1408 files changed, 281519 insertions, 20026 deletions

diff --git a/src/SConscript b/src/SConscript
new file mode 100644
index 0000000000..5440acd135
--- /dev/null
+++ b/src/SConscript
@@ -0,0 +1,11 @@
+Import('*')
+
+SConscript('gallium/SConscript')
+
+if 'mesa' in env['statetrackers']:
+	SConscript('mesa/SConscript')
+
+SConscript('gallium/winsys/SConscript')
+
+SConscript('glut/glx/SConscript')
+SConscript('glew/SConscript')
diff --git a/src/descrip.mms b/src/descrip.mms
deleted file mode 100644
index 71b8ea16ac..0000000000
--- a/src/descrip.mms
+++ /dev/null
@@ -1,43 +0,0 @@
-# Makefile for Mesa for VMS
-# contributed by Jouk Jansen  joukj@hrem.stm.tudelft.nl
-
-.include [-]mms-config.
-
-all :
-	set default [.mesa]
-	$(MMS)$(MMSQUALIFIERS)
-	set default [-]
-.ifdef SHARE
-	$(MMS)$(MMSQUALIFIERS) [-.lib]$(GL_SHAR)
-.endif
-	set default [.glu]
-	$(MMS)$(MMSQUALIFIERS)
-	set default [-.glut.glx]
-	$(MMS)$(MMSQUALIFIERS)
-	set default [--]
-
-[-.lib]$(GL_SHAR) : [-.lib]$(GL_LIB)
-	@ WRITE_ SYS$OUTPUT "  generating libmesa.opt"
-	@ library/extract=* [-.lib]$(GL_LIB)
-	@ OPEN_/WRITE FILE  libmesa.opt
-	@ WRITE_ FILE "!"
-	@ WRITE_ FILE "! libmesa.opt generated by DESCRIP.$(MMS_EXT)" 
-	@ WRITE_ FILE "!"
-	@ WRITE_ FILE "IDENTIFICATION=""mesa5.1"""
-	@ WRITE_ FILE "GSMATCH=LEQUAL,5,1
-	@ WRITE_ FILE "libmesagl.obj"
-	@ write_ file "sys$share:decw$xextlibshr/share"
-	@ write_ file "sys$share:decw$xlibshr/share"
-	@ write_ file "sys$share:pthread$rtl/share"
-	@ CLOSE_ FILE
-	@ $(MMS)$(MMSQUALIFIERS)/ignore=warning mesa_vms
-	@ WRITE_ SYS$OUTPUT "  linking ..."
-	@ LINK_/NODEB/SHARE=[-.lib]$(GL_SHAR)/MAP=libmesa.map/FULL libmesa.opt/opt,\
-	mesa_vms.opt/opt
-	@ delete libmesagl.obj;*
-
-mesa_vms :
-	@ WRITE_ SYS$OUTPUT "  generating libmesa.map ..."
-	@ LINK_/NODEB/NOSHARE/NOEXE/MAP=libmesa.map/FULL libmesa.opt/OPT
-	@ WRITE_ SYS$OUTPUT "  analyzing libmesa.map ..."
-	@ @[-.vms]analyze_map.com libmesa.map mesa_vms.opt
diff --git a/src/driclient/include/driclient.h b/src/driclient/include/driclient.h
new file mode 100644
index 0000000000..d391525039
--- /dev/null
+++ b/src/driclient/include/driclient.h
@@ -0,0 +1,97 @@
+#ifndef driclient_h
+#define driclient_h
+
+#include <stdint.h>
+#include <X11/Xlib.h>
+#include <drm_sarea.h>
+#include "xf86dri.h"
+
+/* TODO: Bring in DRI XML options */
+
+typedef struct dri_version
+{
+	int major;
+	int minor;
+	int patch;
+} dri_version_t;
+
+typedef struct dri_screen
+{
+	Display			*display;
+	unsigned int		num;
+	dri_version_t		ddx, dri, drm;
+	int			draw_lock_id;
+	int			fd;
+	drm_sarea_t		*sarea;
+	void			*drawable_hash;
+	void			*private;
+} dri_screen_t;
+
+struct dri_context;
+
+typedef struct dri_drawable
+{
+	drm_drawable_t		drm_drawable;
+	Drawable		x_drawable;
+	unsigned int		sarea_index;
+	unsigned int		*sarea_stamp;
+	unsigned int		last_sarea_stamp;
+	int			x, y, w, h;
+	int			back_x, back_y;
+	int			num_cliprects, num_back_cliprects;
+	drm_clip_rect_t		*cliprects, *back_cliprects;
+	dri_screen_t		*dri_screen;
+	unsigned int		refcount;
+	void			*private;
+} dri_drawable_t;
+
+typedef struct dri_context
+{
+	XID			id;
+	drm_context_t		drm_context;
+	dri_screen_t		*dri_screen;
+	void			*private;
+} dri_context_t;
+
+typedef struct dri_framebuffer
+{
+	drm_handle_t		drm_handle;
+	int			base, size, stride;
+	int			private_size;
+	void			*private;
+} dri_framebuffer_t;
+
+int driCreateScreen(Display *display, int screen, dri_screen_t **dri_screen, dri_framebuffer_t *dri_framebuf);
+int driDestroyScreen(dri_screen_t *dri_screen);
+int driCreateDrawable(dri_screen_t *dri_screen, Drawable drawable, dri_drawable_t **dri_drawable);
+int driUpdateDrawableInfo(dri_drawable_t *dri_drawable);
+int driDestroyDrawable(dri_drawable_t *dri_drawable);
+int driCreateContext(dri_screen_t *dri_screen, Visual *visual, dri_context_t **dri_context);
+int driDestroyContext(dri_context_t *dri_context);
+
+#define DRI_VALIDATE_DRAWABLE_INFO_ONCE(dri_drawable)					\
+do											\
+{											\
+	if (*(dri_drawable->sarea_stamp) != dri_drawable->last_sarea_stamp)		\
+		driUpdateDrawableInfo(dri_drawable);					\
+} while (0)
+
+#define DRI_VALIDATE_DRAWABLE_INFO(dri_screen, dri_drawable)					\
+do												\
+{												\
+	while (*(dri_drawable->sarea_stamp) != dri_drawable->last_sarea_stamp)			\
+	{											\
+		register unsigned int hwContext = dri_screen->sarea->lock.lock &		\
+		~(DRM_LOCK_HELD | DRM_LOCK_CONT);						\
+		DRM_UNLOCK(dri_screen->fd, &dri_screen->sarea->lock, hwContext);		\
+												\
+		DRM_SPINLOCK(&dri_screen->sarea->drawable_lock, dri_screen->draw_lock_id);	\
+		DRI_VALIDATE_DRAWABLE_INFO_ONCE(dri_drawable);					\
+		DRM_SPINUNLOCK(&dri_screen->sarea->drawable_lock, dri_screen->draw_lock_id);	\
+												\
+		DRM_LIGHT_LOCK(dri_screen->fd, &dri_screen->sarea->lock, hwContext);		\
+	}											\
+} while (0)
+
+#endif
+
diff --git a/src/driclient/include/xf86dri.h b/src/driclient/include/xf86dri.h
new file mode 100644
index 0000000000..baf80a7a9d
--- /dev/null
+++ b/src/driclient/include/xf86dri.h
@@ -0,0 +1,119 @@
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+Copyright 2000 VA Linux Systems, Inc.
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file xf86dri.h
+ * Protocol numbers and function prototypes for DRI X protocol.
+ *
+ * \author Kevin E. Martin <martin@valinux.com>
+ * \author Jens Owen <jens@tungstengraphics.com>
+ * \author Rickard E. (Rik) Faith <faith@valinux.com>
+ */
+
+#ifndef _XF86DRI_H_
+#define _XF86DRI_H_
+
+#include <X11/Xfuncproto.h>
+#include <xf86drm.h>
+
+#define X_XF86DRIQueryVersion			0
+#define X_XF86DRIQueryDirectRenderingCapable	1
+#define X_XF86DRIOpenConnection			2
+#define X_XF86DRICloseConnection		3
+#define X_XF86DRIGetClientDriverName		4
+#define X_XF86DRICreateContext			5
+#define X_XF86DRIDestroyContext			6
+#define X_XF86DRICreateDrawable			7
+#define X_XF86DRIDestroyDrawable		8
+#define X_XF86DRIGetDrawableInfo		9
+#define X_XF86DRIGetDeviceInfo			10
+#define X_XF86DRIAuthConnection                 11
+#define X_XF86DRIOpenFullScreen                 12   /* Deprecated */
+#define X_XF86DRICloseFullScreen                13   /* Deprecated */
+
+#define XF86DRINumberEvents		0
+
+#define XF86DRIClientNotLocal		0
+#define XF86DRIOperationNotSupported	1
+#define XF86DRINumberErrors		(XF86DRIOperationNotSupported + 1)
+
+#ifndef _XF86DRI_SERVER_
+
+_XFUNCPROTOBEGIN
+
+Bool XF86DRIQueryExtension( Display *dpy, int *event_base, int *error_base );
+
+Bool XF86DRIQueryVersion( Display *dpy, int *majorVersion, int *minorVersion,
+    int *patchVersion );
+
+Bool XF86DRIQueryDirectRenderingCapable( Display *dpy, int screen,
+    Bool *isCapable );
+
+Bool XF86DRIOpenConnection( Display *dpy, int screen, drm_handle_t *hSAREA,
+    char **busIDString );
+
+Bool XF86DRIAuthConnection( Display *dpy, int screen, drm_magic_t magic );
+
+Bool XF86DRICloseConnection( Display *dpy, int screen );
+
+Bool XF86DRIGetClientDriverName( Display *dpy, int screen,
+    int *ddxDriverMajorVersion, int *ddxDriverMinorVersion,
+    int *ddxDriverPatchVersion, char **clientDriverName );
+
+Bool XF86DRICreateContext( Display *dpy, int screen, Visual *visual,
+    XID *ptr_to_returned_context_id, drm_context_t *hHWContext );
+
+Bool XF86DRICreateContextWithConfig( Display *dpy, int screen, int configID,
+    XID *ptr_to_returned_context_id, drm_context_t *hHWContext );
+
+Bool XF86DRIDestroyContext( Display *dpy, int screen,
+    XID context_id );
+
+Bool XF86DRICreateDrawable( Display *dpy, int screen,
+    Drawable drawable, drm_drawable_t *hHWDrawable );
+
+Bool XF86DRIDestroyDrawable( Display *dpy, int screen, 
+    Drawable drawable);
+
+Bool XF86DRIGetDrawableInfo( Display *dpy, int screen, Drawable drawable,
+    unsigned int *index, unsigned int *stamp, 
+    int *X, int *Y, int *W, int *H,
+    int *numClipRects, drm_clip_rect_t ** pClipRects,
+    int *backX, int *backY,
+    int *numBackClipRects, drm_clip_rect_t **pBackClipRects );
+
+Bool XF86DRIGetDeviceInfo( Display *dpy, int screen,
+    drm_handle_t *hFrameBuffer, int *fbOrigin, int *fbSize,
+    int *fbStride, int *devPrivateSize, void **pDevPrivate );
+
+_XFUNCPROTOEND
+
+#endif /* _XF86DRI_SERVER_ */
+
+#endif /* _XF86DRI_H_ */
+
diff --git a/src/driclient/src/Makefile b/src/driclient/src/Makefile
new file mode 100644
index 0000000000..34435a2086
--- /dev/null
+++ b/src/driclient/src/Makefile
@@ -0,0 +1,19 @@
+TARGET		= libdriclient.a
+OBJECTS		= driclient.o XF86dri.o
+DRMDIR		?= /usr
+
+CFLAGS		+= -g -Wall -fPIC -I../include -I${DRMDIR}/include -I${DRMDIR}/include/drm
+
+#############################################
+
+.PHONY	= all clean
+
+all: ${TARGET}
+
+${TARGET}: ${OBJECTS}
+	ar rcs $@ $^
+	if ! test -d ../lib; then mkdir ../lib; fi
+	cp ${TARGET} ../lib
+
+clean:
+	rm -rf ${OBJECTS} ${TARGET} ../lib/${TARGET}
diff --git a/src/driclient/src/XF86dri.c b/src/driclient/src/XF86dri.c
new file mode 100644
index 0000000000..9e359a9238
--- /dev/null
+++ b/src/driclient/src/XF86dri.c
@@ -0,0 +1,619 @@
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+Copyright 2000 VA Linux Systems, Inc.
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <martin@valinux.com>
+ *   Jens Owen <jens@tungstengraphics.com>
+ *   Rickard E. (Rik) Faith <faith@valinux.com>
+ *
+ */
+
+/* THIS IS NOT AN X CONSORTIUM STANDARD */
+
+#define NEED_REPLIES
+#include <X11/Xlibint.h>
+#include <X11/extensions/Xext.h>
+#include <X11/extensions/extutil.h>
+#include "xf86dristr.h"
+
+static XExtensionInfo _xf86dri_info_data;
+static XExtensionInfo *xf86dri_info = &_xf86dri_info_data;
+static char xf86dri_extension_name[] = XF86DRINAME;
+
+#define XF86DRICheckExtension(dpy,i,val) \
+  XextCheckExtension (dpy, i, xf86dri_extension_name, val)
+
+/*****************************************************************************
+ *                                                                           *
+ *			   private utility routines                          *
+ *                                                                           *
+ *****************************************************************************/
+
+static int close_display(Display *dpy, XExtCodes *extCodes);
+static /* const */ XExtensionHooks xf86dri_extension_hooks = {
+    NULL,				/* create_gc */
+    NULL,				/* copy_gc */
+    NULL,				/* flush_gc */
+    NULL,				/* free_gc */
+    NULL,				/* create_font */
+    NULL,				/* free_font */
+    close_display,			/* close_display */
+    NULL,				/* wire_to_event */
+    NULL,				/* event_to_wire */
+    NULL,				/* error */
+    NULL,				/* error_string */
+};
+
+static XEXT_GENERATE_FIND_DISPLAY (find_display, xf86dri_info, 
+				   xf86dri_extension_name, 
+				   &xf86dri_extension_hooks, 
+				   0, NULL)
+
+static XEXT_GENERATE_CLOSE_DISPLAY (close_display, xf86dri_info)
+
+
+/*****************************************************************************
+ *                                                                           *
+ *		    public XFree86-DRI Extension routines                    *
+ *                                                                           *
+ *****************************************************************************/
+
+#if 0
+#include <stdio.h>
+#define TRACE(msg)  fprintf(stderr,"XF86DRI%s\n", msg);
+#else
+#define TRACE(msg)
+#endif
+
+#define PUBLIC
+
+PUBLIC Bool XF86DRIQueryExtension (dpy, event_basep, error_basep)
+    Display *dpy;
+    int *event_basep, *error_basep;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+
+    TRACE("QueryExtension...");
+    if (XextHasExtension(info)) {
+	*event_basep = info->codes->first_event;
+	*error_basep = info->codes->first_error;
+        TRACE("QueryExtension... return True");
+	return True;
+    } else {
+        TRACE("QueryExtension... return False");
+	return False;
+    }
+}
+
+PUBLIC Bool XF86DRIQueryVersion(dpy, majorVersion, minorVersion, patchVersion)
+    Display* dpy;
+    int* majorVersion; 
+    int* minorVersion;
+    int* patchVersion;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIQueryVersionReply rep;
+    xXF86DRIQueryVersionReq *req;
+
+    TRACE("QueryVersion...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIQueryVersion, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIQueryVersion;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("QueryVersion... return False");
+	return False;
+    }
+    *majorVersion = rep.majorVersion;
+    *minorVersion = rep.minorVersion;
+    *patchVersion = rep.patchVersion;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("QueryVersion... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIQueryDirectRenderingCapable(dpy, screen, isCapable)
+    Display* dpy;
+    int screen;
+    Bool* isCapable;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIQueryDirectRenderingCapableReply rep;
+    xXF86DRIQueryDirectRenderingCapableReq *req;
+
+    TRACE("QueryDirectRenderingCapable...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIQueryDirectRenderingCapable, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIQueryDirectRenderingCapable;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("QueryDirectRenderingCapable... return False");
+	return False;
+    }
+    *isCapable = rep.isCapable;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("QueryDirectRenderingCapable... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIOpenConnection(dpy, screen, hSAREA, busIdString)
+    Display* dpy;
+    int screen;
+    drm_handle_t * hSAREA;
+    char **busIdString;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIOpenConnectionReply rep;
+    xXF86DRIOpenConnectionReq *req;
+
+    TRACE("OpenConnection...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIOpenConnection, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIOpenConnection;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("OpenConnection... return False");
+	return False;
+    }
+
+    *hSAREA = rep.hSAREALow;
+    if (sizeof(drm_handle_t) == 8) {
+       int shift = 32; /* var to prevent warning on next line */
+       *hSAREA |= ((drm_handle_t) rep.hSAREAHigh) << shift;
+    }
+
+    if (rep.length) {
+        if (!(*busIdString = (char *)Xcalloc(rep.busIdStringLength + 1, 1))) {
+            _XEatData(dpy, ((rep.busIdStringLength+3) & ~3));
+            UnlockDisplay(dpy);
+            SyncHandle();
+            TRACE("OpenConnection... return False");
+            return False;
+        }
+	_XReadPad(dpy, *busIdString, rep.busIdStringLength);
+    } else {
+        *busIdString = NULL;
+    }
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("OpenConnection... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIAuthConnection(dpy, screen, magic)
+    Display* dpy;
+    int screen;
+    drm_magic_t magic;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIAuthConnectionReq *req;
+    xXF86DRIAuthConnectionReply rep;
+
+    TRACE("AuthConnection...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIAuthConnection, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIAuthConnection;
+    req->screen = screen;
+    req->magic = magic;
+    rep.authenticated = 0;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse) || !rep.authenticated) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("AuthConnection... return False");
+	return False;
+    }
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("AuthConnection... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRICloseConnection(dpy, screen)
+    Display* dpy;
+    int screen;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRICloseConnectionReq *req;
+
+    TRACE("CloseConnection...");
+
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRICloseConnection, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRICloseConnection;
+    req->screen = screen;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("CloseConnection... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIGetClientDriverName(dpy, screen, ddxDriverMajorVersion, 
+	ddxDriverMinorVersion, ddxDriverPatchVersion, clientDriverName)
+    Display* dpy;
+    int screen;
+    int* ddxDriverMajorVersion;
+    int* ddxDriverMinorVersion;
+    int* ddxDriverPatchVersion;
+    char** clientDriverName;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIGetClientDriverNameReply rep;
+    xXF86DRIGetClientDriverNameReq *req;
+
+    TRACE("GetClientDriverName...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIGetClientDriverName, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIGetClientDriverName;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetClientDriverName... return False");
+	return False;
+    }
+
+    *ddxDriverMajorVersion = rep.ddxDriverMajorVersion;
+    *ddxDriverMinorVersion = rep.ddxDriverMinorVersion;
+    *ddxDriverPatchVersion = rep.ddxDriverPatchVersion;
+
+    if (rep.length) {
+        if (!(*clientDriverName = (char *)Xcalloc(rep.clientDriverNameLength + 1, 1))) {
+            _XEatData(dpy, ((rep.clientDriverNameLength+3) & ~3));
+            UnlockDisplay(dpy);
+            SyncHandle();
+            TRACE("GetClientDriverName... return False");
+            return False;
+        }
+	_XReadPad(dpy, *clientDriverName, rep.clientDriverNameLength);
+    } else {
+        *clientDriverName = NULL;
+    }
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("GetClientDriverName... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRICreateContextWithConfig(dpy, screen, configID, context,
+	hHWContext)
+    Display* dpy;
+    int screen;
+    int configID;
+    XID* context;
+    drm_context_t * hHWContext;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRICreateContextReply rep;
+    xXF86DRICreateContextReq *req;
+
+    TRACE("CreateContext...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRICreateContext, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRICreateContext;
+    req->visual = configID;
+    req->screen = screen;
+    *context = XAllocID(dpy);
+    req->context = *context;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("CreateContext... return False");
+	return False;
+    }
+    *hHWContext = rep.hHWContext;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("CreateContext... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRICreateContext(dpy, screen, visual, context, hHWContext)
+    Display* dpy;
+    int screen;
+    Visual* visual;
+    XID* context;
+    drm_context_t * hHWContext;
+{
+    return XF86DRICreateContextWithConfig( dpy, screen, visual->visualid,
+					   context, hHWContext );
+}
+
+PUBLIC Bool XF86DRIDestroyContext( Display * ndpy, int screen, 
+    XID context )
+{
+    Display * const dpy = (Display *) ndpy;
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIDestroyContextReq *req;
+
+    TRACE("DestroyContext...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIDestroyContext, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIDestroyContext;
+    req->screen = screen;
+    req->context = context;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("DestroyContext... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRICreateDrawable( Display * ndpy, int screen, 
+    Drawable drawable, drm_drawable_t * hHWDrawable )
+{
+    Display * const dpy = (Display *) ndpy;
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRICreateDrawableReply rep;
+    xXF86DRICreateDrawableReq *req;
+
+    TRACE("CreateDrawable...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRICreateDrawable, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRICreateDrawable;
+    req->screen = screen;
+    req->drawable = drawable;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("CreateDrawable... return False");
+	return False;
+    }
+    *hHWDrawable = rep.hHWDrawable;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("CreateDrawable... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIDestroyDrawable( Display * ndpy, int screen,
+    Drawable drawable )
+{
+    Display * const dpy = (Display *) ndpy;
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIDestroyDrawableReq *req;
+
+    TRACE("DestroyDrawable...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIDestroyDrawable, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIDestroyDrawable;
+    req->screen = screen;
+    req->drawable = drawable;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("DestroyDrawable... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIGetDrawableInfo(Display* dpy, int screen, Drawable drawable,
+    unsigned int* index, unsigned int* stamp,
+    int* X, int* Y, int* W, int* H,
+    int* numClipRects, drm_clip_rect_t ** pClipRects,
+    int* backX, int* backY,
+    int* numBackClipRects, drm_clip_rect_t ** pBackClipRects )
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIGetDrawableInfoReply rep;
+    xXF86DRIGetDrawableInfoReq *req;
+    int total_rects;
+
+    TRACE("GetDrawableInfo...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIGetDrawableInfo, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIGetDrawableInfo;
+    req->screen = screen;
+    req->drawable = drawable;
+
+    if (!_XReply(dpy, (xReply *)&rep, 1, xFalse)) 
+    {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetDrawableInfo... return False");
+	return False;
+    }
+    *index = rep.drawableTableIndex;
+    *stamp = rep.drawableTableStamp;
+    *X = (int)rep.drawableX;
+    *Y = (int)rep.drawableY;
+    *W = (int)rep.drawableWidth;
+    *H = (int)rep.drawableHeight;
+    *numClipRects = rep.numClipRects;
+    total_rects = *numClipRects;
+
+    *backX = rep.backX;
+    *backY = rep.backY;
+    *numBackClipRects = rep.numBackClipRects;
+    total_rects += *numBackClipRects;
+
+#if 0
+    /* Because of the fix in Xserver/GL/dri/xf86dri.c, this check breaks
+     * backwards compatibility (Because of the >> 2 shift) but the fix
+     * enables multi-threaded apps to work.
+     */
+    if (rep.length !=  ((((SIZEOF(xXF86DRIGetDrawableInfoReply) - 
+		       SIZEOF(xGenericReply) + 
+		       total_rects * sizeof(drm_clip_rect_t)) + 3) & ~3) >> 2)) {
+        _XEatData(dpy, rep.length);
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetDrawableInfo... return False");
+        return False;
+    }
+#endif
+
+    if (*numClipRects) {
+       int len = sizeof(drm_clip_rect_t) * (*numClipRects);
+
+       *pClipRects = (drm_clip_rect_t *)Xcalloc(len, 1);
+       if (*pClipRects) 
+	  _XRead(dpy, (char*)*pClipRects, len);
+    } else {
+        *pClipRects = NULL;
+    }
+
+    if (*numBackClipRects) {
+       int len = sizeof(drm_clip_rect_t) * (*numBackClipRects);
+
+       *pBackClipRects = (drm_clip_rect_t *)Xcalloc(len, 1);
+       if (*pBackClipRects) 
+	  _XRead(dpy, (char*)*pBackClipRects, len);
+    } else {
+        *pBackClipRects = NULL;
+    }
+
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("GetDrawableInfo... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIGetDeviceInfo(dpy, screen, hFrameBuffer, 
+	fbOrigin, fbSize, fbStride, devPrivateSize, pDevPrivate)
+    Display* dpy;
+    int screen;
+    drm_handle_t * hFrameBuffer;
+    int* fbOrigin;
+    int* fbSize;
+    int* fbStride;
+    int* devPrivateSize;
+    void** pDevPrivate;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIGetDeviceInfoReply rep;
+    xXF86DRIGetDeviceInfoReq *req;
+
+    TRACE("GetDeviceInfo...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIGetDeviceInfo, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIGetDeviceInfo;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetDeviceInfo... return False");
+	return False;
+    }
+
+    *hFrameBuffer = rep.hFrameBufferLow;
+    if (sizeof(drm_handle_t) == 8) {
+       int shift = 32; /* var to prevent warning on next line */
+       *hFrameBuffer |= ((drm_handle_t) rep.hFrameBufferHigh) << shift;
+    }
+
+    *fbOrigin = rep.framebufferOrigin;
+    *fbSize = rep.framebufferSize;
+    *fbStride = rep.framebufferStride;
+    *devPrivateSize = rep.devPrivateSize;
+
+    if (rep.length) {
+        if (!(*pDevPrivate = (void *)Xcalloc(rep.devPrivateSize, 1))) {
+            _XEatData(dpy, ((rep.devPrivateSize+3) & ~3));
+            UnlockDisplay(dpy);
+            SyncHandle();
+            TRACE("GetDeviceInfo... return False");
+            return False;
+        }
+	_XRead(dpy, (char*)*pDevPrivate, rep.devPrivateSize);
+    } else {
+        *pDevPrivate = NULL;
+    }
+
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("GetDeviceInfo... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIOpenFullScreen(dpy, screen, drawable)
+    Display* dpy;
+    int screen;
+    Drawable drawable;
+{
+    /* This function and the underlying X protocol are deprecated.
+     */
+    (void) dpy;
+    (void) screen;
+    (void) drawable;
+    return False;
+}
+
+PUBLIC Bool XF86DRICloseFullScreen(dpy, screen, drawable)
+    Display* dpy;
+    int screen;
+    Drawable drawable;
+{
+    /* This function and the underlying X protocol are deprecated.
+     */
+    (void) dpy;
+    (void) screen;
+    (void) drawable;
+    return True;
+}
+
diff --git a/src/driclient/src/driclient.c b/src/driclient/src/driclient.c
new file mode 100644
index 0000000000..dc2189afec
--- /dev/null
+++ b/src/driclient/src/driclient.c
@@ -0,0 +1,310 @@
+#include "driclient.h"
+#include <assert.h>
+#include <stdlib.h>
+
+int driCreateScreen(Display *display, int screen, dri_screen_t **dri_screen, dri_framebuffer_t *dri_framebuf)
+{
+	int		evbase, errbase;
+	char		*driver_name;
+	int		newly_opened;
+	drm_magic_t	magic;
+	drmVersionPtr	drm_version;
+	drm_handle_t	sarea_handle;
+	char		*bus_id;
+	dri_screen_t	*dri_scrn;
+
+	assert(display);
+	assert(dri_screen);
+
+	if (!XF86DRIQueryExtension(display, &evbase, &errbase))
+		return 1;
+
+	dri_scrn = calloc(1, sizeof(dri_screen_t));
+
+	if (!dri_scrn)
+		return 1;
+
+	if (!XF86DRIQueryVersion(display, &dri_scrn->dri.major, &dri_scrn->dri.minor, &dri_scrn->dri.patch))
+		goto free_screen;
+
+	dri_scrn->display = display;
+	dri_scrn->num = screen;
+	dri_scrn->draw_lock_id = 1;
+
+	if (!XF86DRIOpenConnection(display, screen, &sarea_handle, &bus_id))
+		goto free_screen;
+
+	dri_scrn->fd = -1;
+	dri_scrn->fd = drmOpenOnce(NULL, bus_id, &newly_opened);
+	XFree(bus_id);
+
+	if (dri_scrn->fd < 0)
+		goto close_connection;
+
+	if (drmGetMagic(dri_scrn->fd, &magic))
+		goto close_drm;
+
+	drm_version = drmGetVersion(dri_scrn->fd);
+
+	if (!drm_version)
+		goto close_drm;
+
+	dri_scrn->drm.major = drm_version->version_major;
+	dri_scrn->drm.minor = drm_version->version_minor;
+	dri_scrn->drm.patch = drm_version->version_patchlevel;
+	drmFreeVersion(drm_version);
+
+	if (!XF86DRIAuthConnection(display, screen, magic))
+		goto close_drm;
+
+	if (!XF86DRIGetClientDriverName
+	(
+		display,
+		screen,
+		&dri_scrn->ddx.major,
+		&dri_scrn->ddx.minor,
+		&dri_scrn->ddx.patch,
+		&driver_name
+	))
+		goto close_drm;
+
+	if (drmMap(dri_scrn->fd, sarea_handle, SAREA_MAX, (drmAddress)&dri_scrn->sarea))
+		goto close_drm;
+
+	dri_scrn->drawable_hash = drmHashCreate();
+
+	if (!dri_scrn->drawable_hash)
+		goto unmap_sarea;
+
+	if (dri_framebuf)
+	{
+		if (!XF86DRIGetDeviceInfo
+		(
+			display,
+			screen, &dri_framebuf->drm_handle,
+			&dri_framebuf->base,
+			&dri_framebuf->size,
+			&dri_framebuf->stride,
+			&dri_framebuf->private_size,
+			&dri_framebuf->private
+		))
+			goto destroy_hash;
+	}
+
+	*dri_screen = dri_scrn;
+
+	return 0;
+
+destroy_hash:
+	drmHashDestroy(dri_scrn->drawable_hash);
+unmap_sarea:
+	drmUnmap(dri_scrn->sarea, SAREA_MAX);
+close_drm:
+	drmCloseOnce(dri_scrn->fd);
+close_connection:
+	XF86DRICloseConnection(display, screen);
+free_screen:
+	free(dri_scrn);
+
+	return 1;
+}
+
+int driDestroyScreen(dri_screen_t *dri_screen)
+{
+	Drawable	draw;
+	dri_drawable_t	*dri_draw;
+
+	assert(dri_screen);
+
+	if (drmHashFirst(dri_screen->drawable_hash, &draw, (void**)&dri_draw))
+	{
+		dri_draw->refcount = 1;
+		driDestroyDrawable(dri_draw);
+
+		while (drmHashNext(dri_screen->drawable_hash, &draw, (void**)&dri_draw))
+		{
+			dri_draw->refcount = 1;
+			driDestroyDrawable(dri_draw);
+		}
+	}
+
+	drmHashDestroy(dri_screen->drawable_hash);
+	drmUnmap(dri_screen->sarea, SAREA_MAX);
+	drmCloseOnce(dri_screen->fd);
+	XF86DRICloseConnection(dri_screen->display, dri_screen->num);
+	free(dri_screen);
+
+	return 0;
+}
+
+int driCreateDrawable(dri_screen_t *dri_screen, Drawable drawable, dri_drawable_t **dri_drawable)
+{
+	int		evbase, errbase;
+	dri_drawable_t	*dri_draw;
+
+	assert(dri_screen);
+	assert(dri_drawable);
+
+	if (!XF86DRIQueryExtension(dri_screen->display, &evbase, &errbase))
+		return 1;
+
+	if (!drmHashLookup(dri_screen->drawable_hash, drawable, (void**)dri_drawable))
+	{
+		/* Found */
+		(*dri_drawable)->refcount++;
+		return 0;
+	}
+
+	dri_draw = calloc(1, sizeof(dri_drawable_t));
+
+	if (!dri_draw)
+		return 1;
+
+	if (!XF86DRICreateDrawable(dri_screen->display, 0, drawable, &dri_draw->drm_drawable))
+	{
+		free(dri_draw);
+		return 1;
+	}
+
+	dri_draw->x_drawable = drawable;
+	dri_draw->sarea_index = 0;
+	dri_draw->sarea_stamp = NULL;
+	dri_draw->last_sarea_stamp = 0;
+	dri_draw->dri_screen = dri_screen;
+	dri_draw->refcount = 1;
+
+	if (drmHashInsert(dri_screen->drawable_hash, drawable, dri_draw))
+	{
+		XF86DRIDestroyDrawable(dri_screen->display, dri_screen->num, drawable);
+		free(dri_draw);
+		return 1;
+	}
+
+	if (!dri_draw->sarea_stamp || *dri_draw->sarea_stamp != dri_draw->last_sarea_stamp)
+	{
+		DRM_SPINLOCK(&dri_screen->sarea->drawable_lock, dri_screen->draw_lock_id);
+
+		if (driUpdateDrawableInfo(dri_draw))
+		{
+			XF86DRIDestroyDrawable(dri_screen->display, dri_screen->num, drawable);
+			free(dri_draw);
+			DRM_SPINUNLOCK(&dri_screen->sarea->drawable_lock, dri_screen->draw_lock_id);
+			return 1;
+		}
+
+		DRM_SPINUNLOCK(&dri_screen->sarea->drawable_lock, dri_screen->draw_lock_id);
+	}
+
+	*dri_drawable = dri_draw;
+
+	return 0;
+}
+
+int driUpdateDrawableInfo(dri_drawable_t *dri_drawable)
+{
+	assert(dri_drawable);
+
+	if (dri_drawable->cliprects)
+	{
+		XFree(dri_drawable->cliprects);
+		dri_drawable->cliprects = NULL;
+	}
+	if (dri_drawable->back_cliprects)
+	{
+		XFree(dri_drawable->back_cliprects);
+		dri_drawable->back_cliprects = NULL;
+	}
+
+	DRM_SPINUNLOCK(&dri_drawable->dri_screen->sarea->drawable_lock, dri_drawable->dri_screen->draw_lock_id);
+
+	if (!XF86DRIGetDrawableInfo
+	(
+		dri_drawable->dri_screen->display,
+		dri_drawable->dri_screen->num,
+		dri_drawable->x_drawable,
+    		&dri_drawable->sarea_index,
+    		&dri_drawable->last_sarea_stamp,
+    		&dri_drawable->x,
+    		&dri_drawable->y,
+    		&dri_drawable->w,
+    		&dri_drawable->h,
+		&dri_drawable->num_cliprects,
+		&dri_drawable->cliprects,
+		&dri_drawable->back_x,
+		&dri_drawable->back_y,
+		&dri_drawable->num_back_cliprects,
+		&dri_drawable->back_cliprects
+	))
+	{
+		dri_drawable->sarea_stamp = &dri_drawable->last_sarea_stamp;
+		dri_drawable->num_cliprects = 0;
+		dri_drawable->cliprects = NULL;
+		dri_drawable->num_back_cliprects = 0;
+		dri_drawable->back_cliprects = 0;
+
+		return 1;
+	}
+	else
+		dri_drawable->sarea_stamp = &dri_drawable->dri_screen->sarea->drawableTable[dri_drawable->sarea_index].stamp;
+
+	DRM_SPINLOCK(&dri_drawable->dri_screen->sarea->drawable_lock, dri_drawable->dri_screen->draw_lock_id);
+
+	return 0;
+}
+
+int driDestroyDrawable(dri_drawable_t *dri_drawable)
+{
+	assert(dri_drawable);
+
+	if (--dri_drawable->refcount == 0)
+	{
+		if (dri_drawable->cliprects)
+			XFree(dri_drawable->cliprects);
+		if (dri_drawable->back_cliprects)
+			XFree(dri_drawable->back_cliprects);
+		drmHashDelete(dri_drawable->dri_screen->drawable_hash, dri_drawable->x_drawable);
+		XF86DRIDestroyDrawable(dri_drawable->dri_screen->display, dri_drawable->dri_screen->num, dri_drawable->x_drawable);
+		free(dri_drawable);
+	}
+
+	return 0;
+}
+
+int driCreateContext(dri_screen_t *dri_screen, Visual *visual, dri_context_t **dri_context)
+{
+	int		evbase, errbase;
+	dri_context_t	*dri_ctx;
+
+	assert(dri_screen);
+	assert(visual);
+	assert(dri_context);
+
+	if (!XF86DRIQueryExtension(dri_screen->display, &evbase, &errbase))
+		return 1;
+
+	dri_ctx = calloc(1, sizeof(dri_context_t));
+
+	if (!dri_ctx)
+		return 1;
+
+	if (!XF86DRICreateContext(dri_screen->display, dri_screen->num, visual, &dri_ctx->id, &dri_ctx->drm_context))
+	{
+		free(dri_ctx);
+		return 1;
+	}
+
+	dri_ctx->dri_screen = dri_screen;
+	*dri_context = dri_ctx;
+
+	return 0;
+}
+
+int driDestroyContext(dri_context_t *dri_context)
+{
+	assert(dri_context);
+
+	XF86DRIDestroyContext(dri_context->dri_screen->display, dri_context->dri_screen->num, dri_context->id);
+	free(dri_context);
+
+	return 0;
+}
diff --git a/src/driclient/src/xf86dristr.h b/src/driclient/src/xf86dristr.h
new file mode 100644
index 0000000000..b834bd1a1a
--- /dev/null
+++ b/src/driclient/src/xf86dristr.h
@@ -0,0 +1,342 @@
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+Copyright 2000 VA Linux Systems, Inc.
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <martin@valinux.com>
+ *   Jens Owen <jens@tungstengraphics.com>
+ *   Rickard E. (Rik) Fiath <faith@valinux.com>
+ *
+ */
+
+#ifndef _XF86DRISTR_H_
+#define _XF86DRISTR_H_
+
+#include "xf86dri.h"
+
+#define XF86DRINAME "XFree86-DRI"
+
+/* The DRI version number.  This was originally set to be the same of the
+ * XFree86 version number.  However, this version is really indepedent of
+ * the XFree86 version.
+ *
+ * Version History:
+ *    4.0.0: Original
+ *    4.0.1: Patch to bump clipstamp when windows are destroyed, 28 May 02
+ *    4.1.0: Add transition from single to multi in DRMInfo rec, 24 Jun 02
+ */
+#define XF86DRI_MAJOR_VERSION	4
+#define XF86DRI_MINOR_VERSION	1
+#define XF86DRI_PATCH_VERSION	0
+
+typedef struct _XF86DRIQueryVersion {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIQueryVersion */
+    CARD16	length B16;
+} xXF86DRIQueryVersionReq;
+#define sz_xXF86DRIQueryVersionReq	4
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD16	majorVersion B16;	/* major version of DRI protocol */
+    CARD16	minorVersion B16;	/* minor version of DRI protocol */
+    CARD32	patchVersion B32;       /* patch version of DRI protocol */
+    CARD32	pad3 B32;
+    CARD32	pad4 B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRIQueryVersionReply;
+#define sz_xXF86DRIQueryVersionReply	32
+
+typedef struct _XF86DRIQueryDirectRenderingCapable {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* X_DRIQueryDirectRenderingCapable */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIQueryDirectRenderingCapableReq;
+#define sz_xXF86DRIQueryDirectRenderingCapableReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    BOOL	isCapable;
+    BOOL	pad2;
+    BOOL	pad3;
+    BOOL	pad4;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+    CARD32	pad7 B32;
+    CARD32	pad8 B32;
+    CARD32	pad9 B32;
+} xXF86DRIQueryDirectRenderingCapableReply;
+#define sz_xXF86DRIQueryDirectRenderingCapableReply	32
+
+typedef struct _XF86DRIOpenConnection {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIOpenConnection */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIOpenConnectionReq;
+#define sz_xXF86DRIOpenConnectionReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hSAREALow B32;
+    CARD32	hSAREAHigh B32;
+    CARD32	busIdStringLength B32;
+    CARD32	pad6 B32;
+    CARD32	pad7 B32;
+    CARD32	pad8 B32;
+} xXF86DRIOpenConnectionReply;
+#define sz_xXF86DRIOpenConnectionReply	32
+
+typedef struct _XF86DRIAuthConnection {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICloseConnection */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32      magic B32;
+} xXF86DRIAuthConnectionReq;
+#define sz_xXF86DRIAuthConnectionReq	12
+
+typedef struct {
+    BYTE        type;
+    BOOL        pad1;
+    CARD16      sequenceNumber B16;
+    CARD32      length B32;
+    CARD32      authenticated B32;
+    CARD32      pad2 B32;
+    CARD32      pad3 B32;
+    CARD32      pad4 B32;
+    CARD32      pad5 B32;
+    CARD32      pad6 B32;
+} xXF86DRIAuthConnectionReply;
+#define zx_xXF86DRIAuthConnectionReply  32
+
+typedef struct _XF86DRICloseConnection {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICloseConnection */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRICloseConnectionReq;
+#define sz_xXF86DRICloseConnectionReq	8
+
+typedef struct _XF86DRIGetClientDriverName {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIGetClientDriverName */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIGetClientDriverNameReq;
+#define sz_xXF86DRIGetClientDriverNameReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	ddxDriverMajorVersion B32;
+    CARD32	ddxDriverMinorVersion B32;
+    CARD32	ddxDriverPatchVersion B32;
+    CARD32	clientDriverNameLength B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRIGetClientDriverNameReply;
+#define sz_xXF86DRIGetClientDriverNameReply	32
+
+typedef struct _XF86DRICreateContext {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICreateContext */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	visual B32;
+    CARD32	context B32;
+} xXF86DRICreateContextReq;
+#define sz_xXF86DRICreateContextReq	16
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hHWContext B32;
+    CARD32	pad2 B32;
+    CARD32	pad3 B32;
+    CARD32	pad4 B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRICreateContextReply;
+#define sz_xXF86DRICreateContextReply	32
+
+typedef struct _XF86DRIDestroyContext {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIDestroyContext */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	context B32;
+} xXF86DRIDestroyContextReq;
+#define sz_xXF86DRIDestroyContextReq	12
+
+typedef struct _XF86DRICreateDrawable {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICreateDrawable */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	drawable B32;
+} xXF86DRICreateDrawableReq;
+#define sz_xXF86DRICreateDrawableReq	12
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hHWDrawable B32;
+    CARD32	pad2 B32;
+    CARD32	pad3 B32;
+    CARD32	pad4 B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRICreateDrawableReply;
+#define sz_xXF86DRICreateDrawableReply	32
+
+typedef struct _XF86DRIDestroyDrawable {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIDestroyDrawable */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	drawable B32;
+} xXF86DRIDestroyDrawableReq;
+#define sz_xXF86DRIDestroyDrawableReq	12
+
+typedef struct _XF86DRIGetDrawableInfo {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIGetDrawableInfo */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	drawable B32;
+} xXF86DRIGetDrawableInfoReq;
+#define sz_xXF86DRIGetDrawableInfoReq	12
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	drawableTableIndex B32;
+    CARD32	drawableTableStamp B32;
+    INT16	drawableX B16;
+    INT16	drawableY B16;
+    INT16	drawableWidth B16;
+    INT16	drawableHeight B16;
+    CARD32	numClipRects B32;
+    INT16       backX B16;
+    INT16       backY B16;
+    CARD32      numBackClipRects B32;
+} xXF86DRIGetDrawableInfoReply;
+
+#define sz_xXF86DRIGetDrawableInfoReply	36
+
+
+typedef struct _XF86DRIGetDeviceInfo {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIGetDeviceInfo */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIGetDeviceInfoReq;
+#define sz_xXF86DRIGetDeviceInfoReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hFrameBufferLow B32;
+    CARD32	hFrameBufferHigh B32;
+    CARD32	framebufferOrigin B32;
+    CARD32	framebufferSize B32;
+    CARD32	framebufferStride B32;
+    CARD32	devPrivateSize B32;
+} xXF86DRIGetDeviceInfoReply;
+#define sz_xXF86DRIGetDeviceInfoReply	32
+
+typedef struct _XF86DRIOpenFullScreen {
+    CARD8       reqType;	/* always DRIReqCode */
+    CARD8       driReqType;	/* always X_DRIOpenFullScreen */
+    CARD16      length B16;
+    CARD32      screen B32;
+    CARD32      drawable B32;
+} xXF86DRIOpenFullScreenReq;
+#define sz_xXF86DRIOpenFullScreenReq    12
+
+typedef struct {
+    BYTE        type;
+    BOOL        pad1;
+    CARD16      sequenceNumber B16;
+    CARD32      length B32;
+    CARD32      isFullScreen B32;
+    CARD32      pad2 B32;
+    CARD32      pad3 B32;
+    CARD32      pad4 B32;
+    CARD32      pad5 B32;
+    CARD32      pad6 B32;
+} xXF86DRIOpenFullScreenReply;
+#define sz_xXF86DRIOpenFullScreenReply  32
+
+typedef struct _XF86DRICloseFullScreen {
+    CARD8       reqType;	/* always DRIReqCode */
+    CARD8       driReqType;	/* always X_DRICloseFullScreen */
+    CARD16      length B16;
+    CARD32      screen B32;
+    CARD32      drawable B32;
+} xXF86DRICloseFullScreenReq;
+#define sz_xXF86DRICloseFullScreenReq   12
+
+typedef struct {
+    BYTE        type;
+    BOOL        pad1;
+    CARD16      sequenceNumber B16;
+    CARD32      length B32;
+    CARD32      pad2 B32;
+    CARD32      pad3 B32;
+    CARD32      pad4 B32;
+    CARD32      pad5 B32;
+    CARD32      pad6 B32;
+    CARD32      pad7 B32;
+} xXF86DRICloseFullScreenReply;
+#define sz_xXF86DRICloseFullScreenReply  32
+
+
+#endif /* _XF86DRISTR_H_ */
diff --git a/src/egl/Makefile b/src/egl/Makefile
index fb4db8f9b9..5b09e178c6 100644
--- a/src/egl/Makefile
+++ b/src/egl/Makefile
@@ -1,8 +1,9 @@
 # src/egl/Makefile
 
 TOP = ../..
+include $(TOP)/configs/current
 
-SUBDIRS = main drivers/demo drivers/dri
+SUBDIRS = main drivers
 
 
 default: subdirs
@@ -15,6 +16,12 @@ subdirs:
 		fi \
 	done
 
+install:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE) install) || exit 1 ; \
+		fi \
+	done
 
 clean:
 	-@for dir in $(SUBDIRS) ; do \
diff --git a/src/egl/drivers/Makefile b/src/egl/drivers/Makefile
new file mode 100644
index 0000000000..dde4ee2255
--- /dev/null
+++ b/src/egl/drivers/Makefile
@@ -0,0 +1,31 @@
+# src/egl/drivers/Makefile
+
+TOP = ../../..
+include $(TOP)/configs/current
+
+SUBDIRS = $(EGL_DRIVERS_DIRS)
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir ; $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+install:
+	@ for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir ; $(MAKE) install) || exit 1 ; \
+		fi \
+	done
+
+clean:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir ; $(MAKE) clean) ; \
+		fi \
+	done
diff --git a/src/egl/drivers/demo/Makefile b/src/egl/drivers/demo/Makefile
index 6ca25aa5d6..d908cdb4b2 100644
--- a/src/egl/drivers/demo/Makefile
+++ b/src/egl/drivers/demo/Makefile
@@ -25,7 +25,9 @@ $(TOP)/$(LIB_DIR)/demodriver.so: $(OBJECTS)
 		-ldflags '$(LDFLAGS)' -install $(TOP)/$(LIB_DIR) \
 		$(OBJECTS)
 
-
+install:
+	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(INSTALL) $(TOP)/$(LIB_DIR)/demodriver.so $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	-rm -f *.o
diff --git a/src/egl/drivers/demo/demo.c b/src/egl/drivers/demo/demo.c
index 45545755c0..1750e976b8 100644
--- a/src/egl/drivers/demo/demo.c
+++ b/src/egl/drivers/demo/demo.c
@@ -21,7 +21,7 @@
 typedef struct demo_driver
 {
    _EGLDriver Base;  /* base class/object */
-   GLuint DemoStuff;
+   unsigned DemoStuff;
 } DemoDriver;
 
 #define DEMO_DRIVER(D) ((DemoDriver *) (D))
@@ -33,7 +33,7 @@ typedef struct demo_driver
 typedef struct demo_surface
 {
    _EGLSurface Base;  /* base class/object */
-   GLuint DemoStuff;
+   unsigned DemoStuff;
 } DemoSurface;
 
 
@@ -43,7 +43,7 @@ typedef struct demo_surface
 typedef struct demo_context
 {
    _EGLContext Base;  /* base class/object */
-   GLuint DemoStuff;
+   unsigned DemoStuff;
 } DemoContext;
 
 
@@ -67,22 +67,22 @@ demoInitialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
 
    /* Create the display's visual configs - silly example */
    for (i = 0; i < 4; i++) {
-      _EGLConfig config;
-      _eglInitConfig(&config, i + 1);
-      _eglSetConfigAttrib(&config, EGL_RED_SIZE, 8);
-      _eglSetConfigAttrib(&config, EGL_GREEN_SIZE, 8);
-      _eglSetConfigAttrib(&config, EGL_BLUE_SIZE, 8);
-      _eglSetConfigAttrib(&config, EGL_ALPHA_SIZE, 8);
-      _eglSetConfigAttrib(&config, EGL_BUFFER_SIZE, 32);
+      _EGLConfig *config = calloc(1, sizeof(_EGLConfig));
+      _eglInitConfig(config, i + 1);
+      _eglSetConfigAttrib(config, EGL_RED_SIZE, 8);
+      _eglSetConfigAttrib(config, EGL_GREEN_SIZE, 8);
+      _eglSetConfigAttrib(config, EGL_BLUE_SIZE, 8);
+      _eglSetConfigAttrib(config, EGL_ALPHA_SIZE, 8);
+      _eglSetConfigAttrib(config, EGL_BUFFER_SIZE, 32);
       if (i & 1) {
-         _eglSetConfigAttrib(&config, EGL_DEPTH_SIZE, 32);
+         _eglSetConfigAttrib(config, EGL_DEPTH_SIZE, 32);
       }
       if (i & 2) {
-         _eglSetConfigAttrib(&config, EGL_STENCIL_SIZE, 8);
+         _eglSetConfigAttrib(config, EGL_STENCIL_SIZE, 8);
       }
-      _eglSetConfigAttrib(&config, EGL_SURFACE_TYPE,
+      _eglSetConfigAttrib(config, EGL_SURFACE_TYPE,
                           (EGL_WINDOW_BIT | EGL_PIXMAP_BIT | EGL_PBUFFER_BIT));
-      _eglAddConfig(disp, &config);
+      _eglAddConfig(disp, config);
    }
 
    drv->Initialized = EGL_TRUE;
@@ -152,9 +152,9 @@ demoCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext
 
    /* generate handle and insert into hash table */
    _eglSaveContext(&c->Base);
-   assert(c->Base.Handle);
+   assert(_eglGetContextHandle(&c->Base));
 
-   return c->Base.Handle;
+   return _eglGetContextHandle(&c->Base);
 }
 
 
@@ -286,7 +286,7 @@ demoMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface rea
  * plug in API functions.
  */
 _EGLDriver *
-_eglMain(_EGLDisplay *dpy)
+_eglMain(_EGLDisplay *dpy, const char *args)
 {
    DemoDriver *demo;
 
diff --git a/src/egl/drivers/dri/Makefile b/src/egl/drivers/dri/Makefile
index af0c40fe2b..be2f9b6bea 100644
--- a/src/egl/drivers/dri/Makefile
+++ b/src/egl/drivers/dri/Makefile
@@ -8,7 +8,7 @@ include $(TOP)/configs/current
 INCLUDE_DIRS = \
 	-I. \
 	-I/usr/include \
-	-I/usr/include/drm \
+	$(shell pkg-config --cflags-only-I libdrm) \
 	-I$(TOP)/include \
 	-I$(TOP)/include/GL/internal \
 	-I$(TOP)/src/mesa \
@@ -29,6 +29,8 @@ SOURCES = egldri.c
 
 OBJECTS = $(SOURCES:.c=.o)
 
+DRM_LIB = `pkg-config --libs libdrm`
+
 
 .c.o:
 	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
@@ -44,8 +46,10 @@ library: $(TOP)/$(LIB_DIR)/libEGLdri.so
 $(TOP)/$(LIB_DIR)/libEGLdri.so: $(OBJECTS)
 	$(MKLIB) -o EGLdri -linker '$(CC)' -ldflags '$(LDFLAGS)' \
 		-major 1 -minor 0 \
-		-install $(TOP)/$(LIB_DIR) -ldl $(OBJECTS)
+		-install $(TOP)/$(LIB_DIR) -ldl $(OBJECTS) $(LIBS)
 
+install: $(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(INSTALL) $(TOP)/$(LIB_DIR)/libEGLdri.so $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	-rm -f *.o
@@ -53,9 +57,10 @@ clean:
 
 depend: $(SOURCES) $(HEADERS)
 	@ echo "running $(MKDEP)"
+	@ rm -f depend
 	@ touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) \
-		$(SOURCES) $(HEADERS) > /dev/null 
+		$(SOURCES) $(HEADERS) >/dev/null 2>/dev/null
 
 include depend
 # DO NOT DELETE
diff --git a/src/egl/drivers/dri/egldri.c b/src/egl/drivers/dri/egldri.c
index cab0be2bd1..57661cc3ab 100644
--- a/src/egl/drivers/dri/egldri.c
+++ b/src/egl/drivers/dri/egldri.c
@@ -1,5 +1,6 @@
 /**
- * Generic EGL driver for DRI.
+ * Generic EGL driver for DRI.  This is basically an "adaptor" driver
+ * that allows libEGL to load/use regular DRI drivers.
  *
  * This file contains all the code needed to interface DRI-based drivers
  * with libEGL.
@@ -23,6 +24,7 @@
 #include "egldisplay.h"
 #include "eglcontext.h"
 #include "eglconfig.h"
+#include "eglconfigutil.h"
 #include "eglsurface.h"
 #include "eglscreen.h"
 #include "eglglobals.h"
@@ -32,18 +34,72 @@
 #include "egldri.h"
 
 const char *sysfs = "/sys/class";
-#define None 0
+
 static const int empty_attribute_list[1] = { None };
 
 
+
+/**
+ * Given a card number, return the name of the DRI driver to use.
+ * This generally means reading the contents of
+ * /sys/class/drm/cardX/dri_library_name, where X is the card number
+ */
+static EGLBoolean
+driver_name_from_card_number(int card, char *driverName, int maxDriverName)
+{
+   char path[2000];
+   FILE *f;
+   int length;
+
+   snprintf(path, sizeof(path), "%s/drm/card%d/dri_library_name", sysfs, card);
+
+   f = fopen(path, "r");
+   if (!f)
+      return EGL_FALSE;
+
+   fgets(driverName, maxDriverName, f);
+   fclose(f);
+
+   if ((length = strlen(driverName)) > 1) {
+      /* remove the trailing newline from sysfs */
+      driverName[length - 1] = '\0';
+      strncat(driverName, "_dri", maxDriverName);
+      return EGL_TRUE;
+   }
+   else {
+      return EGL_FALSE;
+   }   
+}
+
+
+
 /**
  * The bootstrap function.
  * Return a new driDriver object and plug in API functions.
  * This function, in turn, loads a specific DRI driver (ex: r200_dri.so).
  */
 _EGLDriver *
-_eglMain(_EGLDisplay *dpy)
+_eglMain(_EGLDisplay *dpy, const char *args)
 {
+#if 1
+   const int card = args ? atoi(args) : 0;
+   _EGLDriver *driver = NULL;
+   char driverName[1000];
+
+   if (!driver_name_from_card_number(card, driverName, sizeof(driverName))) {
+      _eglLog(_EGL_WARNING,
+              "Unable to determine driver name for card %d\n", card);
+      return NULL;
+   }
+
+   _eglLog(_EGL_DEBUG, "Driver name: %s\n", driverName);
+
+   driver = _eglOpenDriver(dpy, driverName, args);
+
+   return driver;
+
+#else
+
    int length;
    char path[NAME_MAX];
    struct dirent *dirent;
@@ -58,14 +114,19 @@ _eglMain(_EGLDisplay *dpy)
       _eglLog(_EGL_WARNING, "%s DRM devices not found.", path);
       return EGL_FALSE;
    }
+
+   /* loop over dir entries looking for cardX where "X" is in the
+    * dpy->DriverName ":X" string.
+    */
    while ((dirent = readdir(dir))) {
 
       if (strncmp(&dirent->d_name[0], "card", 4) != 0)
          continue;
-      if (strcmp(&dirent->d_name[4], &dpy->Name[1]) != 0)
+      if (strcmp(&dirent->d_name[4], &driverName[1]) != 0)
          continue;
 
-      snprintf(path, sizeof(path), "%s/drm/card%s/dri_library_name", sysfs, &dpy->Name[1]);
+      snprintf(path, sizeof(path), "%s/drm/card%s/dri_library_name",
+               sysfs, &driverName[1]);
       _eglLog(_EGL_INFO, "Opening %s", path);
 #if 1
       file = fopen(path, "r");
@@ -89,6 +150,7 @@ _eglMain(_EGLDisplay *dpy)
    closedir(dir);
 
    return driver;
+#endif
 }
 
 
@@ -141,7 +203,7 @@ _eglDRICreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    /* generate handle and insert into hash table */
    _eglSaveContext(&c->Base);
 
-   return c->Base.Handle;
+   return _eglGetContextHandle(&c->Base);
 }
 
 
@@ -152,13 +214,15 @@ _eglDRIMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw,
    driDisplay *disp = Lookup_driDisplay(dpy);
    driContext *ctx = Lookup_driContext(context);
    EGLBoolean b;
+   __DRIid drawBuf = (__DRIid) draw;
+   __DRIid readBuf = (__DRIid) read;
 
    b = _eglMakeCurrent(drv, dpy, draw, read, context);
    if (!b)
       return EGL_FALSE;
 
    if (ctx) {
-      ctx->driContext.bindContext(disp, 0, read, draw, &ctx->driContext);
+      ctx->driContext.bindContext(disp, 0, drawBuf, readBuf, &ctx->driContext);
    }
    else {
       /* what's this??? */
@@ -190,7 +254,7 @@ _eglDRICreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
 #if 0
       GLcontext *ctx = NULL; /* this _should_ be OK */
 #endif
-      GLvisual visMode;
+      __GLcontextModes visMode;
       _EGLConfig *conf = _eglLookupConfig(drv, dpy, config);
       assert(conf); /* bad config should be caught earlier */
       _eglConfigToContextModesRec(conf, &visMode);
@@ -267,7 +331,8 @@ _eglDRICreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
    _EGLConfig *config = _eglLookupConfig(drv, dpy, cfg);
    driDisplay *disp = Lookup_driDisplay(dpy);
    driSurface *surface;
-   GLvisual visMode;
+   __GLcontextModes visMode;
+   __DRIid drawBuf;
 
    surface = (driSurface *) calloc(1, sizeof(*surface));
    if (!surface) {
@@ -292,8 +357,10 @@ _eglDRICreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
    /* convert EGLConfig to GLvisual */
    _eglConfigToContextModesRec(config, &visMode);
 
+   drawBuf = (__DRIid) _eglGetSurfaceHandle(&surface->Base);
+
    /* Create a new DRI drawable */
-   if (!disp->driScreen.createNewDrawable(disp, &visMode, surface->Base.Handle,
+   if (!disp->driScreen.createNewDrawable(disp, &visMode, drawBuf,
                                           &surface->drawable, GLX_WINDOW_BIT,
                                           empty_attribute_list)) {
       _eglRemoveSurface(&surface->Base);
@@ -715,7 +782,7 @@ __eglGetDrawableInfo(__DRInativeDisplay * ndpy, int screen, __DRIid drawable,
 {
     __DRIscreen *pDRIScreen;
     __DRIscreenPrivate *psp;
-   driSurface *surf = Lookup_driSurface(drawable);
+    driSurface *surf = Lookup_driSurface((EGLSurface) drawable);
 
    pDRIScreen = __eglFindDRIScreen(ndpy, screen);
 
@@ -1019,8 +1086,10 @@ _eglDRICreateDisplay(driDisplay *dpy, __DRIframebuffer *framebuffer)
                             api_ver,
                             & interface_methods,
                             NULL);
-   if (!dpy->driScreen.private)
+   if (!dpy->driScreen.private) {
+      _eglLog(_EGL_WARNING, "egldri.c: DRI create new screen failed");
       return EGL_FALSE;
+   }
 
    DRM_UNLOCK( dpy->drmFD, dpy->pSAREA, dpy->serverContext );
 
@@ -1080,6 +1149,7 @@ _eglDRIInitialize(_EGLDriver *drv, EGLDisplay dpy,
 {
    _EGLDisplay *disp = _eglLookupDisplay(dpy);
    driDisplay *display;
+   const char *driverName = (const char *) disp->NativeDisplay;
 
    assert(disp);
 
@@ -1088,13 +1158,13 @@ _eglDRIInitialize(_EGLDriver *drv, EGLDisplay dpy,
     */
    display = calloc(1, sizeof(*display));
    display->Base = *disp;
-   _eglHashInsert(_eglGlobal.Displays, disp->Handle, display);
+   _eglSaveDisplay(&display->Base);
    free(disp);
 
    *major = 1;
    *minor = 0;
 
-   sscanf(&disp->Name[1], "%d", &display->minor);
+   sscanf(driverName + 1, "%d", &display->minor);
 
    drv->Initialized = EGL_TRUE;
    return EGL_TRUE;
diff --git a/src/egl/drivers/dri/egldri.h b/src/egl/drivers/dri/egldri.h
index 34b12d64fc..54a9a4ea26 100644
--- a/src/egl/drivers/dri/egldri.h
+++ b/src/egl/drivers/dri/egldri.h
@@ -1,11 +1,14 @@
 #ifndef EGLDRI_INCLUDED
 #define EGLDRI_INCLUDED
 
+#include <stdlib.h>
+#include <string.h>
+#include <stdint.h>
 #include "egldisplay.h"
 #include "eglscreen.h"
 #include "eglsurface.h"
 #include "eglcontext.h"
-#include "mtypes.h"
+
 #include "dri_util.h"
 #include "drm_sarea.h"
 
@@ -14,7 +17,7 @@
  */
 typedef struct dri_display
 {
-   _EGLDisplay Base;  /* base class/object */
+   _EGLDisplay Base;  /**< base class */
    void *pFB;
    int drmFD;  /**< \brief DRM device file descriptor */
    int minor;
@@ -32,7 +35,7 @@ typedef struct dri_display
    unsigned long FBStart;   /**< \brief physical address of the framebuffer */
    void *driverClientMsg;
    int driverClientMsgSize;
-   int chipset;
+   unsigned chipset;
    void *driverPrivate;
    drm_magic_t magic;
 
diff --git a/src/egl/drivers/glx/Makefile b/src/egl/drivers/glx/Makefile
new file mode 100644
index 0000000000..090a230ca3
--- /dev/null
+++ b/src/egl/drivers/glx/Makefile
@@ -0,0 +1,76 @@
+# src/egl/drivers/glx/Makefile
+
+# Build XEGL DRI driver loader library:  egl_glx.so
+
+
+TOP = ../../../..
+include $(TOP)/configs/current
+
+
+EXTRA_DEFINES = -DDEFAULT_DRIVER_DIR=\"$(DRI_DRIVER_SEARCH_DIR)\"
+
+DRIVER_NAME = egl_glx.so
+
+
+INCLUDE_DIRS = \
+	-I. \
+	-I/usr/include \
+	$(shell pkg-config --cflags-only-I libdrm) \
+	-I$(TOP)/include \
+	-I$(TOP)/include/GL/internal \
+	-I$(TOP)/src/mesa/glapi \
+	-I$(TOP)/src/mesa/drivers/dri/common \
+	-I$(TOP)/src/mesa/main \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/egl/main \
+	-I$(TOP)/src/glx/x11
+
+SOURCES = egl_glx.c
+
+OBJECTS = $(SOURCES:.c=.o)
+
+DRM_LIB = `pkg-config --libs libdrm`
+
+MISC_LIBS = -ldl -lXext -lGL
+
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(EXTRA_DEFINES) $< -o $@
+
+
+.PHONY: library
+
+
+default: depend library Makefile
+
+
+library: $(TOP)/$(LIB_DIR)/$(DRIVER_NAME)
+
+
+# Make the egl_glx.so library
+$(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(OBJECTS)
+	$(TOP)/bin/mklib -o $(DRIVER_NAME) \
+		-noprefix \
+		-major 1 -minor 0 \
+		-L$(TOP)/$(LIB_DIR) \
+		-install $(TOP)/$(LIB_DIR) \
+		$(OBJECTS) $(DRM_LIB) $(MISC_LIBS)
+
+install: $(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(DESTDIR)$(INSTALL_LIB_DIR)
+
+clean:
+	rm -f *.o
+	rm -f *.so
+	rm -f depend depend.bak
+
+
+depend: $(SOURCES) $(HEADERS)
+	@ echo "running $(MKDEP)"
+	@ rm -f depend
+	@ touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) \
+		$(SOURCES) $(HEADERS) >/dev/null 2>/dev/null
+
+include depend
+# DO NOT DELETE
diff --git a/src/egl/drivers/glx/egl_glx.c b/src/egl/drivers/glx/egl_glx.c
new file mode 100644
index 0000000000..8ba70ba7a6
--- /dev/null
+++ b/src/egl/drivers/glx/egl_glx.c
@@ -0,0 +1,848 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * This is an EGL driver that wraps GLX. This gives the benefit of being
+ * completely agnostic of the direct rendering implementation.
+ *
+ * Authors: Alan Hourihane <alanh@tungstengraphics.com>
+ */
+
+/*
+ * TODO: 
+ *
+ * test eglBind/ReleaseTexImage
+ */
+
+
+#include <assert.h>
+#include <stdio.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include "dlfcn.h"
+#include <X11/Xlib.h>
+#include <GL/gl.h>
+#include "glxclient.h"
+
+#define _EGL_PLATFORM_X
+
+#include "eglconfig.h"
+#include "eglcontext.h"
+#include "egldisplay.h"
+#include "egldriver.h"
+#include "eglglobals.h"
+#include "eglhash.h"
+#include "egllog.h"
+#include "eglsurface.h"
+
+#include <GL/gl.h>
+
+#define CALLOC_STRUCT(T)   (struct T *) calloc(1, sizeof(struct T))
+
+static const EGLint all_apis = (EGL_OPENGL_ES_BIT
+                                | EGL_OPENGL_ES2_BIT
+                                | EGL_OPENVG_BIT
+                                /* | EGL_OPENGL_BIT */); /* can't do */
+
+struct visual_attribs
+{
+   /* X visual attribs */
+   int id;
+   int klass;
+   int depth;
+   int redMask, greenMask, blueMask;
+   int colormapSize;
+   int bitsPerRGB;
+
+   /* GL visual attribs */
+   int supportsGL;
+   int transparentType;
+   int transparentRedValue;
+   int transparentGreenValue;
+   int transparentBlueValue;
+   int transparentAlphaValue;
+   int transparentIndexValue;
+   int bufferSize;
+   int level;
+   int render_type;
+   int doubleBuffer;
+   int stereo;
+   int auxBuffers;
+   int redSize, greenSize, blueSize, alphaSize;
+   int depthSize;
+   int stencilSize;
+   int accumRedSize, accumGreenSize, accumBlueSize, accumAlphaSize;
+   int numSamples, numMultisample;
+   int visualCaveat;
+};
+
+/** subclass of _EGLDriver */
+struct GLX_egl_driver
+{
+   _EGLDriver Base;   /**< base class */
+
+   XVisualInfo *visuals;
+   GLXFBConfig *fbconfigs;
+
+   int glx_maj, glx_min;
+};
+
+
+/** subclass of _EGLContext */
+struct GLX_egl_context
+{
+   _EGLContext Base;   /**< base class */
+
+   GLXContext context;
+};
+
+
+/** subclass of _EGLSurface */
+struct GLX_egl_surface
+{
+   _EGLSurface Base;   /**< base class */
+
+   GLXDrawable drawable;
+};
+
+
+/** subclass of _EGLConfig */
+struct GLX_egl_config
+{
+   _EGLConfig Base;   /**< base class */
+};
+
+/** cast wrapper */
+static struct GLX_egl_driver *
+GLX_egl_driver(_EGLDriver *drv)
+{
+   return (struct GLX_egl_driver *) drv;
+}
+
+static struct GLX_egl_context *
+GLX_egl_context(_EGLContext *ctx)
+{
+   return (struct GLX_egl_context *) ctx;
+}
+
+static struct GLX_egl_surface *
+GLX_egl_surface(_EGLSurface *surf)
+{
+   return (struct GLX_egl_surface *) surf;
+}
+
+static GLboolean
+get_visual_attribs(Display *dpy, XVisualInfo *vInfo,
+                   struct visual_attribs *attribs)
+{
+   const char *ext = glXQueryExtensionsString(dpy, vInfo->screen);
+   int rgba;
+
+   memset(attribs, 0, sizeof(struct visual_attribs));
+
+   attribs->id = vInfo->visualid;
+#if defined(__cplusplus) || defined(c_plusplus)
+   attribs->klass = vInfo->c_class;
+#else
+   attribs->klass = vInfo->class;
+#endif
+   attribs->depth = vInfo->depth;
+   attribs->redMask = vInfo->red_mask;
+   attribs->greenMask = vInfo->green_mask;
+   attribs->blueMask = vInfo->blue_mask;
+   attribs->colormapSize = vInfo->colormap_size;
+   attribs->bitsPerRGB = vInfo->bits_per_rgb;
+
+   if (glXGetConfig(dpy, vInfo, GLX_USE_GL, &attribs->supportsGL) != 0 ||
+       !attribs->supportsGL)
+      return GL_FALSE;
+   glXGetConfig(dpy, vInfo, GLX_BUFFER_SIZE, &attribs->bufferSize);
+   glXGetConfig(dpy, vInfo, GLX_LEVEL, &attribs->level);
+   glXGetConfig(dpy, vInfo, GLX_RGBA, &rgba);
+   if (!rgba)
+      return GL_FALSE;
+   attribs->render_type = GLX_RGBA_BIT;
+   
+   glXGetConfig(dpy, vInfo, GLX_DOUBLEBUFFER, &attribs->doubleBuffer);
+   if (!attribs->doubleBuffer)
+      return GL_FALSE;
+
+   glXGetConfig(dpy, vInfo, GLX_STEREO, &attribs->stereo);
+   glXGetConfig(dpy, vInfo, GLX_AUX_BUFFERS, &attribs->auxBuffers);
+   glXGetConfig(dpy, vInfo, GLX_RED_SIZE, &attribs->redSize);
+   glXGetConfig(dpy, vInfo, GLX_GREEN_SIZE, &attribs->greenSize);
+   glXGetConfig(dpy, vInfo, GLX_BLUE_SIZE, &attribs->blueSize);
+   glXGetConfig(dpy, vInfo, GLX_ALPHA_SIZE, &attribs->alphaSize);
+   glXGetConfig(dpy, vInfo, GLX_DEPTH_SIZE, &attribs->depthSize);
+   glXGetConfig(dpy, vInfo, GLX_STENCIL_SIZE, &attribs->stencilSize);
+   glXGetConfig(dpy, vInfo, GLX_ACCUM_RED_SIZE, &attribs->accumRedSize);
+   glXGetConfig(dpy, vInfo, GLX_ACCUM_GREEN_SIZE, &attribs->accumGreenSize);
+   glXGetConfig(dpy, vInfo, GLX_ACCUM_BLUE_SIZE, &attribs->accumBlueSize);
+   glXGetConfig(dpy, vInfo, GLX_ACCUM_ALPHA_SIZE, &attribs->accumAlphaSize);
+
+   /* get transparent pixel stuff */
+   glXGetConfig(dpy, vInfo,GLX_TRANSPARENT_TYPE, &attribs->transparentType);
+   if (attribs->transparentType == GLX_TRANSPARENT_RGB) {
+     glXGetConfig(dpy, vInfo, GLX_TRANSPARENT_RED_VALUE, &attribs->transparentRedValue);
+     glXGetConfig(dpy, vInfo, GLX_TRANSPARENT_GREEN_VALUE, &attribs->transparentGreenValue);
+     glXGetConfig(dpy, vInfo, GLX_TRANSPARENT_BLUE_VALUE, &attribs->transparentBlueValue);
+     glXGetConfig(dpy, vInfo, GLX_TRANSPARENT_ALPHA_VALUE, &attribs->transparentAlphaValue);
+   }
+   else if (attribs->transparentType == GLX_TRANSPARENT_INDEX) {
+     glXGetConfig(dpy, vInfo, GLX_TRANSPARENT_INDEX_VALUE, &attribs->transparentIndexValue);
+   }
+
+   /* multisample attribs */
+#ifdef GLX_ARB_multisample
+   if (ext && strstr(ext, "GLX_ARB_multisample")) {
+      glXGetConfig(dpy, vInfo, GLX_SAMPLE_BUFFERS_ARB, &attribs->numMultisample);
+      glXGetConfig(dpy, vInfo, GLX_SAMPLES_ARB, &attribs->numSamples);
+   }
+#endif
+   else {
+      attribs->numSamples = 0;
+      attribs->numMultisample = 0;
+   }
+
+#if defined(GLX_EXT_visual_rating)
+   if (ext && strstr(ext, "GLX_EXT_visual_rating")) {
+      glXGetConfig(dpy, vInfo, GLX_VISUAL_CAVEAT_EXT, &attribs->visualCaveat);
+   }
+   else {
+      attribs->visualCaveat = GLX_NONE_EXT;
+   }
+#else
+   attribs->visualCaveat = 0;
+#endif
+
+   return GL_TRUE;
+}
+
+#ifdef GLX_VERSION_1_3
+
+static int
+glx_token_to_visual_class(int visual_type)
+{
+   switch (visual_type) {
+   case GLX_TRUE_COLOR:
+      return TrueColor;
+   case GLX_DIRECT_COLOR:
+      return DirectColor;
+   case GLX_PSEUDO_COLOR:
+      return PseudoColor;
+   case GLX_STATIC_COLOR:
+      return StaticColor;
+   case GLX_GRAY_SCALE:
+      return GrayScale;
+   case GLX_STATIC_GRAY:
+      return StaticGray;
+   case GLX_NONE:
+   default:
+      return None;
+   }
+}
+static GLboolean
+get_fbconfig_attribs(Display *dpy, GLXFBConfig fbconfig,
+		     struct visual_attribs *attribs)
+{
+   int visual_type;
+
+   memset(attribs, 0, sizeof(struct visual_attribs));
+
+   /* We don't use the GLX_FBCONFIG_ID here */
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_VISUAL_ID, &attribs->id);
+
+#if 0
+   attribs->depth = vInfo->depth;
+   attribs->redMask = vInfo->red_mask;
+   attribs->greenMask = vInfo->green_mask;
+   attribs->blueMask = vInfo->blue_mask;
+   attribs->colormapSize = vInfo->colormap_size;
+   attribs->bitsPerRGB = vInfo->bits_per_rgb;
+#endif
+
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_X_VISUAL_TYPE, &visual_type);
+   attribs->klass = glx_token_to_visual_class(visual_type);
+
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_BUFFER_SIZE, &attribs->bufferSize);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_LEVEL, &attribs->level);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_RENDER_TYPE, &attribs->render_type);
+   if (!(attribs->render_type & GLX_RGBA_BIT))
+      return GL_FALSE;
+
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_DOUBLEBUFFER, &attribs->doubleBuffer);
+   if (!attribs->doubleBuffer)
+      return GL_FALSE;
+
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_STEREO, &attribs->stereo);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_AUX_BUFFERS, &attribs->auxBuffers);
+
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_RED_SIZE, &attribs->redSize);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_GREEN_SIZE, &attribs->greenSize);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_BLUE_SIZE, &attribs->blueSize);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_ALPHA_SIZE, &attribs->alphaSize);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_DEPTH_SIZE, &attribs->depthSize);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_STENCIL_SIZE, &attribs->stencilSize);
+
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_ACCUM_RED_SIZE, &attribs->accumRedSize);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_ACCUM_GREEN_SIZE, &attribs->accumGreenSize);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_ACCUM_BLUE_SIZE, &attribs->accumBlueSize);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_ACCUM_ALPHA_SIZE, &attribs->accumAlphaSize);
+
+   /* get transparent pixel stuff */
+   glXGetFBConfigAttrib(dpy, fbconfig,GLX_TRANSPARENT_TYPE, &attribs->transparentType);
+   if (attribs->transparentType == GLX_TRANSPARENT_RGB) {
+     glXGetFBConfigAttrib(dpy, fbconfig, GLX_TRANSPARENT_RED_VALUE, &attribs->transparentRedValue);
+     glXGetFBConfigAttrib(dpy, fbconfig, GLX_TRANSPARENT_GREEN_VALUE, &attribs->transparentGreenValue);
+     glXGetFBConfigAttrib(dpy, fbconfig, GLX_TRANSPARENT_BLUE_VALUE, &attribs->transparentBlueValue);
+     glXGetFBConfigAttrib(dpy, fbconfig, GLX_TRANSPARENT_ALPHA_VALUE, &attribs->transparentAlphaValue);
+   }
+   else if (attribs->transparentType == GLX_TRANSPARENT_INDEX) {
+     glXGetFBConfigAttrib(dpy, fbconfig, GLX_TRANSPARENT_INDEX_VALUE, &attribs->transparentIndexValue);
+   }
+
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_SAMPLE_BUFFERS, &attribs->numMultisample);
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_SAMPLES, &attribs->numSamples);
+
+   glXGetFBConfigAttrib(dpy, fbconfig, GLX_CONFIG_CAVEAT, &attribs->visualCaveat);
+
+   return GL_TRUE;
+}
+
+#endif
+
+static EGLBoolean
+create_configs(_EGLDisplay *disp, struct GLX_egl_driver *GLX_drv)
+{
+   XVisualInfo theTemplate;
+   int numVisuals;
+   long mask;
+   int i;
+   struct visual_attribs attribs;
+
+   GLX_drv->fbconfigs = NULL;
+
+#ifdef GLX_VERSION_1_3
+   /* get list of all fbconfigs on this screen */
+   GLX_drv->fbconfigs = glXGetFBConfigs(disp->Xdpy, DefaultScreen(disp->Xdpy), &numVisuals);
+
+   if (numVisuals == 0) {
+      GLX_drv->fbconfigs = NULL;
+      goto xvisual;
+   }
+
+   for (i = 0; i < numVisuals; i++) {
+      struct GLX_egl_config *config;
+
+      if (!get_fbconfig_attribs(disp->Xdpy, GLX_drv->fbconfigs[i], &attribs))
+         continue;
+
+      config = CALLOC_STRUCT(GLX_egl_config);
+
+      _eglInitConfig(&config->Base, i+1);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_ID, attribs.id);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_BUFFER_SIZE, attribs.bufferSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_RED_SIZE, attribs.redSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_GREEN_SIZE, attribs.greenSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_BLUE_SIZE, attribs.blueSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_ALPHA_SIZE, attribs.alphaSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_DEPTH_SIZE, attribs.depthSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_STENCIL_SIZE, attribs.stencilSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_SAMPLES, attribs.numSamples);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_SAMPLE_BUFFERS, attribs.numMultisample);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_CONFORMANT, all_apis);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_RENDERABLE_TYPE, all_apis);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_SURFACE_TYPE,
+                        (EGL_WINDOW_BIT | EGL_PBUFFER_BIT | EGL_PIXMAP_BIT));
+
+      /* XXX possibly other things to init... */
+
+      _eglAddConfig(disp, &config->Base);
+   }
+
+   goto end;
+#endif
+
+xvisual:
+   /* get list of all visuals on this screen */
+   theTemplate.screen = DefaultScreen(disp->Xdpy);
+   mask = VisualScreenMask;
+   GLX_drv->visuals = XGetVisualInfo(disp->Xdpy, mask, &theTemplate, &numVisuals);
+
+   for (i = 0; i < numVisuals; i++) {
+      struct GLX_egl_config *config;
+
+      if (!get_visual_attribs(disp->Xdpy, &GLX_drv->visuals[i], &attribs))
+	 continue;
+
+      config = CALLOC_STRUCT(GLX_egl_config);
+
+      _eglInitConfig(&config->Base, i+1);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_ID, attribs.id);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_BUFFER_SIZE, attribs.bufferSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_RED_SIZE, attribs.redSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_GREEN_SIZE, attribs.greenSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_BLUE_SIZE, attribs.blueSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_ALPHA_SIZE, attribs.alphaSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_DEPTH_SIZE, attribs.depthSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_STENCIL_SIZE, attribs.stencilSize);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_SAMPLES, attribs.numSamples);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_SAMPLE_BUFFERS, attribs.numMultisample);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_CONFORMANT, all_apis);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_RENDERABLE_TYPE, all_apis);
+      SET_CONFIG_ATTRIB(&config->Base, EGL_SURFACE_TYPE,
+                        (EGL_WINDOW_BIT /*| EGL_PBUFFER_BIT | EGL_PIXMAP_BIT*/));
+
+      /* XXX possibly other things to init... */
+
+      _eglAddConfig(disp, &config->Base);
+   }
+
+end:
+   return EGL_TRUE;
+}
+
+/**
+ * Called via eglInitialize(), GLX_drv->API.Initialize().
+ */
+static EGLBoolean
+GLX_eglInitialize(_EGLDriver *drv, EGLDisplay dpy,
+                   EGLint *minor, EGLint *major)
+{
+   struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+
+   _eglLog(_EGL_DEBUG, "GLX: eglInitialize");
+
+   if (!disp->Xdpy) {
+      disp->Xdpy = XOpenDisplay(NULL);
+      if (!disp->Xdpy) {
+         _eglLog(_EGL_WARNING, "GLX: XOpenDisplay failed");
+         return EGL_FALSE;
+      }
+   } 
+
+   glXQueryVersion(disp->Xdpy, &GLX_drv->glx_maj, &GLX_drv->glx_min);
+   
+   GLX_drv->Base.Initialized = EGL_TRUE;
+
+   GLX_drv->Base.Name = "GLX";
+
+   /* we're supporting EGL 1.4 */
+   *minor = 1;
+   *major = 4;
+
+   create_configs(disp, GLX_drv);
+
+   return EGL_TRUE;
+}
+
+/*
+ * Do some clean-up that normally occurs in XCloseDisplay().
+ * We do this here because we're about to unload a dynamic library
+ * that has added some per-display extension data and callbacks.
+ * If we don't do this here we'll crash in XCloseDisplay() because it'll
+ * try to call functions that went away when the driver library was unloaded.
+ */
+static void
+FreeDisplayExt(Display *dpy)
+{
+   _XExtension *ext, *next;
+
+   for (ext = dpy->ext_procs; ext; ext = next) {
+      next = ext->next;
+      if (ext->close_display) {
+         ext->close_display(dpy, &ext->codes);
+         ext->close_display = NULL;
+      }
+      if (ext->name)
+         Xfree(ext->name);
+      Xfree(ext);
+   }
+   dpy->ext_procs = NULL;
+
+   _XFreeExtData (dpy->ext_data);
+   dpy->ext_data = NULL;
+}
+
+/**
+ * Called via eglTerminate(), drv->API.Terminate().
+ */
+static EGLBoolean
+GLX_eglTerminate(_EGLDriver *drv, EGLDisplay dpy)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+
+   _eglLog(_EGL_DEBUG, "GLX: eglTerminate");
+
+   FreeDisplayExt(disp->Xdpy);
+
+   return EGL_TRUE;
+}
+
+
+/**
+ * Called via eglCreateContext(), drv->API.CreateContext().
+ */
+static EGLContext
+GLX_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+                      EGLContext share_list, const EGLint *attrib_list)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct GLX_egl_context *GLX_ctx = CALLOC_STRUCT(GLX_egl_context);
+   struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
+   struct GLX_egl_context *GLX_ctx_shared = NULL;
+   _EGLConfig *conf;
+
+   if (!GLX_ctx)
+      return EGL_NO_CONTEXT;
+
+   if (!_eglInitContext(drv, dpy, &GLX_ctx->Base, config, attrib_list)) {
+      free(GLX_ctx);
+      return EGL_NO_CONTEXT;
+   }
+
+   if (share_list != EGL_NO_CONTEXT) {
+      _EGLContext *shareCtx = _eglLookupContext(share_list);
+      if (!shareCtx) {
+         _eglError(EGL_BAD_CONTEXT, "eglCreateContext(share_list)");
+         return EGL_FALSE;
+      }
+      GLX_ctx_shared = GLX_egl_context(shareCtx);
+   }
+
+   conf = _eglLookupConfig(drv, dpy, config);
+   assert(conf);
+
+#ifdef GLX_VERSION_1_3
+   if (GLX_drv->fbconfigs)
+      GLX_ctx->context = glXCreateNewContext(disp->Xdpy, GLX_drv->fbconfigs[(int)config-1], GLX_RGBA_TYPE, GLX_ctx_shared ? GLX_ctx_shared->context : NULL, GL_TRUE);
+   else
+#endif
+      GLX_ctx->context = glXCreateContext(disp->Xdpy, &GLX_drv->visuals[(int)config-1], GLX_ctx_shared ? GLX_ctx_shared->context : NULL, GL_TRUE);
+   if (!GLX_ctx->context)
+      return EGL_FALSE;
+
+#if 1
+   /* (maybe?) need to have a direct rendering context */
+   if (!glXIsDirect(disp->Xdpy, GLX_ctx->context))
+      return EGL_FALSE;
+#endif
+
+   return _eglGetContextHandle(&GLX_ctx->Base);
+}
+
+
+/**
+ * Called via eglMakeCurrent(), drv->API.MakeCurrent().
+ */
+static EGLBoolean
+GLX_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
+                    EGLSurface r, EGLContext context)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLContext *ctx = _eglLookupContext(context);
+   _EGLSurface *dsurf = _eglLookupSurface(d);
+   _EGLSurface *rsurf = _eglLookupSurface(r);
+   struct GLX_egl_surface *GLX_dsurf = GLX_egl_surface(dsurf);
+   struct GLX_egl_surface *GLX_rsurf = GLX_egl_surface(rsurf);
+   struct GLX_egl_context *GLX_ctx = GLX_egl_context(ctx);
+
+   if (!_eglMakeCurrent(drv, dpy, d, r, context))
+      return EGL_FALSE;
+
+#ifdef GLX_VERSION_1_3
+   if (!glXMakeContextCurrent(disp->Xdpy, GLX_dsurf ? GLX_dsurf->drawable : 0, GLX_rsurf ? GLX_rsurf->drawable : 0, GLX_ctx ? GLX_ctx->context : NULL))
+#endif
+      if (!glXMakeCurrent(disp->Xdpy, GLX_dsurf ? GLX_dsurf->drawable : 0, GLX_ctx ? GLX_ctx->context : NULL))
+         return EGL_FALSE;
+
+   return EGL_TRUE;
+}
+
+/** Get size of given window */
+static Status
+get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height)
+{
+   Window root;
+   Status stat;
+   int xpos, ypos;
+   unsigned int w, h, bw, depth;
+   stat = XGetGeometry(dpy, d, &root, &xpos, &ypos, &w, &h, &bw, &depth);
+   *width = w;
+   *height = h;
+   return stat;
+}
+
+/**
+ * Called via eglCreateWindowSurface(), drv->API.CreateWindowSurface().
+ */
+static EGLSurface
+GLX_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+                            NativeWindowType window, const EGLint *attrib_list)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct GLX_egl_surface *GLX_surf;
+   uint width, height;
+
+   GLX_surf = CALLOC_STRUCT(GLX_egl_surface);
+   if (!GLX_surf)
+      return EGL_NO_SURFACE;
+
+   if (!_eglInitSurface(drv, dpy, &GLX_surf->Base, EGL_WINDOW_BIT,
+                        config, attrib_list)) {
+      free(GLX_surf);
+      return EGL_FALSE;
+   }
+
+   _eglSaveSurface(&GLX_surf->Base);
+
+   GLX_surf->drawable = window;
+   get_drawable_size(disp->Xdpy, window, &width, &height);
+   GLX_surf->Base.Width = width;
+   GLX_surf->Base.Height = height;
+
+   return _eglGetSurfaceHandle(&GLX_surf->Base);
+}
+
+#ifdef GLX_VERSION_1_3
+static EGLSurface
+GLX_eglCreatePixmapSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, 
+			   NativePixmapType pixmap, const EGLint *attrib_list)
+{
+   struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct GLX_egl_surface *GLX_surf;
+   int i;
+
+   GLX_surf = CALLOC_STRUCT(GLX_egl_surface);
+   if (!GLX_surf)
+      return EGL_NO_SURFACE;
+
+   if (!_eglInitSurface(drv, dpy, &GLX_surf->Base, EGL_PIXMAP_BIT,
+                        config, attrib_list)) {
+      free(GLX_surf);
+      return EGL_FALSE;
+   }
+
+   _eglSaveSurface(&GLX_surf->Base);
+
+   for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
+      switch (attrib_list[i]) {
+         /* no attribs at this time */
+      default:
+         _eglError(EGL_BAD_ATTRIBUTE, "eglCreatePixmapSurface");
+         return EGL_NO_SURFACE;
+      }
+   }
+
+   GLX_surf->drawable = glXCreatePixmap(disp->Xdpy, GLX_drv->fbconfigs[(int)config-1], pixmap, NULL);
+
+   return _eglGetSurfaceHandle(&GLX_surf->Base);
+}
+
+static EGLSurface
+GLX_eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+                            const EGLint *attrib_list)
+{
+   struct GLX_egl_driver *GLX_drv = GLX_egl_driver(drv);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct GLX_egl_surface *GLX_surf;
+   int attribs[5];
+   int i = 0, j = 0;
+
+   GLX_surf = CALLOC_STRUCT(GLX_egl_surface);
+   if (!GLX_surf)
+      return EGL_NO_SURFACE;
+
+   if (!_eglInitSurface(drv, dpy, &GLX_surf->Base, EGL_PBUFFER_BIT,
+                        config, attrib_list)) {
+      free(GLX_surf);
+      return EGL_NO_SURFACE;
+   }
+
+   _eglSaveSurface(&GLX_surf->Base);
+
+   while(attrib_list[i] != EGL_NONE) {
+      switch (attrib_list[i]) {
+         case EGL_WIDTH:
+	    attribs[j++] = GLX_PBUFFER_WIDTH;
+	    attribs[j++] = attrib_list[i+1];
+	    break;
+	 case EGL_HEIGHT:
+	    attribs[j++] = GLX_PBUFFER_HEIGHT;
+	    attribs[j++] = attrib_list[i+1];
+	    break;
+      }
+      i++;
+   }
+   attribs[j++] = 0;
+
+   GLX_surf->drawable = glXCreatePbuffer(disp->Xdpy, GLX_drv->fbconfigs[(int)config-1], attribs);
+
+   return _eglGetSurfaceHandle(&GLX_surf->Base);
+}
+#endif
+
+static EGLBoolean
+GLX_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLSurface *surf = _eglLookupSurface(surface);
+   return EGL_TRUE;
+   if (surf) {
+      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
+      if (surf->IsBound) {
+         surf->DeletePending = EGL_TRUE;
+      }
+      else {
+         free(surf);
+      }
+
+      return EGL_TRUE;
+   }
+   else {
+      _eglError(EGL_BAD_SURFACE, "eglDestroySurface");
+      return EGL_FALSE;
+   }
+}
+
+
+static EGLBoolean
+GLX_eglBindTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface,
+                     EGLint buffer)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLSurface *surf = _eglLookupSurface(surface);
+   struct GLX_egl_surface *GLX_surf = GLX_egl_surface(surf);
+
+   /* buffer ?? */
+   glXBindTexImageEXT(disp->Xdpy, GLX_surf->drawable, GLX_FRONT_LEFT_EXT, NULL);
+
+   return EGL_TRUE;
+}
+
+
+static EGLBoolean
+GLX_eglReleaseTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface,
+                        EGLint buffer)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLSurface *surf = _eglLookupSurface(surface);
+   struct GLX_egl_surface *GLX_surf = GLX_egl_surface(surf);
+
+   /* buffer ?? */
+   glXReleaseTexImageEXT(disp->Xdpy, GLX_surf->drawable, GLX_FRONT_LEFT_EXT);
+
+   return EGL_TRUE;
+}
+
+
+static EGLBoolean
+GLX_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLSurface *surf = _eglLookupSurface(draw);
+   struct GLX_egl_surface *GLX_surf = GLX_egl_surface(surf);
+
+   _eglLog(_EGL_DEBUG, "GLX: EGL SwapBuffers 0x%x",draw);
+
+   /* error checking step: */
+   if (!_eglSwapBuffers(drv, dpy, draw))
+      return EGL_FALSE;
+
+   glXSwapBuffers(disp->Xdpy, GLX_surf->drawable);
+
+   return EGL_TRUE;
+}
+
+/*
+ * Called from eglGetProcAddress() via drv->API.GetProcAddress().
+ */
+static _EGLProc
+GLX_eglGetProcAddress(const char *procname)
+{
+   return (_EGLProc)glXGetProcAddress((const GLubyte *)procname);   
+}
+
+
+/**
+ * This is the main entrypoint into the driver, called by libEGL.
+ * Create a new _EGLDriver object and init its dispatch table.
+ */
+_EGLDriver *
+_eglMain(_EGLDisplay *disp, const char *args)
+{
+   struct GLX_egl_driver *GLX_drv = CALLOC_STRUCT(GLX_egl_driver);
+   char *env;
+   int maj = 0, min = 0;
+
+   if (!GLX_drv)
+      return NULL;
+
+   _eglInitDriverFallbacks(&GLX_drv->Base);
+   GLX_drv->Base.API.Initialize = GLX_eglInitialize;
+   GLX_drv->Base.API.Terminate = GLX_eglTerminate;
+   GLX_drv->Base.API.CreateContext = GLX_eglCreateContext;
+   GLX_drv->Base.API.MakeCurrent = GLX_eglMakeCurrent;
+   GLX_drv->Base.API.CreateWindowSurface = GLX_eglCreateWindowSurface;
+#ifdef GLX_VERSION_1_3
+   if (GLX_drv->glx_maj == 1 && GLX_drv->glx_min >= 3) {
+      GLX_drv->Base.API.CreatePixmapSurface = GLX_eglCreatePixmapSurface;
+      GLX_drv->Base.API.CreatePbufferSurface = GLX_eglCreatePbufferSurface;
+      printf("GLX: Pbuffer and Pixmap support\n");
+   } else {
+      printf("GLX: No pbuffer or pixmap support\n");
+   }
+#endif
+   GLX_drv->Base.API.DestroySurface = GLX_eglDestroySurface;
+   GLX_drv->Base.API.BindTexImage = GLX_eglBindTexImage;
+   GLX_drv->Base.API.ReleaseTexImage = GLX_eglReleaseTexImage;
+   GLX_drv->Base.API.SwapBuffers = GLX_eglSwapBuffers;
+   GLX_drv->Base.API.GetProcAddress = GLX_eglGetProcAddress;
+
+   GLX_drv->Base.ClientAPIsMask = all_apis;
+   GLX_drv->Base.Name = "GLX";
+
+   _eglLog(_EGL_DEBUG, "GLX: main(%s)", args);
+
+   /* set new DRI path to pick up EGL version (which doesn't contain any mesa 
+    * code), but don't override if one is already set.
+    */
+   env = getenv("LIBGL_DRIVERS_PATH");
+   if (env) {
+      if (!strstr(env, "egl")) {
+         sprintf(env, "%s/egl", env);
+         setenv("LIBGL_DRIVERS_PATH", env, 1);
+      }
+   } else
+      setenv("LIBGL_DRIVERS_PATH", DEFAULT_DRIVER_DIR"/egl", 0);
+
+   return &GLX_drv->Base;
+}
diff --git a/src/egl/drivers/xdri/Makefile b/src/egl/drivers/xdri/Makefile
new file mode 100644
index 0000000000..db4b710002
--- /dev/null
+++ b/src/egl/drivers/xdri/Makefile
@@ -0,0 +1,72 @@
+# src/egl/drivers/xdri/Makefile
+
+# Build XEGL DRI driver loader library:  egl_xdri.so
+
+
+TOP = ../../../..
+include $(TOP)/configs/current
+
+
+DRIVER_NAME = egl_xdri.so
+
+
+INCLUDE_DIRS = \
+	-I. \
+	-I/usr/include \
+	$(shell pkg-config --cflags-only-I libdrm) \
+	-I$(TOP)/include \
+	-I$(TOP)/include/GL/internal \
+	-I$(TOP)/src/mesa/glapi \
+	-I$(TOP)/src/mesa/drivers/dri/common \
+	-I$(TOP)/src/egl/main \
+	-I$(TOP)/src/glx/x11
+
+SOURCES = egl_xdri.c
+
+OBJECTS = $(SOURCES:.c=.o)
+
+DRM_LIB = `pkg-config --libs libdrm`
+
+MISC_LIBS = -ldl -lXext -lGL
+
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
+
+
+.PHONY: library
+
+
+default: depend library Makefile
+
+
+library: $(TOP)/$(LIB_DIR)/$(DRIVER_NAME)
+
+
+# Make the egl_xdri.so library
+$(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(OBJECTS)
+	$(TOP)/bin/mklib -o $(DRIVER_NAME) \
+		-noprefix \
+		-major 1 -minor 0 \
+		-L$(TOP)/$(LIB_DIR) \
+		-install $(TOP)/$(LIB_DIR) \
+		$(OBJECTS) $(DRM_LIB) $(MISC_LIBS)
+
+install: $(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(DESTDIR)$(INSTALL_LIB_DIR)
+
+clean:
+	rm -f *.o
+	rm -f *.so
+	rm -f depend depend.bak
+
+
+depend: $(SOURCES) $(HEADERS)
+	@ echo "running $(MKDEP)"
+	@ rm -f depend
+	@ touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) \
+		$(SOURCES) $(HEADERS) >/dev/null 2>/dev/null
+
+include depend
+# DO NOT DELETE
diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c
new file mode 100644
index 0000000000..3b3e312746
--- /dev/null
+++ b/src/egl/drivers/xdri/egl_xdri.c
@@ -0,0 +1,1177 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Code to interface a DRI driver to libEGL.
+ * Note that unlike previous DRI/EGL interfaces, this one is meant to
+ * be used _with_ X.  Applications will use eglCreateWindowSurface()
+ * to render into X-created windows.
+ *
+ * This is an EGL driver that, in turn, loads a regular DRI driver.
+ * There are some dependencies on code in libGL, but those could be
+ * removed with some effort.
+ *
+ * Authors: Brian Paul
+ */
+
+
+#include <assert.h>
+#include <stdio.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include "dlfcn.h"
+#include <X11/Xlib.h>
+#include <GL/gl.h>
+#include "xf86dri.h"
+#include "glxclient.h"
+#include "dri_util.h"
+#include "drm_sarea.h"
+
+#define _EGL_PLATFORM_X
+
+#include "eglconfig.h"
+#include "eglcontext.h"
+#include "egldisplay.h"
+#include "egldriver.h"
+#include "eglglobals.h"
+#include "eglhash.h"
+#include "egllog.h"
+#include "eglsurface.h"
+
+#include <GL/gl.h>
+
+typedef void (*glGetIntegerv_t)(GLenum, GLint *);
+typedef void (*glBindTexture_t)(GLenum, GLuint);
+typedef void (*glCopyTexImage2D_t)(GLenum, GLint, GLenum, GLint, GLint,
+                                   GLint, GLint, GLint);
+
+
+#define CALLOC_STRUCT(T)   (struct T *) calloc(1, sizeof(struct T))
+
+
+/** subclass of _EGLDriver */
+struct xdri_egl_driver
+{
+   _EGLDriver Base;   /**< base class */
+
+   const char *dri_driver_name;  /**< name of DRI driver to load */
+   void *dri_driver_handle;      /**< returned by dlopen(dri_driver_name) */
+
+   __GLXdisplayPrivate *glx_priv;
+
+
+   /* XXX we're not actually using these at this time: */
+   int chipset;
+   int minor;
+   int drmFD;
+
+   __DRIframebuffer framebuffer;
+   drm_handle_t hSAREA;
+   drmAddress pSAREA;
+   char *busID;
+   drm_magic_t magic;
+};
+
+
+/** subclass of _EGLContext */
+struct xdri_egl_context
+{
+   _EGLContext Base;   /**< base class */
+
+   __DRIcontext driContext;
+
+   GLint bound_tex_object;
+};
+
+
+/** subclass of _EGLSurface */
+struct xdri_egl_surface
+{
+   _EGLSurface Base;   /**< base class */
+
+   __DRIid driDrawable;  /**< DRI surface */
+   drm_drawable_t hDrawable;
+};
+
+
+/** subclass of _EGLConfig */
+struct xdri_egl_config
+{
+   _EGLConfig Base;   /**< base class */
+
+   const __GLcontextModes *mode;  /**< corresponding GLX mode */
+};
+
+
+
+/** cast wrapper */
+static struct xdri_egl_driver *
+xdri_egl_driver(_EGLDriver *drv)
+{
+   return (struct xdri_egl_driver *) drv;
+}
+
+
+/** Map EGLSurface handle to xdri_egl_surface object */
+static struct xdri_egl_surface *
+lookup_surface(EGLSurface surf)
+{
+   _EGLSurface *surface = _eglLookupSurface(surf);
+   return (struct xdri_egl_surface *) surface;
+}
+
+
+/** Map EGLContext handle to xdri_egl_context object */
+static struct xdri_egl_context *
+lookup_context(EGLContext c)
+{
+   _EGLContext *context = _eglLookupContext(c);
+   return (struct xdri_egl_context *) context;
+}
+
+static struct xdri_egl_context *
+current_context(void)
+{
+   return (struct xdri_egl_context *) _eglGetCurrentContext();
+}
+
+/** Map EGLConfig handle to xdri_egl_config object */
+static struct xdri_egl_config *
+lookup_config(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config)
+{
+   _EGLConfig *conf = _eglLookupConfig(drv, dpy, config);
+   return (struct xdri_egl_config *) conf;
+}
+
+
+
+/** Get size of given window */
+static Status
+get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height)
+{
+   Window root;
+   Status stat;
+   int xpos, ypos;
+   unsigned int w, h, bw, depth;
+   stat = XGetGeometry(dpy, d, &root, &xpos, &ypos, &w, &h, &bw, &depth);
+   *width = w;
+   *height = h;
+   return stat;
+}
+
+
+/**
+ * Produce a set of EGL configs.
+ * Note that we get the list of GLcontextModes from the GLX library.
+ * This dependency on GLX lib will be removed someday.
+ */
+static void
+create_configs(_EGLDisplay *disp, __GLXdisplayPrivate *glx_priv)
+{
+   static const EGLint all_apis = (EGL_OPENGL_ES_BIT |
+                                   EGL_OPENGL_ES2_BIT |
+                                   EGL_OPENVG_BIT |
+                                   EGL_OPENGL_BIT);
+   __GLXscreenConfigs *scrn = glx_priv->screenConfigs;
+   const __GLcontextModes *m;
+   int id = 1;
+
+   for (m = scrn->configs; m; m = m->next) {
+      /* EGL requires double-buffered configs */
+      if (m->doubleBufferMode) {
+         struct xdri_egl_config *config = CALLOC_STRUCT(xdri_egl_config);
+
+         _eglInitConfig(&config->Base, id++);
+
+         SET_CONFIG_ATTRIB(&config->Base, EGL_BUFFER_SIZE, m->rgbBits);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_RED_SIZE, m->redBits);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_GREEN_SIZE, m->greenBits);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_BLUE_SIZE, m->blueBits);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_ALPHA_SIZE, m->alphaBits);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_DEPTH_SIZE, m->depthBits);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_STENCIL_SIZE, m->stencilBits);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_SAMPLES, m->samples);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_SAMPLE_BUFFERS, m->sampleBuffers);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_ID, m->visualID);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_TYPE, m->visualType);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_CONFORMANT, all_apis);
+         SET_CONFIG_ATTRIB(&config->Base, EGL_RENDERABLE_TYPE, all_apis);
+         /* XXX only window rendering allowed ATM */
+         SET_CONFIG_ATTRIB(&config->Base, EGL_SURFACE_TYPE,
+                           (EGL_WINDOW_BIT | EGL_PBUFFER_BIT));
+
+         /* XXX possibly other things to init... */
+
+         /* Ptr from EGL config to GLcontextMode.  Used in CreateContext(). */
+         config->mode = m;
+
+         _eglAddConfig(disp, &config->Base);
+      }
+   }
+}
+
+
+/**
+ * Called via __DRIinterfaceMethods object
+ */
+static __DRIfuncPtr
+dri_get_proc_address(const char * proc_name)
+{
+   return NULL;
+}
+
+
+static void
+dri_context_modes_destroy(__GLcontextModes *modes)
+{
+   _eglLog(_EGL_DEBUG, "%s", __FUNCTION__);
+
+   while (modes) {
+      __GLcontextModes * const next = modes->next;
+      free(modes);
+      modes = next;
+   }
+}
+
+
+/**
+ * Create a linked list of 'count' GLcontextModes.
+ * These are used during the client/server visual negotiation phase,
+ * then discarded.
+ */
+static __GLcontextModes *
+dri_context_modes_create(unsigned count, size_t minimum_size)
+{
+   /* This code copied from libGLX, and modified */
+   const size_t size = (minimum_size > sizeof(__GLcontextModes))
+       ? minimum_size : sizeof(__GLcontextModes);
+   __GLcontextModes * head = NULL;
+   __GLcontextModes ** next;
+   unsigned   i;
+
+   next = & head;
+   for (i = 0 ; i < count ; i++) {
+      *next = (__GLcontextModes *) calloc(1, size);
+      if (*next == NULL) {
+	 dri_context_modes_destroy(head);
+	 head = NULL;
+	 break;
+      }
+      
+      (*next)->doubleBufferMode = 1;
+      (*next)->visualID = GLX_DONT_CARE;
+      (*next)->visualType = GLX_DONT_CARE;
+      (*next)->visualRating = GLX_NONE;
+      (*next)->transparentPixel = GLX_NONE;
+      (*next)->transparentRed = GLX_DONT_CARE;
+      (*next)->transparentGreen = GLX_DONT_CARE;
+      (*next)->transparentBlue = GLX_DONT_CARE;
+      (*next)->transparentAlpha = GLX_DONT_CARE;
+      (*next)->transparentIndex = GLX_DONT_CARE;
+      (*next)->xRenderable = GLX_DONT_CARE;
+      (*next)->fbconfigID = GLX_DONT_CARE;
+      (*next)->swapMethod = GLX_SWAP_UNDEFINED_OML;
+      (*next)->bindToTextureRgb = GLX_DONT_CARE;
+      (*next)->bindToTextureRgba = GLX_DONT_CARE;
+      (*next)->bindToMipmapTexture = GLX_DONT_CARE;
+      (*next)->bindToTextureTargets = 0;
+      (*next)->yInverted = GLX_DONT_CARE;
+
+      next = & ((*next)->next);
+   }
+
+   return head;
+}
+
+
+static __DRIscreen *
+dri_find_dri_screen(__DRInativeDisplay *ndpy, int scrn)
+{
+   __GLXdisplayPrivate *priv = __glXInitialize(ndpy);
+   __GLXscreenConfigs *scrnConf = priv->screenConfigs;
+   return &scrnConf->driScreen;
+}
+
+
+static GLboolean
+dri_window_exists(__DRInativeDisplay *ndpy, __DRIid draw)
+{
+   return EGL_TRUE;
+}
+
+
+static GLboolean
+dri_create_context(__DRInativeDisplay *ndpy, int screenNum, int configID,
+                   void * contextID, drm_context_t * hw_context)
+{
+   assert(configID >= 0);
+   return XF86DRICreateContextWithConfig(ndpy, screenNum,
+                                         configID, contextID, hw_context);
+}
+
+
+static GLboolean
+dri_destroy_context(__DRInativeDisplay * ndpy, int screen, __DRIid context)
+{
+   return XF86DRIDestroyContext(ndpy, screen, context);
+}
+
+
+static GLboolean
+dri_create_drawable(__DRInativeDisplay * ndpy, int screen,
+                    __DRIid drawable, drm_drawable_t * hHWDrawable)
+{
+   _eglLog(_EGL_DEBUG, "XDRI: %s", __FUNCTION__);
+
+   /* Create DRI drawable for given window ID (drawable) */
+   if (!XF86DRICreateDrawable(ndpy, screen, drawable, hHWDrawable))
+      return EGL_FALSE;
+
+   return EGL_TRUE;
+}
+
+
+static GLboolean
+dri_destroy_drawable(__DRInativeDisplay * ndpy, int screen, __DRIid drawable)
+{
+   _eglLog(_EGL_DEBUG, "XDRI: %s", __FUNCTION__);
+   return XF86DRIDestroyDrawable(ndpy, screen, drawable);
+}
+
+
+static GLboolean
+dri_get_drawable_info(__DRInativeDisplay *ndpy, int scrn,
+                      __DRIid draw, unsigned int * index, unsigned int * stamp,
+                      int * x, int * y, int * width, int * height,
+                      int * numClipRects, drm_clip_rect_t ** pClipRects,
+                      int * backX, int * backY,
+                      int * numBackClipRects,
+                      drm_clip_rect_t ** pBackClipRects)
+{
+   _eglLog(_EGL_DEBUG, "XDRI: %s", __FUNCTION__);
+
+   if (!XF86DRIGetDrawableInfo(ndpy, scrn, draw, index, stamp,
+                               x, y, width, height,
+                               numClipRects, pClipRects,
+                               backX, backY,
+                               numBackClipRects, pBackClipRects)) {
+      return EGL_FALSE;
+   }
+
+   return EGL_TRUE;
+}
+
+
+/**
+ * Table of functions exported by the loader to the driver.
+ */
+static const __DRIinterfaceMethods interface_methods = {
+   dri_get_proc_address,
+
+   dri_context_modes_create,
+   dri_context_modes_destroy,
+
+   dri_find_dri_screen,
+   dri_window_exists,
+
+   dri_create_context,
+   dri_destroy_context,
+
+   dri_create_drawable,
+   dri_destroy_drawable,
+   dri_get_drawable_info,
+
+   NULL,/*__eglGetUST,*/
+   NULL,/*__eglGetMSCRate,*/
+};
+
+
+
+static EGLBoolean
+init_drm(struct xdri_egl_driver *xdri_drv, _EGLDisplay *disp)
+{
+   __DRIversion ddx_version;
+   __DRIversion dri_version;
+   __DRIversion drm_version;
+   drmVersionPtr version;
+   drm_handle_t  hFB;
+   int newlyopened;
+   int status;
+   int scrn = DefaultScreen(disp->Xdpy);
+
+#if 0
+   createNewScreen = (PFNCREATENEWSCREENFUNC)
+      dlsym(xdri_drv->dri_driver_handle, createNewScreenName);
+   if (!createNewScreen) {
+      _eglLog(_EGL_WARNING, "XDRI: Couldn't find %s function in the driver.",
+              createNewScreenName);
+      return EGL_FALSE;
+   }
+   else {
+      _eglLog(_EGL_DEBUG, "XDRI: Found %s", createNewScreenName);
+   }
+#endif
+
+   /*
+    * Get the DRI X extension version.
+    */
+   dri_version.major = 4;
+   dri_version.minor = 0;
+   dri_version.patch = 0;
+
+   if (!XF86DRIOpenConnection(disp->Xdpy, scrn,
+                              &xdri_drv->hSAREA, &xdri_drv->busID)) {
+      _eglLog(_EGL_WARNING, "XF86DRIOpenConnection failed");
+   }
+
+   xdri_drv->drmFD = drmOpenOnce(NULL, xdri_drv->busID, &newlyopened);
+   if (xdri_drv->drmFD < 0) {
+      perror("drmOpenOnce failed: ");
+      return EGL_FALSE;
+   }
+   else {
+      _eglLog(_EGL_DEBUG, "XDRI: drmOpenOnce returned %d", xdri_drv->drmFD);
+   }
+
+
+   if (drmGetMagic(xdri_drv->drmFD, &xdri_drv->magic)) {
+      perror("drmGetMagic failed: ");
+      return EGL_FALSE;
+   }
+
+   version = drmGetVersion(xdri_drv->drmFD);
+   if (version) {
+      drm_version.major = version->version_major;
+      drm_version.minor = version->version_minor;
+      drm_version.patch = version->version_patchlevel;
+      drmFreeVersion(version);
+      _eglLog(_EGL_DEBUG, "XDRI: Got DRM version %d.%d.%d",
+              drm_version.major,
+              drm_version.minor,
+              drm_version.patch);
+   }
+   else {
+      drm_version.major = -1;
+      drm_version.minor = -1;
+      drm_version.patch = -1;
+      _eglLog(_EGL_WARNING, "XDRI: drmGetVersion() failed");
+      return EGL_FALSE;
+   }
+
+   /* Authenticate w/ server.
+    */
+   if (!XF86DRIAuthConnection(disp->Xdpy, scrn, xdri_drv->magic)) {
+      _eglLog(_EGL_WARNING, "XDRI: XF86DRIAuthConnection() failed");
+      return EGL_FALSE;
+   }
+   else {
+      _eglLog(_EGL_DEBUG, "XDRI: XF86DRIAuthConnection() success");
+   }
+
+   /* Get ddx version.
+    */
+   {
+      char *driverName;
+
+      /*
+       * Get device name (like "tdfx") and the ddx version
+       * numbers.  We'll check the version in each DRI driver's
+       * "createNewScreen" function.
+       */
+      if (!XF86DRIGetClientDriverName(disp->Xdpy, scrn,
+                                     &ddx_version.major,
+                                     &ddx_version.minor,
+                                     &ddx_version.patch,
+                                     &driverName)) {
+         _eglLog(_EGL_WARNING, "XDRI: XF86DRIGetClientDriverName failed");
+         return EGL_FALSE;
+      }
+      else {
+         _eglLog(_EGL_DEBUG, "XDRI: XF86DRIGetClientDriverName returned %s", driverName);
+      }
+   }
+
+   /* Get framebuffer info.
+    */
+   {
+      int junk;
+      if (!XF86DRIGetDeviceInfo(disp->Xdpy, scrn,
+                                &hFB,
+                                &junk,
+                                &xdri_drv->framebuffer.size,
+                                &xdri_drv->framebuffer.stride,
+                                &xdri_drv->framebuffer.dev_priv_size,
+                                &xdri_drv->framebuffer.dev_priv)) {
+         _eglLog(_EGL_WARNING, "XDRI: XF86DRIGetDeviceInfo() failed");
+         return EGL_FALSE;
+      }
+      else {
+         _eglLog(_EGL_DEBUG, "XDRI: XF86DRIGetDeviceInfo() success");
+      }
+      xdri_drv->framebuffer.width = DisplayWidth(disp->Xdpy, scrn);
+      xdri_drv->framebuffer.height = DisplayHeight(disp->Xdpy, scrn);
+   }
+
+   /* Map the framebuffer region. (this may not be needed)
+    */
+   status = drmMap(xdri_drv->drmFD, hFB, xdri_drv->framebuffer.size, 
+                   (drmAddressPtr) &xdri_drv->framebuffer.base);
+   if (status != 0) {
+      _eglLog(_EGL_WARNING, "XDRI: drmMap(framebuffer) failed");
+      return EGL_FALSE;
+   }
+   else {
+      _eglLog(_EGL_DEBUG, "XDRI: drmMap(framebuffer) success");
+   }
+
+   /* Map the SAREA region.
+    */
+   status = drmMap(xdri_drv->drmFD, xdri_drv->hSAREA, SAREA_MAX, &xdri_drv->pSAREA);
+   if (status != 0) {
+      _eglLog(_EGL_WARNING, "XDRI: drmMap(sarea) failed");
+      return EGL_FALSE;
+   }
+   else {
+      _eglLog(_EGL_DEBUG, "XDRI: drmMap(sarea) success");
+   }
+
+   return EGL_TRUE;
+}
+
+
+/**
+ * Load the DRI driver named by "xdri_drv->dri_driver_name".
+ * Basically, dlopen() the library to set "xdri_drv->dri_driver_handle".
+ *
+ * Later, we'll call dlsym(createNewScreenName) to get a pointer to
+ * the driver's createNewScreen() function which is the bootstrap function.
+ *
+ * \return EGL_TRUE for success, EGL_FALSE for failure
+ */
+static EGLBoolean
+load_dri_driver(struct xdri_egl_driver *xdri_drv)
+{
+   char filename[100];
+   int flags = RTLD_NOW;
+
+   /* try "egl_xxx_dri.so" first */
+   snprintf(filename, sizeof(filename), "egl_%s.so", xdri_drv->dri_driver_name);
+   _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s)", filename);
+   xdri_drv->dri_driver_handle = dlopen(filename, flags);
+   if (xdri_drv->dri_driver_handle) {
+      _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s) OK", filename);
+      return EGL_TRUE;
+   }
+   else {
+      _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s) fail (%s)", filename, dlerror());
+   }
+
+   /* try regular "xxx_dri.so" next */
+   snprintf(filename, sizeof(filename), "%s.so", xdri_drv->dri_driver_name);
+   _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s)", filename);
+   xdri_drv->dri_driver_handle = dlopen(filename, flags);
+   if (xdri_drv->dri_driver_handle) {
+      _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s) OK", filename);
+      return EGL_TRUE;
+   }
+
+   _eglLog(_EGL_WARNING, "XDRI Could not open %s (%s)", filename, dlerror());
+   return EGL_FALSE;
+}
+
+
+/**
+ * Called via eglInitialize(), xdri_drv->API.Initialize().
+ */
+static EGLBoolean
+xdri_eglInitialize(_EGLDriver *drv, EGLDisplay dpy,
+                   EGLint *minor, EGLint *major)
+{
+   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   static char name[100];
+
+   _eglLog(_EGL_DEBUG, "XDRI: eglInitialize");
+
+   if (!disp->Xdpy) {
+      disp->Xdpy = XOpenDisplay(NULL);
+      if (!disp->Xdpy) {
+         _eglLog(_EGL_WARNING, "XDRI: XOpenDisplay failed");
+         return EGL_FALSE;
+      }
+   }
+
+#if 0
+   /* choose the DRI driver to load */
+   xdri_drv->dri_driver_name = _eglChooseDRMDriver(0);
+   if (!load_dri_driver(xdri_drv))
+      return EGL_FALSE;
+#else
+   (void) load_dri_driver;
+#endif
+
+#if 0
+   if (!init_drm(xdri_drv, disp))
+      return EGL_FALSE;
+#else
+   (void) init_drm;
+#endif
+
+   /*
+    * NOTE: this call to __glXInitialize() bootstraps the whole GLX/DRI
+    * interface, loads the DRI driver, etc.
+    * This replaces the load_dri_driver()  and init_drm() code above.
+    */
+   xdri_drv->glx_priv = __glXInitialize(disp->Xdpy);
+
+   create_configs(disp, xdri_drv->glx_priv);
+
+   xdri_drv->Base.Initialized = EGL_TRUE;
+
+   if (xdri_drv->dri_driver_name)
+      snprintf(name, sizeof(name), "X/DRI:%s", xdri_drv->dri_driver_name);
+   else
+      snprintf(name, sizeof(name), "X/DRI");
+   xdri_drv->Base.Name = name;
+
+   /* we're supporting EGL 1.4 */
+   *minor = 1;
+   *major = 4;
+
+   return EGL_TRUE;
+}
+
+
+/*
+ * Do some clean-up that normally occurs in XCloseDisplay().
+ * We do this here because we're about to unload a dynamic library
+ * that has added some per-display extension data and callbacks.
+ * If we don't do this here we'll crash in XCloseDisplay() because it'll
+ * try to call functions that went away when the driver library was unloaded.
+ */
+static void
+FreeDisplayExt(Display *dpy)
+{
+   _XExtension *ext, *next;
+
+   for (ext = dpy->ext_procs; ext; ext = next) {
+      next = ext->next;
+      if (ext->close_display) {
+         ext->close_display(dpy, &ext->codes);
+         ext->close_display = NULL;
+      }
+      if (ext->name)
+         Xfree(ext->name);
+      Xfree(ext);
+   }
+   dpy->ext_procs = NULL;
+
+   _XFreeExtData (dpy->ext_data);
+   dpy->ext_data = NULL;
+}
+
+
+/**
+ * Called via eglTerminate(), drv->API.Terminate().
+ */
+static EGLBoolean
+xdri_eglTerminate(_EGLDriver *drv, EGLDisplay dpy)
+{
+   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+
+   _eglLog(_EGL_DEBUG, "XDRI: eglTerminate");
+
+   _eglLog(_EGL_DEBUG, "XDRI: Closing %s", xdri_drv->dri_driver_name);
+
+   FreeDisplayExt(disp->Xdpy);
+
+#if 0
+   /* this causes a segfault for some reason */
+   dlclose(xdri_drv->dri_driver_handle);
+#endif
+   xdri_drv->dri_driver_handle = NULL;
+
+   free((void*) xdri_drv->dri_driver_name);
+
+   return EGL_TRUE;
+}
+
+
+/*
+ * Called from eglGetProcAddress() via drv->API.GetProcAddress().
+ */
+static _EGLProc
+xdri_eglGetProcAddress(const char *procname)
+{
+#if 0
+   _EGLDriver *drv = NULL;
+
+   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
+   /*_EGLDisplay *disp = _eglLookupDisplay(dpy);*/
+   _EGLProc *proc = xdri_drv->driScreen.getProcAddress(procname);
+   return proc;
+#elif 1
+   /* This is a bit of a hack to get at the gallium/Mesa state tracker
+    * function st_get_proc_address().  This will probably change at
+    * some point.
+    */
+   _EGLProc (*st_get_proc_addr)(const char *procname);
+   st_get_proc_addr = dlsym(NULL, "st_get_proc_address");
+   if (st_get_proc_addr) {
+      return st_get_proc_addr(procname);
+   }
+   return NULL;   
+#else
+   return NULL;
+#endif
+}
+
+
+/**
+ * Called via eglCreateContext(), drv->API.CreateContext().
+ */
+static EGLContext
+xdri_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+                      EGLContext share_list, const EGLint *attrib_list)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct xdri_egl_config *xdri_config = lookup_config(drv, dpy, config);
+   void *shared = NULL;
+   int renderType = GLX_RGBA_BIT;
+
+   struct xdri_egl_context *xdri_ctx = CALLOC_STRUCT(xdri_egl_context);
+   if (!xdri_ctx)
+      return EGL_NO_CONTEXT;
+
+   if (!_eglInitContext(drv, dpy, &xdri_ctx->Base, config, attrib_list)) {
+      free(xdri_ctx);
+      return EGL_NO_CONTEXT;
+   }
+
+   assert(xdri_config);
+
+   {
+      struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
+      __GLXscreenConfigs *scrnConf = xdri_drv->glx_priv->screenConfigs;
+      xdri_ctx->driContext.private = 
+         scrnConf->driScreen.createNewContext(disp->Xdpy,
+                                              xdri_config->mode, renderType,
+                                              shared, &xdri_ctx->driContext);
+   }
+
+   if (!xdri_ctx->driContext.private) {
+      _eglLog(_EGL_DEBUG, "driScreen.createNewContext failed");
+      free(xdri_ctx);
+      return EGL_NO_CONTEXT;
+   }
+
+   xdri_ctx->driContext.mode = xdri_config->mode;
+
+   return _eglGetContextHandle(&xdri_ctx->Base);
+}
+
+
+/**
+ * Called via eglMakeCurrent(), drv->API.MakeCurrent().
+ */
+static EGLBoolean
+xdri_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
+                    EGLSurface r, EGLContext context)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct xdri_egl_context *xdri_ctx = lookup_context(context);
+   struct xdri_egl_surface *xdri_draw = lookup_surface(d);
+   struct xdri_egl_surface *xdri_read = lookup_surface(r);
+   __DRIid draw = xdri_draw ? xdri_draw->driDrawable : 0;
+   __DRIid read = xdri_read ? xdri_read->driDrawable : 0;
+   int scrn = DefaultScreen(disp->Xdpy);
+
+   if (!_eglMakeCurrent(drv, dpy, d, r, context))
+      return EGL_FALSE;
+
+
+   if (xdri_ctx &&
+       !xdri_ctx->driContext.bindContext(disp->Xdpy, scrn, draw, read,
+                                         &xdri_ctx->driContext)) {
+      return EGL_FALSE;
+   }
+
+   return EGL_TRUE;
+}
+
+
+/**
+ * Called via eglCreateWindowSurface(), drv->API.CreateWindowSurface().
+ */
+static EGLSurface
+xdri_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+                            NativeWindowType window, const EGLint *attrib_list)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct xdri_egl_surface *xdri_surf;
+   int scrn = DefaultScreen(disp->Xdpy);
+   uint width, height;
+
+   xdri_surf = CALLOC_STRUCT(xdri_egl_surface);
+   if (!xdri_surf)
+      return EGL_NO_SURFACE;
+
+   if (!_eglInitSurface(drv, dpy, &xdri_surf->Base, EGL_WINDOW_BIT,
+                        config, attrib_list)) {
+      free(xdri_surf);
+      return EGL_FALSE;
+   }
+
+   if (!XF86DRICreateDrawable(disp->Xdpy, scrn, window, &xdri_surf->hDrawable)) {
+      free(xdri_surf);
+      return EGL_FALSE;
+   }
+
+   xdri_surf->driDrawable = window;
+
+   _eglSaveSurface(&xdri_surf->Base);
+
+   get_drawable_size(disp->Xdpy, window, &width, &height);
+   xdri_surf->Base.Width = width;
+   xdri_surf->Base.Height = height;
+
+   _eglLog(_EGL_DEBUG,
+           "XDRI: CreateWindowSurface win 0x%x  handle %d  hDrawable %d",
+           (int) window, _eglGetSurfaceHandle(&xdri_surf->Base),
+           (int) xdri_surf->hDrawable);
+
+   return _eglGetSurfaceHandle(&xdri_surf->Base);
+}
+
+
+/**
+ * Called via eglCreatePbufferSurface(), drv->API.CreatePbufferSurface().
+ */
+static EGLSurface
+xdri_eglCreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+                             const EGLint *attrib_list)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   struct xdri_egl_surface *xdri_surf;
+   struct xdri_egl_config *xdri_config = lookup_config(drv, dpy, config);
+   int scrn = DefaultScreen(disp->Xdpy);
+   Window window;
+
+   xdri_surf = CALLOC_STRUCT(xdri_egl_surface);
+   if (!xdri_surf)
+      return EGL_NO_SURFACE;
+
+   if (!_eglInitSurface(drv, dpy, &xdri_surf->Base, EGL_PBUFFER_BIT,
+                        config, attrib_list)) {
+      free(xdri_surf);
+      return EGL_FALSE;
+   }
+
+   /* Create a dummy X window */
+   {
+      Window root = RootWindow(disp->Xdpy, scrn);
+      XSetWindowAttributes attr;
+      XVisualInfo *visInfo, visTemplate;
+      unsigned mask;
+      int nvis;
+
+      visTemplate.visualid = xdri_config->mode->visualID;
+      visInfo = XGetVisualInfo(disp->Xdpy, VisualIDMask, &visTemplate, &nvis);
+      if (!visInfo) {
+         return EGL_NO_SURFACE;
+      }
+
+      attr.background_pixel = 0;
+      attr.border_pixel = 0;
+      attr.colormap = XCreateColormap(disp->Xdpy, root,
+                                      visInfo->visual, AllocNone);
+      attr.event_mask = StructureNotifyMask | ExposureMask | KeyPressMask;
+      mask = CWBackPixel | CWBorderPixel | CWColormap | CWEventMask;
+      
+      window = XCreateWindow(disp->Xdpy, root, 0, 0,
+                             xdri_surf->Base.Width, xdri_surf->Base.Height,
+                             0, visInfo->depth, InputOutput,
+                             visInfo->visual, mask, &attr);
+
+      /*XMapWindow(disp->Xdpy, window);*/
+      XFree(visInfo);
+
+      /* set hints and properties */
+      /*
+      sizehints.width = xdri_surf->Base.Width;
+      sizehints.height = xdri_surf->Base.Height;
+      sizehints.flags = USPosition;
+      XSetNormalHints(disp->Xdpy, window, &sizehints);
+      */
+   }
+
+   if (!XF86DRICreateDrawable(disp->Xdpy, scrn, window, &xdri_surf->hDrawable)) {
+      free(xdri_surf);
+      return EGL_FALSE;
+   }
+
+   xdri_surf->driDrawable = window;
+
+   _eglSaveSurface(&xdri_surf->Base);
+
+   _eglLog(_EGL_DEBUG,
+           "XDRI: CreatePbufferSurface handle %d  hDrawable %d",
+           _eglGetSurfaceHandle(&xdri_surf->Base),
+           (int) xdri_surf->hDrawable);
+
+   return _eglGetSurfaceHandle(&xdri_surf->Base);
+}
+
+
+
+static EGLBoolean
+xdri_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
+{
+   struct xdri_egl_surface *xdri_surf = lookup_surface(surface);
+   if (xdri_surf) {
+      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
+      if (xdri_surf->Base.IsBound) {
+         xdri_surf->Base.DeletePending = EGL_TRUE;
+      }
+      else {
+         /*
+         st_unreference_framebuffer(surf->Framebuffer);
+         */
+         free(xdri_surf);
+      }
+      return EGL_TRUE;
+   }
+   else {
+      _eglError(EGL_BAD_SURFACE, "eglDestroySurface");
+      return EGL_FALSE;
+   }
+}
+
+
+static EGLBoolean
+xdri_eglBindTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface,
+                     EGLint buffer)
+{
+   typedef int (*bind_teximage)(__DRInativeDisplay *dpy,
+                                __DRIid surface, __DRIscreen *psc,
+                                int buffer, int target, int format,
+                                int level, int mipmap);
+
+   bind_teximage egl_dri_bind_teximage;
+
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+
+   struct xdri_egl_context *xdri_ctx = current_context();
+   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
+   struct xdri_egl_surface *xdri_surf = lookup_surface(surface);
+
+   __DRIid dri_surf = xdri_surf ? xdri_surf->driDrawable : 0;
+
+   __GLXscreenConfigs *scrnConf = xdri_drv->glx_priv->screenConfigs;
+   __DRIscreen *psc = &scrnConf->driScreen;
+
+   /* this call just does error checking */
+   if (!_eglBindTexImage(drv, dpy, surface, buffer)) {
+      return EGL_FALSE;
+   }
+
+   egl_dri_bind_teximage =
+      (bind_teximage) dlsym(NULL, "egl_dri_bind_teximage");
+   if (egl_dri_bind_teximage) {
+      return egl_dri_bind_teximage(disp->Xdpy, dri_surf, psc,
+                                   buffer,
+                                   xdri_surf->Base.TextureTarget,
+                                   xdri_surf->Base.TextureFormat,
+                                   xdri_surf->Base.MipmapLevel,
+                                   xdri_surf->Base.MipmapTexture);
+   }
+   else {
+      /* fallback path based on glCopyTexImage() */
+      /* Get/save currently bound 2D texobj name */
+      glGetIntegerv_t glGetIntegerv_func = 
+         (glGetIntegerv_t) dlsym(NULL, "glGetIntegerv");
+      GLint curTexObj = 0;
+      if (glGetIntegerv_func) {
+         (*glGetIntegerv_func)(GL_TEXTURE_BINDING_2D, &curTexObj);
+      }
+      xdri_ctx->bound_tex_object = curTexObj;
+   }
+
+   return EGL_FALSE;
+}
+
+
+static EGLBoolean
+xdri_eglReleaseTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface,
+                        EGLint buffer)
+{
+   typedef int (*release_teximage)(__DRInativeDisplay *dpy,
+                                   __DRIid surface, __DRIscreen *psc,
+                                   int buffer, int target, int format,
+                                   int level, int mipmap);
+   release_teximage egl_dri_release_teximage;
+
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+
+   struct xdri_egl_context *xdri_ctx = current_context();
+   struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
+   struct xdri_egl_surface *xdri_surf = lookup_surface(surface);
+
+   __DRIid dri_surf = xdri_surf ? xdri_surf->driDrawable : 0;
+
+   __GLXscreenConfigs *scrnConf = xdri_drv->glx_priv->screenConfigs;
+   __DRIscreen *psc = &scrnConf->driScreen;
+
+   /* this call just does error checking */
+   if (!_eglReleaseTexImage(drv, dpy, surface, buffer)) {
+      return EGL_FALSE;
+   }
+
+   egl_dri_release_teximage =
+      (release_teximage) dlsym(NULL, "egl_dri_release_teximage");
+   if (egl_dri_release_teximage) {
+      return egl_dri_release_teximage(disp->Xdpy, dri_surf, psc,
+                                      buffer,
+                                      xdri_surf->Base.TextureTarget,
+                                      xdri_surf->Base.TextureFormat,
+                                      xdri_surf->Base.MipmapLevel,
+                                      xdri_surf->Base.MipmapTexture);
+   }
+   else {
+      /* fallback path based on glCopyTexImage() */
+      glGetIntegerv_t glGetIntegerv_func = 
+         (glGetIntegerv_t) dlsym(NULL, "glGetIntegerv");
+      glBindTexture_t glBindTexture_func = 
+         (glBindTexture_t) dlsym(NULL, "glBindTexture");
+      glCopyTexImage2D_t glCopyTexImage2D_func = 
+         (glCopyTexImage2D_t) dlsym(NULL, "glCopyTexImage2D");
+      GLint curTexObj;
+      GLenum intFormat;
+      GLint level, width, height;
+
+      if (xdri_surf->Base.TextureFormat == EGL_TEXTURE_RGBA)
+         intFormat = GL_RGBA;
+      else
+         intFormat = GL_RGB;
+      level = xdri_surf->Base.MipmapLevel;
+      width = xdri_surf->Base.Width >> level;
+      height = xdri_surf->Base.Height >> level;
+
+      if (width > 0 && height > 0 &&
+          glGetIntegerv_func && glBindTexture_func && glCopyTexImage2D_func) {
+         glGetIntegerv_func(GL_TEXTURE_BINDING_2D, &curTexObj);
+         /* restore texobj from time of eglBindTexImage() call */
+         if (curTexObj != xdri_ctx->bound_tex_object)
+            glBindTexture_func(GL_TEXTURE_2D, xdri_ctx->bound_tex_object);
+         /* copy pbuffer image to texture */
+         glCopyTexImage2D_func(GL_TEXTURE_2D,
+                               level,
+                               intFormat,
+                               0, 0, width, height, 0);
+         /* restore current texture */
+         if (curTexObj != xdri_ctx->bound_tex_object)
+            glBindTexture_func(GL_TEXTURE_2D, curTexObj);
+      }
+      xdri_ctx->bound_tex_object = -1;
+   }
+
+   return EGL_FALSE;
+}
+
+
+static EGLBoolean
+xdri_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
+{
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+
+   _eglLog(_EGL_DEBUG, "XDRI: EGL SwapBuffers");
+
+   /* error checking step: */
+   if (!_eglSwapBuffers(drv, dpy, draw))
+      return EGL_FALSE;
+
+   {
+      struct xdri_egl_surface *xdri_surf = lookup_surface(draw);
+      struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv);
+      __GLXscreenConfigs *scrnConf = xdri_drv->glx_priv->screenConfigs;
+      __DRIscreen *psc = &scrnConf->driScreen;
+      __DRIdrawable * const pdraw = psc->getDrawable(disp->Xdpy,
+                                                     xdri_surf->driDrawable,
+                                                     psc->private);
+
+      if (pdraw)
+         pdraw->swapBuffers(disp->Xdpy, pdraw->private);
+      else
+         _eglLog(_EGL_WARNING, "pdraw is null in SwapBuffers");
+   }
+
+   return EGL_TRUE;
+}
+
+
+/**
+ * This is the main entrypoint into the driver, called by libEGL.
+ * Create a new _EGLDriver object and init its dispatch table.
+ */
+_EGLDriver *
+_eglMain(_EGLDisplay *disp, const char *args)
+{
+   struct xdri_egl_driver *xdri_drv = CALLOC_STRUCT(xdri_egl_driver);
+   if (!xdri_drv)
+      return NULL;
+
+   /* Tell libGL to prefer the EGL drivers over regular DRI drivers */
+   __glXPreferEGL(1);
+
+   _eglInitDriverFallbacks(&xdri_drv->Base);
+   xdri_drv->Base.API.Initialize = xdri_eglInitialize;
+   xdri_drv->Base.API.Terminate = xdri_eglTerminate;
+
+   xdri_drv->Base.API.GetProcAddress = xdri_eglGetProcAddress;
+
+   xdri_drv->Base.API.CreateContext = xdri_eglCreateContext;
+   xdri_drv->Base.API.MakeCurrent = xdri_eglMakeCurrent;
+   xdri_drv->Base.API.CreateWindowSurface = xdri_eglCreateWindowSurface;
+   xdri_drv->Base.API.CreatePbufferSurface = xdri_eglCreatePbufferSurface;
+   xdri_drv->Base.API.DestroySurface = xdri_eglDestroySurface;
+   xdri_drv->Base.API.BindTexImage = xdri_eglBindTexImage;
+   xdri_drv->Base.API.ReleaseTexImage = xdri_eglReleaseTexImage;
+   xdri_drv->Base.API.SwapBuffers = xdri_eglSwapBuffers;
+
+   xdri_drv->Base.ClientAPIsMask = (EGL_OPENGL_BIT |
+                                    EGL_OPENGL_ES_BIT |
+                                    EGL_OPENGL_ES2_BIT |
+                                    EGL_OPENVG_BIT);
+   xdri_drv->Base.Name = "X/DRI";
+
+   _eglLog(_EGL_DEBUG, "XDRI: main(%s)", args);
+
+   return &xdri_drv->Base;
+}
diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile
index 7ab78197c7..89b4a200bc 100644
--- a/src/egl/main/Makefile
+++ b/src/egl/main/Makefile
@@ -8,34 +8,49 @@ INCLUDE_DIRS = -I$(TOP)/include -I$(TOP)/src/mesa/glapi
 
 HEADERS = \
 	eglconfig.h \
+	eglconfigutil.h \
 	eglcontext.h \
+	egldefines.h \
 	egldisplay.h \
 	egldriver.h \
 	eglglobals.h \
 	egllog.h \
 	eglhash.h \
+	eglmisc.h \
 	eglmode.h \
 	eglscreen.h \
-	eglsurface.h
+	eglstring.h \
+	eglsurface.h \
+	eglx.h
 
 SOURCES = \
 	eglapi.c \
 	eglconfig.c \
+	eglconfigutil.c \
 	eglcontext.c \
 	egldisplay.c \
 	egldriver.c \
 	eglglobals.c \
 	egllog.c \
 	eglhash.c \
+	eglmisc.c \
 	eglmode.c \
 	eglscreen.c \
-	eglsurface.c
+	eglstring.c \
+	eglsurface.c \
+	eglx.c
 
 OBJECTS = $(SOURCES:.c=.o)
 
 
+# Undefined for now
+LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1
+
+LIBS = -lX11
+
+
 .c.o:
-	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(LOCAL_CFLAGS) $< -o $@
 
 
 
@@ -48,9 +63,11 @@ library: $(TOP)/$(LIB_DIR)/libEGL.so
 $(TOP)/$(LIB_DIR)/libEGL.so: $(OBJECTS)
 	$(MKLIB) -o EGL -linker '$(CC)' -ldflags '$(LDFLAGS)' \
 		-major 1 -minor 0 \
-		-install $(TOP)/$(LIB_DIR) -ldl $(OBJECTS)
-
+		-install $(TOP)/$(LIB_DIR) -ldl $(OBJECTS) $(LIBS)
 
+install:
+	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(INSTALL) $(TOP)/$(LIB_DIR)/libEGL.so* $(DESTDIR)$(INSTALL_LIB_DIR)
 
 clean:
 	-rm -f *.o *.so*
@@ -59,9 +76,11 @@ clean:
 
 depend: $(SOURCES) $(HEADERS)
 	@ echo "running $(MKDEP)"
+	@ rm -f depend
 	@ touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) \
-		$(SOURCES) $(HEADERS) > /dev/null 
+		$(SOURCES) $(HEADERS) > /dev/null 2>/dev/null
+
 
 include depend
 # DO NOT DELETE
diff --git a/src/egl/main/README.txt b/src/egl/main/README.txt
new file mode 100644
index 0000000000..b3d253dd13
--- /dev/null
+++ b/src/egl/main/README.txt
@@ -0,0 +1,71 @@
+
+
+Notes about the EGL library:
+
+
+The EGL code here basically consists of two things:
+
+1. An EGL API dispatcher.  This directly routes all the eglFooBar() API
+   calls into driver-specific functions.
+
+2. Fallbacks for EGL API functions.  A driver _could_ implement all the
+   EGL API calls from scratch.  But in many cases, the fallbacks provided
+   in libEGL (such as eglChooseConfig()) will do the job.
+
+
+
+Bootstrapping:
+
+When the apps calls eglOpenDisplay() a device driver is selected and loaded
+(look for dlsym() or LoadLibrary() in egldriver.c).
+
+The driver's _eglMain() function is then called.  This driver function
+allocates, initializes and returns a new _EGLDriver object (usually a
+subclass of that type).
+
+As part of initialization, the dispatch table in _EGLDriver->API must be
+populated with all the EGL entrypoints.  Typically, _eglInitDriverFallbacks()
+can be used to plug in default/fallback functions.  Some functions like
+driver->API.Initialize and driver->API.Terminate _must_ be implemented
+with driver-specific code (no default/fallback function is possible).
+
+
+A bit later, the app will call eglInitialize().  This will get routed
+to the driver->API.Initialize() function.  Any additional driver
+initialization that wasn't done in _eglMain() should be done at this
+point.  Typically, this will involve setting up visual configs, etc.
+
+
+
+Special Functions:
+
+Certain EGL functions _must_ be implemented by the driver.  This includes:
+
+eglCreateContext
+eglCreateWindowSurface
+eglCreatePixmapSurface
+eglCreatePBufferSurface
+eglMakeCurrent
+eglSwapBuffers
+
+Most of the EGLConfig-related functions can be implemented with the
+defaults/fallbacks.  Same thing for the eglGet/Query functions.
+
+
+
+
+Teardown:
+
+When eglTerminate() is called, the driver->API.Terminate() function is
+called.  The driver should clean up after itself.  eglTerminate() will
+then close/unload the driver (shared library).
+
+
+
+
+Subclassing:
+
+The internal libEGL data structures such as _EGLDisplay, _EGLContext,
+_EGLSurface, etc should be considered base classes from which drivers
+will derive subclasses.
+
diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c
index bfa580e6c3..9df938e188 100644
--- a/src/egl/main/eglapi.c
+++ b/src/egl/main/eglapi.c
@@ -43,36 +43,63 @@
 
 
 /**
- * NOTE: displayName is treated as a string in _eglChooseDriver()!!!
- * This will probably change!
- * See _eglChooseDriver() for details!
+ * This is typically the first EGL function that an application calls.
+ * We initialize our global vars and create a private _EGLDisplay object.
  */
-EGLDisplay APIENTRY
-eglGetDisplay(NativeDisplayType displayName)
+EGLDisplay EGLAPIENTRY
+eglGetDisplay(NativeDisplayType nativeDisplay)
 {
    _EGLDisplay *dpy;
    _eglInitGlobals();
-   dpy = _eglNewDisplay(displayName);
-   if (dpy)
-      return dpy->Handle;
-   else
-      return EGL_NO_DISPLAY;
+   dpy = _eglNewDisplay(nativeDisplay);
+   return _eglGetDisplayHandle(dpy);
 }
 
 
-EGLBoolean APIENTRY
+/**
+ * This is typically the second EGL function that an application calls.
+ * Here we load/initialize the actual hardware driver.
+ */
+EGLBoolean EGLAPIENTRY
 eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor)
 {
+   EGLint major_int, minor_int;
+
    if (dpy) {
-      _EGLDriver *drv = _eglChooseDriver(dpy);
-      if (drv)
-         return drv->API.Initialize(drv, dpy, major, minor);
+      EGLBoolean retVal;
+      _EGLDisplay *dpyPriv = _eglLookupDisplay(dpy);
+      if (!dpyPriv) {
+         return EGL_FALSE;
+      }
+      dpyPriv->Driver = _eglOpenDriver(dpyPriv,
+                                       dpyPriv->DriverName,
+                                       dpyPriv->DriverArgs);
+      if (!dpyPriv->Driver) {
+         return EGL_FALSE;
+      }
+      /* Initialize the particular driver now */
+      retVal = dpyPriv->Driver->API.Initialize(dpyPriv->Driver, dpy,
+                                               &major_int, &minor_int);
+
+      dpyPriv->Driver->APImajor = major_int;
+      dpyPriv->Driver->APIminor = minor_int;
+      snprintf(dpyPriv->Driver->Version, sizeof(dpyPriv->Driver->Version),
+               "%d.%d (%s)", major_int, minor_int, dpyPriv->Driver->Name);
+
+      /* Update applications version of major and minor if not NULL */
+      if((major != NULL) && (minor != NULL))
+      {
+         *major = major_int;
+         *minor = minor_int;
+      }
+
+      return retVal;
    }
    return EGL_FALSE;
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglTerminate(EGLDisplay dpy)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -83,7 +110,7 @@ eglTerminate(EGLDisplay dpy)
 }
 
 
-const char * APIENTRY
+const char * EGLAPIENTRY
 eglQueryString(EGLDisplay dpy, EGLint name)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -94,7 +121,7 @@ eglQueryString(EGLDisplay dpy, EGLint name)
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglGetConfigs(EGLDisplay dpy, EGLConfig *configs, EGLint config_size, EGLint *num_config)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -103,7 +130,7 @@ eglGetConfigs(EGLDisplay dpy, EGLConfig *configs, EGLint config_size, EGLint *nu
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglChooseConfig(EGLDisplay dpy, const EGLint *attrib_list, EGLConfig *configs, EGLint config_size, EGLint *num_config)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -111,7 +138,7 @@ eglChooseConfig(EGLDisplay dpy, const EGLint *attrib_list, EGLConfig *configs, E
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglGetConfigAttrib(EGLDisplay dpy, EGLConfig config, EGLint attribute, EGLint *value)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -119,7 +146,7 @@ eglGetConfigAttrib(EGLDisplay dpy, EGLConfig config, EGLint attribute, EGLint *v
 }
 
 
-EGLContext APIENTRY
+EGLContext EGLAPIENTRY
 eglCreateContext(EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -127,7 +154,7 @@ eglCreateContext(EGLDisplay dpy, EGLConfig config, EGLContext share_list, const
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglDestroyContext(EGLDisplay dpy, EGLContext ctx)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -135,7 +162,7 @@ eglDestroyContext(EGLDisplay dpy, EGLContext ctx)
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglMakeCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -143,7 +170,7 @@ eglMakeCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx)
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglQueryContext(EGLDisplay dpy, EGLContext ctx, EGLint attribute, EGLint *value)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -151,7 +178,7 @@ eglQueryContext(EGLDisplay dpy, EGLContext ctx, EGLint attribute, EGLint *value)
 }
 
 
-EGLSurface APIENTRY
+EGLSurface EGLAPIENTRY
 eglCreateWindowSurface(EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -159,7 +186,7 @@ eglCreateWindowSurface(EGLDisplay dpy, EGLConfig config, NativeWindowType window
 }
 
 
-EGLSurface APIENTRY
+EGLSurface EGLAPIENTRY
 eglCreatePixmapSurface(EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -167,7 +194,7 @@ eglCreatePixmapSurface(EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap
 }
 
 
-EGLSurface APIENTRY
+EGLSurface EGLAPIENTRY
 eglCreatePbufferSurface(EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -175,7 +202,7 @@ eglCreatePbufferSurface(EGLDisplay dpy, EGLConfig config, const EGLint *attrib_l
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglDestroySurface(EGLDisplay dpy, EGLSurface surface)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -183,7 +210,7 @@ eglDestroySurface(EGLDisplay dpy, EGLSurface surface)
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglQuerySurface(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint *value)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -191,7 +218,7 @@ eglQuerySurface(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint *va
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglSurfaceAttrib(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint value)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -199,7 +226,7 @@ eglSurfaceAttrib(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint va
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglBindTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -207,7 +234,7 @@ eglBindTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer)
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglReleaseTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -215,7 +242,7 @@ eglReleaseTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer)
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglSwapInterval(EGLDisplay dpy, EGLint interval)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -223,7 +250,7 @@ eglSwapInterval(EGLDisplay dpy, EGLint interval)
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglSwapBuffers(EGLDisplay dpy, EGLSurface draw)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -231,7 +258,7 @@ eglSwapBuffers(EGLDisplay dpy, EGLSurface draw)
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglCopyBuffers(EGLDisplay dpy, EGLSurface surface, NativePixmapType target)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -239,7 +266,7 @@ eglCopyBuffers(EGLDisplay dpy, EGLSurface surface, NativePixmapType target)
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglWaitGL(void)
 {
    EGLDisplay dpy = eglGetCurrentDisplay();
@@ -252,7 +279,7 @@ eglWaitGL(void)
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglWaitNative(EGLint engine)
 {
    EGLDisplay dpy = eglGetCurrentDisplay();
@@ -265,40 +292,31 @@ eglWaitNative(EGLint engine)
 }
 
 
-EGLDisplay APIENTRY
+EGLDisplay EGLAPIENTRY
 eglGetCurrentDisplay(void)
 {
    _EGLDisplay *dpy = _eglGetCurrentDisplay();
-   if (dpy)
-      return dpy->Handle;
-   else
-      return EGL_NO_DISPLAY;
+   return _eglGetDisplayHandle(dpy);
 }
 
 
-EGLContext APIENTRY
+EGLContext EGLAPIENTRY
 eglGetCurrentContext(void)
 {
    _EGLContext *ctx = _eglGetCurrentContext();
-   if (ctx)
-      return ctx->Handle;
-   else
-      return EGL_NO_CONTEXT;
+   return _eglGetContextHandle(ctx);
 }
 
 
-EGLSurface APIENTRY
+EGLSurface EGLAPIENTRY
 eglGetCurrentSurface(EGLint readdraw)
 {
    _EGLSurface *s = _eglGetCurrentSurface(readdraw);
-   if (s)
-      return s->Handle;
-   else
-      return EGL_NO_SURFACE;
+   return _eglGetSurfaceHandle(s);
 }
 
 
-EGLint APIENTRY
+EGLint EGLAPIENTRY
 eglGetError(void)
 {
    _EGLThreadInfo *t = _eglGetCurrentThread();
@@ -308,7 +326,7 @@ eglGetError(void)
 }
 
 
-void (* APIENTRY eglGetProcAddress(const char *procname))()
+void (* EGLAPIENTRY eglGetProcAddress(const char *procname))()
 {
    typedef void (*genericFunc)();
    struct name_function {
@@ -376,12 +394,15 @@ void (* APIENTRY eglGetProcAddress(const char *procname))()
          return (genericFunc) egl_functions[i].function;
       }
    }
-#if 0
-   /* XXX enable this code someday */
-   return (genericFunc) _glapi_get_proc_address(procname);
-#else
+
+   /* now loop over drivers to query their procs */
+   for (i = 0; i < _eglGlobal.NumDrivers; i++) {
+      _EGLProc p = _eglGlobal.Drivers[i]->API.GetProcAddress(procname);
+      if (p)
+         return p;
+   }
+
    return NULL;
-#endif
 }
 
 
@@ -389,7 +410,7 @@ void (* APIENTRY eglGetProcAddress(const char *procname))()
  * EGL_MESA_screen extension
  */
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglChooseModeMESA(EGLDisplay dpy, EGLScreenMESA screen,
                   const EGLint *attrib_list, EGLModeMESA *modes,
                   EGLint modes_size, EGLint *num_modes)
@@ -402,7 +423,7 @@ eglChooseModeMESA(EGLDisplay dpy, EGLScreenMESA screen,
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglGetModesMESA(EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA *modes, EGLint mode_size, EGLint *num_mode)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -413,7 +434,7 @@ eglGetModesMESA(EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA *modes, EGLint
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglGetModeAttribMESA(EGLDisplay dpy, EGLModeMESA mode, EGLint attribute, EGLint *value)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -424,7 +445,7 @@ eglGetModeAttribMESA(EGLDisplay dpy, EGLModeMESA mode, EGLint attribute, EGLint
 }
 
 
-EGLBoolean APIENTRY
+EGLBoolean EGLAPIENTRY
 eglCopyContextMESA(EGLDisplay dpy, EGLContext source, EGLContext dest, EGLint mask)
 {
    _EGLDriver *drv = _eglLookupDriver(dpy);
@@ -508,21 +529,42 @@ eglQueryModeStringMESA(EGLDisplay dpy, EGLModeMESA mode)
 
 #ifdef EGL_VERSION_1_2
 
+
+/**
+ * Specify the client API to use for subsequent calls including:
+ *  eglCreateContext()
+ *  eglGetCurrentContext()
+ *  eglGetCurrentDisplay()
+ *  eglGetCurrentSurface()
+ *  eglMakeCurrent(when the ctx parameter is EGL NO CONTEXT)
+ *  eglWaitClient()
+ *  eglWaitNative()
+ * See section 3.7 "Rendering Context" in the EGL specification for details.
+ */
 EGLBoolean
 eglBindAPI(EGLenum api)
 {
    _EGLThreadInfo *t = _eglGetCurrentThread();
 
    switch (api) {
+#ifdef EGL_VERSION_1_4
+   case EGL_OPENGL_API:
+      if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_BIT) {
+         t->CurrentAPI = api;
+         return EGL_TRUE;
+      }
+      _eglError(EGL_BAD_PARAMETER, "eglBindAPI");
+      return EGL_FALSE;
+#endif
    case EGL_OPENGL_ES_API:
-      if (_eglGlobal.OpenGLESAPISupported) {
+      if (_eglGlobal.ClientAPIsMask & (EGL_OPENGL_ES_BIT | EGL_OPENGL_ES2_BIT)) {
          t->CurrentAPI = api;
          return EGL_TRUE;
       }
       _eglError(EGL_BAD_PARAMETER, "eglBindAPI");
       return EGL_FALSE;
    case EGL_OPENVG_API:
-      if (_eglGlobal.OpenVGAPISupported) {
+      if (_eglGlobal.ClientAPIsMask & EGL_OPENVG_BIT) {
          t->CurrentAPI = api;
          return EGL_TRUE;
       }
@@ -535,6 +577,18 @@ eglBindAPI(EGLenum api)
 }
 
 
+/**
+ * Return the last value set with eglBindAPI().
+ */
+EGLenum
+eglQueryAPI(void)
+{
+   /* returns one of EGL_OPENGL_API, EGL_OPENGL_ES_API or EGL_OPENVG_API */
+   _EGLThreadInfo *t = _eglGetCurrentThread();
+   return t->CurrentAPI;
+}
+
+
 EGLSurface
 eglCreatePbufferFromClientBuffer(EGLDisplay dpy, EGLenum buftype,
                                  EGLClientBuffer buffer, EGLConfig config,
@@ -546,15 +600,6 @@ eglCreatePbufferFromClientBuffer(EGLDisplay dpy, EGLenum buftype,
 }
 
 
-EGLenum
-eglQueryAPI(void)
-{
-   /* returns one of EGL_OPENGL_ES_API or EGL_OPENVG_API */
-   _EGLThreadInfo *t = _eglGetCurrentThread();
-   return t->CurrentAPI;
-}
-
-
 EGLBoolean
 eglReleaseThread(void)
 {
diff --git a/src/egl/main/eglapi.h b/src/egl/main/eglapi.h
index 555aa5dd9e..f6163a0c7a 100644
--- a/src/egl/main/eglapi.h
+++ b/src/egl/main/eglapi.h
@@ -2,10 +2,15 @@
 #define EGLAPI_INCLUDED
 
 /**
- * Typedefs for all EGL API entrypoint functions.
+ * A generic function ptr type
  */
+typedef void (*_EGLProc)();
 
 
+/**
+ * Typedefs for all EGL API entrypoint functions.
+ */
+
 /* driver funcs */
 typedef EGLBoolean (*Initialize_t)(_EGLDriver *, EGLDisplay dpy, EGLint *major, EGLint *minor);
 typedef EGLBoolean (*Terminate_t)(_EGLDriver *, EGLDisplay dpy);
@@ -39,6 +44,9 @@ typedef const char *(*QueryString_t)(_EGLDriver *drv, EGLDisplay dpy, EGLint nam
 typedef EGLBoolean (*WaitGL_t)(_EGLDriver *drv, EGLDisplay dpy);
 typedef EGLBoolean (*WaitNative_t)(_EGLDriver *drv, EGLDisplay dpy, EGLint engine);
 
+typedef _EGLProc (*GetProcAddress_t)(const char *procname);
+
+
 
 #ifdef EGL_MESA_screen_surface
 typedef EGLBoolean (*ChooseModeMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, const EGLint *attrib_list, EGLModeMESA *modes, EGLint modes_size, EGLint *num_modes);
@@ -95,6 +103,7 @@ struct _egl_api
    QueryString_t QueryString;
    WaitGL_t WaitGL;
    WaitNative_t WaitNative;
+   GetProcAddress_t GetProcAddress;
 
    /* EGL_MESA_screen extension */
    ChooseModeMESA_t ChooseModeMESA;
diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c
index 2a62f95b54..b19988f49a 100644
--- a/src/egl/main/eglconfig.c
+++ b/src/egl/main/eglconfig.c
@@ -17,48 +17,6 @@
 #define MIN2(A, B)  (((A) < (B)) ? (A) : (B))
 
 
-/**
- * Convert an _EGLConfig to a __GLcontextModes object.
- * NOTE: This routine may be incomplete - we're only making sure that
- * the fields needed by Mesa (for _mesa_create_context/framebuffer) are
- * set correctly.
- */
-void
-_eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode)
-{
-   memset(mode, 0, sizeof(*mode));
-
-   mode->rgbMode = GL_TRUE; /* no color index */
-   mode->colorIndexMode = GL_FALSE;
-   mode->doubleBufferMode = GL_TRUE;  /* always DB for now */
-   mode->stereoMode = GL_FALSE;
-
-   mode->redBits = GET_CONFIG_ATTRIB(config, EGL_RED_SIZE);
-   mode->greenBits = GET_CONFIG_ATTRIB(config, EGL_GREEN_SIZE);
-   mode->blueBits = GET_CONFIG_ATTRIB(config, EGL_BLUE_SIZE);
-   mode->alphaBits = GET_CONFIG_ATTRIB(config, EGL_ALPHA_SIZE);
-   mode->rgbBits = GET_CONFIG_ATTRIB(config, EGL_BUFFER_SIZE);
-
-   /* no rgba masks - fix? */
-
-   mode->depthBits = GET_CONFIG_ATTRIB(config, EGL_DEPTH_SIZE);
-   mode->haveDepthBuffer = mode->depthBits > 0;
-
-   mode->stencilBits = GET_CONFIG_ATTRIB(config, EGL_STENCIL_SIZE);
-   mode->haveStencilBuffer = mode->stencilBits > 0;
-
-   /* no accum */
-
-   mode->level = GET_CONFIG_ATTRIB(config, EGL_LEVEL);
-   mode->samples = GET_CONFIG_ATTRIB(config, EGL_SAMPLES);
-   mode->sampleBuffers = GET_CONFIG_ATTRIB(config, EGL_SAMPLE_BUFFERS);
-
-   /* surface type - not really needed */
-   mode->visualType = GLX_TRUE_COLOR;
-   mode->renderType = GLX_RGBA_BIT;
-}
-
-
 void
 _eglSetConfigAttrib(_EGLConfig *config, EGLint attr, EGLint val)
 {
@@ -76,7 +34,7 @@ void
 _eglInitConfig(_EGLConfig *config, EGLint id)
 {
    memset(config, 0, sizeof(*config));
-   config->Handle = id;
+   config->Handle = (EGLConfig) id;
    _eglSetConfigAttrib(config, EGL_CONFIG_ID,               id);
    _eglSetConfigAttrib(config, EGL_BIND_TO_TEXTURE_RGB,     EGL_DONT_CARE);
    _eglSetConfigAttrib(config, EGL_BIND_TO_TEXTURE_RGBA,    EGL_DONT_CARE);
@@ -85,9 +43,7 @@ _eglInitConfig(_EGLConfig *config, EGLint id)
    _eglSetConfigAttrib(config, EGL_NATIVE_VISUAL_TYPE,      EGL_DONT_CARE);
    _eglSetConfigAttrib(config, EGL_MIN_SWAP_INTERVAL,       EGL_DONT_CARE);
    _eglSetConfigAttrib(config, EGL_MAX_SWAP_INTERVAL,       EGL_DONT_CARE);
-   _eglSetConfigAttrib(config, EGL_SURFACE_TYPE,            
-                   EGL_SCREEN_BIT_MESA | EGL_PBUFFER_BIT |
-                   EGL_PIXMAP_BIT | EGL_WINDOW_BIT);
+   _eglSetConfigAttrib(config, EGL_SURFACE_TYPE,            EGL_WINDOW_BIT);
    _eglSetConfigAttrib(config, EGL_TRANSPARENT_TYPE,        EGL_NONE);
    _eglSetConfigAttrib(config, EGL_TRANSPARENT_RED_VALUE,   EGL_DONT_CARE);
    _eglSetConfigAttrib(config, EGL_TRANSPARENT_GREEN_VALUE, EGL_DONT_CARE);
@@ -100,7 +56,19 @@ _eglInitConfig(_EGLConfig *config, EGLint id)
 
 
 /**
+ * Return the public handle for an internal _EGLConfig.
+ * This is the inverse of _eglLookupConfig().
+ */
+EGLConfig
+_eglGetConfigHandle(_EGLConfig *config)
+{
+   return config ? config->Handle : 0;
+}
+
+
+/**
  * Given an EGLConfig handle, return the corresponding _EGLConfig object.
+ * This is the inverse of _eglGetConfigHandle().
  */
 _EGLConfig *
 _eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config)
@@ -108,8 +76,8 @@ _eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config)
    EGLint i;
    _EGLDisplay *disp = _eglLookupDisplay(dpy);
    for (i = 0; i < disp->NumConfigs; i++) {
-      if (disp->Configs[i].Handle == config) {
-          return disp->Configs + i;
+      if (disp->Configs[i]->Handle == config) {
+          return disp->Configs[i];
       }
    }
    return NULL;
@@ -118,23 +86,32 @@ _eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config)
 
 /**
  * Add the given _EGLConfig to the given display.
+ * Note that we just save the ptr to the config (we don't copy the config).
  */
 _EGLConfig *
-_eglAddConfig(_EGLDisplay *display, const _EGLConfig *config)
+_eglAddConfig(_EGLDisplay *display, _EGLConfig *config)
 {
-   _EGLConfig *newConfigs;
+   _EGLConfig **newConfigs;
    EGLint n;
 
+   /* do some sanity checks on the config's attribs */
+   assert(GET_CONFIG_ATTRIB(config, EGL_CONFIG_ID) > 0);
+   assert(GET_CONFIG_ATTRIB(config, EGL_RENDERABLE_TYPE) != 0x0);
+   assert(GET_CONFIG_ATTRIB(config, EGL_SURFACE_TYPE) != 0x0);
+   assert(GET_CONFIG_ATTRIB(config, EGL_RED_SIZE) > 0);
+   assert(GET_CONFIG_ATTRIB(config, EGL_GREEN_SIZE) > 0);
+   assert(GET_CONFIG_ATTRIB(config, EGL_BLUE_SIZE) > 0);
+
    n = display->NumConfigs;
 
-   newConfigs = (_EGLConfig *) realloc(display->Configs,
-                                       (n + 1) * sizeof(_EGLConfig));
+   /* realloc array of ptrs */
+   newConfigs = (_EGLConfig **) realloc(display->Configs,
+                                        (n + 1) * sizeof(_EGLConfig *));
    if (newConfigs) {
       display->Configs = newConfigs;
-      display->Configs[n] = *config; /* copy struct */
-      display->Configs[n].Handle = n;
+      display->Configs[n] = config;
       display->NumConfigs++;
-      return display->Configs + n;
+      return config;
    }
    else {
       return NULL;
@@ -176,7 +153,7 @@ _eglParseConfigAttribs(_EGLConfig *config, const EGLint *attrib_list)
       }
       else if (attr == EGL_RENDERABLE_TYPE) {
          EGLint renType = attrib_list[++i];
-         if (renType & ~(EGL_OPENGL_ES_BIT | EGL_OPENVG_BIT)) {
+         if (renType & ~(EGL_OPENGL_ES_BIT | EGL_OPENGL_ES2_BIT | EGL_OPENVG_BIT)) {
             _eglError(EGL_BAD_ATTRIBUTE, "eglChooseConfig");
             return EGL_FALSE;
          }
@@ -360,8 +337,8 @@ _eglChooseConfig(_EGLDriver *drv, EGLDisplay dpy, const EGLint *attrib_list,
 
    /* make array of pointers to qualifying configs */
    for (i = count = 0; i < disp->NumConfigs && count < config_size; i++) {
-      if (_eglConfigQualifies(disp->Configs + i, &criteria)) {
-         configList[count++] = disp->Configs + i;
+      if (_eglConfigQualifies(disp->Configs[i], &criteria)) {
+         configList[count++] = disp->Configs[i];
       }
    }
 
@@ -369,8 +346,10 @@ _eglChooseConfig(_EGLDriver *drv, EGLDisplay dpy, const EGLint *attrib_list,
    qsort(configList, count, sizeof(_EGLConfig *), _eglCompareConfigs);
 
    /* copy config handles to output array */
-   for (i = 0; i < count; i++) {
-      configs[i] = configList[i]->Handle;
+   if (configs) {
+      for (i = 0; i < count; i++) {
+         configs[i] = configList[i]->Handle;
+      }
    }
 
    free(configList);
@@ -419,7 +398,7 @@ _eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs,
       EGLint i;
       *num_config = MIN2(disp->NumConfigs, config_size);
       for (i = 0; i < *num_config; i++) {
-         configs[i] = disp->Configs[i].Handle;
+         configs[i] = disp->Configs[i]->Handle;
       }
    }
    else {
@@ -429,210 +408,3 @@ _eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs,
 
    return EGL_TRUE;
 }
-
-
-/**
- * Creates a set of \c __GLcontextModes that a driver will expose.
- * 
- * A set of \c __GLcontextModes will be created based on the supplied
- * parameters.  The number of modes processed will be 2 *
- * \c num_depth_stencil_bits * \c num_db_modes.
- * 
- * For the most part, data is just copied from \c depth_bits, \c stencil_bits,
- * \c db_modes, and \c visType into each \c __GLcontextModes element.
- * However, the meanings of \c fb_format and \c fb_type require further
- * explanation.  The \c fb_format specifies which color components are in
- * each pixel and what the default order is.  For example, \c GL_RGB specifies
- * that red, green, blue are available and red is in the "most significant"
- * position and blue is in the "least significant".  The \c fb_type specifies
- * the bit sizes of each component and the actual ordering.  For example, if
- * \c GL_UNSIGNED_SHORT_5_6_5_REV is specified with \c GL_RGB, bits [15:11]
- * are the blue value, bits [10:5] are the green value, and bits [4:0] are
- * the red value.
- * 
- * One sublte issue is the combination of \c GL_RGB  or \c GL_BGR and either
- * of the \c GL_UNSIGNED_INT_8_8_8_8 modes.  The resulting mask values in the
- * \c __GLcontextModes structure is \b identical to the \c GL_RGBA or
- * \c GL_BGRA case, except the \c alphaMask is zero.  This means that, as
- * far as this routine is concerned, \c GL_RGB with \c GL_UNSIGNED_INT_8_8_8_8
- * still uses 32-bits.
- *
- * If in doubt, look at the tables used in the function.
- * 
- * \param ptr_to_modes  Pointer to a pointer to a linked list of
- *                      \c __GLcontextModes.  Upon completion, a pointer to
- *                      the next element to be process will be stored here.
- *                      If the function fails and returns \c GL_FALSE, this
- *                      value will be unmodified, but some elements in the
- *                      linked list may be modified.
- * \param fb_format     Format of the framebuffer.  Currently only \c GL_RGB,
- *                      \c GL_RGBA, \c GL_BGR, and \c GL_BGRA are supported.
- * \param fb_type       Type of the pixels in the framebuffer.  Currently only
- *                      \c GL_UNSIGNED_SHORT_5_6_5, 
- *                      \c GL_UNSIGNED_SHORT_5_6_5_REV,
- *                      \c GL_UNSIGNED_INT_8_8_8_8, and
- *                      \c GL_UNSIGNED_INT_8_8_8_8_REV are supported.
- * \param depth_bits    Array of depth buffer sizes to be exposed.
- * \param stencil_bits  Array of stencil buffer sizes to be exposed.
- * \param num_depth_stencil_bits  Number of entries in both \c depth_bits and
- *                      \c stencil_bits.
- * \param db_modes      Array of buffer swap modes.  If an element has a
- *                      value of \c GLX_NONE, then it represents a
- *                      single-buffered mode.  Other valid values are
- *                      \c GLX_SWAP_EXCHANGE_OML, \c GLX_SWAP_COPY_OML, and
- *                      \c GLX_SWAP_UNDEFINED_OML.  See the
- *                      GLX_OML_swap_method extension spec for more details.
- * \param num_db_modes  Number of entries in \c db_modes.
- * \param visType       GLX visual type.  Usually either \c GLX_TRUE_COLOR or
- *                      \c GLX_DIRECT_COLOR.
- * 
- * \returns
- * \c GL_TRUE on success or \c GL_FALSE on failure.  Currently the only
- * cause of failure is a bad parameter (i.e., unsupported \c fb_format or
- * \c fb_type).
- * 
- * \todo
- * There is currently no way to support packed RGB modes (i.e., modes with
- * exactly 3 bytes per pixel) or floating-point modes.  This could probably
- * be done by creating some new, private enums with clever names likes
- * \c GL_UNSIGNED_3BYTE_8_8_8, \c GL_4FLOAT_32_32_32_32, 
- * \c GL_4HALF_16_16_16_16, etc.  We can cross that bridge when we come to it.
- */
-GLboolean
-_eglFillInConfigs(_EGLConfig * configs,
-                  GLenum fb_format, GLenum fb_type,
-                  const uint8_t * depth_bits, const uint8_t * stencil_bits,
-                  unsigned num_depth_stencil_bits,
-                  const GLenum * db_modes, unsigned num_db_modes,
-                  int visType)
-{
-   static const uint8_t bits_table[3][4] = {
-            /* R  G  B  A */
-            { 5, 6, 5, 0 },  /* Any GL_UNSIGNED_SHORT_5_6_5 */
-            { 8, 8, 8, 0 },  /* Any RGB with any GL_UNSIGNED_INT_8_8_8_8 */
-            { 8, 8, 8, 8 }  /* Any RGBA with any GL_UNSIGNED_INT_8_8_8_8 */
-         };
-
-   /* The following arrays are all indexed by the fb_type masked with 0x07.
-    * Given the four supported fb_type values, this results in valid array
-    * indices of 3, 4, 5, and 7.
-    */
-   static const uint32_t masks_table_rgb[8][4] = {
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000},  /* 5_6_5       */
-            {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000},  /* 5_6_5_REV   */
-            {0xFF000000, 0x00FF0000, 0x0000FF00, 0x00000000},  /* 8_8_8_8     */
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x000000FF, 0x0000FF00, 0x00FF0000, 0x00000000}  /* 8_8_8_8_REV */
-         };
-
-   static const uint32_t masks_table_rgba[8][4] = {
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000},  /* 5_6_5       */
-            {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000},  /* 5_6_5_REV   */
-            {0xFF000000, 0x00FF0000, 0x0000FF00, 0x000000FF},  /* 8_8_8_8     */
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x000000FF, 0x0000FF00, 0x00FF0000, 0xFF000000},  /* 8_8_8_8_REV */
-         };
-
-   static const uint32_t masks_table_bgr[8][4] = {
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000},  /* 5_6_5       */
-            {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000},  /* 5_6_5_REV   */
-            {0x0000FF00, 0x00FF0000, 0xFF000000, 0x00000000},  /* 8_8_8_8     */
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x00FF0000, 0x0000FF00, 0x000000FF, 0x00000000},  /* 8_8_8_8_REV */
-         };
-
-   static const uint32_t masks_table_bgra[8][4] = {
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000},  /* 5_6_5       */
-            {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000},  /* 5_6_5_REV   */
-            {0x0000FF00, 0x00FF0000, 0xFF000000, 0x000000FF},  /* 8_8_8_8     */
-            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
-            {0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000},  /* 8_8_8_8_REV */
-         };
-
-   static const uint8_t bytes_per_pixel[8] = {
-            0, 0, 0, 2, 2, 4, 0, 4
-         };
-
-   const uint8_t * bits;
-   const uint32_t * masks;
-   const int index = fb_type & 0x07;
-   _EGLConfig *config;
-   unsigned i;
-   unsigned j;
-   unsigned k;
-
-   if ( bytes_per_pixel[index] == 0 ) {
-      _eglLog(_EGL_INFO,
-              "[%s:%u] Framebuffer type 0x%04x has 0 bytes per pixel.",
-              __FUNCTION__, __LINE__, fb_type);
-      return GL_FALSE;
-   }
-
-   /* Valid types are GL_UNSIGNED_SHORT_5_6_5 and GL_UNSIGNED_INT_8_8_8_8 and
-    * the _REV versions.
-    *
-    * Valid formats are GL_RGBA, GL_RGB, and GL_BGRA.
-    */
-   switch ( fb_format ) {
-   case GL_RGB:
-      bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[1];
-      masks = masks_table_rgb[index];
-      break;
-
-   case GL_RGBA:
-      bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[2];
-      masks = masks_table_rgba[index];
-      break;
-
-   case GL_BGR:
-      bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[1];
-      masks = masks_table_bgr[index];
-      break;
-
-   case GL_BGRA:
-      bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[2];
-      masks = masks_table_bgra[index];
-      break;
-
-   default:
-      _eglLog(_EGL_WARNING,
-              "[%s:%u] Framebuffer format 0x%04x is not GL_RGB, GL_RGBA, GL_BGR, or GL_BGRA.",
-              __FUNCTION__, __LINE__, fb_format);
-      return GL_FALSE;
-   }
-
-   config = configs;
-   for (k = 0; k < num_depth_stencil_bits; k++) {
-      for (i = 0; i < num_db_modes; i++)  {
-         for (j = 0; j < 2; j++) {
-            _eglSetConfigAttrib(config, EGL_RED_SIZE, bits[0]);
-            _eglSetConfigAttrib(config, EGL_GREEN_SIZE, bits[1]);
-            _eglSetConfigAttrib(config, EGL_BLUE_SIZE, bits[2]);
-            _eglSetConfigAttrib(config, EGL_ALPHA_SIZE, bits[3]);
-            _eglSetConfigAttrib(config, EGL_BUFFER_SIZE,
-                                bits[0] + bits[1] + bits[2] + bits[3]);
-
-            _eglSetConfigAttrib(config, EGL_STENCIL_SIZE, stencil_bits[k]);
-            _eglSetConfigAttrib(config, EGL_DEPTH_SIZE, depth_bits[i]);
-
-            _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_SCREEN_BIT_MESA |
-                         EGL_PBUFFER_BIT | EGL_PIXMAP_BIT | EGL_WINDOW_BIT);
-
-            config++;
-         }
-      }
-   }
-   return GL_TRUE;
-}
diff --git a/src/egl/main/eglconfig.h b/src/egl/main/eglconfig.h
index 441b3f4d20..db1c4c10e0 100644
--- a/src/egl/main/eglconfig.h
+++ b/src/egl/main/eglconfig.h
@@ -3,10 +3,10 @@
 
 
 #include "egltypedefs.h"
-#include "GL/internal/glcore.h"
+#include <GLES/gl.h>
 
 
-#define MAX_ATTRIBS 100
+#define MAX_ATTRIBS 128
 #define FIRST_ATTRIB EGL_BUFFER_SIZE
 
 
@@ -17,7 +17,11 @@ struct _egl_config
 };
 
 
-#define SET_CONFIG_ATTRIB(CONF, ATTR, VAL) ((CONF)->Attrib[(ATTR) - FIRST_ATTRIB] = VAL)
+#define SET_CONFIG_ATTRIB(CONF, ATTR, VAL) \
+   assert((ATTR) - FIRST_ATTRIB < MAX_ATTRIBS); \
+   ((CONF)->Attrib[(ATTR) - FIRST_ATTRIB] = VAL)
+
+
 #define GET_CONFIG_ATTRIB(CONF, ATTR) ((CONF)->Attrib[(ATTR) - FIRST_ATTRIB])
 
 
@@ -25,12 +29,16 @@ extern void
 _eglInitConfig(_EGLConfig *config, EGLint id);
 
 
+extern EGLConfig
+_eglGetConfigHandle(_EGLConfig *config);
+
+
 extern _EGLConfig *
 _eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config);
 
 
 extern _EGLConfig *
-_eglAddConfig(_EGLDisplay *display, const _EGLConfig *config);
+_eglAddConfig(_EGLDisplay *display, _EGLConfig *config);
 
 
 extern EGLBoolean
@@ -52,16 +60,5 @@ _eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs, EGLint confi
 extern void
 _eglSetConfigAttrib(_EGLConfig *config, EGLint attr, EGLint val);
 
-extern GLboolean
-_eglFillInConfigs( _EGLConfig *configs,
-		GLenum fb_format, GLenum fb_type,
-		const uint8_t * depth_bits, const uint8_t * stencil_bits,
-		unsigned num_depth_stencil_bits,
-		const GLenum * db_modes, unsigned num_db_modes,
-		int visType );
-                
-extern void
-_eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode);
-
 
 #endif /* EGLCONFIG_INCLUDED */
diff --git a/src/egl/main/eglconfigutil.c b/src/egl/main/eglconfigutil.c
new file mode 100644
index 0000000000..7061df691b
--- /dev/null
+++ b/src/egl/main/eglconfigutil.c
@@ -0,0 +1,258 @@
+/**
+ * Extra utility functions related to EGL configs.
+ */
+
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "eglconfigutil.h"
+#include "egllog.h"
+
+
+/**
+ * Convert an _EGLConfig to a __GLcontextModes object.
+ * NOTE: This routine may be incomplete - we're only making sure that
+ * the fields needed by Mesa (for _mesa_create_context/framebuffer) are
+ * set correctly.
+ */
+void
+_eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode)
+{
+   memset(mode, 0, sizeof(*mode));
+
+   mode->rgbMode = GL_TRUE; /* no color index */
+   mode->colorIndexMode = GL_FALSE;
+   mode->doubleBufferMode = GL_TRUE;  /* always DB for now */
+   mode->stereoMode = GL_FALSE;
+
+   mode->redBits = GET_CONFIG_ATTRIB(config, EGL_RED_SIZE);
+   mode->greenBits = GET_CONFIG_ATTRIB(config, EGL_GREEN_SIZE);
+   mode->blueBits = GET_CONFIG_ATTRIB(config, EGL_BLUE_SIZE);
+   mode->alphaBits = GET_CONFIG_ATTRIB(config, EGL_ALPHA_SIZE);
+   mode->rgbBits = GET_CONFIG_ATTRIB(config, EGL_BUFFER_SIZE);
+
+   /* no rgba masks - fix? */
+
+   mode->depthBits = GET_CONFIG_ATTRIB(config, EGL_DEPTH_SIZE);
+   mode->haveDepthBuffer = mode->depthBits > 0;
+
+   mode->stencilBits = GET_CONFIG_ATTRIB(config, EGL_STENCIL_SIZE);
+   mode->haveStencilBuffer = mode->stencilBits > 0;
+
+   /* no accum */
+
+   mode->level = GET_CONFIG_ATTRIB(config, EGL_LEVEL);
+   mode->samples = GET_CONFIG_ATTRIB(config, EGL_SAMPLES);
+   mode->sampleBuffers = GET_CONFIG_ATTRIB(config, EGL_SAMPLE_BUFFERS);
+
+   /* surface type - not really needed */
+   mode->visualType = GLX_TRUE_COLOR;
+   mode->renderType = GLX_RGBA_BIT;
+}
+
+
+
+/**
+ * Creates a set of \c _EGLConfigs that a driver will expose.
+ * 
+ * A set of \c __GLcontextModes will be created based on the supplied
+ * parameters.  The number of modes processed will be 2 *
+ * \c num_depth_stencil_bits * \c num_db_modes.
+ * 
+ * For the most part, data is just copied from \c depth_bits, \c stencil_bits,
+ * \c db_modes, and \c visType into each \c __GLcontextModes element.
+ * However, the meanings of \c fb_format and \c fb_type require further
+ * explanation.  The \c fb_format specifies which color components are in
+ * each pixel and what the default order is.  For example, \c GL_RGB specifies
+ * that red, green, blue are available and red is in the "most significant"
+ * position and blue is in the "least significant".  The \c fb_type specifies
+ * the bit sizes of each component and the actual ordering.  For example, if
+ * \c GL_UNSIGNED_SHORT_5_6_5_REV is specified with \c GL_RGB, bits [15:11]
+ * are the blue value, bits [10:5] are the green value, and bits [4:0] are
+ * the red value.
+ * 
+ * One sublte issue is the combination of \c GL_RGB  or \c GL_BGR and either
+ * of the \c GL_UNSIGNED_INT_8_8_8_8 modes.  The resulting mask values in the
+ * \c __GLcontextModes structure is \b identical to the \c GL_RGBA or
+ * \c GL_BGRA case, except the \c alphaMask is zero.  This means that, as
+ * far as this routine is concerned, \c GL_RGB with \c GL_UNSIGNED_INT_8_8_8_8
+ * still uses 32-bits.
+ *
+ * If in doubt, look at the tables used in the function.
+ * 
+ * \param configs       the array of configs generated
+ * \param fb_format     Format of the framebuffer.  Currently only \c GL_RGB,
+ *                      \c GL_RGBA, \c GL_BGR, and \c GL_BGRA are supported.
+ * \param fb_type       Type of the pixels in the framebuffer.  Currently only
+ *                      \c GL_UNSIGNED_SHORT_5_6_5, 
+ *                      \c GL_UNSIGNED_SHORT_5_6_5_REV,
+ *                      \c GL_UNSIGNED_INT_8_8_8_8, and
+ *                      \c GL_UNSIGNED_INT_8_8_8_8_REV are supported.
+ * \param depth_bits    Array of depth buffer sizes to be exposed.
+ * \param stencil_bits  Array of stencil buffer sizes to be exposed.
+ * \param num_depth_stencil_bits  Number of entries in both \c depth_bits and
+ *                      \c stencil_bits.
+ * \param db_modes      Array of buffer swap modes.  If an element has a
+ *                      value of \c GLX_NONE, then it represents a
+ *                      single-buffered mode.  Other valid values are
+ *                      \c GLX_SWAP_EXCHANGE_OML, \c GLX_SWAP_COPY_OML, and
+ *                      \c GLX_SWAP_UNDEFINED_OML.  See the
+ *                      GLX_OML_swap_method extension spec for more details.
+ * \param num_db_modes  Number of entries in \c db_modes.
+ * \param visType       GLX visual type.  Usually either \c GLX_TRUE_COLOR or
+ *                      \c GLX_DIRECT_COLOR.
+ * 
+ * \returns
+ * \c GL_TRUE on success or \c GL_FALSE on failure.  Currently the only
+ * cause of failure is a bad parameter (i.e., unsupported \c fb_format or
+ * \c fb_type).
+ * 
+ * \todo
+ * There is currently no way to support packed RGB modes (i.e., modes with
+ * exactly 3 bytes per pixel) or floating-point modes.  This could probably
+ * be done by creating some new, private enums with clever names likes
+ * \c GL_UNSIGNED_3BYTE_8_8_8, \c GL_4FLOAT_32_32_32_32, 
+ * \c GL_4HALF_16_16_16_16, etc.  We can cross that bridge when we come to it.
+ */
+EGLBoolean
+_eglFillInConfigs(_EGLConfig * configs,
+                  GLenum fb_format, GLenum fb_type,
+                  const uint8_t * depth_bits, const uint8_t * stencil_bits,
+                  unsigned num_depth_stencil_bits,
+                  const GLenum * db_modes, unsigned num_db_modes,
+                  int visType)
+{
+   static const uint8_t bits_table[3][4] = {
+            /* R  G  B  A */
+            { 5, 6, 5, 0 },  /* Any GL_UNSIGNED_SHORT_5_6_5 */
+            { 8, 8, 8, 0 },  /* Any RGB with any GL_UNSIGNED_INT_8_8_8_8 */
+            { 8, 8, 8, 8 }  /* Any RGBA with any GL_UNSIGNED_INT_8_8_8_8 */
+         };
+
+   /* The following arrays are all indexed by the fb_type masked with 0x07.
+    * Given the four supported fb_type values, this results in valid array
+    * indices of 3, 4, 5, and 7.
+    */
+   static const uint32_t masks_table_rgb[8][4] = {
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000},  /* 5_6_5       */
+            {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000},  /* 5_6_5_REV   */
+            {0xFF000000, 0x00FF0000, 0x0000FF00, 0x00000000},  /* 8_8_8_8     */
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x000000FF, 0x0000FF00, 0x00FF0000, 0x00000000}  /* 8_8_8_8_REV */
+         };
+
+   static const uint32_t masks_table_rgba[8][4] = {
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000},  /* 5_6_5       */
+            {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000},  /* 5_6_5_REV   */
+            {0xFF000000, 0x00FF0000, 0x0000FF00, 0x000000FF},  /* 8_8_8_8     */
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x000000FF, 0x0000FF00, 0x00FF0000, 0xFF000000},  /* 8_8_8_8_REV */
+         };
+
+   static const uint32_t masks_table_bgr[8][4] = {
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000},  /* 5_6_5       */
+            {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000},  /* 5_6_5_REV   */
+            {0x0000FF00, 0x00FF0000, 0xFF000000, 0x00000000},  /* 8_8_8_8     */
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x00FF0000, 0x0000FF00, 0x000000FF, 0x00000000},  /* 8_8_8_8_REV */
+         };
+
+   static const uint32_t masks_table_bgra[8][4] = {
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000},  /* 5_6_5       */
+            {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000},  /* 5_6_5_REV   */
+            {0x0000FF00, 0x00FF0000, 0xFF000000, 0x000000FF},  /* 8_8_8_8     */
+            {0x00000000, 0x00000000, 0x00000000, 0x00000000},
+            {0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000},  /* 8_8_8_8_REV */
+         };
+
+   static const uint8_t bytes_per_pixel[8] = {
+            0, 0, 0, 2, 2, 4, 0, 4
+         };
+
+   const uint8_t * bits;
+   const uint32_t * masks;
+   const int index = fb_type & 0x07;
+   _EGLConfig *config;
+   unsigned i;
+   unsigned j;
+   unsigned k;
+
+   if ( bytes_per_pixel[index] == 0 ) {
+      _eglLog(_EGL_INFO,
+              "[%s:%u] Framebuffer type 0x%04x has 0 bytes per pixel.",
+              __FUNCTION__, __LINE__, fb_type);
+      return GL_FALSE;
+   }
+
+   /* Valid types are GL_UNSIGNED_SHORT_5_6_5 and GL_UNSIGNED_INT_8_8_8_8 and
+    * the _REV versions.
+    *
+    * Valid formats are GL_RGBA, GL_RGB, and GL_BGRA.
+    */
+   switch ( fb_format ) {
+   case GL_RGB:
+      bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[1];
+      masks = masks_table_rgb[index];
+      break;
+
+   case GL_RGBA:
+      bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[2];
+      masks = masks_table_rgba[index];
+      break;
+
+#if 0
+   case GL_BGR:
+      bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[1];
+      masks = masks_table_bgr[index];
+      break;
+
+   case GL_BGRA:
+      bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[2];
+      masks = masks_table_bgra[index];
+      break;
+#endif
+
+   default:
+      _eglLog(_EGL_WARNING,
+              "[%s:%u] Framebuffer format 0x%04x is not GL_RGB, GL_RGBA, GL_BGR, or GL_BGRA.",
+              __FUNCTION__, __LINE__, fb_format);
+      return GL_FALSE;
+   }
+
+   config = configs;
+   for (k = 0; k < num_depth_stencil_bits; k++) {
+      for (i = 0; i < num_db_modes; i++)  {
+         for (j = 0; j < 2; j++) {
+            _eglSetConfigAttrib(config, EGL_RED_SIZE, bits[0]);
+            _eglSetConfigAttrib(config, EGL_GREEN_SIZE, bits[1]);
+            _eglSetConfigAttrib(config, EGL_BLUE_SIZE, bits[2]);
+            _eglSetConfigAttrib(config, EGL_ALPHA_SIZE, bits[3]);
+            _eglSetConfigAttrib(config, EGL_BUFFER_SIZE,
+                                bits[0] + bits[1] + bits[2] + bits[3]);
+
+            _eglSetConfigAttrib(config, EGL_STENCIL_SIZE, stencil_bits[k]);
+            _eglSetConfigAttrib(config, EGL_DEPTH_SIZE, depth_bits[i]);
+
+            _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_SCREEN_BIT_MESA |
+                         EGL_PBUFFER_BIT | EGL_PIXMAP_BIT | EGL_WINDOW_BIT);
+
+            config++;
+         }
+      }
+   }
+   return GL_TRUE;
+}
+
diff --git a/src/egl/main/eglconfigutil.h b/src/egl/main/eglconfigutil.h
new file mode 100644
index 0000000000..c477b94737
--- /dev/null
+++ b/src/egl/main/eglconfigutil.h
@@ -0,0 +1,26 @@
+
+#ifndef EGLCONFIGUTIL_INCLUDED
+#define EGLCONFIGUTIL_INCLUDED
+
+#include "eglconfig.h"
+#include "GL/internal/glcore.h"
+#if (!defined(WIN32) && !defined(_WIN32_WCE))
+#include "stdint.h"
+#endif
+
+
+extern void
+_eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode);
+
+
+extern EGLBoolean
+_eglFillInConfigs( _EGLConfig *configs,
+                   EGLenum fb_format, EGLenum fb_type,
+                   const uint8_t * depth_bits, const uint8_t * stencil_bits,
+                   unsigned num_depth_stencil_bits,
+                   const EGLenum * db_modes, unsigned num_db_modes,
+                   int visType );
+
+
+
+#endif /* EGLCONFIGUTIL_INCLUDED */
diff --git a/src/egl/main/eglcontext.c b/src/egl/main/eglcontext.c
index 374c006dae..461679db09 100644
--- a/src/egl/main/eglcontext.c
+++ b/src/egl/main/eglcontext.c
@@ -6,12 +6,12 @@
 #include "egldisplay.h"
 #include "egldriver.h"
 #include "eglglobals.h"
-#include "eglhash.h"
 #include "eglsurface.h"
 
 
 /**
- * Initialize the given _EGLContext object to defaults.
+ * Initialize the given _EGLContext object to defaults and/or the values
+ * in the attrib_list.
  */
 EGLBoolean
 _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
@@ -20,42 +20,54 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx,
    _EGLConfig *conf;
    _EGLDisplay *display = _eglLookupDisplay(dpy);
    EGLint i;
+   const EGLenum api = eglQueryAPI();
+
+   if (api == EGL_NONE) {
+      _eglError(EGL_BAD_MATCH, "eglCreateContext(no client API)");
+      return EGL_FALSE;
+   }
 
    conf = _eglLookupConfig(drv, dpy, config);
    if (!conf) {
-      _eglError(EGL_BAD_CONFIG, "eglCreateContext");
+      _eglError(EGL_BAD_CONFIG, "_eglInitContext");
       return EGL_FALSE;
    }
 
+   memset(ctx, 0, sizeof(_EGLContext));
+
+   ctx->ClientVersion = 1; /* the default, per EGL spec */
+
    for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
       switch (attrib_list[i]) {
-         /* no attribs defined for now */
+      case EGL_CONTEXT_CLIENT_VERSION:
+         i++;
+         ctx->ClientVersion = attrib_list[i];
+         break;
       default:
-         _eglError(EGL_BAD_ATTRIBUTE, "eglCreateContext");
-         return EGL_NO_CONTEXT;
+         _eglError(EGL_BAD_ATTRIBUTE, "_eglInitContext");
+         return EGL_FALSE;
       }
    }
 
-   memset(ctx, 0, sizeof(_EGLContext));
    ctx->Display = display;
    ctx->Config = conf;
    ctx->DrawSurface = EGL_NO_SURFACE;
    ctx->ReadSurface = EGL_NO_SURFACE;
+   ctx->ClientAPI = api;
 
    return EGL_TRUE;
 }
 
 
-/*
- * Assign an EGLContext handle to the _EGLContext object then put it into
- * the hash table.
+/**
+ * Save a new _EGLContext into the hash table.
  */
 void
 _eglSaveContext(_EGLContext *ctx)
 {
-   assert(ctx);
-   ctx->Handle = _eglHashGenKey(_eglGlobal.Contexts);
-   _eglHashInsert(_eglGlobal.Contexts, ctx->Handle, ctx);
+   /* no-op.
+    * Public EGLContext handle and private _EGLContext are the same.
+    */
 }
 
 
@@ -65,19 +77,34 @@ _eglSaveContext(_EGLContext *ctx)
 void
 _eglRemoveContext(_EGLContext *ctx)
 {
-   _eglHashRemove(_eglGlobal.Contexts, ctx->Handle);
+   /* no-op.
+    * Public EGLContext handle and private _EGLContext are the same.
+    */
+}
+
+
+/**
+ * Return the public handle for the given private context ptr.
+ * This is the inverse of _eglLookupContext().
+ */
+EGLContext
+_eglGetContextHandle(_EGLContext *ctx)
+{
+   /* just a cast! */
+   return (EGLContext) ctx;
 }
 
 
 /**
  * Return the _EGLContext object that corresponds to the given
  * EGLContext handle.
+ * This is the inverse of _eglGetContextHandle().
  */
 _EGLContext *
 _eglLookupContext(EGLContext ctx)
 {
-   _EGLContext *c = (_EGLContext *) _eglHashLookup(_eglGlobal.Contexts, ctx);
-   return c;
+   /* just a cast since EGLContext is just a void ptr */
+   return (_EGLContext *) ctx;
 }
 
 
@@ -112,7 +139,7 @@ _eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
    }
 
    _eglSaveContext(context);
-   return context->Handle;
+   return (EGLContext) context;
 #endif
    return EGL_NO_CONTEXT;
 }
@@ -126,7 +153,6 @@ _eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx)
 {
    _EGLContext *context = _eglLookupContext(ctx);
    if (context) {
-      _eglHashRemove(_eglGlobal.Contexts, ctx);
       if (context->IsBound) {
          context->DeletePending = EGL_TRUE;
       }
@@ -163,8 +189,11 @@ _eglQueryContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx,
 #ifdef EGL_VERSION_1_2
    case EGL_CONTEXT_CLIENT_TYPE:
       *value = c->ClientAPI;
-      return EGL_FALSE;
+      return EGL_TRUE;
 #endif /* EGL_VERSION_1_2 */
+   case EGL_CONTEXT_CLIENT_VERSION:
+      *value = c->ClientVersion;
+      return EGL_TRUE;
    default:
       _eglError(EGL_BAD_ATTRIBUTE, "eglQueryContext");
       return EGL_FALSE;
@@ -239,7 +268,7 @@ _eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d,
             ctx = NULL;
          }
          /* really delete context now */
-         drv->API.DestroyContext(drv, dpy, oldContext->Handle);
+         drv->API.DestroyContext(drv, dpy, _eglGetContextHandle(oldContext));
       }
    }
 
diff --git a/src/egl/main/eglcontext.h b/src/egl/main/eglcontext.h
index 82bfde151f..34fee9c637 100644
--- a/src/egl/main/eglcontext.h
+++ b/src/egl/main/eglcontext.h
@@ -11,8 +11,6 @@
  */
 struct _egl_context
 {
-   EGLContext Handle;  /* The public/opaque handle which names this object */
-
    _EGLDisplay *Display; /* who do I belong to? */
 
    _EGLConfig *Config;
@@ -22,9 +20,9 @@ struct _egl_context
 
    EGLBoolean IsBound;
    EGLBoolean DeletePending;
-#ifdef EGL_VERSION_1_2
-   EGLint ClientAPI;  /* Either EGL_OPENGL_ES_API or EGL_OPENVG_API */
-#endif /* EGL_VERSION_1_2 */
+
+   EGLint ClientAPI; /**< EGL_OPENGL_ES_API, EGL_OPENGL_API, EGL_OPENVG_API */
+   EGLint ClientVersion; /**< 1 = OpenGLES 1.x, 2 = OpenGLES 2.x */
 };
 
 
@@ -41,6 +39,10 @@ extern void
 _eglRemoveContext(_EGLContext *ctx);
 
 
+extern EGLContext
+_eglGetContextHandle(_EGLContext *ctx);
+
+
 extern _EGLContext *
 _eglLookupContext(EGLContext ctx);
  
diff --git a/src/egl/main/egldefines.h b/src/egl/main/egldefines.h
new file mode 100644
index 0000000000..8fc2301b79
--- /dev/null
+++ b/src/egl/main/egldefines.h
@@ -0,0 +1,45 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+
+/**
+ * Internal EGL defines
+ */
+
+
+#ifndef EGLDEFINES_INCLUDED
+#define EGLDEFINES_INCLUDED
+
+
+#define _EGL_MAX_EXTENSIONS_LEN 1000
+
+#define _EGL_VENDOR_STRING "Mesa Project"
+
+
+
+#endif /* EGLDEFINES_INCLUDED */
diff --git a/src/egl/main/egldisplay.c b/src/egl/main/egldisplay.c
index 074a85bf26..47a2323eaf 100644
--- a/src/egl/main/egldisplay.c
+++ b/src/egl/main/egldisplay.c
@@ -1,52 +1,88 @@
+
+/**
+ * Functions related to EGLDisplay.
+ */
+
+#include <assert.h>
 #include <stdlib.h>
 #include <string.h>
 #include "eglcontext.h"
 #include "egldisplay.h"
+#include "egldriver.h"
 #include "eglglobals.h"
 #include "eglhash.h"
-
-
-static char *
-my_strdup(const char *s)
-{
-   int l = strlen(s);
-   char *s2 = malloc(l + 1);
-   strcpy(s2, s);
-   return s2;
-}
+#include "eglstring.h"
 
 
 /**
- * We're assuming that the NativeDisplayType parameter is actually
- * a string.
- * Return a new _EGLDisplay object for the given displayName
+ * Allocate a new _EGLDisplay object for the given nativeDisplay handle.
+ * We'll also try to determine the device driver name at this time.
+ *
+ * Note that nativeDisplay may be an X Display ptr, or a string.
  */
 _EGLDisplay *
-_eglNewDisplay(NativeDisplayType displayName)
+_eglNewDisplay(NativeDisplayType nativeDisplay)
 {
    _EGLDisplay *dpy = (_EGLDisplay *) calloc(1, sizeof(_EGLDisplay));
    if (dpy) {
-      dpy->Handle = _eglHashGenKey(_eglGlobal.Displays);
-      _eglHashInsert(_eglGlobal.Displays, dpy->Handle, dpy);
-      if (displayName)
-         dpy->Name = my_strdup(displayName);
-      else
-         dpy->Name = NULL;
-      dpy->Driver = NULL;  /* this gets set later */
+      EGLuint key = _eglHashGenKey(_eglGlobal.Displays);
+
+      dpy->Handle = (EGLDisplay) key;
+      _eglHashInsert(_eglGlobal.Displays, key, dpy);
+
+      dpy->NativeDisplay = nativeDisplay;
+#if defined(_EGL_PLATFORM_X)
+      dpy->Xdpy = (Display *) nativeDisplay;
+#endif
+
+      dpy->DriverName = _eglChooseDriver(dpy);
+      if (!dpy->DriverName) {
+         free(dpy);
+         return NULL;
+      }
    }
    return dpy;
 }
 
 
 /**
+ * Return the public handle for an internal _EGLDisplay.
+ * This is the inverse of _eglLookupDisplay().
+ */
+EGLDisplay
+_eglGetDisplayHandle(_EGLDisplay *display)
+{
+   if (display)
+      return display->Handle;
+   else
+      return EGL_NO_DISPLAY;
+}
+
+ 
+/**
  * Return the _EGLDisplay object that corresponds to the given public/
  * opaque display handle.
+ * This is the inverse of _eglGetDisplayHandle().
  */
 _EGLDisplay *
 _eglLookupDisplay(EGLDisplay dpy)
 {
-   _EGLDisplay *d = (_EGLDisplay *) _eglHashLookup(_eglGlobal.Displays, dpy);
-   return d;
+   EGLuint key = (EGLuint) dpy;
+   if (!_eglGlobal.Displays)
+      return NULL;
+   return (_EGLDisplay *) _eglHashLookup(_eglGlobal.Displays, key);
+}
+
+
+void
+_eglSaveDisplay(_EGLDisplay *dpy)
+{
+   EGLuint key = _eglHashGenKey(_eglGlobal.Displays);
+   assert(dpy);
+   assert(!dpy->Handle);
+   dpy->Handle = (EGLDisplay) key;
+   assert(dpy->Handle);
+   _eglHashInsert(_eglGlobal.Displays, key, dpy);
 }
 
 
@@ -61,11 +97,25 @@ _eglGetCurrentDisplay(void)
 }
 
 
+/**
+ * Free all the data hanging of an _EGLDisplay object, but not
+ * the object itself.
+ */
 void
 _eglCleanupDisplay(_EGLDisplay *disp)
 {
-   /* XXX incomplete */
+   EGLint i;
+
+   for (i = 0; i < disp->NumConfigs; i++) {
+      free(disp->Configs[i]);
+   }
    free(disp->Configs);
-   free(disp->Name);
-   /* driver deletes _EGLDisplay */
+   disp->Configs = NULL;
+
+   /* XXX incomplete */
+
+   free((void *) disp->DriverName);
+   disp->DriverName = NULL;
+
+   /* driver deletes the _EGLDisplay object */
 }
diff --git a/src/egl/main/egldisplay.h b/src/egl/main/egldisplay.h
index 1a03fdd4ad..ff623ee1c6 100644
--- a/src/egl/main/egldisplay.h
+++ b/src/egl/main/egldisplay.h
@@ -1,22 +1,31 @@
 #ifndef EGLDISPLAY_INCLUDED
 #define EGLDISPLAY_INCLUDED
 
+#ifdef _EGL_PLATFORM_X
+#include <X11/Xlib.h>
+#endif
 
 #include "egltypedefs.h"
 
 
 struct _egl_display 
 {
+   EGLNativeDisplayType NativeDisplay;
    EGLDisplay Handle;
 
-   char *Name;
+   const char *DriverName;
+   const char *DriverArgs;
    _EGLDriver *Driver;
 
    EGLint NumScreens;
    _EGLScreen **Screens;  /* array [NumScreens] */
 
    EGLint NumConfigs;
-   _EGLConfig *Configs;  /* array [NumConfigs] */
+   _EGLConfig **Configs;  /* array [NumConfigs] of ptr to _EGLConfig */
+
+#ifdef _EGL_PLATFORM_X
+   Display *Xdpy;
+#endif
 };
 
 
@@ -24,10 +33,18 @@ extern _EGLDisplay *
 _eglNewDisplay(NativeDisplayType displayName);
 
 
+EGLDisplay
+_eglGetDisplayHandle(_EGLDisplay *display);
+
+
 extern _EGLDisplay *
 _eglLookupDisplay(EGLDisplay dpy);
 
 
+extern void
+_eglSaveDisplay(_EGLDisplay *dpy);
+
+
 extern _EGLDisplay *
 _eglGetCurrentDisplay(void);
 
diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c
index bda06dd827..43b1f51903 100644
--- a/src/egl/main/egldriver.c
+++ b/src/egl/main/egldriver.c
@@ -1,112 +1,277 @@
+/**
+ * Functions for choosing and opening/loading device drivers.
+ */
+
+
 #include <assert.h>
-#include <dlfcn.h>
-#include <stdio.h>
 #include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
 #include "eglconfig.h"
 #include "eglcontext.h"
+#include "egldefines.h"
 #include "egldisplay.h"
 #include "egldriver.h"
 #include "eglglobals.h"
 #include "egllog.h"
+#include "eglmisc.h"
 #include "eglmode.h"
 #include "eglscreen.h"
+#include "eglstring.h"
 #include "eglsurface.h"
 
+#if defined(_EGL_PLATFORM_X)
+#include <dlfcn.h>
+#include "eglx.h"
+#elif defined(_EGL_PLATFORM_WINDOWS)
+/* Use static linking on Windows for now */
+#define WINDOWS_STATIC_LINK
+#endif
+
+/**
+ * Wrappers for dlopen/dlclose()
+ */
+#if defined(_EGL_PLATFORM_WINDOWS)
+#ifdef WINDOWS_STATIC_LINK
+   static const char *DefaultDriverName = "Windows EGL Static Library";
+#else
+   /* XXX Need to decide how to do dynamic name lookup on Windows */
+   static const char *DefaultDriverName = "TBD";
+#endif
+   static const char *SysFS = NULL;
+   typedef HMODULE lib_handle;
+
+   static HMODULE
+   open_library(const char *filename)
+   {
+#ifdef WINDOWS_STATIC_LINK
+      return 0;
+#else
+      return LoadLibrary(filename);
+#endif
+   }
+
+   static void
+   close_library(HMODULE lib)
+   {
+#ifdef WINDOWS_STATIC_LINK
+#else
+      FreeLibrary(lib);
+#endif
+   }
+
+#elif defined(_EGL_PLATFORM_X)
+   static const char *DefaultDriverName = ":0";
+   static const char *SysFS = "/sys/class";
+
+   typedef void * lib_handle;
+
+   static void *
+   open_library(const char *filename)
+   {
+      return dlopen(filename, RTLD_LAZY);
+   }
+
+   static void
+   close_library(void *lib)
+   {
+      dlclose(lib);
+   }
+   
+#endif
+
+/**
+ * Given a card number, use sysfs to determine the DRI driver name.
+ */
+const char *
+_eglChooseDRMDriver(int card)
+{
+#if 0
+   return _eglstrdup("libEGLdri");
+#else
+   char path[2000], driverName[2000];
+   FILE *f;
+   int length;
+
+   snprintf(path, sizeof(path), "%s/drm/card%d/dri_library_name", SysFS, card);
+
+   f = fopen(path, "r");
+   if (!f)
+      return NULL;
 
-const char *DefaultDriverName = "demodriver";
+   fgets(driverName, sizeof(driverName), f);
+   fclose(f);
 
+   if ((length = strlen(driverName)) > 1) {
+      /* remove the trailing newline from sysfs */
+      driverName[length - 1] = '\0';
+      strncat(driverName, "_dri", sizeof(driverName));
+      return _eglstrdup(driverName);
+   }
+   else {
+      return NULL;
+   }   
+#endif
+}
 
 /**
- * Choose and open/init the hardware driver for the given EGLDisplay.
- * Previously, the EGLDisplay was created with _eglNewDisplay() where
- * we recorded the user's NativeDisplayType parameter.
+ * XXX this function is totally subject change!!!
+ *
+ *
+ * Determine/return the name of the driver to use for the given _EGLDisplay.
  *
- * Now we'll use the NativeDisplayType value.
+ * Try to be clever and determine if nativeDisplay is an Xlib Display
+ * ptr or a string (naming a driver or screen number, etc).
  *
- * Currently, the native display value is treated as a string.
  * If the first character is ':' we interpret it as a screen or card index
  * number (i.e. ":0" or ":1", etc)
  * Else if the first character is '!' we interpret it as specific driver name
  * (i.e. "!r200" or "!i830".
+ *
+ * Whatever follows ':' is copied and put into dpy->DriverArgs.
+ *
+ * The caller may free() the returned string.
  */
-_EGLDriver *
-_eglChooseDriver(EGLDisplay display)
+const char *
+_eglChooseDriver(_EGLDisplay *dpy)
 {
-   _EGLDisplay *dpy = _eglLookupDisplay(display);
-   _EGLDriver *drv;
-   const char *driverName = DefaultDriverName;
-   const char *name;
+   /* Under Windows, the NativeDisplay is an HDC handle, therefore */
+   /* it can't be interpreted as a string or a pointer. */
+#if defined(_EGL_PLATFORM_WINDOWS)
+   const char *displayString = NULL;
+#else
+   const char *displayString = (const char *) dpy->NativeDisplay;
+#endif
+   const char *driverName = NULL;
 
-   assert(dpy);
+   (void) DefaultDriverName;
 
-   name = dpy->Name;
-   if (!name) {
-      /* use default */
+#if defined(_EGL_PLATFORM_X)
+   /* First, if the EGL_DRIVER env var is set, use that */
+   driverName = getenv("EGL_DRIVER");
+   if (driverName)
+      return _eglstrdup(driverName);
+#endif
+
+#if 0
+   if (!displayString) {
+      /* choose a default */
+      displayString = DefaultDriverName;
    }
-   else if (name[0] == ':' && (name[1] >= '0' && name[1] <= '9') && !name[2]) {
-      /* XXX probe hardware here to determine which driver to open */
-      driverName = "libEGLdri";
+#endif
+   /* extract default DriverArgs = whatever follows ':' */
+   if (displayString &&
+       (displayString[0] == '!' ||
+        displayString[0] == ':')) {
+      const char *args = strchr(displayString, ':');
+      if (args)
+         dpy->DriverArgs = _eglstrdup(args + 1);
    }
-   else if (name[0] == '!') {
-      /* use specified driver name */
-      driverName = name + 1;
+
+   /* determine driver name now */
+   if (displayString && displayString[0] == ':' &&
+       (displayString[1] >= '0' && displayString[1] <= '9') &&
+       !displayString[2]) {
+      int card = atoi(displayString + 1);
+      driverName = _eglChooseDRMDriver(card);
    }
-   else {
-      /* Maybe display was returned by XOpenDisplay? */
-      _eglLog(_EGL_FATAL, "eglChooseDriver() bad name");
+   else if (displayString && displayString[0] == '!') {
+      /* use user-specified driver name */
+      driverName = _eglstrdup(displayString + 1);
+      /* truncate driverName at ':' if present */
+      {
+         char *args = strchr(driverName, ':');
+         if (args) {
+            *args = 0;
+         }
+      }
+   }
+   else 
+   {
+      /* NativeDisplay is not a string! */
+#if defined(_EGL_PLATFORM_X)
+      driverName = _xeglChooseDriver(dpy);
+#else
+      driverName = DefaultDriverName;
+#endif
    }
 
-   _eglLog(_EGL_INFO, "eglChooseDriver() choosing %s", driverName);
-
-   drv = _eglOpenDriver(dpy, driverName);
-   dpy->Driver = drv;
-
-   return drv;
+   return driverName;
 }
 
 
 /**
  * Open/load the named driver and call its bootstrap function: _eglMain().
+ * By the time this function is called, the dpy->DriverName should have
+ * been determined.
+ *
  * \return  new _EGLDriver object.
  */
 _EGLDriver *
-_eglOpenDriver(_EGLDisplay *dpy, const char *driverName)
+_eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args)
 {
    _EGLDriver *drv;
    _EGLMain_t mainFunc;
-   void *lib;
+   lib_handle lib;
    char driverFilename[1000];
 
+   assert(driverName);
+
+#if defined(_EGL_PLATFORM_WINDOWS)
+/* Use static linking on Windows for now */
+#ifdef WINDOWS_STATIC_LINK
+   lib = 0;
+   mainFunc = (_EGLMain_t)_eglMain;
+#else
+   /* XXX untested */
+   sprintf(driverFilename, "%s.dll", driverName);
+   _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename);
+   lib = open_library(driverFilename);
+   if (!lib) {
+      _eglLog(_EGL_WARNING, "Could not open %s",
+              driverFilename);
+      return NULL;
+   }
+   mainFunc = (_EGLMain_t) GetProcAddress(lib, "_eglMain");
+#endif
+#elif defined(_EGL_PLATFORM_X)
    /* XXX also prepend a directory path??? */
    sprintf(driverFilename, "%s.so", driverName);
-
    _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename);
-   lib = dlopen(driverFilename, RTLD_NOW);
+   lib = open_library(driverFilename);
    if (!lib) {
       _eglLog(_EGL_WARNING, "Could not open %s (%s)",
               driverFilename, dlerror());
       return NULL;
    }
-
    mainFunc = (_EGLMain_t) dlsym(lib, "_eglMain");
+#endif
+
    if (!mainFunc) {
       _eglLog(_EGL_WARNING, "_eglMain not found in %s", driverFilename);
-      dlclose(lib);
+      close_library(lib);
       return NULL;
    }
 
-   drv = mainFunc(dpy);
+   drv = mainFunc(dpy, args);
    if (!drv) {
-      dlclose(lib);
+      close_library(lib);
       return NULL;
    }
+
    /* with a recurvise open you want the inner most handle */
-   if (!drv->LibHandle)
+   if (!drv->LibHandle) {
       drv->LibHandle = lib;
-   else
-      dlclose(lib);
+   }
+   else {
+      close_library(lib);
+   }
+
+   /* update the global notion of supported APIs */
+   _eglGlobal.ClientAPIsMask |= drv->ClientAPIsMask;
+
+   _eglSaveDriver(drv);
 
-   drv->Display = dpy;
    return drv;
 }
 
@@ -117,19 +282,31 @@ _eglCloseDriver(_EGLDriver *drv, EGLDisplay dpy)
    void *handle = drv->LibHandle;
    EGLBoolean b;
 
-   _eglLog(_EGL_INFO, "Closing driver");
+   _eglLog(_EGL_DEBUG, "Closing %s", drv->Name);
 
    /*
     * XXX check for currently bound context/surfaces and delete them?
     */
 
    b = drv->API.Terminate(drv, dpy);
-   dlclose(handle);
+
+   close_library(handle);
+
    return b;
 }
 
 
 /**
+ * Save the given driver pointer in the list of all known drivers.
+ */
+void
+_eglSaveDriver(_EGLDriver *drv)
+{
+   _eglGlobal.Drivers[ _eglGlobal.NumDrivers++ ] = drv;
+}
+
+
+/**
  * Given a display handle, return the _EGLDriver for that display.
  */
 _EGLDriver *
@@ -201,72 +378,48 @@ _eglInitDriverFallbacks(_EGLDriver *drv)
 }
 
 
+
 /**
- * Examine the individual extension enable/disable flags and recompute
- * the driver's Extensions string.
+ * Try to determine which EGL APIs (OpenGL, OpenGL ES, OpenVG, etc)
+ * are supported on the system by looking for standard library names.
  */
-static void
-_eglUpdateExtensionsString(_EGLDriver *drv)
-{
-   drv->Extensions.String[0] = 0;
-
-   if (drv->Extensions.MESA_screen_surface)
-      strcat(drv->Extensions.String, "EGL_MESA_screen_surface ");
-   if (drv->Extensions.MESA_copy_context)
-      strcat(drv->Extensions.String, "EGL_MESA_copy_context ");
-   assert(strlen(drv->Extensions.String) < MAX_EXTENSIONS_LEN);
-}
-
-
-
-const char *
-_eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name)
+EGLint
+_eglFindAPIs(void)
 {
-   (void) drv;
-   (void) dpy;
-   switch (name) {
-   case EGL_VENDOR:
-      return "Mesa Project";
-   case EGL_VERSION:
-      return "1.0";
-   case EGL_EXTENSIONS:
-      _eglUpdateExtensionsString(drv);
-      return drv->Extensions.String;
-#ifdef EGL_VERSION_1_2
-   case EGL_CLIENT_APIS:
-      /* XXX need to initialize somewhere */
-      return drv->ClientAPIs;
+   EGLint mask = 0x0;
+   lib_handle lib;
+#if defined(_EGL_PLATFORM_WINDOWS)
+   /* XXX not sure about these names */
+   const char *es1_libname = "libGLESv1_CM.dll";
+   const char *es2_libname = "libGLESv2.dll";
+   const char *gl_libname = "OpenGL32.dll";
+   const char *vg_libname = "libOpenVG.dll";
+#elif defined(_EGL_PLATFORM_X)
+   const char *es1_libname = "libGLESv1_CM.so";
+   const char *es2_libname = "libGLESv2.so";
+   const char *gl_libname = "libGL.so";
+   const char *vg_libname = "libOpenVG.so";
 #endif
-   default:
-      _eglError(EGL_BAD_PARAMETER, "eglQueryString");
-      return NULL;
-   }
-}
 
+   if ((lib = open_library(es1_libname))) {
+      close_library(lib);
+      mask |= EGL_OPENGL_ES_BIT;
+   }
 
-EGLBoolean
-_eglWaitGL(_EGLDriver *drv, EGLDisplay dpy)
-{
-   /* just a placeholder */
-   (void) drv;
-   (void) dpy;
-   return EGL_TRUE;
-}
+   if ((lib = open_library(es2_libname))) {
+      close_library(lib);
+      mask |= EGL_OPENGL_ES2_BIT;
+   }
 
+   if ((lib = open_library(gl_libname))) {
+      close_library(lib);
+      mask |= EGL_OPENGL_BIT;
+   }
 
-EGLBoolean
-_eglWaitNative(_EGLDriver *drv, EGLDisplay dpy, EGLint engine)
-{
-   /* just a placeholder */
-   (void) drv;
-   (void) dpy;
-   switch (engine) {
-   case EGL_CORE_NATIVE_ENGINE:
-      break;
-   default:
-      _eglError(EGL_BAD_PARAMETER, "eglWaitNative(engine)");
-      return EGL_FALSE;
+   if ((lib = open_library(vg_libname))) {
+      close_library(lib);
+      mask |= EGL_OPENVG_BIT;
    }
 
-   return EGL_TRUE;
+   return mask;
 }
diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h
index 88526e973d..4066c6ec1d 100644
--- a/src/egl/main/egldriver.h
+++ b/src/egl/main/egldriver.h
@@ -4,9 +4,7 @@
 
 #include "egltypedefs.h"
 #include "eglapi.h"
-
-/* should probably use a dynamic-length string, but this will do */
-#define MAX_EXTENSIONS_LEN 1000
+#include "egldefines.h"
 
 
 /**
@@ -17,7 +15,7 @@ struct _egl_extensions
    EGLBoolean MESA_screen_surface;
    EGLBoolean MESA_copy_context;
 
-   char String[MAX_EXTENSIONS_LEN];
+   char String[_EGL_MAX_EXTENSIONS_LEN];
 };
 
 
@@ -26,37 +24,48 @@ struct _egl_extensions
  */
 struct _egl_driver
 {
-   EGLBoolean Initialized; /* set by driver after initialized */
+   EGLBoolean Initialized; /**< set by driver after initialized */
+
+   void *LibHandle; /**< dlopen handle */
 
-   void *LibHandle; /* dlopen handle */
+   const char *Name;  /**< name of this driver */
 
-   _EGLDisplay *Display;
+   int APImajor, APIminor; /**< as returned by eglInitialize() */
+   char Version[1000];       /**< initialized from APImajor/minor, Name */
 
-   int ABIversion;
-   int APImajor, APIminor; /* returned through eglInitialize */
-   const char *ClientAPIs;
+   /** Bitmask of supported APIs (EGL_xx_BIT) set by the driver during init */
+   EGLint ClientAPIsMask;
 
-   _EGLAPI API;
+   _EGLAPI API;  /**< EGL API dispatch table */
 
    _EGLExtensions Extensions;
+
+   int LargestPbuffer;
 };
 
 
-extern _EGLDriver *_eglMain(_EGLDisplay *dpy);
+extern _EGLDriver *_eglMain(_EGLDisplay *dpy, const char *args);
 
 
-extern _EGLDriver *
-_eglChooseDriver(EGLDisplay dpy);
+extern const char *
+_eglChooseDRMDriver(int card);
+
+extern const char *
+_eglChooseDriver(_EGLDisplay *dpy);
 
 
 extern _EGLDriver *
-_eglOpenDriver(_EGLDisplay *dpy, const char *driverName);
+_eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args);
 
 
 extern EGLBoolean
 _eglCloseDriver(_EGLDriver *drv, EGLDisplay dpy);
 
 
+extern void
+_eglSaveDriver(_EGLDriver *drv);
+
+
 extern _EGLDriver *
 _eglLookupDriver(EGLDisplay d);
 
@@ -65,17 +74,8 @@ extern void
 _eglInitDriverFallbacks(_EGLDriver *drv);
 
 
-extern const char *
-_eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name);
-
-
-extern EGLBoolean
-_eglWaitGL(_EGLDriver *drv, EGLDisplay dpy);
-
-
-extern EGLBoolean
-_eglWaitNative(_EGLDriver *drv, EGLDisplay dpy, EGLint engine);
-
+extern EGLint
+_eglFindAPIs(void);
 
 
 #endif /* EGLDRIVER_INCLUDED */
diff --git a/src/egl/main/eglglobals.c b/src/egl/main/eglglobals.c
index 608311d749..b770e55dbd 100644
--- a/src/egl/main/eglglobals.c
+++ b/src/egl/main/eglglobals.c
@@ -2,9 +2,10 @@
 #include <stdlib.h>
 #include "eglglobals.h"
 
-
-struct _egl_global _eglGlobal = { .Initialized = EGL_FALSE };
-
+struct _egl_global _eglGlobal = 
+{
+   EGL_FALSE
+};
 
 /**
  * Init the fields in the _eglGlobal struct
@@ -15,13 +16,11 @@ _eglInitGlobals(void)
 {
    if (!_eglGlobal.Initialized) {
       _eglGlobal.Displays = _eglNewHashTable();
-      _eglGlobal.Contexts = _eglNewHashTable();
       _eglGlobal.Surfaces = _eglNewHashTable();
       _eglGlobal.FreeScreenHandle = 1;
       _eglGlobal.Initialized = EGL_TRUE;
 
-      _eglGlobal.OpenGLESAPISupported = EGL_TRUE;
-      _eglGlobal.OpenVGAPISupported = EGL_FALSE;
+      _eglGlobal.ClientAPIsMask = 0x0;
 
       /* XXX temporary */
       _eglGlobal.ThreadInfo = _eglNewThreadInfo();
@@ -37,7 +36,6 @@ _eglDestroyGlobals(void)
 {
    /* XXX TODO walk over table entries, deleting each */
    _eglDeleteHashTable(_eglGlobal.Displays);
-   _eglDeleteHashTable(_eglGlobal.Contexts);
    _eglDeleteHashTable(_eglGlobal.Surfaces);
 }
 
@@ -52,7 +50,7 @@ _eglNewThreadInfo(void)
    if (t) {
       t->CurrentContext = EGL_NO_CONTEXT;
       t->LastError = EGL_SUCCESS;
-      t->CurrentAPI = EGL_NONE;
+      t->CurrentAPI = EGL_OPENGL_ES_API;  /* default, per EGL spec */
    }
    return t;
 }
diff --git a/src/egl/main/eglglobals.h b/src/egl/main/eglglobals.h
index c16baa2d6b..14d8ea487a 100644
--- a/src/egl/main/eglglobals.h
+++ b/src/egl/main/eglglobals.h
@@ -24,17 +24,20 @@ struct _egl_global
    EGLBoolean Initialized;
 
    _EGLHashtable *Displays;
-   _EGLHashtable *Contexts;
    _EGLHashtable *Surfaces;
 
    EGLScreenMESA FreeScreenHandle;
 
-   /* XXX these may be temporary */
-   EGLBoolean OpenGLESAPISupported;
-   EGLBoolean OpenVGAPISupported;
+   /* bitmaks of supported APIs (supported by _some_ driver) */
+   EGLint ClientAPIsMask;
+
+   char ClientAPIs[1000];   /**< updated by eglQueryString */
 
    /* XXX temporary - should be thread-specific data (TSD) */
    _EGLThreadInfo *ThreadInfo;
+
+   EGLint NumDrivers;
+   _EGLDriver *Drivers[10];
 };
 
 
diff --git a/src/egl/main/egllog.c b/src/egl/main/egllog.c
index dc1daaa996..1d7a0a388c 100644
--- a/src/egl/main/egllog.c
+++ b/src/egl/main/egllog.c
@@ -1,5 +1,7 @@
 /**
  * Logging facility for debug/info messages.
+ * _EGL_FATAL messages are printed to stderr
+ * The EGL_LOG_LEVEL var controls the output of other warning/info/debug msgs.
  */
 
 
@@ -10,37 +12,41 @@
 #include "egllog.h"
 
 #define MAXSTRING 1000
-#define FALLBACK_LOG_LEVEL      _EGL_DEBUG
-#define FALLBACK_LOG_LEVEL_STR  "debug"
+#define FALLBACK_LOG_LEVEL      _EGL_WARNING
+#define FALLBACK_LOG_LEVEL_STR  "warning"
 
 static EGLint ReportingLevel = -1;
 
 
 static void
-log_level_initialize (void)
+log_level_initialize(void)
 {
-   char *log_env = getenv ("EGL_LOG_LEVEL");
+#if defined(_EGL_PLATFORM_X)  
+   char *log_env = getenv("EGL_LOG_LEVEL");
+#else
+   char *log_env = NULL;
+#endif
 
    if (log_env == NULL) {
       ReportingLevel = FALLBACK_LOG_LEVEL;
    }
-   else if (strcasecmp (log_env, "fatal") == 0) {
+   else if (strcasecmp(log_env, "fatal") == 0) {
       ReportingLevel = _EGL_FATAL;
    }
-   else if (strcasecmp (log_env, "warning") == 0) {
+   else if (strcasecmp(log_env, "warning") == 0) {
       ReportingLevel = _EGL_WARNING;
    }
-   else if (strcasecmp (log_env, "info") == 0) {
+   else if (strcasecmp(log_env, "info") == 0) {
       ReportingLevel = _EGL_INFO;
    }
-   else if (strcasecmp (log_env, "debug") == 0) {
+   else if (strcasecmp(log_env, "debug") == 0) {
       ReportingLevel = _EGL_DEBUG;
    }
    else {
-      fprintf (stderr, "Unrecognized EGL_LOG_LEVEL environment variable value. "
-               "Expected one of \"fatal\", \"warning\", \"info\", \"debug\". "
-               "Got \"%s\". Falling back to \"%s\".\n",
-               log_env, FALLBACK_LOG_LEVEL_STR);
+      fprintf(stderr, "Unrecognized EGL_LOG_LEVEL environment variable value. "
+              "Expected one of \"fatal\", \"warning\", \"info\", \"debug\". "
+              "Got \"%s\". Falling back to \"%s\".\n",
+              log_env, FALLBACK_LOG_LEVEL_STR);
       ReportingLevel = FALLBACK_LOG_LEVEL;
    }
 }
@@ -59,7 +65,7 @@ _eglLog(EGLint level, const char *fmtStr, ...)
    static int log_level_initialized = 0;
 
    if (!log_level_initialized) {
-      log_level_initialize ();
+      log_level_initialize();
       log_level_initialized = 1;
    }
 
@@ -85,7 +91,7 @@ _eglLog(EGLint level, const char *fmtStr, ...)
       vsnprintf(msg, MAXSTRING, fmtStr, args);
       va_end(args);
 
-      fprintf(stderr, "EGL %s: %s\n", levelStr, msg);
+      fprintf(stderr, "libEGL %s: %s\n", levelStr, msg);
 
       if (level == _EGL_FATAL) {
          exit(1); /* or abort()? */
diff --git a/src/egl/main/eglmisc.c b/src/egl/main/eglmisc.c
new file mode 100644
index 0000000000..b5bdc3ea4b
--- /dev/null
+++ b/src/egl/main/eglmisc.c
@@ -0,0 +1,129 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Small/misc EGL functions
+ */
+
+
+#include <assert.h>
+#include <string.h>
+#include "eglglobals.h"
+#include "eglmisc.h"
+
+
+/**
+ * Examine the individual extension enable/disable flags and recompute
+ * the driver's Extensions string.
+ */
+static void
+_eglUpdateExtensionsString(_EGLDriver *drv)
+{
+   drv->Extensions.String[0] = 0;
+
+   if (drv->Extensions.MESA_screen_surface)
+      strcat(drv->Extensions.String, "EGL_MESA_screen_surface ");
+   if (drv->Extensions.MESA_copy_context)
+      strcat(drv->Extensions.String, "EGL_MESA_copy_context ");
+   assert(strlen(drv->Extensions.String) < _EGL_MAX_EXTENSIONS_LEN);
+}
+
+
+static void
+_eglUpdateAPIsString(_EGLDriver *drv)
+{
+   _eglGlobal.ClientAPIs[0] = 0;
+
+   if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_BIT)
+      strcat(_eglGlobal.ClientAPIs, "OpenGL ");
+
+   if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_ES_BIT)
+      strcat(_eglGlobal.ClientAPIs, "OpenGL_ES ");
+
+   if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_ES2_BIT)
+      strcat(_eglGlobal.ClientAPIs, "OpenGL_ES2 ");
+
+   if (_eglGlobal.ClientAPIsMask & EGL_OPENVG_BIT)
+      strcat(_eglGlobal.ClientAPIs, "OpenVG ");
+
+   assert(strlen(_eglGlobal.ClientAPIs) < sizeof(_eglGlobal.ClientAPIs));
+}
+
+
+
+const char *
+_eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name)
+{
+   (void) drv;
+   (void) dpy;
+   switch (name) {
+   case EGL_VENDOR:
+      return _EGL_VENDOR_STRING;
+   case EGL_VERSION:
+      return drv->Version;
+   case EGL_EXTENSIONS:
+      _eglUpdateExtensionsString(drv);
+      return drv->Extensions.String;
+#ifdef EGL_VERSION_1_2
+   case EGL_CLIENT_APIS:
+      _eglUpdateAPIsString(drv);
+      return _eglGlobal.ClientAPIs;
+#endif
+   default:
+      _eglError(EGL_BAD_PARAMETER, "eglQueryString");
+      return NULL;
+   }
+}
+
+
+EGLBoolean
+_eglWaitGL(_EGLDriver *drv, EGLDisplay dpy)
+{
+   /* just a placeholder */
+   (void) drv;
+   (void) dpy;
+   return EGL_TRUE;
+}
+
+
+EGLBoolean
+_eglWaitNative(_EGLDriver *drv, EGLDisplay dpy, EGLint engine)
+{
+   /* just a placeholder */
+   (void) drv;
+   (void) dpy;
+   switch (engine) {
+   case EGL_CORE_NATIVE_ENGINE:
+      break;
+   default:
+      _eglError(EGL_BAD_PARAMETER, "eglWaitNative(engine)");
+      return EGL_FALSE;
+   }
+
+   return EGL_TRUE;
+}
diff --git a/src/egl/main/eglmisc.h b/src/egl/main/eglmisc.h
new file mode 100644
index 0000000000..4e2a40ea99
--- /dev/null
+++ b/src/egl/main/eglmisc.h
@@ -0,0 +1,47 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef EGLMISC_INCLUDED
+#define EGLMISC_INCLUDED
+
+#include "egldriver.h"
+
+
+extern const char *
+_eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name);
+
+
+extern EGLBoolean
+_eglWaitGL(_EGLDriver *drv, EGLDisplay dpy);
+
+
+extern EGLBoolean
+_eglWaitNative(_EGLDriver *drv, EGLDisplay dpy, EGLint engine);
+
+
+#endif /* EGLMISC_INCLUDED */
diff --git a/src/egl/main/eglmode.h b/src/egl/main/eglmode.h
index e70da85759..52d4875676 100644
--- a/src/egl/main/eglmode.h
+++ b/src/egl/main/eglmode.h
@@ -4,6 +4,9 @@
 #include "egltypedefs.h"
 
 
+#define EGL_NO_MODE_MESA 0
+
+
 /**
  * Data structure which corresponds to an EGLModeMESA.
  */
diff --git a/src/egl/main/eglstring.c b/src/egl/main/eglstring.c
new file mode 100644
index 0000000000..ba7406158c
--- /dev/null
+++ b/src/egl/main/eglstring.c
@@ -0,0 +1,24 @@
+/**
+ * String utils.
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "eglstring.h"
+
+
+char *
+_eglstrdup(const char *s)
+{
+   if (s) {
+      int l = strlen(s);
+      char *s2 = malloc(l + 1);
+      if (s2)
+         strcpy(s2, s);
+      return s2;
+   }
+   return NULL;
+}
+
+
+
diff --git a/src/egl/main/eglstring.h b/src/egl/main/eglstring.h
new file mode 100644
index 0000000000..10468636e8
--- /dev/null
+++ b/src/egl/main/eglstring.h
@@ -0,0 +1,9 @@
+#ifndef EGLSTRING_INCLUDED
+#define EGLSTRING_INCLUDED
+
+
+extern char *
+_eglstrdup(const char *s);
+
+
+#endif /* EGLSTRING_INCLUDED */
diff --git a/src/egl/main/eglsurface.c b/src/egl/main/eglsurface.c
index 874f318e96..6905acac50 100644
--- a/src/egl/main/eglsurface.c
+++ b/src/egl/main/eglsurface.c
@@ -8,6 +8,7 @@
 #include <string.h>
 #include "eglcontext.h"
 #include "eglconfig.h"
+#include "egldriver.h"
 #include "eglglobals.h"
 #include "eglhash.h"
 #include "egllog.h"
@@ -60,6 +61,12 @@ _eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
       return EGL_FALSE;
    }
 
+   if ((GET_CONFIG_ATTRIB(conf, EGL_SURFACE_TYPE) & type) == 0) {
+      /* The config can't be used to create a surface of this type */
+      _eglError(EGL_BAD_CONFIG, func);
+      return EGL_FALSE;
+   }
+
    /*
     * Parse attribute list.  Different kinds of surfaces support different
     * attributes.
@@ -175,7 +182,7 @@ _eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
       }
    }
 
-   if (width <= 0 || height <= 0) {
+   if (width < 0 || height < 0) {
       _eglError(EGL_BAD_ATTRIBUTE, func);
       return EGL_FALSE;
    }
@@ -207,25 +214,47 @@ _eglInitSurface(_EGLDriver *drv, EGLDisplay dpy,
 void
 _eglSaveSurface(_EGLSurface *surf)
 {
+   EGLuint key = _eglHashGenKey(_eglGlobal.Surfaces);
    assert(surf);
    assert(!surf->Handle);
-   surf->Handle = _eglHashGenKey(_eglGlobal.Contexts);
+   surf->Handle = (EGLSurface) key;
    assert(surf->Handle);
-   _eglHashInsert(_eglGlobal.Surfaces, surf->Handle, surf);
+   _eglHashInsert(_eglGlobal.Surfaces, key, surf);
 }
 
 
 void
 _eglRemoveSurface(_EGLSurface *surf)
 {
-   _eglHashRemove(_eglGlobal.Surfaces, surf->Handle);
+   _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surf->Handle);
+}
+
+
+
+/**
+ * Return the public handle for an internal _EGLSurface.
+ * This is the inverse of _eglLookupSurface().
+ */
+EGLSurface
+_eglGetSurfaceHandle(_EGLSurface *surface)
+{
+   if (surface)
+      return surface->Handle;
+   else
+      return EGL_NO_SURFACE;
 }
 
 
+/**
+ * Return the private _EGLSurface which corresponds to a public EGLSurface
+ * handle.
+ * This is the inverse of _eglGetSurfaceHandle().
+ */
 _EGLSurface *
 _eglLookupSurface(EGLSurface surf)
 {
-   _EGLSurface *c = (_EGLSurface *) _eglHashLookup(_eglGlobal.Surfaces, surf);
+   _EGLSurface *c = (_EGLSurface *) _eglHashLookup(_eglGlobal.Surfaces,
+                                                   (EGLuint) surf);
    return c;
 }
 
@@ -254,12 +283,7 @@ _eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
    /* Basically just do error checking here.  Drivers have to do the
     * actual buffer swap.
     */
-   _EGLContext *context = _eglGetCurrentContext();
    _EGLSurface *surface = _eglLookupSurface(draw);
-   if (context && context->DrawSurface != surface) {
-      _eglError(EGL_BAD_SURFACE, "eglSwapBuffers");
-      return EGL_FALSE;
-   }
    if (surface == NULL) {
       _eglError(EGL_BAD_SURFACE, "eglSwapBuffers");
       return EGL_FALSE;
@@ -297,7 +321,9 @@ _eglQuerySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
    case EGL_CONFIG_ID:
       *value = GET_CONFIG_ATTRIB(surface->Config, EGL_CONFIG_ID);
       return EGL_TRUE;
-   /*XXX case EGL_LARGEST_PBUFFER:*/
+   case EGL_LARGEST_PBUFFER:
+      *value = drv->LargestPbuffer;
+      return EGL_TRUE;
    case EGL_SURFACE_TYPE:
       *value = surface->Type;
       return EGL_TRUE;
@@ -439,7 +465,7 @@ _eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
 {
    _EGLSurface *surf = _eglLookupSurface(surface);
    if (surf) {
-      _eglHashRemove(_eglGlobal.Surfaces, surface);
+      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
       if (surf->IsBound) {
          surf->DeletePending = EGL_TRUE;
       }
@@ -459,7 +485,8 @@ _eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
  * Default fallback routine - drivers might override this.
  */
 EGLBoolean
-_eglSurfaceAttrib(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf, EGLint attribute, EGLint value)
+_eglSurfaceAttrib(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
+                  EGLint attribute, EGLint value)
 {
    _EGLSurface *surface = _eglLookupSurface(surf);
 
@@ -481,18 +508,67 @@ _eglSurfaceAttrib(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf, EGLint attri
 
 
 EGLBoolean
-_eglBindTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, EGLint buffer)
+_eglBindTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
+                 EGLint buffer)
 {
-   /* XXX unfinished */
-   return EGL_FALSE;
+   /* Just do basic error checking and return success/fail.
+    * Drivers must implement the real stuff.
+    */
+   _EGLSurface *surface = _eglLookupSurface(surf);
+
+   if (!surface || surface->Type != EGL_PBUFFER_BIT) {
+      _eglError(EGL_BAD_SURFACE, "eglBindTexImage");
+      return EGL_FALSE;
+   }
+
+   if (surface->TextureFormat == EGL_NO_TEXTURE) {
+      _eglError(EGL_BAD_MATCH, "eglBindTexImage");
+      return EGL_FALSE;
+   }
+
+   if (buffer != EGL_BACK_BUFFER) {
+      _eglError(EGL_BAD_PARAMETER, "eglBindTexImage");
+      return EGL_FALSE;
+   }
+
+   surface->BoundToTexture = EGL_TRUE;
+
+   return EGL_TRUE;
 }
 
 
 EGLBoolean
-_eglReleaseTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface, EGLint buffer)
+_eglReleaseTexImage(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf,
+                    EGLint buffer)
 {
-   /* XXX unfinished */
-   return EGL_FALSE;
+   /* Just do basic error checking and return success/fail.
+    * Drivers must implement the real stuff.
+    */
+   _EGLSurface *surface = _eglLookupSurface(surf);
+
+   if (!surface || surface->Type != EGL_PBUFFER_BIT) {
+      _eglError(EGL_BAD_SURFACE, "eglBindTexImage");
+      return EGL_FALSE;
+   }
+
+   if (surface->TextureFormat == EGL_NO_TEXTURE) {
+      _eglError(EGL_BAD_MATCH, "eglBindTexImage");
+      return EGL_FALSE;
+   }
+
+   if (buffer != EGL_BACK_BUFFER) {
+      _eglError(EGL_BAD_PARAMETER, "eglReleaseTexImage");
+      return EGL_FALSE;
+   }
+
+   if (!surface->BoundToTexture) {
+      _eglError(EGL_BAD_SURFACE, "eglReleaseTexImage");
+      return EGL_FALSE;
+   }
+
+   surface->BoundToTexture = EGL_FALSE;
+
+   return EGL_TRUE;
 }
 
 
diff --git a/src/egl/main/eglsurface.h b/src/egl/main/eglsurface.h
index 79abeca0b2..50f965b5cb 100644
--- a/src/egl/main/eglsurface.h
+++ b/src/egl/main/eglsurface.h
@@ -16,6 +16,7 @@ struct _egl_surface
    /* May need reference counting here */
    EGLBoolean IsBound;
    EGLBoolean DeletePending;
+   EGLBoolean BoundToTexture;
 
    EGLint Type; /* one of EGL_WINDOW_BIT, EGL_PIXMAP_BIT or EGL_PBUFFER_BIT */
    EGLint Width, Height;
@@ -51,6 +52,10 @@ extern void
 _eglRemoveSurface(_EGLSurface *surf);
 
 
+extern EGLSurface
+_eglGetSurfaceHandle(_EGLSurface *surface);
+
+
 extern _EGLSurface *
 _eglLookupSurface(EGLSurface surf);
  
diff --git a/src/egl/main/egltypedefs.h b/src/egl/main/egltypedefs.h
index fa8cb496f8..9fbc55352c 100644
--- a/src/egl/main/egltypedefs.h
+++ b/src/egl/main/egltypedefs.h
@@ -1,8 +1,10 @@
 #ifndef EGLTYPEDEFS_INCLUDED
 #define EGLTYPEDEFS_INCLUDED
 
+#define EGL_EGLEXT_PROTOTYPES
 
-#include <GLES/egl.h>
+#include <EGL/egl.h>
+#include <EGL/eglext.h>
 
 
 typedef struct _egl_api _EGLAPI;
@@ -26,9 +28,7 @@ typedef struct _egl_surface _EGLSurface;
 typedef struct _egl_thread_info _EGLThreadInfo;
 
 
-typedef void (*_EGLProc)();
-
-typedef _EGLDriver *(*_EGLMain_t)(_EGLDisplay *dpy);
+typedef _EGLDriver *(*_EGLMain_t)(_EGLDisplay *dpy, const char *args);
 
 
 #endif /* EGLTYPEDEFS_INCLUDED */
diff --git a/src/egl/main/eglx.c b/src/egl/main/eglx.c
new file mode 100644
index 0000000000..50acc3a24f
--- /dev/null
+++ b/src/egl/main/eglx.c
@@ -0,0 +1,100 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * X-specific EGL code.
+ *
+ * Any glue code needed to make EGL work with X is placed in this file.
+ */
+
+
+#include <assert.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <X11/Xlib.h>
+
+#include "egldriver.h"
+#include "egllog.h"
+#include "eglstring.h"
+#include "eglx.h"
+
+
+static const char *DefaultGLXDriver = "egl_glx";
+static const char *DefaultSoftDriver = "egl_softpipe";
+
+
+/**
+ * Given an X Display ptr (at dpy->Xdpy) try to determine the appropriate
+ * device driver.  Return its name.
+ *
+ * This boils down to whether to use the egl_glx.so driver which will
+ * load a DRI driver or the egl_softpipe.so driver that'll do software
+ * rendering on Xlib.
+ */
+const char *
+_xeglChooseDriver(_EGLDisplay *dpy)
+{
+#ifdef _EGL_PLATFORM_X
+   _XPrivDisplay xdpy;
+   int screen;
+   const char *driverName;
+
+   assert(dpy);
+
+   if (!dpy->Xdpy) {
+      dpy->Xdpy = XOpenDisplay(NULL);
+      if (!dpy->Xdpy) {
+         /* can't open X display -> can't use X-based driver */
+         return NULL;
+      }
+   }
+   xdpy = (_XPrivDisplay) dpy->Xdpy;
+
+   assert(dpy->Xdpy);
+
+   screen = DefaultScreen(dpy->Xdpy);
+
+   /* See if we can choose a DRI/DRM driver */
+   driverName = _eglChooseDRMDriver(screen);
+   if (driverName) {
+      free((void *) driverName);
+      driverName = _eglstrdup(DefaultGLXDriver);
+   }
+   else {
+      driverName = _eglstrdup(DefaultSoftDriver);
+   }
+
+   _eglLog(_EGL_DEBUG, "_xeglChooseDriver: %s", driverName);
+
+   return driverName;
+#else
+   return NULL;
+#endif
+}
+
+
diff --git a/src/egl/main/eglx.h b/src/egl/main/eglx.h
new file mode 100644
index 0000000000..4323d55838
--- /dev/null
+++ b/src/egl/main/eglx.h
@@ -0,0 +1,12 @@
+#ifndef EGLX_INCLUDED
+#define EGLX_INCLUDED
+
+
+#include "egldisplay.h"
+
+
+extern const char *
+_xeglChooseDriver(_EGLDisplay *dpy);
+
+
+#endif /* EGLX_INCLUDED */
diff --git a/src/gallium/Makefile b/src/gallium/Makefile
new file mode 100644
index 0000000000..875314f6c3
--- /dev/null
+++ b/src/gallium/Makefile
@@ -0,0 +1,26 @@
+TOP = ../..
+include $(TOP)/configs/current
+
+
+SUBDIRS = $(GALLIUM_DIRS)
+# Note winsys/ needs to be built after src/mesa
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	rm -f `find . -name \*.[oa]`
+	rm -f `find . -name depend`
+
+
+# Dummy install target
+install:
diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template
new file mode 100644
index 0000000000..655e949ca2
--- /dev/null
+++ b/src/gallium/Makefile.template
@@ -0,0 +1,64 @@
+# -*-makefile-*-
+
+
+# We still have a dependency on the "dri" buffer manager.  Most likely
+# the interface can be reused in non-dri environments, and also as a
+# frontend to simpler memory managers.
+#
+COMMON_SOURCES = 
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+          $(CPP_SOURCES:.cpp=.o) \
+	  $(ASM_SOURCES:.S=.o) 
+
+
+### Include directories
+INCLUDES = \
+	-I. \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/include \
+        $(DRIVER_INCLUDES)
+
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.cpp.o:
+	$(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES)  $< -o $@
+
+
+##### TARGETS #####
+
+default: depend symlinks lib$(LIBNAME).a
+
+
+lib$(LIBNAME).a: $(OBJECTS) Makefile $(TOP)/src/gallium/Makefile.template
+	$(TOP)/bin/mklib -o $(LIBNAME) -static $(OBJECTS) $(DRIVER_LIBS)
+
+
+depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \
+		$(ASM_SOURCES) 2> /dev/null
+
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` `find ../include`
+
+
+# Remove .o and backup files
+clean::
+	-rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS)
+	-rm -f depend depend.bak
+
+
+include depend
diff --git a/src/gallium/README.portability b/src/gallium/README.portability
new file mode 100644
index 0000000000..adecf4bb79
--- /dev/null
+++ b/src/gallium/README.portability
@@ -0,0 +1,109 @@
+	      CROSS-PLATFORM PORTABILITY GUIDELINES FOR GALLIUM3D 
+
+
+= General Considerations =
+
+The state tracker and winsys driver support a rather limited number of
+platforms. However, the pipe drivers are meant to run in a wide number of
+platforms. Hence the pipe drivers, the auxiliary modules, and all public
+headers in general, should strictly follow these guidelines to ensure
+
+
+= Compiler Support =
+
+* Include the p_compiler.h.
+
+* Don't use the 'inline' keyword, use the INLINE macro in p_compiler.h instead.
+
+* Cast explicitly when converting to integer types of smaller sizes.
+
+* Cast explicitly when converting between float, double and integral types.
+
+* Don't use named struct initializers.
+
+* Don't use variable number of macro arguments. Use static inline functions
+instead.
+
+* Don't use C99 features.
+
+= Standard Library =
+
+* Avoid including standard library headers. Most standard library functions are
+not available in Windows Kernel Mode. Use the appropriate p_*.h include.
+
+== Memory Allocation ==
+
+* Use MALLOC, CALLOC, FREE instead of the malloc, calloc, free functions.
+
+* Use align_pointer() function defined in u_memory.h for aligning pointers
+ in a portable way.
+
+== Debugging ==
+
+* Use the functions/macros in p_debug.h.
+
+* Don't include assert.h, call abort, printf, etc.
+
+
+= Code Style =
+
+== Inherantice in C ==
+
+The main thing we do is mimic inheritance by structure containment.
+
+Here's a silly made-up example:
+
+/* base class */
+struct buffer
+{
+  int size;
+  void (*validate)(struct buffer *buf);
+};
+
+/* sub-class of bufffer */
+struct texture_buffer
+{
+  struct buffer base;  /* the base class, MUST COME FIRST! */
+  int format;
+  int width, height;
+};
+
+
+Then, we'll typically have cast-wrapper functions to convert base-class 
+pointers to sub-class pointers where needed:
+
+static inline struct vertex_buffer *vertex_buffer(struct buffer *buf)
+{
+  return (struct vertex_buffer *) buf;
+}
+
+
+To create/init a sub-classed object:
+
+struct buffer *create_texture_buffer(int w, int h, int format)
+{
+  struct texture_buffer *t = malloc(sizeof(*t));
+  t->format = format;
+  t->width = w;
+  t->height = h;
+  t->base.size = w * h;
+  t->base.validate = tex_validate;
+  return &t->base;
+}
+
+Example sub-class method:
+
+void tex_validate(struct buffer *buf)
+{
+  struct texture_buffer *tb = texture_buffer(buf);
+  assert(tb->format);
+  assert(tb->width);
+  assert(tb->height);
+}
+
+
+Note that we typically do not use typedefs to make "class names"; we use
+'struct whatever' everywhere.
+
+Gallium's pipe_context and the subclassed psb_context, etc are prime examples 
+of this.  There's also many examples in Mesa and the Mesa state tracker.
diff --git a/src/gallium/SConscript b/src/gallium/SConscript
new file mode 100644
index 0000000000..0c632ac2b8
--- /dev/null
+++ b/src/gallium/SConscript
@@ -0,0 +1,34 @@
+import os
+
+Import('*')
+
+env = env.Clone()
+
+auxiliaries = []
+
+Export('auxiliaries')
+
+
+if llvm:
+	SConscript(['auxiliary/gallivm/SConscript'])
+
+SConscript([
+	# NOTE: order matters!
+	'auxiliary/util/SConscript',
+	'auxiliary/rtasm/SConscript',
+	'auxiliary/tgsi/SConscript',
+	'auxiliary/cso_cache/SConscript',
+	'auxiliary/translate/SConscript',
+	'auxiliary/draw/SConscript',
+	'auxiliary/pipebuffer/SConscript',
+	'auxiliary/indices/SConscript',
+])
+
+for driver in env['drivers']:
+	SConscript(os.path.join('drivers', driver, 'SConscript'))
+
+SConscript('state_trackers/python/SConscript')
+SConscript('state_trackers/glx/xlib/SConscript')
+
+if platform == 'windows':
+	SConscript('state_trackers/wgl/SConscript')
diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile
new file mode 100644
index 0000000000..eaa0f2fe4e
--- /dev/null
+++ b/src/gallium/auxiliary/Makefile
@@ -0,0 +1,20 @@
+TOP = ../../..
+include $(TOP)/configs/current
+
+
+SUBDIRS = $(GALLIUM_AUXILIARY_DIRS)
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	rm -f `find . -name \*.[oa]`
diff --git a/src/gallium/auxiliary/cso_cache/Makefile b/src/gallium/auxiliary/cso_cache/Makefile
new file mode 100644
index 0000000000..6bd6602088
--- /dev/null
+++ b/src/gallium/auxiliary/cso_cache/Makefile
@@ -0,0 +1,14 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = cso_cache
+
+C_SOURCES = \
+	cso_context.c \
+	cso_cache.c \
+	cso_hash.c
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/cso_cache/SConscript b/src/gallium/auxiliary/cso_cache/SConscript
new file mode 100644
index 0000000000..651e68a191
--- /dev/null
+++ b/src/gallium/auxiliary/cso_cache/SConscript
@@ -0,0 +1,11 @@
+Import('*')
+
+cso_cache = env.ConvenienceLibrary(
+	target = 'cso_cache',
+	source = [
+		'cso_context.c',
+		'cso_cache.c',
+		'cso_hash.c',
+	])
+
+auxiliaries.insert(0, cso_cache)
diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c
new file mode 100644
index 0000000000..6b1754ea00
--- /dev/null
+++ b/src/gallium/auxiliary/cso_cache/cso_cache.c
@@ -0,0 +1,406 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/* Authors:  Zack Rusin <zack@tungstengraphics.com>
+ */
+
+#include "pipe/p_debug.h"
+
+#include "util/u_memory.h"
+
+#include "cso_cache.h"
+#include "cso_hash.h"
+
+
+struct cso_cache {
+   struct cso_hash *blend_hash;
+   struct cso_hash *depth_stencil_hash;
+   struct cso_hash *fs_hash;
+   struct cso_hash *vs_hash;
+   struct cso_hash *rasterizer_hash;
+   struct cso_hash *sampler_hash;
+   int    max_size;
+
+   cso_sanitize_callback sanitize_cb;
+   void                 *sanitize_data;
+};
+
+#if 1
+static unsigned hash_key(const void *key, unsigned key_size)
+{
+   unsigned *ikey = (unsigned *)key;
+   unsigned hash = 0, i;
+
+   assert(key_size % 4 == 0);
+
+   /* I'm sure this can be improved on:
+    */
+   for (i = 0; i < key_size/4; i++)
+      hash ^= ikey[i];
+
+   return hash;
+}
+#else
+static unsigned hash_key(const unsigned char *p, int n)
+{
+   unsigned h = 0;
+   unsigned g;
+
+   while (n--) {
+      h = (h << 4) + *p++;
+      if ((g = (h & 0xf0000000)) != 0)
+         h ^= g >> 23;
+      h &= ~g;
+   }
+   return h;
+}
+#endif
+
+unsigned cso_construct_key(void *item, int item_size)
+{
+   return hash_key((item), item_size);
+}
+
+static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_type type)
+{
+   struct cso_hash *hash = 0;
+
+   switch(type) {
+   case CSO_BLEND:
+      hash = sc->blend_hash;
+      break;
+   case CSO_SAMPLER:
+      hash = sc->sampler_hash;
+      break;
+   case CSO_DEPTH_STENCIL_ALPHA:
+      hash = sc->depth_stencil_hash;
+      break;
+   case CSO_RASTERIZER:
+      hash = sc->rasterizer_hash;
+      break;
+   case CSO_FRAGMENT_SHADER:
+      hash = sc->fs_hash;
+      break;
+   case CSO_VERTEX_SHADER:
+      hash = sc->vs_hash;
+      break;
+   }
+
+   return hash;
+}
+
+static int _cso_size_for_type(enum cso_cache_type type)
+{
+   switch(type) {
+   case CSO_BLEND:
+      return sizeof(struct pipe_blend_state);
+   case CSO_SAMPLER:
+      return sizeof(struct pipe_sampler_state);
+   case CSO_DEPTH_STENCIL_ALPHA:
+      return sizeof(struct pipe_depth_stencil_alpha_state);
+   case CSO_RASTERIZER:
+      return sizeof(struct pipe_rasterizer_state);
+   case CSO_FRAGMENT_SHADER:
+      return sizeof(struct pipe_shader_state);
+   case CSO_VERTEX_SHADER:
+      return sizeof(struct pipe_shader_state);
+   }
+   return 0;
+}
+
+
+static void delete_blend_state(void *state, void *data)
+{
+   struct cso_blend *cso = (struct cso_blend *)state;
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+}
+
+static void delete_depth_stencil_state(void *state, void *data)
+{
+   struct cso_depth_stencil_alpha *cso = (struct cso_depth_stencil_alpha *)state;
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+}
+
+static void delete_sampler_state(void *state, void *data)
+{
+   struct cso_sampler *cso = (struct cso_sampler *)state;
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+}
+
+static void delete_rasterizer_state(void *state, void *data)
+{
+   struct cso_rasterizer *cso = (struct cso_rasterizer *)state;
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+}
+
+static void delete_fs_state(void *state, void *data)
+{
+   struct cso_fragment_shader *cso = (struct cso_fragment_shader *)state;
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+}
+
+static void delete_vs_state(void *state, void *data)
+{
+   struct cso_vertex_shader *cso = (struct cso_vertex_shader *)state;
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+}
+
+
+static INLINE void delete_cso(void *state, enum cso_cache_type type)
+{
+   switch (type) {
+   case CSO_BLEND:
+      delete_blend_state(state, 0);
+      break;
+   case CSO_SAMPLER:
+      delete_sampler_state(state, 0);
+      break;
+   case CSO_DEPTH_STENCIL_ALPHA:
+      delete_depth_stencil_state(state, 0);
+      break;
+   case CSO_RASTERIZER:
+      delete_rasterizer_state(state, 0);
+      break;
+   case CSO_FRAGMENT_SHADER:
+      delete_fs_state(state, 0);
+      break;
+   case CSO_VERTEX_SHADER:
+      delete_vs_state(state, 0);
+      break;
+   default:
+      assert(0);
+      FREE(state);
+   }
+}
+
+
+static INLINE void sanitize_hash(struct cso_cache *sc,
+                                 struct cso_hash *hash,
+                                 enum cso_cache_type type,
+                                 int max_size)
+{
+   if (sc->sanitize_cb)
+      sc->sanitize_cb(hash, type, max_size, sc->sanitize_data);
+}
+
+
+static INLINE void sanitize_cb(struct cso_hash *hash, enum cso_cache_type type,
+                               int max_size, void *user_data)
+{
+   /* if we're approach the maximum size, remove fourth of the entries
+    * otherwise every subsequent call will go through the same */
+   int hash_size = cso_hash_size(hash);
+   int max_entries = (max_size > hash_size) ? max_size : hash_size;
+   int to_remove =  (max_size < max_entries) * max_entries/4;
+   if (hash_size > max_size)
+      to_remove += hash_size - max_size;
+   while (to_remove) {
+      /*remove elements until we're good */
+      /*fixme: currently we pick the nodes to remove at random*/
+      struct cso_hash_iter iter = cso_hash_first_node(hash);
+      void  *cso = cso_hash_take(hash, cso_hash_iter_key(iter));
+      delete_cso(cso, type);
+      --to_remove;
+   }
+}
+
+struct cso_hash_iter
+cso_insert_state(struct cso_cache *sc,
+                 unsigned hash_key, enum cso_cache_type type,
+                 void *state)
+{
+   struct cso_hash *hash = _cso_hash_for_type(sc, type);
+   sanitize_hash(sc, hash, type, sc->max_size);
+
+   return cso_hash_insert(hash, hash_key, state);
+}
+
+struct cso_hash_iter
+cso_find_state(struct cso_cache *sc,
+               unsigned hash_key, enum cso_cache_type type)
+{
+   struct cso_hash *hash = _cso_hash_for_type(sc, type);
+
+   return cso_hash_find(hash, hash_key);
+}
+
+
+void *cso_hash_find_data_from_template( struct cso_hash *hash,
+				        unsigned hash_key, 
+				        void *templ,
+				        int size )
+{
+   struct cso_hash_iter iter = cso_hash_find(hash, hash_key);
+   while (!cso_hash_iter_is_null(iter)) {
+      void *iter_data = cso_hash_iter_data(iter);
+      if (!memcmp(iter_data, templ, size)) {
+	 /* We found a match
+	  */
+         return iter_data;
+      }
+      iter = cso_hash_iter_next(iter);
+   }
+   return NULL;
+}
+
+
+struct cso_hash_iter cso_find_state_template(struct cso_cache *sc,
+                                             unsigned hash_key, enum cso_cache_type type,
+                                             void *templ)
+{
+   struct cso_hash_iter iter = cso_find_state(sc, hash_key, type);
+   int size = _cso_size_for_type(type);
+   while (!cso_hash_iter_is_null(iter)) {
+      void *iter_data = cso_hash_iter_data(iter);
+      if (!memcmp(iter_data, templ, size))
+         return iter;
+      iter = cso_hash_iter_next(iter);
+   }
+   return iter;
+}
+
+void * cso_take_state(struct cso_cache *sc,
+                      unsigned hash_key, enum cso_cache_type type)
+{
+   struct cso_hash *hash = _cso_hash_for_type(sc, type);
+   return cso_hash_take(hash, hash_key);
+}
+
+struct cso_cache *cso_cache_create(void)
+{
+   struct cso_cache *sc = MALLOC_STRUCT(cso_cache);
+   if (sc == NULL)
+      return NULL;
+
+   sc->max_size           = 4096;
+   sc->blend_hash         = cso_hash_create();
+   sc->sampler_hash       = cso_hash_create();
+   sc->depth_stencil_hash = cso_hash_create();
+   sc->rasterizer_hash    = cso_hash_create();
+   sc->fs_hash            = cso_hash_create();
+   sc->vs_hash            = cso_hash_create();
+   sc->sanitize_cb        = sanitize_cb;
+   sc->sanitize_data      = 0;
+
+   return sc;
+}
+
+void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type,
+                        cso_state_callback func, void *user_data)
+{
+   struct cso_hash *hash = 0;
+   struct cso_hash_iter iter;
+
+   switch (type) {
+   case CSO_BLEND:
+      hash = sc->blend_hash;
+      break;
+   case CSO_SAMPLER:
+      hash = sc->sampler_hash;
+      break;
+   case CSO_DEPTH_STENCIL_ALPHA:
+      hash = sc->depth_stencil_hash;
+      break;
+   case CSO_RASTERIZER:
+      hash = sc->rasterizer_hash;
+      break;
+   case CSO_FRAGMENT_SHADER:
+      hash = sc->fs_hash;
+      break;
+   case CSO_VERTEX_SHADER:
+      hash = sc->vs_hash;
+      break;
+   }
+
+   iter = cso_hash_first_node(hash);
+   while (!cso_hash_iter_is_null(iter)) {
+      void *state = cso_hash_iter_data(iter);
+      iter = cso_hash_iter_next(iter);
+      if (state) {
+         func(state, user_data);
+      }
+   }
+}
+
+void cso_cache_delete(struct cso_cache *sc)
+{
+   assert(sc);
+   /* delete driver data */
+   cso_for_each_state(sc, CSO_BLEND, delete_blend_state, 0);
+   cso_for_each_state(sc, CSO_DEPTH_STENCIL_ALPHA, delete_depth_stencil_state, 0);
+   cso_for_each_state(sc, CSO_FRAGMENT_SHADER, delete_fs_state, 0);
+   cso_for_each_state(sc, CSO_VERTEX_SHADER, delete_vs_state, 0);
+   cso_for_each_state(sc, CSO_RASTERIZER, delete_rasterizer_state, 0);
+   cso_for_each_state(sc, CSO_SAMPLER, delete_sampler_state, 0);
+
+   cso_hash_delete(sc->blend_hash);
+   cso_hash_delete(sc->sampler_hash);
+   cso_hash_delete(sc->depth_stencil_hash);
+   cso_hash_delete(sc->rasterizer_hash);
+   cso_hash_delete(sc->fs_hash);
+   cso_hash_delete(sc->vs_hash);
+   FREE(sc);
+}
+
+void cso_set_maximum_cache_size(struct cso_cache *sc, int number)
+{
+   sc->max_size = number;
+
+   sanitize_hash(sc, sc->blend_hash, CSO_BLEND, sc->max_size);
+   sanitize_hash(sc, sc->depth_stencil_hash, CSO_DEPTH_STENCIL_ALPHA,
+                 sc->max_size);
+   sanitize_hash(sc, sc->fs_hash, CSO_FRAGMENT_SHADER, sc->max_size);
+   sanitize_hash(sc, sc->vs_hash, CSO_VERTEX_SHADER, sc->max_size);
+   sanitize_hash(sc, sc->rasterizer_hash, CSO_RASTERIZER, sc->max_size);
+   sanitize_hash(sc, sc->sampler_hash, CSO_SAMPLER, sc->max_size);
+}
+
+int cso_maximum_cache_size(const struct cso_cache *sc)
+{
+   return sc->max_size;
+}
+
+void cso_cache_set_sanitize_callback(struct cso_cache *sc,
+                                     cso_sanitize_callback cb,
+                                     void *user_data)
+{
+   sc->sanitize_cb   = cb;
+   sc->sanitize_data = user_data;
+}
+
diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h
new file mode 100644
index 0000000000..6b5c230e8f
--- /dev/null
+++ b/src/gallium/auxiliary/cso_cache/cso_cache.h
@@ -0,0 +1,176 @@
+/**************************************************************************
+ *
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /**
+  * @file
+  * Constant State Object (CSO) cache.
+  *
+  * The basic idea is that the states are created via the
+  * create_state/bind_state/delete_state semantics. The driver is expected to
+  * perform as much of the Gallium state translation to whatever its internal
+  * representation is during the create call. Gallium then has a caching
+  * mechanism where it stores the created states. When the pipeline needs an
+  * actual state change, a bind call is issued. In the bind call the driver
+  * gets its already translated representation.
+  *
+  * Those semantics mean that the driver doesn't do the repeated translations
+  * of states on every frame, but only once, when a new state is actually
+  * created.
+  *
+  * Even on hardware that doesn't do any kind of state cache, it makes the
+  * driver look a lot neater, plus it avoids all the redundant state
+  * translations on every frame.
+  *
+  * Currently our constant state objects are:
+  * - alpha test
+  * - blend
+  * - depth stencil
+  * - fragment shader
+  * - rasterizer (old setup)
+  * - sampler
+  * - vertex shader
+  *
+  * Things that are not constant state objects include:
+  * - blend_color
+  * - clip_state
+  * - clear_color_state
+  * - constant_buffer
+  * - feedback_state
+  * - framebuffer_state
+  * - polygon_stipple
+  * - scissor_state
+  * - texture_state
+  * - viewport_state
+  *
+  * @author Zack Rusin <zack@tungstengraphics.com>
+  */
+
+#ifndef CSO_CACHE_H
+#define CSO_CACHE_H
+
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+
+/* cso_hash.h is necessary for cso_hash_iter, as MSVC requires structures
+ * returned by value to be fully defined */
+#include "cso_hash.h"
+
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+enum cso_cache_type {
+   CSO_BLEND,
+   CSO_SAMPLER,
+   CSO_DEPTH_STENCIL_ALPHA,
+   CSO_RASTERIZER,
+   CSO_FRAGMENT_SHADER,
+   CSO_VERTEX_SHADER
+};
+
+typedef void (*cso_state_callback)(void *ctx, void *obj);
+
+typedef void (*cso_sanitize_callback)(struct cso_hash *hash,
+                                      enum cso_cache_type type,
+                                      int max_size,
+                                      void *user_data);
+
+struct cso_cache;
+
+struct cso_blend {
+   struct pipe_blend_state state;
+   void *data;
+   cso_state_callback delete_state;
+   struct pipe_context *context;
+};
+
+struct cso_depth_stencil_alpha {
+   struct pipe_depth_stencil_alpha_state state;
+   void *data;
+   cso_state_callback delete_state;
+   struct pipe_context *context;
+};
+
+struct cso_rasterizer {
+   struct pipe_rasterizer_state state;
+   void *data;
+   cso_state_callback delete_state;
+   struct pipe_context *context;
+};
+
+struct cso_fragment_shader {
+   struct pipe_shader_state state;
+   void *data;
+   cso_state_callback delete_state;
+   struct pipe_context *context;
+};
+
+struct cso_vertex_shader {
+   struct pipe_shader_state state;
+   void *data;
+   cso_state_callback delete_state;
+   struct pipe_context *context;
+};
+
+struct cso_sampler {
+   struct pipe_sampler_state state;
+   void *data;
+   cso_state_callback delete_state;
+   struct pipe_context *context;
+};
+
+unsigned cso_construct_key(void *item, int item_size);
+
+struct cso_cache *cso_cache_create(void);
+void cso_cache_delete(struct cso_cache *sc);
+
+void cso_cache_set_sanitize_callback(struct cso_cache *sc,
+                                     cso_sanitize_callback cb,
+                                     void *user_data);
+
+struct cso_hash_iter cso_insert_state(struct cso_cache *sc,
+                                      unsigned hash_key, enum cso_cache_type type,
+                                      void *state);
+struct cso_hash_iter cso_find_state(struct cso_cache *sc,
+                                    unsigned hash_key, enum cso_cache_type type);
+struct cso_hash_iter cso_find_state_template(struct cso_cache *sc,
+                                             unsigned hash_key, enum cso_cache_type type,
+                                             void *templ);
+void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type,
+                        cso_state_callback func, void *user_data);
+void * cso_take_state(struct cso_cache *sc, unsigned hash_key,
+                      enum cso_cache_type type);
+
+void cso_set_maximum_cache_size(struct cso_cache *sc, int number);
+int cso_maximum_cache_size(const struct cso_cache *sc);
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c
new file mode 100644
index 0000000000..a9157aad71
--- /dev/null
+++ b/src/gallium/auxiliary/cso_cache/cso_context.c
@@ -0,0 +1,866 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /**
+  * @file
+  * 
+  * Wrap the cso cache & hash mechanisms in a simplified
+  * pipe-driver-specific interface.
+  *
+  * @author Zack Rusin <zack@tungstengraphics.com>
+  * @author Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "pipe/p_state.h"
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "tgsi/tgsi_parse.h"
+
+#include "cso_cache/cso_context.h"
+#include "cso_cache/cso_cache.h"
+#include "cso_cache/cso_hash.h"
+
+struct cso_context {
+   struct pipe_context *pipe;
+   struct cso_cache *cache;
+
+   struct {
+      void *samplers[PIPE_MAX_SAMPLERS];
+      unsigned nr_samplers;
+   } hw;
+
+   void *samplers[PIPE_MAX_SAMPLERS];
+   unsigned nr_samplers;
+
+   void *samplers_saved[PIPE_MAX_SAMPLERS];
+   unsigned nr_samplers_saved;
+
+   struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+   uint nr_textures;
+
+   struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS];
+   uint nr_textures_saved;
+
+   /** Current and saved state.
+    * The saved state is used as a 1-deep stack.
+    */
+   void *blend, *blend_saved;
+   void *depth_stencil, *depth_stencil_saved;
+   void *rasterizer, *rasterizer_saved;
+   void *fragment_shader, *fragment_shader_saved;
+   void *vertex_shader, *vertex_shader_saved;
+
+   struct pipe_framebuffer_state fb, fb_saved;
+   struct pipe_viewport_state vp, vp_saved;
+   struct pipe_blend_color blend_color;
+};
+
+
+static void
+free_framebuffer_state(struct pipe_framebuffer_state *fb);
+
+
+static boolean delete_blend_state(struct cso_context *ctx, void *state)
+{
+   struct cso_blend *cso = (struct cso_blend *)state;
+
+   if (ctx->blend == cso->data)
+      return FALSE;
+
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+   return TRUE;
+}
+
+static boolean delete_depth_stencil_state(struct cso_context *ctx, void *state)
+{
+   struct cso_depth_stencil_alpha *cso = (struct cso_depth_stencil_alpha *)state;
+
+   if (ctx->depth_stencil == cso->data)
+      return FALSE;
+
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+
+   return TRUE;
+}
+
+static boolean delete_sampler_state(struct cso_context *ctx, void *state)
+{
+   struct cso_sampler *cso = (struct cso_sampler *)state;
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+   return TRUE;
+}
+
+static boolean delete_rasterizer_state(struct cso_context *ctx, void *state)
+{
+   struct cso_rasterizer *cso = (struct cso_rasterizer *)state;
+
+   if (ctx->rasterizer == cso->data)
+      return FALSE;
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+   return TRUE;
+}
+
+static boolean delete_fs_state(struct cso_context *ctx, void *state)
+{
+   struct cso_fragment_shader *cso = (struct cso_fragment_shader *)state;
+   if (ctx->fragment_shader == cso->data)
+      return FALSE;
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+   return TRUE;
+}
+
+static boolean delete_vs_state(struct cso_context *ctx, void *state)
+{
+   struct cso_vertex_shader *cso = (struct cso_vertex_shader *)state;
+   if (ctx->vertex_shader == cso->data)
+      return TRUE;
+   if (cso->delete_state)
+      cso->delete_state(cso->context, cso->data);
+   FREE(state);
+   return FALSE;
+}
+
+
+static INLINE boolean delete_cso(struct cso_context *ctx,
+                                 void *state, enum cso_cache_type type)
+{
+   switch (type) {
+   case CSO_BLEND:
+      return delete_blend_state(ctx, state);
+      break;
+   case CSO_SAMPLER:
+      return delete_sampler_state(ctx, state);
+      break;
+   case CSO_DEPTH_STENCIL_ALPHA:
+      return delete_depth_stencil_state(ctx, state);
+      break;
+   case CSO_RASTERIZER:
+      return delete_rasterizer_state(ctx, state);
+      break;
+   case CSO_FRAGMENT_SHADER:
+      return delete_fs_state(ctx, state);
+      break;
+   case CSO_VERTEX_SHADER:
+      return delete_vs_state(ctx, state);
+      break;
+   default:
+      assert(0);
+      FREE(state);
+   }
+   return FALSE;
+}
+
+static INLINE void sanitize_hash(struct cso_hash *hash, enum cso_cache_type type,
+                                 int max_size, void *user_data)
+{
+   struct cso_context *ctx = (struct cso_context *)user_data;
+   /* if we're approach the maximum size, remove fourth of the entries
+    * otherwise every subsequent call will go through the same */
+   int hash_size = cso_hash_size(hash);
+   int max_entries = (max_size > hash_size) ? max_size : hash_size;
+   int to_remove =  (max_size < max_entries) * max_entries/4;
+   struct cso_hash_iter iter = cso_hash_first_node(hash);
+   if (hash_size > max_size)
+      to_remove += hash_size - max_size;
+   while (to_remove) {
+      /*remove elements until we're good */
+      /*fixme: currently we pick the nodes to remove at random*/
+      void *cso = cso_hash_iter_data(iter);
+      if (delete_cso(ctx, cso, type)) {
+         iter = cso_hash_erase(hash, iter);
+         --to_remove;
+      } else
+         iter = cso_hash_iter_next(iter);
+   }
+}
+
+
+struct cso_context *cso_create_context( struct pipe_context *pipe )
+{
+   struct cso_context *ctx = CALLOC_STRUCT(cso_context);
+   if (ctx == NULL)
+      goto out;
+
+   ctx->cache = cso_cache_create();
+   if (ctx->cache == NULL)
+      goto out;
+   cso_cache_set_sanitize_callback(ctx->cache,
+                                   sanitize_hash,
+                                   ctx);
+
+   ctx->pipe = pipe;
+
+   /* Enable for testing: */
+   if (0) cso_set_maximum_cache_size( ctx->cache, 4 );
+
+   return ctx;
+
+out:
+   cso_destroy_context( ctx );      
+   return NULL;
+}
+
+
+/**
+ * Prior to context destruction, this function unbinds all state objects.
+ */
+void cso_release_all( struct cso_context *ctx )
+{
+   unsigned i;
+   
+   if (ctx->pipe) {
+      ctx->pipe->bind_blend_state( ctx->pipe, NULL );
+      ctx->pipe->bind_rasterizer_state( ctx->pipe, NULL );
+      ctx->pipe->bind_sampler_states( ctx->pipe, 0, NULL );
+      ctx->pipe->bind_depth_stencil_alpha_state( ctx->pipe, NULL );
+      ctx->pipe->bind_fs_state( ctx->pipe, NULL );
+      ctx->pipe->bind_vs_state( ctx->pipe, NULL );
+   }
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      pipe_texture_reference(&ctx->textures[i], NULL);
+      pipe_texture_reference(&ctx->textures_saved[i], NULL);
+   }
+
+   free_framebuffer_state(&ctx->fb);
+   free_framebuffer_state(&ctx->fb_saved);
+
+   if (ctx->cache) {
+      cso_cache_delete( ctx->cache );
+      ctx->cache = NULL;
+   }
+}
+
+
+void cso_destroy_context( struct cso_context *ctx )
+{
+   if (ctx) {
+      //cso_release_all( ctx );
+      FREE( ctx );
+   }
+}
+
+
+/* Those function will either find the state of the given template
+ * in the cache or they will create a new state from the given
+ * template, insert it in the cache and return it.
+ */
+
+/*
+ * If the driver returns 0 from the create method then they will assign
+ * the data member of the cso to be the template itself.
+ */
+
+enum pipe_error cso_set_blend(struct cso_context *ctx,
+                              const struct pipe_blend_state *templ)
+{
+   unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_blend_state));
+   struct cso_hash_iter iter = cso_find_state_template(ctx->cache,
+                                                       hash_key, CSO_BLEND,
+                                                       (void*)templ);
+   void *handle;
+
+   if (cso_hash_iter_is_null(iter)) {
+      struct cso_blend *cso = MALLOC(sizeof(struct cso_blend));
+      if (!cso)
+         return PIPE_ERROR_OUT_OF_MEMORY;
+
+      memcpy(&cso->state, templ, sizeof(*templ));
+      cso->data = ctx->pipe->create_blend_state(ctx->pipe, &cso->state);
+      cso->delete_state = (cso_state_callback)ctx->pipe->delete_blend_state;
+      cso->context = ctx->pipe;
+
+      iter = cso_insert_state(ctx->cache, hash_key, CSO_BLEND, cso);
+      if (cso_hash_iter_is_null(iter)) {
+         FREE(cso);
+         return PIPE_ERROR_OUT_OF_MEMORY;
+      }
+
+      handle = cso->data;
+   }
+   else {
+      handle = ((struct cso_blend *)cso_hash_iter_data(iter))->data;
+   }
+
+   if (ctx->blend != handle) {
+      ctx->blend = handle;
+      ctx->pipe->bind_blend_state(ctx->pipe, handle);
+   }
+   return PIPE_OK;
+}
+
+void cso_save_blend(struct cso_context *ctx)
+{
+   assert(!ctx->blend_saved);
+   ctx->blend_saved = ctx->blend;
+}
+
+void cso_restore_blend(struct cso_context *ctx)
+{
+   if (ctx->blend != ctx->blend_saved) {
+      ctx->blend = ctx->blend_saved;
+      ctx->pipe->bind_blend_state(ctx->pipe, ctx->blend_saved);
+   }
+   ctx->blend_saved = NULL;
+}
+
+
+
+enum pipe_error cso_single_sampler(struct cso_context *ctx,
+                                   unsigned idx,
+                                   const struct pipe_sampler_state *templ)
+{
+   void *handle = NULL;
+
+   if (templ != NULL) {
+      unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_sampler_state));
+      struct cso_hash_iter iter = cso_find_state_template(ctx->cache,
+                                                          hash_key, CSO_SAMPLER,
+                                                          (void*)templ);
+
+      if (cso_hash_iter_is_null(iter)) {
+         struct cso_sampler *cso = MALLOC(sizeof(struct cso_sampler));
+         if (!cso)
+            return PIPE_ERROR_OUT_OF_MEMORY;
+
+         memcpy(&cso->state, templ, sizeof(*templ));
+         cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state);
+         cso->delete_state = (cso_state_callback)ctx->pipe->delete_sampler_state;
+         cso->context = ctx->pipe;
+
+         iter = cso_insert_state(ctx->cache, hash_key, CSO_SAMPLER, cso);
+         if (cso_hash_iter_is_null(iter)) {
+            FREE(cso);
+            return PIPE_ERROR_OUT_OF_MEMORY;
+         }
+
+         handle = cso->data;
+      }
+      else {
+         handle = ((struct cso_sampler *)cso_hash_iter_data(iter))->data;
+      }
+   }
+
+   ctx->samplers[idx] = handle;
+   return PIPE_OK;
+}
+
+void cso_single_sampler_done( struct cso_context *ctx )
+{
+   unsigned i;
+
+   /* find highest non-null sampler */
+   for (i = PIPE_MAX_SAMPLERS; i > 0; i--) {
+      if (ctx->samplers[i - 1] != NULL)
+         break;
+   }
+
+   ctx->nr_samplers = i;
+
+   if (ctx->hw.nr_samplers != ctx->nr_samplers ||
+       memcmp(ctx->hw.samplers,
+              ctx->samplers,
+              ctx->nr_samplers * sizeof(void *)) != 0) 
+   {
+      memcpy(ctx->hw.samplers, ctx->samplers, ctx->nr_samplers * sizeof(void *));
+      ctx->hw.nr_samplers = ctx->nr_samplers;
+
+      ctx->pipe->bind_sampler_states(ctx->pipe, ctx->nr_samplers, ctx->samplers);
+   }
+}
+
+/*
+ * If the function encouters any errors it will return the
+ * last one. Done to always try to set as many samplers
+ * as possible.
+ */
+enum pipe_error cso_set_samplers( struct cso_context *ctx,
+                                  unsigned nr,
+                                  const struct pipe_sampler_state **templates )
+{
+   unsigned i;
+   enum pipe_error temp, error = PIPE_OK;
+
+   /* TODO: fastpath
+    */
+
+   for (i = 0; i < nr; i++) {
+      temp = cso_single_sampler( ctx, i, templates[i] );
+      if (temp != PIPE_OK)
+         error = temp;
+   }
+
+   for ( ; i < ctx->nr_samplers; i++) {
+      temp = cso_single_sampler( ctx, i, NULL );
+      if (temp != PIPE_OK)
+         error = temp;
+   }
+
+   cso_single_sampler_done( ctx );
+
+   return error;
+}
+
+void cso_save_samplers(struct cso_context *ctx)
+{
+   ctx->nr_samplers_saved = ctx->nr_samplers;
+   memcpy(ctx->samplers_saved, ctx->samplers, sizeof(ctx->samplers));
+}
+
+void cso_restore_samplers(struct cso_context *ctx)
+{
+   ctx->nr_samplers = ctx->nr_samplers_saved;
+   memcpy(ctx->samplers, ctx->samplers_saved, sizeof(ctx->samplers));
+   cso_single_sampler_done( ctx );
+}
+
+
+enum pipe_error cso_set_sampler_textures( struct cso_context *ctx,
+                                          uint count,
+                                          struct pipe_texture **textures )
+{
+   uint i;
+
+   ctx->nr_textures = count;
+
+   for (i = 0; i < count; i++)
+      pipe_texture_reference(&ctx->textures[i], textures[i]);
+   for ( ; i < PIPE_MAX_SAMPLERS; i++)
+      pipe_texture_reference(&ctx->textures[i], NULL);
+
+   ctx->pipe->set_sampler_textures(ctx->pipe, count, textures);
+
+   return PIPE_OK;
+}
+
+void cso_save_sampler_textures( struct cso_context *ctx )
+{
+   uint i;
+
+   ctx->nr_textures_saved = ctx->nr_textures;
+   for (i = 0; i < ctx->nr_textures; i++) {
+      assert(!ctx->textures_saved[i]);
+      pipe_texture_reference(&ctx->textures_saved[i], ctx->textures[i]);
+   }
+}
+
+void cso_restore_sampler_textures( struct cso_context *ctx )
+{
+   uint i;
+
+   ctx->nr_textures = ctx->nr_textures_saved;
+
+   for (i = 0; i < ctx->nr_textures; i++) {
+      pipe_texture_reference(&ctx->textures[i], NULL);
+      ctx->textures[i] = ctx->textures_saved[i];
+      ctx->textures_saved[i] = NULL;
+   }
+   for ( ; i < PIPE_MAX_SAMPLERS; i++)
+      pipe_texture_reference(&ctx->textures[i], NULL);
+
+   ctx->pipe->set_sampler_textures(ctx->pipe, ctx->nr_textures, ctx->textures);
+
+   ctx->nr_textures_saved = 0;
+}
+
+
+
+enum pipe_error cso_set_depth_stencil_alpha(struct cso_context *ctx,
+                                            const struct pipe_depth_stencil_alpha_state *templ)
+{
+   unsigned hash_key = cso_construct_key((void*)templ,
+                                         sizeof(struct pipe_depth_stencil_alpha_state));
+   struct cso_hash_iter iter = cso_find_state_template(ctx->cache,
+                                                       hash_key, 
+						       CSO_DEPTH_STENCIL_ALPHA,
+                                                       (void*)templ);
+   void *handle;
+
+   if (cso_hash_iter_is_null(iter)) {
+      struct cso_depth_stencil_alpha *cso = MALLOC(sizeof(struct cso_depth_stencil_alpha));
+      if (!cso)
+         return PIPE_ERROR_OUT_OF_MEMORY;
+
+      memcpy(&cso->state, templ, sizeof(*templ));
+      cso->data = ctx->pipe->create_depth_stencil_alpha_state(ctx->pipe, &cso->state);
+      cso->delete_state = (cso_state_callback)ctx->pipe->delete_depth_stencil_alpha_state;
+      cso->context = ctx->pipe;
+
+      iter = cso_insert_state(ctx->cache, hash_key, CSO_DEPTH_STENCIL_ALPHA, cso);
+      if (cso_hash_iter_is_null(iter)) {
+         FREE(cso);
+         return PIPE_ERROR_OUT_OF_MEMORY;
+      }
+
+      handle = cso->data;
+   }
+   else {
+      handle = ((struct cso_depth_stencil_alpha *)cso_hash_iter_data(iter))->data;
+   }
+
+   if (ctx->depth_stencil != handle) {
+      ctx->depth_stencil = handle;
+      ctx->pipe->bind_depth_stencil_alpha_state(ctx->pipe, handle);
+   }
+   return PIPE_OK;
+}
+
+void cso_save_depth_stencil_alpha(struct cso_context *ctx)
+{
+   assert(!ctx->depth_stencil_saved);
+   ctx->depth_stencil_saved = ctx->depth_stencil;
+}
+
+void cso_restore_depth_stencil_alpha(struct cso_context *ctx)
+{
+   if (ctx->depth_stencil != ctx->depth_stencil_saved) {
+      ctx->depth_stencil = ctx->depth_stencil_saved;
+      ctx->pipe->bind_depth_stencil_alpha_state(ctx->pipe, ctx->depth_stencil_saved);
+   }
+   ctx->depth_stencil_saved = NULL;
+}
+
+
+
+enum pipe_error cso_set_rasterizer(struct cso_context *ctx,
+                                   const struct pipe_rasterizer_state *templ)
+{
+   unsigned hash_key = cso_construct_key((void*)templ,
+                                         sizeof(struct pipe_rasterizer_state));
+   struct cso_hash_iter iter = cso_find_state_template(ctx->cache,
+                                                       hash_key, CSO_RASTERIZER,
+                                                       (void*)templ);
+   void *handle = NULL;
+
+   if (cso_hash_iter_is_null(iter)) {
+      struct cso_rasterizer *cso = MALLOC(sizeof(struct cso_rasterizer));
+      if (!cso)
+         return PIPE_ERROR_OUT_OF_MEMORY;
+
+      memcpy(&cso->state, templ, sizeof(*templ));
+      cso->data = ctx->pipe->create_rasterizer_state(ctx->pipe, &cso->state);
+      cso->delete_state = (cso_state_callback)ctx->pipe->delete_rasterizer_state;
+      cso->context = ctx->pipe;
+
+      iter = cso_insert_state(ctx->cache, hash_key, CSO_RASTERIZER, cso);
+      if (cso_hash_iter_is_null(iter)) {
+         FREE(cso);
+         return PIPE_ERROR_OUT_OF_MEMORY;
+      }
+
+      handle = cso->data;
+   }
+   else {
+      handle = ((struct cso_rasterizer *)cso_hash_iter_data(iter))->data;
+   }
+
+   if (ctx->rasterizer != handle) {
+      ctx->rasterizer = handle;
+      ctx->pipe->bind_rasterizer_state(ctx->pipe, handle);
+   }
+   return PIPE_OK;
+}
+
+void cso_save_rasterizer(struct cso_context *ctx)
+{
+   assert(!ctx->rasterizer_saved);
+   ctx->rasterizer_saved = ctx->rasterizer;
+}
+
+void cso_restore_rasterizer(struct cso_context *ctx)
+{
+   if (ctx->rasterizer != ctx->rasterizer_saved) {
+      ctx->rasterizer = ctx->rasterizer_saved;
+      ctx->pipe->bind_rasterizer_state(ctx->pipe, ctx->rasterizer_saved);
+   }
+   ctx->rasterizer_saved = NULL;
+}
+
+
+
+enum pipe_error cso_set_fragment_shader_handle(struct cso_context *ctx,
+                                               void *handle )
+{
+   if (ctx->fragment_shader != handle) {
+      ctx->fragment_shader = handle;
+      ctx->pipe->bind_fs_state(ctx->pipe, handle);
+   }
+   return PIPE_OK;
+}
+
+void cso_delete_fragment_shader(struct cso_context *ctx, void *handle )
+{
+   if (handle == ctx->fragment_shader) {
+      /* unbind before deleting */
+      ctx->pipe->bind_fs_state(ctx->pipe, NULL);
+      ctx->fragment_shader = NULL;
+   }
+   ctx->pipe->delete_fs_state(ctx->pipe, handle);
+}
+
+/* Not really working:
+ */
+#if 0
+enum pipe_error cso_set_fragment_shader(struct cso_context *ctx,
+                                        const struct pipe_shader_state *templ)
+{
+   const struct tgsi_token *tokens = templ->tokens;
+   unsigned num_tokens = tgsi_num_tokens(tokens);
+   size_t tokens_size = num_tokens*sizeof(struct tgsi_token);
+   unsigned hash_key = cso_construct_key((void*)tokens, tokens_size);
+   struct cso_hash_iter iter = cso_find_state_template(ctx->cache,
+                                                       hash_key, 
+                                                       CSO_FRAGMENT_SHADER,
+                                                       (void*)tokens);
+   void *handle = NULL;
+
+   if (cso_hash_iter_is_null(iter)) {
+      struct cso_fragment_shader *cso = MALLOC(sizeof(struct cso_fragment_shader) + tokens_size);
+      struct tgsi_token *cso_tokens = (struct tgsi_token *)((char *)cso + sizeof(*cso));
+
+      if (!cso)
+         return PIPE_ERROR_OUT_OF_MEMORY;
+
+      memcpy(cso_tokens, tokens, tokens_size);
+      cso->state.tokens = cso_tokens;
+      cso->data = ctx->pipe->create_fs_state(ctx->pipe, &cso->state);
+      cso->delete_state = (cso_state_callback)ctx->pipe->delete_fs_state;
+      cso->context = ctx->pipe;
+
+      iter = cso_insert_state(ctx->cache, hash_key, CSO_FRAGMENT_SHADER, cso);
+      if (cso_hash_iter_is_null(iter)) {
+         FREE(cso);
+         return PIPE_ERROR_OUT_OF_MEMORY;
+      }
+
+      handle = cso->data;
+   }
+   else {
+      handle = ((struct cso_fragment_shader *)cso_hash_iter_data(iter))->data;
+   }
+
+   return cso_set_fragment_shader_handle( ctx, handle );
+}
+#endif
+
+void cso_save_fragment_shader(struct cso_context *ctx)
+{
+   assert(!ctx->fragment_shader_saved);
+   ctx->fragment_shader_saved = ctx->fragment_shader;
+}
+
+void cso_restore_fragment_shader(struct cso_context *ctx)
+{
+   if (ctx->fragment_shader_saved != ctx->fragment_shader) {
+      ctx->pipe->bind_fs_state(ctx->pipe, ctx->fragment_shader_saved);
+      ctx->fragment_shader = ctx->fragment_shader_saved;
+   }
+   ctx->fragment_shader_saved = NULL;
+}
+
+
+enum pipe_error cso_set_vertex_shader_handle(struct cso_context *ctx,
+                                             void *handle )
+{
+   if (ctx->vertex_shader != handle) {
+      ctx->vertex_shader = handle;
+      ctx->pipe->bind_vs_state(ctx->pipe, handle);
+   }
+   return PIPE_OK;
+}
+
+void cso_delete_vertex_shader(struct cso_context *ctx, void *handle )
+{
+   if (handle == ctx->vertex_shader) {
+      /* unbind before deleting */
+      ctx->pipe->bind_vs_state(ctx->pipe, NULL);
+      ctx->vertex_shader = NULL;
+   }
+   ctx->pipe->delete_vs_state(ctx->pipe, handle);
+}
+
+
+/* Not really working:
+ */
+#if 0
+enum pipe_error cso_set_vertex_shader(struct cso_context *ctx,
+                                      const struct pipe_shader_state *templ)
+{
+   unsigned hash_key = cso_construct_key((void*)templ,
+                                         sizeof(struct pipe_shader_state));
+   struct cso_hash_iter iter = cso_find_state_template(ctx->cache,
+                                                       hash_key, CSO_VERTEX_SHADER,
+                                                       (void*)templ);
+   void *handle = NULL;
+
+   if (cso_hash_iter_is_null(iter)) {
+      struct cso_vertex_shader *cso = MALLOC(sizeof(struct cso_vertex_shader));
+
+      if (!cso)
+         return PIPE_ERROR_OUT_OF_MEMORY;
+
+      memcpy(cso->state, templ, sizeof(*templ));
+      cso->data = ctx->pipe->create_vs_state(ctx->pipe, &cso->state);
+      cso->delete_state = (cso_state_callback)ctx->pipe->delete_vs_state;
+      cso->context = ctx->pipe;
+
+      iter = cso_insert_state(ctx->cache, hash_key, CSO_VERTEX_SHADER, cso);
+      if (cso_hash_iter_is_null(iter)) {
+         FREE(cso);
+         return PIPE_ERROR_OUT_OF_MEMORY;
+      }
+
+      handle = cso->data;
+   }
+   else {
+      handle = ((struct cso_vertex_shader *)cso_hash_iter_data(iter))->data;
+   }
+
+   return cso_set_vertex_shader_handle( ctx, handle );
+}
+#endif
+
+
+
+void cso_save_vertex_shader(struct cso_context *ctx)
+{
+   assert(!ctx->vertex_shader_saved);
+   ctx->vertex_shader_saved = ctx->vertex_shader;
+}
+
+void cso_restore_vertex_shader(struct cso_context *ctx)
+{
+   if (ctx->vertex_shader_saved != ctx->vertex_shader) {
+      ctx->pipe->bind_vs_state(ctx->pipe, ctx->vertex_shader_saved);
+      ctx->vertex_shader = ctx->vertex_shader_saved;
+   }
+   ctx->vertex_shader_saved = NULL;
+}
+
+
+/**
+ * Copy framebuffer state from src to dst with refcounting of surfaces.
+ */
+static void
+copy_framebuffer_state(struct pipe_framebuffer_state *dst,
+                       const struct pipe_framebuffer_state *src)
+{
+   uint i;
+
+   dst->width = src->width;
+   dst->height = src->height;
+   dst->nr_cbufs = src->nr_cbufs;
+   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
+      pipe_surface_reference(&dst->cbufs[i], src->cbufs[i]);
+   }
+   pipe_surface_reference(&dst->zsbuf, src->zsbuf);
+}
+
+
+static void
+free_framebuffer_state(struct pipe_framebuffer_state *fb)
+{
+   uint i;
+
+   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
+      pipe_surface_reference(&fb->cbufs[i], NULL);
+   }
+   pipe_surface_reference(&fb->zsbuf, NULL);
+}
+
+
+enum pipe_error cso_set_framebuffer(struct cso_context *ctx,
+                                    const struct pipe_framebuffer_state *fb)
+{
+   if (memcmp(&ctx->fb, fb, sizeof(*fb)) != 0) {
+      copy_framebuffer_state(&ctx->fb, fb);
+      ctx->pipe->set_framebuffer_state(ctx->pipe, fb);
+   }
+   return PIPE_OK;
+}
+
+void cso_save_framebuffer(struct cso_context *ctx)
+{
+   copy_framebuffer_state(&ctx->fb_saved, &ctx->fb);
+}
+
+void cso_restore_framebuffer(struct cso_context *ctx)
+{
+   if (memcmp(&ctx->fb, &ctx->fb_saved, sizeof(ctx->fb))) {
+      copy_framebuffer_state(&ctx->fb, &ctx->fb_saved);
+      ctx->pipe->set_framebuffer_state(ctx->pipe, &ctx->fb);
+      free_framebuffer_state(&ctx->fb_saved);
+   }
+}
+
+
+enum pipe_error cso_set_viewport(struct cso_context *ctx,
+                                 const struct pipe_viewport_state *vp)
+{
+   if (memcmp(&ctx->vp, vp, sizeof(*vp))) {
+      ctx->vp = *vp;
+      ctx->pipe->set_viewport_state(ctx->pipe, vp);
+   }
+   return PIPE_OK;
+}
+
+void cso_save_viewport(struct cso_context *ctx)
+{
+   ctx->vp_saved = ctx->vp;
+}
+
+
+void cso_restore_viewport(struct cso_context *ctx)
+{
+   if (memcmp(&ctx->vp, &ctx->vp_saved, sizeof(ctx->vp))) {
+      ctx->vp = ctx->vp_saved;
+      ctx->pipe->set_viewport_state(ctx->pipe, &ctx->vp);
+   }
+}
+
+
+
+
+enum pipe_error cso_set_blend_color(struct cso_context *ctx,
+                                    const struct pipe_blend_color *bc)
+{
+   if (memcmp(&ctx->blend_color, bc, sizeof(ctx->blend_color))) {
+      ctx->blend_color = *bc;
+      ctx->pipe->set_blend_color(ctx->pipe, bc);
+   }
+   return PIPE_OK;
+}
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h
new file mode 100644
index 0000000000..b04e98bfa1
--- /dev/null
+++ b/src/gallium/auxiliary/cso_cache/cso_context.h
@@ -0,0 +1,145 @@
+/**************************************************************************
+ *
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#ifndef CSO_CONTEXT_H
+#define CSO_CONTEXT_H
+
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_error.h"
+
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+struct cso_context;
+
+struct cso_context *cso_create_context( struct pipe_context *pipe );
+
+void cso_release_all( struct cso_context *ctx );
+
+void cso_destroy_context( struct cso_context *cso );
+
+
+
+enum pipe_error cso_set_blend( struct cso_context *cso,
+                               const struct pipe_blend_state *blend );
+void cso_save_blend(struct cso_context *cso);
+void cso_restore_blend(struct cso_context *cso);
+
+
+
+enum pipe_error cso_set_depth_stencil_alpha( struct cso_context *cso,
+                                             const struct pipe_depth_stencil_alpha_state *dsa );
+void cso_save_depth_stencil_alpha(struct cso_context *cso);
+void cso_restore_depth_stencil_alpha(struct cso_context *cso);
+
+
+
+enum pipe_error cso_set_rasterizer( struct cso_context *cso,
+                                    const struct pipe_rasterizer_state *rasterizer );
+void cso_save_rasterizer(struct cso_context *cso);
+void cso_restore_rasterizer(struct cso_context *cso);
+
+
+
+enum pipe_error cso_set_samplers( struct cso_context *cso,
+                                  unsigned count,
+                                  const struct pipe_sampler_state **states );
+void cso_save_samplers(struct cso_context *cso);
+void cso_restore_samplers(struct cso_context *cso);
+
+/* Alternate interface to support state trackers that like to modify
+ * samplers one at a time:
+ */
+enum pipe_error cso_single_sampler( struct cso_context *cso,
+                                    unsigned nr,
+                                    const struct pipe_sampler_state *states );
+
+void cso_single_sampler_done( struct cso_context *cso );
+
+
+
+enum pipe_error cso_set_sampler_textures( struct cso_context *cso,
+                                          uint count,
+                                          struct pipe_texture **textures );
+void cso_save_sampler_textures( struct cso_context *cso );
+void cso_restore_sampler_textures( struct cso_context *cso );
+
+
+
+/* These aren't really sensible -- most of the time the api provides
+ * object semantics for shaders anyway, and the cases where it doesn't
+ * (eg mesa's internall-generated texenv programs), it will be up to
+ * the state tracker to implement their own specialized caching.
+ */
+enum pipe_error cso_set_fragment_shader_handle(struct cso_context *ctx,
+                                               void *handle );
+void cso_delete_fragment_shader(struct cso_context *ctx, void *handle );
+/*
+enum pipe_error cso_set_fragment_shader( struct cso_context *cso,
+                                         const struct pipe_shader_state *shader );
+*/
+void cso_save_fragment_shader(struct cso_context *cso);
+void cso_restore_fragment_shader(struct cso_context *cso);
+
+
+enum pipe_error cso_set_vertex_shader_handle(struct cso_context *ctx,
+                                             void *handle );
+void cso_delete_vertex_shader(struct cso_context *ctx, void *handle );
+/*
+enum pipe_error cso_set_vertex_shader( struct cso_context *cso,
+                                       const struct pipe_shader_state *shader );
+*/
+void cso_save_vertex_shader(struct cso_context *cso);
+void cso_restore_vertex_shader(struct cso_context *cso);
+
+
+
+enum pipe_error cso_set_framebuffer(struct cso_context *cso,
+                                    const struct pipe_framebuffer_state *fb);
+void cso_save_framebuffer(struct cso_context *cso);
+void cso_restore_framebuffer(struct cso_context *cso);
+
+
+enum pipe_error cso_set_viewport(struct cso_context *cso,
+                                 const struct pipe_viewport_state *vp);
+void cso_save_viewport(struct cso_context *cso);
+void cso_restore_viewport(struct cso_context *cso);
+
+
+enum pipe_error cso_set_blend_color(struct cso_context *cso,
+                                    const struct pipe_blend_color *bc);
+
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c
new file mode 100644
index 0000000000..4e7664f9bf
--- /dev/null
+++ b/src/gallium/auxiliary/cso_cache/cso_hash.c
@@ -0,0 +1,439 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin <zack@tungstengraphics.com>
+  */
+
+#include "pipe/p_debug.h"
+#include "util/u_memory.h"
+
+#include "cso_hash.h"
+
+#define MAX(a, b) ((a > b) ? (a) : (b))
+
+static const int MinNumBits = 4;
+
+static const unsigned char prime_deltas[] = {
+   0,  0,  1,  3,  1,  5,  3,  3,  1,  9,  7,  5,  3,  9, 25,  3,
+   1, 21,  3, 21,  7, 15,  9,  5,  3, 29, 15,  0,  0,  0,  0,  0
+};
+
+static int primeForNumBits(int numBits)
+{
+   return (1 << numBits) + prime_deltas[numBits];
+}
+
+/*
+    Returns the smallest integer n such that
+    primeForNumBits(n) >= hint.
+*/
+static int countBits(int hint)
+{
+   int numBits = 0;
+   int bits = hint;
+
+   while (bits > 1) {
+      bits >>= 1;
+      numBits++;
+   }
+
+   if (numBits >= (int)sizeof(prime_deltas)) {
+      numBits = sizeof(prime_deltas) - 1;
+   } else if (primeForNumBits(numBits) < hint) {
+      ++numBits;
+   }
+   return numBits;
+}
+
+struct cso_node {
+   struct cso_node *next;
+   unsigned key;
+   void *value;
+};
+
+struct cso_hash_data {
+   struct cso_node *fakeNext;
+   struct cso_node **buckets;
+   int size;
+   int nodeSize;
+   short userNumBits;
+   short numBits;
+   int numBuckets;
+};
+
+struct cso_hash {
+   union {
+      struct cso_hash_data *d;
+      struct cso_node      *e;
+   } data;
+};
+
+static void *cso_data_allocate_node(struct cso_hash_data *hash)
+{
+   return MALLOC(hash->nodeSize);
+}
+
+static void cso_free_node(struct cso_node *node)
+{
+   FREE(node);
+}
+
+static struct cso_node *
+cso_hash_create_node(struct cso_hash *hash,
+                      unsigned akey, void *avalue,
+                      struct cso_node **anextNode)
+{
+   struct cso_node *node = cso_data_allocate_node(hash->data.d);
+
+   if (!node)
+      return NULL;
+
+   node->key = akey;
+   node->value = avalue;
+
+   node->next = (struct cso_node*)(*anextNode);
+   *anextNode = node;
+   ++hash->data.d->size;
+   return node;
+}
+
+static void cso_data_rehash(struct cso_hash_data *hash, int hint)
+{
+   if (hint < 0) {
+      hint = countBits(-hint);
+      if (hint < MinNumBits)
+         hint = MinNumBits;
+      hash->userNumBits = (short)hint;
+      while (primeForNumBits(hint) < (hash->size >> 1))
+         ++hint;
+   } else if (hint < MinNumBits) {
+      hint = MinNumBits;
+   }
+
+   if (hash->numBits != hint) {
+      struct cso_node *e = (struct cso_node *)(hash);
+      struct cso_node **oldBuckets = hash->buckets;
+      int oldNumBuckets = hash->numBuckets;
+      int  i = 0;
+
+      hash->numBits = (short)hint;
+      hash->numBuckets = primeForNumBits(hint);
+      hash->buckets = MALLOC(sizeof(struct cso_node*) * hash->numBuckets);
+      for (i = 0; i < hash->numBuckets; ++i)
+         hash->buckets[i] = e;
+
+      for (i = 0; i < oldNumBuckets; ++i) {
+         struct cso_node *firstNode = oldBuckets[i];
+         while (firstNode != e) {
+            unsigned h = firstNode->key;
+            struct cso_node *lastNode = firstNode;
+            struct cso_node *afterLastNode;
+            struct cso_node **beforeFirstNode;
+            
+            while (lastNode->next != e && lastNode->next->key == h)
+               lastNode = lastNode->next;
+
+            afterLastNode = lastNode->next;
+            beforeFirstNode = &hash->buckets[h % hash->numBuckets];
+            while (*beforeFirstNode != e)
+               beforeFirstNode = &(*beforeFirstNode)->next;
+            lastNode->next = *beforeFirstNode;
+            *beforeFirstNode = firstNode;
+            firstNode = afterLastNode;
+         }
+      }
+      FREE(oldBuckets);
+   }
+}
+
+static void cso_data_might_grow(struct cso_hash_data *hash)
+{
+   if (hash->size >= hash->numBuckets)
+      cso_data_rehash(hash, hash->numBits + 1);
+}
+
+static void cso_data_has_shrunk(struct cso_hash_data *hash)
+{
+   if (hash->size <= (hash->numBuckets >> 3) &&
+       hash->numBits > hash->userNumBits) {
+      int max = MAX(hash->numBits-2, hash->userNumBits);
+      cso_data_rehash(hash,  max);
+   }
+}
+
+static struct cso_node *cso_data_first_node(struct cso_hash_data *hash)
+{
+   struct cso_node *e = (struct cso_node *)(hash);
+   struct cso_node **bucket = hash->buckets;
+   int n = hash->numBuckets;
+   while (n--) {
+      if (*bucket != e)
+         return *bucket;
+      ++bucket;
+   }
+   return e;
+}
+
+static struct cso_node **cso_hash_find_node(struct cso_hash *hash, unsigned akey)
+{
+   struct cso_node **node;
+
+   if (hash->data.d->numBuckets) {
+      node = (struct cso_node **)(&hash->data.d->buckets[akey % hash->data.d->numBuckets]);
+      assert(*node == hash->data.e || (*node)->next);
+      while (*node != hash->data.e && (*node)->key != akey)
+         node = &(*node)->next;
+   } else {
+      node = (struct cso_node **)((const struct cso_node * const *)(&hash->data.e));
+   }
+   return node;
+}
+
+struct cso_hash_iter cso_hash_insert(struct cso_hash *hash,
+                                       unsigned key, void *data)
+{
+   cso_data_might_grow(hash->data.d);
+
+   {
+      struct cso_node **nextNode = cso_hash_find_node(hash, key);
+      struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode);
+      if (!node) {
+         struct cso_hash_iter null_iter = {hash, 0};
+         return null_iter;
+      }
+
+      {
+         struct cso_hash_iter iter = {hash, node};
+         return iter;
+      }
+   }
+}
+
+struct cso_hash * cso_hash_create(void)
+{
+   struct cso_hash *hash = MALLOC_STRUCT(cso_hash);
+   if (!hash)
+      return NULL;
+
+   hash->data.d = MALLOC_STRUCT(cso_hash_data);
+   if (!hash->data.d) {
+      FREE(hash);
+      return NULL;
+   }
+
+   hash->data.d->fakeNext = 0;
+   hash->data.d->buckets = 0;
+   hash->data.d->size = 0;
+   hash->data.d->nodeSize = sizeof(struct cso_node);
+   hash->data.d->userNumBits = (short)MinNumBits;
+   hash->data.d->numBits = 0;
+   hash->data.d->numBuckets = 0;
+
+   return hash;
+}
+
+void cso_hash_delete(struct cso_hash *hash)
+{
+   struct cso_node *e_for_x = (struct cso_node *)(hash->data.d);
+   struct cso_node **bucket = (struct cso_node **)(hash->data.d->buckets);
+   int n = hash->data.d->numBuckets;
+   while (n--) {
+      struct cso_node *cur = *bucket++;
+      while (cur != e_for_x) {
+         struct cso_node *next = cur->next;
+         cso_free_node(cur);
+         cur = next;
+      }
+   }
+   FREE(hash->data.d->buckets);
+   FREE(hash->data.d);
+   FREE(hash);
+}
+
+struct cso_hash_iter cso_hash_find(struct cso_hash *hash,
+                                     unsigned key)
+{
+   struct cso_node **nextNode = cso_hash_find_node(hash, key);
+   struct cso_hash_iter iter = {hash, *nextNode};
+   return iter;
+}
+
+unsigned cso_hash_iter_key(struct cso_hash_iter iter)
+{
+   if (!iter.node || iter.hash->data.e == iter.node)
+      return 0;
+   return iter.node->key;
+}
+
+void * cso_hash_iter_data(struct cso_hash_iter iter)
+{
+   if (!iter.node || iter.hash->data.e == iter.node)
+      return 0;
+   return iter.node->value;
+}
+
+static struct cso_node *cso_hash_data_next(struct cso_node *node)
+{
+   union {
+      struct cso_node *next;
+      struct cso_node *e;
+      struct cso_hash_data *d;
+   } a;
+   int start;
+   struct cso_node **bucket;
+   int n;
+
+   a.next = node->next;
+   if (!a.next) {
+      debug_printf("iterating beyond the last element\n");
+      return 0;
+   }
+   if (a.next->next)
+      return a.next;
+
+   start = (node->key % a.d->numBuckets) + 1;
+   bucket = a.d->buckets + start;
+   n = a.d->numBuckets - start;
+   while (n--) {
+      if (*bucket != a.e)
+         return *bucket;
+      ++bucket;
+   }
+   return a.e;
+}
+
+
+static struct cso_node *cso_hash_data_prev(struct cso_node *node)
+{
+   union {
+      struct cso_node *e;
+      struct cso_hash_data *d;
+   } a;
+   int start;
+   struct cso_node *sentinel;
+   struct cso_node **bucket;
+
+   a.e = node;
+   while (a.e->next)
+      a.e = a.e->next;
+
+   if (node == a.e)
+      start = a.d->numBuckets - 1;
+   else
+      start = node->key % a.d->numBuckets;
+
+   sentinel = node;
+   bucket = a.d->buckets + start;
+   while (start >= 0) {
+      if (*bucket != sentinel) {
+         struct cso_node *prev = *bucket;
+         while (prev->next != sentinel)
+            prev = prev->next;
+         return prev;
+      }
+
+      sentinel = a.e;
+      --bucket;
+      --start;
+   }
+   debug_printf("iterating backward beyond first element\n");
+   return a.e;
+}
+
+struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter)
+{
+   struct cso_hash_iter next = {iter.hash, cso_hash_data_next(iter.node)};
+   return next;
+}
+
+int cso_hash_iter_is_null(struct cso_hash_iter iter)
+{
+   if (!iter.node || iter.node == iter.hash->data.e)
+      return 1;
+   return 0;
+}
+
+void * cso_hash_take(struct cso_hash *hash,
+                      unsigned akey)
+{
+   struct cso_node **node = cso_hash_find_node(hash, akey);
+   if (*node != hash->data.e) {
+      void *t = (*node)->value;
+      struct cso_node *next = (*node)->next;
+      cso_free_node(*node);
+      *node = next;
+      --hash->data.d->size;
+      cso_data_has_shrunk(hash->data.d);
+      return t;
+   }
+   return 0;
+}
+
+struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter)
+{
+   struct cso_hash_iter prev = {iter.hash,
+                                 cso_hash_data_prev(iter.node)};
+   return prev;
+}
+
+struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash)
+{
+   struct cso_hash_iter iter = {hash, cso_data_first_node(hash->data.d)};
+   return iter;
+}
+
+int cso_hash_size(struct cso_hash *hash)
+{
+   return hash->data.d->size;
+}
+
+struct cso_hash_iter cso_hash_erase(struct cso_hash *hash, struct cso_hash_iter iter)
+{
+   struct cso_hash_iter ret = iter;
+   struct cso_node *node = iter.node;
+   struct cso_node **node_ptr;
+
+   if (node == hash->data.e)
+      return iter;
+
+   ret = cso_hash_iter_next(ret);
+   node_ptr = (struct cso_node**)(&hash->data.d->buckets[node->key % hash->data.d->numBuckets]);
+   while (*node_ptr != node)
+      node_ptr = &(*node_ptr)->next;
+   *node_ptr = node->next;
+   cso_free_node(node);
+   --hash->data.d->size;
+   return ret;
+}
+
+boolean cso_hash_contains(struct cso_hash *hash, unsigned key)
+{
+   struct cso_node **node = cso_hash_find_node(hash, key);
+   return (*node != hash->data.e);
+}
diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h
new file mode 100644
index 0000000000..5891c325fa
--- /dev/null
+++ b/src/gallium/auxiliary/cso_cache/cso_hash.h
@@ -0,0 +1,129 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Hash table implementation.
+ * 
+ * This file provides a hash implementation that is capable of dealing
+ * with collisions. It stores colliding entries in linked list. All
+ * functions operating on the hash return an iterator. The iterator
+ * itself points to the collision list. If there wasn't any collision
+ * the list will have just one entry, otherwise client code should
+ * iterate over the entries to find the exact entry among ones that
+ * had the same key (e.g. memcmp could be used on the data to check
+ * that)
+ * 
+ * @author Zack Rusin <zack@tungstengraphics.com>
+ */
+
+#ifndef CSO_HASH_H
+#define CSO_HASH_H
+
+#include "pipe/p_compiler.h"
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+
+struct cso_hash;
+struct cso_node;
+
+
+struct cso_hash_iter {
+   struct cso_hash *hash;
+   struct cso_node  *node;
+};
+
+
+struct cso_hash *cso_hash_create(void);
+void             cso_hash_delete(struct cso_hash *hash);
+
+
+int              cso_hash_size(struct cso_hash *hash);
+
+
+/**
+ * Adds a data with the given key to the hash. If entry with the given
+ * key is already in the hash, this current entry is instered before it
+ * in the collision list.
+ * Function returns iterator pointing to the inserted item in the hash.
+ */
+struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key,
+                                     void *data);
+/**
+ * Removes the item pointed to by the current iterator from the hash.
+ * Note that the data itself is not erased and if it was a malloc'ed pointer
+ * it will have to be freed after calling this function by the callee.
+ * Function returns iterator pointing to the item after the removed one in
+ * the hash.
+ */
+struct cso_hash_iter cso_hash_erase(struct cso_hash *hash, struct cso_hash_iter iter);
+
+void  *cso_hash_take(struct cso_hash *hash, unsigned key);
+
+
+
+struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash);
+
+/**
+ * Return an iterator pointing to the first entry in the collision list.
+ */
+struct cso_hash_iter cso_hash_find(struct cso_hash *hash, unsigned key);
+
+/**
+ * Returns true if a value with the given key exists in the hash
+ */
+boolean   cso_hash_contains(struct cso_hash *hash, unsigned key);
+
+
+int       cso_hash_iter_is_null(struct cso_hash_iter iter);
+unsigned  cso_hash_iter_key(struct cso_hash_iter iter);
+void     *cso_hash_iter_data(struct cso_hash_iter iter);
+
+
+struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter);
+struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter);
+
+
+/**
+ * Convenience routine to iterate over the collision list while doing a memory
+ * comparison to see which entry in the list is a direct copy of our template
+ * and returns that entry.
+ */
+void *cso_hash_find_data_from_template( struct cso_hash *hash,
+				        unsigned hash_key,
+				        void *templ,
+				        int size );
+
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile
new file mode 100644
index 0000000000..bdbf5a08ed
--- /dev/null
+++ b/src/gallium/auxiliary/draw/Makefile
@@ -0,0 +1,50 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = draw
+
+C_SOURCES = \
+	draw_context.c \
+	draw_pipe.c \
+	draw_pipe_aaline.c \
+	draw_pipe_aapoint.c \
+	draw_pipe_clip.c \
+	draw_pipe_cull.c \
+	draw_pipe_flatshade.c \
+	draw_pipe_offset.c \
+	draw_pipe_pstipple.c \
+	draw_pipe_stipple.c \
+	draw_pipe_twoside.c \
+	draw_pipe_unfilled.c \
+	draw_pipe_util.c \
+	draw_pipe_validate.c \
+	draw_pipe_vbuf.c \
+	draw_pipe_wide_line.c \
+	draw_pipe_wide_point.c \
+	draw_pt.c \
+	draw_pt_elts.c \
+	draw_pt_emit.c \
+	draw_pt_fetch.c \
+	draw_pt_fetch_emit.c \
+	draw_pt_fetch_shade_emit.c \
+	draw_pt_fetch_shade_pipeline.c \
+	draw_pt_post_vs.c \
+        draw_pt_util.c \
+        draw_pt_varray.c \
+	draw_pt_vcache.c \
+	draw_vertex.c \
+	draw_vs.c \
+	draw_vs_varient.c \
+	draw_vs_aos.c \
+	draw_vs_aos_io.c \
+	draw_vs_aos_machine.c \
+	draw_vs_exec.c \
+	draw_vs_llvm.c \
+	draw_vs_ppc.c  \
+	draw_vs_sse.c 
+
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript
new file mode 100644
index 0000000000..5f05aa324a
--- /dev/null
+++ b/src/gallium/auxiliary/draw/SConscript
@@ -0,0 +1,46 @@
+Import('*')
+
+draw = env.ConvenienceLibrary(
+	target = 'draw',
+	source = [
+		'draw_context.c',
+		'draw_pipe.c',
+		'draw_pipe_aaline.c',
+		'draw_pipe_aapoint.c',
+		'draw_pipe_clip.c',
+		'draw_pipe_cull.c',
+		'draw_pipe_flatshade.c',
+		'draw_pipe_offset.c',
+		'draw_pipe_pstipple.c',
+		'draw_pipe_stipple.c',
+		'draw_pipe_twoside.c',
+		'draw_pipe_unfilled.c',
+		'draw_pipe_util.c',
+		'draw_pipe_validate.c',
+		'draw_pipe_vbuf.c',
+		'draw_pipe_wide_line.c',
+		'draw_pipe_wide_point.c',
+		'draw_pt.c',
+		'draw_pt_elts.c',
+		'draw_pt_emit.c',
+		'draw_pt_fetch.c',
+		'draw_pt_fetch_emit.c',
+		'draw_pt_fetch_shade_emit.c',
+		'draw_pt_fetch_shade_pipeline.c',
+		'draw_pt_post_vs.c',
+		'draw_pt_util.c',
+		'draw_pt_varray.c',
+		'draw_pt_vcache.c',
+		'draw_vertex.c',
+		'draw_vs.c',
+		'draw_vs_aos.c',
+		'draw_vs_aos_io.c',
+		'draw_vs_aos_machine.c',
+		'draw_vs_exec.c',
+		'draw_vs_llvm.c',
+		'draw_vs_ppc.c',
+		'draw_vs_sse.c',
+		'draw_vs_varient.c'
+	])
+
+auxiliaries.insert(0, draw)
diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
new file mode 100644
index 0000000000..7bd4a2e221
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -0,0 +1,429 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+#include "draw_context.h"
+#include "draw_vbuf.h"
+#include "draw_vs.h"
+#include "draw_pt.h"
+#include "draw_pipe.h"
+
+
+struct draw_context *draw_create( void )
+{
+   struct draw_context *draw = CALLOC_STRUCT( draw_context );
+   if (draw == NULL)
+      goto fail;
+
+   ASSIGN_4V( draw->plane[0], -1,  0,  0, 1 );
+   ASSIGN_4V( draw->plane[1],  1,  0,  0, 1 );
+   ASSIGN_4V( draw->plane[2],  0, -1,  0, 1 );
+   ASSIGN_4V( draw->plane[3],  0,  1,  0, 1 );
+   ASSIGN_4V( draw->plane[4],  0,  0,  1, 1 ); /* yes these are correct */
+   ASSIGN_4V( draw->plane[5],  0,  0, -1, 1 ); /* mesa's a bit wonky */
+   draw->nr_planes = 6;
+
+
+   draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */
+
+
+   if (!draw_pipeline_init( draw ))
+      goto fail;
+
+   if (!draw_pt_init( draw ))
+      goto fail;
+
+   if (!draw_vs_init( draw ))
+      goto fail;
+
+   return draw;
+
+fail:
+   draw_destroy( draw );   
+   return NULL;
+}
+
+
+void draw_destroy( struct draw_context *draw )
+{
+   if (!draw)
+      return;
+
+
+
+   /* Not so fast -- we're just borrowing this at the moment.
+    * 
+   if (draw->render)
+      draw->render->destroy( draw->render );
+   */
+
+   draw_pipeline_destroy( draw );
+   draw_pt_destroy( draw );
+   draw_vs_destroy( draw );
+
+   FREE( draw );
+}
+
+
+
+void draw_flush( struct draw_context *draw )
+{
+   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+}
+
+
+/**
+ * Specify the Minimum Resolvable Depth factor for polygon offset.
+ * This factor potentially depends on the number of Z buffer bits,
+ * the rasterization algorithm and the arithmetic performed on Z
+ * values between vertex shading and rasterization.  It will vary
+ * from one driver to another.
+ */
+void draw_set_mrd(struct draw_context *draw, double mrd)
+{
+   draw->mrd = mrd;
+}
+
+
+/**
+ * Register new primitive rasterization/rendering state.
+ * This causes the drawing pipeline to be rebuilt.
+ */
+void draw_set_rasterizer_state( struct draw_context *draw,
+                                const struct pipe_rasterizer_state *raster )
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+
+   draw->rasterizer = raster;
+   draw->bypass_clipping =
+      ((draw->rasterizer && draw->rasterizer->bypass_clipping) ||
+       draw->driver.bypass_clipping);
+}
+
+
+void draw_set_driver_clipping( struct draw_context *draw,
+                               boolean bypass_clipping )
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+
+   draw->driver.bypass_clipping = bypass_clipping;
+   draw->bypass_clipping = (draw->rasterizer->bypass_clipping || 
+                            draw->driver.bypass_clipping);
+}
+
+
+/** 
+ * Plug in the primitive rendering/rasterization stage (which is the last
+ * stage in the drawing pipeline).
+ * This is provided by the device driver.
+ */
+void draw_set_rasterize_stage( struct draw_context *draw,
+                               struct draw_stage *stage )
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+
+   draw->pipeline.rasterize = stage;
+}
+
+
+/**
+ * Set the draw module's clipping state.
+ */
+void draw_set_clip_state( struct draw_context *draw,
+                          const struct pipe_clip_state *clip )
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+
+   assert(clip->nr <= PIPE_MAX_CLIP_PLANES);
+   memcpy(&draw->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0]));
+   draw->nr_planes = 6 + clip->nr;
+}
+
+
+/**
+ * Set the draw module's viewport state.
+ */
+void draw_set_viewport_state( struct draw_context *draw,
+                              const struct pipe_viewport_state *viewport )
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+   draw->viewport = *viewport; /* struct copy */
+   draw->identity_viewport = (viewport->scale[0] == 1.0f &&
+                              viewport->scale[1] == 1.0f &&
+                              viewport->scale[2] == 1.0f &&
+                              viewport->scale[3] == 1.0f &&
+                              viewport->translate[0] == 0.0f &&
+                              viewport->translate[1] == 0.0f &&
+                              viewport->translate[2] == 0.0f &&
+                              viewport->translate[3] == 0.0f);
+
+   draw_vs_set_viewport( draw, viewport );
+}
+
+
+
+void
+draw_set_vertex_buffers(struct draw_context *draw,
+                        unsigned count,
+                        const struct pipe_vertex_buffer *buffers)
+{
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(draw->pt.vertex_buffer, buffers, count * sizeof(buffers[0]));
+   draw->pt.nr_vertex_buffers = count;
+}
+
+
+void
+draw_set_vertex_elements(struct draw_context *draw,
+                         unsigned count,
+                         const struct pipe_vertex_element *elements)
+{
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(draw->pt.vertex_element, elements, count * sizeof(elements[0]));
+   draw->pt.nr_vertex_elements = count;
+}
+
+
+/**
+ * Tell drawing context where to find mapped vertex buffers.
+ */
+void
+draw_set_mapped_vertex_buffer(struct draw_context *draw,
+                              unsigned attr, const void *buffer)
+{
+   draw->pt.user.vbuffer[attr] = buffer;
+}
+
+
+void
+draw_set_mapped_constant_buffer(struct draw_context *draw,
+                                const void *buffer, 
+                                unsigned size )
+{
+   draw->pt.user.constants = buffer;
+   draw_vs_set_constants( draw, (const float (*)[4])buffer, size );
+}
+
+
+/**
+ * Tells the draw module to draw points with triangles if their size
+ * is greater than this threshold.
+ */
+void
+draw_wide_point_threshold(struct draw_context *draw, float threshold)
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+   draw->pipeline.wide_point_threshold = threshold;
+}
+
+
+/**
+ * Tells the draw module to draw lines with triangles if their width
+ * is greater than this threshold.
+ */
+void
+draw_wide_line_threshold(struct draw_context *draw, float threshold)
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+   draw->pipeline.wide_line_threshold = threshold;
+}
+
+
+/**
+ * Tells the draw module whether or not to implement line stipple.
+ */
+void
+draw_enable_line_stipple(struct draw_context *draw, boolean enable)
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+   draw->pipeline.line_stipple = enable;
+}
+
+
+/**
+ * Tells draw module whether to convert points to quads for sprite mode.
+ */
+void
+draw_enable_point_sprites(struct draw_context *draw, boolean enable)
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+   draw->pipeline.point_sprite = enable;
+}
+
+
+void
+draw_set_force_passthrough( struct draw_context *draw, boolean enable )
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+   draw->force_passthrough = enable;
+}
+
+
+/**
+ * Ask the draw module for the location/slot of the given vertex attribute in
+ * a post-transformed vertex.
+ *
+ * With this function, drivers that use the draw module should have no reason
+ * to track the current vertex shader.
+ *
+ * Note that the draw module may sometimes generate vertices with extra
+ * attributes (such as texcoords for AA lines).  The driver can call this
+ * function to find those attributes.
+ *
+ * Zero is returned if the attribute is not found since this is
+ * a don't care / undefined situtation.  Returning -1 would be a bit more
+ * work for the drivers.
+ */
+int
+draw_find_vs_output(const struct draw_context *draw,
+                    uint semantic_name, uint semantic_index)
+{
+   const struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+   uint i;
+   for (i = 0; i < vs->info.num_outputs; i++) {
+      if (vs->info.output_semantic_name[i] == semantic_name &&
+          vs->info.output_semantic_index[i] == semantic_index)
+         return i;
+   }
+
+   /* XXX there may be more than one extra vertex attrib.
+    * For example, simulated gl_FragCoord and gl_PointCoord.
+    */
+   if (draw->extra_vp_outputs.semantic_name == semantic_name &&
+       draw->extra_vp_outputs.semantic_index == semantic_index) {
+      return draw->extra_vp_outputs.slot;
+   }
+   return 0;
+}
+
+
+/**
+ * Return number of vertex shader outputs.
+ */
+uint
+draw_num_vs_outputs(const struct draw_context *draw)
+{
+   uint count = draw->vs.vertex_shader->info.num_outputs;
+   if (draw->extra_vp_outputs.slot > 0)
+      count++;
+   return count;
+}
+
+
+/**
+ * Provide TGSI sampler objects for vertex shaders that use texture fetches.
+ * This might only be used by software drivers for the time being.
+ */
+void
+draw_texture_samplers(struct draw_context *draw,
+                      uint num_samplers,
+                      struct tgsi_sampler **samplers)
+{
+   draw->vs.num_samplers = num_samplers;
+   draw->vs.samplers = samplers;
+}
+
+
+
+
+void draw_set_render( struct draw_context *draw, 
+		      struct vbuf_render *render )
+{
+   draw->render = render;
+}
+
+void draw_set_edgeflags( struct draw_context *draw,
+                         const unsigned *edgeflag )
+{
+   draw->pt.user.edgeflag = edgeflag;
+}
+
+
+
+
+/**
+ * Tell the drawing context about the index/element buffer to use
+ * (ala glDrawElements)
+ * If no element buffer is to be used (i.e. glDrawArrays) then this
+ * should be called with eltSize=0 and elements=NULL.
+ *
+ * \param draw  the drawing context
+ * \param eltSize  size of each element (1, 2 or 4 bytes)
+ * \param elements  the element buffer ptr
+ */
+void
+draw_set_mapped_element_buffer_range( struct draw_context *draw,
+                                      unsigned eltSize,
+                                      unsigned min_index,
+                                      unsigned max_index,
+                                      const void *elements )
+{
+   draw->pt.user.elts = elements;
+   draw->pt.user.eltSize = eltSize;
+   draw->pt.user.min_index = min_index;
+   draw->pt.user.max_index = max_index;
+}
+
+
+void
+draw_set_mapped_element_buffer( struct draw_context *draw,
+                                unsigned eltSize,
+                                const void *elements )
+{
+   draw->pt.user.elts = elements;
+   draw->pt.user.eltSize = eltSize;
+   draw->pt.user.min_index = 0;
+   draw->pt.user.max_index = 0xffffffff;
+}
+
+ 
+/* Revamp me please:
+ */
+void draw_do_flush( struct draw_context *draw, unsigned flags )
+{
+   if (!draw->suspend_flushing)
+   {
+      assert(!draw->flushing); /* catch inadvertant recursion */
+
+      draw->flushing = TRUE;
+
+      draw_pipeline_flush( draw, flags );
+
+      draw->reduced_prim = ~0; /* is reduced_prim needed any more? */
+      
+      draw->flushing = FALSE;
+   }
+}
diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h
new file mode 100644
index 0000000000..d529e4e9a2
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_context.h
@@ -0,0 +1,182 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief  Public interface into the drawing module.
+ */
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#ifndef DRAW_CONTEXT_H
+#define DRAW_CONTEXT_H
+
+
+#include "pipe/p_state.h"
+
+
+struct pipe_context;
+struct draw_context;
+struct draw_stage;
+struct draw_vertex_shader;
+struct tgsi_sampler;
+
+
+struct draw_context *draw_create( void );
+
+void draw_destroy( struct draw_context *draw );
+
+void draw_set_viewport_state( struct draw_context *draw,
+                              const struct pipe_viewport_state *viewport );
+
+void draw_set_clip_state( struct draw_context *pipe,
+                          const struct pipe_clip_state *clip );
+
+void draw_set_rasterizer_state( struct draw_context *draw,
+                                const struct pipe_rasterizer_state *raster );
+
+void draw_set_rasterize_stage( struct draw_context *draw,
+                               struct draw_stage *stage );
+
+void draw_wide_point_threshold(struct draw_context *draw, float threshold);
+
+void draw_wide_line_threshold(struct draw_context *draw, float threshold);
+
+void draw_enable_line_stipple(struct draw_context *draw, boolean enable);
+
+void draw_enable_point_sprites(struct draw_context *draw, boolean enable);
+
+void draw_set_mrd(struct draw_context *draw, double mrd);
+
+boolean
+draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe);
+
+boolean
+draw_install_aapoint_stage(struct draw_context *draw, struct pipe_context *pipe);
+
+boolean
+draw_install_pstipple_stage(struct draw_context *draw, struct pipe_context *pipe);
+
+
+int
+draw_find_vs_output(const struct draw_context *draw,
+                    uint semantic_name, uint semantic_index);
+
+uint
+draw_num_vs_outputs(const struct draw_context *draw);
+
+
+void
+draw_texture_samplers(struct draw_context *draw,
+                      uint num_samplers,
+                      struct tgsi_sampler **samplers);
+
+
+
+/*
+ * Vertex shader functions
+ */
+
+struct draw_vertex_shader *
+draw_create_vertex_shader(struct draw_context *draw,
+                          const struct pipe_shader_state *shader);
+void draw_bind_vertex_shader(struct draw_context *draw,
+                             struct draw_vertex_shader *dvs);
+void draw_delete_vertex_shader(struct draw_context *draw,
+                               struct draw_vertex_shader *dvs);
+
+
+
+/*
+ * Vertex data functions
+ */
+
+void draw_set_vertex_buffers(struct draw_context *draw,
+                             unsigned count,
+                             const struct pipe_vertex_buffer *buffers);
+
+void draw_set_vertex_elements(struct draw_context *draw,
+			      unsigned count,
+                              const struct pipe_vertex_element *elements);
+
+void
+draw_set_mapped_element_buffer_range( struct draw_context *draw,
+                                      unsigned eltSize,
+                                      unsigned min_index,
+                                      unsigned max_index,
+                                      const void *elements );
+
+void draw_set_mapped_element_buffer( struct draw_context *draw,
+                                     unsigned eltSize, 
+                                     const void *elements );
+
+void draw_set_mapped_vertex_buffer(struct draw_context *draw,
+                                   unsigned attr, const void *buffer);
+
+void draw_set_mapped_constant_buffer(struct draw_context *draw,
+                                     const void *buffer,
+                                     unsigned size );
+
+void draw_set_edgeflags( struct draw_context *draw,
+                         const unsigned *edgeflag );
+
+
+/***********************************************************************
+ * draw_prim.c 
+ */
+
+void draw_arrays(struct draw_context *draw, unsigned prim,
+		 unsigned start, unsigned count);
+
+void draw_flush(struct draw_context *draw);
+
+
+/*******************************************************************************
+ * Driver backend interface 
+ */
+struct vbuf_render;
+void draw_set_render( struct draw_context *draw, 
+		      struct vbuf_render *render );
+
+void draw_set_driver_clipping( struct draw_context *draw,
+                               boolean bypass_clipping );
+
+void draw_set_force_passthrough( struct draw_context *draw, 
+                                 boolean enable );
+
+/*******************************************************************************
+ * Draw pipeline 
+ */
+boolean draw_need_pipeline(const struct draw_context *draw,
+                           const struct pipe_rasterizer_state *rasterizer,
+                           unsigned prim );
+
+
+
+#endif /* DRAW_CONTEXT_H */
diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c
new file mode 100644
index 0000000000..3cde9d36d3
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe.c
@@ -0,0 +1,287 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "draw/draw_private.h"
+#include "draw/draw_pipe.h"
+
+
+
+boolean draw_pipeline_init( struct draw_context *draw )
+{
+   /* create pipeline stages */
+   draw->pipeline.wide_line  = draw_wide_line_stage( draw );
+   draw->pipeline.wide_point = draw_wide_point_stage( draw );
+   draw->pipeline.stipple   = draw_stipple_stage( draw );
+   draw->pipeline.unfilled  = draw_unfilled_stage( draw );
+   draw->pipeline.twoside   = draw_twoside_stage( draw );
+   draw->pipeline.offset    = draw_offset_stage( draw );
+   draw->pipeline.clip      = draw_clip_stage( draw );
+   draw->pipeline.flatshade = draw_flatshade_stage( draw );
+   draw->pipeline.cull      = draw_cull_stage( draw );
+   draw->pipeline.validate  = draw_validate_stage( draw );
+   draw->pipeline.first     = draw->pipeline.validate;
+
+   if (!draw->pipeline.wide_line ||
+       !draw->pipeline.wide_point ||
+       !draw->pipeline.stipple ||
+       !draw->pipeline.unfilled ||
+       !draw->pipeline.twoside ||
+       !draw->pipeline.offset ||
+       !draw->pipeline.clip ||
+       !draw->pipeline.flatshade ||
+       !draw->pipeline.cull ||
+       !draw->pipeline.validate)
+      return FALSE;
+
+   /* these defaults are oriented toward the needs of softpipe */
+   draw->pipeline.wide_point_threshold = 1000000.0; /* infinity */
+   draw->pipeline.wide_line_threshold = 1.0;
+   draw->pipeline.line_stipple = TRUE;
+   draw->pipeline.point_sprite = TRUE;
+
+   return TRUE;
+}
+
+
+void draw_pipeline_destroy( struct draw_context *draw )
+{
+   if (draw->pipeline.wide_line)
+      draw->pipeline.wide_line->destroy( draw->pipeline.wide_line );
+   if (draw->pipeline.wide_point)
+      draw->pipeline.wide_point->destroy( draw->pipeline.wide_point );
+   if (draw->pipeline.stipple)
+      draw->pipeline.stipple->destroy( draw->pipeline.stipple );
+   if (draw->pipeline.unfilled)
+      draw->pipeline.unfilled->destroy( draw->pipeline.unfilled );
+   if (draw->pipeline.twoside)
+      draw->pipeline.twoside->destroy( draw->pipeline.twoside );
+   if (draw->pipeline.offset)
+      draw->pipeline.offset->destroy( draw->pipeline.offset );
+   if (draw->pipeline.clip)
+      draw->pipeline.clip->destroy( draw->pipeline.clip );
+   if (draw->pipeline.flatshade)
+      draw->pipeline.flatshade->destroy( draw->pipeline.flatshade );
+   if (draw->pipeline.cull)
+      draw->pipeline.cull->destroy( draw->pipeline.cull );
+   if (draw->pipeline.validate)
+      draw->pipeline.validate->destroy( draw->pipeline.validate );
+   if (draw->pipeline.aaline)
+      draw->pipeline.aaline->destroy( draw->pipeline.aaline );
+   if (draw->pipeline.aapoint)
+      draw->pipeline.aapoint->destroy( draw->pipeline.aapoint );
+   if (draw->pipeline.pstipple)
+      draw->pipeline.pstipple->destroy( draw->pipeline.pstipple );
+   if (draw->pipeline.rasterize)
+      draw->pipeline.rasterize->destroy( draw->pipeline.rasterize );
+}
+
+
+
+
+
+
+
+static void do_point( struct draw_context *draw,
+		      const char *v0 )
+{
+   struct prim_header prim;
+   
+   prim.flags = 0;
+   prim.pad = 0;
+   prim.v[0] = (struct vertex_header *)v0;
+
+   draw->pipeline.first->point( draw->pipeline.first, &prim );
+}
+
+
+static void do_line( struct draw_context *draw,
+                     ushort flags,
+		     const char *v0,
+		     const char *v1 )
+{
+   struct prim_header prim;
+   
+   prim.flags = flags;
+   prim.pad = 0;
+   prim.v[0] = (struct vertex_header *)v0;
+   prim.v[1] = (struct vertex_header *)v1;
+
+   draw->pipeline.first->line( draw->pipeline.first, &prim );
+}
+
+
+static void do_triangle( struct draw_context *draw,
+                         ushort flags,
+			 char *v0,
+			 char *v1,
+			 char *v2 )
+{
+   struct prim_header prim;
+   
+   prim.v[0] = (struct vertex_header *)v0;
+   prim.v[1] = (struct vertex_header *)v1;
+   prim.v[2] = (struct vertex_header *)v2;
+   prim.flags = flags;
+   prim.pad = 0;
+
+   draw->pipeline.first->tri( draw->pipeline.first, &prim );
+}
+
+
+
+
+/* Code to run the pipeline on a fairly arbitary collection of vertices.
+ *
+ * Vertex headers must be pre-initialized with the
+ * UNDEFINED_VERTEX_ID, this code will cause that id to become
+ * overwritten, so it may have to be reset if there is the intention
+ * to reuse the vertices.
+ *
+ * This code provides a callback to reset the vertex id's which the
+ * draw_vbuf.c code uses when it has to perform a flush.
+ */
+void draw_pipeline_run( struct draw_context *draw,
+                        unsigned prim,
+                        struct vertex_header *vertices,
+                        unsigned vertex_count,
+                        unsigned stride,
+                        const ushort *elts,
+                        unsigned count )
+{
+   char *verts = (char *)vertices;
+   unsigned i;
+
+   draw->pipeline.verts = verts;
+   draw->pipeline.vertex_stride = stride;
+   draw->pipeline.vertex_count = vertex_count;
+   
+   switch (prim) {
+   case PIPE_PRIM_POINTS:
+      for (i = 0; i < count; i++) 
+         do_point( draw, 
+                   verts + stride * elts[i] );
+      break;
+   case PIPE_PRIM_LINES:
+      for (i = 0; i+1 < count; i += 2) 
+         do_line( draw, 
+                  elts[i+0],  /* flags */
+                  verts + stride * (elts[i+0] & ~DRAW_PIPE_FLAG_MASK),
+                  verts + stride * elts[i+1]);
+      break;
+   case PIPE_PRIM_TRIANGLES:
+      for (i = 0; i+2 < count; i += 3)
+         do_triangle( draw, 
+                      elts[i+0],  /* flags */
+                      verts + stride * (elts[i+0] & ~DRAW_PIPE_FLAG_MASK),
+                      verts + stride * elts[i+1],
+                      verts + stride * elts[i+2]);
+      break;
+   }
+   
+   draw->pipeline.verts = NULL;
+   draw->pipeline.vertex_count = 0;
+}
+
+#define QUAD(i0,i1,i2,i3)                                        \
+   do_triangle( draw,                                            \
+                ( DRAW_PIPE_RESET_STIPPLE |                      \
+                  DRAW_PIPE_EDGE_FLAG_0 |                        \
+                  DRAW_PIPE_EDGE_FLAG_2 ),                       \
+                 verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \
+                 verts + stride * (i1),                          \
+                verts + stride * (i3));                          \
+      do_triangle( draw,                                         \
+                   ( DRAW_PIPE_EDGE_FLAG_0 |                     \
+                     DRAW_PIPE_EDGE_FLAG_1 ),                    \
+                 verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK), \
+                 verts + stride * (i2),                          \
+                 verts + stride * (i3))
+
+#define TRIANGLE(flags,i0,i1,i2)                                 \
+   do_triangle( draw,                                            \
+                flags,  /* flags */                              \
+                 verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \
+                 verts + stride * (i1),                          \
+                 verts + stride * (i2))
+
+#define LINE(flags,i0,i1)                                   \
+   do_line( draw,                                           \
+            flags,                                          \
+            verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \
+            verts + stride * (i1))
+
+#define POINT(i0)                               \
+   do_point( draw,                              \
+             verts + stride * i0 )
+
+#define FUNC pipe_run_linear
+#define ARGS                                    \
+    struct draw_context *draw,                  \
+    unsigned prim,                              \
+    struct vertex_header *vertices,             \
+    unsigned stride
+
+#define LOCAL_VARS                                           \
+   char *verts = (char *)vertices;                           \
+   boolean flatfirst = (draw->rasterizer->flatshade &&       \
+                        draw->rasterizer->flatshade_first);  \
+   unsigned i;                                               \
+   ushort flags
+
+#define FLUSH
+
+#include "draw_pt_decompose.h"
+
+void draw_pipeline_run_linear( struct draw_context *draw,
+                               unsigned prim,
+                               struct vertex_header *vertices,
+                               unsigned count,
+                               unsigned stride )
+{
+   char *verts = (char *)vertices;
+   draw->pipeline.verts = verts;
+   draw->pipeline.vertex_stride = stride;
+   draw->pipeline.vertex_count = count;
+
+   pipe_run_linear(draw, prim, vertices, stride, count);
+
+   draw->pipeline.verts = NULL;
+   draw->pipeline.vertex_count = 0;
+}
+
+
+void draw_pipeline_flush( struct draw_context *draw, 
+                          unsigned flags )
+{
+   draw->pipeline.first->flush( draw->pipeline.first, flags );
+   draw->pipeline.first = draw->pipeline.validate;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe.h b/src/gallium/auxiliary/draw/draw_pipe.h
new file mode 100644
index 0000000000..dbad8f98ac
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe.h
@@ -0,0 +1,125 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#ifndef DRAW_PIPE_H
+#define DRAW_PIPE_H
+
+#include "pipe/p_compiler.h"
+#include "draw_private.h"       /* for sizeof(vertex_header) */
+
+
+/**
+ * Basic info for a point/line/triangle primitive.
+ */
+struct prim_header {
+   float det;                 /**< front/back face determinant */
+   ushort flags;
+   ushort pad;
+   struct vertex_header *v[3];  /**< 1 to 3 vertex pointers */
+};
+
+
+
+/**
+ * Base class for all primitive drawing stages.
+ */
+struct draw_stage
+{
+   struct draw_context *draw;   /**< parent context */
+
+   struct draw_stage *next;     /**< next stage in pipeline */
+
+   struct vertex_header **tmp;  /**< temp vert storage, such as for clipping */
+   unsigned nr_tmps;
+
+   void (*point)( struct draw_stage *,
+		  struct prim_header * );
+
+   void (*line)( struct draw_stage *,
+		 struct prim_header * );
+
+   void (*tri)( struct draw_stage *,
+		struct prim_header * );
+
+   void (*flush)( struct draw_stage *,
+		  unsigned flags );
+
+   void (*reset_stipple_counter)( struct draw_stage * );
+
+   void (*destroy)( struct draw_stage * );
+};
+
+
+extern struct draw_stage *draw_unfilled_stage( struct draw_context *context );
+extern struct draw_stage *draw_twoside_stage( struct draw_context *context );
+extern struct draw_stage *draw_offset_stage( struct draw_context *context );
+extern struct draw_stage *draw_clip_stage( struct draw_context *context );
+extern struct draw_stage *draw_flatshade_stage( struct draw_context *context );
+extern struct draw_stage *draw_cull_stage( struct draw_context *context );
+extern struct draw_stage *draw_stipple_stage( struct draw_context *context );
+extern struct draw_stage *draw_wide_line_stage( struct draw_context *context );
+extern struct draw_stage *draw_wide_point_stage( struct draw_context *context );
+extern struct draw_stage *draw_validate_stage( struct draw_context *context );
+
+
+extern void draw_free_temp_verts( struct draw_stage *stage );
+extern boolean draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr );
+
+extern void draw_reset_vertex_ids( struct draw_context *draw );
+
+void draw_pipe_passthrough_tri(struct draw_stage *stage, struct prim_header *header);
+void draw_pipe_passthrough_line(struct draw_stage *stage, struct prim_header *header);
+void draw_pipe_passthrough_point(struct draw_stage *stage, struct prim_header *header);
+
+
+
+/**
+ * Get a writeable copy of a vertex.
+ * \param stage  drawing stage info
+ * \param vert  the vertex to copy (source)
+ * \param idx  index into stage's tmp[] array to put the copy (dest)
+ * \return  pointer to the copied vertex
+ */
+static INLINE struct vertex_header *
+dup_vert( struct draw_stage *stage,
+	  const struct vertex_header *vert,
+	  unsigned idx )
+{   
+   struct vertex_header *tmp = stage->tmp[idx];
+   const uint vsize = sizeof(struct vertex_header)
+      + stage->draw->vs.num_vs_outputs * 4 * sizeof(float);
+   memcpy(tmp, vert, vsize);
+   tmp->vertex_id = UNDEFINED_VERTEX_ID;
+   return tmp;
+}
+
+#endif
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
new file mode 100644
index 0000000000..20841bb5d6
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
@@ -0,0 +1,914 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * AA line stage:  AA lines are converted to texture mapped triangles.
+ *
+ * Authors:  Brian Paul
+ */
+
+
+#include "pipe/p_inlines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "tgsi/tgsi_transform.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "draw_context.h"
+#include "draw_private.h"
+#include "draw_pipe.h"
+
+
+/**
+ * Max texture level for the alpha texture used for antialiasing
+ */
+#define MAX_TEXTURE_LEVEL  5   /* 32 x 32 */
+
+
+/**
+ * Subclass of pipe_shader_state to carry extra fragment shader info.
+ */
+struct aaline_fragment_shader
+{
+   struct pipe_shader_state state;
+   void *driver_fs;
+   void *aaline_fs;
+   void *aapoint_fs; /* not yet */
+   void *sprite_fs; /* not yet */
+   uint sampler_unit;
+   int generic_attrib;  /**< texcoord/generic used for texture */
+};
+
+
+/**
+ * Subclass of draw_stage
+ */
+struct aaline_stage
+{
+   struct draw_stage stage;
+
+   float half_line_width;
+
+   /** For AA lines, this is the vertex attrib slot for the new texcoords */
+   uint tex_slot;
+   /** position, not necessarily output zero */
+   uint pos_slot;
+
+   void *sampler_cso;
+   struct pipe_texture *texture;
+   uint num_samplers;
+   uint num_textures;
+
+
+   /*
+    * Currently bound state
+    */
+   struct aaline_fragment_shader *fs;
+   struct {
+      void *sampler[PIPE_MAX_SAMPLERS];
+      struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
+   } state;
+
+   /*
+    * Driver interface/override functions
+    */
+   void * (*driver_create_fs_state)(struct pipe_context *,
+                                    const struct pipe_shader_state *);
+   void (*driver_bind_fs_state)(struct pipe_context *, void *);
+   void (*driver_delete_fs_state)(struct pipe_context *, void *);
+
+   void (*driver_bind_sampler_states)(struct pipe_context *, unsigned,
+                                      void **);
+   void (*driver_set_sampler_textures)(struct pipe_context *, unsigned,
+                                       struct pipe_texture **);
+
+   struct pipe_context *pipe;
+};
+
+
+
+/**
+ * Subclass of tgsi_transform_context, used for transforming the
+ * user's fragment shader to add the special AA instructions.
+ */
+struct aa_transform_context {
+   struct tgsi_transform_context base;
+   uint tempsUsed;  /**< bitmask */
+   int colorOutput; /**< which output is the primary color */
+   uint samplersUsed;  /**< bitfield of samplers used */
+   int freeSampler;  /** an available sampler for the pstipple */
+   int maxInput, maxGeneric;  /**< max input index found */
+   int colorTemp, texTemp;  /**< temp registers */
+   boolean firstInstruction;
+};
+
+
+/**
+ * TGSI declaration transform callback.
+ * Look for a free sampler, a free input attrib, and two free temp regs.
+ */
+static void
+aa_transform_decl(struct tgsi_transform_context *ctx,
+                  struct tgsi_full_declaration *decl)
+{
+   struct aa_transform_context *aactx = (struct aa_transform_context *) ctx;
+
+   if (decl->Declaration.File == TGSI_FILE_OUTPUT &&
+       decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR &&
+       decl->Semantic.SemanticIndex == 0) {
+      aactx->colorOutput = decl->DeclarationRange.First;
+   }
+   else if (decl->Declaration.File == TGSI_FILE_SAMPLER) {
+      uint i;
+      for (i = decl->DeclarationRange.First;
+           i <= decl->DeclarationRange.Last; i++) {
+         aactx->samplersUsed |= 1 << i;
+      }
+   }
+   else if (decl->Declaration.File == TGSI_FILE_INPUT) {
+      if ((int) decl->DeclarationRange.Last > aactx->maxInput)
+         aactx->maxInput = decl->DeclarationRange.Last;
+      if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC &&
+           (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) {
+         aactx->maxGeneric = decl->Semantic.SemanticIndex;
+      }
+   }
+   else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) {
+      uint i;
+      for (i = decl->DeclarationRange.First;
+           i <= decl->DeclarationRange.Last; i++) {
+         aactx->tempsUsed |= (1 << i);
+      }
+   }
+
+   ctx->emit_declaration(ctx, decl);
+}
+
+
+/**
+ * Find the lowest zero bit in the given word, or -1 if bitfield is all ones.
+ */
+static int
+free_bit(uint bitfield)
+{
+   int i;
+   for (i = 0; i < 32; i++) {
+      if ((bitfield & (1 << i)) == 0)
+         return i;
+   }
+   return -1;
+}
+
+
+/**
+ * TGSI instruction transform callback.
+ * Replace writes to result.color w/ a temp reg.
+ * Upon END instruction, insert texture sampling code for antialiasing.
+ */
+static void
+aa_transform_inst(struct tgsi_transform_context *ctx,
+                  struct tgsi_full_instruction *inst)
+{
+   struct aa_transform_context *aactx = (struct aa_transform_context *) ctx;
+
+   if (aactx->firstInstruction) {
+      /* emit our new declarations before the first instruction */
+
+      struct tgsi_full_declaration decl;
+      uint i;
+
+      /* find free sampler */
+      aactx->freeSampler = free_bit(aactx->samplersUsed);
+      if (aactx->freeSampler >= PIPE_MAX_SAMPLERS)
+         aactx->freeSampler = PIPE_MAX_SAMPLERS - 1;
+
+      /* find two free temp regs */
+      for (i = 0; i < 32; i++) {
+         if ((aactx->tempsUsed & (1 << i)) == 0) {
+            /* found a free temp */
+            if (aactx->colorTemp < 0)
+               aactx->colorTemp  = i;
+            else if (aactx->texTemp < 0)
+               aactx->texTemp  = i;
+            else
+               break;
+         }
+      }
+      assert(aactx->colorTemp >= 0);
+      assert(aactx->texTemp >= 0);
+
+      /* declare new generic input/texcoord */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_INPUT;
+      /* XXX this could be linear... */
+      decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE;
+      decl.Declaration.Semantic = 1;
+      decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
+      decl.Semantic.SemanticIndex = aactx->maxGeneric + 1;
+      decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = aactx->maxInput + 1;
+      ctx->emit_declaration(ctx, &decl);
+
+      /* declare new sampler */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_SAMPLER;
+      decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = aactx->freeSampler;
+      ctx->emit_declaration(ctx, &decl);
+
+      /* declare new temp regs */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_TEMPORARY;
+      decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = aactx->texTemp;
+      ctx->emit_declaration(ctx, &decl);
+
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_TEMPORARY;
+      decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = aactx->colorTemp;
+      ctx->emit_declaration(ctx, &decl);
+
+      aactx->firstInstruction = FALSE;
+   }
+
+   if (inst->Instruction.Opcode == TGSI_OPCODE_END &&
+       aactx->colorOutput != -1) {
+      struct tgsi_full_instruction newInst;
+
+      /* TEX */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_TEX;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = aactx->texTemp;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->maxInput + 1;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->freeSampler;
+
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* MOV rgb */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MOV;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* MUL alpha */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->texTemp;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* END */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_END;
+      newInst.Instruction.NumDstRegs = 0;
+      newInst.Instruction.NumSrcRegs = 0;
+      ctx->emit_instruction(ctx, &newInst);
+   }
+   else {
+      /* Not an END instruction.
+       * Look for writes to result.color and replace with colorTemp reg.
+       */
+      uint i;
+
+      for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
+         struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
+         if (dst->DstRegister.File == TGSI_FILE_OUTPUT &&
+             dst->DstRegister.Index == aactx->colorOutput) {
+            dst->DstRegister.File = TGSI_FILE_TEMPORARY;
+            dst->DstRegister.Index = aactx->colorTemp;
+         }
+      }
+
+      ctx->emit_instruction(ctx, inst);
+   }
+}
+
+
+/**
+ * Generate the frag shader we'll use for drawing AA lines.
+ * This will be the user's shader plus some texture/modulate instructions.
+ */
+static boolean
+generate_aaline_fs(struct aaline_stage *aaline)
+{
+   const struct pipe_shader_state *orig_fs = &aaline->fs->state;
+   struct pipe_shader_state aaline_fs;
+   struct aa_transform_context transform;
+
+#define MAX 1000
+
+   aaline_fs = *orig_fs; /* copy to init */
+   aaline_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX);
+   if (aaline_fs.tokens == NULL)
+      return FALSE;
+
+   memset(&transform, 0, sizeof(transform));
+   transform.colorOutput = -1;
+   transform.maxInput = -1;
+   transform.maxGeneric = -1;
+   transform.colorTemp = -1;
+   transform.texTemp = -1;
+   transform.firstInstruction = TRUE;
+   transform.base.transform_instruction = aa_transform_inst;
+   transform.base.transform_declaration = aa_transform_decl;
+
+   tgsi_transform_shader(orig_fs->tokens,
+                         (struct tgsi_token *) aaline_fs.tokens,
+                         MAX, &transform.base);
+
+#if 0 /* DEBUG */
+   tgsi_dump(orig_fs->tokens, 0);
+   tgsi_dump(aaline_fs.tokens, 0);
+#endif
+
+   aaline->fs->sampler_unit = transform.freeSampler;
+
+   aaline->fs->aaline_fs
+      = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs);
+   if (aaline->fs->aaline_fs == NULL)
+      return FALSE;
+
+   aaline->fs->generic_attrib = transform.maxGeneric + 1;
+   return TRUE;
+}
+
+
+/**
+ * Create the texture map we'll use for antialiasing the lines.
+ */
+static boolean
+aaline_create_texture(struct aaline_stage *aaline)
+{
+   struct pipe_context *pipe = aaline->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture texTemp;
+   uint level;
+
+   memset(&texTemp, 0, sizeof(texTemp));
+   texTemp.target = PIPE_TEXTURE_2D;
+   texTemp.format = PIPE_FORMAT_A8_UNORM; /* XXX verify supported by driver! */
+   texTemp.last_level = MAX_TEXTURE_LEVEL;
+   texTemp.width[0] = 1 << MAX_TEXTURE_LEVEL;
+   texTemp.height[0] = 1 << MAX_TEXTURE_LEVEL;
+   texTemp.depth[0] = 1;
+   pf_get_block(texTemp.format, &texTemp.block);
+
+   aaline->texture = screen->texture_create(screen, &texTemp);
+   if (!aaline->texture)
+      return FALSE;
+
+   /* Fill in mipmap images.
+    * Basically each level is solid opaque, except for the outermost
+    * texels which are zero.  Special case the 1x1 and 2x2 levels.
+    */
+   for (level = 0; level <= MAX_TEXTURE_LEVEL; level++) {
+      struct pipe_surface *surface;
+      const uint size = aaline->texture->width[level];
+      ubyte *data;
+      uint i, j;
+
+      assert(aaline->texture->width[level] == aaline->texture->height[level]);
+
+      /* This texture is new, no need to flush. 
+       */
+      surface = screen->get_tex_surface(screen, aaline->texture, 0, level, 0,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+      data = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_WRITE);
+      if (data == NULL)
+         return FALSE;
+
+      for (i = 0; i < size; i++) {
+         for (j = 0; j < size; j++) {
+            ubyte d;
+            if (size == 1) {
+               d = 255;
+            }
+            else if (size == 2) {
+               d = 200; /* tuneable */
+            }
+            else if (i == 0 || j == 0 || i == size - 1 || j == size - 1) {
+               d = 0;
+            }
+            else {
+               d = 255;
+            }
+            data[i * surface->stride + j] = d;
+         }
+      }
+
+      /* unmap */
+      screen->surface_unmap(screen, surface);
+      screen->tex_surface_release(screen, &surface);
+   }
+   return TRUE;
+}
+
+
+/**
+ * Create the sampler CSO that'll be used for antialiasing.
+ * By using a mipmapped texture, we don't have to generate a different
+ * texture image for each line size.
+ */
+static boolean
+aaline_create_sampler(struct aaline_stage *aaline)
+{
+   struct pipe_sampler_state sampler;
+   struct pipe_context *pipe = aaline->pipe;
+
+   memset(&sampler, 0, sizeof(sampler));
+   sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   sampler.min_mip_filter = PIPE_TEX_MIPFILTER_LINEAR;
+   sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR;
+   sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR;
+   sampler.normalized_coords = 1;
+   sampler.min_lod = 0.0f;
+   sampler.max_lod = MAX_TEXTURE_LEVEL;
+
+   aaline->sampler_cso = pipe->create_sampler_state(pipe, &sampler);
+   if (aaline->sampler_cso == NULL)
+      return FALSE;
+
+   return TRUE;
+}
+
+
+/**
+ * When we're about to draw our first AA line in a batch, this function is
+ * called to tell the driver to bind our modified fragment shader.
+ */
+static boolean
+bind_aaline_fragment_shader(struct aaline_stage *aaline)
+{
+   struct draw_context *draw = aaline->stage.draw;
+
+   if (!aaline->fs->aaline_fs && 
+       !generate_aaline_fs(aaline))
+      return FALSE;
+
+   draw->suspend_flushing = TRUE;
+   aaline->driver_bind_fs_state(aaline->pipe, aaline->fs->aaline_fs);
+   draw->suspend_flushing = FALSE;
+
+   return TRUE;
+}
+
+
+
+static INLINE struct aaline_stage *
+aaline_stage( struct draw_stage *stage )
+{
+   return (struct aaline_stage *) stage;
+}
+
+
+/**
+ * Draw a wide line by drawing a quad, using geometry which will
+ * fullfill GL's antialiased line requirements.
+ */
+static void
+aaline_line(struct draw_stage *stage, struct prim_header *header)
+{
+   const struct aaline_stage *aaline = aaline_stage(stage);
+   const float half_width = aaline->half_line_width;
+   struct prim_header tri;
+   struct vertex_header *v[8];
+   uint texPos = aaline->tex_slot;
+   uint posPos = aaline->pos_slot;
+   float *pos, *tex;
+   float dx = header->v[1]->data[posPos][0] - header->v[0]->data[posPos][0];
+   float dy = header->v[1]->data[posPos][1] - header->v[0]->data[posPos][1];
+   double a = atan2(dy, dx);
+   float c_a = (float) cos(a), s_a = (float) sin(a);
+   uint i;
+
+   /* XXX the ends of lines aren't quite perfect yet, but probably passable */
+   dx = 0.5F * half_width;
+   dy = half_width;
+
+   /* allocate/dup new verts */
+   for (i = 0; i < 8; i++) {
+      v[i] = dup_vert(stage, header->v[i/4], i);
+   }
+
+   /*
+    * Quad strip for line from v0 to v1 (*=endpoints):
+    *
+    *  1   3                     5   7
+    *  +---+---------------------+---+
+    *  |                             |
+    *  | *v0                     v1* |
+    *  |                             |
+    *  +---+---------------------+---+
+    *  0   2                     4   6
+    */
+
+   /* new verts */
+   pos = v[0]->data[posPos];
+   pos[0] += (-dx * c_a -  dy * s_a);
+   pos[1] += (-dx * s_a +  dy * c_a);
+
+   pos = v[1]->data[posPos];
+   pos[0] += (-dx * c_a - -dy * s_a);
+   pos[1] += (-dx * s_a + -dy * c_a);
+
+   pos = v[2]->data[posPos];
+   pos[0] += ( dx * c_a -  dy * s_a);
+   pos[1] += ( dx * s_a +  dy * c_a);
+
+   pos = v[3]->data[posPos];
+   pos[0] += ( dx * c_a - -dy * s_a);
+   pos[1] += ( dx * s_a + -dy * c_a);
+
+   pos = v[4]->data[posPos];
+   pos[0] += (-dx * c_a -  dy * s_a);
+   pos[1] += (-dx * s_a +  dy * c_a);
+
+   pos = v[5]->data[posPos];
+   pos[0] += (-dx * c_a - -dy * s_a);
+   pos[1] += (-dx * s_a + -dy * c_a);
+
+   pos = v[6]->data[posPos];
+   pos[0] += ( dx * c_a -  dy * s_a);
+   pos[1] += ( dx * s_a +  dy * c_a);
+
+   pos = v[7]->data[posPos];
+   pos[0] += ( dx * c_a - -dy * s_a);
+   pos[1] += ( dx * s_a + -dy * c_a);
+
+   /* new texcoords */
+   tex = v[0]->data[texPos];
+   ASSIGN_4V(tex, 0, 0, 0, 1);
+
+   tex = v[1]->data[texPos];
+   ASSIGN_4V(tex, 0, 1, 0, 1);
+
+   tex = v[2]->data[texPos];
+   ASSIGN_4V(tex, .5, 0, 0, 1);
+
+   tex = v[3]->data[texPos];
+   ASSIGN_4V(tex, .5, 1, 0, 1);
+
+   tex = v[4]->data[texPos];
+   ASSIGN_4V(tex, .5, 0, 0, 1);
+
+   tex = v[5]->data[texPos];
+   ASSIGN_4V(tex, .5, 1, 0, 1);
+
+   tex = v[6]->data[texPos];
+   ASSIGN_4V(tex, 1, 0, 0, 1);
+
+   tex = v[7]->data[texPos];
+   ASSIGN_4V(tex, 1, 1, 0, 1);
+
+   /* emit 6 tris for the quad strip */
+   tri.v[0] = v[2];  tri.v[1] = v[1];  tri.v[2] = v[0];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[3];  tri.v[1] = v[1];  tri.v[2] = v[2];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[4];  tri.v[1] = v[3];  tri.v[2] = v[2];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[5];  tri.v[1] = v[3];  tri.v[2] = v[4];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[6];  tri.v[1] = v[5];  tri.v[2] = v[4];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[7];  tri.v[1] = v[5];  tri.v[2] = v[6];
+   stage->next->tri( stage->next, &tri );
+}
+
+
+static void
+aaline_first_line(struct draw_stage *stage, struct prim_header *header)
+{
+   auto struct aaline_stage *aaline = aaline_stage(stage);
+   struct draw_context *draw = stage->draw;
+   struct pipe_context *pipe = aaline->pipe;
+   uint num_samplers;
+
+   assert(draw->rasterizer->line_smooth);
+
+   if (draw->rasterizer->line_width <= 3.0)
+      aaline->half_line_width = 1.5f;
+   else
+      aaline->half_line_width = 0.5f * draw->rasterizer->line_width;
+
+   /*
+    * Bind (generate) our fragprog, sampler and texture
+    */
+   if (!bind_aaline_fragment_shader(aaline)) {
+      stage->line = draw_pipe_passthrough_line;
+      stage->line(stage, header);
+      return;
+   }
+
+   /* update vertex attrib info */
+   aaline->tex_slot = draw->vs.num_vs_outputs;
+   aaline->pos_slot = draw->vs.position_output;
+
+   /* advertise the extra post-transformed vertex attribute */
+   draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
+   draw->extra_vp_outputs.semantic_index = aaline->fs->generic_attrib;
+   draw->extra_vp_outputs.slot = aaline->tex_slot;
+
+   /* how many samplers? */
+   /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */
+   num_samplers = MAX2(aaline->num_textures, aaline->num_samplers);
+   num_samplers = MAX2(num_samplers, aaline->fs->sampler_unit + 1);
+
+   aaline->state.sampler[aaline->fs->sampler_unit] = aaline->sampler_cso;
+   pipe_texture_reference(&aaline->state.texture[aaline->fs->sampler_unit],
+                          aaline->texture);
+
+   draw->suspend_flushing = TRUE;
+   aaline->driver_bind_sampler_states(pipe, num_samplers, aaline->state.sampler);
+   aaline->driver_set_sampler_textures(pipe, num_samplers, aaline->state.texture);
+   draw->suspend_flushing = FALSE;
+
+   /* now really draw first line */
+   stage->line = aaline_line;
+   stage->line(stage, header);
+}
+
+
+static void
+aaline_flush(struct draw_stage *stage, unsigned flags)
+{
+   struct draw_context *draw = stage->draw;
+   struct aaline_stage *aaline = aaline_stage(stage);
+   struct pipe_context *pipe = aaline->pipe;
+
+   stage->line = aaline_first_line;
+   stage->next->flush( stage->next, flags );
+
+   /* restore original frag shader, texture, sampler state */
+   draw->suspend_flushing = TRUE;
+   aaline->driver_bind_fs_state(pipe, aaline->fs->driver_fs);
+   aaline->driver_bind_sampler_states(pipe, aaline->num_samplers,
+                                      aaline->state.sampler);
+   aaline->driver_set_sampler_textures(pipe, aaline->num_textures,
+                                       aaline->state.texture);
+   draw->suspend_flushing = FALSE;
+
+   draw->extra_vp_outputs.slot = 0;
+}
+
+
+static void
+aaline_reset_stipple_counter(struct draw_stage *stage)
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void
+aaline_destroy(struct draw_stage *stage)
+{
+   struct aaline_stage *aaline = aaline_stage(stage);
+   uint i;
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      pipe_texture_reference(&aaline->state.texture[i], NULL);
+   }
+
+   if (aaline->sampler_cso)
+      aaline->pipe->delete_sampler_state(aaline->pipe, aaline->sampler_cso);
+
+   if (aaline->texture)
+      pipe_texture_release(&aaline->texture);
+
+   draw_free_temp_verts( stage );
+
+   FREE( stage );
+}
+
+
+static struct aaline_stage *
+draw_aaline_stage(struct draw_context *draw)
+{
+   struct aaline_stage *aaline = CALLOC_STRUCT(aaline_stage);
+   if (aaline == NULL)
+      return NULL;
+
+   if (!draw_alloc_temp_verts( &aaline->stage, 8 ))
+      goto fail;
+
+   aaline->stage.draw = draw;
+   aaline->stage.next = NULL;
+   aaline->stage.point = draw_pipe_passthrough_point;
+   aaline->stage.line = aaline_first_line;
+   aaline->stage.tri = draw_pipe_passthrough_tri;
+   aaline->stage.flush = aaline_flush;
+   aaline->stage.reset_stipple_counter = aaline_reset_stipple_counter;
+   aaline->stage.destroy = aaline_destroy;
+
+   return aaline;
+
+ fail:
+   if (aaline)
+      aaline_destroy(&aaline->stage);
+
+   return NULL;
+}
+
+
+static struct aaline_stage *
+aaline_stage_from_pipe(struct pipe_context *pipe)
+{
+   struct draw_context *draw = (struct draw_context *) pipe->draw;
+   return aaline_stage(draw->pipeline.aaline);
+}
+
+
+/**
+ * This function overrides the driver's create_fs_state() function and
+ * will typically be called by the state tracker.
+ */
+static void *
+aaline_create_fs_state(struct pipe_context *pipe,
+                       const struct pipe_shader_state *fs)
+{
+   struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
+   struct aaline_fragment_shader *aafs = CALLOC_STRUCT(aaline_fragment_shader);
+   if (aafs == NULL)
+      return NULL;
+
+   aafs->state = *fs;
+
+   /* pass-through */
+   aafs->driver_fs = aaline->driver_create_fs_state(aaline->pipe, fs);
+
+   return aafs;
+}
+
+
+static void
+aaline_bind_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
+   struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs;
+
+   /* save current */
+   aaline->fs = aafs;
+   /* pass-through */
+   aaline->driver_bind_fs_state(aaline->pipe,
+                                (aafs ? aafs->driver_fs : NULL));
+}
+
+
+static void
+aaline_delete_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
+   struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs;
+   /* pass-through */
+   aaline->driver_delete_fs_state(aaline->pipe, aafs->driver_fs);
+   FREE(aafs);
+}
+
+
+static void
+aaline_bind_sampler_states(struct pipe_context *pipe,
+                           unsigned num, void **sampler)
+{
+   struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
+
+   /* save current */
+   memcpy(aaline->state.sampler, sampler, num * sizeof(void *));
+   aaline->num_samplers = num;
+
+   /* pass-through */
+   aaline->driver_bind_sampler_states(aaline->pipe, num, sampler);
+}
+
+
+static void
+aaline_set_sampler_textures(struct pipe_context *pipe,
+                            unsigned num, struct pipe_texture **texture)
+{
+   struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
+   uint i;
+
+   /* save current */
+   for (i = 0; i < num; i++) {
+      pipe_texture_reference(&aaline->state.texture[i], texture[i]);
+   }
+   for ( ; i < PIPE_MAX_SAMPLERS; i++) {
+      pipe_texture_reference(&aaline->state.texture[i], NULL);
+   }
+   aaline->num_textures = num;
+
+   /* pass-through */
+   aaline->driver_set_sampler_textures(aaline->pipe, num, texture);
+}
+
+
+/**
+ * Called by drivers that want to install this AA line prim stage
+ * into the draw module's pipeline.  This will not be used if the
+ * hardware has native support for AA lines.
+ */
+boolean
+draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe)
+{
+   struct aaline_stage *aaline;
+
+   pipe->draw = (void *) draw;
+
+   /*
+    * Create / install AA line drawing / prim stage
+    */
+   aaline = draw_aaline_stage( draw );
+   if (!aaline)
+      goto fail;
+
+   aaline->pipe = pipe;
+
+   /* create special texture, sampler state */
+   if (!aaline_create_texture(aaline))
+      goto fail;
+
+   if (!aaline_create_sampler(aaline))
+      goto fail;
+
+   /* save original driver functions */
+   aaline->driver_create_fs_state = pipe->create_fs_state;
+   aaline->driver_bind_fs_state = pipe->bind_fs_state;
+   aaline->driver_delete_fs_state = pipe->delete_fs_state;
+
+   aaline->driver_bind_sampler_states = pipe->bind_sampler_states;
+   aaline->driver_set_sampler_textures = pipe->set_sampler_textures;
+
+   /* override the driver's functions */
+   pipe->create_fs_state = aaline_create_fs_state;
+   pipe->bind_fs_state = aaline_bind_fs_state;
+   pipe->delete_fs_state = aaline_delete_fs_state;
+
+   pipe->bind_sampler_states = aaline_bind_sampler_states;
+   pipe->set_sampler_textures = aaline_set_sampler_textures;
+   
+   /* Install once everything is known to be OK:
+    */
+   draw->pipeline.aaline = &aaline->stage;
+
+   return TRUE;
+
+ fail:
+   if (aaline)
+      aaline->stage.destroy( &aaline->stage );
+   
+   return FALSE;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
new file mode 100644
index 0000000000..2c1cacbdb4
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
@@ -0,0 +1,875 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * AA point stage:  AA points are converted to quads and rendered with a
+ * special fragment shader.  Another approach would be to use a texture
+ * map image of a point, but experiments indicate the quality isn't nearly
+ * as good as this approach.
+ *
+ * Note: this looks a lot like draw_aaline.c but there's actually little
+ * if any code that can be shared.
+ *
+ * Authors:  Brian Paul
+ */
+
+
+#include "pipe/p_inlines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/tgsi_transform.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "draw_context.h"
+#include "draw_vs.h"
+#include "draw_pipe.h"
+
+
+/*
+ * Enabling NORMALIZE might give _slightly_ better results.
+ * Basically, it controls whether we compute distance as d=sqrt(x*x+y*y) or
+ * d=x*x+y*y.  Since we're working with a unit circle, the later seems
+ * close enough and saves some costly instructions.
+ */
+#define NORMALIZE 0
+
+
+/**
+ * Subclass of pipe_shader_state to carry extra fragment shader info.
+ */
+struct aapoint_fragment_shader
+{
+   struct pipe_shader_state state;
+   void *driver_fs;   /**< the regular shader */
+   void *aapoint_fs;  /**< the aa point-augmented shader */
+   int generic_attrib; /**< The generic input attrib/texcoord we'll use */
+};
+
+
+/**
+ * Subclass of draw_stage
+ */
+struct aapoint_stage
+{
+   struct draw_stage stage;
+
+   int psize_slot;
+   float radius;
+
+   /** this is the vertex attrib slot for the new texcoords */
+   uint tex_slot;
+   uint pos_slot;
+
+   /*
+    * Currently bound state
+    */
+   struct aapoint_fragment_shader *fs;
+
+   /*
+    * Driver interface/override functions
+    */
+   void * (*driver_create_fs_state)(struct pipe_context *,
+                                    const struct pipe_shader_state *);
+   void (*driver_bind_fs_state)(struct pipe_context *, void *);
+   void (*driver_delete_fs_state)(struct pipe_context *, void *);
+
+   struct pipe_context *pipe;
+};
+
+
+
+/**
+ * Subclass of tgsi_transform_context, used for transforming the
+ * user's fragment shader to add the special AA instructions.
+ */
+struct aa_transform_context {
+   struct tgsi_transform_context base;
+   uint tempsUsed;  /**< bitmask */
+   int colorOutput; /**< which output is the primary color */
+   int maxInput, maxGeneric;  /**< max input index found */
+   int tmp0, colorTemp;  /**< temp registers */
+   boolean firstInstruction;
+};
+
+
+/**
+ * TGSI declaration transform callback.
+ * Look for two free temp regs and available input reg for new texcoords.
+ */
+static void
+aa_transform_decl(struct tgsi_transform_context *ctx,
+                  struct tgsi_full_declaration *decl)
+{
+   struct aa_transform_context *aactx = (struct aa_transform_context *) ctx;
+
+   if (decl->Declaration.File == TGSI_FILE_OUTPUT &&
+       decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR &&
+       decl->Semantic.SemanticIndex == 0) {
+      aactx->colorOutput = decl->DeclarationRange.First;
+   }
+   else if (decl->Declaration.File == TGSI_FILE_INPUT) {
+      if ((int) decl->DeclarationRange.Last > aactx->maxInput)
+         aactx->maxInput = decl->DeclarationRange.Last;
+      if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC &&
+           (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) {
+         aactx->maxGeneric = decl->Semantic.SemanticIndex;
+      }
+   }
+   else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) {
+      uint i;
+      for (i = decl->DeclarationRange.First;
+           i <= decl->DeclarationRange.Last; i++) {
+         aactx->tempsUsed |= (1 << i);
+      }
+   }
+
+   ctx->emit_declaration(ctx, decl);
+}
+
+
+/**
+ * TGSI instruction transform callback.
+ * Replace writes to result.color w/ a temp reg.
+ * Upon END instruction, insert texture sampling code for antialiasing.
+ */
+static void
+aa_transform_inst(struct tgsi_transform_context *ctx,
+                  struct tgsi_full_instruction *inst)
+{
+   struct aa_transform_context *aactx = (struct aa_transform_context *) ctx;
+   struct tgsi_full_instruction newInst;
+
+   if (aactx->firstInstruction) {
+      /* emit our new declarations before the first instruction */
+
+      struct tgsi_full_declaration decl;
+      const int texInput = aactx->maxInput + 1;
+      int tmp0;
+      uint i;
+
+      /* find two free temp regs */
+      for (i = 0; i < 32; i++) {
+         if ((aactx->tempsUsed & (1 << i)) == 0) {
+            /* found a free temp */
+            if (aactx->tmp0 < 0)
+               aactx->tmp0 = i;
+            else if (aactx->colorTemp < 0)
+               aactx->colorTemp = i;
+            else
+               break;
+         }
+      }
+
+      assert(aactx->colorTemp != aactx->tmp0);
+
+      tmp0 = aactx->tmp0;
+
+      /* declare new generic input/texcoord */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_INPUT;
+      /* XXX this could be linear... */
+      decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE;
+      decl.Declaration.Semantic = 1;
+      decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
+      decl.Semantic.SemanticIndex = aactx->maxGeneric + 1;
+      decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = texInput;
+      ctx->emit_declaration(ctx, &decl);
+
+      /* declare new temp regs */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_TEMPORARY;
+      decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = tmp0;
+      ctx->emit_declaration(ctx, &decl);
+
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_TEMPORARY;
+      decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = aactx->colorTemp;
+      ctx->emit_declaration(ctx, &decl);
+
+      aactx->firstInstruction = FALSE;
+
+
+      /*
+       * Emit code to compute fragment coverage, kill if outside point radius
+       *
+       * Temp reg0 usage:
+       *  t0.x = distance of fragment from center point
+       *  t0.y = boolean, is t0.x > 1.0, also misc temp usage
+       *  t0.z = temporary for computing 1/(1-k) value
+       *  t0.w = final coverage value
+       */
+
+      /* MUL t0.xy, tex, tex;  # compute x^2, y^2 */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XY;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = texInput;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* ADD t0.x, t0.x, t0.y;  # x^2 + y^2 */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_ADD;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+      ctx->emit_instruction(ctx, &newInst);
+
+#if NORMALIZE  /* OPTIONAL normalization of length */
+      /* RSQ t0.x, t0.x; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_RSQ;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* RCP t0.x, t0.x; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_RCP;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      ctx->emit_instruction(ctx, &newInst);
+#endif
+
+      /* SGT t0.y, t0.xxxx, t0.wwww;  # bool b = d > 1 (NOTE t0.w == 1) */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_SGT;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* KIL -tmp0.yyyy;   # if -tmp0.y < 0, KILL */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_KIL;
+      newInst.Instruction.NumDstRegs = 0;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.Negate = 1;
+      ctx->emit_instruction(ctx, &newInst);
+
+
+      /* compute coverage factor = (1-d)/(1-k) */
+
+      /* SUB t0.z, tex.w, tex.z;  # m = 1 - k */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_SUB;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = texInput;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* RCP t0.z, t0.z;  # t0.z = 1 / m */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_RCP;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Z;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* SUB t0.y, 1, t0.x;  # d = 1 - d */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_SUB;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = texInput;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* MUL t0.w, t0.y, t0.z;   # coverage = d * m */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Z;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* SLE t0.y, t0.x, tex.z;  # bool b = distance <= k */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_SLE;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Z;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* CMP t0.w, -t0.y, tex.w, t0.w;
+       *  # if -t0.y < 0 then
+       *       t0.w = 1
+       *    else
+       *       t0.w = t0.w
+       */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_CMP;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+      newInst.Instruction.NumSrcRegs = 3;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.Negate = 1;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_W;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_W;
+      newInst.FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[2].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[2].SrcRegister.SwizzleX = TGSI_SWIZZLE_W;
+      newInst.FullSrcRegisters[2].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
+      newInst.FullSrcRegisters[2].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
+      newInst.FullSrcRegisters[2].SrcRegister.SwizzleW = TGSI_SWIZZLE_W;
+      ctx->emit_instruction(ctx, &newInst);
+
+   }
+
+   if (inst->Instruction.Opcode == TGSI_OPCODE_END) {
+      /* add alpha modulation code at tail of program */
+
+      /* MOV result.color.xyz, colorTemp; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MOV;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* MUL result.color.w, colorTemp, tmp0.w; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->tmp0;
+      ctx->emit_instruction(ctx, &newInst);
+   }
+   else {
+      /* Not an END instruction.
+       * Look for writes to result.color and replace with colorTemp reg.
+       */
+      uint i;
+
+      for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
+         struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
+         if (dst->DstRegister.File == TGSI_FILE_OUTPUT &&
+             dst->DstRegister.Index == aactx->colorOutput) {
+            dst->DstRegister.File = TGSI_FILE_TEMPORARY;
+            dst->DstRegister.Index = aactx->colorTemp;
+         }
+      }
+   }
+
+   ctx->emit_instruction(ctx, inst);
+}
+
+
+/**
+ * Generate the frag shader we'll use for drawing AA points.
+ * This will be the user's shader plus some texture/modulate instructions.
+ */
+static boolean
+generate_aapoint_fs(struct aapoint_stage *aapoint)
+{
+   const struct pipe_shader_state *orig_fs = &aapoint->fs->state;
+   struct pipe_shader_state aapoint_fs;
+   struct aa_transform_context transform;
+
+#define MAX 1000
+
+   aapoint_fs = *orig_fs; /* copy to init */
+   aapoint_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX);
+   if (aapoint_fs.tokens == NULL)
+      return FALSE;
+
+   memset(&transform, 0, sizeof(transform));
+   transform.colorOutput = -1;
+   transform.maxInput = -1;
+   transform.maxGeneric = -1;
+   transform.colorTemp = -1;
+   transform.tmp0 = -1;
+   transform.firstInstruction = TRUE;
+   transform.base.transform_instruction = aa_transform_inst;
+   transform.base.transform_declaration = aa_transform_decl;
+
+   tgsi_transform_shader(orig_fs->tokens,
+                         (struct tgsi_token *) aapoint_fs.tokens,
+                         MAX, &transform.base);
+
+#if 0 /* DEBUG */
+   printf("draw_aapoint, orig shader:\n");
+   tgsi_dump(orig_fs->tokens, 0);
+   printf("draw_aapoint, new shader:\n");
+   tgsi_dump(aapoint_fs.tokens, 0);
+#endif
+
+   aapoint->fs->aapoint_fs
+      = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs);
+   if (aapoint->fs->aapoint_fs == NULL)
+      return FALSE;
+
+   aapoint->fs->generic_attrib = transform.maxGeneric + 1;
+
+   return TRUE;
+}
+
+
+/**
+ * When we're about to draw our first AA point in a batch, this function is
+ * called to tell the driver to bind our modified fragment shader.
+ */
+static boolean
+bind_aapoint_fragment_shader(struct aapoint_stage *aapoint)
+{
+   struct draw_context *draw = aapoint->stage.draw;
+
+   if (!aapoint->fs->aapoint_fs &&
+       !generate_aapoint_fs(aapoint))
+      return FALSE;
+
+   draw->suspend_flushing = TRUE;
+   aapoint->driver_bind_fs_state(aapoint->pipe, aapoint->fs->aapoint_fs);
+   draw->suspend_flushing = FALSE;
+
+   return TRUE;
+}
+
+
+
+static INLINE struct aapoint_stage *
+aapoint_stage( struct draw_stage *stage )
+{
+   return (struct aapoint_stage *) stage;
+}
+
+
+
+
+/**
+ * Draw an AA point by drawing a quad.
+ */
+static void
+aapoint_point(struct draw_stage *stage, struct prim_header *header)
+{
+   const struct aapoint_stage *aapoint = aapoint_stage(stage);
+   struct prim_header tri;
+   struct vertex_header *v[4];
+   uint texPos = aapoint->tex_slot;
+   uint pos_slot = aapoint->pos_slot;
+   float radius, *pos, *tex;
+   uint i;
+   float k;
+
+   if (aapoint->psize_slot >= 0) {
+      radius = 0.5f * header->v[0]->data[aapoint->psize_slot][0];
+   }
+   else {
+      radius = aapoint->radius;
+   }
+
+   /*
+    * Note: the texcoords (generic attrib, really) we use are special:
+    * The S and T components simply vary from -1 to +1.
+    * The R component is k, below.
+    * The Q component is 1.0 and will used as a handy constant in the
+    * fragment shader.
+    */
+
+   /*
+    * k is the threshold distance from the point's center at which
+    * we begin alpha attenuation (the coverage value).
+    * Operating within a unit circle, we'll compute the fragment's
+    * distance 'd' from the center point using the texcoords.
+    * IF d > 1.0 THEN
+    *    KILL fragment
+    * ELSE IF d > k THEN
+    *    compute coverage in [0,1] proportional to d in [k, 1].
+    * ELSE
+    *    coverage = 1.0;  // full coverage
+    * ENDIF
+    *
+    * Note: the ELSEIF and ELSE clauses are actually implemented with CMP to
+    * avoid using IF/ELSE/ENDIF TGSI opcodes.
+    */
+
+#if !NORMALIZE
+   k = 1.0f / radius;
+   k = 1.0f - 2.0f * k + k * k;
+#else
+   k = 1.0f - 1.0f / radius;
+#endif
+
+   /* allocate/dup new verts */
+   for (i = 0; i < 4; i++) {
+      v[i] = dup_vert(stage, header->v[0], i);
+   }
+
+   /* new verts */
+   pos = v[0]->data[pos_slot];
+   pos[0] -= radius;
+   pos[1] -= radius;
+
+   pos = v[1]->data[pos_slot];
+   pos[0] += radius;
+   pos[1] -= radius;
+
+   pos = v[2]->data[pos_slot];
+   pos[0] += radius;
+   pos[1] += radius;
+
+   pos = v[3]->data[pos_slot];
+   pos[0] -= radius;
+   pos[1] += radius;
+
+   /* new texcoords */
+   tex = v[0]->data[texPos];
+   ASSIGN_4V(tex, -1, -1, k, 1);
+
+   tex = v[1]->data[texPos];
+   ASSIGN_4V(tex,  1, -1, k, 1);
+
+   tex = v[2]->data[texPos];
+   ASSIGN_4V(tex,  1,  1, k, 1);
+
+   tex = v[3]->data[texPos];
+   ASSIGN_4V(tex, -1,  1, k, 1);
+
+   /* emit 2 tris for the quad strip */
+   tri.v[0] = v[0];
+   tri.v[1] = v[1];
+   tri.v[2] = v[2];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[0];
+   tri.v[1] = v[2];
+   tri.v[2] = v[3];
+   stage->next->tri( stage->next, &tri );
+}
+
+
+static void
+aapoint_first_point(struct draw_stage *stage, struct prim_header *header)
+{
+   auto struct aapoint_stage *aapoint = aapoint_stage(stage);
+   struct draw_context *draw = stage->draw;
+
+   assert(draw->rasterizer->point_smooth);
+
+   if (draw->rasterizer->point_size <= 2.0)
+      aapoint->radius = 1.0;
+   else
+      aapoint->radius = 0.5f * draw->rasterizer->point_size;
+
+   /*
+    * Bind (generate) our fragprog.
+    */
+   bind_aapoint_fragment_shader(aapoint);
+
+   /* update vertex attrib info */
+   aapoint->tex_slot = draw->vs.num_vs_outputs;
+   assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */
+
+   aapoint->pos_slot = draw->vs.position_output;
+
+   draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
+   draw->extra_vp_outputs.semantic_index = aapoint->fs->generic_attrib;
+   draw->extra_vp_outputs.slot = aapoint->tex_slot;
+
+   /* find psize slot in post-transform vertex */
+   aapoint->psize_slot = -1;
+   if (draw->rasterizer->point_size_per_vertex) {
+      /* find PSIZ vertex output */
+      const struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+      uint i;
+      for (i = 0; i < vs->info.num_outputs; i++) {
+         if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) {
+            aapoint->psize_slot = i;
+            break;
+         }
+      }
+   }
+
+   /* now really draw first point */
+   stage->point = aapoint_point;
+   stage->point(stage, header);
+}
+
+
+static void
+aapoint_flush(struct draw_stage *stage, unsigned flags)
+{
+   struct draw_context *draw = stage->draw;
+   struct aapoint_stage *aapoint = aapoint_stage(stage);
+   struct pipe_context *pipe = aapoint->pipe;
+
+   stage->point = aapoint_first_point;
+   stage->next->flush( stage->next, flags );
+
+   /* restore original frag shader */
+   draw->suspend_flushing = TRUE;
+   aapoint->driver_bind_fs_state(pipe, aapoint->fs->driver_fs);
+   draw->suspend_flushing = FALSE;
+
+   draw->extra_vp_outputs.slot = 0;
+}
+
+
+static void
+aapoint_reset_stipple_counter(struct draw_stage *stage)
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void
+aapoint_destroy(struct draw_stage *stage)
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+static struct aapoint_stage *
+draw_aapoint_stage(struct draw_context *draw)
+{
+   struct aapoint_stage *aapoint = CALLOC_STRUCT(aapoint_stage);
+   if (aapoint == NULL)
+      goto fail;
+
+   if (!draw_alloc_temp_verts( &aapoint->stage, 4 ))
+      goto fail;
+
+   aapoint->stage.draw = draw;
+   aapoint->stage.next = NULL;
+   aapoint->stage.point = aapoint_first_point;
+   aapoint->stage.line = draw_pipe_passthrough_line;
+   aapoint->stage.tri = draw_pipe_passthrough_tri;
+   aapoint->stage.flush = aapoint_flush;
+   aapoint->stage.reset_stipple_counter = aapoint_reset_stipple_counter;
+   aapoint->stage.destroy = aapoint_destroy;
+
+   return aapoint;
+
+ fail:
+   if (aapoint)
+      aapoint_destroy(&aapoint->stage);
+
+   return NULL;
+
+}
+
+
+static struct aapoint_stage *
+aapoint_stage_from_pipe(struct pipe_context *pipe)
+{
+   struct draw_context *draw = (struct draw_context *) pipe->draw;
+   return aapoint_stage(draw->pipeline.aapoint);
+}
+
+
+/**
+ * This function overrides the driver's create_fs_state() function and
+ * will typically be called by the state tracker.
+ */
+static void *
+aapoint_create_fs_state(struct pipe_context *pipe,
+                       const struct pipe_shader_state *fs)
+{
+   struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe);
+   struct aapoint_fragment_shader *aafs = CALLOC_STRUCT(aapoint_fragment_shader);
+   if (aafs == NULL) 
+      return NULL;
+
+   aafs->state = *fs;
+
+   /* pass-through */
+   aafs->driver_fs = aapoint->driver_create_fs_state(aapoint->pipe, fs);
+
+   return aafs;
+}
+
+
+static void
+aapoint_bind_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe);
+   struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs;
+   /* save current */
+   aapoint->fs = aafs;
+   /* pass-through */
+   aapoint->driver_bind_fs_state(aapoint->pipe,
+                                 (aafs ? aafs->driver_fs : NULL));
+}
+
+
+static void
+aapoint_delete_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe);
+   struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs;
+   /* pass-through */
+   aapoint->driver_delete_fs_state(aapoint->pipe, aafs->driver_fs);
+   FREE(aafs);
+}
+
+
+/**
+ * Called by drivers that want to install this AA point prim stage
+ * into the draw module's pipeline.  This will not be used if the
+ * hardware has native support for AA points.
+ */
+boolean
+draw_install_aapoint_stage(struct draw_context *draw,
+                           struct pipe_context *pipe)
+{
+   struct aapoint_stage *aapoint;
+
+   pipe->draw = (void *) draw;
+
+   /*
+    * Create / install AA point drawing / prim stage
+    */
+   aapoint = draw_aapoint_stage( draw );
+   if (aapoint == NULL)
+      goto fail;
+
+   aapoint->pipe = pipe;
+
+   /* save original driver functions */
+   aapoint->driver_create_fs_state = pipe->create_fs_state;
+   aapoint->driver_bind_fs_state = pipe->bind_fs_state;
+   aapoint->driver_delete_fs_state = pipe->delete_fs_state;
+
+   /* override the driver's functions */
+   pipe->create_fs_state = aapoint_create_fs_state;
+   pipe->bind_fs_state = aapoint_bind_fs_state;
+   pipe->delete_fs_state = aapoint_delete_fs_state;
+
+   draw->pipeline.aapoint = &aapoint->stage;
+
+   return TRUE;
+
+ fail:
+   if (aapoint)
+      aapoint->stage.destroy( &aapoint->stage );
+
+   return FALSE;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c
new file mode 100644
index 0000000000..3265dcd154
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c
@@ -0,0 +1,515 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief  Clipping stage
+ *
+ * \author  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "pipe/p_shader_tokens.h"
+
+#include "draw_vs.h"
+#include "draw_pipe.h"
+
+
+#ifndef IS_NEGATIVE
+#define IS_NEGATIVE(X) ((X) < 0.0)
+#endif
+
+#ifndef DIFFERENT_SIGNS
+#define DIFFERENT_SIGNS(x, y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F)
+#endif
+
+#ifndef MAX_CLIPPED_VERTICES
+#define MAX_CLIPPED_VERTICES ((2 * (6 + PIPE_MAX_CLIP_PLANES))+1)
+#endif
+
+
+
+struct clipper {
+   struct draw_stage stage;      /**< base class */
+
+   /* Basically duplicate some of the flatshading logic here:
+    */
+   boolean flat;
+   uint num_color_attribs;
+   uint color_attribs[4];  /* front/back primary/secondary colors */
+
+   float (*plane)[4];
+};
+
+
+/* This is a bit confusing:
+ */
+static INLINE struct clipper *clipper_stage( struct draw_stage *stage )
+{
+   return (struct clipper *)stage;
+}
+
+
+#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT)))
+
+
+/* All attributes are float[4], so this is easy:
+ */
+static void interp_attr( float *fdst,
+			 float t,
+			 const float *fin,
+			 const float *fout )
+{  
+   fdst[0] = LINTERP( t, fout[0], fin[0] );
+   fdst[1] = LINTERP( t, fout[1], fin[1] );
+   fdst[2] = LINTERP( t, fout[2], fin[2] );
+   fdst[3] = LINTERP( t, fout[3], fin[3] );
+}
+
+static void copy_colors( struct draw_stage *stage,
+			 struct vertex_header *dst,
+			 const struct vertex_header *src )
+{
+   const struct clipper *clipper = clipper_stage(stage);
+   uint i;
+   for (i = 0; i < clipper->num_color_attribs; i++) {
+      const uint attr = clipper->color_attribs[i];
+      COPY_4FV(dst->data[attr], src->data[attr]);
+   }
+}
+
+
+
+/* Interpolate between two vertices to produce a third.  
+ */
+static void interp( const struct clipper *clip,
+		    struct vertex_header *dst,
+		    float t,
+		    const struct vertex_header *out, 
+		    const struct vertex_header *in )
+{
+   const unsigned nr_attrs = clip->stage.draw->vs.num_vs_outputs;
+   const unsigned pos_attr = clip->stage.draw->vs.position_output;
+   unsigned j;
+
+   /* Vertex header.
+    */
+   {
+      dst->clipmask = 0;
+      dst->edgeflag = 0;        /* will get overwritten later */
+      dst->pad = 0;
+      dst->vertex_id = UNDEFINED_VERTEX_ID;
+   }
+
+   /* Clip coordinates:  interpolate normally
+    */
+   {
+      interp_attr(dst->clip, t, in->clip, out->clip);
+   }
+
+   /* Do the projective divide and insert window coordinates:
+    */
+   {
+      const float *pos = dst->clip;
+      const float *scale = clip->stage.draw->viewport.scale;
+      const float *trans = clip->stage.draw->viewport.translate;
+      const float oow = 1.0f / pos[3];
+
+      dst->data[pos_attr][0] = pos[0] * oow * scale[0] + trans[0];
+      dst->data[pos_attr][1] = pos[1] * oow * scale[1] + trans[1];
+      dst->data[pos_attr][2] = pos[2] * oow * scale[2] + trans[2];
+      dst->data[pos_attr][3] = oow;
+   }
+
+   /* Other attributes
+    */
+   for (j = 0; j < nr_attrs; j++) {
+      if (j != pos_attr)
+         interp_attr(dst->data[j], t, in->data[j], out->data[j]);
+   }
+}
+
+
+static void emit_poly( struct draw_stage *stage,
+		       struct vertex_header **inlist,
+		       unsigned n,
+		       const struct prim_header *origPrim)
+{
+   struct prim_header header;
+   unsigned i;
+
+   const ushort edge_first  = DRAW_PIPE_EDGE_FLAG_2;
+   const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0;
+   const ushort edge_last   = DRAW_PIPE_EDGE_FLAG_1;
+
+   /* later stages may need the determinant, but only the sign matters */
+   header.det = origPrim->det;
+   header.flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle;
+   header.pad = 0;
+
+   for (i = 2; i < n; i++, header.flags = edge_middle) {
+      header.v[0] = inlist[i-1];
+      header.v[1] = inlist[i];
+      header.v[2] = inlist[0];	/* keep in v[2] for flatshading */
+
+      if (i == n-1)
+        header.flags |= edge_last;
+
+      if (0) {
+         const struct draw_vertex_shader *vs = stage->draw->vs.vertex_shader;
+         uint j, k;
+         debug_printf("Clipped tri:\n");
+         for (j = 0; j < 3; j++) {
+            for (k = 0; k < vs->info.num_outputs; k++) {
+               debug_printf("  Vert %d: Attr %d:  %f %f %f %f\n", j, k,
+                            header.v[j]->data[k][0],
+                            header.v[j]->data[k][1],
+                            header.v[j]->data[k][2],
+                            header.v[j]->data[k][3]);
+            }
+         }
+      }
+
+      stage->next->tri( stage->next, &header );
+   }
+}
+
+static INLINE float
+dot4(const float *a, const float *b)
+{
+   return (a[0]*b[0] +
+           a[1]*b[1] +
+           a[2]*b[2] +
+           a[3]*b[3]);
+}
+
+
+/* Clip a triangle against the viewport and user clip planes.
+ */
+static void
+do_clip_tri( struct draw_stage *stage, 
+	     struct prim_header *header,
+	     unsigned clipmask )
+{
+   struct clipper *clipper = clipper_stage( stage );
+   struct vertex_header *a[MAX_CLIPPED_VERTICES];
+   struct vertex_header *b[MAX_CLIPPED_VERTICES];
+   struct vertex_header **inlist = a;
+   struct vertex_header **outlist = b;
+   unsigned tmpnr = 0;
+   unsigned n = 3;
+   unsigned i;
+
+   inlist[0] = header->v[0];
+   inlist[1] = header->v[1];
+   inlist[2] = header->v[2];
+
+   while (clipmask && n >= 3) {
+      const unsigned plane_idx = ffs(clipmask)-1;
+      const float *plane = clipper->plane[plane_idx];
+      struct vertex_header *vert_prev = inlist[0];
+      float dp_prev = dot4( vert_prev->clip, plane );
+      unsigned outcount = 0;
+
+      clipmask &= ~(1<<plane_idx);
+
+      inlist[n] = inlist[0]; /* prevent rotation of vertices */
+
+      for (i = 1; i <= n; i++) {
+	 struct vertex_header *vert = inlist[i];
+
+	 float dp = dot4( vert->clip, plane );
+
+	 if (!IS_NEGATIVE(dp_prev)) {
+	    outlist[outcount++] = vert_prev;
+	 }
+
+	 if (DIFFERENT_SIGNS(dp, dp_prev)) {
+	    struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++];
+	    outlist[outcount++] = new_vert;
+
+	    if (IS_NEGATIVE(dp)) {
+	       /* Going out of bounds.  Avoid division by zero as we
+		* know dp != dp_prev from DIFFERENT_SIGNS, above.
+		*/
+	       float t = dp / (dp - dp_prev);
+	       interp( clipper, new_vert, t, vert, vert_prev );
+	       
+	       /* Force edgeflag true in this case:
+		*/
+	       new_vert->edgeflag = 1;
+	    } else {
+	       /* Coming back in.
+		*/
+	       float t = dp_prev / (dp_prev - dp);
+	       interp( clipper, new_vert, t, vert_prev, vert );
+
+	       /* Copy starting vert's edgeflag:
+		*/
+	       new_vert->edgeflag = vert_prev->edgeflag;
+	    }
+	 }
+
+	 vert_prev = vert;
+	 dp_prev = dp;
+      }
+
+      {
+	 struct vertex_header **tmp = inlist;
+	 inlist = outlist;
+	 outlist = tmp;
+	 n = outcount;
+      }
+   }
+
+   /* If flat-shading, copy color to new provoking vertex.
+    */
+   if (clipper->flat && inlist[0] != header->v[2]) {
+      if (1) {
+	 inlist[0] = dup_vert(stage, inlist[0], tmpnr++);
+      }
+
+      copy_colors(stage, inlist[0], header->v[2]);
+   }
+
+
+
+   /* Emit the polygon as triangles to the setup stage:
+    */
+   if (n >= 3)
+      emit_poly( stage, inlist, n, header );
+}
+
+
+/* Clip a line against the viewport and user clip planes.
+ */
+static void
+do_clip_line( struct draw_stage *stage,
+	      struct prim_header *header,
+	      unsigned clipmask )
+{
+   const struct clipper *clipper = clipper_stage( stage );
+   struct vertex_header *v0 = header->v[0];
+   struct vertex_header *v1 = header->v[1];
+   const float *pos0 = v0->clip;
+   const float *pos1 = v1->clip;
+   float t0 = 0.0F;
+   float t1 = 0.0F;
+   struct prim_header newprim;
+
+   while (clipmask) {
+      const unsigned plane_idx = ffs(clipmask)-1;
+      const float *plane = clipper->plane[plane_idx];
+      const float dp0 = dot4( pos0, plane );
+      const float dp1 = dot4( pos1, plane );
+
+      if (dp1 < 0.0F) {
+	 float t = dp1 / (dp1 - dp0);
+         t1 = MAX2(t1, t);
+      } 
+
+      if (dp0 < 0.0F) {
+	 float t = dp0 / (dp0 - dp1);
+         t0 = MAX2(t0, t);
+      }
+
+      if (t0 + t1 >= 1.0F)
+	 return; /* discard */
+
+      clipmask &= ~(1 << plane_idx);  /* turn off this plane's bit */
+   }
+
+   if (v0->clipmask) {
+      interp( clipper, stage->tmp[0], t0, v0, v1 );
+
+      if (clipper->flat)
+	 copy_colors(stage, stage->tmp[0], v0);
+
+      newprim.v[0] = stage->tmp[0];
+   }
+   else {
+      newprim.v[0] = v0;
+   }
+
+   if (v1->clipmask) {
+      interp( clipper, stage->tmp[1], t1, v1, v0 );
+      newprim.v[1] = stage->tmp[1];
+   }
+   else {
+      newprim.v[1] = v1;
+   }
+
+   stage->next->line( stage->next, &newprim );
+}
+
+
+static void
+clip_point( struct draw_stage *stage, 
+	    struct prim_header *header )
+{
+   if (header->v[0]->clipmask == 0) 
+      stage->next->point( stage->next, header );
+}
+
+
+static void
+clip_line( struct draw_stage *stage,
+	   struct prim_header *header )
+{
+   unsigned clipmask = (header->v[0]->clipmask | 
+                        header->v[1]->clipmask);
+   
+   if (clipmask == 0) {
+      /* no clipping needed */
+      stage->next->line( stage->next, header );
+   }
+   else if ((header->v[0]->clipmask &
+             header->v[1]->clipmask) == 0) {
+      do_clip_line(stage, header, clipmask);
+   }
+   /* else, totally clipped */
+}
+
+
+static void
+clip_tri( struct draw_stage *stage,
+	  struct prim_header *header )
+{
+   unsigned clipmask = (header->v[0]->clipmask | 
+                        header->v[1]->clipmask | 
+                        header->v[2]->clipmask);
+   
+   if (clipmask == 0) {
+      /* no clipping needed */
+      stage->next->tri( stage->next, header );
+   }
+   else if ((header->v[0]->clipmask & 
+             header->v[1]->clipmask & 
+             header->v[2]->clipmask) == 0) {
+      do_clip_tri(stage, header, clipmask);
+   }
+}
+
+/* Update state.  Could further delay this until we hit the first
+ * primitive that really requires clipping.
+ */
+static void 
+clip_init_state( struct draw_stage *stage )
+{
+   struct clipper *clipper = clipper_stage( stage );
+
+   clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE;
+
+   if (clipper->flat) {
+      const struct draw_vertex_shader *vs = stage->draw->vs.vertex_shader;
+      uint i;
+
+      clipper->num_color_attribs = 0;
+      for (i = 0; i < vs->info.num_outputs; i++) {
+	 if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR ||
+	     vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) {
+	    clipper->color_attribs[clipper->num_color_attribs++] = i;
+	 }
+      }
+   }
+   
+   stage->tri = clip_tri;
+   stage->line = clip_line;
+}
+
+
+
+static void clip_first_tri( struct draw_stage *stage,
+			    struct prim_header *header )
+{
+   clip_init_state( stage );
+   stage->tri( stage, header );
+}
+
+static void clip_first_line( struct draw_stage *stage,
+			     struct prim_header *header )
+{
+   clip_init_state( stage );
+   stage->line( stage, header );
+}
+
+
+static void clip_flush( struct draw_stage *stage, 
+			     unsigned flags )
+{
+   stage->tri = clip_first_tri;
+   stage->line = clip_first_line;
+   stage->next->flush( stage->next, flags );
+}
+
+
+static void clip_reset_stipple_counter( struct draw_stage *stage )
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void clip_destroy( struct draw_stage *stage )
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+/**
+ * Allocate a new clipper stage.
+ * \return pointer to new stage object
+ */
+struct draw_stage *draw_clip_stage( struct draw_context *draw )
+{
+   struct clipper *clipper = CALLOC_STRUCT(clipper);
+   if (clipper == NULL)
+      goto fail;
+
+   if (!draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 ))
+      goto fail;
+
+   clipper->stage.draw = draw;
+   clipper->stage.point = clip_point;
+   clipper->stage.line = clip_first_line;
+   clipper->stage.tri = clip_first_tri;
+   clipper->stage.flush = clip_flush;
+   clipper->stage.reset_stipple_counter = clip_reset_stipple_counter;
+   clipper->stage.destroy = clip_destroy;
+
+   clipper->plane = draw->plane;
+
+   return &clipper->stage;
+
+ fail:
+   if (clipper)
+      clipper->stage.destroy( &clipper->stage );
+
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c
new file mode 100644
index 0000000000..053be5f050
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c
@@ -0,0 +1,147 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief  Drawing stage for polygon culling
+ */
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "util/u_memory.h"
+#include "pipe/p_defines.h"
+#include "draw_pipe.h"
+
+
+struct cull_stage {
+   struct draw_stage stage;
+   unsigned winding;  /**< which winding(s) to cull (one of PIPE_WINDING_x) */
+};
+
+
+static INLINE struct cull_stage *cull_stage( struct draw_stage *stage )
+{
+   return (struct cull_stage *)stage;
+}
+
+
+
+
+static void cull_tri( struct draw_stage *stage,
+		      struct prim_header *header )
+{
+   const unsigned pos = stage->draw->vs.position_output;
+
+   /* Window coords: */
+   const float *v0 = header->v[0]->data[pos];
+   const float *v1 = header->v[1]->data[pos];
+   const float *v2 = header->v[2]->data[pos];
+
+   /* edge vectors e = v0 - v2, f = v1 - v2 */
+   const float ex = v0[0] - v2[0];
+   const float ey = v0[1] - v2[1];
+   const float fx = v1[0] - v2[0];
+   const float fy = v1[1] - v2[1];
+   
+   /* det = cross(e,f).z */
+   header->det = ex * fy - ey * fx;
+
+   if (header->det != 0) {
+      /* if (det < 0 then Z points toward camera and triangle is 
+       * counter-clockwise winding.
+       */
+      unsigned winding = (header->det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW;
+
+      if ((winding & cull_stage(stage)->winding) == 0) {
+         /* triangle is not culled, pass to next stage */
+	 stage->next->tri( stage->next, header );
+      }
+   }
+}
+
+static void cull_first_tri( struct draw_stage *stage, 
+			    struct prim_header *header )
+{
+   struct cull_stage *cull = cull_stage(stage);
+
+   cull->winding = stage->draw->rasterizer->cull_mode;
+
+   stage->tri = cull_tri;
+   stage->tri( stage, header );
+}
+
+
+
+static void cull_flush( struct draw_stage *stage, unsigned flags )
+{
+   stage->tri = cull_first_tri;
+   stage->next->flush( stage->next, flags );
+}
+
+static void cull_reset_stipple_counter( struct draw_stage *stage )
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void cull_destroy( struct draw_stage *stage )
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+/**
+ * Create a new polygon culling stage.
+ */
+struct draw_stage *draw_cull_stage( struct draw_context *draw )
+{
+   struct cull_stage *cull = CALLOC_STRUCT(cull_stage);
+   if (cull == NULL)
+      goto fail;
+
+   if (!draw_alloc_temp_verts( &cull->stage, 0 ))
+      goto fail;
+
+   cull->stage.draw = draw;
+   cull->stage.next = NULL;
+   cull->stage.point = draw_pipe_passthrough_point;
+   cull->stage.line = draw_pipe_passthrough_line;
+   cull->stage.tri = cull_first_tri;
+   cull->stage.flush = cull_flush;
+   cull->stage.reset_stipple_counter = cull_reset_stipple_counter;
+   cull->stage.destroy = cull_destroy;
+
+   return &cull->stage;
+
+ fail:
+   if (cull)
+      cull->stage.destroy( &cull->stage );
+
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c
new file mode 100644
index 0000000000..43d1fecc4d
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c
@@ -0,0 +1,281 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "draw_vs.h"
+#include "draw_pipe.h"
+
+
+/** subclass of draw_stage */
+struct flat_stage
+{
+   struct draw_stage stage;
+
+   uint num_color_attribs;
+   uint color_attribs[2];  /* front/back primary colors */
+
+   uint num_spec_attribs;
+   uint spec_attribs[2];  /* front/back secondary colors */
+};
+
+#define COPY_3FV( DST, SRC )         \
+do {                                \
+   (DST)[0] = (SRC)[0];             \
+   (DST)[1] = (SRC)[1];             \
+   (DST)[2] = (SRC)[2];             \
+} while (0)
+
+
+static INLINE struct flat_stage *
+flat_stage(struct draw_stage *stage)
+{
+   return (struct flat_stage *) stage;
+}
+
+
+/** Copy all the color attributes from 'src' vertex to 'dst' vertex */
+static INLINE void copy_colors( struct draw_stage *stage,
+                                struct vertex_header *dst,
+                                const struct vertex_header *src )
+{
+   const struct flat_stage *flat = flat_stage(stage);
+   uint i;
+
+   for (i = 0; i < flat->num_color_attribs; i++) {
+      const uint attr = flat->color_attribs[i];
+      COPY_4FV(dst->data[attr], src->data[attr]);
+   }
+
+   for (i = 0; i < flat->num_spec_attribs; i++) {
+      const uint attr = flat->spec_attribs[i];
+      COPY_3FV(dst->data[attr], src->data[attr]);
+   }
+}
+
+
+/** Copy all the color attributes from src vertex to dst0 & dst1 vertices */
+static INLINE void copy_colors2( struct draw_stage *stage,
+                                 struct vertex_header *dst0,
+                                 struct vertex_header *dst1,
+                                 const struct vertex_header *src )
+{
+   const struct flat_stage *flat = flat_stage(stage);
+   uint i;
+   for (i = 0; i < flat->num_color_attribs; i++) {
+      const uint attr = flat->color_attribs[i];
+      COPY_4FV(dst0->data[attr], src->data[attr]);
+      COPY_4FV(dst1->data[attr], src->data[attr]);
+   }
+
+   for (i = 0; i < flat->num_spec_attribs; i++) {
+      const uint attr = flat->spec_attribs[i];
+      COPY_3FV(dst0->data[attr], src->data[attr]);
+      COPY_3FV(dst1->data[attr], src->data[attr]);
+   }
+}
+
+
+/**
+ * Flatshade tri.  Required for clipping and when unfilled tris are
+ * active, otherwise handled by hardware.
+ */
+static void flatshade_tri_0( struct draw_stage *stage,
+                             struct prim_header *header )
+{
+   struct prim_header tmp;
+
+   tmp.det = header->det;
+   tmp.flags = header->flags;
+   tmp.pad = header->pad;
+   tmp.v[0] = header->v[0];
+   tmp.v[1] = dup_vert(stage, header->v[1], 0);
+   tmp.v[2] = dup_vert(stage, header->v[2], 1);
+
+   copy_colors2(stage, tmp.v[1], tmp.v[2], tmp.v[0]);
+   
+   stage->next->tri( stage->next, &tmp );
+}
+
+
+static void flatshade_tri_2( struct draw_stage *stage,
+                             struct prim_header *header )
+{
+   struct prim_header tmp;
+
+   tmp.det = header->det;
+   tmp.flags = header->flags;
+   tmp.pad = header->pad;
+   tmp.v[0] = dup_vert(stage, header->v[0], 0);
+   tmp.v[1] = dup_vert(stage, header->v[1], 1);
+   tmp.v[2] = header->v[2];
+
+   copy_colors2(stage, tmp.v[0], tmp.v[1], tmp.v[2]);
+   
+   stage->next->tri( stage->next, &tmp );
+}
+
+
+
+
+
+/**
+ * Flatshade line.  Required for clipping.
+ */
+static void flatshade_line_0( struct draw_stage *stage,
+                              struct prim_header *header )
+{
+   struct prim_header tmp;
+
+   tmp.v[0] = header->v[0];
+   tmp.v[1] = dup_vert(stage, header->v[1], 0);
+
+   copy_colors(stage, tmp.v[1], tmp.v[0]);
+   
+   stage->next->line( stage->next, &tmp );
+}
+
+static void flatshade_line_1( struct draw_stage *stage,
+                              struct prim_header *header )
+{
+   struct prim_header tmp;
+
+   tmp.v[0] = dup_vert(stage, header->v[0], 0);
+   tmp.v[1] = header->v[1];
+
+   copy_colors(stage, tmp.v[0], tmp.v[1]);
+   
+   stage->next->line( stage->next, &tmp );
+}
+
+
+
+
+static void flatshade_init_state( struct draw_stage *stage )
+{
+   struct flat_stage *flat = flat_stage(stage);
+   const struct draw_vertex_shader *vs = stage->draw->vs.vertex_shader;
+   uint i;
+
+   /* Find which vertex shader outputs are colors, make a list */
+   flat->num_color_attribs = 0;
+   flat->num_spec_attribs = 0;
+   for (i = 0; i < vs->info.num_outputs; i++) {
+      if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR ||
+          vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) {
+         if (vs->info.output_semantic_index[i] == 0)
+            flat->color_attribs[flat->num_color_attribs++] = i;
+         else
+            flat->spec_attribs[flat->num_spec_attribs++] = i;
+      }
+   }
+
+   /* Choose flatshade routine according to provoking vertex:
+    */
+   if (stage->draw->rasterizer->flatshade_first) {
+      stage->line = flatshade_line_0;
+      stage->tri = flatshade_tri_0;
+   }
+   else {
+      stage->line = flatshade_line_1;
+      stage->tri = flatshade_tri_2;
+   }
+}
+
+static void flatshade_first_tri( struct draw_stage *stage,
+				 struct prim_header *header )
+{
+   flatshade_init_state( stage );
+   stage->tri( stage, header );
+}
+
+static void flatshade_first_line( struct draw_stage *stage,
+				  struct prim_header *header )
+{
+   flatshade_init_state( stage );
+   stage->line( stage, header );
+}
+
+
+static void flatshade_flush( struct draw_stage *stage, 
+			     unsigned flags )
+{
+   stage->tri = flatshade_first_tri;
+   stage->line = flatshade_first_line;
+   stage->next->flush( stage->next, flags );
+}
+
+
+static void flatshade_reset_stipple_counter( struct draw_stage *stage )
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void flatshade_destroy( struct draw_stage *stage )
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+/**
+ * Create flatshading drawing stage.
+ */
+struct draw_stage *draw_flatshade_stage( struct draw_context *draw )
+{
+   struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage);
+   if (flatshade == NULL)
+      goto fail;
+
+   if (!draw_alloc_temp_verts( &flatshade->stage, 2 ))
+      goto fail;
+
+   flatshade->stage.draw = draw;
+   flatshade->stage.next = NULL;
+   flatshade->stage.point = draw_pipe_passthrough_point;
+   flatshade->stage.line = flatshade_first_line;
+   flatshade->stage.tri = flatshade_first_tri;
+   flatshade->stage.flush = flatshade_flush;
+   flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter;
+   flatshade->stage.destroy = flatshade_destroy;
+
+   return &flatshade->stage;
+
+ fail:
+   if (flatshade)
+      flatshade->stage.destroy( &flatshade->stage );
+
+   return NULL;
+}
+
+
diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c
new file mode 100644
index 0000000000..62c31532d0
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c
@@ -0,0 +1,184 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief  polygon offset state
+ *
+ * \author  Keith Whitwell <keith@tungstengraphics.com>
+ * \author  Brian Paul
+ */
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "draw_pipe.h"
+
+
+
+struct offset_stage {
+   struct draw_stage stage;
+
+   float scale;
+   float units;
+};
+
+
+
+static INLINE struct offset_stage *offset_stage( struct draw_stage *stage )
+{
+   return (struct offset_stage *) stage;
+}
+
+
+
+
+
+/**
+ * Offset tri Z.  Some hardware can handle this, but not usually when
+ * doing unfilled rendering.
+ */
+static void do_offset_tri( struct draw_stage *stage,
+			   struct prim_header *header )
+{
+   const unsigned pos = stage->draw->vs.position_output;
+   struct offset_stage *offset = offset_stage(stage);   
+   float inv_det = 1.0f / header->det;
+
+   /* Window coords:
+    */
+   float *v0 = header->v[0]->data[pos];
+   float *v1 = header->v[1]->data[pos];
+   float *v2 = header->v[2]->data[pos];
+
+   /* edge vectors e = v0 - v2, f = v1 - v2 */
+   float ex = v0[0] - v2[0];
+   float ey = v0[1] - v2[1];
+   float ez = v0[2] - v2[2];
+   float fx = v1[0] - v2[0];
+   float fy = v1[1] - v2[1];
+   float fz = v1[2] - v2[2];
+
+   /* (a,b) = cross(e,f).xy */
+   float a = ey*fz - ez*fy;
+   float b = ez*fx - ex*fz;
+
+   float dzdx = fabsf(a * inv_det);
+   float dzdy = fabsf(b * inv_det);
+
+   float zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale;
+
+   /*
+    * Note: we're applying the offset and clamping per-vertex.
+    * Ideally, the offset is applied per-fragment prior to fragment shading.
+    */
+   v0[2] = CLAMP(v0[2] + zoffset, 0.0f, 1.0f);
+   v1[2] = CLAMP(v1[2] + zoffset, 0.0f, 1.0f);
+   v2[2] = CLAMP(v2[2] + zoffset, 0.0f, 1.0f);
+
+   stage->next->tri( stage->next, header );
+}
+
+
+static void offset_tri( struct draw_stage *stage,
+			struct prim_header *header )
+{
+   struct prim_header tmp;
+
+   tmp.det = header->det;
+   tmp.flags = header->flags;
+   tmp.pad = header->pad;
+   tmp.v[0] = dup_vert(stage, header->v[0], 0);
+   tmp.v[1] = dup_vert(stage, header->v[1], 1);
+   tmp.v[2] = dup_vert(stage, header->v[2], 2);
+
+   do_offset_tri( stage, &tmp );
+}
+
+
+static void offset_first_tri( struct draw_stage *stage, 
+			      struct prim_header *header )
+{
+   struct offset_stage *offset = offset_stage(stage);
+
+   offset->units = (float) (stage->draw->rasterizer->offset_units * stage->draw->mrd);
+   offset->scale = stage->draw->rasterizer->offset_scale;
+
+   stage->tri = offset_tri;
+   stage->tri( stage, header );
+}
+
+
+
+
+static void offset_flush( struct draw_stage *stage,
+			  unsigned flags )
+{
+   stage->tri = offset_first_tri;
+   stage->next->flush( stage->next, flags );
+}
+
+
+static void offset_reset_stipple_counter( struct draw_stage *stage )
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void offset_destroy( struct draw_stage *stage )
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+/**
+ * Create polygon offset drawing stage.
+ */
+struct draw_stage *draw_offset_stage( struct draw_context *draw )
+{
+   struct offset_stage *offset = CALLOC_STRUCT(offset_stage);
+   if (offset == NULL)
+      goto fail;
+
+   draw_alloc_temp_verts( &offset->stage, 3 );
+
+   offset->stage.draw = draw;
+   offset->stage.next = NULL;
+   offset->stage.point = draw_pipe_passthrough_point;
+   offset->stage.line = draw_pipe_passthrough_line;
+   offset->stage.tri = offset_first_tri;
+   offset->stage.flush = offset_flush;
+   offset->stage.reset_stipple_counter = offset_reset_stipple_counter;
+   offset->stage.destroy = offset_destroy;
+
+   return &offset->stage;
+
+ fail:
+   if (offset)
+      offset->stage.destroy( &offset->stage );
+
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
new file mode 100644
index 0000000000..a0f9716dac
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c
@@ -0,0 +1,770 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Polygon stipple stage:  implement polygon stipple with texture map and
+ * fragment program.  The fragment program samples the texture and does
+ * a fragment kill for the stipple-failing fragments.
+ *
+ * Authors:  Brian Paul
+ */
+
+
+#include "pipe/p_inlines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "tgsi/tgsi_transform.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "draw_context.h"
+#include "draw_pipe.h"
+
+
+
+/**
+ * Subclass of pipe_shader_state to carry extra fragment shader info.
+ */
+struct pstip_fragment_shader
+{
+   struct pipe_shader_state state;
+   void *driver_fs;
+   void *pstip_fs;
+   uint sampler_unit;
+};
+
+
+/**
+ * Subclass of draw_stage
+ */
+struct pstip_stage
+{
+   struct draw_stage stage;
+
+   void *sampler_cso;
+   struct pipe_texture *texture;
+   uint num_samplers;
+   uint num_textures;
+
+   /*
+    * Currently bound state
+    */
+   struct pstip_fragment_shader *fs;
+   struct {
+      void *samplers[PIPE_MAX_SAMPLERS];
+      struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+      const struct pipe_poly_stipple *stipple;
+   } state;
+
+   /*
+    * Driver interface/override functions
+    */
+   void * (*driver_create_fs_state)(struct pipe_context *,
+                                    const struct pipe_shader_state *);
+   void (*driver_bind_fs_state)(struct pipe_context *, void *);
+   void (*driver_delete_fs_state)(struct pipe_context *, void *);
+
+   void (*driver_bind_sampler_states)(struct pipe_context *, unsigned, void **);
+
+   void (*driver_set_sampler_textures)(struct pipe_context *, unsigned,
+                                       struct pipe_texture **);
+
+   void (*driver_set_polygon_stipple)(struct pipe_context *,
+                                      const struct pipe_poly_stipple *);
+
+   struct pipe_context *pipe;
+};
+
+
+
+/**
+ * Subclass of tgsi_transform_context, used for transforming the
+ * user's fragment shader to add the special AA instructions.
+ */
+struct pstip_transform_context {
+   struct tgsi_transform_context base;
+   uint tempsUsed;  /**< bitmask */
+   int wincoordInput;
+   int maxInput;
+   uint samplersUsed;  /**< bitfield of samplers used */
+   int freeSampler;  /** an available sampler for the pstipple */
+   int texTemp;  /**< temp registers */
+   int numImmed;
+   boolean firstInstruction;
+};
+
+
+/**
+ * TGSI declaration transform callback.
+ * Look for a free sampler, a free input attrib, and two free temp regs.
+ */
+static void
+pstip_transform_decl(struct tgsi_transform_context *ctx,
+                     struct tgsi_full_declaration *decl)
+{
+   struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx;
+
+   if (decl->Declaration.File == TGSI_FILE_SAMPLER) {
+      uint i;
+      for (i = decl->DeclarationRange.First;
+           i <= decl->DeclarationRange.Last; i++) {
+         pctx->samplersUsed |= 1 << i;
+      }
+   }
+   else if (decl->Declaration.File == TGSI_FILE_INPUT) {
+      pctx->maxInput = MAX2(pctx->maxInput, (int) decl->DeclarationRange.Last);
+      if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION)
+         pctx->wincoordInput = (int) decl->DeclarationRange.First;
+   }
+   else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) {
+      uint i;
+      for (i = decl->DeclarationRange.First;
+           i <= decl->DeclarationRange.Last; i++) {
+         pctx->tempsUsed |= (1 << i);
+      }
+   }
+
+   ctx->emit_declaration(ctx, decl);
+}
+
+
+static void
+pstip_transform_immed(struct tgsi_transform_context *ctx,
+                      struct tgsi_full_immediate *immed)
+{
+   struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx;
+   pctx->numImmed++;
+}
+
+
+/**
+ * Find the lowest zero bit in the given word, or -1 if bitfield is all ones.
+ */
+static int
+free_bit(uint bitfield)
+{
+   int i;
+   for (i = 0; i < 32; i++) {
+      if ((bitfield & (1 << i)) == 0)
+         return i;
+   }
+   return -1;
+}
+
+
+/**
+ * TGSI instruction transform callback.
+ * Replace writes to result.color w/ a temp reg.
+ * Upon END instruction, insert texture sampling code for antialiasing.
+ */
+static void
+pstip_transform_inst(struct tgsi_transform_context *ctx,
+                     struct tgsi_full_instruction *inst)
+{
+   struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx;
+
+   if (pctx->firstInstruction) {
+      /* emit our new declarations before the first instruction */
+
+      struct tgsi_full_declaration decl;
+      struct tgsi_full_instruction newInst;
+      uint i;
+      int wincoordInput;
+
+      /* find free sampler */
+      pctx->freeSampler = free_bit(pctx->samplersUsed);
+      if (pctx->freeSampler >= PIPE_MAX_SAMPLERS)
+         pctx->freeSampler = PIPE_MAX_SAMPLERS - 1;
+
+      if (pctx->wincoordInput < 0)
+         wincoordInput = pctx->maxInput + 1;
+      else
+         wincoordInput = pctx->wincoordInput;
+
+      /* find one free temp reg */
+      for (i = 0; i < 32; i++) {
+         if ((pctx->tempsUsed & (1 << i)) == 0) {
+            /* found a free temp */
+            if (pctx->texTemp < 0)
+               pctx->texTemp  = i;
+            else
+               break;
+         }
+      }
+      assert(pctx->texTemp >= 0);
+
+      if (pctx->wincoordInput < 0) {
+         /* declare new position input reg */
+         decl = tgsi_default_full_declaration();
+         decl.Declaration.File = TGSI_FILE_INPUT;
+         decl.Declaration.Interpolate = TGSI_INTERPOLATE_LINEAR; /* XXX? */
+         decl.Declaration.Semantic = 1;
+         decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION;
+         decl.Semantic.SemanticIndex = 0;
+         decl.DeclarationRange.First = 
+            decl.DeclarationRange.Last = wincoordInput;
+         ctx->emit_declaration(ctx, &decl);
+      }
+
+      /* declare new sampler */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_SAMPLER;
+      decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = pctx->freeSampler;
+      ctx->emit_declaration(ctx, &decl);
+
+      /* declare new temp regs */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_TEMPORARY;
+      decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = pctx->texTemp;
+      ctx->emit_declaration(ctx, &decl);
+
+      /* emit immediate = {1/32, 1/32, 1, 1}
+       * The index/position of this immediate will be pctx->numImmed
+       */
+      {
+         static const float value[4] = { 1.0/32, 1.0/32, 1.0, 1.0 };
+         struct tgsi_full_immediate immed;
+         uint size = 4;
+         immed = tgsi_default_full_immediate();
+         immed.Immediate.NrTokens = 1 + size; /* one for the token itself */
+         immed.u.Pointer = (void *) value;
+         ctx->emit_immediate(ctx, &immed);
+      }
+
+      pctx->firstInstruction = FALSE;
+
+
+      /* 
+       * Insert new MUL/TEX/KILP instructions at start of program
+       * Take gl_FragCoord, divide by 32 (stipple size), sample the
+       * texture and kill fragment if needed.
+       *
+       * We'd like to use non-normalized texcoords to index into a RECT
+       * texture, but we can only use GL_REPEAT wrap mode with normalized
+       * texcoords.  Darn.
+       */
+
+      /* MUL texTemp, INPUT[wincoord], 1/32; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = wincoordInput;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_IMMEDIATE;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->numImmed;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* TEX texTemp, texTemp, sampler; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_TEX;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->freeSampler;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* KIL -texTemp;   # if -texTemp < 0, KILL fragment */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_KIL;
+      newInst.Instruction.NumDstRegs = 0;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp;
+      newInst.FullSrcRegisters[0].SrcRegister.Negate = 1;
+      ctx->emit_instruction(ctx, &newInst);
+   }
+
+   /* emit this instruction */
+   ctx->emit_instruction(ctx, inst);
+}
+
+
+/**
+ * Generate the frag shader we'll use for doing polygon stipple.
+ * This will be the user's shader prefixed with a TEX and KIL instruction.
+ */
+static boolean
+generate_pstip_fs(struct pstip_stage *pstip)
+{
+   const struct pipe_shader_state *orig_fs = &pstip->fs->state;
+   /*struct draw_context *draw = pstip->stage.draw;*/
+   struct pipe_shader_state pstip_fs;
+   struct pstip_transform_context transform;
+
+#define MAX 1000
+
+   pstip_fs = *orig_fs; /* copy to init */
+   pstip_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX);
+   if (pstip_fs.tokens == NULL)
+      return FALSE;
+
+   memset(&transform, 0, sizeof(transform));
+   transform.wincoordInput = -1;
+   transform.maxInput = -1;
+   transform.texTemp = -1;
+   transform.firstInstruction = TRUE;
+   transform.base.transform_instruction = pstip_transform_inst;
+   transform.base.transform_declaration = pstip_transform_decl;
+   transform.base.transform_immediate = pstip_transform_immed;
+
+   tgsi_transform_shader(orig_fs->tokens,
+                         (struct tgsi_token *) pstip_fs.tokens,
+                         MAX, &transform.base);
+
+#if 0 /* DEBUG */
+   tgsi_dump(orig_fs->tokens, 0);
+   tgsi_dump(pstip_fs.tokens, 0);
+#endif
+
+   pstip->fs->sampler_unit = transform.freeSampler;
+   assert(pstip->fs->sampler_unit < PIPE_MAX_SAMPLERS);
+
+   pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs);
+
+   return TRUE;
+}
+
+
+/**
+ * Load texture image with current stipple pattern.
+ */
+static void
+pstip_update_texture(struct pstip_stage *pstip)
+{
+   static const uint bit31 = 1 << 31;
+   struct pipe_context *pipe = pstip->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surface;
+   const uint *stipple = pstip->state.stipple->stipple;
+   uint i, j;
+   ubyte *data;
+
+   /* XXX: want to avoid flushing just because we use stipple: 
+    */
+   pipe->flush( pipe, PIPE_FLUSH_TEXTURE_CACHE, NULL );
+
+   surface = screen->get_tex_surface(screen, pstip->texture, 0, 0, 0,
+                                     PIPE_BUFFER_USAGE_CPU_WRITE);
+   data = screen->surface_map(screen, surface,
+                              PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /*
+    * Load alpha texture.
+    * Note: 0 means keep the fragment, 255 means kill it.
+    * We'll negate the texel value and use KILP which kills if value
+    * is negative.
+    */
+   for (i = 0; i < 32; i++) {
+      for (j = 0; j < 32; j++) {
+         if (stipple[i] & (bit31 >> j)) {
+            /* fragment "on" */
+            data[i * surface->stride + j] = 0;
+         }
+         else {
+            /* fragment "off" */
+            data[i * surface->stride + j] = 255;
+         }
+      }
+   }
+
+   /* unmap */
+   screen->surface_unmap(screen, surface);
+   screen->tex_surface_release(screen, &surface);
+}
+
+
+/**
+ * Create the texture map we'll use for stippling.
+ */
+static boolean
+pstip_create_texture(struct pstip_stage *pstip)
+{
+   struct pipe_context *pipe = pstip->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture texTemp;
+
+   memset(&texTemp, 0, sizeof(texTemp));
+   texTemp.target = PIPE_TEXTURE_2D;
+   texTemp.format = PIPE_FORMAT_A8_UNORM; /* XXX verify supported by driver! */
+   texTemp.last_level = 0;
+   texTemp.width[0] = 32;
+   texTemp.height[0] = 32;
+   texTemp.depth[0] = 1;
+   pf_get_block(texTemp.format, &texTemp.block);
+
+   pstip->texture = screen->texture_create(screen, &texTemp);
+   if (pstip->texture == NULL)
+      return FALSE;
+
+   return TRUE;
+}
+
+
+/**
+ * Create the sampler CSO that'll be used for stippling.
+ */
+static boolean
+pstip_create_sampler(struct pstip_stage *pstip)
+{
+   struct pipe_sampler_state sampler;
+   struct pipe_context *pipe = pstip->pipe;
+
+   memset(&sampler, 0, sizeof(sampler));
+   sampler.wrap_s = PIPE_TEX_WRAP_REPEAT;
+   sampler.wrap_t = PIPE_TEX_WRAP_REPEAT;
+   sampler.wrap_r = PIPE_TEX_WRAP_REPEAT;
+   sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+   sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+   sampler.normalized_coords = 1;
+   sampler.min_lod = 0.0f;
+   sampler.max_lod = 0.0f;
+
+   pstip->sampler_cso = pipe->create_sampler_state(pipe, &sampler);
+   if (pstip->sampler_cso == NULL)
+      return FALSE;
+   
+   return TRUE;
+}
+
+
+/**
+ * When we're about to draw our first stipple polygon in a batch, this function
+ * is called to tell the driver to bind our modified fragment shader.
+ */
+static boolean
+bind_pstip_fragment_shader(struct pstip_stage *pstip)
+{
+   struct draw_context *draw = pstip->stage.draw;
+   if (!pstip->fs->pstip_fs &&
+       !generate_pstip_fs(pstip))
+      return FALSE;
+
+   draw->suspend_flushing = TRUE;
+   pstip->driver_bind_fs_state(pstip->pipe, pstip->fs->pstip_fs);
+   draw->suspend_flushing = FALSE;
+   return TRUE;
+}
+
+
+static INLINE struct pstip_stage *
+pstip_stage( struct draw_stage *stage )
+{
+   return (struct pstip_stage *) stage;
+}
+
+
+static void
+pstip_first_tri(struct draw_stage *stage, struct prim_header *header)
+{
+   struct pstip_stage *pstip = pstip_stage(stage);
+   struct pipe_context *pipe = pstip->pipe;
+   struct draw_context *draw = stage->draw;
+   uint num_samplers;
+
+   assert(stage->draw->rasterizer->poly_stipple_enable);
+
+   /* bind our fragprog */
+   if (!bind_pstip_fragment_shader(pstip)) {
+      stage->tri = draw_pipe_passthrough_tri;
+      stage->tri(stage, header);
+      return;
+   }
+      
+
+   /* how many samplers? */
+   /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */
+   num_samplers = MAX2(pstip->num_textures, pstip->num_samplers);
+   num_samplers = MAX2(num_samplers, pstip->fs->sampler_unit + 1);
+
+   /* plug in our sampler, texture */
+   pstip->state.samplers[pstip->fs->sampler_unit] = pstip->sampler_cso;
+   pipe_texture_reference(&pstip->state.textures[pstip->fs->sampler_unit],
+                          pstip->texture);
+
+   assert(num_samplers <= PIPE_MAX_SAMPLERS);
+
+   draw->suspend_flushing = TRUE;
+   pstip->driver_bind_sampler_states(pipe, num_samplers, pstip->state.samplers);
+   pstip->driver_set_sampler_textures(pipe, num_samplers, pstip->state.textures);
+   draw->suspend_flushing = FALSE;
+
+   /* now really draw first triangle */
+   stage->tri = draw_pipe_passthrough_tri;
+   stage->tri(stage, header);
+}
+
+
+static void
+pstip_flush(struct draw_stage *stage, unsigned flags)
+{
+   struct draw_context *draw = stage->draw;
+   struct pstip_stage *pstip = pstip_stage(stage);
+   struct pipe_context *pipe = pstip->pipe;
+
+   stage->tri = pstip_first_tri;
+   stage->next->flush( stage->next, flags );
+
+   /* restore original frag shader, texture, sampler state */
+   draw->suspend_flushing = TRUE;
+   pstip->driver_bind_fs_state(pipe, pstip->fs->driver_fs);
+   pstip->driver_bind_sampler_states(pipe, pstip->num_samplers,
+                                     pstip->state.samplers);
+   pstip->driver_set_sampler_textures(pipe, pstip->num_textures,
+                                      pstip->state.textures);
+   draw->suspend_flushing = FALSE;
+}
+
+
+static void
+pstip_reset_stipple_counter(struct draw_stage *stage)
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void
+pstip_destroy(struct draw_stage *stage)
+{
+   struct pstip_stage *pstip = pstip_stage(stage);
+   uint i;
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      pipe_texture_reference(&pstip->state.textures[i], NULL);
+   }
+
+   pstip->pipe->delete_sampler_state(pstip->pipe, pstip->sampler_cso);
+
+   pipe_texture_release(&pstip->texture);
+
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+static struct pstip_stage *
+draw_pstip_stage(struct draw_context *draw)
+{
+   struct pstip_stage *pstip = CALLOC_STRUCT(pstip_stage);
+
+   draw_alloc_temp_verts( &pstip->stage, 8 );
+
+   pstip->stage.draw = draw;
+   pstip->stage.next = NULL;
+   pstip->stage.point = draw_pipe_passthrough_point;
+   pstip->stage.line = draw_pipe_passthrough_line;
+   pstip->stage.tri = pstip_first_tri;
+   pstip->stage.flush = pstip_flush;
+   pstip->stage.reset_stipple_counter = pstip_reset_stipple_counter;
+   pstip->stage.destroy = pstip_destroy;
+
+   return pstip;
+}
+
+
+static struct pstip_stage *
+pstip_stage_from_pipe(struct pipe_context *pipe)
+{
+   struct draw_context *draw = (struct draw_context *) pipe->draw;
+   return pstip_stage(draw->pipeline.pstipple);
+}
+
+
+/**
+ * This function overrides the driver's create_fs_state() function and
+ * will typically be called by the state tracker.
+ */
+static void *
+pstip_create_fs_state(struct pipe_context *pipe,
+                       const struct pipe_shader_state *fs)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   struct pstip_fragment_shader *aafs = CALLOC_STRUCT(pstip_fragment_shader);
+
+   if (aafs) {
+      aafs->state = *fs;
+
+      /* pass-through */
+      aafs->driver_fs = pstip->driver_create_fs_state(pstip->pipe, fs);
+   }
+
+   return aafs;
+}
+
+
+static void
+pstip_bind_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs;
+   /* save current */
+   pstip->fs = aafs;
+   /* pass-through */
+   pstip->driver_bind_fs_state(pstip->pipe,
+                               (aafs ? aafs->driver_fs : NULL));
+}
+
+
+static void
+pstip_delete_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs;
+   /* pass-through */
+   pstip->driver_delete_fs_state(pstip->pipe, aafs->driver_fs);
+   FREE(aafs);
+}
+
+
+static void
+pstip_bind_sampler_states(struct pipe_context *pipe,
+                          unsigned num, void **sampler)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   uint i;
+
+   /* save current */
+   memcpy(pstip->state.samplers, sampler, num * sizeof(void *));
+   for (i = num; i < PIPE_MAX_SAMPLERS; i++) {
+      pstip->state.samplers[i] = NULL;
+   }
+
+   pstip->num_samplers = num;
+   /* pass-through */
+   pstip->driver_bind_sampler_states(pstip->pipe, num, sampler);
+}
+
+
+static void
+pstip_set_sampler_textures(struct pipe_context *pipe,
+                           unsigned num, struct pipe_texture **texture)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   uint i;
+
+   /* save current */
+   for (i = 0; i < num; i++) {
+      pipe_texture_reference(&pstip->state.textures[i], texture[i]);
+   }
+   for (; i < PIPE_MAX_SAMPLERS; i++) {
+      pipe_texture_reference(&pstip->state.textures[i], NULL);
+   }
+
+   pstip->num_textures = num;
+
+   /* pass-through */
+   pstip->driver_set_sampler_textures(pstip->pipe, num, texture);
+}
+
+
+static void
+pstip_set_polygon_stipple(struct pipe_context *pipe,
+                          const struct pipe_poly_stipple *stipple)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+
+   /* save current */
+   pstip->state.stipple = stipple;
+
+   /* pass-through */
+   pstip->driver_set_polygon_stipple(pstip->pipe, stipple);
+
+   pstip_update_texture(pstip);
+}
+
+
+/**
+ * Called by drivers that want to install this polygon stipple stage
+ * into the draw module's pipeline.  This will not be used if the
+ * hardware has native support for polygon stipple.
+ */
+boolean
+draw_install_pstipple_stage(struct draw_context *draw,
+                            struct pipe_context *pipe)
+{
+   struct pstip_stage *pstip;
+
+   pipe->draw = (void *) draw;
+
+   /*
+    * Create / install pgon stipple drawing / prim stage
+    */
+   pstip = draw_pstip_stage( draw );
+   if (pstip == NULL)
+      goto fail;
+
+   draw->pipeline.pstipple = &pstip->stage;
+
+   pstip->pipe = pipe;
+
+   /* create special texture, sampler state */
+   if (!pstip_create_texture(pstip))
+      goto fail;
+
+   if (!pstip_create_sampler(pstip))
+      goto fail;
+
+   /* save original driver functions */
+   pstip->driver_create_fs_state = pipe->create_fs_state;
+   pstip->driver_bind_fs_state = pipe->bind_fs_state;
+   pstip->driver_delete_fs_state = pipe->delete_fs_state;
+
+   pstip->driver_bind_sampler_states = pipe->bind_sampler_states;
+   pstip->driver_set_sampler_textures = pipe->set_sampler_textures;
+   pstip->driver_set_polygon_stipple = pipe->set_polygon_stipple;
+
+   /* override the driver's functions */
+   pipe->create_fs_state = pstip_create_fs_state;
+   pipe->bind_fs_state = pstip_bind_fs_state;
+   pipe->delete_fs_state = pstip_delete_fs_state;
+
+   pipe->bind_sampler_states = pstip_bind_sampler_states;
+   pipe->set_sampler_textures = pstip_set_sampler_textures;
+   pipe->set_polygon_stipple = pstip_set_polygon_stipple;
+
+   return TRUE;
+
+ fail:
+   if (pstip)
+      pstip->stage.destroy( &pstip->stage );
+
+   return FALSE;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c
new file mode 100644
index 0000000000..b65e2aa102
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c
@@ -0,0 +1,250 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/* Implement line stipple by cutting lines up into smaller lines.
+ * There are hundreds of ways to implement line stipple, this is one
+ * choice that should work in all situations, requires no state
+ * manipulations, but with a penalty in terms of large amounts of
+ * generated geometry.
+ */
+
+
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "draw/draw_pipe.h"
+
+
+/** Subclass of draw_stage */
+struct stipple_stage {
+   struct draw_stage stage;
+   float counter;
+   uint pattern;
+   uint factor;
+};
+
+
+static INLINE struct stipple_stage *
+stipple_stage(struct draw_stage *stage)
+{
+   return (struct stipple_stage *) stage;
+}
+
+
+/**
+ * Compute interpolated vertex attributes for 'dst' at position 't' 
+ * between 'v0' and 'v1'.
+ * XXX using linear interpolation for all attribs at this time.
+ */
+static void
+screen_interp( struct draw_context *draw,
+               struct vertex_header *dst,
+               float t,
+               const struct vertex_header *v0, 
+               const struct vertex_header *v1 )
+{
+   uint attr;
+   for (attr = 0; attr < draw->vs.num_vs_outputs; attr++) {
+      const float *val0 = v0->data[attr];
+      const float *val1 = v1->data[attr];
+      float *newv = dst->data[attr];
+      uint i;
+      for (i = 0; i < 4; i++) {
+         newv[i] = val0[i] + t * (val1[i] - val0[i]);
+      }
+   }
+}
+
+
+static void
+emit_segment(struct draw_stage *stage, struct prim_header *header,
+             float t0, float t1)
+{
+   struct vertex_header *v0new = dup_vert(stage, header->v[0], 0);
+   struct vertex_header *v1new = dup_vert(stage, header->v[1], 1);
+   struct prim_header newprim = *header;
+
+   if (t0 > 0.0) {
+      screen_interp( stage->draw, v0new, t0, header->v[0], header->v[1] );
+      newprim.v[0] = v0new;
+   }
+
+   if (t1 < 1.0) {
+      screen_interp( stage->draw, v1new, t1, header->v[0], header->v[1] );
+      newprim.v[1] = v1new;
+   }
+
+   stage->next->line( stage->next, &newprim );
+}
+
+
+static INLINE unsigned
+stipple_test(int counter, ushort pattern, int factor)
+{
+   int b = (counter / factor) & 0xf;
+   return (1 << b) & pattern;
+}
+
+
+static void
+stipple_line(struct draw_stage *stage, struct prim_header *header)
+{
+   struct stipple_stage *stipple = stipple_stage(stage);
+   struct vertex_header *v0 = header->v[0];
+   struct vertex_header *v1 = header->v[1];
+   const unsigned pos = stage->draw->vs.position_output;
+   const float *pos0 = v0->data[pos];
+   const float *pos1 = v1->data[pos];
+   float start = 0;
+   int state = 0;
+
+   float x0 = pos0[0];
+   float x1 = pos1[0];
+   float y0 = pos0[1];
+   float y1 = pos1[1];
+
+   float dx = x0 > x1 ? x0 - x1 : x1 - x0;
+   float dy = y0 > y1 ? y0 - y1 : y1 - y0;
+
+   float length = MAX2(dx, dy);
+   int i;
+
+   if (header->flags & DRAW_PIPE_RESET_STIPPLE)
+      stipple->counter = 0;
+
+
+   /* XXX ToDo: intead of iterating pixel-by-pixel, use a look-up table.
+    */
+   for (i = 0; i < length; i++) {
+      int result = stipple_test( (int) stipple->counter+i,
+                                 (ushort) stipple->pattern, stipple->factor );
+      if (result != state) {
+         /* changing from "off" to "on" or vice versa */
+	 if (state) {
+	    if (start != i) {
+               /* finishing an "on" segment */
+	       emit_segment( stage, header, start / length, i / length );
+            }
+	 }
+	 else {
+            /* starting an "on" segment */
+	    start = (float) i;
+	 }
+	 state = result;	   
+      }
+   }
+
+   if (state && start < length)
+      emit_segment( stage, header, start / length, 1.0 );
+
+   stipple->counter += length;
+}
+
+
+static void
+reset_stipple_counter(struct draw_stage *stage)
+{
+   struct stipple_stage *stipple = stipple_stage(stage);
+   stipple->counter = 0;
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+static void
+stipple_reset_point(struct draw_stage *stage, struct prim_header *header)
+{
+   struct stipple_stage *stipple = stipple_stage(stage);
+   stipple->counter = 0;
+   stage->next->point(stage->next, header);
+}
+
+static void
+stipple_reset_tri(struct draw_stage *stage, struct prim_header *header)
+{
+   struct stipple_stage *stipple = stipple_stage(stage);
+   stipple->counter = 0;
+   stage->next->tri(stage->next, header);
+}
+
+
+static void
+stipple_first_line(struct draw_stage *stage, 
+		   struct prim_header *header)
+{
+   struct stipple_stage *stipple = stipple_stage(stage);
+   struct draw_context *draw = stage->draw;
+
+   stipple->pattern = draw->rasterizer->line_stipple_pattern;
+   stipple->factor = draw->rasterizer->line_stipple_factor + 1;
+
+   stage->line = stipple_line;
+   stage->line( stage, header );
+}
+
+
+static void
+stipple_flush(struct draw_stage *stage, unsigned flags)
+{
+   stage->line = stipple_first_line;
+   stage->next->flush( stage->next, flags );
+}
+
+
+
+
+static void 
+stipple_destroy( struct draw_stage *stage )
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+/**
+ * Create line stippler stage
+ */
+struct draw_stage *draw_stipple_stage( struct draw_context *draw )
+{
+   struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage);
+
+   draw_alloc_temp_verts( &stipple->stage, 2 );
+
+   stipple->stage.draw = draw;
+   stipple->stage.next = NULL;
+   stipple->stage.point = stipple_reset_point;
+   stipple->stage.line = stipple_first_line;
+   stipple->stage.tri = stipple_reset_tri;
+   stipple->stage.reset_stipple_counter = reset_stipple_counter;
+   stipple->stage.flush = stipple_flush;
+   stipple->stage.destroy = stipple_destroy;
+
+   return &stipple->stage;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_twoside.c b/src/gallium/auxiliary/draw/draw_pipe_twoside.c
new file mode 100644
index 0000000000..c329d92339
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_twoside.c
@@ -0,0 +1,199 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw_vs.h"
+#include "draw_pipe.h"
+
+struct twoside_stage {
+   struct draw_stage stage;
+   float sign;         /**< +1 or -1 */
+   uint attrib_front0, attrib_back0;
+   uint attrib_front1, attrib_back1;
+};
+
+
+static INLINE struct twoside_stage *twoside_stage( struct draw_stage *stage )
+{
+   return (struct twoside_stage *)stage;
+}
+
+
+
+
+/**
+ * Copy back color(s) to front color(s).
+ */
+static INLINE struct vertex_header *
+copy_bfc( struct twoside_stage *twoside, 
+          const struct vertex_header *v,
+          unsigned idx )
+{   
+   struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx );
+   
+   if (twoside->attrib_back0) {
+      COPY_4FV(tmp->data[twoside->attrib_front0],
+               tmp->data[twoside->attrib_back0]);
+   }
+   if (twoside->attrib_back1) {
+      COPY_4FV(tmp->data[twoside->attrib_front1],
+               tmp->data[twoside->attrib_back1]);
+   }
+
+   return tmp;
+}
+
+
+/* Twoside tri:
+ */
+static void twoside_tri( struct draw_stage *stage,
+			 struct prim_header *header )
+{
+   struct twoside_stage *twoside = twoside_stage(stage);
+
+   if (header->det * twoside->sign < 0.0) {
+      /* this is a back-facing triangle */
+      struct prim_header tmp;
+
+      tmp.det = header->det;
+      tmp.flags = header->flags;
+      tmp.pad = header->pad;
+      /* copy back attribs to front attribs */
+      tmp.v[0] = copy_bfc(twoside, header->v[0], 0);
+      tmp.v[1] = copy_bfc(twoside, header->v[1], 1);
+      tmp.v[2] = copy_bfc(twoside, header->v[2], 2);
+
+      stage->next->tri( stage->next, &tmp );
+   }
+   else {
+      stage->next->tri( stage->next, header );
+   }
+}
+
+
+
+static void twoside_first_tri( struct draw_stage *stage, 
+			       struct prim_header *header )
+{
+   struct twoside_stage *twoside = twoside_stage(stage);
+   const struct draw_vertex_shader *vs = stage->draw->vs.vertex_shader;
+   uint i;
+
+   twoside->attrib_front0 = 0;
+   twoside->attrib_front1 = 0;
+   twoside->attrib_back0 = 0;
+   twoside->attrib_back1 = 0;
+
+   /* Find which vertex shader outputs are front/back colors */
+   for (i = 0; i < vs->info.num_outputs; i++) {
+      if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR) {
+         if (vs->info.output_semantic_index[i] == 0)
+            twoside->attrib_front0 = i;
+         else
+            twoside->attrib_front1 = i;
+      }
+      if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) {
+         if (vs->info.output_semantic_index[i] == 0)
+            twoside->attrib_back0 = i;
+         else
+            twoside->attrib_back1 = i;
+      }
+   }
+
+   if (!twoside->attrib_back0)
+      twoside->attrib_front0 = 0;
+
+   if (!twoside->attrib_back1)
+      twoside->attrib_front1 = 0;
+
+   /*
+    * We'll multiply the primitive's determinant by this sign to determine
+    * if the triangle is back-facing (negative).
+    * sign = -1 for CCW, +1 for CW
+    */
+   twoside->sign = (stage->draw->rasterizer->front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f;
+
+   stage->tri = twoside_tri;
+   stage->tri( stage, header );
+}
+
+
+static void twoside_flush( struct draw_stage *stage, unsigned flags )
+{
+   stage->tri = twoside_first_tri;
+   stage->next->flush( stage->next, flags );
+}
+
+
+static void twoside_reset_stipple_counter( struct draw_stage *stage )
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void twoside_destroy( struct draw_stage *stage )
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+/**
+ * Create twoside pipeline stage.
+ */
+struct draw_stage *draw_twoside_stage( struct draw_context *draw )
+{
+   struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage);
+   if (twoside == NULL)
+      goto fail;
+
+   if (!draw_alloc_temp_verts( &twoside->stage, 3 ))
+      goto fail;
+
+   twoside->stage.draw = draw;
+   twoside->stage.next = NULL;
+   twoside->stage.point = draw_pipe_passthrough_point;
+   twoside->stage.line = draw_pipe_passthrough_line;
+   twoside->stage.tri = twoside_first_tri;
+   twoside->stage.flush = twoside_flush;
+   twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter;
+   twoside->stage.destroy = twoside_destroy;
+
+   return &twoside->stage;
+
+ fail:
+   if (twoside)
+      twoside->stage.destroy( &twoside->stage );
+
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c
new file mode 100644
index 0000000000..68835fd1a5
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c
@@ -0,0 +1,203 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief  Drawing stage for handling glPolygonMode(line/point).
+ * Convert triangles to points or lines as needed.
+ */
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "util/u_memory.h"
+#include "pipe/p_defines.h"
+#include "draw_private.h"
+#include "draw_pipe.h"
+
+
+struct unfilled_stage {
+   struct draw_stage stage;
+
+   /** [0] = front face, [1] = back face.
+    * legal values:  PIPE_POLYGON_MODE_FILL, PIPE_POLYGON_MODE_LINE,
+    * and PIPE_POLYGON_MODE_POINT,
+    */
+   unsigned mode[2];
+};
+
+
+static INLINE struct unfilled_stage *unfilled_stage( struct draw_stage *stage )
+{
+   return (struct unfilled_stage *)stage;
+}
+
+
+
+static void point( struct draw_stage *stage,
+		   struct vertex_header *v0 )
+{
+   struct prim_header tmp;
+   tmp.v[0] = v0;
+   stage->next->point( stage->next, &tmp );
+}
+
+static void line( struct draw_stage *stage,
+		  struct vertex_header *v0,
+		  struct vertex_header *v1 )
+{
+   struct prim_header tmp;
+   tmp.v[0] = v0;
+   tmp.v[1] = v1;
+   stage->next->line( stage->next, &tmp );
+}
+
+
+static void points( struct draw_stage *stage,
+		    struct prim_header *header )
+{
+   struct vertex_header *v0 = header->v[0];
+   struct vertex_header *v1 = header->v[1];
+   struct vertex_header *v2 = header->v[2];
+
+   if ((header->flags & DRAW_PIPE_EDGE_FLAG_0) && v0->edgeflag) point( stage, v0 );
+   if ((header->flags & DRAW_PIPE_EDGE_FLAG_1) && v1->edgeflag) point( stage, v1 );
+   if ((header->flags & DRAW_PIPE_EDGE_FLAG_2) && v2->edgeflag) point( stage, v2 );
+}
+
+
+static void lines( struct draw_stage *stage,
+		   struct prim_header *header )
+{
+   struct vertex_header *v0 = header->v[0];
+   struct vertex_header *v1 = header->v[1];
+   struct vertex_header *v2 = header->v[2];
+
+   if (header->flags & DRAW_PIPE_RESET_STIPPLE)
+      stage->next->reset_stipple_counter( stage->next );
+
+   if ((header->flags & DRAW_PIPE_EDGE_FLAG_2) && v2->edgeflag) line( stage, v2, v0 );
+   if ((header->flags & DRAW_PIPE_EDGE_FLAG_0) && v0->edgeflag) line( stage, v0, v1 );
+   if ((header->flags & DRAW_PIPE_EDGE_FLAG_1) && v1->edgeflag) line( stage, v1, v2 );
+}
+
+
+/* Unfilled tri:  
+ *
+ * Note edgeflags in the vertex struct is not sufficient as we will
+ * need to manipulate them when decomposing primitives.  
+ * 
+ * We currently keep the vertex edgeflag and primitive edgeflag mask
+ * separate until the last possible moment.
+ */
+static void unfilled_tri( struct draw_stage *stage,
+			  struct prim_header *header )
+{
+   struct unfilled_stage *unfilled = unfilled_stage(stage);
+   unsigned mode = unfilled->mode[header->det >= 0.0];
+  
+   switch (mode) {
+   case PIPE_POLYGON_MODE_FILL:
+      stage->next->tri( stage->next, header );
+      break;
+   case PIPE_POLYGON_MODE_LINE:
+      lines( stage, header );
+      break;
+   case PIPE_POLYGON_MODE_POINT:
+      points( stage, header );
+      break;
+   default:
+      assert(0);
+   }   
+}
+
+
+static void unfilled_first_tri( struct draw_stage *stage, 
+				struct prim_header *header )
+{
+   struct unfilled_stage *unfilled = unfilled_stage(stage);
+
+   unfilled->mode[0] = stage->draw->rasterizer->fill_ccw; /* front */
+   unfilled->mode[1] = stage->draw->rasterizer->fill_cw;  /* back */
+
+   stage->tri = unfilled_tri;
+   stage->tri( stage, header );
+}
+
+
+
+static void unfilled_flush( struct draw_stage *stage,
+			    unsigned flags )
+{
+   stage->next->flush( stage->next, flags );
+
+   stage->tri = unfilled_first_tri;
+}
+
+
+static void unfilled_reset_stipple_counter( struct draw_stage *stage )
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void unfilled_destroy( struct draw_stage *stage )
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+/**
+ * Create unfilled triangle stage.
+ */
+struct draw_stage *draw_unfilled_stage( struct draw_context *draw )
+{
+   struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage);
+   if (unfilled == NULL)
+      goto fail;
+
+   if (!draw_alloc_temp_verts( &unfilled->stage, 0 ))
+      goto fail;
+
+   unfilled->stage.draw = draw;
+   unfilled->stage.next = NULL;
+   unfilled->stage.tmp = NULL;
+   unfilled->stage.point = draw_pipe_passthrough_point;
+   unfilled->stage.line = draw_pipe_passthrough_line;
+   unfilled->stage.tri = unfilled_first_tri;
+   unfilled->stage.flush = unfilled_flush;
+   unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter;
+   unfilled->stage.destroy = unfilled_destroy;
+
+   return &unfilled->stage;
+
+ fail:
+   if (unfilled)
+      unfilled->stage.destroy( &unfilled->stage );
+
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_util.c b/src/gallium/auxiliary/draw/draw_pipe_util.c
new file mode 100644
index 0000000000..e22e5fed0c
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_util.c
@@ -0,0 +1,137 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "util/u_memory.h"
+#include "draw/draw_private.h"
+#include "draw/draw_pipe.h"
+
+
+
+void
+draw_pipe_passthrough_point(struct draw_stage *stage, struct prim_header *header)
+{
+   stage->next->point(stage->next, header);
+}
+
+void
+draw_pipe_passthrough_line(struct draw_stage *stage, struct prim_header *header)
+{
+   stage->next->line(stage->next, header);
+}
+
+void
+draw_pipe_passthrough_tri(struct draw_stage *stage, struct prim_header *header)
+{
+   stage->next->tri(stage->next, header);
+}
+
+
+
+
+
+/* This is only used for temporary verts.
+ */
+#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float))
+
+
+/**
+ * Allocate space for temporary post-transform vertices, such as for clipping.
+ */
+boolean draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr )
+{
+   assert(!stage->tmp);
+
+   stage->tmp = NULL;
+   stage->nr_tmps = nr;
+
+   if (nr != 0)
+   {
+      unsigned i;
+      ubyte *store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr );
+
+      if (store == NULL)
+         return FALSE;
+
+      stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr );
+      if (stage->tmp == NULL) {
+         FREE(store);
+         return FALSE;
+      }
+         
+      for (i = 0; i < nr; i++)
+         stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE);
+   }
+
+   return TRUE;
+}
+
+
+void draw_free_temp_verts( struct draw_stage *stage )
+{
+   if (stage->tmp) {
+      FREE( stage->tmp[0] );
+      FREE( stage->tmp );
+      stage->tmp = NULL;
+   }
+}
+
+
+/* Reset vertex ids.  This is basically a type of flush.
+ *
+ * Called only from draw_pipe_vbuf.c
+ */
+void draw_reset_vertex_ids(struct draw_context *draw)
+{
+   struct draw_stage *stage = draw->pipeline.first;
+   
+   while (stage) {
+      unsigned i;
+
+      for (i = 0; i < stage->nr_tmps; i++)
+	 stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID;
+
+      stage = stage->next;
+   }
+
+   if (draw->pipeline.verts)
+   {
+      unsigned i;
+      char *verts = draw->pipeline.verts;
+      unsigned stride = draw->pipeline.vertex_stride;
+
+      for (i = 0; i < draw->pipeline.vertex_count; i++) {
+         ((struct vertex_header *)verts)->vertex_id = UNDEFINED_VERTEX_ID;
+         verts += stride;
+      }
+   }
+}
+
diff --git a/src/gallium/auxiliary/draw/draw_pipe_validate.c b/src/gallium/auxiliary/draw/draw_pipe_validate.c
new file mode 100644
index 0000000000..03e842ce08
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_validate.c
@@ -0,0 +1,330 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "util/u_memory.h"
+#include "pipe/p_defines.h"
+#include "draw_private.h"
+#include "draw_pipe.h"
+#include "draw_context.h"
+#include "draw_vbuf.h"
+
+static boolean points( unsigned prim )
+{
+   return (prim == PIPE_PRIM_POINTS);
+}
+
+static boolean lines( unsigned prim )
+{
+   return (prim == PIPE_PRIM_LINES ||
+           prim == PIPE_PRIM_LINE_STRIP ||
+           prim == PIPE_PRIM_LINE_LOOP);
+}
+
+static boolean triangles( unsigned prim )
+{
+   return prim >= PIPE_PRIM_TRIANGLES;
+}
+
+/**
+ * Default version of a function to check if we need any special
+ * pipeline stages, or whether prims/verts can go through untouched.
+ * Don't test for bypass clipping or vs modes, this function is just
+ * about the primitive pipeline stages.
+ *
+ * This can be overridden by the driver.
+ */
+boolean
+draw_need_pipeline(const struct draw_context *draw,
+                   const struct pipe_rasterizer_state *rasterizer,
+                   unsigned int prim )
+{
+   /* If the driver has overridden this, use that version: 
+    */
+   if (draw->render &&
+       draw->render->need_pipeline) 
+   {
+      return draw->render->need_pipeline( draw->render,
+                                          rasterizer,
+                                          prim );
+   }
+
+   /* Don't have to worry about triangles turning into lines/points
+    * and triggering the pipeline, because we have to trigger the
+    * pipeline *anyway* if unfilled mode is active.
+    */
+   if (lines(prim)) 
+   {
+      /* line stipple */
+      if (rasterizer->line_stipple_enable && draw->pipeline.line_stipple)
+         return TRUE;
+
+      /* wide lines */
+      if (rasterizer->line_width > draw->pipeline.wide_line_threshold)
+         return TRUE;
+
+      /* AA lines */
+      if (rasterizer->line_smooth && draw->pipeline.aaline)
+         return TRUE;
+   }
+
+   if (points(prim))
+   {
+      /* large points */
+      if (rasterizer->point_size > draw->pipeline.wide_point_threshold)
+         return TRUE;
+
+      /* AA points */
+      if (rasterizer->point_smooth && draw->pipeline.aapoint)
+         return TRUE;
+
+      /* point sprites */
+      if (rasterizer->point_sprite && draw->pipeline.point_sprite)
+         return TRUE;
+   }
+
+
+   if (triangles(prim)) 
+   {
+      /* polygon stipple */
+      if (rasterizer->poly_stipple_enable && draw->pipeline.pstipple)
+         return TRUE;
+
+      /* unfilled polygons */
+      if (rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL ||
+          rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL)
+         return TRUE;
+      
+      /* polygon offset */
+      if (rasterizer->offset_cw || rasterizer->offset_ccw)
+         return TRUE;
+
+      /* two-side lighting */
+      if (rasterizer->light_twoside)
+         return TRUE;
+   }
+
+   /* polygon cull - this is difficult - hardware can cull just fine
+    * most of the time (though sometimes CULL_NEITHER is unsupported.
+    * 
+    * Generally this isn't a reason to require the pipeline, though.
+    *
+   if (rasterizer->cull_mode)
+      return TRUE;
+    */
+
+   return FALSE;
+}
+
+
+
+/**
+ * Rebuild the rendering pipeline.
+ */
+static struct draw_stage *validate_pipeline( struct draw_stage *stage )
+{
+   struct draw_context *draw = stage->draw;
+   struct draw_stage *next = draw->pipeline.rasterize;
+   int need_det = 0;
+   int precalc_flat = 0;
+   boolean wide_lines, wide_points;
+
+   /* Set the validate's next stage to the rasterize stage, so that it
+    * can be found later if needed for flushing.
+    */
+   stage->next = next;
+
+   /* drawing wide lines? */
+   wide_lines = (draw->rasterizer->line_width > draw->pipeline.wide_line_threshold
+                 && !draw->rasterizer->line_smooth);
+
+   /* drawing large points? */
+   if (draw->rasterizer->point_sprite && draw->pipeline.point_sprite)
+      wide_points = TRUE;
+   else if (draw->rasterizer->point_smooth && draw->pipeline.aapoint)
+      wide_points = FALSE;
+   else if (draw->rasterizer->point_size > draw->pipeline.wide_point_threshold)
+      wide_points = TRUE;
+   else
+      wide_points = FALSE;
+
+   /*
+    * NOTE: we build up the pipeline in end-to-start order.
+    *
+    * TODO: make the current primitive part of the state and build
+    * shorter pipelines for lines & points.
+    */
+
+   if (draw->rasterizer->line_smooth && draw->pipeline.aaline) {
+      draw->pipeline.aaline->next = next;
+      next = draw->pipeline.aaline;
+   }
+
+   if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) {
+      draw->pipeline.aapoint->next = next;
+      next = draw->pipeline.aapoint;
+   }
+
+   if (wide_lines) {
+      draw->pipeline.wide_line->next = next;
+      next = draw->pipeline.wide_line;
+      precalc_flat = 1;
+   }
+
+   if (wide_points || draw->rasterizer->point_sprite) {
+      draw->pipeline.wide_point->next = next;
+      next = draw->pipeline.wide_point;
+   }
+
+   if (draw->rasterizer->line_stipple_enable && draw->pipeline.line_stipple) {
+      draw->pipeline.stipple->next = next;
+      next = draw->pipeline.stipple;
+      precalc_flat = 1;		/* only needed for lines really */
+   }
+
+   if (draw->rasterizer->poly_stipple_enable
+       && draw->pipeline.pstipple) {
+      draw->pipeline.pstipple->next = next;
+      next = draw->pipeline.pstipple;
+   }
+
+   if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL ||
+       draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) {
+      draw->pipeline.unfilled->next = next;
+      next = draw->pipeline.unfilled;
+      precalc_flat = 1;		/* only needed for triangles really */
+      need_det = 1;
+   }
+
+   if (draw->rasterizer->flatshade && precalc_flat) {
+      draw->pipeline.flatshade->next = next;
+      next = draw->pipeline.flatshade;
+   }
+	 
+   if (draw->rasterizer->offset_cw ||
+       draw->rasterizer->offset_ccw) {
+      draw->pipeline.offset->next = next;
+      next = draw->pipeline.offset;
+      need_det = 1;
+   }
+
+   if (draw->rasterizer->light_twoside) {
+      draw->pipeline.twoside->next = next;
+      next = draw->pipeline.twoside;
+      need_det = 1;
+   }
+
+   /* Always run the cull stage as we calculate determinant there
+    * also.  
+    *
+    * This can actually be a win as culling out the triangles can lead
+    * to less work emitting vertices, smaller vertex buffers, etc.
+    * It's difficult to say whether this will be true in general.
+    */
+   if (need_det || draw->rasterizer->cull_mode) {
+      draw->pipeline.cull->next = next;
+      next = draw->pipeline.cull;
+   }
+
+   /* Clip stage
+    */
+   if (!draw->bypass_clipping)
+   {
+      draw->pipeline.clip->next = next;
+      next = draw->pipeline.clip;
+   }
+
+   
+   draw->pipeline.first = next;
+   return next;
+}
+
+static void validate_tri( struct draw_stage *stage, 
+			  struct prim_header *header )
+{
+   struct draw_stage *pipeline = validate_pipeline( stage );
+   pipeline->tri( pipeline, header );
+}
+
+static void validate_line( struct draw_stage *stage, 
+			   struct prim_header *header )
+{
+   struct draw_stage *pipeline = validate_pipeline( stage );
+   pipeline->line( pipeline, header );
+}
+
+static void validate_point( struct draw_stage *stage, 
+			    struct prim_header *header )
+{
+   struct draw_stage *pipeline = validate_pipeline( stage );
+   pipeline->point( pipeline, header );
+}
+
+static void validate_reset_stipple_counter( struct draw_stage *stage )
+{
+   struct draw_stage *pipeline = validate_pipeline( stage );
+   pipeline->reset_stipple_counter( pipeline );
+}
+
+static void validate_flush( struct draw_stage *stage, 
+			    unsigned flags )
+{
+   /* May need to pass a backend flush on to the rasterize stage.
+    */
+   if (stage->next)
+      stage->next->flush( stage->next, flags );
+}
+
+
+static void validate_destroy( struct draw_stage *stage )
+{
+   FREE( stage );
+}
+
+
+/**
+ * Create validate pipeline stage.
+ */
+struct draw_stage *draw_validate_stage( struct draw_context *draw )
+{
+   struct draw_stage *stage = CALLOC_STRUCT(draw_stage);
+   if (stage == NULL)
+      return NULL;
+
+   stage->draw = draw;
+   stage->next = NULL;
+   stage->point = validate_point;
+   stage->line = validate_line;
+   stage->tri = validate_tri;
+   stage->flush = validate_flush;
+   stage->reset_stipple_counter = validate_reset_stipple_counter;
+   stage->destroy = validate_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
new file mode 100644
index 0000000000..9153bc2f86
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c
@@ -0,0 +1,497 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Vertex buffer drawing stage.
+ * 
+ * \author José Fonseca <jrfonsec@tungstengraphics.com>
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_debug.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "draw_vbuf.h"
+#include "draw_private.h"
+#include "draw_vertex.h"
+#include "draw_pipe.h"
+#include "translate/translate.h"
+#include "translate/translate_cache.h"
+
+
+/**
+ * Vertex buffer emit stage.
+ */
+struct vbuf_stage {
+   struct draw_stage stage; /**< This must be first (base class) */
+
+   struct vbuf_render *render;
+   
+   const struct vertex_info *vinfo;
+   
+   /** Vertex size in bytes */
+   unsigned vertex_size;
+
+   struct translate *translate;
+   
+   /* FIXME: we have no guarantee that 'unsigned' is 32bit */
+
+   /** Vertices in hardware format */
+   unsigned *vertices;
+   unsigned *vertex_ptr;
+   unsigned max_vertices;
+   unsigned nr_vertices;
+   
+   /** Indices */
+   ushort *indices;
+   unsigned max_indices;
+   unsigned nr_indices;
+
+   /* Cache point size somewhere it's address won't change:
+    */
+   float point_size;
+
+   struct translate_cache *cache;
+};
+
+
+/**
+ * Basically a cast wrapper.
+ */
+static INLINE struct vbuf_stage *
+vbuf_stage( struct draw_stage *stage )
+{
+   assert(stage);
+   return (struct vbuf_stage *)stage;
+}
+
+
+static void vbuf_flush_indices( struct vbuf_stage *vbuf );
+static void vbuf_flush_vertices( struct vbuf_stage *vbuf );
+static void vbuf_alloc_vertices( struct vbuf_stage *vbuf );
+
+
+static INLINE boolean 
+overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz )
+{
+   unsigned long used = (unsigned long) ((char *)ptr - (char *)map);
+   return (used + bytes) > bufsz;
+}
+
+
+static INLINE void 
+check_space( struct vbuf_stage *vbuf, unsigned nr )
+{
+   if (vbuf->nr_vertices + nr > vbuf->max_vertices ) {
+      vbuf_flush_vertices(vbuf);
+      vbuf_alloc_vertices(vbuf);
+   }
+
+   if (vbuf->nr_indices + nr > vbuf->max_indices )
+      vbuf_flush_indices(vbuf);
+}
+
+
+
+
+/**
+ * Extract the needed fields from post-transformed vertex and emit
+ * a hardware(driver) vertex.
+ * Recall that the vertices are constructed by the 'draw' module and
+ * have a couple of slots at the beginning (1-dword header, 4-dword
+ * clip pos) that we ignore here.  We only use the vertex->data[] fields.
+ */
+static INLINE ushort 
+emit_vertex( struct vbuf_stage *vbuf,
+             struct vertex_header *vertex )
+{
+   if(vertex->vertex_id == UNDEFINED_VERTEX_ID) {      
+      /* Hmm - vertices are emitted one at a time - better make sure
+       * set_buffer is efficient.  Consider a special one-shot mode for
+       * translate.
+       */
+      /* Note: we really do want data[0] here, not data[pos]: 
+       */
+      vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0);
+      vbuf->translate->run(vbuf->translate, 0, 1, vbuf->vertex_ptr);
+
+      if (0) draw_dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr);
+      
+      vbuf->vertex_ptr += vbuf->vertex_size/4;
+      vertex->vertex_id = vbuf->nr_vertices++;
+   }
+
+   return (ushort)vertex->vertex_id;
+}
+
+
+static void 
+vbuf_tri( struct draw_stage *stage,
+          struct prim_header *prim )
+{
+   struct vbuf_stage *vbuf = vbuf_stage( stage );
+   unsigned i;
+
+   check_space( vbuf, 3 );
+
+   for (i = 0; i < 3; i++) {
+      vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] );
+   }
+}
+
+
+static void 
+vbuf_line( struct draw_stage *stage, 
+           struct prim_header *prim )
+{
+   struct vbuf_stage *vbuf = vbuf_stage( stage );
+   unsigned i;
+
+   check_space( vbuf, 2 );
+
+   for (i = 0; i < 2; i++) {
+      vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] );
+   }   
+}
+
+
+static void 
+vbuf_point( struct draw_stage *stage, 
+            struct prim_header *prim )
+{
+   struct vbuf_stage *vbuf = vbuf_stage( stage );
+
+   check_space( vbuf, 1 );
+
+   vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[0] );
+}
+
+
+
+
+/**
+ * Set the prim type for subsequent vertices.
+ * This may result in a new vertex size.  The existing vbuffer (if any)
+ * will be flushed if needed and a new one allocated.
+ */
+static void
+vbuf_set_prim( struct vbuf_stage *vbuf, uint prim )
+{
+   struct translate_key hw_key;
+   unsigned dst_offset;
+   unsigned i;
+
+   vbuf->render->set_primitive(vbuf->render, prim);
+
+   /* Must do this after set_primitive() above:
+    * 
+    * XXX: need some state managment to track when this needs to be
+    * recalculated.  The driver should tell us whether there was a
+    * state change.
+    */
+   vbuf->vinfo = vbuf->render->get_vertex_info(vbuf->render);
+
+   if (vbuf->vertex_size != vbuf->vinfo->size * sizeof(float)) {
+      vbuf_flush_vertices(vbuf);
+      vbuf->vertex_size = vbuf->vinfo->size * sizeof(float);
+   }
+
+   /* Translate from pipeline vertices to hw vertices.
+    */
+   dst_offset = 0;
+
+   for (i = 0; i < vbuf->vinfo->num_attribs; i++) {
+      unsigned emit_sz = 0;
+      unsigned src_buffer = 0;
+      unsigned output_format;
+      unsigned src_offset = (vbuf->vinfo->attrib[i].src_index * 4 * sizeof(float) );
+
+      switch (vbuf->vinfo->attrib[i].emit) {
+      case EMIT_4F:
+	 output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+	 emit_sz = 4 * sizeof(float);
+	 break;
+      case EMIT_3F:
+	 output_format = PIPE_FORMAT_R32G32B32_FLOAT;
+	 emit_sz = 3 * sizeof(float);
+	 break;
+      case EMIT_2F:
+	 output_format = PIPE_FORMAT_R32G32_FLOAT;
+	 emit_sz = 2 * sizeof(float);
+	 break;
+      case EMIT_1F:
+	 output_format = PIPE_FORMAT_R32_FLOAT;
+	 emit_sz = 1 * sizeof(float);
+	 break;
+      case EMIT_1F_PSIZE:
+	 output_format = PIPE_FORMAT_R32_FLOAT;
+	 emit_sz = 1 * sizeof(float);
+	 src_buffer = 1;
+	 src_offset = 0;
+	 break;
+      case EMIT_4UB:
+	 output_format = PIPE_FORMAT_B8G8R8A8_UNORM;
+	 emit_sz = 4 * sizeof(ubyte);
+         break;
+      default:
+	 assert(0);
+	 output_format = PIPE_FORMAT_NONE;
+	 emit_sz = 0;
+	 break;
+      }
+      
+      hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+      hw_key.element[i].input_buffer = src_buffer;
+      hw_key.element[i].input_offset = src_offset;
+      hw_key.element[i].output_format = output_format;
+      hw_key.element[i].output_offset = dst_offset;
+
+      dst_offset += emit_sz;
+   }
+
+   hw_key.nr_elements = vbuf->vinfo->num_attribs;
+   hw_key.output_stride = vbuf->vinfo->size * 4;
+
+   /* Don't bother with caching at this stage:
+    */
+   if (!vbuf->translate ||
+       translate_key_compare(&vbuf->translate->key, &hw_key) != 0) 
+   {
+      translate_key_sanitize(&hw_key);
+      vbuf->translate = translate_cache_find(vbuf->cache, &hw_key);
+
+      vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0);
+   }
+
+   vbuf->point_size = vbuf->stage.draw->rasterizer->point_size;
+
+   /* Allocate new buffer?
+    */
+   if (!vbuf->vertices)
+      vbuf_alloc_vertices(vbuf);
+}
+
+
+static void 
+vbuf_first_tri( struct draw_stage *stage,
+                struct prim_header *prim )
+{
+   struct vbuf_stage *vbuf = vbuf_stage( stage );
+
+   vbuf_flush_indices( vbuf );   
+   stage->tri = vbuf_tri;
+   vbuf_set_prim(vbuf, PIPE_PRIM_TRIANGLES);
+   stage->tri( stage, prim );
+}
+
+
+static void 
+vbuf_first_line( struct draw_stage *stage,
+                 struct prim_header *prim )
+{
+   struct vbuf_stage *vbuf = vbuf_stage( stage );
+
+   vbuf_flush_indices( vbuf );
+   stage->line = vbuf_line;
+   vbuf_set_prim(vbuf, PIPE_PRIM_LINES);
+   stage->line( stage, prim );
+}
+
+
+static void 
+vbuf_first_point( struct draw_stage *stage,
+                  struct prim_header *prim )
+{
+   struct vbuf_stage *vbuf = vbuf_stage( stage );
+
+   vbuf_flush_indices( vbuf );
+   stage->point = vbuf_point;
+   vbuf_set_prim(vbuf, PIPE_PRIM_POINTS);
+   stage->point( stage, prim );
+}
+
+
+static void 
+vbuf_flush_indices( struct vbuf_stage *vbuf ) 
+{
+   if(!vbuf->nr_indices)
+      return;
+   
+   assert((uint) (vbuf->vertex_ptr - vbuf->vertices) == 
+          vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned));
+
+   vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices);
+   
+   vbuf->nr_indices = 0;
+}
+
+
+/**
+ * Flush existing vertex buffer and allocate a new one.
+ * 
+ * XXX: We separate flush-on-index-full and flush-on-vb-full, but may 
+ * raise issues uploading vertices if the hardware wants to flush when
+ * we flush.
+ */
+static void 
+vbuf_flush_vertices( struct vbuf_stage *vbuf )
+{
+   if(vbuf->vertices) {      
+      vbuf_flush_indices(vbuf);
+      
+      /* Reset temporary vertices ids */
+      if(vbuf->nr_vertices)
+	 draw_reset_vertex_ids( vbuf->stage.draw );
+      
+      /* Free the vertex buffer */
+      vbuf->render->release_vertices(vbuf->render,
+                                     vbuf->vertices,
+                                     vbuf->vertex_size,
+                                     vbuf->nr_vertices);
+      vbuf->max_vertices = vbuf->nr_vertices = 0;
+      vbuf->vertex_ptr = vbuf->vertices = NULL;
+      
+   }
+}
+   
+
+static void 
+vbuf_alloc_vertices( struct vbuf_stage *vbuf )
+{
+   assert(!vbuf->nr_indices);
+   assert(!vbuf->vertices);
+   
+   /* Allocate a new vertex buffer */
+   vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size;
+
+   /* even number */
+   vbuf->max_vertices = vbuf->max_vertices & ~1;
+
+   if(vbuf->max_vertices >= UNDEFINED_VERTEX_ID)
+      vbuf->max_vertices = UNDEFINED_VERTEX_ID - 1;
+
+   /* Must always succeed -- driver gives us a
+    * 'max_vertex_buffer_bytes' which it guarantees it can allocate,
+    * and it will flush itself if necessary to do so.  If this does
+    * fail, we are basically without usable hardware.
+    */
+   vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render,
+							     (ushort) vbuf->vertex_size,
+							     (ushort) vbuf->max_vertices);
+   vbuf->vertex_ptr = vbuf->vertices;
+}
+
+
+
+static void 
+vbuf_flush( struct draw_stage *stage, unsigned flags )
+{
+   struct vbuf_stage *vbuf = vbuf_stage( stage );
+
+   vbuf_flush_indices( vbuf );
+
+   stage->point = vbuf_first_point;
+   stage->line = vbuf_first_line;
+   stage->tri = vbuf_first_tri;
+
+   if (flags & DRAW_FLUSH_BACKEND)
+      vbuf_flush_vertices( vbuf );
+}
+
+
+static void 
+vbuf_reset_stipple_counter( struct draw_stage *stage )
+{
+   /* XXX: Need to do something here for hardware with linestipple.
+    */
+   (void) stage;
+}
+
+
+static void vbuf_destroy( struct draw_stage *stage )
+{
+   struct vbuf_stage *vbuf = vbuf_stage( stage );
+
+   if(vbuf->indices)
+      align_free( vbuf->indices );
+   
+   if (vbuf->render)
+      vbuf->render->destroy( vbuf->render );
+
+   if (vbuf->cache)
+      translate_cache_destroy(vbuf->cache);
+
+   FREE( stage );
+}
+
+
+/**
+ * Create a new primitive vbuf/render stage.
+ */
+struct draw_stage *draw_vbuf_stage( struct draw_context *draw,
+                                    struct vbuf_render *render )
+{
+   struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage);
+   if (vbuf == NULL)
+      goto fail;
+   
+   vbuf->stage.draw = draw;
+   vbuf->stage.point = vbuf_first_point;
+   vbuf->stage.line = vbuf_first_line;
+   vbuf->stage.tri = vbuf_first_tri;
+   vbuf->stage.flush = vbuf_flush;
+   vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter;
+   vbuf->stage.destroy = vbuf_destroy;
+   
+   vbuf->render = render;
+   vbuf->max_indices = MAX2(render->max_indices, UNDEFINED_VERTEX_ID-1);
+
+   vbuf->indices = (ushort *) align_malloc( vbuf->max_indices * 
+					    sizeof(vbuf->indices[0]), 
+					    16 );
+   if (!vbuf->indices)
+      goto fail;
+
+   vbuf->cache = translate_cache_create();
+   if (!vbuf->cache) 
+      goto fail;
+      
+   
+   vbuf->vertices = NULL;
+   vbuf->vertex_ptr = vbuf->vertices;
+   
+   return &vbuf->stage;
+
+ fail:
+   if (vbuf)
+      vbuf_destroy(&vbuf->stage);
+   
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
new file mode 100644
index 0000000000..184e363594
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
@@ -0,0 +1,180 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "draw_private.h"
+#include "draw_pipe.h"
+
+
+struct wideline_stage {
+   struct draw_stage stage;
+
+   float half_line_width;
+};
+
+
+
+static INLINE struct wideline_stage *wideline_stage( struct draw_stage *stage )
+{
+   return (struct wideline_stage *)stage;
+}
+
+
+
+/**
+ * Draw a wide line by drawing a quad (two triangles).
+ * XXX need to disable polygon stipple.
+ */
+static void wideline_line( struct draw_stage *stage,
+                           struct prim_header *header )
+{
+   /*const struct wideline_stage *wide = wideline_stage(stage);*/
+   const unsigned pos = stage->draw->vs.position_output;
+   const float half_width = 0.5f * stage->draw->rasterizer->line_width;
+
+   struct prim_header tri;
+
+   struct vertex_header *v0 = dup_vert(stage, header->v[0], 0);
+   struct vertex_header *v1 = dup_vert(stage, header->v[0], 1);
+   struct vertex_header *v2 = dup_vert(stage, header->v[1], 2);
+   struct vertex_header *v3 = dup_vert(stage, header->v[1], 3);
+
+   float *pos0 = v0->data[pos];
+   float *pos1 = v1->data[pos];
+   float *pos2 = v2->data[pos];
+   float *pos3 = v3->data[pos];
+
+   const float dx = fabsf(pos0[0] - pos2[0]);
+   const float dy = fabsf(pos0[1] - pos2[1]);
+
+   /* small tweak to meet GL specification */
+   const float bias = 0.125f;
+
+   /*
+    * Draw wide line as a quad (two tris) by "stretching" the line along
+    * X or Y.
+    * We need to tweak coords in several ways to be conformant here.
+    */
+
+   if (dx > dy) {
+      /* x-major line */
+      pos0[1] = pos0[1] - half_width - bias;
+      pos1[1] = pos1[1] + half_width - bias;
+      pos2[1] = pos2[1] - half_width - bias;
+      pos3[1] = pos3[1] + half_width - bias;
+      if (pos0[0] < pos2[0]) {
+         /* left to right line */
+         pos0[0] -= 0.5f;
+         pos1[0] -= 0.5f;
+         pos2[0] -= 0.5f;
+         pos3[0] -= 0.5f;
+      }
+      else {
+         /* right to left line */
+         pos0[0] += 0.5f;
+         pos1[0] += 0.5f;
+         pos2[0] += 0.5f;
+         pos3[0] += 0.5f;
+      }
+   }
+   else {
+      /* y-major line */
+      pos0[0] = pos0[0] - half_width + bias;
+      pos1[0] = pos1[0] + half_width + bias;
+      pos2[0] = pos2[0] - half_width + bias;
+      pos3[0] = pos3[0] + half_width + bias;
+      if (pos0[1] < pos2[1]) {
+         /* top to bottom line */
+         pos0[1] -= 0.5f;
+         pos1[1] -= 0.5f;
+         pos2[1] -= 0.5f;
+         pos3[1] -= 0.5f;
+      }
+      else {
+         /* bottom to top line */
+         pos0[1] += 0.5f;
+         pos1[1] += 0.5f;
+         pos2[1] += 0.5f;
+         pos3[1] += 0.5f;
+      }
+   }
+
+   tri.det = header->det;  /* only the sign matters */
+   tri.v[0] = v0;
+   tri.v[1] = v2;
+   tri.v[2] = v3;
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v0;
+   tri.v[1] = v3;
+   tri.v[2] = v1;
+   stage->next->tri( stage->next, &tri );
+}
+
+
+static void wideline_flush( struct draw_stage *stage, unsigned flags )
+{
+   stage->next->flush( stage->next, flags );
+}
+
+
+static void wideline_reset_stipple_counter( struct draw_stage *stage )
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void wideline_destroy( struct draw_stage *stage )
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+struct draw_stage *draw_wide_line_stage( struct draw_context *draw )
+{
+   struct wideline_stage *wide = CALLOC_STRUCT(wideline_stage);
+
+   draw_alloc_temp_verts( &wide->stage, 4 );
+
+   wide->stage.draw = draw;
+   wide->stage.next = NULL;
+   wide->stage.point = draw_pipe_passthrough_point;
+   wide->stage.line = wideline_line;
+   wide->stage.tri = draw_pipe_passthrough_tri;
+   wide->stage.flush = wideline_flush;
+   wide->stage.reset_stipple_counter = wideline_reset_stipple_counter;
+   wide->stage.destroy = wideline_destroy;
+
+   return &wide->stage;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
new file mode 100644
index 0000000000..e1af9e56a2
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
@@ -0,0 +1,297 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw_vs.h"
+#include "draw_pipe.h"
+
+
+struct widepoint_stage {
+   struct draw_stage stage;
+
+   float half_point_size;
+   float point_size_min;
+   float point_size_max;
+
+   float xbias;
+   float ybias;
+
+   uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS];
+   uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS];
+   uint num_texcoords;
+
+   int psize_slot;
+
+   int point_coord_fs_input;  /**< input for pointcoord (and fog) */
+};
+
+
+
+static INLINE struct widepoint_stage *
+widepoint_stage( struct draw_stage *stage )
+{
+   return (struct widepoint_stage *)stage;
+}
+
+
+
+
+/**
+ * Set the vertex texcoords for sprite mode.
+ * Coords may be left untouched or set to a right-side-up or upside-down
+ * orientation.
+ */
+static void set_texcoords(const struct widepoint_stage *wide,
+                          struct vertex_header *v, const float tc[4])
+{
+   uint i;
+   for (i = 0; i < wide->num_texcoords; i++) {
+      if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) {
+         uint j = wide->texcoord_slot[i];
+         v->data[j][0] = tc[0];
+         if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT)
+            v->data[j][1] = 1.0f - tc[1];
+         else
+            v->data[j][1] = tc[1];
+         v->data[j][2] = tc[2];
+         v->data[j][3] = tc[3];
+      }
+   }
+
+   if (wide->point_coord_fs_input >= 0) {
+      /* put gl_PointCoord into extra vertex output's zw components */
+      uint k = wide->stage.draw->extra_vp_outputs.slot;
+      v->data[k][2] = tc[0];
+      v->data[k][3] = tc[1];
+   }
+}
+
+
+/* If there are lots of sprite points (and why wouldn't there be?) it
+ * would probably be more sensible to change hardware setup to
+ * optimize this rather than doing the whole thing in software like
+ * this.
+ */
+static void widepoint_point( struct draw_stage *stage,
+                             struct prim_header *header )
+{
+   const struct widepoint_stage *wide = widepoint_stage(stage);
+   const unsigned pos = stage->draw->vs.position_output;
+   const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite;
+   float half_size;
+   float left_adj, right_adj, bot_adj, top_adj;
+
+   struct prim_header tri;
+
+   /* four dups of original vertex */
+   struct vertex_header *v0 = dup_vert(stage, header->v[0], 0);
+   struct vertex_header *v1 = dup_vert(stage, header->v[0], 1);
+   struct vertex_header *v2 = dup_vert(stage, header->v[0], 2);
+   struct vertex_header *v3 = dup_vert(stage, header->v[0], 3);
+
+   float *pos0 = v0->data[pos];
+   float *pos1 = v1->data[pos];
+   float *pos2 = v2->data[pos];
+   float *pos3 = v3->data[pos];
+
+   /* point size is either per-vertex or fixed size */
+   if (wide->psize_slot >= 0) {
+      half_size = header->v[0]->data[wide->psize_slot][0];
+
+      /* XXX: temporary -- do this in the vertex shader??
+       */
+      half_size = CLAMP(half_size,
+                        wide->point_size_min,
+                        wide->point_size_max);
+      
+      half_size *= 0.5f; 
+   }
+   else {
+      half_size = wide->half_point_size;
+   }
+
+   left_adj = -half_size + wide->xbias;
+   right_adj = half_size + wide->xbias;
+   bot_adj = half_size + wide->ybias;
+   top_adj = -half_size + wide->ybias;
+
+   pos0[0] += left_adj;
+   pos0[1] += top_adj;
+
+   pos1[0] += left_adj;
+   pos1[1] += bot_adj;
+
+   pos2[0] += right_adj;
+   pos2[1] += top_adj;
+
+   pos3[0] += right_adj;
+   pos3[1] += bot_adj;
+
+   if (sprite) {
+      static const float tex00[4] = { 0, 0, 0, 1 };
+      static const float tex01[4] = { 0, 1, 0, 1 };
+      static const float tex11[4] = { 1, 1, 0, 1 };
+      static const float tex10[4] = { 1, 0, 0, 1 };
+      set_texcoords( wide, v0, tex00 );
+      set_texcoords( wide, v1, tex01 );
+      set_texcoords( wide, v2, tex10 );
+      set_texcoords( wide, v3, tex11 );
+   }
+
+   tri.det = header->det;  /* only the sign matters */
+   tri.v[0] = v0;
+   tri.v[1] = v2;
+   tri.v[2] = v3;
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v0;
+   tri.v[1] = v3;
+   tri.v[2] = v1;
+   stage->next->tri( stage->next, &tri );
+}
+
+
+static void widepoint_first_point( struct draw_stage *stage, 
+			      struct prim_header *header )
+{
+   struct widepoint_stage *wide = widepoint_stage(stage);
+   struct draw_context *draw = stage->draw;
+
+   wide->half_point_size = 0.5f * draw->rasterizer->point_size;
+   wide->point_size_min = draw->rasterizer->point_size_min;
+   wide->point_size_max = draw->rasterizer->point_size_max;
+   wide->xbias = 0.0;
+   wide->ybias = 0.0;
+
+   if (draw->rasterizer->gl_rasterization_rules) {
+      wide->xbias = 0.125;
+   }
+
+   /* XXX we won't know the real size if it's computed by the vertex shader! */
+   if ((draw->rasterizer->point_size > draw->pipeline.wide_point_threshold) ||
+       (draw->rasterizer->point_sprite && draw->pipeline.point_sprite)) {
+      stage->point = widepoint_point;
+   }
+   else {
+      stage->point = draw_pipe_passthrough_point;
+   }
+
+   if (draw->rasterizer->point_sprite) {
+      /* find vertex shader texcoord outputs */
+      const struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+      uint i, j = 0;
+      for (i = 0; i < vs->info.num_outputs; i++) {
+         if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) {
+            wide->texcoord_slot[j] = i;
+            wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j];
+            j++;
+         }
+      }
+      wide->num_texcoords = j;
+
+      /* find fragment shader PointCoord/Fog input */
+      wide->point_coord_fs_input = 0; /* XXX fix this! */
+
+      /* setup extra vp output (point coord implemented as a texcoord) */
+      draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
+      draw->extra_vp_outputs.semantic_index = 0;
+      draw->extra_vp_outputs.slot = draw->vs.num_vs_outputs;
+   }
+   else {
+      wide->point_coord_fs_input = -1;
+      draw->extra_vp_outputs.slot = 0;
+   }
+
+   wide->psize_slot = -1;
+   if (draw->rasterizer->point_size_per_vertex) {
+      /* find PSIZ vertex output */
+      const struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+      uint i;
+      for (i = 0; i < vs->info.num_outputs; i++) {
+         if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) {
+            wide->psize_slot = i;
+            break;
+         }
+      }
+   }
+   
+   stage->point( stage, header );
+}
+
+
+static void widepoint_flush( struct draw_stage *stage, unsigned flags )
+{
+   stage->point = widepoint_first_point;
+   stage->next->flush( stage->next, flags );
+}
+
+
+static void widepoint_reset_stipple_counter( struct draw_stage *stage )
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void widepoint_destroy( struct draw_stage *stage )
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+struct draw_stage *draw_wide_point_stage( struct draw_context *draw )
+{
+   struct widepoint_stage *wide = CALLOC_STRUCT(widepoint_stage);
+   if (wide == NULL)
+      goto fail;
+
+   if (!draw_alloc_temp_verts( &wide->stage, 4 ))
+      goto fail;
+
+   wide->stage.draw = draw;
+   wide->stage.next = NULL;
+   wide->stage.point = widepoint_first_point;
+   wide->stage.line = draw_pipe_passthrough_line;
+   wide->stage.tri = draw_pipe_passthrough_tri;
+   wide->stage.flush = widepoint_flush;
+   wide->stage.reset_stipple_counter = widepoint_reset_stipple_counter;
+   wide->stage.destroy = widepoint_destroy;
+
+   return &wide->stage;
+
+ fail:
+   if (wide)
+      wide->stage.destroy( &wide->stage );
+   
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
new file mode 100644
index 0000000000..81e4eae401
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -0,0 +1,318 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Private data structures, etc for the draw module.
+ */
+
+
+/**
+ * Authors:
+ * Keith Whitwell <keith@tungstengraphics.com>
+ * Brian Paul
+ */
+
+
+#ifndef DRAW_PRIVATE_H
+#define DRAW_PRIVATE_H
+
+
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+
+#include "tgsi/tgsi_exec.h"
+#include "tgsi/tgsi_scan.h"
+
+
+struct pipe_context;
+struct gallivm_prog;
+struct gallivm_cpu_engine;
+struct draw_vertex_shader;
+struct draw_context;
+struct draw_stage;
+struct vbuf_render;
+
+
+/**
+ * Basic vertex info.
+ * Carry some useful information around with the vertices in the prim pipe.  
+ */
+struct vertex_header {
+   unsigned clipmask:12;
+   unsigned edgeflag:1;
+   unsigned pad:3;
+   unsigned vertex_id:16;
+
+   float clip[4];
+
+   /* This will probably become float (*data)[4] soon:
+    */
+   float data[][4];
+};
+
+/* NOTE: It should match vertex_id size above */
+#define UNDEFINED_VERTEX_ID 0xffff
+
+
+/**
+ * Private context for the drawing module.
+ */
+struct draw_context
+{
+   /** Drawing/primitive pipeline stages */
+   struct {
+      struct draw_stage *first;  /**< one of the following */
+
+      struct draw_stage *validate; 
+
+      /* stages (in logical order) */
+      struct draw_stage *flatshade;
+      struct draw_stage *clip;
+      struct draw_stage *cull;
+      struct draw_stage *twoside;
+      struct draw_stage *offset;
+      struct draw_stage *unfilled;
+      struct draw_stage *stipple;
+      struct draw_stage *aapoint;
+      struct draw_stage *aaline;
+      struct draw_stage *pstipple;
+      struct draw_stage *wide_line;
+      struct draw_stage *wide_point;
+      struct draw_stage *rasterize;
+
+      float wide_point_threshold; /**< convert pnts to tris if larger than this */
+      float wide_line_threshold;  /**< convert lines to tris if wider than this */
+      boolean line_stipple;       /**< do line stipple? */
+      boolean point_sprite;       /**< convert points to quads for sprites? */
+
+      /* Temporary storage while the pipeline is being run:
+       */
+      char *verts;
+      unsigned vertex_stride;
+      unsigned vertex_count;
+   } pipeline;
+
+
+   struct vbuf_render *render;
+
+   /* Support prototype passthrough path:
+    */
+   struct {
+      struct {
+         struct draw_pt_middle_end *fetch_emit;
+         struct draw_pt_middle_end *fetch_shade_emit;
+         struct draw_pt_middle_end *general;
+      } middle;
+
+      struct {
+         struct draw_pt_front_end *vcache;
+         struct draw_pt_front_end *varray;
+      } front;
+
+      struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+      unsigned nr_vertex_buffers;
+
+      struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
+      unsigned nr_vertex_elements;
+
+      /* user-space vertex data, buffers */
+      struct {
+         const unsigned *edgeflag;
+
+         /** vertex element/index buffer (ex: glDrawElements) */
+         const void *elts;
+         /** bytes per index (0, 1, 2 or 4) */
+         unsigned eltSize;
+         unsigned min_index;
+         unsigned max_index;
+         
+         /** vertex arrays */
+         const void *vbuffer[PIPE_MAX_ATTRIBS];
+         
+         /** constant buffer (for vertex shader) */
+         const void *constants;
+      } user;
+
+      boolean test_fse;         /* enable FSE even though its not correct (eg for softpipe) */
+      boolean no_fse;           /* disable FSE even when it is correct */
+   } pt;
+
+   struct {
+      boolean bypass_clipping;
+      boolean bypass_vs;
+   } driver;
+
+   boolean flushing;         /**< debugging/sanity */
+   boolean suspend_flushing; /**< internally set */
+   boolean bypass_clipping;  /**< set if either api or driver bypass_clipping true */
+
+   boolean force_passthrough; /**< never clip or shade */
+
+   double mrd;  /**< minimum resolvable depth value, for polygon offset */
+
+   /* pipe state that we need: */
+   const struct pipe_rasterizer_state *rasterizer;
+   struct pipe_viewport_state viewport;
+   boolean identity_viewport;
+
+   struct {
+      struct draw_vertex_shader *vertex_shader;
+      uint num_vs_outputs;  /**< convenience, from vertex_shader */
+      uint position_output;
+
+      /** TGSI program interpreter runtime state */
+      struct tgsi_exec_machine machine;
+
+      uint num_samplers;
+      struct tgsi_sampler **samplers;
+
+      /* This (and the tgsi_exec_machine struct) probably need to be moved somewhere private.
+       */
+      struct gallivm_cpu_engine *engine;   
+
+      /* Here's another one:
+       */
+      struct aos_machine *aos_machine; 
+
+
+      const float (*aligned_constants)[4];
+
+      const float (*aligned_constant_storage)[4];
+      unsigned const_storage_size;
+
+
+      struct translate *fetch;
+      struct translate_cache *fetch_cache;
+      struct translate *emit;
+      struct translate_cache *emit_cache;
+   } vs;
+
+   /* Clip derived state:
+    */
+   float plane[12][4];
+   unsigned nr_planes;
+
+   /* If a prim stage introduces new vertex attributes, they'll be stored here
+    */
+   struct {
+      uint semantic_name;
+      uint semantic_index;
+      int slot;
+   } extra_vp_outputs;
+
+   unsigned reduced_prim;
+
+   void *driver_private;
+};
+
+
+/*******************************************************************************
+ * Vertex shader code:
+ */
+boolean draw_vs_init( struct draw_context *draw );
+void draw_vs_destroy( struct draw_context *draw );
+
+void draw_vs_set_viewport( struct draw_context *, 
+                           const struct pipe_viewport_state * );
+
+void draw_vs_set_constants( struct draw_context *,
+                            const float (*constants)[4],
+                            unsigned size );
+
+
+
+
+/*******************************************************************************
+ * Vertex processing (was passthrough) code:
+ */
+boolean draw_pt_init( struct draw_context *draw );
+void draw_pt_destroy( struct draw_context *draw );
+void draw_pt_reset_vertex_ids( struct draw_context *draw );
+
+
+/*******************************************************************************
+ * Primitive processing (pipeline) code: 
+ */
+
+boolean draw_pipeline_init( struct draw_context *draw );
+void draw_pipeline_destroy( struct draw_context *draw );
+
+
+
+
+
+/* We use the top few bits in the elts[] parameter to convey a little
+ * API information.  This limits the number of vertices we can address
+ * to only 4096 -- if that becomes a problem, we can switch to 32-bit
+ * draw indices.
+ *
+ * These flags expected at first vertex of lines & triangles when
+ * unfilled and/or line stipple modes are operational.
+ */
+#define DRAW_PIPE_MAX_VERTICES  (0x1<<12)
+#define DRAW_PIPE_EDGE_FLAG_0   (0x1<<12)
+#define DRAW_PIPE_EDGE_FLAG_1   (0x2<<12)
+#define DRAW_PIPE_EDGE_FLAG_2   (0x4<<12)
+#define DRAW_PIPE_EDGE_FLAG_ALL (0x7<<12)
+#define DRAW_PIPE_RESET_STIPPLE (0x8<<12)
+#define DRAW_PIPE_FLAG_MASK     (0xf<<12)
+
+void draw_pipeline_run( struct draw_context *draw,
+                        unsigned prim,
+                        struct vertex_header *vertices,
+                        unsigned vertex_count,
+                        unsigned stride,
+                        const ushort *elts,
+                        unsigned count );
+
+void draw_pipeline_run_linear( struct draw_context *draw,
+                               unsigned prim,
+                               struct vertex_header *vertices,
+                               unsigned count,
+                               unsigned stride );
+
+
+
+void draw_pipeline_flush( struct draw_context *draw, 
+                          unsigned flags );
+
+
+
+/*******************************************************************************
+ * Flushing 
+ */
+
+#define DRAW_FLUSH_STATE_CHANGE              0x8
+#define DRAW_FLUSH_BACKEND                   0x10
+
+
+void draw_do_flush( struct draw_context *draw, unsigned flags );
+
+
+
+
+#endif /* DRAW_PRIVATE_H */
diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c
new file mode 100644
index 0000000000..4e5ffa0930
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt.c
@@ -0,0 +1,323 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_pt.h"
+#include "draw/draw_vs.h"
+#include "tgsi/tgsi_dump.h"
+#include "util/u_math.h"
+
+static unsigned trim( unsigned count, unsigned first, unsigned incr )
+{
+   if (count < first)
+      return 0;
+   return count - (count - first) % incr; 
+}
+
+
+
+/* Overall we split things into:
+ *     - frontend -- prepare fetch_elts, draw_elts - eg vcache
+ *     - middle   -- fetch, shade, cliptest, viewport
+ *     - pipeline -- the prim pipeline: clipping, wide lines, etc 
+ *     - backend  -- the vbuf_render provided by the driver.
+ */
+static boolean
+draw_pt_arrays(struct draw_context *draw, 
+               unsigned prim,
+               unsigned start, 
+               unsigned count)
+{
+   struct draw_pt_front_end *frontend = NULL;
+   struct draw_pt_middle_end *middle = NULL;
+   unsigned opt = 0;
+
+   /* Sanitize primitive length:
+    */
+   {
+      unsigned first, incr;
+      draw_pt_split_prim(prim, &first, &incr);
+      count = trim(count, first, incr); 
+      if (count < first)
+         return TRUE;
+   }
+
+   if (!draw->force_passthrough) {
+      if (!draw->render) {
+         opt |= PT_PIPELINE;
+      }
+      
+      if (draw_need_pipeline(draw,
+                             draw->rasterizer,
+                             prim)) {
+         opt |= PT_PIPELINE;
+      }
+
+      if (!draw->bypass_clipping && !draw->pt.test_fse) {
+         opt |= PT_CLIPTEST;
+      }
+      
+      if (!draw->rasterizer->bypass_vs) {
+         opt |= PT_SHADE;
+      }
+   }
+      
+   if (opt == 0) 
+      middle = draw->pt.middle.fetch_emit;
+   else if (opt == PT_SHADE && !draw->pt.no_fse)
+      middle = draw->pt.middle.fetch_shade_emit;
+   else
+      middle = draw->pt.middle.general;
+
+
+   /* Pick the right frontend
+    */
+   if (draw->pt.user.elts || (opt & PT_PIPELINE)) {
+      frontend = draw->pt.front.vcache;
+   } else {
+      frontend = draw->pt.front.varray;
+   }
+
+   frontend->prepare( frontend, prim, middle, opt );
+
+   frontend->run(frontend, 
+                 draw_pt_elt_func(draw),
+                 draw_pt_elt_ptr(draw, start),
+                 count);
+
+   frontend->finish( frontend );
+
+   return TRUE;
+}
+
+
+boolean draw_pt_init( struct draw_context *draw )
+{
+   draw->pt.test_fse = debug_get_bool_option("DRAW_FSE", FALSE);
+   draw->pt.no_fse = debug_get_bool_option("DRAW_NO_FSE", FALSE);
+
+   draw->pt.front.vcache = draw_pt_vcache( draw );
+   if (!draw->pt.front.vcache)
+      return FALSE;
+
+   draw->pt.front.varray = draw_pt_varray(draw);
+   if (!draw->pt.front.varray)
+      return FALSE;
+
+   draw->pt.middle.fetch_emit = draw_pt_fetch_emit( draw );
+   if (!draw->pt.middle.fetch_emit)
+      return FALSE;
+
+   draw->pt.middle.fetch_shade_emit = draw_pt_middle_fse( draw );
+   if (!draw->pt.middle.fetch_shade_emit)
+      return FALSE;
+
+   draw->pt.middle.general = draw_pt_fetch_pipeline_or_emit( draw );
+   if (!draw->pt.middle.general)
+      return FALSE;
+
+   return TRUE;
+}
+
+
+void draw_pt_destroy( struct draw_context *draw )
+{
+   if (draw->pt.middle.general) {
+      draw->pt.middle.general->destroy( draw->pt.middle.general );
+      draw->pt.middle.general = NULL;
+   }
+
+   if (draw->pt.middle.fetch_emit) {
+      draw->pt.middle.fetch_emit->destroy( draw->pt.middle.fetch_emit );
+      draw->pt.middle.fetch_emit = NULL;
+   }
+
+   if (draw->pt.middle.fetch_shade_emit) {
+      draw->pt.middle.fetch_shade_emit->destroy( draw->pt.middle.fetch_shade_emit );
+      draw->pt.middle.fetch_shade_emit = NULL;
+   }
+
+   if (draw->pt.front.vcache) {
+      draw->pt.front.vcache->destroy( draw->pt.front.vcache );
+      draw->pt.front.vcache = NULL;
+   }
+
+   if (draw->pt.front.varray) {
+      draw->pt.front.varray->destroy( draw->pt.front.varray );
+      draw->pt.front.varray = NULL;
+   }
+}
+
+
+/**
+ * Debug- print the first 'count' vertices.
+ */
+static void
+draw_print_arrays(struct draw_context *draw, uint prim, int start, uint count)
+{
+   uint i;
+
+   debug_printf("Draw arrays(prim = %u, start = %u, count = %u)\n",
+                prim, start, count);
+
+   for (i = 0; i < count; i++) {
+      uint ii, j;
+
+      if (draw->pt.user.elts) {
+         /* indexed arrays */
+         switch (draw->pt.user.eltSize) {
+         case 1:
+            {
+               const ubyte *elem = (const ubyte *) draw->pt.user.elts;
+               ii = elem[start + i];
+            }
+            break;
+         case 2:
+            {
+               const ushort *elem = (const ushort *) draw->pt.user.elts;
+               ii = elem[start + i];
+            }
+            break;
+         case 4:
+            {
+               const uint *elem = (const uint *) draw->pt.user.elts;
+               ii = elem[start + i];
+            }
+            break;
+         default:
+            assert(0);
+         }
+         debug_printf("Element[%u + %u] -> Vertex %u:\n", start, i, ii);
+      }
+      else {
+         /* non-indexed arrays */
+         ii = start + i;
+         debug_printf("Vertex %u:\n", ii);
+      }
+
+      for (j = 0; j < draw->pt.nr_vertex_elements; j++) {
+         uint buf = draw->pt.vertex_element[j].vertex_buffer_index;
+         ubyte *ptr = (ubyte *) draw->pt.user.vbuffer[buf];
+         ptr += draw->pt.vertex_buffer[buf].stride * ii;
+         ptr += draw->pt.vertex_element[j].src_offset;
+
+         debug_printf("  Attr %u: ", j);
+         switch (draw->pt.vertex_element[j].src_format) {
+         case PIPE_FORMAT_R32_FLOAT:
+            {
+               float *v = (float *) ptr;
+               debug_printf("%f  @ %p\n", v[0], (void *) v);
+            }
+            break;
+         case PIPE_FORMAT_R32G32_FLOAT:
+            {
+               float *v = (float *) ptr;
+               debug_printf("%f %f  @ %p\n", v[0], v[1], (void *) v);
+            }
+            break;
+         case PIPE_FORMAT_R32G32B32_FLOAT:
+            {
+               float *v = (float *) ptr;
+               debug_printf("%f %f %f  @ %p\n", v[0], v[1], v[2], (void *) v);
+            }
+            break;
+         case PIPE_FORMAT_R32G32B32A32_FLOAT:
+            {
+               float *v = (float *) ptr;
+               debug_printf("%f %f %f %f  @ %p\n", v[0], v[1], v[2], v[3],
+                            (void *) v);
+            }
+            break;
+         default:
+            debug_printf("other format (fix me)\n");
+            ;
+         }
+      }
+   }
+}
+
+
+/**
+ * Draw vertex arrays
+ * This is the main entrypoint into the drawing module.
+ * \param prim  one of PIPE_PRIM_x
+ * \param start  index of first vertex to draw
+ * \param count  number of vertices to draw
+ */
+void
+draw_arrays(struct draw_context *draw, unsigned prim,
+            unsigned start, unsigned count)
+{
+   unsigned reduced_prim = draw_pt_reduced_prim(prim);
+   if (reduced_prim != draw->reduced_prim) {
+      draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+      draw->reduced_prim = reduced_prim;
+   }
+
+   if (0)
+      draw_print_arrays(draw, prim, start, MIN2(count, 20));
+
+#if 0
+   {
+      int i;
+      debug_printf("draw_arrays(prim=%u start=%u count=%u):\n",
+                   prim, start, count);
+      tgsi_dump(draw->vs.vertex_shader->state.tokens, 0);
+      debug_printf("Elements:\n");
+      for (i = 0; i < draw->pt.nr_vertex_elements; i++) {
+         debug_printf("  format=%s comps=%u\n",
+                      pf_name(draw->pt.vertex_element[i].src_format),
+                      draw->pt.vertex_element[i].nr_components);
+      }
+      debug_printf("Buffers:\n");
+      for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
+         debug_printf("  stride=%u offset=%u ptr=%p\n",
+                      draw->pt.vertex_buffer[i].stride,
+                      draw->pt.vertex_buffer[i].buffer_offset,
+                      draw->pt.user.vbuffer[i]);
+      }
+   }
+#endif
+
+   /* drawing done here: */
+   draw_pt_arrays(draw, prim, start, count);
+}
+
+boolean draw_pt_get_edgeflag( struct draw_context *draw,
+                              unsigned idx )
+{
+   if (draw->pt.user.edgeflag)
+      return (draw->pt.user.edgeflag[idx/32] & (1 << (idx%32))) != 0;
+   else
+      return 1;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h
new file mode 100644
index 0000000000..c02f229110
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt.h
@@ -0,0 +1,235 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#ifndef DRAW_PT_H
+#define DRAW_PT_H
+
+#include "pipe/p_compiler.h"
+
+typedef unsigned (*pt_elt_func)( const void *elts, unsigned idx );
+
+struct draw_pt_middle_end;
+struct draw_context;
+
+
+#define PT_SHADE      0x1
+#define PT_CLIPTEST   0x2
+#define PT_PIPELINE   0x4
+#define PT_MAX_MIDDLE 0x8
+
+
+/* The "front end" - prepare sets of fetch, draw elements for the
+ * middle end.
+ *
+ * Currenly one version of this:
+ *    - vcache - catchall implementation, decomposes to TRI/LINE/POINT prims
+ * Later:
+ *    - varray, varray_split
+ *    - velement, velement_split
+ *
+ * Currenly only using the vcache version.
+ */
+struct draw_pt_front_end {
+   void (*prepare)( struct draw_pt_front_end *,
+                    unsigned prim,
+                    struct draw_pt_middle_end *,
+		    unsigned opt );
+
+   void (*run)( struct draw_pt_front_end *,
+                pt_elt_func elt_func,
+                const void *elt_ptr,
+                unsigned count );
+
+   void (*finish)( struct draw_pt_front_end * );
+   void (*destroy)( struct draw_pt_front_end * );
+};
+
+
+/* The "middle end" - prepares actual hardware vertices for the
+ * hardware backend.
+ *
+ * Currently two versions of this:
+ *     - fetch, vertex shade, cliptest, prim-pipeline
+ *     - fetch, emit (ie passthrough)
+ */
+struct draw_pt_middle_end {
+   void (*prepare)( struct draw_pt_middle_end *,
+                    unsigned prim,
+		    unsigned opt,
+                    unsigned *max_vertices );
+
+   void (*run)( struct draw_pt_middle_end *,
+                const unsigned *fetch_elts,
+                unsigned fetch_count,
+                const ushort *draw_elts,
+                unsigned draw_count );
+
+   void (*run_linear)(struct draw_pt_middle_end *,
+                      unsigned start,
+                      unsigned count);
+
+   /* Transform all vertices in a linear range and then draw them with
+    * the supplied element list.  May fail and return FALSE.
+    */
+   boolean (*run_linear_elts)( struct draw_pt_middle_end *,
+                            unsigned fetch_start,
+                            unsigned fetch_count,
+                            const ushort *draw_elts,
+                            unsigned draw_count );
+
+   int (*get_max_vertex_count)( struct draw_pt_middle_end * );
+
+   void (*finish)( struct draw_pt_middle_end * );
+   void (*destroy)( struct draw_pt_middle_end * );
+};
+
+
+/* The "back end" - supplied by the driver, defined in draw_vbuf.h.
+ */
+struct vbuf_render;
+struct vertex_header;
+
+
+/* Helper functions.
+ */
+pt_elt_func draw_pt_elt_func( struct draw_context *draw );
+const void *draw_pt_elt_ptr( struct draw_context *draw,
+                             unsigned start );
+
+/* Frontends: 
+ *
+ * Currently only the general-purpose vcache implementation, could add
+ * a special case for tiny vertex buffers.
+ */
+struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw );
+struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw);
+
+
+/* Middle-ends:
+ *
+ * Currently one general-purpose case which can do all possibilities,
+ * at the slight expense of creating a vertex_header in some cases
+ * unecessarily.
+ *
+ * The special case fetch_emit code avoids pipeline vertices
+ * altogether and builds hardware vertices directly from API
+ * vertex_elements.
+ */
+struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw );
+struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw );
+struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *draw);
+
+
+/* More helpers:
+ */
+boolean draw_pt_get_edgeflag( struct draw_context *draw,
+                              unsigned idx );
+
+
+/*******************************************************************************
+ * HW vertex emit:
+ */
+struct pt_emit;
+
+void draw_pt_emit_prepare( struct pt_emit *emit,
+			   unsigned prim,
+                           unsigned *max_vertices );
+
+void draw_pt_emit( struct pt_emit *emit,
+		   const float (*vertex_data)[4],
+		   unsigned vertex_count,
+		   unsigned stride,
+		   const ushort *elts,
+		   unsigned count );
+
+void draw_pt_emit_linear( struct pt_emit *emit,
+                          const float (*vertex_data)[4],
+                          unsigned vertex_count,
+                          unsigned stride,
+                          unsigned start,
+                          unsigned count );
+
+void draw_pt_emit_destroy( struct pt_emit *emit );
+
+struct pt_emit *draw_pt_emit_create( struct draw_context *draw );
+
+
+/*******************************************************************************
+ * API vertex fetch:
+ */
+
+struct pt_fetch;
+void draw_pt_fetch_prepare( struct pt_fetch *fetch,
+			    unsigned vertex_size );
+
+void draw_pt_fetch_run( struct pt_fetch *fetch,
+			const unsigned *elts,
+			unsigned count,
+			char *verts );
+
+void draw_pt_fetch_run_linear( struct pt_fetch *fetch,
+                               unsigned start,
+                               unsigned count,
+                               char *verts );
+
+void draw_pt_fetch_destroy( struct pt_fetch *fetch );
+
+struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw );
+
+/*******************************************************************************
+ * Post-VS: cliptest, rhw, viewport
+ */
+struct pt_post_vs;
+
+boolean draw_pt_post_vs_run( struct pt_post_vs *pvs,
+			     struct vertex_header *pipeline_verts,
+			     unsigned stride,
+			     unsigned count );
+
+void draw_pt_post_vs_prepare( struct pt_post_vs *pvs,
+			      boolean bypass_clipping,
+			      boolean identity_viewport,
+			      boolean opengl );
+
+struct pt_post_vs *draw_pt_post_vs_create( struct draw_context *draw );
+
+void draw_pt_post_vs_destroy( struct pt_post_vs *pvs );
+
+
+/*******************************************************************************
+ * Utils: 
+ */
+void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr);
+unsigned draw_pt_reduced_prim(unsigned prim);
+
+
+#endif
diff --git a/src/gallium/auxiliary/draw/draw_pt_decompose.h b/src/gallium/auxiliary/draw/draw_pt_decompose.h
new file mode 100644
index 0000000000..3fb0695687
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_decompose.h
@@ -0,0 +1,153 @@
+
+
+static void FUNC( ARGS,
+                  unsigned count )
+{
+   LOCAL_VARS;
+
+   switch (prim) {
+   case PIPE_PRIM_POINTS:
+      for (i = 0; i < count; i ++) {
+	 POINT( (i + 0) );
+      }
+      break;
+
+   case PIPE_PRIM_LINES:
+      for (i = 0; i+1 < count; i += 2) {
+         LINE( DRAW_PIPE_RESET_STIPPLE,
+               (i + 0),
+               (i + 1));
+      }
+      break;
+
+   case PIPE_PRIM_LINE_LOOP:
+      if (count >= 2) {
+         flags = DRAW_PIPE_RESET_STIPPLE;
+
+         for (i = 1; i < count; i++, flags = 0) {
+            LINE( flags,
+                  (i - 1),
+                  (i ));
+         }
+
+	 LINE( flags,
+               (i - 1),
+               (0 ));
+      }
+      break;
+
+   case PIPE_PRIM_LINE_STRIP:
+      flags = DRAW_PIPE_RESET_STIPPLE;
+      for (i = 1; i < count; i++, flags = 0) {
+         LINE( flags,
+               (i - 1),
+               (i ));
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLES:
+      for (i = 0; i+2 < count; i += 3) {
+         TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                   (i + 0),
+                   (i + 1),
+                   (i + 2 ));
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      if (flatfirst) {
+         for (i = 0; i+2 < count; i++) {
+            TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                      (i + 0),
+                      (i + 1 + (i&1)),
+                      (i + 2 - (i&1)));
+         }
+      }
+      else {
+         for (i = 0; i+2 < count; i++) {
+            TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                      (i + 0 + (i&1)),
+                      (i + 1 - (i&1)),
+                      (i + 2 ));
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_FAN:
+      if (count >= 3) {
+         if (flatfirst) {
+            for (i = 0; i+2 < count; i++) {
+               TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                         (i + 1),
+                         (i + 2),
+                         (0 ));
+            }
+         }
+         else {
+            for (i = 0; i+2 < count; i++) {
+               TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                         (0),
+                         (i + 1),
+                         (i + 2 ));
+            }
+         }
+      }
+      break;
+
+
+   case PIPE_PRIM_QUADS:
+      for (i = 0; i+3 < count; i += 4) {
+         QUAD( (i + 0),
+               (i + 1),
+               (i + 2),
+               (i + 3));
+      }
+      break;
+
+   case PIPE_PRIM_QUAD_STRIP:
+      for (i = 0; i+3 < count; i += 2) {
+         QUAD( (i + 2),
+               (i + 0),
+               (i + 1),
+               (i + 3));
+      }
+      break;
+
+   case PIPE_PRIM_POLYGON:
+      {
+         /* These bitflags look a little odd because we submit the
+          * vertices as (1,2,0) to satisfy flatshade requirements.
+          */
+         const ushort edge_first  = DRAW_PIPE_EDGE_FLAG_2;
+         const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0;
+         const ushort edge_last   = DRAW_PIPE_EDGE_FLAG_1;
+
+         flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle;
+
+	 for (i = 0; i+2 < count; i++, flags = edge_middle) {
+
+            if (i + 3 == count)
+               flags |= edge_last;
+
+	    TRIANGLE( flags,
+                      (i + 1),
+                      (i + 2),
+                      (0));
+	 }
+      }
+      break;
+
+   default:
+      assert(0);
+      break;
+   }
+
+   FLUSH;
+}
+
+
+#undef TRIANGLE
+#undef QUAD
+#undef POINT
+#undef LINE
+#undef FUNC
diff --git a/src/gallium/auxiliary/draw/draw_pt_elts.c b/src/gallium/auxiliary/draw/draw_pt_elts.c
new file mode 100644
index 0000000000..b7780fb507
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_elts.c
@@ -0,0 +1,88 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "draw/draw_pt.h"
+#include "draw/draw_private.h"
+
+/* Neat get_elt func that also works for varrays drawing by encoding
+ * the start value into a pointer.  
+ */
+
+static unsigned elt_uint( const void *elts, unsigned idx )
+{
+   return *(((const uint *)elts) + idx);
+}
+
+static unsigned elt_ushort( const void *elts, unsigned idx )
+{
+   return *(((const ushort *)elts) + idx);
+}
+
+static unsigned elt_ubyte( const void *elts, unsigned idx )
+{
+   return *(((const ubyte *)elts) + idx);
+}
+
+static unsigned elt_vert( const void *elts, unsigned idx )
+{
+   return (const ubyte *)elts - (const ubyte *)NULL + idx;
+}
+
+pt_elt_func draw_pt_elt_func( struct draw_context *draw )
+{
+   switch (draw->pt.user.eltSize) {
+   case 0: return &elt_vert;
+   case 1: return &elt_ubyte;
+   case 2: return &elt_ushort; 
+   case 4: return &elt_uint;
+   default: return NULL;
+   }
+}     
+
+const void *draw_pt_elt_ptr( struct draw_context *draw,
+                             unsigned start )
+{
+   const char *elts = draw->pt.user.elts;
+
+   switch (draw->pt.user.eltSize) {
+   case 0: 
+      return (const void *)(((const ubyte *)NULL) + start);
+   case 1: 
+      return (const void *)(((const ubyte *)elts) + start);
+   case 2: 
+      return (const void *)(((const ushort *)elts) + start);
+   case 4: 
+      return (const void *)(((const uint *)elts) + start);
+   default:
+      return NULL;
+   }
+}
diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c
new file mode 100644
index 0000000000..232dfdaed2
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_emit.c
@@ -0,0 +1,308 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_memory.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vbuf.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_pt.h"
+#include "translate/translate.h"
+#include "translate/translate_cache.h"
+
+struct pt_emit {
+   struct draw_context *draw;
+
+   struct translate *translate;
+
+   struct translate_cache *cache;
+   unsigned prim;
+
+   const struct vertex_info *vinfo;
+};
+
+void draw_pt_emit_prepare( struct pt_emit *emit,
+			   unsigned prim,
+                           unsigned *max_vertices )
+{
+   struct draw_context *draw = emit->draw;
+   const struct vertex_info *vinfo;
+   unsigned dst_offset;
+   struct translate_key hw_key;
+   unsigned i;
+   boolean ok;
+   
+   /* XXX: need to flush to get prim_vbuf.c to release its allocation?? 
+    */
+   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+
+
+   /* XXX: may need to defensively reset this later on as clipping can
+    * clobber this state in the render backend.
+    */
+   emit->prim = prim;
+
+   ok = draw->render->set_primitive(draw->render, emit->prim);
+   if (!ok) {
+      assert(0);
+      return;
+   }
+
+   /* Must do this after set_primitive() above:
+    */
+   emit->vinfo = vinfo = draw->render->get_vertex_info(draw->render);
+
+
+   /* Translate from pipeline vertices to hw vertices.
+    */
+   dst_offset = 0;
+   for (i = 0; i < vinfo->num_attribs; i++) {
+      unsigned emit_sz = 0;
+      unsigned src_buffer = 0;
+      unsigned output_format;
+      unsigned src_offset = (vinfo->attrib[i].src_index * 4 * sizeof(float) );
+
+
+         
+      switch (vinfo->attrib[i].emit) {
+      case EMIT_4F:
+	 output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+	 emit_sz = 4 * sizeof(float);
+	 break;
+      case EMIT_3F:
+	 output_format = PIPE_FORMAT_R32G32B32_FLOAT;
+	 emit_sz = 3 * sizeof(float);
+	 break;
+      case EMIT_2F:
+	 output_format = PIPE_FORMAT_R32G32_FLOAT;
+	 emit_sz = 2 * sizeof(float);
+	 break;
+      case EMIT_1F:
+	 output_format = PIPE_FORMAT_R32_FLOAT;
+	 emit_sz = 1 * sizeof(float);
+	 break;
+      case EMIT_1F_PSIZE:
+	 output_format = PIPE_FORMAT_R32_FLOAT;
+	 emit_sz = 1 * sizeof(float);
+	 src_buffer = 1;
+	 src_offset = 0;
+	 break;
+      case EMIT_4UB:
+	 output_format = PIPE_FORMAT_B8G8R8A8_UNORM;
+	 emit_sz = 4 * sizeof(ubyte);
+         break;
+      default:
+	 assert(0);
+	 output_format = PIPE_FORMAT_NONE;
+	 emit_sz = 0;
+	 break;
+      }
+      
+      hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+      hw_key.element[i].input_buffer = src_buffer;
+      hw_key.element[i].input_offset = src_offset;
+      hw_key.element[i].output_format = output_format;
+      hw_key.element[i].output_offset = dst_offset;
+
+      dst_offset += emit_sz;
+   }
+
+   hw_key.nr_elements = vinfo->num_attribs;
+   hw_key.output_stride = vinfo->size * 4;
+
+   if (!emit->translate ||
+       translate_key_compare(&emit->translate->key, &hw_key) != 0)
+   {
+      translate_key_sanitize(&hw_key);
+      emit->translate = translate_cache_find(emit->cache, &hw_key);
+   }
+
+   *max_vertices = (draw->render->max_vertex_buffer_bytes / 
+                    (vinfo->size * 4));
+
+   /* even number */
+   *max_vertices = *max_vertices & ~1;
+}
+
+
+void draw_pt_emit( struct pt_emit *emit,
+		   const float (*vertex_data)[4],
+		   unsigned vertex_count,
+		   unsigned stride,
+		   const ushort *elts,
+		   unsigned count )
+{
+   struct draw_context *draw = emit->draw;
+   struct translate *translate = emit->translate;
+   struct vbuf_render *render = draw->render;
+   void *hw_verts;
+
+   /* XXX: need to flush to get prim_vbuf.c to release its allocation?? 
+    */
+   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+
+   if (vertex_count >= UNDEFINED_VERTEX_ID) {
+      assert(0);
+      return;
+   }
+
+   /* XXX: and work out some way to coordinate the render primitive
+    * between vbuf.c and here...
+    */
+   if (!draw->render->set_primitive(draw->render, emit->prim)) {
+      assert(0);
+      return;
+   }
+
+   hw_verts = render->allocate_vertices(render,
+					(ushort)translate->key.output_stride,
+					(ushort)vertex_count);
+   if (!hw_verts) {
+      assert(0);
+      return;
+   }
+
+   translate->set_buffer(translate, 
+			 0, 
+			 vertex_data,
+			 stride );
+
+   translate->set_buffer(translate, 
+			 1, 
+			 &draw->rasterizer->point_size,
+			 0);
+
+   translate->run( translate,
+		   0, 
+		   vertex_count,
+		   hw_verts );
+
+   render->draw(render,
+		elts,
+		count);
+
+   render->release_vertices(render,
+			    hw_verts,
+			    translate->key.output_stride,
+			    vertex_count);
+}
+
+
+void draw_pt_emit_linear(struct pt_emit *emit,
+                         const float (*vertex_data)[4],
+                         unsigned vertex_count,
+                         unsigned stride,
+                         unsigned start,
+                         unsigned count)
+{
+   struct draw_context *draw = emit->draw;
+   struct translate *translate = emit->translate;
+   struct vbuf_render *render = draw->render;
+   void *hw_verts;
+
+#if 0
+   debug_printf("Linear emit\n");
+#endif
+   /* XXX: need to flush to get prim_vbuf.c to release its allocation?? 
+    */
+   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+
+   if (count >= UNDEFINED_VERTEX_ID) {
+      assert(0);
+      return;
+   }
+
+   /* XXX: and work out some way to coordinate the render primitive
+    * between vbuf.c and here...
+    */
+   if (!draw->render->set_primitive(draw->render, emit->prim)) {
+      assert(0);
+      return;
+   }
+
+   hw_verts = render->allocate_vertices(render,
+					(ushort)translate->key.output_stride,
+					(ushort)count);
+   if (!hw_verts) {
+      assert(0);
+      return;
+   }
+
+   translate->set_buffer(translate, 0,
+			 vertex_data, stride);
+
+   translate->set_buffer(translate, 1,
+			 &draw->rasterizer->point_size,
+			 0);
+
+   translate->run(translate,
+                  0,
+                  vertex_count,
+                  hw_verts);
+
+   if (0) {
+      unsigned i;
+      for (i = 0; i < vertex_count; i++) {
+         debug_printf("\n\n%s vertex %d:\n", __FUNCTION__, i);
+         draw_dump_emitted_vertex( emit->vinfo, 
+                                   (const uint8_t *)hw_verts + 
+                                   translate->key.output_stride * i );
+      }
+   }
+
+
+   render->draw_arrays(render, start, count);
+
+   render->release_vertices(render,
+			    hw_verts,
+			    translate->key.output_stride,
+			    vertex_count);
+}
+
+struct pt_emit *draw_pt_emit_create( struct draw_context *draw )
+{
+   struct pt_emit *emit = CALLOC_STRUCT(pt_emit);
+   if (!emit)
+      return NULL;
+
+   emit->draw = draw;
+   emit->cache = translate_cache_create();
+   if (!emit->cache) {
+      FREE(emit);
+      return NULL;
+   }
+
+   return emit;
+}
+
+void draw_pt_emit_destroy( struct pt_emit *emit )
+{
+   if (emit->cache)
+      translate_cache_destroy(emit->cache);
+
+   FREE(emit);
+}
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c
new file mode 100644
index 0000000000..058caf7dcc
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c
@@ -0,0 +1,228 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_memory.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vbuf.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_pt.h"
+#include "translate/translate.h"
+#include "translate/translate_cache.h"
+
+
+struct pt_fetch {
+   struct draw_context *draw;
+
+   struct translate *translate;
+
+   unsigned vertex_size;
+   boolean need_edgeflags;
+
+   struct translate_cache *cache;
+};
+
+/* Perform the fetch from API vertex elements & vertex buffers, to a
+ * contiguous set of float[4] attributes as required for the
+ * vertex_shader->run_linear() method.
+ *
+ * This is used in all cases except pure passthrough
+ * (draw_pt_fetch_emit.c) which has its own version to translate
+ * directly to hw vertices.
+ *
+ */
+void draw_pt_fetch_prepare( struct pt_fetch *fetch,
+			    unsigned vertex_size )
+{
+   struct draw_context *draw = fetch->draw;
+   unsigned i, nr = 0;
+   unsigned dst_offset = 0;
+   struct translate_key key;
+
+   fetch->vertex_size = vertex_size;
+
+   /* Always emit/leave space for a vertex header.
+    *
+    * It's worth considering whether the vertex headers should contain
+    * a pointer to the 'data', rather than having it inline.
+    * Something to look at after we've fully switched over to the pt
+    * paths.
+    */
+   {
+      /* Need to set header->vertex_id = 0xffff somehow.
+       */
+      key.element[nr].input_format = PIPE_FORMAT_R32_FLOAT;
+      key.element[nr].input_buffer = draw->pt.nr_vertex_buffers;
+      key.element[nr].input_offset = 0;
+      key.element[nr].output_format = PIPE_FORMAT_R32_FLOAT;
+      key.element[nr].output_offset = dst_offset;
+      dst_offset += 1 * sizeof(float);
+      nr++;
+
+
+      /* Just leave the clip[] array untouched.
+       */
+      dst_offset += 4 * sizeof(float);
+   }
+      
+
+   for (i = 0; i < draw->pt.nr_vertex_elements; i++) {
+      key.element[nr].input_format = draw->pt.vertex_element[i].src_format;
+      key.element[nr].input_buffer = draw->pt.vertex_element[i].vertex_buffer_index;
+      key.element[nr].input_offset = draw->pt.vertex_element[i].src_offset;
+      key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+      key.element[nr].output_offset = dst_offset;
+
+      dst_offset += 4 * sizeof(float);
+      nr++;
+   }
+
+   assert(dst_offset <= vertex_size);
+
+   key.nr_elements = nr;
+   key.output_stride = vertex_size;
+
+
+   if (!fetch->translate ||
+       translate_key_compare(&fetch->translate->key, &key) != 0)
+   {
+      translate_key_sanitize(&key);
+      fetch->translate = translate_cache_find(fetch->cache, &key);
+
+      {
+	 static struct vertex_header vh = { 0, 1, 0, 0xffff };
+	 fetch->translate->set_buffer(fetch->translate,
+				      draw->pt.nr_vertex_buffers,
+				      &vh,
+				      0);
+      }
+   }
+
+   fetch->need_edgeflags = ((draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL ||
+                             draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) &&
+                            draw->pt.user.edgeflag);
+}
+
+
+
+
+void draw_pt_fetch_run( struct pt_fetch *fetch,
+			const unsigned *elts,
+			unsigned count,
+			char *verts )
+{
+   struct draw_context *draw = fetch->draw;
+   struct translate *translate = fetch->translate;
+   unsigned i;
+
+   for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
+      translate->set_buffer(translate, 
+			    i, 
+			    ((char *)draw->pt.user.vbuffer[i] + 
+			     draw->pt.vertex_buffer[i].buffer_offset),
+			    draw->pt.vertex_buffer[i].stride );
+   }
+
+   translate->run_elts( translate,
+			elts, 
+			count,
+			verts );
+
+   /* Edgeflags are hard to fit into a translate program, populate
+    * them separately if required.  In the setup above they are
+    * defaulted to one, so only need this if there is reason to change
+    * that default:
+    */
+   if (fetch->need_edgeflags) {
+      for (i = 0; i < count; i++) {
+         struct vertex_header *vh = (struct vertex_header *)(verts + i * fetch->vertex_size);
+         vh->edgeflag = draw_pt_get_edgeflag( draw, elts[i] );
+      }
+   }
+}
+
+
+void draw_pt_fetch_run_linear( struct pt_fetch *fetch,
+                               unsigned start,
+                               unsigned count,
+                               char *verts )
+{
+   struct draw_context *draw = fetch->draw;
+   struct translate *translate = fetch->translate;
+   unsigned i;
+
+   for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
+      translate->set_buffer(translate,
+			    i,
+			    ((char *)draw->pt.user.vbuffer[i] +
+			     draw->pt.vertex_buffer[i].buffer_offset),
+			    draw->pt.vertex_buffer[i].stride );
+   }
+
+   translate->run( translate,
+                   start,
+                   count,
+                   verts );
+
+   /* Edgeflags are hard to fit into a translate program, populate
+    * them separately if required.  In the setup above they are
+    * defaulted to one, so only need this if there is reason to change
+    * that default:
+    */
+   if (fetch->need_edgeflags) {
+      for (i = 0; i < count; i++) {
+         struct vertex_header *vh = (struct vertex_header *)(verts + i * fetch->vertex_size);
+         vh->edgeflag = draw_pt_get_edgeflag( draw, start + i );
+      }
+   }
+}
+
+
+struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw )
+{
+   struct pt_fetch *fetch = CALLOC_STRUCT(pt_fetch);
+   if (!fetch)
+      return NULL;
+
+   fetch->draw = draw;
+   fetch->cache = translate_cache_create();
+   if (!fetch->cache) {
+      FREE(fetch);
+      return NULL;
+   }
+
+   return fetch;
+}
+
+void draw_pt_fetch_destroy( struct pt_fetch *fetch )
+{
+   if (fetch->cache)
+      translate_cache_destroy(fetch->cache);
+
+   FREE(fetch);
+}
+
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
new file mode 100644
index 0000000000..dcb7744b17
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c
@@ -0,0 +1,426 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "util/u_memory.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vbuf.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_pt.h"
+#include "translate/translate.h"
+#include "translate/translate_cache.h"
+
+/* The simplest 'middle end' in the new vertex code.  
+ * 
+ * The responsibilities of a middle end are to:
+ *  - perform vertex fetch using
+ *       - draw vertex element/buffer state
+ *       - a list of fetch indices we received as an input
+ *  - run the vertex shader
+ *  - cliptest, 
+ *  - clip coord calculation 
+ *  - viewport transformation
+ *  - if necessary, run the primitive pipeline, passing it:
+ *       - a linear array of vertex_header vertices constructed here
+ *       - a set of draw indices we received as an input
+ *  - otherwise, drive the hw backend,
+ *       - allocate space for hardware format vertices
+ *       - translate the vertex-shader output vertices to hw format
+ *       - calling the backend draw functions.
+ *
+ * For convenience, we provide a helper function to drive the hardware
+ * backend given similar inputs to those required to run the pipeline.
+ *
+ * In the case of passthrough mode, many of these actions are disabled
+ * or noops, so we end up doing:
+ *
+ *  - perform vertex fetch
+ *  - drive the hw backend
+ *
+ * IE, basically just vertex fetch to post-vs-format vertices,
+ * followed by a call to the backend helper function.
+ */
+
+
+struct fetch_emit_middle_end {
+   struct draw_pt_middle_end base;
+   struct draw_context *draw;
+   
+   struct translate *translate;
+   const struct vertex_info *vinfo;
+
+   /* Cache point size somewhere it's address won't change:
+    */
+   float point_size;
+
+   struct translate_cache *cache;
+};
+
+
+
+
+static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
+                                unsigned prim,
+				unsigned opt,
+                                unsigned *max_vertices )
+{
+   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
+   struct draw_context *draw = feme->draw;
+   const struct vertex_info *vinfo;
+   unsigned i, dst_offset;
+   boolean ok;
+   struct translate_key key;
+
+
+   ok = draw->render->set_primitive( draw->render, 
+                                     prim );
+   if (!ok) {
+      assert(0);
+      return;
+   }
+   
+   /* Must do this after set_primitive() above:
+    */
+   vinfo = feme->vinfo = draw->render->get_vertex_info(draw->render);
+   
+   
+
+   /* Transform from API vertices to HW vertices, skipping the
+    * pipeline_vertex intermediate step.
+    */
+   dst_offset = 0;
+   memset(&key, 0, sizeof(key));
+
+   for (i = 0; i < vinfo->num_attribs; i++) {
+      const struct pipe_vertex_element *src = &draw->pt.vertex_element[vinfo->attrib[i].src_index];
+
+      unsigned emit_sz = 0;
+      unsigned input_format = src->src_format;
+      unsigned input_buffer = src->vertex_buffer_index;
+      unsigned input_offset = src->src_offset;
+      unsigned output_format;
+
+      switch (vinfo->attrib[i].emit) {
+      case EMIT_4F:
+	 output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+	 emit_sz = 4 * sizeof(float);
+         break;
+      case EMIT_3F:
+	 output_format = PIPE_FORMAT_R32G32B32_FLOAT;
+	 emit_sz = 3 * sizeof(float);
+         break;
+      case EMIT_2F:
+	 output_format = PIPE_FORMAT_R32G32_FLOAT;
+	 emit_sz = 2 * sizeof(float);
+         break;
+      case EMIT_1F:
+	 output_format = PIPE_FORMAT_R32_FLOAT;
+	 emit_sz = 1 * sizeof(float);
+         break;
+      case EMIT_1F_PSIZE:
+	 input_format = PIPE_FORMAT_R32_FLOAT;
+	 input_buffer = draw->pt.nr_vertex_buffers;
+	 input_offset = 0;
+	 output_format = PIPE_FORMAT_R32_FLOAT;
+	 emit_sz = 1 * sizeof(float);
+         break;
+      default:
+         assert(0);
+	 output_format = PIPE_FORMAT_NONE;
+	 emit_sz = 0;
+	 continue;
+      }
+
+      key.element[i].input_format = input_format;
+      key.element[i].input_buffer = input_buffer;
+      key.element[i].input_offset = input_offset;
+      key.element[i].output_format = output_format;
+      key.element[i].output_offset = dst_offset;
+      
+      dst_offset += emit_sz;
+   }
+
+   key.nr_elements = vinfo->num_attribs;
+   key.output_stride = vinfo->size * 4;
+
+   /* Don't bother with caching at this stage:
+    */
+   if (!feme->translate ||
+       translate_key_compare(&feme->translate->key, &key) != 0) 
+   {
+      translate_key_sanitize(&key);
+      feme->translate = translate_cache_find(feme->cache,
+                                             &key);
+
+
+      feme->translate->set_buffer(feme->translate, 
+				  draw->pt.nr_vertex_buffers, 
+				  &feme->point_size,
+				  0);
+   }
+   
+   feme->point_size = draw->rasterizer->point_size;
+
+   for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
+      feme->translate->set_buffer(feme->translate, 
+                                  i, 
+                                  ((char *)draw->pt.user.vbuffer[i] + 
+                                   draw->pt.vertex_buffer[i].buffer_offset),
+                                  draw->pt.vertex_buffer[i].stride );
+   }
+
+   *max_vertices = (draw->render->max_vertex_buffer_bytes / 
+                    (vinfo->size * 4));
+
+   /* Return an even number of verts.
+    * This prevents "parity" errors when splitting long triangle strips which
+    * can lead to front/back culling mix-ups.
+    * Every other triangle in a strip has an alternate front/back orientation
+    * so splitting at an odd position can cause the orientation of subsequent
+    * triangles to get reversed.
+    */
+   *max_vertices = *max_vertices & ~1;
+}
+
+
+
+
+
+static void fetch_emit_run( struct draw_pt_middle_end *middle,
+                            const unsigned *fetch_elts,
+                            unsigned fetch_count,
+                            const ushort *draw_elts,
+                            unsigned draw_count )
+{
+   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
+   struct draw_context *draw = feme->draw;
+   void *hw_verts;
+   
+   /* XXX: need to flush to get prim_vbuf.c to release its allocation?? 
+    */
+   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+
+   if (fetch_count >= UNDEFINED_VERTEX_ID) {
+      assert(0);
+      return;
+   }
+
+   hw_verts = draw->render->allocate_vertices( draw->render,
+                                               (ushort)feme->translate->key.output_stride,
+                                               (ushort)fetch_count );
+   if (!hw_verts) {
+      assert(0);
+      return;
+   }
+         
+					
+   /* Single routine to fetch vertices and emit HW verts.
+    */
+   feme->translate->run_elts( feme->translate, 
+			      fetch_elts,
+			      fetch_count,
+			      hw_verts );
+
+   if (0) {
+      unsigned i;
+      for (i = 0; i < fetch_count; i++) {
+         debug_printf("\n\nvertex %d:\n", i);
+         draw_dump_emitted_vertex( feme->vinfo, 
+                                   (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
+      }
+   }
+
+   /* XXX: Draw arrays path to avoid re-emitting index list again and
+    * again.
+    */
+   draw->render->draw( draw->render, 
+                       draw_elts, 
+                       draw_count );
+
+   /* Done -- that was easy, wasn't it: 
+    */
+   draw->render->release_vertices( draw->render, 
+                                   hw_verts, 
+                                   feme->translate->key.output_stride, 
+                                   fetch_count );
+
+}
+
+
+static void fetch_emit_run_linear( struct draw_pt_middle_end *middle,
+                                   unsigned start,
+                                   unsigned count )
+{
+   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
+   struct draw_context *draw = feme->draw;
+   void *hw_verts;
+
+   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
+    */
+   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+
+   if (count >= UNDEFINED_VERTEX_ID) {
+      assert(0);
+      return;
+   }
+
+   hw_verts = draw->render->allocate_vertices( draw->render,
+                                               (ushort)feme->translate->key.output_stride,
+                                               (ushort)count );
+   if (!hw_verts) {
+      assert(0);
+      return;
+   }
+
+   /* Single routine to fetch vertices and emit HW verts.
+    */
+   feme->translate->run( feme->translate,
+                         start,
+                         count,
+                         hw_verts );
+
+   if (0) {
+      unsigned i;
+      for (i = 0; i < count; i++) {
+         debug_printf("\n\nvertex %d:\n", i);
+         draw_dump_emitted_vertex( feme->vinfo,
+                                   (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
+      }
+   }
+
+   /* XXX: Draw arrays path to avoid re-emitting index list again and
+    * again.
+    */
+   draw->render->draw_arrays( draw->render,
+                              0, /*start*/
+                              count );
+
+   /* Done -- that was easy, wasn't it:
+    */
+   draw->render->release_vertices( draw->render,
+                                   hw_verts,
+                                   feme->translate->key.output_stride,
+                                   count );
+
+}
+
+
+static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle,
+                                        unsigned start,
+                                        unsigned count,
+                                        const ushort *draw_elts,
+                                        unsigned draw_count )
+{
+   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
+   struct draw_context *draw = feme->draw;
+   void *hw_verts;
+
+   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
+    */
+   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+
+   if (count >= UNDEFINED_VERTEX_ID)
+      return FALSE;
+
+   hw_verts = draw->render->allocate_vertices( draw->render,
+                                               (ushort)feme->translate->key.output_stride,
+                                               (ushort)count );
+   if (!hw_verts) 
+      return FALSE;
+
+   /* Single routine to fetch vertices and emit HW verts.
+    */
+   feme->translate->run( feme->translate,
+                         start,
+                         count,
+                         hw_verts );
+
+   /* XXX: Draw arrays path to avoid re-emitting index list again and
+    * again.
+    */
+   draw->render->draw( draw->render, 
+                       draw_elts, 
+                       draw_count );
+
+   /* Done -- that was easy, wasn't it:
+    */
+   draw->render->release_vertices( draw->render,
+                                   hw_verts,
+                                   feme->translate->key.output_stride,
+                                   count );
+
+   return TRUE;
+}
+
+
+
+
+static void fetch_emit_finish( struct draw_pt_middle_end *middle )
+{
+   /* nothing to do */
+}
+
+static void fetch_emit_destroy( struct draw_pt_middle_end *middle )
+{
+   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
+
+   if (feme->cache)
+      translate_cache_destroy(feme->cache);
+
+   FREE(middle);
+}
+
+
+struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw )
+{
+   struct fetch_emit_middle_end *fetch_emit = CALLOC_STRUCT( fetch_emit_middle_end );
+   if (fetch_emit == NULL)
+      return NULL;
+
+   fetch_emit->cache = translate_cache_create();
+   if (!fetch_emit->cache) {
+      FREE(fetch_emit);
+      return NULL;
+   }
+
+   fetch_emit->base.prepare    = fetch_emit_prepare;
+   fetch_emit->base.run        = fetch_emit_run;
+   fetch_emit->base.run_linear = fetch_emit_run_linear;
+   fetch_emit->base.run_linear_elts = fetch_emit_run_linear_elts;
+   fetch_emit->base.finish     = fetch_emit_finish;
+   fetch_emit->base.destroy    = fetch_emit_destroy;
+
+   fetch_emit->draw = draw;
+     
+   return &fetch_emit->base;
+}
+
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c
new file mode 100644
index 0000000000..84ffe3296a
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c
@@ -0,0 +1,422 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vbuf.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_pt.h"
+#include "draw/draw_vs.h"
+
+#include "translate/translate.h"
+
+struct fetch_shade_emit;
+
+
+/* Prototype fetch, shade, emit-hw-verts all in one go.
+ */
+struct fetch_shade_emit {
+   struct draw_pt_middle_end base;
+   struct draw_context *draw;
+
+
+   /* Temporaries:
+    */
+   const float *constants;
+   unsigned pitch[PIPE_MAX_ATTRIBS];
+   const ubyte *src[PIPE_MAX_ATTRIBS];
+   unsigned prim;
+
+   struct draw_vs_varient_key key;
+   struct draw_vs_varient *active;
+
+
+   const struct vertex_info *vinfo;
+};
+
+
+
+			       
+static void fse_prepare( struct draw_pt_middle_end *middle,
+                         unsigned prim, 
+                         unsigned opt,
+                         unsigned *max_vertices )
+{
+   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
+   struct draw_context *draw = fse->draw;
+   unsigned num_vs_inputs = draw->vs.vertex_shader->info.num_inputs;
+   const struct vertex_info *vinfo;
+   unsigned i;
+   unsigned nr_vbs = 0;
+   
+
+   if (!draw->render->set_primitive( draw->render, 
+                                     prim )) {
+      assert(0);
+      return;
+   }
+
+   /* Must do this after set_primitive() above:
+    */
+   fse->vinfo = vinfo = draw->render->get_vertex_info(draw->render);
+   
+
+
+   fse->key.output_stride = vinfo->size * 4;
+   fse->key.nr_outputs = vinfo->num_attribs;
+   fse->key.nr_inputs = num_vs_inputs;
+
+   fse->key.nr_elements = MAX2(fse->key.nr_outputs,     /* outputs - translate to hw format */
+                               fse->key.nr_inputs);     /* inputs - fetch from api format */
+
+   fse->key.viewport = !draw->identity_viewport;
+   fse->key.clip = !draw->bypass_clipping;
+   fse->key.const_vbuffers = 0;
+
+   memset(fse->key.element, 0, 
+          fse->key.nr_elements * sizeof(fse->key.element[0]));
+
+   for (i = 0; i < num_vs_inputs; i++) {
+      const struct pipe_vertex_element *src = &draw->pt.vertex_element[i];
+      fse->key.element[i].in.format = src->src_format;
+
+      /* Consider ignoring these, ie make generated programs
+       * independent of this state:
+       */
+      fse->key.element[i].in.buffer = src->vertex_buffer_index;
+      fse->key.element[i].in.offset = src->src_offset;
+      nr_vbs = MAX2(nr_vbs, src->vertex_buffer_index + 1);
+   }
+   
+   for (i = 0; i < 5 && i < nr_vbs; i++) {
+      if (draw->pt.vertex_buffer[i].stride == 0)
+         fse->key.const_vbuffers |= (1<<i);
+   }
+
+   if (0) debug_printf("%s: lookup const_vbuffers: %x\n", __FUNCTION__, fse->key.const_vbuffers);
+   
+   {
+      unsigned dst_offset = 0;
+
+      for (i = 0; i < vinfo->num_attribs; i++) {
+         unsigned emit_sz = 0;
+
+         switch (vinfo->attrib[i].emit) {
+         case EMIT_4F:
+            emit_sz = 4 * sizeof(float);
+            break;
+         case EMIT_3F:
+            emit_sz = 3 * sizeof(float);
+            break;
+         case EMIT_2F:
+            emit_sz = 2 * sizeof(float);
+            break;
+         case EMIT_1F:
+            emit_sz = 1 * sizeof(float);
+            break;
+         case EMIT_1F_PSIZE:
+            emit_sz = 1 * sizeof(float);
+            break;
+         case EMIT_4UB:
+            emit_sz = 4 * sizeof(ubyte);
+            break;
+         default:
+            assert(0);
+            break;
+         }
+
+         /* The elements in the key correspond to vertex shader output
+          * numbers, not to positions in the hw vertex description --
+          * that's handled by the output_offset field.
+          */
+         fse->key.element[i].out.format = vinfo->attrib[i].emit;
+         fse->key.element[i].out.vs_output = vinfo->attrib[i].src_index;
+         fse->key.element[i].out.offset = dst_offset;
+      
+         dst_offset += emit_sz;
+         assert(fse->key.output_stride >= dst_offset);
+      }
+   }
+
+   
+   fse->active = draw_vs_lookup_varient( draw->vs.vertex_shader, 
+                                         &fse->key );
+
+   if (!fse->active) {
+      assert(0);
+      return ;
+   }
+
+   if (0) debug_printf("%s: found const_vbuffers: %x\n", __FUNCTION__, 
+                       fse->active->key.const_vbuffers);
+
+   /* Now set buffer pointers:
+    */
+   for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
+      fse->active->set_buffer( fse->active, 
+                               i, 
+                               ((const ubyte *) draw->pt.user.vbuffer[i] + 
+                                draw->pt.vertex_buffer[i].buffer_offset),
+                              draw->pt.vertex_buffer[i].stride );
+   }
+
+   *max_vertices = (draw->render->max_vertex_buffer_bytes / 
+                    (vinfo->size * 4));
+
+   /* Return an even number of verts.
+    * This prevents "parity" errors when splitting long triangle strips which
+    * can lead to front/back culling mix-ups.
+    * Every other triangle in a strip has an alternate front/back orientation
+    * so splitting at an odd position can cause the orientation of subsequent
+    * triangles to get reversed.
+    */
+   *max_vertices = *max_vertices & ~1;
+
+   /* Probably need to do this somewhere (or fix exec shader not to
+    * need it):
+    */
+   if (1) {
+      struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+      vs->prepare(vs, draw);
+   }
+   
+
+   //return TRUE;
+}
+
+
+
+
+
+
+
+static void fse_run_linear( struct draw_pt_middle_end *middle, 
+                            unsigned start, 
+                            unsigned count )
+{
+   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
+   struct draw_context *draw = fse->draw;
+   char *hw_verts;
+
+   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
+    */
+   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+
+   if (count >= UNDEFINED_VERTEX_ID) {
+      assert(0);
+      return;
+   }
+
+   hw_verts = draw->render->allocate_vertices( draw->render,
+                                               (ushort)fse->key.output_stride,
+                                               (ushort)count );
+
+   if (!hw_verts) {
+      assert(0);
+      return;
+   }
+
+   /* Single routine to fetch vertices, run shader and emit HW verts.
+    * Clipping is done elsewhere -- either by the API or on hardware,
+    * or for some other reason not required...
+    */
+   fse->active->run_linear( fse->active, 
+                            start, count,
+                            hw_verts );
+
+   /* Draw arrays path to avoid re-emitting index list again and
+    * again.
+    */
+   draw->render->draw_arrays( draw->render,
+                              0,
+                              count );
+   
+   if (0) {
+      unsigned i;
+      for (i = 0; i < count; i++) {
+         debug_printf("\n\n%s vertex %d: (stride %d, offset %d)\n", __FUNCTION__, i,
+                      fse->key.output_stride,
+                      fse->key.output_stride * i);
+
+         draw_dump_emitted_vertex( fse->vinfo, 
+                                   (const uint8_t *)hw_verts + fse->key.output_stride * i );
+      }
+   }
+
+
+   draw->render->release_vertices( draw->render, 
+				   hw_verts, 
+				   fse->key.output_stride, 
+				   count );
+}
+
+
+static void
+fse_run(struct draw_pt_middle_end *middle,
+        const unsigned *fetch_elts,
+        unsigned fetch_count,
+        const ushort *draw_elts,
+        unsigned draw_count )
+{
+   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
+   struct draw_context *draw = fse->draw;
+   void *hw_verts;
+   
+   /* XXX: need to flush to get prim_vbuf.c to release its allocation?? 
+    */
+   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+
+   if (fetch_count >= UNDEFINED_VERTEX_ID) {
+      assert(0);
+      return;
+   }
+
+   hw_verts = draw->render->allocate_vertices( draw->render,
+                                               (ushort)fse->key.output_stride,
+                                               (ushort)fetch_count );
+   if (!hw_verts) {
+      assert(0);
+      return;
+   }
+         
+					
+   /* Single routine to fetch vertices, run shader and emit HW verts.
+    */
+   fse->active->run_elts( fse->active, 
+                          fetch_elts,
+                          fetch_count,
+                          hw_verts );
+
+   draw->render->draw( draw->render, 
+                       draw_elts, 
+                       draw_count );
+
+   if (0) {
+      unsigned i;
+      for (i = 0; i < fetch_count; i++) {
+         debug_printf("\n\n%s vertex %d:\n", __FUNCTION__, i);
+         draw_dump_emitted_vertex( fse->vinfo, 
+                                   (const uint8_t *)hw_verts + 
+                                   fse->key.output_stride * i );
+      }
+   }
+
+
+   draw->render->release_vertices( draw->render, 
+                                   hw_verts, 
+                                   fse->key.output_stride, 
+                                   fetch_count );
+
+}
+
+
+
+static boolean fse_run_linear_elts( struct draw_pt_middle_end *middle, 
+                                 unsigned start, 
+                                 unsigned count,
+                                 const ushort *draw_elts,
+                                 unsigned draw_count )
+{
+   struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle;
+   struct draw_context *draw = fse->draw;
+   char *hw_verts;
+
+   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
+    */
+   draw_do_flush( draw, DRAW_FLUSH_BACKEND );
+
+   if (count >= UNDEFINED_VERTEX_ID)
+      return FALSE;
+
+   hw_verts = draw->render->allocate_vertices( draw->render,
+                                               (ushort)fse->key.output_stride,
+                                               (ushort)count );
+
+   if (!hw_verts) {
+      return FALSE;
+   }
+
+   /* Single routine to fetch vertices, run shader and emit HW verts.
+    * Clipping is done elsewhere -- either by the API or on hardware,
+    * or for some other reason not required...
+    */
+   fse->active->run_linear( fse->active, 
+                            start, count,
+                            hw_verts );
+
+
+   draw->render->draw( draw->render, 
+                       draw_elts, 
+                       draw_count );
+   
+
+
+   draw->render->release_vertices( draw->render, 
+				   hw_verts, 
+				   fse->key.output_stride, 
+				   count );
+
+   return TRUE;
+}
+
+
+
+static void fse_finish( struct draw_pt_middle_end *middle )
+{
+}
+
+
+static void
+fse_destroy( struct draw_pt_middle_end *middle ) 
+{
+   FREE(middle);
+}
+
+struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw )
+{
+   struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit);
+   if (!fse)
+      return NULL;
+
+   fse->base.prepare = fse_prepare;
+   fse->base.run = fse_run;
+   fse->base.run_linear = fse_run_linear;
+   fse->base.run_linear_elts = fse_run_linear_elts;
+   fse->base.finish = fse_finish;
+   fse->base.destroy = fse_destroy;
+   fse->draw = draw;
+
+   return &fse->base;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
new file mode 100644
index 0000000000..ec3b41c320
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
@@ -0,0 +1,396 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vbuf.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_pt.h"
+#include "draw/draw_vs.h"
+#include "translate/translate.h"
+
+
+struct fetch_pipeline_middle_end {
+   struct draw_pt_middle_end base;
+   struct draw_context *draw;
+
+   struct pt_emit *emit;
+   struct pt_fetch *fetch;
+   struct pt_post_vs *post_vs;
+
+   unsigned vertex_data_offset;
+   unsigned vertex_size;
+   unsigned prim;
+   unsigned opt;
+};
+
+
+static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle,
+                                    unsigned prim,
+				    unsigned opt,
+                                    unsigned *max_vertices )
+{
+   struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+   struct draw_context *draw = fpme->draw;
+   struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+
+   /* Add one to num_outputs because the pipeline occasionally tags on
+    * an additional texcoord, eg for AA lines.
+    */
+   unsigned nr = MAX2( vs->info.num_inputs,
+		       vs->info.num_outputs + 1 );
+
+   fpme->prim = prim;
+   fpme->opt = opt;
+
+   /* Always leave room for the vertex header whether we need it or
+    * not.  It's hard to get rid of it in particular because of the
+    * viewport code in draw_pt_post_vs.c.  
+    */
+   fpme->vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float);
+
+   
+
+   draw_pt_fetch_prepare( fpme->fetch, 
+			  fpme->vertex_size );
+
+   /* XXX: it's not really gl rasterization rules we care about here,
+    * but gl vs dx9 clip spaces.
+    */
+   draw_pt_post_vs_prepare( fpme->post_vs,
+			    (boolean)draw->bypass_clipping,
+			    (boolean)draw->identity_viewport,
+			    (boolean)draw->rasterizer->gl_rasterization_rules );
+			    
+
+   if (!(opt & PT_PIPELINE)) {
+      draw_pt_emit_prepare( fpme->emit, 
+			    prim,
+                            max_vertices );
+
+      *max_vertices = MAX2( *max_vertices,
+                            DRAW_PIPE_MAX_VERTICES );
+   }
+   else {
+      *max_vertices = DRAW_PIPE_MAX_VERTICES; 
+   }
+
+   /* return even number */
+   *max_vertices = *max_vertices & ~1;
+
+   /* No need to prepare the shader.
+    */
+   vs->prepare(vs, draw);
+}
+
+
+
+static void fetch_pipeline_run( struct draw_pt_middle_end *middle,
+                                const unsigned *fetch_elts,
+                                unsigned fetch_count,
+                                const ushort *draw_elts,
+                                unsigned draw_count )
+{
+   struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+   struct draw_context *draw = fpme->draw;
+   struct draw_vertex_shader *shader = draw->vs.vertex_shader;
+   unsigned opt = fpme->opt;
+   unsigned alloc_count = align( fetch_count, 4 );
+
+   struct vertex_header *pipeline_verts = 
+      (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count);
+
+   if (!pipeline_verts) {
+      /* Not much we can do here - just skip the rendering.
+       */
+      assert(0);
+      return;
+   }
+
+   /* Fetch into our vertex buffer
+    */
+   draw_pt_fetch_run( fpme->fetch,
+		      fetch_elts, 
+		      fetch_count,
+		      (char *)pipeline_verts );
+
+   /* Run the shader, note that this overwrites the data[] parts of
+    * the pipeline verts.  If there is no shader, ie a bypass shader,
+    * then the inputs == outputs, and are already in the correct
+    * place.
+    */
+   if (opt & PT_SHADE)
+   {
+      shader->run_linear(shader, 
+			 (const float (*)[4])pipeline_verts->data,
+			 (      float (*)[4])pipeline_verts->data,
+			 (const float (*)[4])draw->pt.user.constants,
+			 fetch_count,
+			 fpme->vertex_size,
+			 fpme->vertex_size);
+   }
+
+   if (draw_pt_post_vs_run( fpme->post_vs,
+			    pipeline_verts,
+			    fetch_count,
+			    fpme->vertex_size ))
+   {
+      opt |= PT_PIPELINE;
+   }
+
+   /* Do we need to run the pipeline?
+    */
+   if (opt & PT_PIPELINE) {
+      draw_pipeline_run( fpme->draw,
+                         fpme->prim,
+                         pipeline_verts,
+                         fetch_count,
+                         fpme->vertex_size,
+                         draw_elts,
+                         draw_count );
+   }
+   else {
+      draw_pt_emit( fpme->emit,
+		    (const float (*)[4])pipeline_verts->data,
+		    fetch_count,
+		    fpme->vertex_size,
+		    draw_elts,
+		    draw_count );
+   }
+
+
+   FREE(pipeline_verts);
+}
+
+
+static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle,
+                                       unsigned start,
+                                       unsigned count)
+{
+   struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+   struct draw_context *draw = fpme->draw;
+   struct draw_vertex_shader *shader = draw->vs.vertex_shader;
+   unsigned opt = fpme->opt;
+   unsigned alloc_count = align( count, 4 );
+
+   struct vertex_header *pipeline_verts =
+      (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count);
+
+   if (!pipeline_verts) {
+      /* Not much we can do here - just skip the rendering.
+       */
+      assert(0);
+      return;
+   }
+
+   /* Fetch into our vertex buffer
+    */
+   draw_pt_fetch_run_linear( fpme->fetch,
+                             start,
+                             count,
+                             (char *)pipeline_verts );
+
+   /* Run the shader, note that this overwrites the data[] parts of
+    * the pipeline verts.  If there is no shader, ie a bypass shader,
+    * then the inputs == outputs, and are already in the correct
+    * place.
+    */
+   if (opt & PT_SHADE)
+   {
+      shader->run_linear(shader,
+			 (const float (*)[4])pipeline_verts->data,
+			 (      float (*)[4])pipeline_verts->data,
+			 (const float (*)[4])draw->pt.user.constants,
+			 count,
+			 fpme->vertex_size,
+			 fpme->vertex_size);
+   }
+
+   if (draw_pt_post_vs_run( fpme->post_vs,
+			    pipeline_verts,
+			    count,
+			    fpme->vertex_size ))
+   {
+      opt |= PT_PIPELINE;
+   }
+
+   /* Do we need to run the pipeline?
+    */
+   if (opt & PT_PIPELINE) {
+      draw_pipeline_run_linear( fpme->draw,
+                                fpme->prim,
+                                pipeline_verts,
+                                count,
+                                fpme->vertex_size);
+   }
+   else {
+      draw_pt_emit_linear( fpme->emit,
+                           (const float (*)[4])pipeline_verts->data,
+                           count,
+                           fpme->vertex_size,
+                           0, /*start*/
+                           count );
+   }
+
+   FREE(pipeline_verts);
+}
+
+
+
+static boolean fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle,
+                                            unsigned start,
+                                            unsigned count,
+                                            const ushort *draw_elts,
+                                            unsigned draw_count )
+{
+   struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+   struct draw_context *draw = fpme->draw;
+   struct draw_vertex_shader *shader = draw->vs.vertex_shader;
+   unsigned opt = fpme->opt;
+   unsigned alloc_count = align( count, 4 );
+
+   struct vertex_header *pipeline_verts =
+      (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count);
+
+   if (!pipeline_verts) 
+      return FALSE;
+
+   /* Fetch into our vertex buffer
+    */
+   draw_pt_fetch_run_linear( fpme->fetch,
+                             start,
+                             count,
+                             (char *)pipeline_verts );
+
+   /* Run the shader, note that this overwrites the data[] parts of
+    * the pipeline verts.  If there is no shader, ie a bypass shader,
+    * then the inputs == outputs, and are already in the correct
+    * place.
+    */
+   if (opt & PT_SHADE)
+   {
+      shader->run_linear(shader,
+			 (const float (*)[4])pipeline_verts->data,
+			 (      float (*)[4])pipeline_verts->data,
+			 (const float (*)[4])draw->pt.user.constants,
+			 count,
+			 fpme->vertex_size,
+			 fpme->vertex_size);
+   }
+
+   if (draw_pt_post_vs_run( fpme->post_vs,
+			    pipeline_verts,
+			    count,
+			    fpme->vertex_size ))
+   {
+      opt |= PT_PIPELINE;
+   }
+
+   /* Do we need to run the pipeline?
+    */
+   if (opt & PT_PIPELINE) {
+      draw_pipeline_run( fpme->draw,
+                         fpme->prim,
+                         pipeline_verts,
+                         count,
+                         fpme->vertex_size,
+                         draw_elts,
+                         draw_count );
+   }
+   else {
+      draw_pt_emit( fpme->emit,
+		    (const float (*)[4])pipeline_verts->data,
+		    count,
+		    fpme->vertex_size,
+		    draw_elts,
+		    draw_count );
+   }
+
+   FREE(pipeline_verts);
+   return TRUE;
+}
+
+
+
+static void fetch_pipeline_finish( struct draw_pt_middle_end *middle )
+{
+   /* nothing to do */
+}
+
+static void fetch_pipeline_destroy( struct draw_pt_middle_end *middle )
+{
+   struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
+
+   if (fpme->fetch)
+      draw_pt_fetch_destroy( fpme->fetch );
+
+   if (fpme->emit)
+      draw_pt_emit_destroy( fpme->emit );
+
+   if (fpme->post_vs)
+      draw_pt_post_vs_destroy( fpme->post_vs );
+
+   FREE(middle);
+}
+
+
+struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit( struct draw_context *draw )
+{
+   struct fetch_pipeline_middle_end *fpme = CALLOC_STRUCT( fetch_pipeline_middle_end );
+   if (!fpme)
+      goto fail;
+
+   fpme->base.prepare        = fetch_pipeline_prepare;
+   fpme->base.run            = fetch_pipeline_run;
+   fpme->base.run_linear     = fetch_pipeline_linear_run;
+   fpme->base.run_linear_elts = fetch_pipeline_linear_run_elts;
+   fpme->base.finish         = fetch_pipeline_finish;
+   fpme->base.destroy        = fetch_pipeline_destroy;
+
+   fpme->draw = draw;
+
+   fpme->fetch = draw_pt_fetch_create( draw );
+   if (!fpme->fetch)
+      goto fail;
+
+   fpme->post_vs = draw_pt_post_vs_create( draw );
+   if (!fpme->post_vs)
+      goto fail;
+
+   fpme->emit = draw_pt_emit_create( draw );
+   if (!fpme->emit) 
+      goto fail;
+
+   return &fpme->base;
+
+ fail:
+   if (fpme)
+      fetch_pipeline_destroy( &fpme->base );
+
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c
new file mode 100644
index 0000000000..96dc706b99
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c
@@ -0,0 +1,233 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_memory.h"
+#include "pipe/p_context.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vbuf.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_pt.h"
+
+struct pt_post_vs {
+   struct draw_context *draw;
+
+   boolean (*run)( struct pt_post_vs *pvs,
+		struct vertex_header *vertices,
+		unsigned count,
+		unsigned stride );
+};
+
+
+
+static INLINE float
+dot4(const float *a, const float *b)
+{
+   return (a[0]*b[0] +
+           a[1]*b[1] +
+           a[2]*b[2] +
+           a[3]*b[3]);
+}
+
+
+
+static INLINE unsigned
+compute_clipmask_gl(const float *clip, /*const*/ float plane[][4], unsigned nr)
+{
+   unsigned mask = 0x0;
+   unsigned i;
+
+#if 0
+   debug_printf("compute clipmask %f %f %f %f\n",
+                clip[0], clip[1], clip[2], clip[3]);
+   assert(clip[3] != 0.0);
+#endif
+
+   /* Do the hardwired planes first:
+    */
+   if (-clip[0] + clip[3] < 0) mask |= (1<<0);
+   if ( clip[0] + clip[3] < 0) mask |= (1<<1);
+   if (-clip[1] + clip[3] < 0) mask |= (1<<2);
+   if ( clip[1] + clip[3] < 0) mask |= (1<<3);
+   if ( clip[2] + clip[3] < 0) mask |= (1<<4); /* match mesa clipplane numbering - for now */
+   if (-clip[2] + clip[3] < 0) mask |= (1<<5); /* match mesa clipplane numbering - for now */
+
+   /* Followed by any remaining ones:
+    */
+   for (i = 6; i < nr; i++) {
+      if (dot4(clip, plane[i]) < 0) 
+         mask |= (1<<i);
+   }
+
+   return mask;
+}
+
+
+/* The normal case - cliptest, rhw divide, viewport transform.
+ *
+ * Also handle identity viewport here at the expense of a few wasted
+ * instructions
+ */
+static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs,
+					  struct vertex_header *vertices,
+					  unsigned count,
+					  unsigned stride )
+{
+   struct vertex_header *out = vertices;
+   const float *scale = pvs->draw->viewport.scale;
+   const float *trans = pvs->draw->viewport.translate;
+   const unsigned pos = pvs->draw->vs.position_output;
+   unsigned clipped = 0;
+   unsigned j;
+
+   if (0) debug_printf("%s\n");
+
+   for (j = 0; j < count; j++) {
+      float *position = out->data[pos];
+
+      out->clip[0] = position[0];
+      out->clip[1] = position[1];
+      out->clip[2] = position[2];
+      out->clip[3] = position[3];
+
+      out->vertex_id = 0xffff;
+      out->clipmask = compute_clipmask_gl(out->clip, 
+					  pvs->draw->plane,
+					  pvs->draw->nr_planes);
+      clipped += out->clipmask;
+
+      if (out->clipmask == 0)
+      {
+	 /* divide by w */
+	 float w = 1.0f / position[3];
+
+	 /* Viewport mapping */
+	 position[0] = position[0] * w * scale[0] + trans[0];
+	 position[1] = position[1] * w * scale[1] + trans[1];
+	 position[2] = position[2] * w * scale[2] + trans[2];
+	 position[3] = w;
+#if 0
+         debug_printf("post viewport: %f %f %f %f\n",
+                      position[0],
+                      position[1],
+                      position[2],
+                      position[3]);
+#endif
+      }
+
+      out = (struct vertex_header *)( (char *)out + stride );
+   }
+
+   return clipped != 0;
+}
+
+
+
+/* If bypass_clipping is set, skip cliptest and rhw divide.
+ */
+static boolean post_vs_viewport( struct pt_post_vs *pvs,
+			      struct vertex_header *vertices,
+			      unsigned count,
+			      unsigned stride )
+{
+   struct vertex_header *out = vertices;
+   const float *scale = pvs->draw->viewport.scale;
+   const float *trans = pvs->draw->viewport.translate;
+   const unsigned pos = pvs->draw->vs.position_output;
+   unsigned j;
+
+   if (0) debug_printf("%s\n", __FUNCTION__);
+   for (j = 0; j < count; j++) {
+      float *position = out->data[pos];
+
+      /* Viewport mapping only, no cliptest/rhw divide
+       */
+      position[0] = position[0] * scale[0] + trans[0];
+      position[1] = position[1] * scale[1] + trans[1];
+      position[2] = position[2] * scale[2] + trans[2];
+
+      out = (struct vertex_header *)((char *)out + stride);
+   }
+   
+   return FALSE;
+}
+
+
+/* If bypass_clipping is set and we have an identity viewport, nothing
+ * to do.
+ */
+static boolean post_vs_none( struct pt_post_vs *pvs,
+			     struct vertex_header *vertices,
+			     unsigned count,
+			     unsigned stride )
+{
+   if (0) debug_printf("%s\n", __FUNCTION__);
+   return FALSE;
+}
+
+boolean draw_pt_post_vs_run( struct pt_post_vs *pvs,
+			     struct vertex_header *pipeline_verts,
+			     unsigned count,
+			     unsigned stride )
+{
+   return pvs->run( pvs, pipeline_verts, count, stride );
+}
+
+
+void draw_pt_post_vs_prepare( struct pt_post_vs *pvs,
+			      boolean bypass_clipping,
+			      boolean identity_viewport,
+			      boolean opengl )
+{
+   if (bypass_clipping) {
+      if (identity_viewport)
+	 pvs->run = post_vs_none;
+      else
+	 pvs->run = post_vs_viewport;
+   }
+   else {
+      //if (opengl) 
+      pvs->run = post_vs_cliptest_viewport_gl;
+   }
+}
+
+
+struct pt_post_vs *draw_pt_post_vs_create( struct draw_context *draw )
+{
+   struct pt_post_vs *pvs = CALLOC_STRUCT( pt_post_vs );
+   if (!pvs)
+      return NULL;
+
+   pvs->draw = draw;
+   
+   return pvs;
+}
+
+void draw_pt_post_vs_destroy( struct pt_post_vs *pvs )
+{
+   FREE(pvs);
+}
diff --git a/src/gallium/auxiliary/draw/draw_pt_util.c b/src/gallium/auxiliary/draw/draw_pt_util.c
new file mode 100644
index 0000000000..3bc7939c55
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_util.c
@@ -0,0 +1,102 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_pt.h"
+
+void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr)
+{
+   switch (prim) {
+   case PIPE_PRIM_POINTS:
+      *first = 1;
+      *incr = 1;
+      break;
+   case PIPE_PRIM_LINES:
+      *first = 2;
+      *incr = 2;
+      break;
+   case PIPE_PRIM_LINE_STRIP:
+   case PIPE_PRIM_LINE_LOOP:
+      *first = 2;
+      *incr = 1;
+      break;
+   case PIPE_PRIM_TRIANGLES:
+      *first = 3;
+      *incr = 3;
+      break;
+   case PIPE_PRIM_TRIANGLE_STRIP:
+   case PIPE_PRIM_TRIANGLE_FAN:
+   case PIPE_PRIM_POLYGON:
+      *first = 3;
+      *incr = 1;
+      break;
+   case PIPE_PRIM_QUADS:
+      *first = 4;
+      *incr = 4;
+      break;
+   case PIPE_PRIM_QUAD_STRIP:
+      *first = 4;
+      *incr = 2;
+      break;
+   default:
+      assert(0);
+      *first = 0;
+      *incr = 1;		/* set to one so that count % incr works */
+      break;
+   }
+}
+
+
+unsigned draw_pt_reduced_prim(unsigned prim)
+{
+   switch (prim) {
+   case PIPE_PRIM_POINTS:
+      return PIPE_PRIM_POINTS;
+   case PIPE_PRIM_LINES:
+   case PIPE_PRIM_LINE_STRIP:
+   case PIPE_PRIM_LINE_LOOP:
+      return PIPE_PRIM_LINES;
+   case PIPE_PRIM_TRIANGLES:
+   case PIPE_PRIM_TRIANGLE_STRIP:
+   case PIPE_PRIM_TRIANGLE_FAN:
+   case PIPE_PRIM_POLYGON:
+   case PIPE_PRIM_QUADS:
+   case PIPE_PRIM_QUAD_STRIP:
+      return PIPE_PRIM_TRIANGLES;
+   default:
+      assert(0);
+      return PIPE_PRIM_POINTS;
+   }
+}
+
+
diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c
new file mode 100644
index 0000000000..d0e16c9bc3
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_varray.c
@@ -0,0 +1,193 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_pt.h"
+
+#define FETCH_MAX 256
+#define DRAW_MAX (FETCH_MAX+8)
+
+struct varray_frontend {
+   struct draw_pt_front_end base;
+   struct draw_context *draw;
+
+   ushort draw_elts[DRAW_MAX];
+   unsigned fetch_elts[FETCH_MAX];
+
+   unsigned driver_fetch_max;
+   unsigned fetch_max;
+
+   struct draw_pt_middle_end *middle;
+
+   unsigned input_prim;
+   unsigned output_prim;
+};
+
+
+static void varray_flush_linear(struct varray_frontend *varray,
+                                unsigned start, unsigned count)
+{
+   if (count) {
+      assert(varray->middle->run_linear);
+      varray->middle->run_linear(varray->middle, start, count);
+   }
+}
+
+static void varray_line_loop_segment(struct varray_frontend *varray,
+                                     unsigned start,
+                                     unsigned segment_start,
+                                     unsigned segment_count,
+                                     boolean end )
+{
+   assert(segment_count < varray->fetch_max);
+   if (segment_count >= 1) {
+      unsigned nr = 0, i;
+
+      for (i = 0; i < segment_count; i++) 
+         varray->fetch_elts[nr++] = start + segment_start + i;
+
+      if (end) 
+         varray->fetch_elts[nr++] = start;
+
+      assert(nr <= FETCH_MAX);
+
+      varray->middle->run(varray->middle, 
+                          varray->fetch_elts,
+                          nr,
+                          varray->draw_elts, /* ie. linear */
+                          nr);
+   }
+}
+
+
+
+static void varray_fan_segment(struct varray_frontend *varray,
+                               unsigned start, 
+                               unsigned segment_start,
+                               unsigned segment_count )
+{
+   assert(segment_count < varray->fetch_max);
+   if (segment_count >= 2) {
+      unsigned nr = 0, i;
+
+      if (segment_start != 0)
+         varray->fetch_elts[nr++] = start;
+
+      for (i = 0 ; i < segment_count; i++) 
+         varray->fetch_elts[nr++] = start + segment_start + i;
+
+      assert(nr <= FETCH_MAX);
+
+      varray->middle->run(varray->middle, 
+                          varray->fetch_elts,
+                          nr,
+                          varray->draw_elts, /* ie. linear */
+                          nr);
+   }
+}
+
+
+
+
+#define FUNC varray_run
+#include "draw_pt_varray_tmp_linear.h"
+
+static unsigned decompose_prim[PIPE_PRIM_POLYGON + 1] = {
+   PIPE_PRIM_POINTS,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_LINE_STRIP,        /* decomposed LINELOOP */
+   PIPE_PRIM_LINE_STRIP,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLE_STRIP,
+   PIPE_PRIM_TRIANGLE_FAN, 
+   PIPE_PRIM_QUADS,
+   PIPE_PRIM_QUAD_STRIP,
+   PIPE_PRIM_POLYGON
+};
+
+
+
+static void varray_prepare(struct draw_pt_front_end *frontend,
+                           unsigned prim,
+                           struct draw_pt_middle_end *middle,
+                           unsigned opt)
+{
+   struct varray_frontend *varray = (struct varray_frontend *)frontend;
+
+   varray->base.run = varray_run;
+
+   varray->input_prim = prim;
+   varray->output_prim = decompose_prim[prim];
+
+   varray->middle = middle;
+   middle->prepare(middle, varray->output_prim, opt, &varray->driver_fetch_max );
+
+   /* check that the max is even */
+   assert((varray->driver_fetch_max & 1) == 0);
+
+   varray->fetch_max = MIN2(FETCH_MAX, varray->driver_fetch_max);
+}
+
+
+
+
+static void varray_finish(struct draw_pt_front_end *frontend)
+{
+   struct varray_frontend *varray = (struct varray_frontend *)frontend;
+   varray->middle->finish(varray->middle);
+   varray->middle = NULL;
+}
+
+static void varray_destroy(struct draw_pt_front_end *frontend)
+{
+   FREE(frontend);
+}
+
+
+struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw)
+{
+   ushort i;
+   struct varray_frontend *varray = CALLOC_STRUCT(varray_frontend);
+   if (varray == NULL)
+      return NULL;
+
+   varray->base.prepare = varray_prepare;
+   varray->base.run     = NULL;
+   varray->base.finish  = varray_finish;
+   varray->base.destroy = varray_destroy;
+   varray->draw = draw;
+
+   for (i = 0; i < DRAW_MAX; i++) {
+      varray->draw_elts[i] = i;
+   }
+
+   return &varray->base;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h
new file mode 100644
index 0000000000..7c722457c3
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h
@@ -0,0 +1,238 @@
+
+static void FUNC(struct draw_pt_front_end *frontend,
+                 pt_elt_func get_elt,
+                 const void *elts,
+                 unsigned count)
+{
+   struct varray_frontend *varray = (struct varray_frontend *)frontend;
+   struct draw_context *draw = varray->draw;
+   unsigned start = (unsigned)elts;
+
+   boolean flatfirst = (draw->rasterizer->flatshade &&
+                        draw->rasterizer->flatshade_first);
+   unsigned i, j;
+   ushort flags;
+   unsigned first, incr;
+
+   varray->fetch_start = start;
+
+   draw_pt_split_prim(varray->input_prim, &first, &incr);
+
+#if 0
+   debug_printf("%s (%d) %d/%d\n", __FUNCTION__,
+                varray->input_prim,
+                start, count);
+#endif
+
+   switch (varray->input_prim) {
+   case PIPE_PRIM_POINTS:
+      for (j = 0; j + first <= count; j += i) {
+         unsigned end = MIN2(FETCH_MAX, count - j);
+         end -= (end % incr);
+         for (i = 0; i < end; i++) {
+            POINT(varray, i + 0);
+         }
+         i = end;
+         fetch_init(varray, end);
+         varray_flush(varray);
+      }
+      break;
+
+   case PIPE_PRIM_LINES:
+      for (j = 0; j + first <= count; j += i) {
+         unsigned end = MIN2(FETCH_MAX, count - j);
+         end -= (end % incr);
+         for (i = 0; i+1 < end; i += 2) {
+            LINE(varray, DRAW_PIPE_RESET_STIPPLE,
+                 i + 0, i + 1);
+         }
+         i = end;
+         fetch_init(varray, end);
+         varray_flush(varray);
+      }
+      break;
+
+   case PIPE_PRIM_LINE_LOOP:
+      if (count >= 2) {
+         flags = DRAW_PIPE_RESET_STIPPLE;
+
+         for (j = 0; j + first <= count; j += i) {
+            unsigned end = MIN2(FETCH_MAX, count - j);
+            end -= (end % incr);
+            for (i = 1; i < end; i++, flags = 0) {
+               LINE(varray, flags, i - 1, i);
+            }
+            LINE(varray, flags, i - 1, 0);
+            i = end;
+            fetch_init(varray, end);
+            varray_flush(varray);
+         }
+      }
+      break;
+
+   case PIPE_PRIM_LINE_STRIP:
+      flags = DRAW_PIPE_RESET_STIPPLE;
+      for (j = 0; j + first <= count; j += i) {
+         unsigned end = MIN2(FETCH_MAX, count - j);
+         end -= (end % incr);
+         for (i = 1; i < end; i++, flags = 0) {
+            LINE(varray, flags, i - 1, i);
+         }
+         i = end;
+         fetch_init(varray, end);
+         varray_flush(varray);
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLES:
+      for (j = 0; j + first <= count; j += i) {
+         unsigned end = MIN2(FETCH_MAX, count - j);
+         end -= (end % incr);
+         for (i = 0; i+2 < end; i += 3) {
+            TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                     i + 0, i + 1, i + 2);
+         }
+         i = end;
+         fetch_init(varray, end);
+         varray_flush(varray);
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      if (flatfirst) {
+         for (j = 0; j + first <= count; j += i) {
+            unsigned end = MIN2(FETCH_MAX, count - j);
+            end -= (end % incr);
+            for (i = 0; i+2 < end; i++) {
+               TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                        i + 0, i + 1 + (i&1), i + 2 - (i&1));
+            }
+            i = end;
+            fetch_init(varray, end);
+            varray_flush(varray);
+            if (j + first + i <= count) {
+               varray->fetch_start -= 2;
+               i -= 2;
+            }
+         }
+      }
+      else {
+         for (j = 0; j + first <= count; j += i) {
+            unsigned end = MIN2(FETCH_MAX, count - j);
+            end -= (end  % incr);
+            for (i = 0; i + 2 < end; i++) {
+               TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
+                        i + 0 + (i&1), i + 1 - (i&1), i + 2);
+            }
+            i = end;
+            fetch_init(varray, end);
+            varray_flush(varray);
+            if (j + first + i <= count) {
+               varray->fetch_start -= 2;
+               i -= 2;
+            }
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_FAN:
+      if (count >= 3) {
+         if (flatfirst) {
+            flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+            for (j = 0; j + first <= count; j += i) {
+               unsigned end = MIN2(FETCH_MAX, count - j);
+               end -= (end % incr);
+               for (i = 0; i+2 < end; i++) {
+                  TRIANGLE(varray, flags, i + 1, i + 2, 0);
+               }
+               i = end;
+               fetch_init(varray, end);
+               varray_flush(varray);
+            }
+         }
+         else {
+            flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+            for (j = 0; j + first <= count; j += i) {
+               unsigned end = MIN2(FETCH_MAX, count - j);
+               end -= (end % incr);
+               for (i = 0; i+2 < end; i++) {
+                  TRIANGLE(varray, flags, 0, i + 1, i + 2);
+               }
+               i = end;
+               fetch_init(varray, end);
+               varray_flush(varray);
+            }
+         }
+      }
+      break;
+
+   case PIPE_PRIM_QUADS:
+      for (j = 0; j + first <= count; j += i) {
+         unsigned end = MIN2(FETCH_MAX, count - j);
+         end -= (end % incr);
+         for (i = 0; i+3 < end; i += 4) {
+            QUAD(varray, i + 0, i + 1, i + 2, i + 3);
+         }
+         i = end;
+         fetch_init(varray, end);
+         varray_flush(varray);
+      }
+      break;
+
+   case PIPE_PRIM_QUAD_STRIP:
+      for (j = 0; j + first <= count; j += i) {
+         unsigned end = MIN2(FETCH_MAX, count - j);
+         end -= (end % incr);
+         for (i = 0; i+3 < end; i += 2) {
+            QUAD(varray, i + 2, i + 0, i + 1, i + 3);
+         }
+         i = end;
+         fetch_init(varray, end);
+         varray_flush(varray);
+         if (j + first + i <= count) {
+            varray->fetch_start -= 2;
+            i -= 2;
+         }
+      }
+      break;
+
+   case PIPE_PRIM_POLYGON:
+   {
+      /* These bitflags look a little odd because we submit the
+       * vertices as (1,2,0) to satisfy flatshade requirements.
+       */
+      const ushort edge_first  = DRAW_PIPE_EDGE_FLAG_2;
+      const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0;
+      const ushort edge_last   = DRAW_PIPE_EDGE_FLAG_1;
+
+      flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle;
+      for (j = 0; j + first <= count; j += i) {
+         unsigned end = MIN2(FETCH_MAX, count - j);
+         end -= (end % incr);
+         for (i = 0; i+2 < end; i++, flags = edge_middle) {
+
+            if (i + 3 == count)
+               flags |= edge_last;
+
+            TRIANGLE(varray, flags, i + 1, i + 2, 0);
+         }
+         i = end;
+         fetch_init(varray, end);
+         varray_flush(varray);
+      }
+   }
+   break;
+
+   default:
+      assert(0);
+      break;
+   }
+
+   varray_flush(varray);
+}
+
+#undef TRIANGLE
+#undef QUAD
+#undef POINT
+#undef LINE
+#undef FUNC
diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
new file mode 100644
index 0000000000..55a8e6521d
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
@@ -0,0 +1,91 @@
+static unsigned trim( unsigned count, unsigned first, unsigned incr )
+{
+   return count - (count - first) % incr; 
+}
+
+static void FUNC(struct draw_pt_front_end *frontend,
+                 pt_elt_func get_elt,
+                 const void *elts,
+                 unsigned count)
+{
+   struct varray_frontend *varray = (struct varray_frontend *)frontend;
+   unsigned start = (unsigned)elts;
+
+   unsigned j;
+   unsigned first, incr;
+
+   draw_pt_split_prim(varray->input_prim, &first, &incr);
+   
+   /* Sanitize primitive length:
+    */
+   count = trim(count, first, incr); 
+   if (count < first)
+      return;
+
+#if 0
+   debug_printf("%s (%d) %d/%d\n", __FUNCTION__,
+                varray->input_prim,
+                start, count);
+#endif
+
+   switch (varray->input_prim) {
+   case PIPE_PRIM_POINTS:
+   case PIPE_PRIM_LINES:
+   case PIPE_PRIM_TRIANGLES:
+   case PIPE_PRIM_LINE_STRIP:
+   case PIPE_PRIM_TRIANGLE_STRIP:
+   case PIPE_PRIM_QUADS:
+   case PIPE_PRIM_QUAD_STRIP:
+      for (j = 0; j < count;) {
+         unsigned remaining = count - j;
+         unsigned nr = trim( MIN2(varray->driver_fetch_max, remaining), first, incr );
+         varray_flush_linear(varray, start + j, nr);
+         j += nr;
+         if (nr != remaining) 
+            j -= (first - incr);
+      }
+      break;
+
+   case PIPE_PRIM_LINE_LOOP:
+      /* Always have to decompose as we've stated that this will be
+       * emitted as a line-strip.
+       */
+      for (j = 0; j < count;) {
+         unsigned remaining = count - j;
+         unsigned nr = trim( MIN2(varray->fetch_max-1, remaining), first, incr );
+         varray_line_loop_segment(varray, start, j, nr, nr == remaining);
+         j += nr;
+         if (nr != remaining) 
+            j -= (first - incr);
+      }
+      break;
+
+
+   case PIPE_PRIM_POLYGON:
+   case PIPE_PRIM_TRIANGLE_FAN: 
+      if (count < varray->driver_fetch_max) {
+         varray_flush_linear(varray, start, count);
+      }
+      else {
+         for ( j = 0; j < count;) {
+            unsigned remaining = count - j;
+            unsigned nr = trim( MIN2(varray->fetch_max-1, remaining), first, incr );
+            varray_fan_segment(varray, start, j, nr);
+            j += nr;
+            if (nr != remaining) 
+               j -= (first - incr);
+         }
+      }
+      break;
+
+   default:
+      assert(0);
+      break;
+   }
+}
+
+#undef TRIANGLE
+#undef QUAD
+#undef POINT
+#undef LINE
+#undef FUNC
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c
new file mode 100644
index 0000000000..5d268a2226
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c
@@ -0,0 +1,501 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "util/u_memory.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_pt.h"
+
+
+#define CACHE_MAX 256
+#define FETCH_MAX 256
+#define DRAW_MAX (16*1024)
+
+struct vcache_frontend {
+   struct draw_pt_front_end base;
+   struct draw_context *draw;
+
+   unsigned in[CACHE_MAX];
+   ushort out[CACHE_MAX];
+
+   ushort draw_elts[DRAW_MAX];
+   unsigned fetch_elts[FETCH_MAX];
+
+   unsigned draw_count;
+   unsigned fetch_count;
+   unsigned fetch_max;
+   
+   struct draw_pt_middle_end *middle;
+
+   unsigned input_prim;
+   unsigned output_prim;
+
+   unsigned middle_prim;
+   unsigned opt;
+};
+
+static INLINE void 
+vcache_flush( struct vcache_frontend *vcache )
+{
+   if (vcache->middle_prim != vcache->output_prim) {
+      vcache->middle_prim = vcache->output_prim;
+      vcache->middle->prepare( vcache->middle, 
+                               vcache->middle_prim, 
+                               vcache->opt, 
+                               &vcache->fetch_max );
+   }
+
+   if (vcache->draw_count) {
+      vcache->middle->run( vcache->middle,
+                           vcache->fetch_elts,
+                           vcache->fetch_count,
+                           vcache->draw_elts,
+                           vcache->draw_count );
+   }
+
+   memset(vcache->in, ~0, sizeof(vcache->in));
+   vcache->fetch_count = 0;
+   vcache->draw_count = 0;
+}
+
+static INLINE void 
+vcache_check_flush( struct vcache_frontend *vcache )
+{
+   if ( vcache->draw_count + 6 >= DRAW_MAX ||
+        vcache->fetch_count + 4 >= FETCH_MAX )
+   {
+      vcache_flush( vcache );
+   }
+}
+
+
+static INLINE void 
+vcache_elt( struct vcache_frontend *vcache,
+            unsigned felt,
+            ushort flags )
+{
+   unsigned idx = felt % CACHE_MAX;
+
+   if (vcache->in[idx] != felt) {
+      assert(vcache->fetch_count < FETCH_MAX);
+
+      vcache->in[idx] = felt;
+      vcache->out[idx] = (ushort)vcache->fetch_count;
+      vcache->fetch_elts[vcache->fetch_count++] = felt;
+   }
+
+   vcache->draw_elts[vcache->draw_count++] = vcache->out[idx] | flags;
+}
+
+
+                   
+static INLINE void 
+vcache_triangle( struct vcache_frontend *vcache,
+                 unsigned i0,
+                 unsigned i1,
+                 unsigned i2 )
+{
+   vcache_elt(vcache, i0, 0);
+   vcache_elt(vcache, i1, 0);
+   vcache_elt(vcache, i2, 0);
+   vcache_check_flush(vcache);
+}
+
+			  
+static INLINE void 
+vcache_triangle_flags( struct vcache_frontend *vcache,
+                       ushort flags,
+                       unsigned i0,
+                       unsigned i1,
+                       unsigned i2 )
+{
+   vcache_elt(vcache, i0, flags);
+   vcache_elt(vcache, i1, 0);
+   vcache_elt(vcache, i2, 0);
+   vcache_check_flush(vcache);
+}
+
+static INLINE void 
+vcache_line( struct vcache_frontend *vcache,
+             unsigned i0,
+             unsigned i1 )
+{
+   vcache_elt(vcache, i0, 0);
+   vcache_elt(vcache, i1, 0);
+   vcache_check_flush(vcache);
+}
+
+
+static INLINE void 
+vcache_line_flags( struct vcache_frontend *vcache,
+                   ushort flags,
+                   unsigned i0,
+                   unsigned i1 )
+{
+   vcache_elt(vcache, i0, flags);
+   vcache_elt(vcache, i1, 0);
+   vcache_check_flush(vcache);
+}
+
+
+static INLINE void 
+vcache_point( struct vcache_frontend *vcache,
+              unsigned i0 )
+{
+   vcache_elt(vcache, i0, 0);
+   vcache_check_flush(vcache);
+}
+
+static INLINE void 
+vcache_quad( struct vcache_frontend *vcache,
+             unsigned i0,
+             unsigned i1,
+             unsigned i2,
+             unsigned i3 )
+{
+   vcache_triangle( vcache, i0, i1, i3 );
+   vcache_triangle( vcache, i1, i2, i3 );
+}
+
+static INLINE void 
+vcache_ef_quad( struct vcache_frontend *vcache,
+                unsigned i0,
+                unsigned i1,
+                unsigned i2,
+                unsigned i3 )
+{
+   vcache_triangle_flags( vcache,
+                          ( DRAW_PIPE_RESET_STIPPLE |
+                            DRAW_PIPE_EDGE_FLAG_0 |
+                            DRAW_PIPE_EDGE_FLAG_2 ),
+                          i0, i1, i3 );
+
+   vcache_triangle_flags( vcache,
+                          ( DRAW_PIPE_EDGE_FLAG_0 |
+                            DRAW_PIPE_EDGE_FLAG_1 ),
+                          i1, i2, i3 );
+}
+
+/* At least for now, we're back to using a template include file for
+ * this.  The two paths aren't too different though - it may be
+ * possible to reunify them.
+ */
+#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle_flags(vc,flags,i0,i1,i2)
+#define QUAD(vc,i0,i1,i2,i3)        vcache_ef_quad(vc,i0,i1,i2,i3)
+#define LINE(vc,flags,i0,i1)        vcache_line_flags(vc,flags,i0,i1)
+#define POINT(vc,i0)                vcache_point(vc,i0)
+#define FUNC vcache_run_extras
+#include "draw_pt_vcache_tmp.h"
+
+#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle(vc,i0,i1,i2)
+#define QUAD(vc,i0,i1,i2,i3)        vcache_quad(vc,i0,i1,i2,i3)
+#define LINE(vc,flags,i0,i1)        vcache_line(vc,i0,i1)
+#define POINT(vc,i0)                vcache_point(vc,i0)
+#define FUNC vcache_run
+#include "draw_pt_vcache_tmp.h"
+
+static INLINE void 
+rebase_uint_elts( const unsigned *src,
+                  unsigned count,
+                  int delta,
+                  ushort *dest )
+{
+   unsigned i;
+
+   for (i = 0; i < count; i++) 
+      dest[i] = (ushort)(src[i] + delta);
+}
+
+static INLINE void 
+rebase_ushort_elts( const ushort *src,
+                    unsigned count,
+                    int delta,
+                                ushort *dest )
+{
+   unsigned i;
+
+   for (i = 0; i < count; i++) 
+      dest[i] = (ushort)(src[i] + delta);
+}
+
+static INLINE void 
+rebase_ubyte_elts( const ubyte *src,
+                   unsigned count,
+                   int delta,
+                   ushort *dest )
+{
+   unsigned i;
+
+   for (i = 0; i < count; i++) 
+      dest[i] = (ushort)(src[i] + delta);
+}
+
+
+
+static INLINE void 
+translate_uint_elts( const unsigned *src,
+                     unsigned count,
+                     ushort *dest )
+{
+   unsigned i;
+
+   for (i = 0; i < count; i++) 
+      dest[i] = (ushort)(src[i]);
+}
+
+static INLINE void 
+translate_ushort_elts( const ushort *src,
+                       unsigned count,
+                       ushort *dest )
+{
+   unsigned i;
+
+   for (i = 0; i < count; i++) 
+      dest[i] = (ushort)(src[i]);
+}
+
+static INLINE void 
+translate_ubyte_elts( const ubyte *src,
+                      unsigned count,
+                      ushort *dest )
+{
+   unsigned i;
+
+   for (i = 0; i < count; i++) 
+      dest[i] = (ushort)(src[i]);
+}
+
+
+
+
+#if 0
+static INLINE enum pipe_format 
+format_from_get_elt( pt_elt_func get_elt )
+{
+   switch (draw->pt.user.eltSize) {
+   case 1: return PIPE_FORMAT_R8_UNORM;
+   case 2: return PIPE_FORMAT_R16_UNORM;
+   case 4: return PIPE_FORMAT_R32_UNORM;
+   default: return PIPE_FORMAT_NONE;
+   }
+}
+#endif
+
+static INLINE void 
+vcache_check_run( struct draw_pt_front_end *frontend, 
+                  pt_elt_func get_elt,
+                  const void *elts,
+                  unsigned draw_count )
+{
+   struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; 
+   struct draw_context *draw = vcache->draw;
+   unsigned min_index = draw->pt.user.min_index;
+   unsigned max_index = draw->pt.user.max_index;
+   unsigned index_size = draw->pt.user.eltSize;
+   unsigned fetch_count = max_index + 1 - min_index;
+   const ushort *transformed_elts;
+   ushort *storage = NULL;
+   boolean ok = FALSE;
+
+
+   if (0) debug_printf("fetch_count %d fetch_max %d draw_count %d\n", fetch_count, 
+                       vcache->fetch_max,
+                       draw_count);
+      
+   if (max_index == 0xffffffff ||
+       fetch_count > draw_count) {
+      if (0) debug_printf("fail\n");
+      goto fail;
+   }
+      
+   if (vcache->middle_prim != vcache->input_prim) {
+      vcache->middle_prim = vcache->input_prim;
+      vcache->middle->prepare( vcache->middle, 
+                               vcache->middle_prim, 
+                               vcache->opt, 
+                               &vcache->fetch_max );
+   }
+
+
+   if (min_index == 0 &&
+       index_size == 2) 
+   {
+      transformed_elts = (const ushort *)elts;
+   }
+   else 
+   {
+      storage = MALLOC( draw_count * sizeof(ushort) );
+      if (!storage)
+         goto fail;
+      
+      if (min_index == 0) {
+         switch(index_size) {
+         case 1:
+            translate_ubyte_elts( (const ubyte *)elts,
+                                  draw_count,
+                                  storage );
+            break;
+
+         case 2:
+            translate_ushort_elts( (const ushort *)elts,
+                                   draw_count,
+                                   storage );
+            break;
+
+         case 4:
+            translate_uint_elts( (const uint *)elts,
+                                 draw_count,
+                                 storage );
+            break;
+
+         default:
+            assert(0);
+            return;
+         }
+      }
+      else {
+         switch(index_size) {
+         case 1:
+            rebase_ubyte_elts( (const ubyte *)elts,
+                                  draw_count,
+                                  0 - (int)min_index,
+                                  storage );
+            break;
+
+         case 2:
+            rebase_ushort_elts( (const ushort *)elts,
+                                   draw_count,
+                                   0 - (int)min_index,
+                                   storage );
+            break;
+
+         case 4:
+            rebase_uint_elts( (const uint *)elts,
+                                 draw_count,
+                                 0 - (int)min_index,
+                                 storage );
+            break;
+
+         default:
+            assert(0);
+            return;
+         }
+      }
+      transformed_elts = storage;
+   }
+
+   if (fetch_count < UNDEFINED_VERTEX_ID)
+      ok = vcache->middle->run_linear_elts( vcache->middle,
+                                            min_index, /* start */
+                                            fetch_count,
+                                            transformed_elts,
+                                            draw_count );
+   
+   FREE(storage);
+
+   if (ok)
+      return;
+
+   debug_printf("failed to execute atomic draw elts for %d/%d, splitting up\n",
+                fetch_count, draw_count);
+
+ fail:
+   vcache_run( frontend, get_elt, elts, draw_count );
+}
+
+
+
+
+static void 
+vcache_prepare( struct draw_pt_front_end *frontend,
+                unsigned prim,
+                struct draw_pt_middle_end *middle,
+                unsigned opt )
+{
+   struct vcache_frontend *vcache = (struct vcache_frontend *)frontend;
+
+   if (opt & PT_PIPELINE)
+   {
+      vcache->base.run = vcache_run_extras;
+   }
+   else 
+   {
+      vcache->base.run = vcache_check_run;
+   }
+
+   vcache->input_prim = prim;
+   vcache->output_prim = draw_pt_reduced_prim(prim);
+
+   vcache->middle = middle;
+   vcache->opt = opt;
+
+   /* Have to run prepare here, but try and guess a good prim for
+    * doing so:
+    */
+   vcache->middle_prim = (opt & PT_PIPELINE) ? vcache->output_prim : vcache->input_prim;
+   middle->prepare( middle, vcache->middle_prim, opt, &vcache->fetch_max );
+}
+
+
+
+
+static void 
+vcache_finish( struct draw_pt_front_end *frontend )
+{
+   struct vcache_frontend *vcache = (struct vcache_frontend *)frontend;
+   vcache->middle->finish( vcache->middle );
+   vcache->middle = NULL;
+}
+
+static void 
+vcache_destroy( struct draw_pt_front_end *frontend )
+{
+   FREE(frontend);
+}
+
+
+struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw )
+{
+   struct vcache_frontend *vcache = CALLOC_STRUCT( vcache_frontend );
+   if (vcache == NULL)
+      return NULL;
+ 
+   vcache->base.prepare = vcache_prepare;
+   vcache->base.run     = NULL;
+   vcache->base.finish  = vcache_finish;
+   vcache->base.destroy = vcache_destroy;
+   vcache->draw = draw;
+   
+   memset(vcache->in, ~0, sizeof(vcache->in));
+  
+   return &vcache->base;
+}
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
new file mode 100644
index 0000000000..ec05bbeab4
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
@@ -0,0 +1,177 @@
+
+
+static void FUNC( struct draw_pt_front_end *frontend, 
+                  pt_elt_func get_elt,
+                  const void *elts,
+                  unsigned count )
+{
+   struct vcache_frontend *vcache = (struct vcache_frontend *)frontend;
+   struct draw_context *draw = vcache->draw;
+
+   boolean flatfirst = (draw->rasterizer->flatshade && 
+                        draw->rasterizer->flatshade_first);
+   unsigned i;
+   ushort flags;
+
+   if (0) debug_printf("%s %d\n", __FUNCTION__, count);
+
+
+   switch (vcache->input_prim) {
+   case PIPE_PRIM_POINTS:
+      for (i = 0; i < count; i ++) {
+	 POINT( vcache,
+                get_elt(elts, i + 0) );
+      }
+      break;
+
+   case PIPE_PRIM_LINES:
+      for (i = 0; i+1 < count; i += 2) {
+         LINE( vcache, 
+               DRAW_PIPE_RESET_STIPPLE,
+               get_elt(elts, i + 0),
+               get_elt(elts, i + 1));
+      }
+      break;
+
+   case PIPE_PRIM_LINE_LOOP:  
+      if (count >= 2) {
+         flags = DRAW_PIPE_RESET_STIPPLE;
+
+         for (i = 1; i < count; i++, flags = 0) {
+            LINE( vcache, 
+                  flags,
+                  get_elt(elts, i - 1),
+                  get_elt(elts, i ));
+         }
+
+	 LINE( vcache, 
+               flags,
+               get_elt(elts, i - 1),
+               get_elt(elts, 0 ));
+      }
+      break;
+
+   case PIPE_PRIM_LINE_STRIP:
+      flags = DRAW_PIPE_RESET_STIPPLE;
+      for (i = 1; i < count; i++, flags = 0) {
+         LINE( vcache, 
+               flags,
+               get_elt(elts, i - 1),
+               get_elt(elts, i ));
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLES:
+      for (i = 0; i+2 < count; i += 3) {
+         TRIANGLE( vcache,
+                   DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
+                   get_elt(elts, i + 0),
+                   get_elt(elts, i + 1),
+                   get_elt(elts, i + 2 ));
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      if (flatfirst) {
+         for (i = 0; i+2 < count; i++) {
+            TRIANGLE( vcache,
+                      DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
+                      get_elt(elts, i + 0),
+                      get_elt(elts, i + 1 + (i&1)),
+                      get_elt(elts, i + 2 - (i&1)));
+         }
+      }
+      else {
+         for (i = 0; i+2 < count; i++) {
+            TRIANGLE( vcache,
+                      DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
+                      get_elt(elts, i + 0 + (i&1)),
+                      get_elt(elts, i + 1 - (i&1)),
+                      get_elt(elts, i + 2 ));
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_FAN:
+      if (count >= 3) {
+         if (flatfirst) {
+            for (i = 0; i+2 < count; i++) {
+               TRIANGLE( vcache,
+                         DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
+                         get_elt(elts, i + 1),
+                         get_elt(elts, i + 2),
+                         get_elt(elts, 0 ));
+            }
+         }
+         else {
+            for (i = 0; i+2 < count; i++) {
+               TRIANGLE( vcache,
+                         DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
+                         get_elt(elts, 0),
+                         get_elt(elts, i + 1),
+                         get_elt(elts, i + 2 ));
+            }
+         }
+      }
+      break;
+
+
+   case PIPE_PRIM_QUADS:
+      for (i = 0; i+3 < count; i += 4) {
+         QUAD( vcache,
+               get_elt(elts, i + 0),
+               get_elt(elts, i + 1),
+               get_elt(elts, i + 2),
+               get_elt(elts, i + 3));
+      }
+      break;
+
+   case PIPE_PRIM_QUAD_STRIP:
+      for (i = 0; i+3 < count; i += 2) {
+         QUAD( vcache,
+               get_elt(elts, i + 2),
+               get_elt(elts, i + 0),
+               get_elt(elts, i + 1),
+               get_elt(elts, i + 3));
+      }
+      break;
+
+   case PIPE_PRIM_POLYGON:
+      {
+         /* These bitflags look a little odd because we submit the
+          * vertices as (1,2,0) to satisfy flatshade requirements.  
+          */
+         const ushort edge_first  = DRAW_PIPE_EDGE_FLAG_2;
+         const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0;
+         const ushort edge_last   = DRAW_PIPE_EDGE_FLAG_1;
+
+         flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle;
+
+	 for (i = 0; i+2 < count; i++, flags = edge_middle) {
+
+            if (i + 3 == count)
+               flags |= edge_last;
+
+	    TRIANGLE( vcache,
+                      flags,
+                      get_elt(elts, i + 1),
+                      get_elt(elts, i + 2),
+                      get_elt(elts, 0));
+	 }
+      }
+      break;
+
+   default:
+      assert(0);
+      break;
+   }
+   
+   vcache_flush( vcache );
+}
+
+
+#undef TRIANGLE
+#undef QUAD
+#undef POINT
+#undef LINE
+#undef FUNC
diff --git a/src/gallium/auxiliary/draw/draw_vbuf.h b/src/gallium/auxiliary/draw/draw_vbuf.h
new file mode 100644
index 0000000000..a1c4c14445
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vbuf.h
@@ -0,0 +1,127 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Vertex buffer drawing stage.
+ * 
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ * \author Jose Fonseca <jrfonsec@tungstengraphics.com>
+ */
+
+#ifndef DRAW_VBUF_H_
+#define DRAW_VBUF_H_
+
+
+#include "pipe/p_compiler.h"
+
+
+struct pipe_rasterizer_state;
+struct draw_context;
+struct vertex_info;
+
+
+/**
+ * Interface for hardware vertex buffer rendering.
+ */
+struct vbuf_render {
+
+   /**
+    * Driver limits.  May be tuned lower to improve cache hits on
+    * index list.
+    */
+   unsigned max_indices;
+   unsigned max_vertex_buffer_bytes;
+
+   /**
+    * Query if the hardware driver needs assistance for a particular
+    * combination of rasterizer state and primitive.
+    *
+    * Currently optional.
+    */
+   boolean (*need_pipeline)(const struct vbuf_render *render,
+                            const struct pipe_rasterizer_state *rasterizer,
+                            unsigned int prim );
+
+
+   /**
+    * Get the hardware vertex format.
+    * 
+    * XXX: have this in draw_context instead?
+    */ 
+   const struct vertex_info *(*get_vertex_info)( struct vbuf_render * );
+	 
+   /**
+    * Request a destination for vertices.
+    * Hardware renderers will use ttm memory, others will just malloc
+    * something.
+    */
+   void *(*allocate_vertices)( struct vbuf_render *,
+			       ushort vertex_size,
+			       ushort nr_vertices );
+
+   /**
+    * Notify the renderer of the current primitive when it changes.
+    * Must succeed for TRIANGLES, LINES and POINTS.  Other prims at
+    * the discretion of the driver, for the benefit of the passthrough
+    * path.
+    */
+   boolean (*set_primitive)( struct vbuf_render *, unsigned prim );
+
+   /**
+    * DrawElements, note indices are ushort.  The driver must complete
+    * this call, if necessary splitting the index list itself.
+    */
+   void (*draw)( struct vbuf_render *,
+		 const ushort *indices,
+		 uint nr_indices );
+
+   /* Draw Arrays path too.
+    */
+   void (*draw_arrays)( struct vbuf_render *,
+			unsigned start,
+			uint nr );
+
+   /**
+    * Called when vbuf is done with this set of vertices:
+    */
+   void (*release_vertices)( struct vbuf_render *,
+			     void *vertices, 
+			     unsigned vertex_size,
+			     unsigned vertices_used );
+
+   void (*destroy)( struct vbuf_render * );
+};
+
+
+
+struct draw_stage *
+draw_vbuf_stage( struct draw_context *draw,
+                 struct vbuf_render *render );
+
+
+#endif /*DRAW_VBUF_H_*/
diff --git a/src/gallium/auxiliary/draw/draw_vertex.c b/src/gallium/auxiliary/draw/draw_vertex.c
new file mode 100644
index 0000000000..3214213e44
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vertex.c
@@ -0,0 +1,129 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/*
+ * Functions for specifying the post-transformation vertex layout.
+ *
+ * Author:
+ *    Brian Paul
+ *    Keith Whitwell
+ */
+
+
+#include "draw/draw_private.h"
+#include "draw/draw_vertex.h"
+
+
+/**
+ * Compute the size of a vertex, in dwords/floats, to update the
+ * vinfo->size field.
+ */
+void
+draw_compute_vertex_size(struct vertex_info *vinfo)
+{
+   uint i;
+
+   vinfo->size = 0;
+   for (i = 0; i < vinfo->num_attribs; i++) {
+      switch (vinfo->attrib[i].emit) {
+      case EMIT_OMIT:
+         break;
+      case EMIT_4UB:
+         /* fall-through */
+      case EMIT_1F_PSIZE:
+         /* fall-through */
+      case EMIT_1F:
+         vinfo->size += 1;
+         break;
+      case EMIT_2F:
+         vinfo->size += 2;
+         break;
+      case EMIT_3F:
+         vinfo->size += 3;
+         break;
+      case EMIT_4F:
+         vinfo->size += 4;
+         break;
+      default:
+         assert(0);
+      }
+   }
+}
+
+
+void
+draw_dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data)
+{
+   unsigned i, j;
+
+   for (i = 0; i < vinfo->num_attribs; i++) {
+      j = vinfo->attrib[i].src_index;
+      switch (vinfo->attrib[i].emit) {
+      case EMIT_OMIT:
+         debug_printf("EMIT_OMIT:");
+         break;
+      case EMIT_1F:
+         debug_printf("EMIT_1F:\t");
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         break;
+      case EMIT_1F_PSIZE:
+         debug_printf("EMIT_1F_PSIZE:\t");
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         break;
+      case EMIT_2F:
+         debug_printf("EMIT_2F:\t");
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         break;
+      case EMIT_3F:
+         debug_printf("EMIT_3F:\t");
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         data += sizeof(float);
+         break;
+      case EMIT_4F:
+         debug_printf("EMIT_4F:\t");
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         debug_printf("%f ", *(float *)data); data += sizeof(float);
+         break;
+      case EMIT_4UB:
+         debug_printf("EMIT_4UB:\t");
+         debug_printf("%u ", *data++);
+         debug_printf("%u ", *data++);
+         debug_printf("%u ", *data++);
+         debug_printf("%u ", *data++);
+         break;
+      default:
+         assert(0);
+      }
+      debug_printf("\n");
+   }
+   debug_printf("\n");
+}
diff --git a/src/gallium/auxiliary/draw/draw_vertex.h b/src/gallium/auxiliary/draw/draw_vertex.h
new file mode 100644
index 0000000000..c143cf2372
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vertex.h
@@ -0,0 +1,163 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Post-transform vertex format info.  The vertex_info struct is used by
+ * the draw_vbuf code to emit hardware-specific vertex layouts into hw
+ * vertex buffers.
+ *
+ * Author:
+ *    Brian Paul
+ */
+
+
+#ifndef DRAW_VERTEX_H
+#define DRAW_VERTEX_H
+
+
+#include "pipe/p_state.h"
+
+
+/**
+ * Vertex attribute emit modes
+ */
+enum attrib_emit {
+   EMIT_OMIT,      /**< don't emit the attribute */
+   EMIT_1F,
+   EMIT_1F_PSIZE,  /**< insert constant point size */
+   EMIT_2F,
+   EMIT_3F,
+   EMIT_4F,
+   EMIT_4UB  /**< XXX may need variations for RGBA vs BGRA, etc */
+};
+
+
+/**
+ * Attribute interpolation mode
+ */
+enum interp_mode {
+   INTERP_NONE,      /**< never interpolate vertex header info */
+   INTERP_POS,       /**< special case for frag position */
+   INTERP_CONSTANT,
+   INTERP_LINEAR,
+   INTERP_PERSPECTIVE
+};
+
+
+/**
+ * Information about hardware/rasterization vertex layout.
+ */
+struct vertex_info
+{
+   uint num_attribs;
+   uint hwfmt[4];      /**< hardware format info for this format */
+   uint size;          /**< total vertex size in dwords */
+   
+   /* Keep this small and at the end of the struct to allow quick
+    * memcmp() comparisons.
+    */
+   struct {
+      unsigned interp_mode:4;      /**< INTERP_x */
+      unsigned emit:4;             /**< EMIT_x */
+      unsigned src_index:8;          /**< map to post-xform attribs */
+   } attrib[PIPE_MAX_SHADER_INPUTS];
+};
+
+static INLINE int
+draw_vinfo_size( const struct vertex_info *a )
+{
+   return ((const char *)&a->attrib[a->num_attribs] -
+           (const char *)a);
+}
+
+static INLINE int
+draw_vinfo_compare( const struct vertex_info *a,
+                    const struct vertex_info *b )
+{
+   unsigned sizea = draw_vinfo_size( a );
+   return memcmp( a, b, sizea );
+}
+
+static INLINE void
+draw_vinfo_copy( struct vertex_info *dst,
+                 const struct vertex_info *src )
+{
+   unsigned size = draw_vinfo_size( src );
+   memcpy( dst, src, size );
+}
+
+
+
+/**
+ * Add another attribute to the given vertex_info object.
+ * \param src_index  indicates which post-transformed vertex attrib slot
+ *                   corresponds to this attribute.
+ * \return slot in which the attribute was added
+ */
+static INLINE uint
+draw_emit_vertex_attr(struct vertex_info *vinfo,
+                      enum attrib_emit emit, 
+                      enum interp_mode interp, /* only used by softpipe??? */
+                      uint src_index)
+{
+   const uint n = vinfo->num_attribs;
+   assert(n < PIPE_MAX_SHADER_INPUTS);
+   vinfo->attrib[n].emit = emit;
+   vinfo->attrib[n].interp_mode = interp;
+   vinfo->attrib[n].src_index = src_index;
+   vinfo->num_attribs++;
+   return n;
+}
+
+
+extern void draw_compute_vertex_size(struct vertex_info *vinfo);
+
+void draw_dump_emitted_vertex(const struct vertex_info *vinfo, 
+                              const uint8_t *data);
+
+
+static INLINE unsigned draw_translate_vinfo_format(unsigned format )
+{
+   switch (format) {
+   case EMIT_1F:
+   case EMIT_1F_PSIZE:
+      return PIPE_FORMAT_R32_FLOAT;
+   case EMIT_2F:
+      return PIPE_FORMAT_R32G32_FLOAT;
+   case EMIT_3F:
+      return PIPE_FORMAT_R32G32B32_FLOAT;
+   case EMIT_4F:
+      return PIPE_FORMAT_R32G32B32A32_FLOAT;
+   case EMIT_4UB:
+      return PIPE_FORMAT_R8G8B8A8_UNORM;
+   default:
+      return PIPE_FORMAT_NONE;
+   }
+}
+
+
+#endif /* DRAW_VERTEX_H */
diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c
new file mode 100644
index 0000000000..c057cd67fd
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs.c
@@ -0,0 +1,267 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "pipe/p_shader_tokens.h"
+
+#include "draw_private.h"
+#include "draw_context.h"
+#include "draw_vs.h"
+
+#include "translate/translate.h"
+#include "translate/translate_cache.h"
+
+
+
+
+void draw_vs_set_constants( struct draw_context *draw,
+                            const float (*constants)[4],
+                            unsigned size )
+{
+   if (((uintptr_t)constants) & 0xf) {
+      if (size > draw->vs.const_storage_size) {
+         if (draw->vs.aligned_constant_storage)
+            align_free((void *)draw->vs.aligned_constant_storage);
+         draw->vs.aligned_constant_storage = align_malloc( size, 16 );
+      }
+      memcpy( (void*)draw->vs.aligned_constant_storage,
+              constants, 
+              size );
+      constants = draw->vs.aligned_constant_storage;
+   }
+      
+   draw->vs.aligned_constants = constants;
+   draw_vs_aos_machine_constants( draw->vs.aos_machine, constants );
+}
+
+
+void draw_vs_set_viewport( struct draw_context *draw,
+                           const struct pipe_viewport_state *viewport )
+{
+   draw_vs_aos_machine_viewport( draw->vs.aos_machine, viewport );
+}
+
+
+
+struct draw_vertex_shader *
+draw_create_vertex_shader(struct draw_context *draw,
+                          const struct pipe_shader_state *shader)
+{
+   struct draw_vertex_shader *vs;
+
+   vs = draw_create_vs_llvm( draw, shader );
+   if (!vs) {
+      vs = draw_create_vs_sse( draw, shader );
+      if (!vs) {
+         vs = draw_create_vs_ppc( draw, shader );
+         if (!vs) {
+            vs = draw_create_vs_exec( draw, shader );
+         }
+      }
+   }
+
+   if (vs)
+   {
+      uint i;
+      for (i = 0; i < vs->info.num_outputs; i++) {
+         if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_POSITION &&
+             vs->info.output_semantic_index[i] == 0)
+            vs->position_output = i;
+      }
+   }
+
+   assert(vs);
+   return vs;
+}
+
+
+void
+draw_bind_vertex_shader(struct draw_context *draw,
+                        struct draw_vertex_shader *dvs)
+{
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+   
+   if (dvs) 
+   {
+      draw->vs.vertex_shader = dvs;
+      draw->vs.num_vs_outputs = dvs->info.num_outputs;
+      draw->vs.position_output = dvs->position_output;
+      dvs->prepare( dvs, draw );
+   }
+   else {
+      draw->vs.vertex_shader = NULL;
+      draw->vs.num_vs_outputs = 0;
+   }
+}
+
+
+void
+draw_delete_vertex_shader(struct draw_context *draw,
+                          struct draw_vertex_shader *dvs)
+{
+   unsigned i;
+
+   for (i = 0; i < dvs->nr_varients; i++) 
+      dvs->varient[i]->destroy( dvs->varient[i] );
+
+   dvs->nr_varients = 0;
+
+   dvs->delete( dvs );
+}
+
+
+
+boolean 
+draw_vs_init( struct draw_context *draw )
+{
+   tgsi_exec_machine_init(&draw->vs.machine);
+
+   /* FIXME: give this machine thing a proper constructor:
+    */
+   draw->vs.machine.Inputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16);
+   if (!draw->vs.machine.Inputs)
+      return FALSE;
+
+   draw->vs.machine.Outputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16);
+   if (!draw->vs.machine.Outputs)
+      return FALSE;
+
+   draw->vs.emit_cache = translate_cache_create();
+   if (!draw->vs.emit_cache) 
+      return FALSE;
+      
+   draw->vs.fetch_cache = translate_cache_create();
+   if (!draw->vs.fetch_cache) 
+      return FALSE;
+
+   draw->vs.aos_machine = draw_vs_aos_machine();
+#ifdef PIPE_ARCH_X86
+   if (!draw->vs.aos_machine)
+      return FALSE;
+#endif
+      
+   return TRUE;
+}
+
+void
+draw_vs_destroy( struct draw_context *draw )
+{
+   if (draw->vs.machine.Inputs)
+      align_free(draw->vs.machine.Inputs);
+
+   if (draw->vs.machine.Outputs)
+      align_free(draw->vs.machine.Outputs);
+
+   if (draw->vs.fetch_cache)
+      translate_cache_destroy(draw->vs.fetch_cache);
+
+   if (draw->vs.emit_cache)
+      translate_cache_destroy(draw->vs.emit_cache);
+
+   if (draw->vs.aos_machine)
+      draw_vs_aos_machine_destroy(draw->vs.aos_machine);
+
+   if (draw->vs.aligned_constant_storage)
+      align_free((void*)draw->vs.aligned_constant_storage);
+
+   tgsi_exec_machine_free_data(&draw->vs.machine);
+
+}
+
+
+struct draw_vs_varient *
+draw_vs_lookup_varient( struct draw_vertex_shader *vs,
+                        const struct draw_vs_varient_key *key )
+{
+   struct draw_vs_varient *varient;
+   unsigned i;
+
+   /* Lookup existing varient: 
+    */
+   for (i = 0; i < vs->nr_varients; i++)
+      if (draw_vs_varient_key_compare(key, &vs->varient[i]->key) == 0)
+         return vs->varient[i];
+   
+   /* Else have to create a new one: 
+    */
+   varient = vs->create_varient( vs, key );
+   if (varient == NULL)
+      return NULL;
+
+   /* Add it to our list, could be smarter: 
+    */
+   if (vs->nr_varients < Elements(vs->varient)) {
+      vs->varient[vs->nr_varients++] = varient;
+   }
+   else {
+      vs->last_varient++;
+      vs->last_varient %= Elements(vs->varient);
+      vs->varient[vs->last_varient]->destroy(vs->varient[vs->last_varient]);
+      vs->varient[vs->last_varient] = varient;
+   }
+
+   /* Done 
+    */
+   return varient;
+}
+
+
+struct translate *
+draw_vs_get_fetch( struct draw_context *draw,
+                   struct translate_key *key )
+{
+   if (!draw->vs.fetch ||
+       translate_key_compare(&draw->vs.fetch->key, key) != 0) 
+   {
+      translate_key_sanitize(key);
+      draw->vs.fetch = translate_cache_find(draw->vs.fetch_cache, key);
+   }
+   
+   return draw->vs.fetch;
+}
+
+struct translate *
+draw_vs_get_emit( struct draw_context *draw,
+                  struct translate_key *key )
+{
+   if (!draw->vs.emit ||
+       translate_key_compare(&draw->vs.emit->key, key) != 0) 
+   {
+      translate_key_sanitize(key);
+      draw->vs.emit = translate_cache_find(draw->vs.emit_cache, key);
+   }
+   
+   return draw->vs.emit;
+}
diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h
new file mode 100644
index 0000000000..89ae158751
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs.h
@@ -0,0 +1,224 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef DRAW_VS_H
+#define DRAW_VS_H
+
+#include "draw_context.h"
+#include "draw_private.h"
+
+
+struct draw_context;
+struct pipe_shader_state;
+
+struct draw_varient_input 
+{
+   enum pipe_format format;
+   unsigned buffer;
+   unsigned offset; 
+};
+
+struct draw_varient_output
+{
+   enum pipe_format format;     /* output format */
+   unsigned vs_output:8;        /* which vertex shader output is this? */
+   unsigned offset:24;          /* offset into output vertex */
+};
+
+struct draw_varient_element {
+   struct draw_varient_input in;
+   struct draw_varient_output out;
+};
+
+struct draw_vs_varient_key {
+   unsigned output_stride;
+   unsigned nr_elements:8;      /* max2(nr_inputs, nr_outputs) */
+   unsigned nr_inputs:8;
+   unsigned nr_outputs:8;
+   unsigned viewport:1;
+   unsigned clip:1;
+   unsigned const_vbuffers:5;
+   struct draw_varient_element element[PIPE_MAX_ATTRIBS];
+};
+
+struct draw_vs_varient;
+
+
+struct draw_vs_varient {
+   struct draw_vs_varient_key key;
+
+   struct draw_vertex_shader *vs;
+
+   void (*set_buffer)( struct draw_vs_varient *,
+                      unsigned i,
+                      const void *ptr,
+                      unsigned stride );
+
+   void (PIPE_CDECL *run_linear)( struct draw_vs_varient *shader,
+                                  unsigned start,
+                                  unsigned count,
+                                  void *output_buffer );
+
+   void (PIPE_CDECL *run_elts)( struct draw_vs_varient *shader,
+                                const unsigned *elts,
+                                unsigned count,
+                                void *output_buffer );
+
+   void (*destroy)( struct draw_vs_varient * );
+};
+
+
+/**
+ * Private version of the compiled vertex_shader
+ */
+struct draw_vertex_shader {
+   struct draw_context *draw;
+
+   /* This member will disappear shortly:
+    */
+   struct pipe_shader_state   state;
+
+   struct tgsi_shader_info info;
+   unsigned position_output;
+
+   /* Extracted from shader:
+    */
+   const float (*immediates)[4];
+
+   /* 
+    */
+   struct draw_vs_varient *varient[16];
+   unsigned nr_varients;
+   unsigned last_varient;
+   struct draw_vs_varient *(*create_varient)( struct draw_vertex_shader *shader,
+                                              const struct draw_vs_varient_key *key );
+
+
+   void (*prepare)( struct draw_vertex_shader *shader,
+		    struct draw_context *draw );
+
+   /* Run the shader - this interface will get cleaned up in the
+    * future:
+    */
+   void (*run_linear)( struct draw_vertex_shader *shader,
+		       const float (*input)[4],
+		       float (*output)[4],
+		       const float (*constants)[4],
+		       unsigned count,
+		       unsigned input_stride,
+		       unsigned output_stride );
+
+
+   void (*delete)( struct draw_vertex_shader * );
+};
+
+
+struct draw_vs_varient *
+draw_vs_lookup_varient( struct draw_vertex_shader *base,
+                        const struct draw_vs_varient_key *key );
+
+
+/********************************************************************************
+ * Internal functions:
+ */
+
+struct draw_vertex_shader *
+draw_create_vs_exec(struct draw_context *draw,
+		    const struct pipe_shader_state *templ);
+
+struct draw_vertex_shader *
+draw_create_vs_sse(struct draw_context *draw,
+		   const struct pipe_shader_state *templ);
+
+struct draw_vertex_shader *
+draw_create_vs_ppc(struct draw_context *draw,
+		   const struct pipe_shader_state *templ);
+
+struct draw_vertex_shader *
+draw_create_vs_llvm(struct draw_context *draw,
+		    const struct pipe_shader_state *templ);
+
+
+
+struct draw_vs_varient_key;
+struct draw_vertex_shader;
+
+struct draw_vs_varient *draw_vs_varient_aos_sse( struct draw_vertex_shader *vs,
+                                                 const struct draw_vs_varient_key *key );
+
+
+
+/********************************************************************************
+ * Helpers for vs implementations that don't do their own fetch/emit varients.
+ * Means these can be shared between shaders.
+ */
+struct translate;
+struct translate_key;
+
+struct translate *draw_vs_get_fetch( struct draw_context *draw,
+                                     struct translate_key *key );
+
+
+struct translate *draw_vs_get_emit( struct draw_context *draw,
+                                    struct translate_key *key );
+
+struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs,
+                                                 const struct draw_vs_varient_key *key );
+
+
+
+static INLINE int draw_vs_varient_keysize( const struct draw_vs_varient_key *key )
+{
+   return 2 * sizeof(int) + key->nr_elements * sizeof(struct draw_varient_element);
+}
+
+static INLINE int draw_vs_varient_key_compare( const struct draw_vs_varient_key *a,
+                                         const struct draw_vs_varient_key *b )
+{
+   int keysize = draw_vs_varient_keysize(a);
+   return memcmp(a, b, keysize);
+}
+
+
+struct aos_machine *draw_vs_aos_machine( void );
+void draw_vs_aos_machine_destroy( struct aos_machine *machine );
+
+void draw_vs_aos_machine_constants( struct aos_machine *machine,
+                                    const float (*constants)[4] );
+
+void draw_vs_aos_machine_viewport( struct aos_machine *machine,
+                                   const struct pipe_viewport_state *viewport );
+
+
+#define MAX_TGSI_VERTICES 4
+   
+
+
+#endif
diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c
new file mode 100644
index 0000000000..0c693a4a65
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_aos.c
@@ -0,0 +1,2252 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.3
+ *
+ * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * Translate tgsi vertex programs to x86/x87/SSE/SSE2 machine code
+ * using the rtasm runtime assembler.  Based on the old
+ * t_vb_arb_program_sse.c
+ */
+
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_debug.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi/tgsi_exec.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "draw_vs.h"
+#include "draw_vs_aos.h"
+
+#include "rtasm/rtasm_x86sse.h"
+
+#ifdef PIPE_ARCH_X86
+#define DISASSEM 0
+#define FAST_MATH 1
+
+static const char *files[] =
+{
+   "NULL",
+   "CONST",
+   "IN",
+   "OUT",
+   "TEMP",
+   "SAMP",
+   "ADDR",
+   "IMM",
+   "INTERNAL",
+};
+
+static INLINE boolean eq( struct x86_reg a,
+			    struct x86_reg b )
+{
+   return (a.file == b.file &&
+	   a.idx == b.idx &&
+	   a.mod == b.mod &&
+	   a.disp == b.disp);
+}
+      
+struct x86_reg aos_get_x86( struct aos_compilation *cp,
+                            unsigned which_reg, /* quick hack */
+                            unsigned value )
+{
+   struct x86_reg reg;
+
+   if (which_reg == 0)
+      reg = cp->temp_EBP;
+   else
+      reg = cp->tmp_EAX;
+
+   if (cp->x86_reg[which_reg] != value) {
+      unsigned offset;
+
+      switch (value) {
+      case X86_IMMEDIATES:
+         assert(which_reg == 0);
+         offset = Offset(struct aos_machine, immediates);
+         break;
+      case X86_CONSTANTS:
+         assert(which_reg == 1);
+         offset = Offset(struct aos_machine, constants);
+         break;
+      case X86_BUFFERS:
+         assert(which_reg == 0);
+         offset = Offset(struct aos_machine, buffer);
+         break;
+      default:
+         assert(0);
+         offset = 0;
+      }
+
+
+      x86_mov(cp->func, reg, 
+              x86_make_disp(cp->machine_EDX, offset));
+
+      cp->x86_reg[which_reg] = value;
+   }
+
+   return reg;
+}
+
+
+static struct x86_reg get_reg_ptr(struct aos_compilation *cp,
+                                  unsigned file,
+				  unsigned idx )
+{
+   struct x86_reg ptr = cp->machine_EDX;
+
+   switch (file) {
+   case TGSI_FILE_INPUT:
+      assert(idx < MAX_INPUTS);
+      return x86_make_disp(ptr, Offset(struct aos_machine, input[idx]));
+
+   case TGSI_FILE_OUTPUT:
+      return x86_make_disp(ptr, Offset(struct aos_machine, output[idx]));
+
+   case TGSI_FILE_TEMPORARY:
+      assert(idx < MAX_TEMPS);
+      return x86_make_disp(ptr, Offset(struct aos_machine, temp[idx]));
+
+   case AOS_FILE_INTERNAL:
+      assert(idx < MAX_INTERNALS);
+      return x86_make_disp(ptr, Offset(struct aos_machine, internal[idx]));
+
+   case TGSI_FILE_IMMEDIATE: 
+      assert(idx < MAX_IMMEDIATES);  /* just a sanity check */
+      return x86_make_disp(aos_get_x86(cp, 0, X86_IMMEDIATES), idx * 4 * sizeof(float));
+
+   case TGSI_FILE_CONSTANT: 
+      assert(idx < MAX_CONSTANTS);  /* just a sanity check */
+      return x86_make_disp(aos_get_x86(cp, 1, X86_CONSTANTS), idx * 4 * sizeof(float));
+
+   default:
+      ERROR(cp, "unknown reg file");
+      return x86_make_reg(0,0);
+   }
+}
+		
+
+
+#define X87_CW_EXCEPTION_INV_OP       (1<<0)
+#define X87_CW_EXCEPTION_DENORM_OP    (1<<1)
+#define X87_CW_EXCEPTION_ZERO_DIVIDE  (1<<2)
+#define X87_CW_EXCEPTION_OVERFLOW     (1<<3)
+#define X87_CW_EXCEPTION_UNDERFLOW    (1<<4)
+#define X87_CW_EXCEPTION_PRECISION    (1<<5)
+#define X87_CW_PRECISION_SINGLE       (0<<8)
+#define X87_CW_PRECISION_RESERVED     (1<<8)
+#define X87_CW_PRECISION_DOUBLE       (2<<8)
+#define X87_CW_PRECISION_DOUBLE_EXT   (3<<8)
+#define X87_CW_PRECISION_MASK         (3<<8)
+#define X87_CW_ROUND_NEAREST          (0<<10)
+#define X87_CW_ROUND_DOWN             (1<<10)
+#define X87_CW_ROUND_UP               (2<<10)
+#define X87_CW_ROUND_ZERO             (3<<10)
+#define X87_CW_ROUND_MASK             (3<<10)
+#define X87_CW_INFINITY               (1<<12)
+
+
+
+
+static void spill( struct aos_compilation *cp, unsigned idx )
+{
+   if (!cp->xmm[idx].dirty ||
+       (cp->xmm[idx].file != TGSI_FILE_INPUT && /* inputs are fetched into xmm & set dirty */
+        cp->xmm[idx].file != TGSI_FILE_OUTPUT &&
+        cp->xmm[idx].file != TGSI_FILE_TEMPORARY)) {
+      ERROR(cp, "invalid spill");
+      return;
+   }
+   else {
+      struct x86_reg oldval = get_reg_ptr(cp,
+                                          cp->xmm[idx].file,
+                                          cp->xmm[idx].idx);
+     
+      if (0) debug_printf("\nspill %s[%d]", 
+                          files[cp->xmm[idx].file],
+                          cp->xmm[idx].idx);
+ 
+      assert(cp->xmm[idx].dirty);
+      sse_movaps(cp->func, oldval, x86_make_reg(file_XMM, idx));
+      cp->xmm[idx].dirty = 0;
+   }
+}
+
+
+void aos_spill_all( struct aos_compilation *cp )
+{
+   unsigned i;
+
+   for (i = 0; i < 8; i++) {
+      if (cp->xmm[i].dirty) 
+         spill(cp, i);
+      aos_release_xmm_reg(cp, i);
+   }
+}
+
+
+static struct x86_reg get_xmm_writable( struct aos_compilation *cp,
+                                        struct x86_reg reg )
+{
+   if (reg.file != file_XMM ||
+       cp->xmm[reg.idx].file != TGSI_FILE_NULL)
+   {
+      struct x86_reg tmp = aos_get_xmm_reg(cp);
+      sse_movaps(cp->func, tmp, reg);
+      reg = tmp;
+   }
+
+   cp->xmm[reg.idx].last_used = cp->insn_counter;
+   return reg;
+}
+
+static struct x86_reg get_xmm( struct aos_compilation *cp,
+                               struct x86_reg reg )
+{
+   if (reg.file != file_XMM) 
+   {
+      struct x86_reg tmp = aos_get_xmm_reg(cp);
+      sse_movaps(cp->func, tmp, reg);
+      reg = tmp;
+   }
+
+   cp->xmm[reg.idx].last_used = cp->insn_counter;
+   return reg;
+}
+
+
+/* Allocate an empty xmm register, either as a temporary or later to
+ * "adopt" as a shader reg.
+ */
+struct x86_reg aos_get_xmm_reg( struct aos_compilation *cp )
+{
+   unsigned i;
+   unsigned oldest = 0;
+   boolean found = FALSE;
+
+   for (i = 0; i < 8; i++) 
+      if (cp->xmm[i].last_used != cp->insn_counter &&
+          cp->xmm[i].file == TGSI_FILE_NULL) {
+	 oldest = i;
+         found = TRUE;
+      }
+
+   if (!found) {
+      for (i = 0; i < 8; i++) 
+         if (cp->xmm[i].last_used < cp->xmm[oldest].last_used)
+            oldest = i;
+   }
+
+   /* Need to write out the old value?
+    */
+   if (cp->xmm[oldest].dirty) 
+      spill(cp, oldest);
+
+   assert(cp->xmm[oldest].last_used != cp->insn_counter);
+
+   cp->xmm[oldest].file = TGSI_FILE_NULL;
+   cp->xmm[oldest].idx = 0;
+   cp->xmm[oldest].dirty = 0;
+   cp->xmm[oldest].last_used = cp->insn_counter;
+   return x86_make_reg(file_XMM, oldest);
+}
+
+void aos_release_xmm_reg( struct aos_compilation *cp,
+                          unsigned idx )
+{
+   cp->xmm[idx].file = TGSI_FILE_NULL;
+   cp->xmm[idx].idx = 0;
+   cp->xmm[idx].dirty = 0;
+   cp->xmm[idx].last_used = 0;
+}
+
+
+
+     
+/* Mark an xmm reg as holding the current copy of a shader reg.
+ */
+void aos_adopt_xmm_reg( struct aos_compilation *cp,
+                        struct x86_reg reg,
+                        unsigned file,
+                        unsigned idx,
+                        unsigned dirty )
+{
+   unsigned i;
+
+   if (reg.file != file_XMM) {
+      assert(0);
+      return;
+   }
+
+
+   /* If any xmm reg thinks it holds this shader reg, break the
+    * illusion.
+    */
+   for (i = 0; i < 8; i++) {
+      if (cp->xmm[i].file == file && 
+          cp->xmm[i].idx == idx) 
+      {
+         /* If an xmm reg is already holding this shader reg, take into account its
+          * dirty flag...
+          */
+         dirty |= cp->xmm[i].dirty;
+         aos_release_xmm_reg(cp, i);
+      }
+   }
+
+   cp->xmm[reg.idx].file = file;
+   cp->xmm[reg.idx].idx = idx;
+   cp->xmm[reg.idx].dirty = dirty;
+   cp->xmm[reg.idx].last_used = cp->insn_counter;
+}
+
+
+/* Return a pointer to the in-memory copy of the reg, making sure it is uptodate.
+ */
+static struct x86_reg aos_get_shader_reg_ptr( struct aos_compilation *cp, 
+                                              unsigned file,
+                                              unsigned idx )
+{
+   unsigned i;
+
+   /* Ensure the in-memory copy of this reg is up-to-date
+    */
+   for (i = 0; i < 8; i++) {
+      if (cp->xmm[i].file == file && 
+          cp->xmm[i].idx == idx &&
+          cp->xmm[i].dirty) {
+         spill(cp, i);
+      }
+   }
+
+   return get_reg_ptr( cp, file, idx );
+}
+
+
+/* As above, but return a pointer.  Note - this pointer may alias
+ * those returned by get_arg_ptr().
+ */
+static struct x86_reg get_dst_ptr( struct aos_compilation *cp, 
+                                   const struct tgsi_full_dst_register *dst )
+{
+   unsigned file = dst->DstRegister.File;
+   unsigned idx = dst->DstRegister.Index;
+   unsigned i;
+   
+
+   /* Ensure in-memory copy of this reg is up-to-date and invalidate
+    * any xmm copies.
+    */
+   for (i = 0; i < 8; i++) {
+      if (cp->xmm[i].file == file &&
+          cp->xmm[i].idx == idx)
+      {
+         if (cp->xmm[i].dirty) 
+            spill(cp, i);
+         
+         aos_release_xmm_reg(cp, i);
+      }
+   }
+
+   return get_reg_ptr( cp, file, idx );
+}
+
+
+
+
+
+/* Return an XMM reg if the argument is resident, otherwise return a
+ * base+offset pointer to the saved value.
+ */
+struct x86_reg aos_get_shader_reg( struct aos_compilation *cp, 
+                                   unsigned file,
+                                   unsigned idx )
+{
+   unsigned i;
+
+   for (i = 0; i < 8; i++) {
+      if (cp->xmm[i].file == file &&
+	  cp->xmm[i].idx  == idx) 
+      {
+	 cp->xmm[i].last_used = cp->insn_counter;
+	 return x86_make_reg(file_XMM, i);
+      }
+   }
+
+   /* If not found in the XMM register file, return an indirect
+    * reference to the in-memory copy:
+    */
+   return get_reg_ptr( cp, file, idx );
+}
+
+
+
+static struct x86_reg aos_get_shader_reg_xmm( struct aos_compilation *cp, 
+                                              unsigned file,
+                                              unsigned idx )
+{
+   struct x86_reg reg = get_xmm( cp,
+                                 aos_get_shader_reg( cp, file, idx ) );
+
+   aos_adopt_xmm_reg( cp,
+                      reg,
+                      file,
+                      idx,
+                      FALSE );
+   
+   return reg;
+}
+
+
+
+struct x86_reg aos_get_internal_xmm( struct aos_compilation *cp,
+                                     unsigned imm )
+{
+   return aos_get_shader_reg_xmm( cp, AOS_FILE_INTERNAL, imm );
+}
+
+
+struct x86_reg aos_get_internal( struct aos_compilation *cp,
+                                 unsigned imm )
+{
+   return aos_get_shader_reg( cp, AOS_FILE_INTERNAL, imm );
+}
+
+
+
+
+
+/* Emulate pshufd insn in regular SSE, if necessary:
+ */
+static void emit_pshufd( struct aos_compilation *cp,
+			 struct x86_reg dst,
+			 struct x86_reg arg0,
+			 ubyte shuf )
+{
+   if (cp->have_sse2) {
+      sse2_pshufd(cp->func, dst, arg0, shuf);
+   }
+   else {
+      if (!eq(dst, arg0)) 
+	 sse_movaps(cp->func, dst, arg0);
+
+      sse_shufps(cp->func, dst, dst, shuf);
+   }
+}
+
+/* load masks (pack into negs??)
+ * pshufd - shuffle according to writemask
+ * and - result, mask
+ * nand - dest, mask
+ * or - dest, result
+ */
+static boolean mask_write( struct aos_compilation *cp,
+                           struct x86_reg dst,
+                           struct x86_reg result,
+                           unsigned mask )
+{
+   struct x86_reg imm_swz = aos_get_internal_xmm(cp, IMM_SWZ);
+   struct x86_reg tmp = aos_get_xmm_reg(cp);
+   
+   emit_pshufd(cp, tmp, imm_swz, 
+               SHUF((mask & 1) ? 2 : 3,
+                    (mask & 2) ? 2 : 3,
+                    (mask & 4) ? 2 : 3,
+                    (mask & 8) ? 2 : 3));
+
+   sse_andps(cp->func, dst, tmp);
+   sse_andnps(cp->func, tmp, result);
+   sse_orps(cp->func, dst, tmp);
+
+   aos_release_xmm_reg(cp, tmp.idx);
+   return TRUE;
+}
+
+
+
+
+/* Helper for writemask:
+ */
+static boolean emit_shuf_copy2( struct aos_compilation *cp,
+				  struct x86_reg dst,
+				  struct x86_reg arg0,
+				  struct x86_reg arg1,
+				  ubyte shuf )
+{
+   struct x86_reg tmp = aos_get_xmm_reg(cp);
+
+   emit_pshufd(cp, dst, arg1, shuf);
+   emit_pshufd(cp, tmp, arg0, shuf);
+   sse_shufps(cp->func, dst, tmp, SHUF(X, Y, Z, W));
+   emit_pshufd(cp, dst, dst, shuf);
+
+   aos_release_xmm_reg(cp, tmp.idx);
+   return TRUE;
+}
+
+
+
+#define SSE_SWIZZLE_NOOP ((0<<0) | (1<<2) | (2<<4) | (3<<6))
+
+
+/* Locate a source register and perform any required (simple) swizzle.  
+ * 
+ * Just fail on complex swizzles at this point.
+ */
+static struct x86_reg fetch_src( struct aos_compilation *cp, 
+                                 const struct tgsi_full_src_register *src ) 
+{
+   struct x86_reg arg0 = aos_get_shader_reg(cp, 
+                                            src->SrcRegister.File, 
+                                            src->SrcRegister.Index);
+   unsigned i;
+   ubyte swz = 0;
+   unsigned negs = 0;
+   unsigned abs = 0;
+
+   for (i = 0; i < 4; i++) {
+      unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( src, i );
+      unsigned neg = tgsi_util_get_full_src_register_sign_mode( src, i );
+
+      switch (swizzle) {
+      case TGSI_EXTSWIZZLE_ZERO:
+      case TGSI_EXTSWIZZLE_ONE:
+         ERROR(cp, "not supporting full swizzles yet in tgsi_aos_sse2");
+         break;
+
+      default:
+         swz |= (swizzle & 0x3) << (i * 2);
+         break;
+      }
+
+      switch (neg) {
+      case TGSI_UTIL_SIGN_TOGGLE:
+         negs |= (1<<i);
+         break;
+         
+      case TGSI_UTIL_SIGN_KEEP:
+         break;
+
+      case TGSI_UTIL_SIGN_CLEAR:
+         abs |= (1<<i);
+         break;
+
+      default:
+         ERROR(cp, "unsupported sign-mode");
+         break;
+      }
+   }
+
+   if (swz != SSE_SWIZZLE_NOOP || negs != 0 || abs != 0) {
+      struct x86_reg dst = aos_get_xmm_reg(cp);
+
+      if (swz != SSE_SWIZZLE_NOOP)
+         emit_pshufd(cp, dst, arg0, swz);
+      else
+         sse_movaps(cp->func, dst, arg0);
+
+      if (negs && negs != 0xf) {
+         struct x86_reg imm_swz = aos_get_internal_xmm(cp, IMM_SWZ);
+         struct x86_reg tmp = aos_get_xmm_reg(cp);
+
+         /* Load 1,-1,0,0
+          * Use neg as arg to pshufd
+          * Multiply
+          */
+         emit_pshufd(cp, tmp, imm_swz, 
+                     SHUF((negs & 1) ? 1 : 0,
+                          (negs & 2) ? 1 : 0,
+                          (negs & 4) ? 1 : 0,
+                          (negs & 8) ? 1 : 0));
+         sse_mulps(cp->func, dst, tmp);
+
+         aos_release_xmm_reg(cp, tmp.idx);
+      }
+      else if (negs) {
+         struct x86_reg imm_negs = aos_get_internal_xmm(cp, IMM_NEGS);
+         sse_mulps(cp->func, dst, imm_negs);
+      }
+
+
+      if (abs && abs != 0xf) {
+         ERROR(cp, "unsupported partial abs");
+      }
+      else if (abs) {
+         struct x86_reg neg = aos_get_internal(cp, IMM_NEGS);
+         struct x86_reg tmp = aos_get_xmm_reg(cp);
+
+         sse_movaps(cp->func, tmp, dst);
+         sse_mulps(cp->func, tmp, neg);
+         sse_maxps(cp->func, dst, tmp);
+
+         aos_release_xmm_reg(cp, tmp.idx);
+      }
+
+      return dst;
+   }
+      
+   return arg0;
+}
+
+static void x87_fld_src( struct aos_compilation *cp, 
+                         const struct tgsi_full_src_register *src,
+                         unsigned channel ) 
+{
+   struct x86_reg arg0 = aos_get_shader_reg_ptr(cp, 
+                                                src->SrcRegister.File, 
+                                                src->SrcRegister.Index);
+
+   unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( src, channel );
+   unsigned neg = tgsi_util_get_full_src_register_sign_mode( src, channel );
+
+   switch (swizzle) {
+   case TGSI_EXTSWIZZLE_ZERO:
+      x87_fldz( cp->func );
+      break;
+
+   case TGSI_EXTSWIZZLE_ONE:
+      x87_fld1( cp->func );
+      break;
+
+   default:
+      x87_fld( cp->func, x86_make_disp(arg0, (swizzle & 3) * sizeof(float)) );
+      break;
+   }
+   
+
+   switch (neg) {
+   case TGSI_UTIL_SIGN_TOGGLE:
+      /* Flip the sign:
+       */
+      x87_fchs( cp->func );
+      break;
+         
+   case TGSI_UTIL_SIGN_KEEP:
+      break;
+
+   case TGSI_UTIL_SIGN_CLEAR:
+      x87_fabs( cp->func );
+      break;
+
+   case TGSI_UTIL_SIGN_SET:
+      x87_fabs( cp->func );
+      x87_fchs( cp->func );
+      break;
+
+   default:
+      ERROR(cp, "unsupported sign-mode");
+      break;
+   }
+}
+
+
+
+
+
+
+/* Used to implement write masking.  This and most of the other instructions
+ * here would be easier to implement if there had been a translation
+ * to a 2 argument format (dst/arg0, arg1) at the shader level before
+ * attempting to translate to x86/sse code.
+ */
+static void store_dest( struct aos_compilation *cp, 
+                        const struct tgsi_full_dst_register *reg,
+                        struct x86_reg result )
+{
+   struct x86_reg dst;
+
+   switch (reg->DstRegister.WriteMask) {
+   case 0:
+      return;
+   
+   case TGSI_WRITEMASK_XYZW:
+      aos_adopt_xmm_reg(cp, 
+                        get_xmm_writable(cp, result), 
+                        reg->DstRegister.File,
+                        reg->DstRegister.Index,
+                        TRUE);
+      return;
+   default: 
+      break;
+   }
+
+   dst = aos_get_shader_reg_xmm(cp, 
+                                reg->DstRegister.File,
+                                reg->DstRegister.Index);
+
+   switch (reg->DstRegister.WriteMask) {
+   case TGSI_WRITEMASK_X:
+      sse_movss(cp->func, dst, get_xmm(cp, result));
+      break;
+      
+   case TGSI_WRITEMASK_ZW:
+      sse_shufps(cp->func, dst, get_xmm(cp, result), SHUF(X, Y, Z, W));
+      break;
+
+   case TGSI_WRITEMASK_XY: 
+      result = get_xmm_writable(cp, result);
+      sse_shufps(cp->func, result, dst, SHUF(X, Y, Z, W));
+      dst = result;
+      break;
+
+   case TGSI_WRITEMASK_YZW: 
+      result = get_xmm_writable(cp, result);
+      sse_movss(cp->func, result, dst);
+      dst = result;
+      break;
+
+   default:
+      mask_write(cp, dst, result, reg->DstRegister.WriteMask);
+      break;
+   }
+
+   aos_adopt_xmm_reg(cp, 
+                     dst, 
+                     reg->DstRegister.File,
+                     reg->DstRegister.Index,
+                     TRUE);
+
+}
+
+static void inject_scalar( struct aos_compilation *cp,
+                           struct x86_reg dst,
+                           struct x86_reg result,
+                           ubyte swizzle )
+{
+   sse_shufps(cp->func, dst, dst, swizzle);
+   sse_movss(cp->func, dst, result);
+   sse_shufps(cp->func, dst, dst, swizzle);
+}
+
+
+static void store_scalar_dest( struct aos_compilation *cp, 
+                               const struct tgsi_full_dst_register *reg,
+                               struct x86_reg result )
+{
+   unsigned writemask = reg->DstRegister.WriteMask;
+   struct x86_reg dst;
+
+   if (writemask != TGSI_WRITEMASK_X &&
+       writemask != TGSI_WRITEMASK_Y &&
+       writemask != TGSI_WRITEMASK_Z &&
+       writemask != TGSI_WRITEMASK_W &&
+       writemask != 0) 
+   {
+      result = get_xmm_writable(cp, result); /* already true, right? */
+      sse_shufps(cp->func, result, result, SHUF(X,X,X,X));
+      store_dest(cp, reg, result);
+      return;
+   }
+
+   result = get_xmm(cp, result);
+   dst = aos_get_shader_reg_xmm(cp, 
+                                reg->DstRegister.File,
+                                reg->DstRegister.Index);
+
+
+
+   switch (reg->DstRegister.WriteMask) {
+   case TGSI_WRITEMASK_X:
+      sse_movss(cp->func, dst, result);
+      break;
+
+   case TGSI_WRITEMASK_Y:
+      inject_scalar(cp, dst, result, SHUF(Y, X, Z, W));
+      break;
+
+   case TGSI_WRITEMASK_Z:
+      inject_scalar(cp, dst, result, SHUF(Z, Y, X, W));
+      break;
+
+   case TGSI_WRITEMASK_W:
+      inject_scalar(cp, dst, result, SHUF(W, Y, Z, X));
+      break;
+
+   default:
+      break;
+   }
+
+   aos_adopt_xmm_reg(cp, 
+                     dst, 
+                     reg->DstRegister.File,
+                     reg->DstRegister.Index,
+                     TRUE);
+}
+   
+
+
+static void x87_fst_or_nop( struct x86_function *func,
+                            unsigned writemask,
+                            unsigned channel,
+                            struct x86_reg ptr )
+{
+   assert(ptr.file == file_REG32);
+   if (writemask & (1<<channel)) 
+      x87_fst( func, x86_make_disp(ptr, channel * sizeof(float)) );
+}
+
+static void x87_fstp_or_pop( struct x86_function *func,
+                             unsigned writemask,
+                             unsigned channel,
+                             struct x86_reg ptr )
+{
+   assert(ptr.file == file_REG32);
+   if (writemask & (1<<channel)) 
+      x87_fstp( func, x86_make_disp(ptr, channel * sizeof(float)) );
+   else
+      x87_fstp( func, x86_make_reg( file_x87, 0 ));
+}
+
+
+
+/* 
+ */
+static void x87_fstp_dest4( struct aos_compilation *cp,
+                            const struct tgsi_full_dst_register *dst )
+{
+   struct x86_reg ptr = get_dst_ptr(cp, dst); 
+   unsigned writemask = dst->DstRegister.WriteMask;
+
+   x87_fst_or_nop(cp->func, writemask, 0, ptr);
+   x87_fst_or_nop(cp->func, writemask, 1, ptr);
+   x87_fst_or_nop(cp->func, writemask, 2, ptr);
+   x87_fstp_or_pop(cp->func, writemask, 3, ptr);
+}
+
+/* Save current x87 state and put it into single precision mode.
+ */
+static void save_fpu_state( struct aos_compilation *cp )
+{
+   x87_fnstcw( cp->func, x86_make_disp(cp->machine_EDX, 
+                                       Offset(struct aos_machine, fpu_restore)));
+}
+
+static void restore_fpu_state( struct aos_compilation *cp )
+{
+   x87_fnclex(cp->func);
+   x87_fldcw( cp->func, x86_make_disp(cp->machine_EDX, 
+                                      Offset(struct aos_machine, fpu_restore)));
+}
+
+static void set_fpu_round_neg_inf( struct aos_compilation *cp )
+{
+   if (cp->fpucntl != FPU_RND_NEG) {
+      cp->fpucntl = FPU_RND_NEG;
+      x87_fnclex(cp->func);
+      x87_fldcw( cp->func, x86_make_disp(cp->machine_EDX, 
+                                         Offset(struct aos_machine, fpu_rnd_neg_inf)));
+   }
+}
+
+static void set_fpu_round_nearest( struct aos_compilation *cp )
+{
+   if (cp->fpucntl != FPU_RND_NEAREST) {
+      cp->fpucntl = FPU_RND_NEAREST;
+      x87_fnclex(cp->func);
+      x87_fldcw( cp->func, x86_make_disp(cp->machine_EDX, 
+                                         Offset(struct aos_machine, fpu_rnd_nearest)));
+   }
+}
+
+#if 0
+static void x87_emit_ex2( struct aos_compilation *cp )
+{
+   struct x86_reg st0 = x86_make_reg(file_x87, 0);
+   struct x86_reg st1 = x86_make_reg(file_x87, 1);
+   int stack = cp->func->x87_stack;
+
+//   set_fpu_round_neg_inf( cp );
+
+   x87_fld(cp->func, st0);      /* a a */
+   x87_fprndint( cp->func );	/* int(a) a*/
+   x87_fsubr(cp->func, st1, st0);    /* int(a) frc(a) */
+   x87_fxch(cp->func, st1);     /* frc(a) int(a) */
+   x87_f2xm1(cp->func);         /* (2^frc(a))-1 int(a) */
+   x87_fld1(cp->func);          /* 1 (2^frc(a))-1 int(a) */
+   x87_faddp(cp->func, st1);	/* 2^frac(a) int(a)  */
+   x87_fscale(cp->func);	/* (2^frac(a)*2^int(int(a))) int(a) */
+                                /* 2^a int(a) */
+   x87_fstp(cp->func, st1);     /* 2^a */
+
+   assert( stack == cp->func->x87_stack);
+      
+}
+#endif
+
+#if 0
+static void PIPE_CDECL print_reg( const char *msg,
+                                  const float *reg )
+{
+   debug_printf("%s: %f %f %f %f\n", msg, reg[0], reg[1], reg[2], reg[3]);
+}
+#endif
+
+#if 0
+static void emit_print( struct aos_compilation *cp,
+                        const char *message, /* must point to a static string! */
+                        unsigned file,
+                        unsigned idx )
+{
+   struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX );
+   struct x86_reg arg = aos_get_shader_reg_ptr( cp, file, idx );
+   unsigned i;
+
+   /* There shouldn't be anything on the x87 stack.  Can add this
+    * capacity later if need be.
+    */
+   assert(cp->func->x87_stack == 0);
+
+   /* For absolute correctness, need to spill/invalidate all XMM regs
+    * too.  We're obviously not concerned about performance on this
+    * debug path, so here goes:
+    */
+   for (i = 0; i < 8; i++) {
+      if (cp->xmm[i].dirty) 
+         spill(cp, i);
+
+      aos_release_xmm_reg(cp, i);
+   }
+
+   /* Push caller-save (ie scratch) regs.  
+    */
+   x86_cdecl_caller_push_regs( cp->func );
+
+
+   /* Push the arguments:
+    */
+   x86_lea( cp->func, ecx, arg );
+   x86_push( cp->func, ecx );
+   x86_push_imm32( cp->func, (int)message );
+
+   /* Call the helper.  Could call debug_printf directly, but
+    * print_reg is a nice place to put a breakpoint if need be.
+    */
+   x86_mov_reg_imm( cp->func, ecx, (int)print_reg );
+   x86_call( cp->func, ecx );
+   x86_pop( cp->func, ecx );
+   x86_pop( cp->func, ecx );
+
+   /* Pop caller-save regs 
+    */
+   x86_cdecl_caller_pop_regs( cp->func );
+
+   /* Done... 
+    */
+}
+#endif
+
+/**
+ * The traditional instructions.  All operate on internal registers
+ * and ignore write masks and swizzling issues.
+ */
+
+static boolean emit_ABS( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg neg = aos_get_internal(cp, IMM_NEGS);
+   struct x86_reg tmp = aos_get_xmm_reg(cp);
+
+   sse_movaps(cp->func, tmp, arg0);
+   sse_mulps(cp->func, tmp, neg);
+   sse_maxps(cp->func, tmp, arg0);
+   
+   store_dest(cp, &op->FullDstRegisters[0], tmp);
+   return TRUE;
+}
+
+static boolean emit_ADD( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+
+   sse_addps(cp->func, dst, arg1);
+
+   store_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+static boolean emit_COS( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   x87_fld_src(cp, &op->FullSrcRegisters[0], 0);
+   x87_fcos(cp->func);
+   x87_fstp_dest4(cp, &op->FullDstRegisters[0]);
+   return TRUE;
+}
+
+/* The dotproduct instructions don't really do that well in sse:
+ * XXX: produces wrong results -- disabled.
+ */
+static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg tmp = aos_get_xmm_reg(cp); 
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+
+   sse_mulps(cp->func, dst, arg1);
+   /* Now the hard bit: sum the first 3 values:
+    */ 
+   sse_movhlps(cp->func, tmp, dst);
+   sse_addss(cp->func, dst, tmp); /* a*x+c*z, b*y, ?, ? */
+   emit_pshufd(cp, tmp, dst, SHUF(Y,X,W,Z));
+   sse_addss(cp->func, dst, tmp);
+   
+   aos_release_xmm_reg(cp, tmp.idx);
+   store_scalar_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg tmp = aos_get_xmm_reg(cp);      
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+
+   sse_mulps(cp->func, dst, arg1);
+   
+   /* Now the hard bit: sum the values:
+    */ 
+   sse_movhlps(cp->func, tmp, dst);
+   sse_addps(cp->func, dst, tmp); /* a*x+c*z, b*y+d*w, a*x+c*z, b*y+d*w */
+   emit_pshufd(cp, tmp, dst, SHUF(Y,X,W,Z));
+   sse_addss(cp->func, dst, tmp);
+
+   aos_release_xmm_reg(cp, tmp.idx);
+   store_scalar_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+static boolean emit_DPH( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg tmp = aos_get_xmm_reg(cp);
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+
+   sse_mulps(cp->func, dst, arg1);
+
+   /* Now the hard bit: sum the values (from DP3):
+    */ 
+   sse_movhlps(cp->func, tmp, dst);
+   sse_addss(cp->func, dst, tmp); /* a*x+c*z, b*y, ?, ? */
+   emit_pshufd(cp, tmp, dst, SHUF(Y,X,W,Z));
+   sse_addss(cp->func, dst, tmp);
+   emit_pshufd(cp, tmp, arg1, SHUF(W,W,W,W));
+   sse_addss(cp->func, dst, tmp);
+
+   aos_release_xmm_reg(cp, tmp.idx);
+   store_scalar_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+static boolean emit_DST( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+    struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+    struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+    struct x86_reg dst = aos_get_xmm_reg(cp);
+    struct x86_reg tmp = aos_get_xmm_reg(cp);
+    struct x86_reg ones = aos_get_internal(cp, IMM_ONES);
+
+/*    dst[0] = 1.0     * 1.0F; */
+/*    dst[1] = arg0[1] * arg1[1]; */
+/*    dst[2] = arg0[2] * 1.0; */
+/*    dst[3] = 1.0     * arg1[3]; */
+
+    emit_shuf_copy2(cp, dst, arg0, ones, SHUF(X,W,Z,Y));
+    emit_shuf_copy2(cp, tmp, arg1, ones, SHUF(X,Z,Y,W));
+    sse_mulps(cp->func, dst, tmp);
+
+    aos_release_xmm_reg(cp, tmp.idx);
+    store_dest(cp, &op->FullDstRegisters[0], dst);
+    return TRUE;
+}
+
+static boolean emit_LG2( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   x87_fld1(cp->func);		/* 1 */
+   x87_fld_src(cp, &op->FullSrcRegisters[0], 0);	/* a0 1 */
+   x87_fyl2x(cp->func);	/* log2(a0) */
+   x87_fstp_dest4(cp, &op->FullDstRegisters[0]);
+   return TRUE;
+}
+
+#if 0
+static boolean emit_EX2( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   x87_fld_src(cp, &op->FullSrcRegisters[0], 0);
+   x87_emit_ex2(cp);
+   x87_fstp_dest4(cp, &op->FullDstRegisters[0]);
+   return TRUE;
+}
+#endif
+
+
+static boolean emit_FLR( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); 
+   unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask;
+   int i;
+
+   set_fpu_round_neg_inf( cp );
+
+   /* Load all sources first to avoid aliasing
+    */
+   for (i = 3; i >= 0; i--) {
+      if (writemask & (1<<i)) {
+         x87_fld_src(cp, &op->FullSrcRegisters[0], i);   
+      }
+   }
+
+   for (i = 0; i < 4; i++) {
+      if (writemask & (1<<i)) {
+         x87_fprndint( cp->func );   
+         x87_fstp(cp->func, x86_make_disp(dst, i*4));
+      }
+   }
+
+   return TRUE;
+}
+
+
+static boolean emit_RND( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); 
+   unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask;
+   int i;
+
+   set_fpu_round_nearest( cp );
+
+   /* Load all sources first to avoid aliasing
+    */
+   for (i = 3; i >= 0; i--) {
+      if (writemask & (1<<i)) {
+         x87_fld_src(cp, &op->FullSrcRegisters[0], i);   
+      }
+   }
+
+   for (i = 0; i < 4; i++) {
+      if (writemask & (1<<i)) {
+         x87_fprndint( cp->func );   
+         x87_fstp(cp->func, x86_make_disp(dst, i*4));
+      }
+   }
+
+   return TRUE;
+}
+
+
+static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); 
+   struct x86_reg st0 = x86_make_reg(file_x87, 0);
+   struct x86_reg st1 = x86_make_reg(file_x87, 1);
+   unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask;
+   int i;
+
+   set_fpu_round_neg_inf( cp );
+
+   /* suck all the source values onto the stack before writing out any
+    * dst, which may alias...
+    */
+   for (i = 3; i >= 0; i--) {
+      if (writemask & (1<<i)) {
+         x87_fld_src(cp, &op->FullSrcRegisters[0], i);   
+      }
+   }
+
+   for (i = 0; i < 4; i++) {
+      if (writemask & (1<<i)) {
+         x87_fld(cp->func, st0);     /* a a */
+         x87_fprndint( cp->func );   /* flr(a) a */
+         x87_fsubp(cp->func, st1);  /* frc(a) */
+         x87_fstp(cp->func, x86_make_disp(dst, i*4));
+      }
+   }
+
+   return TRUE;
+}
+
+
+
+
+
+
+static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX );
+   unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask;
+   unsigned lit_count = cp->lit_count++;
+   struct x86_reg result, arg0;
+   unsigned i;
+
+#if 1
+   /* For absolute correctness, need to spill/invalidate all XMM regs
+    * too.  
+    */
+   for (i = 0; i < 8; i++) {
+      if (cp->xmm[i].dirty) 
+         spill(cp, i);
+      aos_release_xmm_reg(cp, i);
+   }
+#endif
+
+   if (writemask != TGSI_WRITEMASK_XYZW) 
+      result = x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, tmp[0]));
+   else 
+      result = get_dst_ptr(cp, &op->FullDstRegisters[0]);    
+
+   
+   arg0 = fetch_src( cp, &op->FullSrcRegisters[0] );
+   if (arg0.file == file_XMM) {
+      struct x86_reg tmp = x86_make_disp(cp->machine_EDX, 
+                                         Offset(struct aos_machine, tmp[1]));
+      sse_movaps( cp->func, tmp, arg0 );
+      arg0 = tmp;
+   }
+                  
+      
+
+   /* Push caller-save (ie scratch) regs.  
+    */
+   x86_cdecl_caller_push_regs( cp->func );
+
+   /* Push the arguments:
+    */
+   x86_push_imm32( cp->func, lit_count );
+
+   x86_lea( cp->func, ecx, arg0 );
+   x86_push( cp->func, ecx );
+
+   x86_lea( cp->func, ecx, result );
+   x86_push( cp->func, ecx );
+
+   x86_push( cp->func, cp->machine_EDX );
+
+   if (lit_count < MAX_LIT_INFO) {
+      x86_mov( cp->func, ecx, x86_make_disp( cp->machine_EDX, 
+                                             Offset(struct aos_machine, lit_info) + 
+                                             lit_count * sizeof(struct lit_info) + 
+                                             Offset(struct lit_info, func)));
+   }
+   else {
+      x86_mov_reg_imm( cp->func, ecx, (int)aos_do_lit );
+   }
+
+   x86_call( cp->func, ecx );
+            
+   x86_pop( cp->func, ecx );    /* fixme... */
+   x86_pop( cp->func, ecx );
+   x86_pop( cp->func, ecx );
+   x86_pop( cp->func, ecx );
+
+   x86_cdecl_caller_pop_regs( cp->func );
+
+   if (writemask != TGSI_WRITEMASK_XYZW) {
+      store_dest( cp, 
+                  &op->FullDstRegisters[0],
+                  get_xmm_writable( cp, result ) );
+   }
+
+   return TRUE;
+}
+
+#if 0   
+static boolean emit_inline_LIT( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); 
+   unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask;
+
+   if (writemask & TGSI_WRITEMASK_YZ) {
+      struct x86_reg st1 = x86_make_reg(file_x87, 1);
+      struct x86_reg st2 = x86_make_reg(file_x87, 2);
+
+      /* a1' = a1 <= 0 ? 1 : a1;  
+       */
+      x87_fldz(cp->func);                           /* 1 0  */
+#if 1
+      x87_fld1(cp->func);                           /* 1 0  */
+#else
+      /* Correct but slow due to fp exceptions generated in fyl2x - fix me.
+       */
+      x87_fldz(cp->func);                           /* 1 0  */
+#endif
+      x87_fld_src(cp, &op->FullSrcRegisters[0], 1); /* a1 1 0  */
+      x87_fcomi(cp->func, st2);	                    /* a1 1 0  */
+      x87_fcmovb(cp->func, st1);                    /* a1' 1 0  */
+      x87_fstp(cp->func, st1);                      /* a1' 0  */
+      x87_fstp(cp->func, st1);                      /* a1'  */
+
+      x87_fld_src(cp, &op->FullSrcRegisters[0], 3); /* a3 a1'  */
+      x87_fxch(cp->func, st1);                      /* a1' a3  */
+      
+
+      /* Compute pow(a1, a3)
+       */
+      x87_fyl2x(cp->func);	/* a3*log2(a1)      */
+      x87_emit_ex2( cp );       /* 2^(a3*log2(a1))   */
+
+
+      /* a0' = max2(a0, 0):
+       */
+      x87_fldz(cp->func);                           /* 0 r2 */
+      x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* a0 0 r2 */
+      x87_fcomi(cp->func, st1);	
+      x87_fcmovb(cp->func, st1);                    /* a0' 0 r2 */
+
+      x87_fst_or_nop(cp->func, writemask, 1, dst); /* result[1] = a0' */
+
+      x87_fcomi(cp->func, st1);  /* a0' 0 r2 */
+      x87_fcmovnbe(cp->func, st2); /* r2' 0' r2 */
+
+      x87_fstp_or_pop(cp->func, writemask, 2, dst); /* 0 r2 */
+      x87_fpop(cp->func);       /* r2 */
+      x87_fpop(cp->func);
+   }
+
+   if (writemask & TGSI_WRITEMASK_XW) {
+      x87_fld1(cp->func);
+      x87_fst_or_nop(cp->func, writemask, 0, dst);
+      x87_fstp_or_pop(cp->func, writemask, 3, dst);
+   }
+
+   return TRUE;
+}
+#endif
+
+
+
+static boolean emit_MAX( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+
+   sse_maxps(cp->func, dst, arg1);
+
+   store_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+
+static boolean emit_MIN( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+
+   sse_minps(cp->func, dst, arg1);
+
+   store_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+static boolean emit_MOV( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+
+   /* potentially nothing to do */
+
+   store_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+static boolean emit_MUL( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+
+   sse_mulps(cp->func, dst, arg1);
+
+   store_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+
+static boolean emit_MAD( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg arg2 = fetch_src(cp, &op->FullSrcRegisters[2]);
+
+   /* If we can't clobber old contents of arg0, get a temporary & copy
+    * it there, then clobber it...
+    */
+   arg0 = get_xmm_writable(cp, arg0);
+
+   sse_mulps(cp->func, arg0, arg1);
+   sse_addps(cp->func, arg0, arg2);
+   store_dest(cp, &op->FullDstRegisters[0], arg0);
+   return TRUE;
+}
+
+
+
+/* A wrapper for powf().
+ * Makes sure it is cdecl and operates on floats.
+ */
+static float PIPE_CDECL _powerf( float x, float y )
+{
+#if FAST_MATH
+   return util_fast_pow(x, y);
+#else
+   return powf( x, y );
+#endif
+}
+
+#if FAST_MATH
+static float PIPE_CDECL _exp2(float x)
+{
+   return util_fast_exp2(x);
+}
+#endif
+
+
+/* Really not sufficient -- need to check for conditions that could
+ * generate inf/nan values, which will slow things down hugely.
+ */
+static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+#if 0
+   x87_fld_src(cp, &op->FullSrcRegisters[1], 0);  /* a1.x */
+   x87_fld_src(cp, &op->FullSrcRegisters[0], 0);	/* a0.x a1.x */
+   x87_fyl2x(cp->func);	                                /* a1*log2(a0) */
+
+   x87_emit_ex2( cp );		/* 2^(a1*log2(a0)) */
+
+   x87_fstp_dest4(cp, &op->FullDstRegisters[0]);
+#else
+   uint i;
+
+   /* For absolute correctness, need to spill/invalidate all XMM regs
+    * too.  
+    */
+   for (i = 0; i < 8; i++) {
+      if (cp->xmm[i].dirty) 
+         spill(cp, i);
+      aos_release_xmm_reg(cp, i);
+   }
+
+   /* Push caller-save (ie scratch) regs.  
+    */
+   x86_cdecl_caller_push_regs( cp->func );
+
+   x86_lea( cp->func, cp->stack_ESP, x86_make_disp(cp->stack_ESP, -8) );
+
+   x87_fld_src( cp, &op->FullSrcRegisters[1], 0 );
+   x87_fstp( cp->func, x86_make_disp( cp->stack_ESP, 4 ) );
+   x87_fld_src( cp, &op->FullSrcRegisters[0], 0 );
+   x87_fstp( cp->func, x86_make_disp( cp->stack_ESP, 0 ) );
+
+   /* tmp_EAX has been pushed & will be restored below */
+   x86_mov_reg_imm( cp->func, cp->tmp_EAX, (unsigned long) _powerf );
+   x86_call( cp->func, cp->tmp_EAX );
+
+   x86_lea( cp->func, cp->stack_ESP, x86_make_disp(cp->stack_ESP, 8) );
+
+   x86_cdecl_caller_pop_regs( cp->func );
+
+   /* Note retval on x87 stack:
+    */
+   cp->func->x87_stack++;
+
+   x87_fstp_dest4( cp, &op->FullDstRegisters[0] );
+#endif
+   return TRUE;
+}
+
+
+#if FAST_MATH
+static boolean emit_EXPBASE2( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   uint i;
+
+   /* For absolute correctness, need to spill/invalidate all XMM regs
+    * too.  
+    */
+   for (i = 0; i < 8; i++) {
+      if (cp->xmm[i].dirty) 
+         spill(cp, i);
+      aos_release_xmm_reg(cp, i);
+   }
+
+   /* Push caller-save (ie scratch) regs.  
+    */
+   x86_cdecl_caller_push_regs( cp->func );
+
+   x86_lea( cp->func, cp->stack_ESP, x86_make_disp(cp->stack_ESP, -4) );
+
+   x87_fld_src( cp, &op->FullSrcRegisters[0], 0 );
+   x87_fstp( cp->func, x86_make_disp( cp->stack_ESP, 0 ) );
+
+   /* tmp_EAX has been pushed & will be restored below */
+   x86_mov_reg_imm( cp->func, cp->tmp_EAX, (unsigned long) _exp2 );
+   x86_call( cp->func, cp->tmp_EAX );
+
+   x86_lea( cp->func, cp->stack_ESP, x86_make_disp(cp->stack_ESP, 4) );
+
+   x86_cdecl_caller_pop_regs( cp->func );
+
+   /* Note retval on x87 stack:
+    */
+   cp->func->x87_stack++;
+
+   x87_fstp_dest4( cp, &op->FullDstRegisters[0] );
+
+   return TRUE;
+}
+#endif
+
+
+static boolean emit_RCP( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg dst = aos_get_xmm_reg(cp);
+
+   if (cp->have_sse2) {
+      sse2_rcpss(cp->func, dst, arg0);
+      /* extend precision here...
+       */
+   }
+   else {
+      struct x86_reg ones = aos_get_internal(cp, IMM_ONES);
+      sse_movss(cp->func, dst, ones);
+      sse_divss(cp->func, dst, arg0);
+   }
+
+   store_scalar_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+
+/* Although rsqrtps() and rcpps() are low precision on some/all SSE
+ * implementations, it is possible to improve its precision at
+ * fairly low cost, using a newton/raphson step, as below:
+ * 
+ * x1 = 2 * rcpps(a) - a * rcpps(a) * rcpps(a)
+ * x1 = 0.5 * rsqrtps(a) * [3.0 - (a * rsqrtps(a))* rsqrtps(a)]
+ * or:
+ *   x1 = rsqrtps(a) * [1.5 - .5 * a * rsqrtps(a) * rsqrtps(a)]
+ * 
+ *
+ * See: http://softwarecommunity.intel.com/articles/eng/1818.htm
+ */
+static boolean emit_RSQ( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+
+   if (0) {
+      struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+      struct x86_reg r = aos_get_xmm_reg(cp);
+      sse_rsqrtss(cp->func, r, arg0);
+      store_scalar_dest(cp, &op->FullDstRegisters[0], r);
+      return TRUE;
+   }
+   else {
+      struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+      struct x86_reg r = aos_get_xmm_reg(cp);
+
+      struct x86_reg neg_half       = get_reg_ptr( cp, AOS_FILE_INTERNAL, IMM_RSQ );
+      struct x86_reg one_point_five = x86_make_disp( neg_half, 4 );
+      struct x86_reg src            = get_xmm_writable( cp, arg0 );
+      
+      sse_rsqrtss( cp->func, r, src  );             /* rsqrtss(a) */
+      sse_mulss(   cp->func, src, neg_half  );      /* -.5 * a */
+      sse_mulss(   cp->func, src,  r );             /* -.5 * a * r */
+      sse_mulss(   cp->func, src,  r );             /* -.5 * a * r * r */
+      sse_addss(   cp->func, src, one_point_five ); /* 1.5 - .5 * a * r * r */
+      sse_mulss(   cp->func, r,  src );             /* r * (1.5 - .5 * a * r * r) */
+
+      store_scalar_dest(cp, &op->FullDstRegisters[0], r);
+      return TRUE;
+   }
+}
+
+
+static boolean emit_SGE( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg ones = aos_get_internal(cp, IMM_ONES);
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+
+   sse_cmpps(cp->func, dst, arg1, cc_NotLessThan);
+   sse_andps(cp->func, dst, ones);
+
+   store_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+static boolean emit_SIN( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   x87_fld_src(cp, &op->FullSrcRegisters[0], 0);
+   x87_fsin(cp->func);
+   x87_fstp_dest4(cp, &op->FullDstRegisters[0]);
+   return TRUE;
+}
+
+
+
+static boolean emit_SLT( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg ones = aos_get_internal(cp, IMM_ONES);
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+   
+   sse_cmpps(cp->func, dst, arg1, cc_LessThan);
+   sse_andps(cp->func, dst, ones);
+
+   store_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+static boolean emit_SUB( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg dst = get_xmm_writable(cp, arg0);
+
+   sse_subps(cp->func, dst, arg1);
+
+   store_dest(cp, &op->FullDstRegisters[0], dst);
+   return TRUE;
+}
+
+static boolean emit_TRUNC( struct aos_compilation *cp, const struct tgsi_full_instruction *op )
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg tmp0 = aos_get_xmm_reg(cp);
+
+   sse2_cvttps2dq(cp->func, tmp0, arg0);
+   sse2_cvtdq2ps(cp->func, tmp0, tmp0);
+
+   store_dest(cp, &op->FullDstRegisters[0], tmp0);
+   return TRUE;
+}
+
+static boolean emit_XPD( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) 
+{
+   struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]);
+   struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]);
+   struct x86_reg tmp0 = aos_get_xmm_reg(cp);
+   struct x86_reg tmp1 = aos_get_xmm_reg(cp);
+
+   emit_pshufd(cp, tmp1, arg1, SHUF(Y, Z, X, W));
+   sse_mulps(cp->func, tmp1, arg0);
+   emit_pshufd(cp, tmp0, arg0, SHUF(Y, Z, X, W));
+   sse_mulps(cp->func, tmp0, arg1);
+   sse_subps(cp->func, tmp1, tmp0);
+   sse_shufps(cp->func, tmp1, tmp1, SHUF(Y, Z, X, W));
+
+/*    dst[2] = arg0[0] * arg1[1] - arg0[1] * arg1[0]; */
+/*    dst[0] = arg0[1] * arg1[2] - arg0[2] * arg1[1]; */
+/*    dst[1] = arg0[2] * arg1[0] - arg0[0] * arg1[2]; */
+/*    dst[3] is undef */
+
+
+   aos_release_xmm_reg(cp, tmp0.idx);
+   store_dest(cp, &op->FullDstRegisters[0], tmp1);
+   return TRUE;
+}
+
+
+
+static boolean
+emit_instruction( struct aos_compilation *cp,
+                  struct tgsi_full_instruction *inst )
+{
+   x87_assert_stack_empty(cp->func);
+
+   switch( inst->Instruction.Opcode ) {
+   case TGSI_OPCODE_MOV:
+      return emit_MOV( cp, inst );
+
+   case TGSI_OPCODE_LIT:
+      return emit_LIT(cp, inst);
+
+   case TGSI_OPCODE_RCP:
+      return emit_RCP(cp, inst);
+
+   case TGSI_OPCODE_RSQ:
+      return emit_RSQ(cp, inst);
+
+   case TGSI_OPCODE_EXP:
+      /*return emit_EXP(cp, inst);*/
+      return FALSE;
+
+   case TGSI_OPCODE_LOG:
+      /*return emit_LOG(cp, inst);*/
+      return FALSE;
+
+   case TGSI_OPCODE_MUL:
+      return emit_MUL(cp, inst);
+
+   case TGSI_OPCODE_ADD:
+      return emit_ADD(cp, inst);
+
+   case TGSI_OPCODE_DP3:
+      return emit_DP3(cp, inst);
+
+   case TGSI_OPCODE_DP4:
+      return emit_DP4(cp, inst);
+
+   case TGSI_OPCODE_DST:
+      return emit_DST(cp, inst);
+
+   case TGSI_OPCODE_MIN:
+      return emit_MIN(cp, inst);
+
+   case TGSI_OPCODE_MAX:
+      return emit_MAX(cp, inst);
+
+   case TGSI_OPCODE_SLT:
+      return emit_SLT(cp, inst);
+
+   case TGSI_OPCODE_SGE:
+      return emit_SGE(cp, inst);
+
+   case TGSI_OPCODE_MAD:
+      return emit_MAD(cp, inst);
+
+   case TGSI_OPCODE_SUB:
+      return emit_SUB(cp, inst);
+ 
+   case TGSI_OPCODE_LERP:
+//      return emit_LERP(cp, inst);
+      return FALSE;
+
+   case TGSI_OPCODE_FRAC:
+      return emit_FRC(cp, inst);
+
+   case TGSI_OPCODE_CLAMP:
+//      return emit_CLAMP(cp, inst);
+      return FALSE;
+
+   case TGSI_OPCODE_FLOOR:
+      return emit_FLR(cp, inst);
+
+   case TGSI_OPCODE_ROUND:
+      return emit_RND(cp, inst);
+
+   case TGSI_OPCODE_EXPBASE2:
+#if FAST_MATH
+      return emit_EXPBASE2(cp, inst);
+#elif 0
+      /* this seems to fail for "larger" exponents.
+       * See glean tvertProg1's EX2 test.
+       */
+      return emit_EX2(cp, inst);
+#else
+      return FALSE;
+#endif
+
+   case TGSI_OPCODE_LOGBASE2:
+      return emit_LG2(cp, inst);
+
+   case TGSI_OPCODE_POWER:
+      return emit_POW(cp, inst);
+
+   case TGSI_OPCODE_CROSSPRODUCT:
+      return emit_XPD(cp, inst);
+
+   case TGSI_OPCODE_ABS:
+      return emit_ABS(cp, inst);
+
+   case TGSI_OPCODE_DPH:
+      return emit_DPH(cp, inst);
+
+   case TGSI_OPCODE_COS:
+      return emit_COS(cp, inst);
+
+   case TGSI_OPCODE_SIN:
+      return emit_SIN(cp, inst);
+
+   case TGSI_OPCODE_TRUNC:
+      return emit_TRUNC(cp, inst);
+
+   case TGSI_OPCODE_END:
+      return TRUE;
+
+   default:
+      return FALSE;
+   }
+}
+
+
+static boolean emit_viewport( struct aos_compilation *cp )
+{
+   struct x86_reg pos = aos_get_shader_reg_xmm(cp, 
+                                               TGSI_FILE_OUTPUT, 
+                                               cp->vaos->draw->vs.position_output );
+
+   struct x86_reg scale = x86_make_disp(cp->machine_EDX, 
+                                        Offset(struct aos_machine, scale));
+
+   struct x86_reg translate = x86_make_disp(cp->machine_EDX, 
+                                        Offset(struct aos_machine, translate));
+
+   sse_mulps(cp->func, pos, scale);
+   sse_addps(cp->func, pos, translate);
+
+   aos_adopt_xmm_reg( cp,
+                      pos,
+                      TGSI_FILE_OUTPUT,
+                      cp->vaos->draw->vs.position_output,
+                      TRUE );
+   return TRUE;
+}
+
+
+/* This is useful to be able to see the results on softpipe.  Doesn't
+ * do proper clipping, just assumes the backend can do it during
+ * rasterization -- for debug only...
+ */
+static boolean emit_rhw_viewport( struct aos_compilation *cp )
+{
+   struct x86_reg tmp = aos_get_xmm_reg(cp);
+   struct x86_reg pos = aos_get_shader_reg_xmm(cp, 
+                                               TGSI_FILE_OUTPUT, 
+                                               cp->vaos->draw->vs.position_output);
+
+   struct x86_reg scale = x86_make_disp(cp->machine_EDX, 
+                                        Offset(struct aos_machine, scale));
+
+   struct x86_reg translate = x86_make_disp(cp->machine_EDX, 
+                                        Offset(struct aos_machine, translate));
+
+
+
+   emit_pshufd(cp, tmp, pos, SHUF(W, W, W, W));
+   sse2_rcpss(cp->func, tmp, tmp);
+   sse_shufps(cp->func, tmp, tmp, SHUF(X, X, X, X));
+   
+   sse_mulps(cp->func, pos, scale);
+   sse_mulps(cp->func, pos, tmp);
+   sse_addps(cp->func, pos, translate);
+
+   /* Set pos[3] = w 
+    */
+   mask_write(cp, pos, tmp, TGSI_WRITEMASK_W);
+
+   aos_adopt_xmm_reg( cp,
+                      pos,
+                      TGSI_FILE_OUTPUT,
+                      cp->vaos->draw->vs.position_output,
+                      TRUE );
+   return TRUE;
+}
+
+
+#if 0
+static boolean note_immediate( struct aos_compilation *cp,
+                               struct tgsi_full_immediate *imm )
+{
+   unsigned pos = cp->num_immediates++;
+   unsigned j;
+
+   for (j = 0; j < imm->Immediate.NrTokens - 1; j++) {
+      cp->vaos->machine->immediate[pos][j] = imm->u.ImmediateFloat32[j].Float;
+   }
+
+   return TRUE;
+}
+#endif
+
+
+
+
+static void find_last_write_outputs( struct aos_compilation *cp )
+{
+   struct tgsi_parse_context parse;
+   unsigned this_instruction = 0;
+   unsigned i;
+
+   tgsi_parse_init( &parse, cp->vaos->base.vs->state.tokens );
+
+   while (!tgsi_parse_end_of_tokens( &parse )) {
+      
+      tgsi_parse_token( &parse );
+
+      if (parse.FullToken.Token.Type != TGSI_TOKEN_TYPE_INSTRUCTION) 
+         continue;
+
+      for (i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++) {
+         if (parse.FullToken.FullInstruction.FullDstRegisters[i].DstRegister.File ==
+             TGSI_FILE_OUTPUT) 
+         {
+            unsigned idx = parse.FullToken.FullInstruction.FullDstRegisters[i].DstRegister.Index;
+            cp->output_last_write[idx] = this_instruction;
+         }
+      }
+
+      this_instruction++;
+   }
+
+   tgsi_parse_free( &parse );
+}
+
+
+#define ARG_MACHINE    1
+#define ARG_START_ELTS 2
+#define ARG_COUNT      3
+#define ARG_OUTBUF     4
+
+
+static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient,
+                                     boolean linear )
+{ 
+   struct tgsi_parse_context parse;
+   struct aos_compilation cp;
+   unsigned fixup, label;
+
+   util_init_math();
+
+   tgsi_parse_init( &parse, varient->base.vs->state.tokens );
+
+   memset(&cp, 0, sizeof(cp));
+
+   cp.insn_counter = 1;
+   cp.vaos = varient;
+   cp.have_sse2 = 1;
+   cp.func = &varient->func[ linear ? 0 : 1 ];
+
+   cp.tmp_EAX       = x86_make_reg(file_REG32, reg_AX);
+   cp.idx_EBX      = x86_make_reg(file_REG32, reg_BX);
+   cp.outbuf_ECX    = x86_make_reg(file_REG32, reg_CX);
+   cp.machine_EDX   = x86_make_reg(file_REG32, reg_DX);
+   cp.count_ESI     = x86_make_reg(file_REG32, reg_SI);
+   cp.temp_EBP     = x86_make_reg(file_REG32, reg_BP);
+   cp.stack_ESP = x86_make_reg( file_REG32, reg_SP );
+
+   x86_init_func(cp.func);
+
+   find_last_write_outputs(&cp);
+
+   x86_push(cp.func, cp.idx_EBX);
+   x86_push(cp.func, cp.count_ESI);
+   x86_push(cp.func, cp.temp_EBP);
+
+
+   /* Load arguments into regs:
+    */
+   x86_mov(cp.func, cp.machine_EDX, x86_fn_arg(cp.func, ARG_MACHINE));
+   x86_mov(cp.func, cp.idx_EBX, x86_fn_arg(cp.func, ARG_START_ELTS));
+   x86_mov(cp.func, cp.count_ESI, x86_fn_arg(cp.func, ARG_COUNT));
+   x86_mov(cp.func, cp.outbuf_ECX, x86_fn_arg(cp.func, ARG_OUTBUF));
+
+
+   /* Compare count to zero and possibly bail.
+    */
+   x86_xor(cp.func, cp.tmp_EAX, cp.tmp_EAX);
+   x86_cmp(cp.func, cp.count_ESI, cp.tmp_EAX);
+   fixup = x86_jcc_forward(cp.func, cc_E);
+
+
+   save_fpu_state( &cp );
+   set_fpu_round_nearest( &cp );
+
+   aos_init_inputs( &cp, linear );
+
+   cp.x86_reg[0] = 0;
+   cp.x86_reg[1] = 0;
+   
+   /* Note address for loop jump 
+    */
+   label = x86_get_label(cp.func);
+   {
+      /* Fetch inputs...  TODO:  fetch lazily...
+       */
+      if (!aos_fetch_inputs( &cp, linear ))
+         goto fail;
+
+      /* Emit the shader:
+       */
+      while( !tgsi_parse_end_of_tokens( &parse ) && !cp.error ) 
+      {
+         tgsi_parse_token( &parse );
+
+         switch (parse.FullToken.Token.Type) {
+         case TGSI_TOKEN_TYPE_IMMEDIATE:
+#if 0
+            if (!note_immediate( &cp, &parse.FullToken.FullImmediate ))
+               goto fail;
+#endif
+            break;
+
+         case TGSI_TOKEN_TYPE_INSTRUCTION:
+            if (DISASSEM)
+               tgsi_dump_instruction( &parse.FullToken.FullInstruction, cp.insn_counter );
+
+            if (!emit_instruction( &cp, &parse.FullToken.FullInstruction ))
+               goto fail;
+            break;
+         }
+
+         x87_assert_stack_empty(cp.func);
+         cp.insn_counter++;
+
+         if (DISASSEM)
+            debug_printf("\n");
+      }
+
+   
+      {
+         unsigned i;
+         for (i = 0; i < 8; i++) {
+            if (cp.xmm[i].file != TGSI_FILE_OUTPUT) {
+               cp.xmm[i].file = TGSI_FILE_NULL;
+               cp.xmm[i].dirty = 0;
+            }
+         }
+      }
+
+      if (cp.error)
+         goto fail;
+
+      if (cp.vaos->base.key.clip) {
+         /* not really handling clipping, just do the rhw so we can
+          * see the results...
+          */
+         emit_rhw_viewport(&cp); 
+      }
+      else if (cp.vaos->base.key.viewport) {
+         emit_viewport(&cp);
+      }
+
+      /* Emit output...  TODO: do this eagerly after the last write to a
+       * given output.
+       */
+      if (!aos_emit_outputs( &cp ))
+         goto fail;
+
+
+      /* Next vertex:
+       */
+      x86_lea(cp.func, 
+              cp.outbuf_ECX, 
+              x86_make_disp(cp.outbuf_ECX, 
+                            cp.vaos->base.key.output_stride));
+
+      /* Incr index
+       */   
+      aos_incr_inputs( &cp, linear );
+   }
+   /* decr count, loop if not zero
+    */
+   x86_dec(cp.func, cp.count_ESI);
+   x86_jcc(cp.func, cc_NZ, label);
+
+   restore_fpu_state(&cp);
+
+   /* Land forward jump here:
+    */
+   x86_fixup_fwd_jump(cp.func, fixup);
+
+   /* Exit mmx state?
+    */
+   if (cp.func->need_emms)
+      mmx_emms(cp.func);
+
+   x86_pop(cp.func, cp.temp_EBP);
+   x86_pop(cp.func, cp.count_ESI);
+   x86_pop(cp.func, cp.idx_EBX);
+
+   x87_assert_stack_empty(cp.func);
+   x86_ret(cp.func);
+
+   tgsi_parse_free( &parse );
+   return !cp.error;
+
+ fail:
+   tgsi_parse_free( &parse );
+   return FALSE;
+}
+
+
+
+static void vaos_set_buffer( struct draw_vs_varient *varient,
+                             unsigned buf,
+                             const void *ptr,
+                             unsigned stride )
+{
+   struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient;
+
+   if (buf < vaos->nr_vb) {
+      vaos->buffer[buf].base_ptr = (char *)ptr;
+      vaos->buffer[buf].stride = stride;
+   }
+
+   if (0) debug_printf("%s %d/%d: %p %d\n", __FUNCTION__, buf, vaos->nr_vb, ptr, stride);
+}
+
+
+
+static void PIPE_CDECL vaos_run_elts( struct draw_vs_varient *varient,
+                                      const unsigned *elts,
+                                      unsigned count,
+                                      void *output_buffer )
+{
+   struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient;
+   struct aos_machine *machine = vaos->draw->vs.aos_machine;
+
+   if (0) debug_printf("%s %d\n", __FUNCTION__, count);
+
+   machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size;
+   machine->constants = vaos->draw->vs.aligned_constants;
+   machine->immediates = vaos->base.vs->immediates;
+   machine->buffer = vaos->buffer;
+
+   vaos->gen_run_elts( machine,
+                       elts,
+                       count,
+                       output_buffer );
+}
+
+static void PIPE_CDECL vaos_run_linear( struct draw_vs_varient *varient,
+                                        unsigned start,
+                                        unsigned count,
+                                        void *output_buffer )
+{
+   struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient;
+   struct aos_machine *machine = vaos->draw->vs.aos_machine;
+
+   if (0) debug_printf("%s %d %d const: %x\n", __FUNCTION__, start, count, 
+                       vaos->base.key.const_vbuffers);
+
+   machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size;
+   machine->constants = vaos->draw->vs.aligned_constants;
+   machine->immediates = vaos->base.vs->immediates;
+   machine->buffer = vaos->buffer;
+
+   vaos->gen_run_linear( machine,
+                         start,
+                         count,
+                         output_buffer );
+
+   /* Sanity spot checks to make sure we didn't trash our constants */
+   assert(machine->internal[IMM_ONES][0] == 1.0f);
+   assert(machine->internal[IMM_IDENTITY][0] == 0.0f);
+   assert(machine->internal[IMM_NEGS][0] == -1.0f);
+}
+
+
+
+static void vaos_destroy( struct draw_vs_varient *varient )
+{
+   struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient;
+
+   FREE( vaos->buffer );
+
+   x86_release_func( &vaos->func[0] );
+   x86_release_func( &vaos->func[1] );
+
+   FREE(vaos);
+}
+
+
+
+static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs,
+                                                 const struct draw_vs_varient_key *key )
+{
+   unsigned i;
+   struct draw_vs_varient_aos_sse *vaos = CALLOC_STRUCT(draw_vs_varient_aos_sse);
+
+   if (!vaos)
+      goto fail;
+   
+   vaos->base.key = *key;
+   vaos->base.vs = vs;
+   vaos->base.set_buffer = vaos_set_buffer;
+   vaos->base.destroy = vaos_destroy;
+   vaos->base.run_linear = vaos_run_linear;
+   vaos->base.run_elts = vaos_run_elts;
+
+   vaos->draw = vs->draw;
+
+   for (i = 0; i < key->nr_inputs; i++) 
+      vaos->nr_vb = MAX2( vaos->nr_vb, key->element[i].in.buffer + 1 );
+
+   vaos->buffer = MALLOC( vaos->nr_vb * sizeof(vaos->buffer[0]) );
+   if (!vaos->buffer)
+      goto fail;
+
+   if (0)
+      debug_printf("nr_vb: %d const: %x\n", vaos->nr_vb, vaos->base.key.const_vbuffers);
+
+#if 0
+   tgsi_dump(vs->state.tokens, 0);
+#endif
+
+   if (!build_vertex_program( vaos, TRUE ))
+      goto fail;
+
+   if (!build_vertex_program( vaos, FALSE ))
+      goto fail;
+
+   vaos->gen_run_linear = (vaos_run_linear_func)x86_get_func(&vaos->func[0]);
+   if (!vaos->gen_run_linear)
+      goto fail;
+
+   vaos->gen_run_elts = (vaos_run_elts_func)x86_get_func(&vaos->func[1]);
+   if (!vaos->gen_run_elts)
+      goto fail;
+
+   return &vaos->base;
+
+ fail:
+   if (vaos && vaos->buffer)
+      FREE(vaos->buffer);
+
+   if (vaos)
+      x86_release_func( &vaos->func[0] );
+
+   if (vaos)
+      x86_release_func( &vaos->func[1] );
+
+   FREE(vaos);
+   
+   return NULL;
+}
+
+
+struct draw_vs_varient *draw_vs_varient_aos_sse( struct draw_vertex_shader *vs,
+                                                 const struct draw_vs_varient_key *key )
+{
+   struct draw_vs_varient *varient = varient_aos_sse( vs, key );
+
+   if (varient == NULL) {
+      varient = draw_vs_varient_generic( vs, key );
+   }
+
+   return varient;
+}
+
+
+
+#endif /* PIPE_ARCH_X86 */
diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h
new file mode 100644
index 0000000000..264387517b
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_aos.h
@@ -0,0 +1,253 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef DRAW_VS_AOS_H
+#define DRAW_VS_AOS_H
+
+#include "pipe/p_config.h"
+
+#ifdef PIPE_ARCH_X86
+
+struct tgsi_token;
+struct x86_function;
+
+#include "pipe/p_state.h"
+#include "rtasm/rtasm_x86sse.h"
+
+
+
+
+
+#define X    0
+#define Y    1
+#define Z    2
+#define W    3
+
+#define MAX_INPUTS     PIPE_MAX_ATTRIBS
+#define MAX_OUTPUTS    PIPE_MAX_SHADER_OUTPUTS
+#define MAX_TEMPS      TGSI_EXEC_NUM_TEMPS
+#define MAX_CONSTANTS  1024  /** only used for sanity checking */
+#define MAX_IMMEDIATES 1024  /** only used for sanity checking */
+#define MAX_INTERNALS  8     /** see IMM_x values below */
+
+#define AOS_FILE_INTERNAL TGSI_FILE_COUNT
+
+#define FPU_RND_NEG    1
+#define FPU_RND_NEAREST 2
+
+struct aos_machine;
+typedef void (PIPE_CDECL *lit_func)( struct aos_machine *,
+                                    float *result,
+                                    const float *in,
+                                    unsigned count );
+
+void PIPE_CDECL aos_do_lit( struct aos_machine *machine,
+                            float *result,
+                            const float *in,
+                            unsigned count );
+
+struct shine_tab {
+   float exponent;
+   float values[258];
+   unsigned last_used;
+};
+
+struct lit_info {
+   lit_func func;
+   struct shine_tab *shine_tab;
+};
+
+#define MAX_SHINE_TAB    4
+#define MAX_LIT_INFO     16
+
+struct aos_buffer {
+   const void *base_ptr;
+   unsigned stride;
+   void *ptr;                   /* updated per vertex */
+};
+
+
+
+
+/* This is the temporary storage used by all the aos_sse vs varients.
+ * Create one per context and reuse by passing a pointer in at
+ * vs_varient creation??
+ */
+struct aos_machine {
+   float input    [MAX_INPUTS    ][4];
+   float output   [MAX_OUTPUTS   ][4];
+   float temp     [MAX_TEMPS     ][4];
+   float internal [MAX_INTERNALS ][4];
+
+   float scale[4];              /* viewport */
+   float translate[4];          /* viewport */
+
+   float tmp[2][4];             /* scratch space for LIT */
+
+   struct shine_tab shine_tab[MAX_SHINE_TAB];
+   struct lit_info  lit_info[MAX_LIT_INFO];
+   unsigned now;
+   
+
+   ushort fpu_rnd_nearest;
+   ushort fpu_rnd_neg_inf;
+   ushort fpu_restore;
+   ushort fpucntl;              /* one of FPU_* above */
+
+   const float (*immediates)[4];     /* points to shader data */
+   const float (*constants)[4];      /* points to draw data */
+
+   const struct aos_buffer *buffer; /* points to ? */
+};
+
+
+
+
+struct aos_compilation {
+   struct x86_function *func;
+   struct draw_vs_varient_aos_sse *vaos;
+
+   unsigned insn_counter;
+   unsigned num_immediates;
+   unsigned count;
+   unsigned lit_count;
+
+   struct {
+      unsigned idx:16;
+      unsigned file:8;
+      unsigned dirty:8;
+      unsigned last_used;
+   } xmm[8];
+
+   unsigned x86_reg[2];                /* one of X86_* */
+
+   boolean input_fetched[PIPE_MAX_ATTRIBS];
+   unsigned output_last_write[PIPE_MAX_ATTRIBS];
+
+   boolean have_sse2;
+   boolean error;
+   short fpucntl;
+
+   /* these are actually known values, but putting them in a struct
+    * like this is helpful to keep them in sync across the file.
+    */
+   struct x86_reg tmp_EAX;
+   struct x86_reg idx_EBX;     /* either start+i or &elt[i] */
+   struct x86_reg outbuf_ECX;
+   struct x86_reg machine_EDX;
+   struct x86_reg count_ESI;    /* decrements to zero */
+   struct x86_reg temp_EBP;
+   struct x86_reg stack_ESP;
+};
+
+struct x86_reg aos_get_xmm_reg( struct aos_compilation *cp );
+void aos_release_xmm_reg( struct aos_compilation *cp, unsigned idx );
+
+void aos_adopt_xmm_reg( struct aos_compilation *cp,
+                        struct x86_reg reg,
+                        unsigned file,
+                        unsigned idx,
+                        unsigned dirty );
+
+void aos_spill_all( struct aos_compilation *cp );
+
+struct x86_reg aos_get_shader_reg( struct aos_compilation *cp, 
+                                   unsigned file,
+                                   unsigned idx );
+
+boolean aos_init_inputs( struct aos_compilation *cp, boolean linear );
+boolean aos_fetch_inputs( struct aos_compilation *cp, boolean linear );
+boolean aos_incr_inputs( struct aos_compilation *cp, boolean linear );
+
+boolean aos_emit_outputs( struct aos_compilation *cp );
+
+
+#define IMM_ONES     0              /* 1, 1,1,1 */
+#define IMM_SWZ      1              /* 1,-1,0, 0xffffffff */
+#define IMM_IDENTITY 2              /* 0, 0,0,1 */
+#define IMM_INV_255  3              /* 1/255, 1/255, 1/255, 1/255 */
+#define IMM_255      4              /* 255, 255, 255, 255 */
+#define IMM_NEGS     5              /* -1,-1,-1,-1 */
+#define IMM_RSQ      6              /* -.5,1.5,_,_ */
+#define IMM_PSIZE    7              /* not really an immediate - updated each run */
+
+struct x86_reg aos_get_internal( struct aos_compilation *cp,
+                                 unsigned imm );
+struct x86_reg aos_get_internal_xmm( struct aos_compilation *cp,
+                                     unsigned imm );
+
+
+#define ERROR(cp, msg)                                                  \
+do {                                                                    \
+   if (0) debug_printf("%s: x86 translation failed: %s\n", __FUNCTION__, msg); \
+   cp->error = 1;                                                       \
+} while (0)
+
+
+#define X86_NULL       0
+#define X86_IMMEDIATES 1
+#define X86_CONSTANTS  2
+#define X86_BUFFERS    3
+
+struct x86_reg aos_get_x86( struct aos_compilation *cp,
+                            unsigned which_reg,
+                            unsigned value );
+
+
+typedef void (PIPE_CDECL *vaos_run_elts_func)( struct aos_machine *,
+                                               const unsigned *elts,
+                                               unsigned count,
+                                               void *output_buffer);
+
+typedef void (PIPE_CDECL *vaos_run_linear_func)( struct aos_machine *,
+                                                unsigned start,
+                                                unsigned count,
+                                                void *output_buffer);
+
+
+struct draw_vs_varient_aos_sse {
+   struct draw_vs_varient base;
+   struct draw_context *draw;
+
+   struct aos_buffer *buffer;
+   unsigned nr_vb;
+
+   vaos_run_linear_func gen_run_linear;
+   vaos_run_elts_func gen_run_elts;
+
+
+   struct x86_function func[2];
+};
+
+
+#endif
+
+#endif 
+
diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c
new file mode 100644
index 0000000000..39f75b50b7
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c
@@ -0,0 +1,462 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi/tgsi_exec.h"
+#include "draw_vs.h"
+#include "draw_vs_aos.h"
+#include "draw_vertex.h"
+
+#include "rtasm/rtasm_x86sse.h"
+
+#ifdef PIPE_ARCH_X86
+
+/* Note - don't yet have to worry about interacting with the code in
+ * draw_vs_aos.c as there is no intermingling of generated code...
+ * That may have to change, we'll see.
+ */
+static void emit_load_R32G32B32A32( struct aos_compilation *cp, 			   
+				    struct x86_reg data,
+				    struct x86_reg src_ptr )
+{
+   sse_movups(cp->func, data, src_ptr);
+}
+
+static void emit_load_R32G32B32( struct aos_compilation *cp, 			   
+				 struct x86_reg data,
+				 struct x86_reg src_ptr )
+{
+#if 1
+   sse_movss(cp->func, data, x86_make_disp(src_ptr, 8));
+   /* data = z ? ? ? */
+   sse_shufps(cp->func, data, aos_get_internal_xmm( cp, IMM_IDENTITY ), SHUF(X,Y,Z,W) );
+   /* data = z ? 0 1 */
+   sse_shufps(cp->func, data, data, SHUF(Y,Z,X,W) );
+   /* data = ? 0 z 1 */
+   sse_movlps(cp->func, data, src_ptr);
+   /* data = x y z 1 */
+#else
+   sse_movups(cp->func, data, src_ptr);
+   /* data = x y z ? */
+   sse2_pshufd(cp->func, data, data, SHUF(W,X,Y,Z) );
+   /* data = ? x y z */
+   sse_movss(cp->func, data, aos_get_internal_xmm( cp, IMM_ONES ) );
+   /* data = 1 x y z */
+   sse2_pshufd(cp->func, data, data, SHUF(Y,Z,W,X) );
+   /* data = x y z 1 */
+#endif
+}
+
+static void emit_load_R32G32( struct aos_compilation *cp, 
+			   struct x86_reg data,
+			   struct x86_reg src_ptr )
+{
+   sse_movups(cp->func, data, aos_get_internal_xmm( cp, IMM_IDENTITY ) );
+   sse_movlps(cp->func, data, src_ptr);
+}
+
+
+static void emit_load_R32( struct aos_compilation *cp, 
+			   struct x86_reg data,
+			   struct x86_reg src_ptr )
+{
+   sse_movss(cp->func, data, src_ptr);
+   sse_orps(cp->func, data, aos_get_internal_xmm( cp, IMM_IDENTITY ) );
+}
+
+
+static void emit_load_R8G8B8A8_UNORM( struct aos_compilation *cp,
+				       struct x86_reg data,
+				       struct x86_reg src_ptr )
+{
+   sse_movss(cp->func, data, src_ptr);
+   sse2_punpcklbw(cp->func, data, aos_get_internal_xmm( cp, IMM_IDENTITY ));
+   sse2_punpcklbw(cp->func, data, aos_get_internal_xmm( cp, IMM_IDENTITY ));
+   sse2_cvtdq2ps(cp->func, data, data);
+   sse_mulps(cp->func, data, aos_get_internal(cp, IMM_INV_255));
+}
+
+
+
+/* Extended swizzles?  Maybe later.
+ */  
+static void emit_swizzle( struct aos_compilation *cp,
+			  struct x86_reg dest,
+			  struct x86_reg src,
+			  ubyte shuffle )
+{
+   sse_shufps(cp->func, dest, src, shuffle);
+}
+
+
+
+static boolean get_buffer_ptr( struct aos_compilation *cp,
+                               boolean linear,
+                               unsigned buf_idx,
+                               struct x86_reg elt,
+                               struct x86_reg ptr)
+{
+   struct x86_reg buf = x86_make_disp(aos_get_x86( cp, 0, X86_BUFFERS ), 
+                                      buf_idx * sizeof(struct aos_buffer));
+
+   struct x86_reg buf_stride = x86_make_disp(buf, 
+                                             Offset(struct aos_buffer, stride));
+   if (linear) {
+      struct x86_reg buf_ptr = x86_make_disp(buf, 
+                                             Offset(struct aos_buffer, ptr));
+
+
+      /* Calculate pointer to current attrib:
+       */
+      x86_mov(cp->func, ptr, buf_ptr);
+      x86_mov(cp->func, elt, buf_stride);
+      x86_add(cp->func, elt, ptr);
+      if (buf_idx == 0) sse_prefetchnta(cp->func, x86_make_disp(elt, 192));
+      x86_mov(cp->func, buf_ptr, elt);
+   }
+   else {
+      struct x86_reg buf_base_ptr = x86_make_disp(buf, 
+                                                  Offset(struct aos_buffer, base_ptr));
+
+
+      /* Calculate pointer to current attrib:
+       */
+      x86_mov(cp->func, ptr, buf_stride);
+      x86_imul(cp->func, ptr, elt);
+      x86_add(cp->func, ptr, buf_base_ptr);
+   }
+
+   cp->insn_counter++;
+
+   return TRUE;
+}
+
+
+static boolean load_input( struct aos_compilation *cp,
+                           unsigned idx,
+                           struct x86_reg bufptr )
+{
+   unsigned format = cp->vaos->base.key.element[idx].in.format;
+   unsigned offset = cp->vaos->base.key.element[idx].in.offset;
+   struct x86_reg dataXMM = aos_get_xmm_reg(cp);
+
+   /* Figure out source pointer address:
+    */
+   struct x86_reg src = x86_make_disp(bufptr, offset);
+
+   aos_adopt_xmm_reg( cp,
+                      dataXMM,
+                      TGSI_FILE_INPUT,
+                      idx,
+                      TRUE );
+
+   switch (format) {
+   case PIPE_FORMAT_R32_FLOAT:
+      emit_load_R32(cp, dataXMM, src);
+      break;
+   case PIPE_FORMAT_R32G32_FLOAT:
+      emit_load_R32G32(cp, dataXMM, src);
+      break;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      emit_load_R32G32B32(cp, dataXMM, src);
+      break;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      emit_load_R32G32B32A32(cp, dataXMM, src);
+      break;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      emit_load_R8G8B8A8_UNORM(cp, dataXMM, src);
+      emit_swizzle(cp, dataXMM, dataXMM, SHUF(Z,Y,X,W));
+      break;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      emit_load_R8G8B8A8_UNORM(cp, dataXMM, src);
+      break;
+   default:
+      ERROR(cp, "unhandled input format");
+      return FALSE;
+   }
+
+   return TRUE;
+}
+
+static boolean load_inputs( struct aos_compilation *cp,
+                            unsigned buffer,
+                            struct x86_reg ptr )
+{
+   unsigned i;
+
+   for (i = 0; i < cp->vaos->base.key.nr_inputs; i++) {
+      if (cp->vaos->base.key.element[i].in.buffer == buffer) {
+
+         if (!load_input( cp, i, ptr ))
+            return FALSE;
+
+         cp->insn_counter++;
+      }
+   }
+   
+   return TRUE;
+}
+
+boolean aos_init_inputs( struct aos_compilation *cp, boolean linear )
+{
+   unsigned i;
+   for (i = 0; i < cp->vaos->nr_vb; i++) {
+      struct x86_reg buf = x86_make_disp(aos_get_x86( cp, 0, X86_BUFFERS ), 
+                                         i * sizeof(struct aos_buffer));
+
+      struct x86_reg buf_base_ptr = x86_make_disp(buf, 
+                                                  Offset(struct aos_buffer, base_ptr));
+
+      if (cp->vaos->base.key.const_vbuffers & (1<<i)) {
+         struct x86_reg ptr = cp->tmp_EAX;
+
+         x86_mov(cp->func, ptr, buf_base_ptr);
+
+         /* Load all inputs for this constant vertex buffer
+          */
+         load_inputs( cp, i, x86_deref(ptr) );
+         
+         /* Then just force them out to aos_machine.input[]
+          */
+         aos_spill_all( cp );
+
+      }
+      else if (linear) {
+
+         struct x86_reg elt = cp->idx_EBX;
+         struct x86_reg ptr = cp->tmp_EAX;
+
+         struct x86_reg buf_stride = x86_make_disp(buf, 
+                                                   Offset(struct aos_buffer, stride));
+
+         struct x86_reg buf_ptr = x86_make_disp(buf, 
+                                                Offset(struct aos_buffer, ptr));
+
+
+         /* Calculate pointer to current attrib:
+          */
+         x86_mov(cp->func, ptr, buf_stride);
+         x86_imul(cp->func, ptr, elt);
+         x86_add(cp->func, ptr, buf_base_ptr);
+
+
+         /* In the linear case, keep the buffer pointer instead of the
+          * index number.
+          */
+         if (cp->vaos->nr_vb == 1) 
+            x86_mov( cp->func, elt, ptr );
+         else
+            x86_mov( cp->func, buf_ptr, ptr );
+
+         cp->insn_counter++;
+      }
+   }
+
+   return TRUE;
+}
+
+boolean aos_fetch_inputs( struct aos_compilation *cp, boolean linear )
+{
+   unsigned j;
+
+   for (j = 0; j < cp->vaos->nr_vb; j++) {
+      if (cp->vaos->base.key.const_vbuffers & (1<<j)) {
+         /* just retreive pre-transformed input */
+      }
+      else if (linear && cp->vaos->nr_vb == 1) {
+         load_inputs( cp, 0, cp->idx_EBX );
+      }
+      else {
+         struct x86_reg elt = linear ? cp->idx_EBX : x86_deref(cp->idx_EBX);
+         struct x86_reg ptr = cp->tmp_EAX;
+
+         if (!get_buffer_ptr( cp, linear, j, elt, ptr ))
+            return FALSE;
+
+         if (!load_inputs( cp, j, ptr ))
+            return FALSE;
+      }
+   }
+
+   return TRUE;
+}
+
+boolean aos_incr_inputs( struct aos_compilation *cp, boolean linear )
+{
+   if (linear && cp->vaos->nr_vb == 1) {
+      struct x86_reg stride = x86_make_disp(aos_get_x86( cp, 0, X86_BUFFERS ), 
+                                            (0 * sizeof(struct aos_buffer) + 
+                                             Offset(struct aos_buffer, stride)));
+
+      x86_add(cp->func, cp->idx_EBX, stride);
+      sse_prefetchnta(cp->func, x86_make_disp(cp->idx_EBX, 192));
+   }
+   else if (linear) {
+      /* Nothing to do */
+   } 
+   else {
+      x86_lea(cp->func, cp->idx_EBX, x86_make_disp(cp->idx_EBX, 4));
+   }
+
+   return TRUE;
+}
+
+
+
+
+
+
+static void emit_store_R32G32B32A32( struct aos_compilation *cp, 			   
+				     struct x86_reg dst_ptr,
+				     struct x86_reg dataXMM )
+{
+   sse_movups(cp->func, dst_ptr, dataXMM);
+}
+
+static void emit_store_R32G32B32( struct aos_compilation *cp, 
+				  struct x86_reg dst_ptr,
+				  struct x86_reg dataXMM )
+{
+   sse_movlps(cp->func, dst_ptr, dataXMM);
+   sse_shufps(cp->func, dataXMM, dataXMM, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */
+   sse_movss(cp->func, x86_make_disp(dst_ptr,8), dataXMM);
+}
+
+static void emit_store_R32G32( struct aos_compilation *cp, 
+			       struct x86_reg dst_ptr,
+			       struct x86_reg dataXMM )
+{
+   sse_movlps(cp->func, dst_ptr, dataXMM);
+}
+
+static void emit_store_R32( struct aos_compilation *cp, 
+			    struct x86_reg dst_ptr,
+			    struct x86_reg dataXMM )
+{
+   sse_movss(cp->func, dst_ptr, dataXMM);
+}
+
+
+
+static void emit_store_R8G8B8A8_UNORM( struct aos_compilation *cp,
+				       struct x86_reg dst_ptr,
+				       struct x86_reg dataXMM )
+{
+   sse_mulps(cp->func, dataXMM, aos_get_internal(cp, IMM_255));
+   sse2_cvtps2dq(cp->func, dataXMM, dataXMM);
+   sse2_packssdw(cp->func, dataXMM, dataXMM);
+   sse2_packuswb(cp->func, dataXMM, dataXMM);
+   sse_movss(cp->func, dst_ptr, dataXMM);
+}
+
+
+
+
+
+static boolean emit_output( struct aos_compilation *cp,
+                            struct x86_reg ptr,
+                            struct x86_reg dataXMM, 
+                            unsigned format )
+{
+   switch (format) {
+   case EMIT_1F:
+   case EMIT_1F_PSIZE:
+      emit_store_R32(cp, ptr, dataXMM);
+      break;
+   case EMIT_2F:
+      emit_store_R32G32(cp, ptr, dataXMM);
+      break;
+   case EMIT_3F:
+      emit_store_R32G32B32(cp, ptr, dataXMM);
+      break;
+   case EMIT_4F:
+      emit_store_R32G32B32A32(cp, ptr, dataXMM);
+      break;
+   case EMIT_4UB:
+      if (1) {
+         emit_swizzle(cp, dataXMM, dataXMM, SHUF(Z,Y,X,W));
+         emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM);
+      }
+      else {
+         emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM);
+      }
+      break;
+   default:
+      ERROR(cp, "unhandled output format");
+      return FALSE;
+   }
+
+   return TRUE;
+}
+
+
+
+boolean aos_emit_outputs( struct aos_compilation *cp )
+{
+   unsigned i;
+   
+   for (i = 0; i < cp->vaos->base.key.nr_outputs; i++) {
+      unsigned format = cp->vaos->base.key.element[i].out.format;
+      unsigned offset = cp->vaos->base.key.element[i].out.offset;
+      unsigned vs_output = cp->vaos->base.key.element[i].out.vs_output;
+
+      struct x86_reg data;
+
+      if (format == EMIT_1F_PSIZE) {
+         data = aos_get_internal_xmm( cp, IMM_PSIZE );
+      }
+      else {
+         data = aos_get_shader_reg( cp, 
+                                    TGSI_FILE_OUTPUT,
+                                    vs_output );
+      }
+
+      if (data.file != file_XMM) {
+         struct x86_reg tmp = aos_get_xmm_reg( cp );
+         sse_movaps(cp->func, tmp, data);
+         data = tmp;
+      }
+      
+      if (!emit_output( cp, 
+                        x86_make_disp( cp->outbuf_ECX, offset ),
+                        data, 
+                        format ))
+         return FALSE;
+
+      aos_release_xmm_reg( cp, data.idx );
+
+      cp->insn_counter++;
+   }
+
+   return TRUE;
+}
+
+#endif
diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c
new file mode 100644
index 0000000000..b358bd2df4
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c
@@ -0,0 +1,324 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "pipe/p_config.h"
+
+
+#include "pipe/p_shader_tokens.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi/tgsi_exec.h"
+#include "draw_vs.h"
+#include "draw_vs_aos.h"
+#include "draw_vertex.h"
+
+#ifdef PIPE_ARCH_X86
+
+#include "rtasm/rtasm_x86sse.h"
+
+
+#define X87_CW_EXCEPTION_INV_OP       (1<<0)
+#define X87_CW_EXCEPTION_DENORM_OP    (1<<1)
+#define X87_CW_EXCEPTION_ZERO_DIVIDE  (1<<2)
+#define X87_CW_EXCEPTION_OVERFLOW     (1<<3)
+#define X87_CW_EXCEPTION_UNDERFLOW    (1<<4)
+#define X87_CW_EXCEPTION_PRECISION    (1<<5)
+#define X87_CW_PRECISION_SINGLE       (0<<8)
+#define X87_CW_PRECISION_RESERVED     (1<<8)
+#define X87_CW_PRECISION_DOUBLE       (2<<8)
+#define X87_CW_PRECISION_DOUBLE_EXT   (3<<8)
+#define X87_CW_PRECISION_MASK         (3<<8)
+#define X87_CW_ROUND_NEAREST          (0<<10)
+#define X87_CW_ROUND_DOWN             (1<<10)
+#define X87_CW_ROUND_UP               (2<<10)
+#define X87_CW_ROUND_ZERO             (3<<10)
+#define X87_CW_ROUND_MASK             (3<<10)
+#define X87_CW_INFINITY               (1<<12)
+
+
+void PIPE_CDECL aos_do_lit( struct aos_machine *machine,
+                            float *result,
+                            const float *in,
+                            unsigned count )
+{
+   if (in[0] > 0) 
+   {
+      if (in[1] <= 0.0) 
+      {
+         result[0] = 1.0F;
+         result[1] = in[0];
+         result[2] = 1.0;
+         result[3] = 1.0F;
+      }
+      else
+      {
+         const float epsilon = 1.0F / 256.0F;    
+         float exponent = CLAMP(in[3], -(128.0F - epsilon), (128.0F - epsilon));
+         result[0] = 1.0F;
+         result[1] = in[0];
+         result[2] = powf(in[1], exponent);
+         result[3] = 1.0;
+      }
+   }
+   else 
+   {
+      result[0] = 1.0F;
+      result[1] = 0.0;
+      result[2] = 0.0;
+      result[3] = 1.0F;
+   }
+}
+
+
+static void PIPE_CDECL do_lit_lut( struct aos_machine *machine,
+                                   float *result,
+                                   const float *in,
+                                   unsigned count )
+{
+   if (in[0] > 0) 
+   {
+      if (in[1] <= 0.0) 
+      {
+         result[0] = 1.0F;
+         result[1] = in[0];
+         result[2] = 1.0;
+         result[3] = 1.0F;
+         return;
+      }
+      
+      if (machine->lit_info[count].shine_tab->exponent != in[3]) {
+         machine->lit_info[count].func = aos_do_lit;
+         goto no_luck;
+      }
+
+      if (in[1] <= 1.0)
+      {
+         const float *tab = machine->lit_info[count].shine_tab->values;
+         float f = in[1] * 256;
+         int k = (int)f;
+         float frac = f - (float)k;
+         
+         result[0] = 1.0F;
+         result[1] = in[0];
+         result[2] = tab[k] + frac*(tab[k+1]-tab[k]);
+         result[3] = 1.0;
+         return;
+      }
+      
+   no_luck:
+      {
+         const float epsilon = 1.0F / 256.0F;    
+         float exponent = CLAMP(in[3], -(128.0F - epsilon), (128.0F - epsilon));
+         result[0] = 1.0F;
+         result[1] = in[0];
+         result[2] = powf(in[1], exponent);
+         result[3] = 1.0;
+      }
+   }
+   else 
+   {
+      result[0] = 1.0F;
+      result[1] = 0.0;
+      result[2] = 0.0;
+      result[3] = 1.0F;
+   }
+}
+
+
+static void do_populate_lut( struct shine_tab *tab,
+                             float unclamped_exponent )
+{
+   const float epsilon = 1.0F / 256.0F;    
+   float exponent = CLAMP(unclamped_exponent, -(128.0F - epsilon), (128.0F - epsilon));
+   unsigned i;
+
+   tab->exponent = unclamped_exponent; /* for later comparison */
+   
+   tab->values[0] = 0;
+   if (exponent == 0) {
+      for (i = 1; i < 258; i++) {
+         tab->values[i] = 1.0;
+      }      
+   }
+   else {
+      for (i = 1; i < 258; i++) {
+         tab->values[i] = powf((float)i * epsilon, exponent);
+      }
+   }
+}
+
+
+
+
+static void PIPE_CDECL populate_lut( struct aos_machine *machine,
+                                     float *result,
+                                     const float *in,
+                                     unsigned count )
+{
+   unsigned i, tab;
+
+   /* Search for an existing table for this value.  Note that without
+    * static analysis we don't really know if in[3] will be constant,
+    * but it usually is...
+    */
+   for (tab = 0; tab < 4; tab++) {
+      if (machine->shine_tab[tab].exponent == in[3]) {
+         goto found;
+      }
+   }
+
+   for (tab = 0, i = 1; i < 4; i++) {
+      if (machine->shine_tab[i].last_used < machine->shine_tab[tab].last_used)
+         tab = i;
+   }
+
+   if (machine->shine_tab[tab].last_used == machine->now) {
+      /* No unused tables (this is not a ffvertex program...).  Just
+       * call pow each time:
+       */
+      machine->lit_info[count].func = aos_do_lit;
+      machine->lit_info[count].func( machine, result, in, count );
+      return;
+   }
+   else {
+      do_populate_lut( &machine->shine_tab[tab], in[3] );
+   }
+
+ found:
+   machine->shine_tab[tab].last_used = machine->now;
+   machine->lit_info[count].shine_tab = &machine->shine_tab[tab];
+   machine->lit_info[count].func = do_lit_lut;
+   machine->lit_info[count].func( machine, result, in, count );
+}
+
+
+void draw_vs_aos_machine_constants( struct aos_machine *machine,
+                                    const float (*constants)[4] )
+{
+   machine->constants = constants;
+
+   {
+      unsigned i;
+      for (i = 0; i < MAX_LIT_INFO; i++) {
+         machine->lit_info[i].func = populate_lut;
+         machine->now++;
+      }
+   }
+}
+
+
+void draw_vs_aos_machine_viewport( struct aos_machine *machine,
+                                   const struct pipe_viewport_state *viewport )
+{
+   memcpy(machine->scale, viewport->scale, 4 * sizeof(float));
+   memcpy(machine->translate, viewport->translate, 4 * sizeof(float));
+}
+
+
+
+void draw_vs_aos_machine_destroy( struct aos_machine *machine )
+{
+   align_free(machine);
+}
+
+struct aos_machine *draw_vs_aos_machine( void )
+{
+   struct aos_machine *machine;
+   unsigned i;
+   float inv = 1.0f/255.0f;
+   float f255 = 255.0f;
+
+   machine = align_malloc(sizeof(struct aos_machine), 16);
+   if (!machine)
+      return NULL;
+
+   memset(machine, 0, sizeof(*machine));
+
+   ASSIGN_4V(machine->internal[IMM_SWZ],       1.0f,  -1.0f,  0.0f, 1.0f);
+   *(unsigned *)&machine->internal[IMM_SWZ][3] = 0xffffffff;
+
+   ASSIGN_4V(machine->internal[IMM_ONES],      1.0f,  1.0f,  1.0f,  1.0f);
+   ASSIGN_4V(machine->internal[IMM_NEGS],     -1.0f, -1.0f, -1.0f, -1.0f);
+   ASSIGN_4V(machine->internal[IMM_IDENTITY],  0.0f,  0.0f,  0.0f,  1.0f);
+   ASSIGN_4V(machine->internal[IMM_INV_255],   inv,   inv,   inv,   inv);
+   ASSIGN_4V(machine->internal[IMM_255],       f255,  f255,  f255,  f255);
+   ASSIGN_4V(machine->internal[IMM_RSQ],       -.5f,  1.5f,  0.0f,  0.0f);
+
+
+   machine->fpu_rnd_nearest = (X87_CW_EXCEPTION_INV_OP |
+                               X87_CW_EXCEPTION_DENORM_OP |
+                               X87_CW_EXCEPTION_ZERO_DIVIDE |
+                               X87_CW_EXCEPTION_OVERFLOW |
+                               X87_CW_EXCEPTION_UNDERFLOW |
+                               X87_CW_EXCEPTION_PRECISION |
+                               (1<<6) |
+                               X87_CW_ROUND_NEAREST |
+                               X87_CW_PRECISION_DOUBLE_EXT);
+
+   assert(machine->fpu_rnd_nearest == 0x37f);
+                               
+   machine->fpu_rnd_neg_inf = (X87_CW_EXCEPTION_INV_OP |
+                               X87_CW_EXCEPTION_DENORM_OP |
+                               X87_CW_EXCEPTION_ZERO_DIVIDE |
+                               X87_CW_EXCEPTION_OVERFLOW |
+                               X87_CW_EXCEPTION_UNDERFLOW |
+                               X87_CW_EXCEPTION_PRECISION |
+                               (1<<6) |
+                               X87_CW_ROUND_DOWN |
+                               X87_CW_PRECISION_DOUBLE_EXT);
+
+   for (i = 0; i < MAX_SHINE_TAB; i++)
+      do_populate_lut( &machine->shine_tab[i], 1.0f );
+
+   return machine;
+}
+
+#else
+
+void draw_vs_aos_machine_viewport( struct aos_machine *machine,
+                                   const struct pipe_viewport_state *viewport )
+{
+}
+
+void draw_vs_aos_machine_constants( struct aos_machine *machine,
+                                    const float (*constants)[4] )
+{
+}
+
+void draw_vs_aos_machine_destroy( struct aos_machine *machine )
+{
+}
+
+struct aos_machine *draw_vs_aos_machine( void )
+{
+   return NULL;
+}
+#endif
+
diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c
new file mode 100644
index 0000000000..b3200df811
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_exec.c
@@ -0,0 +1,200 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "draw_private.h"
+#include "draw_context.h"
+#include "draw_vs.h"
+
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_scan.h"
+
+
+struct exec_vertex_shader {
+   struct draw_vertex_shader base;
+   struct tgsi_exec_machine *machine;
+};
+
+static struct exec_vertex_shader *exec_vertex_shader( struct draw_vertex_shader *vs )
+{
+   return (struct exec_vertex_shader *)vs;
+}
+
+
+/* Not required for run_linear.
+ */
+static void
+vs_exec_prepare( struct draw_vertex_shader *shader,
+		 struct draw_context *draw )
+{
+   struct exec_vertex_shader *evs = exec_vertex_shader(shader);
+
+   /* Specify the vertex program to interpret/execute.
+    * Avoid rebinding when possible.
+    */
+   if (evs->machine->Tokens != shader->state.tokens) {
+      tgsi_exec_machine_bind_shader(evs->machine,
+                                    shader->state.tokens,
+                                    draw->vs.num_samplers,
+                                    draw->vs.samplers);
+   }
+}
+
+
+
+
+/* Simplified vertex shader interface for the pt paths.  Given the
+ * complexity of code-generating all the above operations together,
+ * it's time to try doing all the other stuff separately.
+ */
+static void
+vs_exec_run_linear( struct draw_vertex_shader *shader,
+		    const float (*input)[4],
+		    float (*output)[4],
+		    const float (*constants)[4],
+		    unsigned count,
+		    unsigned input_stride,
+		    unsigned output_stride )
+{
+   struct exec_vertex_shader *evs = exec_vertex_shader(shader);
+   struct tgsi_exec_machine *machine = evs->machine;
+   unsigned int i, j;
+   unsigned slot;
+
+   machine->Consts = constants;
+
+   for (i = 0; i < count; i += MAX_TGSI_VERTICES) {
+      unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i);
+
+      /* Swizzle inputs.  
+       */
+      for (j = 0; j < max_vertices; j++) {
+#if 0
+         debug_printf("%d) Input vert:\n", i + j);
+         for (slot = 0; slot < shader->info.num_inputs; slot++) {
+            debug_printf("\t%d: %f %f %f %f\n", slot,
+			 input[slot][0],
+			 input[slot][1],
+			 input[slot][2],
+			 input[slot][3]);
+         }
+#endif
+
+         for (slot = 0; slot < shader->info.num_inputs; slot++) {
+            machine->Inputs[slot].xyzw[0].f[j] = input[slot][0];
+            machine->Inputs[slot].xyzw[1].f[j] = input[slot][1];
+            machine->Inputs[slot].xyzw[2].f[j] = input[slot][2];
+            machine->Inputs[slot].xyzw[3].f[j] = input[slot][3];
+         }
+
+	 input = (const float (*)[4])((const char *)input + input_stride);
+      } 
+
+      tgsi_set_exec_mask(machine,
+                         1,
+                         max_vertices > 1,
+                         max_vertices > 2,
+                         max_vertices > 3);
+
+      /* run interpreter */
+      tgsi_exec_machine_run( machine );
+
+      /* Unswizzle all output results.  
+       */
+      for (j = 0; j < max_vertices; j++) {
+         for (slot = 0; slot < shader->info.num_outputs; slot++) {
+            output[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
+            output[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
+            output[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
+            output[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
+
+         }
+
+#if 0
+	 debug_printf("%d) Post xform vert:\n", i + j);
+	 for (slot = 0; slot < shader->info.num_outputs; slot++) {
+	    debug_printf("\t%d: %f %f %f %f\n", slot,
+			 output[slot][0],
+			 output[slot][1],
+			 output[slot][2],
+			 output[slot][3]);
+         }
+#endif
+
+	 output = (float (*)[4])((char *)output + output_stride);
+      } 
+
+   }
+}
+
+
+
+
+static void
+vs_exec_delete( struct draw_vertex_shader *dvs )
+{
+   FREE((void*) dvs->state.tokens);
+   FREE( dvs );
+}
+
+
+struct draw_vertex_shader *
+draw_create_vs_exec(struct draw_context *draw,
+		    const struct pipe_shader_state *state)
+{
+   struct exec_vertex_shader *vs = CALLOC_STRUCT( exec_vertex_shader );
+
+   if (vs == NULL) 
+      return NULL;
+
+   /* we make a private copy of the tokens */
+   vs->base.state.tokens = tgsi_dup_tokens(state->tokens);
+   if (!vs->base.state.tokens) {
+      FREE(vs);
+      return NULL;
+   }
+
+   tgsi_scan_shader(state->tokens, &vs->base.info);
+
+   vs->base.draw = draw;
+   vs->base.prepare = vs_exec_prepare;
+   vs->base.run_linear = vs_exec_run_linear;
+   vs->base.delete = vs_exec_delete;
+   vs->base.create_varient = draw_vs_varient_generic;
+   vs->machine = &draw->vs.machine;
+
+   return &vs->base;
+}
diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c
new file mode 100644
index 0000000000..727977bc3a
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c
@@ -0,0 +1,161 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw_private.h"
+#include "draw_context.h"
+#include "draw_vs.h"
+
+#include "tgsi/tgsi_parse.h"
+
+#ifdef MESA_LLVM
+
+#include "gallivm/gallivm.h"
+
+struct draw_llvm_vertex_shader {
+   struct draw_vertex_shader base;
+   struct gallivm_prog *llvm_prog;
+   struct tgsi_exec_machine *machine;
+};
+
+
+static void
+vs_llvm_prepare( struct draw_vertex_shader *base,
+		 struct draw_context *draw )
+{
+}
+
+
+
+
+static void
+vs_llvm_run_linear( struct draw_vertex_shader *base,
+		   const float (*input)[4],
+		   float (*output)[4],
+		   const float (*constants)[4],
+		   unsigned count,
+		   unsigned input_stride,
+		   unsigned output_stride )
+{
+   struct draw_llvm_vertex_shader *shader =
+      (struct draw_llvm_vertex_shader *)base;
+
+   gallivm_cpu_vs_exec(shader->llvm_prog, shader->machine,
+                       input, base->info.num_inputs, output, base->info.num_outputs,
+                       constants, count, input_stride, output_stride);
+}
+
+
+
+static void
+vs_llvm_delete( struct draw_vertex_shader *base )
+{
+   struct draw_llvm_vertex_shader *shader =
+      (struct draw_llvm_vertex_shader *)base;
+
+   /* Do something to free compiled shader:
+    */
+
+   FREE( (void*) shader->base.state.tokens );
+   FREE( shader );
+}
+
+
+
+
+struct draw_vertex_shader *
+draw_create_vs_llvm(struct draw_context *draw,
+		    const struct pipe_shader_state *templ)
+{
+   struct draw_llvm_vertex_shader *vs;
+
+   vs = CALLOC_STRUCT( draw_llvm_vertex_shader );
+   if (vs == NULL)
+      return NULL;
+
+   /* we make a private copy of the tokens */
+   vs->base.state.tokens = tgsi_dup_tokens(templ->tokens);
+   if (!vs->base.state.tokens) {
+      FREE(vs);
+      return NULL;
+   }
+
+   tgsi_scan_shader(vs->base.state.tokens, &vs->base.info);
+
+   vs->base.draw = draw;
+   vs->base.prepare = vs_llvm_prepare;
+   vs->base.create_varient = draw_vs_varient_generic;
+   vs->base.run_linear = vs_llvm_run_linear;
+   vs->base.delete = vs_llvm_delete;
+   vs->machine = &draw->vs.machine;
+
+   {
+      struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS);
+      gallivm_ir_set_layout(ir, GALLIVM_SOA);
+      gallivm_ir_set_components(ir, 4);
+      gallivm_ir_fill_from_tgsi(ir, vs->base.state.tokens);
+      vs->llvm_prog = gallivm_ir_compile(ir);
+      gallivm_ir_delete(ir);
+   }
+
+   draw->vs.engine = gallivm_global_cpu_engine();
+
+   /* XXX: Why are there two versions of this?  Shouldn't creating the
+    *      engine be a separate operation to compiling a shader?
+    */
+   if (!draw->vs.engine) {
+      draw->vs.engine = gallivm_cpu_engine_create(vs->llvm_prog);
+   }
+   else {
+      gallivm_cpu_jit_compile(draw->vs.engine, vs->llvm_prog);
+   }
+
+   return &vs->base;
+}
+
+
+
+
+
+#else
+
+struct draw_vertex_shader *
+draw_create_vs_llvm(struct draw_context *draw,
+                          const struct pipe_shader_state *shader)
+{
+   return NULL;
+}
+
+#endif
diff --git a/src/gallium/auxiliary/draw/draw_vs_ppc.c b/src/gallium/auxiliary/draw/draw_vs_ppc.c
new file mode 100644
index 0000000000..d35db57d57
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_ppc.c
@@ -0,0 +1,244 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_config.h"
+
+#include "draw_vs.h"
+
+#if defined(PIPE_ARCH_PPC)
+
+#include "pipe/p_shader_tokens.h"
+
+#include "draw_private.h"
+#include "draw_context.h"
+
+#include "rtasm/rtasm_cpu.h"
+#include "rtasm/rtasm_ppc.h"
+#include "tgsi/tgsi_ppc.h"
+#include "tgsi/tgsi_parse.h"
+
+
+
+typedef void (PIPE_CDECL *codegen_function) (float (*inputs)[4][4],
+                                             float (*outputs)[4][4],
+                                             float (*temps)[4][4],
+                                             float (*immeds)[4],
+                                             float (*consts)[4],
+                                             const float *builtins);
+
+
+struct draw_ppc_vertex_shader {
+   struct draw_vertex_shader base;
+   struct ppc_function ppc_program;
+
+   codegen_function func;
+};
+
+
+static void
+vs_ppc_prepare( struct draw_vertex_shader *base,
+		struct draw_context *draw )
+{
+   /* nothing */
+}
+
+
+/**
+ * Simplified vertex shader interface for the pt paths.  Given the
+ * complexity of code-generating all the above operations together,
+ * it's time to try doing all the other stuff separately.
+ */
+static void
+vs_ppc_run_linear( struct draw_vertex_shader *base,
+		   const float (*input)[4],
+		   float (*output)[4],
+		   const float (*constants)[4],
+		   unsigned count,
+		   unsigned input_stride,
+		   unsigned output_stride )
+{
+   struct draw_ppc_vertex_shader *shader = (struct draw_ppc_vertex_shader *)base;
+   unsigned int i;
+
+#define MAX_VERTICES 4
+
+   /* loop over verts */
+   for (i = 0; i < count; i += MAX_VERTICES) {
+      const uint max_vertices = MIN2(MAX_VERTICES, count - i);
+      float inputs_soa[PIPE_MAX_SHADER_INPUTS][4][4] ALIGN16_ATTRIB;
+      float outputs_soa[PIPE_MAX_SHADER_OUTPUTS][4][4] ALIGN16_ATTRIB;
+      float temps_soa[TGSI_EXEC_NUM_TEMPS][4][4] ALIGN16_ATTRIB;
+      uint attr;
+
+      /* convert (up to) four input verts to SoA format */
+      for (attr = 0; attr < base->info.num_inputs; attr++) {
+         const float *vIn = (const float *) input;
+         uint vert;
+         for (vert = 0; vert < max_vertices; vert++) {
+#if 0
+            if (attr==0)
+               printf("Input v%d a%d: %f %f %f %f\n",
+                      vert, attr, vIn[0], vIn[1], vIn[2], vIn[3]);
+#endif
+            inputs_soa[attr][0][vert] = vIn[attr * 4 + 0];
+            inputs_soa[attr][1][vert] = vIn[attr * 4 + 1];
+            inputs_soa[attr][2][vert] = vIn[attr * 4 + 2];
+            inputs_soa[attr][3][vert] = vIn[attr * 4 + 3];
+            vIn += input_stride / 4;
+         }
+      }
+
+      /* run compiled shader
+       */
+      shader->func(inputs_soa, outputs_soa, temps_soa,
+		   (float (*)[4]) shader->base.immediates,
+		   (float (*)[4]) constants,
+                   ppc_builtin_constants);
+
+      /* convert (up to) four output verts from SoA back to AoS format */
+      for (attr = 0; attr < base->info.num_outputs; attr++) {
+         float *vOut = (float *) output;
+         uint vert;
+         for (vert = 0; vert < max_vertices; vert++) {
+            vOut[attr * 4 + 0] = outputs_soa[attr][0][vert];
+            vOut[attr * 4 + 1] = outputs_soa[attr][1][vert];
+            vOut[attr * 4 + 2] = outputs_soa[attr][2][vert];
+            vOut[attr * 4 + 3] = outputs_soa[attr][3][vert];
+#if 0
+            if (attr==0)
+               printf("Output v%d a%d: %f %f %f %f\n",
+                      vert, attr, vOut[0], vOut[1], vOut[2], vOut[3]);
+#endif
+            vOut += output_stride / 4;
+         }
+      }
+
+      /* advance to next group of four input/output verts */
+      input = (const float (*)[4])((const char *)input + input_stride * max_vertices);
+      output = (float (*)[4])((char *)output + output_stride * max_vertices);
+   }
+}
+
+
+static void
+vs_ppc_delete( struct draw_vertex_shader *base )
+{
+   struct draw_ppc_vertex_shader *shader = (struct draw_ppc_vertex_shader *)base;
+   
+   ppc_release_func( &shader->ppc_program );
+
+   align_free( (void *) shader->base.immediates );
+
+   FREE( (void*) shader->base.state.tokens );
+   FREE( shader );
+}
+
+
+struct draw_vertex_shader *
+draw_create_vs_ppc(struct draw_context *draw,
+                   const struct pipe_shader_state *templ)
+{
+   struct draw_ppc_vertex_shader *vs;
+
+   vs = CALLOC_STRUCT( draw_ppc_vertex_shader );
+   if (vs == NULL) 
+      return NULL;
+
+   /* we make a private copy of the tokens */
+   vs->base.state.tokens = tgsi_dup_tokens(templ->tokens);
+   if (!vs->base.state.tokens)
+      goto fail;
+
+   tgsi_scan_shader(templ->tokens, &vs->base.info);
+
+   vs->base.draw = draw;
+#if 0
+   if (1)
+      vs->base.create_varient = draw_vs_varient_aos_ppc;
+   else
+#endif
+      vs->base.create_varient = draw_vs_varient_generic;
+   vs->base.prepare = vs_ppc_prepare;
+   vs->base.run_linear = vs_ppc_run_linear;
+   vs->base.delete = vs_ppc_delete;
+   
+   vs->base.immediates = align_malloc(TGSI_EXEC_NUM_IMMEDIATES * 4 *
+                                      sizeof(float), 16);
+
+   ppc_init_func( &vs->ppc_program );
+
+#if 0
+   ppc_print_code(&vs->ppc_program, TRUE);
+   ppc_indent(&vs->ppc_program, 8);
+#endif
+
+   if (!tgsi_emit_ppc( (struct tgsi_token *) vs->base.state.tokens,
+			&vs->ppc_program, 
+                       (float (*)[4]) vs->base.immediates, 
+                        TRUE )) 
+      goto fail;
+      
+   vs->func = (codegen_function) ppc_get_func( &vs->ppc_program );
+   if (!vs->func) {
+      goto fail;
+   }
+   
+   return &vs->base;
+
+fail:
+   /*
+   debug_error("tgsi_emit_ppc() failed, falling back to interpreter\n");
+   */
+
+   ppc_release_func( &vs->ppc_program );
+   
+   FREE(vs);
+   return NULL;
+}
+
+
+
+#else /* PIPE_ARCH_PPC */
+
+
+struct draw_vertex_shader *
+draw_create_vs_ppc( struct draw_context *draw,
+		    const struct pipe_shader_state *templ )
+{
+   return (void *) 0;
+}
+
+
+#endif /* PIPE_ARCH_PPC */
diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c
new file mode 100644
index 0000000000..77ba5152f9
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_sse.c
@@ -0,0 +1,226 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_config.h"
+
+#include "draw_vs.h"
+
+#if defined(PIPE_ARCH_X86)
+
+#include "pipe/p_shader_tokens.h"
+
+#include "draw_private.h"
+#include "draw_context.h"
+
+#include "rtasm/rtasm_cpu.h"
+#include "rtasm/rtasm_x86sse.h"
+#include "tgsi/tgsi_sse2.h"
+#include "tgsi/tgsi_parse.h"
+
+#define SSE_MAX_VERTICES 4
+
+typedef void (PIPE_CDECL *codegen_function) (
+   const struct tgsi_exec_vector *input, /* 1 */
+   struct tgsi_exec_vector *output, /* 2 */
+   float (*constant)[4],        /* 3 */
+   struct tgsi_exec_vector *temporary, /* 4 */
+   float (*immediates)[4],      /* 5 */
+   const float (*aos_input)[4], /* 6 */
+   uint num_inputs,             /* 7 */
+   uint input_stride,           /* 8 */
+   float (*aos_output)[4],      /* 9 */
+   uint num_outputs,            /* 10 */
+   uint output_stride );        /* 11 */
+
+struct draw_sse_vertex_shader {
+   struct draw_vertex_shader base;
+   struct x86_function sse2_program;
+
+   codegen_function func;
+   
+   struct tgsi_exec_machine *machine;
+};
+
+
+static void
+vs_sse_prepare( struct draw_vertex_shader *base,
+		struct draw_context *draw )
+{
+}
+
+
+
+/* Simplified vertex shader interface for the pt paths.  Given the
+ * complexity of code-generating all the above operations together,
+ * it's time to try doing all the other stuff separately.
+ */
+static void
+vs_sse_run_linear( struct draw_vertex_shader *base,
+		   const float (*input)[4],
+		   float (*output)[4],
+		   const float (*constants)[4],
+		   unsigned count,
+		   unsigned input_stride,
+		   unsigned output_stride )
+{
+   struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base;
+   struct tgsi_exec_machine *machine = shader->machine;
+   unsigned int i;
+
+   /* By default, execute all channels.  XXX move this inside the loop
+    * below when we support shader conditionals/loops.
+    */
+   tgsi_set_exec_mask(machine, 1, 1, 1, 1);
+
+   for (i = 0; i < count; i += MAX_TGSI_VERTICES) {
+      unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i);
+
+      if (max_vertices < 4) {
+         /* disable the unused execution channels */
+         tgsi_set_exec_mask(machine,
+                            1,
+                            max_vertices > 1,
+                            max_vertices > 2,
+                            0);
+      }
+
+      /* run compiled shader
+       */
+      shader->func(machine->Inputs,
+		   machine->Outputs,
+		   (float (*)[4])constants,
+		   machine->Temps,
+		   (float (*)[4])shader->base.immediates,
+                   input,
+                   base->info.num_inputs,
+                   input_stride,
+                   output,
+                   base->info.num_outputs,
+                   output_stride );
+
+      input = (const float (*)[4])((const char *)input + input_stride * max_vertices);
+      output = (float (*)[4])((char *)output + output_stride * max_vertices);
+   }
+}
+
+
+
+
+static void
+vs_sse_delete( struct draw_vertex_shader *base )
+{
+   struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base;
+   
+   x86_release_func( &shader->sse2_program );
+
+   align_free( (void *) shader->base.immediates );
+
+   FREE( (void*) shader->base.state.tokens );
+   FREE( shader );
+}
+
+
+struct draw_vertex_shader *
+draw_create_vs_sse(struct draw_context *draw,
+                          const struct pipe_shader_state *templ)
+{
+   struct draw_sse_vertex_shader *vs;
+
+   if (!rtasm_cpu_has_sse2())
+      return NULL;
+
+   vs = CALLOC_STRUCT( draw_sse_vertex_shader );
+   if (vs == NULL) 
+      return NULL;
+
+   /* we make a private copy of the tokens */
+   vs->base.state.tokens = tgsi_dup_tokens(templ->tokens);
+   if (!vs->base.state.tokens)
+      goto fail;
+
+   tgsi_scan_shader(templ->tokens, &vs->base.info);
+
+   vs->base.draw = draw;
+   if (1)
+      vs->base.create_varient = draw_vs_varient_aos_sse;
+   else
+      vs->base.create_varient = draw_vs_varient_generic;
+   vs->base.prepare = vs_sse_prepare;
+   vs->base.run_linear = vs_sse_run_linear;
+   vs->base.delete = vs_sse_delete;
+   
+   vs->base.immediates = align_malloc(TGSI_EXEC_NUM_IMMEDIATES * 4 *
+                                      sizeof(float), 16);
+
+   vs->machine = &draw->vs.machine;
+   
+   x86_init_func( &vs->sse2_program );
+
+   if (!tgsi_emit_sse2( (struct tgsi_token *) vs->base.state.tokens,
+			&vs->sse2_program, 
+                        (float (*)[4])vs->base.immediates, 
+                        TRUE )) 
+      goto fail;
+      
+   vs->func = (codegen_function) x86_get_func( &vs->sse2_program );
+   if (!vs->func) {
+      goto fail;
+   }
+   
+   return &vs->base;
+
+fail:
+   debug_error("tgsi_emit_sse2() failed, falling back to interpreter\n");
+
+   x86_release_func( &vs->sse2_program );
+   
+   FREE(vs);
+   return NULL;
+}
+
+
+
+#else
+
+struct draw_vertex_shader *
+draw_create_vs_sse( struct draw_context *draw,
+		    const struct pipe_shader_state *templ )
+{
+   return (void *) 0;
+}
+
+
+#endif
+
diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c
new file mode 100644
index 0000000000..7ee567d478
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_vs_varient.c
@@ -0,0 +1,322 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vbuf.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_vs.h"
+#include "translate/translate.h"
+#include "translate/translate_cache.h"
+
+/* A first pass at incorporating vertex fetch/emit functionality into 
+ */
+struct draw_vs_varient_generic {
+   struct draw_vs_varient base;
+
+   struct draw_vertex_shader *shader;
+   struct draw_context *draw;
+   
+   /* Basic plan is to run these two translate functions before/after
+    * the vertex shader's existing run_linear() routine to simulate
+    * the inclusion of this functionality into the shader...  
+    * 
+    * Next will look at actually including it.
+    */
+   struct translate *fetch;
+   struct translate *emit;
+
+   unsigned temp_vertex_stride;
+};
+
+
+
+
+
+static void vsvg_set_buffer( struct draw_vs_varient *varient,
+                             unsigned buffer,
+                             const void *ptr,
+                             unsigned stride )
+{
+   struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient;
+
+   vsvg->fetch->set_buffer(vsvg->fetch, 
+                           buffer, 
+                           ptr, 
+                           stride);
+}
+
+
+/* Mainly for debug at this stage:
+ */
+static void do_rhw_viewport( struct draw_vs_varient_generic *vsvg,
+                             unsigned count,
+                             void *output_buffer )
+{
+   char *ptr = (char *)output_buffer;
+   const float *scale = vsvg->base.vs->draw->viewport.scale;
+   const float *trans = vsvg->base.vs->draw->viewport.translate;
+   unsigned stride = vsvg->temp_vertex_stride;
+   unsigned j;
+
+   ptr += vsvg->base.vs->position_output * 4 * sizeof(float);
+
+   for (j = 0; j < count; j++, ptr += stride) {
+      float *data = (float *)ptr;
+      float w = 1.0f / data[3];
+
+      data[0] = data[0] * w * scale[0] + trans[0];
+      data[1] = data[1] * w * scale[1] + trans[1];
+      data[2] = data[2] * w * scale[2] + trans[2];
+      data[3] = w;
+   }
+}
+
+static void do_viewport( struct draw_vs_varient_generic *vsvg,
+                         unsigned count,
+                         void *output_buffer )
+{
+   char *ptr = (char *)output_buffer;
+   const float *scale = vsvg->base.vs->draw->viewport.scale;
+   const float *trans = vsvg->base.vs->draw->viewport.translate;
+   unsigned stride = vsvg->temp_vertex_stride;
+   unsigned j;
+
+   ptr += vsvg->base.vs->position_output * 4 * sizeof(float);
+
+   for (j = 0; j < count; j++, ptr += stride) {
+      float *data = (float *)ptr;
+
+      data[0] = data[0] * scale[0] + trans[0];
+      data[1] = data[1] * scale[1] + trans[1];
+      data[2] = data[2] * scale[2] + trans[2];
+   }
+}
+                         
+
+static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient,
+                                      const unsigned *elts,
+                                      unsigned count,
+                                      void *output_buffer)
+{
+   struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient;
+   unsigned temp_vertex_stride = vsvg->temp_vertex_stride;
+   void *temp_buffer = MALLOC( align(count,4) * temp_vertex_stride );
+   
+   if (0) debug_printf("%s %d \n", __FUNCTION__,  count);
+			
+   /* Want to do this in small batches for cache locality?
+    */
+   
+   vsvg->fetch->run_elts( vsvg->fetch, 
+                          elts,
+                          count,
+                          temp_buffer );
+
+   vsvg->base.vs->run_linear( vsvg->base.vs, 
+                              temp_buffer,
+                              temp_buffer,
+                              (const float (*)[4])vsvg->base.vs->draw->pt.user.constants,
+                              count,
+                              temp_vertex_stride, 
+                              temp_vertex_stride);
+
+
+   if (vsvg->base.key.clip) {
+      /* not really handling clipping, just do the rhw so we can
+       * see the results...
+       */
+      do_rhw_viewport( vsvg,
+                       count,
+                       temp_buffer );
+   }
+   else if (vsvg->base.key.viewport) {
+      do_viewport( vsvg,
+                   count,
+                   temp_buffer );
+   }
+
+
+   vsvg->emit->set_buffer( vsvg->emit,
+                           0, 
+                           temp_buffer,
+                           temp_vertex_stride );
+
+   vsvg->emit->set_buffer( vsvg->emit, 
+                           1,
+                           &vsvg->draw->rasterizer->point_size,
+                           0);
+
+   vsvg->emit->run( vsvg->emit,
+                    0, count,
+                    output_buffer );
+
+   FREE(temp_buffer);
+}
+
+
+static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient,
+                                        unsigned start,
+                                        unsigned count,
+                                        void *output_buffer )
+{
+   struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient;
+   unsigned temp_vertex_stride = vsvg->temp_vertex_stride;
+   void *temp_buffer = MALLOC( align(count,4) * temp_vertex_stride );
+	
+   if (0) debug_printf("%s %d %d (sz %d, %d)\n", __FUNCTION__, start, count,
+                       vsvg->base.key.output_stride,
+                       temp_vertex_stride);
+
+   vsvg->fetch->run( vsvg->fetch, 
+                     start,
+                     count,
+                     temp_buffer );
+
+   vsvg->base.vs->run_linear( vsvg->base.vs, 
+                              temp_buffer,
+                              temp_buffer,
+                              (const float (*)[4])vsvg->base.vs->draw->pt.user.constants,
+                              count,
+                              temp_vertex_stride, 
+                              temp_vertex_stride);
+
+   if (vsvg->base.key.clip) {
+      /* not really handling clipping, just do the rhw so we can
+       * see the results...
+       */
+      do_rhw_viewport( vsvg,
+                       count,
+                       temp_buffer );
+   }
+   else if (vsvg->base.key.viewport) {
+      do_viewport( vsvg,
+                   count,
+                   temp_buffer );
+   }
+
+   vsvg->emit->set_buffer( vsvg->emit,
+                           0, 
+                           temp_buffer,
+                           temp_vertex_stride );
+   
+   vsvg->emit->set_buffer( vsvg->emit, 
+                           1,
+                           &vsvg->draw->rasterizer->point_size,
+                           0);
+   
+   vsvg->emit->run( vsvg->emit,
+                    0, count,
+                    output_buffer );
+
+   FREE(temp_buffer);
+}
+
+
+
+
+
+static void vsvg_destroy( struct draw_vs_varient *varient )
+{
+   FREE(varient);
+}
+
+
+struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs,
+                                                 const struct draw_vs_varient_key *key )
+{
+   unsigned i;
+   struct translate_key fetch, emit;
+
+   struct draw_vs_varient_generic *vsvg = CALLOC_STRUCT( draw_vs_varient_generic );
+   if (vsvg == NULL)
+      return NULL;
+
+   vsvg->base.key = *key;
+   vsvg->base.vs = vs;
+   vsvg->base.set_buffer    = vsvg_set_buffer;
+   vsvg->base.run_elts      = vsvg_run_elts;
+   vsvg->base.run_linear    = vsvg_run_linear;
+   vsvg->base.destroy       = vsvg_destroy;
+
+   vsvg->draw = vs->draw;
+
+   vsvg->temp_vertex_stride = MAX2(key->nr_inputs,
+                                   vsvg->base.vs->info.num_outputs) * 4 * sizeof(float);
+
+   /* Build free-standing fetch and emit functions:
+    */
+   fetch.nr_elements = key->nr_inputs;
+   fetch.output_stride = vsvg->temp_vertex_stride;
+   for (i = 0; i < key->nr_inputs; i++) {
+      fetch.element[i].input_format = key->element[i].in.format;
+      fetch.element[i].input_buffer = key->element[i].in.buffer;
+      fetch.element[i].input_offset = key->element[i].in.offset;
+      fetch.element[i].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+      fetch.element[i].output_offset = i * 4 * sizeof(float);
+      assert(fetch.element[i].output_offset < fetch.output_stride);
+   }
+
+
+   emit.nr_elements = key->nr_outputs;
+   emit.output_stride = key->output_stride;
+   for (i = 0; i < key->nr_outputs; i++) {
+      if (key->element[i].out.format != EMIT_1F_PSIZE)
+      {      
+         emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+         emit.element[i].input_buffer = 0;
+         emit.element[i].input_offset = key->element[i].out.vs_output * 4 * sizeof(float);
+         emit.element[i].output_format = draw_translate_vinfo_format(key->element[i].out.format);
+         emit.element[i].output_offset = key->element[i].out.offset;
+         assert(emit.element[i].input_offset <= fetch.output_stride);
+      }
+      else {
+         emit.element[i].input_format = PIPE_FORMAT_R32_FLOAT;
+         emit.element[i].input_buffer = 1;
+         emit.element[i].input_offset = 0;
+         emit.element[i].output_format = PIPE_FORMAT_R32_FLOAT;
+         emit.element[i].output_offset = key->element[i].out.offset;
+      }
+   }
+
+   vsvg->fetch = draw_vs_get_fetch( vs->draw, &fetch );
+   vsvg->emit = draw_vs_get_emit( vs->draw, &emit );
+
+   return &vsvg->base;
+}
+
+
+
+
+
diff --git a/src/gallium/auxiliary/gallivm/Makefile b/src/gallium/auxiliary/gallivm/Makefile
new file mode 100644
index 0000000000..5a96d94ec3
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/Makefile
@@ -0,0 +1,92 @@
+# -*-makefile-*-
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = gallivm
+
+
+GALLIVM_SOURCES = \
+        gallivm.cpp  \
+        gallivm_cpu.cpp \
+        instructions.cpp  \
+        loweringpass.cpp \
+        tgsitollvm.cpp \
+        storage.cpp \
+        storagesoa.cpp \
+        instructionssoa.cpp
+
+INC_SOURCES = gallivm_builtins.cpp gallivmsoabuiltins.cpp
+
+CPP_SOURCES = \
+	$(GALLIVM_SOURCES)
+
+C_SOURCES =
+ASM_SOURCES =
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+          $(CPP_SOURCES:.cpp=.o) \
+	  $(ASM_SOURCES:.S=.o)
+
+### Include directories
+INCLUDES = \
+	-I. \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/include
+
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.cpp.o:
+	$(CXX) -c $(INCLUDES) $(LLVM_CXXFLAGS) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES)  $< -o $@
+
+##### TARGETS #####
+
+default:: depend symlinks $(LIBNAME)
+
+
+$(LIBNAME): $(OBJECTS) Makefile
+	$(TOP)/bin/mklib -o $@ -static $(OBJECTS)
+
+
+depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \
+		$(ASM_SOURCES) $(INC_SOURCES) 2> /dev/null
+
+
+gallivm_builtins.cpp: llvm_builtins.c
+	clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp1.bin
+	(echo "static const unsigned char llvm_builtins_data[] = {"; od -txC temp1.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/,0x00};/") >$@
+	rm temp1.bin
+
+gallivmsoabuiltins.cpp: soabuiltins.c
+	clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp2.bin
+	(echo "static const unsigned char soabuiltins_data[] = {"; od -txC temp2.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/,0x00};/") >$@
+	rm temp2.bin
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` `find ../include`
+
+
+# Remove .o and backup files
+clean:
+	-rm -f *.o */*.o *~ *.so *~ server/*.o
+	-rm -f depend depend.bak
+	-rm -f gallivm_builtins.cpp
+	-rm -f gallivmsoabuiltins.cpp
+
+symlinks:
+
+
+include depend
diff --git a/src/gallium/auxiliary/gallivm/SConscript b/src/gallium/auxiliary/gallivm/SConscript
new file mode 100644
index 0000000000..c0aa51b90a
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/SConscript
@@ -0,0 +1,16 @@
+Import('*')
+
+gallivm = env.ConvenienceLibrary(
+	target = 'gallivm',
+	source = [
+        'gallivm.cpp',
+        'gallivm_cpu.cpp',
+        'instructions.cpp',
+        'loweringpass.cpp',
+        'tgsitollvm.cpp',
+        'storage.cpp',
+        'storagesoa.cpp',
+        'instructionssoa.cpp',
+	])
+
+auxiliaries.insert(0, gallivm)
diff --git a/src/gallium/auxiliary/gallivm/gallivm.cpp b/src/gallium/auxiliary/gallivm/gallivm.cpp
new file mode 100644
index 0000000000..f4af5cc8ad
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/gallivm.cpp
@@ -0,0 +1,332 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin zack@tungstengraphics.com
+  */
+#ifdef MESA_LLVM
+
+#include "gallivm.h"
+#include "gallivm_p.h"
+
+#include "instructions.h"
+#include "loweringpass.h"
+#include "storage.h"
+#include "tgsitollvm.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/tgsi_exec.h"
+#include "tgsi/tgsi_dump.h"
+
+#include <llvm/Module.h>
+#include <llvm/CallingConv.h>
+#include <llvm/Constants.h>
+#include <llvm/DerivedTypes.h>
+#include <llvm/Instructions.h>
+#include <llvm/ModuleProvider.h>
+#include <llvm/Pass.h>
+#include <llvm/PassManager.h>
+#include <llvm/Attributes.h>
+#include <llvm/Support/PatternMatch.h>
+#include <llvm/ExecutionEngine/JIT.h>
+#include <llvm/ExecutionEngine/Interpreter.h>
+#include <llvm/ExecutionEngine/GenericValue.h>
+#include <llvm/Support/MemoryBuffer.h>
+#include <llvm/LinkAllPasses.h>
+#include <llvm/Analysis/Verifier.h>
+#include <llvm/Analysis/LoopPass.h>
+#include <llvm/Target/TargetData.h>
+#include <llvm/Bitcode/ReaderWriter.h>
+#include <llvm/Transforms/Utils/Cloning.h>
+
+#include <sstream>
+#include <fstream>
+#include <iostream>
+
+static int GLOBAL_ID = 0;
+
+using namespace llvm;
+
+static inline
+void AddStandardCompilePasses(PassManager &PM)
+{
+   PM.add(new LoweringPass());
+   PM.add(createVerifierPass());                  // Verify that input is correct
+
+   PM.add(createLowerSetJmpPass());          // Lower llvm.setjmp/.longjmp
+
+   //PM.add(createStripSymbolsPass(true));
+
+   PM.add(createRaiseAllocationsPass());     // call %malloc -> malloc inst
+   PM.add(createCFGSimplificationPass());    // Clean up disgusting code
+   PM.add(createPromoteMemoryToRegisterPass());// Kill useless allocas
+   PM.add(createGlobalOptimizerPass());      // Optimize out global vars
+   PM.add(createGlobalDCEPass());            // Remove unused fns and globs
+   PM.add(createIPConstantPropagationPass());// IP Constant Propagation
+   PM.add(createDeadArgEliminationPass());   // Dead argument elimination
+   PM.add(createInstructionCombiningPass()); // Clean up after IPCP & DAE
+   PM.add(createCFGSimplificationPass());    // Clean up after IPCP & DAE
+
+   PM.add(createPruneEHPass());              // Remove dead EH info
+
+   PM.add(createFunctionInliningPass());   // Inline small functions
+   PM.add(createArgumentPromotionPass());    // Scalarize uninlined fn args
+
+   PM.add(createTailDuplicationPass());      // Simplify cfg by copying code
+   PM.add(createInstructionCombiningPass()); // Cleanup for scalarrepl.
+   PM.add(createCFGSimplificationPass());    // Merge & remove BBs
+   PM.add(createScalarReplAggregatesPass()); // Break up aggregate allocas
+   PM.add(createInstructionCombiningPass()); // Combine silly seq's
+   PM.add(createCondPropagationPass());      // Propagate conditionals
+
+   PM.add(createTailCallEliminationPass());  // Eliminate tail calls
+   PM.add(createCFGSimplificationPass());    // Merge & remove BBs
+   PM.add(createReassociatePass());          // Reassociate expressions
+   PM.add(createLoopRotatePass());
+   PM.add(createLICMPass());                 // Hoist loop invariants
+   PM.add(createLoopUnswitchPass());         // Unswitch loops.
+   PM.add(createLoopIndexSplitPass());       // Index split loops.
+   PM.add(createInstructionCombiningPass()); // Clean up after LICM/reassoc
+   PM.add(createIndVarSimplifyPass());       // Canonicalize indvars
+   PM.add(createLoopUnrollPass());           // Unroll small loops
+   PM.add(createInstructionCombiningPass()); // Clean up after the unroller
+   PM.add(createGVNPass());                  // Remove redundancies
+   PM.add(createSCCPPass());                 // Constant prop with SCCP
+
+   // Run instcombine after redundancy elimination to exploit opportunities
+   // opened up by them.
+   PM.add(createInstructionCombiningPass());
+   PM.add(createCondPropagationPass());      // Propagate conditionals
+
+   PM.add(createDeadStoreEliminationPass()); // Delete dead stores
+   PM.add(createAggressiveDCEPass());        // SSA based 'Aggressive DCE'
+   PM.add(createCFGSimplificationPass());    // Merge & remove BBs
+   PM.add(createSimplifyLibCallsPass());     // Library Call Optimizations
+   PM.add(createDeadTypeEliminationPass());  // Eliminate dead types
+   PM.add(createConstantMergePass());        // Merge dup global constants
+}
+
+void gallivm_prog_delete(struct gallivm_prog *prog)
+{
+   delete prog->module;
+   prog->module = 0;
+   prog->function = 0;
+   free(prog);
+}
+
+static inline void
+constant_interpolation(float (*inputs)[16][4],
+                       const struct tgsi_interp_coef *coefs,
+                       unsigned attrib,
+                       unsigned chan)
+{
+   unsigned i;
+
+   for (i = 0; i < QUAD_SIZE; ++i) {
+      inputs[i][attrib][chan] = coefs[attrib].a0[chan];
+   }
+}
+
+static inline void
+linear_interpolation(float (*inputs)[16][4],
+                     const struct tgsi_interp_coef *coefs,
+                     unsigned attrib,
+                     unsigned chan)
+{
+   unsigned i;
+
+   for( i = 0; i < QUAD_SIZE; i++ ) {
+      const float x = inputs[i][0][0];
+      const float y = inputs[i][0][1];
+
+      inputs[i][attrib][chan] =
+         coefs[attrib].a0[chan] +
+         coefs[attrib].dadx[chan] * x +
+         coefs[attrib].dady[chan] * y;
+   }
+}
+
+static inline void
+perspective_interpolation(float (*inputs)[16][4],
+                          const struct tgsi_interp_coef *coefs,
+                          unsigned attrib,
+                          unsigned chan )
+{
+   unsigned i;
+
+   for( i = 0; i < QUAD_SIZE; i++ ) {
+      const float x = inputs[i][0][0];
+      const float y = inputs[i][0][1];
+      /* WPOS.w here is really 1/w */
+      const float w = 1.0f / inputs[i][0][3];
+      assert(inputs[i][0][3] != 0.0);
+
+      inputs[i][attrib][chan] =
+         (coefs[attrib].a0[chan] +
+          coefs[attrib].dadx[chan] * x +
+          coefs[attrib].dady[chan] * y) * w;
+   }
+}
+
+void gallivm_ir_dump(struct gallivm_ir *ir, const char *file_prefix)
+{
+   if (!ir || !ir->module)
+      return;
+
+   if (file_prefix) {
+      std::ostringstream stream;
+      stream << file_prefix;
+      stream << ir->id;
+      stream << ".ll";
+      std::string name = stream.str();
+      std::ofstream out(name.c_str());
+      if (!out) {
+         std::cerr<<"Can't open file : "<<stream.str()<<std::endl;;
+         return;
+      }
+      out << (*ir->module);
+      out.close();
+   } else {
+      const llvm::Module::FunctionListType &funcs = ir->module->getFunctionList();
+      llvm::Module::FunctionListType::const_iterator itr;
+      std::cout<<"; ---------- Start shader "<<ir->id<<std::endl;
+      for (itr = funcs.begin(); itr != funcs.end(); ++itr) {
+         const llvm::Function &func = (*itr);
+         std::string name = func.getName();
+         const llvm::Function *found = 0;
+         if (name.find("vs_shader") != std::string::npos ||
+             name.find("fs_shader") != std::string::npos ||
+             name.find("function") != std::string::npos)
+            found = &func;
+         if (found) {
+            std::cout<<*found<<std::endl;
+         }
+      }
+      std::cout<<"; ---------- End shader "<<ir->id<<std::endl;
+   }
+}
+
+
+void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog,
+                                     float (*inputs)[16][4],
+                                     const struct tgsi_interp_coef *coef)
+{
+   for (int i = 0; i < prog->num_interp; ++i) {
+      const gallivm_interpolate &interp = prog->interpolators[i];
+      switch (interp.type) {
+      case TGSI_INTERPOLATE_CONSTANT:
+         constant_interpolation(inputs, coef, interp.attrib, interp.chan);
+         break;
+
+      case TGSI_INTERPOLATE_LINEAR:
+         linear_interpolation(inputs, coef, interp.attrib, interp.chan);
+         break;
+
+      case TGSI_INTERPOLATE_PERSPECTIVE:
+         perspective_interpolation(inputs, coef, interp.attrib, interp.chan);
+         break;
+
+      default:
+         assert( 0 );
+      }
+   }
+}
+
+
+struct gallivm_ir * gallivm_ir_new(enum gallivm_shader_type type)
+{
+   struct gallivm_ir *ir =
+      (struct gallivm_ir *)calloc(1, sizeof(struct gallivm_ir));
+   ++GLOBAL_ID;
+   ir->id   = GLOBAL_ID;
+   ir->type = type;
+
+   return ir;
+}
+
+void gallivm_ir_set_layout(struct gallivm_ir *ir,
+                           enum gallivm_vector_layout layout)
+{
+   ir->layout = layout;
+}
+
+void gallivm_ir_set_components(struct gallivm_ir *ir, int num)
+{
+   ir->num_components = num;
+}
+
+void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir,
+                               const struct tgsi_token *tokens)
+{
+   std::cout << "Creating llvm from: " <<std::endl;
+   tgsi_dump(tokens, 0);
+
+   llvm::Module *mod = tgsi_to_llvmir(ir, tokens);
+   ir->module = mod;
+   gallivm_ir_dump(ir, 0);
+}
+
+void gallivm_ir_delete(struct gallivm_ir *ir)
+{
+   delete ir->module;
+   free(ir);
+}
+
+struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir)
+{
+   struct gallivm_prog *prog =
+      (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog));
+
+   std::cout << "Before optimizations:"<<std::endl;
+   ir->module->dump();
+   std::cout<<"-------------------------------"<<std::endl;
+
+   PassManager veri;
+   veri.add(createVerifierPass());
+   veri.run(*ir->module);
+   llvm::Module *mod = llvm::CloneModule(ir->module);
+   prog->num_consts = ir->num_consts;
+   memcpy(prog->interpolators, ir->interpolators, sizeof(prog->interpolators));
+   prog->num_interp = ir->num_interp;
+
+   /* Run optimization passes over it */
+   PassManager passes;
+   passes.add(new TargetData(mod));
+   AddStandardCompilePasses(passes);
+   passes.run(*mod);
+   prog->module = mod;
+
+   std::cout << "After optimizations:"<<std::endl;
+   mod->dump();
+
+   return prog;
+}
+
+#endif /* MESA_LLVM */
diff --git a/src/gallium/auxiliary/gallivm/gallivm.h b/src/gallium/auxiliary/gallivm/gallivm.h
new file mode 100644
index 0000000000..36a64a7747
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/gallivm.h
@@ -0,0 +1,118 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin zack@tungstengraphics.com
+  */
+
+#ifndef GALLIVM_H
+#define GALLIVM_H
+
+/*
+  LLVM representation consists of two stages - layout independent
+  intermediate representation gallivm_ir and driver specific
+  gallivm_prog. TGSI is first being translated into gallivm_ir
+  after that driver can set number of options on gallivm_ir and
+  have it compiled into gallivm_prog. gallivm_prog can be either
+  executed (assuming there's LLVM JIT backend for the current
+  target) or machine code generation can be done (assuming there's
+  a LLVM code generator for thecurrent target)
+ */
+#if defined __cplusplus
+extern "C" {
+#endif
+
+#include "pipe/p_state.h"
+
+#ifdef MESA_LLVM
+
+struct tgsi_token;
+
+struct gallivm_ir;
+struct gallivm_prog;
+struct gallivm_cpu_engine;
+struct tgsi_interp_coef;
+struct tgsi_sampler;
+struct tgsi_exec_vector;
+
+enum gallivm_shader_type {
+   GALLIVM_VS,
+   GALLIVM_FS
+};
+
+enum gallivm_vector_layout {
+   GALLIVM_AOS,
+   GALLIVM_SOA
+};
+
+struct gallivm_ir *gallivm_ir_new(enum gallivm_shader_type type);
+void               gallivm_ir_set_layout(struct gallivm_ir *ir,
+                                         enum gallivm_vector_layout layout);
+void               gallivm_ir_set_components(struct gallivm_ir *ir, int num);
+void               gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir,
+                                             const struct tgsi_token *tokens);
+void               gallivm_ir_delete(struct gallivm_ir *ir);
+
+
+struct gallivm_prog *gallivm_ir_compile(struct gallivm_ir *ir);
+
+void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog,
+                                     float (*inputs)[PIPE_MAX_SHADER_INPUTS][4],
+                                     const struct tgsi_interp_coef *coefs);
+void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix);
+
+
+struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog);
+struct gallivm_cpu_engine *gallivm_global_cpu_engine();
+int gallivm_cpu_vs_exec(struct gallivm_prog *prog,
+                        struct tgsi_exec_machine *machine,
+                        const float (*input)[4],
+                        unsigned num_inputs,
+                        float (*output)[4],
+                        unsigned num_outputs,
+                        const float (*constants)[4],
+                        unsigned count,
+                        unsigned input_stride,
+                        unsigned output_stride);
+int gallivm_cpu_fs_exec(struct gallivm_prog *prog,
+                        float x, float y,
+                        float (*dests)[PIPE_MAX_SHADER_INPUTS][4],
+                        float (*inputs)[PIPE_MAX_SHADER_INPUTS][4],
+                        float (*consts)[4],
+                        struct tgsi_sampler *samplers);
+void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog);
+void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee);
+
+
+#endif /* MESA_LLVM */
+
+#if defined __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp b/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp
new file mode 100644
index 0000000000..634bac0150
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp
@@ -0,0 +1,140 @@
+static const unsigned char llvm_builtins_data[] = {
+0x42,0x43,0xc0,0xde,0x21,0x0c,0x00,0x00,0x27,0x02,0x00,0x00,0x01,0x10,0x00,0x00,
+0x10,0x00,0x00,0x00,0x07,0x81,0x23,0x91,0x41,0xc8,0x04,0x49,0x06,0x10,0x32,0x39,
+0x92,0x01,0x84,0x0c,0x25,0x05,0x08,0x19,0x1e,0x04,0x8b,0x62,0x80,0x14,0x45,0x02,
+0x42,0x92,0x0b,0x42,0xa4,0x10,0x32,0x14,0x38,0x08,0x18,0x49,0x0a,0x32,0x44,0x24,
+0x48,0x0a,0x90,0x21,0x23,0x44,0x72,0x80,0x8c,0x14,0x21,0x86,0x0a,0x8a,0x0a,0x64,
+0x0c,0x1f,0x00,0x00,0x49,0x18,0x00,0x00,0x03,0x00,0x00,0x00,0x0b,0x84,0xff,0xff,
+0xff,0xff,0x1f,0xc0,0x00,0x00,0x00,0x00,0x51,0x20,0x00,0x00,0x12,0x00,0x00,0x00,
+0x32,0x22,0x48,0x09,0x20,0x65,0x82,0x84,0x00,0x26,0x45,0x48,0x05,0x09,0x26,0x45,
+0xc6,0x05,0x42,0x52,0x26,0x08,0xae,0x19,0x80,0x61,0x04,0x02,0x98,0x23,0x00,0x83,
+0x29,0x80,0x21,0x00,0xb2,0x73,0x04,0x01,0x51,0x8a,0xf4,0x08,0x92,0xa4,0x39,0x47,
+0x80,0x50,0x2b,0x03,0x00,0xa0,0x08,0x21,0x5c,0x46,0x2b,0x44,0x08,0x21,0xd4,0x40,
+0x14,0x01,0x80,0x11,0x80,0x22,0x88,0x00,0x13,0x30,0x7c,0xc0,0x03,0x3b,0xf8,0x05,
+0x3b,0xa0,0x83,0x36,0xa8,0x07,0x77,0x58,0x07,0x77,0x78,0x87,0x7b,0x70,0x87,0x36,
+0x60,0x87,0x74,0x70,0x87,0x7a,0xc0,0x87,0x36,0x38,0x07,0x77,0xa8,0x87,0x0d,0xf7,
+0x50,0x0e,0x6d,0x00,0x0f,0x7a,0x60,0x07,0x74,0xa0,0x07,0x76,0x40,0x07,0x7a,0x60,
+0x07,0x74,0xd0,0x06,0xe9,0x10,0x07,0x7a,0x80,0x07,0x7a,0x80,0x07,0x6d,0x90,0x0e,
+0x78,0xa0,0x07,0x78,0xa0,0x07,0x78,0xd0,0x06,0xe9,0x10,0x07,0x76,0xa0,0x07,0x71,
+0x60,0x07,0x7a,0x10,0x07,0x76,0xd0,0x06,0xe9,0x30,0x07,0x72,0xa0,0x07,0x73,0x20,
+0x07,0x7a,0x30,0x07,0x72,0xd0,0x06,0xe9,0x60,0x07,0x74,0xa0,0x07,0x76,0x40,0x07,
+0x7a,0x60,0x07,0x74,0xd0,0x06,0xe6,0x30,0x07,0x72,0xa0,0x07,0x73,0x20,0x07,0x7a,
+0x30,0x07,0x72,0xd0,0x06,0xe6,0x60,0x07,0x74,0xa0,0x07,0x76,0x40,0x07,0x7a,0x60,
+0x07,0x74,0xd0,0x06,0xf6,0x60,0x07,0x74,0xa0,0x07,0x76,0x40,0x07,0x7a,0x60,0x07,
+0x74,0xd0,0x06,0xf6,0x10,0x07,0x72,0x80,0x07,0x7a,0x10,0x07,0x72,0x80,0x07,0x7a,
+0x10,0x07,0x72,0x80,0x07,0x6d,0x10,0x0e,0x70,0xa0,0x07,0x70,0xa0,0x07,0x76,0x40,
+0x07,0x6d,0x60,0x0e,0x78,0x00,0x07,0x7a,0x10,0x07,0x72,0x80,0x07,0x7a,0x10,0x07,
+0x72,0x80,0x07,0x3a,0x0f,0x84,0x48,0x20,0x23,0x24,0x40,0x00,0x62,0x67,0x88,0x9f,
+0x19,0x92,0x24,0x00,0x10,0x04,0x00,0x00,0x00,0x43,0x92,0x04,0x08,0x00,0x00,0x00,
+0x00,0x60,0x48,0xa2,0x00,0x40,0x10,0x00,0x00,0x00,0x0c,0x49,0x16,0x00,0x08,0x02,
+0x00,0x00,0x80,0x21,0x89,0x02,0x00,0x41,0x00,0x00,0x00,0x30,0x24,0x61,0x80,0x00,
+0x00,0x00,0x00,0x00,0x86,0x24,0x07,0x10,0x00,0x00,0x00,0x00,0xc0,0x90,0x44,0x01,
+0x80,0x20,0x00,0x00,0x00,0x18,0x92,0x1c,0x40,0x00,0x00,0x00,0x00,0x00,0x43,0x12,
+0x05,0x00,0x82,0x00,0x00,0x00,0x60,0x48,0x52,0x00,0x40,0x10,0x00,0x00,0x00,0x64,
+0x81,0x00,0x00,0x00,0x10,0x00,0x00,0x00,0x32,0x1e,0x98,0x10,0x19,0x11,0x4c,0x90,
+0x8c,0x09,0x26,0x47,0xc6,0x04,0x43,0x8a,0x8a,0x59,0x8b,0x43,0x50,0xd2,0x09,0x02,
+0x81,0xd2,0x73,0x50,0xc9,0x0c,0x2a,0x99,0x41,0x25,0x33,0xa8,0x64,0x56,0x28,0x66,
+0x2d,0x0e,0x41,0xcf,0x2a,0x15,0x04,0x4a,0xcf,0x41,0x25,0x33,0xa8,0x64,0x06,0x95,
+0xcc,0xa0,0x92,0x59,0x01,0x00,0x00,0x00,0x53,0x82,0x26,0x0c,0x04,0x00,0x00,0x00,
+0x22,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00,0x05,0x00,0x00,0x00,
+0x04,0xc6,0x08,0x40,0x10,0x04,0xe1,0x70,0x18,0x23,0x00,0x41,0x10,0x84,0xc3,0x60,
+0x04,0x00,0x00,0x00,0xc3,0x0d,0xce,0x43,0x4c,0x37,0x3c,0x8e,0x34,0xdc,0x30,0x41,
+0xc2,0x74,0x03,0x34,0x51,0xc3,0x0d,0x4d,0x44,0x4c,0x37,0x44,0x8d,0x35,0x56,0x01,
+0x04,0xc3,0x55,0x21,0x16,0x0e,0x04,0x00,0x0f,0x00,0x00,0x00,0xd6,0x10,0x00,0xe6,
+0x10,0x04,0x76,0x81,0x00,0x3e,0x30,0x0c,0x91,0x4f,0x1b,0x05,0x21,0x30,0x8f,0x6d,
+0x13,0x48,0xe0,0x03,0xc3,0x10,0xf9,0xb4,0x55,0x20,0x81,0x0f,0x0c,0x43,0xe4,0xd7,
+0x66,0x41,0x08,0xcc,0xa3,0x1f,0x40,0x41,0x34,0x53,0x84,0x99,0xc4,0x20,0x30,0x8f,
+0x61,0x10,0x02,0xb0,0x2c,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00,
+0x27,0x00,0x00,0x00,0x13,0x04,0x43,0x2c,0x10,0x00,0x00,0x00,0x08,0x00,0x00,0x00,
+0x24,0x8a,0xa0,0x0c,0x46,0x00,0x4a,0x80,0xc2,0x1c,0x84,0x55,0x55,0xd6,0x1c,0x84,
+0x45,0x51,0x16,0x81,0x19,0x80,0x11,0x80,0x31,0x02,0x10,0x04,0x41,0xfc,0x03,0x00,
+0x63,0x08,0x0d,0x34,0xdc,0x70,0x55,0xc2,0x2c,0x43,0x20,0x60,0x73,0x0c,0xd3,0x15,
+0x8d,0x21,0x34,0xd1,0x18,0x42,0xf3,0x8c,0x55,0x00,0x81,0xa0,0x6d,0x73,0x0c,0x19,
+0xe7,0x60,0x87,0x52,0x38,0x10,0x00,0x00,0x10,0x00,0x00,0x00,0x27,0x50,0x20,0x05,
+0xd1,0x0c,0x17,0x60,0x20,0xc5,0x74,0x10,0x8d,0x65,0x14,0x13,0xf3,0xd4,0xb4,0x6d,
+0x14,0x13,0xf3,0xd4,0xb8,0x69,0x14,0x13,0xf3,0xd4,0xb6,0x75,0x14,0x13,0xf3,0xd4,
+0xba,0x35,0x0c,0x13,0xf3,0xd8,0x05,0x31,0x31,0x8f,0x6e,0x1c,0x84,0x00,0x2c,0xcb,
+0x01,0x14,0x44,0x33,0x45,0x98,0x61,0x0c,0x02,0xf3,0x00,0x00,0x00,0x00,0x00,0x00,
+0x61,0x20,0x00,0x00,0x81,0x00,0x00,0x00,0x13,0x04,0x4d,0x2c,0x10,0x00,0x00,0x00,
+0x04,0x00,0x00,0x00,0x24,0xca,0x60,0x04,0xa0,0x04,0x8a,0x80,0xc2,0x0c,0x00,0x91,
+0x11,0x00,0x00,0x00,0x63,0x08,0x4d,0x64,0x16,0xc1,0xe1,0x86,0xab,0x22,0x66,0x19,
+0x02,0x01,0x1b,0x43,0x70,0xa2,0x59,0x82,0x61,0x0c,0xe1,0x89,0x66,0x09,0x86,0x81,
+0x0a,0x20,0x0b,0x34,0x61,0x8e,0x81,0xda,0xa2,0x31,0x84,0x46,0xb2,0x8e,0xe0,0x70,
+0x83,0x57,0x11,0xb3,0x0c,0x44,0xf1,0x8d,0x21,0x38,0xd2,0x2c,0x81,0x31,0x86,0xf0,
+0x48,0xb3,0x04,0xc6,0x40,0x05,0x00,0x06,0x44,0x18,0x14,0x73,0x0c,0x9c,0x18,0x48,
+0x63,0x08,0xcd,0x64,0x64,0x40,0x70,0xb8,0xa1,0x0c,0x2a,0x62,0x96,0xe1,0x40,0xcc,
+0x60,0x0c,0xc1,0x99,0x66,0x09,0x92,0x31,0x84,0x67,0x9a,0x25,0x48,0x06,0x2a,0x80,
+0x33,0x38,0xd0,0x00,0x99,0x63,0x18,0x83,0x34,0x98,0xc6,0x10,0x1a,0xc8,0xd6,0x80,
+0xe0,0x70,0x03,0x1b,0x54,0xc4,0x2c,0x83,0xb2,0xb4,0xc1,0x18,0x82,0x03,0xcd,0x12,
+0x30,0x63,0x08,0x0f,0x34,0x4b,0xc0,0x0c,0x54,0x00,0x6e,0xa0,0xbc,0xc1,0x32,0xc7,
+0xa0,0x06,0x70,0x00,0x61,0x1c,0x84,0x03,0x01,0x00,0x00,0x00,0x4e,0x00,0x00,0x00,
+0x76,0x52,0x4c,0xcc,0x73,0xd3,0x24,0x05,0x64,0xec,0xcd,0x8d,0xcc,0xe5,0x87,0x46,
+0xc6,0x50,0x8a,0x89,0x79,0xee,0xdb,0x54,0x8a,0x89,0x79,0xee,0xdd,0x1a,0x88,0x89,
+0x79,0x68,0x73,0x20,0x26,0xe6,0xa9,0xed,0x81,0x98,0x98,0xc7,0x36,0x0b,0x62,0x62,
+0x9e,0xdb,0x32,0x88,0x89,0x79,0x72,0xd3,0x20,0x26,0xe6,0xd9,0x8d,0x83,0x98,0x98,
+0xa7,0xb7,0x95,0x62,0x62,0x9e,0xbb,0x27,0x2d,0x20,0x63,0x6f,0x6e,0x64,0x2e,0x3a,
+0x34,0x35,0x56,0x62,0x08,0x4e,0x53,0xd9,0xba,0xb5,0x14,0x02,0xf3,0xe0,0xf5,0x25,
+0x2c,0x82,0xd3,0x0c,0xbe,0xe0,0x34,0xd3,0x8d,0x9b,0x88,0x21,0x38,0xcd,0x60,0xd7,
+0x24,0x01,0x63,0xec,0xcd,0x8d,0xcc,0x45,0x87,0x44,0x80,0x8c,0xbd,0xb9,0x91,0xb9,
+0xfc,0xc4,0xd0,0x90,0x02,0x8c,0xb1,0x37,0x37,0x32,0x97,0x1f,0x73,0x29,0x26,0xe6,
+0xc1,0x71,0x7b,0x29,0x26,0xe6,0xc1,0x77,0xfb,0x28,0x04,0xe6,0xa9,0x6f,0x52,0x01,
+0x32,0xf6,0xe6,0x46,0xe6,0xa2,0x13,0x73,0x63,0x18,0x83,0xc0,0x3c,0xb6,0x41,0x08,
+0x4e,0x33,0x58,0x47,0x31,0x31,0x4f,0x5d,0x1f,0xc3,0x22,0x38,0xcd,0xe0,0x0b,0x4e,
+0x33,0xe1,0xbc,0xa5,0x18,0x82,0xd3,0x0c,0x77,0x6e,0x20,0xc5,0xc4,0x3c,0xb5,0x4e,
+0x3a,0x40,0xc6,0xde,0xdc,0xc8,0x5c,0x7e,0x64,0x70,0x2c,0xa4,0x98,0x98,0xa7,0xee,
+0x6f,0x20,0x11,0x9c,0x66,0xf0,0x05,0xa7,0x99,0xec,0x82,0x10,0x9c,0xa6,0x32,0x93,
+0x42,0x60,0x1e,0x7b,0xb7,0x98,0x62,0x62,0x9e,0xbc,0x36,0x16,0x43,0x70,0x9a,0x0a,
+0xa7,0x6d,0xa4,0x98,0x98,0xc7,0xbe,0x8d,0xa4,0x98,0x98,0xc7,0xce,0x0d,0xc6,0x10,
+0x9c,0x66,0xc0,0x7b,0x12,0x02,0x32,0xf6,0xe6,0x46,0xe6,0xa2,0x33,0x13,0x73,0x06,
+0x8b,0xe0,0x34,0x83,0x2f,0x38,0xcd,0x64,0xd3,0x07,0x50,0x10,0xcd,0x14,0x61,0xe6,
+0x61,0x08,0x4e,0x53,0xd5,0x36,0x00,0x00,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00,
+0x4a,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00,0x07,0x00,0x00,0x00,
+0x24,0xca,0x60,0x04,0xa0,0x04,0x8a,0x80,0xc2,0x0c,0x00,0xb9,0x61,0x0c,0x04,0x10,
+0x1e,0xe1,0x19,0xc6,0x40,0x02,0xe1,0x11,0x1e,0x00,0x00,0x00,0x63,0x08,0xcd,0x63,
+0x15,0xc1,0x31,0x84,0x06,0xb2,0x8b,0xe0,0x18,0x42,0x13,0x59,0x46,0x70,0x0c,0xa1,
+0x71,0x6c,0x23,0x38,0x16,0x02,0x04,0xc7,0x64,0x61,0x1a,0x37,0x16,0x01,0x04,0x48,
+0x35,0xc7,0x20,0x79,0xcf,0x58,0x04,0x10,0x20,0xd5,0x1c,0xc3,0x07,0x06,0xd0,0x58,
+0x04,0x10,0x20,0xd5,0x1c,0x43,0x18,0x88,0x41,0x34,0x16,0x01,0x04,0x48,0x35,0xc7,
+0x30,0x06,0x64,0xe0,0x98,0x47,0xd0,0xc0,0x80,0xa0,0x89,0x01,0x41,0x23,0x03,0x82,
+0x63,0x21,0x40,0x70,0x50,0x66,0x70,0x06,0x68,0x90,0x06,0x58,0x06,0xe1,0x40,0x00,
+0x25,0x00,0x00,0x00,0x56,0x52,0x4c,0xcc,0x73,0xd3,0x56,0x41,0x4c,0xcc,0x53,0xdb,
+0x05,0x31,0x31,0xcf,0x6d,0x19,0xc4,0xc4,0x3c,0xba,0x6d,0x10,0x13,0xf3,0xf4,0xd6,
+0x41,0x08,0xc0,0xb2,0x18,0x46,0x21,0x38,0x4d,0x85,0x9b,0x46,0x21,0x38,0x4d,0xb5,
+0x9b,0x8a,0x21,0x00,0xcb,0x82,0xdf,0x66,0x62,0x08,0x4e,0x53,0xdd,0xb7,0x9d,0x18,
+0x82,0xd3,0x54,0xb7,0x6e,0x28,0x86,0xe0,0x34,0xd5,0xdd,0xdb,0x47,0x31,0x31,0x4f,
+0x9d,0x9b,0x87,0x21,0x00,0xcb,0x52,0xdf,0x06,0x62,0x08,0xc0,0xb2,0xd4,0xbc,0x59,
+0x10,0x82,0xd3,0x54,0x96,0x62,0x08,0x4e,0x53,0xe1,0xb6,0x85,0x14,0x13,0xf3,0xd8,
+0xb4,0x8d,0x14,0x13,0xf3,0xd8,0xb9,0x89,0x18,0x02,0xb0,0x2c,0xf6,0x6d,0x24,0x86,
+0x00,0x2c,0x8b,0xcd,0x1b,0x87,0x21,0x38,0x4d,0x55,0xd3,0xd6,0x30,0x54,0xc0,0x72,
+0x00,0x05,0xd1,0x4c,0x11,0x06,0x00,0x00,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00,
+0x19,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00,0x03,0x00,0x00,0x00,
+0x24,0x4a,0x60,0x04,0x80,0xc2,0x0c,0x00,0x00,0x00,0x00,0x00,0x63,0x08,0xcd,0x33,
+0x16,0x01,0x04,0x48,0x34,0xc7,0x00,0x49,0xcf,0x58,0x04,0x10,0x28,0xd1,0x1c,0xc3,
+0x44,0x39,0x58,0x85,0x03,0x01,0x00,0x00,0x0a,0x00,0x00,0x00,0x26,0x41,0x08,0xc0,
+0xb2,0x18,0x45,0x21,0x00,0xcb,0xb2,0x5b,0x04,0x31,0x31,0x8f,0x6d,0x13,0xc4,0xc4,
+0x3c,0xb9,0x35,0x0c,0x15,0xb0,0x58,0x05,0x31,0x31,0x4f,0x7f,0x00,0x05,0xd1,0x4c,
+0x11,0x06,0x00,0x00,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00,0x1b,0x00,0x00,0x00,
+0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00,0x03,0x00,0x00,0x00,0x24,0xca,0x60,0x04,
+0xa0,0x04,0x8a,0x80,0xc2,0x0c,0x00,0x00,0x63,0x08,0xcd,0x33,0x16,0x01,0x04,0xca,
+0x34,0xc7,0x20,0x51,0xcf,0x1c,0x43,0x45,0x41,0x73,0x0c,0x16,0x15,0xcd,0x31,0x5c,
+0x94,0x83,0x58,0x38,0x10,0x00,0x00,0x00,0x0b,0x00,0x00,0x00,0x76,0x51,0x4c,0xcc,
+0x53,0xdb,0x86,0x51,0x4c,0xcc,0x53,0xe7,0x36,0x41,0x4c,0xcc,0x63,0x5b,0x05,0x31,
+0x31,0x8f,0x6e,0x16,0xc4,0xc4,0x3c,0xbd,0x51,0x10,0x02,0xb0,0x2c,0xd6,0x30,0x54,
+0xc0,0x72,0x00,0x05,0xd1,0x4c,0x11,0x06,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00,
+0x2c,0x00,0x00,0x00,0x13,0x04,0x45,0x2c,0x10,0x00,0x00,0x00,0x03,0x00,0x00,0x00,
+0x24,0xca,0xa0,0x04,0x46,0x00,0x8a,0x80,0xc0,0x08,0x00,0x00,0x63,0x08,0x0d,0x34,
+0xdc,0x30,0x49,0xc4,0x2c,0x03,0x11,0x50,0x63,0x08,0xcd,0x33,0xdc,0x50,0x49,0xc4,
+0x2c,0x03,0x21,0x58,0x63,0x08,0x4d,0x34,0xdc,0x70,0x49,0xc4,0x2c,0x03,0x31,0x60,
+0x63,0x08,0x8d,0x33,0xdc,0x90,0x49,0x84,0x69,0x22,0x70,0xc3,0x27,0x1c,0x08,0x00,
+0x17,0x00,0x00,0x00,0x96,0x51,0x4c,0xcc,0x53,0xdf,0x66,0x41,0x08,0xcc,0x83,0xdb,
+0x04,0x31,0x31,0x4f,0x6d,0x15,0xc4,0xc4,0x3c,0xb7,0x61,0x10,0x02,0xf3,0xf0,0x76,
+0x41,0x4c,0xcc,0xb3,0x1f,0x81,0x11,0x11,0x13,0x15,0x35,0x37,0x90,0x2c,0x4e,0xf4,
+0x47,0x87,0x54,0xd7,0x17,0x70,0x2c,0x4e,0xf4,0x47,0x87,0x74,0x02,0xc8,0xe2,0x44,
+0x7f,0x74,0x48,0xb9,0x69,0x14,0x02,0xf3,0xd4,0xb8,0x6d,0x18,0x11,0x31,0x55,0xc0,
+0x62,0x0d,0x43,0x05,0x2c,0x07,0x50,0x10,0xcd,0x14,0x61,0x46,0x31,0x08,0xcc,0x03,
+0x00,0x00,0x00,0x00,0x71,0x20,0x00,0x00,0x12,0x00,0x00,0x00,0x66,0x40,0x54,0x82,
+0x23,0x19,0xc3,0xa0,0x20,0x8b,0x1d,0x18,0x4f,0x84,0x34,0x53,0x61,0x03,0xc4,0xe3,
+0x58,0x85,0x05,0x14,0xbe,0x34,0x45,0xb5,0x21,0x10,0x82,0x23,0x15,0x46,0x30,0x2c,
+0xc8,0x64,0x02,0x06,0xf0,0x3c,0x91,0x73,0x19,0x00,0xe1,0x4b,0x53,0x64,0x0a,0x84,
+0x84,0x34,0x85,0x25,0x0c,0x92,0x20,0x59,0xc1,0x20,0x30,0x8f,0x2d,0x10,0x95,0x84,
+0x34,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};
diff --git a/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp b/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp
new file mode 100644
index 0000000000..1bd00a0c2a
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp
@@ -0,0 +1,243 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin zack@tungstengraphics.com
+  */
+#ifdef MESA_LLVM
+
+#include "gallivm.h"
+#include "gallivm_p.h"
+
+#include "instructions.h"
+#include "loweringpass.h"
+#include "storage.h"
+#include "tgsitollvm.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/tgsi_exec.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include <llvm/Module.h>
+#include <llvm/CallingConv.h>
+#include <llvm/Constants.h>
+#include <llvm/DerivedTypes.h>
+#include <llvm/Instructions.h>
+#include <llvm/ModuleProvider.h>
+#include <llvm/Pass.h>
+#include <llvm/PassManager.h>
+#include <llvm/Attributes.h>
+#include <llvm/Support/PatternMatch.h>
+#include <llvm/ExecutionEngine/JIT.h>
+#include <llvm/ExecutionEngine/Interpreter.h>
+#include <llvm/ExecutionEngine/GenericValue.h>
+#include <llvm/Support/MemoryBuffer.h>
+#include <llvm/LinkAllPasses.h>
+#include <llvm/Analysis/Verifier.h>
+#include <llvm/Analysis/LoopPass.h>
+#include <llvm/Target/TargetData.h>
+#include <llvm/Bitcode/ReaderWriter.h>
+#include <llvm/Transforms/Utils/Cloning.h>
+
+#include <sstream>
+#include <fstream>
+#include <iostream>
+
+struct gallivm_cpu_engine {
+   llvm::ExecutionEngine *engine;
+};
+
+static struct gallivm_cpu_engine *CPU = 0;
+
+typedef int (*fragment_shader_runner)(float x, float y,
+                                      float (*dests)[16][4],
+                                      float (*inputs)[16][4],
+                                      int num_attribs,
+                                      float (*consts)[4], int num_consts,
+                                      struct tgsi_sampler *samplers);
+
+int gallivm_cpu_fs_exec(struct gallivm_prog *prog,
+                        float fx, float fy,
+                        float (*dests)[16][4],
+                        float (*inputs)[16][4],
+                        float (*consts)[4],
+                        struct tgsi_sampler *samplers)
+{
+   fragment_shader_runner runner = reinterpret_cast<fragment_shader_runner>(prog->function);
+   assert(runner);
+
+   return runner(fx, fy, dests, inputs, prog->num_interp,
+                 consts, prog->num_consts,
+                 samplers);
+}
+
+static inline llvm::Function *func_for_shader(struct gallivm_prog *prog)
+{
+   llvm::Module *mod = prog->module;
+   llvm::Function *func = 0;
+
+   switch (prog->type) {
+   case GALLIVM_VS:
+      func = mod->getFunction("vs_shader");
+      break;
+   case GALLIVM_FS:
+      func = mod->getFunction("fs_shader");
+      break;
+   default:
+      assert(!"Unknown shader type!");
+      break;
+   }
+   return func;
+}
+
+/*!
+  This function creates a CPU based execution engine for the given gallivm_prog.
+  gallivm_cpu_engine should be used as a singleton throughout the library. Before
+  executing gallivm_prog_exec one needs to call gallivm_cpu_jit_compile.
+  The gallivm_prog instance which is being passed to the constructor is being
+  automatically JIT compiled so one shouldn't call gallivm_cpu_jit_compile
+  with it again.
+ */
+struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog)
+{
+   struct gallivm_cpu_engine *cpu = (struct gallivm_cpu_engine *)
+                                    calloc(1, sizeof(struct gallivm_cpu_engine));
+   llvm::Module *mod = static_cast<llvm::Module*>(prog->module);
+   llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod);
+   llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false);
+   ee->DisableLazyCompilation();
+   cpu->engine = ee;
+
+   llvm::Function *func = func_for_shader(prog);
+
+   prog->function = ee->getPointerToFunction(func);
+   CPU = cpu;
+   return cpu;
+}
+
+
+/*!
+  This function JIT compiles the given gallivm_prog with the given cpu based execution engine.
+  The reference to the generated machine code entry point will be stored
+  in the gallivm_prog program. After executing this function one can call gallivm_prog_exec
+  in order to execute the gallivm_prog on the CPU.
+ */
+void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog)
+{
+   llvm::Module *mod = static_cast<llvm::Module*>(prog->module);
+   llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod);
+   llvm::ExecutionEngine *ee = cpu->engine;
+   assert(ee);
+   /*FIXME : why was this disabled ? we need it for pow/sqrt/... */
+   ee->DisableLazyCompilation(false);
+   ee->addModuleProvider(mp);
+
+   llvm::Function *func = func_for_shader(prog);
+   prog->function = ee->getPointerToFunction(func);
+}
+
+void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu)
+{
+   free(cpu);
+}
+
+struct gallivm_cpu_engine * gallivm_global_cpu_engine()
+{
+   return CPU;
+}
+
+
+typedef void (*vertex_shader_runner)(void *ainputs,
+                                     void *dests,
+                                     float (*aconsts)[4]);
+
+#define MAX_TGSI_VERTICES 4
+/*!
+  This function is used to execute the gallivm_prog in software. Before calling
+  this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile
+  function.
+ */
+int gallivm_cpu_vs_exec(struct gallivm_prog *prog,
+                        struct tgsi_exec_machine *machine,
+                        const float (*input)[4],
+                        unsigned num_inputs,
+                        float (*output)[4],
+                        unsigned num_outputs,
+                        const float (*constants)[4],
+                        unsigned count,
+                        unsigned input_stride,
+                        unsigned output_stride )
+{
+   unsigned int i, j;
+   unsigned slot;
+   vertex_shader_runner runner = reinterpret_cast<vertex_shader_runner>(prog->function);
+   assert(runner);
+
+   for (i = 0; i < count; i += MAX_TGSI_VERTICES) {
+      unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i);
+
+      /* Swizzle inputs.
+       */
+      for (j = 0; j < max_vertices; j++) {
+	 for (slot = 0; slot < num_inputs; slot++) {
+	    machine->Inputs[slot].xyzw[0].f[j] = input[slot][0];
+	    machine->Inputs[slot].xyzw[1].f[j] = input[slot][1];
+	    machine->Inputs[slot].xyzw[2].f[j] = input[slot][2];
+	    machine->Inputs[slot].xyzw[3].f[j] = input[slot][3];
+	 }
+
+	 input = (const float (*)[4])((const char *)input + input_stride);
+      }
+
+      /* run shader */
+      runner(machine->Inputs,
+             machine->Outputs,
+             (float (*)[4]) constants);
+
+      /* Unswizzle all output results
+       */
+      for (j = 0; j < max_vertices; j++) {
+         for (slot = 0; slot < num_outputs; slot++) {
+            output[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
+            output[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
+            output[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
+            output[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
+         }
+         output = (float (*)[4])((char *)output + output_stride);
+      }
+   }
+
+   return 0;
+}
+
+#endif
diff --git a/src/gallium/auxiliary/gallivm/gallivm_p.h b/src/gallium/auxiliary/gallivm/gallivm_p.h
new file mode 100644
index 0000000000..d2c5852bdf
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/gallivm_p.h
@@ -0,0 +1,110 @@
+#ifndef GALLIVM_P_H
+#define GALLIVM_P_H
+
+#ifdef MESA_LLVM
+
+#include "gallivm.h"
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_compiler.h"
+
+namespace llvm {
+   class Module;
+}
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+enum gallivm_shader_type;
+enum gallivm_vector_layout;
+
+struct gallivm_interpolate {
+   int attrib;
+   int chan;
+   int type;
+};
+
+struct gallivm_ir {
+   llvm::Module *module;
+   int id;
+   enum gallivm_shader_type type;
+   enum gallivm_vector_layout layout;
+   int num_components;
+   int   num_consts;
+
+   /* FIXME: this might not be enough for some shaders */
+   struct gallivm_interpolate interpolators[32*4];
+   int   num_interp;
+};
+
+struct gallivm_prog {
+   llvm::Module *module;
+   void *function;
+
+   int   id;
+   enum gallivm_shader_type type;
+
+   int   num_consts;
+
+   /* FIXME: this might not be enough for some shaders */
+   struct gallivm_interpolate interpolators[32*4];
+   int   num_interp;
+};
+
+static INLINE void gallivm_swizzle_components(int swizzle,
+                                              int *xc, int *yc,
+                                              int *zc, int *wc)
+{
+   int x = swizzle / 1000; swizzle -= x * 1000;
+   int y = swizzle / 100;  swizzle -= y * 100;
+   int z = swizzle / 10;   swizzle -= z * 10;
+   int w = swizzle;
+
+   if (xc) *xc = x;
+   if (yc) *yc = y;
+   if (zc) *zc = z;
+   if (wc) *wc = w;
+}
+
+static INLINE boolean gallivm_is_swizzle(int swizzle)
+{
+   const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 +
+                          TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W;
+   return swizzle != NO_SWIZZLE;
+}
+
+static INLINE int gallivm_x_swizzle(int swizzle)
+{
+   int x;
+   gallivm_swizzle_components(swizzle, &x, 0, 0, 0);
+   return x;
+}
+
+static INLINE int gallivm_y_swizzle(int swizzle)
+{
+   int y;
+   gallivm_swizzle_components(swizzle, 0, &y, 0, 0);
+   return y;
+}
+
+static INLINE int gallivm_z_swizzle(int swizzle)
+{
+   int z;
+   gallivm_swizzle_components(swizzle, 0, 0, &z, 0);
+   return z;
+}
+
+static INLINE int gallivm_w_swizzle(int swizzle)
+{
+   int w;
+   gallivm_swizzle_components(swizzle, 0, 0, 0, &w);
+   return w;
+}
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* MESA_LLVM */
+
+#endif
diff --git a/src/gallium/auxiliary/gallivm/instructions.cpp b/src/gallium/auxiliary/gallivm/instructions.cpp
new file mode 100644
index 0000000000..ee8162efce
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/instructions.cpp
@@ -0,0 +1,1193 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin zack@tungstengraphics.com
+  */
+#ifdef MESA_LLVM
+
+#include "instructions.h"
+
+#include "storage.h"
+
+#include "util/u_memory.h"
+
+#include <llvm/CallingConv.h>
+#include <llvm/Constants.h>
+#include <llvm/DerivedTypes.h>
+#include <llvm/Function.h>
+#include <llvm/InstrTypes.h>
+#include <llvm/Instructions.h>
+#include <llvm/Attributes.h>
+#include <llvm/Support/MemoryBuffer.h>
+#include <llvm/Bitcode/ReaderWriter.h>
+
+#include <sstream>
+#include <fstream>
+#include <iostream>
+
+using namespace llvm;
+
+#include "gallivm_builtins.cpp"
+
+#if 0
+llvm::Value *arrayFromChannels(std::vector<llvm::Value*> &vals)
+{
+   VectorType *vectorType = VectorType::get(Type::FloatTy, 4);
+   ArrayType *vectorArray = ArrayType::get(vectorType, 4);
+}
+#endif
+
+static inline std::string createFuncName(int label)
+{
+   std::ostringstream stream;
+   stream << "function";
+   stream << label;
+   return stream.str();
+}
+
+Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block,
+                           Storage *storage)
+   :  m_mod(mod), m_func(func), m_builder(block), m_idx(0),
+      m_storage(storage)
+{
+   m_floatVecType = VectorType::get(Type::FloatTy, 4);
+
+   m_llvmFSqrt = 0;
+   m_llvmFAbs  = 0;
+   m_llvmPow   = 0;
+   m_llvmFloor = 0;
+   m_llvmFlog  = 0;
+   m_llvmFexp  = 0;
+   m_llvmLit  = 0;
+   m_fmtPtr = 0;
+
+   MemoryBuffer *buffer = MemoryBuffer::getMemBuffer(
+      (const char*)&llvm_builtins_data[0],
+      (const char*)&llvm_builtins_data[Elements(llvm_builtins_data)-1]);
+   m_mod = ParseBitcodeFile(buffer);
+}
+
+llvm::BasicBlock * Instructions::currentBlock() const
+{
+   return m_builder.GetInsertBlock();
+}
+
+llvm::Value * Instructions::abs(llvm::Value *in)
+{
+   std::vector<llvm::Value*> vec = extractVector(in);
+   Value *xabs  = callFAbs(vec[0]);
+   Value *yabs  = callFAbs(vec[1]);
+   Value *zabs  = callFAbs(vec[2]);
+   Value *wabs  = callFAbs(vec[3]);
+   return vectorFromVals(xabs, yabs, zabs, wabs);
+}
+
+llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2)
+{
+   return m_builder.CreateAdd(in1, in2, name("add"));
+}
+
+llvm::Value * Instructions::arl(llvm::Value *in)
+{
+   return floor(in);
+}
+
+void Instructions::beginLoop()
+{
+   BasicBlock *begin = BasicBlock::Create(name("loop"), m_func,0);
+   BasicBlock *end = BasicBlock::Create(name("endloop"), m_func,0);
+
+   m_builder.CreateBr(begin);
+   Loop loop;
+   loop.begin = begin;
+   loop.end   = end;
+   m_builder.SetInsertPoint(begin);
+   m_loopStack.push(loop);
+}
+
+void Instructions::bgnSub(unsigned label)
+{
+   llvm::Function *func = findFunction(label);
+
+   Function::arg_iterator args = func->arg_begin();
+   Value *ptr_INPUT = args++;
+   ptr_INPUT->setName("INPUT");
+   m_storage->pushArguments(ptr_INPUT);
+
+   llvm::BasicBlock *entry = BasicBlock::Create("entry", func, 0);
+
+   m_func = func;
+   m_builder.SetInsertPoint(entry);
+}
+
+void Instructions::brk()
+{
+   assert(!m_loopStack.empty());
+   BasicBlock *unr = BasicBlock::Create(name("unreachable"), m_func,0);
+   m_builder.CreateBr(m_loopStack.top().end);
+   m_builder.SetInsertPoint(unr);
+}
+
+void Instructions::cal(int label, llvm::Value *input)
+{
+   std::vector<Value*> params;
+   params.push_back(input);
+   llvm::Function *func = findFunction(label);
+
+   m_builder.CreateCall(func, params.begin(), params.end());
+}
+
+llvm::Value * Instructions::ceil(llvm::Value *in)
+{
+   std::vector<llvm::Value*> vec = extractVector(in);
+   return vectorFromVals(callCeil(vec[0]), callCeil(vec[1]),
+                         callCeil(vec[2]), callCeil(vec[3]));
+}
+
+llvm::Value * Instructions::clamp(llvm::Value *in1)
+{
+   llvm::Value *zero = constVector(0.0f, 0.0f, 0.0f, 0.0f);
+   llvm::Value *one = constVector(1.0f, 1.0f, 1.0f, 1.0f);
+   return min( max(zero, in1), one);
+}
+
+llvm::Value * Instructions::cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3)
+{
+   llvm::Function *func = m_mod->getFunction("cmp");
+   assert(func);
+
+   std::vector<Value*> params;
+   params.push_back(in1);
+   params.push_back(in2);
+   params.push_back(in3);
+   CallInst *call = m_builder.CreateCall(func, params.begin(), params.end(), name("cmpres"));
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value * Instructions::cnd(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3)
+{
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+   std::vector<llvm::Value*> vec3 = extractVector(in3);
+   Constant *half = ConstantFP::get(APFloat(0.5f));
+
+   Value *xcmp  = m_builder.CreateFCmpOGT(vec1[0], half, name("xcmp"));
+   Value *selx = m_builder.CreateSelect(xcmp, vec2[0], vec3[0],
+                                        name("selx"));
+
+   Value *ycmp  = m_builder.CreateFCmpOGT(vec1[1], half, name("ycmp"));
+   Value *sely = m_builder.CreateSelect(ycmp, vec2[1], vec3[1],
+                                        name("sely"));
+
+   Value *zcmp  = m_builder.CreateFCmpOGT(vec1[2], half, name("zcmp"));
+   Value *selz = m_builder.CreateSelect(zcmp, vec2[2], vec3[2],
+                                        name("selz"));
+
+   Value *wcmp  = m_builder.CreateFCmpOGT(vec1[3], half, name("wcmp"));
+   Value *selw = m_builder.CreateSelect(wcmp, vec2[3], vec3[3],
+                                        name("selw"));
+
+   return vectorFromVals(selx, sely, selz, selw);
+}
+
+llvm::Value * Instructions::cnd0(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3)
+{
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+   std::vector<llvm::Value*> vec3 = extractVector(in3);
+   Constant *zero = Constant::getNullValue(Type::FloatTy);
+
+   Value *xcmp  = m_builder.CreateFCmpOGE(vec1[0], zero, name("xcmp"));
+   Value *selx = m_builder.CreateSelect(xcmp, vec2[0], vec3[0],
+                                        name("selx"));
+
+   Value *ycmp  = m_builder.CreateFCmpOGE(vec1[1], zero, name("ycmp"));
+   Value *sely = m_builder.CreateSelect(ycmp, vec2[1], vec3[1],
+                                        name("sely"));
+
+   Value *zcmp  = m_builder.CreateFCmpOGE(vec1[2], zero, name("zcmp"));
+   Value *selz = m_builder.CreateSelect(zcmp, vec2[2], vec3[2],
+                                        name("selz"));
+
+   Value *wcmp  = m_builder.CreateFCmpOGE(vec1[3], zero, name("wcmp"));
+   Value *selw = m_builder.CreateSelect(wcmp, vec2[3], vec3[3],
+                                        name("selw"));
+
+   return vectorFromVals(selx, sely, selz, selw);
+}
+
+llvm::Value * Instructions::cos(llvm::Value *in)
+{
+#if 0
+   llvm::Function *func = m_mod->getFunction("vcos");
+   assert(func);
+
+   CallInst *call = m_builder.CreateCall(func, in, name("cosres"));
+   call->setTailCall(false);
+   return call;
+#else
+   std::vector<llvm::Value*> elems = extractVector(in);
+   Function *func = m_mod->getFunction("cosf");
+   assert(func);
+   CallInst *cos = m_builder.CreateCall(func, elems[0], name("cosres"));
+   cos->setCallingConv(CallingConv::C);
+   cos->setTailCall(true);
+   return vectorFromVals(cos, cos, cos, cos);
+#endif
+}
+
+llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2)
+{
+   Value *x1 = m_builder.CreateExtractElement(in1,
+                                              m_storage->constantInt(0),
+                                              name("x1"));
+   Value *y1 = m_builder.CreateExtractElement(in1,
+                                              m_storage->constantInt(1),
+                                              name("y1"));
+   Value *z1 = m_builder.CreateExtractElement(in1,
+                                              m_storage->constantInt(2),
+                                              name("z1"));
+
+   Value *x2 = m_builder.CreateExtractElement(in2,
+                                              m_storage->constantInt(0),
+                                              name("x2"));
+   Value *y2 = m_builder.CreateExtractElement(in2,
+                                              m_storage->constantInt(1),
+                                              name("y2"));
+   Value *z2 = m_builder.CreateExtractElement(in2,
+                                              m_storage->constantInt(2),
+                                              name("z2"));
+   Value *y1z2 = mul(y1, z2);
+   Value *z1y2 = mul(z1, y2);
+
+   Value *z1x2 = mul(z1, x2);
+   Value *x1z2 = mul(x1, z2);
+
+   Value *x1y2 = mul(x1, y2);
+   Value *y1x2 = mul(y1, x2);
+
+   return vectorFromVals(sub(y1z2, z1y2), sub(z1x2, x1z2), sub(x1y2, y1x2));
+}
+
+llvm::Value * Instructions::ddx(llvm::Value *in)
+{
+   // FIXME
+   assert(0);
+}
+
+llvm::Value * Instructions::ddy(llvm::Value *in)
+{
+   // FIXME
+   assert(0);
+}
+
+llvm::Value * Instructions::div(llvm::Value *in1, llvm::Value *in2)
+{
+   return m_builder.CreateFDiv(in1, in2, name("div"));
+}
+
+llvm::Value * Instructions::dot2add(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3)
+{
+   Value *mulRes = mul(in1, in2);
+   Value *x = m_builder.CreateExtractElement(mulRes,
+                                                          m_storage->constantInt(0),
+                                                          name("extractx"));
+   Value *y = m_builder.CreateExtractElement(mulRes,
+                                                          m_storage->constantInt(1),
+                                                          name("extracty"));
+   Value *z = m_builder.CreateExtractElement(in3,
+                                                          m_storage->constantInt(2),
+                                                          name("extractz"));
+   Value *xy = m_builder.CreateAdd(x, y,name("xy"));
+   Value *dot2add = m_builder.CreateAdd(xy, z, name("dot2add"));
+   return vectorFromVals(dot2add, dot2add, dot2add, dot2add);
+}
+
+llvm::Value * Instructions::dp2(llvm::Value *in1, llvm::Value *in2)
+{
+   Value *mulRes = mul(in1, in2);
+   Value *x = m_builder.CreateExtractElement(mulRes,
+                                                          m_storage->constantInt(0),
+                                                          name("extractx"));
+   Value *y = m_builder.CreateExtractElement(mulRes,
+                                                          m_storage->constantInt(1),
+                                                          name("extracty"));
+   Value *xy = m_builder.CreateAdd(x, y,name("xy"));
+   return vectorFromVals(xy, xy, xy, xy);
+}
+
+llvm::Value * Instructions::dp3(llvm::Value *in1, llvm::Value *in2)
+{
+   Value *mulRes = mul(in1, in2);
+   Value *x = m_builder.CreateExtractElement(mulRes,
+                                                          m_storage->constantInt(0),
+                                                          name("extractx"));
+   Value *y = m_builder.CreateExtractElement(mulRes,
+                                                          m_storage->constantInt(1),
+                                                          name("extracty"));
+   Value *z = m_builder.CreateExtractElement(mulRes,
+                                                          m_storage->constantInt(2),
+                                                          name("extractz"));
+   Value *xy = m_builder.CreateAdd(x, y,name("xy"));
+   Value *dot3 = m_builder.CreateAdd(xy, z, name("dot3"));
+   return vectorFromVals(dot3, dot3, dot3, dot3);
+}
+
+llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2)
+{
+   Value *mulRes = mul(in1, in2);
+   std::vector<llvm::Value*> vec = extractVector(mulRes);
+   Value *xy = m_builder.CreateAdd(vec[0], vec[1], name("xy"));
+   Value *xyz = m_builder.CreateAdd(xy, vec[2], name("xyz"));
+   Value *dot4 = m_builder.CreateAdd(xyz, vec[3], name("dot4"));
+   return vectorFromVals(dot4, dot4, dot4, dot4);
+}
+
+llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2)
+{
+   Value *mulRes = mul(in1, in2);
+   std::vector<llvm::Value*> vec1 = extractVector(mulRes);
+   Value *xy = m_builder.CreateAdd(vec1[0], vec1[1], name("xy"));
+   Value *xyz = m_builder.CreateAdd(xy, vec1[2], name("xyz"));
+   Value *dph = m_builder.CreateAdd(xyz, vec1[3], name("dph"));
+   return vectorFromVals(dph, dph, dph, dph);
+}
+
+llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2)
+{
+   Value *y1 = m_builder.CreateExtractElement(in1,
+                                              m_storage->constantInt(1),
+                                              name("y1"));
+   Value *z = m_builder.CreateExtractElement(in1,
+                                             m_storage->constantInt(2),
+                                             name("z"));
+   Value *y2 = m_builder.CreateExtractElement(in2,
+                                              m_storage->constantInt(1),
+                                              name("y2"));
+   Value *w = m_builder.CreateExtractElement(in2,
+                                             m_storage->constantInt(3),
+                                             name("w"));
+   Value *ry = m_builder.CreateMul(y1, y2, name("tyuy"));
+   return vectorFromVals(ConstantFP::get(APFloat(1.f)),
+                         ry, z, w);
+}
+
+void Instructions::elseop()
+{
+   assert(!m_ifStack.empty());
+   BasicBlock *ifend = BasicBlock::Create(name("ifend"), m_func,0);
+   m_builder.CreateBr(ifend);
+   m_builder.SetInsertPoint(m_ifStack.top());
+   currentBlock()->setName(name("ifelse"));
+   m_ifStack.pop();
+   m_ifStack.push(ifend);
+}
+
+void Instructions::endif()
+{
+   assert(!m_ifStack.empty());
+   m_builder.CreateBr(m_ifStack.top());
+   m_builder.SetInsertPoint(m_ifStack.top());
+   m_ifStack.pop();
+}
+
+void Instructions::endLoop()
+{
+   assert(!m_loopStack.empty());
+   Loop loop = m_loopStack.top();
+   m_builder.CreateBr(loop.begin);
+   loop.end->moveAfter(currentBlock());
+   m_builder.SetInsertPoint(loop.end);
+   m_loopStack.pop();
+}
+
+void Instructions::end()
+{
+   m_builder.CreateRetVoid();
+}
+
+void Instructions::endSub()
+{
+   m_func = 0;
+   m_builder.SetInsertPoint(0);
+}
+
+llvm::Value * Instructions::exp(llvm::Value *in)
+{
+   std::vector<llvm::Value*> vec = extractVector(in);
+   return vectorFromVals(callFExp(vec[0]), callFExp(vec[1]),
+                             callFExp(vec[2]), callFExp(vec[3]));
+}
+
+llvm::Value * Instructions::ex2(llvm::Value *in)
+{
+   llvm::Value *val = callPow(ConstantFP::get(APFloat(2.f)),
+                              m_builder.CreateExtractElement(
+                                 in, m_storage->constantInt(0),
+                                 name("x1")));
+   return vectorFromVals(val, val, val, val);
+}
+
+llvm::Value * Instructions::floor(llvm::Value *in)
+{
+   std::vector<llvm::Value*> vec = extractVector(in);
+   return vectorFromVals(callFloor(vec[0]), callFloor(vec[1]),
+                         callFloor(vec[2]), callFloor(vec[3]));
+}
+
+llvm::Value * Instructions::frc(llvm::Value *in)
+{
+   llvm::Value *flr = floor(in);
+   return sub(in, flr);
+}
+
+void Instructions::ifop(llvm::Value *in)
+{
+   BasicBlock *ifthen = BasicBlock::Create(name("ifthen"), m_func,0);
+   BasicBlock *ifend = BasicBlock::Create(name("ifthenend"), m_func,0);
+
+   //BasicBlock *yblock = new BasicBlock(name("yblock"), m_func,0);
+   //BasicBlock *zblock = new BasicBlock(name("zblock"), m_func,0);
+   //BasicBlock *wblock = new BasicBlock(name("wblock"), m_func,0);
+
+   Constant *float0 = Constant::getNullValue(Type::FloatTy);
+
+   Value *x = m_builder.CreateExtractElement(in, m_storage->constantInt(0),
+                                             name("extractx"));
+   Value *xcmp = m_builder.CreateFCmpUNE(x, float0, name("xcmp"));
+   m_builder.CreateCondBr(xcmp, ifthen, ifend);
+   //m_builder.SetInsertPoint(yblock);
+
+   m_builder.SetInsertPoint(ifthen);
+   m_ifStack.push(ifend);
+}
+
+llvm::Value * Instructions::kil(llvm::Value *in)
+{
+   llvm::Function *func = m_mod->getFunction("kil");
+   assert(func);
+
+   CallInst *call = m_builder.CreateCall(func, in, name("kilpres"));
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2,
+                                 llvm::Value *in3)
+{
+   llvm::Value *m = mul(in1, in2);
+   llvm::Value *vec1 = constVector(1.f, 1.f, 1.f, 1.f);
+   llvm::Value *s = sub(vec1, in1);
+   return add(m, mul(s, in3));
+}
+
+llvm::Value * Instructions::lg2(llvm::Value *in)
+{
+   std::vector<llvm::Value*> vec = extractVector(in);
+   llvm::Value *const_vec = constVector(1.442695f, 1.442695f,
+                                        1.442695f, 1.442695f);
+   return mul(vectorFromVals(callFLog(vec[0]), callFLog(vec[1]),
+                             callFLog(vec[2]), callFLog(vec[3])), const_vec);
+}
+
+llvm::Value * Instructions::lit(llvm::Value *in)
+{
+   if (!m_llvmLit) {
+      m_llvmLit = m_mod->getFunction("lit");
+   }
+   CallInst *call = m_builder.CreateCall(m_llvmLit, in, name("litres"));
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value * Instructions::log(llvm::Value *in)
+{
+   std::vector<llvm::Value*> vec = extractVector(in);
+   return vectorFromVals(callFLog(vec[0]), callFLog(vec[1]),
+                             callFLog(vec[2]), callFLog(vec[3]));
+}
+
+llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2,
+                                 llvm::Value *in3)
+{
+   Value *mulRes = mul(in1, in2);
+   return add(mulRes, in3);
+}
+
+llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2)
+{
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+
+   Value *xcmp  = m_builder.CreateFCmpOGT(vec1[0], vec2[0],
+                                          name("xcmp"));
+   Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0],
+                                        name("selx"));
+
+   Value *ycmp  = m_builder.CreateFCmpOGT(vec1[1], vec2[1],
+                                          name("ycmp"));
+   Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1],
+                                        name("sely"));
+
+   Value *zcmp  = m_builder.CreateFCmpOGT(vec1[2], vec2[2],
+                                          name("zcmp"));
+   Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2],
+                                        name("selz"));
+
+   Value *wcmp  = m_builder.CreateFCmpOGT(vec1[3], vec2[3],
+                                          name("wcmp"));
+   Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3],
+                                        name("selw"));
+
+   return vectorFromVals(selx, sely, selz, selw);
+}
+
+llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2)
+{
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+
+   Value *xcmp  = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp"));
+   Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0],
+                                        name("selx"));
+
+   Value *ycmp  = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp"));
+   Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1],
+                                        name("sely"));
+
+   Value *zcmp  = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp"));
+   Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2],
+                                        name("selz"));
+
+   Value *wcmp  = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp"));
+   Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3],
+                                        name("selw"));
+
+   return vectorFromVals(selx, sely, selz, selw);
+}
+
+llvm::Value * Instructions::mul(llvm::Value *in1, llvm::Value *in2)
+{
+   return m_builder.CreateMul(in1, in2, name("mul"));
+}
+
+llvm::Value * Instructions::neg(llvm::Value *in)
+{
+   Value *neg = m_builder.CreateNeg(in, name("neg"));
+   return neg;
+}
+
+llvm::Value * Instructions::nrm(llvm::Value *in)
+{
+   llvm::Value *v = rsq(in);
+   return mul(v, in);
+}
+
+llvm::Value * Instructions::pow(llvm::Value *in1, llvm::Value *in2)
+{
+   Value *x1 = m_builder.CreateExtractElement(in1,
+                                              m_storage->constantInt(0),
+                                              name("x1"));
+   Value *x2 = m_builder.CreateExtractElement(in2,
+                                              m_storage->constantInt(0),
+                                              name("x2"));
+   llvm::Value *val = callPow(x1, x2);
+   return vectorFromVals(val, val, val, val);
+}
+
+llvm::Value * Instructions::rcp(llvm::Value *in1)
+{
+   Value *x1 = m_builder.CreateExtractElement(in1,
+                                              m_storage->constantInt(0),
+                                              name("x1"));
+   Value *res = m_builder.CreateFDiv(ConstantFP::get(APFloat(1.f)),
+                                     x1, name("rcp"));
+   return vectorFromVals(res, res, res, res);
+}
+
+llvm::Value * Instructions::rsq(llvm::Value *in1)
+{
+   Value *x = m_builder.CreateExtractElement(in1,
+                                             m_storage->constantInt(0),
+                                             name("extractx"));
+   Value *abs  = callFAbs(x);
+   Value *sqrt = callFSqrt(abs);
+
+   Value *rsqrt = m_builder.CreateFDiv(ConstantFP::get(APFloat(1.f)),
+                                       sqrt,
+                                       name("rsqrt"));
+   return vectorFromVals(rsqrt, rsqrt, rsqrt, rsqrt);
+}
+
+llvm::Value * Instructions::scs(llvm::Value *in)
+{
+   llvm::Function *func = m_mod->getFunction("scs");
+   assert(func);
+
+   CallInst *call = m_builder.CreateCall(func, in, name("scsres"));
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value * Instructions::seq(llvm::Value *in1, llvm::Value *in2)
+{
+   Constant *const1f = ConstantFP::get(APFloat(1.000000e+00f));
+   Constant *const0f = Constant::getNullValue(Type::FloatTy);
+
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+
+   Value *xcmp = m_builder.CreateFCmpOEQ(vec1[0], vec2[0], name("xcmp"));
+   Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel"));
+
+   Value *ycmp = m_builder.CreateFCmpOEQ(vec1[1], vec2[1], name("ycmp"));
+   Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel"));
+
+   Value *zcmp = m_builder.CreateFCmpOEQ(vec1[2], vec2[2], name("zcmp"));
+   Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel"));
+
+   Value *wcmp = m_builder.CreateFCmpOEQ(vec1[3], vec2[3], name("wcmp"));
+   Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel"));
+
+   return vectorFromVals(x, y, z, w);
+}
+
+llvm::Value * Instructions::sfl(llvm::Value *in1, llvm::Value *in2)
+{
+   Constant *const0f = Constant::getNullValue(Type::FloatTy);
+
+   return vectorFromVals(const0f, const0f, const0f, const0f);
+}
+
+llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2)
+{
+   Constant *const1f = ConstantFP::get(APFloat(1.000000e+00f));
+   Constant *const0f = Constant::getNullValue(Type::FloatTy);
+
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+
+   Value *xcmp = m_builder.CreateFCmpOGE(vec1[0], vec2[0], name("xcmp"));
+   Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel"));
+
+   Value *ycmp = m_builder.CreateFCmpOGE(vec1[1], vec2[1], name("ycmp"));
+   Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel"));
+
+   Value *zcmp = m_builder.CreateFCmpOGE(vec1[2], vec2[2], name("zcmp"));
+   Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel"));
+
+   Value *wcmp = m_builder.CreateFCmpOGE(vec1[3], vec2[3], name("wcmp"));
+   Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel"));
+
+   return vectorFromVals(x, y, z, w);
+}
+
+llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2)
+{
+   Constant *const1f = ConstantFP::get(APFloat(1.000000e+00f));
+   Constant *const0f = Constant::getNullValue(Type::FloatTy);
+
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+   Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], name("xcmp"));
+   Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel"));
+
+   Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], name("ycmp"));
+   Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel"));
+
+   Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], name("zcmp"));
+   Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel"));
+
+   Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], name("wcmp"));
+   Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel"));
+
+   return vectorFromVals(x, y, z, w);
+}
+
+llvm::Value * Instructions::sin(llvm::Value *in)
+{
+   llvm::Function *func = m_mod->getFunction("vsin");
+   assert(func);
+
+   CallInst *call = m_builder.CreateCall(func, in, name("sinres"));
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value * Instructions::sle(llvm::Value *in1, llvm::Value *in2)
+{
+   Constant *const1f = ConstantFP::get(APFloat(1.000000e+00f));
+   Constant *const0f = Constant::getNullValue(Type::FloatTy);
+
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+
+   Value *xcmp = m_builder.CreateFCmpOLE(vec1[0], vec2[0], name("xcmp"));
+   Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel"));
+
+   Value *ycmp = m_builder.CreateFCmpOLE(vec1[1], vec2[1], name("ycmp"));
+   Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel"));
+
+   Value *zcmp = m_builder.CreateFCmpOLE(vec1[2], vec2[2], name("zcmp"));
+   Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel"));
+
+   Value *wcmp = m_builder.CreateFCmpOLE(vec1[3], vec2[3], name("wcmp"));
+   Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel"));
+
+   return vectorFromVals(x, y, z, w);
+}
+
+llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2)
+{
+   Constant *const1f = ConstantFP::get(APFloat(1.000000e+00f));
+   Constant *const0f = Constant::getNullValue(Type::FloatTy);
+
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+
+   Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp"));
+   Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel"));
+
+   Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp"));
+   Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel"));
+
+   Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp"));
+   Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel"));
+
+   Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp"));
+   Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel"));
+
+   return vectorFromVals(x, y, z, w);
+}
+
+llvm::Value * Instructions::sne(llvm::Value *in1, llvm::Value *in2)
+{
+   Constant *const1f = ConstantFP::get(APFloat(1.000000e+00f));
+   Constant *const0f = Constant::getNullValue(Type::FloatTy);
+
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+
+   Value *xcmp = m_builder.CreateFCmpONE(vec1[0], vec2[0], name("xcmp"));
+   Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel"));
+
+   Value *ycmp = m_builder.CreateFCmpONE(vec1[1], vec2[1], name("ycmp"));
+   Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel"));
+
+   Value *zcmp = m_builder.CreateFCmpONE(vec1[2], vec2[2], name("zcmp"));
+   Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel"));
+
+   Value *wcmp = m_builder.CreateFCmpONE(vec1[3], vec2[3], name("wcmp"));
+   Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel"));
+
+   return vectorFromVals(x, y, z, w);
+}
+
+llvm::Value * Instructions::str(llvm::Value *in1, llvm::Value *in2)
+{
+   Constant *const1f = ConstantFP::get(APFloat(1.000000e+00f));
+
+   return vectorFromVals(const1f, const1f, const1f, const1f);
+}
+
+llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2)
+{
+   Value *res = m_builder.CreateSub(in1, in2, name("sub"));
+   return res;
+}
+
+llvm::Value * Instructions::trunc(llvm::Value *in)
+{
+   std::vector<llvm::Value*> vec = extractVector(in);
+   Value *icastx = m_builder.CreateFPToSI(vec[0], IntegerType::get(32),
+                                          name("ftoix"));
+   Value *icasty = m_builder.CreateFPToSI(vec[1], IntegerType::get(32),
+                                          name("ftoiy"));
+   Value *icastz = m_builder.CreateFPToSI(vec[2], IntegerType::get(32),
+                                          name("ftoiz"));
+   Value *icastw = m_builder.CreateFPToSI(vec[3], IntegerType::get(32),
+                                          name("ftoiw"));
+   Value *fx = m_builder.CreateSIToFP(icastx, Type::FloatTy,
+                                      name("fx"));
+   Value *fy = m_builder.CreateSIToFP(icasty, Type::FloatTy,
+                                      name("fy"));
+   Value *fz = m_builder.CreateSIToFP(icastz, Type::FloatTy,
+                                      name("fz"));
+   Value *fw = m_builder.CreateSIToFP(icastw, Type::FloatTy,
+                                      name("fw"));
+   return vectorFromVals(fx, fy, fz, fw);
+}
+
+llvm::Value * Instructions::x2d(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3)
+{
+   std::vector<llvm::Value*> vec1 = extractVector(in1);
+   std::vector<llvm::Value*> vec2 = extractVector(in2);
+   std::vector<llvm::Value*> vec3 = extractVector(in3);
+
+   Value *x2x3 = m_builder.CreateMul( vec2[0], vec3[0], name("x2x3"));
+   Value *y2y3 = m_builder.CreateMul( vec2[1], vec3[1], name("y2y3"));
+   Value *x1px2x3 = m_builder.CreateAdd (vec1[0], x2x3, name("x1 + x2x3"));
+   Value *x1px2x3py2y3 = m_builder.CreateAdd (x1px2x3, y2y3, name("x1 + x2x3 + y2y3"));
+
+   Value *x2z3 = m_builder.CreateMul( vec2[0], vec3[2], name("x2z3"));
+   Value *y2w3 = m_builder.CreateMul( vec2[1], vec3[3], name("y2w3"));
+   Value *y1px2z3 = m_builder.CreateAdd (vec1[1], x2z3, name("y1 + x2z3"));
+   Value *y1px2z3py2w3 = m_builder.CreateAdd (y1px2z3, y2w3, name("y1 + x2z3 + y2w3"));
+
+   return vectorFromVals(x1px2x3py2y3, y1px2z3py2w3, x1px2x3py2y3, y1px2z3py2w3);
+}
+
+void Instructions::printVector(llvm::Value *val)
+{
+   static const char *frmt = "Vector is [%f, %f, %f, %f]\x0A";
+
+   if (!m_fmtPtr) {
+      Constant *format = ConstantArray::get(frmt, true);
+      ArrayType *arrayTy = ArrayType::get(IntegerType::get(8), strlen(frmt) + 1);
+      GlobalVariable* globalFormat = new GlobalVariable(
+         /*Type=*/arrayTy,
+         /*isConstant=*/true,
+         /*Linkage=*/GlobalValue::InternalLinkage,
+         /*Initializer=*/0, // has initializer, specified below
+         /*Name=*/name(".str"),
+         m_mod);
+      globalFormat->setInitializer(format);
+
+      Constant* const_int0 = Constant::getNullValue(IntegerType::get(32));
+      std::vector<Constant*> const_ptr_21_indices;
+      const_ptr_21_indices.push_back(const_int0);
+      const_ptr_21_indices.push_back(const_int0);
+      m_fmtPtr = ConstantExpr::getGetElementPtr(globalFormat,
+                                                &const_ptr_21_indices[0], const_ptr_21_indices.size());
+   }
+
+   Function *func_printf = m_mod->getFunction("printf");
+   if (!func_printf)
+      func_printf = declarePrintf();
+   assert(func_printf);
+   std::vector<llvm::Value*> vec = extractVector(val);
+   Value *dx = m_builder.CreateFPExt(vec[0], Type::DoubleTy, name("dx"));
+   Value *dy = m_builder.CreateFPExt(vec[1], Type::DoubleTy, name("dy"));
+   Value *dz = m_builder.CreateFPExt(vec[2], Type::DoubleTy, name("dz"));
+   Value *dw = m_builder.CreateFPExt(vec[3], Type::DoubleTy, name("dw"));
+   std::vector<Value*> params;
+   params.push_back(m_fmtPtr);
+   params.push_back(dx);
+   params.push_back(dy);
+   params.push_back(dz);
+   params.push_back(dw);
+   CallInst *call = m_builder.CreateCall(func_printf, params.begin(), params.end(),
+                                         name("printf"));
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(true);
+}
+
+const char * Instructions::name(const char *prefix)
+{
+   ++m_idx;
+   snprintf(m_name, 32, "%s%d", prefix, m_idx);
+   return m_name;
+}
+
+llvm::Value * Instructions::callCeil(llvm::Value *val)
+{
+   if (!m_llvmCeil) {
+      // predeclare the intrinsic
+      std::vector<const Type*> ceilArgs;
+      ceilArgs.push_back(Type::FloatTy);
+      AttrListPtr ceilPal;
+      FunctionType* ceilType = FunctionType::get(
+         /*Result=*/Type::FloatTy,
+         /*Params=*/ceilArgs,
+         /*isVarArg=*/false);
+      m_llvmCeil = Function::Create(
+         /*Type=*/ceilType,
+         /*Linkage=*/GlobalValue::ExternalLinkage,
+         /*Name=*/"ceilf", m_mod);
+      m_llvmCeil->setCallingConv(CallingConv::C);
+      m_llvmCeil->setAttributes(ceilPal);
+   }
+   CallInst *call =  m_builder.CreateCall(m_llvmCeil, val,
+                                          name("ceilf"));
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value *Instructions::callFAbs(llvm::Value *val)
+{
+   if (!m_llvmFAbs) {
+      // predeclare the intrinsic
+      std::vector<const Type*> fabsArgs;
+      fabsArgs.push_back(Type::FloatTy);
+      AttrListPtr fabsPal;
+      FunctionType* fabsType = FunctionType::get(
+         /*Result=*/Type::FloatTy,
+         /*Params=*/fabsArgs,
+         /*isVarArg=*/false);
+      m_llvmFAbs = Function::Create(
+         /*Type=*/fabsType,
+         /*Linkage=*/GlobalValue::ExternalLinkage,
+         /*Name=*/"fabs", m_mod);
+      m_llvmFAbs->setCallingConv(CallingConv::C);
+      m_llvmFAbs->setAttributes(fabsPal);
+   }
+   CallInst *call = m_builder.CreateCall(m_llvmFAbs, val,
+                                         name("fabs"));
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value * Instructions::callFExp(llvm::Value *val)
+{
+   if (!m_llvmFexp) {
+      // predeclare the intrinsic
+      std::vector<const Type*> fexpArgs;
+      fexpArgs.push_back(Type::FloatTy);
+      AttrListPtr fexpPal;
+      FunctionType* fexpType = FunctionType::get(
+         /*Result=*/Type::FloatTy,
+         /*Params=*/fexpArgs,
+         /*isVarArg=*/false);
+      m_llvmFexp = Function::Create(
+         /*Type=*/fexpType,
+         /*Linkage=*/GlobalValue::ExternalLinkage,
+         /*Name=*/"expf", m_mod);
+      m_llvmFexp->setCallingConv(CallingConv::C);
+      m_llvmFexp->setAttributes(fexpPal);
+   }
+   CallInst *call = m_builder.CreateCall(m_llvmFexp, val,
+                                         name("expf"));
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value * Instructions::callFLog(llvm::Value *val)
+{
+   if (!m_llvmFlog) {
+      // predeclare the intrinsic
+      std::vector<const Type*> flogArgs;
+      flogArgs.push_back(Type::FloatTy);
+      AttrListPtr flogPal;
+      FunctionType* flogType = FunctionType::get(
+         /*Result=*/Type::FloatTy,
+         /*Params=*/flogArgs,
+         /*isVarArg=*/false);
+      m_llvmFlog = Function::Create(
+         /*Type=*/flogType,
+         /*Linkage=*/GlobalValue::ExternalLinkage,
+         /*Name=*/"logf", m_mod);
+      m_llvmFlog->setCallingConv(CallingConv::C);
+      m_llvmFlog->setAttributes(flogPal);
+   }
+   CallInst *call = m_builder.CreateCall(m_llvmFlog, val,
+                                         name("logf"));
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value * Instructions::callFloor(llvm::Value *val)
+{
+   if (!m_llvmFloor) {
+      // predeclare the intrinsic
+      std::vector<const Type*> floorArgs;
+      floorArgs.push_back(Type::FloatTy);
+      AttrListPtr floorPal;
+      FunctionType* floorType = FunctionType::get(
+         /*Result=*/Type::FloatTy,
+         /*Params=*/floorArgs,
+         /*isVarArg=*/false);
+      m_llvmFloor = Function::Create(
+         /*Type=*/floorType,
+         /*Linkage=*/GlobalValue::ExternalLinkage,
+         /*Name=*/"floorf", m_mod);
+      m_llvmFloor->setCallingConv(CallingConv::C);
+      m_llvmFloor->setAttributes(floorPal);
+   }
+   CallInst *call =  m_builder.CreateCall(m_llvmFloor, val,
+                                          name("floorf"));
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value *Instructions::callFSqrt(llvm::Value *val)
+{
+   if (!m_llvmFSqrt) {
+      // predeclare the intrinsic
+      std::vector<const Type*> fsqrtArgs;
+      fsqrtArgs.push_back(Type::FloatTy);
+      AttrListPtr fsqrtPal;
+      FunctionType* fsqrtType = FunctionType::get(
+         /*Result=*/Type::FloatTy,
+         /*Params=*/fsqrtArgs,
+         /*isVarArg=*/false);
+      m_llvmFSqrt = Function::Create(
+         /*Type=*/fsqrtType,
+         /*Linkage=*/GlobalValue::ExternalLinkage,
+         /*Name=*/"llvm.sqrt.f32", m_mod);
+      m_llvmFSqrt->setCallingConv(CallingConv::C);
+      m_llvmFSqrt->setAttributes(fsqrtPal);
+   }
+   CallInst *call = m_builder.CreateCall(m_llvmFSqrt, val,
+                                         name("sqrt"));
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2)
+{
+   if (!m_llvmPow) {
+      // predeclare the intrinsic
+      std::vector<const Type*> powArgs;
+      powArgs.push_back(Type::FloatTy);
+      powArgs.push_back(Type::FloatTy);
+      AttrListPtr powPal;
+      FunctionType* powType = FunctionType::get(
+         /*Result=*/Type::FloatTy,
+         /*Params=*/powArgs,
+         /*isVarArg=*/false);
+      m_llvmPow = Function::Create(
+         /*Type=*/powType,
+         /*Linkage=*/GlobalValue::ExternalLinkage,
+         /*Name=*/"llvm.pow.f32", m_mod);
+      m_llvmPow->setCallingConv(CallingConv::C);
+      m_llvmPow->setAttributes(powPal);
+   }
+   std::vector<Value*> params;
+   params.push_back(val1);
+   params.push_back(val2);
+   CallInst *call = m_builder.CreateCall(m_llvmPow, params.begin(), params.end(),
+                                         name("pow"));
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+   return call;
+}
+
+llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y,
+                                           llvm::Value *z, llvm::Value *w)
+{
+   Constant *const_vec = Constant::getNullValue(m_floatVecType);
+   Value *res = m_builder.CreateInsertElement(const_vec, x,
+                                              m_storage->constantInt(0),
+                                              name("vecx"));
+   res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1),
+                               name("vecxy"));
+   res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2),
+                               name("vecxyz"));
+   if (w)
+      res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3),
+                                          name("vecxyzw"));
+   return res;
+}
+
+llvm::Value * Instructions::constVector(float x, float y, float z, float w)
+{
+   std::vector<Constant*> vec(4);
+   vec[0] = ConstantFP::get(APFloat(x));
+   vec[1] = ConstantFP::get(APFloat(y));
+   vec[2] = ConstantFP::get(APFloat(z));
+   vec[3] = ConstantFP::get(APFloat(w));
+   return ConstantVector::get(m_floatVecType, vec);
+}
+
+llvm::Function * Instructions::declarePrintf()
+{
+   std::vector<const Type*> args;
+   AttrListPtr params;
+   FunctionType* funcTy = FunctionType::get(
+      /*Result=*/IntegerType::get(32),
+      /*Params=*/args,
+      /*isVarArg=*/true);
+   Function* func_printf = Function::Create(
+      /*Type=*/funcTy,
+      /*Linkage=*/GlobalValue::ExternalLinkage,
+      /*Name=*/"printf", m_mod);
+   func_printf->setCallingConv(CallingConv::C);
+   func_printf->setAttributes(params);
+   return func_printf;
+}
+
+llvm::Function * Instructions::declareFunc(int label)
+{
+   PointerType *vecPtr = PointerType::getUnqual(m_floatVecType);
+   std::vector<const Type*> args;
+   args.push_back(vecPtr);
+   args.push_back(vecPtr);
+   args.push_back(vecPtr);
+   args.push_back(vecPtr);
+   AttrListPtr params;
+   FunctionType *funcType = FunctionType::get(
+      /*Result=*/Type::VoidTy,
+      /*Params=*/args,
+      /*isVarArg=*/false);
+   std::string name = createFuncName(label);
+   Function *func = Function::Create(
+      /*Type=*/funcType,
+      /*Linkage=*/GlobalValue::ExternalLinkage,
+      /*Name=*/name.c_str(), m_mod);
+   func->setCallingConv(CallingConv::C);
+   func->setAttributes(params);
+   return func;
+}
+
+llvm::Function * Instructions::findFunction(int label)
+{
+   llvm::Function *func = m_functions[label];
+   if (!func) {
+      func = declareFunc(label);
+      m_functions[label] = func;
+   }
+   return func;
+}
+
+std::vector<llvm::Value*> Instructions::extractVector(llvm::Value *vec)
+{
+   std::vector<llvm::Value*> elems(4);
+   elems[0] = m_builder.CreateExtractElement(vec, m_storage->constantInt(0),
+                                             name("x"));
+   elems[1] = m_builder.CreateExtractElement(vec, m_storage->constantInt(1),
+                                             name("y"));
+   elems[2] = m_builder.CreateExtractElement(vec, m_storage->constantInt(2),
+                                             name("z"));
+   elems[3] = m_builder.CreateExtractElement(vec, m_storage->constantInt(3),
+                                             name("w"));
+   return elems;
+}
+
+
+#endif //MESA_LLVM
+
+
diff --git a/src/gallium/auxiliary/gallivm/instructions.h b/src/gallium/auxiliary/gallivm/instructions.h
new file mode 100644
index 0000000000..e18571251e
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/instructions.h
@@ -0,0 +1,175 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin zack@tungstengraphics.com
+  */
+
+#ifndef INSTRUCTIONS_H
+#define INSTRUCTIONS_H
+
+#include <llvm/BasicBlock.h>
+#include <llvm/Module.h>
+#include <llvm/Value.h>
+#include <llvm/Support/IRBuilder.h>
+
+#include <map>
+#include <stack>
+
+namespace llvm {
+   class VectorType;
+   class Function;
+}
+
+class Storage;
+
+class Instructions
+{
+public:
+   Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block,
+                Storage *storage);
+
+   llvm::BasicBlock *currentBlock() const;
+
+   llvm::Value *abs(llvm::Value *in1);
+   llvm::Value *add(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *arl(llvm::Value *in1);
+   void         beginLoop();
+   void         bgnSub(unsigned);
+   void         brk();
+   void         cal(int label, llvm::Value *input);
+   llvm::Value *ceil(llvm::Value *in);
+   llvm::Value *clamp(llvm::Value *in);
+   llvm::Value *cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3);
+   llvm::Value *cnd(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3);
+   llvm::Value *cnd0(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3);
+   llvm::Value *cos(llvm::Value *in);
+   llvm::Value *cross(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *ddx(llvm::Value *in);
+   llvm::Value *ddy(llvm::Value *in);
+   llvm::Value *div(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *dot2add(llvm::Value *in, llvm::Value *in2, llvm::Value *in3);
+   llvm::Value *dp2(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *dph(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *dst(llvm::Value *in1, llvm::Value *in2);
+   void         elseop();
+   void         endif();
+   void         endLoop();
+   void         end();
+   void         endSub();
+   llvm::Value *exp(llvm::Value *in);
+   llvm::Value *ex2(llvm::Value *in);
+   llvm::Value *floor(llvm::Value *in);
+   llvm::Value *frc(llvm::Value *in);
+   void         ifop(llvm::Value *in);
+   llvm::Value *kil(llvm::Value *in);
+   llvm::Value *lerp(llvm::Value *in1, llvm::Value *in2,
+                     llvm::Value *in3);
+   llvm::Value *lg2(llvm::Value *in);
+   llvm::Value *lit(llvm::Value *in);
+   llvm::Value *log(llvm::Value *in);
+   llvm::Value *madd(llvm::Value *in1, llvm::Value *in2,
+                     llvm::Value *in3);
+   llvm::Value *max(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *min(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *mul(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *neg(llvm::Value *in);
+   llvm::Value *nrm(llvm::Value *in);
+   llvm::Value *pow(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *rcp(llvm::Value *in);
+   llvm::Value *rsq(llvm::Value *in);
+   llvm::Value *scs(llvm::Value *in);
+   llvm::Value *seq(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *sfl(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *sge(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *sin(llvm::Value *in);
+   llvm::Value *sle(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *slt(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *sne(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *str(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *sub(llvm::Value *in1, llvm::Value *in2);
+   llvm::Value *trunc(llvm::Value *in);
+   llvm::Value *x2d(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3);
+
+   void printVector(llvm::Value *val);
+private:
+   const char *name(const char *prefix);
+
+   llvm::Value *callCeil(llvm::Value *val);
+   llvm::Value *callFAbs(llvm::Value *val);
+   llvm::Value *callFExp(llvm::Value *val);
+   llvm::Value *callFLog(llvm::Value *val);
+   llvm::Value *callFloor(llvm::Value *val);
+   llvm::Value *callFSqrt(llvm::Value *val);
+   llvm::Value *callPow(llvm::Value *val1, llvm::Value *val2);
+
+   llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y,
+                               llvm::Value *z, llvm::Value *w=0);
+
+   llvm::Value *constVector(float x, float y, float z, float w);
+
+   llvm::Function *declarePrintf();
+   llvm::Function *declareFunc(int label);
+
+   llvm::Function *findFunction(int label);
+
+   std::vector<llvm::Value*> extractVector(llvm::Value *vec);
+private:
+   llvm::Module             *m_mod;
+   llvm::Function           *m_func;
+   char                      m_name[32];
+   llvm::IRBuilder<>         m_builder;
+   int                       m_idx;
+
+   llvm::VectorType *m_floatVecType;
+
+   llvm::Function   *m_llvmCeil;
+   llvm::Function   *m_llvmFSqrt;
+   llvm::Function   *m_llvmFAbs;
+   llvm::Function   *m_llvmPow;
+   llvm::Function   *m_llvmFloor;
+   llvm::Function   *m_llvmFlog;
+   llvm::Function   *m_llvmFexp;
+   llvm::Function   *m_llvmLit;
+
+   llvm::Constant   *m_fmtPtr;
+
+   std::stack<llvm::BasicBlock*> m_ifStack;
+   struct Loop {
+      llvm::BasicBlock *begin;
+      llvm::BasicBlock *end;
+   };
+   std::stack<Loop> m_loopStack;
+   std::map<int, llvm::Function*> m_functions;
+   Storage *m_storage;
+};
+
+#endif
diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp
new file mode 100644
index 0000000000..925e948763
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp
@@ -0,0 +1,523 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+#include "instructionssoa.h"
+
+#include "storagesoa.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "util/u_memory.h"
+
+#include <llvm/CallingConv.h>
+#include <llvm/Constants.h>
+#include <llvm/Module.h>
+#include <llvm/Function.h>
+#include <llvm/Instructions.h>
+#include <llvm/Transforms/Utils/Cloning.h>
+#include <llvm/Attributes.h>
+#include <llvm/Support/MemoryBuffer.h>
+#include <llvm/Bitcode/ReaderWriter.h>
+
+
+#include <iostream>
+
+
+/* disable some warnings. this file is autogenerated */
+#if defined(__GNUC__)
+#pragma GCC diagnostic ignored "-Wunused-variable"
+#endif
+using namespace llvm;
+#include "gallivmsoabuiltins.cpp"
+#if defined(__GNUC__)
+#pragma GCC diagnostic warning "-Wunused-variable"
+#endif
+
+InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func,
+                                 llvm::BasicBlock *block, StorageSoa *storage)
+   : m_builder(block),
+     m_storage(storage),
+     m_idx(0)
+{
+   createFunctionMap();
+   createBuiltins();
+}
+
+const char * InstructionsSoa::name(const char *prefix) const
+{
+   ++m_idx;
+   snprintf(m_name, 32, "%s%d", prefix, m_idx);
+   return m_name;
+}
+
+llvm::Value * InstructionsSoa::vectorFromVals(llvm::Value *x, llvm::Value *y,
+                                              llvm::Value *z, llvm::Value *w)
+{
+   VectorType  *vectorType = VectorType::get(Type::FloatTy, 4);
+   Constant *constVector = Constant::getNullValue(vectorType);
+   Value *res = m_builder.CreateInsertElement(constVector, x,
+                                              m_storage->constantInt(0),
+                                              name("vecx"));
+   res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1),
+                               name("vecxy"));
+   res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2),
+                               name("vecxyz"));
+   if (w)
+      res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3),
+                                          name("vecxyzw"));
+   return res;
+}
+
+void InstructionsSoa::end()
+{
+   m_builder.CreateRetVoid();
+}
+
+std::vector<llvm::Value*> InstructionsSoa::extractVector(llvm::Value *vector)
+{
+   std::vector<llvm::Value*> res(4);
+   res[0] = m_builder.CreateExtractElement(vector,
+                                           m_storage->constantInt(0),
+                                           name("extract1X"));
+   res[1] = m_builder.CreateExtractElement(vector,
+                                           m_storage->constantInt(1),
+                                           name("extract2X"));
+   res[2] = m_builder.CreateExtractElement(vector,
+                                           m_storage->constantInt(2),
+                                           name("extract3X"));
+   res[3] = m_builder.CreateExtractElement(vector,
+                                           m_storage->constantInt(3),
+                                           name("extract4X"));
+
+   return res;
+}
+
+llvm::IRBuilder<>* InstructionsSoa::getIRBuilder()
+{
+   return &m_builder;
+}
+
+void InstructionsSoa::createFunctionMap()
+{
+   m_functionsMap[TGSI_OPCODE_ABS]   = "abs";
+   m_functionsMap[TGSI_OPCODE_DP3]   = "dp3";
+   m_functionsMap[TGSI_OPCODE_DP4]   = "dp4";
+   m_functionsMap[TGSI_OPCODE_MIN]   = "min";
+   m_functionsMap[TGSI_OPCODE_MAX]   = "max";
+   m_functionsMap[TGSI_OPCODE_POWER] = "pow";
+   m_functionsMap[TGSI_OPCODE_LIT]   = "lit";
+   m_functionsMap[TGSI_OPCODE_RSQ]   = "rsq";
+   m_functionsMap[TGSI_OPCODE_SLT]   = "slt";
+}
+
+void InstructionsSoa::createDependencies()
+{
+   {
+      std::vector<std::string> powDeps(2);
+      powDeps[0] = "powf";
+      powDeps[1] = "powvec";
+      m_builtinDependencies["pow"] = powDeps;
+   }
+   {
+      std::vector<std::string> absDeps(2);
+      absDeps[0] = "fabsf";
+      absDeps[1] = "absvec";
+      m_builtinDependencies["abs"] = absDeps;
+   }
+   {
+      std::vector<std::string> maxDeps(1);
+      maxDeps[0] = "maxvec";
+      m_builtinDependencies["max"] = maxDeps;
+   }
+   {
+      std::vector<std::string> minDeps(1);
+      minDeps[0] = "minvec";
+      m_builtinDependencies["min"] = minDeps;
+   }
+   {
+      std::vector<std::string> litDeps(4);
+      litDeps[0] = "minvec";
+      litDeps[1] = "maxvec";
+      litDeps[2] = "powf";
+      litDeps[3] = "powvec";
+      m_builtinDependencies["lit"] = litDeps;
+   }
+   {
+      std::vector<std::string> rsqDeps(4);
+      rsqDeps[0] = "sqrtf";
+      rsqDeps[1] = "sqrtvec";
+      rsqDeps[2] = "fabsf";
+      rsqDeps[3] = "absvec";
+      m_builtinDependencies["rsq"] = rsqDeps;
+   }
+}
+
+llvm::Function * InstructionsSoa::function(int op)
+{
+    if (m_functions.find(op) != m_functions.end())
+       return m_functions[op];
+
+    std::string name = m_functionsMap[op];
+
+    std::cout <<"For op = "<<op<<", func is '"<<name<<"'"<<std::endl;
+
+    std::vector<std::string> deps = m_builtinDependencies[name];
+    for (unsigned int i = 0; i < deps.size(); ++i) {
+       llvm::Function *func = m_builtins->getFunction(deps[i]);
+       std::cout <<"\tinjecting dep = '"<<func->getName()<<"'"<<std::endl;
+       injectFunction(func);
+    }
+
+    llvm::Function *originalFunc = m_builtins->getFunction(name);
+    injectFunction(originalFunc, op);
+    return m_functions[op];
+}
+
+llvm::Module * InstructionsSoa::currentModule() const
+{
+   BasicBlock *block = m_builder.GetInsertBlock();
+   if (!block || !block->getParent())
+      return 0;
+
+   return block->getParent()->getParent();
+}
+
+void InstructionsSoa::createBuiltins()
+{
+   std::string ErrMsg;
+   MemoryBuffer *buffer = MemoryBuffer::getMemBuffer(
+      (const char*)&soabuiltins_data[0],
+      (const char*)&soabuiltins_data[Elements(soabuiltins_data) - 1]);
+   m_builtins = ParseBitcodeFile(buffer, &ErrMsg);
+   std::cout<<"Builtins created at "<<m_builtins<<" ("<<ErrMsg<<")"<<std::endl;
+   assert(m_builtins);
+   createDependencies();
+}
+
+
+std::vector<llvm::Value*> InstructionsSoa::abs(const std::vector<llvm::Value*> in1)
+{
+   llvm::Function *func = function(TGSI_OPCODE_ABS);
+   return callBuiltin(func, in1);
+}
+
+std::vector<llvm::Value*> InstructionsSoa::add(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = m_builder.CreateAdd(in1[0], in2[0], name("addx"));
+   res[1] = m_builder.CreateAdd(in1[1], in2[1], name("addy"));
+   res[2] = m_builder.CreateAdd(in1[2], in2[2], name("addz"));
+   res[3] = m_builder.CreateAdd(in1[3], in2[3], name("addw"));
+
+   return res;
+}
+
+std::vector<llvm::Value*> InstructionsSoa::arl(const std::vector<llvm::Value*> in)
+{
+   std::vector<llvm::Value*> res(4);
+
+   //Extract x's
+   llvm::Value *x1 = m_builder.CreateExtractElement(in[0],
+                                                    m_storage->constantInt(0),
+                                                    name("extractX"));
+   //cast it to an unsigned int
+   x1 = m_builder.CreateFPToUI(x1, IntegerType::get(32), name("x1IntCast"));
+
+   res[0] = x1;//vectorFromVals(x1, x2, x3, x4);
+   //only x is valid. the others shouldn't be necessary
+   /*
+   res[1] = Constant::getNullValue(m_floatVecType);
+   res[2] = Constant::getNullValue(m_floatVecType);
+   res[3] = Constant::getNullValue(m_floatVecType);
+   */
+
+   return res;
+}
+
+std::vector<llvm::Value*> InstructionsSoa::dp3(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   llvm::Function *func = function(TGSI_OPCODE_DP3);
+   return callBuiltin(func, in1, in2);
+}
+
+std::vector<llvm::Value*> InstructionsSoa::lit(const std::vector<llvm::Value*> in)
+{
+   llvm::Function *func = function(TGSI_OPCODE_LIT);
+   return callBuiltin(func, in);
+}
+
+std::vector<llvm::Value*> InstructionsSoa::madd(const std::vector<llvm::Value*> in1,
+                                                const std::vector<llvm::Value*> in2,
+                                                const std::vector<llvm::Value*> in3)
+{
+   std::vector<llvm::Value*> res = mul(in1, in2);
+   return add(res, in3);
+}
+
+std::vector<llvm::Value*> InstructionsSoa::max(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   llvm::Function *func = function(TGSI_OPCODE_MAX);
+   return callBuiltin(func, in1, in2);
+}
+
+std::vector<llvm::Value*> InstructionsSoa::min(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   llvm::Function *func = function(TGSI_OPCODE_MIN);
+   return callBuiltin(func, in1, in2);
+}
+
+std::vector<llvm::Value*> InstructionsSoa::mul(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = m_builder.CreateMul(in1[0], in2[0], name("mulx"));
+   res[1] = m_builder.CreateMul(in1[1], in2[1], name("muly"));
+   res[2] = m_builder.CreateMul(in1[2], in2[2], name("mulz"));
+   res[3] = m_builder.CreateMul(in1[3], in2[3], name("mulw"));
+
+   return res;
+}
+
+std::vector<llvm::Value*> InstructionsSoa::pow(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   llvm::Function *func = function(TGSI_OPCODE_POWER);
+   return callBuiltin(func, in1, in2);
+}
+
+std::vector<llvm::Value*> InstructionsSoa::rsq(const std::vector<llvm::Value*> in)
+{
+   llvm::Function *func = function(TGSI_OPCODE_RSQ);
+   return callBuiltin(func, in);
+}
+
+std::vector<llvm::Value*> InstructionsSoa::slt(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   llvm::Function *func = function(TGSI_OPCODE_SLT);
+   return callBuiltin(func, in1, in2);
+}
+
+std::vector<llvm::Value*> InstructionsSoa::sub(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = m_builder.CreateSub(in1[0], in2[0], name("subx"));
+   res[1] = m_builder.CreateSub(in1[1], in2[1], name("suby"));
+   res[2] = m_builder.CreateSub(in1[2], in2[2], name("subz"));
+   res[3] = m_builder.CreateSub(in1[3], in2[3], name("subw"));
+
+   return res;
+}
+
+void checkFunction(Function *func)
+{
+   for (Function::const_iterator BI = func->begin(), BE = func->end();
+        BI != BE; ++BI) {
+      const BasicBlock &BB = *BI;
+      for (BasicBlock::const_iterator II = BB.begin(), IE = BB.end();
+           II != IE; ++II) {
+         const Instruction &I = *II;
+         std::cout<< "Instr = "<<I;
+         for (unsigned op = 0, E = I.getNumOperands(); op != E; ++op) {
+            const Value *Op = I.getOperand(op);
+            std::cout<< "\top = "<<Op<<"("<<op<<")"<<std::endl;
+            //I->setOperand(op, V);
+  }
+      }
+   }
+}
+
+llvm::Value * InstructionsSoa::allocaTemp()
+{
+   VectorType *vector   = VectorType::get(Type::FloatTy, 4);
+   ArrayType  *vecArray = ArrayType::get(vector, 4);
+   AllocaInst *alloca = new AllocaInst(vecArray, name("tmpRes"),
+                                       m_builder.GetInsertBlock());
+
+   std::vector<Value*> indices;
+   indices.push_back(m_storage->constantInt(0));
+   indices.push_back(m_storage->constantInt(0));
+   GetElementPtrInst *getElem = GetElementPtrInst::Create(alloca,
+                                                          indices.begin(),
+                                                          indices.end(),
+                                                          name("allocaPtr"),
+                                                          m_builder.GetInsertBlock());
+   return getElem;
+}
+
+std::vector<llvm::Value*> InstructionsSoa::allocaToResult(llvm::Value *allocaPtr)
+{
+   GetElementPtrInst *xElemPtr =  GetElementPtrInst::Create(allocaPtr,
+                                                            m_storage->constantInt(0),
+                                                            name("xPtr"),
+                                                            m_builder.GetInsertBlock());
+   GetElementPtrInst *yElemPtr =  GetElementPtrInst::Create(allocaPtr,
+                                                            m_storage->constantInt(1),
+                                                            name("yPtr"),
+                                                            m_builder.GetInsertBlock());
+   GetElementPtrInst *zElemPtr =  GetElementPtrInst::Create(allocaPtr,
+                                                            m_storage->constantInt(2),
+                                                            name("zPtr"),
+                                                            m_builder.GetInsertBlock());
+   GetElementPtrInst *wElemPtr =  GetElementPtrInst::Create(allocaPtr,
+                                                            m_storage->constantInt(3),
+                                                            name("wPtr"),
+                                                            m_builder.GetInsertBlock());
+
+   std::vector<llvm::Value*> res(4);
+   res[0] = new LoadInst(xElemPtr, name("xRes"), false, m_builder.GetInsertBlock());
+   res[1] = new LoadInst(yElemPtr, name("yRes"), false, m_builder.GetInsertBlock());
+   res[2] = new LoadInst(zElemPtr, name("zRes"), false, m_builder.GetInsertBlock());
+   res[3] = new LoadInst(wElemPtr, name("wRes"), false, m_builder.GetInsertBlock());
+
+   return res;
+}
+
+std::vector<llvm::Value*> InstructionsSoa::dp4(const std::vector<llvm::Value*> in1,
+                                               const std::vector<llvm::Value*> in2)
+{
+   llvm::Function *func = function(TGSI_OPCODE_DP4);
+   return callBuiltin(func, in1, in2);
+}
+
+std::vector<Value*> InstructionsSoa::callBuiltin(llvm::Function *func, const std::vector<llvm::Value*> in1)
+{
+   std::vector<Value*> params;
+
+   llvm::Value *allocaPtr = allocaTemp();
+   params.push_back(allocaPtr);
+   params.push_back(in1[0]);
+   params.push_back(in1[1]);
+   params.push_back(in1[2]);
+   params.push_back(in1[3]);
+   CallInst *call = m_builder.CreateCall(func, params.begin(), params.end());
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+
+   return allocaToResult(allocaPtr);
+}
+
+std::vector<Value*> InstructionsSoa::callBuiltin(llvm::Function *func, const std::vector<llvm::Value*> in1,
+                                                 const std::vector<llvm::Value*> in2)
+{
+   std::vector<Value*> params;
+
+   llvm::Value *allocaPtr = allocaTemp();
+   params.push_back(allocaPtr);
+   params.push_back(in1[0]);
+   params.push_back(in1[1]);
+   params.push_back(in1[2]);
+   params.push_back(in1[3]);
+   params.push_back(in2[0]);
+   params.push_back(in2[1]);
+   params.push_back(in2[2]);
+   params.push_back(in2[3]);
+   CallInst *call = m_builder.CreateCall(func, params.begin(), params.end());
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+
+   return allocaToResult(allocaPtr);
+}
+
+std::vector<Value*> InstructionsSoa::callBuiltin(llvm::Function *func, const std::vector<llvm::Value*> in1,
+                                                 const std::vector<llvm::Value*> in2,
+                                                 const std::vector<llvm::Value*> in3)
+{
+   std::vector<Value*> params;
+
+   llvm::Value *allocaPtr = allocaTemp();
+   params.push_back(allocaPtr);
+   params.push_back(in1[0]);
+   params.push_back(in1[1]);
+   params.push_back(in1[2]);
+   params.push_back(in1[3]);
+   params.push_back(in2[0]);
+   params.push_back(in2[1]);
+   params.push_back(in2[2]);
+   params.push_back(in2[3]);
+   params.push_back(in3[0]);
+   params.push_back(in3[1]);
+   params.push_back(in3[2]);
+   params.push_back(in3[3]);
+   CallInst *call = m_builder.CreateCall(func, params.begin(), params.end());
+   call->setCallingConv(CallingConv::C);
+   call->setTailCall(false);
+
+   return allocaToResult(allocaPtr);
+}
+
+void InstructionsSoa::injectFunction(llvm::Function *originalFunc, int op)
+{
+   assert(originalFunc);
+   std::cout << "injecting function originalFunc " <<originalFunc->getName() <<std::endl;
+   if (op != TGSI_OPCODE_LAST) {
+      /* in this case it's possible the function has been already
+       * injected as part of the dependency chain, which gets
+       * injected below */
+      llvm::Function *func = currentModule()->getFunction(originalFunc->getName());
+      if (func) {
+         m_functions[op] = func;
+         return;
+      }
+   }
+   llvm::Function *func = 0;
+   if (originalFunc->isDeclaration()) {
+      func = Function::Create(originalFunc->getFunctionType(), GlobalValue::ExternalLinkage,
+                              originalFunc->getName(), currentModule());
+      func->setCallingConv(CallingConv::C);
+      const AttrListPtr pal;
+      func->setAttributes(pal);
+      currentModule()->dump();
+   } else {
+      DenseMap<const Value*, Value *> val;
+      val[m_builtins->getFunction("fabsf")] = currentModule()->getFunction("fabsf");
+      val[m_builtins->getFunction("powf")] = currentModule()->getFunction("powf");
+      val[m_builtins->getFunction("sqrtf")] = currentModule()->getFunction("sqrtf");
+      func = CloneFunction(originalFunc, val);
+#if 0
+      std::cout <<" replacing "<<m_builtins->getFunction("powf")
+                <<", with " <<currentModule()->getFunction("powf")<<std::endl;
+      std::cout<<"1111-------------------------------"<<std::endl;
+      checkFunction(originalFunc);
+      std::cout<<"2222-------------------------------"<<std::endl;
+      checkFunction(func);
+      std::cout <<"XXXX = " <<val[m_builtins->getFunction("powf")]<<std::endl;
+#endif
+      currentModule()->getFunctionList().push_back(func);
+   }
+   if (op != TGSI_OPCODE_LAST) {
+      m_functions[op] = func;
+   }
+}
+
+
diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.h b/src/gallium/auxiliary/gallivm/instructionssoa.h
new file mode 100644
index 0000000000..d6831e0a6b
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/instructionssoa.h
@@ -0,0 +1,116 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef INSTRUCTIONSSOA_H
+#define INSTRUCTIONSSOA_H
+
+#include <pipe/p_shader_tokens.h>
+#include <llvm/Support/IRBuilder.h>
+
+#include <map>
+#include <vector>
+
+namespace llvm {
+   class Module;
+   class Function;
+   class BasicBlock;
+   class Value;
+}
+class StorageSoa;
+
+class InstructionsSoa
+{
+public:
+   InstructionsSoa(llvm::Module *mod, llvm::Function *func,
+                   llvm::BasicBlock *block, StorageSoa *storage);
+
+   std::vector<llvm::Value*> abs(const std::vector<llvm::Value*> in1);
+   std::vector<llvm::Value*> arl(const std::vector<llvm::Value*> in);
+   std::vector<llvm::Value*> add(const std::vector<llvm::Value*> in1,
+                                 const std::vector<llvm::Value*> in2);
+   std::vector<llvm::Value*> dp3(const std::vector<llvm::Value*> in1,
+                                 const std::vector<llvm::Value*> in2);
+   std::vector<llvm::Value*> dp4(const std::vector<llvm::Value*> in1,
+                                 const std::vector<llvm::Value*> in2);
+   std::vector<llvm::Value*> lit(const std::vector<llvm::Value*> in);
+   std::vector<llvm::Value*> madd(const std::vector<llvm::Value*> in1,
+                                  const std::vector<llvm::Value*> in2,
+                                  const std::vector<llvm::Value*> in3);
+   std::vector<llvm::Value*> max(const std::vector<llvm::Value*> in1,
+                                 const std::vector<llvm::Value*> in2);
+   std::vector<llvm::Value*> min(const std::vector<llvm::Value*> in1,
+                                 const std::vector<llvm::Value*> in2);
+   std::vector<llvm::Value*> mul(const std::vector<llvm::Value*> in1,
+                                 const std::vector<llvm::Value*> in2);
+   std::vector<llvm::Value*> pow(const std::vector<llvm::Value*> in1,
+                                 const std::vector<llvm::Value*> in2);
+   std::vector<llvm::Value*> rsq(const std::vector<llvm::Value*> in1);
+   std::vector<llvm::Value*> slt(const std::vector<llvm::Value*> in1,
+                                 const std::vector<llvm::Value*> in2);
+   std::vector<llvm::Value*> sub(const std::vector<llvm::Value*> in1,
+                                 const std::vector<llvm::Value*> in2);
+   void         end();
+
+   std::vector<llvm::Value*> extractVector(llvm::Value *vector);
+   llvm::IRBuilder<>*  getIRBuilder();
+private:
+   const char * name(const char *prefix) const;
+   llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y,
+                               llvm::Value *z, llvm::Value *w);
+   void createFunctionMap();
+   void createBuiltins();
+   void createDependencies();
+   llvm::Function *function(int);
+   llvm::Module *currentModule() const;
+   llvm::Value *allocaTemp();
+   std::vector<llvm::Value*> allocaToResult(llvm::Value *allocaPtr);
+   std::vector<llvm::Value*> callBuiltin(llvm::Function *func,
+                                         const std::vector<llvm::Value*> in1);
+   std::vector<llvm::Value*> callBuiltin(llvm::Function *func,
+                                         const std::vector<llvm::Value*> in1,
+                                         const std::vector<llvm::Value*> in2);
+   std::vector<llvm::Value*> callBuiltin(llvm::Function *func,
+                                         const std::vector<llvm::Value*> in1,
+                                         const std::vector<llvm::Value*> in2,
+                                         const std::vector<llvm::Value*> in3);
+   void injectFunction(llvm::Function *originalFunc, int op = TGSI_OPCODE_LAST);
+private:
+   llvm::IRBuilder<>  m_builder;
+   StorageSoa *m_storage;
+
+   std::map<int, std::string> m_functionsMap;
+   std::map<int, llvm::Function*> m_functions;
+   llvm::Module *m_builtins;
+   std::map<std::string, std::vector<std::string> > m_builtinDependencies;
+
+private:
+   mutable int  m_idx;
+   mutable char m_name[32];
+};
+
+
+#endif
diff --git a/src/gallium/auxiliary/gallivm/llvm_builtins.c b/src/gallium/auxiliary/gallivm/llvm_builtins.c
new file mode 100644
index 0000000000..d5a003a48b
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/llvm_builtins.c
@@ -0,0 +1,114 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin zack@tungstengraphics.com
+  */
+typedef __attribute__(( ext_vector_type(4) )) float float4;
+
+extern float powf(float a, float b);
+
+inline float approx(float a, float b)
+{
+    if (b < -128.0f) b = -128.0f;
+    if (b > 128.0f)   b = 128.0f;
+    if (a < 0) a = 0;
+    return powf(a, b);
+}
+
+inline float4 lit(float4 tmp)
+{
+    float4 result;
+    result.x = 1.0;
+    result.w = 1.0;
+    if (tmp.x > 0) {
+        result.y = tmp.x;
+        result.z = approx(tmp.y, tmp.w);
+    } else {
+        result.y = 0;
+        result.z = 0;
+    }
+    return result;
+}
+
+inline float4 cmp(float4 tmp0, float4 tmp1, float4 tmp2)
+{
+   float4 result;
+
+   result.x = (tmp0.x < 0.0) ? tmp1.x : tmp2.x;
+   result.y = (tmp0.y < 0.0) ? tmp1.y : tmp2.y;
+   result.z = (tmp0.z < 0.0) ? tmp1.z : tmp2.z;
+   result.w = (tmp0.w < 0.0) ? tmp1.w : tmp2.w;
+
+   return result;
+}
+
+extern float cosf(float  val);
+extern float sinf(float  val);
+
+inline float4 vcos(float4 val)
+{
+   float4 result;
+   printf("VEC IN   is %f %f %f %f\n", val.x, val.y, val.z, val.w);
+   result.x = cosf(val.x);
+   result.y = cosf(val.x);
+   result.z = cosf(val.x);
+   result.w = cosf(val.x);
+   printf("VEC OUT  is %f %f %f %f\n", result.x, result.y, result.z, result.w);
+   return result;
+}
+
+inline float4 scs(float4 val)
+{
+   float4 result;
+   float tmp = val.x;
+   result.x = cosf(tmp);
+   result.y = sinf(tmp);
+   return result;
+}
+
+
+inline float4 vsin(float4 val)
+{
+   float4 result;
+   float tmp = val.x;
+   float res = sinf(tmp);
+   result.x = res;
+   result.y = res;
+   result.z = res;
+   result.w = res;
+   return result;
+}
+
+inline int kil(float4 val)
+{
+   if (val.x < 0 || val.y < 0 || val.z < 0 || val.w < 0)
+      return 1;
+   else
+      return 0;
+}
diff --git a/src/gallium/auxiliary/gallivm/loweringpass.cpp b/src/gallium/auxiliary/gallivm/loweringpass.cpp
new file mode 100644
index 0000000000..556dbec366
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/loweringpass.cpp
@@ -0,0 +1,17 @@
+#include "loweringpass.h"
+
+using namespace llvm;
+
+char LoweringPass::ID = 0;
+RegisterPass<LoweringPass> X("lowering", "Lowering Pass");
+
+LoweringPass::LoweringPass()
+   :  ModulePass((intptr_t)&ID)
+{
+}
+
+bool LoweringPass::runOnModule(Module &m)
+{
+   llvm::cerr << "Hello: " << m.getModuleIdentifier() << "\n";
+   return false;
+}
diff --git a/src/gallium/auxiliary/gallivm/loweringpass.h b/src/gallium/auxiliary/gallivm/loweringpass.h
new file mode 100644
index 0000000000..f62dcf6ba7
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/loweringpass.h
@@ -0,0 +1,15 @@
+#ifndef LOWERINGPASS_H
+#define LOWERINGPASS_H
+
+#include "llvm/Pass.h"
+#include "llvm/Module.h"
+
+struct LoweringPass : public llvm::ModulePass
+{
+   static char ID;
+   LoweringPass();
+
+   virtual bool runOnModule(llvm::Module &m);
+};
+
+#endif
diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c
new file mode 100644
index 0000000000..cb85e1734e
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/soabuiltins.c
@@ -0,0 +1,210 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * This file is compiled with clang into the LLVM bitcode
+  *
+  * Authors:
+  *   Zack Rusin zack@tungstengraphics.com
+  */
+typedef __attribute__(( ext_vector_type(4) )) float float4;
+
+
+extern float fabsf(float val);
+
+/* helpers */
+
+float4 absvec(float4 vec)
+{
+   float4 res;
+   res.x = fabsf(vec.x);
+   res.y = fabsf(vec.y);
+   res.z = fabsf(vec.z);
+   res.w = fabsf(vec.w);
+
+   return res;
+}
+
+float4 maxvec(float4 a, float4 b)
+{
+   return (float4){(a.x > b.x) ? a.x : b.x,
+         (a.y > b.y) ? a.y : b.y,
+         (a.z > b.z) ? a.z : b.z,
+         (a.w > b.w) ? a.w : b.w};
+}
+
+float4 minvec(float4 a, float4 b)
+{
+   return (float4){(a.x < b.x) ? a.x : b.x,
+         (a.y < b.y) ? a.y : b.y,
+         (a.z < b.z) ? a.z : b.z,
+         (a.w < b.w) ? a.w : b.w};
+}
+
+extern float powf(float num, float p);
+extern float sqrtf(float x);
+
+float4 powvec(float4 vec, float4 q)
+{
+   float4 p;
+   p.x = powf(vec.x, q.x);
+   p.y = powf(vec.y, q.y);
+   p.z = powf(vec.z, q.z);
+   p.w = powf(vec.w, q.w);
+   return p;
+}
+
+float4 sqrtvec(float4 vec)
+{
+   float4 p;
+   p.x = sqrtf(vec.x);
+   p.y = sqrtf(vec.y);
+   p.z = sqrtf(vec.z);
+   p.w = sqrtf(vec.w);
+   return p;
+}
+
+float4 sltvec(float4 v1, float4 v2)
+{
+   float4 p;
+   p.x = (v1.x < v2.x) ? 1.0 : 0.0;
+   p.y = (v1.y < v2.y) ? 1.0 : 0.0;
+   p.z = (v1.z < v2.z) ? 1.0 : 0.0;
+   p.w = (v1.w < v2.w) ? 1.0 : 0.0;
+   return p;
+}
+
+
+/* instructions */
+
+void abs(float4 *res,
+         float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w)
+{
+   res[0] = absvec(tmp0x);
+   res[1] = absvec(tmp0y);
+   res[2] = absvec(tmp0z);
+   res[3] = absvec(tmp0w);
+}
+
+void dp3(float4 *res,
+         float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w,
+         float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w)
+{
+   float4 dot = (tmp0x * tmp1x) + (tmp0y * tmp1y) +
+                (tmp0z * tmp1z);
+
+   res[0] = dot;
+   res[1] = dot;
+   res[2] = dot;
+   res[3] = dot;
+}
+
+void dp4(float4 *res,
+         float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w,
+         float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w)
+{
+   float4 dot = (tmp0x * tmp1x) + (tmp0y * tmp1y) +
+                (tmp0z * tmp1z) + (tmp0w * tmp1w);
+
+   res[0] = dot;
+   res[1] = dot;
+   res[2] = dot;
+   res[3] = dot;
+}
+
+void lit(float4 *res,
+         float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w)
+{
+   const float4 zerovec = (float4) {0.0, 0.0, 0.0, 0.0};
+   const float4 min128 = (float4) {-128.f, -128.f, -128.f, -128.f};
+   const float4 plus128 = (float4) {128.f,  128.f,  128.f,  128.f};
+
+   res[0] = (float4){1.0, 1.0, 1.0, 1.0};
+   if (tmp0x.x > 0) {
+      float4 tmpy = maxvec(tmp0y, zerovec);
+      float4 tmpw = minvec(tmp0w, plus128);
+      tmpw = maxvec(tmpw, min128);
+      res[1] = tmp0x;
+      res[2] = powvec(tmpy, tmpw);
+   } else {
+      res[1] = zerovec;
+      res[2] = zerovec;
+   }
+   res[3] = (float4){1.0, 1.0, 1.0, 1.0};
+}
+
+void min(float4 *res,
+         float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w,
+         float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w)
+{
+   res[0] = minvec(tmp0x, tmp1x);
+   res[1] = minvec(tmp0y, tmp1y);
+   res[2] = minvec(tmp0z, tmp1z);
+   res[3] = minvec(tmp0w, tmp1w);
+}
+
+
+void max(float4 *res,
+         float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w,
+         float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w)
+{
+   res[0] = maxvec(tmp0x, tmp1x);
+   res[1] = maxvec(tmp0y, tmp1y);
+   res[2] = maxvec(tmp0z, tmp1z);
+   res[3] = maxvec(tmp0w, tmp1w);
+}
+
+void pow(float4 *res,
+         float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w,
+         float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w)
+{
+   res[0] = powvec(tmp0x, tmp1x);
+   res[1] = res[0];
+   res[2] = res[0];
+   res[3] = res[0];
+}
+
+void rsq(float4 *res,
+         float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w)
+{
+   const float4 onevec = (float4) {1., 1., 1., 1.};
+   res[0] = onevec/sqrtvec(absvec(tmp0x));
+   res[1] = onevec/sqrtvec(absvec(tmp0y));
+   res[2] = onevec/sqrtvec(absvec(tmp0z));
+   res[3] = onevec/sqrtvec(absvec(tmp0w));
+}
+
+void slt(float4 *res,
+         float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w,
+         float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w)
+{
+   res[0] = sltvec(tmp0x, tmp1x);
+   res[1] = sltvec(tmp0y, tmp1y);
+   res[2] = sltvec(tmp0z, tmp1z);
+   res[3] = sltvec(tmp0w, tmp1w);
+}
+
diff --git a/src/gallium/auxiliary/gallivm/storage.cpp b/src/gallium/auxiliary/gallivm/storage.cpp
new file mode 100644
index 0000000000..73df24c976
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/storage.cpp
@@ -0,0 +1,364 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin zack@tungstengraphics.com
+  */
+#ifdef MESA_LLVM
+
+#include "storage.h"
+
+#include "gallivm_p.h"
+
+#include "pipe/p_shader_tokens.h"
+#include <llvm/BasicBlock.h>
+#include <llvm/Module.h>
+#include <llvm/Value.h>
+
+#include <llvm/CallingConv.h>
+#include <llvm/Constants.h>
+#include <llvm/DerivedTypes.h>
+#include <llvm/InstrTypes.h>
+#include <llvm/Instructions.h>
+
+using namespace llvm;
+
+Storage::Storage(llvm::BasicBlock *block, llvm::Value *input)
+   : m_block(block),
+     m_INPUT(input),
+     m_addrs(32),
+     m_idx(0)
+{
+   m_floatVecType = VectorType::get(Type::FloatTy, 4);
+   m_intVecType   = VectorType::get(IntegerType::get(32), 4);
+
+   m_undefFloatVec = UndefValue::get(m_floatVecType);
+   m_undefIntVec   = UndefValue::get(m_intVecType);
+   m_extSwizzleVec = 0;
+
+   m_numConsts = 0;
+}
+
+//can only build vectors with all members in the [0, 9] range
+llvm::Constant *Storage::shuffleMask(int vec)
+{
+   if (!m_extSwizzleVec) {
+      std::vector<Constant*> elems;
+      elems.push_back(ConstantFP::get(APFloat(0.f)));
+      elems.push_back(ConstantFP::get(APFloat(1.f)));
+      elems.push_back(ConstantFP::get(APFloat(0.f)));
+      elems.push_back(ConstantFP::get(APFloat(1.f)));
+      m_extSwizzleVec = ConstantVector::get(m_floatVecType, elems);
+   }
+
+   if (m_intVecs.find(vec) != m_intVecs.end()) {
+      return m_intVecs[vec];
+   }
+   int origVec = vec;
+   Constant* const_vec = 0;
+   if (origVec == 0) {
+      const_vec = Constant::getNullValue(m_intVecType);
+   } else {
+      int x = gallivm_x_swizzle(vec);
+      int y = gallivm_y_swizzle(vec);
+      int z = gallivm_z_swizzle(vec);
+      int w = gallivm_w_swizzle(vec);
+      std::vector<Constant*> elems;
+      elems.push_back(constantInt(x));
+      elems.push_back(constantInt(y));
+      elems.push_back(constantInt(z));
+      elems.push_back(constantInt(w));
+      const_vec = ConstantVector::get(m_intVecType, elems);
+   }
+
+   m_intVecs[origVec] = const_vec;
+   return const_vec;
+}
+
+llvm::ConstantInt *Storage::constantInt(int idx)
+{
+   if (m_constInts.find(idx) != m_constInts.end()) {
+      return m_constInts[idx];
+   }
+   ConstantInt *const_int = ConstantInt::get(APInt(32,  idx));
+   m_constInts[idx] = const_int;
+   return const_int;
+}
+
+llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx)
+{
+   Value *val = element(InputsArg, idx, indIdx);
+   LoadInst *load = new LoadInst(val, name("input"), false, m_block);
+   load->setAlignment(8);
+
+   return load;
+}
+
+llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx)
+{
+   m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts;
+
+   Value *elem = element(ConstsArg, idx, indIdx);
+   LoadInst *load = new LoadInst(elem, name("const"), false, m_block);
+   load->setAlignment(8);
+   return load;
+}
+
+llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle)
+{
+   Constant *mask = shuffleMask(shuffle);
+   ShuffleVectorInst *res =
+      new ShuffleVectorInst(vec, m_extSwizzleVec, mask,
+                            name("shuffle"), m_block);
+   return res;
+}
+
+
+llvm::Value *Storage::tempElement(int idx, llvm::Value *indIdx)
+{
+   Value *elem = element(TempsArg, idx, indIdx);
+
+   LoadInst *load = new LoadInst(elem, name("temp"), false, m_block);
+   load->setAlignment(8);
+
+   return load;
+}
+
+void Storage::setTempElement(int idx, llvm::Value *val, int mask)
+{
+   if (mask != TGSI_WRITEMASK_XYZW) {
+      llvm::Value *templ = 0;
+      if (m_tempWriteMap[idx])
+         templ = tempElement(idx);
+      val = maskWrite(val, mask, templ);
+   }
+   Value *elem = element(TempsArg, idx);
+   StoreInst *st = new StoreInst(val, elem, false, m_block);
+   st->setAlignment(8);
+   m_tempWriteMap[idx] = true;
+}
+
+void Storage::setOutputElement(int dstIdx, llvm::Value *val, int mask)
+{
+   if (mask != TGSI_WRITEMASK_XYZW) {
+      llvm::Value *templ = 0;
+      if (m_destWriteMap[dstIdx])
+         templ = outputElement(dstIdx);
+      val = maskWrite(val, mask, templ);
+   }
+
+   Value *elem = element(DestsArg, dstIdx);
+   StoreInst *st = new StoreInst(val, elem, false, m_block);
+   st->setAlignment(8);
+   m_destWriteMap[dstIdx] = true;
+}
+
+llvm::Value *Storage::maskWrite(llvm::Value *src, int mask, llvm::Value *templ)
+{
+   llvm::Value *dst = templ;
+   if (!dst)
+      dst = Constant::getNullValue(m_floatVecType);
+   if ((mask & TGSI_WRITEMASK_X)) {
+      llvm::Value *x = new ExtractElementInst(src, unsigned(0),
+                                              name("x"), m_block);
+      dst = InsertElementInst::Create(dst, x, unsigned(0),
+                                      name("dstx"), m_block);
+   }
+   if ((mask & TGSI_WRITEMASK_Y)) {
+      llvm::Value *y = new ExtractElementInst(src, unsigned(1),
+                                              name("y"), m_block);
+      dst = InsertElementInst::Create(dst, y, unsigned(1),
+                                      name("dsty"), m_block);
+   }
+   if ((mask & TGSI_WRITEMASK_Z)) {
+      llvm::Value *z = new ExtractElementInst(src, unsigned(2),
+                                              name("z"), m_block);
+      dst = InsertElementInst::Create(dst, z, unsigned(2),
+                                      name("dstz"), m_block);
+   }
+   if ((mask & TGSI_WRITEMASK_W)) {
+      llvm::Value *w = new ExtractElementInst(src, unsigned(3),
+                                              name("w"), m_block);
+      dst = InsertElementInst::Create(dst, w, unsigned(3),
+                                      name("dstw"), m_block);
+   }
+   return dst;
+}
+
+const char * Storage::name(const char *prefix)
+{
+   ++m_idx;
+   snprintf(m_name, 32, "%s%d", prefix, m_idx);
+   return m_name;
+}
+
+int Storage::numConsts() const
+{
+   return m_numConsts;
+}
+
+llvm::Value * Storage::addrElement(int idx) const
+{
+   Value *ret = m_addrs[idx];
+   if (!ret)
+      return m_undefFloatVec;
+   return ret;
+}
+
+void Storage::setAddrElement(int idx, llvm::Value *val, int mask)
+{
+   if (mask != TGSI_WRITEMASK_XYZW) {
+      llvm::Value *templ = m_addrs[idx];
+      val = maskWrite(val, mask, templ);
+   }
+   m_addrs[idx] = val;
+}
+
+llvm::Value * Storage::extractIndex(llvm::Value *vec)
+{
+   llvm::Value *x = new ExtractElementInst(vec, unsigned(0),
+                                           name("x"), m_block);
+   return new FPToSIInst(x, IntegerType::get(32), name("intidx"), m_block);
+}
+
+void Storage::setCurrentBlock(llvm::BasicBlock *block)
+{
+   m_block = block;
+}
+
+llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx)
+{
+   Value *elem = element(DestsArg, idx, indIdx);
+   LoadInst *load = new LoadInst(elem, name("output"), false, m_block);
+   load->setAlignment(8);
+
+   return load;
+}
+
+llvm::Value * Storage::inputPtr() const
+{
+   return m_INPUT;
+}
+
+void Storage::pushArguments(llvm::Value *input)
+{
+   m_argStack.push(m_INPUT);
+
+   m_INPUT = input;
+}
+
+void Storage::popArguments()
+{
+   m_INPUT = m_argStack.top();
+   m_argStack.pop();
+}
+
+void Storage::pushTemps()
+{
+   m_extSwizzleVec = 0;
+}
+
+void Storage::popTemps()
+{
+}
+
+llvm::Value * Storage::immediateElement(int idx)
+{
+   return m_immediates[idx];
+}
+
+void Storage::addImmediate(float *val)
+{
+   std::vector<Constant*> vec(4);
+   vec[0] = ConstantFP::get(APFloat(val[0]));
+   vec[1] = ConstantFP::get(APFloat(val[1]));
+   vec[2] = ConstantFP::get(APFloat(val[2]));
+   vec[3] = ConstantFP::get(APFloat(val[3]));
+   m_immediates.push_back(ConstantVector::get(m_floatVecType, vec));
+}
+
+
+llvm::Value * Storage::elemPtr(Args arg)
+{
+   std::vector<Value*> indices;
+   indices.push_back(constantInt(0));
+   indices.push_back(constantInt(static_cast<int>(arg)));
+   GetElementPtrInst *getElem = GetElementPtrInst::Create(m_INPUT,
+                                                          indices.begin(),
+                                                          indices.end(),
+                                                          name("input_ptr"),
+                                                          m_block);
+   return new LoadInst(getElem, name("input_field"), false, m_block);
+}
+
+llvm::Value * Storage::elemIdx(llvm::Value *ptr, int idx,
+                               llvm::Value *indIdx )
+{
+   GetElementPtrInst *getElem = 0;
+
+   if (indIdx) {
+      getElem = GetElementPtrInst::Create(ptr,
+                                      BinaryOperator::Create(Instruction::Add,
+                                                             indIdx,
+                                                             constantInt(idx),
+                                                             name("add"),
+                                                             m_block),
+                                      name("field"),
+                                      m_block);
+   } else {
+      getElem = GetElementPtrInst::Create(ptr,
+                                      constantInt(idx),
+                                      name("field"),
+                                      m_block);
+   }
+   return getElem;
+}
+
+llvm::Value * Storage::element(Args arg, int idx, llvm::Value *indIdx )
+{
+   Value *val = elemPtr(arg);
+   return elemIdx(val, idx, indIdx);
+}
+
+void Storage::setKilElement(llvm::Value *val)
+{
+   std::vector<Value*> indices;
+   indices.push_back(constantInt(0));
+   indices.push_back(constantInt(static_cast<int>(KilArg)));
+   GetElementPtrInst *elem = GetElementPtrInst::Create(m_INPUT,
+                                                   indices.begin(),
+                                                   indices.end(),
+                                                   name("kil_ptr"),
+                                                   m_block);
+   StoreInst *st = new StoreInst(val, elem, false, m_block);
+   st->setAlignment(8);
+}
+
+#endif //MESA_LLVM
+
+
diff --git a/src/gallium/auxiliary/gallivm/storage.h b/src/gallium/auxiliary/gallivm/storage.h
new file mode 100644
index 0000000000..8574f7554e
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/storage.h
@@ -0,0 +1,133 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin zack@tungstengraphics.com
+  */
+
+#ifndef STORAGE_H
+#define STORAGE_H
+
+#include <map>
+#include <set>
+#include <stack>
+#include <vector>
+
+namespace llvm {
+   class BasicBlock;
+   class Constant;
+   class ConstantInt;
+   class LoadInst;
+   class Value;
+   class VectorType;
+}
+
+class Storage
+{
+public:
+   Storage(llvm::BasicBlock *block,
+           llvm::Value *input);
+
+   llvm::Value *inputPtr() const;
+
+   void setCurrentBlock(llvm::BasicBlock *block);
+
+   llvm::ConstantInt *constantInt(int);
+   llvm::Constant *shuffleMask(int vec);
+   llvm::Value *inputElement(int idx, llvm::Value *indIdx =0);
+   llvm::Value *constElement(int idx, llvm::Value *indIdx =0);
+   llvm::Value *outputElement(int idx, llvm::Value *indIdx =0);
+   llvm::Value *tempElement(int idx, llvm::Value *indIdx =0);
+   llvm::Value *immediateElement(int idx);
+
+   void setOutputElement(int dstIdx, llvm::Value *val, int mask);
+   void setTempElement(int idx, llvm::Value *val, int mask);
+
+   llvm::Value *addrElement(int idx) const;
+   void setAddrElement(int idx, llvm::Value *val, int mask);
+
+   void setKilElement(llvm::Value *val);
+
+   llvm::Value *shuffleVector(llvm::Value *vec, int shuffle);
+
+   llvm::Value *extractIndex(llvm::Value *vec);
+
+   int numConsts() const;
+
+   void pushArguments(llvm::Value *input);
+   void popArguments();
+   void pushTemps();
+   void popTemps();
+
+   void addImmediate(float *val);
+
+private:
+   llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ);
+   const char *name(const char *prefix);
+
+   enum Args {
+      DestsArg   = 0,
+      InputsArg  = 1,
+      TempsArg   = 2,
+      ConstsArg  = 3,
+      KilArg     = 4
+   };
+   llvm::Value *elemPtr(Args arg);
+   llvm::Value *elemIdx(llvm::Value *ptr, int idx,
+                        llvm::Value *indIdx = 0);
+   llvm::Value *element(Args arg, int idx, llvm::Value *indIdx = 0);
+
+private:
+   llvm::BasicBlock *m_block;
+   llvm::Value *m_INPUT;
+
+   std::map<int, llvm::ConstantInt*> m_constInts;
+   std::map<int, llvm::Constant*>    m_intVecs;
+   std::vector<llvm::Value*>         m_addrs;
+   std::vector<llvm::Constant*>      m_immediates;
+
+   llvm::VectorType *m_floatVecType;
+   llvm::VectorType *m_intVecType;
+
+   char        m_name[32];
+   int         m_idx;
+
+   int         m_numConsts;
+
+   std::map<int, bool > m_destWriteMap;
+   std::map<int, bool > m_tempWriteMap;
+
+   llvm::Value      *m_undefFloatVec;
+   llvm::Value      *m_undefIntVec;
+   llvm::Value      *m_extSwizzleVec;
+
+   std::stack<llvm::Value*> m_argStack;
+   std::stack<std::vector<llvm::Value*> > m_tempStack;
+};
+
+#endif
diff --git a/src/gallium/auxiliary/gallivm/storagesoa.cpp b/src/gallium/auxiliary/gallivm/storagesoa.cpp
new file mode 100644
index 0000000000..e1e5cabcf5
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/storagesoa.cpp
@@ -0,0 +1,438 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "storagesoa.h"
+
+#include "gallivm_p.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_debug.h"
+
+#include <llvm/BasicBlock.h>
+#include <llvm/Module.h>
+#include <llvm/Value.h>
+
+#include <llvm/CallingConv.h>
+#include <llvm/Constants.h>
+#include <llvm/DerivedTypes.h>
+#include <llvm/InstrTypes.h>
+#include <llvm/Instructions.h>
+
+using namespace llvm;
+
+
+StorageSoa::StorageSoa(llvm::BasicBlock *block,
+                       llvm::Value *input,
+                       llvm::Value *output,
+                       llvm::Value *consts)
+   : m_block(block),
+     m_input(input),
+     m_output(output),
+     m_consts(consts),
+     m_immediates(0),
+     m_idx(0)
+{
+}
+
+void StorageSoa::addImmediate(float *vec)
+{
+   std::vector<float> vals(4);
+   vals[0] = vec[0];
+   vals[1] = vec[1];
+   vals[2] = vec[2];
+   vals[3] = vec[3];
+   m_immediatesToFlush.push_back(vals);
+}
+
+void StorageSoa::declareImmediates()
+{
+   if (m_immediatesToFlush.empty())
+      return;
+
+   VectorType *vectorType = VectorType::get(Type::FloatTy, 4);
+   ArrayType  *vectorChannels = ArrayType::get(vectorType, 4);
+   ArrayType  *arrayType = ArrayType::get(vectorChannels, m_immediatesToFlush.size());
+
+   m_immediates = new GlobalVariable(
+      /*Type=*/arrayType,
+      /*isConstant=*/false,
+      /*Linkage=*/GlobalValue::ExternalLinkage,
+      /*Initializer=*/0, // has initializer, specified below
+      /*Name=*/name("immediates"),
+      currentModule());
+
+   std::vector<Constant*> arrayVals;
+   for (unsigned int i = 0; i < m_immediatesToFlush.size(); ++i) {
+      std::vector<float> vec = m_immediatesToFlush[i];
+      std::vector<float> vals(4);
+      std::vector<Constant*> channelArray;
+
+      vals[0] = vec[0]; vals[1] = vec[1]; vals[2] = vec[2]; vals[3] = vec[3];
+      llvm::Constant *xChannel = createConstGlobalVector(vals);
+
+      vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1];
+      llvm::Constant *yChannel = createConstGlobalVector(vals);
+
+      vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2];
+      llvm::Constant *zChannel = createConstGlobalVector(vals);
+
+      vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3];
+      llvm::Constant *wChannel = createConstGlobalVector(vals);
+      channelArray.push_back(xChannel);
+      channelArray.push_back(yChannel);
+      channelArray.push_back(zChannel);
+      channelArray.push_back(wChannel);
+      Constant *constChannels = ConstantArray::get(vectorChannels,
+                                                   channelArray);
+      arrayVals.push_back(constChannels);
+   }
+   Constant *constArray = ConstantArray::get(arrayType, arrayVals);
+   m_immediates->setInitializer(constArray);
+
+   m_immediatesToFlush.clear();
+}
+
+llvm::Value *StorageSoa::addrElement(int idx) const
+{
+   std::map<int, llvm::Value*>::const_iterator itr = m_addresses.find(idx);
+   if (itr == m_addresses.end()) {
+      debug_printf("Trying to access invalid shader 'address'\n");
+      return 0;
+   }
+   llvm::Value * res = (*itr).second;
+
+   res = new LoadInst(res, name("addr"), false, m_block);
+
+   return res;
+}
+
+std::vector<llvm::Value*> StorageSoa::inputElement(llvm::Value *idx)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = element(m_input, idx, 0);
+   res[1] = element(m_input, idx, 1);
+   res[2] = element(m_input, idx, 2);
+   res[3] = element(m_input, idx, 3);
+
+   return res;
+}
+
+llvm::Value* StorageSoa::unpackConstElement(llvm::IRBuilder<>* m_builder, llvm::Value* vector, int cc)
+{
+   std::vector<llvm::Value*> x(4);
+   x[0] = m_builder->CreateExtractElement(vector,
+                                           constantInt(cc),
+                                           name("x"));
+
+   VectorType  *vectorType = VectorType::get(Type::FloatTy, 4);
+   Constant *constVector = Constant::getNullValue(vectorType);
+   Value *res = m_builder->CreateInsertElement(constVector, x[0],
+                                              constantInt(0),
+                                              name("vecx"));
+   res = m_builder->CreateInsertElement(res, x[0], constantInt(1),
+                               name("vecxx"));
+   res = m_builder->CreateInsertElement(res, x[0], constantInt(2),
+                               name("vecxxx"));
+   res = m_builder->CreateInsertElement(res, x[0], constantInt(3),
+                               name("vecxxxx"));
+   return res;
+}
+
+std::vector<llvm::Value*> StorageSoa::constElement(llvm::IRBuilder<>* m_builder, llvm::Value *idx)
+{
+   llvm::Value* res;
+   std::vector<llvm::Value*> res2(4);
+   llvm::Value *xChannel;
+
+   xChannel = elementPointer(m_consts, idx, 0);
+
+   res = alignedArrayLoad(xChannel);
+
+   res2[0]=unpackConstElement(m_builder, res,0);
+   res2[1]=unpackConstElement(m_builder, res,1);
+   res2[2]=unpackConstElement(m_builder, res,2);
+   res2[3]=unpackConstElement(m_builder, res,3);
+
+   return res2;
+}
+
+std::vector<llvm::Value*> StorageSoa::outputElement(llvm::Value *idx)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = element(m_output, idx, 0);
+   res[1] = element(m_output, idx, 1);
+   res[2] = element(m_output, idx, 2);
+   res[3] = element(m_output, idx, 3);
+
+   return res;
+}
+
+std::vector<llvm::Value*> StorageSoa::tempElement(llvm::IRBuilder<>* m_builder, int idx)
+{
+   std::vector<llvm::Value*> res(4);
+   llvm::Value *temp = m_temps[idx];
+
+   res[0] = element(temp, constantInt(0), 0);
+   res[1] = element(temp, constantInt(0), 1);
+   res[2] = element(temp, constantInt(0), 2);
+   res[3] = element(temp, constantInt(0), 3);
+
+   return res;
+}
+
+std::vector<llvm::Value*> StorageSoa::immediateElement(llvm::Value *idx)
+{
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = element(m_immediates, idx, 0);
+   res[1] = element(m_immediates, idx, 1);
+   res[2] = element(m_immediates, idx, 2);
+   res[3] = element(m_immediates, idx, 3);
+
+   return res;
+}
+
+llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, llvm::Value *index,
+                                         int channel) const
+{
+   std::vector<Value*> indices;
+   if (m_immediates == ptr)
+      indices.push_back(constantInt(0));
+   indices.push_back(index);
+   indices.push_back(constantInt(channel));
+
+   GetElementPtrInst *getElem = GetElementPtrInst::Create(ptr,
+                                                          indices.begin(),
+                                                          indices.end(),
+                                                          name("ptr"),
+                                                          m_block);
+   return getElem;
+}
+
+llvm::Value * StorageSoa::element(llvm::Value *ptr, llvm::Value *index,
+                                  int channel) const
+{
+   llvm::Value *res = elementPointer(ptr, index, channel);
+   LoadInst *load = new LoadInst(res, name("element"), false, m_block);
+   //load->setAlignment(8);
+   return load;
+}
+
+const char * StorageSoa::name(const char *prefix) const
+{
+   ++m_idx;
+   snprintf(m_name, 32, "%s%d", prefix, m_idx);
+   return m_name;
+}
+
+llvm::ConstantInt * StorageSoa::constantInt(int idx) const
+{
+   if (m_constInts.find(idx) != m_constInts.end()) {
+      return m_constInts[idx];
+   }
+   ConstantInt *constInt = ConstantInt::get(APInt(32,  idx));
+   m_constInts[idx] = constInt;
+   return constInt;
+}
+
+llvm::Value *StorageSoa::alignedArrayLoad(llvm::Value *val)
+{
+   VectorType  *vectorType = VectorType::get(Type::FloatTy, 4);
+   PointerType *vectorPtr  = PointerType::get(vectorType, 0);
+
+   CastInst *cast = new BitCastInst(val, vectorPtr, name("toVector"), m_block);
+   LoadInst *load = new LoadInst(cast, name("alignLoad"), false, m_block);
+   load->setAlignment(8);
+   return load;
+}
+
+llvm::Module * StorageSoa::currentModule() const
+{
+    if (!m_block || !m_block->getParent())
+       return 0;
+
+    return m_block->getParent()->getParent();
+}
+
+llvm::Constant * StorageSoa::createConstGlobalFloat(const float val)
+{
+   Constant*c = ConstantFP::get(APFloat(val));
+   return c;
+}
+
+llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector<float> &vec)
+{
+   VectorType *vectorType = VectorType::get(Type::FloatTy, 4);
+   std::vector<Constant*> immValues;
+   ConstantFP *constx = ConstantFP::get(APFloat(vec[0]));
+   ConstantFP *consty = ConstantFP::get(APFloat(vec[1]));
+   ConstantFP *constz = ConstantFP::get(APFloat(vec[2]));
+   ConstantFP *constw = ConstantFP::get(APFloat(vec[3]));
+   immValues.push_back(constx);
+   immValues.push_back(consty);
+   immValues.push_back(constz);
+   immValues.push_back(constw);
+   Constant  *constVector = ConstantVector::get(vectorType, immValues);
+
+   return constVector;
+}
+
+std::vector<llvm::Value*> StorageSoa::load(enum tgsi_file_type type, int idx, int swizzle,
+                                           llvm::IRBuilder<>* m_builder,llvm::Value *indIdx)
+{
+   std::vector<llvm::Value*> val(4);
+
+   //if we have an indirect index, always use that
+   //   if not use the integer offset to create one
+   llvm::Value *realIndex = 0;
+   if (indIdx)
+      realIndex = indIdx;
+   else
+      realIndex = constantInt(idx);
+   debug_printf("XXXXXXXXX realIdx = %p, indIdx = %p\n", realIndex, indIdx);
+
+   switch(type) {
+   case TGSI_FILE_INPUT:
+      val = inputElement(realIndex);
+      break;
+   case TGSI_FILE_OUTPUT:
+      val = outputElement(realIndex);
+      break;
+   case TGSI_FILE_TEMPORARY:
+      val = tempElement(m_builder, idx);
+      break;
+   case TGSI_FILE_CONSTANT:
+      val = constElement(m_builder, realIndex);
+      break;
+   case TGSI_FILE_IMMEDIATE:
+      val = immediateElement(realIndex);
+      break;
+   case TGSI_FILE_ADDRESS:
+      debug_printf("Address not handled in the load phase!\n");
+      assert(0);
+      break;
+   default:
+      debug_printf("Unknown load!\n");
+      assert(0);
+      break;
+   }
+   if (!gallivm_is_swizzle(swizzle))
+      return val;
+
+   std::vector<llvm::Value*> res(4);
+
+   res[0] = val[gallivm_x_swizzle(swizzle)];
+   res[1] = val[gallivm_y_swizzle(swizzle)];
+   res[2] = val[gallivm_z_swizzle(swizzle)];
+   res[3] = val[gallivm_w_swizzle(swizzle)];
+   return res;
+}
+
+llvm::Value * StorageSoa::allocaTemp(llvm::IRBuilder<>* m_builder)
+{
+   VectorType *vector   = VectorType::get(Type::FloatTy, 4);
+   ArrayType  *vecArray = ArrayType::get(vector, 4);
+   AllocaInst *alloca = new AllocaInst(vecArray, "temp",
+                                       m_builder->GetInsertBlock());
+
+   return alloca;
+}
+
+
+void StorageSoa::store(enum tgsi_file_type type, int idx, const std::vector<llvm::Value*> &val,
+                       int mask, llvm::IRBuilder<>* m_builder)
+{
+   llvm::Value *out = 0;
+   llvm::Value *realIndex = 0;
+   switch(type) {
+   case TGSI_FILE_OUTPUT:
+      out = m_output;
+      realIndex = constantInt(idx);
+      break;
+   case TGSI_FILE_TEMPORARY:
+      // if that temp doesn't already exist, alloca it
+      if (m_temps.find(idx) == m_temps.end())
+         m_temps[idx] = allocaTemp(m_builder);
+
+      out = m_temps[idx];
+
+      realIndex = constantInt(0);
+      break;
+   case TGSI_FILE_INPUT:
+      out = m_input;
+      realIndex = constantInt(idx);
+      break;
+   case TGSI_FILE_ADDRESS: {
+      llvm::Value *addr = m_addresses[idx];
+      if (!addr) {
+         addAddress(idx);
+         addr = m_addresses[idx];
+         assert(addr);
+      }
+      new StoreInst(val[0], addr, false, m_block);
+      return;
+      break;
+   }
+   default:
+      debug_printf("Can't save output of this type: %d !\n", type);
+      assert(0);
+      break;
+   }
+   if ((mask & TGSI_WRITEMASK_X)) {
+      llvm::Value *xChannel = elementPointer(out, realIndex, 0);
+      new StoreInst(val[0], xChannel, false, m_block);
+   }
+   if ((mask & TGSI_WRITEMASK_Y)) {
+      llvm::Value *yChannel = elementPointer(out, realIndex, 1);
+      new StoreInst(val[1], yChannel, false, m_block);
+   }
+   if ((mask & TGSI_WRITEMASK_Z)) {
+      llvm::Value *zChannel = elementPointer(out, realIndex, 2);
+      new StoreInst(val[2], zChannel, false, m_block);
+   }
+   if ((mask & TGSI_WRITEMASK_W)) {
+      llvm::Value *wChannel = elementPointer(out, realIndex, 3);
+      new StoreInst(val[3], wChannel, false, m_block);
+   }
+}
+
+void StorageSoa::addAddress(int idx)
+{
+   GlobalVariable *val = new GlobalVariable(
+      /*Type=*/IntegerType::get(32),
+      /*isConstant=*/false,
+      /*Linkage=*/GlobalValue::ExternalLinkage,
+      /*Initializer=*/0, // has initializer, specified below
+      /*Name=*/name("address"),
+      currentModule());
+   val->setInitializer(Constant::getNullValue(IntegerType::get(32)));
+
+   debug_printf("adding to %d\n", idx);
+   m_addresses[idx] = val;
+}
diff --git a/src/gallium/auxiliary/gallivm/storagesoa.h b/src/gallium/auxiliary/gallivm/storagesoa.h
new file mode 100644
index 0000000000..56886f85e7
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/storagesoa.h
@@ -0,0 +1,107 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef STORAGESOA_H
+#define STORAGESOA_H
+
+#include <pipe/p_shader_tokens.h>
+#include <llvm/Support/IRBuilder.h>
+
+#include <vector>
+#include <list>
+#include <map>
+
+namespace llvm {
+   class BasicBlock;
+   class Constant;
+   class ConstantInt;
+   class GlobalVariable;
+   class LoadInst;
+   class Value;
+   class VectorType;
+   class Module;
+}
+
+class StorageSoa
+{
+public:
+   StorageSoa(llvm::BasicBlock *block,
+              llvm::Value *input,
+              llvm::Value *output,
+              llvm::Value *consts);
+
+
+   std::vector<llvm::Value*> load(enum tgsi_file_type type, int idx, int swizzle, 
+                                  llvm::IRBuilder<>* m_builder, llvm::Value *indIdx =0);
+   void store(enum tgsi_file_type type, int idx, const std::vector<llvm::Value*> &val,
+              int mask, llvm::IRBuilder<>* m_builder);
+
+   void addImmediate(float *vec);
+   void declareImmediates();
+
+   void addAddress(int idx);
+
+   llvm::Value  * addrElement(int idx) const;
+
+   llvm::ConstantInt *constantInt(int) const;
+private:
+   llvm::Value *elementPointer(llvm::Value *ptr, llvm::Value *indIdx,
+                               int channel) const;
+   llvm::Value *element(llvm::Value *ptr, llvm::Value *idx,
+                        int channel) const;
+   const char *name(const char *prefix) const;
+   llvm::Value  *alignedArrayLoad(llvm::Value *val);
+   llvm::Module *currentModule() const;
+   llvm::Constant  *createConstGlobalFloat(const float val);
+   llvm::Constant  *createConstGlobalVector(const std::vector<float> &vec);
+
+   std::vector<llvm::Value*> inputElement(llvm::Value *indIdx);
+   llvm::Value* unpackConstElement(llvm::IRBuilder<>* m_builder, llvm::Value *indIdx, int cc);
+   std::vector<llvm::Value*> constElement(llvm::IRBuilder<>* m_builder, llvm::Value *indIdx);
+   std::vector<llvm::Value*> outputElement(llvm::Value *indIdx);
+   std::vector<llvm::Value*> tempElement(llvm::IRBuilder<>* m_builder, int idx);
+   std::vector<llvm::Value*> immediateElement(llvm::Value *indIdx);
+private:
+   llvm::BasicBlock *m_block;
+
+   llvm::Value *m_input;
+   llvm::Value *m_output;
+   llvm::Value *m_consts;
+   std::map<int, llvm::Value*> m_temps;
+   llvm::GlobalVariable *m_immediates;
+
+   std::map<int, llvm::Value*> m_addresses;
+
+   std::vector<std::vector<float> > m_immediatesToFlush;
+   llvm::Value * allocaTemp(llvm::IRBuilder<>* m_builder);
+
+   mutable std::map<int, llvm::ConstantInt*> m_constInts;
+   mutable char        m_name[32];
+   mutable int         m_idx;
+};
+
+#endif
diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp
new file mode 100644
index 0000000000..5b08200d14
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp
@@ -0,0 +1,1164 @@
+#include "tgsitollvm.h"
+
+#include "gallivm.h"
+#include "gallivm_p.h"
+
+#include "storage.h"
+#include "instructions.h"
+#include "storagesoa.h"
+#include "instructionssoa.h"
+
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_exec.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi/tgsi_build.h"
+#include "tgsi/tgsi_dump.h"
+
+
+#include <llvm/Module.h>
+#include <llvm/CallingConv.h>
+#include <llvm/Constants.h>
+#include <llvm/DerivedTypes.h>
+#include <llvm/Instructions.h>
+#include <llvm/ModuleProvider.h>
+#include <llvm/Pass.h>
+#include <llvm/PassManager.h>
+#include <llvm/Attributes.h>
+#include <llvm/Support/PatternMatch.h>
+#include <llvm/ExecutionEngine/JIT.h>
+#include <llvm/ExecutionEngine/Interpreter.h>
+#include <llvm/ExecutionEngine/GenericValue.h>
+#include <llvm/Support/MemoryBuffer.h>
+#include <llvm/LinkAllPasses.h>
+#include <llvm/Analysis/Verifier.h>
+#include <llvm/Analysis/LoopPass.h>
+#include <llvm/Target/TargetData.h>
+#include <llvm/Bitcode/ReaderWriter.h>
+#include <llvm/Transforms/Utils/Cloning.h>
+
+
+#include <sstream>
+#include <fstream>
+#include <iostream>
+
+using namespace llvm;
+
+static inline FunctionType *vertexShaderFunctionType()
+{
+   //Function takes three arguments,
+   // the calling code has to make sure the types it will
+   // pass are castable to the following:
+   // [4 x <4 x float>] inputs,
+   // [4 x <4 x float>] output,
+   // [4 x [1 x float]] consts,
+
+   std::vector<const Type*> funcArgs;
+   VectorType *vectorType = VectorType::get(Type::FloatTy, 4);
+   ArrayType *vectorArray = ArrayType::get(vectorType, 4);
+   PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0);
+
+   ArrayType   *floatArray     = ArrayType::get(Type::FloatTy, 4);
+   ArrayType   *constsArray    = ArrayType::get(floatArray, 1);
+   PointerType *constsArrayPtr = PointerType::get(constsArray, 0);
+
+   funcArgs.push_back(vectorArrayPtr);//inputs
+   funcArgs.push_back(vectorArrayPtr);//output
+   funcArgs.push_back(constsArrayPtr);//consts
+
+   FunctionType *functionType = FunctionType::get(
+      /*Result=*/Type::VoidTy,
+      /*Params=*/funcArgs,
+      /*isVarArg=*/false);
+
+   return functionType;
+}
+
+static inline void
+add_interpolator(struct gallivm_ir *ir,
+                 struct gallivm_interpolate *interp)
+{
+   ir->interpolators[ir->num_interp] = *interp;
+   ++ir->num_interp;
+}
+
+static void
+translate_declaration(struct gallivm_ir *prog,
+                      llvm::Module *module,
+                      Storage *storage,
+                      struct tgsi_full_declaration *decl,
+                      struct tgsi_full_declaration *fd)
+{
+   if (decl->Declaration.File == TGSI_FILE_INPUT) {
+      unsigned first, last, mask;
+      uint interp_method;
+
+      first = decl->DeclarationRange.First;
+      last = decl->DeclarationRange.Last;
+      mask = decl->Declaration.UsageMask;
+
+      /* Do not touch WPOS.xy */
+      if (first == 0) {
+         mask &= ~TGSI_WRITEMASK_XY;
+         if (mask == TGSI_WRITEMASK_NONE) {
+            first++;
+            if (first > last) {
+               return;
+            }
+         }
+      }
+
+      interp_method = decl->Declaration.Interpolate;
+
+      if (mask == TGSI_WRITEMASK_XYZW) {
+         unsigned i, j;
+
+         for (i = first; i <= last; i++) {
+            for (j = 0; j < NUM_CHANNELS; j++) {
+               //interp( mach, i, j );
+               struct gallivm_interpolate interp;
+               interp.type = interp_method;
+               interp.attrib = i;
+               interp.chan = j;
+               add_interpolator(prog, &interp);
+            }
+         }
+      } else {
+         unsigned i, j;
+         for( j = 0; j < NUM_CHANNELS; j++ ) {
+            if( mask & (1 << j) ) {
+               for( i = first; i <= last; i++ ) {
+                  struct gallivm_interpolate interp;
+                  interp.type = interp_method;
+                  interp.attrib = i;
+                  interp.chan = j;
+                  add_interpolator(prog, &interp);
+               }
+            }
+         }
+      }
+   }
+}
+
+static void
+translate_declarationir(struct gallivm_ir *,
+                      llvm::Module *,
+                      StorageSoa *storage,
+                      struct tgsi_full_declaration *decl,
+                      struct tgsi_full_declaration *)
+{
+   if (decl->Declaration.File == TGSI_FILE_ADDRESS) {
+      int idx = decl->DeclarationRange.First;
+      storage->addAddress(idx);
+   }
+}
+
+static void
+translate_immediate(Storage *storage,
+                    struct tgsi_full_immediate *imm)
+{
+   float vec[4];
+   int i;
+   for (i = 0; i < imm->Immediate.NrTokens - 1; ++i) {
+      switch (imm->Immediate.DataType) {
+      case TGSI_IMM_FLOAT32:
+         vec[i] = imm->u.ImmediateFloat32[i].Float;
+         break;
+      default:
+         assert(0);
+      }
+   }
+   storage->addImmediate(vec);
+}
+
+
+static void
+translate_immediateir(StorageSoa *storage,
+                      struct tgsi_full_immediate *imm)
+{
+   float vec[4];
+   int i;
+   for (i = 0; i < imm->Immediate.NrTokens - 1; ++i) {
+      switch (imm->Immediate.DataType) {
+      case TGSI_IMM_FLOAT32:
+         vec[i] = imm->u.ImmediateFloat32[i].Float;
+         break;
+      default:
+         assert(0);
+      }
+   }
+   storage->addImmediate(vec);
+}
+
+static inline int
+swizzleInt(struct tgsi_full_src_register *src)
+{
+   int swizzle = 0;
+   int start = 1000;
+
+   for (int k = 0; k < 4; ++k) {
+      swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start;
+      start /= 10;
+   }
+   return swizzle;
+}
+
+static inline llvm::Value *
+swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src,
+              Storage *storage)
+{
+   int swizzle = swizzleInt(src);
+
+   if (gallivm_is_swizzle(swizzle)) {
+      /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/
+      val = storage->shuffleVector(val, swizzle);
+   }
+   return val;
+}
+
+static void
+translate_instruction(llvm::Module *module,
+                      Storage *storage,
+                      Instructions *instr,
+                      struct tgsi_full_instruction *inst,
+                      struct tgsi_full_instruction *fi,
+                      unsigned instno)
+{
+   llvm::Value *inputs[4];
+   inputs[0] = 0;
+   inputs[1] = 0;
+   inputs[2] = 0;
+   inputs[3] = 0;
+
+   for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) {
+      struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i];
+      llvm::Value *val = 0;
+      llvm::Value *indIdx = 0;
+
+      if (src->SrcRegister.Indirect) {
+         indIdx = storage->addrElement(src->SrcRegisterInd.Index);
+         indIdx = storage->extractIndex(indIdx);
+      }
+      if (src->SrcRegister.File == TGSI_FILE_CONSTANT) {
+         val = storage->constElement(src->SrcRegister.Index, indIdx);
+      } else if (src->SrcRegister.File == TGSI_FILE_INPUT) {
+         val = storage->inputElement(src->SrcRegister.Index, indIdx);
+      } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) {
+         val = storage->tempElement(src->SrcRegister.Index);
+      } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) {
+         val = storage->outputElement(src->SrcRegister.Index, indIdx);
+      } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) {
+         val = storage->immediateElement(src->SrcRegister.Index);
+      } else {
+         fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File);
+         return;
+      }
+
+      inputs[i] = swizzleVector(val, src, storage);
+   }
+
+   /*if (inputs[0])
+     instr->printVector(inputs[0]);
+     if (inputs[1])
+     instr->printVector(inputs[1]);*/
+   llvm::Value *out = 0;
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_ARL: {
+      out = instr->arl(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_MOV: {
+      out = inputs[0];
+   }
+      break;
+   case TGSI_OPCODE_LIT: {
+      out = instr->lit(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_RCP: {
+      out = instr->rcp(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_RSQ: {
+      out = instr->rsq(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_EXP: {
+      out = instr->exp(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_LOG: {
+      out = instr->log(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_MUL: {
+      out = instr->mul(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_ADD: {
+      out = instr->add(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_DP3: {
+      out = instr->dp3(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_DP4: {
+      out = instr->dp4(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_DST: {
+      out = instr->dst(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_MIN: {
+      out = instr->min(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_MAX: {
+      out = instr->max(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_SLT: {
+      out = instr->slt(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_SGE: {
+      out = instr->sge(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_MAD: {
+      out = instr->madd(inputs[0], inputs[1], inputs[2]);
+   }
+      break;
+   case TGSI_OPCODE_SUB: {
+      out = instr->sub(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_LERP: {
+      out = instr->lerp(inputs[0], inputs[1], inputs[2]);
+   }
+      break;
+   case TGSI_OPCODE_CND: {
+      out = instr->cnd(inputs[0], inputs[1], inputs[2]);
+   }
+      break;
+   case TGSI_OPCODE_CND0: {
+      out = instr->cnd0(inputs[0], inputs[1], inputs[2]);
+   }
+      break;
+   case TGSI_OPCODE_DOT2ADD: {
+      out = instr->dot2add(inputs[0], inputs[1], inputs[2]);
+   }
+      break;
+   case TGSI_OPCODE_INDEX:
+      break;
+   case TGSI_OPCODE_NEGATE: {
+      out = instr->neg(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_FRAC: {
+      out = instr->frc(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_CLAMP: {
+      out = instr->clamp(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_FLOOR: {
+      out = instr->floor(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_ROUND:
+      break;
+   case TGSI_OPCODE_EXPBASE2: {
+      out = instr->ex2(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_LOGBASE2: {
+      out = instr->lg2(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_POWER: {
+      out = instr->pow(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_CROSSPRODUCT: {
+      out = instr->cross(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_MULTIPLYMATRIX:
+      break;
+   case TGSI_OPCODE_ABS: {
+      out = instr->abs(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_RCC:
+      break;
+   case TGSI_OPCODE_DPH: {
+      out = instr->dph(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_COS: {
+      out = instr->cos(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_DDX: {
+      out = instr->ddx(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_DDY: {
+      out = instr->ddy(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_KILP:
+      break;
+   case TGSI_OPCODE_PK2H:
+      break;
+   case TGSI_OPCODE_PK2US:
+      break;
+   case TGSI_OPCODE_PK4B:
+      break;
+   case TGSI_OPCODE_PK4UB:
+      break;
+   case TGSI_OPCODE_RFL:
+      break;
+   case TGSI_OPCODE_SEQ: {
+      out = instr->seq(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_SFL: {
+      out = instr->sfl(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_SGT: {
+      out = instr->sgt(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_SIN: {
+      out = instr->sin(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_SLE: {
+      out = instr->sle(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_SNE: {
+      out = instr->sne(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_STR: {
+      out = instr->str(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_TEX:
+      break;
+   case TGSI_OPCODE_TXD:
+      break;
+   case TGSI_OPCODE_UP2H:
+      break;
+   case TGSI_OPCODE_UP2US:
+      break;
+   case TGSI_OPCODE_UP4B:
+      break;
+   case TGSI_OPCODE_UP4UB:
+      break;
+   case TGSI_OPCODE_X2D: {
+      out = instr->x2d(inputs[0], inputs[1], inputs[2]);
+   }
+      break;
+   case TGSI_OPCODE_ARA:
+      break;
+   case TGSI_OPCODE_ARR:
+      break;
+   case TGSI_OPCODE_BRA:
+      break;
+   case TGSI_OPCODE_CAL: {
+      instr->cal(inst->InstructionExtLabel.Label, storage->inputPtr());
+      return;
+   }
+      break;
+   case TGSI_OPCODE_RET: {
+      instr->end();
+      return;
+   }
+      break;
+   case TGSI_OPCODE_SSG:
+      break;
+   case TGSI_OPCODE_CMP: {
+      out = instr->cmp(inputs[0], inputs[1], inputs[2]);
+   }
+      break;
+   case TGSI_OPCODE_SCS: {
+      out = instr->scs(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_TXB:
+      break;
+   case TGSI_OPCODE_NRM4:
+   case TGSI_OPCODE_NRM: {
+      out = instr->nrm(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_DIV: {
+      out = instr->div(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_DP2: {
+      out = instr->dp2(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_TXL:
+      break;
+   case TGSI_OPCODE_BRK: {
+      instr->brk();
+      return;
+   }
+      break;
+   case TGSI_OPCODE_IF: {
+      instr->ifop(inputs[0]);
+      storage->setCurrentBlock(instr->currentBlock());
+      return;  //just update the state
+   }
+      break;
+   case TGSI_OPCODE_LOOP:
+      break;
+   case TGSI_OPCODE_REP:
+      break;
+   case TGSI_OPCODE_ELSE: {
+      instr->elseop();
+      storage->setCurrentBlock(instr->currentBlock());
+      return; //only state update
+   }
+      break;
+   case TGSI_OPCODE_ENDIF: {
+      instr->endif();
+      storage->setCurrentBlock(instr->currentBlock());
+      return; //just update the state
+   }
+      break;
+   case TGSI_OPCODE_ENDLOOP:
+      break;
+   case TGSI_OPCODE_ENDREP:
+      break;
+   case TGSI_OPCODE_PUSHA:
+      break;
+   case TGSI_OPCODE_POPA:
+      break;
+   case TGSI_OPCODE_CEIL:
+      break;
+   case TGSI_OPCODE_I2F:
+      break;
+   case TGSI_OPCODE_NOT:
+      break;
+   case TGSI_OPCODE_TRUNC: {
+      out = instr->trunc(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_SHL:
+      break;
+   case TGSI_OPCODE_SHR:
+      break;
+   case TGSI_OPCODE_AND:
+      break;
+   case TGSI_OPCODE_OR:
+      break;
+   case TGSI_OPCODE_MOD:
+      break;
+   case TGSI_OPCODE_XOR:
+      break;
+   case TGSI_OPCODE_SAD:
+      break;
+   case TGSI_OPCODE_TXF:
+      break;
+   case TGSI_OPCODE_TXQ:
+      break;
+   case TGSI_OPCODE_CONT:
+      break;
+   case TGSI_OPCODE_EMIT:
+      break;
+   case TGSI_OPCODE_ENDPRIM:
+      break;
+   case TGSI_OPCODE_BGNLOOP2: {
+      instr->beginLoop();
+      storage->setCurrentBlock(instr->currentBlock());
+      return;
+   }
+      break;
+   case TGSI_OPCODE_BGNSUB: {
+      instr->bgnSub(instno);
+      storage->setCurrentBlock(instr->currentBlock());
+      storage->pushTemps();
+      return;
+   }
+      break;
+   case TGSI_OPCODE_ENDLOOP2: {
+      instr->endLoop();
+      storage->setCurrentBlock(instr->currentBlock());
+      return;
+   }
+      break;
+   case TGSI_OPCODE_ENDSUB: {
+      instr->endSub();
+      storage->setCurrentBlock(instr->currentBlock());
+      storage->popArguments();
+      storage->popTemps();
+      return;
+   }
+      break;
+   case TGSI_OPCODE_NOISE1:
+      break;
+   case TGSI_OPCODE_NOISE2:
+      break;
+   case TGSI_OPCODE_NOISE3:
+      break;
+   case TGSI_OPCODE_NOISE4:
+      break;
+   case TGSI_OPCODE_NOP:
+      break;
+   case TGSI_OPCODE_M4X3:
+      break;
+   case TGSI_OPCODE_M3X4:
+      break;
+   case TGSI_OPCODE_M3X3:
+      break;
+   case TGSI_OPCODE_M3X2:
+      break;
+   case TGSI_OPCODE_CALLNZ:
+      break;
+   case TGSI_OPCODE_IFC:
+      break;
+   case TGSI_OPCODE_BREAKC:
+      break;
+   case TGSI_OPCODE_KIL: {
+      out = instr->kil(inputs[0]);
+      storage->setKilElement(out);
+      return;
+   }
+      break;
+   case TGSI_OPCODE_END:
+      instr->end();
+      return;
+      break;
+   default:
+      fprintf(stderr, "ERROR: Unknown opcode %d\n",
+              inst->Instruction.Opcode);
+      assert(0);
+      break;
+   }
+
+   if (!out) {
+      fprintf(stderr, "ERROR: unsupported opcode %d\n",
+              inst->Instruction.Opcode);
+      assert(!"Unsupported opcode");
+   }
+
+   /* # not sure if we need this */
+   switch( inst->Instruction.Saturate ) {
+   case TGSI_SAT_NONE:
+      break;
+   case TGSI_SAT_ZERO_ONE:
+      /*TXT( "_SAT" );*/
+      break;
+   case TGSI_SAT_MINUS_PLUS_ONE:
+      /*TXT( "_SAT[-1,1]" );*/
+      break;
+   default:
+      assert( 0 );
+   }
+
+   /* store results  */
+   for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) {
+      struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
+
+      if (dst->DstRegister.File == TGSI_FILE_OUTPUT) {
+         storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
+      } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) {
+         storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
+      } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) {
+         storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask);
+      } else {
+         fprintf(stderr, "ERROR: unsupported LLVM destination!");
+         assert(!"wrong destination");
+      }
+   }
+}
+
+
+static void
+translate_instructionir(llvm::Module *module,
+                        StorageSoa *storage,
+                        InstructionsSoa *instr,
+                        struct tgsi_full_instruction *inst,
+                        struct tgsi_full_instruction *fi,
+                        unsigned instno)
+{
+   std::vector< std::vector<llvm::Value*> > inputs(inst->Instruction.NumSrcRegs);
+
+   for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) {
+      struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i];
+      std::vector<llvm::Value*> val;
+      llvm::Value *indIdx = 0;
+      int swizzle = swizzleInt(src);
+
+      if (src->SrcRegister.Indirect) {
+         indIdx = storage->addrElement(src->SrcRegisterInd.Index);
+      }
+      val = storage->load((enum tgsi_file_type)src->SrcRegister.File,
+                          src->SrcRegister.Index, swizzle, instr->getIRBuilder(), indIdx);
+
+      inputs[i] = val;
+   }
+
+   std::vector<llvm::Value*> out(4);
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_ARL: {
+      out = instr->arl(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_MOV: {
+      out = inputs[0];
+   }
+      break;
+   case TGSI_OPCODE_LIT: {
+      out = instr->lit(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_RCP: {
+   }
+      break;
+   case TGSI_OPCODE_RSQ: {
+      out = instr->rsq(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_EXP:
+      break;
+   case TGSI_OPCODE_LOG:
+      break;
+   case TGSI_OPCODE_MUL: {
+      out = instr->mul(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_ADD: {
+      out = instr->add(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_DP3: {
+      out = instr->dp3(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_DP4: {
+      out = instr->dp4(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_DST: {
+   }
+      break;
+   case TGSI_OPCODE_MIN: {
+      out = instr->min(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_MAX: {
+      out = instr->max(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_SLT: {
+      out = instr->slt(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_SGE: {
+   }
+      break;
+   case TGSI_OPCODE_MAD: {
+      out = instr->madd(inputs[0], inputs[1], inputs[2]);
+   }
+      break;
+   case TGSI_OPCODE_SUB: {
+      out = instr->sub(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_LERP: {
+   }
+      break;
+   case TGSI_OPCODE_CND:
+      break;
+   case TGSI_OPCODE_CND0:
+      break;
+   case TGSI_OPCODE_DOT2ADD:
+      break;
+   case TGSI_OPCODE_INDEX:
+      break;
+   case TGSI_OPCODE_NEGATE:
+      break;
+   case TGSI_OPCODE_FRAC: {
+   }
+      break;
+   case TGSI_OPCODE_CLAMP:
+      break;
+   case TGSI_OPCODE_FLOOR: {
+   }
+      break;
+   case TGSI_OPCODE_ROUND:
+      break;
+   case TGSI_OPCODE_EXPBASE2: {
+   }
+      break;
+   case TGSI_OPCODE_LOGBASE2: {
+   }
+      break;
+   case TGSI_OPCODE_POWER: {
+      out = instr->pow(inputs[0], inputs[1]);
+   }
+      break;
+   case TGSI_OPCODE_CROSSPRODUCT: {
+   }
+      break;
+   case TGSI_OPCODE_MULTIPLYMATRIX:
+      break;
+   case TGSI_OPCODE_ABS: {
+      out = instr->abs(inputs[0]);
+   }
+      break;
+   case TGSI_OPCODE_RCC:
+      break;
+   case TGSI_OPCODE_DPH: {
+   }
+      break;
+   case TGSI_OPCODE_COS: {
+   }
+      break;
+   case TGSI_OPCODE_DDX:
+      break;
+   case TGSI_OPCODE_DDY:
+      break;
+   case TGSI_OPCODE_KILP:
+      break;
+   case TGSI_OPCODE_PK2H:
+      break;
+   case TGSI_OPCODE_PK2US:
+      break;
+   case TGSI_OPCODE_PK4B:
+      break;
+   case TGSI_OPCODE_PK4UB:
+      break;
+   case TGSI_OPCODE_RFL:
+      break;
+   case TGSI_OPCODE_SEQ:
+      break;
+   case TGSI_OPCODE_SFL:
+      break;
+   case TGSI_OPCODE_SGT: {
+   }
+      break;
+   case TGSI_OPCODE_SIN: {
+   }
+      break;
+   case TGSI_OPCODE_SLE:
+      break;
+   case TGSI_OPCODE_SNE:
+      break;
+   case TGSI_OPCODE_STR:
+      break;
+   case TGSI_OPCODE_TEX:
+      break;
+   case TGSI_OPCODE_TXD:
+      break;
+   case TGSI_OPCODE_UP2H:
+      break;
+   case TGSI_OPCODE_UP2US:
+      break;
+   case TGSI_OPCODE_UP4B:
+      break;
+   case TGSI_OPCODE_UP4UB:
+      break;
+   case TGSI_OPCODE_X2D:
+      break;
+   case TGSI_OPCODE_ARA:
+      break;
+   case TGSI_OPCODE_ARR:
+      break;
+   case TGSI_OPCODE_BRA:
+      break;
+   case TGSI_OPCODE_CAL: {
+   }
+      break;
+   case TGSI_OPCODE_RET: {
+   }
+      break;
+   case TGSI_OPCODE_SSG:
+      break;
+   case TGSI_OPCODE_CMP: {
+   }
+      break;
+   case TGSI_OPCODE_SCS: {
+   }
+      break;
+   case TGSI_OPCODE_TXB:
+      break;
+   case TGSI_OPCODE_NRM:
+      break;
+   case TGSI_OPCODE_DIV:
+      break;
+   case TGSI_OPCODE_DP2:
+      break;
+   case TGSI_OPCODE_TXL:
+      break;
+   case TGSI_OPCODE_BRK: {
+   }
+      break;
+   case TGSI_OPCODE_IF: {
+   }
+      break;
+   case TGSI_OPCODE_LOOP:
+      break;
+   case TGSI_OPCODE_REP:
+      break;
+   case TGSI_OPCODE_ELSE: {
+   }
+      break;
+   case TGSI_OPCODE_ENDIF: {
+   }
+      break;
+   case TGSI_OPCODE_ENDLOOP:
+      break;
+   case TGSI_OPCODE_ENDREP:
+      break;
+   case TGSI_OPCODE_PUSHA:
+      break;
+   case TGSI_OPCODE_POPA:
+      break;
+   case TGSI_OPCODE_CEIL:
+      break;
+   case TGSI_OPCODE_I2F:
+      break;
+   case TGSI_OPCODE_NOT:
+      break;
+   case TGSI_OPCODE_TRUNC: {
+   }
+      break;
+   case TGSI_OPCODE_SHL:
+      break;
+   case TGSI_OPCODE_SHR:
+      break;
+   case TGSI_OPCODE_AND:
+      break;
+   case TGSI_OPCODE_OR:
+      break;
+   case TGSI_OPCODE_MOD:
+      break;
+   case TGSI_OPCODE_XOR:
+      break;
+   case TGSI_OPCODE_SAD:
+      break;
+   case TGSI_OPCODE_TXF:
+      break;
+   case TGSI_OPCODE_TXQ:
+      break;
+   case TGSI_OPCODE_CONT:
+      break;
+   case TGSI_OPCODE_EMIT:
+      break;
+   case TGSI_OPCODE_ENDPRIM:
+      break;
+   case TGSI_OPCODE_BGNLOOP2: {
+   }
+      break;
+   case TGSI_OPCODE_BGNSUB: {
+   }
+      break;
+   case TGSI_OPCODE_ENDLOOP2: {
+   }
+      break;
+   case TGSI_OPCODE_ENDSUB: {
+   }
+      break;
+   case TGSI_OPCODE_NOISE1:
+      break;
+   case TGSI_OPCODE_NOISE2:
+      break;
+   case TGSI_OPCODE_NOISE3:
+      break;
+   case TGSI_OPCODE_NOISE4:
+      break;
+   case TGSI_OPCODE_NOP:
+      break;
+   case TGSI_OPCODE_M4X3:
+      break;
+   case TGSI_OPCODE_M3X4:
+      break;
+   case TGSI_OPCODE_M3X3:
+      break;
+   case TGSI_OPCODE_M3X2:
+      break;
+   case TGSI_OPCODE_NRM4:
+      break;
+   case TGSI_OPCODE_CALLNZ:
+      break;
+   case TGSI_OPCODE_IFC:
+      break;
+   case TGSI_OPCODE_BREAKC:
+      break;
+   case TGSI_OPCODE_KIL: {
+   }
+      break;
+   case TGSI_OPCODE_END:
+      instr->end();
+      return;
+      break;
+   default:
+      fprintf(stderr, "ERROR: Unknown opcode %d\n",
+              inst->Instruction.Opcode);
+      assert(0);
+      break;
+   }
+
+   if (!out[0]) {
+      fprintf(stderr, "ERROR: unsupported opcode %d\n",
+              inst->Instruction.Opcode);
+      assert(!"Unsupported opcode");
+   }
+
+   /* store results  */
+   for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) {
+      struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
+      storage->store((enum tgsi_file_type)dst->DstRegister.File,
+                     dst->DstRegister.Index, out, dst->DstRegister.WriteMask,
+		     instr->getIRBuilder() );
+   }
+}
+
+llvm::Module *
+tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens)
+{
+   llvm::Module *mod = new Module("shader");
+   struct tgsi_parse_context parse;
+   struct tgsi_full_instruction fi;
+   struct tgsi_full_declaration fd;
+   unsigned instno = 0;
+   Function* shader = mod->getFunction("execute_shader");
+   std::ostringstream stream;
+   if (ir->type == GALLIVM_VS) {
+      stream << "vs_shader";
+   } else {
+      stream << "fs_shader";
+   }
+   stream << ir->id;
+   std::string func_name = stream.str();
+   shader->setName(func_name.c_str());
+
+   Function::arg_iterator args = shader->arg_begin();
+   Value *ptr_INPUT = args++;
+   ptr_INPUT->setName("input");
+
+   BasicBlock *label_entry = BasicBlock::Create("entry", shader, 0);
+
+   tgsi_parse_init(&parse, tokens);
+
+   fi = tgsi_default_full_instruction();
+   fd = tgsi_default_full_declaration();
+   Storage storage(label_entry, ptr_INPUT);
+   Instructions instr(mod, shader, label_entry, &storage);
+   while(!tgsi_parse_end_of_tokens(&parse)) {
+      tgsi_parse_token(&parse);
+
+      switch (parse.FullToken.Token.Type) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         translate_declaration(ir, mod, &storage,
+                               &parse.FullToken.FullDeclaration,
+                               &fd);
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         translate_immediate(&storage,
+                             &parse.FullToken.FullImmediate);
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         translate_instruction(mod, &storage, &instr,
+                               &parse.FullToken.FullInstruction,
+                               &fi, instno);
+         ++instno;
+         break;
+
+      default:
+         assert(0);
+      }
+   }
+
+   tgsi_parse_free(&parse);
+
+   ir->num_consts = storage.numConsts();
+   return mod;
+}
+
+llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir,
+                              const struct tgsi_token *tokens)
+{
+   llvm::Module *mod = new Module("shader");
+   struct tgsi_parse_context parse;
+   struct tgsi_full_instruction fi;
+   struct tgsi_full_declaration fd;
+   unsigned instno = 0;
+   std::ostringstream stream;
+   if (ir->type == GALLIVM_VS) {
+      stream << "vs_shader";
+   } else {
+      stream << "fs_shader";
+   }
+   //stream << ir->id;
+   std::string func_name = stream.str();
+   Function *shader = llvm::cast<Function>(mod->getOrInsertFunction(
+                                              func_name.c_str(),
+                                              vertexShaderFunctionType()));
+
+   Function::arg_iterator args = shader->arg_begin();
+   Value *input = args++;
+   input->setName("inputs");
+   Value *output = args++;
+   output->setName("outputs");
+   Value *consts = args++;
+   consts->setName("consts");
+
+   BasicBlock *label_entry = BasicBlock::Create("entry", shader, 0);
+
+   tgsi_parse_init(&parse, tokens);
+
+   fi = tgsi_default_full_instruction();
+   fd = tgsi_default_full_declaration();
+
+   StorageSoa storage(label_entry, input, output, consts);
+   InstructionsSoa instr(mod, shader, label_entry, &storage);
+
+   while(!tgsi_parse_end_of_tokens(&parse)) {
+      tgsi_parse_token(&parse);
+
+      switch (parse.FullToken.Token.Type) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         translate_declarationir(ir, mod, &storage,
+                                 &parse.FullToken.FullDeclaration,
+                                 &fd);
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         translate_immediateir(&storage,
+                             &parse.FullToken.FullImmediate);
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         storage.declareImmediates();
+         translate_instructionir(mod, &storage, &instr,
+                                 &parse.FullToken.FullInstruction,
+                                 &fi, instno);
+         ++instno;
+         break;
+
+      default:
+         assert(0);
+      }
+   }
+
+   tgsi_parse_free(&parse);
+
+   return mod;
+}
diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.h b/src/gallium/auxiliary/gallivm/tgsitollvm.h
new file mode 100644
index 0000000000..7ada04d629
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/tgsitollvm.h
@@ -0,0 +1,20 @@
+#ifndef TGSITOLLVM_H
+#define TGSITOLLVM_H
+
+
+namespace llvm {
+   class Module;
+}
+
+struct gallivm_ir;
+struct tgsi_token;
+
+
+llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir,
+                            const struct tgsi_token *tokens);
+
+
+llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir,
+                              const struct tgsi_token *tokens);
+
+#endif
diff --git a/src/gallium/auxiliary/indices/Makefile b/src/gallium/auxiliary/indices/Makefile
new file mode 100644
index 0000000000..8fa61d265e
--- /dev/null
+++ b/src/gallium/auxiliary/indices/Makefile
@@ -0,0 +1,16 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = indices
+
+C_SOURCES = \
+	u_indices_gen.c 
+
+include ../../Makefile.template
+
+u_indices_gen.c: u_indices_gen.py
+	python $< > $@
+
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/indices/SConscript b/src/gallium/auxiliary/indices/SConscript
new file mode 100644
index 0000000000..65a43a9f64
--- /dev/null
+++ b/src/gallium/auxiliary/indices/SConscript
@@ -0,0 +1,17 @@
+Import('*')
+
+env.CodeGenerate(
+	target = 'u_indices_gen.c', 
+	script = 'u_indices_gen.py', 
+	source = [],
+	command = 'python $SCRIPT > $TARGET'
+)
+
+indices = env.ConvenienceLibrary(
+	target = 'indices',
+	source = [
+#               'u_indices.c',
+               'u_indices_gen.c',
+	])
+
+auxiliaries.insert(0, indices)
diff --git a/src/gallium/auxiliary/indices/u_indices.c b/src/gallium/auxiliary/indices/u_indices.c
new file mode 100644
index 0000000000..0cf7d88653
--- /dev/null
+++ b/src/gallium/auxiliary/indices/u_indices.c
@@ -0,0 +1,253 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "u_indices.h"
+#include "u_indices_priv.h"
+
+static void translate_memcpy_ushort( const void *in,
+                                     unsigned nr,
+                                     void *out )
+{
+   memcpy(out, in, nr*sizeof(short));
+}
+                              
+static void translate_memcpy_uint( const void *in,
+                                   unsigned nr,
+                                   void *out )
+{
+   memcpy(out, in, nr*sizeof(int));
+}
+                              
+
+int u_index_translator( unsigned hw_mask,
+                        unsigned prim,
+                        unsigned in_index_size,
+                        unsigned nr,
+                        unsigned in_pv,
+                        unsigned out_pv,
+                        unsigned *out_prim,
+                        unsigned *out_index_size,
+                        unsigned *out_nr,
+                        u_translate_func *out_translate )
+{
+   unsigned in_idx;
+   unsigned out_idx;
+   int ret = U_TRANSLATE_NORMAL;
+
+   u_index_init();
+
+   in_idx = in_size_idx(in_index_size);
+   *out_index_size = (in_index_size == 4) ? 4 : 2;
+   out_idx = out_size_idx(*out_index_size);
+
+   if ((hw_mask & (1<<prim)) && 
+       in_index_size == *out_index_size &&
+       in_pv == out_pv) 
+   {
+      if (in_index_size == 4)
+         *out_translate = translate_memcpy_uint;
+      else
+         *out_translate = translate_memcpy_ushort;
+
+      *out_prim = prim;
+      *out_nr = nr;
+
+      return U_TRANSLATE_MEMCPY;
+   }
+   else {
+      switch (prim) {
+      case PIPE_PRIM_POINTS:
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_POINTS;
+         *out_nr = nr;
+         break;
+
+      case PIPE_PRIM_LINES:
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_LINES;
+         *out_nr = nr;
+         break;
+
+      case PIPE_PRIM_LINE_STRIP:
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_LINES;
+         *out_nr = (nr - 1) * 2;
+         break;
+
+      case PIPE_PRIM_LINE_LOOP:
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_LINES;
+         *out_nr = nr * 2;
+         break;
+
+      case PIPE_PRIM_TRIANGLES:
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = nr;
+         break;
+
+      case PIPE_PRIM_TRIANGLE_STRIP:
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = (nr - 2) * 3;
+         break;
+
+      case PIPE_PRIM_TRIANGLE_FAN:
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = (nr - 2) * 3;
+         break;
+
+      case PIPE_PRIM_QUADS:
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = (nr / 4) * 6;
+         break;
+
+      case PIPE_PRIM_QUAD_STRIP:
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = (nr - 2) * 3;
+         break;
+
+      case PIPE_PRIM_POLYGON:
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = (nr - 2) * 3;
+         break;
+
+      default:
+         assert(0);
+         *out_translate = translate[in_idx][out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_POINTS;
+         *out_nr = nr;
+         return U_TRANSLATE_ERROR;
+      }
+   }
+
+   return ret;
+}
+
+
+
+
+
+int u_index_generator( unsigned hw_mask,
+                       unsigned prim,
+                       unsigned start,
+                       unsigned nr,
+                       unsigned in_pv,
+                       unsigned out_pv,
+                       unsigned *out_prim,
+                       unsigned *out_index_size,
+                       unsigned *out_nr,
+                       u_generate_func *out_generate )
+
+{
+   unsigned out_idx;
+
+   u_index_init();
+
+   *out_index_size = ((start + nr) > 0xfffe) ? 4 : 2;
+   out_idx = out_size_idx(*out_index_size);
+
+   if ((hw_mask & (1<<prim)) && 
+       (in_pv == out_pv)) {
+       
+      *out_generate = generate[out_idx][in_pv][out_pv][PIPE_PRIM_POINTS];
+      *out_prim = prim;
+      *out_nr = nr;
+      return U_GENERATE_LINEAR;
+   }
+   else {
+      switch (prim) {
+      case PIPE_PRIM_POINTS:
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_POINTS;
+         *out_nr = nr;
+         return U_GENERATE_REUSABLE;
+
+      case PIPE_PRIM_LINES:
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_LINES;
+         *out_nr = nr;
+         return U_GENERATE_REUSABLE;
+
+      case PIPE_PRIM_LINE_STRIP:
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_LINES;
+         *out_nr = (nr - 1) * 2;
+         return U_GENERATE_REUSABLE;
+
+      case PIPE_PRIM_LINE_LOOP:
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_LINES;
+         *out_nr = nr * 2;
+         return U_GENERATE_ONE_OFF;
+
+      case PIPE_PRIM_TRIANGLES:
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = nr;
+         return U_GENERATE_REUSABLE;
+
+      case PIPE_PRIM_TRIANGLE_STRIP:
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = (nr - 2) * 3;
+         return U_GENERATE_REUSABLE;
+
+      case PIPE_PRIM_TRIANGLE_FAN:
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = (nr - 2) * 3;
+         return U_GENERATE_REUSABLE;
+
+      case PIPE_PRIM_QUADS:
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = (nr / 4) * 6;
+         return U_GENERATE_REUSABLE;
+
+      case PIPE_PRIM_QUAD_STRIP:
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = (nr - 2) * 3;
+         return U_GENERATE_REUSABLE;
+
+      case PIPE_PRIM_POLYGON:
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_TRIANGLES;
+         *out_nr = (nr - 2) * 3;
+         return U_GENERATE_REUSABLE;
+
+      default:
+         assert(0);
+         *out_generate = generate[out_idx][in_pv][out_pv][prim];
+         *out_prim = PIPE_PRIM_POINTS;
+         *out_nr = nr;
+         return U_TRANSLATE_ERROR;
+      }
+   }
+}
diff --git a/src/gallium/auxiliary/indices/u_indices.h b/src/gallium/auxiliary/indices/u_indices.h
new file mode 100644
index 0000000000..abf5a3037d
--- /dev/null
+++ b/src/gallium/auxiliary/indices/u_indices.h
@@ -0,0 +1,83 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef U_INDICES_H
+#define U_INDICES_H
+
+#include "pipe/p_compiler.h"
+
+#define PV_FIRST      0
+#define PV_LAST       1
+#define PV_COUNT      2
+
+typedef void (*u_translate_func)( const void *in,
+                                  unsigned nr,
+                                  void *out );
+
+typedef void (*u_generate_func)( unsigned nr,
+                                 void *out );
+
+
+/* Return codes describe the translate/generate operation.  Caller may
+ * be able to reuse translated indices under some circumstances.
+ */
+#define U_TRANSLATE_ERROR  -1
+#define U_TRANSLATE_NORMAL  1
+#define U_TRANSLATE_MEMCPY  2
+#define U_GENERATE_LINEAR   3
+#define U_GENERATE_REUSABLE 4
+#define U_GENERATE_ONE_OFF  5
+
+
+void u_index_init( void );
+
+int u_index_translator( unsigned hw_mask,
+                        unsigned prim,
+                        unsigned in_index_size,
+                        unsigned nr,
+                        unsigned in_pv,   /* API */
+                        unsigned out_pv,  /* hardware */
+                        unsigned *out_prim,
+                        unsigned *out_index_size,
+                        unsigned *out_nr,
+                        u_translate_func *out_translate );
+
+/* Note that even when generating it is necessary to know what the
+ * API's PV is, as the indices generated will depend on whether it is
+ * the same as hardware or not, and in the case of triangle strips,
+ * whether it is first or last.
+ */
+int u_index_generator( unsigned hw_mask,
+                       unsigned prim,
+                       unsigned start,
+                       unsigned nr,
+                       unsigned in_pv,   /* API */
+                       unsigned out_pv,  /* hardware */
+                       unsigned *out_prim,
+                       unsigned *out_index_size,
+                       unsigned *out_nr,
+                       u_generate_func *out_generate );
+
+
+#endif
diff --git a/src/gallium/auxiliary/indices/u_indices_gen.c b/src/gallium/auxiliary/indices/u_indices_gen.c
new file mode 100644
index 0000000000..4c05b3eedb
--- /dev/null
+++ b/src/gallium/auxiliary/indices/u_indices_gen.c
@@ -0,0 +1,5129 @@
+/* File automatically generated by indices.py */
+
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+
+/**
+ * @file
+ * Functions to translate and generate index lists
+ */
+
+#include "indices/u_indices.h"
+#include "indices/u_indices_priv.h"
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+
+
+static unsigned out_size_idx( unsigned index_size )
+{
+   switch (index_size) {
+   case 4: return OUT_UINT;
+   case 2: return OUT_USHORT;
+   default: assert(0); return OUT_USHORT;
+   }
+}
+
+static unsigned in_size_idx( unsigned index_size )
+{
+   switch (index_size) {
+   case 4: return IN_UINT;
+   case 2: return IN_USHORT;
+   case 1: return IN_UBYTE;
+   default: assert(0); return IN_UBYTE;
+   }
+}
+
+
+static u_translate_func translate[IN_COUNT][OUT_COUNT][PV_COUNT][PV_COUNT][PRIM_COUNT];
+static u_generate_func  generate[OUT_COUNT][PV_COUNT][PV_COUNT][PRIM_COUNT];
+
+
+
+static void generate_points_ushort_first2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)(i);
+   }
+}
+static void generate_lines_ushort_first2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)(i);
+      (out+i)[1] = (ushort)(i+1);
+   }
+}
+static void generate_linestrip_ushort_first2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)(i);
+      (out+j)[1] = (ushort)(i+1);
+   }
+}
+static void generate_lineloop_ushort_first2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)(i);
+      (out+j)[1] = (ushort)(i+1);
+   }
+      (out+j)[0] = (ushort)(i);
+      (out+j)[1] = (ushort)(0);
+}
+static void generate_tris_ushort_first2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)(i);
+      (out+i)[1] = (ushort)(i+1);
+      (out+i)[2] = (ushort)(i+2);
+   }
+}
+static void generate_tristrip_ushort_first2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(i);
+      (out+j)[1] = (ushort)(i+1+(i&1));
+      (out+j)[2] = (ushort)(i+2-(i&1));
+   }
+}
+static void generate_trifan_ushort_first2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(0);
+      (out+j)[1] = (ushort)(i+1);
+      (out+j)[2] = (ushort)(i+2);
+   }
+}
+static void generate_quads_ushort_first2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)(i+0);
+      (out+j+0)[1] = (ushort)(i+1);
+      (out+j+0)[2] = (ushort)(i+3);
+      (out+j+3)[0] = (ushort)(i+1);
+      (out+j+3)[1] = (ushort)(i+2);
+      (out+j+3)[2] = (ushort)(i+3);
+   }
+}
+static void generate_quadstrip_ushort_first2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)(i+2);
+      (out+j+0)[1] = (ushort)(i+0);
+      (out+j+0)[2] = (ushort)(i+3);
+      (out+j+3)[0] = (ushort)(i+0);
+      (out+j+3)[1] = (ushort)(i+1);
+      (out+j+3)[2] = (ushort)(i+3);
+   }
+}
+static void generate_polygon_ushort_first2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(0);
+      (out+j)[1] = (ushort)(i+1);
+      (out+j)[2] = (ushort)(i+2);
+   }
+}
+static void generate_points_ushort_first2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)(i);
+   }
+}
+static void generate_lines_ushort_first2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)(i+1);
+      (out+i)[1] = (ushort)(i);
+   }
+}
+static void generate_linestrip_ushort_first2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)(i+1);
+      (out+j)[1] = (ushort)(i);
+   }
+}
+static void generate_lineloop_ushort_first2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)(i+1);
+      (out+j)[1] = (ushort)(i);
+   }
+      (out+j)[0] = (ushort)(0);
+      (out+j)[1] = (ushort)(i);
+}
+static void generate_tris_ushort_first2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)(i+1);
+      (out+i)[1] = (ushort)(i+2);
+      (out+i)[2] = (ushort)(i);
+   }
+}
+static void generate_tristrip_ushort_first2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(i+1+(i&1));
+      (out+j)[1] = (ushort)(i+2-(i&1));
+      (out+j)[2] = (ushort)(i);
+   }
+}
+static void generate_trifan_ushort_first2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(i+1);
+      (out+j)[1] = (ushort)(i+2);
+      (out+j)[2] = (ushort)(0);
+   }
+}
+static void generate_quads_ushort_first2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)(i+1);
+      (out+j+0)[1] = (ushort)(i+3);
+      (out+j+0)[2] = (ushort)(i+0);
+      (out+j+3)[0] = (ushort)(i+2);
+      (out+j+3)[1] = (ushort)(i+3);
+      (out+j+3)[2] = (ushort)(i+1);
+   }
+}
+static void generate_quadstrip_ushort_first2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)(i+0);
+      (out+j+0)[1] = (ushort)(i+3);
+      (out+j+0)[2] = (ushort)(i+2);
+      (out+j+3)[0] = (ushort)(i+1);
+      (out+j+3)[1] = (ushort)(i+3);
+      (out+j+3)[2] = (ushort)(i+0);
+   }
+}
+static void generate_polygon_ushort_first2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(i+1);
+      (out+j)[1] = (ushort)(i+2);
+      (out+j)[2] = (ushort)(0);
+   }
+}
+static void generate_points_ushort_last2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)(i);
+   }
+}
+static void generate_lines_ushort_last2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)(i+1);
+      (out+i)[1] = (ushort)(i);
+   }
+}
+static void generate_linestrip_ushort_last2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)(i+1);
+      (out+j)[1] = (ushort)(i);
+   }
+}
+static void generate_lineloop_ushort_last2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)(i+1);
+      (out+j)[1] = (ushort)(i);
+   }
+      (out+j)[0] = (ushort)(0);
+      (out+j)[1] = (ushort)(i);
+}
+static void generate_tris_ushort_last2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)(i+2);
+      (out+i)[1] = (ushort)(i);
+      (out+i)[2] = (ushort)(i+1);
+   }
+}
+static void generate_tristrip_ushort_last2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(i+2);
+      (out+j)[1] = (ushort)(i+(i&1));
+      (out+j)[2] = (ushort)(i+1-(i&1));
+   }
+}
+static void generate_trifan_ushort_last2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(i+2);
+      (out+j)[1] = (ushort)(0);
+      (out+j)[2] = (ushort)(i+1);
+   }
+}
+static void generate_quads_ushort_last2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)(i+3);
+      (out+j+0)[1] = (ushort)(i+0);
+      (out+j+0)[2] = (ushort)(i+1);
+      (out+j+3)[0] = (ushort)(i+3);
+      (out+j+3)[1] = (ushort)(i+1);
+      (out+j+3)[2] = (ushort)(i+2);
+   }
+}
+static void generate_quadstrip_ushort_last2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)(i+3);
+      (out+j+0)[1] = (ushort)(i+2);
+      (out+j+0)[2] = (ushort)(i+0);
+      (out+j+3)[0] = (ushort)(i+3);
+      (out+j+3)[1] = (ushort)(i+0);
+      (out+j+3)[2] = (ushort)(i+1);
+   }
+}
+static void generate_polygon_ushort_last2first(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(0);
+      (out+j)[1] = (ushort)(i+1);
+      (out+j)[2] = (ushort)(i+2);
+   }
+}
+static void generate_points_ushort_last2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)(i);
+   }
+}
+static void generate_lines_ushort_last2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)(i);
+      (out+i)[1] = (ushort)(i+1);
+   }
+}
+static void generate_linestrip_ushort_last2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)(i);
+      (out+j)[1] = (ushort)(i+1);
+   }
+}
+static void generate_lineloop_ushort_last2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)(i);
+      (out+j)[1] = (ushort)(i+1);
+   }
+      (out+j)[0] = (ushort)(i);
+      (out+j)[1] = (ushort)(0);
+}
+static void generate_tris_ushort_last2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)(i);
+      (out+i)[1] = (ushort)(i+1);
+      (out+i)[2] = (ushort)(i+2);
+   }
+}
+static void generate_tristrip_ushort_last2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(i+(i&1));
+      (out+j)[1] = (ushort)(i+1-(i&1));
+      (out+j)[2] = (ushort)(i+2);
+   }
+}
+static void generate_trifan_ushort_last2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(0);
+      (out+j)[1] = (ushort)(i+1);
+      (out+j)[2] = (ushort)(i+2);
+   }
+}
+static void generate_quads_ushort_last2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)(i+0);
+      (out+j+0)[1] = (ushort)(i+1);
+      (out+j+0)[2] = (ushort)(i+3);
+      (out+j+3)[0] = (ushort)(i+1);
+      (out+j+3)[1] = (ushort)(i+2);
+      (out+j+3)[2] = (ushort)(i+3);
+   }
+}
+static void generate_quadstrip_ushort_last2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)(i+2);
+      (out+j+0)[1] = (ushort)(i+0);
+      (out+j+0)[2] = (ushort)(i+3);
+      (out+j+3)[0] = (ushort)(i+0);
+      (out+j+3)[1] = (ushort)(i+1);
+      (out+j+3)[2] = (ushort)(i+3);
+   }
+}
+static void generate_polygon_ushort_last2last(
+    unsigned nr,
+    void *_out )
+{
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)(i+1);
+      (out+j)[1] = (ushort)(i+2);
+      (out+j)[2] = (ushort)(0);
+   }
+}
+static void generate_points_uint_first2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)(i);
+   }
+}
+static void generate_lines_uint_first2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)(i);
+      (out+i)[1] = (uint)(i+1);
+   }
+}
+static void generate_linestrip_uint_first2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)(i);
+      (out+j)[1] = (uint)(i+1);
+   }
+}
+static void generate_lineloop_uint_first2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)(i);
+      (out+j)[1] = (uint)(i+1);
+   }
+      (out+j)[0] = (uint)(i);
+      (out+j)[1] = (uint)(0);
+}
+static void generate_tris_uint_first2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)(i);
+      (out+i)[1] = (uint)(i+1);
+      (out+i)[2] = (uint)(i+2);
+   }
+}
+static void generate_tristrip_uint_first2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(i);
+      (out+j)[1] = (uint)(i+1+(i&1));
+      (out+j)[2] = (uint)(i+2-(i&1));
+   }
+}
+static void generate_trifan_uint_first2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(0);
+      (out+j)[1] = (uint)(i+1);
+      (out+j)[2] = (uint)(i+2);
+   }
+}
+static void generate_quads_uint_first2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)(i+0);
+      (out+j+0)[1] = (uint)(i+1);
+      (out+j+0)[2] = (uint)(i+3);
+      (out+j+3)[0] = (uint)(i+1);
+      (out+j+3)[1] = (uint)(i+2);
+      (out+j+3)[2] = (uint)(i+3);
+   }
+}
+static void generate_quadstrip_uint_first2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)(i+2);
+      (out+j+0)[1] = (uint)(i+0);
+      (out+j+0)[2] = (uint)(i+3);
+      (out+j+3)[0] = (uint)(i+0);
+      (out+j+3)[1] = (uint)(i+1);
+      (out+j+3)[2] = (uint)(i+3);
+   }
+}
+static void generate_polygon_uint_first2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(0);
+      (out+j)[1] = (uint)(i+1);
+      (out+j)[2] = (uint)(i+2);
+   }
+}
+static void generate_points_uint_first2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)(i);
+   }
+}
+static void generate_lines_uint_first2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)(i+1);
+      (out+i)[1] = (uint)(i);
+   }
+}
+static void generate_linestrip_uint_first2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)(i+1);
+      (out+j)[1] = (uint)(i);
+   }
+}
+static void generate_lineloop_uint_first2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)(i+1);
+      (out+j)[1] = (uint)(i);
+   }
+      (out+j)[0] = (uint)(0);
+      (out+j)[1] = (uint)(i);
+}
+static void generate_tris_uint_first2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)(i+1);
+      (out+i)[1] = (uint)(i+2);
+      (out+i)[2] = (uint)(i);
+   }
+}
+static void generate_tristrip_uint_first2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(i+1+(i&1));
+      (out+j)[1] = (uint)(i+2-(i&1));
+      (out+j)[2] = (uint)(i);
+   }
+}
+static void generate_trifan_uint_first2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(i+1);
+      (out+j)[1] = (uint)(i+2);
+      (out+j)[2] = (uint)(0);
+   }
+}
+static void generate_quads_uint_first2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)(i+1);
+      (out+j+0)[1] = (uint)(i+3);
+      (out+j+0)[2] = (uint)(i+0);
+      (out+j+3)[0] = (uint)(i+2);
+      (out+j+3)[1] = (uint)(i+3);
+      (out+j+3)[2] = (uint)(i+1);
+   }
+}
+static void generate_quadstrip_uint_first2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)(i+0);
+      (out+j+0)[1] = (uint)(i+3);
+      (out+j+0)[2] = (uint)(i+2);
+      (out+j+3)[0] = (uint)(i+1);
+      (out+j+3)[1] = (uint)(i+3);
+      (out+j+3)[2] = (uint)(i+0);
+   }
+}
+static void generate_polygon_uint_first2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(i+1);
+      (out+j)[1] = (uint)(i+2);
+      (out+j)[2] = (uint)(0);
+   }
+}
+static void generate_points_uint_last2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)(i);
+   }
+}
+static void generate_lines_uint_last2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)(i+1);
+      (out+i)[1] = (uint)(i);
+   }
+}
+static void generate_linestrip_uint_last2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)(i+1);
+      (out+j)[1] = (uint)(i);
+   }
+}
+static void generate_lineloop_uint_last2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)(i+1);
+      (out+j)[1] = (uint)(i);
+   }
+      (out+j)[0] = (uint)(0);
+      (out+j)[1] = (uint)(i);
+}
+static void generate_tris_uint_last2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)(i+2);
+      (out+i)[1] = (uint)(i);
+      (out+i)[2] = (uint)(i+1);
+   }
+}
+static void generate_tristrip_uint_last2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(i+2);
+      (out+j)[1] = (uint)(i+(i&1));
+      (out+j)[2] = (uint)(i+1-(i&1));
+   }
+}
+static void generate_trifan_uint_last2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(i+2);
+      (out+j)[1] = (uint)(0);
+      (out+j)[2] = (uint)(i+1);
+   }
+}
+static void generate_quads_uint_last2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)(i+3);
+      (out+j+0)[1] = (uint)(i+0);
+      (out+j+0)[2] = (uint)(i+1);
+      (out+j+3)[0] = (uint)(i+3);
+      (out+j+3)[1] = (uint)(i+1);
+      (out+j+3)[2] = (uint)(i+2);
+   }
+}
+static void generate_quadstrip_uint_last2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)(i+3);
+      (out+j+0)[1] = (uint)(i+2);
+      (out+j+0)[2] = (uint)(i+0);
+      (out+j+3)[0] = (uint)(i+3);
+      (out+j+3)[1] = (uint)(i+0);
+      (out+j+3)[2] = (uint)(i+1);
+   }
+}
+static void generate_polygon_uint_last2first(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(0);
+      (out+j)[1] = (uint)(i+1);
+      (out+j)[2] = (uint)(i+2);
+   }
+}
+static void generate_points_uint_last2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)(i);
+   }
+}
+static void generate_lines_uint_last2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)(i);
+      (out+i)[1] = (uint)(i+1);
+   }
+}
+static void generate_linestrip_uint_last2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)(i);
+      (out+j)[1] = (uint)(i+1);
+   }
+}
+static void generate_lineloop_uint_last2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)(i);
+      (out+j)[1] = (uint)(i+1);
+   }
+      (out+j)[0] = (uint)(i);
+      (out+j)[1] = (uint)(0);
+}
+static void generate_tris_uint_last2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)(i);
+      (out+i)[1] = (uint)(i+1);
+      (out+i)[2] = (uint)(i+2);
+   }
+}
+static void generate_tristrip_uint_last2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(i+(i&1));
+      (out+j)[1] = (uint)(i+1-(i&1));
+      (out+j)[2] = (uint)(i+2);
+   }
+}
+static void generate_trifan_uint_last2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(0);
+      (out+j)[1] = (uint)(i+1);
+      (out+j)[2] = (uint)(i+2);
+   }
+}
+static void generate_quads_uint_last2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)(i+0);
+      (out+j+0)[1] = (uint)(i+1);
+      (out+j+0)[2] = (uint)(i+3);
+      (out+j+3)[0] = (uint)(i+1);
+      (out+j+3)[1] = (uint)(i+2);
+      (out+j+3)[2] = (uint)(i+3);
+   }
+}
+static void generate_quadstrip_uint_last2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)(i+2);
+      (out+j+0)[1] = (uint)(i+0);
+      (out+j+0)[2] = (uint)(i+3);
+      (out+j+3)[0] = (uint)(i+0);
+      (out+j+3)[1] = (uint)(i+1);
+      (out+j+3)[2] = (uint)(i+3);
+   }
+}
+static void generate_polygon_uint_last2last(
+    unsigned nr,
+    void *_out )
+{
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)(i+1);
+      (out+j)[1] = (uint)(i+2);
+      (out+j)[2] = (uint)(0);
+   }
+}
+static void translate_points_ubyte2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_ubyte2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_linestrip_ubyte2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_lineloop_ubyte2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[0];
+}
+static void translate_tris_ubyte2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+      (out+i)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_tristrip_ubyte2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1+(i&1)];
+      (out+j)[2] = (ushort)in[i+2-(i&1)];
+   }
+}
+static void translate_trifan_ubyte2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_quads_ubyte2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+0];
+      (out+j+0)[1] = (ushort)in[i+1];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+1];
+      (out+j+3)[1] = (ushort)in[i+2];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_quadstrip_ubyte2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+2];
+      (out+j+0)[1] = (ushort)in[i+0];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+0];
+      (out+j+3)[1] = (ushort)in[i+1];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_polygon_ubyte2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_points_ubyte2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_ubyte2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i+1];
+      (out+i)[1] = (ushort)in[i];
+   }
+}
+static void translate_linestrip_ubyte2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+}
+static void translate_lineloop_ubyte2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i];
+}
+static void translate_tris_ubyte2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i+1];
+      (out+i)[1] = (ushort)in[i+2];
+      (out+i)[2] = (ushort)in[i];
+   }
+}
+static void translate_tristrip_ubyte2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1+(i&1)];
+      (out+j)[1] = (ushort)in[i+2-(i&1)];
+      (out+j)[2] = (ushort)in[i];
+   }
+}
+static void translate_trifan_ubyte2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i+2];
+      (out+j)[2] = (ushort)in[0];
+   }
+}
+static void translate_quads_ubyte2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+1];
+      (out+j+0)[1] = (ushort)in[i+3];
+      (out+j+0)[2] = (ushort)in[i+0];
+      (out+j+3)[0] = (ushort)in[i+2];
+      (out+j+3)[1] = (ushort)in[i+3];
+      (out+j+3)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_quadstrip_ubyte2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+0];
+      (out+j+0)[1] = (ushort)in[i+3];
+      (out+j+0)[2] = (ushort)in[i+2];
+      (out+j+3)[0] = (ushort)in[i+1];
+      (out+j+3)[1] = (ushort)in[i+3];
+      (out+j+3)[2] = (ushort)in[i+0];
+   }
+}
+static void translate_polygon_ubyte2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i+2];
+      (out+j)[2] = (ushort)in[0];
+   }
+}
+static void translate_points_ubyte2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_ubyte2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i+1];
+      (out+i)[1] = (ushort)in[i];
+   }
+}
+static void translate_linestrip_ubyte2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+}
+static void translate_lineloop_ubyte2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i];
+}
+static void translate_tris_ubyte2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i+2];
+      (out+i)[1] = (ushort)in[i];
+      (out+i)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_tristrip_ubyte2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+2];
+      (out+j)[1] = (ushort)in[i+(i&1)];
+      (out+j)[2] = (ushort)in[i+1-(i&1)];
+   }
+}
+static void translate_trifan_ubyte2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+2];
+      (out+j)[1] = (ushort)in[0];
+      (out+j)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_quads_ubyte2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+3];
+      (out+j+0)[1] = (ushort)in[i+0];
+      (out+j+0)[2] = (ushort)in[i+1];
+      (out+j+3)[0] = (ushort)in[i+3];
+      (out+j+3)[1] = (ushort)in[i+1];
+      (out+j+3)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_quadstrip_ubyte2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+3];
+      (out+j+0)[1] = (ushort)in[i+2];
+      (out+j+0)[2] = (ushort)in[i+0];
+      (out+j+3)[0] = (ushort)in[i+3];
+      (out+j+3)[1] = (ushort)in[i+0];
+      (out+j+3)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_polygon_ubyte2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_points_ubyte2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_ubyte2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_linestrip_ubyte2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_lineloop_ubyte2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[0];
+}
+static void translate_tris_ubyte2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+      (out+i)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_tristrip_ubyte2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+(i&1)];
+      (out+j)[1] = (ushort)in[i+1-(i&1)];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_trifan_ubyte2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_quads_ubyte2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+0];
+      (out+j+0)[1] = (ushort)in[i+1];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+1];
+      (out+j+3)[1] = (ushort)in[i+2];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_quadstrip_ubyte2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+2];
+      (out+j+0)[1] = (ushort)in[i+0];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+0];
+      (out+j+3)[1] = (ushort)in[i+1];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_polygon_ubyte2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i+2];
+      (out+j)[2] = (ushort)in[0];
+   }
+}
+static void translate_points_ubyte2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_ubyte2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+   }
+}
+static void translate_linestrip_ubyte2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+}
+static void translate_lineloop_ubyte2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[0];
+}
+static void translate_tris_ubyte2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+      (out+i)[2] = (uint)in[i+2];
+   }
+}
+static void translate_tristrip_ubyte2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1+(i&1)];
+      (out+j)[2] = (uint)in[i+2-(i&1)];
+   }
+}
+static void translate_trifan_ubyte2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_quads_ubyte2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+0];
+      (out+j+0)[1] = (uint)in[i+1];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+1];
+      (out+j+3)[1] = (uint)in[i+2];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_quadstrip_ubyte2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+2];
+      (out+j+0)[1] = (uint)in[i+0];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+0];
+      (out+j+3)[1] = (uint)in[i+1];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_polygon_ubyte2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_points_ubyte2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_ubyte2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i+1];
+      (out+i)[1] = (uint)in[i];
+   }
+}
+static void translate_linestrip_ubyte2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+}
+static void translate_lineloop_ubyte2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i];
+}
+static void translate_tris_ubyte2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i+1];
+      (out+i)[1] = (uint)in[i+2];
+      (out+i)[2] = (uint)in[i];
+   }
+}
+static void translate_tristrip_ubyte2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1+(i&1)];
+      (out+j)[1] = (uint)in[i+2-(i&1)];
+      (out+j)[2] = (uint)in[i];
+   }
+}
+static void translate_trifan_ubyte2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i+2];
+      (out+j)[2] = (uint)in[0];
+   }
+}
+static void translate_quads_ubyte2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+1];
+      (out+j+0)[1] = (uint)in[i+3];
+      (out+j+0)[2] = (uint)in[i+0];
+      (out+j+3)[0] = (uint)in[i+2];
+      (out+j+3)[1] = (uint)in[i+3];
+      (out+j+3)[2] = (uint)in[i+1];
+   }
+}
+static void translate_quadstrip_ubyte2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+0];
+      (out+j+0)[1] = (uint)in[i+3];
+      (out+j+0)[2] = (uint)in[i+2];
+      (out+j+3)[0] = (uint)in[i+1];
+      (out+j+3)[1] = (uint)in[i+3];
+      (out+j+3)[2] = (uint)in[i+0];
+   }
+}
+static void translate_polygon_ubyte2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i+2];
+      (out+j)[2] = (uint)in[0];
+   }
+}
+static void translate_points_ubyte2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_ubyte2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i+1];
+      (out+i)[1] = (uint)in[i];
+   }
+}
+static void translate_linestrip_ubyte2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+}
+static void translate_lineloop_ubyte2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i];
+}
+static void translate_tris_ubyte2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i+2];
+      (out+i)[1] = (uint)in[i];
+      (out+i)[2] = (uint)in[i+1];
+   }
+}
+static void translate_tristrip_ubyte2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+2];
+      (out+j)[1] = (uint)in[i+(i&1)];
+      (out+j)[2] = (uint)in[i+1-(i&1)];
+   }
+}
+static void translate_trifan_ubyte2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+2];
+      (out+j)[1] = (uint)in[0];
+      (out+j)[2] = (uint)in[i+1];
+   }
+}
+static void translate_quads_ubyte2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+3];
+      (out+j+0)[1] = (uint)in[i+0];
+      (out+j+0)[2] = (uint)in[i+1];
+      (out+j+3)[0] = (uint)in[i+3];
+      (out+j+3)[1] = (uint)in[i+1];
+      (out+j+3)[2] = (uint)in[i+2];
+   }
+}
+static void translate_quadstrip_ubyte2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+3];
+      (out+j+0)[1] = (uint)in[i+2];
+      (out+j+0)[2] = (uint)in[i+0];
+      (out+j+3)[0] = (uint)in[i+3];
+      (out+j+3)[1] = (uint)in[i+0];
+      (out+j+3)[2] = (uint)in[i+1];
+   }
+}
+static void translate_polygon_ubyte2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_points_ubyte2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_ubyte2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+   }
+}
+static void translate_linestrip_ubyte2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+}
+static void translate_lineloop_ubyte2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[0];
+}
+static void translate_tris_ubyte2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+      (out+i)[2] = (uint)in[i+2];
+   }
+}
+static void translate_tristrip_ubyte2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+(i&1)];
+      (out+j)[1] = (uint)in[i+1-(i&1)];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_trifan_ubyte2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_quads_ubyte2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+0];
+      (out+j+0)[1] = (uint)in[i+1];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+1];
+      (out+j+3)[1] = (uint)in[i+2];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_quadstrip_ubyte2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+2];
+      (out+j+0)[1] = (uint)in[i+0];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+0];
+      (out+j+3)[1] = (uint)in[i+1];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_polygon_ubyte2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ubyte*in = (const ubyte*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i+2];
+      (out+j)[2] = (uint)in[0];
+   }
+}
+static void translate_points_ushort2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_ushort2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_linestrip_ushort2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_lineloop_ushort2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[0];
+}
+static void translate_tris_ushort2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+      (out+i)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_tristrip_ushort2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1+(i&1)];
+      (out+j)[2] = (ushort)in[i+2-(i&1)];
+   }
+}
+static void translate_trifan_ushort2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_quads_ushort2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+0];
+      (out+j+0)[1] = (ushort)in[i+1];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+1];
+      (out+j+3)[1] = (ushort)in[i+2];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_quadstrip_ushort2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+2];
+      (out+j+0)[1] = (ushort)in[i+0];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+0];
+      (out+j+3)[1] = (ushort)in[i+1];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_polygon_ushort2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_points_ushort2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_ushort2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i+1];
+      (out+i)[1] = (ushort)in[i];
+   }
+}
+static void translate_linestrip_ushort2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+}
+static void translate_lineloop_ushort2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i];
+}
+static void translate_tris_ushort2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i+1];
+      (out+i)[1] = (ushort)in[i+2];
+      (out+i)[2] = (ushort)in[i];
+   }
+}
+static void translate_tristrip_ushort2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1+(i&1)];
+      (out+j)[1] = (ushort)in[i+2-(i&1)];
+      (out+j)[2] = (ushort)in[i];
+   }
+}
+static void translate_trifan_ushort2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i+2];
+      (out+j)[2] = (ushort)in[0];
+   }
+}
+static void translate_quads_ushort2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+1];
+      (out+j+0)[1] = (ushort)in[i+3];
+      (out+j+0)[2] = (ushort)in[i+0];
+      (out+j+3)[0] = (ushort)in[i+2];
+      (out+j+3)[1] = (ushort)in[i+3];
+      (out+j+3)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_quadstrip_ushort2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+0];
+      (out+j+0)[1] = (ushort)in[i+3];
+      (out+j+0)[2] = (ushort)in[i+2];
+      (out+j+3)[0] = (ushort)in[i+1];
+      (out+j+3)[1] = (ushort)in[i+3];
+      (out+j+3)[2] = (ushort)in[i+0];
+   }
+}
+static void translate_polygon_ushort2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i+2];
+      (out+j)[2] = (ushort)in[0];
+   }
+}
+static void translate_points_ushort2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_ushort2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i+1];
+      (out+i)[1] = (ushort)in[i];
+   }
+}
+static void translate_linestrip_ushort2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+}
+static void translate_lineloop_ushort2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i];
+}
+static void translate_tris_ushort2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i+2];
+      (out+i)[1] = (ushort)in[i];
+      (out+i)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_tristrip_ushort2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+2];
+      (out+j)[1] = (ushort)in[i+(i&1)];
+      (out+j)[2] = (ushort)in[i+1-(i&1)];
+   }
+}
+static void translate_trifan_ushort2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+2];
+      (out+j)[1] = (ushort)in[0];
+      (out+j)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_quads_ushort2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+3];
+      (out+j+0)[1] = (ushort)in[i+0];
+      (out+j+0)[2] = (ushort)in[i+1];
+      (out+j+3)[0] = (ushort)in[i+3];
+      (out+j+3)[1] = (ushort)in[i+1];
+      (out+j+3)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_quadstrip_ushort2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+3];
+      (out+j+0)[1] = (ushort)in[i+2];
+      (out+j+0)[2] = (ushort)in[i+0];
+      (out+j+3)[0] = (ushort)in[i+3];
+      (out+j+3)[1] = (ushort)in[i+0];
+      (out+j+3)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_polygon_ushort2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_points_ushort2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_ushort2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_linestrip_ushort2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_lineloop_ushort2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[0];
+}
+static void translate_tris_ushort2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+      (out+i)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_tristrip_ushort2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+(i&1)];
+      (out+j)[1] = (ushort)in[i+1-(i&1)];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_trifan_ushort2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_quads_ushort2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+0];
+      (out+j+0)[1] = (ushort)in[i+1];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+1];
+      (out+j+3)[1] = (ushort)in[i+2];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_quadstrip_ushort2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+2];
+      (out+j+0)[1] = (ushort)in[i+0];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+0];
+      (out+j+3)[1] = (ushort)in[i+1];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_polygon_ushort2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i+2];
+      (out+j)[2] = (ushort)in[0];
+   }
+}
+static void translate_points_ushort2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_ushort2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+   }
+}
+static void translate_linestrip_ushort2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+}
+static void translate_lineloop_ushort2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[0];
+}
+static void translate_tris_ushort2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+      (out+i)[2] = (uint)in[i+2];
+   }
+}
+static void translate_tristrip_ushort2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1+(i&1)];
+      (out+j)[2] = (uint)in[i+2-(i&1)];
+   }
+}
+static void translate_trifan_ushort2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_quads_ushort2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+0];
+      (out+j+0)[1] = (uint)in[i+1];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+1];
+      (out+j+3)[1] = (uint)in[i+2];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_quadstrip_ushort2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+2];
+      (out+j+0)[1] = (uint)in[i+0];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+0];
+      (out+j+3)[1] = (uint)in[i+1];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_polygon_ushort2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_points_ushort2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_ushort2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i+1];
+      (out+i)[1] = (uint)in[i];
+   }
+}
+static void translate_linestrip_ushort2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+}
+static void translate_lineloop_ushort2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i];
+}
+static void translate_tris_ushort2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i+1];
+      (out+i)[1] = (uint)in[i+2];
+      (out+i)[2] = (uint)in[i];
+   }
+}
+static void translate_tristrip_ushort2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1+(i&1)];
+      (out+j)[1] = (uint)in[i+2-(i&1)];
+      (out+j)[2] = (uint)in[i];
+   }
+}
+static void translate_trifan_ushort2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i+2];
+      (out+j)[2] = (uint)in[0];
+   }
+}
+static void translate_quads_ushort2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+1];
+      (out+j+0)[1] = (uint)in[i+3];
+      (out+j+0)[2] = (uint)in[i+0];
+      (out+j+3)[0] = (uint)in[i+2];
+      (out+j+3)[1] = (uint)in[i+3];
+      (out+j+3)[2] = (uint)in[i+1];
+   }
+}
+static void translate_quadstrip_ushort2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+0];
+      (out+j+0)[1] = (uint)in[i+3];
+      (out+j+0)[2] = (uint)in[i+2];
+      (out+j+3)[0] = (uint)in[i+1];
+      (out+j+3)[1] = (uint)in[i+3];
+      (out+j+3)[2] = (uint)in[i+0];
+   }
+}
+static void translate_polygon_ushort2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i+2];
+      (out+j)[2] = (uint)in[0];
+   }
+}
+static void translate_points_ushort2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_ushort2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i+1];
+      (out+i)[1] = (uint)in[i];
+   }
+}
+static void translate_linestrip_ushort2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+}
+static void translate_lineloop_ushort2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i];
+}
+static void translate_tris_ushort2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i+2];
+      (out+i)[1] = (uint)in[i];
+      (out+i)[2] = (uint)in[i+1];
+   }
+}
+static void translate_tristrip_ushort2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+2];
+      (out+j)[1] = (uint)in[i+(i&1)];
+      (out+j)[2] = (uint)in[i+1-(i&1)];
+   }
+}
+static void translate_trifan_ushort2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+2];
+      (out+j)[1] = (uint)in[0];
+      (out+j)[2] = (uint)in[i+1];
+   }
+}
+static void translate_quads_ushort2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+3];
+      (out+j+0)[1] = (uint)in[i+0];
+      (out+j+0)[2] = (uint)in[i+1];
+      (out+j+3)[0] = (uint)in[i+3];
+      (out+j+3)[1] = (uint)in[i+1];
+      (out+j+3)[2] = (uint)in[i+2];
+   }
+}
+static void translate_quadstrip_ushort2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+3];
+      (out+j+0)[1] = (uint)in[i+2];
+      (out+j+0)[2] = (uint)in[i+0];
+      (out+j+3)[0] = (uint)in[i+3];
+      (out+j+3)[1] = (uint)in[i+0];
+      (out+j+3)[2] = (uint)in[i+1];
+   }
+}
+static void translate_polygon_ushort2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_points_ushort2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_ushort2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+   }
+}
+static void translate_linestrip_ushort2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+}
+static void translate_lineloop_ushort2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[0];
+}
+static void translate_tris_ushort2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+      (out+i)[2] = (uint)in[i+2];
+   }
+}
+static void translate_tristrip_ushort2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+(i&1)];
+      (out+j)[1] = (uint)in[i+1-(i&1)];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_trifan_ushort2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_quads_ushort2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+0];
+      (out+j+0)[1] = (uint)in[i+1];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+1];
+      (out+j+3)[1] = (uint)in[i+2];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_quadstrip_ushort2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+2];
+      (out+j+0)[1] = (uint)in[i+0];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+0];
+      (out+j+3)[1] = (uint)in[i+1];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_polygon_ushort2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const ushort*in = (const ushort*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i+2];
+      (out+j)[2] = (uint)in[0];
+   }
+}
+static void translate_points_uint2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_uint2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_linestrip_uint2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_lineloop_uint2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[0];
+}
+static void translate_tris_uint2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+      (out+i)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_tristrip_uint2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1+(i&1)];
+      (out+j)[2] = (ushort)in[i+2-(i&1)];
+   }
+}
+static void translate_trifan_uint2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_quads_uint2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+0];
+      (out+j+0)[1] = (ushort)in[i+1];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+1];
+      (out+j+3)[1] = (ushort)in[i+2];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_quadstrip_uint2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+2];
+      (out+j+0)[1] = (ushort)in[i+0];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+0];
+      (out+j+3)[1] = (ushort)in[i+1];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_polygon_uint2ushort_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_points_uint2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_uint2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i+1];
+      (out+i)[1] = (ushort)in[i];
+   }
+}
+static void translate_linestrip_uint2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+}
+static void translate_lineloop_uint2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i];
+}
+static void translate_tris_uint2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i+1];
+      (out+i)[1] = (ushort)in[i+2];
+      (out+i)[2] = (ushort)in[i];
+   }
+}
+static void translate_tristrip_uint2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1+(i&1)];
+      (out+j)[1] = (ushort)in[i+2-(i&1)];
+      (out+j)[2] = (ushort)in[i];
+   }
+}
+static void translate_trifan_uint2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i+2];
+      (out+j)[2] = (ushort)in[0];
+   }
+}
+static void translate_quads_uint2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+1];
+      (out+j+0)[1] = (ushort)in[i+3];
+      (out+j+0)[2] = (ushort)in[i+0];
+      (out+j+3)[0] = (ushort)in[i+2];
+      (out+j+3)[1] = (ushort)in[i+3];
+      (out+j+3)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_quadstrip_uint2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+0];
+      (out+j+0)[1] = (ushort)in[i+3];
+      (out+j+0)[2] = (ushort)in[i+2];
+      (out+j+3)[0] = (ushort)in[i+1];
+      (out+j+3)[1] = (ushort)in[i+3];
+      (out+j+3)[2] = (ushort)in[i+0];
+   }
+}
+static void translate_polygon_uint2ushort_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i+2];
+      (out+j)[2] = (ushort)in[0];
+   }
+}
+static void translate_points_uint2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_uint2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i+1];
+      (out+i)[1] = (ushort)in[i];
+   }
+}
+static void translate_linestrip_uint2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+}
+static void translate_lineloop_uint2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i];
+   }
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i];
+}
+static void translate_tris_uint2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i+2];
+      (out+i)[1] = (ushort)in[i];
+      (out+i)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_tristrip_uint2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+2];
+      (out+j)[1] = (ushort)in[i+(i&1)];
+      (out+j)[2] = (ushort)in[i+1-(i&1)];
+   }
+}
+static void translate_trifan_uint2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+2];
+      (out+j)[1] = (ushort)in[0];
+      (out+j)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_quads_uint2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+3];
+      (out+j+0)[1] = (ushort)in[i+0];
+      (out+j+0)[2] = (ushort)in[i+1];
+      (out+j+3)[0] = (ushort)in[i+3];
+      (out+j+3)[1] = (ushort)in[i+1];
+      (out+j+3)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_quadstrip_uint2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+3];
+      (out+j+0)[1] = (ushort)in[i+2];
+      (out+j+0)[2] = (ushort)in[i+0];
+      (out+j+3)[0] = (ushort)in[i+3];
+      (out+j+3)[1] = (ushort)in[i+0];
+      (out+j+3)[2] = (ushort)in[i+1];
+   }
+}
+static void translate_polygon_uint2ushort_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_points_uint2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (ushort)in[i];
+   }
+}
+static void translate_lines_uint2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_linestrip_uint2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+}
+static void translate_lineloop_uint2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[i+1];
+   }
+      (out+j)[0] = (ushort)in[i];
+      (out+j)[1] = (ushort)in[0];
+}
+static void translate_tris_uint2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (ushort)in[i];
+      (out+i)[1] = (ushort)in[i+1];
+      (out+i)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_tristrip_uint2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+(i&1)];
+      (out+j)[1] = (ushort)in[i+1-(i&1)];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_trifan_uint2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[0];
+      (out+j)[1] = (ushort)in[i+1];
+      (out+j)[2] = (ushort)in[i+2];
+   }
+}
+static void translate_quads_uint2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (ushort)in[i+0];
+      (out+j+0)[1] = (ushort)in[i+1];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+1];
+      (out+j+3)[1] = (ushort)in[i+2];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_quadstrip_uint2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (ushort)in[i+2];
+      (out+j+0)[1] = (ushort)in[i+0];
+      (out+j+0)[2] = (ushort)in[i+3];
+      (out+j+3)[0] = (ushort)in[i+0];
+      (out+j+3)[1] = (ushort)in[i+1];
+      (out+j+3)[2] = (ushort)in[i+3];
+   }
+}
+static void translate_polygon_uint2ushort_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  ushort *out = (ushort*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (ushort)in[i+1];
+      (out+j)[1] = (ushort)in[i+2];
+      (out+j)[2] = (ushort)in[0];
+   }
+}
+static void translate_points_uint2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_uint2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+   }
+}
+static void translate_linestrip_uint2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+}
+static void translate_lineloop_uint2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[0];
+}
+static void translate_tris_uint2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+      (out+i)[2] = (uint)in[i+2];
+   }
+}
+static void translate_tristrip_uint2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1+(i&1)];
+      (out+j)[2] = (uint)in[i+2-(i&1)];
+   }
+}
+static void translate_trifan_uint2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_quads_uint2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+0];
+      (out+j+0)[1] = (uint)in[i+1];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+1];
+      (out+j+3)[1] = (uint)in[i+2];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_quadstrip_uint2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+2];
+      (out+j+0)[1] = (uint)in[i+0];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+0];
+      (out+j+3)[1] = (uint)in[i+1];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_polygon_uint2uint_first2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_points_uint2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_uint2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i+1];
+      (out+i)[1] = (uint)in[i];
+   }
+}
+static void translate_linestrip_uint2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+}
+static void translate_lineloop_uint2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i];
+}
+static void translate_tris_uint2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i+1];
+      (out+i)[1] = (uint)in[i+2];
+      (out+i)[2] = (uint)in[i];
+   }
+}
+static void translate_tristrip_uint2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1+(i&1)];
+      (out+j)[1] = (uint)in[i+2-(i&1)];
+      (out+j)[2] = (uint)in[i];
+   }
+}
+static void translate_trifan_uint2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i+2];
+      (out+j)[2] = (uint)in[0];
+   }
+}
+static void translate_quads_uint2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+1];
+      (out+j+0)[1] = (uint)in[i+3];
+      (out+j+0)[2] = (uint)in[i+0];
+      (out+j+3)[0] = (uint)in[i+2];
+      (out+j+3)[1] = (uint)in[i+3];
+      (out+j+3)[2] = (uint)in[i+1];
+   }
+}
+static void translate_quadstrip_uint2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+0];
+      (out+j+0)[1] = (uint)in[i+3];
+      (out+j+0)[2] = (uint)in[i+2];
+      (out+j+3)[0] = (uint)in[i+1];
+      (out+j+3)[1] = (uint)in[i+3];
+      (out+j+3)[2] = (uint)in[i+0];
+   }
+}
+static void translate_polygon_uint2uint_first2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i+2];
+      (out+j)[2] = (uint)in[0];
+   }
+}
+static void translate_points_uint2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_uint2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i+1];
+      (out+i)[1] = (uint)in[i];
+   }
+}
+static void translate_linestrip_uint2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+}
+static void translate_lineloop_uint2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i];
+   }
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i];
+}
+static void translate_tris_uint2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i+2];
+      (out+i)[1] = (uint)in[i];
+      (out+i)[2] = (uint)in[i+1];
+   }
+}
+static void translate_tristrip_uint2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+2];
+      (out+j)[1] = (uint)in[i+(i&1)];
+      (out+j)[2] = (uint)in[i+1-(i&1)];
+   }
+}
+static void translate_trifan_uint2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+2];
+      (out+j)[1] = (uint)in[0];
+      (out+j)[2] = (uint)in[i+1];
+   }
+}
+static void translate_quads_uint2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+3];
+      (out+j+0)[1] = (uint)in[i+0];
+      (out+j+0)[2] = (uint)in[i+1];
+      (out+j+3)[0] = (uint)in[i+3];
+      (out+j+3)[1] = (uint)in[i+1];
+      (out+j+3)[2] = (uint)in[i+2];
+   }
+}
+static void translate_quadstrip_uint2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+3];
+      (out+j+0)[1] = (uint)in[i+2];
+      (out+j+0)[2] = (uint)in[i+0];
+      (out+j+3)[0] = (uint)in[i+3];
+      (out+j+3)[1] = (uint)in[i+0];
+      (out+j+3)[2] = (uint)in[i+1];
+   }
+}
+static void translate_polygon_uint2uint_last2first(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_points_uint2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i++) { 
+      (out+i)[0] = (uint)in[i];
+   }
+}
+static void translate_lines_uint2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=2) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+   }
+}
+static void translate_linestrip_uint2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+}
+static void translate_lineloop_uint2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr - 2; j+=2, i++) { 
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[i+1];
+   }
+      (out+j)[0] = (uint)in[i];
+      (out+j)[1] = (uint)in[0];
+}
+static void translate_tris_uint2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (i = 0; i < nr; i+=3) { 
+      (out+i)[0] = (uint)in[i];
+      (out+i)[1] = (uint)in[i+1];
+      (out+i)[2] = (uint)in[i+2];
+   }
+}
+static void translate_tristrip_uint2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+(i&1)];
+      (out+j)[1] = (uint)in[i+1-(i&1)];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_trifan_uint2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[0];
+      (out+j)[1] = (uint)in[i+1];
+      (out+j)[2] = (uint)in[i+2];
+   }
+}
+static void translate_quads_uint2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=4) { 
+      (out+j+0)[0] = (uint)in[i+0];
+      (out+j+0)[1] = (uint)in[i+1];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+1];
+      (out+j+3)[1] = (uint)in[i+2];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_quadstrip_uint2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=6, i+=2) { 
+      (out+j+0)[0] = (uint)in[i+2];
+      (out+j+0)[1] = (uint)in[i+0];
+      (out+j+0)[2] = (uint)in[i+3];
+      (out+j+3)[0] = (uint)in[i+0];
+      (out+j+3)[1] = (uint)in[i+1];
+      (out+j+3)[2] = (uint)in[i+3];
+   }
+}
+static void translate_polygon_uint2uint_last2last(
+    const void * _in,
+    unsigned nr,
+    void *_out )
+{
+  const uint*in = (const uint*)_in;
+  uint *out = (uint*)_out;
+  unsigned i, j;
+  (void)j;
+  for (j = i = 0; j < nr; j+=3, i++) { 
+      (out+j)[0] = (uint)in[i+1];
+      (out+j)[1] = (uint)in[i+2];
+      (out+j)[2] = (uint)in[0];
+   }
+}
+void u_index_init( void )
+{
+  static int firsttime = 1;
+  if (!firsttime) return;
+  firsttime = 0;
+generate[OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_POINTS] = generate_points_ushort_first2first;
+generate[OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINES] = generate_lines_ushort_first2first;
+generate[OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = generate_linestrip_ushort_first2first;
+generate[OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = generate_lineloop_ushort_first2first;
+generate[OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLES] = generate_tris_ushort_first2first;
+generate[OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = generate_trifan_ushort_first2first;
+generate[OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = generate_tristrip_ushort_first2first;
+generate[OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUADS] = generate_quads_ushort_first2first;
+generate[OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = generate_quadstrip_ushort_first2first;
+generate[OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_POLYGON] = generate_polygon_ushort_first2first;
+generate[OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_POINTS] = generate_points_ushort_first2last;
+generate[OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINES] = generate_lines_ushort_first2last;
+generate[OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_STRIP] = generate_linestrip_ushort_first2last;
+generate[OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_LOOP] = generate_lineloop_ushort_first2last;
+generate[OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLES] = generate_tris_ushort_first2last;
+generate[OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = generate_trifan_ushort_first2last;
+generate[OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = generate_tristrip_ushort_first2last;
+generate[OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_QUADS] = generate_quads_ushort_first2last;
+generate[OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = generate_quadstrip_ushort_first2last;
+generate[OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_POLYGON] = generate_polygon_ushort_first2last;
+generate[OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_POINTS] = generate_points_ushort_last2first;
+generate[OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINES] = generate_lines_ushort_last2first;
+generate[OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = generate_linestrip_ushort_last2first;
+generate[OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = generate_lineloop_ushort_last2first;
+generate[OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLES] = generate_tris_ushort_last2first;
+generate[OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = generate_trifan_ushort_last2first;
+generate[OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = generate_tristrip_ushort_last2first;
+generate[OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_QUADS] = generate_quads_ushort_last2first;
+generate[OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = generate_quadstrip_ushort_last2first;
+generate[OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_POLYGON] = generate_polygon_ushort_last2first;
+generate[OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_POINTS] = generate_points_ushort_last2last;
+generate[OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINES] = generate_lines_ushort_last2last;
+generate[OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_STRIP] = generate_linestrip_ushort_last2last;
+generate[OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_LOOP] = generate_lineloop_ushort_last2last;
+generate[OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLES] = generate_tris_ushort_last2last;
+generate[OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = generate_trifan_ushort_last2last;
+generate[OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = generate_tristrip_ushort_last2last;
+generate[OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_QUADS] = generate_quads_ushort_last2last;
+generate[OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = generate_quadstrip_ushort_last2last;
+generate[OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_POLYGON] = generate_polygon_ushort_last2last;
+generate[OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_POINTS] = generate_points_uint_first2first;
+generate[OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINES] = generate_lines_uint_first2first;
+generate[OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = generate_linestrip_uint_first2first;
+generate[OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = generate_lineloop_uint_first2first;
+generate[OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLES] = generate_tris_uint_first2first;
+generate[OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = generate_trifan_uint_first2first;
+generate[OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = generate_tristrip_uint_first2first;
+generate[OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUADS] = generate_quads_uint_first2first;
+generate[OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = generate_quadstrip_uint_first2first;
+generate[OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_POLYGON] = generate_polygon_uint_first2first;
+generate[OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_POINTS] = generate_points_uint_first2last;
+generate[OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINES] = generate_lines_uint_first2last;
+generate[OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_STRIP] = generate_linestrip_uint_first2last;
+generate[OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_LOOP] = generate_lineloop_uint_first2last;
+generate[OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLES] = generate_tris_uint_first2last;
+generate[OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = generate_trifan_uint_first2last;
+generate[OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = generate_tristrip_uint_first2last;
+generate[OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_QUADS] = generate_quads_uint_first2last;
+generate[OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = generate_quadstrip_uint_first2last;
+generate[OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_POLYGON] = generate_polygon_uint_first2last;
+generate[OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_POINTS] = generate_points_uint_last2first;
+generate[OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINES] = generate_lines_uint_last2first;
+generate[OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = generate_linestrip_uint_last2first;
+generate[OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = generate_lineloop_uint_last2first;
+generate[OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLES] = generate_tris_uint_last2first;
+generate[OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = generate_trifan_uint_last2first;
+generate[OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = generate_tristrip_uint_last2first;
+generate[OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_QUADS] = generate_quads_uint_last2first;
+generate[OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = generate_quadstrip_uint_last2first;
+generate[OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_POLYGON] = generate_polygon_uint_last2first;
+generate[OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_POINTS] = generate_points_uint_last2last;
+generate[OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINES] = generate_lines_uint_last2last;
+generate[OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_STRIP] = generate_linestrip_uint_last2last;
+generate[OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_LOOP] = generate_lineloop_uint_last2last;
+generate[OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLES] = generate_tris_uint_last2last;
+generate[OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = generate_trifan_uint_last2last;
+generate[OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = generate_tristrip_uint_last2last;
+generate[OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_QUADS] = generate_quads_uint_last2last;
+generate[OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = generate_quadstrip_uint_last2last;
+generate[OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_POLYGON] = generate_polygon_uint_last2last;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_ubyte2ushort_first2first;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_ubyte2ushort_first2first;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ubyte2ushort_first2first;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ubyte2ushort_first2first;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_ubyte2ushort_first2first;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ubyte2ushort_first2first;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ubyte2ushort_first2first;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_ubyte2ushort_first2first;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ubyte2ushort_first2first;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_ubyte2ushort_first2first;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_ubyte2ushort_first2last;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_ubyte2ushort_first2last;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ubyte2ushort_first2last;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ubyte2ushort_first2last;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_ubyte2ushort_first2last;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ubyte2ushort_first2last;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ubyte2ushort_first2last;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_ubyte2ushort_first2last;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ubyte2ushort_first2last;
+translate[IN_UBYTE][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_ubyte2ushort_first2last;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_ubyte2ushort_last2first;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_ubyte2ushort_last2first;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ubyte2ushort_last2first;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ubyte2ushort_last2first;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_ubyte2ushort_last2first;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ubyte2ushort_last2first;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ubyte2ushort_last2first;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_ubyte2ushort_last2first;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ubyte2ushort_last2first;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_ubyte2ushort_last2first;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_ubyte2ushort_last2last;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_ubyte2ushort_last2last;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ubyte2ushort_last2last;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ubyte2ushort_last2last;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_ubyte2ushort_last2last;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ubyte2ushort_last2last;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ubyte2ushort_last2last;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_ubyte2ushort_last2last;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ubyte2ushort_last2last;
+translate[IN_UBYTE][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_ubyte2ushort_last2last;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_ubyte2uint_first2first;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_ubyte2uint_first2first;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ubyte2uint_first2first;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ubyte2uint_first2first;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_ubyte2uint_first2first;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ubyte2uint_first2first;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ubyte2uint_first2first;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_ubyte2uint_first2first;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ubyte2uint_first2first;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_ubyte2uint_first2first;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_ubyte2uint_first2last;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_ubyte2uint_first2last;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ubyte2uint_first2last;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ubyte2uint_first2last;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_ubyte2uint_first2last;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ubyte2uint_first2last;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ubyte2uint_first2last;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_ubyte2uint_first2last;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ubyte2uint_first2last;
+translate[IN_UBYTE][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_ubyte2uint_first2last;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_ubyte2uint_last2first;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_ubyte2uint_last2first;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ubyte2uint_last2first;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ubyte2uint_last2first;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_ubyte2uint_last2first;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ubyte2uint_last2first;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ubyte2uint_last2first;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_ubyte2uint_last2first;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ubyte2uint_last2first;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_ubyte2uint_last2first;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_ubyte2uint_last2last;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_ubyte2uint_last2last;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ubyte2uint_last2last;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ubyte2uint_last2last;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_ubyte2uint_last2last;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ubyte2uint_last2last;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ubyte2uint_last2last;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_ubyte2uint_last2last;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ubyte2uint_last2last;
+translate[IN_UBYTE][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_ubyte2uint_last2last;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_ushort2ushort_first2first;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_ushort2ushort_first2first;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ushort2ushort_first2first;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ushort2ushort_first2first;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_ushort2ushort_first2first;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ushort2ushort_first2first;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ushort2ushort_first2first;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_ushort2ushort_first2first;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ushort2ushort_first2first;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_ushort2ushort_first2first;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_ushort2ushort_first2last;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_ushort2ushort_first2last;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ushort2ushort_first2last;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ushort2ushort_first2last;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_ushort2ushort_first2last;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ushort2ushort_first2last;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ushort2ushort_first2last;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_ushort2ushort_first2last;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ushort2ushort_first2last;
+translate[IN_USHORT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_ushort2ushort_first2last;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_ushort2ushort_last2first;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_ushort2ushort_last2first;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ushort2ushort_last2first;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ushort2ushort_last2first;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_ushort2ushort_last2first;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ushort2ushort_last2first;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ushort2ushort_last2first;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_ushort2ushort_last2first;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ushort2ushort_last2first;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_ushort2ushort_last2first;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_ushort2ushort_last2last;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_ushort2ushort_last2last;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ushort2ushort_last2last;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ushort2ushort_last2last;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_ushort2ushort_last2last;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ushort2ushort_last2last;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ushort2ushort_last2last;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_ushort2ushort_last2last;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ushort2ushort_last2last;
+translate[IN_USHORT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_ushort2ushort_last2last;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_ushort2uint_first2first;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_ushort2uint_first2first;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ushort2uint_first2first;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ushort2uint_first2first;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_ushort2uint_first2first;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ushort2uint_first2first;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ushort2uint_first2first;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_ushort2uint_first2first;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ushort2uint_first2first;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_ushort2uint_first2first;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_ushort2uint_first2last;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_ushort2uint_first2last;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ushort2uint_first2last;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ushort2uint_first2last;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_ushort2uint_first2last;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ushort2uint_first2last;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ushort2uint_first2last;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_ushort2uint_first2last;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ushort2uint_first2last;
+translate[IN_USHORT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_ushort2uint_first2last;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_ushort2uint_last2first;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_ushort2uint_last2first;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ushort2uint_last2first;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ushort2uint_last2first;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_ushort2uint_last2first;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ushort2uint_last2first;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ushort2uint_last2first;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_ushort2uint_last2first;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ushort2uint_last2first;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_ushort2uint_last2first;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_ushort2uint_last2last;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_ushort2uint_last2last;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_ushort2uint_last2last;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_ushort2uint_last2last;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_ushort2uint_last2last;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_ushort2uint_last2last;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_ushort2uint_last2last;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_ushort2uint_last2last;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_ushort2uint_last2last;
+translate[IN_USHORT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_ushort2uint_last2last;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_uint2ushort_first2first;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_uint2ushort_first2first;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_uint2ushort_first2first;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_uint2ushort_first2first;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_uint2ushort_first2first;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_uint2ushort_first2first;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_uint2ushort_first2first;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_uint2ushort_first2first;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_uint2ushort_first2first;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_uint2ushort_first2first;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_uint2ushort_first2last;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_uint2ushort_first2last;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_uint2ushort_first2last;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_uint2ushort_first2last;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_uint2ushort_first2last;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_uint2ushort_first2last;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_uint2ushort_first2last;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_uint2ushort_first2last;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_uint2ushort_first2last;
+translate[IN_UINT][OUT_USHORT][PV_FIRST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_uint2ushort_first2last;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_uint2ushort_last2first;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_uint2ushort_last2first;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_uint2ushort_last2first;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_uint2ushort_last2first;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_uint2ushort_last2first;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_uint2ushort_last2first;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_uint2ushort_last2first;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_uint2ushort_last2first;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_uint2ushort_last2first;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_uint2ushort_last2first;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_uint2ushort_last2last;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_uint2ushort_last2last;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_uint2ushort_last2last;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_uint2ushort_last2last;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_uint2ushort_last2last;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_uint2ushort_last2last;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_uint2ushort_last2last;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_uint2ushort_last2last;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_uint2ushort_last2last;
+translate[IN_UINT][OUT_USHORT][PV_LAST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_uint2ushort_last2last;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_uint2uint_first2first;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_uint2uint_first2first;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_uint2uint_first2first;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_uint2uint_first2first;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_uint2uint_first2first;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_uint2uint_first2first;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_uint2uint_first2first;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_uint2uint_first2first;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_uint2uint_first2first;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_uint2uint_first2first;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_uint2uint_first2last;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_uint2uint_first2last;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_uint2uint_first2last;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_uint2uint_first2last;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_uint2uint_first2last;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_uint2uint_first2last;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_uint2uint_first2last;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_uint2uint_first2last;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_uint2uint_first2last;
+translate[IN_UINT][OUT_UINT][PV_FIRST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_uint2uint_first2last;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_POINTS] = translate_points_uint2uint_last2first;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINES] = translate_lines_uint2uint_last2first;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_uint2uint_last2first;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_uint2uint_last2first;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLES] = translate_tris_uint2uint_last2first;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_uint2uint_last2first;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_uint2uint_last2first;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_QUADS] = translate_quads_uint2uint_last2first;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_uint2uint_last2first;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_FIRST][PIPE_PRIM_POLYGON] = translate_polygon_uint2uint_last2first;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_POINTS] = translate_points_uint2uint_last2last;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINES] = translate_lines_uint2uint_last2last;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_STRIP] = translate_linestrip_uint2uint_last2last;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_LINE_LOOP] = translate_lineloop_uint2uint_last2last;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLES] = translate_tris_uint2uint_last2last;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_FAN] = translate_trifan_uint2uint_last2last;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_TRIANGLE_STRIP] = translate_tristrip_uint2uint_last2last;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_QUADS] = translate_quads_uint2uint_last2last;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_QUAD_STRIP] = translate_quadstrip_uint2uint_last2last;
+translate[IN_UINT][OUT_UINT][PV_LAST][PV_LAST][PIPE_PRIM_POLYGON] = translate_polygon_uint2uint_last2last;
+}
+#include "indices/u_indices.c"
diff --git a/src/gallium/auxiliary/indices/u_indices_gen.py b/src/gallium/auxiliary/indices/u_indices_gen.py
new file mode 100644
index 0000000000..0dc58d0cd0
--- /dev/null
+++ b/src/gallium/auxiliary/indices/u_indices_gen.py
@@ -0,0 +1,319 @@
+#!/usr/bin/env python
+copyright = '''
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VMWARE AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+'''
+
+GENERATE, UBYTE, USHORT, UINT = 'generate', 'ubyte', 'ushort', 'uint'
+FIRST, LAST = 'first', 'last'
+
+INTYPES = (GENERATE, UBYTE, USHORT, UINT)
+OUTTYPES = (USHORT, UINT)
+PVS=(FIRST, LAST)
+PRIMS=('points', 
+       'lines', 
+       'linestrip', 
+       'lineloop', 
+       'tris', 
+       'trifan', 
+       'tristrip', 
+       'quads', 
+       'quadstrip', 
+       'polygon')
+
+LONGPRIMS=('PIPE_PRIM_POINTS', 
+           'PIPE_PRIM_LINES', 
+           'PIPE_PRIM_LINE_STRIP', 
+           'PIPE_PRIM_LINE_LOOP', 
+           'PIPE_PRIM_TRIANGLES', 
+           'PIPE_PRIM_TRIANGLE_FAN', 
+           'PIPE_PRIM_TRIANGLE_STRIP', 
+           'PIPE_PRIM_QUADS', 
+           'PIPE_PRIM_QUAD_STRIP', 
+           'PIPE_PRIM_POLYGON')
+
+longprim = dict(zip(PRIMS, LONGPRIMS))
+intype_idx = dict(ubyte='IN_UBYTE', ushort='IN_USHORT', uint='IN_UINT')
+outtype_idx = dict(ushort='OUT_USHORT', uint='OUT_UINT')
+pv_idx = dict(first='PV_FIRST', last='PV_LAST')
+
+
+def prolog():
+    print '''/* File automatically generated by indices.py */'''
+    print copyright
+    print r'''
+
+/**
+ * @file
+ * Functions to translate and generate index lists
+ */
+
+#include "indices/u_indices.h"
+#include "indices/u_indices_priv.h"
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+
+
+static unsigned out_size_idx( unsigned index_size )
+{
+   switch (index_size) {
+   case 4: return OUT_UINT;
+   case 2: return OUT_USHORT;
+   default: assert(0); return OUT_USHORT;
+   }
+}
+
+static unsigned in_size_idx( unsigned index_size )
+{
+   switch (index_size) {
+   case 4: return IN_UINT;
+   case 2: return IN_USHORT;
+   case 1: return IN_UBYTE;
+   default: assert(0); return IN_UBYTE;
+   }
+}
+
+
+static u_translate_func translate[IN_COUNT][OUT_COUNT][PV_COUNT][PV_COUNT][PRIM_COUNT];
+static u_generate_func  generate[OUT_COUNT][PV_COUNT][PV_COUNT][PRIM_COUNT];
+
+
+'''
+
+def vert( intype, outtype, v0 ):
+    if intype == GENERATE:
+        return '(' + outtype + ')(' + v0 + ')'
+    else:
+        return '(' + outtype + ')in[' + v0 + ']'
+
+def point( intype, outtype, ptr, v0 ):
+    print '      (' + ptr + ')[0] = ' + vert( intype, outtype, v0 ) + ';'
+
+def line( intype, outtype, ptr, v0, v1 ):
+    print '      (' + ptr + ')[0] = ' + vert( intype, outtype, v0 ) + ';'
+    print '      (' + ptr + ')[1] = ' + vert( intype, outtype, v1 ) + ';'
+
+def tri( intype, outtype, ptr, v0, v1, v2 ):
+    print '      (' + ptr + ')[0] = ' + vert( intype, outtype, v0 ) + ';'
+    print '      (' + ptr + ')[1] = ' + vert( intype, outtype, v1 ) + ';'
+    print '      (' + ptr + ')[2] = ' + vert( intype, outtype, v2 ) + ';'
+
+def do_point( intype, outtype, ptr, v0 ):
+    point( intype, outtype, ptr, v0 )
+
+def do_line( intype, outtype, ptr, v0, v1, inpv, outpv ):
+    if inpv == outpv:
+        line( intype, outtype, ptr, v0, v1 )
+    else:
+        line( intype, outtype, ptr, v1, v0 )
+
+def do_tri( intype, outtype, ptr, v0, v1, v2, inpv, outpv ):
+    if inpv == outpv:
+        tri( intype, outtype, ptr, v0, v1, v2 )
+    else: 
+        if inpv == FIRST:
+            tri( intype, outtype, ptr, v1, v2, v0 )
+        else:
+            tri( intype, outtype, ptr, v2, v0, v1 )
+
+def do_quad( intype, outtype, ptr, v0, v1, v2, v3, inpv, outpv ):
+    do_tri( intype, outtype, ptr+'+0',  v0, v1, v3, inpv, outpv );
+    do_tri( intype, outtype, ptr+'+3',  v1, v2, v3, inpv, outpv );
+
+def name(intype, outtype, inpv, outpv, prim):
+    if intype == GENERATE:
+        return 'generate_' + prim + '_' + outtype + '_' + inpv + '2' + outpv
+    else:
+        return 'translate_' + prim + '_' + intype + '2' + outtype + '_' + inpv + '2' + outpv
+
+def preamble(intype, outtype, inpv, outpv, prim):
+    print 'static void ' + name( intype, outtype, inpv, outpv, prim ) + '('
+    if intype != GENERATE:
+        print '    const void * _in,'
+    print '    unsigned nr,'
+    print '    void *_out )'
+    print '{'
+    if intype != GENERATE:
+        print '  const ' + intype + '*in = (const ' + intype + '*)_in;'
+    print '  ' + outtype + ' *out = (' + outtype + '*)_out;'
+    print '  unsigned i, j;'
+    print '  (void)j;'
+
+def postamble():
+    print '}'
+
+
+def points(intype, outtype, inpv, outpv):
+    preamble(intype, outtype, inpv, outpv, prim='points')
+    print '  for (i = 0; i < nr; i++) { '
+    do_point( intype, outtype, 'out+i',  'i' );
+    print '   }'
+    postamble()
+
+def lines(intype, outtype, inpv, outpv):
+    preamble(intype, outtype, inpv, outpv, prim='lines')
+    print '  for (i = 0; i < nr; i+=2) { '
+    do_line( intype, outtype, 'out+i',  'i', 'i+1', inpv, outpv );
+    print '   }'
+    postamble()
+
+def linestrip(intype, outtype, inpv, outpv):
+    preamble(intype, outtype, inpv, outpv, prim='linestrip')
+    print '  for (j = i = 0; j < nr; j+=2, i++) { '
+    do_line( intype, outtype, 'out+j',  'i', 'i+1', inpv, outpv );
+    print '   }'
+    postamble()
+
+def lineloop(intype, outtype, inpv, outpv):
+    preamble(intype, outtype, inpv, outpv, prim='lineloop')
+    print '  for (j = i = 0; j < nr - 2; j+=2, i++) { '
+    do_line( intype, outtype, 'out+j',  'i', 'i+1', inpv, outpv );
+    print '   }'
+    do_line( intype, outtype, 'out+j',  'i', '0', inpv, outpv );
+    postamble()
+
+def tris(intype, outtype, inpv, outpv):
+    preamble(intype, outtype, inpv, outpv, prim='tris')
+    print '  for (i = 0; i < nr; i+=3) { '
+    do_tri( intype, outtype, 'out+i',  'i', 'i+1', 'i+2', inpv, outpv );
+    print '   }'
+    postamble()
+
+
+def tristrip(intype, outtype, inpv, outpv):
+    preamble(intype, outtype, inpv, outpv, prim='tristrip')
+    print '  for (j = i = 0; j < nr; j+=3, i++) { '
+    if inpv == FIRST:
+        do_tri( intype, outtype, 'out+j',  'i', 'i+1+(i&1)', 'i+2-(i&1)', inpv, outpv );
+    else:
+        do_tri( intype, outtype, 'out+j',  'i+(i&1)', 'i+1-(i&1)', 'i+2', inpv, outpv );
+    print '   }'
+    postamble()
+
+
+def trifan(intype, outtype, inpv, outpv):
+    preamble(intype, outtype, inpv, outpv, prim='trifan')
+    print '  for (j = i = 0; j < nr; j+=3, i++) { '
+    do_tri( intype, outtype, 'out+j',  '0', 'i+1', 'i+2', inpv, outpv );
+    print '   }'
+    postamble()
+
+
+
+def polygon(intype, outtype, inpv, outpv):
+    preamble(intype, outtype, inpv, outpv, prim='polygon')
+    print '  for (j = i = 0; j < nr; j+=3, i++) { '
+    if inpv == FIRST:
+        do_tri( intype, outtype, 'out+j',  '0', 'i+1', 'i+2', inpv, outpv );
+    else:
+        do_tri( intype, outtype, 'out+j',  'i+1', 'i+2', '0', inpv, outpv );
+    print '   }'
+    postamble()
+
+
+def quads(intype, outtype, inpv, outpv):
+    preamble(intype, outtype, inpv, outpv, prim='quads')
+    print '  for (j = i = 0; j < nr; j+=6, i+=4) { '
+    do_quad( intype, outtype, 'out+j', 'i+0', 'i+1', 'i+2', 'i+3', inpv, outpv );
+    print '   }'
+    postamble()
+
+
+def quadstrip(intype, outtype, inpv, outpv):
+    preamble(intype, outtype, inpv, outpv, prim='quadstrip')
+    print '  for (j = i = 0; j < nr; j+=6, i+=2) { '
+    do_quad( intype, outtype, 'out+j', 'i+2', 'i+0', 'i+1', 'i+3', inpv, outpv );
+    print '   }'
+    postamble()
+
+
+def emit_funcs():
+    for intype in INTYPES:
+        for outtype in OUTTYPES:
+            for inpv in (FIRST, LAST):
+                for outpv in (FIRST, LAST):
+                    points(intype, outtype, inpv, outpv)
+                    lines(intype, outtype, inpv, outpv)
+                    linestrip(intype, outtype, inpv, outpv)
+                    lineloop(intype, outtype, inpv, outpv)
+                    tris(intype, outtype, inpv, outpv)
+                    tristrip(intype, outtype, inpv, outpv)
+                    trifan(intype, outtype, inpv, outpv)
+                    quads(intype, outtype, inpv, outpv)
+                    quadstrip(intype, outtype, inpv, outpv)
+                    polygon(intype, outtype, inpv, outpv)
+
+def init(intype, outtype, inpv, outpv, prim):
+    if intype == GENERATE:
+        print ('generate[' + 
+               outtype_idx[outtype] + 
+               '][' + pv_idx[inpv] + 
+               '][' + pv_idx[outpv] + 
+               '][' + longprim[prim] + 
+               '] = ' + name( intype, outtype, inpv, outpv, prim ) + ';')
+    else:
+        print ('translate[' + 
+               intype_idx[intype] + 
+               '][' + outtype_idx[outtype] + 
+               '][' + pv_idx[inpv] + 
+               '][' + pv_idx[outpv] + 
+               '][' + longprim[prim] + 
+               '] = ' + name( intype, outtype, inpv, outpv, prim ) + ';')
+
+
+def emit_all_inits():
+    for intype in INTYPES:
+        for outtype in OUTTYPES:
+            for inpv in PVS:
+                for outpv in PVS:
+                    for prim in PRIMS:
+                        init(intype, outtype, inpv, outpv, prim)
+
+def emit_init():
+    print 'void u_index_init( void )'
+    print '{'
+    print '  static int firsttime = 1;'
+    print '  if (!firsttime) return;'
+    print '  firsttime = 0;'
+    emit_all_inits()
+    print '}'
+
+
+    
+
+def epilog():
+    print '#include "indices/u_indices.c"'
+
+
+def main():
+    prolog()
+    emit_funcs()
+    emit_init()
+    epilog()
+
+
+if __name__ == '__main__':
+    main()
diff --git a/src/gallium/auxiliary/indices/u_indices_priv.h b/src/gallium/auxiliary/indices/u_indices_priv.h
new file mode 100644
index 0000000000..9c3298c24d
--- /dev/null
+++ b/src/gallium/auxiliary/indices/u_indices_priv.h
@@ -0,0 +1,43 @@
+/*
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef U_INDICES_PRIV_H
+#define U_INDICES_PRIV_H
+
+#include "pipe/p_compiler.h"
+#include "u_indices.h"
+
+#define IN_UBYTE      0
+#define IN_USHORT     1
+#define IN_UINT       2
+#define IN_COUNT      3
+
+#define OUT_USHORT    0
+#define OUT_UINT      1
+#define OUT_COUNT     2
+
+
+#define PRIM_COUNT   (PIPE_PRIM_POLYGON + 1)
+
+#endif
diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile
new file mode 100644
index 0000000000..3b501c51df
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/Makefile
@@ -0,0 +1,22 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = pipebuffer
+
+C_SOURCES = \
+	pb_buffer_fenced.c \
+	pb_buffer_malloc.c \
+	pb_bufmgr_alt.c \
+	pb_bufmgr_cache.c \
+	pb_bufmgr_debug.c \
+	pb_bufmgr_fenced.c \
+	pb_bufmgr_mm.c \
+	pb_bufmgr_ondemand.c \
+	pb_bufmgr_pool.c \
+	pb_bufmgr_slab.c \
+	pb_validate.c
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/pipebuffer/SConscript b/src/gallium/auxiliary/pipebuffer/SConscript
new file mode 100644
index 0000000000..8e9f06abe4
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/SConscript
@@ -0,0 +1,19 @@
+Import('*')
+
+pipebuffer = env.ConvenienceLibrary(
+	target = 'pipebuffer',
+	source = [
+		'pb_buffer_fenced.c',
+		'pb_buffer_malloc.c',
+		'pb_bufmgr_alt.c',
+		'pb_bufmgr_cache.c',
+		'pb_bufmgr_debug.c',
+		'pb_bufmgr_fenced.c',
+		'pb_bufmgr_mm.c',
+		'pb_bufmgr_ondemand.c',
+		'pb_bufmgr_pool.c',
+		'pb_bufmgr_slab.c',
+		'pb_validate.c',
+	])
+
+auxiliaries.insert(0, pipebuffer)
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h
new file mode 100644
index 0000000000..d8f1f02d68
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h
@@ -0,0 +1,292 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Generic code for buffers.
+ * 
+ * Behind a pipe buffle handle there can be DMA buffers, client (or user) 
+ * buffers, regular malloced buffers, etc. This file provides an abstract base 
+ * buffer handle that allows the driver to cope with all those kinds of buffers 
+ * in a more flexible way.
+ * 
+ * There is no obligation of a winsys driver to use this library. And a pipe
+ * driver should be completly agnostic about it.
+ * 
+ * \author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef PB_BUFFER_H_
+#define PB_BUFFER_H_
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_error.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+struct pb_vtbl;
+struct pb_validate;
+
+
+/**
+ * Buffer description.
+ * 
+ * Used when allocating the buffer.
+ */
+struct pb_desc
+{
+   unsigned alignment;
+   unsigned usage;
+};
+
+
+/**
+ * Base class for all pb_* buffers.
+ */
+struct pb_buffer 
+{
+   struct pipe_buffer base;
+
+   /**
+    * Pointer to the virtual function table.
+    *
+    * Avoid accessing this table directly. Use the inline functions below 
+    * instead to avoid mistakes. 
+    */
+   const struct pb_vtbl *vtbl;
+};
+
+
+/**
+ * Virtual function table for the buffer storage operations.
+ * 
+ * Note that creation is not done through this table.
+ */
+struct pb_vtbl
+{
+   void (*destroy)( struct pb_buffer *buf );
+
+   /** 
+    * Map the entire data store of a buffer object into the client's address.
+    * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. 
+    */
+   void *(*map)( struct pb_buffer *buf, 
+                 unsigned flags );
+   
+   void (*unmap)( struct pb_buffer *buf );
+
+   enum pipe_error (*validate)( struct pb_buffer *buf, 
+                                struct pb_validate *vl,
+                                unsigned flags );
+
+   void (*fence)( struct pb_buffer *buf, 
+                  struct pipe_fence_handle *fence );
+
+   /**
+    * Get the base buffer and the offset.
+    * 
+    * A buffer can be subdivided in smaller buffers. This method should return
+    * the underlaying buffer, and the relative offset.
+    * 
+    * Buffers without an underlaying base buffer should return themselves, with 
+    * a zero offset.
+    * 
+    * Note that this will increase the reference count of the base buffer.
+    */
+   void (*get_base_buffer)( struct pb_buffer *buf,
+                            struct pb_buffer **base_buf,
+                            unsigned *offset );
+   
+};
+
+
+static INLINE struct pipe_buffer *
+pb_pipe_buffer( struct pb_buffer *pbuf )
+{
+   assert(pbuf);
+   return &pbuf->base;
+}
+
+
+static INLINE struct pb_buffer *
+pb_buffer( struct pipe_buffer *buf )
+{
+   assert(buf);
+   /* Could add a magic cookie check on debug builds.
+    */
+   return (struct pb_buffer *)buf;
+}
+
+
+/* Accessor functions for pb->vtbl:
+ */
+static INLINE void *
+pb_map(struct pb_buffer *buf, 
+       unsigned flags)
+{
+   assert(buf);
+   if(!buf)
+      return NULL;
+   assert(buf->base.refcount > 0);
+   return buf->vtbl->map(buf, flags);
+}
+
+
+static INLINE void 
+pb_unmap(struct pb_buffer *buf)
+{
+   assert(buf);
+   if(!buf)
+      return;
+   assert(buf->base.refcount > 0);
+   buf->vtbl->unmap(buf);
+}
+
+
+static INLINE void
+pb_get_base_buffer( struct pb_buffer *buf,
+		    struct pb_buffer **base_buf,
+		    unsigned *offset )
+{
+   assert(buf);
+   if(!buf) {
+      base_buf = NULL;
+      offset = 0;
+      return;
+   }
+   assert(buf->base.refcount > 0);
+   assert(buf->vtbl->get_base_buffer);
+   buf->vtbl->get_base_buffer(buf, base_buf, offset);
+   assert(*base_buf);
+   assert(*offset < (*base_buf)->base.size);
+}
+
+
+static INLINE enum pipe_error 
+pb_validate(struct pb_buffer *buf, struct pb_validate *vl, unsigned flags)
+{
+   assert(buf);
+   if(!buf)
+      return PIPE_ERROR;
+   assert(buf->vtbl->validate);
+   return buf->vtbl->validate(buf, vl, flags);
+}
+
+
+static INLINE void 
+pb_fence(struct pb_buffer *buf, struct pipe_fence_handle *fence)
+{
+   assert(buf);
+   if(!buf)
+      return;
+   assert(buf->vtbl->fence);
+   buf->vtbl->fence(buf, fence);
+}
+
+
+static INLINE void 
+pb_destroy(struct pb_buffer *buf)
+{
+   assert(buf);
+   if(!buf)
+      return;
+   assert(buf->base.refcount == 0);
+   buf->vtbl->destroy(buf);
+}
+
+
+/* XXX: thread safety issues!
+ */
+static INLINE void
+pb_reference(struct pb_buffer **dst,
+             struct pb_buffer *src)
+{
+   if (src) {
+      assert(src->base.refcount);
+      src->base.refcount++;
+   }
+
+   if (*dst) {
+      assert((*dst)->base.refcount);
+      if(--(*dst)->base.refcount == 0)
+         pb_destroy( *dst );
+   }
+
+   *dst = src;
+}
+
+
+/**
+ * Utility function to check whether the provided alignment is consistent with
+ * the requested or not.
+ */
+static INLINE boolean
+pb_check_alignment(size_t requested, size_t provided)
+{
+   if(!requested)
+      return TRUE;
+   if(requested > provided)
+      return FALSE;
+   if(provided % requested != 0)
+      return FALSE;
+   return TRUE;
+}
+
+
+/**
+ * Utility function to check whether the provided alignment is consistent with
+ * the requested or not.
+ */
+static INLINE boolean
+pb_check_usage(unsigned requested, unsigned provided)
+{
+   return (requested & provided) == requested ? TRUE : FALSE;
+}
+
+
+/**
+ * Malloc-based buffer to store data that can't be used by the graphics 
+ * hardware.
+ */
+struct pb_buffer *
+pb_malloc_buffer_create(size_t size, 
+                        const struct pb_desc *desc);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*PB_BUFFER_H_*/
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
new file mode 100644
index 0000000000..f9e6226436
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
@@ -0,0 +1,501 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Implementation of fenced buffers.
+ * 
+ * \author Jose Fonseca <jrfonseca-at-tungstengraphics-dot-com>
+ * \author Thomas Hellström <thomas-at-tungstengraphics-dot-com>
+ */
+
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_OS_LINUX)
+#include <unistd.h>
+#include <sched.h>
+#endif
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_error.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_thread.h"
+#include "util/u_memory.h"
+#include "util/u_double_list.h"
+
+#include "pb_buffer.h"
+#include "pb_buffer_fenced.h"
+
+
+
+/**
+ * Convenience macro (type safe).
+ */
+#define SUPER(__derived) (&(__derived)->base)
+
+
+struct fenced_buffer_list
+{
+   pipe_mutex mutex;
+   
+   struct pb_fence_ops *ops;
+   
+   size_t numDelayed;
+   
+   struct list_head delayed;
+};
+
+
+/**
+ * Wrapper around a pipe buffer which adds fencing and reference counting.
+ */
+struct fenced_buffer
+{
+   struct pb_buffer base;
+   
+   struct pb_buffer *buffer;
+
+   /* FIXME: protect access with mutex */
+
+   /**
+    * A bitmask of PIPE_BUFFER_USAGE_CPU/GPU_READ/WRITE describing the current
+    * buffer usage.
+    */
+   unsigned flags;
+
+   unsigned mapcount;
+   struct pb_validate *vl;
+   unsigned validation_flags;
+   struct pipe_fence_handle *fence;
+
+   struct list_head head;
+   struct fenced_buffer_list *list;
+};
+
+
+static INLINE struct fenced_buffer *
+fenced_buffer(struct pb_buffer *buf)
+{
+   assert(buf);
+   return (struct fenced_buffer *)buf;
+}
+
+
+static INLINE void
+_fenced_buffer_add(struct fenced_buffer *fenced_buf)
+{
+   struct fenced_buffer_list *fenced_list = fenced_buf->list;
+
+   assert(fenced_buf->base.base.refcount);
+   assert(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE);
+   assert(fenced_buf->fence);
+
+   assert(!fenced_buf->head.prev);
+   assert(!fenced_buf->head.next);
+   LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed);
+   ++fenced_list->numDelayed;
+}
+
+
+/**
+ * Actually destroy the buffer.
+ */
+static INLINE void
+_fenced_buffer_destroy(struct fenced_buffer *fenced_buf)
+{
+   assert(!fenced_buf->base.base.refcount);
+   assert(!fenced_buf->fence);
+   pb_reference(&fenced_buf->buffer, NULL);
+   FREE(fenced_buf);
+}
+
+
+static INLINE void
+_fenced_buffer_remove(struct fenced_buffer_list *fenced_list,
+                      struct fenced_buffer *fenced_buf)
+{
+   struct pb_fence_ops *ops = fenced_list->ops;
+
+   assert(fenced_buf->fence);
+   assert(fenced_buf->list == fenced_list);
+   
+   ops->fence_reference(ops, &fenced_buf->fence, NULL);
+   fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE;
+   
+   assert(fenced_buf->head.prev);
+   assert(fenced_buf->head.next);
+   LIST_DEL(&fenced_buf->head);
+#ifdef DEBUG
+   fenced_buf->head.prev = NULL;
+   fenced_buf->head.next = NULL;
+#endif
+   
+   assert(fenced_list->numDelayed);
+   --fenced_list->numDelayed;
+   
+   if(!fenced_buf->base.base.refcount)
+      _fenced_buffer_destroy(fenced_buf);
+}
+
+
+static INLINE enum pipe_error
+_fenced_buffer_finish(struct fenced_buffer *fenced_buf)
+{
+   struct fenced_buffer_list *fenced_list = fenced_buf->list;
+   struct pb_fence_ops *ops = fenced_list->ops;
+
+#if 0
+   debug_warning("waiting for GPU");
+#endif
+
+   assert(fenced_buf->fence);
+   if(fenced_buf->fence) {
+      if(ops->fence_finish(ops, fenced_buf->fence, 0) != 0) {
+	 return PIPE_ERROR;
+      }
+      /* Remove from the fenced list */
+      /* TODO: remove consequents */
+      _fenced_buffer_remove(fenced_list, fenced_buf);
+   }
+
+   fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE;
+   return PIPE_OK;
+}
+
+
+/**
+ * Free as many fenced buffers from the list head as possible. 
+ */
+static void
+_fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, 
+                               int wait)
+{
+   struct pb_fence_ops *ops = fenced_list->ops;
+   struct list_head *curr, *next;
+   struct fenced_buffer *fenced_buf;
+   struct pipe_fence_handle *prev_fence = NULL;
+
+   curr = fenced_list->delayed.next;
+   next = curr->next;
+   while(curr != &fenced_list->delayed) {
+      fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
+
+      if(fenced_buf->fence != prev_fence) {
+	 int signaled;
+	 if (wait)
+	    signaled = ops->fence_finish(ops, fenced_buf->fence, 0);
+	 else
+	    signaled = ops->fence_signalled(ops, fenced_buf->fence, 0);
+	 if (signaled != 0)
+	    break;
+	 prev_fence = fenced_buf->fence;
+      }
+      else {
+	 assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0);
+      }
+
+      _fenced_buffer_remove(fenced_list, fenced_buf);
+
+      curr = next; 
+      next = curr->next;
+   }
+}
+
+
+static void
+fenced_buffer_destroy(struct pb_buffer *buf)
+{
+   struct fenced_buffer *fenced_buf = fenced_buffer(buf);   
+   struct fenced_buffer_list *fenced_list = fenced_buf->list;
+
+   pipe_mutex_lock(fenced_list->mutex);
+   assert(fenced_buf->base.base.refcount == 0);
+   if (fenced_buf->fence) {
+      struct pb_fence_ops *ops = fenced_list->ops;
+      if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) {
+	 struct list_head *curr, *prev;
+	 curr = &fenced_buf->head;
+	 prev = curr->prev;
+	 do {
+	    fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
+	    assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0);
+	    _fenced_buffer_remove(fenced_list, fenced_buf);
+	    curr = prev;
+	    prev = curr->prev;
+	 } while (curr != &fenced_list->delayed);
+      }	  
+      else {
+	 /* delay destruction */
+      }
+   }
+   else {
+      _fenced_buffer_destroy(fenced_buf);
+   }
+   pipe_mutex_unlock(fenced_list->mutex);
+}
+
+
+static void *
+fenced_buffer_map(struct pb_buffer *buf, 
+                  unsigned flags)
+{
+   struct fenced_buffer *fenced_buf = fenced_buffer(buf);
+   void *map;
+
+   assert(flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE);
+   assert(!(flags & ~PIPE_BUFFER_USAGE_CPU_READ_WRITE));
+   flags &= PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+   
+   /* Check for GPU read/write access */
+   if(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) {
+      /* Wait for the GPU to finish writing */
+      _fenced_buffer_finish(fenced_buf);
+   }
+
+#if 0
+   /* Check for CPU write access (read is OK) */
+   if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) {
+      /* this is legal -- just for debugging */
+      debug_warning("concurrent CPU writes");
+   }
+#endif
+   
+   map = pb_map(fenced_buf->buffer, flags);
+   if(map) {
+      ++fenced_buf->mapcount;
+      fenced_buf->flags |= flags;
+   }
+
+   return map;
+}
+
+
+static void
+fenced_buffer_unmap(struct pb_buffer *buf)
+{
+   struct fenced_buffer *fenced_buf = fenced_buffer(buf);
+   assert(fenced_buf->mapcount);
+   if(fenced_buf->mapcount) {
+      pb_unmap(fenced_buf->buffer);
+      --fenced_buf->mapcount;
+      if(!fenced_buf->mapcount)
+	 fenced_buf->flags &= ~PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+   }
+}
+
+
+static enum pipe_error
+fenced_buffer_validate(struct pb_buffer *buf,
+                       struct pb_validate *vl,
+                       unsigned flags)
+{
+   struct fenced_buffer *fenced_buf = fenced_buffer(buf);
+   enum pipe_error ret;
+   
+   if(!vl) {
+      /* invalidate */
+      fenced_buf->vl = NULL;
+      fenced_buf->validation_flags = 0;
+      return PIPE_OK;
+   }
+   
+   assert(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE);
+   assert(!(flags & ~PIPE_BUFFER_USAGE_GPU_READ_WRITE));
+   flags &= PIPE_BUFFER_USAGE_GPU_READ_WRITE;
+
+   /* Buffer cannot be validated in two different lists */ 
+   if(fenced_buf->vl && fenced_buf->vl != vl)
+      return PIPE_ERROR_RETRY;
+   
+   /* Do not validate if buffer is still mapped */
+   if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) {
+      /* TODO: wait for the thread that mapped the buffer to unmap it */
+      return PIPE_ERROR_RETRY;
+   }
+
+   if(fenced_buf->vl == vl &&
+      (fenced_buf->validation_flags & flags) == flags) {
+      /* Nothing to do -- buffer already validated */
+      return PIPE_OK;
+   }
+
+   /* Final sanity checking */
+   assert(!(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE));
+   assert(!fenced_buf->mapcount);
+   
+   ret = pb_validate(fenced_buf->buffer, vl, flags);
+   if (ret != PIPE_OK)
+      return ret;
+   
+   fenced_buf->vl = vl;
+   fenced_buf->validation_flags |= flags;
+   
+   return PIPE_OK;
+}
+
+
+static void
+fenced_buffer_fence(struct pb_buffer *buf,
+                    struct pipe_fence_handle *fence)
+{
+   struct fenced_buffer *fenced_buf;
+   struct fenced_buffer_list *fenced_list;
+   struct pb_fence_ops *ops;
+
+   fenced_buf = fenced_buffer(buf);
+   fenced_list = fenced_buf->list;
+   ops = fenced_list->ops;
+   
+   if(fence == fenced_buf->fence) {
+      /* Nothing to do */
+      return;
+   }
+
+   assert(fenced_buf->vl);
+   assert(fenced_buf->validation_flags);
+   
+   pipe_mutex_lock(fenced_list->mutex);
+   if (fenced_buf->fence)
+      _fenced_buffer_remove(fenced_list, fenced_buf);
+   if (fence) {
+      ops->fence_reference(ops, &fenced_buf->fence, fence);
+      fenced_buf->flags |= fenced_buf->validation_flags;
+      _fenced_buffer_add(fenced_buf);
+   }
+   pipe_mutex_unlock(fenced_list->mutex);
+   
+   pb_fence(fenced_buf->buffer, fence);
+
+   fenced_buf->vl = NULL;
+   fenced_buf->validation_flags = 0;
+}
+
+
+static void
+fenced_buffer_get_base_buffer(struct pb_buffer *buf,
+                              struct pb_buffer **base_buf,
+                              unsigned *offset)
+{
+   struct fenced_buffer *fenced_buf = fenced_buffer(buf);
+   pb_get_base_buffer(fenced_buf->buffer, base_buf, offset);
+}
+
+
+static const struct pb_vtbl 
+fenced_buffer_vtbl = {
+      fenced_buffer_destroy,
+      fenced_buffer_map,
+      fenced_buffer_unmap,
+      fenced_buffer_validate,
+      fenced_buffer_fence,
+      fenced_buffer_get_base_buffer
+};
+
+
+struct pb_buffer *
+fenced_buffer_create(struct fenced_buffer_list *fenced_list, 
+                     struct pb_buffer *buffer)
+{
+   struct fenced_buffer *buf;
+   
+   if(!buffer)
+      return NULL;
+   
+   buf = CALLOC_STRUCT(fenced_buffer);
+   if(!buf) {
+      pb_reference(&buffer, NULL);
+      return NULL;
+   }
+   
+   buf->base.base.refcount = 1;
+   buf->base.base.alignment = buffer->base.alignment;
+   buf->base.base.usage = buffer->base.usage;
+   buf->base.base.size = buffer->base.size;
+   
+   buf->base.vtbl = &fenced_buffer_vtbl;
+   buf->buffer = buffer;
+   buf->list = fenced_list;
+   
+   return &buf->base;
+}
+
+
+struct fenced_buffer_list *
+fenced_buffer_list_create(struct pb_fence_ops *ops) 
+{
+   struct fenced_buffer_list *fenced_list;
+
+   fenced_list = CALLOC_STRUCT(fenced_buffer_list);
+   if (!fenced_list)
+      return NULL;
+
+   fenced_list->ops = ops;
+
+   LIST_INITHEAD(&fenced_list->delayed);
+
+   fenced_list->numDelayed = 0;
+   
+   pipe_mutex_init(fenced_list->mutex);
+
+   return fenced_list;
+}
+
+
+void
+fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, 
+                              int wait)
+{
+   pipe_mutex_lock(fenced_list->mutex);
+   _fenced_buffer_list_check_free(fenced_list, wait);
+   pipe_mutex_unlock(fenced_list->mutex);
+}
+
+
+void
+fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list)
+{
+   pipe_mutex_lock(fenced_list->mutex);
+
+   /* Wait on outstanding fences */
+   while (fenced_list->numDelayed) {
+      pipe_mutex_unlock(fenced_list->mutex);
+#if defined(PIPE_OS_LINUX)
+      sched_yield();
+#endif
+      _fenced_buffer_list_check_free(fenced_list, 1);
+      pipe_mutex_lock(fenced_list->mutex);
+   }
+
+   pipe_mutex_unlock(fenced_list->mutex);
+   
+   fenced_list->ops->destroy(fenced_list->ops);
+   
+   FREE(fenced_list);
+}
+
+
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h
new file mode 100644
index 0000000000..d1c9d4c17d
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h
@@ -0,0 +1,135 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Buffer fencing.
+ * 
+ * "Fenced buffers" is actually a misnomer. They should be referred as 
+ * "fenceable buffers", i.e, buffers that can be fenced, but I couldn't find
+ * the word "fenceable" in the dictionary.
+ * 
+ * A "fenced buffer" is a decorator around a normal buffer, which adds two 
+ * special properties:
+ * - the ability for the destruction to be delayed by a fence;
+ * - reference counting.
+ * 
+ * Usually DMA buffers have a life-time that will extend the life-time of its 
+ * handle. The end-of-life is dictated by the fence signalling. 
+ * 
+ * Between the handle's destruction, and the fence signalling, the buffer is 
+ * stored in a fenced buffer list.
+ * 
+ * \author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef PB_BUFFER_FENCED_H_
+#define PB_BUFFER_FENCED_H_
+
+
+#include "pipe/p_debug.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+struct pipe_buffer;
+struct pipe_fence_handle;
+
+
+/**
+ * List of buffers which are awaiting fence signalling.
+ */
+struct fenced_buffer_list;
+
+
+struct pb_fence_ops
+{
+   void (*destroy)( struct pb_fence_ops *ops );
+
+   /** Set ptr = fence, with reference counting */
+   void (*fence_reference)( struct pb_fence_ops *ops,
+                            struct pipe_fence_handle **ptr,
+                            struct pipe_fence_handle *fence );
+
+   /**
+    * Checks whether the fence has been signalled.
+    * \param flags  driver-specific meaning
+    * \return zero on success.
+    */
+   int (*fence_signalled)( struct pb_fence_ops *ops,
+                           struct pipe_fence_handle *fence,
+                           unsigned flag );
+
+   /**
+    * Wait for the fence to finish.
+    * \param flags  driver-specific meaning
+    * \return zero on success.
+    */
+   int (*fence_finish)( struct pb_fence_ops *ops,
+                        struct pipe_fence_handle *fence,
+                        unsigned flag );
+};
+
+
+/**
+ * Create a fenced buffer list.
+ * 
+ * See also fenced_bufmgr_create for a more convenient way to use this.
+ */
+struct fenced_buffer_list *
+fenced_buffer_list_create(struct pb_fence_ops *ops);
+
+
+/**
+ * Walk the fenced buffer list to check and free signalled buffers.
+ */ 
+void
+fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, 
+                              int wait);
+
+void
+fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list);
+
+
+/**
+ * Wrap a buffer in a fenced buffer.
+ * 
+ * NOTE: this will not increase the buffer reference count.
+ */
+struct pb_buffer *
+fenced_buffer_create(struct fenced_buffer_list *fenced, 
+                     struct pb_buffer *buffer);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*PB_BUFFER_FENCED_H_*/
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c
new file mode 100644
index 0000000000..53f497cfb0
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c
@@ -0,0 +1,186 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Implementation of malloc-based buffers to store data that can't be processed
+ * by the hardware. 
+ * 
+ * \author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_debug.h"
+#include "util/u_memory.h"
+#include "pb_buffer.h"
+#include "pb_bufmgr.h"
+
+
+struct malloc_buffer 
+{
+   struct pb_buffer base;
+   void *data;
+};
+
+
+extern const struct pb_vtbl malloc_buffer_vtbl;
+
+static INLINE struct malloc_buffer *
+malloc_buffer(struct pb_buffer *buf)
+{
+   assert(buf);
+   assert(buf->vtbl == &malloc_buffer_vtbl);
+   return (struct malloc_buffer *)buf;
+}
+
+
+static void
+malloc_buffer_destroy(struct pb_buffer *buf)
+{
+   align_free(malloc_buffer(buf)->data);
+   FREE(buf);
+}
+
+
+static void *
+malloc_buffer_map(struct pb_buffer *buf, 
+                  unsigned flags)
+{
+   return malloc_buffer(buf)->data;
+}
+
+
+static void
+malloc_buffer_unmap(struct pb_buffer *buf)
+{
+   /* No-op */
+}
+
+
+static enum pipe_error 
+malloc_buffer_validate(struct pb_buffer *buf, 
+                       struct pb_validate *vl,
+                       unsigned flags)
+{
+   assert(0);
+   return PIPE_ERROR;
+}
+
+
+static void
+malloc_buffer_fence(struct pb_buffer *buf, 
+                    struct pipe_fence_handle *fence)
+{
+   assert(0);
+}
+
+
+static void
+malloc_buffer_get_base_buffer(struct pb_buffer *buf,
+                              struct pb_buffer **base_buf,
+                              unsigned *offset)
+{
+   *base_buf = buf;
+   *offset = 0;
+}
+
+
+const struct pb_vtbl 
+malloc_buffer_vtbl = {
+      malloc_buffer_destroy,
+      malloc_buffer_map,
+      malloc_buffer_unmap,
+      malloc_buffer_validate,
+      malloc_buffer_fence,
+      malloc_buffer_get_base_buffer
+};
+
+
+struct pb_buffer *
+pb_malloc_buffer_create(size_t size,
+                   	const struct pb_desc *desc) 
+{
+   struct malloc_buffer *buf;
+   
+   /* TODO: do a single allocation */
+   
+   buf = CALLOC_STRUCT(malloc_buffer);
+   if(!buf)
+      return NULL;
+
+   buf->base.base.refcount = 1;
+   buf->base.base.alignment = desc->alignment;
+   buf->base.base.usage = desc->usage;
+   buf->base.base.size = size;
+   buf->base.vtbl = &malloc_buffer_vtbl;
+
+   buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment);
+   if(!buf->data) {
+      FREE(buf);
+      return NULL;
+   }
+
+   return &buf->base;
+}
+
+
+static struct pb_buffer *
+pb_malloc_bufmgr_create_buffer(struct pb_manager *mgr, 
+                               size_t size,
+                               const struct pb_desc *desc) 
+{
+   return pb_malloc_buffer_create(size, desc);
+}
+
+
+static void
+pb_malloc_bufmgr_flush(struct pb_manager *mgr) 
+{
+   /* No-op */
+}
+
+
+static void
+pb_malloc_bufmgr_destroy(struct pb_manager *mgr) 
+{
+   /* No-op */
+}
+
+
+static struct pb_manager 
+pb_malloc_bufmgr = {
+   pb_malloc_bufmgr_destroy,
+   pb_malloc_bufmgr_create_buffer,
+   pb_malloc_bufmgr_flush
+};
+
+
+struct pb_manager *
+pb_malloc_bufmgr_create(void) 
+{
+  return &pb_malloc_bufmgr;
+}
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
new file mode 100644
index 0000000000..fec8db91c7
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
@@ -0,0 +1,213 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Buffer management.
+ * 
+ * A buffer manager does only one basic thing: it creates buffers. Actually,
+ * "buffer factory" would probably a more accurate description.
+ * 
+ * You can chain buffer managers so that you can have a finer grained memory
+ * management and pooling.
+ * 
+ * For example, for a simple batch buffer manager you would chain:
+ * - the native buffer manager, which provides DMA memory from the graphics
+ * memory space;
+ * - the pool buffer manager, which keep around a pool of equally sized buffers
+ * to avoid latency associated with the native buffer manager; 
+ * - the fenced buffer manager, which will delay buffer destruction until the 
+ * the moment the card finishing processing it. 
+ * 
+ * \author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef PB_BUFMGR_H_
+#define PB_BUFMGR_H_
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_error.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+struct pb_desc;
+struct pipe_buffer;
+
+
+/** 
+ * Abstract base class for all buffer managers.
+ */
+struct pb_manager
+{
+   void
+   (*destroy)( struct pb_manager *mgr );
+
+   struct pb_buffer *
+   (*create_buffer)( struct pb_manager *mgr, 
+	             size_t size,
+	             const struct pb_desc *desc);
+
+   /**
+    * Flush all temporary-held buffers.
+    * 
+    * Used mostly to aid debugging memory issues or to clean up resources when 
+    * the drivers are long lived.
+    */
+   void
+   (*flush)( struct pb_manager *mgr );
+};
+
+
+/**
+ * Malloc buffer provider.
+ * 
+ * Simple wrapper around pb_malloc_buffer_create for convenience.
+ */
+struct pb_manager *
+pb_malloc_bufmgr_create(void);
+
+
+/** 
+ * Static buffer pool sub-allocator.
+ * 
+ * Manages the allocation of equally sized buffers. It does so by allocating
+ * a single big buffer and divide it equally sized buffers. 
+ * 
+ * It is meant to manage the allocation of batch buffer pools.
+ */
+struct pb_manager *
+pool_bufmgr_create(struct pb_manager *provider, 
+                   size_t n, size_t size,
+                   const struct pb_desc *desc);
+
+
+/** 
+ * Static sub-allocator based the old memory manager.
+ * 
+ * It managers buffers of different sizes. It does so by allocating a buffer
+ * with the size of the heap, and then using the old mm memory manager to manage
+ * that heap. 
+ */
+struct pb_manager *
+mm_bufmgr_create(struct pb_manager *provider, 
+                 size_t size, size_t align2);
+
+/**
+ * Same as mm_bufmgr_create.
+ * 
+ * Buffer will be release when the manager is destroyed.
+ */
+struct pb_manager *
+mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, 
+                             size_t size, size_t align2);
+
+
+/**
+ * Slab sub-allocator.
+ */
+struct pb_manager *
+pb_slab_manager_create(struct pb_manager *provider,
+                       size_t bufSize,
+                       size_t slabSize,
+                       const struct pb_desc *desc);
+
+/**
+ * Allow a range of buffer size, by aggregating multiple slabs sub-allocators 
+ * with different bucket sizes.
+ */
+struct pb_manager *
+pb_slab_range_manager_create(struct pb_manager *provider,
+                             size_t minBufSize,
+                             size_t maxBufSize,
+                             size_t slabSize,
+                             const struct pb_desc *desc);
+
+
+/** 
+ * Time-based buffer cache.
+ *
+ * This manager keeps a cache of destroyed buffers during a time interval. 
+ */
+struct pb_manager *
+pb_cache_manager_create(struct pb_manager *provider, 
+                     	unsigned usecs); 
+
+
+struct pb_fence_ops;
+
+/** 
+ * Fenced buffer manager.
+ *
+ * This manager is just meant for convenience. It wraps the buffers returned
+ * by another manager in fenced buffers, so that  
+ * 
+ * NOTE: the buffer manager that provides the buffers will be destroyed
+ * at the same time.
+ */
+struct pb_manager *
+fenced_bufmgr_create(struct pb_manager *provider,
+                     struct pb_fence_ops *ops);
+
+
+struct pb_manager *
+pb_alt_manager_create(struct pb_manager *provider1, 
+                      struct pb_manager *provider2);
+
+
+/** 
+ * Ondemand buffer manager.
+ * 
+ * Buffers are created in malloc'ed memory (fast and cached), and the constents
+ * is transfered to a buffer from the provider (typically in slow uncached 
+ * memory) when there is an attempt to validate the buffer.
+ * 
+ * Ideal for situations where one does not know before hand whether a given
+ * buffer will effectively be used by the hardware or not. 
+ */
+struct pb_manager *
+pb_ondemand_manager_create(struct pb_manager *provider); 
+
+
+/** 
+ * Debug buffer manager to detect buffer under- and overflows.
+ *
+ * Band size should be a multiple of the largest alignment
+ */
+struct pb_manager *
+pb_debug_manager_create(struct pb_manager *provider, size_t band_size); 
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*PB_BUFMGR_H_*/
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c
new file mode 100644
index 0000000000..c956924cc7
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c
@@ -0,0 +1,120 @@
+/**************************************************************************
+ *
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Allocate buffers from two alternative buffer providers.
+ * 
+ * \author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "util/u_memory.h"
+
+#include "pb_buffer.h"
+#include "pb_bufmgr.h"
+
+
+struct pb_alt_manager
+{
+   struct pb_manager base;
+
+   struct pb_manager *provider1;
+   struct pb_manager *provider2;
+};
+
+
+static INLINE struct pb_alt_manager *
+pb_alt_manager(struct pb_manager *mgr)
+{
+   assert(mgr);
+   return (struct pb_alt_manager *)mgr;
+}
+
+
+static struct pb_buffer *
+pb_alt_manager_create_buffer(struct pb_manager *_mgr, 
+                             size_t size,
+                             const struct pb_desc *desc)
+{
+   struct pb_alt_manager *mgr = pb_alt_manager(_mgr);
+   struct pb_buffer *buf;
+   
+   buf = mgr->provider1->create_buffer(mgr->provider1, size, desc);
+   if(buf)
+      return buf;
+   
+   buf = mgr->provider2->create_buffer(mgr->provider2, size, desc);
+   return buf;
+}
+
+
+static void
+pb_alt_manager_flush(struct pb_manager *_mgr)
+{
+   struct pb_alt_manager *mgr = pb_alt_manager(_mgr);
+   
+   assert(mgr->provider1->flush);
+   if(mgr->provider1->flush)
+      mgr->provider1->flush(mgr->provider1);
+   
+   assert(mgr->provider2->flush);
+   if(mgr->provider2->flush)
+      mgr->provider2->flush(mgr->provider2);
+}
+
+
+static void
+pb_alt_manager_destroy(struct pb_manager *mgr)
+{
+   FREE(mgr);
+}
+
+
+struct pb_manager *
+pb_alt_manager_create(struct pb_manager *provider1, 
+                      struct pb_manager *provider2)
+{
+   struct pb_alt_manager *mgr;
+
+   if(!provider1 || !provider2)
+      return NULL;
+   
+   mgr = CALLOC_STRUCT(pb_alt_manager);
+   if (!mgr)
+      return NULL;
+
+   mgr->base.destroy = pb_alt_manager_destroy;
+   mgr->base.create_buffer = pb_alt_manager_create_buffer;
+   mgr->base.flush = pb_alt_manager_flush;
+   mgr->provider1 = provider1;
+   mgr->provider2 = provider2;
+      
+   return &mgr->base;
+}
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c
new file mode 100644
index 0000000000..a168853713
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c
@@ -0,0 +1,384 @@
+/**************************************************************************
+ *
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Buffer cache.
+ * 
+ * \author Jose Fonseca <jrfonseca-at-tungstengraphics-dot-com>
+ * \author Thomas Hellström <thomas-at-tungstengraphics-dot-com>
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_thread.h"
+#include "util/u_memory.h"
+#include "util/u_double_list.h"
+#include "util/u_time.h"
+
+#include "pb_buffer.h"
+#include "pb_bufmgr.h"
+
+
+/**
+ * Convenience macro (type safe).
+ */
+#define SUPER(__derived) (&(__derived)->base)
+
+
+struct pb_cache_manager;
+
+
+/**
+ * Wrapper around a pipe buffer which adds delayed destruction.
+ */
+struct pb_cache_buffer
+{
+   struct pb_buffer base;
+   
+   struct pb_buffer *buffer;
+   struct pb_cache_manager *mgr;
+
+   /** Caching time interval */
+   struct util_time start, end;
+
+   struct list_head head;
+};
+
+
+struct pb_cache_manager
+{
+   struct pb_manager base;
+
+   struct pb_manager *provider;
+   unsigned usecs;
+   
+   pipe_mutex mutex;
+   
+   struct list_head delayed;
+   size_t numDelayed;
+};
+
+
+static INLINE struct pb_cache_buffer *
+pb_cache_buffer(struct pb_buffer *buf)
+{
+   assert(buf);
+   return (struct pb_cache_buffer *)buf;
+}
+
+
+static INLINE struct pb_cache_manager *
+pb_cache_manager(struct pb_manager *mgr)
+{
+   assert(mgr);
+   return (struct pb_cache_manager *)mgr;
+}
+
+
+/**
+ * Actually destroy the buffer.
+ */
+static INLINE void
+_pb_cache_buffer_destroy(struct pb_cache_buffer *buf)
+{
+   struct pb_cache_manager *mgr = buf->mgr;
+
+   LIST_DEL(&buf->head);
+   assert(mgr->numDelayed);
+   --mgr->numDelayed;
+   assert(!buf->base.base.refcount);
+   pb_reference(&buf->buffer, NULL);
+   FREE(buf);
+}
+
+
+/**
+ * Free as many cache buffers from the list head as possible. 
+ */
+static void
+_pb_cache_buffer_list_check_free(struct pb_cache_manager *mgr)
+{
+   struct list_head *curr, *next;
+   struct pb_cache_buffer *buf;
+   struct util_time now;
+   
+   util_time_get(&now);
+   
+   curr = mgr->delayed.next;
+   next = curr->next;
+   while(curr != &mgr->delayed) {
+      buf = LIST_ENTRY(struct pb_cache_buffer, curr, head);
+
+      if(!util_time_timeout(&buf->start, &buf->end, &now))
+	 break;
+	 
+      _pb_cache_buffer_destroy(buf);
+
+      curr = next; 
+      next = curr->next;
+   }
+}
+
+
+static void
+pb_cache_buffer_destroy(struct pb_buffer *_buf)
+{
+   struct pb_cache_buffer *buf = pb_cache_buffer(_buf);   
+   struct pb_cache_manager *mgr = buf->mgr;
+
+   pipe_mutex_lock(mgr->mutex);
+   assert(buf->base.base.refcount == 0);
+   
+   _pb_cache_buffer_list_check_free(mgr);
+   
+   util_time_get(&buf->start);
+   util_time_add(&buf->start, mgr->usecs, &buf->end);
+   LIST_ADDTAIL(&buf->head, &mgr->delayed);
+   ++mgr->numDelayed;
+   pipe_mutex_unlock(mgr->mutex);
+}
+
+
+static void *
+pb_cache_buffer_map(struct pb_buffer *_buf, 
+                  unsigned flags)
+{
+   struct pb_cache_buffer *buf = pb_cache_buffer(_buf);   
+   return pb_map(buf->buffer, flags);
+}
+
+
+static void
+pb_cache_buffer_unmap(struct pb_buffer *_buf)
+{
+   struct pb_cache_buffer *buf = pb_cache_buffer(_buf);   
+   pb_unmap(buf->buffer);
+}
+
+
+static enum pipe_error 
+pb_cache_buffer_validate(struct pb_buffer *_buf, 
+                         struct pb_validate *vl,
+                         unsigned flags)
+{
+   struct pb_cache_buffer *buf = pb_cache_buffer(_buf);
+   return pb_validate(buf->buffer, vl, flags);
+}
+
+
+static void
+pb_cache_buffer_fence(struct pb_buffer *_buf, 
+                      struct pipe_fence_handle *fence)
+{
+   struct pb_cache_buffer *buf = pb_cache_buffer(_buf);
+   pb_fence(buf->buffer, fence);
+}
+
+
+static void
+pb_cache_buffer_get_base_buffer(struct pb_buffer *_buf,
+                              struct pb_buffer **base_buf,
+                              unsigned *offset)
+{
+   struct pb_cache_buffer *buf = pb_cache_buffer(_buf);
+   pb_get_base_buffer(buf->buffer, base_buf, offset);
+}
+
+
+const struct pb_vtbl 
+pb_cache_buffer_vtbl = {
+      pb_cache_buffer_destroy,
+      pb_cache_buffer_map,
+      pb_cache_buffer_unmap,
+      pb_cache_buffer_validate,
+      pb_cache_buffer_fence,
+      pb_cache_buffer_get_base_buffer
+};
+
+
+static INLINE boolean
+pb_cache_is_buffer_compat(struct pb_cache_buffer *buf,  
+                          size_t size,
+                          const struct pb_desc *desc)
+{
+   if(buf->base.base.size < size)
+      return FALSE;
+
+   /* be lenient with size */
+   if(buf->base.base.size >= 2*size)
+      return FALSE;
+   
+   if(!pb_check_alignment(desc->alignment, buf->base.base.alignment))
+      return FALSE;
+   
+   if(!pb_check_usage(desc->usage, buf->base.base.usage))
+      return FALSE;
+   
+   return TRUE;
+}
+
+
+static struct pb_buffer *
+pb_cache_manager_create_buffer(struct pb_manager *_mgr, 
+                               size_t size,
+                               const struct pb_desc *desc)
+{
+   struct pb_cache_manager *mgr = pb_cache_manager(_mgr);
+   struct pb_cache_buffer *buf;
+   struct pb_cache_buffer *curr_buf;
+   struct list_head *curr, *next;
+   struct util_time now;
+   
+   pipe_mutex_lock(mgr->mutex);
+
+   buf = NULL;
+   curr = mgr->delayed.next;
+   next = curr->next;
+   
+   /* search in the expired buffers, freeing them in the process */
+   util_time_get(&now);
+   while(curr != &mgr->delayed) {
+      curr_buf = LIST_ENTRY(struct pb_cache_buffer, curr, head);
+      if(!buf && pb_cache_is_buffer_compat(curr_buf, size, desc))
+	 buf = curr_buf;
+      else if(util_time_timeout(&curr_buf->start, &curr_buf->end, &now))
+	 _pb_cache_buffer_destroy(curr_buf);
+      else
+         /* This buffer (and all hereafter) are still hot in cache */
+         break;
+      curr = next; 
+      next = curr->next;
+   }
+
+   /* keep searching in the hot buffers */
+   if(!buf) {
+      while(curr != &mgr->delayed) {
+         curr_buf = LIST_ENTRY(struct pb_cache_buffer, curr, head);
+         if(pb_cache_is_buffer_compat(curr_buf, size, desc)) {
+            buf = curr_buf;
+            break;
+         }
+         /* no need to check the timeout here */
+         curr = next;
+         next = curr->next;
+      }
+   }
+   
+   if(buf) {
+      LIST_DEL(&buf->head);
+      pipe_mutex_unlock(mgr->mutex);
+      ++buf->base.base.refcount;
+      return &buf->base;
+   }
+   
+   pipe_mutex_unlock(mgr->mutex);
+
+   buf = CALLOC_STRUCT(pb_cache_buffer);
+   if(!buf)
+      return NULL;
+   
+   buf->buffer = mgr->provider->create_buffer(mgr->provider, size, desc);
+   if(!buf->buffer) {
+      FREE(buf);
+      return NULL;
+   }
+   
+   assert(buf->buffer->base.refcount >= 1);
+   assert(pb_check_alignment(desc->alignment, buf->buffer->base.alignment));
+   assert(pb_check_usage(desc->usage, buf->buffer->base.usage));
+   assert(buf->buffer->base.size >= size);
+   
+   buf->base.base.refcount = 1;
+   buf->base.base.alignment = buf->buffer->base.alignment;
+   buf->base.base.usage = buf->buffer->base.usage;
+   buf->base.base.size = buf->buffer->base.size;
+   
+   buf->base.vtbl = &pb_cache_buffer_vtbl;
+   buf->mgr = mgr;
+   
+   return &buf->base;
+}
+
+
+static void
+pb_cache_manager_flush(struct pb_manager *_mgr)
+{
+   struct pb_cache_manager *mgr = pb_cache_manager(_mgr);
+   struct list_head *curr, *next;
+   struct pb_cache_buffer *buf;
+
+   pipe_mutex_lock(mgr->mutex);
+   curr = mgr->delayed.next;
+   next = curr->next;
+   while(curr != &mgr->delayed) {
+      buf = LIST_ENTRY(struct pb_cache_buffer, curr, head);
+      _pb_cache_buffer_destroy(buf);
+      curr = next; 
+      next = curr->next;
+   }
+   pipe_mutex_unlock(mgr->mutex);
+   
+   assert(mgr->provider->flush);
+   if(mgr->provider->flush)
+      mgr->provider->flush(mgr->provider);
+}
+
+
+static void
+pb_cache_manager_destroy(struct pb_manager *mgr)
+{
+   pb_cache_manager_flush(mgr);
+   FREE(mgr);
+}
+
+
+struct pb_manager *
+pb_cache_manager_create(struct pb_manager *provider, 
+                     	unsigned usecs) 
+{
+   struct pb_cache_manager *mgr;
+
+   if(!provider)
+      return NULL;
+   
+   mgr = CALLOC_STRUCT(pb_cache_manager);
+   if (!mgr)
+      return NULL;
+
+   mgr->base.destroy = pb_cache_manager_destroy;
+   mgr->base.create_buffer = pb_cache_manager_create_buffer;
+   mgr->base.flush = pb_cache_manager_flush;
+   mgr->provider = provider;
+   mgr->usecs = usecs;
+   LIST_INITHEAD(&mgr->delayed);
+   mgr->numDelayed = 0;
+   pipe_mutex_init(mgr->mutex);
+      
+   return &mgr->base;
+}
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c
new file mode 100644
index 0000000000..26d9c24aec
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c
@@ -0,0 +1,390 @@
+/**************************************************************************
+ *
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Debug buffer manager to detect buffer under- and overflows.
+ * 
+ * \author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_thread.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "util/u_double_list.h"
+#include "util/u_time.h"
+
+#include "pb_buffer.h"
+#include "pb_bufmgr.h"
+
+
+#ifdef DEBUG
+
+
+/**
+ * Convenience macro (type safe).
+ */
+#define SUPER(__derived) (&(__derived)->base)
+
+
+struct pb_debug_manager;
+
+
+/**
+ * Wrapper around a pipe buffer which adds delayed destruction.
+ */
+struct pb_debug_buffer
+{
+   struct pb_buffer base;
+   
+   struct pb_buffer *buffer;
+   struct pb_debug_manager *mgr;
+   
+   size_t underflow_size;
+   size_t overflow_size;
+};
+
+
+struct pb_debug_manager
+{
+   struct pb_manager base;
+
+   struct pb_manager *provider;
+
+   size_t band_size;
+};
+
+
+static INLINE struct pb_debug_buffer *
+pb_debug_buffer(struct pb_buffer *buf)
+{
+   assert(buf);
+   return (struct pb_debug_buffer *)buf;
+}
+
+
+static INLINE struct pb_debug_manager *
+pb_debug_manager(struct pb_manager *mgr)
+{
+   assert(mgr);
+   return (struct pb_debug_manager *)mgr;
+}
+
+
+static const uint8_t random_pattern[32] = {
+   0xaf, 0xcf, 0xa5, 0xa2, 0xc2, 0x63, 0x15, 0x1a, 
+   0x7e, 0xe2, 0x7e, 0x84, 0x15, 0x49, 0xa2, 0x1e,
+   0x49, 0x63, 0xf5, 0x52, 0x74, 0x66, 0x9e, 0xc4, 
+   0x6d, 0xcf, 0x2c, 0x4a, 0x74, 0xe6, 0xfd, 0x94
+};
+
+
+static INLINE void 
+fill_random_pattern(uint8_t *dst, size_t size)
+{
+   size_t i = 0;
+   while(size--) {
+      *dst++ = random_pattern[i++];
+      i &= sizeof(random_pattern) - 1;
+   }
+}
+
+
+static INLINE boolean 
+check_random_pattern(const uint8_t *dst, size_t size, 
+                     size_t *min_ofs, size_t *max_ofs) 
+{
+   boolean result = TRUE;
+   size_t i;
+   *min_ofs = size;
+   *max_ofs = 0;
+   for(i = 0; i < size; ++i) {
+      if(*dst++ != random_pattern[i % sizeof(random_pattern)]) {
+         *min_ofs = MIN2(*min_ofs, i);
+         *max_ofs = MAX2(*max_ofs, i);
+	 result = FALSE;
+      }
+   }
+   return result;
+}
+
+
+static void
+pb_debug_buffer_fill(struct pb_debug_buffer *buf)
+{
+   uint8_t *map;
+   
+   map = pb_map(buf->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+   assert(map);
+   if(map) {
+      fill_random_pattern(map, buf->underflow_size);
+      fill_random_pattern(map + buf->underflow_size + buf->base.base.size, 
+                          buf->overflow_size);
+      pb_unmap(buf->buffer);
+   }
+}
+
+
+/**
+ * Check for under/over flows.
+ * 
+ * Should be called with the buffer unmaped.
+ */
+static void
+pb_debug_buffer_check(struct pb_debug_buffer *buf)
+{
+   uint8_t *map;
+   
+   map = pb_map(buf->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+   assert(map);
+   if(map) {
+      boolean underflow, overflow;
+      size_t min_ofs, max_ofs;
+      
+      underflow = !check_random_pattern(map, buf->underflow_size, 
+                                        &min_ofs, &max_ofs);
+      if(underflow) {
+         debug_printf("buffer underflow (offset -%u%s to -%u bytes) detected\n",
+                      buf->underflow_size - min_ofs,
+                      min_ofs == 0 ? "+" : "",
+                      buf->underflow_size - max_ofs);
+      }
+      
+      overflow = !check_random_pattern(map + buf->underflow_size + buf->base.base.size, 
+                                       buf->overflow_size, 
+                                       &min_ofs, &max_ofs);
+      if(overflow) {
+         debug_printf("buffer overflow (size %u plus offset %u to %u%s bytes) detected\n",
+                      buf->base.base.size,
+                      min_ofs,
+                      max_ofs,
+                      max_ofs == buf->overflow_size - 1 ? "+" : "");
+      }
+      
+      debug_assert(!underflow && !overflow);
+
+      /* re-fill if not aborted */
+      if(underflow)
+         fill_random_pattern(map, buf->underflow_size);
+      if(overflow)
+         fill_random_pattern(map + buf->underflow_size + buf->base.base.size, 
+                             buf->overflow_size);
+
+      pb_unmap(buf->buffer);
+   }
+}
+
+
+static void
+pb_debug_buffer_destroy(struct pb_buffer *_buf)
+{
+   struct pb_debug_buffer *buf = pb_debug_buffer(_buf);  
+   
+   assert(!buf->base.base.refcount);
+   
+   pb_debug_buffer_check(buf);
+
+   pb_reference(&buf->buffer, NULL);
+   FREE(buf);
+}
+
+
+static void *
+pb_debug_buffer_map(struct pb_buffer *_buf, 
+                    unsigned flags)
+{
+   struct pb_debug_buffer *buf = pb_debug_buffer(_buf);
+   void *map;
+   
+   pb_debug_buffer_check(buf);
+
+   map = pb_map(buf->buffer, flags);
+   if(!map)
+      return NULL;
+   
+   return (uint8_t *)map + buf->underflow_size;
+}
+
+
+static void
+pb_debug_buffer_unmap(struct pb_buffer *_buf)
+{
+   struct pb_debug_buffer *buf = pb_debug_buffer(_buf);   
+   pb_unmap(buf->buffer);
+   
+   pb_debug_buffer_check(buf);
+}
+
+
+static void
+pb_debug_buffer_get_base_buffer(struct pb_buffer *_buf,
+                                struct pb_buffer **base_buf,
+                                unsigned *offset)
+{
+   struct pb_debug_buffer *buf = pb_debug_buffer(_buf);
+   pb_get_base_buffer(buf->buffer, base_buf, offset);
+   *offset += buf->underflow_size;
+}
+
+
+static enum pipe_error 
+pb_debug_buffer_validate(struct pb_buffer *_buf, 
+                         struct pb_validate *vl,
+                         unsigned flags)
+{
+   struct pb_debug_buffer *buf = pb_debug_buffer(_buf);
+   
+   pb_debug_buffer_check(buf);
+
+   return pb_validate(buf->buffer, vl, flags);
+}
+
+
+static void
+pb_debug_buffer_fence(struct pb_buffer *_buf, 
+                      struct pipe_fence_handle *fence)
+{
+   struct pb_debug_buffer *buf = pb_debug_buffer(_buf);
+   pb_fence(buf->buffer, fence);
+}
+
+
+const struct pb_vtbl 
+pb_debug_buffer_vtbl = {
+      pb_debug_buffer_destroy,
+      pb_debug_buffer_map,
+      pb_debug_buffer_unmap,
+      pb_debug_buffer_validate,
+      pb_debug_buffer_fence,
+      pb_debug_buffer_get_base_buffer
+};
+
+
+static struct pb_buffer *
+pb_debug_manager_create_buffer(struct pb_manager *_mgr, 
+                               size_t size,
+                               const struct pb_desc *desc)
+{
+   struct pb_debug_manager *mgr = pb_debug_manager(_mgr);
+   struct pb_debug_buffer *buf;
+   struct pb_desc real_desc;
+   size_t real_size;
+   
+   buf = CALLOC_STRUCT(pb_debug_buffer);
+   if(!buf)
+      return NULL;
+   
+   real_size = size + 2*mgr->band_size;
+   real_desc = *desc;
+   real_desc.usage |= PIPE_BUFFER_USAGE_CPU_WRITE;
+   real_desc.usage |= PIPE_BUFFER_USAGE_CPU_READ;
+
+   buf->buffer = mgr->provider->create_buffer(mgr->provider, 
+                                              real_size, 
+                                              &real_desc);
+   if(!buf->buffer) {
+      FREE(buf);
+      return NULL;
+   }
+   
+   assert(buf->buffer->base.refcount >= 1);
+   assert(pb_check_alignment(real_desc.alignment, buf->buffer->base.alignment));
+   assert(pb_check_usage(real_desc.usage, buf->buffer->base.usage));
+   assert(buf->buffer->base.size >= real_size);
+   
+   buf->base.base.refcount = 1;
+   buf->base.base.alignment = desc->alignment;
+   buf->base.base.usage = desc->usage;
+   buf->base.base.size = size;
+   
+   buf->base.vtbl = &pb_debug_buffer_vtbl;
+   buf->mgr = mgr;
+
+   buf->underflow_size = mgr->band_size;
+   buf->overflow_size = buf->buffer->base.size - buf->underflow_size - size;
+   
+   pb_debug_buffer_fill(buf);
+   
+   return &buf->base;
+}
+
+
+static void
+pb_debug_manager_flush(struct pb_manager *_mgr)
+{
+   struct pb_debug_manager *mgr = pb_debug_manager(_mgr);
+   assert(mgr->provider->flush);
+   if(mgr->provider->flush)
+      mgr->provider->flush(mgr->provider);
+}
+
+
+static void
+pb_debug_manager_destroy(struct pb_manager *_mgr)
+{
+   struct pb_debug_manager *mgr = pb_debug_manager(_mgr);
+   mgr->provider->destroy(mgr->provider);
+   FREE(mgr);
+}
+
+
+struct pb_manager *
+pb_debug_manager_create(struct pb_manager *provider, size_t band_size) 
+{
+   struct pb_debug_manager *mgr;
+
+   if(!provider)
+      return NULL;
+   
+   mgr = CALLOC_STRUCT(pb_debug_manager);
+   if (!mgr)
+      return NULL;
+
+   mgr->base.destroy = pb_debug_manager_destroy;
+   mgr->base.create_buffer = pb_debug_manager_create_buffer;
+   mgr->base.flush = pb_debug_manager_flush;
+   mgr->provider = provider;
+   mgr->band_size = band_size;
+      
+   return &mgr->base;
+}
+
+
+#else /* !DEBUG */
+
+
+struct pb_manager *
+pb_debug_manager_create(struct pb_manager *provider, size_t band_size) 
+{
+   return provider;
+}
+
+
+#endif /* !DEBUG */
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c
new file mode 100644
index 0000000000..47e9fee533
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c
@@ -0,0 +1,148 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/**
+ * \file
+ * A buffer manager that wraps buffers in fenced buffers.
+ * 
+ * \author Jose Fonseca <jrfonseca@tungstengraphics.dot.com>
+ */
+
+
+#include "pipe/p_debug.h"
+#include "util/u_memory.h"
+
+#include "pb_buffer.h"
+#include "pb_buffer_fenced.h"
+#include "pb_bufmgr.h"
+
+
+struct fenced_pb_manager
+{
+   struct pb_manager base;
+
+   struct pb_manager *provider;
+   
+   struct fenced_buffer_list *fenced_list;
+};
+
+
+static INLINE struct fenced_pb_manager *
+fenced_pb_manager(struct pb_manager *mgr)
+{
+   assert(mgr);
+   return (struct fenced_pb_manager *)mgr;
+}
+
+
+static struct pb_buffer *
+fenced_bufmgr_create_buffer(struct pb_manager *mgr, 
+                            size_t size,
+                            const struct pb_desc *desc)
+{
+   struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr);
+   struct pb_buffer *buf;
+   struct pb_buffer *fenced_buf;
+
+   /* check for free buffers before allocating new ones */
+   fenced_buffer_list_check_free(fenced_mgr->fenced_list, 0);
+   
+   buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc);
+   if(!buf) {
+      /* try harder to get a buffer */
+      fenced_buffer_list_check_free(fenced_mgr->fenced_list, 1);
+      
+      buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc);
+      if(!buf) {
+         /* give up */
+         return NULL;
+      }
+   }
+   
+   fenced_buf = fenced_buffer_create(fenced_mgr->fenced_list, buf);
+   if(!fenced_buf) {
+      pb_reference(&buf, NULL);
+   }
+   
+   return fenced_buf;
+}
+
+
+static void
+fenced_bufmgr_flush(struct pb_manager *mgr)
+{
+   struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr);
+
+   fenced_buffer_list_check_free(fenced_mgr->fenced_list, TRUE);
+
+   assert(fenced_mgr->provider->flush);
+   if(fenced_mgr->provider->flush)
+      fenced_mgr->provider->flush(fenced_mgr->provider);
+}
+
+
+static void
+fenced_bufmgr_destroy(struct pb_manager *mgr)
+{
+   struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr);
+
+   fenced_buffer_list_destroy(fenced_mgr->fenced_list);
+
+   if(fenced_mgr->provider)
+      fenced_mgr->provider->destroy(fenced_mgr->provider);
+   
+   FREE(fenced_mgr);
+}
+
+
+struct pb_manager *
+fenced_bufmgr_create(struct pb_manager *provider, 
+                     struct pb_fence_ops *ops) 
+{
+   struct fenced_pb_manager *fenced_mgr;
+
+   if(!provider)
+      return NULL;
+   
+   fenced_mgr = CALLOC_STRUCT(fenced_pb_manager);
+   if (!fenced_mgr)
+      return NULL;
+
+   fenced_mgr->base.destroy = fenced_bufmgr_destroy;
+   fenced_mgr->base.create_buffer = fenced_bufmgr_create_buffer;
+   fenced_mgr->base.flush = fenced_bufmgr_flush;
+
+   fenced_mgr->provider = provider;
+   fenced_mgr->fenced_list = fenced_buffer_list_create(ops);
+   if(!fenced_mgr->fenced_list) {
+      FREE(fenced_mgr);
+      return NULL;
+   }
+      
+   return &fenced_mgr->base;
+}
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c
new file mode 100644
index 0000000000..2f5a5d8ea0
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c
@@ -0,0 +1,322 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * Buffer manager using the old texture memory manager.
+ * 
+ * \author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_defines.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_thread.h"
+#include "util/u_memory.h"
+#include "util/u_double_list.h"
+#include "util/u_mm.h"
+#include "pb_buffer.h"
+#include "pb_bufmgr.h"
+
+
+/**
+ * Convenience macro (type safe).
+ */
+#define SUPER(__derived) (&(__derived)->base)
+
+
+struct mm_pb_manager
+{
+   struct pb_manager base;
+   
+   pipe_mutex mutex;
+   
+   size_t size;
+   struct mem_block *heap;
+   
+   size_t align2;
+   
+   struct pb_buffer *buffer;
+   void *map;
+};
+
+
+static INLINE struct mm_pb_manager *
+mm_pb_manager(struct pb_manager *mgr)
+{
+   assert(mgr);
+   return (struct mm_pb_manager *)mgr;
+}
+
+
+struct mm_buffer
+{
+   struct pb_buffer base;
+   
+   struct mm_pb_manager *mgr;
+   
+   struct mem_block *block;
+};
+
+
+static INLINE struct mm_buffer *
+mm_buffer(struct pb_buffer *buf)
+{
+   assert(buf);
+   return (struct mm_buffer *)buf;
+}
+
+
+static void
+mm_buffer_destroy(struct pb_buffer *buf)
+{
+   struct mm_buffer *mm_buf = mm_buffer(buf);
+   struct mm_pb_manager *mm = mm_buf->mgr;
+   
+   assert(buf->base.refcount == 0);
+   
+   pipe_mutex_lock(mm->mutex);
+   u_mmFreeMem(mm_buf->block);
+   FREE(buf);
+   pipe_mutex_unlock(mm->mutex);
+}
+
+
+static void *
+mm_buffer_map(struct pb_buffer *buf,
+              unsigned flags)
+{
+   struct mm_buffer *mm_buf = mm_buffer(buf);
+   struct mm_pb_manager *mm = mm_buf->mgr;
+
+   return (unsigned char *) mm->map + mm_buf->block->ofs;
+}
+
+
+static void
+mm_buffer_unmap(struct pb_buffer *buf)
+{
+   /* No-op */
+}
+
+
+static enum pipe_error 
+mm_buffer_validate(struct pb_buffer *buf, 
+                   struct pb_validate *vl,
+                   unsigned flags)
+{
+   struct mm_buffer *mm_buf = mm_buffer(buf);
+   struct mm_pb_manager *mm = mm_buf->mgr;
+   return pb_validate(mm->buffer, vl, flags);
+}
+
+
+static void
+mm_buffer_fence(struct pb_buffer *buf, 
+                struct pipe_fence_handle *fence)
+{
+   struct mm_buffer *mm_buf = mm_buffer(buf);
+   struct mm_pb_manager *mm = mm_buf->mgr;
+   pb_fence(mm->buffer, fence);
+}
+
+
+static void
+mm_buffer_get_base_buffer(struct pb_buffer *buf,
+                          struct pb_buffer **base_buf,
+                          unsigned *offset)
+{
+   struct mm_buffer *mm_buf = mm_buffer(buf);
+   struct mm_pb_manager *mm = mm_buf->mgr;
+   pb_get_base_buffer(mm->buffer, base_buf, offset);
+   *offset += mm_buf->block->ofs;
+}
+
+
+static const struct pb_vtbl 
+mm_buffer_vtbl = {
+      mm_buffer_destroy,
+      mm_buffer_map,
+      mm_buffer_unmap,
+      mm_buffer_validate,
+      mm_buffer_fence,
+      mm_buffer_get_base_buffer
+};
+
+
+static struct pb_buffer *
+mm_bufmgr_create_buffer(struct pb_manager *mgr, 
+                        size_t size,
+                        const struct pb_desc *desc)
+{
+   struct mm_pb_manager *mm = mm_pb_manager(mgr);
+   struct mm_buffer *mm_buf;
+
+   /* We don't handle alignments larger then the one initially setup */
+   assert(pb_check_alignment(desc->alignment, 1 << mm->align2));
+   if(!pb_check_alignment(desc->alignment, 1 << mm->align2))
+      return NULL;
+   
+   pipe_mutex_lock(mm->mutex);
+
+   mm_buf = CALLOC_STRUCT(mm_buffer);
+   if (!mm_buf) {
+      pipe_mutex_unlock(mm->mutex);
+      return NULL;
+   }
+
+   mm_buf->base.base.refcount = 1;
+   mm_buf->base.base.alignment = desc->alignment;
+   mm_buf->base.base.usage = desc->usage;
+   mm_buf->base.base.size = size;
+   
+   mm_buf->base.vtbl = &mm_buffer_vtbl;
+   
+   mm_buf->mgr = mm;
+   
+   mm_buf->block = u_mmAllocMem(mm->heap, size, mm->align2, 0);
+   if(!mm_buf->block) {
+      debug_printf("warning: heap full\n");
+#if 0
+      mmDumpMemInfo(mm->heap);
+#endif
+      
+      mm_buf->block = u_mmAllocMem(mm->heap, size, mm->align2, 0);
+      if(!mm_buf->block) {
+         FREE(mm_buf);
+         pipe_mutex_unlock(mm->mutex);
+         return NULL;
+      }
+   }
+   
+   /* Some sanity checks */
+   assert(0 <= (unsigned)mm_buf->block->ofs && (unsigned)mm_buf->block->ofs < mm->size);
+   assert(size <= (unsigned)mm_buf->block->size && (unsigned)mm_buf->block->ofs + (unsigned)mm_buf->block->size <= mm->size);
+   
+   pipe_mutex_unlock(mm->mutex);
+   return SUPER(mm_buf);
+}
+
+
+static void
+mm_bufmgr_flush(struct pb_manager *mgr)
+{
+   /* No-op */
+}
+
+
+static void
+mm_bufmgr_destroy(struct pb_manager *mgr)
+{
+   struct mm_pb_manager *mm = mm_pb_manager(mgr);
+   
+   pipe_mutex_lock(mm->mutex);
+
+   u_mmDestroy(mm->heap);
+   
+   pb_unmap(mm->buffer);
+   pb_reference(&mm->buffer, NULL);
+   
+   pipe_mutex_unlock(mm->mutex);
+   
+   FREE(mgr);
+}
+
+
+struct pb_manager *
+mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, 
+                             size_t size, size_t align2) 
+{
+   struct mm_pb_manager *mm;
+
+   if(!buffer)
+      return NULL;
+   
+   mm = CALLOC_STRUCT(mm_pb_manager);
+   if (!mm)
+      return NULL;
+
+   mm->base.destroy = mm_bufmgr_destroy;
+   mm->base.create_buffer = mm_bufmgr_create_buffer;
+   mm->base.flush = mm_bufmgr_flush;
+
+   mm->size = size;
+   mm->align2 = align2; /* 64-byte alignment */
+
+   pipe_mutex_init(mm->mutex);
+
+   mm->buffer = buffer; 
+
+   mm->map = pb_map(mm->buffer, 
+		    PIPE_BUFFER_USAGE_CPU_READ |
+		    PIPE_BUFFER_USAGE_CPU_WRITE);
+   if(!mm->map)
+      goto failure;
+
+   mm->heap = u_mmInit(0, size); 
+   if (!mm->heap)
+      goto failure;
+
+   return SUPER(mm);
+   
+failure:
+if(mm->heap)
+   u_mmDestroy(mm->heap);
+   if(mm->map)
+      pb_unmap(mm->buffer);
+   if(mm)
+      FREE(mm);
+   return NULL;
+}
+
+
+struct pb_manager *
+mm_bufmgr_create(struct pb_manager *provider, 
+                 size_t size, size_t align2) 
+{
+   struct pb_buffer *buffer;
+   struct pb_manager *mgr;
+   struct pb_desc desc;
+
+   if(!provider)
+      return NULL;
+   
+   memset(&desc, 0, sizeof(desc));
+   desc.alignment = 1 << align2;
+   
+   buffer = provider->create_buffer(provider, size, &desc); 
+   if (!buffer)
+      return NULL;
+   
+   mgr = mm_bufmgr_create_from_buffer(buffer, size, align2);
+   if (!mgr) {
+      pb_reference(&buffer, NULL);
+      return NULL;
+   }
+
+  return mgr;
+}
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c
new file mode 100644
index 0000000000..ba02a84e62
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_ondemand.c
@@ -0,0 +1,303 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * A variation of malloc buffers which get transferred to real graphics memory
+ * when there is an attempt to validate them. 
+ * 
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_debug.h"
+#include "util/u_memory.h"
+#include "pb_buffer.h"
+#include "pb_bufmgr.h"
+
+
+struct pb_ondemand_manager;
+
+
+struct pb_ondemand_buffer 
+{
+   struct pb_buffer base;
+   
+   struct pb_ondemand_manager *mgr;
+   
+   /** Regular malloc'ed memory */
+   void *data;
+   unsigned mapcount;
+   
+   /** Real buffer */
+   struct pb_buffer *buffer;
+   size_t size;
+   struct pb_desc desc;
+};
+
+
+struct pb_ondemand_manager
+{
+   struct pb_manager base;
+   
+   struct pb_manager *provider;
+};
+
+
+extern const struct pb_vtbl pb_ondemand_buffer_vtbl;
+
+static INLINE struct pb_ondemand_buffer *
+pb_ondemand_buffer(struct pb_buffer *buf)
+{
+   assert(buf);
+   assert(buf->vtbl == &pb_ondemand_buffer_vtbl);
+   return (struct pb_ondemand_buffer *)buf;
+}
+
+static INLINE struct pb_ondemand_manager *
+pb_ondemand_manager(struct pb_manager *mgr)
+{
+   assert(mgr);
+   return (struct pb_ondemand_manager *)mgr;
+}
+
+
+static void
+pb_ondemand_buffer_destroy(struct pb_buffer *_buf)
+{
+   struct pb_ondemand_buffer *buf = pb_ondemand_buffer(_buf);
+   
+   pb_reference(&buf->buffer, NULL);
+   
+   align_free(buf->data);
+   
+   FREE(buf);
+}
+
+
+static void *
+pb_ondemand_buffer_map(struct pb_buffer *_buf, 
+                       unsigned flags)
+{
+   struct pb_ondemand_buffer *buf = pb_ondemand_buffer(_buf);
+
+   if(buf->buffer) {
+      assert(!buf->data);
+      return pb_map(buf->buffer, flags);
+   }
+   else {
+      assert(buf->data);
+      ++buf->mapcount;
+      return buf->data;
+   }
+}
+
+
+static void
+pb_ondemand_buffer_unmap(struct pb_buffer *_buf)
+{
+   struct pb_ondemand_buffer *buf = pb_ondemand_buffer(_buf);
+
+   if(buf->buffer) {
+      assert(!buf->data);
+      pb_unmap(buf->buffer);
+   }
+   else {
+      assert(buf->data);
+      assert(buf->mapcount);
+      if(buf->mapcount)
+         --buf->mapcount;
+   }
+}
+
+
+static enum pipe_error 
+pb_ondemand_buffer_instantiate(struct pb_ondemand_buffer *buf)
+{
+   if(!buf->buffer) {
+      struct pb_manager *provider = buf->mgr->provider;
+      uint8_t *map;
+      
+      assert(!buf->mapcount);
+      
+      buf->buffer = provider->create_buffer(provider, buf->size, &buf->desc);
+      if(!buf->buffer)
+         return PIPE_ERROR_OUT_OF_MEMORY;
+      
+      map = pb_map(buf->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+      if(!map) {
+         pb_reference(&buf->buffer, NULL);
+         return PIPE_ERROR;
+      }
+      
+      memcpy(map, buf->data, buf->size);
+      
+      pb_unmap(buf->buffer);
+      
+      if(!buf->mapcount) {
+         FREE(buf->data);
+         buf->data = NULL;
+      }
+   }
+   
+   return PIPE_OK;
+}
+
+static enum pipe_error 
+pb_ondemand_buffer_validate(struct pb_buffer *_buf, 
+                            struct pb_validate *vl,
+                            unsigned flags)
+{
+   struct pb_ondemand_buffer *buf = pb_ondemand_buffer(_buf);
+   enum pipe_error ret; 
+
+   assert(!buf->mapcount);
+   if(buf->mapcount)
+      return PIPE_ERROR;
+
+   ret = pb_ondemand_buffer_instantiate(buf);
+   if(ret != PIPE_OK)
+      return ret;
+   
+   return pb_validate(buf->buffer, vl, flags);
+}
+
+
+static void
+pb_ondemand_buffer_fence(struct pb_buffer *_buf, 
+                         struct pipe_fence_handle *fence)
+{
+   struct pb_ondemand_buffer *buf = pb_ondemand_buffer(_buf);
+   
+   assert(buf->buffer);
+   if(!buf->buffer)
+      return;
+   
+   pb_fence(buf->buffer, fence);
+}
+
+
+static void
+pb_ondemand_buffer_get_base_buffer(struct pb_buffer *_buf,
+                                   struct pb_buffer **base_buf,
+                                   unsigned *offset)
+{
+   struct pb_ondemand_buffer *buf = pb_ondemand_buffer(_buf);
+
+   if(pb_ondemand_buffer_instantiate(buf) != PIPE_OK) {
+      assert(0);
+      *base_buf = &buf->base;
+      *offset = 0;
+      return;
+   }
+
+   pb_get_base_buffer(buf->buffer, base_buf, offset);
+}
+
+
+const struct pb_vtbl 
+pb_ondemand_buffer_vtbl = {
+      pb_ondemand_buffer_destroy,
+      pb_ondemand_buffer_map,
+      pb_ondemand_buffer_unmap,
+      pb_ondemand_buffer_validate,
+      pb_ondemand_buffer_fence,
+      pb_ondemand_buffer_get_base_buffer
+};
+
+
+static struct pb_buffer *
+pb_ondemand_manager_create_buffer(struct pb_manager *_mgr, 
+                                  size_t size,
+                                  const struct pb_desc *desc) 
+{
+   struct pb_ondemand_manager *mgr = pb_ondemand_manager(_mgr);
+   struct pb_ondemand_buffer *buf;
+   
+   buf = CALLOC_STRUCT(pb_ondemand_buffer);
+   if(!buf)
+      return NULL;
+
+   buf->base.base.refcount = 1;
+   buf->base.base.alignment = desc->alignment;
+   buf->base.base.usage = desc->usage;
+   buf->base.base.size = size;
+   buf->base.vtbl = &pb_ondemand_buffer_vtbl;
+   
+   buf->mgr = mgr;
+   
+   buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment);
+   if(!buf->data) {
+      FREE(buf);
+      return NULL;
+   }
+   
+   buf->size = size;
+   buf->desc = *desc;
+
+   return &buf->base;
+}
+
+
+static void
+pb_ondemand_manager_flush(struct pb_manager *_mgr) 
+{
+   struct pb_ondemand_manager *mgr = pb_ondemand_manager(_mgr);
+   
+   mgr->provider->flush(mgr->provider);
+}
+
+
+static void
+pb_ondemand_manager_destroy(struct pb_manager *_mgr) 
+{
+   struct pb_ondemand_manager *mgr = pb_ondemand_manager(_mgr);
+
+   FREE(mgr);
+}
+
+
+struct pb_manager *
+pb_ondemand_manager_create(struct pb_manager *provider) 
+{
+   struct pb_ondemand_manager *mgr;
+
+   if(!provider)
+      return NULL;
+   
+   mgr = CALLOC_STRUCT(pb_ondemand_manager);
+   if(!mgr)
+      return NULL;
+   
+   mgr->base.destroy = pb_ondemand_manager_destroy;
+   mgr->base.create_buffer = pb_ondemand_manager_create_buffer;
+   mgr->base.flush = pb_ondemand_manager_flush;
+   
+   mgr->provider = provider;
+
+   return &mgr->base;
+}
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c
new file mode 100644
index 0000000000..a6ff37653e
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c
@@ -0,0 +1,321 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/**
+ * \file
+ * Batch buffer pool management.
+ * 
+ * \author Jose Fonseca <jrfonseca-at-tungstengraphics-dot-com>
+ * \author Thomas Hellström <thomas-at-tungstengraphics-dot-com>
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_thread.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "util/u_double_list.h"
+
+#include "pb_buffer.h"
+#include "pb_bufmgr.h"
+
+
+/**
+ * Convenience macro (type safe).
+ */
+#define SUPER(__derived) (&(__derived)->base)
+
+
+struct pool_pb_manager
+{
+   struct pb_manager base;
+   
+   pipe_mutex mutex;
+   
+   size_t bufSize;
+   size_t bufAlign;
+   
+   size_t numFree;
+   size_t numTot;
+   
+   struct list_head free;
+   
+   struct pb_buffer *buffer;
+   void *map;
+   
+   struct pool_buffer *bufs;
+};
+
+
+static INLINE struct pool_pb_manager *
+pool_pb_manager(struct pb_manager *mgr)
+{
+   assert(mgr);
+   return (struct pool_pb_manager *)mgr;
+}
+
+
+struct pool_buffer
+{
+   struct pb_buffer base;
+   
+   struct pool_pb_manager *mgr;
+   
+   struct list_head head;
+   
+   size_t start;
+};
+
+
+static INLINE struct pool_buffer *
+pool_buffer(struct pb_buffer *buf)
+{
+   assert(buf);
+   return (struct pool_buffer *)buf;
+}
+
+
+
+static void
+pool_buffer_destroy(struct pb_buffer *buf)
+{
+   struct pool_buffer *pool_buf = pool_buffer(buf);
+   struct pool_pb_manager *pool = pool_buf->mgr;
+   
+   assert(pool_buf->base.base.refcount == 0);
+
+   pipe_mutex_lock(pool->mutex);
+   LIST_ADD(&pool_buf->head, &pool->free);
+   pool->numFree++;
+   pipe_mutex_unlock(pool->mutex);
+}
+
+
+static void *
+pool_buffer_map(struct pb_buffer *buf, unsigned flags)
+{
+   struct pool_buffer *pool_buf = pool_buffer(buf);
+   struct pool_pb_manager *pool = pool_buf->mgr;
+   void *map;
+
+   pipe_mutex_lock(pool->mutex);
+   map = (unsigned char *) pool->map + pool_buf->start;
+   pipe_mutex_unlock(pool->mutex);
+   return map;
+}
+
+
+static void
+pool_buffer_unmap(struct pb_buffer *buf)
+{
+   /* No-op */
+}
+
+
+static enum pipe_error 
+pool_buffer_validate(struct pb_buffer *buf, 
+                     struct pb_validate *vl,
+                     unsigned flags)
+{
+   struct pool_buffer *pool_buf = pool_buffer(buf);
+   struct pool_pb_manager *pool = pool_buf->mgr;
+   return pb_validate(pool->buffer, vl, flags);
+}
+
+
+static void
+pool_buffer_fence(struct pb_buffer *buf, 
+                  struct pipe_fence_handle *fence)
+{
+   struct pool_buffer *pool_buf = pool_buffer(buf);
+   struct pool_pb_manager *pool = pool_buf->mgr;
+   pb_fence(pool->buffer, fence);
+}
+
+
+static void
+pool_buffer_get_base_buffer(struct pb_buffer *buf,
+                            struct pb_buffer **base_buf,
+                            unsigned *offset)
+{
+   struct pool_buffer *pool_buf = pool_buffer(buf);
+   struct pool_pb_manager *pool = pool_buf->mgr;
+   pb_get_base_buffer(pool->buffer, base_buf, offset);
+   *offset += pool_buf->start;
+}
+
+
+static const struct pb_vtbl 
+pool_buffer_vtbl = {
+      pool_buffer_destroy,
+      pool_buffer_map,
+      pool_buffer_unmap,
+      pool_buffer_validate,
+      pool_buffer_fence,
+      pool_buffer_get_base_buffer
+};
+
+
+static struct pb_buffer *
+pool_bufmgr_create_buffer(struct pb_manager *mgr,
+                          size_t size,
+                          const struct pb_desc *desc)
+{
+   struct pool_pb_manager *pool = pool_pb_manager(mgr);
+   struct pool_buffer *pool_buf;
+   struct list_head *item;
+
+   assert(size == pool->bufSize);
+   assert(pool->bufAlign % desc->alignment == 0);
+   
+   pipe_mutex_lock(pool->mutex);
+
+   if (pool->numFree == 0) {
+      pipe_mutex_unlock(pool->mutex);
+      debug_printf("warning: out of fixed size buffer objects\n");
+      return NULL;
+   }
+
+   item = pool->free.next;
+
+   if (item == &pool->free) {
+      pipe_mutex_unlock(pool->mutex);
+      debug_printf("error: fixed size buffer pool corruption\n");
+      return NULL;
+   }
+
+   LIST_DEL(item);
+   --pool->numFree;
+
+   pipe_mutex_unlock(pool->mutex);
+   
+   pool_buf = LIST_ENTRY(struct pool_buffer, item, head);
+   assert(pool_buf->base.base.refcount == 0);
+   pool_buf->base.base.refcount = 1;
+   pool_buf->base.base.alignment = desc->alignment;
+   pool_buf->base.base.usage = desc->usage;
+   
+   return SUPER(pool_buf);
+}
+
+
+static void
+pool_bufmgr_flush(struct pb_manager *mgr)
+{
+   /* No-op */
+}
+
+
+static void
+pool_bufmgr_destroy(struct pb_manager *mgr)
+{
+   struct pool_pb_manager *pool = pool_pb_manager(mgr);
+   pipe_mutex_lock(pool->mutex);
+
+   FREE(pool->bufs);
+   
+   pb_unmap(pool->buffer);
+   pb_reference(&pool->buffer, NULL);
+   
+   pipe_mutex_unlock(pool->mutex);
+   
+   FREE(mgr);
+}
+
+
+struct pb_manager *
+pool_bufmgr_create(struct pb_manager *provider, 
+                   size_t numBufs, 
+                   size_t bufSize,
+                   const struct pb_desc *desc) 
+{
+   struct pool_pb_manager *pool;
+   struct pool_buffer *pool_buf;
+   size_t i;
+
+   if(!provider)
+      return NULL;
+   
+   pool = CALLOC_STRUCT(pool_pb_manager);
+   if (!pool)
+      return NULL;
+
+   pool->base.destroy = pool_bufmgr_destroy;
+   pool->base.create_buffer = pool_bufmgr_create_buffer;
+   pool->base.flush = pool_bufmgr_flush;
+
+   LIST_INITHEAD(&pool->free);
+
+   pool->numTot = numBufs;
+   pool->numFree = numBufs;
+   pool->bufSize = bufSize;
+   pool->bufAlign = desc->alignment; 
+   
+   pipe_mutex_init(pool->mutex);
+
+   pool->buffer = provider->create_buffer(provider, numBufs*bufSize, desc); 
+   if (!pool->buffer)
+      goto failure;
+
+   pool->map = pb_map(pool->buffer,
+                          PIPE_BUFFER_USAGE_CPU_READ |
+                          PIPE_BUFFER_USAGE_CPU_WRITE);
+   if(!pool->map)
+      goto failure;
+
+   pool->bufs = (struct pool_buffer *)CALLOC(numBufs, sizeof(*pool->bufs));
+   if (!pool->bufs)
+      goto failure;
+
+   pool_buf = pool->bufs;
+   for (i = 0; i < numBufs; ++i) {
+      pool_buf->base.base.refcount = 0;
+      pool_buf->base.base.alignment = 0;
+      pool_buf->base.base.usage = 0;
+      pool_buf->base.base.size = bufSize;
+      pool_buf->base.vtbl = &pool_buffer_vtbl;
+      pool_buf->mgr = pool;
+      pool_buf->start = i * bufSize;
+      LIST_ADDTAIL(&pool_buf->head, &pool->free);
+      pool_buf++;
+   }
+
+   return SUPER(pool);
+   
+failure:
+   if(pool->bufs)
+      FREE(pool->bufs);
+   if(pool->map)
+      pb_unmap(pool->buffer);
+   if(pool->buffer)
+      pb_reference(&pool->buffer, NULL);
+   if(pool)
+      FREE(pool);
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c
new file mode 100644
index 0000000000..9b9fedccb4
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c
@@ -0,0 +1,584 @@
+/**************************************************************************
+ *
+ * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, FREE of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * S-lab pool implementation.
+ * 
+ * @sa http://en.wikipedia.org/wiki/Slab_allocation
+ * 
+ * @author Thomas Hellstrom <thomas-at-tungstengraphics-dot-com>
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_error.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_thread.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "util/u_double_list.h"
+#include "util/u_time.h"
+
+#include "pb_buffer.h"
+#include "pb_bufmgr.h"
+
+
+struct pb_slab;
+
+
+/**
+ * Buffer in a slab.
+ * 
+ * Sub-allocation of a contiguous buffer.
+ */
+struct pb_slab_buffer
+{
+   struct pb_buffer base;
+   
+   struct pb_slab *slab;
+   
+   struct list_head head;
+   
+   unsigned mapCount;
+   
+   /** Offset relative to the start of the slab buffer. */
+   size_t start;
+   
+   /** Use when validating, to signal that all mappings are finished */
+   /* TODO: Actually validation does not reach this stage yet */
+   pipe_condvar event;
+};
+
+
+/**
+ * Slab -- a contiguous piece of memory. 
+ */
+struct pb_slab
+{
+   struct list_head head;
+   struct list_head freeBuffers;
+   size_t numBuffers;
+   size_t numFree;
+   
+   struct pb_slab_buffer *buffers;
+   struct pb_slab_manager *mgr;
+   
+   /** Buffer from the provider */
+   struct pb_buffer *bo;
+   
+   void *virtual;   
+};
+
+
+/**
+ * It adds/removes slabs as needed in order to meet the allocation/destruction 
+ * of individual buffers.
+ */
+struct pb_slab_manager 
+{
+   struct pb_manager base;
+   
+   /** From where we get our buffers */
+   struct pb_manager *provider;
+   
+   /** Size of the buffers we hand on downstream */
+   size_t bufSize;
+   
+   /** Size of the buffers we request upstream */
+   size_t slabSize;
+   
+   /** 
+    * Alignment, usage to be used to allocate the slab buffers.
+    * 
+    * We can only provide buffers which are consistent (in alignment, usage) 
+    * with this description.   
+    */
+   struct pb_desc desc;
+
+   /** 
+    * Partial slabs
+    * 
+    * Full slabs are not stored in any list. Empty slabs are destroyed 
+    * immediatly.
+    */
+   struct list_head slabs;
+   
+   pipe_mutex mutex;
+};
+
+
+/**
+ * Wrapper around several slabs, therefore capable of handling buffers of 
+ * multiple sizes. 
+ * 
+ * This buffer manager just dispatches buffer allocations to the appropriate slab
+ * manager, according to the requested buffer size, or by passes the slab 
+ * managers altogether for even greater sizes.
+ * 
+ * The data of this structure remains constant after
+ * initialization and thus needs no mutex protection.
+ */
+struct pb_slab_range_manager 
+{
+   struct pb_manager base;
+
+   struct pb_manager *provider;
+   
+   size_t minBufSize;
+   size_t maxBufSize;
+   
+   /** @sa pb_slab_manager::desc */ 
+   struct pb_desc desc;
+   
+   unsigned numBuckets;
+   size_t *bucketSizes;
+   
+   /** Array of pb_slab_manager, one for each bucket size */
+   struct pb_manager **buckets;
+};
+
+
+static INLINE struct pb_slab_buffer *
+pb_slab_buffer(struct pb_buffer *buf)
+{
+   assert(buf);
+   return (struct pb_slab_buffer *)buf;
+}
+
+
+static INLINE struct pb_slab_manager *
+pb_slab_manager(struct pb_manager *mgr)
+{
+   assert(mgr);
+   return (struct pb_slab_manager *)mgr;
+}
+
+
+static INLINE struct pb_slab_range_manager *
+pb_slab_range_manager(struct pb_manager *mgr)
+{
+   assert(mgr);
+   return (struct pb_slab_range_manager *)mgr;
+}
+
+
+/**
+ * Delete a buffer from the slab delayed list and put
+ * it on the slab FREE list.
+ */
+static void
+pb_slab_buffer_destroy(struct pb_buffer *_buf)
+{
+   struct pb_slab_buffer *buf = pb_slab_buffer(_buf);
+   struct pb_slab *slab = buf->slab;
+   struct pb_slab_manager *mgr = slab->mgr;
+   struct list_head *list = &buf->head;
+
+   pipe_mutex_lock(mgr->mutex);
+   
+   assert(buf->base.base.refcount == 0);
+   
+   buf->mapCount = 0;
+
+   LIST_DEL(list);
+   LIST_ADDTAIL(list, &slab->freeBuffers);
+   slab->numFree++;
+
+   if (slab->head.next == &slab->head)
+      LIST_ADDTAIL(&slab->head, &mgr->slabs);
+
+   /* If the slab becomes totally empty, free it */
+   if (slab->numFree == slab->numBuffers) {
+      list = &slab->head;
+      LIST_DELINIT(list);
+      pb_reference(&slab->bo, NULL);
+      FREE(slab->buffers);
+      FREE(slab);
+   }
+
+   pipe_mutex_unlock(mgr->mutex);
+}
+
+
+static void *
+pb_slab_buffer_map(struct pb_buffer *_buf, 
+                   unsigned flags)
+{
+   struct pb_slab_buffer *buf = pb_slab_buffer(_buf);
+
+   ++buf->mapCount;
+   return (void *) ((uint8_t *) buf->slab->virtual + buf->start);
+}
+
+
+static void
+pb_slab_buffer_unmap(struct pb_buffer *_buf)
+{
+   struct pb_slab_buffer *buf = pb_slab_buffer(_buf);
+
+   --buf->mapCount;
+   if (buf->mapCount == 0) 
+       pipe_condvar_broadcast(buf->event);
+}
+
+
+static enum pipe_error 
+pb_slab_buffer_validate(struct pb_buffer *_buf, 
+                         struct pb_validate *vl,
+                         unsigned flags)
+{
+   struct pb_slab_buffer *buf = pb_slab_buffer(_buf);
+   return pb_validate(buf->slab->bo, vl, flags);
+}
+
+
+static void
+pb_slab_buffer_fence(struct pb_buffer *_buf, 
+                      struct pipe_fence_handle *fence)
+{
+   struct pb_slab_buffer *buf = pb_slab_buffer(_buf);
+   pb_fence(buf->slab->bo, fence);
+}
+
+
+static void
+pb_slab_buffer_get_base_buffer(struct pb_buffer *_buf,
+                               struct pb_buffer **base_buf,
+                               unsigned *offset)
+{
+   struct pb_slab_buffer *buf = pb_slab_buffer(_buf);
+   pb_get_base_buffer(buf->slab->bo, base_buf, offset);
+   *offset += buf->start;
+}
+
+
+static const struct pb_vtbl 
+pb_slab_buffer_vtbl = {
+      pb_slab_buffer_destroy,
+      pb_slab_buffer_map,
+      pb_slab_buffer_unmap,
+      pb_slab_buffer_validate,
+      pb_slab_buffer_fence,
+      pb_slab_buffer_get_base_buffer
+};
+
+
+/**
+ * Create a new slab.
+ * 
+ * Called when we ran out of free slabs.
+ */
+static enum pipe_error
+pb_slab_create(struct pb_slab_manager *mgr)
+{
+   struct pb_slab *slab;
+   struct pb_slab_buffer *buf;
+   unsigned numBuffers;
+   unsigned i;
+   enum pipe_error ret;
+
+   slab = CALLOC_STRUCT(pb_slab);
+   if (!slab)
+      return PIPE_ERROR_OUT_OF_MEMORY;
+
+   slab->bo = mgr->provider->create_buffer(mgr->provider, mgr->slabSize, &mgr->desc);
+   if(!slab->bo) {
+      ret = PIPE_ERROR_OUT_OF_MEMORY;
+      goto out_err0;
+   }
+
+   /* Note down the slab virtual address. All mappings are accessed directly 
+    * through this address so it is required that the buffer is pinned. */
+   slab->virtual = pb_map(slab->bo, 
+                          PIPE_BUFFER_USAGE_CPU_READ |
+                          PIPE_BUFFER_USAGE_CPU_WRITE);
+   if(!slab->virtual) {
+      ret = PIPE_ERROR_OUT_OF_MEMORY;
+      goto out_err1;
+   }
+   pb_unmap(slab->bo);
+
+   numBuffers = slab->bo->base.size / mgr->bufSize;
+
+   slab->buffers = CALLOC(numBuffers, sizeof(*slab->buffers));
+   if (!slab->buffers) {
+      ret = PIPE_ERROR_OUT_OF_MEMORY;
+      goto out_err1;
+   }
+
+   LIST_INITHEAD(&slab->head);
+   LIST_INITHEAD(&slab->freeBuffers);
+   slab->numBuffers = numBuffers;
+   slab->numFree = 0;
+   slab->mgr = mgr;
+
+   buf = slab->buffers;
+   for (i=0; i < numBuffers; ++i) {
+      buf->base.base.refcount = 0;
+      buf->base.base.size = mgr->bufSize;
+      buf->base.base.alignment = 0;
+      buf->base.base.usage = 0;
+      buf->base.vtbl = &pb_slab_buffer_vtbl;
+      buf->slab = slab;
+      buf->start = i* mgr->bufSize;
+      buf->mapCount = 0;
+      pipe_condvar_init(buf->event);
+      LIST_ADDTAIL(&buf->head, &slab->freeBuffers);
+      slab->numFree++;
+      buf++;
+   }
+
+   /* Add this slab to the list of partial slabs */
+   LIST_ADDTAIL(&slab->head, &mgr->slabs);
+
+   return PIPE_OK;
+
+out_err1: 
+   pb_reference(&slab->bo, NULL);
+out_err0: 
+   FREE(slab);
+   return ret;
+}
+
+
+static struct pb_buffer *
+pb_slab_manager_create_buffer(struct pb_manager *_mgr,
+                              size_t size,
+                              const struct pb_desc *desc)
+{
+   struct pb_slab_manager *mgr = pb_slab_manager(_mgr);
+   static struct pb_slab_buffer *buf;
+   struct pb_slab *slab;
+   struct list_head *list;
+
+   /* check size */
+   assert(size <= mgr->bufSize);
+   if(size > mgr->bufSize)
+      return NULL;
+   
+   /* check if we can provide the requested alignment */
+   assert(pb_check_alignment(desc->alignment, mgr->desc.alignment));
+   if(!pb_check_alignment(desc->alignment, mgr->desc.alignment))
+      return NULL;
+   assert(pb_check_alignment(desc->alignment, mgr->bufSize));
+   if(!pb_check_alignment(desc->alignment, mgr->bufSize))
+      return NULL;
+
+   assert(pb_check_usage(desc->usage, mgr->desc.usage));
+   if(!pb_check_usage(desc->usage, mgr->desc.usage))
+      return NULL;
+
+   pipe_mutex_lock(mgr->mutex);
+   
+   /* Create a new slab, if we run out of partial slabs */
+   if (mgr->slabs.next == &mgr->slabs) {
+      (void) pb_slab_create(mgr);
+      if (mgr->slabs.next == &mgr->slabs) {
+	 pipe_mutex_unlock(mgr->mutex);
+	 return NULL;
+      }
+   }
+   
+   /* Allocate the buffer from a partial (or just created) slab */
+   list = mgr->slabs.next;
+   slab = LIST_ENTRY(struct pb_slab, list, head);
+   
+   /* If totally full remove from the partial slab list */
+   if (--slab->numFree == 0)
+      LIST_DELINIT(list);
+
+   list = slab->freeBuffers.next;
+   LIST_DELINIT(list);
+
+   pipe_mutex_unlock(mgr->mutex);
+   buf = LIST_ENTRY(struct pb_slab_buffer, list, head);
+   
+   ++buf->base.base.refcount;
+   buf->base.base.alignment = desc->alignment;
+   buf->base.base.usage = desc->usage;
+   
+   return &buf->base;
+}
+
+
+static void
+pb_slab_manager_flush(struct pb_manager *_mgr)
+{
+   struct pb_slab_manager *mgr = pb_slab_manager(_mgr);
+
+   assert(mgr->provider->flush);
+   if(mgr->provider->flush)
+      mgr->provider->flush(mgr->provider);
+}
+
+
+static void
+pb_slab_manager_destroy(struct pb_manager *_mgr)
+{
+   struct pb_slab_manager *mgr = pb_slab_manager(_mgr);
+
+   /* TODO: cleanup all allocated buffers */
+   FREE(mgr);
+}
+
+
+struct pb_manager *
+pb_slab_manager_create(struct pb_manager *provider,
+                       size_t bufSize,
+                       size_t slabSize,
+                       const struct pb_desc *desc)
+{
+   struct pb_slab_manager *mgr;
+
+   mgr = CALLOC_STRUCT(pb_slab_manager);
+   if (!mgr)
+      return NULL;
+
+   mgr->base.destroy = pb_slab_manager_destroy;
+   mgr->base.create_buffer = pb_slab_manager_create_buffer;
+   mgr->base.flush = pb_slab_manager_flush;
+
+   mgr->provider = provider;
+   mgr->bufSize = bufSize;
+   mgr->slabSize = slabSize;
+   mgr->desc = *desc;
+
+   LIST_INITHEAD(&mgr->slabs);
+   
+   pipe_mutex_init(mgr->mutex);
+
+   return &mgr->base;
+}
+
+
+static struct pb_buffer *
+pb_slab_range_manager_create_buffer(struct pb_manager *_mgr,
+                                    size_t size,
+                                    const struct pb_desc *desc)
+{
+   struct pb_slab_range_manager *mgr = pb_slab_range_manager(_mgr);
+   size_t bufSize;
+   unsigned i;
+
+   bufSize = mgr->minBufSize;
+   for (i = 0; i < mgr->numBuckets; ++i) {
+      if(bufSize >= size)
+	 return mgr->buckets[i]->create_buffer(mgr->buckets[i], size, desc);
+      bufSize *= 2;
+   }
+
+   /* Fall back to allocate a buffer object directly from the provider. */
+   return mgr->provider->create_buffer(mgr->provider, size, desc);
+}
+
+
+static void
+pb_slab_range_manager_flush(struct pb_manager *_mgr)
+{
+   struct pb_slab_range_manager *mgr = pb_slab_range_manager(_mgr);
+
+   /* Individual slabs don't hold any temporary buffers so no need to call them */
+   
+   assert(mgr->provider->flush);
+   if(mgr->provider->flush)
+      mgr->provider->flush(mgr->provider);
+}
+
+
+static void
+pb_slab_range_manager_destroy(struct pb_manager *_mgr)
+{
+   struct pb_slab_range_manager *mgr = pb_slab_range_manager(_mgr);
+   unsigned i;
+   
+   for (i = 0; i < mgr->numBuckets; ++i)
+      mgr->buckets[i]->destroy(mgr->buckets[i]);
+   FREE(mgr->buckets);
+   FREE(mgr->bucketSizes);
+   FREE(mgr);
+}
+
+
+struct pb_manager *
+pb_slab_range_manager_create(struct pb_manager *provider,
+                             size_t minBufSize,
+                             size_t maxBufSize,
+                             size_t slabSize,
+                             const struct pb_desc *desc)
+{
+   struct pb_slab_range_manager *mgr;
+   size_t bufSize;
+   unsigned i;
+
+   if(!provider)
+      return NULL;
+   
+   mgr = CALLOC_STRUCT(pb_slab_range_manager);
+   if (!mgr)
+      goto out_err0;
+
+   mgr->base.destroy = pb_slab_range_manager_destroy;
+   mgr->base.create_buffer = pb_slab_range_manager_create_buffer;
+   mgr->base.flush = pb_slab_range_manager_flush;
+
+   mgr->provider = provider;
+   mgr->minBufSize = minBufSize;
+   mgr->maxBufSize = maxBufSize;
+
+   mgr->numBuckets = 1;
+   bufSize = minBufSize;
+   while(bufSize < maxBufSize) {
+      bufSize *= 2;
+      ++mgr->numBuckets;
+   }
+   
+   mgr->buckets = CALLOC(mgr->numBuckets, sizeof(*mgr->buckets));
+   if (!mgr->buckets)
+      goto out_err1;
+
+   bufSize = minBufSize;
+   for (i = 0; i < mgr->numBuckets; ++i) {
+      mgr->buckets[i] = pb_slab_manager_create(provider, bufSize, slabSize, desc);
+      if(!mgr->buckets[i])
+	 goto out_err2;
+      bufSize *= 2;
+   }
+
+   return &mgr->base;
+
+out_err2: 
+   for (i = 0; i < mgr->numBuckets; ++i)
+      if(mgr->buckets[i])
+	    mgr->buckets[i]->destroy(mgr->buckets[i]);
+   FREE(mgr->buckets);
+out_err1: 
+   FREE(mgr);
+out_err0:
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.c b/src/gallium/auxiliary/pipebuffer/pb_validate.c
new file mode 100644
index 0000000000..94532bb4ce
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_validate.c
@@ -0,0 +1,193 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Buffer validation.
+ * 
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_error.h"
+#include "util/u_memory.h"
+#include "pipe/p_debug.h"
+
+#include "pb_buffer.h"
+#include "pb_buffer_fenced.h"
+#include "pb_validate.h"
+
+
+#define PB_VALIDATE_INITIAL_SIZE 1 /* 512 */ 
+
+
+struct pb_validate_entry
+{
+   struct pb_buffer *buf;
+   unsigned flags;
+};
+
+
+struct pb_validate
+{
+   struct pb_validate_entry *entries;
+   unsigned used;
+   unsigned size;
+};
+
+
+enum pipe_error
+pb_validate_add_buffer(struct pb_validate *vl,
+                       struct pb_buffer *buf,
+                       unsigned flags)
+{
+   assert(buf);
+   if(!buf)
+      return PIPE_ERROR;
+
+   assert(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE);
+   assert(!(flags & ~PIPE_BUFFER_USAGE_GPU_READ_WRITE));
+   flags &= PIPE_BUFFER_USAGE_GPU_READ_WRITE;
+
+   /* We only need to store one reference for each buffer, so avoid storing
+    * consecutive references for the same buffer. It might not be the most 
+    * common pattern, but it is easy to implement.
+    */
+   if(vl->used && vl->entries[vl->used - 1].buf == buf) {
+      vl->entries[vl->used - 1].flags |= flags;
+      return PIPE_OK;
+   }
+   
+   /* Grow the table */
+   if(vl->used == vl->size) {
+      unsigned new_size;
+      struct pb_validate_entry *new_entries;
+      
+      new_size = vl->size * 2;
+      if(!new_size)
+	 return PIPE_ERROR_OUT_OF_MEMORY;
+
+      new_entries = (struct pb_validate_entry *)REALLOC(vl->entries,
+                                                        vl->size*sizeof(struct pb_validate_entry),
+                                                        new_size*sizeof(struct pb_validate_entry));
+      if(!new_entries)
+         return PIPE_ERROR_OUT_OF_MEMORY;
+      
+      memset(new_entries + vl->size, 0, (new_size - vl->size)*sizeof(struct pb_validate_entry));
+      
+      vl->size = new_size;
+      vl->entries = new_entries;
+   }
+   
+   assert(!vl->entries[vl->used].buf);
+   pb_reference(&vl->entries[vl->used].buf, buf);
+   vl->entries[vl->used].flags = flags;
+   ++vl->used;
+   
+   return PIPE_OK;
+}
+
+
+enum pipe_error
+pb_validate_foreach(struct pb_validate *vl,
+                    enum pipe_error (*callback)(struct pb_buffer *buf, void *data),
+                    void *data)
+{
+   unsigned i;
+   for(i = 0; i < vl->used; ++i) {
+      enum pipe_error ret;
+      ret = callback(vl->entries[i].buf, data);
+      if(ret != PIPE_OK)
+         return ret;
+   }
+   return PIPE_OK;
+}
+
+
+enum pipe_error
+pb_validate_validate(struct pb_validate *vl) 
+{
+   unsigned i;
+   
+   for(i = 0; i < vl->used; ++i) {
+      enum pipe_error ret;
+      ret = pb_validate(vl->entries[i].buf, vl, vl->entries[i].flags);
+      if(ret != PIPE_OK) {
+         while(i--)
+            pb_validate(vl->entries[i].buf, NULL, 0);
+         return ret;
+      }
+   }
+
+   return PIPE_OK;
+}
+
+
+void
+pb_validate_fence(struct pb_validate *vl,
+                  struct pipe_fence_handle *fence)
+{
+   unsigned i;
+   for(i = 0; i < vl->used; ++i) {
+      pb_fence(vl->entries[i].buf, fence);
+      pb_reference(&vl->entries[i].buf, NULL);
+   }
+   vl->used = 0;
+}
+
+
+void
+pb_validate_destroy(struct pb_validate *vl)
+{
+   unsigned i;
+   for(i = 0; i < vl->used; ++i)
+      pb_reference(&vl->entries[i].buf, NULL);
+   FREE(vl->entries);
+   FREE(vl);
+}
+
+
+struct pb_validate *
+pb_validate_create()
+{
+   struct pb_validate *vl;
+   
+   vl = CALLOC_STRUCT(pb_validate);
+   if(!vl)
+      return NULL;
+   
+   vl->size = PB_VALIDATE_INITIAL_SIZE;
+   vl->entries = (struct pb_validate_entry *)CALLOC(vl->size, sizeof(struct pb_validate_entry));
+   if(!vl->entries) {
+      FREE(vl);
+      return NULL;
+   }
+
+   return vl;
+}
+
diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.h b/src/gallium/auxiliary/pipebuffer/pb_validate.h
new file mode 100644
index 0000000000..dfb84df1ce
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/pb_validate.h
@@ -0,0 +1,97 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Buffer validation.
+ * 
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef PB_VALIDATE_H_
+#define PB_VALIDATE_H_
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_error.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+struct pb_buffer;
+struct pipe_fence_handle;
+
+
+/**
+ * Buffer validation list.
+ * 
+ * It holds a list of buffers to be validated and fenced when flushing.
+ */
+struct pb_validate;
+
+
+enum pipe_error
+pb_validate_add_buffer(struct pb_validate *vl,
+                       struct pb_buffer *buf,
+                       unsigned flags);
+
+enum pipe_error
+pb_validate_foreach(struct pb_validate *vl,
+                    enum pipe_error (*callback)(struct pb_buffer *buf, void *data),
+                    void *data);
+
+/**
+ * Validate all buffers for hardware access.
+ * 
+ * Should be called right before issuing commands to the hardware.
+ */
+enum pipe_error
+pb_validate_validate(struct pb_validate *vl);
+
+/**
+ * Fence all buffers and clear the list.
+ * 
+ * Should be called right after issuing commands to the hardware.
+ */
+void
+pb_validate_fence(struct pb_validate *vl,
+                  struct pipe_fence_handle *fence);
+
+struct pb_validate *
+pb_validate_create(void);
+
+void
+pb_validate_destroy(struct pb_validate *vl);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /*PB_VALIDATE_H_*/
diff --git a/src/gallium/auxiliary/rtasm/Makefile b/src/gallium/auxiliary/rtasm/Makefile
new file mode 100644
index 0000000000..252dc5274a
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/Makefile
@@ -0,0 +1,16 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = rtasm
+
+C_SOURCES = \
+	rtasm_cpu.c \
+	rtasm_execmem.c \
+	rtasm_x86sse.c \
+	rtasm_ppc.c \
+	rtasm_ppc_spe.c
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/rtasm/SConscript b/src/gallium/auxiliary/rtasm/SConscript
new file mode 100644
index 0000000000..eb48368acc
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/SConscript
@@ -0,0 +1,13 @@
+Import('*')
+
+rtasm = env.ConvenienceLibrary(
+	target = 'rtasm',
+	source = [
+		'rtasm_cpu.c',
+		'rtasm_execmem.c',
+		'rtasm_x86sse.c',
+		'rtasm_ppc.c',
+		'rtasm_ppc_spe.c',
+	])
+
+auxiliaries.insert(0, rtasm)
diff --git a/src/gallium/auxiliary/rtasm/rtasm_cpu.c b/src/gallium/auxiliary/rtasm/rtasm_cpu.c
new file mode 100644
index 0000000000..5499018b21
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/rtasm_cpu.c
@@ -0,0 +1,65 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_debug.h"
+#include "rtasm_cpu.h"
+
+
+static boolean rtasm_sse_enabled(void)
+{
+   static boolean firsttime = 1;
+   static boolean enabled;
+   
+   /* This gets called quite often at the moment:
+    */
+   if (firsttime) {
+      enabled =  !debug_get_bool_option("GALLIUM_NOSSE", FALSE);
+      firsttime = FALSE;
+   }
+   return enabled;
+}
+
+int rtasm_cpu_has_sse(void)
+{
+   /* FIXME: actually detect this at run-time */
+#if defined(PIPE_ARCH_X86)
+   return rtasm_sse_enabled();
+#else
+   return 0;
+#endif
+}
+
+int rtasm_cpu_has_sse2(void) 
+{
+   /* FIXME: actually detect this at run-time */
+#if defined(PIPE_ARCH_X86)
+   return rtasm_sse_enabled();
+#else
+   return 0;
+#endif
+}
diff --git a/src/gallium/auxiliary/rtasm/rtasm_cpu.h b/src/gallium/auxiliary/rtasm/rtasm_cpu.h
new file mode 100644
index 0000000000..ebc71634fd
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/rtasm_cpu.h
@@ -0,0 +1,42 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Runtime detection of CPU capabilities.
+ */
+
+#ifndef _RTASM_CPU_H_
+#define _RTASM_CPU_H_
+
+
+int rtasm_cpu_has_sse(void);
+
+int rtasm_cpu_has_sse2(void);
+
+
+#endif /* _RTASM_CPU_H_ */
diff --git a/src/gallium/auxiliary/rtasm/rtasm_execmem.c b/src/gallium/auxiliary/rtasm/rtasm_execmem.c
new file mode 100644
index 0000000000..be7433baf8
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/rtasm_execmem.c
@@ -0,0 +1,137 @@
+/**************************************************************************
+ *
+ * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * \file exemem.c
+ * Functions for allocating executable memory.
+ *
+ * \author Keith Whitwell
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_thread.h"
+#include "util/u_memory.h"
+
+#include "rtasm_execmem.h"
+
+
+#if defined(PIPE_OS_LINUX)
+
+
+/*
+ * Allocate a large block of memory which can hold code then dole it out
+ * in pieces by means of the generic memory manager code.
+ */
+
+#include <unistd.h>
+#include <sys/mman.h>
+#include "pipe/p_thread.h"
+#include "util/u_mm.h"
+
+#define EXEC_HEAP_SIZE (10*1024*1024)
+
+pipe_static_mutex(exec_mutex);
+
+static struct mem_block *exec_heap = NULL;
+static unsigned char *exec_mem = NULL;
+
+
+static void
+init_heap(void)
+{
+   if (!exec_heap)
+      exec_heap = u_mmInit( 0, EXEC_HEAP_SIZE );
+   
+   if (!exec_mem)
+      exec_mem = (unsigned char *) mmap(0, EXEC_HEAP_SIZE, 
+					PROT_EXEC | PROT_READ | PROT_WRITE, 
+					MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+}
+
+
+void *
+rtasm_exec_malloc(size_t size)
+{
+   struct mem_block *block = NULL;
+   void *addr = NULL;
+
+   pipe_mutex_lock(exec_mutex);
+
+   init_heap();
+
+   if (exec_heap) {
+      size = (size + 31) & ~31;  /* next multiple of 32 bytes */
+      block = u_mmAllocMem( exec_heap, size, 5, 0 ); /* 5 -> 32-byte alignment */
+   }
+
+   if (block)
+      addr = exec_mem + block->ofs;
+   else 
+      debug_printf("rtasm_exec_malloc failed\n");
+   
+   pipe_mutex_unlock(exec_mutex);
+   
+   return addr;
+}
+
+ 
+void 
+rtasm_exec_free(void *addr)
+{
+   pipe_mutex_lock(exec_mutex);
+
+   if (exec_heap) {
+      struct mem_block *block = u_mmFindBlock(exec_heap, (unsigned char *)addr - exec_mem);
+   
+      if (block)
+	 u_mmFreeMem(block);
+   }
+
+   pipe_mutex_unlock(exec_mutex);
+}
+
+
+#else /* PIPE_OS_LINUX */
+
+/*
+ * Just use regular memory.
+ */
+
+void *
+rtasm_exec_malloc(size_t size)
+{
+   return MALLOC( size );
+}
+
+ 
+void 
+rtasm_exec_free(void *addr)
+{
+   FREE(addr);
+}
+
+
+#endif /* PIPE_OS_LINUX */
diff --git a/src/gallium/auxiliary/rtasm/rtasm_execmem.h b/src/gallium/auxiliary/rtasm/rtasm_execmem.h
new file mode 100644
index 0000000000..155c6d34e0
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/rtasm_execmem.h
@@ -0,0 +1,45 @@
+/**************************************************************************
+ *
+ * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file exemem.c
+ * Functions for allocating executable memory.
+ *
+ * \author Keith Whitwell
+ */
+
+#ifndef _RTASM_EXECMEM_H_
+#define _RTASM_EXECMEM_H_
+
+#include "pipe/p_compiler.h"
+
+
+extern void *
+rtasm_exec_malloc( size_t size );
+
+
+extern void 
+rtasm_exec_free( void *addr );
+
+
+#endif
diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc.c b/src/gallium/auxiliary/rtasm/rtasm_ppc.c
new file mode 100644
index 0000000000..1bb9026205
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/rtasm_ppc.c
@@ -0,0 +1,1077 @@
+/**************************************************************************
+ *
+ * Copyright (C) 2008 Tungsten Graphics, Inc.   All Rights Reserved.
+ * Copyright (C) 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * PPC code generation.
+ * For reference, see http://www.power.org/resources/reading/PowerISA_V2.05.pdf
+ * ABI info: http://www.cs.utsa.edu/~whaley/teach/cs6463FHPO/LEC/lec12_ho.pdf
+ *
+ * Other PPC refs:
+ * http://www-01.ibm.com/chips/techlib/techlib.nsf/techdocs/852569B20050FF778525699600719DF2
+ * http://www.ibm.com/developerworks/eserver/library/es-archguide-v2.html
+ * http://www.freescale.com/files/product/doc/MPCFPE32B.pdf
+ *
+ * \author Brian Paul
+ */
+
+
+#include <stdio.h>
+#include "util/u_memory.h"
+#include "pipe/p_debug.h"
+#include "rtasm_execmem.h"
+#include "rtasm_ppc.h"
+
+
+void
+ppc_init_func(struct ppc_function *p)
+{
+   uint i;
+
+   memset(p, 0, sizeof(*p));
+
+   p->num_inst = 0;
+   p->max_inst = 100; /* first guess at buffer size */
+   p->store = rtasm_exec_malloc(p->max_inst * PPC_INST_SIZE);
+   p->reg_used = 0x0;
+   p->fp_used = 0x0;
+   p->vec_used = 0x0;
+
+   p->print = FALSE;
+   p->indent = 0;
+
+   /* only allow using gp registers 3..12 for now */
+   for (i = 0; i < 3; i++)
+      ppc_reserve_register(p, i);
+   for (i = 12; i < PPC_NUM_REGS; i++)
+      ppc_reserve_register(p, i);
+}
+
+
+void
+ppc_release_func(struct ppc_function *p)
+{
+   assert(p->num_inst <= p->max_inst);
+   if (p->store != NULL) {
+      rtasm_exec_free(p->store);
+   }
+   p->store = NULL;
+}
+
+
+uint
+ppc_num_instructions(const struct ppc_function *p)
+{
+   return p->num_inst;
+}
+
+
+void (*ppc_get_func(struct ppc_function *p))(void)
+{
+#if 0
+   DUMP_END();
+   if (DISASSEM && p->store)
+      debug_printf("disassemble %p %p\n", p->store, p->csr);
+
+   if (p->store == p->error_overflow)
+      return (void (*)(void)) NULL;
+   else
+#endif
+      return (void (*)(void)) p->store;
+}
+
+
+void
+ppc_dump_func(const struct ppc_function *p)
+{
+   uint i;
+   for (i = 0; i < p->num_inst; i++) {
+      debug_printf("%3u: 0x%08x\n", i, p->store[i]);
+   }
+}
+
+
+void
+ppc_print_code(struct ppc_function *p, boolean enable)
+{
+   p->print = enable;
+}
+
+
+void
+ppc_indent(struct ppc_function *p, int spaces)
+{
+   p->indent += spaces;
+}
+
+
+static void
+indent(const struct ppc_function *p)
+{
+   int i;
+   for (i = 0; i < p->indent; i++) {
+      putchar(' ');
+   }
+}
+
+
+void
+ppc_comment(struct ppc_function *p, int rel_indent, const char *s)
+{
+   if (p->print) {
+      p->indent += rel_indent;
+      indent(p);
+      p->indent -= rel_indent;
+      printf("# %s\n", s);
+   }
+}
+
+
+/**
+ * Mark a register as being unavailable.
+ */
+int
+ppc_reserve_register(struct ppc_function *p, int reg)
+{
+   assert(reg < PPC_NUM_REGS);
+   p->reg_used |= (1 << reg);
+   return reg;
+}
+
+
+/**
+ * Allocate a general purpose register.
+ * \return register index or -1 if none left.
+ */
+int
+ppc_allocate_register(struct ppc_function *p)
+{
+   unsigned i;
+   for (i = 0; i < PPC_NUM_REGS; i++) {
+      const uint64_t mask = 1 << i;
+      if ((p->reg_used & mask) == 0) {
+         p->reg_used |= mask;
+         return i;
+      }
+   }
+   printf("OUT OF PPC registers!\n");
+   return -1;
+}
+
+
+/**
+ * Mark the given general purpose register as "unallocated".
+ */
+void
+ppc_release_register(struct ppc_function *p, int reg)
+{
+   assert(reg < PPC_NUM_REGS);
+   assert(p->reg_used & (1 << reg));
+   p->reg_used &= ~(1 << reg);
+}
+
+
+/**
+ * Allocate a floating point register.
+ * \return register index or -1 if none left.
+ */
+int
+ppc_allocate_fp_register(struct ppc_function *p)
+{
+   unsigned i;
+   for (i = 0; i < PPC_NUM_FP_REGS; i++) {
+      const uint64_t mask = 1 << i;
+      if ((p->fp_used & mask) == 0) {
+         p->fp_used |= mask;
+         return i;
+      }
+   }
+   printf("OUT OF PPC FP registers!\n");
+   return -1;
+}
+
+
+/**
+ * Mark the given floating point register as "unallocated".
+ */
+void
+ppc_release_fp_register(struct ppc_function *p, int reg)
+{
+   assert(reg < PPC_NUM_FP_REGS);
+   assert(p->fp_used & (1 << reg));
+   p->fp_used &= ~(1 << reg);
+}
+
+
+/**
+ * Allocate a vector register.
+ * \return register index or -1 if none left.
+ */
+int
+ppc_allocate_vec_register(struct ppc_function *p)
+{
+   unsigned i;
+   for (i = 0; i < PPC_NUM_VEC_REGS; i++) {
+      const uint64_t mask = 1 << i;
+      if ((p->vec_used & mask) == 0) {
+         p->vec_used |= mask;
+         return i;
+      }
+   }
+   printf("OUT OF PPC VEC registers!\n");
+   return -1;
+}
+
+
+/**
+ * Mark the given vector register as "unallocated".
+ */
+void
+ppc_release_vec_register(struct ppc_function *p, int reg)
+{
+   assert(reg < PPC_NUM_VEC_REGS);
+   assert(p->vec_used & (1 << reg));
+   p->vec_used &= ~(1 << reg);
+}
+
+
+/**
+ * Append instruction to instruction buffer.  Grow buffer if out of room.
+ */
+static void
+emit_instruction(struct ppc_function *p, uint32_t inst_bits)
+{
+   if (!p->store)
+      return;  /* out of memory, drop the instruction */
+
+   if (p->num_inst == p->max_inst) {
+      /* allocate larger buffer */
+      uint32_t *newbuf;
+      p->max_inst *= 2;  /* 2x larger */
+      newbuf = rtasm_exec_malloc(p->max_inst * PPC_INST_SIZE);
+      if (newbuf) {
+         memcpy(newbuf, p->store, p->num_inst * PPC_INST_SIZE);
+      }
+      rtasm_exec_free(p->store);
+      p->store = newbuf;
+      if (!p->store) {
+         /* out of memory */
+         p->num_inst = 0;
+         return;
+      }
+   }
+
+   p->store[p->num_inst++] = inst_bits;
+}
+
+
+union vx_inst {
+   uint32_t bits;
+   struct {
+      unsigned op:6;
+      unsigned vD:5;
+      unsigned vA:5;
+      unsigned vB:5;
+      unsigned op2:11;
+   } inst;
+};
+
+static INLINE void
+emit_vx(struct ppc_function *p, uint op2, uint vD, uint vA, uint vB,
+        const char *format, boolean transpose)
+{
+   union vx_inst inst;
+   inst.inst.op = 4;
+   inst.inst.vD = vD;
+   inst.inst.vA = vA;
+   inst.inst.vB = vB;
+   inst.inst.op2 = op2;
+   emit_instruction(p, inst.bits);
+   if (p->print) {
+      indent(p);
+      if (transpose)
+         printf(format, vD, vB, vA);
+      else
+         printf(format, vD, vA, vB);
+   }
+}
+
+
+union vxr_inst {
+   uint32_t bits;
+   struct {
+      unsigned op:6;
+      unsigned vD:5;
+      unsigned vA:5;
+      unsigned vB:5;
+      unsigned rC:1;
+      unsigned op2:10;
+   } inst;
+};
+
+static INLINE void
+emit_vxr(struct ppc_function *p, uint op2, uint vD, uint vA, uint vB,
+         const char *format)
+{
+   union vxr_inst inst;
+   inst.inst.op = 4;
+   inst.inst.vD = vD;
+   inst.inst.vA = vA;
+   inst.inst.vB = vB;
+   inst.inst.rC = 0;
+   inst.inst.op2 = op2;
+   emit_instruction(p, inst.bits);
+   if (p->print) {
+      indent(p);
+      printf(format, vD, vA, vB);
+   }
+}
+
+
+union va_inst {
+   uint32_t bits;
+   struct {
+      unsigned op:6;
+      unsigned vD:5;
+      unsigned vA:5;
+      unsigned vB:5;
+      unsigned vC:5;
+      unsigned op2:6;
+   } inst;
+};
+
+static INLINE void
+emit_va(struct ppc_function *p, uint op2, uint vD, uint vA, uint vB, uint vC,
+        const char *format)
+{
+   union va_inst inst;
+   inst.inst.op = 4;
+   inst.inst.vD = vD;
+   inst.inst.vA = vA;
+   inst.inst.vB = vB;
+   inst.inst.vC = vC;
+   inst.inst.op2 = op2;
+   emit_instruction(p, inst.bits);
+   if (p->print) {
+      indent(p);
+      printf(format, vD, vA, vB, vC);
+   }
+}
+
+
+union i_inst {
+   uint32_t bits;
+   struct {
+      unsigned op:6;
+      unsigned li:24;
+      unsigned aa:1;
+      unsigned lk:1;
+   } inst;
+};
+
+static INLINE void
+emit_i(struct ppc_function *p, uint op, uint li, uint aa, uint lk)
+{
+   union i_inst inst;
+   inst.inst.op = op;
+   inst.inst.li = li;
+   inst.inst.aa = aa;
+   inst.inst.lk = lk;
+   emit_instruction(p, inst.bits);
+}
+
+
+union xl_inst {
+   uint32_t bits;
+   struct {
+      unsigned op:6;
+      unsigned bo:5;
+      unsigned bi:5;
+      unsigned unused:3;
+      unsigned bh:2;
+      unsigned op2:10;
+      unsigned lk:1;
+   } inst;
+};
+
+static INLINE void
+emit_xl(struct ppc_function *p, uint op, uint bo, uint bi, uint bh,
+        uint op2, uint lk)
+{
+   union xl_inst inst;
+   inst.inst.op = op;
+   inst.inst.bo = bo;
+   inst.inst.bi = bi;
+   inst.inst.unused = 0x0;
+   inst.inst.bh = bh;
+   inst.inst.op2 = op2;
+   inst.inst.lk = lk;
+   emit_instruction(p, inst.bits);
+}
+
+static INLINE void
+dump_xl(const char *name, uint inst)
+{
+   union xl_inst i;
+
+   i.bits = inst;
+   debug_printf("%s = 0x%08x\n", name, inst);
+   debug_printf(" op: %d 0x%x\n", i.inst.op, i.inst.op);
+   debug_printf(" bo: %d 0x%x\n", i.inst.bo, i.inst.bo);
+   debug_printf(" bi: %d 0x%x\n", i.inst.bi, i.inst.bi);
+   debug_printf(" unused: %d 0x%x\n", i.inst.unused, i.inst.unused);
+   debug_printf(" bh: %d 0x%x\n", i.inst.bh, i.inst.bh);
+   debug_printf(" op2: %d 0x%x\n", i.inst.op2, i.inst.op2);
+   debug_printf(" lk: %d 0x%x\n", i.inst.lk, i.inst.lk);
+}
+
+
+union x_inst {
+   uint32_t bits;
+   struct {
+      unsigned op:6;
+      unsigned vrs:5;
+      unsigned ra:5;
+      unsigned rb:5;
+      unsigned op2:10;
+      unsigned unused:1;
+   } inst;
+};
+
+static INLINE void
+emit_x(struct ppc_function *p, uint op, uint vrs, uint ra, uint rb, uint op2,
+       const char *format)
+{
+   union x_inst inst;
+   inst.inst.op = op;
+   inst.inst.vrs = vrs;
+   inst.inst.ra = ra;
+   inst.inst.rb = rb;
+   inst.inst.op2 = op2;
+   inst.inst.unused = 0x0;
+   emit_instruction(p, inst.bits);
+   if (p->print) {
+      indent(p);
+      printf(format, vrs, ra, rb);
+   }
+}
+
+
+union d_inst {
+   uint32_t bits;
+   struct {
+      unsigned op:6;
+      unsigned rt:5;
+      unsigned ra:5;
+      unsigned si:16;
+   } inst;
+};
+
+static INLINE void
+emit_d(struct ppc_function *p, uint op, uint rt, uint ra, int si,
+       const char *format, boolean transpose)
+{
+   union d_inst inst;
+   assert(si >= -32768);
+   assert(si <= 32767);
+   inst.inst.op = op;
+   inst.inst.rt = rt;
+   inst.inst.ra = ra;
+   inst.inst.si = (unsigned) (si & 0xffff);
+   emit_instruction(p, inst.bits);
+   if (p->print) {
+      indent(p);
+      if (transpose)
+         printf(format, rt, si, ra);
+      else
+         printf(format, rt, ra, si);
+   }
+}
+
+
+union a_inst {
+   uint32_t bits;
+   struct {
+      unsigned op:6;
+      unsigned frt:5;
+      unsigned fra:5;
+      unsigned frb:5;
+      unsigned unused:5;
+      unsigned op2:5;
+      unsigned rc:1;
+   } inst;
+};
+
+static INLINE void
+emit_a(struct ppc_function *p, uint op, uint frt, uint fra, uint frb, uint op2,
+       uint rc, const char *format)
+{
+   union a_inst inst;
+   inst.inst.op = op;
+   inst.inst.frt = frt;
+   inst.inst.fra = fra;
+   inst.inst.frb = frb;
+   inst.inst.unused = 0x0;
+   inst.inst.op2 = op2;
+   inst.inst.rc = rc;
+   emit_instruction(p, inst.bits);
+   if (p->print) {
+      indent(p);
+      printf(format, frt, fra, frb);
+   }
+}
+
+
+union xo_inst {
+   uint32_t bits;
+   struct {
+      unsigned op:6;
+      unsigned rt:5;
+      unsigned ra:5;
+      unsigned rb:5;
+      unsigned oe:1;
+      unsigned op2:9;
+      unsigned rc:1;
+   } inst;
+};
+
+static INLINE void
+emit_xo(struct ppc_function *p, uint op, uint rt, uint ra, uint rb, uint oe,
+        uint op2, uint rc, const char *format)
+{
+   union xo_inst inst;
+   inst.inst.op = op;
+   inst.inst.rt = rt;
+   inst.inst.ra = ra;
+   inst.inst.rb = rb;
+   inst.inst.oe = oe;
+   inst.inst.op2 = op2;
+   inst.inst.rc = rc;
+   emit_instruction(p, inst.bits);
+   if (p->print) {
+      indent(p);
+      printf(format, rt, ra, rb);
+   }
+}
+
+
+
+
+
+/**
+ ** float vector arithmetic
+ **/
+
+/** vector float add */
+void
+ppc_vaddfp(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vx(p, 10, vD, vA, vB, "vaddfp\t%u, v%u, v%u\n", FALSE);
+}
+
+/** vector float substract */
+void
+ppc_vsubfp(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vx(p, 74, vD, vA, vB, "vsubfp\tv%u, v%u, v%u\n", FALSE);
+}
+
+/** vector float min */
+void
+ppc_vminfp(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vx(p, 1098, vD, vA, vB, "vminfp\tv%u, v%u, v%u\n", FALSE);
+}
+
+/** vector float max */
+void
+ppc_vmaxfp(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vx(p, 1034, vD, vA, vB, "vmaxfp\tv%u, v%u, v%u\n", FALSE);
+}
+
+/** vector float mult add: vD = vA * vB + vC */
+void
+ppc_vmaddfp(struct ppc_function *p, uint vD, uint vA, uint vB, uint vC)
+{
+   /* note arg order */
+   emit_va(p, 46, vD, vA, vC, vB, "vmaddfp\tv%u, v%u, v%u, v%u\n");
+}
+
+/** vector float negative mult subtract: vD = vA - vB * vC */
+void
+ppc_vnmsubfp(struct ppc_function *p, uint vD, uint vA, uint vB, uint vC)
+{
+   /* note arg order */
+   emit_va(p, 47, vD, vB, vA, vC, "vnmsubfp\tv%u, v%u, v%u, v%u\n");
+}
+
+/** vector float compare greater than */
+void
+ppc_vcmpgtfpx(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vxr(p, 710, vD, vA, vB, "vcmpgtfpx\tv%u, v%u, v%u");
+}
+
+/** vector float compare greater than or equal to */
+void
+ppc_vcmpgefpx(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vxr(p, 454, vD, vA, vB, "vcmpgefpx\tv%u, v%u, v%u");
+}
+
+/** vector float compare equal */
+void
+ppc_vcmpeqfpx(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vxr(p, 198, vD, vA, vB, "vcmpeqfpx\tv%u, v%u, v%u");
+}
+
+/** vector float 2^x */
+void
+ppc_vexptefp(struct ppc_function *p, uint vD, uint vB)
+{
+   emit_vx(p, 394, vD, 0, vB, "vexptefp\tv%u, 0%u, v%u\n", FALSE);
+}
+
+/** vector float log2(x) */
+void
+ppc_vlogefp(struct ppc_function *p, uint vD, uint vB)
+{
+   emit_vx(p, 458, vD, 0, vB, "vlogefp\tv%u, 0%u, v%u\n", FALSE);
+}
+
+/** vector float reciprocol */
+void
+ppc_vrefp(struct ppc_function *p, uint vD, uint vB)
+{
+   emit_vx(p, 266, vD, 0, vB, "vrefp\tv%u, 0%u, v%u\n", FALSE);
+}
+
+/** vector float reciprocol sqrt estimate */
+void
+ppc_vrsqrtefp(struct ppc_function *p, uint vD, uint vB)
+{
+   emit_vx(p, 330, vD, 0, vB, "vrsqrtefp\tv%u, 0%u, v%u\n", FALSE);
+}
+
+/** vector float round to negative infinity */
+void
+ppc_vrfim(struct ppc_function *p, uint vD, uint vB)
+{
+   emit_vx(p, 714, vD, 0, vB, "vrfim\tv%u, 0%u, v%u\n", FALSE);
+}
+
+/** vector float round to positive infinity */
+void
+ppc_vrfip(struct ppc_function *p, uint vD, uint vB)
+{
+   emit_vx(p, 650, vD, 0, vB, "vrfip\tv%u, 0%u, v%u\n", FALSE);
+}
+
+/** vector float round to nearest int */
+void
+ppc_vrfin(struct ppc_function *p, uint vD, uint vB)
+{
+   emit_vx(p, 522, vD, 0, vB, "vrfin\tv%u, 0%u, v%u\n", FALSE);
+}
+
+/** vector float round to int toward zero */
+void
+ppc_vrfiz(struct ppc_function *p, uint vD, uint vB)
+{
+   emit_vx(p, 586, vD, 0, vB, "vrfiz\tv%u, 0%u, v%u\n", FALSE);
+}
+
+/** vector store: store vR at mem[rA+rB] */
+void
+ppc_stvx(struct ppc_function *p, uint vR, uint rA, uint rB)
+{
+   emit_x(p, 31, vR, rA, rB, 231, "stvx\tv%u, r%u, r%u\n");
+}
+
+/** vector load: vR = mem[rA+rB] */
+void
+ppc_lvx(struct ppc_function *p, uint vR, uint rA, uint rB)
+{
+   emit_x(p, 31, vR, rA, rB, 103, "lvx\tv%u, r%u, r%u\n");
+}
+
+/** load vector element word: vR = mem_word[ra+rb] */
+void
+ppc_lvewx(struct ppc_function *p, uint vR, uint rA, uint rB)
+{
+   emit_x(p, 31, vR, rA, rB, 71, "lvewx\tv%u, r%u, r%u\n");
+}
+
+
+
+
+/**
+ ** vector bitwise operations
+ **/
+
+/** vector and */
+void
+ppc_vand(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vx(p, 1028, vD, vA, vB, "vand\tv%u, v%u, v%u\n", FALSE);
+}
+
+/** vector and complement */
+void
+ppc_vandc(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vx(p, 1092, vD, vA, vB, "vandc\tv%u, v%u, v%u\n", FALSE);
+}
+
+/** vector or */
+void
+ppc_vor(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vx(p, 1156, vD, vA, vB, "vor\tv%u, v%u, v%u\n", FALSE);
+}
+
+/** vector nor */
+void
+ppc_vnor(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vx(p, 1284, vD, vA, vB, "vnor\tv%u, v%u, v%u\n", FALSE);
+}
+
+/** vector xor */
+void
+ppc_vxor(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vx(p, 1220, vD, vA, vB, "vxor\tv%u, v%u, v%u\n", FALSE);
+}
+
+/** Pseudo-instruction: vector move */
+void
+ppc_vmove(struct ppc_function *p, uint vD, uint vA)
+{
+   boolean print = p->print;
+   p->print = FALSE;
+   ppc_vor(p, vD, vA, vA);
+   if (print) {
+      indent(p);
+      printf("vor\tv%u, v%u, v%u \t# v%u = v%u\n", vD, vA, vA, vD, vA);
+   }
+   p->print = print;
+}
+
+/** Set vector register to {0,0,0,0} */
+void
+ppc_vzero(struct ppc_function *p, uint vr)
+{
+   boolean print = p->print;
+   p->print = FALSE;
+   ppc_vxor(p, vr, vr, vr);
+   if (print) {
+      indent(p);
+      printf("vxor\tv%u, v%u, v%u \t# v%u = {0,0,0,0}\n", vr, vr, vr, vr);
+   }
+   p->print = print;
+}
+
+
+
+
+/**
+ ** Vector shuffle / select / splat / etc
+ **/
+
+/** vector permute */
+void
+ppc_vperm(struct ppc_function *p, uint vD, uint vA, uint vB, uint vC)
+{
+   emit_va(p, 43, vD, vA, vB, vC, "vperm\tr%u, r%u, r%u, r%u");
+}
+
+/** vector select */
+void
+ppc_vsel(struct ppc_function *p, uint vD, uint vA, uint vB, uint vC)
+{
+   emit_va(p, 42, vD, vA, vB, vC, "vsel\tr%u, r%u, r%u, r%u");
+}
+
+/** vector splat byte */
+void
+ppc_vspltb(struct ppc_function *p, uint vD, uint vB, uint imm)
+{
+   emit_vx(p, 42, vD, imm, vB, "vspltb\tv%u, v%u, %u\n", TRUE);
+}
+
+/** vector splat half word */
+void
+ppc_vsplthw(struct ppc_function *p, uint vD, uint vB, uint imm)
+{
+   emit_vx(p, 588, vD, imm, vB, "vsplthw\tv%u, v%u, %u\n", TRUE);
+}
+
+/** vector splat word */
+void
+ppc_vspltw(struct ppc_function *p, uint vD, uint vB, uint imm)
+{
+   emit_vx(p, 652, vD, imm, vB, "vspltw\tv%u, v%u, %u\n", TRUE);
+}
+
+/** vector splat signed immediate word */
+void
+ppc_vspltisw(struct ppc_function *p, uint vD, int imm)
+{
+   assert(imm >= -16);
+   assert(imm < 15);
+   emit_vx(p, 908, vD, imm, 0, "vspltisw\tv%u, %d, %u\n", FALSE);
+}
+
+/** vector shift left word: vD[word] = vA[word] << (vB[word] & 0x1f) */
+void
+ppc_vslw(struct ppc_function *p, uint vD, uint vA, uint vB)
+{
+   emit_vx(p, 388, vD, vA, vB, "vslw\tv%u, v%u, v%u\n", FALSE);
+}
+
+
+
+
+/**
+ ** integer arithmetic
+ **/
+
+/** rt = ra + imm */
+void
+ppc_addi(struct ppc_function *p, uint rt, uint ra, int imm)
+{
+   emit_d(p, 14, rt, ra, imm, "addi\tr%u, r%u, %d\n", FALSE);
+}
+
+/** rt = ra + (imm << 16) */
+void
+ppc_addis(struct ppc_function *p, uint rt, uint ra, int imm)
+{
+   emit_d(p, 15, rt, ra, imm, "addis\tr%u, r%u, %d\n", FALSE);
+}
+
+/** rt = ra + rb */
+void
+ppc_add(struct ppc_function *p, uint rt, uint ra, uint rb)
+{
+   emit_xo(p, 31, rt, ra, rb, 0, 266, 0, "add\tr%u, r%u, r%u\n");
+}
+
+/** rt = ra AND ra */
+void
+ppc_and(struct ppc_function *p, uint rt, uint ra, uint rb)
+{
+   emit_x(p, 31, ra, rt, rb, 28, "and\tr%u, r%u, r%u\n");  /* note argument order */
+}
+
+/** rt = ra AND imm */
+void
+ppc_andi(struct ppc_function *p, uint rt, uint ra, int imm)
+{
+   /* note argument order */
+   emit_d(p, 28, ra, rt, imm, "andi\tr%u, r%u, %d\n", FALSE);
+}
+
+/** rt = ra OR ra */
+void
+ppc_or(struct ppc_function *p, uint rt, uint ra, uint rb)
+{
+   emit_x(p, 31, ra, rt, rb, 444, "or\tr%u, r%u, r%u\n");  /* note argument order */
+}
+
+/** rt = ra OR imm */
+void
+ppc_ori(struct ppc_function *p, uint rt, uint ra, int imm)
+{
+   /* note argument order */
+   emit_d(p, 24, ra, rt, imm, "ori\tr%u, r%u, %d\n", FALSE);
+}
+
+/** rt = ra XOR ra */
+void
+ppc_xor(struct ppc_function *p, uint rt, uint ra, uint rb)
+{
+   emit_x(p, 31, ra, rt, rb, 316, "xor\tr%u, r%u, r%u\n");  /* note argument order */
+}
+
+/** rt = ra XOR imm */
+void
+ppc_xori(struct ppc_function *p, uint rt, uint ra, int imm)
+{
+   /* note argument order */
+   emit_d(p, 26, ra, rt, imm, "xori\tr%u, r%u, %d\n", FALSE);
+}
+
+/** pseudo instruction: move: rt = ra */
+void
+ppc_mr(struct ppc_function *p, uint rt, uint ra)
+{
+   ppc_or(p, rt, ra, ra);
+}
+
+/** pseudo instruction: load immediate: rt = imm */
+void
+ppc_li(struct ppc_function *p, uint rt, int imm)
+{
+   boolean print = p->print;
+   p->print = FALSE;
+   ppc_addi(p, rt, 0, imm);
+   if (print) {
+      indent(p);
+      printf("addi\tr%u, r0, %d \t# r%u = %d\n", rt, imm, rt, imm);
+   }
+   p->print = print;
+}
+
+/** rt = imm << 16 */
+void
+ppc_lis(struct ppc_function *p, uint rt, int imm)
+{
+   ppc_addis(p, rt, 0, imm);
+}
+
+/** rt = imm */
+void
+ppc_load_int(struct ppc_function *p, uint rt, int imm)
+{
+   ppc_lis(p, rt, (imm >> 16));          /* rt = imm >> 16 */
+   ppc_ori(p, rt, rt, (imm & 0xffff));   /* rt = rt | (imm & 0xffff) */
+}
+
+
+
+
+/**
+ ** integer load/store
+ **/
+
+/** store rs at memory[(ra)+d],
+ * then update ra = (ra)+d
+ */
+void
+ppc_stwu(struct ppc_function *p, uint rs, uint ra, int d)
+{
+   emit_d(p, 37, rs, ra, d, "stwu\tr%u, %d(r%u)\n", TRUE);
+}
+
+/** store rs at memory[(ra)+d] */
+void
+ppc_stw(struct ppc_function *p, uint rs, uint ra, int d)
+{
+   emit_d(p, 36, rs, ra, d, "stw\tr%u, %d(r%u)\n", TRUE);
+}
+
+/** Load rt = mem[(ra)+d];  then zero set high 32 bits to zero. */
+void
+ppc_lwz(struct ppc_function *p, uint rt, uint ra, int d)
+{
+   emit_d(p, 32, rt, ra, d, "lwz\tr%u, %d(r%u)\n", TRUE);
+}
+
+
+
+/**
+ ** Float (non-vector) arithmetic
+ **/
+
+/** add: frt = fra + frb */
+void
+ppc_fadd(struct ppc_function *p, uint frt, uint fra, uint frb)
+{
+   emit_a(p, 63, frt, fra, frb, 21, 0, "fadd\tf%u, f%u, f%u\n");
+}
+
+/** sub: frt = fra - frb */
+void
+ppc_fsub(struct ppc_function *p, uint frt, uint fra, uint frb)
+{
+   emit_a(p, 63, frt, fra, frb, 20, 0, "fsub\tf%u, f%u, f%u\n");
+}
+
+/** convert to int: rt = (int) ra */
+void
+ppc_fctiwz(struct ppc_function *p, uint rt, uint fra)
+{
+   emit_x(p, 63, rt, 0, fra, 15, "fctiwz\tr%u, r%u, r%u\n");
+}
+
+/** store frs at mem[(ra)+offset] */
+void
+ppc_stfs(struct ppc_function *p, uint frs, uint ra, int offset)
+{
+   emit_d(p, 52, frs, ra, offset, "stfs\tr%u, %d(r%u)\n", TRUE);
+}
+
+/** store frs at mem[(ra)+(rb)] */
+void
+ppc_stfiwx(struct ppc_function *p, uint frs, uint ra, uint rb)
+{
+   emit_x(p, 31, frs, ra, rb, 983, "stfiwx\tr%u, r%u, r%u\n");
+}
+
+/** load frt = mem[(ra)+offset] */
+void
+ppc_lfs(struct ppc_function *p, uint frt, uint ra, int offset)
+{
+   emit_d(p, 48, frt, ra, offset, "stfs\tr%u, %d(r%u)\n", TRUE);
+}
+
+
+
+
+
+/**
+ ** branch instructions
+ **/
+
+/** BLR: Branch to link register (p. 35) */
+void
+ppc_blr(struct ppc_function *p)
+{
+   emit_i(p, 18, 0, 0, 1);
+   if (p->print) {
+      indent(p);
+      printf("blr\n");
+   }
+}
+
+/** Branch Conditional to Link Register (p. 36) */
+void
+ppc_bclr(struct ppc_function *p, uint condOp, uint branchHint, uint condReg)
+{
+   emit_xl(p, 19, condOp, condReg, branchHint, 16, 0);
+   if (p->print) {
+      indent(p);
+      printf("bclr\t%u %u %u\n", condOp, branchHint, condReg);
+   }
+}
+
+/** Pseudo instruction: return from subroutine */
+void
+ppc_return(struct ppc_function *p)
+{
+   ppc_bclr(p, BRANCH_COND_ALWAYS, BRANCH_HINT_SUB_RETURN, 0);
+}
diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc.h b/src/gallium/auxiliary/rtasm/rtasm_ppc.h
new file mode 100644
index 0000000000..93e5f5187d
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/rtasm_ppc.h
@@ -0,0 +1,342 @@
+/**************************************************************************
+ *
+ * Copyright (C) 2008 Tungsten Graphics, Inc.   All Rights Reserved.
+ * Copyright (C) 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * PPC code generation.
+ * \author Brian Paul
+ */
+
+
+#ifndef RTASM_PPC_H
+#define RTASM_PPC_H
+
+
+#include "pipe/p_compiler.h"
+
+
+#define PPC_INST_SIZE 4  /**< 4 bytes / instruction */
+
+#define PPC_NUM_REGS 32
+#define PPC_NUM_FP_REGS 32
+#define PPC_NUM_VEC_REGS 32
+
+/** Stack pointer register */
+#define PPC_REG_SP 1
+
+/** Branch conditions */
+#define BRANCH_COND_ALWAYS       0x14  /* binary 1z1zz (z=ignored) */
+
+/** Branch hints */
+#define BRANCH_HINT_SUB_RETURN   0x0   /* binary 00 */
+
+
+struct ppc_function
+{
+   uint32_t *store;  /**< instruction buffer */
+   uint num_inst;
+   uint max_inst;
+   uint32_t reg_used;   /** used/free general-purpose registers bitmask */
+   uint32_t fp_used;   /** used/free floating point registers bitmask */
+   uint32_t vec_used;   /** used/free vector registers bitmask */
+   int indent;
+   boolean print;
+};
+
+
+
+extern void ppc_init_func(struct ppc_function *p);
+extern void ppc_release_func(struct ppc_function *p);
+extern uint ppc_num_instructions(const struct ppc_function *p);
+extern void (*ppc_get_func( struct ppc_function *p ))( void );
+extern void ppc_dump_func(const struct ppc_function *p);
+
+extern void ppc_print_code(struct ppc_function *p, boolean enable);
+extern void ppc_indent(struct ppc_function *p, int spaces);
+extern void ppc_comment(struct ppc_function *p, int rel_indent, const char *s);
+
+extern int ppc_reserve_register(struct ppc_function *p, int reg);
+extern int ppc_allocate_register(struct ppc_function *p);
+extern void ppc_release_register(struct ppc_function *p, int reg);
+extern int ppc_allocate_fp_register(struct ppc_function *p);
+extern void ppc_release_fp_register(struct ppc_function *p, int reg);
+extern int ppc_allocate_vec_register(struct ppc_function *p);
+extern void ppc_release_vec_register(struct ppc_function *p, int reg);
+
+
+
+/**
+ ** float vector arithmetic
+ **/
+
+/** vector float add */
+extern void
+ppc_vaddfp(struct ppc_function *p,uint vD, uint vA, uint vB);
+
+/** vector float substract */
+extern void
+ppc_vsubfp(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** vector float min */
+extern void
+ppc_vminfp(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** vector float max */
+extern void
+ppc_vmaxfp(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** vector float mult add: vD = vA * vB + vC */
+extern void
+ppc_vmaddfp(struct ppc_function *p, uint vD, uint vA, uint vB, uint vC);
+
+/** vector float negative mult subtract: vD = vA - vB * vC */
+extern void
+ppc_vnmsubfp(struct ppc_function *p, uint vD, uint vA, uint vB, uint vC);
+
+/** vector float compare greater than */
+extern void
+ppc_vcmpgtfpx(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** vector float compare greater than or equal to */
+extern void
+ppc_vcmpgefpx(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** vector float compare equal */
+extern void
+ppc_vcmpeqfpx(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** vector float 2^x */
+extern void
+ppc_vexptefp(struct ppc_function *p, uint vD, uint vB);
+
+/** vector float log2(x) */
+extern void
+ppc_vlogefp(struct ppc_function *p, uint vD, uint vB);
+
+/** vector float reciprocol */
+extern void
+ppc_vrefp(struct ppc_function *p, uint vD, uint vB);
+
+/** vector float reciprocol sqrt estimate */
+extern void
+ppc_vrsqrtefp(struct ppc_function *p, uint vD, uint vB);
+
+/** vector float round to negative infinity */
+extern void
+ppc_vrfim(struct ppc_function *p, uint vD, uint vB);
+
+/** vector float round to positive infinity */
+extern void
+ppc_vrfip(struct ppc_function *p, uint vD, uint vB);
+
+/** vector float round to nearest int */
+extern void
+ppc_vrfin(struct ppc_function *p, uint vD, uint vB);
+
+/** vector float round to int toward zero */
+extern void
+ppc_vrfiz(struct ppc_function *p, uint vD, uint vB);
+
+
+/** vector store: store vR at mem[vA+vB] */
+extern void
+ppc_stvx(struct ppc_function *p, uint vR, uint vA, uint vB);
+
+/** vector load: vR = mem[vA+vB] */
+extern void
+ppc_lvx(struct ppc_function *p, uint vR, uint vA, uint vB);
+
+/** load vector element word: vR = mem_word[vA+vB] */
+extern void
+ppc_lvewx(struct ppc_function *p, uint vR, uint vA, uint vB);
+
+
+
+/**
+ ** vector bitwise operations
+ **/
+
+
+/** vector and */
+extern void
+ppc_vand(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** vector and complement */
+extern void
+ppc_vandc(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** vector or */
+extern void
+ppc_vor(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** vector nor */
+extern void
+ppc_vnor(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** vector xor */
+extern void
+ppc_vxor(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+/** Pseudo-instruction: vector move */
+extern void
+ppc_vmove(struct ppc_function *p, uint vD, uint vA);
+
+/** Set vector register to {0,0,0,0} */
+extern void
+ppc_vzero(struct ppc_function *p, uint vr);
+
+
+
+/**
+ ** Vector shuffle / select / splat / etc
+ **/
+
+/** vector permute */
+extern void
+ppc_vperm(struct ppc_function *p, uint vD, uint vA, uint vB, uint vC);
+
+/** vector select */
+extern void
+ppc_vsel(struct ppc_function *p, uint vD, uint vA, uint vB, uint vC);
+
+/** vector splat byte */
+extern void
+ppc_vspltb(struct ppc_function *p, uint vD, uint vB, uint imm);
+
+/** vector splat half word */
+extern void
+ppc_vsplthw(struct ppc_function *p, uint vD, uint vB, uint imm);
+
+/** vector splat word */
+extern void
+ppc_vspltw(struct ppc_function *p, uint vD, uint vB, uint imm);
+
+/** vector splat signed immediate word */
+extern void
+ppc_vspltisw(struct ppc_function *p, uint vD, int imm);
+
+/** vector shift left word: vD[word] = vA[word] << (vB[word] & 0x1f) */
+extern void
+ppc_vslw(struct ppc_function *p, uint vD, uint vA, uint vB);
+
+
+
+/**
+ ** scalar arithmetic
+ **/
+
+extern void
+ppc_add(struct ppc_function *p, uint rt, uint ra, uint rb);
+
+extern void
+ppc_addi(struct ppc_function *p, uint rt, uint ra, int imm);
+
+extern void
+ppc_addis(struct ppc_function *p, uint rt, uint ra, int imm);
+
+extern void
+ppc_and(struct ppc_function *p, uint rt, uint ra, uint rb);
+
+extern void
+ppc_andi(struct ppc_function *p, uint rt, uint ra, int imm);
+
+extern void
+ppc_or(struct ppc_function *p, uint rt, uint ra, uint rb);
+
+extern void
+ppc_ori(struct ppc_function *p, uint rt, uint ra, int imm);
+
+extern void
+ppc_xor(struct ppc_function *p, uint rt, uint ra, uint rb);
+
+extern void
+ppc_xori(struct ppc_function *p, uint rt, uint ra, int imm);
+
+extern void
+ppc_mr(struct ppc_function *p, uint rt, uint ra);
+
+extern void
+ppc_li(struct ppc_function *p, uint rt, int imm);
+
+extern void
+ppc_lis(struct ppc_function *p, uint rt, int imm);
+
+extern void
+ppc_load_int(struct ppc_function *p, uint rt, int imm);
+
+
+
+/**
+ ** scalar load/store
+ **/
+
+extern void
+ppc_stwu(struct ppc_function *p, uint rs, uint ra, int d);
+
+extern void
+ppc_stw(struct ppc_function *p, uint rs, uint ra, int d);
+
+extern void
+ppc_lwz(struct ppc_function *p, uint rs, uint ra, int d);
+
+
+
+/**
+ ** Float (non-vector) arithmetic
+ **/
+
+extern void
+ppc_fadd(struct ppc_function *p, uint frt, uint fra, uint frb);
+
+extern void
+ppc_fsub(struct ppc_function *p, uint frt, uint fra, uint frb);
+
+extern void
+ppc_fctiwz(struct ppc_function *p, uint rt, uint ra);
+
+extern void
+ppc_stfs(struct ppc_function *p, uint frs, uint ra, int offset);
+
+extern void
+ppc_stfiwx(struct ppc_function *p, uint frs, uint ra, uint rb);
+
+extern void
+ppc_lfs(struct ppc_function *p, uint frt, uint ra, int offset);
+
+
+
+/**
+ ** branch instructions
+ **/
+
+extern void
+ppc_blr(struct ppc_function *p);
+
+void
+ppc_bclr(struct ppc_function *p, uint condOp, uint branchHint, uint condReg);
+
+extern void
+ppc_return(struct ppc_function *p);
+
+
+#endif /* RTASM_PPC_H */
diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c
new file mode 100644
index 0000000000..53a0e722cf
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c
@@ -0,0 +1,1067 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file
+ * Real-time assembly generation interface for Cell B.E. SPEs.
+ *
+ * \author Ian Romanick <idr@us.ibm.com>
+ * \author Brian Paul
+ */
+
+
+#include <stdio.h>
+#include "pipe/p_compiler.h"
+#include "util/u_memory.h"
+#include "rtasm_ppc_spe.h"
+
+
+#ifdef GALLIUM_CELL
+/**
+ * SPE instruction types
+ *
+ * There are 6 primary instruction encodings used on the Cell's SPEs.  Each of
+ * the following unions encodes one type.
+ *
+ * \bug
+ * If, at some point, we start generating SPE code from a little-endian host
+ * these unions will not work.
+ */
+/*@{*/
+/**
+ * Encode one output register with two input registers
+ */
+union spe_inst_RR {
+    uint32_t bits;
+    struct {
+	unsigned op:11;
+	unsigned rB:7;
+	unsigned rA:7;
+	unsigned rT:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with three input registers
+ */
+union spe_inst_RRR {
+    uint32_t bits;
+    struct {
+	unsigned op:4;
+	unsigned rT:7;
+	unsigned rB:7;
+	unsigned rA:7;
+	unsigned rC:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with one input reg. and a 7-bit signed immed
+ */
+union spe_inst_RI7 {
+    uint32_t bits;
+    struct {
+	unsigned op:11;
+	unsigned i7:7;
+	unsigned rA:7;
+	unsigned rT:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with one input reg. and an 8-bit signed immed
+ */
+union spe_inst_RI8 {
+    uint32_t bits;
+    struct {
+	unsigned op:10;
+	unsigned i8:8;
+	unsigned rA:7;
+	unsigned rT:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with one input reg. and a 10-bit signed immed
+ */
+union spe_inst_RI10 {
+    uint32_t bits;
+    struct {
+	unsigned op:8;
+	unsigned i10:10;
+	unsigned rA:7;
+	unsigned rT:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with a 16-bit signed immediate
+ */
+union spe_inst_RI16 {
+    uint32_t bits;
+    struct {
+	unsigned op:9;
+	unsigned i16:16;
+	unsigned rT:7;
+    } inst;
+};
+
+
+/**
+ * Encode one output register with a 18-bit signed immediate
+ */
+union spe_inst_RI18 {
+    uint32_t bits;
+    struct {
+	unsigned op:7;
+	unsigned i18:18;
+	unsigned rT:7;
+    } inst;
+};
+/*@}*/
+
+
+static void
+indent(const struct spe_function *p)
+{
+   int i;
+   for (i = 0; i < p->indent; i++) {
+      putchar(' ');
+   }
+}
+
+
+static const char *
+rem_prefix(const char *longname)
+{
+   return longname + 4;
+}
+
+
+static const char *
+reg_name(int reg)
+{
+   switch (reg) {
+   case SPE_REG_SP:
+      return "$sp";
+   case SPE_REG_RA:
+      return "$lr";
+   default:
+      {
+         /* cycle through four buffers to handle multiple calls per printf */
+         static char buf[4][10];
+         static int b = 0;
+         b = (b + 1) % 4;
+         sprintf(buf[b], "$%d", reg);
+         return buf[b];
+      }
+   }
+}
+
+
+static void
+emit_instruction(struct spe_function *p, uint32_t inst_bits)
+{
+   if (!p->store)
+      return;  /* out of memory, drop the instruction */
+
+   if (p->num_inst == p->max_inst) {
+      /* allocate larger buffer */
+      uint32_t *newbuf;
+      p->max_inst *= 2;  /* 2x larger */
+      newbuf = align_malloc(p->max_inst * SPE_INST_SIZE, 16);
+      if (newbuf) {
+         memcpy(newbuf, p->store, p->num_inst * SPE_INST_SIZE);
+      }
+      align_free(p->store);
+      p->store = newbuf;
+      if (!p->store) {
+         /* out of memory */
+         p->num_inst = 0;
+         return;
+      }
+   }
+
+   p->store[p->num_inst++] = inst_bits;
+}
+
+
+
+static void emit_RR(struct spe_function *p, unsigned op, int rT,
+		    int rA, int rB, const char *name)
+{
+    union spe_inst_RR inst;
+    inst.inst.op = op;
+    inst.inst.rB = rB;
+    inst.inst.rA = rA;
+    inst.inst.rT = rT;
+    emit_instruction(p, inst.bits);
+    if (p->print) {
+       indent(p);
+       printf("%s\t%s, %s, %s\n",
+              rem_prefix(name), reg_name(rT), reg_name(rA), reg_name(rB));
+    }
+}
+
+
+static void emit_RRR(struct spe_function *p, unsigned op, int rT,
+                     int rA, int rB, int rC, const char *name)
+{
+    union spe_inst_RRR inst;
+    inst.inst.op = op;
+    inst.inst.rT = rT;
+    inst.inst.rB = rB;
+    inst.inst.rA = rA;
+    inst.inst.rC = rC;
+    emit_instruction(p, inst.bits);
+    if (p->print) {
+       indent(p);
+       printf("%s\t%s, %s, %s, %s\n", rem_prefix(name), reg_name(rT),
+              reg_name(rA), reg_name(rB), reg_name(rC));
+    }
+}
+
+
+static void emit_RI7(struct spe_function *p, unsigned op, int rT,
+		     int rA, int imm, const char *name)
+{
+    union spe_inst_RI7 inst;
+    inst.inst.op = op;
+    inst.inst.i7 = imm;
+    inst.inst.rA = rA;
+    inst.inst.rT = rT;
+    emit_instruction(p, inst.bits);
+    if (p->print) {
+       indent(p);
+       printf("%s\t%s, %s, 0x%x\n",
+              rem_prefix(name), reg_name(rT), reg_name(rA), imm);
+    }
+}
+
+
+
+static void emit_RI8(struct spe_function *p, unsigned op, int rT,
+		     int rA, int imm, const char *name)
+{
+    union spe_inst_RI8 inst;
+    inst.inst.op = op;
+    inst.inst.i8 = imm;
+    inst.inst.rA = rA;
+    inst.inst.rT = rT;
+    emit_instruction(p, inst.bits);
+    if (p->print) {
+       indent(p);
+       printf("%s\t%s, %s, 0x%x\n",
+              rem_prefix(name), reg_name(rT), reg_name(rA), imm);
+    }
+}
+
+
+
+static void emit_RI10(struct spe_function *p, unsigned op, int rT,
+		      int rA, int imm, const char *name)
+{
+    union spe_inst_RI10 inst;
+    inst.inst.op = op;
+    inst.inst.i10 = imm;
+    inst.inst.rA = rA;
+    inst.inst.rT = rT;
+    emit_instruction(p, inst.bits);
+    if (p->print) {
+       indent(p);
+       printf("%s\t%s, %s, 0x%x\n",
+              rem_prefix(name), reg_name(rT), reg_name(rA), imm);
+    }
+}
+
+
+/** As above, but do range checking on signed immediate value */
+static void emit_RI10s(struct spe_function *p, unsigned op, int rT,
+                       int rA, int imm, const char *name)
+{
+    assert(imm <= 511);
+    assert(imm >= -512);
+    emit_RI10(p, op, rT, rA, imm, name);
+}
+
+
+static void emit_RI16(struct spe_function *p, unsigned op, int rT,
+		      int imm, const char *name)
+{
+    union spe_inst_RI16 inst;
+    inst.inst.op = op;
+    inst.inst.i16 = imm;
+    inst.inst.rT = rT;
+    emit_instruction(p, inst.bits);
+    if (p->print) {
+       indent(p);
+       printf("%s\t%s, 0x%x\n", rem_prefix(name), reg_name(rT), imm);
+    }
+}
+
+
+static void emit_RI18(struct spe_function *p, unsigned op, int rT,
+		      int imm, const char *name)
+{
+    union spe_inst_RI18 inst;
+    inst.inst.op = op;
+    inst.inst.i18 = imm;
+    inst.inst.rT = rT;
+    emit_instruction(p, inst.bits);
+    if (p->print) {
+       indent(p);
+       printf("%s\t%s, 0x%x\n", rem_prefix(name), reg_name(rT), imm);
+    }
+}
+
+
+#define EMIT(_name, _op) \
+void _name (struct spe_function *p) \
+{ \
+   emit_RR(p, _op, 0, 0, 0, __FUNCTION__); \
+}
+
+#define EMIT_(_name, _op) \
+void _name (struct spe_function *p, int rT) \
+{ \
+   emit_RR(p, _op, rT, 0, 0, __FUNCTION__); \
+}
+
+#define EMIT_R(_name, _op) \
+void _name (struct spe_function *p, int rT, int rA) \
+{ \
+   emit_RR(p, _op, rT, rA, 0, __FUNCTION__);                 \
+}
+
+#define EMIT_RR(_name, _op) \
+void _name (struct spe_function *p, int rT, int rA, int rB) \
+{ \
+   emit_RR(p, _op, rT, rA, rB, __FUNCTION__);                \
+}
+
+#define EMIT_RRR(_name, _op) \
+void _name (struct spe_function *p, int rT, int rA, int rB, int rC) \
+{ \
+   emit_RRR(p, _op, rT, rA, rB, rC, __FUNCTION__);           \
+}
+
+#define EMIT_RI7(_name, _op) \
+void _name (struct spe_function *p, int rT, int rA, int imm) \
+{ \
+   emit_RI7(p, _op, rT, rA, imm, __FUNCTION__);              \
+}
+
+#define EMIT_RI8(_name, _op, bias) \
+void _name (struct spe_function *p, int rT, int rA, int imm) \
+{ \
+   emit_RI8(p, _op, rT, rA, bias - imm, __FUNCTION__);       \
+}
+
+#define EMIT_RI10(_name, _op) \
+void _name (struct spe_function *p, int rT, int rA, int imm) \
+{ \
+   emit_RI10(p, _op, rT, rA, imm, __FUNCTION__);             \
+}
+
+#define EMIT_RI10s(_name, _op) \
+void _name (struct spe_function *p, int rT, int rA, int imm) \
+{ \
+   emit_RI10s(p, _op, rT, rA, imm, __FUNCTION__);             \
+}
+
+#define EMIT_RI16(_name, _op) \
+void _name (struct spe_function *p, int rT, int imm) \
+{ \
+   emit_RI16(p, _op, rT, imm, __FUNCTION__);                 \
+}
+
+#define EMIT_RI18(_name, _op) \
+void _name (struct spe_function *p, int rT, int imm) \
+{ \
+   emit_RI18(p, _op, rT, imm, __FUNCTION__);                 \
+}
+
+#define EMIT_I16(_name, _op) \
+void _name (struct spe_function *p, int imm) \
+{ \
+   emit_RI16(p, _op, 0, imm, __FUNCTION__);                  \
+}
+
+#include "rtasm_ppc_spe.h"
+
+
+
+/**
+ * Initialize an spe_function.
+ * \param code_size  initial size of instruction buffer to allocate, in bytes.
+ *                   If zero, use a default.
+ */
+void spe_init_func(struct spe_function *p, unsigned code_size)
+{
+    uint i;
+
+    if (!code_size)
+       code_size = 64;
+
+    p->num_inst = 0;
+    p->max_inst = code_size / SPE_INST_SIZE;
+    p->store = align_malloc(code_size, 16);
+
+    p->set_count = 0;
+    memset(p->regs, 0, SPE_NUM_REGS * sizeof(p->regs[0]));
+
+    /* Conservatively treat R0 - R2 and R80 - R127 as non-volatile.
+     */
+    p->regs[0] = p->regs[1] = p->regs[2] = 1;
+    for (i = 80; i <= 127; i++) {
+      p->regs[i] = 1;
+    }
+
+    p->print = FALSE;
+    p->indent = 0;
+}
+
+
+void spe_release_func(struct spe_function *p)
+{
+    assert(p->num_inst <= p->max_inst);
+    if (p->store != NULL) {
+        align_free(p->store);
+    }
+    p->store = NULL;
+}
+
+
+/** Return current code size in bytes. */
+unsigned spe_code_size(const struct spe_function *p)
+{
+   return p->num_inst * SPE_INST_SIZE;
+}
+
+
+/**
+ * Allocate a SPE register.
+ * \return register index or -1 if none left.
+ */
+int spe_allocate_available_register(struct spe_function *p)
+{
+   unsigned i;
+   for (i = 0; i < SPE_NUM_REGS; i++) {
+      if (p->regs[i] == 0) {
+         p->regs[i] = 1;
+         return i;
+      }
+   }
+
+   return -1;
+}
+
+
+/**
+ * Mark the given SPE register as "allocated".
+ */
+int spe_allocate_register(struct spe_function *p, int reg)
+{
+   assert(reg < SPE_NUM_REGS);
+   assert(p->regs[reg] == 0);
+   p->regs[reg] = 1;
+   return reg;
+}
+
+
+/**
+ * Mark the given SPE register as "unallocated".  Note that this should
+ * only be used on registers allocated in the current register set; an
+ * assertion will fail if an attempt is made to deallocate a register
+ * allocated in an earlier register set.
+ */
+void spe_release_register(struct spe_function *p, int reg)
+{
+   assert(reg >= 0);
+   assert(reg < SPE_NUM_REGS);
+   assert(p->regs[reg] == 1);
+
+   p->regs[reg] = 0;
+}
+
+/**
+ * Start a new set of registers.  This can be called if
+ * it will be difficult later to determine exactly what
+ * registers were actually allocated during a code generation
+ * sequence, and you really just want to deallocate all of them.
+ */
+void spe_allocate_register_set(struct spe_function *p)
+{
+   uint i;
+
+   /* Keep track of the set count.  If it ever wraps around to 0, 
+    * we're in trouble.
+    */
+   p->set_count++;
+   assert(p->set_count > 0);
+
+   /* Increment the allocation count of all registers currently
+    * allocated.  Then any registers that are allocated in this set
+    * will be the only ones with a count of 1; they'll all be released
+    * when the register set is released.
+    */
+   for (i = 0; i < SPE_NUM_REGS; i++) {
+      if (p->regs[i] > 0)
+         p->regs[i]++;
+   }
+}
+
+void spe_release_register_set(struct spe_function *p)
+{
+   uint i;
+
+   /* If the set count drops below zero, we're in trouble. */
+   assert(p->set_count > 0);
+   p->set_count--;
+
+   /* Drop the allocation level of all registers.  Any allocated
+    * during this register set will drop to 0 and then become
+    * available.
+    */
+   for (i = 0; i < SPE_NUM_REGS; i++) {
+      if (p->regs[i] > 0)
+         p->regs[i]--;
+   }
+}
+
+
+unsigned
+spe_get_registers_used(const struct spe_function *p, ubyte used[])
+{
+   unsigned i, num = 0;
+   /* only count registers in the range available to callers */
+   for (i = 2; i < 80; i++) {
+      if (p->regs[i]) {
+         used[num++] = i;
+      }
+   }
+   return num;
+}
+
+
+void
+spe_print_code(struct spe_function *p, boolean enable)
+{
+   p->print = enable;
+}
+
+
+void
+spe_indent(struct spe_function *p, int spaces)
+{
+   p->indent += spaces;
+}
+
+
+void
+spe_comment(struct spe_function *p, int rel_indent, const char *s)
+{
+   if (p->print) {
+      p->indent += rel_indent;
+      indent(p);
+      p->indent -= rel_indent;
+      printf("# %s\n", s);
+   }
+}
+
+
+/**
+ * Load quad word.
+ * NOTE: offset is in bytes and the least significant 4 bits must be zero!
+ */
+void spe_lqd(struct spe_function *p, int rT, int rA, int offset)
+{
+   const boolean pSave = p->print;
+
+   /* offset must be a multiple of 16 */
+   assert(offset % 16 == 0);
+   /* offset must fit in 10-bit signed int field, after shifting */
+   assert((offset >> 4) <= 511);
+   assert((offset >> 4) >= -512);
+
+   p->print = FALSE;
+   emit_RI10(p, 0x034, rT, rA, offset >> 4, "spe_lqd");
+   p->print = pSave;
+
+   if (p->print) {
+      indent(p);
+      printf("lqd\t%s, %d(%s)\n", reg_name(rT), offset, reg_name(rA));
+   }
+}
+
+
+/**
+ * Store quad word.
+ * NOTE: offset is in bytes and the least significant 4 bits must be zero!
+ */
+void spe_stqd(struct spe_function *p, int rT, int rA, int offset)
+{
+   const boolean pSave = p->print;
+
+   /* offset must be a multiple of 16 */
+   assert(offset % 16 == 0);
+   /* offset must fit in 10-bit signed int field, after shifting */
+   assert((offset >> 4) <= 511);
+   assert((offset >> 4) >= -512);
+
+   p->print = FALSE;
+   emit_RI10(p, 0x024, rT, rA, offset >> 4, "spe_stqd");
+   p->print = pSave;
+
+   if (p->print) {
+      indent(p);
+      printf("stqd\t%s, %d(%s)\n", reg_name(rT), offset, reg_name(rA));
+   }
+}
+
+
+/**
+ * For branch instructions:
+ * \param d  if 1, disable interupts if branch is taken
+ * \param e  if 1, enable interupts if branch is taken
+ * If d and e are both zero, don't change interupt status (right?)
+ */
+
+/** Branch Indirect to address in rA */
+void spe_bi(struct spe_function *p, int rA, int d, int e)
+{
+   emit_RI7(p, 0x1a8, 0, rA, (d << 5) | (e << 4), __FUNCTION__);
+}
+
+/** Interupt Return */
+void spe_iret(struct spe_function *p, int rA, int d, int e)
+{
+   emit_RI7(p, 0x1aa, 0, rA, (d << 5) | (e << 4), __FUNCTION__);
+}
+
+/** Branch indirect and set link on external data */
+void spe_bisled(struct spe_function *p, int rT, int rA, int d,
+		int e)
+{
+   emit_RI7(p, 0x1ab, rT, rA, (d << 5) | (e << 4), __FUNCTION__);
+}
+
+/** Branch indirect and set link.  Save PC in rT, jump to rA. */
+void spe_bisl(struct spe_function *p, int rT, int rA, int d,
+		int e)
+{
+   emit_RI7(p, 0x1a9, rT, rA, (d << 5) | (e << 4), __FUNCTION__);
+}
+
+/** Branch indirect if zero word.  If rT.word[0]==0, jump to rA. */
+void spe_biz(struct spe_function *p, int rT, int rA, int d, int e)
+{
+   emit_RI7(p, 0x128, rT, rA, (d << 5) | (e << 4), __FUNCTION__);
+}
+
+/** Branch indirect if non-zero word.  If rT.word[0]!=0, jump to rA. */
+void spe_binz(struct spe_function *p, int rT, int rA, int d, int e)
+{
+   emit_RI7(p, 0x129, rT, rA, (d << 5) | (e << 4), __FUNCTION__);
+}
+
+/** Branch indirect if zero halfword.  If rT.halfword[1]==0, jump to rA. */
+void spe_bihz(struct spe_function *p, int rT, int rA, int d, int e)
+{
+   emit_RI7(p, 0x12a, rT, rA, (d << 5) | (e << 4), __FUNCTION__);
+}
+
+/** Branch indirect if non-zero halfword.  If rT.halfword[1]!=0, jump to rA. */
+void spe_bihnz(struct spe_function *p, int rT, int rA, int d, int e)
+{
+   emit_RI7(p, 0x12b, rT, rA, (d << 5) | (e << 4), __FUNCTION__);
+}
+
+
+/* Hint-for-branch instructions
+ */
+#if 0
+hbr;
+hbra;
+hbrr;
+#endif
+
+
+/* Control instructions
+ */
+#if 0
+stop;
+EMIT_RR  (spe_stopd, 0x140);
+EMIT_    (spe_nop,   0x201);
+sync;
+EMIT_    (spe_dsync, 0x003);
+EMIT_R   (spe_mfspr, 0x00c);
+EMIT_R   (spe_mtspr, 0x10c);
+#endif
+
+
+/**
+ ** Helper / "macro" instructions.
+ ** Use somewhat verbose names as a reminder that these aren't native
+ ** SPE instructions.
+ **/
+
+
+void
+spe_load_float(struct spe_function *p, int rT, float x)
+{
+   if (x == 0.0f) {
+      spe_il(p, rT, 0x0);
+   }
+   else if (x == 0.5f) {
+      spe_ilhu(p, rT, 0x3f00);
+   }
+   else if (x == 1.0f) {
+      spe_ilhu(p, rT, 0x3f80);
+   }
+   else if (x == -1.0f) {
+      spe_ilhu(p, rT, 0xbf80);
+   }
+   else {
+      union {
+         float f;
+         unsigned u;
+      } bits;
+      bits.f = x;
+      spe_ilhu(p, rT, bits.u >> 16);
+      spe_iohl(p, rT, bits.u & 0xffff);
+   }
+}
+
+
+void
+spe_load_int(struct spe_function *p, int rT, int i)
+{
+   if (-32768 <= i && i <= 32767) {
+      spe_il(p, rT, i);
+   }
+   else {
+      spe_ilhu(p, rT, i >> 16);
+      if (i & 0xffff)
+         spe_iohl(p, rT, i & 0xffff);
+   }
+}
+
+void spe_load_uint(struct spe_function *p, int rT, uint ui)
+{
+   /* If the whole value is in the lower 18 bits, use ila, which
+    * doesn't sign-extend.  Otherwise, if the two halfwords of
+    * the constant are identical, use ilh.  Otherwise, if every byte of
+    * the desired value is 0x00 or 0xff, we can use Form Select Mask for
+    * Bytes Immediate (fsmbi) to load the value in a single instruction.
+    * Otherwise, in the general case, we have to use ilhu followed by iohl.
+    */
+   if ((ui & 0x0003ffff) == ui) {
+      spe_ila(p, rT, ui);
+   }
+   else if ((ui >> 16) == (ui & 0xffff)) {
+      spe_ilh(p, rT, ui & 0xffff);
+   }
+   else if (
+      ((ui & 0x000000ff) == 0 || (ui & 0x000000ff) == 0x000000ff) &&
+      ((ui & 0x0000ff00) == 0 || (ui & 0x0000ff00) == 0x0000ff00) &&
+      ((ui & 0x00ff0000) == 0 || (ui & 0x00ff0000) == 0x00ff0000) &&
+      ((ui & 0xff000000) == 0 || (ui & 0xff000000) == 0xff000000)
+   ) {
+      uint mask = 0;
+      /* fsmbi duplicates each bit in the given mask eight times,
+       * using a 16-bit value to initialize a 16-byte quadword.
+       * Each 4-bit nybble of the mask corresponds to a full word
+       * of the result; look at the value and figure out the mask
+       * (replicated for each word in the quadword), and then
+       * form the "select mask" to get the value.
+       */
+      if ((ui & 0x000000ff) == 0x000000ff) mask |= 0x1111;
+      if ((ui & 0x0000ff00) == 0x0000ff00) mask |= 0x2222;
+      if ((ui & 0x00ff0000) == 0x00ff0000) mask |= 0x4444;
+      if ((ui & 0xff000000) == 0xff000000) mask |= 0x8888;
+      spe_fsmbi(p, rT, mask);
+   }
+   else {
+      /* The general case: this usually uses two instructions, but
+       * may use only one if the low-order 16 bits of each word are 0.
+       */
+      spe_ilhu(p, rT, ui >> 16);
+      if (ui & 0xffff)
+         spe_iohl(p, rT, ui & 0xffff);
+   }
+}
+
+/**
+ * This function is constructed identically to spe_xor_uint() below.
+ * Changes to one should be made in the other.
+ */
+void
+spe_and_uint(struct spe_function *p, int rT, int rA, uint ui)
+{
+   /* If we can, emit a single instruction, either And Byte Immediate
+    * (which uses the same constant across each byte), And Halfword Immediate
+    * (which sign-extends a 10-bit immediate to 16 bits and uses that
+    * across each halfword), or And Word Immediate (which sign-extends
+    * a 10-bit immediate to 32 bits).
+    *
+    * Otherwise, we'll need to use a temporary register.
+    */
+   uint tmp;
+
+   /* If the upper 23 bits are all 0s or all 1s, sign extension
+    * will work and we can use And Word Immediate
+    */
+   tmp = ui & 0xfffffe00;
+   if (tmp == 0xfffffe00 || tmp  == 0) {
+      spe_andi(p, rT, rA, ui & 0x000003ff);
+      return;
+   }
+   
+   /* If the ui field is symmetric along halfword boundaries and
+    * the upper 7 bits of each halfword are all 0s or 1s, we
+    * can use And Halfword Immediate
+    */
+   tmp = ui & 0xfe00fe00;
+   if ((tmp == 0xfe00fe00 || tmp == 0) && ((ui >> 16) == (ui & 0x0000ffff))) {
+      spe_andhi(p, rT, rA, ui & 0x000003ff);
+      return;
+   }
+
+   /* If the ui field is symmetric in each byte, then we can use
+    * the And Byte Immediate instruction.
+    */
+   tmp = ui & 0x000000ff;
+   if ((ui >> 24) == tmp && ((ui >> 16) & 0xff) == tmp && ((ui >> 8) & 0xff) == tmp) {
+      spe_andbi(p, rT, rA, tmp);
+      return;
+   }
+
+   /* Otherwise, we'll have to use a temporary register. */
+   int tmp_reg = spe_allocate_available_register(p);
+   spe_load_uint(p, tmp_reg, ui);
+   spe_and(p, rT, rA, tmp_reg);
+   spe_release_register(p, tmp_reg);
+}
+
+
+/**
+ * This function is constructed identically to spe_and_uint() above.
+ * Changes to one should be made in the other.
+ */
+void
+spe_xor_uint(struct spe_function *p, int rT, int rA, uint ui)
+{
+   /* If we can, emit a single instruction, either Exclusive Or Byte 
+    * Immediate (which uses the same constant across each byte), Exclusive 
+    * Or Halfword Immediate (which sign-extends a 10-bit immediate to 
+    * 16 bits and uses that across each halfword), or Exclusive Or Word 
+    * Immediate (which sign-extends a 10-bit immediate to 32 bits).
+    *
+    * Otherwise, we'll need to use a temporary register.
+    */
+   uint tmp;
+
+   /* If the upper 23 bits are all 0s or all 1s, sign extension
+    * will work and we can use Exclusive Or Word Immediate
+    */
+   tmp = ui & 0xfffffe00;
+   if (tmp == 0xfffffe00 || tmp  == 0) {
+      spe_xori(p, rT, rA, ui & 0x000003ff);
+      return;
+   }
+   
+   /* If the ui field is symmetric along halfword boundaries and
+    * the upper 7 bits of each halfword are all 0s or 1s, we
+    * can use Exclusive Or Halfword Immediate
+    */
+   tmp = ui & 0xfe00fe00;
+   if ((tmp == 0xfe00fe00 || tmp == 0) && ((ui >> 16) == (ui & 0x0000ffff))) {
+      spe_xorhi(p, rT, rA, ui & 0x000003ff);
+      return;
+   }
+
+   /* If the ui field is symmetric in each byte, then we can use
+    * the Exclusive Or Byte Immediate instruction.
+    */
+   tmp = ui & 0x000000ff;
+   if ((ui >> 24) == tmp && ((ui >> 16) & 0xff) == tmp && ((ui >> 8) & 0xff) == tmp) {
+      spe_xorbi(p, rT, rA, tmp);
+      return;
+   }
+
+   /* Otherwise, we'll have to use a temporary register. */
+   int tmp_reg = spe_allocate_available_register(p);
+   spe_load_uint(p, tmp_reg, ui);
+   spe_xor(p, rT, rA, tmp_reg);
+   spe_release_register(p, tmp_reg);
+}
+
+void
+spe_compare_equal_uint(struct spe_function *p, int rT, int rA, uint ui)
+{
+   /* If the comparison value is 9 bits or less, it fits inside a
+    * Compare Equal Word Immediate instruction.
+    */
+   if ((ui & 0x000001ff) == ui) {
+      spe_ceqi(p, rT, rA, ui);
+   }
+   /* Otherwise, we're going to have to load a word first. */
+   else {
+      int tmp_reg = spe_allocate_available_register(p);
+      spe_load_uint(p, tmp_reg, ui);
+      spe_ceq(p, rT, rA, tmp_reg);
+      spe_release_register(p, tmp_reg);
+   }
+}
+
+void
+spe_compare_greater_uint(struct spe_function *p, int rT, int rA, uint ui)
+{
+   /* If the comparison value is 10 bits or less, it fits inside a
+    * Compare Logical Greater Than Word Immediate instruction.
+    */
+   if ((ui & 0x000003ff) == ui) {
+      spe_clgti(p, rT, rA, ui);
+   }
+   /* Otherwise, we're going to have to load a word first. */
+   else {
+      int tmp_reg = spe_allocate_available_register(p);
+      spe_load_uint(p, tmp_reg, ui);
+      spe_clgt(p, rT, rA, tmp_reg);
+      spe_release_register(p, tmp_reg);
+   }
+}
+
+void
+spe_splat(struct spe_function *p, int rT, int rA)
+{
+   /* Use a temporary, just in case rT == rA */
+   int tmp_reg = spe_allocate_available_register(p);
+   /* Duplicate bytes 0, 1, 2, and 3 across the whole register */
+   spe_ila(p, tmp_reg, 0x00010203);
+   spe_shufb(p, rT, rA, rA, tmp_reg);
+   spe_release_register(p, tmp_reg);
+}
+
+
+void
+spe_complement(struct spe_function *p, int rT, int rA)
+{
+   spe_nor(p, rT, rA, rA);
+}
+
+
+void
+spe_move(struct spe_function *p, int rT, int rA)
+{
+   /* Use different instructions depending on the instruction address
+    * to take advantage of the dual pipelines.
+    */
+   if (p->num_inst & 1)
+      spe_shlqbyi(p, rT, rA, 0);  /* odd pipe */
+   else
+      spe_ori(p, rT, rA, 0);  /* even pipe */
+}
+
+
+void
+spe_zero(struct spe_function *p, int rT)
+{
+   spe_xor(p, rT, rT, rT);
+}
+
+
+void
+spe_splat_word(struct spe_function *p, int rT, int rA, int word)
+{
+   assert(word >= 0);
+   assert(word <= 3);
+
+   if (word == 0) {
+      int tmp1 = rT;
+      spe_ila(p, tmp1, 66051);
+      spe_shufb(p, rT, rA, rA, tmp1);
+   }
+   else {
+      /* XXX review this, we may not need the rotqbyi instruction */
+      int tmp1 = rT;
+      int tmp2 = spe_allocate_available_register(p);
+
+      spe_ila(p, tmp1, 66051);
+      spe_rotqbyi(p, tmp2, rA, 4 * word);
+      spe_shufb(p, rT, tmp2, tmp2, tmp1);
+
+      spe_release_register(p, tmp2);
+   }
+}
+
+/**
+ * For each 32-bit float element of rA and rB, choose the smaller of the
+ * two, compositing them into the rT register.
+ * 
+ * The Float Compare Greater Than (fcgt) instruction will put 1s into
+ * compare_reg where rA > rB, and 0s where rA <= rB.
+ *
+ * Then the Select Bits (selb) instruction will take bits from rA where
+ * compare_reg is 0, and from rB where compare_reg is 1; i.e., from rA
+ * where rA <= rB and from rB where rB > rA, which is exactly the
+ * "min" operation.
+ *
+ * The compare_reg could in many cases be the same as rT, unless
+ * rT == rA || rt == rB.  But since this is common in constructions
+ * like "x = min(x, a)", we always allocate a new register to be safe.
+ */
+void 
+spe_float_min(struct spe_function *p, int rT, int rA, int rB)
+{
+   int compare_reg = spe_allocate_available_register(p);
+   spe_fcgt(p, compare_reg, rA, rB);
+   spe_selb(p, rT, rA, rB, compare_reg);
+   spe_release_register(p, compare_reg);
+}
+
+/**
+ * For each 32-bit float element of rA and rB, choose the greater of the
+ * two, compositing them into the rT register.
+ * 
+ * The logic is similar to that of spe_float_min() above; the only
+ * difference is that the registers on spe_selb() have been reversed,
+ * so that the larger of the two is selected instead of the smaller.
+ */
+void 
+spe_float_max(struct spe_function *p, int rT, int rA, int rB)
+{
+   int compare_reg = spe_allocate_available_register(p);
+   spe_fcgt(p, compare_reg, rA, rB);
+   spe_selb(p, rT, rB, rA, compare_reg);
+   spe_release_register(p, compare_reg);
+}
+
+#endif /* GALLIUM_CELL */
diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h
new file mode 100644
index 0000000000..65d9c77415
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h
@@ -0,0 +1,433 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file
+ * Real-time assembly generation interface for Cell B.E. SPEs.
+ * For details, see /opt/cell/sdk/docs/arch/SPU_ISA_v1.2_27Jan2007_pub.pdf
+ *
+ * \author Ian Romanick <idr@us.ibm.com>
+ * \author Brian Paul
+ */
+
+#ifndef RTASM_PPC_SPE_H
+#define RTASM_PPC_SPE_H
+
+/** 4 bytes per instruction */
+#define SPE_INST_SIZE 4
+
+/** number of general-purpose SIMD registers */
+#define SPE_NUM_REGS  128
+
+/** Return Address register (aka $lr / Link Register) */
+#define SPE_REG_RA  0
+
+/** Stack Pointer register (aka $sp) */
+#define SPE_REG_SP  1
+
+
+struct spe_function
+{
+   uint32_t *store;  /**< instruction buffer */
+   uint num_inst;
+   uint max_inst;
+
+   /**
+    * The "set count" reflects the number of nested register sets
+    * are allowed.  In the unlikely case that we exceed the set count,
+    * register allocation will start to be confused, which is critical
+    * enough that we check for it.
+    */
+   unsigned char set_count;
+
+   /** 
+    * Flags for used and unused registers.  Each byte corresponds to a
+    * register; a 0 in that byte means that the register is available.
+    * A value of 1 means that the register was allocated in the current
+    * register set.  Any other value N means that the register was allocated
+    * N register sets ago.
+     *
+     * \sa
+     * spe_allocate_register, spe_allocate_available_register,
+     * spe_allocate_register_set, spe_release_register_set, spe_release_register, 
+     */
+    unsigned char regs[SPE_NUM_REGS];
+
+    boolean print; /**< print/dump instructions as they're emitted? */
+    int indent;    /**< number of spaces to indent */
+};
+
+
+extern void spe_init_func(struct spe_function *p, uint code_size);
+extern void spe_release_func(struct spe_function *p);
+extern uint spe_code_size(const struct spe_function *p);
+
+extern int spe_allocate_available_register(struct spe_function *p);
+extern int spe_allocate_register(struct spe_function *p, int reg);
+extern void spe_release_register(struct spe_function *p, int reg);
+extern void spe_allocate_register_set(struct spe_function *p);
+extern void spe_release_register_set(struct spe_function *p);
+
+extern uint spe_get_registers_used(const struct spe_function *p, ubyte used[]);
+
+extern void spe_print_code(struct spe_function *p, boolean enable);
+extern void spe_indent(struct spe_function *p, int spaces);
+extern void spe_comment(struct spe_function *p, int rel_indent, const char *s);
+
+
+#endif /* RTASM_PPC_SPE_H */
+
+#ifndef EMIT
+#define EMIT(_name, _op) \
+    extern void _name (struct spe_function *p);
+#define EMIT_(_name, _op) \
+    extern void _name (struct spe_function *p, int rT);
+#define EMIT_R(_name, _op) \
+    extern void _name (struct spe_function *p, int rT, int rA);
+#define EMIT_RR(_name, _op) \
+    extern void _name (struct spe_function *p, int rT, int rA, int rB);
+#define EMIT_RRR(_name, _op) \
+    extern void _name (struct spe_function *p, int rT, int rA, int rB, int rC);
+#define EMIT_RI7(_name, _op) \
+    extern void _name (struct spe_function *p, int rT, int rA, int imm);
+#define EMIT_RI8(_name, _op, bias) \
+    extern void _name (struct spe_function *p, int rT, int rA, int imm);
+#define EMIT_RI10(_name, _op) \
+    extern void _name (struct spe_function *p, int rT, int rA, int imm);
+#define EMIT_RI10s(_name, _op) \
+    extern void _name (struct spe_function *p, int rT, int rA, int imm);
+#define EMIT_RI16(_name, _op) \
+    extern void _name (struct spe_function *p, int rT, int imm);
+#define EMIT_RI18(_name, _op) \
+    extern void _name (struct spe_function *p, int rT, int imm);
+#define EMIT_I16(_name, _op) \
+    extern void _name (struct spe_function *p, int imm);
+#define UNDEF_EMIT_MACROS
+#endif /* EMIT */
+
+
+/* Memory load / store instructions
+ */
+EMIT_RR  (spe_lqx,  0x1c4)
+EMIT_RI16(spe_lqa,  0x061)
+EMIT_RI16(spe_lqr,  0x067)
+EMIT_RR  (spe_stqx, 0x144)
+EMIT_RI16(spe_stqa, 0x041)
+EMIT_RI16(spe_stqr, 0x047)
+EMIT_RI7 (spe_cbd,  0x1f4)
+EMIT_RR  (spe_cbx,  0x1d4)
+EMIT_RI7 (spe_chd,  0x1f5)
+EMIT_RI7 (spe_chx,  0x1d5)
+EMIT_RI7 (spe_cwd,  0x1f6)
+EMIT_RI7 (spe_cwx,  0x1d6)
+EMIT_RI7 (spe_cdd,  0x1f7)
+EMIT_RI7 (spe_cdx,  0x1d7)
+
+
+/* Constant formation instructions
+ */
+EMIT_RI16(spe_ilh,   0x083)
+EMIT_RI16(spe_ilhu,  0x082)
+EMIT_RI16(spe_il,    0x081)
+EMIT_RI18(spe_ila,   0x021)
+EMIT_RI16(spe_iohl,  0x0c1)
+EMIT_RI16(spe_fsmbi, 0x065)
+
+
+
+/* Integer and logical instructions
+ */
+EMIT_RR  (spe_ah,      0x0c8)
+EMIT_RI10(spe_ahi,     0x01d)
+EMIT_RR  (spe_a,       0x0c0)
+EMIT_RI10s(spe_ai,      0x01c)
+EMIT_RR  (spe_sfh,     0x048)
+EMIT_RI10(spe_sfhi,    0x00d)
+EMIT_RR  (spe_sf,      0x040)
+EMIT_RI10(spe_sfi,     0x00c)
+EMIT_RR  (spe_addx,    0x340)
+EMIT_RR  (spe_cg,      0x0c2)
+EMIT_RR  (spe_cgx,     0x342)
+EMIT_RR  (spe_sfx,     0x341)
+EMIT_RR  (spe_bg,      0x042)
+EMIT_RR  (spe_bgx,     0x343)
+EMIT_RR  (spe_mpy,     0x3c4)
+EMIT_RR  (spe_mpyu,    0x3cc)
+EMIT_RI10(spe_mpyi,    0x074)
+EMIT_RI10(spe_mpyui,   0x075)
+EMIT_RRR (spe_mpya,    0x00c)
+EMIT_RR  (spe_mpyh,    0x3c5)
+EMIT_RR  (spe_mpys,    0x3c7)
+EMIT_RR  (spe_mpyhh,   0x3c6)
+EMIT_RR  (spe_mpyhha,  0x346)
+EMIT_RR  (spe_mpyhhu,  0x3ce)
+EMIT_RR  (spe_mpyhhau, 0x34e)
+EMIT_R   (spe_clz,     0x2a5)
+EMIT_R   (spe_cntb,    0x2b4)
+EMIT_R   (spe_fsmb,    0x1b6)
+EMIT_R   (spe_fsmh,    0x1b5)
+EMIT_R   (spe_fsm,     0x1b4)
+EMIT_R   (spe_gbb,     0x1b2)
+EMIT_R   (spe_gbh,     0x1b1)
+EMIT_R   (spe_gb,      0x1b0)
+EMIT_RR  (spe_avgb,    0x0d3)
+EMIT_RR  (spe_absdb,   0x053)
+EMIT_RR  (spe_sumb,    0x253)
+EMIT_R   (spe_xsbh,    0x2b6)
+EMIT_R   (spe_xshw,    0x2ae)
+EMIT_R   (spe_xswd,    0x2a6)
+EMIT_RR  (spe_and,     0x0c1)
+EMIT_RR  (spe_andc,    0x2c1)
+EMIT_RI10s(spe_andbi,   0x016)
+EMIT_RI10s(spe_andhi,   0x015)
+EMIT_RI10s(spe_andi,    0x014)
+EMIT_RR  (spe_or,      0x041)
+EMIT_RR  (spe_orc,     0x2c9)
+EMIT_RI10s(spe_orbi,    0x006)
+EMIT_RI10s(spe_orhi,    0x005)
+EMIT_RI10s(spe_ori,     0x004)
+EMIT_R   (spe_orx,     0x1f0)
+EMIT_RR  (spe_xor,     0x241)
+EMIT_RI10s(spe_xorbi,   0x046)
+EMIT_RI10s(spe_xorhi,   0x045)
+EMIT_RI10s(spe_xori,    0x044)
+EMIT_RR  (spe_nand,    0x0c9)
+EMIT_RR  (spe_nor,     0x049)
+EMIT_RR  (spe_eqv,     0x249)
+EMIT_RRR (spe_selb,    0x008)
+EMIT_RRR (spe_shufb,   0x00b)
+
+
+/* Shift and rotate instructions
+ */
+EMIT_RR  (spe_shlh,      0x05f)
+EMIT_RI7 (spe_shlhi,     0x07f)
+EMIT_RR  (spe_shl,       0x05b)
+EMIT_RI7 (spe_shli,      0x07b)
+EMIT_RR  (spe_shlqbi,    0x1db)
+EMIT_RI7 (spe_shlqbii,   0x1fb)
+EMIT_RR  (spe_shlqby,    0x1df)
+EMIT_RI7 (spe_shlqbyi,   0x1ff)
+EMIT_RR  (spe_shlqbybi,  0x1cf)
+EMIT_RR  (spe_roth,      0x05c)
+EMIT_RI7 (spe_rothi,     0x07c)
+EMIT_RR  (spe_rot,       0x058)
+EMIT_RI7 (spe_roti,      0x078)
+EMIT_RR  (spe_rotqby,    0x1dc)
+EMIT_RI7 (spe_rotqbyi,   0x1fc)
+EMIT_RR  (spe_rotqbybi,  0x1cc)
+EMIT_RR  (spe_rotqbi,    0x1d8)
+EMIT_RI7 (spe_rotqbii,   0x1f8)
+EMIT_RR  (spe_rothm,     0x05d)
+EMIT_RI7 (spe_rothmi,    0x07d)
+EMIT_RR  (spe_rotm,      0x059)
+EMIT_RI7 (spe_rotmi,     0x079)
+EMIT_RR  (spe_rotqmby,   0x1dd)
+EMIT_RI7 (spe_rotqmbyi,  0x1fd)
+EMIT_RR  (spe_rotqmbybi, 0x1cd)
+EMIT_RR  (spe_rotqmbi,   0x1c9)
+EMIT_RI7 (spe_rotqmbii,  0x1f9)
+EMIT_RR  (spe_rotmah,    0x05e)
+EMIT_RI7 (spe_rotmahi,   0x07e)
+EMIT_RR  (spe_rotma,     0x05a)
+EMIT_RI7 (spe_rotmai,    0x07a)
+
+
+/* Compare, branch, and halt instructions
+ */
+EMIT_RR  (spe_heq,       0x3d8)
+EMIT_RI10(spe_heqi,      0x07f)
+EMIT_RR  (spe_hgt,       0x258)
+EMIT_RI10(spe_hgti,      0x04f)
+EMIT_RR  (spe_hlgt,      0x2d8)
+EMIT_RI10(spe_hlgti,     0x05f)
+EMIT_RR  (spe_ceqb,      0x3d0)
+EMIT_RI10(spe_ceqbi,     0x07e)
+EMIT_RR  (spe_ceqh,      0x3c8)
+EMIT_RI10(spe_ceqhi,     0x07d)
+EMIT_RR  (spe_ceq,       0x3c0)
+EMIT_RI10(spe_ceqi,      0x07c)
+EMIT_RR  (spe_cgtb,      0x250)
+EMIT_RI10(spe_cgtbi,     0x04e)
+EMIT_RR  (spe_cgth,      0x248)
+EMIT_RI10(spe_cgthi,     0x04d)
+EMIT_RR  (spe_cgt,       0x240)
+EMIT_RI10(spe_cgti,      0x04c)
+EMIT_RR  (spe_clgtb,     0x2d0)
+EMIT_RI10(spe_clgtbi,    0x05e)
+EMIT_RR  (spe_clgth,     0x2c8)
+EMIT_RI10(spe_clgthi,    0x05d)
+EMIT_RR  (spe_clgt,      0x2c0)
+EMIT_RI10(spe_clgti,     0x05c)
+EMIT_I16 (spe_br,        0x064)
+EMIT_I16 (spe_bra,       0x060)
+EMIT_RI16(spe_brsl,      0x066)
+EMIT_RI16(spe_brasl,     0x062)
+EMIT_RI16(spe_brnz,      0x042)
+EMIT_RI16(spe_brz,       0x040)
+EMIT_RI16(spe_brhnz,     0x046)
+EMIT_RI16(spe_brhz,      0x044)
+
+/* Control instructions
+ */
+EMIT     (spe_lnop,      0x001)
+
+extern void
+spe_lqd(struct spe_function *p, int rT, int rA, int offset);
+
+extern void
+spe_stqd(struct spe_function *p, int rT, int rA, int offset);
+
+extern void spe_bi(struct spe_function *p, int rA, int d, int e);
+extern void spe_iret(struct spe_function *p, int rA, int d, int e);
+extern void spe_bisled(struct spe_function *p, int rT, int rA,
+    int d, int e);
+extern void spe_bisl(struct spe_function *p, int rT, int rA,
+    int d, int e);
+extern void spe_biz(struct spe_function *p, int rT, int rA,
+    int d, int e);
+extern void spe_binz(struct spe_function *p, int rT, int rA,
+    int d, int e);
+extern void spe_bihz(struct spe_function *p, int rT, int rA,
+    int d, int e);
+extern void spe_bihnz(struct spe_function *p, int rT, int rA,
+    int d, int e);
+
+
+/** Load/splat immediate float into rT. */
+extern void
+spe_load_float(struct spe_function *p, int rT, float x);
+
+/** Load/splat immediate int into rT. */
+extern void
+spe_load_int(struct spe_function *p, int rT, int i);
+
+/** Load/splat immediate unsigned int into rT. */
+extern void
+spe_load_uint(struct spe_function *p, int rT, uint ui);
+
+/** And immediate value into rT. */
+extern void
+spe_and_uint(struct spe_function *p, int rT, int rA, uint ui);
+
+/** Xor immediate value into rT. */
+extern void
+spe_xor_uint(struct spe_function *p, int rT, int rA, uint ui);
+
+/** Compare equal with immediate value. */
+extern void
+spe_compare_equal_uint(struct spe_function *p, int rT, int rA, uint ui);
+
+/** Compare greater with immediate value. */
+extern void
+spe_compare_greater_uint(struct spe_function *p, int rT, int rA, uint ui);
+
+/** Replicate word 0 of rA across rT. */
+extern void
+spe_splat(struct spe_function *p, int rT, int rA);
+
+/** rT = complement_all_bits(rA). */
+extern void
+spe_complement(struct spe_function *p, int rT, int rA);
+
+/** rT = rA. */
+extern void
+spe_move(struct spe_function *p, int rT, int rA);
+
+/** rT = {0,0,0,0}. */
+extern void
+spe_zero(struct spe_function *p, int rT);
+
+/** rT = splat(rA, word) */
+extern void
+spe_splat_word(struct spe_function *p, int rT, int rA, int word);
+
+/** rT = float min(rA, rB) */
+extern void
+spe_float_min(struct spe_function *p, int rT, int rA, int rB);
+
+/** rT = float max(rA, rB) */
+extern void
+spe_float_max(struct spe_function *p, int rT, int rA, int rB);
+
+
+/* Floating-point instructions
+ */
+EMIT_RR  (spe_fa,         0x2c4)
+EMIT_RR  (spe_dfa,        0x2cc)
+EMIT_RR  (spe_fs,         0x2c5)
+EMIT_RR  (spe_dfs,        0x2cd)
+EMIT_RR  (spe_fm,         0x2c6)
+EMIT_RR  (spe_dfm,        0x2ce)
+EMIT_RRR (spe_fma,        0x00e)
+EMIT_RR  (spe_dfma,       0x35c)
+EMIT_RRR (spe_fnms,       0x00d)
+EMIT_RR  (spe_dfnms,      0x35e)
+EMIT_RRR (spe_fms,        0x00f)
+EMIT_RR  (spe_dfms,       0x35d)
+EMIT_RR  (spe_dfnma,      0x35f)
+EMIT_R   (spe_frest,      0x1b8)
+EMIT_R   (spe_frsqest,    0x1b9)
+EMIT_RR  (spe_fi,         0x3d4)
+EMIT_RI8 (spe_csflt,      0x1da, 155)
+EMIT_RI8 (spe_cflts,      0x1d8, 173)
+EMIT_RI8 (spe_cuflt,      0x1db, 155)
+EMIT_RI8 (spe_cfltu,      0x1d9, 173)
+EMIT_R   (spe_frds,       0x3b9)
+EMIT_R   (spe_fesd,       0x3b8)
+EMIT_RR  (spe_dfceq,      0x3c3)
+EMIT_RR  (spe_dfcmeq,     0x3cb)
+EMIT_RR  (spe_dfcgt,      0x2c3)
+EMIT_RR  (spe_dfcmgt,     0x2cb)
+EMIT_RI7 (spe_dftsv,      0x3bf)
+EMIT_RR  (spe_fceq,       0x3c2)
+EMIT_RR  (spe_fcmeq,      0x3ca)
+EMIT_RR  (spe_fcgt,       0x2c2)
+EMIT_RR  (spe_fcmgt,      0x2ca)
+EMIT_R   (spe_fscrwr,     0x3ba)
+EMIT_    (spe_fscrrd,     0x398)
+
+
+/* Channel instructions
+ */
+EMIT_R   (spe_rdch,       0x00d)
+EMIT_R   (spe_rdchcnt,    0x00f)
+EMIT_R   (spe_wrch,       0x10d)
+
+
+#ifdef UNDEF_EMIT_MACROS
+#undef EMIT
+#undef EMIT_
+#undef EMIT_R
+#undef EMIT_RR
+#undef EMIT_RRR
+#undef EMIT_RI7
+#undef EMIT_RI8
+#undef EMIT_RI10
+#undef EMIT_RI10s
+#undef EMIT_RI16
+#undef EMIT_RI18
+#undef EMIT_I16
+#undef UNDEF_EMIT_MACROS
+#endif /* EMIT_ */
diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
new file mode 100644
index 0000000000..99ee74cf14
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c
@@ -0,0 +1,1748 @@
+/**************************************************************************
+ *
+ * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_ARCH_X86)
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "util/u_pointer.h"
+
+#include "rtasm_execmem.h"
+#include "rtasm_x86sse.h"
+
+#define DISASSEM 0
+#define X86_TWOB 0x0f
+
+
+#define DUMP_SSE  0
+
+
+void x86_print_reg( struct x86_reg reg )
+{
+   if (reg.mod != mod_REG) 
+      debug_printf( "[" );
+      
+   switch( reg.file ) {
+   case file_REG32:
+      switch( reg.idx ) {
+      case reg_AX: debug_printf( "EAX" ); break;
+      case reg_CX: debug_printf( "ECX" ); break;
+      case reg_DX: debug_printf( "EDX" ); break;
+      case reg_BX: debug_printf( "EBX" ); break;
+      case reg_SP: debug_printf( "ESP" ); break;
+      case reg_BP: debug_printf( "EBP" ); break;
+      case reg_SI: debug_printf( "ESI" ); break;
+      case reg_DI: debug_printf( "EDI" ); break;
+      }
+      break;
+   case file_MMX:
+      debug_printf( "MMX%u", reg.idx );
+      break;
+   case file_XMM:
+      debug_printf( "XMM%u", reg.idx );
+      break;
+   case file_x87:
+      debug_printf( "fp%u", reg.idx );
+      break;
+   }
+
+   if (reg.mod == mod_DISP8 ||
+       reg.mod == mod_DISP32)
+      debug_printf("+%d", reg.disp);
+
+   if (reg.mod != mod_REG) 
+      debug_printf( "]" );
+}
+
+#if DUMP_SSE
+
+#define DUMP_START() debug_printf( "\n" )
+#define DUMP_END() debug_printf( "\n" )
+
+#define DUMP() do {                             \
+   const char *foo = __FUNCTION__;              \
+   while (*foo && *foo != '_')                  \
+      foo++;                                    \
+   if  (*foo)                                   \
+      foo++;                                    \
+   debug_printf( "\n% 4x% 15s ", p->csr - p->store, foo );             \
+} while (0)
+
+#define DUMP_I( I ) do {                        \
+   DUMP();                                      \
+   debug_printf( "%u", I );                     \
+} while( 0 )
+
+#define DUMP_R( R0 ) do {                       \
+   DUMP();                                      \
+   x86_print_reg( R0 );                            \
+} while( 0 )
+
+#define DUMP_RR( R0, R1 ) do {                  \
+   DUMP();                                      \
+   x86_print_reg( R0 );                            \
+   debug_printf( ", " );                        \
+   x86_print_reg( R1 );                            \
+} while( 0 )
+
+#define DUMP_RI( R0, I ) do {                   \
+   DUMP();                                      \
+   x86_print_reg( R0 );                            \
+   debug_printf( ", %u", I );                   \
+} while( 0 )
+
+#define DUMP_RRI( R0, R1, I ) do {              \
+   DUMP();                                      \
+   x86_print_reg( R0 );                            \
+   debug_printf( ", " );                        \
+   x86_print_reg( R1 );                            \
+   debug_printf( ", %u", I );                   \
+} while( 0 )
+
+#else
+
+#define DUMP_START()
+#define DUMP_END()
+#define DUMP( )
+#define DUMP_I( I )
+#define DUMP_R( R0 )
+#define DUMP_RR( R0, R1 )
+#define DUMP_RI( R0, I )
+#define DUMP_RRI( R0, R1, I )
+
+#endif
+
+
+static void do_realloc( struct x86_function *p )
+{
+   if (p->store == p->error_overflow) {
+      p->csr = p->store;
+   }
+   else if (p->size == 0) {
+      p->size = 1024;
+      p->store = rtasm_exec_malloc(p->size);
+      p->csr = p->store;
+   }
+   else {
+      uintptr_t used = pointer_to_uintptr( p->csr ) - pointer_to_uintptr( p->store );
+      unsigned char *tmp = p->store;
+      p->size *= 2;
+      p->store = rtasm_exec_malloc(p->size);
+
+      if (p->store) {
+         memcpy(p->store, tmp, used);
+         p->csr = p->store + used;
+      }
+      else {
+         p->csr = p->store;
+      }
+
+      rtasm_exec_free(tmp);
+   }
+
+   if (p->store == NULL) {
+      p->store = p->csr = p->error_overflow;
+      p->size = sizeof(p->error_overflow);
+   }
+}
+
+/* Emit bytes to the instruction stream:
+ */
+static unsigned char *reserve( struct x86_function *p, int bytes )
+{
+   if (p->csr + bytes - p->store > (int) p->size)
+      do_realloc(p);
+
+   {
+      unsigned char *csr = p->csr;
+      p->csr += bytes;
+      return csr;
+   }
+}
+
+
+
+static void emit_1b( struct x86_function *p, char b0 )
+{
+   char *csr = (char *)reserve(p, 1);
+   *csr = b0;
+}
+
+static void emit_1i( struct x86_function *p, int i0 )
+{
+   int *icsr = (int *)reserve(p, sizeof(i0));
+   *icsr = i0;
+}
+
+static void emit_1ub( struct x86_function *p, unsigned char b0 )
+{
+   unsigned char *csr = reserve(p, 1);
+   *csr++ = b0;
+}
+
+static void emit_2ub( struct x86_function *p, unsigned char b0, unsigned char b1 )
+{
+   unsigned char *csr = reserve(p, 2);
+   *csr++ = b0;
+   *csr++ = b1;
+}
+
+static void emit_3ub( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2 )
+{
+   unsigned char *csr = reserve(p, 3);
+   *csr++ = b0;
+   *csr++ = b1;
+   *csr++ = b2;
+}
+
+
+/* Build a modRM byte + possible displacement.  No treatment of SIB
+ * indexing.  BZZT - no way to encode an absolute address.
+ *
+ * This is the "/r" field in the x86 manuals...
+ */
+static void emit_modrm( struct x86_function *p, 
+			struct x86_reg reg, 
+			struct x86_reg regmem )
+{
+   unsigned char val = 0;
+   
+   assert(reg.mod == mod_REG);
+   
+   val |= regmem.mod << 6;     	/* mod field */
+   val |= reg.idx << 3;		/* reg field */
+   val |= regmem.idx;		/* r/m field */
+   
+   emit_1ub(p, val);
+
+   /* Oh-oh we've stumbled into the SIB thing.
+    */
+   if (regmem.file == file_REG32 &&
+       regmem.idx == reg_SP &&
+       regmem.mod != mod_REG) {
+      emit_1ub(p, 0x24);		/* simplistic! */
+   }
+
+   switch (regmem.mod) {
+   case mod_REG:
+   case mod_INDIRECT:
+      break;
+   case mod_DISP8:
+      emit_1b(p, (char) regmem.disp);
+      break;
+   case mod_DISP32:
+      emit_1i(p, regmem.disp);
+      break;
+   default:
+      assert(0);
+      break;
+   }
+}
+
+/* Emits the "/0".."/7" specialized versions of the modrm ("/r") bytes.
+ */
+static void emit_modrm_noreg( struct x86_function *p,
+			      unsigned op,
+			      struct x86_reg regmem )
+{
+   struct x86_reg dummy = x86_make_reg(file_REG32, op);
+   emit_modrm(p, dummy, regmem);
+}
+
+/* Many x86 instructions have two opcodes to cope with the situations
+ * where the destination is a register or memory reference
+ * respectively.  This function selects the correct opcode based on
+ * the arguments presented.
+ */
+static void emit_op_modrm( struct x86_function *p,
+			   unsigned char op_dst_is_reg, 
+			   unsigned char op_dst_is_mem,
+			   struct x86_reg dst,
+			   struct x86_reg src )
+{  
+   switch (dst.mod) {
+   case mod_REG:
+      emit_1ub(p, op_dst_is_reg);
+      emit_modrm(p, dst, src);
+      break;
+   case mod_INDIRECT:
+   case mod_DISP32:
+   case mod_DISP8:
+      assert(src.mod == mod_REG);
+      emit_1ub(p, op_dst_is_mem);
+      emit_modrm(p, src, dst);
+      break;
+   default:
+      assert(0);
+      break;
+   }
+}
+
+
+
+
+
+
+
+/* Create and manipulate registers and regmem values:
+ */
+struct x86_reg x86_make_reg( enum x86_reg_file file,
+			     enum x86_reg_name idx )
+{
+   struct x86_reg reg;
+
+   reg.file = file;
+   reg.idx = idx;
+   reg.mod = mod_REG;
+   reg.disp = 0;
+
+   return reg;
+}
+
+struct x86_reg x86_make_disp( struct x86_reg reg,
+			      int disp )
+{
+   assert(reg.file == file_REG32);
+
+   if (reg.mod == mod_REG)
+      reg.disp = disp;
+   else
+      reg.disp += disp;
+
+   if (reg.disp == 0 && reg.idx != reg_BP)
+      reg.mod = mod_INDIRECT;
+   else if (reg.disp <= 127 && reg.disp >= -128)
+      reg.mod = mod_DISP8;
+   else
+      reg.mod = mod_DISP32;
+
+   return reg;
+}
+
+struct x86_reg x86_deref( struct x86_reg reg )
+{
+   return x86_make_disp(reg, 0);
+}
+
+struct x86_reg x86_get_base_reg( struct x86_reg reg )
+{
+   return x86_make_reg( reg.file, reg.idx );
+}
+
+int x86_get_label( struct x86_function *p )
+{
+   return p->csr - p->store;
+}
+
+
+
+/***********************************************************************
+ * x86 instructions
+ */
+
+
+void x86_jcc( struct x86_function *p,
+	      enum x86_cc cc,
+	      int label )
+{
+   int offset = label - (x86_get_label(p) + 2);
+   DUMP_I(cc);
+   
+   if (offset < 0) {
+      /*assert(p->csr - p->store > -offset);*/
+      if (p->csr - p->store <= -offset) {
+         /* probably out of memory (using the error_overflow buffer) */
+         return;
+      }
+   }
+
+   if (offset <= 127 && offset >= -128) {
+      emit_1ub(p, 0x70 + cc);
+      emit_1b(p, (char) offset);
+   }
+   else {
+      offset = label - (x86_get_label(p) + 6);
+      emit_2ub(p, 0x0f, 0x80 + cc);
+      emit_1i(p, offset);
+   }
+}
+
+/* Always use a 32bit offset for forward jumps:
+ */
+int x86_jcc_forward( struct x86_function *p,
+                     enum x86_cc cc )
+{
+   DUMP_I(cc);
+   emit_2ub(p, 0x0f, 0x80 + cc);
+   emit_1i(p, 0);
+   return x86_get_label(p);
+}
+
+int x86_jmp_forward( struct x86_function *p)
+{
+   DUMP();
+   emit_1ub(p, 0xe9);
+   emit_1i(p, 0);
+   return x86_get_label(p);
+}
+
+int x86_call_forward( struct x86_function *p)
+{
+   DUMP();
+
+   emit_1ub(p, 0xe8);
+   emit_1i(p, 0);
+   return x86_get_label(p);
+}
+
+/* Fixup offset from forward jump:
+ */
+void x86_fixup_fwd_jump( struct x86_function *p,
+			 int fixup )
+{
+   *(int *)(p->store + fixup - 4) = x86_get_label(p) - fixup;
+}
+
+void x86_jmp( struct x86_function *p, int label)
+{
+   DUMP_I( label );
+   emit_1ub(p, 0xe9);
+   emit_1i(p, label - x86_get_label(p) - 4);
+}
+
+void x86_call( struct x86_function *p, struct x86_reg reg)
+{
+   DUMP_R( reg );
+   emit_1ub(p, 0xff);
+   emit_modrm_noreg(p, 2, reg);
+}
+
+
+void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm )
+{
+   DUMP_RI( dst, imm );
+   assert(dst.file == file_REG32);
+   assert(dst.mod == mod_REG);
+   emit_1ub(p, 0xb8 + dst.idx);
+   emit_1i(p, imm);
+}
+
+/**
+ * Immediate group 1 instructions.
+ */
+static INLINE void 
+x86_group1_imm( struct x86_function *p, 
+                unsigned op, struct x86_reg dst, int imm )
+{
+   assert(dst.file == file_REG32);
+   assert(dst.mod == mod_REG);
+   if(-0x80 <= imm && imm < 0x80) {
+      emit_1ub(p, 0x83);
+      emit_modrm_noreg(p, op, dst);
+      emit_1b(p, (char)imm);
+   }
+   else {
+      emit_1ub(p, 0x81);
+      emit_modrm_noreg(p, op, dst);
+      emit_1i(p, imm);
+   }
+}
+
+void x86_add_imm( struct x86_function *p, struct x86_reg dst, int imm )
+{
+   DUMP_RI( dst, imm );
+   x86_group1_imm(p, 0, dst, imm);
+}
+
+void x86_or_imm( struct x86_function *p, struct x86_reg dst, int imm )
+{
+   DUMP_RI( dst, imm );
+   x86_group1_imm(p, 1, dst, imm);
+}
+
+void x86_and_imm( struct x86_function *p, struct x86_reg dst, int imm )
+{
+   DUMP_RI( dst, imm );
+   x86_group1_imm(p, 4, dst, imm);
+}
+
+void x86_sub_imm( struct x86_function *p, struct x86_reg dst, int imm )
+{
+   DUMP_RI( dst, imm );
+   x86_group1_imm(p, 5, dst, imm);
+}
+
+void x86_xor_imm( struct x86_function *p, struct x86_reg dst, int imm )
+{
+   DUMP_RI( dst, imm );
+   x86_group1_imm(p, 6, dst, imm);
+}
+
+void x86_cmp_imm( struct x86_function *p, struct x86_reg dst, int imm )
+{
+   DUMP_RI( dst, imm );
+   x86_group1_imm(p, 7, dst, imm);
+}
+
+
+void x86_push( struct x86_function *p,
+	       struct x86_reg reg )
+{
+   DUMP_R( reg );
+   if (reg.mod == mod_REG)
+      emit_1ub(p, 0x50 + reg.idx);
+   else 
+   {
+      emit_1ub(p, 0xff);
+      emit_modrm_noreg(p, 6, reg);
+   }
+
+
+   p->stack_offset += 4;
+}
+
+void x86_push_imm32( struct x86_function *p,
+                     int imm32 )
+{
+   DUMP_I( imm32 );
+   emit_1ub(p, 0x68);
+   emit_1i(p,  imm32);
+
+   p->stack_offset += 4;
+}
+
+
+void x86_pop( struct x86_function *p,
+	      struct x86_reg reg )
+{
+   DUMP_R( reg );
+   assert(reg.mod == mod_REG);
+   emit_1ub(p, 0x58 + reg.idx);
+   p->stack_offset -= 4;
+}
+
+void x86_inc( struct x86_function *p,
+	      struct x86_reg reg )
+{
+   DUMP_R( reg );
+   assert(reg.mod == mod_REG);
+   emit_1ub(p, 0x40 + reg.idx);
+}
+
+void x86_dec( struct x86_function *p,
+	      struct x86_reg reg )
+{
+   DUMP_R( reg );
+   assert(reg.mod == mod_REG);
+   emit_1ub(p, 0x48 + reg.idx);
+}
+
+void x86_ret( struct x86_function *p )
+{
+   DUMP();
+   assert(p->stack_offset == 0);
+   emit_1ub(p, 0xc3);
+}
+
+void x86_retw( struct x86_function *p, unsigned short imm )
+{
+   DUMP();
+   emit_3ub(p, 0xc2, imm & 0xff, (imm >> 8) & 0xff);
+}
+
+void x86_sahf( struct x86_function *p )
+{
+   DUMP();
+   emit_1ub(p, 0x9e);
+}
+
+void x86_mov( struct x86_function *p,
+	      struct x86_reg dst,
+	      struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_op_modrm( p, 0x8b, 0x89, dst, src );
+}
+
+void x86_xor( struct x86_function *p,
+	      struct x86_reg dst,
+	      struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_op_modrm( p, 0x33, 0x31, dst, src );
+}
+
+void x86_cmp( struct x86_function *p,
+	      struct x86_reg dst,
+	      struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_op_modrm( p, 0x3b, 0x39, dst, src );
+}
+
+void x86_lea( struct x86_function *p,
+	      struct x86_reg dst,
+	      struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_1ub(p, 0x8d);
+   emit_modrm( p, dst, src );
+}
+
+void x86_test( struct x86_function *p,
+	       struct x86_reg dst,
+	       struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_1ub(p, 0x85);
+   emit_modrm( p, dst, src );
+}
+
+void x86_add( struct x86_function *p,
+	       struct x86_reg dst,
+	       struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_op_modrm(p, 0x03, 0x01, dst, src );
+}
+
+/* Calculate EAX * src, results in EDX:EAX.
+ */
+void x86_mul( struct x86_function *p,
+	       struct x86_reg src )
+{
+   DUMP_R(  src );
+   emit_1ub(p, 0xf7);
+   emit_modrm_noreg(p, 4, src );
+}
+
+
+void x86_imul( struct x86_function *p,
+	       struct x86_reg dst,
+	       struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0xAF);
+   emit_modrm(p, dst, src);
+}
+
+
+void x86_sub( struct x86_function *p,
+	       struct x86_reg dst,
+	       struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_op_modrm(p, 0x2b, 0x29, dst, src );
+}
+
+void x86_or( struct x86_function *p,
+             struct x86_reg dst,
+             struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_op_modrm( p, 0x0b, 0x09, dst, src );
+}
+
+void x86_and( struct x86_function *p,
+              struct x86_reg dst,
+              struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_op_modrm( p, 0x23, 0x21, dst, src );
+}
+
+
+
+/***********************************************************************
+ * SSE instructions
+ */
+
+void sse_prefetchnta( struct x86_function *p, struct x86_reg ptr)
+{
+   DUMP_R( ptr );
+   assert(ptr.mod != mod_REG);
+   emit_2ub(p, 0x0f, 0x18);
+   emit_modrm_noreg(p, 0, ptr);
+}
+
+void sse_prefetch0( struct x86_function *p, struct x86_reg ptr)
+{
+   DUMP_R( ptr );
+   assert(ptr.mod != mod_REG);
+   emit_2ub(p, 0x0f, 0x18);
+   emit_modrm_noreg(p, 1, ptr);
+}
+
+void sse_prefetch1( struct x86_function *p, struct x86_reg ptr)
+{
+   DUMP_R( ptr );
+   assert(ptr.mod != mod_REG);
+   emit_2ub(p, 0x0f, 0x18);
+   emit_modrm_noreg(p, 2, ptr);
+}
+
+void sse_movntps( struct x86_function *p, 
+                  struct x86_reg dst,
+                  struct x86_reg src)
+{
+   DUMP_RR( dst, src );
+
+   assert(dst.mod != mod_REG);
+   assert(src.mod == mod_REG);
+   emit_2ub(p, 0x0f, 0x2b);
+   emit_modrm(p, src, dst);
+}
+
+
+
+
+void sse_movss( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, 0xF3, X86_TWOB);
+   emit_op_modrm( p, 0x10, 0x11, dst, src );
+}
+
+void sse_movaps( struct x86_function *p,
+		 struct x86_reg dst,
+		 struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_1ub(p, X86_TWOB);
+   emit_op_modrm( p, 0x28, 0x29, dst, src );
+}
+
+void sse_movups( struct x86_function *p,
+		 struct x86_reg dst,
+		 struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_1ub(p, X86_TWOB);
+   emit_op_modrm( p, 0x10, 0x11, dst, src );
+}
+
+void sse_movhps( struct x86_function *p,
+		 struct x86_reg dst,
+		 struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   assert(dst.mod != mod_REG || src.mod != mod_REG);
+   emit_1ub(p, X86_TWOB);
+   emit_op_modrm( p, 0x16, 0x17, dst, src ); /* cf movlhps */
+}
+
+void sse_movlps( struct x86_function *p,
+		 struct x86_reg dst,
+		 struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   assert(dst.mod != mod_REG || src.mod != mod_REG);
+   emit_1ub(p, X86_TWOB);
+   emit_op_modrm( p, 0x12, 0x13, dst, src ); /* cf movhlps */
+}
+
+void sse_maxps( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x5F);
+   emit_modrm( p, dst, src );
+}
+
+void sse_maxss( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0xF3, X86_TWOB, 0x5F);
+   emit_modrm( p, dst, src );
+}
+
+void sse_divss( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0xF3, X86_TWOB, 0x5E);
+   emit_modrm( p, dst, src );
+}
+
+void sse_minps( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x5D);
+   emit_modrm( p, dst, src );
+}
+
+void sse_subps( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x5C);
+   emit_modrm( p, dst, src );
+}
+
+void sse_mulps( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x59);
+   emit_modrm( p, dst, src );
+}
+
+void sse_mulss( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0xF3, X86_TWOB, 0x59);
+   emit_modrm( p, dst, src );
+}
+
+void sse_addps( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x58);
+   emit_modrm( p, dst, src );
+}
+
+void sse_addss( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0xF3, X86_TWOB, 0x58);
+   emit_modrm( p, dst, src );
+}
+
+void sse_andnps( struct x86_function *p,
+                 struct x86_reg dst,
+                 struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x55);
+   emit_modrm( p, dst, src );
+}
+
+void sse_andps( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x54);
+   emit_modrm( p, dst, src );
+}
+
+void sse_rsqrtps( struct x86_function *p,
+                  struct x86_reg dst,
+                  struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x52);
+   emit_modrm( p, dst, src );
+}
+
+void sse_rsqrtss( struct x86_function *p,
+		  struct x86_reg dst,
+		  struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0xF3, X86_TWOB, 0x52);
+   emit_modrm( p, dst, src );
+
+}
+
+void sse_movhlps( struct x86_function *p,
+		  struct x86_reg dst,
+		  struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   assert(dst.mod == mod_REG && src.mod == mod_REG);
+   emit_2ub(p, X86_TWOB, 0x12);
+   emit_modrm( p, dst, src );
+}
+
+void sse_movlhps( struct x86_function *p,
+		  struct x86_reg dst,
+		  struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   assert(dst.mod == mod_REG && src.mod == mod_REG);
+   emit_2ub(p, X86_TWOB, 0x16);
+   emit_modrm( p, dst, src );
+}
+
+void sse_orps( struct x86_function *p,
+               struct x86_reg dst,
+               struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x56);
+   emit_modrm( p, dst, src );
+}
+
+void sse_xorps( struct x86_function *p,
+                struct x86_reg dst,
+                struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x57);
+   emit_modrm( p, dst, src );
+}
+
+void sse_cvtps2pi( struct x86_function *p,
+		   struct x86_reg dst,
+		   struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   assert(dst.file == file_MMX && 
+	  (src.file == file_XMM || src.mod != mod_REG));
+
+   p->need_emms = 1;
+
+   emit_2ub(p, X86_TWOB, 0x2d);
+   emit_modrm( p, dst, src );
+}
+
+void sse2_cvtdq2ps( struct x86_function *p,
+		   struct x86_reg dst,
+		   struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x5b);
+   emit_modrm( p, dst, src );
+}
+
+
+/* Shufps can also be used to implement a reduced swizzle when dest ==
+ * arg0.
+ */
+void sse_shufps( struct x86_function *p,
+		 struct x86_reg dst,
+		 struct x86_reg src,
+		 unsigned char shuf) 
+{
+   DUMP_RRI( dst, src, shuf );
+   emit_2ub(p, X86_TWOB, 0xC6);
+   emit_modrm(p, dst, src);
+   emit_1ub(p, shuf); 
+}
+
+void sse_unpckhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub( p, X86_TWOB, 0x15 );
+   emit_modrm( p, dst, src );
+}
+
+void sse_unpcklps( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub( p, X86_TWOB, 0x14 );
+   emit_modrm( p, dst, src );
+}
+
+void sse_cmpps( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src,
+		enum sse_cc cc) 
+{
+   DUMP_RRI( dst, src, cc );
+   emit_2ub(p, X86_TWOB, 0xC2);
+   emit_modrm(p, dst, src);
+   emit_1ub(p, cc); 
+}
+
+void sse_pmovmskb( struct x86_function *p,
+                   struct x86_reg dst,
+                   struct x86_reg src)
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0x66, X86_TWOB, 0xD7);
+   emit_modrm(p, dst, src);
+}
+
+/***********************************************************************
+ * SSE2 instructions
+ */
+
+/**
+ * Perform a reduced swizzle:
+ */
+void sse2_pshufd( struct x86_function *p,
+		  struct x86_reg dst,
+		  struct x86_reg src,
+		  unsigned char shuf) 
+{
+   DUMP_RRI( dst, src, shuf );
+   emit_3ub(p, 0x66, X86_TWOB, 0x70);
+   emit_modrm(p, dst, src);
+   emit_1ub(p, shuf); 
+}
+
+void sse2_cvttps2dq( struct x86_function *p,
+                     struct x86_reg dst,
+                     struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub( p, 0xF3, X86_TWOB, 0x5B );
+   emit_modrm( p, dst, src );
+}
+
+void sse2_cvtps2dq( struct x86_function *p,
+		    struct x86_reg dst,
+		    struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0x66, X86_TWOB, 0x5B);
+   emit_modrm( p, dst, src );
+}
+
+void sse2_packssdw( struct x86_function *p,
+		    struct x86_reg dst,
+		    struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0x66, X86_TWOB, 0x6B);
+   emit_modrm( p, dst, src );
+}
+
+void sse2_packsswb( struct x86_function *p,
+		    struct x86_reg dst,
+		    struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0x66, X86_TWOB, 0x63);
+   emit_modrm( p, dst, src );
+}
+
+void sse2_packuswb( struct x86_function *p,
+		    struct x86_reg dst,
+		    struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0x66, X86_TWOB, 0x67);
+   emit_modrm( p, dst, src );
+}
+
+void sse2_punpcklbw( struct x86_function *p,
+		    struct x86_reg dst,
+		    struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0x66, X86_TWOB, 0x60);
+   emit_modrm( p, dst, src );
+}
+
+
+void sse2_rcpps( struct x86_function *p,
+                 struct x86_reg dst,
+                 struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, X86_TWOB, 0x53);
+   emit_modrm( p, dst, src );
+}
+
+void sse2_rcpss( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_3ub(p, 0xF3, X86_TWOB, 0x53);
+   emit_modrm( p, dst, src );
+}
+
+void sse2_movd( struct x86_function *p,
+		struct x86_reg dst,
+		struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   emit_2ub(p, 0x66, X86_TWOB);
+   emit_op_modrm( p, 0x6e, 0x7e, dst, src );
+}
+
+
+
+
+/***********************************************************************
+ * x87 instructions
+ */
+static void note_x87_pop( struct x86_function *p )
+{
+   p->x87_stack--;
+   assert(p->x87_stack >= 0);
+}
+
+static void note_x87_push( struct x86_function *p )
+{
+   p->x87_stack++;
+   assert(p->x87_stack <= 7);
+}
+
+void x87_assert_stack_empty( struct x86_function *p )
+{
+   assert (p->x87_stack == 0);
+}
+
+
+void x87_fist( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   emit_1ub(p, 0xdb);
+   emit_modrm_noreg(p, 2, dst);
+}
+
+void x87_fistp( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   emit_1ub(p, 0xdb);
+   emit_modrm_noreg(p, 3, dst);
+   note_x87_pop(p);
+}
+
+void x87_fild( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   emit_1ub(p, 0xdf);
+   emit_modrm_noreg(p, 0, arg);
+   note_x87_push(p);
+}
+
+void x87_fldz( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xee);
+   note_x87_push(p);
+}
+
+
+void x87_fldcw( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   assert(arg.file == file_REG32);
+   assert(arg.mod != mod_REG);
+   emit_1ub(p, 0xd9);
+   emit_modrm_noreg(p, 5, arg);
+}
+
+void x87_fld1( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xe8);
+   note_x87_push(p);
+}
+
+void x87_fldl2e( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xea);
+   note_x87_push(p);
+}
+
+void x87_fldln2( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xed);
+   note_x87_push(p);
+}
+
+void x87_fwait( struct x86_function *p )
+{
+   DUMP();
+   emit_1ub(p, 0x9b);
+}
+
+void x87_fnclex( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xdb, 0xe2);
+}
+
+void x87_fclex( struct x86_function *p )
+{
+   x87_fwait(p);
+   x87_fnclex(p);
+}
+
+void x87_fcmovb( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   assert(arg.file == file_x87);
+   emit_2ub(p, 0xda, 0xc0+arg.idx);
+}
+
+void x87_fcmove( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   assert(arg.file == file_x87);
+   emit_2ub(p, 0xda, 0xc8+arg.idx);
+}
+
+void x87_fcmovbe( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   assert(arg.file == file_x87);
+   emit_2ub(p, 0xda, 0xd0+arg.idx);
+}
+
+void x87_fcmovnb( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   assert(arg.file == file_x87);
+   emit_2ub(p, 0xdb, 0xc0+arg.idx);
+}
+
+void x87_fcmovne( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   assert(arg.file == file_x87);
+   emit_2ub(p, 0xdb, 0xc8+arg.idx);
+}
+
+void x87_fcmovnbe( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   assert(arg.file == file_x87);
+   emit_2ub(p, 0xdb, 0xd0+arg.idx);
+}
+
+
+
+static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86_reg arg,
+			  unsigned char dst0ub0,
+			  unsigned char dst0ub1,
+			  unsigned char arg0ub0,
+			  unsigned char arg0ub1,
+			  unsigned char argmem_noreg)
+{
+   assert(dst.file == file_x87);
+
+   if (arg.file == file_x87) {
+      if (dst.idx == 0) 
+	 emit_2ub(p, dst0ub0, dst0ub1+arg.idx);
+      else if (arg.idx == 0) 
+	 emit_2ub(p, arg0ub0, arg0ub1+arg.idx);
+      else
+	 assert(0);
+   }
+   else if (dst.idx == 0) {
+      assert(arg.file == file_REG32);
+      emit_1ub(p, 0xd8);
+      emit_modrm_noreg(p, argmem_noreg, arg);
+   }
+   else
+      assert(0);
+}
+
+void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   x87_arith_op(p, dst, src, 
+		0xd8, 0xc8,
+		0xdc, 0xc8,
+		4);
+}
+
+void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   x87_arith_op(p, dst, src, 
+		0xd8, 0xe0,
+		0xdc, 0xe8,
+		4);
+}
+
+void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   x87_arith_op(p, dst, src, 
+		0xd8, 0xe8,
+		0xdc, 0xe0,
+		5);
+}
+
+void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   x87_arith_op(p, dst, src, 
+		0xd8, 0xc0,
+		0xdc, 0xc0,
+		0);
+}
+
+void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   x87_arith_op(p, dst, src, 
+		0xd8, 0xf0,
+		0xdc, 0xf8,
+		6);
+}
+
+void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   x87_arith_op(p, dst, src, 
+		0xd8, 0xf8,
+		0xdc, 0xf0,
+		7);
+}
+
+void x87_fmulp( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   assert(dst.file == file_x87);
+   assert(dst.idx >= 1);
+   emit_2ub(p, 0xde, 0xc8+dst.idx);
+   note_x87_pop(p);
+}
+
+void x87_fsubp( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   assert(dst.file == file_x87);
+   assert(dst.idx >= 1);
+   emit_2ub(p, 0xde, 0xe8+dst.idx);
+   note_x87_pop(p);
+}
+
+void x87_fsubrp( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   assert(dst.file == file_x87);
+   assert(dst.idx >= 1);
+   emit_2ub(p, 0xde, 0xe0+dst.idx);
+   note_x87_pop(p);
+}
+
+void x87_faddp( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   assert(dst.file == file_x87);
+   assert(dst.idx >= 1);
+   emit_2ub(p, 0xde, 0xc0+dst.idx);
+   note_x87_pop(p);
+}
+
+void x87_fdivp( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   assert(dst.file == file_x87);
+   assert(dst.idx >= 1);
+   emit_2ub(p, 0xde, 0xf8+dst.idx);
+   note_x87_pop(p);
+}
+
+void x87_fdivrp( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   assert(dst.file == file_x87);
+   assert(dst.idx >= 1);
+   emit_2ub(p, 0xde, 0xf0+dst.idx);
+   note_x87_pop(p);
+}
+
+void x87_ftst( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xe4);
+}
+
+void x87_fucom( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   assert(arg.file == file_x87);
+   emit_2ub(p, 0xdd, 0xe0+arg.idx);
+}
+
+void x87_fucomp( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   assert(arg.file == file_x87);
+   emit_2ub(p, 0xdd, 0xe8+arg.idx);
+   note_x87_pop(p);
+}
+
+void x87_fucompp( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xda, 0xe9);
+   note_x87_pop(p);             /* pop twice */
+   note_x87_pop(p);             /* pop twice */
+}
+
+void x87_fxch( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   assert(arg.file == file_x87);
+   emit_2ub(p, 0xd9, 0xc8+arg.idx);
+}
+
+void x87_fabs( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xe1);
+}
+
+void x87_fchs( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xe0);
+}
+
+void x87_fcos( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xff);
+}
+
+
+void x87_fprndint( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xfc);
+}
+
+void x87_fscale( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xfd);
+}
+
+void x87_fsin( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xfe);
+}
+
+void x87_fsincos( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xfb);
+}
+
+void x87_fsqrt( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xfa);
+}
+
+void x87_fxtract( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xf4);
+}
+
+/* st0 = (2^st0)-1
+ *
+ * Restrictions: -1.0 <= st0 <= 1.0
+ */
+void x87_f2xm1( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xf0);
+}
+
+/* st1 = st1 * log2(st0);
+ * pop_stack;
+ */
+void x87_fyl2x( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xf1);
+   note_x87_pop(p);
+}
+
+/* st1 = st1 * log2(st0 + 1.0);
+ * pop_stack;
+ *
+ * A fast operation, with restrictions: -.29 < st0 < .29 
+ */
+void x87_fyl2xp1( struct x86_function *p )
+{
+   DUMP();
+   emit_2ub(p, 0xd9, 0xf9);
+   note_x87_pop(p);
+}
+
+
+void x87_fld( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   if (arg.file == file_x87) 
+      emit_2ub(p, 0xd9, 0xc0 + arg.idx);
+   else {
+      emit_1ub(p, 0xd9);
+      emit_modrm_noreg(p, 0, arg);
+   }
+   note_x87_push(p);
+}
+
+void x87_fst( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   if (dst.file == file_x87) 
+      emit_2ub(p, 0xdd, 0xd0 + dst.idx);
+   else {
+      emit_1ub(p, 0xd9);
+      emit_modrm_noreg(p, 2, dst);
+   }
+}
+
+void x87_fstp( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   if (dst.file == file_x87) 
+      emit_2ub(p, 0xdd, 0xd8 + dst.idx);
+   else {
+      emit_1ub(p, 0xd9);
+      emit_modrm_noreg(p, 3, dst);
+   }
+   note_x87_pop(p);
+}
+
+void x87_fpop( struct x86_function *p )
+{
+   x87_fstp( p, x86_make_reg( file_x87, 0 ));
+}
+
+
+void x87_fcom( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   if (dst.file == file_x87) 
+      emit_2ub(p, 0xd8, 0xd0 + dst.idx);
+   else {
+      emit_1ub(p, 0xd8);
+      emit_modrm_noreg(p, 2, dst);
+   }
+}
+
+
+void x87_fcomp( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   if (dst.file == file_x87) 
+      emit_2ub(p, 0xd8, 0xd8 + dst.idx);
+   else {
+      emit_1ub(p, 0xd8);
+      emit_modrm_noreg(p, 3, dst);
+   }
+   note_x87_pop(p);
+}
+
+void x87_fcomi( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   emit_2ub(p, 0xdb, 0xf0+arg.idx);
+}
+
+void x87_fcomip( struct x86_function *p, struct x86_reg arg )
+{
+   DUMP_R( arg );
+   emit_2ub(p, 0xdb, 0xf0+arg.idx);
+   note_x87_pop(p);
+}
+
+
+void x87_fnstsw( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   assert(dst.file == file_REG32);
+
+   if (dst.idx == reg_AX &&
+       dst.mod == mod_REG) 
+      emit_2ub(p, 0xdf, 0xe0);
+   else {
+      emit_1ub(p, 0xdd);
+      emit_modrm_noreg(p, 7, dst);
+   }
+}
+
+
+void x87_fnstcw( struct x86_function *p, struct x86_reg dst )
+{
+   DUMP_R( dst );
+   assert(dst.file == file_REG32);
+
+   emit_1ub(p, 0x9b);           /* WAIT -- needed? */
+   emit_1ub(p, 0xd9);
+   emit_modrm_noreg(p, 7, dst);
+}
+
+
+
+
+/***********************************************************************
+ * MMX instructions
+ */
+
+void mmx_emms( struct x86_function *p )
+{
+   DUMP();
+   assert(p->need_emms);
+   emit_2ub(p, 0x0f, 0x77);
+   p->need_emms = 0;
+}
+
+void mmx_packssdw( struct x86_function *p,
+		   struct x86_reg dst,
+		   struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   assert(dst.file == file_MMX && 
+	  (src.file == file_MMX || src.mod != mod_REG));
+
+   p->need_emms = 1;
+
+   emit_2ub(p, X86_TWOB, 0x6b);
+   emit_modrm( p, dst, src );
+}
+
+void mmx_packuswb( struct x86_function *p,
+		   struct x86_reg dst,
+		   struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   assert(dst.file == file_MMX && 
+	  (src.file == file_MMX || src.mod != mod_REG));
+
+   p->need_emms = 1;
+
+   emit_2ub(p, X86_TWOB, 0x67);
+   emit_modrm( p, dst, src );
+}
+
+void mmx_movd( struct x86_function *p,
+	       struct x86_reg dst,
+	       struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   p->need_emms = 1;
+   emit_1ub(p, X86_TWOB);
+   emit_op_modrm( p, 0x6e, 0x7e, dst, src );
+}
+
+void mmx_movq( struct x86_function *p,
+	       struct x86_reg dst,
+	       struct x86_reg src )
+{
+   DUMP_RR( dst, src );
+   p->need_emms = 1;
+   emit_1ub(p, X86_TWOB);
+   emit_op_modrm( p, 0x6f, 0x7f, dst, src );
+}
+
+
+/***********************************************************************
+ * Helper functions
+ */
+
+
+void x86_cdecl_caller_push_regs( struct x86_function *p )
+{
+   x86_push(p, x86_make_reg(file_REG32, reg_AX));
+   x86_push(p, x86_make_reg(file_REG32, reg_CX));
+   x86_push(p, x86_make_reg(file_REG32, reg_DX));
+}
+
+void x86_cdecl_caller_pop_regs( struct x86_function *p )
+{
+   x86_pop(p, x86_make_reg(file_REG32, reg_DX));
+   x86_pop(p, x86_make_reg(file_REG32, reg_CX));
+   x86_pop(p, x86_make_reg(file_REG32, reg_AX));
+}
+
+
+/* Retreive a reference to one of the function arguments, taking into
+ * account any push/pop activity:
+ */
+struct x86_reg x86_fn_arg( struct x86_function *p,
+			   unsigned arg )
+{
+   return x86_make_disp(x86_make_reg(file_REG32, reg_SP), 
+			p->stack_offset + arg * 4);	/* ??? */
+}
+
+
+void x86_init_func( struct x86_function *p )
+{
+   p->size = 0;
+   p->store = NULL;
+   p->csr = p->store;
+   DUMP_START();
+}
+
+void x86_init_func_size( struct x86_function *p, unsigned code_size )
+{
+   p->size = code_size;
+   p->store = rtasm_exec_malloc(code_size);
+   if (p->store == NULL) {
+      p->store = p->error_overflow;
+   }
+   p->csr = p->store;
+   DUMP_START();
+}
+
+void x86_release_func( struct x86_function *p )
+{
+   if (p->store && p->store != p->error_overflow)
+      rtasm_exec_free(p->store);
+
+   p->store = NULL;
+   p->csr = NULL;
+   p->size = 0;
+}
+
+
+void (*x86_get_func( struct x86_function *p ))(void)
+{
+   DUMP_END();
+   if (DISASSEM && p->store)
+      debug_printf("disassemble %p %p\n", p->store, p->csr);
+
+   if (p->store == p->error_overflow)
+      return (void (*)(void)) NULL;
+   else
+      return (void (*)(void)) p->store;
+}
+
+#else
+
+void x86sse_dummy( void )
+{
+}
+
+#endif
diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
new file mode 100644
index 0000000000..1b5eaaca85
--- /dev/null
+++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h
@@ -0,0 +1,319 @@
+/**************************************************************************
+ *
+ * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef _RTASM_X86SSE_H_
+#define _RTASM_X86SSE_H_
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_ARCH_X86)
+
+/* It is up to the caller to ensure that instructions issued are
+ * suitable for the host cpu.  There are no checks made in this module
+ * for mmx/sse/sse2 support on the cpu.
+ */
+struct x86_reg {
+   unsigned file:3;
+   unsigned idx:3;
+   unsigned mod:2;		/* mod_REG if this is just a register */
+   int      disp:24;		/* only +/- 23bits of offset - should be enough... */
+};
+
+struct x86_function {
+   unsigned size;
+   unsigned char *store;
+   unsigned char *csr;
+
+   unsigned stack_offset:16;
+   unsigned need_emms:8;
+   int x87_stack:8;
+
+   unsigned char error_overflow[4];
+};
+
+enum x86_reg_file {
+   file_REG32,
+   file_MMX,
+   file_XMM,
+   file_x87
+};
+
+/* Values for mod field of modr/m byte
+ */
+enum x86_reg_mod {
+   mod_INDIRECT,
+   mod_DISP8,
+   mod_DISP32,
+   mod_REG
+};
+
+enum x86_reg_name {
+   reg_AX,
+   reg_CX,
+   reg_DX,
+   reg_BX,
+   reg_SP,
+   reg_BP,
+   reg_SI,
+   reg_DI
+};
+
+
+enum x86_cc {
+   cc_O,			/* overflow */
+   cc_NO,			/* not overflow */
+   cc_NAE,			/* not above or equal / carry */
+   cc_AE,			/* above or equal / not carry */
+   cc_E,			/* equal / zero */
+   cc_NE			/* not equal / not zero */
+};
+
+enum sse_cc {
+   cc_Equal,
+   cc_LessThan,
+   cc_LessThanEqual,
+   cc_Unordered,
+   cc_NotEqual,
+   cc_NotLessThan,
+   cc_NotLessThanEqual,
+   cc_Ordered
+};
+
+#define cc_Z  cc_E
+#define cc_NZ cc_NE
+
+/* Begin/end/retreive function creation:
+ */
+
+
+void x86_init_func( struct x86_function *p );
+void x86_init_func_size( struct x86_function *p, unsigned code_size );
+void x86_release_func( struct x86_function *p );
+void (*x86_get_func( struct x86_function *p ))( void );
+
+/* Debugging:
+ */
+void x86_print_reg( struct x86_reg reg );
+
+
+/* Create and manipulate registers and regmem values:
+ */
+struct x86_reg x86_make_reg( enum x86_reg_file file,
+			     enum x86_reg_name idx );
+
+struct x86_reg x86_make_disp( struct x86_reg reg,
+			      int disp );
+
+struct x86_reg x86_deref( struct x86_reg reg );
+
+struct x86_reg x86_get_base_reg( struct x86_reg reg );
+
+
+/* Labels, jumps and fixup:
+ */
+int x86_get_label( struct x86_function *p );
+
+void x86_jcc( struct x86_function *p,
+	      enum x86_cc cc,
+	      int label );
+
+int x86_jcc_forward( struct x86_function *p,
+			  enum x86_cc cc );
+
+int x86_jmp_forward( struct x86_function *p);
+
+int x86_call_forward( struct x86_function *p);
+
+void x86_fixup_fwd_jump( struct x86_function *p,
+			 int fixup );
+
+void x86_jmp( struct x86_function *p, int label );
+
+/* void x86_call( struct x86_function *p, void (*label)() ); */
+void x86_call( struct x86_function *p, struct x86_reg reg);
+
+void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm );
+void x86_add_imm( struct x86_function *p, struct x86_reg dst, int imm );
+void x86_or_imm( struct x86_function *p, struct x86_reg dst, int imm );
+void x86_and_imm( struct x86_function *p, struct x86_reg dst, int imm );
+void x86_sub_imm( struct x86_function *p, struct x86_reg dst, int imm );
+void x86_xor_imm( struct x86_function *p, struct x86_reg dst, int imm );
+void x86_cmp_imm( struct x86_function *p, struct x86_reg dst, int imm );
+
+
+/* Macro for sse_shufps() and sse2_pshufd():
+ */
+#define SHUF(_x,_y,_z,_w)       (((_x)<<0) | ((_y)<<2) | ((_z)<<4) | ((_w)<<6))
+#define SHUF_NOOP               RSW(0,1,2,3)
+#define GET_SHUF(swz, idx)      (((swz) >> ((idx)*2)) & 0x3)
+
+void mmx_emms( struct x86_function *p );
+void mmx_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void mmx_movq( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void mmx_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void mmx_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+
+void sse2_cvtps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse2_cvttps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse2_cvtdq2ps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0,
+                  unsigned char shuf );
+void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+
+
+void sse_prefetchnta( struct x86_function *p, struct x86_reg ptr);
+void sse_prefetch0( struct x86_function *p, struct x86_reg ptr);
+void sse_prefetch1( struct x86_function *p, struct x86_reg ptr);
+
+void sse_movntps( struct x86_function *p, struct x86_reg dst, struct x86_reg src);
+
+void sse_addps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_addss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src,
+                enum sse_cc cc );
+void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_movaps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_movhlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_movhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_movlhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_movlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_movss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_movups( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_mulps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_mulss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_orps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0,
+                 unsigned char shuf );
+void sse_unpckhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_unpcklps( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src );
+void sse2_punpcklbw( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+
+void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_cmp( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_dec( struct x86_function *p, struct x86_reg reg );
+void x86_inc( struct x86_function *p, struct x86_reg reg );
+void x86_lea( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_mov( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_mul( struct x86_function *p, struct x86_reg src );
+void x86_imul( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_or( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_pop( struct x86_function *p, struct x86_reg reg );
+void x86_push( struct x86_function *p, struct x86_reg reg );
+void x86_push_imm32( struct x86_function *p, int imm );
+void x86_ret( struct x86_function *p );
+void x86_retw( struct x86_function *p, unsigned short imm );
+void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_sahf( struct x86_function *p );
+
+
+void x86_cdecl_caller_push_regs( struct x86_function *p );
+void x86_cdecl_caller_pop_regs( struct x86_function *p );
+
+void x87_assert_stack_empty( struct x86_function *p );
+
+void x87_f2xm1( struct x86_function *p );
+void x87_fabs( struct x86_function *p );
+void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg arg );
+void x87_faddp( struct x86_function *p, struct x86_reg dst );
+void x87_fchs( struct x86_function *p );
+void x87_fclex( struct x86_function *p );
+void x87_fcmovb( struct x86_function *p, struct x86_reg src );
+void x87_fcmovbe( struct x86_function *p, struct x86_reg src );
+void x87_fcmove( struct x86_function *p, struct x86_reg src );
+void x87_fcmovnb( struct x86_function *p, struct x86_reg src );
+void x87_fcmovnbe( struct x86_function *p, struct x86_reg src );
+void x87_fcmovne( struct x86_function *p, struct x86_reg src );
+void x87_fcom( struct x86_function *p, struct x86_reg dst );
+void x87_fcomi( struct x86_function *p, struct x86_reg dst );
+void x87_fcomip( struct x86_function *p, struct x86_reg dst );
+void x87_fcomp( struct x86_function *p, struct x86_reg dst );
+void x87_fcos( struct x86_function *p );
+void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg arg );
+void x87_fdivp( struct x86_function *p, struct x86_reg dst );
+void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg );
+void x87_fdivrp( struct x86_function *p, struct x86_reg dst );
+void x87_fild( struct x86_function *p, struct x86_reg arg );
+void x87_fist( struct x86_function *p, struct x86_reg dst );
+void x87_fistp( struct x86_function *p, struct x86_reg dst );
+void x87_fld( struct x86_function *p, struct x86_reg arg );
+void x87_fld1( struct x86_function *p );
+void x87_fldcw( struct x86_function *p, struct x86_reg arg );
+void x87_fldl2e( struct x86_function *p );
+void x87_fldln2( struct x86_function *p );
+void x87_fldz( struct x86_function *p );
+void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg arg );
+void x87_fmulp( struct x86_function *p, struct x86_reg dst );
+void x87_fnclex( struct x86_function *p );
+void x87_fprndint( struct x86_function *p );
+void x87_fpop( struct x86_function *p );
+void x87_fscale( struct x86_function *p );
+void x87_fsin( struct x86_function *p );
+void x87_fsincos( struct x86_function *p );
+void x87_fsqrt( struct x86_function *p );
+void x87_fst( struct x86_function *p, struct x86_reg dst );
+void x87_fstp( struct x86_function *p, struct x86_reg dst );
+void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg arg );
+void x87_fsubp( struct x86_function *p, struct x86_reg dst );
+void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg );
+void x87_fsubrp( struct x86_function *p, struct x86_reg dst );
+void x87_ftst( struct x86_function *p );
+void x87_fxch( struct x86_function *p, struct x86_reg dst );
+void x87_fxtract( struct x86_function *p );
+void x87_fyl2x( struct x86_function *p );
+void x87_fyl2xp1( struct x86_function *p );
+void x87_fwait( struct x86_function *p );
+void x87_fnstcw( struct x86_function *p, struct x86_reg dst );
+void x87_fnstsw( struct x86_function *p, struct x86_reg dst );
+void x87_fucompp( struct x86_function *p );
+void x87_fucomp( struct x86_function *p, struct x86_reg arg );
+void x87_fucom( struct x86_function *p, struct x86_reg arg );
+
+
+
+/* Retreive a reference to one of the function arguments, taking into
+ * account any push/pop activity.  Note - doesn't track explict
+ * manipulation of ESP by other instructions.
+ */
+struct x86_reg x86_fn_arg( struct x86_function *p, unsigned arg );
+
+#endif
+#endif
diff --git a/src/gallium/auxiliary/sct/Makefile b/src/gallium/auxiliary/sct/Makefile
new file mode 100644
index 0000000000..516d1756cf
--- /dev/null
+++ b/src/gallium/auxiliary/sct/Makefile
@@ -0,0 +1,12 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = sct
+
+C_SOURCES = \
+	sct.c
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/sct/SConscript b/src/gallium/auxiliary/sct/SConscript
new file mode 100644
index 0000000000..76927d973f
--- /dev/null
+++ b/src/gallium/auxiliary/sct/SConscript
@@ -0,0 +1,9 @@
+Import('*')
+
+sct = env.ConvenienceLibrary(
+	target = 'sct',
+	source = [
+		'sct.c'
+	])
+
+auxiliaries.insert(0, sct)
diff --git a/src/gallium/auxiliary/sct/sct.c b/src/gallium/auxiliary/sct/sct.c
new file mode 100644
index 0000000000..49bb7ea92e
--- /dev/null
+++ b/src/gallium/auxiliary/sct/sct.c
@@ -0,0 +1,454 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "util/u_memory.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "sct.h"
+
+
+struct texture_list
+{
+   struct pipe_texture *texture;
+   struct texture_list *next;
+};
+
+
+
+#define MAX_SURFACES  ((PIPE_MAX_COLOR_BUFS) + 1)
+
+struct sct_context
+{
+   const struct pipe_context *context;
+
+   /** surfaces the context is drawing into */
+   struct pipe_surface *surfaces[MAX_SURFACES];
+
+   /** currently bound textures */
+   struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+
+   /** previously bound textures, used but not flushed */
+   struct texture_list *textures_used;
+
+   boolean needs_flush;
+
+   struct sct_context *next;
+};
+
+
+
+struct sct_surface
+{
+   const struct pipe_surface *surface;
+
+   /** list of contexts drawing to this surface */
+   struct sct_context_list *contexts;
+
+   struct sct_surface *next;
+};
+
+
+
+/**
+ * Find the surface_info for the given pipe_surface
+ */
+static struct sct_surface *
+find_surface_info(struct surface_context_tracker *sct,
+                  const struct pipe_surface *surface)
+{
+   struct sct_surface *si;
+   for (si = sct->surfaces; si; si = si->next)
+      if (si->surface == surface)
+         return si;
+   return NULL;
+}
+
+
+/**
+ * As above, but create new surface_info if surface is new.
+ */
+static struct sct_surface *
+find_create_surface_info(struct surface_context_tracker *sct,
+                         const struct pipe_surface *surface)
+{
+   struct sct_surface *si = find_surface_info(sct, surface);
+   if (si)
+      return si;
+
+   /* alloc new */
+   si = CALLOC_STRUCT(sct_surface);
+   if (si) {
+      si->surface = surface;
+
+      /* insert at head */
+      si->next = sct->surfaces;
+      sct->surfaces = si;
+   }
+
+   return si;
+}
+
+
+/**
+ * Find a context_info for the given context.
+ */
+static struct sct_context *
+find_context_info(struct surface_context_tracker *sct,
+                  const struct pipe_context *context)
+{
+   struct sct_context *ci;
+   for (ci = sct->contexts; ci; ci = ci->next)
+      if (ci->context == context)
+         return ci;
+   return NULL;
+}
+
+
+/**
+ * As above, but create new context_info if context is new.
+ */
+static struct sct_context *
+find_create_context_info(struct surface_context_tracker *sct,
+                         const struct pipe_context *context)
+{
+   struct sct_context *ci = find_context_info(sct, context);
+   if (ci)
+      return ci;
+
+   /* alloc new */
+   ci = CALLOC_STRUCT(sct_context);
+   if (ci) {
+      ci->context = context;
+
+      /* insert at head */
+      ci->next = sct->contexts;
+      sct->contexts = ci;
+   }
+
+   return ci;
+}
+
+
+/**
+ * Is the context already bound to the surface?
+ */
+static boolean
+find_surface_context(const struct sct_surface *si,
+                     const struct pipe_context *context)
+{
+   const struct sct_context_list *cl;
+   for (cl = si->contexts; cl; cl = cl->next) {
+      if (cl->context == context) {
+         return TRUE;
+      }
+   }
+   return FALSE;
+}
+
+
+/**
+ * Add a context to the list of contexts associated with a surface.
+ */
+static void
+add_context_to_surface(struct sct_surface *si,
+                       const struct pipe_context *context)
+{
+   struct sct_context_list *cl = CALLOC_STRUCT(sct_context_list);
+   if (cl) {
+      cl->context = context;
+      /* insert at head of list of contexts */
+      cl->next = si->contexts;
+      si->contexts = cl;
+   }
+}
+
+
+/**
+ * Remove a context from the list of contexts associated with a surface.
+ */
+static void
+remove_context_from_surface(struct sct_surface *si,
+                            const struct pipe_context *context)
+{
+   struct sct_context_list *prev = NULL, *curr, *next;
+
+   for (curr = si->contexts; curr; curr = next) {
+      if (curr->context == context) {
+         /* remove */
+         if (prev)
+            prev->next = curr->next;
+         else
+            si->contexts = curr->next;
+         next = curr->next;
+         FREE(curr);
+      }
+      else {
+         prev = curr;
+         next = curr->next;
+      }
+   }
+}
+
+
+/**
+ * Unbind context from surface.
+ */
+static void
+unbind_context_surface(struct surface_context_tracker *sct,
+                       struct pipe_context *context,
+                       struct pipe_surface *surface)
+{
+   struct sct_surface *si = find_surface_info(sct, surface);
+   if (si) {
+      remove_context_from_surface(si, context);
+   }
+}
+
+
+/**
+ * Bind context to a set of surfaces (color + Z).
+ * Like MakeCurrent().
+ */
+void
+sct_bind_surfaces(struct surface_context_tracker *sct,
+                  struct pipe_context *context,
+                  uint num_surf,
+                  struct pipe_surface **surfaces)
+{
+   struct sct_context *ci = find_create_context_info(sct, context);
+   uint i;
+
+   if (!ci) {
+      return; /* out of memory */
+   }
+
+   /* unbind currently bound surfaces */
+   for (i = 0; i < MAX_SURFACES; i++) {
+      if (ci->surfaces[i]) {
+         unbind_context_surface(sct, context, ci->surfaces[i]);
+      }
+   }
+
+   /* bind new surfaces */
+   for (i = 0; i < num_surf; i++) {
+      struct sct_surface *si = find_create_surface_info(sct, surfaces[i]);
+      if (!find_surface_context(si, context)) {
+         add_context_to_surface(si, context);
+      }
+   }
+}
+
+
+/**
+ * Return list of contexts bound to a surface.
+ */
+const struct sct_context_list *
+sct_get_surface_contexts(struct surface_context_tracker *sct,
+                         const struct pipe_surface *surface)
+{
+   const struct sct_surface *si = find_surface_info(sct, surface);
+   return si->contexts;
+}
+
+
+
+static boolean
+find_texture(const struct sct_context *ci,
+             const struct pipe_texture *texture)
+{
+   const struct texture_list *tl;
+
+   for (tl = ci->textures_used; tl; tl = tl->next) {
+      if (tl->texture == texture) {
+         return TRUE;
+      }
+   }
+   return FALSE;
+}
+
+
+/**
+ * Add the given texture to the context's list of used textures.
+ */
+static void
+add_texture_used(struct sct_context *ci,
+                 struct pipe_texture *texture)
+{
+   if (!find_texture(ci, texture)) {
+      /* add to list */
+      struct texture_list *tl = CALLOC_STRUCT(texture_list);
+      if (tl) {
+         pipe_texture_reference(&tl->texture, texture);
+         /* insert at head */
+         tl->next = ci->textures_used;
+         ci->textures_used = tl;
+      }
+   }
+}
+
+
+/**
+ * Bind a texture to a rendering context.
+ */
+void
+sct_bind_texture(struct surface_context_tracker *sct,
+                 struct pipe_context *context,
+                 uint unit,
+                 struct pipe_texture *tex)
+{
+   struct sct_context *ci = find_context_info(sct, context);
+
+   if (ci->textures[unit] != tex) {
+      /* put texture on the 'used' list */
+      add_texture_used(ci, tex);
+      /* bind new */
+      pipe_texture_reference(&ci->textures[unit], tex);
+   }
+}
+
+
+/**
+ * Check if the given texture has been used by the rendering context
+ * since the last call to sct_flush_textures().
+ */
+boolean
+sct_is_texture_used(struct surface_context_tracker *sct,
+                    const struct pipe_context *context,
+                    const struct pipe_texture *texture)
+{
+   const struct sct_context *ci = find_context_info(sct, context);
+   return find_texture(ci, texture);
+}
+
+
+/**
+ * To be called when the image contents of a texture are changed, such
+ * as for gl[Copy]TexSubImage().
+ * XXX this may not be needed
+ */
+void
+sct_update_texture(struct pipe_texture *tex)
+{
+
+}
+
+
+/**
+ * When a scene is flushed/rendered we can release the list of
+ * used textures.
+ */
+void
+sct_flush_textures(struct surface_context_tracker *sct,
+                   struct pipe_context *context)
+{
+   struct sct_context *ci = find_context_info(sct, context);
+   struct texture_list *tl, *next;
+   uint i;
+
+   for (tl = ci->textures_used; tl; tl = next) {
+      next = tl->next;
+      pipe_texture_release(&tl->texture);
+      FREE(tl);
+   }
+   ci->textures_used = NULL;
+
+   /* put the currently bound textures on the 'used' list */
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      add_texture_used(ci, ci->textures[i]);
+   }
+}
+
+
+
+void
+sct_destroy_context(struct surface_context_tracker *sct,
+                    struct pipe_context *context)
+{
+   /* XXX should we require an unbinding first? */
+   {
+      struct sct_surface *si;
+      for (si = sct->surfaces; si; si = si->next) {
+         remove_context_from_surface(si, context);
+      }
+   }
+
+   /* remove context from context_info list */
+   {
+      struct sct_context *ci, *next, *prev = NULL;
+      for (ci = sct->contexts; ci; ci = next) {
+         next = ci->next;
+         if (ci->context == context) {
+            if (prev)
+               prev->next = ci->next;
+            else
+               sct->contexts = ci->next;
+            FREE(ci);
+         }
+         else {
+            prev = ci;
+         }
+      }
+   }
+
+}
+
+
+void
+sct_destroy_surface(struct surface_context_tracker *sct,
+                    struct pipe_surface *surface)
+{
+   if (1) {
+      /* debug/sanity: no context should be bound to surface */
+      struct sct_context *ci;
+      uint i;
+      for (ci = sct->contexts; ci; ci = ci->next) {
+         for (i = 0; i < MAX_SURFACES; i++) {
+            assert(ci->surfaces[i] != surface);
+         }
+      }
+   }
+
+   /* remove surface from sct_surface list */
+   {
+      struct sct_surface *si, *next, *prev = NULL;
+      for (si = sct->surfaces; si; si = next) {
+         next = si->next;
+         if (si->surface == surface) {
+            /* unlink */
+            if (prev)
+               prev->next = si->next;
+            else
+               sct->surfaces = si->next;
+            FREE(si);
+         }
+         else {
+            prev = si;
+         }
+      }
+   }
+}
diff --git a/src/gallium/auxiliary/sct/sct.h b/src/gallium/auxiliary/sct/sct.h
new file mode 100644
index 0000000000..cf7c4d3bdf
--- /dev/null
+++ b/src/gallium/auxiliary/sct/sct.h
@@ -0,0 +1,123 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Surface/Context Tracking
+ *
+ * For some drivers, we need to monitor the binding between contexts and
+ * surfaces/textures.
+ * This code may evolve quite a bit...
+ */
+
+
+#ifndef SCT_H
+#define SCT_H
+
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+
+struct pipe_context;
+struct pipe_surface;
+
+struct sct_context;
+struct sct_surface;
+
+
+/**
+ * Per-device info, basically
+ */
+struct surface_context_tracker
+{
+   struct sct_context *contexts;
+   struct sct_surface *surfaces;
+};
+
+
+
+/**
+ * Simple linked list of contexts
+ */
+struct sct_context_list
+{
+   const struct pipe_context *context;
+   struct sct_context_list *next;
+};
+
+
+
+extern void
+sct_bind_surfaces(struct surface_context_tracker *sct,
+                  struct pipe_context *context,
+                  uint num_surf,
+                  struct pipe_surface **surfaces);
+
+
+extern void
+sct_bind_texture(struct surface_context_tracker *sct,
+                 struct pipe_context *context,
+                 uint unit,
+                 struct pipe_texture *texture);
+
+
+extern void
+sct_update_texture(struct pipe_texture *tex);
+
+
+extern boolean
+sct_is_texture_used(struct surface_context_tracker *sct,
+                    const struct pipe_context *context,
+                    const struct pipe_texture *texture);
+
+extern void
+sct_flush_textures(struct surface_context_tracker *sct,
+                   struct pipe_context *context);
+
+
+extern const struct sct_context_list *
+sct_get_surface_contexts(struct surface_context_tracker *sct,
+                         const struct pipe_surface *surf);
+
+
+extern void
+sct_destroy_context(struct surface_context_tracker *sct,
+                    struct pipe_context *context);
+
+
+extern void
+sct_destroy_surface(struct surface_context_tracker *sct,
+                    struct pipe_surface *surface);
+
+
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif /* SCT_H */
diff --git a/src/gallium/auxiliary/sct/usage.c b/src/gallium/auxiliary/sct/usage.c
new file mode 100644
index 0000000000..6227f19962
--- /dev/null
+++ b/src/gallium/auxiliary/sct/usage.c
@@ -0,0 +1,61 @@
+/* surface / context tracking */
+
+
+/*
+
+context A:
+  render to texture T
+
+context B:
+  texture from T
+
+-----------------------
+
+flush surface:
+  which contexts are bound to the surface?
+
+-----------------------
+
+glTexSubImage():
+  which contexts need to be flushed?
+
+ */
+
+
+/*
+
+in MakeCurrent():
+
+  call sct_bind_surfaces(context, list of surfaces) to update the
+  dependencies between context and surfaces
+
+
+in SurfaceFlush(), or whatever it is in D3D:
+
+  call sct_get_surface_contexts(surface) to get a list of contexts
+  which are currently bound to the surface.
+
+
+
+in BindTexture():
+
+  call sct_bind_texture(context, texture) to indicate that the texture
+  is used in the scene.
+
+
+in glTexSubImage() or RenderToTexture():
+
+  call sct_is_texture_used(context, texture) to determine if the texture
+  has been used in the scene, but the scene's not flushed.  If TRUE is
+  returned it means the scene has to be rendered/flushed before the contents
+  of the texture can be changed.
+
+
+in psb_scene_flush/terminate():
+
+  call sct_flush_textures(context) to tell the SCT that the textures which
+  were used in the scene can be released.
+
+
+
+*/
diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile
new file mode 100644
index 0000000000..d7df9490cf
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/Makefile
@@ -0,0 +1,24 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = tgsi
+
+C_SOURCES = \
+	tgsi_sanity.c \
+	tgsi_build.c \
+	tgsi_dump.c \
+	tgsi_exec.c \
+	tgsi_info.c \
+	tgsi_iterate.c \
+	tgsi_parse.c \
+	tgsi_ppc.c \
+	tgsi_scan.c \
+	tgsi_sse2.c \
+	tgsi_text.c \
+	tgsi_transform.c \
+	tgsi_util.c
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript
new file mode 100644
index 0000000000..8200cce42f
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/SConscript
@@ -0,0 +1,22 @@
+Import('*')
+
+tgsi = env.ConvenienceLibrary(
+	target = 'tgsi',
+	source = [
+		'tgsi_build.c',
+		'tgsi_dump.c',
+		'tgsi_dump_c.c',
+		'tgsi_exec.c',
+		'tgsi_info.c',
+		'tgsi_iterate.c',
+		'tgsi_parse.c',
+		'tgsi_sanity.c',
+		'tgsi_scan.c',
+		'tgsi_ppc.c',
+		'tgsi_sse2.c',
+		'tgsi_text.c',
+		'tgsi_transform.c',
+		'tgsi_util.c',
+	])
+
+auxiliaries.insert(0, tgsi)
diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c
new file mode 100644
index 0000000000..17886540cf
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_build.c
@@ -0,0 +1,1335 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_debug.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi_build.h"
+#include "tgsi_parse.h"
+
+/*
+ * version
+ */
+
+struct tgsi_version
+tgsi_build_version( void )
+{
+   struct tgsi_version  version;
+
+   version.MajorVersion = 1;
+   version.MinorVersion = 1;
+   version.Padding = 0;
+
+   return version;
+}
+
+/*
+ * header
+ */
+
+struct tgsi_header
+tgsi_build_header( void )
+{
+   struct tgsi_header header;
+
+   header.HeaderSize = 1;
+   header.BodySize = 0;
+
+   return header;
+}
+
+static void
+header_headersize_grow( struct tgsi_header *header )
+{
+   assert( header->HeaderSize < 0xFF );
+   assert( header->BodySize == 0 );
+
+   header->HeaderSize++;
+}
+
+static void
+header_bodysize_grow( struct tgsi_header *header )
+{
+   assert( header->BodySize < 0xFFFFFF );
+
+   header->BodySize++;
+}
+
+struct tgsi_processor
+tgsi_default_processor( void )
+{
+   struct tgsi_processor processor;
+
+   processor.Processor = TGSI_PROCESSOR_FRAGMENT;
+   processor.Padding = 0;
+
+   return processor;
+}
+
+struct tgsi_processor
+tgsi_build_processor(
+   unsigned type,
+   struct tgsi_header *header )
+{
+   struct tgsi_processor processor;
+
+   processor = tgsi_default_processor();
+   processor.Processor = type;
+
+   header_headersize_grow( header );
+
+   return processor;
+}
+
+/*
+ * declaration
+ */
+
+struct tgsi_declaration
+tgsi_default_declaration( void )
+{
+   struct tgsi_declaration declaration;
+
+   declaration.Type = TGSI_TOKEN_TYPE_DECLARATION;
+   declaration.NrTokens = 1;
+   declaration.File = TGSI_FILE_NULL;
+   declaration.UsageMask = TGSI_WRITEMASK_XYZW;
+   declaration.Interpolate = TGSI_INTERPOLATE_CONSTANT;
+   declaration.Semantic = 0;
+   declaration.Centroid = 0;
+   declaration.Invariant = 0;
+   declaration.Padding = 0;
+   declaration.Extended = 0;
+
+   return declaration;
+}
+
+struct tgsi_declaration
+tgsi_build_declaration(
+   unsigned file,
+   unsigned usage_mask,
+   unsigned interpolate,
+   unsigned semantic,
+   unsigned centroid,
+   unsigned invariant,
+   struct tgsi_header *header )
+{
+   struct tgsi_declaration declaration;
+
+   assert( file <= TGSI_FILE_IMMEDIATE );
+   assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE );
+
+   declaration = tgsi_default_declaration();
+   declaration.File = file;
+   declaration.UsageMask = usage_mask;
+   declaration.Interpolate = interpolate;
+   declaration.Semantic = semantic;
+   declaration.Centroid = centroid;
+   declaration.Invariant = invariant;
+
+   header_bodysize_grow( header );
+
+   return declaration;
+}
+
+static void
+declaration_grow(
+   struct tgsi_declaration *declaration,
+   struct tgsi_header *header )
+{
+   assert( declaration->NrTokens < 0xFF );
+
+   declaration->NrTokens++;
+
+   header_bodysize_grow( header );
+}
+
+struct tgsi_full_declaration
+tgsi_default_full_declaration( void )
+{
+   struct tgsi_full_declaration  full_declaration;
+
+   full_declaration.Declaration  = tgsi_default_declaration();
+   full_declaration.DeclarationRange = tgsi_default_declaration_range();
+   full_declaration.Semantic = tgsi_default_declaration_semantic();
+
+   return full_declaration;
+}
+
+unsigned
+tgsi_build_full_declaration(
+   const struct tgsi_full_declaration *full_decl,
+   struct tgsi_token *tokens,
+   struct tgsi_header *header,
+   unsigned maxsize )
+{
+   unsigned size = 0;
+   struct tgsi_declaration *declaration;
+   struct tgsi_declaration_range *dr;
+
+   if( maxsize <= size )
+     return 0;
+   declaration = (struct tgsi_declaration *) &tokens[size];
+   size++;
+
+   *declaration = tgsi_build_declaration(
+      full_decl->Declaration.File,
+      full_decl->Declaration.UsageMask,
+      full_decl->Declaration.Interpolate,
+      full_decl->Declaration.Semantic,
+      full_decl->Declaration.Centroid,
+      full_decl->Declaration.Invariant,
+      header );
+
+   if (maxsize <= size)
+      return 0;
+   dr = (struct tgsi_declaration_range *) &tokens[size];
+   size++;
+
+   *dr = tgsi_build_declaration_range(
+      full_decl->DeclarationRange.First,
+      full_decl->DeclarationRange.Last,
+      declaration,
+      header );
+
+   if( full_decl->Declaration.Semantic ) {
+      struct tgsi_declaration_semantic *ds;
+
+      if( maxsize <= size )
+         return  0;
+      ds = (struct tgsi_declaration_semantic *) &tokens[size];
+      size++;
+
+      *ds = tgsi_build_declaration_semantic(
+         full_decl->Semantic.SemanticName,
+         full_decl->Semantic.SemanticIndex,
+         declaration,
+         header );
+   }
+
+   return size;
+}
+
+struct tgsi_declaration_range
+tgsi_default_declaration_range( void )
+{
+   struct tgsi_declaration_range dr;
+
+   dr.First = 0;
+   dr.Last = 0;
+
+   return dr;
+}
+
+struct tgsi_declaration_range
+tgsi_build_declaration_range(
+   unsigned first,
+   unsigned last,
+   struct tgsi_declaration *declaration,
+   struct tgsi_header *header )
+{
+   struct tgsi_declaration_range declaration_range;
+
+   assert( last >= first );
+   assert( last <= 0xFFFF );
+
+   declaration_range = tgsi_default_declaration_range();
+   declaration_range.First = first;
+   declaration_range.Last = last;
+
+   declaration_grow( declaration, header );
+
+   return declaration_range;
+}
+
+struct tgsi_declaration_semantic
+tgsi_default_declaration_semantic( void )
+{
+   struct tgsi_declaration_semantic ds;
+
+   ds.SemanticName = TGSI_SEMANTIC_POSITION;
+   ds.SemanticIndex = 0;
+   ds.Padding = 0;
+
+   return ds;
+}
+
+struct tgsi_declaration_semantic
+tgsi_build_declaration_semantic(
+   unsigned semantic_name,
+   unsigned semantic_index,
+   struct tgsi_declaration *declaration,
+   struct tgsi_header *header )
+{
+   struct tgsi_declaration_semantic ds;
+
+   assert( semantic_name <= TGSI_SEMANTIC_COUNT );
+   assert( semantic_index <= 0xFFFF );
+
+   ds = tgsi_default_declaration_semantic();
+   ds.SemanticName = semantic_name;
+   ds.SemanticIndex = semantic_index;
+
+   declaration_grow( declaration, header );
+
+   return ds;
+}
+
+/*
+ * immediate
+ */
+
+struct tgsi_immediate
+tgsi_default_immediate( void )
+{
+   struct tgsi_immediate immediate;
+
+   immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE;
+   immediate.NrTokens = 1;
+   immediate.DataType = TGSI_IMM_FLOAT32;
+   immediate.Padding = 0;
+   immediate.Extended = 0;
+
+   return immediate;
+}
+
+struct tgsi_immediate
+tgsi_build_immediate(
+   struct tgsi_header *header )
+{
+   struct tgsi_immediate immediate;
+
+   immediate = tgsi_default_immediate();
+
+   header_bodysize_grow( header );
+
+   return immediate;
+}
+
+struct tgsi_full_immediate
+tgsi_default_full_immediate( void )
+{
+   struct tgsi_full_immediate fullimm;
+
+   fullimm.Immediate = tgsi_default_immediate();
+   fullimm.u.Pointer = (void *) 0;
+
+   return fullimm;
+}
+
+static void
+immediate_grow(
+   struct tgsi_immediate *immediate,
+   struct tgsi_header *header )
+{
+   assert( immediate->NrTokens < 0xFF );
+
+   immediate->NrTokens++;
+
+   header_bodysize_grow( header );
+}
+
+struct tgsi_immediate_float32
+tgsi_build_immediate_float32(
+   float value,
+   struct tgsi_immediate *immediate,
+   struct tgsi_header *header )
+{
+   struct tgsi_immediate_float32 immediate_float32;
+
+   immediate_float32.Float = value;
+
+   immediate_grow( immediate, header );
+
+   return immediate_float32;
+}
+
+unsigned
+tgsi_build_full_immediate(
+   const struct tgsi_full_immediate *full_imm,
+   struct tgsi_token *tokens,
+   struct tgsi_header *header,
+   unsigned maxsize )
+{
+   unsigned size = 0, i;
+   struct tgsi_immediate *immediate;
+
+   if( maxsize <= size )
+      return 0;
+   immediate = (struct tgsi_immediate *) &tokens[size];
+   size++;
+
+   *immediate = tgsi_build_immediate( header );
+
+   for( i = 0; i < full_imm->Immediate.NrTokens - 1; i++ ) {
+      struct tgsi_immediate_float32 *if32;
+
+      if( maxsize <= size )
+         return  0;
+      if32 = (struct tgsi_immediate_float32 *) &tokens[size];
+      size++;
+
+      *if32 = tgsi_build_immediate_float32(
+         full_imm->u.ImmediateFloat32[i].Float,
+         immediate,
+         header );
+   }
+
+   return size;
+}
+
+/*
+ * instruction
+ */
+
+struct tgsi_instruction
+tgsi_default_instruction( void )
+{
+   struct tgsi_instruction instruction;
+
+   instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION;
+   instruction.NrTokens = 1;
+   instruction.Opcode = TGSI_OPCODE_MOV;
+   instruction.Saturate = TGSI_SAT_NONE;
+   instruction.NumDstRegs = 1;
+   instruction.NumSrcRegs = 1;
+   instruction.Padding  = 0;
+   instruction.Extended = 0;
+
+   return instruction;
+}
+
+struct tgsi_instruction
+tgsi_build_instruction(
+   unsigned opcode,
+   unsigned saturate,
+   unsigned num_dst_regs,
+   unsigned num_src_regs,
+   struct tgsi_header *header )
+{
+   struct tgsi_instruction instruction;
+
+   assert (opcode <= TGSI_OPCODE_LAST);
+   assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE);
+   assert (num_dst_regs <= 3);
+   assert (num_src_regs <= 15);
+
+   instruction = tgsi_default_instruction();
+   instruction.Opcode = opcode;
+   instruction.Saturate = saturate;
+   instruction.NumDstRegs = num_dst_regs;
+   instruction.NumSrcRegs = num_src_regs;
+
+   header_bodysize_grow( header );
+
+   return instruction;
+}
+
+static void
+instruction_grow(
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   assert (instruction->NrTokens <   0xFF);
+
+   instruction->NrTokens++;
+
+   header_bodysize_grow( header );
+}
+
+struct tgsi_full_instruction
+tgsi_default_full_instruction( void )
+{
+   struct tgsi_full_instruction full_instruction;
+   unsigned i;
+
+   full_instruction.Instruction = tgsi_default_instruction();
+   full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv();
+   full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label();
+   full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture();
+   for( i = 0;  i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) {
+      full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register();
+   }
+   for( i = 0;  i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) {
+      full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register();
+   }
+
+   return full_instruction;
+}
+
+unsigned
+tgsi_build_full_instruction(
+   const struct tgsi_full_instruction *full_inst,
+   struct  tgsi_token *tokens,
+   struct  tgsi_header *header,
+   unsigned  maxsize )
+{
+   unsigned size = 0;
+   unsigned i;
+   struct tgsi_instruction *instruction;
+   struct tgsi_token *prev_token;
+
+   if( maxsize <= size )
+      return 0;
+   instruction = (struct tgsi_instruction *) &tokens[size];
+   size++;
+
+   *instruction = tgsi_build_instruction(
+      full_inst->Instruction.Opcode,
+      full_inst->Instruction.Saturate,
+      full_inst->Instruction.NumDstRegs,
+      full_inst->Instruction.NumSrcRegs,
+      header );
+   prev_token = (struct tgsi_token  *) instruction;
+
+   if( tgsi_compare_instruction_ext_nv(
+         full_inst->InstructionExtNv,
+         tgsi_default_instruction_ext_nv() ) ) {
+      struct tgsi_instruction_ext_nv *instruction_ext_nv;
+
+      if( maxsize <= size )
+         return 0;
+      instruction_ext_nv =
+         (struct  tgsi_instruction_ext_nv *) &tokens[size];
+      size++;
+
+      *instruction_ext_nv  = tgsi_build_instruction_ext_nv(
+         full_inst->InstructionExtNv.Precision,
+         full_inst->InstructionExtNv.CondDstIndex,
+         full_inst->InstructionExtNv.CondFlowIndex,
+         full_inst->InstructionExtNv.CondMask,
+         full_inst->InstructionExtNv.CondSwizzleX,
+         full_inst->InstructionExtNv.CondSwizzleY,
+         full_inst->InstructionExtNv.CondSwizzleZ,
+         full_inst->InstructionExtNv.CondSwizzleW,
+         full_inst->InstructionExtNv.CondDstUpdate,
+         full_inst->InstructionExtNv.CondFlowEnable,
+         prev_token,
+         instruction,
+         header );
+      prev_token = (struct tgsi_token  *) instruction_ext_nv;
+   }
+
+   if( tgsi_compare_instruction_ext_label(
+         full_inst->InstructionExtLabel,
+         tgsi_default_instruction_ext_label() ) ) {
+      struct tgsi_instruction_ext_label *instruction_ext_label;
+
+      if( maxsize <= size )
+         return 0;
+      instruction_ext_label =
+         (struct  tgsi_instruction_ext_label *) &tokens[size];
+      size++;
+
+      *instruction_ext_label = tgsi_build_instruction_ext_label(
+         full_inst->InstructionExtLabel.Label,
+         prev_token,
+         instruction,
+         header );
+      prev_token = (struct tgsi_token  *) instruction_ext_label;
+   }
+
+   if( tgsi_compare_instruction_ext_texture(
+         full_inst->InstructionExtTexture,
+         tgsi_default_instruction_ext_texture() ) ) {
+      struct tgsi_instruction_ext_texture *instruction_ext_texture;
+
+      if( maxsize <= size )
+         return 0;
+      instruction_ext_texture =
+         (struct  tgsi_instruction_ext_texture *) &tokens[size];
+      size++;
+
+      *instruction_ext_texture = tgsi_build_instruction_ext_texture(
+         full_inst->InstructionExtTexture.Texture,
+         prev_token,
+         instruction,
+         header   );
+      prev_token = (struct tgsi_token  *) instruction_ext_texture;
+   }
+
+   for( i = 0;  i <   full_inst->Instruction.NumDstRegs; i++ ) {
+      const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i];
+      struct tgsi_dst_register *dst_register;
+      struct tgsi_token *prev_token;
+
+      if( maxsize <= size )
+         return 0;
+      dst_register = (struct tgsi_dst_register *) &tokens[size];
+      size++;
+
+      *dst_register = tgsi_build_dst_register(
+         reg->DstRegister.File,
+         reg->DstRegister.WriteMask,
+         reg->DstRegister.Index,
+         instruction,
+         header );
+      prev_token = (struct tgsi_token  *) dst_register;
+
+      if( tgsi_compare_dst_register_ext_concode(
+            reg->DstRegisterExtConcode,
+            tgsi_default_dst_register_ext_concode() ) ) {
+         struct tgsi_dst_register_ext_concode *dst_register_ext_concode;
+
+         if( maxsize <= size )
+            return 0;
+         dst_register_ext_concode =
+            (struct  tgsi_dst_register_ext_concode *) &tokens[size];
+         size++;
+
+         *dst_register_ext_concode =   tgsi_build_dst_register_ext_concode(
+            reg->DstRegisterExtConcode.CondMask,
+            reg->DstRegisterExtConcode.CondSwizzleX,
+            reg->DstRegisterExtConcode.CondSwizzleY,
+            reg->DstRegisterExtConcode.CondSwizzleZ,
+            reg->DstRegisterExtConcode.CondSwizzleW,
+            reg->DstRegisterExtConcode.CondSrcIndex,
+            prev_token,
+            instruction,
+            header );
+         prev_token = (struct tgsi_token  *) dst_register_ext_concode;
+      }
+
+      if( tgsi_compare_dst_register_ext_modulate(
+            reg->DstRegisterExtModulate,
+            tgsi_default_dst_register_ext_modulate() ) ) {
+         struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate;
+
+         if( maxsize <= size )
+            return 0;
+         dst_register_ext_modulate =
+            (struct  tgsi_dst_register_ext_modulate *) &tokens[size];
+         size++;
+
+         *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate(
+            reg->DstRegisterExtModulate.Modulate,
+            prev_token,
+            instruction,
+            header );
+         prev_token = (struct tgsi_token  *) dst_register_ext_modulate;
+      }
+   }
+
+   for( i = 0;  i < full_inst->Instruction.NumSrcRegs; i++ ) {
+      const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i];
+      struct tgsi_src_register *src_register;
+      struct tgsi_token *prev_token;
+
+      if( maxsize <= size )
+         return 0;
+      src_register = (struct tgsi_src_register *)  &tokens[size];
+      size++;
+
+      *src_register = tgsi_build_src_register(
+         reg->SrcRegister.File,
+         reg->SrcRegister.SwizzleX,
+         reg->SrcRegister.SwizzleY,
+         reg->SrcRegister.SwizzleZ,
+         reg->SrcRegister.SwizzleW,
+         reg->SrcRegister.Negate,
+         reg->SrcRegister.Indirect,
+         reg->SrcRegister.Dimension,
+         reg->SrcRegister.Index,
+         instruction,
+         header );
+      prev_token = (struct tgsi_token  *) src_register;
+
+      if( tgsi_compare_src_register_ext_swz(
+            reg->SrcRegisterExtSwz,
+            tgsi_default_src_register_ext_swz() ) ) {
+         struct tgsi_src_register_ext_swz *src_register_ext_swz;
+
+         /* Use of the extended swizzle requires the simple swizzle to be identity.
+          */
+         assert( reg->SrcRegister.SwizzleX == TGSI_SWIZZLE_X );
+         assert( reg->SrcRegister.SwizzleY == TGSI_SWIZZLE_Y );
+         assert( reg->SrcRegister.SwizzleZ == TGSI_SWIZZLE_Z );
+         assert( reg->SrcRegister.SwizzleW == TGSI_SWIZZLE_W );
+         assert( reg->SrcRegister.Negate == FALSE );
+
+         if( maxsize <= size )
+            return 0;
+         src_register_ext_swz =
+            (struct  tgsi_src_register_ext_swz *) &tokens[size];
+         size++;
+
+         *src_register_ext_swz = tgsi_build_src_register_ext_swz(
+            reg->SrcRegisterExtSwz.ExtSwizzleX,
+            reg->SrcRegisterExtSwz.ExtSwizzleY,
+            reg->SrcRegisterExtSwz.ExtSwizzleZ,
+            reg->SrcRegisterExtSwz.ExtSwizzleW,
+            reg->SrcRegisterExtSwz.NegateX,
+            reg->SrcRegisterExtSwz.NegateY,
+            reg->SrcRegisterExtSwz.NegateZ,
+            reg->SrcRegisterExtSwz.NegateW,
+            prev_token,
+            instruction,
+            header );
+         prev_token = (struct tgsi_token  *) src_register_ext_swz;
+      }
+
+      if( tgsi_compare_src_register_ext_mod(
+            reg->SrcRegisterExtMod,
+            tgsi_default_src_register_ext_mod() ) ) {
+         struct tgsi_src_register_ext_mod *src_register_ext_mod;
+
+         if( maxsize <= size )
+            return 0;
+         src_register_ext_mod =
+            (struct  tgsi_src_register_ext_mod *) &tokens[size];
+         size++;
+
+         *src_register_ext_mod = tgsi_build_src_register_ext_mod(
+            reg->SrcRegisterExtMod.Complement,
+            reg->SrcRegisterExtMod.Bias,
+            reg->SrcRegisterExtMod.Scale2X,
+            reg->SrcRegisterExtMod.Absolute,
+            reg->SrcRegisterExtMod.Negate,
+            prev_token,
+            instruction,
+            header );
+         prev_token = (struct tgsi_token  *) src_register_ext_mod;
+      }
+
+      if( reg->SrcRegister.Indirect ) {
+         struct  tgsi_src_register *ind;
+
+         if( maxsize <= size )
+            return 0;
+         ind = (struct tgsi_src_register *) &tokens[size];
+         size++;
+
+         *ind = tgsi_build_src_register(
+            reg->SrcRegisterInd.File,
+            reg->SrcRegisterInd.SwizzleX,
+            reg->SrcRegisterInd.SwizzleY,
+            reg->SrcRegisterInd.SwizzleZ,
+            reg->SrcRegisterInd.SwizzleW,
+            reg->SrcRegisterInd.Negate,
+            reg->SrcRegisterInd.Indirect,
+            reg->SrcRegisterInd.Dimension,
+            reg->SrcRegisterInd.Index,
+            instruction,
+            header );
+      }
+
+      if( reg->SrcRegister.Dimension ) {
+         struct  tgsi_dimension *dim;
+
+         assert( !reg->SrcRegisterDim.Dimension );
+
+         if( maxsize <= size )
+            return 0;
+         dim = (struct tgsi_dimension *) &tokens[size];
+         size++;
+
+         *dim = tgsi_build_dimension(
+            reg->SrcRegisterDim.Indirect,
+            reg->SrcRegisterDim.Index,
+            instruction,
+            header );
+
+         if( reg->SrcRegisterDim.Indirect ) {
+            struct tgsi_src_register *ind;
+
+            if( maxsize <= size )
+               return 0;
+            ind = (struct tgsi_src_register *) &tokens[size];
+            size++;
+
+            *ind = tgsi_build_src_register(
+               reg->SrcRegisterDimInd.File,
+               reg->SrcRegisterDimInd.SwizzleX,
+               reg->SrcRegisterDimInd.SwizzleY,
+               reg->SrcRegisterDimInd.SwizzleZ,
+               reg->SrcRegisterDimInd.SwizzleW,
+               reg->SrcRegisterDimInd.Negate,
+               reg->SrcRegisterDimInd.Indirect,
+               reg->SrcRegisterDimInd.Dimension,
+               reg->SrcRegisterDimInd.Index,
+               instruction,
+               header );
+         }
+      }
+   }
+
+   return size;
+}
+
+struct tgsi_instruction_ext_nv
+tgsi_default_instruction_ext_nv( void )
+{
+   struct tgsi_instruction_ext_nv instruction_ext_nv;
+
+   instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV;
+   instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT;
+   instruction_ext_nv.CondDstIndex = 0;
+   instruction_ext_nv.CondFlowIndex = 0;
+   instruction_ext_nv.CondMask = TGSI_CC_TR;
+   instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X;
+   instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y;
+   instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z;
+   instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W;
+   instruction_ext_nv.CondDstUpdate = 0;
+   instruction_ext_nv.CondFlowEnable = 0;
+   instruction_ext_nv.Padding = 0;
+   instruction_ext_nv.Extended = 0;
+
+   return instruction_ext_nv;
+}
+
+
+/** test for inequality of 32-bit values pointed to by a and b */
+static INLINE boolean
+compare32(const void *a, const void *b)
+{
+   return *((uint32_t *) a) != *((uint32_t *) b);
+}
+
+
+unsigned
+tgsi_compare_instruction_ext_nv(
+   struct tgsi_instruction_ext_nv a,
+   struct tgsi_instruction_ext_nv b )
+{
+   a.Padding = b.Padding = 0;
+   a.Extended = b.Extended = 0;
+   return compare32(&a, &b);
+}
+
+struct tgsi_instruction_ext_nv
+tgsi_build_instruction_ext_nv(
+   unsigned precision,
+   unsigned cond_dst_index,
+   unsigned cond_flow_index,
+   unsigned cond_mask,
+   unsigned cond_swizzle_x,
+   unsigned cond_swizzle_y,
+   unsigned cond_swizzle_z,
+   unsigned cond_swizzle_w,
+   unsigned cond_dst_update,
+   unsigned cond_flow_enable,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_instruction_ext_nv instruction_ext_nv;
+
+   instruction_ext_nv = tgsi_default_instruction_ext_nv();
+   instruction_ext_nv.Precision = precision;
+   instruction_ext_nv.CondDstIndex = cond_dst_index;
+   instruction_ext_nv.CondFlowIndex = cond_flow_index;
+   instruction_ext_nv.CondMask = cond_mask;
+   instruction_ext_nv.CondSwizzleX = cond_swizzle_x;
+   instruction_ext_nv.CondSwizzleY = cond_swizzle_y;
+   instruction_ext_nv.CondSwizzleZ = cond_swizzle_z;
+   instruction_ext_nv.CondSwizzleW = cond_swizzle_w;
+   instruction_ext_nv.CondDstUpdate = cond_dst_update;
+   instruction_ext_nv.CondFlowEnable = cond_flow_enable;
+
+   prev_token->Extended = 1;
+   instruction_grow( instruction, header );
+
+   return instruction_ext_nv;
+}
+
+struct tgsi_instruction_ext_label
+tgsi_default_instruction_ext_label( void )
+{
+   struct tgsi_instruction_ext_label instruction_ext_label;
+
+   instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL;
+   instruction_ext_label.Label = 0;
+   instruction_ext_label.Padding = 0;
+   instruction_ext_label.Extended = 0;
+
+   return instruction_ext_label;
+}
+
+unsigned
+tgsi_compare_instruction_ext_label(
+   struct tgsi_instruction_ext_label a,
+   struct tgsi_instruction_ext_label b )
+{
+   a.Padding = b.Padding = 0;
+   a.Extended = b.Extended = 0;
+   return compare32(&a, &b);
+}
+
+struct tgsi_instruction_ext_label
+tgsi_build_instruction_ext_label(
+   unsigned label,
+   struct tgsi_token  *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_instruction_ext_label instruction_ext_label;
+
+   instruction_ext_label = tgsi_default_instruction_ext_label();
+   instruction_ext_label.Label = label;
+
+   prev_token->Extended = 1;
+   instruction_grow( instruction, header );
+
+   return instruction_ext_label;
+}
+
+struct tgsi_instruction_ext_texture
+tgsi_default_instruction_ext_texture( void )
+{
+   struct tgsi_instruction_ext_texture instruction_ext_texture;
+
+   instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE;
+   instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN;
+   instruction_ext_texture.Padding = 0;
+   instruction_ext_texture.Extended = 0;
+
+   return instruction_ext_texture;
+}
+
+unsigned
+tgsi_compare_instruction_ext_texture(
+   struct tgsi_instruction_ext_texture a,
+   struct tgsi_instruction_ext_texture b )
+{
+   a.Padding = b.Padding = 0;
+   a.Extended = b.Extended = 0;
+   return compare32(&a, &b);
+}
+
+struct tgsi_instruction_ext_texture
+tgsi_build_instruction_ext_texture(
+   unsigned texture,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_instruction_ext_texture instruction_ext_texture;
+
+   instruction_ext_texture = tgsi_default_instruction_ext_texture();
+   instruction_ext_texture.Texture = texture;
+
+   prev_token->Extended = 1;
+   instruction_grow( instruction, header );
+
+   return instruction_ext_texture;
+}
+
+struct tgsi_src_register
+tgsi_default_src_register( void )
+{
+   struct tgsi_src_register src_register;
+
+   src_register.File = TGSI_FILE_NULL;
+   src_register.SwizzleX = TGSI_SWIZZLE_X;
+   src_register.SwizzleY = TGSI_SWIZZLE_Y;
+   src_register.SwizzleZ = TGSI_SWIZZLE_Z;
+   src_register.SwizzleW = TGSI_SWIZZLE_W;
+   src_register.Negate = 0;
+   src_register.Indirect = 0;
+   src_register.Dimension = 0;
+   src_register.Index = 0;
+   src_register.Extended = 0;
+
+   return src_register;
+}
+
+struct tgsi_src_register
+tgsi_build_src_register(
+   unsigned file,
+   unsigned swizzle_x,
+   unsigned swizzle_y,
+   unsigned swizzle_z,
+   unsigned swizzle_w,
+   unsigned negate,
+   unsigned indirect,
+   unsigned dimension,
+   int index,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_src_register   src_register;
+
+   assert( file <= TGSI_FILE_IMMEDIATE );
+   assert( swizzle_x <= TGSI_SWIZZLE_W );
+   assert( swizzle_y <= TGSI_SWIZZLE_W );
+   assert( swizzle_z <= TGSI_SWIZZLE_W );
+   assert( swizzle_w <= TGSI_SWIZZLE_W );
+   assert( negate <= 1 );
+   assert( index >= -0x8000 && index <= 0x7FFF );
+
+   src_register = tgsi_default_src_register();
+   src_register.File = file;
+   src_register.SwizzleX = swizzle_x;
+   src_register.SwizzleY = swizzle_y;
+   src_register.SwizzleZ = swizzle_z;
+   src_register.SwizzleW = swizzle_w;
+   src_register.Negate = negate;
+   src_register.Indirect = indirect;
+   src_register.Dimension = dimension;
+   src_register.Index = index;
+
+   instruction_grow( instruction, header );
+
+   return src_register;
+}
+
+struct tgsi_full_src_register
+tgsi_default_full_src_register( void )
+{
+   struct tgsi_full_src_register full_src_register;
+
+   full_src_register.SrcRegister = tgsi_default_src_register();
+   full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz();
+   full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod();
+   full_src_register.SrcRegisterInd = tgsi_default_src_register();
+   full_src_register.SrcRegisterDim = tgsi_default_dimension();
+   full_src_register.SrcRegisterDimInd = tgsi_default_src_register();
+
+   return full_src_register;
+}
+
+struct tgsi_src_register_ext_swz
+tgsi_default_src_register_ext_swz( void )
+{
+   struct tgsi_src_register_ext_swz src_register_ext_swz;
+
+   src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ;
+   src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X;
+   src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y;
+   src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z;
+   src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W;
+   src_register_ext_swz.NegateX = 0;
+   src_register_ext_swz.NegateY = 0;
+   src_register_ext_swz.NegateZ = 0;
+   src_register_ext_swz.NegateW = 0;
+   src_register_ext_swz.Padding = 0;
+   src_register_ext_swz.Extended = 0;
+
+   return src_register_ext_swz;
+}
+
+unsigned
+tgsi_compare_src_register_ext_swz(
+   struct tgsi_src_register_ext_swz a,
+   struct tgsi_src_register_ext_swz b )
+{
+   a.Padding = b.Padding = 0;
+   a.Extended = b.Extended = 0;
+   return compare32(&a, &b);
+}
+
+struct tgsi_src_register_ext_swz
+tgsi_build_src_register_ext_swz(
+   unsigned ext_swizzle_x,
+   unsigned ext_swizzle_y,
+   unsigned ext_swizzle_z,
+   unsigned ext_swizzle_w,
+   unsigned negate_x,
+   unsigned negate_y,
+   unsigned negate_z,
+   unsigned negate_w,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_src_register_ext_swz src_register_ext_swz;
+
+   assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE );
+   assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE );
+   assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE );
+   assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE );
+   assert( negate_x <= 1 );
+   assert( negate_y <= 1 );
+   assert( negate_z <= 1 );
+   assert( negate_w <= 1 );
+
+   src_register_ext_swz = tgsi_default_src_register_ext_swz();
+   src_register_ext_swz.ExtSwizzleX = ext_swizzle_x;
+   src_register_ext_swz.ExtSwizzleY = ext_swizzle_y;
+   src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z;
+   src_register_ext_swz.ExtSwizzleW = ext_swizzle_w;
+   src_register_ext_swz.NegateX = negate_x;
+   src_register_ext_swz.NegateY = negate_y;
+   src_register_ext_swz.NegateZ = negate_z;
+   src_register_ext_swz.NegateW = negate_w;
+
+   prev_token->Extended = 1;
+   instruction_grow( instruction, header );
+
+   return src_register_ext_swz;
+}
+
+struct tgsi_src_register_ext_mod
+tgsi_default_src_register_ext_mod( void )
+{
+   struct tgsi_src_register_ext_mod src_register_ext_mod;
+
+   src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD;
+   src_register_ext_mod.Complement = 0;
+   src_register_ext_mod.Bias = 0;
+   src_register_ext_mod.Scale2X = 0;
+   src_register_ext_mod.Absolute = 0;
+   src_register_ext_mod.Negate = 0;
+   src_register_ext_mod.Padding = 0;
+   src_register_ext_mod.Extended = 0;
+
+   return src_register_ext_mod;
+}
+
+unsigned
+tgsi_compare_src_register_ext_mod(
+   struct tgsi_src_register_ext_mod a,
+   struct tgsi_src_register_ext_mod b )
+{
+   a.Padding = b.Padding = 0;
+   a.Extended = b.Extended = 0;
+   return compare32(&a, &b);
+}
+
+struct tgsi_src_register_ext_mod
+tgsi_build_src_register_ext_mod(
+   unsigned complement,
+   unsigned bias,
+   unsigned scale_2x,
+   unsigned absolute,
+   unsigned negate,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_src_register_ext_mod src_register_ext_mod;
+
+   assert( complement <= 1 );
+   assert( bias <= 1 );
+   assert( scale_2x <= 1 );
+   assert( absolute <= 1 );
+   assert( negate <= 1 );
+
+   src_register_ext_mod = tgsi_default_src_register_ext_mod();
+   src_register_ext_mod.Complement = complement;
+   src_register_ext_mod.Bias = bias;
+   src_register_ext_mod.Scale2X = scale_2x;
+   src_register_ext_mod.Absolute = absolute;
+   src_register_ext_mod.Negate = negate;
+
+   prev_token->Extended = 1;
+   instruction_grow( instruction, header );
+
+   return src_register_ext_mod;
+}
+
+struct tgsi_dimension
+tgsi_default_dimension( void )
+{
+   struct tgsi_dimension dimension;
+
+   dimension.Indirect = 0;
+   dimension.Dimension = 0;
+   dimension.Padding = 0;
+   dimension.Index = 0;
+   dimension.Extended = 0;
+
+   return dimension;
+}
+
+struct tgsi_dimension
+tgsi_build_dimension(
+   unsigned indirect,
+   unsigned index,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_dimension dimension;
+
+   dimension = tgsi_default_dimension();
+   dimension.Indirect = indirect;
+   dimension.Index = index;
+
+   instruction_grow( instruction, header );
+
+   return dimension;
+}
+
+struct tgsi_dst_register
+tgsi_default_dst_register( void )
+{
+   struct tgsi_dst_register dst_register;
+
+   dst_register.File = TGSI_FILE_NULL;
+   dst_register.WriteMask = TGSI_WRITEMASK_XYZW;
+   dst_register.Indirect = 0;
+   dst_register.Dimension = 0;
+   dst_register.Index = 0;
+   dst_register.Padding = 0;
+   dst_register.Extended = 0;
+
+   return dst_register;
+}
+
+struct tgsi_dst_register
+tgsi_build_dst_register(
+   unsigned file,
+   unsigned mask,
+   int index,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_dst_register dst_register;
+
+   assert( file <= TGSI_FILE_IMMEDIATE );
+   assert( mask <= TGSI_WRITEMASK_XYZW );
+   assert( index >= -32768 && index <= 32767 );
+
+   dst_register = tgsi_default_dst_register();
+   dst_register.File = file;
+   dst_register.WriteMask = mask;
+   dst_register.Index = index;
+
+   instruction_grow( instruction, header );
+
+   return dst_register;
+}
+
+struct tgsi_full_dst_register
+tgsi_default_full_dst_register( void )
+{
+   struct tgsi_full_dst_register full_dst_register;
+
+   full_dst_register.DstRegister = tgsi_default_dst_register();
+   full_dst_register.DstRegisterExtConcode =
+      tgsi_default_dst_register_ext_concode();
+   full_dst_register.DstRegisterExtModulate =
+      tgsi_default_dst_register_ext_modulate();
+
+   return full_dst_register;
+}
+
+struct tgsi_dst_register_ext_concode
+tgsi_default_dst_register_ext_concode( void )
+{
+   struct tgsi_dst_register_ext_concode dst_register_ext_concode;
+
+   dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE;
+   dst_register_ext_concode.CondMask = TGSI_CC_TR;
+   dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X;
+   dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y;
+   dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z;
+   dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W;
+   dst_register_ext_concode.CondSrcIndex = 0;
+   dst_register_ext_concode.Padding = 0;
+   dst_register_ext_concode.Extended = 0;
+
+   return dst_register_ext_concode;
+}
+
+unsigned
+tgsi_compare_dst_register_ext_concode(
+   struct tgsi_dst_register_ext_concode a,
+   struct tgsi_dst_register_ext_concode b )
+{
+   a.Padding = b.Padding = 0;
+   a.Extended = b.Extended = 0;
+   return compare32(&a, &b);
+}
+
+struct tgsi_dst_register_ext_concode
+tgsi_build_dst_register_ext_concode(
+   unsigned cc,
+   unsigned swizzle_x,
+   unsigned swizzle_y,
+   unsigned swizzle_z,
+   unsigned swizzle_w,
+   int index,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_dst_register_ext_concode dst_register_ext_concode;
+
+   assert( cc <= TGSI_CC_FL );
+   assert( swizzle_x <= TGSI_SWIZZLE_W );
+   assert( swizzle_y <= TGSI_SWIZZLE_W );
+   assert( swizzle_z <= TGSI_SWIZZLE_W );
+   assert( swizzle_w <= TGSI_SWIZZLE_W );
+   assert( index >= -32768 && index <= 32767 );
+
+   dst_register_ext_concode = tgsi_default_dst_register_ext_concode();
+   dst_register_ext_concode.CondMask = cc;
+   dst_register_ext_concode.CondSwizzleX = swizzle_x;
+   dst_register_ext_concode.CondSwizzleY = swizzle_y;
+   dst_register_ext_concode.CondSwizzleZ = swizzle_z;
+   dst_register_ext_concode.CondSwizzleW = swizzle_w;
+   dst_register_ext_concode.CondSrcIndex = index;
+
+   prev_token->Extended = 1;
+   instruction_grow( instruction, header );
+
+   return dst_register_ext_concode;
+}
+
+struct tgsi_dst_register_ext_modulate
+tgsi_default_dst_register_ext_modulate( void )
+{
+   struct tgsi_dst_register_ext_modulate dst_register_ext_modulate;
+
+   dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE;
+   dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X;
+   dst_register_ext_modulate.Padding = 0;
+   dst_register_ext_modulate.Extended = 0;
+
+   return dst_register_ext_modulate;
+}
+
+unsigned
+tgsi_compare_dst_register_ext_modulate(
+   struct tgsi_dst_register_ext_modulate a,
+   struct tgsi_dst_register_ext_modulate b )
+{
+   a.Padding = b.Padding = 0;
+   a.Extended = b.Extended = 0;
+   return compare32(&a, &b);
+}
+
+struct tgsi_dst_register_ext_modulate
+tgsi_build_dst_register_ext_modulate(
+   unsigned modulate,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header )
+{
+   struct tgsi_dst_register_ext_modulate dst_register_ext_modulate;
+
+   assert( modulate <= TGSI_MODULATE_EIGHTH );
+
+   dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate();
+   dst_register_ext_modulate.Modulate = modulate;
+
+   prev_token->Extended = 1;
+   instruction_grow( instruction, header );
+
+   return dst_register_ext_modulate;
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.h b/src/gallium/auxiliary/tgsi/tgsi_build.h
new file mode 100644
index 0000000000..0fd6fabd83
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_build.h
@@ -0,0 +1,338 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_BUILD_H
+#define TGSI_BUILD_H
+
+
+struct tgsi_token;
+
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+/*
+ * version
+ */
+
+struct tgsi_version
+tgsi_build_version( void );
+
+/*
+ * header
+ */
+
+struct tgsi_header
+tgsi_build_header( void );
+
+struct tgsi_processor
+tgsi_default_processor( void );
+
+struct tgsi_processor
+tgsi_build_processor(
+   unsigned processor,
+   struct tgsi_header *header );
+
+/*
+ * declaration
+ */
+
+struct tgsi_declaration
+tgsi_default_declaration( void );
+
+struct tgsi_declaration
+tgsi_build_declaration(
+   unsigned file,
+   unsigned usage_mask,
+   unsigned interpolate,
+   unsigned semantic,
+   unsigned centroid,
+   unsigned invariant,
+   struct tgsi_header *header );
+
+struct tgsi_full_declaration
+tgsi_default_full_declaration( void );
+
+unsigned
+tgsi_build_full_declaration(
+   const struct tgsi_full_declaration *full_decl,
+   struct tgsi_token *tokens,
+   struct tgsi_header *header,
+   unsigned maxsize );
+
+struct tgsi_declaration_range
+tgsi_default_declaration_range( void );
+
+struct tgsi_declaration_range
+tgsi_build_declaration_range(
+   unsigned first,
+   unsigned last,
+   struct tgsi_declaration *declaration,
+   struct tgsi_header *header );
+
+struct tgsi_declaration_semantic
+tgsi_default_declaration_semantic( void );
+
+struct tgsi_declaration_semantic
+tgsi_build_declaration_semantic(
+   unsigned semantic_name,
+   unsigned semantic_index,
+   struct tgsi_declaration *declaration,
+   struct tgsi_header *header );
+
+/*
+ * immediate
+ */
+
+struct tgsi_immediate
+tgsi_default_immediate( void );
+
+struct tgsi_immediate
+tgsi_build_immediate(
+   struct tgsi_header *header );
+
+struct tgsi_full_immediate
+tgsi_default_full_immediate( void );
+
+struct tgsi_immediate_float32
+tgsi_build_immediate_float32(
+   float value,
+   struct tgsi_immediate *immediate,
+   struct tgsi_header *header );
+
+unsigned
+tgsi_build_full_immediate(
+   const struct tgsi_full_immediate *full_imm,
+   struct tgsi_token *tokens,
+   struct tgsi_header *header,
+   unsigned maxsize );
+
+/*
+ * instruction
+ */
+
+struct tgsi_instruction
+tgsi_default_instruction( void );
+
+struct tgsi_instruction
+tgsi_build_instruction(
+   unsigned opcode,
+   unsigned saturate,
+   unsigned num_dst_regs,
+   unsigned num_src_regs,
+   struct tgsi_header *header );
+
+struct tgsi_full_instruction
+tgsi_default_full_instruction( void );
+
+unsigned
+tgsi_build_full_instruction(
+   const struct tgsi_full_instruction *full_inst,
+   struct tgsi_token *tokens,
+   struct tgsi_header *header,
+   unsigned maxsize );
+
+struct tgsi_instruction_ext_nv
+tgsi_default_instruction_ext_nv( void );
+
+unsigned
+tgsi_compare_instruction_ext_nv(
+   struct tgsi_instruction_ext_nv a,
+   struct tgsi_instruction_ext_nv b );
+
+struct tgsi_instruction_ext_nv
+tgsi_build_instruction_ext_nv(
+   unsigned precision,
+   unsigned cond_dst_index,
+   unsigned cond_flow_index,
+   unsigned cond_mask,
+   unsigned cond_swizzle_x,
+   unsigned cond_swizzle_y,
+   unsigned cond_swizzle_z,
+   unsigned cond_swizzle_w,
+   unsigned cond_dst_update,
+   unsigned cond_flow_enable,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header );
+
+struct tgsi_instruction_ext_label
+tgsi_default_instruction_ext_label( void );
+
+unsigned
+tgsi_compare_instruction_ext_label(
+   struct tgsi_instruction_ext_label a,
+   struct tgsi_instruction_ext_label b );
+
+struct tgsi_instruction_ext_label
+tgsi_build_instruction_ext_label(
+   unsigned label,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header );
+
+struct tgsi_instruction_ext_texture
+tgsi_default_instruction_ext_texture( void );
+
+unsigned
+tgsi_compare_instruction_ext_texture(
+   struct tgsi_instruction_ext_texture a,
+   struct tgsi_instruction_ext_texture b );
+
+struct tgsi_instruction_ext_texture
+tgsi_build_instruction_ext_texture(
+   unsigned texture,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header );
+
+struct tgsi_src_register
+tgsi_default_src_register( void );
+
+struct tgsi_src_register
+tgsi_build_src_register(
+   unsigned file,
+   unsigned swizzle_x,
+   unsigned swizzle_y,
+   unsigned swizzle_z,
+   unsigned swizzle_w,
+   unsigned negate,
+   unsigned indirect,
+   unsigned dimension,
+   int index,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header );
+
+struct tgsi_full_src_register
+tgsi_default_full_src_register( void );
+
+struct tgsi_src_register_ext_swz
+tgsi_default_src_register_ext_swz( void );
+
+unsigned
+tgsi_compare_src_register_ext_swz(
+   struct tgsi_src_register_ext_swz a,
+   struct tgsi_src_register_ext_swz b );
+
+struct tgsi_src_register_ext_swz
+tgsi_build_src_register_ext_swz(
+   unsigned ext_swizzle_x,
+   unsigned ext_swizzle_y,
+   unsigned ext_swizzle_z,
+   unsigned ext_swizzle_w,
+   unsigned negate_x,
+   unsigned negate_y,
+   unsigned negate_z,
+   unsigned negate_w,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header );
+
+struct tgsi_src_register_ext_mod
+tgsi_default_src_register_ext_mod( void );
+
+unsigned
+tgsi_compare_src_register_ext_mod(
+   struct tgsi_src_register_ext_mod a,
+   struct tgsi_src_register_ext_mod b );
+
+struct tgsi_src_register_ext_mod
+tgsi_build_src_register_ext_mod(
+   unsigned complement,
+   unsigned bias,
+   unsigned scale_2x,
+   unsigned absolute,
+   unsigned negate,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header );
+
+struct tgsi_dimension
+tgsi_default_dimension( void );
+
+struct tgsi_dimension
+tgsi_build_dimension(
+   unsigned indirect,
+   unsigned index,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header );
+
+struct tgsi_dst_register
+tgsi_default_dst_register( void );
+
+struct tgsi_dst_register
+tgsi_build_dst_register(
+   unsigned file,
+   unsigned mask,
+   int index,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header );
+
+struct tgsi_full_dst_register
+tgsi_default_full_dst_register( void );
+
+struct tgsi_dst_register_ext_concode
+tgsi_default_dst_register_ext_concode( void );
+
+unsigned
+tgsi_compare_dst_register_ext_concode(
+   struct tgsi_dst_register_ext_concode a,
+   struct tgsi_dst_register_ext_concode b );
+
+struct tgsi_dst_register_ext_concode
+tgsi_build_dst_register_ext_concode(
+   unsigned cc,
+   unsigned swizzle_x,
+   unsigned swizzle_y,
+   unsigned swizzle_z,
+   unsigned swizzle_w,
+   int index,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header );
+
+struct tgsi_dst_register_ext_modulate
+tgsi_default_dst_register_ext_modulate( void );
+
+unsigned
+tgsi_compare_dst_register_ext_modulate(
+   struct tgsi_dst_register_ext_modulate a,
+   struct tgsi_dst_register_ext_modulate b );
+
+struct tgsi_dst_register_ext_modulate
+tgsi_build_dst_register_ext_modulate(
+   unsigned modulate,
+   struct tgsi_token *prev_token,
+   struct tgsi_instruction *instruction,
+   struct tgsi_header *header );
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_BUILD_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c
new file mode 100644
index 0000000000..ab2b1f2c58
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c
@@ -0,0 +1,558 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_debug.h"
+#include "util/u_string.h"
+#include "tgsi_dump.h"
+#include "tgsi_info.h"
+#include "tgsi_iterate.h"
+
+struct dump_ctx
+{
+   struct tgsi_iterate_context iter;
+
+   uint instno;
+   
+   void (*printf)(struct dump_ctx *ctx, const char *format, ...);
+};
+
+static void 
+dump_ctx_printf(struct dump_ctx *ctx, const char *format, ...)
+{
+   va_list ap;
+   (void)ctx;
+   va_start(ap, format);
+   debug_vprintf(format, ap);
+   va_end(ap);
+}
+
+static void
+dump_enum(
+   struct dump_ctx *ctx,
+   uint e,
+   const char **enums,
+   uint enum_count )
+{
+   if (e >= enum_count)
+      ctx->printf( ctx, "%u", e );
+   else
+      ctx->printf( ctx, "%s", enums[e] );
+}
+
+#define EOL()           ctx->printf( ctx, "\n" )
+#define TXT(S)          ctx->printf( ctx, "%s", S )
+#define CHR(C)          ctx->printf( ctx, "%c", C )
+#define UIX(I)          ctx->printf( ctx, "0x%x", I )
+#define UID(I)          ctx->printf( ctx, "%u", I )
+#define INSTID(I)          ctx->printf( ctx, "% 3u", I )
+#define SID(I)          ctx->printf( ctx, "%d", I )
+#define FLT(F)          ctx->printf( ctx, "%10.4f", F )
+#define ENM(E,ENUMS)    dump_enum( ctx, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) )
+
+static const char *processor_type_names[] =
+{
+   "FRAG",
+   "VERT",
+   "GEOM"
+};
+
+static const char *file_names[] =
+{
+   "NULL",
+   "CONST",
+   "IN",
+   "OUT",
+   "TEMP",
+   "SAMP",
+   "ADDR",
+   "IMM"
+};
+
+static const char *interpolate_names[] =
+{
+   "CONSTANT",
+   "LINEAR",
+   "PERSPECTIVE"
+};
+
+static const char *semantic_names[] =
+{
+   "POSITION",
+   "COLOR",
+   "BCOLOR",
+   "FOG",
+   "PSIZE",
+   "GENERIC",
+   "NORMAL"
+};
+
+static const char *immediate_type_names[] =
+{
+   "FLT32"
+};
+
+static const char *swizzle_names[] =
+{
+   "x",
+   "y",
+   "z",
+   "w"
+};
+
+static const char *texture_names[] =
+{
+   "UNKNOWN",
+   "1D",
+   "2D",
+   "3D",
+   "CUBE",
+   "RECT",
+   "SHADOW1D",
+   "SHADOW2D",
+   "SHADOWRECT"
+};
+
+static const char *extswizzle_names[] =
+{
+   "x",
+   "y",
+   "z",
+   "w",
+   "0",
+   "1"
+};
+
+static const char *modulate_names[TGSI_MODULATE_COUNT] =
+{
+   "",
+   "_2X",
+   "_4X",
+   "_8X",
+   "_D2",
+   "_D4",
+   "_D8"
+};
+
+static void
+_dump_register(
+   struct dump_ctx *ctx,
+   uint file,
+   int first,
+   int last )
+{
+   ENM( file, file_names );
+   CHR( '[' );
+   SID( first );
+   if (first != last) {
+      TXT( ".." );
+      SID( last );
+   }
+   CHR( ']' );
+}
+
+static void
+_dump_register_ind(
+   struct dump_ctx *ctx,
+   uint file,
+   int index,
+   uint ind_file,
+   int ind_index,
+   uint ind_swizzle )
+{
+   ENM( file, file_names );
+   CHR( '[' );
+   ENM( ind_file, file_names );
+   CHR( '[' );
+   SID( ind_index );
+   TXT( "]." );
+   ENM( ind_swizzle, swizzle_names );
+   if (index != 0) {
+      if (index > 0)
+         CHR( '+' );
+      SID( index );
+   }
+   CHR( ']' );
+}
+
+static void
+_dump_writemask(
+   struct dump_ctx *ctx,
+   uint writemask )
+{
+   if (writemask != TGSI_WRITEMASK_XYZW) {
+      CHR( '.' );
+      if (writemask & TGSI_WRITEMASK_X)
+         CHR( 'x' );
+      if (writemask & TGSI_WRITEMASK_Y)
+         CHR( 'y' );
+      if (writemask & TGSI_WRITEMASK_Z)
+         CHR( 'z' );
+      if (writemask & TGSI_WRITEMASK_W)
+         CHR( 'w' );
+   }
+}
+
+static boolean
+iter_declaration(
+   struct tgsi_iterate_context *iter,
+   struct tgsi_full_declaration *decl )
+{
+   struct dump_ctx *ctx = (struct dump_ctx *)iter;
+
+   TXT( "DCL " );
+
+   _dump_register(
+      ctx,
+      decl->Declaration.File,
+      decl->DeclarationRange.First,
+      decl->DeclarationRange.Last );
+   _dump_writemask(
+      ctx,
+      decl->Declaration.UsageMask );
+
+   if (decl->Declaration.Semantic) {
+      TXT( ", " );
+      ENM( decl->Semantic.SemanticName, semantic_names );
+      if (decl->Semantic.SemanticIndex != 0 ||
+          decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC) {
+         CHR( '[' );
+         UID( decl->Semantic.SemanticIndex );
+         CHR( ']' );
+      }
+   }
+
+   TXT( ", " );
+   ENM( decl->Declaration.Interpolate, interpolate_names );
+
+   if (decl->Declaration.Centroid) {
+      TXT( ", CENTROID" );
+   }
+
+   if (decl->Declaration.Invariant) {
+      TXT( ", INVARIANT" );
+   }
+
+   EOL();
+
+   return TRUE;
+}
+
+void
+tgsi_dump_declaration(
+   const struct tgsi_full_declaration *decl )
+{
+   struct dump_ctx ctx;
+
+   ctx.printf = dump_ctx_printf;
+
+   iter_declaration( &ctx.iter, (struct tgsi_full_declaration *)decl );
+}
+
+static boolean
+iter_immediate(
+   struct tgsi_iterate_context *iter,
+   struct tgsi_full_immediate *imm )
+{
+   struct dump_ctx *ctx = (struct dump_ctx *) iter;
+
+   uint i;
+
+   TXT( "IMM " );
+   ENM( imm->Immediate.DataType, immediate_type_names );
+
+   TXT( " { " );
+   for (i = 0; i < imm->Immediate.NrTokens - 1; i++) {
+      switch (imm->Immediate.DataType) {
+      case TGSI_IMM_FLOAT32:
+         FLT( imm->u.ImmediateFloat32[i].Float );
+         break;
+      default:
+         assert( 0 );
+      }
+
+      if (i < imm->Immediate.NrTokens - 2)
+         TXT( ", " );
+   }
+   TXT( " }" );
+
+   EOL();
+
+   return TRUE;
+}
+
+void
+tgsi_dump_immediate(
+   const struct tgsi_full_immediate *imm )
+{
+   struct dump_ctx ctx;
+
+   ctx.printf = dump_ctx_printf;
+
+   iter_immediate( &ctx.iter, (struct tgsi_full_immediate *)imm );
+}
+
+static boolean
+iter_instruction(
+   struct tgsi_iterate_context *iter,
+   struct tgsi_full_instruction *inst )
+{
+   struct dump_ctx *ctx = (struct dump_ctx *) iter;
+   uint instno = ctx->instno++;
+   
+   uint i;
+   boolean first_reg = TRUE;
+
+   INSTID( instno );
+   TXT( ": " );
+   TXT( tgsi_get_opcode_info( inst->Instruction.Opcode )->mnemonic );
+
+   switch (inst->Instruction.Saturate) {
+   case TGSI_SAT_NONE:
+      break;
+   case TGSI_SAT_ZERO_ONE:
+      TXT( "_SAT" );
+      break;
+   case TGSI_SAT_MINUS_PLUS_ONE:
+      TXT( "_SATNV" );
+      break;
+   default:
+      assert( 0 );
+   }
+
+   for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
+      const struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
+
+      if (!first_reg)
+         CHR( ',' );
+      CHR( ' ' );
+
+      _dump_register(
+         ctx,
+         dst->DstRegister.File,
+         dst->DstRegister.Index,
+         dst->DstRegister.Index );
+      ENM( dst->DstRegisterExtModulate.Modulate, modulate_names );
+      _dump_writemask( ctx, dst->DstRegister.WriteMask );
+
+      first_reg = FALSE;
+   }
+
+   for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
+      const struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i];
+
+      if (!first_reg)
+         CHR( ',' );
+      CHR( ' ' );
+
+      if (src->SrcRegisterExtMod.Negate)
+         TXT( "-(" );
+      if (src->SrcRegisterExtMod.Absolute)
+         CHR( '|' );
+      if (src->SrcRegisterExtMod.Scale2X)
+         TXT( "2*(" );
+      if (src->SrcRegisterExtMod.Bias)
+         CHR( '(' );
+      if (src->SrcRegisterExtMod.Complement)
+         TXT( "1-(" );
+      if (src->SrcRegister.Negate)
+         CHR( '-' );
+
+      if (src->SrcRegister.Indirect) {
+         _dump_register_ind(
+            ctx,
+            src->SrcRegister.File,
+            src->SrcRegister.Index,
+            src->SrcRegisterInd.File,
+            src->SrcRegisterInd.Index,
+            src->SrcRegisterInd.SwizzleX );
+      }
+      else {
+         _dump_register(
+            ctx,
+            src->SrcRegister.File,
+            src->SrcRegister.Index,
+            src->SrcRegister.Index );
+      }
+
+      if (src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X ||
+          src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y ||
+          src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z ||
+          src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W) {
+         CHR( '.' );
+         ENM( src->SrcRegister.SwizzleX, swizzle_names );
+         ENM( src->SrcRegister.SwizzleY, swizzle_names );
+         ENM( src->SrcRegister.SwizzleZ, swizzle_names );
+         ENM( src->SrcRegister.SwizzleW, swizzle_names );
+      }
+      if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X ||
+          src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y ||
+          src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z ||
+          src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) {
+         CHR( '.' );
+         if (src->SrcRegisterExtSwz.NegateX)
+            TXT("-");
+         ENM( src->SrcRegisterExtSwz.ExtSwizzleX, extswizzle_names );
+         if (src->SrcRegisterExtSwz.NegateY)
+            TXT("-");
+         ENM( src->SrcRegisterExtSwz.ExtSwizzleY, extswizzle_names );
+         if (src->SrcRegisterExtSwz.NegateZ)
+            TXT("-");
+         ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, extswizzle_names );
+         if (src->SrcRegisterExtSwz.NegateW)
+            TXT("-");
+         ENM( src->SrcRegisterExtSwz.ExtSwizzleW, extswizzle_names );
+      }
+
+      if (src->SrcRegisterExtMod.Complement)
+         CHR( ')' );
+      if (src->SrcRegisterExtMod.Bias)
+         TXT( ")-.5" );
+      if (src->SrcRegisterExtMod.Scale2X)
+         CHR( ')' );
+      if (src->SrcRegisterExtMod.Absolute)
+         CHR( '|' );
+      if (src->SrcRegisterExtMod.Negate)
+         CHR( ')' );
+
+      first_reg = FALSE;
+   }
+
+   if (inst->InstructionExtTexture.Texture != TGSI_TEXTURE_UNKNOWN) {
+      TXT( ", " );
+      ENM( inst->InstructionExtTexture.Texture, texture_names );
+   }
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_IF:
+   case TGSI_OPCODE_ELSE:
+   case TGSI_OPCODE_BGNLOOP2:
+   case TGSI_OPCODE_ENDLOOP2:
+   case TGSI_OPCODE_CAL:
+      TXT( " :" );
+      UID( inst->InstructionExtLabel.Label );
+      break;
+   }
+
+   EOL();
+
+   return TRUE;
+}
+
+void
+tgsi_dump_instruction(
+   const struct tgsi_full_instruction *inst,
+   uint instno )
+{
+   struct dump_ctx ctx;
+
+   ctx.instno = instno;
+   ctx.printf = dump_ctx_printf;
+
+   iter_instruction( &ctx.iter, (struct tgsi_full_instruction *)inst );
+}
+
+static boolean
+prolog(
+   struct tgsi_iterate_context *iter )
+{
+   struct dump_ctx *ctx = (struct dump_ctx *) iter;
+   ENM( iter->processor.Processor, processor_type_names );
+   UID( iter->version.MajorVersion );
+   CHR( '.' );
+   UID( iter->version.MinorVersion );
+   EOL();
+   return TRUE;
+}
+
+void
+tgsi_dump(
+   const struct tgsi_token *tokens,
+   uint flags )
+{
+   struct dump_ctx ctx;
+
+   ctx.iter.prolog = prolog;
+   ctx.iter.iterate_instruction = iter_instruction;
+   ctx.iter.iterate_declaration = iter_declaration;
+   ctx.iter.iterate_immediate = iter_immediate;
+   ctx.iter.epilog = NULL;
+
+   ctx.instno = 0;
+   ctx.printf = dump_ctx_printf;
+
+   tgsi_iterate_shader( tokens, &ctx.iter );
+}
+
+struct str_dump_ctx
+{
+   struct dump_ctx base;
+   char *str;
+   char *ptr;
+   size_t left;
+};
+
+static void
+str_dump_ctx_printf(struct dump_ctx *ctx, const char *format, ...)
+{
+   struct str_dump_ctx *sctx = (struct str_dump_ctx *)ctx;
+   
+   if(sctx->left > 1) {
+      size_t written;
+      va_list ap;
+      va_start(ap, format);
+      written = util_vsnprintf(sctx->ptr, sctx->left, format, ap);
+      va_end(ap);
+      sctx->ptr += written;
+      sctx->left -= written;
+   }
+}
+
+void
+tgsi_dump_str(
+   const struct tgsi_token *tokens,
+   uint flags,
+   char *str,
+   size_t size)
+{
+   struct str_dump_ctx ctx;
+
+   ctx.base.iter.prolog = prolog;
+   ctx.base.iter.iterate_instruction = iter_instruction;
+   ctx.base.iter.iterate_declaration = iter_declaration;
+   ctx.base.iter.iterate_immediate = iter_immediate;
+   ctx.base.iter.epilog = NULL;
+
+   ctx.base.instno = 0;
+   ctx.base.printf = &str_dump_ctx_printf;
+
+   ctx.str = str;
+   ctx.str[0] = 0;
+   ctx.ptr = str;
+   ctx.left = size;
+
+   tgsi_iterate_shader( tokens, &ctx.base.iter );
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.h b/src/gallium/auxiliary/tgsi/tgsi_dump.h
new file mode 100644
index 0000000000..ad1e647ec9
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.h
@@ -0,0 +1,70 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_DUMP_H
+#define TGSI_DUMP_H
+
+#include "pipe/p_shader_tokens.h"
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+void
+tgsi_dump_str(
+   const struct tgsi_token *tokens,
+   uint flags,
+   char *str,
+   size_t size);
+
+void
+tgsi_dump(
+   const struct tgsi_token *tokens,
+   uint flags );
+
+struct tgsi_full_immediate;
+struct tgsi_full_instruction;
+struct tgsi_full_declaration;
+
+void
+tgsi_dump_immediate(
+   const struct tgsi_full_immediate *imm );
+
+void
+tgsi_dump_instruction(
+   const struct tgsi_full_instruction *inst,
+   uint instno );
+
+void
+tgsi_dump_declaration(
+   const struct tgsi_full_declaration *decl );
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_DUMP_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c
new file mode 100644
index 0000000000..2ecf1e2f14
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c
@@ -0,0 +1,718 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_debug.h"
+#include "util/u_string.h"
+#include "tgsi_dump_c.h"
+#include "tgsi_build.h"
+#include "tgsi_info.h"
+#include "tgsi_parse.h"
+
+static void
+dump_enum(
+   const unsigned    e,
+   const char        **enums,
+   const unsigned    enums_count )
+{
+   if (e >= enums_count) {
+      debug_printf( "%u", e );
+   }
+   else {
+      debug_printf( "%s", enums[e] );
+   }
+}
+
+#define EOL()           debug_printf( "\n" )
+#define TXT(S)          debug_printf( "%s", S )
+#define CHR(C)          debug_printf( "%c", C )
+#define UIX(I)          debug_printf( "0x%x", I )
+#define UID(I)          debug_printf( "%u", I )
+#define SID(I)          debug_printf( "%d", I )
+#define FLT(F)          debug_printf( "%10.4f", F )
+#define ENM(E,ENUMS)    dump_enum( E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) )
+
+static const char *TGSI_PROCESSOR_TYPES[] =
+{
+   "PROCESSOR_FRAGMENT",
+   "PROCESSOR_VERTEX",
+   "PROCESSOR_GEOMETRY"
+};
+
+static const char *TGSI_TOKEN_TYPES[] =
+{
+   "TOKEN_TYPE_DECLARATION",
+   "TOKEN_TYPE_IMMEDIATE",
+   "TOKEN_TYPE_INSTRUCTION"
+};
+
+static const char *TGSI_FILES[] =
+{
+   "FILE_NULL",
+   "FILE_CONSTANT",
+   "FILE_INPUT",
+   "FILE_OUTPUT",
+   "FILE_TEMPORARY",
+   "FILE_SAMPLER",
+   "FILE_ADDRESS",
+   "FILE_IMMEDIATE"
+};
+
+static const char *TGSI_INTERPOLATES[] =
+{
+   "INTERPOLATE_CONSTANT",
+   "INTERPOLATE_LINEAR",
+   "INTERPOLATE_PERSPECTIVE"
+};
+
+static const char *TGSI_SEMANTICS[] =
+{
+   "SEMANTIC_POSITION",
+   "SEMANTIC_COLOR",
+   "SEMANTIC_BCOLOR",
+   "SEMANTIC_FOG",
+   "SEMANTIC_PSIZE",
+   "SEMANTIC_GENERIC",
+   "SEMANTIC_NORMAL"
+};
+
+static const char *TGSI_IMMS[] =
+{
+   "IMM_FLOAT32"
+};
+
+static const char *TGSI_SATS[] =
+{
+   "SAT_NONE",
+   "SAT_ZERO_ONE",
+   "SAT_MINUS_PLUS_ONE"
+};
+
+static const char *TGSI_INSTRUCTION_EXTS[] =
+{
+   "INSTRUCTION_EXT_TYPE_NV",
+   "INSTRUCTION_EXT_TYPE_LABEL",
+   "INSTRUCTION_EXT_TYPE_TEXTURE"
+};
+
+static const char *TGSI_PRECISIONS[] =
+{
+   "PRECISION_DEFAULT",
+   "PRECISION_FLOAT32",
+   "PRECISION_FLOAT16",
+   "PRECISION_FIXED12"
+};
+
+static const char *TGSI_CCS[] =
+{
+   "CC_GT",
+   "CC_EQ",
+   "CC_LT",
+   "CC_UN",
+   "CC_GE",
+   "CC_LE",
+   "CC_NE",
+   "CC_TR",
+   "CC_FL"
+};
+
+static const char *TGSI_SWIZZLES[] =
+{
+   "SWIZZLE_X",
+   "SWIZZLE_Y",
+   "SWIZZLE_Z",
+   "SWIZZLE_W"
+};
+
+static const char *TGSI_TEXTURES[] =
+{
+   "TEXTURE_UNKNOWN",
+   "TEXTURE_1D",
+   "TEXTURE_2D",
+   "TEXTURE_3D",
+   "TEXTURE_CUBE",
+   "TEXTURE_RECT",
+   "TEXTURE_SHADOW1D",
+   "TEXTURE_SHADOW2D",
+   "TEXTURE_SHADOWRECT"
+};
+
+static const char *TGSI_SRC_REGISTER_EXTS[] =
+{
+   "SRC_REGISTER_EXT_TYPE_SWZ",
+   "SRC_REGISTER_EXT_TYPE_MOD"
+};
+
+static const char *TGSI_EXTSWIZZLES[] =
+{
+   "EXTSWIZZLE_X",
+   "EXTSWIZZLE_Y",
+   "EXTSWIZZLE_Z",
+   "EXTSWIZZLE_W",
+   "EXTSWIZZLE_ZERO",
+   "EXTSWIZZLE_ONE"
+};
+
+static const char *TGSI_WRITEMASKS[] =
+{
+   "0",
+   "WRITEMASK_X",
+   "WRITEMASK_Y",
+   "WRITEMASK_XY",
+   "WRITEMASK_Z",
+   "WRITEMASK_XZ",
+   "WRITEMASK_YZ",
+   "WRITEMASK_XYZ",
+   "WRITEMASK_W",
+   "WRITEMASK_XW",
+   "WRITEMASK_YW",
+   "WRITEMASK_XYW",
+   "WRITEMASK_ZW",
+   "WRITEMASK_XZW",
+   "WRITEMASK_YZW",
+   "WRITEMASK_XYZW"
+};
+
+static const char *TGSI_DST_REGISTER_EXTS[] =
+{
+   "DST_REGISTER_EXT_TYPE_CONDCODE",
+   "DST_REGISTER_EXT_TYPE_MODULATE"
+};
+
+static const char *TGSI_MODULATES[] =
+{
+   "MODULATE_1X",
+   "MODULATE_2X",
+   "MODULATE_4X",
+   "MODULATE_8X",
+   "MODULATE_HALF",
+   "MODULATE_QUARTER",
+   "MODULATE_EIGHTH"
+};
+
+static void
+dump_declaration_verbose(
+   struct tgsi_full_declaration  *decl,
+   unsigned                      ignored,
+   unsigned                      deflt,
+   struct tgsi_full_declaration  *fd )
+{
+   TXT( "\nFile       : " );
+   ENM( decl->Declaration.File, TGSI_FILES );
+   if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) {
+      TXT( "\nUsageMask  : " );
+      if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) {
+         CHR( 'X' );
+      }
+      if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) {
+         CHR( 'Y' );
+      }
+      if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) {
+         CHR( 'Z' );
+      }
+      if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) {
+         CHR( 'W' );
+      }
+   }
+   if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) {
+      TXT( "\nInterpolate: " );
+      ENM( decl->Declaration.Interpolate, TGSI_INTERPOLATES );
+   }
+   if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) {
+      TXT( "\nSemantic   : " );
+      UID( decl->Declaration.Semantic );
+   }
+   if( ignored ) {
+      TXT( "\nPadding    : " );
+      UIX( decl->Declaration.Padding );
+   }
+
+   EOL();
+   TXT( "\nFirst: " );
+   UID( decl->DeclarationRange.First );
+   TXT( "\nLast : " );
+   UID( decl->DeclarationRange.Last );
+
+   if( decl->Declaration.Semantic ) {
+      EOL();
+      TXT( "\nSemanticName : " );
+      ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS );
+      TXT( "\nSemanticIndex: " );
+      UID( decl->Semantic.SemanticIndex );
+      if( ignored ) {
+         TXT( "\nPadding      : " );
+         UIX( decl->Semantic.Padding );
+      }
+   }
+}
+
+static void
+dump_immediate_verbose(
+   struct tgsi_full_immediate *imm,
+   unsigned                   ignored )
+{
+   unsigned i;
+
+   TXT( "\nDataType   : " );
+   ENM( imm->Immediate.DataType, TGSI_IMMS );
+   if( ignored ) {
+      TXT( "\nPadding    : " );
+      UIX( imm->Immediate.Padding );
+   }
+
+   for( i = 0; i < imm->Immediate.NrTokens - 1; i++ ) {
+      EOL();
+      switch( imm->Immediate.DataType ) {
+      case TGSI_IMM_FLOAT32:
+         TXT( "\nFloat: " );
+         FLT( imm->u.ImmediateFloat32[i].Float );
+         break;
+
+      default:
+         assert( 0 );
+      }
+   }
+}
+
+static void
+dump_instruction_verbose(
+   struct tgsi_full_instruction  *inst,
+   unsigned                      ignored,
+   unsigned                      deflt,
+   struct tgsi_full_instruction  *fi )
+{
+   unsigned i;
+
+   TXT( "\nOpcode     : OPCODE_" );
+   TXT( tgsi_get_opcode_info( inst->Instruction.Opcode )->mnemonic );
+   if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) {
+      TXT( "\nSaturate   : " );
+      ENM( inst->Instruction.Saturate, TGSI_SATS );
+   }
+   if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) {
+      TXT( "\nNumDstRegs : " );
+      UID( inst->Instruction.NumDstRegs );
+   }
+   if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) {
+      TXT( "\nNumSrcRegs : " );
+      UID( inst->Instruction.NumSrcRegs );
+   }
+   if( ignored ) {
+      TXT( "\nPadding    : " );
+      UIX( inst->Instruction.Padding );
+   }
+
+   if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) {
+      EOL();
+      TXT( "\nType          : " );
+      ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS );
+      if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) {
+         TXT( "\nPrecision     : " );
+         ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS );
+      }
+      if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) {
+         TXT( "\nCondDstIndex  : " );
+         UID( inst->InstructionExtNv.CondDstIndex );
+      }
+      if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) {
+         TXT( "\nCondFlowIndex : " );
+         UID( inst->InstructionExtNv.CondFlowIndex );
+      }
+      if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) {
+         TXT( "\nCondMask      : " );
+         ENM( inst->InstructionExtNv.CondMask, TGSI_CCS );
+      }
+      if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) {
+         TXT( "\nCondSwizzleX  : " );
+         ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES );
+      }
+      if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) {
+         TXT( "\nCondSwizzleY  : " );
+         ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES );
+      }
+      if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) {
+         TXT( "\nCondSwizzleZ  : " );
+         ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES );
+      }
+      if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) {
+         TXT( "\nCondSwizzleW  : " );
+         ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES );
+      }
+      if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) {
+         TXT( "\nCondDstUpdate : " );
+         UID( inst->InstructionExtNv.CondDstUpdate );
+      }
+      if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) {
+         TXT( "\nCondFlowEnable: " );
+         UID( inst->InstructionExtNv.CondFlowEnable );
+      }
+      if( ignored ) {
+         TXT( "\nPadding       : " );
+         UIX( inst->InstructionExtNv.Padding );
+         if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) {
+            TXT( "\nExtended      : " );
+            UID( inst->InstructionExtNv.Extended );
+         }
+      }
+   }
+
+   if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) {
+      EOL();
+      TXT( "\nType    : " );
+      ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS );
+      if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) {
+         TXT( "\nLabel   : " );
+         UID( inst->InstructionExtLabel.Label );
+      }
+      if( ignored ) {
+         TXT( "\nPadding : " );
+         UIX( inst->InstructionExtLabel.Padding );
+         if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) {
+            TXT( "\nExtended: " );
+            UID( inst->InstructionExtLabel.Extended );
+         }
+      }
+   }
+
+   if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) {
+      EOL();
+      TXT( "\nType    : " );
+      ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS );
+      if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) {
+         TXT( "\nTexture : " );
+         ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES );
+      }
+      if( ignored ) {
+         TXT( "\nPadding : " );
+         UIX( inst->InstructionExtTexture.Padding );
+         if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) {
+            TXT( "\nExtended: " );
+            UID( inst->InstructionExtTexture.Extended );
+         }
+      }
+   }
+
+   for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) {
+      struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
+      struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i];
+
+      EOL();
+      TXT( "\nFile     : " );
+      ENM( dst->DstRegister.File, TGSI_FILES );
+      if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) {
+         TXT( "\nWriteMask: " );
+         ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS );
+      }
+      if( ignored ) {
+         if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) {
+            TXT( "\nIndirect : " );
+            UID( dst->DstRegister.Indirect );
+         }
+         if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) {
+            TXT( "\nDimension: " );
+            UID( dst->DstRegister.Dimension );
+         }
+      }
+      if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) {
+         TXT( "\nIndex    : " );
+         SID( dst->DstRegister.Index );
+      }
+      if( ignored ) {
+         TXT( "\nPadding  : " );
+         UIX( dst->DstRegister.Padding );
+         if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) {
+            TXT( "\nExtended : " );
+            UID( dst->DstRegister.Extended );
+         }
+      }
+
+      if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) {
+         EOL();
+         TXT( "\nType        : " );
+         ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS );
+         if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) {
+            TXT( "\nCondMask    : " );
+            ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS );
+         }
+         if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) {
+            TXT( "\nCondSwizzleX: " );
+            ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES );
+         }
+         if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) {
+            TXT( "\nCondSwizzleY: " );
+            ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES );
+         }
+         if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) {
+            TXT( "\nCondSwizzleZ: " );
+            ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES );
+         }
+         if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) {
+            TXT( "\nCondSwizzleW: " );
+            ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES );
+         }
+         if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) {
+            TXT( "\nCondSrcIndex: " );
+            UID( dst->DstRegisterExtConcode.CondSrcIndex );
+         }
+         if( ignored ) {
+            TXT( "\nPadding     : " );
+            UIX( dst->DstRegisterExtConcode.Padding );
+            if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) {
+               TXT( "\nExtended    : " );
+               UID( dst->DstRegisterExtConcode.Extended );
+            }
+         }
+      }
+
+      if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) {
+         EOL();
+         TXT( "\nType    : " );
+         ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS );
+         if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) {
+            TXT( "\nModulate: " );
+            ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES );
+         }
+         if( ignored ) {
+            TXT( "\nPadding : " );
+            UIX( dst->DstRegisterExtModulate.Padding );
+            if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) {
+               TXT( "\nExtended: " );
+               UID( dst->DstRegisterExtModulate.Extended );
+            }
+         }
+      }
+   }
+
+   for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) {
+      struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i];
+      struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i];
+
+      EOL();
+      TXT( "\nFile     : ");
+      ENM( src->SrcRegister.File, TGSI_FILES );
+      if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) {
+         TXT( "\nSwizzleX : " );
+         ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES );
+      }
+      if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) {
+         TXT( "\nSwizzleY : " );
+         ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES );
+      }
+      if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) {
+         TXT( "\nSwizzleZ : " );
+         ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES );
+      }
+      if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) {
+         TXT( "\nSwizzleW : " );
+         ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES );
+      }
+      if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) {
+         TXT( "\nNegate   : " );
+         UID( src->SrcRegister.Negate );
+      }
+      if( ignored ) {
+         if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) {
+            TXT( "\nIndirect : " );
+            UID( src->SrcRegister.Indirect );
+         }
+         if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) {
+            TXT( "\nDimension: " );
+            UID( src->SrcRegister.Dimension );
+         }
+      }
+      if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) {
+         TXT( "\nIndex    : " );
+         SID( src->SrcRegister.Index );
+      }
+      if( ignored ) {
+         if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) {
+            TXT( "\nExtended : " );
+            UID( src->SrcRegister.Extended );
+         }
+      }
+
+      if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) {
+         EOL();
+         TXT( "\nType       : " );
+         ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS );
+         if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) {
+            TXT( "\nExtSwizzleX: " );
+            ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES );
+         }
+         if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) {
+            TXT( "\nExtSwizzleY: " );
+            ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES );
+         }
+         if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) {
+            TXT( "\nExtSwizzleZ: " );
+            ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES );
+         }
+         if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) {
+            TXT( "\nExtSwizzleW: " );
+            ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES );
+         }
+         if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) {
+            TXT( "\nNegateX   : " );
+            UID( src->SrcRegisterExtSwz.NegateX );
+         }
+         if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) {
+            TXT( "\nNegateY   : " );
+            UID( src->SrcRegisterExtSwz.NegateY );
+         }
+         if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) {
+            TXT( "\nNegateZ   : " );
+            UID( src->SrcRegisterExtSwz.NegateZ );
+         }
+         if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) {
+            TXT( "\nNegateW   : " );
+            UID( src->SrcRegisterExtSwz.NegateW );
+         }
+         if( ignored ) {
+            TXT( "\nPadding   : " );
+            UIX( src->SrcRegisterExtSwz.Padding );
+            if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) {
+               TXT( "\nExtended   : " );
+               UID( src->SrcRegisterExtSwz.Extended );
+            }
+         }
+      }
+
+      if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) {
+         EOL();
+         TXT( "\nType     : " );
+         ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS );
+         if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) {
+            TXT( "\nComplement: " );
+            UID( src->SrcRegisterExtMod.Complement );
+         }
+         if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) {
+            TXT( "\nBias     : " );
+            UID( src->SrcRegisterExtMod.Bias );
+         }
+         if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) {
+            TXT( "\nScale2X   : " );
+            UID( src->SrcRegisterExtMod.Scale2X );
+         }
+         if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) {
+            TXT( "\nAbsolute  : " );
+            UID( src->SrcRegisterExtMod.Absolute );
+         }
+         if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) {
+            TXT( "\nNegate   : " );
+            UID( src->SrcRegisterExtMod.Negate );
+         }
+         if( ignored ) {
+            TXT( "\nPadding   : " );
+            UIX( src->SrcRegisterExtMod.Padding );
+            if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) {
+               TXT( "\nExtended  : " );
+               UID( src->SrcRegisterExtMod.Extended );
+            }
+         }
+      }
+   }
+}
+
+void
+tgsi_dump_c(
+   const struct tgsi_token *tokens,
+   uint flags )
+{
+   struct tgsi_parse_context parse;
+   struct tgsi_full_instruction fi;
+   struct tgsi_full_declaration fd;
+   uint ignored = flags & TGSI_DUMP_C_IGNORED;
+   uint deflt = flags & TGSI_DUMP_C_DEFAULT;
+
+   tgsi_parse_init( &parse, tokens );
+
+   TXT( "tgsi-dump begin -----------------" );
+
+   TXT( "\nMajorVersion: " );
+   UID( parse.FullVersion.Version.MajorVersion );
+   TXT( "\nMinorVersion: " );
+   UID( parse.FullVersion.Version.MinorVersion );
+   EOL();
+
+   TXT( "\nHeaderSize: " );
+   UID( parse.FullHeader.Header.HeaderSize );
+   TXT( "\nBodySize  : " );
+   UID( parse.FullHeader.Header.BodySize );
+   TXT( "\nProcessor : " );
+   ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES );
+   EOL();
+
+   fi = tgsi_default_full_instruction();
+   fd = tgsi_default_full_declaration();
+
+   while( !tgsi_parse_end_of_tokens( &parse ) ) {
+      tgsi_parse_token( &parse );
+
+      TXT( "\nType       : " );
+      ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES );
+      if( ignored ) {
+         TXT( "\nSize       : " );
+         UID( parse.FullToken.Token.NrTokens );
+         if( deflt || parse.FullToken.Token.Extended ) {
+            TXT( "\nExtended   : " );
+            UID( parse.FullToken.Token.Extended );
+         }
+      }
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         dump_declaration_verbose(
+            &parse.FullToken.FullDeclaration,
+            ignored,
+            deflt,
+            &fd );
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         dump_immediate_verbose(
+            &parse.FullToken.FullImmediate,
+            ignored );
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         dump_instruction_verbose(
+            &parse.FullToken.FullInstruction,
+            ignored,
+            deflt,
+            &fi );
+         break;
+
+      default:
+         assert( 0 );
+      }
+
+      EOL();
+   }
+
+   TXT( "\ntgsi-dump end -------------------\n" );
+
+   tgsi_parse_free( &parse );
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.h b/src/gallium/auxiliary/tgsi/tgsi_dump_c.h
new file mode 100644
index 0000000000..d91cd35b3b
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.h
@@ -0,0 +1,49 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_DUMP_C_H
+#define TGSI_DUMP_C_H
+
+#include "pipe/p_shader_tokens.h"
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+#define TGSI_DUMP_C_IGNORED 1
+#define TGSI_DUMP_C_DEFAULT 2
+
+void
+tgsi_dump_c(
+   const struct tgsi_token *tokens,
+   uint flags );
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_DUMP_C_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
new file mode 100644
index 0000000000..ab641efb60
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -0,0 +1,2960 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * TGSI interpreter/executor.
+ *
+ * Flow control information:
+ *
+ * Since we operate on 'quads' (4 pixels or 4 vertices in parallel)
+ * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special
+ * care since a condition may be true for some quad components but false
+ * for other components.
+ *
+ * We basically execute all statements (even if they're in the part of
+ * an IF/ELSE clause that's "not taken") and use a special mask to
+ * control writing to destination registers.  This is the ExecMask.
+ * See store_dest().
+ *
+ * The ExecMask is computed from three other masks (CondMask, LoopMask and
+ * ContMask) which are controlled by the flow control instructions (namely:
+ * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT).
+ *
+ *
+ * Authors:
+ *   Michal Krol
+ *   Brian Paul
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi_exec.h"
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#define FAST_MATH 1
+
+#define TILE_TOP_LEFT     0
+#define TILE_TOP_RIGHT    1
+#define TILE_BOTTOM_LEFT  2
+#define TILE_BOTTOM_RIGHT 3
+
+#define CHAN_X  0
+#define CHAN_Y  1
+#define CHAN_Z  2
+#define CHAN_W  3
+
+/*
+ * Shorthand locations of various utility registers (_I = Index, _C = Channel)
+ */
+#define TEMP_0_I           TGSI_EXEC_TEMP_00000000_I
+#define TEMP_0_C           TGSI_EXEC_TEMP_00000000_C
+#define TEMP_7F_I          TGSI_EXEC_TEMP_7FFFFFFF_I
+#define TEMP_7F_C          TGSI_EXEC_TEMP_7FFFFFFF_C
+#define TEMP_80_I          TGSI_EXEC_TEMP_80000000_I
+#define TEMP_80_C          TGSI_EXEC_TEMP_80000000_C
+#define TEMP_FF_I          TGSI_EXEC_TEMP_FFFFFFFF_I
+#define TEMP_FF_C          TGSI_EXEC_TEMP_FFFFFFFF_C
+#define TEMP_1_I           TGSI_EXEC_TEMP_ONE_I
+#define TEMP_1_C           TGSI_EXEC_TEMP_ONE_C
+#define TEMP_2_I           TGSI_EXEC_TEMP_TWO_I
+#define TEMP_2_C           TGSI_EXEC_TEMP_TWO_C
+#define TEMP_128_I         TGSI_EXEC_TEMP_128_I
+#define TEMP_128_C         TGSI_EXEC_TEMP_128_C
+#define TEMP_M128_I        TGSI_EXEC_TEMP_MINUS_128_I
+#define TEMP_M128_C        TGSI_EXEC_TEMP_MINUS_128_C
+#define TEMP_KILMASK_I     TGSI_EXEC_TEMP_KILMASK_I
+#define TEMP_KILMASK_C     TGSI_EXEC_TEMP_KILMASK_C
+#define TEMP_OUTPUT_I      TGSI_EXEC_TEMP_OUTPUT_I
+#define TEMP_OUTPUT_C      TGSI_EXEC_TEMP_OUTPUT_C
+#define TEMP_PRIMITIVE_I   TGSI_EXEC_TEMP_PRIMITIVE_I
+#define TEMP_PRIMITIVE_C   TGSI_EXEC_TEMP_PRIMITIVE_C
+#define TEMP_CC_I          TGSI_EXEC_TEMP_CC_I
+#define TEMP_CC_C          TGSI_EXEC_TEMP_CC_C
+#define TEMP_3_I           TGSI_EXEC_TEMP_THREE_I
+#define TEMP_3_C           TGSI_EXEC_TEMP_THREE_C
+#define TEMP_HALF_I        TGSI_EXEC_TEMP_HALF_I
+#define TEMP_HALF_C        TGSI_EXEC_TEMP_HALF_C
+#define TEMP_R0            TGSI_EXEC_TEMP_R0
+
+#define IS_CHANNEL_ENABLED(INST, CHAN)\
+   ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN)))
+
+#define IS_CHANNEL_ENABLED2(INST, CHAN)\
+   ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN)))
+
+#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\
+   for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)\
+      if (IS_CHANNEL_ENABLED( INST, CHAN ))
+
+#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\
+   for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)\
+      if (IS_CHANNEL_ENABLED2( INST, CHAN ))
+
+
+/** The execution mask depends on the conditional mask and the loop mask */
+#define UPDATE_EXEC_MASK(MACH) \
+      MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask
+
+/**
+ * Initialize machine state by expanding tokens to full instructions,
+ * allocating temporary storage, setting up constants, etc.
+ * After this, we can call tgsi_exec_machine_run() many times.
+ */
+void 
+tgsi_exec_machine_bind_shader(
+   struct tgsi_exec_machine *mach,
+   const struct tgsi_token *tokens,
+   uint numSamplers,
+   struct tgsi_sampler **samplers)
+{
+   uint k;
+   struct tgsi_parse_context parse;
+   struct tgsi_exec_labels *labels = &mach->Labels;
+   struct tgsi_full_instruction *instructions;
+   struct tgsi_full_declaration *declarations;
+   uint maxInstructions = 10, numInstructions = 0;
+   uint maxDeclarations = 10, numDeclarations = 0;
+   uint instno = 0;
+
+#if 0
+   tgsi_dump(tokens, 0);
+#endif
+
+   util_init_math();
+
+   mach->Tokens = tokens;
+   mach->Samplers = samplers;
+
+   k = tgsi_parse_init (&parse, mach->Tokens);
+   if (k != TGSI_PARSE_OK) {
+      debug_printf( "Problem parsing!\n" );
+      return;
+   }
+
+   mach->Processor = parse.FullHeader.Processor.Processor;
+   mach->ImmLimit = 0;
+   labels->count = 0;
+
+   declarations = (struct tgsi_full_declaration *)
+      MALLOC( maxDeclarations * sizeof(struct tgsi_full_declaration) );
+
+   if (!declarations) {
+      return;
+   }
+
+   instructions = (struct tgsi_full_instruction *)
+      MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) );
+
+   if (!instructions) {
+      FREE( declarations );
+      return;
+   }
+
+   while( !tgsi_parse_end_of_tokens( &parse ) ) {
+      uint pointer = parse.Position;
+      uint i;
+
+      tgsi_parse_token( &parse );
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         /* save expanded declaration */
+         if (numDeclarations == maxDeclarations) {
+            declarations = REALLOC(declarations,
+                                   maxDeclarations
+                                   * sizeof(struct tgsi_full_declaration),
+                                   (maxDeclarations + 10)
+                                   * sizeof(struct tgsi_full_declaration));
+            maxDeclarations += 10;
+         }
+         memcpy(declarations + numDeclarations,
+                &parse.FullToken.FullDeclaration,
+                sizeof(declarations[0]));
+         numDeclarations++;
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         {
+            uint size = parse.FullToken.FullImmediate.Immediate.NrTokens - 1;
+            assert( size % 4 == 0 );
+            assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES );
+
+            for( i = 0; i < size; i++ ) {
+               mach->Imms[mach->ImmLimit + i / 4][i % 4] = 
+		  parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float;
+            }
+            mach->ImmLimit += size / 4;
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         assert( labels->count < MAX_LABELS );
+
+         labels->labels[labels->count][0] = instno;
+         labels->labels[labels->count][1] = pointer;
+         labels->count++;
+
+         /* save expanded instruction */
+         if (numInstructions == maxInstructions) {
+            instructions = REALLOC(instructions,
+                                   maxInstructions
+                                   * sizeof(struct tgsi_full_instruction),
+                                   (maxInstructions + 10)
+                                   * sizeof(struct tgsi_full_instruction));
+            maxInstructions += 10;
+         }
+         memcpy(instructions + numInstructions,
+                &parse.FullToken.FullInstruction,
+                sizeof(instructions[0]));
+         numInstructions++;
+         break;
+
+      default:
+         assert( 0 );
+      }
+   }
+   tgsi_parse_free (&parse);
+
+   if (mach->Declarations) {
+      FREE( mach->Declarations );
+   }
+   mach->Declarations = declarations;
+   mach->NumDeclarations = numDeclarations;
+
+   if (mach->Instructions) {
+      FREE( mach->Instructions );
+   }
+   mach->Instructions = instructions;
+   mach->NumInstructions = numInstructions;
+}
+
+
+void
+tgsi_exec_machine_init(
+   struct tgsi_exec_machine *mach )
+{
+   uint i;
+
+   mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps);
+   mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR];
+
+   /* Setup constants. */
+   for( i = 0; i < 4; i++ ) {
+      mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000;
+      mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF;
+      mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000;
+      mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF;
+      mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f;
+      mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f;
+      mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f;
+      mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f;
+      mach->Temps[TEMP_3_I].xyzw[TEMP_3_C].f[i] = 3.0f;
+      mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C].f[i] = 0.5f;
+   }
+}
+
+
+void
+tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach)
+{
+   if (mach->Instructions) {
+      FREE(mach->Instructions);
+      mach->Instructions = NULL;
+      mach->NumInstructions = 0;
+   }
+   if (mach->Declarations) {
+      FREE(mach->Declarations);
+      mach->Declarations = NULL;
+      mach->NumDeclarations = 0;
+   }
+}
+
+
+static void
+micro_abs(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = fabsf( src->f[0] );
+   dst->f[1] = fabsf( src->f[1] );
+   dst->f[2] = fabsf( src->f[2] );
+   dst->f[3] = fabsf( src->f[3] );
+}
+
+static void
+micro_add(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->f[0] = src0->f[0] + src1->f[0];
+   dst->f[1] = src0->f[1] + src1->f[1];
+   dst->f[2] = src0->f[2] + src1->f[2];
+   dst->f[3] = src0->f[3] + src1->f[3];
+}
+
+#if 0
+static void
+micro_iadd(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->i[0] = src0->i[0] + src1->i[0];
+   dst->i[1] = src0->i[1] + src1->i[1];
+   dst->i[2] = src0->i[2] + src1->i[2];
+   dst->i[3] = src0->i[3] + src1->i[3];
+}
+#endif
+
+static void
+micro_and(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->u[0] = src0->u[0] & src1->u[0];
+   dst->u[1] = src0->u[1] & src1->u[1];
+   dst->u[2] = src0->u[2] & src1->u[2];
+   dst->u[3] = src0->u[3] & src1->u[3];
+}
+
+static void
+micro_ceil(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = ceilf( src->f[0] );
+   dst->f[1] = ceilf( src->f[1] );
+   dst->f[2] = ceilf( src->f[2] );
+   dst->f[3] = ceilf( src->f[3] );
+}
+
+static void
+micro_cos(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = cosf( src->f[0] );
+   dst->f[1] = cosf( src->f[1] );
+   dst->f[2] = cosf( src->f[2] );
+   dst->f[3] = cosf( src->f[3] );
+}
+
+static void
+micro_ddx(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] =
+   dst->f[1] =
+   dst->f[2] =
+   dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT];
+}
+
+static void
+micro_ddy(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] =
+   dst->f[1] =
+   dst->f[2] =
+   dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT];
+}
+
+static void
+micro_div(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   if (src1->f[0] != 0) {
+      dst->f[0] = src0->f[0] / src1->f[0];
+   }
+   if (src1->f[1] != 0) {
+      dst->f[1] = src0->f[1] / src1->f[1];
+   }
+   if (src1->f[2] != 0) {
+      dst->f[2] = src0->f[2] / src1->f[2];
+   }
+   if (src1->f[3] != 0) {
+      dst->f[3] = src0->f[3] / src1->f[3];
+   }
+}
+
+#if 0
+static void
+micro_udiv(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->u[0] = src0->u[0] / src1->u[0];
+   dst->u[1] = src0->u[1] / src1->u[1];
+   dst->u[2] = src0->u[2] / src1->u[2];
+   dst->u[3] = src0->u[3] / src1->u[3];
+}
+#endif
+
+static void
+micro_eq(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1,
+   const union tgsi_exec_channel *src2,
+   const union tgsi_exec_channel *src3 )
+{
+   dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0];
+   dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1];
+   dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2];
+   dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3];
+}
+
+#if 0
+static void
+micro_ieq(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1,
+   const union tgsi_exec_channel *src2,
+   const union tgsi_exec_channel *src3 )
+{
+   dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0];
+   dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1];
+   dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2];
+   dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3];
+}
+#endif
+
+static void
+micro_exp2(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src)
+{
+#if FAST_MATH
+   dst->f[0] = util_fast_exp2( src->f[0] );
+   dst->f[1] = util_fast_exp2( src->f[1] );
+   dst->f[2] = util_fast_exp2( src->f[2] );
+   dst->f[3] = util_fast_exp2( src->f[3] );
+#else
+   dst->f[0] = powf( 2.0f, src->f[0] );
+   dst->f[1] = powf( 2.0f, src->f[1] );
+   dst->f[2] = powf( 2.0f, src->f[2] );
+   dst->f[3] = powf( 2.0f, src->f[3] );
+#endif
+}
+
+#if 0
+static void
+micro_f2ut(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->u[0] = (uint) src->f[0];
+   dst->u[1] = (uint) src->f[1];
+   dst->u[2] = (uint) src->f[2];
+   dst->u[3] = (uint) src->f[3];
+}
+#endif
+
+static void
+micro_flr(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = floorf( src->f[0] );
+   dst->f[1] = floorf( src->f[1] );
+   dst->f[2] = floorf( src->f[2] );
+   dst->f[3] = floorf( src->f[3] );
+}
+
+static void
+micro_frc(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = src->f[0] - floorf( src->f[0] );
+   dst->f[1] = src->f[1] - floorf( src->f[1] );
+   dst->f[2] = src->f[2] - floorf( src->f[2] );
+   dst->f[3] = src->f[3] - floorf( src->f[3] );
+}
+
+static void
+micro_ge(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1,
+   const union tgsi_exec_channel *src2,
+   const union tgsi_exec_channel *src3 )
+{
+   dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0];
+   dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1];
+   dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2];
+   dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3];
+}
+
+static void
+micro_i2f(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = (float) src->i[0];
+   dst->f[1] = (float) src->i[1];
+   dst->f[2] = (float) src->i[2];
+   dst->f[3] = (float) src->i[3];
+}
+
+static void
+micro_lg2(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+#if FAST_MATH
+   dst->f[0] = util_fast_log2( src->f[0] );
+   dst->f[1] = util_fast_log2( src->f[1] );
+   dst->f[2] = util_fast_log2( src->f[2] );
+   dst->f[3] = util_fast_log2( src->f[3] );
+#else
+   dst->f[0] = logf( src->f[0] ) * 1.442695f;
+   dst->f[1] = logf( src->f[1] ) * 1.442695f;
+   dst->f[2] = logf( src->f[2] ) * 1.442695f;
+   dst->f[3] = logf( src->f[3] ) * 1.442695f;
+#endif
+}
+
+static void
+micro_le(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1,
+   const union tgsi_exec_channel *src2,
+   const union tgsi_exec_channel *src3 )
+{
+   dst->f[0] = src0->f[0] <= src1->f[0] ? src2->f[0] : src3->f[0];
+   dst->f[1] = src0->f[1] <= src1->f[1] ? src2->f[1] : src3->f[1];
+   dst->f[2] = src0->f[2] <= src1->f[2] ? src2->f[2] : src3->f[2];
+   dst->f[3] = src0->f[3] <= src1->f[3] ? src2->f[3] : src3->f[3];
+}
+
+static void
+micro_lt(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1,
+   const union tgsi_exec_channel *src2,
+   const union tgsi_exec_channel *src3 )
+{
+   dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0];
+   dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1];
+   dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2];
+   dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3];
+}
+
+#if 0
+static void
+micro_ilt(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1,
+   const union tgsi_exec_channel *src2,
+   const union tgsi_exec_channel *src3 )
+{
+   dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0];
+   dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1];
+   dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2];
+   dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3];
+}
+#endif
+
+#if 0
+static void
+micro_ult(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1,
+   const union tgsi_exec_channel *src2,
+   const union tgsi_exec_channel *src3 )
+{
+   dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0];
+   dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1];
+   dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2];
+   dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3];
+}
+#endif
+
+static void
+micro_max(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0];
+   dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1];
+   dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2];
+   dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3];
+}
+
+#if 0
+static void
+micro_imax(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0];
+   dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1];
+   dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2];
+   dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3];
+}
+#endif
+
+#if 0
+static void
+micro_umax(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0];
+   dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1];
+   dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2];
+   dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3];
+}
+#endif
+
+static void
+micro_min(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0];
+   dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1];
+   dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2];
+   dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3];
+}
+
+#if 0
+static void
+micro_imin(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0];
+   dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1];
+   dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2];
+   dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3];
+}
+#endif
+
+#if 0
+static void
+micro_umin(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0];
+   dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1];
+   dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2];
+   dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3];
+}
+#endif
+
+#if 0
+static void
+micro_umod(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->u[0] = src0->u[0] % src1->u[0];
+   dst->u[1] = src0->u[1] % src1->u[1];
+   dst->u[2] = src0->u[2] % src1->u[2];
+   dst->u[3] = src0->u[3] % src1->u[3];
+}
+#endif
+
+static void
+micro_mul(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->f[0] = src0->f[0] * src1->f[0];
+   dst->f[1] = src0->f[1] * src1->f[1];
+   dst->f[2] = src0->f[2] * src1->f[2];
+   dst->f[3] = src0->f[3] * src1->f[3];
+}
+
+#if 0
+static void
+micro_imul(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->i[0] = src0->i[0] * src1->i[0];
+   dst->i[1] = src0->i[1] * src1->i[1];
+   dst->i[2] = src0->i[2] * src1->i[2];
+   dst->i[3] = src0->i[3] * src1->i[3];
+}
+#endif
+
+#if 0
+static void
+micro_imul64(
+   union tgsi_exec_channel *dst0,
+   union tgsi_exec_channel *dst1,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst1->i[0] = src0->i[0] * src1->i[0];
+   dst1->i[1] = src0->i[1] * src1->i[1];
+   dst1->i[2] = src0->i[2] * src1->i[2];
+   dst1->i[3] = src0->i[3] * src1->i[3];
+   dst0->i[0] = 0;
+   dst0->i[1] = 0;
+   dst0->i[2] = 0;
+   dst0->i[3] = 0;
+}
+#endif
+
+#if 0
+static void
+micro_umul64(
+   union tgsi_exec_channel *dst0,
+   union tgsi_exec_channel *dst1,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst1->u[0] = src0->u[0] * src1->u[0];
+   dst1->u[1] = src0->u[1] * src1->u[1];
+   dst1->u[2] = src0->u[2] * src1->u[2];
+   dst1->u[3] = src0->u[3] * src1->u[3];
+   dst0->u[0] = 0;
+   dst0->u[1] = 0;
+   dst0->u[2] = 0;
+   dst0->u[3] = 0;
+}
+#endif
+
+
+#if 0
+static void
+micro_movc(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1,
+   const union tgsi_exec_channel *src2 )
+{
+   dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0];
+   dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1];
+   dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2];
+   dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3];
+}
+#endif
+
+static void
+micro_neg(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = -src->f[0];
+   dst->f[1] = -src->f[1];
+   dst->f[2] = -src->f[2];
+   dst->f[3] = -src->f[3];
+}
+
+#if 0
+static void
+micro_ineg(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->i[0] = -src->i[0];
+   dst->i[1] = -src->i[1];
+   dst->i[2] = -src->i[2];
+   dst->i[3] = -src->i[3];
+}
+#endif
+
+static void
+micro_not(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->u[0] = ~src->u[0];
+   dst->u[1] = ~src->u[1];
+   dst->u[2] = ~src->u[2];
+   dst->u[3] = ~src->u[3];
+}
+
+static void
+micro_or(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->u[0] = src0->u[0] | src1->u[0];
+   dst->u[1] = src0->u[1] | src1->u[1];
+   dst->u[2] = src0->u[2] | src1->u[2];
+   dst->u[3] = src0->u[3] | src1->u[3];
+}
+
+static void
+micro_pow(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+#if FAST_MATH
+   dst->f[0] = util_fast_pow( src0->f[0], src1->f[0] );
+   dst->f[1] = util_fast_pow( src0->f[1], src1->f[1] );
+   dst->f[2] = util_fast_pow( src0->f[2], src1->f[2] );
+   dst->f[3] = util_fast_pow( src0->f[3], src1->f[3] );
+#else
+   dst->f[0] = powf( src0->f[0], src1->f[0] );
+   dst->f[1] = powf( src0->f[1], src1->f[1] );
+   dst->f[2] = powf( src0->f[2], src1->f[2] );
+   dst->f[3] = powf( src0->f[3], src1->f[3] );
+#endif
+}
+
+static void
+micro_rnd(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = floorf( src->f[0] + 0.5f );
+   dst->f[1] = floorf( src->f[1] + 0.5f );
+   dst->f[2] = floorf( src->f[2] + 0.5f );
+   dst->f[3] = floorf( src->f[3] + 0.5f );
+}
+
+static void
+micro_sgn(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = src->f[0] < 0.0f ? -1.0f : src->f[0] > 0.0f ? 1.0f : 0.0f;
+   dst->f[1] = src->f[1] < 0.0f ? -1.0f : src->f[1] > 0.0f ? 1.0f : 0.0f;
+   dst->f[2] = src->f[2] < 0.0f ? -1.0f : src->f[2] > 0.0f ? 1.0f : 0.0f;
+   dst->f[3] = src->f[3] < 0.0f ? -1.0f : src->f[3] > 0.0f ? 1.0f : 0.0f;
+}
+
+static void
+micro_shl(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->i[0] = src0->i[0] << src1->i[0];
+   dst->i[1] = src0->i[1] << src1->i[1];
+   dst->i[2] = src0->i[2] << src1->i[2];
+   dst->i[3] = src0->i[3] << src1->i[3];
+}
+
+static void
+micro_ishr(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->i[0] = src0->i[0] >> src1->i[0];
+   dst->i[1] = src0->i[1] >> src1->i[1];
+   dst->i[2] = src0->i[2] >> src1->i[2];
+   dst->i[3] = src0->i[3] >> src1->i[3];
+}
+
+static void
+micro_trunc(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0 )
+{
+   dst->f[0] = (float) (int) src0->f[0];
+   dst->f[1] = (float) (int) src0->f[1];
+   dst->f[2] = (float) (int) src0->f[2];
+   dst->f[3] = (float) (int) src0->f[3];
+}
+
+#if 0
+static void
+micro_ushr(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->u[0] = src0->u[0] >> src1->u[0];
+   dst->u[1] = src0->u[1] >> src1->u[1];
+   dst->u[2] = src0->u[2] >> src1->u[2];
+   dst->u[3] = src0->u[3] >> src1->u[3];
+}
+#endif
+
+static void
+micro_sin(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = sinf( src->f[0] );
+   dst->f[1] = sinf( src->f[1] );
+   dst->f[2] = sinf( src->f[2] );
+   dst->f[3] = sinf( src->f[3] );
+}
+
+static void
+micro_sqrt( union tgsi_exec_channel *dst,
+            const union tgsi_exec_channel *src )
+{
+   dst->f[0] = sqrtf( src->f[0] );
+   dst->f[1] = sqrtf( src->f[1] );
+   dst->f[2] = sqrtf( src->f[2] );
+   dst->f[3] = sqrtf( src->f[3] );
+}
+
+static void
+micro_sub(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->f[0] = src0->f[0] - src1->f[0];
+   dst->f[1] = src0->f[1] - src1->f[1];
+   dst->f[2] = src0->f[2] - src1->f[2];
+   dst->f[3] = src0->f[3] - src1->f[3];
+}
+
+#if 0
+static void
+micro_u2f(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src )
+{
+   dst->f[0] = (float) src->u[0];
+   dst->f[1] = (float) src->u[1];
+   dst->f[2] = (float) src->u[2];
+   dst->f[3] = (float) src->u[3];
+}
+#endif
+
+static void
+micro_xor(
+   union tgsi_exec_channel *dst,
+   const union tgsi_exec_channel *src0,
+   const union tgsi_exec_channel *src1 )
+{
+   dst->u[0] = src0->u[0] ^ src1->u[0];
+   dst->u[1] = src0->u[1] ^ src1->u[1];
+   dst->u[2] = src0->u[2] ^ src1->u[2];
+   dst->u[3] = src0->u[3] ^ src1->u[3];
+}
+
+static void
+fetch_src_file_channel(
+   const struct tgsi_exec_machine *mach,
+   const uint file,
+   const uint swizzle,
+   const union tgsi_exec_channel *index,
+   union tgsi_exec_channel *chan )
+{
+   switch( swizzle ) {
+   case TGSI_EXTSWIZZLE_X:
+   case TGSI_EXTSWIZZLE_Y:
+   case TGSI_EXTSWIZZLE_Z:
+   case TGSI_EXTSWIZZLE_W:
+      switch( file ) {
+      case TGSI_FILE_CONSTANT:
+         assert(mach->Consts);
+         if (index->i[0] < 0)
+            chan->f[0] = 0.0f;
+         else
+            chan->f[0] = mach->Consts[index->i[0]][swizzle];
+         if (index->i[1] < 0)
+            chan->f[1] = 0.0f;
+         else
+            chan->f[1] = mach->Consts[index->i[1]][swizzle];
+         if (index->i[2] < 0)
+            chan->f[2] = 0.0f;
+         else
+            chan->f[2] = mach->Consts[index->i[2]][swizzle];
+         if (index->i[3] < 0)
+            chan->f[3] = 0.0f;
+         else
+            chan->f[3] = mach->Consts[index->i[3]][swizzle];
+         break;
+
+      case TGSI_FILE_INPUT:
+         chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0];
+         chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1];
+         chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2];
+         chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3];
+         break;
+
+      case TGSI_FILE_TEMPORARY:
+         assert(index->i[0] < TGSI_EXEC_NUM_TEMPS);
+         chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0];
+         chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1];
+         chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2];
+         chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3];
+         break;
+
+      case TGSI_FILE_IMMEDIATE:
+         assert( index->i[0] < (int) mach->ImmLimit );
+         chan->f[0] = mach->Imms[index->i[0]][swizzle];
+         assert( index->i[1] < (int) mach->ImmLimit );
+         chan->f[1] = mach->Imms[index->i[1]][swizzle];
+         assert( index->i[2] < (int) mach->ImmLimit );
+         chan->f[2] = mach->Imms[index->i[2]][swizzle];
+         assert( index->i[3] < (int) mach->ImmLimit );
+         chan->f[3] = mach->Imms[index->i[3]][swizzle];
+         break;
+
+      case TGSI_FILE_ADDRESS:
+         chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0];
+         chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1];
+         chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2];
+         chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3];
+         break;
+
+      case TGSI_FILE_OUTPUT:
+         /* vertex/fragment output vars can be read too */
+         chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0];
+         chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1];
+         chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2];
+         chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3];
+         break;
+
+      default:
+         assert( 0 );
+      }
+      break;
+
+   case TGSI_EXTSWIZZLE_ZERO:
+      *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C];
+      break;
+
+   case TGSI_EXTSWIZZLE_ONE:
+      *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C];
+      break;
+
+   default:
+      assert( 0 );
+   }
+}
+
+static void
+fetch_source(
+   const struct tgsi_exec_machine *mach,
+   union tgsi_exec_channel *chan,
+   const struct tgsi_full_src_register *reg,
+   const uint chan_index )
+{
+   union tgsi_exec_channel index;
+   uint swizzle;
+
+   /* We start with a direct index into a register file.
+    *
+    *    file[1],
+    *    where:
+    *       file = SrcRegister.File
+    *       [1] = SrcRegister.Index
+    */
+   index.i[0] =
+   index.i[1] =
+   index.i[2] =
+   index.i[3] = reg->SrcRegister.Index;
+
+   /* There is an extra source register that indirectly subscripts
+    * a register file. The direct index now becomes an offset
+    * that is being added to the indirect register.
+    *
+    *    file[ind[2].x+1],
+    *    where:
+    *       ind = SrcRegisterInd.File
+    *       [2] = SrcRegisterInd.Index
+    *       .x = SrcRegisterInd.SwizzleX
+    */
+   if (reg->SrcRegister.Indirect) {
+      union tgsi_exec_channel index2;
+      union tgsi_exec_channel indir_index;
+      const uint execmask = mach->ExecMask;
+      uint i;
+
+      /* which address register (always zero now) */
+      index2.i[0] =
+      index2.i[1] =
+      index2.i[2] =
+      index2.i[3] = reg->SrcRegisterInd.Index;
+
+      /* get current value of address register[swizzle] */
+      swizzle = tgsi_util_get_src_register_swizzle( &reg->SrcRegisterInd, CHAN_X );
+      fetch_src_file_channel(
+         mach,
+         reg->SrcRegisterInd.File,
+         swizzle,
+         &index2,
+         &indir_index );
+
+      /* add value of address register to the offset */
+      index.i[0] += (int) indir_index.f[0];
+      index.i[1] += (int) indir_index.f[1];
+      index.i[2] += (int) indir_index.f[2];
+      index.i[3] += (int) indir_index.f[3];
+
+      /* for disabled execution channels, zero-out the index to
+       * avoid using a potential garbage value.
+       */
+      for (i = 0; i < QUAD_SIZE; i++) {
+         if ((execmask & (1 << i)) == 0)
+            index.i[i] = 0;
+      }
+   }
+
+   /* There is an extra source register that is a second
+    * subscript to a register file. Effectively it means that
+    * the register file is actually a 2D array of registers.
+    *
+    *    file[1][3] == file[1*sizeof(file[1])+3],
+    *    where:
+    *       [3] = SrcRegisterDim.Index
+    */
+   if (reg->SrcRegister.Dimension) {
+      /* The size of the first-order array depends on the register file type.
+       * We need to multiply the index to the first array to get an effective,
+       * "flat" index that points to the beginning of the second-order array.
+       */
+      switch (reg->SrcRegister.File) {
+      case TGSI_FILE_INPUT:
+         index.i[0] *= TGSI_EXEC_MAX_INPUT_ATTRIBS;
+         index.i[1] *= TGSI_EXEC_MAX_INPUT_ATTRIBS;
+         index.i[2] *= TGSI_EXEC_MAX_INPUT_ATTRIBS;
+         index.i[3] *= TGSI_EXEC_MAX_INPUT_ATTRIBS;
+         break;
+      case TGSI_FILE_CONSTANT:
+         index.i[0] *= TGSI_EXEC_MAX_CONST_BUFFER;
+         index.i[1] *= TGSI_EXEC_MAX_CONST_BUFFER;
+         index.i[2] *= TGSI_EXEC_MAX_CONST_BUFFER;
+         index.i[3] *= TGSI_EXEC_MAX_CONST_BUFFER;
+         break;
+      default:
+         assert( 0 );
+      }
+
+      index.i[0] += reg->SrcRegisterDim.Index;
+      index.i[1] += reg->SrcRegisterDim.Index;
+      index.i[2] += reg->SrcRegisterDim.Index;
+      index.i[3] += reg->SrcRegisterDim.Index;
+
+      /* Again, the second subscript index can be addressed indirectly
+       * identically to the first one.
+       * Nothing stops us from indirectly addressing the indirect register,
+       * but there is no need for that, so we won't exercise it.
+       *
+       *    file[1][ind[4].y+3],
+       *    where:
+       *       ind = SrcRegisterDimInd.File
+       *       [4] = SrcRegisterDimInd.Index
+       *       .y = SrcRegisterDimInd.SwizzleX
+       */
+      if (reg->SrcRegisterDim.Indirect) {
+         union tgsi_exec_channel index2;
+         union tgsi_exec_channel indir_index;
+         const uint execmask = mach->ExecMask;
+         uint i;
+
+         index2.i[0] =
+         index2.i[1] =
+         index2.i[2] =
+         index2.i[3] = reg->SrcRegisterDimInd.Index;
+
+         swizzle = tgsi_util_get_src_register_swizzle( &reg->SrcRegisterDimInd, CHAN_X );
+         fetch_src_file_channel(
+            mach,
+            reg->SrcRegisterDimInd.File,
+            swizzle,
+            &index2,
+            &indir_index );
+
+         index.i[0] += (int) indir_index.f[0];
+         index.i[1] += (int) indir_index.f[1];
+         index.i[2] += (int) indir_index.f[2];
+         index.i[3] += (int) indir_index.f[3];
+
+         /* for disabled execution channels, zero-out the index to
+          * avoid using a potential garbage value.
+          */
+         for (i = 0; i < QUAD_SIZE; i++) {
+            if ((execmask & (1 << i)) == 0)
+               index.i[i] = 0;
+         }
+      }
+
+      /* If by any chance there was a need for a 3D array of register
+       * files, we would have to check whether SrcRegisterDim is followed
+       * by a dimension register and continue the saga.
+       */
+   }
+
+   swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index );
+   fetch_src_file_channel(
+      mach,
+      reg->SrcRegister.File,
+      swizzle,
+      &index,
+      chan );
+
+   switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) {
+   case TGSI_UTIL_SIGN_CLEAR:
+      micro_abs( chan, chan );
+      break;
+
+   case TGSI_UTIL_SIGN_SET:
+      micro_abs( chan, chan );
+      micro_neg( chan, chan );
+      break;
+
+   case TGSI_UTIL_SIGN_TOGGLE:
+      micro_neg( chan, chan );
+      break;
+
+   case TGSI_UTIL_SIGN_KEEP:
+      break;
+   }
+
+   if (reg->SrcRegisterExtMod.Complement) {
+      micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan );
+   }
+}
+
+static void
+store_dest(
+   struct tgsi_exec_machine *mach,
+   const union tgsi_exec_channel *chan,
+   const struct tgsi_full_dst_register *reg,
+   const struct tgsi_full_instruction *inst,
+   uint chan_index )
+{
+   uint i;
+   union tgsi_exec_channel null;
+   union tgsi_exec_channel *dst;
+   uint execmask = mach->ExecMask;
+
+   switch (reg->DstRegister.File) {
+   case TGSI_FILE_NULL:
+      dst = &null;
+      break;
+
+   case TGSI_FILE_OUTPUT:
+      dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0]
+                           + reg->DstRegister.Index].xyzw[chan_index];
+      break;
+
+   case TGSI_FILE_TEMPORARY:
+      assert( reg->DstRegister.Index < TGSI_EXEC_NUM_TEMPS );
+      dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index];
+      break;
+
+   case TGSI_FILE_ADDRESS:
+      dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index];
+      break;
+
+   default:
+      assert( 0 );
+      return;
+   }
+
+   if (inst->InstructionExtNv.CondFlowEnable) {
+      union tgsi_exec_channel *cc = &mach->Temps[TEMP_CC_I].xyzw[TEMP_CC_C];
+      uint swizzle;
+      uint shift;
+      uint mask;
+      uint test;
+
+      /* Only CC0 supported.
+       */
+      assert( inst->InstructionExtNv.CondFlowIndex < 1 );
+
+      switch (chan_index) {
+      case CHAN_X:
+         swizzle = inst->InstructionExtNv.CondSwizzleX;
+         break;
+      case CHAN_Y:
+         swizzle = inst->InstructionExtNv.CondSwizzleY;
+         break;
+      case CHAN_Z:
+         swizzle = inst->InstructionExtNv.CondSwizzleZ;
+         break;
+      case CHAN_W:
+         swizzle = inst->InstructionExtNv.CondSwizzleW;
+         break;
+      default:
+         assert( 0 );
+         return;
+      }
+
+      switch (swizzle) {
+      case TGSI_SWIZZLE_X:
+         shift = TGSI_EXEC_CC_X_SHIFT;
+         mask = TGSI_EXEC_CC_X_MASK;
+         break;
+      case TGSI_SWIZZLE_Y:
+         shift = TGSI_EXEC_CC_Y_SHIFT;
+         mask = TGSI_EXEC_CC_Y_MASK;
+         break;
+      case TGSI_SWIZZLE_Z:
+         shift = TGSI_EXEC_CC_Z_SHIFT;
+         mask = TGSI_EXEC_CC_Z_MASK;
+         break;
+      case TGSI_SWIZZLE_W:
+         shift = TGSI_EXEC_CC_W_SHIFT;
+         mask = TGSI_EXEC_CC_W_MASK;
+         break;
+      default:
+         assert( 0 );
+         return;
+      }
+
+      switch (inst->InstructionExtNv.CondMask) {
+      case TGSI_CC_GT:
+         test = ~(TGSI_EXEC_CC_GT << shift) & mask;
+         for (i = 0; i < QUAD_SIZE; i++)
+            if (cc->u[i] & test)
+               execmask &= ~(1 << i);
+         break;
+
+      case TGSI_CC_EQ:
+         test = ~(TGSI_EXEC_CC_EQ << shift) & mask;
+         for (i = 0; i < QUAD_SIZE; i++)
+            if (cc->u[i] & test)
+               execmask &= ~(1 << i);
+         break;
+
+      case TGSI_CC_LT:
+         test = ~(TGSI_EXEC_CC_LT << shift) & mask;
+         for (i = 0; i < QUAD_SIZE; i++)
+            if (cc->u[i] & test)
+               execmask &= ~(1 << i);
+         break;
+
+      case TGSI_CC_GE:
+         test = ~((TGSI_EXEC_CC_GT | TGSI_EXEC_CC_EQ) << shift) & mask;
+         for (i = 0; i < QUAD_SIZE; i++)
+            if (cc->u[i] & test)
+               execmask &= ~(1 << i);
+         break;
+
+      case TGSI_CC_LE:
+         test = ~((TGSI_EXEC_CC_LT | TGSI_EXEC_CC_EQ) << shift) & mask;
+         for (i = 0; i < QUAD_SIZE; i++)
+            if (cc->u[i] & test)
+               execmask &= ~(1 << i);
+         break;
+
+      case TGSI_CC_NE:
+         test = ~((TGSI_EXEC_CC_GT | TGSI_EXEC_CC_LT | TGSI_EXEC_CC_UN) << shift) & mask;
+         for (i = 0; i < QUAD_SIZE; i++)
+            if (cc->u[i] & test)
+               execmask &= ~(1 << i);
+         break;
+
+      case TGSI_CC_TR:
+         break;
+
+      case TGSI_CC_FL:
+         for (i = 0; i < QUAD_SIZE; i++)
+            execmask &= ~(1 << i);
+         break;
+
+      default:
+         assert( 0 );
+         return;
+      }
+   }
+
+   switch (inst->Instruction.Saturate) {
+   case TGSI_SAT_NONE:
+      for (i = 0; i < QUAD_SIZE; i++)
+         if (execmask & (1 << i))
+            dst->i[i] = chan->i[i];
+      break;
+
+   case TGSI_SAT_ZERO_ONE:
+      for (i = 0; i < QUAD_SIZE; i++)
+         if (execmask & (1 << i)) {
+            if (chan->f[i] < 0.0f)
+               dst->f[i] = 0.0f;
+            else if (chan->f[i] > 1.0f)
+               dst->f[i] = 1.0f;
+            else
+               dst->i[i] = chan->i[i];
+         }
+      break;
+
+   case TGSI_SAT_MINUS_PLUS_ONE:
+      for (i = 0; i < QUAD_SIZE; i++)
+         if (execmask & (1 << i)) {
+            if (chan->f[i] < -1.0f)
+               dst->f[i] = -1.0f;
+            else if (chan->f[i] > 1.0f)
+               dst->f[i] = 1.0f;
+            else
+               dst->i[i] = chan->i[i];
+         }
+      break;
+
+   default:
+      assert( 0 );
+   }
+
+   if (inst->InstructionExtNv.CondDstUpdate) {
+      union tgsi_exec_channel *cc = &mach->Temps[TEMP_CC_I].xyzw[TEMP_CC_C];
+      uint shift;
+      uint mask;
+
+      /* Only CC0 supported.
+       */
+      assert( inst->InstructionExtNv.CondDstIndex < 1 );
+
+      switch (chan_index) {
+      case CHAN_X:
+         shift = TGSI_EXEC_CC_X_SHIFT;
+         mask = ~TGSI_EXEC_CC_X_MASK;
+         break;
+      case CHAN_Y:
+         shift = TGSI_EXEC_CC_Y_SHIFT;
+         mask = ~TGSI_EXEC_CC_Y_MASK;
+         break;
+      case CHAN_Z:
+         shift = TGSI_EXEC_CC_Z_SHIFT;
+         mask = ~TGSI_EXEC_CC_Z_MASK;
+         break;
+      case CHAN_W:
+         shift = TGSI_EXEC_CC_W_SHIFT;
+         mask = ~TGSI_EXEC_CC_W_MASK;
+         break;
+      default:
+         assert( 0 );
+         return;
+      }
+
+      for (i = 0; i < QUAD_SIZE; i++)
+         if (execmask & (1 << i)) {
+            cc->u[i] &= mask;
+            if (dst->f[i] < 0.0f)
+               cc->u[i] |= TGSI_EXEC_CC_LT << shift;
+            else if (dst->f[i] > 0.0f)
+               cc->u[i] |= TGSI_EXEC_CC_GT << shift;
+            else if (dst->f[i] == 0.0f)
+               cc->u[i] |= TGSI_EXEC_CC_EQ << shift;
+            else
+               cc->u[i] |= TGSI_EXEC_CC_UN << shift;
+         }
+   }
+}
+
+#define FETCH(VAL,INDEX,CHAN)\
+    fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN)
+
+#define STORE(VAL,INDEX,CHAN)\
+    store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN )
+
+
+/**
+ * Execute ARB-style KIL which is predicated by a src register.
+ * Kill fragment if any of the four values is less than zero.
+ */
+static void
+exec_kil(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   uint uniquemask;
+   uint chan_index;
+   uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */
+   union tgsi_exec_channel r[1];
+
+   /* This mask stores component bits that were already tested. Note that
+    * we test if the value is less than zero, so 1.0 and 0.0 need not to be
+    * tested. */
+   uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE);
+
+   for (chan_index = 0; chan_index < 4; chan_index++)
+   {
+      uint swizzle;
+      uint i;
+
+      /* unswizzle channel */
+      swizzle = tgsi_util_get_full_src_register_extswizzle (
+                        &inst->FullSrcRegisters[0],
+                        chan_index);
+
+      /* check if the component has not been already tested */
+      if (uniquemask & (1 << swizzle))
+         continue;
+      uniquemask |= 1 << swizzle;
+
+      FETCH(&r[0], 0, chan_index);
+      for (i = 0; i < 4; i++)
+         if (r[0].f[i] < 0.0f)
+            kilmask |= 1 << i;
+   }
+
+   mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask;
+}
+
+/**
+ * Execute NVIDIA-style KIL which is predicated by a condition code.
+ * Kill fragment if the condition code is TRUE.
+ */
+static void
+exec_kilp(struct tgsi_exec_machine *mach,
+          const struct tgsi_full_instruction *inst)
+{
+   uint kilmask; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */
+
+   if (inst->InstructionExtNv.CondFlowEnable) {
+      uint swizzle[4];
+      uint chan_index;
+
+      kilmask = 0x0;
+
+      swizzle[0] = inst->InstructionExtNv.CondSwizzleX;
+      swizzle[1] = inst->InstructionExtNv.CondSwizzleY;
+      swizzle[2] = inst->InstructionExtNv.CondSwizzleZ;
+      swizzle[3] = inst->InstructionExtNv.CondSwizzleW;
+
+      for (chan_index = 0; chan_index < 4; chan_index++)
+      {
+         uint i;
+
+         for (i = 0; i < 4; i++) {
+            /* TODO: evaluate the condition code */
+            if (0)
+               kilmask |= 1 << i;
+         }
+      }
+   }
+   else {
+      /* "unconditional" kil */
+      kilmask = mach->ExecMask;
+   }
+   mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask;
+}
+
+
+/*
+ * Fetch a four texture samples using STR texture coordinates.
+ */
+static void
+fetch_texel( struct tgsi_sampler *sampler,
+             const union tgsi_exec_channel *s,
+             const union tgsi_exec_channel *t,
+             const union tgsi_exec_channel *p,
+             float lodbias,  /* XXX should be float[4] */
+             union tgsi_exec_channel *r,
+             union tgsi_exec_channel *g,
+             union tgsi_exec_channel *b,
+             union tgsi_exec_channel *a )
+{
+   uint j;
+   float rgba[NUM_CHANNELS][QUAD_SIZE];
+
+   sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba);
+
+   for (j = 0; j < 4; j++) {
+      r->f[j] = rgba[0][j];
+      g->f[j] = rgba[1][j];
+      b->f[j] = rgba[2][j];
+      a->f[j] = rgba[3][j];
+   }
+}
+
+
+static void
+exec_tex(struct tgsi_exec_machine *mach,
+         const struct tgsi_full_instruction *inst,
+         boolean biasLod,
+         boolean projected)
+{
+   const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index;
+   union tgsi_exec_channel r[4];
+   uint chan_index;
+   float lodBias;
+
+   /*   debug_printf("Sampler %u unit %u\n", sampler, unit); */
+
+   switch (inst->InstructionExtTexture.Texture) {
+   case TGSI_TEXTURE_1D:
+
+      FETCH(&r[0], 0, CHAN_X);
+
+      if (projected) {
+         FETCH(&r[1], 0, CHAN_W);
+         micro_div( &r[0], &r[0], &r[1] );
+      }
+
+      if (biasLod) {
+         FETCH(&r[1], 0, CHAN_W);
+         lodBias = r[2].f[0];
+      }
+      else
+         lodBias = 0.0;
+
+      fetch_texel(mach->Samplers[unit],
+                  &r[0], NULL, NULL, lodBias,  /* S, T, P, BIAS */
+                  &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */
+      break;
+
+   case TGSI_TEXTURE_2D:
+   case TGSI_TEXTURE_RECT:
+
+      FETCH(&r[0], 0, CHAN_X);
+      FETCH(&r[1], 0, CHAN_Y);
+      FETCH(&r[2], 0, CHAN_Z);
+
+      if (projected) {
+         FETCH(&r[3], 0, CHAN_W);
+         micro_div( &r[0], &r[0], &r[3] );
+         micro_div( &r[1], &r[1], &r[3] );
+         micro_div( &r[2], &r[2], &r[3] );
+      }
+
+      if (biasLod) {
+         FETCH(&r[3], 0, CHAN_W);
+         lodBias = r[3].f[0];
+      }
+      else
+         lodBias = 0.0;
+
+      fetch_texel(mach->Samplers[unit],
+                  &r[0], &r[1], &r[2], lodBias,  /* inputs */
+                  &r[0], &r[1], &r[2], &r[3]);  /* outputs */
+      break;
+
+   case TGSI_TEXTURE_3D:
+   case TGSI_TEXTURE_CUBE:
+
+      FETCH(&r[0], 0, CHAN_X);
+      FETCH(&r[1], 0, CHAN_Y);
+      FETCH(&r[2], 0, CHAN_Z);
+
+      if (projected) {
+         FETCH(&r[3], 0, CHAN_W);
+         micro_div( &r[0], &r[0], &r[3] );
+         micro_div( &r[1], &r[1], &r[3] );
+         micro_div( &r[2], &r[2], &r[3] );
+      }
+
+      if (biasLod) {
+         FETCH(&r[3], 0, CHAN_W);
+         lodBias = r[3].f[0];
+      }
+      else
+         lodBias = 0.0;
+
+      fetch_texel(mach->Samplers[unit],
+                  &r[0], &r[1], &r[2], lodBias,
+                  &r[0], &r[1], &r[2], &r[3]);
+      break;
+
+   default:
+      assert (0);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+      STORE( &r[chan_index], 0, chan_index );
+   }
+}
+
+
+/**
+ * Evaluate a constant-valued coefficient at the position of the
+ * current quad.
+ */
+static void
+eval_constant_coef(
+   struct tgsi_exec_machine *mach,
+   unsigned attrib,
+   unsigned chan )
+{
+   unsigned i;
+
+   for( i = 0; i < QUAD_SIZE; i++ ) {
+      mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan];
+   }
+}
+
+/**
+ * Evaluate a linear-valued coefficient at the position of the
+ * current quad.
+ */
+static void
+eval_linear_coef(
+   struct tgsi_exec_machine *mach,
+   unsigned attrib,
+   unsigned chan )
+{
+   const float x = mach->QuadPos.xyzw[0].f[0];
+   const float y = mach->QuadPos.xyzw[1].f[0];
+   const float dadx = mach->InterpCoefs[attrib].dadx[chan];
+   const float dady = mach->InterpCoefs[attrib].dady[chan];
+   const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y;
+   mach->Inputs[attrib].xyzw[chan].f[0] = a0;
+   mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx;
+   mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady;
+   mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady;
+}
+
+/**
+ * Evaluate a perspective-valued coefficient at the position of the
+ * current quad.
+ */
+static void
+eval_perspective_coef(
+   struct tgsi_exec_machine *mach,
+   unsigned attrib,
+   unsigned chan )
+{
+   const float x = mach->QuadPos.xyzw[0].f[0];
+   const float y = mach->QuadPos.xyzw[1].f[0];
+   const float dadx = mach->InterpCoefs[attrib].dadx[chan];
+   const float dady = mach->InterpCoefs[attrib].dady[chan];
+   const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y;
+   const float *w = mach->QuadPos.xyzw[3].f;
+   /* divide by W here */
+   mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0];
+   mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1];
+   mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2];
+   mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3];
+}
+
+
+typedef void (* eval_coef_func)(
+   struct tgsi_exec_machine *mach,
+   unsigned attrib,
+   unsigned chan );
+
+static void
+exec_declaration(
+   struct tgsi_exec_machine *mach,
+   const struct tgsi_full_declaration *decl )
+{
+   if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) {
+      if( decl->Declaration.File == TGSI_FILE_INPUT ) {
+         unsigned first, last, mask;
+         eval_coef_func eval;
+
+         first = decl->DeclarationRange.First;
+         last = decl->DeclarationRange.Last;
+         mask = decl->Declaration.UsageMask;
+
+         switch( decl->Declaration.Interpolate ) {
+         case TGSI_INTERPOLATE_CONSTANT:
+            eval = eval_constant_coef;
+            break;
+
+         case TGSI_INTERPOLATE_LINEAR:
+            eval = eval_linear_coef;
+            break;
+
+         case TGSI_INTERPOLATE_PERSPECTIVE:
+            eval = eval_perspective_coef;
+            break;
+
+         default:
+            eval = NULL;
+            assert( 0 );
+         }
+
+         if( mask == TGSI_WRITEMASK_XYZW ) {
+            unsigned i, j;
+
+            for( i = first; i <= last; i++ ) {
+               for( j = 0; j < NUM_CHANNELS; j++ ) {
+                  eval( mach, i, j );
+               }
+            }
+         }
+         else {
+            unsigned i, j;
+
+            for( j = 0; j < NUM_CHANNELS; j++ ) {
+               if( mask & (1 << j) ) {
+                  for( i = first; i <= last; i++ ) {
+                     eval( mach, i, j );
+                  }
+               }
+            }
+         }
+      }
+   }
+}
+
+static void
+exec_instruction(
+   struct tgsi_exec_machine *mach,
+   const struct tgsi_full_instruction *inst,
+   int *pc )
+{
+   uint chan_index;
+   union tgsi_exec_channel r[8];
+
+   (*pc)++;
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_ARL:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_flr( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_MOV:
+   case TGSI_OPCODE_SWZ:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_LIT:
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+	 STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X );
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+	 FETCH( &r[0], 0, CHAN_X );
+	 if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+	    micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
+	    STORE( &r[0], 0, CHAN_Y );
+	 }
+
+	 if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+	    FETCH( &r[1], 0, CHAN_Y );
+	    micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
+
+	    FETCH( &r[2], 0, CHAN_W );
+	    micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] );
+	    micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] );
+	    micro_pow( &r[1], &r[1], &r[2] );
+	    micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
+	    STORE( &r[0], 0, CHAN_Z );
+	 }
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
+	 STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_RCP:
+   /* TGSI_OPCODE_RECIP */
+      FETCH( &r[0], 0, CHAN_X );
+      micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] );
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_RSQ:
+   /* TGSI_OPCODE_RECIPSQRT */
+      FETCH( &r[0], 0, CHAN_X );
+      micro_sqrt( &r[0], &r[0] );
+      micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] );
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_EXP:
+      FETCH( &r[0], 0, CHAN_X );
+      micro_flr( &r[1], &r[0] );  /* r1 = floor(r0) */
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+         micro_exp2( &r[2], &r[1] );       /* r2 = 2 ^ r1 */
+         STORE( &r[2], 0, CHAN_X );        /* store r2 */
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+         micro_sub( &r[2], &r[0], &r[1] ); /* r2 = r0 - r1 */
+         STORE( &r[2], 0, CHAN_Y );        /* store r2 */
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+         micro_exp2( &r[2], &r[0] );       /* r2 = 2 ^ r0 */
+         STORE( &r[2], 0, CHAN_Z );        /* store r2 */
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
+         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_LOG:
+      FETCH( &r[0], 0, CHAN_X );
+      micro_abs( &r[2], &r[0] );  /* r2 = abs(r0) */
+      micro_lg2( &r[1], &r[2] );  /* r1 = lg2(r2) */
+      micro_flr( &r[0], &r[1] );  /* r0 = floor(r1) */
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+         STORE( &r[0], 0, CHAN_X );
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+         micro_exp2( &r[0], &r[0] );       /* r0 = 2 ^ r0 */
+         micro_div( &r[0], &r[2], &r[0] ); /* r0 = r2 / r0 */
+         STORE( &r[0], 0, CHAN_Y );
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+         STORE( &r[1], 0, CHAN_Z );
+      }
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
+         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_MUL:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index )
+      {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+
+         micro_mul( &r[0], &r[0], &r[1] );
+
+         STORE(&r[0], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_ADD:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_add( &r[0], &r[0], &r[1] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DP3:
+   /* TGSI_OPCODE_DOT3 */
+      FETCH( &r[0], 0, CHAN_X );
+      FETCH( &r[1], 1, CHAN_X );
+      micro_mul( &r[0], &r[0], &r[1] );
+
+      FETCH( &r[1], 0, CHAN_Y );
+      FETCH( &r[2], 1, CHAN_Y );
+      micro_mul( &r[1], &r[1], &r[2] );
+      micro_add( &r[0], &r[0], &r[1] );
+
+      FETCH( &r[1], 0, CHAN_Z );
+      FETCH( &r[2], 1, CHAN_Z );
+      micro_mul( &r[1], &r[1], &r[2] );
+      micro_add( &r[0], &r[0], &r[1] );
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+    case TGSI_OPCODE_DP4:
+    /* TGSI_OPCODE_DOT4 */
+       FETCH(&r[0], 0, CHAN_X);
+       FETCH(&r[1], 1, CHAN_X);
+
+       micro_mul( &r[0], &r[0], &r[1] );
+
+       FETCH(&r[1], 0, CHAN_Y);
+       FETCH(&r[2], 1, CHAN_Y);
+
+       micro_mul( &r[1], &r[1], &r[2] );
+       micro_add( &r[0], &r[0], &r[1] );
+
+       FETCH(&r[1], 0, CHAN_Z);
+       FETCH(&r[2], 1, CHAN_Z);
+
+       micro_mul( &r[1], &r[1], &r[2] );
+       micro_add( &r[0], &r[0], &r[1] );
+
+       FETCH(&r[1], 0, CHAN_W);
+       FETCH(&r[2], 1, CHAN_W);
+
+       micro_mul( &r[1], &r[1], &r[2] );
+       micro_add( &r[0], &r[0], &r[1] );
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DST:
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+	 STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X );
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+	 FETCH( &r[0], 0, CHAN_Y );
+	 FETCH( &r[1], 1, CHAN_Y);
+	 micro_mul( &r[0], &r[0], &r[1] );
+	 STORE( &r[0], 0, CHAN_Y );
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+	 FETCH( &r[0], 0, CHAN_Z );
+	 STORE( &r[0], 0, CHAN_Z );
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
+	 FETCH( &r[0], 1, CHAN_W );
+	 STORE( &r[0], 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_MIN:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+
+         /* XXX use micro_min()?? */
+         micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] );
+
+         STORE(&r[0], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_MAX:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+
+         /* XXX use micro_max()?? */
+         micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] );
+
+         STORE(&r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SLT:
+   /* TGSI_OPCODE_SETLT */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SGE:
+   /* TGSI_OPCODE_SETGE */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_MAD:
+   /* TGSI_OPCODE_MADD */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_mul( &r[0], &r[0], &r[1] );
+         FETCH( &r[1], 2, chan_index );
+         micro_add( &r[0], &r[0], &r[1] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SUB:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+
+         micro_sub( &r[0], &r[0], &r[1] );
+
+         STORE(&r[0], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_LERP:
+   /* TGSI_OPCODE_LRP */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+         FETCH(&r[2], 2, chan_index);
+
+         micro_sub( &r[1], &r[1], &r[2] );
+         micro_mul( &r[0], &r[0], &r[1] );
+         micro_add( &r[0], &r[0], &r[2] );
+
+         STORE(&r[0], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_CND:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_CND0:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_DOT2ADD:
+      /* TGSI_OPCODE_DP2A */
+      FETCH( &r[0], 0, CHAN_X );
+      FETCH( &r[1], 1, CHAN_X );
+      micro_mul( &r[0], &r[0], &r[1] );
+
+      FETCH( &r[1], 0, CHAN_Y );
+      FETCH( &r[2], 1, CHAN_Y );
+      micro_mul( &r[1], &r[1], &r[2] );
+      micro_add( &r[0], &r[0], &r[1] );
+
+      FETCH( &r[2], 2, CHAN_X );
+      micro_add( &r[0], &r[0], &r[2] );
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_INDEX:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_NEGATE:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_FRAC:
+   /* TGSI_OPCODE_FRC */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_frc( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_CLAMP:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_FLOOR:
+   /* TGSI_OPCODE_FLR */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_flr( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_ROUND:
+   case TGSI_OPCODE_ARR:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_rnd( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_EXPBASE2:
+    /* TGSI_OPCODE_EX2 */
+      FETCH(&r[0], 0, CHAN_X);
+
+#if FAST_MATH
+      micro_exp2( &r[0], &r[0] );
+#else
+      micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] );
+#endif
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_LOGBASE2:
+   /* TGSI_OPCODE_LG2 */
+      FETCH( &r[0], 0, CHAN_X );
+      micro_lg2( &r[0], &r[0] );
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_POWER:
+      /* TGSI_OPCODE_POW */
+      FETCH(&r[0], 0, CHAN_X);
+      FETCH(&r[1], 1, CHAN_X);
+
+      micro_pow( &r[0], &r[0], &r[1] );
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_CROSSPRODUCT:
+      /* TGSI_OPCODE_XPD */
+      FETCH(&r[0], 0, CHAN_Y);
+      FETCH(&r[1], 1, CHAN_Z);
+
+      micro_mul( &r[2], &r[0], &r[1] );
+
+      FETCH(&r[3], 0, CHAN_Z);
+      FETCH(&r[4], 1, CHAN_Y);
+
+      micro_mul( &r[5], &r[3], &r[4] );
+      micro_sub( &r[2], &r[2], &r[5] );
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+         STORE( &r[2], 0, CHAN_X );
+      }
+
+      FETCH(&r[2], 1, CHAN_X);
+
+      micro_mul( &r[3], &r[3], &r[2] );
+
+      FETCH(&r[5], 0, CHAN_X);
+
+      micro_mul( &r[1], &r[1], &r[5] );
+      micro_sub( &r[3], &r[3], &r[1] );
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+         STORE( &r[3], 0, CHAN_Y );
+      }
+
+      micro_mul( &r[5], &r[5], &r[4] );
+      micro_mul( &r[0], &r[0], &r[2] );
+      micro_sub( &r[5], &r[5], &r[0] );
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+         STORE( &r[5], 0, CHAN_Z );
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
+         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
+      }
+      break;
+
+    case TGSI_OPCODE_MULTIPLYMATRIX:
+       assert (0);
+       break;
+
+    case TGSI_OPCODE_ABS:
+       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+          FETCH(&r[0], 0, chan_index);
+
+          micro_abs( &r[0], &r[0] );
+
+          STORE(&r[0], 0, chan_index);
+       }
+       break;
+
+   case TGSI_OPCODE_RCC:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_DPH:
+      FETCH(&r[0], 0, CHAN_X);
+      FETCH(&r[1], 1, CHAN_X);
+
+      micro_mul( &r[0], &r[0], &r[1] );
+
+      FETCH(&r[1], 0, CHAN_Y);
+      FETCH(&r[2], 1, CHAN_Y);
+
+      micro_mul( &r[1], &r[1], &r[2] );
+      micro_add( &r[0], &r[0], &r[1] );
+
+      FETCH(&r[1], 0, CHAN_Z);
+      FETCH(&r[2], 1, CHAN_Z);
+
+      micro_mul( &r[1], &r[1], &r[2] );
+      micro_add( &r[0], &r[0], &r[1] );
+
+      FETCH(&r[1], 1, CHAN_W);
+
+      micro_add( &r[0], &r[0], &r[1] );
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_COS:
+      FETCH(&r[0], 0, CHAN_X);
+
+      micro_cos( &r[0], &r[0] );
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DDX:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_ddx( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DDY:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_ddy( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_KILP:
+      exec_kilp (mach, inst);
+      break;
+
+   case TGSI_OPCODE_KIL:
+      exec_kil (mach, inst);
+      break;
+
+   case TGSI_OPCODE_PK2H:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_PK2US:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_PK4B:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_PK4UB:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_RFL:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_SEQ:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_eq( &r[0], &r[0], &r[1],
+                   &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C],
+                   &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SFL:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_SGT:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SIN:
+      FETCH( &r[0], 0, CHAN_X );
+      micro_sin( &r[0], &r[0] );
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SLE:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SNE:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_STR:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_TEX:
+      /* simple texture lookup */
+      /* src[0] = texcoord */
+      /* src[1] = sampler unit */
+      exec_tex(mach, inst, FALSE, FALSE);
+      break;
+
+   case TGSI_OPCODE_TXB:
+      /* Texture lookup with lod bias */
+      /* src[0] = texcoord (src[0].w = LOD bias) */
+      /* src[1] = sampler unit */
+      exec_tex(mach, inst, TRUE, FALSE);
+      break;
+
+   case TGSI_OPCODE_TXD:
+      /* Texture lookup with explict partial derivatives */
+      /* src[0] = texcoord */
+      /* src[1] = d[strq]/dx */
+      /* src[2] = d[strq]/dy */
+      /* src[3] = sampler unit */
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_TXL:
+      /* Texture lookup with explit LOD */
+      /* src[0] = texcoord (src[0].w = LOD) */
+      /* src[1] = sampler unit */
+      exec_tex(mach, inst, TRUE, FALSE);
+      break;
+
+   case TGSI_OPCODE_TXP:
+      /* Texture lookup with projection */
+      /* src[0] = texcoord (src[0].w = projection) */
+      /* src[1] = sampler unit */
+      exec_tex(mach, inst, FALSE, TRUE);
+      break;
+
+   case TGSI_OPCODE_UP2H:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_UP2US:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_UP4B:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_UP4UB:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_X2D:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_ARA:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_BRA:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_CAL:
+      /* skip the call if no execution channels are enabled */
+      if (mach->ExecMask) {
+         /* do the call */
+
+         /* push the Cond, Loop, Cont stacks */
+         assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING);
+         mach->CondStack[mach->CondStackTop++] = mach->CondMask;
+         assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+         mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
+         assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+         mach->ContStack[mach->ContStackTop++] = mach->ContMask;
+
+         assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING);
+         mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask;
+
+         /* note that PC was already incremented above */
+         mach->CallStack[mach->CallStackTop++] = *pc;
+         *pc = inst->InstructionExtLabel.Label;
+      }
+      break;
+
+   case TGSI_OPCODE_RET:
+      mach->FuncMask &= ~mach->ExecMask;
+      UPDATE_EXEC_MASK(mach);
+
+      if (mach->FuncMask == 0x0) {
+         /* really return now (otherwise, keep executing */
+
+         if (mach->CallStackTop == 0) {
+            /* returning from main() */
+            *pc = -1;
+            return;
+         }
+         *pc = mach->CallStack[--mach->CallStackTop];
+
+         /* pop the Cond, Loop, Cont stacks */
+         assert(mach->CondStackTop > 0);
+         mach->CondMask = mach->CondStack[--mach->CondStackTop];
+         assert(mach->LoopStackTop > 0);
+         mach->LoopMask = mach->LoopStack[--mach->LoopStackTop];
+         assert(mach->ContStackTop > 0);
+         mach->ContMask = mach->ContStack[--mach->ContStackTop];
+         assert(mach->FuncStackTop > 0);
+         mach->FuncMask = mach->FuncStack[--mach->FuncStackTop];
+
+         UPDATE_EXEC_MASK(mach);
+      }
+      break;
+
+   case TGSI_OPCODE_SSG:
+   /* TGSI_OPCODE_SGN */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_sgn( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_CMP:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+         FETCH(&r[2], 2, chan_index);
+
+         micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] );
+
+         STORE(&r[0], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_SCS:
+      if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) {
+         FETCH( &r[0], 0, CHAN_X );
+      }
+      if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) {
+         micro_cos( &r[1], &r[0] );
+         STORE( &r[1], 0, CHAN_X );
+      }
+      if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) {
+         micro_sin( &r[1], &r[0] );
+         STORE( &r[1], 0, CHAN_Y );
+      }
+      if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) {
+         STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z );
+      }
+      if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) {
+         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_NRM:
+      /* 3-component vector normalize */
+      {
+         union tgsi_exec_channel tmp, dot;
+
+         /* tmp = dp3(src0, src0): */
+         FETCH( &r[0], 0, CHAN_X );
+         micro_mul( &tmp, &r[0], &r[0] );
+
+         FETCH( &r[1], 0, CHAN_Y );
+         micro_mul( &dot, &r[1], &r[1] );
+         micro_add( &tmp, &tmp, &dot );
+
+         FETCH( &r[2], 0, CHAN_Z );
+         micro_mul( &dot, &r[2], &r[2] );
+         micro_add( &tmp, &tmp, &dot );
+
+         /* tmp = 1 / sqrt(tmp) */
+         micro_sqrt( &tmp, &tmp );
+         micro_div( &tmp, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &tmp );
+
+         /* note: w channel is undefined */
+         FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+            /* chan = chan * tmp */
+            micro_mul( &r[chan_index], &tmp, &r[chan_index] );
+            STORE( &r[chan_index], 0, chan_index );
+         }
+      }
+      break;
+
+   case TGSI_OPCODE_NRM4:
+      /* 4-component vector normalize */
+      {
+         union tgsi_exec_channel tmp, dot;
+
+         /* tmp = dp4(src0, src0): */
+         FETCH( &r[0], 0, CHAN_X );
+         micro_mul( &tmp, &r[0], &r[0] );
+
+         FETCH( &r[1], 0, CHAN_Y );
+         micro_mul( &dot, &r[1], &r[1] );
+         micro_add( &tmp, &tmp, &dot );
+
+         FETCH( &r[2], 0, CHAN_Z );
+         micro_mul( &dot, &r[2], &r[2] );
+         micro_add( &tmp, &tmp, &dot );
+
+         FETCH( &r[3], 0, CHAN_W );
+         micro_mul( &dot, &r[3], &r[3] );
+         micro_add( &tmp, &tmp, &dot );
+
+         /* tmp = 1 / sqrt(tmp) */
+         micro_sqrt( &tmp, &tmp );
+         micro_div( &tmp, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &tmp );
+
+         FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+            /* chan = chan * tmp */
+            micro_mul( &r[chan_index], &tmp, &r[chan_index] );
+            STORE( &r[chan_index], 0, chan_index );
+         }
+      }
+      break;
+
+   case TGSI_OPCODE_DIV:
+      assert( 0 );
+      break;
+
+   case TGSI_OPCODE_DP2:
+      FETCH( &r[0], 0, CHAN_X );
+      FETCH( &r[1], 1, CHAN_X );
+      micro_mul( &r[0], &r[0], &r[1] );
+
+      FETCH( &r[1], 0, CHAN_Y );
+      FETCH( &r[2], 1, CHAN_Y );
+      micro_mul( &r[1], &r[1], &r[2] );
+      micro_add( &r[0], &r[0], &r[1] );
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_IF:
+      /* push CondMask */
+      assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING);
+      mach->CondStack[mach->CondStackTop++] = mach->CondMask;
+      FETCH( &r[0], 0, CHAN_X );
+      /* update CondMask */
+      if( ! r[0].u[0] ) {
+         mach->CondMask &= ~0x1;
+      }
+      if( ! r[0].u[1] ) {
+         mach->CondMask &= ~0x2;
+      }
+      if( ! r[0].u[2] ) {
+         mach->CondMask &= ~0x4;
+      }
+      if( ! r[0].u[3] ) {
+         mach->CondMask &= ~0x8;
+      }
+      UPDATE_EXEC_MASK(mach);
+      /* Todo: If CondMask==0, jump to ELSE */
+      break;
+
+   case TGSI_OPCODE_ELSE:
+      /* invert CondMask wrt previous mask */
+      {
+         uint prevMask;
+         assert(mach->CondStackTop > 0);
+         prevMask = mach->CondStack[mach->CondStackTop - 1];
+         mach->CondMask = ~mach->CondMask & prevMask;
+         UPDATE_EXEC_MASK(mach);
+         /* Todo: If CondMask==0, jump to ENDIF */
+      }
+      break;
+
+   case TGSI_OPCODE_ENDIF:
+      /* pop CondMask */
+      assert(mach->CondStackTop > 0);
+      mach->CondMask = mach->CondStack[--mach->CondStackTop];
+      UPDATE_EXEC_MASK(mach);
+      break;
+
+   case TGSI_OPCODE_END:
+      /* halt execution */
+      *pc = -1;
+      break;
+
+   case TGSI_OPCODE_REP:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_ENDREP:
+       assert (0);
+       break;
+
+   case TGSI_OPCODE_PUSHA:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_POPA:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_CEIL:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_ceil( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_I2F:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_i2f( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_NOT:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_not( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_TRUNC:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         micro_trunc( &r[0], &r[0] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SHL:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_shl( &r[0], &r[0], &r[1] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SHR:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_ishr( &r[0], &r[0], &r[1] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_AND:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_and( &r[0], &r[0], &r[1] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_OR:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_or( &r[0], &r[0], &r[1] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_MOD:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_XOR:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         micro_xor( &r[0], &r[0], &r[1] );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SAD:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_TXF:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_TXQ:
+      assert (0);
+      break;
+
+   case TGSI_OPCODE_EMIT:
+      mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16;
+      mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++;
+      break;
+
+   case TGSI_OPCODE_ENDPRIM:
+      mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++;
+      mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0;
+      break;
+
+   case TGSI_OPCODE_LOOP:
+      /* fall-through (for now) */
+   case TGSI_OPCODE_BGNLOOP2:
+      /* push LoopMask and ContMasks */
+      assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+      mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
+      assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+      mach->ContStack[mach->ContStackTop++] = mach->ContMask;
+      break;
+
+   case TGSI_OPCODE_ENDLOOP:
+      /* fall-through (for now at least) */
+   case TGSI_OPCODE_ENDLOOP2:
+      /* Restore ContMask, but don't pop */
+      assert(mach->ContStackTop > 0);
+      mach->ContMask = mach->ContStack[mach->ContStackTop - 1];
+      UPDATE_EXEC_MASK(mach);
+      if (mach->ExecMask) {
+         /* repeat loop: jump to instruction just past BGNLOOP */
+         *pc = inst->InstructionExtLabel.Label + 1;
+      }
+      else {
+         /* exit loop: pop LoopMask */
+         assert(mach->LoopStackTop > 0);
+         mach->LoopMask = mach->LoopStack[--mach->LoopStackTop];
+         /* pop ContMask */
+         assert(mach->ContStackTop > 0);
+         mach->ContMask = mach->ContStack[--mach->ContStackTop];
+      }
+      UPDATE_EXEC_MASK(mach);
+      break;
+
+   case TGSI_OPCODE_BRK:
+      /* turn off loop channels for each enabled exec channel */
+      mach->LoopMask &= ~mach->ExecMask;
+      /* Todo: if mach->LoopMask == 0, jump to end of loop */
+      UPDATE_EXEC_MASK(mach);
+      break;
+
+   case TGSI_OPCODE_CONT:
+      /* turn off cont channels for each enabled exec channel */
+      mach->ContMask &= ~mach->ExecMask;
+      /* Todo: if mach->LoopMask == 0, jump to end of loop */
+      UPDATE_EXEC_MASK(mach);
+      break;
+
+   case TGSI_OPCODE_BGNSUB:
+      /* no-op */
+      break;
+
+   case TGSI_OPCODE_ENDSUB:
+      /* no-op */
+      break;
+
+   case TGSI_OPCODE_NOISE1:
+      assert( 0 );
+      break;
+
+   case TGSI_OPCODE_NOISE2:
+      assert( 0 );
+      break;
+
+   case TGSI_OPCODE_NOISE3:
+      assert( 0 );
+      break;
+
+   case TGSI_OPCODE_NOISE4:
+      assert( 0 );
+      break;
+
+   case TGSI_OPCODE_NOP:
+      break;
+
+   default:
+      assert( 0 );
+   }
+}
+
+
+/**
+ * Run TGSI interpreter.
+ * \return bitmask of "alive" quad components
+ */
+uint
+tgsi_exec_machine_run( struct tgsi_exec_machine *mach )
+{
+   uint i;
+   int pc = 0;
+
+   mach->CondMask = 0xf;
+   mach->LoopMask = 0xf;
+   mach->ContMask = 0xf;
+   mach->FuncMask = 0xf;
+   mach->ExecMask = 0xf;
+
+   mach->CondStackTop = 0; /* temporarily subvert this assertion */
+   assert(mach->CondStackTop == 0);
+   assert(mach->LoopStackTop == 0);
+   assert(mach->ContStackTop == 0);
+   assert(mach->CallStackTop == 0);
+
+   mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0;
+   mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0;
+
+   if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) {
+      mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0;
+      mach->Primitives[0] = 0;
+   }
+
+   for (i = 0; i < QUAD_SIZE; i++) {
+      mach->Temps[TEMP_CC_I].xyzw[TEMP_CC_C].u[i] =
+         (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_X_SHIFT) |
+         (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_Y_SHIFT) |
+         (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_Z_SHIFT) |
+         (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_W_SHIFT);
+   }
+
+   /* execute declarations (interpolants) */
+   for (i = 0; i < mach->NumDeclarations; i++) {
+      exec_declaration( mach, mach->Declarations+i );
+   }
+
+   /* execute instructions, until pc is set to -1 */
+   while (pc != -1) {
+      assert(pc < (int) mach->NumInstructions);
+      exec_instruction( mach, mach->Instructions + pc, &pc );
+   }
+
+#if 0
+   /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */
+   if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) {
+      /*
+       * Scale back depth component.
+       */
+      for (i = 0; i < 4; i++)
+         mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF;
+   }
+#endif
+
+   return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0];
+}
+
+
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h
new file mode 100644
index 0000000000..4ffd4efbff
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h
@@ -0,0 +1,300 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_EXEC_H
+#define TGSI_EXEC_H
+
+#include "pipe/p_compiler.h"
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+#define MAX_LABELS 1024
+
+#define NUM_CHANNELS 4  /* R,G,B,A */
+#define QUAD_SIZE    4  /* 4 pixel/quad */
+
+/**
+  * Registers may be treated as float, signed int or unsigned int.
+  */
+union tgsi_exec_channel
+{
+   float    f[QUAD_SIZE];
+   int      i[QUAD_SIZE];
+   unsigned u[QUAD_SIZE];
+};
+
+/**
+  * A vector[RGBA] of channels[4 pixels]
+  */
+struct tgsi_exec_vector
+{
+   union tgsi_exec_channel xyzw[NUM_CHANNELS];
+};
+
+/**
+ * For fragment programs, information for computing fragment input
+ * values from plane equation of the triangle/line.
+ */
+struct tgsi_interp_coef
+{
+   float a0[NUM_CHANNELS];	/* in an xyzw layout */
+   float dadx[NUM_CHANNELS];
+   float dady[NUM_CHANNELS];
+};
+
+/**
+ * Information for sampling textures, which must be implemented
+ * by code outside the TGSI executor.
+ */
+struct tgsi_sampler
+{
+   /** Get samples for four fragments in a quad */
+   void (*get_samples)(struct tgsi_sampler *sampler,
+                       const float s[QUAD_SIZE],
+                       const float t[QUAD_SIZE],
+                       const float p[QUAD_SIZE],
+                       float lodbias,
+                       float rgba[NUM_CHANNELS][QUAD_SIZE]);
+};
+
+/**
+ * For branching/calling subroutines.
+ */
+struct tgsi_exec_labels
+{
+   unsigned labels[MAX_LABELS][2];
+   unsigned count;
+};
+
+
+#define TGSI_EXEC_NUM_TEMPS       128
+#define TGSI_EXEC_NUM_TEMP_EXTRAS   6
+#define TGSI_EXEC_NUM_IMMEDIATES  256
+
+/*
+ * Locations of various utility registers (_I = Index, _C = Channel)
+ */
+#define TGSI_EXEC_TEMP_00000000_I   (TGSI_EXEC_NUM_TEMPS + 0)
+#define TGSI_EXEC_TEMP_00000000_C   0
+
+#define TGSI_EXEC_TEMP_7FFFFFFF_I   (TGSI_EXEC_NUM_TEMPS + 0)
+#define TGSI_EXEC_TEMP_7FFFFFFF_C   1
+
+#define TGSI_EXEC_TEMP_80000000_I   (TGSI_EXEC_NUM_TEMPS + 0)
+#define TGSI_EXEC_TEMP_80000000_C   2
+
+#define TGSI_EXEC_TEMP_FFFFFFFF_I   (TGSI_EXEC_NUM_TEMPS + 0)
+#define TGSI_EXEC_TEMP_FFFFFFFF_C   3
+
+#define TGSI_EXEC_TEMP_ONE_I        (TGSI_EXEC_NUM_TEMPS + 1)
+#define TGSI_EXEC_TEMP_ONE_C        0
+
+#define TGSI_EXEC_TEMP_TWO_I        (TGSI_EXEC_NUM_TEMPS + 1)
+#define TGSI_EXEC_TEMP_TWO_C        1
+
+#define TGSI_EXEC_TEMP_128_I        (TGSI_EXEC_NUM_TEMPS + 1)
+#define TGSI_EXEC_TEMP_128_C        2
+
+#define TGSI_EXEC_TEMP_MINUS_128_I  (TGSI_EXEC_NUM_TEMPS + 1)
+#define TGSI_EXEC_TEMP_MINUS_128_C  3
+
+#define TGSI_EXEC_TEMP_KILMASK_I    (TGSI_EXEC_NUM_TEMPS + 2)
+#define TGSI_EXEC_TEMP_KILMASK_C    0
+
+#define TGSI_EXEC_TEMP_OUTPUT_I     (TGSI_EXEC_NUM_TEMPS + 2)
+#define TGSI_EXEC_TEMP_OUTPUT_C     1
+
+#define TGSI_EXEC_TEMP_PRIMITIVE_I  (TGSI_EXEC_NUM_TEMPS + 2)
+#define TGSI_EXEC_TEMP_PRIMITIVE_C  2
+
+/* NVIDIA condition code (CC) vector
+ */
+#define TGSI_EXEC_CC_GT       0x01
+#define TGSI_EXEC_CC_EQ       0x02
+#define TGSI_EXEC_CC_LT       0x04
+#define TGSI_EXEC_CC_UN       0x08
+
+#define TGSI_EXEC_CC_X_MASK   0x000000ff
+#define TGSI_EXEC_CC_X_SHIFT  0
+#define TGSI_EXEC_CC_Y_MASK   0x0000ff00
+#define TGSI_EXEC_CC_Y_SHIFT  8
+#define TGSI_EXEC_CC_Z_MASK   0x00ff0000
+#define TGSI_EXEC_CC_Z_SHIFT  16
+#define TGSI_EXEC_CC_W_MASK   0xff000000
+#define TGSI_EXEC_CC_W_SHIFT  24
+
+#define TGSI_EXEC_TEMP_CC_I         (TGSI_EXEC_NUM_TEMPS + 2)
+#define TGSI_EXEC_TEMP_CC_C         3
+
+#define TGSI_EXEC_TEMP_THREE_I      (TGSI_EXEC_NUM_TEMPS + 3)
+#define TGSI_EXEC_TEMP_THREE_C      0
+
+#define TGSI_EXEC_TEMP_HALF_I       (TGSI_EXEC_NUM_TEMPS + 3)
+#define TGSI_EXEC_TEMP_HALF_C       1
+
+/* execution mask, each value is either 0 or ~0 */
+#define TGSI_EXEC_MASK_I            (TGSI_EXEC_NUM_TEMPS + 3)
+#define TGSI_EXEC_MASK_C            2
+
+#define TGSI_EXEC_TEMP_R0           (TGSI_EXEC_NUM_TEMPS + 4)
+
+#define TGSI_EXEC_TEMP_ADDR         (TGSI_EXEC_NUM_TEMPS + 5)
+
+
+#define TGSI_EXEC_MAX_COND_NESTING  20
+#define TGSI_EXEC_MAX_LOOP_NESTING  20
+#define TGSI_EXEC_MAX_CALL_NESTING  20
+
+/* The maximum number of input attributes per vertex. For 2D
+ * input register files, this is the stride between two 1D
+ * arrays.
+ */
+#define TGSI_EXEC_MAX_INPUT_ATTRIBS 17
+
+/* The maximum number of constant vectors per constant buffer.
+ */
+#define TGSI_EXEC_MAX_CONST_BUFFER  4096
+
+/**
+ * Run-time virtual machine state for executing TGSI shader.
+ */
+struct tgsi_exec_machine
+{
+   /* Total = program temporaries + internal temporaries
+    *         + 1 padding to align to 16 bytes
+    */
+   struct tgsi_exec_vector       _Temps[TGSI_EXEC_NUM_TEMPS +
+                                        TGSI_EXEC_NUM_TEMP_EXTRAS + 1];
+
+   /*
+    * This will point to _Temps after aligning to 16B boundary.
+    */
+   struct tgsi_exec_vector       *Temps;
+   struct tgsi_exec_vector       *Addrs;
+
+   struct tgsi_sampler           **Samplers;
+
+   float                         Imms[TGSI_EXEC_NUM_IMMEDIATES][4];
+   unsigned                      ImmLimit;
+   const float                   (*Consts)[4];
+   struct tgsi_exec_vector       *Inputs;
+   struct tgsi_exec_vector       *Outputs;
+   const struct tgsi_token       *Tokens;
+   unsigned                      Processor;
+
+   /* GEOMETRY processor only. */
+   unsigned                      *Primitives;
+
+   /* FRAGMENT processor only. */
+   const struct tgsi_interp_coef *InterpCoefs;
+   struct tgsi_exec_vector       QuadPos;
+
+   /* Conditional execution masks */
+   uint CondMask;  /**< For IF/ELSE/ENDIF */
+   uint LoopMask;  /**< For BGNLOOP/ENDLOOP */
+   uint ContMask;  /**< For loop CONT statements */
+   uint FuncMask;  /**< For function calls */
+   uint ExecMask;  /**< = CondMask & LoopMask */
+
+   /** Condition mask stack (for nested conditionals) */
+   uint CondStack[TGSI_EXEC_MAX_COND_NESTING];
+   int CondStackTop;
+
+   /** Loop mask stack (for nested loops) */
+   uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING];
+   int LoopStackTop;
+
+   /** Loop continue mask stack (see comments in tgsi_exec.c) */
+   uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING];
+   int ContStackTop;
+
+   /** Function execution mask stack (for executing subroutine code) */
+   uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING];
+   int FuncStackTop;
+
+   /** Function call stack for saving/restoring the program counter */
+   uint CallStack[TGSI_EXEC_MAX_CALL_NESTING];
+   int CallStackTop;
+
+   struct tgsi_full_instruction *Instructions;
+   uint NumInstructions;
+
+   struct tgsi_full_declaration *Declarations;
+   uint NumDeclarations;
+
+   struct tgsi_exec_labels Labels;
+};
+
+void
+tgsi_exec_machine_init(
+   struct tgsi_exec_machine *mach );
+
+
+void 
+tgsi_exec_machine_bind_shader(
+   struct tgsi_exec_machine *mach,
+   const struct tgsi_token *tokens,
+   uint numSamplers,
+   struct tgsi_sampler **samplers);
+
+uint
+tgsi_exec_machine_run(
+   struct tgsi_exec_machine *mach );
+
+
+void
+tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach);
+
+
+static INLINE void
+tgsi_set_kill_mask(struct tgsi_exec_machine *mach, unsigned mask)
+{
+   mach->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0] =
+      mask;
+}
+
+
+/** Set execution mask values prior to executing the shader */
+static INLINE void
+tgsi_set_exec_mask(struct tgsi_exec_machine *mach,
+                   boolean ch0, boolean ch1, boolean ch2, boolean ch3)
+{
+   int *mask = mach->Temps[TGSI_EXEC_MASK_I].xyzw[TGSI_EXEC_MASK_C].i;
+   mask[0] = ch0 ? ~0 : 0;
+   mask[1] = ch1 ? ~0 : 0;
+   mask[2] = ch2 ? ~0 : 0;
+   mask[3] = ch3 ? ~0 : 0;
+}
+
+
+#if defined __cplusplus
+} /* extern "C" */
+#endif
+
+#endif /* TGSI_EXEC_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c
new file mode 100644
index 0000000000..68c7a6b7f5
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_info.c
@@ -0,0 +1,161 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_debug.h"
+#include "tgsi_info.h"
+
+static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] =
+{
+   { 1, 1, 0, 0, "ARL" },
+   { 1, 1, 0, 0, "MOV" },
+   { 1, 1, 0, 0, "LIT" },
+   { 1, 1, 0, 0, "RCP" },
+   { 1, 1, 0, 0, "RSQ" },
+   { 1, 1, 0, 0, "EXP" },
+   { 1, 1, 0, 0, "LOG" },
+   { 1, 2, 0, 0, "MUL" },
+   { 1, 2, 0, 0, "ADD" },
+   { 1, 2, 0, 0, "DP3" },
+   { 1, 2, 0, 0, "DP4" },
+   { 1, 2, 0, 0, "DST" },
+   { 1, 2, 0, 0, "MIN" },
+   { 1, 2, 0, 0, "MAX" },
+   { 1, 2, 0, 0, "SLT" },
+   { 1, 2, 0, 0, "SGE" },
+   { 1, 3, 0, 0, "MAD" },
+   { 1, 2, 0, 0, "SUB" },
+   { 1, 3, 0, 0, "LERP" },
+   { 1, 3, 0, 0, "CND" },
+   { 1, 3, 0, 0, "CND0" },
+   { 1, 3, 0, 0, "DOT2ADD" },
+   { 1, 2, 0, 0, "INDEX" },
+   { 1, 1, 0, 0, "NEGATE" },
+   { 1, 1, 0, 0, "FRAC" },
+   { 1, 3, 0, 0, "CLAMP" },
+   { 1, 1, 0, 0, "FLOOR" },
+   { 1, 1, 0, 0, "ROUND" },
+   { 1, 1, 0, 0, "EXPBASE2" },
+   { 1, 1, 0, 0, "LOGBASE2" },
+   { 1, 2, 0, 0, "POWER" },
+   { 1, 2, 0, 0, "CROSSPRODUCT" },
+   { 1, 2, 0, 0, "MULTIPLYMATRIX" },
+   { 1, 1, 0, 0, "ABS" },
+   { 1, 1, 0, 0, "RCC" },
+   { 1, 2, 0, 0, "DPH" },
+   { 1, 1, 0, 0, "COS" },
+   { 1, 1, 0, 0, "DDX" },
+   { 1, 1, 0, 0, "DDY" },
+   { 0, 0, 0, 0, "KILP" },
+   { 1, 1, 0, 0, "PK2H" },
+   { 1, 1, 0, 0, "PK2US" },
+   { 1, 1, 0, 0, "PK4B" },
+   { 1, 1, 0, 0, "PK4UB" },
+   { 1, 2, 0, 0, "RFL" },
+   { 1, 2, 0, 0, "SEQ" },
+   { 1, 2, 0, 0, "SFL" },
+   { 1, 2, 0, 0, "SGT" },
+   { 1, 1, 0, 0, "SIN" },
+   { 1, 2, 0, 0, "SLE" },
+   { 1, 2, 0, 0, "SNE" },
+   { 1, 2, 0, 0, "STR" },
+   { 1, 2, 1, 0, "TEX" },
+   { 1, 4, 1, 0, "TXD" },
+   { 1, 2, 1, 0, "TXP" },
+   { 1, 1, 0, 0, "UP2H" },
+   { 1, 1, 0, 0, "UP2US" },
+   { 1, 1, 0, 0, "UP4B" },
+   { 1, 1, 0, 0, "UP4UB" },
+   { 1, 3, 0, 0, "X2D" },
+   { 1, 1, 0, 0, "ARA" },
+   { 1, 1, 0, 0, "ARR" },
+   { 0, 1, 0, 0, "BRA" },
+   { 0, 0, 0, 1, "CAL" },
+   { 0, 0, 0, 0, "RET" },
+   { 1, 1, 0, 0, "SSG" },
+   { 1, 3, 0, 0, "CMP" },
+   { 1, 1, 0, 0, "SCS" },
+   { 1, 2, 1, 0, "TXB" },
+   { 1, 1, 0, 0, "NRM" },
+   { 1, 2, 0, 0, "DIV" },
+   { 1, 2, 0, 0, "DP2" },
+   { 1, 2, 1, 0, "TXL" },
+   { 0, 0, 0, 0, "BRK" },
+   { 0, 1, 0, 1, "IF" },
+   { 0, 0, 0, 0, "LOOP" },
+   { 0, 1, 0, 0, "REP" },
+   { 0, 0, 0, 1, "ELSE" },
+   { 0, 0, 0, 0, "ENDIF" },
+   { 0, 0, 0, 0, "ENDLOOP" },
+   { 0, 0, 0, 0, "ENDREP" },
+   { 0, 1, 0, 0, "PUSHA" },
+   { 1, 0, 0, 0, "POPA" },
+   { 1, 1, 0, 0, "CEIL" },
+   { 1, 1, 0, 0, "I2F" },
+   { 1, 1, 0, 0, "NOT" },
+   { 1, 1, 0, 0, "TRUNC" },
+   { 1, 2, 0, 0, "SHL" },
+   { 1, 2, 0, 0, "SHR" },
+   { 1, 2, 0, 0, "AND" },
+   { 1, 2, 0, 0, "OR" },
+   { 1, 2, 0, 0, "MOD" },
+   { 1, 2, 0, 0, "XOR" },
+   { 1, 3, 0, 0, "SAD" },
+   { 1, 2, 1, 0, "TXF" },
+   { 1, 2, 1, 0, "TXQ" },
+   { 0, 0, 0, 0, "CONT" },
+   { 0, 0, 0, 0, "EMIT" },
+   { 0, 0, 0, 0, "ENDPRIM" },
+   { 0, 0, 0, 1, "BGNLOOP2" },
+   { 0, 0, 0, 0, "BGNSUB" },
+   { 0, 0, 0, 1, "ENDLOOP2" },
+   { 0, 0, 0, 0, "ENDSUB" },
+   { 1, 1, 0, 0, "NOISE1" },
+   { 1, 1, 0, 0, "NOISE2" },
+   { 1, 1, 0, 0, "NOISE3" },
+   { 1, 1, 0, 0, "NOISE4" },
+   { 0, 0, 0, 0, "NOP" },
+   { 1, 2, 0, 0, "M4X3" },
+   { 1, 2, 0, 0, "M3X4" },
+   { 1, 2, 0, 0, "M3X3" },
+   { 1, 2, 0, 0, "M3X2" },
+   { 1, 1, 0, 0, "NRM4" },
+   { 0, 1, 0, 0, "CALLNZ" },
+   { 0, 1, 0, 0, "IFC" },
+   { 0, 1, 0, 0, "BREAKC" },
+   { 0, 1, 0, 0, "KIL" },
+   { 0, 0, 0, 0, "END" },
+   { 1, 1, 0, 0, "SWZ" }
+};
+
+const struct tgsi_opcode_info *
+tgsi_get_opcode_info( uint opcode )
+{
+   if (opcode < TGSI_OPCODE_LAST)
+      return &opcode_info[opcode];
+   assert( 0 );
+   return NULL;
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.h b/src/gallium/auxiliary/tgsi/tgsi_info.h
new file mode 100644
index 0000000000..7230bdaae3
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_info.h
@@ -0,0 +1,53 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_INFO_H
+#define TGSI_INFO_H
+
+#include "pipe/p_shader_tokens.h"
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+struct tgsi_opcode_info
+{
+   uint num_dst;
+   uint num_src;
+   boolean is_tex;
+   boolean is_branch;
+   const char *mnemonic;
+};
+
+const struct tgsi_opcode_info *
+tgsi_get_opcode_info( uint opcode );
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_INFO_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_iterate.c b/src/gallium/auxiliary/tgsi/tgsi_iterate.c
new file mode 100644
index 0000000000..5371a88b96
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_iterate.c
@@ -0,0 +1,85 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_debug.h"
+#include "tgsi_iterate.h"
+
+boolean
+tgsi_iterate_shader(
+   const struct tgsi_token *tokens,
+   struct tgsi_iterate_context *ctx )
+{
+   struct tgsi_parse_context parse;
+
+   if (tgsi_parse_init( &parse, tokens ) != TGSI_PARSE_OK)
+      return FALSE;
+
+   ctx->processor = parse.FullHeader.Processor;
+   ctx->version = parse.FullVersion.Version;
+
+   if (ctx->prolog)
+      if (!ctx->prolog( ctx ))
+         goto fail;
+
+   while (!tgsi_parse_end_of_tokens( &parse )) {
+      tgsi_parse_token( &parse );
+
+      switch (parse.FullToken.Token.Type) {
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         if (ctx->iterate_instruction)
+            if (!ctx->iterate_instruction( ctx, &parse.FullToken.FullInstruction ))
+               goto fail;
+         break;
+
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         if (ctx->iterate_declaration)
+            if (!ctx->iterate_declaration( ctx, &parse.FullToken.FullDeclaration ))
+               goto fail;
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         if (ctx->iterate_immediate)
+            if (!ctx->iterate_immediate( ctx, &parse.FullToken.FullImmediate ))
+               goto fail;
+         break;
+
+      default:
+         assert( 0 );
+      }
+   }
+
+   if (ctx->epilog)
+      if (!ctx->epilog( ctx ))
+         goto fail;
+
+   tgsi_parse_free( &parse );
+   return TRUE;
+
+fail:
+   tgsi_parse_free( &parse );
+   return FALSE;
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_iterate.h b/src/gallium/auxiliary/tgsi/tgsi_iterate.h
new file mode 100644
index 0000000000..ec7b85bf63
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_iterate.h
@@ -0,0 +1,76 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_ITERATE_H
+#define TGSI_ITERATE_H
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+struct tgsi_iterate_context
+{
+   boolean
+   (* prolog)(
+      struct tgsi_iterate_context *ctx );
+
+   boolean
+   (* iterate_instruction)(
+      struct tgsi_iterate_context *ctx,
+      struct tgsi_full_instruction *inst );
+
+   boolean
+   (* iterate_declaration)(
+      struct tgsi_iterate_context *ctx,
+      struct tgsi_full_declaration *decl );
+
+   boolean
+   (* iterate_immediate)(
+      struct tgsi_iterate_context *ctx,
+      struct tgsi_full_immediate *imm );
+
+   boolean
+   (* epilog)(
+      struct tgsi_iterate_context *ctx );
+
+   struct tgsi_processor processor;
+   struct tgsi_version version;
+};
+
+boolean
+tgsi_iterate_shader(
+   const struct tgsi_token *tokens,
+   struct tgsi_iterate_context *ctx );
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_ITERATE_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.c b/src/gallium/auxiliary/tgsi/tgsi_parse.c
new file mode 100644
index 0000000000..d374b16f9a
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c
@@ -0,0 +1,344 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_debug.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi_parse.h"
+#include "tgsi_build.h"
+#include "util/u_memory.h"
+
+void
+tgsi_full_token_init(
+   union tgsi_full_token *full_token )
+{
+   full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION;
+}
+
+void
+tgsi_full_token_free(
+   union tgsi_full_token *full_token )
+{
+   if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) {
+      FREE( (void *) full_token->FullImmediate.u.Pointer );
+   }
+}
+
+unsigned
+tgsi_parse_init(
+   struct tgsi_parse_context *ctx,
+   const struct tgsi_token *tokens )
+{
+   ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0];
+   if( ctx->FullVersion.Version.MajorVersion > 1 ) {
+      return TGSI_PARSE_ERROR;
+   }
+
+   ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1];
+   if( ctx->FullHeader.Header.HeaderSize >= 2 ) {
+      ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2];
+   }
+   else {
+      ctx->FullHeader.Processor = tgsi_default_processor();
+   }
+
+   ctx->Tokens = tokens;
+   ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize;
+
+   tgsi_full_token_init( &ctx->FullToken );
+
+   return TGSI_PARSE_OK;
+}
+
+void
+tgsi_parse_free(
+   struct tgsi_parse_context *ctx )
+{
+   tgsi_full_token_free( &ctx->FullToken );
+}
+
+boolean
+tgsi_parse_end_of_tokens(
+   struct tgsi_parse_context *ctx )
+{
+   return ctx->Position >=
+      1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize;
+}
+
+
+/**
+ * This function is used to avoid and work-around type punning/aliasing
+ * warnings.  The warnings seem harmless on x86 but on PPC they cause
+ * real failures.
+ */
+static INLINE void
+copy_token(void *dst, const void *src)
+{
+   memcpy(dst, src, 4);
+}
+
+
+/**
+ * Get next 4-byte token, return it at address specified by 'token'
+ */
+static void
+next_token(
+   struct tgsi_parse_context *ctx,
+   void *token )
+{
+   assert( !tgsi_parse_end_of_tokens( ctx ) );
+   copy_token(token, &ctx->Tokens[ctx->Position]);
+   ctx->Position++;
+}
+
+
+void
+tgsi_parse_token(
+   struct tgsi_parse_context *ctx )
+{
+   struct tgsi_token token;
+   unsigned i;
+
+   tgsi_full_token_free( &ctx->FullToken );
+   tgsi_full_token_init( &ctx->FullToken );
+
+   next_token( ctx, &token );
+
+   switch( token.Type ) {
+   case TGSI_TOKEN_TYPE_DECLARATION:
+   {
+      struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration;
+
+      *decl = tgsi_default_full_declaration();
+      copy_token(&decl->Declaration, &token);
+
+      next_token( ctx, &decl->DeclarationRange );
+
+      if( decl->Declaration.Semantic ) {
+         next_token( ctx, &decl->Semantic );
+      }
+
+      break;
+   }
+
+   case TGSI_TOKEN_TYPE_IMMEDIATE:
+   {
+      struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate;
+
+      *imm = tgsi_default_full_immediate();
+      copy_token(&imm->Immediate, &token);
+      assert( !imm->Immediate.Extended );
+
+      switch (imm->Immediate.DataType) {
+      case TGSI_IMM_FLOAT32:
+         imm->u.Pointer = MALLOC(
+            sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.NrTokens - 1) );
+         for( i = 0; i < imm->Immediate.NrTokens - 1; i++ ) {
+            next_token( ctx, (struct tgsi_immediate_float32 *) &imm->u.ImmediateFloat32[i] );
+         }
+         break;
+
+      default:
+         assert( 0 );
+      }
+
+      break;
+   }
+
+   case TGSI_TOKEN_TYPE_INSTRUCTION:
+   {
+      struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction;
+      unsigned extended;
+
+      *inst = tgsi_default_full_instruction();
+      copy_token(&inst->Instruction, &token);
+      extended = inst->Instruction.Extended;
+
+      while( extended ) {
+         struct tgsi_src_register_ext token;
+
+         next_token( ctx, &token );
+
+         switch( token.Type ) {
+         case TGSI_INSTRUCTION_EXT_TYPE_NV:
+            copy_token(&inst->InstructionExtNv, &token);
+            break;
+
+         case TGSI_INSTRUCTION_EXT_TYPE_LABEL:
+            copy_token(&inst->InstructionExtLabel, &token);
+            break;
+
+         case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE:
+            copy_token(&inst->InstructionExtTexture, &token);
+            break;
+
+         default:
+            assert( 0 );
+         }
+
+         extended = token.Extended;
+      }
+
+      assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS );
+
+      for(  i = 0; i < inst->Instruction.NumDstRegs; i++ ) {
+         unsigned extended;
+
+         next_token( ctx, &inst->FullDstRegisters[i].DstRegister );
+
+         /*
+          * No support for indirect or multi-dimensional addressing.
+          */
+         assert( !inst->FullDstRegisters[i].DstRegister.Indirect );
+         assert( !inst->FullDstRegisters[i].DstRegister.Dimension );
+
+         extended = inst->FullDstRegisters[i].DstRegister.Extended;
+
+         while( extended ) {
+            struct tgsi_src_register_ext token;
+
+            next_token( ctx, &token );
+
+            switch( token.Type ) {
+            case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE:
+               copy_token(&inst->FullDstRegisters[i].DstRegisterExtConcode,
+                          &token);
+               break;
+
+            case TGSI_DST_REGISTER_EXT_TYPE_MODULATE:
+               copy_token(&inst->FullDstRegisters[i].DstRegisterExtModulate,
+                          &token);
+               break;
+
+            default:
+               assert( 0 );
+            }
+
+            extended = token.Extended;
+         }
+      }
+
+      assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS );
+
+      for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) {
+         unsigned extended;
+
+         next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister );
+
+         extended = inst->FullSrcRegisters[i].SrcRegister.Extended;
+
+         while( extended ) {
+            struct tgsi_src_register_ext token;
+
+            next_token( ctx, &token );
+
+            switch( token.Type ) {
+            case TGSI_SRC_REGISTER_EXT_TYPE_SWZ:
+               copy_token(&inst->FullSrcRegisters[i].SrcRegisterExtSwz,
+                          &token);
+               break;
+
+            case TGSI_SRC_REGISTER_EXT_TYPE_MOD:
+               copy_token(&inst->FullSrcRegisters[i].SrcRegisterExtMod,
+                          &token);
+               break;
+
+            default:
+               assert( 0 );
+            }
+
+            extended = token.Extended;
+         }
+
+         if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) {
+            next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd );
+
+            /*
+             * No support for indirect or multi-dimensional addressing.
+             */
+            assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect );
+            assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension );
+            assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended );
+         }
+
+         if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) {
+            next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim );
+
+            /*
+             * No support for multi-dimensional addressing.
+             */
+            assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension );
+            assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended );
+
+            if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) {
+               next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd );
+
+               /*
+               * No support for indirect or multi-dimensional addressing.
+               */
+               assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect );
+               assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension );
+               assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended );
+            }
+         }
+      }
+
+      break;
+   }
+
+   default:
+      assert( 0 );
+   }
+}
+
+
+unsigned
+tgsi_num_tokens(const struct tgsi_token *tokens)
+{
+   struct tgsi_parse_context ctx;
+   if (tgsi_parse_init(&ctx, tokens) == TGSI_PARSE_OK) {
+      unsigned len = (ctx.FullHeader.Header.HeaderSize +
+                      ctx.FullHeader.Header.BodySize +
+                      1);
+      return len;
+   }
+   return 0;
+}
+
+
+/**
+ * Make a new copy of a token array.
+ */
+struct tgsi_token *
+tgsi_dup_tokens(const struct tgsi_token *tokens)
+{
+   unsigned n = tgsi_num_tokens(tokens);
+   unsigned bytes = n * sizeof(struct tgsi_token);
+   struct tgsi_token *new_tokens = (struct tgsi_token *) MALLOC(bytes);
+   if (new_tokens)
+      memcpy(new_tokens, tokens, bytes);
+   return new_tokens;
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.h b/src/gallium/auxiliary/tgsi/tgsi_parse.h
new file mode 100644
index 0000000000..054350712d
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_parse.h
@@ -0,0 +1,151 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_PARSE_H
+#define TGSI_PARSE_H
+
+#include "pipe/p_shader_tokens.h"
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+struct tgsi_full_version
+{
+   struct tgsi_version  Version;
+};
+
+struct tgsi_full_header
+{
+   struct tgsi_header      Header;
+   struct tgsi_processor   Processor;
+};
+
+struct tgsi_full_dst_register
+{
+   struct tgsi_dst_register               DstRegister;
+   struct tgsi_dst_register_ext_concode   DstRegisterExtConcode;
+   struct tgsi_dst_register_ext_modulate  DstRegisterExtModulate;
+};
+
+struct tgsi_full_src_register
+{
+   struct tgsi_src_register         SrcRegister;
+   struct tgsi_src_register_ext_swz SrcRegisterExtSwz;
+   struct tgsi_src_register_ext_mod SrcRegisterExtMod;
+   struct tgsi_src_register         SrcRegisterInd;
+   struct tgsi_dimension            SrcRegisterDim;
+   struct tgsi_src_register         SrcRegisterDimInd;
+};
+
+struct tgsi_full_declaration
+{
+   struct tgsi_declaration Declaration;
+   struct tgsi_declaration_range DeclarationRange;
+   struct tgsi_declaration_semantic Semantic;
+};
+
+struct tgsi_full_immediate
+{
+   struct tgsi_immediate   Immediate;
+   union
+   {
+      const void                          *Pointer;
+      const struct tgsi_immediate_float32 *ImmediateFloat32;
+   } u;
+};
+
+#define TGSI_FULL_MAX_DST_REGISTERS 2
+#define TGSI_FULL_MAX_SRC_REGISTERS 4 /* TXD has 4 */
+
+struct tgsi_full_instruction
+{
+   struct tgsi_instruction             Instruction;
+   struct tgsi_instruction_ext_nv      InstructionExtNv;
+   struct tgsi_instruction_ext_label   InstructionExtLabel;
+   struct tgsi_instruction_ext_texture InstructionExtTexture;
+   struct tgsi_full_dst_register       FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS];
+   struct tgsi_full_src_register       FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS];
+};
+
+union tgsi_full_token
+{
+   struct tgsi_token             Token;
+   struct tgsi_full_declaration  FullDeclaration;
+   struct tgsi_full_immediate    FullImmediate;
+   struct tgsi_full_instruction  FullInstruction;
+};
+
+void
+tgsi_full_token_init(
+   union tgsi_full_token *full_token );
+
+void
+tgsi_full_token_free(
+   union tgsi_full_token *full_token );
+
+struct tgsi_parse_context
+{
+   const struct tgsi_token    *Tokens;
+   unsigned                   Position;
+   struct tgsi_full_version   FullVersion;
+   struct tgsi_full_header    FullHeader;
+   union tgsi_full_token      FullToken;
+};
+
+#define TGSI_PARSE_OK      0
+#define TGSI_PARSE_ERROR   1
+
+unsigned
+tgsi_parse_init(
+   struct tgsi_parse_context *ctx,
+   const struct tgsi_token *tokens );
+
+void
+tgsi_parse_free(
+   struct tgsi_parse_context *ctx );
+
+boolean
+tgsi_parse_end_of_tokens(
+   struct tgsi_parse_context *ctx );
+
+void
+tgsi_parse_token(
+   struct tgsi_parse_context *ctx );
+
+unsigned
+tgsi_num_tokens(const struct tgsi_token *tokens);
+
+struct tgsi_token *
+tgsi_dup_tokens(const struct tgsi_token *tokens);
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_PARSE_H */
+
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ppc.c b/src/gallium/auxiliary/tgsi/tgsi_ppc.c
new file mode 100644
index 0000000000..f365030e52
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_ppc.c
@@ -0,0 +1,1363 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * TGSI to PowerPC code generation.
+ */
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_ARCH_PPC)
+
+#include "pipe/p_debug.h"
+#include "pipe/p_shader_tokens.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "util/u_sse.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi_dump.h"
+#include "tgsi_exec.h"
+#include "tgsi_ppc.h"
+#include "rtasm/rtasm_ppc.h"
+
+
+/**
+ * Since it's pretty much impossible to form PPC vector immediates, load
+ * them from memory here:
+ */
+const float ppc_builtin_constants[] ALIGN16_ATTRIB = {
+   1.0f, -128.0f, 128.0, 0.0
+};
+
+
+#define FOR_EACH_CHANNEL( CHAN )\
+   for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)
+
+#define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\
+   ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN)))
+
+#define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\
+   if (IS_DST0_CHANNEL_ENABLED( INST, CHAN ))
+
+#define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\
+   FOR_EACH_CHANNEL( CHAN )\
+      IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )
+
+#define CHAN_X 0
+#define CHAN_Y 1
+#define CHAN_Z 2
+#define CHAN_W 3
+
+
+/**
+ * How many TGSI temps should be implemented with real PPC vector registers
+ * rather than memory.
+ */
+#define MAX_PPC_TEMPS 3
+
+
+/**
+ * Context/state used during code gen.
+ */
+struct gen_context
+{
+   struct ppc_function *f;
+   int inputs_reg;    /**< GP register pointing to input params */
+   int outputs_reg;   /**< GP register pointing to output params */
+   int temps_reg;     /**< GP register pointing to temporary "registers" */
+   int immed_reg;     /**< GP register pointing to immediates buffer */
+   int const_reg;     /**< GP register pointing to constants buffer */
+   int builtins_reg;  /**< GP register pointint to built-in constants */
+
+   int offset_reg;    /**< used to reduce redundant li instructions */
+   int offset_value;
+
+   int one_vec;       /**< vector register with {1.0, 1.0, 1.0, 1.0} */
+   int bit31_vec;     /**< vector register with {1<<31, 1<<31, 1<<31, 1<<31} */
+
+   /**
+    * Map TGSI temps to PPC vector temps.
+    * We have 32 PPC vector regs.  Use 16 of them for storing 4 TGSI temps.
+    * XXX currently only do this for TGSI temps [0..MAX_PPC_TEMPS-1].
+    */
+   int temps_map[MAX_PPC_TEMPS][4];
+
+   /**
+    * Cache of src registers.
+    * This is used to avoid redundant load instructions.
+    */
+   struct {
+      struct tgsi_full_src_register src;
+      uint chan;
+      uint vec;
+   } regs[12];  /* 3 src regs, 4 channels */
+   uint num_regs;
+};
+
+
+/**
+ * Initialize code generation context.
+ */
+static void
+init_gen_context(struct gen_context *gen, struct ppc_function *func)
+{
+   uint i;
+
+   memset(gen, 0, sizeof(*gen));
+   gen->f = func;
+   gen->inputs_reg = ppc_reserve_register(func, 3);   /* first function param */
+   gen->outputs_reg = ppc_reserve_register(func, 4);  /* second function param */
+   gen->temps_reg = ppc_reserve_register(func, 5);    /* ... */
+   gen->immed_reg = ppc_reserve_register(func, 6);
+   gen->const_reg = ppc_reserve_register(func, 7);
+   gen->builtins_reg = ppc_reserve_register(func, 8);
+   gen->one_vec = -1;
+   gen->bit31_vec = -1;
+   gen->offset_reg = -1;
+   gen->offset_value = -9999999;
+   for (i = 0; i < MAX_PPC_TEMPS; i++) {
+      gen->temps_map[i][0] = ppc_allocate_vec_register(gen->f);
+      gen->temps_map[i][1] = ppc_allocate_vec_register(gen->f);
+      gen->temps_map[i][2] = ppc_allocate_vec_register(gen->f);
+      gen->temps_map[i][3] = ppc_allocate_vec_register(gen->f);
+   }
+}
+
+
+/**
+ * Is the given TGSI register stored as a real PPC vector register?
+ */
+static boolean
+is_ppc_vec_temporary(const struct tgsi_full_src_register *reg)
+{
+   return (reg->SrcRegister.File == TGSI_FILE_TEMPORARY &&
+           reg->SrcRegister.Index < MAX_PPC_TEMPS);
+}
+
+
+/**
+ * Is the given TGSI register stored as a real PPC vector register?
+ */
+static boolean
+is_ppc_vec_temporary_dst(const struct tgsi_full_dst_register *reg)
+{
+   return (reg->DstRegister.File == TGSI_FILE_TEMPORARY &&
+           reg->DstRegister.Index < MAX_PPC_TEMPS);
+}
+
+
+
+/**
+ * All PPC vector load/store instructions form an effective address
+ * by adding the contents of two registers.  For example:
+ *    lvx v2,r8,r9   # v2 = memory[r8 + r9]
+ *    stvx v2,r8,r9  # memory[r8 + r9] = v2;
+ * So our lvx/stvx instructions are typically preceded by an 'li' instruction
+ * to load r9 (above) with an immediate (an offset).
+ * This code emits that 'li' instruction, but only if the offset value is
+ * different than the previous 'li'.
+ * This optimization seems to save about 10% in the instruction count.
+ * Note that we need to unconditionally emit an 'li' inside basic blocks
+ * (such as inside loops).
+ */
+static int
+emit_li_offset(struct gen_context *gen, int offset)
+{
+   if (gen->offset_reg <= 0) {
+      /* allocate a GP register for storing load/store offset */
+      gen->offset_reg = ppc_allocate_register(gen->f);
+   }
+
+   /* emit new 'li' if offset is changing */
+   if (gen->offset_value < 0 || gen->offset_value != offset) {
+      gen->offset_value = offset;
+      ppc_li(gen->f, gen->offset_reg, offset);
+   }
+
+   return gen->offset_reg;
+}
+
+
+/**
+ * Forces subsequent emit_li_offset() calls to emit an 'li'.
+ * To be called at the top of basic blocks.
+ */
+static void
+reset_li_offset(struct gen_context *gen)
+{
+   gen->offset_value = -9999999;
+}
+
+
+
+/**
+ * Load the given vector register with {value, value, value, value}.
+ * The value must be in the ppu_builtin_constants[] array.
+ * We wouldn't need this if there was a simple way to load PPC vector
+ * registers with immediate values!
+ */
+static void
+load_constant_vec(struct gen_context *gen, int dst_vec, float value)
+{
+   uint pos;
+   for (pos = 0; pos < Elements(ppc_builtin_constants); pos++) {
+      if (ppc_builtin_constants[pos] == value) {
+         int offset = pos * 4;
+         int offset_reg = emit_li_offset(gen, offset);
+
+         /* Load 4-byte word into vector register.
+          * The vector slot depends on the effective address we load from.
+          * We know that our builtins start at a 16-byte boundary so we
+          * know that 'swizzle' tells us which vector slot will have the
+          * loaded word.  The other vector slots will be undefined.
+          */
+         ppc_lvewx(gen->f, dst_vec, gen->builtins_reg, offset_reg);
+         /* splat word[pos % 4] across the vector reg */
+         ppc_vspltw(gen->f, dst_vec, dst_vec, pos % 4);
+         return;
+      }
+   }
+   assert(0 && "Need to add new constant to ppc_builtin_constants array");
+}
+
+
+/**
+ * Return index of vector register containing {1.0, 1.0, 1.0, 1.0}.
+ */
+static int
+gen_one_vec(struct gen_context *gen)
+{
+   if (gen->one_vec < 0) {
+      gen->one_vec = ppc_allocate_vec_register(gen->f);
+      load_constant_vec(gen, gen->one_vec, 1.0f);
+   }
+   return gen->one_vec;
+}
+
+/**
+ * Return index of vector register containing {1<<31, 1<<31, 1<<31, 1<<31}.
+ */
+static int
+gen_get_bit31_vec(struct gen_context *gen)
+{
+   if (gen->bit31_vec < 0) {
+      gen->bit31_vec = ppc_allocate_vec_register(gen->f);
+      ppc_vspltisw(gen->f, gen->bit31_vec, -1);
+      ppc_vslw(gen->f, gen->bit31_vec, gen->bit31_vec, gen->bit31_vec);
+   }
+   return gen->bit31_vec;
+}
+
+
+/**
+ * Register fetch.  Return PPC vector register with result.
+ */
+static int
+emit_fetch(struct gen_context *gen,
+           const struct tgsi_full_src_register *reg,
+           const unsigned chan_index)
+{
+   uint swizzle = tgsi_util_get_full_src_register_extswizzle(reg, chan_index);
+   int dst_vec = -1;
+
+   switch (swizzle) {
+   case TGSI_EXTSWIZZLE_X:
+   case TGSI_EXTSWIZZLE_Y:
+   case TGSI_EXTSWIZZLE_Z:
+   case TGSI_EXTSWIZZLE_W:
+      switch (reg->SrcRegister.File) {
+      case TGSI_FILE_INPUT:
+         {
+            int offset = (reg->SrcRegister.Index * 4 + swizzle) * 16;
+            int offset_reg = emit_li_offset(gen, offset);
+            dst_vec = ppc_allocate_vec_register(gen->f);
+            ppc_lvx(gen->f, dst_vec, gen->inputs_reg, offset_reg);
+         }
+         break;
+      case TGSI_FILE_TEMPORARY:
+         if (is_ppc_vec_temporary(reg)) {
+            /* use PPC vec register */
+            dst_vec = gen->temps_map[reg->SrcRegister.Index][swizzle];
+         }
+         else {
+            /* use memory-based temp register "file" */
+            int offset = (reg->SrcRegister.Index * 4 + swizzle) * 16;
+            int offset_reg = emit_li_offset(gen, offset);
+            dst_vec = ppc_allocate_vec_register(gen->f);
+            ppc_lvx(gen->f, dst_vec, gen->temps_reg, offset_reg);
+         }
+         break;
+      case TGSI_FILE_IMMEDIATE:
+         {
+            int offset = (reg->SrcRegister.Index * 4 + swizzle) * 4;
+            int offset_reg = emit_li_offset(gen, offset);
+            dst_vec = ppc_allocate_vec_register(gen->f);
+            /* Load 4-byte word into vector register.
+             * The vector slot depends on the effective address we load from.
+             * We know that our immediates start at a 16-byte boundary so we
+             * know that 'swizzle' tells us which vector slot will have the
+             * loaded word.  The other vector slots will be undefined.
+             */
+            ppc_lvewx(gen->f, dst_vec, gen->immed_reg, offset_reg);
+            /* splat word[swizzle] across the vector reg */
+            ppc_vspltw(gen->f, dst_vec, dst_vec, swizzle);
+         }
+         break;
+      case TGSI_FILE_CONSTANT:
+         {
+            int offset = (reg->SrcRegister.Index * 4 + swizzle) * 4;
+            int offset_reg = emit_li_offset(gen, offset);
+            dst_vec = ppc_allocate_vec_register(gen->f);
+            /* Load 4-byte word into vector register.
+             * The vector slot depends on the effective address we load from.
+             * We know that our constants start at a 16-byte boundary so we
+             * know that 'swizzle' tells us which vector slot will have the
+             * loaded word.  The other vector slots will be undefined.
+             */
+            ppc_lvewx(gen->f, dst_vec, gen->const_reg, offset_reg);
+            /* splat word[swizzle] across the vector reg */
+            ppc_vspltw(gen->f, dst_vec, dst_vec, swizzle);
+         }
+         break;
+      default:
+         assert( 0 );
+      }
+      break;
+   case TGSI_EXTSWIZZLE_ZERO:
+      ppc_vzero(gen->f, dst_vec);
+      break;
+   case TGSI_EXTSWIZZLE_ONE:
+      {
+         int one_vec = gen_one_vec(gen);
+         dst_vec = ppc_allocate_vec_register(gen->f);
+         ppc_vmove(gen->f, dst_vec, one_vec);
+      }
+      break;
+   default:
+      assert( 0 );
+   }
+
+   assert(dst_vec >= 0);
+
+   {
+      uint sign_op = tgsi_util_get_full_src_register_sign_mode(reg, chan_index);
+      if (sign_op != TGSI_UTIL_SIGN_KEEP) {
+         int bit31_vec = gen_get_bit31_vec(gen);
+         int dst_vec2;
+
+         if (is_ppc_vec_temporary(reg)) {
+            /* need to use a new temp */
+            dst_vec2 = ppc_allocate_vec_register(gen->f);
+         }
+         else {
+            dst_vec2 = dst_vec;
+         }
+
+         switch (sign_op) {
+         case TGSI_UTIL_SIGN_CLEAR:
+            /* vec = vec & ~bit31 */
+            ppc_vandc(gen->f, dst_vec2, dst_vec, bit31_vec);
+            break;
+         case TGSI_UTIL_SIGN_SET:
+            /* vec = vec | bit31 */
+            ppc_vor(gen->f, dst_vec2, dst_vec, bit31_vec);
+            break;
+         case TGSI_UTIL_SIGN_TOGGLE:
+            /* vec = vec ^ bit31 */
+            ppc_vxor(gen->f, dst_vec2, dst_vec, bit31_vec);
+            break;
+         default:
+            assert(0);
+         }
+         return dst_vec2;
+      }
+   }
+
+   return dst_vec;
+}
+
+
+
+/**
+ * Test if two TGSI src registers refer to the same memory location.
+ * We use this to avoid redundant register loads.
+ */
+static boolean
+equal_src_locs(const struct tgsi_full_src_register *a, uint chan_a,
+               const struct tgsi_full_src_register *b, uint chan_b)
+{
+   int swz_a, swz_b;
+   int sign_a, sign_b;
+   if (a->SrcRegister.File != b->SrcRegister.File)
+      return FALSE;
+   if (a->SrcRegister.Index != b->SrcRegister.Index)
+      return FALSE;
+   swz_a = tgsi_util_get_full_src_register_extswizzle(a, chan_a);
+   swz_b = tgsi_util_get_full_src_register_extswizzle(b, chan_b);
+   if (swz_a != swz_b)
+      return FALSE;
+   sign_a = tgsi_util_get_full_src_register_sign_mode(a, chan_a);
+   sign_b = tgsi_util_get_full_src_register_sign_mode(b, chan_b);
+   if (sign_a != sign_b)
+      return FALSE;
+   return TRUE;
+}
+
+
+/**
+ * Given a TGSI src register and channel index, return the PPC vector
+ * register containing the value.  We use a cache to prevent re-loading
+ * the same register multiple times.
+ * \return index of PPC vector register with the desired src operand
+ */
+static int
+get_src_vec(struct gen_context *gen,
+            struct tgsi_full_instruction *inst, int src_reg, uint chan)
+{
+   const const struct tgsi_full_src_register *src = 
+      &inst->FullSrcRegisters[src_reg];
+   int vec;
+   uint i;
+
+   /* check the cache */
+   for (i = 0; i < gen->num_regs; i++) {
+      if (equal_src_locs(&gen->regs[i].src, gen->regs[i].chan, src, chan)) {
+         /* cache hit */
+         assert(gen->regs[i].vec >= 0);
+         return gen->regs[i].vec;
+      }
+   }
+
+   /* cache miss: allocate new vec reg and emit fetch/load code */
+   vec = emit_fetch(gen, src, chan);
+   gen->regs[gen->num_regs].src = *src;
+   gen->regs[gen->num_regs].chan = chan;
+   gen->regs[gen->num_regs].vec = vec;
+   gen->num_regs++;
+
+   assert(gen->num_regs <= Elements(gen->regs));
+
+   assert(vec >= 0);
+
+   return vec;
+}
+
+
+/**
+ * Clear the src operand cache.  To be called at the end of each emit function.
+ */
+static void
+release_src_vecs(struct gen_context *gen)
+{
+   uint i;
+   for (i = 0; i < gen->num_regs; i++) {
+      const const struct tgsi_full_src_register src = gen->regs[i].src;
+      if (!is_ppc_vec_temporary(&src)) {
+         ppc_release_vec_register(gen->f, gen->regs[i].vec);
+      }
+   }
+   gen->num_regs = 0;
+}
+
+
+
+static int
+get_dst_vec(struct gen_context *gen, 
+            const struct tgsi_full_instruction *inst,
+            unsigned chan_index)
+{
+   const struct tgsi_full_dst_register *reg = &inst->FullDstRegisters[0];
+
+   if (is_ppc_vec_temporary_dst(reg)) {
+      int vec = gen->temps_map[reg->DstRegister.Index][chan_index];
+      return vec;
+   }
+   else {
+      return ppc_allocate_vec_register(gen->f);
+   }
+}
+
+
+/**
+ * Register store.  Store 'src_vec' at location indicated by 'reg'.
+ * \param free_vec  Should the src_vec be released when done?
+ */
+static void
+emit_store(struct gen_context *gen,
+           int src_vec,
+           const struct tgsi_full_instruction *inst,
+           unsigned chan_index,
+           boolean free_vec)
+{
+   const struct tgsi_full_dst_register *reg = &inst->FullDstRegisters[0];
+
+   switch (reg->DstRegister.File) {
+   case TGSI_FILE_OUTPUT:
+      {
+         int offset = (reg->DstRegister.Index * 4 + chan_index) * 16;
+         int offset_reg = emit_li_offset(gen, offset);
+         ppc_stvx(gen->f, src_vec, gen->outputs_reg, offset_reg);
+      }
+      break;
+   case TGSI_FILE_TEMPORARY:
+      if (is_ppc_vec_temporary_dst(reg)) {
+         if (!free_vec) {
+            int dst_vec = gen->temps_map[reg->DstRegister.Index][chan_index];
+            if (dst_vec != src_vec)
+               ppc_vmove(gen->f, dst_vec, src_vec);
+         }
+         free_vec = FALSE;
+      }
+      else {
+         int offset = (reg->DstRegister.Index * 4 + chan_index) * 16;
+         int offset_reg = emit_li_offset(gen, offset);
+         ppc_stvx(gen->f, src_vec, gen->temps_reg, offset_reg);
+      }
+      break;
+#if 0
+   case TGSI_FILE_ADDRESS:
+      emit_addrs(
+         func,
+         xmm,
+         reg->DstRegister.Index,
+         chan_index );
+      break;
+#endif
+   default:
+      assert( 0 );
+   }
+
+#if 0
+   switch( inst->Instruction.Saturate ) {
+   case TGSI_SAT_NONE:
+      break;
+
+   case TGSI_SAT_ZERO_ONE:
+      /* assert( 0 ); */
+      break;
+
+   case TGSI_SAT_MINUS_PLUS_ONE:
+      assert( 0 );
+      break;
+   }
+#endif
+
+   if (free_vec)
+      ppc_release_vec_register(gen->f, src_vec);
+}
+
+
+static void
+emit_scalar_unaryop(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   int v0, v1;
+   uint chan_index;
+
+   v0 = get_src_vec(gen, inst, 0, CHAN_X);
+   v1 = ppc_allocate_vec_register(gen->f);
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_RSQ:
+      /* v1 = 1.0 / sqrt(v0) */
+      ppc_vrsqrtefp(gen->f, v1, v0);
+      break;
+   case TGSI_OPCODE_RCP:
+      /* v1 = 1.0 / v0 */
+      ppc_vrefp(gen->f, v1, v0);
+      break;
+   default:
+      assert(0);
+   }
+
+   FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+      emit_store(gen, v1, inst, chan_index, FALSE);
+   }
+
+   release_src_vecs(gen);
+   ppc_release_vec_register(gen->f, v1);
+}
+
+
+static void
+emit_unaryop(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   uint chan_index;
+
+   FOR_EACH_DST0_ENABLED_CHANNEL(*inst, chan_index) {
+      int v0 = get_src_vec(gen, inst, 0, chan_index);   /* v0 = srcreg[0] */
+      int v1 = get_dst_vec(gen, inst, chan_index);
+      switch (inst->Instruction.Opcode) {
+      case TGSI_OPCODE_ABS:
+         /* turn off the most significant bit of each vector float word */
+         {
+            int bit31_vec = gen_get_bit31_vec(gen);
+            ppc_vandc(gen->f, v1, v0, bit31_vec); /* v1 = v0 & ~bit31 */
+         }
+         break;
+      case TGSI_OPCODE_FLOOR:
+         ppc_vrfim(gen->f, v1, v0);         /* v1 = floor(v0) */
+         break;
+      case TGSI_OPCODE_FRAC:
+         ppc_vrfim(gen->f, v1, v0);      /* tmp = floor(v0) */
+         ppc_vsubfp(gen->f, v1, v0, v1); /* v1 = v0 - v1 */
+         break;
+      case TGSI_OPCODE_EXPBASE2:
+         ppc_vexptefp(gen->f, v1, v0);     /* v1 = 2^v0 */
+         break;
+      case TGSI_OPCODE_LOGBASE2:
+         /* XXX this may be broken! */
+         ppc_vlogefp(gen->f, v1, v0);      /* v1 = log2(v0) */
+         break;
+      case TGSI_OPCODE_MOV:
+      case TGSI_OPCODE_SWZ:
+         if (v0 != v1)
+            ppc_vmove(gen->f, v1, v0);
+         break;
+      default:
+         assert(0);
+      }
+      emit_store(gen, v1, inst, chan_index, TRUE);  /* store v0 */
+   }
+
+   release_src_vecs(gen);
+}
+
+
+static void
+emit_binop(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   int zero_vec = -1;
+   uint chan;
+
+   if (inst->Instruction.Opcode == TGSI_OPCODE_MUL) {
+      zero_vec = ppc_allocate_vec_register(gen->f);
+      ppc_vzero(gen->f, zero_vec);
+   }
+
+   FOR_EACH_DST0_ENABLED_CHANNEL(*inst, chan) {
+      /* fetch src operands */
+      int v0 = get_src_vec(gen, inst, 0, chan);
+      int v1 = get_src_vec(gen, inst, 1, chan);
+      int v2 = get_dst_vec(gen, inst, chan);
+
+      /* emit binop */
+      switch (inst->Instruction.Opcode) {
+      case TGSI_OPCODE_ADD:
+         ppc_vaddfp(gen->f, v2, v0, v1);
+         break;
+      case TGSI_OPCODE_SUB:
+         ppc_vsubfp(gen->f, v2, v0, v1);
+         break;
+      case TGSI_OPCODE_MUL:
+         ppc_vmaddfp(gen->f, v2, v0, v1, zero_vec);
+         break;
+      case TGSI_OPCODE_MIN:
+         ppc_vminfp(gen->f, v2, v0, v1);
+         break;
+      case TGSI_OPCODE_MAX:
+         ppc_vmaxfp(gen->f, v2, v0, v1);
+         break;
+      default:
+         assert(0);
+      }
+
+      /* store v2 */
+      emit_store(gen, v2, inst, chan, TRUE);
+   }
+
+   if (inst->Instruction.Opcode == TGSI_OPCODE_MUL)
+      ppc_release_vec_register(gen->f, zero_vec);
+
+   release_src_vecs(gen);
+}
+
+
+static void
+emit_triop(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   uint chan;
+
+   FOR_EACH_DST0_ENABLED_CHANNEL(*inst, chan) {
+      /* fetch src operands */
+      int v0 = get_src_vec(gen, inst, 0, chan);
+      int v1 = get_src_vec(gen, inst, 1, chan);
+      int v2 = get_src_vec(gen, inst, 2, chan);
+      int v3 = get_dst_vec(gen, inst, chan);
+
+      /* emit ALU */
+      switch (inst->Instruction.Opcode) {
+      case TGSI_OPCODE_MAD:
+         ppc_vmaddfp(gen->f, v3, v0, v1, v2);   /* v3 = v0 * v1 + v2 */
+         break;
+      case TGSI_OPCODE_LRP:
+         ppc_vsubfp(gen->f, v3, v1, v2);        /* v3 = v1 - v2 */
+         ppc_vmaddfp(gen->f, v3, v0, v3, v2);   /* v3 = v0 * v3 + v2 */
+         break;
+      default:
+         assert(0);
+      }
+
+      /* store v3 */
+      emit_store(gen, v3, inst, chan, TRUE);
+   }
+
+   release_src_vecs(gen);
+}
+
+
+/**
+ * Vector comparisons, resulting in 1.0 or 0.0 values.
+ */
+static void
+emit_inequality(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   uint chan;
+   int one_vec = gen_one_vec(gen);
+
+   FOR_EACH_DST0_ENABLED_CHANNEL(*inst, chan) {
+      /* fetch src operands */
+      int v0 = get_src_vec(gen, inst, 0, chan);
+      int v1 = get_src_vec(gen, inst, 1, chan);
+      int v2 = get_dst_vec(gen, inst, chan);
+      boolean complement = FALSE;
+
+      switch (inst->Instruction.Opcode) {
+      case TGSI_OPCODE_SNE:
+         complement = TRUE;
+         /* fall-through */
+      case TGSI_OPCODE_SEQ:
+         ppc_vcmpeqfpx(gen->f, v2, v0, v1); /* v2 = v0 == v1 ? ~0 : 0 */
+         break;
+
+      case TGSI_OPCODE_SGE:
+         complement = TRUE;
+         /* fall-through */
+      case TGSI_OPCODE_SLT:
+         ppc_vcmpgtfpx(gen->f, v2, v1, v0); /* v2 = v1 > v0 ? ~0 : 0 */
+         break;
+
+      case TGSI_OPCODE_SLE:
+         complement = TRUE;
+         /* fall-through */
+      case TGSI_OPCODE_SGT:
+         ppc_vcmpgtfpx(gen->f, v2, v0, v1); /* v2 = v0 > v1 ? ~0 : 0 */
+         break;
+      default:
+         assert(0);
+      }
+
+      /* v2 is now {0,0,0,0} or {~0,~0,~0,~0} */
+
+      if (complement)
+         ppc_vandc(gen->f, v2, one_vec, v2);    /* v2 = one_vec & ~v2 */
+      else
+         ppc_vand(gen->f, v2, one_vec, v2);     /* v2 = one_vec & v2 */
+
+      /* store v2 */
+      emit_store(gen, v2, inst, chan, TRUE);
+   }
+
+   release_src_vecs(gen);
+}
+
+
+static void
+emit_dotprod(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   int v0, v1, v2;
+   uint chan_index;
+
+   v2 = ppc_allocate_vec_register(gen->f);
+
+   ppc_vzero(gen->f, v2);                  /* v2 = {0, 0, 0, 0} */
+
+   v0 = get_src_vec(gen, inst, 0, CHAN_X); /* v0 = src0.XXXX */
+   v1 = get_src_vec(gen, inst, 1, CHAN_X); /* v1 = src1.XXXX */
+   ppc_vmaddfp(gen->f, v2, v0, v1, v2);    /* v2 = v0 * v1 + v2 */
+
+   v0 = get_src_vec(gen, inst, 0, CHAN_Y); /* v0 = src0.YYYY */
+   v1 = get_src_vec(gen, inst, 1, CHAN_Y); /* v1 = src1.YYYY */
+   ppc_vmaddfp(gen->f, v2, v0, v1, v2);    /* v2 = v0 * v1 + v2 */
+
+   v0 = get_src_vec(gen, inst, 0, CHAN_Z); /* v0 = src0.ZZZZ */
+   v1 = get_src_vec(gen, inst, 1, CHAN_Z); /* v1 = src1.ZZZZ */
+   ppc_vmaddfp(gen->f, v2, v0, v1, v2);    /* v2 = v0 * v1 + v2 */
+
+   if (inst->Instruction.Opcode == TGSI_OPCODE_DP4) {
+      v0 = get_src_vec(gen, inst, 0, CHAN_W); /* v0 = src0.WWWW */
+      v1 = get_src_vec(gen, inst, 1, CHAN_W); /* v1 = src1.WWWW */
+      ppc_vmaddfp(gen->f, v2, v0, v1, v2);    /* v2 = v0 * v1 + v2 */
+   }
+   else if (inst->Instruction.Opcode == TGSI_OPCODE_DPH) {
+      v1 = get_src_vec(gen, inst, 1, CHAN_W); /* v1 = src1.WWWW */
+      ppc_vaddfp(gen->f, v2, v2, v1);         /* v2 = v2 + v1 */
+   }
+
+   FOR_EACH_DST0_ENABLED_CHANNEL(*inst, chan_index) {
+      emit_store(gen, v2, inst, chan_index, FALSE);  /* store v2, free v2 later */
+   }
+
+   release_src_vecs(gen);
+
+   ppc_release_vec_register(gen->f, v2);
+}
+
+
+/** Approximation for vr = pow(va, vb) */
+static void
+ppc_vec_pow(struct ppc_function *f, int vr, int va, int vb)
+{
+   /* pow(a,b) ~= exp2(log2(a) * b) */
+   int t_vec = ppc_allocate_vec_register(f);
+   int zero_vec = ppc_allocate_vec_register(f);
+
+   ppc_vzero(f, zero_vec);
+
+   ppc_vlogefp(f, t_vec, va);                   /* t = log2(va) */
+   ppc_vmaddfp(f, t_vec, t_vec, vb, zero_vec);  /* t = t * vb + zero */
+   ppc_vexptefp(f, vr, t_vec);                  /* vr = 2^t */
+
+   ppc_release_vec_register(f, t_vec);
+   ppc_release_vec_register(f, zero_vec);
+}
+
+
+static void
+emit_lit(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   int one_vec = gen_one_vec(gen);
+
+   /* Compute X */
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X)) {
+      emit_store(gen, one_vec, inst, CHAN_X, FALSE);
+   }
+
+   /* Compute Y, Z */
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y) ||
+       IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Z)) {
+      int x_vec;
+      int zero_vec = ppc_allocate_vec_register(gen->f);
+
+      x_vec = get_src_vec(gen, inst, 0, CHAN_X);  /* x_vec = src[0].x */
+
+      ppc_vzero(gen->f, zero_vec);                /* zero = {0,0,0,0} */
+      ppc_vmaxfp(gen->f, x_vec, x_vec, zero_vec); /* x_vec = max(x_vec, 0) */
+
+      if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y)) {
+         emit_store(gen, x_vec, inst, CHAN_Y, FALSE);
+      }
+
+      if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Z)) {
+         int y_vec, w_vec;
+         int z_vec = ppc_allocate_vec_register(gen->f);
+         int pow_vec = ppc_allocate_vec_register(gen->f);
+         int pos_vec = ppc_allocate_vec_register(gen->f);
+         int p128_vec = ppc_allocate_vec_register(gen->f);
+         int n128_vec = ppc_allocate_vec_register(gen->f);
+
+         y_vec = get_src_vec(gen, inst, 0, CHAN_Y);  /* y_vec = src[0].y */
+         ppc_vmaxfp(gen->f, y_vec, y_vec, zero_vec); /* y_vec = max(y_vec, 0) */
+
+         w_vec = get_src_vec(gen, inst, 0, CHAN_W);  /* w_vec = src[0].w */
+
+         /* clamp W to [-128, 128] */
+         load_constant_vec(gen, p128_vec, 128.0f);
+         load_constant_vec(gen, n128_vec, -128.0f);
+         ppc_vmaxfp(gen->f, w_vec, w_vec, n128_vec); /* w = max(w, -128) */
+         ppc_vminfp(gen->f, w_vec, w_vec, p128_vec); /* w = min(w, 128) */
+
+         /* if temp.x > 0
+          *    z = pow(tmp.y, tmp.w)
+          * else
+          *    z = 0.0
+          */
+         ppc_vec_pow(gen->f, pow_vec, y_vec, w_vec);      /* pow = pow(y, w) */
+         ppc_vcmpgtfpx(gen->f, pos_vec, x_vec, zero_vec); /* pos = x > 0 */
+         ppc_vand(gen->f, z_vec, pow_vec, pos_vec);       /* z = pow & pos */
+
+         emit_store(gen, z_vec, inst, CHAN_Z, FALSE);
+
+         ppc_release_vec_register(gen->f, z_vec);
+         ppc_release_vec_register(gen->f, pow_vec);
+         ppc_release_vec_register(gen->f, pos_vec);
+         ppc_release_vec_register(gen->f, p128_vec);
+         ppc_release_vec_register(gen->f, n128_vec);
+      }
+
+      ppc_release_vec_register(gen->f, zero_vec);
+   }
+
+   /* Compute W */
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_W)) {
+      emit_store(gen, one_vec, inst, CHAN_W, FALSE);
+   }
+
+   release_src_vecs(gen);
+}
+
+
+static void
+emit_exp(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   const int one_vec = gen_one_vec(gen);
+   int src_vec;
+
+   /* get src arg */
+   src_vec = get_src_vec(gen, inst, 0, CHAN_X);
+
+   /* Compute X = 2^floor(src) */
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X)) {
+      int dst_vec = get_dst_vec(gen, inst, CHAN_X);
+      int tmp_vec = ppc_allocate_vec_register(gen->f);
+      ppc_vrfim(gen->f, tmp_vec, src_vec);             /* tmp = floor(src); */
+      ppc_vexptefp(gen->f, dst_vec, tmp_vec);          /* dst = 2 ^ tmp */
+      emit_store(gen, dst_vec, inst, CHAN_X, TRUE);
+      ppc_release_vec_register(gen->f, tmp_vec);
+   }
+
+   /* Compute Y = src - floor(src) */
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y)) {
+      int dst_vec = get_dst_vec(gen, inst, CHAN_Y);
+      int tmp_vec = ppc_allocate_vec_register(gen->f);
+      ppc_vrfim(gen->f, tmp_vec, src_vec);             /* tmp = floor(src); */
+      ppc_vsubfp(gen->f, dst_vec, src_vec, tmp_vec);   /* dst = src - tmp */
+      emit_store(gen, dst_vec, inst, CHAN_Y, TRUE);
+      ppc_release_vec_register(gen->f, tmp_vec);
+   }
+
+   /* Compute Z = RoughApprox2ToX(src) */
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Z)) {
+      int dst_vec = get_dst_vec(gen, inst, CHAN_Z);
+      ppc_vexptefp(gen->f, dst_vec, src_vec);          /* dst = 2 ^ src */
+      emit_store(gen, dst_vec, inst, CHAN_Z, TRUE);
+   }
+
+   /* Compute W = 1.0 */
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_W)) {
+      emit_store(gen, one_vec, inst, CHAN_W, FALSE);
+   }
+
+   release_src_vecs(gen);
+}
+
+
+static void
+emit_log(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   const int bit31_vec = gen_get_bit31_vec(gen);
+   const int one_vec = gen_one_vec(gen);
+   int src_vec, abs_vec;
+
+   /* get src arg */
+   src_vec = get_src_vec(gen, inst, 0, CHAN_X);
+
+   /* compute abs(src) */
+   abs_vec = ppc_allocate_vec_register(gen->f);
+   ppc_vandc(gen->f, abs_vec, src_vec, bit31_vec);     /* abs = src & ~bit31 */
+
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X) &&
+       IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y)) {
+
+      /* compute tmp = floor(log2(abs)) */
+      int tmp_vec = ppc_allocate_vec_register(gen->f);
+      ppc_vlogefp(gen->f, tmp_vec, abs_vec);           /* tmp = log2(abs) */
+      ppc_vrfim(gen->f, tmp_vec, tmp_vec);             /* tmp = floor(tmp); */
+
+      /* Compute X = tmp */
+      if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X)) {
+         emit_store(gen, tmp_vec, inst, CHAN_X, FALSE);
+      }
+      
+      /* Compute Y = abs / 2^tmp */
+      if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y)) {
+         const int zero_vec = ppc_allocate_vec_register(gen->f);
+         ppc_vzero(gen->f, zero_vec);
+         ppc_vexptefp(gen->f, tmp_vec, tmp_vec);       /* tmp = 2 ^ tmp */
+         ppc_vrefp(gen->f, tmp_vec, tmp_vec);          /* tmp = 1 / tmp */
+         /* tmp = abs * tmp + zero */
+         ppc_vmaddfp(gen->f, tmp_vec, abs_vec, tmp_vec, zero_vec);
+         emit_store(gen, tmp_vec, inst, CHAN_Y, FALSE);
+         ppc_release_vec_register(gen->f, zero_vec);
+      }
+
+      ppc_release_vec_register(gen->f, tmp_vec);
+   }
+
+   /* Compute Z = RoughApproxLog2(abs) */
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Z)) {
+      int dst_vec = get_dst_vec(gen, inst, CHAN_Z);
+      ppc_vlogefp(gen->f, dst_vec, abs_vec);           /* dst = log2(abs) */
+      emit_store(gen, dst_vec, inst, CHAN_Z, TRUE);
+   }
+
+   /* Compute W = 1.0 */
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_W)) {
+      emit_store(gen, one_vec, inst, CHAN_W, FALSE);
+   }
+
+   ppc_release_vec_register(gen->f, abs_vec);
+   release_src_vecs(gen);
+}
+
+
+static void
+emit_pow(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   int s0_vec = get_src_vec(gen, inst, 0, CHAN_X);
+   int s1_vec = get_src_vec(gen, inst, 1, CHAN_X);
+   int pow_vec = ppc_allocate_vec_register(gen->f);
+   int chan;
+
+   ppc_vec_pow(gen->f, pow_vec, s0_vec, s1_vec);
+
+   FOR_EACH_DST0_ENABLED_CHANNEL(*inst, chan) {
+      emit_store(gen, pow_vec, inst, chan, FALSE);
+   }
+
+   ppc_release_vec_register(gen->f, pow_vec);
+
+   release_src_vecs(gen);
+}
+
+
+static void
+emit_xpd(struct gen_context *gen, struct tgsi_full_instruction *inst)
+{
+   int x0_vec, y0_vec, z0_vec;
+   int x1_vec, y1_vec, z1_vec;
+   int zero_vec, tmp_vec;
+   int tmp2_vec;
+
+   zero_vec = ppc_allocate_vec_register(gen->f);
+   ppc_vzero(gen->f, zero_vec);
+
+   tmp_vec = ppc_allocate_vec_register(gen->f);
+   tmp2_vec = ppc_allocate_vec_register(gen->f);
+
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y) ||
+       IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Z)) {
+      x0_vec = get_src_vec(gen, inst, 0, CHAN_X);
+      x1_vec = get_src_vec(gen, inst, 1, CHAN_X);
+   }
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X) ||
+       IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Z)) {
+      y0_vec = get_src_vec(gen, inst, 0, CHAN_Y);
+      y1_vec = get_src_vec(gen, inst, 1, CHAN_Y);
+   }
+   if (IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X) ||
+       IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y)) {
+      z0_vec = get_src_vec(gen, inst, 0, CHAN_Z);
+      z1_vec = get_src_vec(gen, inst, 1, CHAN_Z);
+   }
+
+   IF_IS_DST0_CHANNEL_ENABLED(*inst, CHAN_X) {
+      /* tmp = y0 * z1 */
+      ppc_vmaddfp(gen->f, tmp_vec, y0_vec, z1_vec, zero_vec);
+      /* tmp = tmp - z0 * y1*/
+      ppc_vnmsubfp(gen->f, tmp_vec, tmp_vec, z0_vec, y1_vec);
+      emit_store(gen, tmp_vec, inst, CHAN_X, FALSE);
+   }
+   IF_IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Y) {
+      /* tmp = z0 * x1 */
+      ppc_vmaddfp(gen->f, tmp_vec, z0_vec, x1_vec, zero_vec);
+      /* tmp = tmp - x0 * z1 */
+      ppc_vnmsubfp(gen->f, tmp_vec, tmp_vec, x0_vec, z1_vec);
+      emit_store(gen, tmp_vec, inst, CHAN_Y, FALSE);
+   }
+   IF_IS_DST0_CHANNEL_ENABLED(*inst, CHAN_Z) {
+      /* tmp = x0 * y1 */
+      ppc_vmaddfp(gen->f, tmp_vec, x0_vec, y1_vec, zero_vec);
+      /* tmp = tmp - y0 * x1 */
+      ppc_vnmsubfp(gen->f, tmp_vec, tmp_vec, y0_vec, x1_vec);
+      emit_store(gen, tmp_vec, inst, CHAN_Z, FALSE);
+   }
+   /* W is undefined */
+
+   ppc_release_vec_register(gen->f, tmp_vec);
+   ppc_release_vec_register(gen->f, zero_vec);
+   release_src_vecs(gen);
+}
+
+static int
+emit_instruction(struct gen_context *gen,
+                 struct tgsi_full_instruction *inst)
+{
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_MOV:
+   case TGSI_OPCODE_SWZ:
+   case TGSI_OPCODE_ABS:
+   case TGSI_OPCODE_FLOOR:
+   case TGSI_OPCODE_FRAC:
+   case TGSI_OPCODE_EXPBASE2:
+   case TGSI_OPCODE_LOGBASE2:
+      emit_unaryop(gen, inst);
+      break;
+   case TGSI_OPCODE_RSQ:
+   case TGSI_OPCODE_RCP:
+      emit_scalar_unaryop(gen, inst);
+      break;
+   case TGSI_OPCODE_ADD:
+   case TGSI_OPCODE_SUB:
+   case TGSI_OPCODE_MUL:
+   case TGSI_OPCODE_MIN:
+   case TGSI_OPCODE_MAX:
+      emit_binop(gen, inst);
+      break;
+   case TGSI_OPCODE_SEQ:
+   case TGSI_OPCODE_SNE:
+   case TGSI_OPCODE_SLT:
+   case TGSI_OPCODE_SGT:
+   case TGSI_OPCODE_SLE:
+   case TGSI_OPCODE_SGE:
+      emit_inequality(gen, inst);
+      break;
+   case TGSI_OPCODE_MAD:
+   case TGSI_OPCODE_LRP:
+      emit_triop(gen, inst);
+      break;
+   case TGSI_OPCODE_DP3:
+   case TGSI_OPCODE_DP4:
+   case TGSI_OPCODE_DPH:
+      emit_dotprod(gen, inst);
+      break;
+   case TGSI_OPCODE_LIT:
+      emit_lit(gen, inst);
+      break;
+   case TGSI_OPCODE_LOG:
+      emit_log(gen, inst);
+      break;
+   case TGSI_OPCODE_EXP:
+      emit_exp(gen, inst);
+      break;
+   case TGSI_OPCODE_POW:
+      emit_pow(gen, inst);
+      break;
+   case TGSI_OPCODE_XPD:
+      emit_xpd(gen, inst);
+      break;
+   case TGSI_OPCODE_END:
+      /* normal end */
+      return 1;
+   default:
+      return 0;
+   }
+   return 1;
+}
+
+
+static void
+emit_declaration(
+   struct ppc_function *func,
+   struct tgsi_full_declaration *decl )
+{
+   if( decl->Declaration.File == TGSI_FILE_INPUT ) {
+#if 0
+      unsigned first, last, mask;
+      unsigned i, j;
+
+      first = decl->DeclarationRange.First;
+      last = decl->DeclarationRange.Last;
+      mask = decl->Declaration.UsageMask;
+
+      for( i = first; i <= last; i++ ) {
+         for( j = 0; j < NUM_CHANNELS; j++ ) {
+            if( mask & (1 << j) ) {
+               switch( decl->Declaration.Interpolate ) {
+               case TGSI_INTERPOLATE_CONSTANT:
+                  emit_coef_a0( func, 0, i, j );
+                  emit_inputs( func, 0, i, j );
+                  break;
+
+               case TGSI_INTERPOLATE_LINEAR:
+                  emit_tempf( func, 0, 0, TGSI_SWIZZLE_X );
+                  emit_coef_dadx( func, 1, i, j );
+                  emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y );
+                  emit_coef_dady( func, 3, i, j );
+                  emit_mul( func, 0, 1 );    /* x * dadx */
+                  emit_coef_a0( func, 4, i, j );
+                  emit_mul( func, 2, 3 );    /* y * dady */
+                  emit_add( func, 0, 4 );    /* x * dadx + a0 */
+                  emit_add( func, 0, 2 );    /* x * dadx + y * dady + a0 */
+                  emit_inputs( func, 0, i, j );
+                  break;
+
+               case TGSI_INTERPOLATE_PERSPECTIVE:
+                  emit_tempf( func, 0, 0, TGSI_SWIZZLE_X );
+                  emit_coef_dadx( func, 1, i, j );
+                  emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y );
+                  emit_coef_dady( func, 3, i, j );
+                  emit_mul( func, 0, 1 );    /* x * dadx */
+                  emit_tempf( func, 4, 0, TGSI_SWIZZLE_W );
+                  emit_coef_a0( func, 5, i, j );
+                  emit_rcp( func, 4, 4 );    /* 1.0 / w */
+                  emit_mul( func, 2, 3 );    /* y * dady */
+                  emit_add( func, 0, 5 );    /* x * dadx + a0 */
+                  emit_add( func, 0, 2 );    /* x * dadx + y * dady + a0 */
+                  emit_mul( func, 0, 4 );    /* (x * dadx + y * dady + a0) / w */
+                  emit_inputs( func, 0, i, j );
+                  break;
+
+               default:
+                  assert( 0 );
+		  break;
+               }
+            }
+         }
+      }
+#endif
+   }
+}
+
+
+
+static void
+emit_prologue(struct ppc_function *func)
+{
+   /* XXX set up stack frame */
+}
+
+
+static void
+emit_epilogue(struct ppc_function *func)
+{
+   ppc_comment(func, -4, "Epilogue:");
+   ppc_return(func);
+   /* XXX restore prev stack frame */
+#if 0
+   debug_printf("PPC: Emitted %u instructions\n", func->num_inst);
+#endif
+}
+
+
+
+/**
+ * Translate a TGSI vertex/fragment shader to PPC code.
+ *
+ * \param tokens  the TGSI input shader
+ * \param func  the output PPC code/function
+ * \param immediates  buffer to place immediates, later passed to PPC func
+ * \return TRUE for success, FALSE if translation failed
+ */
+boolean
+tgsi_emit_ppc(const struct tgsi_token *tokens,
+              struct ppc_function *func,
+              float (*immediates)[4],
+              boolean do_swizzles )
+{
+   static int use_ppc_asm = -1;
+   struct tgsi_parse_context parse;
+   /*boolean instruction_phase = FALSE;*/
+   unsigned ok = 1;
+   uint num_immediates = 0;
+   struct gen_context gen;
+   uint ic = 0;
+
+   if (use_ppc_asm < 0) {
+      /* If GALLIUM_NOPPC is set, don't use PPC codegen */
+      use_ppc_asm = !debug_get_bool_option("GALLIUM_NOPPC", FALSE);
+   }
+   if (!use_ppc_asm)
+      return FALSE;
+
+   if (0) {
+      debug_printf("\n********* TGSI->PPC ********\n");
+      tgsi_dump(tokens, 0);
+   }
+
+   util_init_math();
+
+   init_gen_context(&gen, func);
+
+   emit_prologue(func);
+
+   tgsi_parse_init( &parse, tokens );
+
+   while (!tgsi_parse_end_of_tokens(&parse) && ok) {
+      tgsi_parse_token(&parse);
+
+      switch (parse.FullToken.Token.Type) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
+            emit_declaration(func, &parse.FullToken.FullDeclaration );
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         if (func->print) {
+            _debug_printf("# ");
+            ic++;
+            tgsi_dump_instruction(&parse.FullToken.FullInstruction, ic);
+         }
+
+         ok = emit_instruction(&gen, &parse.FullToken.FullInstruction);
+
+	 if (!ok) {
+	    debug_printf("failed to translate tgsi opcode %d to PPC (%s)\n", 
+			 parse.FullToken.FullInstruction.Instruction.Opcode,
+                         parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ?
+                         "vertex shader" : "fragment shader");
+	 }
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         /* splat each immediate component into a float[4] vector for SoA */
+         {
+            const uint size = parse.FullToken.FullImmediate.Immediate.NrTokens - 1;
+            uint i;
+            assert(size <= 4);
+            assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES);
+            for (i = 0; i < size; i++) {
+               immediates[num_immediates][i] =
+		  parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float;
+            }
+            num_immediates++;
+         }
+         break;
+
+      default:
+	 ok = 0;
+         assert( 0 );
+      }
+   }
+
+   emit_epilogue(func);
+
+   tgsi_parse_free( &parse );
+
+   if (ppc_num_instructions(func) == 0) {
+      /* ran out of memory for instructions */
+      ok = FALSE;
+   }
+
+   if (!ok)
+      debug_printf("TGSI->PPC translation failed\n");
+
+   return ok;
+}
+
+#endif /* PIPE_ARCH_PPC */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ppc.h b/src/gallium/auxiliary/tgsi/tgsi_ppc.h
new file mode 100644
index 0000000000..829ec075e7
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_ppc.h
@@ -0,0 +1,51 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_PPC_H
+#define TGSI_PPC_H
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+struct tgsi_token;
+struct ppc_function;
+
+extern const float ppc_builtin_constants[];
+
+
+boolean
+tgsi_emit_ppc(const struct tgsi_token *tokens,
+              struct ppc_function *function,
+              float (*immediates)[4],
+              boolean do_swizzles);
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_PPC_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
new file mode 100644
index 0000000000..bc7b941b78
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
@@ -0,0 +1,364 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_debug.h"
+#include "tgsi_sanity.h"
+#include "tgsi_info.h"
+#include "tgsi_iterate.h"
+
+#define MAX_REGISTERS 256
+
+typedef uint reg_flag;
+
+#define BITS_IN_REG_FLAG (sizeof( reg_flag ) * 8)
+
+struct sanity_check_ctx
+{
+   struct tgsi_iterate_context iter;
+
+   reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REGISTERS / BITS_IN_REG_FLAG];
+   reg_flag regs_used[TGSI_FILE_COUNT][MAX_REGISTERS / BITS_IN_REG_FLAG];
+   boolean regs_ind_used[TGSI_FILE_COUNT];
+   uint num_imms;
+   uint num_instructions;
+   uint index_of_END;
+
+   uint errors;
+   uint warnings;
+};
+
+static void
+report_error(
+   struct sanity_check_ctx *ctx,
+   const char *format,
+   ... )
+{
+   va_list args;
+
+   debug_printf( "Error  : " );
+   va_start( args, format );
+   _debug_vprintf( format, args );
+   va_end( args );
+   debug_printf( "\n" );
+   ctx->errors++;
+}
+
+static void
+report_warning(
+   struct sanity_check_ctx *ctx,
+   const char *format,
+   ... )
+{
+   va_list args;
+
+   debug_printf( "Warning: " );
+   va_start( args, format );
+   _debug_vprintf( format, args );
+   va_end( args );
+   debug_printf( "\n" );
+   ctx->warnings++;
+}
+
+static boolean
+check_file_name(
+   struct sanity_check_ctx *ctx,
+   uint file )
+{
+   if (file <= TGSI_FILE_NULL || file >= TGSI_FILE_COUNT) {
+      report_error( ctx, "Invalid register file name" );
+      return FALSE;
+   }
+   return TRUE;
+}
+
+static boolean
+is_register_declared(
+   struct sanity_check_ctx *ctx,
+   uint file,
+   int index )
+{
+   assert( index >= 0 && index < MAX_REGISTERS );
+
+   return (ctx->regs_decl[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE;
+}
+
+static boolean
+is_any_register_declared(
+   struct sanity_check_ctx *ctx,
+   uint file )
+{
+   uint i;
+
+   for (i = 0; i < MAX_REGISTERS / BITS_IN_REG_FLAG; i++)
+      if (ctx->regs_decl[file][i])
+         return TRUE;
+   return FALSE;
+}
+
+static boolean
+is_register_used(
+   struct sanity_check_ctx *ctx,
+   uint file,
+   int index )
+{
+   assert( index < MAX_REGISTERS );
+
+   return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE;
+}
+
+static const char *file_names[] =
+{
+   "NULL",
+   "CONST",
+   "IN",
+   "OUT",
+   "TEMP",
+   "SAMP",
+   "ADDR",
+   "IMM"
+};
+
+static boolean
+check_register_usage(
+   struct sanity_check_ctx *ctx,
+   uint file,
+   int index,
+   const char *name,
+   boolean indirect_access )
+{
+   if (!check_file_name( ctx, file ))
+      return FALSE;
+
+   if (indirect_access) {
+      /* Note that 'index' is an offset relative to the value of the
+       * address register.  No range checking done here.
+       */
+      if (!is_any_register_declared( ctx, file ))
+         report_error( ctx, "%s: Undeclared %s register", file_names[file], name );
+      ctx->regs_ind_used[file] = TRUE;
+   }
+   else {
+      if (index < 0 || index > MAX_REGISTERS) {
+         report_error( ctx, "%s[%i]: Invalid index %s",
+                       file_names[file], index, name );
+         return FALSE;
+      }
+
+      if (!is_register_declared( ctx, file, index ))
+         report_error( ctx, "%s[%d]: Undeclared %s register", file_names[file], index, name );
+      ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG));
+   }
+   return TRUE;
+}
+
+static boolean
+iter_instruction(
+   struct tgsi_iterate_context *iter,
+   struct tgsi_full_instruction *inst )
+{
+   struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;
+   const struct tgsi_opcode_info *info;
+   uint i;
+
+   if (inst->Instruction.Opcode == TGSI_OPCODE_END) {
+      if (ctx->index_of_END != ~0) {
+         report_error( ctx, "Too many END instructions" );
+      }
+      ctx->index_of_END = ctx->num_instructions;
+   }
+
+   info = tgsi_get_opcode_info( inst->Instruction.Opcode );
+   if (info == NULL) {
+      report_error( ctx, "Invalid instruction opcode" );
+      return TRUE;
+   }
+
+   if (info->num_dst != inst->Instruction.NumDstRegs) {
+      report_error( ctx, "Invalid number of destination operands" );
+   }
+   if (info->num_src != inst->Instruction.NumSrcRegs) {
+      report_error( ctx, "Invalid number of source operands" );
+   }
+
+   /* Check destination and source registers' validity.
+    * Mark the registers as used.
+    */
+   for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
+      check_register_usage(
+         ctx,
+         inst->FullDstRegisters[i].DstRegister.File,
+         inst->FullDstRegisters[i].DstRegister.Index,
+         "destination",
+         FALSE );
+   }
+   for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
+      check_register_usage(
+         ctx,
+         inst->FullSrcRegisters[i].SrcRegister.File,
+         inst->FullSrcRegisters[i].SrcRegister.Index,
+         "source",
+         (boolean)inst->FullSrcRegisters[i].SrcRegister.Indirect );
+      if (inst->FullSrcRegisters[i].SrcRegister.Indirect) {
+         uint file;
+         int index;
+
+         file = inst->FullSrcRegisters[i].SrcRegisterInd.File;
+         index = inst->FullSrcRegisters[i].SrcRegisterInd.Index;
+         check_register_usage(
+            ctx,
+            file,
+            index,
+            "indirect",
+            FALSE );
+         if (file != TGSI_FILE_ADDRESS || index != 0)
+            report_warning( ctx, "Indirect register not ADDR[0]" );
+      }
+   }
+
+   ctx->num_instructions++;
+
+   return TRUE;
+}
+
+static boolean
+iter_declaration(
+   struct tgsi_iterate_context *iter,
+   struct tgsi_full_declaration *decl )
+{
+   struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;
+   uint file;
+   uint i;
+
+   /* No declarations allowed after the first instruction.
+    */
+   if (ctx->num_instructions > 0)
+      report_error( ctx, "Instruction expected but declaration found" );
+
+   /* Check registers' validity.
+    * Mark the registers as declared.
+    */
+   file = decl->Declaration.File;
+   if (!check_file_name( ctx, file ))
+      return TRUE;
+   for (i = decl->DeclarationRange.First; i <= decl->DeclarationRange.Last; i++) {
+      if (is_register_declared( ctx, file, i ))
+         report_error( ctx, "The same register declared twice" );
+      ctx->regs_decl[file][i / BITS_IN_REG_FLAG] |= (1 << (i % BITS_IN_REG_FLAG));
+   }
+
+   return TRUE;
+}
+
+static boolean
+iter_immediate(
+   struct tgsi_iterate_context *iter,
+   struct tgsi_full_immediate *imm )
+{
+   struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;
+
+   assert( ctx->num_imms < MAX_REGISTERS );
+
+   /* No immediates allowed after the first instruction.
+    */
+   if (ctx->num_instructions > 0)
+      report_error( ctx, "Instruction expected but immediate found" );
+
+   /* Mark the register as declared.
+    */
+   ctx->regs_decl[TGSI_FILE_IMMEDIATE][ctx->num_imms / BITS_IN_REG_FLAG] |= (1 << (ctx->num_imms % BITS_IN_REG_FLAG));
+   ctx->num_imms++;
+
+   /* Check data type validity.
+    */
+   if (imm->Immediate.DataType != TGSI_IMM_FLOAT32) {
+      report_error( ctx, "Invalid immediate data type" );
+      return TRUE;
+   }
+
+   return TRUE;
+}
+
+static boolean
+epilog(
+   struct tgsi_iterate_context *iter )
+{
+   struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;
+   uint file;
+
+   /* There must be an END instruction somewhere.
+    */
+   if (ctx->index_of_END == ~0) {
+      report_error( ctx, "Missing END instruction" );
+   }
+
+   /* Check if all declared registers were used.
+    */
+   for (file = TGSI_FILE_NULL; file < TGSI_FILE_COUNT; file++) {
+      uint i;
+
+      for (i = 0; i < MAX_REGISTERS; i++) {
+         if (is_register_declared( ctx, file, i ) && !is_register_used( ctx, file, i ) && !ctx->regs_ind_used[file]) {
+            report_warning( ctx, "Register never used" );
+         }
+      }
+   }
+
+   /* Print totals, if any.
+    */
+   if (ctx->errors || ctx->warnings)
+      debug_printf( "%u errors, %u warnings\n", ctx->errors, ctx->warnings );
+
+   return TRUE;
+}
+
+boolean
+tgsi_sanity_check(
+   struct tgsi_token *tokens )
+{
+   struct sanity_check_ctx ctx;
+
+   ctx.iter.prolog = NULL;
+   ctx.iter.iterate_instruction = iter_instruction;
+   ctx.iter.iterate_declaration = iter_declaration;
+   ctx.iter.iterate_immediate = iter_immediate;
+   ctx.iter.epilog = epilog;
+
+   memset( ctx.regs_decl, 0, sizeof( ctx.regs_decl ) );
+   memset( ctx.regs_used, 0, sizeof( ctx.regs_used ) );
+   memset( ctx.regs_ind_used, 0, sizeof( ctx.regs_ind_used ) );
+   ctx.num_imms = 0;
+   ctx.num_instructions = 0;
+   ctx.index_of_END = ~0;
+
+   ctx.errors = 0;
+   ctx.warnings = 0;
+
+   if (!tgsi_iterate_shader( tokens, &ctx.iter ))
+      return FALSE;
+
+   return ctx.errors == 0;
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.h b/src/gallium/auxiliary/tgsi/tgsi_sanity.h
new file mode 100644
index 0000000000..ca45e94c7a
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.h
@@ -0,0 +1,49 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_SANITY_H
+#define TGSI_SANITY_H
+
+#include "pipe/p_shader_tokens.h"
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+/* Check the given token stream for errors and common mistakes.
+ * Diagnostic messages are printed out to the debug output.
+ * Returns TRUE if there are no errors, even though there could be some warnings.
+ */
+boolean
+tgsi_sanity_check(
+   struct tgsi_token *tokens );
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_SANITY_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c
new file mode 100644
index 0000000000..d02205a63e
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c
@@ -0,0 +1,249 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2008 VMware, Inc.  All rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * TGSI program scan utility.
+ * Used to determine which registers and instructions are used by a shader.
+ *
+ * Authors:  Brian Paul
+ */
+
+
+#include "util/u_math.h"
+#include "tgsi/tgsi_build.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_scan.h"
+
+
+
+
+/**
+ * Scan the given TGSI shader to collect information such as number of
+ * registers used, special instructions used, etc.
+ * \return info  the result of the scan
+ */
+void
+tgsi_scan_shader(const struct tgsi_token *tokens,
+                 struct tgsi_shader_info *info)
+{
+   uint procType, i;
+   struct tgsi_parse_context parse;
+
+   memset(info, 0, sizeof(*info));
+   for (i = 0; i < TGSI_FILE_COUNT; i++)
+      info->file_max[i] = -1;
+
+   /**
+    ** Setup to begin parsing input shader
+    **/
+   if (tgsi_parse_init( &parse, tokens ) != TGSI_PARSE_OK) {
+      debug_printf("tgsi_parse_init() failed in tgsi_scan_shader()!\n");
+      return;
+   }
+   procType = parse.FullHeader.Processor.Processor;
+   assert(procType == TGSI_PROCESSOR_FRAGMENT ||
+          procType == TGSI_PROCESSOR_VERTEX ||
+          procType == TGSI_PROCESSOR_GEOMETRY);
+
+
+   /**
+    ** Loop over incoming program tokens/instructions
+    */
+   while( !tgsi_parse_end_of_tokens( &parse ) ) {
+
+      info->num_tokens++;
+
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         {
+            const struct tgsi_full_instruction *fullinst
+               = &parse.FullToken.FullInstruction;
+
+            assert(fullinst->Instruction.Opcode < TGSI_OPCODE_LAST);
+            info->opcode_count[fullinst->Instruction.Opcode]++;
+
+            /* special case: scan fragment shaders for use of the fog
+             * input/attribute.  The X component is fog, the Y component
+             * is the front/back-face flag.
+             */
+            if (procType == TGSI_PROCESSOR_FRAGMENT) {
+               uint i;
+               for (i = 0; i < fullinst->Instruction.NumSrcRegs; i++) {
+                  const struct tgsi_full_src_register *src =
+                     &fullinst->FullSrcRegisters[i];
+                  if (src->SrcRegister.File == TGSI_FILE_INPUT) {
+                     const int ind = src->SrcRegister.Index;
+                     if (info->input_semantic_name[ind] == TGSI_SEMANTIC_FOG) {
+                        if (src->SrcRegister.SwizzleX == TGSI_SWIZZLE_X) {
+                           info->uses_fogcoord = TRUE;
+                        }
+                        else if (src->SrcRegister.SwizzleX == TGSI_SWIZZLE_Y) {
+                           info->uses_frontfacing = TRUE;
+                        }
+                     }
+                  }
+               }
+            }
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         {
+            const struct tgsi_full_declaration *fulldecl
+               = &parse.FullToken.FullDeclaration;
+            const uint file = fulldecl->Declaration.File;
+            uint reg;
+            for (reg = fulldecl->DeclarationRange.First;
+                 reg <= fulldecl->DeclarationRange.Last;
+                 reg++) {
+
+               /* only first 32 regs will appear in this bitfield */
+               info->file_mask[file] |= (1 << reg);
+               info->file_count[file]++;
+               info->file_max[file] = MAX2(info->file_max[file], (int)reg);
+
+               if (file == TGSI_FILE_INPUT) {
+                  info->input_semantic_name[reg] = (ubyte)fulldecl->Semantic.SemanticName;
+                  info->input_semantic_index[reg] = (ubyte)fulldecl->Semantic.SemanticIndex;
+                  info->num_inputs++;
+               }
+               else if (file == TGSI_FILE_OUTPUT) {
+                  info->output_semantic_name[reg] = (ubyte)fulldecl->Semantic.SemanticName;
+                  info->output_semantic_index[reg] = (ubyte)fulldecl->Semantic.SemanticIndex;
+                  info->num_outputs++;
+               }
+
+               /* special case */
+               if (procType == TGSI_PROCESSOR_FRAGMENT &&
+                   file == TGSI_FILE_OUTPUT &&
+                   fulldecl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) {
+                  info->writes_z = TRUE;
+               }
+            }
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         info->immediate_count++;
+         break;
+
+      default:
+         assert( 0 );
+      }
+   }
+
+   info->uses_kill = (info->opcode_count[TGSI_OPCODE_KIL] ||
+                      info->opcode_count[TGSI_OPCODE_KILP]);
+
+   tgsi_parse_free (&parse);
+}
+
+
+
+/**
+ * Check if the given shader is a "passthrough" shader consisting of only
+ * MOV instructions of the form:  MOV OUT[n], IN[n]
+ *  
+ */
+boolean
+tgsi_is_passthrough_shader(const struct tgsi_token *tokens)
+{
+   struct tgsi_parse_context parse;
+
+   /**
+    ** Setup to begin parsing input shader
+    **/
+   if (tgsi_parse_init(&parse, tokens) != TGSI_PARSE_OK) {
+      debug_printf("tgsi_parse_init() failed in tgsi_is_passthrough_shader()!\n");
+      return FALSE;
+   }
+
+   /**
+    ** Loop over incoming program tokens/instructions
+    */
+   while (!tgsi_parse_end_of_tokens(&parse)) {
+
+      tgsi_parse_token(&parse);
+
+      switch (parse.FullToken.Token.Type) {
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         {
+            struct tgsi_full_instruction *fullinst =
+               &parse.FullToken.FullInstruction;
+            const struct tgsi_full_src_register *src =
+               &fullinst->FullSrcRegisters[0];
+            const struct tgsi_full_dst_register *dst =
+               &fullinst->FullDstRegisters[0];
+
+            /* Do a whole bunch of checks for a simple move */
+            if (fullinst->Instruction.Opcode != TGSI_OPCODE_MOV ||
+                src->SrcRegister.File != TGSI_FILE_INPUT ||
+                dst->DstRegister.File != TGSI_FILE_OUTPUT ||
+                src->SrcRegister.Index != dst->DstRegister.Index ||
+
+                src->SrcRegister.Negate ||
+                src->SrcRegisterExtMod.Negate ||
+                src->SrcRegisterExtMod.Absolute ||
+                src->SrcRegisterExtMod.Scale2X ||
+                src->SrcRegisterExtMod.Bias ||
+                src->SrcRegisterExtMod.Complement ||
+
+                src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X ||
+                src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y ||
+                src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z ||
+                src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ||
+
+                src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X ||
+                src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y ||
+                src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z ||
+                src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W ||
+
+                dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW)
+            {
+               tgsi_parse_free(&parse);
+               return FALSE;
+            }
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         /* fall-through */
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         /* fall-through */
+      default:
+         ; /* no-op */
+      }
+   }
+
+   tgsi_parse_free(&parse);
+
+   /* if we get here, it's a pass-through shader */
+   return TRUE;
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.h b/src/gallium/auxiliary/tgsi/tgsi_scan.h
new file mode 100644
index 0000000000..2c1a75bc81
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.h
@@ -0,0 +1,75 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_SCAN_H
+#define TGSI_SCAN_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+
+
+/**
+ * Shader summary info
+ */
+struct tgsi_shader_info
+{
+   uint num_tokens;
+
+   ubyte num_inputs;
+   ubyte num_outputs;
+   ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */
+   ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS];
+   ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */
+   ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS];
+
+   uint file_mask[TGSI_FILE_COUNT];  /**< bitmask of declared registers */
+   uint file_count[TGSI_FILE_COUNT];  /**< number of declared registers */
+   int file_max[TGSI_FILE_COUNT];  /**< highest index of declared registers */
+
+   uint immediate_count; /**< number of immediates declared */
+
+   uint opcode_count[TGSI_OPCODE_LAST];  /**< opcode histogram */
+
+   boolean writes_z;  /**< does fragment shader write Z value? */
+   boolean uses_kill;  /**< KIL or KILP instruction used? */
+   boolean uses_fogcoord; /**< fragment shader uses fog coord? */
+   boolean uses_frontfacing; /**< fragment shader uses front/back-face flag? */
+};
+
+
+extern void
+tgsi_scan_shader(const struct tgsi_token *tokens,
+                 struct tgsi_shader_info *info);
+
+
+extern boolean
+tgsi_is_passthrough_shader(const struct tgsi_token *tokens);
+
+
+#endif /* TGSI_SCAN_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
new file mode 100644
index 0000000000..481ba89c5e
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c
@@ -0,0 +1,2723 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_ARCH_X86)
+
+#include "pipe/p_debug.h"
+#include "pipe/p_shader_tokens.h"
+#include "util/u_math.h"
+#if defined(PIPE_ARCH_SSE)
+#include "util/u_sse.h"
+#endif
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi_exec.h"
+#include "tgsi_sse2.h"
+
+#include "rtasm/rtasm_x86sse.h"
+
+/* for 1/sqrt()
+ *
+ * This costs about 100fps (close to 10%) in gears:
+ */
+#define HIGH_PRECISION 1
+
+#define FAST_MATH 1
+
+
+#define FOR_EACH_CHANNEL( CHAN )\
+   for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)
+
+#define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\
+   ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN)))
+
+#define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\
+   if (IS_DST0_CHANNEL_ENABLED( INST, CHAN ))
+
+#define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\
+   FOR_EACH_CHANNEL( CHAN )\
+      IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )
+
+#define CHAN_X 0
+#define CHAN_Y 1
+#define CHAN_Z 2
+#define CHAN_W 3
+
+#define TEMP_ONE_I   TGSI_EXEC_TEMP_ONE_I
+#define TEMP_ONE_C   TGSI_EXEC_TEMP_ONE_C
+
+#define TEMP_R0   TGSI_EXEC_TEMP_R0
+#define TEMP_ADDR TGSI_EXEC_TEMP_ADDR
+#define TEMP_EXEC_MASK_I TGSI_EXEC_MASK_I
+#define TEMP_EXEC_MASK_C TGSI_EXEC_MASK_C
+
+
+/**
+ * X86 utility functions.
+ */
+
+static struct x86_reg
+make_xmm(
+   unsigned xmm )
+{
+   return x86_make_reg(
+      file_XMM,
+      (enum x86_reg_name) xmm );
+}
+
+/**
+ * X86 register mapping helpers.
+ */
+
+static struct x86_reg
+get_const_base( void )
+{
+   return x86_make_reg(
+      file_REG32,
+      reg_CX );
+}
+
+static struct x86_reg
+get_input_base( void )
+{
+   return x86_make_reg(
+      file_REG32,
+      reg_AX );
+}
+
+static struct x86_reg
+get_output_base( void )
+{
+   return x86_make_reg(
+      file_REG32,
+      reg_DX );
+}
+
+static struct x86_reg
+get_temp_base( void )
+{
+   return x86_make_reg(
+      file_REG32,
+      reg_BX );
+}
+
+static struct x86_reg
+get_coef_base( void )
+{
+   return get_output_base();
+}
+
+static struct x86_reg
+get_immediate_base( void )
+{
+   return x86_make_reg(
+      file_REG32,
+      reg_DI );
+}
+
+
+/**
+ * Data access helpers.
+ */
+
+
+static struct x86_reg
+get_immediate(
+   unsigned vec,
+   unsigned chan )
+{
+   return x86_make_disp(
+      get_immediate_base(),
+      (vec * 4 + chan) * 4 );
+}
+
+static struct x86_reg
+get_const(
+   unsigned vec,
+   unsigned chan )
+{
+   return x86_make_disp(
+      get_const_base(),
+      (vec * 4 + chan) * 4 );
+}
+
+static struct x86_reg
+get_input(
+   unsigned vec,
+   unsigned chan )
+{
+   return x86_make_disp(
+      get_input_base(),
+      (vec * 4 + chan) * 16 );
+}
+
+static struct x86_reg
+get_output(
+   unsigned vec,
+   unsigned chan )
+{
+   return x86_make_disp(
+      get_output_base(),
+      (vec * 4 + chan) * 16 );
+}
+
+static struct x86_reg
+get_temp(
+   unsigned vec,
+   unsigned chan )
+{
+   return x86_make_disp(
+      get_temp_base(),
+      (vec * 4 + chan) * 16 );
+}
+
+static struct x86_reg
+get_coef(
+   unsigned vec,
+   unsigned chan,
+   unsigned member )
+{
+   return x86_make_disp(
+      get_coef_base(),
+      ((vec * 3 + member) * 4 + chan) * 4 );
+}
+
+
+static void
+emit_ret(
+   struct x86_function  *func )
+{
+   x86_ret( func );
+}
+
+
+/**
+ * Data fetch helpers.
+ */
+
+/**
+ * Copy a shader constant to xmm register
+ * \param xmm  the destination xmm register
+ * \param vec  the src const buffer index
+ * \param chan  src channel to fetch (X, Y, Z or W)
+ */
+static void
+emit_const(
+   struct x86_function *func,
+   uint xmm,
+   int vec,
+   uint chan,
+   uint indirect,
+   uint indirectFile,
+   int indirectIndex )
+{
+   if (indirect) {
+      /* 'vec' is the offset from the address register's value.
+       * We're loading CONST[ADDR+vec] into an xmm register.
+       */
+      struct x86_reg r0 = get_input_base();
+      struct x86_reg r1 = get_output_base();
+      uint i;
+
+      assert( indirectFile == TGSI_FILE_ADDRESS );
+      assert( indirectIndex == 0 );
+
+      x86_push( func, r0 );
+      x86_push( func, r1 );
+
+      /*
+       * Loop over the four pixels or vertices in the quad.
+       * Get the value of the address (offset) register for pixel/vertex[i],
+       * add it to the src offset and index into the constant buffer.
+       * Note that we're working on SOA data.
+       * If any of the pixel/vertex execution channels are unused their
+       * values will be garbage.  It's very important that we don't use
+       * those garbage values as indexes into the constant buffer since
+       * that'll cause segfaults.
+       * The solution is to bitwise-AND the offset with the execution mask
+       * register whose values are either 0 or ~0.
+       * The caller must setup the execution mask register to indicate
+       * which channels are valid/alive before running the shader.
+       * The execution mask will also figure into loops and conditionals
+       * someday.
+       */
+      for (i = 0; i < QUAD_SIZE; i++) {
+         /* r1 = address register[i] */
+         x86_mov( func, r1, x86_make_disp( get_temp( TEMP_ADDR, CHAN_X ), i * 4 ) );
+         /* r0 = execution mask[i] */
+         x86_mov( func, r0, x86_make_disp( get_temp( TEMP_EXEC_MASK_I, TEMP_EXEC_MASK_C ), i * 4 ) );
+         /* r1 = r1 & r0 */
+         x86_and( func, r1, r0 );
+         /* r0 = 'vec', the offset */
+         x86_lea( func, r0, get_const( vec, chan ) );
+
+         /* Quick hack to multiply r1 by 16 -- need to add SHL to rtasm.
+          */
+         x86_add( func, r1, r1 );
+         x86_add( func, r1, r1 );
+         x86_add( func, r1, r1 );
+         x86_add( func, r1, r1 );
+
+         x86_add( func, r0, r1 );  /* r0 = r0 + r1 */
+         x86_mov( func, r1, x86_deref( r0 ) );
+         x86_mov( func, x86_make_disp( get_temp( TEMP_R0, CHAN_X ), i * 4 ), r1 );
+      }
+
+      x86_pop( func, r1 );
+      x86_pop( func, r0 );
+
+      sse_movaps(
+         func,
+         make_xmm( xmm ),
+         get_temp( TEMP_R0, CHAN_X ) );
+   }
+   else {
+      /* 'vec' is the index into the src register file, such as TEMP[vec] */
+      assert( vec >= 0 );
+
+      sse_movss(
+         func,
+         make_xmm( xmm ),
+         get_const( vec, chan ) );
+      sse_shufps(
+         func,
+         make_xmm( xmm ),
+         make_xmm( xmm ),
+         SHUF( 0, 0, 0, 0 ) );
+   }
+}
+
+static void
+emit_immediate(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   sse_movss(
+      func,
+      make_xmm( xmm ),
+      get_immediate( vec, chan ) );
+   sse_shufps(
+      func,
+      make_xmm( xmm ),
+      make_xmm( xmm ),
+      SHUF( 0, 0, 0, 0 ) );
+}
+
+
+/**
+ * Copy a shader input to xmm register
+ * \param xmm  the destination xmm register
+ * \param vec  the src input attrib
+ * \param chan  src channel to fetch (X, Y, Z or W)
+ */
+static void
+emit_inputf(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   sse_movups(
+      func,
+      make_xmm( xmm ),
+      get_input( vec, chan ) );
+}
+
+/**
+ * Store an xmm register to a shader output
+ * \param xmm  the source xmm register
+ * \param vec  the dest output attrib
+ * \param chan  src dest channel to store (X, Y, Z or W)
+ */
+static void
+emit_output(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   sse_movups(
+      func,
+      get_output( vec, chan ),
+      make_xmm( xmm ) );
+}
+
+/**
+ * Copy a shader temporary to xmm register
+ * \param xmm  the destination xmm register
+ * \param vec  the src temp register
+ * \param chan  src channel to fetch (X, Y, Z or W)
+ */
+static void
+emit_tempf(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   sse_movaps(
+      func,
+      make_xmm( xmm ),
+      get_temp( vec, chan ) );
+}
+
+/**
+ * Load an xmm register with an input attrib coefficient (a0, dadx or dady)
+ * \param xmm  the destination xmm register
+ * \param vec  the src input/attribute coefficient index
+ * \param chan  src channel to fetch (X, Y, Z or W)
+ * \param member  0=a0, 1=dadx, 2=dady
+ */
+static void
+emit_coef(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan,
+   unsigned member )
+{
+   sse_movss(
+      func,
+      make_xmm( xmm ),
+      get_coef( vec, chan, member ) );
+   sse_shufps(
+      func,
+      make_xmm( xmm ),
+      make_xmm( xmm ),
+      SHUF( 0, 0, 0, 0 ) );
+}
+
+/**
+ * Data store helpers.
+ */
+
+static void
+emit_inputs(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   sse_movups(
+      func,
+      get_input( vec, chan ),
+      make_xmm( xmm ) );
+}
+
+static void
+emit_temps(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   sse_movaps(
+      func,
+      get_temp( vec, chan ),
+      make_xmm( xmm ) );
+}
+
+static void
+emit_addrs(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   assert( vec == 0 );
+
+   emit_temps(
+      func,
+      xmm,
+      vec + TGSI_EXEC_TEMP_ADDR,
+      chan );
+}
+
+/**
+ * Coefficent fetch helpers.
+ */
+
+static void
+emit_coef_a0(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   emit_coef(
+      func,
+      xmm,
+      vec,
+      chan,
+      0 );
+}
+
+static void
+emit_coef_dadx(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   emit_coef(
+      func,
+      xmm,
+      vec,
+      chan,
+      1 );
+}
+
+static void
+emit_coef_dady(
+   struct x86_function *func,
+   unsigned xmm,
+   unsigned vec,
+   unsigned chan )
+{
+   emit_coef(
+      func,
+      xmm,
+      vec,
+      chan,
+      2 );
+}
+
+/**
+ * Function call helpers.
+ */
+
+/**
+ * NOTE: In gcc, if the destination uses the SSE intrinsics, then it must be 
+ * defined with __attribute__((force_align_arg_pointer)), as we do not guarantee
+ * that the stack pointer is 16 byte aligned, as expected.
+ */
+static void
+emit_func_call_dst(
+   struct x86_function *func,
+   unsigned xmm_save,
+   unsigned xmm_dst,
+   void (PIPE_CDECL *code)() )
+{
+   struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX );
+   unsigned i, n;
+   unsigned xmm_mask;
+   
+   /* Bitmask of the xmm registers to save */
+   xmm_mask = (1 << xmm_save) - 1;
+   xmm_mask &= ~(1 << xmm_dst);
+
+   sse_movaps(
+      func,
+      get_temp( TEMP_R0, 0 ),
+      make_xmm( xmm_dst ) );
+
+   x86_push(
+      func,
+      x86_make_reg( file_REG32, reg_AX) );
+   x86_push(
+      func,
+      x86_make_reg( file_REG32, reg_CX) );
+   x86_push(
+      func,
+      x86_make_reg( file_REG32, reg_DX) );
+   
+   for(i = 0, n = 0; i < 8; ++i)
+      if(xmm_mask & (1 << i))
+         ++n;
+   
+   x86_sub_imm(
+      func, 
+      x86_make_reg( file_REG32, reg_SP ),
+      n*16);
+
+   for(i = 0, n = 0; i < 8; ++i)
+      if(xmm_mask & (1 << i)) {
+         sse_movups(
+            func,
+            x86_make_disp( x86_make_reg( file_REG32, reg_SP ), n*16 ),
+            make_xmm( i ) );
+         ++n;
+      }
+   
+   x86_lea(
+      func,
+      ecx,
+      get_temp( TEMP_R0, 0 ) );
+   
+   x86_push( func, ecx );
+   x86_mov_reg_imm( func, ecx, (unsigned long) code );
+   x86_call( func, ecx );
+   x86_pop(func, ecx );
+   
+   for(i = 0, n = 0; i < 8; ++i)
+      if(xmm_mask & (1 << i)) {
+         sse_movups(
+            func,
+            make_xmm( i ),
+            x86_make_disp( x86_make_reg( file_REG32, reg_SP ), n*16 ) );
+         ++n;
+      }
+   
+   x86_add_imm(
+      func, 
+      x86_make_reg( file_REG32, reg_SP ),
+      n*16);
+
+   /* Restore GP registers in a reverse order.
+    */
+   x86_pop(
+      func,
+      x86_make_reg( file_REG32, reg_DX) );
+   x86_pop(
+      func,
+      x86_make_reg( file_REG32, reg_CX) );
+   x86_pop(
+      func,
+      x86_make_reg( file_REG32, reg_AX) );
+
+   sse_movaps(
+      func,
+      make_xmm( xmm_dst ),
+      get_temp( TEMP_R0, 0 ) );
+}
+
+static void
+emit_func_call_dst_src(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst,
+   unsigned xmm_src,
+   void (PIPE_CDECL *code)() )
+{
+   sse_movaps(
+      func,
+      get_temp( TEMP_R0, 1 ),
+      make_xmm( xmm_src ) );
+
+   emit_func_call_dst(
+      func,
+      xmm_save,
+      xmm_dst,
+      code );
+}
+
+
+#if defined(PIPE_ARCH_SSE)
+
+/*
+ * Fast SSE2 implementation of special math functions.
+ */
+
+#define POLY0(x, c0) _mm_set1_ps(c0)
+#define POLY1(x, c0, c1) _mm_add_ps(_mm_mul_ps(POLY0(x, c1), x), _mm_set1_ps(c0))
+#define POLY2(x, c0, c1, c2) _mm_add_ps(_mm_mul_ps(POLY1(x, c1, c2), x), _mm_set1_ps(c0))
+#define POLY3(x, c0, c1, c2, c3) _mm_add_ps(_mm_mul_ps(POLY2(x, c1, c2, c3), x), _mm_set1_ps(c0))
+#define POLY4(x, c0, c1, c2, c3, c4) _mm_add_ps(_mm_mul_ps(POLY3(x, c1, c2, c3, c4), x), _mm_set1_ps(c0))
+#define POLY5(x, c0, c1, c2, c3, c4, c5) _mm_add_ps(_mm_mul_ps(POLY4(x, c1, c2, c3, c4, c5), x), _mm_set1_ps(c0))
+
+#define EXP_POLY_DEGREE 3
+#define LOG_POLY_DEGREE 5
+
+/**
+ * See http://www.devmaster.net/forums/showthread.php?p=43580
+ */
+static INLINE __m128 
+exp2f4(__m128 x)
+{
+   __m128i ipart;
+   __m128 fpart, expipart, expfpart;
+
+   x = _mm_min_ps(x, _mm_set1_ps( 129.00000f));
+   x = _mm_max_ps(x, _mm_set1_ps(-126.99999f));
+
+   /* ipart = int(x - 0.5) */
+   ipart = _mm_cvtps_epi32(_mm_sub_ps(x, _mm_set1_ps(0.5f)));
+
+   /* fpart = x - ipart */
+   fpart = _mm_sub_ps(x, _mm_cvtepi32_ps(ipart));
+
+   /* expipart = (float) (1 << ipart) */
+   expipart = _mm_castsi128_ps(_mm_slli_epi32(_mm_add_epi32(ipart, _mm_set1_epi32(127)), 23));
+
+   /* minimax polynomial fit of 2**x, in range [-0.5, 0.5[ */
+#if EXP_POLY_DEGREE == 5
+   expfpart = POLY5(fpart, 9.9999994e-1f, 6.9315308e-1f, 2.4015361e-1f, 5.5826318e-2f, 8.9893397e-3f, 1.8775767e-3f);
+#elif EXP_POLY_DEGREE == 4
+   expfpart = POLY4(fpart, 1.0000026f, 6.9300383e-1f, 2.4144275e-1f, 5.2011464e-2f, 1.3534167e-2f);
+#elif EXP_POLY_DEGREE == 3
+   expfpart = POLY3(fpart, 9.9992520e-1f, 6.9583356e-1f, 2.2606716e-1f, 7.8024521e-2f);
+#elif EXP_POLY_DEGREE == 2
+   expfpart = POLY2(fpart, 1.0017247f, 6.5763628e-1f, 3.3718944e-1f);
+#else
+#error
+#endif
+
+   return _mm_mul_ps(expipart, expfpart);
+}
+
+
+/**
+ * See http://www.devmaster.net/forums/showthread.php?p=43580
+ */
+static INLINE __m128 
+log2f4(__m128 x)
+{
+   __m128i expmask = _mm_set1_epi32(0x7f800000);
+   __m128i mantmask = _mm_set1_epi32(0x007fffff);
+   __m128 one = _mm_set1_ps(1.0f);
+
+   __m128i i = _mm_castps_si128(x);
+
+   /* exp = (float) exponent(x) */
+   __m128 exp = _mm_cvtepi32_ps(_mm_sub_epi32(_mm_srli_epi32(_mm_and_si128(i, expmask), 23), _mm_set1_epi32(127)));
+
+   /* mant = (float) mantissa(x) */
+   __m128 mant = _mm_or_ps(_mm_castsi128_ps(_mm_and_si128(i, mantmask)), one);
+
+   __m128 logmant;
+
+   /* Minimax polynomial fit of log2(x)/(x - 1), for x in range [1, 2[ 
+    * These coefficients can be generate with 
+    * http://www.boost.org/doc/libs/1_36_0/libs/math/doc/sf_and_dist/html/math_toolkit/toolkit/internals2/minimax.html
+    */
+#if LOG_POLY_DEGREE == 6
+   logmant = POLY5(mant, 3.11578814719469302614f, -3.32419399085241980044f, 2.59883907202499966007f, -1.23152682416275988241f, 0.318212422185251071475f, -0.0344359067839062357313f);
+#elif LOG_POLY_DEGREE == 5
+   logmant = POLY4(mant, 2.8882704548164776201f, -2.52074962577807006663f, 1.48116647521213171641f, -0.465725644288844778798f, 0.0596515482674574969533f);
+#elif LOG_POLY_DEGREE == 4
+   logmant = POLY3(mant, 2.61761038894603480148f, -1.75647175389045657003f, 0.688243882994381274313f, -0.107254423828329604454f);
+#elif LOG_POLY_DEGREE == 3
+   logmant = POLY2(mant, 2.28330284476918490682f, -1.04913055217340124191f, 0.204446009836232697516f);
+#else
+#error
+#endif
+
+   /* This effectively increases the polynomial degree by one, but ensures that log2(1) == 0*/
+   logmant = _mm_mul_ps(logmant, _mm_sub_ps(mant, one));
+
+   return _mm_add_ps(logmant, exp);
+}
+
+
+static INLINE __m128
+powf4(__m128 x, __m128 y)
+{
+   return exp2f4(_mm_mul_ps(log2f4(x), y));
+}
+
+#endif /* PIPE_ARCH_SSE */
+
+
+
+/**
+ * Low-level instruction translators.
+ */
+
+static void
+emit_abs(
+   struct x86_function *func,
+   unsigned xmm )
+{
+   sse_andps(
+      func,
+      make_xmm( xmm ),
+      get_temp(
+         TGSI_EXEC_TEMP_7FFFFFFF_I,
+         TGSI_EXEC_TEMP_7FFFFFFF_C ) );
+}
+
+static void
+emit_add(
+   struct x86_function *func,
+   unsigned xmm_dst,
+   unsigned xmm_src )
+{
+   sse_addps(
+      func,
+      make_xmm( xmm_dst ),
+      make_xmm( xmm_src ) );
+}
+
+static void PIPE_CDECL
+cos4f(
+   float *store )
+{
+   store[0] = cosf( store[0] );
+   store[1] = cosf( store[1] );
+   store[2] = cosf( store[2] );
+   store[3] = cosf( store[3] );
+}
+
+static void
+emit_cos(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst )
+{
+   emit_func_call_dst(
+      func,
+      xmm_save, 
+      xmm_dst,
+      cos4f );
+}
+
+static void PIPE_CDECL
+#if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_SSE)
+__attribute__((force_align_arg_pointer))
+#endif
+ex24f(
+   float *store )
+{
+#if defined(PIPE_ARCH_SSE)
+   _mm_store_ps(&store[0], exp2f4( _mm_load_ps(&store[0]) ));
+#else
+   store[0] = util_fast_exp2( store[0] );
+   store[1] = util_fast_exp2( store[1] );
+   store[2] = util_fast_exp2( store[2] );
+   store[3] = util_fast_exp2( store[3] );
+#endif
+}
+
+static void
+emit_ex2(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst )
+{
+   emit_func_call_dst(
+      func,
+      xmm_save,
+      xmm_dst,
+      ex24f );
+}
+
+static void
+emit_f2it(
+   struct x86_function *func,
+   unsigned xmm )
+{
+   sse2_cvttps2dq(
+      func,
+      make_xmm( xmm ),
+      make_xmm( xmm ) );
+}
+
+static void
+emit_i2f(
+   struct x86_function *func,
+   unsigned xmm )
+{
+   sse2_cvtdq2ps(
+      func,
+      make_xmm( xmm ),
+      make_xmm( xmm ) );
+}
+
+static void PIPE_CDECL
+flr4f(
+   float *store )
+{
+   store[0] = floorf( store[0] );
+   store[1] = floorf( store[1] );
+   store[2] = floorf( store[2] );
+   store[3] = floorf( store[3] );
+}
+
+static void
+emit_flr(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst )
+{
+   emit_func_call_dst(
+      func,
+      xmm_save,
+      xmm_dst,
+      flr4f );
+}
+
+static void PIPE_CDECL
+frc4f(
+   float *store )
+{
+   store[0] -= floorf( store[0] );
+   store[1] -= floorf( store[1] );
+   store[2] -= floorf( store[2] );
+   store[3] -= floorf( store[3] );
+}
+
+static void
+emit_frc(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst )
+{
+   emit_func_call_dst(
+      func,
+      xmm_save,
+      xmm_dst,
+      frc4f );
+}
+
+static void PIPE_CDECL
+#if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_SSE)
+__attribute__((force_align_arg_pointer))
+#endif
+lg24f(
+   float *store )
+{
+#if defined(PIPE_ARCH_SSE)
+   _mm_store_ps(&store[0], log2f4( _mm_load_ps(&store[0]) ));
+#else
+   store[0] = util_fast_log2( store[0] );
+   store[1] = util_fast_log2( store[1] );
+   store[2] = util_fast_log2( store[2] );
+   store[3] = util_fast_log2( store[3] );
+#endif
+}
+
+static void
+emit_lg2(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst )
+{
+   emit_func_call_dst(
+      func,
+      xmm_save,
+      xmm_dst,
+      lg24f );
+}
+
+static void
+emit_MOV(
+   struct x86_function *func,
+   unsigned xmm_dst,
+   unsigned xmm_src )
+{
+   sse_movups(
+      func,
+      make_xmm( xmm_dst ),
+      make_xmm( xmm_src ) );
+}
+
+static void
+emit_mul (struct x86_function *func,
+          unsigned xmm_dst,
+          unsigned xmm_src)
+{
+   sse_mulps(
+      func,
+      make_xmm( xmm_dst ),
+      make_xmm( xmm_src ) );
+}
+
+static void
+emit_neg(
+   struct x86_function *func,
+   unsigned xmm )
+{
+   sse_xorps(
+      func,
+      make_xmm( xmm ),
+      get_temp(
+         TGSI_EXEC_TEMP_80000000_I,
+         TGSI_EXEC_TEMP_80000000_C ) );
+}
+
+static void PIPE_CDECL
+#if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_SSE)
+__attribute__((force_align_arg_pointer))
+#endif
+pow4f(
+   float *store )
+{
+#if defined(PIPE_ARCH_SSE)
+   _mm_store_ps(&store[0], powf4( _mm_load_ps(&store[0]), _mm_load_ps(&store[4]) ));
+#else
+   store[0] = util_fast_pow( store[0], store[4] );
+   store[1] = util_fast_pow( store[1], store[5] );
+   store[2] = util_fast_pow( store[2], store[6] );
+   store[3] = util_fast_pow( store[3], store[7] );
+#endif
+}
+
+static void
+emit_pow(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst,
+   unsigned xmm_src )
+{
+   emit_func_call_dst_src(
+      func,
+      xmm_save,
+      xmm_dst,
+      xmm_src,
+      pow4f );
+}
+
+static void
+emit_rcp (
+   struct x86_function *func,
+   unsigned xmm_dst,
+   unsigned xmm_src )
+{
+   /* On Intel CPUs at least, this is only accurate to 12 bits -- not
+    * good enough.  Need to either emit a proper divide or use the
+    * iterative technique described below in emit_rsqrt().
+    */
+   sse2_rcpps(
+      func,
+      make_xmm( xmm_dst ),
+      make_xmm( xmm_src ) );
+}
+
+static void PIPE_CDECL
+rnd4f(
+   float *store )
+{
+   store[0] = floorf( store[0] + 0.5f );
+   store[1] = floorf( store[1] + 0.5f );
+   store[2] = floorf( store[2] + 0.5f );
+   store[3] = floorf( store[3] + 0.5f );
+}
+
+static void
+emit_rnd(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst )
+{
+   emit_func_call_dst(
+      func,
+      xmm_save,
+      xmm_dst,
+      rnd4f );
+}
+
+static void
+emit_rsqrt(
+   struct x86_function *func,
+   unsigned xmm_dst,
+   unsigned xmm_src )
+{
+#if HIGH_PRECISION
+   /* Although rsqrtps() and rcpps() are low precision on some/all SSE
+    * implementations, it is possible to improve its precision at
+    * fairly low cost, using a newton/raphson step, as below:
+    * 
+    * x1 = 2 * rcpps(a) - a * rcpps(a) * rcpps(a)
+    * x1 = 0.5 * rsqrtps(a) * [3.0 - (a * rsqrtps(a))* rsqrtps(a)]
+    *
+    * See: http://softwarecommunity.intel.com/articles/eng/1818.htm
+    */
+   {
+      struct x86_reg dst = make_xmm( xmm_dst );
+      struct x86_reg src = make_xmm( xmm_src );
+      struct x86_reg tmp0 = make_xmm( 2 );
+      struct x86_reg tmp1 = make_xmm( 3 );
+
+      assert( xmm_dst != xmm_src );
+      assert( xmm_dst != 2 && xmm_dst != 3 );
+      assert( xmm_src != 2 && xmm_src != 3 );
+
+      sse_movaps(  func, dst,  get_temp( TGSI_EXEC_TEMP_HALF_I, TGSI_EXEC_TEMP_HALF_C ) );
+      sse_movaps(  func, tmp0, get_temp( TGSI_EXEC_TEMP_THREE_I, TGSI_EXEC_TEMP_THREE_C ) );
+      sse_rsqrtps( func, tmp1, src  );
+      sse_mulps(   func, src,  tmp1 );
+      sse_mulps(   func, dst,  tmp1 );
+      sse_mulps(   func, src,  tmp1 );
+      sse_subps(   func, tmp0, src  );
+      sse_mulps(   func, dst,  tmp0 );
+   }
+#else
+   /* On Intel CPUs at least, this is only accurate to 12 bits -- not
+    * good enough.
+    */
+   sse_rsqrtps(
+      func,
+      make_xmm( xmm_dst ),
+      make_xmm( xmm_src ) );
+#endif
+}
+
+static void
+emit_setsign(
+   struct x86_function *func,
+   unsigned xmm )
+{
+   sse_orps(
+      func,
+      make_xmm( xmm ),
+      get_temp(
+         TGSI_EXEC_TEMP_80000000_I,
+         TGSI_EXEC_TEMP_80000000_C ) );
+}
+
+static void PIPE_CDECL
+sgn4f(
+   float *store )
+{
+   store[0] = store[0] < 0.0f ? -1.0f : store[0] > 0.0f ? 1.0f : 0.0f;
+   store[1] = store[1] < 0.0f ? -1.0f : store[1] > 0.0f ? 1.0f : 0.0f;
+   store[2] = store[2] < 0.0f ? -1.0f : store[2] > 0.0f ? 1.0f : 0.0f;
+   store[3] = store[3] < 0.0f ? -1.0f : store[3] > 0.0f ? 1.0f : 0.0f;
+}
+
+static void
+emit_sgn(
+   struct x86_function *func,
+   unsigned xmm_save, 
+   unsigned xmm_dst )
+{
+   emit_func_call_dst(
+      func,
+      xmm_save,
+      xmm_dst,
+      sgn4f );
+}
+
+static void PIPE_CDECL
+sin4f(
+   float *store )
+{
+   store[0] = sinf( store[0] );
+   store[1] = sinf( store[1] );
+   store[2] = sinf( store[2] );
+   store[3] = sinf( store[3] );
+}
+
+static void
+emit_sin (struct x86_function *func,
+          unsigned xmm_save, 
+          unsigned xmm_dst)
+{
+   emit_func_call_dst(
+      func,
+      xmm_save,
+      xmm_dst,
+      sin4f );
+}
+
+static void
+emit_sub(
+   struct x86_function *func,
+   unsigned xmm_dst,
+   unsigned xmm_src )
+{
+   sse_subps(
+      func,
+      make_xmm( xmm_dst ),
+      make_xmm( xmm_src ) );
+}
+
+/**
+ * Register fetch.
+ */
+
+static void
+emit_fetch(
+   struct x86_function *func,
+   unsigned xmm,
+   const struct tgsi_full_src_register *reg,
+   const unsigned chan_index )
+{
+   unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index );
+
+   switch (swizzle) {
+   case TGSI_EXTSWIZZLE_X:
+   case TGSI_EXTSWIZZLE_Y:
+   case TGSI_EXTSWIZZLE_Z:
+   case TGSI_EXTSWIZZLE_W:
+      switch (reg->SrcRegister.File) {
+      case TGSI_FILE_CONSTANT:
+         emit_const(
+            func,
+            xmm,
+            reg->SrcRegister.Index,
+            swizzle,
+            reg->SrcRegister.Indirect,
+            reg->SrcRegisterInd.File,
+            reg->SrcRegisterInd.Index );
+         break;
+
+      case TGSI_FILE_IMMEDIATE:
+         emit_immediate(
+            func,
+            xmm,
+            reg->SrcRegister.Index,
+            swizzle );
+         break;
+
+      case TGSI_FILE_INPUT:
+         emit_inputf(
+            func,
+            xmm,
+            reg->SrcRegister.Index,
+            swizzle );
+         break;
+
+      case TGSI_FILE_TEMPORARY:
+         emit_tempf(
+            func,
+            xmm,
+            reg->SrcRegister.Index,
+            swizzle );
+         break;
+
+      default:
+         assert( 0 );
+      }
+      break;
+
+   case TGSI_EXTSWIZZLE_ZERO:
+      emit_tempf(
+         func,
+         xmm,
+         TGSI_EXEC_TEMP_00000000_I,
+         TGSI_EXEC_TEMP_00000000_C );
+      break;
+
+   case TGSI_EXTSWIZZLE_ONE:
+      emit_tempf(
+         func,
+         xmm,
+         TEMP_ONE_I,
+         TEMP_ONE_C );
+      break;
+
+   default:
+      assert( 0 );
+   }
+
+   switch( tgsi_util_get_full_src_register_sign_mode( reg, chan_index ) ) {
+   case TGSI_UTIL_SIGN_CLEAR:
+      emit_abs( func, xmm );
+      break;
+
+   case TGSI_UTIL_SIGN_SET:
+      emit_setsign( func, xmm );
+      break;
+
+   case TGSI_UTIL_SIGN_TOGGLE:
+      emit_neg( func, xmm );
+      break;
+
+   case TGSI_UTIL_SIGN_KEEP:
+      break;
+   }
+}
+
+#define FETCH( FUNC, INST, XMM, INDEX, CHAN )\
+   emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN )
+
+/**
+ * Register store.
+ */
+
+static void
+emit_store(
+   struct x86_function *func,
+   unsigned xmm,
+   const struct tgsi_full_dst_register *reg,
+   const struct tgsi_full_instruction *inst,
+   unsigned chan_index )
+{
+   switch( reg->DstRegister.File ) {
+   case TGSI_FILE_OUTPUT:
+      emit_output(
+         func,
+         xmm,
+         reg->DstRegister.Index,
+         chan_index );
+      break;
+
+   case TGSI_FILE_TEMPORARY:
+      emit_temps(
+         func,
+         xmm,
+         reg->DstRegister.Index,
+         chan_index );
+      break;
+
+   case TGSI_FILE_ADDRESS:
+      emit_addrs(
+         func,
+         xmm,
+         reg->DstRegister.Index,
+         chan_index );
+      break;
+
+   default:
+      assert( 0 );
+   }
+
+   switch( inst->Instruction.Saturate ) {
+   case TGSI_SAT_NONE:
+      break;
+
+   case TGSI_SAT_ZERO_ONE:
+      /* assert( 0 ); */
+      break;
+
+   case TGSI_SAT_MINUS_PLUS_ONE:
+      assert( 0 );
+      break;
+   }
+}
+
+#define STORE( FUNC, INST, XMM, INDEX, CHAN )\
+   emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN )
+
+/**
+ * High-level instruction translators.
+ */
+
+static void
+emit_kil(
+   struct x86_function *func,
+   const struct tgsi_full_src_register *reg )
+{
+   unsigned uniquemask;
+   unsigned registers[4];
+   unsigned nextregister = 0;
+   unsigned firstchan = ~0;
+   unsigned chan_index;
+
+   /* This mask stores component bits that were already tested. Note that
+    * we test if the value is less than zero, so 1.0 and 0.0 need not to be
+    * tested. */
+   uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE);
+
+   FOR_EACH_CHANNEL( chan_index ) {
+      unsigned swizzle;
+
+      /* unswizzle channel */
+      swizzle = tgsi_util_get_full_src_register_extswizzle(
+         reg,
+         chan_index );
+
+      /* check if the component has not been already tested */
+      if( !(uniquemask & (1 << swizzle)) ) {
+         uniquemask |= 1 << swizzle;
+
+         /* allocate register */
+         registers[chan_index] = nextregister;
+         emit_fetch(
+            func,
+            nextregister,
+            reg,
+            chan_index );
+         nextregister++;
+
+         /* mark the first channel used */
+         if( firstchan == ~0 ) {
+            firstchan = chan_index;
+         }
+      }
+   }
+
+   x86_push(
+      func,
+      x86_make_reg( file_REG32, reg_AX ) );
+   x86_push(
+      func,
+      x86_make_reg( file_REG32, reg_DX ) );
+
+   FOR_EACH_CHANNEL( chan_index ) {
+      if( uniquemask & (1 << chan_index) ) {
+         sse_cmpps(
+            func,
+            make_xmm( registers[chan_index] ),
+            get_temp(
+               TGSI_EXEC_TEMP_00000000_I,
+               TGSI_EXEC_TEMP_00000000_C ),
+            cc_LessThan );
+
+         if( chan_index == firstchan ) {
+            sse_pmovmskb(
+               func,
+               x86_make_reg( file_REG32, reg_AX ),
+               make_xmm( registers[chan_index] ) );
+         }
+         else {
+            sse_pmovmskb(
+               func,
+               x86_make_reg( file_REG32, reg_DX ),
+               make_xmm( registers[chan_index] ) );
+            x86_or(
+               func,
+               x86_make_reg( file_REG32, reg_AX ),
+               x86_make_reg( file_REG32, reg_DX ) );
+         }
+      }
+   }
+
+   x86_or(
+      func,
+      get_temp(
+         TGSI_EXEC_TEMP_KILMASK_I,
+         TGSI_EXEC_TEMP_KILMASK_C ),
+      x86_make_reg( file_REG32, reg_AX ) );
+
+   x86_pop(
+      func,
+      x86_make_reg( file_REG32, reg_DX ) );
+   x86_pop(
+      func,
+      x86_make_reg( file_REG32, reg_AX ) );
+}
+
+
+static void
+emit_kilp(
+   struct x86_function *func )
+{
+   /* XXX todo / fix me */
+}
+
+
+static void
+emit_setcc(
+   struct x86_function *func,
+   struct tgsi_full_instruction *inst,
+   enum sse_cc cc )
+{
+   unsigned chan_index;
+
+   FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+      FETCH( func, *inst, 0, 0, chan_index );
+      FETCH( func, *inst, 1, 1, chan_index );
+      sse_cmpps(
+         func,
+         make_xmm( 0 ),
+         make_xmm( 1 ),
+         cc );
+      sse_andps(
+         func,
+         make_xmm( 0 ),
+         get_temp(
+            TEMP_ONE_I,
+            TEMP_ONE_C ) );
+      STORE( func, *inst, 0, 0, chan_index );
+   }
+}
+
+static void
+emit_cmp(
+   struct x86_function *func,
+   struct tgsi_full_instruction *inst )
+{
+   unsigned chan_index;
+
+   FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+      FETCH( func, *inst, 0, 0, chan_index );
+      FETCH( func, *inst, 1, 1, chan_index );
+      FETCH( func, *inst, 2, 2, chan_index );
+      sse_cmpps(
+         func,
+         make_xmm( 0 ),
+         get_temp(
+            TGSI_EXEC_TEMP_00000000_I,
+            TGSI_EXEC_TEMP_00000000_C ),
+         cc_LessThan );
+      sse_andps(
+         func,
+         make_xmm( 1 ),
+         make_xmm( 0 ) );
+      sse_andnps(
+         func,
+         make_xmm( 0 ),
+         make_xmm( 2 ) );
+      sse_orps(
+         func,
+         make_xmm( 0 ),
+         make_xmm( 1 ) );
+      STORE( func, *inst, 0, 0, chan_index );
+   }
+}
+
+static int
+emit_instruction(
+   struct x86_function *func,
+   struct tgsi_full_instruction *inst )
+{
+   unsigned chan_index;
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_ARL:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         emit_f2it( func, 0 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_MOV:
+   case TGSI_OPCODE_SWZ:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_LIT:
+      if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ||
+          IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) {
+         emit_tempf(
+            func,
+            0,
+            TEMP_ONE_I,
+            TEMP_ONE_C);
+         if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ) {
+            STORE( func, *inst, 0, 0, CHAN_X );
+         }
+         if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) {
+            STORE( func, *inst, 0, 0, CHAN_W );
+         }
+      }
+      if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ||
+          IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) {
+         if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) {
+            FETCH( func, *inst, 0, 0, CHAN_X );
+            sse_maxps(
+               func,
+               make_xmm( 0 ),
+               get_temp(
+                  TGSI_EXEC_TEMP_00000000_I,
+                  TGSI_EXEC_TEMP_00000000_C ) );
+            STORE( func, *inst, 0, 0, CHAN_Y );
+         }
+         if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) {
+            /* XMM[1] = SrcReg[0].yyyy */
+            FETCH( func, *inst, 1, 0, CHAN_Y );
+            /* XMM[1] = max(XMM[1], 0) */
+            sse_maxps(
+               func,
+               make_xmm( 1 ),
+               get_temp(
+                  TGSI_EXEC_TEMP_00000000_I,
+                  TGSI_EXEC_TEMP_00000000_C ) );
+            /* XMM[2] = SrcReg[0].wwww */
+            FETCH( func, *inst, 2, 0, CHAN_W );
+            /* XMM[2] = min(XMM[2], 128.0) */
+            sse_minps(
+               func,
+               make_xmm( 2 ),
+               get_temp(
+                  TGSI_EXEC_TEMP_128_I,
+                  TGSI_EXEC_TEMP_128_C ) );
+            /* XMM[2] = max(XMM[2], -128.0) */
+            sse_maxps(
+               func,
+               make_xmm( 2 ),
+               get_temp(
+                  TGSI_EXEC_TEMP_MINUS_128_I,
+                  TGSI_EXEC_TEMP_MINUS_128_C ) );
+            emit_pow( func, 3, 1, 2 );
+            FETCH( func, *inst, 0, 0, CHAN_X );
+            sse_xorps(
+               func,
+               make_xmm( 2 ),
+               make_xmm( 2 ) );
+            sse_cmpps(
+               func,
+               make_xmm( 2 ),
+               make_xmm( 0 ),
+               cc_LessThanEqual );
+            sse_andps(
+               func,
+               make_xmm( 2 ),
+               make_xmm( 1 ) );
+            STORE( func, *inst, 2, 0, CHAN_Z );
+         }
+      }
+      break;
+
+   case TGSI_OPCODE_RCP:
+   /* TGSI_OPCODE_RECIP */
+      FETCH( func, *inst, 0, 0, CHAN_X );
+      emit_rcp( func, 0, 0 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_RSQ:
+   /* TGSI_OPCODE_RECIPSQRT */
+      FETCH( func, *inst, 0, 0, CHAN_X );
+      emit_rsqrt( func, 1, 0 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 1, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_EXP:
+      if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ||
+          IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ||
+          IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+         FETCH( func, *inst, 0, 0, CHAN_X );
+         if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ||
+             IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+            emit_MOV( func, 1, 0 );
+            emit_flr( func, 2, 1 );
+            /* dst.x = ex2(floor(src.x)) */
+            if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X )) {
+               emit_MOV( func, 2, 1 );
+               emit_ex2( func, 3, 2 );
+               STORE( func, *inst, 2, 0, CHAN_X );
+            }
+            /* dst.y = src.x - floor(src.x) */
+            if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+               emit_MOV( func, 2, 0 );
+               emit_sub( func, 2, 1 );
+               STORE( func, *inst, 2, 0, CHAN_Y );
+            }
+         }
+         /* dst.z = ex2(src.x) */
+         if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+            emit_ex2( func, 3, 0 );
+            STORE( func, *inst, 0, 0, CHAN_Z );
+         }
+      }
+      /* dst.w = 1.0 */
+      if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W )) {
+         emit_tempf( func, 0, TEMP_ONE_I, TEMP_ONE_C );
+         STORE( func, *inst, 0, 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_LOG:
+      if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ||
+          IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ||
+          IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+         FETCH( func, *inst, 0, 0, CHAN_X );
+         emit_abs( func, 0 );
+         emit_MOV( func, 1, 0 );
+         emit_lg2( func, 2, 1 );
+         /* dst.z = lg2(abs(src.x)) */
+         if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+            STORE( func, *inst, 1, 0, CHAN_Z );
+         }
+         if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ||
+             IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+            emit_flr( func, 2, 1 );
+            /* dst.x = floor(lg2(abs(src.x))) */
+            if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X )) {
+               STORE( func, *inst, 1, 0, CHAN_X );
+            }
+            /* dst.x = abs(src)/ex2(floor(lg2(abs(src.x)))) */
+            if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+               emit_ex2( func, 2, 1 );
+               emit_rcp( func, 1, 1 );
+               emit_mul( func, 0, 1 );
+               STORE( func, *inst, 0, 0, CHAN_Y );
+            }
+         }
+      }
+      /* dst.w = 1.0 */
+      if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W )) {
+         emit_tempf( func, 0, TEMP_ONE_I, TEMP_ONE_C );
+         STORE( func, *inst, 0, 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_MUL:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         FETCH( func, *inst, 1, 1, chan_index );
+         emit_mul( func, 0, 1 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_ADD:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         FETCH( func, *inst, 1, 1, chan_index );
+         emit_add( func, 0, 1 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DP3:
+   /* TGSI_OPCODE_DOT3 */
+      FETCH( func, *inst, 0, 0, CHAN_X );
+      FETCH( func, *inst, 1, 1, CHAN_X );
+      emit_mul( func, 0, 1 );
+      FETCH( func, *inst, 1, 0, CHAN_Y );
+      FETCH( func, *inst, 2, 1, CHAN_Y );
+      emit_mul( func, 1, 2 );
+      emit_add( func, 0, 1 );
+      FETCH( func, *inst, 1, 0, CHAN_Z );
+      FETCH( func, *inst, 2, 1, CHAN_Z );
+      emit_mul( func, 1, 2 );
+      emit_add( func, 0, 1 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DP4:
+   /* TGSI_OPCODE_DOT4 */
+      FETCH( func, *inst, 0, 0, CHAN_X );
+      FETCH( func, *inst, 1, 1, CHAN_X );
+      emit_mul( func, 0, 1 );
+      FETCH( func, *inst, 1, 0, CHAN_Y );
+      FETCH( func, *inst, 2, 1, CHAN_Y );
+      emit_mul( func, 1, 2 );
+      emit_add( func, 0, 1 );
+      FETCH( func, *inst, 1, 0, CHAN_Z );
+      FETCH( func, *inst, 2, 1, CHAN_Z );
+      emit_mul(func, 1, 2 );
+      emit_add(func, 0, 1 );
+      FETCH( func, *inst, 1, 0, CHAN_W );
+      FETCH( func, *inst, 2, 1, CHAN_W );
+      emit_mul( func, 1, 2 );
+      emit_add( func, 0, 1 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DST:
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) {
+         emit_tempf(
+            func,
+            0,
+            TEMP_ONE_I,
+            TEMP_ONE_C );
+         STORE( func, *inst, 0, 0, CHAN_X );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) {
+         FETCH( func, *inst, 0, 0, CHAN_Y );
+         FETCH( func, *inst, 1, 1, CHAN_Y );
+         emit_mul( func, 0, 1 );
+         STORE( func, *inst, 0, 0, CHAN_Y );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) {
+         FETCH( func, *inst, 0, 0, CHAN_Z );
+         STORE( func, *inst, 0, 0, CHAN_Z );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) {
+         FETCH( func, *inst, 0, 1, CHAN_W );
+         STORE( func, *inst, 0, 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_MIN:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         FETCH( func, *inst, 1, 1, chan_index );
+         sse_minps(
+            func,
+            make_xmm( 0 ),
+            make_xmm( 1 ) );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_MAX:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         FETCH( func, *inst, 1, 1, chan_index );
+         sse_maxps(
+            func,
+            make_xmm( 0 ),
+            make_xmm( 1 ) );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SLT:
+   /* TGSI_OPCODE_SETLT */
+      emit_setcc( func, inst, cc_LessThan );
+      break;
+
+   case TGSI_OPCODE_SGE:
+   /* TGSI_OPCODE_SETGE */
+      emit_setcc( func, inst, cc_NotLessThan );
+      break;
+
+   case TGSI_OPCODE_MAD:
+   /* TGSI_OPCODE_MADD */
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         FETCH( func, *inst, 1, 1, chan_index );
+         FETCH( func, *inst, 2, 2, chan_index );
+         emit_mul( func, 0, 1 );
+         emit_add( func, 0, 2 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SUB:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         FETCH( func, *inst, 1, 1, chan_index );
+         emit_sub( func, 0, 1 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_LERP:
+   /* TGSI_OPCODE_LRP */
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         FETCH( func, *inst, 1, 1, chan_index );
+         FETCH( func, *inst, 2, 2, chan_index );
+         emit_sub( func, 1, 2 );
+         emit_mul( func, 0, 1 );
+         emit_add( func, 0, 2 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_CND:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_CND0:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_DOT2ADD:
+   /* TGSI_OPCODE_DP2A */
+      FETCH( func, *inst, 0, 0, CHAN_X );  /* xmm0 = src[0].x */
+      FETCH( func, *inst, 1, 1, CHAN_X );  /* xmm1 = src[1].x */
+      emit_mul( func, 0, 1 );              /* xmm0 = xmm0 * xmm1 */
+      FETCH( func, *inst, 1, 0, CHAN_Y );  /* xmm1 = src[0].y */
+      FETCH( func, *inst, 2, 1, CHAN_Y );  /* xmm2 = src[1].y */
+      emit_mul( func, 1, 2 );              /* xmm1 = xmm1 * xmm2 */
+      emit_add( func, 0, 1 );              /* xmm0 = xmm0 + xmm1 */
+      FETCH( func, *inst, 1, 2, CHAN_X );  /* xmm1 = src[2].x */
+      emit_add( func, 0, 1 );              /* xmm0 = xmm0 + xmm1 */
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );  /* dest[ch] = xmm0 */
+      }
+      break;
+
+   case TGSI_OPCODE_INDEX:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_NEGATE:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_FRAC:
+   /* TGSI_OPCODE_FRC */
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         emit_frc( func, 0, 0 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_CLAMP:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_FLOOR:
+   /* TGSI_OPCODE_FLR */
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         emit_flr( func, 0, 0 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_ROUND:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         emit_rnd( func, 0, 0 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_EXPBASE2:
+   /* TGSI_OPCODE_EX2 */
+      FETCH( func, *inst, 0, 0, CHAN_X );
+      emit_ex2( func, 0, 0 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_LOGBASE2:
+   /* TGSI_OPCODE_LG2 */
+      FETCH( func, *inst, 0, 0, CHAN_X );
+      emit_lg2( func, 0, 0 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_POWER:
+   /* TGSI_OPCODE_POW */
+      FETCH( func, *inst, 0, 0, CHAN_X );
+      FETCH( func, *inst, 1, 1, CHAN_X );
+      emit_pow( func, 0, 0, 1 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_CROSSPRODUCT:
+   /* TGSI_OPCODE_XPD */
+      if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ||
+          IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) {
+         FETCH( func, *inst, 1, 1, CHAN_Z );
+         FETCH( func, *inst, 3, 0, CHAN_Z );
+      }
+      if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ||
+          IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) {
+         FETCH( func, *inst, 0, 0, CHAN_Y );
+         FETCH( func, *inst, 4, 1, CHAN_Y );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) {
+         emit_MOV( func, 2, 0 );
+         emit_mul( func, 2, 1 );
+         emit_MOV( func, 5, 3 );
+         emit_mul( func, 5, 4 );
+         emit_sub( func, 2, 5 );
+         STORE( func, *inst, 2, 0, CHAN_X );
+      }
+      if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ||
+          IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) {
+         FETCH( func, *inst, 2, 1, CHAN_X );
+         FETCH( func, *inst, 5, 0, CHAN_X );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) {
+         emit_mul( func, 3, 2 );
+         emit_mul( func, 1, 5 );
+         emit_sub( func, 3, 1 );
+         STORE( func, *inst, 3, 0, CHAN_Y );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) {
+         emit_mul( func, 5, 4 );
+         emit_mul( func, 0, 2 );
+         emit_sub( func, 5, 0 );
+         STORE( func, *inst, 5, 0, CHAN_Z );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) {
+	 emit_tempf(
+	    func,
+	    0,
+	    TEMP_ONE_I,
+	    TEMP_ONE_C );
+         STORE( func, *inst, 0, 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_MULTIPLYMATRIX:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ABS:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         emit_abs( func, 0) ;
+
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_RCC:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_DPH:
+      FETCH( func, *inst, 0, 0, CHAN_X );
+      FETCH( func, *inst, 1, 1, CHAN_X );
+      emit_mul( func, 0, 1 );
+      FETCH( func, *inst, 1, 0, CHAN_Y );
+      FETCH( func, *inst, 2, 1, CHAN_Y );
+      emit_mul( func, 1, 2 );
+      emit_add( func, 0, 1 );
+      FETCH( func, *inst, 1, 0, CHAN_Z );
+      FETCH( func, *inst, 2, 1, CHAN_Z );
+      emit_mul( func, 1, 2 );
+      emit_add( func, 0, 1 );
+      FETCH( func, *inst, 1, 1, CHAN_W );
+      emit_add( func, 0, 1 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_COS:
+      FETCH( func, *inst, 0, 0, CHAN_X );
+      emit_cos( func, 0, 0 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DDX:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_DDY:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_KILP:
+      /* predicated kill */
+      emit_kilp( func );
+      return 0; /* XXX fix me */
+      break;
+
+   case TGSI_OPCODE_KIL:
+      /* conditional kill */
+      emit_kil( func, &inst->FullSrcRegisters[0] );
+      break;
+
+   case TGSI_OPCODE_PK2H:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_PK2US:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_PK4B:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_PK4UB:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_RFL:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SEQ:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SFL:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SGT:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SIN:
+      FETCH( func, *inst, 0, 0, CHAN_X );
+      emit_sin( func, 0, 0 );
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SLE:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SNE:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_STR:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_TEX:
+      if (0) {
+	 /* Disable dummy texture code: 
+	  */
+	 emit_tempf(
+	    func,
+	    0,
+	    TEMP_ONE_I,
+	    TEMP_ONE_C );
+	 FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+	    STORE( func, *inst, 0, 0, chan_index );
+	 }
+      }
+      else {
+	 return 0;
+      }
+      break;
+
+   case TGSI_OPCODE_TXD:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_UP2H:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_UP2US:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_UP4B:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_UP4UB:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_X2D:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ARA:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ARR:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         emit_rnd( func, 0, 0 );
+         emit_f2it( func, 0 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_BRA:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_CAL:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_RET:
+      emit_ret( func );
+      break;
+
+   case TGSI_OPCODE_END:
+      break;
+
+   case TGSI_OPCODE_SSG:
+   /* TGSI_OPCODE_SGN */
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         emit_sgn( func, 0, 0 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_CMP:
+      emit_cmp (func, inst);
+      break;
+
+   case TGSI_OPCODE_SCS:
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) {
+         FETCH( func, *inst, 0, 0, CHAN_X );
+         emit_cos( func, 0, 0 );
+         STORE( func, *inst, 0, 0, CHAN_X );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) {
+         FETCH( func, *inst, 0, 0, CHAN_X );
+         emit_sin( func, 0, 0 );
+         STORE( func, *inst, 0, 0, CHAN_Y );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) {
+	 emit_tempf(
+	    func,
+	    0,
+	    TGSI_EXEC_TEMP_00000000_I,
+	    TGSI_EXEC_TEMP_00000000_C );
+         STORE( func, *inst, 0, 0, CHAN_Z );
+      }
+      IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) {
+	 emit_tempf(
+	    func,
+	    0,
+	    TEMP_ONE_I,
+	    TEMP_ONE_C );
+         STORE( func, *inst, 0, 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_TXB:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_NRM:
+      /* fall-through */
+   case TGSI_OPCODE_NRM4:
+      /* 3 or 4-component normalization */
+      {
+         uint dims = (inst->Instruction.Opcode == TGSI_OPCODE_NRM) ? 3 : 4;
+         /* note: cannot use xmm regs 2/3 here (see emit_rsqrt() above) */
+         FETCH( func, *inst, 4, 0, CHAN_X );    /* xmm4 = src[0].x */
+         FETCH( func, *inst, 5, 0, CHAN_Y );    /* xmm5 = src[0].y */
+         FETCH( func, *inst, 6, 0, CHAN_Z );    /* xmm6 = src[0].z */
+         if (dims == 4) {
+            FETCH( func, *inst, 7, 0, CHAN_W ); /* xmm7 = src[0].w */
+         }
+         emit_MOV( func, 0, 4 );                /* xmm0 = xmm3 */
+         emit_mul( func, 0, 4 );                /* xmm0 *= xmm3 */
+         emit_MOV( func, 1, 5 );                /* xmm1 = xmm4 */
+         emit_mul( func, 1, 5 );                /* xmm1 *= xmm4 */
+         emit_add( func, 0, 1 );                /* xmm0 += xmm1 */
+         emit_MOV( func, 1, 6 );                /* xmm1 = xmm5 */
+         emit_mul( func, 1, 6 );                /* xmm1 *= xmm5 */
+         emit_add( func, 0, 1 );                /* xmm0 += xmm1 */
+         if (dims == 4) {
+            emit_MOV( func, 1, 7 );             /* xmm1 = xmm7 */
+            emit_mul( func, 1, 7 );             /* xmm1 *= xmm7 */
+            emit_add( func, 0, 0 );             /* xmm0 += xmm1 */
+         }
+         emit_rsqrt( func, 1, 0 );              /* xmm1 = 1/sqrt(xmm0) */
+         FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+            if (chan_index < dims) {
+               emit_mul( func, 4+chan_index, 1); /* xmm[4+ch] *= xmm1 */
+               STORE( func, *inst, 4+chan_index, 0, chan_index );
+            }
+         }
+      }
+      break;
+
+   case TGSI_OPCODE_DIV:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_DP2:
+      FETCH( func, *inst, 0, 0, CHAN_X );  /* xmm0 = src[0].x */
+      FETCH( func, *inst, 1, 1, CHAN_X );  /* xmm1 = src[1].x */
+      emit_mul( func, 0, 1 );              /* xmm0 = xmm0 * xmm1 */
+      FETCH( func, *inst, 1, 0, CHAN_Y );  /* xmm1 = src[0].y */
+      FETCH( func, *inst, 2, 1, CHAN_Y );  /* xmm2 = src[1].y */
+      emit_mul( func, 1, 2 );              /* xmm1 = xmm1 * xmm2 */
+      emit_add( func, 0, 1 );              /* xmm0 = xmm0 + xmm1 */
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( func, *inst, 0, 0, chan_index );  /* dest[ch] = xmm0 */
+      }
+      break;
+
+   case TGSI_OPCODE_TXL:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_BRK:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_IF:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_LOOP:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_REP:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ELSE:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ENDIF:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ENDLOOP:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ENDREP:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_PUSHA:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_POPA:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_CEIL:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_I2F:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_NOT:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_TRUNC:
+      FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( func, *inst, 0, 0, chan_index );
+         emit_f2it( func, 0 );
+         emit_i2f( func, 0 );
+         STORE( func, *inst, 0, 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SHL:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SHR:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_AND:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_OR:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_MOD:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_XOR:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_SAD:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_TXF:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_TXQ:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_CONT:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_EMIT:
+      return 0;
+      break;
+
+   case TGSI_OPCODE_ENDPRIM:
+      return 0;
+      break;
+
+   default:
+      return 0;
+   }
+   
+   return 1;
+}
+
+static void
+emit_declaration(
+   struct x86_function *func,
+   struct tgsi_full_declaration *decl )
+{
+   if( decl->Declaration.File == TGSI_FILE_INPUT ) {
+      unsigned first, last, mask;
+      unsigned i, j;
+
+      first = decl->DeclarationRange.First;
+      last = decl->DeclarationRange.Last;
+      mask = decl->Declaration.UsageMask;
+
+      for( i = first; i <= last; i++ ) {
+         for( j = 0; j < NUM_CHANNELS; j++ ) {
+            if( mask & (1 << j) ) {
+               switch( decl->Declaration.Interpolate ) {
+               case TGSI_INTERPOLATE_CONSTANT:
+                  emit_coef_a0( func, 0, i, j );
+                  emit_inputs( func, 0, i, j );
+                  break;
+
+               case TGSI_INTERPOLATE_LINEAR:
+                  emit_tempf( func, 0, 0, TGSI_SWIZZLE_X );
+                  emit_coef_dadx( func, 1, i, j );
+                  emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y );
+                  emit_coef_dady( func, 3, i, j );
+                  emit_mul( func, 0, 1 );    /* x * dadx */
+                  emit_coef_a0( func, 4, i, j );
+                  emit_mul( func, 2, 3 );    /* y * dady */
+                  emit_add( func, 0, 4 );    /* x * dadx + a0 */
+                  emit_add( func, 0, 2 );    /* x * dadx + y * dady + a0 */
+                  emit_inputs( func, 0, i, j );
+                  break;
+
+               case TGSI_INTERPOLATE_PERSPECTIVE:
+                  emit_tempf( func, 0, 0, TGSI_SWIZZLE_X );
+                  emit_coef_dadx( func, 1, i, j );
+                  emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y );
+                  emit_coef_dady( func, 3, i, j );
+                  emit_mul( func, 0, 1 );    /* x * dadx */
+                  emit_tempf( func, 4, 0, TGSI_SWIZZLE_W );
+                  emit_coef_a0( func, 5, i, j );
+                  emit_rcp( func, 4, 4 );    /* 1.0 / w */
+                  emit_mul( func, 2, 3 );    /* y * dady */
+                  emit_add( func, 0, 5 );    /* x * dadx + a0 */
+                  emit_add( func, 0, 2 );    /* x * dadx + y * dady + a0 */
+                  emit_mul( func, 0, 4 );    /* (x * dadx + y * dady + a0) / w */
+                  emit_inputs( func, 0, i, j );
+                  break;
+
+               default:
+                  assert( 0 );
+		  break;
+               }
+            }
+         }
+      }
+   }
+}
+
+static void aos_to_soa( struct x86_function *func, 
+                        uint arg_aos,
+                        uint arg_soa, 
+                        uint arg_num, 
+                        uint arg_stride )
+{
+   struct x86_reg soa_input = x86_make_reg( file_REG32, reg_AX );
+   struct x86_reg aos_input = x86_make_reg( file_REG32, reg_BX );
+   struct x86_reg num_inputs = x86_make_reg( file_REG32, reg_CX );
+   struct x86_reg stride = x86_make_reg( file_REG32, reg_DX );
+   int inner_loop;
+
+
+   /* Save EBX */
+   x86_push( func, x86_make_reg( file_REG32, reg_BX ) );
+
+   x86_mov( func, aos_input,  x86_fn_arg( func, arg_aos ) );
+   x86_mov( func, soa_input,  x86_fn_arg( func, arg_soa ) );
+   x86_mov( func, num_inputs, x86_fn_arg( func, arg_num ) );
+   x86_mov( func, stride,     x86_fn_arg( func, arg_stride ) );
+
+   /* do */
+   inner_loop = x86_get_label( func );
+   {
+      x86_push( func, aos_input );
+      sse_movlps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) );
+      sse_movlps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) );
+      x86_add( func, aos_input, stride );
+      sse_movhps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) );
+      sse_movhps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) );
+      x86_add( func, aos_input, stride );
+      sse_movlps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) );
+      sse_movlps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) );
+      x86_add( func, aos_input, stride );
+      sse_movhps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) );
+      sse_movhps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) );
+      x86_pop( func, aos_input );
+
+      sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) );
+      sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) );
+      sse_shufps( func, make_xmm( 0 ), make_xmm( 1 ), 0x88 );
+      sse_shufps( func, make_xmm( 2 ), make_xmm( 1 ), 0xdd );
+      sse_shufps( func, make_xmm( 3 ), make_xmm( 4 ), 0x88 );
+      sse_shufps( func, make_xmm( 5 ), make_xmm( 4 ), 0xdd );
+
+      sse_movups( func, x86_make_disp( soa_input, 0 ), make_xmm( 0 ) );
+      sse_movups( func, x86_make_disp( soa_input, 16 ), make_xmm( 2 ) );
+      sse_movups( func, x86_make_disp( soa_input, 32 ), make_xmm( 3 ) );
+      sse_movups( func, x86_make_disp( soa_input, 48 ), make_xmm( 5 ) );
+
+      /* Advance to next input */
+      x86_lea( func, aos_input, x86_make_disp(aos_input, 16) );
+      x86_lea( func, soa_input, x86_make_disp(soa_input, 64) );
+   }
+   /* while --num_inputs */
+   x86_dec( func, num_inputs );
+   x86_jcc( func, cc_NE, inner_loop );
+
+   /* Restore EBX */
+   x86_pop( func, aos_input );
+}
+
+static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, uint stride )
+{
+   struct x86_reg soa_output;
+   struct x86_reg aos_output;
+   struct x86_reg num_outputs;
+   struct x86_reg temp;
+   int inner_loop;
+
+   soa_output = x86_make_reg( file_REG32, reg_AX );
+   aos_output = x86_make_reg( file_REG32, reg_BX );
+   num_outputs = x86_make_reg( file_REG32, reg_CX );
+   temp = x86_make_reg( file_REG32, reg_DX );
+
+   /* Save EBX */
+   x86_push( func, aos_output );
+
+   x86_mov( func, soa_output, x86_fn_arg( func, soa ) );
+   x86_mov( func, aos_output, x86_fn_arg( func, aos ) );
+   x86_mov( func, num_outputs, x86_fn_arg( func, num ) );
+
+   /* do */
+   inner_loop = x86_get_label( func );
+   {
+      sse_movups( func, make_xmm( 0 ), x86_make_disp( soa_output, 0 ) );
+      sse_movups( func, make_xmm( 1 ), x86_make_disp( soa_output, 16 ) );
+      sse_movups( func, make_xmm( 3 ), x86_make_disp( soa_output, 32 ) );
+      sse_movups( func, make_xmm( 4 ), x86_make_disp( soa_output, 48 ) );
+
+      sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) );
+      sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) );
+      sse_unpcklps( func, make_xmm( 0 ), make_xmm( 1 ) );
+      sse_unpckhps( func, make_xmm( 2 ), make_xmm( 1 ) );
+      sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) );
+      sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) );
+
+      x86_mov( func, temp, x86_fn_arg( func, stride ) );
+      x86_push( func, aos_output );
+      sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) );
+      sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) );
+      x86_add( func, aos_output, temp );
+      sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) );
+      sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) );
+      x86_add( func, aos_output, temp );
+      sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) );
+      sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) );
+      x86_add( func, aos_output, temp );
+      sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) );
+      sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) );
+      x86_pop( func, aos_output );
+
+      /* Advance to next output */
+      x86_lea( func, aos_output, x86_make_disp(aos_output, 16) );
+      x86_lea( func, soa_output, x86_make_disp(soa_output, 64) );
+   }
+   /* while --num_outputs */
+   x86_dec( func, num_outputs );
+   x86_jcc( func, cc_NE, inner_loop );
+
+   /* Restore EBX */
+   x86_pop( func, aos_output );
+}
+
+/**
+ * Translate a TGSI vertex/fragment shader to SSE2 code.
+ * Slightly different things are done for vertex vs. fragment shaders.
+ *
+ * Note that fragment shaders are responsible for interpolating shader
+ * inputs. Because on x86 we have only 4 GP registers, and here we
+ * have 5 shader arguments (input, output, const, temp and coef), the
+ * code is split into two phases -- DECLARATION and INSTRUCTION phase.
+ * GP register holding the output argument is aliased with the coeff
+ * argument, as outputs are not needed in the DECLARATION phase.
+ *
+ * \param tokens  the TGSI input shader
+ * \param func  the output SSE code/function
+ * \param immediates  buffer to place immediates, later passed to SSE func
+ * \param return  1 for success, 0 if translation failed
+ */
+unsigned
+tgsi_emit_sse2(
+   const struct tgsi_token *tokens,
+   struct x86_function *func,
+   float (*immediates)[4],
+   boolean do_swizzles )
+{
+   struct tgsi_parse_context parse;
+   boolean instruction_phase = FALSE;
+   unsigned ok = 1;
+   uint num_immediates = 0;
+
+   util_init_math();
+
+   func->csr = func->store;
+
+   tgsi_parse_init( &parse, tokens );
+
+   /* Can't just use EDI, EBX without save/restoring them:
+    */
+   x86_push(
+      func,
+      get_immediate_base() );
+
+   x86_push(
+      func,
+      get_temp_base() );
+
+
+   /*
+    * Different function args for vertex/fragment shaders:
+    */
+   if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
+      /* DECLARATION phase, do not load output argument. */
+      x86_mov(
+         func,
+         get_input_base(),
+         x86_fn_arg( func, 1 ) );
+      /* skipping outputs argument here */
+      x86_mov(
+         func,
+         get_const_base(),
+         x86_fn_arg( func, 3 ) );
+      x86_mov(
+         func,
+         get_temp_base(),
+         x86_fn_arg( func, 4 ) );
+      x86_mov(
+         func,
+         get_coef_base(),
+         x86_fn_arg( func, 5 ) );
+      x86_mov(
+         func,
+         get_immediate_base(),
+         x86_fn_arg( func, 6 ) );
+   }
+   else {
+      assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX);
+
+      if (do_swizzles)
+         aos_to_soa( func, 
+                     6,         /* aos_input */
+                     1,         /* machine->input */
+                     7,         /* num_inputs */
+                     8 );       /* input_stride */
+
+      x86_mov(
+         func,
+         get_input_base(),
+         x86_fn_arg( func, 1 ) );
+      x86_mov(
+         func,
+         get_output_base(),
+         x86_fn_arg( func, 2 ) );
+      x86_mov(
+         func,
+         get_const_base(),
+         x86_fn_arg( func, 3 ) );
+      x86_mov(
+         func,
+         get_temp_base(),
+         x86_fn_arg( func, 4 ) );
+      x86_mov(
+         func,
+         get_immediate_base(),
+         x86_fn_arg( func, 5 ) );
+   }
+
+   while( !tgsi_parse_end_of_tokens( &parse ) && ok ) {
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
+            emit_declaration(
+               func,
+               &parse.FullToken.FullDeclaration );
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
+            if( !instruction_phase ) {
+               /* INSTRUCTION phase, overwrite coeff with output. */
+               instruction_phase = TRUE;
+               x86_mov(
+                  func,
+                  get_output_base(),
+                  x86_fn_arg( func, 2 ) );
+            }
+         }
+
+         ok = emit_instruction(
+            func,
+            &parse.FullToken.FullInstruction );
+
+	 if (!ok) {
+	    debug_printf("failed to translate tgsi opcode %d to SSE (%s)\n", 
+			 parse.FullToken.FullInstruction.Instruction.Opcode,
+                         parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ?
+                         "vertex shader" : "fragment shader");
+	 }
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         /* simply copy the immediate values into the next immediates[] slot */
+         {
+            const uint size = parse.FullToken.FullImmediate.Immediate.NrTokens - 1;
+            uint i;
+            assert(size <= 4);
+            assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES);
+            for( i = 0; i < size; i++ ) {
+               immediates[num_immediates][i] =
+		  parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float;
+            }
+#if 0
+            debug_printf("SSE FS immediate[%d] = %f %f %f %f\n",
+                   num_immediates,
+                   immediates[num_immediates][0],
+                   immediates[num_immediates][1],
+                   immediates[num_immediates][2],
+                   immediates[num_immediates][3]);
+#endif
+            num_immediates++;
+         }
+         break;
+
+      default:
+	 ok = 0;
+         assert( 0 );
+      }
+   }
+
+   if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) {
+      if (do_swizzles)
+         soa_to_aos( func, 9, 2, 10, 11 );
+   }
+
+   /* Can't just use EBX, EDI without save/restoring them:
+    */
+   x86_pop(
+      func,
+      get_temp_base() );
+
+   x86_pop(
+      func,
+      get_immediate_base() );
+
+   emit_ret( func );
+
+   tgsi_parse_free( &parse );
+
+   return ok;
+}
+
+#endif /* PIPE_ARCH_X86 */
+
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/tgsi_sse2.h
new file mode 100644
index 0000000000..af838b2a25
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.h
@@ -0,0 +1,49 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_SSE2_H
+#define TGSI_SSE2_H
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+struct tgsi_token;
+struct x86_function;
+
+unsigned
+tgsi_emit_sse2(
+   const struct tgsi_token *tokens,
+   struct x86_function *function,
+   float (*immediates)[4],
+   boolean do_swizzles );
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_SSE2_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c
new file mode 100644
index 0000000000..1e822fbbea
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.c
@@ -0,0 +1,1097 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_debug.h"
+#include "tgsi_text.h"
+#include "tgsi_build.h"
+#include "tgsi_info.h"
+#include "tgsi_parse.h"
+#include "tgsi_sanity.h"
+#include "tgsi_util.h"
+
+static boolean is_alpha_underscore( const char *cur )
+{
+   return
+      (*cur >= 'a' && *cur <= 'z') ||
+      (*cur >= 'A' && *cur <= 'Z') ||
+      *cur == '_';
+}
+
+static boolean is_digit( const char *cur )
+{
+   return *cur >= '0' && *cur <= '9';
+}
+
+static boolean is_digit_alpha_underscore( const char *cur )
+{
+   return is_digit( cur ) || is_alpha_underscore( cur );
+}
+
+static boolean uprcase( char c )
+{
+   if (c >= 'a' && c <= 'z')
+      return c += 'A' - 'a';
+   return c;
+}
+
+static boolean str_match_no_case( const char **pcur, const char *str )
+{
+   const char *cur = *pcur;
+
+   while (*str != '\0' && *str == uprcase( *cur )) {
+      str++;
+      cur++;
+   }
+   if (*str == '\0') {
+      *pcur = cur;
+      return TRUE;
+   }
+   return FALSE;
+}
+
+/* Eat zero or more whitespaces.
+ */
+static void eat_opt_white( const char **pcur )
+{
+   while (**pcur == ' ' || **pcur == '\t' || **pcur == '\n')
+      (*pcur)++;
+}
+
+/* Eat one or more whitespaces.
+ * Return TRUE if at least one whitespace eaten.
+ */
+static boolean eat_white( const char **pcur )
+{
+   const char *cur = *pcur;
+
+   eat_opt_white( pcur );
+   return *pcur > cur;
+}
+
+/* Parse unsigned integer.
+ * No checks for overflow.
+ */
+static boolean parse_uint( const char **pcur, uint *val )
+{
+   const char *cur = *pcur;
+
+   if (is_digit( cur )) {
+      *val = *cur++ - '0';
+      while (is_digit( cur ))
+         *val = *val * 10 + *cur++ - '0';
+      *pcur = cur;
+      return TRUE;
+   }
+   return FALSE;
+}
+
+/* Parse floating point.
+ */
+static boolean parse_float( const char **pcur, float *val )
+{
+   const char *cur = *pcur;
+   boolean integral_part = FALSE;
+   boolean fractional_part = FALSE;
+
+   *val = (float) atof( cur );
+
+   if (*cur == '-' || *cur == '+')
+      cur++;
+   if (is_digit( cur )) {
+      cur++;
+      integral_part = TRUE;
+      while (is_digit( cur ))
+         cur++;
+   }
+   if (*cur == '.') {
+      cur++;
+      if (is_digit( cur )) {
+         cur++;
+         fractional_part = TRUE;
+         while (is_digit( cur ))
+            cur++;
+      }
+   }
+   if (!integral_part && !fractional_part)
+      return FALSE;
+   if (uprcase( *cur ) == 'E') {
+      cur++;
+      if (*cur == '-' || *cur == '+')
+         cur++;
+      if (is_digit( cur )) {
+         cur++;
+         while (is_digit( cur ))
+            cur++;
+      }
+      else
+         return FALSE;
+   }
+   *pcur = cur;
+   return TRUE;
+}
+
+struct translate_ctx
+{
+   const char *text;
+   const char *cur;
+   struct tgsi_token *tokens;
+   struct tgsi_token *tokens_cur;
+   struct tgsi_token *tokens_end;
+   struct tgsi_header *header;
+};
+
+static void report_error( struct translate_ctx *ctx, const char *msg )
+{
+   debug_printf( "\nError: %s", msg );
+}
+
+/* Parse shader header.
+ * Return TRUE for one of the following headers.
+ *    FRAG1.1
+ *    GEOM1.1
+ *    VERT1.1
+ */
+static boolean parse_header( struct translate_ctx *ctx )
+{
+   uint processor;
+
+   if (str_match_no_case( &ctx->cur, "FRAG1.1" ))
+      processor = TGSI_PROCESSOR_FRAGMENT;
+   else if (str_match_no_case( &ctx->cur, "VERT1.1" ))
+      processor = TGSI_PROCESSOR_VERTEX;
+   else if (str_match_no_case( &ctx->cur, "GEOM1.1" ))
+      processor = TGSI_PROCESSOR_GEOMETRY;
+   else {
+      report_error( ctx, "Unknown header" );
+      return FALSE;
+   }
+
+   if (ctx->tokens_cur >= ctx->tokens_end)
+      return FALSE;
+   *(struct tgsi_version *) ctx->tokens_cur++ = tgsi_build_version();
+
+   if (ctx->tokens_cur >= ctx->tokens_end)
+      return FALSE;
+   ctx->header = (struct tgsi_header *) ctx->tokens_cur++;
+   *ctx->header = tgsi_build_header();
+
+   if (ctx->tokens_cur >= ctx->tokens_end)
+      return FALSE;
+   *(struct tgsi_processor *) ctx->tokens_cur++ = tgsi_build_processor( processor, ctx->header );
+
+   return TRUE;
+}
+
+static boolean parse_label( struct translate_ctx *ctx, uint *val )
+{
+   const char *cur = ctx->cur;
+
+   if (parse_uint( &cur, val )) {
+      eat_opt_white( &cur );
+      if (*cur == ':') {
+         cur++;
+         ctx->cur = cur;
+         return TRUE;
+      }
+   }
+   return FALSE;
+}
+
+static const char *file_names[TGSI_FILE_COUNT] =
+{
+   "NULL",
+   "CONST",
+   "IN",
+   "OUT",
+   "TEMP",
+   "SAMP",
+   "ADDR",
+   "IMM"
+};
+
+static boolean
+parse_file( const char **pcur, uint *file )
+{
+   uint i;
+
+   for (i = 0; i < TGSI_FILE_COUNT; i++) {
+      const char *cur = *pcur;
+
+      if (str_match_no_case( &cur, file_names[i] )) {
+         if (!is_digit_alpha_underscore( cur )) {
+            *pcur = cur;
+            *file = i;
+            return TRUE;
+         }
+      }
+   }
+   return FALSE;
+}
+
+static boolean
+parse_opt_writemask(
+   struct translate_ctx *ctx,
+   uint *writemask )
+{
+   const char *cur;
+
+   cur = ctx->cur;
+   eat_opt_white( &cur );
+   if (*cur == '.') {
+      cur++;
+      *writemask = TGSI_WRITEMASK_NONE;
+      eat_opt_white( &cur );
+      if (uprcase( *cur ) == 'X') {
+         cur++;
+         *writemask |= TGSI_WRITEMASK_X;
+      }
+      if (uprcase( *cur ) == 'Y') {
+         cur++;
+         *writemask |= TGSI_WRITEMASK_Y;
+      }
+      if (uprcase( *cur ) == 'Z') {
+         cur++;
+         *writemask |= TGSI_WRITEMASK_Z;
+      }
+      if (uprcase( *cur ) == 'W') {
+         cur++;
+         *writemask |= TGSI_WRITEMASK_W;
+      }
+
+      if (*writemask == TGSI_WRITEMASK_NONE) {
+         report_error( ctx, "Writemask expected" );
+         return FALSE;
+      }
+
+      ctx->cur = cur;
+   }
+   else {
+      *writemask = TGSI_WRITEMASK_XYZW;
+   }
+   return TRUE;
+}
+
+/* <register_file_bracket> ::= <file> `['
+ */
+static boolean
+parse_register_file_bracket(
+   struct translate_ctx *ctx,
+   uint *file )
+{
+   if (!parse_file( &ctx->cur, file )) {
+      report_error( ctx, "Unknown register file" );
+      return FALSE;
+   }
+   eat_opt_white( &ctx->cur );
+   if (*ctx->cur != '[') {
+      report_error( ctx, "Expected `['" );
+      return FALSE;
+   }
+   ctx->cur++;
+   return TRUE;
+}
+
+/* <register_file_bracket_index> ::= <register_file_bracket> <uint>
+ */
+static boolean
+parse_register_file_bracket_index(
+   struct translate_ctx *ctx,
+   uint *file,
+   int *index )
+{
+   uint uindex;
+
+   if (!parse_register_file_bracket( ctx, file ))
+      return FALSE;
+   eat_opt_white( &ctx->cur );
+   if (!parse_uint( &ctx->cur, &uindex )) {
+      report_error( ctx, "Expected literal unsigned integer" );
+      return FALSE;
+   }
+   *index = (int) uindex;
+   return TRUE;
+}
+
+/* Parse destination register operand.
+ *    <register_dst> ::= <register_file_bracket_index> `]'
+ */
+static boolean
+parse_register_dst(
+   struct translate_ctx *ctx,
+   uint *file,
+   int *index )
+{
+   if (!parse_register_file_bracket_index( ctx, file, index ))
+      return FALSE;
+   eat_opt_white( &ctx->cur );
+   if (*ctx->cur != ']') {
+      report_error( ctx, "Expected `]'" );
+      return FALSE;
+   }
+   ctx->cur++;
+   return TRUE;
+}
+
+/* Parse source register operand.
+ *    <register_src> ::= <register_file_bracket_index> `]' |
+ *                       <register_file_bracket> <register_dst> `]' |
+ *                       <register_file_bracket> <register_dst> `+' <uint> `]' |
+ *                       <register_file_bracket> <register_dst> `-' <uint> `]'
+ */
+static boolean
+parse_register_src(
+   struct translate_ctx *ctx,
+   uint *file,
+   int *index,
+   uint *ind_file,
+   int *ind_index )
+{
+   const char *cur;
+   uint uindex;
+
+   if (!parse_register_file_bracket( ctx, file ))
+      return FALSE;
+   eat_opt_white( &ctx->cur );
+   cur = ctx->cur;
+   if (parse_file( &cur, ind_file )) {
+      if (!parse_register_dst( ctx, ind_file, ind_index ))
+         return FALSE;
+      eat_opt_white( &ctx->cur );
+      if (*ctx->cur == '+' || *ctx->cur == '-') {
+         boolean negate;
+
+         negate = *ctx->cur == '-';
+         ctx->cur++;
+         eat_opt_white( &ctx->cur );
+         if (!parse_uint( &ctx->cur, &uindex )) {
+            report_error( ctx, "Expected literal unsigned integer" );
+            return FALSE;
+         }
+         if (negate)
+            *index = -(int) uindex;
+         else
+            *index = (int) uindex;
+      }
+      else {
+         *index = 0;
+      }
+   }
+   else {
+      if (!parse_uint( &ctx->cur, &uindex )) {
+         report_error( ctx, "Expected literal unsigned integer" );
+         return FALSE;
+      }
+      *index = (int) uindex;
+      *ind_file = TGSI_FILE_NULL;
+      *ind_index = 0;
+   }
+   eat_opt_white( &ctx->cur );
+   if (*ctx->cur != ']') {
+      report_error( ctx, "Expected `]'" );
+      return FALSE;
+   }
+   ctx->cur++;
+   return TRUE;
+}
+
+/* Parse register declaration.
+ *    <register_dcl> ::= <register_file_bracket_index> `]' |
+ *                       <register_file_bracket_index> `..' <index> `]'
+ */
+static boolean
+parse_register_dcl(
+   struct translate_ctx *ctx,
+   uint *file,
+   int *first,
+   int *last )
+{
+   if (!parse_register_file_bracket_index( ctx, file, first ))
+      return FALSE;
+   eat_opt_white( &ctx->cur );
+   if (ctx->cur[0] == '.' && ctx->cur[1] == '.') {
+      uint uindex;
+
+      ctx->cur += 2;
+      eat_opt_white( &ctx->cur );
+      if (!parse_uint( &ctx->cur, &uindex )) {
+         report_error( ctx, "Expected literal integer" );
+         return FALSE;
+      }
+      *last = (int) uindex;
+      eat_opt_white( &ctx->cur );
+   }
+   else {
+      *last = *first;
+   }
+   if (*ctx->cur != ']') {
+      report_error( ctx, "Expected `]' or `..'" );
+      return FALSE;
+   }
+   ctx->cur++;
+   return TRUE;
+}
+
+static const char *modulate_names[TGSI_MODULATE_COUNT] =
+{
+   "_1X",
+   "_2X",
+   "_4X",
+   "_8X",
+   "_D2",
+   "_D4",
+   "_D8"
+};
+
+static boolean
+parse_dst_operand(
+   struct translate_ctx *ctx,
+   struct tgsi_full_dst_register *dst )
+{
+   uint file;
+   int index;
+   uint writemask;
+   const char *cur;
+
+   if (!parse_register_dst( ctx, &file, &index ))
+      return FALSE;
+
+   cur = ctx->cur;
+   eat_opt_white( &cur );
+   if (*cur == '_') {
+      uint i;
+
+      for (i = 0; i < TGSI_MODULATE_COUNT; i++) {
+         if (str_match_no_case( &cur, modulate_names[i] )) {
+            if (!is_digit_alpha_underscore( cur )) {
+               dst->DstRegisterExtModulate.Modulate = i;
+               ctx->cur = cur;
+               break;
+            }
+         }
+      }
+   }
+
+   if (!parse_opt_writemask( ctx, &writemask ))
+      return FALSE;
+
+   dst->DstRegister.File = file;
+   dst->DstRegister.Index = index;
+   dst->DstRegister.WriteMask = writemask;
+   return TRUE;
+}
+
+static boolean
+parse_optional_swizzle(
+   struct translate_ctx *ctx,
+   uint swizzle[4],
+   boolean *parsed_swizzle,
+   boolean *parsed_extswizzle )
+{
+   const char *cur = ctx->cur;
+
+   *parsed_swizzle = FALSE;
+   *parsed_extswizzle = FALSE;
+
+   eat_opt_white( &cur );
+   if (*cur == '.') {
+      uint i;
+
+      cur++;
+      eat_opt_white( &cur );
+      for (i = 0; i < 4; i++) {
+         if (uprcase( *cur ) == 'X')
+            swizzle[i] = TGSI_SWIZZLE_X;
+         else if (uprcase( *cur ) == 'Y')
+            swizzle[i] = TGSI_SWIZZLE_Y;
+         else if (uprcase( *cur ) == 'Z')
+            swizzle[i] = TGSI_SWIZZLE_Z;
+         else if (uprcase( *cur ) == 'W')
+            swizzle[i] = TGSI_SWIZZLE_W;
+         else {
+            if (*cur == '0')
+               swizzle[i] = TGSI_EXTSWIZZLE_ZERO;
+            else if (*cur == '1')
+               swizzle[i] = TGSI_EXTSWIZZLE_ONE;
+            else {
+               report_error( ctx, "Expected register swizzle component `x', `y', `z', `w', `0' or `1'" );
+               return FALSE;
+            }
+            *parsed_extswizzle = TRUE;
+         }
+         cur++;
+      }
+      *parsed_swizzle = TRUE;
+      ctx->cur = cur;
+   }
+   return TRUE;
+}
+
+static boolean
+parse_src_operand(
+   struct translate_ctx *ctx,
+   struct tgsi_full_src_register *src )
+{
+   const char *cur;
+   float value;
+   uint file;
+   int index;
+   uint ind_file;
+   int ind_index;
+   uint swizzle[4];
+   boolean parsed_swizzle;
+   boolean parsed_extswizzle;
+
+   if (*ctx->cur == '-') {
+      cur = ctx->cur;
+      cur++;
+      eat_opt_white( &cur );
+      if (*cur == '(') {
+         cur++;
+         src->SrcRegisterExtMod.Negate = 1;
+         eat_opt_white( &cur );
+         ctx->cur = cur;
+      }
+   }
+
+   if (*ctx->cur == '|') {
+      ctx->cur++;
+      eat_opt_white( &ctx->cur );
+      src->SrcRegisterExtMod.Absolute = 1;
+   }
+
+   if (*ctx->cur == '-') {
+      ctx->cur++;
+      eat_opt_white( &ctx->cur );
+      src->SrcRegister.Negate = 1;
+   }
+
+   cur = ctx->cur;
+   if (parse_float( &cur, &value )) {
+      if (value == 2.0f) {
+         eat_opt_white( &cur );
+         if (*cur != '*') {
+            report_error( ctx, "Expected `*'" );
+            return FALSE;
+         }
+         cur++;
+         if (*cur != '(') {
+            report_error( ctx, "Expected `('" );
+            return FALSE;
+         }
+         cur++;
+         src->SrcRegisterExtMod.Scale2X = 1;
+         eat_opt_white( &cur );
+         ctx->cur = cur;
+      }
+   }
+
+   if (*ctx->cur == '(') {
+      ctx->cur++;
+      eat_opt_white( &ctx->cur );
+      src->SrcRegisterExtMod.Bias = 1;
+   }
+
+   cur = ctx->cur;
+   if (parse_float( &cur, &value )) {
+      if (value == 1.0f) {
+         eat_opt_white( &cur );
+         if (*cur != '-') {
+            report_error( ctx, "Expected `-'" );
+            return FALSE;
+         }
+         cur++;
+         if (*cur != '(') {
+            report_error( ctx, "Expected `('" );
+            return FALSE;
+         }
+         cur++;
+         src->SrcRegisterExtMod.Complement = 1;
+         eat_opt_white( &cur );
+         ctx->cur = cur;
+      }
+   }
+
+   if (!parse_register_src( ctx, &file, &index, &ind_file, &ind_index ))
+      return FALSE;
+   src->SrcRegister.File = file;
+   src->SrcRegister.Index = index;
+   if (ind_file != TGSI_FILE_NULL) {
+      src->SrcRegister.Indirect = 1;
+      src->SrcRegisterInd.File = ind_file;
+      src->SrcRegisterInd.Index = ind_index;
+   }
+
+   /* Parse optional swizzle.
+    */
+   if (parse_optional_swizzle( ctx, swizzle, &parsed_swizzle, &parsed_extswizzle )) {
+      if (parsed_extswizzle) {
+         assert( parsed_swizzle );
+
+         src->SrcRegisterExtSwz.ExtSwizzleX = swizzle[0];
+         src->SrcRegisterExtSwz.ExtSwizzleY = swizzle[1];
+         src->SrcRegisterExtSwz.ExtSwizzleZ = swizzle[2];
+         src->SrcRegisterExtSwz.ExtSwizzleW = swizzle[3];
+      }
+      else if (parsed_swizzle) {
+         src->SrcRegister.SwizzleX = swizzle[0];
+         src->SrcRegister.SwizzleY = swizzle[1];
+         src->SrcRegister.SwizzleZ = swizzle[2];
+         src->SrcRegister.SwizzleW = swizzle[3];
+      }
+   }
+
+   if (src->SrcRegisterExtMod.Complement) {
+      eat_opt_white( &ctx->cur );
+      if (*ctx->cur != ')') {
+         report_error( ctx, "Expected `)'" );
+         return FALSE;
+      }
+      ctx->cur++;
+   }
+
+   if (src->SrcRegisterExtMod.Bias) {
+      eat_opt_white( &ctx->cur );
+      if (*ctx->cur != ')') {
+         report_error( ctx, "Expected `)'" );
+         return FALSE;
+      }
+      ctx->cur++;
+      eat_opt_white( &ctx->cur );
+      if (*ctx->cur != '-') {
+         report_error( ctx, "Expected `-'" );
+         return FALSE;
+      }
+      ctx->cur++;
+      eat_opt_white( &ctx->cur );
+      if (!parse_float( &ctx->cur, &value )) {
+         report_error( ctx, "Expected literal floating point" );
+         return FALSE;
+      }
+      if (value != 0.5f) {
+         report_error( ctx, "Expected 0.5" );
+         return FALSE;
+      }
+   }
+
+   if (src->SrcRegisterExtMod.Scale2X) {
+      eat_opt_white( &ctx->cur );
+      if (*ctx->cur != ')') {
+         report_error( ctx, "Expected `)'" );
+         return FALSE;
+      }
+      ctx->cur++;
+   }
+
+   if (src->SrcRegisterExtMod.Absolute) {
+      eat_opt_white( &ctx->cur );
+      if (*ctx->cur != '|') {
+         report_error( ctx, "Expected `|'" );
+         return FALSE;
+      }
+      ctx->cur++;
+   }
+
+   if (src->SrcRegisterExtMod.Negate) {
+      eat_opt_white( &ctx->cur );
+      if (*ctx->cur != ')') {
+         report_error( ctx, "Expected `)'" );
+         return FALSE;
+      }
+      ctx->cur++;
+   }
+
+   return TRUE;
+}
+
+static const char *texture_names[TGSI_TEXTURE_COUNT] =
+{
+   "UNKNOWN",
+   "1D",
+   "2D",
+   "3D",
+   "CUBE",
+   "RECT",
+   "SHADOW1D",
+   "SHADOW2D",
+   "SHADOWRECT"
+};
+
+static boolean
+parse_instruction(
+   struct translate_ctx *ctx,
+   boolean has_label )
+{
+   uint i;
+   uint saturate = TGSI_SAT_NONE;
+   const struct tgsi_opcode_info *info;
+   struct tgsi_full_instruction inst;
+   uint advance;
+
+   /* Parse instruction name.
+    */
+   eat_opt_white( &ctx->cur );
+   for (i = 0; i < TGSI_OPCODE_LAST; i++) {
+      const char *cur = ctx->cur;
+
+      info = tgsi_get_opcode_info( i );
+      if (str_match_no_case( &cur, info->mnemonic )) {
+         if (str_match_no_case( &cur, "_SATNV" ))
+            saturate = TGSI_SAT_MINUS_PLUS_ONE;
+         else if (str_match_no_case( &cur, "_SAT" ))
+            saturate = TGSI_SAT_ZERO_ONE;
+
+         if (info->num_dst + info->num_src + info->is_tex == 0) {
+            if (!is_digit_alpha_underscore( cur )) {
+               ctx->cur = cur;
+               break;
+            }
+         }
+         else if (*cur == '\0' || eat_white( &cur )) {
+            ctx->cur = cur;
+            break;
+         }
+      }
+   }
+   if (i == TGSI_OPCODE_LAST) {
+      if (has_label)
+         report_error( ctx, "Unknown opcode" );
+      else
+         report_error( ctx, "Expected `DCL', `IMM' or a label" );
+      return FALSE;
+   }
+
+   inst = tgsi_default_full_instruction();
+   inst.Instruction.Opcode = i;
+   inst.Instruction.Saturate = saturate;
+   inst.Instruction.NumDstRegs = info->num_dst;
+   inst.Instruction.NumSrcRegs = info->num_src;
+
+   /* Parse instruction operands.
+    */
+   for (i = 0; i < info->num_dst + info->num_src + info->is_tex; i++) {
+      if (i > 0) {
+         eat_opt_white( &ctx->cur );
+         if (*ctx->cur != ',') {
+            report_error( ctx, "Expected `,'" );
+            return FALSE;
+         }
+         ctx->cur++;
+         eat_opt_white( &ctx->cur );
+      }
+
+      if (i < info->num_dst) {
+         if (!parse_dst_operand( ctx, &inst.FullDstRegisters[i] ))
+            return FALSE;
+      }
+      else if (i < info->num_dst + info->num_src) {
+         if (!parse_src_operand( ctx, &inst.FullSrcRegisters[i - info->num_dst] ))
+            return FALSE;
+      }
+      else {
+         uint j;
+
+         for (j = 0; j < TGSI_TEXTURE_COUNT; j++) {
+            if (str_match_no_case( &ctx->cur, texture_names[j] )) {
+               if (!is_digit_alpha_underscore( ctx->cur )) {
+                  inst.InstructionExtTexture.Texture = j;
+                  break;
+               }
+            }
+         }
+         if (j == TGSI_TEXTURE_COUNT) {
+            report_error( ctx, "Expected texture target" );
+            return FALSE;
+         }
+      }
+   }
+
+   if (info->is_branch) {
+      uint target;
+
+      eat_opt_white( &ctx->cur );
+      if (*ctx->cur != ':') {
+         report_error( ctx, "Expected `:'" );
+         return FALSE;
+      }
+      ctx->cur++;
+      eat_opt_white( &ctx->cur );
+      if (!parse_uint( &ctx->cur, &target )) {
+         report_error( ctx, "Expected a label" );
+         return FALSE;
+      }
+      inst.InstructionExtLabel.Label = target;
+   }
+
+   advance = tgsi_build_full_instruction(
+      &inst,
+      ctx->tokens_cur,
+      ctx->header,
+      (uint) (ctx->tokens_end - ctx->tokens_cur) );
+   if (advance == 0)
+      return FALSE;
+   ctx->tokens_cur += advance;
+
+   return TRUE;
+}
+
+static const char *semantic_names[TGSI_SEMANTIC_COUNT] =
+{
+   "POSITION",
+   "COLOR",
+   "BCOLOR",
+   "FOG",
+   "PSIZE",
+   "GENERIC",
+   "NORMAL"
+};
+
+static const char *interpolate_names[TGSI_INTERPOLATE_COUNT] =
+{
+   "CONSTANT",
+   "LINEAR",
+   "PERSPECTIVE"
+};
+
+static boolean parse_declaration( struct translate_ctx *ctx )
+{
+   struct tgsi_full_declaration decl;
+   uint file;
+   int first;
+   int last;
+   uint writemask;
+   const char *cur;
+   uint advance;
+
+   if (!eat_white( &ctx->cur )) {
+      report_error( ctx, "Syntax error" );
+      return FALSE;
+   }
+   if (!parse_register_dcl( ctx, &file, &first, &last ))
+      return FALSE;
+   if (!parse_opt_writemask( ctx, &writemask ))
+      return FALSE;
+
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = file;
+   decl.Declaration.UsageMask = writemask;
+   decl.DeclarationRange.First = first;
+   decl.DeclarationRange.Last = last;
+
+   cur = ctx->cur;
+   eat_opt_white( &cur );
+   if (*cur == ',') {
+      uint i;
+
+      cur++;
+      eat_opt_white( &cur );
+      for (i = 0; i < TGSI_SEMANTIC_COUNT; i++) {
+         if (str_match_no_case( &cur, semantic_names[i] )) {
+            const char *cur2 = cur;
+            uint index;
+
+            if (is_digit_alpha_underscore( cur ))
+               continue;
+            eat_opt_white( &cur2 );
+            if (*cur2 == '[') {
+               cur2++;
+               eat_opt_white( &cur2 );
+               if (!parse_uint( &cur2, &index )) {
+                  report_error( ctx, "Expected literal integer" );
+                  return FALSE;
+               }
+               eat_opt_white( &cur2 );
+               if (*cur2 != ']') {
+                  report_error( ctx, "Expected `]'" );
+                  return FALSE;
+               }
+               cur2++;
+
+               decl.Semantic.SemanticIndex = index;
+
+               cur = cur2;
+            }
+
+            decl.Declaration.Semantic = 1;
+            decl.Semantic.SemanticName = i;
+
+            ctx->cur = cur;
+            break;
+         }
+      }
+   }
+
+   cur = ctx->cur;
+   eat_opt_white( &cur );
+   if (*cur == ',') {
+      uint i;
+
+      cur++;
+      eat_opt_white( &cur );
+      for (i = 0; i < TGSI_INTERPOLATE_COUNT; i++) {
+         if (str_match_no_case( &cur, interpolate_names[i] )) {
+            if (is_digit_alpha_underscore( cur ))
+               continue;
+            decl.Declaration.Interpolate = i;
+
+            ctx->cur = cur;
+            break;
+         }
+      }
+      if (i == TGSI_INTERPOLATE_COUNT) {
+         report_error( ctx, "Expected semantic or interpolate attribute" );
+         return FALSE;
+      }
+   }
+
+   advance = tgsi_build_full_declaration(
+      &decl,
+      ctx->tokens_cur,
+      ctx->header,
+      (uint) (ctx->tokens_end - ctx->tokens_cur) );
+   if (advance == 0)
+      return FALSE;
+   ctx->tokens_cur += advance;
+
+   return TRUE;
+}
+
+static boolean parse_immediate( struct translate_ctx *ctx )
+{
+   struct tgsi_full_immediate imm;
+   uint i;
+   float values[4];
+   uint advance;
+
+   if (!eat_white( &ctx->cur )) {
+      report_error( ctx, "Syntax error" );
+      return FALSE;
+   }
+   if (!str_match_no_case( &ctx->cur, "FLT32" ) || is_digit_alpha_underscore( ctx->cur )) {
+      report_error( ctx, "Expected `FLT32'" );
+      return FALSE;
+   }
+   eat_opt_white( &ctx->cur );
+   if (*ctx->cur != '{') {
+      report_error( ctx, "Expected `{'" );
+      return FALSE;
+   }
+   ctx->cur++;
+   for (i = 0; i < 4; i++) {
+      eat_opt_white( &ctx->cur );
+      if (i > 0) {
+         if (*ctx->cur != ',') {
+            report_error( ctx, "Expected `,'" );
+            return FALSE;
+         }
+         ctx->cur++;
+         eat_opt_white( &ctx->cur );
+      }
+      if (!parse_float( &ctx->cur, &values[i] )) {
+         report_error( ctx, "Expected literal floating point" );
+         return FALSE;
+      }
+   }
+   eat_opt_white( &ctx->cur );
+   if (*ctx->cur != '}') {
+      report_error( ctx, "Expected `}'" );
+      return FALSE;
+   }
+   ctx->cur++;
+
+   imm = tgsi_default_full_immediate();
+   imm.Immediate.NrTokens += 4;
+   imm.Immediate.DataType = TGSI_IMM_FLOAT32;
+   imm.u.Pointer = values;
+
+   advance = tgsi_build_full_immediate(
+      &imm,
+      ctx->tokens_cur,
+      ctx->header,
+      (uint) (ctx->tokens_end - ctx->tokens_cur) );
+   if (advance == 0)
+      return FALSE;
+   ctx->tokens_cur += advance;
+
+   return TRUE;
+}
+
+static boolean translate( struct translate_ctx *ctx )
+{
+   eat_opt_white( &ctx->cur );
+   if (!parse_header( ctx ))
+      return FALSE;
+
+   while (*ctx->cur != '\0') {
+      uint label_val = 0;
+
+      if (!eat_white( &ctx->cur )) {
+         report_error( ctx, "Syntax error" );
+         return FALSE;
+      }
+
+      if (*ctx->cur == '\0')
+         break;
+
+      if (parse_label( ctx, &label_val )) {
+         if (!parse_instruction( ctx, TRUE ))
+            return FALSE;
+      }
+      else if (str_match_no_case( &ctx->cur, "DCL" )) {
+         if (!parse_declaration( ctx ))
+            return FALSE;
+      }
+      else if (str_match_no_case( &ctx->cur, "IMM" )) {
+         if (!parse_immediate( ctx ))
+            return FALSE;
+      }
+      else if (!parse_instruction( ctx, FALSE )) {
+         return FALSE;
+      }
+   }
+
+   return TRUE;
+}
+
+boolean
+tgsi_text_translate(
+   const char *text,
+   struct tgsi_token *tokens,
+   uint num_tokens )
+{
+   struct translate_ctx ctx;
+
+   ctx.text = text;
+   ctx.cur = text;
+   ctx.tokens = tokens;
+   ctx.tokens_cur = tokens;
+   ctx.tokens_end = tokens + num_tokens;
+
+   if (!translate( &ctx ))
+      return FALSE;
+
+   return tgsi_sanity_check( tokens );
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.h b/src/gallium/auxiliary/tgsi/tgsi_text.h
new file mode 100644
index 0000000000..8eeeeef140
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.h
@@ -0,0 +1,47 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_TEXT_H
+#define TGSI_TEXT_H
+
+#include "pipe/p_shader_tokens.h"
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+boolean
+tgsi_text_translate(
+   const char *text,
+   struct tgsi_token *tokens,
+   uint num_tokens );
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_TEXT_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.c b/src/gallium/auxiliary/tgsi/tgsi_transform.c
new file mode 100644
index 0000000000..ea87da31e5
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_transform.c
@@ -0,0 +1,200 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * TGSI program transformation utility.
+ *
+ * Authors:  Brian Paul
+ */
+
+#include "pipe/p_debug.h"
+
+#include "tgsi_transform.h"
+
+
+
+static void
+emit_instruction(struct tgsi_transform_context *ctx,
+                 const struct tgsi_full_instruction *inst)
+{
+   uint ti = ctx->ti;
+
+   ti += tgsi_build_full_instruction(inst,
+                                     ctx->tokens_out + ti,
+                                     ctx->header,
+                                     ctx->max_tokens_out - ti);
+   ctx->ti = ti;
+}
+
+
+static void
+emit_declaration(struct tgsi_transform_context *ctx,
+                 const struct tgsi_full_declaration *decl)
+{
+   uint ti = ctx->ti;
+
+   ti += tgsi_build_full_declaration(decl,
+                                     ctx->tokens_out + ti,
+                                     ctx->header,
+                                     ctx->max_tokens_out - ti);
+   ctx->ti = ti;
+}
+
+
+static void
+emit_immediate(struct tgsi_transform_context *ctx,
+               const struct tgsi_full_immediate *imm)
+{
+   uint ti = ctx->ti;
+
+   ti += tgsi_build_full_immediate(imm,
+                                   ctx->tokens_out + ti,
+                                   ctx->header,
+                                   ctx->max_tokens_out - ti);
+   ctx->ti = ti;
+}
+
+
+
+/**
+ * Apply user-defined transformations to the input shader to produce
+ * the output shader.
+ * For example, a register search-and-replace operation could be applied
+ * by defining a transform_instruction() callback that examined and changed
+ * the instruction src/dest regs.
+ *
+ * \return number of tokens emitted
+ */
+int
+tgsi_transform_shader(const struct tgsi_token *tokens_in,
+                      struct tgsi_token *tokens_out,
+                      uint max_tokens_out,
+                      struct tgsi_transform_context *ctx)
+{
+   uint procType;
+
+   /* input shader */
+   struct tgsi_parse_context parse;
+
+   /* output shader */
+   struct tgsi_processor *processor;
+
+
+   /**
+    ** callback context init
+    **/
+   ctx->emit_instruction = emit_instruction;
+   ctx->emit_declaration = emit_declaration;
+   ctx->emit_immediate = emit_immediate;
+   ctx->tokens_out = tokens_out;
+   ctx->max_tokens_out = max_tokens_out;
+
+
+   /**
+    ** Setup to begin parsing input shader
+    **/
+   if (tgsi_parse_init( &parse, tokens_in ) != TGSI_PARSE_OK) {
+      debug_printf("tgsi_parse_init() failed in tgsi_transform_shader()!\n");
+      return -1;
+   }
+   procType = parse.FullHeader.Processor.Processor;
+   assert(procType == TGSI_PROCESSOR_FRAGMENT ||
+          procType == TGSI_PROCESSOR_VERTEX ||
+          procType == TGSI_PROCESSOR_GEOMETRY);
+
+
+   /**
+    **  Setup output shader
+    **/
+   *(struct tgsi_version *) &tokens_out[0] = tgsi_build_version();
+
+   ctx->header = (struct tgsi_header *) (tokens_out + 1);
+   *ctx->header = tgsi_build_header();
+
+   processor = (struct tgsi_processor *) (tokens_out + 2);
+   *processor = tgsi_build_processor( procType, ctx->header );
+
+   ctx->ti = 3;
+
+
+   /**
+    ** Loop over incoming program tokens/instructions
+    */
+   while( !tgsi_parse_end_of_tokens( &parse ) ) {
+
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         {
+            struct tgsi_full_instruction *fullinst
+               = &parse.FullToken.FullInstruction;
+
+            if (ctx->transform_instruction)
+               ctx->transform_instruction(ctx, fullinst);
+            else
+               ctx->emit_instruction(ctx, fullinst);
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         {
+            struct tgsi_full_declaration *fulldecl
+               = &parse.FullToken.FullDeclaration;
+
+            if (ctx->transform_declaration)
+               ctx->transform_declaration(ctx, fulldecl);
+            else
+               ctx->emit_declaration(ctx, fulldecl);
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         {
+            struct tgsi_full_immediate *fullimm
+               = &parse.FullToken.FullImmediate;
+
+            if (ctx->transform_immediate)
+               ctx->transform_immediate(ctx, fullimm);
+            else
+               ctx->emit_immediate(ctx, fullimm);
+         }
+         break;
+
+      default:
+         assert( 0 );
+      }
+   }
+
+   if (ctx->epilog) {
+      ctx->epilog(ctx);
+   }
+
+   tgsi_parse_free (&parse);
+
+   return ctx->ti;
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.h b/src/gallium/auxiliary/tgsi/tgsi_transform.h
new file mode 100644
index 0000000000..a121adbaef
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_transform.h
@@ -0,0 +1,92 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_TRANSFORM_H
+#define TGSI_TRANSFORM_H
+
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_build.h"
+
+
+
+/**
+ * Subclass this to add caller-specific data
+ */
+struct tgsi_transform_context
+{
+/**** PUBLIC ***/
+
+   /**
+    * User-defined callbacks invoked per instruction.
+    */
+   void (*transform_instruction)(struct tgsi_transform_context *ctx,
+                                 struct tgsi_full_instruction *inst);
+
+   void (*transform_declaration)(struct tgsi_transform_context *ctx,
+                                 struct tgsi_full_declaration *decl);
+
+   void (*transform_immediate)(struct tgsi_transform_context *ctx,
+                               struct tgsi_full_immediate *imm);
+
+   /**
+    * Called at end of input program to allow caller to append extra
+    * instructions.  Return number of tokens emitted.
+    */
+   void (*epilog)(struct tgsi_transform_context *ctx);
+
+
+/*** PRIVATE ***/
+
+   /**
+    * These are setup by tgsi_transform_shader() and cannot be overridden.
+    * Meant to be called from in the above user callback functions.
+    */
+   void (*emit_instruction)(struct tgsi_transform_context *ctx,
+                            const struct tgsi_full_instruction *inst);
+   void (*emit_declaration)(struct tgsi_transform_context *ctx,
+                            const struct tgsi_full_declaration *decl);
+   void (*emit_immediate)(struct tgsi_transform_context *ctx,
+                          const struct tgsi_full_immediate *imm);
+
+   struct tgsi_header *header;
+   uint max_tokens_out;
+   struct tgsi_token *tokens_out;
+   uint ti;
+};
+
+
+
+extern int
+tgsi_transform_shader(const struct tgsi_token *tokens_in,
+                      struct tgsi_token *tokens_out,
+                      uint max_tokens_out,
+                      struct tgsi_transform_context *ctx);
+
+
+#endif /* TGSI_TRANSFORM_H */
diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.c b/src/gallium/auxiliary/tgsi/tgsi_util.c
new file mode 100644
index 0000000000..50101a9bb0
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_util.c
@@ -0,0 +1,299 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_debug.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi_parse.h"
+#include "tgsi_build.h"
+#include "tgsi_util.h"
+
+union pointer_hack
+{
+   void *pointer;
+   uint64_t uint64;
+};
+
+void *
+tgsi_align_128bit(
+   void *unaligned )
+{
+   union pointer_hack ph;
+
+   ph.uint64 = 0;
+   ph.pointer = unaligned;
+   ph.uint64 = (ph.uint64 + 15) & ~15;
+   return ph.pointer;
+}
+
+unsigned
+tgsi_util_get_src_register_swizzle(
+   const struct tgsi_src_register *reg,
+   unsigned component )
+{
+   switch( component ) {
+   case 0:
+      return reg->SwizzleX;
+   case 1:
+      return reg->SwizzleY;
+   case 2:
+      return reg->SwizzleZ;
+   case 3:
+      return reg->SwizzleW;
+   default:
+      assert( 0 );
+   }
+   return 0;
+}
+
+unsigned
+tgsi_util_get_src_register_extswizzle(
+   const struct tgsi_src_register_ext_swz *reg,
+   unsigned component )
+{
+   switch( component ) {
+   case 0:
+      return reg->ExtSwizzleX;
+   case 1:
+      return reg->ExtSwizzleY;
+   case 2:
+      return reg->ExtSwizzleZ;
+   case 3:
+      return reg->ExtSwizzleW;
+   default:
+      assert( 0 );
+   }
+   return 0;
+}
+
+unsigned
+tgsi_util_get_full_src_register_extswizzle(
+   const struct tgsi_full_src_register  *reg,
+   unsigned component )
+{
+   unsigned swizzle;
+
+   /*
+    * First, calculate  the   extended swizzle for a given channel. This will give
+    * us either a channel index into the simple swizzle or  a constant 1 or   0.
+    */
+   swizzle = tgsi_util_get_src_register_extswizzle(
+      &reg->SrcRegisterExtSwz,
+      component );
+
+   assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X);
+   assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y);
+   assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z);
+   assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W);
+   assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W);
+   assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W);
+
+   /*
+    * Second, calculate the simple  swizzle  for   the   unswizzled channel index.
+    * Leave the constants intact, they are   not   affected by the   simple swizzle.
+    */
+   if( swizzle <= TGSI_SWIZZLE_W ) {
+      swizzle = tgsi_util_get_src_register_swizzle(
+         &reg->SrcRegister,
+         swizzle );
+   }
+
+   return swizzle;
+}
+
+void
+tgsi_util_set_src_register_swizzle(
+   struct tgsi_src_register *reg,
+   unsigned swizzle,
+   unsigned component )
+{
+   switch( component ) {
+   case 0:
+      reg->SwizzleX = swizzle;
+      break;
+   case 1:
+      reg->SwizzleY = swizzle;
+      break;
+   case 2:
+      reg->SwizzleZ = swizzle;
+      break;
+   case 3:
+      reg->SwizzleW = swizzle;
+      break;
+   default:
+      assert( 0 );
+   }
+}
+
+void
+tgsi_util_set_src_register_extswizzle(
+   struct tgsi_src_register_ext_swz *reg,
+   unsigned swizzle,
+   unsigned component )
+{
+   switch( component ) {
+   case 0:
+      reg->ExtSwizzleX = swizzle;
+      break;
+   case 1:
+      reg->ExtSwizzleY = swizzle;
+      break;
+   case 2:
+      reg->ExtSwizzleZ = swizzle;
+      break;
+   case 3:
+      reg->ExtSwizzleW = swizzle;
+      break;
+   default:
+      assert( 0 );
+   }
+}
+
+unsigned
+tgsi_util_get_src_register_extnegate(
+   const  struct tgsi_src_register_ext_swz *reg,
+   unsigned component )
+{
+   switch( component ) {
+   case 0:
+      return reg->NegateX;
+   case 1:
+      return reg->NegateY;
+   case 2:
+      return reg->NegateZ;
+   case 3:
+      return reg->NegateW;
+   default:
+      assert( 0 );
+   }
+   return 0;
+}
+
+void
+tgsi_util_set_src_register_extnegate(
+   struct tgsi_src_register_ext_swz *reg,
+   unsigned negate,
+   unsigned component )
+{
+   switch( component ) {
+   case 0:
+      reg->NegateX = negate;
+      break;
+   case 1:
+      reg->NegateY = negate;
+      break;
+   case 2:
+      reg->NegateZ = negate;
+      break;
+   case 3:
+      reg->NegateW = negate;
+      break;
+   default:
+      assert( 0 );
+   }
+}
+
+unsigned
+tgsi_util_get_full_src_register_sign_mode(
+   const struct  tgsi_full_src_register *reg,
+   unsigned component )
+{
+   unsigned sign_mode;
+
+   if( reg->SrcRegisterExtMod.Absolute ) {
+      /* Consider only the post-abs negation. */
+
+      if( reg->SrcRegisterExtMod.Negate ) {
+         sign_mode = TGSI_UTIL_SIGN_SET;
+      }
+      else {
+         sign_mode = TGSI_UTIL_SIGN_CLEAR;
+      }
+   }
+   else {
+      /* Accumulate the three negations. */
+
+      unsigned negate;
+
+      negate = reg->SrcRegister.Negate;
+      if( tgsi_util_get_src_register_extnegate( &reg->SrcRegisterExtSwz, component ) ) {
+         negate = !negate;
+      }
+      if( reg->SrcRegisterExtMod.Negate ) {
+         negate = !negate;
+      }
+
+      if( negate ) {
+         sign_mode = TGSI_UTIL_SIGN_TOGGLE;
+      }
+      else {
+         sign_mode = TGSI_UTIL_SIGN_KEEP;
+      }
+   }
+
+   return sign_mode;
+}
+
+void
+tgsi_util_set_full_src_register_sign_mode(
+   struct tgsi_full_src_register *reg,
+   unsigned sign_mode )
+{
+   reg->SrcRegisterExtSwz.NegateX = 0;
+   reg->SrcRegisterExtSwz.NegateY = 0;
+   reg->SrcRegisterExtSwz.NegateZ = 0;
+   reg->SrcRegisterExtSwz.NegateW = 0;
+
+   switch (sign_mode)
+   {
+   case TGSI_UTIL_SIGN_CLEAR:
+      reg->SrcRegister.Negate = 0;
+      reg->SrcRegisterExtMod.Absolute = 1;
+      reg->SrcRegisterExtMod.Negate = 0;
+      break;
+
+   case TGSI_UTIL_SIGN_SET:
+      reg->SrcRegister.Negate = 0;
+      reg->SrcRegisterExtMod.Absolute = 1;
+      reg->SrcRegisterExtMod.Negate = 1;
+      break;
+
+   case TGSI_UTIL_SIGN_TOGGLE:
+      reg->SrcRegister.Negate = 1;
+      reg->SrcRegisterExtMod.Absolute = 0;
+      reg->SrcRegisterExtMod.Negate = 0;
+      break;
+
+   case TGSI_UTIL_SIGN_KEEP:
+      reg->SrcRegister.Negate = 0;
+      reg->SrcRegisterExtMod.Absolute = 0;
+      reg->SrcRegisterExtMod.Negate = 0;
+      break;
+
+   default:
+      assert( 0 );
+   }
+}
diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.h b/src/gallium/auxiliary/tgsi/tgsi_util.h
new file mode 100644
index 0000000000..7877f34558
--- /dev/null
+++ b/src/gallium/auxiliary/tgsi/tgsi_util.h
@@ -0,0 +1,96 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_UTIL_H
+#define TGSI_UTIL_H
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+void *
+tgsi_align_128bit(
+   void *unaligned );
+
+unsigned
+tgsi_util_get_src_register_swizzle(
+   const struct tgsi_src_register *reg,
+   unsigned component );
+
+unsigned
+tgsi_util_get_src_register_extswizzle(
+   const struct tgsi_src_register_ext_swz *reg,
+   unsigned component);
+
+unsigned
+tgsi_util_get_full_src_register_extswizzle(
+   const struct tgsi_full_src_register *reg,
+   unsigned component );
+
+void
+tgsi_util_set_src_register_swizzle(
+   struct tgsi_src_register *reg,
+   unsigned swizzle,
+   unsigned component );
+
+void
+tgsi_util_set_src_register_extswizzle(
+   struct tgsi_src_register_ext_swz *reg,
+   unsigned swizzle,
+   unsigned component );
+
+unsigned
+tgsi_util_get_src_register_extnegate(
+   const struct tgsi_src_register_ext_swz *reg,
+   unsigned component );
+
+void
+tgsi_util_set_src_register_extnegate(
+   struct tgsi_src_register_ext_swz *reg,
+   unsigned negate,
+   unsigned component );
+
+#define TGSI_UTIL_SIGN_CLEAR    0   /* Force positive */
+#define TGSI_UTIL_SIGN_SET      1   /* Force negative */
+#define TGSI_UTIL_SIGN_TOGGLE   2   /* Negate */
+#define TGSI_UTIL_SIGN_KEEP     3   /* No change */
+
+unsigned
+tgsi_util_get_full_src_register_sign_mode(
+   const struct tgsi_full_src_register *reg,
+   unsigned component );
+
+void
+tgsi_util_set_full_src_register_sign_mode(
+   struct tgsi_full_src_register *reg,
+   unsigned sign_mode );
+
+#if defined __cplusplus
+}
+#endif
+
+#endif /* TGSI_UTIL_H */
diff --git a/src/gallium/auxiliary/translate/Makefile b/src/gallium/auxiliary/translate/Makefile
new file mode 100644
index 0000000000..ad2a5b705e
--- /dev/null
+++ b/src/gallium/auxiliary/translate/Makefile
@@ -0,0 +1,15 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = translate
+
+C_SOURCES = \
+	translate_generic.c \
+	translate_sse.c \
+	translate.c \
+        translate_cache.c
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/translate/SConscript b/src/gallium/auxiliary/translate/SConscript
new file mode 100644
index 0000000000..9553a67537
--- /dev/null
+++ b/src/gallium/auxiliary/translate/SConscript
@@ -0,0 +1,12 @@
+Import('*')
+
+translate = env.ConvenienceLibrary(
+	target = 'translate',
+	source = [
+		'translate_generic.c',
+		'translate_sse.c',
+		'translate.c',
+		'translate_cache.c',
+	])
+
+auxiliaries.insert(0, translate)
diff --git a/src/gallium/auxiliary/translate/translate.c b/src/gallium/auxiliary/translate/translate.c
new file mode 100644
index 0000000000..7678903f75
--- /dev/null
+++ b/src/gallium/auxiliary/translate/translate.c
@@ -0,0 +1,48 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "pipe/p_config.h"
+#include "pipe/p_state.h"
+#include "translate.h"
+
+struct translate *translate_create( const struct translate_key *key )
+{
+   struct translate *translate = NULL;
+
+#if defined(PIPE_ARCH_X86)
+   translate = translate_sse2_create( key );
+   if (translate)
+      return translate;
+#endif
+
+   return translate_generic_create( key );
+}
diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h
new file mode 100644
index 0000000000..34526eb061
--- /dev/null
+++ b/src/gallium/auxiliary/translate/translate.h
@@ -0,0 +1,127 @@
+/*
+ * Copyright 2008 Tungsten Graphics, inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/**
+ * Vertex fetch/store/convert code.  This functionality is used in two places:
+ * 1. Vertex fetch/convert - to grab vertex data from incoming vertex
+ *    arrays and convert to format needed by vertex shaders.
+ * 2. Vertex store/emit - to convert simple float[][4] vertex attributes
+ *    (which is the organization used throughout the draw/prim pipeline) to
+ *    hardware-specific formats and emit into hardware vertex buffers.
+ *
+ *
+ * Authors:
+ *    Keith Whitwell <keithw@tungstengraphics.com>
+ */
+
+#ifndef _TRANSLATE_H
+#define _TRANSLATE_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+
+struct translate_element 
+{
+   enum pipe_format input_format;
+   enum pipe_format output_format;
+   unsigned input_buffer:8;
+   unsigned input_offset:24;
+   unsigned output_offset;
+};
+
+
+struct translate_key {
+   unsigned output_stride;
+   unsigned nr_elements;
+   struct translate_element element[PIPE_MAX_ATTRIBS + 1];
+};
+
+
+struct translate {
+   struct translate_key key;
+
+   void (*release)( struct translate * );
+
+   void (*set_buffer)( struct translate *,
+		       unsigned i,
+		       const void *ptr,
+		       unsigned stride );
+
+   void (PIPE_CDECL *run_elts)( struct translate *,
+                                const unsigned *elts,
+                                unsigned count,
+                                void *output_buffer);
+
+   void (PIPE_CDECL *run)( struct translate *,
+                           unsigned start,
+                           unsigned count,
+                           void *output_buffer);
+};
+
+
+
+#if 0
+struct translate_context *translate_context_create( void );
+void translate_context_destroy( struct translate_context * );
+
+struct translate *translate_lookup_or_create( struct translate_context *tctx,
+					      const struct translate_key *key );
+#endif
+
+
+struct translate *translate_create( const struct translate_key *key );
+
+static INLINE int translate_keysize( const struct translate_key *key )
+{
+   return 2 * sizeof(int) + key->nr_elements * sizeof(struct translate_element);
+}
+
+static INLINE int translate_key_compare( const struct translate_key *a,
+                                         const struct translate_key *b )
+{
+   int keysize = translate_keysize(a);
+   return memcmp(a, b, keysize);
+}
+
+
+static INLINE void translate_key_sanitize( struct translate_key *a )
+{
+   int keysize = translate_keysize(a);
+   char *ptr = (char *)a;
+   memset(ptr + keysize, 0, sizeof(*a) - keysize);
+}
+
+
+/*******************************************************************************
+ *  Private:
+ */
+struct translate *translate_sse2_create( const struct translate_key *key );
+
+struct translate *translate_generic_create( const struct translate_key *key );
+
+
+#endif
diff --git a/src/gallium/auxiliary/translate/translate_cache.c b/src/gallium/auxiliary/translate/translate_cache.c
new file mode 100644
index 0000000000..d8069a149c
--- /dev/null
+++ b/src/gallium/auxiliary/translate/translate_cache.c
@@ -0,0 +1,102 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_memory.h"
+#include "pipe/p_state.h"
+#include "translate.h"
+#include "translate_cache.h"
+
+#include "cso_cache/cso_cache.h"
+#include "cso_cache/cso_hash.h"
+
+struct translate_cache {
+   struct cso_hash *hash;
+};
+
+struct translate_cache * translate_cache_create( void )
+{
+   struct translate_cache *cache = MALLOC_STRUCT(translate_cache);
+   cache->hash = cso_hash_create();
+   return cache;
+}
+
+
+static INLINE void delete_translates(struct translate_cache *cache)
+{
+   struct cso_hash *hash = cache->hash;
+   struct cso_hash_iter iter = cso_hash_first_node(hash);
+   while (!cso_hash_iter_is_null(iter)) {
+      struct translate *state = (struct translate*)cso_hash_iter_data(iter);
+      iter = cso_hash_iter_next(iter);
+      if (state) {
+         state->release(state);
+      }
+   }
+}
+
+void translate_cache_destroy(struct translate_cache *cache)
+{
+   delete_translates(cache);
+   cso_hash_delete(cache->hash);
+   FREE(cache);
+}
+
+
+static INLINE unsigned translate_hash_key_size(struct translate_key *key)
+{
+   unsigned size = sizeof(struct translate_key) -
+                   sizeof(struct translate_element) * (PIPE_MAX_ATTRIBS - key->nr_elements);
+   return size;
+}
+
+static INLINE unsigned create_key(struct translate_key *key)
+{
+   unsigned hash_key;
+   unsigned size = translate_hash_key_size(key);
+   /*debug_printf("key size = %d, (els = %d)\n",
+     size, key->nr_elements);*/
+   hash_key = cso_construct_key(key, size);
+   return hash_key;
+}
+
+struct translate * translate_cache_find(struct translate_cache *cache,
+                                        struct translate_key *key)
+{
+   unsigned hash_key = create_key(key);
+   struct translate *translate = (struct translate*)
+      cso_hash_find_data_from_template(cache->hash,
+                                       hash_key,
+                                       key, sizeof(*key));
+
+   if (!translate) {
+      /* create/insert */
+      translate = translate_create(key);
+      cso_hash_insert(cache->hash, hash_key, translate);
+   }
+
+   return translate;
+}
diff --git a/src/gallium/auxiliary/translate/translate_cache.h b/src/gallium/auxiliary/translate/translate_cache.h
new file mode 100644
index 0000000000..7dba871e57
--- /dev/null
+++ b/src/gallium/auxiliary/translate/translate_cache.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright 2008 Tungsten Graphics, inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+#ifndef _TRANSLATE_CACHE_H
+#define _TRANSLATE_CACHE_H
+
+
+/*******************************************************************************
+ * Translate cache.
+ * Simply used to cache created translates. Avoids unecessary creation of
+ * translate's if one suitable for a given translate_key has already been
+ * created.
+ *
+ * Note: this functionality depends and requires the CSO module.
+ */
+struct translate_cache;
+
+struct translate_key;
+struct translate;
+
+struct translate_cache *translate_cache_create( void );
+void translate_cache_destroy(struct translate_cache *cache);
+
+/**
+ * Will try to find a translate structure matched by the given key.
+ * If such a structure doesn't exist in the cache the function
+ * will automatically create it, insert it in the cache and
+ * return the created version.
+ *
+ */
+struct translate *translate_cache_find(struct translate_cache *cache,
+                                       struct translate_key *key);
+
+#endif
diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c
new file mode 100644
index 0000000000..8d39b64c6c
--- /dev/null
+++ b/src/gallium/auxiliary/translate/translate_generic.c
@@ -0,0 +1,700 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "util/u_memory.h"
+#include "pipe/p_state.h"
+#include "translate.h"
+
+
+#define DRAW_DBG 0
+
+typedef void (*fetch_func)(const void *ptr, float *attrib);
+typedef void (*emit_func)(const float *attrib, void *ptr);
+
+
+
+struct translate_generic {
+   struct translate translate;
+
+   struct {
+      fetch_func fetch;
+      unsigned buffer;
+      unsigned input_offset;
+
+      emit_func emit;
+      unsigned output_offset;
+      
+      char *input_ptr;
+      unsigned input_stride;
+
+   } attrib[PIPE_MAX_ATTRIBS];
+
+   unsigned nr_attrib;
+};
+
+
+static struct translate_generic *translate_generic( struct translate *translate )
+{
+   return (struct translate_generic *)translate;
+}
+
+/**
+ * Fetch a float[4] vertex attribute from memory, doing format/type
+ * conversion as needed.
+ *
+ * This is probably needed/dupliocated elsewhere, eg format
+ * conversion, texture sampling etc.
+ */
+#define ATTRIB( NAME, SZ, TYPE, FROM, TO )		\
+static void						\
+fetch_##NAME(const void *ptr, float *attrib)		\
+{							\
+   const float defaults[4] = { 0.0f,0.0f,0.0f,1.0f };	\
+   unsigned i;						\
+							\
+   for (i = 0; i < SZ; i++) {				\
+      attrib[i] = FROM(i);				\
+   }							\
+							\
+   for (; i < 4; i++) {					\
+      attrib[i] = defaults[i];				\
+   }							\
+}							\
+							\
+static void						\
+emit_##NAME(const float *attrib, void *ptr)		\
+{  \
+   unsigned i;						\
+   TYPE *out = (TYPE *)ptr;				\
+							\
+   for (i = 0; i < SZ; i++) {				\
+      out[i] = TO(attrib[i]);				\
+   }							\
+}
+
+
+#define FROM_64_FLOAT(i)   ((float) ((double *) ptr)[i])
+#define FROM_32_FLOAT(i)   (((float *) ptr)[i])
+
+#define FROM_8_USCALED(i)  ((float) ((unsigned char *) ptr)[i])
+#define FROM_16_USCALED(i) ((float) ((unsigned short *) ptr)[i])
+#define FROM_32_USCALED(i) ((float) ((unsigned int *) ptr)[i])
+
+#define FROM_8_SSCALED(i)  ((float) ((char *) ptr)[i])
+#define FROM_16_SSCALED(i) ((float) ((short *) ptr)[i])
+#define FROM_32_SSCALED(i) ((float) ((int *) ptr)[i])
+
+#define FROM_8_UNORM(i)    ((float) ((unsigned char *) ptr)[i] / 255.0f)
+#define FROM_16_UNORM(i)   ((float) ((unsigned short *) ptr)[i] / 65535.0f)
+#define FROM_32_UNORM(i)   ((float) ((unsigned int *) ptr)[i] / 4294967295.0f)
+
+#define FROM_8_SNORM(i)    ((float) ((char *) ptr)[i] / 127.0f)
+#define FROM_16_SNORM(i)   ((float) ((short *) ptr)[i] / 32767.0f)
+#define FROM_32_SNORM(i)   ((float) ((int *) ptr)[i] / 2147483647.0f)
+
+#define FROM_32_FIXED(i)   (((int *) ptr)[i] / 65536.0f)
+
+#define TO_64_FLOAT(x)   ((double) x)
+#define TO_32_FLOAT(x)   (x)
+
+#define TO_8_USCALED(x)  ((unsigned char) x)
+#define TO_16_USCALED(x) ((unsigned short) x)
+#define TO_32_USCALED(x) ((unsigned int) x)
+
+#define TO_8_SSCALED(x)  ((char) x)
+#define TO_16_SSCALED(x) ((short) x)
+#define TO_32_SSCALED(x) ((int) x)
+
+#define TO_8_UNORM(x)    ((unsigned char) (x * 255.0f))
+#define TO_16_UNORM(x)   ((unsigned short) (x * 65535.0f))
+#define TO_32_UNORM(x)   ((unsigned int) (x * 4294967295.0f))
+
+#define TO_8_SNORM(x)    ((char) (x * 127.0f))
+#define TO_16_SNORM(x)   ((short) (x * 32767.0f))
+#define TO_32_SNORM(x)   ((int) (x * 2147483647.0f))
+
+#define TO_32_FIXED(x)   ((int) (x * 65536.0f))
+
+
+
+ATTRIB( R64G64B64A64_FLOAT,   4, double, FROM_64_FLOAT, TO_64_FLOAT )
+ATTRIB( R64G64B64_FLOAT,      3, double, FROM_64_FLOAT, TO_64_FLOAT )
+ATTRIB( R64G64_FLOAT,         2, double, FROM_64_FLOAT, TO_64_FLOAT )
+ATTRIB( R64_FLOAT,            1, double, FROM_64_FLOAT, TO_64_FLOAT )
+
+ATTRIB( R32G32B32A32_FLOAT,   4, float, FROM_32_FLOAT, TO_32_FLOAT )
+ATTRIB( R32G32B32_FLOAT,      3, float, FROM_32_FLOAT, TO_32_FLOAT )
+ATTRIB( R32G32_FLOAT,         2, float, FROM_32_FLOAT, TO_32_FLOAT )
+ATTRIB( R32_FLOAT,            1, float, FROM_32_FLOAT, TO_32_FLOAT )
+
+ATTRIB( R32G32B32A32_USCALED, 4, unsigned, FROM_32_USCALED, TO_32_USCALED )
+ATTRIB( R32G32B32_USCALED,    3, unsigned, FROM_32_USCALED, TO_32_USCALED )
+ATTRIB( R32G32_USCALED,       2, unsigned, FROM_32_USCALED, TO_32_USCALED )
+ATTRIB( R32_USCALED,          1, unsigned, FROM_32_USCALED, TO_32_USCALED )
+
+ATTRIB( R32G32B32A32_SSCALED, 4, int, FROM_32_SSCALED, TO_32_SSCALED )
+ATTRIB( R32G32B32_SSCALED,    3, int, FROM_32_SSCALED, TO_32_SSCALED )
+ATTRIB( R32G32_SSCALED,       2, int, FROM_32_SSCALED, TO_32_SSCALED )
+ATTRIB( R32_SSCALED,          1, int, FROM_32_SSCALED, TO_32_SSCALED )
+
+ATTRIB( R32G32B32A32_UNORM, 4, unsigned, FROM_32_UNORM, TO_32_UNORM )
+ATTRIB( R32G32B32_UNORM,    3, unsigned, FROM_32_UNORM, TO_32_UNORM )
+ATTRIB( R32G32_UNORM,       2, unsigned, FROM_32_UNORM, TO_32_UNORM )
+ATTRIB( R32_UNORM,          1, unsigned, FROM_32_UNORM, TO_32_UNORM )
+
+ATTRIB( R32G32B32A32_SNORM, 4, int, FROM_32_SNORM, TO_32_SNORM )
+ATTRIB( R32G32B32_SNORM,    3, int, FROM_32_SNORM, TO_32_SNORM )
+ATTRIB( R32G32_SNORM,       2, int, FROM_32_SNORM, TO_32_SNORM )
+ATTRIB( R32_SNORM,          1, int, FROM_32_SNORM, TO_32_SNORM )
+
+ATTRIB( R16G16B16A16_USCALED, 4, ushort, FROM_16_USCALED, TO_16_USCALED )
+ATTRIB( R16G16B16_USCALED,    3, ushort, FROM_16_USCALED, TO_16_USCALED )
+ATTRIB( R16G16_USCALED,       2, ushort, FROM_16_USCALED, TO_16_USCALED )
+ATTRIB( R16_USCALED,          1, ushort, FROM_16_USCALED, TO_16_USCALED )
+
+ATTRIB( R16G16B16A16_SSCALED, 4, short, FROM_16_SSCALED, TO_16_SSCALED )
+ATTRIB( R16G16B16_SSCALED,    3, short, FROM_16_SSCALED, TO_16_SSCALED )
+ATTRIB( R16G16_SSCALED,       2, short, FROM_16_SSCALED, TO_16_SSCALED )
+ATTRIB( R16_SSCALED,          1, short, FROM_16_SSCALED, TO_16_SSCALED )
+
+ATTRIB( R16G16B16A16_UNORM, 4, ushort, FROM_16_UNORM, TO_16_UNORM )
+ATTRIB( R16G16B16_UNORM,    3, ushort, FROM_16_UNORM, TO_16_UNORM )
+ATTRIB( R16G16_UNORM,       2, ushort, FROM_16_UNORM, TO_16_UNORM )
+ATTRIB( R16_UNORM,          1, ushort, FROM_16_UNORM, TO_16_UNORM )
+
+ATTRIB( R16G16B16A16_SNORM, 4, short, FROM_16_SNORM, TO_16_SNORM )
+ATTRIB( R16G16B16_SNORM,    3, short, FROM_16_SNORM, TO_16_SNORM )
+ATTRIB( R16G16_SNORM,       2, short, FROM_16_SNORM, TO_16_SNORM )
+ATTRIB( R16_SNORM,          1, short, FROM_16_SNORM, TO_16_SNORM )
+
+ATTRIB( R8G8B8A8_USCALED,   4, ubyte, FROM_8_USCALED, TO_8_USCALED )
+ATTRIB( R8G8B8_USCALED,     3, ubyte, FROM_8_USCALED, TO_8_USCALED )
+ATTRIB( R8G8_USCALED,       2, ubyte, FROM_8_USCALED, TO_8_USCALED )
+ATTRIB( R8_USCALED,         1, ubyte, FROM_8_USCALED, TO_8_USCALED )
+
+ATTRIB( R8G8B8A8_SSCALED,  4, char, FROM_8_SSCALED, TO_8_SSCALED )
+ATTRIB( R8G8B8_SSCALED,    3, char, FROM_8_SSCALED, TO_8_SSCALED )
+ATTRIB( R8G8_SSCALED,      2, char, FROM_8_SSCALED, TO_8_SSCALED )
+ATTRIB( R8_SSCALED,        1, char, FROM_8_SSCALED, TO_8_SSCALED )
+
+ATTRIB( R8G8B8A8_UNORM,  4, ubyte, FROM_8_UNORM, TO_8_UNORM )
+ATTRIB( R8G8B8_UNORM,    3, ubyte, FROM_8_UNORM, TO_8_UNORM )
+ATTRIB( R8G8_UNORM,      2, ubyte, FROM_8_UNORM, TO_8_UNORM )
+ATTRIB( R8_UNORM,        1, ubyte, FROM_8_UNORM, TO_8_UNORM )
+
+ATTRIB( R8G8B8A8_SNORM,  4, char, FROM_8_SNORM, TO_8_SNORM )
+ATTRIB( R8G8B8_SNORM,    3, char, FROM_8_SNORM, TO_8_SNORM )
+ATTRIB( R8G8_SNORM,      2, char, FROM_8_SNORM, TO_8_SNORM )
+ATTRIB( R8_SNORM,        1, char, FROM_8_SNORM, TO_8_SNORM )
+
+ATTRIB( A8R8G8B8_UNORM,       4, ubyte, FROM_8_UNORM, TO_8_UNORM )
+//ATTRIB( R8G8B8A8_UNORM,       4, ubyte, FROM_8_UNORM, TO_8_UNORM )
+
+ATTRIB( R32G32B32A32_FIXED,   4, int, FROM_32_FIXED, TO_32_FIXED )
+ATTRIB( R32G32B32_FIXED,      3, int, FROM_32_FIXED, TO_32_FIXED )
+ATTRIB( R32G32_FIXED,         2, int, FROM_32_FIXED, TO_32_FIXED )
+ATTRIB( R32_FIXED,            1, int, FROM_32_FIXED, TO_32_FIXED )
+
+
+
+static void
+fetch_B8G8R8A8_UNORM(const void *ptr, float *attrib)
+{
+   attrib[2] = FROM_8_UNORM(0);
+   attrib[1] = FROM_8_UNORM(1);
+   attrib[0] = FROM_8_UNORM(2);
+   attrib[3] = FROM_8_UNORM(3);
+}
+
+static void
+emit_B8G8R8A8_UNORM( const float *attrib, void *ptr)
+{
+   ubyte *out = (ubyte *)ptr;
+   out[2] = TO_8_UNORM(attrib[0]);
+   out[1] = TO_8_UNORM(attrib[1]);
+   out[0] = TO_8_UNORM(attrib[2]);
+   out[3] = TO_8_UNORM(attrib[3]);
+}
+
+static void 
+fetch_NULL( const void *ptr, float *attrib )
+{
+   attrib[0] = 0;
+   attrib[1] = 0;
+   attrib[2] = 0;
+   attrib[3] = 1;
+}
+
+static void 
+emit_NULL( const float *attrib, void *ptr )
+{
+   /* do nothing is the only sensible option */
+}
+
+static fetch_func get_fetch_func( enum pipe_format format )
+{
+   switch (format) {
+   case PIPE_FORMAT_R64_FLOAT:
+      return &fetch_R64_FLOAT;
+   case PIPE_FORMAT_R64G64_FLOAT:
+      return &fetch_R64G64_FLOAT;
+   case PIPE_FORMAT_R64G64B64_FLOAT:
+      return &fetch_R64G64B64_FLOAT;
+   case PIPE_FORMAT_R64G64B64A64_FLOAT:
+      return &fetch_R64G64B64A64_FLOAT;
+
+   case PIPE_FORMAT_R32_FLOAT:
+      return &fetch_R32_FLOAT;
+   case PIPE_FORMAT_R32G32_FLOAT:
+      return &fetch_R32G32_FLOAT;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      return &fetch_R32G32B32_FLOAT;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      return &fetch_R32G32B32A32_FLOAT;
+
+   case PIPE_FORMAT_R32_UNORM:
+      return &fetch_R32_UNORM;
+   case PIPE_FORMAT_R32G32_UNORM:
+      return &fetch_R32G32_UNORM;
+   case PIPE_FORMAT_R32G32B32_UNORM:
+      return &fetch_R32G32B32_UNORM;
+   case PIPE_FORMAT_R32G32B32A32_UNORM:
+      return &fetch_R32G32B32A32_UNORM;
+
+   case PIPE_FORMAT_R32_USCALED:
+      return &fetch_R32_USCALED;
+   case PIPE_FORMAT_R32G32_USCALED:
+      return &fetch_R32G32_USCALED;
+   case PIPE_FORMAT_R32G32B32_USCALED:
+      return &fetch_R32G32B32_USCALED;
+   case PIPE_FORMAT_R32G32B32A32_USCALED:
+      return &fetch_R32G32B32A32_USCALED;
+
+   case PIPE_FORMAT_R32_SNORM:
+      return &fetch_R32_SNORM;
+   case PIPE_FORMAT_R32G32_SNORM:
+      return &fetch_R32G32_SNORM;
+   case PIPE_FORMAT_R32G32B32_SNORM:
+      return &fetch_R32G32B32_SNORM;
+   case PIPE_FORMAT_R32G32B32A32_SNORM:
+      return &fetch_R32G32B32A32_SNORM;
+
+   case PIPE_FORMAT_R32_SSCALED:
+      return &fetch_R32_SSCALED;
+   case PIPE_FORMAT_R32G32_SSCALED:
+      return &fetch_R32G32_SSCALED;
+   case PIPE_FORMAT_R32G32B32_SSCALED:
+      return &fetch_R32G32B32_SSCALED;
+   case PIPE_FORMAT_R32G32B32A32_SSCALED:
+      return &fetch_R32G32B32A32_SSCALED;
+
+   case PIPE_FORMAT_R16_UNORM:
+      return &fetch_R16_UNORM;
+   case PIPE_FORMAT_R16G16_UNORM:
+      return &fetch_R16G16_UNORM;
+   case PIPE_FORMAT_R16G16B16_UNORM:
+      return &fetch_R16G16B16_UNORM;
+   case PIPE_FORMAT_R16G16B16A16_UNORM:
+      return &fetch_R16G16B16A16_UNORM;
+
+   case PIPE_FORMAT_R16_USCALED:
+      return &fetch_R16_USCALED;
+   case PIPE_FORMAT_R16G16_USCALED:
+      return &fetch_R16G16_USCALED;
+   case PIPE_FORMAT_R16G16B16_USCALED:
+      return &fetch_R16G16B16_USCALED;
+   case PIPE_FORMAT_R16G16B16A16_USCALED:
+      return &fetch_R16G16B16A16_USCALED;
+
+   case PIPE_FORMAT_R16_SNORM:
+      return &fetch_R16_SNORM;
+   case PIPE_FORMAT_R16G16_SNORM:
+      return &fetch_R16G16_SNORM;
+   case PIPE_FORMAT_R16G16B16_SNORM:
+      return &fetch_R16G16B16_SNORM;
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      return &fetch_R16G16B16A16_SNORM;
+
+   case PIPE_FORMAT_R16_SSCALED:
+      return &fetch_R16_SSCALED;
+   case PIPE_FORMAT_R16G16_SSCALED:
+      return &fetch_R16G16_SSCALED;
+   case PIPE_FORMAT_R16G16B16_SSCALED:
+      return &fetch_R16G16B16_SSCALED;
+   case PIPE_FORMAT_R16G16B16A16_SSCALED:
+      return &fetch_R16G16B16A16_SSCALED;
+
+   case PIPE_FORMAT_R8_UNORM:
+      return &fetch_R8_UNORM;
+   case PIPE_FORMAT_R8G8_UNORM:
+      return &fetch_R8G8_UNORM;
+   case PIPE_FORMAT_R8G8B8_UNORM:
+      return &fetch_R8G8B8_UNORM;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      return &fetch_R8G8B8A8_UNORM;
+
+   case PIPE_FORMAT_R8_USCALED:
+      return &fetch_R8_USCALED;
+   case PIPE_FORMAT_R8G8_USCALED:
+      return &fetch_R8G8_USCALED;
+   case PIPE_FORMAT_R8G8B8_USCALED:
+      return &fetch_R8G8B8_USCALED;
+   case PIPE_FORMAT_R8G8B8A8_USCALED:
+      return &fetch_R8G8B8A8_USCALED;
+
+   case PIPE_FORMAT_R8_SNORM:
+      return &fetch_R8_SNORM;
+   case PIPE_FORMAT_R8G8_SNORM:
+      return &fetch_R8G8_SNORM;
+   case PIPE_FORMAT_R8G8B8_SNORM:
+      return &fetch_R8G8B8_SNORM;
+   case PIPE_FORMAT_R8G8B8A8_SNORM:
+      return &fetch_R8G8B8A8_SNORM;
+
+   case PIPE_FORMAT_R8_SSCALED:
+      return &fetch_R8_SSCALED;
+   case PIPE_FORMAT_R8G8_SSCALED:
+      return &fetch_R8G8_SSCALED;
+   case PIPE_FORMAT_R8G8B8_SSCALED:
+      return &fetch_R8G8B8_SSCALED;
+   case PIPE_FORMAT_R8G8B8A8_SSCALED:
+      return &fetch_R8G8B8A8_SSCALED;
+
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      return &fetch_A8R8G8B8_UNORM;
+
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      return &fetch_B8G8R8A8_UNORM;
+
+   case PIPE_FORMAT_R32_FIXED:
+      return &fetch_R32_FIXED;
+   case PIPE_FORMAT_R32G32_FIXED:
+      return &fetch_R32G32_FIXED;
+   case PIPE_FORMAT_R32G32B32_FIXED:
+      return &fetch_R32G32B32_FIXED;
+   case PIPE_FORMAT_R32G32B32A32_FIXED:
+      return &fetch_R32G32B32A32_FIXED;
+
+   default:
+      assert(0); 
+      return &fetch_NULL;
+   }
+}
+
+
+
+
+static emit_func get_emit_func( enum pipe_format format )
+{
+   /* silence warnings */
+   (void) emit_R32G32B32A32_FIXED;
+   (void) emit_R32G32B32_FIXED;
+   (void) emit_R32G32_FIXED;
+   (void) emit_R32_FIXED;
+
+   switch (format) {
+   case PIPE_FORMAT_R64_FLOAT:
+      return &emit_R64_FLOAT;
+   case PIPE_FORMAT_R64G64_FLOAT:
+      return &emit_R64G64_FLOAT;
+   case PIPE_FORMAT_R64G64B64_FLOAT:
+      return &emit_R64G64B64_FLOAT;
+   case PIPE_FORMAT_R64G64B64A64_FLOAT:
+      return &emit_R64G64B64A64_FLOAT;
+
+   case PIPE_FORMAT_R32_FLOAT:
+      return &emit_R32_FLOAT;
+   case PIPE_FORMAT_R32G32_FLOAT:
+      return &emit_R32G32_FLOAT;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      return &emit_R32G32B32_FLOAT;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      return &emit_R32G32B32A32_FLOAT;
+
+   case PIPE_FORMAT_R32_UNORM:
+      return &emit_R32_UNORM;
+   case PIPE_FORMAT_R32G32_UNORM:
+      return &emit_R32G32_UNORM;
+   case PIPE_FORMAT_R32G32B32_UNORM:
+      return &emit_R32G32B32_UNORM;
+   case PIPE_FORMAT_R32G32B32A32_UNORM:
+      return &emit_R32G32B32A32_UNORM;
+
+   case PIPE_FORMAT_R32_USCALED:
+      return &emit_R32_USCALED;
+   case PIPE_FORMAT_R32G32_USCALED:
+      return &emit_R32G32_USCALED;
+   case PIPE_FORMAT_R32G32B32_USCALED:
+      return &emit_R32G32B32_USCALED;
+   case PIPE_FORMAT_R32G32B32A32_USCALED:
+      return &emit_R32G32B32A32_USCALED;
+
+   case PIPE_FORMAT_R32_SNORM:
+      return &emit_R32_SNORM;
+   case PIPE_FORMAT_R32G32_SNORM:
+      return &emit_R32G32_SNORM;
+   case PIPE_FORMAT_R32G32B32_SNORM:
+      return &emit_R32G32B32_SNORM;
+   case PIPE_FORMAT_R32G32B32A32_SNORM:
+      return &emit_R32G32B32A32_SNORM;
+
+   case PIPE_FORMAT_R32_SSCALED:
+      return &emit_R32_SSCALED;
+   case PIPE_FORMAT_R32G32_SSCALED:
+      return &emit_R32G32_SSCALED;
+   case PIPE_FORMAT_R32G32B32_SSCALED:
+      return &emit_R32G32B32_SSCALED;
+   case PIPE_FORMAT_R32G32B32A32_SSCALED:
+      return &emit_R32G32B32A32_SSCALED;
+
+   case PIPE_FORMAT_R16_UNORM:
+      return &emit_R16_UNORM;
+   case PIPE_FORMAT_R16G16_UNORM:
+      return &emit_R16G16_UNORM;
+   case PIPE_FORMAT_R16G16B16_UNORM:
+      return &emit_R16G16B16_UNORM;
+   case PIPE_FORMAT_R16G16B16A16_UNORM:
+      return &emit_R16G16B16A16_UNORM;
+
+   case PIPE_FORMAT_R16_USCALED:
+      return &emit_R16_USCALED;
+   case PIPE_FORMAT_R16G16_USCALED:
+      return &emit_R16G16_USCALED;
+   case PIPE_FORMAT_R16G16B16_USCALED:
+      return &emit_R16G16B16_USCALED;
+   case PIPE_FORMAT_R16G16B16A16_USCALED:
+      return &emit_R16G16B16A16_USCALED;
+
+   case PIPE_FORMAT_R16_SNORM:
+      return &emit_R16_SNORM;
+   case PIPE_FORMAT_R16G16_SNORM:
+      return &emit_R16G16_SNORM;
+   case PIPE_FORMAT_R16G16B16_SNORM:
+      return &emit_R16G16B16_SNORM;
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      return &emit_R16G16B16A16_SNORM;
+
+   case PIPE_FORMAT_R16_SSCALED:
+      return &emit_R16_SSCALED;
+   case PIPE_FORMAT_R16G16_SSCALED:
+      return &emit_R16G16_SSCALED;
+   case PIPE_FORMAT_R16G16B16_SSCALED:
+      return &emit_R16G16B16_SSCALED;
+   case PIPE_FORMAT_R16G16B16A16_SSCALED:
+      return &emit_R16G16B16A16_SSCALED;
+
+   case PIPE_FORMAT_R8_UNORM:
+      return &emit_R8_UNORM;
+   case PIPE_FORMAT_R8G8_UNORM:
+      return &emit_R8G8_UNORM;
+   case PIPE_FORMAT_R8G8B8_UNORM:
+      return &emit_R8G8B8_UNORM;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      return &emit_R8G8B8A8_UNORM;
+
+   case PIPE_FORMAT_R8_USCALED:
+      return &emit_R8_USCALED;
+   case PIPE_FORMAT_R8G8_USCALED:
+      return &emit_R8G8_USCALED;
+   case PIPE_FORMAT_R8G8B8_USCALED:
+      return &emit_R8G8B8_USCALED;
+   case PIPE_FORMAT_R8G8B8A8_USCALED:
+      return &emit_R8G8B8A8_USCALED;
+
+   case PIPE_FORMAT_R8_SNORM:
+      return &emit_R8_SNORM;
+   case PIPE_FORMAT_R8G8_SNORM:
+      return &emit_R8G8_SNORM;
+   case PIPE_FORMAT_R8G8B8_SNORM:
+      return &emit_R8G8B8_SNORM;
+   case PIPE_FORMAT_R8G8B8A8_SNORM:
+      return &emit_R8G8B8A8_SNORM;
+
+   case PIPE_FORMAT_R8_SSCALED:
+      return &emit_R8_SSCALED;
+   case PIPE_FORMAT_R8G8_SSCALED:
+      return &emit_R8G8_SSCALED;
+   case PIPE_FORMAT_R8G8B8_SSCALED:
+      return &emit_R8G8B8_SSCALED;
+   case PIPE_FORMAT_R8G8B8A8_SSCALED:
+      return &emit_R8G8B8A8_SSCALED;
+
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      return &emit_A8R8G8B8_UNORM;
+
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      return &emit_B8G8R8A8_UNORM;
+
+   default:
+      assert(0); 
+      return &emit_NULL;
+   }
+}
+
+
+
+/**
+ * Fetch vertex attributes for 'count' vertices.
+ */
+static void PIPE_CDECL generic_run_elts( struct translate *translate,
+                                         const unsigned *elts,
+                                         unsigned count,
+                                         void *output_buffer )
+{
+   struct translate_generic *tg = translate_generic(translate);
+   char *vert = output_buffer;
+   unsigned nr_attrs = tg->nr_attrib;
+   unsigned attr;
+   unsigned i;
+
+   /* loop over vertex attributes (vertex shader inputs)
+    */
+   for (i = 0; i < count; i++) {
+      unsigned elt = *elts++;
+
+      for (attr = 0; attr < nr_attrs; attr++) {
+	 float data[4];
+
+	 const char *src = (tg->attrib[attr].input_ptr + 
+			    tg->attrib[attr].input_stride * elt);
+
+	 char *dst = (vert + 
+		      tg->attrib[attr].output_offset);
+
+	 tg->attrib[attr].fetch( src, data );
+
+         if (0) debug_printf("vert %d/%d attr %d: %f %f %f %f\n",
+                             i, elt, attr, data[0], data[1], data[2], data[3]);
+
+	 tg->attrib[attr].emit( data, dst );
+      }
+      
+      vert += tg->translate.key.output_stride;
+   }
+}
+
+
+
+static void PIPE_CDECL generic_run( struct translate *translate,
+                                    unsigned start,
+                                    unsigned count,
+                                    void *output_buffer )
+{
+   struct translate_generic *tg = translate_generic(translate);
+   char *vert = output_buffer;
+   unsigned nr_attrs = tg->nr_attrib;
+   unsigned attr;
+   unsigned i;
+
+   /* loop over vertex attributes (vertex shader inputs)
+    */
+   for (i = 0; i < count; i++) {
+      unsigned elt = start + i;
+
+      for (attr = 0; attr < nr_attrs; attr++) {
+	 float data[4];
+
+	 const char *src = (tg->attrib[attr].input_ptr + 
+			    tg->attrib[attr].input_stride * elt);
+
+	 char *dst = (vert + 
+		      tg->attrib[attr].output_offset);
+
+	 tg->attrib[attr].fetch( src, data );
+
+         if (0) debug_printf("vert %d attr %d: %f %f %f %f\n",
+                             i, attr, data[0], data[1], data[2], data[3]);
+
+	 tg->attrib[attr].emit( data, dst );
+      }
+      
+      vert += tg->translate.key.output_stride;
+   }
+}
+
+
+			       
+static void generic_set_buffer( struct translate *translate,
+				unsigned buf,
+				const void *ptr,
+				unsigned stride )
+{
+   struct translate_generic *tg = translate_generic(translate);
+   unsigned i;
+
+   for (i = 0; i < tg->nr_attrib; i++) {
+      if (tg->attrib[i].buffer == buf) {
+	 tg->attrib[i].input_ptr = ((char *)ptr +
+				    tg->attrib[i].input_offset);
+	 tg->attrib[i].input_stride = stride;
+      }
+   }
+}
+
+
+static void generic_release( struct translate *translate )
+{
+   /* Refcount?
+    */
+   FREE(translate);
+}
+
+struct translate *translate_generic_create( const struct translate_key *key )
+{
+   struct translate_generic *tg = CALLOC_STRUCT(translate_generic);
+   unsigned i;
+
+   if (tg == NULL)
+      return NULL;
+
+   tg->translate.key = *key;
+   tg->translate.release = generic_release;
+   tg->translate.set_buffer = generic_set_buffer;
+   tg->translate.run_elts = generic_run_elts;
+   tg->translate.run = generic_run;
+
+   for (i = 0; i < key->nr_elements; i++) {
+
+      tg->attrib[i].fetch = get_fetch_func(key->element[i].input_format);
+      tg->attrib[i].buffer = key->element[i].input_buffer;
+      tg->attrib[i].input_offset = key->element[i].input_offset;
+
+      tg->attrib[i].emit = get_emit_func(key->element[i].output_format);
+      tg->attrib[i].output_offset = key->element[i].output_offset;
+
+   }
+
+   tg->nr_attrib = key->nr_elements;
+
+
+   return &tg->translate;
+}
diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c
new file mode 100644
index 0000000000..b62db8d8f3
--- /dev/null
+++ b/src/gallium/auxiliary/translate/translate_sse.c
@@ -0,0 +1,706 @@
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keithw@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_config.h"
+#include "pipe/p_compiler.h"
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "translate.h"
+
+
+#if defined(PIPE_ARCH_X86)
+
+#include "rtasm/rtasm_cpu.h"
+#include "rtasm/rtasm_x86sse.h"
+
+
+#define X    0
+#define Y    1
+#define Z    2
+#define W    3
+
+
+typedef void (PIPE_CDECL *run_func)( struct translate *translate,
+                                     unsigned start,
+                                     unsigned count,
+                                     void *output_buffer );
+
+typedef void (PIPE_CDECL *run_elts_func)( struct translate *translate,
+                                          const unsigned *elts,
+                                          unsigned count,
+                                          void *output_buffer );
+
+struct translate_buffer {
+   const void *base_ptr;
+   unsigned stride;
+   void *ptr;                   /* updated per vertex */
+};
+
+
+struct translate_sse {
+   struct translate translate;
+
+   struct x86_function linear_func;
+   struct x86_function elt_func;
+   struct x86_function *func;
+
+   boolean loaded_identity;
+   boolean loaded_255;
+   boolean loaded_inv_255;
+
+   float identity[4];
+   float float_255[4];
+   float inv_255[4];
+
+   struct translate_buffer buffer[PIPE_MAX_ATTRIBS];
+   unsigned nr_buffers;
+
+   run_func      gen_run;
+   run_elts_func gen_run_elts;
+
+   /* these are actually known values, but putting them in a struct
+    * like this is helpful to keep them in sync across the file.
+    */
+   struct x86_reg tmp_EAX;
+   struct x86_reg idx_EBX;     /* either start+i or &elt[i] */
+   struct x86_reg outbuf_ECX;
+   struct x86_reg machine_EDX;
+   struct x86_reg count_ESI;    /* decrements to zero */
+};
+
+static int get_offset( const void *a, const void *b )
+{
+   return (const char *)b - (const char *)a;
+}
+
+
+
+static struct x86_reg get_identity( struct translate_sse *p )
+{
+   struct x86_reg reg = x86_make_reg(file_XMM, 6);
+
+   if (!p->loaded_identity) {
+      p->loaded_identity = TRUE;
+      p->identity[0] = 0;
+      p->identity[1] = 0;
+      p->identity[2] = 0;
+      p->identity[3] = 1;
+
+      sse_movups(p->func, reg, 
+		 x86_make_disp(p->machine_EDX, 
+			       get_offset(p, &p->identity[0])));
+   }
+
+   return reg;
+}
+
+static struct x86_reg get_255( struct translate_sse *p )
+{
+   struct x86_reg reg = x86_make_reg(file_XMM, 7);
+
+   if (!p->loaded_255) {
+      p->loaded_255 = TRUE;
+      p->float_255[0] =
+	 p->float_255[1] =
+	 p->float_255[2] =
+	 p->float_255[3] = 255.0f;
+
+      sse_movups(p->func, reg, 
+		 x86_make_disp(p->machine_EDX, 
+			       get_offset(p, &p->float_255[0])));
+   }
+
+   return reg;
+}
+
+static struct x86_reg get_inv_255( struct translate_sse *p )
+{
+   struct x86_reg reg = x86_make_reg(file_XMM, 5);
+
+   if (!p->loaded_inv_255) {
+      p->loaded_inv_255 = TRUE;
+      p->inv_255[0] =
+	 p->inv_255[1] =
+	 p->inv_255[2] =
+	 p->inv_255[3] = 1.0f / 255.0f;
+
+      sse_movups(p->func, reg, 
+		 x86_make_disp(p->machine_EDX, 
+			       get_offset(p, &p->inv_255[0])));
+   }
+
+   return reg;
+}
+
+
+static void emit_load_R32G32B32A32( struct translate_sse *p, 			   
+				    struct x86_reg data,
+				    struct x86_reg arg0 )
+{
+   sse_movups(p->func, data, arg0);
+}
+
+static void emit_load_R32G32B32( struct translate_sse *p, 			   
+				 struct x86_reg data,
+				 struct x86_reg arg0 )
+{
+   /* Have to jump through some hoops:
+    *
+    * c 0 0 0
+    * c 0 0 1
+    * 0 0 c 1
+    * a b c 1
+    */
+   sse_movss(p->func, data, x86_make_disp(arg0, 8));
+   sse_shufps(p->func, data, get_identity(p), SHUF(X,Y,Z,W) );
+   sse_shufps(p->func, data, data, SHUF(Y,Z,X,W) );
+   sse_movlps(p->func, data, arg0);
+}
+
+static void emit_load_R32G32( struct translate_sse *p, 
+			   struct x86_reg data,
+			   struct x86_reg arg0 )
+{
+   /* 0 0 0 1
+    * a b 0 1
+    */
+   sse_movups(p->func, data, get_identity(p) );
+   sse_movlps(p->func, data, arg0);
+}
+
+
+static void emit_load_R32( struct translate_sse *p, 
+			   struct x86_reg data,
+			   struct x86_reg arg0 )
+{
+   /* a 0 0 0
+    * a 0 0 1
+    */
+   sse_movss(p->func, data, arg0);
+   sse_orps(p->func, data, get_identity(p) );
+}
+
+
+static void emit_load_R8G8B8A8_UNORM( struct translate_sse *p,
+				       struct x86_reg data,
+				       struct x86_reg src )
+{
+
+   /* Load and unpack twice:
+    */
+   sse_movss(p->func, data, src);
+   sse2_punpcklbw(p->func, data, get_identity(p));
+   sse2_punpcklbw(p->func, data, get_identity(p));
+
+   /* Convert to float:
+    */
+   sse2_cvtdq2ps(p->func, data, data);
+
+
+   /* Scale by 1/255.0
+    */
+   sse_mulps(p->func, data, get_inv_255(p));
+}
+
+
+
+
+static void emit_store_R32G32B32A32( struct translate_sse *p, 			   
+				     struct x86_reg dest,
+				     struct x86_reg dataXMM )
+{
+   sse_movups(p->func, dest, dataXMM);
+}
+
+static void emit_store_R32G32B32( struct translate_sse *p, 
+				  struct x86_reg dest,
+				  struct x86_reg dataXMM )
+{
+   /* Emit two, shuffle, emit one.
+    */
+   sse_movlps(p->func, dest, dataXMM);
+   sse_shufps(p->func, dataXMM, dataXMM, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */
+   sse_movss(p->func, x86_make_disp(dest,8), dataXMM);
+}
+
+static void emit_store_R32G32( struct translate_sse *p, 
+			       struct x86_reg dest,
+			       struct x86_reg dataXMM )
+{
+   sse_movlps(p->func, dest, dataXMM);
+}
+
+static void emit_store_R32( struct translate_sse *p, 
+			    struct x86_reg dest,
+			    struct x86_reg dataXMM )
+{
+   sse_movss(p->func, dest, dataXMM);
+}
+
+
+
+static void emit_store_R8G8B8A8_UNORM( struct translate_sse *p,
+				       struct x86_reg dest,
+				       struct x86_reg dataXMM )
+{
+   /* Scale by 255.0
+    */
+   sse_mulps(p->func, dataXMM, get_255(p));
+
+   /* Pack and emit:
+    */
+   sse2_cvtps2dq(p->func, dataXMM, dataXMM);
+   sse2_packssdw(p->func, dataXMM, dataXMM);
+   sse2_packuswb(p->func, dataXMM, dataXMM);
+   sse_movss(p->func, dest, dataXMM);
+}
+
+
+
+
+
+/* Extended swizzles?  Maybe later.
+ */  
+static void emit_swizzle( struct translate_sse *p,
+			  struct x86_reg dest,
+			  struct x86_reg src,
+			  unsigned char shuffle )
+{
+   sse_shufps(p->func, dest, src, shuffle);
+}
+
+
+static boolean translate_attr( struct translate_sse *p,
+			       const struct translate_element *a,
+			       struct x86_reg srcECX,
+			       struct x86_reg dstEAX)
+{
+   struct x86_reg dataXMM = x86_make_reg(file_XMM, 0);
+
+   switch (a->input_format) {
+   case PIPE_FORMAT_R32_FLOAT:
+      emit_load_R32(p, dataXMM, srcECX);
+      break;
+   case PIPE_FORMAT_R32G32_FLOAT:
+      emit_load_R32G32(p, dataXMM, srcECX);
+      break;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      emit_load_R32G32B32(p, dataXMM, srcECX);
+      break;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      emit_load_R32G32B32A32(p, dataXMM, srcECX);
+      break;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      emit_load_R8G8B8A8_UNORM(p, dataXMM, srcECX);
+      emit_swizzle(p, dataXMM, dataXMM, SHUF(Z,Y,X,W));
+      break;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      emit_load_R8G8B8A8_UNORM(p, dataXMM, srcECX);
+      break;
+   default:
+      return FALSE;
+   }
+
+   switch (a->output_format) {
+   case PIPE_FORMAT_R32_FLOAT:
+      emit_store_R32(p, dstEAX, dataXMM);
+      break;
+   case PIPE_FORMAT_R32G32_FLOAT:
+      emit_store_R32G32(p, dstEAX, dataXMM);
+      break;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      emit_store_R32G32B32(p, dstEAX, dataXMM);
+      break;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      emit_store_R32G32B32A32(p, dstEAX, dataXMM);
+      break;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      emit_swizzle(p, dataXMM, dataXMM, SHUF(Z,Y,X,W));
+      emit_store_R8G8B8A8_UNORM(p, dstEAX, dataXMM);
+      break;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      emit_store_R8G8B8A8_UNORM(p, dstEAX, dataXMM);
+      break;
+   default:
+      return FALSE;
+   }
+
+   return TRUE;
+}
+
+
+static boolean init_inputs( struct translate_sse *p,
+                            boolean linear )
+{
+   unsigned i;
+   if (linear) {
+      for (i = 0; i < p->nr_buffers; i++) {
+         struct x86_reg buf_stride   = x86_make_disp(p->machine_EDX,
+                                                     get_offset(p, &p->buffer[i].stride));
+         struct x86_reg buf_ptr      = x86_make_disp(p->machine_EDX,
+                                                     get_offset(p, &p->buffer[i].ptr));
+         struct x86_reg buf_base_ptr = x86_make_disp(p->machine_EDX,
+                                                     get_offset(p, &p->buffer[i].base_ptr));
+         struct x86_reg elt = p->idx_EBX;
+         struct x86_reg tmp = p->tmp_EAX;
+
+
+         /* Calculate pointer to first attrib:
+          */
+         x86_mov(p->func, tmp, buf_stride);
+         x86_imul(p->func, tmp, elt);
+         x86_add(p->func, tmp, buf_base_ptr);
+
+
+         /* In the linear case, keep the buffer pointer instead of the
+          * index number.
+          */
+         if (p->nr_buffers == 1) 
+            x86_mov( p->func, elt, tmp );
+         else
+            x86_mov( p->func, buf_ptr, tmp );
+      }
+   }
+
+   return TRUE;
+}
+
+
+static struct x86_reg get_buffer_ptr( struct translate_sse *p,
+                                      boolean linear,
+                                      unsigned buf_idx,
+                                      struct x86_reg elt )
+{
+   if (linear && p->nr_buffers == 1) {
+      return p->idx_EBX;
+   }
+   else if (linear) {
+      struct x86_reg ptr = p->tmp_EAX;
+      struct x86_reg buf_ptr = 
+         x86_make_disp(p->machine_EDX, 
+                       get_offset(p, &p->buffer[buf_idx].ptr));
+      
+      x86_mov(p->func, ptr, buf_ptr);
+      return ptr;
+   }
+   else {
+      struct x86_reg ptr = p->tmp_EAX;
+
+      struct x86_reg buf_stride = 
+         x86_make_disp(p->machine_EDX, 
+                       get_offset(p, &p->buffer[buf_idx].stride));
+
+      struct x86_reg buf_base_ptr = 
+         x86_make_disp(p->machine_EDX, 
+                       get_offset(p, &p->buffer[buf_idx].base_ptr));
+
+
+
+      /* Calculate pointer to current attrib:
+       */
+      x86_mov(p->func, ptr, buf_stride);
+      x86_imul(p->func, ptr, elt);
+      x86_add(p->func, ptr, buf_base_ptr);
+      return ptr;
+   }
+}
+
+
+
+static boolean incr_inputs( struct translate_sse *p, 
+                            boolean linear )
+{
+   if (linear && p->nr_buffers == 1) {
+      struct x86_reg stride = x86_make_disp(p->machine_EDX,
+                                            get_offset(p, &p->buffer[0].stride));
+
+      x86_add(p->func, p->idx_EBX, stride);
+      sse_prefetchnta(p->func, x86_make_disp(p->idx_EBX, 192));
+   }
+   else if (linear) {
+      unsigned i;
+
+      /* Is this worthwhile??
+       */
+      for (i = 0; i < p->nr_buffers; i++) {
+         struct x86_reg buf_ptr = x86_make_disp(p->machine_EDX,
+                                                get_offset(p, &p->buffer[i].ptr));
+         struct x86_reg buf_stride = x86_make_disp(p->machine_EDX,
+                                                   get_offset(p, &p->buffer[i].stride));
+
+         x86_mov(p->func, p->tmp_EAX, buf_ptr);
+         x86_add(p->func, p->tmp_EAX, buf_stride);
+         if (i == 0) sse_prefetchnta(p->func, x86_make_disp(p->tmp_EAX, 192));
+         x86_mov(p->func, buf_ptr, p->tmp_EAX);
+      }
+   } 
+   else {
+      x86_lea(p->func, p->idx_EBX, x86_make_disp(p->idx_EBX, 4));
+   }
+   
+   return TRUE;
+}
+
+
+/* Build run( struct translate *machine,
+ *            unsigned start,
+ *            unsigned count,
+ *            void *output_buffer )
+ * or
+ *  run_elts( struct translate *machine,
+ *            unsigned *elts,
+ *            unsigned count,
+ *            void *output_buffer )
+ *
+ *  Lots of hardcoding
+ *
+ * EAX -- pointer to current output vertex
+ * ECX -- pointer to current attribute 
+ * 
+ */
+static boolean build_vertex_emit( struct translate_sse *p,
+				  struct x86_function *func,
+				  boolean linear )
+{
+   int fixup, label;
+   unsigned j;
+
+   p->tmp_EAX       = x86_make_reg(file_REG32, reg_AX);
+   p->idx_EBX       = x86_make_reg(file_REG32, reg_BX);
+   p->outbuf_ECX    = x86_make_reg(file_REG32, reg_CX);
+   p->machine_EDX   = x86_make_reg(file_REG32, reg_DX);
+   p->count_ESI     = x86_make_reg(file_REG32, reg_SI);
+
+   p->func = func;
+   p->loaded_inv_255 = FALSE;
+   p->loaded_255 = FALSE;
+   p->loaded_identity = FALSE;
+
+   x86_init_func(p->func);
+
+   /* Push a few regs?
+    */
+   x86_push(p->func, p->idx_EBX);
+   x86_push(p->func, p->count_ESI);
+
+   /* Load arguments into regs:
+    */
+   x86_mov(p->func, p->machine_EDX, x86_fn_arg(p->func, 1));
+   x86_mov(p->func, p->idx_EBX, x86_fn_arg(p->func, 2));
+   x86_mov(p->func, p->count_ESI, x86_fn_arg(p->func, 3));
+   x86_mov(p->func, p->outbuf_ECX, x86_fn_arg(p->func, 4));
+
+   /* Get vertex count, compare to zero
+    */
+   x86_xor(p->func, p->tmp_EAX, p->tmp_EAX);
+   x86_cmp(p->func, p->count_ESI, p->tmp_EAX);
+   fixup = x86_jcc_forward(p->func, cc_E);
+
+   /* always load, needed or not:
+    */
+   init_inputs(p, linear);
+
+   /* Note address for loop jump
+    */
+   label = x86_get_label(p->func);
+   {
+      struct x86_reg elt = linear ? p->idx_EBX : x86_deref(p->idx_EBX);
+      int last_vb = -1;
+      struct x86_reg vb;
+
+      for (j = 0; j < p->translate.key.nr_elements; j++) {
+         const struct translate_element *a = &p->translate.key.element[j];
+
+         /* Figure out source pointer address:
+          */
+         if (a->input_buffer != last_vb) {
+            last_vb = a->input_buffer;
+            vb = get_buffer_ptr(p, linear, a->input_buffer, elt);
+         }
+         
+         if (!translate_attr( p, a, 
+                              x86_make_disp(vb, a->input_offset), 
+                              x86_make_disp(p->outbuf_ECX, a->output_offset)))
+            return FALSE;
+      }
+
+      /* Next output vertex:
+       */
+      x86_lea(p->func, 
+              p->outbuf_ECX, 
+              x86_make_disp(p->outbuf_ECX, 
+                            p->translate.key.output_stride));
+
+      /* Incr index
+       */ 
+      incr_inputs( p, linear );
+   }
+
+   /* decr count, loop if not zero
+    */
+   x86_dec(p->func, p->count_ESI);
+   x86_jcc(p->func, cc_NZ, label);
+
+   /* Exit mmx state?
+    */
+   if (p->func->need_emms)
+      mmx_emms(p->func);
+
+   /* Land forward jump here:
+    */
+   x86_fixup_fwd_jump(p->func, fixup);
+
+   /* Pop regs and return
+    */
+   
+   x86_pop(p->func, p->count_ESI);
+   x86_pop(p->func, p->idx_EBX);
+   x86_ret(p->func);
+
+   return TRUE;
+}
+
+
+
+
+
+
+			       
+static void translate_sse_set_buffer( struct translate *translate,
+				unsigned buf,
+				const void *ptr,
+				unsigned stride )
+{
+   struct translate_sse *p = (struct translate_sse *)translate;
+
+   if (buf < p->nr_buffers) {
+      p->buffer[buf].base_ptr = (char *)ptr;
+      p->buffer[buf].stride = stride;
+   }
+
+   if (0) debug_printf("%s %d/%d: %p %d\n", 
+                       __FUNCTION__, buf, 
+                       p->nr_buffers, 
+                       ptr, stride);
+}
+
+
+static void translate_sse_release( struct translate *translate )
+{
+   struct translate_sse *p = (struct translate_sse *)translate;
+
+   x86_release_func( &p->linear_func );
+   x86_release_func( &p->elt_func );
+
+   FREE(p);
+}
+
+static void PIPE_CDECL translate_sse_run_elts( struct translate *translate,
+			      const unsigned *elts,
+			      unsigned count,
+			      void *output_buffer )
+{
+   struct translate_sse *p = (struct translate_sse *)translate;
+
+   p->gen_run_elts( translate,
+		    elts,
+		    count,
+		    output_buffer );
+}
+
+static void PIPE_CDECL translate_sse_run( struct translate *translate,
+			 unsigned start,
+			 unsigned count,
+			 void *output_buffer )
+{
+   struct translate_sse *p = (struct translate_sse *)translate;
+
+   p->gen_run( translate,
+	       start,
+	       count,
+	       output_buffer );
+}
+
+
+struct translate *translate_sse2_create( const struct translate_key *key )
+{
+   struct translate_sse *p = NULL;
+   unsigned i;
+
+   if (!rtasm_cpu_has_sse() || !rtasm_cpu_has_sse2())
+      goto fail;
+
+   p = CALLOC_STRUCT( translate_sse );
+   if (p == NULL) 
+      goto fail;
+
+   p->translate.key = *key;
+   p->translate.release = translate_sse_release;
+   p->translate.set_buffer = translate_sse_set_buffer;
+   p->translate.run_elts = translate_sse_run_elts;
+   p->translate.run = translate_sse_run;
+
+   for (i = 0; i < key->nr_elements; i++) 
+      p->nr_buffers = MAX2( p->nr_buffers, key->element[i].input_buffer + 1 );
+
+   if (0) debug_printf("nr_buffers: %d\n", p->nr_buffers);
+
+   if (!build_vertex_emit(p, &p->linear_func, TRUE))
+      goto fail;
+
+   if (!build_vertex_emit(p, &p->elt_func, FALSE))
+      goto fail;
+
+   p->gen_run = (run_func)x86_get_func(&p->linear_func);
+   if (p->gen_run == NULL)
+      goto fail;
+
+   p->gen_run_elts = (run_elts_func)x86_get_func(&p->elt_func);
+   if (p->gen_run_elts == NULL)
+      goto fail;
+
+   return &p->translate;
+
+ fail:
+   if (p)
+      translate_sse_release( &p->translate );
+
+   return NULL;
+}
+
+
+
+#else
+
+struct translate *translate_sse2_create( const struct translate_key *key )
+{
+   return NULL;
+}
+
+#endif
diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile
new file mode 100644
index 0000000000..44c2377721
--- /dev/null
+++ b/src/gallium/auxiliary/util/Makefile
@@ -0,0 +1,32 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = util
+
+C_SOURCES = \
+	p_debug.c \
+	u_blit.c \
+	u_cache.c \
+	u_draw_quad.c \
+	u_gen_mipmap.c \
+	u_handle_table.c \
+	u_hash_table.c \
+	u_hash.c \
+	u_keymap.c \
+	u_linear.c \
+	u_math.c \
+	u_mm.c \
+	u_rect.c \
+	u_simple_shaders.c \
+	u_snprintf.c \
+	u_stream_stdc.c \
+	u_stream_wd.c \
+	u_tile.c \
+	u_time.c \
+	u_timed_winsys.c \
+	u_simple_screen.c
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript
new file mode 100644
index 0000000000..35f683fb8e
--- /dev/null
+++ b/src/gallium/auxiliary/util/SConscript
@@ -0,0 +1,30 @@
+Import('*')
+
+util = env.ConvenienceLibrary(
+	target = 'util',
+	source = [
+		'p_debug.c',
+		'p_debug_mem.c',
+		'p_debug_prof.c',
+		'u_blit.c',
+		'u_cache.c',
+		'u_draw_quad.c',
+		'u_gen_mipmap.c',
+		'u_handle_table.c',
+		'u_hash.c',
+		'u_hash_table.c',
+		'u_keymap.c',
+		'u_math.c',
+		'u_mm.c',
+		'u_rect.c',
+		'u_simple_shaders.c',
+		'u_snprintf.c',
+		'u_stream_stdc.c',
+		'u_stream_wd.c',
+		'u_tile.c',
+		'u_time.c',
+		'u_timed_winsys.c',
+		'u_simple_screen.c',
+	])
+
+auxiliaries.insert(0, util)
diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c
new file mode 100644
index 0000000000..f373f941dd
--- /dev/null
+++ b/src/gallium/auxiliary/util/p_debug.c
@@ -0,0 +1,782 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright (c) 2008 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "pipe/p_config.h" 
+
+#include <stdarg.h>
+
+
+#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY
+
+#include <windows.h>
+#include <winddi.h>
+
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+
+#include <stdio.h> 
+#include <stdlib.h> 
+#include <windows.h> 
+#include <types.h> 
+
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+
+#ifndef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN      // Exclude rarely-used stuff from Windows headers
+#endif
+#include <windows.h>
+
+#else
+
+#include <stdio.h>
+#include <stdlib.h>
+
+#endif
+
+#include "pipe/p_compiler.h" 
+#include "pipe/p_debug.h" 
+#include "pipe/p_format.h" 
+#include "pipe/p_state.h" 
+#include "pipe/p_inlines.h" 
+#include "util/u_memory.h" 
+#include "util/u_string.h" 
+#include "util/u_stream.h" 
+#include "util/u_math.h" 
+#include "util/u_tile.h" 
+
+
+#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY
+static INLINE void 
+_EngDebugPrint(const char *format, ...)
+{
+   va_list ap;
+   va_start(ap, format);
+   EngDebugPrint("", (PCHAR)format, ap);
+   va_end(ap);
+}
+#endif
+
+
+void _debug_vprintf(const char *format, va_list ap)
+{
+#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+   /* EngDebugPrint does not handle float point arguments, so we need to use
+    * our own vsnprintf implementation. It is also very slow, so buffer until
+    * we find a newline. */
+   static char buf[512] = {'\0'};
+   size_t len = strlen(buf);
+   int ret = util_vsnprintf(buf + len, sizeof(buf) - len, format, ap);
+   if(ret > (int)(sizeof(buf) - len - 1) || util_strchr(buf + len, '\n')) {
+      _EngDebugPrint("%s", buf);
+      buf[0] = '\0';
+   }
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   /* EngDebugPrint does not handle float point arguments, so we need to use
+    * our own vsnprintf implementation. It is also very slow, so buffer until
+    * we find a newline. */
+   static char buf[512 + 1] = {'\0'};
+   size_t len = strlen(buf);
+   int ret = util_vsnprintf(buf + len, sizeof(buf) - len, format, ap);
+   if(ret > (int)(sizeof(buf) - len - 1) || util_strchr(buf + len, '\n')) {
+      OutputDebugStringA(buf);
+      buf[0] = '\0';
+   }
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+   wchar_t *wide_format;
+   long wide_str_len;   
+   char buf[512];   
+   int ret;   
+#if (_WIN32_WCE < 600)
+   ret = vsprintf(buf, format, ap);   
+   if(ret < 0){   
+       sprintf(buf, "Cant handle debug print!");   
+       ret = 25;
+   }
+#else
+   ret = vsprintf_s(buf, 512, format, ap);   
+   if(ret < 0){   
+       sprintf_s(buf, 512, "Cant handle debug print!");   
+       ret = 25;
+   }
+#endif
+   buf[ret] = '\0';   
+   /* Format is ascii - needs to be converted to wchar_t for printing */   
+   wide_str_len = MultiByteToWideChar(CP_ACP, 0, (const char *) buf, -1, NULL, 0);   
+   wide_format = (wchar_t *) malloc((wide_str_len+1) * sizeof(wchar_t));   
+   if (wide_format) {   
+      MultiByteToWideChar(CP_ACP, 0, (const char *) buf, -1,   
+            wide_format, wide_str_len);   
+      NKDbgPrintfW(wide_format, wide_format);   
+      free(wide_format);   
+   } 
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+   /* TODO */
+#else /* !PIPE_SUBSYSTEM_WINDOWS */
+#ifdef DEBUG
+   vfprintf(stderr, format, ap);
+#endif
+#endif
+}
+
+
+#ifdef DEBUG
+void debug_print_blob( const char *name,
+                       const void *blob,
+                       unsigned size )
+{
+   const unsigned *ublob = (const unsigned *)blob;
+   unsigned i;
+
+   debug_printf("%s (%d dwords%s)\n", name, size/4,
+                size%4 ? "... plus a few bytes" : "");
+
+   for (i = 0; i < size/4; i++) {
+      debug_printf("%d:\t%08x\n", i, ublob[i]);
+   }
+}
+#endif
+
+
+void _debug_break(void) 
+{
+#if defined(PIPE_ARCH_X86) && defined(PIPE_CC_GCC)
+   __asm("int3");
+#elif defined(PIPE_ARCH_X86) && defined(PIPE_CC_MSVC)
+   _asm {int 3};
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+   EngDebugBreak();
+#else
+   abort();
+#endif
+}
+
+
+#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY
+static const char *
+find(const char *start, const char *end, char c) 
+{
+   const char *p;
+   for(p = start; !end || p != end; ++p) {
+      if(*p == c)
+	 return p;
+      if(*p < 32)
+	 break;
+   }
+   return NULL;
+}
+
+static int 
+compare(const char *start, const char *end, const char *s)
+{
+   const char *p, *q;
+   for(p = start, q = s; p != end && *q != '\0'; ++p, ++q) {
+      if(*p != *q)
+	 return 0;
+   }
+   return p == end && *q == '\0';
+}
+
+static void 
+copy(char *dst, const char *start, const char *end, size_t n) 
+{
+   const char *p;
+   char *q;
+   for(p = start, q = dst, n = n - 1; p != end && n; ++p, ++q, --n)
+      *q = *p;
+   *q = '\0';
+}
+#endif
+
+
+static INLINE const char *
+_debug_get_option(const char *name)
+{
+#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+   /* EngMapFile creates the file if it does not exists, so it must either be
+    * disabled on release versions (or put in a less conspicuous place). */
+#ifdef DEBUG
+   const char *result = NULL;
+   ULONG_PTR iFile = 0;
+   const void *pMap = NULL;
+   const char *sol, *eol, *sep;
+   static char output[1024];
+   
+   pMap = EngMapFile(L"\\??\\c:\\gallium.cfg", 0, &iFile);
+   if(pMap) {
+      sol = (const char *)pMap;
+      while(1) {
+	 /* TODO: handle LF line endings */
+	 eol = find(sol, NULL, '\r');
+	 if(!eol || eol == sol)
+	    break;
+	 sep = find(sol, eol, '=');
+	 if(!sep)
+	    break;
+	 if(compare(sol, sep, name)) {
+	    copy(output, sep + 1, eol, sizeof(output));
+	    result = output;
+	    break;
+	 }
+	 sol = eol + 2;
+      }
+      EngUnmapFile(iFile);
+   }
+   return result;
+#else
+   return NULL;
+#endif
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) || defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) 
+   /* TODO: implement */
+   return NULL;
+#else
+   return getenv(name);
+#endif
+}
+
+const char *
+debug_get_option(const char *name, const char *dfault)
+{
+   const char *result;
+
+   result = _debug_get_option(name);
+   if(!result)
+      result = dfault;
+      
+   debug_printf("%s: %s = %s\n", __FUNCTION__, name, result ? result : "(null)");
+   
+   return result;
+}
+
+boolean
+debug_get_bool_option(const char *name, boolean dfault)
+{
+   const char *str = _debug_get_option(name);
+   boolean result;
+   
+   if(str == NULL)
+      result = dfault;
+   else if(!util_strcmp(str, "n"))
+      result = FALSE;
+   else if(!util_strcmp(str, "no"))
+      result = FALSE;
+   else if(!util_strcmp(str, "0"))
+      result = FALSE;
+   else if(!util_strcmp(str, "f"))
+      result = FALSE;
+   else if(!util_strcmp(str, "false"))
+      result = FALSE;
+   else
+      result = TRUE;
+
+   debug_printf("%s: %s = %s\n", __FUNCTION__, name, result ? "TRUE" : "FALSE");
+   
+   return result;
+}
+
+
+long
+debug_get_num_option(const char *name, long dfault)
+{
+   long result;
+   const char *str;
+   
+   str = _debug_get_option(name);
+   if(!str)
+      result = dfault;
+   else {
+      long sign;
+      char c;
+      c = *str++;
+      if(c == '-') {
+	 sign = -1;
+	 c = *str++;
+      } 
+      else {
+	 sign = 1;
+      }
+      result = 0;
+      while('0' <= c && c <= '9') {
+	 result = result*10 + (c - '0');
+	 c = *str++;
+      }
+      result *= sign;
+   }
+   
+   debug_printf("%s: %s = %li\n", __FUNCTION__, name, result);
+
+   return result;
+}
+
+
+unsigned long
+debug_get_flags_option(const char *name, 
+                       const struct debug_named_value *flags,
+                       unsigned long dfault)
+{
+   unsigned long result;
+   const char *str;
+   
+   str = _debug_get_option(name);
+   if(!str)
+      result = dfault;
+   else if (!util_strcmp(str, "help")) {
+      result = dfault;
+      while (flags->name) {
+         debug_printf("%s: help for %s: %s [0x%lx]\n", __FUNCTION__, name, flags->name, flags->value);
+         flags++;
+      }
+   }
+   else {
+      result = 0;
+      while( flags->name ) {
+	 if (!util_strcmp(str, "all") || util_strstr(str, flags->name ))
+	    result |= flags->value;
+	 ++flags;
+      }
+   }
+
+   if (str) {
+      debug_printf("%s: %s = 0x%lx (%s)\n", __FUNCTION__, name, result, str);
+   }
+   else {
+      debug_printf("%s: %s = 0x%lx\n", __FUNCTION__, name, result);
+   }
+
+   return result;
+}
+
+
+void _debug_assert_fail(const char *expr, 
+                        const char *file, 
+                        unsigned line, 
+                        const char *function) 
+{
+   _debug_printf("%s:%u:%s: Assertion `%s' failed.\n", file, line, function, expr);
+#if defined(PIPE_OS_WINDOWS) && !defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   if (debug_get_bool_option("GALLIUM_ABORT_ON_ASSERT", FALSE))
+#else
+   if (debug_get_bool_option("GALLIUM_ABORT_ON_ASSERT", TRUE))
+#endif
+      debug_break();
+   else
+      _debug_printf("continuing...\n");
+}
+
+
+const char *
+debug_dump_enum(const struct debug_named_value *names, 
+                unsigned long value)
+{
+   static char rest[64];
+   
+   while(names->name) {
+      if(names->value == value)
+	 return names->name;
+      ++names;
+   }
+
+   util_snprintf(rest, sizeof(rest), "0x%08lx", value);
+   return rest;
+}
+
+
+const char *
+debug_dump_enum_noprefix(const struct debug_named_value *names, 
+                         const char *prefix,
+                         unsigned long value)
+{
+   static char rest[64];
+   
+   while(names->name) {
+      if(names->value == value) {
+         const char *name = names->name;
+         while (*name == *prefix) {
+            name++;
+            prefix++;
+         }
+         return name;
+      }
+      ++names;
+   }
+
+   
+
+   util_snprintf(rest, sizeof(rest), "0x%08lx", value);
+   return rest;
+}
+
+
+const char *
+debug_dump_flags(const struct debug_named_value *names, 
+                 unsigned long value)
+{
+   static char output[4096];
+   static char rest[256];
+   int first = 1;
+
+   output[0] = '\0';
+
+   while(names->name) {
+      if((names->value & value) == names->value) {
+	 if (!first)
+	    util_strncat(output, "|", sizeof(output));
+	 else
+	    first = 0;
+	 util_strncat(output, names->name, sizeof(output));
+	 value &= ~names->value;
+      }
+      ++names;
+   }
+   
+   if (value) {
+      if (!first)
+	 util_strncat(output, "|", sizeof(output));
+      else
+	 first = 0;
+      
+      util_snprintf(rest, sizeof(rest), "0x%08lx", value);
+      util_strncat(output, rest, sizeof(output));
+   }
+   
+   if(first)
+      return "0";
+   
+   return output;
+}
+
+
+static const struct debug_named_value pipe_format_names[] = {
+#ifdef DEBUG
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_NONE),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A8R8G8B8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_X8R8G8B8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_B8G8R8A8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_B8G8R8X8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A1R5G5B5_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A4R4G4B4_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R5G6B5_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A2B10G10R10_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_L8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_I8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A8L8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_L16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_YCBCR),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_YCBCR_REV),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_Z16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_Z32_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_Z32_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_S8Z24_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_Z24S8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_X8Z24_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_Z24X8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_S8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R64_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R64G64_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R64G64B64_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R64G64B64A64_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_FLOAT),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_UNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_USCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_B6G5R5_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A8B8G8R8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_X8B8G8R8_SNORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SSCALED),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_L8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A8L8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_A8R8G8B8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_X8R8G8B8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_B8G8R8A8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_B8G8R8X8_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_X8UB8UG8SR8S_NORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_B6UG5SR5S_NORM),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_RGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_RGBA),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT3_RGBA),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT5_RGBA),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_SRGB),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_SRGBA),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT3_SRGBA),
+   DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT5_SRGBA),
+#endif
+   DEBUG_NAMED_VALUE_END
+};
+
+#ifdef DEBUG
+void debug_print_format(const char *msg, unsigned fmt )
+{
+   debug_printf("%s: %s\n", msg, debug_dump_enum(pipe_format_names, fmt)); 
+}
+#endif
+
+const char *pf_name( enum pipe_format format )
+{
+   return debug_dump_enum(pipe_format_names, format);
+}
+
+
+#ifdef DEBUG
+void debug_dump_image(const char *prefix,
+                      unsigned format, unsigned cpp,
+                      unsigned width, unsigned height,
+                      unsigned stride,
+                      const void *data)     
+{
+#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY
+   static unsigned no = 0; 
+   char filename[256];
+   WCHAR wfilename[sizeof(filename)];
+   ULONG_PTR iFile = 0;
+   struct {
+      unsigned format;
+      unsigned cpp;
+      unsigned width;
+      unsigned height;
+   } header;
+   unsigned char *pMap = NULL;
+   unsigned i;
+
+   util_snprintf(filename, sizeof(filename), "\\??\\c:\\%03u%s.raw", ++no, prefix);
+   for(i = 0; i < sizeof(filename); ++i)
+      wfilename[i] = (WCHAR)filename[i];
+   
+   pMap = (unsigned char *)EngMapFile(wfilename, sizeof(header) + height*width*cpp, &iFile);
+   if(!pMap)
+      return;
+   
+   header.format = format;
+   header.cpp = cpp;
+   header.width = width;
+   header.height = height;
+   memcpy(pMap, &header, sizeof(header));
+   pMap += sizeof(header);
+   
+   for(i = 0; i < height; ++i) {
+      memcpy(pMap, (unsigned char *)data + stride*i, cpp*width);
+      pMap += cpp*width;
+   }
+      
+   EngUnmapFile(iFile);
+#endif
+}
+
+void debug_dump_surface(const char *prefix,
+                        struct pipe_surface *surface)     
+{
+   unsigned surface_usage;
+   void *data;
+
+   if (!surface)
+      goto error1;
+
+   /* XXX: force mappable surface */
+   surface_usage = surface->usage;
+   surface->usage |= PIPE_BUFFER_USAGE_CPU_READ;
+   
+   data = pipe_surface_map(surface, 
+                           PIPE_BUFFER_USAGE_CPU_READ);
+   if(!data)
+      goto error2;
+   
+   debug_dump_image(prefix, 
+                    surface->format,
+                    surface->block.size, 
+                    surface->nblocksx,
+                    surface->nblocksy,
+                    surface->stride,
+                    data);
+   
+   pipe_surface_unmap(surface);
+error2:
+   surface->usage = surface_usage;
+error1:
+   ;
+}
+
+
+#pragma pack(push,2)
+struct bmp_file_header {
+   uint16_t bfType;
+   uint32_t bfSize;
+   uint16_t bfReserved1;
+   uint16_t bfReserved2;
+   uint32_t bfOffBits;
+};
+#pragma pack(pop)
+
+struct bmp_info_header {
+   uint32_t biSize;
+   int32_t biWidth;
+   int32_t biHeight;
+   uint16_t biPlanes;
+   uint16_t biBitCount;
+   uint32_t biCompression;
+   uint32_t biSizeImage;
+   int32_t biXPelsPerMeter;
+   int32_t biYPelsPerMeter;
+   uint32_t biClrUsed;
+   uint32_t biClrImportant;
+};
+
+struct bmp_rgb_quad {
+   uint8_t rgbBlue;
+   uint8_t rgbGreen;
+   uint8_t rgbRed;
+   uint8_t rgbAlpha;
+};
+
+void 
+debug_dump_surface_bmp(const char *filename,
+                       struct pipe_surface *surface)
+{
+#ifndef PIPE_SUBSYSTEM_WINDOWS_MINIPORT
+   struct util_stream *stream;
+   unsigned surface_usage;
+   struct bmp_file_header bmfh;
+   struct bmp_info_header bmih;
+   float *rgba;
+   unsigned x, y;
+
+   if (!surface)
+      goto error1;
+
+   rgba = MALLOC(surface->width*4*sizeof(float));
+   if(!rgba)
+      goto error1;
+   
+   bmfh.bfType = 0x4d42;
+   bmfh.bfSize = 14 + 40 + surface->height*surface->width*4;
+   bmfh.bfReserved1 = 0;
+   bmfh.bfReserved2 = 0;
+   bmfh.bfOffBits = 14 + 40;
+   
+   bmih.biSize = 40;
+   bmih.biWidth = surface->width;
+   bmih.biHeight = surface->height;
+   bmih.biPlanes = 1;
+   bmih.biBitCount = 32;
+   bmih.biCompression = 0;
+   bmih.biSizeImage = surface->height*surface->width*4;
+   bmih.biXPelsPerMeter = 0;
+   bmih.biYPelsPerMeter = 0;
+   bmih.biClrUsed = 0;
+   bmih.biClrImportant = 0;
+   
+   stream = util_stream_create(filename, bmfh.bfSize);
+   if(!stream)
+      goto error2;
+   
+   util_stream_write(stream, &bmfh, 14);
+   util_stream_write(stream, &bmih, 40);
+   
+   /* XXX: force mappable surface */
+   surface_usage = surface->usage;
+   surface->usage |= PIPE_BUFFER_USAGE_CPU_READ;
+
+   y = surface->height;
+   while(y--) {
+      pipe_get_tile_rgba(surface,
+                         0, y, surface->width, 1,
+                         rgba);
+      for(x = 0; x < surface->width; ++x)
+      {
+         struct bmp_rgb_quad pixel;
+         pixel.rgbRed   = float_to_ubyte(rgba[x*4 + 0]);
+         pixel.rgbGreen = float_to_ubyte(rgba[x*4 + 1]);
+         pixel.rgbBlue  = float_to_ubyte(rgba[x*4 + 2]);
+         pixel.rgbAlpha = float_to_ubyte(rgba[x*4 + 3]);
+         util_stream_write(stream, &pixel, 4);
+      }  
+   }
+   
+   surface->usage = surface_usage;
+
+   util_stream_close(stream);
+error2:
+   FREE(rgba);
+error1:
+   ;
+#endif
+}
+
+#endif
diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c
new file mode 100644
index 0000000000..250fd60f63
--- /dev/null
+++ b/src/gallium/auxiliary/util/p_debug_mem.c
@@ -0,0 +1,311 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Memory debugging.
+ * 
+ * @author José Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#include "pipe/p_config.h" 
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+#include <windows.h>
+#include <winddi.h>
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+#include <wdm.h>
+#else
+#include <stdio.h>
+#include <stdlib.h>
+#endif
+
+#include "pipe/p_debug.h" 
+#include "util/u_double_list.h" 
+
+
+#define DEBUG_MEMORY_MAGIC 0x6e34090aU 
+
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && !defined(WINCE)
+#define real_malloc(_size) EngAllocMem(0, _size, 'D3AG')
+#define real_free(_ptr) EngFreeMem(_ptr)
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+#define real_malloc(_size) ExAllocatePool(0, _size)
+#define real_free(_ptr) ExFreePool(_ptr)
+#else
+#define real_malloc(_size) malloc(_size)
+#define real_free(_ptr) free(_ptr)
+#endif
+
+
+struct debug_memory_header 
+{
+   struct list_head head;
+   
+   unsigned long no;
+   const char *file;
+   unsigned line;
+   const char *function;
+   size_t size;
+   unsigned magic;
+};
+
+struct debug_memory_footer
+{
+   unsigned magic;
+};
+
+
+static struct list_head list = { &list, &list };
+
+static unsigned long last_no = 0;
+
+
+static INLINE struct debug_memory_header *
+header_from_data(void *data)
+{
+   if(data)
+      return (struct debug_memory_header *)((char *)data - sizeof(struct debug_memory_header));
+   else
+      return NULL;
+}
+
+static INLINE void *
+data_from_header(struct debug_memory_header *hdr)
+{
+   if(hdr)
+      return (void *)((char *)hdr + sizeof(struct debug_memory_header));
+   else
+      return NULL;
+}
+
+static INLINE struct debug_memory_footer *
+footer_from_header(struct debug_memory_header *hdr)
+{
+   if(hdr)
+      return (struct debug_memory_footer *)((char *)hdr + sizeof(struct debug_memory_header) + hdr->size);
+   else
+      return NULL;
+}
+
+
+void *
+debug_malloc(const char *file, unsigned line, const char *function,
+             size_t size) 
+{
+   struct debug_memory_header *hdr;
+   struct debug_memory_footer *ftr;
+   
+   hdr = real_malloc(sizeof(*hdr) + size + sizeof(*ftr));
+   if(!hdr) {
+      debug_printf("%s:%u:%s: out of memory when trying to allocate %lu bytes\n",
+                   file, line, function,
+                   (long unsigned)size);
+      return NULL;
+   }
+ 
+   hdr->no = last_no++;
+   hdr->file = file;
+   hdr->line = line;
+   hdr->function = function;
+   hdr->size = size;
+   hdr->magic = DEBUG_MEMORY_MAGIC;
+
+   ftr = footer_from_header(hdr);
+   ftr->magic = DEBUG_MEMORY_MAGIC;
+   
+   LIST_ADDTAIL(&hdr->head, &list);
+   
+   return data_from_header(hdr);
+}
+
+void
+debug_free(const char *file, unsigned line, const char *function,
+           void *ptr) 
+{
+   struct debug_memory_header *hdr;
+   struct debug_memory_footer *ftr;
+   
+   if(!ptr)
+      return;
+   
+   hdr = header_from_data(ptr);
+   if(hdr->magic != DEBUG_MEMORY_MAGIC) {
+      debug_printf("%s:%u:%s: freeing bad or corrupted memory %p\n",
+                   file, line, function,
+                   ptr);
+      debug_assert(0);
+      return;
+   }
+
+   ftr = footer_from_header(hdr);
+   if(ftr->magic != DEBUG_MEMORY_MAGIC) {
+      debug_printf("%s:%u:%s: buffer overflow %p\n",
+                   hdr->file, hdr->line, hdr->function,
+                   ptr);
+      debug_assert(0);
+   }
+
+   LIST_DEL(&hdr->head);
+   hdr->magic = 0;
+   ftr->magic = 0;
+   
+   real_free(hdr);
+}
+
+void *
+debug_calloc(const char *file, unsigned line, const char *function,
+             size_t count, size_t size )
+{
+   void *ptr = debug_malloc( file, line, function, count * size );
+   if( ptr )
+      memset( ptr, 0, count * size );
+   return ptr;
+}
+
+void *
+debug_realloc(const char *file, unsigned line, const char *function,
+              void *old_ptr, size_t old_size, size_t new_size )
+{
+   struct debug_memory_header *old_hdr, *new_hdr;
+   struct debug_memory_footer *old_ftr, *new_ftr;
+   void *new_ptr;
+   
+   if(!old_ptr)
+      return debug_malloc( file, line, function, new_size );
+   
+   if(!new_size) {
+      debug_free( file, line, function, old_ptr );
+      return NULL;
+   }
+   
+   old_hdr = header_from_data(old_ptr);
+   if(old_hdr->magic != DEBUG_MEMORY_MAGIC) {
+      debug_printf("%s:%u:%s: reallocating bad or corrupted memory %p\n",
+                   file, line, function,
+                   old_ptr);
+      debug_assert(0);
+      return NULL;
+   }
+   
+   old_ftr = footer_from_header(old_hdr);
+   if(old_ftr->magic != DEBUG_MEMORY_MAGIC) {
+      debug_printf("%s:%u:%s: buffer overflow %p\n",
+                   old_hdr->file, old_hdr->line, old_hdr->function,
+                   old_ptr);
+      debug_assert(0);
+   }
+
+   /* alloc new */
+   new_hdr = real_malloc(sizeof(*new_hdr) + new_size + sizeof(*new_ftr));
+   if(!new_hdr) {
+      debug_printf("%s:%u:%s: out of memory when trying to allocate %lu bytes\n",
+                   file, line, function,
+                   (long unsigned)new_size);
+      return NULL;
+   }
+   new_hdr->no = old_hdr->no;
+   new_hdr->file = old_hdr->file;
+   new_hdr->line = old_hdr->line;
+   new_hdr->function = old_hdr->function;
+   new_hdr->size = new_size;
+   new_hdr->magic = DEBUG_MEMORY_MAGIC;
+   
+   new_ftr = footer_from_header(new_hdr);
+   new_ftr->magic = DEBUG_MEMORY_MAGIC;
+   
+   LIST_REPLACE(&old_hdr->head, &new_hdr->head);
+
+   /* copy data */
+   new_ptr = data_from_header(new_hdr);
+   memcpy( new_ptr, old_ptr, old_size < new_size ? old_size : new_size );
+
+   /* free old */
+   old_hdr->magic = 0;
+   old_ftr->magic = 0;
+   real_free(old_hdr);
+
+   return new_ptr;
+}
+
+unsigned long
+debug_memory_begin(void)
+{
+   return last_no;
+}
+
+void 
+debug_memory_end(unsigned long start_no)
+{
+   size_t total_size = 0;
+   struct list_head *entry;
+
+   if(start_no == last_no)
+      return;
+
+   entry = list.prev;
+   for (; entry != &list; entry = entry->prev) {
+      struct debug_memory_header *hdr;
+      void *ptr;
+      struct debug_memory_footer *ftr;
+
+      hdr = LIST_ENTRY(struct debug_memory_header, entry, head);
+      ptr = data_from_header(hdr);
+      ftr = footer_from_header(hdr);
+
+      if(hdr->magic != DEBUG_MEMORY_MAGIC) {
+         debug_printf("%s:%u:%s: bad or corrupted memory %p\n",
+                      hdr->file, hdr->line, hdr->function,
+                      ptr);
+         debug_assert(0);
+      }
+
+      if((start_no <= hdr->no && hdr->no < last_no) ||
+	 (last_no < start_no && (hdr->no < last_no || start_no <= hdr->no))) {
+	 debug_printf("%s:%u:%s: %u bytes at %p not freed\n",
+		      hdr->file, hdr->line, hdr->function,
+		      hdr->size, ptr);
+	 total_size += hdr->size;
+      }
+
+      if(ftr->magic != DEBUG_MEMORY_MAGIC) {
+         debug_printf("%s:%u:%s: buffer overflow %p\n",
+                      hdr->file, hdr->line, hdr->function,
+                      ptr);
+         debug_assert(0);
+      }
+   }
+
+   if(total_size) {
+      debug_printf("Total of %u KB of system memory apparently leaked\n",
+		   (total_size + 1023)/1024);
+   }
+   else {
+      debug_printf("No memory leaks detected.\n");
+   }
+}
diff --git a/src/gallium/auxiliary/util/p_debug_prof.c b/src/gallium/auxiliary/util/p_debug_prof.c
new file mode 100644
index 0000000000..5f9772ef91
--- /dev/null
+++ b/src/gallium/auxiliary/util/p_debug_prof.c
@@ -0,0 +1,320 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Poor-man profiling.
+ * 
+ * @author José Fonseca <jrfonseca@tungstengraphics.com>
+ * 
+ * @sa http://blogs.msdn.com/joshpoley/archive/2008/03/12/poor-man-s-profiler.aspx
+ * @sa http://www.johnpanzer.com/aci_cuj/index.html
+ */
+
+#include "pipe/p_config.h" 
+
+#if defined(PROFILE) && defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+
+#include <windows.h>
+#include <winddi.h>
+
+#include "pipe/p_debug.h" 
+#include "util/u_string.h" 
+
+
+#define PROFILE_TABLE_SIZE (1024*1024)
+#define FILE_NAME_SIZE 256
+
+struct debug_profile_entry
+{
+   uintptr_t caller;
+   uintptr_t callee;
+   uint64_t samples;
+};
+
+static unsigned long enabled = 0;
+
+static WCHAR wFileName[FILE_NAME_SIZE] = L"\\??\\c:\\00000000.prof";
+static ULONG_PTR iFile = 0;
+
+static struct debug_profile_entry *table = NULL;
+static unsigned long free_table_entries = 0;
+static unsigned long max_table_entries = 0;
+
+uint64_t start_stamp = 0;
+uint64_t end_stamp = 0;
+
+
+static void
+debug_profile_entry(uintptr_t caller, uintptr_t callee, uint64_t samples)
+{
+   unsigned hash = ( caller + callee ) & PROFILE_TABLE_SIZE - 1;
+   
+   while(1) {
+      if(table[hash].caller == 0 && table[hash].callee == 0) {
+         table[hash].caller = caller;
+         table[hash].callee = callee;
+         table[hash].samples = samples;
+         --free_table_entries;
+         break;
+      }
+      else if(table[hash].caller == caller && table[hash].callee == callee) {
+         table[hash].samples += samples;
+         break;
+      }
+      else {
+         ++hash;
+      }
+   }
+}
+
+
+static uintptr_t caller_stack[1024];
+static unsigned last_caller = 0;
+
+
+static int64_t delta(void) {
+   int64_t result = end_stamp - start_stamp;
+   if(result > UINT64_C(0xffffffff))
+      result = 0;
+   return result;
+}
+
+
+static void __cdecl 
+debug_profile_enter(uintptr_t callee)
+{
+   uintptr_t caller = last_caller ? caller_stack[last_caller - 1] : 0;
+                
+   if (caller)
+      debug_profile_entry(caller, 0, delta());
+   debug_profile_entry(caller, callee, 1);
+   caller_stack[last_caller++] = callee;
+}
+
+
+static void __cdecl
+debug_profile_exit(uintptr_t callee)
+{
+   debug_profile_entry(callee, 0, delta());
+   if(last_caller)
+      --last_caller;
+}
+   
+   
+/**
+ * Called at the start of every method or function.
+ * 
+ * @sa http://msdn.microsoft.com/en-us/library/c63a9b7h.aspx
+ */
+void __declspec(naked) __cdecl 
+_penter(void) {
+   _asm {
+      push eax
+      mov eax, [enabled]
+      test eax, eax
+      jz skip
+
+      push edx
+      
+      rdtsc
+      mov dword ptr [end_stamp], eax
+      mov dword ptr [end_stamp+4], edx
+
+      xor eax, eax
+      mov [enabled], eax
+
+      mov eax, [esp+8]
+
+      push ebx
+      push ecx
+      push ebp
+      push edi
+      push esi
+
+      push eax
+      call debug_profile_enter
+      add esp, 4
+
+      pop esi
+      pop edi
+      pop ebp
+      pop ecx
+      pop ebx
+
+      mov eax, 1
+      mov [enabled], eax 
+
+      rdtsc
+      mov dword ptr [start_stamp], eax
+      mov dword ptr [start_stamp+4], edx
+      
+      pop edx
+skip:
+      pop eax
+      ret
+   }
+}
+
+
+/**
+ * Called at the end of Calls the end of every method or function.
+ * 
+ * @sa http://msdn.microsoft.com/en-us/library/xc11y76y.aspx
+ */
+void __declspec(naked) __cdecl 
+_pexit(void) {
+   _asm {
+      push eax
+      mov eax, [enabled]
+      test eax, eax
+      jz skip
+
+      push edx
+      
+      rdtsc
+      mov dword ptr [end_stamp], eax
+      mov dword ptr [end_stamp+4], edx
+
+      xor eax, eax
+      mov [enabled], eax
+
+      mov eax, [esp+8]
+
+      push ebx
+      push ecx
+      push ebp
+      push edi
+      push esi
+
+      push eax
+      call debug_profile_exit
+      add esp, 4
+
+      pop esi
+      pop edi
+      pop ebp
+      pop ecx
+      pop ebx
+
+      mov eax, 1
+      mov [enabled], eax 
+
+      rdtsc
+      mov dword ptr [start_stamp], eax
+      mov dword ptr [start_stamp+4], edx
+      
+      pop edx
+skip:
+      pop eax
+      ret
+   }
+}
+
+
+/**
+ * Reference function for calibration. 
+ */
+void __declspec(naked) 
+__debug_profile_reference(void) {
+   _asm {
+      call _penter
+      call _pexit
+      ret
+   }
+}
+
+
+void
+debug_profile_start(void)
+{
+   WCHAR *p;
+
+   // increment starting from the less significant digit
+   p = &wFileName[14];
+   while(1) {
+      if(*p == '9') {
+         *p-- = '0';
+      }
+      else {
+         *p += 1;
+         break;
+      }
+   }
+
+   table = EngMapFile(wFileName, 
+                      PROFILE_TABLE_SIZE*sizeof(struct debug_profile_entry), 
+                      &iFile);
+   if(table) {
+      unsigned i;
+      
+      free_table_entries = max_table_entries = PROFILE_TABLE_SIZE;
+      memset(table, 0, PROFILE_TABLE_SIZE*sizeof(struct debug_profile_entry));
+      
+      table[0].caller = (uintptr_t)&__debug_profile_reference;
+      table[0].callee = 0;
+      table[0].samples = 0;
+      --free_table_entries;
+
+      _asm {
+         push edx
+         push eax
+      
+         rdtsc
+         mov dword ptr [start_stamp], eax
+         mov dword ptr [start_stamp+4], edx
+         
+         pop edx
+         pop eax
+      }
+
+      last_caller = 0;
+      
+      enabled = 1;
+
+      for(i = 0; i < 8; ++i) {
+         _asm {
+            call __debug_profile_reference
+         }
+      }
+   }
+}
+
+
+void 
+debug_profile_stop(void)
+{
+   enabled = 0;
+
+   if(iFile)
+      EngUnmapFile(iFile);
+   iFile = 0;
+   table = NULL;
+   free_table_entries = max_table_entries = 0;
+}
+
+#endif /* PROFILE */
diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c
new file mode 100644
index 0000000000..841e9c01e7
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_blit.c
@@ -0,0 +1,555 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Copy/blit pixel rect between surfaces
+ *  
+ * @author Brian Paul
+ */
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_blit.h"
+#include "util/u_draw_quad.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "util/u_simple_shaders.h"
+
+#include "cso_cache/cso_context.h"
+
+
+struct blit_state
+{
+   struct pipe_context *pipe;
+   struct cso_context *cso;
+
+   struct pipe_blend_state blend;
+   struct pipe_depth_stencil_alpha_state depthstencil;
+   struct pipe_rasterizer_state rasterizer;
+   struct pipe_sampler_state sampler;
+   struct pipe_viewport_state viewport;
+
+   struct pipe_shader_state vert_shader;
+   struct pipe_shader_state frag_shader;
+   void *vs;
+   void *fs;
+
+   struct pipe_buffer *vbuf;  /**< quad vertices */
+   unsigned vbuf_slot;
+
+   float vertices[4][2][4];   /**< vertex/texcoords for quad */
+};
+
+
+/**
+ * Create state object for blit.
+ * Intended to be created once and re-used for many blit() calls.
+ */
+struct blit_state *
+util_create_blit(struct pipe_context *pipe, struct cso_context *cso)
+{
+   struct blit_state *ctx;
+   uint i;
+
+   ctx = CALLOC_STRUCT(blit_state);
+   if (!ctx)
+      return NULL;
+
+   ctx->pipe = pipe;
+   ctx->cso = cso;
+
+   /* disabled blending/masking */
+   memset(&ctx->blend, 0, sizeof(ctx->blend));
+   ctx->blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+   ctx->blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+   ctx->blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   ctx->blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   ctx->blend.colormask = PIPE_MASK_RGBA;
+
+   /* no-op depth/stencil/alpha */
+   memset(&ctx->depthstencil, 0, sizeof(ctx->depthstencil));
+
+   /* rasterizer */
+   memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer));
+   ctx->rasterizer.front_winding = PIPE_WINDING_CW;
+   ctx->rasterizer.cull_mode = PIPE_WINDING_NONE;
+   ctx->rasterizer.bypass_clipping = 1;
+   /*ctx->rasterizer.bypass_vs = 1;*/
+   ctx->rasterizer.gl_rasterization_rules = 1;
+
+   /* samplers */
+   memset(&ctx->sampler, 0, sizeof(ctx->sampler));
+   ctx->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   ctx->sampler.min_img_filter = 0; /* set later */
+   ctx->sampler.mag_img_filter = 0; /* set later */
+   ctx->sampler.normalized_coords = 1;
+
+   /* viewport (identity, we setup vertices in wincoords) */
+   ctx->viewport.scale[0] = 1.0;
+   ctx->viewport.scale[1] = 1.0;
+   ctx->viewport.scale[2] = 1.0;
+   ctx->viewport.scale[3] = 1.0;
+   ctx->viewport.translate[0] = 0.0;
+   ctx->viewport.translate[1] = 0.0;
+   ctx->viewport.translate[2] = 0.0;
+   ctx->viewport.translate[3] = 0.0;
+
+   /* vertex shader */
+   {
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_GENERIC };
+      const uint semantic_indexes[] = { 0, 0 };
+      ctx->vs = util_make_vertex_passthrough_shader(pipe, 2, semantic_names,
+                                                    semantic_indexes,
+                                                    &ctx->vert_shader);
+   }
+
+   /* fragment shader */
+   ctx->fs = util_make_fragment_tex_shader(pipe, &ctx->frag_shader);
+   ctx->vbuf = NULL;
+
+   /* init vertex data that doesn't change */
+   for (i = 0; i < 4; i++) {
+      ctx->vertices[i][0][3] = 1.0f; /* w */
+      ctx->vertices[i][1][2] = 0.0f; /* r */
+      ctx->vertices[i][1][3] = 1.0f; /* q */
+   }
+
+   return ctx;
+}
+
+
+/**
+ * Destroy a blit context
+ */
+void
+util_destroy_blit(struct blit_state *ctx)
+{
+   struct pipe_context *pipe = ctx->pipe;
+
+   pipe->delete_vs_state(pipe, ctx->vs);
+   pipe->delete_fs_state(pipe, ctx->fs);
+
+   FREE((void*) ctx->vert_shader.tokens);
+   FREE((void*) ctx->frag_shader.tokens);
+
+   pipe_buffer_reference(pipe->screen, &ctx->vbuf, NULL);
+
+   FREE(ctx);
+}
+
+
+static unsigned get_next_slot( struct blit_state *ctx )
+{
+   const unsigned max_slots = 4096 / sizeof ctx->vertices;
+
+   if (ctx->vbuf_slot >= max_slots) 
+      util_blit_flush( ctx );
+
+   if (!ctx->vbuf) {
+      ctx->vbuf = pipe_buffer_create(ctx->pipe->screen,
+                                     32,
+                                     PIPE_BUFFER_USAGE_VERTEX,
+                                     max_slots * sizeof ctx->vertices);
+   }
+   
+   return ctx->vbuf_slot++ * sizeof ctx->vertices;
+}
+                               
+
+
+/**
+ * Setup vertex data for the textured quad we'll draw.
+ * Note: y=0=top
+ */
+static unsigned
+setup_vertex_data(struct blit_state *ctx,
+                  float x0, float y0, float x1, float y1, float z)
+{
+   void *buf;
+   unsigned offset;
+
+   ctx->vertices[0][0][0] = x0;
+   ctx->vertices[0][0][1] = y0;
+   ctx->vertices[0][0][2] = z;
+   ctx->vertices[0][1][0] = 0.0f; /*s*/
+   ctx->vertices[0][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[1][0][0] = x1;
+   ctx->vertices[1][0][1] = y0;
+   ctx->vertices[1][0][2] = z;
+   ctx->vertices[1][1][0] = 1.0f; /*s*/
+   ctx->vertices[1][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[2][0][0] = x1;
+   ctx->vertices[2][0][1] = y1;
+   ctx->vertices[2][0][2] = z;
+   ctx->vertices[2][1][0] = 1.0f;
+   ctx->vertices[2][1][1] = 1.0f;
+
+   ctx->vertices[3][0][0] = x0;
+   ctx->vertices[3][0][1] = y1;
+   ctx->vertices[3][0][2] = z;
+   ctx->vertices[3][1][0] = 0.0f;
+   ctx->vertices[3][1][1] = 1.0f;
+
+   offset = get_next_slot( ctx );
+
+   buf = pipe_buffer_map(ctx->pipe->screen, ctx->vbuf,
+                         PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   memcpy((char *)buf + offset, ctx->vertices, sizeof(ctx->vertices));
+
+   pipe_buffer_unmap(ctx->pipe->screen, ctx->vbuf);
+
+   return offset;
+}
+
+
+/**
+ * Setup vertex data for the textured quad we'll draw.
+ * Note: y=0=top
+ */
+static unsigned
+setup_vertex_data_tex(struct blit_state *ctx,
+                      float x0, float y0, float x1, float y1,
+                      float s0, float t0, float s1, float t1,
+                      float z)
+{
+   void *buf;
+   unsigned offset;
+
+   ctx->vertices[0][0][0] = x0;
+   ctx->vertices[0][0][1] = y0;
+   ctx->vertices[0][0][2] = z;
+   ctx->vertices[0][1][0] = s0; /*s*/
+   ctx->vertices[0][1][1] = t0; /*t*/
+
+   ctx->vertices[1][0][0] = x1;
+   ctx->vertices[1][0][1] = y0;
+   ctx->vertices[1][0][2] = z;
+   ctx->vertices[1][1][0] = s1; /*s*/
+   ctx->vertices[1][1][1] = t0; /*t*/
+
+   ctx->vertices[2][0][0] = x1;
+   ctx->vertices[2][0][1] = y1;
+   ctx->vertices[2][0][2] = z;
+   ctx->vertices[2][1][0] = s1;
+   ctx->vertices[2][1][1] = t1;
+
+   ctx->vertices[3][0][0] = x0;
+   ctx->vertices[3][0][1] = y1;
+   ctx->vertices[3][0][2] = z;
+   ctx->vertices[3][1][0] = s0;
+   ctx->vertices[3][1][1] = t1;
+
+   offset = get_next_slot( ctx );
+
+   buf = pipe_buffer_map(ctx->pipe->screen, ctx->vbuf,
+                         PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   memcpy((char *)buf + offset, ctx->vertices, sizeof(ctx->vertices));
+
+   pipe_buffer_unmap(ctx->pipe->screen, ctx->vbuf);
+
+   return offset;
+}
+/**
+ * Copy pixel block from src surface to dst surface.
+ * Overlapping regions are acceptable.
+ * XXX need some control over blitting Z and/or stencil.
+ */
+void
+util_blit_pixels(struct blit_state *ctx,
+                 struct pipe_surface *src,
+                 int srcX0, int srcY0,
+                 int srcX1, int srcY1,
+                 struct pipe_surface *dst,
+                 int dstX0, int dstY0,
+                 int dstX1, int dstY1,
+                 float z, uint filter)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture texTemp, *tex;
+   struct pipe_surface *texSurf;
+   struct pipe_framebuffer_state fb;
+   const int srcW = abs(srcX1 - srcX0);
+   const int srcH = abs(srcY1 - srcY0);
+   const int srcLeft = MIN2(srcX0, srcX1);
+   const int srcTop = MIN2(srcY0, srcY1);
+   unsigned offset;
+
+   assert(filter == PIPE_TEX_MIPFILTER_NEAREST ||
+          filter == PIPE_TEX_MIPFILTER_LINEAR);
+
+   if (srcLeft != srcX0) {
+      /* left-right flip */
+      int tmp = dstX0;
+      dstX0 = dstX1;
+      dstX1 = tmp;
+   }
+
+   if (srcTop != srcY0) {
+      /* up-down flip */
+      int tmp = dstY0;
+      dstY0 = dstY1;
+      dstY1 = tmp;
+   }
+
+   assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_SAMPLER, 0));
+   assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_SAMPLER, 0));
+
+   if(dst->format == src->format && (dstX1 - dstX0) == srcW && (dstY1 - dstY0) == srcH) {
+      /* FIXME: this will most surely fail for overlapping rectangles */
+      pipe->surface_copy(pipe, FALSE,
+			 dst, dstX0, dstY0,   /* dest */
+			 src, srcX0, srcY0, /* src */
+			 srcW, srcH);     /* size */
+      return;
+   }
+   
+   assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_RENDER_TARGET, 0));
+
+   /*
+    * XXX for now we're always creating a temporary texture.
+    * Strictly speaking that's not always needed.
+    */
+
+   /* create temp texture */
+   memset(&texTemp, 0, sizeof(texTemp));
+   texTemp.target = PIPE_TEXTURE_2D;
+   texTemp.format = src->format;
+   texTemp.last_level = 0;
+   texTemp.width[0] = srcW;
+   texTemp.height[0] = srcH;
+   texTemp.depth[0] = 1;
+   texTemp.compressed = 0;
+   pf_get_block(src->format, &texTemp.block);
+
+   tex = screen->texture_create(screen, &texTemp);
+   if (!tex)
+      return;
+
+   texSurf = screen->get_tex_surface(screen, tex, 0, 0, 0, 
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   /* load temp texture */
+   pipe->surface_copy(pipe, FALSE,
+                      texSurf, 0, 0,   /* dest */
+                      src, srcLeft, srcTop, /* src */
+                      srcW, srcH);     /* size */
+
+   /* free the surface, update the texture if necessary.
+    */
+   screen->tex_surface_release(screen, &texSurf);
+
+   /* save state (restored below) */
+   cso_save_blend(ctx->cso);
+   cso_save_depth_stencil_alpha(ctx->cso);
+   cso_save_rasterizer(ctx->cso);
+   cso_save_samplers(ctx->cso);
+   cso_save_sampler_textures(ctx->cso);
+   cso_save_framebuffer(ctx->cso);
+   cso_save_fragment_shader(ctx->cso);
+   cso_save_vertex_shader(ctx->cso);
+   cso_save_viewport(ctx->cso);
+
+   /* set misc state we care about */
+   cso_set_blend(ctx->cso, &ctx->blend);
+   cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil);
+   cso_set_rasterizer(ctx->cso, &ctx->rasterizer);
+   cso_set_viewport(ctx->cso, &ctx->viewport);
+
+   /* sampler */
+   ctx->sampler.min_img_filter = filter;
+   ctx->sampler.mag_img_filter = filter;
+   cso_single_sampler(ctx->cso, 0, &ctx->sampler);
+   cso_single_sampler_done(ctx->cso);
+
+   /* texture */
+   cso_set_sampler_textures(ctx->cso, 1, &tex);
+
+   /* shaders */
+   cso_set_fragment_shader_handle(ctx->cso, ctx->fs);
+   cso_set_vertex_shader_handle(ctx->cso, ctx->vs);
+
+   /* drawing dest */
+   memset(&fb, 0, sizeof(fb));
+   fb.width = dst->width;
+   fb.height = dst->height;
+   fb.nr_cbufs = 1;
+   fb.cbufs[0] = dst;
+   cso_set_framebuffer(ctx->cso, &fb);
+
+   /* draw quad */
+   offset = setup_vertex_data(ctx,
+                              (float) dstX0, (float) dstY0, 
+                              (float) dstX1, (float) dstY1, z);
+
+   util_draw_vertex_buffer(ctx->pipe, ctx->vbuf, offset,
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           2); /* attribs/vert */
+
+   /* restore state we changed */
+   cso_restore_blend(ctx->cso);
+   cso_restore_depth_stencil_alpha(ctx->cso);
+   cso_restore_rasterizer(ctx->cso);
+   cso_restore_samplers(ctx->cso);
+   cso_restore_sampler_textures(ctx->cso);
+   cso_restore_framebuffer(ctx->cso);
+   cso_restore_fragment_shader(ctx->cso);
+   cso_restore_vertex_shader(ctx->cso);
+   cso_restore_viewport(ctx->cso);
+
+   screen->texture_release(screen, &tex);
+}
+
+
+/* Release vertex buffer at end of frame to avoid synchronous
+ * rendering.
+ */
+void util_blit_flush( struct blit_state *ctx )
+{
+   pipe_buffer_reference(ctx->pipe->screen, &ctx->vbuf, NULL);
+   ctx->vbuf_slot = 0;
+} 
+
+
+
+/**
+ * Copy pixel block from src texture to dst surface.
+ * Overlapping regions are acceptable.
+ *
+ * XXX Should support selection of level.
+ * XXX need some control over blitting Z and/or stencil.
+ */
+void
+util_blit_pixels_tex(struct blit_state *ctx,
+                 struct pipe_texture *tex,
+                 int srcX0, int srcY0,
+                 int srcX1, int srcY1,
+                 struct pipe_surface *dst,
+                 int dstX0, int dstY0,
+                 int dstX1, int dstY1,
+                 float z, uint filter)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_framebuffer_state fb;
+   float s0, t0, s1, t1;
+   unsigned offset;
+
+   assert(filter == PIPE_TEX_MIPFILTER_NEAREST ||
+          filter == PIPE_TEX_MIPFILTER_LINEAR);
+
+   assert(tex->width[0] != 0);
+   assert(tex->height[0] != 0);
+
+   s0 = srcX0 / (float)tex->width[0];
+   s1 = srcX1 / (float)tex->width[0];
+   t0 = srcY0 / (float)tex->height[0];
+   t1 = srcY1 / (float)tex->height[0];
+
+   assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_RENDER_TARGET, 0));
+
+   /* save state (restored below) */
+   cso_save_blend(ctx->cso);
+   cso_save_depth_stencil_alpha(ctx->cso);
+   cso_save_rasterizer(ctx->cso);
+   cso_save_samplers(ctx->cso);
+   cso_save_sampler_textures(ctx->cso);
+   cso_save_framebuffer(ctx->cso);
+   cso_save_fragment_shader(ctx->cso);
+   cso_save_vertex_shader(ctx->cso);
+   cso_save_viewport(ctx->cso);
+
+   /* set misc state we care about */
+   cso_set_blend(ctx->cso, &ctx->blend);
+   cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil);
+   cso_set_rasterizer(ctx->cso, &ctx->rasterizer);
+   cso_set_viewport(ctx->cso, &ctx->viewport);
+
+   /* sampler */
+   ctx->sampler.min_img_filter = filter;
+   ctx->sampler.mag_img_filter = filter;
+   cso_single_sampler(ctx->cso, 0, &ctx->sampler);
+   cso_single_sampler_done(ctx->cso);
+
+   /* texture */
+   cso_set_sampler_textures(ctx->cso, 1, &tex);
+
+   /* shaders */
+   cso_set_fragment_shader_handle(ctx->cso, ctx->fs);
+   cso_set_vertex_shader_handle(ctx->cso, ctx->vs);
+
+   /* drawing dest */
+   memset(&fb, 0, sizeof(fb));
+   fb.width = dst->width;
+   fb.height = dst->height;
+   fb.nr_cbufs = 1;
+   fb.cbufs[0] = dst;
+   cso_set_framebuffer(ctx->cso, &fb);
+
+   /* draw quad */
+   offset = setup_vertex_data_tex(ctx,
+                                  (float) dstX0, (float) dstY0,
+                                  (float) dstX1, (float) dstY1,
+                                  s0, t0, s1, t1,
+                                  z);
+
+   util_draw_vertex_buffer(ctx->pipe, 
+                           ctx->vbuf, offset,
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           2); /* attribs/vert */
+
+   /* restore state we changed */
+   cso_restore_blend(ctx->cso);
+   cso_restore_depth_stencil_alpha(ctx->cso);
+   cso_restore_rasterizer(ctx->cso);
+   cso_restore_samplers(ctx->cso);
+   cso_restore_sampler_textures(ctx->cso);
+   cso_restore_framebuffer(ctx->cso);
+   cso_restore_fragment_shader(ctx->cso);
+   cso_restore_vertex_shader(ctx->cso);
+   cso_restore_viewport(ctx->cso);
+}
diff --git a/src/gallium/auxiliary/util/u_blit.h b/src/gallium/auxiliary/util/u_blit.h
new file mode 100644
index 0000000000..c35beceda8
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_blit.h
@@ -0,0 +1,82 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#ifndef U_BLIT_H
+#define U_BLIT_H
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+   
+struct pipe_context;
+struct pipe_surface;
+struct pipe_texture;
+struct cso_context;
+
+
+struct blit_state;
+
+
+extern struct blit_state *
+util_create_blit(struct pipe_context *pipe, struct cso_context *cso);
+
+extern void
+util_destroy_blit(struct blit_state *ctx);
+
+extern void
+util_blit_pixels(struct blit_state *ctx,
+                 struct pipe_surface *src,
+                 int srcX0, int srcY0,
+                 int srcX1, int srcY1,
+                 struct pipe_surface *dst,
+                 int dstX0, int dstY0,
+                 int dstX1, int dstY1,
+                 float z, uint filter);
+
+extern void
+util_blit_pixels_tex(struct blit_state *ctx,
+                     struct pipe_texture *tex,
+                     int srcX0, int srcY0,
+                     int srcX1, int srcY1,
+                     struct pipe_surface *dst,
+                     int dstX0, int dstY0,
+                     int dstX1, int dstY1,
+                     float z, uint filter);
+
+/* Call at end of frame to avoid synchronous rendering.
+ */
+extern void
+util_blit_flush( struct blit_state *ctx );
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_cache.c b/src/gallium/auxiliary/util/u_cache.c
new file mode 100644
index 0000000000..0a1a64259f
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_cache.c
@@ -0,0 +1,209 @@
+/**************************************************************************
+ *
+ * Copyright 2008 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Simple cache implementation.
+ *
+ * We simply have fixed size array, and destroy previous values on collision. 
+ * 
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "util/u_cache.h"
+
+
+struct util_cache_entry
+{
+   void *key;
+   void *value;
+   
+#ifdef DEBUG
+   unsigned count;
+#endif
+};
+
+
+struct util_cache
+{
+   /** Hash function */
+   uint32_t (*hash)(const void *key);
+   
+   /** Compare two keys */
+   int (*compare)(const void *key1, const void *key2);
+
+   /** Destroy a (key, value) pair */
+   void (*destroy)(void *key, void *value);
+
+   uint32_t size;
+   
+   struct util_cache_entry *entries;
+   
+#ifdef DEBUG
+   unsigned count;
+#endif
+};
+
+
+struct util_cache *
+util_cache_create(uint32_t (*hash)(const void *key),
+                  int (*compare)(const void *key1, const void *key2),
+                  void (*destroy)(void *key, void *value),
+                  uint32_t size)
+{
+   struct util_cache *cache;
+   
+   cache = CALLOC_STRUCT(util_cache);
+   if(!cache)
+      return NULL;
+   
+   cache->hash = hash;
+   cache->compare = compare;
+   cache->destroy = destroy;
+   cache->size = size;
+   
+   cache->entries = CALLOC(size, sizeof(struct util_cache_entry));
+   if(!cache->entries) {
+      FREE(cache);
+      return NULL;
+   }
+   
+   return cache;
+}
+
+
+static INLINE struct util_cache_entry *
+util_cache_entry_get(struct util_cache *cache,
+                     const void *key)
+{
+   uint32_t hash;
+   
+   hash = cache->hash(key);
+
+   return &cache->entries[hash % cache->size];
+}
+
+static INLINE void
+util_cache_entry_destroy(struct util_cache *cache,
+                         struct util_cache_entry *entry)
+{
+   void *key = entry->key;
+   void *value = entry->value;
+   
+   entry->key = NULL;
+   entry->value = NULL;
+   
+   if(key || value)
+      if(cache->destroy)
+         cache->destroy(key, value);
+}
+
+
+void
+util_cache_set(struct util_cache *cache,
+               void *key,
+               void *value)
+{
+   struct util_cache_entry *entry;
+
+   assert(cache);
+
+   entry = util_cache_entry_get(cache, key);
+   util_cache_entry_destroy(cache, entry);
+   
+#ifdef DEBUG
+   ++entry->count;
+   ++cache->count;
+#endif
+   
+   entry->key = key;
+   entry->value = value;
+}
+
+
+void *
+util_cache_get(struct util_cache *cache, 
+               const void *key)
+{
+   struct util_cache_entry *entry;
+
+   assert(cache);
+
+   entry = util_cache_entry_get(cache, key);
+   if(!entry->key && !entry->value)
+      return NULL;
+   
+   if(cache->compare(key, entry->key) != 0)
+      return NULL;
+   
+   return entry->value;
+}
+
+
+void 
+util_cache_clear(struct util_cache *cache)
+{
+   uint32_t i;
+
+   assert(cache);
+   
+   for(i = 0; i < cache->size; ++i)
+      util_cache_entry_destroy(cache, &cache->entries[i]);
+}
+
+
+void
+util_cache_destroy(struct util_cache *cache)
+{
+   assert(cache);
+
+#ifdef DEBUG
+   if(cache->count >= 20*cache->size) {
+      /* Normal approximation of the Poisson distribution */
+      double mean = (double)cache->count/(double)cache->size;
+      double stddev = sqrt(mean);
+      unsigned i;
+      for(i = 0; i < cache->size; ++i) {
+         double z = fabs(cache->count - mean)/stddev;
+         /* This assert should not fail 99.9999998027% of the times, unless 
+          * the hash function is a poor one */
+         assert(z <= 6.0);
+      }
+   }
+#endif
+      
+   util_cache_clear(cache);
+   
+   FREE(cache->entries);
+   FREE(cache);
+}
diff --git a/src/gallium/auxiliary/util/u_cache.h b/src/gallium/auxiliary/util/u_cache.h
new file mode 100644
index 0000000000..8a612c6585
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_cache.h
@@ -0,0 +1,87 @@
+/**************************************************************************
+ *
+ * Copyright 2008 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Simple cache.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#ifndef U_CACHE_H_
+#define U_CACHE_H_
+
+
+#include "pipe/p_compiler.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+   
+/**
+ * Least Recently Used (LRU) cache.
+ */
+struct util_cache;
+
+
+/**
+ * Create a cache.
+ * 
+ * @param hash hash function
+ * @param compare should return 0 for two equal keys
+ * @param destroy destruction callback (optional)
+ * @param size maximum number of entries
+ */
+struct util_cache *
+util_cache_create(uint32_t (*hash)(const void *key),
+                  int (*compare)(const void *key1, const void *key2),
+                  void (*destroy)(void *key, void *value),
+                  uint32_t size);
+
+void
+util_cache_set(struct util_cache *cache,
+               void *key,
+               void *value);
+
+void *
+util_cache_get(struct util_cache *cache, 
+               const void *key);
+
+void
+util_cache_clear(struct util_cache *cache);
+
+void
+util_cache_destroy(struct util_cache *cache);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* U_CACHE_H_ */
diff --git a/src/gallium/auxiliary/util/u_cpu_detect.c b/src/gallium/auxiliary/util/u_cpu_detect.c
new file mode 100644
index 0000000000..d9f2f8fc28
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_cpu_detect.c
@@ -0,0 +1,506 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Dennis Smit
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/*
+ * Based on the work of Eric Anholt <anholt@FreeBSD.org>
+ */
+
+/* FIXME: clean this entire file up */
+
+#include "u_cpu_detect.h"
+
+#ifdef __linux__
+#define OS_LINUX
+#endif
+#ifdef WIN32
+#define OS_WIN32
+#endif
+
+#if defined(ARCH_POWERPC)
+#if defined(OS_DARWIN)
+#include <sys/sysctl.h>
+#else
+#include <signal.h>
+#include <setjmp.h>
+#endif
+#endif
+
+#if defined(OS_NETBSD) || defined(OS_OPENBSD)
+#include <sys/param.h>
+#include <sys/sysctl.h>
+#include <machine/cpu.h>
+#endif
+
+#if defined(OS_FREEBSD)
+#include <sys/types.h>
+#include <sys/sysctl.h>
+#endif
+
+#if defined(OS_LINUX)
+#include <signal.h>
+#endif
+
+#if defined(OS_WIN32)
+#include <windows.h>
+#endif
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <string.h>
+
+
+static struct cpu_detect_caps __cpu_detect_caps;
+static int __cpu_detect_initialized = 0;
+
+static int has_cpuid(void);
+static int cpuid(unsigned int ax, unsigned int *p);
+
+/* The sigill handlers */
+#if defined(ARCH_X86) /* x86 (linux katmai handler check thing) */
+#if defined(OS_LINUX) && defined(_POSIX_SOURCE) && defined(X86_FXSR_MAGIC)
+static void sigill_handler_sse(int signal, struct sigcontext sc)
+{
+	/* Both the "xorps %%xmm0,%%xmm0" and "divps %xmm0,%%xmm1"
+	 * instructions are 3 bytes long.  We must increment the instruction
+	 * pointer manually to avoid repeated execution of the offending
+	 * instruction.
+	 *
+	 * If the SIGILL is caused by a divide-by-zero when unmasked
+	 * exceptions aren't supported, the SIMD FPU status and control
+	 * word will be restored at the end of the test, so we don't need
+	 * to worry about doing it here.  Besides, we may not be able to...
+	 */
+	sc.eip += 3;
+
+	__cpu_detect_caps.hasSSE=0;
+}
+
+static void sigfpe_handler_sse(int signal, struct sigcontext sc)
+{
+	if (sc.fpstate->magic != 0xffff) {
+		/* Our signal context has the extended FPU state, so reset the
+		 * divide-by-zero exception mask and clear the divide-by-zero
+		 * exception bit.
+		 */
+		sc.fpstate->mxcsr |= 0x00000200;
+		sc.fpstate->mxcsr &= 0xfffffffb;
+	} else {
+		/* If we ever get here, we're completely hosed.
+		*/
+	}
+}
+#endif
+#endif /* OS_LINUX && _POSIX_SOURCE && X86_FXSR_MAGIC */
+
+#if defined(OS_WIN32)
+LONG CALLBACK win32_sig_handler_sse(EXCEPTION_POINTERS* ep)
+{
+	if(ep->ExceptionRecord->ExceptionCode==EXCEPTION_ILLEGAL_INSTRUCTION){
+		ep->ContextRecord->Eip +=3;
+		__cpu_detect_caps.hasSSE=0;
+		return EXCEPTION_CONTINUE_EXECUTION;
+	}
+	return EXCEPTION_CONTINUE_SEARCH;
+}
+#endif /* OS_WIN32 */
+
+
+#if defined(ARCH_POWERPC) && !defined(OS_DARWIN)
+static sigjmp_buf __lv_powerpc_jmpbuf;
+static volatile sig_atomic_t __lv_powerpc_canjump = 0;
+
+static void sigill_handler (int sig);
+
+static void sigill_handler (int sig)
+{
+	if (!__lv_powerpc_canjump) {
+		signal (sig, SIG_DFL);
+		raise (sig);
+	}
+
+	__lv_powerpc_canjump = 0;
+	siglongjmp(__lv_powerpc_jmpbuf, 1);
+}
+
+static void check_os_altivec_support(void)
+{
+#if defined(OS_DARWIN)
+	int sels[2] = {CTL_HW, HW_VECTORUNIT};
+	int has_vu = 0;
+	int len = sizeof (has_vu);
+	int err;
+
+	err = sysctl(sels, 2, &has_vu, &len, NULL, 0);
+
+	if (err == 0) {
+		if (has_vu != 0) {
+			__cpu_detect_caps.hasAltiVec = 1;
+		}
+	}
+#else /* !OS_DARWIN */
+	/* no Darwin, do it the brute-force way */
+	/* this is borrowed from the libmpeg2 library */
+	signal(SIGILL, sigill_handler);
+	if (sigsetjmp(__lv_powerpc_jmpbuf, 1)) {
+		signal(SIGILL, SIG_DFL);
+	} else {
+		__lv_powerpc_canjump = 1;
+
+		__asm __volatile
+			("mtspr 256, %0\n\t"
+			 "vand %%v0, %%v0, %%v0"
+			 :
+			 : "r" (-1));
+
+		signal(SIGILL, SIG_DFL);
+		__cpu_detect_caps.hasAltiVec = 1;
+	}
+#endif
+}
+#endif
+
+/* If we're running on a processor that can do SSE, let's see if we
+ * are allowed to or not.  This will catch 2.4.0 or later kernels that
+ * haven't been configured for a Pentium III but are running on one,
+ * and RedHat patched 2.2 kernels that have broken exception handling
+ * support for user space apps that do SSE.
+ */
+static void check_os_katmai_support(void)
+{
+#if defined(ARCH_X86)
+#if defined(OS_FREEBSD)
+	int has_sse=0, ret;
+	int len = sizeof (has_sse);
+
+	ret = sysctlbyname("hw.instruction_sse", &has_sse, &len, NULL, 0);
+	if (ret || !has_sse)
+		__cpu_detect_caps.hasSSE=0;
+
+#elif defined(OS_NETBSD) || defined(OS_OPENBSD)
+	int has_sse, has_sse2, ret, mib[2];
+	int varlen;
+
+	mib[0] = CTL_MACHDEP;
+	mib[1] = CPU_SSE;
+	varlen = sizeof (has_sse);
+
+	ret = sysctl(mib, 2, &has_sse, &varlen, NULL, 0);
+	if (ret < 0 || !has_sse) {
+		__cpu_detect_caps.hasSSE = 0;
+	} else {
+		__cpu_detect_caps.hasSSE = 1;
+	}
+
+	mib[1] = CPU_SSE2;
+	varlen = sizeof (has_sse2);
+	ret = sysctl(mib, 2, &has_sse2, &varlen, NULL, 0);
+	if (ret < 0 || !has_sse2) {
+		__cpu_detect_caps.hasSSE2 = 0;
+	} else {
+		__cpu_detect_caps.hasSSE2 = 1;
+	}
+	__cpu_detect_caps.hasSSE = 0; /* FIXME ?!?!? */
+
+#elif defined(OS_WIN32)
+	LPTOP_LEVEL_EXCEPTION_FILTER exc_fil;
+	if (__cpu_detect_caps.hasSSE) {
+		exc_fil = SetUnhandledExceptionFilter(win32_sig_handler_sse);
+		__asm __volatile ("xorps %xmm0, %xmm0");
+		SetUnhandledExceptionFilter(exc_fil);
+	}
+#elif defined(OS_LINUX)
+	struct sigaction saved_sigill;
+	struct sigaction saved_sigfpe;
+
+	/* Save the original signal handlers.
+	*/
+	sigaction(SIGILL, NULL, &saved_sigill);
+	sigaction(SIGFPE, NULL, &saved_sigfpe);
+
+	signal(SIGILL, (void (*)(int))sigill_handler_sse);
+	signal(SIGFPE, (void (*)(int))sigfpe_handler_sse);
+
+	/* Emulate test for OSFXSR in CR4.  The OS will set this bit if it
+	 * supports the extended FPU save and restore required for SSE.  If
+	 * we execute an SSE instruction on a PIII and get a SIGILL, the OS
+	 * doesn't support Streaming SIMD Exceptions, even if the processor
+	 * does.
+	 */
+	if (__cpu_detect_caps.hasSSE) {
+		__asm __volatile ("xorps %xmm1, %xmm0");
+	}
+
+	/* Emulate test for OSXMMEXCPT in CR4.  The OS will set this bit if
+	 * it supports unmasked SIMD FPU exceptions.  If we unmask the
+	 * exceptions, do a SIMD divide-by-zero and get a SIGILL, the OS
+	 * doesn't support unmasked SIMD FPU exceptions.  If we get a SIGFPE
+	 * as expected, we're okay but we need to clean up after it.
+	 *
+	 * Are we being too stringent in our requirement that the OS support
+	 * unmasked exceptions?  Certain RedHat 2.2 kernels enable SSE by
+	 * setting CR4.OSFXSR but don't support unmasked exceptions.  Win98
+	 * doesn't even support them.  We at least know the user-space SSE
+	 * support is good in kernels that do support unmasked exceptions,
+	 * and therefore to be safe I'm going to leave this test in here.
+	 */
+	if (__cpu_detect_caps.hasSSE) {
+		//      test_os_katmai_exception_support();
+	}
+
+	/* Restore the original signal handlers.
+	*/
+	sigaction(SIGILL, &saved_sigill, NULL);
+	sigaction(SIGFPE, &saved_sigfpe, NULL);
+
+#else
+	/* We can't use POSIX signal handling to test the availability of
+	 * SSE, so we disable it by default.
+	 */
+	__cpu_detect_caps.hasSSE = 0;
+#endif /* __linux__ */
+#endif
+}
+
+
+static int has_cpuid(void)
+{
+#if defined(ARCH_X86)
+	int a, c;
+
+	__asm __volatile
+		("pushf\n"
+		 "popl %0\n"
+		 "movl %0, %1\n"
+		 "xorl $0x200000, %0\n"
+		 "push %0\n"
+		 "popf\n"
+		 "pushf\n"
+		 "popl %0\n"
+		 : "=a" (a), "=c" (c)
+		 :
+		 : "cc");
+
+	return a != c;
+#else
+	return 0;
+#endif
+}
+
+static int cpuid(unsigned int ax, unsigned int *p)
+{
+#if defined(ARCH_X86)
+	unsigned int flags;
+
+	__asm __volatile
+		("movl %%ebx, %%esi\n\t"
+		 "cpuid\n\t"
+		 "xchgl %%ebx, %%esi"
+		 : "=a" (p[0]), "=S" (p[1]),
+		 "=c" (p[2]), "=d" (p[3])
+		 : "0" (ax));
+
+	return 0;
+#else
+	return -1;
+#endif
+}
+
+void cpu_detect_initialize()
+{
+	unsigned int regs[4];
+	unsigned int regs2[4];
+
+	int mib[2], ncpu;
+	int len;
+
+	memset(&__cpu_detect_caps, 0, sizeof (struct cpu_detect_caps));
+
+	/* Check for arch type */
+#if defined(ARCH_MIPS)
+	__cpu_detect_caps.type = CPU_DETECT_TYPE_MIPS;
+#elif defined(ARCH_ALPHA)
+	__cpu_detect_caps.type = CPU_DETECT_TYPE_ALPHA;
+#elif defined(ARCH_SPARC)
+	__cpu_detect_caps.type = CPU_DETECT_TYPE_SPARC;
+#elif defined(ARCH_X86)
+	__cpu_detect_caps.type = CPU_DETECT_TYPE_X86;
+#elif defined(ARCH_POWERPC)
+	__cpu_detect_caps.type = CPU_DETECT_TYPE_POWERPC;
+#else
+	__cpu_detect_caps.type = CPU_DETECT_TYPE_OTHER;
+#endif
+
+	/* Count the number of CPUs in system */
+#if !defined(OS_WIN32) && !defined(OS_UNKNOWN) && defined(_SC_NPROCESSORS_ONLN)
+	__cpu_detect_caps.nrcpu = sysconf(_SC_NPROCESSORS_ONLN);
+	if (__cpu_detect_caps.nrcpu == -1)
+		__cpu_detect_caps.nrcpu = 1;
+
+#elif defined(OS_NETBSD) || defined(OS_FREEBSD) || defined(OS_OPENBSD)
+
+	mib[0] = CTL_HW;
+	mib[1] = HW_NCPU;
+
+	len = sizeof (ncpu);
+	sysctl(mib, 2, &ncpu, &len, NULL, 0);
+	__cpu_detect_caps.nrcpu = ncpu;
+
+#else
+	__cpu_detect_caps.nrcpu = 1;
+#endif
+
+#if defined(ARCH_X86)
+	/* No cpuid, old 486 or lower */
+	if (has_cpuid() == 0)
+		return;
+
+	__cpu_detect_caps.cacheline = 32;
+
+	/* Get max cpuid level */
+	cpuid(0x00000000, regs);
+
+	if (regs[0] >= 0x00000001) {
+		unsigned int cacheline;
+
+		cpuid (0x00000001, regs2);
+
+		__cpu_detect_caps.x86cpuType = (regs2[0] >> 8) & 0xf;
+		if (__cpu_detect_caps.x86cpuType == 0xf)
+		    __cpu_detect_caps.x86cpuType = 8 + ((regs2[0] >> 20) & 255); /* use extended family (P4, IA64) */
+
+		/* general feature flags */
+		__cpu_detect_caps.hasTSC  = (regs2[3] & (1 << 8  )) >>  8; /* 0x0000010 */
+		__cpu_detect_caps.hasMMX  = (regs2[3] & (1 << 23 )) >> 23; /* 0x0800000 */
+		__cpu_detect_caps.hasSSE  = (regs2[3] & (1 << 25 )) >> 25; /* 0x2000000 */
+		__cpu_detect_caps.hasSSE2 = (regs2[3] & (1 << 26 )) >> 26; /* 0x4000000 */
+		__cpu_detect_caps.hasSSE3 = (regs2[2] & (1));	       /* 0x0000001 */
+		__cpu_detect_caps.hasSSSE3 = (regs2[2] & (1 << 9 )) >> 9;   /* 0x0000020 */
+		__cpu_detect_caps.hasMMX2 = __cpu_detect_caps.hasSSE; /* SSE cpus supports mmxext too */
+
+		cacheline = ((regs2[1] >> 8) & 0xFF) * 8;
+		if (cacheline > 0)
+			__cpu_detect_caps.cacheline = cacheline;
+	}
+
+	cpuid(0x80000000, regs);
+
+	if (regs[0] >= 0x80000001) {
+
+		cpuid(0x80000001, regs2);
+
+		__cpu_detect_caps.hasMMX  |= (regs2[3] & (1 << 23 )) >> 23; /* 0x0800000 */
+		__cpu_detect_caps.hasMMX2 |= (regs2[3] & (1 << 22 )) >> 22; /* 0x400000 */
+		__cpu_detect_caps.has3DNow    = (regs2[3] & (1 << 31 )) >> 31; /* 0x80000000 */
+		__cpu_detect_caps.has3DNowExt = (regs2[3] & (1 << 30 )) >> 30;
+	}
+
+	if (regs[0] >= 0x80000006) {
+		cpuid(0x80000006, regs2);
+		__cpu_detect_caps.cacheline = regs2[2] & 0xFF;
+	}
+
+
+#if defined(OS_LINUX) || defined(OS_FREEBSD) || defined(OS_NETBSD) || defined(OS_CYGWIN) || defined(OS_OPENBSD)
+	if (__cpu_detect_caps.hasSSE)
+		check_os_katmai_support();
+
+	if (!__cpu_detect_caps.hasSSE) {
+		__cpu_detect_caps.hasSSE2 = 0;
+		__cpu_detect_caps.hasSSE3 = 0;
+		__cpu_detect_caps.hasSSSE3 = 0;
+	}
+#else
+	__cpu_detect_caps.hasSSE = 0;
+	__cpu_detect_caps.hasSSE2 = 0;
+	__cpu_detect_caps.hasSSE3 = 0;
+	__cpu_detect_caps.hasSSSE3 = 0;
+#endif
+#endif /* ARCH_X86 */
+
+#if defined(ARCH_POWERPC)
+	check_os_altivec_support();
+#endif /* ARCH_POWERPC */
+
+	__cpu_detect_initialized = 1;
+}
+
+struct cpu_detect_caps *cpu_detect_get_caps()
+{
+	return &__cpu_detect_caps;
+}
+
+/* The getters and setters for feature flags */
+int cpu_detect_get_tsc()
+{
+	return __cpu_detect_caps.hasTSC;
+}
+
+int cpu_detect_get_mmx()
+{
+	return __cpu_detect_caps.hasMMX;
+}
+
+int cpu_detect_get_mmx2()
+{
+	return __cpu_detect_caps.hasMMX2;
+}
+
+int cpu_detect_get_sse()
+{
+	return __cpu_detect_caps.hasSSE;
+}
+
+int cpu_detect_get_sse2()
+{
+	return __cpu_detect_caps.hasSSE2;
+}
+
+int cpu_detect_get_sse3()
+{
+	return __cpu_detect_caps.hasSSE3;
+}
+
+int cpu_detect_get_ssse3()
+{
+	return __cpu_detect_caps.hasSSSE3;
+}
+
+int cpu_detect_get_3dnow()
+{
+	return __cpu_detect_caps.has3DNow;
+}
+
+int cpu_detect_get_3dnow2()
+{
+	return __cpu_detect_caps.has3DNowExt;
+}
+
+int cpu_detect_get_altivec()
+{
+	return __cpu_detect_caps.hasAltiVec;
+}
+
diff --git a/src/gallium/auxiliary/util/u_cpu_detect.h b/src/gallium/auxiliary/util/u_cpu_detect.h
new file mode 100644
index 0000000000..1612d49286
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_cpu_detect.h
@@ -0,0 +1,78 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Dennis Smit
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ ***************************************************************************/
+
+/*
+ * Based on the work of Eric Anholt <anholt@FreeBSD.org>
+ */
+
+#ifndef _CPU_DETECT_H
+#define _CPU_DETECT_H
+
+typedef enum {
+	CPU_DETECT_TYPE_MIPS,
+	CPU_DETECT_TYPE_ALPHA,
+	CPU_DETECT_TYPE_SPARC,
+	CPU_DETECT_TYPE_X86,
+	CPU_DETECT_TYPE_POWERPC,
+	CPU_DETECT_TYPE_OTHER
+} cpu_detect_type;
+
+struct cpu_detect_caps {
+	cpu_detect_type	type;
+	int		nrcpu;
+
+	/* Feature flags */
+	int		x86cpuType;
+	int		cacheline;
+
+	int		hasTSC;
+	int		hasMMX;
+	int		hasMMX2;
+	int		hasSSE;
+	int		hasSSE2;
+	int		hasSSE3;
+	int		hasSSSE3;
+	int		has3DNow;
+	int		has3DNowExt;
+	int		hasAltiVec;
+};
+
+/* prototypes */
+void cpu_detect_initialize(void);
+struct cpu_detect_caps *cpu_detect_get_caps(void);
+
+int cpu_detect_get_tsc(void);
+int cpu_detect_get_mmx(void);
+int cpu_detect_get_mmx2(void);
+int cpu_detect_get_sse(void);
+int cpu_detect_get_sse2(void);
+int cpu_detect_get_sse3(void);
+int cpu_detect_get_ssse3(void);
+int cpu_detect_get_3dnow(void);
+int cpu_detect_get_3dnow2(void);
+int cpu_detect_get_altivec(void);
+
+#endif /* _CPU_DETECT_H */
diff --git a/src/gallium/auxiliary/util/u_double_list.h b/src/gallium/auxiliary/util/u_double_list.h
new file mode 100644
index 0000000000..d108d92e52
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_double_list.h
@@ -0,0 +1,99 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND. USA.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, 
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ **************************************************************************/
+
+/**
+ * \file
+ * List macros heavily inspired by the Linux kernel
+ * list handling. No list looping yet.
+ * 
+ * Is not threadsafe, so common operations need to
+ * be protected using an external mutex.
+ */
+
+#ifndef _U_DOUBLE_LIST_H_
+#define _U_DOUBLE_LIST_H_
+
+
+#include <stddef.h>
+
+
+struct list_head
+{
+    struct list_head *prev;
+    struct list_head *next;
+};
+
+
+#define LIST_INITHEAD(__item)			\
+  do {						\
+    (__item)->prev = (__item);			\
+    (__item)->next = (__item);			\
+  } while (0)
+
+#define LIST_ADD(__item, __list)		\
+  do {						\
+    (__item)->prev = (__list);			\
+    (__item)->next = (__list)->next;		\
+    (__list)->next->prev = (__item);		\
+    (__list)->next = (__item);			\
+  } while (0)
+
+#define LIST_ADDTAIL(__item, __list)		\
+  do {						\
+    (__item)->next = (__list);			\
+    (__item)->prev = (__list)->prev;		\
+    (__list)->prev->next = (__item);		\
+    (__list)->prev = (__item);			\
+  } while(0)
+
+#define LIST_REPLACE(__from, __to)		\
+  do {						\
+    (__to)->prev = (__from)->prev;		\
+    (__to)->next = (__from)->next;		\
+    (__from)->next->prev = (__to);		\
+    (__from)->prev->next = (__to);		\
+  } while (0)
+
+#define LIST_DEL(__item)			\
+  do {						\
+    (__item)->prev->next = (__item)->next;	\
+    (__item)->next->prev = (__item)->prev;	\
+  } while(0)
+
+#define LIST_DELINIT(__item)			\
+  do {						\
+    (__item)->prev->next = (__item)->next;	\
+    (__item)->next->prev = (__item)->prev;	\
+    (__item)->next = (__item);			\
+    (__item)->prev = (__item);			\
+  } while(0)
+
+#define LIST_ENTRY(__type, __item, __field)   \
+    ((__type *)(((char *)(__item)) - offsetof(__type, __field)))
+
+
+#endif /*_U_DOUBLE_LIST_H_*/
diff --git a/src/gallium/auxiliary/util/u_draw_quad.c b/src/gallium/auxiliary/util/u_draw_quad.c
new file mode 100644
index 0000000000..f282f3d289
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_draw_quad.c
@@ -0,0 +1,132 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_draw_quad.h"
+
+
+/**
+ * Draw a simple vertex buffer / primitive.
+ * Limited to float[4] vertex attribs, tightly packed.
+ */
+void 
+util_draw_vertex_buffer(struct pipe_context *pipe,
+                        struct pipe_buffer *vbuf,
+                        uint offset,
+                        uint prim_type,
+                        uint num_verts,
+                        uint num_attribs)
+{
+   struct pipe_vertex_buffer vbuffer;
+   struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS];
+   uint i;
+
+   assert(num_attribs <= PIPE_MAX_ATTRIBS);
+
+   /* tell pipe about the vertex buffer */
+   vbuffer.buffer = vbuf;
+   vbuffer.stride = num_attribs * 4 * sizeof(float);  /* vertex size */
+   vbuffer.buffer_offset = offset;
+   pipe->set_vertex_buffers(pipe, 1, &vbuffer);
+
+   /* tell pipe about the vertex attributes */
+   for (i = 0; i < num_attribs; i++) {
+      velements[i].src_offset = i * 4 * sizeof(float);
+      velements[i].vertex_buffer_index = 0;
+      velements[i].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+      velements[i].nr_components = 4;
+   }
+   pipe->set_vertex_elements(pipe, num_attribs, velements);
+
+   /* draw */
+   pipe->draw_arrays(pipe, prim_type, 0, num_verts);
+}
+
+
+
+/**
+ * Draw screen-aligned textured quad.
+ * Note: this function allocs/destroys a vertex buffer and isn't especially
+ * efficient.
+ */
+void 
+util_draw_texquad(struct pipe_context *pipe,
+                  float x0, float y0, float x1, float y1, float z)
+{
+   struct pipe_buffer *vbuf;
+   uint numAttribs = 2, vertexBytes, i, j;
+
+   vertexBytes = 4 * (4 * numAttribs * sizeof(float));
+
+   /* XXX create one-time */
+   vbuf = pipe_buffer_create(pipe->screen, 32,
+                             PIPE_BUFFER_USAGE_VERTEX, vertexBytes);
+   if (vbuf) {
+      float *v = (float *) pipe_buffer_map(pipe->screen, vbuf,
+                                           PIPE_BUFFER_USAGE_CPU_WRITE);
+      if (v) {
+         /*
+          * Load vertex buffer
+          */
+         for (i = j = 0; i < 4; i++) {
+            v[j + 2] = z;   /* z */
+            v[j + 3] = 1.0; /* w */
+            v[j + 6] = 0.0; /* r */
+            v[j + 7] = 1.0; /* q */
+            j += 8;
+         }
+
+         v[0] = x0;
+         v[1] = y0;
+         v[4] = 0.0; /*s*/
+         v[5] = 0.0; /*t*/
+
+         v[8] = x1;
+         v[9] = y0;
+         v[12] = 1.0;
+         v[13] = 0.0;
+
+         v[16] = x1;
+         v[17] = y1;
+         v[20] = 1.0;
+         v[21] = 1.0;
+
+         v[24] = x0;
+         v[25] = y1;
+         v[28] = 0.0;
+         v[29] = 1.0;
+
+         pipe_buffer_unmap(pipe->screen, vbuf);
+         util_draw_vertex_buffer(pipe, vbuf, 0, PIPE_PRIM_TRIANGLE_FAN, 4, 2);
+      }
+
+      pipe_buffer_reference(pipe->screen, &vbuf, NULL);
+   }
+}
diff --git a/src/gallium/auxiliary/util/u_draw_quad.h b/src/gallium/auxiliary/util/u_draw_quad.h
new file mode 100644
index 0000000000..00d3f5b715
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_draw_quad.h
@@ -0,0 +1,54 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef U_DRAWQUAD_H
+#define U_DRAWQUAD_H
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct pipe_buffer;
+
+extern void 
+util_draw_vertex_buffer(struct pipe_context *pipe,
+                        struct pipe_buffer *vbuf, uint offset,
+                        uint num_attribs, uint num_verts, uint prim_type);
+
+
+extern void 
+util_draw_texquad(struct pipe_context *pipe,
+                  float x0, float y0, float x1, float y1, float z);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
new file mode 100644
index 0000000000..2b4cdab6cf
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -0,0 +1,1559 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2008  VMware, Inc.  All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Mipmap generation utility
+ *  
+ * @author Brian Paul
+ */
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_memory.h"
+#include "util/u_draw_quad.h"
+#include "util/u_gen_mipmap.h"
+#include "util/u_simple_shaders.h"
+
+#include "tgsi/tgsi_build.h"
+#include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_parse.h"
+
+#include "cso_cache/cso_context.h"
+
+
+struct gen_mipmap_state
+{
+   struct pipe_context *pipe;
+   struct cso_context *cso;
+
+   struct pipe_blend_state blend;
+   struct pipe_depth_stencil_alpha_state depthstencil;
+   struct pipe_rasterizer_state rasterizer;
+   struct pipe_sampler_state sampler;
+   struct pipe_viewport_state viewport;
+
+   struct pipe_shader_state vert_shader;
+   struct pipe_shader_state frag_shader;
+   void *vs;
+   void *fs;
+
+   struct pipe_buffer *vbuf;  /**< quad vertices */
+   unsigned vbuf_slot;
+
+   float vertices[4][2][4];   /**< vertex/texcoords for quad */
+};
+
+
+
+enum dtype
+{
+   UBYTE,
+   UBYTE_3_3_2,
+   USHORT,
+   USHORT_4_4_4_4,
+   USHORT_5_6_5,
+   USHORT_1_5_5_5_REV,
+   UINT,
+   FLOAT,
+   HALF_FLOAT
+};
+
+
+typedef ushort half_float;
+
+
+static half_float
+float_to_half(float f)
+{
+   /* XXX fix this */
+   return 0;
+}
+
+static float
+half_to_float(half_float h)
+{
+   /* XXX fix this */
+   return 0.0f;
+}
+
+
+
+
+/**
+ * \name Support macros for do_row and do_row_3d
+ *
+ * The macro madness is here for two reasons.  First, it compacts the code
+ * slightly.  Second, it makes it much easier to adjust the specifics of the
+ * filter to tune the rounding characteristics.
+ */
+/*@{*/
+#define DECLARE_ROW_POINTERS(t, e) \
+      const t(*rowA)[e] = (const t(*)[e]) srcRowA; \
+      const t(*rowB)[e] = (const t(*)[e]) srcRowB; \
+      const t(*rowC)[e] = (const t(*)[e]) srcRowC; \
+      const t(*rowD)[e] = (const t(*)[e]) srcRowD; \
+      t(*dst)[e] = (t(*)[e]) dstRow
+
+#define DECLARE_ROW_POINTERS0(t) \
+      const t *rowA = (const t *) srcRowA; \
+      const t *rowB = (const t *) srcRowB; \
+      const t *rowC = (const t *) srcRowC; \
+      const t *rowD = (const t *) srcRowD; \
+      t *dst = (t *) dstRow
+
+#define FILTER_SUM_3D(Aj, Ak, Bj, Bk, Cj, Ck, Dj, Dk) \
+   ((unsigned) Aj + (unsigned) Ak \
+    + (unsigned) Bj + (unsigned) Bk \
+    + (unsigned) Cj + (unsigned) Ck \
+    + (unsigned) Dj + (unsigned) Dk \
+    + 4) >> 3
+
+#define FILTER_3D(e) \
+   do { \
+      dst[i][e] = FILTER_SUM_3D(rowA[j][e], rowA[k][e], \
+                                rowB[j][e], rowB[k][e], \
+                                rowC[j][e], rowC[k][e], \
+                                rowD[j][e], rowD[k][e]); \
+   } while(0)
+   
+#define FILTER_F_3D(e) \
+   do { \
+      dst[i][e] = (rowA[j][e] + rowA[k][e] \
+                   + rowB[j][e] + rowB[k][e] \
+                   + rowC[j][e] + rowC[k][e] \
+                   + rowD[j][e] + rowD[k][e]) * 0.125F; \
+   } while(0)
+
+#define FILTER_HF_3D(e) \
+   do { \
+      const float aj = half_to_float(rowA[j][e]); \
+      const float ak = half_to_float(rowA[k][e]); \
+      const float bj = half_to_float(rowB[j][e]); \
+      const float bk = half_to_float(rowB[k][e]); \
+      const float cj = half_to_float(rowC[j][e]); \
+      const float ck = half_to_float(rowC[k][e]); \
+      const float dj = half_to_float(rowD[j][e]); \
+      const float dk = half_to_float(rowD[k][e]); \
+      dst[i][e] = float_to_half((aj + ak + bj + bk + cj + ck + dj + dk) \
+                                      * 0.125F); \
+   } while(0)
+/*@}*/
+
+
+/**
+ * Average together two rows of a source image to produce a single new
+ * row in the dest image.  It's legal for the two source rows to point
+ * to the same data.  The source width must be equal to either the
+ * dest width or two times the dest width.
+ * \param datatype  GL_UNSIGNED_BYTE, GL_UNSIGNED_SHORT, GL_FLOAT, etc.
+ * \param comps  number of components per pixel (1..4)
+ */
+static void
+do_row(enum dtype datatype, uint comps, int srcWidth,
+       const void *srcRowA, const void *srcRowB,
+       int dstWidth, void *dstRow)
+{
+   const uint k0 = (srcWidth == dstWidth) ? 0 : 1;
+   const uint colStride = (srcWidth == dstWidth) ? 1 : 2;
+
+   assert(comps >= 1);
+   assert(comps <= 4);
+
+   /* This assertion is no longer valid with non-power-of-2 textures
+   assert(srcWidth == dstWidth || srcWidth == 2 * dstWidth);
+   */
+
+   if (datatype == UBYTE && comps == 4) {
+      uint i, j, k;
+      const ubyte(*rowA)[4] = (const ubyte(*)[4]) srcRowA;
+      const ubyte(*rowB)[4] = (const ubyte(*)[4]) srcRowB;
+      ubyte(*dst)[4] = (ubyte(*)[4]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4;
+         dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4;
+         dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4;
+         dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4;
+      }
+   }
+   else if (datatype == UBYTE && comps == 3) {
+      uint i, j, k;
+      const ubyte(*rowA)[3] = (const ubyte(*)[3]) srcRowA;
+      const ubyte(*rowB)[3] = (const ubyte(*)[3]) srcRowB;
+      ubyte(*dst)[3] = (ubyte(*)[3]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4;
+         dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4;
+         dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4;
+      }
+   }
+   else if (datatype == UBYTE && comps == 2) {
+      uint i, j, k;
+      const ubyte(*rowA)[2] = (const ubyte(*)[2]) srcRowA;
+      const ubyte(*rowB)[2] = (const ubyte(*)[2]) srcRowB;
+      ubyte(*dst)[2] = (ubyte(*)[2]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) >> 2;
+         dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) >> 2;
+      }
+   }
+   else if (datatype == UBYTE && comps == 1) {
+      uint i, j, k;
+      const ubyte *rowA = (const ubyte *) srcRowA;
+      const ubyte *rowB = (const ubyte *) srcRowB;
+      ubyte *dst = (ubyte *) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) >> 2;
+      }
+   }
+
+   else if (datatype == USHORT && comps == 4) {
+      uint i, j, k;
+      const ushort(*rowA)[4] = (const ushort(*)[4]) srcRowA;
+      const ushort(*rowB)[4] = (const ushort(*)[4]) srcRowB;
+      ushort(*dst)[4] = (ushort(*)[4]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4;
+         dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4;
+         dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4;
+         dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4;
+      }
+   }
+   else if (datatype == USHORT && comps == 3) {
+      uint i, j, k;
+      const ushort(*rowA)[3] = (const ushort(*)[3]) srcRowA;
+      const ushort(*rowB)[3] = (const ushort(*)[3]) srcRowB;
+      ushort(*dst)[3] = (ushort(*)[3]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4;
+         dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4;
+         dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4;
+      }
+   }
+   else if (datatype == USHORT && comps == 2) {
+      uint i, j, k;
+      const ushort(*rowA)[2] = (const ushort(*)[2]) srcRowA;
+      const ushort(*rowB)[2] = (const ushort(*)[2]) srcRowB;
+      ushort(*dst)[2] = (ushort(*)[2]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4;
+         dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4;
+      }
+   }
+   else if (datatype == USHORT && comps == 1) {
+      uint i, j, k;
+      const ushort *rowA = (const ushort *) srcRowA;
+      const ushort *rowB = (const ushort *) srcRowB;
+      ushort *dst = (ushort *) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) / 4;
+      }
+   }
+
+   else if (datatype == FLOAT && comps == 4) {
+      uint i, j, k;
+      const float(*rowA)[4] = (const float(*)[4]) srcRowA;
+      const float(*rowB)[4] = (const float(*)[4]) srcRowB;
+      float(*dst)[4] = (float(*)[4]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i][0] = (rowA[j][0] + rowA[k][0] +
+                      rowB[j][0] + rowB[k][0]) * 0.25F;
+         dst[i][1] = (rowA[j][1] + rowA[k][1] +
+                      rowB[j][1] + rowB[k][1]) * 0.25F;
+         dst[i][2] = (rowA[j][2] + rowA[k][2] +
+                      rowB[j][2] + rowB[k][2]) * 0.25F;
+         dst[i][3] = (rowA[j][3] + rowA[k][3] +
+                      rowB[j][3] + rowB[k][3]) * 0.25F;
+      }
+   }
+   else if (datatype == FLOAT && comps == 3) {
+      uint i, j, k;
+      const float(*rowA)[3] = (const float(*)[3]) srcRowA;
+      const float(*rowB)[3] = (const float(*)[3]) srcRowB;
+      float(*dst)[3] = (float(*)[3]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i][0] = (rowA[j][0] + rowA[k][0] +
+                      rowB[j][0] + rowB[k][0]) * 0.25F;
+         dst[i][1] = (rowA[j][1] + rowA[k][1] +
+                      rowB[j][1] + rowB[k][1]) * 0.25F;
+         dst[i][2] = (rowA[j][2] + rowA[k][2] +
+                      rowB[j][2] + rowB[k][2]) * 0.25F;
+      }
+   }
+   else if (datatype == FLOAT && comps == 2) {
+      uint i, j, k;
+      const float(*rowA)[2] = (const float(*)[2]) srcRowA;
+      const float(*rowB)[2] = (const float(*)[2]) srcRowB;
+      float(*dst)[2] = (float(*)[2]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i][0] = (rowA[j][0] + rowA[k][0] +
+                      rowB[j][0] + rowB[k][0]) * 0.25F;
+         dst[i][1] = (rowA[j][1] + rowA[k][1] +
+                      rowB[j][1] + rowB[k][1]) * 0.25F;
+      }
+   }
+   else if (datatype == FLOAT && comps == 1) {
+      uint i, j, k;
+      const float *rowA = (const float *) srcRowA;
+      const float *rowB = (const float *) srcRowB;
+      float *dst = (float *) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) * 0.25F;
+      }
+   }
+
+#if 0
+   else if (datatype == HALF_FLOAT && comps == 4) {
+      uint i, j, k, comp;
+      const half_float(*rowA)[4] = (const half_float(*)[4]) srcRowA;
+      const half_float(*rowB)[4] = (const half_float(*)[4]) srcRowB;
+      half_float(*dst)[4] = (half_float(*)[4]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         for (comp = 0; comp < 4; comp++) {
+            float aj, ak, bj, bk;
+            aj = half_to_float(rowA[j][comp]);
+            ak = half_to_float(rowA[k][comp]);
+            bj = half_to_float(rowB[j][comp]);
+            bk = half_to_float(rowB[k][comp]);
+            dst[i][comp] = float_to_half((aj + ak + bj + bk) * 0.25F);
+         }
+      }
+   }
+   else if (datatype == HALF_FLOAT && comps == 3) {
+      uint i, j, k, comp;
+      const half_float(*rowA)[3] = (const half_float(*)[3]) srcRowA;
+      const half_float(*rowB)[3] = (const half_float(*)[3]) srcRowB;
+      half_float(*dst)[3] = (half_float(*)[3]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         for (comp = 0; comp < 3; comp++) {
+            float aj, ak, bj, bk;
+            aj = half_to_float(rowA[j][comp]);
+            ak = half_to_float(rowA[k][comp]);
+            bj = half_to_float(rowB[j][comp]);
+            bk = half_to_float(rowB[k][comp]);
+            dst[i][comp] = float_to_half((aj + ak + bj + bk) * 0.25F);
+         }
+      }
+   }
+   else if (datatype == HALF_FLOAT && comps == 2) {
+      uint i, j, k, comp;
+      const half_float(*rowA)[2] = (const half_float(*)[2]) srcRowA;
+      const half_float(*rowB)[2] = (const half_float(*)[2]) srcRowB;
+      half_float(*dst)[2] = (half_float(*)[2]) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         for (comp = 0; comp < 2; comp++) {
+            float aj, ak, bj, bk;
+            aj = half_to_float(rowA[j][comp]);
+            ak = half_to_float(rowA[k][comp]);
+            bj = half_to_float(rowB[j][comp]);
+            bk = half_to_float(rowB[k][comp]);
+            dst[i][comp] = float_to_half((aj + ak + bj + bk) * 0.25F);
+         }
+      }
+   }
+   else if (datatype == HALF_FLOAT && comps == 1) {
+      uint i, j, k;
+      const half_float *rowA = (const half_float *) srcRowA;
+      const half_float *rowB = (const half_float *) srcRowB;
+      half_float *dst = (half_float *) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         float aj, ak, bj, bk;
+         aj = half_to_float(rowA[j]);
+         ak = half_to_float(rowA[k]);
+         bj = half_to_float(rowB[j]);
+         bk = half_to_float(rowB[k]);
+         dst[i] = float_to_half((aj + ak + bj + bk) * 0.25F);
+      }
+   }
+#endif
+
+   else if (datatype == UINT && comps == 1) {
+      uint i, j, k;
+      const uint *rowA = (const uint *) srcRowA;
+      const uint *rowB = (const uint *) srcRowB;
+      uint *dst = (uint *) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4;
+      }
+   }
+
+   else if (datatype == USHORT_5_6_5 && comps == 3) {
+      uint i, j, k;
+      const ushort *rowA = (const ushort *) srcRowA;
+      const ushort *rowB = (const ushort *) srcRowB;
+      ushort *dst = (ushort *) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const int rowAr0 = rowA[j] & 0x1f;
+         const int rowAr1 = rowA[k] & 0x1f;
+         const int rowBr0 = rowB[j] & 0x1f;
+         const int rowBr1 = rowB[k] & 0x1f;
+         const int rowAg0 = (rowA[j] >> 5) & 0x3f;
+         const int rowAg1 = (rowA[k] >> 5) & 0x3f;
+         const int rowBg0 = (rowB[j] >> 5) & 0x3f;
+         const int rowBg1 = (rowB[k] >> 5) & 0x3f;
+         const int rowAb0 = (rowA[j] >> 11) & 0x1f;
+         const int rowAb1 = (rowA[k] >> 11) & 0x1f;
+         const int rowBb0 = (rowB[j] >> 11) & 0x1f;
+         const int rowBb1 = (rowB[k] >> 11) & 0x1f;
+         const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2;
+         const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2;
+         const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2;
+         dst[i] = (blue << 11) | (green << 5) | red;
+      }
+   }
+   else if (datatype == USHORT_4_4_4_4 && comps == 4) {
+      uint i, j, k;
+      const ushort *rowA = (const ushort *) srcRowA;
+      const ushort *rowB = (const ushort *) srcRowB;
+      ushort *dst = (ushort *) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const int rowAr0 = rowA[j] & 0xf;
+         const int rowAr1 = rowA[k] & 0xf;
+         const int rowBr0 = rowB[j] & 0xf;
+         const int rowBr1 = rowB[k] & 0xf;
+         const int rowAg0 = (rowA[j] >> 4) & 0xf;
+         const int rowAg1 = (rowA[k] >> 4) & 0xf;
+         const int rowBg0 = (rowB[j] >> 4) & 0xf;
+         const int rowBg1 = (rowB[k] >> 4) & 0xf;
+         const int rowAb0 = (rowA[j] >> 8) & 0xf;
+         const int rowAb1 = (rowA[k] >> 8) & 0xf;
+         const int rowBb0 = (rowB[j] >> 8) & 0xf;
+         const int rowBb1 = (rowB[k] >> 8) & 0xf;
+         const int rowAa0 = (rowA[j] >> 12) & 0xf;
+         const int rowAa1 = (rowA[k] >> 12) & 0xf;
+         const int rowBa0 = (rowB[j] >> 12) & 0xf;
+         const int rowBa1 = (rowB[k] >> 12) & 0xf;
+         const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2;
+         const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2;
+         const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2;
+         const int alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2;
+         dst[i] = (alpha << 12) | (blue << 8) | (green << 4) | red;
+      }
+   }
+   else if (datatype == USHORT_1_5_5_5_REV && comps == 4) {
+      uint i, j, k;
+      const ushort *rowA = (const ushort *) srcRowA;
+      const ushort *rowB = (const ushort *) srcRowB;
+      ushort *dst = (ushort *) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const int rowAr0 = rowA[j] & 0x1f;
+         const int rowAr1 = rowA[k] & 0x1f;
+         const int rowBr0 = rowB[j] & 0x1f;
+         const int rowBr1 = rowB[k] & 0x1f;
+         const int rowAg0 = (rowA[j] >> 5) & 0x1f;
+         const int rowAg1 = (rowA[k] >> 5) & 0x1f;
+         const int rowBg0 = (rowB[j] >> 5) & 0x1f;
+         const int rowBg1 = (rowB[k] >> 5) & 0x1f;
+         const int rowAb0 = (rowA[j] >> 10) & 0x1f;
+         const int rowAb1 = (rowA[k] >> 10) & 0x1f;
+         const int rowBb0 = (rowB[j] >> 10) & 0x1f;
+         const int rowBb1 = (rowB[k] >> 10) & 0x1f;
+         const int rowAa0 = (rowA[j] >> 15) & 0x1;
+         const int rowAa1 = (rowA[k] >> 15) & 0x1;
+         const int rowBa0 = (rowB[j] >> 15) & 0x1;
+         const int rowBa1 = (rowB[k] >> 15) & 0x1;
+         const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2;
+         const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2;
+         const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2;
+         const int alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2;
+         dst[i] = (alpha << 15) | (blue << 10) | (green << 5) | red;
+      }
+   }
+   else if (datatype == UBYTE_3_3_2 && comps == 3) {
+      uint i, j, k;
+      const ubyte *rowA = (const ubyte *) srcRowA;
+      const ubyte *rowB = (const ubyte *) srcRowB;
+      ubyte *dst = (ubyte *) dstRow;
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const int rowAr0 = rowA[j] & 0x3;
+         const int rowAr1 = rowA[k] & 0x3;
+         const int rowBr0 = rowB[j] & 0x3;
+         const int rowBr1 = rowB[k] & 0x3;
+         const int rowAg0 = (rowA[j] >> 2) & 0x7;
+         const int rowAg1 = (rowA[k] >> 2) & 0x7;
+         const int rowBg0 = (rowB[j] >> 2) & 0x7;
+         const int rowBg1 = (rowB[k] >> 2) & 0x7;
+         const int rowAb0 = (rowA[j] >> 5) & 0x7;
+         const int rowAb1 = (rowA[k] >> 5) & 0x7;
+         const int rowBb0 = (rowB[j] >> 5) & 0x7;
+         const int rowBb1 = (rowB[k] >> 5) & 0x7;
+         const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2;
+         const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2;
+         const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2;
+         dst[i] = (blue << 5) | (green << 2) | red;
+      }
+   }
+   else {
+      debug_printf("bad format in do_row()");
+   }
+}
+
+
+/**
+ * Average together four rows of a source image to produce a single new
+ * row in the dest image.  It's legal for the two source rows to point
+ * to the same data.  The source width must be equal to either the
+ * dest width or two times the dest width.
+ *
+ * \param datatype  GL pixel type \c GL_UNSIGNED_BYTE, \c GL_UNSIGNED_SHORT,
+ *                  \c GL_FLOAT, etc.
+ * \param comps     number of components per pixel (1..4)
+ * \param srcWidth  Width of a row in the source data
+ * \param srcRowA   Pointer to one of the rows of source data
+ * \param srcRowB   Pointer to one of the rows of source data
+ * \param srcRowC   Pointer to one of the rows of source data
+ * \param srcRowD   Pointer to one of the rows of source data
+ * \param dstWidth  Width of a row in the destination data
+ * \param srcRowA   Pointer to the row of destination data
+ */
+static void
+do_row_3D(enum dtype datatype, uint comps, int srcWidth,
+          const void *srcRowA, const void *srcRowB,
+          const void *srcRowC, const void *srcRowD,
+          int dstWidth, void *dstRow)
+{
+   const uint k0 = (srcWidth == dstWidth) ? 0 : 1;
+   const uint colStride = (srcWidth == dstWidth) ? 1 : 2;
+   uint i, j, k;
+
+   assert(comps >= 1);
+   assert(comps <= 4);
+
+   if ((datatype == UBYTE) && (comps == 4)) {
+      DECLARE_ROW_POINTERS(ubyte, 4);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+         FILTER_3D(2);
+         FILTER_3D(3);
+      }
+   }
+   else if ((datatype == UBYTE) && (comps == 3)) {
+      DECLARE_ROW_POINTERS(ubyte, 3);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+         FILTER_3D(2);
+      }
+   }
+   else if ((datatype == UBYTE) && (comps == 2)) {
+      DECLARE_ROW_POINTERS(ubyte, 2);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+      }
+   }
+   else if ((datatype == UBYTE) && (comps == 1)) {
+      DECLARE_ROW_POINTERS(ubyte, 1);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+      }
+   }
+   else if ((datatype == USHORT) && (comps == 4)) {
+      DECLARE_ROW_POINTERS(ushort, 4);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+         FILTER_3D(2);
+         FILTER_3D(3);
+      }
+   }
+   else if ((datatype == USHORT) && (comps == 3)) {
+      DECLARE_ROW_POINTERS(ushort, 3);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+         FILTER_3D(2);
+      }
+   }
+   else if ((datatype == USHORT) && (comps == 2)) {
+      DECLARE_ROW_POINTERS(ushort, 2);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+      }
+   }
+   else if ((datatype == USHORT) && (comps == 1)) {
+      DECLARE_ROW_POINTERS(ushort, 1);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+      }
+   }
+   else if ((datatype == FLOAT) && (comps == 4)) {
+      DECLARE_ROW_POINTERS(float, 4);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_F_3D(0);
+         FILTER_F_3D(1);
+         FILTER_F_3D(2);
+         FILTER_F_3D(3);
+      }
+   }
+   else if ((datatype == FLOAT) && (comps == 3)) {
+      DECLARE_ROW_POINTERS(float, 3);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_F_3D(0);
+         FILTER_F_3D(1);
+         FILTER_F_3D(2);
+      }
+   }
+   else if ((datatype == FLOAT) && (comps == 2)) {
+      DECLARE_ROW_POINTERS(float, 2);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_F_3D(0);
+         FILTER_F_3D(1);
+      }
+   }
+   else if ((datatype == FLOAT) && (comps == 1)) {
+      DECLARE_ROW_POINTERS(float, 1);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_F_3D(0);
+      }
+   }
+   else if ((datatype == HALF_FLOAT) && (comps == 4)) {
+      DECLARE_ROW_POINTERS(half_float, 4);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_HF_3D(0);
+         FILTER_HF_3D(1);
+         FILTER_HF_3D(2);
+         FILTER_HF_3D(3);
+      }
+   }
+   else if ((datatype == HALF_FLOAT) && (comps == 3)) {
+      DECLARE_ROW_POINTERS(half_float, 4);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_HF_3D(0);
+         FILTER_HF_3D(1);
+         FILTER_HF_3D(2);
+      }
+   }
+   else if ((datatype == HALF_FLOAT) && (comps == 2)) {
+      DECLARE_ROW_POINTERS(half_float, 4);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_HF_3D(0);
+         FILTER_HF_3D(1);
+      }
+   }
+   else if ((datatype == HALF_FLOAT) && (comps == 1)) {
+      DECLARE_ROW_POINTERS(half_float, 4);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_HF_3D(0);
+      }
+   }
+   else if ((datatype == UINT) && (comps == 1)) {
+      const uint *rowA = (const uint *) srcRowA;
+      const uint *rowB = (const uint *) srcRowB;
+      const uint *rowC = (const uint *) srcRowC;
+      const uint *rowD = (const uint *) srcRowD;
+      float *dst = (float *) dstRow;
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const uint64_t tmp = (((uint64_t) rowA[j] + (uint64_t) rowA[k])
+                               + ((uint64_t) rowB[j] + (uint64_t) rowB[k])
+                               + ((uint64_t) rowC[j] + (uint64_t) rowC[k])
+                               + ((uint64_t) rowD[j] + (uint64_t) rowD[k]));
+         dst[i] = (float)((double) tmp * 0.125);
+      }
+   }
+   else if ((datatype == USHORT_5_6_5) && (comps == 3)) {
+      DECLARE_ROW_POINTERS0(ushort);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const int rowAr0 = rowA[j] & 0x1f;
+         const int rowAr1 = rowA[k] & 0x1f;
+         const int rowBr0 = rowB[j] & 0x1f;
+         const int rowBr1 = rowB[k] & 0x1f;
+         const int rowCr0 = rowC[j] & 0x1f;
+         const int rowCr1 = rowC[k] & 0x1f;
+         const int rowDr0 = rowD[j] & 0x1f;
+         const int rowDr1 = rowD[k] & 0x1f;
+         const int rowAg0 = (rowA[j] >> 5) & 0x3f;
+         const int rowAg1 = (rowA[k] >> 5) & 0x3f;
+         const int rowBg0 = (rowB[j] >> 5) & 0x3f;
+         const int rowBg1 = (rowB[k] >> 5) & 0x3f;
+         const int rowCg0 = (rowC[j] >> 5) & 0x3f;
+         const int rowCg1 = (rowC[k] >> 5) & 0x3f;
+         const int rowDg0 = (rowD[j] >> 5) & 0x3f;
+         const int rowDg1 = (rowD[k] >> 5) & 0x3f;
+         const int rowAb0 = (rowA[j] >> 11) & 0x1f;
+         const int rowAb1 = (rowA[k] >> 11) & 0x1f;
+         const int rowBb0 = (rowB[j] >> 11) & 0x1f;
+         const int rowBb1 = (rowB[k] >> 11) & 0x1f;
+         const int rowCb0 = (rowC[j] >> 11) & 0x1f;
+         const int rowCb1 = (rowC[k] >> 11) & 0x1f;
+         const int rowDb0 = (rowD[j] >> 11) & 0x1f;
+         const int rowDb1 = (rowD[k] >> 11) & 0x1f;
+         const int r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
+                                       rowCr0, rowCr1, rowDr0, rowDr1);
+         const int g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
+                                       rowCg0, rowCg1, rowDg0, rowDg1);
+         const int b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
+                                       rowCb0, rowCb1, rowDb0, rowDb1);
+         dst[i] = (b << 11) | (g << 5) | r;
+      }
+   }
+   else if ((datatype == USHORT_4_4_4_4) && (comps == 4)) {
+      DECLARE_ROW_POINTERS0(ushort);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const int rowAr0 = rowA[j] & 0xf;
+         const int rowAr1 = rowA[k] & 0xf;
+         const int rowBr0 = rowB[j] & 0xf;
+         const int rowBr1 = rowB[k] & 0xf;
+         const int rowCr0 = rowC[j] & 0xf;
+         const int rowCr1 = rowC[k] & 0xf;
+         const int rowDr0 = rowD[j] & 0xf;
+         const int rowDr1 = rowD[k] & 0xf;
+         const int rowAg0 = (rowA[j] >> 4) & 0xf;
+         const int rowAg1 = (rowA[k] >> 4) & 0xf;
+         const int rowBg0 = (rowB[j] >> 4) & 0xf;
+         const int rowBg1 = (rowB[k] >> 4) & 0xf;
+         const int rowCg0 = (rowC[j] >> 4) & 0xf;
+         const int rowCg1 = (rowC[k] >> 4) & 0xf;
+         const int rowDg0 = (rowD[j] >> 4) & 0xf;
+         const int rowDg1 = (rowD[k] >> 4) & 0xf;
+         const int rowAb0 = (rowA[j] >> 8) & 0xf;
+         const int rowAb1 = (rowA[k] >> 8) & 0xf;
+         const int rowBb0 = (rowB[j] >> 8) & 0xf;
+         const int rowBb1 = (rowB[k] >> 8) & 0xf;
+         const int rowCb0 = (rowC[j] >> 8) & 0xf;
+         const int rowCb1 = (rowC[k] >> 8) & 0xf;
+         const int rowDb0 = (rowD[j] >> 8) & 0xf;
+         const int rowDb1 = (rowD[k] >> 8) & 0xf;
+         const int rowAa0 = (rowA[j] >> 12) & 0xf;
+         const int rowAa1 = (rowA[k] >> 12) & 0xf;
+         const int rowBa0 = (rowB[j] >> 12) & 0xf;
+         const int rowBa1 = (rowB[k] >> 12) & 0xf;
+         const int rowCa0 = (rowC[j] >> 12) & 0xf;
+         const int rowCa1 = (rowC[k] >> 12) & 0xf;
+         const int rowDa0 = (rowD[j] >> 12) & 0xf;
+         const int rowDa1 = (rowD[k] >> 12) & 0xf;
+         const int r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
+                                       rowCr0, rowCr1, rowDr0, rowDr1);
+         const int g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
+                                       rowCg0, rowCg1, rowDg0, rowDg1);
+         const int b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
+                                       rowCb0, rowCb1, rowDb0, rowDb1);
+         const int a = FILTER_SUM_3D(rowAa0, rowAa1, rowBa0, rowBa1,
+                                       rowCa0, rowCa1, rowDa0, rowDa1);
+
+         dst[i] = (a << 12) | (b << 8) | (g << 4) | r;
+      }
+   }
+   else if ((datatype == USHORT_1_5_5_5_REV) && (comps == 4)) {
+      DECLARE_ROW_POINTERS0(ushort);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const int rowAr0 = rowA[j] & 0x1f;
+         const int rowAr1 = rowA[k] & 0x1f;
+         const int rowBr0 = rowB[j] & 0x1f;
+         const int rowBr1 = rowB[k] & 0x1f;
+         const int rowCr0 = rowC[j] & 0x1f;
+         const int rowCr1 = rowC[k] & 0x1f;
+         const int rowDr0 = rowD[j] & 0x1f;
+         const int rowDr1 = rowD[k] & 0x1f;
+         const int rowAg0 = (rowA[j] >> 5) & 0x1f;
+         const int rowAg1 = (rowA[k] >> 5) & 0x1f;
+         const int rowBg0 = (rowB[j] >> 5) & 0x1f;
+         const int rowBg1 = (rowB[k] >> 5) & 0x1f;
+         const int rowCg0 = (rowC[j] >> 5) & 0x1f;
+         const int rowCg1 = (rowC[k] >> 5) & 0x1f;
+         const int rowDg0 = (rowD[j] >> 5) & 0x1f;
+         const int rowDg1 = (rowD[k] >> 5) & 0x1f;
+         const int rowAb0 = (rowA[j] >> 10) & 0x1f;
+         const int rowAb1 = (rowA[k] >> 10) & 0x1f;
+         const int rowBb0 = (rowB[j] >> 10) & 0x1f;
+         const int rowBb1 = (rowB[k] >> 10) & 0x1f;
+         const int rowCb0 = (rowC[j] >> 10) & 0x1f;
+         const int rowCb1 = (rowC[k] >> 10) & 0x1f;
+         const int rowDb0 = (rowD[j] >> 10) & 0x1f;
+         const int rowDb1 = (rowD[k] >> 10) & 0x1f;
+         const int rowAa0 = (rowA[j] >> 15) & 0x1;
+         const int rowAa1 = (rowA[k] >> 15) & 0x1;
+         const int rowBa0 = (rowB[j] >> 15) & 0x1;
+         const int rowBa1 = (rowB[k] >> 15) & 0x1;
+         const int rowCa0 = (rowC[j] >> 15) & 0x1;
+         const int rowCa1 = (rowC[k] >> 15) & 0x1;
+         const int rowDa0 = (rowD[j] >> 15) & 0x1;
+         const int rowDa1 = (rowD[k] >> 15) & 0x1;
+         const int r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
+                                       rowCr0, rowCr1, rowDr0, rowDr1);
+         const int g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
+                                       rowCg0, rowCg1, rowDg0, rowDg1);
+         const int b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
+                                       rowCb0, rowCb1, rowDb0, rowDb1);
+         const int a = FILTER_SUM_3D(rowAa0, rowAa1, rowBa0, rowBa1,
+                                       rowCa0, rowCa1, rowDa0, rowDa1);
+
+         dst[i] = (a << 15) | (b << 10) | (g << 5) | r;
+      }
+   }
+   else if ((datatype == UBYTE_3_3_2) && (comps == 3)) {
+      DECLARE_ROW_POINTERS0(ushort);
+
+      for (i = j = 0, k = k0; i < (uint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const int rowAr0 = rowA[j] & 0x3;
+         const int rowAr1 = rowA[k] & 0x3;
+         const int rowBr0 = rowB[j] & 0x3;
+         const int rowBr1 = rowB[k] & 0x3;
+         const int rowCr0 = rowC[j] & 0x3;
+         const int rowCr1 = rowC[k] & 0x3;
+         const int rowDr0 = rowD[j] & 0x3;
+         const int rowDr1 = rowD[k] & 0x3;
+         const int rowAg0 = (rowA[j] >> 2) & 0x7;
+         const int rowAg1 = (rowA[k] >> 2) & 0x7;
+         const int rowBg0 = (rowB[j] >> 2) & 0x7;
+         const int rowBg1 = (rowB[k] >> 2) & 0x7;
+         const int rowCg0 = (rowC[j] >> 2) & 0x7;
+         const int rowCg1 = (rowC[k] >> 2) & 0x7;
+         const int rowDg0 = (rowD[j] >> 2) & 0x7;
+         const int rowDg1 = (rowD[k] >> 2) & 0x7;
+         const int rowAb0 = (rowA[j] >> 5) & 0x7;
+         const int rowAb1 = (rowA[k] >> 5) & 0x7;
+         const int rowBb0 = (rowB[j] >> 5) & 0x7;
+         const int rowBb1 = (rowB[k] >> 5) & 0x7;
+         const int rowCb0 = (rowC[j] >> 5) & 0x7;
+         const int rowCb1 = (rowC[k] >> 5) & 0x7;
+         const int rowDb0 = (rowD[j] >> 5) & 0x7;
+         const int rowDb1 = (rowD[k] >> 5) & 0x7;
+         const int r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
+                                       rowCr0, rowCr1, rowDr0, rowDr1);
+         const int g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
+                                       rowCg0, rowCg1, rowDg0, rowDg1);
+         const int b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
+                                       rowCb0, rowCb1, rowDb0, rowDb1);
+         dst[i] = (b << 5) | (g << 2) | r;
+      }
+   }
+   else {
+      debug_printf("bad format in do_row_3D()");
+   }
+}
+
+
+
+static void
+format_to_type_comps(enum pipe_format pformat,
+                     enum dtype *datatype, uint *comps)
+{
+   switch (pformat) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
+      *datatype = UBYTE;
+      *comps = 4;
+      return;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      *datatype = USHORT_1_5_5_5_REV;
+      *comps = 4;
+      return;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      *datatype = USHORT_4_4_4_4;
+      *comps = 4;
+      return;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      *datatype = USHORT_5_6_5;
+      *comps = 3;
+      return;
+   case PIPE_FORMAT_L8_UNORM:
+   case PIPE_FORMAT_A8_UNORM:
+   case PIPE_FORMAT_I8_UNORM:
+      *datatype = UBYTE;
+      *comps = 1;
+      return;
+   case PIPE_FORMAT_A8L8_UNORM:
+      *datatype = UBYTE;
+      *comps = 2;
+      return;
+   default:
+      assert(0);
+      *datatype = UBYTE;
+      *comps = 0;
+      break;
+   }
+}
+
+
+static void
+reduce_1d(enum pipe_format pformat,
+          int srcWidth, const ubyte *srcPtr,
+          int dstWidth, ubyte *dstPtr)
+{
+   enum dtype datatype;
+   uint comps;
+
+   format_to_type_comps(pformat, &datatype, &comps);
+
+   /* we just duplicate the input row, kind of hack, saves code */
+   do_row(datatype, comps,
+          srcWidth, srcPtr, srcPtr,
+          dstWidth, dstPtr);
+}
+
+
+/**
+ * Strides are in bytes.  If zero, it'll be computed as width * bpp.
+ */
+static void
+reduce_2d(enum pipe_format pformat,
+          int srcWidth, int srcHeight,
+          int srcRowStride, const ubyte *srcPtr,
+          int dstWidth, int dstHeight,
+          int dstRowStride, ubyte *dstPtr)
+{
+   enum dtype datatype;
+   uint comps;
+   const int bpt = pf_get_size(pformat);
+   const ubyte *srcA, *srcB;
+   ubyte *dst;
+   int row;
+
+   format_to_type_comps(pformat, &datatype, &comps);
+
+   if (!srcRowStride)
+      srcRowStride = bpt * srcWidth;
+
+   if (!dstRowStride)
+      dstRowStride = bpt * dstWidth;
+
+   /* Compute src and dst pointers */
+   srcA = srcPtr;
+   if (srcHeight > 1) 
+      srcB = srcA + srcRowStride;
+   else
+      srcB = srcA;
+   dst = dstPtr;
+
+   for (row = 0; row < dstHeight; row++) {
+      do_row(datatype, comps,
+             srcWidth, srcA, srcB,
+             dstWidth, dst);
+      srcA += 2 * srcRowStride;
+      srcB += 2 * srcRowStride;
+      dst += dstRowStride;
+   }
+}
+
+
+static void
+reduce_3d(enum pipe_format pformat,
+          int srcWidth, int srcHeight, int srcDepth,
+          int srcRowStride, const ubyte *srcPtr,
+          int dstWidth, int dstHeight, int dstDepth,
+          int dstRowStride, ubyte *dstPtr)
+{
+   const int bpt = pf_get_size(pformat);
+   const int border = 0;
+   int img, row;
+   int bytesPerSrcImage, bytesPerDstImage;
+   int bytesPerSrcRow, bytesPerDstRow;
+   int srcImageOffset, srcRowOffset;
+   enum dtype datatype;
+   uint comps;
+
+   format_to_type_comps(pformat, &datatype, &comps);
+
+   bytesPerSrcImage = srcWidth * srcHeight * bpt;
+   bytesPerDstImage = dstWidth * dstHeight * bpt;
+
+   bytesPerSrcRow = srcWidth * bpt;
+   bytesPerDstRow = dstWidth * bpt;
+
+   /* Offset between adjacent src images to be averaged together */
+   srcImageOffset = (srcDepth == dstDepth) ? 0 : bytesPerSrcImage;
+
+   /* Offset between adjacent src rows to be averaged together */
+   srcRowOffset = (srcHeight == dstHeight) ? 0 : srcWidth * bpt;
+
+   /*
+    * Need to average together up to 8 src pixels for each dest pixel.
+    * Break that down into 3 operations:
+    *   1. take two rows from source image and average them together.
+    *   2. take two rows from next source image and average them together.
+    *   3. take the two averaged rows and average them for the final dst row.
+    */
+
+   /*
+   _mesa_printf("mip3d %d x %d x %d  ->  %d x %d x %d\n",
+          srcWidth, srcHeight, srcDepth, dstWidth, dstHeight, dstDepth);
+   */
+
+   for (img = 0; img < dstDepth; img++) {
+      /* first source image pointer, skipping border */
+      const ubyte *imgSrcA = srcPtr
+         + (bytesPerSrcImage + bytesPerSrcRow + border) * bpt * border
+         + img * (bytesPerSrcImage + srcImageOffset);
+      /* second source image pointer, skipping border */
+      const ubyte *imgSrcB = imgSrcA + srcImageOffset;
+      /* address of the dest image, skipping border */
+      ubyte *imgDst = dstPtr
+         + (bytesPerDstImage + bytesPerDstRow + border) * bpt * border
+         + img * bytesPerDstImage;
+
+      /* setup the four source row pointers and the dest row pointer */
+      const ubyte *srcImgARowA = imgSrcA;
+      const ubyte *srcImgARowB = imgSrcA + srcRowOffset;
+      const ubyte *srcImgBRowA = imgSrcB;
+      const ubyte *srcImgBRowB = imgSrcB + srcRowOffset;
+      ubyte *dstImgRow = imgDst;
+
+      for (row = 0; row < dstHeight; row++) {
+         do_row_3D(datatype, comps, srcWidth, 
+                   srcImgARowA, srcImgARowB,
+                   srcImgBRowA, srcImgBRowB,
+                   dstWidth, dstImgRow);
+
+         /* advance to next rows */
+         srcImgARowA += bytesPerSrcRow + srcRowOffset;
+         srcImgARowB += bytesPerSrcRow + srcRowOffset;
+         srcImgBRowA += bytesPerSrcRow + srcRowOffset;
+         srcImgBRowB += bytesPerSrcRow + srcRowOffset;
+         dstImgRow += bytesPerDstRow;
+      }
+   }
+}
+
+
+
+
+static void
+make_1d_mipmap(struct gen_mipmap_state *ctx,
+               struct pipe_texture *pt,
+               uint face, uint baseLevel, uint lastLevel)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const uint zslice = 0;
+   uint dstLevel;
+
+   for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
+      const uint srcLevel = dstLevel - 1;
+      struct pipe_surface *srcSurf, *dstSurf;
+      void *srcMap, *dstMap;
+      
+      srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+
+      dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      srcMap = ((ubyte *) pipe_surface_map(srcSurf,
+                                           PIPE_BUFFER_USAGE_CPU_READ)
+                + srcSurf->offset);
+      dstMap = ((ubyte *) pipe_surface_map(dstSurf,
+                                           PIPE_BUFFER_USAGE_CPU_WRITE)
+                + dstSurf->offset);
+
+      reduce_1d(pt->format,
+                srcSurf->width, srcMap,
+                dstSurf->width, dstMap);
+
+      pipe_surface_unmap(srcSurf);
+      pipe_surface_unmap(dstSurf);
+
+      pipe_surface_reference(&srcSurf, NULL);
+      pipe_surface_reference(&dstSurf, NULL);
+   }
+}
+
+
+static void
+make_2d_mipmap(struct gen_mipmap_state *ctx,
+               struct pipe_texture *pt,
+               uint face, uint baseLevel, uint lastLevel)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const uint zslice = 0;
+   uint dstLevel;
+   
+   assert(pt->block.width == 1);
+   assert(pt->block.height == 1);
+
+   for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
+      const uint srcLevel = dstLevel - 1;
+      struct pipe_surface *srcSurf, *dstSurf;
+      ubyte *srcMap, *dstMap;
+      
+      srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+      dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      srcMap = ((ubyte *) pipe_surface_map(srcSurf,
+                                           PIPE_BUFFER_USAGE_CPU_READ)
+                + srcSurf->offset);
+      dstMap = ((ubyte *) pipe_surface_map(dstSurf,
+                                           PIPE_BUFFER_USAGE_CPU_WRITE)
+                + dstSurf->offset);
+
+      reduce_2d(pt->format,
+                srcSurf->width, srcSurf->height,
+                srcSurf->stride, srcMap,
+                dstSurf->width, dstSurf->height,
+                dstSurf->stride, dstMap);
+
+      pipe_surface_unmap(srcSurf);
+      pipe_surface_unmap(dstSurf);
+
+      pipe_surface_reference(&srcSurf, NULL);
+      pipe_surface_reference(&dstSurf, NULL);
+   }
+}
+
+
+static void
+make_3d_mipmap(struct gen_mipmap_state *ctx,
+               struct pipe_texture *pt,
+               uint face, uint baseLevel, uint lastLevel)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   uint dstLevel, zslice = 0;
+
+   assert(pt->block.width == 1);
+   assert(pt->block.height == 1);
+
+   for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
+      const uint srcLevel = dstLevel - 1;
+      struct pipe_surface *srcSurf, *dstSurf;
+      ubyte *srcMap, *dstMap;
+      
+      srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+      dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      srcMap = ((ubyte *) pipe_surface_map(srcSurf,
+                                           PIPE_BUFFER_USAGE_CPU_READ)
+                + srcSurf->offset);
+      dstMap = ((ubyte *) pipe_surface_map(dstSurf,
+                                           PIPE_BUFFER_USAGE_CPU_WRITE)
+                + dstSurf->offset);
+
+#if 0
+      reduce_3d(pt->format,
+                srcSurf->width, srcSurf->height,
+                srcSurf->stride, srcMap,
+                dstSurf->width, dstSurf->height,
+                dstSurf->stride, dstMap);
+#else
+      (void) reduce_3d;
+#endif
+
+      pipe_surface_unmap(srcSurf);
+      pipe_surface_unmap(dstSurf);
+
+      pipe_surface_reference(&srcSurf, NULL);
+      pipe_surface_reference(&dstSurf, NULL);
+   }
+}
+
+
+static void
+fallback_gen_mipmap(struct gen_mipmap_state *ctx,
+                    struct pipe_texture *pt,
+                    uint face, uint baseLevel, uint lastLevel)
+{
+   switch (pt->target) {
+   case PIPE_TEXTURE_1D:
+      make_1d_mipmap(ctx, pt, face, baseLevel, lastLevel);
+      break;
+   case PIPE_TEXTURE_2D:
+   case PIPE_TEXTURE_CUBE:
+      make_2d_mipmap(ctx, pt, face, baseLevel, lastLevel);
+      break;
+   case PIPE_TEXTURE_3D:
+      make_3d_mipmap(ctx, pt, face, baseLevel, lastLevel);
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * Create a mipmap generation context.
+ * The idea is to create one of these and re-use it each time we need to
+ * generate a mipmap.
+ */
+struct gen_mipmap_state *
+util_create_gen_mipmap(struct pipe_context *pipe,
+                       struct cso_context *cso)
+{
+   struct gen_mipmap_state *ctx;
+   uint i;
+
+   ctx = CALLOC_STRUCT(gen_mipmap_state);
+   if (!ctx)
+      return NULL;
+
+   ctx->pipe = pipe;
+   ctx->cso = cso;
+
+   /* disabled blending/masking */
+   memset(&ctx->blend, 0, sizeof(ctx->blend));
+   ctx->blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+   ctx->blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+   ctx->blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   ctx->blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   ctx->blend.colormask = PIPE_MASK_RGBA;
+
+   /* no-op depth/stencil/alpha */
+   memset(&ctx->depthstencil, 0, sizeof(ctx->depthstencil));
+
+   /* rasterizer */
+   memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer));
+   ctx->rasterizer.front_winding = PIPE_WINDING_CW;
+   ctx->rasterizer.cull_mode = PIPE_WINDING_NONE;
+   ctx->rasterizer.bypass_clipping = 1;
+   /*ctx->rasterizer.bypass_vs = 1;*/
+   ctx->rasterizer.gl_rasterization_rules = 1;
+
+   /* sampler state */
+   memset(&ctx->sampler, 0, sizeof(ctx->sampler));
+   ctx->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   ctx->sampler.normalized_coords = 1;
+
+   /* viewport state (identity, verts are in wincoords) */
+   ctx->viewport.scale[0] = 1.0;
+   ctx->viewport.scale[1] = 1.0;
+   ctx->viewport.scale[2] = 1.0;
+   ctx->viewport.scale[3] = 1.0;
+   ctx->viewport.translate[0] = 0.0;
+   ctx->viewport.translate[1] = 0.0;
+   ctx->viewport.translate[2] = 0.0;
+   ctx->viewport.translate[3] = 0.0;
+
+   /* vertex shader */
+   {
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_GENERIC };
+      const uint semantic_indexes[] = { 0, 0 };
+      ctx->vs = util_make_vertex_passthrough_shader(pipe, 2, semantic_names,
+                                                    semantic_indexes,
+                                                    &ctx->vert_shader);
+   }
+
+   /* fragment shader */
+   ctx->fs = util_make_fragment_tex_shader(pipe, &ctx->frag_shader);
+
+   /* vertex data that doesn't change */
+   for (i = 0; i < 4; i++) {
+      ctx->vertices[i][0][2] = 0.0f; /* z */
+      ctx->vertices[i][0][3] = 1.0f; /* w */
+      ctx->vertices[i][1][2] = 0.0f; /* r */
+      ctx->vertices[i][1][3] = 1.0f; /* q */
+   }
+
+   /* Note: the actual vertex buffer is allocated as needed below */
+
+   return ctx;
+}
+
+
+/**
+ * Get next "slot" of vertex space in the vertex buffer.
+ * We're allocating one large vertex buffer and using it piece by piece.
+ */
+static unsigned
+get_next_slot(struct gen_mipmap_state *ctx)
+{
+   const unsigned max_slots = 4096 / sizeof ctx->vertices;
+
+   if (ctx->vbuf_slot >= max_slots) 
+      util_gen_mipmap_flush( ctx );
+
+   if (!ctx->vbuf) {
+      ctx->vbuf = pipe_buffer_create(ctx->pipe->screen,
+                                     32,
+                                     PIPE_BUFFER_USAGE_VERTEX,
+                                     max_slots * sizeof ctx->vertices);
+   }
+   
+   return ctx->vbuf_slot++ * sizeof ctx->vertices;
+}
+
+
+static unsigned
+set_vertex_data(struct gen_mipmap_state *ctx, float width, float height)
+{
+   void *buf;
+   unsigned offset;
+
+   ctx->vertices[0][0][0] = 0.0f; /*x*/
+   ctx->vertices[0][0][1] = 0.0f; /*y*/
+   ctx->vertices[0][1][0] = 0.0f; /*s*/
+   ctx->vertices[0][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[1][0][0] = width;
+   ctx->vertices[1][0][1] = 0.0f;
+   ctx->vertices[1][1][0] = 1.0f;
+   ctx->vertices[1][1][1] = 0.0f;
+
+   ctx->vertices[2][0][0] = width;
+   ctx->vertices[2][0][1] = height;
+   ctx->vertices[2][1][0] = 1.0f;
+   ctx->vertices[2][1][1] = 1.0f;
+
+   ctx->vertices[3][0][0] = 0.0f;
+   ctx->vertices[3][0][1] = height;
+   ctx->vertices[3][1][0] = 0.0f;
+   ctx->vertices[3][1][1] = 1.0f;
+
+   offset = get_next_slot( ctx );
+
+   buf = pipe_buffer_map(ctx->pipe->screen, ctx->vbuf,
+                         PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   memcpy((char *)buf + offset, ctx->vertices, sizeof(ctx->vertices));
+
+   pipe_buffer_unmap(ctx->pipe->screen, ctx->vbuf);
+
+   return offset;
+}
+
+
+
+/**
+ * Destroy a mipmap generation context
+ */
+void
+util_destroy_gen_mipmap(struct gen_mipmap_state *ctx)
+{
+   struct pipe_context *pipe = ctx->pipe;
+
+   pipe->delete_vs_state(pipe, ctx->vs);
+   pipe->delete_fs_state(pipe, ctx->fs);
+
+   FREE((void*) ctx->vert_shader.tokens);
+   FREE((void*) ctx->frag_shader.tokens);
+
+   pipe_buffer_reference(pipe->screen, &ctx->vbuf, NULL);
+
+   FREE(ctx);
+}
+
+
+
+/* Release vertex buffer at end of frame to avoid synchronous
+ * rendering.
+ */
+void util_gen_mipmap_flush( struct gen_mipmap_state *ctx )
+{
+   pipe_buffer_reference(ctx->pipe->screen, &ctx->vbuf, NULL);
+   ctx->vbuf_slot = 0;
+} 
+
+
+/**
+ * Generate mipmap images.  It's assumed all needed texture memory is
+ * already allocated.
+ *
+ * \param pt  the texture to generate mipmap levels for
+ * \param face  which cube face to generate mipmaps for (0 for non-cube maps)
+ * \param baseLevel  the first mipmap level to use as a src
+ * \param lastLevel  the last mipmap level to generate
+ * \param filter  the minification filter used to generate mipmap levels with
+ * \param filter  one of PIPE_TEX_FILTER_LINEAR, PIPE_TEX_FILTER_NEAREST
+ */
+void
+util_gen_mipmap(struct gen_mipmap_state *ctx,
+                struct pipe_texture *pt,
+                uint face, uint baseLevel, uint lastLevel, uint filter)
+{
+   struct pipe_context *pipe = ctx->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_framebuffer_state fb;
+   uint dstLevel;
+   uint zslice = 0;
+   uint offset;
+
+   /* check if we can render in the texture's format */
+   if (!screen->is_format_supported(screen, pt->format, PIPE_TEXTURE_2D,
+                                    PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) {
+      fallback_gen_mipmap(ctx, pt, face, baseLevel, lastLevel);
+      return;
+   }
+
+   /* save state (restored below) */
+   cso_save_blend(ctx->cso);
+   cso_save_depth_stencil_alpha(ctx->cso);
+   cso_save_rasterizer(ctx->cso);
+   cso_save_samplers(ctx->cso);
+   cso_save_sampler_textures(ctx->cso);
+   cso_save_framebuffer(ctx->cso);
+   cso_save_fragment_shader(ctx->cso);
+   cso_save_vertex_shader(ctx->cso);
+   cso_save_viewport(ctx->cso);
+
+   /* bind our state */
+   cso_set_blend(ctx->cso, &ctx->blend);
+   cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil);
+   cso_set_rasterizer(ctx->cso, &ctx->rasterizer);
+   cso_set_viewport(ctx->cso, &ctx->viewport);
+
+   cso_set_fragment_shader_handle(ctx->cso, ctx->fs);
+   cso_set_vertex_shader_handle(ctx->cso, ctx->vs);
+
+   /* init framebuffer state */
+   memset(&fb, 0, sizeof(fb));
+   fb.nr_cbufs = 1;
+
+   /* set min/mag to same filter for faster sw speed */
+   ctx->sampler.mag_img_filter = filter;
+   ctx->sampler.min_img_filter = filter;
+
+   /*
+    * XXX for small mipmap levels, it may be faster to use the software
+    * fallback path...
+    */
+   for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
+      const uint srcLevel = dstLevel - 1;
+
+      struct pipe_surface *surf = 
+         screen->get_tex_surface(screen, pt, face, dstLevel, zslice,
+                                 PIPE_BUFFER_USAGE_GPU_WRITE);
+
+      /*
+       * Setup framebuffer / dest surface
+       */
+      fb.cbufs[0] = surf;
+      fb.width = pt->width[dstLevel];
+      fb.height = pt->height[dstLevel];
+      cso_set_framebuffer(ctx->cso, &fb);
+
+      /*
+       * Setup sampler state
+       * Note: we should only have to set the min/max LOD clamps to ensure
+       * we grab texels from the right mipmap level.  But some hardware
+       * has trouble with min clamping so we also set the lod_bias to
+       * try to work around that.
+       */
+      ctx->sampler.min_lod = ctx->sampler.max_lod = (float) srcLevel;
+      ctx->sampler.lod_bias = (float) srcLevel;
+      cso_single_sampler(ctx->cso, 0, &ctx->sampler);
+      cso_single_sampler_done(ctx->cso);
+
+      cso_set_sampler_textures(ctx->cso, 1, &pt);
+
+      /* quad coords in window coords (bypassing clipping, viewport mapping) */
+      offset = set_vertex_data(ctx,
+                               (float) pt->width[dstLevel],
+                               (float) pt->height[dstLevel]);
+
+      util_draw_vertex_buffer(ctx->pipe, 
+                              ctx->vbuf,
+                              offset,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+      /* need to signal that the texture has changed _after_ rendering to it */
+      pipe_surface_reference( &surf, NULL );
+   }
+
+   /* restore state we changed */
+   cso_restore_blend(ctx->cso);
+   cso_restore_depth_stencil_alpha(ctx->cso);
+   cso_restore_rasterizer(ctx->cso);
+   cso_restore_samplers(ctx->cso);
+   cso_restore_sampler_textures(ctx->cso);
+   cso_restore_framebuffer(ctx->cso);
+   cso_restore_fragment_shader(ctx->cso);
+   cso_restore_vertex_shader(ctx->cso);
+   cso_restore_viewport(ctx->cso);
+}
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.h b/src/gallium/auxiliary/util/u_gen_mipmap.h
new file mode 100644
index 0000000000..54608f9466
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.h
@@ -0,0 +1,70 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef U_GENMIPMAP_H
+#define U_GENMIPMAP_H
+
+#include "pipe/p_state.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+   
+struct pipe_context;
+struct pipe_texture;
+struct cso_context;
+
+struct gen_mipmap_state;
+
+
+extern struct gen_mipmap_state *
+util_create_gen_mipmap(struct pipe_context *pipe, struct cso_context *cso);
+
+
+extern void
+util_destroy_gen_mipmap(struct gen_mipmap_state *ctx);
+
+/* Release vertex buffer at end of frame to avoid synchronous
+ * rendering.
+ */
+extern void 
+util_gen_mipmap_flush( struct gen_mipmap_state *ctx );
+
+
+extern void
+util_gen_mipmap(struct gen_mipmap_state *ctx,
+                struct pipe_texture *pt,
+                uint face, uint baseLevel, uint lastLevel, uint filter);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c
new file mode 100644
index 0000000000..2d15932ce3
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_handle_table.c
@@ -0,0 +1,293 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Generic handle table implementation.
+ *  
+ * @author José Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+
+#include "util/u_memory.h"
+#include "util/u_handle_table.h"
+
+
+#define HANDLE_TABLE_INITIAL_SIZE 16  
+
+
+struct handle_table
+{
+   /** Object array. Empty handles have a null object */
+   void **objects;
+   
+   /** Number of objects the handle can currently hold */
+   unsigned size;
+   /** Number of consecutive objects allocated at the start of the table */
+   unsigned filled;
+   
+   /** Optional object destructor */
+   void (*destroy)(void *object);
+};
+
+
+struct handle_table *
+handle_table_create(void)
+{
+   struct handle_table *ht;
+   
+   ht = MALLOC_STRUCT(handle_table);
+   if(!ht)
+      return NULL;
+   
+   ht->objects = (void **)CALLOC(HANDLE_TABLE_INITIAL_SIZE, sizeof(void *));
+   if(!ht->objects) {
+      FREE(ht);
+      return NULL;
+   }
+   
+   ht->size = HANDLE_TABLE_INITIAL_SIZE;
+   ht->filled = 0;
+   
+   ht->destroy = NULL;
+   
+   return ht;
+}
+
+
+void
+handle_table_set_destroy(struct handle_table *ht,
+                         void (*destroy)(void *object))
+{
+   assert(ht);
+   ht->destroy = destroy;
+}
+
+
+/**
+ * Resize the table if necessary 
+ */
+static INLINE int
+handle_table_resize(struct handle_table *ht,
+                    unsigned minimum_size)
+{
+   unsigned new_size;
+   void **new_objects;
+
+   if(ht->size > minimum_size)
+      return ht->size;
+
+   new_size = ht->size;
+   while(!(new_size > minimum_size))
+      new_size *= 2;
+   assert(new_size);
+   
+   new_objects = (void **)REALLOC((void *)ht->objects,
+				  ht->size*sizeof(void *),
+				  new_size*sizeof(void *));
+   if(!new_objects)
+      return 0;
+   
+   memset(new_objects + ht->size, 0, (new_size - ht->size)*sizeof(void *));
+   
+   ht->size = new_size;
+   ht->objects = new_objects;
+   
+   return ht->size;
+}
+
+
+static INLINE void
+handle_table_clear(struct handle_table *ht, 
+                   unsigned index)
+{
+   void *object;
+   
+   /* The order here is important so that the object being destroyed is not
+    * present in the table when seen by the destroy callback, because the 
+    * destroy callback may directly or indirectly call the other functions in 
+    * this module.
+    */
+
+   object = ht->objects[index];
+   if(object) {
+      ht->objects[index] = NULL;
+      
+      if(ht->destroy)
+         ht->destroy(object);
+   }   
+}
+
+
+unsigned
+handle_table_add(struct handle_table *ht, 
+                 void *object)
+{
+   unsigned index;
+   unsigned handle;
+   
+   assert(ht);
+   assert(object);
+   if(!object)
+      return 0;
+
+   /* linear search for an empty handle */
+   while(ht->filled < ht->size) {
+      if(!ht->objects[ht->filled])
+	 break;
+      ++ht->filled;
+   }
+  
+   index = ht->filled;
+   handle = index + 1;
+   
+   /* check integer overflow */
+   if(!handle)
+      return 0;
+   
+   /* grow the table if necessary */
+   if(!handle_table_resize(ht, index))
+      return 0;
+
+   assert(!ht->objects[index]);
+   ht->objects[index] = object;
+   ++ht->filled;
+   
+   return handle;
+}
+
+
+unsigned
+handle_table_set(struct handle_table *ht, 
+                 unsigned handle,
+                 void *object)
+{
+   unsigned index;
+   
+   assert(ht);
+   assert(handle);
+   if(!handle)
+      return 0;
+
+   assert(object);
+   if(!object)
+      return 0;
+   
+   index = handle - 1;
+
+   /* grow the table if necessary */
+   if(!handle_table_resize(ht, index))
+      return 0;
+
+   handle_table_clear(ht, index);
+
+   ht->objects[index] = object;
+   
+   return handle;
+}
+
+
+void *
+handle_table_get(struct handle_table *ht, 
+                 unsigned handle)
+{
+   void *object;
+   
+   assert(ht);
+   assert(handle);
+   if(!handle || handle > ht->size)
+      return NULL;
+
+   object = ht->objects[handle - 1];
+   
+   return object;
+}
+
+
+void
+handle_table_remove(struct handle_table *ht, 
+                    unsigned handle)
+{
+   void *object;
+   unsigned index;
+   
+   assert(ht);
+   assert(handle);
+   if(!handle || handle > ht->size)
+      return;
+
+   index = handle - 1;
+   object = ht->objects[index];
+   if(!object)
+      return;
+   
+   handle_table_clear(ht, index);
+
+   if(index < ht->filled)
+      ht->filled = index;
+}
+
+
+unsigned
+handle_table_get_next_handle(struct handle_table *ht, 
+                             unsigned handle)
+{
+   unsigned index;
+   
+   for(index = handle; index < ht->size; ++index) {
+      if(ht->objects[index])
+	 return index + 1;
+   }
+
+   return 0;
+}
+
+
+unsigned
+handle_table_get_first_handle(struct handle_table *ht)
+{
+   return handle_table_get_next_handle(ht, 0);
+}
+
+
+void
+handle_table_destroy(struct handle_table *ht)
+{
+   unsigned index;
+   assert(ht);
+
+   if(ht->destroy)
+      for(index = 0; index < ht->size; ++index)
+         handle_table_clear(ht, index);
+   
+   FREE(ht->objects);
+   FREE(ht);
+}
+
diff --git a/src/gallium/auxiliary/util/u_handle_table.h b/src/gallium/auxiliary/util/u_handle_table.h
new file mode 100644
index 0000000000..d080135c9f
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_handle_table.h
@@ -0,0 +1,116 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Generic handle table.
+ *  
+ * @author José Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef U_HANDLE_TABLE_H_
+#define U_HANDLE_TABLE_H_
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+   
+/**
+ * Abstract data type to map integer handles to objects.
+ * 
+ * Also referred as "pointer array".
+ */
+struct handle_table;
+
+
+struct handle_table *
+handle_table_create(void);
+
+
+/**
+ * Set an optional destructor callback.
+ * 
+ * If set, it will be called during handle_table_remove and 
+ * handle_table_destroy calls.
+ */
+void
+handle_table_set_destroy(struct handle_table *ht,
+                         void (*destroy)(void *object));
+
+
+/**
+ * Add a new object.
+ * 
+ * Returns a zero handle on failure (out of memory).
+ */
+unsigned
+handle_table_add(struct handle_table *ht, 
+                 void *object);
+
+/**
+ * Returns zero on failure (out of memory).
+ */
+unsigned
+handle_table_set(struct handle_table *ht, 
+                 unsigned handle,
+                 void *object);
+
+/**
+ * Fetch an existing object.
+ * 
+ * Returns NULL for an invalid handle.
+ */
+void *
+handle_table_get(struct handle_table *ht, 
+                 unsigned handle);
+
+
+void
+handle_table_remove(struct handle_table *ht, 
+                    unsigned handle);
+
+
+void
+handle_table_destroy(struct handle_table *ht);
+
+
+unsigned
+handle_table_get_first_handle(struct handle_table *ht);
+
+
+unsigned
+handle_table_get_next_handle(struct handle_table *ht,
+                             unsigned handle);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* U_HANDLE_TABLE_H_ */
diff --git a/src/gallium/auxiliary/util/u_hash.c b/src/gallium/auxiliary/util/u_hash.c
new file mode 100644
index 0000000000..b67653ec70
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_hash.c
@@ -0,0 +1,121 @@
+/**************************************************************************
+ *
+ * Copyright 2008 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Hash functions implementation.
+ * 
+ * @author Jose Fonseca
+ */
+
+
+#include "u_hash.h"
+
+
+static const uint32_t 
+util_crc32_table[256] = {
+   0x00000000, 0x77073096, 0xee0e612c, 0x990951ba, 
+   0x076dc419, 0x706af48f, 0xe963a535, 0x9e6495a3, 
+   0x0edb8832, 0x79dcb8a4, 0xe0d5e91e, 0x97d2d988, 
+   0x09b64c2b, 0x7eb17cbd, 0xe7b82d07, 0x90bf1d91, 
+   0x1db71064, 0x6ab020f2, 0xf3b97148, 0x84be41de, 
+   0x1adad47d, 0x6ddde4eb, 0xf4d4b551, 0x83d385c7, 
+   0x136c9856, 0x646ba8c0, 0xfd62f97a, 0x8a65c9ec, 
+   0x14015c4f, 0x63066cd9, 0xfa0f3d63, 0x8d080df5, 
+   0x3b6e20c8, 0x4c69105e, 0xd56041e4, 0xa2677172, 
+   0x3c03e4d1, 0x4b04d447, 0xd20d85fd, 0xa50ab56b, 
+   0x35b5a8fa, 0x42b2986c, 0xdbbbc9d6, 0xacbcf940, 
+   0x32d86ce3, 0x45df5c75, 0xdcd60dcf, 0xabd13d59, 
+   0x26d930ac, 0x51de003a, 0xc8d75180, 0xbfd06116, 
+   0x21b4f4b5, 0x56b3c423, 0xcfba9599, 0xb8bda50f, 
+   0x2802b89e, 0x5f058808, 0xc60cd9b2, 0xb10be924, 
+   0x2f6f7c87, 0x58684c11, 0xc1611dab, 0xb6662d3d, 
+   0x76dc4190, 0x01db7106, 0x98d220bc, 0xefd5102a, 
+   0x71b18589, 0x06b6b51f, 0x9fbfe4a5, 0xe8b8d433, 
+   0x7807c9a2, 0x0f00f934, 0x9609a88e, 0xe10e9818, 
+   0x7f6a0dbb, 0x086d3d2d, 0x91646c97, 0xe6635c01, 
+   0x6b6b51f4, 0x1c6c6162, 0x856530d8, 0xf262004e, 
+   0x6c0695ed, 0x1b01a57b, 0x8208f4c1, 0xf50fc457, 
+   0x65b0d9c6, 0x12b7e950, 0x8bbeb8ea, 0xfcb9887c, 
+   0x62dd1ddf, 0x15da2d49, 0x8cd37cf3, 0xfbd44c65, 
+   0x4db26158, 0x3ab551ce, 0xa3bc0074, 0xd4bb30e2, 
+   0x4adfa541, 0x3dd895d7, 0xa4d1c46d, 0xd3d6f4fb, 
+   0x4369e96a, 0x346ed9fc, 0xad678846, 0xda60b8d0, 
+   0x44042d73, 0x33031de5, 0xaa0a4c5f, 0xdd0d7cc9, 
+   0x5005713c, 0x270241aa, 0xbe0b1010, 0xc90c2086, 
+   0x5768b525, 0x206f85b3, 0xb966d409, 0xce61e49f, 
+   0x5edef90e, 0x29d9c998, 0xb0d09822, 0xc7d7a8b4, 
+   0x59b33d17, 0x2eb40d81, 0xb7bd5c3b, 0xc0ba6cad, 
+   0xedb88320, 0x9abfb3b6, 0x03b6e20c, 0x74b1d29a, 
+   0xead54739, 0x9dd277af, 0x04db2615, 0x73dc1683, 
+   0xe3630b12, 0x94643b84, 0x0d6d6a3e, 0x7a6a5aa8, 
+   0xe40ecf0b, 0x9309ff9d, 0x0a00ae27, 0x7d079eb1, 
+   0xf00f9344, 0x8708a3d2, 0x1e01f268, 0x6906c2fe, 
+   0xf762575d, 0x806567cb, 0x196c3671, 0x6e6b06e7, 
+   0xfed41b76, 0x89d32be0, 0x10da7a5a, 0x67dd4acc, 
+   0xf9b9df6f, 0x8ebeeff9, 0x17b7be43, 0x60b08ed5, 
+   0xd6d6a3e8, 0xa1d1937e, 0x38d8c2c4, 0x4fdff252, 
+   0xd1bb67f1, 0xa6bc5767, 0x3fb506dd, 0x48b2364b, 
+   0xd80d2bda, 0xaf0a1b4c, 0x36034af6, 0x41047a60, 
+   0xdf60efc3, 0xa867df55, 0x316e8eef, 0x4669be79, 
+   0xcb61b38c, 0xbc66831a, 0x256fd2a0, 0x5268e236, 
+   0xcc0c7795, 0xbb0b4703, 0x220216b9, 0x5505262f, 
+   0xc5ba3bbe, 0xb2bd0b28, 0x2bb45a92, 0x5cb36a04, 
+   0xc2d7ffa7, 0xb5d0cf31, 0x2cd99e8b, 0x5bdeae1d, 
+   0x9b64c2b0, 0xec63f226, 0x756aa39c, 0x026d930a, 
+   0x9c0906a9, 0xeb0e363f, 0x72076785, 0x05005713, 
+   0x95bf4a82, 0xe2b87a14, 0x7bb12bae, 0x0cb61b38, 
+   0x92d28e9b, 0xe5d5be0d, 0x7cdcefb7, 0x0bdbdf21, 
+   0x86d3d2d4, 0xf1d4e242, 0x68ddb3f8, 0x1fda836e, 
+   0x81be16cd, 0xf6b9265b, 0x6fb077e1, 0x18b74777, 
+   0x88085ae6, 0xff0f6a70, 0x66063bca, 0x11010b5c, 
+   0x8f659eff, 0xf862ae69, 0x616bffd3, 0x166ccf45, 
+   0xa00ae278, 0xd70dd2ee, 0x4e048354, 0x3903b3c2, 
+   0xa7672661, 0xd06016f7, 0x4969474d, 0x3e6e77db, 
+   0xaed16a4a, 0xd9d65adc, 0x40df0b66, 0x37d83bf0, 
+   0xa9bcae53, 0xdebb9ec5, 0x47b2cf7f, 0x30b5ffe9, 
+   0xbdbdf21c, 0xcabac28a, 0x53b39330, 0x24b4a3a6, 
+   0xbad03605, 0xcdd70693, 0x54de5729, 0x23d967bf, 
+   0xb3667a2e, 0xc4614ab8, 0x5d681b02, 0x2a6f2b94, 
+   0xb40bbe37, 0xc30c8ea1, 0x5a05df1b, 0x2d02ef8d
+};
+
+
+/**
+ * @sa http://www.w3.org/TR/PNG/#D-CRCAppendix
+ */
+uint32_t
+util_hash_crc32(const void *data, size_t size)
+{
+   uint8_t *p = (uint8_t *)data;
+   uint32_t crc = 0xffffffff;
+ 
+   while (size--)
+      crc = util_crc32_table[(crc ^ *p++) & 0xff] ^ (crc >> 8);
+   
+   return crc;
+}
diff --git a/src/gallium/auxiliary/util/u_hash.h b/src/gallium/auxiliary/util/u_hash.h
new file mode 100644
index 0000000000..8d92b07c85
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_hash.h
@@ -0,0 +1,55 @@
+/**************************************************************************
+ *
+ * Copyright 2008 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Hash functions.
+ * 
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#ifndef U_HASH_H_
+#define U_HASH_H_
+
+
+#include "pipe/p_compiler.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+   
+uint32_t
+util_hash_crc32(const void *data, size_t size);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* U_HASH_H_ */
diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c
new file mode 100644
index 0000000000..0bc8de9632
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_hash_table.c
@@ -0,0 +1,280 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * General purpose hash table implementation.
+ * 
+ * Just uses the cso_hash for now, but it might be better switch to a linear 
+ * probing hash table implementation at some point -- as it is said they have 
+ * better lookup and cache performance and it appears to be possible to write 
+ * a lock-free implementation of such hash tables . 
+ * 
+ * @author José Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+
+#include "cso_cache/cso_hash.h"
+
+#include "util/u_memory.h"
+#include "util/u_hash_table.h"
+
+
+struct hash_table
+{
+   struct cso_hash *cso;   
+   
+   /** Hash function */
+   unsigned (*hash)(void *key);
+   
+   /** Compare two keys */
+   int (*compare)(void *key1, void *key2);
+   
+   /* TODO: key, value destructors? */
+};
+
+
+struct hash_table_item
+{
+   void *key;
+   void *value;
+};
+
+
+static INLINE struct hash_table_item *
+hash_table_item(struct cso_hash_iter iter)
+{
+   return (struct hash_table_item *)cso_hash_iter_data(iter);
+}
+
+
+struct hash_table *
+hash_table_create(unsigned (*hash)(void *key),
+                  int (*compare)(void *key1, void *key2))
+{
+   struct hash_table *ht;
+   
+   ht = MALLOC_STRUCT(hash_table);
+   if(!ht)
+      return NULL;
+   
+   ht->cso = cso_hash_create();
+   if(!ht->cso) {
+      FREE(ht);
+      return NULL;
+   }
+   
+   ht->hash = hash;
+   ht->compare = compare;
+   
+   return ht;
+}
+
+
+static INLINE struct cso_hash_iter
+hash_table_find_iter(struct hash_table *ht,
+                     void *key, 
+                     unsigned key_hash)
+{
+   struct cso_hash_iter iter;
+   struct hash_table_item *item;
+   
+   iter = cso_hash_find(ht->cso, key_hash);
+   while (!cso_hash_iter_is_null(iter)) {
+      item = (struct hash_table_item *)cso_hash_iter_data(iter);
+      if (!ht->compare(item->key, key))
+         break;
+      iter = cso_hash_iter_next(iter);
+   }
+   
+   return iter;
+}
+
+
+static INLINE struct hash_table_item *
+hash_table_find_item(struct hash_table *ht,
+                     void *key, 
+                     unsigned key_hash)
+{
+   struct cso_hash_iter iter;
+   struct hash_table_item *item;
+   
+   iter = cso_hash_find(ht->cso, key_hash);
+   while (!cso_hash_iter_is_null(iter)) {
+      item = (struct hash_table_item *)cso_hash_iter_data(iter);
+      if (!ht->compare(item->key, key))
+         return item;
+      iter = cso_hash_iter_next(iter);
+   }
+   
+   return NULL;
+}
+
+
+enum pipe_error
+hash_table_set(struct hash_table *ht,
+               void *key,
+               void *value)
+{
+   unsigned key_hash;
+   struct hash_table_item *item;
+   struct cso_hash_iter iter;
+
+   assert(ht);
+
+   key_hash = ht->hash(key);
+
+   item = hash_table_find_item(ht, key, key_hash);
+   if(item) {
+      /* TODO: key/value destruction? */
+      item->value = value;
+      return PIPE_OK;
+   }
+   
+   item = MALLOC_STRUCT(hash_table_item);
+   if(!item)
+      return PIPE_ERROR_OUT_OF_MEMORY;
+   
+   item->key = key;
+   item->value = value;
+   
+   iter = cso_hash_insert(ht->cso, key_hash, item);
+   if(cso_hash_iter_is_null(iter)) {
+      FREE(item);
+      return PIPE_ERROR_OUT_OF_MEMORY;
+   }
+
+   return PIPE_OK;
+}
+
+
+void *
+hash_table_get(struct hash_table *ht, 
+               void *key)
+{
+   unsigned key_hash;
+   struct hash_table_item *item;
+
+   assert(ht);
+
+   key_hash = ht->hash(key);
+
+   item = hash_table_find_item(ht, key, key_hash);
+   if(!item)
+      return NULL;
+   
+   return item->value;
+}
+
+
+void
+hash_table_remove(struct hash_table *ht, 
+                  void *key)
+{
+   unsigned key_hash;
+   struct cso_hash_iter iter;
+   struct hash_table_item *item;
+
+   assert(ht);
+
+   key_hash = ht->hash(key);
+
+   iter = hash_table_find_iter(ht, key, key_hash);
+   if(cso_hash_iter_is_null(iter))
+      return;
+   
+   item = hash_table_item(iter);
+   assert(item);
+   FREE(item);
+   
+   cso_hash_erase(ht->cso, iter);
+}
+
+
+void 
+hash_table_clear(struct hash_table *ht)
+{
+   struct cso_hash_iter iter;
+   struct hash_table_item *item;
+
+   assert(ht);
+   
+   iter = cso_hash_first_node(ht->cso);
+   while (!cso_hash_iter_is_null(iter)) {
+      item = (struct hash_table_item *)cso_hash_take(ht->cso, cso_hash_iter_key(iter));
+      FREE(item);
+      iter = cso_hash_first_node(ht->cso);
+   }
+}
+
+
+enum pipe_error
+hash_table_foreach(struct hash_table *ht,
+                   enum pipe_error (*callback)(void *key, void *value, void *data),
+                   void *data)
+{
+   struct cso_hash_iter iter;
+   struct hash_table_item *item;
+   enum pipe_error result;
+   
+   assert(ht);
+   
+   iter = cso_hash_first_node(ht->cso);
+   while (!cso_hash_iter_is_null(iter)) {
+      item = (struct hash_table_item *)cso_hash_iter_data(iter);
+      result = callback(item->key, item->value, data);
+      if(result != PIPE_OK)
+	 return result;
+      iter = cso_hash_iter_next(iter);
+   }
+
+   return PIPE_OK;
+}
+
+
+void
+hash_table_destroy(struct hash_table *ht)
+{
+   struct cso_hash_iter iter;
+   struct hash_table_item *item;
+   
+   assert(ht);
+   
+   iter = cso_hash_first_node(ht->cso);
+   while (!cso_hash_iter_is_null(iter)) {
+      item = (struct hash_table_item *)cso_hash_iter_data(iter);
+      FREE(item);
+      iter = cso_hash_iter_next(iter);
+   }
+
+   cso_hash_delete(ht->cso);
+   
+   FREE(ht);
+}
diff --git a/src/gallium/auxiliary/util/u_hash_table.h b/src/gallium/auxiliary/util/u_hash_table.h
new file mode 100644
index 0000000000..feee881582
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_hash_table.h
@@ -0,0 +1,95 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * General purpose hash table.
+ *  
+ * @author José Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef U_HASH_TABLE_H_
+#define U_HASH_TABLE_H_
+
+
+#include "pipe/p_error.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+   
+/**
+ * Generic purpose hash table.
+ */
+struct hash_table;
+
+
+/**
+ * Create an hash table.
+ * 
+ * @param hash hash function
+ * @param compare should return 0 for two equal keys.
+ */
+struct hash_table *
+hash_table_create(unsigned (*hash)(void *key),
+                  int (*compare)(void *key1, void *key2));
+
+
+enum pipe_error
+hash_table_set(struct hash_table *ht,
+               void *key,
+               void *value);
+
+void *
+hash_table_get(struct hash_table *ht, 
+               void *key);
+
+
+void
+hash_table_remove(struct hash_table *ht, 
+                  void *key);
+
+
+void
+hash_table_clear(struct hash_table *ht);
+
+
+enum pipe_error
+hash_table_foreach(struct hash_table *ht,
+                   enum pipe_error (*callback)(void *key, void *value, void *data),
+                   void *data);
+
+void
+hash_table_destroy(struct hash_table *ht);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* U_HASH_TABLE_H_ */
diff --git a/src/gallium/auxiliary/util/u_keymap.c b/src/gallium/auxiliary/util/u_keymap.c
new file mode 100644
index 0000000000..01b17ddb1b
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_keymap.c
@@ -0,0 +1,309 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * Key lookup/associative container.
+ *
+ * Like Jose's u_hash_table, based on CSO cache code for now.
+ *
+ * Author: Brian Paul
+ */
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_error.h"
+
+#include "cso_cache/cso_hash.h"
+
+#include "util/u_memory.h"
+#include "util/u_keymap.h"
+
+
+struct keymap
+{
+   struct cso_hash *cso;   
+   unsigned key_size;
+   unsigned max_entries; /* XXX not obeyed net */
+   unsigned num_entries;
+   keymap_delete_func delete_func;
+};
+
+
+struct keymap_item
+{
+   void *key, *value;
+};
+
+
+/**
+ * This the default key-delete function used when the client doesn't
+ * provide one.
+ */
+static void
+default_delete_func(const struct keymap *map,
+                    const void *key, void *data, void *user)
+{
+   FREE((void*) data);
+}
+
+
+static INLINE struct keymap_item *
+hash_table_item(struct cso_hash_iter iter)
+{
+   return (struct keymap_item *) cso_hash_iter_data(iter);
+}
+
+
+/**
+ * Return 4-byte hash key for a block of bytes.
+ */
+static unsigned
+hash(const void *key, unsigned keySize)
+{
+   unsigned i, hash;
+
+   keySize /= 4; /* convert from bytes to uints */
+
+   hash = 0;
+   for (i = 0; i < keySize; i++) {
+      hash ^= (i + 1) * ((const unsigned *) key)[i];
+   }
+
+   /*hash = hash ^ (hash >> 11) ^ (hash >> 22);*/
+
+   return hash;
+}
+
+
+/**
+ * Create a new map.
+ * \param keySize  size of the keys in bytes
+ * \param maxEntries  max number of entries to allow (~0 = infinity)
+ * \param deleteFunc  optional callback to call when entries
+ *                    are deleted/replaced
+ */
+struct keymap *
+util_new_keymap(unsigned keySize, unsigned maxEntries,
+                 keymap_delete_func deleteFunc)
+{
+   struct keymap *map = MALLOC_STRUCT(keymap);
+   if (!map)
+      return NULL;
+   
+   map->cso = cso_hash_create();
+   if (!map->cso) {
+      FREE(map);
+      return NULL;
+   }
+   
+   map->max_entries = maxEntries;
+   map->num_entries = 0;
+   map->key_size = keySize;
+   map->delete_func = deleteFunc ? deleteFunc : default_delete_func;
+
+   return map;
+}
+
+
+/**
+ * Delete/free a keymap and all entries.  The deleteFunc that was given at
+ * create time will be called for each entry.
+ * \param user  user-provided pointer passed through to the delete callback
+ */
+void
+util_delete_keymap(struct keymap *map, void *user)
+{
+   util_keymap_remove_all(map, user);
+   cso_hash_delete(map->cso);
+   FREE(map);
+}
+
+
+static INLINE struct cso_hash_iter
+hash_table_find_iter(const struct keymap *map, const void *key,
+                     unsigned key_hash)
+{
+   struct cso_hash_iter iter;
+   struct keymap_item *item;
+   
+   iter = cso_hash_find(map->cso, key_hash);
+   while (!cso_hash_iter_is_null(iter)) {
+      item = (struct keymap_item *) cso_hash_iter_data(iter);
+      if (!memcmp(item->key, key, map->key_size))
+         break;
+      iter = cso_hash_iter_next(iter);
+   }
+   
+   return iter;
+}
+
+
+static INLINE struct keymap_item *
+hash_table_find_item(const struct keymap *map, const void *key,
+                     unsigned key_hash)
+{
+   struct cso_hash_iter iter = hash_table_find_iter(map, key, key_hash);
+   if (cso_hash_iter_is_null(iter)) {
+      return NULL;
+   }
+   else {
+      return hash_table_item(iter);
+   }
+}
+
+
+/**
+ * Insert a new key + data pointer into the table.
+ * Note: we create a copy of the key, but not the data!
+ * If the key is already present in the table, replace the existing
+ * entry (calling the delete callback on the previous entry).
+ * If the maximum capacity of the map is reached an old entry
+ * will be deleted (the delete callback will be called).
+ */
+boolean
+util_keymap_insert(struct keymap *map, const void *key,
+                   const void *data, void *user)
+{
+   unsigned key_hash;
+   struct keymap_item *item;
+   struct cso_hash_iter iter;
+
+   assert(map);
+
+   key_hash = hash(key, map->key_size);
+
+   item = hash_table_find_item(map, key, key_hash);
+   if (item) {
+      /* call delete callback for old entry/item */
+      map->delete_func(map, item->key, item->value, user);
+      item->value = (void *) data;
+      return TRUE;
+   }
+   
+   item = MALLOC_STRUCT(keymap_item);
+   if (!item)
+      return FALSE;
+
+   item->key = mem_dup(key, map->key_size);
+   item->value = (void *) data;
+   
+   iter = cso_hash_insert(map->cso, key_hash, item);
+   if (cso_hash_iter_is_null(iter)) {
+      FREE(item);
+      return FALSE;
+   }
+
+   map->num_entries++;
+
+   return TRUE;
+}
+
+
+/**
+ * Look up a key in the map and return the associated data pointer.
+ */
+const void *
+util_keymap_lookup(const struct keymap *map, const void *key)
+{
+   unsigned key_hash;
+   struct keymap_item *item;
+
+   assert(map);
+
+   key_hash = hash(key, map->key_size);
+
+   item = hash_table_find_item(map, key, key_hash);
+   if (!item)
+      return NULL;
+   
+   return item->value;
+}
+
+
+/**
+ * Remove an entry from the map.
+ * The delete callback will be called if the given key/entry is found.
+ * \param user  passed to the delete callback as the last param.
+ */
+void
+util_keymap_remove(struct keymap *map, const void *key, void *user)
+{
+   unsigned key_hash;
+   struct cso_hash_iter iter;
+   struct keymap_item *item;
+
+   assert(map);
+
+   key_hash = hash(key, map->key_size);
+
+   iter = hash_table_find_iter(map, key, key_hash);
+   if (cso_hash_iter_is_null(iter))
+      return;
+   
+   item = hash_table_item(iter);
+   assert(item);
+   map->delete_func(map, item->key, item->value, user);
+   FREE(item->key);
+   FREE(item);
+   
+   map->num_entries--;
+
+   cso_hash_erase(map->cso, iter);
+}
+
+
+/**
+ * Remove all entries from the map, calling the delete callback for each.
+ * \param user  passed to the delete callback as the last param.
+ */
+void
+util_keymap_remove_all(struct keymap *map, void *user)
+{
+   struct cso_hash_iter iter;
+   struct keymap_item *item;
+
+   assert(map);
+   
+   iter = cso_hash_first_node(map->cso);
+   while (!cso_hash_iter_is_null(iter)) {
+      item = (struct keymap_item *)
+         cso_hash_take(map->cso, cso_hash_iter_key(iter));
+      map->delete_func(map, item->key, item->value, user);
+      FREE(item->key);
+      FREE(item);
+      iter = cso_hash_first_node(map->cso);
+   }
+}
+
+
+extern void
+util_keymap_info(const struct keymap *map)
+{
+   debug_printf("Keymap %p: %u of max %u entries\n",
+                (void *) map, map->num_entries, map->max_entries);
+}
diff --git a/src/gallium/auxiliary/util/u_keymap.h b/src/gallium/auxiliary/util/u_keymap.h
new file mode 100644
index 0000000000..8d60a76fc3
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_keymap.h
@@ -0,0 +1,68 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef U_KEYMAP_H
+#define U_KEYMAP_H
+
+#include "pipe/p_compiler.h"
+
+
+/** opaque keymap type */
+struct keymap;
+
+
+/** Delete/callback function type */
+typedef void (*keymap_delete_func)(const struct keymap *map,
+                                   const void *key, void *data,
+                                   void *user);
+
+
+extern struct keymap *
+util_new_keymap(unsigned keySize, unsigned maxEntries,
+                keymap_delete_func deleteFunc);
+
+extern void
+util_delete_keymap(struct keymap *map, void *user);
+
+extern boolean
+util_keymap_insert(struct keymap *map, const void *key,
+                   const void *data, void *user);
+
+extern const void *
+util_keymap_lookup(const struct keymap *map, const void *key);
+
+extern void
+util_keymap_remove(struct keymap *map, const void *key, void *user);
+
+extern void
+util_keymap_remove_all(struct keymap *map, void *user);
+
+extern void
+util_keymap_info(const struct keymap *map);
+
+
+#endif /* U_KEYMAP_H */
diff --git a/src/gallium/auxiliary/util/u_linear.c b/src/gallium/auxiliary/util/u_linear.c
new file mode 100644
index 0000000000..e999cefe74
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_linear.c
@@ -0,0 +1,70 @@
+
+#include "pipe/p_debug.h"
+#include "u_linear.h"
+
+void
+pipe_linear_to_tile(size_t src_stride, const void *src_ptr,
+		    struct pipe_tile_info *t, void *dst_ptr)
+{
+   int x, y, z;
+   char *ptr;
+   size_t bytes = t->cols * t->block.size;
+   char *dst_ptr2 = (char *) dst_ptr;
+
+   assert(pipe_linear_check_tile(t));
+
+   /* lets write lineary to the tiled buffer */
+   for (y = 0; y < t->tiles_y; y++) {
+      for (x = 0; x < t->tiles_x; x++) {
+	 /* this inner loop could be replace with SSE magic */
+	 ptr = (char*)src_ptr + src_stride * t->rows * y + bytes * x;
+	 for (z = 0; z < t->rows; z++) {
+	    memcpy(dst_ptr2, ptr, bytes);
+	    dst_ptr2 += bytes;
+	    ptr += src_stride;
+	 }
+      }
+   }
+}
+
+void pipe_linear_from_tile(struct pipe_tile_info *t, const void *src_ptr,
+			   size_t dst_stride, void *dst_ptr)
+{
+   int x, y, z;
+   char *ptr;
+   size_t bytes = t->cols * t->block.size;
+   const char *src_ptr2 = (const char *) src_ptr;
+
+   /* lets read lineary from the tiled buffer */
+   for (y = 0; y < t->tiles_y; y++) {
+      for (x = 0; x < t->tiles_x; x++) {
+	 /* this inner loop could be replace with SSE magic */
+	 ptr = (char*)dst_ptr + dst_stride * t->rows * y + bytes * x;
+	 for (z = 0; z < t->rows; z++) {
+	    memcpy(ptr, src_ptr2, bytes);
+	    src_ptr2 += bytes;
+	    ptr += dst_stride;
+	 }
+      }
+   }
+}
+
+void
+pipe_linear_fill_info(struct pipe_tile_info *t,
+		      const struct pipe_format_block *block,
+		      unsigned tile_width, unsigned tile_height,
+		      unsigned tiles_x, unsigned tiles_y)
+{
+   t->block = *block;
+
+   t->tile.width = tile_width;
+   t->tile.height = tile_height;
+   t->cols = t->tile.width / t->block.width;
+   t->rows = t->tile.height / t->block.height;
+   t->tile.size = t->cols * t->rows * t->block.size;
+
+   t->tiles_x = tiles_x;
+   t->tiles_y = tiles_y;
+   t->stride = t->cols * t->tiles_x * t->block.size;
+   t->size = t->tiles_x * t->tiles_y * t->tile.size;
+}
diff --git a/src/gallium/auxiliary/util/u_linear.h b/src/gallium/auxiliary/util/u_linear.h
new file mode 100644
index 0000000000..1589f029bc
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_linear.h
@@ -0,0 +1,61 @@
+
+#ifndef U_LINEAR_H
+#define U_LINEAR_H
+
+#include "pipe/p_format.h"
+
+struct pipe_tile_info
+{
+   unsigned size;
+   unsigned stride;
+
+   /* The number of tiles */
+   unsigned tiles_x;
+   unsigned tiles_y;
+
+   /* size of each tile expressed in blocks */
+   unsigned cols;
+   unsigned rows;
+
+   /* Describe the tile in pixels */
+   struct pipe_format_block tile;
+
+   /* Describe each block within the tile */
+   struct pipe_format_block block;
+};
+
+void pipe_linear_to_tile(size_t src_stride, const void *src_ptr,
+			 struct pipe_tile_info *t, void  *dst_ptr);
+
+void pipe_linear_from_tile(struct pipe_tile_info *t, const void *src_ptr,
+			   size_t dst_stride, void *dst_ptr);
+
+/**
+ * Convenience function to fillout a pipe_tile_info struct.
+ * @t info to fill out.
+ * @block block info about pixel layout
+ * @tile_width the width of the tile in pixels
+ * @tile_height the height of the tile in pixels
+ * @tiles_x number of tiles in x axis
+ * @tiles_y number of tiles in y axis
+ */
+void pipe_linear_fill_info(struct pipe_tile_info *t,
+			   const struct pipe_format_block *block,
+			   unsigned tile_width, unsigned tile_height,
+			   unsigned tiles_x, unsigned tiles_y);
+
+static INLINE boolean pipe_linear_check_tile(const struct pipe_tile_info *t)
+{
+   if (t->tile.size != t->block.size * t->cols * t->rows)
+      return FALSE;
+
+   if (t->stride != t->block.size * t->cols * t->tiles_x)
+      return FALSE;
+
+   if (t->size < t->stride * t->rows * t->tiles_y)
+      return FALSE;
+
+   return TRUE;
+}
+
+#endif /* U_LINEAR_H */
diff --git a/src/gallium/auxiliary/util/u_math.c b/src/gallium/auxiliary/util/u_math.c
new file mode 100644
index 0000000000..2811475fa0
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_math.c
@@ -0,0 +1,73 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+
+#include "util/u_math.h"
+
+
+/** 2^x, for x in [-1.0, 1.0) */
+float pow2_table[POW2_TABLE_SIZE];
+
+
+static void
+init_pow2_table(void)
+{
+   int i;
+   for (i = 0; i < POW2_TABLE_SIZE; i++)
+      pow2_table[i] = (float) pow(2.0, (i - POW2_TABLE_OFFSET) / POW2_TABLE_SCALE);
+}
+
+
+/** log2(x), for x in [1.0, 2.0) */
+float log2_table[LOG2_TABLE_SIZE];
+
+
+static void 
+init_log2_table(void)
+{
+   unsigned i;
+   for (i = 0; i < LOG2_TABLE_SIZE; i++)
+      log2_table[i] = (float) log2(1.0 + i * (1.0 / LOG2_TABLE_SCALE));
+}
+
+
+/**
+ * One time init for math utilities.
+ */
+void
+util_init_math(void)
+{
+   static boolean initialized = FALSE;
+   if (!initialized) {
+      init_pow2_table();
+      init_log2_table();
+      initialized = TRUE;
+   }
+}
+
+
diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h
new file mode 100644
index 0000000000..ab6f39ac31
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_math.h
@@ -0,0 +1,437 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Math utilities and approximations for common math functions.
+ * Reduced precision is usually acceptable in shaders...
+ *
+ * "fast" is used in the names of functions which are low-precision,
+ * or at least lower-precision than the normal C lib functions.
+ */
+
+
+#ifndef U_MATH_H
+#define U_MATH_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+__inline double ceil(double val)
+{
+   double ceil_val;
+
+   if((val - (long) val) == 0) {
+      ceil_val = val;
+   }
+   else {
+      if(val > 0) {
+         ceil_val = (long) val + 1;
+      }
+      else {
+         ceil_val = (long) val;
+      }
+   }
+
+   return ceil_val;
+}
+
+#ifndef PIPE_SUBSYSTEM_WINDOWS_CE_OGL
+__inline double floor(double val)
+{
+   double floor_val;
+
+   if((val - (long) val) == 0) {
+      floor_val = val;
+   }
+   else {
+      if(val > 0) {
+         floor_val = (long) val;
+      }
+      else {
+         floor_val = (long) val - 1;
+      }
+   }
+
+   return floor_val;
+}
+#endif
+
+#pragma function(pow)
+__inline double __cdecl pow(double val, double exponent)
+{
+   /* XXX */
+   assert(0);
+   return 0;
+}
+
+#pragma function(log)
+__inline double __cdecl log(double val)
+{
+   /* XXX */
+   assert(0);
+   return 0;
+}
+
+#pragma function(atan2)
+__inline double __cdecl atan2(double val)
+{
+   /* XXX */
+   assert(0);
+   return 0;
+}
+#else
+#include <math.h>
+#include <stdarg.h>
+#endif
+
+
+#if defined(_MSC_VER) 
+
+#if _MSC_VER < 1400 && !defined(__cplusplus) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+ 
+static INLINE float cosf( float f ) 
+{
+   return (float) cos( (double) f );
+}
+
+static INLINE float sinf( float f ) 
+{
+   return (float) sin( (double) f );
+}
+
+static INLINE float ceilf( float f ) 
+{
+   return (float) ceil( (double) f );
+}
+
+static INLINE float floorf( float f ) 
+{
+   return (float) floor( (double) f );
+}
+
+static INLINE float powf( float f, float g ) 
+{
+   return (float) pow( (double) f, (double) g );
+}
+
+static INLINE float sqrtf( float f ) 
+{
+   return (float) sqrt( (double) f );
+}
+
+static INLINE float fabsf( float f ) 
+{
+   return (float) fabs( (double) f );
+}
+
+static INLINE float logf( float f ) 
+{
+   return (float) log( (double) f );
+}
+
+#else
+/* Work-around an extra semi-colon in VS 2005 logf definition */
+#ifdef logf
+#undef logf
+#define logf(x) ((float)log((double)(x)))
+#endif /* logf */
+#endif
+
+static INLINE double log2( double x )
+{
+   const double invln2 = 1.442695041;
+   return log( x ) * invln2;
+}
+
+#endif /* _MSC_VER */
+
+
+
+
+
+#define POW2_TABLE_SIZE_LOG2 9
+#define POW2_TABLE_SIZE (1 << POW2_TABLE_SIZE_LOG2)
+#define POW2_TABLE_OFFSET (POW2_TABLE_SIZE/2)
+#define POW2_TABLE_SCALE ((float)(POW2_TABLE_SIZE/2))
+extern float pow2_table[POW2_TABLE_SIZE];
+
+
+
+extern void
+util_init_math(void);
+
+
+union fi {
+   float f;
+   int32_t i;
+   uint32_t ui;
+};
+
+
+/**
+ * Fast version of 2^x
+ * Identity: exp2(a + b) = exp2(a) * exp2(b)
+ * Let ipart = int(x)
+ * Let fpart = x - ipart;
+ * So, exp2(x) = exp2(ipart) * exp2(fpart)
+ * Compute exp2(ipart) with i << ipart
+ * Compute exp2(fpart) with lookup table.
+ */
+static INLINE float
+util_fast_exp2(float x)
+{
+   int32_t ipart;
+   float fpart, mpart;
+   union fi epart;
+   
+   if(x > 129.00000f)
+      return 3.402823466e+38f;
+   
+   if(x < -126.99999f)
+      return 0.0f;
+
+   ipart = (int32_t) x;
+   fpart = x - (float) ipart;
+   
+   /* same as
+    *   epart.f = (float) (1 << ipart)
+    * but faster and without integer overflow for ipart > 31 */
+   epart.i = (ipart + 127 ) << 23;
+   
+   mpart = pow2_table[POW2_TABLE_OFFSET + (int)(fpart * POW2_TABLE_SCALE)];
+   
+   return epart.f * mpart;
+}
+
+
+/**
+ * Fast approximation to exp(x).
+ */
+static INLINE float
+util_fast_exp(float x)
+{
+   const float k = 1.44269f; /* = log2(e) */
+   return util_fast_exp2(k * x);
+}
+
+
+#define LOG2_TABLE_SIZE_LOG2 16
+#define LOG2_TABLE_SCALE (1 << LOG2_TABLE_SIZE_LOG2)
+#define LOG2_TABLE_SIZE (LOG2_TABLE_SCALE + 1)
+extern float log2_table[LOG2_TABLE_SIZE];
+
+
+static INLINE float
+util_fast_log2(float x)
+{
+   union fi num;
+   float epart, mpart;
+   num.f = x;
+   epart = (float)(((num.i & 0x7f800000) >> 23) - 127);
+   /* mpart = log2_table[mantissa*LOG2_TABLE_SCALE + 0.5] */
+   mpart = log2_table[((num.i & 0x007fffff) + (1 << (22 - LOG2_TABLE_SIZE_LOG2))) >> (23 - LOG2_TABLE_SIZE_LOG2)];
+   return epart + mpart;
+}
+
+
+static INLINE float
+util_fast_pow(float x, float y)
+{
+   return util_fast_exp2(util_fast_log2(x) * y);
+}
+
+
+
+/**
+ * Floor(x), returned as int.
+ */
+static INLINE int
+util_ifloor(float f)
+{
+   int ai, bi;
+   double af, bf;
+   union fi u;
+   af = (3 << 22) + 0.5 + (double)f;
+   bf = (3 << 22) + 0.5 - (double)f;
+   u.f = (float) af;  ai = u.i;
+   u.f = (float) bf;  bi = u.i;
+   return (ai - bi) >> 1;
+}
+
+
+/**
+ * Round float to nearest int.
+ */
+static INLINE int
+util_iround(float f)
+{
+#if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86) 
+   int r;
+   __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st");
+   return r;
+#elif defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86)
+   int r;
+   _asm {
+	 fld f
+	 fistp r
+	}
+   return r;
+#else
+   if (f >= 0.0f)
+      return (int) (f + 0.5f);
+   else
+      return (int) (f - 0.5f);
+#endif
+}
+
+
+
+#if defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86)
+/**
+ * Find first bit set in word.  Least significant bit is 1.
+ * Return 0 if no bits set.
+ */
+static INLINE
+unsigned ffs( unsigned u )
+{
+   unsigned i;
+
+   if( u == 0 ) {
+      return 0;
+   }
+
+   __asm bsf eax, [u]
+   __asm inc eax
+   __asm mov [i], eax
+
+   return i;
+}
+#endif
+
+#ifdef __MINGW32__
+#define ffs __builtin_ffs
+#endif
+
+
+/**
+ * Return float bits.
+ */
+static INLINE unsigned
+fui( float f )
+{
+   union fi fi;
+   fi.f = f;
+   return fi.ui;
+}
+
+
+
+static INLINE float
+ubyte_to_float(ubyte ub)
+{
+   return (float) ub * (1.0f / 255.0f);
+}
+
+
+/**
+ * Convert float in [0,1] to ubyte in [0,255] with clamping.
+ */
+static INLINE ubyte
+float_to_ubyte(float f)
+{
+   const int ieee_0996 = 0x3f7f0000;   /* 0.996 or so */
+   union fi tmp;
+
+   tmp.f = f;
+   if (tmp.i < 0) {
+      return (ubyte) 0;
+   }
+   else if (tmp.i >= ieee_0996) {
+      return (ubyte) 255;
+   }
+   else {
+      tmp.f = tmp.f * (255.0f/256.0f) + 32768.0f;
+      return (ubyte) tmp.i;
+   }
+}
+
+
+
+#define CLAMP( X, MIN, MAX )  ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) )
+
+#define MIN2( A, B )   ( (A)<(B) ? (A) : (B) )
+#define MAX2( A, B )   ( (A)>(B) ? (A) : (B) )
+
+
+static INLINE int
+align(int value, int alignment)
+{
+   return (value + alignment - 1) & ~(alignment - 1);
+}
+
+
+#ifndef COPY_4V
+#define COPY_4V( DST, SRC )         \
+do {                                \
+   (DST)[0] = (SRC)[0];             \
+   (DST)[1] = (SRC)[1];             \
+   (DST)[2] = (SRC)[2];             \
+   (DST)[3] = (SRC)[3];             \
+} while (0)
+#endif
+
+
+#ifndef COPY_4FV
+#define COPY_4FV( DST, SRC )  COPY_4V(DST, SRC)
+#endif
+
+
+#ifndef ASSIGN_4V
+#define ASSIGN_4V( DST, V0, V1, V2, V3 ) \
+do {                                     \
+   (DST)[0] = (V0);                      \
+   (DST)[1] = (V1);                      \
+   (DST)[2] = (V2);                      \
+   (DST)[3] = (V3);                      \
+} while (0)
+#endif
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* U_MATH_H */
diff --git a/src/gallium/auxiliary/util/u_memory.h b/src/gallium/auxiliary/util/u_memory.h
new file mode 100644
index 0000000000..1a6b596421
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_memory.h
@@ -0,0 +1,234 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Memory functions
+ */
+
+
+#ifndef U_MEMORY_H
+#define U_MEMORY_H
+
+
+#include "util/u_pointer.h"
+#include "pipe/p_debug.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/* Define ENOMEM for WINCE */ 
+#if (_WIN32_WCE < 600)
+#ifndef ENOMEM
+#define ENOMEM 12
+#endif
+#endif
+
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && defined(DEBUG) 
+
+/* memory debugging */
+
+#include "pipe/p_debug.h"
+
+#define MALLOC( _size ) \
+   debug_malloc( __FILE__, __LINE__, __FUNCTION__, _size )
+#define CALLOC( _count, _size ) \
+   debug_calloc(__FILE__, __LINE__, __FUNCTION__, _count, _size )
+#define FREE( _ptr ) \
+   debug_free( __FILE__, __LINE__, __FUNCTION__,  _ptr )
+#define REALLOC( _ptr, _old_size, _size ) \
+   debug_realloc( __FILE__, __LINE__, __FUNCTION__,  _ptr, _old_size, _size )
+
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+
+void * __stdcall
+EngAllocMem(
+    unsigned long Flags,
+    unsigned long MemSize,
+    unsigned long Tag );
+
+void __stdcall
+EngFreeMem(
+    void *Mem );
+
+#define MALLOC( _size ) EngAllocMem( 0, _size, 'D3AG' )
+#define _FREE( _ptr ) EngFreeMem( _ptr )
+
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+
+void *
+ExAllocatePool(
+    unsigned long PoolType, 
+    size_t NumberOfBytes);
+
+void 
+ExFreePool(void *P);
+
+#define MALLOC(_size) ExAllocatePool(0, _size)
+#define _FREE(_ptr) ExFreePool(_ptr)
+
+#else
+
+#define MALLOC( SIZE )  malloc( SIZE )
+#define CALLOC( COUNT, SIZE )   calloc( COUNT, SIZE )
+#define FREE( PTR )  free( PTR )
+#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE )  realloc( OLDPTR, NEWSIZE )
+
+#endif
+
+
+#ifndef CALLOC
+static INLINE void *
+CALLOC( unsigned count, unsigned size )
+{
+   void *ptr = MALLOC( count * size );
+   if( ptr ) {
+      memset( ptr, 0, count * size );
+   }
+   return ptr;
+}
+#endif /* !CALLOC */
+
+#ifndef FREE
+static INLINE void
+FREE( void *ptr )
+{
+   if( ptr ) {
+      _FREE( ptr );
+   }
+}
+#endif /* !FREE */
+
+#ifndef REALLOC
+static INLINE void *
+REALLOC( void *old_ptr, unsigned old_size, unsigned new_size )
+{
+   void *new_ptr = NULL;
+
+   if (new_size != 0) {
+      unsigned copy_size = old_size < new_size ? old_size : new_size;
+      new_ptr = MALLOC( new_size );
+      if (new_ptr && old_ptr && copy_size) {
+         memcpy( new_ptr, old_ptr, copy_size );
+      }
+   }
+
+   FREE( old_ptr );
+   return new_ptr;
+}
+#endif /* !REALLOC */
+
+
+#define MALLOC_STRUCT(T)   (struct T *) MALLOC(sizeof(struct T))
+
+#define CALLOC_STRUCT(T)   (struct T *) CALLOC(1, sizeof(struct T))
+
+#define CALLOC_VARIANT_LENGTH_STRUCT(T,more_size)   ((struct T *) CALLOC(1, sizeof(struct T) + more_size))
+
+
+/**
+ * Return memory on given byte alignment
+ */
+static INLINE void *
+align_malloc(size_t bytes, uint alignment)
+{
+#if defined(HAVE_POSIX_MEMALIGN)
+   void *mem;
+   alignment = (alignment + (uint)sizeof(void*) - 1) & ~((uint)sizeof(void*) - 1);
+   if(posix_memalign(& mem, alignment, bytes) != 0)
+      return NULL;
+   return mem;
+#else
+   char *ptr, *buf;
+
+   assert( alignment > 0 );
+
+   ptr = (char *) MALLOC(bytes + alignment + sizeof(void *));
+   if (!ptr)
+      return NULL;
+
+   buf = (char *) align_pointer( ptr + sizeof(void *), alignment );
+   *(char **)(buf - sizeof(void *)) = ptr;
+
+   return buf;
+#endif /* defined(HAVE_POSIX_MEMALIGN) */
+}
+
+/**
+ * Free memory returned by align_malloc().
+ */
+static INLINE void
+align_free(void *ptr)
+{
+#if defined(HAVE_POSIX_MEMALIGN)
+   FREE(ptr);
+#else
+   void **cubbyHole = (void **) ((char *) ptr - sizeof(void *));
+   void *realAddr = *cubbyHole;
+   FREE(realAddr);
+#endif /* defined(HAVE_POSIX_MEMALIGN) */
+}
+
+
+/**
+ * Duplicate a block of memory.
+ */
+static INLINE void *
+mem_dup(const void *src, uint size)
+{
+   void *dup = MALLOC(size);
+   if (dup)
+      memcpy(dup, src, size);
+   return dup;
+}
+
+
+/**
+ * Number of elements in an array.
+ */
+#ifndef Elements
+#define Elements(x) (sizeof(x)/sizeof((x)[0]))
+#endif
+
+
+/**
+ * Offset of a field in a struct, in bytes.
+ */
+#define Offset(TYPE, MEMBER) ((unsigned)&(((TYPE *)NULL)->MEMBER))
+
+
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif /* U_MEMORY_H */
diff --git a/src/gallium/auxiliary/util/u_mm.c b/src/gallium/auxiliary/util/u_mm.c
new file mode 100644
index 0000000000..45ce257b5e
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_mm.c
@@ -0,0 +1,287 @@
+/**************************************************************************
+ *
+ * Copyright (C) 1999 Wittawat Yamwong
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * WITTAWAT YAMWONG, OR ANY OTHER CONTRIBUTORS BE LIABLE FOR ANY CLAIM, 
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE 
+ * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+
+#include "util/u_memory.h"
+#include "util/u_mm.h"
+
+
+void
+u_mmDumpMemInfo(const struct mem_block *heap)
+{
+   debug_printf("Memory heap %p:\n", (void *)heap);
+   if (heap == 0) {
+      debug_printf("  heap == 0\n");
+   } else {
+      const struct mem_block *p;
+
+      for(p = heap->next; p != heap; p = p->next) {
+	 debug_printf("  Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size,
+		 p->free ? 'F':'.',
+		 p->reserved ? 'R':'.');
+      }
+
+      debug_printf("\nFree list:\n");
+
+      for(p = heap->next_free; p != heap; p = p->next_free) {
+	 debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size,
+		 p->free ? 'F':'.',
+		 p->reserved ? 'R':'.');
+      }
+
+   }
+   debug_printf("End of memory blocks\n");
+}
+
+struct mem_block *
+u_mmInit(int ofs, int size)
+{
+   struct mem_block *heap, *block;
+  
+   if (size <= 0) 
+      return NULL;
+
+   heap = CALLOC_STRUCT(mem_block);
+   if (!heap) 
+      return NULL;
+   
+   block = CALLOC_STRUCT(mem_block);
+   if (!block) {
+      FREE(heap);
+      return NULL;
+   }
+
+   heap->next = block;
+   heap->prev = block;
+   heap->next_free = block;
+   heap->prev_free = block;
+
+   block->heap = heap;
+   block->next = heap;
+   block->prev = heap;
+   block->next_free = heap;
+   block->prev_free = heap;
+
+   block->ofs = ofs;
+   block->size = size;
+   block->free = 1;
+
+   return heap;
+}
+
+
+static struct mem_block *
+SliceBlock(struct mem_block *p, 
+           int startofs, int size, 
+           int reserved, int alignment)
+{
+   struct mem_block *newblock;
+
+   /* break left  [p, newblock, p->next], then p = newblock */
+   if (startofs > p->ofs) {
+      newblock = CALLOC_STRUCT(mem_block);
+      if (!newblock)
+	 return NULL;
+      newblock->ofs = startofs;
+      newblock->size = p->size - (startofs - p->ofs);
+      newblock->free = 1;
+      newblock->heap = p->heap;
+
+      newblock->next = p->next;
+      newblock->prev = p;
+      p->next->prev = newblock;
+      p->next = newblock;
+
+      newblock->next_free = p->next_free;
+      newblock->prev_free = p;
+      p->next_free->prev_free = newblock;
+      p->next_free = newblock;
+
+      p->size -= newblock->size;
+      p = newblock;
+   }
+
+   /* break right, also [p, newblock, p->next] */
+   if (size < p->size) {
+      newblock = CALLOC_STRUCT(mem_block);
+      if (!newblock)
+	 return NULL;
+      newblock->ofs = startofs + size;
+      newblock->size = p->size - size;
+      newblock->free = 1;
+      newblock->heap = p->heap;
+
+      newblock->next = p->next;
+      newblock->prev = p;
+      p->next->prev = newblock;
+      p->next = newblock;
+
+      newblock->next_free = p->next_free;
+      newblock->prev_free = p;
+      p->next_free->prev_free = newblock;
+      p->next_free = newblock;
+	 
+      p->size = size;
+   }
+
+   /* p = middle block */
+   p->free = 0;
+
+   /* Remove p from the free list: 
+    */
+   p->next_free->prev_free = p->prev_free;
+   p->prev_free->next_free = p->next_free;
+
+   p->next_free = 0;
+   p->prev_free = 0;
+
+   p->reserved = reserved;
+   return p;
+}
+
+
+struct mem_block *
+u_mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch)
+{
+   struct mem_block *p;
+   const int mask = (1 << align2)-1;
+   int startofs = 0;
+   int endofs;
+
+   assert(size >= 0);
+   assert(align2 >= 0);
+   assert(align2 <= 12); /* sanity check, 2^12 (4KB) enough? */
+
+   if (!heap || align2 < 0 || size <= 0)
+      return NULL;
+
+   for (p = heap->next_free; p != heap; p = p->next_free) {
+      assert(p->free);
+
+      startofs = (p->ofs + mask) & ~mask;
+      if ( startofs < startSearch ) {
+	 startofs = startSearch;
+      }
+      endofs = startofs+size;
+      if (endofs <= (p->ofs+p->size))
+	 break;
+   }
+
+   if (p == heap) 
+      return NULL;
+
+   assert(p->free);
+   p = SliceBlock(p,startofs,size,0,mask+1);
+
+   return p;
+}
+
+
+struct mem_block *
+u_mmFindBlock(struct mem_block *heap, int start)
+{
+   struct mem_block *p;
+
+   for (p = heap->next; p != heap; p = p->next) {
+      if (p->ofs == start) 
+	 return p;
+   }
+
+   return NULL;
+}
+
+
+static INLINE int
+Join2Blocks(struct mem_block *p)
+{
+   /* XXX there should be some assertions here */
+
+   /* NOTE: heap->free == 0 */
+
+   if (p->free && p->next->free) {
+      struct mem_block *q = p->next;
+
+      assert(p->ofs + p->size == q->ofs);
+      p->size += q->size;
+
+      p->next = q->next;
+      q->next->prev = p;
+
+      q->next_free->prev_free = q->prev_free; 
+      q->prev_free->next_free = q->next_free;
+     
+      FREE(q);
+      return 1;
+   }
+   return 0;
+}
+
+int
+u_mmFreeMem(struct mem_block *b)
+{
+   if (!b)
+      return 0;
+
+   if (b->free) {
+      debug_printf("block already free\n");
+      return -1;
+   }
+   if (b->reserved) {
+      debug_printf("block is reserved\n");
+      return -1;
+   }
+
+   b->free = 1;
+   b->next_free = b->heap->next_free;
+   b->prev_free = b->heap;
+   b->next_free->prev_free = b;
+   b->prev_free->next_free = b;
+
+   Join2Blocks(b);
+   if (b->prev != b->heap)
+      Join2Blocks(b->prev);
+
+   return 0;
+}
+
+
+void
+u_mmDestroy(struct mem_block *heap)
+{
+   struct mem_block *p;
+
+   if (!heap)
+      return;
+
+   for (p = heap->next; p != heap; ) {
+      struct mem_block *next = p->next;
+      FREE(p);
+      p = next;
+   }
+
+   FREE(heap);
+}
diff --git a/src/gallium/auxiliary/util/u_mm.h b/src/gallium/auxiliary/util/u_mm.h
new file mode 100644
index 0000000000..ce20e48763
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_mm.h
@@ -0,0 +1,91 @@
+/**************************************************************************
+ *
+ * Copyright (C) 1999 Wittawat Yamwong
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * KEITH WHITWELL, OR ANY OTHER CONTRIBUTORS BE LIABLE FOR ANY CLAIM, 
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE 
+ * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Memory manager code.  Primarily used by device drivers to manage texture
+ * heaps, etc.
+ */
+
+
+#ifndef _U_MM_H_
+#define _U_MM_H_
+
+
+struct mem_block {
+   struct mem_block *next, *prev;
+   struct mem_block *next_free, *prev_free;
+   struct mem_block *heap;
+   int ofs,size;
+   unsigned int free:1;
+   unsigned int reserved:1;
+};
+
+
+
+/** 
+ * input: total size in bytes
+ * return: a heap pointer if OK, NULL if error
+ */
+extern struct mem_block *u_mmInit(int ofs, int size);
+
+/**
+ * Allocate 'size' bytes with 2^align2 bytes alignment,
+ * restrict the search to free memory after 'startSearch'
+ * depth and back buffers should be in different 4mb banks
+ * to get better page hits if possible
+ * input:	size = size of block
+ *       	align2 = 2^align2 bytes alignment
+ *		startSearch = linear offset from start of heap to begin search
+ * return: pointer to the allocated block, 0 if error
+ */
+extern struct mem_block *u_mmAllocMem(struct mem_block *heap, int size, int align2, 
+                            int startSearch);
+
+/**
+ * Free block starts at offset
+ * input: pointer to a block
+ * return: 0 if OK, -1 if error
+ */
+extern int u_mmFreeMem(struct mem_block *b);
+
+/**
+ * Free block starts at offset
+ * input: pointer to a heap, start offset
+ * return: pointer to a block
+ */
+extern struct mem_block *u_mmFindBlock(struct mem_block *heap, int start);
+
+/**
+ * destroy MM
+ */
+extern void u_mmDestroy(struct mem_block *mmInit);
+
+/**
+ * For debuging purpose.
+ */
+extern void u_mmDumpMemInfo(const struct mem_block *mmInit);
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h
new file mode 100644
index 0000000000..e0e8aa8e9f
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_pack_color.h
@@ -0,0 +1,472 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Functions to produce packed colors/Z from floats.
+ */
+
+
+#ifndef U_PACK_COLOR_H
+#define U_PACK_COLOR_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "util/u_math.h"
+
+
+/**
+ * Pack ubyte R,G,B,A into dest pixel.
+ */
+static INLINE void
+util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a,
+                   enum pipe_format format, void *dest)
+{
+   switch (format) {
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (r << 24) | (g << 16) | (b << 8) | a;
+      }
+      return;
+   case PIPE_FORMAT_R8G8B8X8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (r << 24) | (g << 16) | (b << 8) | 0xff;
+      }
+      return;
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (a << 24) | (r << 16) | (g << 8) | b;
+      }
+      return;
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (0xff << 24) | (r << 16) | (g << 8) | b;
+      }
+      return;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (b << 24) | (g << 16) | (r << 8) | a;
+      }
+      return;
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (b << 24) | (g << 16) | (r << 8) | 0xff;
+      }
+      return;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      {
+         ushort *d = (ushort *) dest;
+         *d = ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3);
+      }
+      return;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      {
+         ushort *d = (ushort *) dest;
+         *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 2) | (b >> 3);
+      }
+      return;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      {
+         ushort *d = (ushort *) dest;
+         *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4);
+      }
+      return;
+   case PIPE_FORMAT_A8_UNORM:
+      {
+         ubyte *d = (ubyte *) dest;
+         *d = a;
+      }
+      return;
+   case PIPE_FORMAT_L8_UNORM:
+   case PIPE_FORMAT_I8_UNORM:
+      {
+         ubyte *d = (ubyte *) dest;
+         *d = r;
+      }
+      return;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      {
+         float *d = (float *) dest;
+         d[0] = (float)r / 255.0f;
+         d[1] = (float)g / 255.0f;
+         d[2] = (float)b / 255.0f;
+         d[3] = (float)a / 255.0f;
+      }
+      return;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      {
+         float *d = (float *) dest;
+         d[0] = (float)r / 255.0f;
+         d[1] = (float)g / 255.0f;
+         d[2] = (float)b / 255.0f;
+      }
+      return;
+
+   /* XXX lots more cases to add */
+   default:
+      debug_print_format("gallium: unhandled format in util_pack_color_ub()", format);
+      assert(0);
+   }
+}
+ 
+
+/**
+ * Unpack RGBA from a packed pixel, returning values as ubytes in [0,255].
+ */
+static INLINE void
+util_unpack_color_ub(enum pipe_format format, const void *src,
+                     ubyte *r, ubyte *g, ubyte *b, ubyte *a)
+{
+   switch (format) {
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      {
+         uint p = ((const uint *) src)[0];
+         *r = (ubyte) ((p >> 24) & 0xff);
+         *g = (ubyte) ((p >> 16) & 0xff);
+         *b = (ubyte) ((p >>  8) & 0xff);
+         *a = (ubyte) ((p >>  0) & 0xff);
+      }
+      return;
+   case PIPE_FORMAT_R8G8B8X8_UNORM:
+      {
+         uint p = ((const uint *) src)[0];
+         *r = (ubyte) ((p >> 24) & 0xff);
+         *g = (ubyte) ((p >> 16) & 0xff);
+         *b = (ubyte) ((p >>  8) & 0xff);
+         *a = (ubyte) 0xff;
+      }
+      return;
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      {
+         uint p = ((const uint *) src)[0];
+         *r = (ubyte) ((p >> 16) & 0xff);
+         *g = (ubyte) ((p >>  8) & 0xff);
+         *b = (ubyte) ((p >>  0) & 0xff);
+         *a = (ubyte) ((p >> 24) & 0xff);
+      }
+      return;
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+      {
+         uint p = ((const uint *) src)[0];
+         *r = (ubyte) ((p >> 16) & 0xff);
+         *g = (ubyte) ((p >>  8) & 0xff);
+         *b = (ubyte) ((p >>  0) & 0xff);
+         *a = (ubyte) 0xff;
+      }
+      return;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      {
+         uint p = ((const uint *) src)[0];
+         *r = (ubyte) ((p >>  8) & 0xff);
+         *g = (ubyte) ((p >> 16) & 0xff);
+         *b = (ubyte) ((p >> 24) & 0xff);
+         *a = (ubyte) ((p >>  0) & 0xff);
+      }
+      return;
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
+      {
+         uint p = ((const uint *) src)[0];
+         *r = (ubyte) ((p >>  8) & 0xff);
+         *g = (ubyte) ((p >> 16) & 0xff);
+         *b = (ubyte) ((p >> 24) & 0xff);
+         *a = (ubyte) 0xff;
+      }
+      return;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      {
+         ushort p = ((const ushort *) src)[0];
+         *r = (ubyte) (((p >> 8) & 0xf8) | ((p >> 13) & 0x7));
+         *g = (ubyte) (((p >> 3) & 0xfc) | ((p >>  9) & 0x3));
+         *b = (ubyte) (((p << 3) & 0xf8) | ((p >>  2) & 0x7));
+         *a = (ubyte) 0xff;
+      }
+      return;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      {
+         ushort p = ((const ushort *) src)[0];
+         *r = (ubyte) (((p >>  7) & 0xf8) | ((p >> 12) & 0x7));
+         *g = (ubyte) (((p >>  2) & 0xf8) | ((p >>  7) & 0x7));
+         *b = (ubyte) (((p <<  3) & 0xf8) | ((p >>  2) & 0x7));
+         *a = (ubyte) (0xff * (p >> 15));
+      }
+      return;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      {
+         ushort p = ((const ushort *) src)[0];
+         *r = (ubyte) (((p >> 4) & 0xf0) | ((p >>  8) & 0xf));
+         *g = (ubyte) (((p >> 0) & 0xf0) | ((p >>  4) & 0xf));
+         *b = (ubyte) (((p << 4) & 0xf0) | ((p >>  0) & 0xf));
+         *a = (ubyte) (((p >> 8) & 0xf0) | ((p >> 12) & 0xf));
+      }
+      return;
+   case PIPE_FORMAT_A8_UNORM:
+      {
+         ubyte p = ((const ubyte *) src)[0];
+         *r = *g = *b = (ubyte) 0xff;
+         *a = p;
+      }
+      return;
+   case PIPE_FORMAT_L8_UNORM:
+      {
+         ubyte p = ((const ubyte *) src)[0];
+         *r = *g = *b = p;
+         *a = (ubyte) 0xff;
+      }
+      return;
+   case PIPE_FORMAT_I8_UNORM:
+      {
+         ubyte p = ((const ubyte *) src)[0];
+         *r = *g = *b = *a = p;
+      }
+      return;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      {
+         const float *p = (const float *) src;
+         *r = float_to_ubyte(p[0]);
+         *g = float_to_ubyte(p[1]);
+         *b = float_to_ubyte(p[2]);
+         *a = float_to_ubyte(p[3]);
+      }
+      return;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      {
+         const float *p = (const float *) src;
+         *r = float_to_ubyte(p[0]);
+         *g = float_to_ubyte(p[1]);
+         *b = float_to_ubyte(p[2]);
+         *a = (ubyte) 0xff;
+      }
+      return;
+
+   case PIPE_FORMAT_R32G32_FLOAT:
+      {
+         const float *p = (const float *) src;
+         *r = float_to_ubyte(p[0]);
+         *g = float_to_ubyte(p[1]);
+         *b = *a = (ubyte) 0xff;
+      }
+      return;
+
+   case PIPE_FORMAT_R32_FLOAT:
+      {
+         const float *p = (const float *) src;
+         *r = float_to_ubyte(p[0]);
+         *g = *b = *a = (ubyte) 0xff;
+      }
+      return;
+
+   /* XXX lots more cases to add */
+   default:
+      debug_print_format("gallium: unhandled format in util_unpack_color_ub()",
+                         format);
+      assert(0);
+   }
+}
+ 
+
+
+/**
+ * Note rgba outside [0,1] will be clamped for int pixel formats.
+ */
+static INLINE void
+util_pack_color(const float rgba[4], enum pipe_format format, void *dest)
+{
+   ubyte r, g, b, a;
+
+   if (pf_size_x(format) <= 8) {
+      /* format uses 8-bit components or less */
+      r = float_to_ubyte(rgba[0]);
+      g = float_to_ubyte(rgba[1]);
+      b = float_to_ubyte(rgba[2]);
+      a = float_to_ubyte(rgba[3]);
+   }
+
+   switch (format) {
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (r << 24) | (g << 16) | (b << 8) | a;
+      }
+      return;
+   case PIPE_FORMAT_R8G8B8X8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (r << 24) | (g << 16) | (b << 8) | 0xff;
+      }
+      return;
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (a << 24) | (r << 16) | (g << 8) | b;
+      }
+      return;
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (0xff << 24) | (r << 16) | (g << 8) | b;
+      }
+      return;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (b << 24) | (g << 16) | (r << 8) | a;
+      }
+      return;
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
+      {
+         uint *d = (uint *) dest;
+         *d = (b << 24) | (g << 16) | (r << 8) | 0xff;
+      }
+      return;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      {
+         ushort *d = (ushort *) dest;
+         *d = ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3);
+      }
+      return;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      {
+         ushort *d = (ushort *) dest;
+         *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 2) | (b >> 3);
+      }
+      return;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      {
+         ushort *d = (ushort *) dest;
+         *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4);
+      }
+      return;
+   case PIPE_FORMAT_A8_UNORM:
+      {
+         ubyte *d = (ubyte *) dest;
+         *d = a;
+      }
+      return;
+   case PIPE_FORMAT_L8_UNORM:
+   case PIPE_FORMAT_I8_UNORM:
+      {
+         ubyte *d = (ubyte *) dest;
+         *d = r;
+      }
+      return;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      {
+         float *d = (float *) dest;
+         d[0] = rgba[0];
+         d[1] = rgba[1];
+         d[2] = rgba[2];
+         d[3] = rgba[3];
+      }
+      return;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      {
+         float *d = (float *) dest;
+         d[0] = rgba[0];
+         d[1] = rgba[1];
+         d[2] = rgba[2];
+      }
+      return;
+   /* XXX lots more cases to add */
+   default:
+      debug_print_format("gallium: unhandled format in util_pack_color()", format);
+      assert(0);
+   }
+}
+ 
+
+/**
+ * Note: it's assumed that z is in [0,1]
+ */
+static INLINE uint
+util_pack_z(enum pipe_format format, double z)
+{
+   if (z == 0.0)
+      return 0;
+
+   switch (format) {
+   case PIPE_FORMAT_Z16_UNORM:
+      if (z == 1.0)
+         return 0xffff;
+      return (uint) (z * 0xffff);
+   case PIPE_FORMAT_Z32_UNORM:
+      /* special-case to avoid overflow */
+      if (z == 1.0)
+         return 0xffffffff;
+      return (uint) (z * 0xffffffff);
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+      if (z == 1.0)
+         return 0xffffff;
+      return (uint) (z * 0xffffff);
+   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_Z24X8_UNORM:
+      if (z == 1.0)
+         return 0xffffff00;
+      return ((uint) (z * 0xffffff)) << 8;
+   default:
+      debug_print_format("gallium: unhandled format in util_pack_z()", format);
+      assert(0);
+      return 0;
+   }
+}
+
+
+/**
+ * Pack 4 ubytes into a 4-byte word
+ */
+static INLINE unsigned
+pack_ub4(ubyte b0, ubyte b1, ubyte b2, ubyte b3)
+{
+   return ((((unsigned int)b0) << 0) |
+	   (((unsigned int)b1) << 8) |
+	   (((unsigned int)b2) << 16) |
+	   (((unsigned int)b3) << 24));
+}
+
+
+/**
+ * Pack/convert 4 floats into one 4-byte word.
+ */
+static INLINE unsigned
+pack_ui32_float4(float a, float b, float c, float d)
+{
+   return pack_ub4( float_to_ubyte(a),
+		    float_to_ubyte(b),
+		    float_to_ubyte(c),
+		    float_to_ubyte(d) );
+}
+
+
+
+#endif /* U_PACK_COLOR_H */
diff --git a/src/gallium/auxiliary/util/u_pointer.h b/src/gallium/auxiliary/util/u_pointer.h
new file mode 100644
index 0000000000..e1af9f11cb
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_pointer.h
@@ -0,0 +1,107 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef U_POINTER_H
+#define U_POINTER_H
+
+#include "pipe/p_compiler.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+static INLINE intptr_t
+pointer_to_intptr( const void *p )
+{
+   union {
+      const void *p;
+      intptr_t i;
+   } pi;
+   pi.p = p;
+   return pi.i;
+}
+
+static INLINE void *
+intptr_to_pointer( intptr_t i )
+{
+   union {
+      void *p;
+      intptr_t i;
+   } pi;
+   pi.i = i;
+   return pi.p;
+}
+
+static INLINE uintptr_t
+pointer_to_uintptr( const void *ptr )
+{
+   union {
+      const void *p;
+      uintptr_t u;
+   } pu;
+   pu.p = ptr;
+   return pu.u;
+}
+
+static INLINE void *
+uintptr_to_pointer( uintptr_t u )
+{
+   union {
+      void *p;
+      uintptr_t u;
+   } pu;
+   pu.u = u;
+   return pu.p;
+}
+
+/**
+ * Return a pointer aligned to next multiple of N bytes.
+ */
+static INLINE void *
+align_pointer( const void *unaligned, uintptr_t alignment )
+{
+   uintptr_t aligned = (pointer_to_uintptr( unaligned ) + alignment - 1) & ~(alignment - 1);
+   return uintptr_to_pointer( aligned );
+}
+
+
+/**
+ * Return a pointer aligned to next multiple of 16 bytes.
+ */
+static INLINE void *
+align16( void *unaligned )
+{
+   return align_pointer( unaligned, 16 );
+}
+
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* U_POINTER_H */
diff --git a/src/gallium/auxiliary/util/u_prim.h b/src/gallium/auxiliary/util/u_prim.h
new file mode 100644
index 0000000000..d7c3995dbf
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_prim.h
@@ -0,0 +1,138 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#ifndef U_BLIT_H
+#define U_BLIT_H
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "pipe/p_defines.h"
+
+static INLINE boolean u_validate_pipe_prim( unsigned pipe_prim, unsigned nr )
+{
+   boolean ok = TRUE;
+
+   switch (pipe_prim) {
+   case PIPE_PRIM_POINTS:
+      ok = (nr >= 1);
+      break;
+   case PIPE_PRIM_LINES:
+      ok = (nr >= 2);
+      break;
+   case PIPE_PRIM_LINE_STRIP:
+   case PIPE_PRIM_LINE_LOOP:
+      ok = (nr >= 2);
+      break;
+   case PIPE_PRIM_TRIANGLES:
+      ok = (nr >= 3);
+      break;
+   case PIPE_PRIM_TRIANGLE_STRIP:
+   case PIPE_PRIM_TRIANGLE_FAN:
+   case PIPE_PRIM_POLYGON:
+      ok = (nr >= 3);
+      break;
+   case PIPE_PRIM_QUADS:
+      ok = (nr >= 4);
+      break;
+   case PIPE_PRIM_QUAD_STRIP:
+      ok = (nr >= 4);
+      break;
+   default:
+      ok = 0;
+      break;
+   }
+
+   return ok;
+}
+
+
+static INLINE boolean u_trim_pipe_prim( unsigned pipe_prim, unsigned *nr )
+{
+   boolean ok = TRUE;
+
+   switch (pipe_prim) {
+   case PIPE_PRIM_POINTS:
+      ok = (*nr >= 1);
+      break;
+   case PIPE_PRIM_LINES:
+      ok = (*nr >= 2);
+      *nr -= (*nr % 2);
+      break;
+   case PIPE_PRIM_LINE_STRIP:
+   case PIPE_PRIM_LINE_LOOP:
+      ok = (*nr >= 2);
+      break;
+   case PIPE_PRIM_TRIANGLES:
+      ok = (*nr >= 3);
+      *nr -= (*nr % 3);
+      break;
+   case PIPE_PRIM_TRIANGLE_STRIP:
+   case PIPE_PRIM_TRIANGLE_FAN:
+   case PIPE_PRIM_POLYGON:
+      ok = (*nr >= 3);
+      break;
+   case PIPE_PRIM_QUADS:
+      ok = (*nr >= 4);
+      *nr -= (*nr % 4);
+      break;
+   case PIPE_PRIM_QUAD_STRIP:
+      ok = (*nr >= 4);
+      *nr -= (*nr % 2);
+      break;
+   default:
+      ok = 0;
+      break;
+   }
+
+   if (!ok)
+      *nr = 0;
+
+   return ok;
+}
+
+
+static INLINE boolean u_reduced_prim( unsigned pipe_prim )
+{
+   switch (pipe_prim) {
+   case PIPE_PRIM_POINTS:
+      return PIPE_PRIM_POINTS;
+
+   case PIPE_PRIM_LINES:
+   case PIPE_PRIM_LINE_STRIP:
+   case PIPE_PRIM_LINE_LOOP:
+      return PIPE_PRIM_LINES;
+
+   default:
+      return PIPE_PRIM_TRIANGLES;
+   }
+}
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c
new file mode 100644
index 0000000000..fe81a685be
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_rect.c
@@ -0,0 +1,329 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Rectangle-related helper functions.
+ */
+
+
+#include "pipe/p_defines.h"
+#include "pipe/p_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "util/u_rect.h"
+
+
+/**
+ * Copy 2D rect from one place to another.
+ * Position and sizes are in pixels.
+ * src_pitch may be negative to do vertical flip of pixels from source.
+ */
+void
+pipe_copy_rect(ubyte * dst,
+               const struct pipe_format_block *block,
+               unsigned dst_stride,
+               unsigned dst_x,
+               unsigned dst_y,
+               unsigned width,
+               unsigned height,
+               const ubyte * src,
+               int src_stride,
+               unsigned src_x, 
+               int src_y)
+{
+   unsigned i;
+   int src_stride_pos = src_stride < 0 ? -src_stride : src_stride;
+
+   assert(block->size > 0);
+   assert(block->width > 0);
+   assert(block->height > 0);
+   assert(src_x >= 0);
+   assert(src_y >= 0);
+   assert(dst_x >= 0);
+   assert(dst_y >= 0);
+
+   dst_x /= block->width;
+   dst_y /= block->height;
+   width = (width + block->width - 1)/block->width;
+   height = (height + block->height - 1)/block->height;
+   src_x /= block->width;
+   src_y /= block->height;
+   
+   dst += dst_x * block->size;
+   src += src_x * block->size;
+   dst += dst_y * dst_stride;
+   src += src_y * src_stride_pos;
+   width *= block->size;
+
+   if (width == dst_stride && width == src_stride)
+      memcpy(dst, src, height * width);
+   else {
+      for (i = 0; i < height; i++) {
+         memcpy(dst, src, width);
+         dst += dst_stride;
+         src += src_stride;
+      }
+   }
+}
+
+void
+pipe_fill_rect(ubyte * dst,
+               const struct pipe_format_block *block,
+               unsigned dst_stride,
+               unsigned dst_x,
+               unsigned dst_y,
+               unsigned width,
+               unsigned height,
+               uint32_t value)
+{
+   unsigned i, j;
+   unsigned width_size;
+
+   assert(block->size > 0);
+   assert(block->width > 0);
+   assert(block->height > 0);
+   assert(dst_x >= 0);
+   assert(dst_y >= 0);
+
+   dst_x /= block->width;
+   dst_y /= block->height;
+   width = (width + block->width - 1)/block->width;
+   height = (height + block->height - 1)/block->height;
+   
+   dst += dst_x * block->size;
+   dst += dst_y * dst_stride;
+   width_size = width * block->size;
+   
+   switch (block->size) {
+   case 1:
+      if(dst_stride == width_size)
+	 memset(dst, (ubyte) value, height * width_size);
+      else {
+	 for (i = 0; i < height; i++) {
+	    memset(dst, (ubyte) value, width_size);
+	    dst += dst_stride;
+	 }
+      }
+      break;
+   case 2:
+      for (i = 0; i < height; i++) {
+	 uint16_t *row = (uint16_t *)dst;
+	 for (j = 0; j < width; j++)
+	    *row++ = (uint16_t) value;
+	 dst += dst_stride;
+      }
+      break;
+   case 4:
+      for (i = 0; i < height; i++) {
+	 uint32_t *row = (uint32_t *)dst;
+	 for (j = 0; j < width; j++)
+	    *row++ = value;
+	 dst += dst_stride;
+      }
+      break;
+   default:
+	 assert(0);
+	 break;
+   }
+}
+
+
+
+/**
+ * Fallback function for pipe->surface_copy().
+ * Note: (X,Y)=(0,0) is always the upper-left corner.
+ * if do_flip, flip the image vertically on its way from src rect to dst rect.
+ * XXX should probably put this in new u_surface.c file...
+ */
+void
+util_surface_copy(struct pipe_context *pipe,
+                  boolean do_flip,
+                  struct pipe_surface *dst,
+                  unsigned dst_x, unsigned dst_y,
+                  struct pipe_surface *src,
+                  unsigned src_x, unsigned src_y, 
+                  unsigned w, unsigned h)
+{
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *new_src = NULL, *new_dst = NULL;
+   void *dst_map;
+   const void *src_map;
+
+   assert(dst->block.size == src->block.size);
+   assert(dst->block.width == src->block.width);
+   assert(dst->block.height == src->block.height);
+
+   if ((src->usage & PIPE_BUFFER_USAGE_CPU_READ) == 0) {
+      /* Need to create new src surface which is CPU readable */
+      assert(src->texture);
+      if (!src->texture)
+         return;
+      new_src = screen->get_tex_surface(screen,
+                                        src->texture,
+                                        src->face,
+                                        src->level,
+                                        src->zslice,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+      src = new_src;
+   }
+
+   if ((dst->usage & PIPE_BUFFER_USAGE_CPU_WRITE) == 0) {
+      /* Need to create new dst surface which is CPU writable */
+      assert(dst->texture);
+      if (!dst->texture)
+         return;
+      new_dst = screen->get_tex_surface(screen,
+                                        dst->texture,
+                                        dst->face,
+                                        dst->level,
+                                        dst->zslice,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+      dst = new_dst;
+   }
+
+   src_map = pipe->screen->surface_map(screen,
+                                       src, PIPE_BUFFER_USAGE_CPU_READ);
+   dst_map = pipe->screen->surface_map(screen,
+                                       dst, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   assert(src_map);
+   assert(dst_map);
+
+   if (src_map && dst_map) {
+      /* If do_flip, invert src_y position and pass negative src stride */
+      pipe_copy_rect(dst_map,
+                     &dst->block,
+                     dst->stride,
+                     dst_x, dst_y,
+                     w, h,
+                     src_map,
+                     do_flip ? -(int) src->stride : src->stride,
+                     src_x,
+                     do_flip ? src_y + h - 1 : src_y);
+   }
+
+   pipe->screen->surface_unmap(pipe->screen, src);
+   pipe->screen->surface_unmap(pipe->screen, dst);
+
+   if (new_src)
+      screen->tex_surface_release(screen, &new_src);
+   if (new_dst)
+      screen->tex_surface_release(screen, &new_dst);
+}
+
+
+
+static void *
+get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y)
+{
+   return (char *)dst_map
+      + y / dst->block.height * dst->stride
+      + x / dst->block.width * dst->block.size;
+}
+
+
+#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8))
+
+
+/**
+ * Fallback for pipe->surface_fill() function.
+ * XXX should probably put this in new u_surface.c file...
+ */
+void
+util_surface_fill(struct pipe_context *pipe,
+                  struct pipe_surface *dst,
+                  unsigned dstx, unsigned dsty,
+                  unsigned width, unsigned height, unsigned value)
+{
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *new_dst = NULL;
+   void *dst_map;
+
+   if ((dst->usage & PIPE_BUFFER_USAGE_CPU_WRITE) == 0) {
+      /* Need to create new dst surface which is CPU writable */
+      assert(dst->texture);
+      if (!dst->texture)
+         return;
+      new_dst = screen->get_tex_surface(screen,
+                                        dst->texture,
+                                        dst->face,
+                                        dst->level,
+                                        dst->zslice,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+      dst = new_dst;
+   }
+
+   dst_map = pipe->screen->surface_map(screen,
+                                       dst, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   assert(dst_map);
+
+   if (dst_map) {
+      assert(dst->stride > 0);
+
+      switch (dst->block.size) {
+      case 1:
+      case 2:
+      case 4:
+         pipe_fill_rect(dst_map, &dst->block, dst->stride,
+                        dstx, dsty, width, height, value);
+         break;
+      case 8:
+         {
+            /* expand the 4-byte clear value to an 8-byte value */
+            ushort *row = (ushort *) get_pointer(dst, dst_map, dstx, dsty);
+            ushort val0 = UBYTE_TO_USHORT((value >>  0) & 0xff);
+            ushort val1 = UBYTE_TO_USHORT((value >>  8) & 0xff);
+            ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff);
+            ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff);
+            unsigned i, j;
+            val0 = (val0 << 8) | val0;
+            val1 = (val1 << 8) | val1;
+            val2 = (val2 << 8) | val2;
+            val3 = (val3 << 8) | val3;
+            for (i = 0; i < height; i++) {
+               for (j = 0; j < width; j++) {
+                  row[j*4+0] = val0;
+                  row[j*4+1] = val1;
+                  row[j*4+2] = val2;
+                  row[j*4+3] = val3;
+               }
+               row += dst->stride/2;
+            }
+         }
+         break;
+      default:
+         assert(0);
+         break;
+      }
+   }
+
+   pipe->screen->surface_unmap(pipe->screen, dst);
+
+   if (new_dst)
+      screen->tex_surface_release(screen, &new_dst);
+}
diff --git a/src/gallium/auxiliary/util/u_rect.h b/src/gallium/auxiliary/util/u_rect.h
new file mode 100644
index 0000000000..59e842e16d
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_rect.h
@@ -0,0 +1,72 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Pipe copy/fill rect helpers.
+ */
+
+
+#ifndef U_RECT_H
+#define U_RECT_H
+
+
+#include "pipe/p_format.h"
+
+struct pipe_context;
+struct pipe_surface;
+
+
+extern void
+pipe_copy_rect(ubyte * dst, const struct pipe_format_block *block,
+               unsigned dst_stride, unsigned dst_x, unsigned dst_y,
+               unsigned width, unsigned height, const ubyte * src,
+               int src_stride, unsigned src_x, int src_y);
+
+extern void
+pipe_fill_rect(ubyte * dst, const struct pipe_format_block *block,
+               unsigned dst_stride, unsigned dst_x, unsigned dst_y,
+               unsigned width, unsigned height, uint32_t value);
+
+
+extern void
+util_surface_copy(struct pipe_context *pipe,
+                  boolean do_flip,
+                  struct pipe_surface *dst,
+                  unsigned dst_x, unsigned dst_y,
+                  struct pipe_surface *src,
+                  unsigned src_x, unsigned src_y, 
+                  unsigned w, unsigned h);
+
+extern void
+util_surface_fill(struct pipe_context *pipe,
+                  struct pipe_surface *dst,
+                  unsigned dstx, unsigned dsty,
+                  unsigned width, unsigned height, unsigned value);
+
+
+#endif /* U_RECT_H */
diff --git a/src/gallium/auxiliary/util/u_simple_list.h b/src/gallium/auxiliary/util/u_simple_list.h
new file mode 100644
index 0000000000..f5f43b0faa
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_simple_list.h
@@ -0,0 +1,197 @@
+/**
+ * \file simple_list.h
+ * Simple macros for type-safe, intrusive lists.
+ *
+ *  Intended to work with a list sentinal which is created as an empty
+ *  list.  Insert & delete are O(1).
+ *  
+ * \author
+ *  (C) 1997, Keith Whitwell
+ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2001  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef _U_SIMPLE_LIST_H_
+#define _U_SIMPLE_LIST_H_
+
+/**
+ * Remove an element from list.
+ *
+ * \param elem element to remove.
+ */
+#define remove_from_list(elem)			\
+do {						\
+   (elem)->next->prev = (elem)->prev;		\
+   (elem)->prev->next = (elem)->next;		\
+} while (0)
+
+/**
+ * Insert an element to the list head.
+ *
+ * \param list list.
+ * \param elem element to insert.
+ */
+#define insert_at_head(list, elem)		\
+do {						\
+   (elem)->prev = list;				\
+   (elem)->next = (list)->next;			\
+   (list)->next->prev = elem;			\
+   (list)->next = elem;				\
+} while(0)
+
+/**
+ * Insert an element to the list tail.
+ *
+ * \param list list.
+ * \param elem element to insert.
+ */
+#define insert_at_tail(list, elem)		\
+do {						\
+   (elem)->next = list;				\
+   (elem)->prev = (list)->prev;			\
+   (list)->prev->next = elem;			\
+   (list)->prev = elem;				\
+} while(0)
+
+/**
+ * Move an element to the list head.
+ *
+ * \param list list.
+ * \param elem element to move.
+ */
+#define move_to_head(list, elem)		\
+do {						\
+   remove_from_list(elem);			\
+   insert_at_head(list, elem);			\
+} while (0)
+
+/**
+ * Move an element to the list tail.
+ *
+ * \param list list.
+ * \param elem element to move.
+ */
+#define move_to_tail(list, elem)		\
+do {						\
+   remove_from_list(elem);			\
+   insert_at_tail(list, elem);			\
+} while (0)
+
+/**
+ * Make a empty list empty.
+ *
+ * \param sentinal list (sentinal element).
+ */
+#define make_empty_list(sentinal)		\
+do {						\
+   (sentinal)->next = sentinal;			\
+   (sentinal)->prev = sentinal;			\
+} while (0)
+
+/**
+ * Get list first element.
+ *
+ * \param list list.
+ *
+ * \return pointer to first element.
+ */
+#define first_elem(list)       ((list)->next)
+
+/**
+ * Get list last element.
+ *
+ * \param list list.
+ *
+ * \return pointer to last element.
+ */
+#define last_elem(list)        ((list)->prev)
+
+/**
+ * Get next element.
+ *
+ * \param elem element.
+ *
+ * \return pointer to next element.
+ */
+#define next_elem(elem)        ((elem)->next)
+
+/**
+ * Get previous element.
+ *
+ * \param elem element.
+ *
+ * \return pointer to previous element.
+ */
+#define prev_elem(elem)        ((elem)->prev)
+
+/**
+ * Test whether element is at end of the list.
+ * 
+ * \param list list.
+ * \param elem element.
+ * 
+ * \return non-zero if element is at end of list, or zero otherwise.
+ */
+#define at_end(list, elem)     ((elem) == (list))
+
+/**
+ * Test if a list is empty.
+ * 
+ * \param list list.
+ * 
+ * \return non-zero if list empty, or zero otherwise.
+ */
+#define is_empty_list(list)    ((list)->next == (list))
+
+/**
+ * Walk through the elements of a list.
+ *
+ * \param ptr pointer to the current element.
+ * \param list list.
+ *
+ * \note It should be followed by a { } block or a single statement, as in a \c
+ * for loop.
+ */
+#define foreach(ptr, list)     \
+        for( ptr=(list)->next ;  ptr!=list ;  ptr=(ptr)->next )
+
+/**
+ * Walk through the elements of a list.
+ *
+ * Same as #foreach but lets you unlink the current value during a list
+ * traversal.  Useful for freeing a list, element by element.
+ * 
+ * \param ptr pointer to the current element.
+ * \param t temporary pointer.
+ * \param list list.
+ *
+ * \note It should be followed by a { } block or a single statement, as in a \c
+ * for loop.
+ */
+#define foreach_s(ptr, t, list)   \
+        for(ptr=(list)->next,t=(ptr)->next; list != ptr; ptr=t, t=(t)->next)
+
+#endif /* _U_SIMPLE_LIST_H_ */
diff --git a/src/gallium/auxiliary/util/u_simple_screen.c b/src/gallium/auxiliary/util/u_simple_screen.c
new file mode 100644
index 0000000000..089bbbc48a
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_simple_screen.c
@@ -0,0 +1,143 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "u_simple_screen.h"
+
+#include "pipe/p_screen.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+
+static struct pipe_buffer *
+pass_buffer_create(struct pipe_screen *screen,
+                   unsigned alignment,
+                   unsigned usage,
+                   unsigned size)
+{
+   return screen->winsys->buffer_create(screen->winsys,
+                                        alignment, usage, size);
+}
+
+static struct pipe_buffer *
+pass_user_buffer_create(struct pipe_screen *screen,
+                        void *ptr,
+                        unsigned bytes)
+{
+   return screen->winsys->user_buffer_create(screen->winsys,
+                                             ptr, bytes);
+}
+
+static struct pipe_buffer *
+pass_surface_buffer_create(struct pipe_screen *screen,
+                           unsigned width, unsigned height,
+                           enum pipe_format format,
+                           unsigned usage,
+                           unsigned *stride)
+{
+   return screen->winsys->surface_buffer_create(screen->winsys,
+                                                width, height,
+                                                format, usage, stride);
+}
+
+static void *
+pass_buffer_map(struct pipe_screen *screen,
+                struct pipe_buffer *buf,
+                unsigned usage)
+{
+   return screen->winsys->buffer_map(screen->winsys,
+                                     buf, usage);
+}
+
+static void
+pass_buffer_unmap(struct pipe_screen *screen,
+                  struct pipe_buffer *buf)
+{
+   screen->winsys->buffer_unmap(screen->winsys, buf);
+}
+
+static void
+pass_buffer_destroy(struct pipe_screen *screen,
+                    struct pipe_buffer *buf)
+{
+   screen->winsys->buffer_destroy(screen->winsys, buf);
+}
+
+
+static void
+pass_flush_frontbuffer(struct pipe_screen *screen,
+                       struct pipe_surface *surf,
+                       void *context_private)
+{
+   screen->winsys->flush_frontbuffer(screen->winsys,
+                                     surf, context_private);
+}
+
+static void
+pass_fence_reference(struct pipe_screen *screen,
+                     struct pipe_fence_handle **ptr,
+                     struct pipe_fence_handle *fence)
+{
+   screen->winsys->fence_reference(screen->winsys,
+                                   ptr, fence);
+}
+
+static int
+pass_fence_signalled(struct pipe_screen *screen,
+                     struct pipe_fence_handle *fence,
+                     unsigned flag)
+{
+   return screen->winsys->fence_signalled(screen->winsys,
+                                          fence, flag);
+}
+
+static int
+pass_fence_finish(struct pipe_screen *screen,
+                  struct pipe_fence_handle *fence,
+                  unsigned flag)
+{
+   return screen->winsys->fence_finish(screen->winsys,
+                                       fence, flag);
+}
+
+void u_simple_screen_init(struct pipe_screen *screen)
+{
+   screen->buffer_create = pass_buffer_create;
+   screen->user_buffer_create = pass_user_buffer_create;
+   screen->surface_buffer_create = pass_surface_buffer_create;
+
+   screen->buffer_map = pass_buffer_map;
+   screen->buffer_unmap = pass_buffer_unmap;
+   screen->buffer_destroy = pass_buffer_destroy;
+   screen->flush_frontbuffer = pass_flush_frontbuffer;
+   screen->fence_reference = pass_fence_reference;
+   screen->fence_signalled = pass_fence_signalled;
+   screen->fence_finish = pass_fence_finish;
+}
+
+const char* u_simple_screen_winsys_name(struct pipe_screen *screen)
+{
+   return screen->winsys->get_name(screen->winsys);
+}
diff --git a/src/gallium/auxiliary/util/u_simple_screen.h b/src/gallium/auxiliary/util/u_simple_screen.h
new file mode 100644
index 0000000000..6612a8a7c0
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_simple_screen.h
@@ -0,0 +1,47 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef U_SIMPLE_SCREEN_H
+#define U_SIMPLE_SCREEN_H
+
+struct pipe_screen;
+struct pipe_winsys;
+
+/**
+ * The following function initializes a simple passthrough screen.
+ *
+ * All the relevant screen function pointers will forwarded to the
+ * winsys.
+ */
+void u_simple_screen_init(struct pipe_screen *screen);
+
+/**
+ * Returns the name of the winsys associated with this screen.
+ */
+const char* u_simple_screen_winsys_name(struct pipe_screen *screen);
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c
new file mode 100644
index 0000000000..706155e99a
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_simple_shaders.c
@@ -0,0 +1,361 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Simple vertex/fragment shader generators.
+ *  
+ * @author Brian Paul
+ */
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_memory.h"
+#include "util/u_simple_shaders.h"
+
+#include "tgsi/tgsi_build.h"
+#include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_parse.h"
+
+
+
+/**
+ * Make simple vertex pass-through shader.
+ */
+void *
+util_make_vertex_passthrough_shader(struct pipe_context *pipe,
+                                    uint num_attribs,
+                                    const uint *semantic_names,
+                                    const uint *semantic_indexes,
+                                    struct pipe_shader_state *shader)
+                                    
+{
+   uint maxTokens = 100;
+   struct tgsi_token *tokens;
+   struct tgsi_header *header;
+   struct tgsi_processor *processor;
+   struct tgsi_full_declaration decl;
+   struct tgsi_full_instruction inst;
+   const uint procType = TGSI_PROCESSOR_VERTEX;
+   uint ti, i;
+
+   tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0]));
+
+   /* shader header
+    */
+   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
+
+   header = (struct tgsi_header *) &tokens[1];
+   *header = tgsi_build_header();
+
+   processor = (struct tgsi_processor *) &tokens[2];
+   *processor = tgsi_build_processor( procType, header );
+
+   ti = 3;
+
+   /* declare inputs */
+   for (i = 0; i < num_attribs; i++) {
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_INPUT;
+
+      decl.Declaration.Semantic = 1;
+      decl.Semantic.SemanticName = semantic_names[i];
+      decl.Semantic.SemanticIndex = semantic_indexes[i];
+
+      decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = i;
+      ti += tgsi_build_full_declaration(&decl,
+                                        &tokens[ti],
+                                        header,
+                                        maxTokens - ti);
+   }
+
+   /* declare outputs */
+   for (i = 0; i < num_attribs; i++) {
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_OUTPUT;
+      decl.Declaration.Semantic = 1;
+      decl.Semantic.SemanticName = semantic_names[i];
+      decl.Semantic.SemanticIndex = semantic_indexes[i];
+      decl.DeclarationRange.First = 
+         decl.DeclarationRange.Last = i;
+      ti += tgsi_build_full_declaration(&decl,
+                                        &tokens[ti],
+                                        header,
+                                        maxTokens - ti);
+   }
+
+   /* emit MOV instructions */
+   for (i = 0; i < num_attribs; i++) {
+      /* MOVE out[i], in[i]; */
+      inst = tgsi_default_full_instruction();
+      inst.Instruction.Opcode = TGSI_OPCODE_MOV;
+      inst.Instruction.NumDstRegs = 1;
+      inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+      inst.FullDstRegisters[0].DstRegister.Index = i;
+      inst.Instruction.NumSrcRegs = 1;
+      inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+      inst.FullSrcRegisters[0].SrcRegister.Index = i;
+      ti += tgsi_build_full_instruction(&inst,
+                                        &tokens[ti],
+                                        header,
+                                        maxTokens - ti );
+   }
+
+   /* END instruction */
+   inst = tgsi_default_full_instruction();
+   inst.Instruction.Opcode = TGSI_OPCODE_END;
+   inst.Instruction.NumDstRegs = 0;
+   inst.Instruction.NumSrcRegs = 0;
+   ti += tgsi_build_full_instruction(&inst,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti );
+
+#if 0 /*debug*/
+   tgsi_dump(tokens, 0);
+#endif
+
+   shader->tokens = tokens;
+   /*shader->num_tokens = ti;*/
+
+   return pipe->create_vs_state(pipe, shader);
+}
+
+
+
+
+/**
+ * Make simple fragment texture shader:
+ *  TEX OUT[0], IN[0], SAMP[0], 2D;
+ *  END;
+ */
+void *
+util_make_fragment_tex_shader(struct pipe_context *pipe,
+                              struct pipe_shader_state *shader)
+{
+   uint maxTokens = 100;
+   struct tgsi_token *tokens;
+   struct tgsi_header *header;
+   struct tgsi_processor *processor;
+   struct tgsi_full_declaration decl;
+   struct tgsi_full_instruction inst;
+   const uint procType = TGSI_PROCESSOR_FRAGMENT;
+   uint ti;
+
+   tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0]));
+
+   /* shader header
+    */
+   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
+
+   header = (struct tgsi_header *) &tokens[1];
+   *header = tgsi_build_header();
+
+   processor = (struct tgsi_processor *) &tokens[2];
+   *processor = tgsi_build_processor( procType, header );
+
+   ti = 3;
+
+   /* declare TEX[0] input */
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_INPUT;
+   /* XXX this could be linear... */
+   decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE;
+   decl.Declaration.Semantic = 1;
+   decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
+   decl.Semantic.SemanticIndex = 0;
+   decl.DeclarationRange.First = 
+   decl.DeclarationRange.Last = 0;
+   ti += tgsi_build_full_declaration(&decl,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti);
+
+   /* declare color[0] output */
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_OUTPUT;
+   decl.Declaration.Semantic = 1;
+   decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
+   decl.Semantic.SemanticIndex = 0;
+   decl.DeclarationRange.First = 
+   decl.DeclarationRange.Last = 0;
+   ti += tgsi_build_full_declaration(&decl,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti);
+
+   /* declare sampler */
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_SAMPLER;
+   decl.DeclarationRange.First = 
+   decl.DeclarationRange.Last = 0;
+   ti += tgsi_build_full_declaration(&decl,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti);
+
+   /* TEX instruction */
+   inst = tgsi_default_full_instruction();
+   inst.Instruction.Opcode = TGSI_OPCODE_TEX;
+   inst.Instruction.NumDstRegs = 1;
+   inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+   inst.FullDstRegisters[0].DstRegister.Index = 0;
+   inst.Instruction.NumSrcRegs = 2;
+   inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
+   inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+   inst.FullSrcRegisters[0].SrcRegister.Index = 0;
+   inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+   inst.FullSrcRegisters[1].SrcRegister.Index = 0;
+   ti += tgsi_build_full_instruction(&inst,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti );
+
+   /* END instruction */
+   inst = tgsi_default_full_instruction();
+   inst.Instruction.Opcode = TGSI_OPCODE_END;
+   inst.Instruction.NumDstRegs = 0;
+   inst.Instruction.NumSrcRegs = 0;
+   ti += tgsi_build_full_instruction(&inst,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti );
+
+#if 0 /*debug*/
+   tgsi_dump(tokens, 0);
+#endif
+
+   shader->tokens = tokens;
+   /*shader->num_tokens = ti;*/
+
+   return pipe->create_fs_state(pipe, shader);
+}
+
+
+
+
+
+/**
+ * Make simple fragment color pass-through shader.
+ */
+void *
+util_make_fragment_passthrough_shader(struct pipe_context *pipe,
+                                      struct pipe_shader_state *shader)
+{
+   uint maxTokens = 40;
+   struct tgsi_token *tokens;
+   struct tgsi_header *header;
+   struct tgsi_processor *processor;
+   struct tgsi_full_declaration decl;
+   struct tgsi_full_instruction inst;
+   const uint procType = TGSI_PROCESSOR_FRAGMENT;
+   uint ti;
+
+   tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0]));
+
+   /* shader header
+    */
+   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
+
+   header = (struct tgsi_header *) &tokens[1];
+   *header = tgsi_build_header();
+
+   processor = (struct tgsi_processor *) &tokens[2];
+   *processor = tgsi_build_processor( procType, header );
+
+   ti = 3;
+
+   /* declare input */
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_INPUT;
+   decl.Declaration.Semantic = 1;
+   decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
+   decl.Semantic.SemanticIndex = 0;
+   decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = 0;
+   ti += tgsi_build_full_declaration(&decl,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti);
+
+   /* declare output */
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_OUTPUT;
+   decl.Declaration.Semantic = 1;
+   decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
+   decl.Semantic.SemanticIndex = 0;
+   decl.DeclarationRange.First = 
+      decl.DeclarationRange.Last = 0;
+   ti += tgsi_build_full_declaration(&decl,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti);
+
+
+   /* MOVE out[0], in[0]; */
+   inst = tgsi_default_full_instruction();
+   inst.Instruction.Opcode = TGSI_OPCODE_MOV;
+   inst.Instruction.NumDstRegs = 1;
+   inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+   inst.FullDstRegisters[0].DstRegister.Index = 0;
+   inst.Instruction.NumSrcRegs = 1;
+   inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+   inst.FullSrcRegisters[0].SrcRegister.Index = 0;
+   ti += tgsi_build_full_instruction(&inst,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti );
+
+   /* END instruction */
+   inst = tgsi_default_full_instruction();
+   inst.Instruction.Opcode = TGSI_OPCODE_END;
+   inst.Instruction.NumDstRegs = 0;
+   inst.Instruction.NumSrcRegs = 0;
+   ti += tgsi_build_full_instruction(&inst,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti );
+
+   assert(ti < maxTokens);
+
+#if 0 /*debug*/
+   tgsi_dump(tokens, 0);
+#endif
+
+   shader->tokens = tokens;
+   /*shader->num_tokens = ti;*/
+
+   return pipe->create_fs_state(pipe, shader);
+}
+
diff --git a/src/gallium/auxiliary/util/u_simple_shaders.h b/src/gallium/auxiliary/util/u_simple_shaders.h
new file mode 100644
index 0000000000..8ca4977d71
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_simple_shaders.h
@@ -0,0 +1,68 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#ifndef U_SIMPLE_SHADERS_H
+#define U_SIMPLE_SHADERS_H
+
+
+#include "pipe/p_compiler.h"
+
+
+struct pipe_context;
+struct pipe_shader_state;
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+extern void *
+util_make_vertex_passthrough_shader(struct pipe_context *pipe,
+                                    uint num_attribs,
+                                    const uint *semantic_names,
+                                    const uint *semantic_indexes,
+                                    struct pipe_shader_state *shader);
+
+
+extern void *
+util_make_fragment_tex_shader(struct pipe_context *pipe,
+                              struct pipe_shader_state *shader);
+
+
+extern void *
+util_make_fragment_passthrough_shader(struct pipe_context *pipe,
+                                      struct pipe_shader_state *shader);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_snprintf.c b/src/gallium/auxiliary/util/u_snprintf.c
new file mode 100644
index 0000000000..0d54299b28
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_snprintf.c
@@ -0,0 +1,1497 @@
+/*
+ * Copyright (c) 1995 Patrick Powell.
+ *
+ * This code is based on code written by Patrick Powell <papowell@astart.com>.
+ * It may be used for any purpose as long as this notice remains intact on all
+ * source code distributions.
+ */
+
+/*
+ * Copyright (c) 2008 Holger Weiss.
+ *
+ * This version of the code is maintained by Holger Weiss <holger@jhweiss.de>.
+ * My changes to the code may freely be used, modified and/or redistributed for
+ * any purpose.  It would be nice if additions and fixes to this file (including
+ * trivial code cleanups) would be sent back in order to let me include them in
+ * the version available at <http://www.jhweiss.de/software/snprintf.html>.
+ * However, this is not a requirement for using or redistributing (possibly
+ * modified) versions of this file, nor is leaving this notice intact mandatory.
+ */
+
+/*
+ * History
+ *
+ * 2008-01-20 Holger Weiss <holger@jhweiss.de> for C99-snprintf 1.1:
+ *
+ * 	Fixed the detection of infinite floating point values on IRIX (and
+ * 	possibly other systems) and applied another few minor cleanups.
+ *
+ * 2008-01-06 Holger Weiss <holger@jhweiss.de> for C99-snprintf 1.0:
+ *
+ * 	Added a lot of new features, fixed many bugs, and incorporated various
+ * 	improvements done by Andrew Tridgell <tridge@samba.org>, Russ Allbery
+ * 	<rra@stanford.edu>, Hrvoje Niksic <hniksic@xemacs.org>, Damien Miller
+ * 	<djm@mindrot.org>, and others for the Samba, INN, Wget, and OpenSSH
+ * 	projects.  The additions include: support the "e", "E", "g", "G", and
+ * 	"F" conversion specifiers (and use conversion style "f" or "F" for the
+ * 	still unsupported "a" and "A" specifiers); support the "hh", "ll", "j",
+ * 	"t", and "z" length modifiers; support the "#" flag and the (non-C99)
+ * 	"'" flag; use localeconv(3) (if available) to get both the current
+ * 	locale's decimal point character and the separator between groups of
+ * 	digits; fix the handling of various corner cases of field width and
+ * 	precision specifications; fix various floating point conversion bugs;
+ * 	handle infinite and NaN floating point values; don't attempt to write to
+ * 	the output buffer (which may be NULL) if a size of zero was specified;
+ * 	check for integer overflow of the field width, precision, and return
+ * 	values and during the floating point conversion; use the OUTCHAR() macro
+ * 	instead of a function for better performance; provide asprintf(3) and
+ * 	vasprintf(3) functions; add new test cases.  The replacement functions
+ * 	have been renamed to use an "rpl_" prefix, the function calls in the
+ * 	main project (and in this file) must be redefined accordingly for each
+ * 	replacement function which is needed (by using Autoconf or other means).
+ * 	Various other minor improvements have been applied and the coding style
+ * 	was cleaned up for consistency.
+ *
+ * 2007-07-23 Holger Weiss <holger@jhweiss.de> for Mutt 1.5.13:
+ *
+ * 	C99 compliant snprintf(3) and vsnprintf(3) functions return the number
+ * 	of characters that would have been written to a sufficiently sized
+ * 	buffer (excluding the '\0').  The original code simply returned the
+ * 	length of the resulting output string, so that's been fixed.
+ *
+ * 1998-03-05 Michael Elkins <me@mutt.org> for Mutt 0.90.8:
+ *
+ * 	The original code assumed that both snprintf(3) and vsnprintf(3) were
+ * 	missing.  Some systems only have snprintf(3) but not vsnprintf(3), so
+ * 	the code is now broken down under HAVE_SNPRINTF and HAVE_VSNPRINTF.
+ *
+ * 1998-01-27 Thomas Roessler <roessler@does-not-exist.org> for Mutt 0.89i:
+ *
+ * 	The PGP code was using unsigned hexadecimal formats.  Unfortunately,
+ * 	unsigned formats simply didn't work.
+ *
+ * 1997-10-22 Brandon Long <blong@fiction.net> for Mutt 0.87.1:
+ *
+ * 	Ok, added some minimal floating point support, which means this probably
+ * 	requires libm on most operating systems.  Don't yet support the exponent
+ * 	(e,E) and sigfig (g,G).  Also, fmtint() was pretty badly broken, it just
+ * 	wasn't being exercised in ways which showed it, so that's been fixed.
+ * 	Also, formatted the code to Mutt conventions, and removed dead code left
+ * 	over from the original.  Also, there is now a builtin-test, run with:
+ * 	gcc -DTEST_SNPRINTF -o snprintf snprintf.c -lm && ./snprintf
+ *
+ * 2996-09-15 Brandon Long <blong@fiction.net> for Mutt 0.43:
+ *
+ * 	This was ugly.  It is still ugly.  I opted out of floating point
+ * 	numbers, but the formatter understands just about everything from the
+ * 	normal C string format, at least as far as I can tell from the Solaris
+ * 	2.5 printf(3S) man page.
+ */
+
+/*
+ * ToDo
+ *
+ * - Add wide character support.
+ * - Add support for "%a" and "%A" conversions.
+ * - Create test routines which predefine the expected results.  Our test cases
+ *   usually expose bugs in system implementations rather than in ours :-)
+ */
+
+/*
+ * Usage
+ *
+ * 1) The following preprocessor macros should be defined to 1 if the feature or
+ *    file in question is available on the target system (by using Autoconf or
+ *    other means), though basic functionality should be available as long as
+ *    HAVE_STDARG_H and HAVE_STDLIB_H are defined correctly:
+ *
+ *    	HAVE_VSNPRINTF
+ *    	HAVE_SNPRINTF
+ *    	HAVE_VASPRINTF
+ *    	HAVE_ASPRINTF
+ *    	HAVE_STDARG_H
+ *    	HAVE_STDDEF_H
+ *    	HAVE_STDINT_H
+ *    	HAVE_STDLIB_H
+ *    	HAVE_INTTYPES_H
+ *    	HAVE_LOCALE_H
+ *    	HAVE_LOCALECONV
+ *    	HAVE_LCONV_DECIMAL_POINT
+ *    	HAVE_LCONV_THOUSANDS_SEP
+ *    	HAVE_LONG_DOUBLE
+ *    	HAVE_LONG_LONG_INT
+ *    	HAVE_UNSIGNED_LONG_LONG_INT
+ *    	HAVE_INTMAX_T
+ *    	HAVE_UINTMAX_T
+ *    	HAVE_UINTPTR_T
+ *    	HAVE_PTRDIFF_T
+ *    	HAVE_VA_COPY
+ *    	HAVE___VA_COPY
+ *
+ * 2) The calls to the functions which should be replaced must be redefined
+ *    throughout the project files (by using Autoconf or other means):
+ *
+ *    	#define vsnprintf rpl_vsnprintf
+ *    	#define snprintf rpl_snprintf
+ *    	#define vasprintf rpl_vasprintf
+ *    	#define asprintf rpl_asprintf
+ *
+ * 3) The required replacement functions should be declared in some header file
+ *    included throughout the project files:
+ *
+ *    	#if HAVE_CONFIG_H
+ *    	#include <config.h>
+ *    	#endif
+ *    	#if HAVE_STDARG_H
+ *    	#include <stdarg.h>
+ *    	#if !HAVE_VSNPRINTF
+ *    	int rpl_vsnprintf(char *, size_t, const char *, va_list);
+ *    	#endif
+ *    	#if !HAVE_SNPRINTF
+ *    	int rpl_snprintf(char *, size_t, const char *, ...);
+ *    	#endif
+ *    	#if !HAVE_VASPRINTF
+ *    	int rpl_vasprintf(char **, const char *, va_list);
+ *    	#endif
+ *    	#if !HAVE_ASPRINTF
+ *    	int rpl_asprintf(char **, const char *, ...);
+ *    	#endif
+ *    	#endif
+ *
+ * Autoconf macros for handling step 1 and step 2 are available at
+ * <http://www.jhweiss.de/software/snprintf.html>.
+ */
+
+#include "pipe/p_config.h"
+
+#if HAVE_CONFIG_H
+#include <config.h>
+#else
+#ifdef WIN32
+#define vsnprintf util_vsnprintf
+#define snprintf util_snprintf
+#define HAVE_VSNPRINTF 0
+#define HAVE_SNPRINTF 0
+#define HAVE_VASPRINTF 1 /* not needed */
+#define HAVE_ASPRINTF 1 /* not needed */
+#define HAVE_STDARG_H 1
+#define HAVE_STDDEF_H 1
+#define HAVE_STDINT_H 0
+#define HAVE_STDLIB_H 1
+#define HAVE_INTTYPES_H 0
+#define HAVE_LOCALE_H 0
+#define HAVE_LOCALECONV 0
+#define HAVE_LCONV_DECIMAL_POINT 0
+#define HAVE_LCONV_THOUSANDS_SEP 0
+#define HAVE_LONG_DOUBLE 0
+#define HAVE_LONG_LONG_INT 1
+#define HAVE_UNSIGNED_LONG_LONG_INT 1
+#define HAVE_INTMAX_T 0
+#define HAVE_UINTMAX_T 0
+#define HAVE_UINTPTR_T 1
+#define HAVE_PTRDIFF_T 1
+#define HAVE_VA_COPY 0
+#define HAVE___VA_COPY 0
+#else
+#define HAVE_VSNPRINTF 1
+#define HAVE_SNPRINTF 1
+#define HAVE_VASPRINTF 1
+#define HAVE_ASPRINTF 1
+#endif
+#endif	/* HAVE_CONFIG_H */
+
+#if !HAVE_SNPRINTF || !HAVE_VSNPRINTF || !HAVE_ASPRINTF || !HAVE_VASPRINTF
+#include <stdio.h>	/* For NULL, size_t, vsnprintf(3), and vasprintf(3). */
+#ifdef VA_START
+#undef VA_START
+#endif	/* defined(VA_START) */
+#ifdef VA_SHIFT
+#undef VA_SHIFT
+#endif	/* defined(VA_SHIFT) */
+#if HAVE_STDARG_H
+#include <stdarg.h>
+#define VA_START(ap, last) va_start(ap, last)
+#define VA_SHIFT(ap, value, type) /* No-op for ANSI C. */
+#else	/* Assume <varargs.h> is available. */
+#include <varargs.h>
+#define VA_START(ap, last) va_start(ap)	/* "last" is ignored. */
+#define VA_SHIFT(ap, value, type) value = va_arg(ap, type)
+#endif	/* HAVE_STDARG_H */
+
+#if !HAVE_VASPRINTF
+#if HAVE_STDLIB_H
+#include <stdlib.h>	/* For malloc(3). */
+#endif	/* HAVE_STDLIB_H */
+#ifdef VA_COPY
+#undef VA_COPY
+#endif	/* defined(VA_COPY) */
+#ifdef VA_END_COPY
+#undef VA_END_COPY
+#endif	/* defined(VA_END_COPY) */
+#if HAVE_VA_COPY
+#define VA_COPY(dest, src) va_copy(dest, src)
+#define VA_END_COPY(ap) va_end(ap)
+#elif HAVE___VA_COPY
+#define VA_COPY(dest, src) __va_copy(dest, src)
+#define VA_END_COPY(ap) va_end(ap)
+#else
+#define VA_COPY(dest, src) (void)mymemcpy(&dest, &src, sizeof(va_list))
+#define VA_END_COPY(ap) /* No-op. */
+#define NEED_MYMEMCPY 1
+static void *mymemcpy(void *, void *, size_t);
+#endif	/* HAVE_VA_COPY */
+#endif	/* !HAVE_VASPRINTF */
+
+#if !HAVE_VSNPRINTF
+#include <limits.h>	/* For *_MAX. */
+#if HAVE_INTTYPES_H
+#include <inttypes.h>	/* For intmax_t (if not defined in <stdint.h>). */
+#endif	/* HAVE_INTTYPES_H */
+#if HAVE_LOCALE_H
+#include <locale.h>	/* For localeconv(3). */
+#endif	/* HAVE_LOCALE_H */
+#if HAVE_STDDEF_H
+#include <stddef.h>	/* For ptrdiff_t. */
+#endif	/* HAVE_STDDEF_H */
+#if HAVE_STDINT_H
+#include <stdint.h>	/* For intmax_t. */
+#endif	/* HAVE_STDINT_H */
+
+/* Support for unsigned long long int.  We may also need ULLONG_MAX. */
+#ifndef ULONG_MAX	/* We may need ULONG_MAX as a fallback. */
+#ifdef UINT_MAX
+#define ULONG_MAX UINT_MAX
+#else
+#define ULONG_MAX INT_MAX
+#endif	/* defined(UINT_MAX) */
+#endif	/* !defined(ULONG_MAX) */
+#ifdef ULLONG
+#undef ULLONG
+#endif	/* defined(ULLONG) */
+#if HAVE_UNSIGNED_LONG_LONG_INT
+#define ULLONG unsigned long long int
+#ifndef ULLONG_MAX
+#define ULLONG_MAX ULONG_MAX
+#endif	/* !defined(ULLONG_MAX) */
+#else
+#define ULLONG unsigned long int
+#ifdef ULLONG_MAX
+#undef ULLONG_MAX
+#endif	/* defined(ULLONG_MAX) */
+#define ULLONG_MAX ULONG_MAX
+#endif	/* HAVE_LONG_LONG_INT */
+
+/* Support for uintmax_t.  We also need UINTMAX_MAX. */
+#ifdef UINTMAX_T
+#undef UINTMAX_T
+#endif	/* defined(UINTMAX_T) */
+#if HAVE_UINTMAX_T || defined(uintmax_t)
+#define UINTMAX_T uintmax_t
+#ifndef UINTMAX_MAX
+#define UINTMAX_MAX ULLONG_MAX
+#endif	/* !defined(UINTMAX_MAX) */
+#else
+#define UINTMAX_T ULLONG
+#ifdef UINTMAX_MAX
+#undef UINTMAX_MAX
+#endif	/* defined(UINTMAX_MAX) */
+#define UINTMAX_MAX ULLONG_MAX
+#endif	/* HAVE_UINTMAX_T || defined(uintmax_t) */
+
+/* Support for long double. */
+#ifndef LDOUBLE
+#if HAVE_LONG_DOUBLE
+#define LDOUBLE long double
+#else
+#define LDOUBLE double
+#endif	/* HAVE_LONG_DOUBLE */
+#endif	/* !defined(LDOUBLE) */
+
+/* Support for long long int. */
+#ifndef LLONG
+#if HAVE_LONG_LONG_INT
+#define LLONG long long int
+#else
+#define LLONG long int
+#endif	/* HAVE_LONG_LONG_INT */
+#endif	/* !defined(LLONG) */
+
+/* Support for intmax_t. */
+#ifndef INTMAX_T
+#if HAVE_INTMAX_T || defined(intmax_t)
+#define INTMAX_T intmax_t
+#else
+#define INTMAX_T LLONG
+#endif	/* HAVE_INTMAX_T || defined(intmax_t) */
+#endif	/* !defined(INTMAX_T) */
+
+/* Support for uintptr_t. */
+#ifndef UINTPTR_T
+#if HAVE_UINTPTR_T || defined(uintptr_t)
+#define UINTPTR_T uintptr_t
+#else
+#define UINTPTR_T unsigned long int
+#endif	/* HAVE_UINTPTR_T || defined(uintptr_t) */
+#endif	/* !defined(UINTPTR_T) */
+
+/* WinCE5.0 does not have uintptr_t defined */ 
+#if (_WIN32_WCE < 600) 
+#ifdef UINTPTR_T 
+#undef UINTPTR_T 
+#endif 
+#define UINTPTR_T unsigned long int 
+#endif 
+
+
+/* Support for ptrdiff_t. */
+#ifndef PTRDIFF_T
+#if HAVE_PTRDIFF_T || defined(ptrdiff_t)
+#define PTRDIFF_T ptrdiff_t
+#else
+#define PTRDIFF_T long int
+#endif	/* HAVE_PTRDIFF_T || defined(ptrdiff_t) */
+#endif	/* !defined(PTRDIFF_T) */
+
+/*
+ * We need an unsigned integer type corresponding to ptrdiff_t (cf. C99:
+ * 7.19.6.1, 7).  However, we'll simply use PTRDIFF_T and convert it to an
+ * unsigned type if necessary.  This should work just fine in practice.
+ */
+#ifndef UPTRDIFF_T
+#define UPTRDIFF_T PTRDIFF_T
+#endif	/* !defined(UPTRDIFF_T) */
+
+/*
+ * We need a signed integer type corresponding to size_t (cf. C99: 7.19.6.1, 7).
+ * However, we'll simply use size_t and convert it to a signed type if
+ * necessary.  This should work just fine in practice.
+ */
+#ifndef SSIZE_T
+#define SSIZE_T size_t
+#endif	/* !defined(SSIZE_T) */
+
+/* Either ERANGE or E2BIG should be available everywhere. */
+#ifndef ERANGE
+#define ERANGE E2BIG
+#endif	/* !defined(ERANGE) */
+#ifndef EOVERFLOW
+#define EOVERFLOW ERANGE
+#endif	/* !defined(EOVERFLOW) */
+
+/*
+ * Buffer size to hold the octal string representation of UINT128_MAX without
+ * nul-termination ("3777777777777777777777777777777777777777777").
+ */
+#ifdef MAX_CONVERT_LENGTH
+#undef MAX_CONVERT_LENGTH
+#endif	/* defined(MAX_CONVERT_LENGTH) */
+#define MAX_CONVERT_LENGTH      43
+
+/* Format read states. */
+#define PRINT_S_DEFAULT         0
+#define PRINT_S_FLAGS           1
+#define PRINT_S_WIDTH           2
+#define PRINT_S_DOT             3
+#define PRINT_S_PRECISION       4
+#define PRINT_S_MOD             5
+#define PRINT_S_CONV            6
+
+/* Format flags. */
+#define PRINT_F_MINUS           (1 << 0)
+#define PRINT_F_PLUS            (1 << 1)
+#define PRINT_F_SPACE           (1 << 2)
+#define PRINT_F_NUM             (1 << 3)
+#define PRINT_F_ZERO            (1 << 4)
+#define PRINT_F_QUOTE           (1 << 5)
+#define PRINT_F_UP              (1 << 6)
+#define PRINT_F_UNSIGNED        (1 << 7)
+#define PRINT_F_TYPE_G          (1 << 8)
+#define PRINT_F_TYPE_E          (1 << 9)
+
+/* Conversion flags. */
+#define PRINT_C_CHAR            1
+#define PRINT_C_SHORT           2
+#define PRINT_C_LONG            3
+#define PRINT_C_LLONG           4
+#define PRINT_C_LDOUBLE         5
+#define PRINT_C_SIZE            6
+#define PRINT_C_PTRDIFF         7
+#define PRINT_C_INTMAX          8
+
+#ifndef MAX
+#define MAX(x, y) ((x >= y) ? x : y)
+#endif	/* !defined(MAX) */
+#ifndef CHARTOINT
+#define CHARTOINT(ch) (ch - '0')
+#endif	/* !defined(CHARTOINT) */
+#ifndef ISDIGIT
+#define ISDIGIT(ch) ('0' <= (unsigned char)ch && (unsigned char)ch <= '9')
+#endif	/* !defined(ISDIGIT) */
+#ifndef ISNAN
+#define ISNAN(x) (x != x)
+#endif	/* !defined(ISNAN) */
+#ifndef ISINF
+#define ISINF(x) (x != 0.0 && x + x == x)
+#endif	/* !defined(ISINF) */
+
+#ifdef OUTCHAR
+#undef OUTCHAR
+#endif	/* defined(OUTCHAR) */
+#define OUTCHAR(str, len, size, ch)                                          \
+do {                                                                         \
+	if (len + 1 < size)                                                  \
+		str[len] = ch;                                               \
+	(len)++;                                                             \
+} while (/* CONSTCOND */ 0)
+
+static void fmtstr(char *, size_t *, size_t, const char *, int, int, int);
+static void fmtint(char *, size_t *, size_t, INTMAX_T, int, int, int, int);
+static void fmtflt(char *, size_t *, size_t, LDOUBLE, int, int, int, int *);
+static void printsep(char *, size_t *, size_t);
+static int getnumsep(int);
+static int getexponent(LDOUBLE);
+static int convert(UINTMAX_T, char *, size_t, int, int);
+static UINTMAX_T cast(LDOUBLE);
+static UINTMAX_T myround(LDOUBLE);
+static LDOUBLE mypow10(int);
+
+int
+util_vsnprintf(char *str, size_t size, const char *format, va_list args)
+{
+	LDOUBLE fvalue;
+	INTMAX_T value;
+	unsigned char cvalue;
+	const char *strvalue;
+	INTMAX_T *intmaxptr;
+	PTRDIFF_T *ptrdiffptr;
+	SSIZE_T *sizeptr;
+	LLONG *llongptr;
+	long int *longptr;
+	int *intptr;
+	short int *shortptr;
+	signed char *charptr;
+	size_t len = 0;
+	int overflow = 0;
+	int base = 0;
+	int cflags = 0;
+	int flags = 0;
+	int width = 0;
+	int precision = -1;
+	int state = PRINT_S_DEFAULT;
+	char ch = *format++;
+
+	/*
+	 * C99 says: "If `n' is zero, nothing is written, and `s' may be a null
+	 * pointer." (7.19.6.5, 2)  We're forgiving and allow a NULL pointer
+	 * even if a size larger than zero was specified.  At least NetBSD's
+	 * snprintf(3) does the same, as well as other versions of this file.
+	 * (Though some of these versions will write to a non-NULL buffer even
+	 * if a size of zero was specified, which violates the standard.)
+	 */
+	if (str == NULL && size != 0)
+		size = 0;
+
+	while (ch != '\0')
+		switch (state) {
+		case PRINT_S_DEFAULT:
+			if (ch == '%')
+				state = PRINT_S_FLAGS;
+			else
+				OUTCHAR(str, len, size, ch);
+			ch = *format++;
+			break;
+		case PRINT_S_FLAGS:
+			switch (ch) {
+			case '-':
+				flags |= PRINT_F_MINUS;
+				ch = *format++;
+				break;
+			case '+':
+				flags |= PRINT_F_PLUS;
+				ch = *format++;
+				break;
+			case ' ':
+				flags |= PRINT_F_SPACE;
+				ch = *format++;
+				break;
+			case '#':
+				flags |= PRINT_F_NUM;
+				ch = *format++;
+				break;
+			case '0':
+				flags |= PRINT_F_ZERO;
+				ch = *format++;
+				break;
+			case '\'':	/* SUSv2 flag (not in C99). */
+				flags |= PRINT_F_QUOTE;
+				ch = *format++;
+				break;
+			default:
+				state = PRINT_S_WIDTH;
+				break;
+			}
+			break;
+		case PRINT_S_WIDTH:
+			if (ISDIGIT(ch)) {
+				ch = CHARTOINT(ch);
+				if (width > (INT_MAX - ch) / 10) {
+					overflow = 1;
+					goto out;
+				}
+				width = 10 * width + ch;
+				ch = *format++;
+			} else if (ch == '*') {
+				/*
+				 * C99 says: "A negative field width argument is
+				 * taken as a `-' flag followed by a positive
+				 * field width." (7.19.6.1, 5)
+				 */
+				if ((width = va_arg(args, int)) < 0) {
+					flags |= PRINT_F_MINUS;
+					width = -width;
+				}
+				ch = *format++;
+				state = PRINT_S_DOT;
+			} else
+				state = PRINT_S_DOT;
+			break;
+		case PRINT_S_DOT:
+			if (ch == '.') {
+				state = PRINT_S_PRECISION;
+				ch = *format++;
+			} else
+				state = PRINT_S_MOD;
+			break;
+		case PRINT_S_PRECISION:
+			if (precision == -1)
+				precision = 0;
+			if (ISDIGIT(ch)) {
+				ch = CHARTOINT(ch);
+				if (precision > (INT_MAX - ch) / 10) {
+					overflow = 1;
+					goto out;
+				}
+				precision = 10 * precision + ch;
+				ch = *format++;
+			} else if (ch == '*') {
+				/*
+				 * C99 says: "A negative precision argument is
+				 * taken as if the precision were omitted."
+				 * (7.19.6.1, 5)
+				 */
+				if ((precision = va_arg(args, int)) < 0)
+					precision = -1;
+				ch = *format++;
+				state = PRINT_S_MOD;
+			} else
+				state = PRINT_S_MOD;
+			break;
+		case PRINT_S_MOD:
+			switch (ch) {
+			case 'h':
+				ch = *format++;
+				if (ch == 'h') {	/* It's a char. */
+					ch = *format++;
+					cflags = PRINT_C_CHAR;
+				} else
+					cflags = PRINT_C_SHORT;
+				break;
+			case 'l':
+				ch = *format++;
+				if (ch == 'l') {	/* It's a long long. */
+					ch = *format++;
+					cflags = PRINT_C_LLONG;
+				} else
+					cflags = PRINT_C_LONG;
+				break;
+			case 'L':
+				cflags = PRINT_C_LDOUBLE;
+				ch = *format++;
+				break;
+			case 'j':
+				cflags = PRINT_C_INTMAX;
+				ch = *format++;
+				break;
+			case 't':
+				cflags = PRINT_C_PTRDIFF;
+				ch = *format++;
+				break;
+			case 'z':
+				cflags = PRINT_C_SIZE;
+				ch = *format++;
+				break;
+			}
+			state = PRINT_S_CONV;
+			break;
+		case PRINT_S_CONV:
+			switch (ch) {
+			case 'd':
+				/* FALLTHROUGH */
+			case 'i':
+				switch (cflags) {
+				case PRINT_C_CHAR:
+					value = (signed char)va_arg(args, int);
+					break;
+				case PRINT_C_SHORT:
+					value = (short int)va_arg(args, int);
+					break;
+				case PRINT_C_LONG:
+					value = va_arg(args, long int);
+					break;
+				case PRINT_C_LLONG:
+					value = va_arg(args, LLONG);
+					break;
+				case PRINT_C_SIZE:
+					value = va_arg(args, SSIZE_T);
+					break;
+				case PRINT_C_INTMAX:
+					value = va_arg(args, INTMAX_T);
+					break;
+				case PRINT_C_PTRDIFF:
+					value = va_arg(args, PTRDIFF_T);
+					break;
+				default:
+					value = va_arg(args, int);
+					break;
+				}
+				fmtint(str, &len, size, value, 10, width,
+				    precision, flags);
+				break;
+			case 'X':
+				flags |= PRINT_F_UP;
+				/* FALLTHROUGH */
+			case 'x':
+				base = 16;
+				/* FALLTHROUGH */
+			case 'o':
+				if (base == 0)
+					base = 8;
+				/* FALLTHROUGH */
+			case 'u':
+				if (base == 0)
+					base = 10;
+				flags |= PRINT_F_UNSIGNED;
+				switch (cflags) {
+				case PRINT_C_CHAR:
+					value = (unsigned char)va_arg(args,
+					    unsigned int);
+					break;
+				case PRINT_C_SHORT:
+					value = (unsigned short int)va_arg(args,
+					    unsigned int);
+					break;
+				case PRINT_C_LONG:
+					value = va_arg(args, unsigned long int);
+					break;
+				case PRINT_C_LLONG:
+					value = va_arg(args, ULLONG);
+					break;
+				case PRINT_C_SIZE:
+					value = va_arg(args, size_t);
+					break;
+				case PRINT_C_INTMAX:
+					value = va_arg(args, UINTMAX_T);
+					break;
+				case PRINT_C_PTRDIFF:
+					value = va_arg(args, UPTRDIFF_T);
+					break;
+				default:
+					value = va_arg(args, unsigned int);
+					break;
+				}
+				fmtint(str, &len, size, value, base, width,
+				    precision, flags);
+				break;
+			case 'A':
+				/* Not yet supported, we'll use "%F". */
+				/* FALLTHROUGH */
+			case 'F':
+				flags |= PRINT_F_UP;
+			case 'a':
+				/* Not yet supported, we'll use "%f". */
+				/* FALLTHROUGH */
+			case 'f':
+				if (cflags == PRINT_C_LDOUBLE)
+					fvalue = va_arg(args, LDOUBLE);
+				else
+					fvalue = va_arg(args, double);
+				fmtflt(str, &len, size, fvalue, width,
+				    precision, flags, &overflow);
+				if (overflow)
+					goto out;
+				break;
+			case 'E':
+				flags |= PRINT_F_UP;
+				/* FALLTHROUGH */
+			case 'e':
+				flags |= PRINT_F_TYPE_E;
+				if (cflags == PRINT_C_LDOUBLE)
+					fvalue = va_arg(args, LDOUBLE);
+				else
+					fvalue = va_arg(args, double);
+				fmtflt(str, &len, size, fvalue, width,
+				    precision, flags, &overflow);
+				if (overflow)
+					goto out;
+				break;
+			case 'G':
+				flags |= PRINT_F_UP;
+				/* FALLTHROUGH */
+			case 'g':
+				flags |= PRINT_F_TYPE_G;
+				if (cflags == PRINT_C_LDOUBLE)
+					fvalue = va_arg(args, LDOUBLE);
+				else
+					fvalue = va_arg(args, double);
+				/*
+				 * If the precision is zero, it is treated as
+				 * one (cf. C99: 7.19.6.1, 8).
+				 */
+				if (precision == 0)
+					precision = 1;
+				fmtflt(str, &len, size, fvalue, width,
+				    precision, flags, &overflow);
+				if (overflow)
+					goto out;
+				break;
+			case 'c':
+				cvalue = (unsigned char)va_arg(args, int);
+				OUTCHAR(str, len, size, cvalue);
+				break;
+			case 's':
+				strvalue = va_arg(args, char *);
+				fmtstr(str, &len, size, strvalue, width,
+				    precision, flags);
+				break;
+			case 'p':
+				/*
+				 * C99 says: "The value of the pointer is
+				 * converted to a sequence of printing
+				 * characters, in an implementation-defined
+				 * manner." (C99: 7.19.6.1, 8)
+				 */
+				if ((strvalue = va_arg(args, void *)) == NULL)
+					/*
+					 * We use the glibc format.  BSD prints
+					 * "0x0", SysV "0".
+					 */
+					fmtstr(str, &len, size, "(nil)", width,
+					    -1, flags);
+				else {
+					/*
+					 * We use the BSD/glibc format.  SysV
+					 * omits the "0x" prefix (which we emit
+					 * using the PRINT_F_NUM flag).
+					 */
+					flags |= PRINT_F_NUM;
+					flags |= PRINT_F_UNSIGNED;
+					fmtint(str, &len, size,
+					    (UINTPTR_T)strvalue, 16, width,
+					    precision, flags);
+				}
+				break;
+			case 'n':
+				switch (cflags) {
+				case PRINT_C_CHAR:
+					charptr = va_arg(args, signed char *);
+					*charptr = (signed char)len;
+					break;
+				case PRINT_C_SHORT:
+					shortptr = va_arg(args, short int *);
+					*shortptr = (short int)len;
+					break;
+				case PRINT_C_LONG:
+					longptr = va_arg(args, long int *);
+					*longptr = (long int)len;
+					break;
+				case PRINT_C_LLONG:
+					llongptr = va_arg(args, LLONG *);
+					*llongptr = (LLONG)len;
+					break;
+				case PRINT_C_SIZE:
+					/*
+					 * C99 says that with the "z" length
+					 * modifier, "a following `n' conversion
+					 * specifier applies to a pointer to a
+					 * signed integer type corresponding to
+					 * size_t argument." (7.19.6.1, 7)
+					 */
+					sizeptr = va_arg(args, SSIZE_T *);
+					*sizeptr = len;
+					break;
+				case PRINT_C_INTMAX:
+					intmaxptr = va_arg(args, INTMAX_T *);
+					*intmaxptr = len;
+					break;
+				case PRINT_C_PTRDIFF:
+					ptrdiffptr = va_arg(args, PTRDIFF_T *);
+					*ptrdiffptr = len;
+					break;
+				default:
+					intptr = va_arg(args, int *);
+					*intptr = len;
+					break;
+				}
+				break;
+			case '%':	/* Print a "%" character verbatim. */
+				OUTCHAR(str, len, size, ch);
+				break;
+			default:	/* Skip other characters. */
+				break;
+			}
+			ch = *format++;
+			state = PRINT_S_DEFAULT;
+			base = cflags = flags = width = 0;
+			precision = -1;
+			break;
+		}
+out:
+	if (len < size)
+		str[len] = '\0';
+	else if (size > 0)
+		str[size - 1] = '\0';
+
+	if (overflow || len >= INT_MAX) {
+		return -1;
+	}
+	return (int)len;
+}
+
+static void
+fmtstr(char *str, size_t *len, size_t size, const char *value, int width,
+       int precision, int flags)
+{
+	int padlen, strln;	/* Amount to pad. */
+	int noprecision = (precision == -1);
+
+	if (value == NULL)	/* We're forgiving. */
+		value = "(null)";
+
+	/* If a precision was specified, don't read the string past it. */
+	for (strln = 0; value[strln] != '\0' &&
+	    (noprecision || strln < precision); strln++)
+		continue;
+
+	if ((padlen = width - strln) < 0)
+		padlen = 0;
+	if (flags & PRINT_F_MINUS)	/* Left justify. */
+		padlen = -padlen;
+
+	while (padlen > 0) {	/* Leading spaces. */
+		OUTCHAR(str, *len, size, ' ');
+		padlen--;
+	}
+	while (*value != '\0' && (noprecision || precision-- > 0)) {
+		OUTCHAR(str, *len, size, *value);
+		value++;
+	}
+	while (padlen < 0) {	/* Trailing spaces. */
+		OUTCHAR(str, *len, size, ' ');
+		padlen++;
+	}
+}
+
+static void
+fmtint(char *str, size_t *len, size_t size, INTMAX_T value, int base, int width,
+       int precision, int flags)
+{
+	UINTMAX_T uvalue;
+	char iconvert[MAX_CONVERT_LENGTH];
+	char sign = 0;
+	char hexprefix = 0;
+	int spadlen = 0;	/* Amount to space pad. */
+	int zpadlen = 0;	/* Amount to zero pad. */
+	int pos;
+	int separators = (flags & PRINT_F_QUOTE);
+	int noprecision = (precision == -1);
+
+	if (flags & PRINT_F_UNSIGNED)
+		uvalue = value;
+	else {
+		uvalue = (value >= 0) ? value : -value;
+		if (value < 0)
+			sign = '-';
+		else if (flags & PRINT_F_PLUS)	/* Do a sign. */
+			sign = '+';
+		else if (flags & PRINT_F_SPACE)
+			sign = ' ';
+	}
+
+	pos = convert(uvalue, iconvert, sizeof(iconvert), base,
+	    flags & PRINT_F_UP);
+
+	if (flags & PRINT_F_NUM && uvalue != 0) {
+		/*
+		 * C99 says: "The result is converted to an `alternative form'.
+		 * For `o' conversion, it increases the precision, if and only
+		 * if necessary, to force the first digit of the result to be a
+		 * zero (if the value and precision are both 0, a single 0 is
+		 * printed).  For `x' (or `X') conversion, a nonzero result has
+		 * `0x' (or `0X') prefixed to it." (7.19.6.1, 6)
+		 */
+		switch (base) {
+		case 8:
+			if (precision <= pos)
+				precision = pos + 1;
+			break;
+		case 16:
+			hexprefix = (flags & PRINT_F_UP) ? 'X' : 'x';
+			break;
+		}
+	}
+
+	if (separators)	/* Get the number of group separators we'll print. */
+		separators = getnumsep(pos);
+
+	zpadlen = precision - pos - separators;
+	spadlen = width                         /* Minimum field width. */
+	    - separators                        /* Number of separators. */
+	    - MAX(precision, pos)               /* Number of integer digits. */
+	    - ((sign != 0) ? 1 : 0)             /* Will we print a sign? */
+	    - ((hexprefix != 0) ? 2 : 0);       /* Will we print a prefix? */
+
+	if (zpadlen < 0)
+		zpadlen = 0;
+	if (spadlen < 0)
+		spadlen = 0;
+
+	/*
+	 * C99 says: "If the `0' and `-' flags both appear, the `0' flag is
+	 * ignored.  For `d', `i', `o', `u', `x', and `X' conversions, if a
+	 * precision is specified, the `0' flag is ignored." (7.19.6.1, 6)
+	 */
+	if (flags & PRINT_F_MINUS)	/* Left justify. */
+		spadlen = -spadlen;
+	else if (flags & PRINT_F_ZERO && noprecision) {
+		zpadlen += spadlen;
+		spadlen = 0;
+	}
+	while (spadlen > 0) {	/* Leading spaces. */
+		OUTCHAR(str, *len, size, ' ');
+		spadlen--;
+	}
+	if (sign != 0)	/* Sign. */
+		OUTCHAR(str, *len, size, sign);
+	if (hexprefix != 0) {	/* A "0x" or "0X" prefix. */
+		OUTCHAR(str, *len, size, '0');
+		OUTCHAR(str, *len, size, hexprefix);
+	}
+	while (zpadlen > 0) {	/* Leading zeros. */
+		OUTCHAR(str, *len, size, '0');
+		zpadlen--;
+	}
+	while (pos > 0) {	/* The actual digits. */
+		pos--;
+		OUTCHAR(str, *len, size, iconvert[pos]);
+		if (separators > 0 && pos > 0 && pos % 3 == 0)
+			printsep(str, len, size);
+	}
+	while (spadlen < 0) {	/* Trailing spaces. */
+		OUTCHAR(str, *len, size, ' ');
+		spadlen++;
+	}
+}
+
+static void
+fmtflt(char *str, size_t *len, size_t size, LDOUBLE fvalue, int width,
+       int precision, int flags, int *overflow)
+{
+	LDOUBLE ufvalue;
+	UINTMAX_T intpart;
+	UINTMAX_T fracpart;
+	UINTMAX_T mask;
+	const char *infnan = NULL;
+	char iconvert[MAX_CONVERT_LENGTH];
+	char fconvert[MAX_CONVERT_LENGTH];
+	char econvert[4];	/* "e-12" (without nul-termination). */
+	char esign = 0;
+	char sign = 0;
+	int leadfraczeros = 0;
+	int exponent = 0;
+	int emitpoint = 0;
+	int omitzeros = 0;
+	int omitcount = 0;
+	int padlen = 0;
+	int epos = 0;
+	int fpos = 0;
+	int ipos = 0;
+	int separators = (flags & PRINT_F_QUOTE);
+	int estyle = (flags & PRINT_F_TYPE_E);
+#if HAVE_LOCALECONV && HAVE_LCONV_DECIMAL_POINT
+	struct lconv *lc = localeconv();
+#endif	/* HAVE_LOCALECONV && HAVE_LCONV_DECIMAL_POINT */
+
+	/*
+	 * AIX' man page says the default is 0, but C99 and at least Solaris'
+	 * and NetBSD's man pages say the default is 6, and sprintf(3) on AIX
+	 * defaults to 6.
+	 */
+	if (precision == -1)
+		precision = 6;
+
+	if (fvalue < 0.0)
+		sign = '-';
+	else if (flags & PRINT_F_PLUS)	/* Do a sign. */
+		sign = '+';
+	else if (flags & PRINT_F_SPACE)
+		sign = ' ';
+
+	if (ISNAN(fvalue))
+		infnan = (flags & PRINT_F_UP) ? "NAN" : "nan";
+	else if (ISINF(fvalue))
+		infnan = (flags & PRINT_F_UP) ? "INF" : "inf";
+
+	if (infnan != NULL) {
+		if (sign != 0)
+			iconvert[ipos++] = sign;
+		while (*infnan != '\0')
+			iconvert[ipos++] = *infnan++;
+		fmtstr(str, len, size, iconvert, width, ipos, flags);
+		return;
+	}
+
+	/* "%e" (or "%E") or "%g" (or "%G") conversion. */
+	if (flags & PRINT_F_TYPE_E || flags & PRINT_F_TYPE_G) {
+		if (flags & PRINT_F_TYPE_G) {
+			/*
+			 * For "%g" (and "%G") conversions, the precision
+			 * specifies the number of significant digits, which
+			 * includes the digits in the integer part.  The
+			 * conversion will or will not be using "e-style" (like
+			 * "%e" or "%E" conversions) depending on the precision
+			 * and on the exponent.  However, the exponent can be
+			 * affected by rounding the converted value, so we'll
+			 * leave this decision for later.  Until then, we'll
+			 * assume that we're going to do an "e-style" conversion
+			 * (in order to get the exponent calculated).  For
+			 * "e-style", the precision must be decremented by one.
+			 */
+			precision--;
+			/*
+			 * For "%g" (and "%G") conversions, trailing zeros are
+			 * removed from the fractional portion of the result
+			 * unless the "#" flag was specified.
+			 */
+			if (!(flags & PRINT_F_NUM))
+				omitzeros = 1;
+		}
+		exponent = getexponent(fvalue);
+		estyle = 1;
+	}
+
+again:
+	/*
+	 * Sorry, we only support 9, 19, or 38 digits (that is, the number of
+	 * digits of the 32-bit, the 64-bit, or the 128-bit UINTMAX_MAX value
+	 * minus one) past the decimal point due to our conversion method.
+	 */
+	switch (sizeof(UINTMAX_T)) {
+	case 16:
+		if (precision > 38)
+			precision = 38;
+		break;
+	case 8:
+		if (precision > 19)
+			precision = 19;
+		break;
+	default:
+		if (precision > 9)
+			precision = 9;
+		break;
+	}
+
+	ufvalue = (fvalue >= 0.0) ? fvalue : -fvalue;
+	if (estyle)	/* We want exactly one integer digit. */
+		ufvalue /= mypow10(exponent);
+
+	if ((intpart = cast(ufvalue)) == UINTMAX_MAX) {
+		*overflow = 1;
+		return;
+	}
+
+	/*
+	 * Factor of ten with the number of digits needed for the fractional
+	 * part.  For example, if the precision is 3, the mask will be 1000.
+	 */
+#if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+	mask = (unsigned long)mypow10(precision);
+#else
+	mask = (UINTMAX_T)mypow10(precision);
+#endif
+	/*
+	 * We "cheat" by converting the fractional part to integer by
+	 * multiplying by a factor of ten.
+	 */
+	if ((fracpart = myround(mask * (ufvalue - intpart))) >= mask) {
+		/*
+		 * For example, ufvalue = 2.99962, intpart = 2, and mask = 1000
+		 * (because precision = 3).  Now, myround(1000 * 0.99962) will
+		 * return 1000.  So, the integer part must be incremented by one
+		 * and the fractional part must be set to zero.
+		 */
+		intpart++;
+		fracpart = 0;
+		if (estyle && intpart == 10) {
+			/*
+			 * The value was rounded up to ten, but we only want one
+			 * integer digit if using "e-style".  So, the integer
+			 * part must be set to one and the exponent must be
+			 * incremented by one.
+			 */
+			intpart = 1;
+			exponent++;
+		}
+	}
+
+	/*
+	 * Now that we know the real exponent, we can check whether or not to
+	 * use "e-style" for "%g" (and "%G") conversions.  If we don't need
+	 * "e-style", the precision must be adjusted and the integer and
+	 * fractional parts must be recalculated from the original value.
+	 *
+	 * C99 says: "Let P equal the precision if nonzero, 6 if the precision
+	 * is omitted, or 1 if the precision is zero.  Then, if a conversion
+	 * with style `E' would have an exponent of X:
+	 *
+	 * - if P > X >= -4, the conversion is with style `f' (or `F') and
+	 *   precision P - (X + 1).
+	 *
+	 * - otherwise, the conversion is with style `e' (or `E') and precision
+	 *   P - 1." (7.19.6.1, 8)
+	 *
+	 * Note that we had decremented the precision by one.
+	 */
+	if (flags & PRINT_F_TYPE_G && estyle &&
+	    precision + 1 > exponent && exponent >= -4) {
+		precision -= exponent;
+		estyle = 0;
+		goto again;
+	}
+
+	if (estyle) {
+		if (exponent < 0) {
+			exponent = -exponent;
+			esign = '-';
+		} else
+			esign = '+';
+
+		/*
+		 * Convert the exponent.  The sizeof(econvert) is 4.  So, the
+		 * econvert buffer can hold e.g. "e+99" and "e-99".  We don't
+		 * support an exponent which contains more than two digits.
+		 * Therefore, the following stores are safe.
+		 */
+		epos = convert(exponent, econvert, 2, 10, 0);
+		/*
+		 * C99 says: "The exponent always contains at least two digits,
+		 * and only as many more digits as necessary to represent the
+		 * exponent." (7.19.6.1, 8)
+		 */
+		if (epos == 1)
+			econvert[epos++] = '0';
+		econvert[epos++] = esign;
+		econvert[epos++] = (flags & PRINT_F_UP) ? 'E' : 'e';
+	}
+
+	/* Convert the integer part and the fractional part. */
+	ipos = convert(intpart, iconvert, sizeof(iconvert), 10, 0);
+	if (fracpart != 0)	/* convert() would return 1 if fracpart == 0. */
+		fpos = convert(fracpart, fconvert, sizeof(fconvert), 10, 0);
+
+	leadfraczeros = precision - fpos;
+
+	if (omitzeros) {
+		if (fpos > 0)	/* Omit trailing fractional part zeros. */
+			while (omitcount < fpos && fconvert[omitcount] == '0')
+				omitcount++;
+		else {	/* The fractional part is zero, omit it completely. */
+			omitcount = precision;
+			leadfraczeros = 0;
+		}
+		precision -= omitcount;
+	}
+
+	/*
+	 * Print a decimal point if either the fractional part is non-zero
+	 * and/or the "#" flag was specified.
+	 */
+	if (precision > 0 || flags & PRINT_F_NUM)
+		emitpoint = 1;
+	if (separators)	/* Get the number of group separators we'll print. */
+		separators = getnumsep(ipos);
+
+	padlen = width                  /* Minimum field width. */
+	    - ipos                      /* Number of integer digits. */
+	    - epos                      /* Number of exponent characters. */
+	    - precision                 /* Number of fractional digits. */
+	    - separators                /* Number of group separators. */
+	    - (emitpoint ? 1 : 0)       /* Will we print a decimal point? */
+	    - ((sign != 0) ? 1 : 0);    /* Will we print a sign character? */
+
+	if (padlen < 0)
+		padlen = 0;
+
+	/*
+	 * C99 says: "If the `0' and `-' flags both appear, the `0' flag is
+	 * ignored." (7.19.6.1, 6)
+	 */
+	if (flags & PRINT_F_MINUS)	/* Left justifty. */
+		padlen = -padlen;
+	else if (flags & PRINT_F_ZERO && padlen > 0) {
+		if (sign != 0) {	/* Sign. */
+			OUTCHAR(str, *len, size, sign);
+			sign = 0;
+		}
+		while (padlen > 0) {	/* Leading zeros. */
+			OUTCHAR(str, *len, size, '0');
+			padlen--;
+		}
+	}
+	while (padlen > 0) {	/* Leading spaces. */
+		OUTCHAR(str, *len, size, ' ');
+		padlen--;
+	}
+	if (sign != 0)	/* Sign. */
+		OUTCHAR(str, *len, size, sign);
+	while (ipos > 0) {	/* Integer part. */
+		ipos--;
+		OUTCHAR(str, *len, size, iconvert[ipos]);
+		if (separators > 0 && ipos > 0 && ipos % 3 == 0)
+			printsep(str, len, size);
+	}
+	if (emitpoint) {	/* Decimal point. */
+#if HAVE_LOCALECONV && HAVE_LCONV_DECIMAL_POINT
+		if (lc->decimal_point != NULL && *lc->decimal_point != '\0')
+			OUTCHAR(str, *len, size, *lc->decimal_point);
+		else	/* We'll always print some decimal point character. */
+#endif	/* HAVE_LOCALECONV && HAVE_LCONV_DECIMAL_POINT */
+			OUTCHAR(str, *len, size, '.');
+	}
+	while (leadfraczeros > 0) {	/* Leading fractional part zeros. */
+		OUTCHAR(str, *len, size, '0');
+		leadfraczeros--;
+	}
+	while (fpos > omitcount) {	/* The remaining fractional part. */
+		fpos--;
+		OUTCHAR(str, *len, size, fconvert[fpos]);
+	}
+	while (epos > 0) {	/* Exponent. */
+		epos--;
+		OUTCHAR(str, *len, size, econvert[epos]);
+	}
+	while (padlen < 0) {	/* Trailing spaces. */
+		OUTCHAR(str, *len, size, ' ');
+		padlen++;
+	}
+}
+
+static void
+printsep(char *str, size_t *len, size_t size)
+{
+#if HAVE_LOCALECONV && HAVE_LCONV_THOUSANDS_SEP
+	struct lconv *lc = localeconv();
+	int i;
+
+	if (lc->thousands_sep != NULL)
+		for (i = 0; lc->thousands_sep[i] != '\0'; i++)
+			OUTCHAR(str, *len, size, lc->thousands_sep[i]);
+	else
+#endif	/* HAVE_LOCALECONV && HAVE_LCONV_THOUSANDS_SEP */
+		OUTCHAR(str, *len, size, ',');
+}
+
+static int
+getnumsep(int digits)
+{
+	int separators = (digits - ((digits % 3 == 0) ? 1 : 0)) / 3;
+#if HAVE_LOCALECONV && HAVE_LCONV_THOUSANDS_SEP
+	int strln;
+	struct lconv *lc = localeconv();
+
+	/* We support an arbitrary separator length (including zero). */
+	if (lc->thousands_sep != NULL) {
+		for (strln = 0; lc->thousands_sep[strln] != '\0'; strln++)
+			continue;
+		separators *= strln;
+	}
+#endif	/* HAVE_LOCALECONV && HAVE_LCONV_THOUSANDS_SEP */
+	return separators;
+}
+
+static int
+getexponent(LDOUBLE value)
+{
+	LDOUBLE tmp = (value >= 0.0) ? value : -value;
+	int exponent = 0;
+
+	/*
+	 * We check for 99 > exponent > -99 in order to work around possible
+	 * endless loops which could happen (at least) in the second loop (at
+	 * least) if we're called with an infinite value.  However, we checked
+	 * for infinity before calling this function using our ISINF() macro, so
+	 * this might be somewhat paranoid.
+	 */
+	while (tmp < 1.0 && tmp > 0.0 && --exponent > -99)
+		tmp *= 10;
+	while (tmp >= 10.0 && ++exponent < 99)
+		tmp /= 10;
+
+	return exponent;
+}
+
+static int
+convert(UINTMAX_T value, char *buf, size_t size, int base, int caps)
+{
+	const char *digits = caps ? "0123456789ABCDEF" : "0123456789abcdef";
+	size_t pos = 0;
+
+	/* We return an unterminated buffer with the digits in reverse order. */
+	do {
+		buf[pos++] = digits[value % base];
+		value /= base;
+	} while (value != 0 && pos < size);
+
+	return (int)pos;
+}
+
+static UINTMAX_T
+cast(LDOUBLE value)
+{
+	UINTMAX_T result;
+
+	/*
+	 * We check for ">=" and not for ">" because if UINTMAX_MAX cannot be
+	 * represented exactly as an LDOUBLE value (but is less than LDBL_MAX),
+	 * it may be increased to the nearest higher representable value for the
+	 * comparison (cf. C99: 6.3.1.4, 2).  It might then equal the LDOUBLE
+	 * value although converting the latter to UINTMAX_T would overflow.
+	 */
+	if (value >= UINTMAX_MAX)
+		return UINTMAX_MAX;
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+	result = (unsigned long)value;
+#else
+	result = (UINTMAX_T)value;
+#endif
+	/*
+	 * At least on NetBSD/sparc64 3.0.2 and 4.99.30, casting long double to
+	 * an integer type converts e.g. 1.9 to 2 instead of 1 (which violates
+	 * the standard).  Sigh.
+	 */
+	return (result <= value) ? result : result - 1;
+}
+
+static UINTMAX_T
+myround(LDOUBLE value)
+{
+	UINTMAX_T intpart = cast(value);
+
+	return ((value -= intpart) < 0.5) ? intpart : intpart + 1;
+}
+
+static LDOUBLE
+mypow10(int exponent)
+{
+	LDOUBLE result = 1;
+
+	while (exponent > 0) {
+		result *= 10;
+		exponent--;
+	}
+	while (exponent < 0) {
+		result /= 10;
+		exponent++;
+	}
+	return result;
+}
+#endif	/* !HAVE_VSNPRINTF */
+
+#if !HAVE_VASPRINTF
+#if NEED_MYMEMCPY
+void *
+mymemcpy(void *dst, void *src, size_t len)
+{
+	const char *from = src;
+	char *to = dst;
+
+	/* No need for optimization, we use this only to replace va_copy(3). */
+	while (len-- > 0)
+		*to++ = *from++;
+	return dst;
+}
+#endif	/* NEED_MYMEMCPY */
+
+int
+util_vasprintf(char **ret, const char *format, va_list ap)
+{
+	size_t size;
+	int len;
+	va_list aq;
+
+	VA_COPY(aq, ap);
+	len = vsnprintf(NULL, 0, format, aq);
+	VA_END_COPY(aq);
+	if (len < 0 || (*ret = malloc(size = len + 1)) == NULL)
+		return -1;
+	return vsnprintf(*ret, size, format, ap);
+}
+#endif	/* !HAVE_VASPRINTF */
+
+#if !HAVE_SNPRINTF
+#if HAVE_STDARG_H
+int
+util_snprintf(char *str, size_t size, const char *format, ...)
+#else
+int
+util_snprintf(va_alist) va_dcl
+#endif	/* HAVE_STDARG_H */
+{
+#if !HAVE_STDARG_H
+	char *str;
+	size_t size;
+	char *format;
+#endif	/* HAVE_STDARG_H */
+	va_list ap;
+	int len;
+
+	VA_START(ap, format);
+	VA_SHIFT(ap, str, char *);
+	VA_SHIFT(ap, size, size_t);
+	VA_SHIFT(ap, format, const char *);
+	len = vsnprintf(str, size, format, ap);
+	va_end(ap);
+	return len;
+}
+#endif	/* !HAVE_SNPRINTF */
+
+#if !HAVE_ASPRINTF
+#if HAVE_STDARG_H
+int
+util_asprintf(char **ret, const char *format, ...)
+#else
+int
+util_asprintf(va_alist) va_dcl
+#endif	/* HAVE_STDARG_H */
+{
+#if !HAVE_STDARG_H
+	char **ret;
+	char *format;
+#endif	/* HAVE_STDARG_H */
+	va_list ap;
+	int len;
+
+	VA_START(ap, format);
+	VA_SHIFT(ap, ret, char **);
+	VA_SHIFT(ap, format, const char *);
+	len = vasprintf(ret, format, ap);
+	va_end(ap);
+	return len;
+}
+#endif	/* !HAVE_ASPRINTF */
+#else	/* Dummy declaration to avoid empty translation unit warnings. */
+int main(void);
+#endif	/* !HAVE_SNPRINTF || !HAVE_VSNPRINTF || !HAVE_ASPRINTF || [...] */
+
+
+/* vim: set joinspaces textwidth=80: */
diff --git a/src/gallium/auxiliary/util/u_sse.h b/src/gallium/auxiliary/util/u_sse.h
new file mode 100644
index 0000000000..e2a8491e62
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_sse.h
@@ -0,0 +1,77 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * SSE intrinsics portability header.
+ * 
+ * Although the SSE intrinsics are support by all modern x86 and x86-64 
+ * compilers, there are some intrisincs missing in some implementations 
+ * (especially older MSVC versions). This header abstracts that away.
+ */
+
+#ifndef U_SSE_H_
+#define U_SSE_H_
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_ARCH_SSE)
+
+#include <xmmintrin.h>
+#include <emmintrin.h>
+
+
+/* MSVC before VC8 does not support the _mm_castxxx_yyy */
+#if defined(_MSC_VER) && _MSC_VER < 1500
+
+union __declspec(align(16)) m128_types {
+   __m128 m128;
+   __m128i m128i;
+   __m128d m128d;
+};
+
+static __inline __m128
+_mm_castsi128_ps(__m128i a)
+{
+   union m128_types u;
+   u.m128i = a;
+   return u.m128;
+}
+
+static __inline __m128i
+_mm_castps_si128(__m128 a)
+{
+   union m128_types u;
+   u.m128 = a;
+   return u.m128i;
+}
+
+#endif /* defined(_MSC_VER) && _MSC_VER < 1500 */
+
+#endif /* PIPE_ARCH_X86 || PIPE_ARCH_X86_64 */
+
+#endif /* U_SSE_H_ */
diff --git a/src/gallium/auxiliary/util/u_stream.h b/src/gallium/auxiliary/util/u_stream.h
new file mode 100644
index 0000000000..a9d0f0121a
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_stream.h
@@ -0,0 +1,61 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Cross-platform sequential access stream abstraction.
+ */
+
+#ifndef U_STREAM_H
+#define U_STREAM_H
+
+
+#include "pipe/p_compiler.h"
+
+
+struct util_stream;
+
+
+/**
+ * Create a stream
+ * @param filename relative or absolute path (necessary for windows)  
+ * @param optional maximum file size (0 for a growable size).
+ */
+struct util_stream *
+util_stream_create(const char *filename, size_t max_size);
+
+boolean
+util_stream_write(struct util_stream *stream, const void *data, size_t size);
+
+void
+util_stream_flush(struct util_stream *stream);
+
+void
+util_stream_close(struct util_stream *stream);
+
+
+#endif /* U_STREAM_H */
diff --git a/src/gallium/auxiliary/util/u_stream_stdc.c b/src/gallium/auxiliary/util/u_stream_stdc.c
new file mode 100644
index 0000000000..ca80bef0f3
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_stream_stdc.c
@@ -0,0 +1,106 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Stream implementation based on the Standard C Library.
+ */
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_OS_LINUX) || defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+
+#include <stdio.h>
+
+#include "util/u_memory.h"
+
+#include "u_stream.h"
+
+
+struct util_stream 
+{
+   FILE *file;
+};
+
+
+struct util_stream *
+util_stream_create(const char *filename, size_t max_size)
+{
+   struct util_stream *stream;
+   
+   (void)max_size;
+   
+   stream = CALLOC_STRUCT(util_stream);
+   if(!stream)
+      goto error1;
+   
+   stream->file = fopen(filename, "w");
+   if(!stream->file)
+      goto error2;
+   
+   return stream;
+   
+error2:
+   FREE(stream);
+error1:
+   return NULL;
+}
+
+
+boolean
+util_stream_write(struct util_stream *stream, const void *data, size_t size)
+{
+   if(!stream)
+      return FALSE;
+   
+   return fwrite(data, size, 1, stream->file) == size ? TRUE : FALSE;
+}
+
+
+void
+util_stream_flush(struct util_stream *stream) 
+{
+   if(!stream)
+      return;
+   
+   fflush(stream->file);
+}
+
+
+void
+util_stream_close(struct util_stream *stream) 
+{
+   if(!stream)
+      return;
+   
+   fclose(stream->file);
+
+   FREE(stream);
+}
+
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_stream_wd.c b/src/gallium/auxiliary/util/u_stream_wd.c
new file mode 100644
index 0000000000..864489e775
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_stream_wd.c
@@ -0,0 +1,224 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Stream implementation for the Windows Display driver.
+ */
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+
+#include <windows.h>
+#include <winddi.h>
+
+#include "util/u_memory.h"
+#include "util/u_string.h"
+
+#include "u_stream.h"
+
+
+#define MAP_FILE_SIZE (4*1024*1024)
+
+
+struct util_stream 
+{
+   char filename[MAX_PATH + 1];
+   WCHAR wFileName[MAX_PATH + 1];
+   boolean growable;
+   size_t map_size;
+   ULONG_PTR iFile;
+   char *pMap;
+   size_t written;
+   unsigned suffix;
+};
+
+
+static INLINE boolean
+util_stream_map(struct util_stream *stream)
+{
+   ULONG BytesInUnicodeString;
+   static char filename[MAX_PATH + 1];
+   unsigned filename_len;
+
+   if(stream->growable)
+      filename_len = util_snprintf(filename,
+                                   sizeof(filename),
+                                   "%s.%04x",
+                                   stream->filename,
+                                   stream->suffix++);
+   else
+      filename_len = util_snprintf(filename,
+                                   sizeof(filename),
+                                   "%s",
+                                   stream->filename);
+
+   EngMultiByteToUnicodeN(
+         stream->wFileName,
+         sizeof(stream->wFileName),
+         &BytesInUnicodeString,
+         filename,
+         filename_len);
+   
+   stream->pMap = EngMapFile(stream->wFileName, stream->map_size, &stream->iFile);
+   if(!stream->pMap)
+      return FALSE;
+   
+   memset(stream->pMap, 0, stream->map_size);
+   stream->written = 0;
+   
+   return TRUE;
+}
+
+
+static INLINE void
+util_stream_unmap(struct util_stream *stream)
+{
+   EngUnmapFile(stream->iFile);
+   if(stream->written < stream->map_size) {
+      /* Truncate file size */
+      stream->pMap = EngMapFile(stream->wFileName, stream->written, &stream->iFile);
+      if(stream->pMap)
+         EngUnmapFile(stream->iFile);
+   }
+   
+   stream->pMap = NULL;
+}
+
+
+static INLINE void
+util_stream_full_qualified_filename(char *dst, size_t size, const char *src)
+{
+   boolean need_drive, need_root;
+   
+   if((('A' <= src[0] && src[0] <= 'Z') || ('a' <= src[0] && src[0] <= 'z')) && src[1] == ':') {
+      need_drive = FALSE;
+      need_root = src[2] == '\\' ? FALSE : TRUE;
+   }
+   else {
+      need_drive = TRUE;
+      need_root = src[0] == '\\' ? FALSE : TRUE;
+   }
+   
+   util_snprintf(dst, size, 
+                 "\\??\\%s%s%s",
+                 need_drive ? "C:" : "",
+                 need_root ? "\\" : "",
+                 src);
+}
+
+
+struct util_stream *
+util_stream_create(const char *filename, size_t max_size)
+{
+   struct util_stream *stream;
+   
+   stream = CALLOC_STRUCT(util_stream);
+   if(!stream)
+      goto error1;
+   
+   util_stream_full_qualified_filename(stream->filename,
+                                       sizeof(stream->filename),
+                                       filename);
+   
+   if(max_size) {
+      stream->growable = FALSE;
+      stream->map_size = max_size;
+   }
+   else {
+      stream->growable = TRUE;
+      stream->map_size = MAP_FILE_SIZE;
+   }
+   
+   if(!util_stream_map(stream))
+      goto error2;
+   
+   return stream;
+   
+error2:
+   FREE(stream);
+error1:
+   return NULL;
+}
+
+
+static INLINE void
+util_stream_copy(struct util_stream *stream, const char *data, size_t size)
+{
+   assert(stream->written + size <= stream->map_size);
+   memcpy(stream->pMap + stream->written, data, size);
+   stream->written += size;
+}
+
+
+boolean
+util_stream_write(struct util_stream *stream, const void *data, size_t size)
+{
+   if(!stream)
+      return FALSE;
+   
+   if(!stream->pMap)
+      return FALSE;
+   
+   while(stream->written + size > stream->map_size) {
+      size_t step = stream->map_size - stream->written;
+      util_stream_copy(stream, data, step);
+      data = (const char *)data + step;
+      size -= step;
+      
+      util_stream_unmap(stream);
+      if(!stream->growable || !util_stream_map(stream))
+         return FALSE;
+   }
+
+   util_stream_copy(stream, data, size);
+   
+   return TRUE;
+}
+
+
+void
+util_stream_flush(struct util_stream *stream) 
+{
+   (void)stream;
+}
+
+
+void
+util_stream_close(struct util_stream *stream) 
+{
+   if(!stream)
+      return;
+   
+   util_stream_unmap(stream);
+
+   FREE(stream);
+}
+
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_string.h b/src/gallium/auxiliary/util/u_string.h
new file mode 100644
index 0000000000..08c89bbf77
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_string.h
@@ -0,0 +1,220 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Platform independent functions for string manipulation.
+ *
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef U_STRING_H_
+#define U_STRING_H_
+
+#if !defined(WIN32) && !defined(XF86_LIBC_H)
+#include <stdio.h>
+#endif
+#include <stddef.h>
+#include <stdarg.h>
+
+#include "pipe/p_compiler.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+#ifdef WIN32
+
+int util_vsnprintf(char *, size_t, const char *, va_list);
+int util_snprintf(char *str, size_t size, const char *format, ...);
+
+static INLINE void
+util_vsprintf(char *str, const char *format, va_list ap)
+{
+   util_vsnprintf(str, (size_t)-1, format, ap);
+}
+
+static INLINE void
+util_sprintf(char *str, const char *format, ...)
+{
+   va_list ap;
+   va_start(ap, format);
+   util_vsnprintf(str, (size_t)-1, format, ap);
+   va_end(ap);
+}
+
+static INLINE char *
+util_strchr(const char *s, char c)
+{
+   while(*s) {
+      if(*s == c)
+	 return (char *)s;
+      ++s;
+   }
+   return NULL;
+}
+
+static INLINE char*
+util_strncat(char *dst, const char *src, size_t n)
+{
+   char *p = dst + strlen(dst);
+   const char *q = src;
+   size_t i;
+
+   for (i = 0; i < n && *q != '\0'; ++i)
+       *p++ = *q++;
+   *p = '\0';
+
+   return dst;
+}
+
+static INLINE int
+util_strcmp(const char *s1, const char *s2)
+{
+   unsigned char u1, u2;
+
+   while (1) {
+      u1 = (unsigned char) *s1++;
+      u2 = (unsigned char) *s2++;
+      if (u1 != u2)
+	 return u1 - u2;
+      if (u1 == '\0')
+	 return 0;
+   }
+   return 0;
+}
+
+static INLINE int
+util_strncmp(const char *s1, const char *s2, size_t n)
+{
+   unsigned char u1, u2;
+
+   while (n-- > 0) {
+      u1 = (unsigned char) *s1++;
+      u2 = (unsigned char) *s2++;
+      if (u1 != u2)
+	 return u1 - u2;
+      if (u1 == '\0')
+	 return 0;
+   }
+   return 0;
+}
+
+static INLINE char *
+util_strstr(const char *haystack, const char *needle)
+{
+   const char *p = haystack;
+   int len = strlen(needle);
+
+   for (; (p = util_strchr(p, *needle)) != 0; p++) {
+      if (util_strncmp(p, needle, len) == 0) {
+	 return (char *)p;
+      }
+   }
+   return NULL;
+}
+
+static INLINE void *
+util_memmove(void *dest, const void *src, size_t n)
+{
+   char *p = (char *)dest;
+   const char *q = (const char *)src;
+   if (dest < src) {
+      while (n--)
+	 *p++ = *q++;
+   }
+   else
+   {
+      p += n;
+      q += n;
+      while (n--)
+	 *--p = *--q;
+   }
+   return dest;
+}
+
+
+#else
+
+#define util_vsnprintf vsnprintf
+#define util_snprintf snprintf
+#define util_vsprintf vsprintf
+#define util_sprintf sprintf
+#define util_strchr strchr
+#define util_strcmp strcmp
+#define util_strncmp strncmp
+#define util_strncat strncat
+#define util_strstr strstr
+#define util_memmove memmove
+
+#endif
+
+
+/**
+ * Printable string buffer
+ */
+struct util_strbuf
+{
+   char *str;
+   char *ptr;
+   size_t left;
+};
+
+
+static INLINE void
+util_strbuf_init(struct util_strbuf *sbuf, char *str, size_t size) 
+{
+   sbuf->str = str;
+   sbuf->str[0] = 0;
+   sbuf->ptr = sbuf->str;
+   sbuf->left = size;
+}
+
+
+static INLINE void
+util_strbuf_printf(struct util_strbuf *sbuf, const char *format, ...)
+{
+   if(sbuf->left > 1) {
+      size_t written;
+      va_list ap;
+      va_start(ap, format);
+      written = util_vsnprintf(sbuf->ptr, sbuf->left, format, ap);
+      va_end(ap);
+      sbuf->ptr += written;
+      sbuf->left -= written;
+   }
+}
+
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* U_STRING_H_ */
diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c
new file mode 100644
index 0000000000..32f6b072a0
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_tile.c
@@ -0,0 +1,1198 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * RGBA/float tile get/put functions.
+ * Usable both by drivers and state trackers.
+ * Surfaces should already be in a mapped state.
+ */
+
+
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "util/u_rect.h"
+#include "util/u_tile.h"
+
+
+/**
+ * Move raw block of pixels from surface to user memory.
+ * This should be usable by any hw driver that has mappable surfaces.
+ */
+void
+pipe_get_tile_raw(struct pipe_surface *ps,
+                  uint x, uint y, uint w, uint h,
+                  void *dst, int dst_stride)
+{
+   const void *src;
+
+   if (dst_stride == 0)
+      dst_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size;
+
+   if (pipe_clip_tile(x, y, &w, &h, ps))
+      return;
+
+   src = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ);
+   assert(src);
+   if(!src)
+      return;
+
+   pipe_copy_rect(dst, &ps->block, dst_stride, 0, 0, w, h, src, ps->stride, x, y);
+
+   pipe_surface_unmap(ps);
+}
+
+
+/**
+ * Move raw block of pixels from user memory to surface.
+ * This should be usable by any hw driver that has mappable surfaces.
+ */
+void
+pipe_put_tile_raw(struct pipe_surface *ps,
+                  uint x, uint y, uint w, uint h,
+                  const void *src, int src_stride)
+{
+   void *dst;
+
+   if (src_stride == 0)
+      src_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size;
+
+   if (pipe_clip_tile(x, y, &w, &h, ps))
+      return;
+
+   dst = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE);
+   assert(dst);
+   if(!dst)
+      return;
+
+   pipe_copy_rect(dst, &ps->block, ps->stride, x, y, w, h, src, src_stride, 0, 0);
+
+   pipe_surface_unmap(ps);
+}
+
+
+
+
+/** Convert short in [-32768,32767] to GLfloat in [-1.0,1.0] */
+#define SHORT_TO_FLOAT(S)   ((2.0F * (S) + 1.0F) * (1.0F/65535.0F))
+
+#define UNCLAMPED_FLOAT_TO_SHORT(us, f)  \
+   us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) )
+
+
+
+/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/
+
+static void
+a8r8g8b8_get_tile_rgba(const unsigned *src,
+                       unsigned w, unsigned h,
+                       float *p,
+                       unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         const unsigned pixel = *src++;
+         pRow[0] = ubyte_to_float((pixel >> 16) & 0xff);
+         pRow[1] = ubyte_to_float((pixel >>  8) & 0xff);
+         pRow[2] = ubyte_to_float((pixel >>  0) & 0xff);
+         pRow[3] = ubyte_to_float((pixel >> 24) & 0xff);
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+a8r8g8b8_put_tile_rgba(unsigned *dst,
+                       unsigned w, unsigned h,
+                       const float *p,
+                       unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         unsigned r, g, b, a;
+         r = float_to_ubyte(pRow[0]);
+         g = float_to_ubyte(pRow[1]);
+         b = float_to_ubyte(pRow[2]);
+         a = float_to_ubyte(pRow[3]);
+         *dst++ = (a << 24) | (r << 16) | (g << 8) | b;
+      }
+      p += src_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/
+
+static void
+x8r8g8b8_get_tile_rgba(const unsigned *src,
+                       unsigned w, unsigned h,
+                       float *p,
+                       unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         const unsigned pixel = *src++;
+         pRow[0] = ubyte_to_float((pixel >> 16) & 0xff);
+         pRow[1] = ubyte_to_float((pixel >>  8) & 0xff);
+         pRow[2] = ubyte_to_float((pixel >>  0) & 0xff);
+         pRow[3] = ubyte_to_float(0xff);
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+x8r8g8b8_put_tile_rgba(unsigned *dst,
+                       unsigned w, unsigned h,
+                       const float *p,
+                       unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         unsigned r, g, b;
+         r = float_to_ubyte(pRow[0]);
+         g = float_to_ubyte(pRow[1]);
+         b = float_to_ubyte(pRow[2]);
+         *dst++ = (0xff << 24) | (r << 16) | (g << 8) | b;
+      }
+      p += src_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/
+
+static void
+b8g8r8a8_get_tile_rgba(const unsigned *src,
+                       unsigned w, unsigned h,
+                       float *p,
+                       unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         const unsigned pixel = *src++;
+         pRow[0] = ubyte_to_float((pixel >>  8) & 0xff);
+         pRow[1] = ubyte_to_float((pixel >> 16) & 0xff);
+         pRow[2] = ubyte_to_float((pixel >> 24) & 0xff);
+         pRow[3] = ubyte_to_float((pixel >>  0) & 0xff);
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+b8g8r8a8_put_tile_rgba(unsigned *dst,
+                       unsigned w, unsigned h,
+                       const float *p,
+                       unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         unsigned r, g, b, a;
+         r = float_to_ubyte(pRow[0]);
+         g = float_to_ubyte(pRow[1]);
+         b = float_to_ubyte(pRow[2]);
+         a = float_to_ubyte(pRow[3]);
+         *dst++ = (b << 24) | (g << 16) | (r << 8) | a;
+      }
+      p += src_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/
+
+static void
+a1r5g5b5_get_tile_rgba(const ushort *src,
+                       unsigned w, unsigned h,
+                       float *p,
+                       unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         const ushort pixel = *src++;
+         pRow[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f);
+         pRow[1] = ((pixel >>  5) & 0x1f) * (1.0f / 31.0f);
+         pRow[2] = ((pixel      ) & 0x1f) * (1.0f / 31.0f);
+         pRow[3] = ((pixel >> 15)       ) * 1.0f;
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+a1r5g5b5_put_tile_rgba(ushort *dst,
+                       unsigned w, unsigned h,
+                       const float *p,
+                       unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         unsigned r, g, b, a;
+         r = float_to_ubyte(pRow[0]);
+         g = float_to_ubyte(pRow[1]);
+         b = float_to_ubyte(pRow[2]);
+         a = float_to_ubyte(pRow[3]);
+         r = r >> 3;  /* 5 bits */
+         g = g >> 3;  /* 5 bits */
+         b = b >> 3;  /* 5 bits */
+         a = a >> 7;  /* 1 bit */
+         *dst++ = (a << 15) | (r << 10) | (g << 5) | b;
+      }
+      p += src_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/
+
+static void
+a4r4g4b4_get_tile_rgba(const ushort *src,
+                       unsigned w, unsigned h,
+                       float *p,
+                       unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         const ushort pixel = *src++;
+         pRow[0] = ((pixel >>  8) & 0xf) * (1.0f / 15.0f);
+         pRow[1] = ((pixel >>  4) & 0xf) * (1.0f / 15.0f);
+         pRow[2] = ((pixel      ) & 0xf) * (1.0f / 15.0f);
+         pRow[3] = ((pixel >> 12)      ) * (1.0f / 15.0f);
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+a4r4g4b4_put_tile_rgba(ushort *dst,
+                       unsigned w, unsigned h,
+                       const float *p,
+                       unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         unsigned r, g, b, a;
+         r = float_to_ubyte(pRow[0]);
+         g = float_to_ubyte(pRow[1]);
+         b = float_to_ubyte(pRow[2]);
+         a = float_to_ubyte(pRow[3]);
+         r >>= 4;
+         g >>= 4;
+         b >>= 4;
+         a >>= 4;
+         *dst++ = (a << 12) | (r << 16) | (g << 4) | b;
+      }
+      p += src_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_R5G6B5_UNORM ***/
+
+static void
+r5g6b5_get_tile_rgba(const ushort *src,
+                     unsigned w, unsigned h,
+                     float *p,
+                     unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         const ushort pixel = *src++;
+         pRow[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f);
+         pRow[1] = ((pixel >>  5) & 0x3f) * (1.0f / 63.0f);
+         pRow[2] = ((pixel      ) & 0x1f) * (1.0f / 31.0f);
+         pRow[3] = 1.0f;
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+r5g6b5_put_tile_rgba(ushort *dst,
+                     unsigned w, unsigned h,
+                     const float *p,
+                     unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0);
+         uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0);
+         uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0);
+         *dst++ = (r << 11) | (g << 5) | (b);
+      }
+      p += src_stride;
+   }
+}
+
+
+
+/*** PIPE_FORMAT_Z16_UNORM ***/
+
+/**
+ * Return each Z value as four floats in [0,1].
+ */
+static void
+z16_get_tile_rgba(const ushort *src,
+                  unsigned w, unsigned h,
+                  float *p,
+                  unsigned dst_stride)
+{
+   const float scale = 1.0f / 65535.0f;
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         pRow[0] =
+         pRow[1] =
+         pRow[2] =
+         pRow[3] = *src++ * scale;
+      }
+      p += dst_stride;
+   }
+}
+
+
+
+
+/*** PIPE_FORMAT_L8_UNORM ***/
+
+static void
+l8_get_tile_rgba(const ubyte *src,
+                 unsigned w, unsigned h,
+                 float *p,
+                 unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, src++, pRow += 4) {
+         pRow[0] =
+         pRow[1] =
+         pRow[2] = ubyte_to_float(*src);
+         pRow[3] = 1.0;
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+l8_put_tile_rgba(ubyte *dst,
+                 unsigned w, unsigned h,
+                 const float *p,
+                 unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         unsigned r;
+         r = float_to_ubyte(pRow[0]);
+         *dst++ = (ubyte) r;
+      }
+      p += src_stride;
+   }
+}
+
+
+
+/*** PIPE_FORMAT_A8_UNORM ***/
+
+static void
+a8_get_tile_rgba(const ubyte *src,
+                 unsigned w, unsigned h,
+                 float *p,
+                 unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, src++, pRow += 4) {
+         pRow[0] =
+         pRow[1] =
+         pRow[2] = 0.0;
+         pRow[3] = ubyte_to_float(*src);
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+a8_put_tile_rgba(ubyte *dst,
+                 unsigned w, unsigned h,
+                 const float *p,
+                 unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         unsigned a;
+         a = float_to_ubyte(pRow[3]);
+         *dst++ = (ubyte) a;
+      }
+      p += src_stride;
+   }
+}
+
+
+
+/*** PIPE_FORMAT_R16_SNORM ***/
+
+static void
+r16_get_tile_rgba(const short *src,
+                  unsigned w, unsigned h,
+                  float *p,
+                  unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, src++, pRow += 4) {
+         pRow[0] = SHORT_TO_FLOAT(src[0]);
+         pRow[1] =
+         pRow[2] = 0.0;
+         pRow[3] = 1.0;
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+r16_put_tile_rgba(short *dst,
+                  unsigned w, unsigned h,
+                  const float *p,
+                  unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, dst++, pRow += 4) {
+         UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]);
+      }
+      p += src_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/
+
+static void
+r16g16b16a16_get_tile_rgba(const short *src,
+                           unsigned w, unsigned h,
+                           float *p,
+                           unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, src += 4, pRow += 4) {
+         pRow[0] = SHORT_TO_FLOAT(src[0]);
+         pRow[1] = SHORT_TO_FLOAT(src[1]);
+         pRow[2] = SHORT_TO_FLOAT(src[2]);
+         pRow[3] = SHORT_TO_FLOAT(src[3]);
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+r16g16b16a16_put_tile_rgba(short *dst,
+                           unsigned w, unsigned h,
+                           const float *p,
+                           unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, dst += 4, pRow += 4) {
+         UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]);
+         UNCLAMPED_FLOAT_TO_SHORT(dst[1], pRow[1]);
+         UNCLAMPED_FLOAT_TO_SHORT(dst[2], pRow[2]);
+         UNCLAMPED_FLOAT_TO_SHORT(dst[3], pRow[3]);
+      }
+      p += src_stride;
+   }
+}
+
+
+
+/*** PIPE_FORMAT_I8_UNORM ***/
+
+static void
+i8_get_tile_rgba(const ubyte *src,
+                 unsigned w, unsigned h,
+                 float *p,
+                 unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, src++, pRow += 4) {
+         pRow[0] =
+         pRow[1] =
+         pRow[2] =
+         pRow[3] = ubyte_to_float(*src);
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+i8_put_tile_rgba(ubyte *dst,
+                 unsigned w, unsigned h,
+                 const float *p,
+                 unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         unsigned r;
+         r = float_to_ubyte(pRow[0]);
+         *dst++ = (ubyte) r;
+      }
+      p += src_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_A8L8_UNORM ***/
+
+static void
+a8l8_get_tile_rgba(const ushort *src,
+                   unsigned w, unsigned h,
+                   float *p,
+                   unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         ushort p = *src++;
+         pRow[0] =
+         pRow[1] =
+         pRow[2] = ubyte_to_float(p & 0xff);
+         pRow[3] = ubyte_to_float(p >> 8);
+      }
+      p += dst_stride;
+   }
+}
+
+
+static void
+a8l8_put_tile_rgba(ushort *dst,
+                   unsigned w, unsigned h,
+                   const float *p,
+                   unsigned src_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      const float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         unsigned r, a;
+         r = float_to_ubyte(pRow[0]);
+         a = float_to_ubyte(pRow[3]);
+         *dst++ = (a << 8) | r;
+      }
+      p += src_stride;
+   }
+}
+
+
+
+
+/*** PIPE_FORMAT_Z32_UNORM ***/
+
+/**
+ * Return each Z value as four floats in [0,1].
+ */
+static void
+z32_get_tile_rgba(const unsigned *src,
+                  unsigned w, unsigned h,
+                  float *p,
+                  unsigned dst_stride)
+{
+   const double scale = 1.0 / (double) 0xffffffff;
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         pRow[0] =
+         pRow[1] =
+         pRow[2] =
+         pRow[3] = (float) (*src++ * scale);
+      }
+      p += dst_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_S8Z24_UNORM ***/
+
+/**
+ * Return Z component as four float in [0,1].  Stencil part ignored.
+ */
+static void
+s8z24_get_tile_rgba(const unsigned *src,
+                    unsigned w, unsigned h,
+                    float *p,
+                    unsigned dst_stride)
+{
+   const double scale = 1.0 / ((1 << 24) - 1);
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         pRow[0] =
+         pRow[1] =
+         pRow[2] =
+         pRow[3] = (float) (scale * (*src++ & 0xffffff));
+      }
+      p += dst_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_Z24S8_UNORM ***/
+
+/**
+ * Return Z component as four float in [0,1].  Stencil part ignored.
+ */
+static void
+z24s8_get_tile_rgba(const unsigned *src,
+                    unsigned w, unsigned h,
+                    float *p,
+                    unsigned dst_stride)
+{
+   const double scale = 1.0 / ((1 << 24) - 1);
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         pRow[0] =
+         pRow[1] =
+         pRow[2] =
+         pRow[3] = (float) (scale * (*src++ >> 8));
+      }
+      p += dst_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_Z32_FLOAT ***/
+
+/**
+ * Return each Z value as four floats in [0,1].
+ */
+static void
+z32f_get_tile_rgba(const float *src,
+                   unsigned w, unsigned h,
+                   float *p,
+                   unsigned dst_stride)
+{
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      for (j = 0; j < w; j++, pRow += 4) {
+         pRow[0] =
+         pRow[1] =
+         pRow[2] =
+         pRow[3] = *src++;
+      }
+      p += dst_stride;
+   }
+}
+
+
+/*** PIPE_FORMAT_YCBCR / PIPE_FORMAT_YCBCR_REV ***/
+
+/**
+ * Convert YCbCr (or YCrCb) to RGBA.
+ */
+static void
+ycbcr_get_tile_rgba(const ushort *src,
+                    unsigned w, unsigned h,
+                    float *p,
+                    unsigned dst_stride,
+                    boolean rev)
+{
+   const float scale = 1.0f / 255.0f;
+   unsigned i, j;
+
+   for (i = 0; i < h; i++) {
+      float *pRow = p;
+      /* do two texels at a time */
+      for (j = 0; j < (w & ~1); j += 2, src += 2) {
+         const ushort t0 = src[0];
+         const ushort t1 = src[1];
+         const ubyte y0 = (t0 >> 8) & 0xff;  /* luminance */
+         const ubyte y1 = (t1 >> 8) & 0xff;  /* luminance */
+         ubyte cb, cr;
+         float r, g, b;
+
+         if (rev) {
+            cb = t1 & 0xff;         /* chroma U */
+            cr = t0 & 0xff;         /* chroma V */
+         }
+         else {
+            cb = t0 & 0xff;         /* chroma U */
+            cr = t1 & 0xff;         /* chroma V */
+         }
+
+         /* even pixel: y0,cr,cb */
+         r = 1.164f * (y0-16) + 1.596f * (cr-128);
+         g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128);
+         b = 1.164f * (y0-16) + 2.018f * (cb-128);
+         pRow[0] = r * scale;
+         pRow[1] = g * scale;
+         pRow[2] = b * scale;
+         pRow[3] = 1.0f;
+         pRow += 4;
+
+         /* odd pixel: use y1,cr,cb */
+         r = 1.164f * (y1-16) + 1.596f * (cr-128);
+         g = 1.164f * (y1-16) - 0.813f * (cr-128) - 0.391f * (cb-128);
+         b = 1.164f * (y1-16) + 2.018f * (cb-128);
+         pRow[0] = r * scale;
+         pRow[1] = g * scale;
+         pRow[2] = b * scale;
+         pRow[3] = 1.0f;
+         pRow += 4;
+
+      }
+      /* do the last texel */
+      if (w & 1) {
+         const ushort t0 = src[0];
+         const ushort t1 = src[1];
+         const ubyte y0 = (t0 >> 8) & 0xff;  /* luminance */
+         ubyte cb, cr;
+         float r, g, b;
+
+         if (rev) {
+            cb = t1 & 0xff;         /* chroma U */
+            cr = t0 & 0xff;         /* chroma V */
+         }
+         else {
+            cb = t0 & 0xff;         /* chroma U */
+            cr = t1 & 0xff;         /* chroma V */
+         }
+
+         /* even pixel: y0,cr,cb */
+         r = 1.164f * (y0-16) + 1.596f * (cr-128);
+         g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128);
+         b = 1.164f * (y0-16) + 2.018f * (cb-128);
+         pRow[0] = r * scale;
+         pRow[1] = g * scale;
+         pRow[2] = b * scale;
+         pRow[3] = 1.0f;
+         pRow += 4;
+      }
+      p += dst_stride;
+   }
+}
+
+
+void
+pipe_tile_raw_to_rgba(enum pipe_format format,
+                      void *src,
+                      uint w, uint h,
+                      float *dst, unsigned dst_stride)
+{
+   switch (format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      a8r8g8b8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+      x8r8g8b8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      b8g8r8a8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      a1r5g5b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      a4r4g4b4_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      r5g6b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_L8_UNORM:
+      l8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_A8_UNORM:
+      a8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_I8_UNORM:
+      i8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_A8L8_UNORM:
+      a8l8_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_R16_SNORM:
+      r16_get_tile_rgba((short *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      r16g16b16a16_get_tile_rgba((short *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_Z16_UNORM:
+      z16_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_Z32_UNORM:
+      z32_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+      s8z24_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_Z24S8_UNORM:
+      z24s8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_Z32_FLOAT:
+      z32f_get_tile_rgba((float *) src, w, h, dst, dst_stride);
+      break;
+   case PIPE_FORMAT_YCBCR:
+      ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, FALSE);
+      break;
+   case PIPE_FORMAT_YCBCR_REV:
+      ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, TRUE);
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+void
+pipe_get_tile_rgba(struct pipe_surface *ps,
+                   uint x, uint y, uint w, uint h,
+                   float *p)
+{
+   unsigned dst_stride = w * 4;
+   void *packed;
+
+   if (pipe_clip_tile(x, y, &w, &h, ps))
+      return;
+
+   packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size);
+
+   if (!packed)
+      return;
+
+   if(ps->format == PIPE_FORMAT_YCBCR || ps->format == PIPE_FORMAT_YCBCR_REV)
+      assert((x & 1) == 0);
+
+   pipe_get_tile_raw(ps, x, y, w, h, packed, 0);
+
+   pipe_tile_raw_to_rgba(ps->format, packed, w, h, p, dst_stride);
+
+   FREE(packed);
+}
+
+
+void
+pipe_put_tile_rgba(struct pipe_surface *ps,
+                   uint x, uint y, uint w, uint h,
+                   const float *p)
+{
+   unsigned src_stride = w * 4;
+   void *packed;
+
+   if (pipe_clip_tile(x, y, &w, &h, ps))
+      return;
+
+   packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size);
+
+   if (!packed)
+      return;
+
+   switch (ps->format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+      x8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      r5g6b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      assert(0);
+      break;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      a4r4g4b4_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_L8_UNORM:
+      l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_A8_UNORM:
+      a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_I8_UNORM:
+      i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_A8L8_UNORM:
+      a8l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_R16_SNORM:
+      r16_put_tile_rgba((short *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride);
+      break;
+   case PIPE_FORMAT_Z16_UNORM:
+      /*z16_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/
+      break;
+   case PIPE_FORMAT_Z32_UNORM:
+      /*z32_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/
+      break;
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+      /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/
+      break;
+   case PIPE_FORMAT_Z24S8_UNORM:
+      /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/
+      break;
+   default:
+      assert(0);
+   }
+
+   pipe_put_tile_raw(ps, x, y, w, h, packed, 0);
+
+   FREE(packed);
+}
+
+
+/**
+ * Get a block of Z values, converted to 32-bit range.
+ */
+void
+pipe_get_tile_z(struct pipe_surface *ps,
+                uint x, uint y, uint w, uint h,
+                uint *z)
+{
+   const uint dstStride = w;
+   ubyte *map;
+   uint *pDest = z;
+   uint i, j;
+
+   if (pipe_clip_tile(x, y, &w, &h, ps))
+      return;
+
+   map = (ubyte *)pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ);
+   if (!map) {
+      assert(0);
+      return;
+   }
+
+   switch (ps->format) {
+   case PIPE_FORMAT_Z32_UNORM:
+      {
+         const uint *pSrc
+            = (const uint *)(map  + y * ps->stride + x*4);
+         for (i = 0; i < h; i++) {
+            memcpy(pDest, pSrc, 4 * w);
+            pDest += dstStride;
+            pSrc += ps->stride/4;
+         }
+      }
+      break;
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+      {
+         const uint *pSrc
+            = (const uint *)(map + y * ps->stride + x*4);
+         for (i = 0; i < h; i++) {
+            for (j = 0; j < w; j++) {
+               /* convert 24-bit Z to 32-bit Z */
+               pDest[j] = (pSrc[j] << 8) | (pSrc[j] & 0xff);
+            }
+            pDest += dstStride;
+            pSrc += ps->stride/4;
+         }
+      }
+      break;
+   case PIPE_FORMAT_Z16_UNORM:
+      {
+         const ushort *pSrc
+            = (const ushort *)(map + y * ps->stride + x*2);
+         for (i = 0; i < h; i++) {
+            for (j = 0; j < w; j++) {
+               /* convert 16-bit Z to 32-bit Z */
+               pDest[j] = (pSrc[j] << 16) | pSrc[j];
+            }
+            pDest += dstStride;
+            pSrc += ps->stride/2;
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+
+   pipe_surface_unmap(ps);
+}
+
+
+void
+pipe_put_tile_z(struct pipe_surface *ps,
+                uint x, uint y, uint w, uint h,
+                const uint *zSrc)
+{
+   const uint srcStride = w;
+   const uint *pSrc = zSrc;
+   ubyte *map;
+   uint i, j;
+
+   if (pipe_clip_tile(x, y, &w, &h, ps))
+      return;
+
+   map = (ubyte *)pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE);
+   if (!map) {
+      assert(0);
+      return;
+   }
+
+   switch (ps->format) {
+   case PIPE_FORMAT_Z32_UNORM:
+      {
+         uint *pDest = (uint *) (map + y * ps->stride + x*4);
+         for (i = 0; i < h; i++) {
+            memcpy(pDest, pSrc, 4 * w);
+            pDest += ps->stride/4;
+            pSrc += srcStride;
+         }
+      }
+      break;
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+      {
+         uint *pDest = (uint *) (map + y * ps->stride + x*4);
+         for (i = 0; i < h; i++) {
+            for (j = 0; j < w; j++) {
+               /* convert 32-bit Z to 24-bit Z (0 stencil) */
+               pDest[j] = pSrc[j] >> 8;
+            }
+            pDest += ps->stride/4;
+            pSrc += srcStride;
+         }
+      }
+      break;
+   case PIPE_FORMAT_Z16_UNORM:
+      {
+         ushort *pDest = (ushort *) (map + y * ps->stride + x*2);
+         for (i = 0; i < h; i++) {
+            for (j = 0; j < w; j++) {
+               /* convert 32-bit Z to 16-bit Z */
+               pDest[j] = pSrc[j] >> 16;
+            }
+            pDest += ps->stride/2;
+            pSrc += srcStride;
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+
+   pipe_surface_unmap(ps);
+}
+
+
diff --git a/src/gallium/auxiliary/util/u_tile.h b/src/gallium/auxiliary/util/u_tile.h
new file mode 100644
index 0000000000..a8ac805308
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_tile.h
@@ -0,0 +1,101 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef P_TILE_H
+#define P_TILE_H
+
+#include "pipe/p_compiler.h"
+
+struct pipe_surface;
+
+
+/**
+ * Clip tile against surface dims.
+ * \return TRUE if tile is totally clipped, FALSE otherwise
+ */
+static INLINE boolean
+pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_surface *ps)
+{
+   if (x >= ps->width)
+      return TRUE;
+   if (y >= ps->height)
+      return TRUE;
+   if (x + *w > ps->width)
+      *w = ps->width - x;
+   if (y + *h > ps->height)
+      *h = ps->height - y;
+   return FALSE;
+}
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void
+pipe_get_tile_raw(struct pipe_surface *ps,
+                  uint x, uint y, uint w, uint h,
+                  void *p, int dst_stride);
+
+void
+pipe_put_tile_raw(struct pipe_surface *ps,
+                  uint x, uint y, uint w, uint h,
+                  const void *p, int src_stride);
+
+
+void
+pipe_get_tile_rgba(struct pipe_surface *ps,
+                   uint x, uint y, uint w, uint h,
+                   float *p);
+
+void
+pipe_put_tile_rgba(struct pipe_surface *ps,
+                   uint x, uint y, uint w, uint h,
+                   const float *p);
+
+
+void
+pipe_get_tile_z(struct pipe_surface *ps,
+                uint x, uint y, uint w, uint h,
+                uint *z);
+
+void
+pipe_put_tile_z(struct pipe_surface *ps,
+                uint x, uint y, uint w, uint h,
+                const uint *z);
+
+void
+pipe_tile_raw_to_rgba(enum pipe_format format,
+                      void *src,
+                      uint w, uint h,
+                      float *dst, unsigned dst_stride);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c
new file mode 100644
index 0000000000..dde2c74fa8
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_time.c
@@ -0,0 +1,220 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * OS independent time-manipulation functions.
+ * 
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_OS_LINUX)
+#include <sys/time.h>
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+#include <windows.h>
+#include <winddi.h>
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+#include <windows.h>
+extern VOID KeQuerySystemTime(PLARGE_INTEGER);
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+#include <windows.h>
+#else
+#error Unsupported OS
+#endif
+
+#include "util/u_time.h"
+
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+
+static int64_t frequency = 0;
+
+static INLINE void 
+util_time_get_frequency(void)
+{
+   if(!frequency) {
+#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+      LONGLONG temp;
+      EngQueryPerformanceFrequency(&temp);
+      frequency = temp;
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+      LARGE_INTEGER temp;
+      QueryPerformanceFrequency(&temp);
+      frequency = temp.QuadPart;
+#endif
+   }
+}
+#endif
+
+
+void 
+util_time_get(struct util_time *t)
+{
+#if defined(PIPE_OS_LINUX)
+   gettimeofday(&t->tv, NULL);
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+   LONGLONG temp;
+   EngQueryPerformanceCounter(&temp);
+   t->counter = temp;
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+   /* Updated every 10 miliseconds, measured in units of 100 nanoseconds.
+    * http://msdn.microsoft.com/en-us/library/ms801642.aspx */
+   LARGE_INTEGER temp;
+   KeQuerySystemTime(&temp);
+   t->counter = temp.QuadPart;
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+   LARGE_INTEGER temp;
+   QueryPerformanceCounter(&temp);
+   t->counter = temp.QuadPart;
+#endif
+}
+
+
+void 
+util_time_add(const struct util_time *t1,
+              int64_t usecs,
+              struct util_time *t2)
+{
+#if defined(PIPE_OS_LINUX)
+   t2->tv.tv_sec = t1->tv.tv_sec + usecs / 1000000;
+   t2->tv.tv_usec = t1->tv.tv_usec + usecs % 1000000;
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+   util_time_get_frequency();
+   t2->counter = t1->counter + (usecs * frequency + INT64_C(999999))/INT64_C(1000000);
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+   /* 1 tick = 100 nano seconds. */
+   t2->counter = t1->counter + usecs * 10;
+#else
+   LARGE_INTEGER temp;
+   LONGLONG freq;
+   freq = temp.QuadPart;
+   t2->counter = t1->counter + (usecs * freq)/1000000L;
+#endif
+}
+
+
+int64_t
+util_time_diff(const struct util_time *t1, 
+               const struct util_time *t2)
+{
+#if defined(PIPE_OS_LINUX)
+   return (t2->tv.tv_usec - t1->tv.tv_usec) + 
+          (t2->tv.tv_sec - t1->tv.tv_sec)*1000000;
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+   util_time_get_frequency();
+   return (t2->counter - t1->counter)*INT64_C(1000000)/frequency;
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+   return (t2->counter - t1->counter)/10;
+#endif
+}
+
+
+
+uint64_t
+util_time_micros( void )
+{
+   struct util_time t1;
+   
+   util_time_get(&t1);
+   
+#if defined(PIPE_OS_LINUX)
+   return t1.tv.tv_usec + t1.tv.tv_sec*1000000LL;
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+   util_time_get_frequency();
+   return t1.counter*INT64_C(1000000)/frequency;
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+   return t1.counter/10;
+#endif
+}
+
+
+
+/**
+ * Compare two time values.
+ * 
+ * Not publicly available because it does not take in account wrap-arounds. 
+ * Use util_time_timeout instead.
+ */
+static INLINE int
+util_time_compare(const struct util_time *t1, 
+                  const struct util_time *t2)
+{
+#if defined(PIPE_OS_LINUX)
+   if (t1->tv.tv_sec < t2->tv.tv_sec)
+      return -1;
+   else if(t1->tv.tv_sec > t2->tv.tv_sec)
+      return 1;
+   else if (t1->tv.tv_usec < t2->tv.tv_usec)
+      return -1;
+   else if(t1->tv.tv_usec > t2->tv.tv_usec)
+      return 1;
+   else 
+      return 0;
+#elif defined(PIPE_OS_WINDOWS)
+   if (t1->counter < t2->counter)
+      return -1;
+   else if(t1->counter > t2->counter)
+      return 1;
+   else 
+      return 0;
+#endif
+}
+
+
+boolean 
+util_time_timeout(const struct util_time *start, 
+                  const struct util_time *end,
+                  const struct util_time *curr) 
+{
+   if(util_time_compare(start, end) <= 0)
+      return !(util_time_compare(start, curr) <= 0 && util_time_compare(curr, end) < 0);
+   else
+      return !(util_time_compare(start, curr) <= 0 || util_time_compare(curr, end) < 0);
+}
+
+
+#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+void util_time_sleep(unsigned usecs)
+{
+   LONGLONG start, curr, end;
+   
+   EngQueryPerformanceCounter(&start);
+   
+   if(!frequency)
+      EngQueryPerformanceFrequency(&frequency);
+   
+   end = start + (usecs * frequency + 999999LL)/1000000LL;
+   
+   do {
+      EngQueryPerformanceCounter(&curr);
+   } while(start <= curr && curr < end || 
+	   end < start && (curr < end || start <= curr));
+}
+#endif
diff --git a/src/gallium/auxiliary/util/u_time.h b/src/gallium/auxiliary/util/u_time.h
new file mode 100644
index 0000000000..35d97d16c7
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_time.h
@@ -0,0 +1,104 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * OS independent time-manipulation functions.
+ * 
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef U_TIME_H_
+#define U_TIME_H_
+
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_OS_LINUX)
+#include <time.h> /* timeval */
+#include <unistd.h> /* usleep */
+#endif
+
+#include "pipe/p_compiler.h"
+
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+
+/**
+ * Time abstraction.
+ * 
+ * Do not access this structure directly. Use the provided function instead.
+ */
+struct util_time 
+{
+#if defined(PIPE_OS_LINUX)
+   struct timeval tv;
+#else
+   int64_t counter;
+#endif
+};
+   
+
+void 
+util_time_get(struct util_time *t);
+
+void 
+util_time_add(const struct util_time *t1,
+              int64_t usecs,
+              struct util_time *t2);
+
+uint64_t
+util_time_micros( void );
+
+int64_t
+util_time_diff(const struct util_time *t1, 
+               const struct util_time *t2);
+
+/**
+ * Returns non-zero when the timeout expires.
+ */
+boolean 
+util_time_timeout(const struct util_time *start, 
+                  const struct util_time *end,
+                  const struct util_time *curr);
+
+#if defined(PIPE_OS_LINUX)
+#define util_time_sleep usleep
+#else
+void
+util_time_sleep(unsigned usecs);
+#endif
+
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif /* U_TIME_H_ */
diff --git a/src/gallium/auxiliary/util/u_timed_winsys.c b/src/gallium/auxiliary/util/u_timed_winsys.c
new file mode 100644
index 0000000000..f237e12d73
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_timed_winsys.c
@@ -0,0 +1,316 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+/*
+ * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com>
+ */
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "u_timed_winsys.h"
+#include "util/u_memory.h"
+#include "util/u_time.h"
+
+
+struct timed_winsys {
+   struct pipe_winsys base;
+   struct pipe_winsys *backend;
+   uint64_t last_dump;
+   struct {
+      const char *name_key;
+      double total;
+      unsigned calls;
+   } funcs[13];
+};
+
+
+static struct timed_winsys *timed_winsys( struct pipe_winsys *winsys )
+{
+   return (struct timed_winsys *)winsys;
+}
+
+
+static uint64_t time_start( void )
+{
+   return util_time_micros();
+}
+
+
+static void time_display( struct pipe_winsys *winsys )
+{
+   struct timed_winsys *tws = timed_winsys(winsys);
+   unsigned i;
+   double overall = 0;
+
+   for (i = 0; i < Elements(tws->funcs); i++) {
+      if (tws->funcs[i].name_key) {
+         debug_printf("*** %-25s %5.3fms (%d calls, avg %.3fms)\n", 
+                      tws->funcs[i].name_key,
+                      tws->funcs[i].total,
+                      tws->funcs[i].calls,
+                      tws->funcs[i].total / tws->funcs[i].calls);
+         overall += tws->funcs[i].total;
+         tws->funcs[i].calls = 0;
+         tws->funcs[i].total = 0;
+      }
+   }
+
+   debug_printf("*** %-25s %5.3fms\n", 
+                "OVERALL WINSYS",
+                overall);
+}
+
+static void time_finish( struct pipe_winsys *winsys,
+                         long long startval, 
+                         unsigned idx,
+                         const char *name ) 
+{
+   struct timed_winsys *tws = timed_winsys(winsys);
+   uint64_t endval = util_time_micros();
+   double elapsed = (endval - startval)/1000.0;
+
+   if (endval - startval > 1000LL) 
+      debug_printf("*** %s %.3f\n", name, elapsed );
+
+   assert( tws->funcs[idx].name_key == name ||
+           tws->funcs[idx].name_key == NULL);
+
+   tws->funcs[idx].name_key = name;
+   tws->funcs[idx].total += elapsed;
+   tws->funcs[idx].calls++;
+
+   if (endval - tws->last_dump > 10LL * 1000LL * 1000LL) {
+      time_display( winsys );
+      tws->last_dump = endval;
+   }
+}
+
+
+/* Pipe has no concept of pools, but the psb driver passes a flag that
+ * can be mapped onto pools in the backend.
+ */
+static struct pipe_buffer *
+timed_buffer_create(struct pipe_winsys *winsys, 
+                    unsigned alignment, 
+                    unsigned usage, 
+                    unsigned size )
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   struct pipe_buffer *buf =
+      backend->buffer_create( backend, alignment, usage, size );
+
+   time_finish(winsys, start, 0, __FUNCTION__);
+   
+   return buf;
+}
+
+
+
+
+static struct pipe_buffer *
+timed_user_buffer_create(struct pipe_winsys *winsys,
+                             void *data, 
+                             unsigned bytes) 
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   struct pipe_buffer *buf = backend->user_buffer_create( backend, data, bytes );
+
+   time_finish(winsys, start, 1, __FUNCTION__);
+   
+   return buf;
+}
+
+
+static void *
+timed_buffer_map(struct pipe_winsys *winsys,
+                     struct pipe_buffer *buf,
+                     unsigned flags)
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   void *map = backend->buffer_map( backend, buf, flags );
+
+   time_finish(winsys, start, 2, __FUNCTION__);
+   
+   return map;
+}
+
+
+static void
+timed_buffer_unmap(struct pipe_winsys *winsys,
+                       struct pipe_buffer *buf)
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   backend->buffer_unmap( backend, buf );
+
+   time_finish(winsys, start, 3, __FUNCTION__);
+}
+
+
+static void
+timed_buffer_destroy(struct pipe_winsys *winsys,
+                         struct pipe_buffer *buf)
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   backend->buffer_destroy( backend, buf );
+
+   time_finish(winsys, start, 4, __FUNCTION__);
+}
+
+
+static void
+timed_flush_frontbuffer( struct pipe_winsys *winsys,
+                         struct pipe_surface *surf,
+                         void *context_private)
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   backend->flush_frontbuffer( backend, surf, context_private );
+
+   time_finish(winsys, start, 5, __FUNCTION__);
+}
+
+
+
+
+static struct pipe_buffer *
+timed_surface_buffer_create(struct pipe_winsys *winsys,
+                              unsigned width, unsigned height,
+                              enum pipe_format format, 
+                              unsigned usage,
+                              unsigned *stride)
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   struct pipe_buffer *ret = backend->surface_buffer_create( backend, width, height, 
+                                                             format, usage, stride );
+
+   time_finish(winsys, start, 7, __FUNCTION__);
+   
+   return ret;
+}
+
+
+static const char *
+timed_get_name( struct pipe_winsys *winsys )
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   const char *ret = backend->get_name( backend );
+
+   time_finish(winsys, start, 9, __FUNCTION__);
+   
+   return ret;
+}
+
+static void
+timed_fence_reference(struct pipe_winsys *winsys,
+                    struct pipe_fence_handle **ptr,
+                    struct pipe_fence_handle *fence)
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   backend->fence_reference( backend, ptr, fence );
+
+   time_finish(winsys, start, 10, __FUNCTION__);
+}
+
+
+static int
+timed_fence_signalled( struct pipe_winsys *winsys,
+                       struct pipe_fence_handle *fence,
+                       unsigned flag )
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   int ret = backend->fence_signalled( backend, fence, flag );
+
+   time_finish(winsys, start, 11, __FUNCTION__);
+   
+   return ret;
+}
+
+static int
+timed_fence_finish( struct pipe_winsys *winsys,
+                     struct pipe_fence_handle *fence,
+                     unsigned flag )
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   uint64_t start = time_start();
+
+   int ret = backend->fence_finish( backend, fence, flag );
+
+   time_finish(winsys, start, 12, __FUNCTION__);
+   
+   return ret;
+}
+
+static void
+timed_winsys_destroy( struct pipe_winsys *winsys )
+{
+   struct pipe_winsys *backend = timed_winsys(winsys)->backend;
+   backend->destroy( backend );
+   FREE(winsys);
+}
+
+
+
+struct pipe_winsys *u_timed_winsys_create( struct pipe_winsys *backend )
+{
+   struct timed_winsys *ws = CALLOC_STRUCT(timed_winsys);
+   
+   ws->base.user_buffer_create = timed_user_buffer_create;
+   ws->base.buffer_map = timed_buffer_map;
+   ws->base.buffer_unmap = timed_buffer_unmap;
+   ws->base.buffer_destroy = timed_buffer_destroy;
+   ws->base.buffer_create = timed_buffer_create;
+   ws->base.surface_buffer_create = timed_surface_buffer_create;
+   ws->base.flush_frontbuffer = timed_flush_frontbuffer;
+   ws->base.get_name = timed_get_name;
+   ws->base.fence_reference = timed_fence_reference;
+   ws->base.fence_signalled = timed_fence_signalled;
+   ws->base.fence_finish = timed_fence_finish;
+   ws->base.destroy = timed_winsys_destroy;
+   
+   ws->backend = backend;
+
+   return &ws->base;
+}
+
diff --git a/src/gallium/auxiliary/util/u_timed_winsys.h b/src/gallium/auxiliary/util/u_timed_winsys.h
new file mode 100644
index 0000000000..542365112d
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_timed_winsys.h
@@ -0,0 +1,41 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+/*
+ * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com>
+ */
+
+
+#ifndef U_TIMED_WINSYS_H
+#define U_TIMED_WINSYS_H
+
+
+struct pipe_winsys;
+struct pipe_winsys *u_timed_winsys_create( struct pipe_winsys *backend );
+
+
+#endif
diff --git a/src/gallium/drivers/Makefile b/src/gallium/drivers/Makefile
new file mode 100644
index 0000000000..6161cb6ff8
--- /dev/null
+++ b/src/gallium/drivers/Makefile
@@ -0,0 +1,20 @@
+TOP = ../../..
+include $(TOP)/configs/current
+
+
+SUBDIRS = $(GALLIUM_DRIVER_DIRS)
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	rm -f `find . -name \*.[oa]`
diff --git a/src/gallium/drivers/cell/Makefile b/src/gallium/drivers/cell/Makefile
new file mode 100644
index 0000000000..47aef7b05f
--- /dev/null
+++ b/src/gallium/drivers/cell/Makefile
@@ -0,0 +1,12 @@
+# Cell Gallium driver Makefile
+
+
+default:
+	( cd spu ; make )
+	( cd ppu ; make )
+
+
+
+clean:
+	( cd spu ; make clean )
+	( cd ppu ; make clean )
diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h
new file mode 100644
index 0000000000..1f6860da11
--- /dev/null
+++ b/src/gallium/drivers/cell/common.h
@@ -0,0 +1,376 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Types and tokens which are common to the SPU and PPU code.
+ */
+
+
+#ifndef CELL_COMMON_H
+#define CELL_COMMON_H
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+
+
+/** The standard assert macro doesn't seem to work reliably */
+#define ASSERT(x) \
+   if (!(x)) { \
+      ubyte *p = NULL; \
+      fprintf(stderr, "%s:%d: %s(): assertion %s failed.\n", \
+              __FILE__, __LINE__, __FUNCTION__, #x);             \
+      *p = 0; \
+      exit(1); \
+   }
+
+
+
+#define JOIN(x, y) JOIN_AGAIN(x, y)
+#define JOIN_AGAIN(x, y) x ## y
+
+#define STATIC_ASSERT(e) \
+{typedef char JOIN(assertion_failed_at_line_, __LINE__) [(e) ? 1 : -1];}
+
+
+
+/** for sanity checking */
+#define ASSERT_ALIGN16(ptr) \
+  ASSERT((((unsigned long) (ptr)) & 0xf) == 0);
+
+
+/** round up value to next multiple of 4 */
+#define ROUNDUP4(k)  (((k) + 0x3) & ~0x3)
+
+/** round up value to next multiple of 8 */
+#define ROUNDUP8(k)  (((k) + 0x7) & ~0x7)
+
+/** round up value to next multiple of 16 */
+#define ROUNDUP16(k)  (((k) + 0xf) & ~0xf)
+
+
+#define CELL_MAX_SPUS 8
+
+#define CELL_MAX_SAMPLERS 4
+#define CELL_MAX_TEXTURE_LEVELS 12  /* 2k x 2k */
+#define CELL_MAX_CONSTANTS 32  /**< number of float[4] constants */
+#define CELL_MAX_WIDTH 1024    /**< max framebuffer width */
+#define CELL_MAX_HEIGHT 1024   /**< max framebuffer width */
+
+#define TILE_SIZE 32
+
+
+/**
+ * The low byte of a mailbox word contains the command opcode.
+ * Remaining higher bytes are command specific.
+ */
+#define CELL_CMD_OPCODE_MASK 0xff
+
+#define CELL_CMD_EXIT                 1
+#define CELL_CMD_CLEAR_SURFACE        2
+#define CELL_CMD_FINISH               3
+#define CELL_CMD_RENDER               4
+#define CELL_CMD_BATCH                5
+#define CELL_CMD_RELEASE_VERTS        6
+#define CELL_CMD_STATE_FRAMEBUFFER   10
+#define CELL_CMD_STATE_FRAGMENT_OPS  11
+#define CELL_CMD_STATE_SAMPLER       12
+#define CELL_CMD_STATE_TEXTURE       13
+#define CELL_CMD_STATE_VERTEX_INFO   14
+#define CELL_CMD_STATE_VIEWPORT      15
+#define CELL_CMD_STATE_UNIFORMS      16
+#define CELL_CMD_STATE_VS_ARRAY_INFO 17
+#define CELL_CMD_STATE_BIND_VS       18
+#define CELL_CMD_STATE_FRAGMENT_PROGRAM 19
+#define CELL_CMD_STATE_ATTRIB_FETCH  20
+#define CELL_CMD_STATE_FS_CONSTANTS  21
+#define CELL_CMD_STATE_RASTERIZER    22
+#define CELL_CMD_VS_EXECUTE          23
+#define CELL_CMD_FLUSH_BUFFER_RANGE  24
+#define CELL_CMD_FENCE               25
+
+
+/** Command/batch buffers */
+#define CELL_NUM_BUFFERS 4
+#define CELL_BUFFER_SIZE (4*1024)  /**< 16KB would be the max */
+
+#define CELL_BUFFER_STATUS_FREE 10
+#define CELL_BUFFER_STATUS_USED 20
+
+/** Debug flags */
+#define CELL_DEBUG_CHECKER              (1 << 0)
+#define CELL_DEBUG_ASM                  (1 << 1)
+#define CELL_DEBUG_SYNC                 (1 << 2)
+#define CELL_DEBUG_FRAGMENT_OPS         (1 << 3)
+#define CELL_DEBUG_FRAGMENT_OP_FALLBACK (1 << 4)
+#define CELL_DEBUG_CMD                  (1 << 5)
+#define CELL_DEBUG_CACHE                (1 << 6)
+
+#define CELL_FENCE_IDLE      0
+#define CELL_FENCE_EMITTED   1
+#define CELL_FENCE_SIGNALLED 2
+
+#define CELL_FACING_FRONT    0
+#define CELL_FACING_BACK     1
+
+struct cell_fence
+{
+   /** There's a 16-byte status qword per SPU */
+   volatile uint status[CELL_MAX_SPUS][4];
+};
+
+#ifdef __SPU__
+typedef vector unsigned int opcode_t;
+#else
+typedef unsigned int opcode_t[4];
+#endif
+
+/**
+ * Fence command sent to SPUs.  In response, the SPUs will write
+ * CELL_FENCE_STATUS_SIGNALLED back to the fence status word in main memory.
+ */
+struct cell_command_fence
+{
+   opcode_t opcode;      /**< CELL_CMD_FENCE */
+   struct cell_fence *fence;
+   uint32_t pad_[3];
+};
+
+
+/**
+ * Command to specify per-fragment operations state and generated code.
+ * Note that this is a variant-length structure, allocated with as 
+ * much memory as needed to hold the generated code; the "code"
+ * field *must* be the last field in the structure.  Also, the entire
+ * length of the structure (including the variant code field) must be
+ * a multiple of 8 bytes; we require that this structure itself be
+ * a multiple of 8 bytes, and that the generated code also be a multiple
+ * of 8 bytes.
+ *
+ * Also note that the dsa, blend, blend_color fields are really only needed
+ * for the fallback/C per-pixel code.  They're not used when we generate
+ * dynamic SPU fragment code (which is the normal case), and will eventually
+ * be removed from this structure.
+ */
+struct cell_command_fragment_ops
+{
+   opcode_t opcode;      /**< CELL_CMD_STATE_FRAGMENT_OPS */
+
+   /* Fields for the fallback case */
+   struct pipe_depth_stencil_alpha_state dsa;
+   struct pipe_blend_state blend;
+   struct pipe_blend_color blend_color;
+
+   /* Fields for the generated SPU code */
+   unsigned total_code_size;
+   unsigned front_code_index;
+   unsigned back_code_index;
+   /* this field has variant length, and must be the last field in 
+    * the structure
+    */
+   unsigned code[0];
+};
+
+
+/** Max instructions for fragment programs */
+#define SPU_MAX_FRAGMENT_PROGRAM_INSTS 512
+
+/**
+ * Command to send a fragment program to SPUs.
+ */
+struct cell_command_fragment_program
+{
+   opcode_t opcode;      /**< CELL_CMD_STATE_FRAGMENT_PROGRAM */
+   uint num_inst;        /**< Number of instructions */
+   uint32_t pad[3];
+   unsigned code[SPU_MAX_FRAGMENT_PROGRAM_INSTS];
+};
+
+
+/**
+ * Tell SPUs about the framebuffer size, location
+ */
+struct cell_command_framebuffer
+{
+   opcode_t opcode;     /**< CELL_CMD_STATE_FRAMEBUFFER */
+   int width, height;
+   void *color_start, *depth_start;
+   enum pipe_format color_format, depth_format;
+   uint32_t pad_[2];
+};
+
+
+/**
+ * Tell SPUs about rasterizer state.
+ */
+struct cell_command_rasterizer
+{
+   opcode_t opcode;    /**< CELL_CMD_STATE_RASTERIZER */
+   struct pipe_rasterizer_state rasterizer;
+};
+
+
+/**
+ * Clear framebuffer to the given value/color.
+ */
+struct cell_command_clear_surface
+{
+   opcode_t opcode;     /**< CELL_CMD_CLEAR_SURFACE */
+   uint surface; /**< Temporary: 0=color, 1=Z */
+   uint value;
+   uint32_t pad[2];
+};
+
+
+/**
+ * Array info used by the vertex shader's vertex puller.
+ */
+struct cell_array_info
+{
+   uint64_t base;      /**< Base address of the 0th element. */
+   uint attr;          /**< Attribute that this state is for. */
+   uint pitch;         /**< Byte pitch from one entry to the next. */
+   uint size;
+   uint function_offset;
+};
+
+
+struct cell_attribute_fetch_code
+{
+   uint64_t base;
+   uint size;
+};
+
+
+struct cell_buffer_range
+{
+   uint64_t base;
+   unsigned size;
+};
+
+
+struct cell_shader_info
+{
+   uint64_t declarations;
+   uint64_t instructions;
+   uint64_t  immediates;
+
+   unsigned num_outputs;
+   unsigned num_declarations;
+   unsigned num_instructions;
+   unsigned num_immediates;
+};
+
+
+#define SPU_VERTS_PER_BATCH 64
+struct cell_command_vs
+{
+   opcode_t opcode;       /**< CELL_CMD_VS_EXECUTE */
+   uint64_t vOut[SPU_VERTS_PER_BATCH];
+   unsigned num_elts;
+   unsigned elts[SPU_VERTS_PER_BATCH];
+   float plane[12][4];
+   unsigned nr_planes;
+   unsigned nr_attrs;
+};
+
+
+struct cell_command_render
+{
+   opcode_t opcode;   /**< CELL_CMD_RENDER */
+   uint prim_type;    /**< PIPE_PRIM_x */
+   uint num_verts;
+   uint vertex_size;  /**< bytes per vertex */
+   uint num_indexes;
+   uint vertex_buf;  /**< which cell->buffer[] contains the vertex data */
+   float xmin, ymin, xmax, ymax;  /* XXX another dummy field */
+   uint min_index;
+   boolean inline_verts;
+   uint32_t pad_[1];
+};
+
+
+struct cell_command_release_verts
+{
+   opcode_t opcode;         /**< CELL_CMD_RELEASE_VERTS */
+   uint vertex_buf;    /**< in [0, CELL_NUM_BUFFERS-1] */
+   uint32_t pad_[3];
+};
+
+
+struct cell_command_sampler
+{
+   opcode_t opcode;         /**< CELL_CMD_STATE_SAMPLER */
+   uint unit;
+   struct pipe_sampler_state state;
+   uint32_t pad_[1];
+};
+
+
+struct cell_command_texture
+{
+   opcode_t opcode;     /**< CELL_CMD_STATE_TEXTURE */
+   uint target;         /**< PIPE_TEXTURE_x */
+   uint unit;
+   void *start[CELL_MAX_TEXTURE_LEVELS];   /**< Address in main memory */
+   ushort width[CELL_MAX_TEXTURE_LEVELS];
+   ushort height[CELL_MAX_TEXTURE_LEVELS];
+   ushort depth[CELL_MAX_TEXTURE_LEVELS];
+};
+
+
+#define MAX_SPU_FUNCTIONS 12
+/**
+ * Used to tell the PPU about the address of particular functions in the
+ * SPU's address space.
+ */
+struct cell_spu_function_info
+{
+   uint num;
+   char names[MAX_SPU_FUNCTIONS][16];
+   uint addrs[MAX_SPU_FUNCTIONS];
+   char pad[12];   /**< Pad struct to multiple of 16 bytes (256 currently) */
+};
+
+
+/** This is the object passed to spe_create_thread() */
+struct cell_init_info
+{
+   unsigned id;
+   unsigned num_spus;
+   unsigned debug_flags;  /**< mask of CELL_DEBUG_x flags */
+   float inv_timebase;    /**< 1.0/timebase, for perf measurement */
+
+   /** Buffers for command batches, vertex/index data */
+   ubyte *buffers[CELL_NUM_BUFFERS];
+   uint *buffer_status;  /**< points at cell_context->buffer_status */
+
+   struct cell_spu_function_info *spu_functions;
+} ALIGN16_ATTRIB;
+
+
+#endif /* CELL_COMMON_H */
diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile
new file mode 100644
index 0000000000..c92f8e5cba
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/Makefile
@@ -0,0 +1,86 @@
+# Gallium3D Cell driver: PPU code
+
+# This makefile builds the libcell.a library which gets pulled into
+# the main libGL.so library
+
+
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+
+# This is the "top-level" cell PPU driver code, will get pulled into libGL.so
+# by the winsys Makefile.
+CELL_LIB = ../libcell.a
+
+
+# This is the SPU code.  We'd like to be able to put this into the libcell.a
+# archive with the PPU code, but nesting .a libs doesn't seem to work.
+# So, it's pulled into libGL.so in gallium/winsys/xlib/Makefile
+SPU_CODE_MODULE = ../spu/g3d_spu.a
+
+
+SOURCES = \
+	cell_batch.c \
+	cell_clear.c \
+	cell_context.c \
+	cell_draw_arrays.c \
+	cell_fence.c \
+	cell_flush.c \
+	cell_gen_fragment.c \
+	cell_gen_fp.c \
+	cell_state_derived.c \
+	cell_state_emit.c \
+	cell_state_shader.c \
+	cell_pipe_state.c \
+	cell_screen.c \
+	cell_state_vertex.c \
+	cell_spu.c \
+	cell_surface.c \
+	cell_texture.c \
+	cell_vbuf.c \
+	cell_vertex_fetch.c \
+	cell_vertex_shader.c
+
+
+OBJECTS = $(SOURCES:.c=.o) \
+
+INCLUDE_DIRS = \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
+
+
+.c.s:
+	$(CC) -S $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
+
+
+default: $(CELL_LIB)
+
+
+$(CELL_LIB): $(OBJECTS) $(SPU_CODE_MODULE)
+#	ar -ru $(CELL_LIB) $(OBJECTS) $(SPU_CODE_MODULE) # doesn't work
+	ar -ru $(CELL_LIB) $(OBJECTS)
+
+#$(PROG): $(PPU_OBJECTS)
+#	$(CC) -o $(PROG) $(PPU_OBJECTS) $(SPU_CODE_MODULE) $(PPU_LFLAGS)
+
+
+
+clean:
+	rm -f *.o *~ $(CELL_LIB)
+
+
+
+depend: $(SOURCES)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(SOURCES) 2> /dev/null
+
+include depend
+
+
+
diff --git a/src/gallium/drivers/cell/ppu/cell_batch.c b/src/gallium/drivers/cell/ppu/cell_batch.c
new file mode 100644
index 0000000000..fe144f8b84
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_batch.c
@@ -0,0 +1,260 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "cell_context.h"
+#include "cell_batch.h"
+#include "cell_fence.h"
+#include "cell_spu.h"
+
+
+
+/**
+ * Search the buffer pool for an empty/free buffer and return its index.
+ * Buffers are used for storing vertex data, state and commands which
+ * will be sent to the SPUs.
+ * If no empty buffers are available, wait for one.
+ * \return buffer index in [0, CELL_NUM_BUFFERS-1]
+ */
+uint
+cell_get_empty_buffer(struct cell_context *cell)
+{
+   static uint prev_buffer = 0;
+   uint buf = (prev_buffer + 1) % CELL_NUM_BUFFERS;
+   uint tries = 0;
+
+   /* Find a buffer that's marked as free by all SPUs */
+   while (1) {
+      uint spu, num_free = 0;
+
+      for (spu = 0; spu < cell->num_spus; spu++) {
+         if (cell->buffer_status[spu][buf][0] == CELL_BUFFER_STATUS_FREE) {
+            num_free++;
+
+            if (num_free == cell->num_spus) {
+               /* found a free buffer, now mark status as used */
+               for (spu = 0; spu < cell->num_spus; spu++) {
+                  cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_USED;
+               }
+               /*
+               printf("PPU: ALLOC BUFFER %u, %u tries\n", buf, tries);
+               */
+               prev_buffer = buf;
+
+               /* release tex buffer associated w/ prev use of this batch buf */
+               cell_free_fenced_buffers(cell, &cell->fenced_buffers[buf]);
+
+               return buf;
+            }
+         }
+         else {
+            break;
+         }
+      }
+
+      /* try next buf */
+      buf = (buf + 1) % CELL_NUM_BUFFERS;
+
+      tries++;
+      if (tries == 100) {
+         /*
+         printf("PPU WAITING for buffer...\n");
+         */
+      }
+   }
+}
+
+
+/**
+ * Append a fence command to the current batch buffer.
+ * Note that we're sure there's always room for this because of the
+ * adjusted size check in cell_batch_free_space().
+ */
+static void
+emit_fence(struct cell_context *cell)
+{
+   const uint batch = cell->cur_batch;
+   const uint size = cell->buffer_size[batch];
+   struct cell_command_fence *fence_cmd;
+   struct cell_fence *fence = &cell->fenced_buffers[batch].fence;
+   uint i;
+
+   /* set fence status to emitted, not yet signalled */
+   for (i = 0; i < cell->num_spus; i++) {
+      fence->status[i][0] = CELL_FENCE_EMITTED;
+   }
+
+   STATIC_ASSERT(sizeof(struct cell_command_fence) % 16 == 0);
+   ASSERT(size % 16 == 0);
+   ASSERT(size + sizeof(struct cell_command_fence) <= CELL_BUFFER_SIZE);
+
+   fence_cmd = (struct cell_command_fence *) (cell->buffer[batch] + size);
+   fence_cmd->opcode[0] = CELL_CMD_FENCE;
+   fence_cmd->fence = fence;
+
+   /* update batch buffer size */
+   cell->buffer_size[batch] = size + sizeof(struct cell_command_fence);
+}
+
+
+/**
+ * Flush the current batch buffer to the SPUs.
+ * An empty buffer will be found and set as the new current batch buffer
+ * for subsequent commands/data.
+ */
+void
+cell_batch_flush(struct cell_context *cell)
+{
+   static boolean flushing = FALSE;
+   uint batch = cell->cur_batch;
+   uint size = cell->buffer_size[batch];
+   uint spu, cmd_word;
+
+   assert(!flushing);
+
+   if (size == 0)
+      return;
+
+   /* Before we use this batch buffer, make sure any fenced texture buffers
+    * are released.
+    */
+   if (cell->fenced_buffers[batch].head) {
+      emit_fence(cell);
+      size = cell->buffer_size[batch];
+   }
+
+   flushing = TRUE;
+
+   assert(batch < CELL_NUM_BUFFERS);
+
+   /*
+   printf("cell_batch_dispatch: buf %u at %p, size %u\n",
+          batch, &cell->buffer[batch][0], size);
+   */
+     
+   /*
+    * Build "BATCH" command and send to all SPUs.
+    */
+   cmd_word = CELL_CMD_BATCH | (batch << 8) | (size << 16);
+
+   for (spu = 0; spu < cell->num_spus; spu++) {
+      assert(cell->buffer_status[spu][batch][0] == CELL_BUFFER_STATUS_USED);
+      send_mbox_message(cell_global.spe_contexts[spu], cmd_word);
+   }
+
+   /* When the SPUs are done copying the buffer into their locals stores
+    * they'll write a BUFFER_STATUS_FREE message into the buffer_status[]
+    * array indicating that the PPU can re-use the buffer.
+    */
+
+   batch = cell_get_empty_buffer(cell);
+
+   cell->buffer_size[batch] = 0;  /* empty */
+   cell->cur_batch = batch;
+
+   flushing = FALSE;
+}
+
+
+/**
+ * Return the number of bytes free in the current batch buffer.
+ */
+uint
+cell_batch_free_space(const struct cell_context *cell)
+{
+   uint free = CELL_BUFFER_SIZE - cell->buffer_size[cell->cur_batch];
+   free -= sizeof(struct cell_command_fence);
+   return free;
+}
+
+
+/**
+ * Allocate space in the current batch buffer for 'bytes' space.
+ * Bytes must be a multiple of 16 bytes.  Allocation will be 16 byte aligned.
+ * \return address in batch buffer to put data
+ */
+void *
+cell_batch_alloc16(struct cell_context *cell, uint bytes)
+{
+   void *pos;
+   uint size;
+
+   ASSERT(bytes % 16 == 0);
+   ASSERT(bytes <= CELL_BUFFER_SIZE);
+   ASSERT(cell->cur_batch >= 0);
+
+#ifdef ASSERT
+   {
+      uint spu;
+      for (spu = 0; spu < cell->num_spus; spu++) {
+         ASSERT(cell->buffer_status[spu][cell->cur_batch][0]
+                 == CELL_BUFFER_STATUS_USED);
+      }
+   }
+#endif
+
+   size = cell->buffer_size[cell->cur_batch];
+
+   if (bytes > cell_batch_free_space(cell)) {
+      cell_batch_flush(cell);
+      size = 0;
+   }
+
+   ASSERT(size % 16 == 0);
+   ASSERT(size + bytes <= CELL_BUFFER_SIZE);
+
+   pos = (void *) (cell->buffer[cell->cur_batch] + size);
+
+   cell->buffer_size[cell->cur_batch] = size + bytes;
+
+   return pos;
+}
+
+
+/**
+ * One-time init of batch buffers.
+ */
+void
+cell_init_batch_buffers(struct cell_context *cell)
+{
+   uint spu, buf;
+
+   /* init command, vertex/index buffer info */
+   for (buf = 0; buf < CELL_NUM_BUFFERS; buf++) {
+      cell->buffer_size[buf] = 0;
+
+      /* init batch buffer status values,
+       * mark 0th buffer as used, rest as free.
+       */
+      for (spu = 0; spu < cell->num_spus; spu++) {
+         if (buf == 0)
+            cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_USED;
+         else
+            cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_FREE;
+      }
+   }
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_batch.h b/src/gallium/drivers/cell/ppu/cell_batch.h
new file mode 100644
index 0000000000..290136031a
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_batch.h
@@ -0,0 +1,54 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef CELL_BATCH_H
+#define CELL_BATCH_H
+
+#include "pipe/p_compiler.h"
+
+
+struct cell_context;
+
+
+extern uint
+cell_get_empty_buffer(struct cell_context *cell);
+
+extern void
+cell_batch_flush(struct cell_context *cell);
+
+extern uint
+cell_batch_free_space(const struct cell_context *cell);
+
+extern void *
+cell_batch_alloc16(struct cell_context *cell, uint bytes);
+
+extern void
+cell_init_batch_buffers(struct cell_context *cell);
+
+
+#endif /* CELL_BATCH_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_clear.c b/src/gallium/drivers/cell/ppu/cell_clear.c
new file mode 100644
index 0000000000..c2e276988c
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_clear.c
@@ -0,0 +1,123 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Authors
+ *  Brian Paul
+ */
+
+#include <stdio.h>
+#include <assert.h>
+#include <stdint.h>
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+#include "util/u_pack_color.h"
+#include "cell/common.h"
+#include "cell_clear.h"
+#include "cell_context.h"
+#include "cell_batch.h"
+#include "cell_flush.h"
+#include "cell_spu.h"
+#include "cell_state.h"
+
+
+/**
+ * Convert packed pixel from one format to another.
+ */
+static unsigned
+convert_color(enum pipe_format srcFormat, unsigned srcColor,
+              enum pipe_format dstFormat)
+{
+   ubyte r, g, b, a;
+   unsigned dstColor;
+
+   util_unpack_color_ub(srcFormat, &srcColor, &r, &g, &b, &a);
+   util_pack_color_ub(r, g, b, a, dstFormat, &dstColor);
+
+   return dstColor;
+}
+
+
+
+/**
+ * Called via pipe->clear()
+ */
+void
+cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps,
+                   unsigned clearValue)
+{
+   struct pipe_screen *screen = pipe->screen;
+   struct cell_context *cell = cell_context(pipe);
+   uint surfIndex;
+
+   if (cell->dirty)
+      cell_update_derived(cell);
+
+
+   if (!cell->cbuf_map[0])
+      cell->cbuf_map[0] = screen->surface_map(screen, ps,
+                                              PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   if (ps == cell->framebuffer.zsbuf) {
+      /* clear z/stencil buffer */
+      surfIndex = 1;
+   }
+   else {
+      /* clear color buffer */
+      surfIndex = 0;
+
+      if (ps->format != PIPE_FORMAT_A8R8G8B8_UNORM) {
+         clearValue = convert_color(PIPE_FORMAT_A8R8G8B8_UNORM, clearValue,
+                                    ps->format);
+      }
+   }
+
+
+   /* Build a CLEAR command and place it in the current batch buffer */
+   {
+      STATIC_ASSERT(sizeof(struct cell_command_clear_surface) % 16 == 0);
+      struct cell_command_clear_surface *clr
+         = (struct cell_command_clear_surface *)
+         cell_batch_alloc16(cell, sizeof(*clr));
+      clr->opcode[0] = CELL_CMD_CLEAR_SURFACE;
+      clr->surface = surfIndex;
+      clr->value = clearValue;
+   }
+
+   /* Technically, the surface's contents are now known and cleared,
+    * so we could set the status to PIPE_SURFACE_STATUS_CLEAR.  But
+    * it turns out it's quite painful to recognize when any particular
+    * surface goes from PIPE_SURFACE_STATUS_CLEAR to 
+    * PIPE_SURFACE_STATUS_DEFINED (i.e. with known contents), because
+    * the drawing commands could be operating on numerous draw buffers,
+    * which we'd have to iterate through to set all their stati...
+    * For now, we cheat a bit and set the surface's status to DEFINED
+    * right here.  Later we should revisit this and set the status to
+    * CLEAR here, and find a better place to set the status to DEFINED.
+    */
+   ps->status = PIPE_SURFACE_STATUS_DEFINED;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_clear.h b/src/gallium/drivers/cell/ppu/cell_clear.h
new file mode 100644
index 0000000000..ff47d43f4c
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_clear.h
@@ -0,0 +1,43 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef CELL_CLEAR_H
+#define CELL_CLEAR_H
+
+
+struct pipe_context;
+struct pipe_surface;
+
+
+extern void
+cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps,
+                   unsigned clearValue);
+
+
+
+#endif /* CELL_CLEAR_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c
new file mode 100644
index 0000000000..ae82ded334
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_context.c
@@ -0,0 +1,183 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Authors
+ *  Brian Paul
+ */
+
+
+#include <stdio.h>
+
+#include "pipe/p_defines.h"
+#include "pipe/p_format.h"
+#include "util/u_memory.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_screen.h"
+
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+
+#include "cell/common.h"
+#include "cell_batch.h"
+#include "cell_clear.h"
+#include "cell_context.h"
+#include "cell_draw_arrays.h"
+#include "cell_fence.h"
+#include "cell_flush.h"
+#include "cell_state.h"
+#include "cell_surface.h"
+#include "cell_spu.h"
+#include "cell_pipe_state.h"
+#include "cell_texture.h"
+#include "cell_vbuf.h"
+
+
+
+static void
+cell_destroy_context( struct pipe_context *pipe )
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   util_delete_keymap(cell->fragment_ops_cache, NULL);
+
+   cell_spu_exit(cell);
+
+   align_free(cell);
+}
+
+
+static struct draw_context *
+cell_draw_create(struct cell_context *cell)
+{
+   struct draw_context *draw = draw_create();
+
+#if 0 /* broken */
+   if (getenv("GALLIUM_CELL_VS")) {
+      /* plug in SPU-based vertex transformation code */
+      draw->shader_queue_flush = cell_vertex_shader_queue_flush;
+      draw->driver_private = cell;
+   }
+#endif
+
+   return draw;
+}
+
+
+static const struct debug_named_value cell_debug_flags[] = {
+   {"checker", CELL_DEBUG_CHECKER},/**< modulate tile clear color by SPU ID */
+   {"asm", CELL_DEBUG_ASM},        /**< dump SPU asm code */
+   {"sync", CELL_DEBUG_SYNC},      /**< SPUs do synchronous DMA */
+   {"fragops", CELL_DEBUG_FRAGMENT_OPS}, /**< SPUs emit fragment ops debug messages*/
+   {"fragopfallback", CELL_DEBUG_FRAGMENT_OP_FALLBACK}, /**< SPUs use reference implementation for fragment ops*/
+   {"cmd", CELL_DEBUG_CMD},       /**< SPUs dump command buffer info */
+   {"cache", CELL_DEBUG_CACHE},   /**< report texture cache stats on exit */
+   {NULL, 0}
+};
+
+
+struct pipe_context *
+cell_create_context(struct pipe_screen *screen,
+                    struct cell_winsys *cws)
+{
+   struct cell_context *cell;
+   uint i;
+
+   /* some fields need to be 16-byte aligned, so align the whole object */
+   cell = (struct cell_context*) align_malloc(sizeof(struct cell_context), 16);
+   if (!cell)
+      return NULL;
+
+   memset(cell, 0, sizeof(*cell));
+
+   cell->winsys = cws;
+   cell->pipe.winsys = screen->winsys;
+   cell->pipe.screen = screen;
+   cell->pipe.destroy = cell_destroy_context;
+
+   cell->pipe.clear = cell_clear_surface;
+   cell->pipe.flush = cell_flush;
+
+#if 0
+   cell->pipe.begin_query = cell_begin_query;
+   cell->pipe.end_query = cell_end_query;
+   cell->pipe.wait_query = cell_wait_query;
+#endif
+
+   cell_init_draw_functions(cell);
+   cell_init_state_functions(cell);
+   cell_init_shader_functions(cell);
+   cell_init_surface_functions(cell);
+   cell_init_vertex_functions(cell);
+
+   cell->draw = cell_draw_create(cell);
+
+   /* Create cache of fragment ops generated code */
+   cell->fragment_ops_cache =
+      util_new_keymap(sizeof(struct cell_fragment_ops_key), ~0, NULL);
+
+   cell_init_vbuf(cell);
+
+   draw_set_rasterize_stage(cell->draw, cell->vbuf);
+
+   /* convert all points/lines to tris for the time being */
+   draw_wide_point_threshold(cell->draw, 0.0);
+   draw_wide_line_threshold(cell->draw, 0.0);
+
+   /* get env vars or read config file to get debug flags */
+   cell->debug_flags = debug_get_flags_option("CELL_DEBUG", 
+                                              cell_debug_flags, 
+                                              0 );
+
+   for (i = 0; i < CELL_NUM_BUFFERS; i++)
+      cell_fence_init(&cell->fenced_buffers[i].fence);
+
+
+   /*
+    * SPU stuff
+    */
+   /* This call only works with SDK 3.0.  Anyone still using 2.1??? */
+   cell->num_cells = spe_cpu_info_get(SPE_COUNT_PHYSICAL_CPU_NODES, -1);
+   cell->num_spus = spe_cpu_info_get(SPE_COUNT_USABLE_SPES, -1);
+   if (cell->debug_flags) {
+      printf("Cell: found %d Cell(s) with %u SPUs\n",
+             cell->num_cells, cell->num_spus);
+   }
+   if (getenv("CELL_NUM_SPUS")) {
+      cell->num_spus = atoi(getenv("CELL_NUM_SPUS"));
+      assert(cell->num_spus > 0);
+   }
+
+   cell_start_spus(cell);
+
+   cell_init_batch_buffers(cell);
+
+   /* make sure SPU initializations are done before proceeding */
+   cell_flush_int(cell, CELL_FLUSH_WAIT);
+
+   return &cell->pipe;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h
new file mode 100644
index 0000000000..eb1397bb3f
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_context.h
@@ -0,0 +1,205 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef CELL_CONTEXT_H
+#define CELL_CONTEXT_H
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_vbuf.h"
+#include "cell_winsys.h"
+#include "cell/common.h"
+#include "rtasm/rtasm_ppc_spe.h"
+#include "tgsi/tgsi_scan.h"
+#include "util/u_keymap.h"
+
+
+struct cell_vbuf_render;
+
+
+/**
+ * Cell vertex shader state, subclass of pipe_shader_state.
+ */
+struct cell_vertex_shader_state
+{
+   struct pipe_shader_state shader;
+   struct tgsi_shader_info info;
+   void *draw_data;
+};
+
+
+/**
+ * Cell fragment shader state, subclass of pipe_shader_state.
+ */
+struct cell_fragment_shader_state
+{
+   struct pipe_shader_state shader;
+   struct tgsi_shader_info info;
+   struct spe_function code;
+   void *data;
+};
+
+
+/**
+ * Key for mapping per-fragment state to cached SPU machine code.
+ *  keymap(cell_fragment_ops_key) => cell_command_fragment_ops
+ */
+struct cell_fragment_ops_key
+{
+   struct pipe_blend_state blend;
+   struct pipe_blend_color blend_color;
+   struct pipe_depth_stencil_alpha_state dsa;
+   enum pipe_format color_format;
+   enum pipe_format zs_format;
+};
+
+
+struct cell_buffer_node;
+
+/**
+ * Fenced buffer list.  List of buffers which can be unreferenced after
+ * the fence has been executed/signalled.
+ */
+struct cell_buffer_list
+{
+   struct cell_fence fence ALIGN16_ATTRIB;
+   struct cell_buffer_node *head;
+};
+
+
+/**
+ * Per-context state, subclass of pipe_context.
+ */
+struct cell_context
+{
+   struct pipe_context pipe;
+
+   struct cell_winsys *winsys;
+
+   const struct pipe_blend_state *blend;
+   const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS];
+   uint num_samplers;
+   const struct pipe_depth_stencil_alpha_state *depth_stencil;
+   const struct pipe_rasterizer_state *rasterizer;
+   const struct cell_vertex_shader_state *vs;
+   const struct cell_fragment_shader_state *fs;
+
+   struct spe_function logic_op;
+
+   struct pipe_blend_color blend_color;
+   struct pipe_clip_state clip;
+   struct pipe_constant_buffer constants[2];
+   struct pipe_framebuffer_state framebuffer;
+   struct pipe_poly_stipple poly_stipple;
+   struct pipe_scissor_state scissor;
+   struct cell_texture *texture[PIPE_MAX_SAMPLERS];
+   uint num_textures;
+   struct pipe_viewport_state viewport;
+   struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+   uint num_vertex_buffers;
+   struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
+   uint num_vertex_elements;
+
+   ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS];
+   ubyte *zsbuf_map;
+
+   struct pipe_surface *tex_surf;
+   uint *tex_map;
+
+   uint dirty;
+   uint dirty_textures;  /* bitmask of texture units */
+   uint dirty_samplers;  /* bitmask of sampler units */
+
+   /** Cache of code generated for per-fragment ops */
+   struct keymap *fragment_ops_cache;
+
+   /** The primitive drawing context */
+   struct draw_context *draw;
+   struct draw_stage *render_stage;
+
+   /** For post-transformed vertex buffering: */
+   struct cell_vbuf_render *vbuf_render;
+   struct draw_stage *vbuf;
+
+   struct vertex_info vertex_info;
+
+   /** Mapped constant buffers */
+   void *mapped_constants[PIPE_SHADER_TYPES];
+
+   struct cell_spu_function_info spu_functions ALIGN16_ATTRIB;
+
+   uint num_cells, num_spus;
+
+   /** Buffers for command batches, vertex/index data */
+   uint buffer_size[CELL_NUM_BUFFERS];
+   ubyte buffer[CELL_NUM_BUFFERS][CELL_BUFFER_SIZE] ALIGN16_ATTRIB;
+
+   int cur_batch;  /**< which buffer is being filled w/ commands */
+
+   /** [4] to ensure 16-byte alignment for each status word */
+   uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BUFFERS][4] ALIGN16_ATTRIB;
+
+
+   /** Associated with each command/batch buffer is a list of pipe_buffers
+    * that are fenced.  When the last command in a buffer is executed, the
+    * fence will be signalled, indicating that any pipe_buffers preceeding
+    * that fence can be unreferenced (and probably freed).
+    */
+   struct cell_buffer_list fenced_buffers[CELL_NUM_BUFFERS];
+
+
+   struct spe_function attrib_fetch;
+   unsigned attrib_fetch_offsets[PIPE_MAX_ATTRIBS];
+
+   unsigned debug_flags;
+};
+
+
+
+
+static INLINE struct cell_context *
+cell_context(struct pipe_context *pipe)
+{
+   return (struct cell_context *) pipe;
+}
+
+
+extern struct pipe_context *
+cell_create_context(struct pipe_screen *screen, struct cell_winsys *cws);
+
+extern void
+cell_vertex_shader_queue_flush(struct draw_context *draw);
+
+
+/* XXX find a better home for this */
+extern void cell_update_vertex_fetch(struct draw_context *draw);
+
+
+#endif /* CELL_CONTEXT_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
new file mode 100644
index 0000000000..644496db40
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
@@ -0,0 +1,191 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Brian Paul
+ *    Keith Whitwell
+ */
+
+
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+
+#include "cell_context.h"
+#include "cell_draw_arrays.h"
+#include "cell_state.h"
+#include "cell_flush.h"
+
+#include "draw/draw_context.h"
+
+
+
+static void
+cell_map_constant_buffers(struct cell_context *sp)
+{
+   struct pipe_winsys *ws = sp->pipe.winsys;
+   uint i;
+   for (i = 0; i < 2; i++) {
+      if (sp->constants[i].buffer && sp->constants[i].buffer->size) {
+         sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer,
+                                                   PIPE_BUFFER_USAGE_CPU_READ);
+         cell_flush_buffer_range(sp, sp->mapped_constants[i], 
+                                 sp->constants[i].buffer->size);
+      }
+   }
+
+   draw_set_mapped_constant_buffer(sp->draw,
+                                   sp->mapped_constants[PIPE_SHADER_VERTEX],
+                                   sp->constants[PIPE_SHADER_VERTEX].buffer->size);
+}
+
+static void
+cell_unmap_constant_buffers(struct cell_context *sp)
+{
+   struct pipe_winsys *ws = sp->pipe.winsys;
+   uint i;
+   for (i = 0; i < 2; i++) {
+      if (sp->constants[i].buffer && sp->constants[i].buffer->size)
+         ws->buffer_unmap(ws, sp->constants[i].buffer);
+      sp->mapped_constants[i] = NULL;
+   }
+}
+
+
+
+/**
+ * Draw vertex arrays, with optional indexing.
+ * Basically, map the vertex buffers (and drawing surfaces), then hand off
+ * the drawing to the 'draw' module.
+ *
+ * XXX should the element buffer be specified/bound with a separate function?
+ */
+static boolean
+cell_draw_range_elements(struct pipe_context *pipe,
+                         struct pipe_buffer *indexBuffer,
+                         unsigned indexSize,
+                         unsigned min_index,
+                         unsigned max_index,
+                         unsigned mode, unsigned start, unsigned count)
+{
+   struct cell_context *sp = cell_context(pipe);
+   struct draw_context *draw = sp->draw;
+   unsigned i;
+
+   if (sp->dirty)
+      cell_update_derived( sp );
+
+#if 0
+   cell_map_surfaces(sp);
+#endif
+   cell_map_constant_buffers(sp);
+
+   /*
+    * Map vertex buffers
+    */
+   for (i = 0; i < sp->num_vertex_buffers; i++) {
+      void *buf = pipe_buffer_map(pipe->screen,
+                                           sp->vertex_buffer[i].buffer,
+                                           PIPE_BUFFER_USAGE_CPU_READ);
+      cell_flush_buffer_range(sp, buf, sp->vertex_buffer[i].buffer->size);
+      draw_set_mapped_vertex_buffer(draw, i, buf);
+   }
+   /* Map index buffer, if present */
+   if (indexBuffer) {
+      void *mapped_indexes = pipe_buffer_map(pipe->screen,
+                                                      indexBuffer,
+                                                      PIPE_BUFFER_USAGE_CPU_READ);
+      draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
+   }
+   else {
+      /* no index/element buffer */
+      draw_set_mapped_element_buffer(draw, 0, NULL);
+   }
+
+
+   /* draw! */
+   draw_arrays(draw, mode, start, count);
+
+   /*
+    * unmap vertex/index buffers - will cause draw module to flush
+    */
+   for (i = 0; i < sp->num_vertex_buffers; i++) {
+      draw_set_mapped_vertex_buffer(draw, i, NULL);
+      pipe_buffer_unmap(pipe->screen, sp->vertex_buffer[i].buffer);
+   }
+   if (indexBuffer) {
+      draw_set_mapped_element_buffer(draw, 0, NULL);
+      pipe_buffer_unmap(pipe->screen, indexBuffer);
+   }
+
+   /* Note: leave drawing surfaces mapped */
+   cell_unmap_constant_buffers(sp);
+
+   return TRUE;
+}
+
+
+static boolean
+cell_draw_elements(struct pipe_context *pipe,
+                   struct pipe_buffer *indexBuffer,
+                   unsigned indexSize,
+                   unsigned mode, unsigned start, unsigned count)
+{
+   return cell_draw_range_elements( pipe, indexBuffer,
+                                    indexSize,
+                                    0, 0xffffffff,
+                                    mode, start, count );
+}
+
+
+static boolean
+cell_draw_arrays(struct pipe_context *pipe, unsigned mode,
+                     unsigned start, unsigned count)
+{
+   return cell_draw_elements(pipe, NULL, 0, mode, start, count);
+}
+
+
+static void
+cell_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags)
+{
+   struct cell_context *cell = cell_context(pipe);
+   draw_set_edgeflags(cell->draw, edgeflags);
+}
+
+
+
+void
+cell_init_draw_functions(struct cell_context *cell)
+{
+   cell->pipe.draw_arrays = cell_draw_arrays;
+   cell->pipe.draw_elements = cell_draw_elements;
+   cell->pipe.draw_range_elements = cell_draw_range_elements;
+   cell->pipe.set_edgeflags = cell_set_edgeflags;
+}
+
diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.h b/src/gallium/drivers/cell/ppu/cell_draw_arrays.h
new file mode 100644
index 0000000000..148873aa67
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.h
@@ -0,0 +1,36 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef CELL_DRAW_ARRAYS_H
+#define CELL_DRAW_ARRAYS_H
+
+
+extern void
+cell_init_draw_functions(struct cell_context *cell);
+
+
+#endif /* CELL_DRAW_ARRAYS_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_fence.c b/src/gallium/drivers/cell/ppu/cell_fence.c
new file mode 100644
index 0000000000..867b5dcaa0
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_fence.c
@@ -0,0 +1,168 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include <unistd.h>
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "cell_context.h"
+#include "cell_batch.h"
+#include "cell_fence.h"
+#include "cell_texture.h"
+
+
+void
+cell_fence_init(struct cell_fence *fence)
+{
+   uint i;
+   ASSERT_ALIGN16(fence->status);
+   for (i = 0; i < CELL_MAX_SPUS; i++) {
+      fence->status[i][0] = CELL_FENCE_IDLE;
+   }
+}
+
+
+boolean
+cell_fence_signalled(const struct cell_context *cell,
+                     const struct cell_fence *fence)
+{
+   uint i;
+   for (i = 0; i < cell->num_spus; i++) {
+      if (fence->status[i][0] != CELL_FENCE_SIGNALLED)
+         return FALSE;
+      /*assert(fence->status[i][0] == CELL_FENCE_EMITTED);*/
+   }
+   return TRUE;
+}
+
+
+void
+cell_fence_finish(const struct cell_context *cell,
+                  const struct cell_fence *fence)
+{
+   while (!cell_fence_signalled(cell, fence)) {
+      usleep(10);
+   }
+
+#ifdef DEBUG
+   {
+      uint i;
+      for (i = 0; i < cell->num_spus; i++) {
+         assert(fence->status[i][0] == CELL_FENCE_SIGNALLED);
+      }
+   }
+#endif
+}
+
+
+
+
+struct cell_buffer_node
+{
+   struct pipe_buffer *buffer;
+   struct cell_buffer_node *next;
+};
+
+
+static void
+cell_add_buffer_to_list(struct cell_context *cell,
+                        struct cell_buffer_list *list,
+                        struct pipe_buffer *buffer)
+{
+   struct pipe_screen *ps = cell->pipe.screen;
+   struct cell_buffer_node *node = CALLOC_STRUCT(cell_buffer_node);
+   /* create new list node which references the buffer, insert at head */
+   if (node) {
+      pipe_buffer_reference(ps, &node->buffer, buffer);
+      node->next = list->head;
+      list->head = node;
+   }
+}
+
+
+/**
+ * Wait for completion of the given fence, then unreference any buffers
+ * on the list.
+ * This typically unrefs/frees texture buffers after any rendering which uses
+ * them has completed.
+ */
+void
+cell_free_fenced_buffers(struct cell_context *cell,
+                         struct cell_buffer_list *list)
+{
+   if (list->head) {
+      struct pipe_screen *ps = cell->pipe.screen;
+      struct cell_buffer_node *node;
+
+      cell_fence_finish(cell, &list->fence);
+
+      /* traverse the list, unreferencing buffers, freeing nodes */
+      node = list->head;
+      while (node) {
+         struct cell_buffer_node *next = node->next;
+         assert(node->buffer);
+         pipe_buffer_unmap(ps, node->buffer);
+#if 0
+         printf("Unref buffer %p\n", node->buffer);
+         if (node->buffer->refcount == 1)
+            printf("   Delete!\n");
+#endif
+         pipe_buffer_reference(ps, &node->buffer, NULL);
+         FREE(node);
+         node = next;
+      }
+      list->head = NULL;
+   }
+}
+
+
+/**
+ * This should be called for each render command.
+ * Any texture buffers that are current bound will be added to a fenced
+ * list to be freed later when the fence is executed/signalled.
+ */
+void
+cell_add_fenced_textures(struct cell_context *cell)
+{
+   struct cell_buffer_list *list = &cell->fenced_buffers[cell->cur_batch];
+   uint i;
+
+   for (i = 0; i < cell->num_textures; i++) {
+      struct cell_texture *ct = cell->texture[i];
+      if (ct) {
+         uint level;
+         for (level = 0; level < CELL_MAX_TEXTURE_LEVELS; level++) {
+            if (ct->tiled_buffer[level]) {
+#if 0
+               printf("Adding texture %p buffer %p to list\n",
+                      ct, ct->tiled_buffer[level]);
+#endif
+               cell_add_buffer_to_list(cell, list, ct->tiled_buffer[level]);
+            }
+         }
+      }
+   }
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_fence.h b/src/gallium/drivers/cell/ppu/cell_fence.h
new file mode 100644
index 0000000000..536b4ba411
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_fence.h
@@ -0,0 +1,57 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef CELL_FENCE_H
+#define CELL_FENCE_H
+
+
+extern void
+cell_fence_init(struct cell_fence *fence);
+
+
+extern boolean
+cell_fence_signalled(const struct cell_context *cell,
+                     const struct cell_fence *fence);
+
+
+extern void
+cell_fence_finish(const struct cell_context *cell,
+                  const struct cell_fence *fence);
+
+
+
+extern void
+cell_free_fenced_buffers(struct cell_context *cell,
+                         struct cell_buffer_list *list);
+
+
+extern void
+cell_add_fenced_textures(struct cell_context *cell);
+
+
+#endif /* CELL_FENCE_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_flush.c b/src/gallium/drivers/cell/ppu/cell_flush.c
new file mode 100644
index 0000000000..8275c9dc9c
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_flush.c
@@ -0,0 +1,112 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "cell_context.h"
+#include "cell_batch.h"
+#include "cell_flush.h"
+#include "cell_spu.h"
+#include "cell_render.h"
+#include "draw/draw_context.h"
+
+
+/**
+ * Called via pipe->flush()
+ */
+void
+cell_flush(struct pipe_context *pipe, unsigned flags,
+           struct pipe_fence_handle **fence)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   if (fence) {
+      *fence = NULL;
+      /* XXX: Implement real fencing */
+      flags |= CELL_FLUSH_WAIT;
+   }
+
+   if (flags & (PIPE_FLUSH_SWAPBUFFERS | PIPE_FLUSH_RENDER_CACHE))
+      flags |= CELL_FLUSH_WAIT;
+
+   draw_flush( cell->draw );
+   cell_flush_int(cell, flags);
+}
+
+
+/**
+ * Cell internal flush function.  Send the current batch buffer to all SPUs.
+ * If flags & CELL_FLUSH_WAIT, do not return until the SPUs are idle.
+ * \param flags  bitmask of flags CELL_FLUSH_WAIT, or zero
+ */
+void
+cell_flush_int(struct cell_context *cell, unsigned flags)
+{
+   static boolean flushing = FALSE;  /* recursion catcher */
+   uint i;
+
+   ASSERT(!flushing);
+   flushing = TRUE;
+
+   if (flags & CELL_FLUSH_WAIT) {
+      STATIC_ASSERT(sizeof(opcode_t) % 16 == 0);
+      opcode_t *cmd = (opcode_t*) cell_batch_alloc16(cell, sizeof(opcode_t));
+      *cmd[0] = CELL_CMD_FINISH;
+   }
+
+   cell_batch_flush(cell);
+
+#if 0
+   /* Send CMD_FINISH to all SPUs */
+   for (i = 0; i < cell->num_spus; i++) {
+      send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FINISH);
+   }
+#endif
+
+   if (flags & CELL_FLUSH_WAIT) {
+      /* Wait for ack */
+      for (i = 0; i < cell->num_spus; i++) {
+         uint k = wait_mbox_message(cell_global.spe_contexts[i]);
+         assert(k == CELL_CMD_FINISH);
+      }
+   }
+
+   flushing = FALSE;
+}
+
+
+void
+cell_flush_buffer_range(struct cell_context *cell, void *ptr,
+			unsigned size)
+{
+   STATIC_ASSERT((sizeof(opcode_t) + sizeof(struct cell_buffer_range)) % 16 == 0);
+   uint32_t *batch = (uint32_t*)cell_batch_alloc16(cell, 
+      sizeof(opcode_t) + sizeof(struct cell_buffer_range));
+   struct cell_buffer_range *br = (struct cell_buffer_range *) &batch[4];
+   batch[0] = CELL_CMD_FLUSH_BUFFER_RANGE;
+   br->base = (uintptr_t) ptr;
+   br->size = size;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_flush.h b/src/gallium/drivers/cell/ppu/cell_flush.h
new file mode 100644
index 0000000000..509ae6239a
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_flush.h
@@ -0,0 +1,45 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef CELL_FLUSH
+#define CELL_FLUSH
+
+#define CELL_FLUSH_WAIT 0x80000000
+
+extern void
+cell_flush(struct pipe_context *pipe, unsigned flags,
+           struct pipe_fence_handle **fence);
+
+extern void
+cell_flush_int(struct cell_context *cell, unsigned flags);
+
+extern void
+cell_flush_buffer_range(struct cell_context *cell, void *ptr,
+			unsigned size);
+
+#endif
diff --git a/src/gallium/drivers/cell/ppu/cell_gen_fp.c b/src/gallium/drivers/cell/ppu/cell_gen_fp.c
new file mode 100644
index 0000000000..5a889a6119
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_gen_fp.c
@@ -0,0 +1,2046 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2009 VMware, Inc.  All rights reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+
+/**
+ * Generate SPU fragment program/shader code.
+ *
+ * Note that we generate SOA-style code here.  So each TGSI instruction
+ * operates on four pixels (and is translated into four SPU instructions,
+ * generally speaking).
+ *
+ * \author Brian Paul
+ */
+
+#include <math.h>
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi/tgsi_exec.h"
+#include "tgsi/tgsi_dump.h"
+#include "rtasm/rtasm_ppc_spe.h"
+#include "util/u_memory.h"
+#include "cell_context.h"
+#include "cell_gen_fp.h"
+
+
+#define MAX_TEMPS 16
+#define MAX_IMMED  8
+
+#define CHAN_X  0
+#define CHAN_Y  1
+#define CHAN_Z  2
+#define CHAN_W  3
+
+/**
+ * Context needed during code generation.
+ */
+struct codegen
+{
+   struct cell_context *cell;
+   int inputs_reg;      /**< 1st function parameter */
+   int outputs_reg;     /**< 2nd function parameter */
+   int constants_reg;   /**< 3rd function parameter */
+   int temp_regs[MAX_TEMPS][4]; /**< maps TGSI temps to SPE registers */
+   int imm_regs[MAX_IMMED][4];  /**< maps TGSI immediates to SPE registers */
+
+   int num_imm;  /**< number of immediates */
+
+   int one_reg;         /**< register containing {1.0, 1.0, 1.0, 1.0} */
+
+   int addr_reg;        /**< address register, integer values */
+
+   /** Per-instruction temps / intermediate temps */
+   int num_itemps;
+   int itemps[12];
+
+   /** Current IF/ELSE/ENDIF nesting level */
+   int if_nesting;
+   /** Current BGNLOOP/ENDLOOP nesting level */
+   int loop_nesting;
+   /** Location of start of current loop */
+   int loop_start;
+
+   /** Index of if/conditional mask register */
+   int cond_mask_reg;
+   /** Index of loop mask register */
+   int loop_mask_reg;
+
+   /** Index of master execution mask register */
+   int exec_mask_reg;
+
+   /** KIL mask: indicates which fragments have been killed */
+   int kill_mask_reg;
+
+   int frame_size;  /**< Stack frame size, in words */
+
+   struct spe_function *f;
+   boolean error;
+};
+
+
+/**
+ * Allocate an intermediate temporary register.
+ */
+static int
+get_itemp(struct codegen *gen)
+{
+   int t = spe_allocate_available_register(gen->f);
+   assert(gen->num_itemps < Elements(gen->itemps));
+   gen->itemps[gen->num_itemps++] = t;
+   return t;
+}
+
+/**
+ * Free all intermediate temporary registers.  To be called after each
+ * instruction has been emitted.
+ */
+static void
+free_itemps(struct codegen *gen)
+{
+   int i;
+   for (i = 0; i < gen->num_itemps; i++) {
+      spe_release_register(gen->f, gen->itemps[i]);
+   }
+   gen->num_itemps = 0;
+}
+
+
+/**
+ * Return index of an SPE register containing {1.0, 1.0, 1.0, 1.0}.
+ * The register is allocated and initialized upon the first call.
+ */
+static int
+get_const_one_reg(struct codegen *gen)
+{
+   if (gen->one_reg <= 0) {
+      gen->one_reg = spe_allocate_available_register(gen->f);
+
+      spe_indent(gen->f, 4);
+      spe_comment(gen->f, -4, "init constant reg = 1.0:");
+
+      /* one = {1.0, 1.0, 1.0, 1.0} */
+      spe_load_float(gen->f, gen->one_reg, 1.0f);
+
+      spe_indent(gen->f, -4);
+   }
+
+   return gen->one_reg;
+}
+
+
+/**
+ * Return index of the address register.
+ * Used for indirect register loads/stores.
+ */
+static int
+get_address_reg(struct codegen *gen)
+{
+   if (gen->addr_reg <= 0) {
+      gen->addr_reg = spe_allocate_available_register(gen->f);
+
+      spe_indent(gen->f, 4);
+      spe_comment(gen->f, -4, "init address reg = 0:");
+
+      /* init addr = {0, 0, 0, 0} */
+      spe_zero(gen->f, gen->addr_reg);
+
+      spe_indent(gen->f, -4);
+   }
+
+   return gen->addr_reg;
+}
+
+
+/**
+ * Return index of the master execution mask.
+ * The register is allocated an initialized upon the first call.
+ *
+ * The master execution mask controls which pixels in a quad are
+ * modified, according to surrounding conditionals, loops, etc.
+ */
+static int
+get_exec_mask_reg(struct codegen *gen)
+{
+   if (gen->exec_mask_reg <= 0) {
+      gen->exec_mask_reg = spe_allocate_available_register(gen->f);
+
+      /* XXX this may not be needed */
+      spe_comment(gen->f, 0*-4, "initialize master execution mask = ~0");
+      spe_load_int(gen->f, gen->exec_mask_reg, ~0);
+   }
+
+   return gen->exec_mask_reg;
+}
+
+
+/** Return index of the conditional (if/else) execution mask register */
+static int
+get_cond_mask_reg(struct codegen *gen)
+{
+   if (gen->cond_mask_reg <= 0) {
+      gen->cond_mask_reg = spe_allocate_available_register(gen->f);
+   }
+
+   return gen->cond_mask_reg;
+}
+
+
+/** Return index of the loop execution mask register */
+static int
+get_loop_mask_reg(struct codegen *gen)
+{
+   if (gen->loop_mask_reg <= 0) {
+      gen->loop_mask_reg = spe_allocate_available_register(gen->f);
+   }
+
+   return gen->loop_mask_reg;
+}
+
+
+
+static boolean
+is_register_src(struct codegen *gen, int channel,
+                const struct tgsi_full_src_register *src)
+{
+   int swizzle = tgsi_util_get_full_src_register_extswizzle(src, channel);
+   int sign_op = tgsi_util_get_full_src_register_sign_mode(src, channel);
+
+   if (swizzle > TGSI_SWIZZLE_W || sign_op != TGSI_UTIL_SIGN_KEEP) {
+      return FALSE;
+   }
+   if (src->SrcRegister.File == TGSI_FILE_TEMPORARY ||
+       src->SrcRegister.File == TGSI_FILE_IMMEDIATE) {
+      return TRUE;
+   }
+   return FALSE;
+}
+
+  
+static boolean
+is_memory_dst(struct codegen *gen, int channel,
+              const struct tgsi_full_dst_register *dst)
+{
+   if (dst->DstRegister.File == TGSI_FILE_OUTPUT) {
+      return TRUE;
+   }
+   else {
+      return FALSE;
+   }
+}
+
+  
+/**
+ * Return the index of the SPU temporary containing the named TGSI
+ * source register.  If the TGSI register is a TGSI_FILE_TEMPORARY we
+ * just return the corresponding SPE register.  If the TGIS register
+ * is TGSI_FILE_INPUT/CONSTANT/IMMEDIATE we allocate a new SPE register
+ * and emit an SPE load instruction.
+ */
+static int
+get_src_reg(struct codegen *gen,
+            int channel,
+            const struct tgsi_full_src_register *src)
+{
+   int reg = -1;
+   int swizzle = tgsi_util_get_full_src_register_extswizzle(src, channel);
+   boolean reg_is_itemp = FALSE;
+   uint sign_op;
+
+   assert(swizzle >= TGSI_SWIZZLE_X);
+   assert(swizzle <= TGSI_EXTSWIZZLE_ONE);
+
+   if (swizzle == TGSI_EXTSWIZZLE_ONE) {
+      /* Load const one float and early out */
+      reg = get_const_one_reg(gen);
+   }
+   else if (swizzle == TGSI_EXTSWIZZLE_ZERO) {
+      /* Load const zero float and early out */
+      reg = get_itemp(gen);
+      spe_xor(gen->f, reg, reg, reg);
+   }
+   else {
+      int index = src->SrcRegister.Index;
+
+      assert(swizzle < 4);
+
+      if (src->SrcRegister.Indirect) {
+         /* XXX unfinished */
+      }
+
+      switch (src->SrcRegister.File) {
+      case TGSI_FILE_TEMPORARY:
+         reg = gen->temp_regs[index][swizzle];
+         break;
+      case TGSI_FILE_INPUT:
+         {
+            /* offset is measured in quadwords, not bytes */
+            int offset = index * 4 + swizzle;
+            reg = get_itemp(gen);
+            reg_is_itemp = TRUE;
+            /* Load:  reg = memory[(machine_reg) + offset] */
+            spe_lqd(gen->f, reg, gen->inputs_reg, offset * 16);
+         }
+         break;
+      case TGSI_FILE_IMMEDIATE:
+         reg = gen->imm_regs[index][swizzle];
+         break;
+      case TGSI_FILE_CONSTANT:
+         {
+            /* offset is measured in quadwords, not bytes */
+            int offset = index * 4 + swizzle;
+            reg = get_itemp(gen);
+            reg_is_itemp = TRUE;
+            /* Load:  reg = memory[(machine_reg) + offset] */
+            spe_lqd(gen->f, reg, gen->constants_reg, offset * 16);
+         }
+         break;
+      default:
+         assert(0);
+      }
+   }
+
+   /*
+    * Handle absolute value, negate or set-negative of src register.
+    */
+   sign_op = tgsi_util_get_full_src_register_sign_mode(src, channel);
+   if (sign_op != TGSI_UTIL_SIGN_KEEP) {
+      /*
+       * All sign ops are done by manipulating bit 31, the IEEE float sign bit.
+       */
+      const int bit31mask_reg = get_itemp(gen);
+      int result_reg;
+
+      if (reg_is_itemp) {
+         /* re-use 'reg' for the result */
+         result_reg = reg;
+      }
+      else {
+         /* alloc a new reg for the result */
+         result_reg = get_itemp(gen);
+      }
+
+      /* mask with bit 31 set, the rest cleared */
+      spe_load_uint(gen->f, bit31mask_reg, (1 << 31));
+
+      if (sign_op == TGSI_UTIL_SIGN_CLEAR) {
+         spe_andc(gen->f, result_reg, reg, bit31mask_reg);
+      }
+      else if (sign_op == TGSI_UTIL_SIGN_SET) {
+         spe_and(gen->f, result_reg, reg, bit31mask_reg);
+      }
+      else {
+         assert(sign_op == TGSI_UTIL_SIGN_TOGGLE);
+         spe_xor(gen->f, result_reg, reg, bit31mask_reg);
+      }
+
+      reg = result_reg;
+   }
+
+   return reg;
+}
+
+
+/**
+ * Return the index of an SPE register to use for the given TGSI register.
+ * If the TGSI register is TGSI_FILE_TEMPORARAY, the index of the
+ * corresponding SPE register is returned.  If the TGSI register is
+ * TGSI_FILE_OUTPUT we allocate an intermediate temporary register.
+ * See store_dest_reg() below...
+ */
+static int
+get_dst_reg(struct codegen *gen,
+            int channel,
+            const struct tgsi_full_dst_register *dest)
+{
+   int reg = -1;
+
+   switch (dest->DstRegister.File) {
+   case TGSI_FILE_TEMPORARY:
+      if (gen->if_nesting > 0 || gen->loop_nesting > 0)
+         reg = get_itemp(gen);
+      else
+         reg = gen->temp_regs[dest->DstRegister.Index][channel];
+      break;
+   case TGSI_FILE_OUTPUT:
+      reg = get_itemp(gen);
+      break;
+   default:
+      assert(0);
+   }
+
+   return reg;
+}
+
+
+/**
+ * When a TGSI instruction is writing to an output register, this
+ * function emits the SPE store instruction to store the value_reg.
+ * \param value_reg  the SPE register containing the value to store.
+ *                   This would have been returned by get_dst_reg().
+ */
+static void
+store_dest_reg(struct codegen *gen,
+               int value_reg, int channel,
+               const struct tgsi_full_dst_register *dest)
+{
+   /*
+    * XXX need to implement dst reg clamping/saturation
+    */
+#if 0
+   switch (inst->Instruction.Saturate) {
+   case TGSI_SAT_NONE:
+      break;
+   case TGSI_SAT_ZERO_ONE:
+      break;
+   case TGSI_SAT_MINUS_PLUS_ONE:
+      break;
+   default:
+      assert( 0 );
+   }
+#endif
+
+   switch (dest->DstRegister.File) {
+   case TGSI_FILE_TEMPORARY:
+      if (gen->if_nesting > 0 || gen->loop_nesting > 0) {
+         int d_reg = gen->temp_regs[dest->DstRegister.Index][channel];
+         int exec_reg = get_exec_mask_reg(gen);
+         /* Mix d with new value according to exec mask:
+          * d[i] = mask_reg[i] ? value_reg : d_reg
+          */
+         spe_selb(gen->f, d_reg, d_reg, value_reg, exec_reg);
+      }
+      else {
+         /* we're not inside a condition or loop: do nothing special */
+
+      }
+      break;
+   case TGSI_FILE_OUTPUT:
+      {
+         /* offset is measured in quadwords, not bytes */
+         int offset = dest->DstRegister.Index * 4 + channel;
+         if (gen->if_nesting > 0 || gen->loop_nesting > 0) {
+            int exec_reg = get_exec_mask_reg(gen);
+            int curval_reg = get_itemp(gen);
+            /* First read the current value from memory:
+             * Load:  curval = memory[(machine_reg) + offset]
+             */
+            spe_lqd(gen->f, curval_reg, gen->outputs_reg, offset * 16);
+            /* Mix curval with newvalue according to exec mask:
+             * d[i] = mask_reg[i] ? value_reg : d_reg
+             */
+            spe_selb(gen->f, curval_reg, curval_reg, value_reg, exec_reg);
+            /* Store: memory[(machine_reg) + offset] = curval */
+            spe_stqd(gen->f, curval_reg, gen->outputs_reg, offset * 16);
+         }
+         else {
+            /* Store: memory[(machine_reg) + offset] = reg */
+            spe_stqd(gen->f, value_reg, gen->outputs_reg, offset * 16);
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+
+static void
+emit_prologue(struct codegen *gen)
+{
+   gen->frame_size = 1024; /* XXX temporary, should be dynamic */
+
+   spe_comment(gen->f, 0, "Function prologue:");
+
+   /* save $lr on stack     # stqd $lr,16($sp) */
+   spe_stqd(gen->f, SPE_REG_RA, SPE_REG_SP, 16);
+
+   if (gen->frame_size >= 512) {
+      /* offset is too large for ai instruction */
+      int offset_reg = spe_allocate_available_register(gen->f);
+      int sp_reg = spe_allocate_available_register(gen->f);
+      /* offset = -framesize */
+      spe_load_int(gen->f, offset_reg, -gen->frame_size);
+      /* sp = $sp */
+      spe_move(gen->f, sp_reg, SPE_REG_SP);
+      /* $sp = $sp + offset_reg */
+      spe_a(gen->f, SPE_REG_SP, SPE_REG_SP, offset_reg);
+      /* save $sp in stack frame */
+      spe_stqd(gen->f, sp_reg, SPE_REG_SP, 0);
+      /* clean up */
+      spe_release_register(gen->f, offset_reg);
+      spe_release_register(gen->f, sp_reg);
+   }
+   else {
+      /* save stack pointer    # stqd $sp,-frameSize($sp) */
+      spe_stqd(gen->f, SPE_REG_SP, SPE_REG_SP, -gen->frame_size);
+
+      /* adjust stack pointer  # ai $sp,$sp,-frameSize */
+      spe_ai(gen->f, SPE_REG_SP, SPE_REG_SP, -gen->frame_size);
+   }
+}
+
+
+static void
+emit_epilogue(struct codegen *gen)
+{
+   const int return_reg = 3;
+
+   spe_comment(gen->f, 0, "Function epilogue:");
+
+   spe_comment(gen->f, 0, "return the killed mask");
+   if (gen->kill_mask_reg > 0) {
+      /* shader called KIL, return the "alive" mask */
+      spe_move(gen->f, return_reg, gen->kill_mask_reg);
+   }
+   else {
+      /* return {0,0,0,0} */
+      spe_load_uint(gen->f, return_reg, 0);
+   }
+
+   spe_comment(gen->f, 0, "restore stack and return");
+   if (gen->frame_size >= 512) {
+      /* offset is too large for ai instruction */
+      int offset_reg = spe_allocate_available_register(gen->f);
+      /* offset = framesize */
+      spe_load_int(gen->f, offset_reg, gen->frame_size);
+      /* $sp = $sp + offset */
+      spe_a(gen->f, SPE_REG_SP, SPE_REG_SP, offset_reg);
+      /* clean up */
+      spe_release_register(gen->f, offset_reg);
+   }
+   else {
+      /* restore stack pointer    # ai $sp,$sp,frameSize */
+      spe_ai(gen->f, SPE_REG_SP, SPE_REG_SP, gen->frame_size);
+   }
+
+   /* restore $lr              # lqd $lr,16($sp) */
+   spe_lqd(gen->f, SPE_REG_RA, SPE_REG_SP, 16);
+
+   /* return from function call */
+   spe_bi(gen->f, SPE_REG_RA, 0, 0);
+}
+
+
+#define FOR_EACH_ENABLED_CHANNEL(inst, ch) \
+   for (ch = 0; ch < 4; ch++) \
+      if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << ch))
+
+
+static boolean
+emit_ARL(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch = 0, src_reg, addr_reg;
+
+   src_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+   addr_reg = get_address_reg(gen);
+
+   /* convert float to int */
+   spe_cflts(gen->f, addr_reg, src_reg, 0);
+
+   free_itemps(gen);
+
+   return TRUE;
+}
+
+
+static boolean
+emit_MOV(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, src_reg[4], dst_reg[4];
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      src_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      dst_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      if (is_register_src(gen, ch, &inst->FullSrcRegisters[0]) &&
+          is_memory_dst(gen, ch, &inst->FullDstRegisters[0])) {
+         /* special-case: register to memory store */
+         store_dest_reg(gen, src_reg[ch], ch, &inst->FullDstRegisters[0]);
+      }
+      else {
+         spe_move(gen->f, dst_reg[ch], src_reg[ch]);
+         store_dest_reg(gen, dst_reg[ch], ch, &inst->FullDstRegisters[0]);
+      }
+   }
+
+   free_itemps(gen);
+
+   return TRUE;
+}
+
+/**
+ * Emit binary operation
+ */
+static boolean
+emit_binop(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, s1_reg[4], s2_reg[4], d_reg[4];
+
+   /* Loop over Red/Green/Blue/Alpha channels, fetch src operands */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      s2_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]);
+      d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+   }
+
+   /* Loop over Red/Green/Blue/Alpha channels, do the op, store results */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      /* Emit actual SPE instruction: d = s1 + s2 */
+      switch (inst->Instruction.Opcode) {
+      case TGSI_OPCODE_ADD:
+         spe_fa(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch]);
+         break;
+      case TGSI_OPCODE_SUB:
+         spe_fs(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch]);
+         break;
+      case TGSI_OPCODE_MUL:
+         spe_fm(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch]);
+         break;
+      default:
+         ;
+      }
+   }
+
+   /* Store the result (a no-op for TGSI_FILE_TEMPORARY dests) */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]);
+   }
+
+   /* Free any intermediate temps we allocated */
+   free_itemps(gen);
+
+   return TRUE;
+}
+
+
+/**
+ * Emit multiply add.  See emit_ADD for comments.
+ */
+static boolean
+emit_MAD(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, s1_reg[4], s2_reg[4], s3_reg[4], d_reg[4];
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      s2_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]);
+      s3_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[2]);
+      d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+   }
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_fma(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch], s3_reg[ch]);
+   }
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]);
+   }
+   free_itemps(gen);
+   return TRUE;
+}
+
+
+/**
+ * Emit linear interpolate.  See emit_ADD for comments.
+ */
+static boolean
+emit_LERP(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, s1_reg[4], s2_reg[4], s3_reg[4], d_reg[4], tmp_reg[4];
+
+   /* setup/get src/dst/temp regs */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      s2_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]);
+      s3_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[2]);
+      d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+      tmp_reg[ch] = get_itemp(gen);
+   }
+
+   /* d = s3 + s1(s2 - s3) */
+   /* do all subtracts, then all fma, then all stores to better pipeline */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_fs(gen->f, tmp_reg[ch], s2_reg[ch], s3_reg[ch]);
+   }
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_fma(gen->f, d_reg[ch], tmp_reg[ch], s1_reg[ch], s3_reg[ch]);
+   }
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]);
+   }
+   free_itemps(gen);
+   return TRUE;
+}
+
+
+
+/**
+ * Emit reciprocal or recip sqrt.
+ */
+static boolean
+emit_RCP_RSQ(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, s1_reg[4], d_reg[4], tmp_reg[4];
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+      tmp_reg[ch] = get_itemp(gen);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      if (inst->Instruction.Opcode == TGSI_OPCODE_RCP) {
+         /* tmp = 1/s1 */
+         spe_frest(gen->f, tmp_reg[ch], s1_reg[ch]);
+      }
+      else {
+         /* tmp = 1/sqrt(s1) */
+         spe_frsqest(gen->f, tmp_reg[ch], s1_reg[ch]);
+      }
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      /* d = float_interp(s1, tmp) */
+      spe_fi(gen->f, d_reg[ch], s1_reg[ch], tmp_reg[ch]);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+
+/**
+ * Emit absolute value.  See emit_ADD for comments.
+ */
+static boolean
+emit_ABS(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, s1_reg[4], d_reg[4];
+   const int bit31mask_reg = get_itemp(gen);
+
+   /* mask with bit 31 set, the rest cleared */  
+   spe_load_uint(gen->f, bit31mask_reg, (1 << 31));
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+   }
+
+   /* d = sign bit cleared in s1 */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_andc(gen->f, d_reg[ch], s1_reg[ch], bit31mask_reg);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+/**
+ * Emit 3 component dot product.  See emit_ADD for comments.
+ */
+static boolean
+emit_DP3(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch;
+   int s1x_reg, s1y_reg, s1z_reg;
+   int s2x_reg, s2y_reg, s2z_reg;
+   int t0_reg = get_itemp(gen), t1_reg = get_itemp(gen);
+
+   s1x_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]);
+   s2x_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[1]);
+   s1y_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]);
+   s2y_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[1]);
+   s1z_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]);
+   s2z_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[1]);
+
+   /* t0 = x0 * x1 */
+   spe_fm(gen->f, t0_reg, s1x_reg, s2x_reg);
+
+   /* t1 = y0 * y1 */
+   spe_fm(gen->f, t1_reg, s1y_reg, s2y_reg);
+
+   /* t0 = z0 * z1 + t0 */
+   spe_fma(gen->f, t0_reg, s1z_reg, s2z_reg, t0_reg);
+
+   /* t0 = t0 + t1 */
+   spe_fa(gen->f, t0_reg, t0_reg, t1_reg);
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+      spe_move(gen->f, d_reg, t0_reg);
+      store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+/**
+ * Emit 4 component dot product.  See emit_ADD for comments.
+ */
+static boolean
+emit_DP4(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch;
+   int s0x_reg, s0y_reg, s0z_reg, s0w_reg;
+   int s1x_reg, s1y_reg, s1z_reg, s1w_reg;
+   int t0_reg = get_itemp(gen), t1_reg = get_itemp(gen);
+
+   s0x_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]);
+   s1x_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[1]);
+   s0y_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]);
+   s1y_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[1]);
+   s0z_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]);
+   s1z_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[1]);
+   s0w_reg = get_src_reg(gen, CHAN_W, &inst->FullSrcRegisters[0]);
+   s1w_reg = get_src_reg(gen, CHAN_W, &inst->FullSrcRegisters[1]);
+
+   /* t0 = x0 * x1 */
+   spe_fm(gen->f, t0_reg, s0x_reg, s1x_reg);
+
+   /* t1 = y0 * y1 */
+   spe_fm(gen->f, t1_reg, s0y_reg, s1y_reg);
+
+   /* t0 = z0 * z1 + t0 */
+   spe_fma(gen->f, t0_reg, s0z_reg, s1z_reg, t0_reg);
+
+   /* t1 = w0 * w1 + t1 */
+   spe_fma(gen->f, t1_reg, s0w_reg, s1w_reg, t1_reg);
+
+   /* t0 = t0 + t1 */
+   spe_fa(gen->f, t0_reg, t0_reg, t1_reg);
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+      spe_move(gen->f, d_reg, t0_reg);
+      store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+/**
+ * Emit homogeneous dot product.  See emit_ADD for comments.
+ */
+static boolean
+emit_DPH(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   /* XXX rewrite this function to look more like DP3/DP4 */
+   int ch;
+   int s1_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]);
+   int s2_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[1]);
+   int tmp_reg = get_itemp(gen);
+
+   /* t = x0 * x1 */
+   spe_fm(gen->f, tmp_reg, s1_reg, s2_reg);
+
+   s1_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]);
+   s2_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[1]);
+   /* t = y0 * y1 + t */
+   spe_fma(gen->f, tmp_reg, s1_reg, s2_reg, tmp_reg);
+
+   s1_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]);
+   s2_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[1]);
+   /* t = z0 * z1 + t */
+   spe_fma(gen->f, tmp_reg, s1_reg, s2_reg, tmp_reg);
+
+   s2_reg = get_src_reg(gen, CHAN_W, &inst->FullSrcRegisters[1]);
+   /* t = w1 + t */
+   spe_fa(gen->f, tmp_reg, s2_reg, tmp_reg);
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+      spe_move(gen->f, d_reg, tmp_reg);
+      store_dest_reg(gen, tmp_reg, ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+/**
+ * Emit 3-component vector normalize.
+ */
+static boolean
+emit_NRM3(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch;
+   int src_reg[3];
+   int t0_reg = get_itemp(gen), t1_reg = get_itemp(gen);
+
+   src_reg[0] = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]);
+   src_reg[1] = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]);
+   src_reg[2] = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]);
+
+   /* t0 = x * x */
+   spe_fm(gen->f, t0_reg, src_reg[0], src_reg[0]);
+
+   /* t1 = y * y */
+   spe_fm(gen->f, t1_reg, src_reg[1], src_reg[1]);
+
+   /* t0 = z * z + t0 */
+   spe_fma(gen->f, t0_reg, src_reg[2], src_reg[2], t0_reg);
+
+   /* t0 = t0 + t1 */
+   spe_fa(gen->f, t0_reg, t0_reg, t1_reg);
+
+   /* t1 = 1.0 / sqrt(t0) */
+   spe_frsqest(gen->f, t1_reg, t0_reg);
+   spe_fi(gen->f, t1_reg, t0_reg, t1_reg);
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+      /* dst = src[ch] * t1 */
+      spe_fm(gen->f, d_reg, src_reg[ch], t1_reg);
+      store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+
+/**
+ * Emit cross product.  See emit_ADD for comments.
+ */
+static boolean
+emit_XPD(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int s1_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]);
+   int s2_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[1]);
+   int tmp_reg = get_itemp(gen);
+
+   /* t = z0 * y1 */
+   spe_fm(gen->f, tmp_reg, s1_reg, s2_reg);
+
+   s1_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]);
+   s2_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[1]);
+   /* t = y0 * z1 - t */
+   spe_fms(gen->f, tmp_reg, s1_reg, s2_reg, tmp_reg);
+
+   if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << CHAN_X)) {
+      store_dest_reg(gen, tmp_reg, CHAN_X, &inst->FullDstRegisters[0]);
+   }
+
+   s1_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]);
+   s2_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[1]);
+   /* t = x0 * z1 */
+   spe_fm(gen->f, tmp_reg, s1_reg, s2_reg);
+
+   s1_reg = get_src_reg(gen, CHAN_Z, &inst->FullSrcRegisters[0]);
+   s2_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[1]);
+   /* t = z0 * x1 - t */
+   spe_fms(gen->f, tmp_reg, s1_reg, s2_reg, tmp_reg);
+
+   if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << CHAN_Y)) {
+      store_dest_reg(gen, tmp_reg, CHAN_Y, &inst->FullDstRegisters[0]);
+   }
+
+   s1_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[0]);
+   s2_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[1]);
+   /* t = y0 * x1 */
+   spe_fm(gen->f, tmp_reg, s1_reg, s2_reg);
+
+   s1_reg = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[0]);
+   s2_reg = get_src_reg(gen, CHAN_Y, &inst->FullSrcRegisters[1]);
+   /* t = x0 * y1 - t */
+   spe_fms(gen->f, tmp_reg, s1_reg, s2_reg, tmp_reg);
+
+   if (inst->FullDstRegisters[0].DstRegister.WriteMask & (1 << CHAN_Z)) {
+      store_dest_reg(gen, tmp_reg, CHAN_Z, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+
+/**
+ * Emit inequality instruction.
+ * Note that the SPE fcgt instruction produces 0x0 and 0xffffffff as
+ * the result but OpenGL/TGSI needs 0.0 and 1.0 results.
+ * We can easily convert 0x0/0xffffffff to 0.0/1.0 with a bitwise AND.
+ */
+static boolean
+emit_inequality(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, s1_reg[4], s2_reg[4], d_reg[4], one_reg;
+   bool complement = FALSE;
+
+   one_reg = get_const_one_reg(gen);
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      s2_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]);
+      d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      switch (inst->Instruction.Opcode) {
+      case TGSI_OPCODE_SGT:
+         spe_fcgt(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch]);
+         break;
+      case TGSI_OPCODE_SLT:
+         spe_fcgt(gen->f, d_reg[ch], s2_reg[ch], s1_reg[ch]);
+         break;
+      case TGSI_OPCODE_SGE:
+         spe_fcgt(gen->f, d_reg[ch], s2_reg[ch], s1_reg[ch]);
+         complement = TRUE;
+         break;
+      case TGSI_OPCODE_SLE:
+         spe_fcgt(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch]);
+         complement = TRUE;
+         break;
+      case TGSI_OPCODE_SEQ:
+         spe_fceq(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch]);
+         break;
+      case TGSI_OPCODE_SNE:
+         spe_fceq(gen->f, d_reg[ch], s1_reg[ch], s2_reg[ch]);
+         complement = TRUE;
+         break;
+      default:
+         assert(0);
+      }
+   }
+
+   /* convert d from 0x0/0xffffffff to 0.0/1.0 */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      /* d = d & one_reg */
+      if (complement)
+         spe_andc(gen->f, d_reg[ch], one_reg, d_reg[ch]);
+      else
+         spe_and(gen->f, d_reg[ch], one_reg, d_reg[ch]);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+
+/**
+ * Emit compare.
+ */
+static boolean
+emit_CMP(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch;
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      int s1_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      int s2_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]);
+      int s3_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[2]);
+      int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+      int zero_reg = get_itemp(gen);
+   
+      spe_zero(gen->f, zero_reg);
+
+      /* d = (s1 < 0) ? s2 : s3 */
+      spe_fcgt(gen->f, d_reg, zero_reg, s1_reg);
+      spe_selb(gen->f, d_reg, s3_reg, s2_reg, d_reg);
+
+      store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]);
+      free_itemps(gen);
+   }
+
+   return TRUE;
+}
+
+/**
+ * Emit trunc.  
+ * Convert float to signed int
+ * Convert signed int to float
+ */
+static boolean
+emit_TRUNC(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, s1_reg[4], d_reg[4];
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+   }
+
+   /* Convert float to int */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_cflts(gen->f, d_reg[ch], s1_reg[ch], 0);
+   }
+
+   /* Convert int to float */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_csflt(gen->f, d_reg[ch], d_reg[ch], 0);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+
+/**
+ * Emit floor.  
+ * If negative int subtract one
+ * Convert float to signed int
+ * Convert signed int to float
+ */
+static boolean
+emit_FLR(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, s1_reg[4], d_reg[4], tmp_reg[4], zero_reg, one_reg;
+
+   zero_reg = get_itemp(gen);
+   spe_zero(gen->f, zero_reg);
+   one_reg = get_const_one_reg(gen);
+   
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+      tmp_reg[ch] = get_itemp(gen);
+   }
+
+   /* If negative, subtract 1.0 */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_fcgt(gen->f, tmp_reg[ch], zero_reg, s1_reg[ch]);
+   }
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_selb(gen->f, tmp_reg[ch], zero_reg, one_reg, tmp_reg[ch]);
+   }
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_fs(gen->f, tmp_reg[ch], s1_reg[ch], tmp_reg[ch]);
+   }
+
+   /* Convert float to int */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_cflts(gen->f, tmp_reg[ch], tmp_reg[ch], 0);
+   }
+
+   /* Convert int to float */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_csflt(gen->f, d_reg[ch], tmp_reg[ch], 0);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+
+/**
+ * Compute frac = Input - FLR(Input)
+ */
+static boolean
+emit_FRC(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, s1_reg[4], d_reg[4], tmp_reg[4], zero_reg, one_reg;
+
+   zero_reg = get_itemp(gen);
+   spe_zero(gen->f, zero_reg);
+   one_reg = get_const_one_reg(gen);
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+      tmp_reg[ch] = get_itemp(gen);
+   }
+
+   /* If negative, subtract 1.0 */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_fcgt(gen->f, tmp_reg[ch], zero_reg, s1_reg[ch]);
+   }
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_selb(gen->f, tmp_reg[ch], zero_reg, one_reg, tmp_reg[ch]);
+   }
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_fs(gen->f, tmp_reg[ch], s1_reg[ch], tmp_reg[ch]);
+   }
+
+   /* Convert float to int */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_cflts(gen->f, tmp_reg[ch], tmp_reg[ch], 0);
+   }
+
+   /* Convert int to float */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_csflt(gen->f, tmp_reg[ch], tmp_reg[ch], 0);
+   }
+
+   /* d = s1 - FLR(s1) */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_fs(gen->f, d_reg[ch], s1_reg[ch], tmp_reg[ch]);
+   }
+
+   /* store result */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+
+#if 0
+static void
+print_functions(struct cell_context *cell)
+{
+   struct cell_spu_function_info *funcs = &cell->spu_functions;
+   uint i;
+   for (i = 0; i < funcs->num; i++) {
+      printf("SPU func %u: %s at %u\n",
+             i, funcs->names[i], funcs->addrs[i]);
+   }
+}
+#endif
+
+
+static uint
+lookup_function(struct cell_context *cell, const char *funcname)
+{
+   const struct cell_spu_function_info *funcs = &cell->spu_functions;
+   uint i, addr = 0;
+   for (i = 0; i < funcs->num; i++) {
+      if (strcmp(funcs->names[i], funcname) == 0) {
+         addr = funcs->addrs[i];
+      }
+   }
+   assert(addr && "spu function not found");
+   return addr / 4;  /* discard 2 least significant bits */
+}
+
+
+/**
+ * Emit code to call a SPU function.
+ * Used to implement instructions like SIN/COS/POW/TEX/etc.
+ * If scalar, only the X components of the src regs are used, and the
+ * result is replicated across the dest register's XYZW components.
+ */
+static boolean
+emit_function_call(struct codegen *gen,
+                   const struct tgsi_full_instruction *inst,
+                   char *funcname, uint num_args, boolean scalar)
+{
+   const uint addr = lookup_function(gen->cell, funcname);
+   char comment[100];
+   int s_regs[3];
+   int func_called = FALSE;
+   uint a, ch;
+   int retval_reg = -1;
+
+   assert(num_args <= 3);
+
+   snprintf(comment, sizeof(comment), "CALL %s:", funcname);
+   spe_comment(gen->f, -4, comment);
+
+   if (scalar) {
+      for (a = 0; a < num_args; a++) {
+         s_regs[a] = get_src_reg(gen, CHAN_X, &inst->FullSrcRegisters[a]);
+      }
+      /* we'll call the function, put the return value in this register,
+       * then replicate it across all write-enabled components in d_reg.
+       */
+      retval_reg = spe_allocate_available_register(gen->f);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      int d_reg;
+      ubyte usedRegs[SPE_NUM_REGS];
+      uint i, numUsed;
+
+      if (!scalar) {
+         for (a = 0; a < num_args; a++) {
+            s_regs[a] = get_src_reg(gen, ch, &inst->FullSrcRegisters[a]);
+         }
+      }
+
+      d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+
+      if (!scalar || !func_called) {
+         /* for a scalar function, we'll really only call the function once */
+
+         numUsed = spe_get_registers_used(gen->f, usedRegs);
+         assert(numUsed < gen->frame_size / 16 - 2);
+
+         /* save registers to stack */
+         for (i = 0; i < numUsed; i++) {
+            uint reg = usedRegs[i];
+            int offset = 2 + i;
+            spe_stqd(gen->f, reg, SPE_REG_SP, 16 * offset);
+         }
+
+         /* setup function arguments */
+         for (a = 0; a < num_args; a++) {
+            spe_move(gen->f, 3 + a, s_regs[a]);
+         }
+
+         /* branch to function, save return addr */
+         spe_brasl(gen->f, SPE_REG_RA, addr);
+
+         /* save function's return value */
+         if (scalar)
+            spe_move(gen->f, retval_reg, 3);
+         else
+            spe_move(gen->f, d_reg, 3);
+
+         /* restore registers from stack */
+         for (i = 0; i < numUsed; i++) {
+            uint reg = usedRegs[i];
+            if (reg != d_reg && reg != retval_reg) {
+               int offset = 2 + i;
+               spe_lqd(gen->f, reg, SPE_REG_SP, 16 * offset);
+            }
+         }
+
+         func_called = TRUE;
+      }
+
+      if (scalar) {
+         spe_move(gen->f, d_reg, retval_reg);
+      }
+
+      store_dest_reg(gen, d_reg, ch, &inst->FullDstRegisters[0]);
+      free_itemps(gen);
+   }
+
+   if (scalar) {
+      spe_release_register(gen->f, retval_reg);
+   }
+
+   return TRUE;
+}
+
+
+static boolean
+emit_TEX(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   const uint target = inst->InstructionExtTexture.Texture;
+   const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index;
+   uint addr;
+   int ch;
+   int coord_regs[4], d_regs[4];
+
+   switch (target) {
+   case TGSI_TEXTURE_1D:
+   case TGSI_TEXTURE_2D:
+      addr = lookup_function(gen->cell, "spu_tex_2d");
+      break;
+   case TGSI_TEXTURE_3D:
+      addr = lookup_function(gen->cell, "spu_tex_3d");
+      break;
+   case TGSI_TEXTURE_CUBE:
+      addr = lookup_function(gen->cell, "spu_tex_cube");
+      break;
+   default:
+      ASSERT(0 && "unsupported texture target");
+      return FALSE;
+   }
+
+   assert(inst->FullSrcRegisters[1].SrcRegister.File == TGSI_FILE_SAMPLER);
+
+   spe_comment(gen->f, -4, "CALL tex:");
+
+   /* get src/dst reg info */
+   for (ch = 0; ch < 4; ch++) {
+      coord_regs[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      d_regs[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+   }
+
+   {
+      ubyte usedRegs[SPE_NUM_REGS];
+      uint i, numUsed;
+
+      numUsed = spe_get_registers_used(gen->f, usedRegs);
+      assert(numUsed < gen->frame_size / 16 - 2);
+
+      /* save registers to stack */
+      for (i = 0; i < numUsed; i++) {
+         uint reg = usedRegs[i];
+         int offset = 2 + i;
+         spe_stqd(gen->f, reg, SPE_REG_SP, 16 * offset);
+      }
+
+      /* setup function arguments (XXX depends on target) */
+      for (i = 0; i < 4; i++) {
+         spe_move(gen->f, 3 + i, coord_regs[i]);
+      }
+      spe_load_uint(gen->f, 7, unit); /* sampler unit */
+
+      /* branch to function, save return addr */
+      spe_brasl(gen->f, SPE_REG_RA, addr);
+
+      /* save function's return values (four pixel's colors) */
+      for (i = 0; i < 4; i++) {
+         spe_move(gen->f, d_regs[i], 3 + i);
+      }
+
+      /* restore registers from stack */
+      for (i = 0; i < numUsed; i++) {
+         uint reg = usedRegs[i];
+         if (reg != d_regs[0] &&
+             reg != d_regs[1] &&
+             reg != d_regs[2] &&
+             reg != d_regs[3]) {
+            int offset = 2 + i;
+            spe_lqd(gen->f, reg, SPE_REG_SP, 16 * offset);
+         }
+      }
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_regs[ch], ch, &inst->FullDstRegisters[0]);
+      free_itemps(gen);
+   }
+
+   return TRUE;
+}
+
+
+/**
+ * KILL if any of src reg values are less than zero.
+ */
+static boolean
+emit_KIL(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch;
+   int s_regs[4], kil_reg = -1, cmp_reg, zero_reg;
+
+   spe_comment(gen->f, -4, "CALL kil:");
+
+   /* zero = {0,0,0,0} */
+   zero_reg = get_itemp(gen);
+   spe_zero(gen->f, zero_reg);
+
+   cmp_reg = get_itemp(gen);
+
+   /* get src regs */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s_regs[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+   }
+
+   /* test if any src regs are < 0 */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      if (kil_reg >= 0) {
+         /* cmp = 0 > src ? : ~0 : 0 */
+         spe_fcgt(gen->f, cmp_reg, zero_reg, s_regs[ch]);
+         /* kil = kil | cmp */
+         spe_or(gen->f, kil_reg, kil_reg, cmp_reg);
+      }
+      else {
+         kil_reg = get_itemp(gen);
+         /* kil = 0 > src ? : ~0 : 0 */
+         spe_fcgt(gen->f, kil_reg, zero_reg, s_regs[ch]);
+      }
+   }
+
+   if (gen->if_nesting || gen->loop_nesting) {
+      /* may have been a conditional kil */
+      spe_and(gen->f, kil_reg, kil_reg, gen->exec_mask_reg);
+   }
+
+   /* allocate the kill mask reg if needed */
+   if (gen->kill_mask_reg <= 0) {
+      gen->kill_mask_reg = spe_allocate_available_register(gen->f);
+      spe_move(gen->f, gen->kill_mask_reg, kil_reg);
+   }
+   else {
+      spe_or(gen->f, gen->kill_mask_reg, gen->kill_mask_reg, kil_reg);
+   }
+
+   free_itemps(gen);
+
+   return TRUE;
+}
+
+
+
+/**
+ * Emit min or max.
+ */
+static boolean
+emit_MIN_MAX(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int ch, s0_reg[4], s1_reg[4], d_reg[4], tmp_reg[4];
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      s0_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      s1_reg[ch] = get_src_reg(gen, ch, &inst->FullSrcRegisters[1]);
+      d_reg[ch] = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+      tmp_reg[ch] = get_itemp(gen);         
+   }
+
+   /* d = (s0 > s1) ? s0 : s1 */
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      if (inst->Instruction.Opcode == TGSI_OPCODE_MAX)
+         spe_fcgt(gen->f, tmp_reg[ch], s0_reg[ch], s1_reg[ch]);
+      else
+         spe_fcgt(gen->f, tmp_reg[ch], s1_reg[ch], s0_reg[ch]);
+   }
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      spe_selb(gen->f, d_reg[ch], s1_reg[ch], s0_reg[ch], tmp_reg[ch]);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      store_dest_reg(gen, d_reg[ch], ch, &inst->FullDstRegisters[0]);
+   }
+
+   free_itemps(gen);
+   return TRUE;
+}
+
+
+/**
+ * Emit code to update the execution mask.
+ * This needs to be done whenever the execution status of a conditional
+ * or loop is changed.
+ */
+static void
+emit_update_exec_mask(struct codegen *gen)
+{
+   const int exec_reg = get_exec_mask_reg(gen);
+   const int cond_reg = gen->cond_mask_reg;
+   const int loop_reg = gen->loop_mask_reg;
+
+   spe_comment(gen->f, 0, "Update master execution mask");
+
+   if (gen->if_nesting > 0 && gen->loop_nesting > 0) {
+      /* exec_mask = cond_mask & loop_mask */
+      assert(cond_reg > 0);
+      assert(loop_reg > 0);
+      spe_and(gen->f, exec_reg, cond_reg, loop_reg);
+   }
+   else if (gen->if_nesting > 0) {
+      assert(cond_reg > 0);
+      spe_move(gen->f, exec_reg, cond_reg);
+   }
+   else if (gen->loop_nesting > 0) {
+      assert(loop_reg > 0);
+      spe_move(gen->f, exec_reg, loop_reg);
+   }
+   else {
+      spe_load_int(gen->f, exec_reg, ~0x0);
+   }
+}
+
+
+static boolean
+emit_IF(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   const int channel = 0;
+   int cond_reg;
+
+   cond_reg = get_cond_mask_reg(gen);
+
+   /* XXX push cond exec mask */
+
+   spe_comment(gen->f,  0, "init conditional exec mask = ~0:");
+   spe_load_int(gen->f, cond_reg, ~0);
+
+   /* update conditional execution mask with the predicate register */
+   int tmp_reg = get_itemp(gen);
+   int s1_reg = get_src_reg(gen, channel, &inst->FullSrcRegisters[0]);
+
+   /* tmp = (s1_reg == 0) */
+   spe_ceqi(gen->f, tmp_reg, s1_reg, 0);
+   /* tmp = !tmp */
+   spe_complement(gen->f, tmp_reg, tmp_reg);
+   /* cond_mask = cond_mask & tmp */
+   spe_and(gen->f, cond_reg, cond_reg, tmp_reg);
+
+   gen->if_nesting++;
+
+   /* update the master execution mask */
+   emit_update_exec_mask(gen);
+
+   free_itemps(gen);
+
+   return TRUE;
+}
+
+
+static boolean
+emit_ELSE(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   const int cond_reg = get_cond_mask_reg(gen);
+
+   spe_comment(gen->f, 0, "cond exec mask = !cond exec mask");
+   spe_complement(gen->f, cond_reg, cond_reg);
+   emit_update_exec_mask(gen);
+
+   return TRUE;
+}
+
+
+static boolean
+emit_ENDIF(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   /* XXX todo: pop cond exec mask */
+
+   gen->if_nesting--;
+
+   emit_update_exec_mask(gen);
+
+   return TRUE;
+}
+
+
+static boolean
+emit_BGNLOOP(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   int exec_reg, loop_reg;
+
+   exec_reg = get_exec_mask_reg(gen);
+   loop_reg = get_loop_mask_reg(gen);
+
+   /* XXX push loop_exec mask */
+
+   spe_comment(gen->f,  0*-4, "initialize loop exec mask = ~0");
+   spe_load_int(gen->f, loop_reg, ~0x0);
+
+   gen->loop_nesting++;
+   gen->loop_start = spe_code_size(gen->f);  /* in bytes */
+
+   return TRUE;
+}
+
+
+static boolean
+emit_ENDLOOP(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   const int loop_reg = get_loop_mask_reg(gen);
+   const int tmp_reg = get_itemp(gen);
+   int offset;
+
+   /* tmp_reg = exec[0] | exec[1] | exec[2] | exec[3] */
+   spe_orx(gen->f, tmp_reg, loop_reg);
+
+   offset = gen->loop_start - spe_code_size(gen->f); /* in bytes */
+
+   /* branch back to top of loop if tmp_reg != 0 */
+   spe_brnz(gen->f, tmp_reg, offset / 4);
+
+   /* XXX pop loop_exec mask */
+
+   gen->loop_nesting--;
+
+   emit_update_exec_mask(gen);
+
+   return TRUE;
+}
+
+
+static boolean
+emit_BRK(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   const int exec_reg = get_exec_mask_reg(gen);
+   const int loop_reg = get_loop_mask_reg(gen);
+
+   assert(gen->loop_nesting > 0);
+
+   spe_comment(gen->f, 0, "loop exec mask &= ~master exec mask");
+   spe_andc(gen->f, loop_reg, loop_reg, exec_reg);
+
+   emit_update_exec_mask(gen);
+
+   return TRUE;
+}
+
+
+static boolean
+emit_CONT(struct codegen *gen, const struct tgsi_full_instruction *inst)
+{
+   assert(gen->loop_nesting > 0);
+
+   return TRUE;
+}
+
+
+static boolean
+emit_DDX_DDY(struct codegen *gen, const struct tgsi_full_instruction *inst,
+             boolean ddx)
+{
+   int ch;
+
+   FOR_EACH_ENABLED_CHANNEL(inst, ch) {
+      int s_reg = get_src_reg(gen, ch, &inst->FullSrcRegisters[0]);
+      int d_reg = get_dst_reg(gen, ch, &inst->FullDstRegisters[0]);
+
+      int t1_reg = get_itemp(gen);
+      int t2_reg = get_itemp(gen);
+
+      spe_splat_word(gen->f, t1_reg, s_reg, 0); /* upper-left pixel */
+      if (ddx) {
+         spe_splat_word(gen->f, t2_reg, s_reg, 1); /* upper-right pixel */
+      }
+      else {
+         spe_splat_word(gen->f, t2_reg, s_reg, 2); /* lower-left pixel */
+      }
+      spe_fs(gen->f, d_reg, t2_reg, t1_reg);
+
+      free_itemps(gen);
+   }
+
+   return TRUE;
+}
+
+
+
+
+/**
+ * Emit END instruction.
+ * We just return from the shader function at this point.
+ *
+ * Note that there may be more code after this that would be
+ * called by TGSI_OPCODE_CALL.
+ */
+static boolean
+emit_END(struct codegen *gen)
+{
+   emit_epilogue(gen);
+   return TRUE;
+}
+
+
+/**
+ * Emit code for the given instruction.  Just a big switch stmt.
+ */
+static boolean
+emit_instruction(struct codegen *gen,
+                 const struct tgsi_full_instruction *inst)
+{
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_ARL:
+      return emit_ARL(gen, inst);
+   case TGSI_OPCODE_MOV:
+   case TGSI_OPCODE_SWZ:
+      return emit_MOV(gen, inst);
+   case TGSI_OPCODE_ADD:
+   case TGSI_OPCODE_SUB:
+   case TGSI_OPCODE_MUL:
+      return emit_binop(gen, inst);
+   case TGSI_OPCODE_MAD:
+      return emit_MAD(gen, inst);
+   case TGSI_OPCODE_LERP:
+      return emit_LERP(gen, inst);
+   case TGSI_OPCODE_DP3:
+      return emit_DP3(gen, inst);
+   case TGSI_OPCODE_DP4:
+      return emit_DP4(gen, inst);
+   case TGSI_OPCODE_DPH:
+      return emit_DPH(gen, inst);
+   case TGSI_OPCODE_NRM:
+      return emit_NRM3(gen, inst);
+   case TGSI_OPCODE_XPD:
+      return emit_XPD(gen, inst);
+   case TGSI_OPCODE_RCP:
+   case TGSI_OPCODE_RSQ:
+      return emit_RCP_RSQ(gen, inst);
+   case TGSI_OPCODE_ABS:
+      return emit_ABS(gen, inst);
+   case TGSI_OPCODE_SGT:
+   case TGSI_OPCODE_SLT:
+   case TGSI_OPCODE_SGE:
+   case TGSI_OPCODE_SLE:
+   case TGSI_OPCODE_SEQ:
+   case TGSI_OPCODE_SNE:
+      return emit_inequality(gen, inst);
+   case TGSI_OPCODE_CMP:
+      return emit_CMP(gen, inst);
+   case TGSI_OPCODE_MIN:
+   case TGSI_OPCODE_MAX:
+      return emit_MIN_MAX(gen, inst);
+   case TGSI_OPCODE_TRUNC:
+      return emit_TRUNC(gen, inst);
+   case TGSI_OPCODE_FLR:
+      return emit_FLR(gen, inst);
+   case TGSI_OPCODE_FRC:
+      return emit_FRC(gen, inst);
+   case TGSI_OPCODE_END:
+      return emit_END(gen);
+
+   case TGSI_OPCODE_COS:
+      return emit_function_call(gen, inst, "spu_cos", 1, TRUE);
+   case TGSI_OPCODE_SIN:
+      return emit_function_call(gen, inst, "spu_sin", 1, TRUE);
+   case TGSI_OPCODE_POW:
+      return emit_function_call(gen, inst, "spu_pow", 2, TRUE);
+   case TGSI_OPCODE_EXPBASE2:
+      return emit_function_call(gen, inst, "spu_exp2", 1, TRUE);
+   case TGSI_OPCODE_LOGBASE2:
+      return emit_function_call(gen, inst, "spu_log2", 1, TRUE);
+   case TGSI_OPCODE_TEX:
+      /* fall-through for now */
+   case TGSI_OPCODE_TXD:
+      /* fall-through for now */
+   case TGSI_OPCODE_TXB:
+      /* fall-through for now */
+   case TGSI_OPCODE_TXL:
+      /* fall-through for now */
+   case TGSI_OPCODE_TXP:
+      return emit_TEX(gen, inst);
+   case TGSI_OPCODE_KIL:
+      return emit_KIL(gen, inst);
+
+   case TGSI_OPCODE_IF:
+      return emit_IF(gen, inst);
+   case TGSI_OPCODE_ELSE:
+      return emit_ELSE(gen, inst);
+   case TGSI_OPCODE_ENDIF:
+      return emit_ENDIF(gen, inst);
+
+   case TGSI_OPCODE_BGNLOOP2:
+      return emit_BGNLOOP(gen, inst);
+   case TGSI_OPCODE_ENDLOOP2:
+      return emit_ENDLOOP(gen, inst);
+   case TGSI_OPCODE_BRK:
+      return emit_BRK(gen, inst);
+   case TGSI_OPCODE_CONT:
+      return emit_CONT(gen, inst);
+
+   case TGSI_OPCODE_DDX:
+      return emit_DDX_DDY(gen, inst, TRUE);
+   case TGSI_OPCODE_DDY:
+      return emit_DDX_DDY(gen, inst, FALSE);
+
+   /* XXX lots more cases to do... */
+
+   default:
+      fprintf(stderr, "Cell: unimplemented TGSI instruction %d!\n",
+              inst->Instruction.Opcode);
+      return FALSE;
+   }
+
+   return TRUE;
+}
+
+
+
+/**
+ * Emit code for a TGSI immediate value (vector of four floats).
+ * This involves register allocation and initialization.
+ * XXX the initialization should be done by a "prepare" stage, not
+ * per quad execution!
+ */
+static boolean
+emit_immediate(struct codegen *gen, const struct tgsi_full_immediate *immed)
+{
+   int ch;
+
+   assert(gen->num_imm < MAX_TEMPS);
+
+   for (ch = 0; ch < 4; ch++) {
+      float val = immed->u.ImmediateFloat32[ch].Float;
+
+      if (ch > 0 && val == immed->u.ImmediateFloat32[ch - 1].Float) {
+         /* re-use previous register */
+         gen->imm_regs[gen->num_imm][ch] = gen->imm_regs[gen->num_imm][ch - 1];
+      }
+      else {
+         char str[100];
+         int reg = spe_allocate_available_register(gen->f);
+
+         if (reg < 0)
+            return FALSE;
+
+         sprintf(str, "init $%d = %f", reg, val);
+         spe_comment(gen->f, 0, str);
+
+         /* update immediate map */
+         gen->imm_regs[gen->num_imm][ch] = reg;
+
+         /* emit initializer instruction */
+         spe_load_float(gen->f, reg, val);
+      }
+   }
+
+   gen->num_imm++;
+
+   return TRUE;
+}
+
+
+
+/**
+ * Emit "code" for a TGSI declaration.
+ * We only care about TGSI TEMPORARY register declarations at this time.
+ * For each TGSI TEMPORARY we allocate four SPE registers.
+ */
+static boolean
+emit_declaration(struct cell_context *cell,
+                 struct codegen *gen, const struct tgsi_full_declaration *decl)
+{
+   int i, ch;
+
+   switch (decl->Declaration.File) {
+   case TGSI_FILE_TEMPORARY:
+      for (i = decl->DeclarationRange.First;
+           i <= decl->DeclarationRange.Last;
+           i++) {
+         assert(i < MAX_TEMPS);
+         for (ch = 0; ch < 4; ch++) {
+            gen->temp_regs[i][ch] = spe_allocate_available_register(gen->f);
+            if (gen->temp_regs[i][ch] < 0)
+               return FALSE; /* out of regs */
+         }
+
+         /* XXX if we run out of SPE registers, we need to spill
+          * to SPU memory.  someday...
+          */
+
+         {
+            char buf[100];
+            sprintf(buf, "TGSI temp[%d] maps to SPU regs [$%d $%d $%d $%d]", i,
+                    gen->temp_regs[i][0], gen->temp_regs[i][1],
+                    gen->temp_regs[i][2], gen->temp_regs[i][3]);
+            spe_comment(gen->f, 0, buf);
+         }
+      }
+      break;
+   default:
+      ; /* ignore */
+   }
+
+   return TRUE;
+}
+
+
+
+/**
+ * Translate TGSI shader code to SPE instructions.  This is done when
+ * the state tracker gives us a new shader (via pipe->create_fs_state()).
+ *
+ * \param cell    the rendering context (in)
+ * \param tokens  the TGSI shader (in)
+ * \param f       the generated function (out)
+ */
+boolean
+cell_gen_fragment_program(struct cell_context *cell,
+                          const struct tgsi_token *tokens,
+                          struct spe_function *f)
+{
+   struct tgsi_parse_context parse;
+   struct codegen gen;
+   uint ic = 0;
+
+   memset(&gen, 0, sizeof(gen));
+   gen.cell = cell;
+   gen.f = f;
+
+   /* For SPE function calls: reg $3 = first param, $4 = second param, etc. */
+   gen.inputs_reg = 3;     /* pointer to inputs array */
+   gen.outputs_reg = 4;    /* pointer to outputs array */
+   gen.constants_reg = 5;  /* pointer to constants array */
+
+   spe_init_func(f, SPU_MAX_FRAGMENT_PROGRAM_INSTS * SPE_INST_SIZE);
+   spe_allocate_register(f, gen.inputs_reg);
+   spe_allocate_register(f, gen.outputs_reg);
+   spe_allocate_register(f, gen.constants_reg);
+
+   if (cell->debug_flags & CELL_DEBUG_ASM) {
+      spe_print_code(f, TRUE);
+      spe_indent(f, 2*8);
+      printf("Begin %s\n", __FUNCTION__);
+      tgsi_dump(tokens, 0);
+   }
+
+   tgsi_parse_init(&parse, tokens);
+
+   emit_prologue(&gen);
+
+   while (!tgsi_parse_end_of_tokens(&parse) && !gen.error) {
+      tgsi_parse_token(&parse);
+
+      switch (parse.FullToken.Token.Type) {
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         if (f->print) {
+            _debug_printf("    # ");
+            tgsi_dump_immediate(&parse.FullToken.FullImmediate);
+         }
+         if (!emit_immediate(&gen, &parse.FullToken.FullImmediate))
+            gen.error = TRUE;
+         break;
+
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         if (f->print) {
+            _debug_printf("    # ");
+            tgsi_dump_declaration(&parse.FullToken.FullDeclaration);
+         }
+         if (!emit_declaration(cell, &gen, &parse.FullToken.FullDeclaration))
+            gen.error = TRUE;
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         if (f->print) {
+            _debug_printf("    # ");
+            ic++;
+            tgsi_dump_instruction(&parse.FullToken.FullInstruction, ic);
+         }
+         if (!emit_instruction(&gen, &parse.FullToken.FullInstruction))
+            gen.error = TRUE;
+         break;
+
+      default:
+         assert(0);
+      }
+   }
+
+   if (gen.error) {
+      /* terminate the SPE code */
+      return emit_END(&gen);
+   }
+
+   if (cell->debug_flags & CELL_DEBUG_ASM) {
+      printf("cell_gen_fragment_program nr instructions: %d\n", f->num_inst);
+      printf("End %s\n", __FUNCTION__);
+   }
+
+   tgsi_parse_free( &parse );
+
+   return !gen.error;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_gen_fp.h b/src/gallium/drivers/cell/ppu/cell_gen_fp.h
new file mode 100644
index 0000000000..99faea7046
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_gen_fp.h
@@ -0,0 +1,42 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+
+#ifndef CELL_GEN_FP_H
+#define CELL_GEN_FP_H
+
+
+
+extern boolean
+cell_gen_fragment_program(struct cell_context *cell,
+                          const struct tgsi_token *tokens,
+                          struct spe_function *f);
+
+
+#endif /* CELL_GEN_FP_H */
+
diff --git a/src/gallium/drivers/cell/ppu/cell_gen_fragment.c b/src/gallium/drivers/cell/ppu/cell_gen_fragment.c
new file mode 100644
index 0000000000..66d4b3b6a3
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_gen_fragment.c
@@ -0,0 +1,2181 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Generate SPU per-fragment code (actually per-quad code).
+ * \author Brian Paul
+ * \author Bob Ellison
+ */
+
+
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "rtasm/rtasm_ppc_spe.h"
+#include "cell_context.h"
+#include "cell_gen_fragment.h"
+
+
+
+/** Do extra optimizations? */
+#define OPTIMIZATIONS 1
+
+
+/**
+ * Generate SPE code to perform Z/depth testing.
+ *
+ * \param dsa         Gallium depth/stencil/alpha state to gen code for
+ * \param f           SPE function to append instruction onto.
+ * \param mask_reg    register containing quad/pixel "alive" mask (in/out)
+ * \param ifragZ_reg  register containing integer fragment Z values (in)
+ * \param ifbZ_reg    register containing integer frame buffer Z values (in/out)
+ * \param zmask_reg   register containing result of Z test/comparison (out)
+ *
+ * Returns TRUE if the Z-buffer needs to be updated.
+ */
+static boolean
+gen_depth_test(struct spe_function *f,
+               const struct pipe_depth_stencil_alpha_state *dsa,
+               int mask_reg, int ifragZ_reg, int ifbZ_reg, int zmask_reg)
+{
+   /* NOTE: we use clgt below, not cgt, because we want to compare _unsigned_
+    * quantities.  This only makes a difference for 32-bit Z values though.
+    */
+   ASSERT(dsa->depth.enabled);
+
+   switch (dsa->depth.func) {
+   case PIPE_FUNC_EQUAL:
+      /* zmask = (ifragZ == ref) */
+      spe_ceq(f, zmask_reg, ifragZ_reg, ifbZ_reg);
+      /* mask = (mask & zmask) */
+      spe_and(f, mask_reg, mask_reg, zmask_reg);
+      break;
+
+   case PIPE_FUNC_NOTEQUAL:
+      /* zmask = (ifragZ == ref) */
+      spe_ceq(f, zmask_reg, ifragZ_reg, ifbZ_reg);
+      /* mask = (mask & ~zmask) */
+      spe_andc(f, mask_reg, mask_reg, zmask_reg);
+      break;
+
+   case PIPE_FUNC_GREATER:
+      /* zmask = (ifragZ > ref) */
+      spe_clgt(f, zmask_reg, ifragZ_reg, ifbZ_reg);
+      /* mask = (mask & zmask) */
+      spe_and(f, mask_reg, mask_reg, zmask_reg);
+      break;
+
+   case PIPE_FUNC_LESS:
+      /* zmask = (ref > ifragZ) */
+      spe_clgt(f, zmask_reg, ifbZ_reg, ifragZ_reg);
+      /* mask = (mask & zmask) */
+      spe_and(f, mask_reg, mask_reg, zmask_reg);
+      break;
+
+   case PIPE_FUNC_LEQUAL:
+      /* zmask = (ifragZ > ref) */
+      spe_clgt(f, zmask_reg, ifragZ_reg, ifbZ_reg);
+      /* mask = (mask & ~zmask) */
+      spe_andc(f, mask_reg, mask_reg, zmask_reg);
+      break;
+
+   case PIPE_FUNC_GEQUAL:
+      /* zmask = (ref > ifragZ) */
+      spe_clgt(f, zmask_reg, ifbZ_reg, ifragZ_reg);
+      /* mask = (mask & ~zmask) */
+      spe_andc(f, mask_reg, mask_reg, zmask_reg);
+      break;
+
+   case PIPE_FUNC_NEVER:
+      spe_il(f, mask_reg, 0);  /* mask = {0,0,0,0} */
+      spe_move(f, zmask_reg, mask_reg);  /* zmask = mask */
+      break;
+
+   case PIPE_FUNC_ALWAYS:
+      /* mask unchanged */
+      spe_il(f, zmask_reg, ~0);  /* zmask = {~0,~0,~0,~0} */
+      break;
+
+   default:
+      ASSERT(0);
+      break;
+   }
+
+   if (dsa->depth.writemask) {
+      /*
+       * If (ztest passed) {
+       *    framebufferZ = fragmentZ;
+       * }
+       * OR,
+       * framebufferZ = (ztest_passed ? fragmentZ : framebufferZ;
+       */
+      spe_selb(f, ifbZ_reg, ifbZ_reg, ifragZ_reg, mask_reg);
+      return TRUE;
+   }
+
+   return FALSE;
+}
+
+
+/**
+ * Generate SPE code to perform alpha testing.
+ *
+ * \param dsa        Gallium depth/stencil/alpha state to gen code for
+ * \param f          SPE function to append instruction onto.
+ * \param mask_reg   register containing quad/pixel "alive" mask (in/out)
+ * \param fragA_reg  register containing four fragment alpha values (in)
+ */
+static void
+gen_alpha_test(const struct pipe_depth_stencil_alpha_state *dsa,
+               struct spe_function *f, int mask_reg, int fragA_reg)
+{
+   int ref_reg = spe_allocate_available_register(f);
+   int amask_reg = spe_allocate_available_register(f);
+
+   ASSERT(dsa->alpha.enabled);
+
+   if ((dsa->alpha.func != PIPE_FUNC_NEVER) &&
+       (dsa->alpha.func != PIPE_FUNC_ALWAYS)) {
+      /* load/splat the alpha reference float value */
+      spe_load_float(f, ref_reg, dsa->alpha.ref_value);
+   }
+
+   /* emit code to do the alpha comparison, updating 'mask' */
+   switch (dsa->alpha.func) {
+   case PIPE_FUNC_EQUAL:
+      /* amask = (fragA == ref) */
+      spe_fceq(f, amask_reg, fragA_reg, ref_reg);
+      /* mask = (mask & amask) */
+      spe_and(f, mask_reg, mask_reg, amask_reg);
+      break;
+
+   case PIPE_FUNC_NOTEQUAL:
+      /* amask = (fragA == ref) */
+      spe_fceq(f, amask_reg, fragA_reg, ref_reg);
+      /* mask = (mask & ~amask) */
+      spe_andc(f, mask_reg, mask_reg, amask_reg);
+      break;
+
+   case PIPE_FUNC_GREATER:
+      /* amask = (fragA > ref) */
+      spe_fcgt(f, amask_reg, fragA_reg, ref_reg);
+      /* mask = (mask & amask) */
+      spe_and(f, mask_reg, mask_reg, amask_reg);
+      break;
+
+   case PIPE_FUNC_LESS:
+      /* amask = (ref > fragA) */
+      spe_fcgt(f, amask_reg, ref_reg, fragA_reg);
+      /* mask = (mask & amask) */
+      spe_and(f, mask_reg, mask_reg, amask_reg);
+      break;
+
+   case PIPE_FUNC_LEQUAL:
+      /* amask = (fragA > ref) */
+      spe_fcgt(f, amask_reg, fragA_reg, ref_reg);
+      /* mask = (mask & ~amask) */
+      spe_andc(f, mask_reg, mask_reg, amask_reg);
+      break;
+
+   case PIPE_FUNC_GEQUAL:
+      /* amask = (ref > fragA) */
+      spe_fcgt(f, amask_reg, ref_reg, fragA_reg);
+      /* mask = (mask & ~amask) */
+      spe_andc(f, mask_reg, mask_reg, amask_reg);
+      break;
+
+   case PIPE_FUNC_NEVER:
+      spe_il(f, mask_reg, 0);  /* mask = [0,0,0,0] */
+      break;
+
+   case PIPE_FUNC_ALWAYS:
+      /* no-op, mask unchanged */
+      break;
+
+   default:
+      ASSERT(0);
+      break;
+   }
+
+#if OPTIMIZATIONS
+   /* if mask == {0,0,0,0} we're all done, return */
+   {
+      /* re-use amask reg here */
+      int tmp_reg = amask_reg;
+      /* tmp[0] = (mask[0] | mask[1] | mask[2] | mask[3]) */
+      spe_orx(f, tmp_reg, mask_reg);
+      /* if tmp[0] == 0 then return from function call */
+      spe_biz(f, tmp_reg, SPE_REG_RA, 0, 0);
+   }
+#endif
+
+   spe_release_register(f, ref_reg);
+   spe_release_register(f, amask_reg);
+}
+
+
+/**
+ * This pair of functions is used inline to allocate and deallocate
+ * optional constant registers.  Once a constant is discovered to be 
+ * needed, we will likely need it again, so we don't want to deallocate
+ * it and have to allocate and load it again unnecessarily.
+ */
+static INLINE void
+setup_optional_register(struct spe_function *f,
+                        int *r)
+{
+   if (*r < 0)
+      *r = spe_allocate_available_register(f);
+}
+
+static INLINE void
+release_optional_register(struct spe_function *f,
+                          int r)
+{
+   if (r >= 0)
+      spe_release_register(f, r);
+}
+
+static INLINE void
+setup_const_register(struct spe_function *f,
+                     int *r,
+                     float value)
+{
+   if (*r >= 0)
+      return;
+   setup_optional_register(f, r);
+   spe_load_float(f, *r, value);
+}
+
+static INLINE void
+release_const_register(struct spe_function *f,
+                       int r)
+{
+   release_optional_register(f, r);
+}
+
+
+
+/**
+ * Unpack/convert framebuffer colors from four 32-bit packed colors
+ * (fbRGBA) to four float RGBA vectors (fbR, fbG, fbB, fbA).
+ * Each 8-bit color component is expanded into a float in [0.0, 1.0].
+ */
+static void
+unpack_colors(struct spe_function *f,
+              enum pipe_format color_format,
+              int fbRGBA_reg,
+              int fbR_reg, int fbG_reg, int fbB_reg, int fbA_reg)
+{
+   int mask0_reg = spe_allocate_available_register(f);
+   int mask1_reg = spe_allocate_available_register(f);
+   int mask2_reg = spe_allocate_available_register(f);
+   int mask3_reg = spe_allocate_available_register(f);
+
+   spe_load_int(f, mask0_reg, 0xff);
+   spe_load_int(f, mask1_reg, 0xff00);
+   spe_load_int(f, mask2_reg, 0xff0000);
+   spe_load_int(f, mask3_reg, 0xff000000);
+
+   spe_comment(f, 0, "Unpack framebuffer colors, convert to floats");
+
+   switch (color_format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      /* fbB = fbRGBA & mask */
+      spe_and(f, fbB_reg, fbRGBA_reg, mask0_reg);
+
+      /* fbG = fbRGBA & mask */
+      spe_and(f, fbG_reg, fbRGBA_reg, mask1_reg);
+
+      /* fbR = fbRGBA & mask */
+      spe_and(f, fbR_reg, fbRGBA_reg, mask2_reg);
+
+      /* fbA = fbRGBA & mask */
+      spe_and(f, fbA_reg, fbRGBA_reg, mask3_reg);
+
+      /* fbG = fbG >> 8 */
+      spe_roti(f, fbG_reg, fbG_reg, -8);
+
+      /* fbR = fbR >> 16 */
+      spe_roti(f, fbR_reg, fbR_reg, -16);
+
+      /* fbA = fbA >> 24 */
+      spe_roti(f, fbA_reg, fbA_reg, -24);
+      break;
+
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      /* fbA = fbRGBA & mask */
+      spe_and(f, fbA_reg, fbRGBA_reg, mask0_reg);
+
+      /* fbR = fbRGBA & mask */
+      spe_and(f, fbR_reg, fbRGBA_reg, mask1_reg);
+
+      /* fbG = fbRGBA & mask */
+      spe_and(f, fbG_reg, fbRGBA_reg, mask2_reg);
+
+      /* fbB = fbRGBA & mask */
+      spe_and(f, fbB_reg, fbRGBA_reg, mask3_reg);
+
+      /* fbR = fbR >> 8 */
+      spe_roti(f, fbR_reg, fbR_reg, -8);
+
+      /* fbG = fbG >> 16 */
+      spe_roti(f, fbG_reg, fbG_reg, -16);
+
+      /* fbB = fbB >> 24 */
+      spe_roti(f, fbB_reg, fbB_reg, -24);
+      break;
+
+   default:
+      ASSERT(0);
+   }
+
+   /* convert int[4] in [0,255] to float[4] in [0.0, 1.0] */
+   spe_cuflt(f, fbR_reg, fbR_reg, 8);
+   spe_cuflt(f, fbG_reg, fbG_reg, 8);
+   spe_cuflt(f, fbB_reg, fbB_reg, 8);
+   spe_cuflt(f, fbA_reg, fbA_reg, 8);
+
+   spe_release_register(f, mask0_reg);
+   spe_release_register(f, mask1_reg);
+   spe_release_register(f, mask2_reg);
+   spe_release_register(f, mask3_reg);
+}
+
+
+/**
+ * Generate SPE code to implement the given blend mode for a quad of pixels.
+ * \param f          SPE function to append instruction onto.
+ * \param fragR_reg  register with fragment red values (float) (in/out)
+ * \param fragG_reg  register with fragment green values (float) (in/out)
+ * \param fragB_reg  register with fragment blue values (float) (in/out)
+ * \param fragA_reg  register with fragment alpha values (float) (in/out)
+ * \param fbRGBA_reg register with packed framebuffer colors (integer) (in)
+ */
+static void
+gen_blend(const struct pipe_blend_state *blend,
+          const struct pipe_blend_color *blend_color,
+          struct spe_function *f,
+          enum pipe_format color_format,
+          int fragR_reg, int fragG_reg, int fragB_reg, int fragA_reg,
+          int fbRGBA_reg)
+{
+   int term1R_reg = spe_allocate_available_register(f);
+   int term1G_reg = spe_allocate_available_register(f);
+   int term1B_reg = spe_allocate_available_register(f);
+   int term1A_reg = spe_allocate_available_register(f);
+
+   int term2R_reg = spe_allocate_available_register(f);
+   int term2G_reg = spe_allocate_available_register(f);
+   int term2B_reg = spe_allocate_available_register(f);
+   int term2A_reg = spe_allocate_available_register(f);
+
+   int fbR_reg = spe_allocate_available_register(f);
+   int fbG_reg = spe_allocate_available_register(f);
+   int fbB_reg = spe_allocate_available_register(f);
+   int fbA_reg = spe_allocate_available_register(f);
+
+   int tmp_reg = spe_allocate_available_register(f);
+
+   /* Optional constant registers we might or might not end up using;
+    * if we do use them, make sure we only allocate them once by
+    * keeping a flag on each one.
+    */
+   int one_reg = -1;
+   int constR_reg = -1, constG_reg = -1, constB_reg = -1, constA_reg = -1;
+
+   ASSERT(blend->blend_enable);
+
+   /* packed RGBA -> float colors */
+   unpack_colors(f, color_format, fbRGBA_reg,
+                 fbR_reg, fbG_reg, fbB_reg, fbA_reg);
+
+   /*
+    * Compute Src RGB terms.  We're actually looking for the value
+    * of (the appropriate RGB factors) * (the incoming source RGB color),
+    * because in some cases (like PIPE_BLENDFACTOR_ONE and 
+    * PIPE_BLENDFACTOR_ZERO) we can avoid doing unnecessary math.
+    */
+   switch (blend->rgb_src_factor) {
+   case PIPE_BLENDFACTOR_ONE:
+      /* factors = (1,1,1), so term = (R,G,B) */
+      spe_move(f, term1R_reg, fragR_reg);
+      spe_move(f, term1G_reg, fragG_reg);
+      spe_move(f, term1B_reg, fragB_reg);
+      break;
+   case PIPE_BLENDFACTOR_ZERO:
+      /* factors = (0,0,0), so term = (0,0,0) */
+      spe_load_float(f, term1R_reg, 0.0f);
+      spe_load_float(f, term1G_reg, 0.0f);
+      spe_load_float(f, term1B_reg, 0.0f);
+      break;
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+      /* factors = (R,G,B), so term = (R*R, G*G, B*B) */
+      spe_fm(f, term1R_reg, fragR_reg, fragR_reg);
+      spe_fm(f, term1G_reg, fragG_reg, fragG_reg);
+      spe_fm(f, term1B_reg, fragB_reg, fragB_reg);
+      break;
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      /* factors = (A,A,A), so term = (R*A, G*A, B*A) */
+      spe_fm(f, term1R_reg, fragR_reg, fragA_reg);
+      spe_fm(f, term1G_reg, fragG_reg, fragA_reg);
+      spe_fm(f, term1B_reg, fragB_reg, fragA_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+      /* factors = (1-R,1-G,1-B), so term = (R*(1-R), G*(1-G), B*(1-B)) 
+       * or in other words term = (R-R*R, G-G*G, B-B*B)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term1R_reg, fragR_reg, fragR_reg, fragR_reg);
+      spe_fnms(f, term1G_reg, fragG_reg, fragG_reg, fragG_reg);
+      spe_fnms(f, term1B_reg, fragB_reg, fragB_reg, fragB_reg);
+      break;
+   case PIPE_BLENDFACTOR_DST_COLOR:
+      /* factors = (Rfb,Gfb,Bfb), so term = (R*Rfb, G*Gfb, B*Bfb) */
+      spe_fm(f, term1R_reg, fragR_reg, fbR_reg);
+      spe_fm(f, term1G_reg, fragG_reg, fbG_reg);
+      spe_fm(f, term1B_reg, fragB_reg, fbB_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+      /* factors = (1-Rfb,1-Gfb,1-Bfb), so term = (R*(1-Rfb),G*(1-Gfb),B*(1-Bfb))
+       * or term = (R-R*Rfb, G-G*Gfb, B-B*Bfb)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term1R_reg, fragR_reg, fbR_reg, fragR_reg);
+      spe_fnms(f, term1G_reg, fragG_reg, fbG_reg, fragG_reg);
+      spe_fnms(f, term1B_reg, fragB_reg, fbB_reg, fragB_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+      /* factors = (1-A,1-A,1-A), so term = (R*(1-A),G*(1-A),B*(1-A))
+       * or term = (R-R*A,G-G*A,B-B*A)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term1R_reg, fragR_reg, fragA_reg, fragR_reg);
+      spe_fnms(f, term1G_reg, fragG_reg, fragA_reg, fragG_reg);
+      spe_fnms(f, term1B_reg, fragB_reg, fragA_reg, fragB_reg);
+      break;
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      /* factors = (Afb, Afb, Afb), so term = (R*Afb, G*Afb, B*Afb) */
+      spe_fm(f, term1R_reg, fragR_reg, fbA_reg);
+      spe_fm(f, term1G_reg, fragG_reg, fbA_reg);
+      spe_fm(f, term1B_reg, fragB_reg, fbA_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      /* factors = (1-Afb, 1-Afb, 1-Afb), so term = (R*(1-Afb),G*(1-Afb),B*(1-Afb)) 
+       * or term = (R-R*Afb,G-G*Afb,b-B*Afb)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term1R_reg, fragR_reg, fbA_reg, fragR_reg);
+      spe_fnms(f, term1G_reg, fragG_reg, fbA_reg, fragG_reg);
+      spe_fnms(f, term1B_reg, fragB_reg, fbA_reg, fragB_reg);
+      break;
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+      /* We need the optional constant color registers */
+      setup_const_register(f, &constR_reg, blend_color->color[0]);
+      setup_const_register(f, &constG_reg, blend_color->color[1]);
+      setup_const_register(f, &constB_reg, blend_color->color[2]);
+      /* now, factor = (Rc,Gc,Bc), so term = (R*Rc,G*Gc,B*Bc) */
+      spe_fm(f, term1R_reg, fragR_reg, constR_reg);
+      spe_fm(f, term1G_reg, fragG_reg, constG_reg);
+      spe_fm(f, term1B_reg, fragB_reg, constB_reg);
+      break;
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      /* we'll need the optional constant alpha register */
+      setup_const_register(f, &constA_reg, blend_color->color[3]);
+      /* factor = (Ac,Ac,Ac), so term = (R*Ac,G*Ac,B*Ac) */
+      spe_fm(f, term1R_reg, fragR_reg, constA_reg);
+      spe_fm(f, term1G_reg, fragG_reg, constA_reg);
+      spe_fm(f, term1B_reg, fragB_reg, constA_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+      /* We need the optional constant color registers */
+      setup_const_register(f, &constR_reg, blend_color->color[0]);
+      setup_const_register(f, &constG_reg, blend_color->color[1]);
+      setup_const_register(f, &constB_reg, blend_color->color[2]);
+      /* factor = (1-Rc,1-Gc,1-Bc), so term = (R*(1-Rc),G*(1-Gc),B*(1-Bc)) 
+       * or term = (R-R*Rc, G-G*Gc, B-B*Bc)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term1R_reg, fragR_reg, constR_reg, fragR_reg);
+      spe_fnms(f, term1G_reg, fragG_reg, constG_reg, fragG_reg);
+      spe_fnms(f, term1B_reg, fragB_reg, constB_reg, fragB_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+      /* We need the optional constant color registers */
+      setup_const_register(f, &constR_reg, blend_color->color[0]);
+      setup_const_register(f, &constG_reg, blend_color->color[1]);
+      setup_const_register(f, &constB_reg, blend_color->color[2]);
+      /* factor = (1-Ac,1-Ac,1-Ac), so term = (R*(1-Ac),G*(1-Ac),B*(1-Ac))
+       * or term = (R-R*Ac,G-G*Ac,B-B*Ac)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term1R_reg, fragR_reg, constA_reg, fragR_reg);
+      spe_fnms(f, term1G_reg, fragG_reg, constA_reg, fragG_reg);
+      spe_fnms(f, term1B_reg, fragB_reg, constA_reg, fragB_reg);
+      break;
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+      /* We'll need the optional {1,1,1,1} register */
+      setup_const_register(f, &one_reg, 1.0f);
+      /* factor = (min(A,1-Afb),min(A,1-Afb),min(A,1-Afb)), so 
+       * term = (R*min(A,1-Afb), G*min(A,1-Afb), B*min(A,1-Afb))
+       * We could expand the term (as a*min(b,c) == min(a*b,a*c)
+       * as long as a is positive), but then we'd have to do three
+       * spe_float_min() functions instead of one, so this is simpler.
+       */
+      /* tmp = 1 - Afb */
+      spe_fs(f, tmp_reg, one_reg, fbA_reg);
+      /* tmp = min(A,tmp) */
+      spe_float_min(f, tmp_reg, fragA_reg, tmp_reg);
+      /* term = R*tmp */
+      spe_fm(f, term1R_reg, fragR_reg, tmp_reg);
+      spe_fm(f, term1G_reg, fragG_reg, tmp_reg);
+      spe_fm(f, term1B_reg, fragB_reg, tmp_reg);
+      break;
+
+      /* These are special D3D cases involving a second color output
+       * from the fragment shader.  I'm not sure we can support them
+       * yet... XXX
+       */
+   case PIPE_BLENDFACTOR_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:
+   case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+
+   default:
+      ASSERT(0);
+   }
+
+   /*
+    * Compute Src Alpha term.  Like the above, we're looking for
+    * the full term A*factor, not just the factor itself, because
+    * in many cases we can avoid doing unnecessary multiplies.
+    */
+   switch (blend->alpha_src_factor) {
+   case PIPE_BLENDFACTOR_ZERO:
+      /* factor = 0, so term = 0 */
+      spe_load_float(f, term1A_reg, 0.0f);
+      break;
+
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: /* fall through */
+   case PIPE_BLENDFACTOR_ONE:
+      /* factor = 1, so term = A */
+      spe_move(f, term1A_reg, fragA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+      /* factor = A, so term = A*A */
+      spe_fm(f, term1A_reg, fragA_reg, fragA_reg);
+      break;
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      spe_fm(f, term1A_reg, fragA_reg, fragA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+      /* factor = 1-A, so term = A*(1-A) = A-A*A */
+      /* fnms(a,b,c,d) computes a = d - b*c */
+      spe_fnms(f, term1A_reg, fragA_reg, fragA_reg, fragA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_DST_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_DST_COLOR:
+      /* factor = Afb, so term = A*Afb */
+      spe_fm(f, term1A_reg, fragA_reg, fbA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+      /* factor = 1-Afb, so term = A*(1-Afb) = A - A*Afb */
+      /* fnms(a,b,c,d) computes a = d - b*c */
+      spe_fnms(f, term1A_reg, fragA_reg, fbA_reg, fragA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_CONST_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+      /* We need the optional constA_reg register */
+      setup_const_register(f, &constA_reg, blend_color->color[3]);
+      /* factor = Ac, so term = A*Ac */
+      spe_fm(f, term1A_reg, fragA_reg, constA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+      /* We need the optional constA_reg register */
+      setup_const_register(f, &constA_reg, blend_color->color[3]);
+      /* factor = 1-Ac, so term = A*(1-Ac) = A-A*Ac */
+      /* fnms(a,b,c,d) computes a = d - b*c */
+      spe_fnms(f, term1A_reg, fragA_reg, constA_reg, fragA_reg);
+      break;
+
+      /* These are special D3D cases involving a second color output
+       * from the fragment shader.  I'm not sure we can support them
+       * yet... XXX
+       */
+   case PIPE_BLENDFACTOR_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:
+   case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+   default:
+      ASSERT(0);
+   }
+
+   /*
+    * Compute Dest RGB term.  Like the above, we're looking for
+    * the full term (Rfb,Gfb,Bfb)*(factor), not just the factor itself, because
+    * in many cases we can avoid doing unnecessary multiplies.
+    */
+   switch (blend->rgb_dst_factor) {
+   case PIPE_BLENDFACTOR_ONE:
+      /* factors = (1,1,1), so term = (Rfb,Gfb,Bfb) */
+      spe_move(f, term2R_reg, fbR_reg);
+      spe_move(f, term2G_reg, fbG_reg);
+      spe_move(f, term2B_reg, fbB_reg);
+      break;
+   case PIPE_BLENDFACTOR_ZERO:
+      /* factor s= (0,0,0), so term = (0,0,0) */
+      spe_load_float(f, term2R_reg, 0.0f);
+      spe_load_float(f, term2G_reg, 0.0f);
+      spe_load_float(f, term2B_reg, 0.0f);
+      break;
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+      /* factors = (R,G,B), so term = (R*Rfb, G*Gfb, B*Bfb) */
+      spe_fm(f, term2R_reg, fbR_reg, fragR_reg);
+      spe_fm(f, term2G_reg, fbG_reg, fragG_reg);
+      spe_fm(f, term2B_reg, fbB_reg, fragB_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+      /* factors = (1-R,1-G,1-B), so term = (Rfb*(1-R), Gfb*(1-G), Bfb*(1-B)) 
+       * or in other words term = (Rfb-Rfb*R, Gfb-Gfb*G, Bfb-Bfb*B)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term2R_reg, fragR_reg, fbR_reg, fbR_reg);
+      spe_fnms(f, term2G_reg, fragG_reg, fbG_reg, fbG_reg);
+      spe_fnms(f, term2B_reg, fragB_reg, fbB_reg, fbB_reg);
+      break;
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      /* factors = (A,A,A), so term = (Rfb*A, Gfb*A, Bfb*A) */
+      spe_fm(f, term2R_reg, fbR_reg, fragA_reg);
+      spe_fm(f, term2G_reg, fbG_reg, fragA_reg);
+      spe_fm(f, term2B_reg, fbB_reg, fragA_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+      /* factors = (1-A,1-A,1-A) so term = (Rfb-Rfb*A,Gfb-Gfb*A,Bfb-Bfb*A) */
+      /* fnms(a,b,c,d) computes a = d - b*c */
+      spe_fnms(f, term2R_reg, fbR_reg, fragA_reg, fbR_reg);
+      spe_fnms(f, term2G_reg, fbG_reg, fragA_reg, fbG_reg);
+      spe_fnms(f, term2B_reg, fbB_reg, fragA_reg, fbB_reg);
+      break;
+   case PIPE_BLENDFACTOR_DST_COLOR:
+      /* factors = (Rfb,Gfb,Bfb), so term = (Rfb*Rfb, Gfb*Gfb, Bfb*Bfb) */
+      spe_fm(f, term2R_reg, fbR_reg, fbR_reg);
+      spe_fm(f, term2G_reg, fbG_reg, fbG_reg);
+      spe_fm(f, term2B_reg, fbB_reg, fbB_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+      /* factors = (1-Rfb,1-Gfb,1-Bfb), so term = (Rfb*(1-Rfb),Gfb*(1-Gfb),Bfb*(1-Bfb))
+       * or term = (Rfb-Rfb*Rfb, Gfb-Gfb*Gfb, Bfb-Bfb*Bfb)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term2R_reg, fbR_reg, fbR_reg, fbR_reg);
+      spe_fnms(f, term2G_reg, fbG_reg, fbG_reg, fbG_reg);
+      spe_fnms(f, term2B_reg, fbB_reg, fbB_reg, fbB_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      /* factors = (Afb, Afb, Afb), so term = (Rfb*Afb, Gfb*Afb, Bfb*Afb) */
+      spe_fm(f, term2R_reg, fbR_reg, fbA_reg);
+      spe_fm(f, term2G_reg, fbG_reg, fbA_reg);
+      spe_fm(f, term2B_reg, fbB_reg, fbA_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      /* factors = (1-Afb, 1-Afb, 1-Afb), so term = (Rfb*(1-Afb),Gfb*(1-Afb),Bfb*(1-Afb)) 
+       * or term = (Rfb-Rfb*Afb,Gfb-Gfb*Afb,Bfb-Bfb*Afb)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term2R_reg, fbR_reg, fbA_reg, fbR_reg);
+      spe_fnms(f, term2G_reg, fbG_reg, fbA_reg, fbG_reg);
+      spe_fnms(f, term2B_reg, fbB_reg, fbA_reg, fbB_reg);
+      break;
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+      /* We need the optional constant color registers */
+      setup_const_register(f, &constR_reg, blend_color->color[0]);
+      setup_const_register(f, &constG_reg, blend_color->color[1]);
+      setup_const_register(f, &constB_reg, blend_color->color[2]);
+      /* now, factor = (Rc,Gc,Bc), so term = (Rfb*Rc,Gfb*Gc,Bfb*Bc) */
+      spe_fm(f, term2R_reg, fbR_reg, constR_reg);
+      spe_fm(f, term2G_reg, fbG_reg, constG_reg);
+      spe_fm(f, term2B_reg, fbB_reg, constB_reg);
+      break;
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      /* we'll need the optional constant alpha register */
+      setup_const_register(f, &constA_reg, blend_color->color[3]);
+      /* factor = (Ac,Ac,Ac), so term = (Rfb*Ac,Gfb*Ac,Bfb*Ac) */
+      spe_fm(f, term2R_reg, fbR_reg, constA_reg);
+      spe_fm(f, term2G_reg, fbG_reg, constA_reg);
+      spe_fm(f, term2B_reg, fbB_reg, constA_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+      /* We need the optional constant color registers */
+      setup_const_register(f, &constR_reg, blend_color->color[0]);
+      setup_const_register(f, &constG_reg, blend_color->color[1]);
+      setup_const_register(f, &constB_reg, blend_color->color[2]);
+      /* factor = (1-Rc,1-Gc,1-Bc), so term = (Rfb*(1-Rc),Gfb*(1-Gc),Bfb*(1-Bc)) 
+       * or term = (Rfb-Rfb*Rc, Gfb-Gfb*Gc, Bfb-Bfb*Bc)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term2R_reg, fbR_reg, constR_reg, fbR_reg);
+      spe_fnms(f, term2G_reg, fbG_reg, constG_reg, fbG_reg);
+      spe_fnms(f, term2B_reg, fbB_reg, constB_reg, fbB_reg);
+      break;
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+      /* We need the optional constant color registers */
+      setup_const_register(f, &constR_reg, blend_color->color[0]);
+      setup_const_register(f, &constG_reg, blend_color->color[1]);
+      setup_const_register(f, &constB_reg, blend_color->color[2]);
+      /* factor = (1-Ac,1-Ac,1-Ac), so term = (Rfb*(1-Ac),Gfb*(1-Ac),Bfb*(1-Ac))
+       * or term = (Rfb-Rfb*Ac,Gfb-Gfb*Ac,Bfb-Bfb*Ac)
+       * fnms(a,b,c,d) computes a = d - b*c
+       */
+      spe_fnms(f, term2R_reg, fbR_reg, constA_reg, fbR_reg);
+      spe_fnms(f, term2G_reg, fbG_reg, constA_reg, fbG_reg);
+      spe_fnms(f, term2B_reg, fbB_reg, constA_reg, fbB_reg);
+      break;
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: /* not supported for dest RGB */
+      ASSERT(0);
+      break;
+
+      /* These are special D3D cases involving a second color output
+       * from the fragment shader.  I'm not sure we can support them
+       * yet... XXX
+       */
+   case PIPE_BLENDFACTOR_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:
+   case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+
+   default:
+      ASSERT(0);
+   }
+
+   /*
+    * Compute Dest Alpha term.  Like the above, we're looking for
+    * the full term Afb*factor, not just the factor itself, because
+    * in many cases we can avoid doing unnecessary multiplies.
+    */
+   switch (blend->alpha_dst_factor) {
+   case PIPE_BLENDFACTOR_ONE:
+      /* factor = 1, so term = Afb */
+      spe_move(f, term2A_reg, fbA_reg);
+      break;
+   case PIPE_BLENDFACTOR_ZERO:
+      /* factor = 0, so term = 0 */
+      spe_load_float(f, term2A_reg, 0.0f);
+      break;
+
+   case PIPE_BLENDFACTOR_SRC_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+      /* factor = A, so term = Afb*A */
+      spe_fm(f, term2A_reg, fbA_reg, fragA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+      /* factor = 1-A, so term = Afb*(1-A) = Afb-Afb*A */
+      /* fnms(a,b,c,d) computes a = d - b*c */
+      spe_fnms(f, term2A_reg, fbA_reg, fragA_reg, fbA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_DST_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_DST_COLOR:
+      /* factor = Afb, so term = Afb*Afb */
+      spe_fm(f, term2A_reg, fbA_reg, fbA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+      /* factor = 1-Afb, so term = Afb*(1-Afb) = Afb - Afb*Afb */
+      /* fnms(a,b,c,d) computes a = d - b*c */
+      spe_fnms(f, term2A_reg, fbA_reg, fbA_reg, fbA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_CONST_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+      /* We need the optional constA_reg register */
+      setup_const_register(f, &constA_reg, blend_color->color[3]);
+      /* factor = Ac, so term = Afb*Ac */
+      spe_fm(f, term2A_reg, fbA_reg, constA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA: /* fall through */
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+      /* We need the optional constA_reg register */
+      setup_const_register(f, &constA_reg, blend_color->color[3]);
+      /* factor = 1-Ac, so term = Afb*(1-Ac) = Afb-Afb*Ac */
+      /* fnms(a,b,c,d) computes a = d - b*c */
+      spe_fnms(f, term2A_reg, fbA_reg, constA_reg, fbA_reg);
+      break;
+
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: /* not supported for dest alpha */
+      ASSERT(0);
+      break;
+
+      /* These are special D3D cases involving a second color output
+       * from the fragment shader.  I'm not sure we can support them
+       * yet... XXX
+       */
+   case PIPE_BLENDFACTOR_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:
+   case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+   default:
+      ASSERT(0);
+   }
+
+   /*
+    * Combine Src/Dest RGB terms as per the blend equation.
+    */
+   switch (blend->rgb_func) {
+   case PIPE_BLEND_ADD:
+      spe_fa(f, fragR_reg, term1R_reg, term2R_reg);
+      spe_fa(f, fragG_reg, term1G_reg, term2G_reg);
+      spe_fa(f, fragB_reg, term1B_reg, term2B_reg);
+      break;
+   case PIPE_BLEND_SUBTRACT:
+      spe_fs(f, fragR_reg, term1R_reg, term2R_reg);
+      spe_fs(f, fragG_reg, term1G_reg, term2G_reg);
+      spe_fs(f, fragB_reg, term1B_reg, term2B_reg);
+      break;
+   case PIPE_BLEND_REVERSE_SUBTRACT:
+      spe_fs(f, fragR_reg, term2R_reg, term1R_reg);
+      spe_fs(f, fragG_reg, term2G_reg, term1G_reg);
+      spe_fs(f, fragB_reg, term2B_reg, term1B_reg);
+      break;
+   case PIPE_BLEND_MIN:
+      spe_float_min(f, fragR_reg, term1R_reg, term2R_reg);
+      spe_float_min(f, fragG_reg, term1G_reg, term2G_reg);
+      spe_float_min(f, fragB_reg, term1B_reg, term2B_reg);
+      break;
+   case PIPE_BLEND_MAX:
+      spe_float_max(f, fragR_reg, term1R_reg, term2R_reg);
+      spe_float_max(f, fragG_reg, term1G_reg, term2G_reg);
+      spe_float_max(f, fragB_reg, term1B_reg, term2B_reg);
+      break;
+   default:
+      ASSERT(0);
+   }
+
+   /*
+    * Combine Src/Dest A term
+    */
+   switch (blend->alpha_func) {
+   case PIPE_BLEND_ADD:
+      spe_fa(f, fragA_reg, term1A_reg, term2A_reg);
+      break;
+   case PIPE_BLEND_SUBTRACT:
+      spe_fs(f, fragA_reg, term1A_reg, term2A_reg);
+      break;
+   case PIPE_BLEND_REVERSE_SUBTRACT:
+      spe_fs(f, fragA_reg, term2A_reg, term1A_reg);
+      break;
+   case PIPE_BLEND_MIN:
+      spe_float_min(f, fragA_reg, term1A_reg, term2A_reg);
+      break;
+   case PIPE_BLEND_MAX:
+      spe_float_max(f, fragA_reg, term1A_reg, term2A_reg);
+      break;
+   default:
+      ASSERT(0);
+   }
+
+   spe_release_register(f, term1R_reg);
+   spe_release_register(f, term1G_reg);
+   spe_release_register(f, term1B_reg);
+   spe_release_register(f, term1A_reg);
+
+   spe_release_register(f, term2R_reg);
+   spe_release_register(f, term2G_reg);
+   spe_release_register(f, term2B_reg);
+   spe_release_register(f, term2A_reg);
+
+   spe_release_register(f, fbR_reg);
+   spe_release_register(f, fbG_reg);
+   spe_release_register(f, fbB_reg);
+   spe_release_register(f, fbA_reg);
+
+   spe_release_register(f, tmp_reg);
+
+   /* Free any optional registers that actually got used */
+   release_const_register(f, one_reg);
+   release_const_register(f, constR_reg);
+   release_const_register(f, constG_reg);
+   release_const_register(f, constB_reg);
+   release_const_register(f, constA_reg);
+}
+
+
+static void
+gen_logicop(const struct pipe_blend_state *blend,
+            struct spe_function *f,
+            int fragRGBA_reg, int fbRGBA_reg)
+{
+   /* We've got four 32-bit RGBA packed pixels in each of
+    * fragRGBA_reg and fbRGBA_reg, not sets of floating-point
+    * reds, greens, blues, and alphas.
+    * */
+   ASSERT(blend->logicop_enable);
+
+   switch(blend->logicop_func) {
+      case PIPE_LOGICOP_CLEAR: /* 0 */
+         spe_zero(f, fragRGBA_reg);
+         break;
+      case PIPE_LOGICOP_NOR: /* ~(s | d) */
+         spe_nor(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg);
+         break;
+      case PIPE_LOGICOP_AND_INVERTED: /* ~s & d */
+         /* andc R, A, B computes R = A & ~B */
+         spe_andc(f, fragRGBA_reg, fbRGBA_reg, fragRGBA_reg);
+         break;
+      case PIPE_LOGICOP_COPY_INVERTED: /* ~s */
+         spe_complement(f, fragRGBA_reg, fragRGBA_reg);
+         break;
+      case PIPE_LOGICOP_AND_REVERSE: /* s & ~d */
+         /* andc R, A, B computes R = A & ~B */
+         spe_andc(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg);
+         break;
+      case PIPE_LOGICOP_INVERT: /* ~d */
+         /* Note that (A nor A) == ~(A|A) == ~A */
+         spe_nor(f, fragRGBA_reg, fbRGBA_reg, fbRGBA_reg);
+         break;
+      case PIPE_LOGICOP_XOR: /* s ^ d */
+         spe_xor(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg);
+         break;
+      case PIPE_LOGICOP_NAND: /* ~(s & d) */
+         spe_nand(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg);
+         break;
+      case PIPE_LOGICOP_AND: /* s & d */
+         spe_and(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg);
+         break;
+      case PIPE_LOGICOP_EQUIV: /* ~(s ^ d) */
+         spe_xor(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg);
+         spe_complement(f, fragRGBA_reg, fragRGBA_reg);
+         break;
+      case PIPE_LOGICOP_NOOP: /* d */
+         spe_move(f, fragRGBA_reg, fbRGBA_reg);
+         break;
+      case PIPE_LOGICOP_OR_INVERTED: /* ~s | d */
+         /* orc R, A, B computes R = A | ~B */
+         spe_orc(f, fragRGBA_reg, fbRGBA_reg, fragRGBA_reg);
+         break;
+      case PIPE_LOGICOP_COPY: /* s */
+         break;
+      case PIPE_LOGICOP_OR_REVERSE: /* s | ~d */
+         /* orc R, A, B computes R = A | ~B */
+         spe_orc(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg);
+         break;
+      case PIPE_LOGICOP_OR: /* s | d */
+         spe_or(f, fragRGBA_reg, fragRGBA_reg, fbRGBA_reg);
+         break;
+      case PIPE_LOGICOP_SET: /* 1 */
+         spe_load_int(f, fragRGBA_reg, 0xffffffff);
+         break;
+      default:
+         ASSERT(0);
+   }
+}
+
+
+/**
+ * Generate code to pack a quad of float colors into four 32-bit integers.
+ *
+ * \param f             SPE function to append instruction onto.
+ * \param color_format  the dest color packing format
+ * \param r_reg         register containing four red values (in/clobbered)
+ * \param g_reg         register containing four green values (in/clobbered)
+ * \param b_reg         register containing four blue values (in/clobbered)
+ * \param a_reg         register containing four alpha values (in/clobbered)
+ * \param rgba_reg      register to store the packed RGBA colors (out)
+ */
+static void
+gen_pack_colors(struct spe_function *f,
+                enum pipe_format color_format,
+                int r_reg, int g_reg, int b_reg, int a_reg,
+                int rgba_reg)
+{
+   int rg_reg = spe_allocate_available_register(f);
+   int ba_reg = spe_allocate_available_register(f);
+
+   /* Convert float[4] in [0.0,1.0] to int[4] in [0,~0], with clamping */
+   spe_cfltu(f, r_reg, r_reg, 32);
+   spe_cfltu(f, g_reg, g_reg, 32);
+   spe_cfltu(f, b_reg, b_reg, 32);
+   spe_cfltu(f, a_reg, a_reg, 32);
+
+   /* Shift the most significant bytes to the least significant positions.
+    * I.e.: reg = reg >> 24
+    */
+   spe_rotmi(f, r_reg, r_reg, -24);
+   spe_rotmi(f, g_reg, g_reg, -24);
+   spe_rotmi(f, b_reg, b_reg, -24);
+   spe_rotmi(f, a_reg, a_reg, -24);
+
+   /* Shift the color bytes according to the surface format */
+   if (color_format == PIPE_FORMAT_A8R8G8B8_UNORM) {
+      spe_roti(f, g_reg, g_reg, 8);   /* green <<= 8 */
+      spe_roti(f, r_reg, r_reg, 16);  /* red <<= 16 */
+      spe_roti(f, a_reg, a_reg, 24);  /* alpha <<= 24 */
+   }
+   else if (color_format == PIPE_FORMAT_B8G8R8A8_UNORM) {
+      spe_roti(f, r_reg, r_reg, 8);   /* red <<= 8 */
+      spe_roti(f, g_reg, g_reg, 16);  /* green <<= 16 */
+      spe_roti(f, b_reg, b_reg, 24);  /* blue <<= 24 */
+   }
+   else {
+      ASSERT(0);
+   }
+
+   /* Merge red, green, blue, alpha registers to make packed RGBA colors.
+    * Eg: after shifting according to color_format we might have:
+    *     R = {0x00ff0000, 0x00110000, 0x00220000, 0x00330000}
+    *     G = {0x0000ff00, 0x00004400, 0x00005500, 0x00006600}
+    *     B = {0x000000ff, 0x00000077, 0x00000088, 0x00000099}
+    *     A = {0xff000000, 0xaa000000, 0xbb000000, 0xcc000000}
+    * OR-ing all those together gives us four packed colors:
+    *  RGBA = {0xffffffff, 0xaa114477, 0xbb225588, 0xcc336699}
+    */
+   spe_or(f, rg_reg, r_reg, g_reg);
+   spe_or(f, ba_reg, a_reg, b_reg);
+   spe_or(f, rgba_reg, rg_reg, ba_reg);
+
+   spe_release_register(f, rg_reg);
+   spe_release_register(f, ba_reg);
+}
+
+
+static void
+gen_colormask(struct spe_function *f,
+              uint colormask,
+              enum pipe_format color_format,
+              int fragRGBA_reg, int fbRGBA_reg)
+{
+   /* We've got four 32-bit RGBA packed pixels in each of
+    * fragRGBA_reg and fbRGBA_reg, not sets of floating-point
+    * reds, greens, blues, and alphas.  Further, the pixels
+    * are packed according to the given color format, not
+    * necessarily RGBA...
+    */
+   uint r_mask;
+   uint g_mask;
+   uint b_mask;
+   uint a_mask;
+
+   /* Calculate exactly where the bits for any particular color
+    * end up, so we can mask them correctly.
+    */
+   switch(color_format) {
+      case PIPE_FORMAT_A8R8G8B8_UNORM:
+         /* ARGB */
+         a_mask = 0xff000000;
+         r_mask = 0x00ff0000;
+         g_mask = 0x0000ff00;
+         b_mask = 0x000000ff;
+         break;
+      case PIPE_FORMAT_B8G8R8A8_UNORM:
+         /* BGRA */
+         b_mask = 0xff000000;
+         g_mask = 0x00ff0000;
+         r_mask = 0x0000ff00;
+         a_mask = 0x000000ff;
+         break;
+      default:
+         ASSERT(0);
+   }
+
+   /* For each R, G, B, and A component we're supposed to mask out, 
+    * clear its bits.   Then our mask operation later will work 
+    * as expected.
+    */
+   if (!(colormask & PIPE_MASK_R)) {
+      r_mask = 0;
+   }
+   if (!(colormask & PIPE_MASK_G)) {
+      g_mask = 0;
+   }
+   if (!(colormask & PIPE_MASK_B)) {
+      b_mask = 0;
+   }
+   if (!(colormask & PIPE_MASK_A)) {
+      a_mask = 0;
+   }
+
+   /* Get a temporary register to hold the mask that will be applied
+    * to the fragment
+    */
+   int colormask_reg = spe_allocate_available_register(f);
+
+   /* The actual mask we're going to use is an OR of the remaining R, G, B,
+    * and A masks.  Load the result value into our temporary register.
+    */
+   spe_load_uint(f, colormask_reg, r_mask | g_mask | b_mask | a_mask);
+
+   /* Use the mask register to select between the fragment color
+    * values and the frame buffer color values.  Wherever the
+    * mask has a 0 bit, the current frame buffer color should override
+    * the fragment color.  Wherever the mask has a 1 bit, the 
+    * fragment color should persevere.  The Select Bits (selb rt, rA, rB, rM)
+    * instruction will select bits from its first operand rA wherever the
+    * the mask bits rM are 0, and from its second operand rB wherever the
+    * mask bits rM are 1.  That means that the frame buffer color is the
+    * first operand, and the fragment color the second.
+    */
+    spe_selb(f, fragRGBA_reg, fbRGBA_reg, fragRGBA_reg, colormask_reg);
+
+    /* Release the temporary register and we're done */
+    spe_release_register(f, colormask_reg);
+}
+
+
+/**
+ * This function is annoyingly similar to gen_depth_test(), above, except
+ * that instead of comparing two varying values (i.e. fragment and buffer),
+ * we're comparing a varying value with a static value.  As such, we have
+ * access to the Compare Immediate instructions where we don't in 
+ * gen_depth_test(), which is what makes us very different.
+ *
+ * There's some added complexity if there's a non-trivial state->mask
+ * value; then stencil and reference both must be masked
+ *
+ * The return value in the stencil_pass_reg is a bitmask of valid
+ * fragments that also passed the stencil test.  The bitmask of valid
+ * fragments that failed would be found in
+ * (fragment_mask_reg & ~stencil_pass_reg).
+ */
+static void
+gen_stencil_test(struct spe_function *f,
+                 const struct pipe_stencil_state *state, 
+                 uint stencil_max_value,
+                 int fragment_mask_reg,
+                 int fbS_reg, 
+                 int stencil_pass_reg)
+{
+   /* Generate code that puts the set of passing fragments into the
+    * stencil_pass_reg register, taking into account whether each fragment
+    * was active to begin with.
+    */
+   switch (state->func) {
+   case PIPE_FUNC_EQUAL:
+      if (state->valuemask == stencil_max_value) {
+         /* stencil_pass = fragment_mask & (s == reference) */
+         spe_compare_equal_uint(f, stencil_pass_reg, fbS_reg, state->ref_value);
+         spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+      }
+      else {
+         /* stencil_pass = fragment_mask & ((s&mask) == (reference&mask)) */
+         uint tmp_masked_stencil = spe_allocate_available_register(f);
+         spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->valuemask);
+         spe_compare_equal_uint(f, stencil_pass_reg, tmp_masked_stencil,
+                                state->valuemask & state->ref_value);
+         spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+         spe_release_register(f, tmp_masked_stencil);
+      }
+      break;
+
+   case PIPE_FUNC_NOTEQUAL:
+      if (state->valuemask == stencil_max_value) {
+         /* stencil_pass = fragment_mask & ~(s == reference) */
+         spe_compare_equal_uint(f, stencil_pass_reg, fbS_reg, state->ref_value);
+         spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+      }
+      else {
+         /* stencil_pass = fragment_mask & ~((s&mask) == (reference&mask)) */
+         int tmp_masked_stencil = spe_allocate_available_register(f);
+         spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->valuemask);
+         spe_compare_equal_uint(f, stencil_pass_reg, tmp_masked_stencil,
+                                state->valuemask & state->ref_value);
+         spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+         spe_release_register(f, tmp_masked_stencil);
+      }
+      break;
+
+   case PIPE_FUNC_LESS:
+      if (state->valuemask == stencil_max_value) {
+         /* stencil_pass = fragment_mask & (reference < s)  */
+         spe_compare_greater_uint(f, stencil_pass_reg, fbS_reg, state->ref_value);
+         spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+      }
+      else {
+         /* stencil_pass = fragment_mask & ((reference&mask) < (s & mask)) */
+         int tmp_masked_stencil = spe_allocate_available_register(f);
+         spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->valuemask);
+         spe_compare_greater_uint(f, stencil_pass_reg, tmp_masked_stencil,
+                                  state->valuemask & state->ref_value);
+         spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+         spe_release_register(f, tmp_masked_stencil);
+      }
+      break;
+
+   case PIPE_FUNC_GREATER:
+      if (state->valuemask == stencil_max_value) {
+         /* stencil_pass = fragment_mask & (reference > s) */
+         /* There's no convenient Compare Less Than Immediate instruction, so
+          * we'll have to do this one the harder way, by loading a register and 
+          * comparing directly.  Compare Logical Greater Than Word (clgt) 
+          * treats its operands as unsigned - no sign extension.
+          */
+         int tmp_reg = spe_allocate_available_register(f);
+         spe_load_uint(f, tmp_reg, state->ref_value);
+         spe_clgt(f, stencil_pass_reg, tmp_reg, fbS_reg);
+         spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+         spe_release_register(f, tmp_reg);
+      }
+      else {
+         /* stencil_pass = fragment_mask & ((reference&mask) > (s&mask)) */
+         int tmp_reg = spe_allocate_available_register(f);
+         int tmp_masked_stencil = spe_allocate_available_register(f);
+         spe_load_uint(f, tmp_reg, state->valuemask & state->ref_value);
+         spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->valuemask);
+         spe_clgt(f, stencil_pass_reg, tmp_reg, tmp_masked_stencil);
+         spe_and(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+         spe_release_register(f, tmp_reg);
+         spe_release_register(f, tmp_masked_stencil);
+      }
+      break;
+
+   case PIPE_FUNC_GEQUAL:
+      if (state->valuemask == stencil_max_value) {
+         /* stencil_pass = fragment_mask & (reference >= s) 
+          *              = fragment_mask & ~(s > reference) */
+         spe_compare_greater_uint(f, stencil_pass_reg, fbS_reg,
+                                  state->ref_value);
+         spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+      }
+      else {
+         /* stencil_pass = fragment_mask & ~((s&mask) > (reference&mask)) */
+         int tmp_masked_stencil = spe_allocate_available_register(f);
+         spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->valuemask);
+         spe_compare_greater_uint(f, stencil_pass_reg, tmp_masked_stencil,
+                                  state->valuemask & state->ref_value);
+         spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+         spe_release_register(f, tmp_masked_stencil);
+      }
+      break;
+
+   case PIPE_FUNC_LEQUAL:
+      if (state->valuemask == stencil_max_value) {
+         /* stencil_pass = fragment_mask & (reference <= s) ]
+          *               = fragment_mask & ~(reference > s) */
+         /* As above, we have to do this by loading a register */
+         int tmp_reg = spe_allocate_available_register(f);
+         spe_load_uint(f, tmp_reg, state->ref_value);
+         spe_clgt(f, stencil_pass_reg, tmp_reg, fbS_reg);
+         spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+         spe_release_register(f, tmp_reg);
+      }
+      else {
+         /* stencil_pass = fragment_mask & ~((reference&mask) > (s&mask)) */
+         int tmp_reg = spe_allocate_available_register(f);
+         int tmp_masked_stencil = spe_allocate_available_register(f);
+         spe_load_uint(f, tmp_reg, state->ref_value & state->valuemask);
+         spe_and_uint(f, tmp_masked_stencil, fbS_reg, state->valuemask);
+         spe_clgt(f, stencil_pass_reg, tmp_reg, tmp_masked_stencil);
+         spe_andc(f, stencil_pass_reg, fragment_mask_reg, stencil_pass_reg);
+         spe_release_register(f, tmp_reg);
+         spe_release_register(f, tmp_masked_stencil);
+      }
+      break;
+
+   case PIPE_FUNC_NEVER:
+      /* stencil_pass = fragment_mask & 0 = 0 */
+      spe_load_uint(f, stencil_pass_reg, 0);
+      break;
+
+   case PIPE_FUNC_ALWAYS:
+      /* stencil_pass = fragment_mask & 1 = fragment_mask */
+      spe_move(f, stencil_pass_reg, fragment_mask_reg);
+      break;
+   }
+
+   /* The fragments that passed the stencil test are now in stencil_pass_reg.
+    * The fragments that failed would be (fragment_mask_reg & ~stencil_pass_reg).
+    */
+}
+
+
+/**
+ * This function generates code that calculates a set of new stencil values
+ * given the earlier values and the operation to apply.  It does not
+ * apply any tests.  It is intended to be called up to 3 times
+ * (for the stencil fail operation, for the stencil pass-z fail operation,
+ * and for the stencil pass-z pass operation) to collect up to three
+ * possible sets of values, and for the caller to combine them based
+ * on the result of the tests.
+ *
+ * stencil_max_value should be (2^n - 1) where n is the number of bits
+ * in the stencil buffer - in other words, it should be usable as a mask.
+ */
+static void
+gen_stencil_values(struct spe_function *f,
+                   uint stencil_op,
+                   uint stencil_ref_value,
+                   uint stencil_max_value,
+                   int fbS_reg,
+                   int newS_reg)
+{
+   /* The code below assumes that newS_reg and fbS_reg are not the same
+    * register; if they can be, the calculations below will have to use
+    * an additional temporary register.  For now, mark the assumption
+    * with an assertion that will fail if they are the same.
+    */
+   ASSERT(fbS_reg != newS_reg);
+
+   /* The code also assumes the the stencil_max_value is of the form 
+    * 2^n-1 and can therefore be used as a mask for the valid bits in 
+    * addition to a maximum.  Make sure this is the case as well.
+    * The clever math below exploits the fact that incrementing a 
+    * binary number serves to flip all the bits of a number starting at
+    * the LSB and continuing to (and including) the first zero bit
+    * found.  That means that a number and its increment will always
+    * have at least one bit in common (the high order bit, if nothing
+    * else) *unless* the number is zero, *or* the number is of a form
+    * consisting of some number of 1s in the low-order bits followed
+    * by nothing but 0s in the high-order bits.  The latter case
+    * implies it's of the form 2^n-1.
+    */
+   ASSERT(stencil_max_value > 0 && ((stencil_max_value + 1) & stencil_max_value) == 0);
+
+   switch(stencil_op) {
+   case PIPE_STENCIL_OP_KEEP:
+      /* newS = S */
+      spe_move(f, newS_reg, fbS_reg);
+      break;
+
+   case PIPE_STENCIL_OP_ZERO:
+      /* newS = 0 */
+      spe_zero(f, newS_reg);
+      break;
+
+   case PIPE_STENCIL_OP_REPLACE:
+      /* newS = stencil reference value */
+      spe_load_uint(f, newS_reg, stencil_ref_value);
+      break;
+
+   case PIPE_STENCIL_OP_INCR: {
+      /* newS = (s == max ? max : s + 1) */
+      int equals_reg = spe_allocate_available_register(f);
+
+      spe_compare_equal_uint(f, equals_reg, fbS_reg, stencil_max_value);
+      /* Add Word Immediate computes rT = rA + 10-bit signed immediate */
+      spe_ai(f, newS_reg, fbS_reg, 1);
+      /* Select from the current value or the new value based on the equality test */
+      spe_selb(f, newS_reg, newS_reg, fbS_reg, equals_reg);
+
+      spe_release_register(f, equals_reg);
+      break;
+   }
+   case PIPE_STENCIL_OP_DECR: {
+      /* newS = (s == 0 ? 0 : s - 1) */
+      int equals_reg = spe_allocate_available_register(f);
+
+      spe_compare_equal_uint(f, equals_reg, fbS_reg, 0);
+      /* Add Word Immediate with a (-1) value works */
+      spe_ai(f, newS_reg, fbS_reg, -1);
+      /* Select from the current value or the new value based on the equality test */
+      spe_selb(f, newS_reg, newS_reg, fbS_reg, equals_reg);
+
+      spe_release_register(f, equals_reg);
+      break;
+   }
+   case PIPE_STENCIL_OP_INCR_WRAP:
+      /* newS = (s == max ? 0 : s + 1), but since max is 2^n-1, we can
+       * do a normal add and mask off the correct bits 
+       */
+      spe_ai(f, newS_reg, fbS_reg, 1);
+      spe_and_uint(f, newS_reg, newS_reg, stencil_max_value);
+      break;
+
+   case PIPE_STENCIL_OP_DECR_WRAP:
+      /* newS = (s == 0 ? max : s - 1), but we'll pull the same mask trick as above */
+      spe_ai(f, newS_reg, fbS_reg, -1);
+      spe_and_uint(f, newS_reg, newS_reg, stencil_max_value);
+      break;
+
+   case PIPE_STENCIL_OP_INVERT:
+      /* newS = ~s.  We take advantage of the mask/max value to invert only
+       * the valid bits for the field so we don't have to do an extra "and".
+       */
+      spe_xor_uint(f, newS_reg, fbS_reg, stencil_max_value);
+      break;
+
+   default:
+      ASSERT(0);
+   }
+}
+
+
+/**
+ * This function generates code to get all the necessary possible
+ * stencil values.  For each of the output registers (fail_reg,
+ * zfail_reg, and zpass_reg), it either allocates a new register
+ * and calculates a new set of values based on the stencil operation,
+ * or it reuses a register allocation and calculation done for an
+ * earlier (matching) operation, or it reuses the fbS_reg register
+ * (if the stencil operation is KEEP, which doesn't change the 
+ * stencil buffer).
+ *
+ * Since this function allocates a variable number of registers,
+ * to avoid incurring complex logic to free them, they should
+ * be allocated after a spe_allocate_register_set() call
+ * and released by the corresponding spe_release_register_set() call.
+ */
+static void
+gen_get_stencil_values(struct spe_function *f,
+                       const struct pipe_stencil_state *stencil,
+                       const uint depth_enabled,
+                       int fbS_reg, 
+                       int *fail_reg,
+                       int *zfail_reg, 
+                       int *zpass_reg)
+{
+   uint zfail_op;
+
+   /* Stenciling had better be enabled here */
+   ASSERT(stencil->enabled);
+
+   /* If the depth test is not enabled, it is treated as though it always
+    * passes, which means that the zfail_op is not considered - a
+    * failing stencil test triggers the fail_op, and a passing one
+    * triggers the zpass_op
+    *
+    * As an optimization, override calculation of the zfail_op values
+    * if they aren't going to be used.  By setting the value of
+    * the operation to PIPE_STENCIL_OP_KEEP, its value will be assumed
+    * to match the incoming stencil values, and no calculation will
+    * be done.
+    */
+   if (depth_enabled) {
+      zfail_op = stencil->zfail_op;
+   }
+   else {
+      zfail_op = PIPE_STENCIL_OP_KEEP;
+   }
+
+   /* One-sided or front-facing stencil */
+   if (stencil->fail_op == PIPE_STENCIL_OP_KEEP) {
+      *fail_reg = fbS_reg;
+   }
+   else {
+      *fail_reg = spe_allocate_available_register(f);
+      gen_stencil_values(f, stencil->fail_op, stencil->ref_value, 
+         0xff, fbS_reg, *fail_reg);
+   }
+
+   /* Check the possibly overridden value, not the structure value */
+   if (zfail_op == PIPE_STENCIL_OP_KEEP) {
+      *zfail_reg = fbS_reg;
+   }
+   else if (zfail_op == stencil->fail_op) {
+      *zfail_reg = *fail_reg;
+   }
+   else {
+      *zfail_reg = spe_allocate_available_register(f);
+      gen_stencil_values(f, stencil->zfail_op, stencil->ref_value, 
+         0xff, fbS_reg, *zfail_reg);
+   }
+
+   if (stencil->zpass_op == PIPE_STENCIL_OP_KEEP) {
+      *zpass_reg = fbS_reg;
+   }
+   else if (stencil->zpass_op == stencil->fail_op) {
+      *zpass_reg = *fail_reg;
+   }
+   else if (stencil->zpass_op == zfail_op) {
+      *zpass_reg = *zfail_reg;
+   }
+   else {
+      *zpass_reg = spe_allocate_available_register(f);
+      gen_stencil_values(f, stencil->zpass_op, stencil->ref_value, 
+         0xff, fbS_reg, *zpass_reg);
+   }
+}
+
+/**
+ * Note that fbZ_reg may *not* be set on entry, if in fact
+ * the depth test is not enabled.  This function must not use
+ * the register if depth is not enabled.
+ */
+static boolean
+gen_stencil_depth_test(struct spe_function *f, 
+                       const struct pipe_depth_stencil_alpha_state *dsa, 
+                       const uint facing,
+                       const int mask_reg, const int fragZ_reg, 
+                       const int fbZ_reg, const int fbS_reg)
+{
+   /* True if we've generated code that could require writeback to the
+    * depth and/or stencil buffers
+    */
+   boolean modified_buffers = FALSE;
+
+   boolean need_to_calculate_stencil_values;
+   boolean need_to_writemask_stencil_values;
+
+   struct pipe_stencil_state *stencil;
+
+   /* Registers.  We may or may not actually allocate these, depending
+    * on whether the state values indicate that we need them.
+    */
+   int stencil_pass_reg, stencil_fail_reg;
+   int stencil_fail_values, stencil_pass_depth_fail_values, stencil_pass_depth_pass_values;
+   int stencil_writemask_reg;
+   int zmask_reg;
+   int newS_reg;
+
+   /* Stenciling is quite complex: up to six different configurable stencil 
+    * operations/calculations can be required (three each for front-facing
+    * and back-facing fragments).  Many of those operations will likely 
+    * be identical, so there's good reason to try to avoid calculating 
+    * the same values more than once (which unfortunately makes the code less 
+    * straightforward).
+    *
+    * To make register management easier, we start a new 
+    * register set; we can release all the registers in the set at
+    * once, and avoid having to keep track of exactly which registers
+    * we allocate.  We can still allocate and free registers as 
+    * desired (if we know we no longer need a register), but we don't
+    * have to spend the complexity to track the more difficult variant
+    * register usage scenarios.
+    */
+   spe_comment(f, 0, "Allocating stencil register set");
+   spe_allocate_register_set(f);
+
+   /* The facing we're given is the fragment facing; it doesn't
+    * exactly match the stencil facing.  If stencil is enabled,
+    * but two-sided stencil is *not* enabled, we use the same
+    * stencil settings for both front- and back-facing fragments.
+    * We only use the "back-facing" stencil for backfacing fragments
+    * if two-sided stenciling is enabled.
+    */
+   if (facing == CELL_FACING_BACK && dsa->stencil[1].enabled) {
+      stencil = &dsa->stencil[1];
+   }
+   else {
+      stencil = &dsa->stencil[0];
+   }
+
+   /* Calculate the writemask.  If the writemask is trivial (either
+    * all 0s, meaning that we don't need to calculate any stencil values
+    * because they're not going to change the stencil anyway, or all 1s,
+    * meaning that we have to calculate the stencil values but do not
+    * need to mask them), we can avoid generating code.  Don't forget
+    * that we need to consider backfacing stencil, if enabled.
+    *
+    * Note that if the backface stencil is *not* enabled, the backface
+    * stencil will have the same values as the frontface stencil.
+    */
+   if (stencil->fail_op == PIPE_STENCIL_OP_KEEP &&
+       stencil->zfail_op == PIPE_STENCIL_OP_KEEP &&
+       stencil->zpass_op == PIPE_STENCIL_OP_KEEP) {
+       need_to_calculate_stencil_values = FALSE;
+       need_to_writemask_stencil_values = FALSE;
+    }
+    else if (stencil->writemask == 0x0) {
+      /* All changes are writemasked out, so no need to calculate
+       * what those changes might be, and no need to write anything back.
+       */
+      need_to_calculate_stencil_values = FALSE;
+      need_to_writemask_stencil_values = FALSE;
+   }
+   else if (stencil->writemask == 0xff) {
+      /* Still trivial, but a little less so.  We need to write the stencil
+       * values, but we don't need to mask them.
+       */
+      need_to_calculate_stencil_values = TRUE;
+      need_to_writemask_stencil_values = FALSE;
+   }
+   else {
+      /* The general case: calculate, mask, and write */
+      need_to_calculate_stencil_values = TRUE;
+      need_to_writemask_stencil_values = TRUE;
+
+      /* While we're here, generate code that calculates what the
+       * writemask should be.  If backface stenciling is enabled,
+       * and the backface writemask is not the same as the frontface
+       * writemask, we'll have to generate code that merges the
+       * two masks into a single effective mask based on fragment facing.
+       */
+      spe_comment(f, 0, "Computing stencil writemask");
+      stencil_writemask_reg = spe_allocate_available_register(f);
+      spe_load_uint(f, stencil_writemask_reg, dsa->stencil[facing].writemask);
+   }
+
+   /* At least one-sided stenciling must be on.  Generate code that
+    * runs the stencil test on the basic/front-facing stencil, leaving
+    * the mask of passing stencil bits in stencil_pass_reg.  This mask will
+    * be used both to mask the set of active pixels, and also to
+    * determine how the stencil buffer changes.
+    *
+    * This test will *not* change the value in mask_reg (because we don't
+    * yet know whether to apply the two-sided stencil or one-sided stencil).
+    */
+   spe_comment(f, 0, "Running basic stencil test");
+   stencil_pass_reg = spe_allocate_available_register(f);
+   gen_stencil_test(f, stencil, 0xff, mask_reg, fbS_reg, stencil_pass_reg);
+
+   /* Generate code that, given the mask of valid fragments and the
+    * mask of valid fragments that passed the stencil test, computes
+    * the mask of valid fragments that failed the stencil test.  We
+    * have to do this before we run a depth test (because the
+    * depth test should not be performed on fragments that failed the
+    * stencil test, and because the depth test will update the 
+    * mask of valid fragments based on the results of the depth test).
+    */
+   spe_comment(f, 0, "Computing stencil fail mask and updating fragment mask");
+   stencil_fail_reg = spe_allocate_available_register(f);
+   spe_andc(f, stencil_fail_reg, mask_reg, stencil_pass_reg);
+   /* Now remove the stenciled-out pixels from the valid fragment mask,
+    * so we can later use the valid fragment mask in the depth test.
+    */
+   spe_and(f, mask_reg, mask_reg, stencil_pass_reg);
+
+   /* We may not need to calculate stencil values, if the writemask is off */
+   if (need_to_calculate_stencil_values) {
+      /* Generate code that calculates exactly which stencil values we need,
+       * without calculating the same value twice (say, if two different
+       * stencil ops have the same value).  This code will work for one-sided
+       * and two-sided stenciling (so that we take into account that operations
+       * may match between front and back stencils), and will also take into
+       * account whether the depth test is enabled (if the depth test is off,
+       * we don't need any of the zfail results, because the depth test always
+       * is considered to pass if it is disabled).  Any register value that
+       * does not need to be calculated will come back with the same value
+       * that's in fbS_reg.
+       *
+       * This function will allocate a variant number of registers that
+       * will be released as part of the register set.
+       */
+      spe_comment(f, 0, facing == CELL_FACING_FRONT
+                  ? "Computing front-facing stencil values"
+                  : "Computing back-facing stencil values");
+      gen_get_stencil_values(f, stencil, dsa->depth.enabled, fbS_reg, 
+         &stencil_fail_values, &stencil_pass_depth_fail_values, 
+         &stencil_pass_depth_pass_values);
+   }  
+
+   /* We now have all the stencil values we need.  We also need 
+    * the results of the depth test to figure out which
+    * stencil values will become the new stencil values.  (Even if
+    * we aren't actually calculating stencil values, we need to apply
+    * the depth test if it's enabled.)
+    *
+    * The code generated by gen_depth_test() returns the results of the
+    * test in the given register, but also alters the mask_reg based
+    * on the results of the test.
+    */
+   if (dsa->depth.enabled) {
+      spe_comment(f, 0, "Running stencil depth test");
+      zmask_reg = spe_allocate_available_register(f);
+      modified_buffers |= gen_depth_test(f, dsa, mask_reg, fragZ_reg,
+                                         fbZ_reg, zmask_reg);
+   }
+
+   if (need_to_calculate_stencil_values) {
+
+      /* If we need to writemask the stencil values before going into
+       * the stencil buffer, we'll have to use a new register to
+       * hold the new values.  If not, we can just keep using the
+       * current register.
+       */
+      if (need_to_writemask_stencil_values) {
+         newS_reg = spe_allocate_available_register(f);
+         spe_comment(f, 0, "Saving current stencil values for writemasking");
+         spe_move(f, newS_reg, fbS_reg);
+      }
+      else {
+         newS_reg = fbS_reg;
+      }
+
+      /* Merge in the selected stencil fail values */
+      if (stencil_fail_values != fbS_reg) {
+         spe_comment(f, 0, "Loading stencil fail values");
+         spe_selb(f, newS_reg, newS_reg, stencil_fail_values, stencil_fail_reg);
+         modified_buffers = TRUE;
+      }
+
+      /* Same for the stencil pass/depth fail values.  If this calculation
+       * is not needed (say, if depth test is off), then the
+       * stencil_pass_depth_fail_values register will be equal to fbS_reg
+       * and we'll skip the calculation.
+       */
+      if (stencil_pass_depth_fail_values != fbS_reg) {
+         /* We don't actually have a stencil pass/depth fail mask yet.
+          * Calculate it here from the stencil passing mask and the
+          * depth passing mask.  Note that zmask_reg *must* have been
+          * set above if we're here.
+          */
+         uint stencil_pass_depth_fail_mask =
+            spe_allocate_available_register(f);
+
+         spe_comment(f, 0, "Loading stencil pass/depth fail values");
+         spe_andc(f, stencil_pass_depth_fail_mask, stencil_pass_reg, zmask_reg);
+
+         spe_selb(f, newS_reg, newS_reg, stencil_pass_depth_fail_values,
+                  stencil_pass_depth_fail_mask);
+
+         spe_release_register(f, stencil_pass_depth_fail_mask);
+         modified_buffers = TRUE;
+      }
+
+      /* Same for the stencil pass/depth pass mask.  Note that we
+       * *can* get here with zmask_reg being unset (if the depth
+       * test is off but the stencil test is on).  In this case,
+       * we assume the depth test passes, and don't need to mask
+       * the stencil pass mask with the Z mask.
+       */
+      if (stencil_pass_depth_pass_values != fbS_reg) {
+         if (dsa->depth.enabled) {
+            uint stencil_pass_depth_pass_mask = spe_allocate_available_register(f);
+            /* We'll need a separate register */
+            spe_comment(f, 0, "Loading stencil pass/depth pass values");
+            spe_and(f, stencil_pass_depth_pass_mask, stencil_pass_reg, zmask_reg);
+            spe_selb(f, newS_reg, newS_reg, stencil_pass_depth_pass_values, stencil_pass_depth_pass_mask);
+            spe_release_register(f, stencil_pass_depth_pass_mask);
+         }
+         else {
+            /* We can use the same stencil-pass register */
+            spe_comment(f, 0, "Loading stencil pass values");
+            spe_selb(f, newS_reg, newS_reg, stencil_pass_depth_pass_values, stencil_pass_reg);
+         }
+         modified_buffers = TRUE;
+      }
+
+      /* Almost done.  If we need to writemask, do it now, leaving the
+       * results in the fbS_reg register passed in.  If we don't need
+       * to writemask, then the results are *already* in the fbS_reg,
+       * so there's nothing more to do.
+       */
+
+      if (need_to_writemask_stencil_values && modified_buffers) {
+         /* The Select Bytes command makes a fine writemask.  Where
+          * the mask is 0, the first (original) values are retained,
+          * effectively masking out changes.  Where the mask is 1, the
+          * second (new) values are retained, incorporating changes.
+          */
+         spe_comment(f, 0, "Writemasking new stencil values");
+         spe_selb(f, fbS_reg, fbS_reg, newS_reg, stencil_writemask_reg);
+      }
+
+   } /* done calculating stencil values */
+
+   /* The stencil and/or depth values have been applied, and the
+    * mask_reg, fbS_reg, and fbZ_reg values have been updated.
+    * We're all done, except that we've allocated a fair number
+    * of registers that we didn't bother tracking.  Release all
+    * those registers as part of the register set, and go home.
+    */
+   spe_comment(f, 0, "Releasing stencil register set");
+   spe_release_register_set(f);
+
+   /* Return TRUE if we could have modified the stencil and/or
+    * depth buffers.
+    */
+   return modified_buffers;
+}
+
+
+/**
+ * Generate depth and/or stencil test code.
+ * \param cell  context
+ * \param dsa  depth/stencil/alpha state
+ * \param f  spe function to emit
+ * \param facing  either CELL_FACING_FRONT or CELL_FACING_BACK
+ * \param mask_reg  register containing the pixel alive/dead mask
+ * \param depth_tile_reg  register containing address of z/stencil tile
+ * \param quad_offset_reg  offset to quad from start of tile
+ * \param fragZ_reg  register containg fragment Z values
+ */
+static void
+gen_depth_stencil(struct cell_context *cell,
+                  const struct pipe_depth_stencil_alpha_state *dsa,
+                  struct spe_function *f,
+                  uint facing,
+                  int mask_reg,
+                  int depth_tile_reg,
+                  int quad_offset_reg,
+                  int fragZ_reg)
+
+{
+   const enum pipe_format zs_format = cell->framebuffer.zsbuf->format;
+   boolean write_depth_stencil;
+
+   /* framebuffer's combined z/stencil values register */
+   int fbZS_reg = spe_allocate_available_register(f);
+
+   /* Framebufer Z values register */
+   int fbZ_reg = spe_allocate_available_register(f);
+
+   /* Framebuffer stencil values register (may not be used) */
+   int fbS_reg = spe_allocate_available_register(f);
+
+   /* 24-bit mask register (may not be used) */
+   int zmask_reg = spe_allocate_available_register(f);
+
+   /**
+    * The following code:
+    * 1. fetch quad of packed Z/S values from the framebuffer tile.
+    * 2. extract the separate the Z and S values from packed values
+    * 3. convert fragment Z values from float in [0,1] to 32/24/16-bit ints
+    *
+    * The instructions for doing this are interleaved for better performance.
+    */
+   spe_comment(f, 0, "Fetch Z/stencil quad from tile");
+
+   switch(zs_format) {
+   case PIPE_FORMAT_S8Z24_UNORM: /* fall through */
+   case PIPE_FORMAT_X8Z24_UNORM:
+      /* prepare mask to extract Z vals from ZS vals */
+      spe_load_uint(f, zmask_reg, 0x00ffffff);
+
+      /* convert fragment Z from [0,1] to 32-bit ints */
+      spe_cfltu(f, fragZ_reg, fragZ_reg, 32);
+
+      /* Load: fbZS_reg = memory[depth_tile_reg + offset_reg] */
+      spe_lqx(f, fbZS_reg, depth_tile_reg, quad_offset_reg);
+
+      /* right shift 32-bit fragment Z to 24 bits */
+      spe_rotmi(f, fragZ_reg, fragZ_reg, -8);
+
+      /* extract 24-bit Z values from ZS values by masking */
+      spe_and(f, fbZ_reg, fbZS_reg, zmask_reg);
+
+      /* extract 8-bit stencil values by shifting */
+      spe_rotmi(f, fbS_reg, fbZS_reg, -24);
+      break;
+
+   case PIPE_FORMAT_Z24S8_UNORM: /* fall through */
+   case PIPE_FORMAT_Z24X8_UNORM:
+      /* convert fragment Z from [0,1] to 32-bit ints */
+      spe_cfltu(f, fragZ_reg, fragZ_reg, 32);
+
+      /* Load: fbZS_reg = memory[depth_tile_reg + offset_reg] */
+      spe_lqx(f, fbZS_reg, depth_tile_reg, quad_offset_reg);
+
+      /* right shift 32-bit fragment Z to 24 bits */
+      spe_rotmi(f, fragZ_reg, fragZ_reg, -8);
+
+      /* extract 24-bit Z values from ZS values by shifting */
+      spe_rotmi(f, fbZ_reg, fbZS_reg, -8);
+
+      /* extract 8-bit stencil values by masking */
+      spe_and_uint(f, fbS_reg, fbZS_reg, 0x000000ff);
+      break;
+
+   case PIPE_FORMAT_Z32_UNORM:
+      /* Load: fbZ_reg = memory[depth_tile_reg + offset_reg] */
+      spe_lqx(f, fbZ_reg, depth_tile_reg, quad_offset_reg);
+
+      /* convert fragment Z from [0,1] to 32-bit ints */
+      spe_cfltu(f, fragZ_reg, fragZ_reg, 32);
+
+      /* No stencil, so can't do anything there */
+      break;
+
+   case PIPE_FORMAT_Z16_UNORM:
+      /* XXX This code for 16bpp Z is broken! */
+
+      /* Load: fbZS_reg = memory[depth_tile_reg + offset_reg] */
+      spe_lqx(f, fbZS_reg, depth_tile_reg, quad_offset_reg);
+
+      /* Copy over 4 32-bit values */
+      spe_move(f, fbZ_reg, fbZS_reg);
+
+      /* convert Z from [0,1] to 16-bit ints */
+      spe_cfltu(f, fragZ_reg, fragZ_reg, 32);
+      spe_rotmi(f, fragZ_reg, fragZ_reg, -16);
+      /* No stencil */
+      break;
+
+   default:
+      ASSERT(0); /* invalid format */
+   }
+
+   /* If stencil is enabled, use the stencil-specific code
+    * generator to generate both the stencil and depth (if needed)
+    * tests.  Otherwise, if only depth is enabled, generate
+    * a quick depth test.  The test generators themselves will
+    * report back whether the depth/stencil buffer has to be
+    * written back.
+    */
+   if (dsa->stencil[0].enabled) {
+      /* This will perform the stencil and depth tests, and update
+       * the mask_reg, fbZ_reg, and fbS_reg as required by the
+       * tests.
+       */
+      ASSERT(fbS_reg >= 0);
+      spe_comment(f, 0, "Perform stencil test");
+
+      /* Note that fbZ_reg may not be set on entry, if stenciling
+       * is enabled but there's no Z-buffer.  The 
+       * gen_stencil_depth_test() function must ignore the
+       * fbZ_reg register if depth is not enabled.
+       */
+      write_depth_stencil = gen_stencil_depth_test(f, dsa, facing,
+                                                   mask_reg, fragZ_reg,
+                                                   fbZ_reg, fbS_reg);
+   }
+   else if (dsa->depth.enabled) {
+      int zmask_reg = spe_allocate_available_register(f);
+      ASSERT(fbZ_reg >= 0);
+      spe_comment(f, 0, "Perform depth test");
+      write_depth_stencil = gen_depth_test(f, dsa, mask_reg, fragZ_reg,
+                                           fbZ_reg, zmask_reg);
+      spe_release_register(f, zmask_reg);
+   }
+   else {
+      write_depth_stencil = FALSE;
+   }
+
+   if (write_depth_stencil) {
+      /* Merge latest Z and Stencil values into fbZS_reg.
+       * fbZ_reg has four Z vals in bits [23..0] or bits [15..0].
+       * fbS_reg has four 8-bit Z values in bits [7..0].
+       */
+      spe_comment(f, 0, "Store quad's depth/stencil values in tile");
+      if (zs_format == PIPE_FORMAT_S8Z24_UNORM ||
+          zs_format == PIPE_FORMAT_X8Z24_UNORM) {
+         spe_shli(f, fbS_reg, fbS_reg, 24); /* fbS = fbS << 24 */
+         spe_or(f, fbZS_reg, fbS_reg, fbZ_reg); /* fbZS = fbS | fbZ */
+      }
+      else if (zs_format == PIPE_FORMAT_Z24S8_UNORM ||
+               zs_format == PIPE_FORMAT_Z24X8_UNORM) {
+         spe_shli(f, fbZ_reg, fbZ_reg, 8); /* fbZ = fbZ << 8 */
+         spe_or(f, fbZS_reg, fbS_reg, fbZ_reg); /* fbZS = fbS | fbZ */
+      }
+      else if (zs_format == PIPE_FORMAT_Z32_UNORM) {
+         spe_move(f, fbZS_reg, fbZ_reg); /* fbZS = fbZ */
+      }
+      else if (zs_format == PIPE_FORMAT_Z16_UNORM) {
+         spe_move(f, fbZS_reg, fbZ_reg); /* fbZS = fbZ */
+      }
+      else if (zs_format == PIPE_FORMAT_S8_UNORM) {
+         ASSERT(0);   /* XXX to do */
+      }
+      else {
+         ASSERT(0); /* bad zs_format */
+      }
+
+      /* Store: memory[depth_tile_reg + quad_offset_reg] = fbZS */
+      spe_stqx(f, fbZS_reg, depth_tile_reg, quad_offset_reg);
+   }
+
+   /* Don't need these any more */
+   spe_release_register(f, fbZS_reg);
+   spe_release_register(f, fbZ_reg);
+   spe_release_register(f, fbS_reg);
+   spe_release_register(f, zmask_reg);
+}
+
+
+
+/**
+ * Generate SPE code to implement the fragment operations (alpha test,
+ * depth test, stencil test, blending, colormask, and final
+ * framebuffer write) as specified by the current context state.
+ *
+ * Logically, this code will be called after running the fragment
+ * shader.  But under some circumstances we could run some of this
+ * code before the fragment shader to cull fragments/quads that are
+ * totally occluded/discarded.
+ *
+ * XXX we only support PIPE_FORMAT_Z24S8_UNORM z/stencil buffer right now.
+ *
+ * See the spu_default_fragment_ops() function to see how the per-fragment
+ * operations would be done with ordinary C code.
+ * The code we generate here though has no branches, is SIMD, etc and
+ * should be much faster.
+ *
+ * \param cell  the rendering context (in)
+ * \param facing whether the generated code is for front-facing or 
+ *              back-facing fragments
+ * \param f     the generated function (in/out); on input, the function
+ *              must already have been initialized.  On exit, whatever
+ *              instructions within the generated function have had
+ *              the fragment ops appended.
+ */
+void
+cell_gen_fragment_function(struct cell_context *cell,
+                           const uint facing,
+                           struct spe_function *f)
+{
+   const struct pipe_depth_stencil_alpha_state *dsa = cell->depth_stencil;
+   const struct pipe_blend_state *blend = cell->blend;
+   const struct pipe_blend_color *blend_color = &cell->blend_color;
+   const enum pipe_format color_format = cell->framebuffer.cbufs[0]->format;
+
+   /* For SPE function calls: reg $3 = first param, $4 = second param, etc. */
+   const int x_reg = 3;  /* uint */
+   const int y_reg = 4;  /* uint */
+   const int color_tile_reg = 5;  /* tile_t * */
+   const int depth_tile_reg = 6;  /* tile_t * */
+   const int fragZ_reg = 7;   /* vector float */
+   const int fragR_reg = 8;   /* vector float */
+   const int fragG_reg = 9;   /* vector float */
+   const int fragB_reg = 10;  /* vector float */
+   const int fragA_reg = 11;  /* vector float */
+   const int mask_reg = 12;   /* vector uint */
+
+   ASSERT(facing == CELL_FACING_FRONT || facing == CELL_FACING_BACK);
+
+   /* offset of quad from start of tile
+    * XXX assuming 4-byte pixels for color AND Z/stencil!!!!
+    */
+   int quad_offset_reg;
+
+   int fbRGBA_reg;  /**< framebuffer's RGBA colors for quad */
+
+   if (cell->debug_flags & CELL_DEBUG_ASM) {
+      spe_print_code(f, TRUE);
+      spe_indent(f, 8);
+      spe_comment(f, -4, facing == CELL_FACING_FRONT
+                  ? "Begin front-facing per-fragment ops"
+                  : "Begin back-facing per-fragment ops");
+   }
+
+   spe_allocate_register(f, x_reg);
+   spe_allocate_register(f, y_reg);
+   spe_allocate_register(f, color_tile_reg);
+   spe_allocate_register(f, depth_tile_reg);
+   spe_allocate_register(f, fragZ_reg);
+   spe_allocate_register(f, fragR_reg);
+   spe_allocate_register(f, fragG_reg);
+   spe_allocate_register(f, fragB_reg);
+   spe_allocate_register(f, fragA_reg);
+   spe_allocate_register(f, mask_reg);
+
+   quad_offset_reg = spe_allocate_available_register(f);
+   fbRGBA_reg = spe_allocate_available_register(f);
+
+   /* compute offset of quad from start of tile, in bytes */
+   {
+      int x2_reg = spe_allocate_available_register(f);
+      int y2_reg = spe_allocate_available_register(f);
+
+      ASSERT(TILE_SIZE == 32);
+
+      spe_comment(f, 0, "Compute quad offset within tile");
+      spe_rotmi(f, y2_reg, y_reg, -1);  /* y2 = y / 2 */
+      spe_rotmi(f, x2_reg, x_reg, -1);  /* x2 = x / 2 */
+      spe_shli(f, y2_reg, y2_reg, 4);   /* y2 *= 16 */
+      spe_a(f, quad_offset_reg, y2_reg, x2_reg);  /* offset = y2 + x2 */
+      spe_shli(f, quad_offset_reg, quad_offset_reg, 4);   /* offset *= 16 */
+
+      spe_release_register(f, x2_reg);
+      spe_release_register(f, y2_reg);
+   }
+
+   /* Generate the alpha test, if needed. */
+   if (dsa->alpha.enabled) {
+      gen_alpha_test(dsa, f, mask_reg, fragA_reg);
+   }
+
+   /* generate depth and/or stencil test code */
+   if (dsa->depth.enabled || dsa->stencil[0].enabled) {
+      gen_depth_stencil(cell, dsa, f,
+                        facing,
+                        mask_reg,
+                        depth_tile_reg,
+                        quad_offset_reg,
+                        fragZ_reg);
+   }
+
+   /* Get framebuffer quad/colors.  We'll need these for blending,
+    * color masking, and to obey the quad/pixel mask.
+    * Load: fbRGBA_reg = memory[color_tile + quad_offset]
+    * Note: if mask={~0,~0,~0,~0} and we're not blending or colormasking
+    * we could skip this load.
+    */
+   spe_comment(f, 0, "Fetch quad colors from tile");
+   spe_lqx(f, fbRGBA_reg, color_tile_reg, quad_offset_reg);
+
+   if (blend->blend_enable) {
+      spe_comment(f, 0, "Perform blending");
+      gen_blend(blend, blend_color, f, color_format,
+                fragR_reg, fragG_reg, fragB_reg, fragA_reg, fbRGBA_reg);
+   }
+
+   /*
+    * Write fragment colors to framebuffer/tile.
+    * This involves converting the fragment colors from float[4] to the
+    * tile's specific format and obeying the quad/pixel mask.
+    */
+   {
+      int rgba_reg = spe_allocate_available_register(f);
+
+      /* Pack four float colors as four 32-bit int colors */
+      spe_comment(f, 0, "Convert float quad colors to packed int framebuffer colors");
+      gen_pack_colors(f, color_format,
+                      fragR_reg, fragG_reg, fragB_reg, fragA_reg,
+                      rgba_reg);
+
+      if (blend->logicop_enable) {
+         spe_comment(f, 0, "Compute logic op");
+         gen_logicop(blend, f, rgba_reg, fbRGBA_reg);
+      }
+
+      if (blend->colormask != PIPE_MASK_RGBA) {
+         spe_comment(f, 0, "Compute color mask");
+         gen_colormask(f, blend->colormask, color_format, rgba_reg, fbRGBA_reg);
+      }
+
+      /* Mix fragment colors with framebuffer colors using the quad/pixel mask:
+       * if (mask[i])
+       *    rgba[i] = rgba[i];
+       * else
+       *    rgba[i] = framebuffer[i];
+       */
+      spe_selb(f, rgba_reg, fbRGBA_reg, rgba_reg, mask_reg);
+
+      /* Store updated quad in tile:
+       * memory[color_tile + quad_offset] = rgba_reg;
+       */
+      spe_comment(f, 0, "Store quad colors into color tile");
+      spe_stqx(f, rgba_reg, color_tile_reg, quad_offset_reg);
+
+      spe_release_register(f, rgba_reg);
+   }
+
+   //printf("gen_fragment_ops nr instructions: %u\n", f->num_inst);
+
+   spe_bi(f, SPE_REG_RA, 0, 0);  /* return from function call */
+
+   spe_release_register(f, fbRGBA_reg);
+   spe_release_register(f, quad_offset_reg);
+
+   if (cell->debug_flags & CELL_DEBUG_ASM) {
+      char buffer[1024];
+      sprintf(buffer, "End %s-facing per-fragment ops: %d instructions", 
+         facing == CELL_FACING_FRONT ? "front" : "back", f->num_inst);
+      spe_comment(f, -4, buffer);
+   }
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_gen_fragment.h b/src/gallium/drivers/cell/ppu/cell_gen_fragment.h
new file mode 100644
index 0000000000..21b35d1faf
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_gen_fragment.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef CELL_GEN_FRAGMENT_H
+#define CELL_GEN_FRAGMENT_H
+
+
+extern void
+cell_gen_fragment_function(struct cell_context *cell, const uint facing, struct spe_function *f);
+
+
+#endif /* CELL_GEN_FRAGMENT_H */
+
diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c
new file mode 100644
index 0000000000..ca358ed031
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c
@@ -0,0 +1,358 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/* Authors:
+ *  Keith Whitwell <keith@tungstengraphics.com>
+ *  Brian Paul
+ */
+
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "draw/draw_context.h"
+#include "cell_context.h"
+#include "cell_flush.h"
+#include "cell_pipe_state.h"
+#include "cell_state.h"
+#include "cell_texture.h"
+
+
+
+static void *
+cell_create_blend_state(struct pipe_context *pipe,
+                        const struct pipe_blend_state *blend)
+{
+   return mem_dup(blend, sizeof(*blend));
+}
+
+
+static void
+cell_bind_blend_state(struct pipe_context *pipe, void *blend)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   draw_flush(cell->draw);
+
+   cell->blend = (struct pipe_blend_state *) blend;
+   cell->dirty |= CELL_NEW_BLEND;
+}
+
+
+static void
+cell_delete_blend_state(struct pipe_context *pipe, void *blend)
+{
+   FREE(blend);
+}
+
+
+static void
+cell_set_blend_color(struct pipe_context *pipe,
+                     const struct pipe_blend_color *blend_color)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   draw_flush(cell->draw);
+
+   cell->blend_color = *blend_color;
+
+   cell->dirty |= CELL_NEW_BLEND;
+}
+
+
+
+
+static void *
+cell_create_depth_stencil_alpha_state(struct pipe_context *pipe,
+                 const struct pipe_depth_stencil_alpha_state *dsa)
+{
+   return mem_dup(dsa, sizeof(*dsa));
+}
+
+
+static void
+cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe,
+                                    void *dsa)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   draw_flush(cell->draw);
+
+   cell->depth_stencil = (struct pipe_depth_stencil_alpha_state *) dsa;
+   cell->dirty |= CELL_NEW_DEPTH_STENCIL;
+}
+
+
+static void
+cell_delete_depth_stencil_alpha_state(struct pipe_context *pipe, void *dsa)
+{
+   FREE(dsa);
+}
+
+
+static void
+cell_set_clip_state(struct pipe_context *pipe,
+                    const struct pipe_clip_state *clip)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   /* pass the clip state to the draw module */
+   draw_set_clip_state(cell->draw, clip);
+}
+
+
+
+/* Called when driver state tracker notices changes to the viewport
+ * matrix:
+ */
+static void
+cell_set_viewport_state( struct pipe_context *pipe,
+                         const struct pipe_viewport_state *viewport )
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   cell->viewport = *viewport; /* struct copy */
+   cell->dirty |= CELL_NEW_VIEWPORT;
+
+   /* pass the viewport info to the draw module */
+   draw_set_viewport_state(cell->draw, viewport);
+
+   /* Using tnl/ and vf/ modules is temporary while getting started.
+    * Full pipe will have vertex shader, vertex fetch of its own.
+    */
+}
+
+
+static void
+cell_set_scissor_state( struct pipe_context *pipe,
+                        const struct pipe_scissor_state *scissor )
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   memcpy( &cell->scissor, scissor, sizeof(*scissor) );
+   cell->dirty |= CELL_NEW_SCISSOR;
+}
+
+
+static void
+cell_set_polygon_stipple( struct pipe_context *pipe,
+                          const struct pipe_poly_stipple *stipple )
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   memcpy( &cell->poly_stipple, stipple, sizeof(*stipple) );
+   cell->dirty |= CELL_NEW_STIPPLE;
+}
+
+
+
+static void *
+cell_create_rasterizer_state(struct pipe_context *pipe,
+                             const struct pipe_rasterizer_state *rasterizer)
+{
+   return mem_dup(rasterizer, sizeof(*rasterizer));
+}
+
+
+static void
+cell_bind_rasterizer_state(struct pipe_context *pipe, void *rast)
+{
+   struct pipe_rasterizer_state *rasterizer =
+      (struct pipe_rasterizer_state *) rast;
+   struct cell_context *cell = cell_context(pipe);
+
+   /* pass-through to draw module */
+   draw_set_rasterizer_state(cell->draw, rasterizer);
+
+   cell->rasterizer = rasterizer;
+
+   cell->dirty |= CELL_NEW_RASTERIZER;
+}
+
+
+static void
+cell_delete_rasterizer_state(struct pipe_context *pipe, void *rasterizer)
+{
+   FREE(rasterizer);
+}
+
+
+
+static void *
+cell_create_sampler_state(struct pipe_context *pipe,
+                          const struct pipe_sampler_state *sampler)
+{
+   return mem_dup(sampler, sizeof(*sampler));
+}
+
+
+static void
+cell_bind_sampler_states(struct pipe_context *pipe,
+                         unsigned num, void **samplers)
+{
+   struct cell_context *cell = cell_context(pipe);
+   uint i, changed = 0x0;
+
+   assert(num <= CELL_MAX_SAMPLERS);
+
+   draw_flush(cell->draw);
+
+   for (i = 0; i < CELL_MAX_SAMPLERS; i++) {
+      struct pipe_sampler_state *new_samp = i < num ? samplers[i] : NULL;
+      if (cell->sampler[i] != new_samp) {
+         cell->sampler[i] = new_samp;
+         changed |= (1 << i);
+      }
+   }
+
+   if (changed) {
+      cell->dirty |= CELL_NEW_SAMPLER;
+      cell->dirty_samplers |= changed;
+   }
+}
+
+
+static void
+cell_delete_sampler_state(struct pipe_context *pipe,
+                              void *sampler)
+{
+   FREE( sampler );
+}
+
+
+
+static void
+cell_set_sampler_textures(struct pipe_context *pipe,
+                          unsigned num, struct pipe_texture **texture)
+{
+   struct cell_context *cell = cell_context(pipe);
+   uint i, changed = 0x0;
+
+   assert(num <= CELL_MAX_SAMPLERS);
+
+   for (i = 0; i < CELL_MAX_SAMPLERS; i++) {
+      struct pipe_texture *new_tex = i < num ? texture[i] : NULL;
+      if ((struct pipe_texture *) cell->texture[i] != new_tex) {
+         pipe_texture_reference((struct pipe_texture **) &cell->texture[i],
+                                new_tex);
+         changed |= (1 << i);
+      }
+   }
+
+   cell->num_textures = num;
+
+   if (changed) {
+      cell->dirty |= CELL_NEW_TEXTURE;
+      cell->dirty_textures |= changed;
+   }
+}
+
+
+
+static void
+cell_set_framebuffer_state(struct pipe_context *pipe,
+                           const struct pipe_framebuffer_state *fb)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   if (1 /*memcmp(&cell->framebuffer, fb, sizeof(*fb))*/) {
+      struct pipe_surface *csurf = fb->cbufs[0];
+      struct pipe_surface *zsurf = fb->zsbuf;
+      uint i;
+      uint flags = (PIPE_BUFFER_USAGE_GPU_WRITE |
+                    PIPE_BUFFER_USAGE_GPU_READ);
+
+      /* unmap old surfaces */
+      for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
+         if (cell->framebuffer.cbufs[i] && cell->cbuf_map[i]) {
+            pipe_surface_unmap(cell->framebuffer.cbufs[i]);
+            cell->cbuf_map[i] = NULL;
+         }
+      }
+
+      if (cell->framebuffer.zsbuf && cell->zsbuf_map) {
+         pipe_surface_unmap(cell->framebuffer.zsbuf);
+         cell->zsbuf_map = NULL;
+      }
+
+      /* Finish any pending rendering to the current surface before
+       * installing a new surface!
+       */
+      cell_flush_int(cell, CELL_FLUSH_WAIT);
+
+      /* update my state
+       * (this is also where old surfaces will finally get freed)
+       */
+      cell->framebuffer.width = fb->width;
+      cell->framebuffer.height = fb->height;
+      cell->framebuffer.nr_cbufs = fb->nr_cbufs;
+      for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
+         pipe_surface_reference(&cell->framebuffer.cbufs[i], fb->cbufs[i]);
+      }
+      pipe_surface_reference(&cell->framebuffer.zsbuf, fb->zsbuf);
+
+      /* map new surfaces */
+      if (csurf)
+         cell->cbuf_map[0] = pipe_surface_map(csurf, flags);
+
+      if (zsurf)
+         cell->zsbuf_map = pipe_surface_map(zsurf, flags);
+
+      cell->dirty |= CELL_NEW_FRAMEBUFFER;
+   }
+}
+
+
+
+void
+cell_init_state_functions(struct cell_context *cell)
+{
+   cell->pipe.create_blend_state = cell_create_blend_state;
+   cell->pipe.bind_blend_state   = cell_bind_blend_state;
+   cell->pipe.delete_blend_state = cell_delete_blend_state;
+
+   cell->pipe.create_sampler_state = cell_create_sampler_state;
+   cell->pipe.bind_sampler_states = cell_bind_sampler_states;
+   cell->pipe.delete_sampler_state = cell_delete_sampler_state;
+
+   cell->pipe.set_sampler_textures = cell_set_sampler_textures;
+
+   cell->pipe.create_depth_stencil_alpha_state = cell_create_depth_stencil_alpha_state;
+   cell->pipe.bind_depth_stencil_alpha_state   = cell_bind_depth_stencil_alpha_state;
+   cell->pipe.delete_depth_stencil_alpha_state = cell_delete_depth_stencil_alpha_state;
+
+   cell->pipe.create_rasterizer_state = cell_create_rasterizer_state;
+   cell->pipe.bind_rasterizer_state   = cell_bind_rasterizer_state;
+   cell->pipe.delete_rasterizer_state = cell_delete_rasterizer_state;
+
+   cell->pipe.set_blend_color = cell_set_blend_color;
+   cell->pipe.set_clip_state = cell_set_clip_state;
+
+   cell->pipe.set_framebuffer_state = cell_set_framebuffer_state;
+
+   cell->pipe.set_polygon_stipple = cell_set_polygon_stipple;
+   cell->pipe.set_scissor_state = cell_set_scissor_state;
+   cell->pipe.set_viewport_state = cell_set_viewport_state;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.h b/src/gallium/drivers/cell/ppu/cell_pipe_state.h
new file mode 100644
index 0000000000..1889bd52ff
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef CELL_PIPE_STATE_H
+#define CELL_PIPE_STATE_H
+
+
+struct cell_context;
+
+extern void
+cell_init_state_functions(struct cell_context *cell);
+
+
+#endif /* CELL_PIPE_STATE_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_render.c b/src/gallium/drivers/cell/ppu/cell_render.c
new file mode 100644
index 0000000000..79cb8df82f
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_render.c
@@ -0,0 +1,211 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief  Last stage of 'draw' pipeline: send tris to SPUs.
+ * \author  Brian Paul
+ */
+
+#include "cell_context.h"
+#include "cell_render.h"
+#include "cell_spu.h"
+#include "util/u_memory.h"
+#include "draw/draw_private.h"
+
+
+struct render_stage {
+   struct draw_stage stage; /**< This must be first (base class) */
+
+   struct cell_context *cell;
+};
+
+
+static INLINE struct render_stage *
+render_stage(struct draw_stage *stage)
+{
+   return (struct render_stage *) stage;
+}
+
+
+static void render_begin( struct draw_stage *stage )
+{
+#if 0
+   struct render_stage *render = render_stage(stage);
+   struct cell_context *sp = render->cell;
+   const struct pipe_shader_state *fs = &render->cell->fs->shader;
+   render->quad.nr_attrs = render->cell->nr_frag_attrs;
+
+   render->firstFpInput = fs->input_semantic_name[0];
+
+   sp->quad.first->begin(sp->quad.first);
+#endif
+}
+
+
+static void render_end( struct draw_stage *stage )
+{
+}
+
+
+static void reset_stipple_counter( struct draw_stage *stage )
+{
+   struct render_stage *render = render_stage(stage);
+   /*render->cell->line_stipple_counter = 0;*/
+}
+
+
+static void
+render_point(struct draw_stage *stage, struct prim_header *prim)
+{
+}
+
+
+static void
+render_line(struct draw_stage *stage, struct prim_header *prim)
+{
+}
+
+
+/** Write a vertex into the prim buffer */
+static void
+save_vertex(struct cell_prim_buffer *buf, uint pos,
+            const struct vertex_header *vert)
+{
+   uint attr, j;
+
+   for (attr = 0; attr < 2; attr++) {
+      for (j = 0; j < 4; j++) {
+         buf->vertex[pos][attr][j] = vert->data[attr][j];
+      }
+   }
+
+   /* update bounding box */
+   if (vert->data[0][0] < buf->xmin)
+      buf->xmin = vert->data[0][0];
+   if (vert->data[0][0] > buf->xmax)
+      buf->xmax = vert->data[0][0];
+   if (vert->data[0][1] < buf->ymin)
+      buf->ymin = vert->data[0][1];
+   if (vert->data[0][1] > buf->ymax)
+      buf->ymax = vert->data[0][1];
+}
+
+
+static void
+render_tri(struct draw_stage *stage, struct prim_header *prim)
+{
+   struct render_stage *rs = render_stage(stage);
+   struct cell_context *cell = rs->cell;
+   struct cell_prim_buffer *buf = &cell->prim_buffer;
+   uint i;
+
+   if (buf->num_verts + 3 > CELL_MAX_VERTS) {
+      cell_flush_prim_buffer(cell);
+   }
+
+   i = buf->num_verts;
+   assert(i+2 <= CELL_MAX_VERTS);
+   save_vertex(buf, i+0, prim->v[0]);
+   save_vertex(buf, i+1, prim->v[1]);
+   save_vertex(buf, i+2, prim->v[2]);
+   buf->num_verts += 3;
+}
+
+
+/**
+ * Send the a RENDER command to all SPUs to have them render the prims
+ * in the current prim_buffer.
+ */
+void
+cell_flush_prim_buffer(struct cell_context *cell)
+{
+   uint i;
+
+   if (cell->prim_buffer.num_verts == 0)
+      return;
+
+   for (i = 0; i < cell->num_spus; i++) {
+      struct cell_command_render *render = &cell_global.command[i].render;
+      render->prim_type = PIPE_PRIM_TRIANGLES;
+      render->num_verts = cell->prim_buffer.num_verts;
+      render->front_winding = cell->rasterizer->front_winding;
+      render->vertex_size = cell->vertex_info->size * 4;
+      render->xmin = cell->prim_buffer.xmin;
+      render->ymin = cell->prim_buffer.ymin;
+      render->xmax = cell->prim_buffer.xmax;
+      render->ymax = cell->prim_buffer.ymax;
+      render->vertex_data = &cell->prim_buffer.vertex;
+      ASSERT_ALIGN16(render->vertex_data);
+      send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_RENDER);
+   }
+
+   cell->prim_buffer.num_verts = 0;
+
+   cell->prim_buffer.xmin = 1e100;
+   cell->prim_buffer.ymin = 1e100;
+   cell->prim_buffer.xmax = -1e100;
+   cell->prim_buffer.ymax = -1e100;
+
+   /* XXX temporary, need to double-buffer the prim buffer until we get
+    * a real command buffer/list system.
+    */
+   cell_flush(&cell->pipe, 0x0);
+}
+
+
+
+static void render_destroy( struct draw_stage *stage )
+{
+   FREE( stage );
+}
+
+
+/**
+ * Create a new draw/render stage.  This will be plugged into the
+ * draw module as the last pipeline stage.
+ */
+struct draw_stage *cell_draw_render_stage( struct cell_context *cell )
+{
+   struct render_stage *render = CALLOC_STRUCT(render_stage);
+
+   render->cell = cell;
+   render->stage.draw = cell->draw;
+   render->stage.begin = render_begin;
+   render->stage.point = render_point;
+   render->stage.line = render_line;
+   render->stage.tri = render_tri;
+   render->stage.end = render_end;
+   render->stage.reset_stipple_counter = reset_stipple_counter;
+   render->stage.destroy = render_destroy;
+
+   /*
+   render->quad.coef = render->coef;
+   render->quad.posCoef = &render->posCoef;
+   */
+
+   return &render->stage;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_render.h b/src/gallium/drivers/cell/ppu/cell_render.h
new file mode 100644
index 0000000000..826dcbafeb
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_render.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef CELL_RENDER_H
+#define CELL_RENDER_H
+
+struct cell_context;
+struct draw_stage;
+
+extern void
+cell_flush_prim_buffer(struct cell_context *cell);
+
+extern struct draw_stage *cell_draw_render_stage( struct cell_context *cell );
+
+#endif /* CELL_RENDER_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c
new file mode 100644
index 0000000000..512d85d352
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_screen.c
@@ -0,0 +1,176 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "util/u_memory.h"
+#include "util/u_simple_screen.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+
+#include "cell/common.h"
+#include "cell_screen.h"
+#include "cell_texture.h"
+#include "cell_winsys.h"
+
+
+static const char *
+cell_get_vendor(struct pipe_screen *screen)
+{
+   return "Tungsten Graphics, Inc.";
+}
+
+
+static const char *
+cell_get_name(struct pipe_screen *screen)
+{
+   return "Cell";
+}
+
+
+static int
+cell_get_param(struct pipe_screen *screen, int param)
+{
+   switch (param) {
+   case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+      return CELL_MAX_SAMPLERS;
+   case PIPE_CAP_NPOT_TEXTURES:
+      return 1;
+   case PIPE_CAP_TWO_SIDED_STENCIL:
+      return 1;
+   case PIPE_CAP_GLSL:
+      return 1;
+   case PIPE_CAP_S3TC:
+      return 0;
+   case PIPE_CAP_ANISOTROPIC_FILTER:
+      return 0;
+   case PIPE_CAP_POINT_SPRITE:
+      return 1;
+   case PIPE_CAP_MAX_RENDER_TARGETS:
+      return 1;
+   case PIPE_CAP_OCCLUSION_QUERY:
+      return 1;
+   case PIPE_CAP_TEXTURE_SHADOW_MAP:
+      return 10;
+   case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+      return CELL_MAX_TEXTURE_LEVELS;
+   case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+      return 8;  /* max 128x128x128 */
+   case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+      return CELL_MAX_TEXTURE_LEVELS;
+   case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
+      return 1; /* XXX not really true */
+   case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
+      return 0; /* XXX to do */
+   default:
+      return 0;
+   }
+}
+
+
+static float
+cell_get_paramf(struct pipe_screen *screen, int param)
+{
+   switch (param) {
+   case PIPE_CAP_MAX_LINE_WIDTH:
+      /* fall-through */
+   case PIPE_CAP_MAX_LINE_WIDTH_AA:
+      return 255.0; /* arbitrary */
+
+   case PIPE_CAP_MAX_POINT_WIDTH:
+      /* fall-through */
+   case PIPE_CAP_MAX_POINT_WIDTH_AA:
+      return 255.0; /* arbitrary */
+
+   case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+      return 0.0;
+
+   case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+      return 16.0; /* arbitrary */
+
+   default:
+      return 0;
+   }
+}
+
+
+static boolean
+cell_is_format_supported( struct pipe_screen *screen,
+                          enum pipe_format format, 
+                          enum pipe_texture_target target,
+                          unsigned tex_usage, 
+                          unsigned geom_flags )
+{
+   /* cell supports most formats, XXX for now anyway */
+   if (format == PIPE_FORMAT_DXT5_RGBA ||
+       format == PIPE_FORMAT_R8G8B8A8_SRGB)
+      return FALSE;
+   else
+      return TRUE;
+}
+
+
+static void
+cell_destroy_screen( struct pipe_screen *screen )
+{
+   struct pipe_winsys *winsys = screen->winsys;
+
+   if(winsys->destroy)
+      winsys->destroy(winsys);
+
+   FREE(screen);
+}
+
+
+/**
+ * Create a new pipe_screen object
+ * Note: we're not presently subclassing pipe_screen (no cell_screen) but
+ * that would be the place to put SPU thread/context info...
+ */
+struct pipe_screen *
+cell_create_screen(struct pipe_winsys *winsys)
+{
+   struct pipe_screen *screen = CALLOC_STRUCT(pipe_screen);
+
+   if (!screen)
+      return NULL;
+
+   screen->winsys = winsys;
+
+   screen->destroy = cell_destroy_screen;
+
+   screen->get_name = cell_get_name;
+   screen->get_vendor = cell_get_vendor;
+   screen->get_param = cell_get_param;
+   screen->get_paramf = cell_get_paramf;
+   screen->is_format_supported = cell_is_format_supported;
+
+   cell_init_screen_texture_funcs(screen);
+   u_simple_screen_init(screen);
+
+   return screen;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_screen.h b/src/gallium/drivers/cell/ppu/cell_screen.h
new file mode 100644
index 0000000000..c7e15889d6
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_screen.h
@@ -0,0 +1,41 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef CELL_SCREEN_H
+#define CELL_SCREEN_H
+
+
+struct pipe_screen;
+struct pipe_winsys;
+
+
+extern struct pipe_screen *
+cell_create_screen(struct pipe_winsys *winsys);
+
+
+#endif /* CELL_SCREEN_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_spu.c b/src/gallium/drivers/cell/ppu/cell_spu.c
new file mode 100644
index 0000000000..28e5e6d706
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_spu.c
@@ -0,0 +1,219 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Utility/wrappers for communicating with the SPUs.
+ */
+
+
+#include <pthread.h>
+
+#include "cell_spu.h"
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+#include "util/u_memory.h"
+#include "cell/common.h"
+
+
+/*
+helpful headers:
+/opt/ibm/cell-sdk/prototype/src/include/ppu/cbe_mfc.h
+*/
+
+
+/**
+ * Cell/SPU info that's not per-context.
+ */
+struct cell_global_info cell_global;
+
+
+/**
+ * Scan /proc/cpuinfo to determine the timebase for the system.
+ * This is used by the SPUs to convert 'decrementer' ticks to seconds.
+ * There may be a better way to get this value...
+ */
+static unsigned
+get_timebase(void)
+{
+   FILE *f = fopen("/proc/cpuinfo", "r");
+   unsigned timebase;
+
+   assert(f);
+   while (!feof(f)) {
+      char line[80];
+      fgets(line, sizeof(line), f);
+      if (strncmp(line, "timebase", 8) == 0) {
+         char *colon = strchr(line, ':');
+         if (colon) {
+            timebase = atoi(colon + 2);
+            break;
+         }
+      }
+   }
+   fclose(f);
+
+   return timebase;
+}
+
+
+/**
+ * Write a 1-word message to the given SPE mailbox.
+ */
+void
+send_mbox_message(spe_context_ptr_t ctx, unsigned int msg)
+{
+   spe_in_mbox_write(ctx, &msg, 1, SPE_MBOX_ALL_BLOCKING);
+}
+
+
+/**
+ * Wait for a 1-word message to arrive in given mailbox.
+ */
+uint
+wait_mbox_message(spe_context_ptr_t ctx)
+{
+   do {
+      unsigned data;
+      int count = spe_out_mbox_read(ctx, &data, 1);
+
+      if (count == 1) {
+	 return data;
+      }
+      
+      if (count < 0) {
+	 /* error */ ;
+      }
+   } while (1);
+}
+
+
+/**
+ * Called by pthread_create() to spawn an SPU thread.
+ */
+static void *
+cell_thread_function(void *arg)
+{
+   struct cell_init_info *init = (struct cell_init_info *) arg;
+   unsigned entry = SPE_DEFAULT_ENTRY;
+
+   ASSERT_ALIGN16(init);
+
+   if (spe_context_run(cell_global.spe_contexts[init->id], &entry, 0,
+                       init, NULL, NULL) < 0) {
+      fprintf(stderr, "spe_context_run() failed\n");
+      exit(1);
+   }
+
+   pthread_exit(NULL);
+}
+
+
+/**
+ * Create the SPU threads.  This is done once during driver initialization.
+ * This involves setting the the "init" message which is sent to each SPU.
+ * The init message specifies an SPU id, total number of SPUs, location
+ * and number of batch buffers, etc.
+ */
+void
+cell_start_spus(struct cell_context *cell)
+{
+   static boolean one_time_init = FALSE;
+   uint i, j;
+   uint timebase = get_timebase();
+
+   if (one_time_init) {
+      fprintf(stderr, "PPU: Multiple rendering contexts not yet supported "
+	      "on Cell.\n");
+      abort();
+   }
+
+   one_time_init = TRUE;
+
+   assert(cell->num_spus <= CELL_MAX_SPUS);
+
+   ASSERT_ALIGN16(&cell_global.inits[0]);
+   ASSERT_ALIGN16(&cell_global.inits[1]);
+
+   /*
+    * Initialize the global 'inits' structure for each SPU.
+    * A pointer to the init struct will be passed to each SPU.
+    * The SPUs will then each grab their init info with mfc_get().
+    */
+   for (i = 0; i < cell->num_spus; i++) {
+      cell_global.inits[i].id = i;
+      cell_global.inits[i].num_spus = cell->num_spus;
+      cell_global.inits[i].debug_flags = cell->debug_flags;
+      cell_global.inits[i].inv_timebase = 1000.0f / timebase;
+
+      for (j = 0; j < CELL_NUM_BUFFERS; j++) {
+         cell_global.inits[i].buffers[j] = cell->buffer[j];
+      }
+      cell_global.inits[i].buffer_status = &cell->buffer_status[0][0][0];
+
+      cell_global.inits[i].spu_functions = &cell->spu_functions;
+
+      cell_global.spe_contexts[i] = spe_context_create(0, NULL);
+      if (!cell_global.spe_contexts[i]) {
+         fprintf(stderr, "spe_context_create() failed\n");
+         exit(1);
+      }
+
+      if (spe_program_load(cell_global.spe_contexts[i], &g3d_spu)) {
+         fprintf(stderr, "spe_program_load() failed\n");
+         exit(1);
+      }
+      
+      pthread_create(&cell_global.spe_threads[i], /* returned thread handle */
+                     NULL,                        /* pthread attribs */
+                     &cell_thread_function,       /* start routine */
+		     &cell_global.inits[i]);      /* thread argument */
+   }
+}
+
+
+/**
+ * Tell all the SPUs to stop/exit.
+ * This is done when the driver's exiting / cleaning up.
+ */
+void
+cell_spu_exit(struct cell_context *cell)
+{
+   uint i;
+
+   for (i = 0; i < cell->num_spus; i++) {
+      send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_EXIT);
+   }
+
+   /* wait for threads to exit */
+   for (i = 0; i < cell->num_spus; i++) {
+      void *value;
+      pthread_join(cell_global.spe_threads[i], &value);
+      cell_global.spe_threads[i] = 0;
+      cell_global.spe_contexts[i] = 0;
+   }
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_spu.h b/src/gallium/drivers/cell/ppu/cell_spu.h
new file mode 100644
index 0000000000..c93958a9ed
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_spu.h
@@ -0,0 +1,79 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef CELL_SPU
+#define CELL_SPU
+
+
+#include <libspe2.h>
+#include <pthread.h>
+#include "cell/common.h"
+
+#include "cell_context.h"
+
+
+/**
+ * Global vars, for now anyway.
+ */
+struct cell_global_info
+{
+   /**
+    * SPU/SPE handles, etc
+    */
+   spe_context_ptr_t spe_contexts[CELL_MAX_SPUS];
+   pthread_t spe_threads[CELL_MAX_SPUS];
+
+   /**
+    * Data sent to SPUs at start-up
+    */
+   struct cell_init_info inits[CELL_MAX_SPUS];
+};
+
+
+extern struct cell_global_info cell_global;
+
+
+/** This is the handle for the actual SPE code */
+extern spe_program_handle_t g3d_spu;
+
+
+extern void
+send_mbox_message(spe_context_ptr_t ctx, unsigned int msg);
+
+extern uint
+wait_mbox_message(spe_context_ptr_t ctx);
+
+
+extern void
+cell_start_spus(struct cell_context *cell);
+
+
+extern void
+cell_spu_exit(struct cell_context *cell);
+
+
+#endif /* CELL_SPU */
diff --git a/src/gallium/drivers/cell/ppu/cell_state.h b/src/gallium/drivers/cell/ppu/cell_state.h
new file mode 100644
index 0000000000..b193170f9c
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_state.h
@@ -0,0 +1,65 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef CELL_STATE_H
+#define CELL_STATE_H
+
+
+#define CELL_NEW_VIEWPORT      0x1
+#define CELL_NEW_RASTERIZER    0x2
+#define CELL_NEW_FS            0x4
+#define CELL_NEW_BLEND         0x8
+#define CELL_NEW_CLIP          0x10
+#define CELL_NEW_SCISSOR       0x20
+#define CELL_NEW_STIPPLE       0x40
+#define CELL_NEW_FRAMEBUFFER   0x80
+#define CELL_NEW_ALPHA_TEST    0x100
+#define CELL_NEW_DEPTH_STENCIL 0x200
+#define CELL_NEW_SAMPLER       0x400
+#define CELL_NEW_TEXTURE       0x800
+#define CELL_NEW_VERTEX        0x1000
+#define CELL_NEW_VS            0x2000
+#define CELL_NEW_VS_CONSTANTS  0x4000
+#define CELL_NEW_FS_CONSTANTS  0x8000
+#define CELL_NEW_VERTEX_INFO   0x10000
+
+
+extern void
+cell_update_derived( struct cell_context *softpipe );
+
+
+extern void
+cell_init_shader_functions(struct cell_context *cell);
+
+
+extern void
+cell_init_vertex_functions(struct cell_context *cell);
+
+
+#endif /* CELL_STATE_H */
+
diff --git a/src/gallium/drivers/cell/ppu/cell_state_derived.c b/src/gallium/drivers/cell/ppu/cell_state_derived.c
new file mode 100644
index 0000000000..efc4f78364
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_state_derived.c
@@ -0,0 +1,170 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vertex.h"
+#include "cell_context.h"
+#include "cell_batch.h"
+#include "cell_state.h"
+#include "cell_state_emit.h"
+
+
+/**
+ * Determine how to map vertex program outputs to fragment program inputs.
+ * Basically, this will be used when computing the triangle interpolation
+ * coefficients from the post-transform vertex attributes.
+ */
+static void
+calculate_vertex_layout( struct cell_context *cell )
+{
+   const struct cell_fragment_shader_state *fs = cell->fs;
+   const enum interp_mode colorInterp
+      = cell->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR;
+   struct vertex_info *vinfo = &cell->vertex_info;
+   uint i;
+   int src;
+
+#if 0
+   if (cell->vbuf) {
+      /* if using the post-transform vertex buffer, tell draw_vbuf to
+       * simply emit the whole post-xform vertex as-is:
+       */
+      struct vertex_info *vinfo_vbuf = &cell->vertex_info_vbuf;
+      vinfo_vbuf->num_attribs = 0;
+      draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0);
+      vinfo_vbuf->size = 4 * vs->num_outputs + sizeof(struct vertex_header)/4;
+   }
+#endif
+
+   /* reset vinfo */
+   vinfo->num_attribs = 0;
+
+   /* we always want to emit vertex pos */
+   src = draw_find_vs_output(cell->draw, TGSI_SEMANTIC_POSITION, 0);
+   assert(src >= 0);
+   draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src);
+
+
+   /*
+    * Loop over fragment shader inputs, searching for the matching output
+    * from the vertex shader.
+    */
+   for (i = 0; i < fs->info.num_inputs; i++) {
+      switch (fs->info.input_semantic_name[i]) {
+      case TGSI_SEMANTIC_POSITION:
+         /* already done above */
+         break;
+
+      case TGSI_SEMANTIC_COLOR:
+         src = draw_find_vs_output(cell->draw, TGSI_SEMANTIC_COLOR, 
+                                   fs->info.input_semantic_index[i]);
+         assert(src >= 0);
+         draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
+         break;
+
+      case TGSI_SEMANTIC_FOG:
+         src = draw_find_vs_output(cell->draw, TGSI_SEMANTIC_FOG, 0);
+#if 1
+         if (src < 0) /* XXX temp hack, try demos/fogcoord.c with this */
+            src = 0;
+#endif
+         assert(src >= 0);
+         draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_PERSPECTIVE, src);
+         break;
+
+      case TGSI_SEMANTIC_GENERIC:
+         /* this includes texcoords and varying vars */
+         src = draw_find_vs_output(cell->draw, TGSI_SEMANTIC_GENERIC,
+                              fs->info.input_semantic_index[i]);
+         assert(src >= 0);
+         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+         break;
+
+      default:
+         assert(0);
+      }
+   }
+
+   draw_compute_vertex_size(vinfo);
+
+   /* XXX only signal this if format really changes */
+   cell->dirty |= CELL_NEW_VERTEX_INFO;
+}
+
+
+#if 0
+/**
+ * Recompute cliprect from scissor bounds, scissor enable and surface size.
+ */
+static void
+compute_cliprect(struct cell_context *sp)
+{
+   uint surfWidth = sp->framebuffer.width;
+   uint surfHeight = sp->framebuffer.height;
+
+   if (sp->rasterizer->scissor) {
+      /* clip to scissor rect */
+      sp->cliprect.minx = MAX2(sp->scissor.minx, 0);
+      sp->cliprect.miny = MAX2(sp->scissor.miny, 0);
+      sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth);
+      sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight);
+   }
+   else {
+      /* clip to surface bounds */
+      sp->cliprect.minx = 0;
+      sp->cliprect.miny = 0;
+      sp->cliprect.maxx = surfWidth;
+      sp->cliprect.maxy = surfHeight;
+   }
+}
+#endif
+
+
+
+/**
+ * Update derived state, send current state to SPUs prior to rendering.
+ */
+void cell_update_derived( struct cell_context *cell )
+{
+   if (cell->dirty & (CELL_NEW_RASTERIZER |
+                      CELL_NEW_FS |
+                      CELL_NEW_VS))
+      calculate_vertex_layout( cell );
+
+#if 0
+   if (cell->dirty & (CELL_NEW_SCISSOR |
+                      CELL_NEW_DEPTH_STENCIL_ALPHA |
+                      CELL_NEW_FRAMEBUFFER))
+      compute_cliprect(cell);
+#endif
+
+   cell_emit_state(cell);
+
+   cell->dirty = 0;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c
new file mode 100644
index 0000000000..ff529fe22c
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c
@@ -0,0 +1,340 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+#include "cell_context.h"
+#include "cell_gen_fragment.h"
+#include "cell_state.h"
+#include "cell_state_emit.h"
+#include "cell_batch.h"
+#include "cell_texture.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+
+
+/**
+ * Find/create a cell_command_fragment_ops object corresponding to the
+ * current blend/stencil/z/colormask/etc. state.
+ */
+static struct cell_command_fragment_ops *
+lookup_fragment_ops(struct cell_context *cell)
+{
+   struct cell_fragment_ops_key key;
+   struct cell_command_fragment_ops *ops;
+
+   /*
+    * Build key
+    */
+   memset(&key, 0, sizeof(key));
+   key.blend = *cell->blend;
+   key.blend_color = cell->blend_color;
+   key.dsa = *cell->depth_stencil;
+
+   if (cell->framebuffer.cbufs[0])
+      key.color_format = cell->framebuffer.cbufs[0]->format;
+   else
+      key.color_format = PIPE_FORMAT_NONE;
+
+   if (cell->framebuffer.zsbuf)
+      key.zs_format = cell->framebuffer.zsbuf->format;
+   else
+      key.zs_format = PIPE_FORMAT_NONE;
+
+   /*
+    * Look up key in cache.
+    */
+   ops = (struct cell_command_fragment_ops *)
+      util_keymap_lookup(cell->fragment_ops_cache, &key);
+
+   /*
+    * If not found, create/save new fragment ops command.
+    */
+   if (!ops) {
+      struct spe_function spe_code_front, spe_code_back;
+      unsigned int facing_dependent, total_code_size;
+
+      if (0)
+         debug_printf("**** Create New Fragment Ops\n");
+
+      /* Prepare the buffer that will hold the generated code.  The
+       * "0" passed in for the size means that the SPE code will
+       * use a default size.
+       */
+      spe_init_func(&spe_code_front, 0);
+      spe_init_func(&spe_code_back, 0);
+
+      /* Generate new code.  Always generate new code for both front-facing
+       * and back-facing fragments, even if it's the same code in both
+       * cases.
+       */
+      cell_gen_fragment_function(cell, CELL_FACING_FRONT, &spe_code_front);
+      cell_gen_fragment_function(cell, CELL_FACING_BACK, &spe_code_back);
+
+      /* Make sure the code is a multiple of 8 bytes long; this is
+       * required to ensure that the dual pipe instruction alignment
+       * is correct.  It's also important for the SPU unpacking,
+       * which assumes 8-byte boundaries.
+       */
+      unsigned int front_code_size = spe_code_size(&spe_code_front);
+      while (front_code_size % 8 != 0) {
+         spe_lnop(&spe_code_front);
+         front_code_size = spe_code_size(&spe_code_front);
+      }
+      unsigned int back_code_size = spe_code_size(&spe_code_back);
+      while (back_code_size % 8 != 0) {
+         spe_lnop(&spe_code_back);
+         back_code_size = spe_code_size(&spe_code_back);
+      }
+
+      /* Determine whether the code we generated is facing-dependent, by
+       * determining whether the generated code is different for the front-
+       * and back-facing fragments.
+       */
+      if (front_code_size == back_code_size && memcmp(spe_code_front.store, spe_code_back.store, front_code_size) == 0) {
+         /* Code is identical; only need one copy. */
+         facing_dependent = 0;
+         total_code_size = front_code_size;
+      }
+      else {
+         /* Code is different for front-facing and back-facing fragments.
+          * Need to send both copies.
+          */
+         facing_dependent = 1;
+         total_code_size = front_code_size + back_code_size;
+      }
+
+      /* alloc new fragment ops command.  Note that this structure
+       * has variant length based on the total code size required.
+       */
+      ops = CALLOC_VARIANT_LENGTH_STRUCT(cell_command_fragment_ops, total_code_size);
+      /* populate the new cell_command_fragment_ops object */
+      ops->opcode[0] = CELL_CMD_STATE_FRAGMENT_OPS;
+      ops->total_code_size = total_code_size;
+      ops->front_code_index = 0;
+      memcpy(ops->code, spe_code_front.store, front_code_size);
+      if (facing_dependent) {
+        /* We have separate front- and back-facing code.  Append the
+         * back-facing code to the buffer.  Be careful because the code
+         * size is in bytes, but the buffer is of unsigned elements.
+         */
+        ops->back_code_index = front_code_size / sizeof(spe_code_front.store[0]);
+        memcpy(ops->code + ops->back_code_index, spe_code_back.store, back_code_size);
+      }
+      else {
+        /* Use the same code for front- and back-facing fragments */
+        ops->back_code_index = ops->front_code_index;
+      }
+
+      /* Set the fields for the fallback case.  Note that these fields
+       * (and the whole fallback case) will eventually go away.
+       */
+      ops->dsa = *cell->depth_stencil;
+      ops->blend = *cell->blend;
+      ops->blend_color = cell->blend_color;
+
+      /* insert cell_command_fragment_ops object into keymap/cache */
+      util_keymap_insert(cell->fragment_ops_cache, &key, ops, NULL);
+
+      /* release rtasm buffer */
+      spe_release_func(&spe_code_front);
+      spe_release_func(&spe_code_back);
+   }
+   else {
+      if (0)
+         debug_printf("**** Re-use Fragment Ops\n");
+   }
+
+   return ops;
+}
+
+
+
+static void
+emit_state_cmd(struct cell_context *cell, uint cmd,
+               const void *state, uint state_size)
+{
+   uint32_t *dst = (uint32_t *) 
+       cell_batch_alloc16(cell, ROUNDUP16(sizeof(opcode_t) + state_size));
+   *dst = cmd;
+   memcpy(dst + 4, state, state_size);
+}
+
+
+/**
+ * For state marked as 'dirty', construct a state-update command block
+ * and insert it into the current batch buffer.
+ */
+void
+cell_emit_state(struct cell_context *cell)
+{
+   if (cell->dirty & CELL_NEW_FRAMEBUFFER) {
+      struct pipe_surface *cbuf = cell->framebuffer.cbufs[0];
+      struct pipe_surface *zbuf = cell->framebuffer.zsbuf;
+      STATIC_ASSERT(sizeof(struct cell_command_framebuffer) % 16 == 0);
+      struct cell_command_framebuffer *fb
+         = cell_batch_alloc16(cell, sizeof(*fb));
+      fb->opcode[0] = CELL_CMD_STATE_FRAMEBUFFER;
+      fb->color_start = cell->cbuf_map[0];
+      fb->color_format = cbuf->format;
+      fb->depth_start = cell->zsbuf_map;
+      fb->depth_format = zbuf ? zbuf->format : PIPE_FORMAT_NONE;
+      fb->width = cell->framebuffer.width;
+      fb->height = cell->framebuffer.height;
+#if 0
+      printf("EMIT color format %s\n", pf_name(fb->color_format));
+      printf("EMIT depth format %s\n", pf_name(fb->depth_format));
+#endif
+   }
+
+   if (cell->dirty & (CELL_NEW_RASTERIZER)) {
+      STATIC_ASSERT(sizeof(struct cell_command_rasterizer) % 16 == 0);
+      struct cell_command_rasterizer *rast =
+         cell_batch_alloc16(cell, sizeof(*rast));
+      rast->opcode[0] = CELL_CMD_STATE_RASTERIZER;
+      rast->rasterizer = *cell->rasterizer;
+   }
+
+   if (cell->dirty & (CELL_NEW_FS)) {
+      /* Send new fragment program to SPUs */
+      STATIC_ASSERT(sizeof(struct cell_command_fragment_program) % 16 == 0);
+      struct cell_command_fragment_program *fp
+            = cell_batch_alloc16(cell, sizeof(*fp));
+      fp->opcode[0] = CELL_CMD_STATE_FRAGMENT_PROGRAM;
+      fp->num_inst = cell->fs->code.num_inst;
+      memcpy(&fp->code, cell->fs->code.store,
+             SPU_MAX_FRAGMENT_PROGRAM_INSTS * SPE_INST_SIZE);
+      if (0) {
+         int i;
+         printf("PPU Emit CELL_CMD_STATE_FRAGMENT_PROGRAM:\n");
+         for (i = 0; i < fp->num_inst; i++) {
+            printf(" %3d: 0x%08x\n", i, fp->code[i]);
+         }
+      }
+   }
+
+   if (cell->dirty & (CELL_NEW_FS_CONSTANTS)) {
+      const uint shader = PIPE_SHADER_FRAGMENT;
+      const uint num_const = cell->constants[shader].buffer->size / sizeof(float);
+      uint i, j;
+      float *buf = cell_batch_alloc16(cell, ROUNDUP16(32 + num_const * sizeof(float)));
+      uint32_t *ibuf = (uint32_t *) buf;
+      const float *constants = pipe_buffer_map(cell->pipe.screen,
+                                               cell->constants[shader].buffer,
+                                               PIPE_BUFFER_USAGE_CPU_READ);
+      ibuf[0] = CELL_CMD_STATE_FS_CONSTANTS;
+      ibuf[4] = num_const;
+      j = 8;
+      for (i = 0; i < num_const; i++) {
+         buf[j++] = constants[i];
+      }
+      pipe_buffer_unmap(cell->pipe.screen, cell->constants[shader].buffer);
+   }
+
+   if (cell->dirty & (CELL_NEW_FRAMEBUFFER |
+                      CELL_NEW_DEPTH_STENCIL |
+                      CELL_NEW_BLEND)) {
+      struct cell_command_fragment_ops *fops, *fops_cmd;
+      /* Note that cell_command_fragment_ops is a variant-sized record */
+      fops = lookup_fragment_ops(cell);
+      fops_cmd = cell_batch_alloc16(cell, ROUNDUP16(sizeof(*fops_cmd) + fops->total_code_size));
+      memcpy(fops_cmd, fops, sizeof(*fops) + fops->total_code_size);
+   }
+
+   if (cell->dirty & CELL_NEW_SAMPLER) {
+      uint i;
+      for (i = 0; i < CELL_MAX_SAMPLERS; i++) {
+         if (cell->dirty_samplers & (1 << i)) {
+            if (cell->sampler[i]) {
+               STATIC_ASSERT(sizeof(struct cell_command_sampler) % 16 == 0);
+               struct cell_command_sampler *sampler
+                  = cell_batch_alloc16(cell, sizeof(*sampler));
+               sampler->opcode[0] = CELL_CMD_STATE_SAMPLER;
+               sampler->unit = i;
+               sampler->state = *cell->sampler[i];
+            }
+         }
+      }
+      cell->dirty_samplers = 0x0;
+   }
+
+   if (cell->dirty & CELL_NEW_TEXTURE) {
+      uint i;
+      for (i = 0;i < CELL_MAX_SAMPLERS; i++) {
+         if (cell->dirty_textures & (1 << i)) {
+            STATIC_ASSERT(sizeof(struct cell_command_texture) % 16 == 0);
+            struct cell_command_texture *texture
+               =  (struct cell_command_texture *)cell_batch_alloc16(cell, sizeof(*texture));
+            texture->opcode[0] = CELL_CMD_STATE_TEXTURE;
+            texture->unit = i;
+            if (cell->texture[i]) {
+               uint level;
+               for (level = 0; level < CELL_MAX_TEXTURE_LEVELS; level++) {
+                  texture->start[level] = cell->texture[i]->tiled_mapped[level];
+                  texture->width[level] = cell->texture[i]->base.width[level];
+                  texture->height[level] = cell->texture[i]->base.height[level];
+                  texture->depth[level] = cell->texture[i]->base.depth[level];
+               }
+               texture->target = cell->texture[i]->base.target;
+            }
+            else {
+               uint level;
+               for (level = 0; level < CELL_MAX_TEXTURE_LEVELS; level++) {
+                  texture->start[level] = NULL;
+                  texture->width[level] = 0;
+                  texture->height[level] = 0;
+                  texture->depth[level] = 0;
+               }
+               texture->target = 0;
+            }
+         }
+      }
+      cell->dirty_textures = 0x0;
+   }
+
+   if (cell->dirty & CELL_NEW_VERTEX_INFO) {
+      emit_state_cmd(cell, CELL_CMD_STATE_VERTEX_INFO,
+                     &cell->vertex_info, sizeof(struct vertex_info));
+   }
+
+#if 0
+   if (cell->dirty & CELL_NEW_VS) {
+      const struct draw_context *const draw = cell->draw;
+      struct cell_shader_info info;
+
+      info.num_outputs = draw_num_vs_outputs(draw);
+      info.declarations = (uintptr_t) draw->vs.machine.Declarations;
+      info.num_declarations = draw->vs.machine.NumDeclarations;
+      info.instructions = (uintptr_t) draw->vs.machine.Instructions;
+      info.num_instructions = draw->vs.machine.NumInstructions;
+      info.immediates = (uintptr_t) draw->vs.machine.Imms;
+      info.num_immediates = draw->vs.machine.ImmLimit / 4;
+
+      emit_state_cmd(cell, CELL_CMD_STATE_BIND_VS, &info, sizeof(info));
+   }
+#endif
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.h b/src/gallium/drivers/cell/ppu/cell_state_emit.h
new file mode 100644
index 0000000000..59f8affe8d
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_state_emit.h
@@ -0,0 +1,36 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef CELL_STATE_EMIT_H
+#define CELL_STATE_EMIT_H
+
+
+extern void
+cell_emit_state(struct cell_context *cell);
+
+
+#endif /* CELL_STATE_EMIT_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c
new file mode 100644
index 0000000000..d97c22b2ef
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c
@@ -0,0 +1,1430 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file
+ * Generate code to perform all per-fragment operations.
+ *
+ * Code generated by these functions perform both alpha, depth, and stencil
+ * testing as well as alpha blending.
+ *
+ * \note
+ * Occlusion query is not supported, but this is the right place to add that
+ * support.
+ *
+ * \author Ian Romanick <idr@us.ibm.com>
+ */
+
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "cell_context.h"
+
+#include "rtasm/rtasm_ppc_spe.h"
+
+
+/**
+ * Generate code to perform alpha testing.
+ *
+ * The code generated by this function uses the register specificed by
+ * \c mask as both an input and an output.
+ *
+ * \param dsa    Current alpha-test state
+ * \param f      Function to which code should be appended
+ * \param mask   Index of register containing active fragment mask
+ * \param alphas Index of register containing per-fragment alpha values
+ *
+ * \note Emits a maximum of 6 instructions.
+ */
+static void
+emit_alpha_test(struct pipe_depth_stencil_alpha_state *dsa,
+                struct spe_function *f, int mask, int alphas)
+{
+   /* If the alpha function is either NEVER or ALWAYS, there is no need to
+    * load the reference value into a register.  ALWAYS is a fairly common
+    * case, and this optimization saves 2 instructions.
+    */
+   if (dsa->alpha.enabled
+       && (dsa->alpha.func != PIPE_FUNC_NEVER)
+       && (dsa->alpha.func != PIPE_FUNC_ALWAYS)) {
+      int ref = spe_allocate_available_register(f);
+      int tmp_a = spe_allocate_available_register(f);
+      int tmp_b = spe_allocate_available_register(f);
+      union {
+         float f;
+         unsigned u;
+      } ref_val;
+      boolean complement = FALSE;
+
+      ref_val.f = dsa->alpha.ref;
+
+      spe_il(f, ref, ref_val.u & 0x0000ffff);
+      spe_ilh(f, ref, ref_val.u >> 16);
+
+      switch (dsa->alpha.func) {
+      case PIPE_FUNC_NOTEQUAL:
+         complement = TRUE;
+         /* FALLTHROUGH */
+
+      case PIPE_FUNC_EQUAL:
+         spe_fceq(f, tmp_a, ref, alphas);
+         break;
+
+      case PIPE_FUNC_LEQUAL:
+         complement = TRUE;
+         /* FALLTHROUGH */
+
+      case PIPE_FUNC_GREATER:
+         spe_fcgt(f, tmp_a, ref, alphas);
+         break;
+
+      case PIPE_FUNC_LESS:
+         complement = TRUE;
+         /* FALLTHROUGH */
+
+      case PIPE_FUNC_GEQUAL:
+         spe_fcgt(f, tmp_a, ref, alphas);
+         spe_fceq(f, tmp_b, ref, alphas);
+         spe_or(f, tmp_a, tmp_b, tmp_a);
+         break;
+
+      case PIPE_FUNC_ALWAYS:
+      case PIPE_FUNC_NEVER:
+      default:
+         assert(0);
+         break;
+      }
+
+      if (complement) {
+         spe_andc(f, mask, mask, tmp_a);
+      } else {
+         spe_and(f, mask, mask, tmp_a);
+      }
+
+      spe_release_register(f, ref);
+      spe_release_register(f, tmp_a);
+      spe_release_register(f, tmp_b);
+   } else if (dsa->alpha.enabled && (dsa->alpha.func == PIPE_FUNC_NEVER)) {
+      spe_il(f, mask, 0);
+   }
+}
+
+
+/**
+ * Generate code to perform Z testing.  Four Z values are tested at once.
+ * \param dsa        Current depth-test state
+ * \param f          Function to which code should be appended
+ * \param mask       Index of register to contain depth-pass mask
+ * \param stored     Index of register containing values from depth buffer
+ * \param calculated Index of register containing per-fragment depth values
+ *
+ * \return
+ * If the calculated depth comparison mask is the actual mask, \c FALSE is
+ * returned.  If the calculated depth comparison mask is the compliment of
+ * the actual mask, \c TRUE is returned.
+ *
+ * \note Emits a maximum of 3 instructions.
+ */
+static boolean
+emit_depth_test(struct pipe_depth_stencil_alpha_state *dsa,
+                struct spe_function *f, int mask, int stored, int calculated)
+{
+   unsigned func = (dsa->depth.enabled)
+       ? dsa->depth.func : PIPE_FUNC_ALWAYS;
+   int tmp = spe_allocate_available_register(f);
+   boolean compliment = FALSE;
+
+   switch (func) {
+   case PIPE_FUNC_NEVER:
+      spe_il(f, mask, 0);
+      break;
+
+   case PIPE_FUNC_NOTEQUAL:
+      compliment = TRUE;
+      /* FALLTHROUGH */
+   case PIPE_FUNC_EQUAL:
+      spe_ceq(f, mask, calculated, stored);
+      break;
+
+   case PIPE_FUNC_LEQUAL:
+      compliment = TRUE;
+      /* FALLTHROUGH */
+   case PIPE_FUNC_GREATER:
+      spe_clgt(f, mask, calculated, stored);
+      break;
+
+   case PIPE_FUNC_LESS:
+      compliment = TRUE;
+      /* FALLTHROUGH */
+   case PIPE_FUNC_GEQUAL:
+      spe_clgt(f, mask, calculated, stored);
+      spe_ceq(f, tmp, calculated, stored);
+      spe_or(f, mask, mask, tmp);
+      break;
+
+   case PIPE_FUNC_ALWAYS:
+      spe_il(f, mask, ~0);
+      break;
+
+   default:
+      assert(0);
+      break;
+   }
+
+   spe_release_register(f, tmp);
+   return compliment;
+}
+
+
+/**
+ * Generate code to apply the stencil operation (after testing).
+ * \note Emits a maximum of 5 instructions.
+ *
+ * \warning
+ * Since \c out and \c in might be the same register, this routine cannot
+ * generate code that uses \c out as a temporary.
+ */
+static void
+emit_stencil_op(struct spe_function *f,
+                int out, int in, int mask, unsigned op, unsigned ref)
+{
+   const int clamp = spe_allocate_available_register(f);
+   const int clamp_mask = spe_allocate_available_register(f);
+   const int result = spe_allocate_available_register(f);
+
+   switch(op) {
+   case PIPE_STENCIL_OP_KEEP:
+      assert(0);
+   case PIPE_STENCIL_OP_ZERO:
+      spe_il(f, result, 0);
+      break;
+   case PIPE_STENCIL_OP_REPLACE:
+      spe_il(f, result, ref);
+      break;
+   case PIPE_STENCIL_OP_INCR:
+      /* clamp = [0xff, 0xff, 0xff, 0xff] */
+      spe_il(f, clamp, 0x0ff);
+      /* result[i] = in[i] + 1 */
+      spe_ai(f, result, in, 1);
+      /* clamp_mask[i] = (result[i] > 0xff) */
+      spe_clgti(f, clamp_mask, result, 0x0ff);
+      /* result[i] = clamp_mask[i] ? clamp[i] : result[i] */
+      spe_selb(f, result, result, clamp, clamp_mask);
+      break;
+   case PIPE_STENCIL_OP_DECR:
+      spe_il(f, clamp, 0);
+      spe_ai(f, result, in, -1);
+
+      /* If "(s-1) < 0" in signed arithemtic, then "(s-1) > MAX" in unsigned
+       * arithmetic.
+       */
+      spe_clgti(f, clamp_mask, result, 0x0ff);
+      spe_selb(f, result, result, clamp, clamp_mask);
+      break;
+   case PIPE_STENCIL_OP_INCR_WRAP:
+      spe_ai(f, result, in, 1);
+      break;
+   case PIPE_STENCIL_OP_DECR_WRAP:
+      spe_ai(f, result, in, -1);
+      break;
+   case PIPE_STENCIL_OP_INVERT:
+      spe_nor(f, result, in, in);
+      break;
+   default:
+      assert(0);
+   }
+
+   spe_selb(f, out, in, result, mask);
+
+   spe_release_register(f, result);
+   spe_release_register(f, clamp_mask);
+   spe_release_register(f, clamp);
+}
+
+
+/**
+ * Generate code to do stencil test.  Four pixels are tested at once.
+ * \param dsa        Depth / stencil test state
+ * \param face       0 for front face, 1 for back face
+ * \param f          Function to append instructions to
+ * \param mask       Register containing mask of fragments passing the
+ *                   alpha test
+ * \param depth_mask Register containing mask of fragments passing the
+ *                   depth test
+ * \param depth_compliment  Is \c depth_mask the compliment of the actual mask?
+ * \param stencil    Register containing values from stencil buffer
+ * \param depth_pass Register to store mask of fragments passing stencil test
+ *                   and depth test
+ *
+ * \note
+ * Emits a maximum of 10 + (3 * 5) = 25 instructions.
+ */
+static int
+emit_stencil_test(struct pipe_depth_stencil_alpha_state *dsa,
+                  unsigned face,
+                  struct spe_function *f,
+                  int mask,
+                  int depth_mask,
+                  boolean depth_complement,
+                  int stencil,
+                  int depth_pass)
+{
+   int stencil_fail = spe_allocate_available_register(f);
+   int depth_fail = spe_allocate_available_register(f);
+   int stencil_mask = spe_allocate_available_register(f);
+   int stencil_pass = spe_allocate_available_register(f);
+   int face_stencil = spe_allocate_available_register(f);
+   int stencil_src = stencil;
+   const unsigned ref = (dsa->stencil[face].ref_value
+                         & dsa->stencil[face].valuemask);
+   boolean complement = FALSE;
+   int stored;
+   int tmp = spe_allocate_available_register(f);
+
+
+   if ((dsa->stencil[face].func != PIPE_FUNC_NEVER)
+       && (dsa->stencil[face].func != PIPE_FUNC_ALWAYS)
+       && (dsa->stencil[face].valuemask != 0x0ff)) {
+      stored = spe_allocate_available_register(f);
+      spe_andi(f, stored, stencil, dsa->stencil[face].valuemask);
+   } else {
+      stored = stencil;
+   }
+
+
+   switch (dsa->stencil[face].func) {
+   case PIPE_FUNC_NEVER:
+      spe_il(f, stencil_mask, 0);   /* stencil_mask[0..3] = [0,0,0,0] */
+      break;
+
+   case PIPE_FUNC_NOTEQUAL:
+      complement = TRUE;
+      /* FALLTHROUGH */
+   case PIPE_FUNC_EQUAL:
+      /* stencil_mask[i] = (stored[i] == ref) */
+      spe_ceqi(f, stencil_mask, stored, ref);
+      break;
+
+   case PIPE_FUNC_LEQUAL:
+      complement = TRUE;
+      /* FALLTHROUGH */
+   case PIPE_FUNC_GREATER:
+      complement = TRUE;
+      /* stencil_mask[i] = (stored[i] > ref) */
+      spe_clgti(f, stencil_mask, stored, ref);
+      break;
+
+   case PIPE_FUNC_LESS:
+      complement = TRUE;
+      /* FALLTHROUGH */
+   case PIPE_FUNC_GEQUAL:
+      /* stencil_mask[i] = (stored[i] > ref) */
+      spe_clgti(f, stencil_mask, stored, ref);
+      /* tmp[i] = (stored[i] == ref) */
+      spe_ceqi(f, tmp, stored, ref);
+      /* stencil_mask[i] = stencil_mask[i] | tmp[i] */
+      spe_or(f, stencil_mask, stencil_mask, tmp);
+      break;
+
+   case PIPE_FUNC_ALWAYS:
+      /* See comment below. */
+      break;
+
+   default:
+      assert(0);
+      break;
+   }
+
+   if (stored != stencil) {
+      spe_release_register(f, stored);
+   }
+   spe_release_register(f, tmp);
+
+
+   /* ALWAYS is a very common stencil-test, so some effort is applied to
+    * optimize that case.  The stencil-pass mask is the same as the input
+    * fragment mask.  This makes the stencil-test (above) a no-op, and the
+    * input fragment mask can be "renamed" the stencil-pass mask.
+    */
+   if (dsa->stencil[face].func == PIPE_FUNC_ALWAYS) {
+      spe_release_register(f, stencil_pass);
+      stencil_pass = mask;
+   } else {
+      if (complement) {
+         spe_andc(f, stencil_pass, mask, stencil_mask);
+      } else {
+         spe_and(f, stencil_pass, mask, stencil_mask);
+      }
+   }
+
+   if (depth_complement) {
+      spe_andc(f, depth_pass, stencil_pass, depth_mask);
+   } else {
+      spe_and(f, depth_pass, stencil_pass, depth_mask);
+   }
+
+
+   /* Conditionally emit code to update the stencil value under various
+    * condititons.  Note that there is no need to generate code under the
+    * following circumstances:
+    *
+    * - Stencil write mask is zero.
+    * - For stencil-fail if the stencil test is ALWAYS
+    * - For depth-fail if the stencil test is NEVER
+    * - For depth-pass if the stencil test is NEVER
+    * - Any of the 3 conditions if the operation is KEEP
+    */
+   if (dsa->stencil[face].writemask != 0) {
+      if ((dsa->stencil[face].func != PIPE_FUNC_ALWAYS)
+          && (dsa->stencil[face].fail_op != PIPE_STENCIL_OP_KEEP)) {
+         if (complement) {
+            spe_and(f, stencil_fail, mask, stencil_mask);
+         } else {
+            spe_andc(f, stencil_fail, mask, stencil_mask);
+         }
+
+         emit_stencil_op(f, face_stencil, stencil_src, stencil_fail,
+                         dsa->stencil[face].fail_op,
+                         dsa->stencil[face].ref_value);
+
+         stencil_src = face_stencil;
+      }
+
+      if ((dsa->stencil[face].func != PIPE_FUNC_NEVER)
+          && (dsa->stencil[face].zfail_op != PIPE_STENCIL_OP_KEEP)) {
+         if (depth_complement) {
+            spe_and(f, depth_fail, stencil_pass, depth_mask);
+         } else {
+            spe_andc(f, depth_fail, stencil_pass, depth_mask);
+         }
+
+         emit_stencil_op(f, face_stencil, stencil_src, depth_fail,
+                         dsa->stencil[face].zfail_op,
+                         dsa->stencil[face].ref_value);
+         stencil_src = face_stencil;
+      }
+
+      if ((dsa->stencil[face].func != PIPE_FUNC_NEVER)
+          && (dsa->stencil[face].zpass_op != PIPE_STENCIL_OP_KEEP)) {
+         emit_stencil_op(f, face_stencil, stencil_src, depth_pass,
+                         dsa->stencil[face].zpass_op,
+                         dsa->stencil[face].ref_value);
+         stencil_src = face_stencil;
+      }
+   }
+
+   spe_release_register(f, stencil_fail);
+   spe_release_register(f, depth_fail);
+   spe_release_register(f, stencil_mask);
+   if (stencil_pass != mask) {
+      spe_release_register(f, stencil_pass);
+   }
+
+   /* If all of the stencil operations were KEEP or the stencil write mask was
+    * zero, "stencil_src" will still be set to "stencil".  In this case
+    * release the "face_stencil" register.  Otherwise apply the stencil write
+    * mask to select bits from the calculated stencil value and the previous
+    * stencil value.
+    */
+   if (stencil_src == stencil) {
+      spe_release_register(f, face_stencil);
+   } else if (dsa->stencil[face].writemask != 0x0ff) {
+      int tmp = spe_allocate_available_register(f);
+
+      spe_il(f, tmp, dsa->stencil[face].writemask);
+      spe_selb(f, stencil_src, stencil, stencil_src, tmp);
+
+      spe_release_register(f, tmp);
+   }
+
+   return stencil_src;
+}
+
+
+void
+cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa)
+{
+   struct pipe_depth_stencil_alpha_state *const dsa = &cdsa->base;
+   struct spe_function *const f = &cdsa->code;
+
+   /* This code generates a maximum of 6 (alpha test) + 3 (depth test)
+    * + 25 (front stencil) + 25 (back stencil) + 4 = 63 instructions.  Round
+    * up to 64 to make it a happy power-of-two.
+    */
+   spe_init_func(f, SPE_INST_SIZE * 64);
+
+
+   /* Allocate registers for the function's input parameters.  Cleverly (and
+    * clever code is usually dangerous, but I couldn't resist) the generated
+    * function returns a structure.  Returned structures start with register
+    * 3, and the structure fields are ordered to match up exactly with the
+    * input parameters.
+    */
+   int mask = spe_allocate_register(f, 3);
+   int depth = spe_allocate_register(f, 4);
+   int stencil = spe_allocate_register(f, 5);
+   int zvals = spe_allocate_register(f, 6);
+   int frag_a = spe_allocate_register(f, 7);
+   int facing = spe_allocate_register(f, 8);
+
+   int depth_mask = spe_allocate_available_register(f);
+
+   boolean depth_complement;
+
+
+   emit_alpha_test(dsa, f, mask, frag_a);
+
+   depth_complement = emit_depth_test(dsa, f, depth_mask, depth, zvals);
+
+   if (dsa->stencil[0].enabled) {
+      const int front_depth_pass = spe_allocate_available_register(f);
+      int front_stencil = emit_stencil_test(dsa, 0, f, mask,
+                                            depth_mask, depth_complement,
+                                            stencil, front_depth_pass);
+
+      if (dsa->stencil[1].enabled) {
+         const int back_depth_pass = spe_allocate_available_register(f);
+         int back_stencil = emit_stencil_test(dsa, 1, f, mask,
+                                              depth_mask,  depth_complement,
+                                              stencil, back_depth_pass);
+
+         /* If the front facing stencil value and the back facing stencil
+          * value are stored in the same register, there is no need to select
+          * a value based on the facing.  This can happen if the stencil value
+          * was not modified due to the write masks being zero, the stencil
+          * operations being KEEP, etc.
+          */
+         if (front_stencil != back_stencil) {
+            spe_selb(f, stencil, back_stencil, front_stencil, facing);
+         }
+
+         if (back_stencil != stencil) {
+            spe_release_register(f, back_stencil);
+         }
+
+         if (front_stencil != stencil) {
+            spe_release_register(f, front_stencil);
+         }
+
+         spe_selb(f, mask, back_depth_pass, front_depth_pass, facing);
+
+         spe_release_register(f, back_depth_pass);
+      } else {
+         if (front_stencil != stencil) {
+            spe_or(f, stencil, front_stencil, front_stencil);
+            spe_release_register(f, front_stencil);
+         }
+         spe_or(f, mask, front_depth_pass, front_depth_pass);
+      }
+
+      spe_release_register(f, front_depth_pass);
+   } else if (dsa->depth.enabled) {
+      if (depth_complement) {
+         spe_andc(f, mask, mask, depth_mask);
+      } else {
+         spe_and(f, mask, mask, depth_mask);
+      }
+   }
+
+   if (dsa->depth.writemask) {
+         spe_selb(f, depth, depth, zvals, mask);
+   }
+
+   spe_bi(f, 0, 0, 0);  /* return from function call */
+
+
+#if 0
+   {
+      const uint32_t *p = f->store;
+      unsigned i;
+
+      printf("# alpha (%sabled)\n",
+             (dsa->alpha.enabled) ? "en" : "dis");
+      printf("#    func: %u\n", dsa->alpha.func);
+      printf("#    ref: %.2f\n", dsa->alpha.ref);
+
+      printf("# depth (%sabled)\n",
+             (dsa->depth.enabled) ? "en" : "dis");
+      printf("#    func: %u\n", dsa->depth.func);
+
+      for (i = 0; i < 2; i++) {
+         printf("# %s stencil (%sabled)\n",
+                (i == 0) ? "front" : "back",
+                (dsa->stencil[i].enabled) ? "en" : "dis");
+
+         printf("#    func: %u\n", dsa->stencil[i].func);
+         printf("#    op (sf, zf, zp): %u %u %u\n",
+                dsa->stencil[i].fail_op,
+                dsa->stencil[i].zfail_op,
+                dsa->stencil[i].zpass_op);
+         printf("#    ref value / value mask / write mask: %02x %02x %02x\n",
+                dsa->stencil[i].ref_value,
+                dsa->stencil[i].valuemask,
+                dsa->stencil[i].writemask);
+      }
+
+      printf("\t.text\n");
+      for (/* empty */; p < f->csr; p++) {
+         printf("\t.long\t0x%04x\n", *p);
+      }
+      fflush(stdout);
+   }
+#endif
+}
+
+
+/**
+ * \note Emits a maximum of 3 instructions
+ */
+static int
+emit_alpha_factor_calculation(struct spe_function *f,
+                              unsigned factor,
+                              int src_alpha, int dst_alpha, int const_alpha)
+{
+   int factor_reg;
+   int tmp;
+
+
+   switch (factor) {
+   case PIPE_BLENDFACTOR_ONE:
+      factor_reg = -1;
+      break;
+
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      factor_reg = spe_allocate_available_register(f);
+
+      spe_or(f, factor_reg, src_alpha, src_alpha);
+      break;
+
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      factor_reg = dst_alpha;
+      break;
+
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+      factor_reg = -1;
+      break;
+
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+      factor_reg = spe_allocate_available_register(f);
+
+      tmp = spe_allocate_available_register(f);
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+      spe_fs(f, factor_reg, tmp, const_alpha);
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      factor_reg = const_alpha;
+      break;
+
+   case PIPE_BLENDFACTOR_ZERO:
+      factor_reg = -1;
+      break;
+
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+      tmp = spe_allocate_available_register(f);
+      factor_reg = spe_allocate_available_register(f);
+
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+      spe_fs(f, factor_reg, tmp, src_alpha);
+
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      tmp = spe_allocate_available_register(f);
+      factor_reg = spe_allocate_available_register(f);
+
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+      spe_fs(f, factor_reg, tmp, dst_alpha);
+
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+   default:
+      assert(0);
+      factor_reg = -1;
+      break;
+   }
+
+   return factor_reg;
+}
+
+
+/**
+ * \note Emits a maximum of 6 instructions
+ */
+static void
+emit_color_factor_calculation(struct spe_function *f,
+                              unsigned sF, unsigned mask,
+                              const int *src,
+                              const int *dst,
+                              const int *const_color,
+                              int *factor)
+{
+   int tmp;
+   unsigned i;
+
+
+   factor[0] = -1;
+   factor[1] = -1;
+   factor[2] = -1;
+   factor[3] = -1;
+
+   switch (sF) {
+   case PIPE_BLENDFACTOR_ONE:
+      break;
+
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+      for (i = 0; i < 3; ++i) {
+         if ((mask & (1U << i)) != 0) {
+            factor[i] = spe_allocate_available_register(f);
+            spe_or(f, factor[i], src[i], src[i]);
+         }
+      }
+      break;
+
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      factor[0] = spe_allocate_available_register(f);
+      factor[1] = factor[0];
+      factor[2] = factor[0];
+
+      spe_or(f, factor[0], src[3], src[3]);
+      break;
+
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      factor[0] = dst[3];
+      factor[1] = dst[3];
+      factor[2] = dst[3];
+      break;
+
+   case PIPE_BLENDFACTOR_DST_COLOR:
+      factor[0] = dst[0];
+      factor[1] = dst[1];
+      factor[2] = dst[2];
+      break;
+
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+      tmp = spe_allocate_available_register(f);
+      factor[0] = spe_allocate_available_register(f);
+      factor[1] = factor[0];
+      factor[2] = factor[0];
+
+      /* Alpha saturate means min(As, 1-Ad).
+       */
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+      spe_fs(f, tmp, tmp, dst[3]);
+      spe_fcgt(f, factor[0], tmp, src[3]);
+      spe_selb(f, factor[0], src[3], tmp, factor[0]);
+
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+      tmp = spe_allocate_available_register(f);
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+
+      for (i = 0; i < 3; i++) {
+         factor[i] = spe_allocate_available_register(f);
+
+         spe_fs(f, factor[i], tmp, const_color[i]);
+      }
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+      for (i = 0; i < 3; i++) {
+         factor[i] = const_color[i];
+      }
+      break;
+
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+      factor[0] = spe_allocate_available_register(f);
+      factor[1] = factor[0];
+      factor[2] = factor[0];
+
+      tmp = spe_allocate_available_register(f);
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+      spe_fs(f, factor[0], tmp, const_color[3]);
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      factor[0] = const_color[3];
+      factor[1] = factor[0];
+      factor[2] = factor[0];
+      break;
+
+   case PIPE_BLENDFACTOR_ZERO:
+      break;
+
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+      tmp = spe_allocate_available_register(f);
+
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+
+      for (i = 0; i < 3; ++i) {
+         if ((mask & (1U << i)) != 0) {
+            factor[i] = spe_allocate_available_register(f);
+            spe_fs(f, factor[i], tmp, src[i]);
+         }
+      }
+
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+      tmp = spe_allocate_available_register(f);
+      factor[0] = spe_allocate_available_register(f);
+      factor[1] = factor[0];
+      factor[2] = factor[0];
+
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+      spe_fs(f, factor[0], tmp, src[3]);
+
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      tmp = spe_allocate_available_register(f);
+      factor[0] = spe_allocate_available_register(f);
+      factor[1] = factor[0];
+      factor[2] = factor[0];
+
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+      spe_fs(f, factor[0], tmp, dst[3]);
+
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+      tmp = spe_allocate_available_register(f);
+
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+
+      for (i = 0; i < 3; ++i) {
+         if ((mask & (1U << i)) != 0) {
+            factor[i] = spe_allocate_available_register(f);
+            spe_fs(f, factor[i], tmp, dst[i]);
+         }
+      }
+
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_SRC1_ALPHA:
+   case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+   case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+   default:
+      assert(0);
+   }
+}
+
+
+static void
+emit_blend_calculation(struct spe_function *f,
+                       unsigned func, unsigned sF, unsigned dF,
+                       int src, int src_factor, int dst, int dst_factor)
+{
+   int tmp = spe_allocate_available_register(f);
+
+   switch (func) {
+   case PIPE_BLEND_ADD:
+      if (sF == PIPE_BLENDFACTOR_ONE) {
+         if (dF == PIPE_BLENDFACTOR_ZERO) {
+            /* Do nothing. */
+         } else if (dF == PIPE_BLENDFACTOR_ONE) {
+            spe_fa(f, src, src, dst);
+         }
+      } else if (sF == PIPE_BLENDFACTOR_ZERO) {
+         if (dF == PIPE_BLENDFACTOR_ZERO) {
+            spe_il(f, src, 0);
+         } else if (dF == PIPE_BLENDFACTOR_ONE) {
+            spe_or(f, src, dst, dst);
+         } else {
+            spe_fm(f, src, dst, dst_factor);
+         }
+      } else if (dF == PIPE_BLENDFACTOR_ZERO) {
+         spe_fm(f, src, src, src_factor);
+      } else {
+         spe_fm(f, tmp, dst, dst_factor);
+         spe_fma(f, src, src, src_factor, tmp);
+      }
+      break;
+
+   case PIPE_BLEND_SUBTRACT:
+      if (sF == PIPE_BLENDFACTOR_ONE) {
+         if (dF == PIPE_BLENDFACTOR_ZERO) {
+            /* Do nothing. */
+         } else if (dF == PIPE_BLENDFACTOR_ONE) {
+            spe_fs(f, src, src, dst);
+         }
+      } else if (sF == PIPE_BLENDFACTOR_ZERO) {
+         if (dF == PIPE_BLENDFACTOR_ZERO) {
+            spe_il(f, src, 0);
+         } else if (dF == PIPE_BLENDFACTOR_ONE) {
+            spe_il(f, tmp, 0);
+            spe_fs(f, src, tmp, dst);
+         } else {
+            spe_fm(f, src, dst, dst_factor);
+         }
+      } else if (dF == PIPE_BLENDFACTOR_ZERO) {
+         spe_fm(f, src, src, src_factor);
+      } else {
+         spe_fm(f, tmp, dst, dst_factor);
+         spe_fms(f, src, src, src_factor, tmp);
+      }
+      break;
+
+   case PIPE_BLEND_REVERSE_SUBTRACT:
+      if (sF == PIPE_BLENDFACTOR_ONE) {
+         if (dF == PIPE_BLENDFACTOR_ZERO) {
+            spe_il(f, tmp, 0);
+            spe_fs(f, src, tmp, src);
+         } else if (dF == PIPE_BLENDFACTOR_ONE) {
+            spe_fs(f, src, dst, src);
+         }
+      } else if (sF == PIPE_BLENDFACTOR_ZERO) {
+         if (dF == PIPE_BLENDFACTOR_ZERO) {
+            spe_il(f, src, 0);
+         } else if (dF == PIPE_BLENDFACTOR_ONE) {
+            spe_or(f, src, dst, dst);
+         } else {
+            spe_fm(f, src, dst, dst_factor);
+         }
+      } else if (dF == PIPE_BLENDFACTOR_ZERO) {
+         spe_fm(f, src, src, src_factor);
+      } else {
+         spe_fm(f, tmp, src, src_factor);
+         spe_fms(f, src, src, dst_factor, tmp);
+      }
+      break;
+
+   case PIPE_BLEND_MIN:
+      spe_cgt(f, tmp, src, dst);
+      spe_selb(f, src, src, dst, tmp);
+      break;
+
+   case PIPE_BLEND_MAX:
+      spe_cgt(f, tmp, src, dst);
+      spe_selb(f, src, dst, src, tmp);
+      break;
+
+   default:
+      assert(0);
+   }
+
+   spe_release_register(f, tmp);
+}
+
+
+/**
+ * Generate code to perform alpha blending on the SPE
+ */
+void
+cell_generate_alpha_blend(struct cell_blend_state *cb)
+{
+   struct pipe_blend_state *const b = &cb->base;
+   struct spe_function *const f = &cb->code;
+
+   /* This code generates a maximum of 3 (source alpha factor)
+    * + 3 (destination alpha factor) + (3 * 6) (source color factor)
+    * + (3 * 6) (destination color factor) + (4 * 2) (blend equation)
+    * + 4 (fragment mask) + 1 (return) = 55 instlructions.  Round up to 64 to
+    * make it a happy power-of-two.
+    */
+   spe_init_func(f, SPE_INST_SIZE * 64);
+
+
+   const int frag[4] = {
+      spe_allocate_register(f, 3),
+      spe_allocate_register(f, 4),
+      spe_allocate_register(f, 5),
+      spe_allocate_register(f, 6),
+   };
+   const int pixel[4] = {
+      spe_allocate_register(f, 7),
+      spe_allocate_register(f, 8),
+      spe_allocate_register(f, 9),
+      spe_allocate_register(f, 10),
+   };
+   const int const_color[4] = {
+      spe_allocate_register(f, 11),
+      spe_allocate_register(f, 12),
+      spe_allocate_register(f, 13),
+      spe_allocate_register(f, 14),
+   };
+   unsigned func[4];
+   unsigned sF[4];
+   unsigned dF[4];
+   unsigned i;
+   int src_factor[4];
+   int dst_factor[4];
+
+
+   /* Does the selected blend mode make use of the source / destination
+    * color (RGB) blend factors?
+    */
+   boolean need_color_factor = b->blend_enable
+       && (b->rgb_func != PIPE_BLEND_MIN)
+       && (b->rgb_func != PIPE_BLEND_MAX);
+
+   /* Does the selected blend mode make use of the source / destination
+    * alpha blend factors?
+    */
+   boolean need_alpha_factor = b->blend_enable
+       && (b->alpha_func != PIPE_BLEND_MIN)
+       && (b->alpha_func != PIPE_BLEND_MAX);
+
+
+   if (b->blend_enable) {
+      sF[0] = b->rgb_src_factor;
+      sF[1] = sF[0];
+      sF[2] = sF[0];
+      switch (b->alpha_src_factor & 0x0f) {
+      case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+         sF[3] = PIPE_BLENDFACTOR_ONE;
+         break;
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+      case PIPE_BLENDFACTOR_DST_COLOR:
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+      case PIPE_BLENDFACTOR_SRC1_COLOR:
+         sF[3] = b->alpha_src_factor + 1;
+         break;
+      default:
+         sF[3] = b->alpha_src_factor;
+      }
+
+      dF[0] = b->rgb_dst_factor;
+      dF[1] = dF[0];
+      dF[2] = dF[0];
+      switch (b->alpha_dst_factor & 0x0f) {
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+      case PIPE_BLENDFACTOR_DST_COLOR:
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+      case PIPE_BLENDFACTOR_SRC1_COLOR:
+         dF[3] = b->alpha_dst_factor + 1;
+         break;
+      default:
+         dF[3] = b->alpha_dst_factor;
+      }
+
+      func[0] = b->rgb_func;
+      func[1] = func[0];
+      func[2] = func[0];
+      func[3] = b->alpha_func;
+   } else {
+      sF[0] = PIPE_BLENDFACTOR_ONE;
+      sF[1] = PIPE_BLENDFACTOR_ONE;
+      sF[2] = PIPE_BLENDFACTOR_ONE;
+      sF[3] = PIPE_BLENDFACTOR_ONE;
+      dF[0] = PIPE_BLENDFACTOR_ZERO;
+      dF[1] = PIPE_BLENDFACTOR_ZERO;
+      dF[2] = PIPE_BLENDFACTOR_ZERO;
+      dF[3] = PIPE_BLENDFACTOR_ZERO;
+
+      func[0] = PIPE_BLEND_ADD;
+      func[1] = PIPE_BLEND_ADD;
+      func[2] = PIPE_BLEND_ADD;
+      func[3] = PIPE_BLEND_ADD;
+   }
+
+
+   /* If alpha writing is enabled and the alpha blend mode requires use of
+    * the alpha factor, calculate the alpha factor.
+    */
+   if (((b->colormask & 8) != 0) && need_alpha_factor) {
+      src_factor[3] = emit_alpha_factor_calculation(f, sF[3], const_color[3],
+                                                    frag[3], pixel[3]);
+
+      /* If the alpha destination blend factor is the same as the alpha source
+       * blend factor, re-use the previously calculated value.
+       */
+      dst_factor[3] = (dF[3] == sF[3])
+          ? src_factor[3]
+          : emit_alpha_factor_calculation(f, dF[3], const_color[3],
+                                          frag[3], pixel[3]);
+   }
+
+
+   if (sF[0] == sF[3]) {
+      src_factor[0] = src_factor[3];
+      src_factor[1] = src_factor[3];
+      src_factor[2] = src_factor[3];
+   } else if (sF[0] == dF[3]) {
+      src_factor[0] = dst_factor[3];
+      src_factor[1] = dst_factor[3];
+      src_factor[2] = dst_factor[3];
+   } else if (need_color_factor) {
+      emit_color_factor_calculation(f,
+                                    b->rgb_src_factor,
+                                    b->colormask,
+                                    frag, pixel, const_color, src_factor);
+   }
+
+
+   if (dF[0] == sF[3]) {
+      dst_factor[0] = src_factor[3];
+      dst_factor[1] = src_factor[3];
+      dst_factor[2] = src_factor[3];
+   } else if (dF[0] == dF[3]) {
+      dst_factor[0] = dst_factor[3];
+      dst_factor[1] = dst_factor[3];
+      dst_factor[2] = dst_factor[3];
+   } else if (dF[0] == sF[0]) {
+      dst_factor[0] = src_factor[0];
+      dst_factor[1] = src_factor[1];
+      dst_factor[2] = src_factor[2];
+   } else if (need_color_factor) {
+      emit_color_factor_calculation(f,
+                                    b->rgb_dst_factor,
+                                    b->colormask,
+                                    frag, pixel, const_color, dst_factor);
+   }
+
+
+
+   for (i = 0; i < 4; ++i) {
+      if ((b->colormask & (1U << i)) != 0) {
+         emit_blend_calculation(f,
+                                func[i], sF[i], dF[i],
+                                frag[i], src_factor[i],
+                                pixel[i], dst_factor[i]);
+      }
+   }
+
+   spe_bi(f, 0, 0, 0);
+
+#if 0
+   {
+      const uint32_t *p = f->store;
+
+      printf("# %u instructions\n", f->csr - f->store);
+      printf("# blend (%sabled)\n",
+             (cb->base.blend_enable) ? "en" : "dis");
+      printf("#    RGB func / sf / df: %u %u %u\n",
+             cb->base.rgb_func,
+             cb->base.rgb_src_factor,
+             cb->base.rgb_dst_factor);
+      printf("#    ALP func / sf / df: %u %u %u\n",
+             cb->base.alpha_func,
+             cb->base.alpha_src_factor,
+             cb->base.alpha_dst_factor);
+
+      printf("\t.text\n");
+      for (/* empty */; p < f->csr; p++) {
+         printf("\t.long\t0x%04x\n", *p);
+      }
+      fflush(stdout);
+   }
+#endif
+}
+
+
+static int
+PC_OFFSET(const struct spe_function *f, const void *d)
+{
+   const intptr_t pc = (intptr_t) &f->store[f->num_inst];
+   const intptr_t ea = ~0x0f & (intptr_t) d;
+
+   return (ea - pc) >> 2;
+}
+
+
+/**
+ * Generate code to perform color conversion and logic op
+ *
+ * \bug
+ * The code generated by this function should also perform dithering.
+ *
+ * \bug
+ * The code generated by this function should also perform color-write
+ * masking.
+ *
+ * \bug
+ * Only two framebuffer formats are supported at this time.
+ */
+void
+cell_generate_logic_op(struct spe_function *f,
+                       const struct pipe_blend_state *blend,
+                       struct pipe_surface *surf)
+{
+   const unsigned logic_op = (blend->logicop_enable)
+       ? blend->logicop_func : PIPE_LOGICOP_COPY;
+
+   /* This code generates a maximum of 37 instructions.  An additional 32
+    * bytes (equiv. to 8 instructions) are needed for data storage.  Round up
+    * to 64 to make it a happy power-of-two.
+    */
+   spe_init_func(f, SPE_INST_SIZE * 64);
+
+
+   /* Pixel colors in framebuffer format in AoS layout.
+    */
+   const int pixel[4] = {
+      spe_allocate_register(f, 3),
+      spe_allocate_register(f, 4),
+      spe_allocate_register(f, 5),
+      spe_allocate_register(f, 6),
+   };
+
+   /* Fragment colors stored as floats in SoA layout.
+    */
+   const int frag[4] = {
+      spe_allocate_register(f, 7),
+      spe_allocate_register(f, 8),
+      spe_allocate_register(f, 9),
+      spe_allocate_register(f, 10),
+   };
+
+   const int mask = spe_allocate_register(f, 11);
+
+
+   /* Short-circuit the noop and invert cases.
+    */
+   if ((logic_op == PIPE_LOGICOP_NOOP) || (blend->colormask == 0)) {
+      spe_bi(f, 0, 0, 0);
+      return;
+   } else if (logic_op == PIPE_LOGICOP_INVERT) {
+      spe_nor(f, pixel[0], pixel[0], pixel[0]);
+      spe_nor(f, pixel[1], pixel[1], pixel[1]);
+      spe_nor(f, pixel[2], pixel[2], pixel[2]);
+      spe_nor(f, pixel[3], pixel[3], pixel[3]);
+      spe_bi(f, 0, 0, 0);
+      return;
+   }
+
+
+   const int tmp[4] = {
+      spe_allocate_available_register(f),
+      spe_allocate_available_register(f),
+      spe_allocate_available_register(f),
+      spe_allocate_available_register(f),
+   };
+
+   const int shuf_xpose_hi = spe_allocate_available_register(f);
+   const int shuf_xpose_lo = spe_allocate_available_register(f);
+   const int shuf_color = spe_allocate_available_register(f);
+
+
+   /* Pointer to the begining of the function's private data area.
+    */
+   uint32_t *const data = ((uint32_t *) f->store) + (64 - 8);
+
+
+   /* Convert fragment colors to framebuffer format in AoS layout.
+    */
+   switch (surf->format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      data[0] = 0x00010203;
+      data[1] = 0x10111213;
+      data[2] = 0x04050607;
+      data[3] = 0x14151617;
+      data[4] = 0x0c000408;
+      data[5] = 0x80808080;
+      data[6] = 0x80808080;
+      data[7] = 0x80808080;
+      break;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      data[0] = 0x03020100;
+      data[1] = 0x13121110;
+      data[2] = 0x07060504;
+      data[3] = 0x17161514;
+      data[4] = 0x0804000c;
+      data[5] = 0x80808080;
+      data[6] = 0x80808080;
+      data[7] = 0x80808080;
+      break;
+   default:
+      fprintf(stderr, "CELL: Bad pixel format in cell_generate_logic_op()");
+      ASSERT(0);
+   }
+
+   spe_ilh(f, tmp[0], 0x0808);
+   spe_lqr(f, shuf_xpose_hi, PC_OFFSET(f, data+0));
+   spe_lqr(f, shuf_color, PC_OFFSET(f, data+4));
+   spe_a(f, shuf_xpose_lo, shuf_xpose_hi, tmp[0]);
+
+   spe_shufb(f, tmp[0], frag[0], frag[2], shuf_xpose_hi);
+   spe_shufb(f, tmp[1], frag[0], frag[2], shuf_xpose_lo);
+   spe_shufb(f, tmp[2], frag[1], frag[3], shuf_xpose_hi);
+   spe_shufb(f, tmp[3], frag[1], frag[3], shuf_xpose_lo);
+
+   spe_shufb(f, frag[0], tmp[0], tmp[2], shuf_xpose_hi);
+   spe_shufb(f, frag[1], tmp[0], tmp[2], shuf_xpose_lo);
+   spe_shufb(f, frag[2], tmp[1], tmp[3], shuf_xpose_hi);
+   spe_shufb(f, frag[3], tmp[1], tmp[3], shuf_xpose_lo);
+
+   spe_cfltu(f, frag[0], frag[0], 32);
+   spe_cfltu(f, frag[1], frag[1], 32);
+   spe_cfltu(f, frag[2], frag[2], 32);
+   spe_cfltu(f, frag[3], frag[3], 32);
+
+   spe_shufb(f, frag[0], frag[0], pixel[0], shuf_color);
+   spe_shufb(f, frag[1], frag[1], pixel[1], shuf_color);
+   spe_shufb(f, frag[2], frag[2], pixel[2], shuf_color);
+   spe_shufb(f, frag[3], frag[3], pixel[3], shuf_color);
+
+
+   /* If logic op is enabled, perform the requested logical operation on the
+    * converted fragment colors and the pixel colors.
+    */
+   switch (logic_op) {
+   case PIPE_LOGICOP_CLEAR:
+      spe_il(f, frag[0], 0);
+      spe_il(f, frag[1], 0);
+      spe_il(f, frag[2], 0);
+      spe_il(f, frag[3], 0);
+      break;
+   case PIPE_LOGICOP_NOR:
+      spe_nor(f, frag[0], frag[0], pixel[0]);
+      spe_nor(f, frag[1], frag[1], pixel[1]);
+      spe_nor(f, frag[2], frag[2], pixel[2]);
+      spe_nor(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_AND_INVERTED:
+      spe_andc(f, frag[0], pixel[0], frag[0]);
+      spe_andc(f, frag[1], pixel[1], frag[1]);
+      spe_andc(f, frag[2], pixel[2], frag[2]);
+      spe_andc(f, frag[3], pixel[3], frag[3]);
+      break;
+   case PIPE_LOGICOP_COPY_INVERTED:
+      spe_nor(f, frag[0], frag[0], frag[0]);
+      spe_nor(f, frag[1], frag[1], frag[1]);
+      spe_nor(f, frag[2], frag[2], frag[2]);
+      spe_nor(f, frag[3], frag[3], frag[3]);
+      break;
+   case PIPE_LOGICOP_AND_REVERSE:
+      spe_andc(f, frag[0], frag[0], pixel[0]);
+      spe_andc(f, frag[1], frag[1], pixel[1]);
+      spe_andc(f, frag[2], frag[2], pixel[2]);
+      spe_andc(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_XOR:
+      spe_xor(f, frag[0], frag[0], pixel[0]);
+      spe_xor(f, frag[1], frag[1], pixel[1]);
+      spe_xor(f, frag[2], frag[2], pixel[2]);
+      spe_xor(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_NAND:
+      spe_nand(f, frag[0], frag[0], pixel[0]);
+      spe_nand(f, frag[1], frag[1], pixel[1]);
+      spe_nand(f, frag[2], frag[2], pixel[2]);
+      spe_nand(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_AND:
+      spe_and(f, frag[0], frag[0], pixel[0]);
+      spe_and(f, frag[1], frag[1], pixel[1]);
+      spe_and(f, frag[2], frag[2], pixel[2]);
+      spe_and(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_EQUIV:
+      spe_eqv(f, frag[0], frag[0], pixel[0]);
+      spe_eqv(f, frag[1], frag[1], pixel[1]);
+      spe_eqv(f, frag[2], frag[2], pixel[2]);
+      spe_eqv(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_OR_INVERTED:
+      spe_orc(f, frag[0], pixel[0], frag[0]);
+      spe_orc(f, frag[1], pixel[1], frag[1]);
+      spe_orc(f, frag[2], pixel[2], frag[2]);
+      spe_orc(f, frag[3], pixel[3], frag[3]);
+      break;
+   case PIPE_LOGICOP_COPY:
+      break;
+   case PIPE_LOGICOP_OR_REVERSE:
+      spe_orc(f, frag[0], frag[0], pixel[0]);
+      spe_orc(f, frag[1], frag[1], pixel[1]);
+      spe_orc(f, frag[2], frag[2], pixel[2]);
+      spe_orc(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_OR:
+      spe_or(f, frag[0], frag[0], pixel[0]);
+      spe_or(f, frag[1], frag[1], pixel[1]);
+      spe_or(f, frag[2], frag[2], pixel[2]);
+      spe_or(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_SET:
+      spe_il(f, frag[0], ~0);
+      spe_il(f, frag[1], ~0);
+      spe_il(f, frag[2], ~0);
+      spe_il(f, frag[3], ~0);
+      break;
+
+   /* These two cases are short-circuited above.
+    */
+   case PIPE_LOGICOP_INVERT:
+   case PIPE_LOGICOP_NOOP:
+   default:
+      assert(0);
+   }
+
+
+   /* Apply fragment mask.
+    */
+   spe_ilh(f, tmp[0], 0x0000);
+   spe_ilh(f, tmp[1], 0x0404);
+   spe_ilh(f, tmp[2], 0x0808);
+   spe_ilh(f, tmp[3], 0x0c0c);
+
+   spe_shufb(f, tmp[0], mask, mask, tmp[0]);
+   spe_shufb(f, tmp[1], mask, mask, tmp[1]);
+   spe_shufb(f, tmp[2], mask, mask, tmp[2]);
+   spe_shufb(f, tmp[3], mask, mask, tmp[3]);
+
+   spe_selb(f, pixel[0], pixel[0], frag[0], tmp[0]);
+   spe_selb(f, pixel[1], pixel[1], frag[1], tmp[1]);
+   spe_selb(f, pixel[2], pixel[2], frag[2], tmp[2]);
+   spe_selb(f, pixel[3], pixel[3], frag[3], tmp[3]);
+
+   spe_bi(f, 0, 0, 0);
+
+#if 0
+   {
+      const uint32_t *p = f->store;
+      unsigned i;
+
+      printf("# %u instructions\n", f->csr - f->store);
+
+      printf("\t.text\n");
+      for (i = 0; i < 64; i++) {
+         printf("\t.long\t0x%04x\n", p[i]);
+      }
+      fflush(stdout);
+   }
+#endif
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h
new file mode 100644
index 0000000000..a8267a5133
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h
@@ -0,0 +1,39 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef CELL_STATE_PER_FRAGMENT_H
+#define CELL_STATE_PER_FRAGMENT_H
+
+extern void
+cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa);
+
+extern void
+cell_generate_alpha_blend(struct cell_blend_state *cb);
+
+extern void
+cell_generate_logic_op(struct spe_function *f,
+                       const struct pipe_blend_state *blend,
+                       struct pipe_surface *surf);
+
+#endif /* CELL_STATE_PER_FRAGMENT_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c
new file mode 100644
index 0000000000..bf517ea563
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c
@@ -0,0 +1,219 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "draw/draw_context.h"
+#include "tgsi/tgsi_parse.h"
+
+#include "cell_context.h"
+#include "cell_state.h"
+#include "cell_gen_fp.h"
+
+
+/** cast wrapper */
+static INLINE struct cell_fragment_shader_state *
+cell_fragment_shader_state(void *shader)
+{
+   return (struct cell_fragment_shader_state *) shader;
+}
+
+
+/** cast wrapper */
+static INLINE struct cell_vertex_shader_state *
+cell_vertex_shader_state(void *shader)
+{
+   return (struct cell_vertex_shader_state *) shader;
+}
+
+
+/**
+ * Create fragment shader state.
+ * Called via pipe->create_fs_state()
+ */
+static void *
+cell_create_fs_state(struct pipe_context *pipe,
+                     const struct pipe_shader_state *templ)
+{
+   struct cell_context *cell = cell_context(pipe);
+   struct cell_fragment_shader_state *cfs;
+
+   cfs = CALLOC_STRUCT(cell_fragment_shader_state);
+   if (!cfs)
+      return NULL;
+
+   cfs->shader.tokens = tgsi_dup_tokens(templ->tokens);
+   if (!cfs->shader.tokens) {
+      FREE(cfs);
+      return NULL;
+   }
+
+   tgsi_scan_shader(templ->tokens, &cfs->info);
+
+   cell_gen_fragment_program(cell, cfs->shader.tokens, &cfs->code);
+
+   return cfs;
+}
+
+
+/**
+ * Called via pipe->bind_fs_state()
+ */
+static void
+cell_bind_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   cell->fs = cell_fragment_shader_state(fs);
+
+   cell->dirty |= CELL_NEW_FS;
+}
+
+
+/**
+ * Called via pipe->delete_fs_state()
+ */
+static void
+cell_delete_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct cell_fragment_shader_state *cfs = cell_fragment_shader_state(fs);
+
+   spe_release_func(&cfs->code);
+
+   FREE((void *) cfs->shader.tokens);
+   FREE(cfs);
+}
+
+
+/**
+ * Create vertex shader state.
+ * Called via pipe->create_vs_state()
+ */
+static void *
+cell_create_vs_state(struct pipe_context *pipe,
+                     const struct pipe_shader_state *templ)
+{
+   struct cell_context *cell = cell_context(pipe);
+   struct cell_vertex_shader_state *cvs;
+
+   cvs = CALLOC_STRUCT(cell_vertex_shader_state);
+   if (!cvs)
+      return NULL;
+
+   cvs->shader.tokens = tgsi_dup_tokens(templ->tokens);
+   if (!cvs->shader.tokens) {
+      FREE(cvs);
+      return NULL;
+   }
+
+   tgsi_scan_shader(templ->tokens, &cvs->info);
+
+   cvs->draw_data = draw_create_vertex_shader(cell->draw, &cvs->shader);
+   if (cvs->draw_data == NULL) {
+      FREE( (void *) cvs->shader.tokens );
+      FREE( cvs );
+      return NULL;
+   }
+
+   return cvs;
+}
+
+
+/**
+ * Called via pipe->bind_vs_state()
+ */
+static void
+cell_bind_vs_state(struct pipe_context *pipe, void *vs)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   cell->vs = cell_vertex_shader_state(vs);
+
+   draw_bind_vertex_shader(cell->draw,
+                           (cell->vs ? cell->vs->draw_data : NULL));
+
+   cell->dirty |= CELL_NEW_VS;
+}
+
+
+/**
+ * Called via pipe->delete_vs_state()
+ */
+static void
+cell_delete_vs_state(struct pipe_context *pipe, void *vs)
+{
+   struct cell_context *cell = cell_context(pipe);
+   struct cell_vertex_shader_state *cvs = cell_vertex_shader_state(vs);
+
+   draw_delete_vertex_shader(cell->draw, cvs->draw_data);
+   FREE( (void *) cvs->shader.tokens );
+   FREE( cvs );
+}
+
+
+/**
+ * Called via pipe->set_constant_buffer()
+ */
+static void
+cell_set_constant_buffer(struct pipe_context *pipe,
+                         uint shader, uint index,
+                         const struct pipe_constant_buffer *buf)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   assert(shader < PIPE_SHADER_TYPES);
+   assert(index == 0);
+
+   draw_flush(cell->draw);
+
+   /* note: reference counting */
+   pipe_buffer_reference(pipe->screen,
+                         &cell->constants[shader].buffer,
+                         buf->buffer);
+
+   if (shader == PIPE_SHADER_VERTEX)
+      cell->dirty |= CELL_NEW_VS_CONSTANTS;
+   else if (shader == PIPE_SHADER_FRAGMENT)
+      cell->dirty |= CELL_NEW_FS_CONSTANTS;
+}
+
+
+void
+cell_init_shader_functions(struct cell_context *cell)
+{
+   cell->pipe.create_fs_state = cell_create_fs_state;
+   cell->pipe.bind_fs_state   = cell_bind_fs_state;
+   cell->pipe.delete_fs_state = cell_delete_fs_state;
+
+   cell->pipe.create_vs_state = cell_create_vs_state;
+   cell->pipe.bind_vs_state   = cell_bind_vs_state;
+   cell->pipe.delete_vs_state = cell_delete_vs_state;
+
+   cell->pipe.set_constant_buffer = cell_set_constant_buffer;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_state_vertex.c b/src/gallium/drivers/cell/ppu/cell_state_vertex.c
new file mode 100644
index 0000000000..fbe55c8472
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_state_vertex.c
@@ -0,0 +1,79 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "cell_context.h"
+#include "cell_state.h"
+
+#include "draw/draw_context.h"
+
+
+static void
+cell_set_vertex_elements(struct pipe_context *pipe,
+                         unsigned count,
+                         const struct pipe_vertex_element *elements)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(cell->vertex_element, elements, count * sizeof(elements[0]));
+   cell->num_vertex_elements = count;
+
+   cell->dirty |= CELL_NEW_VERTEX;
+
+   draw_set_vertex_elements(cell->draw, count, elements);
+}
+
+
+static void
+cell_set_vertex_buffers(struct pipe_context *pipe,
+                        unsigned count,
+                        const struct pipe_vertex_buffer *buffers)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(cell->vertex_buffer, buffers, count * sizeof(buffers[0]));
+   cell->num_vertex_buffers = count;
+
+   cell->dirty |= CELL_NEW_VERTEX;
+
+   draw_set_vertex_buffers(cell->draw, count, buffers);
+}
+
+
+void
+cell_init_vertex_functions(struct cell_context *cell)
+{
+   cell->pipe.set_vertex_buffers = cell_set_vertex_buffers;
+   cell->pipe.set_vertex_elements = cell_set_vertex_elements;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c
new file mode 100644
index 0000000000..c9203fee08
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_surface.c
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "util/u_rect.h"
+#include "cell_context.h"
+#include "cell_surface.h"
+
+
+void
+cell_init_surface_functions(struct cell_context *cell)
+{
+   cell->pipe.surface_copy = util_surface_copy;
+   cell->pipe.surface_fill = util_surface_fill;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_surface.h b/src/gallium/drivers/cell/ppu/cell_surface.h
new file mode 100644
index 0000000000..9e58f32944
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_surface.h
@@ -0,0 +1,42 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef CELL_SURFACE_H
+#define CELL_SURFACE_H
+
+
+struct cell_context;
+
+
+extern void
+cell_init_surface_functions(struct cell_context *cell);
+
+
+#endif /* SP_SURFACE_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c
new file mode 100644
index 0000000000..9ba995ab7d
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_texture.c
@@ -0,0 +1,531 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Michel Dänzer <michel@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "cell_context.h"
+#include "cell_state.h"
+#include "cell_texture.h"
+
+
+
+static unsigned
+minify(unsigned d)
+{
+   return MAX2(1, d>>1);
+}
+
+
+static void
+cell_texture_layout(struct cell_texture *ct)
+{
+   struct pipe_texture *pt = &ct->base;
+   unsigned level;
+   unsigned width = pt->width[0];
+   unsigned height = pt->height[0];
+   unsigned depth = pt->depth[0];
+
+   ct->buffer_size = 0;
+
+   for ( level = 0 ; level <= pt->last_level ; level++ ) {
+      unsigned size;
+      unsigned w_tile, h_tile;
+
+      assert(level < CELL_MAX_TEXTURE_LEVELS);
+
+      /* width, height, rounded up to tile size */
+      w_tile = align(width, TILE_SIZE);
+      h_tile = align(height, TILE_SIZE);
+
+      pt->width[level] = width;
+      pt->height[level] = height;
+      pt->depth[level] = depth;
+      pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w_tile);  
+      pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h_tile);  
+
+      ct->stride[level] = pt->nblocksx[level] * pt->block.size;
+
+      ct->level_offset[level] = ct->buffer_size;
+
+      size = pt->nblocksx[level] * pt->nblocksy[level] * pt->block.size;
+      if (pt->target == PIPE_TEXTURE_CUBE)
+         size *= 6;
+      else
+         size *= depth;
+
+      ct->buffer_size += size;
+
+      width  = minify(width);
+      height = minify(height);
+      depth = minify(depth);
+   }
+}
+
+
+static struct pipe_texture *
+cell_texture_create(struct pipe_screen *screen,
+                    const struct pipe_texture *templat)
+{
+   struct pipe_winsys *ws = screen->winsys;
+   struct cell_texture *ct = CALLOC_STRUCT(cell_texture);
+   if (!ct)
+      return NULL;
+
+   ct->base = *templat;
+   ct->base.refcount = 1;
+   ct->base.screen = screen;
+
+   cell_texture_layout(ct);
+
+   ct->buffer = ws->buffer_create(ws, 32, PIPE_BUFFER_USAGE_PIXEL,
+                                   ct->buffer_size);
+
+   if (!ct->buffer) {
+      FREE(ct);
+      return NULL;
+   }
+
+   return &ct->base;
+}
+
+
+static void
+cell_texture_release(struct pipe_screen *screen,
+                     struct pipe_texture **pt)
+{
+   if (!*pt)
+      return;
+
+   /*
+   DBG("%s %p refcount will be %d\n",
+       __FUNCTION__, (void *) *pt, (*pt)->refcount - 1);
+   */
+   if (--(*pt)->refcount <= 0) {
+      /* Delete this texture now.
+       * But note that the underlying pipe_buffer may linger...
+       */
+      struct cell_texture *ct = cell_texture(*pt);
+      uint i;
+
+      /*
+      DBG("%s deleting %p\n", __FUNCTION__, (void *) ct);
+      */
+
+      pipe_buffer_reference(screen, &ct->buffer, NULL);
+
+      for (i = 0; i < CELL_MAX_TEXTURE_LEVELS; i++) {
+         /* Unreference the tiled image buffer.
+          * It may not actually be deleted until a fence is hit.
+          */
+         if (ct->tiled_buffer[i]) {
+            ct->tiled_mapped[i] = NULL;
+            pipe_buffer_reference(screen, &ct->tiled_buffer[i], NULL);
+         }
+      }
+
+      FREE(ct);
+   }
+   *pt = NULL;
+}
+
+
+
+/**
+ * Convert image from linear layout to tiled layout.  4-byte pixels.
+ */
+static void
+twiddle_image_uint(uint w, uint h, uint tile_size, uint *dst,
+                   uint src_stride, const uint *src)
+{
+   const uint tile_size2 = tile_size * tile_size;
+   const uint h_t = (h + tile_size - 1) / tile_size;
+   const uint w_t = (w + tile_size - 1) / tile_size;
+
+   uint it, jt;  /* tile counters */
+   uint i, j;    /* intra-tile counters */
+
+   src_stride /= 4; /* convert from bytes to pixels */
+
+   /* loop over dest tiles */
+   for (it = 0; it < h_t; it++) {
+      for (jt = 0; jt < w_t; jt++) {
+         /* start of dest tile: */
+         uint *tdst = dst + (it * w_t + jt) * tile_size2;
+
+         /* compute size of this tile (may be smaller than tile_size) */
+         /* XXX note: a compiler bug was found here. That's why the code
+          * looks as it does.
+          */
+         uint tile_width = w - jt * tile_size;
+         tile_width = MIN2(tile_width, tile_size);
+         uint tile_height = h - it * tile_size;
+         tile_height = MIN2(tile_height, tile_size);
+
+         /* loop over texels in the tile */
+         for (i = 0; i < tile_height; i++) {
+            for (j = 0; j < tile_width; j++) {
+               const uint srci = it * tile_size + i;
+               const uint srcj = jt * tile_size + j;
+               ASSERT(srci < h);
+               ASSERT(srcj < w);
+               tdst[i * tile_size + j] = src[srci * src_stride + srcj];
+            }
+         }
+      }
+   }
+}
+
+
+/**
+ * For Cell.  Basically, rearrange the pixels/quads from this layout:
+ *  +--+--+--+--+
+ *  |p0|p1|p2|p3|....
+ *  +--+--+--+--+
+ *
+ * to this layout:
+ *  +--+--+
+ *  |p0|p1|....
+ *  +--+--+
+ *  |p2|p3|
+ *  +--+--+
+ */
+static void
+twiddle_tile(const uint *tileIn, uint *tileOut)
+{
+   int y, x;
+
+   for (y = 0; y < TILE_SIZE; y+=2) {
+      for (x = 0; x < TILE_SIZE; x+=2) {
+         int k = 4 * (y/2 * TILE_SIZE/2 + x/2);
+         tileOut[y * TILE_SIZE + (x + 0)] = tileIn[k];
+         tileOut[y * TILE_SIZE + (x + 1)] = tileIn[k+1];
+         tileOut[(y + 1) * TILE_SIZE + (x + 0)] = tileIn[k+2];
+         tileOut[(y + 1) * TILE_SIZE + (x + 1)] = tileIn[k+3];
+      }
+   }
+}
+
+
+/**
+ * Convert image from tiled layout to linear layout.  4-byte pixels.
+ */
+static void
+untwiddle_image_uint(uint w, uint h, uint tile_size, uint *dst,
+                     uint dst_stride, const uint *src)
+{
+   const uint tile_size2 = tile_size * tile_size;
+   const uint h_t = (h + tile_size - 1) / tile_size;
+   const uint w_t = (w + tile_size - 1) / tile_size;
+   uint *tile_buf;
+   uint it, jt;  /* tile counters */
+   uint i, j;    /* intra-tile counters */
+
+   dst_stride /= 4; /* convert from bytes to pixels */
+
+   tile_buf = align_malloc(tile_size * tile_size * 4, 16);
+   
+   /* loop over src tiles */
+   for (it = 0; it < h_t; it++) {
+      for (jt = 0; jt < w_t; jt++) {
+         /* start of src tile: */
+         const uint *tsrc = src + (it * w_t + jt) * tile_size2;
+         
+         twiddle_tile(tsrc, tile_buf);
+         tsrc = tile_buf;
+
+         /* compute size of this tile (may be smaller than tile_size) */
+         /* XXX note: a compiler bug was found here. That's why the code
+          * looks as it does.
+          */
+         uint tile_width = w - jt * tile_size;
+         tile_width = MIN2(tile_width, tile_size);
+         uint tile_height = h - it * tile_size;
+         tile_height = MIN2(tile_height, tile_size);
+
+         /* loop over texels in the tile */
+         for (i = 0; i < tile_height; i++) {
+            for (j = 0; j < tile_width; j++) {
+               uint dsti = it * tile_size + i;
+               uint dstj = jt * tile_size + j;
+               ASSERT(dsti < h);
+               ASSERT(dstj < w);
+               dst[dsti * dst_stride + dstj] = tsrc[i * tile_size + j];
+            }
+         }
+      }
+   }
+
+   align_free(tile_buf);
+}
+
+
+/**
+ * Convert linear texture image data to tiled format for SPU usage.
+ */
+static void
+cell_twiddle_texture(struct pipe_screen *screen,
+                     struct pipe_surface *surface)
+{
+   struct cell_texture *ct = cell_texture(surface->texture);
+   const uint level = surface->level;
+   const uint texWidth = ct->base.width[level];
+   const uint texHeight = ct->base.height[level];
+   const uint bufWidth = align(texWidth, TILE_SIZE);
+   const uint bufHeight = align(texHeight, TILE_SIZE);
+   const void *map = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_READ);
+   const uint *src = (const uint *) map;
+
+   switch (ct->base.format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+   case PIPE_FORMAT_S8Z24_UNORM:
+      {
+         int numFaces = ct->base.target == PIPE_TEXTURE_CUBE ? 6 : 1;
+         int offset = bufWidth * bufHeight * 4 * surface->face;
+         uint *dst;
+
+         if (!ct->tiled_buffer[level]) {
+            /* allocate buffer for tiled data now */
+            struct pipe_winsys *ws = screen->winsys;
+            uint bytes = bufWidth * bufHeight * 4 * numFaces;
+            ct->tiled_buffer[level] =
+               ws->buffer_create(ws, 16, PIPE_BUFFER_USAGE_PIXEL, bytes);
+            /* and map it */
+            ct->tiled_mapped[level] =
+               ws->buffer_map(ws, ct->tiled_buffer[level],
+                              PIPE_BUFFER_USAGE_GPU_READ);
+         }
+         dst = (uint *) ((ubyte *) ct->tiled_mapped[level] + offset);
+
+         twiddle_image_uint(texWidth, texHeight, TILE_SIZE, dst,
+                            surface->stride, src);
+      }
+      break;
+   default:
+      printf("Cell: twiddle unsupported texture format %s\n",
+             pf_name(ct->base.format));
+   }
+
+   screen->surface_unmap(screen, surface);
+}
+
+
+/**
+ * Convert SPU tiled texture image data to linear format for app usage.
+ */
+static void
+cell_untwiddle_texture(struct pipe_screen *screen,
+                     struct pipe_surface *surface)
+{
+   struct cell_texture *ct = cell_texture(surface->texture);
+   const uint level = surface->level;
+   const uint texWidth = ct->base.width[level];
+   const uint texHeight = ct->base.height[level];
+   const void *map = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_READ);
+   const uint *src = (const uint *) ((const ubyte *) map + surface->offset);
+
+   switch (ct->base.format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+   case PIPE_FORMAT_S8Z24_UNORM:
+      {
+         int numFaces = ct->base.target == PIPE_TEXTURE_CUBE ? 6 : 1;
+         int offset = surface->stride * texHeight * 4 * surface->face;
+         uint *dst;
+
+         if (!ct->untiled_data[level]) {
+            ct->untiled_data[level] =
+               align_malloc(surface->stride * texHeight * 4 * numFaces, 16);
+         }
+
+         dst = (uint *) ((ubyte *) ct->untiled_data[level] + offset);
+
+         untwiddle_image_uint(texWidth, texHeight, TILE_SIZE, dst,
+                              surface->stride, src);
+      }
+      break;
+   default:
+      {
+         ct->untiled_data[level] = NULL;
+         printf("Cell: untwiddle unsupported texture format %s\n",
+                pf_name(ct->base.format));
+      }
+   }
+
+   screen->surface_unmap(screen, surface);
+}
+
+
+static struct pipe_surface *
+cell_get_tex_surface(struct pipe_screen *screen,
+                     struct pipe_texture *pt,
+                     unsigned face, unsigned level, unsigned zslice,
+                     unsigned usage)
+{
+   struct cell_texture *ct = cell_texture(pt);
+   struct pipe_surface *ps;
+
+   ps = CALLOC_STRUCT(pipe_surface);
+   if (ps) {
+      ps->refcount = 1;
+      pipe_texture_reference(&ps->texture, pt);
+      ps->format = pt->format;
+      ps->block = pt->block;
+      ps->width = pt->width[level];
+      ps->height = pt->height[level];
+      ps->nblocksx = pt->nblocksx[level];
+      ps->nblocksy = pt->nblocksy[level];
+      ps->stride = ct->stride[level];
+      ps->offset = ct->level_offset[level];
+      ps->usage = usage;
+
+      /* XXX may need to override usage flags (see sp_texture.c) */
+
+      pipe_texture_reference(&ps->texture, pt); 
+      ps->face = face;
+      ps->level = level;
+      ps->zslice = zslice;
+
+      if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) {
+         ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) *
+            ps->nblocksy *
+            ps->stride;
+      }
+      else {
+         assert(face == 0);
+         assert(zslice == 0);
+      }
+
+      if (ps->usage & PIPE_BUFFER_USAGE_CPU_READ) {
+         /* convert from tiled to linear layout */
+         cell_untwiddle_texture(screen, ps);
+      }
+   }
+   return ps;
+}
+
+
+static void 
+cell_tex_surface_release(struct pipe_screen *screen, 
+                         struct pipe_surface **s)
+{
+   struct cell_texture *ct = cell_texture((*s)->texture);
+   const uint level = (*s)->level;
+   struct pipe_surface *surf = *s;
+
+   if ((surf->usage & PIPE_BUFFER_USAGE_CPU_READ) && (ct->untiled_data[level]))
+   {
+      align_free(ct->untiled_data[level]);
+      ct->untiled_data[level] = NULL;
+   }
+
+   if ((ct->base.tex_usage & PIPE_TEXTURE_USAGE_SAMPLER) &&
+       (surf->usage & PIPE_BUFFER_USAGE_CPU_WRITE)) {
+      /* convert from linear to tiled layout */
+      cell_twiddle_texture(screen, surf);
+   }
+
+   /* XXX if done rendering to teximage, re-tile */
+
+   if (--surf->refcount == 0) {
+      pipe_texture_reference(&surf->texture, NULL);
+      FREE(surf);
+   }
+   *s = NULL;
+}
+
+
+static void *
+cell_surface_map(struct pipe_screen *screen,
+                 struct pipe_surface *surface,
+                 unsigned flags)
+{
+   ubyte *map;
+   struct cell_texture *ct = cell_texture(surface->texture);
+   const uint level = surface->level;
+
+   assert(ct);
+
+#if 0
+   if (flags & ~surface->usage) {
+      assert(0);
+      return NULL;
+   }
+#endif
+
+   map = pipe_buffer_map( screen, ct->buffer, flags );
+   if (map == NULL) {
+      return NULL;
+   }
+   else {
+      if ((surface->usage & PIPE_BUFFER_USAGE_CPU_READ) &&
+          (ct->untiled_data[level])) {
+         return (void *) ((ubyte *) ct->untiled_data[level] + surface->offset);
+      }
+      else {
+         return (void *) (map + surface->offset);
+      }
+   }
+}
+
+
+static void
+cell_surface_unmap(struct pipe_screen *screen,
+                   struct pipe_surface *surface)
+{
+   struct cell_texture *ct = cell_texture(surface->texture);
+
+   assert(ct);
+
+   pipe_buffer_unmap( screen, ct->buffer );
+}
+
+
+
+void
+cell_init_screen_texture_funcs(struct pipe_screen *screen)
+{
+   screen->texture_create = cell_texture_create;
+   screen->texture_release = cell_texture_release;
+
+   screen->get_tex_surface = cell_get_tex_surface;
+   screen->tex_surface_release = cell_tex_surface_release;
+
+   screen->surface_map = cell_surface_map;
+   screen->surface_unmap = cell_surface_unmap;
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_texture.h b/src/gallium/drivers/cell/ppu/cell_texture.h
new file mode 100644
index 0000000000..7018b0c9bf
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_texture.h
@@ -0,0 +1,72 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef CELL_TEXTURE_H
+#define CELL_TEXTURE_H
+
+
+struct cell_context;
+struct pipe_texture;
+
+
+/**
+ * Subclass of pipe_texture
+ */
+struct cell_texture
+{
+   struct pipe_texture base;
+
+   unsigned long level_offset[CELL_MAX_TEXTURE_LEVELS];
+   unsigned long stride[CELL_MAX_TEXTURE_LEVELS];
+
+   /* The data is held here:
+    */
+   struct pipe_buffer *buffer;
+   unsigned long buffer_size;
+
+   /** Texture data in tiled layout is held here */
+   struct pipe_buffer *tiled_buffer[CELL_MAX_TEXTURE_LEVELS];
+   /** Mapped, tiled texture data */
+   void *tiled_mapped[CELL_MAX_TEXTURE_LEVELS];
+   void *untiled_data[CELL_MAX_TEXTURE_LEVELS];
+};
+
+
+/** cast wrapper */
+static INLINE struct cell_texture *
+cell_texture(struct pipe_texture *pt)
+{
+   return (struct cell_texture *) pt;
+}
+
+
+
+extern void
+cell_init_screen_texture_funcs(struct pipe_screen *screen);
+
+
+#endif /* CELL_TEXTURE_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_vbuf.c b/src/gallium/drivers/cell/ppu/cell_vbuf.c
new file mode 100644
index 0000000000..ab54e79689
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_vbuf.c
@@ -0,0 +1,309 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Vertex buffer code.  The draw module transforms vertices to window
+ * coords, etc. and emits the vertices into buffer supplied by this module.
+ * When a vertex buffer is full, or we flush, we'll send the vertex data
+ * to the SPUs.
+ *
+ * Authors
+ *  Brian Paul
+ */
+
+
+#include "cell_batch.h"
+#include "cell_context.h"
+#include "cell_fence.h"
+#include "cell_flush.h"
+#include "cell_spu.h"
+#include "cell_vbuf.h"
+#include "draw/draw_vbuf.h"
+#include "util/u_memory.h"
+
+
+/** Allow vertex data to be inlined after RENDER command */
+#define ALLOW_INLINE_VERTS 1
+
+
+/**
+ * Subclass of vbuf_render because we need a cell_context pointer in
+ * a few places.
+ */
+struct cell_vbuf_render
+{
+   struct vbuf_render base;
+   struct cell_context *cell;
+   uint prim;            /**< PIPE_PRIM_x */
+   uint vertex_size;     /**< in bytes */
+   void *vertex_buffer;  /**< just for debug, really */
+   uint vertex_buf;      /**< in [0, CELL_NUM_BUFFERS-1] */
+};
+
+
+/** cast wrapper */
+static struct cell_vbuf_render *
+cell_vbuf_render(struct vbuf_render *vbr)
+{
+   return (struct cell_vbuf_render *) vbr;
+}
+
+
+
+static const struct vertex_info *
+cell_vbuf_get_vertex_info(struct vbuf_render *vbr)
+{
+   struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr);
+   return &cvbr->cell->vertex_info;
+}
+
+
+static void *
+cell_vbuf_allocate_vertices(struct vbuf_render *vbr,
+                            ushort vertex_size, ushort nr_vertices)
+{
+   struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr);
+   /*printf("Alloc verts %u * %u\n", vertex_size, nr_vertices);*/
+
+   assert(cvbr->vertex_buf == ~0);
+   cvbr->vertex_buf = cell_get_empty_buffer(cvbr->cell);
+   cvbr->vertex_buffer = cvbr->cell->buffer[cvbr->vertex_buf];
+   cvbr->vertex_size = vertex_size;
+   return cvbr->vertex_buffer;
+}
+
+
+static void
+cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, 
+                           unsigned vertex_size, unsigned vertices_used)
+{
+   struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr);
+   struct cell_context *cell = cvbr->cell;
+
+   /*
+   printf("%s vertex_buf = %u  count = %u\n",
+          __FUNCTION__, cvbr->vertex_buf, vertices_used);
+   */
+
+   /* Make sure texture buffers aren't released until we're done rendering
+    * with them.
+    */
+   cell_add_fenced_textures(cell);
+
+   /* Tell SPUs they can release the vert buf */
+   if (cvbr->vertex_buf != ~0U) {
+      STATIC_ASSERT(sizeof(struct cell_command_release_verts) % 16 == 0);
+      struct cell_command_release_verts *release
+         = (struct cell_command_release_verts *)
+         cell_batch_alloc16(cell, sizeof(struct cell_command_release_verts));
+      release->opcode[0] = CELL_CMD_RELEASE_VERTS;
+      release->vertex_buf = cvbr->vertex_buf;
+   }
+
+   cvbr->vertex_buf = ~0;
+   cell_flush_int(cell, 0x0);
+
+   assert(vertices == cvbr->vertex_buffer);
+   cvbr->vertex_buffer = NULL;
+}
+
+
+
+static boolean
+cell_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim)
+{
+   struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr);
+   cvbr->prim = prim;
+   /*printf("cell_set_prim %u\n", prim);*/
+   return TRUE;
+}
+
+
+static void
+cell_vbuf_draw(struct vbuf_render *vbr,
+	       const ushort *indices,
+               uint nr_indices)
+{
+   struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr);
+   struct cell_context *cell = cvbr->cell;
+   float xmin, ymin, xmax, ymax;
+   uint i;
+   uint nr_vertices = 0, min_index = ~0;
+   const void *vertices = cvbr->vertex_buffer;
+   const uint vertex_size = cvbr->vertex_size;
+
+   for (i = 0; i < nr_indices; i++) {
+      if (indices[i] > nr_vertices)
+         nr_vertices = indices[i];
+      if (indices[i] < min_index)
+         min_index = indices[i];
+   }
+   nr_vertices++;
+
+#if 0
+   /*if (min_index > 0)*/
+      printf("%s min_index = %u\n", __FUNCTION__, min_index);
+#endif
+
+#if 0
+   printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u\n",
+          nr_indices, nr_vertices);
+   printf("  ");
+   for (i = 0; i < nr_indices; i += 3) {
+      printf("%u %u %u, ", indices[i+0], indices[i+1], indices[i+2]);
+   }
+   printf("\n");
+#elif 0
+   printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u  indexes = [%u %u %u ...]\n",
+          nr_indices, nr_vertices,
+          indices[0], indices[1], indices[2]);
+   printf("ind space = %u, vert space = %u, space = %u\n",
+          nr_indices * 2,
+          nr_vertices * 4 * cell->vertex_info.size,
+          cell_batch_free_space(cell));
+#endif
+
+   /* compute x/y bounding box */
+   xmin = ymin = 1e50;
+   xmax = ymax = -1e50;
+   for (i = min_index; i < nr_vertices; i++) {
+      const float *v = (float *) ((ubyte *) vertices + i * vertex_size);
+      if (v[0] < xmin)
+         xmin = v[0];
+      if (v[0] > xmax)
+         xmax = v[0];
+      if (v[1] < ymin)
+         ymin = v[1];
+      if (v[1] > ymax)
+         ymax = v[1];
+   }
+#if 0
+   printf("PPU Bounds %g, %g .. %g, %g\n", xmin, ymin, xmax, ymax);
+   fflush(stdout);
+#endif
+
+   if (cvbr->prim != PIPE_PRIM_TRIANGLES)
+      return; /* only render tris for now */
+
+   /* build/insert batch RENDER command */
+   {
+      const uint index_bytes = ROUNDUP16(nr_indices * 2);
+      const uint vertex_bytes = ROUNDUP16(nr_vertices * 4 * cell->vertex_info.size);
+      STATIC_ASSERT(sizeof(struct cell_command_render) % 16 == 0);
+      const uint batch_size = sizeof(struct cell_command_render) + index_bytes;
+
+      struct cell_command_render *render
+         = (struct cell_command_render *)
+         cell_batch_alloc16(cell, batch_size);
+
+      render->opcode[0] = CELL_CMD_RENDER;
+      render->prim_type = cvbr->prim;
+
+      render->num_indexes = nr_indices;
+      render->min_index = min_index;
+
+      /* append indices after render command */
+      memcpy(render + 1, indices, nr_indices * 2);
+
+      /* if there's room, append vertices after the indices, else leave
+       * vertices in the original/separate buffer.
+       */
+      render->vertex_size = 4 * cell->vertex_info.size;
+      render->num_verts = nr_vertices;
+      if (ALLOW_INLINE_VERTS &&
+          min_index == 0 &&
+          vertex_bytes + 16 <= cell_batch_free_space(cell)) {
+         /* vertex data inlined, after indices, at 16-byte boundary */
+         void *dst = cell_batch_alloc16(cell, vertex_bytes);
+         memcpy(dst, vertices, vertex_bytes);
+         render->inline_verts = TRUE;
+         render->vertex_buf = ~0;
+      }
+      else {
+         /* vertex data in separate buffer */
+         render->inline_verts = FALSE;
+         ASSERT(cvbr->vertex_buf >= 0);
+         render->vertex_buf = cvbr->vertex_buf;
+      }
+
+      render->xmin = xmin;
+      render->ymin = ymin;
+      render->xmax = xmax;
+      render->ymax = ymax;
+   }
+
+#if 0
+   /* helpful for debug */
+   cell_flush_int(cell, CELL_FLUSH_WAIT);
+#endif
+}
+
+
+static void
+cell_vbuf_destroy(struct vbuf_render *vbr)
+{
+   struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr);
+   cvbr->cell->vbuf_render = NULL;
+   FREE(cvbr);
+}
+
+
+/**
+ * Initialize the post-transform vertex buffer information for the given
+ * context.
+ */
+void
+cell_init_vbuf(struct cell_context *cell)
+{
+   assert(cell->draw);
+
+   cell->vbuf_render = CALLOC_STRUCT(cell_vbuf_render);
+
+   /* The max number of indexes is what can fix into a batch buffer,
+    * minus the render and release-verts commands.
+    */
+   cell->vbuf_render->base.max_indices
+      = (CELL_BUFFER_SIZE
+         - sizeof(struct cell_command_render)
+         - sizeof(struct cell_command_release_verts))
+      / sizeof(ushort);
+   cell->vbuf_render->base.max_vertex_buffer_bytes = CELL_BUFFER_SIZE;
+
+   cell->vbuf_render->base.get_vertex_info = cell_vbuf_get_vertex_info;
+   cell->vbuf_render->base.allocate_vertices = cell_vbuf_allocate_vertices;
+   cell->vbuf_render->base.set_primitive = cell_vbuf_set_primitive;
+   cell->vbuf_render->base.draw = cell_vbuf_draw;
+   cell->vbuf_render->base.release_vertices = cell_vbuf_release_vertices;
+   cell->vbuf_render->base.destroy = cell_vbuf_destroy;
+
+   cell->vbuf_render->cell = cell;
+#if 1
+   cell->vbuf_render->vertex_buf = ~0;
+#endif
+
+   cell->vbuf = draw_vbuf_stage(cell->draw, &cell->vbuf_render->base);
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_vbuf.h b/src/gallium/drivers/cell/ppu/cell_vbuf.h
new file mode 100644
index 0000000000..d265cbf770
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_vbuf.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef CELL_VBUF_H
+#define CELL_VBUF_H
+
+
+struct cell_context;
+
+extern void
+cell_init_vbuf(struct cell_context *cell);
+
+
+#endif /* CELL_VBUF_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c
new file mode 100644
index 0000000000..9cba537d9e
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c
@@ -0,0 +1,346 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <inttypes.h>
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_format.h"
+
+#include "../auxiliary/draw/draw_context.h"
+#include "../auxiliary/draw/draw_private.h"
+
+#include "cell_context.h"
+#include "rtasm/rtasm_ppc_spe.h"
+
+
+/**
+ * Emit a 4x4 matrix transpose operation
+ *
+ * \param p         Function that the transpose operation is to be appended to
+ * \param row0      Register containing row 0 of the source matrix
+ * \param row1      Register containing row 1 of the source matrix
+ * \param row2      Register containing row 2 of the source matrix
+ * \param row3      Register containing row 3 of the source matrix
+ * \param dest_ptr  Register containing the address of the destination matrix
+ * \param shuf_ptr  Register containing the address of the shuffled data
+ * \param count     Number of colums to actually be written to the destination
+ *
+ * \note
+ * This function assumes that the registers named by \c row0, \c row1,
+ * \c row2, and \c row3 are scratch and can be modified by the generated code.
+ * Furthermore, these registers will be released, via calls to
+ * \c release_register, by this function.
+ * 
+ * \note
+ * This function requires that four temporary are available on entry.
+ */
+static void
+emit_matrix_transpose(struct spe_function *p,
+		      unsigned row0, unsigned row1, unsigned row2,
+		      unsigned row3, unsigned dest_ptr,
+		      unsigned shuf_ptr, unsigned count)
+{
+   int shuf_hi = spe_allocate_available_register(p);
+   int shuf_lo = spe_allocate_available_register(p);
+   int t1 = spe_allocate_available_register(p);
+   int t2 = spe_allocate_available_register(p);
+   int t3;
+   int t4;
+   int col0;
+   int col1;
+   int col2;
+   int col3;
+
+
+   spe_lqd(p, shuf_hi, shuf_ptr, 3*16);
+   spe_lqd(p, shuf_lo, shuf_ptr, 4*16);
+   spe_shufb(p, t1, row0, row2, shuf_hi);
+   spe_shufb(p, t2, row0, row2, shuf_lo);
+
+
+   /* row0 and row2 are now no longer needed.  Re-use those registers as
+    * temporaries.
+    */
+   t3 = row0;
+   t4 = row2;
+
+   spe_shufb(p, t3, row1, row3, shuf_hi);
+   spe_shufb(p, t4, row1, row3, shuf_lo);
+
+
+   /* row1 and row3 are now no longer needed.  Re-use those registers as
+    * temporaries.
+    */
+   col0 = row1;
+   col1 = row3;
+
+   spe_shufb(p, col0, t1, t3, shuf_hi);
+   if (count > 1) {
+      spe_shufb(p, col1, t1, t3, shuf_lo);
+   }
+
+   /* t1 and t3 are now no longer needed.  Re-use those registers as
+    * temporaries.
+    */
+   col2 = t1;
+   col3 = t3;
+
+   if (count > 2) {
+      spe_shufb(p, col2, t2, t4, shuf_hi);
+   }
+
+   if (count > 3) {
+      spe_shufb(p, col3, t2, t4, shuf_lo);
+   }
+
+
+   /* Store the results.  Remember that the stqd instruction is encoded using
+    * the qword offset (stand-alone assemblers to the byte-offset to
+    * qword-offset conversion for you), so the byte-offset needs be divided by
+    * 16.
+    */
+   switch (count) {
+   case 4:
+      spe_stqd(p, col3, dest_ptr, 3 * 16);
+   case 3:
+      spe_stqd(p, col2, dest_ptr, 2 * 16);
+   case 2:
+      spe_stqd(p, col1, dest_ptr, 1 * 16);
+   case 1:
+      spe_stqd(p, col0, dest_ptr, 0 * 16);
+   }
+
+
+   /* Release all of the temporary registers used.
+    */
+   spe_release_register(p, col0);
+   spe_release_register(p, col1);
+   spe_release_register(p, col2);
+   spe_release_register(p, col3);
+   spe_release_register(p, shuf_hi);
+   spe_release_register(p, shuf_lo);
+   spe_release_register(p, t2);
+   spe_release_register(p, t4);
+}
+
+
+#if 0
+/* This appears to not be used currently */
+static void
+emit_fetch(struct spe_function *p,
+	   unsigned in_ptr, unsigned *offset,
+	   unsigned out_ptr, unsigned shuf_ptr,
+	   enum pipe_format format)
+{
+   const unsigned count = (pf_size_x(format) != 0) + (pf_size_y(format) != 0)
+       + (pf_size_z(format) != 0) + (pf_size_w(format) != 0);
+   const unsigned type = pf_type(format);
+   const unsigned bytes = pf_size_x(format);
+
+   int v0 = spe_allocate_available_register(p);
+   int v1 = spe_allocate_available_register(p);
+   int v2 = spe_allocate_available_register(p);
+   int v3 = spe_allocate_available_register(p);
+   int tmp = spe_allocate_available_register(p);
+   int float_zero = -1;
+   int float_one = -1;
+   float scale_signed = 0.0;
+   float scale_unsigned = 0.0;
+
+   spe_lqd(p, v0, in_ptr, (0 + offset[0]) * 16);
+   spe_lqd(p, v1, in_ptr, (1 + offset[0]) * 16);
+   spe_lqd(p, v2, in_ptr, (2 + offset[0]) * 16);
+   spe_lqd(p, v3, in_ptr, (3 + offset[0]) * 16);
+   offset[0] += 4;
+   
+   switch (bytes) {
+   case 1:
+      scale_signed = 1.0f / 127.0f;
+      scale_unsigned = 1.0f / 255.0f;
+      spe_lqd(p, tmp, shuf_ptr, 1 * 16);
+      spe_shufb(p, v0, v0, v0, tmp);
+      spe_shufb(p, v1, v1, v1, tmp);
+      spe_shufb(p, v2, v2, v2, tmp);
+      spe_shufb(p, v3, v3, v3, tmp);
+      break;
+   case 2:
+      scale_signed = 1.0f / 32767.0f;
+      scale_unsigned = 1.0f / 65535.0f;
+      spe_lqd(p, tmp, shuf_ptr, 2 * 16);
+      spe_shufb(p, v0, v0, v0, tmp);
+      spe_shufb(p, v1, v1, v1, tmp);
+      spe_shufb(p, v2, v2, v2, tmp);
+      spe_shufb(p, v3, v3, v3, tmp);
+      break;
+   case 4:
+      scale_signed = 1.0f / 2147483647.0f;
+      scale_unsigned = 1.0f / 4294967295.0f;
+      break;
+   default:
+      assert(0);
+      break;
+   }
+
+   switch (type) {
+   case PIPE_FORMAT_TYPE_FLOAT:
+      break;
+   case PIPE_FORMAT_TYPE_UNORM:
+      spe_ilhu(p, tmp, ((unsigned) scale_unsigned) >> 16);
+      spe_iohl(p, tmp, ((unsigned) scale_unsigned) & 0x0ffff);
+      spe_cuflt(p, v0, v0, 0);
+      spe_fm(p, v0, v0, tmp);
+      break;
+   case PIPE_FORMAT_TYPE_SNORM:
+      spe_ilhu(p, tmp, ((unsigned) scale_signed) >> 16);
+      spe_iohl(p, tmp, ((unsigned) scale_signed) & 0x0ffff);
+      spe_csflt(p, v0, v0, 0);
+      spe_fm(p, v0, v0, tmp);
+      break;
+   case PIPE_FORMAT_TYPE_USCALED:
+      spe_cuflt(p, v0, v0, 0);
+      break;
+   case PIPE_FORMAT_TYPE_SSCALED:
+      spe_csflt(p, v0, v0, 0);
+      break;
+   }
+
+
+   if (count < 4) {
+      float_one = spe_allocate_available_register(p);
+      spe_il(p, float_one, 1);
+      spe_cuflt(p, float_one, float_one, 0);
+      
+      if (count < 3) {
+	 float_zero = spe_allocate_available_register(p);
+	 spe_il(p, float_zero, 0);
+      }
+   }
+
+   spe_release_register(p, tmp);
+
+   emit_matrix_transpose(p, v0, v1, v2, v3, out_ptr, shuf_ptr, count);
+
+   switch (count) {
+   case 1:
+      spe_stqd(p, float_zero, out_ptr, 1 * 16);
+   case 2:
+      spe_stqd(p, float_zero, out_ptr, 2 * 16);
+   case 3:
+      spe_stqd(p, float_one, out_ptr, 3 * 16);
+   }
+
+   if (float_zero != -1) {
+      spe_release_register(p, float_zero);
+   }
+
+   if (float_one != -1) {
+      spe_release_register(p, float_one);
+   }
+}
+#endif
+
+
+void cell_update_vertex_fetch(struct draw_context *draw)
+{
+#if 0
+   struct cell_context *const cell =
+       (struct cell_context *) draw->driver_private;
+   struct spe_function *p = &cell->attrib_fetch;
+   unsigned function_index[PIPE_MAX_ATTRIBS];
+   unsigned unique_attr_formats;
+   int out_ptr;
+   int in_ptr;
+   int shuf_ptr;
+   unsigned i;
+   unsigned j;
+
+
+   /* Determine how many unique input attribute formats there are.  At the
+    * same time, store the index of the lowest numbered attribute that has
+    * the same format as any non-unique format.
+    */
+   unique_attr_formats = 1;
+   function_index[0] = 0;
+   for (i = 1; i < draw->vertex_fetch.nr_attrs; i++) {
+      const enum pipe_format curr_fmt = draw->vertex_element[i].src_format;
+
+      for (j = 0; j < i; j++) {
+	 if (curr_fmt == draw->vertex_element[j].src_format) {
+	    break;
+	 }
+      }
+      
+      if (j == i) {
+	 unique_attr_formats++;
+      }
+
+      function_index[i] = j;
+   }
+
+
+   /* Each fetch function can be a maximum of 34 instructions (note: this is
+    * actually a slight over-estimate).
+    */
+   spe_init_func(p, 34 * SPE_INST_SIZE * unique_attr_formats);
+
+
+   /* Allocate registers for the function's input parameters.
+    */
+   out_ptr = spe_allocate_register(p, 3);
+   in_ptr = spe_allocate_register(p, 4);
+   shuf_ptr = spe_allocate_register(p, 5);
+
+
+   /* Generate code for the individual attribute fetch functions.
+    */
+   for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) {
+      unsigned offset;
+
+      if (function_index[i] == i) {
+	 cell->attrib_fetch_offsets[i] = (unsigned) ((void *) p->csr 
+						     - (void *) p->store);
+
+	 offset = 0;
+	 emit_fetch(p, in_ptr, &offset, out_ptr, shuf_ptr,
+		    draw->vertex_element[i].src_format);
+	 spe_bi(p, 0, 0, 0);
+
+	 /* Round up to the next 16-byte boundary.
+	  */
+	 if ((((unsigned) p->store) & 0x0f) != 0) {
+	    const unsigned align = ((unsigned) p->store) & 0x0f;
+	    p->store = (uint32_t *) (((void *) p->store) + align);
+	 }
+      } else {
+	 /* Use the same function entry-point as a previously seen attribute
+	  * with the same format.
+	  */
+	 cell->attrib_fetch_offsets[i] = 
+	     cell->attrib_fetch_offsets[function_index[i]];
+      }
+   }
+#else
+   assert(0);
+#endif
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c
new file mode 100644
index 0000000000..403cf6d50f
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c
@@ -0,0 +1,146 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file cell_vertex_shader.c
+ * Vertex shader interface routines for Cell.
+ *
+ * \author Ian Romanick <idr@us.ibm.com>
+ */
+
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_math.h"
+
+#include "cell_context.h"
+#include "cell_draw_arrays.h"
+#include "cell_flush.h"
+#include "cell_spu.h"
+#include "cell_batch.h"
+
+#include "cell/common.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+
+/**
+ * Run the vertex shader on all vertices in the vertex queue.
+ * Called by the draw module when the vertx cache needs to be flushed.
+ */
+void
+cell_vertex_shader_queue_flush(struct draw_context *draw)
+{
+#if 0
+   struct cell_context *const cell =
+       (struct cell_context *) draw->driver_private;
+   struct cell_command_vs *const vs = &cell_global.command[0].vs;
+   uint64_t *batch;
+   struct cell_array_info *array_info;
+   unsigned i, j;
+   struct cell_attribute_fetch_code *cf;
+
+   assert(draw->vs.queue_nr != 0);
+
+   /* XXX: do this on statechange: 
+    */
+   draw_update_vertex_fetch(draw);
+   cell_update_vertex_fetch(draw);
+
+
+   batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*cf));
+   batch[0] = CELL_CMD_STATE_ATTRIB_FETCH;
+   cf = (struct cell_attribute_fetch_code *) (&batch[1]);
+   cf->base = (uint64_t) cell->attrib_fetch.store;
+   cf->size = ROUNDUP16((unsigned)((void *) cell->attrib_fetch.csr 
+				   - (void *) cell->attrib_fetch.store));
+
+
+   for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) {
+      const enum pipe_format format = draw->vertex_element[i].src_format;
+      const unsigned count = ((pf_size_x(format) != 0)
+			      + (pf_size_y(format) != 0)
+			      + (pf_size_z(format) != 0)
+			      + (pf_size_w(format) != 0));
+      const unsigned size = pf_size_x(format) * count;
+
+      batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*array_info));
+
+      batch[0] = CELL_CMD_STATE_VS_ARRAY_INFO;
+
+      array_info = (struct cell_array_info *) &batch[1];
+      assert(draw->vertex_fetch.src_ptr[i] != NULL);
+      array_info->base = (uintptr_t) draw->vertex_fetch.src_ptr[i];
+      array_info->attr = i;
+      array_info->pitch = draw->vertex_fetch.pitch[i];
+      array_info->size = size;
+      array_info->function_offset = cell->attrib_fetch_offsets[i];
+   }
+
+   batch = cell_batch_alloc(cell, sizeof(batch[0])
+                            + sizeof(struct pipe_viewport_state));
+   batch[0] = CELL_CMD_STATE_VIEWPORT;
+   (void) memcpy(&batch[1], &draw->viewport,
+                 sizeof(struct pipe_viewport_state));
+
+   {
+      uint64_t uniforms = (uintptr_t) draw->user.constants;
+
+      batch = cell_batch_alloc(cell, 2 *sizeof(batch[0]));
+      batch[0] = CELL_CMD_STATE_UNIFORMS;
+      batch[1] = uniforms;
+   }
+
+   cell_batch_flush(cell);
+
+   vs->opcode = CELL_CMD_VS_EXECUTE;
+   vs->nr_attrs = draw->vertex_fetch.nr_attrs;
+
+   (void) memcpy(vs->plane, draw->plane, sizeof(draw->plane));
+   vs->nr_planes = draw->nr_planes;
+
+   for (i = 0; i < draw->vs.queue_nr; i += SPU_VERTS_PER_BATCH) {
+      const unsigned n = MIN2(SPU_VERTS_PER_BATCH, draw->vs.queue_nr - i);
+
+      for (j = 0; j < n; j++) {
+         vs->elts[j] = draw->vs.queue[i + j].elt;
+         vs->vOut[j] = (uintptr_t) draw->vs.queue[i + j].vertex;
+      }
+
+      for (/* empty */; j < SPU_VERTS_PER_BATCH; j++) {
+         vs->elts[j] = vs->elts[0];
+         vs->vOut[j] = (uintptr_t) draw->vs.queue[i + j].vertex;
+      }
+
+      vs->num_elts = n;
+      send_mbox_message(cell_global.spe_contexts[0], CELL_CMD_VS_EXECUTE);
+
+      cell_flush_int(cell, CELL_FLUSH_WAIT);
+   }
+
+   draw->vs.post_nr = draw->vs.queue_nr;
+   draw->vs.queue_nr = 0;
+#else
+   assert(0);
+#endif
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_winsys.h b/src/gallium/drivers/cell/ppu/cell_winsys.h
new file mode 100644
index 0000000000..ae2af5696b
--- /dev/null
+++ b/src/gallium/drivers/cell/ppu/cell_winsys.h
@@ -0,0 +1,50 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef CELL_WINSYS_H
+#define CELL_WINSYS_H
+
+#include "pipe/p_compiler.h"
+
+
+/**
+ * Very simple winsys at this time.
+ * Will probably eventually add SPU control info.
+ */
+struct cell_winsys
+{
+   uint preferredFormat;
+};
+
+
+extern struct cell_winsys *
+cell_get_winsys(uint format);
+
+
+
+#endif
diff --git a/src/gallium/drivers/cell/spu/.gitignore b/src/gallium/drivers/cell/spu/.gitignore
new file mode 100644
index 0000000000..2be9a2d324
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/.gitignore
@@ -0,0 +1 @@
+g3d_spu
diff --git a/src/gallium/drivers/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile
new file mode 100644
index 0000000000..116453b79c
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/Makefile
@@ -0,0 +1,82 @@
+# Gallium3D Cell driver: SPU code
+
+# This makefile builds the g3d_spu.a file that's linked into the
+# PPU code/library.
+
+
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+
+PROG = g3d
+
+PROG_SPU = $(PROG)_spu
+PROG_SPU_A = $(PROG)_spu.a
+PROG_SPU_EMBED_O = $(PROG)_spu-embed.o
+
+
+SOURCES = \
+	spu_command.c \
+	spu_dcache.c \
+	spu_funcs.c \
+	spu_main.c \
+	spu_per_fragment_op.c \
+	spu_render.c \
+	spu_texture.c \
+	spu_tile.c \
+	spu_tri.c
+
+OLD_SOURCES = \
+	spu_exec.c \
+	spu_util.c \
+	spu_vertex_fetch.c \
+	spu_vertex_shader.c
+
+
+SPU_OBJECTS = $(SOURCES:.c=.o) \
+
+SPU_ASM_OUT = $(SOURCES:.c=.s) \
+
+INCLUDE_DIRS = \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers
+
+
+.c.o:
+	$(SPU_CC) $(SPU_CFLAGS) -c $<
+
+.c.s:
+	$(SPU_CC) $(SPU_CFLAGS) -O3 -S $<
+
+
+# The .a file will be linked into the main/PPU executable
+default: $(PROG_SPU_A)
+
+$(PROG_SPU_A): $(PROG_SPU_EMBED_O)
+	$(SPU_AR) $(SPU_AR_FLAGS) $(PROG_SPU_A) $(PROG_SPU_EMBED_O)
+
+$(PROG_SPU_EMBED_O): $(PROG_SPU)
+	$(SPU_EMBED) $(SPU_EMBED_FLAGS) $(PROG_SPU) $(PROG_SPU) $(PROG_SPU_EMBED_O)
+
+$(PROG_SPU): $(SPU_OBJECTS)
+	$(SPU_CC) -o $(PROG_SPU) $(SPU_OBJECTS) $(SPU_LFLAGS)
+
+
+
+asmfiles: $(SPU_ASM_OUT)
+
+
+clean:
+	rm -f *~ *.o *.a *.d *.s $(PROG_SPU)
+
+
+
+depend: $(SOURCES)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(SOURCES) 2> /dev/null
+
+include depend
+
diff --git a/src/gallium/drivers/cell/spu/spu_colorpack.h b/src/gallium/drivers/cell/spu/spu_colorpack.h
new file mode 100644
index 0000000000..d7ce005524
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_colorpack.h
@@ -0,0 +1,145 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+
+#ifndef SPU_COLORPACK_H
+#define SPU_COLORPACK_H
+
+
+#include <transpose_matrix4x4.h>
+#include <spu_intrinsics.h>
+
+
+static INLINE unsigned int
+spu_pack_R8G8B8A8(vector float rgba)
+{
+  vector unsigned int out = spu_convtu(rgba, 32);
+
+  out = spu_shuffle(out, out, ((vector unsigned char) {
+                                  0, 4, 8, 12, 0, 0, 0, 0, 
+                                  0, 0, 0, 0, 0, 0, 0, 0 }) );
+
+  return spu_extract(out, 0);
+}
+
+
+static INLINE unsigned int
+spu_pack_A8R8G8B8(vector float rgba)
+{
+  vector unsigned int out = spu_convtu(rgba, 32);
+  out = spu_shuffle(out, out, ((vector unsigned char) {
+                                  12, 0, 4, 8, 0, 0, 0, 0, 
+                                  0, 0, 0, 0, 0, 0, 0, 0}) );
+  return spu_extract(out, 0);
+}
+
+
+static INLINE unsigned int
+spu_pack_B8G8R8A8(vector float rgba)
+{
+  vector unsigned int out = spu_convtu(rgba, 32);
+  out = spu_shuffle(out, out, ((vector unsigned char) {
+                                  8, 4, 0, 12, 0, 0, 0, 0, 
+                                  0, 0, 0, 0, 0, 0, 0, 0}) );
+  return spu_extract(out, 0);
+}
+
+
+static INLINE unsigned int
+spu_pack_color_shuffle(vector float rgba, vector unsigned char shuffle)
+{
+  vector unsigned int out = spu_convtu(rgba, 32);
+  out = spu_shuffle(out, out, shuffle);
+  return spu_extract(out, 0);
+}
+
+
+static INLINE vector float
+spu_unpack_B8G8R8A8(uint color)
+{
+   vector unsigned int color_u4 = spu_splats(color);
+   color_u4 = spu_shuffle(color_u4, color_u4,
+                          ((vector unsigned char) {
+                             2, 2, 2, 2,
+                             1, 1, 1, 1,
+                             0, 0, 0, 0,
+                             3, 3, 3, 3}) );
+   return spu_convtf(color_u4, 32);
+}
+
+
+static INLINE vector float
+spu_unpack_A8R8G8B8(uint color)
+{
+   vector unsigned int color_u4 = spu_splats(color);
+   color_u4 = spu_shuffle(color_u4, color_u4,
+                          ((vector unsigned char) {
+                             1, 1, 1, 1,
+                             2, 2, 2, 2,
+                             3, 3, 3, 3,
+                             0, 0, 0, 0}) );
+   return spu_convtf(color_u4, 32);
+}
+
+
+/**
+ * \param color_in - array of 32-bit packed ARGB colors
+ * \param color_out - returns float colors in RRRR, GGGG, BBBB, AAAA order
+ */
+static INLINE void
+spu_unpack_A8R8G8B8_transpose4(const vector unsigned int color_in[4],
+                               vector float color_out[4])
+{
+   vector unsigned int c0;
+
+   c0 = spu_shuffle(color_in[0], color_in[0],
+                    ((vector unsigned char) {
+                       1, 1, 1, 1,  2, 2, 2, 2,  3, 3, 3, 3,  0, 0, 0, 0}) );
+   color_out[0] = spu_convtf(c0, 32);
+
+   c0 = spu_shuffle(color_in[1], color_in[1],
+                    ((vector unsigned char) {
+                       1, 1, 1, 1,  2, 2, 2, 2,  3, 3, 3, 3,  0, 0, 0, 0}) );
+   color_out[1] = spu_convtf(c0, 32);
+
+   c0 = spu_shuffle(color_in[2], color_in[2],
+                    ((vector unsigned char) {
+                       1, 1, 1, 1,  2, 2, 2, 2,  3, 3, 3, 3,  0, 0, 0, 0}) );
+   color_out[2] = spu_convtf(c0, 32);
+
+   c0 = spu_shuffle(color_in[3], color_in[3],
+                    ((vector unsigned char) {
+                       1, 1, 1, 1,  2, 2, 2, 2,  3, 3, 3, 3,  0, 0, 0, 0}) );
+   color_out[3] = spu_convtf(c0, 32);
+
+   _transpose_matrix4x4(color_out, color_out);
+}
+
+
+
+#endif /* SPU_COLORPACK_H */
diff --git a/src/gallium/drivers/cell/spu/spu_command.c b/src/gallium/drivers/cell/spu/spu_command.c
new file mode 100644
index 0000000000..5c0179d954
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_command.c
@@ -0,0 +1,815 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * SPU command processing code
+ */
+
+
+#include <stdio.h>
+#include <libmisc.h>
+
+#include "pipe/p_defines.h"
+
+#include "spu_command.h"
+#include "spu_main.h"
+#include "spu_render.h"
+#include "spu_per_fragment_op.h"
+#include "spu_texture.h"
+#include "spu_tile.h"
+#include "spu_vertex_shader.h"
+#include "spu_dcache.h"
+#include "cell/common.h"
+
+
+struct spu_vs_context draw;
+
+
+/**
+ * Buffers containing dynamically generated SPU code:
+ */
+static unsigned char attribute_fetch_code_buffer[136 * PIPE_MAX_ATTRIBS]
+    ALIGN16_ATTRIB;
+
+
+
+static INLINE int
+align(int value, int alignment)
+{
+   return (value + alignment - 1) & ~(alignment - 1);
+}
+
+
+
+/**
+ * Tell the PPU that this SPU has finished copying a buffer to
+ * local store and that it may be reused by the PPU.
+ * This is done by writting a 16-byte batch-buffer-status block back into
+ * main memory (in cell_context->buffer_status[]).
+ */
+static void
+release_buffer(uint buffer)
+{
+   /* Evidently, using less than a 16-byte status doesn't work reliably */
+   static const vector unsigned int status = {CELL_BUFFER_STATUS_FREE,
+                                              CELL_BUFFER_STATUS_FREE,
+                                              CELL_BUFFER_STATUS_FREE,
+                                              CELL_BUFFER_STATUS_FREE};
+   const uint index = 4 * (spu.init.id * CELL_NUM_BUFFERS + buffer);
+   uint *dst = spu.init.buffer_status + index;
+
+   ASSERT(buffer < CELL_NUM_BUFFERS);
+
+   mfc_put((void *) &status,    /* src in local memory */
+           (unsigned int) dst,  /* dst in main memory */
+           sizeof(status),      /* size */
+           TAG_MISC,            /* tag is unimportant */
+           0, /* tid */
+           0  /* rid */);
+}
+
+
+/**
+ * Write CELL_FENCE_SIGNALLED back to the fence status qword in main memory.
+ * There's a qword of status per SPU.
+ */
+static void
+cmd_fence(struct cell_command_fence *fence_cmd)
+{
+   static const vector unsigned int status = {CELL_FENCE_SIGNALLED,
+                                              CELL_FENCE_SIGNALLED,
+                                              CELL_FENCE_SIGNALLED,
+                                              CELL_FENCE_SIGNALLED};
+   uint *dst = (uint *) fence_cmd->fence;
+   dst += 4 * spu.init.id;  /* main store/memory address, not local store */
+   ASSERT_ALIGN16(dst);
+   mfc_put((void *) &status,    /* src in local memory */
+           (unsigned int) dst,  /* dst in main memory */
+           sizeof(status),      /* size */
+           TAG_FENCE,           /* tag */
+           0, /* tid */
+           0  /* rid */);
+}
+
+
+static void
+cmd_clear_surface(const struct cell_command_clear_surface *clear)
+{
+   D_PRINTF(CELL_DEBUG_CMD, "CLEAR SURF %u to 0x%08x\n", clear->surface, clear->value);
+
+   if (clear->surface == 0) {
+      spu.fb.color_clear_value = clear->value;
+      if (spu.init.debug_flags & CELL_DEBUG_CHECKER) {
+         uint x = (spu.init.id << 4) | (spu.init.id << 12) |
+            (spu.init.id << 20) | (spu.init.id << 28);
+         spu.fb.color_clear_value ^= x;
+      }
+   }
+   else {
+      spu.fb.depth_clear_value = clear->value;
+   }
+
+#define CLEAR_OPT 1
+#if CLEAR_OPT
+
+   /* Simply set all tiles' status to CLEAR.
+    * When we actually begin rendering into a tile, we'll initialize it to
+    * the clear value.  If any tiles go untouched during the frame,
+    * really_clear_tiles() will set them to the clear value.
+    */
+   if (clear->surface == 0) {
+      memset(spu.ctile_status, TILE_STATUS_CLEAR, sizeof(spu.ctile_status));
+   }
+   else {
+      memset(spu.ztile_status, TILE_STATUS_CLEAR, sizeof(spu.ztile_status));
+   }
+
+#else
+
+   /*
+    * This path clears the whole framebuffer to the clear color right now.
+    */
+
+   /*
+   printf("SPU: %s num=%d w=%d h=%d\n",
+          __FUNCTION__, num_tiles, spu.fb.width_tiles, spu.fb.height_tiles);
+   */
+
+   /* init a single tile to the clear value */
+   if (clear->surface == 0) {
+      clear_c_tile(&spu.ctile);
+   }
+   else {
+      clear_z_tile(&spu.ztile);
+   }
+
+   /* walk over my tiles, writing the 'clear' tile's data */
+   {
+      const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles;
+      uint i;
+      for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) {
+         uint tx = i % spu.fb.width_tiles;
+         uint ty = i / spu.fb.width_tiles;
+         if (clear->surface == 0)
+            put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0);
+         else
+            put_tile(tx, ty, &spu.ztile, TAG_SURFACE_CLEAR, 1);
+      }
+   }
+
+   if (spu.init.debug_flags & CELL_DEBUG_SYNC) {
+      wait_on_mask(1 << TAG_SURFACE_CLEAR);
+   }
+
+#endif /* CLEAR_OPT */
+
+   D_PRINTF(CELL_DEBUG_CMD, "CLEAR SURF done\n");
+}
+
+
+static void
+cmd_release_verts(const struct cell_command_release_verts *release)
+{
+   D_PRINTF(CELL_DEBUG_CMD, "RELEASE VERTS %u\n", release->vertex_buf);
+   ASSERT(release->vertex_buf != ~0U);
+   release_buffer(release->vertex_buf);
+}
+
+
+/**
+ * Process a CELL_CMD_STATE_FRAGMENT_OPS command.
+ * This involves installing new fragment ops SPU code.
+ * If this function is never called, we'll use a regular C fallback function
+ * for fragment processing.
+ */
+static void
+cmd_state_fragment_ops(const struct cell_command_fragment_ops *fops)
+{
+   D_PRINTF(CELL_DEBUG_CMD, "CMD_STATE_FRAGMENT_OPS\n");
+
+   /* Copy state info (for fallback case only - this will eventually
+    * go away when the fallback case goes away)
+    */
+   memcpy(&spu.depth_stencil_alpha, &fops->dsa, sizeof(fops->dsa));
+   memcpy(&spu.blend, &fops->blend, sizeof(fops->blend));
+   memcpy(&spu.blend_color, &fops->blend_color, sizeof(fops->blend_color));
+
+   /* Make sure the SPU knows which buffers it's expected to read when
+    * it's told to pull tiles.
+    */
+   spu.read_depth_stencil = (spu.depth_stencil_alpha.depth.enabled || spu.depth_stencil_alpha.stencil[0].enabled);
+
+   /* If we're forcing the fallback code to be used (for debug purposes),
+    * install that.  Otherwise install the incoming SPU code.
+    */
+   if ((spu.init.debug_flags & CELL_DEBUG_FRAGMENT_OP_FALLBACK) != 0) {
+      static unsigned int warned = 0;
+      if (!warned) {
+         fprintf(stderr, "Cell Warning: using fallback per-fragment code\n");
+         warned = 1;
+      }
+      /* The following two lines aren't really necessary if you
+       * know the debug flags won't change during a run, and if you
+       * know that the function pointers are initialized correctly.
+       * We set them here to allow a person to change the debug
+       * flags during a run (from inside a debugger).
+       */
+      spu.fragment_ops[CELL_FACING_FRONT] = spu_fallback_fragment_ops;
+      spu.fragment_ops[CELL_FACING_BACK] = spu_fallback_fragment_ops;
+      return;
+   }
+
+   /* Make sure the SPU code buffer is large enough to hold the incoming code.
+    * Note that we *don't* use align_malloc() and align_free(), because
+    * those utility functions are *not* available in SPU code.
+    * */
+   if (spu.fragment_ops_code_size < fops->total_code_size) {
+      if (spu.fragment_ops_code != NULL) {
+         free(spu.fragment_ops_code);
+      }
+      spu.fragment_ops_code_size = fops->total_code_size;
+      spu.fragment_ops_code = malloc(fops->total_code_size);
+      if (spu.fragment_ops_code == NULL) {
+         /* Whoops. */
+         fprintf(stderr, "CELL Warning: failed to allocate fragment ops code (%d bytes) - using fallback\n", fops->total_code_size);
+         spu.fragment_ops_code = NULL;
+         spu.fragment_ops_code_size = 0;
+         spu.fragment_ops[CELL_FACING_FRONT] = spu_fallback_fragment_ops;
+         spu.fragment_ops[CELL_FACING_BACK] = spu_fallback_fragment_ops;
+         return;
+      }
+   }
+
+   /* Copy the SPU code from the command buffer to the spu buffer */
+   memcpy(spu.fragment_ops_code, fops->code, fops->total_code_size);
+
+   /* Set the pointers for the front-facing and back-facing fragments
+    * to the specified offsets within the code.  Note that if the
+    * front-facing and back-facing code are the same, they'll have
+    * the same offset.
+    */
+   spu.fragment_ops[CELL_FACING_FRONT] = (spu_fragment_ops_func) &spu.fragment_ops_code[fops->front_code_index];
+   spu.fragment_ops[CELL_FACING_BACK] = (spu_fragment_ops_func) &spu.fragment_ops_code[fops->back_code_index];
+}
+
+static void
+cmd_state_fragment_program(const struct cell_command_fragment_program *fp)
+{
+   D_PRINTF(CELL_DEBUG_CMD, "CMD_STATE_FRAGMENT_PROGRAM\n");
+   /* Copy SPU code from batch buffer to spu buffer */
+   memcpy(spu.fragment_program_code, fp->code,
+          SPU_MAX_FRAGMENT_PROGRAM_INSTS * 4);
+#if 01
+   /* Point function pointer at new code */
+   spu.fragment_program = (spu_fragment_program_func)spu.fragment_program_code;
+#endif
+}
+
+
+static uint
+cmd_state_fs_constants(const qword *buffer, uint pos)
+{
+   const uint num_const = spu_extract((vector unsigned int)buffer[pos+1], 0);
+   const float *constants = (const float *) &buffer[pos+2];
+   uint i;
+
+   D_PRINTF(CELL_DEBUG_CMD, "CMD_STATE_FS_CONSTANTS (%u)\n", num_const);
+
+   /* Expand each float to float[4] for SOA execution */
+   for (i = 0; i < num_const; i++) {
+      D_PRINTF(CELL_DEBUG_CMD, "  const[%u] = %f\n", i, constants[i]);
+      spu.constants[i] = spu_splats(constants[i]);
+   }
+
+   /* return new buffer pos (in 16-byte words) */
+   return pos + 2 + (ROUNDUP16(num_const * sizeof(float)) / 16);
+}
+
+
+static void
+cmd_state_framebuffer(const struct cell_command_framebuffer *cmd)
+{
+   D_PRINTF(CELL_DEBUG_CMD, "FRAMEBUFFER: %d x %d at %p, cformat 0x%x  zformat 0x%x\n",
+             cmd->width,
+             cmd->height,
+             cmd->color_start,
+             cmd->color_format,
+             cmd->depth_format);
+
+   ASSERT_ALIGN16(cmd->color_start);
+   ASSERT_ALIGN16(cmd->depth_start);
+
+   spu.fb.color_start = cmd->color_start;
+   spu.fb.depth_start = cmd->depth_start;
+   spu.fb.color_format = cmd->color_format;
+   spu.fb.depth_format = cmd->depth_format;
+   spu.fb.width = cmd->width;
+   spu.fb.height = cmd->height;
+   spu.fb.width_tiles = (spu.fb.width + TILE_SIZE - 1) / TILE_SIZE;
+   spu.fb.height_tiles = (spu.fb.height + TILE_SIZE - 1) / TILE_SIZE;
+
+   switch (spu.fb.depth_format) {
+   case PIPE_FORMAT_Z32_UNORM:
+      spu.fb.zsize = 4;
+      spu.fb.zscale = (float) 0xffffffffu;
+      break;
+   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_Z24X8_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+      spu.fb.zsize = 4;
+      spu.fb.zscale = (float) 0x00ffffffu;
+      break;
+   case PIPE_FORMAT_Z16_UNORM:
+      spu.fb.zsize = 2;
+      spu.fb.zscale = (float) 0xffffu;
+      break;
+   default:
+      spu.fb.zsize = 0;
+      break;
+   }
+}
+
+
+/**
+ * Tex texture mask_s/t and scale_s/t fields depend on the texture size and
+ * sampler wrap modes.
+ */
+static void
+update_tex_masks(struct spu_texture *texture,
+                 const struct pipe_sampler_state *sampler)
+{
+   uint i;
+
+   for (i = 0; i < CELL_MAX_TEXTURE_LEVELS; i++) {
+      int width = texture->level[i].width;
+      int height = texture->level[i].height;
+
+      if (sampler->wrap_s == PIPE_TEX_WRAP_REPEAT)
+         texture->level[i].mask_s = spu_splats(width - 1);
+      else
+         texture->level[i].mask_s = spu_splats(~0);
+
+      if (sampler->wrap_t == PIPE_TEX_WRAP_REPEAT)
+         texture->level[i].mask_t = spu_splats(height - 1);
+      else
+         texture->level[i].mask_t = spu_splats(~0);
+
+      if (sampler->normalized_coords) {
+         texture->level[i].scale_s = spu_splats((float) width);
+         texture->level[i].scale_t = spu_splats((float) height);
+      }
+      else {
+         texture->level[i].scale_s = spu_splats(1.0f);
+         texture->level[i].scale_t = spu_splats(1.0f);
+      }
+   }
+}
+
+
+static void
+cmd_state_sampler(const struct cell_command_sampler *sampler)
+{
+   uint unit = sampler->unit;
+
+   D_PRINTF(CELL_DEBUG_CMD, "SAMPLER [%u]\n", unit);
+
+   spu.sampler[unit] = sampler->state;
+
+   switch (spu.sampler[unit].min_img_filter) {
+   case PIPE_TEX_FILTER_LINEAR:
+      spu.min_sample_texture_2d[unit] = sample_texture_2d_bilinear;
+      break;
+   case PIPE_TEX_FILTER_ANISO:
+      /* fall-through, for now */
+   case PIPE_TEX_FILTER_NEAREST:
+      spu.min_sample_texture_2d[unit] = sample_texture_2d_nearest;
+      break;
+   default:
+      ASSERT(0);
+   }
+
+   switch (spu.sampler[sampler->unit].mag_img_filter) {
+   case PIPE_TEX_FILTER_LINEAR:
+      spu.mag_sample_texture_2d[unit] = sample_texture_2d_bilinear;
+      break;
+   case PIPE_TEX_FILTER_ANISO:
+      /* fall-through, for now */
+   case PIPE_TEX_FILTER_NEAREST:
+      spu.mag_sample_texture_2d[unit] = sample_texture_2d_nearest;
+      break;
+   default:
+      ASSERT(0);
+   }
+
+   switch (spu.sampler[sampler->unit].min_mip_filter) {
+   case PIPE_TEX_MIPFILTER_NEAREST:
+   case PIPE_TEX_MIPFILTER_LINEAR:
+      spu.sample_texture_2d[unit] = sample_texture_2d_lod;
+      break;
+   case PIPE_TEX_MIPFILTER_NONE:
+      spu.sample_texture_2d[unit] = spu.mag_sample_texture_2d[unit];
+      break;
+   default:
+      ASSERT(0);
+   }
+
+   update_tex_masks(&spu.texture[unit], &spu.sampler[unit]);
+}
+
+
+static void
+cmd_state_texture(const struct cell_command_texture *texture)
+{
+   const uint unit = texture->unit;
+   uint i;
+
+   D_PRINTF(CELL_DEBUG_CMD, "TEXTURE [%u]\n", texture->unit);
+
+   spu.texture[unit].max_level = 0;
+   spu.texture[unit].target = texture->target;
+
+   for (i = 0; i < CELL_MAX_TEXTURE_LEVELS; i++) {
+      uint width = texture->width[i];
+      uint height = texture->height[i];
+      uint depth = texture->depth[i];
+
+      D_PRINTF(CELL_DEBUG_CMD, "  LEVEL %u: at %p  size[0] %u x %u\n", i,
+             texture->start[i], texture->width[i], texture->height[i]);
+
+      spu.texture[unit].level[i].start = texture->start[i];
+      spu.texture[unit].level[i].width = width;
+      spu.texture[unit].level[i].height = height;
+      spu.texture[unit].level[i].depth = depth;
+
+      spu.texture[unit].level[i].tiles_per_row =
+         (width + TILE_SIZE - 1) / TILE_SIZE;
+
+      spu.texture[unit].level[i].bytes_per_image =
+         4 * align(width, TILE_SIZE) * align(height, TILE_SIZE) * depth;
+
+      spu.texture[unit].level[i].max_s = spu_splats((int) width - 1);
+      spu.texture[unit].level[i].max_t = spu_splats((int) height - 1);
+
+      if (texture->start[i])
+         spu.texture[unit].max_level = i;
+   }
+
+   update_tex_masks(&spu.texture[unit], &spu.sampler[unit]);
+}
+
+
+static void
+cmd_state_vertex_info(const struct vertex_info *vinfo)
+{
+   D_PRINTF(CELL_DEBUG_CMD, "VERTEX_INFO num_attribs=%u\n", vinfo->num_attribs);
+   ASSERT(vinfo->num_attribs >= 1);
+   ASSERT(vinfo->num_attribs <= 8);
+   memcpy(&spu.vertex_info, vinfo, sizeof(*vinfo));
+}
+
+
+static void
+cmd_state_vs_array_info(const struct cell_array_info *vs_info)
+{
+   const unsigned attr = vs_info->attr;
+
+   ASSERT(attr < PIPE_MAX_ATTRIBS);
+   draw.vertex_fetch.src_ptr[attr] = vs_info->base;
+   draw.vertex_fetch.pitch[attr] = vs_info->pitch;
+   draw.vertex_fetch.size[attr] = vs_info->size;
+   draw.vertex_fetch.code_offset[attr] = vs_info->function_offset;
+   draw.vertex_fetch.dirty = 1;
+}
+
+
+static void
+cmd_state_attrib_fetch(const struct cell_attribute_fetch_code *code)
+{
+   mfc_get(attribute_fetch_code_buffer,
+           (unsigned int) code->base,  /* src */
+           code->size,
+           TAG_BATCH_BUFFER,
+           0, /* tid */
+           0  /* rid */);
+   wait_on_mask(1 << TAG_BATCH_BUFFER);
+
+   draw.vertex_fetch.code = attribute_fetch_code_buffer;
+}
+
+
+static void
+cmd_finish(void)
+{
+   D_PRINTF(CELL_DEBUG_CMD, "FINISH\n");
+   really_clear_tiles(0);
+   /* wait for all outstanding DMAs to finish */
+   mfc_write_tag_mask(~0);
+   mfc_read_tag_status_all();
+   /* send mbox message to PPU */
+   spu_write_out_mbox(CELL_CMD_FINISH);
+}
+
+
+/**
+ * Execute a batch of commands which was sent to us by the PPU.
+ * See the cell_emit_state.c code to see where the commands come from.
+ *
+ * The opcode param encodes the location of the buffer and its size.
+ */
+static void
+cmd_batch(uint opcode)
+{
+   const uint buf = (opcode >> 8) & 0xff;
+   uint size = (opcode >> 16);
+   qword buffer[CELL_BUFFER_SIZE / 16] ALIGN16_ATTRIB;
+   const unsigned usize = ROUNDUP16(size) / sizeof(buffer[0]);
+   uint pos;
+
+   D_PRINTF(CELL_DEBUG_CMD, "BATCH buffer %u, len %u, from %p\n",
+             buf, size, spu.init.buffers[buf]);
+
+   ASSERT((opcode & CELL_CMD_OPCODE_MASK) == CELL_CMD_BATCH);
+
+   ASSERT_ALIGN16(spu.init.buffers[buf]);
+
+   size = ROUNDUP16(size);
+
+   ASSERT_ALIGN16(spu.init.buffers[buf]);
+
+   mfc_get(buffer,  /* dest */
+           (unsigned int) spu.init.buffers[buf],  /* src */
+           size,
+           TAG_BATCH_BUFFER,
+           0, /* tid */
+           0  /* rid */);
+   wait_on_mask(1 << TAG_BATCH_BUFFER);
+
+   /* Tell PPU we're done copying the buffer to local store */
+   D_PRINTF(CELL_DEBUG_CMD, "release batch buf %u\n", buf);
+   release_buffer(buf);
+
+   /*
+    * Loop over commands in the batch buffer
+    */
+   for (pos = 0; pos < usize; /* no incr */) {
+      switch (si_to_uint(buffer[pos])) {
+      /*
+       * rendering commands
+       */
+      case CELL_CMD_CLEAR_SURFACE:
+         {
+            struct cell_command_clear_surface *clr
+               = (struct cell_command_clear_surface *) &buffer[pos];
+            cmd_clear_surface(clr);
+            pos += sizeof(*clr) / 16;
+         }
+         break;
+      case CELL_CMD_RENDER:
+         {
+            struct cell_command_render *render
+               = (struct cell_command_render *) &buffer[pos];
+            uint pos_incr;
+            cmd_render(render, &pos_incr);
+            pos += ((pos_incr+1)&~1) / 2; // should 'fix' cmd_render return
+         }
+         break;
+      /*
+       * state-update commands
+       */
+      case CELL_CMD_STATE_FRAMEBUFFER:
+         {
+            struct cell_command_framebuffer *fb
+               = (struct cell_command_framebuffer *) &buffer[pos];
+            cmd_state_framebuffer(fb);
+            pos += sizeof(*fb) / 16;
+         }
+         break;
+      case CELL_CMD_STATE_FRAGMENT_OPS:
+         {
+            struct cell_command_fragment_ops *fops
+               = (struct cell_command_fragment_ops *) &buffer[pos];
+            cmd_state_fragment_ops(fops);
+            /* This is a variant-sized command */
+            pos += ROUNDUP16(sizeof(*fops) + fops->total_code_size) / 16;
+         }
+         break;
+      case CELL_CMD_STATE_FRAGMENT_PROGRAM:
+         {
+            struct cell_command_fragment_program *fp
+               = (struct cell_command_fragment_program *) &buffer[pos];
+            cmd_state_fragment_program(fp);
+            pos += sizeof(*fp) / 16;
+         }
+         break;
+      case CELL_CMD_STATE_FS_CONSTANTS:
+         pos = cmd_state_fs_constants(buffer, pos);
+         break;
+      case CELL_CMD_STATE_RASTERIZER:
+         {
+            struct cell_command_rasterizer *rast =
+               (struct cell_command_rasterizer *) &buffer[pos];
+            spu.rasterizer = rast->rasterizer;
+            pos += sizeof(*rast) / 16;
+         }
+         break;
+      case CELL_CMD_STATE_SAMPLER:
+         {
+            struct cell_command_sampler *sampler
+               = (struct cell_command_sampler *) &buffer[pos];
+            cmd_state_sampler(sampler);
+            pos += sizeof(*sampler) / 16;
+         }
+         break;
+      case CELL_CMD_STATE_TEXTURE:
+         {
+            struct cell_command_texture *texture
+               = (struct cell_command_texture *) &buffer[pos];
+            cmd_state_texture(texture);
+            pos += sizeof(*texture) / 16;
+         }
+         break;
+      case CELL_CMD_STATE_VERTEX_INFO:
+         cmd_state_vertex_info((struct vertex_info *) &buffer[pos+1]);
+         pos += 1 + ROUNDUP16(sizeof(struct vertex_info)) / 16;
+         break;
+      case CELL_CMD_STATE_VIEWPORT:
+         (void) memcpy(& draw.viewport, &buffer[pos+1],
+                       sizeof(struct pipe_viewport_state));
+         pos += 1 + ROUNDUP16(sizeof(struct pipe_viewport_state)) / 16;
+         break;
+      case CELL_CMD_STATE_UNIFORMS:
+         draw.constants = (const float (*)[4]) (uintptr_t)spu_extract((vector unsigned int)buffer[pos+1],0);
+         pos += 2;
+         break;
+      case CELL_CMD_STATE_VS_ARRAY_INFO:
+         cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]);
+         pos += 1 + ROUNDUP16(sizeof(struct cell_array_info)) / 16;
+         break;
+      case CELL_CMD_STATE_BIND_VS:
+#if 0
+         spu_bind_vertex_shader(&draw,
+                                (struct cell_shader_info *) &buffer[pos+1]);
+#endif
+         pos += 1 + ROUNDUP16(sizeof(struct cell_shader_info)) / 16;
+         break;
+      case CELL_CMD_STATE_ATTRIB_FETCH:
+         cmd_state_attrib_fetch((struct cell_attribute_fetch_code *)
+                                &buffer[pos+1]);
+         pos += 1 + ROUNDUP16(sizeof(struct cell_attribute_fetch_code)) / 16;
+         break;
+      /*
+       * misc commands
+       */
+      case CELL_CMD_FINISH:
+         cmd_finish();
+         pos += 1;
+         break;
+      case CELL_CMD_FENCE:
+         {
+            struct cell_command_fence *fence_cmd =
+               (struct cell_command_fence *) &buffer[pos];
+            cmd_fence(fence_cmd);
+            pos += sizeof(*fence_cmd) / 16;
+         }
+         break;
+      case CELL_CMD_RELEASE_VERTS:
+         {
+            struct cell_command_release_verts *release
+               = (struct cell_command_release_verts *) &buffer[pos];
+            cmd_release_verts(release);
+            pos += sizeof(*release) / 16;
+         }
+         break;
+      case CELL_CMD_FLUSH_BUFFER_RANGE: {
+	 struct cell_buffer_range *br = (struct cell_buffer_range *)
+	     &buffer[pos+1];
+
+	 spu_dcache_mark_dirty((unsigned) br->base, br->size);
+         pos += 1 + ROUNDUP16(sizeof(struct cell_buffer_range)) / 16;
+	 break;
+      }
+      default:
+         printf("SPU %u: bad opcode: 0x%x\n", spu.init.id, si_to_uint(buffer[pos]));
+         ASSERT(0);
+         break;
+      }
+   }
+
+   D_PRINTF(CELL_DEBUG_CMD, "BATCH complete\n");
+}
+
+
+#define PERF 0
+
+
+/**
+ * Main loop for SPEs: Get a command, execute it, repeat.
+ */
+void
+command_loop(void)
+{
+   int exitFlag = 0;
+   uint t0, t1;
+
+   D_PRINTF(CELL_DEBUG_CMD, "Enter command loop\n");
+
+   while (!exitFlag) {
+      unsigned opcode;
+
+      D_PRINTF(CELL_DEBUG_CMD, "Wait for cmd...\n");
+
+      if (PERF)
+         spu_write_decrementer(~0);
+
+      /* read/wait from mailbox */
+      opcode = (unsigned int) spu_read_in_mbox();
+      D_PRINTF(CELL_DEBUG_CMD, "got cmd 0x%x\n", opcode);
+
+      if (PERF)
+         t0 = spu_read_decrementer();
+
+      switch (opcode & CELL_CMD_OPCODE_MASK) {
+      case CELL_CMD_EXIT:
+         D_PRINTF(CELL_DEBUG_CMD, "EXIT\n");
+         exitFlag = 1;
+         break;
+      case CELL_CMD_VS_EXECUTE:
+#if 0
+         spu_execute_vertex_shader(&draw, &cmd.vs);
+#endif
+         break;
+      case CELL_CMD_BATCH:
+         cmd_batch(opcode);
+         break;
+      default:
+         printf("Bad opcode 0x%x!\n", opcode & CELL_CMD_OPCODE_MASK);
+      }
+
+      if (PERF) {
+         t1 = spu_read_decrementer();
+         printf("wait mbox time: %gms   batch time: %gms\n",
+                (~0u - t0) * spu.init.inv_timebase,
+                (t0 - t1) * spu.init.inv_timebase);
+      }
+   }
+
+   D_PRINTF(CELL_DEBUG_CMD, "Exit command loop\n");
+
+   if (spu.init.debug_flags & CELL_DEBUG_CACHE)
+      spu_dcache_report();
+}
+
+/* Initialize this module; we manage the fragment ops buffer here. */
+void
+spu_command_init(void)
+{
+   /* Install default/fallback fragment processing function.
+    * This will normally be overriden by a code-gen'd function
+    * unless CELL_FORCE_FRAGMENT_OPS_FALLBACK is set.
+    */
+   spu.fragment_ops[CELL_FACING_FRONT] = spu_fallback_fragment_ops;
+   spu.fragment_ops[CELL_FACING_BACK] = spu_fallback_fragment_ops;
+
+   /* Set up the basic empty buffer for code-gen'ed fragment ops */
+   spu.fragment_ops_code = NULL;
+   spu.fragment_ops_code_size = 0;
+}
+
+void
+spu_command_close(void)
+{
+   /* Deallocate the code-gen buffer for fragment ops, and reset the
+    * fragment ops functions to their initial setting (just to leave
+    * things in a good state).
+    */
+   if (spu.fragment_ops_code != NULL) {
+      free(spu.fragment_ops_code);
+   }
+   spu_command_init();
+}
diff --git a/src/gallium/drivers/cell/spu/spu_command.h b/src/gallium/drivers/cell/spu/spu_command.h
new file mode 100644
index 0000000000..83dcdade28
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_command.h
@@ -0,0 +1,35 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+extern void
+command_loop(void);
+
+extern void
+spu_command_init(void);
+
+extern void
+spu_command_close(void);
diff --git a/src/gallium/drivers/cell/spu/spu_dcache.c b/src/gallium/drivers/cell/spu/spu_dcache.c
new file mode 100644
index 0000000000..a6d67634fd
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_dcache.c
@@ -0,0 +1,145 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "cell/common.h"
+#include "spu_main.h"
+#include "spu_dcache.h"
+
+#define CACHELINE_LOG2SIZE    7
+#define LINE_SIZE             (1U << 7)
+#define ALIGN_MASK            (~(LINE_SIZE - 1))
+
+#define CACHE_NAME            data
+#define CACHED_TYPE           qword
+#define CACHE_TYPE            CACHE_TYPE_RO
+#define CACHE_SET_TAGID(set)  (((set) & 0x03) + TAG_DCACHE0)
+#define CACHE_LOG2NNWAY       2
+#define CACHE_LOG2NSETS       6
+#ifdef DEBUG
+#define CACHE_STATS           1
+#endif
+#include <cache-api.h>
+
+/* Yes folks, this is ugly.
+ */
+#undef CACHE_NWAY
+#undef CACHE_NSETS
+#define CACHE_NAME            data
+#define CACHE_NWAY            4
+#define CACHE_NSETS           (1U << 6)
+
+
+/**
+ * Fetch between arbitrary number of bytes from an unaligned address
+ *
+ * \param dst   Destination data buffer
+ * \param ea    Main memory effective address of source data
+ * \param size  Number of bytes to read
+ *
+ * \warning
+ * As is hinted by the type of the \c dst pointer, this function writes
+ * multiples of 16-bytes.
+ */
+void
+spu_dcache_fetch_unaligned(qword *dst, unsigned ea, unsigned size)
+{
+   const int shift = ea & 0x0f;
+   const unsigned read_size = ROUNDUP16(size + shift);
+   const unsigned last_read = ROUNDUP16(ea + size);
+   const qword *const last_write = dst + (ROUNDUP16(size) / 16);
+   unsigned i;
+
+
+   if (shift == 0) {
+      /* Data is already aligned.  Fetch directly into the destination buffer.
+       */
+      for (i = 0; i < size; i += 16) {
+         *(dst++) = cache_rd(data, ea + i);
+      }
+   } else {
+      qword hi;
+
+
+      /* Please exercise extreme caution when modifying this code.  This code
+       * must not read past the end of the page containing the source data,
+       * and it must not write more than ((size + 15) / 16) qwords to the
+       * destination buffer.
+       */
+      ea &= ~0x0f;
+      hi = cache_rd(data, ea);
+      for (i = 16; i < read_size; i += 16) {
+         qword lo = cache_rd(data, ea + i);
+
+         *(dst++) = si_or((qword) spu_slqwbyte(hi, shift),
+                          (qword) spu_rlmaskqwbyte(lo, shift - 16));
+         hi = lo;
+      }
+
+      if (dst != last_write) {
+         *(dst++) = si_or((qword) spu_slqwbyte(hi, shift), si_il(0));
+      }
+   }
+   
+   ASSERT((ea + i) == last_read);
+   ASSERT(dst == last_write);
+}
+
+
+/**
+ * Notify the cache that a range of main memory may have been modified
+ */
+void
+spu_dcache_mark_dirty(unsigned ea, unsigned size)
+{
+   unsigned i;
+   const unsigned aligned_start = (ea & ALIGN_MASK);
+   const unsigned aligned_end = (ea + size + (LINE_SIZE - 1)) 
+       & ALIGN_MASK;
+
+
+   for (i = 0; i < (CACHE_NWAY * CACHE_NSETS); i++) {
+      const unsigned entry = __cache_dir[i];
+      const unsigned addr = entry & ~0x0f;
+
+      __cache_dir[i] = ((addr >= aligned_start) && (addr < aligned_end))
+          ? (entry & ~CACHELINE_VALID) : entry;
+   }
+}
+
+
+/**
+ * Print cache utilization report
+ */
+void
+spu_dcache_report(void)
+{
+#ifdef CACHE_STATS
+   if (spu.init.id == 0) {
+      printf("SPU 0: Texture cache report:\n");
+      cache_pr_stats(data);
+   }
+#endif
+}
+
+
diff --git a/src/gallium/drivers/cell/spu/spu_dcache.h b/src/gallium/drivers/cell/spu/spu_dcache.h
new file mode 100644
index 0000000000..39a19eb31b
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_dcache.h
@@ -0,0 +1,37 @@
+/*
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef SPU_DCACHE_H
+#define SPU_DCACHE_H
+
+extern void
+spu_dcache_fetch_unaligned(qword *dst, unsigned ea, unsigned size);
+
+extern void
+spu_dcache_mark_dirty(unsigned ea, unsigned size);
+
+extern void
+spu_dcache_report(void);
+
+#endif /* SPU_DCACHE_H */
diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c
new file mode 100644
index 0000000000..e27df2dfb3
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_exec.c
@@ -0,0 +1,1933 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * TGSI interpretor/executor.
+ *
+ * Flow control information:
+ *
+ * Since we operate on 'quads' (4 pixels or 4 vertices in parallel)
+ * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special
+ * care since a condition may be true for some quad components but false
+ * for other components.
+ *
+ * We basically execute all statements (even if they're in the part of
+ * an IF/ELSE clause that's "not taken") and use a special mask to
+ * control writing to destination registers.  This is the ExecMask.
+ * See store_dest().
+ *
+ * The ExecMask is computed from three other masks (CondMask, LoopMask and
+ * ContMask) which are controlled by the flow control instructions (namely:
+ * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT).
+ *
+ *
+ * Authors:
+ *   Michal Krol
+ *   Brian Paul
+ */
+
+#include <transpose_matrix4x4.h>
+#include <simdmath/ceilf4.h>
+#include <simdmath/cosf4.h>
+#include <simdmath/divf4.h>
+#include <simdmath/floorf4.h>
+#include <simdmath/log2f4.h>
+#include <simdmath/powf4.h>
+#include <simdmath/sinf4.h>
+#include <simdmath/sqrtf4.h>
+#include <simdmath/truncf4.h>
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+#include "spu_exec.h"
+#include "spu_main.h"
+#include "spu_vertex_shader.h"
+#include "spu_dcache.h"
+#include "cell/common.h"
+
+#define TILE_TOP_LEFT     0
+#define TILE_TOP_RIGHT    1
+#define TILE_BOTTOM_LEFT  2
+#define TILE_BOTTOM_RIGHT 3
+
+/*
+ * Shorthand locations of various utility registers (_I = Index, _C = Channel)
+ */
+#define TEMP_0_I           TGSI_EXEC_TEMP_00000000_I
+#define TEMP_0_C           TGSI_EXEC_TEMP_00000000_C
+#define TEMP_7F_I          TGSI_EXEC_TEMP_7FFFFFFF_I
+#define TEMP_7F_C          TGSI_EXEC_TEMP_7FFFFFFF_C
+#define TEMP_80_I          TGSI_EXEC_TEMP_80000000_I
+#define TEMP_80_C          TGSI_EXEC_TEMP_80000000_C
+#define TEMP_FF_I          TGSI_EXEC_TEMP_FFFFFFFF_I
+#define TEMP_FF_C          TGSI_EXEC_TEMP_FFFFFFFF_C
+#define TEMP_1_I           TGSI_EXEC_TEMP_ONE_I
+#define TEMP_1_C           TGSI_EXEC_TEMP_ONE_C
+#define TEMP_2_I           TGSI_EXEC_TEMP_TWO_I
+#define TEMP_2_C           TGSI_EXEC_TEMP_TWO_C
+#define TEMP_128_I         TGSI_EXEC_TEMP_128_I
+#define TEMP_128_C         TGSI_EXEC_TEMP_128_C
+#define TEMP_M128_I        TGSI_EXEC_TEMP_MINUS_128_I
+#define TEMP_M128_C        TGSI_EXEC_TEMP_MINUS_128_C
+#define TEMP_KILMASK_I     TGSI_EXEC_TEMP_KILMASK_I
+#define TEMP_KILMASK_C     TGSI_EXEC_TEMP_KILMASK_C
+#define TEMP_OUTPUT_I      TGSI_EXEC_TEMP_OUTPUT_I
+#define TEMP_OUTPUT_C      TGSI_EXEC_TEMP_OUTPUT_C
+#define TEMP_PRIMITIVE_I   TGSI_EXEC_TEMP_PRIMITIVE_I
+#define TEMP_PRIMITIVE_C   TGSI_EXEC_TEMP_PRIMITIVE_C
+#define TEMP_R0            TGSI_EXEC_TEMP_R0
+
+#define FOR_EACH_CHANNEL(CHAN)\
+   for (CHAN = 0; CHAN < 4; CHAN++)
+
+#define IS_CHANNEL_ENABLED(INST, CHAN)\
+   ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN)))
+
+#define IS_CHANNEL_ENABLED2(INST, CHAN)\
+   ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN)))
+
+#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\
+   FOR_EACH_CHANNEL( CHAN )\
+      if (IS_CHANNEL_ENABLED( INST, CHAN ))
+
+#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\
+   FOR_EACH_CHANNEL( CHAN )\
+      if (IS_CHANNEL_ENABLED2( INST, CHAN ))
+
+
+/** The execution mask depends on the conditional mask and the loop mask */
+#define UPDATE_EXEC_MASK(MACH) \
+      MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask
+
+
+#define CHAN_X  0
+#define CHAN_Y  1
+#define CHAN_Z  2
+#define CHAN_W  3
+
+
+
+/**
+ * Initialize machine state by expanding tokens to full instructions,
+ * allocating temporary storage, setting up constants, etc.
+ * After this, we can call spu_exec_machine_run() many times.
+ */
+void
+spu_exec_machine_init(struct spu_exec_machine *mach,
+                      uint numSamplers,
+                      struct spu_sampler *samplers,
+                      unsigned processor)
+{
+   const qword zero = si_il(0);
+   const qword not_zero = si_il(~0);
+
+   (void) numSamplers;
+   mach->Samplers = samplers;
+   mach->Processor = processor;
+   mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS];
+
+   /* Setup constants. */
+   mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q = zero;
+   mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].q = not_zero;
+   mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].q = si_shli(not_zero, -1);
+   mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].q = si_shli(not_zero, 31);
+
+   mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q = (qword) spu_splats(1.0f);
+   mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].q = (qword) spu_splats(2.0f);
+   mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].q = (qword) spu_splats(128.0f);
+   mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].q = (qword) spu_splats(-128.0f);
+}
+
+
+static INLINE qword
+micro_abs(qword src)
+{
+   return si_rotmi(si_shli(src, 1), -1);
+}
+
+static INLINE qword
+micro_ceil(qword src)
+{
+   return (qword) _ceilf4((vec_float4) src);
+}
+
+static INLINE qword
+micro_cos(qword src)
+{
+   return (qword) _cosf4((vec_float4) src);
+}
+
+static const qword br_shuf = {
+   TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1,
+   TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3,
+   TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1,
+   TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3,
+   TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1,
+   TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3,
+   TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1,
+   TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3,
+};
+
+static const qword bl_shuf = {
+   TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1,
+   TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3,
+   TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1,
+   TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3,
+   TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1,
+   TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3,
+   TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1,
+   TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3,
+};
+
+static const qword tl_shuf = {
+   TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1,
+   TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3,
+   TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1,
+   TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3,
+   TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1,
+   TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3,
+   TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1,
+   TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3,
+};
+
+static qword
+micro_ddx(qword src)
+{
+   qword bottom_right = si_shufb(src, src, br_shuf);
+   qword bottom_left = si_shufb(src, src, bl_shuf);
+
+   return si_fs(bottom_right, bottom_left);
+}
+
+static qword
+micro_ddy(qword src)
+{
+   qword top_left = si_shufb(src, src, tl_shuf);
+   qword bottom_left = si_shufb(src, src, bl_shuf);
+
+   return si_fs(top_left, bottom_left);
+}
+
+static INLINE qword
+micro_div(qword src0, qword src1)
+{
+   return (qword) _divf4((vec_float4) src0, (vec_float4) src1);
+}
+
+static qword
+micro_flr(qword src)
+{
+   return (qword) _floorf4((vec_float4) src);
+}
+
+static qword
+micro_frc(qword src)
+{
+   return si_fs(src, (qword) _floorf4((vec_float4) src));
+}
+
+static INLINE qword
+micro_ge(qword src0, qword src1)
+{
+   return si_or(si_fceq(src0, src1), si_fcgt(src0, src1));
+}
+
+static qword
+micro_lg2(qword src)
+{
+   return (qword) _log2f4((vec_float4) src);
+}
+
+static INLINE qword
+micro_lt(qword src0, qword src1)
+{
+   const qword tmp = si_or(si_fceq(src0, src1), si_fcgt(src0, src1));
+
+   return si_xori(tmp, 0xff);
+}
+
+static INLINE qword
+micro_max(qword src0, qword src1)
+{
+   return si_selb(src1, src0, si_fcgt(src0, src1));
+}
+
+static INLINE qword
+micro_min(qword src0, qword src1)
+{
+   return si_selb(src0, src1, si_fcgt(src0, src1));
+}
+
+static qword
+micro_neg(qword src)
+{
+   return si_xor(src, (qword) spu_splats(0x80000000));
+}
+
+static qword
+micro_set_sign(qword src)
+{
+   return si_or(src, (qword) spu_splats(0x80000000));
+}
+
+static qword
+micro_pow(qword src0, qword src1)
+{
+   return (qword) _powf4((vec_float4) src0, (vec_float4) src1);
+}
+
+static qword
+micro_rnd(qword src)
+{
+   const qword half = (qword) spu_splats(0.5f);
+
+   /* May be able to use _roundf4.  There may be some difference, though.
+    */
+   return (qword) _floorf4((vec_float4) si_fa(src, half));
+}
+
+static INLINE qword
+micro_ishr(qword src0, qword src1)
+{
+   return si_rotma(src0, si_sfi(src1, 0));
+}
+
+static qword
+micro_trunc(qword src)
+{
+   return (qword) _truncf4((vec_float4) src);
+}
+
+static qword
+micro_sin(qword src)
+{
+   return (qword) _sinf4((vec_float4) src);
+}
+
+static INLINE qword
+micro_sqrt(qword src)
+{
+   return (qword) _sqrtf4((vec_float4) src);
+}
+
+static void
+fetch_src_file_channel(
+   const struct spu_exec_machine *mach,
+   const uint file,
+   const uint swizzle,
+   const union spu_exec_channel *index,
+   union spu_exec_channel *chan )
+{
+   switch( swizzle ) {
+   case TGSI_EXTSWIZZLE_X:
+   case TGSI_EXTSWIZZLE_Y:
+   case TGSI_EXTSWIZZLE_Z:
+   case TGSI_EXTSWIZZLE_W:
+      switch( file ) {
+      case TGSI_FILE_CONSTANT: {
+         unsigned i;
+
+         for (i = 0; i < 4; i++) {
+            const float *ptr = mach->Consts[index->i[i]];
+            float tmp[4];
+
+            spu_dcache_fetch_unaligned((qword *) tmp,
+                                       (uintptr_t)(ptr + swizzle),
+                                       sizeof(float));
+
+            chan->f[i] = tmp[0];
+         }
+         break;
+      }
+
+      case TGSI_FILE_INPUT:
+         chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0];
+         chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1];
+         chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2];
+         chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3];
+         break;
+
+      case TGSI_FILE_TEMPORARY:
+         chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0];
+         chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1];
+         chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2];
+         chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3];
+         break;
+
+      case TGSI_FILE_IMMEDIATE:
+         ASSERT( index->i[0] < (int) mach->ImmLimit );
+         ASSERT( index->i[1] < (int) mach->ImmLimit );
+         ASSERT( index->i[2] < (int) mach->ImmLimit );
+         ASSERT( index->i[3] < (int) mach->ImmLimit );
+
+         chan->f[0] = mach->Imms[index->i[0]][swizzle];
+         chan->f[1] = mach->Imms[index->i[1]][swizzle];
+         chan->f[2] = mach->Imms[index->i[2]][swizzle];
+         chan->f[3] = mach->Imms[index->i[3]][swizzle];
+         break;
+
+      case TGSI_FILE_ADDRESS:
+         chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0];
+         chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1];
+         chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2];
+         chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3];
+         break;
+
+      case TGSI_FILE_OUTPUT:
+         /* vertex/fragment output vars can be read too */
+         chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0];
+         chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1];
+         chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2];
+         chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3];
+         break;
+
+      default:
+         ASSERT( 0 );
+      }
+      break;
+
+   case TGSI_EXTSWIZZLE_ZERO:
+      *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C];
+      break;
+
+   case TGSI_EXTSWIZZLE_ONE:
+      *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C];
+      break;
+
+   default:
+      ASSERT( 0 );
+   }
+}
+
+static void
+fetch_source(
+   const struct spu_exec_machine *mach,
+   union spu_exec_channel *chan,
+   const struct tgsi_full_src_register *reg,
+   const uint chan_index )
+{
+   union spu_exec_channel index;
+   uint swizzle;
+
+   index.i[0] =
+   index.i[1] =
+   index.i[2] =
+   index.i[3] = reg->SrcRegister.Index;
+
+   if (reg->SrcRegister.Indirect) {
+      union spu_exec_channel index2;
+      union spu_exec_channel indir_index;
+
+      index2.i[0] =
+      index2.i[1] =
+      index2.i[2] =
+      index2.i[3] = reg->SrcRegisterInd.Index;
+
+      swizzle = tgsi_util_get_src_register_swizzle(&reg->SrcRegisterInd,
+                                                   CHAN_X);
+      fetch_src_file_channel(
+         mach,
+         reg->SrcRegisterInd.File,
+         swizzle,
+         &index2,
+         &indir_index );
+
+      index.q = si_a(index.q, indir_index.q);
+   }
+
+   if( reg->SrcRegister.Dimension ) {
+      switch( reg->SrcRegister.File ) {
+      case TGSI_FILE_INPUT:
+         index.q = si_mpyi(index.q, 17);
+         break;
+      case TGSI_FILE_CONSTANT:
+         index.q = si_shli(index.q, 12);
+         break;
+      default:
+         ASSERT( 0 );
+      }
+
+      index.i[0] += reg->SrcRegisterDim.Index;
+      index.i[1] += reg->SrcRegisterDim.Index;
+      index.i[2] += reg->SrcRegisterDim.Index;
+      index.i[3] += reg->SrcRegisterDim.Index;
+
+      if (reg->SrcRegisterDim.Indirect) {
+         union spu_exec_channel index2;
+         union spu_exec_channel indir_index;
+
+         index2.i[0] =
+         index2.i[1] =
+         index2.i[2] =
+         index2.i[3] = reg->SrcRegisterDimInd.Index;
+
+         swizzle = tgsi_util_get_src_register_swizzle( &reg->SrcRegisterDimInd, CHAN_X );
+         fetch_src_file_channel(
+            mach,
+            reg->SrcRegisterDimInd.File,
+            swizzle,
+            &index2,
+            &indir_index );
+
+         index.q = si_a(index.q, indir_index.q);
+      }
+   }
+
+   swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index );
+   fetch_src_file_channel(
+      mach,
+      reg->SrcRegister.File,
+      swizzle,
+      &index,
+      chan );
+
+   switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) {
+   case TGSI_UTIL_SIGN_CLEAR:
+      chan->q = micro_abs(chan->q);
+      break;
+
+   case TGSI_UTIL_SIGN_SET:
+      chan->q = micro_set_sign(chan->q);
+      break;
+
+   case TGSI_UTIL_SIGN_TOGGLE:
+      chan->q = micro_neg(chan->q);
+      break;
+
+   case TGSI_UTIL_SIGN_KEEP:
+      break;
+   }
+
+   if (reg->SrcRegisterExtMod.Complement) {
+      chan->q = si_fs(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, chan->q);
+   }
+}
+
+static void
+store_dest(
+   struct spu_exec_machine *mach,
+   const union spu_exec_channel *chan,
+   const struct tgsi_full_dst_register *reg,
+   const struct tgsi_full_instruction *inst,
+   uint chan_index )
+{
+   union spu_exec_channel *dst;
+
+   switch( reg->DstRegister.File ) {
+   case TGSI_FILE_NULL:
+      return;
+
+   case TGSI_FILE_OUTPUT:
+      dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0]
+                           + reg->DstRegister.Index].xyzw[chan_index];
+      break;
+
+   case TGSI_FILE_TEMPORARY:
+      dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index];
+      break;
+
+   case TGSI_FILE_ADDRESS:
+      dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index];
+      break;
+
+   default:
+      ASSERT( 0 );
+      return;
+   }
+
+   switch (inst->Instruction.Saturate)
+   {
+   case TGSI_SAT_NONE:
+      if (mach->ExecMask & 0x1)
+         dst->i[0] = chan->i[0];
+      if (mach->ExecMask & 0x2)
+         dst->i[1] = chan->i[1];
+      if (mach->ExecMask & 0x4)
+         dst->i[2] = chan->i[2];
+      if (mach->ExecMask & 0x8)
+         dst->i[3] = chan->i[3];
+      break;
+
+   case TGSI_SAT_ZERO_ONE:
+      /* XXX need to obey ExecMask here */
+      dst->q = micro_max(chan->q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q);
+      dst->q = micro_min(dst->q, mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q);
+      break;
+
+   case TGSI_SAT_MINUS_PLUS_ONE:
+      ASSERT( 0 );
+      break;
+
+   default:
+      ASSERT( 0 );
+   }
+}
+
+#define FETCH(VAL,INDEX,CHAN)\
+    fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN)
+
+#define STORE(VAL,INDEX,CHAN)\
+    store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN )
+
+
+/**
+ * Execute ARB-style KIL which is predicated by a src register.
+ * Kill fragment if any of the four values is less than zero.
+ */
+static void
+exec_kil(struct spu_exec_machine *mach,
+         const struct tgsi_full_instruction *inst)
+{
+   uint uniquemask;
+   uint chan_index;
+   uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */
+   union spu_exec_channel r[1];
+
+   /* This mask stores component bits that were already tested. Note that
+    * we test if the value is less than zero, so 1.0 and 0.0 need not to be
+    * tested. */
+   uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE);
+
+   for (chan_index = 0; chan_index < 4; chan_index++)
+   {
+      uint swizzle;
+      uint i;
+
+      /* unswizzle channel */
+      swizzle = tgsi_util_get_full_src_register_extswizzle (
+                        &inst->FullSrcRegisters[0],
+                        chan_index);
+
+      /* check if the component has not been already tested */
+      if (uniquemask & (1 << swizzle))
+         continue;
+      uniquemask |= 1 << swizzle;
+
+      FETCH(&r[0], 0, chan_index);
+      for (i = 0; i < 4; i++)
+         if (r[0].f[i] < 0.0f)
+            kilmask |= 1 << i;
+   }
+
+   mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask;
+}
+
+/**
+ * Execute NVIDIA-style KIL which is predicated by a condition code.
+ * Kill fragment if the condition code is TRUE.
+ */
+static void
+exec_kilp(struct tgsi_exec_machine *mach,
+          const struct tgsi_full_instruction *inst)
+{
+   uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */
+
+   /* TODO: build kilmask from CC mask */
+
+   mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask;
+}
+
+/*
+ * Fetch a texel using STR texture coordinates.
+ */
+static void
+fetch_texel( struct spu_sampler *sampler,
+             const union spu_exec_channel *s,
+             const union spu_exec_channel *t,
+             const union spu_exec_channel *p,
+             float lodbias,  /* XXX should be float[4] */
+             union spu_exec_channel *r,
+             union spu_exec_channel *g,
+             union spu_exec_channel *b,
+             union spu_exec_channel *a )
+{
+   qword rgba[4];
+   qword out[4];
+
+   sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, 
+			(float (*)[4]) rgba);
+
+   _transpose_matrix4x4((vec_float4 *) out, (vec_float4 *) rgba);
+   r->q = out[0];
+   g->q = out[1];
+   b->q = out[2];
+   a->q = out[3];
+}
+
+
+static void
+exec_tex(struct spu_exec_machine *mach,
+         const struct tgsi_full_instruction *inst,
+         boolean biasLod, boolean projected)
+{
+   const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index;
+   union spu_exec_channel r[8];
+   uint chan_index;
+   float lodBias;
+
+   /*   printf("Sampler %u unit %u\n", sampler, unit); */
+
+   switch (inst->InstructionExtTexture.Texture) {
+   case TGSI_TEXTURE_1D:
+
+      FETCH(&r[0], 0, CHAN_X);
+
+      if (projected) {
+         FETCH(&r[1], 0, CHAN_W);
+         r[0].q = micro_div(r[0].q, r[1].q);
+      }
+
+      if (biasLod) {
+         FETCH(&r[1], 0, CHAN_W);
+         lodBias = r[2].f[0];
+      }
+      else
+         lodBias = 0.0;
+
+      fetch_texel(&mach->Samplers[unit],
+                  &r[0], NULL, NULL, lodBias,  /* S, T, P, BIAS */
+                  &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */
+      break;
+
+   case TGSI_TEXTURE_2D:
+   case TGSI_TEXTURE_RECT:
+
+      FETCH(&r[0], 0, CHAN_X);
+      FETCH(&r[1], 0, CHAN_Y);
+      FETCH(&r[2], 0, CHAN_Z);
+
+      if (projected) {
+         FETCH(&r[3], 0, CHAN_W);
+         r[0].q = micro_div(r[0].q, r[3].q);
+         r[1].q = micro_div(r[1].q, r[3].q);
+         r[2].q = micro_div(r[2].q, r[3].q);
+      }
+
+      if (biasLod) {
+         FETCH(&r[3], 0, CHAN_W);
+         lodBias = r[3].f[0];
+      }
+      else
+         lodBias = 0.0;
+
+      fetch_texel(&mach->Samplers[unit],
+                  &r[0], &r[1], &r[2], lodBias,  /* inputs */
+                  &r[0], &r[1], &r[2], &r[3]);  /* outputs */
+      break;
+
+   case TGSI_TEXTURE_3D:
+   case TGSI_TEXTURE_CUBE:
+
+      FETCH(&r[0], 0, CHAN_X);
+      FETCH(&r[1], 0, CHAN_Y);
+      FETCH(&r[2], 0, CHAN_Z);
+
+      if (projected) {
+         FETCH(&r[3], 0, CHAN_W);
+         r[0].q = micro_div(r[0].q, r[3].q);
+         r[1].q = micro_div(r[1].q, r[3].q);
+         r[2].q = micro_div(r[2].q, r[3].q);
+      }
+
+      if (biasLod) {
+         FETCH(&r[3], 0, CHAN_W);
+         lodBias = r[3].f[0];
+      }
+      else
+         lodBias = 0.0;
+
+      fetch_texel(&mach->Samplers[unit],
+                  &r[0], &r[1], &r[2], lodBias,
+                  &r[0], &r[1], &r[2], &r[3]);
+      break;
+
+   default:
+      ASSERT (0);
+   }
+
+   FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+      STORE( &r[chan_index], 0, chan_index );
+   }
+}
+
+
+
+static void
+constant_interpolation(
+   struct spu_exec_machine *mach,
+   unsigned attrib,
+   unsigned chan )
+{
+   unsigned i;
+
+   for( i = 0; i < QUAD_SIZE; i++ ) {
+      mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan];
+   }
+}
+
+static void
+linear_interpolation(
+   struct spu_exec_machine *mach,
+   unsigned attrib,
+   unsigned chan )
+{
+   const float x = mach->QuadPos.xyzw[0].f[0];
+   const float y = mach->QuadPos.xyzw[1].f[0];
+   const float dadx = mach->InterpCoefs[attrib].dadx[chan];
+   const float dady = mach->InterpCoefs[attrib].dady[chan];
+   const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y;
+   mach->Inputs[attrib].xyzw[chan].f[0] = a0;
+   mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx;
+   mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady;
+   mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady;
+}
+
+static void
+perspective_interpolation(
+   struct spu_exec_machine *mach,
+   unsigned attrib,
+   unsigned chan )
+{
+   const float x = mach->QuadPos.xyzw[0].f[0];
+   const float y = mach->QuadPos.xyzw[1].f[0];
+   const float dadx = mach->InterpCoefs[attrib].dadx[chan];
+   const float dady = mach->InterpCoefs[attrib].dady[chan];
+   const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y;
+   const float *w = mach->QuadPos.xyzw[3].f;
+   /* divide by W here */
+   mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0];
+   mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1];
+   mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2];
+   mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3];
+}
+
+
+typedef void (* interpolation_func)(
+   struct spu_exec_machine *mach,
+   unsigned attrib,
+   unsigned chan );
+
+static void
+exec_declaration(struct spu_exec_machine *mach,
+                 const struct tgsi_full_declaration *decl)
+{
+   if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) {
+      if( decl->Declaration.File == TGSI_FILE_INPUT ) {
+         unsigned first, last, mask;
+         interpolation_func interp;
+
+         first = decl->DeclarationRange.First;
+         last = decl->DeclarationRange.Last;
+         mask = decl->Declaration.UsageMask;
+
+         switch( decl->Declaration.Interpolate ) {
+         case TGSI_INTERPOLATE_CONSTANT:
+            interp = constant_interpolation;
+            break;
+
+         case TGSI_INTERPOLATE_LINEAR:
+            interp = linear_interpolation;
+            break;
+
+         case TGSI_INTERPOLATE_PERSPECTIVE:
+            interp = perspective_interpolation;
+            break;
+
+         default:
+            ASSERT( 0 );
+         }
+
+         if( mask == TGSI_WRITEMASK_XYZW ) {
+            unsigned i, j;
+
+            for( i = first; i <= last; i++ ) {
+               for( j = 0; j < NUM_CHANNELS; j++ ) {
+                  interp( mach, i, j );
+               }
+            }
+         }
+         else {
+            unsigned i, j;
+
+            for( j = 0; j < NUM_CHANNELS; j++ ) {
+               if( mask & (1 << j) ) {
+                  for( i = first; i <= last; i++ ) {
+                     interp( mach, i, j );
+                  }
+               }
+            }
+         }
+      }
+   }
+}
+
+static void
+exec_instruction(
+   struct spu_exec_machine *mach,
+   const struct tgsi_full_instruction *inst,
+   int *pc )
+{
+   uint chan_index;
+   union spu_exec_channel r[8];
+
+   (*pc)++;
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_ARL:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 FETCH( &r[0], 0, chan_index );
+         r[0].q = si_cflts(r[0].q, 0);
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_MOV:
+   case TGSI_OPCODE_SWZ:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_LIT:
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+	 STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X );
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+	 FETCH( &r[0], 0, CHAN_X );
+         if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+            r[0].q = micro_max(r[0].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q);
+	    STORE( &r[0], 0, CHAN_Y );
+	 }
+
+         if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+            FETCH( &r[1], 0, CHAN_Y );
+            r[1].q = micro_max(r[1].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q);
+
+            FETCH( &r[2], 0, CHAN_W );
+            r[2].q = micro_min(r[2].q, mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].q);
+            r[2].q = micro_max(r[2].q, mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].q);
+            r[1].q = micro_pow(r[1].q, r[2].q);
+
+            /* r0 = (r0 > 0.0) ? r1 : 0.0
+             */
+            r[0].q = si_fcgt(r[0].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q);
+            r[0].q = si_selb(mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q, r[1].q,
+                             r[0].q);
+            STORE( &r[0], 0, CHAN_Z );
+         }
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
+	 STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_RCP:
+   /* TGSI_OPCODE_RECIP */
+      FETCH( &r[0], 0, CHAN_X );
+      r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q);
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_RSQ:
+   /* TGSI_OPCODE_RECIPSQRT */
+      FETCH( &r[0], 0, CHAN_X );
+      r[0].q = micro_sqrt(r[0].q);
+      r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q);
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_EXP:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_LOG:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_MUL:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index )
+      {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+
+         r[0].q = si_fm(r[0].q, r[1].q);
+
+         STORE(&r[0], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_ADD:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         r[0].q = si_fa(r[0].q, r[1].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DP3:
+   /* TGSI_OPCODE_DOT3 */
+      FETCH( &r[0], 0, CHAN_X );
+      FETCH( &r[1], 1, CHAN_X );
+      r[0].q = si_fm(r[0].q, r[1].q);
+
+      FETCH( &r[1], 0, CHAN_Y );
+      FETCH( &r[2], 1, CHAN_Y );
+      r[0].q = si_fma(r[1].q, r[2].q, r[0].q);
+
+
+      FETCH( &r[1], 0, CHAN_Z );
+      FETCH( &r[2], 1, CHAN_Z );
+      r[0].q = si_fma(r[1].q, r[2].q, r[0].q);
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+    case TGSI_OPCODE_DP4:
+    /* TGSI_OPCODE_DOT4 */
+       FETCH(&r[0], 0, CHAN_X);
+       FETCH(&r[1], 1, CHAN_X);
+
+      r[0].q = si_fm(r[0].q, r[1].q);
+
+       FETCH(&r[1], 0, CHAN_Y);
+       FETCH(&r[2], 1, CHAN_Y);
+
+      r[0].q = si_fma(r[1].q, r[2].q, r[0].q);
+
+       FETCH(&r[1], 0, CHAN_Z);
+       FETCH(&r[2], 1, CHAN_Z);
+
+      r[0].q = si_fma(r[1].q, r[2].q, r[0].q);
+
+       FETCH(&r[1], 0, CHAN_W);
+       FETCH(&r[2], 1, CHAN_W);
+
+      r[0].q = si_fma(r[1].q, r[2].q, r[0].q);
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DST:
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+	 STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X );
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+	 FETCH( &r[0], 0, CHAN_Y );
+	 FETCH( &r[1], 1, CHAN_Y);
+      r[0].q = si_fm(r[0].q, r[1].q);
+	 STORE( &r[0], 0, CHAN_Y );
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+	 FETCH( &r[0], 0, CHAN_Z );
+	 STORE( &r[0], 0, CHAN_Z );
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
+	 FETCH( &r[0], 1, CHAN_W );
+	 STORE( &r[0], 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_MIN:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+
+         r[0].q = micro_min(r[0].q, r[1].q);
+
+         STORE(&r[0], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_MAX:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+
+         r[0].q = micro_max(r[0].q, r[1].q);
+
+         STORE(&r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SLT:
+   /* TGSI_OPCODE_SETLT */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+
+         r[0].q = micro_ge(r[0].q, r[1].q);
+         r[0].q = si_xori(r[0].q, 0xff);
+
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SGE:
+   /* TGSI_OPCODE_SETGE */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         r[0].q = micro_ge(r[0].q, r[1].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_MAD:
+   /* TGSI_OPCODE_MADD */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         FETCH( &r[2], 2, chan_index );
+         r[0].q = si_fma(r[0].q, r[1].q, r[2].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SUB:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+
+         r[0].q = si_fs(r[0].q, r[1].q);
+
+         STORE(&r[0], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_LERP:
+   /* TGSI_OPCODE_LRP */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+         FETCH(&r[2], 2, chan_index);
+
+         r[1].q = si_fs(r[1].q, r[2].q);
+         r[0].q = si_fma(r[0].q, r[1].q, r[2].q);
+
+         STORE(&r[0], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_CND:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_CND0:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_DOT2ADD:
+      /* TGSI_OPCODE_DP2A */
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_INDEX:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_NEGATE:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_FRAC:
+   /* TGSI_OPCODE_FRC */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         r[0].q = micro_frc(r[0].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_CLAMP:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_FLOOR:
+   /* TGSI_OPCODE_FLR */
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         r[0].q = micro_flr(r[0].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_ROUND:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         r[0].q = micro_rnd(r[0].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_EXPBASE2:
+    /* TGSI_OPCODE_EX2 */
+      FETCH(&r[0], 0, CHAN_X);
+
+      r[0].q = micro_pow(mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].q, r[0].q);
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_LOGBASE2:
+   /* TGSI_OPCODE_LG2 */
+      FETCH( &r[0], 0, CHAN_X );
+      r[0].q = micro_lg2(r[0].q);
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_POWER:
+      /* TGSI_OPCODE_POW */
+      FETCH(&r[0], 0, CHAN_X);
+      FETCH(&r[1], 1, CHAN_X);
+
+      r[0].q = micro_pow(r[0].q, r[1].q);
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_CROSSPRODUCT:
+      /* TGSI_OPCODE_XPD */
+      FETCH(&r[0], 0, CHAN_Y);
+      FETCH(&r[1], 1, CHAN_Z);
+      FETCH(&r[3], 0, CHAN_Z);
+      FETCH(&r[4], 1, CHAN_Y);
+
+      /* r2 = (r0 * r1) - (r3 * r5)
+       */
+      r[2].q = si_fm(r[3].q, r[5].q);
+      r[2].q = si_fms(r[0].q, r[1].q, r[2].q);
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) {
+         STORE( &r[2], 0, CHAN_X );
+      }
+
+      FETCH(&r[2], 1, CHAN_X);
+      FETCH(&r[5], 0, CHAN_X);
+
+      /* r3 = (r3 * r2) - (r1 * r5)
+       */
+      r[1].q = si_fm(r[1].q, r[5].q);
+      r[3].q = si_fms(r[3].q, r[2].q, r[1].q);
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) {
+         STORE( &r[3], 0, CHAN_Y );
+      }
+
+      /* r5 = (r5 * r4) - (r0 * r2)
+       */
+      r[0].q = si_fm(r[0].q, r[2].q);
+      r[5].q = si_fms(r[5].q, r[4].q, r[0].q);
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) {
+         STORE( &r[5], 0, CHAN_Z );
+      }
+
+      if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) {
+         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
+      }
+      break;
+
+    case TGSI_OPCODE_MULTIPLYMATRIX:
+       ASSERT (0);
+       break;
+
+    case TGSI_OPCODE_ABS:
+       FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+          FETCH(&r[0], 0, chan_index);
+
+          r[0].q = micro_abs(r[0].q);
+
+          STORE(&r[0], 0, chan_index);
+       }
+       break;
+
+   case TGSI_OPCODE_RCC:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_DPH:
+      FETCH(&r[0], 0, CHAN_X);
+      FETCH(&r[1], 1, CHAN_X);
+
+      r[0].q = si_fm(r[0].q, r[1].q);
+
+      FETCH(&r[1], 0, CHAN_Y);
+      FETCH(&r[2], 1, CHAN_Y);
+
+      r[0].q = si_fma(r[1].q, r[2].q, r[0].q);
+
+      FETCH(&r[1], 0, CHAN_Z);
+      FETCH(&r[2], 1, CHAN_Z);
+
+      r[0].q = si_fma(r[1].q, r[2].q, r[0].q);
+
+      FETCH(&r[1], 1, CHAN_W);
+
+      r[0].q = si_fa(r[0].q, r[1].q);
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_COS:
+      FETCH(&r[0], 0, CHAN_X);
+
+      r[0].q = micro_cos(r[0].q);
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+	 STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DDX:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         r[0].q = micro_ddx(r[0].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_DDY:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         r[0].q = micro_ddy(r[0].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_KILP:
+      exec_kilp (mach, inst);
+      break;
+
+   case TGSI_OPCODE_KIL:
+      exec_kil (mach, inst);
+      break;
+
+   case TGSI_OPCODE_PK2H:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_PK2US:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_PK4B:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_PK4UB:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_RFL:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_SEQ:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+
+         r[0].q = si_fceq(r[0].q, r[1].q);
+
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SFL:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_SGT:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         r[0].q = si_fcgt(r[0].q, r[1].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SIN:
+      FETCH( &r[0], 0, CHAN_X );
+      r[0].q = micro_sin(r[0].q);
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SLE:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+
+         r[0].q = si_fcgt(r[0].q, r[1].q);
+         r[0].q = si_xori(r[0].q, 0xff);
+
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SNE:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+
+         r[0].q = si_fceq(r[0].q, r[1].q);
+         r[0].q = si_xori(r[0].q, 0xff);
+
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_STR:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_TEX:
+      /* simple texture lookup */
+      /* src[0] = texcoord */
+      /* src[1] = sampler unit */
+      exec_tex(mach, inst, FALSE, FALSE);
+      break;
+
+   case TGSI_OPCODE_TXB:
+      /* Texture lookup with lod bias */
+      /* src[0] = texcoord (src[0].w = load bias) */
+      /* src[1] = sampler unit */
+      exec_tex(mach, inst, TRUE, FALSE);
+      break;
+
+   case TGSI_OPCODE_TXD:
+      /* Texture lookup with explict partial derivatives */
+      /* src[0] = texcoord */
+      /* src[1] = d[strq]/dx */
+      /* src[2] = d[strq]/dy */
+      /* src[3] = sampler unit */
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_TXL:
+      /* Texture lookup with explit LOD */
+      /* src[0] = texcoord (src[0].w = load bias) */
+      /* src[1] = sampler unit */
+      exec_tex(mach, inst, TRUE, FALSE);
+      break;
+
+   case TGSI_OPCODE_TXP:
+      /* Texture lookup with projection */
+      /* src[0] = texcoord (src[0].w = projection) */
+      /* src[1] = sampler unit */
+      exec_tex(mach, inst, TRUE, TRUE);
+      break;
+
+   case TGSI_OPCODE_UP2H:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_UP2US:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_UP4B:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_UP4UB:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_X2D:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_ARA:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_ARR:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_BRA:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_CAL:
+      /* skip the call if no execution channels are enabled */
+      if (mach->ExecMask) {
+         /* do the call */
+
+         /* push the Cond, Loop, Cont stacks */
+         ASSERT(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING);
+         mach->CondStack[mach->CondStackTop++] = mach->CondMask;
+         ASSERT(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+         mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
+         ASSERT(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+         mach->ContStack[mach->ContStackTop++] = mach->ContMask;
+
+         ASSERT(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING);
+         mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask;
+
+         /* note that PC was already incremented above */
+         mach->CallStack[mach->CallStackTop++] = *pc;
+         *pc = inst->InstructionExtLabel.Label;
+      }
+      break;
+
+   case TGSI_OPCODE_RET:
+      mach->FuncMask &= ~mach->ExecMask;
+      UPDATE_EXEC_MASK(mach);
+
+      if (mach->ExecMask == 0x0) {
+         /* really return now (otherwise, keep executing */
+
+         if (mach->CallStackTop == 0) {
+            /* returning from main() */
+            *pc = -1;
+            return;
+         }
+         *pc = mach->CallStack[--mach->CallStackTop];
+
+         /* pop the Cond, Loop, Cont stacks */
+         ASSERT(mach->CondStackTop > 0);
+         mach->CondMask = mach->CondStack[--mach->CondStackTop];
+         ASSERT(mach->LoopStackTop > 0);
+         mach->LoopMask = mach->LoopStack[--mach->LoopStackTop];
+         ASSERT(mach->ContStackTop > 0);
+         mach->ContMask = mach->ContStack[--mach->ContStackTop];
+         ASSERT(mach->FuncStackTop > 0);
+         mach->FuncMask = mach->FuncStack[--mach->FuncStackTop];
+
+         UPDATE_EXEC_MASK(mach);
+      }
+      break;
+
+   case TGSI_OPCODE_SSG:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_CMP:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH(&r[0], 0, chan_index);
+         FETCH(&r[1], 1, chan_index);
+         FETCH(&r[2], 2, chan_index);
+
+         /* r0 = (r0 < 0.0) ? r1 : r2
+          */
+         r[3].q = si_xor(r[3].q, r[3].q);
+         r[0].q = micro_lt(r[0].q, r[3].q);
+         r[0].q = si_selb(r[1].q, r[2].q, r[0].q);
+
+         STORE(&r[0], 0, chan_index);
+      }
+      break;
+
+   case TGSI_OPCODE_SCS:
+      if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) {
+         FETCH( &r[0], 0, CHAN_X );
+      }
+      if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) {
+         r[1].q = micro_cos(r[0].q);
+         STORE( &r[1], 0, CHAN_X );
+      }
+      if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) {
+         r[1].q = micro_sin(r[0].q);
+         STORE( &r[1], 0, CHAN_Y );
+      }
+      if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) {
+         STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z );
+      }
+      if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) {
+         STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W );
+      }
+      break;
+
+   case TGSI_OPCODE_NRM:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_DIV:
+      ASSERT( 0 );
+      break;
+
+   case TGSI_OPCODE_DP2:
+      FETCH( &r[0], 0, CHAN_X );
+      FETCH( &r[1], 1, CHAN_X );
+      r[0].q = si_fm(r[0].q, r[1].q);
+
+      FETCH( &r[1], 0, CHAN_Y );
+      FETCH( &r[2], 1, CHAN_Y );
+      r[0].q = si_fma(r[1].q, r[2].q, r[0].q);
+
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_IF:
+      /* push CondMask */
+      ASSERT(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING);
+      mach->CondStack[mach->CondStackTop++] = mach->CondMask;
+      FETCH( &r[0], 0, CHAN_X );
+      /* update CondMask */
+      if( ! r[0].u[0] ) {
+         mach->CondMask &= ~0x1;
+      }
+      if( ! r[0].u[1] ) {
+         mach->CondMask &= ~0x2;
+      }
+      if( ! r[0].u[2] ) {
+         mach->CondMask &= ~0x4;
+      }
+      if( ! r[0].u[3] ) {
+         mach->CondMask &= ~0x8;
+      }
+      UPDATE_EXEC_MASK(mach);
+      /* Todo: If CondMask==0, jump to ELSE */
+      break;
+
+   case TGSI_OPCODE_ELSE:
+      /* invert CondMask wrt previous mask */
+      {
+         uint prevMask;
+         ASSERT(mach->CondStackTop > 0);
+         prevMask = mach->CondStack[mach->CondStackTop - 1];
+         mach->CondMask = ~mach->CondMask & prevMask;
+         UPDATE_EXEC_MASK(mach);
+         /* Todo: If CondMask==0, jump to ENDIF */
+      }
+      break;
+
+   case TGSI_OPCODE_ENDIF:
+      /* pop CondMask */
+      ASSERT(mach->CondStackTop > 0);
+      mach->CondMask = mach->CondStack[--mach->CondStackTop];
+      UPDATE_EXEC_MASK(mach);
+      break;
+
+   case TGSI_OPCODE_END:
+      /* halt execution */
+      *pc = -1;
+      break;
+
+   case TGSI_OPCODE_REP:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_ENDREP:
+       ASSERT (0);
+       break;
+
+   case TGSI_OPCODE_PUSHA:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_POPA:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_CEIL:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         r[0].q = micro_ceil(r[0].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_I2F:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         r[0].q = si_csflt(r[0].q, 0);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_NOT:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         r[0].q = si_xorbi(r[0].q, 0xff);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_TRUNC:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         r[0].q = micro_trunc(r[0].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SHL:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+
+         r[0].q = si_shl(r[0].q, r[1].q);
+
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SHR:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         r[0].q = micro_ishr(r[0].q, r[1].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_AND:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         r[0].q = si_and(r[0].q, r[1].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_OR:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         r[0].q = si_or(r[0].q, r[1].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_MOD:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_XOR:
+      FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) {
+         FETCH( &r[0], 0, chan_index );
+         FETCH( &r[1], 1, chan_index );
+         r[0].q = si_xor(r[0].q, r[1].q);
+         STORE( &r[0], 0, chan_index );
+      }
+      break;
+
+   case TGSI_OPCODE_SAD:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_TXF:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_TXQ:
+      ASSERT (0);
+      break;
+
+   case TGSI_OPCODE_EMIT:
+      mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16;
+      mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++;
+      break;
+
+   case TGSI_OPCODE_ENDPRIM:
+      mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++;
+      mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0;
+      break;
+
+   case TGSI_OPCODE_LOOP:
+      /* fall-through (for now) */
+   case TGSI_OPCODE_BGNLOOP2:
+      /* push LoopMask and ContMasks */
+      ASSERT(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+      mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask;
+      ASSERT(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING);
+      mach->ContStack[mach->ContStackTop++] = mach->ContMask;
+      break;
+
+   case TGSI_OPCODE_ENDLOOP:
+      /* fall-through (for now at least) */
+   case TGSI_OPCODE_ENDLOOP2:
+      /* Restore ContMask, but don't pop */
+      ASSERT(mach->ContStackTop > 0);
+      mach->ContMask = mach->ContStack[mach->ContStackTop - 1];
+      if (mach->LoopMask) {
+         /* repeat loop: jump to instruction just past BGNLOOP */
+         *pc = inst->InstructionExtLabel.Label + 1;
+      }
+      else {
+         /* exit loop: pop LoopMask */
+         ASSERT(mach->LoopStackTop > 0);
+         mach->LoopMask = mach->LoopStack[--mach->LoopStackTop];
+         /* pop ContMask */
+         ASSERT(mach->ContStackTop > 0);
+         mach->ContMask = mach->ContStack[--mach->ContStackTop];
+      }
+      UPDATE_EXEC_MASK(mach);
+      break;
+
+   case TGSI_OPCODE_BRK:
+      /* turn off loop channels for each enabled exec channel */
+      mach->LoopMask &= ~mach->ExecMask;
+      /* Todo: if mach->LoopMask == 0, jump to end of loop */
+      UPDATE_EXEC_MASK(mach);
+      break;
+
+   case TGSI_OPCODE_CONT:
+      /* turn off cont channels for each enabled exec channel */
+      mach->ContMask &= ~mach->ExecMask;
+      /* Todo: if mach->LoopMask == 0, jump to end of loop */
+      UPDATE_EXEC_MASK(mach);
+      break;
+
+   case TGSI_OPCODE_BGNSUB:
+      /* no-op */
+      break;
+
+   case TGSI_OPCODE_ENDSUB:
+      /* no-op */
+      break;
+
+   case TGSI_OPCODE_NOISE1:
+      ASSERT( 0 );
+      break;
+
+   case TGSI_OPCODE_NOISE2:
+      ASSERT( 0 );
+      break;
+
+   case TGSI_OPCODE_NOISE3:
+      ASSERT( 0 );
+      break;
+
+   case TGSI_OPCODE_NOISE4:
+      ASSERT( 0 );
+      break;
+
+   case TGSI_OPCODE_NOP:
+      break;
+
+   default:
+      ASSERT( 0 );
+   }
+}
+
+
+/**
+ * Run TGSI interpreter.
+ * \return bitmask of "alive" quad components
+ */
+uint
+spu_exec_machine_run( struct spu_exec_machine *mach )
+{
+   uint i;
+   int pc = 0;
+
+   mach->CondMask = 0xf;
+   mach->LoopMask = 0xf;
+   mach->ContMask = 0xf;
+   mach->FuncMask = 0xf;
+   mach->ExecMask = 0xf;
+
+   mach->CondStackTop = 0; /* temporarily subvert this ASSERTion */
+   ASSERT(mach->CondStackTop == 0);
+   ASSERT(mach->LoopStackTop == 0);
+   ASSERT(mach->ContStackTop == 0);
+   ASSERT(mach->CallStackTop == 0);
+
+   mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0;
+   mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0;
+
+   if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) {
+      mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0;
+      mach->Primitives[0] = 0;
+   }
+
+
+   /* execute declarations (interpolants) */
+   if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) {
+      for (i = 0; i < mach->NumDeclarations; i++) {
+         union {
+            struct tgsi_full_declaration decl;
+            qword buffer[ROUNDUP16(sizeof(struct tgsi_full_declaration)) / 16];
+         } d ALIGN16_ATTRIB;
+         unsigned ea = (unsigned) (mach->Declarations + pc);
+
+         spu_dcache_fetch_unaligned(d.buffer, ea, sizeof(d.decl));
+
+         exec_declaration( mach, &d.decl );
+      }
+   }
+
+   /* execute instructions, until pc is set to -1 */
+   while (pc != -1) {
+      union {
+         struct tgsi_full_instruction inst;
+         qword buffer[ROUNDUP16(sizeof(struct tgsi_full_instruction)) / 16];
+      } i ALIGN16_ATTRIB;
+      unsigned ea = (unsigned) (mach->Instructions + pc);
+
+      spu_dcache_fetch_unaligned(i.buffer, ea, sizeof(i.inst));
+      exec_instruction( mach, & i.inst, &pc );
+   }
+
+#if 0
+   /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */
+   if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) {
+      /*
+       * Scale back depth component.
+       */
+      for (i = 0; i < 4; i++)
+         mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF;
+   }
+#endif
+
+   return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0];
+}
+
+
diff --git a/src/gallium/drivers/cell/spu/spu_exec.h b/src/gallium/drivers/cell/spu/spu_exec.h
new file mode 100644
index 0000000000..8605679940
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_exec.h
@@ -0,0 +1,172 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#if !defined SPU_EXEC_H
+#define SPU_EXEC_H
+
+#include "pipe/p_compiler.h"
+#include "tgsi/tgsi_exec.h"
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+/**
+  * Registers may be treated as float, signed int or unsigned int.
+  */
+union spu_exec_channel
+{
+   float    f[QUAD_SIZE];
+   int      i[QUAD_SIZE];
+   unsigned u[QUAD_SIZE];
+   qword    q;
+};
+
+/**
+  * A vector[RGBA] of channels[4 pixels]
+  */
+struct spu_exec_vector
+{
+   union spu_exec_channel xyzw[NUM_CHANNELS];
+};
+
+/**
+ * For fragment programs, information for computing fragment input
+ * values from plane equation of the triangle/line.
+ */
+struct spu_interp_coef
+{
+   float a0[NUM_CHANNELS];	/* in an xyzw layout */
+   float dadx[NUM_CHANNELS];
+   float dady[NUM_CHANNELS];
+};
+
+
+struct softpipe_tile_cache;  /**< Opaque to TGSI */
+
+/**
+ * Information for sampling textures, which must be implemented
+ * by code outside the TGSI executor.
+ */
+struct spu_sampler
+{
+   const struct pipe_sampler_state *state;
+   struct pipe_texture *texture;
+   /** Get samples for four fragments in a quad */
+   void (*get_samples)(struct spu_sampler *sampler,
+                       const float s[QUAD_SIZE],
+                       const float t[QUAD_SIZE],
+                       const float p[QUAD_SIZE],
+                       float lodbias,
+                       float rgba[NUM_CHANNELS][QUAD_SIZE]);
+   void *pipe; /*XXX temporary*/
+   struct softpipe_tile_cache *cache;
+};
+
+
+/**
+ * Run-time virtual machine state for executing TGSI shader.
+ */
+struct spu_exec_machine
+{
+   /*
+    * 32 program temporaries
+    * 4  internal temporaries
+    * 1  address
+    */
+   struct spu_exec_vector       Temps[TGSI_EXEC_NUM_TEMPS 
+                                      + TGSI_EXEC_NUM_TEMP_EXTRAS + 1]
+       ALIGN16_ATTRIB;
+
+   struct spu_exec_vector       *Addrs;
+
+   struct spu_sampler           *Samplers;
+
+   float                         Imms[TGSI_EXEC_NUM_IMMEDIATES][4];
+   unsigned                      ImmLimit;
+   float                         (*Consts)[4];
+   struct spu_exec_vector       *Inputs;
+   struct spu_exec_vector       *Outputs;
+   unsigned                      Processor;
+
+   /* GEOMETRY processor only. */
+   unsigned                      *Primitives;
+
+   /* FRAGMENT processor only. */
+   const struct spu_interp_coef *InterpCoefs;
+   struct spu_exec_vector       QuadPos;
+
+   /* Conditional execution masks */
+   uint CondMask;  /**< For IF/ELSE/ENDIF */
+   uint LoopMask;  /**< For BGNLOOP/ENDLOOP */
+   uint ContMask;  /**< For loop CONT statements */
+   uint FuncMask;  /**< For function calls */
+   uint ExecMask;  /**< = CondMask & LoopMask */
+
+   /** Condition mask stack (for nested conditionals) */
+   uint CondStack[TGSI_EXEC_MAX_COND_NESTING];
+   int CondStackTop;
+
+   /** Loop mask stack (for nested loops) */
+   uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING];
+   int LoopStackTop;
+
+   /** Loop continue mask stack (see comments in tgsi_exec.c) */
+   uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING];
+   int ContStackTop;
+
+   /** Function execution mask stack (for executing subroutine code) */
+   uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING];
+   int FuncStackTop;
+
+   /** Function call stack for saving/restoring the program counter */
+   uint CallStack[TGSI_EXEC_MAX_CALL_NESTING];
+   int CallStackTop;
+
+   struct tgsi_full_instruction *Instructions;
+   uint NumInstructions;
+
+   struct tgsi_full_declaration *Declarations;
+   uint NumDeclarations;
+};
+
+
+extern void
+spu_exec_machine_init(struct spu_exec_machine *mach,
+                      uint numSamplers,
+                      struct spu_sampler *samplers,
+                      unsigned processor);
+
+extern uint
+spu_exec_machine_run( struct spu_exec_machine *mach );
+
+
+#if defined __cplusplus
+} /* extern "C" */
+#endif
+
+#endif /* SPU_EXEC_H */
diff --git a/src/gallium/drivers/cell/spu/spu_funcs.c b/src/gallium/drivers/cell/spu/spu_funcs.c
new file mode 100644
index 0000000000..ff3d609d25
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_funcs.c
@@ -0,0 +1,173 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * SPU functions accessed by shaders.
+ *
+ * Authors: Brian Paul
+ */
+
+
+#include <string.h>
+#include <libmisc.h>
+#include <math.h>
+#include <cos14_v.h>
+#include <sin14_v.h>
+#include <simdmath/exp2f4.h>
+#include <simdmath/log2f4.h>
+#include <simdmath/powf4.h>
+
+#include "cell/common.h"
+#include "spu_main.h"
+#include "spu_funcs.h"
+#include "spu_texture.h"
+
+
+/** For "return"-ing four vectors */
+struct vec_4x4
+{
+   vector float v[4];
+};
+
+
+static vector float
+spu_cos(vector float x)
+{
+   return _cos14_v(x);
+}
+
+static vector float
+spu_sin(vector float x)
+{
+   return _sin14_v(x);
+}
+
+static vector float
+spu_pow(vector float x, vector float y)
+{
+   return _powf4(x, y);
+}
+
+static vector float
+spu_exp2(vector float x)
+{
+   return _exp2f4(x);
+}
+
+static vector float
+spu_log2(vector float x)
+{
+   return _log2f4(x);
+}
+
+
+static struct vec_4x4
+spu_tex_2d(vector float s, vector float t, vector float r, vector float q,
+           unsigned unit)
+{
+   struct vec_4x4 colors;
+   (void) r;
+   (void) q;
+   spu.sample_texture_2d[unit](s, t, unit, 0, 0, colors.v);
+   return colors;
+}
+
+static struct vec_4x4
+spu_tex_3d(vector float s, vector float t, vector float r, vector float q,
+           unsigned unit)
+{
+   struct vec_4x4 colors;
+   (void) r;
+   (void) q;
+   spu.sample_texture_2d[unit](s, t, unit, 0, 0, colors.v);
+   return colors;
+}
+
+static struct vec_4x4
+spu_tex_cube(vector float s, vector float t, vector float r, vector float q,
+           unsigned unit)
+{
+   struct vec_4x4 colors;
+   (void) q;
+   sample_texture_cube(s, t, r, unit, colors.v);
+   return colors;
+}
+
+
+/**
+ * Add named function to list of "exported" functions that will be
+ * made available to the PPU-hosted code generator.
+ */
+static void
+export_func(struct cell_spu_function_info *spu_functions,
+            const char *name, void *addr)
+{
+   uint n = spu_functions->num;
+   ASSERT(strlen(name) < 16);
+   strcpy(spu_functions->names[n], name);
+   spu_functions->addrs[n] = (uint) addr;
+   spu_functions->num++;
+   ASSERT(spu_functions->num <= 16);
+}
+
+
+/**
+ * Return info about the SPU's function to the PPU / main memory.
+ * The PPU needs to know the address of some SPU-side functions so
+ * that we can generate shader code with function calls.
+ */
+void
+return_function_info(void)
+{
+   struct cell_spu_function_info funcs ALIGN16_ATTRIB;
+   int tag = TAG_MISC;
+
+   ASSERT(sizeof(funcs) == 256); /* must be multiple of 16 bytes */
+
+   funcs.num = 0;
+   export_func(&funcs, "spu_cos", &spu_cos);
+   export_func(&funcs, "spu_sin", &spu_sin);
+   export_func(&funcs, "spu_pow", &spu_pow);
+   export_func(&funcs, "spu_exp2", &spu_exp2);
+   export_func(&funcs, "spu_log2", &spu_log2);
+   export_func(&funcs, "spu_tex_2d", &spu_tex_2d);
+   export_func(&funcs, "spu_tex_3d", &spu_tex_3d);
+   export_func(&funcs, "spu_tex_cube", &spu_tex_cube);
+
+   /* Send the function info back to the PPU / main memory */
+   mfc_put((void *) &funcs,  /* src in local store */
+           (unsigned int) spu.init.spu_functions, /* dst in main memory */
+           sizeof(funcs),  /* bytes */
+           tag,
+           0, /* tid */
+           0  /* rid */);
+   wait_on_mask(1 << tag);
+}
+
+
+
diff --git a/src/gallium/drivers/cell/spu/spu_funcs.h b/src/gallium/drivers/cell/spu/spu_funcs.h
new file mode 100644
index 0000000000..3adb6ae99f
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_funcs.h
@@ -0,0 +1,35 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef SPU_FUNCS_H
+#define SPU_FUNCS_H
+
+extern void
+return_function_info(void);
+
+#endif
+
diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c
new file mode 100644
index 0000000000..97c86d194d
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_main.c
@@ -0,0 +1,117 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/* main() for Cell SPU code */
+
+
+#include <stdio.h>
+#include <libmisc.h>
+
+#include "pipe/p_defines.h"
+
+#include "spu_funcs.h"
+#include "spu_command.h"
+#include "spu_main.h"
+#include "spu_per_fragment_op.h"
+#include "spu_texture.h"
+//#include "spu_test.h"
+#include "cell/common.h"
+
+
+/*
+helpful headers:
+/usr/lib/gcc/spu/4.1.1/include/spu_mfcio.h
+/opt/cell/sdk/usr/include/libmisc.h
+*/
+
+struct spu_global spu;
+
+
+static void
+one_time_init(void)
+{
+   memset(spu.ctile_status, TILE_STATUS_DEFINED, sizeof(spu.ctile_status));
+   memset(spu.ztile_status, TILE_STATUS_DEFINED, sizeof(spu.ztile_status));
+   invalidate_tex_cache();
+}
+
+/* In some versions of the SDK the SPE main takes 'unsigned long' as a
+ * parameter.  In others it takes 'unsigned long long'.  Use a define to
+ * select between the two.
+ */
+#ifdef SPU_MAIN_PARAM_LONG_LONG
+typedef unsigned long long main_param_t;
+#else
+typedef unsigned long main_param_t;
+#endif
+
+/**
+ * SPE entrypoint.
+ */
+int
+main(main_param_t speid, main_param_t argp)
+{
+   int tag = 0;
+
+   (void) speid;
+
+   ASSERT(sizeof(tile_t) == TILE_SIZE * TILE_SIZE * 4);
+   ASSERT(sizeof(struct cell_command_render) % 8 == 0);
+   ASSERT(sizeof(struct cell_command_fragment_ops) % 8 == 0);
+   ASSERT(((unsigned long) &spu.fragment_program_code) % 8 == 0);
+
+   one_time_init();
+   spu_command_init();
+
+   D_PRINTF(CELL_DEBUG_CMD, "main() speid=%lu\n", (unsigned long) speid);
+   D_PRINTF(CELL_DEBUG_FRAGMENT_OP_FALLBACK, "using fragment op fallback\n");
+
+   /* get initialization data */
+   mfc_get(&spu.init,  /* dest */
+           (unsigned int) argp, /* src */
+           sizeof(struct cell_init_info), /* bytes */
+           tag,
+           0, /* tid */
+           0  /* rid */);
+   wait_on_mask( 1 << tag );
+
+   if (spu.init.id == 0) {
+      return_function_info();
+   }
+
+#if 0
+   if (spu.init.id==0)
+      spu_test_misc(spu.init.id);
+#endif
+
+   command_loop();
+
+   spu_command_close();
+
+   return 0;
+}
diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h
new file mode 100644
index 0000000000..33767e7c51
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_main.h
@@ -0,0 +1,254 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef SPU_MAIN_H
+#define SPU_MAIN_H
+
+
+#include <spu_mfcio.h>
+
+#include "cell/common.h"
+#include "draw/draw_vertex.h"
+#include "pipe/p_state.h"
+
+
+#if DEBUG
+/* These debug macros use the unusual construction ", ##__VA_ARGS__"
+ * which expands to the expected comma + args if variadic arguments
+ * are supplied, but swallows the comma if there are no variadic
+ * arguments (which avoids syntax errors that would otherwise occur).
+ */
+#define D_PRINTF(flag, format,...) \
+   if (spu.init.debug_flags & (flag)) \
+      printf("SPU %u: " format, spu.init.id, ##__VA_ARGS__)
+#else
+#define D_PRINTF(...)
+#endif
+
+
+/**
+ * A tile is basically a TILE_SIZE x TILE_SIZE block of 4-byte pixels.
+ * The data may be addressed through several different types.
+ */
+typedef union {
+   ushort us[TILE_SIZE][TILE_SIZE];
+   uint   ui[TILE_SIZE][TILE_SIZE];
+   vector unsigned short us8[TILE_SIZE/2][TILE_SIZE/4];
+   vector unsigned int ui4[TILE_SIZE/2][TILE_SIZE/2];
+} tile_t;
+
+
+#define TILE_STATUS_CLEAR   1
+#define TILE_STATUS_DEFINED 2  /**< defined in FB, but not in local store */
+#define TILE_STATUS_CLEAN   3  /**< in local store, but not changed */
+#define TILE_STATUS_DIRTY   4  /**< modified locally, but not put back yet */
+#define TILE_STATUS_GETTING 5  /**< mfc_get() called but not yet arrived */
+
+
+/** Function for sampling textures */
+typedef void (*spu_sample_texture_2d_func)(vector float s,
+                                           vector float t,
+                                           uint unit, uint level, uint face,
+                                           vector float colors[4]);
+
+
+/** Function for performing per-fragment ops */
+typedef void (*spu_fragment_ops_func)(uint x, uint y,
+                                      tile_t *colorTile,
+                                      tile_t *depthStencilTile,
+                                      vector float fragZ,
+                                      vector float fragRed,
+                                      vector float fragGreen,
+                                      vector float fragBlue,
+                                      vector float fragAlpha,
+                                      vector unsigned int mask);
+
+/** Function for running fragment program */
+typedef vector unsigned int (*spu_fragment_program_func)(vector float *inputs,
+                                                         vector float *outputs,
+                                                         vector float *constants);
+
+
+struct spu_framebuffer
+{
+   void *color_start;              /**< addr of color surface in main memory */
+   void *depth_start;              /**< addr of depth surface in main memory */
+   enum pipe_format color_format;
+   enum pipe_format depth_format;
+   uint width, height;             /**< size in pixels */
+   uint width_tiles, height_tiles; /**< width and height in tiles */
+
+   uint color_clear_value;
+   uint depth_clear_value;
+
+   uint zsize;                     /**< 0, 2 or 4 bytes per Z */
+   float zscale;                   /**< 65535.0, 2^24-1 or 2^32-1 */
+} ALIGN16_ATTRIB;
+
+
+/** per-texture level info */
+struct spu_texture_level
+{
+   void *start;
+   ushort width, height, depth;
+   ushort tiles_per_row;
+   uint bytes_per_image;
+   /** texcoord scale factors */
+   vector float scale_s, scale_t, scale_r;
+   /** texcoord masks (if REPEAT then size-1, else ~0) */
+   vector signed int mask_s, mask_t, mask_r;
+   /** texcoord clamp limits */
+   vector signed int max_s, max_t, max_r;
+} ALIGN16_ATTRIB;
+
+
+struct spu_texture
+{
+   struct spu_texture_level level[CELL_MAX_TEXTURE_LEVELS];
+   uint max_level;
+   uint target;  /**< PIPE_TEXTURE_x */
+} ALIGN16_ATTRIB;
+
+
+/**
+ * All SPU global/context state will be in a singleton object of this type:
+ */
+struct spu_global
+{
+   /** One-time init/constant info */
+   struct cell_init_info init;
+
+   /*
+    * Current state
+    */
+   struct spu_framebuffer fb;
+   struct pipe_depth_stencil_alpha_state depth_stencil_alpha;
+   struct pipe_blend_state blend;
+   struct pipe_blend_color blend_color;
+   struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS];
+   struct pipe_rasterizer_state rasterizer;
+   struct spu_texture texture[PIPE_MAX_SAMPLERS];
+   struct vertex_info vertex_info;
+
+   /** Current color and Z tiles */
+   tile_t ctile ALIGN16_ATTRIB;
+   tile_t ztile ALIGN16_ATTRIB;
+
+   /** Read depth/stencil tiles? */
+   boolean read_depth_stencil;
+
+   /** Current tiles' status */
+   ubyte cur_ctile_status, cur_ztile_status;
+
+   /** Status of all tiles in framebuffer */
+   ubyte ctile_status[CELL_MAX_HEIGHT/TILE_SIZE][CELL_MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB;
+   ubyte ztile_status[CELL_MAX_HEIGHT/TILE_SIZE][CELL_MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB;
+
+   /** Current fragment ops machine code, at 8-byte boundary */
+   uint *fragment_ops_code;
+   uint fragment_ops_code_size;
+   /** Current fragment ops functions, 0 = frontfacing, 1 = backfacing */
+   spu_fragment_ops_func fragment_ops[2];
+
+   /** Current fragment program machine code, at 8-byte boundary */
+   uint fragment_program_code[SPU_MAX_FRAGMENT_PROGRAM_INSTS] ALIGN8_ATTRIB;
+   /** Current fragment ops function */
+   spu_fragment_program_func fragment_program;
+
+   /** Current texture sampler function */
+   spu_sample_texture_2d_func sample_texture_2d[CELL_MAX_SAMPLERS];
+   spu_sample_texture_2d_func min_sample_texture_2d[CELL_MAX_SAMPLERS];
+   spu_sample_texture_2d_func mag_sample_texture_2d[CELL_MAX_SAMPLERS];
+
+   /** Fragment program constants */
+   vector float constants[4 * CELL_MAX_CONSTANTS];
+
+} ALIGN16_ATTRIB;
+
+
+extern struct spu_global spu;
+
+
+
+/* DMA TAGS */
+
+#define TAG_SURFACE_CLEAR     10
+#define TAG_VERTEX_BUFFER     11
+#define TAG_READ_TILE_COLOR   12
+#define TAG_READ_TILE_Z       13
+#define TAG_WRITE_TILE_COLOR  14
+#define TAG_WRITE_TILE_Z      15
+#define TAG_INDEX_BUFFER      16
+#define TAG_BATCH_BUFFER      17
+#define TAG_MISC              18
+#define TAG_DCACHE0           20
+#define TAG_DCACHE1           21
+#define TAG_DCACHE2           22
+#define TAG_DCACHE3           23
+#define TAG_FENCE             24
+
+
+static INLINE void
+wait_on_mask(unsigned tagMask)
+{
+   mfc_write_tag_mask( tagMask );
+   /* wait for completion of _any_ DMAs specified by tagMask */
+   mfc_read_tag_status_any();
+}
+
+
+static INLINE void
+wait_on_mask_all(unsigned tagMask)
+{
+   mfc_write_tag_mask( tagMask );
+   /* wait for completion of _any_ DMAs specified by tagMask */
+   mfc_read_tag_status_all();
+}
+
+
+
+
+
+static INLINE void
+memset16(ushort *d, ushort value, uint count)
+{
+   uint i;
+   for (i = 0; i < count; i++)
+      d[i] = value;
+}
+
+
+static INLINE void
+memset32(uint *d, uint value, uint count)
+{
+   uint i;
+   for (i = 0; i < count; i++)
+      d[i] = value;
+}
+
+
+#endif /* SPU_MAIN_H */
diff --git a/src/gallium/drivers/cell/spu/spu_per_fragment_op.c b/src/gallium/drivers/cell/spu/spu_per_fragment_op.c
new file mode 100644
index 0000000000..eba9f95cf1
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_per_fragment_op.c
@@ -0,0 +1,631 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \author Brian Paul
+ */
+
+
+#include <transpose_matrix4x4.h>
+#include "pipe/p_format.h"
+#include "spu_main.h"
+#include "spu_colorpack.h"
+#include "spu_per_fragment_op.h"
+
+
+#define LINEAR_QUAD_LAYOUT 1
+
+
+static INLINE vector float
+spu_min(vector float a, vector float b)
+{
+   vector unsigned int m;
+   m = spu_cmpgt(a, b);    /* m = a > b ? ~0 : 0 */
+   return spu_sel(a, b, m);
+}
+
+
+static INLINE vector float
+spu_max(vector float a, vector float b)
+{
+   vector unsigned int m;
+   m = spu_cmpgt(a, b);    /* m = a > b ? ~0 : 0 */
+   return spu_sel(b, a, m);
+}
+
+
+/**
+ * Called by rasterizer for each quad after the shader has run.  Do
+ * all the per-fragment operations including alpha test, z test,
+ * stencil test, blend, colormask and logicops.  This is a
+ * fallback/debug function.  In reality we'll use a generated function
+ * produced by the PPU.  But this function is useful for
+ * debug/validation.
+ */
+void
+spu_fallback_fragment_ops(uint x, uint y,
+                          tile_t *colorTile,
+                          tile_t *depthStencilTile,
+                          vector float fragZ,
+                          vector float fragR,
+                          vector float fragG,
+                          vector float fragB,
+                          vector float fragA,
+                          vector unsigned int mask)
+{
+   vector float frag_aos[4];
+   unsigned int fbc0, fbc1, fbc2, fbc3 ; /* framebuffer/tile colors */
+   unsigned int fragc0, fragc1, fragc2, fragc3;  /* fragment colors */
+
+   /*
+    * Do alpha test
+    */
+   if (spu.depth_stencil_alpha.alpha.enabled) {
+      vector float ref = spu_splats(spu.depth_stencil_alpha.alpha.ref_value);
+      vector unsigned int amask;
+
+      switch (spu.depth_stencil_alpha.alpha.func) {
+      case PIPE_FUNC_LESS:
+         amask = spu_cmpgt(ref, fragA);  /* mask = (fragA < ref) */
+         break;
+      case PIPE_FUNC_GREATER:
+         amask = spu_cmpgt(fragA, ref);  /* mask = (fragA > ref) */
+         break;
+      case PIPE_FUNC_GEQUAL:
+         amask = spu_cmpgt(ref, fragA);
+         amask = spu_nor(amask, amask);
+         break;
+      case PIPE_FUNC_LEQUAL:
+         amask = spu_cmpgt(fragA, ref);
+         amask = spu_nor(amask, amask);
+         break;
+      case PIPE_FUNC_EQUAL:
+         amask = spu_cmpeq(ref, fragA);
+         break;
+      case PIPE_FUNC_NOTEQUAL:
+         amask = spu_cmpeq(ref, fragA);
+         amask = spu_nor(amask, amask);
+         break;
+      case PIPE_FUNC_ALWAYS:
+         amask = spu_splats(0xffffffffU);
+         break;
+      case PIPE_FUNC_NEVER:
+         amask = spu_splats( 0x0U);
+         break;
+      default:
+         ;
+      }
+
+      mask = spu_and(mask, amask);
+   }
+
+
+   /*
+    * Z and/or stencil testing...
+    */
+   if (spu.depth_stencil_alpha.depth.enabled ||
+       spu.depth_stencil_alpha.stencil[0].enabled) {
+
+      /* get four Z/Stencil values from tile */
+      vector unsigned int mask24 = spu_splats((unsigned int)0x00ffffffU);
+      vector unsigned int ifbZS = depthStencilTile->ui4[y/2][x/2];
+      vector unsigned int ifbZ = spu_and(ifbZS, mask24);
+      vector unsigned int ifbS = spu_andc(ifbZS, mask24);
+
+      if (spu.depth_stencil_alpha.stencil[0].enabled) {
+         /* do stencil test */
+         ASSERT(spu.fb.depth_format == PIPE_FORMAT_S8Z24_UNORM);
+
+      }
+      else if (spu.depth_stencil_alpha.depth.enabled) {
+         /* do depth test */
+
+         ASSERT(spu.fb.depth_format == PIPE_FORMAT_S8Z24_UNORM ||
+                spu.fb.depth_format == PIPE_FORMAT_X8Z24_UNORM);
+
+         vector unsigned int ifragZ;
+         vector unsigned int zmask;
+
+         /* convert four fragZ from float to uint */
+         fragZ = spu_mul(fragZ, spu_splats((float) 0xffffff));
+         ifragZ = spu_convtu(fragZ, 0);
+
+         /* do depth comparison, setting zmask with results */
+         switch (spu.depth_stencil_alpha.depth.func) {
+         case PIPE_FUNC_LESS:
+            zmask = spu_cmpgt(ifbZ, ifragZ);  /* mask = (ifragZ < ifbZ) */
+            break;
+         case PIPE_FUNC_GREATER:
+            zmask = spu_cmpgt(ifragZ, ifbZ);  /* mask = (ifbZ > ifragZ) */
+            break;
+         case PIPE_FUNC_GEQUAL:
+            zmask = spu_cmpgt(ifbZ, ifragZ);
+            zmask = spu_nor(zmask, zmask);
+            break;
+         case PIPE_FUNC_LEQUAL:
+            zmask = spu_cmpgt(ifragZ, ifbZ);
+            zmask = spu_nor(zmask, zmask);
+            break;
+         case PIPE_FUNC_EQUAL:
+            zmask = spu_cmpeq(ifbZ, ifragZ);
+            break;
+         case PIPE_FUNC_NOTEQUAL:
+            zmask = spu_cmpeq(ifbZ, ifragZ);
+            zmask = spu_nor(zmask, zmask);
+            break;
+         case PIPE_FUNC_ALWAYS:
+            zmask = spu_splats(0xffffffffU);
+            break;
+         case PIPE_FUNC_NEVER:
+            zmask = spu_splats( 0x0U);
+            break;
+         default:
+            ;
+         }
+
+         mask = spu_and(mask, zmask);
+
+         /* merge framebuffer Z and fragment Z according to the mask */
+         ifbZ = spu_or(spu_and(ifragZ, mask),
+                       spu_andc(ifbZ, mask));
+      }
+
+      if (spu_extract(spu_orx(mask), 0)) {
+         /* put new fragment Z/Stencil values back into Z/Stencil tile */
+         depthStencilTile->ui4[y/2][x/2] = spu_or(ifbZ, ifbS);
+
+         spu.cur_ztile_status = TILE_STATUS_DIRTY;
+      }
+   }
+
+
+   /*
+    * If we'll need the current framebuffer/tile colors for blending
+    * or logicop or colormask, fetch them now.
+    */
+   if (spu.blend.blend_enable ||
+       spu.blend.logicop_enable ||
+       spu.blend.colormask != 0xf) {
+
+#if LINEAR_QUAD_LAYOUT /* See comments/diagram below */
+      fbc0 = colorTile->ui[y][x*2+0];
+      fbc1 = colorTile->ui[y][x*2+1];
+      fbc2 = colorTile->ui[y][x*2+2];
+      fbc3 = colorTile->ui[y][x*2+3];
+#else
+      fbc0 = colorTile->ui[y+0][x+0];
+      fbc1 = colorTile->ui[y+0][x+1];
+      fbc2 = colorTile->ui[y+1][x+0];
+      fbc3 = colorTile->ui[y+1][x+1];
+#endif
+   }
+
+
+   /*
+    * Do blending
+    */
+   if (spu.blend.blend_enable) {
+      /* blending terms, misc regs */
+      vector float term1r, term1g, term1b, term1a;
+      vector float term2r, term2g, term2b, term2a;
+      vector float one, tmp;
+
+      vector float fbRGBA[4];  /* current framebuffer colors */
+
+      /* convert framebuffer colors from packed int to vector float */
+      {
+         vector float temp[4]; /* float colors in AOS form */
+         switch (spu.fb.color_format) {
+         case PIPE_FORMAT_B8G8R8A8_UNORM:
+            temp[0] = spu_unpack_B8G8R8A8(fbc0);
+            temp[1] = spu_unpack_B8G8R8A8(fbc1);
+            temp[2] = spu_unpack_B8G8R8A8(fbc2);
+            temp[3] = spu_unpack_B8G8R8A8(fbc3);
+            break;
+         case PIPE_FORMAT_A8R8G8B8_UNORM:
+            temp[0] = spu_unpack_A8R8G8B8(fbc0);
+            temp[1] = spu_unpack_A8R8G8B8(fbc1);
+            temp[2] = spu_unpack_A8R8G8B8(fbc2);
+            temp[3] = spu_unpack_A8R8G8B8(fbc3);
+            break;
+         default:
+            ASSERT(0);
+         }
+         _transpose_matrix4x4(fbRGBA, temp); /* fbRGBA = transpose(temp) */
+      }
+
+      /*
+       * Compute Src RGB terms (fragment color * factor)
+       */
+      switch (spu.blend.rgb_src_factor) {
+      case PIPE_BLENDFACTOR_ONE:
+         term1r = fragR;
+         term1g = fragG;
+         term1b = fragB;
+         break;
+      case PIPE_BLENDFACTOR_ZERO:
+         term1r =
+         term1g =
+         term1b = spu_splats(0.0f);
+         break;
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+         term1r = spu_mul(fragR, fragR);
+         term1g = spu_mul(fragG, fragG);
+         term1b = spu_mul(fragB, fragB);
+         break;
+      case PIPE_BLENDFACTOR_SRC_ALPHA:
+         term1r = spu_mul(fragR, fragA);
+         term1g = spu_mul(fragG, fragA);
+         term1b = spu_mul(fragB, fragA);
+         break;
+      case PIPE_BLENDFACTOR_DST_COLOR:
+         term1r = spu_mul(fragR, fbRGBA[0]);
+         term1g = spu_mul(fragG, fbRGBA[1]);
+         term1b = spu_mul(fragB, fbRGBA[1]);
+         break;
+      case PIPE_BLENDFACTOR_DST_ALPHA:
+         term1r = spu_mul(fragR, fbRGBA[3]);
+         term1g = spu_mul(fragG, fbRGBA[3]);
+         term1b = spu_mul(fragB, fbRGBA[3]);
+         break;
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+         term1r = spu_mul(fragR, spu_splats(spu.blend_color.color[0]));
+         term1g = spu_mul(fragG, spu_splats(spu.blend_color.color[1]));
+         term1b = spu_mul(fragB, spu_splats(spu.blend_color.color[2]));
+         break;
+      case PIPE_BLENDFACTOR_CONST_ALPHA:
+         term1r = spu_mul(fragR, spu_splats(spu.blend_color.color[3]));
+         term1g = spu_mul(fragG, spu_splats(spu.blend_color.color[3]));
+         term1b = spu_mul(fragB, spu_splats(spu.blend_color.color[3]));
+         break;
+      /* XXX more cases */
+      default:
+         ASSERT(0);
+      }
+
+      /*
+       * Compute Src Alpha term (fragment alpha * factor)
+       */
+      switch (spu.blend.alpha_src_factor) {
+      case PIPE_BLENDFACTOR_ONE:
+         term1a = fragA;
+         break;
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+         term1a = spu_splats(0.0f);
+         break;
+      case PIPE_BLENDFACTOR_SRC_ALPHA:
+         term1a = spu_mul(fragA, fragA);
+         break;
+      case PIPE_BLENDFACTOR_DST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_DST_ALPHA:
+         term1a = spu_mul(fragA, fbRGBA[3]);
+         break;
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_CONST_ALPHA:
+         term1a = spu_mul(fragR, spu_splats(spu.blend_color.color[3]));
+         break;
+      /* XXX more cases */
+      default:
+         ASSERT(0);
+      }
+
+      /*
+       * Compute Dest RGB terms (framebuffer color * factor)
+       */
+      switch (spu.blend.rgb_dst_factor) {
+      case PIPE_BLENDFACTOR_ONE:
+         term2r = fbRGBA[0];
+         term2g = fbRGBA[1];
+         term2b = fbRGBA[2];
+         break;
+      case PIPE_BLENDFACTOR_ZERO:
+         term2r =
+         term2g =
+         term2b = spu_splats(0.0f);
+         break;
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+         term2r = spu_mul(fbRGBA[0], fragR);
+         term2g = spu_mul(fbRGBA[1], fragG);
+         term2b = spu_mul(fbRGBA[2], fragB);
+         break;
+      case PIPE_BLENDFACTOR_SRC_ALPHA:
+         term2r = spu_mul(fbRGBA[0], fragA);
+         term2g = spu_mul(fbRGBA[1], fragA);
+         term2b = spu_mul(fbRGBA[2], fragA);
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+         one = spu_splats(1.0f);
+         tmp = spu_sub(one, fragA);
+         term2r = spu_mul(fbRGBA[0], tmp);
+         term2g = spu_mul(fbRGBA[1], tmp);
+         term2b = spu_mul(fbRGBA[2], tmp);
+         break;
+      case PIPE_BLENDFACTOR_DST_COLOR:
+         term2r = spu_mul(fbRGBA[0], fbRGBA[0]);
+         term2g = spu_mul(fbRGBA[1], fbRGBA[1]);
+         term2b = spu_mul(fbRGBA[2], fbRGBA[2]);
+         break;
+      case PIPE_BLENDFACTOR_DST_ALPHA:
+         term2r = spu_mul(fbRGBA[0], fbRGBA[3]);
+         term2g = spu_mul(fbRGBA[1], fbRGBA[3]);
+         term2b = spu_mul(fbRGBA[2], fbRGBA[3]);
+         break;
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+         term2r = spu_mul(fbRGBA[0], spu_splats(spu.blend_color.color[0]));
+         term2g = spu_mul(fbRGBA[1], spu_splats(spu.blend_color.color[1]));
+         term2b = spu_mul(fbRGBA[2], spu_splats(spu.blend_color.color[2]));
+         break;
+      case PIPE_BLENDFACTOR_CONST_ALPHA:
+         term2r = spu_mul(fbRGBA[0], spu_splats(spu.blend_color.color[3]));
+         term2g = spu_mul(fbRGBA[1], spu_splats(spu.blend_color.color[3]));
+         term2b = spu_mul(fbRGBA[2], spu_splats(spu.blend_color.color[3]));
+         break;
+       /* XXX more cases */
+      default:
+         ASSERT(0);
+      }
+
+      /*
+       * Compute Dest Alpha term (framebuffer alpha * factor)
+       */
+      switch (spu.blend.alpha_dst_factor) {
+      case PIPE_BLENDFACTOR_ONE:
+         term2a = fbRGBA[3];
+         break;
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+         term2a = spu_splats(0.0f);
+         break;
+      case PIPE_BLENDFACTOR_SRC_ALPHA:
+         term2a = spu_mul(fbRGBA[3], fragA);
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+         one = spu_splats(1.0f);
+         tmp = spu_sub(one, fragA);
+         term2a = spu_mul(fbRGBA[3], tmp);
+         break;
+      case PIPE_BLENDFACTOR_DST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_DST_ALPHA:
+         term2a = spu_mul(fbRGBA[3], fbRGBA[3]);
+         break;
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_CONST_ALPHA:
+         term2a = spu_mul(fbRGBA[3], spu_splats(spu.blend_color.color[3]));
+         break;
+      /* XXX more cases */
+      default:
+         ASSERT(0);
+      }
+
+      /*
+       * Combine Src/Dest RGB terms
+       */
+      switch (spu.blend.rgb_func) {
+      case PIPE_BLEND_ADD:
+         fragR = spu_add(term1r, term2r);
+         fragG = spu_add(term1g, term2g);
+         fragB = spu_add(term1b, term2b);
+         break;
+      case PIPE_BLEND_SUBTRACT:
+         fragR = spu_sub(term1r, term2r);
+         fragG = spu_sub(term1g, term2g);
+         fragB = spu_sub(term1b, term2b);
+         break;
+      case PIPE_BLEND_REVERSE_SUBTRACT:
+         fragR = spu_sub(term2r, term1r);
+         fragG = spu_sub(term2g, term1g);
+         fragB = spu_sub(term2b, term1b);
+         break;
+      case PIPE_BLEND_MIN:
+         fragR = spu_min(term1r, term2r);
+         fragG = spu_min(term1g, term2g);
+         fragB = spu_min(term1b, term2b);
+         break;
+      case PIPE_BLEND_MAX:
+         fragR = spu_max(term1r, term2r);
+         fragG = spu_max(term1g, term2g);
+         fragB = spu_max(term1b, term2b);
+         break;
+      default:
+         ASSERT(0);
+      }
+
+      /*
+       * Combine Src/Dest A term
+       */
+      switch (spu.blend.alpha_func) {
+      case PIPE_BLEND_ADD:
+         fragA = spu_add(term1a, term2a);
+         break;
+      case PIPE_BLEND_SUBTRACT:
+         fragA = spu_sub(term1a, term2a);
+         break;
+      case PIPE_BLEND_REVERSE_SUBTRACT:
+         fragA = spu_sub(term2a, term1a);
+         break;
+      case PIPE_BLEND_MIN:
+         fragA = spu_min(term1a, term2a);
+         break;
+      case PIPE_BLEND_MAX:
+         fragA = spu_max(term1a, term2a);
+         break;
+      default:
+         ASSERT(0);
+      }
+   }
+
+
+   /*
+    * Convert RRRR,GGGG,BBBB,AAAA to RGBA,RGBA,RGBA,RGBA.
+    */
+#if 0
+   /* original code */
+   {
+      vector float frag_soa[4];
+      frag_soa[0] = fragR;
+      frag_soa[1] = fragG;
+      frag_soa[2] = fragB;
+      frag_soa[3] = fragA;
+      _transpose_matrix4x4(frag_aos, frag_soa);
+   }
+#else
+   /* short-cut relying on function parameter layout: */
+   _transpose_matrix4x4(frag_aos, &fragR);
+   (void) fragG;
+   (void) fragB;
+#endif
+
+   /*
+    * Pack fragment float colors into 32-bit RGBA words.
+    */
+   switch (spu.fb.color_format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      fragc0 = spu_pack_A8R8G8B8(frag_aos[0]);
+      fragc1 = spu_pack_A8R8G8B8(frag_aos[1]);
+      fragc2 = spu_pack_A8R8G8B8(frag_aos[2]);
+      fragc3 = spu_pack_A8R8G8B8(frag_aos[3]);
+      break;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      fragc0 = spu_pack_B8G8R8A8(frag_aos[0]);
+      fragc1 = spu_pack_B8G8R8A8(frag_aos[1]);
+      fragc2 = spu_pack_B8G8R8A8(frag_aos[2]);
+      fragc3 = spu_pack_B8G8R8A8(frag_aos[3]);
+      break;
+   default:
+      fprintf(stderr, "SPU: Bad pixel format in spu_default_fragment_ops\n");
+      ASSERT(0);
+   }
+
+
+   /*
+    * Do color masking
+    */
+   if (spu.blend.colormask != 0xf) {
+      uint cmask = 0x0; /* each byte corresponds to a color channel */
+
+      /* Form bitmask depending on color buffer format and colormask bits */
+      switch (spu.fb.color_format) {
+      case PIPE_FORMAT_A8R8G8B8_UNORM:
+         if (spu.blend.colormask & PIPE_MASK_R)
+            cmask |= 0x00ff0000; /* red */
+         if (spu.blend.colormask & PIPE_MASK_G)
+            cmask |= 0x0000ff00; /* green */
+         if (spu.blend.colormask & PIPE_MASK_B)
+            cmask |= 0x000000ff; /* blue */
+         if (spu.blend.colormask & PIPE_MASK_A)
+            cmask |= 0xff000000; /* alpha */
+         break;
+      case PIPE_FORMAT_B8G8R8A8_UNORM:
+         if (spu.blend.colormask & PIPE_MASK_R)
+            cmask |= 0x0000ff00; /* red */
+         if (spu.blend.colormask & PIPE_MASK_G)
+            cmask |= 0x00ff0000; /* green */
+         if (spu.blend.colormask & PIPE_MASK_B)
+            cmask |= 0xff000000; /* blue */
+         if (spu.blend.colormask & PIPE_MASK_A)
+            cmask |= 0x000000ff; /* alpha */
+         break;
+      default:
+         ASSERT(0);
+      }
+
+      /*
+       * Apply color mask to the 32-bit packed colors.
+       * if (cmask[i])
+       *    frag color[i] = frag color[i];
+       * else
+       *    frag color[i] = framebuffer color[i];
+       */
+      fragc0 = (fragc0 & cmask) | (fbc0 & ~cmask);
+      fragc1 = (fragc1 & cmask) | (fbc1 & ~cmask);
+      fragc2 = (fragc2 & cmask) | (fbc2 & ~cmask);
+      fragc3 = (fragc3 & cmask) | (fbc3 & ~cmask);
+   }
+
+
+   /*
+    * Do logic ops
+    */
+   if (spu.blend.logicop_enable) {
+      /* XXX to do */
+      /* apply logicop to 32-bit packed colors (fragcx and fbcx) */
+   }
+
+
+   /*
+    * If mask is non-zero, mark tile as dirty.
+    */
+   if (spu_extract(spu_orx(mask), 0)) {
+      spu.cur_ctile_status = TILE_STATUS_DIRTY;
+   }
+   else {
+      /* write no fragments */
+      return;
+   }
+
+
+   /*
+    * Write new fragment/quad colors to the framebuffer/tile.
+    * Only write pixels where the corresponding mask word is set.
+    */
+#if LINEAR_QUAD_LAYOUT
+   /*
+    * Quad layout:
+    *  +--+--+--+--+
+    *  |p0|p1|p2|p3|...
+    *  +--+--+--+--+
+    */
+   if (spu_extract(mask, 0))
+      colorTile->ui[y][x*2] = fragc0;
+   if (spu_extract(mask, 1))
+      colorTile->ui[y][x*2+1] = fragc1;
+   if (spu_extract(mask, 2))
+      colorTile->ui[y][x*2+2] = fragc2;
+   if (spu_extract(mask, 3))
+      colorTile->ui[y][x*2+3] = fragc3;
+#else
+   /*
+    * Quad layout:
+    *  +--+--+
+    *  |p0|p1|...
+    *  +--+--+
+    *  |p2|p3|...
+    *  +--+--+
+    */
+   if (spu_extract(mask, 0))
+      colorTile->ui[y+0][x+0] = fragc0;
+   if (spu_extract(mask, 1))
+      colorTile->ui[y+0][x+1] = fragc1;
+   if (spu_extract(mask, 2))
+      colorTile->ui[y+1][x+0] = fragc2;
+   if (spu_extract(mask, 3))
+      colorTile->ui[y+1][x+1] = fragc3;
+#endif
+}
diff --git a/src/gallium/drivers/cell/spu/spu_per_fragment_op.h b/src/gallium/drivers/cell/spu/spu_per_fragment_op.h
new file mode 100644
index 0000000000..f817abf046
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_per_fragment_op.h
@@ -0,0 +1,44 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef SPU_PER_FRAGMENT_OP
+#define SPU_PER_FRAGMENT_OP
+
+
+extern void
+spu_fallback_fragment_ops(uint x, uint y,
+                          tile_t *colorTile,
+                          tile_t *depthStencilTile,
+                          vector float fragZ,
+                          vector float fragRed,
+                          vector float fragGreen,
+                          vector float fragBlue,
+                          vector float fragAlpha,
+                          vector unsigned int mask);
+
+
+#endif /* SPU_PER_FRAGMENT_OP */
diff --git a/src/gallium/drivers/cell/spu/spu_render.c b/src/gallium/drivers/cell/spu/spu_render.c
new file mode 100644
index 0000000000..7c225e2f27
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_render.c
@@ -0,0 +1,295 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include <stdio.h>
+#include <libmisc.h>
+#include <spu_mfcio.h>
+
+#include "spu_main.h"
+#include "spu_render.h"
+#include "spu_tri.h"
+#include "spu_tile.h"
+#include "cell/common.h"
+#include "util/u_memory.h"
+
+
+/**
+ * Given a rendering command's bounding box (in pixels) compute the
+ * location of the corresponding screen tile bounding box.
+ */
+static INLINE void
+tile_bounding_box(const struct cell_command_render *render,
+                  uint *txmin, uint *tymin,
+                  uint *box_num_tiles, uint *box_width_tiles)
+{
+#if 0
+   /* Debug: full-window bounding box */
+   uint txmax = spu.fb.width_tiles - 1;
+   uint tymax = spu.fb.height_tiles - 1;
+   *txmin = 0;
+   *tymin = 0;
+   *box_num_tiles = spu.fb.width_tiles * spu.fb.height_tiles;
+   *box_width_tiles = spu.fb.width_tiles;
+   (void) render;
+   (void) txmax;
+   (void) tymax;
+#else
+   uint txmax, tymax, box_height_tiles;
+
+   *txmin = (uint) render->xmin / TILE_SIZE;
+   *tymin = (uint) render->ymin / TILE_SIZE;
+   txmax = (uint) render->xmax / TILE_SIZE;
+   tymax = (uint) render->ymax / TILE_SIZE;
+   if (txmax >= spu.fb.width_tiles)
+      txmax = spu.fb.width_tiles-1;
+   if (tymax >= spu.fb.height_tiles)
+      tymax = spu.fb.height_tiles-1;
+   *box_width_tiles = txmax - *txmin + 1;
+   box_height_tiles = tymax - *tymin + 1;
+   *box_num_tiles = *box_width_tiles * box_height_tiles;
+#endif
+#if 0
+   printf("SPU %u: bounds: %g, %g  ...  %g, %g\n", spu.init.id,
+          render->xmin, render->ymin, render->xmax, render->ymax);
+   printf("SPU %u: tiles:  %u, %u .. %u, %u\n",
+           spu.init.id, *txmin, *tymin, txmax, tymax);
+   ASSERT(render->xmin <= render->xmax);
+   ASSERT(render->ymin <= render->ymax);
+#endif
+}
+
+
+/** Check if the tile at (tx,ty) belongs to this SPU */
+static INLINE boolean
+my_tile(uint tx, uint ty)
+{
+   return (spu.fb.width_tiles * ty + tx) % spu.init.num_spus == spu.init.id;
+}
+
+
+/**
+ * Start fetching non-clear color/Z tiles from main memory
+ */
+static INLINE void
+get_cz_tiles(uint tx, uint ty)
+{
+   if (spu.read_depth_stencil) {
+      if (spu.cur_ztile_status != TILE_STATUS_CLEAR) {
+         //printf("SPU %u: getting Z tile %u, %u\n", spu.init.id, tx, ty);
+         get_tile(tx, ty, &spu.ztile, TAG_READ_TILE_Z, 1);
+         spu.cur_ztile_status = TILE_STATUS_GETTING;
+      }
+   }
+
+   if (spu.cur_ctile_status != TILE_STATUS_CLEAR) {
+      //printf("SPU %u: getting C tile %u, %u\n", spu.init.id, tx, ty);
+      get_tile(tx, ty, &spu.ctile, TAG_READ_TILE_COLOR, 0);
+      spu.cur_ctile_status = TILE_STATUS_GETTING;
+   }
+}
+
+
+/**
+ * Start putting dirty color/Z tiles back to main memory
+ */
+static INLINE void
+put_cz_tiles(uint tx, uint ty)
+{
+   if (spu.cur_ztile_status == TILE_STATUS_DIRTY) {
+      /* tile was modified and needs to be written back */
+      //printf("SPU %u: put dirty Z tile %u, %u\n", spu.init.id, tx, ty);
+      put_tile(tx, ty, &spu.ztile, TAG_WRITE_TILE_Z, 1);
+      spu.cur_ztile_status = TILE_STATUS_DEFINED;
+   }
+   else if (spu.cur_ztile_status == TILE_STATUS_GETTING) {
+      /* tile was never used */
+      spu.cur_ztile_status = TILE_STATUS_DEFINED;
+      //printf("SPU %u: put getting Z tile %u, %u\n", spu.init.id, tx, ty);
+   }
+
+   if (spu.cur_ctile_status == TILE_STATUS_DIRTY) {
+      /* tile was modified and needs to be written back */
+      //printf("SPU %u: put dirty C tile %u, %u\n", spu.init.id, tx, ty);
+      put_tile(tx, ty, &spu.ctile, TAG_WRITE_TILE_COLOR, 0);
+      spu.cur_ctile_status = TILE_STATUS_DEFINED;
+   }
+   else if (spu.cur_ctile_status == TILE_STATUS_GETTING) {
+      /* tile was never used */
+      spu.cur_ctile_status = TILE_STATUS_DEFINED;
+      //printf("SPU %u: put getting C tile %u, %u\n", spu.init.id, tx, ty);
+   }
+}
+
+
+/**
+ * Wait for 'put' of color/z tiles to complete.
+ */
+static INLINE void
+wait_put_cz_tiles(void)
+{
+   wait_on_mask(1 << TAG_WRITE_TILE_COLOR);
+   if (spu.read_depth_stencil) {
+      wait_on_mask(1 << TAG_WRITE_TILE_Z);
+   }
+}
+
+
+/**
+ * Render primitives
+ * \param pos_incr  returns value indicating how may words to skip after
+ *                  this command in the batch buffer
+ */
+void
+cmd_render(const struct cell_command_render *render, uint *pos_incr)
+{
+   /* we'll DMA into these buffers */
+   ubyte vertex_data[CELL_BUFFER_SIZE] ALIGN16_ATTRIB;
+   const uint vertex_size = render->vertex_size; /* in bytes */
+   /*const*/ uint total_vertex_bytes = render->num_verts * vertex_size;
+   uint index_bytes;
+   const ubyte *vertices;
+   const ushort *indexes;
+   uint i, j;
+   uint num_tiles;
+
+   D_PRINTF(CELL_DEBUG_CMD,
+            "RENDER prim=%u num_vert=%u num_ind=%u inline_vert=%u\n",
+            render->prim_type,
+            render->num_verts,
+            render->num_indexes,
+            render->inline_verts);
+
+   ASSERT(sizeof(*render) % 4 == 0);
+   ASSERT(total_vertex_bytes % 16 == 0);
+   ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES);
+   ASSERT(render->num_indexes % 3 == 0);
+
+
+   /* indexes are right after the render command in the batch buffer */
+   indexes = (const ushort *) (render + 1);
+   index_bytes = ROUNDUP8(render->num_indexes * 2);
+   *pos_incr = index_bytes / 8 + sizeof(*render) / 8;
+
+
+   if (render->inline_verts) {
+      /* Vertices are after indexes in batch buffer at next 16-byte addr */
+      vertices = (const ubyte *) render + (*pos_incr * 8);
+      vertices = (const ubyte *) align_pointer((void *) vertices, 16);
+      ASSERT_ALIGN16(vertices);
+      *pos_incr = ((vertices + total_vertex_bytes) - (ubyte *) render) / 8;
+   }
+   else {
+      /* Begin DMA fetch of vertex buffer */
+      ubyte *src = spu.init.buffers[render->vertex_buf];
+      ubyte *dest = vertex_data;
+
+      /* skip vertex data we won't use */
+#if 01
+      src += render->min_index * vertex_size;
+      dest += render->min_index * vertex_size;
+      total_vertex_bytes -= render->min_index * vertex_size;
+#endif
+      ASSERT(total_vertex_bytes % 16 == 0);
+      ASSERT_ALIGN16(dest);
+      ASSERT_ALIGN16(src);
+
+      mfc_get(dest,   /* in vertex_data[] array */
+              (unsigned int) src,  /* src in main memory */
+              total_vertex_bytes,  /* size */
+              TAG_VERTEX_BUFFER,
+              0, /* tid */
+              0  /* rid */);
+
+      vertices = vertex_data;
+
+      wait_on_mask(1 << TAG_VERTEX_BUFFER);
+   }
+
+
+   /**
+    ** find tiles which intersect the prim bounding box
+    **/
+   uint txmin, tymin, box_width_tiles, box_num_tiles;
+   tile_bounding_box(render, &txmin, &tymin,
+                     &box_num_tiles, &box_width_tiles);
+
+
+   /* make sure any pending clears have completed */
+   wait_on_mask(1 << TAG_SURFACE_CLEAR); /* XXX temporary */
+
+
+   num_tiles = 0;
+
+   /**
+    ** loop over tiles, rendering tris
+    **/
+   for (i = 0; i < box_num_tiles; i++) {
+      const uint tx = txmin + i % box_width_tiles;
+      const uint ty = tymin + i / box_width_tiles;
+
+      ASSERT(tx < spu.fb.width_tiles);
+      ASSERT(ty < spu.fb.height_tiles);
+
+      if (!my_tile(tx, ty))
+         continue;
+
+      num_tiles++;
+
+      spu.cur_ctile_status = spu.ctile_status[ty][tx];
+      spu.cur_ztile_status = spu.ztile_status[ty][tx];
+
+      get_cz_tiles(tx, ty);
+
+      uint drawn = 0;
+
+      /* loop over tris */
+      for (j = 0; j < render->num_indexes; j += 3) {
+         const float *v0, *v1, *v2;
+
+         v0 = (const float *) (vertices + indexes[j+0] * vertex_size);
+         v1 = (const float *) (vertices + indexes[j+1] * vertex_size);
+         v2 = (const float *) (vertices + indexes[j+2] * vertex_size);
+
+         drawn += tri_draw(v0, v1, v2, tx, ty);
+      }
+
+      //printf("SPU %u: drew %u of %u\n", spu.init.id, drawn, render->num_indexes/3);
+
+      /* write color/z tiles back to main framebuffer, if dirtied */
+      put_cz_tiles(tx, ty);
+
+      wait_put_cz_tiles(); /* XXX seems unnecessary... */
+
+      spu.ctile_status[ty][tx] = spu.cur_ctile_status;
+      spu.ztile_status[ty][tx] = spu.cur_ztile_status;
+   }
+
+   D_PRINTF(CELL_DEBUG_CMD,
+            "RENDER done (%u tiles hit)\n",
+            num_tiles);
+}
diff --git a/src/gallium/drivers/cell/spu/spu_render.h b/src/gallium/drivers/cell/spu/spu_render.h
new file mode 100644
index 0000000000..493434f087
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_render.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef SPU_RENDER_H
+#define SPU_RENDER_H
+
+#include "cell/common.h"
+
+extern void
+cmd_render(const struct cell_command_render *render, uint *pos_incr);
+
+#endif /* SPU_RENDER_H */
+
diff --git a/src/gallium/drivers/cell/spu/spu_shuffle.h b/src/gallium/drivers/cell/spu/spu_shuffle.h
new file mode 100644
index 0000000000..74f2a0b6d2
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_shuffle.h
@@ -0,0 +1,186 @@
+#ifndef SPU_SHUFFLE_H
+#define SPU_SHUFFLE_H
+
+/*
+ * Generate shuffle patterns with minimal fuss.
+ *
+ * Based on ideas from 
+ * http://www.insomniacgames.com/tech/articles/0408/files/shuffles.pdf
+ *
+ * A-P indicates 0-15th position in first vector
+ * a-p indicates 0-15th position in second vector
+ *
+ * +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
+ * |00|01|02|03|04|05|06|07|08|09|0a|0b|0c|0d|0e|0f|
+ * +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
+ * |          A|          B|          C|          D|
+ * +-----+-----+-----+-----+-----+-----+-----+-----+
+ * |    A|    B|    C|    D|    E|    F|    G|    H|
+ * +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
+ * | A| B| C| D| E| F| G| H| I| J| K| L| M| N| O| P|
+ * +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
+ *
+ * x or X indicates 0xff
+ * 8 indicates 0x80
+ * 0 indicates 0x00
+ *
+ * The macros SHUFFLE4() SHUFFLE8() and SHUFFLE16() provide a const vector 
+ * unsigned char literal suitable for use with spu_shuffle().
+ *
+ * The macros SHUFB4() SHUFB8() and SHUFB16() provide a const qword vector 
+ * literal suitable for use with si_shufb().
+ *
+ *
+ * For example :
+ * SHUFB4(A,A,A,A)
+ * expands to :
+ * ((const qword){0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3})
+ * 
+ * SHUFFLE8(A,B,a,b,C,c,8,8)
+ * expands to :
+ * ((const vector unsigned char){0x00,0x01,0x02,0x03,0x10,0x11,0x12,0x13,
+ *				 0x04,0x05,0x14,0x15,0xe0,0xe0,0xe0,0xe0})
+ *
+ */
+
+#include <spu_intrinsics.h>
+
+#define SHUFFLE_PATTERN_4_A__  0x00, 0x01, 0x02, 0x03
+#define SHUFFLE_PATTERN_4_B__  0x04, 0x05, 0x06, 0x07
+#define SHUFFLE_PATTERN_4_C__  0x08, 0x09, 0x0a, 0x0b
+#define SHUFFLE_PATTERN_4_D__  0x0c, 0x0d, 0x0e, 0x0f
+#define SHUFFLE_PATTERN_4_a__  0x10, 0x11, 0x12, 0x13
+#define SHUFFLE_PATTERN_4_b__  0x14, 0x15, 0x16, 0x17
+#define SHUFFLE_PATTERN_4_c__  0x18, 0x19, 0x1a, 0x1b
+#define SHUFFLE_PATTERN_4_d__  0x1c, 0x1d, 0x1e, 0x1f
+#define SHUFFLE_PATTERN_4_X__  0xc0, 0xc0, 0xc0, 0xc0
+#define SHUFFLE_PATTERN_4_x__  0xc0, 0xc0, 0xc0, 0xc0
+#define SHUFFLE_PATTERN_4_0__  0x80, 0x80, 0x80, 0x80
+#define SHUFFLE_PATTERN_4_8__  0xe0, 0xe0, 0xe0, 0xe0
+
+#define SHUFFLE_VECTOR_4__(A, B, C, D) \
+   SHUFFLE_PATTERN_4_##A##__, \
+   SHUFFLE_PATTERN_4_##B##__, \
+   SHUFFLE_PATTERN_4_##C##__, \
+   SHUFFLE_PATTERN_4_##D##__
+
+#define SHUFFLE4(A, B, C, D) \
+   ((const vector unsigned char){ \
+      SHUFFLE_VECTOR_4__(A, B, C, D) \
+   })
+
+#define SHUFB4(A, B, C, D) \
+   ((const qword){ \
+      SHUFFLE_VECTOR_4__(A, B, C, D) \
+   })
+
+
+#define SHUFFLE_PATTERN_8_A__  0x00, 0x01
+#define SHUFFLE_PATTERN_8_B__  0x02, 0x03
+#define SHUFFLE_PATTERN_8_C__  0x04, 0x05
+#define SHUFFLE_PATTERN_8_D__  0x06, 0x07
+#define SHUFFLE_PATTERN_8_E__  0x08, 0x09
+#define SHUFFLE_PATTERN_8_F__  0x0a, 0x0b
+#define SHUFFLE_PATTERN_8_G__  0x0c, 0x0d
+#define SHUFFLE_PATTERN_8_H__  0x0e, 0x0f
+#define SHUFFLE_PATTERN_8_a__  0x10, 0x11
+#define SHUFFLE_PATTERN_8_b__  0x12, 0x13
+#define SHUFFLE_PATTERN_8_c__  0x14, 0x15
+#define SHUFFLE_PATTERN_8_d__  0x16, 0x17
+#define SHUFFLE_PATTERN_8_e__  0x18, 0x19
+#define SHUFFLE_PATTERN_8_f__  0x1a, 0x1b
+#define SHUFFLE_PATTERN_8_g__  0x1c, 0x1d
+#define SHUFFLE_PATTERN_8_h__  0x1e, 0x1f
+#define SHUFFLE_PATTERN_8_X__  0xc0, 0xc0
+#define SHUFFLE_PATTERN_8_x__  0xc0, 0xc0
+#define SHUFFLE_PATTERN_8_0__  0x80, 0x80
+#define SHUFFLE_PATTERN_8_8__  0xe0, 0xe0
+
+
+#define SHUFFLE_VECTOR_8__(A, B, C, D, E, F, G, H) \
+   SHUFFLE_PATTERN_8_##A##__, \
+   SHUFFLE_PATTERN_8_##B##__, \
+   SHUFFLE_PATTERN_8_##C##__, \
+   SHUFFLE_PATTERN_8_##D##__, \
+   SHUFFLE_PATTERN_8_##E##__, \
+   SHUFFLE_PATTERN_8_##F##__, \
+   SHUFFLE_PATTERN_8_##G##__, \
+   SHUFFLE_PATTERN_8_##H##__
+
+#define SHUFFLE8(A, B, C, D, E, F, G, H) \
+   ((const vector unsigned char){ \
+      SHUFFLE_VECTOR_8__(A, B, C, D, E, F, G, H) \
+   })
+
+#define SHUFB8(A, B, C, D, E, F, G, H) \
+   ((const qword){ \
+      SHUFFLE_VECTOR_8__(A, B, C, D, E, F, G, H) \
+   })
+
+
+#define SHUFFLE_PATTERN_16_A__  0x00
+#define SHUFFLE_PATTERN_16_B__  0x01
+#define SHUFFLE_PATTERN_16_C__  0x02
+#define SHUFFLE_PATTERN_16_D__  0x03
+#define SHUFFLE_PATTERN_16_E__  0x04
+#define SHUFFLE_PATTERN_16_F__  0x05
+#define SHUFFLE_PATTERN_16_G__  0x06
+#define SHUFFLE_PATTERN_16_H__  0x07
+#define SHUFFLE_PATTERN_16_I__  0x08
+#define SHUFFLE_PATTERN_16_J__  0x09
+#define SHUFFLE_PATTERN_16_K__  0x0a
+#define SHUFFLE_PATTERN_16_L__  0x0b
+#define SHUFFLE_PATTERN_16_M__  0x0c
+#define SHUFFLE_PATTERN_16_N__  0x0d
+#define SHUFFLE_PATTERN_16_O__  0x0e
+#define SHUFFLE_PATTERN_16_P__  0x0f
+#define SHUFFLE_PATTERN_16_a__  0x10
+#define SHUFFLE_PATTERN_16_b__  0x11
+#define SHUFFLE_PATTERN_16_c__  0x12
+#define SHUFFLE_PATTERN_16_d__  0x13
+#define SHUFFLE_PATTERN_16_e__  0x14
+#define SHUFFLE_PATTERN_16_f__  0x15
+#define SHUFFLE_PATTERN_16_g__  0x16
+#define SHUFFLE_PATTERN_16_h__  0x17
+#define SHUFFLE_PATTERN_16_i__  0x18
+#define SHUFFLE_PATTERN_16_j__  0x19
+#define SHUFFLE_PATTERN_16_k__  0x1a
+#define SHUFFLE_PATTERN_16_l__  0x1b
+#define SHUFFLE_PATTERN_16_m__  0x1c
+#define SHUFFLE_PATTERN_16_n__  0x1d
+#define SHUFFLE_PATTERN_16_o__  0x1e
+#define SHUFFLE_PATTERN_16_p__  0x1f
+#define SHUFFLE_PATTERN_16_X__  0xc0
+#define SHUFFLE_PATTERN_16_x__  0xc0
+#define SHUFFLE_PATTERN_16_0__  0x80
+#define SHUFFLE_PATTERN_16_8__  0xe0
+
+#define SHUFFLE_VECTOR_16__(A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P) \
+   SHUFFLE_PATTERN_16_##A##__, \
+   SHUFFLE_PATTERN_16_##B##__, \
+   SHUFFLE_PATTERN_16_##C##__, \
+   SHUFFLE_PATTERN_16_##D##__, \
+   SHUFFLE_PATTERN_16_##E##__, \
+   SHUFFLE_PATTERN_16_##F##__, \
+   SHUFFLE_PATTERN_16_##G##__, \
+   SHUFFLE_PATTERN_16_##H##__, \
+   SHUFFLE_PATTERN_16_##I##__, \
+   SHUFFLE_PATTERN_16_##J##__, \
+   SHUFFLE_PATTERN_16_##K##__, \
+   SHUFFLE_PATTERN_16_##L##__, \
+   SHUFFLE_PATTERN_16_##M##__, \
+   SHUFFLE_PATTERN_16_##N##__, \
+   SHUFFLE_PATTERN_16_##O##__, \
+   SHUFFLE_PATTERN_16_##P##__
+
+#define SHUFFLE16(A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P) \
+   ((const vector unsigned char){ \
+      SHUFFLE_VECTOR_16__(A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P) \
+   })
+
+#define SHUFB16(A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P) \
+   ((const qword){ \
+      SHUFFLE_VECTOR_16__(A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P) \
+   })
+
+#endif
diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c
new file mode 100644
index 0000000000..69784c8978
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_texture.c
@@ -0,0 +1,641 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include <math.h>
+
+#include "pipe/p_compiler.h"
+#include "spu_main.h"
+#include "spu_texture.h"
+#include "spu_tile.h"
+#include "spu_colorpack.h"
+#include "spu_dcache.h"
+
+
+/**
+ * Mark all tex cache entries as invalid.
+ */
+void
+invalidate_tex_cache(void)
+{
+   uint lvl;
+   for (lvl = 0; lvl < CELL_MAX_TEXTURE_LEVELS; lvl++) {
+      uint unit = 0;
+      uint bytes = 4 * spu.texture[unit].level[lvl].width
+         * spu.texture[unit].level[lvl].height;
+
+      if (spu.texture[unit].target == PIPE_TEXTURE_CUBE)
+         bytes *= 6;
+      else if (spu.texture[unit].target == PIPE_TEXTURE_3D)
+         bytes *= spu.texture[unit].level[lvl].depth;
+
+      spu_dcache_mark_dirty((unsigned) spu.texture[unit].level[lvl].start, bytes);
+   }
+}
+
+
+/**
+ * Get four texels from locations (x[0], y[0]), (x[1], y[1]) ...
+ *
+ * NOTE: in the typical case of bilinear filtering, the four texels
+ * are in a 2x2 group so we could get by with just two dcache fetches
+ * (two side-by-side texels per fetch).  But when bilinear filtering
+ * wraps around a texture edge, we'll probably need code like we have
+ * now.
+ * FURTHERMORE: since we're rasterizing a quad of 2x2 pixels at a time,
+ * it's quite likely that the four pixels in a quad will need some of the
+ * same texels.  So look into doing texture fetches for four pixels at
+ * a time.
+ */
+static void
+get_four_texels(const struct spu_texture_level *tlevel, uint face,
+                vec_int4 x, vec_int4 y,
+                vec_uint4 *texels)
+{
+   unsigned texture_ea = (uintptr_t) tlevel->start;
+   const vec_int4 tile_x = spu_rlmask(x, -5);  /* tile_x = x / 32 */
+   const vec_int4 tile_y = spu_rlmask(y, -5);  /* tile_y = y / 32 */
+   const qword offset_x = si_andi((qword) x, 0x1f); /* offset_x = x & 0x1f */
+   const qword offset_y = si_andi((qword) y, 0x1f); /* offset_y = y & 0x1f */
+
+   const qword tiles_per_row = (qword) spu_splats(tlevel->tiles_per_row);
+   const qword tile_size = (qword) spu_splats((unsigned) sizeof(tile_t));
+
+   qword tile_offset = si_mpya((qword) tile_y, tiles_per_row, (qword) tile_x);
+   tile_offset = si_mpy((qword) tile_offset, tile_size);
+
+   qword texel_offset = si_a(si_mpyui(offset_y, 32), offset_x);
+   texel_offset = si_mpyui(texel_offset, 4);
+   
+   vec_uint4 offset = (vec_uint4) si_a(tile_offset, texel_offset);
+   
+   texture_ea = texture_ea + face * tlevel->bytes_per_image;
+
+   spu_dcache_fetch_unaligned((qword *) & texels[0],
+                              texture_ea + spu_extract(offset, 0), 4);
+   spu_dcache_fetch_unaligned((qword *) & texels[1],
+                              texture_ea + spu_extract(offset, 1), 4);
+   spu_dcache_fetch_unaligned((qword *) & texels[2],
+                              texture_ea + spu_extract(offset, 2), 4);
+   spu_dcache_fetch_unaligned((qword *) & texels[3],
+                              texture_ea + spu_extract(offset, 3), 4);
+}
+
+
+/** clamp vec to [0, max] */
+static INLINE vector signed int
+spu_clamp(vector signed int vec, vector signed int max)
+{
+   static const vector signed int zero = {0,0,0,0};
+   vector unsigned int c;
+   c = spu_cmpgt(vec, zero);    /* c = vec > zero ? ~0 : 0 */
+   vec = spu_sel(zero, vec, c);
+   c = spu_cmpgt(vec, max);    /* c = vec > max ? ~0 : 0 */
+   vec = spu_sel(vec, max, c);
+   return vec;
+}
+
+
+
+/**
+ * Do nearest texture sampling for four pixels.
+ * \param colors  returned colors in SOA format (rrrr, gggg, bbbb, aaaa).
+ */
+void
+sample_texture_2d_nearest(vector float s, vector float t,
+                          uint unit, uint level, uint face,
+                          vector float colors[4])
+{
+   const struct spu_texture_level *tlevel = &spu.texture[unit].level[level];
+   vector float ss = spu_mul(s, tlevel->scale_s);
+   vector float tt = spu_mul(t, tlevel->scale_t);
+   vector signed int is = spu_convts(ss, 0);
+   vector signed int it = spu_convts(tt, 0);
+   vec_uint4 texels[4];
+
+   /* PIPE_TEX_WRAP_REPEAT */
+   is = spu_and(is, tlevel->mask_s);
+   it = spu_and(it, tlevel->mask_t);
+
+   /* PIPE_TEX_WRAP_CLAMP */
+   is = spu_clamp(is, tlevel->max_s);
+   it = spu_clamp(it, tlevel->max_t);
+
+   get_four_texels(tlevel, face, is, it, texels);
+
+   /* convert four packed ARGBA pixels to float RRRR,GGGG,BBBB,AAAA */
+   spu_unpack_A8R8G8B8_transpose4(texels, colors);
+}
+
+
+/**
+ * Do bilinear texture sampling for four pixels.
+ * \param colors  returned colors in SOA format (rrrr, gggg, bbbb, aaaa).
+ */
+void
+sample_texture_2d_bilinear(vector float s, vector float t,
+                           uint unit, uint level, uint face,
+                           vector float colors[4])
+{
+   const struct spu_texture_level *tlevel = &spu.texture[unit].level[level];
+   static const vector float half = {-0.5f, -0.5f, -0.5f, -0.5f};
+
+   vector float ss = spu_madd(s, tlevel->scale_s, half);
+   vector float tt = spu_madd(t, tlevel->scale_t, half);
+
+   vector signed int is0 = spu_convts(ss, 0);
+   vector signed int it0 = spu_convts(tt, 0);
+
+   /* is + 1, it + 1 */
+   vector signed int is1 = spu_add(is0, 1);
+   vector signed int it1 = spu_add(it0, 1);
+
+   /* PIPE_TEX_WRAP_REPEAT */
+   is0 = spu_and(is0, tlevel->mask_s);
+   it0 = spu_and(it0, tlevel->mask_t);
+   is1 = spu_and(is1, tlevel->mask_s);
+   it1 = spu_and(it1, tlevel->mask_t);
+
+   /* PIPE_TEX_WRAP_CLAMP */
+   is0 = spu_clamp(is0, tlevel->max_s);
+   it0 = spu_clamp(it0, tlevel->max_t);
+   is1 = spu_clamp(is1, tlevel->max_s);
+   it1 = spu_clamp(it1, tlevel->max_t);
+
+   /* get packed int texels */
+   vector unsigned int texels[16];
+   get_four_texels(tlevel, face, is0, it0, texels + 0);  /* upper-left */
+   get_four_texels(tlevel, face, is1, it0, texels + 4);  /* upper-right */
+   get_four_texels(tlevel, face, is0, it1, texels + 8);  /* lower-left */
+   get_four_texels(tlevel, face, is1, it1, texels + 12); /* lower-right */
+
+   /* convert packed int texels to float colors */
+   vector float ftexels[16];
+   spu_unpack_A8R8G8B8_transpose4(texels + 0, ftexels + 0);
+   spu_unpack_A8R8G8B8_transpose4(texels + 4, ftexels + 4);
+   spu_unpack_A8R8G8B8_transpose4(texels + 8, ftexels + 8);
+   spu_unpack_A8R8G8B8_transpose4(texels + 12, ftexels + 12);
+
+   /* Compute weighting factors in [0,1]
+    * Multiply texcoord by 1024, AND with 1023, convert back to float.
+    */
+   vector float ss1024 = spu_mul(ss, spu_splats(1024.0f));
+   vector signed int iss1024 = spu_convts(ss1024, 0);
+   iss1024 = spu_and(iss1024, 1023);
+   vector float sWeights0 = spu_convtf(iss1024, 10);
+
+   vector float tt1024 = spu_mul(tt, spu_splats(1024.0f));
+   vector signed int itt1024 = spu_convts(tt1024, 0);
+   itt1024 = spu_and(itt1024, 1023);
+   vector float tWeights0 = spu_convtf(itt1024, 10);
+
+   /* 1 - sWeight and 1 - tWeight */
+   vector float sWeights1 = spu_sub(spu_splats(1.0f), sWeights0);
+   vector float tWeights1 = spu_sub(spu_splats(1.0f), tWeights0);
+
+   /* reds, for four pixels */
+   ftexels[ 0] = spu_mul(ftexels[ 0], spu_mul(sWeights1, tWeights1)); /*ul*/
+   ftexels[ 4] = spu_mul(ftexels[ 4], spu_mul(sWeights0, tWeights1)); /*ur*/
+   ftexels[ 8] = spu_mul(ftexels[ 8], spu_mul(sWeights1, tWeights0)); /*ll*/
+   ftexels[12] = spu_mul(ftexels[12], spu_mul(sWeights0, tWeights0)); /*lr*/
+   colors[0] = spu_add(spu_add(ftexels[0], ftexels[4]),
+                       spu_add(ftexels[8], ftexels[12]));
+
+   /* greens, for four pixels */
+   ftexels[ 1] = spu_mul(ftexels[ 1], spu_mul(sWeights1, tWeights1)); /*ul*/
+   ftexels[ 5] = spu_mul(ftexels[ 5], spu_mul(sWeights0, tWeights1)); /*ur*/
+   ftexels[ 9] = spu_mul(ftexels[ 9], spu_mul(sWeights1, tWeights0)); /*ll*/
+   ftexels[13] = spu_mul(ftexels[13], spu_mul(sWeights0, tWeights0)); /*lr*/
+   colors[1] = spu_add(spu_add(ftexels[1], ftexels[5]),
+                       spu_add(ftexels[9], ftexels[13]));
+
+   /* blues, for four pixels */
+   ftexels[ 2] = spu_mul(ftexels[ 2], spu_mul(sWeights1, tWeights1)); /*ul*/
+   ftexels[ 6] = spu_mul(ftexels[ 6], spu_mul(sWeights0, tWeights1)); /*ur*/
+   ftexels[10] = spu_mul(ftexels[10], spu_mul(sWeights1, tWeights0)); /*ll*/
+   ftexels[14] = spu_mul(ftexels[14], spu_mul(sWeights0, tWeights0)); /*lr*/
+   colors[2] = spu_add(spu_add(ftexels[2], ftexels[6]),
+                       spu_add(ftexels[10], ftexels[14]));
+
+   /* alphas, for four pixels */
+   ftexels[ 3] = spu_mul(ftexels[ 3], spu_mul(sWeights1, tWeights1)); /*ul*/
+   ftexels[ 7] = spu_mul(ftexels[ 7], spu_mul(sWeights0, tWeights1)); /*ur*/
+   ftexels[11] = spu_mul(ftexels[11], spu_mul(sWeights1, tWeights0)); /*ll*/
+   ftexels[15] = spu_mul(ftexels[15], spu_mul(sWeights0, tWeights0)); /*lr*/
+   colors[3] = spu_add(spu_add(ftexels[3], ftexels[7]),
+                       spu_add(ftexels[11], ftexels[15]));
+}
+
+
+
+/**
+ * Adapted from /opt/cell/sdk/usr/spu/include/transpose_matrix4x4.h
+ */
+static INLINE void
+transpose(vector unsigned int *mOut0,
+          vector unsigned int *mOut1,
+          vector unsigned int *mOut2,
+          vector unsigned int *mOut3,
+          vector unsigned int *mIn)
+{
+  vector unsigned int abcd, efgh, ijkl, mnop;	/* input vectors */
+  vector unsigned int aeim, bfjn, cgko, dhlp;	/* output vectors */
+  vector unsigned int aibj, ckdl, emfn, gohp;	/* intermediate vectors */
+
+  vector unsigned char shufflehi = ((vector unsigned char) {
+					       0x00, 0x01, 0x02, 0x03,
+					       0x10, 0x11, 0x12, 0x13,
+					       0x04, 0x05, 0x06, 0x07,
+					       0x14, 0x15, 0x16, 0x17});
+  vector unsigned char shufflelo = ((vector unsigned char) {
+					       0x08, 0x09, 0x0A, 0x0B,
+					       0x18, 0x19, 0x1A, 0x1B,
+					       0x0C, 0x0D, 0x0E, 0x0F,
+					       0x1C, 0x1D, 0x1E, 0x1F});
+  abcd = *(mIn+0);
+  efgh = *(mIn+1);
+  ijkl = *(mIn+2);
+  mnop = *(mIn+3);
+
+  aibj = spu_shuffle(abcd, ijkl, shufflehi);
+  ckdl = spu_shuffle(abcd, ijkl, shufflelo);
+  emfn = spu_shuffle(efgh, mnop, shufflehi);
+  gohp = spu_shuffle(efgh, mnop, shufflelo);
+
+  aeim = spu_shuffle(aibj, emfn, shufflehi);
+  bfjn = spu_shuffle(aibj, emfn, shufflelo);
+  cgko = spu_shuffle(ckdl, gohp, shufflehi);
+  dhlp = spu_shuffle(ckdl, gohp, shufflelo);
+
+  *mOut0 = aeim;
+  *mOut1 = bfjn;
+  *mOut2 = cgko;
+  *mOut3 = dhlp;
+}
+
+
+/**
+ * Bilinear filtering, using int instead of float arithmetic for computing
+ * sample weights.
+ */
+void
+sample_texture_2d_bilinear_int(vector float s, vector float t,
+                               uint unit, uint level, uint face,
+                               vector float colors[4])
+{
+   const struct spu_texture_level *tlevel = &spu.texture[unit].level[level];
+   static const vector float half = {-0.5f, -0.5f, -0.5f, -0.5f};
+
+   /* Scale texcoords by size of texture, and add half pixel bias */
+   vector float ss = spu_madd(s, tlevel->scale_s, half);
+   vector float tt = spu_madd(t, tlevel->scale_t, half);
+
+   /* convert float coords to fixed-pt coords with 7 fraction bits */
+   vector signed int is = spu_convts(ss, 7);  /* XXX really need floor() here */
+   vector signed int it = spu_convts(tt, 7);  /* XXX really need floor() here */
+
+   /* compute integer texel weights in [0, 127] */
+   vector signed int sWeights0 = spu_and(is, 127);
+   vector signed int tWeights0 = spu_and(it, 127);
+   vector signed int sWeights1 = spu_sub(127, sWeights0);
+   vector signed int tWeights1 = spu_sub(127, tWeights0);
+
+   /* texel coords: is0 = is / 128, it0 = is / 128 */
+   vector signed int is0 = spu_rlmask(is, -7);
+   vector signed int it0 = spu_rlmask(it, -7);
+
+   /* texel coords: i1 = is0 + 1, it1 = it0 + 1 */
+   vector signed int is1 = spu_add(is0, 1);
+   vector signed int it1 = spu_add(it0, 1);
+
+   /* PIPE_TEX_WRAP_REPEAT */
+   is0 = spu_and(is0, tlevel->mask_s);
+   it0 = spu_and(it0, tlevel->mask_t);
+   is1 = spu_and(is1, tlevel->mask_s);
+   it1 = spu_and(it1, tlevel->mask_t);
+
+   /* PIPE_TEX_WRAP_CLAMP */
+   is0 = spu_clamp(is0, tlevel->max_s);
+   it0 = spu_clamp(it0, tlevel->max_t);
+   is1 = spu_clamp(is1, tlevel->max_s);
+   it1 = spu_clamp(it1, tlevel->max_t);
+
+   /* get packed int texels */
+   vector unsigned int texels[16];
+   get_four_texels(tlevel, face, is0, it0, texels + 0);  /* upper-left */
+   get_four_texels(tlevel, face, is1, it0, texels + 4);  /* upper-right */
+   get_four_texels(tlevel, face, is0, it1, texels + 8);  /* lower-left */
+   get_four_texels(tlevel, face, is1, it1, texels + 12); /* lower-right */
+
+   /* twiddle packed 32-bit BGRA pixels into RGBA as four unsigned ints */
+   {
+      static const unsigned char ZERO = 0x80;
+      int i;
+      for (i = 0; i < 16; i++) {
+         texels[i] = spu_shuffle(texels[i], texels[i],
+                                 ((vector unsigned char) {
+                                    ZERO, ZERO, ZERO, 1,
+                                    ZERO, ZERO, ZERO, 2,
+                                    ZERO, ZERO, ZERO, 3,
+                                    ZERO, ZERO, ZERO, 0}));
+      }
+   }
+
+   /* convert RGBA,RGBA,RGBA,RGBA to RRRR,GGGG,BBBB,AAAA */
+   vector unsigned int texel0, texel1, texel2, texel3, texel4, texel5, texel6, texel7,
+      texel8, texel9, texel10, texel11, texel12, texel13, texel14, texel15;
+   transpose(&texel0, &texel1, &texel2, &texel3, texels + 0);
+   transpose(&texel4, &texel5, &texel6, &texel7, texels + 4);
+   transpose(&texel8, &texel9, &texel10, &texel11, texels + 8);
+   transpose(&texel12, &texel13, &texel14, &texel15, texels + 12);
+
+   /* computed weighted colors */
+   vector unsigned int c0, c1, c2, c3, cSum;
+
+   /* red */
+   c0 = (vector unsigned int) si_mpy((qword) texel0, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/
+   c1 = (vector unsigned int) si_mpy((qword) texel4, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/
+   c2 = (vector unsigned int) si_mpy((qword) texel8, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/
+   c3 = (vector unsigned int) si_mpy((qword) texel12, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/
+   cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3));
+   colors[0] = spu_convtf(cSum, 22);
+
+   /* green */
+   c0 = (vector unsigned int) si_mpy((qword) texel1, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/
+   c1 = (vector unsigned int) si_mpy((qword) texel5, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/
+   c2 = (vector unsigned int) si_mpy((qword) texel9, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/
+   c3 = (vector unsigned int) si_mpy((qword) texel13, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/
+   cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3));
+   colors[1] = spu_convtf(cSum, 22);
+
+   /* blue */
+   c0 = (vector unsigned int) si_mpy((qword) texel2, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/
+   c1 = (vector unsigned int) si_mpy((qword) texel6, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/
+   c2 = (vector unsigned int) si_mpy((qword) texel10, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/
+   c3 = (vector unsigned int) si_mpy((qword) texel14, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/
+   cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3));
+   colors[2] = spu_convtf(cSum, 22);
+
+   /* alpha */
+   c0 = (vector unsigned int) si_mpy((qword) texel3, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/
+   c1 = (vector unsigned int) si_mpy((qword) texel7, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/
+   c2 = (vector unsigned int) si_mpy((qword) texel11, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/
+   c3 = (vector unsigned int) si_mpy((qword) texel15, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/
+   cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3));
+   colors[3] = spu_convtf(cSum, 22);
+}
+
+
+
+/**
+ * Compute level of detail factor from texcoords.
+ */
+static INLINE float
+compute_lambda_2d(uint unit, vector float s, vector float t)
+{
+   uint baseLevel = 0;
+   float width = spu.texture[unit].level[baseLevel].width;
+   float height = spu.texture[unit].level[baseLevel].width;
+   float dsdx = width * (spu_extract(s, 1) - spu_extract(s, 0));
+   float dsdy = width * (spu_extract(s, 2) - spu_extract(s, 0));
+   float dtdx = height * (spu_extract(t, 1) - spu_extract(t, 0));
+   float dtdy = height * (spu_extract(t, 2) - spu_extract(t, 0));
+#if 0
+   /* ideal value */
+   float x = dsdx * dsdx + dtdx * dtdx;
+   float y = dsdy * dsdy + dtdy * dtdy;
+   float rho = x > y ? x : y;
+   rho = sqrtf(rho);
+#else
+   /* approximation */
+   dsdx = fabsf(dsdx);
+   dsdy = fabsf(dsdy);
+   dtdx = fabsf(dtdx);
+   dtdy = fabsf(dtdy);
+   float rho = (dsdx + dsdy + dtdx + dtdy) * 0.5;
+#endif
+   float lambda = logf(rho) * 1.442695f; /* compute logbase2(rho) */
+   return lambda;
+}
+
+
+/**
+ * Blend two sets of colors according to weight.
+ */
+static void
+blend_colors(vector float c0[4], const vector float c1[4], float weight)
+{
+   vector float t = spu_splats(weight);
+   vector float dc0 = spu_sub(c1[0], c0[0]);
+   vector float dc1 = spu_sub(c1[1], c0[1]);
+   vector float dc2 = spu_sub(c1[2], c0[2]);
+   vector float dc3 = spu_sub(c1[3], c0[3]);
+   c0[0] = spu_madd(dc0, t, c0[0]);
+   c0[1] = spu_madd(dc1, t, c0[1]);
+   c0[2] = spu_madd(dc2, t, c0[2]);
+   c0[3] = spu_madd(dc3, t, c0[3]);
+}
+
+
+/**
+ * Texture sampling with level of detail selection and possibly mipmap
+ * interpolation.
+ */
+void
+sample_texture_2d_lod(vector float s, vector float t,
+                      uint unit, uint level_ignored, uint face,
+                      vector float colors[4])
+{
+   /*
+    * Note that we're computing a lambda/lod here that's used for all
+    * four pixels in the quad.
+    */
+   float lambda = compute_lambda_2d(unit, s, t);
+
+   (void) face;
+   (void) level_ignored;
+
+   /* apply lod bias */
+   lambda += spu.sampler[unit].lod_bias;
+
+   /* clamp */
+   if (lambda < spu.sampler[unit].min_lod)
+      lambda = spu.sampler[unit].min_lod;
+   else if (lambda > spu.sampler[unit].max_lod)
+      lambda = spu.sampler[unit].max_lod;
+
+   if (lambda <= 0.0f) {
+      /* magnify */
+      spu.mag_sample_texture_2d[unit](s, t, unit, 0, face, colors);
+   }
+   else {
+      /* minify */
+      if (spu.sampler[unit].min_img_filter == PIPE_TEX_FILTER_LINEAR) {
+         /* sample two mipmap levels and interpolate */
+         int level = (int) lambda;
+         if (level > (int) spu.texture[unit].max_level)
+            level = spu.texture[unit].max_level;
+         spu.min_sample_texture_2d[unit](s, t, unit, level, face, colors);
+         if (spu.sampler[unit].min_img_filter == PIPE_TEX_FILTER_LINEAR) {
+            /* sample second mipmap level */
+            float weight = lambda - (float) level;
+            level++;
+            if (level <= (int) spu.texture[unit].max_level) {
+               vector float colors2[4];
+               spu.min_sample_texture_2d[unit](s, t, unit, level, face, colors2);
+               blend_colors(colors, colors2, weight);
+            }
+         }
+      }
+      else {
+         /* sample one mipmap level */
+         int level = (int) (lambda + 0.5f);
+         if (level > (int) spu.texture[unit].max_level)
+            level = spu.texture[unit].max_level;
+         spu.min_sample_texture_2d[unit](s, t, unit, level, face, colors);
+      }
+   }
+}
+
+
+/** XXX need a SIMD version of this */
+static unsigned
+choose_cube_face(float rx, float ry, float rz, float *newS, float *newT)
+{
+   /*
+      major axis
+      direction     target                             sc     tc    ma
+      ----------    -------------------------------    ---    ---   ---
+       +rx          TEXTURE_CUBE_MAP_POSITIVE_X_EXT    -rz    -ry   rx
+       -rx          TEXTURE_CUBE_MAP_NEGATIVE_X_EXT    +rz    -ry   rx
+       +ry          TEXTURE_CUBE_MAP_POSITIVE_Y_EXT    +rx    +rz   ry
+       -ry          TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT    +rx    -rz   ry
+       +rz          TEXTURE_CUBE_MAP_POSITIVE_Z_EXT    +rx    -ry   rz
+       -rz          TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT    -rx    -ry   rz
+   */
+   const float arx = fabsf(rx);
+   const float ary = fabsf(ry);
+   const float arz = fabsf(rz);
+   unsigned face;
+   float sc, tc, ma;
+
+   if (arx > ary && arx > arz) {
+      if (rx >= 0.0F) {
+         face = PIPE_TEX_FACE_POS_X;
+         sc = -rz;
+         tc = -ry;
+         ma = arx;
+      }
+      else {
+         face = PIPE_TEX_FACE_NEG_X;
+         sc = rz;
+         tc = -ry;
+         ma = arx;
+      }
+   }
+   else if (ary > arx && ary > arz) {
+      if (ry >= 0.0F) {
+         face = PIPE_TEX_FACE_POS_Y;
+         sc = rx;
+         tc = rz;
+         ma = ary;
+      }
+      else {
+         face = PIPE_TEX_FACE_NEG_Y;
+         sc = rx;
+         tc = -rz;
+         ma = ary;
+      }
+   }
+   else {
+      if (rz > 0.0F) {
+         face = PIPE_TEX_FACE_POS_Z;
+         sc = rx;
+         tc = -ry;
+         ma = arz;
+      }
+      else {
+         face = PIPE_TEX_FACE_NEG_Z;
+         sc = -rx;
+         tc = -ry;
+         ma = arz;
+      }
+   }
+
+   *newS = (sc / ma + 1.0F) * 0.5F;
+   *newT = (tc / ma + 1.0F) * 0.5F;
+
+   return face;
+}
+
+
+
+void
+sample_texture_cube(vector float s, vector float t, vector float r,
+                    uint unit, vector float colors[4])
+{
+   uint p, faces[4], level = 0;
+   float newS[4], newT[4];
+
+   /* Compute cube faces referenced by the four sets of texcoords.
+    * XXX we should SIMD-ize this.
+    */
+   for (p = 0; p < 4; p++) {      
+      float rx = spu_extract(s, p);
+      float ry = spu_extract(t, p);
+      float rz = spu_extract(r, p);
+      faces[p] = choose_cube_face(rx, ry, rz, &newS[p], &newT[p]);
+   }
+
+   if (faces[0] == faces[1] &&
+       faces[0] == faces[2] &&
+       faces[0] == faces[3]) {
+      /* GOOD!  All four texcoords refer to the same cube face */
+      s = (vector float) {newS[0], newS[1], newS[2], newS[3]};
+      t = (vector float) {newT[0], newT[1], newT[2], newT[3]};
+      spu.sample_texture_2d[unit](s, t, unit, level, faces[0], colors);
+   }
+   else {
+      /* BAD!  The four texcoords refer to different faces */
+      for (p = 0; p < 4; p++) {      
+         vector float c[4];
+
+         spu.sample_texture_2d[unit](spu_splats(newS[p]), spu_splats(newT[p]),
+                                     unit, level, faces[p], c);
+
+         float red = spu_extract(c[0], p);
+         float green = spu_extract(c[1], p);
+         float blue = spu_extract(c[2], p);
+         float alpha = spu_extract(c[3], p);
+
+         colors[0] = spu_insert(red,   colors[0], p);
+         colors[1] = spu_insert(green, colors[1], p);
+         colors[2] = spu_insert(blue,  colors[2], p);
+         colors[3] = spu_insert(alpha, colors[3], p);
+      }
+   }
+}
diff --git a/src/gallium/drivers/cell/spu/spu_texture.h b/src/gallium/drivers/cell/spu/spu_texture.h
new file mode 100644
index 0000000000..7b75b007b5
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_texture.h
@@ -0,0 +1,67 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef SPU_TEXTURE_H
+#define SPU_TEXTURE_H
+
+
+#include "pipe/p_compiler.h"
+
+
+extern void
+invalidate_tex_cache(void);
+
+
+extern void
+sample_texture_2d_nearest(vector float s, vector float t,
+                          uint unit, uint level, uint face,
+                          vector float colors[4]);
+
+
+extern void
+sample_texture_2d_bilinear(vector float s, vector float t,
+                           uint unit, uint level, uint face,
+                           vector float colors[4]);
+
+extern void
+sample_texture_2d_bilinear_int(vector float s, vector float t,
+                               uint unit, uint level, uint face,
+                               vector float colors[4]);
+
+
+extern void
+sample_texture_2d_lod(vector float s, vector float t,
+                      uint unit, uint level, uint face,
+                      vector float colors[4]);
+
+
+extern void
+sample_texture_cube(vector float s, vector float t, vector float r,
+                    uint unit, vector float colors[4]);
+
+
+#endif /* SPU_TEXTURE_H */
diff --git a/src/gallium/drivers/cell/spu/spu_tile.c b/src/gallium/drivers/cell/spu/spu_tile.c
new file mode 100644
index 0000000000..6905015a48
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_tile.c
@@ -0,0 +1,126 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+
+#include "spu_tile.h"
+#include "spu_main.h"
+
+
+/**
+ * Get tile of color or Z values from main memory, put into SPU memory.
+ */
+void
+get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf)
+{
+   const uint offset = ty * spu.fb.width_tiles + tx;
+   const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? spu.fb.zsize : 4);
+   const ubyte *src = zBuf ? spu.fb.depth_start : spu.fb.color_start;
+
+   src += offset * bytesPerTile;
+
+   ASSERT(tx < spu.fb.width_tiles);
+   ASSERT(ty < spu.fb.height_tiles);
+   ASSERT_ALIGN16(tile);
+   /*
+   printf("get_tile:  dest: %p  src: 0x%x  size: %d\n",
+          tile, (unsigned int) src, bytesPerTile);
+   */
+   mfc_get(tile->ui,  /* dest in local memory */
+           (unsigned int) src, /* src in main memory */
+           bytesPerTile,
+           tag,
+           0, /* tid */
+           0  /* rid */);
+}
+
+
+/**
+ * Move tile of color or Z values from SPU memory to main memory.
+ */
+void
+put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf)
+{
+   const uint offset = ty * spu.fb.width_tiles + tx;
+   const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? spu.fb.zsize : 4);
+   ubyte *dst = zBuf ? spu.fb.depth_start : spu.fb.color_start;
+
+   dst += offset * bytesPerTile;
+
+   ASSERT(tx < spu.fb.width_tiles);
+   ASSERT(ty < spu.fb.height_tiles);
+   ASSERT_ALIGN16(tile);
+   /*
+   printf("SPU %u: put_tile:  src: %p  dst: 0x%x  size: %d\n",
+          spu.init.id,
+          tile, (unsigned int) dst, bytesPerTile);
+   */
+   mfc_put((void *) tile->ui,  /* src in local memory */
+           (unsigned int) dst,  /* dst in main memory */
+           bytesPerTile,
+           tag,
+           0, /* tid */
+           0  /* rid */);
+}
+
+
+/**
+ * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled
+ * tiles back to the main framebuffer.
+ */
+void
+really_clear_tiles(uint surfaceIndex)
+{
+   const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles;
+   uint i;
+
+   if (surfaceIndex == 0) {
+      clear_c_tile(&spu.ctile);
+
+      for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) {
+         uint tx = i % spu.fb.width_tiles;
+         uint ty = i / spu.fb.width_tiles;
+         if (spu.ctile_status[ty][tx] == TILE_STATUS_CLEAR) {
+            put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0);
+         }
+      }
+   }
+   else {
+      clear_z_tile(&spu.ztile);
+
+      for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) {
+         uint tx = i % spu.fb.width_tiles;
+         uint ty = i / spu.fb.width_tiles;
+         if (spu.ztile_status[ty][tx] == TILE_STATUS_CLEAR)
+            put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 1);
+      }
+   }
+
+#if 0
+   wait_on_mask(1 << TAG_SURFACE_CLEAR);
+#endif
+}
diff --git a/src/gallium/drivers/cell/spu/spu_tile.h b/src/gallium/drivers/cell/spu/spu_tile.h
new file mode 100644
index 0000000000..7bfb52be8f
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_tile.h
@@ -0,0 +1,75 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef SPU_TILE_H
+#define SPU_TILE_H
+
+
+#include <libmisc.h>
+#include <spu_mfcio.h>
+#include "spu_main.h"
+#include "cell/common.h"
+
+
+
+extern void
+get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf);
+
+extern void
+put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf);
+
+extern void
+really_clear_tiles(uint surfaceIndex);
+
+
+static INLINE void
+clear_c_tile(tile_t *ctile)
+{
+   memset32((uint*) ctile->ui,
+            spu.fb.color_clear_value,
+            TILE_SIZE * TILE_SIZE);
+}
+
+
+static INLINE void
+clear_z_tile(tile_t *ztile)
+{
+   if (spu.fb.zsize == 2) {
+      memset16((ushort*) ztile->us,
+               spu.fb.depth_clear_value,
+               TILE_SIZE * TILE_SIZE);
+   }
+   else {
+      ASSERT(spu.fb.zsize != 0);
+      memset32((uint*) ztile->ui,
+               spu.fb.depth_clear_value,
+               TILE_SIZE * TILE_SIZE);
+   }
+}
+
+
+#endif /* SPU_TILE_H */
diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c
new file mode 100644
index 0000000000..0d9fcb9997
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_tri.c
@@ -0,0 +1,809 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Triangle rendering within a tile.
+ */
+
+#include <transpose_matrix4x4.h>
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "util/u_math.h"
+#include "spu_colorpack.h"
+#include "spu_main.h"
+#include "spu_shuffle.h"
+#include "spu_texture.h"
+#include "spu_tile.h"
+#include "spu_tri.h"
+
+
+/** Masks are uint[4] vectors with each element being 0 or 0xffffffff */
+typedef vector unsigned int mask_t;
+
+
+
+/**
+ * Simplified types taken from other parts of Gallium
+ */
+struct vertex_header {
+   vector float data[1];
+};
+
+
+
+/* XXX fix this */
+#undef CEILF
+#define CEILF(X) ((float) (int) ((X) + 0.99999f))
+
+
+#define QUAD_TOP_LEFT     0
+#define QUAD_TOP_RIGHT    1
+#define QUAD_BOTTOM_LEFT  2
+#define QUAD_BOTTOM_RIGHT 3
+#define MASK_TOP_LEFT     (1 << QUAD_TOP_LEFT)
+#define MASK_TOP_RIGHT    (1 << QUAD_TOP_RIGHT)
+#define MASK_BOTTOM_LEFT  (1 << QUAD_BOTTOM_LEFT)
+#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT)
+#define MASK_ALL          0xf
+
+
+#define DEBUG_VERTS 0
+
+/**
+ * Triangle edge info
+ */
+struct edge {
+   union {
+      struct {
+         float dx;	/**< X(v1) - X(v0), used only during setup */
+         float dy;	/**< Y(v1) - Y(v0), used only during setup */
+      };
+      vec_float4 ds;    /**< vector accessor for dx and dy */
+   };
+   float dxdy;		/**< dx/dy */
+   float sx, sy;	/**< first sample point coord */
+   int lines;		/**< number of lines on this edge */
+};
+
+
+struct interp_coef
+{
+   vector float a0;
+   vector float dadx;
+   vector float dady;
+};
+
+
+/**
+ * Triangle setup info (derived from draw_stage).
+ * Also used for line drawing (taking some liberties).
+ */
+struct setup_stage {
+
+   /* Vertices are just an array of floats making up each attribute in
+    * turn.  Currently fixed at 4 floats, but should change in time.
+    * Codegen will help cope with this.
+    */
+   union {
+      struct {
+         const struct vertex_header *vmin;
+         const struct vertex_header *vmid;
+         const struct vertex_header *vmax;
+         const struct vertex_header *vprovoke;
+      };
+      qword vertex_headers;
+   };
+
+   struct edge ebot;
+   struct edge etop;
+   struct edge emaj;
+
+   float oneOverArea;  /* XXX maybe make into vector? */
+
+   uint facing;
+
+   uint tx, ty;  /**< position of current tile (x, y) */
+
+   int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy;
+
+   struct interp_coef coef[PIPE_MAX_SHADER_INPUTS];
+
+   struct {
+      vec_int4 quad; /**< [0] = row0, [1] = row1; {left[0],left[1],right[0],right[1]} */
+      int y;
+      unsigned y_flags;
+      unsigned mask;     /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */
+   } span;
+};
+
+
+static struct setup_stage setup;
+
+
+/**
+ * Evaluate attribute coefficients (plane equations) to compute
+ * attribute values for the four fragments in a quad.
+ * Eg: four colors will be computed (in AoS format).
+ */
+static INLINE void
+eval_coeff(uint slot, float x, float y, vector float w, vector float result[4])
+{
+   switch (spu.vertex_info.attrib[slot].interp_mode) {
+   case INTERP_CONSTANT:
+      result[QUAD_TOP_LEFT] =
+      result[QUAD_TOP_RIGHT] =
+      result[QUAD_BOTTOM_LEFT] =
+      result[QUAD_BOTTOM_RIGHT] = setup.coef[slot].a0;
+      break;
+   case INTERP_LINEAR:
+      {
+         vector float dadx = setup.coef[slot].dadx;
+         vector float dady = setup.coef[slot].dady;
+         vector float topLeft =
+            spu_add(setup.coef[slot].a0,
+                    spu_add(spu_mul(spu_splats(x), dadx),
+                            spu_mul(spu_splats(y), dady)));
+
+         result[QUAD_TOP_LEFT] = topLeft;
+         result[QUAD_TOP_RIGHT] = spu_add(topLeft, dadx);
+         result[QUAD_BOTTOM_LEFT] = spu_add(topLeft, dady);
+         result[QUAD_BOTTOM_RIGHT] = spu_add(spu_add(topLeft, dadx), dady);
+      }
+      break;
+   case INTERP_PERSPECTIVE:
+      {
+         vector float dadx = setup.coef[slot].dadx;
+         vector float dady = setup.coef[slot].dady;
+         vector float topLeft =
+            spu_add(setup.coef[slot].a0,
+                    spu_add(spu_mul(spu_splats(x), dadx),
+                            spu_mul(spu_splats(y), dady)));
+
+         vector float wInv = spu_re(w);  /* 1.0 / w */
+
+         result[QUAD_TOP_LEFT] = spu_mul(topLeft, wInv);
+         result[QUAD_TOP_RIGHT] = spu_mul(spu_add(topLeft, dadx), wInv);
+         result[QUAD_BOTTOM_LEFT] = spu_mul(spu_add(topLeft, dady), wInv);
+         result[QUAD_BOTTOM_RIGHT] = spu_mul(spu_add(spu_add(topLeft, dadx), dady), wInv);
+      }
+      break;
+   case INTERP_POS:
+   case INTERP_NONE:
+      break;
+   default:
+      ASSERT(0);
+   }
+}
+
+
+/**
+ * As above, but return 4 vectors in SOA format.
+ * XXX this will all be re-written someday.
+ */
+static INLINE void
+eval_coeff_soa(uint slot, float x, float y, vector float w, vector float result[4])
+{
+   eval_coeff(slot, x, y, w, result);
+   _transpose_matrix4x4(result, result);
+}
+
+
+/** Evalute coefficients to get Z for four pixels in a quad */
+static INLINE vector float
+eval_z(float x, float y)
+{
+   const uint slot = 0;
+   const float dzdx = spu_extract(setup.coef[slot].dadx, 2);
+   const float dzdy = spu_extract(setup.coef[slot].dady, 2);
+   const float topLeft = spu_extract(setup.coef[slot].a0, 2) + x * dzdx + y * dzdy;
+   const vector float topLeftv = spu_splats(topLeft);
+   const vector float derivs = (vector float) { 0.0, dzdx, dzdy, dzdx + dzdy };
+   return spu_add(topLeftv, derivs);
+}
+
+
+/** Evalute coefficients to get W for four pixels in a quad */
+static INLINE vector float
+eval_w(float x, float y)
+{
+   const uint slot = 0;
+   const float dwdx = spu_extract(setup.coef[slot].dadx, 3);
+   const float dwdy = spu_extract(setup.coef[slot].dady, 3);
+   const float topLeft = spu_extract(setup.coef[slot].a0, 3) + x * dwdx + y * dwdy;
+   const vector float topLeftv = spu_splats(topLeft);
+   const vector float derivs = (vector float) { 0.0, dwdx, dwdy, dwdx + dwdy };
+   return spu_add(topLeftv, derivs);
+}
+
+
+/**
+ * Emit a quad (pass to next stage).  No clipping is done.
+ * Note: about 1/5 to 1/7 of the time, mask is zero and this function
+ * should be skipped.  But adding the test for that slows things down
+ * overall.
+ */
+static INLINE void
+emit_quad( int x, int y, mask_t mask)
+{
+   /* If any bits in mask are set... */
+   if (spu_extract(spu_orx(mask), 0)) {
+      const int ix = x - setup.cliprect_minx;
+      const int iy = y - setup.cliprect_miny;
+
+      spu.cur_ctile_status = TILE_STATUS_DIRTY;
+      spu.cur_ztile_status = TILE_STATUS_DIRTY;
+
+      {
+         /*
+          * Run fragment shader, execute per-fragment ops, update fb/tile.
+          */
+         vector float inputs[4*4], outputs[2*4];
+         vector float fragZ = eval_z((float) x, (float) y);
+         vector float fragW = eval_w((float) x, (float) y);
+         vector unsigned int kill_mask;
+
+         /* setup inputs */
+#if 0
+         eval_coeff_soa(1, (float) x, (float) y, fragW, inputs);
+#else
+         uint i;
+         for (i = 0; i < spu.vertex_info.num_attribs; i++) {
+            eval_coeff_soa(i+1, (float) x, (float) y, fragW, inputs + i * 4);
+         }
+#endif
+         ASSERT(spu.fragment_program);
+         ASSERT(spu.fragment_ops);
+
+         /* Execute the current fragment program */
+         kill_mask = spu.fragment_program(inputs, outputs, spu.constants);
+
+         mask = spu_andc(mask, kill_mask);
+
+         /* Execute per-fragment/quad operations, including:
+          * alpha test, z test, stencil test, blend and framebuffer writing.
+          * Note that there are two different fragment operations functions
+          * that can be called, one for front-facing fragments, and one
+          * for back-facing fragments.  (Often the two are the same;
+          * but in some cases, like two-sided stenciling, they can be
+          * very different.)  So choose the correct function depending
+          * on the calculated facing.
+          */
+         spu.fragment_ops[setup.facing](ix, iy, &spu.ctile, &spu.ztile,
+                          fragZ,
+                          outputs[0*4+0],
+                          outputs[0*4+1],
+                          outputs[0*4+2],
+                          outputs[0*4+3],
+                          mask);
+      }
+   }
+}
+
+
+/**
+ * Given an X or Y coordinate, return the block/quad coordinate that it
+ * belongs to.
+ */
+static INLINE int
+block(int x)
+{
+   return x & ~1;
+}
+
+
+/**
+ * Render a horizontal span of quads
+ */
+static void
+flush_spans(void)
+{
+   int minleft, maxright;
+
+   const int l0 = spu_extract(setup.span.quad, 0);
+   const int l1 = spu_extract(setup.span.quad, 1);
+   const int r0 = spu_extract(setup.span.quad, 2);
+   const int r1 = spu_extract(setup.span.quad, 3);
+
+   switch (setup.span.y_flags) {
+   case 0x3:
+      /* both odd and even lines written (both quad rows) */
+      minleft = MIN2(l0, l1);
+      maxright = MAX2(r0, r1);
+      break;
+
+   case 0x1:
+      /* only even line written (quad top row) */
+      minleft = l0;
+      maxright = r0;
+      break;
+
+   case 0x2:
+      /* only odd line written (quad bottom row) */
+      minleft = l1;
+      maxright = r1;
+      break;
+
+   default:
+      return;
+   }
+
+   /* OK, we're very likely to need the tile data now.
+    * clear or finish waiting if needed.
+    */
+   if (spu.cur_ctile_status == TILE_STATUS_GETTING) {
+      /* wait for mfc_get() to complete */
+      //printf("SPU: %u: waiting for ctile\n", spu.init.id);
+      wait_on_mask(1 << TAG_READ_TILE_COLOR);
+      spu.cur_ctile_status = TILE_STATUS_CLEAN;
+   }
+   else if (spu.cur_ctile_status == TILE_STATUS_CLEAR) {
+      //printf("SPU %u: clearing C tile %u, %u\n", spu.init.id, setup.tx, setup.ty);
+      clear_c_tile(&spu.ctile);
+      spu.cur_ctile_status = TILE_STATUS_DIRTY;
+   }
+   ASSERT(spu.cur_ctile_status != TILE_STATUS_DEFINED);
+
+   if (spu.read_depth_stencil) {
+      if (spu.cur_ztile_status == TILE_STATUS_GETTING) {
+         /* wait for mfc_get() to complete */
+         //printf("SPU: %u: waiting for ztile\n", spu.init.id);
+         wait_on_mask(1 << TAG_READ_TILE_Z);
+         spu.cur_ztile_status = TILE_STATUS_CLEAN;
+      }
+      else if (spu.cur_ztile_status == TILE_STATUS_CLEAR) {
+         //printf("SPU %u: clearing Z tile %u, %u\n", spu.init.id, setup.tx, setup.ty);
+         clear_z_tile(&spu.ztile);
+         spu.cur_ztile_status = TILE_STATUS_DIRTY;
+      }
+      ASSERT(spu.cur_ztile_status != TILE_STATUS_DEFINED);
+   }
+
+   /* XXX this loop could be moved into the above switch cases... */
+   
+   /* Setup for mask calculation */
+   const vec_int4 quad_LlRr = setup.span.quad;
+   const vec_int4 quad_RrLl = spu_rlqwbyte(quad_LlRr, 8);
+   const vec_int4 quad_LLll = spu_shuffle(quad_LlRr, quad_LlRr, SHUFFLE4(A,A,B,B));
+   const vec_int4 quad_RRrr = spu_shuffle(quad_RrLl, quad_RrLl, SHUFFLE4(A,A,B,B));
+
+   const vec_int4 twos = spu_splats(2);
+
+   const int x = block(minleft);
+   vec_int4 xs = {x, x+1, x, x+1};
+
+   for (; spu_extract(xs, 0) <= block(maxright); xs += twos) {
+      /**
+       * Computes mask to indicate which pixels in the 2x2 quad are actually
+       * inside the triangle's bounds.
+       */
+      
+      /* Calculate ({x,x+1,x,x+1} >= {l[0],l[0],l[1],l[1]}) */
+      const mask_t gt_LLll_xs = spu_cmpgt(quad_LLll, xs);
+      const mask_t gte_xs_LLll = spu_nand(gt_LLll_xs, gt_LLll_xs); 
+      
+      /* Calculate ({r[0],r[0],r[1],r[1]} > {x,x+1,x,x+1}) */
+      const mask_t gt_RRrr_xs = spu_cmpgt(quad_RRrr, xs);
+
+      /* Combine results to create mask */
+      const mask_t mask = spu_and(gte_xs_LLll, gt_RRrr_xs);
+
+      emit_quad(spu_extract(xs, 0), setup.span.y, mask);
+   }
+
+   setup.span.y = 0;
+   setup.span.y_flags = 0;
+   /* Zero right elements */
+   setup.span.quad = spu_shuffle(setup.span.quad, setup.span.quad, SHUFFLE4(A,B,0,0));
+}
+
+
+#if DEBUG_VERTS
+static void
+print_vertex(const struct vertex_header *v)
+{
+   uint i;
+   fprintf(stderr, "  Vertex: (%p)\n", v);
+   for (i = 0; i < spu.vertex_info.num_attribs; i++) {
+      fprintf(stderr, "    %d: %f %f %f %f\n",  i, 
+              spu_extract(v->data[i], 0),
+              spu_extract(v->data[i], 1),
+              spu_extract(v->data[i], 2),
+              spu_extract(v->data[i], 3));
+   }
+}
+#endif
+
+
+/**
+ * Sort vertices from top to bottom.
+ * Compute area and determine front vs. back facing.
+ * Do coarse clip test against tile bounds
+ * \return  FALSE if tri is totally outside tile, TRUE otherwise
+ */
+static boolean
+setup_sort_vertices(const struct vertex_header *v0,
+                    const struct vertex_header *v1,
+                    const struct vertex_header *v2)
+{
+   float area, sign;
+
+#if DEBUG_VERTS
+   if (spu.init.id==0) {
+      fprintf(stderr, "SPU %u: Triangle:\n", spu.init.id);
+      print_vertex(v0);
+      print_vertex(v1);
+      print_vertex(v2);
+   }
+#endif
+
+   /* determine bottom to top order of vertices */
+   {
+      /* A table of shuffle patterns for putting vertex_header pointers into
+         correct order.  Quite magical. */
+      const vec_uchar16 sort_order_patterns[] = {
+         SHUFFLE4(A,B,C,C),
+         SHUFFLE4(C,A,B,C),
+         SHUFFLE4(A,C,B,C),
+         SHUFFLE4(B,C,A,C),
+         SHUFFLE4(B,A,C,C),
+         SHUFFLE4(C,B,A,C) };
+
+      /* The vertex_header pointers, packed for easy shuffling later */
+      const vec_uint4 vs = {(unsigned)v0, (unsigned)v1, (unsigned)v2};
+
+      /* Collate y values into two vectors for comparison.
+         Using only one shuffle constant! ;) */
+      const vec_float4 y_02_ = spu_shuffle(v0->data[0], v2->data[0], SHUFFLE4(0,B,b,C));
+      const vec_float4 y_10_ = spu_shuffle(v1->data[0], v0->data[0], SHUFFLE4(0,B,b,C));
+      const vec_float4 y_012 = spu_shuffle(y_02_, v1->data[0], SHUFFLE4(0,B,b,C));
+      const vec_float4 y_120 = spu_shuffle(y_10_, v2->data[0], SHUFFLE4(0,B,b,C));
+
+      /* Perform comparison: {y0,y1,y2} > {y1,y2,y0} */
+      const vec_uint4 compare = spu_cmpgt(y_012, y_120);
+      /* Compress the result of the comparison into 4 bits */
+      const vec_uint4 gather = spu_gather(compare);
+      /* Subtract one to attain the index into the LUT.  Magical. */
+      const unsigned int index = spu_extract(gather, 0) - 1;
+
+      /* Load the appropriate pattern and construct the desired vector. */
+      setup.vertex_headers = (qword)spu_shuffle(vs, vs, sort_order_patterns[index]);
+
+      /* Using the result of the comparison, set sign.
+         Very magical. */
+      sign = ((si_to_uint(si_cntb((qword)gather)) == 2) ? 1.0f : -1.0f);
+   }
+
+   /* Check if triangle is completely outside the tile bounds */
+   if (spu_extract(setup.vmin->data[0], 1) > setup.cliprect_maxy)
+      return FALSE;
+   if (spu_extract(setup.vmax->data[0], 1) < setup.cliprect_miny)
+      return FALSE;
+   if (spu_extract(setup.vmin->data[0], 0) < setup.cliprect_minx &&
+       spu_extract(setup.vmid->data[0], 0) < setup.cliprect_minx &&
+       spu_extract(setup.vmax->data[0], 0) < setup.cliprect_minx)
+      return FALSE;
+   if (spu_extract(setup.vmin->data[0], 0) > setup.cliprect_maxx &&
+       spu_extract(setup.vmid->data[0], 0) > setup.cliprect_maxx &&
+       spu_extract(setup.vmax->data[0], 0) > setup.cliprect_maxx)
+      return FALSE;
+
+   setup.ebot.ds = spu_sub(setup.vmid->data[0], setup.vmin->data[0]);
+   setup.emaj.ds = spu_sub(setup.vmax->data[0], setup.vmin->data[0]);
+   setup.etop.ds = spu_sub(setup.vmax->data[0], setup.vmid->data[0]);
+
+   /*
+    * Compute triangle's area.  Use 1/area to compute partial
+    * derivatives of attributes later.
+    */
+   area = setup.emaj.dx * setup.ebot.dy - setup.ebot.dx * setup.emaj.dy;
+
+   setup.oneOverArea = 1.0f / area;
+
+   /* The product of area * sign indicates front/back orientation (0/1).
+    * Just in case someone gets the bright idea of switching the front
+    * and back constants without noticing that we're assuming their
+    * values in this operation, also assert that the values are
+    * what we think they are.
+    */
+   ASSERT(CELL_FACING_FRONT == 0);
+   ASSERT(CELL_FACING_BACK == 1);
+   setup.facing = (area * sign > 0.0f)
+      ^ (spu.rasterizer.front_winding == PIPE_WINDING_CW);
+
+   return TRUE;
+}
+
+
+/**
+ * Compute a0 for a constant-valued coefficient (GL_FLAT shading).
+ * The value value comes from vertex->data[slot].
+ * The result will be put into setup.coef[slot].a0.
+ * \param slot  which attribute slot 
+ */
+static INLINE void
+const_coeff4(uint slot)
+{
+   setup.coef[slot].dadx = (vector float) {0.0, 0.0, 0.0, 0.0};
+   setup.coef[slot].dady = (vector float) {0.0, 0.0, 0.0, 0.0};
+   setup.coef[slot].a0 = setup.vprovoke->data[slot];
+}
+
+
+/**
+ * As above, but interp setup all four vector components.
+ */
+static INLINE void
+tri_linear_coeff4(uint slot)
+{
+   const vector float vmin_d = setup.vmin->data[slot];
+   const vector float vmid_d = setup.vmid->data[slot];
+   const vector float vmax_d = setup.vmax->data[slot];
+   const vector float xxxx = spu_splats(spu_extract(setup.vmin->data[0], 0) - 0.5f);
+   const vector float yyyy = spu_splats(spu_extract(setup.vmin->data[0], 1) - 0.5f);
+
+   vector float botda = vmid_d - vmin_d;
+   vector float majda = vmax_d - vmin_d;
+
+   vector float a = spu_sub(spu_mul(spu_splats(setup.ebot.dy), majda),
+                            spu_mul(botda, spu_splats(setup.emaj.dy)));
+   vector float b = spu_sub(spu_mul(spu_splats(setup.emaj.dx), botda),
+                            spu_mul(majda, spu_splats(setup.ebot.dx)));
+
+   setup.coef[slot].dadx = spu_mul(a, spu_splats(setup.oneOverArea));
+   setup.coef[slot].dady = spu_mul(b, spu_splats(setup.oneOverArea));
+
+   vector float tempx = spu_mul(setup.coef[slot].dadx, xxxx);
+   vector float tempy = spu_mul(setup.coef[slot].dady, yyyy);
+                         
+   setup.coef[slot].a0 = spu_sub(vmin_d, spu_add(tempx, tempy));
+}
+
+
+/**
+ * Compute a0, dadx and dady for a perspective-corrected interpolant,
+ * for a triangle.
+ * We basically multiply the vertex value by 1/w before computing
+ * the plane coefficients (a0, dadx, dady).
+ * Later, when we compute the value at a particular fragment position we'll
+ * divide the interpolated value by the interpolated W at that fragment.
+ */
+static void
+tri_persp_coeff4(uint slot)
+{
+   const vector float xxxx = spu_splats(spu_extract(setup.vmin->data[0], 0) - 0.5f);
+   const vector float yyyy = spu_splats(spu_extract(setup.vmin->data[0], 1) - 0.5f);
+
+   const vector float vmin_w = spu_splats(spu_extract(setup.vmin->data[0], 3));
+   const vector float vmid_w = spu_splats(spu_extract(setup.vmid->data[0], 3));
+   const vector float vmax_w = spu_splats(spu_extract(setup.vmax->data[0], 3));
+
+   vector float vmin_d = setup.vmin->data[slot];
+   vector float vmid_d = setup.vmid->data[slot];
+   vector float vmax_d = setup.vmax->data[slot];
+
+   vmin_d = spu_mul(vmin_d, vmin_w);
+   vmid_d = spu_mul(vmid_d, vmid_w);
+   vmax_d = spu_mul(vmax_d, vmax_w);
+
+   vector float botda = vmid_d - vmin_d;
+   vector float majda = vmax_d - vmin_d;
+
+   vector float a = spu_sub(spu_mul(spu_splats(setup.ebot.dy), majda),
+                            spu_mul(botda, spu_splats(setup.emaj.dy)));
+   vector float b = spu_sub(spu_mul(spu_splats(setup.emaj.dx), botda),
+                            spu_mul(majda, spu_splats(setup.ebot.dx)));
+
+   setup.coef[slot].dadx = spu_mul(a, spu_splats(setup.oneOverArea));
+   setup.coef[slot].dady = spu_mul(b, spu_splats(setup.oneOverArea));
+
+   vector float tempx = spu_mul(setup.coef[slot].dadx, xxxx);
+   vector float tempy = spu_mul(setup.coef[slot].dady, yyyy);
+                         
+   setup.coef[slot].a0 = spu_sub(vmin_d, spu_add(tempx, tempy));
+}
+
+
+
+/**
+ * Compute the setup.coef[] array dadx, dady, a0 values.
+ * Must be called after setup.vmin,vmid,vmax,vprovoke are initialized.
+ */
+static void
+setup_tri_coefficients(void)
+{
+   uint i;
+
+   for (i = 0; i < spu.vertex_info.num_attribs; i++) {
+      switch (spu.vertex_info.attrib[i].interp_mode) {
+      case INTERP_NONE:
+         break;
+      case INTERP_CONSTANT:
+         const_coeff4(i);
+         break;
+      case INTERP_POS:
+         /* fall-through */
+      case INTERP_LINEAR:
+         tri_linear_coeff4(i);
+         break;
+      case INTERP_PERSPECTIVE:
+         tri_persp_coeff4(i);
+         break;
+      default:
+         ASSERT(0);
+      }
+   }
+}
+
+
+static void
+setup_tri_edges(void)
+{
+   float vmin_x = spu_extract(setup.vmin->data[0], 0) + 0.5f;
+   float vmid_x = spu_extract(setup.vmid->data[0], 0) + 0.5f;
+
+   float vmin_y = spu_extract(setup.vmin->data[0], 1) - 0.5f;
+   float vmid_y = spu_extract(setup.vmid->data[0], 1) - 0.5f;
+   float vmax_y = spu_extract(setup.vmax->data[0], 1) - 0.5f;
+
+   setup.emaj.sy = CEILF(vmin_y);
+   setup.emaj.lines = (int) CEILF(vmax_y - setup.emaj.sy);
+   setup.emaj.dxdy = setup.emaj.dx / setup.emaj.dy;
+   setup.emaj.sx = vmin_x + (setup.emaj.sy - vmin_y) * setup.emaj.dxdy;
+
+   setup.etop.sy = CEILF(vmid_y);
+   setup.etop.lines = (int) CEILF(vmax_y - setup.etop.sy);
+   setup.etop.dxdy = setup.etop.dx / setup.etop.dy;
+   setup.etop.sx = vmid_x + (setup.etop.sy - vmid_y) * setup.etop.dxdy;
+
+   setup.ebot.sy = CEILF(vmin_y);
+   setup.ebot.lines = (int) CEILF(vmid_y - setup.ebot.sy);
+   setup.ebot.dxdy = setup.ebot.dx / setup.ebot.dy;
+   setup.ebot.sx = vmin_x + (setup.ebot.sy - vmin_y) * setup.ebot.dxdy;
+}
+
+
+/**
+ * Render the upper or lower half of a triangle.
+ * Scissoring/cliprect is applied here too.
+ */
+static void
+subtriangle(struct edge *eleft, struct edge *eright, unsigned lines)
+{
+   const int minx = setup.cliprect_minx;
+   const int maxx = setup.cliprect_maxx;
+   const int miny = setup.cliprect_miny;
+   const int maxy = setup.cliprect_maxy;
+   int y, start_y, finish_y;
+   int sy = (int)eleft->sy;
+
+   ASSERT((int)eleft->sy == (int) eright->sy);
+
+   /* clip top/bottom */
+   start_y = sy;
+   finish_y = sy + lines;
+
+   if (start_y < miny)
+      start_y = miny;
+
+   if (finish_y > maxy)
+      finish_y = maxy;
+
+   start_y -= sy;
+   finish_y -= sy;
+
+   /*
+   _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y);  
+   */
+
+   for (y = start_y; y < finish_y; y++) {
+
+      /* avoid accumulating adds as floats don't have the precision to
+       * accurately iterate large triangle edges that way.  luckily we
+       * can just multiply these days.
+       *
+       * this is all drowned out by the attribute interpolation anyway.
+       */
+      int left = (int)(eleft->sx + y * eleft->dxdy);
+      int right = (int)(eright->sx + y * eright->dxdy);
+
+      /* clip left/right */
+      if (left < minx)
+         left = minx;
+      if (right > maxx)
+         right = maxx;
+
+      if (left < right) {
+         int _y = sy + y;
+         if (block(_y) != setup.span.y) {
+            flush_spans();
+            setup.span.y = block(_y);
+         }
+
+         int offset = _y&1;
+         vec_int4 quad_LlRr = {left, left, right, right};
+         /* Store left and right in 0 or 1 row of quad based on offset */
+         setup.span.quad = spu_sel(quad_LlRr, setup.span.quad, spu_maskw(5<<offset));
+         setup.span.y_flags |= 1<<offset;
+      }
+   }
+
+
+   /* save the values so that emaj can be restarted:
+    */
+   eleft->sx += lines * eleft->dxdy;
+   eright->sx += lines * eright->dxdy;
+   eleft->sy += lines;
+   eright->sy += lines;
+}
+
+
+/**
+ * Draw triangle into tile at (tx, ty) (tile coords)
+ * The tile data should have already been fetched.
+ */
+boolean
+tri_draw(const float *v0, const float *v1, const float *v2,
+         uint tx, uint ty)
+{
+   setup.tx = tx;
+   setup.ty = ty;
+
+   /* set clipping bounds to tile bounds */
+   setup.cliprect_minx = tx * TILE_SIZE;
+   setup.cliprect_miny = ty * TILE_SIZE;
+   setup.cliprect_maxx = (tx + 1) * TILE_SIZE;
+   setup.cliprect_maxy = (ty + 1) * TILE_SIZE;
+
+   if (!setup_sort_vertices((struct vertex_header *) v0,
+                            (struct vertex_header *) v1,
+                            (struct vertex_header *) v2)) {
+      return FALSE; /* totally clipped */
+   }
+
+   setup_tri_coefficients();
+   setup_tri_edges();
+
+   setup.span.y = 0;
+   setup.span.y_flags = 0;
+   /* Zero right elements */
+   setup.span.quad = spu_shuffle(setup.span.quad, setup.span.quad, SHUFFLE4(A,B,0,0));
+
+   if (setup.oneOverArea < 0.0) {
+      /* emaj on left */
+      subtriangle( &setup.emaj, &setup.ebot, setup.ebot.lines );
+      subtriangle( &setup.emaj, &setup.etop, setup.etop.lines );
+   }
+   else {
+      /* emaj on right */
+      subtriangle( &setup.ebot, &setup.emaj, setup.ebot.lines );
+      subtriangle( &setup.etop, &setup.emaj, setup.etop.lines );
+   }
+
+   flush_spans();
+
+   return TRUE;
+}
diff --git a/src/gallium/drivers/cell/spu/spu_tri.h b/src/gallium/drivers/cell/spu/spu_tri.h
new file mode 100644
index 0000000000..aa694dd7c9
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_tri.h
@@ -0,0 +1,37 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef SPU_TRI_H
+#define SPU_TRI_H
+
+
+extern boolean
+tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty);
+
+
+#endif /* SPU_TRI_H */
diff --git a/src/gallium/drivers/cell/spu/spu_util.c b/src/gallium/drivers/cell/spu/spu_util.c
new file mode 100644
index 0000000000..b8a0d4a265
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_util.c
@@ -0,0 +1,167 @@
+
+#include "cell/common.h"
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_debug.h"
+#include "tgsi/tgsi_parse.h"
+//#include "tgsi_build.h"
+#include "tgsi/tgsi_util.h"
+
+unsigned
+tgsi_util_get_src_register_swizzle(
+   const struct tgsi_src_register *reg,
+   unsigned component )
+{
+   switch( component ) {
+   case 0:
+      return reg->SwizzleX;
+   case 1:
+      return reg->SwizzleY;
+   case 2:
+      return reg->SwizzleZ;
+   case 3:
+      return reg->SwizzleW;
+   default:
+      ASSERT( 0 );
+   }
+   return 0;
+}
+
+unsigned
+tgsi_util_get_src_register_extswizzle(
+   const struct tgsi_src_register_ext_swz *reg,
+   unsigned component )
+{
+   switch( component ) {
+   case 0:
+      return reg->ExtSwizzleX;
+   case 1:
+      return reg->ExtSwizzleY;
+   case 2:
+      return reg->ExtSwizzleZ;
+   case 3:
+      return reg->ExtSwizzleW;
+   default:
+      ASSERT( 0 );
+   }
+   return 0;
+}
+
+unsigned
+tgsi_util_get_full_src_register_extswizzle(
+   const struct tgsi_full_src_register  *reg,
+   unsigned component )
+{
+   unsigned swizzle;
+
+   /*
+    * First, calculate  the   extended swizzle for a given channel. This will give
+    * us either a channel index into the simple swizzle or  a constant 1 or   0.
+    */
+   swizzle = tgsi_util_get_src_register_extswizzle(
+      &reg->SrcRegisterExtSwz,
+      component );
+
+   ASSERT (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X);
+   ASSERT (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y);
+   ASSERT (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z);
+   ASSERT (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W);
+   ASSERT (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W);
+   ASSERT (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W);
+
+   /*
+    * Second, calculate the simple  swizzle  for   the   unswizzled channel index.
+    * Leave the constants intact, they are   not   affected by the   simple swizzle.
+    */
+   if( swizzle <= TGSI_SWIZZLE_W ) {
+      swizzle = tgsi_util_get_src_register_swizzle(
+         &reg->SrcRegister,
+         component );
+   }
+
+   return swizzle;
+}
+
+unsigned
+tgsi_util_get_src_register_extnegate(
+   const  struct tgsi_src_register_ext_swz *reg,
+   unsigned component )
+{
+   switch( component ) {
+   case 0:
+      return reg->NegateX;
+   case 1:
+      return reg->NegateY;
+   case 2:
+      return reg->NegateZ;
+   case 3:
+      return reg->NegateW;
+   default:
+      ASSERT( 0 );
+   }
+   return 0;
+}
+
+void
+tgsi_util_set_src_register_extnegate(
+   struct tgsi_src_register_ext_swz *reg,
+   unsigned negate,
+   unsigned component )
+{
+   switch( component ) {
+   case 0:
+      reg->NegateX = negate;
+      break;
+   case 1:
+      reg->NegateY = negate;
+      break;
+   case 2:
+      reg->NegateZ = negate;
+      break;
+   case 3:
+      reg->NegateW = negate;
+      break;
+   default:
+      ASSERT( 0 );
+   }
+}
+
+unsigned
+tgsi_util_get_full_src_register_sign_mode(
+   const struct  tgsi_full_src_register *reg,
+   unsigned component )
+{
+   unsigned sign_mode;
+
+   if( reg->SrcRegisterExtMod.Absolute ) {
+      /* Consider only the post-abs negation. */
+
+      if( reg->SrcRegisterExtMod.Negate ) {
+         sign_mode = TGSI_UTIL_SIGN_SET;
+      }
+      else {
+         sign_mode = TGSI_UTIL_SIGN_CLEAR;
+      }
+   }
+   else {
+      /* Accumulate the three negations. */
+
+      unsigned negate;
+
+      negate = reg->SrcRegister.Negate;
+      if( tgsi_util_get_src_register_extnegate( &reg->SrcRegisterExtSwz, component ) ) {
+         negate = !negate;
+      }
+      if( reg->SrcRegisterExtMod.Negate ) {
+         negate = !negate;
+      }
+
+      if( negate ) {
+         sign_mode = TGSI_UTIL_SIGN_TOGGLE;
+      }
+      else {
+         sign_mode = TGSI_UTIL_SIGN_KEEP;
+      }
+   }
+
+   return sign_mode;
+}
diff --git a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c
new file mode 100644
index 0000000000..03375d84a5
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c
@@ -0,0 +1,145 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * (C) Copyright IBM Corporation 2008
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Ian Romanick <idr@us.ibm.com>
+  */
+
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+#include "spu_exec.h"
+#include "spu_vertex_shader.h"
+#include "spu_main.h"
+#include "spu_dcache.h"
+
+typedef void (*spu_fetch_func)(qword *out, const qword *in,
+			       const qword *shuffle_data);
+
+
+static const qword fetch_shuffle_data[5] ALIGN16_ATTRIB = {
+   /* Shuffle used by CVT_64_FLOAT
+    */
+   {
+      0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13,
+      0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
+   },
+
+   /* Shuffle used by CVT_8_USCALED and CVT_8_SSCALED
+    */
+   {
+      0x00, 0x80, 0x80, 0x80, 0x01, 0x80, 0x80, 0x80,
+      0x02, 0x80, 0x80, 0x80, 0x03, 0x80, 0x80, 0x80,
+   },
+   
+   /* Shuffle used by CVT_16_USCALED and CVT_16_SSCALED
+    */
+   {
+      0x00, 0x01, 0x80, 0x80, 0x02, 0x03, 0x80, 0x80,
+      0x04, 0x05, 0x80, 0x80, 0x06, 0x07, 0x80, 0x80,
+   },
+   
+   /* High value shuffle used by trans4x4.
+    */
+   {
+      0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13,
+      0x04, 0x05, 0x06, 0x07, 0x14, 0x15, 0x16, 0x17
+   },
+
+   /* Low value shuffle used by trans4x4.
+    */
+   {
+      0x08, 0x09, 0x0A, 0x0B, 0x18, 0x19, 0x1A, 0x1B,
+      0x0C, 0x0D, 0x0E, 0x0F, 0x1C, 0x1D, 0x1E, 0x1F
+   }
+};
+
+
+/**
+ * Fetch vertex attributes for 'count' vertices.
+ */
+static void generic_vertex_fetch(struct spu_vs_context *draw,
+                                 struct spu_exec_machine *machine,
+                                 const unsigned *elts,
+                                 unsigned count)
+{
+   unsigned nr_attrs = draw->vertex_fetch.nr_attrs;
+   unsigned attr;
+
+   ASSERT(count <= 4);
+
+#if DRAW_DBG
+   printf("SPU: %s count = %u, nr_attrs = %u\n", 
+          __FUNCTION__, count, nr_attrs);
+#endif
+
+   /* loop over vertex attributes (vertex shader inputs)
+    */
+   for (attr = 0; attr < nr_attrs; attr++) {
+      const unsigned pitch = draw->vertex_fetch.pitch[attr];
+      const uint64_t src = draw->vertex_fetch.src_ptr[attr];
+      const spu_fetch_func fetch = (spu_fetch_func)
+	  (draw->vertex_fetch.code + draw->vertex_fetch.code_offset[attr]);
+      unsigned i;
+      unsigned idx;
+      const unsigned bytes_per_entry = draw->vertex_fetch.size[attr];
+      const unsigned quads_per_entry = (bytes_per_entry + 15) / 16;
+      qword in[2 * 4] ALIGN16_ATTRIB;
+
+
+      /* Fetch four attributes for four vertices.  
+       */
+      idx = 0;
+      for (i = 0; i < count; i++) {
+         const uint64_t addr = src + (elts[i] * pitch);
+
+#if DRAW_DBG
+         printf("SPU: fetching = 0x%llx\n", addr);
+#endif
+
+         spu_dcache_fetch_unaligned(& in[idx], addr, bytes_per_entry);
+         idx += quads_per_entry;
+      }
+
+      /* Be nice and zero out any missing vertices.
+       */
+      (void) memset(& in[idx], 0, (8 - idx) * sizeof(qword));
+
+
+      /* Convert all 4 vertices to vectors of float.
+       */
+      (*fetch)(&machine->Inputs[attr].xyzw[0].q, in, fetch_shuffle_data);
+   }
+}
+
+
+void spu_update_vertex_fetch( struct spu_vs_context *draw )
+{
+   draw->vertex_fetch.fetch_func = generic_vertex_fetch;
+}
diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.c b/src/gallium/drivers/cell/spu/spu_vertex_shader.c
new file mode 100644
index 0000000000..fbe5b34d39
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.c
@@ -0,0 +1,244 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  *   Ian Romanick <idr@us.ibm.com>
+  */
+
+#include <spu_mfcio.h>
+
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+#include "util/u_math.h"
+#include "draw/draw_private.h"
+#include "draw/draw_context.h"
+#include "cell/common.h"
+#include "spu_vertex_shader.h"
+#include "spu_exec.h"
+#include "spu_main.h"
+
+
+#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float))
+
+
+#define CLIP_RIGHT_BIT 0x01
+#define CLIP_LEFT_BIT 0x02
+#define CLIP_TOP_BIT 0x04
+#define CLIP_BOTTOM_BIT 0x08
+#define CLIP_FAR_BIT 0x10
+#define CLIP_NEAR_BIT 0x20
+
+
+static INLINE float
+dot4(const float *a, const float *b)
+{
+   return (a[0]*b[0] +
+           a[1]*b[1] +
+           a[2]*b[2] +
+           a[3]*b[3]);
+}
+
+static INLINE unsigned
+compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr)
+{
+   unsigned mask = 0;
+   unsigned i;
+
+   /* Do the hardwired planes first:
+    */
+   if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT;
+   if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT;
+   if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT;
+   if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT;
+   if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT;
+   if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT;
+
+   /* Followed by any remaining ones:
+    */
+   for (i = 6; i < nr; i++) {
+      if (dot4(clip, plane[i]) < 0) 
+         mask |= (1<<i);
+   }
+
+   return mask;
+}
+
+
+/**
+ * Transform vertices with the current vertex program/shader
+ * Up to four vertices can be shaded at a time.
+ * \param vbuffer  the input vertex data
+ * \param elts  indexes of four input vertices
+ * \param count  number of vertices to shade [1..4]
+ * \param vOut  array of pointers to four output vertices
+ */
+static void
+run_vertex_program(struct spu_vs_context *draw,
+                   unsigned elts[4], unsigned count,
+                   const uint64_t *vOut)
+{
+   struct spu_exec_machine *machine = &draw->machine;
+   unsigned int j;
+
+   ALIGN16_DECL(struct spu_exec_vector, inputs, PIPE_MAX_ATTRIBS);
+   ALIGN16_DECL(struct spu_exec_vector, outputs, PIPE_MAX_ATTRIBS);
+   const float *scale = draw->viewport.scale;
+   const float *trans = draw->viewport.translate;
+
+   ASSERT(count <= 4);
+
+   machine->Processor = TGSI_PROCESSOR_VERTEX;
+
+   ASSERT_ALIGN16(draw->constants);
+   machine->Consts = (float (*)[4]) draw->constants;
+
+   machine->Inputs = ALIGN16_ASSIGN(inputs);
+   machine->Outputs = ALIGN16_ASSIGN(outputs);
+
+   spu_vertex_fetch( draw, machine, elts, count );
+
+   /* run shader */
+   spu_exec_machine_run( machine );
+
+
+   /* store machine results */
+   for (j = 0; j < count; j++) {
+      unsigned slot;
+      float x, y, z, w;
+      unsigned char buffer[sizeof(struct vertex_header)
+          + MAX_VERTEX_SIZE] ALIGN16_ATTRIB;
+      struct vertex_header *const tmpOut =
+          (struct vertex_header *) buffer;
+      const unsigned vert_size = ROUNDUP16(sizeof(struct vertex_header)
+                                           + (sizeof(float) * 4 
+                                              * draw->num_vs_outputs));
+
+      mfc_get(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0);
+      wait_on_mask(1 << TAG_VERTEX_BUFFER);
+
+
+      /* Handle attr[0] (position) specially:
+       *
+       * XXX: Computing the clipmask should be done in the vertex
+       * program as a set of DP4 instructions appended to the
+       * user-provided code.
+       */
+      x = tmpOut->clip[0] = machine->Outputs[0].xyzw[0].f[j];
+      y = tmpOut->clip[1] = machine->Outputs[0].xyzw[1].f[j];
+      z = tmpOut->clip[2] = machine->Outputs[0].xyzw[2].f[j];
+      w = tmpOut->clip[3] = machine->Outputs[0].xyzw[3].f[j];
+
+      tmpOut->clipmask = compute_clipmask(tmpOut->clip, draw->plane,
+					   draw->nr_planes);
+      tmpOut->edgeflag = 1;
+
+      /* divide by w */
+      w = 1.0f / w;
+      x *= w;
+      y *= w;
+      z *= w;
+
+      /* Viewport mapping */
+      tmpOut->data[0][0] = x * scale[0] + trans[0];
+      tmpOut->data[0][1] = y * scale[1] + trans[1];
+      tmpOut->data[0][2] = z * scale[2] + trans[2];
+      tmpOut->data[0][3] = w;
+
+      /* Remaining attributes are packed into sequential post-transform
+       * vertex attrib slots.
+       */
+      for (slot = 1; slot < draw->num_vs_outputs; slot++) {
+         tmpOut->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
+         tmpOut->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
+         tmpOut->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
+         tmpOut->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
+      }
+
+      mfc_put(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0);
+   } /* loop over vertices */
+}
+
+
+unsigned char immediates[(sizeof(float) * 4 * TGSI_EXEC_NUM_IMMEDIATES) + 32]
+    ALIGN16_ATTRIB;
+
+
+void
+spu_bind_vertex_shader(struct spu_vs_context *draw,
+		       struct cell_shader_info *vs)
+{
+   const unsigned immediate_addr = vs->immediates;
+   const unsigned immediate_size = 
+       ROUNDUP16((sizeof(float) * 4 * vs->num_immediates)
+		 + (immediate_addr & 0x0f));
+ 
+
+   mfc_get(immediates, immediate_addr & ~0x0f, immediate_size,
+           TAG_VERTEX_BUFFER, 0, 0);
+
+   draw->machine.Instructions = (struct tgsi_full_instruction *)
+       vs->instructions;
+   draw->machine.NumInstructions = vs->num_instructions;
+
+   draw->machine.Declarations = (struct tgsi_full_declaration *)
+       vs->declarations;
+   draw->machine.NumDeclarations = vs->num_declarations;
+
+   draw->num_vs_outputs = vs->num_outputs;
+
+   /* specify the shader to interpret/execute */
+   spu_exec_machine_init(&draw->machine,
+			 PIPE_MAX_SAMPLERS,
+			 NULL /*samplers*/,
+			 PIPE_SHADER_VERTEX);
+
+   wait_on_mask(1 << TAG_VERTEX_BUFFER);
+
+   (void) memcpy(& draw->machine.Imms, &immediates[immediate_addr & 0x0f],
+                 sizeof(float) * 4 * vs->num_immediates);
+}
+
+
+void
+spu_execute_vertex_shader(struct spu_vs_context *draw,
+                          const struct cell_command_vs *vs)
+{
+   unsigned i;
+
+   (void) memcpy(draw->plane, vs->plane, sizeof(float) * 4 * vs->nr_planes);
+   draw->nr_planes = vs->nr_planes;
+   draw->vertex_fetch.nr_attrs = vs->nr_attrs;
+
+   for (i = 0; i < vs->num_elts; i += 4) {
+      const unsigned batch_size = MIN2(vs->num_elts - i, 4);
+
+      run_vertex_program(draw, & vs->elts[i], batch_size, &vs->vOut[i]);
+   }
+}
diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.h b/src/gallium/drivers/cell/spu/spu_vertex_shader.h
new file mode 100644
index 0000000000..4c74f5e74d
--- /dev/null
+++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.h
@@ -0,0 +1,66 @@
+#ifndef SPU_VERTEX_SHADER_H
+#define SPU_VERTEX_SHADER_H
+
+#include "cell/common.h"
+#include "pipe/p_format.h"
+#include "spu_exec.h"
+
+struct spu_vs_context;
+
+typedef void (*spu_full_fetch_func)( struct spu_vs_context *draw,
+				     struct spu_exec_machine *machine,
+				     const unsigned *elts,
+				     unsigned count );
+
+struct spu_vs_context {
+   struct pipe_viewport_state viewport;
+
+   struct {
+      uint64_t src_ptr[PIPE_MAX_ATTRIBS];
+      unsigned pitch[PIPE_MAX_ATTRIBS];
+      unsigned size[PIPE_MAX_ATTRIBS];
+      unsigned code_offset[PIPE_MAX_ATTRIBS];
+      unsigned nr_attrs;
+      boolean dirty;
+
+      spu_full_fetch_func fetch_func;
+      void *code;
+   } vertex_fetch;
+   
+   /* Clip derived state:
+    */
+   float plane[12][4];
+   unsigned nr_planes;
+
+   struct spu_exec_machine machine;
+   const float (*constants)[4];
+
+   unsigned num_vs_outputs;
+};
+
+extern void spu_update_vertex_fetch(struct spu_vs_context *draw);
+
+static INLINE void spu_vertex_fetch(struct spu_vs_context *draw,
+				    struct spu_exec_machine *machine,
+				    const unsigned *elts,
+				    unsigned count)
+{
+   if (draw->vertex_fetch.dirty) {
+      spu_update_vertex_fetch(draw);
+      draw->vertex_fetch.dirty = 0;
+   }
+   
+   (*draw->vertex_fetch.fetch_func)(draw, machine, elts, count);
+}
+
+struct cell_command_vs;
+
+extern void
+spu_bind_vertex_shader(struct spu_vs_context *draw,
+		       struct cell_shader_info *vs);
+
+extern void
+spu_execute_vertex_shader(struct spu_vs_context *draw,
+			  const struct cell_command_vs *vs);
+
+#endif /* SPU_VERTEX_SHADER_H */
diff --git a/src/gallium/drivers/failover/Makefile b/src/gallium/drivers/failover/Makefile
new file mode 100644
index 0000000000..f08b8df07a
--- /dev/null
+++ b/src/gallium/drivers/failover/Makefile
@@ -0,0 +1,14 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = failover
+
+C_SOURCES = \
+	fo_state.c \
+	fo_state_emit.c \
+	fo_context.c 
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/drivers/failover/SConscript b/src/gallium/drivers/failover/SConscript
new file mode 100644
index 0000000000..f8e9b1b491
--- /dev/null
+++ b/src/gallium/drivers/failover/SConscript
@@ -0,0 +1,13 @@
+Import('*')
+
+env = env.Clone()
+
+failover = env.ConvenienceLibrary(
+	target = 'failover',
+	source = [
+		'fo_state.c',
+		'fo_state_emit.c',
+		'fo_context.c',
+	])
+
+Export('failover')
diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c
new file mode 100644
index 0000000000..0742b27b8f
--- /dev/null
+++ b/src/gallium/drivers/failover/fo_context.c
@@ -0,0 +1,159 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_memory.h"
+#include "pipe/p_context.h"
+
+#include "fo_context.h"
+#include "fo_winsys.h"
+
+
+
+static void failover_destroy( struct pipe_context *pipe )
+{
+   struct failover_context *failover = failover_context( pipe );
+
+   free( failover );
+}
+
+
+
+static boolean failover_draw_elements( struct pipe_context *pipe,
+				       struct pipe_buffer *indexBuffer,
+				       unsigned indexSize,
+				       unsigned prim, unsigned start, unsigned count)
+{
+   struct failover_context *failover = failover_context( pipe );
+
+   /* If there has been any statechange since last time, try hardware
+    * rendering again:
+    */
+   if (failover->dirty) {
+      failover->mode = FO_HW;
+   }
+
+   /* Try hardware:
+    */
+   if (failover->mode == FO_HW) {
+      if (!failover->hw->draw_elements( failover->hw, 
+					indexBuffer, 
+					indexSize, 
+					prim, 
+					start, 
+					count )) {
+
+	 failover->hw->flush( failover->hw, ~0, NULL );
+	 failover->mode = FO_SW;
+      }
+   }
+
+   /* Possibly try software:
+    */
+   if (failover->mode == FO_SW) {
+
+      if (failover->dirty) 
+	 failover_state_emit( failover );
+
+      failover->sw->draw_elements( failover->sw, 
+				   indexBuffer, 
+				   indexSize, 
+				   prim, 
+				   start, 
+				   count );
+
+      /* Be ready to switch back to hardware rendering without an
+       * intervening flush.  Unlikely to be much performance impact to
+       * this:
+       */
+      failover->sw->flush( failover->sw, ~0, NULL );
+   }
+
+   return TRUE;
+}
+
+
+static boolean failover_draw_arrays( struct pipe_context *pipe,
+				     unsigned prim, unsigned start, unsigned count)
+{
+   return failover_draw_elements(pipe, NULL, 0, prim, start, count);
+}
+
+
+
+struct pipe_context *failover_create( struct pipe_context *hw,
+				      struct pipe_context *sw )
+{
+   struct failover_context *failover = CALLOC_STRUCT(failover_context);
+   if (failover == NULL)
+      return NULL;
+
+   failover->hw = hw;
+   failover->sw = sw;
+   failover->pipe.winsys = hw->winsys;
+   failover->pipe.screen = hw->screen;
+   failover->pipe.destroy = failover_destroy;
+#if 0
+   failover->pipe.is_format_supported = hw->is_format_supported;
+   failover->pipe.get_name = hw->get_name;
+   failover->pipe.get_vendor = hw->get_vendor;
+   failover->pipe.get_param = hw->get_param;
+   failover->pipe.get_paramf = hw->get_paramf;
+#endif
+
+   failover->pipe.draw_arrays = failover_draw_arrays;
+   failover->pipe.draw_elements = failover_draw_elements;
+   failover->pipe.clear = hw->clear;
+
+   /* No software occlusion fallback (or other optional functionality)
+    * at this point - if the hardware doesn't support it, don't
+    * advertise it to the application.
+    */
+   failover->pipe.begin_query = hw->begin_query;
+   failover->pipe.end_query = hw->end_query;
+
+   failover_init_state_functions( failover );
+
+   failover->pipe.surface_copy = hw->surface_copy;
+   failover->pipe.surface_fill = hw->surface_fill;
+
+#if 0
+   failover->pipe.texture_create = hw->texture_create;
+   failover->pipe.texture_release = hw->texture_release;
+   failover->pipe.get_tex_surface = hw->get_tex_surface;
+   failover->pipe.texture_update = hw->texture_update;
+#endif
+
+   failover->pipe.flush = hw->flush;
+
+   failover->dirty = 0;
+
+   return &failover->pipe;
+}
+
diff --git a/src/gallium/drivers/failover/fo_context.h b/src/gallium/drivers/failover/fo_context.h
new file mode 100644
index 0000000000..9ba86ba866
--- /dev/null
+++ b/src/gallium/drivers/failover/fo_context.h
@@ -0,0 +1,125 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef FO_CONTEXT_H
+#define FO_CONTEXT_H
+
+#include "pipe/p_state.h"
+#include "pipe/p_context.h"
+
+
+
+#define FO_NEW_VIEWPORT        0x1
+#define FO_NEW_RASTERIZER      0x2
+#define FO_NEW_FRAGMENT_SHADER 0x4
+#define FO_NEW_BLEND           0x8
+#define FO_NEW_CLIP            0x10
+#define FO_NEW_SCISSOR         0x20
+#define FO_NEW_STIPPLE         0x40
+#define FO_NEW_FRAMEBUFFER     0x80
+#define FO_NEW_ALPHA_TEST      0x100
+#define FO_NEW_DEPTH_STENCIL   0x200
+#define FO_NEW_SAMPLER         0x400
+#define FO_NEW_TEXTURE         0x800
+#define FO_NEW_VERTEX          0x2000
+#define FO_NEW_VERTEX_SHADER   0x4000
+#define FO_NEW_BLEND_COLOR     0x8000
+#define FO_NEW_CLEAR_COLOR     0x10000
+#define FO_NEW_VERTEX_BUFFER   0x20000
+#define FO_NEW_VERTEX_ELEMENT  0x40000
+
+
+
+#define FO_HW 0
+#define FO_SW 1
+
+struct fo_state {
+   void *sw_state;
+   void *hw_state;
+};
+struct failover_context {
+   struct pipe_context pipe;  /**< base class */
+
+
+   /* The most recent drawing state as set by the driver:
+    */
+   const struct fo_state     *blend;
+   const struct fo_state     *sampler[PIPE_MAX_SAMPLERS];
+   const struct fo_state     *depth_stencil;
+   const struct fo_state     *rasterizer;
+   const struct fo_state     *fragment_shader;
+   const struct fo_state     *vertex_shader;
+
+   struct pipe_blend_color blend_color;
+   struct pipe_clip_state clip;
+   struct pipe_framebuffer_state framebuffer;
+   struct pipe_poly_stipple poly_stipple;
+   struct pipe_scissor_state scissor;
+   struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
+   struct pipe_viewport_state viewport;
+   struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS];
+   struct pipe_vertex_element vertex_elements[PIPE_MAX_ATTRIBS];
+
+   uint num_vertex_buffers;
+   uint num_vertex_elements;
+
+   void *sw_sampler_state[PIPE_MAX_SAMPLERS];
+   void *hw_sampler_state[PIPE_MAX_SAMPLERS];
+
+   unsigned dirty;
+
+   unsigned num_samplers;
+   unsigned num_textures;
+
+   unsigned mode;
+   struct pipe_context *hw;
+   struct pipe_context *sw;
+};
+
+
+
+void failover_init_state_functions( struct failover_context *failover );
+void failover_state_emit( struct failover_context *failover );
+
+static INLINE struct failover_context *
+failover_context( struct pipe_context *pipe )
+{
+   return (struct failover_context *)pipe;
+}
+
+/* Internal functions
+ */
+void
+failover_set_constant_buffer(struct pipe_context *pipe,
+                             uint shader, uint index,
+                             const struct pipe_constant_buffer *buf);
+
+
+#endif /* FO_CONTEXT_H */
diff --git a/src/gallium/drivers/failover/fo_state.c b/src/gallium/drivers/failover/fo_state.c
new file mode 100644
index 0000000000..6a79706632
--- /dev/null
+++ b/src/gallium/drivers/failover/fo_state.c
@@ -0,0 +1,483 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "pipe/p_inlines.h"
+
+#include "fo_context.h"
+
+
+/* This looks like a lot of work at the moment - we're keeping a
+ * duplicate copy of the state up-to-date.  
+ *
+ * This can change in two ways:
+ * - With constant state objects we would only need to save a pointer,
+ *     not the whole object.
+ * - By adding a callback in the state tracker to re-emit state.  The
+ *     state tracker knows the current state already and can re-emit it 
+ *     without additional complexity.
+ *
+ * This works as a proof-of-concept, but a final version will have
+ * lower overheads.
+ */
+
+
+
+static void *
+failover_create_blend_state( struct pipe_context *pipe,
+                             const struct pipe_blend_state *blend )
+{
+   struct fo_state *state = malloc(sizeof(struct fo_state));
+   struct failover_context *failover = failover_context(pipe);
+
+   state->sw_state = failover->sw->create_blend_state(failover->sw, blend);
+   state->hw_state = failover->hw->create_blend_state(failover->hw, blend);
+
+   return state;
+}
+
+static void
+failover_bind_blend_state( struct pipe_context *pipe,
+                           void *blend )
+{
+   struct failover_context *failover = failover_context(pipe);
+   struct fo_state *state = (struct fo_state *)blend;
+   failover->blend = state;
+   failover->dirty |= FO_NEW_BLEND;
+   failover->sw->bind_blend_state( failover->sw, state->sw_state );
+   failover->hw->bind_blend_state( failover->hw, state->hw_state );
+}
+
+static void
+failover_delete_blend_state( struct pipe_context *pipe,
+                             void *blend )
+{
+   struct fo_state *state = (struct fo_state*)blend;
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->sw->delete_blend_state(failover->sw, state->sw_state);
+   failover->hw->delete_blend_state(failover->hw, state->hw_state);
+   state->sw_state = 0;
+   state->hw_state = 0;
+   free(state);
+}
+
+static void
+failover_set_blend_color( struct pipe_context *pipe,
+			  const struct pipe_blend_color *blend_color )
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->blend_color = *blend_color;
+   failover->dirty |= FO_NEW_BLEND_COLOR;
+   failover->sw->set_blend_color( failover->sw, blend_color );
+   failover->hw->set_blend_color( failover->hw, blend_color );
+}
+
+static void 
+failover_set_clip_state( struct pipe_context *pipe,
+			 const struct pipe_clip_state *clip )
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->clip = *clip;
+   failover->dirty |= FO_NEW_CLIP;
+   failover->sw->set_clip_state( failover->sw, clip );
+   failover->hw->set_clip_state( failover->hw, clip );
+}
+
+
+static void *
+failover_create_depth_stencil_state(struct pipe_context *pipe,
+                              const struct pipe_depth_stencil_alpha_state *templ)
+{
+   struct fo_state *state = malloc(sizeof(struct fo_state));
+   struct failover_context *failover = failover_context(pipe);
+
+   state->sw_state = failover->sw->create_depth_stencil_alpha_state(failover->sw, templ);
+   state->hw_state = failover->hw->create_depth_stencil_alpha_state(failover->hw, templ);
+
+   return state;
+}
+
+static void
+failover_bind_depth_stencil_state(struct pipe_context *pipe,
+                                  void *depth_stencil)
+{
+   struct failover_context *failover = failover_context(pipe);
+   struct fo_state *state = (struct fo_state *)depth_stencil;
+   failover->depth_stencil = state;
+   failover->dirty |= FO_NEW_DEPTH_STENCIL;
+   failover->sw->bind_depth_stencil_alpha_state(failover->sw, state->sw_state);
+   failover->hw->bind_depth_stencil_alpha_state(failover->hw, state->hw_state);
+}
+
+static void
+failover_delete_depth_stencil_state(struct pipe_context *pipe,
+                                    void *ds)
+{
+   struct fo_state *state = (struct fo_state*)ds;
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->sw->delete_depth_stencil_alpha_state(failover->sw, state->sw_state);
+   failover->hw->delete_depth_stencil_alpha_state(failover->hw, state->hw_state);
+   state->sw_state = 0;
+   state->hw_state = 0;
+   free(state);
+}
+
+static void
+failover_set_framebuffer_state(struct pipe_context *pipe,
+			       const struct pipe_framebuffer_state *framebuffer)
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->framebuffer = *framebuffer;
+   failover->dirty |= FO_NEW_FRAMEBUFFER;
+   failover->sw->set_framebuffer_state( failover->sw, framebuffer );
+   failover->hw->set_framebuffer_state( failover->hw, framebuffer );
+}
+
+
+static void *
+failover_create_fs_state(struct pipe_context *pipe,
+                         const struct pipe_shader_state *templ)
+{
+   struct fo_state *state = malloc(sizeof(struct fo_state));
+   struct failover_context *failover = failover_context(pipe);
+
+   state->sw_state = failover->sw->create_fs_state(failover->sw, templ);
+   state->hw_state = failover->hw->create_fs_state(failover->hw, templ);
+
+   return state;
+}
+
+static void
+failover_bind_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct failover_context *failover = failover_context(pipe);
+   struct fo_state *state = (struct fo_state*)fs;
+   failover->fragment_shader = state;
+   failover->dirty |= FO_NEW_FRAGMENT_SHADER;
+   failover->sw->bind_fs_state(failover->sw, state->sw_state);
+   failover->hw->bind_fs_state(failover->hw, state->hw_state);
+}
+
+static void
+failover_delete_fs_state(struct pipe_context *pipe,
+                         void *fs)
+{
+   struct fo_state *state = (struct fo_state*)fs;
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->sw->delete_fs_state(failover->sw, state->sw_state);
+   failover->hw->delete_fs_state(failover->hw, state->hw_state);
+   state->sw_state = 0;
+   state->hw_state = 0;
+   free(state);
+}
+
+static void *
+failover_create_vs_state(struct pipe_context *pipe,
+                         const struct pipe_shader_state *templ)
+{
+   struct fo_state *state = malloc(sizeof(struct fo_state));
+   struct failover_context *failover = failover_context(pipe);
+
+   state->sw_state = failover->sw->create_vs_state(failover->sw, templ);
+   state->hw_state = failover->hw->create_vs_state(failover->hw, templ);
+
+   return state;
+}
+
+static void
+failover_bind_vs_state(struct pipe_context *pipe,
+                       void *vs)
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   struct fo_state *state = (struct fo_state*)vs;
+   failover->vertex_shader = state;
+   failover->dirty |= FO_NEW_VERTEX_SHADER;
+   failover->sw->bind_vs_state(failover->sw, state->sw_state);
+   failover->hw->bind_vs_state(failover->hw, state->hw_state);
+}
+
+static void
+failover_delete_vs_state(struct pipe_context *pipe,
+                         void *vs)
+{
+   struct fo_state *state = (struct fo_state*)vs;
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->sw->delete_vs_state(failover->sw, state->sw_state);
+   failover->hw->delete_vs_state(failover->hw, state->hw_state);
+   state->sw_state = 0;
+   state->hw_state = 0;
+   free(state);
+}
+
+static void 
+failover_set_polygon_stipple( struct pipe_context *pipe,
+			      const struct pipe_poly_stipple *stipple )
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->poly_stipple = *stipple;
+   failover->dirty |= FO_NEW_STIPPLE;
+   failover->sw->set_polygon_stipple( failover->sw, stipple );
+   failover->hw->set_polygon_stipple( failover->hw, stipple );
+}
+
+
+static void *
+failover_create_rasterizer_state(struct pipe_context *pipe,
+                                 const struct pipe_rasterizer_state *templ)
+{
+   struct fo_state *state = malloc(sizeof(struct fo_state));
+   struct failover_context *failover = failover_context(pipe);
+
+   state->sw_state = failover->sw->create_rasterizer_state(failover->sw, templ);
+   state->hw_state = failover->hw->create_rasterizer_state(failover->hw, templ);
+
+   return state;
+}
+
+static void
+failover_bind_rasterizer_state(struct pipe_context *pipe,
+                               void *raster)
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   struct fo_state *state = (struct fo_state*)raster;
+   failover->rasterizer = state;
+   failover->dirty |= FO_NEW_RASTERIZER;
+   failover->sw->bind_rasterizer_state(failover->sw, state->sw_state);
+   failover->hw->bind_rasterizer_state(failover->hw, state->hw_state);
+}
+
+static void
+failover_delete_rasterizer_state(struct pipe_context *pipe,
+                                 void *raster)
+{
+   struct fo_state *state = (struct fo_state*)raster;
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->sw->delete_rasterizer_state(failover->sw, state->sw_state);
+   failover->hw->delete_rasterizer_state(failover->hw, state->hw_state);
+   state->sw_state = 0;
+   state->hw_state = 0;
+   free(state);
+}
+
+
+static void 
+failover_set_scissor_state( struct pipe_context *pipe,
+                                 const struct pipe_scissor_state *scissor )
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->scissor = *scissor;
+   failover->dirty |= FO_NEW_SCISSOR;
+   failover->sw->set_scissor_state( failover->sw, scissor );
+   failover->hw->set_scissor_state( failover->hw, scissor );
+}
+
+
+static void *
+failover_create_sampler_state(struct pipe_context *pipe,
+                              const struct pipe_sampler_state *templ)
+{
+   struct fo_state *state = malloc(sizeof(struct fo_state));
+   struct failover_context *failover = failover_context(pipe);
+
+   state->sw_state = failover->sw->create_sampler_state(failover->sw, templ);
+   state->hw_state = failover->hw->create_sampler_state(failover->hw, templ);
+
+   return state;
+}
+
+static void
+failover_bind_sampler_states(struct pipe_context *pipe,
+                             unsigned num, void **sampler)
+{
+   struct failover_context *failover = failover_context(pipe);
+   struct fo_state *state = (struct fo_state*)sampler;
+   uint i;
+   assert(num <= PIPE_MAX_SAMPLERS);
+   /* Check for no-op */
+   if (num == failover->num_samplers &&
+       !memcmp(failover->sampler, sampler, num * sizeof(void *)))
+      return;
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      failover->sw_sampler_state[i] = i < num ? state[i].sw_state : NULL;
+      failover->hw_sampler_state[i] = i < num ? state[i].hw_state : NULL;
+   }
+   failover->dirty |= FO_NEW_SAMPLER;
+   failover->num_samplers = num;
+   failover->sw->bind_sampler_states(failover->sw, num,
+                                     failover->sw_sampler_state);
+   failover->hw->bind_sampler_states(failover->hw, num,
+                                     failover->hw_sampler_state);
+}
+
+static void
+failover_delete_sampler_state(struct pipe_context *pipe, void *sampler)
+{
+   struct fo_state *state = (struct fo_state*)sampler;
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->sw->delete_sampler_state(failover->sw, state->sw_state);
+   failover->hw->delete_sampler_state(failover->hw, state->hw_state);
+   state->sw_state = 0;
+   state->hw_state = 0;
+   free(state);
+}
+
+
+static void
+failover_set_sampler_textures(struct pipe_context *pipe,
+                              unsigned num,
+                              struct pipe_texture **texture)
+{
+   struct failover_context *failover = failover_context(pipe);
+   uint i;
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == failover->num_textures &&
+       !memcmp(failover->texture, texture, num * sizeof(struct pipe_texture *)))
+      return;
+   for (i = 0; i < num; i++)
+      pipe_texture_reference((struct pipe_texture **) &failover->texture[i],
+                             texture[i]);
+   for (i = num; i < failover->num_textures; i++)
+      pipe_texture_reference((struct pipe_texture **) &failover->texture[i],
+                             NULL);
+   failover->dirty |= FO_NEW_TEXTURE;
+   failover->num_textures = num;
+   failover->sw->set_sampler_textures( failover->sw, num, texture );
+   failover->hw->set_sampler_textures( failover->hw, num, texture );
+}
+
+
+static void 
+failover_set_viewport_state( struct pipe_context *pipe,
+			     const struct pipe_viewport_state *viewport )
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   failover->viewport = *viewport; 
+   failover->dirty |= FO_NEW_VIEWPORT;
+   failover->sw->set_viewport_state( failover->sw, viewport );
+   failover->hw->set_viewport_state( failover->hw, viewport );
+}
+
+
+static void
+failover_set_vertex_buffers(struct pipe_context *pipe,
+                            unsigned count,
+                            const struct pipe_vertex_buffer *vertex_buffers)
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   memcpy(failover->vertex_buffers, vertex_buffers,
+          count * sizeof(vertex_buffers[0]));
+   failover->dirty |= FO_NEW_VERTEX_BUFFER;
+   failover->num_vertex_buffers = count;
+   failover->sw->set_vertex_buffers( failover->sw, count, vertex_buffers );
+   failover->hw->set_vertex_buffers( failover->hw, count, vertex_buffers );
+}
+
+
+static void
+failover_set_vertex_elements(struct pipe_context *pipe,
+                             unsigned count,
+                             const struct pipe_vertex_element *vertex_elements)
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   memcpy(failover->vertex_elements, vertex_elements,
+          count * sizeof(vertex_elements[0]));
+
+   failover->dirty |= FO_NEW_VERTEX_ELEMENT;
+   failover->num_vertex_elements = count;
+   failover->sw->set_vertex_elements( failover->sw, count, vertex_elements );
+   failover->hw->set_vertex_elements( failover->hw, count, vertex_elements );
+}
+
+void
+failover_set_constant_buffer(struct pipe_context *pipe,
+                             uint shader, uint index,
+                             const struct pipe_constant_buffer *buf)
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   assert(shader < PIPE_SHADER_TYPES);
+   assert(index == 0);
+
+   failover->sw->set_constant_buffer(failover->sw, shader, index, buf);
+   failover->hw->set_constant_buffer(failover->hw, shader, index, buf);
+}
+
+
+void
+failover_init_state_functions( struct failover_context *failover )
+{
+   failover->pipe.create_blend_state = failover_create_blend_state;
+   failover->pipe.bind_blend_state   = failover_bind_blend_state;
+   failover->pipe.delete_blend_state = failover_delete_blend_state;
+   failover->pipe.create_sampler_state = failover_create_sampler_state;
+   failover->pipe.bind_sampler_states  = failover_bind_sampler_states;
+   failover->pipe.delete_sampler_state = failover_delete_sampler_state;
+   failover->pipe.create_depth_stencil_alpha_state = failover_create_depth_stencil_state;
+   failover->pipe.bind_depth_stencil_alpha_state   = failover_bind_depth_stencil_state;
+   failover->pipe.delete_depth_stencil_alpha_state = failover_delete_depth_stencil_state;
+   failover->pipe.create_rasterizer_state = failover_create_rasterizer_state;
+   failover->pipe.bind_rasterizer_state = failover_bind_rasterizer_state;
+   failover->pipe.delete_rasterizer_state = failover_delete_rasterizer_state;
+   failover->pipe.create_fs_state = failover_create_fs_state;
+   failover->pipe.bind_fs_state   = failover_bind_fs_state;
+   failover->pipe.delete_fs_state = failover_delete_fs_state;
+   failover->pipe.create_vs_state = failover_create_vs_state;
+   failover->pipe.bind_vs_state   = failover_bind_vs_state;
+   failover->pipe.delete_vs_state = failover_delete_vs_state;
+
+   failover->pipe.set_blend_color = failover_set_blend_color;
+   failover->pipe.set_clip_state = failover_set_clip_state;
+   failover->pipe.set_framebuffer_state = failover_set_framebuffer_state;
+   failover->pipe.set_polygon_stipple = failover_set_polygon_stipple;
+   failover->pipe.set_scissor_state = failover_set_scissor_state;
+   failover->pipe.set_sampler_textures = failover_set_sampler_textures;
+   failover->pipe.set_viewport_state = failover_set_viewport_state;
+   failover->pipe.set_vertex_buffers = failover_set_vertex_buffers;
+   failover->pipe.set_vertex_elements = failover_set_vertex_elements;
+   failover->pipe.set_constant_buffer = failover_set_constant_buffer;
+}
diff --git a/src/gallium/drivers/failover/fo_state_emit.c b/src/gallium/drivers/failover/fo_state_emit.c
new file mode 100644
index 0000000000..bd4fce9d20
--- /dev/null
+++ b/src/gallium/drivers/failover/fo_state_emit.c
@@ -0,0 +1,117 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "fo_context.h"
+
+/* This looks like a lot of work at the moment - we're keeping a
+ * duplicate copy of the state up-to-date.  
+ *
+ * This can change in two ways:
+ * - With constant state objects we would only need to save a pointer,
+ *     not the whole object.
+ * - By adding a callback in the state tracker to re-emit state.  The
+ *     state tracker knows the current state already and can re-emit it 
+ *     without additional complexity.
+ *
+ * This works as a proof-of-concept, but a final version will have
+ * lower overheads.
+ */
+
+
+/* Bring the software pipe uptodate with current state.
+ * 
+ * With constant state objects we would probably just send all state
+ * to both rasterizers all the time???
+ */
+void
+failover_state_emit( struct failover_context *failover )
+{
+   if (failover->dirty & FO_NEW_BLEND)
+      failover->sw->bind_blend_state( failover->sw,
+                                      failover->blend->sw_state );
+
+   if (failover->dirty & FO_NEW_BLEND_COLOR)
+      failover->sw->set_blend_color( failover->sw, &failover->blend_color );
+
+   if (failover->dirty & FO_NEW_CLIP)
+      failover->sw->set_clip_state( failover->sw, &failover->clip );
+
+   if (failover->dirty & FO_NEW_DEPTH_STENCIL)
+      failover->sw->bind_depth_stencil_alpha_state( failover->sw,
+						    failover->depth_stencil->sw_state );
+
+   if (failover->dirty & FO_NEW_FRAMEBUFFER)
+      failover->sw->set_framebuffer_state( failover->sw, &failover->framebuffer );
+
+   if (failover->dirty & FO_NEW_FRAGMENT_SHADER)
+      failover->sw->bind_fs_state( failover->sw,
+                                   failover->fragment_shader->sw_state );
+
+   if (failover->dirty & FO_NEW_VERTEX_SHADER)
+      failover->sw->bind_vs_state( failover->sw,
+                                   failover->vertex_shader->sw_state );
+
+   if (failover->dirty & FO_NEW_STIPPLE)
+      failover->sw->set_polygon_stipple( failover->sw, &failover->poly_stipple );
+
+   if (failover->dirty & FO_NEW_RASTERIZER)
+      failover->sw->bind_rasterizer_state( failover->sw,
+                                           failover->rasterizer->sw_state );
+
+   if (failover->dirty & FO_NEW_SCISSOR)
+      failover->sw->set_scissor_state( failover->sw, &failover->scissor );
+
+   if (failover->dirty & FO_NEW_VIEWPORT)
+      failover->sw->set_viewport_state( failover->sw, &failover->viewport );
+
+   if (failover->dirty & FO_NEW_SAMPLER) {
+      failover->sw->bind_sampler_states( failover->sw, failover->num_samplers,
+                                         failover->sw_sampler_state );
+   }
+
+   if (failover->dirty & FO_NEW_TEXTURE) {
+      failover->sw->set_sampler_textures( failover->sw, failover->num_textures, 
+                                          failover->texture );
+   }
+
+   if (failover->dirty & FO_NEW_VERTEX_BUFFER) {
+      failover->sw->set_vertex_buffers( failover->sw,
+                                        failover->num_vertex_buffers,
+                                        failover->vertex_buffers );
+   }
+
+   if (failover->dirty & FO_NEW_VERTEX_ELEMENT) {
+      failover->sw->set_vertex_elements( failover->sw,
+                                         failover->num_vertex_elements,
+                                         failover->vertex_elements );
+   }
+
+   failover->dirty = 0;
+}
diff --git a/src/gallium/drivers/failover/fo_winsys.h b/src/gallium/drivers/failover/fo_winsys.h
new file mode 100644
index 0000000000..a8ce997a1f
--- /dev/null
+++ b/src/gallium/drivers/failover/fo_winsys.h
@@ -0,0 +1,45 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef FO_WINSYS_H
+#define FO_WINSYS_H
+
+
+/* This is the interface that failover requires any window system
+ * hosting it to implement.  This is the only include file in failover
+ * which is public.
+ */
+
+
+struct pipe_context;
+
+
+struct pipe_context *failover_create( struct pipe_context *hw,
+				      struct pipe_context *sw );
+
+
+#endif /* FO_WINSYS_H */
diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile
new file mode 100644
index 0000000000..41a61a0020
--- /dev/null
+++ b/src/gallium/drivers/i915simple/Makefile
@@ -0,0 +1,31 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = i915simple
+
+C_SOURCES = \
+	i915_blit.c \
+	i915_clear.c \
+	i915_flush.c \
+	i915_context.c \
+	i915_context.c \
+	i915_debug.c \
+	i915_debug_fp.c \
+	i915_state.c \
+	i915_state_immediate.c \
+	i915_state_dynamic.c \
+	i915_state_derived.c \
+	i915_state_emit.c \
+	i915_state_sampler.c \
+	i915_screen.c \
+	i915_prim_emit.c \
+	i915_prim_vbuf.c \
+	i915_texture.c \
+	i915_fpc_emit.c \
+	i915_fpc_translate.c \
+	i915_surface.c 
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/drivers/i915simple/SConscript b/src/gallium/drivers/i915simple/SConscript
new file mode 100644
index 0000000000..2366e1247f
--- /dev/null
+++ b/src/gallium/drivers/i915simple/SConscript
@@ -0,0 +1,29 @@
+Import('*')
+
+env = env.Clone()
+
+i915simple = env.ConvenienceLibrary(
+	target = 'i915simple',
+	source = [
+		'i915_blit.c',
+		'i915_clear.c',
+		'i915_context.c',
+		'i915_debug.c',
+		'i915_debug_fp.c',
+		'i915_flush.c',
+		'i915_fpc_emit.c',
+		'i915_fpc_translate.c',
+		'i915_prim_emit.c',
+		'i915_prim_vbuf.c',
+		'i915_screen.c',
+		'i915_state.c',
+		'i915_state_derived.c',
+		'i915_state_dynamic.c',
+		'i915_state_emit.c',
+		'i915_state_immediate.c',
+		'i915_state_sampler.c',
+		'i915_surface.c',
+		'i915_texture.c',
+	])
+
+Export('i915simple')
diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h
new file mode 100644
index 0000000000..a433cf054d
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_batch.h
@@ -0,0 +1,116 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef I915_BATCH_H
+#define I915_BATCH_H
+
+#include "i915_winsys.h"
+
+struct i915_batchbuffer
+{
+   struct pipe_buffer *buffer;
+   struct i915_winsys *winsys;
+
+   unsigned char *map;
+   unsigned char *ptr;
+
+   size_t size;
+   size_t actual_size;
+
+   size_t relocs;
+   size_t max_relocs;
+};
+
+static INLINE boolean
+i915_batchbuffer_check( struct i915_batchbuffer *batch,
+			size_t dwords,
+			size_t relocs )
+{
+   /** TODO JB: Check relocs */
+   return dwords * 4 <= batch->size - (batch->ptr - batch->map);
+}
+
+static INLINE size_t
+i915_batchbuffer_space( struct i915_batchbuffer *batch )
+{
+   return batch->size - (batch->ptr - batch->map);
+}
+
+static INLINE void
+i915_batchbuffer_dword( struct i915_batchbuffer *batch,
+			unsigned dword )
+{
+   if (i915_batchbuffer_space(batch) < 4)
+      return;
+
+   *(unsigned *)batch->ptr = dword;
+   batch->ptr += 4;
+}
+
+static INLINE void
+i915_batchbuffer_write( struct i915_batchbuffer *batch,
+			void *data,
+			size_t size )
+{
+   if (i915_batchbuffer_space(batch) < size)
+      return;
+
+   memcpy(data, batch->ptr, size);
+   batch->ptr += size;
+}
+
+static INLINE void
+i915_batchbuffer_reloc( struct i915_batchbuffer *batch,
+			struct pipe_buffer *buffer,
+			size_t flags,
+			size_t offset )
+{
+   batch->winsys->batch_reloc( batch->winsys, buffer, flags, offset );
+}
+
+static INLINE void
+i915_batchbuffer_flush( struct i915_batchbuffer *batch,
+			struct pipe_fence_handle **fence )
+{
+   batch->winsys->batch_flush( batch->winsys, fence );
+}
+
+#define BEGIN_BATCH( dwords, relocs ) \
+   (i915_batchbuffer_check( i915->batch, dwords, relocs ))
+
+#define OUT_BATCH( dword ) \
+   i915_batchbuffer_dword( i915->batch, dword )
+
+#define OUT_RELOC( buf, flags, delta ) \
+   i915_batchbuffer_reloc( i915->batch, buf, flags, delta )
+
+#define FLUSH_BATCH(fence) do {				\
+   i915->winsys->batch_flush( i915->winsys, fence );	\
+   i915->hardware_dirty = ~0;				\
+} while (0)
+
+#endif
diff --git a/src/gallium/drivers/i915simple/i915_blit.c b/src/gallium/drivers/i915simple/i915_blit.c
new file mode 100644
index 0000000000..448a4708ce
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_blit.c
@@ -0,0 +1,159 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "i915_context.h"
+#include "i915_winsys.h"
+#include "i915_blit.h"
+#include "i915_reg.h"
+#include "i915_batch.h"
+#include "i915_debug.h"
+
+#define FILE_DEBUG_FLAG DEBUG_BLIT
+
+void
+i915_fill_blit(struct i915_context *i915,
+	       unsigned cpp,
+	       unsigned short dst_pitch,
+	       struct pipe_buffer *dst_buffer,
+	       unsigned dst_offset,
+	       short x, short y, 
+	       short w, short h, 
+	       unsigned color)
+{
+   unsigned BR13, CMD;
+
+
+   I915_DBG(i915,
+       "%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
+       __FUNCTION__,
+       dst_buffer, dst_pitch, dst_offset, x, y, w, h);
+
+   switch (cpp) {
+   case 1:
+   case 2:
+   case 3:
+      BR13 = (((int) dst_pitch) & 0xffff) |
+	 (0xF0 << 16) | (1 << 24);
+      CMD = XY_COLOR_BLT_CMD;
+      break;
+   case 4:
+      BR13 = (((int) dst_pitch) & 0xffff) |
+	 (0xF0 << 16) | (1 << 24) | (1 << 25);
+      CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA |
+             XY_COLOR_BLT_WRITE_RGB);
+      break;
+   default:
+      return;
+   }
+
+   if (!BEGIN_BATCH(6, 1)) {
+      FLUSH_BATCH(NULL);
+      assert(BEGIN_BATCH(6, 1));
+   }
+   OUT_BATCH(CMD);
+   OUT_BATCH(BR13);
+   OUT_BATCH((y << 16) | x);
+   OUT_BATCH(((y + h) << 16) | (x + w));
+   OUT_RELOC( dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset);
+   OUT_BATCH(color);
+   FLUSH_BATCH(NULL);
+}
+
+
+void
+i915_copy_blit( struct i915_context *i915,
+                  unsigned do_flip,
+                  unsigned cpp,
+                  unsigned short src_pitch,
+                  struct pipe_buffer *src_buffer,
+                  unsigned src_offset,
+                  unsigned short dst_pitch,
+                  struct pipe_buffer *dst_buffer,
+                  unsigned dst_offset,
+                  short src_x, short src_y,
+                  short dst_x, short dst_y, 
+		  short w, short h )
+{
+   unsigned CMD, BR13;
+   int dst_y2 = dst_y + h;
+   int dst_x2 = dst_x + w;
+
+
+   I915_DBG(i915,
+       "%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
+       __FUNCTION__,
+       src_buffer, src_pitch, src_offset, src_x, src_y,
+       dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h);
+
+   switch (cpp) {
+   case 1:
+   case 2:
+   case 3:
+      BR13 = (((int) dst_pitch) & 0xffff) |
+	 (0xCC << 16) | (1 << 24);
+      CMD = XY_SRC_COPY_BLT_CMD;
+      break;
+   case 4:
+      BR13 = (((int) dst_pitch) & 0xffff) |
+	 (0xCC << 16) | (1 << 24) | (1 << 25);
+      CMD =
+         (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA |
+          XY_SRC_COPY_BLT_WRITE_RGB);
+      break;
+   default:
+      return;
+   }
+
+   if (dst_y2 < dst_y || 
+       dst_x2 < dst_x) {
+      return;
+   }
+
+   /* Hardware can handle negative pitches but loses the ability to do
+    * proper overlapping blits in that case.  We don't really have a
+    * need for either at this stage.
+    */
+   assert (dst_pitch > 0 && src_pitch > 0);
+
+
+   if (!BEGIN_BATCH(8, 2)) {
+      FLUSH_BATCH(NULL);
+      assert(BEGIN_BATCH(8, 2));
+   }
+   OUT_BATCH(CMD);
+   OUT_BATCH(BR13);
+   OUT_BATCH((dst_y << 16) | dst_x);
+   OUT_BATCH((dst_y2 << 16) | dst_x2);
+   OUT_RELOC(dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset);
+   OUT_BATCH((src_y << 16) | src_x);
+   OUT_BATCH(((int) src_pitch & 0xffff));
+   OUT_RELOC(src_buffer, I915_BUFFER_ACCESS_READ, src_offset);
+   FLUSH_BATCH(NULL);
+}
+
+
diff --git a/src/gallium/drivers/i915simple/i915_blit.h b/src/gallium/drivers/i915simple/i915_blit.h
new file mode 100644
index 0000000000..0bb3453861
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_blit.h
@@ -0,0 +1,55 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef I915_BLIT_H
+#define I915_BLIT_H
+
+#include "i915_context.h"
+
+extern void i915_copy_blit(struct i915_context *i915,
+                           unsigned do_flip,
+			   unsigned cpp,
+			   unsigned short src_pitch,
+			   struct pipe_buffer *src_buffer,
+			   unsigned src_offset,
+			   unsigned short dst_pitch,
+			   struct pipe_buffer *dst_buffer,
+			   unsigned dst_offset,
+			   short srcx, short srcy,
+			   short dstx, short dsty,
+			   short w, short h );
+
+extern void i915_fill_blit(struct i915_context *i915,
+			   unsigned cpp,
+			   unsigned short dst_pitch,
+			   struct pipe_buffer *dst_buffer,
+			   unsigned dst_offset,
+			   short x, short y,
+			   short w, short h, unsigned color);
+
+
+#endif
diff --git a/src/gallium/drivers/i915simple/i915_clear.c b/src/gallium/drivers/i915simple/i915_clear.c
new file mode 100644
index 0000000000..8a2d3ca43f
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_clear.c
@@ -0,0 +1,48 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Brian Paul
+ */
+
+
+#include "pipe/p_defines.h"
+#include "i915_context.h"
+#include "i915_state.h"
+
+
+/**
+ * Clear the given surface to the specified value.
+ * No masking, no scissor (clear entire buffer).
+ */
+void
+i915_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+	   unsigned clearValue)
+{
+   pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue);
+   ps->status = PIPE_SURFACE_STATUS_DEFINED;
+}
diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c
new file mode 100644
index 0000000000..3e3a596884
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_context.c
@@ -0,0 +1,193 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "i915_context.h"
+#include "i915_winsys.h"
+#include "i915_state.h"
+#include "i915_batch.h"
+#include "i915_texture.h"
+#include "i915_reg.h"
+
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+#include "pipe/p_screen.h"
+
+
+static void i915_destroy( struct pipe_context *pipe )
+{
+   struct i915_context *i915 = i915_context( pipe );
+
+   draw_destroy( i915->draw );
+   
+   if(i915->winsys->destroy)
+      i915->winsys->destroy(i915->winsys);
+
+   FREE( i915 );
+}
+
+
+static boolean
+i915_draw_range_elements(struct pipe_context *pipe,
+			     struct pipe_buffer *indexBuffer,
+			     unsigned indexSize,
+			     unsigned min_index,
+			     unsigned max_index,
+			     unsigned prim, unsigned start, unsigned count)
+{
+   struct i915_context *i915 = i915_context( pipe );
+   struct draw_context *draw = i915->draw;
+   unsigned i;
+
+   if (i915->dirty)
+      i915_update_derived( i915 );
+
+   /*
+    * Map vertex buffers
+    */
+   for (i = 0; i < i915->num_vertex_buffers; i++) {
+      void *buf
+         = pipe_buffer_map(pipe->screen,
+                                    i915->vertex_buffer[i].buffer,
+                                    PIPE_BUFFER_USAGE_CPU_READ);
+      draw_set_mapped_vertex_buffer(draw, i, buf);
+   }
+   /* Map index buffer, if present */
+   if (indexBuffer) {
+      void *mapped_indexes
+         = pipe_buffer_map(pipe->screen, indexBuffer,
+                                    PIPE_BUFFER_USAGE_CPU_READ);
+      draw_set_mapped_element_buffer_range(draw, indexSize,
+					   min_index,
+					   max_index,
+					   mapped_indexes);
+   }
+   else {
+      /* no index/element buffer */
+      draw_set_mapped_element_buffer(draw, 0, NULL);
+   }
+
+
+   draw_set_mapped_constant_buffer(draw,
+                                   i915->current.constants[PIPE_SHADER_VERTEX],
+                                   ( i915->current.num_user_constants[PIPE_SHADER_VERTEX] * 
+                                     4 * sizeof(float) ));
+
+   /* draw! */
+   draw_arrays(i915->draw, prim, start, count);
+
+   /*
+    * unmap vertex/index buffers
+    */
+   for (i = 0; i < i915->num_vertex_buffers; i++) {
+      pipe_buffer_unmap(pipe->screen, i915->vertex_buffer[i].buffer);
+      draw_set_mapped_vertex_buffer(draw, i, NULL);
+   }
+   if (indexBuffer) {
+      pipe_buffer_unmap(pipe->screen, indexBuffer);
+      draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL);
+   }
+
+   return TRUE;
+}
+
+static boolean
+i915_draw_elements( struct pipe_context *pipe,
+                    struct pipe_buffer *indexBuffer,
+                    unsigned indexSize,
+                    unsigned prim, unsigned start, unsigned count)
+{
+   return i915_draw_range_elements( pipe, indexBuffer,
+					indexSize,
+					0, 0xffffffff,
+					prim, start, count );
+}
+
+static boolean i915_draw_arrays( struct pipe_context *pipe,
+				 unsigned prim, unsigned start, unsigned count)
+{
+   return i915_draw_elements(pipe, NULL, 0, prim, start, count);
+}
+
+
+
+struct pipe_context *i915_create_context( struct pipe_screen *screen,
+                                          struct pipe_winsys *pipe_winsys,
+                                          struct i915_winsys *i915_winsys )
+{
+   struct i915_context *i915;
+
+   i915 = CALLOC_STRUCT(i915_context);
+   if (i915 == NULL)
+      return NULL;
+
+   i915->winsys = i915_winsys;
+   i915->pipe.winsys = pipe_winsys;
+   i915->pipe.screen = screen;
+
+   i915->pipe.destroy = i915_destroy;
+
+   i915->pipe.clear = i915_clear;
+
+
+   i915->pipe.draw_arrays = i915_draw_arrays;
+   i915->pipe.draw_elements = i915_draw_elements;
+   i915->pipe.draw_range_elements = i915_draw_range_elements;
+
+   /*
+    * Create drawing context and plug our rendering stage into it.
+    */
+   i915->draw = draw_create();
+   assert(i915->draw);
+   if (!debug_get_bool_option("I915_NO_VBUF", FALSE)) {
+      draw_set_rasterize_stage(i915->draw, i915_draw_vbuf_stage(i915));
+   }
+   else {
+      draw_set_rasterize_stage(i915->draw, i915_draw_render_stage(i915));
+   }
+
+   i915_init_surface_functions(i915);
+   i915_init_state_functions(i915);
+   i915_init_flush_functions(i915);
+   i915_init_texture_functions(i915);
+
+   draw_install_aaline_stage(i915->draw, &i915->pipe);
+   draw_install_aapoint_stage(i915->draw, &i915->pipe);
+
+   i915->dirty = ~0;
+   i915->hardware_dirty = ~0;
+
+   /* Batch stream debugging is a bit hacked up at the moment:
+    */
+   i915->batch = i915_winsys->batch_get(i915_winsys);
+   i915->batch->winsys = i915_winsys;
+
+   return &i915->pipe;
+}
+
diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h
new file mode 100644
index 0000000000..3cdabe45f9
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_context.h
@@ -0,0 +1,345 @@
+ /**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef I915_CONTEXT_H
+#define I915_CONTEXT_H
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "draw/draw_vertex.h"
+
+#include "tgsi/tgsi_scan.h"
+
+
+#define I915_TEX_UNITS 8
+
+#define I915_DYNAMIC_MODES4       0
+#define I915_DYNAMIC_DEPTHSCALE_0 1 /* just the header */
+#define I915_DYNAMIC_DEPTHSCALE_1 2 
+#define I915_DYNAMIC_IAB          3
+#define I915_DYNAMIC_BC_0         4 /* just the header */
+#define I915_DYNAMIC_BC_1         5
+#define I915_DYNAMIC_BFO_0        6 
+#define I915_DYNAMIC_BFO_1        7
+#define I915_DYNAMIC_STP_0        8 
+#define I915_DYNAMIC_STP_1        9 
+#define I915_DYNAMIC_SC_ENA_0     10 
+#define I915_DYNAMIC_SC_RECT_0    11 
+#define I915_DYNAMIC_SC_RECT_1    12 
+#define I915_DYNAMIC_SC_RECT_2    13 
+#define I915_MAX_DYNAMIC          14
+
+
+#define I915_IMMEDIATE_S0         0
+#define I915_IMMEDIATE_S1         1
+#define I915_IMMEDIATE_S2         2
+#define I915_IMMEDIATE_S3         3
+#define I915_IMMEDIATE_S4         4
+#define I915_IMMEDIATE_S5         5
+#define I915_IMMEDIATE_S6         6
+#define I915_IMMEDIATE_S7         7
+#define I915_MAX_IMMEDIATE        8
+
+/* These must mach the order of LI0_STATE_* bits, as they will be used
+ * to generate hardware packets:
+ */
+#define I915_CACHE_STATIC         0 
+#define I915_CACHE_DYNAMIC        1 /* handled specially */
+#define I915_CACHE_SAMPLER        2
+#define I915_CACHE_MAP            3
+#define I915_CACHE_PROGRAM        4
+#define I915_CACHE_CONSTANTS      5
+#define I915_MAX_CACHE            6
+
+#define I915_MAX_CONSTANT  32
+
+
+/** See constant_flags[] below */
+#define I915_CONSTFLAG_USER 0x1f
+
+
+/**
+ * Subclass of pipe_shader_state
+ */
+struct i915_fragment_shader
+{
+   struct pipe_shader_state state;
+
+   struct tgsi_shader_info info;
+
+   uint *program;
+   uint program_len;
+
+   /**
+    * constants introduced during translation.
+    * These are placed at the end of the constant buffer and grow toward
+    * the beginning (eg: slot 31, 30 29, ...)
+    * User-provided constants start at 0.
+    * This allows both types of constants to co-exist (until there's too many)
+    * and doesn't require regenerating/changing the fragment program to
+    * shuffle constants around.
+    */
+   uint num_constants;
+   float constants[I915_MAX_CONSTANT][4];
+
+   /**
+    * Status of each constant
+    * if I915_CONSTFLAG_PARAM, the value must be taken from the corresponding
+    * slot of the user's constant buffer. (set by pipe->set_constant_buffer())
+    * Else, the bitmask indicates which components are occupied by immediates.
+    */
+   ubyte constant_flags[I915_MAX_CONSTANT];
+};
+
+
+struct i915_cache_context;
+
+/* Use to calculate differences between state emitted to hardware and
+ * current driver-calculated state.  
+ */
+struct i915_state 
+{
+   unsigned immediate[I915_MAX_IMMEDIATE];
+   unsigned dynamic[I915_MAX_DYNAMIC];
+
+   float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4];
+   /** number of constants passed in through a constant buffer */
+   uint num_user_constants[PIPE_SHADER_TYPES];
+
+   /* texture sampler state */
+   unsigned sampler[I915_TEX_UNITS][3];
+   unsigned sampler_enable_flags;
+   unsigned sampler_enable_nr;
+
+   /* texture image buffers */
+   unsigned texbuffer[I915_TEX_UNITS][2];
+
+   /** Describes the current hardware vertex layout */
+   struct vertex_info vertex_info;
+   
+   unsigned id;			/* track lost context events */
+};
+
+struct i915_blend_state {
+   unsigned iab;
+   unsigned modes4;
+   unsigned LIS5;
+   unsigned LIS6;
+};
+
+struct i915_depth_stencil_state {
+   unsigned stencil_modes4;
+   unsigned bfo[2];
+   unsigned stencil_LIS5;
+   unsigned depth_LIS6;
+};
+
+struct i915_rasterizer_state {
+   int light_twoside : 1;
+   unsigned st;
+   enum interp_mode color_interp;
+
+   unsigned LIS4;
+   unsigned LIS7;
+   unsigned sc[1];
+
+   const struct pipe_rasterizer_state *templ;
+
+   union { float f; unsigned u; } ds[2];
+};
+
+struct i915_sampler_state {
+   unsigned state[3];
+   const struct pipe_sampler_state *templ;
+   unsigned minlod;
+   unsigned maxlod;
+};
+
+
+struct i915_texture {
+   struct pipe_texture base;
+
+   /* Derived from the above:
+    */
+   unsigned stride;
+   unsigned depth_stride;          /* per-image on i945? */
+   unsigned total_nblocksy;
+
+   unsigned tiled;
+
+   unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS];
+
+   /* Explicitly store the offset of each image for each cube face or
+    * depth value.  Pretty much have to accept that hardware formats
+    * are going to be so diverse that there is no unified way to
+    * compute the offsets of depth/cube images within a mipmap level,
+    * so have to store them as a lookup table:
+    */
+   unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS];   /**< array [depth] of offsets */
+
+   /* The data is held here:
+    */
+   struct pipe_buffer *buffer;
+};
+
+struct i915_batchbuffer;
+
+struct i915_context
+{
+   struct pipe_context pipe;
+   struct i915_winsys *winsys;
+   struct draw_context *draw;
+
+   /* The most recent drawing state as set by the driver:
+    */
+   const struct i915_blend_state           *blend;
+   const struct i915_sampler_state         *sampler[PIPE_MAX_SAMPLERS];
+   const struct i915_depth_stencil_state   *depth_stencil;
+   const struct i915_rasterizer_state      *rasterizer;
+
+   struct i915_fragment_shader *fs;
+
+   struct pipe_blend_color blend_color;
+   struct pipe_clip_state clip;
+   struct pipe_constant_buffer constants[PIPE_SHADER_TYPES];
+   struct pipe_framebuffer_state framebuffer;
+   struct pipe_poly_stipple poly_stipple;
+   struct pipe_scissor_state scissor;
+   struct i915_texture *texture[PIPE_MAX_SAMPLERS];
+   struct pipe_viewport_state viewport;
+   struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+
+   unsigned dirty;
+
+   unsigned num_samplers;
+   unsigned num_textures;
+   unsigned num_vertex_elements;
+   unsigned num_vertex_buffers;
+
+   struct i915_batchbuffer *batch;
+
+   /** Vertex buffer */
+   struct pipe_buffer *vbo;
+   size_t vbo_offset;
+   unsigned vbo_flushed;
+
+   struct i915_state current;
+   unsigned hardware_dirty;
+   
+   unsigned debug;
+};
+
+/* A flag for each state_tracker state object:
+ */
+#define I915_NEW_VIEWPORT      0x1
+#define I915_NEW_RASTERIZER    0x2
+#define I915_NEW_FS            0x4
+#define I915_NEW_BLEND         0x8
+#define I915_NEW_CLIP          0x10
+#define I915_NEW_SCISSOR       0x20
+#define I915_NEW_STIPPLE       0x40
+#define I915_NEW_FRAMEBUFFER   0x80
+#define I915_NEW_ALPHA_TEST    0x100
+#define I915_NEW_DEPTH_STENCIL 0x200
+#define I915_NEW_SAMPLER       0x400
+#define I915_NEW_TEXTURE       0x800
+#define I915_NEW_CONSTANTS     0x1000
+#define I915_NEW_VBO           0x2000
+#define I915_NEW_VS            0x4000
+
+
+/* Driver's internally generated state flags:
+ */
+#define I915_NEW_VERTEX_FORMAT    0x10000
+
+
+/* Dirty flags for hardware emit
+ */
+#define I915_HW_STATIC            (1<<I915_CACHE_STATIC)
+#define I915_HW_DYNAMIC           (1<<I915_CACHE_DYNAMIC)
+#define I915_HW_SAMPLER           (1<<I915_CACHE_SAMPLER)
+#define I915_HW_MAP               (1<<I915_CACHE_MAP)
+#define I915_HW_PROGRAM           (1<<I915_CACHE_PROGRAM)
+#define I915_HW_CONSTANTS         (1<<I915_CACHE_CONSTANTS)
+#define I915_HW_IMMEDIATE         (1<<(I915_MAX_CACHE+0))
+#define I915_HW_INVARIENT         (1<<(I915_MAX_CACHE+1))
+
+
+/***********************************************************************
+ * i915_prim_emit.c: 
+ */
+struct draw_stage *i915_draw_render_stage( struct i915_context *i915 );
+
+
+/***********************************************************************
+ * i915_prim_vbuf.c: 
+ */
+struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 );
+
+
+/***********************************************************************
+ * i915_state_emit.c: 
+ */
+void i915_emit_hardware_state(struct i915_context *i915 );
+
+
+
+/***********************************************************************
+ * i915_clear.c: 
+ */
+void i915_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+		unsigned clearValue);
+
+
+/***********************************************************************
+ * i915_surface.c: 
+ */
+void i915_init_surface_functions( struct i915_context *i915 );
+
+void i915_init_state_functions( struct i915_context *i915 );
+void i915_init_flush_functions( struct i915_context *i915 );
+void i915_init_string_functions( struct i915_context *i915 );
+
+
+
+
+/***********************************************************************
+ * Inline conversion functions.  These are better-typed than the
+ * macros used previously:
+ */
+static INLINE struct i915_context *
+i915_context( struct pipe_context *pipe )
+{
+   return (struct i915_context *)pipe;
+}
+
+
+
+#endif
diff --git a/src/gallium/drivers/i915simple/i915_debug.c b/src/gallium/drivers/i915simple/i915_debug.c
new file mode 100644
index 0000000000..a300b61c3b
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_debug.c
@@ -0,0 +1,899 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "i915_reg.h"
+#include "i915_context.h"
+#include "i915_winsys.h"
+#include "i915_debug.h"
+#include "i915_batch.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_debug.h"
+
+
+static void
+PRINTF(
+   struct debug_stream  *stream,
+   const char           *fmt,
+                        ... )
+{
+   va_list  args;
+
+   va_start( args, fmt );
+   debug_vprintf( fmt, args );
+   va_end( args );
+}
+
+
+static boolean debug( struct debug_stream *stream, const char *name, unsigned len )
+{
+   unsigned i;
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   
+   if (len == 0) {
+      PRINTF(stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0]);
+      assert(0);
+      return FALSE;
+   }
+
+   if (stream->print_addresses)
+      PRINTF(stream, "%08x:  ", stream->offset);
+
+
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   for (i = 0; i < len; i++)
+      PRINTF(stream, "\t0x%08x\n",  ptr[i]);   
+   PRINTF(stream, "\n");
+
+   stream->offset += len * sizeof(unsigned);
+   
+   return TRUE;
+}
+
+
+static const char *get_prim_name( unsigned val )
+{
+   switch (val & PRIM3D_MASK) {
+   case PRIM3D_TRILIST: return "TRILIST"; break;
+   case PRIM3D_TRISTRIP: return "TRISTRIP"; break;
+   case PRIM3D_TRISTRIP_RVRSE: return "TRISTRIP_RVRSE"; break;
+   case PRIM3D_TRIFAN: return "TRIFAN"; break;
+   case PRIM3D_POLY: return "POLY"; break;
+   case PRIM3D_LINELIST: return "LINELIST"; break;
+   case PRIM3D_LINESTRIP: return "LINESTRIP"; break;
+   case PRIM3D_RECTLIST: return "RECTLIST"; break;
+   case PRIM3D_POINTLIST: return "POINTLIST"; break;
+   case PRIM3D_DIB: return "DIB"; break;
+   case PRIM3D_CLEAR_RECT: return "CLEAR_RECT"; break;
+   case PRIM3D_ZONE_INIT: return "ZONE_INIT"; break;
+   default: return "????"; break;
+   }
+}
+
+static boolean debug_prim( struct debug_stream *stream, const char *name, 
+			     boolean dump_floats,
+			     unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   const char *prim = get_prim_name( ptr[0] );
+   unsigned i;
+   
+
+
+   PRINTF(stream, "%s %s (%d dwords):\n", name, prim, len);
+   PRINTF(stream, "\t0x%08x\n",  ptr[0]);   
+   for (i = 1; i < len; i++) {
+      if (dump_floats)
+	 PRINTF(stream, "\t0x%08x // %f\n",  ptr[i], *(float *)&ptr[i]);   
+      else
+	 PRINTF(stream, "\t0x%08x\n",  ptr[i]);   
+   }
+
+      
+   PRINTF(stream, "\n");
+
+   stream->offset += len * sizeof(unsigned);
+   
+   return TRUE;
+}
+   
+
+
+
+static boolean debug_program( struct debug_stream *stream, const char *name, unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+
+   if (len == 0) {
+      PRINTF(stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0]);
+      assert(0);
+      return FALSE;
+   }
+
+   if (stream->print_addresses)
+      PRINTF(stream, "%08x:  ", stream->offset);
+
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   i915_disassemble_program( stream, ptr, len );
+
+   stream->offset += len * sizeof(unsigned);
+   return TRUE;
+}
+
+
+static boolean debug_chain( struct debug_stream *stream, const char *name, unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   unsigned old_offset = stream->offset + len * sizeof(unsigned);
+   unsigned i;
+
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   for (i = 0; i < len; i++)
+      PRINTF(stream, "\t0x%08x\n",  ptr[i]);
+
+   stream->offset = ptr[1] & ~0x3;
+   
+   if (stream->offset < old_offset)
+      PRINTF(stream, "\n... skipping backwards from 0x%x --> 0x%x ...\n\n", 
+		   old_offset, stream->offset );
+   else
+      PRINTF(stream, "\n... skipping from 0x%x --> 0x%x ...\n\n", 
+		   old_offset, stream->offset );
+
+
+   return TRUE;
+}
+
+
+static boolean debug_variable_length_prim( struct debug_stream *stream )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   const char *prim = get_prim_name( ptr[0] );
+   unsigned i, len;
+
+   ushort *idx = (ushort *)(ptr+1);
+   for (i = 0; idx[i] != 0xffff; i++)
+      ;
+
+   len = 1+(i+2)/2;
+
+   PRINTF(stream, "3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len);
+   for (i = 0; i < len; i++)
+      PRINTF(stream, "\t0x%08x\n",  ptr[i]);
+   PRINTF(stream, "\n");
+
+   stream->offset += len * sizeof(unsigned);
+   return TRUE;
+}
+
+
+static void
+BITS(
+   struct debug_stream  *stream,
+   unsigned             dw,
+   unsigned             hi,
+   unsigned             lo,
+   const char           *fmt,
+                        ... )
+{
+   va_list  args;
+   unsigned himask = ~0UL >> (31 - (hi));
+
+   PRINTF(stream, "\t\t ");
+
+   va_start( args, fmt );
+   debug_vprintf( fmt, args );
+   va_end( args );
+
+   PRINTF(stream, ": 0x%x\n", ((dw) & himask) >> (lo));
+}
+
+#ifdef DEBUG
+#define MBZ( dw, hi, lo) do {							\
+   unsigned x = (dw) >> (lo);				\
+   unsigned lomask = (1 << (lo)) - 1;			\
+   unsigned himask;					\
+   himask = (1UL << (hi)) - 1;				\
+   assert ((x & himask & ~lomask) == 0);	\
+} while (0)
+#else
+#define MBZ( dw, hi, lo) do {							\
+} while (0)
+#endif
+
+static void
+FLAG(
+   struct debug_stream  *stream,
+   unsigned             dw,
+   unsigned             bit,
+   const char           *fmt,
+                        ... )
+{
+   if (((dw) >> (bit)) & 1) {
+      va_list  args;
+
+      PRINTF(stream, "\t\t ");
+
+      va_start( args, fmt );
+      debug_vprintf( fmt, args );
+      va_end( args );
+
+      PRINTF(stream, "\n");
+   }
+}
+
+static boolean debug_load_immediate( struct debug_stream *stream,
+				       const char *name,
+				       unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   unsigned bits = (ptr[0] >> 4) & 0xff;
+   unsigned j = 0;
+   
+   PRINTF(stream, "%s (%d dwords, flags: %x):\n", name, len, bits);
+   PRINTF(stream, "\t0x%08x\n",  ptr[j++]);
+
+   if (bits & (1<<0)) {
+      PRINTF(stream, "\t  LIS0: 0x%08x\n", ptr[j]);
+      PRINTF(stream, "\t vb address: 0x%08x\n", (ptr[j] & ~0x3));
+      BITS(stream, ptr[j], 0, 0, "vb invalidate disable");
+      j++;
+   }
+   if (bits & (1<<1)) {
+      PRINTF(stream, "\t  LIS1: 0x%08x\n", ptr[j]);
+      BITS(stream, ptr[j], 29, 24, "vb dword width");
+      BITS(stream, ptr[j], 21, 16, "vb dword pitch");
+      BITS(stream, ptr[j], 15, 0, "vb max index");
+      j++;
+   }
+   if (bits & (1<<2)) {
+      int i;
+      PRINTF(stream, "\t  LIS2: 0x%08x\n", ptr[j]);
+      for (i = 0; i < 8; i++) {
+	 unsigned tc = (ptr[j] >> (i * 4)) & 0xf;
+	 if (tc != 0xf)
+	    BITS(stream, tc, 3, 0, "tex coord %d", i);
+      }
+      j++;
+   }
+   if (bits & (1<<3)) {
+      PRINTF(stream, "\t  LIS3: 0x%08x\n", ptr[j]);
+      j++;
+   }
+   if (bits & (1<<4)) {
+      PRINTF(stream, "\t  LIS4: 0x%08x\n", ptr[j]);
+      BITS(stream, ptr[j], 31, 23, "point width");
+      BITS(stream, ptr[j], 22, 19, "line width");
+      FLAG(stream, ptr[j], 18, "alpha flatshade");
+      FLAG(stream, ptr[j], 17, "fog flatshade");
+      FLAG(stream, ptr[j], 16, "spec flatshade");
+      FLAG(stream, ptr[j], 15, "rgb flatshade");
+      BITS(stream, ptr[j], 14, 13, "cull mode");
+      FLAG(stream, ptr[j], 12, "vfmt: point width");
+      FLAG(stream, ptr[j], 11, "vfmt: specular/fog");
+      FLAG(stream, ptr[j], 10, "vfmt: rgba");
+      FLAG(stream, ptr[j], 9, "vfmt: depth offset");
+      BITS(stream, ptr[j], 8, 6, "vfmt: position (2==xyzw)");
+      FLAG(stream, ptr[j], 5, "force dflt diffuse");
+      FLAG(stream, ptr[j], 4, "force dflt specular");
+      FLAG(stream, ptr[j], 3, "local depth offset enable");
+      FLAG(stream, ptr[j], 2, "vfmt: fp32 fog coord");
+      FLAG(stream, ptr[j], 1, "sprite point");
+      FLAG(stream, ptr[j], 0, "antialiasing");
+      j++;
+   }
+   if (bits & (1<<5)) {
+      PRINTF(stream, "\t  LIS5: 0x%08x\n", ptr[j]);
+      BITS(stream, ptr[j], 31, 28, "rgba write disables");
+      FLAG(stream, ptr[j], 27,     "force dflt point width");
+      FLAG(stream, ptr[j], 26,     "last pixel enable");
+      FLAG(stream, ptr[j], 25,     "global z offset enable");
+      FLAG(stream, ptr[j], 24,     "fog enable");
+      BITS(stream, ptr[j], 23, 16, "stencil ref");
+      BITS(stream, ptr[j], 15, 13, "stencil test");
+      BITS(stream, ptr[j], 12, 10, "stencil fail op");
+      BITS(stream, ptr[j], 9, 7,   "stencil pass z fail op");
+      BITS(stream, ptr[j], 6, 4,   "stencil pass z pass op");
+      FLAG(stream, ptr[j], 3,      "stencil write enable");
+      FLAG(stream, ptr[j], 2,      "stencil test enable");
+      FLAG(stream, ptr[j], 1,      "color dither enable");
+      FLAG(stream, ptr[j], 0,      "logiop enable");
+      j++;
+   }
+   if (bits & (1<<6)) {
+      PRINTF(stream, "\t  LIS6: 0x%08x\n", ptr[j]);
+      FLAG(stream, ptr[j], 31,      "alpha test enable");
+      BITS(stream, ptr[j], 30, 28,  "alpha func");
+      BITS(stream, ptr[j], 27, 20,  "alpha ref");
+      FLAG(stream, ptr[j], 19,      "depth test enable");
+      BITS(stream, ptr[j], 18, 16,  "depth func");
+      FLAG(stream, ptr[j], 15,      "blend enable");
+      BITS(stream, ptr[j], 14, 12,  "blend func");
+      BITS(stream, ptr[j], 11, 8,   "blend src factor");
+      BITS(stream, ptr[j], 7,  4,   "blend dst factor");
+      FLAG(stream, ptr[j], 3,       "depth write enable");
+      FLAG(stream, ptr[j], 2,       "color write enable");
+      BITS(stream, ptr[j], 1,  0,   "provoking vertex"); 
+      j++;
+   }
+
+
+   PRINTF(stream, "\n");
+
+   assert(j == len);
+
+   stream->offset += len * sizeof(unsigned);
+   
+   return TRUE;
+}
+ 
+
+
+static boolean debug_load_indirect( struct debug_stream *stream,
+				      const char *name,
+				      unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   unsigned bits = (ptr[0] >> 8) & 0x3f;
+   unsigned i, j = 0;
+   
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   PRINTF(stream, "\t0x%08x\n",  ptr[j++]);
+
+   for (i = 0; i < 6; i++) {
+      if (bits & (1<<i)) {
+	 switch (1<<(8+i)) {
+	 case LI0_STATE_STATIC_INDIRECT:
+	    PRINTF(stream, "        STATIC: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++;
+	    PRINTF(stream, "                0x%08x\n", ptr[j++]);
+	    break;
+	 case LI0_STATE_DYNAMIC_INDIRECT:
+	    PRINTF(stream, "       DYNAMIC: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++;
+	    break;
+	 case LI0_STATE_SAMPLER:
+	    PRINTF(stream, "       SAMPLER: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++;
+	    PRINTF(stream, "                0x%08x\n", ptr[j++]);
+	    break;
+	 case LI0_STATE_MAP:
+	    PRINTF(stream, "           MAP: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++;
+	    PRINTF(stream, "                0x%08x\n", ptr[j++]);
+	    break;
+	 case LI0_STATE_PROGRAM:
+	    PRINTF(stream, "       PROGRAM: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++;
+	    PRINTF(stream, "                0x%08x\n", ptr[j++]);
+	    break;
+	 case LI0_STATE_CONSTANTS:
+	    PRINTF(stream, "     CONSTANTS: 0x%08x | %x\n", ptr[j]&~3, ptr[j]&3); j++;
+	    PRINTF(stream, "                0x%08x\n", ptr[j++]);
+	    break;
+	 default:
+	    assert(0);
+	    break;
+	 }
+      }
+   }
+
+   if (bits == 0) {
+      PRINTF(stream, "\t  DUMMY: 0x%08x\n", ptr[j++]);
+   }
+
+   PRINTF(stream, "\n");
+
+
+   assert(j == len);
+
+   stream->offset += len * sizeof(unsigned);
+   
+   return TRUE;
+}
+ 	
+static void BR13( struct debug_stream *stream,
+		  unsigned val )
+{
+   PRINTF(stream, "\t0x%08x\n",  val);
+   FLAG(stream, val, 30, "clipping enable");
+   BITS(stream, val, 25, 24, "color depth (3==32bpp)");
+   BITS(stream, val, 23, 16, "raster op");
+   BITS(stream, val, 15, 0,  "dest pitch");
+}
+
+
+static void BR22( struct debug_stream *stream,
+		  unsigned val )
+{
+   PRINTF(stream, "\t0x%08x\n",  val);
+   BITS(stream, val, 31, 16, "dest y1");
+   BITS(stream, val, 15, 0,  "dest x1");
+}
+
+static void BR23( struct debug_stream *stream,
+		  unsigned val )
+{
+   PRINTF(stream, "\t0x%08x\n",  val);
+   BITS(stream, val, 31, 16, "dest y2");
+   BITS(stream, val, 15, 0,  "dest x2");
+}
+
+static void BR09( struct debug_stream *stream,
+		  unsigned val )
+{
+   PRINTF(stream, "\t0x%08x -- dest address\n",  val);
+}
+
+static void BR26( struct debug_stream *stream,
+		  unsigned val )
+{
+   PRINTF(stream, "\t0x%08x\n",  val);
+   BITS(stream, val, 31, 16, "src y1");
+   BITS(stream, val, 15, 0,  "src x1");
+}
+
+static void BR11( struct debug_stream *stream,
+		  unsigned val )
+{
+   PRINTF(stream, "\t0x%08x\n",  val);
+   BITS(stream, val, 15, 0,  "src pitch");
+}
+
+static void BR12( struct debug_stream *stream,
+		  unsigned val )
+{
+   PRINTF(stream, "\t0x%08x -- src address\n",  val);
+}
+
+static void BR16( struct debug_stream *stream,
+		  unsigned val )
+{
+   PRINTF(stream, "\t0x%08x -- color\n",  val);
+}
+   
+static boolean debug_copy_blit( struct debug_stream *stream,
+				  const char *name,
+				  unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   int j = 0;
+
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   PRINTF(stream, "\t0x%08x\n",  ptr[j++]);
+   
+   BR13(stream, ptr[j++]);
+   BR22(stream, ptr[j++]);
+   BR23(stream, ptr[j++]);
+   BR09(stream, ptr[j++]);
+   BR26(stream, ptr[j++]);
+   BR11(stream, ptr[j++]);
+   BR12(stream, ptr[j++]);
+
+   stream->offset += len * sizeof(unsigned);
+   assert(j == len);
+   return TRUE;
+}
+
+static boolean debug_color_blit( struct debug_stream *stream,
+				  const char *name,
+				  unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   int j = 0;
+
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   PRINTF(stream, "\t0x%08x\n",  ptr[j++]);
+
+   BR13(stream, ptr[j++]);
+   BR22(stream, ptr[j++]);
+   BR23(stream, ptr[j++]);
+   BR09(stream, ptr[j++]);
+   BR16(stream, ptr[j++]);
+
+   stream->offset += len * sizeof(unsigned);
+   assert(j == len);
+   return TRUE;
+}
+
+static boolean debug_modes4( struct debug_stream *stream,
+				  const char *name,
+				  unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   int j = 0;
+
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   PRINTF(stream, "\t0x%08x\n",  ptr[j]);
+   BITS(stream, ptr[j], 21, 18, "logicop func");
+   FLAG(stream, ptr[j], 17, "stencil test mask modify-enable");
+   FLAG(stream, ptr[j], 16, "stencil write mask modify-enable");
+   BITS(stream, ptr[j], 15, 8, "stencil test mask");
+   BITS(stream, ptr[j], 7, 0,  "stencil write mask");
+   j++;
+
+   stream->offset += len * sizeof(unsigned);
+   assert(j == len);
+   return TRUE;
+}
+
+static boolean debug_map_state( struct debug_stream *stream,
+				  const char *name,
+				  unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   unsigned j = 0;
+
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   PRINTF(stream, "\t0x%08x\n",  ptr[j++]);
+   
+   {
+      PRINTF(stream, "\t0x%08x\n",  ptr[j]);
+      BITS(stream, ptr[j], 15, 0,   "map mask");
+      j++;
+   }
+
+   while (j < len) {
+      {
+	 PRINTF(stream, "\t  TMn.0: 0x%08x\n", ptr[j]);
+	 PRINTF(stream, "\t map address: 0x%08x\n", (ptr[j] & ~0x3));
+	 FLAG(stream, ptr[j], 1, "vertical line stride");
+	 FLAG(stream, ptr[j], 0, "vertical line stride offset");
+	 j++;
+      }
+
+      {
+	 PRINTF(stream, "\t  TMn.1: 0x%08x\n", ptr[j]);
+	 BITS(stream, ptr[j], 31, 21, "height");
+	 BITS(stream, ptr[j], 20, 10, "width");
+	 BITS(stream, ptr[j], 9, 7, "surface format");
+	 BITS(stream, ptr[j], 6, 3, "texel format");
+	 FLAG(stream, ptr[j], 2, "use fence regs");
+	 FLAG(stream, ptr[j], 1, "tiled surface");
+	 FLAG(stream, ptr[j], 0, "tile walk ymajor");
+	 j++;
+      }
+      {
+	 PRINTF(stream, "\t  TMn.2: 0x%08x\n", ptr[j]);
+	 BITS(stream, ptr[j], 31, 21, "dword pitch");
+	 BITS(stream, ptr[j], 20, 15, "cube face enables");
+	 BITS(stream, ptr[j], 14, 9, "max lod");
+	 FLAG(stream, ptr[j], 8,     "mip layout right");
+	 BITS(stream, ptr[j], 7, 0, "depth");
+	 j++;
+      }
+   }
+
+   stream->offset += len * sizeof(unsigned);
+   assert(j == len);
+   return TRUE;
+}
+
+static boolean debug_sampler_state( struct debug_stream *stream,
+				  const char *name,
+				  unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   unsigned j = 0;
+
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   PRINTF(stream, "\t0x%08x\n",  ptr[j++]);
+   
+   {
+      PRINTF(stream, "\t0x%08x\n",  ptr[j]);
+      BITS(stream, ptr[j], 15, 0,   "sampler mask");
+      j++;
+   }
+
+   while (j < len) {
+      {
+	 PRINTF(stream, "\t  TSn.0: 0x%08x\n", ptr[j]);
+	 FLAG(stream, ptr[j], 31, "reverse gamma");
+	 FLAG(stream, ptr[j], 30, "planar to packed");
+	 FLAG(stream, ptr[j], 29, "yuv->rgb");
+	 BITS(stream, ptr[j], 28, 27, "chromakey index");
+	 BITS(stream, ptr[j], 26, 22, "base mip level");
+	 BITS(stream, ptr[j], 21, 20, "mip mode filter");
+	 BITS(stream, ptr[j], 19, 17, "mag mode filter");
+	 BITS(stream, ptr[j], 16, 14, "min mode filter");
+	 BITS(stream, ptr[j], 13, 5,  "lod bias (s4.4)");
+	 FLAG(stream, ptr[j], 4,      "shadow enable");
+	 FLAG(stream, ptr[j], 3,      "max-aniso-4");
+	 BITS(stream, ptr[j], 2, 0,   "shadow func");
+	 j++;
+      }
+
+      {
+	 PRINTF(stream, "\t  TSn.1: 0x%08x\n", ptr[j]);
+	 BITS(stream, ptr[j], 31, 24, "min lod");
+	 MBZ( ptr[j], 23, 18 );
+	 FLAG(stream, ptr[j], 17,     "kill pixel enable");
+	 FLAG(stream, ptr[j], 16,     "keyed tex filter mode");
+	 FLAG(stream, ptr[j], 15,     "chromakey enable");
+	 BITS(stream, ptr[j], 14, 12, "tcx wrap mode");
+	 BITS(stream, ptr[j], 11, 9,  "tcy wrap mode");
+	 BITS(stream, ptr[j], 8,  6,  "tcz wrap mode");
+	 FLAG(stream, ptr[j], 5,      "normalized coords");
+	 BITS(stream, ptr[j], 4,  1,  "map (surface) index");
+	 FLAG(stream, ptr[j], 0,      "EAST deinterlacer enable");
+	 j++;
+      }
+      {
+	 PRINTF(stream, "\t  TSn.2: 0x%08x  (default color)\n", ptr[j]);
+	 j++;
+      }
+   }
+
+   stream->offset += len * sizeof(unsigned);
+   assert(j == len);
+   return TRUE;
+}
+
+static boolean debug_dest_vars( struct debug_stream *stream,
+				  const char *name,
+				  unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   int j = 0;
+
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   PRINTF(stream, "\t0x%08x\n",  ptr[j++]);
+
+   {
+      PRINTF(stream, "\t0x%08x\n",  ptr[j]);
+      FLAG(stream, ptr[j], 31,     "early classic ztest");
+      FLAG(stream, ptr[j], 30,     "opengl tex default color");
+      FLAG(stream, ptr[j], 29,     "bypass iz");
+      FLAG(stream, ptr[j], 28,     "lod preclamp");
+      BITS(stream, ptr[j], 27, 26, "dither pattern");
+      FLAG(stream, ptr[j], 25,     "linear gamma blend");
+      FLAG(stream, ptr[j], 24,     "debug dither");
+      BITS(stream, ptr[j], 23, 20, "dstorg x");
+      BITS(stream, ptr[j], 19, 16, "dstorg y");
+      MBZ (ptr[j], 15, 15 );
+      BITS(stream, ptr[j], 14, 12, "422 write select");
+      BITS(stream, ptr[j], 11, 8,  "cbuf format");
+      BITS(stream, ptr[j], 3, 2,   "zbuf format");
+      FLAG(stream, ptr[j], 1,      "vert line stride");
+      FLAG(stream, ptr[j], 1,      "vert line stride offset");
+      j++;
+   }
+   
+   stream->offset += len * sizeof(unsigned);
+   assert(j == len);
+   return TRUE;
+}
+
+static boolean debug_buf_info( struct debug_stream *stream,
+				  const char *name,
+				  unsigned len )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   int j = 0;
+
+   PRINTF(stream, "%s (%d dwords):\n", name, len);
+   PRINTF(stream, "\t0x%08x\n",  ptr[j++]);
+
+   {
+      PRINTF(stream, "\t0x%08x\n",  ptr[j]);
+      BITS(stream, ptr[j], 28, 28, "aux buffer id");
+      BITS(stream, ptr[j], 27, 24, "buffer id (7=depth, 3=back)");
+      FLAG(stream, ptr[j], 23,     "use fence regs");
+      FLAG(stream, ptr[j], 22,     "tiled surface");
+      FLAG(stream, ptr[j], 21,     "tile walk ymajor");
+      MBZ (ptr[j], 20, 14);
+      BITS(stream, ptr[j], 13, 2,  "dword pitch");
+      MBZ (ptr[j], 2,  0);
+      j++;
+   }
+   
+   PRINTF(stream, "\t0x%08x -- buffer base address\n",  ptr[j++]);
+
+   stream->offset += len * sizeof(unsigned);
+   assert(j == len);
+   return TRUE;
+}
+
+static boolean i915_debug_packet( struct debug_stream *stream )
+{
+   unsigned *ptr = (unsigned *)(stream->ptr + stream->offset);
+   unsigned cmd = *ptr;
+   
+   switch (((cmd >> 29) & 0x7)) {
+   case 0x0:
+      switch ((cmd >> 23) & 0x3f) {
+      case 0x0:
+	 return debug(stream, "MI_NOOP", 1);
+      case 0x3:
+	 return debug(stream, "MI_WAIT_FOR_EVENT", 1);
+      case 0x4:
+	 return debug(stream, "MI_FLUSH", 1);
+      case 0xA:
+	 debug(stream, "MI_BATCH_BUFFER_END", 1);
+	 return FALSE;
+      case 0x22:
+	 return debug(stream, "MI_LOAD_REGISTER_IMM", 3);
+      case 0x31:
+	 return debug_chain(stream, "MI_BATCH_BUFFER_START", 2);
+      default:
+         (void)debug(stream, "UNKNOWN 0x0 case!", 1);
+         assert(0);
+	 break;
+      }
+      break;
+   case 0x1:
+      (void) debug(stream, "UNKNOWN 0x1 case!", 1);
+      assert(0);
+      break;
+   case 0x2:
+      switch ((cmd >> 22) & 0xff) {	 
+      case 0x50:
+	 return debug_color_blit(stream, "XY_COLOR_BLT", (cmd & 0xff) + 2);
+      case 0x53:
+	 return debug_copy_blit(stream, "XY_SRC_COPY_BLT", (cmd & 0xff) + 2);
+      default:
+	 return debug(stream, "blit command", (cmd & 0xff) + 2);
+      }
+      break;
+   case 0x3:
+      switch ((cmd >> 24) & 0x1f) {	 
+      case 0x6:
+	 return debug(stream, "3DSTATE_ANTI_ALIASING", 1);
+      case 0x7:
+	 return debug(stream, "3DSTATE_RASTERIZATION_RULES", 1);
+      case 0x8:
+	 return debug(stream, "3DSTATE_BACKFACE_STENCIL_OPS", 2);
+      case 0x9:
+	 return debug(stream, "3DSTATE_BACKFACE_STENCIL_MASKS", 1);
+      case 0xb:
+	 return debug(stream, "3DSTATE_INDEPENDENT_ALPHA_BLEND", 1);
+      case 0xc:
+	 return debug(stream, "3DSTATE_MODES5", 1);	 
+      case 0xd:
+	 return debug_modes4(stream, "3DSTATE_MODES4", 1);
+      case 0x15:
+	 return debug(stream, "3DSTATE_FOG_COLOR", 1);
+      case 0x16:
+	 return debug(stream, "3DSTATE_COORD_SET_BINDINGS", 1);
+      case 0x1c:
+	 /* 3DState16NP */
+	 switch((cmd >> 19) & 0x1f) {
+	 case 0x10:
+	    return debug(stream, "3DSTATE_SCISSOR_ENABLE", 1);
+	 case 0x11:
+	    return debug(stream, "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE", 1);
+	 default:
+            (void) debug(stream, "UNKNOWN 0x1c case!", 1);
+            assert(0);
+	    break;
+	 }
+	 break;
+      case 0x1d:
+	 /* 3DStateMW */
+	 switch ((cmd >> 16) & 0xff) {
+	 case 0x0:
+	    return debug_map_state(stream, "3DSTATE_MAP_STATE", (cmd & 0x1f) + 2);
+	 case 0x1:
+	    return debug_sampler_state(stream, "3DSTATE_SAMPLER_STATE", (cmd & 0x1f) + 2);
+	 case 0x4:
+	    return debug_load_immediate(stream, "3DSTATE_LOAD_STATE_IMMEDIATE", (cmd & 0xf) + 2);
+	 case 0x5:
+	    return debug_program(stream, "3DSTATE_PIXEL_SHADER_PROGRAM", (cmd & 0x1ff) + 2);
+	 case 0x6:
+	    return debug(stream, "3DSTATE_PIXEL_SHADER_CONSTANTS", (cmd & 0xff) + 2);
+	 case 0x7:
+	    return debug_load_indirect(stream, "3DSTATE_LOAD_INDIRECT", (cmd & 0xff) + 2);
+	 case 0x80:
+	    return debug(stream, "3DSTATE_DRAWING_RECTANGLE", (cmd & 0xffff) + 2);
+	 case 0x81:
+	    return debug(stream, "3DSTATE_SCISSOR_RECTANGLE", (cmd & 0xffff) + 2);
+	 case 0x83:
+	    return debug(stream, "3DSTATE_SPAN_STIPPLE", (cmd & 0xffff) + 2);
+	 case 0x85:
+	    return debug_dest_vars(stream, "3DSTATE_DEST_BUFFER_VARS", (cmd & 0xffff) + 2);
+	 case 0x88:
+	    return debug(stream, "3DSTATE_CONSTANT_BLEND_COLOR", (cmd & 0xffff) + 2);
+	 case 0x89:
+	    return debug(stream, "3DSTATE_FOG_MODE", (cmd & 0xffff) + 2);
+	 case 0x8e:
+	    return debug_buf_info(stream, "3DSTATE_BUFFER_INFO", (cmd & 0xffff) + 2);
+	 case 0x97:
+	    return debug(stream, "3DSTATE_DEPTH_OFFSET_SCALE", (cmd & 0xffff) + 2);
+	 case 0x98:
+	    return debug(stream, "3DSTATE_DEFAULT_Z", (cmd & 0xffff) + 2);
+	 case 0x99:
+	    return debug(stream, "3DSTATE_DEFAULT_DIFFUSE", (cmd & 0xffff) + 2);
+	 case 0x9a:
+	    return debug(stream, "3DSTATE_DEFAULT_SPECULAR", (cmd & 0xffff) + 2);
+	 case 0x9c:
+	    return debug(stream, "3DSTATE_CLEAR_PARAMETERS", (cmd & 0xffff) + 2);
+	 default:
+	    assert(0);
+	    return 0;
+	 }
+	 break;
+      case 0x1e:
+	 if (cmd & (1 << 23))
+	    return debug(stream, "???", (cmd & 0xffff) + 1);
+	 else
+	    return debug(stream, "", 1);
+	 break;
+      case 0x1f:
+	 if ((cmd & (1 << 23)) == 0)	
+	    return debug_prim(stream, "3DPRIM (inline)", 1, (cmd & 0x1ffff) + 2);
+	 else if (cmd & (1 << 17)) 
+	 {
+	    if ((cmd & 0xffff) == 0)
+	       return debug_variable_length_prim(stream);
+	    else
+	       return debug_prim(stream, "3DPRIM (indexed)", 0, (((cmd & 0xffff) + 1) / 2) + 1);
+	 }
+	 else
+	    return debug_prim(stream, "3DPRIM  (indirect sequential)", 0, 2); 
+	 break;
+      default:
+	 return debug(stream, "", 0);
+      }
+   default:
+      assert(0);
+      return 0;
+   }
+
+   assert(0);
+   return 0;
+}
+
+
+
+void
+i915_dump_batchbuffer( struct i915_batchbuffer *batch )
+{
+   struct debug_stream stream;
+   unsigned *start = (unsigned*)batch->map;
+   unsigned *end = (unsigned*)batch->ptr;
+   unsigned long bytes = (unsigned long) (end - start) * 4;
+   boolean done = FALSE;
+
+   stream.offset = 0;
+   stream.ptr = (char *)start;
+   stream.print_addresses = 0;
+
+   if (!start || !end) {
+      debug_printf( "\n\nBATCH: ???\n");
+      return;
+   }
+   
+   debug_printf( "\n\nBATCH: (%d)\n", bytes / 4);
+
+   while (!done &&
+	  stream.offset < bytes)
+   {
+      if (!i915_debug_packet( &stream ))
+	 break;
+
+      assert(stream.offset <= bytes &&
+	     stream.offset >= 0);
+   }
+
+   debug_printf( "END-BATCH\n\n\n");
+}
+
+
diff --git a/src/gallium/drivers/i915simple/i915_debug.h b/src/gallium/drivers/i915simple/i915_debug.h
new file mode 100644
index 0000000000..16ca7277c7
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_debug.h
@@ -0,0 +1,114 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef I915_DEBUG_H
+#define I915_DEBUG_H
+
+#include <stdarg.h>
+
+struct i915_context;
+
+struct debug_stream 
+{
+   unsigned offset;		/* current gtt offset */
+   char *ptr;		/* pointer to gtt offset zero */
+   char *end;		/* pointer to gtt offset zero */
+   unsigned print_addresses;
+};
+
+
+/* Internal functions
+ */
+void i915_disassemble_program(struct debug_stream *stream, 
+			      const unsigned *program, unsigned sz);
+
+void i915_print_ureg(const char *msg, unsigned ureg);
+
+
+#define DEBUG_BATCH	 0x1
+#define DEBUG_BLIT       0x2
+#define DEBUG_BUFFER     0x4
+#define DEBUG_CONSTANTS  0x8
+#define DEBUG_CONTEXT    0x10
+#define DEBUG_DRAW	 0x20
+#define DEBUG_DYNAMIC	 0x40
+#define DEBUG_FLUSH      0x80
+#define DEBUG_MAP	 0x100
+#define DEBUG_PROGRAM	 0x200
+#define DEBUG_REGIONS    0x400
+#define DEBUG_SAMPLER	 0x800
+#define DEBUG_STATIC	 0x1000
+#define DEBUG_SURFACE    0x2000
+#define DEBUG_WINSYS     0x4000
+
+#include "pipe/p_compiler.h"
+
+#if defined(DEBUG) && defined(FILE_DEBUG_FLAG)
+
+#include "pipe/internal/p_winsys_screen.h"
+
+static INLINE void
+I915_DBG(
+   struct i915_context  *i915,
+   const char           *fmt,
+                        ... )
+{
+   if ((i915)->debug & FILE_DEBUG_FLAG) {
+      va_list  args;
+
+      va_start( args, fmt );
+      debug_vprintf( fmt, args );
+      va_end( args );
+   }
+}
+
+#else
+
+static INLINE void
+I915_DBG(
+   struct i915_context  *i915,
+   const char           *fmt,
+                        ... )
+{
+   (void) i915;
+   (void) fmt;
+}
+
+#endif
+
+
+struct i915_batchbuffer;
+
+void i915_dump_batchbuffer( struct i915_batchbuffer *i915 );
+
+void i915_debug_init( struct i915_context *i915 );
+
+
+#endif
diff --git a/src/gallium/drivers/i915simple/i915_debug_fp.c b/src/gallium/drivers/i915simple/i915_debug_fp.c
new file mode 100644
index 0000000000..9c5b117b6d
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_debug_fp.c
@@ -0,0 +1,363 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "i915_reg.h"
+#include "i915_debug.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_memory.h"
+
+
+static void
+PRINTF(
+   struct debug_stream  *stream,
+   const char           *fmt,
+                        ... )
+{
+   va_list  args;
+
+   va_start( args, fmt );
+   debug_vprintf( fmt, args );
+   va_end( args );
+}
+
+
+static const char *opcodes[0x20] = {
+   "NOP",
+   "ADD",
+   "MOV",
+   "MUL",
+   "MAD",
+   "DP2ADD",
+   "DP3",
+   "DP4",
+   "FRC",
+   "RCP",
+   "RSQ",
+   "EXP",
+   "LOG",
+   "CMP",
+   "MIN",
+   "MAX",
+   "FLR",
+   "MOD",
+   "TRC",
+   "SGE",
+   "SLT",
+   "TEXLD",
+   "TEXLDP",
+   "TEXLDB",
+   "TEXKILL",
+   "DCL",
+   "0x1a",
+   "0x1b",
+   "0x1c",
+   "0x1d",
+   "0x1e",
+   "0x1f",
+};
+
+
+static const int args[0x20] = {
+   0,                           /* 0 nop */
+   2,                           /* 1 add */
+   1,                           /* 2 mov */
+   2,                           /* 3 m ul */
+   3,                           /* 4 mad */
+   3,                           /* 5 dp2add */
+   2,                           /* 6 dp3 */
+   2,                           /* 7 dp4 */
+   1,                           /* 8 frc */
+   1,                           /* 9 rcp */
+   1,                           /* a rsq */
+   1,                           /* b exp */
+   1,                           /* c log */
+   3,                           /* d cmp */
+   2,                           /* e min */
+   2,                           /* f max */
+   1,                           /* 10 flr */
+   1,                           /* 11 mod */
+   1,                           /* 12 trc */
+   2,                           /* 13 sge */
+   2,                           /* 14 slt */
+   1,
+   1,
+   1,
+   1,
+   0,
+   0,
+   0,
+   0,
+   0,
+   0,
+   0,
+};
+
+
+static const char *regname[0x8] = {
+   "R",
+   "T",
+   "CONST",
+   "S",
+   "OC",
+   "OD",
+   "U",
+   "UNKNOWN",
+};
+
+static void
+print_reg_type_nr(struct debug_stream *stream, unsigned type, unsigned nr)
+{
+   switch (type) {
+   case REG_TYPE_T:
+      switch (nr) {
+      case T_DIFFUSE:
+         PRINTF(stream, "T_DIFFUSE");
+         return;
+      case T_SPECULAR:
+         PRINTF(stream, "T_SPECULAR");
+         return;
+      case T_FOG_W:
+         PRINTF(stream, "T_FOG_W");
+         return;
+      default:
+         PRINTF(stream, "T_TEX%d", nr);
+         return;
+      }
+   case REG_TYPE_OC:
+      if (nr == 0) {
+         PRINTF(stream, "oC");
+         return;
+      }
+      break;
+   case REG_TYPE_OD:
+      if (nr == 0) {
+         PRINTF(stream, "oD");
+         return;
+      }
+      break;
+   default:
+      break;
+   }
+
+   PRINTF(stream, "%s[%d]", regname[type], nr);
+}
+
+#define REG_SWIZZLE_MASK 0x7777
+#define REG_NEGATE_MASK 0x8888
+
+#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) |	\
+		      (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) |	\
+		      (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) |	\
+		      (SRC_W << A2_SRC2_CHANNEL_W_SHIFT))
+
+
+static void
+print_reg_neg_swizzle(struct debug_stream *stream, unsigned reg)
+{
+   int i;
+
+   if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW &&
+       (reg & REG_NEGATE_MASK) == 0)
+      return;
+
+   PRINTF(stream, ".");
+
+   for (i = 3; i >= 0; i--) {
+      if (reg & (1 << ((i * 4) + 3)))
+         PRINTF(stream, "-");
+
+      switch ((reg >> (i * 4)) & 0x7) {
+      case 0:
+         PRINTF(stream, "x");
+         break;
+      case 1:
+         PRINTF(stream, "y");
+         break;
+      case 2:
+         PRINTF(stream, "z");
+         break;
+      case 3:
+         PRINTF(stream, "w");
+         break;
+      case 4:
+         PRINTF(stream, "0");
+         break;
+      case 5:
+         PRINTF(stream, "1");
+         break;
+      default:
+         PRINTF(stream, "?");
+         break;
+      }
+   }
+}
+
+
+static void
+print_src_reg(struct debug_stream *stream, unsigned dword)
+{
+   unsigned nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK;
+   unsigned type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK;
+   print_reg_type_nr(stream, type, nr);
+   print_reg_neg_swizzle(stream, dword);
+}
+
+
+static void
+print_dest_reg(struct debug_stream *stream, unsigned dword)
+{
+   unsigned nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK;
+   unsigned type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK;
+   print_reg_type_nr(stream, type, nr);
+   if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL)
+      return;
+   PRINTF(stream, ".");
+   if (dword & A0_DEST_CHANNEL_X)
+      PRINTF(stream, "x");
+   if (dword & A0_DEST_CHANNEL_Y)
+      PRINTF(stream, "y");
+   if (dword & A0_DEST_CHANNEL_Z)
+      PRINTF(stream, "z");
+   if (dword & A0_DEST_CHANNEL_W)
+      PRINTF(stream, "w");
+}
+
+
+#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT))
+#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT))
+#define GET_SRC2_REG(r)      (r)
+
+
+static void
+print_arith_op(struct debug_stream *stream, 
+	       unsigned opcode, const unsigned * program)
+{
+   if (opcode != A0_NOP) {
+      print_dest_reg(stream, program[0]);
+      if (program[0] & A0_DEST_SATURATE)
+         PRINTF(stream, " = SATURATE ");
+      else
+         PRINTF(stream, " = ");
+   }
+
+   PRINTF(stream, "%s ", opcodes[opcode]);
+
+   print_src_reg(stream, GET_SRC0_REG(program[0], program[1]));
+   if (args[opcode] == 1) {
+      PRINTF(stream, "\n");
+      return;
+   }
+
+   PRINTF(stream, ", ");
+   print_src_reg(stream, GET_SRC1_REG(program[1], program[2]));
+   if (args[opcode] == 2) {
+      PRINTF(stream, "\n");
+      return;
+   }
+
+   PRINTF(stream, ", ");
+   print_src_reg(stream, GET_SRC2_REG(program[2]));
+   PRINTF(stream, "\n");
+   return;
+}
+
+
+static void
+print_tex_op(struct debug_stream *stream, 
+	     unsigned opcode, const unsigned * program)
+{
+   print_dest_reg(stream, program[0] | A0_DEST_CHANNEL_ALL);
+   PRINTF(stream, " = ");
+
+   PRINTF(stream, "%s ", opcodes[opcode]);
+
+   PRINTF(stream, "S[%d],", program[0] & T0_SAMPLER_NR_MASK);
+
+   print_reg_type_nr(stream, 
+		     (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) &
+                     REG_TYPE_MASK,
+                     (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK);
+   PRINTF(stream, "\n");
+}
+
+static void
+print_texkil_op(struct debug_stream *stream, 
+                unsigned opcode, const unsigned * program)
+{
+   PRINTF(stream, "TEXKIL ");
+
+   print_reg_type_nr(stream, 
+		     (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) &
+                     REG_TYPE_MASK,
+                     (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK);
+   PRINTF(stream, "\n");
+}
+
+static void
+print_dcl_op(struct debug_stream *stream, 
+	     unsigned opcode, const unsigned * program)
+{
+   PRINTF(stream, "%s ", opcodes[opcode]);
+   print_dest_reg(stream, 
+		  program[0] | A0_DEST_CHANNEL_ALL);
+   PRINTF(stream, "\n");
+}
+
+
+void
+i915_disassemble_program(struct debug_stream *stream, 
+			 const unsigned * program, unsigned sz)
+{
+   unsigned i;
+
+   PRINTF(stream, "\t\tBEGIN\n");
+
+   assert((program[0] & 0x1ff) + 2 == sz);
+
+   program++;
+   for (i = 1; i < sz; i += 3, program += 3) {
+      unsigned opcode = program[0] & (0x1f << 24);
+
+      PRINTF(stream, "\t\t");
+
+      if ((int) opcode >= A0_NOP && opcode <= A0_SLT)
+         print_arith_op(stream, opcode >> 24, program);
+      else if (opcode >= T0_TEXLD && opcode < T0_TEXKILL)
+         print_tex_op(stream, opcode >> 24, program);
+      else if (opcode == T0_TEXKILL)
+         print_texkil_op(stream, opcode >> 24, program);
+      else if (opcode == D0_DCL)
+         print_dcl_op(stream, opcode >> 24, program);
+      else
+         PRINTF(stream, "Unknown opcode 0x%x\n", opcode);
+   }
+
+   PRINTF(stream, "\t\tEND\n\n");
+}
+
+
diff --git a/src/gallium/drivers/i915simple/i915_flush.c b/src/gallium/drivers/i915simple/i915_flush.c
new file mode 100644
index 0000000000..472e0ab774
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_flush.c
@@ -0,0 +1,78 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_defines.h"
+#include "draw/draw_context.h"
+#include "i915_context.h"
+#include "i915_reg.h"
+#include "i915_batch.h"
+
+
+static void i915_flush( struct pipe_context *pipe,
+                        unsigned flags,
+                        struct pipe_fence_handle **fence )
+{
+   struct i915_context *i915 = i915_context(pipe);
+
+   draw_flush(i915->draw);
+
+   /* Do we need to emit an MI_FLUSH command to flush the hardware
+    * caches?
+    */
+   if (flags & (PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE)) {
+      unsigned flush = MI_FLUSH;
+      
+      if (!(flags & PIPE_FLUSH_RENDER_CACHE))
+	 flush |= INHIBIT_FLUSH_RENDER_CACHE;
+
+      if (flags & PIPE_FLUSH_TEXTURE_CACHE)
+	 flush |= FLUSH_MAP_CACHE;
+
+      if (!BEGIN_BATCH(1, 0)) {
+	 FLUSH_BATCH(NULL);
+	 assert(BEGIN_BATCH(1, 0));
+      }
+      OUT_BATCH( flush );
+   }
+
+   /* If there are no flags, just flush pending commands to hardware:
+    */
+   FLUSH_BATCH(fence);
+   i915->vbo_flushed = 1;
+}
+
+
+
+void i915_init_flush_functions( struct i915_context *i915 )
+{
+   i915->pipe.flush = i915_flush;
+}
diff --git a/src/gallium/drivers/i915simple/i915_fpc.h b/src/gallium/drivers/i915simple/i915_fpc.h
new file mode 100644
index 0000000000..2f0f99d046
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_fpc.h
@@ -0,0 +1,207 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef I915_FPC_H
+#define I915_FPC_H
+
+
+#include "i915_context.h"
+#include "i915_reg.h"
+
+
+
+#define I915_PROGRAM_SIZE 192
+
+
+
+/**
+ * Program translation state
+ */
+struct i915_fp_compile {
+   struct i915_fragment_shader *shader;  /* the shader we're compiling */
+
+   boolean used_constants[I915_MAX_CONSTANT];
+
+   /** maps TGSI immediate index to constant slot */
+   uint num_immediates;
+   uint immediates_map[I915_MAX_CONSTANT];
+   float immediates[I915_MAX_CONSTANT][4];
+
+   boolean first_instruction;
+
+   uint declarations[I915_PROGRAM_SIZE];
+   uint program[I915_PROGRAM_SIZE];
+
+   uint *csr;            /**< Cursor, points into program. */
+
+   uint *decl;           /**< Cursor, points into declarations. */
+
+   uint decl_s;          /**< flags for which s regs need to be decl'd */
+   uint decl_t;          /**< flags for which t regs need to be decl'd */
+
+   uint temp_flag;       /**< Tracks temporary regs which are in use */
+   uint utemp_flag;      /**< Tracks TYPE_U temporary regs which are in use */
+
+   uint nr_tex_indirect;
+   uint nr_tex_insn;
+   uint nr_alu_insn;
+   uint nr_decl_insn;
+
+   boolean error;      /**< Set if i915_program_error() is called */
+   uint wpos_tex;
+   uint NumNativeInstructions;
+   uint NumNativeAluInstructions;
+   uint NumNativeTexInstructions;
+   uint NumNativeTexIndirections;
+};
+
+
+/* Having zero and one in here makes the definition of swizzle a lot
+ * easier.
+ */
+#define UREG_TYPE_SHIFT               29
+#define UREG_NR_SHIFT                 24
+#define UREG_CHANNEL_X_NEGATE_SHIFT   23
+#define UREG_CHANNEL_X_SHIFT          20
+#define UREG_CHANNEL_Y_NEGATE_SHIFT   19
+#define UREG_CHANNEL_Y_SHIFT          16
+#define UREG_CHANNEL_Z_NEGATE_SHIFT   15
+#define UREG_CHANNEL_Z_SHIFT          12
+#define UREG_CHANNEL_W_NEGATE_SHIFT   11
+#define UREG_CHANNEL_W_SHIFT          8
+#define UREG_CHANNEL_ZERO_NEGATE_MBZ  5
+#define UREG_CHANNEL_ZERO_SHIFT       4
+#define UREG_CHANNEL_ONE_NEGATE_MBZ   1
+#define UREG_CHANNEL_ONE_SHIFT        0
+
+#define UREG_BAD          0xffffffff    /* not a valid ureg */
+
+#define X    SRC_X
+#define Y    SRC_Y
+#define Z    SRC_Z
+#define W    SRC_W
+#define ZERO SRC_ZERO
+#define ONE  SRC_ONE
+
+/* Construct a ureg:
+ */
+#define UREG( type, nr ) (((type)<< UREG_TYPE_SHIFT) |		\
+			  ((nr)  << UREG_NR_SHIFT) |		\
+			  (X     << UREG_CHANNEL_X_SHIFT) |	\
+			  (Y     << UREG_CHANNEL_Y_SHIFT) |	\
+			  (Z     << UREG_CHANNEL_Z_SHIFT) |	\
+			  (W     << UREG_CHANNEL_W_SHIFT) |	\
+			  (ZERO  << UREG_CHANNEL_ZERO_SHIFT) |	\
+			  (ONE   << UREG_CHANNEL_ONE_SHIFT))
+
+#define GET_CHANNEL_SRC( reg, channel ) ((reg<<(channel*4)) & (0xf<<20))
+#define CHANNEL_SRC( src, channel ) (src>>(channel*4))
+
+#define GET_UREG_TYPE(reg) (((reg)>>UREG_TYPE_SHIFT)&REG_TYPE_MASK)
+#define GET_UREG_NR(reg)   (((reg)>>UREG_NR_SHIFT)&REG_NR_MASK)
+
+
+
+#define UREG_XYZW_CHANNEL_MASK 0x00ffff00
+
+/* One neat thing about the UREG representation:  
+ */
+static INLINE int
+swizzle(int reg, uint x, uint y, uint z, uint w)
+{
+   assert(x <= SRC_ONE);
+   assert(y <= SRC_ONE);
+   assert(z <= SRC_ONE);
+   assert(w <= SRC_ONE);
+   return ((reg & ~UREG_XYZW_CHANNEL_MASK) |
+           CHANNEL_SRC(GET_CHANNEL_SRC(reg, x), 0) |
+           CHANNEL_SRC(GET_CHANNEL_SRC(reg, y), 1) |
+           CHANNEL_SRC(GET_CHANNEL_SRC(reg, z), 2) |
+           CHANNEL_SRC(GET_CHANNEL_SRC(reg, w), 3));
+}
+
+
+
+/***********************************************************************
+ * Public interface for the compiler
+ */
+extern void
+i915_translate_fragment_program( struct i915_context *i915,
+                                 struct i915_fragment_shader *fs);
+
+
+
+extern uint i915_get_temp(struct i915_fp_compile *p);
+extern uint i915_get_utemp(struct i915_fp_compile *p);
+extern void i915_release_utemps(struct i915_fp_compile *p);
+
+
+extern uint i915_emit_texld(struct i915_fp_compile *p,
+                              uint dest,
+                              uint destmask,
+                              uint sampler, uint coord, uint op);
+
+extern uint i915_emit_arith(struct i915_fp_compile *p,
+                              uint op,
+                              uint dest,
+                              uint mask,
+                              uint saturate,
+                              uint src0, uint src1, uint src2);
+
+extern uint i915_emit_decl(struct i915_fp_compile *p,
+                             uint type, uint nr, uint d0_flags);
+
+
+extern uint i915_emit_const1f(struct i915_fp_compile *p, float c0);
+
+extern uint i915_emit_const2f(struct i915_fp_compile *p,
+                                float c0, float c1);
+
+extern uint i915_emit_const4fv(struct i915_fp_compile *p,
+                                 const float * c);
+
+extern uint i915_emit_const4f(struct i915_fp_compile *p,
+                                float c0, float c1,
+                                float c2, float c3);
+
+
+/*======================================================================
+ * i915_fpc_debug.c
+ */
+extern void i915_disassemble_program(const uint * program, uint sz);
+
+
+/*======================================================================
+ * i915_fpc_translate.c
+ */
+
+extern void
+i915_program_error(struct i915_fp_compile *p, const char *msg, ...);
+
+
+#endif
diff --git a/src/gallium/drivers/i915simple/i915_fpc_emit.c b/src/gallium/drivers/i915simple/i915_fpc_emit.c
new file mode 100644
index 0000000000..b054ce41d3
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_fpc_emit.c
@@ -0,0 +1,375 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "i915_reg.h"
+#include "i915_context.h"
+#include "i915_fpc.h"
+#include "util/u_math.h"
+
+
+#define A0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT)
+#define D0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT)
+#define T0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT)
+#define A0_SRC0( reg ) (((reg)&UREG_MASK)>>UREG_A0_SRC0_SHIFT_LEFT)
+#define A1_SRC0( reg ) (((reg)&UREG_MASK)<<UREG_A1_SRC0_SHIFT_RIGHT)
+#define A1_SRC1( reg ) (((reg)&UREG_MASK)>>UREG_A1_SRC1_SHIFT_LEFT)
+#define A2_SRC1( reg ) (((reg)&UREG_MASK)<<UREG_A2_SRC1_SHIFT_RIGHT)
+#define A2_SRC2( reg ) (((reg)&UREG_MASK)>>UREG_A2_SRC2_SHIFT_LEFT)
+
+/* These are special, and don't have swizzle/negate bits.
+ */
+#define T0_SAMPLER( reg )     (GET_UREG_NR(reg)<<T0_SAMPLER_NR_SHIFT)
+#define T1_ADDRESS_REG( reg ) ((GET_UREG_NR(reg)<<T1_ADDRESS_REG_NR_SHIFT) | \
+			       (GET_UREG_TYPE(reg)<<T1_ADDRESS_REG_TYPE_SHIFT))
+
+
+/* Macros for translating UREG's into the various register fields used
+ * by the I915 programmable unit.
+ */
+#define UREG_A0_DEST_SHIFT_LEFT  (UREG_TYPE_SHIFT - A0_DEST_TYPE_SHIFT)
+#define UREG_A0_SRC0_SHIFT_LEFT  (UREG_TYPE_SHIFT - A0_SRC0_TYPE_SHIFT)
+#define UREG_A1_SRC0_SHIFT_RIGHT (A1_SRC0_CHANNEL_W_SHIFT - UREG_CHANNEL_W_SHIFT)
+#define UREG_A1_SRC1_SHIFT_LEFT  (UREG_TYPE_SHIFT - A1_SRC1_TYPE_SHIFT)
+#define UREG_A2_SRC1_SHIFT_RIGHT (A2_SRC1_CHANNEL_W_SHIFT - UREG_CHANNEL_W_SHIFT)
+#define UREG_A2_SRC2_SHIFT_LEFT  (UREG_TYPE_SHIFT - A2_SRC2_TYPE_SHIFT)
+
+#define UREG_MASK         0xffffff00
+#define UREG_TYPE_NR_MASK ((REG_TYPE_MASK << UREG_TYPE_SHIFT) | \
+  			   (REG_NR_MASK << UREG_NR_SHIFT))
+
+
+uint
+i915_get_temp(struct i915_fp_compile *p)
+{
+   int bit = ffs(~p->temp_flag);
+   if (!bit) {
+      i915_program_error(p, "i915_get_temp: out of temporaries\n");
+      return 0;
+   }
+
+   p->temp_flag |= 1 << (bit - 1);
+   return bit - 1;
+}
+
+
+static void
+i915_release_temp(struct i915_fp_compile *p, int reg)
+{
+   p->temp_flag &= ~(1 << reg);
+}
+
+
+/**
+ * Get unpreserved temporary, a temp whose value is not preserved between
+ * PS program phases.
+ */
+uint
+i915_get_utemp(struct i915_fp_compile * p)
+{
+   int bit = ffs(~p->utemp_flag);
+   if (!bit) {
+      i915_program_error(p, "i915_get_utemp: out of temporaries\n");
+      return 0;
+   }
+
+   p->utemp_flag |= 1 << (bit - 1);
+   return UREG(REG_TYPE_U, (bit - 1));
+}
+
+void
+i915_release_utemps(struct i915_fp_compile *p)
+{
+   p->utemp_flag = ~0x7;
+}
+
+
+uint
+i915_emit_decl(struct i915_fp_compile *p,
+               uint type, uint nr, uint d0_flags)
+{
+   uint reg = UREG(type, nr);
+
+   if (type == REG_TYPE_T) {
+      if (p->decl_t & (1 << nr))
+         return reg;
+
+      p->decl_t |= (1 << nr);
+   }
+   else if (type == REG_TYPE_S) {
+      if (p->decl_s & (1 << nr))
+         return reg;
+
+      p->decl_s |= (1 << nr);
+   }
+   else
+      return reg;
+
+   *(p->decl++) = (D0_DCL | D0_DEST(reg) | d0_flags);
+   *(p->decl++) = D1_MBZ;
+   *(p->decl++) = D2_MBZ;
+
+   p->nr_decl_insn++;
+   return reg;
+}
+
+uint
+i915_emit_arith(struct i915_fp_compile * p,
+                uint op,
+                uint dest,
+                uint mask,
+                uint saturate, uint src0, uint src1, uint src2)
+{
+   uint c[3];
+   uint nr_const = 0;
+
+   assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST);
+   dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest));
+   assert(dest);
+
+   if (GET_UREG_TYPE(src0) == REG_TYPE_CONST)
+      c[nr_const++] = 0;
+   if (GET_UREG_TYPE(src1) == REG_TYPE_CONST)
+      c[nr_const++] = 1;
+   if (GET_UREG_TYPE(src2) == REG_TYPE_CONST)
+      c[nr_const++] = 2;
+
+   /* Recursively call this function to MOV additional const values
+    * into temporary registers.  Use utemp registers for this -
+    * currently shouldn't be possible to run out, but keep an eye on
+    * this.
+    */
+   if (nr_const > 1) {
+      uint s[3], first, i, old_utemp_flag;
+
+      s[0] = src0;
+      s[1] = src1;
+      s[2] = src2;
+      old_utemp_flag = p->utemp_flag;
+
+      first = GET_UREG_NR(s[c[0]]);
+      for (i = 1; i < nr_const; i++) {
+         if (GET_UREG_NR(s[c[i]]) != first) {
+            uint tmp = i915_get_utemp(p);
+
+            i915_emit_arith(p, A0_MOV, tmp, A0_DEST_CHANNEL_ALL, 0,
+                            s[c[i]], 0, 0);
+            s[c[i]] = tmp;
+         }
+      }
+
+      src0 = s[0];
+      src1 = s[1];
+      src2 = s[2];
+      p->utemp_flag = old_utemp_flag;   /* restore */
+   }
+
+   *(p->csr++) = (op | A0_DEST(dest) | mask | saturate | A0_SRC0(src0));
+   *(p->csr++) = (A1_SRC0(src0) | A1_SRC1(src1));
+   *(p->csr++) = (A2_SRC1(src1) | A2_SRC2(src2));
+
+   p->nr_alu_insn++;
+   return dest;
+}
+
+
+/**
+ * Emit a texture load or texkill instruction.
+ * \param dest  the dest i915 register
+ * \param destmask  the dest register writemask
+ * \param sampler  the i915 sampler register
+ * \param coord  the i915 source texcoord operand
+ * \param opcode  the instruction opcode
+ */
+uint i915_emit_texld( struct i915_fp_compile *p,
+			uint dest,
+			uint destmask,
+			uint sampler,
+			uint coord,
+			uint opcode )
+{
+   const uint k = UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord));
+   int temp = -1;
+
+   if (coord != k) {
+      /* texcoord is swizzled or negated.  Need to allocate a new temporary
+       * register (a utemp / unpreserved temp) won't do.
+       */
+      uint tempReg;
+
+      temp = i915_get_temp(p);           /* get temp reg index */
+      tempReg = UREG(REG_TYPE_R, temp);  /* make i915 register */
+
+      i915_emit_arith( p, A0_MOV,
+                       tempReg, A0_DEST_CHANNEL_ALL, /* dest reg, writemask */
+                       0,                            /* saturate */
+                       coord, 0, 0 );                /* src0, src1, src2 */
+
+      /* new src texcoord is tempReg */
+      coord = tempReg;
+   }
+
+   /* Don't worry about saturate as we only support  
+    */
+   if (destmask != A0_DEST_CHANNEL_ALL) {
+      /* if not writing to XYZW... */
+      uint tmp = i915_get_utemp(p);
+      i915_emit_texld( p, tmp, A0_DEST_CHANNEL_ALL, sampler, coord, opcode );
+      i915_emit_arith( p, A0_MOV, dest, destmask, 0, tmp, 0, 0 );
+      /* XXX release utemp here? */
+   }
+   else {
+      assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST);
+      assert(dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest)));
+
+      /* is the sampler coord a texcoord input reg? */
+      if (GET_UREG_TYPE(coord) != REG_TYPE_T) {
+	 p->nr_tex_indirect++;
+      }
+
+      *(p->csr++) = (opcode | 
+		     T0_DEST( dest ) |
+		     T0_SAMPLER( sampler ));
+
+      *(p->csr++) = T1_ADDRESS_REG( coord );
+      *(p->csr++) = T2_MBZ;
+
+      p->nr_tex_insn++;
+   }
+
+   if (temp >= 0)
+      i915_release_temp(p, temp);
+
+   return dest;
+}
+
+
+uint
+i915_emit_const1f(struct i915_fp_compile * p, float c0)
+{
+   struct i915_fragment_shader *ifs = p->shader;
+   unsigned reg, idx;
+
+   if (c0 == 0.0)
+      return swizzle(UREG(REG_TYPE_R, 0), ZERO, ZERO, ZERO, ZERO);
+   if (c0 == 1.0)
+      return swizzle(UREG(REG_TYPE_R, 0), ONE, ONE, ONE, ONE);
+
+   for (reg = 0; reg < I915_MAX_CONSTANT; reg++) {
+      if (ifs->constant_flags[reg] == I915_CONSTFLAG_USER)
+         continue;
+      for (idx = 0; idx < 4; idx++) {
+         if (!(ifs->constant_flags[reg] & (1 << idx)) ||
+             ifs->constants[reg][idx] == c0) {
+            ifs->constants[reg][idx] = c0;
+            ifs->constant_flags[reg] |= 1 << idx;
+            if (reg + 1 > ifs->num_constants)
+               ifs->num_constants = reg + 1;
+            return swizzle(UREG(REG_TYPE_CONST, reg), idx, ZERO, ZERO, ONE);
+         }
+      }
+   }
+
+   i915_program_error(p, "i915_emit_const1f: out of constants\n");
+   return 0;
+}
+
+uint
+i915_emit_const2f(struct i915_fp_compile * p, float c0, float c1)
+{
+   struct i915_fragment_shader *ifs = p->shader;
+   unsigned reg, idx;
+
+   if (c0 == 0.0)
+      return swizzle(i915_emit_const1f(p, c1), ZERO, X, Z, W);
+   if (c0 == 1.0)
+      return swizzle(i915_emit_const1f(p, c1), ONE, X, Z, W);
+
+   if (c1 == 0.0)
+      return swizzle(i915_emit_const1f(p, c0), X, ZERO, Z, W);
+   if (c1 == 1.0)
+      return swizzle(i915_emit_const1f(p, c0), X, ONE, Z, W);
+
+   for (reg = 0; reg < I915_MAX_CONSTANT; reg++) {
+      if (ifs->constant_flags[reg] == 0xf ||
+          ifs->constant_flags[reg] == I915_CONSTFLAG_USER)
+         continue;
+      for (idx = 0; idx < 3; idx++) {
+         if (!(ifs->constant_flags[reg] & (3 << idx))) {
+            ifs->constants[reg][idx + 0] = c0;
+            ifs->constants[reg][idx + 1] = c1;
+            ifs->constant_flags[reg] |= 3 << idx;
+            if (reg + 1 > ifs->num_constants)
+               ifs->num_constants = reg + 1;
+            return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, ONE);
+         }
+      }
+   }
+
+   i915_program_error(p, "i915_emit_const2f: out of constants\n");
+   return 0;
+}
+
+
+
+uint
+i915_emit_const4f(struct i915_fp_compile * p,
+                  float c0, float c1, float c2, float c3)
+{
+   struct i915_fragment_shader *ifs = p->shader;
+   unsigned reg;
+
+   for (reg = 0; reg < I915_MAX_CONSTANT; reg++) {
+      if (ifs->constant_flags[reg] == 0xf &&
+          ifs->constants[reg][0] == c0 &&
+          ifs->constants[reg][1] == c1 &&
+          ifs->constants[reg][2] == c2 &&
+          ifs->constants[reg][3] == c3) {
+         return UREG(REG_TYPE_CONST, reg);
+      }
+      else if (ifs->constant_flags[reg] == 0) {
+
+         ifs->constants[reg][0] = c0;
+         ifs->constants[reg][1] = c1;
+         ifs->constants[reg][2] = c2;
+         ifs->constants[reg][3] = c3;
+         ifs->constant_flags[reg] = 0xf;
+         if (reg + 1 > ifs->num_constants)
+            ifs->num_constants = reg + 1;
+         return UREG(REG_TYPE_CONST, reg);
+      }
+   }
+
+   i915_program_error(p, "i915_emit_const4f: out of constants\n");
+   return 0;
+}
+
+
+uint
+i915_emit_const4fv(struct i915_fp_compile * p, const float * c)
+{
+   return i915_emit_const4f(p, c[0], c[1], c[2], c[3]);
+}
diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c
new file mode 100644
index 0000000000..d92bdc1bc6
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c
@@ -0,0 +1,1190 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include <stdarg.h>
+
+#include "i915_reg.h"
+#include "i915_context.h"
+#include "i915_fpc.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "util/u_string.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "draw/draw_vertex.h"
+
+
+/**
+ * Simple pass-through fragment shader to use when we don't have
+ * a real shader (or it fails to compile for some reason).
+ */
+static unsigned passthrough[] = 
+{
+   _3DSTATE_PIXEL_SHADER_PROGRAM | ((2*3)-1),
+
+   /* declare input color:
+    */
+   (D0_DCL | 
+    (REG_TYPE_T << D0_TYPE_SHIFT) | 
+    (T_DIFFUSE << D0_NR_SHIFT) | 
+    D0_CHANNEL_ALL),
+   0,
+   0,
+
+   /* move to output color:
+    */
+   (A0_MOV | 
+    (REG_TYPE_OC << A0_DEST_TYPE_SHIFT) | 
+    A0_DEST_CHANNEL_ALL | 
+    (REG_TYPE_T << A0_SRC0_TYPE_SHIFT) |
+    (T_DIFFUSE << A0_SRC0_NR_SHIFT)),
+   0x01230000,			/* .xyzw */
+   0
+};
+
+
+/* 1, -1/3!, 1/5!, -1/7! */
+static const float sin_constants[4] = { 1.0,
+   -1.0f / (3 * 2 * 1),
+   1.0f / (5 * 4 * 3 * 2 * 1),
+   -1.0f / (7 * 6 * 5 * 4 * 3 * 2 * 1)
+};
+
+/* 1, -1/2!, 1/4!, -1/6! */
+static const float cos_constants[4] = { 1.0,
+   -1.0f / (2 * 1),
+   1.0f / (4 * 3 * 2 * 1),
+   -1.0f / (6 * 5 * 4 * 3 * 2 * 1)
+};
+
+
+
+/**
+ * component-wise negation of ureg
+ */
+static INLINE int
+negate(int reg, int x, int y, int z, int w)
+{
+   /* Another neat thing about the UREG representation */
+   return reg ^ (((x & 1) << UREG_CHANNEL_X_NEGATE_SHIFT) |
+                 ((y & 1) << UREG_CHANNEL_Y_NEGATE_SHIFT) |
+                 ((z & 1) << UREG_CHANNEL_Z_NEGATE_SHIFT) |
+                 ((w & 1) << UREG_CHANNEL_W_NEGATE_SHIFT));
+}
+
+
+/**
+ * In the event of a translation failure, we'll generate a simple color
+ * pass-through program.
+ */
+static void
+i915_use_passthrough_shader(struct i915_fragment_shader *fs)
+{
+   fs->program = (uint *) MALLOC(sizeof(passthrough));
+   if (fs->program) {
+      memcpy(fs->program, passthrough, sizeof(passthrough));
+      fs->program_len = Elements(passthrough);
+   }
+   fs->num_constants = 0;
+}
+
+
+void
+i915_program_error(struct i915_fp_compile *p, const char *msg, ...)
+{
+   va_list args;
+   char buffer[1024];
+
+   debug_printf("i915_program_error: ");
+   va_start( args, msg );  
+   util_vsnprintf( buffer, sizeof(buffer), msg, args );
+   va_end( args );
+   debug_printf(buffer);
+   debug_printf("\n");
+
+   p->error = 1;
+}
+
+
+
+/**
+ * Construct a ureg for the given source register.  Will emit
+ * constants, apply swizzling and negation as needed.
+ */
+static uint
+src_vector(struct i915_fp_compile *p,
+           const struct tgsi_full_src_register *source)
+{
+   uint index = source->SrcRegister.Index;
+   uint src = 0, sem_name, sem_ind;
+
+   switch (source->SrcRegister.File) {
+   case TGSI_FILE_TEMPORARY:
+      if (source->SrcRegister.Index >= I915_MAX_TEMPORARY) {
+         i915_program_error(p, "Exceeded max temporary reg");
+         return 0;
+      }
+      src = UREG(REG_TYPE_R, index);
+      break;
+   case TGSI_FILE_INPUT:
+      /* XXX: Packing COL1, FOGC into a single attribute works for
+       * texenv programs, but will fail for real fragment programs
+       * that use these attributes and expect them to be a full 4
+       * components wide.  Could use a texcoord to pass these
+       * attributes if necessary, but that won't work in the general
+       * case.
+       * 
+       * We also use a texture coordinate to pass wpos when possible.
+       */
+
+      sem_name = p->shader->info.input_semantic_name[index];
+      sem_ind = p->shader->info.input_semantic_index[index];
+
+      switch (sem_name) {
+      case TGSI_SEMANTIC_POSITION:
+         debug_printf("SKIP SEM POS\n");
+         /*
+         assert(p->wpos_tex != -1);
+         src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL);
+         */
+         break;
+      case TGSI_SEMANTIC_COLOR:
+         if (sem_ind == 0) {
+            src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL);
+         }
+         else {
+            /* secondary color */
+            assert(sem_ind == 1);
+            src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ);
+            src = swizzle(src, X, Y, Z, ONE);
+         }
+         break;
+      case TGSI_SEMANTIC_FOG:
+         src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W);
+         src = swizzle(src, W, W, W, W);
+         break;
+      case TGSI_SEMANTIC_GENERIC:
+         /* usually a texcoord */
+         src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + sem_ind, D0_CHANNEL_ALL);
+         break;
+      default:
+         i915_program_error(p, "Bad source->Index");
+         return 0;
+      }
+      break;
+
+   case TGSI_FILE_IMMEDIATE:
+      assert(index < p->num_immediates);
+      index = p->immediates_map[index];
+      /* fall-through */
+   case TGSI_FILE_CONSTANT:
+      src = UREG(REG_TYPE_CONST, index);
+      break;
+
+   default:
+      i915_program_error(p, "Bad source->File");
+      return 0;
+   }
+
+   if (source->SrcRegister.Extended) {
+      src = swizzle(src,
+                    source->SrcRegisterExtSwz.ExtSwizzleX,
+                    source->SrcRegisterExtSwz.ExtSwizzleY,
+                    source->SrcRegisterExtSwz.ExtSwizzleZ,
+                    source->SrcRegisterExtSwz.ExtSwizzleW);
+   }
+   else {
+      src = swizzle(src,
+                    source->SrcRegister.SwizzleX,
+                    source->SrcRegister.SwizzleY,
+                    source->SrcRegister.SwizzleZ,
+                    source->SrcRegister.SwizzleW);
+   }
+
+
+   /* There's both negate-all-components and per-component negation.
+    * Try to handle both here.
+    */
+   {
+      int nx = source->SrcRegisterExtSwz.NegateX;
+      int ny = source->SrcRegisterExtSwz.NegateY;
+      int nz = source->SrcRegisterExtSwz.NegateZ;
+      int nw = source->SrcRegisterExtSwz.NegateW;
+      if (source->SrcRegister.Negate) {
+         nx = !nx;
+         ny = !ny;
+         nz = !nz;
+         nw = !nw;
+      }
+      src = negate(src, nx, ny, nz, nw);
+   }
+
+   /* no abs() or post-abs negation */
+#if 0
+   /* XXX assertions disabled to allow arbfplight.c to run */
+   /* XXX enable these assertions, or fix things */
+   assert(!source->SrcRegisterExtMod.Absolute);
+   assert(!source->SrcRegisterExtMod.Negate);
+#endif
+   return src;
+}
+
+
+/**
+ * Construct a ureg for a destination register.
+ */
+static uint
+get_result_vector(struct i915_fp_compile *p,
+                  const struct tgsi_full_dst_register *dest)
+{
+   switch (dest->DstRegister.File) {
+   case TGSI_FILE_OUTPUT:
+      {
+         uint sem_name = p->shader->info.output_semantic_name[dest->DstRegister.Index];
+         switch (sem_name) {
+         case TGSI_SEMANTIC_POSITION:
+            return UREG(REG_TYPE_OD, 0);
+         case TGSI_SEMANTIC_COLOR:
+            return UREG(REG_TYPE_OC, 0);
+         default:
+            i915_program_error(p, "Bad inst->DstReg.Index/semantics");
+            return 0;
+         }
+      }
+   case TGSI_FILE_TEMPORARY:
+      return UREG(REG_TYPE_R, dest->DstRegister.Index);
+   default:
+      i915_program_error(p, "Bad inst->DstReg.File");
+      return 0;
+   }
+}
+
+
+/**
+ * Compute flags for saturation and writemask.
+ */
+static uint
+get_result_flags(const struct tgsi_full_instruction *inst)
+{
+   const uint writeMask
+      = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   uint flags = 0x0;
+
+   if (inst->Instruction.Saturate == TGSI_SAT_ZERO_ONE)
+      flags |= A0_DEST_SATURATE;
+
+   if (writeMask & TGSI_WRITEMASK_X)
+      flags |= A0_DEST_CHANNEL_X;
+   if (writeMask & TGSI_WRITEMASK_Y)
+      flags |= A0_DEST_CHANNEL_Y;
+   if (writeMask & TGSI_WRITEMASK_Z)
+      flags |= A0_DEST_CHANNEL_Z;
+   if (writeMask & TGSI_WRITEMASK_W)
+      flags |= A0_DEST_CHANNEL_W;
+
+   return flags;
+}
+
+
+/**
+ * Convert TGSI_TEXTURE_x token to DO_SAMPLE_TYPE_x token
+ */
+static uint
+translate_tex_src_target(struct i915_fp_compile *p, uint tex)
+{
+   switch (tex) {
+   case TGSI_TEXTURE_1D:
+      return D0_SAMPLE_TYPE_2D;
+   case TGSI_TEXTURE_2D:
+      return D0_SAMPLE_TYPE_2D;
+   case TGSI_TEXTURE_RECT:
+      return D0_SAMPLE_TYPE_2D;
+   case TGSI_TEXTURE_3D:
+      return D0_SAMPLE_TYPE_VOLUME;
+   case TGSI_TEXTURE_CUBE:
+      return D0_SAMPLE_TYPE_CUBE;
+   default:
+      i915_program_error(p, "TexSrc type");
+      return 0;
+   }
+}
+
+
+/**
+ * Generate texel lookup instruction.
+ */
+static void
+emit_tex(struct i915_fp_compile *p,
+         const struct tgsi_full_instruction *inst,
+         uint opcode)
+{
+   uint texture = inst->InstructionExtTexture.Texture;
+   uint unit = inst->FullSrcRegisters[1].SrcRegister.Index;
+   uint tex = translate_tex_src_target( p, texture );
+   uint sampler = i915_emit_decl(p, REG_TYPE_S, unit, tex);
+   uint coord = src_vector( p, &inst->FullSrcRegisters[0]);
+
+   i915_emit_texld( p,
+                    get_result_vector( p, &inst->FullDstRegisters[0] ),
+                    get_result_flags( inst ),
+                    sampler,
+                    coord,
+                    opcode);
+}
+
+
+/**
+ * Generate a simple arithmetic instruction
+ * \param opcode  the i915 opcode
+ * \param numArgs  the number of input/src arguments
+ */
+static void
+emit_simple_arith(struct i915_fp_compile *p,
+                  const struct tgsi_full_instruction *inst,
+                  uint opcode, uint numArgs)
+{
+   uint arg1, arg2, arg3;
+
+   assert(numArgs <= 3);
+
+   arg1 = (numArgs < 1) ? 0 : src_vector( p, &inst->FullSrcRegisters[0] );
+   arg2 = (numArgs < 2) ? 0 : src_vector( p, &inst->FullSrcRegisters[1] );
+   arg3 = (numArgs < 3) ? 0 : src_vector( p, &inst->FullSrcRegisters[2] );
+
+   i915_emit_arith( p,
+                    opcode,
+                    get_result_vector( p, &inst->FullDstRegisters[0]),
+                    get_result_flags( inst ), 0,
+                    arg1,
+                    arg2,
+                    arg3 );
+}
+
+
+/** As above, but swap the first two src regs */
+static void
+emit_simple_arith_swap2(struct i915_fp_compile *p,
+                        const struct tgsi_full_instruction *inst,
+                        uint opcode, uint numArgs)
+{
+   struct tgsi_full_instruction inst2;
+
+   assert(numArgs == 2);
+
+   /* transpose first two registers */
+   inst2 = *inst;
+   inst2.FullSrcRegisters[0] = inst->FullSrcRegisters[1];
+   inst2.FullSrcRegisters[1] = inst->FullSrcRegisters[0];
+
+   emit_simple_arith(p, &inst2, opcode, numArgs);
+}
+
+
+#ifndef M_PI
+#define M_PI 3.14159265358979323846
+#endif
+
+/*
+ * Translate TGSI instruction to i915 instruction.
+ *
+ * Possible concerns:
+ *
+ * SIN, COS -- could use another taylor step?
+ * LIT      -- results seem a little different to sw mesa
+ * LOG      -- different to mesa on negative numbers, but this is conformant.
+ */ 
+static void
+i915_translate_instruction(struct i915_fp_compile *p,
+                           const struct tgsi_full_instruction *inst)
+{
+   uint writemask;
+   uint src0, src1, src2, flags;
+   uint tmp = 0;
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_ABS:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      i915_emit_arith(p,
+                      A0_MAX,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      src0, negate(src0, 1, 1, 1, 1), 0);
+      break;
+
+   case TGSI_OPCODE_ADD:
+      emit_simple_arith(p, inst, A0_ADD, 2);
+      break;
+
+   case TGSI_OPCODE_CMP:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+      src2 = src_vector(p, &inst->FullSrcRegisters[2]);
+      i915_emit_arith(p, A0_CMP, 
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 
+                      0, src0, src2, src1);   /* NOTE: order of src2, src1 */
+      break;
+
+   case TGSI_OPCODE_COS:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      tmp = i915_get_utemp(p);
+
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_X, 0,
+                      src0, i915_emit_const1f(p, 1.0f / (float) (M_PI * 2.0)), 0);
+
+      i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0);
+
+      /* By choosing different taylor constants, could get rid of this mul:
+       */
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_X, 0,
+                      tmp, i915_emit_const1f(p, (float) (M_PI * 2.0)), 0);
+
+      /* 
+       * t0.xy = MUL x.xx11, x.x1111  ; x^2, x, 1, 1
+       * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, 1
+       * t0 = MUL t0.xxz1 t0.z111    ; x^6 x^4 x^2 1
+       * result = DP4 t0, cos_constants
+       */
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_XY, 0,
+                      swizzle(tmp, X, X, ONE, ONE),
+                      swizzle(tmp, X, ONE, ONE, ONE), 0);
+
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_XYZ, 0,
+                      swizzle(tmp, X, Y, X, ONE),
+                      swizzle(tmp, X, X, ONE, ONE), 0);
+
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_XYZ, 0,
+                      swizzle(tmp, X, X, Z, ONE),
+                      swizzle(tmp, Z, ONE, ONE, ONE), 0);
+
+      i915_emit_arith(p,
+                      A0_DP4,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      swizzle(tmp, ONE, Z, Y, X),
+                      i915_emit_const4fv(p, cos_constants), 0);
+      break;
+
+   case TGSI_OPCODE_DP3:
+      emit_simple_arith(p, inst, A0_DP3, 2);
+      break;
+
+   case TGSI_OPCODE_DP4:
+      emit_simple_arith(p, inst, A0_DP4, 2);
+      break;
+
+   case TGSI_OPCODE_DPH:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+
+      i915_emit_arith(p,
+                      A0_DP4,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      swizzle(src0, X, Y, Z, ONE), src1, 0);
+      break;
+
+   case TGSI_OPCODE_DST:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+
+      /* result[0] = 1    * 1;
+       * result[1] = a[1] * b[1];
+       * result[2] = a[2] * 1;
+       * result[3] = 1    * b[3];
+       */
+      i915_emit_arith(p,
+                      A0_MUL,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      swizzle(src0, ONE, Y, Z, ONE),
+                      swizzle(src1, ONE, Y, ONE, W), 0);
+      break;
+
+   case TGSI_OPCODE_END:
+      /* no-op */
+      break;
+
+   case TGSI_OPCODE_EX2:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+
+      i915_emit_arith(p,
+                      A0_EXP,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      swizzle(src0, X, X, X, X), 0, 0);
+      break;
+
+   case TGSI_OPCODE_FLR:
+      emit_simple_arith(p, inst, A0_FLR, 1);
+      break;
+
+   case TGSI_OPCODE_FRC:
+      emit_simple_arith(p, inst, A0_FRC, 1);
+      break;
+
+   case TGSI_OPCODE_KIL:
+      /* kill if src[0].x < 0 || src[0].y < 0 ... */
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      tmp = i915_get_utemp(p);
+
+      i915_emit_texld(p,
+                      tmp,                   /* dest reg: a dummy reg */
+                      A0_DEST_CHANNEL_ALL,   /* dest writemask */
+                      0,                     /* sampler */
+                      src0,                  /* coord*/
+                      T0_TEXKILL);           /* opcode */
+      break;
+
+   case TGSI_OPCODE_KILP:
+      assert(0); /* not tested yet */
+      break;
+
+   case TGSI_OPCODE_LG2:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+
+      i915_emit_arith(p,
+                      A0_LOG,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      swizzle(src0, X, X, X, X), 0, 0);
+      break;
+
+   case TGSI_OPCODE_LIT:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      tmp = i915_get_utemp(p);
+
+      /* tmp = max( a.xyzw, a.00zw )
+       * XXX: Clamp tmp.w to -128..128
+       * tmp.y = log(tmp.y)
+       * tmp.y = tmp.w * tmp.y
+       * tmp.y = exp(tmp.y)
+       * result = cmp (a.11-x1, a.1x01, a.1xy1 )
+       */
+      i915_emit_arith(p, A0_MAX, tmp, A0_DEST_CHANNEL_ALL, 0,
+                      src0, swizzle(src0, ZERO, ZERO, Z, W), 0);
+
+      i915_emit_arith(p, A0_LOG, tmp, A0_DEST_CHANNEL_Y, 0,
+                      swizzle(tmp, Y, Y, Y, Y), 0, 0);
+
+      i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_Y, 0,
+                      swizzle(tmp, ZERO, Y, ZERO, ZERO),
+                      swizzle(tmp, ZERO, W, ZERO, ZERO), 0);
+
+      i915_emit_arith(p, A0_EXP, tmp, A0_DEST_CHANNEL_Y, 0,
+                      swizzle(tmp, Y, Y, Y, Y), 0, 0);
+
+      i915_emit_arith(p, A0_CMP,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      negate(swizzle(tmp, ONE, ONE, X, ONE), 0, 0, 1, 0),
+                      swizzle(tmp, ONE, X, ZERO, ONE),
+                      swizzle(tmp, ONE, X, Y, ONE));
+
+      break;
+
+   case TGSI_OPCODE_LRP:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+      src2 = src_vector(p, &inst->FullSrcRegisters[2]);
+      flags = get_result_flags(inst);
+      tmp = i915_get_utemp(p);
+
+      /* b*a + c*(1-a)
+       *
+       * b*a + c - ca 
+       *
+       * tmp = b*a + c, 
+       * result = (-c)*a + tmp 
+       */
+      i915_emit_arith(p, A0_MAD, tmp,
+                      flags & A0_DEST_CHANNEL_ALL, 0, src1, src0, src2);
+
+      i915_emit_arith(p, A0_MAD,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      flags, 0, negate(src2, 1, 1, 1, 1), src0, tmp);
+      break;
+
+   case TGSI_OPCODE_MAD:
+      emit_simple_arith(p, inst, A0_MAD, 3);
+      break;
+
+   case TGSI_OPCODE_MAX:
+      emit_simple_arith(p, inst, A0_MAX, 2);
+      break;
+
+   case TGSI_OPCODE_MIN:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+      tmp = i915_get_utemp(p);
+      flags = get_result_flags(inst);
+
+      i915_emit_arith(p,
+                      A0_MAX,
+                      tmp, flags & A0_DEST_CHANNEL_ALL, 0,
+                      negate(src0, 1, 1, 1, 1),
+                      negate(src1, 1, 1, 1, 1), 0);
+
+      i915_emit_arith(p,
+                      A0_MOV,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      flags, 0, negate(tmp, 1, 1, 1, 1), 0, 0);
+      break;
+
+   case TGSI_OPCODE_MOV:
+   case TGSI_OPCODE_SWZ:
+      emit_simple_arith(p, inst, A0_MOV, 1);
+      break;
+
+   case TGSI_OPCODE_MUL:
+      emit_simple_arith(p, inst, A0_MUL, 2);
+      break;
+
+   case TGSI_OPCODE_POW:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+      tmp = i915_get_utemp(p);
+      flags = get_result_flags(inst);
+
+      /* XXX: masking on intermediate values, here and elsewhere.
+       */
+      i915_emit_arith(p,
+                      A0_LOG,
+                      tmp, A0_DEST_CHANNEL_X, 0,
+                      swizzle(src0, X, X, X, X), 0, 0);
+
+      i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, tmp, src1, 0);
+
+      i915_emit_arith(p,
+                      A0_EXP,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      flags, 0, swizzle(tmp, X, X, X, X), 0, 0);
+      break;
+      
+   case TGSI_OPCODE_RET:
+      /* XXX: no-op? */
+      break;
+      
+   case TGSI_OPCODE_RCP:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+
+      i915_emit_arith(p,
+                      A0_RCP,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                         get_result_flags(inst), 0,
+                      swizzle(src0, X, X, X, X), 0, 0);
+      break;
+
+   case TGSI_OPCODE_RSQ:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+
+      i915_emit_arith(p,
+                      A0_RSQ,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      swizzle(src0, X, X, X, X), 0, 0);
+      break;
+
+   case TGSI_OPCODE_SCS:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      tmp = i915_get_utemp(p);
+
+      /* 
+       * t0.xy = MUL x.xx11, x.x1111  ; x^2, x, 1, 1
+       * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x
+       * t1 = MUL t0.xyyw t0.yz11    ; x^7 x^5 x^3 x
+       * scs.x = DP4 t1, sin_constants
+       * t1 = MUL t0.xxz1 t0.z111    ; x^6 x^4 x^2 1
+       * scs.y = DP4 t1, cos_constants
+       */
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_XY, 0,
+                      swizzle(src0, X, X, ONE, ONE),
+                      swizzle(src0, X, ONE, ONE, ONE), 0);
+
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_ALL, 0,
+                      swizzle(tmp, X, Y, X, Y),
+                      swizzle(tmp, X, X, ONE, ONE), 0);
+
+      writemask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+
+      if (writemask & TGSI_WRITEMASK_Y) {
+         uint tmp1;
+
+         if (writemask & TGSI_WRITEMASK_X)
+            tmp1 = i915_get_utemp(p);
+         else
+            tmp1 = tmp;
+
+         i915_emit_arith(p,
+                         A0_MUL,
+                         tmp1, A0_DEST_CHANNEL_ALL, 0,
+                         swizzle(tmp, X, Y, Y, W),
+                         swizzle(tmp, X, Z, ONE, ONE), 0);
+
+         i915_emit_arith(p,
+                         A0_DP4,
+                         get_result_vector(p, &inst->FullDstRegisters[0]),
+                         A0_DEST_CHANNEL_Y, 0,
+                         swizzle(tmp1, W, Z, Y, X),
+                         i915_emit_const4fv(p, sin_constants), 0);
+      }
+
+      if (writemask & TGSI_WRITEMASK_X) {
+         i915_emit_arith(p,
+                         A0_MUL,
+                         tmp, A0_DEST_CHANNEL_XYZ, 0,
+                         swizzle(tmp, X, X, Z, ONE),
+                         swizzle(tmp, Z, ONE, ONE, ONE), 0);
+
+         i915_emit_arith(p,
+                         A0_DP4,
+                         get_result_vector(p, &inst->FullDstRegisters[0]),
+                         A0_DEST_CHANNEL_X, 0,
+                         swizzle(tmp, ONE, Z, Y, X),
+                         i915_emit_const4fv(p, cos_constants), 0);
+      }
+      break;
+
+   case TGSI_OPCODE_SGE:
+      emit_simple_arith(p, inst, A0_SGE, 2);
+      break;
+
+   case TGSI_OPCODE_SLE:
+      /* like SGE, but swap reg0, reg1 */
+      emit_simple_arith_swap2(p, inst, A0_SGE, 2);
+      break;
+
+   case TGSI_OPCODE_SIN:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      tmp = i915_get_utemp(p);
+
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_X, 0,
+                      src0, i915_emit_const1f(p, 1.0f / (float) (M_PI * 2.0)), 0);
+
+      i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0);
+
+      /* By choosing different taylor constants, could get rid of this mul:
+       */
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_X, 0,
+                      tmp, i915_emit_const1f(p, (float) (M_PI * 2.0)), 0);
+
+      /* 
+       * t0.xy = MUL x.xx11, x.x1111  ; x^2, x, 1, 1
+       * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x
+       * t1 = MUL t0.xyyw t0.yz11    ; x^7 x^5 x^3 x
+       * result = DP4 t1.wzyx, sin_constants
+       */
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_XY, 0,
+                      swizzle(tmp, X, X, ONE, ONE),
+                      swizzle(tmp, X, ONE, ONE, ONE), 0);
+
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_ALL, 0,
+                      swizzle(tmp, X, Y, X, Y),
+                      swizzle(tmp, X, X, ONE, ONE), 0);
+
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_ALL, 0,
+                      swizzle(tmp, X, Y, Y, W),
+                      swizzle(tmp, X, Z, ONE, ONE), 0);
+
+      i915_emit_arith(p,
+                      A0_DP4,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      swizzle(tmp, W, Z, Y, X),
+                      i915_emit_const4fv(p, sin_constants), 0);
+      break;
+
+   case TGSI_OPCODE_SLT:
+      emit_simple_arith(p, inst, A0_SLT, 2);
+      break;
+
+   case TGSI_OPCODE_SGT:
+      /* like SLT, but swap reg0, reg1 */
+      emit_simple_arith_swap2(p, inst, A0_SLT, 2);
+      break;
+
+   case TGSI_OPCODE_SUB:
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+
+      i915_emit_arith(p,
+                      A0_ADD,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      src0, negate(src1, 1, 1, 1, 1), 0);
+      break;
+
+   case TGSI_OPCODE_TEX:
+      emit_tex(p, inst, T0_TEXLD);
+      break;
+
+   case TGSI_OPCODE_TXB:
+      emit_tex(p, inst, T0_TEXLDB);
+      break;
+
+   case TGSI_OPCODE_TXP:
+      emit_tex(p, inst, T0_TEXLDP);
+      break;
+
+   case TGSI_OPCODE_XPD:
+      /* Cross product:
+       *      result.x = src0.y * src1.z - src0.z * src1.y;
+       *      result.y = src0.z * src1.x - src0.x * src1.z;
+       *      result.z = src0.x * src1.y - src0.y * src1.x;
+       *      result.w = undef;
+       */
+      src0 = src_vector(p, &inst->FullSrcRegisters[0]);
+      src1 = src_vector(p, &inst->FullSrcRegisters[1]);
+      tmp = i915_get_utemp(p);
+
+      i915_emit_arith(p,
+                      A0_MUL,
+                      tmp, A0_DEST_CHANNEL_ALL, 0,
+                      swizzle(src0, Z, X, Y, ONE),
+                      swizzle(src1, Y, Z, X, ONE), 0);
+
+      i915_emit_arith(p,
+                      A0_MAD,
+                      get_result_vector(p, &inst->FullDstRegisters[0]),
+                      get_result_flags(inst), 0,
+                      swizzle(src0, Y, Z, X, ONE),
+                      swizzle(src1, Z, X, Y, ONE),
+                      negate(tmp, 1, 1, 1, 0));
+      break;
+
+   default:
+      i915_program_error(p, "bad opcode %d", inst->Instruction.Opcode);
+      p->error = 1;
+      return;
+   }
+
+   i915_release_utemps(p);
+}
+
+
+/**
+ * Translate TGSI fragment shader into i915 hardware instructions.
+ * \param p  the translation state
+ * \param tokens  the TGSI token array
+ */
+static void
+i915_translate_instructions(struct i915_fp_compile *p,
+                            const struct tgsi_token *tokens)
+{
+   struct i915_fragment_shader *ifs = p->shader;
+   struct tgsi_parse_context parse;
+
+   tgsi_parse_init( &parse, tokens );
+
+   while( !tgsi_parse_end_of_tokens( &parse ) ) {
+
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+         if (parse.FullToken.FullDeclaration.Declaration.File
+                  == TGSI_FILE_CONSTANT) {
+            uint i;
+            for (i = parse.FullToken.FullDeclaration.DeclarationRange.First;
+                 i <= parse.FullToken.FullDeclaration.DeclarationRange.Last;
+                 i++) {
+               assert(ifs->constant_flags[i] == 0x0);
+               ifs->constant_flags[i] = I915_CONSTFLAG_USER;
+               ifs->num_constants = MAX2(ifs->num_constants, i + 1);
+            }
+         }
+         else if (parse.FullToken.FullDeclaration.Declaration.File
+                  == TGSI_FILE_TEMPORARY) {
+            uint i;
+            for (i = parse.FullToken.FullDeclaration.DeclarationRange.First;
+                 i <= parse.FullToken.FullDeclaration.DeclarationRange.Last;
+                 i++) {
+               assert(i < I915_MAX_TEMPORARY);
+               /* XXX just use shader->info->file_mask[TGSI_FILE_TEMPORARY] */
+               p->temp_flag |= (1 << i); /* mark temp as used */
+            }
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         {
+            const struct tgsi_full_immediate *imm
+               = &parse.FullToken.FullImmediate;
+            const uint pos = p->num_immediates++;
+            uint j;
+            for (j = 0; j < imm->Immediate.NrTokens - 1; j++) {
+               p->immediates[pos][j] = imm->u.ImmediateFloat32[j].Float;
+            }
+         }
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         if (p->first_instruction) {
+            /* resolve location of immediates */
+            uint i, j;
+            for (i = 0; i < p->num_immediates; i++) {
+               /* find constant slot for this immediate */
+               for (j = 0; j < I915_MAX_CONSTANT; j++) {
+                  if (ifs->constant_flags[j] == 0x0) {
+                     memcpy(ifs->constants[j],
+                            p->immediates[i],
+                            4 * sizeof(float));
+                     /*printf("immediate %d maps to const %d\n", i, j);*/
+                     ifs->constant_flags[j] = 0xf;  /* all four comps used */
+                     p->immediates_map[i] = j;
+                     ifs->num_constants = MAX2(ifs->num_constants, j + 1);
+                     break;
+                  }
+               }
+            }
+
+            p->first_instruction = FALSE;
+         }
+
+         i915_translate_instruction(p, &parse.FullToken.FullInstruction);
+         break;
+
+      default:
+         assert( 0 );
+      }
+
+   } /* while */
+
+   tgsi_parse_free (&parse);
+}
+
+
+static struct i915_fp_compile *
+i915_init_compile(struct i915_context *i915,
+                  struct i915_fragment_shader *ifs)
+{
+   struct i915_fp_compile *p = CALLOC_STRUCT(i915_fp_compile);
+
+   p->shader = ifs;
+
+   /* Put new constants at end of const buffer, growing downward.
+    * The problem is we don't know how many user-defined constants might
+    * be specified with pipe->set_constant_buffer().
+    * Should pre-scan the user's program to determine the highest-numbered
+    * constant referenced.
+    */
+   ifs->num_constants = 0;
+   memset(ifs->constant_flags, 0, sizeof(ifs->constant_flags));
+
+   p->first_instruction = TRUE;
+
+   p->nr_tex_indirect = 1;      /* correct? */
+   p->nr_tex_insn = 0;
+   p->nr_alu_insn = 0;
+   p->nr_decl_insn = 0;
+
+   p->csr = p->program;
+   p->decl = p->declarations;
+   p->decl_s = 0;
+   p->decl_t = 0;
+   p->temp_flag = ~0x0 << I915_MAX_TEMPORARY;
+   p->utemp_flag = ~0x7;
+
+   p->wpos_tex = -1;
+
+   /* initialize the first program word */
+   *(p->decl++) = _3DSTATE_PIXEL_SHADER_PROGRAM;
+
+   return p;
+}
+
+
+/* Copy compile results to the fragment program struct and destroy the
+ * compilation context.
+ */
+static void
+i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p)
+{
+   struct i915_fragment_shader *ifs = p->shader;
+   unsigned long program_size = (unsigned long) (p->csr - p->program);
+   unsigned long decl_size = (unsigned long) (p->decl - p->declarations);
+
+   if (p->nr_tex_indirect > I915_MAX_TEX_INDIRECT)
+      i915_program_error(p, "Exceeded max nr indirect texture lookups");
+
+   if (p->nr_tex_insn > I915_MAX_TEX_INSN)
+      i915_program_error(p, "Exceeded max TEX instructions");
+
+   if (p->nr_alu_insn > I915_MAX_ALU_INSN)
+      i915_program_error(p, "Exceeded max ALU instructions");
+
+   if (p->nr_decl_insn > I915_MAX_DECL_INSN)
+      i915_program_error(p, "Exceeded max DECL instructions");
+
+   if (p->error) {
+      p->NumNativeInstructions = 0;
+      p->NumNativeAluInstructions = 0;
+      p->NumNativeTexInstructions = 0;
+      p->NumNativeTexIndirections = 0;
+
+      i915_use_passthrough_shader(ifs);
+   }
+   else {
+      p->NumNativeInstructions
+         = p->nr_alu_insn + p->nr_tex_insn + p->nr_decl_insn;
+      p->NumNativeAluInstructions = p->nr_alu_insn;
+      p->NumNativeTexInstructions = p->nr_tex_insn;
+      p->NumNativeTexIndirections = p->nr_tex_indirect;
+
+      /* patch in the program length */
+      p->declarations[0] |= program_size + decl_size - 2;
+
+      /* Copy compilation results to fragment program struct: 
+       */
+      assert(!ifs->program);
+      ifs->program
+         = (uint *) MALLOC((program_size + decl_size) * sizeof(uint));
+      if (ifs->program) {
+         ifs->program_len = program_size + decl_size;
+
+         memcpy(ifs->program,
+                p->declarations, 
+                decl_size * sizeof(uint));
+
+         memcpy(ifs->program + decl_size, 
+                p->program, 
+                program_size * sizeof(uint));
+      }
+   }
+
+   /* Release the compilation struct: 
+    */
+   FREE(p);
+}
+
+
+/**
+ * Find an unused texture coordinate slot to use for fragment WPOS.
+ * Update p->fp->wpos_tex with the result (-1 if no used texcoord slot is found).
+ */
+static void
+i915_find_wpos_space(struct i915_fp_compile *p)
+{
+#if 0
+   const uint inputs
+      = p->shader->inputs_read | (1 << TGSI_ATTRIB_POS); /*XXX hack*/
+   uint i;
+
+   p->wpos_tex = -1;
+
+   if (inputs & (1 << TGSI_ATTRIB_POS)) {
+      for (i = 0; i < I915_TEX_UNITS; i++) {
+	 if ((inputs & (1 << (TGSI_ATTRIB_TEX0 + i))) == 0) {
+	    p->wpos_tex = i;
+	    return;
+	 }
+      }
+
+      i915_program_error(p, "No free texcoord for wpos value");
+   }
+#else
+   if (p->shader->info.input_semantic_name[0] == TGSI_SEMANTIC_POSITION) {
+      /* frag shader using the fragment position input */
+#if 0
+      assert(0);
+#endif
+   }
+#endif
+}
+
+
+
+
+/**
+ * Rather than trying to intercept and jiggle depth writes during
+ * emit, just move the value into its correct position at the end of
+ * the program:
+ */
+static void
+i915_fixup_depth_write(struct i915_fp_compile *p)
+{
+   /* XXX assuming pos/depth is always in output[0] */
+   if (p->shader->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION) {
+      const uint depth = UREG(REG_TYPE_OD, 0);
+
+      i915_emit_arith(p,
+                      A0_MOV,                     /* opcode */
+                      depth,                      /* dest reg */
+                      A0_DEST_CHANNEL_W,          /* write mask */
+                      0,                          /* saturate? */
+                      swizzle(depth, X, Y, Z, Z), /* src0 */
+                      0, 0 /* src1, src2 */);
+   }
+}
+
+
+void
+i915_translate_fragment_program( struct i915_context *i915,
+                                 struct i915_fragment_shader *fs)
+{
+   struct i915_fp_compile *p = i915_init_compile(i915, fs);
+   const struct tgsi_token *tokens = fs->state.tokens;
+
+   i915_find_wpos_space(p);
+
+#if 0
+   tgsi_dump(tokens, 0);
+#endif
+
+   i915_translate_instructions(p, tokens);
+   i915_fixup_depth_write(p);
+
+   i915_fini_compile(i915, p);
+}
diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c
new file mode 100644
index 0000000000..8f1f58b2dd
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_prim_emit.c
@@ -0,0 +1,220 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "draw/draw_pipe.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "util/u_pack_color.h"
+
+#include "i915_context.h"
+#include "i915_winsys.h"
+#include "i915_reg.h"
+#include "i915_state.h"
+#include "i915_batch.h"
+
+
+
+/**
+ * Primitive emit to hardware.  No support for vertex buffers or any
+ * nice fast paths.
+ */
+struct setup_stage {
+   struct draw_stage stage; /**< This must be first (base class) */
+
+   struct i915_context *i915;   
+};
+
+
+
+/**
+ * Basically a cast wrapper.
+ */
+static INLINE struct setup_stage *setup_stage( struct draw_stage *stage )
+{
+   return (struct setup_stage *)stage;
+}
+
+
+/**
+ * Extract the needed fields from vertex_header and emit i915 dwords.
+ * Recall that the vertices are constructed by the 'draw' module and
+ * have a couple of slots at the beginning (1-dword header, 4-dword
+ * clip pos) that we ignore here.
+ */
+static INLINE void
+emit_hw_vertex( struct i915_context *i915,
+                const struct vertex_header *vertex)
+{
+   const struct vertex_info *vinfo = &i915->current.vertex_info;
+   uint i;
+   uint count = 0;  /* for debug/sanity */
+
+   assert(!i915->dirty);
+
+   for (i = 0; i < vinfo->num_attribs; i++) {
+      const uint j = vinfo->attrib[i].src_index;
+      const float *attrib = vertex->data[j];
+      switch (vinfo->attrib[i].emit) {
+      case EMIT_1F:
+         OUT_BATCH( fui(attrib[0]) );
+         count++;
+         break;
+      case EMIT_2F:
+         OUT_BATCH( fui(attrib[0]) );
+         OUT_BATCH( fui(attrib[1]) );
+         count += 2;
+         break;
+      case EMIT_3F:
+         OUT_BATCH( fui(attrib[0]) );
+         OUT_BATCH( fui(attrib[1]) );
+         OUT_BATCH( fui(attrib[2]) );
+         count += 3;
+         break;
+      case EMIT_4F:
+         OUT_BATCH( fui(attrib[0]) );
+         OUT_BATCH( fui(attrib[1]) );
+         OUT_BATCH( fui(attrib[2]) );
+         OUT_BATCH( fui(attrib[3]) );
+         count += 4;
+         break;
+      case EMIT_4UB:
+         OUT_BATCH( pack_ub4(float_to_ubyte( attrib[2] ),
+                             float_to_ubyte( attrib[1] ),
+                             float_to_ubyte( attrib[0] ),
+                             float_to_ubyte( attrib[3] )) );
+         count += 1;
+         break;
+      default:
+         assert(0);
+      }
+   }
+   assert(count == vinfo->size);
+}
+
+
+
+static INLINE void 
+emit_prim( struct draw_stage *stage, 
+	   struct prim_header *prim,
+	   unsigned hwprim,
+	   unsigned nr )
+{
+   struct i915_context *i915 = setup_stage(stage)->i915;
+   unsigned vertex_size;
+   unsigned i;
+
+   if (i915->dirty)
+      i915_update_derived( i915 );
+
+   if (i915->hardware_dirty)
+      i915_emit_hardware_state( i915 );
+
+   /* need to do this after validation! */
+   vertex_size = i915->current.vertex_info.size * 4; /* in bytes */
+   assert(vertex_size >= 12); /* never smaller than 12 bytes */
+
+   if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) {
+      FLUSH_BATCH(NULL);
+
+      /* Make sure state is re-emitted after a flush: 
+       */
+      i915_update_derived( i915 );
+      i915_emit_hardware_state( i915 );
+
+      if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) {
+	 assert(0);
+	 return;
+      }
+   }
+
+   /* Emit each triangle as a single primitive.  I told you this was
+    * simple.
+    */
+   OUT_BATCH(_3DPRIMITIVE | 
+	     hwprim |
+	     ((4 + vertex_size * nr)/4 - 2));
+
+   for (i = 0; i < nr; i++)
+      emit_hw_vertex(i915, prim->v[i]);
+}
+
+
+static void 
+setup_tri( struct draw_stage *stage, struct prim_header *prim )
+{
+   emit_prim( stage, prim, PRIM3D_TRILIST, 3 );
+}
+
+
+static void
+setup_line(struct draw_stage *stage, struct prim_header *prim)
+{
+   emit_prim( stage, prim, PRIM3D_LINELIST, 2 );
+}
+
+
+static void
+setup_point(struct draw_stage *stage, struct prim_header *prim)
+{
+   emit_prim( stage, prim, PRIM3D_POINTLIST, 1 );
+}
+
+
+static void setup_flush( struct draw_stage *stage, unsigned flags )
+{
+}
+
+static void reset_stipple_counter( struct draw_stage *stage )
+{
+}
+
+static void render_destroy( struct draw_stage *stage )
+{
+   FREE( stage );
+}
+
+
+/**
+ * Create a new primitive setup/render stage.  This gets plugged into
+ * the 'draw' module's pipeline.
+ */
+struct draw_stage *i915_draw_render_stage( struct i915_context *i915 )
+{
+   struct setup_stage *setup = CALLOC_STRUCT(setup_stage);
+
+   setup->i915 = i915;
+   setup->stage.draw = i915->draw;
+   setup->stage.point = setup_point;
+   setup->stage.line = setup_line;
+   setup->stage.tri = setup_tri;
+   setup->stage.flush = setup_flush;
+   setup->stage.reset_stipple_counter = reset_stipple_counter;
+   setup->stage.destroy = render_destroy;
+
+   return &setup->stage;
+}
diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c
new file mode 100644
index 0000000000..f49f6d6ed1
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c
@@ -0,0 +1,545 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \file
+ * Build post-transformation, post-clipping vertex buffers and element
+ * lists by hooking into the end of the primitive pipeline and
+ * manipulating the vertex_id field in the vertex headers.
+ *
+ * XXX: work in progress 
+ * 
+ * \author José Fonseca <jrfonseca@tungstengraphics.com>
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "draw/draw_context.h"
+#include "draw/draw_vbuf.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "i915_context.h"
+#include "i915_reg.h"
+#include "i915_winsys.h"
+#include "i915_batch.h"
+#include "i915_state.h"
+
+
+/**
+ * Primitive renderer for i915.
+ */
+struct i915_vbuf_render {
+   struct vbuf_render base;
+
+   struct i915_context *i915;   
+
+   /** Vertex size in bytes */
+   unsigned vertex_size;
+
+   /** Software primitive */
+   unsigned prim;
+
+   /** Hardware primitive */
+   unsigned hwprim;
+
+   /** Genereate a vertex list */
+   unsigned fallback;
+
+   /* Stuff for the vbo */
+   struct pipe_buffer *vbo;
+   size_t vbo_size;
+   size_t vbo_offset;
+   void *vbo_ptr;
+   size_t vbo_alloc_size;
+};
+
+
+/**
+ * Basically a cast wrapper.
+ */
+static INLINE struct i915_vbuf_render *
+i915_vbuf_render( struct vbuf_render *render )
+{
+   assert(render);
+   return (struct i915_vbuf_render *)render;
+}
+
+
+static const struct vertex_info *
+i915_vbuf_render_get_vertex_info( struct vbuf_render *render )
+{
+   struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
+   struct i915_context *i915 = i915_render->i915;
+
+   if (i915->dirty) {
+      /* make sure we have up to date vertex layout */
+      i915_update_derived( i915 );
+   }
+
+   return &i915->current.vertex_info;
+}
+
+
+static void *
+i915_vbuf_render_allocate_vertices( struct vbuf_render *render,
+                                    ushort vertex_size,
+                                    ushort nr_vertices )
+{
+   struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
+   struct i915_context *i915 = i915_render->i915;
+   struct pipe_screen *screen = i915->pipe.screen;
+   size_t size = (size_t)vertex_size * (size_t)nr_vertices;
+
+   /* FIXME: handle failure */
+   assert(!i915->vbo);
+
+   if (i915_render->vbo_size > size + i915_render->vbo_offset && !i915->vbo_flushed) {
+   } else {
+      i915->vbo_flushed = 0;
+      pipe_buffer_reference(screen, &i915_render->vbo, NULL);
+   }
+
+   if (!i915_render->vbo) {
+      i915_render->vbo_size = MAX2(size, i915_render->vbo_alloc_size);
+      i915_render->vbo_offset = 0;
+      i915_render->vbo = pipe_buffer_create(screen,
+                                            64,
+                                            I915_BUFFER_USAGE_LIT_VERTEX,
+                                            i915_render->vbo_size);
+      i915_render->vbo_ptr = pipe_buffer_map(screen,
+                                             i915_render->vbo,
+                                             PIPE_BUFFER_USAGE_CPU_WRITE);
+      pipe_buffer_unmap(screen, i915_render->vbo);
+   }
+
+   i915->vbo = i915_render->vbo;
+   i915->vbo_offset = i915_render->vbo_offset;
+   i915->dirty |= I915_NEW_VBO;
+
+   return (unsigned char *)i915_render->vbo_ptr + i915->vbo_offset;
+}
+
+
+static boolean
+i915_vbuf_render_set_primitive( struct vbuf_render *render, 
+                                unsigned prim )
+{
+   struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
+   i915_render->prim = prim;
+
+   switch(prim) {
+   case PIPE_PRIM_POINTS:
+      i915_render->hwprim = PRIM3D_POINTLIST;
+      i915_render->fallback = 0;
+      return TRUE;
+   case PIPE_PRIM_LINES:
+      i915_render->hwprim = PRIM3D_LINELIST;
+      i915_render->fallback = 0;
+      return TRUE;
+   case PIPE_PRIM_LINE_LOOP:
+      i915_render->hwprim = PRIM3D_LINELIST;
+      i915_render->fallback = PIPE_PRIM_LINE_LOOP;
+      return TRUE;
+   case PIPE_PRIM_LINE_STRIP:
+      i915_render->hwprim = PRIM3D_LINESTRIP;
+      i915_render->fallback = 0;
+      return TRUE;
+   case PIPE_PRIM_TRIANGLES:
+      i915_render->hwprim = PRIM3D_TRILIST;
+      i915_render->fallback = 0;
+      return TRUE;
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      i915_render->hwprim = PRIM3D_TRISTRIP;
+      i915_render->fallback = 0;
+      return TRUE;
+   case PIPE_PRIM_TRIANGLE_FAN:
+      i915_render->hwprim = PRIM3D_TRIFAN;
+      i915_render->fallback = 0;
+      return TRUE;
+   case PIPE_PRIM_QUADS:
+      i915_render->hwprim = PRIM3D_TRILIST;
+      i915_render->fallback = PIPE_PRIM_QUADS;
+      return TRUE;
+   case PIPE_PRIM_QUAD_STRIP:
+      i915_render->hwprim = PRIM3D_TRILIST;
+      i915_render->fallback = PIPE_PRIM_QUAD_STRIP;
+      return TRUE;
+   case PIPE_PRIM_POLYGON:
+      i915_render->hwprim = PRIM3D_POLY;
+      i915_render->fallback = 0;
+      return TRUE;
+   default:
+      /* FIXME: Actually, can handle a lot more just fine... */
+      return FALSE;
+   }
+}
+
+
+
+/**
+ * Used for fallbacks in draw_arrays
+ */
+static void
+draw_arrays_generate_indices( struct vbuf_render *render,
+                              unsigned start, uint nr,
+                              unsigned type )
+{
+   struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
+   struct i915_context *i915 = i915_render->i915;
+   unsigned i;
+   unsigned end = start + nr;
+   switch(type) {
+   case 0:
+      for (i = start; i+1 < end; i += 2)
+	 OUT_BATCH( (i+0) | (i+1) << 16 );
+      if (i < end)
+	 OUT_BATCH( i );
+      break;
+   case PIPE_PRIM_LINE_LOOP:
+      if (nr >= 2) {
+	 for (i = start + 1; i < end; i++)
+	    OUT_BATCH( (i-0) | (i+0) << 16 );
+	 OUT_BATCH( (i-0) | (  start) << 16 );
+      }
+      break;
+   case PIPE_PRIM_QUADS:
+      for (i = start; i + 3 < end; i += 4) {
+	 OUT_BATCH( (i+0) | (i+1) << 16 );
+	 OUT_BATCH( (i+3) | (i+1) << 16 );
+	 OUT_BATCH( (i+2) | (i+3) << 16 );
+      }
+      break;
+   case PIPE_PRIM_QUAD_STRIP:
+      for (i = start; i + 3 < end; i += 2) {
+	 OUT_BATCH( (i+0) | (i+1) << 16 );
+	 OUT_BATCH( (i+3) | (i+2) << 16 );
+	 OUT_BATCH( (i+0) | (i+3) << 16 );
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+static unsigned
+draw_arrays_calc_nr_indices( uint nr, unsigned type )
+{
+   switch (type) {
+   case 0:
+      return nr;
+   case PIPE_PRIM_LINE_LOOP:
+      if (nr >= 2)
+	 return nr * 2;
+      else
+	 return 0;
+   case PIPE_PRIM_QUADS:
+      return (nr / 4) * 6;
+   case PIPE_PRIM_QUAD_STRIP:
+      return ((nr - 2) / 2) * 6;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+static void
+draw_arrays_fallback( struct vbuf_render *render,
+                      unsigned start,
+                      uint nr )
+{
+   struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
+   struct i915_context *i915 = i915_render->i915;
+   unsigned nr_indices;
+
+   if (i915->dirty)
+      i915_update_derived( i915 );
+
+   if (i915->hardware_dirty)
+      i915_emit_hardware_state( i915 );
+
+   nr_indices = draw_arrays_calc_nr_indices( nr, i915_render->fallback );
+   if (!nr_indices)
+      return;
+
+   if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) {
+      FLUSH_BATCH(NULL);
+
+      /* Make sure state is re-emitted after a flush:
+       */
+      i915_update_derived( i915 );
+      i915_emit_hardware_state( i915 );
+      i915->vbo_flushed = 1;
+
+      if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) {
+	 assert(0);
+	 goto out;
+      }
+   }
+   OUT_BATCH( _3DPRIMITIVE |
+	      PRIM_INDIRECT |
+	      i915_render->hwprim |
+	      PRIM_INDIRECT_ELTS |
+	      nr_indices );
+
+   draw_arrays_generate_indices( render, start, nr, i915_render->fallback );
+
+out:
+   return;
+}
+
+static void
+i915_vbuf_render_draw_arrays( struct vbuf_render *render,
+                              unsigned start,
+                              uint nr )
+{
+   struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
+
+   if (i915_render->fallback) {
+      draw_arrays_fallback( render, start, nr );
+      return;
+   }
+
+   /* JB: TODO submit direct cmds */
+   draw_arrays_fallback( render, start, nr );
+}
+
+/**
+ * Used for normal and fallback emitting of indices
+ * If type is zero normal operation assumed.
+ */
+static void
+draw_generate_indices( struct vbuf_render *render,
+                       const ushort *indices,
+                       uint nr_indices,
+                       unsigned type )
+{
+   struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
+   struct i915_context *i915 = i915_render->i915;
+   unsigned i;
+
+   switch(type) {
+   case 0:
+      for (i = 0; i + 1 < nr_indices; i += 2) {
+	 OUT_BATCH( indices[i] | indices[i+1] << 16 );
+      }
+      if (i < nr_indices) {
+	 OUT_BATCH( indices[i] );
+      }
+      break;
+   case PIPE_PRIM_LINE_LOOP:
+      if (nr_indices >= 2) {
+	 for (i = 1; i < nr_indices; i++)
+	    OUT_BATCH( indices[i-1] | indices[i] << 16 );
+	 OUT_BATCH( indices[i-1] | indices[0] << 16 );
+      }
+      break;
+   case PIPE_PRIM_QUADS:
+      for (i = 0; i + 3 < nr_indices; i += 4) {
+	 OUT_BATCH( indices[i+0] | indices[i+1] << 16 );
+	 OUT_BATCH( indices[i+3] | indices[i+1] << 16 );
+	 OUT_BATCH( indices[i+2] | indices[i+3] << 16 );
+      }
+      break;
+   case PIPE_PRIM_QUAD_STRIP:
+      for (i = 0; i + 3 < nr_indices; i += 2) {
+	 OUT_BATCH( indices[i+0] | indices[i+1] << 16 );
+	 OUT_BATCH( indices[i+3] | indices[i+2] << 16 );
+	 OUT_BATCH( indices[i+0] | indices[i+3] << 16 );
+      }
+      break;
+   default:
+      assert(0);
+      break;
+   }
+}
+
+static unsigned
+draw_calc_nr_indices( uint nr_indices, unsigned type )
+{
+   switch (type) {
+   case 0:
+      return nr_indices;
+   case PIPE_PRIM_LINE_LOOP:
+      if (nr_indices >= 2)
+	 return nr_indices * 2;
+      else
+	 return 0;
+   case PIPE_PRIM_QUADS:
+      return (nr_indices / 4) * 6;
+   case PIPE_PRIM_QUAD_STRIP:
+      return ((nr_indices - 2) / 2) * 6;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+static void 
+i915_vbuf_render_draw( struct vbuf_render *render,
+                       const ushort *indices,
+                       uint nr_indices)
+{
+   struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
+   struct i915_context *i915 = i915_render->i915;
+   unsigned save_nr_indices;
+
+   save_nr_indices = nr_indices;
+
+   nr_indices = draw_calc_nr_indices( nr_indices, i915_render->fallback );
+   if (!nr_indices)
+      return;
+
+   if (i915->dirty)
+      i915_update_derived( i915 );
+
+   if (i915->hardware_dirty)
+      i915_emit_hardware_state( i915 );
+
+   if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) {
+      FLUSH_BATCH(NULL);
+
+      /* Make sure state is re-emitted after a flush: 
+       */
+      i915_update_derived( i915 );
+      i915_emit_hardware_state( i915 );
+      i915->vbo_flushed = 1;
+
+      if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) {
+	 assert(0);
+     goto out;
+      }
+   }
+
+   OUT_BATCH( _3DPRIMITIVE |
+	      PRIM_INDIRECT |
+	      i915_render->hwprim |
+	      PRIM_INDIRECT_ELTS |
+	      nr_indices );
+   draw_generate_indices( render,
+			  indices,
+			  save_nr_indices,
+			  i915_render->fallback );
+
+out:
+   return;
+}
+
+
+static void
+i915_vbuf_render_release_vertices( struct vbuf_render *render,
+			           void *vertices, 
+			           unsigned vertex_size,
+			           unsigned vertices_used )
+{
+   struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
+   struct i915_context *i915 = i915_render->i915;
+   size_t size = (size_t)vertex_size * (size_t)vertices_used;
+
+   assert(i915->vbo);
+
+   i915_render->vbo_offset += size;
+   i915->vbo = NULL;
+   i915->dirty |= I915_NEW_VBO;
+}
+
+
+static void
+i915_vbuf_render_destroy( struct vbuf_render *render )
+{
+   struct i915_vbuf_render *i915_render = i915_vbuf_render(render);
+   FREE(i915_render);
+}
+
+
+/**
+ * Create a new primitive render.
+ */
+static struct vbuf_render *
+i915_vbuf_render_create( struct i915_context *i915 )
+{
+   struct i915_vbuf_render *i915_render = CALLOC_STRUCT(i915_vbuf_render);
+   struct pipe_screen *screen = i915->pipe.screen;
+
+   i915_render->i915 = i915;
+   
+   i915_render->base.max_vertex_buffer_bytes = 128*1024;
+   
+   /* NOTE: it must be such that state and vertices indices fit in a single 
+    * batch buffer.
+    */
+   i915_render->base.max_indices = 16*1024;
+
+   i915_render->base.get_vertex_info = i915_vbuf_render_get_vertex_info;
+   i915_render->base.allocate_vertices = i915_vbuf_render_allocate_vertices;
+   i915_render->base.set_primitive = i915_vbuf_render_set_primitive;
+   i915_render->base.draw = i915_vbuf_render_draw;
+   i915_render->base.draw_arrays = i915_vbuf_render_draw_arrays;
+   i915_render->base.release_vertices = i915_vbuf_render_release_vertices;
+   i915_render->base.destroy = i915_vbuf_render_destroy;
+
+   i915_render->vbo_alloc_size = 128 * 4096;
+   i915_render->vbo_size = i915_render->vbo_alloc_size;
+   i915_render->vbo_offset = 0;
+   i915_render->vbo = pipe_buffer_create(screen,
+                                         64,
+                                         I915_BUFFER_USAGE_LIT_VERTEX,
+                                         i915_render->vbo_size);
+   i915_render->vbo_ptr = pipe_buffer_map(screen,
+                                          i915_render->vbo,
+                                          PIPE_BUFFER_USAGE_CPU_WRITE);
+   pipe_buffer_unmap(screen, i915_render->vbo);
+
+   return &i915_render->base;
+}
+
+
+/**
+ * Create a new primitive vbuf/render stage.
+ */
+struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 )
+{
+   struct vbuf_render *render;
+   struct draw_stage *stage;
+   
+   render = i915_vbuf_render_create(i915);
+   if(!render)
+      return NULL;
+   
+   stage = draw_vbuf_stage( i915->draw, render );
+   if(!stage) {
+      render->destroy(render);
+      return NULL;
+   }
+   /** TODO JB: this shouldn't be here */
+   draw_set_render(i915->draw, render);
+
+   return stage;
+}
diff --git a/src/gallium/drivers/i915simple/i915_reg.h b/src/gallium/drivers/i915simple/i915_reg.h
new file mode 100644
index 0000000000..04620fec68
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_reg.h
@@ -0,0 +1,978 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef I915_REG_H
+#define I915_REG_H
+
+
+#define I915_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value)
+
+#define CMD_3D (0x3<<29)
+
+#define PRIM3D_INLINE		(CMD_3D | (0x1f<<24))
+#define PRIM3D_TRILIST		(0x0<<18)
+#define PRIM3D_TRISTRIP 	(0x1<<18)
+#define PRIM3D_TRISTRIP_RVRSE	(0x2<<18)
+#define PRIM3D_TRIFAN		(0x3<<18)
+#define PRIM3D_POLY		(0x4<<18)
+#define PRIM3D_LINELIST 	(0x5<<18)
+#define PRIM3D_LINESTRIP	(0x6<<18)
+#define PRIM3D_RECTLIST 	(0x7<<18)
+#define PRIM3D_POINTLIST	(0x8<<18)
+#define PRIM3D_DIB		(0x9<<18)
+#define PRIM3D_CLEAR_RECT	(0xa<<18)
+#define PRIM3D_ZONE_INIT	(0xd<<18)
+#define PRIM3D_MASK		(0x1f<<18)
+
+/* p137 */
+#define _3DSTATE_AA_CMD			(CMD_3D | (0x06<<24))
+#define AA_LINE_ECAAR_WIDTH_ENABLE	(1<<16)
+#define AA_LINE_ECAAR_WIDTH_0_5 	0
+#define AA_LINE_ECAAR_WIDTH_1_0		(1<<14)
+#define AA_LINE_ECAAR_WIDTH_2_0 	(2<<14)
+#define AA_LINE_ECAAR_WIDTH_4_0 	(3<<14)
+#define AA_LINE_REGION_WIDTH_ENABLE	(1<<8)
+#define AA_LINE_REGION_WIDTH_0_5	0
+#define AA_LINE_REGION_WIDTH_1_0	(1<<6)
+#define AA_LINE_REGION_WIDTH_2_0	(2<<6)
+#define AA_LINE_REGION_WIDTH_4_0	(3<<6)
+
+/* 3DSTATE_BACKFACE_STENCIL_OPS, p138*/
+#define _3DSTATE_BACKFACE_STENCIL_OPS    (CMD_3D | (0x8<<24))
+#define BFO_ENABLE_STENCIL_REF          (1<<23)
+#define BFO_STENCIL_REF_SHIFT           15
+#define BFO_STENCIL_REF_MASK            (0xff<<15)
+#define BFO_ENABLE_STENCIL_FUNCS        (1<<14)
+#define BFO_STENCIL_TEST_SHIFT          11
+#define BFO_STENCIL_TEST_MASK           (0x7<<11)
+#define BFO_STENCIL_FAIL_SHIFT          8
+#define BFO_STENCIL_FAIL_MASK           (0x7<<8)
+#define BFO_STENCIL_PASS_Z_FAIL_SHIFT   5
+#define BFO_STENCIL_PASS_Z_FAIL_MASK    (0x7<<5)
+#define BFO_STENCIL_PASS_Z_PASS_SHIFT   2
+#define BFO_STENCIL_PASS_Z_PASS_MASK    (0x7<<2)
+#define BFO_ENABLE_STENCIL_TWO_SIDE     (1<<1)
+#define BFO_STENCIL_TWO_SIDE            (1<<0)
+
+
+/* 3DSTATE_BACKFACE_STENCIL_MASKS, p140 */
+#define _3DSTATE_BACKFACE_STENCIL_MASKS    (CMD_3D | (0x9<<24))
+#define BFM_ENABLE_STENCIL_TEST_MASK      (1<<17)
+#define BFM_ENABLE_STENCIL_WRITE_MASK     (1<<16)
+#define BFM_STENCIL_TEST_MASK_SHIFT       8
+#define BFM_STENCIL_TEST_MASK_MASK        (0xff<<8)
+#define BFM_STENCIL_WRITE_MASK_SHIFT      0
+#define BFM_STENCIL_WRITE_MASK_MASK       (0xff<<0)
+
+
+
+/* 3DSTATE_BIN_CONTROL p141 */
+
+/* p143 */
+#define _3DSTATE_BUF_INFO_CMD	(CMD_3D | (0x1d<<24) | (0x8e<<16) | 1)
+/* Dword 1 */
+#define BUF_3D_ID_COLOR_BACK	(0x3<<24)
+#define BUF_3D_ID_DEPTH 	(0x7<<24)
+#define BUF_3D_USE_FENCE	(1<<23)
+#define BUF_3D_TILED_SURFACE	(1<<22)
+#define BUF_3D_TILE_WALK_X	0
+#define BUF_3D_TILE_WALK_Y	(1<<21)
+#define BUF_3D_PITCH(x)         (((x)/4)<<2)
+/* Dword 2 */
+#define BUF_3D_ADDR(x)		((x) & ~0x3)
+
+
+/* 3DSTATE_CHROMA_KEY */
+
+/* 3DSTATE_CLEAR_PARAMETERS, p150 */
+#define _3DSTATE_CLEAR_PARAMETERS	(CMD_3D | (0x1d<<24) | (0x9c<<16) | 5)
+/* Dword 1 */
+#define CLEARPARAM_CLEAR_RECT		(1 << 16)
+#define CLEARPARAM_ZONE_INIT		(0 << 16)
+#define CLEARPARAM_WRITE_COLOR		(1 << 2)
+#define CLEARPARAM_WRITE_DEPTH		(1 << 1)
+#define CLEARPARAM_WRITE_STENCIL	(1 << 0)
+
+/* 3DSTATE_CONSTANT_BLEND_COLOR, p153 */
+#define _3DSTATE_CONST_BLEND_COLOR_CMD	(CMD_3D | (0x1d<<24) | (0x88<<16))
+
+
+
+/* 3DSTATE_COORD_SET_BINDINGS, p154 */
+#define _3DSTATE_COORD_SET_BINDINGS      (CMD_3D | (0x16<<24))
+#define CSB_TCB(iunit, eunit)           ((eunit)<<(iunit*3))
+
+/* p156 */
+#define _3DSTATE_DFLT_DIFFUSE_CMD	(CMD_3D | (0x1d<<24) | (0x99<<16))
+
+/* p157 */
+#define _3DSTATE_DFLT_SPEC_CMD		(CMD_3D | (0x1d<<24) | (0x9a<<16))
+
+/* p158 */
+#define _3DSTATE_DFLT_Z_CMD		(CMD_3D | (0x1d<<24) | (0x98<<16))
+
+
+/* 3DSTATE_DEPTH_OFFSET_SCALE, p159 */
+#define _3DSTATE_DEPTH_OFFSET_SCALE       (CMD_3D | (0x1d<<24) | (0x97<<16))
+/* scale in dword 1 */
+
+
+/* 3DSTATE_DEPTH_SUBRECT_DISABLE, p160 */
+#define _3DSTATE_DEPTH_SUBRECT_DISABLE    (CMD_3D | (0x1c<<24) | (0x11<<19) | 0x2)
+
+/* p161 */
+#define _3DSTATE_DST_BUF_VARS_CMD	(CMD_3D | (0x1d<<24) | (0x85<<16))
+/* Dword 1 */
+#define TEX_DEFAULT_COLOR_OGL           (0<<30)
+#define TEX_DEFAULT_COLOR_D3D           (1<<30)
+#define ZR_EARLY_DEPTH                  (1<<29)
+#define LOD_PRECLAMP_OGL                (1<<28)
+#define LOD_PRECLAMP_D3D                (0<<28)
+#define DITHER_FULL_ALWAYS              (0<<26)
+#define DITHER_FULL_ON_FB_BLEND         (1<<26)
+#define DITHER_CLAMPED_ALWAYS           (2<<26)
+#define LINEAR_GAMMA_BLEND_32BPP        (1<<25)
+#define DEBUG_DISABLE_ENH_DITHER        (1<<24)
+#define DSTORG_HORT_BIAS(x)		((x)<<20)
+#define DSTORG_VERT_BIAS(x)		((x)<<16)
+#define COLOR_4_2_2_CHNL_WRT_ALL	0
+#define COLOR_4_2_2_CHNL_WRT_Y		(1<<12)
+#define COLOR_4_2_2_CHNL_WRT_CR		(2<<12)
+#define COLOR_4_2_2_CHNL_WRT_CB		(3<<12)
+#define COLOR_4_2_2_CHNL_WRT_CRCB	(4<<12)
+#define COLOR_BUF_8BIT			0
+#define COLOR_BUF_RGB555 		(1<<8)
+#define COLOR_BUF_RGB565 		(2<<8)
+#define COLOR_BUF_ARGB8888		(3<<8)
+#define DEPTH_FRMT_16_FIXED		0
+#define DEPTH_FRMT_16_FLOAT		(1<<2)
+#define DEPTH_FRMT_24_FIXED_8_OTHER	(2<<2)
+#define VERT_LINE_STRIDE_1		(1<<1)
+#define VERT_LINE_STRIDE_0		(0<<1)
+#define VERT_LINE_STRIDE_OFS_1		1
+#define VERT_LINE_STRIDE_OFS_0		0
+
+/* p166 */
+#define _3DSTATE_DRAW_RECT_CMD		(CMD_3D|(0x1d<<24)|(0x80<<16)|3)
+/* Dword 1 */
+#define DRAW_RECT_DIS_DEPTH_OFS 	(1<<30)
+#define DRAW_DITHER_OFS_X(x)		((x)<<26)
+#define DRAW_DITHER_OFS_Y(x)		((x)<<24)
+/* Dword 2 */
+#define DRAW_YMIN(x)			((x)<<16)
+#define DRAW_XMIN(x)			(x)
+/* Dword 3 */
+#define DRAW_YMAX(x)			((x)<<16)
+#define DRAW_XMAX(x)			(x)
+/* Dword 4 */
+#define DRAW_YORG(x)			((x)<<16)
+#define DRAW_XORG(x)			(x)
+
+
+/* 3DSTATE_FILTER_COEFFICIENTS_4X4, p170 */
+
+/* 3DSTATE_FILTER_COEFFICIENTS_6X5, p172 */
+
+
+/* _3DSTATE_FOG_COLOR, p173 */
+#define _3DSTATE_FOG_COLOR_CMD		(CMD_3D|(0x15<<24))
+#define FOG_COLOR_RED(x)		((x)<<16)
+#define FOG_COLOR_GREEN(x)		((x)<<8)
+#define FOG_COLOR_BLUE(x)		(x)
+
+/* _3DSTATE_FOG_MODE, p174 */
+#define _3DSTATE_FOG_MODE_CMD		(CMD_3D|(0x1d<<24)|(0x89<<16)|2)
+/* Dword 1 */
+#define FMC1_FOGFUNC_MODIFY_ENABLE	(1<<31)
+#define FMC1_FOGFUNC_VERTEX		(0<<28)
+#define FMC1_FOGFUNC_PIXEL_EXP		(1<<28)
+#define FMC1_FOGFUNC_PIXEL_EXP2		(2<<28)
+#define FMC1_FOGFUNC_PIXEL_LINEAR	(3<<28)
+#define FMC1_FOGFUNC_MASK		(3<<28)
+#define FMC1_FOGINDEX_MODIFY_ENABLE     (1<<27)
+#define FMC1_FOGINDEX_Z		        (0<<25)
+#define FMC1_FOGINDEX_W   		(1<<25)
+#define FMC1_C1_C2_MODIFY_ENABLE	(1<<24)
+#define FMC1_DENSITY_MODIFY_ENABLE	(1<<23)
+#define FMC1_C1_ONE      	        (1<<13)
+#define FMC1_C1_MASK		        (0xffff<<4)
+/* Dword 2 */
+#define FMC2_C2_ONE		        (1<<16)
+/* Dword 3 */
+#define FMC3_D_ONE      		(1<<16)
+
+
+
+/* _3DSTATE_INDEPENDENT_ALPHA_BLEND, p177 */
+#define _3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD	(CMD_3D|(0x0b<<24))
+#define IAB_MODIFY_ENABLE	        (1<<23)
+#define IAB_ENABLE       	        (1<<22)
+#define IAB_MODIFY_FUNC         	(1<<21)
+#define IAB_FUNC_SHIFT          	16
+#define IAB_MODIFY_SRC_FACTOR   	(1<<11)
+#define IAB_SRC_FACTOR_SHIFT		6
+#define IAB_SRC_FACTOR_MASK		(BLENDFACT_MASK<<6)
+#define IAB_MODIFY_DST_FACTOR	        (1<<5)
+#define IAB_DST_FACTOR_SHIFT		0
+#define IAB_DST_FACTOR_MASK		(BLENDFACT_MASK<<0)
+
+
+#define BLENDFUNC_ADD			0x0
+#define BLENDFUNC_SUBTRACT		0x1
+#define BLENDFUNC_REVERSE_SUBTRACT	0x2
+#define BLENDFUNC_MIN			0x3
+#define BLENDFUNC_MAX			0x4
+#define BLENDFUNC_MASK			0x7
+
+/* 3DSTATE_LOAD_INDIRECT, p180 */
+
+#define _3DSTATE_LOAD_INDIRECT	        (CMD_3D|(0x1d<<24)|(0x7<<16))
+#define LI0_STATE_STATIC_INDIRECT       (0x01<<8)
+#define LI0_STATE_DYNAMIC_INDIRECT      (0x02<<8)
+#define LI0_STATE_SAMPLER               (0x04<<8)
+#define LI0_STATE_MAP                   (0x08<<8)
+#define LI0_STATE_PROGRAM               (0x10<<8)
+#define LI0_STATE_CONSTANTS             (0x20<<8)
+
+#define SIS0_BUFFER_ADDRESS(x)          ((x)&~0x3)
+#define SIS0_FORCE_LOAD                 (1<<1)
+#define SIS0_BUFFER_VALID               (1<<0)
+#define SIS1_BUFFER_LENGTH(x)           ((x)&0xff)
+
+#define DIS0_BUFFER_ADDRESS(x)          ((x)&~0x3)
+#define DIS0_BUFFER_RESET               (1<<1)
+#define DIS0_BUFFER_VALID               (1<<0)
+
+#define SSB0_BUFFER_ADDRESS(x)          ((x)&~0x3)
+#define SSB0_FORCE_LOAD                 (1<<1)
+#define SSB0_BUFFER_VALID               (1<<0)
+#define SSB1_BUFFER_LENGTH(x)           ((x)&0xff)
+
+#define MSB0_BUFFER_ADDRESS(x)          ((x)&~0x3)
+#define MSB0_FORCE_LOAD                 (1<<1)
+#define MSB0_BUFFER_VALID               (1<<0)
+#define MSB1_BUFFER_LENGTH(x)           ((x)&0xff)
+
+#define PSP0_BUFFER_ADDRESS(x)          ((x)&~0x3)
+#define PSP0_FORCE_LOAD                 (1<<1)
+#define PSP0_BUFFER_VALID               (1<<0)
+#define PSP1_BUFFER_LENGTH(x)           ((x)&0xff)
+
+#define PSC0_BUFFER_ADDRESS(x)          ((x)&~0x3)
+#define PSC0_FORCE_LOAD                 (1<<1)
+#define PSC0_BUFFER_VALID               (1<<0)
+#define PSC1_BUFFER_LENGTH(x)           ((x)&0xff)
+
+
+
+
+
+/* _3DSTATE_RASTERIZATION_RULES */
+#define _3DSTATE_RASTER_RULES_CMD	(CMD_3D|(0x07<<24))
+#define ENABLE_POINT_RASTER_RULE	(1<<15)
+#define OGL_POINT_RASTER_RULE		(1<<13)
+#define ENABLE_TEXKILL_3D_4D            (1<<10)
+#define TEXKILL_3D                      (0<<9)
+#define TEXKILL_4D                      (1<<9)
+#define ENABLE_LINE_STRIP_PROVOKE_VRTX	(1<<8)
+#define ENABLE_TRI_FAN_PROVOKE_VRTX	(1<<5)
+#define LINE_STRIP_PROVOKE_VRTX(x)	((x)<<6)
+#define TRI_FAN_PROVOKE_VRTX(x) 	((x)<<3)
+
+/* _3DSTATE_SCISSOR_ENABLE, p256 */
+#define _3DSTATE_SCISSOR_ENABLE_CMD	(CMD_3D|(0x1c<<24)|(0x10<<19))
+#define ENABLE_SCISSOR_RECT		((1<<1) | 1)
+#define DISABLE_SCISSOR_RECT		(1<<1)
+
+/* _3DSTATE_SCISSOR_RECTANGLE_0, p257 */
+#define _3DSTATE_SCISSOR_RECT_0_CMD	(CMD_3D|(0x1d<<24)|(0x81<<16)|1)
+/* Dword 1 */
+#define SCISSOR_RECT_0_YMIN(x)		((x)<<16)
+#define SCISSOR_RECT_0_XMIN(x)		(x)
+/* Dword 2 */
+#define SCISSOR_RECT_0_YMAX(x)		((x)<<16)
+#define SCISSOR_RECT_0_XMAX(x)		(x)
+
+/* p189 */
+#define _3DSTATE_LOAD_STATE_IMMEDIATE_1   ((0x3<<29)|(0x1d<<24)|(0x04<<16))
+#define I1_LOAD_S(n)                      (1<<(4+n))
+
+#define S0_VB_OFFSET_MASK              0xffffffc
+#define S0_AUTO_CACHE_INV_DISABLE      (1<<0)
+
+#define S1_VERTEX_WIDTH_SHIFT          24
+#define S1_VERTEX_WIDTH_MASK           (0x3f<<24)
+#define S1_VERTEX_PITCH_SHIFT          16
+#define S1_VERTEX_PITCH_MASK           (0x3f<<16)
+
+#define TEXCOORDFMT_2D                 0x0
+#define TEXCOORDFMT_3D                 0x1
+#define TEXCOORDFMT_4D                 0x2
+#define TEXCOORDFMT_1D                 0x3
+#define TEXCOORDFMT_2D_16              0x4
+#define TEXCOORDFMT_4D_16              0x5
+#define TEXCOORDFMT_NOT_PRESENT        0xf
+#define S2_TEXCOORD_FMT0_MASK            0xf
+#define S2_TEXCOORD_FMT1_SHIFT           4
+#define S2_TEXCOORD_FMT(unit, type)    ((type)<<(unit*4))
+#define S2_TEXCOORD_NONE               (~0)
+
+/* S3 not interesting */
+
+#define S4_POINT_WIDTH_SHIFT           23
+#define S4_POINT_WIDTH_MASK            (0x1ff<<23)
+#define S4_LINE_WIDTH_SHIFT            19
+#define S4_LINE_WIDTH_ONE              (0x2<<19)
+#define S4_LINE_WIDTH_MASK             (0xf<<19)
+#define S4_FLATSHADE_ALPHA             (1<<18)
+#define S4_FLATSHADE_FOG               (1<<17)
+#define S4_FLATSHADE_SPECULAR          (1<<16)
+#define S4_FLATSHADE_COLOR             (1<<15)
+#define S4_CULLMODE_BOTH	       (0<<13)
+#define S4_CULLMODE_NONE	       (1<<13)
+#define S4_CULLMODE_CW		       (2<<13)
+#define S4_CULLMODE_CCW		       (3<<13)
+#define S4_CULLMODE_MASK	       (3<<13)
+#define S4_VFMT_POINT_WIDTH            (1<<12)
+#define S4_VFMT_SPEC_FOG               (1<<11)
+#define S4_VFMT_COLOR                  (1<<10)
+#define S4_VFMT_DEPTH_OFFSET           (1<<9)
+#define S4_VFMT_XYZ     	       (1<<6)
+#define S4_VFMT_XYZW     	       (2<<6)
+#define S4_VFMT_XY     		       (3<<6)
+#define S4_VFMT_XYW     	       (4<<6)
+#define S4_VFMT_XYZW_MASK              (7<<6)
+#define S4_FORCE_DEFAULT_DIFFUSE       (1<<5)
+#define S4_FORCE_DEFAULT_SPECULAR      (1<<4)
+#define S4_LOCAL_DEPTH_OFFSET_ENABLE   (1<<3)
+#define S4_VFMT_FOG_PARAM              (1<<2)
+#define S4_SPRITE_POINT_ENABLE         (1<<1)
+#define S4_LINE_ANTIALIAS_ENABLE       (1<<0)
+
+#define S4_VFMT_MASK (S4_VFMT_POINT_WIDTH   | 	\
+		      S4_VFMT_SPEC_FOG      |	\
+		      S4_VFMT_COLOR         |	\
+		      S4_VFMT_DEPTH_OFFSET  |	\
+		      S4_VFMT_XYZW_MASK     |	\
+		      S4_VFMT_FOG_PARAM)
+
+
+#define S5_WRITEDISABLE_ALPHA          (1<<31)
+#define S5_WRITEDISABLE_RED            (1<<30)
+#define S5_WRITEDISABLE_GREEN          (1<<29)
+#define S5_WRITEDISABLE_BLUE           (1<<28)
+#define S5_WRITEDISABLE_MASK           (0xf<<28)
+#define S5_FORCE_DEFAULT_POINT_SIZE    (1<<27)
+#define S5_LAST_PIXEL_ENABLE           (1<<26)
+#define S5_GLOBAL_DEPTH_OFFSET_ENABLE  (1<<25)
+#define S5_FOG_ENABLE                  (1<<24)
+#define S5_STENCIL_REF_SHIFT           16
+#define S5_STENCIL_REF_MASK            (0xff<<16)
+#define S5_STENCIL_TEST_FUNC_SHIFT     13
+#define S5_STENCIL_TEST_FUNC_MASK      (0x7<<13)
+#define S5_STENCIL_FAIL_SHIFT          10
+#define S5_STENCIL_FAIL_MASK           (0x7<<10)
+#define S5_STENCIL_PASS_Z_FAIL_SHIFT   7
+#define S5_STENCIL_PASS_Z_FAIL_MASK    (0x7<<7)
+#define S5_STENCIL_PASS_Z_PASS_SHIFT   4
+#define S5_STENCIL_PASS_Z_PASS_MASK    (0x7<<4)
+#define S5_STENCIL_WRITE_ENABLE        (1<<3)
+#define S5_STENCIL_TEST_ENABLE         (1<<2)
+#define S5_COLOR_DITHER_ENABLE         (1<<1)
+#define S5_LOGICOP_ENABLE              (1<<0)
+
+
+#define S6_ALPHA_TEST_ENABLE           (1<<31)
+#define S6_ALPHA_TEST_FUNC_SHIFT       28
+#define S6_ALPHA_TEST_FUNC_MASK        (0x7<<28)
+#define S6_ALPHA_REF_SHIFT             20
+#define S6_ALPHA_REF_MASK              (0xff<<20)
+#define S6_DEPTH_TEST_ENABLE           (1<<19)
+#define S6_DEPTH_TEST_FUNC_SHIFT       16
+#define S6_DEPTH_TEST_FUNC_MASK        (0x7<<16)
+#define S6_CBUF_BLEND_ENABLE           (1<<15)
+#define S6_CBUF_BLEND_FUNC_SHIFT       12
+#define S6_CBUF_BLEND_FUNC_MASK        (0x7<<12)
+#define S6_CBUF_SRC_BLEND_FACT_SHIFT   8
+#define S6_CBUF_SRC_BLEND_FACT_MASK    (0xf<<8)
+#define S6_CBUF_DST_BLEND_FACT_SHIFT   4
+#define S6_CBUF_DST_BLEND_FACT_MASK    (0xf<<4)
+#define S6_DEPTH_WRITE_ENABLE          (1<<3)
+#define S6_COLOR_WRITE_ENABLE          (1<<2)
+#define S6_TRISTRIP_PV_SHIFT           0
+#define S6_TRISTRIP_PV_MASK            (0x3<<0)
+
+#define S7_DEPTH_OFFSET_CONST_MASK     ~0
+
+
+
+#define DST_BLND_FACT(f) ((f)<<S6_CBUF_DST_BLEND_FACT_SHIFT)
+#define SRC_BLND_FACT(f) ((f)<<S6_CBUF_SRC_BLEND_FACT_SHIFT)
+#define DST_ABLND_FACT(f) ((f)<<IAB_DST_FACTOR_SHIFT)
+#define SRC_ABLND_FACT(f) ((f)<<IAB_SRC_FACTOR_SHIFT)
+
+
+
+
+/* 3DSTATE_MAP_DEINTERLACER_PARAMETERS */
+
+/* 3DSTATE_MAP_PALETTE_LOAD_32, p206 */
+#define _3DSTATE_MAP_PALETTE_LOAD_32    (CMD_3D|(0x1d<<24)|(0x8f<<16))
+/* subsequent dwords up to length (max 16) are ARGB8888 color values */
+
+/* _3DSTATE_MODES_4, p218 */
+#define _3DSTATE_MODES_4_CMD		(CMD_3D|(0x0d<<24))
+#define ENABLE_LOGIC_OP_FUNC		(1<<23)
+#define LOGIC_OP_FUNC(x)		((x)<<18)
+#define LOGICOP_MASK			(0xf<<18)
+#define MODE4_ENABLE_STENCIL_TEST_MASK	((1<<17)|(0xff00))
+#define ENABLE_STENCIL_TEST_MASK	(1<<17)
+#define STENCIL_TEST_MASK(x)		(((x)&0xff)<<8)
+#define MODE4_ENABLE_STENCIL_WRITE_MASK	((1<<16)|(0x00ff))
+#define ENABLE_STENCIL_WRITE_MASK	(1<<16)
+#define STENCIL_WRITE_MASK(x)		((x)&0xff)
+
+/* _3DSTATE_MODES_5, p220 */
+#define _3DSTATE_MODES_5_CMD		(CMD_3D|(0x0c<<24))
+#define PIPELINE_FLUSH_RENDER_CACHE	(1<<18)
+#define PIPELINE_FLUSH_TEXTURE_CACHE	(1<<16)
+
+
+/* p221 */
+#define _3DSTATE_PIXEL_SHADER_CONSTANTS  (CMD_3D|(0x1d<<24)|(0x6<<16))
+#define PS1_REG(n)                      (1<<(n))
+#define PS2_CONST_X(n)                  (n)
+#define PS3_CONST_Y(n)                  (n)
+#define PS4_CONST_Z(n)                  (n)
+#define PS5_CONST_W(n)                  (n)
+
+/* p222 */
+
+
+#define I915_MAX_TEX_INDIRECT 4
+#define I915_MAX_TEX_INSN     32
+#define I915_MAX_ALU_INSN     64
+#define I915_MAX_DECL_INSN    27
+#define I915_MAX_TEMPORARY    16
+
+
+/* Each instruction is 3 dwords long, though most don't require all
+ * this space.  Maximum of 123 instructions.  Smaller maxes per insn
+ * type.
+ */
+#define _3DSTATE_PIXEL_SHADER_PROGRAM    (CMD_3D|(0x1d<<24)|(0x5<<16))
+
+#define REG_TYPE_R                 0    /* temporary regs, no need to
+                                         * dcl, must be written before
+                                         * read -- Preserved between
+                                         * phases. 
+                                         */
+#define REG_TYPE_T                 1    /* Interpolated values, must be
+                                         * dcl'ed before use.
+                                         *
+                                         * 0..7: texture coord,
+                                         * 8: diffuse spec,
+                                         * 9: specular color,
+                                         * 10: fog parameter in w.
+                                         */
+#define REG_TYPE_CONST             2    /* Restriction: only one const
+                                         * can be referenced per
+                                         * instruction, though it may be
+                                         * selected for multiple inputs.
+                                         * Constants not initialized
+                                         * default to zero.
+                                         */
+#define REG_TYPE_S                 3    /* sampler */
+#define REG_TYPE_OC                4    /* output color (rgba) */
+#define REG_TYPE_OD                5    /* output depth (w), xyz are
+                                         * temporaries.  If not written,
+                                         * interpolated depth is used?
+                                         */
+#define REG_TYPE_U                 6    /* unpreserved temporaries */
+#define REG_TYPE_MASK              0x7
+#define REG_NR_MASK                0xf
+
+
+/* REG_TYPE_T:
+ */
+#define T_TEX0     0
+#define T_TEX1     1
+#define T_TEX2     2
+#define T_TEX3     3
+#define T_TEX4     4
+#define T_TEX5     5
+#define T_TEX6     6
+#define T_TEX7     7
+#define T_DIFFUSE  8
+#define T_SPECULAR 9
+#define T_FOG_W    10           /* interpolated fog is in W coord */
+
+/* Arithmetic instructions */
+
+/* .replicate_swizzle == selection and replication of a particular
+ * scalar channel, ie., .xxxx, .yyyy, .zzzz or .wwww 
+ */
+#define A0_NOP    (0x0<<24)     /* no operation */
+#define A0_ADD    (0x1<<24)     /* dst = src0 + src1 */
+#define A0_MOV    (0x2<<24)     /* dst = src0 */
+#define A0_MUL    (0x3<<24)     /* dst = src0 * src1 */
+#define A0_MAD    (0x4<<24)     /* dst = src0 * src1 + src2 */
+#define A0_DP2ADD (0x5<<24)     /* dst.xyzw = src0.xy dot src1.xy + src2.replicate_swizzle */
+#define A0_DP3    (0x6<<24)     /* dst.xyzw = src0.xyz dot src1.xyz */
+#define A0_DP4    (0x7<<24)     /* dst.xyzw = src0.xyzw dot src1.xyzw */
+#define A0_FRC    (0x8<<24)     /* dst = src0 - floor(src0) */
+#define A0_RCP    (0x9<<24)     /* dst.xyzw = 1/(src0.replicate_swizzle) */
+#define A0_RSQ    (0xa<<24)     /* dst.xyzw = 1/(sqrt(abs(src0.replicate_swizzle))) */
+#define A0_EXP    (0xb<<24)     /* dst.xyzw = exp2(src0.replicate_swizzle) */
+#define A0_LOG    (0xc<<24)     /* dst.xyzw = log2(abs(src0.replicate_swizzle)) */
+#define A0_CMP    (0xd<<24)     /* dst = (src0 >= 0.0) ? src1 : src2 */
+#define A0_MIN    (0xe<<24)     /* dst = (src0 < src1) ? src0 : src1 */
+#define A0_MAX    (0xf<<24)     /* dst = (src0 >= src1) ? src0 : src1 */
+#define A0_FLR    (0x10<<24)    /* dst = floor(src0) */
+#define A0_MOD    (0x11<<24)    /* dst = src0 fmod 1.0 */
+#define A0_TRC    (0x12<<24)    /* dst = int(src0) */
+#define A0_SGE    (0x13<<24)    /* dst = src0 >= src1 ? 1.0 : 0.0 */
+#define A0_SLT    (0x14<<24)    /* dst = src0 < src1 ? 1.0 : 0.0 */
+#define A0_DEST_SATURATE                 (1<<22)
+#define A0_DEST_TYPE_SHIFT                19
+/* Allow: R, OC, OD, U */
+#define A0_DEST_NR_SHIFT                 14
+/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */
+#define A0_DEST_CHANNEL_X                (1<<10)
+#define A0_DEST_CHANNEL_Y                (2<<10)
+#define A0_DEST_CHANNEL_Z                (4<<10)
+#define A0_DEST_CHANNEL_W                (8<<10)
+#define A0_DEST_CHANNEL_ALL              (0xf<<10)
+#define A0_DEST_CHANNEL_SHIFT            10
+#define A0_SRC0_TYPE_SHIFT               7
+#define A0_SRC0_NR_SHIFT                 2
+
+#define A0_DEST_CHANNEL_XY              (A0_DEST_CHANNEL_X|A0_DEST_CHANNEL_Y)
+#define A0_DEST_CHANNEL_XYZ             (A0_DEST_CHANNEL_XY|A0_DEST_CHANNEL_Z)
+
+
+#define SRC_X        0
+#define SRC_Y        1
+#define SRC_Z        2
+#define SRC_W        3
+#define SRC_ZERO     4
+#define SRC_ONE      5
+
+#define A1_SRC0_CHANNEL_X_NEGATE         (1<<31)
+#define A1_SRC0_CHANNEL_X_SHIFT          28
+#define A1_SRC0_CHANNEL_Y_NEGATE         (1<<27)
+#define A1_SRC0_CHANNEL_Y_SHIFT          24
+#define A1_SRC0_CHANNEL_Z_NEGATE         (1<<23)
+#define A1_SRC0_CHANNEL_Z_SHIFT          20
+#define A1_SRC0_CHANNEL_W_NEGATE         (1<<19)
+#define A1_SRC0_CHANNEL_W_SHIFT          16
+#define A1_SRC1_TYPE_SHIFT               13
+#define A1_SRC1_NR_SHIFT                 8
+#define A1_SRC1_CHANNEL_X_NEGATE         (1<<7)
+#define A1_SRC1_CHANNEL_X_SHIFT          4
+#define A1_SRC1_CHANNEL_Y_NEGATE         (1<<3)
+#define A1_SRC1_CHANNEL_Y_SHIFT          0
+
+#define A2_SRC1_CHANNEL_Z_NEGATE         (1<<31)
+#define A2_SRC1_CHANNEL_Z_SHIFT          28
+#define A2_SRC1_CHANNEL_W_NEGATE         (1<<27)
+#define A2_SRC1_CHANNEL_W_SHIFT          24
+#define A2_SRC2_TYPE_SHIFT               21
+#define A2_SRC2_NR_SHIFT                 16
+#define A2_SRC2_CHANNEL_X_NEGATE         (1<<15)
+#define A2_SRC2_CHANNEL_X_SHIFT          12
+#define A2_SRC2_CHANNEL_Y_NEGATE         (1<<11)
+#define A2_SRC2_CHANNEL_Y_SHIFT          8
+#define A2_SRC2_CHANNEL_Z_NEGATE         (1<<7)
+#define A2_SRC2_CHANNEL_Z_SHIFT          4
+#define A2_SRC2_CHANNEL_W_NEGATE         (1<<3)
+#define A2_SRC2_CHANNEL_W_SHIFT          0
+
+
+
+/* Texture instructions */
+#define T0_TEXLD     (0x15<<24) /* Sample texture using predeclared
+                                 * sampler and address, and output
+                                 * filtered texel data to destination
+                                 * register */
+#define T0_TEXLDP    (0x16<<24) /* Same as texld but performs a
+                                 * perspective divide of the texture
+                                 * coordinate .xyz values by .w before
+                                 * sampling. */
+#define T0_TEXLDB    (0x17<<24) /* Same as texld but biases the
+                                 * computed LOD by w.  Only S4.6 two's
+                                 * comp is used.  This implies that a
+                                 * float to fixed conversion is
+                                 * done. */
+#define T0_TEXKILL   (0x18<<24) /* Does not perform a sampling
+                                 * operation.  Simply kills the pixel
+                                 * if any channel of the address
+                                 * register is < 0.0. */
+#define T0_DEST_TYPE_SHIFT                19
+/* Allow: R, OC, OD, U */
+/* Note: U (unpreserved) regs do not retain their values between
+ * phases (cannot be used for feedback) 
+ *
+ * Note: oC and OD registers can only be used as the destination of a
+ * texture instruction once per phase (this is an implementation
+ * restriction). 
+ */
+#define T0_DEST_NR_SHIFT                 14
+/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */
+#define T0_SAMPLER_NR_SHIFT              0      /* This field ignored for TEXKILL */
+#define T0_SAMPLER_NR_MASK               (0xf<<0)
+
+#define T1_ADDRESS_REG_TYPE_SHIFT        24     /* Reg to use as texture coord */
+/* Allow R, T, OC, OD -- R, OC, OD are 'dependent' reads, new program phase */
+#define T1_ADDRESS_REG_NR_SHIFT          17
+#define T2_MBZ                           0
+
+/* Declaration instructions */
+#define D0_DCL       (0x19<<24) /* Declare a t (interpolated attrib)
+                                 * register or an s (sampler)
+                                 * register. */
+#define D0_SAMPLE_TYPE_SHIFT              22
+#define D0_SAMPLE_TYPE_2D                 (0x0<<22)
+#define D0_SAMPLE_TYPE_CUBE               (0x1<<22)
+#define D0_SAMPLE_TYPE_VOLUME             (0x2<<22)
+#define D0_SAMPLE_TYPE_MASK               (0x3<<22)
+
+#define D0_TYPE_SHIFT                19
+/* Allow: T, S */
+#define D0_NR_SHIFT                  14
+/* Allow T: 0..10, S: 0..15 */
+#define D0_CHANNEL_X                (1<<10)
+#define D0_CHANNEL_Y                (2<<10)
+#define D0_CHANNEL_Z                (4<<10)
+#define D0_CHANNEL_W                (8<<10)
+#define D0_CHANNEL_ALL              (0xf<<10)
+#define D0_CHANNEL_NONE             (0<<10)
+
+#define D0_CHANNEL_XY               (D0_CHANNEL_X|D0_CHANNEL_Y)
+#define D0_CHANNEL_XYZ              (D0_CHANNEL_XY|D0_CHANNEL_Z)
+
+/* I915 Errata: Do not allow (xz), (xw), (xzw) combinations for diffuse
+ * or specular declarations. 
+ *
+ * For T dcls, only allow: (x), (xy), (xyz), (w), (xyzw) 
+ *
+ * Must be zero for S (sampler) dcls
+ */
+#define D1_MBZ                          0
+#define D2_MBZ                          0
+
+
+
+/* p207 */
+#define _3DSTATE_MAP_STATE               (CMD_3D|(0x1d<<24)|(0x0<<16))
+
+#define MS1_MAPMASK_SHIFT               0
+#define MS1_MAPMASK_MASK                (0x8fff<<0)
+
+#define MS2_UNTRUSTED_SURFACE           (1<<31)
+#define MS2_ADDRESS_MASK                0xfffffffc
+#define MS2_VERTICAL_LINE_STRIDE        (1<<1)
+#define MS2_VERTICAL_OFFSET             (1<<1)
+
+#define MS3_HEIGHT_SHIFT              21
+#define MS3_WIDTH_SHIFT               10
+#define MS3_PALETTE_SELECT            (1<<9)
+#define MS3_MAPSURF_FORMAT_SHIFT      7
+#define MS3_MAPSURF_FORMAT_MASK       (0x7<<7)
+#define    MAPSURF_8BIT		 	   (1<<7)
+#define    MAPSURF_16BIT		   (2<<7)
+#define    MAPSURF_32BIT		   (3<<7)
+#define    MAPSURF_422			   (5<<7)
+#define    MAPSURF_COMPRESSED		   (6<<7)
+#define    MAPSURF_4BIT_INDEXED		   (7<<7)
+#define MS3_MT_FORMAT_MASK         (0x7 << 3)
+#define MS3_MT_FORMAT_SHIFT        3
+#define    MT_4BIT_IDX_ARGB8888	           (7<<3)       /* SURFACE_4BIT_INDEXED */
+#define    MT_8BIT_I8		           (0<<3)       /* SURFACE_8BIT */
+#define    MT_8BIT_L8		           (1<<3)
+#define    MT_8BIT_A8		           (4<<3)
+#define    MT_8BIT_MONO8	           (5<<3)
+#define    MT_16BIT_RGB565 		   (0<<3)       /* SURFACE_16BIT */
+#define    MT_16BIT_ARGB1555		   (1<<3)
+#define    MT_16BIT_ARGB4444		   (2<<3)
+#define    MT_16BIT_AY88		   (3<<3)
+#define    MT_16BIT_88DVDU	           (5<<3)
+#define    MT_16BIT_BUMP_655LDVDU	   (6<<3)
+#define    MT_16BIT_I16	                   (7<<3)
+#define    MT_16BIT_L16	                   (8<<3)
+#define    MT_16BIT_A16	                   (9<<3)
+#define    MT_32BIT_ARGB8888		   (0<<3)       /* SURFACE_32BIT */
+#define    MT_32BIT_ABGR8888		   (1<<3)
+#define    MT_32BIT_XRGB8888		   (2<<3)
+#define    MT_32BIT_XBGR8888		   (3<<3)
+#define    MT_32BIT_QWVU8888		   (4<<3)
+#define    MT_32BIT_AXVU8888		   (5<<3)
+#define    MT_32BIT_LXVU8888	           (6<<3)
+#define    MT_32BIT_XLVU8888	           (7<<3)
+#define    MT_32BIT_ARGB2101010	           (8<<3)
+#define    MT_32BIT_ABGR2101010	           (9<<3)
+#define    MT_32BIT_AWVU2101010	           (0xA<<3)
+#define    MT_32BIT_GR1616	           (0xB<<3)
+#define    MT_32BIT_VU1616	           (0xC<<3)
+#define    MT_32BIT_xI824	           (0xD<<3)
+#define    MT_32BIT_xA824	           (0xE<<3)
+#define    MT_32BIT_xL824	           (0xF<<3)
+#define    MT_422_YCRCB_SWAPY	           (0<<3)       /* SURFACE_422 */
+#define    MT_422_YCRCB_NORMAL	           (1<<3)
+#define    MT_422_YCRCB_SWAPUV	           (2<<3)
+#define    MT_422_YCRCB_SWAPUVY	           (3<<3)
+#define    MT_COMPRESS_DXT1		   (0<<3)       /* SURFACE_COMPRESSED */
+#define    MT_COMPRESS_DXT2_3	           (1<<3)
+#define    MT_COMPRESS_DXT4_5	           (2<<3)
+#define    MT_COMPRESS_FXT1		   (3<<3)
+#define    MT_COMPRESS_DXT1_RGB		   (4<<3)
+#define MS3_USE_FENCE_REGS              (1<<2)
+#define MS3_TILED_SURFACE             (1<<1)
+#define MS3_TILE_WALK                 (1<<0)
+
+#define MS4_PITCH_SHIFT                 21
+#define MS4_CUBE_FACE_ENA_NEGX          (1<<20)
+#define MS4_CUBE_FACE_ENA_POSX          (1<<19)
+#define MS4_CUBE_FACE_ENA_NEGY          (1<<18)
+#define MS4_CUBE_FACE_ENA_POSY          (1<<17)
+#define MS4_CUBE_FACE_ENA_NEGZ          (1<<16)
+#define MS4_CUBE_FACE_ENA_POSZ          (1<<15)
+#define MS4_CUBE_FACE_ENA_MASK          (0x3f<<15)
+#define MS4_MAX_LOD_SHIFT		9
+#define MS4_MAX_LOD_MASK		(0x3f<<9)
+#define MS4_MIP_LAYOUT_LEGACY           (0<<8)
+#define MS4_MIP_LAYOUT_BELOW_LPT        (0<<8)
+#define MS4_MIP_LAYOUT_RIGHT_LPT        (1<<8)
+#define MS4_VOLUME_DEPTH_SHIFT          0
+#define MS4_VOLUME_DEPTH_MASK           (0xff<<0)
+
+/* p244 */
+#define _3DSTATE_SAMPLER_STATE         (CMD_3D|(0x1d<<24)|(0x1<<16))
+
+#define SS1_MAPMASK_SHIFT               0
+#define SS1_MAPMASK_MASK                (0x8fff<<0)
+
+#define SS2_REVERSE_GAMMA_ENABLE        (1<<31)
+#define SS2_PACKED_TO_PLANAR_ENABLE     (1<<30)
+#define SS2_COLORSPACE_CONVERSION       (1<<29)
+#define SS2_CHROMAKEY_SHIFT             27
+#define SS2_BASE_MIP_LEVEL_SHIFT        22
+#define SS2_BASE_MIP_LEVEL_MASK         (0x1f<<22)
+#define SS2_MIP_FILTER_SHIFT            20
+#define SS2_MIP_FILTER_MASK             (0x3<<20)
+#define   MIPFILTER_NONE       	0
+#define   MIPFILTER_NEAREST	1
+#define   MIPFILTER_LINEAR	3
+#define SS2_MAG_FILTER_SHIFT          17
+#define SS2_MAG_FILTER_MASK           (0x7<<17)
+#define   FILTER_NEAREST	0
+#define   FILTER_LINEAR		1
+#define   FILTER_ANISOTROPIC	2
+#define   FILTER_4X4_1    	3
+#define   FILTER_4X4_2    	4
+#define   FILTER_4X4_FLAT 	5
+#define   FILTER_6X5_MONO   	6       /* XXX - check */
+#define SS2_MIN_FILTER_SHIFT          14
+#define SS2_MIN_FILTER_MASK           (0x7<<14)
+#define SS2_LOD_BIAS_SHIFT            5
+#define SS2_LOD_BIAS_ONE              (0x10<<5)
+#define SS2_LOD_BIAS_MASK             (0x1ff<<5)
+/* Shadow requires:
+ *  MT_X8{I,L,A}24 or MT_{I,L,A}16 texture format
+ *  FILTER_4X4_x  MIN and MAG filters
+ */
+#define SS2_SHADOW_ENABLE             (1<<4)
+#define SS2_MAX_ANISO_MASK            (1<<3)
+#define SS2_MAX_ANISO_2               (0<<3)
+#define SS2_MAX_ANISO_4               (1<<3)
+#define SS2_SHADOW_FUNC_SHIFT         0
+#define SS2_SHADOW_FUNC_MASK          (0x7<<0)
+/* SS2_SHADOW_FUNC values: see COMPAREFUNC_* */
+
+#define SS3_MIN_LOD_SHIFT            24
+#define SS3_MIN_LOD_ONE              (0x10<<24)
+#define SS3_MIN_LOD_MASK             (0xff<<24)
+#define SS3_KILL_PIXEL_ENABLE        (1<<17)
+#define SS3_TCX_ADDR_MODE_SHIFT      12
+#define SS3_TCX_ADDR_MODE_MASK       (0x7<<12)
+#define   TEXCOORDMODE_WRAP		0
+#define   TEXCOORDMODE_MIRROR		1
+#define   TEXCOORDMODE_CLAMP_EDGE	2
+#define   TEXCOORDMODE_CUBE       	3
+#define   TEXCOORDMODE_CLAMP_BORDER	4
+#define   TEXCOORDMODE_MIRROR_ONCE      5
+#define SS3_TCY_ADDR_MODE_SHIFT      9
+#define SS3_TCY_ADDR_MODE_MASK       (0x7<<9)
+#define SS3_TCZ_ADDR_MODE_SHIFT      6
+#define SS3_TCZ_ADDR_MODE_MASK       (0x7<<6)
+#define SS3_NORMALIZED_COORDS        (1<<5)
+#define SS3_TEXTUREMAP_INDEX_SHIFT   1
+#define SS3_TEXTUREMAP_INDEX_MASK    (0xf<<1)
+#define SS3_DEINTERLACER_ENABLE      (1<<0)
+
+#define SS4_BORDER_COLOR_MASK        (~0)
+
+/* 3DSTATE_SPAN_STIPPLE, p258
+ */
+#define _3DSTATE_STIPPLE           ((0x3<<29)|(0x1d<<24)|(0x83<<16))
+#define ST1_ENABLE               (1<<16)
+#define ST1_MASK                 (0xffff)
+
+#define _3DSTATE_DEFAULT_Z          ((0x3<<29)|(0x1d<<24)|(0x98<<16))
+#define _3DSTATE_DEFAULT_DIFFUSE    ((0x3<<29)|(0x1d<<24)|(0x99<<16))
+#define _3DSTATE_DEFAULT_SPECULAR   ((0x3<<29)|(0x1d<<24)|(0x9a<<16))
+
+
+#define MI_FLUSH                   ((0<<29)|(4<<23))
+#define FLUSH_MAP_CACHE            (1<<0)
+#define INHIBIT_FLUSH_RENDER_CACHE (1<<2)
+
+
+#define CMD_3D (0x3<<29)
+
+
+#define _3DPRIMITIVE         ((0x3<<29)|(0x1f<<24))
+#define PRIM_INDIRECT            (1<<23)
+#define PRIM_INLINE              (0<<23)
+#define PRIM_INDIRECT_SEQUENTIAL (0<<17)
+#define PRIM_INDIRECT_ELTS       (1<<17)
+
+#define PRIM3D_TRILIST		(0x0<<18)
+#define PRIM3D_TRISTRIP 	(0x1<<18)
+#define PRIM3D_TRISTRIP_RVRSE	(0x2<<18)
+#define PRIM3D_TRIFAN		(0x3<<18)
+#define PRIM3D_POLY		(0x4<<18)
+#define PRIM3D_LINELIST 	(0x5<<18)
+#define PRIM3D_LINESTRIP	(0x6<<18)
+#define PRIM3D_RECTLIST 	(0x7<<18)
+#define PRIM3D_POINTLIST	(0x8<<18)
+#define PRIM3D_DIB		(0x9<<18)
+#define PRIM3D_MASK		(0x1f<<18)
+
+#define I915PACKCOLOR4444(r,g,b,a) \
+  ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4))
+
+#define I915PACKCOLOR1555(r,g,b,a) \
+  ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \
+    ((a) ? 0x8000 : 0))
+
+#define I915PACKCOLOR565(r,g,b) \
+  ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3))
+
+#define I915PACKCOLOR8888(r,g,b,a) \
+  ((a<<24) | (r<<16) | (g<<8) | b)
+
+
+
+
+#define BR00_BITBLT_CLIENT   0x40000000
+#define BR00_OP_COLOR_BLT    0x10000000
+#define BR00_OP_SRC_COPY_BLT 0x10C00000
+#define BR13_SOLID_PATTERN   0x80000000
+
+#define XY_COLOR_BLT_CMD		((2<<29)|(0x50<<22)|0x4)
+#define XY_COLOR_BLT_WRITE_ALPHA	(1<<21)
+#define XY_COLOR_BLT_WRITE_RGB		(1<<20)
+
+#define XY_SRC_COPY_BLT_CMD             ((2<<29)|(0x53<<22)|6)
+#define XY_SRC_COPY_BLT_WRITE_ALPHA     (1<<21)
+#define XY_SRC_COPY_BLT_WRITE_RGB       (1<<20)
+
+#define MI_WAIT_FOR_EVENT               ((0x3<<23))
+#define MI_WAIT_FOR_PLANE_B_FLIP        (1<<6)
+#define MI_WAIT_FOR_PLANE_A_FLIP        (1<<2)
+
+#define MI_BATCH_BUFFER                 (0x30<<23)
+#define MI_BATCH_BUFFER_START           (0x31<<23)
+#define MI_BATCH_BUFFER_END             (0xa<<23)
+
+
+
+#define COMPAREFUNC_ALWAYS		0
+#define COMPAREFUNC_NEVER		0x1
+#define COMPAREFUNC_LESS		0x2
+#define COMPAREFUNC_EQUAL		0x3
+#define COMPAREFUNC_LEQUAL		0x4
+#define COMPAREFUNC_GREATER		0x5
+#define COMPAREFUNC_NOTEQUAL		0x6
+#define COMPAREFUNC_GEQUAL		0x7
+
+#define STENCILOP_KEEP			0
+#define STENCILOP_ZERO			0x1
+#define STENCILOP_REPLACE		0x2
+#define STENCILOP_INCRSAT		0x3
+#define STENCILOP_DECRSAT		0x4
+#define STENCILOP_INCR			0x5
+#define STENCILOP_DECR			0x6
+#define STENCILOP_INVERT		0x7
+
+#define LOGICOP_CLEAR			0
+#define LOGICOP_NOR			0x1
+#define LOGICOP_AND_INV 		0x2
+#define LOGICOP_COPY_INV		0x3
+#define LOGICOP_AND_RVRSE		0x4
+#define LOGICOP_INV			0x5
+#define LOGICOP_XOR			0x6
+#define LOGICOP_NAND			0x7
+#define LOGICOP_AND			0x8
+#define LOGICOP_EQUIV			0x9
+#define LOGICOP_NOOP			0xa
+#define LOGICOP_OR_INV			0xb
+#define LOGICOP_COPY			0xc
+#define LOGICOP_OR_RVRSE		0xd
+#define LOGICOP_OR			0xe
+#define LOGICOP_SET			0xf
+
+#define BLENDFACT_ZERO			0x01
+#define BLENDFACT_ONE			0x02
+#define BLENDFACT_SRC_COLR		0x03
+#define BLENDFACT_INV_SRC_COLR 		0x04
+#define BLENDFACT_SRC_ALPHA		0x05
+#define BLENDFACT_INV_SRC_ALPHA 	0x06
+#define BLENDFACT_DST_ALPHA		0x07
+#define BLENDFACT_INV_DST_ALPHA 	0x08
+#define BLENDFACT_DST_COLR		0x09
+#define BLENDFACT_INV_DST_COLR		0x0a
+#define BLENDFACT_SRC_ALPHA_SATURATE	0x0b
+#define BLENDFACT_CONST_COLOR		0x0c
+#define BLENDFACT_INV_CONST_COLOR	0x0d
+#define BLENDFACT_CONST_ALPHA		0x0e
+#define BLENDFACT_INV_CONST_ALPHA	0x0f
+#define BLENDFACT_MASK          	0x0f
+
+#define PCI_CHIP_I915_G			0x2582
+#define PCI_CHIP_I915_GM		0x2592
+#define PCI_CHIP_I945_G			0x2772
+#define PCI_CHIP_I945_GM		0x27A2
+#define PCI_CHIP_I945_GME		0x27AE
+#define PCI_CHIP_G33_G			0x29C2
+#define PCI_CHIP_Q35_G			0x29B2
+#define PCI_CHIP_Q33_G			0x29D2
+
+
+#endif
diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c
new file mode 100644
index 0000000000..39e48105b3
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_screen.c
@@ -0,0 +1,288 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "util/u_memory.h"
+#include "util/u_simple_screen.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_string.h"
+
+#include "i915_reg.h"
+#include "i915_context.h"
+#include "i915_screen.h"
+#include "i915_texture.h"
+
+
+static const char *
+i915_get_vendor( struct pipe_screen *pscreen )
+{
+   return "Tungsten Graphics, Inc.";
+}
+
+
+static const char *
+i915_get_name( struct pipe_screen *pscreen )
+{
+   static char buffer[128];
+   const char *chipset;
+
+   switch (i915_screen(pscreen)->pci_id) {
+   case PCI_CHIP_I915_G:
+      chipset = "915G";
+      break;
+   case PCI_CHIP_I915_GM:
+      chipset = "915GM";
+      break;
+   case PCI_CHIP_I945_G:
+      chipset = "945G";
+      break;
+   case PCI_CHIP_I945_GM:
+      chipset = "945GM";
+      break;
+   case PCI_CHIP_I945_GME:
+      chipset = "945GME";
+      break;
+   case PCI_CHIP_G33_G:
+      chipset = "G33";
+      break;
+   case PCI_CHIP_Q35_G:
+      chipset = "Q35";
+      break;
+   case PCI_CHIP_Q33_G:
+      chipset = "Q33";
+      break;
+   default:
+      chipset = "unknown";
+      break;
+   }
+
+   util_snprintf(buffer, sizeof(buffer), "i915 (chipset: %s)", chipset);
+   return buffer;
+}
+
+
+static int
+i915_get_param(struct pipe_screen *screen, int param)
+{
+   switch (param) {
+   case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+      return 8;
+   case PIPE_CAP_NPOT_TEXTURES:
+      return 1;
+   case PIPE_CAP_TWO_SIDED_STENCIL:
+      return 1;
+   case PIPE_CAP_GLSL:
+      return 0;
+   case PIPE_CAP_S3TC:
+      return 0;
+   case PIPE_CAP_ANISOTROPIC_FILTER:
+      return 0;
+   case PIPE_CAP_POINT_SPRITE:
+      return 0;
+   case PIPE_CAP_MAX_RENDER_TARGETS:
+      return 1;
+   case PIPE_CAP_OCCLUSION_QUERY:
+      return 0;
+   case PIPE_CAP_TEXTURE_SHADOW_MAP:
+      return 1;
+   case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+      return 11; /* max 1024x1024 */
+   case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+      return 8;  /* max 128x128x128 */
+   case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+      return 11; /* max 1024x1024 */
+   default:
+      return 0;
+   }
+}
+
+
+static float
+i915_get_paramf(struct pipe_screen *screen, int param)
+{
+   switch (param) {
+   case PIPE_CAP_MAX_LINE_WIDTH:
+      /* fall-through */
+   case PIPE_CAP_MAX_LINE_WIDTH_AA:
+      return 7.5;
+
+   case PIPE_CAP_MAX_POINT_WIDTH:
+      /* fall-through */
+   case PIPE_CAP_MAX_POINT_WIDTH_AA:
+      return 255.0;
+
+   case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+      return 4.0;
+
+   case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+      return 16.0;
+
+   default:
+      return 0;
+   }
+}
+
+
+static boolean
+i915_is_format_supported( struct pipe_screen *screen,
+                          enum pipe_format format, 
+                          enum pipe_texture_target target,
+                          unsigned tex_usage, 
+                          unsigned geom_flags )
+{
+   static const enum pipe_format tex_supported[] = {
+      PIPE_FORMAT_R8G8B8A8_UNORM,
+      PIPE_FORMAT_A8R8G8B8_UNORM,
+      PIPE_FORMAT_R5G6B5_UNORM,
+      PIPE_FORMAT_L8_UNORM,
+      PIPE_FORMAT_A8_UNORM,
+      PIPE_FORMAT_I8_UNORM,
+      PIPE_FORMAT_A8L8_UNORM,
+      PIPE_FORMAT_YCBCR,
+      PIPE_FORMAT_YCBCR_REV,
+      PIPE_FORMAT_S8Z24_UNORM,
+      PIPE_FORMAT_NONE  /* list terminator */
+   };
+   static const enum pipe_format surface_supported[] = {
+      PIPE_FORMAT_A8R8G8B8_UNORM,
+      PIPE_FORMAT_R5G6B5_UNORM,
+      PIPE_FORMAT_S8Z24_UNORM,
+      /*PIPE_FORMAT_R16G16B16A16_SNORM,*/
+      PIPE_FORMAT_NONE  /* list terminator */
+   };
+   const enum pipe_format *list;
+   uint i;
+
+   if(tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET)
+      list = surface_supported;
+   else
+      list = tex_supported;
+
+   for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) {
+      if (list[i] == format)
+         return TRUE;
+   }
+
+   return FALSE;
+}
+
+
+static void
+i915_destroy_screen( struct pipe_screen *screen )
+{
+   struct pipe_winsys *winsys = screen->winsys;
+
+   if(winsys->destroy)
+      winsys->destroy(winsys);
+
+   FREE(screen);
+}
+
+
+static void *
+i915_surface_map( struct pipe_screen *screen,
+                  struct pipe_surface *surface,
+                  unsigned flags )
+{
+   struct i915_texture *tex = (struct i915_texture *)surface->texture;
+   char *map = pipe_buffer_map( screen, tex->buffer, flags );
+   if (map == NULL)
+      return NULL;
+
+   if (surface->texture &&
+       (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) 
+   {
+      /* Do something to notify contexts of a texture change.  
+       */
+      /* i915_screen(screen)->timestamp++; */
+   }
+   
+   return map + surface->offset;
+}
+
+static void
+i915_surface_unmap(struct pipe_screen *screen,
+                   struct pipe_surface *surface)
+{
+   struct i915_texture *tex = (struct i915_texture *)surface->texture;
+   pipe_buffer_unmap( screen, tex->buffer );
+}
+
+
+
+/**
+ * Create a new i915_screen object
+ */
+struct pipe_screen *
+i915_create_screen(struct pipe_winsys *winsys, uint pci_id)
+{
+   struct i915_screen *i915screen = CALLOC_STRUCT(i915_screen);
+
+   if (!i915screen)
+      return NULL;
+
+   switch (pci_id) {
+   case PCI_CHIP_I915_G:
+   case PCI_CHIP_I915_GM:
+      i915screen->is_i945 = FALSE;
+      break;
+
+   case PCI_CHIP_I945_G:
+   case PCI_CHIP_I945_GM:
+   case PCI_CHIP_I945_GME:
+   case PCI_CHIP_G33_G:
+   case PCI_CHIP_Q33_G:
+   case PCI_CHIP_Q35_G:
+      i915screen->is_i945 = TRUE;
+      break;
+
+   default:
+      debug_printf("%s: unknown pci id 0x%x, cannot create screen\n", 
+                   __FUNCTION__, pci_id);
+      return NULL;
+   }
+
+   i915screen->pci_id = pci_id;
+
+   i915screen->screen.winsys = winsys;
+
+   i915screen->screen.destroy = i915_destroy_screen;
+
+   i915screen->screen.get_name = i915_get_name;
+   i915screen->screen.get_vendor = i915_get_vendor;
+   i915screen->screen.get_param = i915_get_param;
+   i915screen->screen.get_paramf = i915_get_paramf;
+   i915screen->screen.is_format_supported = i915_is_format_supported;
+   i915screen->screen.surface_map = i915_surface_map;
+   i915screen->screen.surface_unmap = i915_surface_unmap;
+
+   i915_init_screen_texture_functions(&i915screen->screen);
+   u_simple_screen_init(&i915screen->screen);
+
+   return &i915screen->screen;
+}
diff --git a/src/gallium/drivers/i915simple/i915_screen.h b/src/gallium/drivers/i915simple/i915_screen.h
new file mode 100644
index 0000000000..73b0ff05ce
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_screen.h
@@ -0,0 +1,69 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef I915_SCREEN_H
+#define I915_SCREEN_H
+
+
+#include "pipe/p_screen.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/**
+ * Subclass of pipe_screen
+ */
+struct i915_screen
+{
+   struct pipe_screen screen;
+
+   boolean is_i945;
+   uint pci_id;
+};
+
+
+/** cast wrapper */
+static INLINE struct i915_screen *
+i915_screen(struct pipe_screen *pscreen)
+{
+   return (struct i915_screen *) pscreen;
+}
+
+
+extern struct pipe_screen *
+i915_create_screen(struct pipe_winsys *winsys, uint pci_id);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* I915_SCREEN_H */
diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c
new file mode 100644
index 0000000000..273e74002a
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_state.c
@@ -0,0 +1,788 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "draw/draw_context.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "tgsi/tgsi_parse.h"
+
+#include "i915_context.h"
+#include "i915_reg.h"
+#include "i915_state.h"
+#include "i915_state_inlines.h"
+#include "i915_fpc.h"
+
+/* The i915 (and related graphics cores) do not support GL_CLAMP.  The
+ * Intel drivers for "other operating systems" implement GL_CLAMP as
+ * GL_CLAMP_TO_EDGE, so the same is done here.
+ */
+static unsigned
+translate_wrap_mode(unsigned wrap)
+{
+   switch (wrap) {
+   case PIPE_TEX_WRAP_REPEAT:
+      return TEXCOORDMODE_WRAP;
+   case PIPE_TEX_WRAP_CLAMP:
+      return TEXCOORDMODE_CLAMP_EDGE;   /* not quite correct */
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      return TEXCOORDMODE_CLAMP_EDGE;
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      return TEXCOORDMODE_CLAMP_BORDER;
+//   case PIPE_TEX_WRAP_MIRRORED_REPEAT:
+//      return TEXCOORDMODE_MIRROR;
+   default:
+      return TEXCOORDMODE_WRAP;
+   }
+}
+
+static unsigned translate_img_filter( unsigned filter )
+{
+   switch (filter) {
+   case PIPE_TEX_FILTER_NEAREST:
+      return FILTER_NEAREST;
+   case PIPE_TEX_FILTER_LINEAR:
+      return FILTER_LINEAR;
+   case PIPE_TEX_FILTER_ANISO:
+      return FILTER_ANISOTROPIC;
+   default:
+      assert(0);
+      return FILTER_NEAREST;
+   }
+}
+
+static unsigned translate_mip_filter( unsigned filter )
+{
+   switch (filter) {
+   case PIPE_TEX_MIPFILTER_NONE:
+      return MIPFILTER_NONE;
+   case PIPE_TEX_MIPFILTER_NEAREST:
+      return MIPFILTER_NEAREST;
+   case PIPE_TEX_MIPFILTER_LINEAR:
+      return MIPFILTER_LINEAR;
+   default:
+      assert(0);
+      return MIPFILTER_NONE;
+   }
+}
+
+
+/* None of this state is actually used for anything yet.
+ */
+static void *
+i915_create_blend_state(struct pipe_context *pipe,
+                        const struct pipe_blend_state *blend)
+{
+   struct i915_blend_state *cso_data = CALLOC_STRUCT( i915_blend_state );
+
+   {
+      unsigned eqRGB  = blend->rgb_func;
+      unsigned srcRGB = blend->rgb_src_factor;
+      unsigned dstRGB = blend->rgb_dst_factor;
+
+      unsigned eqA    = blend->alpha_func;
+      unsigned srcA   = blend->alpha_src_factor;
+      unsigned dstA   = blend->alpha_dst_factor;
+
+      /* Special handling for MIN/MAX filter modes handled at
+       * state_tracker level.
+       */
+
+      if (srcA != srcRGB ||
+	  dstA != dstRGB ||
+	  eqA != eqRGB) {
+
+	 cso_data->iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD |
+                          IAB_MODIFY_ENABLE |
+                          IAB_ENABLE |
+                          IAB_MODIFY_FUNC |
+                          IAB_MODIFY_SRC_FACTOR |
+                          IAB_MODIFY_DST_FACTOR |
+                          SRC_ABLND_FACT(i915_translate_blend_factor(srcA)) |
+                          DST_ABLND_FACT(i915_translate_blend_factor(dstA)) |
+                          (i915_translate_blend_func(eqA) << IAB_FUNC_SHIFT));
+      }
+      else {
+	 cso_data->iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD |
+                          IAB_MODIFY_ENABLE |
+                          0);
+      }
+   }
+
+   cso_data->modes4 |= (_3DSTATE_MODES_4_CMD |
+                        ENABLE_LOGIC_OP_FUNC |
+                        LOGIC_OP_FUNC(i915_translate_logic_op(blend->logicop_func)));
+
+   if (blend->logicop_enable)
+      cso_data->LIS5 |= S5_LOGICOP_ENABLE;
+
+   if (blend->dither)
+      cso_data->LIS5 |= S5_COLOR_DITHER_ENABLE;
+
+   if ((blend->colormask & PIPE_MASK_R) == 0)
+      cso_data->LIS5 |= S5_WRITEDISABLE_RED;
+
+   if ((blend->colormask & PIPE_MASK_G) == 0)
+      cso_data->LIS5 |= S5_WRITEDISABLE_GREEN;
+
+   if ((blend->colormask & PIPE_MASK_B) == 0)
+      cso_data->LIS5 |= S5_WRITEDISABLE_BLUE;
+
+   if ((blend->colormask & PIPE_MASK_A) == 0)
+      cso_data->LIS5 |= S5_WRITEDISABLE_ALPHA;
+
+   if (blend->blend_enable) {
+      unsigned funcRGB = blend->rgb_func;
+      unsigned srcRGB  = blend->rgb_src_factor;
+      unsigned dstRGB  = blend->rgb_dst_factor;
+
+      cso_data->LIS6 |= (S6_CBUF_BLEND_ENABLE |
+                         SRC_BLND_FACT(i915_translate_blend_factor(srcRGB)) |
+                         DST_BLND_FACT(i915_translate_blend_factor(dstRGB)) |
+                         (i915_translate_blend_func(funcRGB) << S6_CBUF_BLEND_FUNC_SHIFT));
+   }
+
+   return cso_data;
+}
+
+static void i915_bind_blend_state(struct pipe_context *pipe,
+                                  void *blend)
+{
+   struct i915_context *i915 = i915_context(pipe);
+   draw_flush(i915->draw);
+
+   i915->blend = (struct i915_blend_state*)blend;
+
+   i915->dirty |= I915_NEW_BLEND;
+}
+
+
+static void i915_delete_blend_state(struct pipe_context *pipe, void *blend)
+{
+   FREE(blend);
+}
+
+static void i915_set_blend_color( struct pipe_context *pipe,
+			     const struct pipe_blend_color *blend_color )
+{
+   struct i915_context *i915 = i915_context(pipe);
+   draw_flush(i915->draw);
+
+   i915->blend_color = *blend_color;
+
+   i915->dirty |= I915_NEW_BLEND;
+}
+
+static void *
+i915_create_sampler_state(struct pipe_context *pipe,
+                          const struct pipe_sampler_state *sampler)
+{
+   struct i915_sampler_state *cso = CALLOC_STRUCT( i915_sampler_state );
+   const unsigned ws = sampler->wrap_s;
+   const unsigned wt = sampler->wrap_t;
+   const unsigned wr = sampler->wrap_r;
+   unsigned minFilt, magFilt;
+   unsigned mipFilt;
+
+   cso->templ = sampler;
+
+   mipFilt = translate_mip_filter(sampler->min_mip_filter);
+   minFilt = translate_img_filter( sampler->min_img_filter );
+   magFilt = translate_img_filter( sampler->mag_img_filter );
+   
+   if (sampler->max_anisotropy > 2.0) {
+      cso->state[0] |= SS2_MAX_ANISO_4;
+   }
+
+   {
+      int b = (int) (sampler->lod_bias * 16.0);
+      b = CLAMP(b, -256, 255);
+      cso->state[0] |= ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK);
+   }
+
+   /* Shadow:
+    */
+   if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) 
+   {
+      cso->state[0] |= (SS2_SHADOW_ENABLE |
+                        i915_translate_compare_func(sampler->compare_func));
+
+      minFilt = FILTER_4X4_FLAT;
+      magFilt = FILTER_4X4_FLAT;
+   }
+
+   cso->state[0] |= ((minFilt << SS2_MIN_FILTER_SHIFT) |
+                     (mipFilt << SS2_MIP_FILTER_SHIFT) |
+                     (magFilt << SS2_MAG_FILTER_SHIFT));
+
+   cso->state[1] |=
+      ((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) |
+       (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) |
+       (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT));
+
+   if (sampler->normalized_coords)
+      cso->state[1] |= SS3_NORMALIZED_COORDS;
+
+   {
+      int minlod = (int) (16.0 * sampler->min_lod);
+      int maxlod = (int) (16.0 * sampler->max_lod);
+      minlod = CLAMP(minlod, 0, 16 * 11);
+      maxlod = CLAMP(maxlod, 0, 16 * 11);
+
+      if (minlod > maxlod)
+	 maxlod = minlod;
+
+      cso->minlod = minlod;
+      cso->maxlod = maxlod;
+   }
+
+   {
+      ubyte r = float_to_ubyte(sampler->border_color[0]);
+      ubyte g = float_to_ubyte(sampler->border_color[1]);
+      ubyte b = float_to_ubyte(sampler->border_color[2]);
+      ubyte a = float_to_ubyte(sampler->border_color[3]);
+      cso->state[2] = I915PACKCOLOR8888(r, g, b, a);
+   }
+   return cso;
+}
+
+static void i915_bind_sampler_states(struct pipe_context *pipe,
+                                     unsigned num, void **sampler)
+{
+   struct i915_context *i915 = i915_context(pipe);
+   unsigned i;
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == i915->num_samplers &&
+       !memcmp(i915->sampler, sampler, num * sizeof(void *)))
+      return;
+
+   draw_flush(i915->draw);
+
+   for (i = 0; i < num; ++i)
+      i915->sampler[i] = sampler[i];
+   for (i = num; i < PIPE_MAX_SAMPLERS; ++i)
+      i915->sampler[i] = NULL;
+
+   i915->num_samplers = num;
+
+   i915->dirty |= I915_NEW_SAMPLER;
+}
+
+static void i915_delete_sampler_state(struct pipe_context *pipe,
+                                      void *sampler)
+{
+   FREE(sampler);
+}
+
+
+/** XXX move someday?  Or consolidate all these simple state setters
+ * into one file.
+ */
+
+static void *
+i915_create_depth_stencil_state(struct pipe_context *pipe,
+				const struct pipe_depth_stencil_alpha_state *depth_stencil)
+{
+   struct i915_depth_stencil_state *cso = CALLOC_STRUCT( i915_depth_stencil_state );
+
+   {
+      int testmask = depth_stencil->stencil[0].valuemask & 0xff;
+      int writemask = depth_stencil->stencil[0].writemask & 0xff;
+
+      cso->stencil_modes4 |= (_3DSTATE_MODES_4_CMD |
+                              ENABLE_STENCIL_TEST_MASK |
+                              STENCIL_TEST_MASK(testmask) |
+                              ENABLE_STENCIL_WRITE_MASK |
+                              STENCIL_WRITE_MASK(writemask));
+   }
+
+   if (depth_stencil->stencil[0].enabled) {
+      int test = i915_translate_compare_func(depth_stencil->stencil[0].func);
+      int fop  = i915_translate_stencil_op(depth_stencil->stencil[0].fail_op);
+      int dfop = i915_translate_stencil_op(depth_stencil->stencil[0].zfail_op);
+      int dpop = i915_translate_stencil_op(depth_stencil->stencil[0].zpass_op);
+      int ref  = depth_stencil->stencil[0].ref_value & 0xff;
+
+      cso->stencil_LIS5 |= (S5_STENCIL_TEST_ENABLE |
+                            S5_STENCIL_WRITE_ENABLE |
+                            (ref  << S5_STENCIL_REF_SHIFT) |
+                            (test << S5_STENCIL_TEST_FUNC_SHIFT) |
+                            (fop  << S5_STENCIL_FAIL_SHIFT) |
+                            (dfop << S5_STENCIL_PASS_Z_FAIL_SHIFT) |
+                            (dpop << S5_STENCIL_PASS_Z_PASS_SHIFT));
+   }
+
+   if (depth_stencil->stencil[1].enabled) {
+      int test  = i915_translate_compare_func(depth_stencil->stencil[1].func);
+      int fop   = i915_translate_stencil_op(depth_stencil->stencil[1].fail_op);
+      int dfop  = i915_translate_stencil_op(depth_stencil->stencil[1].zfail_op);
+      int dpop  = i915_translate_stencil_op(depth_stencil->stencil[1].zpass_op);
+      int ref   = depth_stencil->stencil[1].ref_value & 0xff;
+      int tmask = depth_stencil->stencil[1].valuemask & 0xff;
+      int wmask = depth_stencil->stencil[1].writemask & 0xff;
+
+      cso->bfo[0] = (_3DSTATE_BACKFACE_STENCIL_OPS |
+                     BFO_ENABLE_STENCIL_FUNCS |
+                     BFO_ENABLE_STENCIL_TWO_SIDE |
+                     BFO_ENABLE_STENCIL_REF |
+                     BFO_STENCIL_TWO_SIDE |
+                     (ref  << BFO_STENCIL_REF_SHIFT) |
+                     (test << BFO_STENCIL_TEST_SHIFT) |
+                     (fop  << BFO_STENCIL_FAIL_SHIFT) |
+                     (dfop << BFO_STENCIL_PASS_Z_FAIL_SHIFT) |
+                     (dpop << BFO_STENCIL_PASS_Z_PASS_SHIFT));
+
+      cso->bfo[1] = (_3DSTATE_BACKFACE_STENCIL_MASKS |
+                     BFM_ENABLE_STENCIL_TEST_MASK |
+                     BFM_ENABLE_STENCIL_WRITE_MASK |
+                     (tmask << BFM_STENCIL_TEST_MASK_SHIFT) |
+                     (wmask << BFM_STENCIL_WRITE_MASK_SHIFT));
+   }
+   else {
+      /* This actually disables two-side stencil: The bit set is a
+       * modify-enable bit to indicate we are changing the two-side
+       * setting.  Then there is a symbolic zero to show that we are
+       * setting the flag to zero/off.
+       */
+      cso->bfo[0] = (_3DSTATE_BACKFACE_STENCIL_OPS |
+                     BFO_ENABLE_STENCIL_TWO_SIDE |
+                     0);
+      cso->bfo[1] = 0;
+   }
+
+   if (depth_stencil->depth.enabled) {
+      int func = i915_translate_compare_func(depth_stencil->depth.func);
+
+      cso->depth_LIS6 |= (S6_DEPTH_TEST_ENABLE |
+                          (func << S6_DEPTH_TEST_FUNC_SHIFT));
+
+      if (depth_stencil->depth.writemask)
+	 cso->depth_LIS6 |= S6_DEPTH_WRITE_ENABLE;
+   }
+
+   if (depth_stencil->alpha.enabled) {
+      int test = i915_translate_compare_func(depth_stencil->alpha.func);
+      ubyte refByte = float_to_ubyte(depth_stencil->alpha.ref_value);
+
+      cso->depth_LIS6 |= (S6_ALPHA_TEST_ENABLE |
+			  (test << S6_ALPHA_TEST_FUNC_SHIFT) |
+			  (((unsigned) refByte) << S6_ALPHA_REF_SHIFT));
+   }
+
+   return cso;
+}
+
+static void i915_bind_depth_stencil_state(struct pipe_context *pipe,
+                                          void *depth_stencil)
+{
+   struct i915_context *i915 = i915_context(pipe);
+   draw_flush(i915->draw);
+
+   i915->depth_stencil = (const struct i915_depth_stencil_state *)depth_stencil;
+
+   i915->dirty |= I915_NEW_DEPTH_STENCIL;
+}
+
+static void i915_delete_depth_stencil_state(struct pipe_context *pipe,
+                                            void *depth_stencil)
+{
+   FREE(depth_stencil);
+}
+
+
+static void i915_set_scissor_state( struct pipe_context *pipe,
+                                 const struct pipe_scissor_state *scissor )
+{
+   struct i915_context *i915 = i915_context(pipe);
+   draw_flush(i915->draw);
+
+   memcpy( &i915->scissor, scissor, sizeof(*scissor) );
+   i915->dirty |= I915_NEW_SCISSOR;
+}
+
+
+static void i915_set_polygon_stipple( struct pipe_context *pipe,
+                                   const struct pipe_poly_stipple *stipple )
+{
+}
+
+
+
+static void *
+i915_create_fs_state(struct pipe_context *pipe,
+                     const struct pipe_shader_state *templ)
+{
+   struct i915_context *i915 = i915_context(pipe);
+   struct i915_fragment_shader *ifs = CALLOC_STRUCT(i915_fragment_shader);
+   if (!ifs)
+      return NULL;
+
+   ifs->state.tokens = tgsi_dup_tokens(templ->tokens);
+
+   tgsi_scan_shader(templ->tokens, &ifs->info);
+
+   /* The shader's compiled to i915 instructions here */
+   i915_translate_fragment_program(i915, ifs);
+
+   return ifs;
+}
+
+static void
+i915_bind_fs_state(struct pipe_context *pipe, void *shader)
+{
+   struct i915_context *i915 = i915_context(pipe);
+   draw_flush(i915->draw);
+
+   i915->fs = (struct i915_fragment_shader*) shader;
+
+   i915->dirty |= I915_NEW_FS;
+}
+
+static
+void i915_delete_fs_state(struct pipe_context *pipe, void *shader)
+{
+   struct i915_fragment_shader *ifs = (struct i915_fragment_shader *) shader;
+
+   if (ifs->program)
+      FREE(ifs->program);
+   ifs->program_len = 0;
+
+   FREE((struct tgsi_token *)ifs->state.tokens);
+
+   FREE(ifs);
+}
+
+
+static void *
+i915_create_vs_state(struct pipe_context *pipe,
+                     const struct pipe_shader_state *templ)
+{
+   struct i915_context *i915 = i915_context(pipe);
+
+   /* just pass-through to draw module */
+   return draw_create_vertex_shader(i915->draw, templ);
+}
+
+static void i915_bind_vs_state(struct pipe_context *pipe, void *shader)
+{
+   struct i915_context *i915 = i915_context(pipe);
+
+   /* just pass-through to draw module */
+   draw_bind_vertex_shader(i915->draw, (struct draw_vertex_shader *) shader);
+
+   i915->dirty |= I915_NEW_VS;
+}
+
+static void i915_delete_vs_state(struct pipe_context *pipe, void *shader)
+{
+   struct i915_context *i915 = i915_context(pipe);
+
+   /* just pass-through to draw module */
+   draw_delete_vertex_shader(i915->draw, (struct draw_vertex_shader *) shader);
+}
+
+static void i915_set_constant_buffer(struct pipe_context *pipe,
+                                     uint shader, uint index,
+                                     const struct pipe_constant_buffer *buf)
+{
+   struct i915_context *i915 = i915_context(pipe);
+   struct pipe_winsys *ws = pipe->winsys;
+   draw_flush(i915->draw);
+
+   assert(shader < PIPE_SHADER_TYPES);
+   assert(index == 0);
+
+   /* Make a copy of shader constants.
+    * During fragment program translation we may add additional
+    * constants to the array.
+    *
+    * We want to consider the situation where some user constants
+    * (ex: a material color) may change frequently but the shader program
+    * stays the same.  In that case we should only be updating the first
+    * N constants, leaving any extras from shader translation alone.
+    */
+   if (buf) {
+      void *mapped;
+      if (buf->buffer && buf->buffer->size &&
+          (mapped = ws->buffer_map(ws, buf->buffer,
+                                    PIPE_BUFFER_USAGE_CPU_READ))) {
+         memcpy(i915->current.constants[shader], mapped, buf->buffer->size);
+         ws->buffer_unmap(ws, buf->buffer);
+         i915->current.num_user_constants[shader]
+            = buf->buffer->size / (4 * sizeof(float));
+      }
+      else {
+         i915->current.num_user_constants[shader] = 0;
+      }
+   }
+
+   i915->dirty |= I915_NEW_CONSTANTS;
+}
+
+
+static void i915_set_sampler_textures(struct pipe_context *pipe,
+                                      unsigned num,
+                                      struct pipe_texture **texture)
+{
+   struct i915_context *i915 = i915_context(pipe);
+   uint i;
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == i915->num_textures &&
+       !memcmp(i915->texture, texture, num * sizeof(struct pipe_texture *)))
+      return;
+
+   /* Fixes wrong texture in texobj with VBUF */
+   draw_flush(i915->draw);
+
+   for (i = 0; i < num; i++)
+      pipe_texture_reference((struct pipe_texture **) &i915->texture[i],
+                             texture[i]);
+
+   for (i = num; i < i915->num_textures; i++)
+      pipe_texture_reference((struct pipe_texture **) &i915->texture[i],
+                             NULL);
+
+   i915->num_textures = num;
+
+   i915->dirty |= I915_NEW_TEXTURE;
+}
+
+
+
+static void i915_set_framebuffer_state(struct pipe_context *pipe,
+				       const struct pipe_framebuffer_state *fb)
+{
+   struct i915_context *i915 = i915_context(pipe);
+   draw_flush(i915->draw);
+
+   i915->framebuffer = *fb; /* struct copy */
+
+   i915->dirty |= I915_NEW_FRAMEBUFFER;
+}
+
+
+
+static void i915_set_clip_state( struct pipe_context *pipe,
+			     const struct pipe_clip_state *clip )
+{
+   struct i915_context *i915 = i915_context(pipe);
+   draw_flush(i915->draw);
+
+   draw_set_clip_state(i915->draw, clip);
+
+   i915->dirty |= I915_NEW_CLIP;
+}
+
+
+
+/* Called when driver state tracker notices changes to the viewport
+ * matrix:
+ */
+static void i915_set_viewport_state( struct pipe_context *pipe,
+				     const struct pipe_viewport_state *viewport )
+{
+   struct i915_context *i915 = i915_context(pipe);
+
+   i915->viewport = *viewport; /* struct copy */
+
+   /* pass the viewport info to the draw module */
+   draw_set_viewport_state(i915->draw, &i915->viewport);
+
+   i915->dirty |= I915_NEW_VIEWPORT;
+}
+
+
+static void *
+i915_create_rasterizer_state(struct pipe_context *pipe,
+                             const struct pipe_rasterizer_state *rasterizer)
+{
+   struct i915_rasterizer_state *cso = CALLOC_STRUCT( i915_rasterizer_state );
+
+   cso->templ = rasterizer;
+   cso->color_interp = rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR;
+   cso->light_twoside = rasterizer->light_twoside;
+   cso->ds[0].u = _3DSTATE_DEPTH_OFFSET_SCALE;
+   cso->ds[1].f = rasterizer->offset_scale;
+   if (rasterizer->poly_stipple_enable) {
+      cso->st |= ST1_ENABLE;
+   }
+
+   if (rasterizer->scissor)
+      cso->sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT;
+   else
+      cso->sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT;
+
+   switch (rasterizer->cull_mode) {
+   case PIPE_WINDING_NONE:
+      cso->LIS4 |= S4_CULLMODE_NONE;
+      break;
+   case PIPE_WINDING_CW:
+      cso->LIS4 |= S4_CULLMODE_CW;
+      break;
+   case PIPE_WINDING_CCW:
+      cso->LIS4 |= S4_CULLMODE_CCW;
+      break;
+   case PIPE_WINDING_BOTH:
+      cso->LIS4 |= S4_CULLMODE_BOTH;
+      break;
+   }
+
+   {
+      int line_width = CLAMP((int)(rasterizer->line_width * 2), 1, 0xf);
+
+      cso->LIS4 |= line_width << S4_LINE_WIDTH_SHIFT;
+
+      if (rasterizer->line_smooth)
+	 cso->LIS4 |= S4_LINE_ANTIALIAS_ENABLE;
+   }
+
+   {
+      int point_size = CLAMP((int) rasterizer->point_size, 1, 0xff);
+
+      cso->LIS4 |= point_size << S4_POINT_WIDTH_SHIFT;
+   }
+
+   if (rasterizer->flatshade) {
+      cso->LIS4 |= (S4_FLATSHADE_ALPHA |
+                    S4_FLATSHADE_COLOR |
+                    S4_FLATSHADE_SPECULAR);
+   }
+
+   cso->LIS7 = fui( rasterizer->offset_units );
+
+
+   return cso;
+}
+
+static void i915_bind_rasterizer_state( struct pipe_context *pipe,
+                                        void *raster )
+{
+   struct i915_context *i915 = i915_context(pipe);
+
+   i915->rasterizer = (struct i915_rasterizer_state *)raster;
+
+   /* pass-through to draw module */
+   draw_set_rasterizer_state(i915->draw,
+                          (i915->rasterizer ? i915->rasterizer->templ : NULL));
+
+   i915->dirty |= I915_NEW_RASTERIZER;
+}
+
+static void i915_delete_rasterizer_state(struct pipe_context *pipe,
+                                         void *raster)
+{
+   FREE(raster);
+}
+
+static void i915_set_vertex_buffers(struct pipe_context *pipe,
+                                    unsigned count,
+                                    const struct pipe_vertex_buffer *buffers)
+{
+   struct i915_context *i915 = i915_context(pipe);
+   /* Because we change state before the draw_set_vertex_buffers call
+    * we need a flush here, just to be sure.
+    */
+   draw_flush(i915->draw);
+
+   memcpy(i915->vertex_buffer, buffers, count * sizeof(buffers[0]));
+   i915->num_vertex_buffers = count;
+
+   /* pass-through to draw module */
+   draw_set_vertex_buffers(i915->draw, count, buffers);
+}
+
+static void i915_set_vertex_elements(struct pipe_context *pipe,
+                                     unsigned count,
+                                     const struct pipe_vertex_element *elements)
+{
+   struct i915_context *i915 = i915_context(pipe);
+   /* Because we change state before the draw_set_vertex_buffers call
+    * we need a flush here, just to be sure.
+    */
+   draw_flush(i915->draw);
+
+   i915->num_vertex_elements = count;
+   /* pass-through to draw module */
+   draw_set_vertex_elements(i915->draw, count, elements);
+}
+
+
+static void i915_set_edgeflags(struct pipe_context *pipe,
+                               const unsigned *bitfield)
+{
+   /* TODO do something here */
+}
+
+void
+i915_init_state_functions( struct i915_context *i915 )
+{
+   i915->pipe.set_edgeflags = i915_set_edgeflags;
+   i915->pipe.create_blend_state = i915_create_blend_state;
+   i915->pipe.bind_blend_state = i915_bind_blend_state;
+   i915->pipe.delete_blend_state = i915_delete_blend_state;
+
+   i915->pipe.create_sampler_state = i915_create_sampler_state;
+   i915->pipe.bind_sampler_states = i915_bind_sampler_states;
+   i915->pipe.delete_sampler_state = i915_delete_sampler_state;
+
+   i915->pipe.create_depth_stencil_alpha_state = i915_create_depth_stencil_state;
+   i915->pipe.bind_depth_stencil_alpha_state = i915_bind_depth_stencil_state;
+   i915->pipe.delete_depth_stencil_alpha_state = i915_delete_depth_stencil_state;
+
+   i915->pipe.create_rasterizer_state = i915_create_rasterizer_state;
+   i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state;
+   i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state;
+   i915->pipe.create_fs_state = i915_create_fs_state;
+   i915->pipe.bind_fs_state = i915_bind_fs_state;
+   i915->pipe.delete_fs_state = i915_delete_fs_state;
+   i915->pipe.create_vs_state = i915_create_vs_state;
+   i915->pipe.bind_vs_state = i915_bind_vs_state;
+   i915->pipe.delete_vs_state = i915_delete_vs_state;
+
+   i915->pipe.set_blend_color = i915_set_blend_color;
+   i915->pipe.set_clip_state = i915_set_clip_state;
+   i915->pipe.set_constant_buffer = i915_set_constant_buffer;
+   i915->pipe.set_framebuffer_state = i915_set_framebuffer_state;
+
+   i915->pipe.set_polygon_stipple = i915_set_polygon_stipple;
+   i915->pipe.set_scissor_state = i915_set_scissor_state;
+   i915->pipe.set_sampler_textures = i915_set_sampler_textures;
+   i915->pipe.set_viewport_state = i915_set_viewport_state;
+   i915->pipe.set_vertex_buffers = i915_set_vertex_buffers;
+   i915->pipe.set_vertex_elements = i915_set_vertex_elements;
+}
diff --git a/src/gallium/drivers/i915simple/i915_state.h b/src/gallium/drivers/i915simple/i915_state.h
new file mode 100644
index 0000000000..86c6b0027d
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_state.h
@@ -0,0 +1,50 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef I915_STATE_H
+#define I915_STATE_H
+
+struct i915_context;
+
+
+struct i915_tracked_state {
+   unsigned dirty;
+   void (*update)( struct i915_context * );
+};
+
+void i915_update_immediate( struct i915_context *i915 );
+void i915_update_dynamic( struct i915_context *i915 );
+void i915_update_derived( struct i915_context *i915 );
+void i915_update_samplers( struct i915_context *i915 );
+void i915_update_textures(struct i915_context *i915);
+
+void i915_emit_hardware_state( struct i915_context *i915 );
+
+#endif
diff --git a/src/gallium/drivers/i915simple/i915_state_derived.c b/src/gallium/drivers/i915simple/i915_state_derived.c
new file mode 100644
index 0000000000..178d4e8781
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_state_derived.c
@@ -0,0 +1,183 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vertex.h"
+#include "i915_context.h"
+#include "i915_state.h"
+#include "i915_reg.h"
+#include "i915_fpc.h"
+
+
+
+/**
+ * Determine the hardware vertex layout.
+ * Depends on vertex/fragment shader state.
+ */
+static void calculate_vertex_layout( struct i915_context *i915 )
+{
+   const struct i915_fragment_shader *fs = i915->fs;
+   const enum interp_mode colorInterp = i915->rasterizer->color_interp;
+   struct vertex_info vinfo;
+   boolean texCoords[8], colors[2], fog, needW;
+   uint i;
+   int src;
+
+   memset(texCoords, 0, sizeof(texCoords));
+   colors[0] = colors[1] = fog = needW = FALSE;
+   memset(&vinfo, 0, sizeof(vinfo));
+
+   /* Determine which fragment program inputs are needed.  Setup HW vertex
+    * layout below, in the HW-specific attribute order.
+    */
+   for (i = 0; i < fs->info.num_inputs; i++) {
+      switch (fs->info.input_semantic_name[i]) {
+      case TGSI_SEMANTIC_POSITION:
+         break;
+      case TGSI_SEMANTIC_COLOR:
+         assert(fs->info.input_semantic_index[i] < 2);
+         colors[fs->info.input_semantic_index[i]] = TRUE;
+         break;
+      case TGSI_SEMANTIC_GENERIC:
+         /* usually a texcoord */
+         {
+            const uint unit = fs->info.input_semantic_index[i];
+            assert(unit < 8);
+            texCoords[unit] = TRUE;
+            needW = TRUE;
+         }
+         break;
+      case TGSI_SEMANTIC_FOG:
+         fog = TRUE;
+         break;
+      default:
+         assert(0);
+      }
+   }
+
+   
+   /* pos */
+   src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_POSITION, 0);
+   if (needW) {
+      draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src);
+      vinfo.hwfmt[0] |= S4_VFMT_XYZW;
+      vinfo.attrib[0].emit = EMIT_4F;
+   }
+   else {
+      draw_emit_vertex_attr(&vinfo, EMIT_3F, INTERP_LINEAR, src);
+      vinfo.hwfmt[0] |= S4_VFMT_XYZ;
+      vinfo.attrib[0].emit = EMIT_3F;
+   }
+
+   /* hardware point size */
+   /* XXX todo */
+
+   /* primary color */
+   if (colors[0]) {
+      src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_COLOR, 0);
+      draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src);
+      vinfo.hwfmt[0] |= S4_VFMT_COLOR;
+   }
+
+   /* secondary color */
+   if (colors[1]) {
+      src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_COLOR, 1);
+      draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src);
+      vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG;
+   }
+
+   /* fog coord, not fog blend factor */
+   if (fog) {
+      src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_FOG, 0);
+      draw_emit_vertex_attr(&vinfo, EMIT_1F, INTERP_PERSPECTIVE, src);
+      vinfo.hwfmt[0] |= S4_VFMT_FOG_PARAM;
+   }
+
+   /* texcoords */
+   for (i = 0; i < 8; i++) {
+      uint hwtc;
+      if (texCoords[i]) {
+         hwtc = TEXCOORDFMT_4D;
+         src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_GENERIC, i);
+         draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+      }
+      else {
+         hwtc = TEXCOORDFMT_NOT_PRESENT;
+      }
+      vinfo.hwfmt[1] |= hwtc << (i * 4);
+   }
+
+   draw_compute_vertex_size(&vinfo);
+
+   if (memcmp(&i915->current.vertex_info, &vinfo, sizeof(vinfo))) {
+      /* Need to set this flag so that the LIS2/4 registers get set.
+       * It also means the i915_update_immediate() function must be called
+       * after this one, in i915_update_derived().
+       */
+      i915->dirty |= I915_NEW_VERTEX_FORMAT;
+
+      memcpy(&i915->current.vertex_info, &vinfo, sizeof(vinfo));
+   }
+}
+
+
+
+
+/* Hopefully this will remain quite simple, otherwise need to pull in
+ * something like the state tracker mechanism.
+ */
+void i915_update_derived( struct i915_context *i915 )
+{
+   if (i915->dirty & (I915_NEW_RASTERIZER | I915_NEW_FS | I915_NEW_VS))
+      calculate_vertex_layout( i915 );
+
+   if (i915->dirty & (I915_NEW_SAMPLER | I915_NEW_TEXTURE))
+      i915_update_samplers(i915);
+
+   if (i915->dirty & I915_NEW_TEXTURE)
+      i915_update_textures(i915);
+
+   if (i915->dirty)
+      i915_update_immediate( i915 );
+
+   if (i915->dirty)
+      i915_update_dynamic( i915 );
+
+   if (i915->dirty & I915_NEW_FS) {
+      i915->hardware_dirty |= I915_HW_PROGRAM; /* XXX right? */
+   }
+
+   /* HW emit currently references framebuffer state directly:
+    */
+   if (i915->dirty & I915_NEW_FRAMEBUFFER)
+      i915->hardware_dirty |= I915_HW_STATIC;
+
+   i915->dirty = 0;
+}
diff --git a/src/gallium/drivers/i915simple/i915_state_dynamic.c b/src/gallium/drivers/i915simple/i915_state_dynamic.c
new file mode 100644
index 0000000000..86126a5a15
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_state_dynamic.c
@@ -0,0 +1,310 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "i915_batch.h"
+#include "i915_state_inlines.h"
+#include "i915_context.h"
+#include "i915_reg.h"
+#include "i915_state.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "util/u_pack_color.h"
+
+#define FILE_DEBUG_FLAG DEBUG_STATE
+
+/* State that we have chosen to store in the DYNAMIC segment of the
+ * i915 indirect state mechanism.  
+ *
+ * Can't cache these in the way we do the static state, as there is no
+ * start/size in the command packet, instead an 'end' value that gets
+ * incremented.
+ *
+ * Additionally, there seems to be a requirement to re-issue the full
+ * (active) state every time a 4kb boundary is crossed.
+ */
+
+static INLINE void set_dynamic_indirect( struct i915_context *i915,
+					 unsigned offset,
+					 const unsigned *src,
+					 unsigned dwords )
+{
+   unsigned i;
+
+   for (i = 0; i < dwords; i++)
+      i915->current.dynamic[offset + i] = src[i];
+
+   i915->hardware_dirty |= I915_HW_DYNAMIC;
+}
+
+
+/***********************************************************************
+ * Modes4: stencil masks and logicop 
+ */
+static void upload_MODES4( struct i915_context *i915 )
+{
+   unsigned modes4 = 0;
+
+   /* I915_NEW_STENCIL */
+   modes4 |= i915->depth_stencil->stencil_modes4;
+   /* I915_NEW_BLEND */
+   modes4 |= i915->blend->modes4;
+
+   /* Always, so that we know when state is in-active: 
+    */
+   set_dynamic_indirect( i915, 
+			 I915_DYNAMIC_MODES4,
+			 &modes4,
+			 1 );
+}
+
+const struct i915_tracked_state i915_upload_MODES4 = {
+   I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL,
+   upload_MODES4
+};
+
+
+
+
+/***********************************************************************
+ */
+
+static void upload_BFO( struct i915_context *i915 )
+{
+   set_dynamic_indirect( i915,
+			 I915_DYNAMIC_BFO_0,
+			 &(i915->depth_stencil->bfo[0]),
+			 2 );
+}
+
+const struct i915_tracked_state i915_upload_BFO = {
+   I915_NEW_DEPTH_STENCIL,
+   upload_BFO
+};
+
+
+/***********************************************************************
+ */
+
+
+static void upload_BLENDCOLOR( struct i915_context *i915 )
+{
+   unsigned bc[2];
+
+   memset( bc, 0, sizeof(bc) );
+
+   /* I915_NEW_BLEND {_COLOR} 
+    */
+   {
+      const float *color = i915->blend_color.color;
+
+      bc[0] = _3DSTATE_CONST_BLEND_COLOR_CMD;
+      bc[1] = pack_ui32_float4( color[0],
+				color[1],
+				color[2], 
+				color[3] );
+   }
+
+   set_dynamic_indirect( i915, 
+			 I915_DYNAMIC_BC_0,
+			 bc,
+			 2 );
+}
+
+const struct i915_tracked_state i915_upload_BLENDCOLOR = {
+   I915_NEW_BLEND,
+   upload_BLENDCOLOR
+};
+
+/***********************************************************************
+ */
+
+
+static void upload_IAB( struct i915_context *i915 )
+{
+   unsigned iab = i915->blend->iab;
+
+
+   set_dynamic_indirect( i915,
+			 I915_DYNAMIC_IAB,
+			 &iab,
+			 1 );
+}
+
+const struct i915_tracked_state i915_upload_IAB = {
+   I915_NEW_BLEND,
+   upload_IAB
+};
+
+
+/***********************************************************************
+ */
+
+
+
+static void upload_DEPTHSCALE( struct i915_context *i915 )
+{
+   set_dynamic_indirect( i915,
+			 I915_DYNAMIC_DEPTHSCALE_0,
+			 &(i915->rasterizer->ds[0].u),
+			 2 );
+}
+
+const struct i915_tracked_state i915_upload_DEPTHSCALE = {
+   I915_NEW_RASTERIZER,
+   upload_DEPTHSCALE
+};
+
+
+
+/***********************************************************************
+ * Polygon stipple
+ *
+ * The i915 supports a 4x4 stipple natively, GL wants 32x32.
+ * Fortunately stipple is usually a repeating pattern.
+ *
+ * XXX: does stipple pattern need to be adjusted according to
+ * the window position?
+ *
+ * XXX: possibly need workaround for conform paths test. 
+ */
+
+static void upload_STIPPLE( struct i915_context *i915 )
+{
+   unsigned st[2];
+
+   st[0] = _3DSTATE_STIPPLE;
+   st[1] = 0;
+
+   /* I915_NEW_RASTERIZER
+    */
+   st[1] |= i915->rasterizer->st;
+
+
+   /* I915_NEW_STIPPLE
+    */
+   {
+      const ubyte *mask = (const ubyte *)i915->poly_stipple.stipple;
+      ubyte p[4];
+
+      p[0] = mask[12] & 0xf;
+      p[1] = mask[8] & 0xf;
+      p[2] = mask[4] & 0xf;
+      p[3] = mask[0] & 0xf;
+
+      /* Not sure what to do about fallbacks, so for now just dont:
+       */
+      st[1] |= ((p[0] << 0) |
+		(p[1] << 4) |
+		(p[2] << 8) | 
+		(p[3] << 12));
+   }
+
+
+   set_dynamic_indirect( i915, 
+			 I915_DYNAMIC_STP_0,
+			 &st[0],
+			 2 );
+}
+
+
+const struct i915_tracked_state i915_upload_STIPPLE = {
+   I915_NEW_RASTERIZER | I915_NEW_STIPPLE,
+   upload_STIPPLE
+};
+
+
+
+/***********************************************************************
+ * Scissor.
+ */
+static void upload_SCISSOR_ENABLE( struct i915_context *i915 )
+{
+   set_dynamic_indirect( i915,
+			 I915_DYNAMIC_SC_ENA_0,
+			 &(i915->rasterizer->sc[0]),
+			 1 );
+}
+
+const struct i915_tracked_state i915_upload_SCISSOR_ENABLE = {
+   I915_NEW_RASTERIZER,
+   upload_SCISSOR_ENABLE
+};
+
+
+
+static void upload_SCISSOR_RECT( struct i915_context *i915 )
+{
+   unsigned x1 = i915->scissor.minx;
+   unsigned y1 = i915->scissor.miny;
+   unsigned x2 = i915->scissor.maxx;
+   unsigned y2 = i915->scissor.maxy;
+   unsigned sc[3];
+ 
+   sc[0] = _3DSTATE_SCISSOR_RECT_0_CMD;
+   sc[1] = (y1 << 16) | (x1 & 0xffff);
+   sc[2] = (y2 << 16) | (x2 & 0xffff);
+
+   set_dynamic_indirect( i915, 
+			 I915_DYNAMIC_SC_RECT_0,
+			 &sc[0],
+			 3 );
+}
+
+
+const struct i915_tracked_state i915_upload_SCISSOR_RECT = {
+   I915_NEW_SCISSOR,
+   upload_SCISSOR_RECT
+};
+
+
+
+
+
+
+static const struct i915_tracked_state *atoms[] = {
+   &i915_upload_MODES4,
+   &i915_upload_BFO,
+   &i915_upload_BLENDCOLOR,
+   &i915_upload_IAB,
+   &i915_upload_DEPTHSCALE,
+   &i915_upload_STIPPLE,
+   &i915_upload_SCISSOR_ENABLE,
+   &i915_upload_SCISSOR_RECT
+};
+
+/* These will be dynamic indirect state commands, but for now just end
+ * up on the batch buffer with everything else.
+ */
+void i915_update_dynamic( struct i915_context *i915 )
+{
+   int i;
+
+   for (i = 0; i < Elements(atoms); i++)
+      if (i915->dirty & atoms[i]->dirty)
+	 atoms[i]->update( i915 );
+}
+
diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c
new file mode 100644
index 0000000000..6558cf1c3e
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_state_emit.c
@@ -0,0 +1,410 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "i915_reg.h"
+#include "i915_context.h"
+#include "i915_winsys.h"
+#include "i915_batch.h"
+#include "i915_reg.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+
+static unsigned translate_format( enum pipe_format format )
+{
+   switch (format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      return COLOR_BUF_ARGB8888;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      return COLOR_BUF_RGB565;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+static unsigned translate_depth_format( enum pipe_format zformat )
+{
+   switch (zformat) {
+   case PIPE_FORMAT_S8Z24_UNORM:
+      return DEPTH_FRMT_24_FIXED_8_OTHER;
+   case PIPE_FORMAT_Z16_UNORM:
+      return DEPTH_FRMT_16_FIXED;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+
+/**
+ * Examine framebuffer state to determine width, height.
+ */
+static boolean
+framebuffer_size(const struct pipe_framebuffer_state *fb,
+                 uint *width, uint *height)
+{
+   if (fb->cbufs[0]) {
+      *width = fb->cbufs[0]->width;
+      *height = fb->cbufs[0]->height;
+      return TRUE;
+   }
+   else if (fb->zsbuf) {
+      *width = fb->zsbuf->width;
+      *height = fb->zsbuf->height;
+      return TRUE;
+   }
+   else {
+      *width = *height = 0;
+      return FALSE;
+   }
+}
+
+
+/* Push the state into the sarea and/or texture memory.
+ */
+void
+i915_emit_hardware_state(struct i915_context *i915 )
+{
+   /* XXX: there must be an easier way */
+   const unsigned dwords = ( 14 + 
+                             7 + 
+                             I915_MAX_DYNAMIC + 
+                             8 + 
+                             2 + I915_TEX_UNITS*3 + 
+                             2 + I915_TEX_UNITS*3 +
+                             2 + I915_MAX_CONSTANT*4 + 
+#if 0
+                             i915->current.program_len + 
+#else
+                             i915->fs->program_len + 
+#endif
+                             6 
+                           ) * 3/2; /* plus 50% margin */
+   const unsigned relocs = ( I915_TEX_UNITS +
+	                     3
+                           ) * 3/2; /* plus 50% margin */
+
+#if 0
+   debug_printf("i915_emit_hardware_state: %d dwords, %d relocs\n", dwords, relocs);
+#endif
+   
+   if(!BEGIN_BATCH(dwords, relocs)) {
+      FLUSH_BATCH(NULL);
+      assert(BEGIN_BATCH(dwords, relocs));
+   }
+
+   /* 14 dwords, 0 relocs */
+   if (i915->hardware_dirty & I915_HW_INVARIENT)
+   {
+      OUT_BATCH(_3DSTATE_AA_CMD |
+		AA_LINE_ECAAR_WIDTH_ENABLE |
+		AA_LINE_ECAAR_WIDTH_1_0 |
+		AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0);
+
+      OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD);
+      OUT_BATCH(0);
+
+      OUT_BATCH(_3DSTATE_DFLT_SPEC_CMD);
+      OUT_BATCH(0);
+      
+      OUT_BATCH(_3DSTATE_DFLT_Z_CMD);
+      OUT_BATCH(0);
+
+      OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS |
+		CSB_TCB(0, 0) |
+		CSB_TCB(1, 1) |
+		CSB_TCB(2, 2) |
+		CSB_TCB(3, 3) |
+		CSB_TCB(4, 4) | 
+		CSB_TCB(5, 5) | 
+		CSB_TCB(6, 6) | 
+		CSB_TCB(7, 7));
+
+      OUT_BATCH(_3DSTATE_RASTER_RULES_CMD |
+		ENABLE_POINT_RASTER_RULE |
+		OGL_POINT_RASTER_RULE |
+		ENABLE_LINE_STRIP_PROVOKE_VRTX |
+		ENABLE_TRI_FAN_PROVOKE_VRTX |
+		LINE_STRIP_PROVOKE_VRTX(1) |
+		TRI_FAN_PROVOKE_VRTX(2) | 
+		ENABLE_TEXKILL_3D_4D | 
+		TEXKILL_4D);
+
+      /* Need to initialize this to zero.
+       */
+      OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | (0));
+      OUT_BATCH(0);
+
+      OUT_BATCH(_3DSTATE_DEPTH_SUBRECT_DISABLE);
+
+      /* disable indirect state for now
+       */
+      OUT_BATCH(_3DSTATE_LOAD_INDIRECT | 0);
+      OUT_BATCH(0);
+   }
+   
+   /* 7 dwords, 1 relocs */
+   if (i915->hardware_dirty & I915_HW_IMMEDIATE)
+   {
+      OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | 
+		I1_LOAD_S(0) |
+		I1_LOAD_S(1) |
+		I1_LOAD_S(2) |
+		I1_LOAD_S(4) |
+		I1_LOAD_S(5) |
+		I1_LOAD_S(6) | 
+		(5));
+      
+      if(i915->vbo)
+         OUT_RELOC(i915->vbo,
+                   I915_BUFFER_ACCESS_READ,
+                   i915->current.immediate[I915_IMMEDIATE_S0]);
+      else
+	 /* FIXME: we should not do this */
+	 OUT_BATCH(0);
+      OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S1]);
+      OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S2]);
+      OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S4]);
+      OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S5]);
+      OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S6]);
+   } 
+   
+   /* I915_MAX_DYNAMIC dwords, 0 relocs */
+   if (i915->hardware_dirty & I915_HW_DYNAMIC) 
+   {
+      int i;
+      for (i = 0; i < I915_MAX_DYNAMIC; i++) {
+	 OUT_BATCH(i915->current.dynamic[i]);
+      }
+   }
+   
+   /* 8 dwords, 2 relocs */
+   if (i915->hardware_dirty & I915_HW_STATIC)
+   {
+      struct pipe_surface *cbuf_surface = i915->framebuffer.cbufs[0];
+      struct pipe_surface *depth_surface = i915->framebuffer.zsbuf;
+
+      if (cbuf_surface) {
+	 unsigned cpitch = cbuf_surface->stride;
+	 unsigned ctile = BUF_3D_USE_FENCE;
+         struct i915_texture *tex = (struct i915_texture *)
+                                    cbuf_surface->texture;
+         struct pipe_buffer *buffer = tex->buffer;
+         assert(tex);
+
+	 if (tex && tex->tiled) {
+	    ctile = BUF_3D_TILED_SURFACE;
+	 }
+
+	 OUT_BATCH(_3DSTATE_BUF_INFO_CMD);
+
+	 OUT_BATCH(BUF_3D_ID_COLOR_BACK |
+		   BUF_3D_PITCH(cpitch) |  /* pitch in bytes */
+		   ctile);
+
+	 OUT_RELOC(tex->buffer,
+		   I915_BUFFER_ACCESS_WRITE,
+		   cbuf_surface->offset);
+      }
+
+      /* What happens if no zbuf??
+       */
+      if (depth_surface) {
+	 unsigned zpitch = depth_surface->stride;
+	 unsigned ztile = BUF_3D_USE_FENCE;
+         struct i915_texture *tex = (struct i915_texture *)
+                                    depth_surface->texture;
+         struct pipe_buffer *buffer = tex->buffer;
+         assert(tex);
+
+	 if (tex && tex->tiled) {
+	    ztile = BUF_3D_TILED_SURFACE;
+	 }
+
+	 OUT_BATCH(_3DSTATE_BUF_INFO_CMD);
+
+	 OUT_BATCH(BUF_3D_ID_DEPTH |
+		   BUF_3D_PITCH(zpitch) |  /* pitch in bytes */
+		   ztile);
+
+	 OUT_RELOC(tex->buffer,
+		   I915_BUFFER_ACCESS_WRITE,
+		   depth_surface->offset);
+      }
+   
+      {
+	 unsigned cformat, zformat = 0;
+      
+	 if (cbuf_surface)
+            cformat = cbuf_surface->format;
+         else
+            cformat = PIPE_FORMAT_A8R8G8B8_UNORM; /* arbitrary */
+         cformat = translate_format(cformat);
+
+	 if (depth_surface) 
+	    zformat = translate_depth_format( i915->framebuffer.zsbuf->format );
+
+	 OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD);
+	 OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */
+		   DSTORG_VERT_BIAS(0x8) | /* .5 */
+		   LOD_PRECLAMP_OGL |
+		   TEX_DEFAULT_COLOR_OGL |
+		   cformat |
+		   zformat );
+      }
+   }
+
+#if 01
+      /* texture images */
+      /* 2 + I915_TEX_UNITS*3 dwords, I915_TEX_UNITS relocs */
+      if (i915->hardware_dirty & (I915_HW_MAP | I915_HW_SAMPLER))
+      {
+         const uint nr = i915->current.sampler_enable_nr;
+         if (nr) {
+            const uint enabled = i915->current.sampler_enable_flags;
+            uint unit;
+            uint count = 0;
+            OUT_BATCH(_3DSTATE_MAP_STATE | (3 * nr));
+            OUT_BATCH(enabled);
+            for (unit = 0; unit < I915_TEX_UNITS; unit++) {
+               if (enabled & (1 << unit)) {
+                  struct pipe_buffer *buf =
+                     i915->texture[unit]->buffer;
+                  uint offset = 0;
+                  assert(buf);
+
+                  count++;
+
+                  OUT_RELOC(buf,
+                            I915_BUFFER_ACCESS_READ,
+                            offset);
+                  OUT_BATCH(i915->current.texbuffer[unit][0]); /* MS3 */
+                  OUT_BATCH(i915->current.texbuffer[unit][1]); /* MS4 */
+               }
+            }
+            assert(count == nr);
+         }
+      }
+#endif
+
+#if 01
+   /* samplers */
+   /* 2 + I915_TEX_UNITS*3 dwords, 0 relocs */
+   if (i915->hardware_dirty & I915_HW_SAMPLER) 
+   {
+      if (i915->current.sampler_enable_nr) {
+	 int i;
+	 
+	 OUT_BATCH( _3DSTATE_SAMPLER_STATE | 
+		    (3 * i915->current.sampler_enable_nr) );
+
+	 OUT_BATCH( i915->current.sampler_enable_flags );
+
+	 for (i = 0; i < I915_TEX_UNITS; i++) {
+	    if (i915->current.sampler_enable_flags & (1<<i)) {
+	       OUT_BATCH( i915->current.sampler[i][0] );
+	       OUT_BATCH( i915->current.sampler[i][1] );
+	       OUT_BATCH( i915->current.sampler[i][2] );
+	    }
+	 }
+      }
+   }
+#endif
+
+   /* constants */
+   /* 2 + I915_MAX_CONSTANT*4 dwords, 0 relocs */
+   if (i915->hardware_dirty & I915_HW_PROGRAM)
+   {
+      /* Collate the user-defined constants with the fragment shader's
+       * immediates according to the constant_flags[] array.
+       */
+      const uint nr = i915->fs->num_constants;
+      if (nr) {
+         uint i;
+
+         OUT_BATCH( _3DSTATE_PIXEL_SHADER_CONSTANTS | (nr * 4) );
+         OUT_BATCH( (1 << (nr - 1)) | ((1 << (nr - 1)) - 1) );
+
+         for (i = 0; i < nr; i++) {
+            const uint *c;
+            if (i915->fs->constant_flags[i] == I915_CONSTFLAG_USER) {
+               /* grab user-defined constant */
+               c = (uint *) i915->current.constants[PIPE_SHADER_FRAGMENT][i];
+            }
+            else {
+               /* emit program constant */
+               c = (uint *) i915->fs->constants[i];
+            }
+#if 0 /* debug */
+            {
+               float *f = (float *) c;
+               printf("Const %2d: %f %f %f %f %s\n", i, f[0], f[1], f[2], f[3],
+                      (i915->fs->constant_flags[i] == I915_CONSTFLAG_USER
+                       ? "user" : "immediate"));
+            }
+#endif
+            OUT_BATCH(*c++);
+            OUT_BATCH(*c++);
+            OUT_BATCH(*c++);
+            OUT_BATCH(*c++);
+         }
+      }
+   }
+
+   /* Fragment program */
+   /* i915->current.program_len dwords, 0 relocs */
+   if (i915->hardware_dirty & I915_HW_PROGRAM)
+   {
+      uint i;
+      /* we should always have, at least, a pass-through program */
+      assert(i915->fs->program_len > 0);
+      for (i = 0; i < i915->fs->program_len; i++) {
+         OUT_BATCH(i915->fs->program[i]);
+      }
+   }
+
+   /* drawing surface size */
+   /* 6 dwords, 0 relocs */
+   {
+      uint w, h;
+      boolean k = framebuffer_size(&i915->framebuffer, &w, &h);
+      (void)k;
+      assert(k);
+
+      OUT_BATCH(_3DSTATE_DRAW_RECT_CMD);
+      OUT_BATCH(0);
+      OUT_BATCH(0);
+      OUT_BATCH(((w - 1) & 0xffff) | ((h - 1) << 16));
+      OUT_BATCH(0);
+      OUT_BATCH(0);
+   }
+
+
+   i915->hardware_dirty = 0;
+}
diff --git a/src/gallium/drivers/i915simple/i915_state_immediate.c b/src/gallium/drivers/i915simple/i915_state_immediate.c
new file mode 100644
index 0000000000..8c16bb4e27
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_state_immediate.c
@@ -0,0 +1,225 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+ 
+#include "i915_state_inlines.h"
+#include "i915_context.h"
+#include "i915_state.h"
+#include "i915_reg.h"
+#include "util/u_memory.h"
+
+
+/* All state expressable with the LOAD_STATE_IMMEDIATE_1 packet.
+ * Would like to opportunistically recombine all these fragments into
+ * a single packet containing only what has changed, but for now emit
+ * as multiple packets.
+ */
+
+
+
+
+/***********************************************************************
+ * S0,S1: Vertex buffer state.  
+ */
+static void upload_S0S1(struct i915_context *i915)
+{
+   unsigned LIS0, LIS1;
+
+   /* INTEL_NEW_VBO */
+   /* TODO: re-use vertex buffers here? */
+   LIS0 = i915->vbo_offset;
+
+   /* INTEL_NEW_VERTEX_SIZE -- do this where the vertex size is calculated! 
+    */
+   {
+      unsigned vertex_size = i915->current.vertex_info.size;
+
+      LIS1 = ((vertex_size << 24) |
+	      (vertex_size << 16));
+   }
+
+   /* INTEL_NEW_VBO */
+   /* TODO: use a vertex generation number to track vbo changes */
+   if (1 ||
+       i915->current.immediate[I915_IMMEDIATE_S0] != LIS0 ||
+       i915->current.immediate[I915_IMMEDIATE_S1] != LIS1) 
+   {
+      i915->current.immediate[I915_IMMEDIATE_S0] = LIS0;
+      i915->current.immediate[I915_IMMEDIATE_S1] = LIS1;
+      i915->hardware_dirty |= I915_HW_IMMEDIATE;
+   }
+}
+
+const struct i915_tracked_state i915_upload_S0S1 = {
+   I915_NEW_VBO | I915_NEW_VERTEX_FORMAT,
+   upload_S0S1
+};
+
+
+
+
+/***********************************************************************
+ * S4: Vertex format, rasterization state
+ */
+static void upload_S2S4(struct i915_context *i915)
+{
+   unsigned LIS2, LIS4;
+
+   /* I915_NEW_VERTEX_FORMAT */
+   {
+      LIS2 = i915->current.vertex_info.hwfmt[1];
+      LIS4 = i915->current.vertex_info.hwfmt[0];
+      /*
+      debug_printf("LIS2: 0x%x  LIS4: 0x%x\n", LIS2, LIS4);
+      */
+      assert(LIS4); /* should never be zero? */
+   }
+
+   LIS4 |= i915->rasterizer->LIS4;
+
+   if (LIS2 != i915->current.immediate[I915_IMMEDIATE_S2] ||
+       LIS4 != i915->current.immediate[I915_IMMEDIATE_S4]) {
+
+      i915->current.immediate[I915_IMMEDIATE_S2] = LIS2;
+      i915->current.immediate[I915_IMMEDIATE_S4] = LIS4;
+      i915->hardware_dirty |= I915_HW_IMMEDIATE;
+   }
+}
+
+
+const struct i915_tracked_state i915_upload_S2S4 = {
+   I915_NEW_RASTERIZER | I915_NEW_VERTEX_FORMAT,
+   upload_S2S4
+};
+
+
+
+/***********************************************************************
+ * 
+ */
+static void upload_S5( struct i915_context *i915 )
+{
+   unsigned LIS5 = 0;
+
+   LIS5 |= i915->depth_stencil->stencil_LIS5;
+
+   LIS5 |= i915->blend->LIS5;
+
+#if 0
+   /* I915_NEW_RASTERIZER */
+   if (i915->state.Polygon->OffsetFill) {
+      LIS5 |= S5_GLOBAL_DEPTH_OFFSET_ENABLE;
+   }
+#endif
+
+
+   if (LIS5 != i915->current.immediate[I915_IMMEDIATE_S5]) {
+      i915->current.immediate[I915_IMMEDIATE_S5] = LIS5;
+      i915->hardware_dirty |= I915_HW_IMMEDIATE;
+   }
+}
+
+const struct i915_tracked_state i915_upload_S5 = {
+   (I915_NEW_DEPTH_STENCIL | I915_NEW_BLEND | I915_NEW_RASTERIZER),
+   upload_S5
+};
+
+
+/***********************************************************************
+ */
+static void upload_S6( struct i915_context *i915 )
+{
+   unsigned LIS6 = (2 << S6_TRISTRIP_PV_SHIFT);
+
+   /* I915_NEW_FRAMEBUFFER
+    */
+   if (i915->framebuffer.cbufs[0])
+      LIS6 |= S6_COLOR_WRITE_ENABLE;
+
+   /* I915_NEW_BLEND
+    */
+   LIS6 |= i915->blend->LIS6;
+
+   /* I915_NEW_DEPTH
+    */
+   LIS6 |= i915->depth_stencil->depth_LIS6;
+
+   if (LIS6 != i915->current.immediate[I915_IMMEDIATE_S6]) {
+      i915->current.immediate[I915_IMMEDIATE_S6] = LIS6;
+      i915->hardware_dirty |= I915_HW_IMMEDIATE;
+   }
+}
+
+const struct i915_tracked_state i915_upload_S6 = {
+   I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL | I915_NEW_FRAMEBUFFER,
+   upload_S6
+};
+
+
+/***********************************************************************
+ */
+static void upload_S7( struct i915_context *i915 )
+{
+   unsigned LIS7;
+
+   /* I915_NEW_RASTERIZER
+    */
+   LIS7 = i915->rasterizer->LIS7;
+
+   if (LIS7 != i915->current.immediate[I915_IMMEDIATE_S7]) {
+      i915->current.immediate[I915_IMMEDIATE_S7] = LIS7;
+      i915->hardware_dirty |= I915_HW_IMMEDIATE;
+   }
+}
+
+const struct i915_tracked_state i915_upload_S7 = {
+   I915_NEW_RASTERIZER,
+   upload_S7
+};
+
+
+static const struct i915_tracked_state *atoms[] = {
+   &i915_upload_S0S1,
+   &i915_upload_S2S4,
+   &i915_upload_S5,
+   &i915_upload_S6,
+   &i915_upload_S7
+};
+
+/* 
+ */
+void i915_update_immediate( struct i915_context *i915 )
+{
+   int i;
+
+   for (i = 0; i < Elements(atoms); i++)
+      if (i915->dirty & atoms[i]->dirty)
+	 atoms[i]->update( i915 );
+}
diff --git a/src/gallium/drivers/i915simple/i915_state_inlines.h b/src/gallium/drivers/i915simple/i915_state_inlines.h
new file mode 100644
index 0000000000..378de8f9c4
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_state_inlines.h
@@ -0,0 +1,230 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef I915_STATE_INLINES_H
+#define I915_STATE_INLINES_H
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_defines.h"
+#include "i915_reg.h"
+
+
+static INLINE unsigned
+i915_translate_compare_func(unsigned func)
+{
+   switch (func) {
+   case PIPE_FUNC_NEVER:
+      return COMPAREFUNC_NEVER;
+   case PIPE_FUNC_LESS:
+      return COMPAREFUNC_LESS;
+   case PIPE_FUNC_LEQUAL:
+      return COMPAREFUNC_LEQUAL;
+   case PIPE_FUNC_GREATER:
+      return COMPAREFUNC_GREATER;
+   case PIPE_FUNC_GEQUAL:
+      return COMPAREFUNC_GEQUAL;
+   case PIPE_FUNC_NOTEQUAL:
+      return COMPAREFUNC_NOTEQUAL;
+   case PIPE_FUNC_EQUAL:
+      return COMPAREFUNC_EQUAL;
+   case PIPE_FUNC_ALWAYS:
+      return COMPAREFUNC_ALWAYS;
+   default:
+      return COMPAREFUNC_ALWAYS;
+   }
+}
+
+static INLINE unsigned
+i915_translate_stencil_op(unsigned op)
+{
+   switch (op) {
+   case PIPE_STENCIL_OP_KEEP:
+      return STENCILOP_KEEP;
+   case PIPE_STENCIL_OP_ZERO:
+      return STENCILOP_ZERO;
+   case PIPE_STENCIL_OP_REPLACE:
+      return STENCILOP_REPLACE;
+   case PIPE_STENCIL_OP_INCR:
+      return STENCILOP_INCRSAT;
+   case PIPE_STENCIL_OP_DECR:
+      return STENCILOP_DECRSAT;
+   case PIPE_STENCIL_OP_INCR_WRAP:
+      return STENCILOP_INCR;
+   case PIPE_STENCIL_OP_DECR_WRAP:
+      return STENCILOP_DECR;
+   case PIPE_STENCIL_OP_INVERT:
+      return STENCILOP_INVERT;
+   default:
+      return STENCILOP_ZERO;
+   }
+}
+
+static INLINE unsigned
+i915_translate_blend_factor(unsigned factor)
+{
+   switch (factor) {
+   case PIPE_BLENDFACTOR_ZERO:
+      return BLENDFACT_ZERO;
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      return BLENDFACT_SRC_ALPHA;
+   case PIPE_BLENDFACTOR_ONE:
+      return BLENDFACT_ONE;
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+      return BLENDFACT_SRC_COLR;
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+      return BLENDFACT_INV_SRC_COLR;
+   case PIPE_BLENDFACTOR_DST_COLOR:
+      return BLENDFACT_DST_COLR;
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+      return BLENDFACT_INV_DST_COLR;
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+      return BLENDFACT_INV_SRC_ALPHA;
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      return BLENDFACT_DST_ALPHA;
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      return BLENDFACT_INV_DST_ALPHA;
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+      return BLENDFACT_SRC_ALPHA_SATURATE;
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+      return BLENDFACT_CONST_COLOR;
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+      return BLENDFACT_INV_CONST_COLOR;
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      return BLENDFACT_CONST_ALPHA;
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+      return BLENDFACT_INV_CONST_ALPHA;
+   default:
+      return BLENDFACT_ZERO;
+   }
+}
+
+static INLINE unsigned
+i915_translate_blend_func(unsigned mode)
+{
+   switch (mode) {
+   case PIPE_BLEND_ADD:
+      return BLENDFUNC_ADD;
+   case PIPE_BLEND_MIN:
+      return BLENDFUNC_MIN;
+   case PIPE_BLEND_MAX:
+      return BLENDFUNC_MAX;
+   case PIPE_BLEND_SUBTRACT:
+      return BLENDFUNC_SUBTRACT;
+   case PIPE_BLEND_REVERSE_SUBTRACT:
+      return BLENDFUNC_REVERSE_SUBTRACT;
+   default:
+      return 0;
+   }
+}
+
+
+static INLINE unsigned
+i915_translate_logic_op(unsigned opcode)
+{
+   switch (opcode) {
+   case PIPE_LOGICOP_CLEAR:
+      return LOGICOP_CLEAR;
+   case PIPE_LOGICOP_AND:
+      return LOGICOP_AND;
+   case PIPE_LOGICOP_AND_REVERSE:
+      return LOGICOP_AND_RVRSE;
+   case PIPE_LOGICOP_COPY:
+      return LOGICOP_COPY;
+   case PIPE_LOGICOP_COPY_INVERTED:
+      return LOGICOP_COPY_INV;
+   case PIPE_LOGICOP_AND_INVERTED:
+      return LOGICOP_AND_INV;
+   case PIPE_LOGICOP_NOOP:
+      return LOGICOP_NOOP;
+   case PIPE_LOGICOP_XOR:
+      return LOGICOP_XOR;
+   case PIPE_LOGICOP_OR:
+      return LOGICOP_OR;
+   case PIPE_LOGICOP_OR_INVERTED:
+      return LOGICOP_OR_INV;
+   case PIPE_LOGICOP_NOR:
+      return LOGICOP_NOR;
+   case PIPE_LOGICOP_EQUIV:
+      return LOGICOP_EQUIV;
+   case PIPE_LOGICOP_INVERT:
+      return LOGICOP_INV;
+   case PIPE_LOGICOP_OR_REVERSE:
+      return LOGICOP_OR_RVRSE;
+   case PIPE_LOGICOP_NAND:
+      return LOGICOP_NAND;
+   case PIPE_LOGICOP_SET:
+      return LOGICOP_SET;
+   default:
+      return LOGICOP_SET;
+   }
+}
+
+
+
+static INLINE boolean i915_validate_vertices( unsigned hw_prim, unsigned nr )
+{
+   boolean ok;
+
+   switch (hw_prim) {
+   case PRIM3D_POINTLIST:
+      ok = (nr >= 1);
+      assert(ok);
+      break;
+   case PRIM3D_LINELIST:
+      ok = (nr >= 2) && (nr % 2) == 0;
+      assert(ok);
+      break;
+   case PRIM3D_LINESTRIP:
+      ok = (nr >= 2);
+      assert(ok);
+      break;
+   case PRIM3D_TRILIST:
+      ok = (nr >= 3) && (nr % 3) == 0;
+      assert(ok);
+      break;
+   case PRIM3D_TRISTRIP:
+      ok = (nr >= 3);
+      assert(ok);
+      break;
+   case PRIM3D_TRIFAN:
+      ok = (nr >= 3);
+      assert(ok);
+      break;
+   case PRIM3D_POLY:
+      ok = (nr >= 3);
+      assert(ok);
+      break;
+   default:
+      assert(0);
+      ok = 0;
+      break;
+   }
+
+   return ok;
+}
+
+#endif
diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c
new file mode 100644
index 0000000000..c09c10601b
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_state_sampler.c
@@ -0,0 +1,299 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "util/u_memory.h"
+
+#include "i915_state_inlines.h"
+#include "i915_context.h"
+#include "i915_reg.h"
+#include "i915_state.h"
+
+
+/*
+ * A note about min_lod & max_lod.
+ *
+ * There is a circular dependancy between the sampler state
+ * and the map state to be submitted to hw.
+ *
+ * Two condition must be meet:
+ * min_lod =< max_lod == true
+ * max_lod =< last_level == true
+ *
+ *
+ * This is all fine and dandy if it where for the fact that max_lod
+ * is set on the map state instead of the sampler state. That is
+ * the max_lod we submit on map is:
+ * max_lod = MIN2(last_level, max_lod);
+ *
+ * So we need to update the map state when we change samplers and
+ * we need to be change the sampler state when map state is changed.
+ * The first part is done by calling i915_update_texture in
+ * i915_update_samplers and the second part is done else where in
+ * code tracking the state changes.
+ */
+
+static void
+i915_update_texture(struct i915_context *i915,
+                    uint unit,
+                    const struct i915_texture *tex,
+                    const struct i915_sampler_state *sampler,
+                    uint state[6]);
+/**
+ * Compute i915 texture sampling state.
+ *
+ * Recalculate all state from scratch.  Perhaps not the most
+ * efficient, but this has gotten complex enough that we need
+ * something which is understandable and reliable.
+ * \param state  returns the 3 words of compute state
+ */
+static void update_sampler(struct i915_context *i915,
+                           uint unit,
+			   const struct i915_sampler_state *sampler,
+			   const struct i915_texture *tex,
+			   unsigned state[3] )
+{
+   const struct pipe_texture *pt = &tex->base;
+   unsigned minlod, lastlod;
+
+   /* Need to do this after updating the maps, which call the
+    * intel_finalize_mipmap_tree and hence can update firstLevel:
+    */
+   state[0] = sampler->state[0];
+   state[1] = sampler->state[1];
+   state[2] = sampler->state[2];
+
+   if (pt->format == PIPE_FORMAT_YCBCR ||
+       pt->format == PIPE_FORMAT_YCBCR_REV)
+      state[0] |= SS2_COLORSPACE_CONVERSION;
+
+   /* 3D textures don't seem to respect the border color.
+    * Fallback if there's ever a danger that they might refer to
+    * it.  
+    * 
+    * Effectively this means fallback on 3D clamp or
+    * clamp_to_border.
+    *
+    * XXX: Check if this is true on i945.  
+    * XXX: Check if this bug got fixed in release silicon.
+    */
+#if 0
+   {
+      const unsigned ws = sampler->templ->wrap_s;
+      const unsigned wt = sampler->templ->wrap_t;
+      const unsigned wr = sampler->templ->wrap_r;
+      if (pt->target == PIPE_TEXTURE_3D &&
+          (sampler->templ->min_img_filter != PIPE_TEX_FILTER_NEAREST ||
+           sampler->templ->mag_img_filter != PIPE_TEX_FILTER_NEAREST) &&
+          (ws == PIPE_TEX_WRAP_CLAMP ||
+           wt == PIPE_TEX_WRAP_CLAMP ||
+           wr == PIPE_TEX_WRAP_CLAMP ||
+           ws == PIPE_TEX_WRAP_CLAMP_TO_BORDER ||
+           wt == PIPE_TEX_WRAP_CLAMP_TO_BORDER || 
+           wr == PIPE_TEX_WRAP_CLAMP_TO_BORDER)) {
+         if (i915->strict_conformance) {
+            assert(0);
+            /* 	    sampler->fallback = true; */
+            /* TODO */
+         }
+      }
+   }
+#endif
+
+   /* See note at the top of file */
+   minlod = sampler->minlod;
+   lastlod = pt->last_level << 4;
+
+   if (lastlod < minlod) {
+      minlod = lastlod;
+   }
+
+   state[1] |= (sampler->minlod << SS3_MIN_LOD_SHIFT);
+   state[1] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT);
+}
+
+
+void i915_update_samplers( struct i915_context *i915 )
+{
+   uint unit;
+
+   i915->current.sampler_enable_nr = 0;
+   i915->current.sampler_enable_flags = 0x0;
+
+   for (unit = 0; unit < i915->num_textures && unit < i915->num_samplers;
+        unit++) {
+      /* determine unit enable/disable by looking for a bound texture */
+      /* could also examine the fragment program? */
+      if (i915->texture[unit]) {
+	 update_sampler( i915,
+	                 unit,
+	                 i915->sampler[unit],       /* sampler state */
+	                 i915->texture[unit],        /* texture */
+	                 i915->current.sampler[unit] /* the result */
+	                 );
+	 i915_update_texture( i915,
+	                      unit,
+	                      i915->texture[unit],          /* texture */
+	                      i915->sampler[unit],          /* sampler state */
+	                      i915->current.texbuffer[unit] );
+
+	 i915->current.sampler_enable_nr++;
+	 i915->current.sampler_enable_flags |= (1 << unit);
+      }
+   }
+
+   i915->hardware_dirty |= I915_HW_SAMPLER | I915_HW_MAP;
+}
+
+
+static uint
+translate_texture_format(enum pipe_format pipeFormat)
+{
+   switch (pipeFormat) {
+   case PIPE_FORMAT_L8_UNORM:
+      return MAPSURF_8BIT | MT_8BIT_L8;
+   case PIPE_FORMAT_I8_UNORM:
+      return MAPSURF_8BIT | MT_8BIT_I8;
+   case PIPE_FORMAT_A8_UNORM:
+      return MAPSURF_8BIT | MT_8BIT_A8;
+   case PIPE_FORMAT_A8L8_UNORM:
+      return MAPSURF_16BIT | MT_16BIT_AY88;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      return MAPSURF_16BIT | MT_16BIT_RGB565;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      return MAPSURF_16BIT | MT_16BIT_ARGB1555;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      return MAPSURF_16BIT | MT_16BIT_ARGB4444;
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      return MAPSURF_32BIT | MT_32BIT_ARGB8888;
+   case PIPE_FORMAT_YCBCR_REV:
+      return (MAPSURF_422 | MT_422_YCRCB_NORMAL);
+   case PIPE_FORMAT_YCBCR:
+      return (MAPSURF_422 | MT_422_YCRCB_SWAPY);
+#if 0
+   case PIPE_FORMAT_RGB_FXT1:
+   case PIPE_FORMAT_RGBA_FXT1:
+      return (MAPSURF_COMPRESSED | MT_COMPRESS_FXT1);
+#endif
+   case PIPE_FORMAT_Z16_UNORM:
+      return (MAPSURF_16BIT | MT_16BIT_L16);
+#if 0
+   case PIPE_FORMAT_RGBA_DXT1:
+   case PIPE_FORMAT_RGB_DXT1:
+      return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT1);
+   case PIPE_FORMAT_RGBA_DXT3:
+      return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3);
+   case PIPE_FORMAT_RGBA_DXT5:
+      return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5);
+#endif
+   case PIPE_FORMAT_S8Z24_UNORM:
+      return (MAPSURF_32BIT | MT_32BIT_xI824);
+   default:
+      debug_printf("i915: translate_texture_format() bad image format %x\n",
+              pipeFormat);
+      assert(0);
+      return 0;
+   }
+}
+
+
+static void
+i915_update_texture(struct i915_context *i915,
+                    uint unit,
+                    const struct i915_texture *tex,
+                    const struct i915_sampler_state *sampler,
+                    uint state[6])
+{
+   const struct pipe_texture *pt = &tex->base;
+   uint format, pitch;
+   const uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0];
+   const uint num_levels = pt->last_level;
+   unsigned max_lod = num_levels * 4;
+   unsigned tiled = MS3_USE_FENCE_REGS;
+
+   assert(tex);
+   assert(width);
+   assert(height);
+   assert(depth);
+
+   format = translate_texture_format(pt->format);
+   pitch = tex->stride;
+
+   assert(format);
+   assert(pitch);
+
+   if (tex->tiled) {
+      assert(!((pitch - 1) & pitch));
+      tiled = MS3_TILED_SURFACE;
+   }
+
+   /* MS3 state */
+   state[0] =
+      (((height - 1) << MS3_HEIGHT_SHIFT)
+       | ((width - 1) << MS3_WIDTH_SHIFT)
+       | format
+       | tiled);
+
+   /*
+    * XXX When min_filter != mag_filter and there's just one mipmap level,
+    * set max_lod = 1 to make sure i915 chooses between min/mag filtering.
+    */
+
+   /* See note at the top of file */
+   if (max_lod > (sampler->maxlod >> 2))
+      max_lod = sampler->maxlod >> 2;
+
+   /* MS4 state */
+   state[1] =
+      ((((pitch / 4) - 1) << MS4_PITCH_SHIFT)
+       | MS4_CUBE_FACE_ENA_MASK
+       | ((max_lod) << MS4_MAX_LOD_SHIFT)
+       | ((depth - 1) << MS4_VOLUME_DEPTH_SHIFT));
+}
+
+
+void
+i915_update_textures(struct i915_context *i915)
+{
+   uint unit;
+
+   for (unit = 0; unit < i915->num_textures && unit < i915->num_samplers;
+        unit++) {
+      /* determine unit enable/disable by looking for a bound texture */
+      /* could also examine the fragment program? */
+      if (i915->texture[unit]) {
+	 i915_update_texture( i915,
+	                      unit,
+	                      i915->texture[unit],          /* texture */
+	                      i915->sampler[unit],          /* sampler state */
+	                      i915->current.texbuffer[unit] );
+      }
+   }
+
+   i915->hardware_dirty |= I915_HW_MAP;
+}
diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c
new file mode 100644
index 0000000000..94e2deaf61
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_surface.c
@@ -0,0 +1,126 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "i915_context.h"
+#include "i915_blit.h"
+#include "i915_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_tile.h"
+#include "util/u_rect.h"
+
+
+/* Assumes all values are within bounds -- no checking at this level -
+ * do it higher up if required.
+ */
+static void
+i915_surface_copy(struct pipe_context *pipe,
+                  boolean do_flip,
+		  struct pipe_surface *dst,
+		  unsigned dstx, unsigned dsty,
+		  struct pipe_surface *src,
+		  unsigned srcx, unsigned srcy, unsigned width, unsigned height)
+{
+   assert( dst != src );
+   assert( dst->block.size == src->block.size );
+   assert( dst->block.width == src->block.height );
+   assert( dst->block.height == src->block.height );
+
+   if (0) {
+      void *dst_map = pipe->screen->surface_map( pipe->screen,
+                                                 dst,
+                                                 PIPE_BUFFER_USAGE_CPU_WRITE );
+      
+      const void *src_map = pipe->screen->surface_map( pipe->screen,
+                                                       src,
+                                                       PIPE_BUFFER_USAGE_CPU_READ );
+      
+      pipe_copy_rect(dst_map,
+                     &dst->block,
+                     dst->stride,
+                     dstx, dsty,
+                     width, height,
+                     src_map,
+                     do_flip ? -(int) src->stride : src->stride,
+                     srcx, do_flip ? height - 1 - srcy : srcy);
+
+      pipe->screen->surface_unmap(pipe->screen, src);
+      pipe->screen->surface_unmap(pipe->screen, dst);
+   }
+   else {
+      struct i915_texture *dst_tex = (struct i915_texture *)dst->texture;
+      struct i915_texture *src_tex = (struct i915_texture *)src->texture;
+      assert(dst->block.width == 1);
+      assert(dst->block.height == 1);
+      i915_copy_blit( i915_context(pipe),
+                      do_flip,
+                      dst->block.size,
+		      (unsigned short) src->stride, src_tex->buffer, src->offset,
+		      (unsigned short) dst->stride, dst_tex->buffer, dst->offset,
+		      (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height );
+   }
+}
+
+
+static void
+i915_surface_fill(struct pipe_context *pipe,
+		  struct pipe_surface *dst,
+		  unsigned dstx, unsigned dsty,
+		  unsigned width, unsigned height, unsigned value)
+{
+   if (0) {
+      void *dst_map = pipe->screen->surface_map( pipe->screen,
+                                                 dst,
+                                                 PIPE_BUFFER_USAGE_CPU_WRITE );
+
+      pipe_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value);
+
+      pipe->screen->surface_unmap(pipe->screen, dst);
+   }
+   else {
+      struct i915_texture *tex = (struct i915_texture *)dst->texture;
+      assert(dst->block.width == 1);
+      assert(dst->block.height == 1);
+      i915_fill_blit( i915_context(pipe),
+		      dst->block.size,
+		      (unsigned short) dst->stride,
+		      tex->buffer, dst->offset,
+		      (short) dstx, (short) dsty,
+		      (short) width, (short) height,
+		      value );
+   }
+}
+
+
+void
+i915_init_surface_functions(struct i915_context *i915)
+{
+   i915->pipe.surface_copy = i915_surface_copy;
+   i915->pipe.surface_fill = i915_surface_fill;
+}
diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c
new file mode 100644
index 0000000000..b2ca3a2286
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_texture.c
@@ -0,0 +1,771 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Michel Dänzer <michel@tungstengraphics.com>
+  */
+
+#include "pipe/p_state.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "i915_context.h"
+#include "i915_texture.h"
+#include "i915_debug.h"
+#include "i915_screen.h"
+
+/*
+ * Helper function and arrays
+ */
+
+/**
+ * Initial offset for Cube map.
+ */
+static const int initial_offsets[6][2] = {
+   {0, 0},
+   {0, 2},
+   {1, 0},
+   {1, 2},
+   {1, 1},
+   {1, 3}
+};
+
+/**
+ * Step offsets for Cube map.
+ */
+static const int step_offsets[6][2] = {
+   {0, 2},
+   {0, 2},
+   {-1, 2},
+   {-1, 2},
+   {-1, 1},
+   {-1, 1}
+};
+
+static unsigned minify( unsigned d )
+{
+   return MAX2(1, d>>1);
+}
+
+static unsigned
+power_of_two(unsigned x)
+{
+   unsigned value = 1;
+   while (value < x)
+      value = value << 1;
+   return value;
+}
+
+static unsigned
+round_up(unsigned n, unsigned multiple)
+{
+   return (n + multiple - 1) & ~(multiple - 1);
+}
+
+
+/*
+ * More advanced helper funcs
+ */
+
+
+static void
+i915_miptree_set_level_info(struct i915_texture *tex,
+                             unsigned level,
+                             unsigned nr_images,
+                             unsigned w, unsigned h, unsigned d)
+{
+   struct pipe_texture *pt = &tex->base;
+
+   assert(level < PIPE_MAX_TEXTURE_LEVELS);
+
+   pt->width[level] = w;
+   pt->height[level] = h;
+   pt->depth[level] = d;
+   
+   pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w);
+   pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h);
+
+   tex->nr_images[level] = nr_images;
+
+   /*
+   DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
+       level, w, h, d, x, y, tex->level_offset[level]);
+   */
+
+   /* Not sure when this would happen, but anyway: 
+    */
+   if (tex->image_offset[level]) {
+      FREE(tex->image_offset[level]);
+      tex->image_offset[level] = NULL;
+   }
+
+   assert(nr_images);
+   assert(!tex->image_offset[level]);
+
+   tex->image_offset[level] = (unsigned *) MALLOC(nr_images * sizeof(unsigned));
+   tex->image_offset[level][0] = 0;
+}
+
+static void
+i915_miptree_set_image_offset(struct i915_texture *tex,
+			      unsigned level, unsigned img, unsigned x, unsigned y)
+{
+   if (img == 0 && level == 0)
+      assert(x == 0 && y == 0);
+
+   assert(img < tex->nr_images[level]);
+
+   tex->image_offset[level][img] = y * tex->stride + x * tex->base.block.size;
+
+   /*
+   printf("%s level %d img %d pos %d,%d image_offset %x\n",
+       __FUNCTION__, level, img, x, y, tex->image_offset[level][img]);
+   */
+}
+
+
+/*
+ * Layout functions
+ */
+
+
+/**
+ * Special case to deal with display targets.
+ */
+static boolean
+i915_displaytarget_layout(struct i915_texture *tex)
+{
+   struct pipe_texture *pt = &tex->base;
+
+   if (pt->last_level > 0 || pt->block.size != 4)
+      return 0;
+
+   i915_miptree_set_level_info( tex, 0, 1,
+                                tex->base.width[0],
+                                tex->base.height[0],
+                                1 );
+   i915_miptree_set_image_offset( tex, 0, 0, 0, 0 );
+
+   if (tex->base.width[0] >= 128) {
+      tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size);
+      tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8);
+#if 0 /* used for tiled display targets */
+      tex->tiled = 1;
+#endif
+   } else {
+      tex->stride = round_up(tex->base.nblocksx[0] * pt->block.size, 64);
+      tex->total_nblocksy = tex->base.nblocksy[0];
+   }
+
+   /*
+   printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
+      tex->base.width[0], tex->base.height[0], pt->block.size,
+      tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy);
+   */
+
+   return 1;
+}
+
+static void
+i945_miptree_layout_2d( struct i915_texture *tex )
+{
+   struct pipe_texture *pt = &tex->base;
+   const int align_x = 2, align_y = 4;
+   unsigned level;
+   unsigned x = 0;
+   unsigned y = 0;
+   unsigned width = pt->width[0];
+   unsigned height = pt->height[0];
+   unsigned nblocksx = pt->nblocksx[0];
+   unsigned nblocksy = pt->nblocksy[0];
+
+   /* used for tiled display targets */
+   if (0)
+      if (i915_displaytarget_layout(tex))
+	 return;
+
+   tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+
+   /* May need to adjust pitch to accomodate the placement of
+    * the 2nd mipmap level.  This occurs when the alignment
+    * constraints of mipmap placement push the right edge of the
+    * 2nd mipmap level out past the width of its parent.
+    */
+   if (pt->last_level > 0) {
+      unsigned mip1_nblocksx 
+	 = align(pf_get_nblocksx(&pt->block, minify(width)), align_x)
+         + pf_get_nblocksx(&pt->block, minify(minify(width)));
+
+      if (mip1_nblocksx > nblocksx)
+	 tex->stride = mip1_nblocksx * pt->block.size;
+   }
+
+   /* Pitch must be a whole number of dwords
+    */
+   tex->stride = align(tex->stride, 64);
+   tex->total_nblocksy = 0;
+
+   for (level = 0; level <= pt->last_level; level++) {
+      i915_miptree_set_level_info(tex, level, 1, width, height, 1);
+      i915_miptree_set_image_offset(tex, level, 0, x, y);
+
+      nblocksy = align(nblocksy, align_y);
+
+      /* Because the images are packed better, the final offset
+       * might not be the maximal one:
+       */
+      tex->total_nblocksy = MAX2(tex->total_nblocksy, y + nblocksy);
+
+      /* Layout_below: step right after second mipmap level.
+       */
+      if (level == 1) {
+	 x += align(nblocksx, align_x);
+      }
+      else {
+	 y += nblocksy;
+      }
+
+      width  = minify(width);
+      height = minify(height);
+      nblocksx = pf_get_nblocksx(&pt->block, width);
+      nblocksy = pf_get_nblocksy(&pt->block, height);
+   }
+}
+
+static void
+i945_miptree_layout_cube(struct i915_texture *tex)
+{
+   struct pipe_texture *pt = &tex->base;
+   unsigned level;
+
+   const unsigned nblocks = pt->nblocksx[0];
+   unsigned face;
+   unsigned width = pt->width[0];
+   unsigned height = pt->height[0];
+
+   /*
+   printf("%s %i, %i\n", __FUNCTION__, pt->width[0], pt->height[0]);
+   */
+
+   assert(width == height); /* cubemap images are square */
+
+   /*
+    * XXX Should only be used for compressed formats. But lets
+    * keep this code active just in case.
+    *
+    * Depending on the size of the largest images, pitch can be
+    * determined either by the old-style packing of cubemap faces,
+    * or the final row of 4x4, 2x2 and 1x1 faces below this.
+    */
+   if (nblocks > 32)
+      tex->stride = round_up(nblocks * pt->block.size * 2, 4);
+   else
+      tex->stride = 14 * 8 * pt->block.size;
+
+   tex->total_nblocksy = nblocks * 4;
+
+   /* Set all the levels to effectively occupy the whole rectangular region.
+   */
+   for (level = 0; level <= pt->last_level; level++) {
+      i915_miptree_set_level_info(tex, level, 6, width, height, 1);
+      width /= 2;
+      height /= 2;
+   }
+
+   for (face = 0; face < 6; face++) {
+      unsigned x = initial_offsets[face][0] * nblocks;
+      unsigned y = initial_offsets[face][1] * nblocks;
+      unsigned d = nblocks;
+
+#if 0 /* Fix and enable this code for compressed formats */
+      if (nblocks == 4 && face >= 4) {
+         y = tex->total_height - 4;
+         x = (face - 4) * 8;
+      }
+      else if (nblocks < 4 && (face > 0)) {
+         y = tex->total_height - 4;
+         x = face * 8;
+      }
+#endif
+
+      for (level = 0; level <= pt->last_level; level++) {
+         i915_miptree_set_image_offset(tex, level, face, x, y);
+
+         d >>= 1;
+
+#if 0 /* Fix and enable this code for compressed formats */
+         switch (d) {
+            case 4:
+               switch (face) {
+                  case PIPE_TEX_FACE_POS_X:
+                  case PIPE_TEX_FACE_NEG_X:
+                     x += step_offsets[face][0] * d;
+                     y += step_offsets[face][1] * d;
+                     break;
+                  case PIPE_TEX_FACE_POS_Y:
+                  case PIPE_TEX_FACE_NEG_Y:
+                     y += 12;
+                     x -= 8;
+                     break;
+                  case PIPE_TEX_FACE_POS_Z:
+                  case PIPE_TEX_FACE_NEG_Z:
+                     y = tex->total_height - 4;
+                     x = (face - 4) * 8;
+                     break;
+               }
+            case 2:
+               y = tex->total_height - 4;
+               x = 16 + face * 8;
+               break;
+
+            case 1:
+               x += 48;
+               break;
+            default:
+#endif
+               x += step_offsets[face][0] * d;
+               y += step_offsets[face][1] * d;
+#if 0
+               break;
+         }
+#endif
+      }
+   }
+}
+
+static boolean
+i915_miptree_layout(struct i915_texture * tex)
+{
+   struct pipe_texture *pt = &tex->base;
+   unsigned level;
+
+   switch (pt->target) {
+   case PIPE_TEXTURE_CUBE: {
+         const unsigned nblocks = pt->nblocksx[0];
+         unsigned face;
+         unsigned width = pt->width[0], height = pt->height[0];
+
+         assert(width == height); /* cubemap images are square */
+
+         /* double pitch for cube layouts */
+         tex->stride = round_up(nblocks * pt->block.size * 2, 4);
+         tex->total_nblocksy = nblocks * 4;
+
+         for (level = 0; level <= pt->last_level; level++) {
+            i915_miptree_set_level_info(tex, level, 6,
+                                         width, height,
+                                         1);
+            width /= 2;
+            height /= 2;
+         }
+
+         for (face = 0; face < 6; face++) {
+            unsigned x = initial_offsets[face][0] * nblocks;
+            unsigned y = initial_offsets[face][1] * nblocks;
+            unsigned d = nblocks;
+
+            for (level = 0; level <= pt->last_level; level++) {
+               i915_miptree_set_image_offset(tex, level, face, x, y);
+               d >>= 1;
+               x += step_offsets[face][0] * d;
+               y += step_offsets[face][1] * d;
+            }
+         }
+         break;
+      }
+   case PIPE_TEXTURE_3D:{
+         unsigned width = pt->width[0];
+         unsigned height = pt->height[0];
+         unsigned depth = pt->depth[0];
+         unsigned nblocksx = pt->nblocksx[0];
+         unsigned nblocksy = pt->nblocksy[0];
+         unsigned stack_nblocksy = 0;
+
+         /* Calculate the size of a single slice. 
+          */
+         tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+
+         /* XXX: hardware expects/requires 9 levels at minimum.
+          */
+         for (level = 0; level <= MAX2(8, pt->last_level);
+              level++) {
+            i915_miptree_set_level_info(tex, level, depth,
+                                        width, height, depth);
+
+
+            stack_nblocksy += MAX2(2, nblocksy);
+
+            width = minify(width);
+            height = minify(height);
+            depth = minify(depth);
+            nblocksx = pf_get_nblocksx(&pt->block, width);
+            nblocksy = pf_get_nblocksy(&pt->block, height);
+         }
+
+         /* Fixup depth image_offsets: 
+          */
+         depth = pt->depth[0];
+         for (level = 0; level <= pt->last_level; level++) {
+            unsigned i;
+            for (i = 0; i < depth; i++) 
+               i915_miptree_set_image_offset(tex, level, i,
+                                             0, i * stack_nblocksy);
+
+            depth = minify(depth);
+         }
+
+
+         /* Multiply slice size by texture depth for total size.  It's
+          * remarkable how wasteful of memory the i915 texture layouts
+          * are.  They are largely fixed in the i945.
+          */
+         tex->total_nblocksy = stack_nblocksy * pt->depth[0];
+         break;
+      }
+
+   default:{
+         unsigned width = pt->width[0];
+         unsigned height = pt->height[0];
+         unsigned nblocksx = pt->nblocksx[0];
+         unsigned nblocksy = pt->nblocksy[0];
+
+         tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+         tex->total_nblocksy = 0;
+
+         for (level = 0; level <= pt->last_level; level++) {
+            i915_miptree_set_level_info(tex, level, 1,
+                                        width, height, 1);
+            i915_miptree_set_image_offset(tex, level, 0,
+                                          0, tex->total_nblocksy);
+
+            nblocksy = round_up(MAX2(2, nblocksy), 2);
+
+	    tex->total_nblocksy += nblocksy;
+
+            width = minify(width);
+            height = minify(height);
+            nblocksx = pf_get_nblocksx(&pt->block, width);
+            nblocksy = pf_get_nblocksy(&pt->block, height);
+         }
+         break;
+      }
+   }
+   /*
+   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__,
+       tex->pitch,
+       tex->total_nblocksy, pt->block.size, tex->stride * tex->total_nblocksy);
+   */
+
+   return TRUE;
+}
+
+
+static boolean
+i945_miptree_layout(struct i915_texture * tex)
+{
+   struct pipe_texture *pt = &tex->base;
+   unsigned level;
+
+   switch (pt->target) {
+   case PIPE_TEXTURE_CUBE:
+      i945_miptree_layout_cube(tex);
+      break;
+   case PIPE_TEXTURE_3D:{
+         unsigned width = pt->width[0];
+         unsigned height = pt->height[0];
+         unsigned depth = pt->depth[0];
+         unsigned nblocksx = pt->nblocksx[0];
+         unsigned nblocksy = pt->nblocksy[0];
+         unsigned pack_x_pitch, pack_x_nr;
+         unsigned pack_y_pitch;
+
+         tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4);
+         tex->total_nblocksy = 0;
+
+         pack_y_pitch = MAX2(pt->nblocksy[0], 2);
+         pack_x_pitch = tex->stride / pt->block.size;
+         pack_x_nr = 1;
+
+         for (level = 0; level <= pt->last_level; level++) {
+            unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6;
+            int x = 0;
+            int y = 0;
+            unsigned q, j;
+
+            i915_miptree_set_level_info(tex, level, nr_images,
+                                        width, height, depth);
+
+            for (q = 0; q < nr_images;) {
+               for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) {
+                  i915_miptree_set_image_offset(tex, level, q, x, y + tex->total_nblocksy);
+                  x += pack_x_pitch;
+               }
+
+               x = 0;
+               y += pack_y_pitch;
+            }
+
+
+            tex->total_nblocksy += y;
+
+            if (pack_x_pitch > 4) {
+               pack_x_pitch >>= 1;
+               pack_x_nr <<= 1;
+               assert(pack_x_pitch * pack_x_nr * pt->block.size <= tex->stride);
+            }
+
+            if (pack_y_pitch > 2) {
+               pack_y_pitch >>= 1;
+            }
+
+            width = minify(width);
+            height = minify(height);
+            depth = minify(depth);
+            nblocksx = pf_get_nblocksx(&pt->block, width);
+            nblocksy = pf_get_nblocksy(&pt->block, height);
+         }
+         break;
+      }
+
+   case PIPE_TEXTURE_1D:
+   case PIPE_TEXTURE_2D:
+//   case PIPE_TEXTURE_RECTANGLE:
+         i945_miptree_layout_2d(tex);
+         break;
+   default:
+      assert(0);
+      return FALSE;
+   }
+
+   /*
+   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__,
+       tex->pitch,
+       tex->total_nblocksy, pt->block.size, tex->stride * tex->total_nblocksy);
+   */
+
+   return TRUE;
+}
+
+
+static struct pipe_texture *
+i915_texture_create(struct pipe_screen *screen,
+                    const struct pipe_texture *templat)
+{
+   struct i915_screen *i915screen = i915_screen(screen);
+   struct pipe_winsys *ws = screen->winsys;
+   struct i915_texture *tex = CALLOC_STRUCT(i915_texture);
+   size_t tex_size;
+
+   if (!tex)
+      return NULL;
+
+   tex->base = *templat;
+   tex->base.refcount = 1;
+   tex->base.screen = screen;
+
+   tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width[0]);
+   tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height[0]);
+   
+   if (i915screen->is_i945) {
+      if (!i945_miptree_layout(tex))
+	 goto fail;
+   } else {
+      if (!i915_miptree_layout(tex))
+	 goto fail;
+   }
+
+   tex_size = tex->stride * tex->total_nblocksy;
+
+   tex->buffer = ws->buffer_create(ws, 64,
+                                    PIPE_BUFFER_USAGE_PIXEL,
+                                    tex_size);
+
+   if (!tex->buffer)
+      goto fail;
+
+#if 0
+   void *ptr = ws->buffer_map(ws, tex->buffer,
+      PIPE_BUFFER_USAGE_CPU_WRITE);
+   memset(ptr, 0x80, tex_size);
+   ws->buffer_unmap(ws, tex->buffer);
+#endif
+
+   return &tex->base;
+
+fail:
+   FREE(tex);
+   return NULL;
+}
+
+
+static void
+i915_texture_release(struct pipe_screen *screen,
+                     struct pipe_texture **pt)
+{
+   if (!*pt)
+      return;
+
+   /*
+   DBG("%s %p refcount will be %d\n",
+       __FUNCTION__, (void *) *pt, (*pt)->refcount - 1);
+   */
+   if (--(*pt)->refcount <= 0) {
+      struct i915_texture *tex = (struct i915_texture *)*pt;
+      uint i;
+
+      /*
+      DBG("%s deleting %p\n", __FUNCTION__, (void *) tex);
+      */
+
+      pipe_buffer_reference(screen, &tex->buffer, NULL);
+
+      for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++)
+         if (tex->image_offset[i])
+            FREE(tex->image_offset[i]);
+
+      FREE(tex);
+   }
+   *pt = NULL;
+}
+
+static struct pipe_surface *
+i915_get_tex_surface(struct pipe_screen *screen,
+                     struct pipe_texture *pt,
+                     unsigned face, unsigned level, unsigned zslice,
+                     unsigned flags)
+{
+   struct i915_texture *tex = (struct i915_texture *)pt;
+   struct pipe_surface *ps;
+   unsigned offset;  /* in bytes */
+
+   if (pt->target == PIPE_TEXTURE_CUBE) {
+      offset = tex->image_offset[level][face];
+   }
+   else if (pt->target == PIPE_TEXTURE_3D) {
+      offset = tex->image_offset[level][zslice];
+   }
+   else {
+      offset = tex->image_offset[level][0];
+      assert(face == 0);
+      assert(zslice == 0);
+   }
+
+   ps = CALLOC_STRUCT(pipe_surface);
+   if (ps) {
+      ps->refcount = 1;
+      pipe_texture_reference(&ps->texture, pt);
+      ps->format = pt->format;
+      ps->width = pt->width[level];
+      ps->height = pt->height[level];
+      ps->block = pt->block;
+      ps->nblocksx = pt->nblocksx[level];
+      ps->nblocksy = pt->nblocksy[level];
+      ps->stride = tex->stride;
+      ps->offset = offset;
+      ps->usage = flags;
+      ps->status = PIPE_SURFACE_STATUS_DEFINED;
+   }
+   return ps;
+}
+
+static struct pipe_texture *
+i915_texture_blanket(struct pipe_screen * screen,
+                     const struct pipe_texture *base,
+                     const unsigned *stride,
+                     struct pipe_buffer *buffer)
+{
+   struct i915_texture *tex;
+   assert(screen);
+
+   /* Only supports one type */
+   if (base->target != PIPE_TEXTURE_2D ||
+       base->last_level != 0 ||
+       base->depth[0] != 1) {
+      return NULL;
+   }
+
+   tex = CALLOC_STRUCT(i915_texture);
+   if (!tex)
+      return NULL;
+
+   tex->base = *base;
+   tex->base.refcount = 1;
+   tex->base.screen = screen;
+
+   tex->stride = stride[0];
+
+   i915_miptree_set_level_info(tex, 0, 1, base->width[0], base->height[0], 1);
+   i915_miptree_set_image_offset(tex, 0, 0, 0, 0);
+
+   pipe_buffer_reference(screen, &tex->buffer, buffer);
+
+   return &tex->base;
+}
+
+void
+i915_init_texture_functions(struct i915_context *i915)
+{
+//   i915->pipe.texture_update = i915_texture_update;
+}
+
+static void
+i915_tex_surface_release(struct pipe_screen *screen,
+                         struct pipe_surface **surface)
+{
+   struct pipe_surface *surf = *surface;
+
+   if (--surf->refcount == 0) {
+
+      /* This really should not be possible, but it's actually
+       * happening quite a bit...  Will fix.
+       */
+      if (surf->status == PIPE_SURFACE_STATUS_CLEAR) {
+         debug_printf("XXX destroying a surface with pending clears...\n");
+         assert(0);
+      }
+
+      pipe_texture_reference(&surf->texture, NULL);
+      FREE(surf);
+   }
+
+   *surface = NULL;
+}
+
+void
+i915_init_screen_texture_functions(struct pipe_screen *screen)
+{
+   screen->texture_create = i915_texture_create;
+   screen->texture_release = i915_texture_release;
+   screen->get_tex_surface = i915_get_tex_surface;
+   screen->texture_blanket = i915_texture_blanket;
+   screen->tex_surface_release = i915_tex_surface_release;
+}
diff --git a/src/gallium/drivers/i915simple/i915_texture.h b/src/gallium/drivers/i915simple/i915_texture.h
new file mode 100644
index 0000000000..7225016a9f
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_texture.h
@@ -0,0 +1,43 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef I915_TEXTURE_H
+#define I915_TEXTURE_H
+
+struct i915_context;
+struct pipe_screen;
+
+
+extern void
+i915_init_texture_functions(struct i915_context *i915);
+
+
+extern void
+i915_init_screen_texture_functions(struct pipe_screen *screen);
+
+
+#endif /* I915_TEXTURE_H */
diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h
new file mode 100644
index 0000000000..81904c2a74
--- /dev/null
+++ b/src/gallium/drivers/i915simple/i915_winsys.h
@@ -0,0 +1,121 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \file
+ * This is the interface that i915simple requires any window system
+ * hosting it to implement.  This is the only include file in i915simple
+ * which is public.
+ * 
+ */
+
+#ifndef I915_WINSYS_H
+#define I915_WINSYS_H
+
+
+#include "pipe/p_defines.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+   
+/* Pipe drivers are (meant to be!) independent of both GL and the
+ * window system.  The window system provides a buffer manager and a
+ * set of additional hooks for things like command buffer submission,
+ * etc.
+ *
+ * There clearly has to be some agreement between the window system
+ * driver and the hardware driver about the format of command buffers,
+ * etc.
+ */
+
+struct i915_batchbuffer;
+struct pipe_buffer;
+struct pipe_fence_handle;
+struct pipe_winsys;
+struct pipe_screen;
+
+
+/**
+ * Additional winsys interface for i915simple.
+ * 
+ * It is an over-simple batchbuffer mechanism.  Will want to improve the
+ * performance of this, perhaps based on the cmdstream stuff.  It
+ * would be pretty impossible to implement swz on top of this
+ * interface.
+ *
+ * Will also need additions/changes to implement static/dynamic
+ * indirect state.
+ */
+struct i915_winsys {
+
+   void (*destroy)( struct i915_winsys *sws );
+   
+   /**
+    * Get the current batch buffer from the winsys.
+    */
+   struct i915_batchbuffer *(*batch_get)( struct i915_winsys *sws );
+
+   /**
+    * Emit a relocation to a buffer.
+    * 
+    * Used not only when the buffer addresses are not pinned, but also to 
+    * ensure refered buffers will not be destroyed until the current batch 
+    * buffer execution is finished.
+    *
+    * The access flags is a combination of I915_BUFFER_ACCESS_WRITE and 
+    * I915_BUFFER_ACCESS_READ macros.
+    */
+   void (*batch_reloc)( struct i915_winsys *sws,
+			struct pipe_buffer *buf,
+			unsigned access_flags,
+			unsigned delta );
+
+   /**
+    * Flush the batch.
+    */
+   void (*batch_flush)( struct i915_winsys *sws,
+                        struct pipe_fence_handle **fence );
+};
+
+#define I915_BUFFER_ACCESS_WRITE   0x1 
+#define I915_BUFFER_ACCESS_READ    0x2
+
+#define I915_BUFFER_USAGE_LIT_VERTEX  (PIPE_BUFFER_USAGE_CUSTOM << 0)
+
+
+struct pipe_context *i915_create_context( struct pipe_screen *,
+                                          struct pipe_winsys *,
+                                          struct i915_winsys * );
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif 
diff --git a/src/gallium/drivers/i965simple/Makefile b/src/gallium/drivers/i965simple/Makefile
new file mode 100644
index 0000000000..e97146e57c
--- /dev/null
+++ b/src/gallium/drivers/i965simple/Makefile
@@ -0,0 +1,54 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = i965simple
+
+C_SOURCES = \
+	brw_blit.c \
+	brw_flush.c \
+	brw_screen.c \
+	brw_surface.c \
+	brw_cc.c \
+	brw_clip.c \
+	brw_clip_line.c \
+	brw_clip_point.c \
+	brw_clip_state.c \
+	brw_clip_tri.c \
+	brw_clip_util.c \
+	brw_context.c \
+	brw_curbe.c \
+	brw_draw.c \
+	brw_draw_upload.c \
+	brw_eu.c \
+	brw_eu_debug.c \
+	brw_eu_emit.c \
+	brw_eu_util.c \
+	brw_gs.c \
+	brw_gs_emit.c \
+	brw_gs_state.c \
+	brw_misc_state.c \
+	brw_sf.c \
+	brw_sf_emit.c \
+	brw_sf_state.c \
+	brw_state.c \
+	brw_state_batch.c \
+	brw_state_cache.c \
+	brw_state_pool.c \
+	brw_state_upload.c \
+	brw_tex_layout.c \
+	brw_urb.c \
+	brw_util.c \
+	brw_vs.c \
+	brw_vs_emit.c \
+	brw_vs_state.c \
+	brw_wm.c \
+	brw_wm_iz.c \
+	brw_wm_decl.c \
+	brw_wm_glsl.c \
+	brw_wm_sampler_state.c \
+	brw_wm_state.c \
+	brw_wm_surface_state.c
+
+include ../../Makefile.template
+
+symlinks:
diff --git a/src/gallium/drivers/i965simple/SConscript b/src/gallium/drivers/i965simple/SConscript
new file mode 100644
index 0000000000..43fc2a4005
--- /dev/null
+++ b/src/gallium/drivers/i965simple/SConscript
@@ -0,0 +1,54 @@
+Import('*')
+
+env = env.Clone()
+
+i965simple = env.ConvenienceLibrary(
+	target = 'i965simple',
+	source = [
+		'brw_blit.c',
+		'brw_cc.c',
+		'brw_clip.c',
+		'brw_clip_line.c',
+		'brw_clip_point.c',
+		'brw_clip_state.c',
+		'brw_clip_tri.c',
+		'brw_clip_util.c',
+		'brw_context.c',
+		'brw_curbe.c',
+		'brw_draw.c',
+		'brw_draw_upload.c',
+		'brw_eu.c',
+		'brw_eu_debug.c',
+		'brw_eu_emit.c',
+		'brw_eu_util.c',
+		'brw_flush.c',
+		'brw_gs.c',
+		'brw_gs_emit.c',
+		'brw_gs_state.c',
+		'brw_misc_state.c',
+		'brw_screen.c',
+		'brw_sf.c',
+		'brw_sf_emit.c',
+		'brw_sf_state.c',
+		'brw_state.c',
+		'brw_state_batch.c',
+		'brw_state_cache.c',
+		'brw_state_pool.c',
+		'brw_state_upload.c',
+		'brw_surface.c',
+		'brw_tex_layout.c',
+		'brw_urb.c',
+		'brw_util.c',
+		'brw_vs.c',
+		'brw_vs_emit.c',
+		'brw_vs_state.c',
+		'brw_wm.c',
+		'brw_wm_decl.c',
+		'brw_wm_glsl.c',
+		'brw_wm_iz.c',
+		'brw_wm_sampler_state.c',
+		'brw_wm_state.c',
+		'brw_wm_surface_state.c',
+	])
+
+Export('i965simple')
diff --git a/src/gallium/drivers/i965simple/brw_batch.h b/src/gallium/drivers/i965simple/brw_batch.h
new file mode 100644
index 0000000000..5f5932a488
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_batch.h
@@ -0,0 +1,59 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef BRW_BATCH_H
+#define BRW_BATCH_H
+
+#include "brw_winsys.h"
+
+#define BATCH_LOCALS
+
+#define INTEL_BATCH_NO_CLIPRECTS 0x1
+#define INTEL_BATCH_CLIPRECTS    0x2
+
+#define BEGIN_BATCH( dwords, relocs ) \
+   brw->winsys->batch_start(brw->winsys, dwords, relocs)
+
+#define OUT_BATCH( dword ) \
+   brw->winsys->batch_dword(brw->winsys, dword)
+
+#define OUT_RELOC( buf, flags, delta ) \
+   brw->winsys->batch_reloc(brw->winsys, buf, flags, delta)
+
+#define ADVANCE_BATCH() \
+   brw->winsys->batch_end( brw->winsys )
+
+/* XXX: this is bogus - need proper handling for out-of-memory in batchbuffer.
+ */
+#define FLUSH_BATCH(fence) do {				\
+   brw->winsys->batch_flush(brw->winsys, fence);	\
+   brw->hardware_dirty = ~0;				\
+} while (0)
+
+#define BRW_BATCH_STRUCT(brw, s) brw_batchbuffer_data( brw->winsys, (s), sizeof(*(s)))
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_blit.c b/src/gallium/drivers/i965simple/brw_blit.c
new file mode 100644
index 0000000000..4d11f8d2ab
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_blit.c
@@ -0,0 +1,218 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include <stdio.h>
+#include <errno.h>
+
+#include "brw_batch.h"
+#include "brw_blit.h"
+#include "brw_context.h"
+#include "brw_reg.h"
+
+#include "pipe/p_context.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+#define FILE_DEBUG_FLAG DEBUG_BLIT
+
+void brw_fill_blit(struct brw_context *brw,
+                   unsigned cpp,
+                   short dst_pitch,
+                   struct pipe_buffer *dst_buffer,
+                   unsigned dst_offset,
+                   boolean dst_tiled,
+                   short x, short y,
+                   short w, short h,
+                   unsigned color)
+{
+   unsigned BR13, CMD;
+   BATCH_LOCALS;
+
+   dst_pitch *= cpp;
+
+   switch(cpp) {
+   case 1:
+   case 2:
+   case 3:
+      BR13 = (0xF0 << 16) | (1<<24);
+      CMD = XY_COLOR_BLT_CMD;
+      break;
+   case 4:
+      BR13 = (0xF0 << 16) | (1<<24) | (1<<25);
+      CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
+      break;
+   default:
+      return;
+   }
+
+   if (dst_tiled) {
+      CMD |= XY_DST_TILED;
+      dst_pitch /= 4;
+   }
+
+   BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS);
+   OUT_BATCH( CMD );
+   OUT_BATCH( dst_pitch | BR13 );
+   OUT_BATCH( (y << 16) | x );
+   OUT_BATCH( ((y+h) << 16) | (x+w) );
+   OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE, dst_offset );
+   OUT_BATCH( color );
+   ADVANCE_BATCH();
+}
+
+static unsigned translate_raster_op(unsigned logicop)
+{
+   switch(logicop) {
+   case PIPE_LOGICOP_CLEAR: return 0x00;
+   case PIPE_LOGICOP_AND: return 0x88;
+   case PIPE_LOGICOP_AND_REVERSE: return 0x44;
+   case PIPE_LOGICOP_COPY: return 0xCC;
+   case PIPE_LOGICOP_AND_INVERTED: return 0x22;
+   case PIPE_LOGICOP_NOOP: return 0xAA;
+   case PIPE_LOGICOP_XOR: return 0x66;
+   case PIPE_LOGICOP_OR: return 0xEE;
+   case PIPE_LOGICOP_NOR: return 0x11;
+   case PIPE_LOGICOP_EQUIV: return 0x99;
+   case PIPE_LOGICOP_INVERT: return 0x55;
+   case PIPE_LOGICOP_OR_REVERSE: return 0xDD;
+   case PIPE_LOGICOP_COPY_INVERTED: return 0x33;
+   case PIPE_LOGICOP_OR_INVERTED: return 0xBB;
+   case PIPE_LOGICOP_NAND: return 0x77;
+   case PIPE_LOGICOP_SET: return 0xFF;
+   default: return 0;
+   }
+}
+
+
+/* Copy BitBlt
+ */
+void brw_copy_blit(struct brw_context *brw,
+                   unsigned do_flip,
+                   unsigned cpp,
+                   short src_pitch,
+                   struct pipe_buffer *src_buffer,
+                   unsigned  src_offset,
+                   boolean src_tiled,
+                   short dst_pitch,
+                   struct pipe_buffer *dst_buffer,
+                   unsigned  dst_offset,
+                   boolean dst_tiled,
+                   short src_x, short src_y,
+                   short dst_x, short dst_y,
+                   short w, short h,
+                   unsigned logic_op)
+{
+   unsigned CMD, BR13;
+   int dst_y2 = dst_y + h;
+   int dst_x2 = dst_x + w;
+   BATCH_LOCALS;
+
+
+   DBG("%s src:buf(%d)/%d %d,%d dst:buf(%d)/%d %d,%d sz:%dx%d op:%d\n",
+       __FUNCTION__,
+       src_buffer, src_pitch, src_x, src_y,
+       dst_buffer, dst_pitch, dst_x, dst_y,
+       w,h,logic_op);
+
+   assert( logic_op - PIPE_LOGICOP_CLEAR >= 0 );
+   assert( logic_op - PIPE_LOGICOP_CLEAR < 0x10 );
+
+   src_pitch *= cpp;
+   dst_pitch *= cpp;
+
+   switch(cpp) {
+   case 1:
+   case 2:
+   case 3:
+      BR13 = (translate_raster_op(logic_op) << 16) | (1<<24);
+      CMD = XY_SRC_COPY_BLT_CMD;
+      break;
+   case 4:
+      BR13 = (translate_raster_op(logic_op) << 16) | (1<<24) |
+	  (1<<25);
+      CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
+      break;
+   default:
+      return;
+   }
+
+   if (src_tiled) {
+      CMD |= XY_SRC_TILED;
+      src_pitch /= 4;
+   }
+
+   if (dst_tiled) {
+      CMD |= XY_DST_TILED;
+      dst_pitch /= 4;
+   }
+
+   if (dst_y2 < dst_y ||
+       dst_x2 < dst_x) {
+      return;
+   }
+
+   dst_pitch &= 0xffff;
+   src_pitch &= 0xffff;
+
+   /* Initial y values don't seem to work with negative pitches.  If
+    * we adjust the offsets manually (below), it seems to work fine.
+    *
+    * On the other hand, if we always adjust, the hardware doesn't
+    * know which blit directions to use, so overlapping copypixels get
+    * the wrong result.
+    */
+   if (dst_pitch > 0 && src_pitch > 0) {
+      BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
+      OUT_BATCH( CMD );
+      OUT_BATCH( dst_pitch | BR13 );
+      OUT_BATCH( (dst_y << 16) | dst_x );
+      OUT_BATCH( (dst_y2 << 16) | dst_x2 );
+      OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE,
+		 dst_offset );
+      OUT_BATCH( (src_y << 16) | src_x );
+      OUT_BATCH( src_pitch );
+      OUT_RELOC( src_buffer, BRW_BUFFER_ACCESS_READ,
+		 src_offset );
+      ADVANCE_BATCH();
+   }
+   else {
+      BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
+      OUT_BATCH( CMD );
+      OUT_BATCH( (dst_pitch & 0xffff) | BR13 );
+      OUT_BATCH( (0 << 16) | dst_x );
+      OUT_BATCH( (h << 16) | dst_x2 );
+      OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE,
+		 dst_offset + dst_y * dst_pitch );
+      OUT_BATCH( (src_pitch & 0xffff) );
+      OUT_RELOC( src_buffer, BRW_BUFFER_ACCESS_READ,
+		 src_offset + src_y * src_pitch );
+      ADVANCE_BATCH();
+   }
+}
+
+
+
diff --git a/src/gallium/drivers/i965simple/brw_blit.h b/src/gallium/drivers/i965simple/brw_blit.h
new file mode 100644
index 0000000000..111c5d91d3
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_blit.h
@@ -0,0 +1,33 @@
+#ifndef BRW_BLIT_H
+#define BRW_BLIT_H
+
+#include "pipe/p_compiler.h"
+
+struct pipe_buffer;
+struct brw_context;
+
+void brw_fill_blit(struct brw_context *intel,
+                   unsigned cpp,
+                   short dst_pitch,
+                   struct pipe_buffer *dst_buffer,
+                   unsigned dst_offset,
+                   boolean dst_tiled,
+                   short x, short y,
+                   short w, short h,
+                   unsigned color);
+void brw_copy_blit(struct brw_context *intel,
+                   unsigned do_flip,
+                   unsigned cpp,
+                   short src_pitch,
+                   struct pipe_buffer *src_buffer,
+                   unsigned  src_offset,
+                   boolean src_tiled,
+                   short dst_pitch,
+                   struct pipe_buffer *dst_buffer,
+                   unsigned  dst_offset,
+                   boolean dst_tiled,
+                   short src_x, short src_y,
+                   short dst_x, short dst_y,
+                   short w, short h,
+                   unsigned logic_op);
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_cc.c b/src/gallium/drivers/i965simple/brw_cc.c
new file mode 100644
index 0000000000..3668123e2e
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_cc.c
@@ -0,0 +1,269 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+#include "brw_util.h"
+
+
+static int brw_translate_compare_func(int func)
+{
+   switch(func) {
+   case PIPE_FUNC_NEVER:
+      return BRW_COMPAREFUNCTION_NEVER;
+   case PIPE_FUNC_LESS:
+      return BRW_COMPAREFUNCTION_LESS;
+   case PIPE_FUNC_LEQUAL:
+      return BRW_COMPAREFUNCTION_LEQUAL;
+   case PIPE_FUNC_GREATER:
+      return BRW_COMPAREFUNCTION_GREATER;
+   case PIPE_FUNC_GEQUAL:
+      return BRW_COMPAREFUNCTION_GEQUAL;
+   case PIPE_FUNC_NOTEQUAL:
+      return BRW_COMPAREFUNCTION_NOTEQUAL;
+   case PIPE_FUNC_EQUAL:
+      return BRW_COMPAREFUNCTION_EQUAL;
+   case PIPE_FUNC_ALWAYS:
+      return BRW_COMPAREFUNCTION_ALWAYS;
+   }
+
+   debug_printf("Unknown value in %s: %x\n", __FUNCTION__, func);
+   return BRW_COMPAREFUNCTION_ALWAYS;
+}
+
+static int brw_translate_stencil_op(int op)
+{
+   switch(op) {
+   case PIPE_STENCIL_OP_KEEP:
+      return BRW_STENCILOP_KEEP;
+   case PIPE_STENCIL_OP_ZERO:
+      return BRW_STENCILOP_ZERO;
+   case PIPE_STENCIL_OP_REPLACE:
+      return BRW_STENCILOP_REPLACE;
+   case PIPE_STENCIL_OP_INCR:
+      return BRW_STENCILOP_INCRSAT;
+   case PIPE_STENCIL_OP_DECR:
+      return BRW_STENCILOP_DECRSAT;
+   case PIPE_STENCIL_OP_INCR_WRAP:
+      return BRW_STENCILOP_INCR;
+   case PIPE_STENCIL_OP_DECR_WRAP:
+      return BRW_STENCILOP_DECR;
+   case PIPE_STENCIL_OP_INVERT:
+      return BRW_STENCILOP_INVERT;
+   default:
+      return BRW_STENCILOP_ZERO;
+   }
+}
+
+
+static int brw_translate_logic_op(int opcode)
+{
+   switch(opcode) {
+   case PIPE_LOGICOP_CLEAR:
+      return BRW_LOGICOPFUNCTION_CLEAR;
+   case PIPE_LOGICOP_AND:
+      return BRW_LOGICOPFUNCTION_AND;
+   case PIPE_LOGICOP_AND_REVERSE:
+      return BRW_LOGICOPFUNCTION_AND_REVERSE;
+   case PIPE_LOGICOP_COPY:
+      return BRW_LOGICOPFUNCTION_COPY;
+   case PIPE_LOGICOP_COPY_INVERTED:
+      return BRW_LOGICOPFUNCTION_COPY_INVERTED;
+   case PIPE_LOGICOP_AND_INVERTED:
+      return BRW_LOGICOPFUNCTION_AND_INVERTED;
+   case PIPE_LOGICOP_NOOP:
+      return BRW_LOGICOPFUNCTION_NOOP;
+   case PIPE_LOGICOP_XOR:
+      return BRW_LOGICOPFUNCTION_XOR;
+   case PIPE_LOGICOP_OR:
+      return BRW_LOGICOPFUNCTION_OR;
+   case PIPE_LOGICOP_OR_INVERTED:
+      return BRW_LOGICOPFUNCTION_OR_INVERTED;
+   case PIPE_LOGICOP_NOR:
+      return BRW_LOGICOPFUNCTION_NOR;
+   case PIPE_LOGICOP_EQUIV:
+      return BRW_LOGICOPFUNCTION_EQUIV;
+   case PIPE_LOGICOP_INVERT:
+      return BRW_LOGICOPFUNCTION_INVERT;
+   case PIPE_LOGICOP_OR_REVERSE:
+      return BRW_LOGICOPFUNCTION_OR_REVERSE;
+   case PIPE_LOGICOP_NAND:
+      return BRW_LOGICOPFUNCTION_NAND;
+   case PIPE_LOGICOP_SET:
+      return BRW_LOGICOPFUNCTION_SET;
+   default:
+      return BRW_LOGICOPFUNCTION_SET;
+   }
+}
+
+
+static void upload_cc_vp( struct brw_context *brw )
+{
+   struct brw_cc_viewport ccv;
+
+   memset(&ccv, 0, sizeof(ccv));
+
+   ccv.min_depth = 0.0;
+   ccv.max_depth = 1.0;
+
+   brw->cc.vp_gs_offset = brw_cache_data( &brw->cache[BRW_CC_VP], &ccv );
+}
+
+const struct brw_tracked_state brw_cc_vp = {
+   .dirty = {
+      .brw = BRW_NEW_SCENE,
+      .cache = 0
+   },
+   .update = upload_cc_vp
+};
+
+
+static void upload_cc_unit( struct brw_context *brw )
+{
+   struct brw_cc_unit_state cc;
+
+   memset(&cc, 0, sizeof(cc));
+
+   /* BRW_NEW_DEPTH_STENCIL */
+   if (brw->attribs.DepthStencil->stencil[0].enabled) {
+      cc.cc0.stencil_enable = brw->attribs.DepthStencil->stencil[0].enabled;
+      cc.cc0.stencil_func = brw_translate_compare_func(brw->attribs.DepthStencil->stencil[0].func);
+      cc.cc0.stencil_fail_op = brw_translate_stencil_op(brw->attribs.DepthStencil->stencil[0].fail_op);
+      cc.cc0.stencil_pass_depth_fail_op = brw_translate_stencil_op(
+         brw->attribs.DepthStencil->stencil[0].zfail_op);
+      cc.cc0.stencil_pass_depth_pass_op = brw_translate_stencil_op(
+         brw->attribs.DepthStencil->stencil[0].zpass_op);
+      cc.cc1.stencil_ref = brw->attribs.DepthStencil->stencil[0].ref_value;
+      cc.cc1.stencil_write_mask = brw->attribs.DepthStencil->stencil[0].writemask;
+      cc.cc1.stencil_test_mask = brw->attribs.DepthStencil->stencil[0].valuemask;
+
+      if (brw->attribs.DepthStencil->stencil[1].enabled) {
+	 cc.cc0.bf_stencil_enable = brw->attribs.DepthStencil->stencil[1].enabled;
+	 cc.cc0.bf_stencil_func = brw_translate_compare_func(
+            brw->attribs.DepthStencil->stencil[1].func);
+	 cc.cc0.bf_stencil_fail_op = brw_translate_stencil_op(
+            brw->attribs.DepthStencil->stencil[1].fail_op);
+	 cc.cc0.bf_stencil_pass_depth_fail_op = brw_translate_stencil_op(
+            brw->attribs.DepthStencil->stencil[1].zfail_op);
+	 cc.cc0.bf_stencil_pass_depth_pass_op = brw_translate_stencil_op(
+            brw->attribs.DepthStencil->stencil[1].zpass_op);
+	 cc.cc1.bf_stencil_ref = brw->attribs.DepthStencil->stencil[1].ref_value;
+	 cc.cc2.bf_stencil_write_mask = brw->attribs.DepthStencil->stencil[1].writemask;
+	 cc.cc2.bf_stencil_test_mask = brw->attribs.DepthStencil->stencil[1].valuemask;
+      }
+
+      /* Not really sure about this:
+       */
+      if (brw->attribs.DepthStencil->stencil[0].writemask ||
+	  brw->attribs.DepthStencil->stencil[1].writemask)
+	 cc.cc0.stencil_write_enable = 1;
+   }
+
+   /* BRW_NEW_BLEND */
+   if (brw->attribs.Blend->logicop_enable) {
+      cc.cc2.logicop_enable = 1;
+      cc.cc5.logicop_func = brw_translate_logic_op( brw->attribs.Blend->logicop_func );
+   }
+   else if (brw->attribs.Blend->blend_enable) {
+      int eqRGB = brw->attribs.Blend->rgb_func;
+      int eqA = brw->attribs.Blend->alpha_func;
+      int srcRGB = brw->attribs.Blend->rgb_src_factor;
+      int dstRGB = brw->attribs.Blend->rgb_dst_factor;
+      int srcA = brw->attribs.Blend->alpha_src_factor;
+      int dstA = brw->attribs.Blend->alpha_dst_factor;
+
+      if (eqRGB == PIPE_BLEND_MIN || eqRGB == PIPE_BLEND_MAX) {
+	 srcRGB = dstRGB = PIPE_BLENDFACTOR_ONE;
+      }
+
+      if (eqA == PIPE_BLEND_MIN || eqA == PIPE_BLEND_MAX) {
+	 srcA = dstA = PIPE_BLENDFACTOR_ONE;
+      }
+
+      cc.cc6.dest_blend_factor = brw_translate_blend_factor(dstRGB);
+      cc.cc6.src_blend_factor = brw_translate_blend_factor(srcRGB);
+      cc.cc6.blend_function = brw_translate_blend_equation( eqRGB );
+
+      cc.cc5.ia_dest_blend_factor = brw_translate_blend_factor(dstA);
+      cc.cc5.ia_src_blend_factor = brw_translate_blend_factor(srcA);
+      cc.cc5.ia_blend_function = brw_translate_blend_equation( eqA );
+
+      cc.cc3.blend_enable = 1;
+      cc.cc3.ia_blend_enable = (srcA != srcRGB ||
+				dstA != dstRGB ||
+				eqA != eqRGB);
+   }
+   
+   /* BRW_NEW_ALPHATEST
+    */
+   if (brw->attribs.DepthStencil->alpha.enabled) {
+      cc.cc3.alpha_test = 1;
+      cc.cc3.alpha_test_func = 
+	 brw_translate_compare_func(brw->attribs.DepthStencil->alpha.func);
+
+      cc.cc7.alpha_ref.ub[0] = float_to_ubyte(brw->attribs.DepthStencil->alpha.ref_value);
+
+      cc.cc3.alpha_test_format = BRW_ALPHATEST_FORMAT_UNORM8;
+   }
+
+   if (brw->attribs.Blend->dither) {
+      cc.cc5.dither_enable = 1;
+      cc.cc6.y_dither_offset = 0;
+      cc.cc6.x_dither_offset = 0;
+   }
+
+   if (brw->attribs.DepthStencil->depth.enabled) {
+      cc.cc2.depth_test = brw->attribs.DepthStencil->depth.enabled;
+      cc.cc2.depth_test_function = brw_translate_compare_func(brw->attribs.DepthStencil->depth.func);
+      cc.cc2.depth_write_enable = brw->attribs.DepthStencil->depth.writemask;
+   }
+
+   /* CACHE_NEW_CC_VP */
+   cc.cc4.cc_viewport_state_offset =  brw->cc.vp_gs_offset >> 5;
+
+   if (BRW_DEBUG & DEBUG_STATS)
+      cc.cc5.statistics_enable = 1;
+
+   brw->cc.state_gs_offset = brw_cache_data( &brw->cache[BRW_CC_UNIT], &cc );
+}
+
+const struct brw_tracked_state brw_cc_unit = {
+   .dirty = {
+      .brw = BRW_NEW_DEPTH_STENCIL | BRW_NEW_BLEND | BRW_NEW_ALPHA_TEST,
+      .cache = CACHE_NEW_CC_VP
+   },
+   .update = upload_cc_unit
+};
+
diff --git a/src/gallium/drivers/i965simple/brw_clip.c b/src/gallium/drivers/i965simple/brw_clip.c
new file mode 100644
index 0000000000..268124cc53
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_clip.c
@@ -0,0 +1,206 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_util.h"
+#include "brw_state.h"
+#include "brw_clip.h"
+
+#define FRONT_UNFILLED_BIT  0x1
+#define BACK_UNFILLED_BIT   0x2
+
+
+static void compile_clip_prog( struct brw_context *brw,
+			     struct brw_clip_prog_key *key )
+{
+   struct brw_clip_compile c;
+   const unsigned *program;
+   unsigned program_size;
+   unsigned delta;
+   unsigned i;
+
+   memset(&c, 0, sizeof(c));
+
+   /* Begin the compilation:
+    */
+   brw_init_compile(&c.func);
+
+   c.func.single_program_flow = 1;
+
+   c.key = *key;
+
+
+   /* Need to locate the two positions present in vertex + header.
+    * These are currently hardcoded:
+    */
+   c.header_position_offset = ATTR_SIZE;
+
+   for (i = 0, delta = REG_SIZE; i < PIPE_MAX_SHADER_OUTPUTS; i++)
+      if (c.key.attrs & (1<<i)) {
+	 c.offset[i] = delta;
+	 delta += ATTR_SIZE;
+      }
+
+   c.nr_attrs = brw_count_bits(c.key.attrs);
+   c.nr_regs = (c.nr_attrs + 1) / 2 + 1;  /* are vertices packed, or reg-aligned? */
+   c.nr_bytes = c.nr_regs * REG_SIZE;
+
+   c.prog_data.clip_mode = c.key.clip_mode; /* XXX */
+
+   /* For some reason the thread is spawned with only 4 channels
+    * unmasked.
+    */
+   brw_set_mask_control(&c.func, BRW_MASK_DISABLE);
+
+
+   /* Would ideally have the option of producing a program which could
+    * do all three:
+    */
+   switch (key->primitive) {
+   case PIPE_PRIM_TRIANGLES:
+#if 0
+      if (key->do_unfilled)
+	 brw_emit_unfilled_clip( &c );
+      else
+#endif
+	 brw_emit_tri_clip( &c );
+      break;
+   case PIPE_PRIM_LINES:
+      brw_emit_line_clip( &c );
+      break;
+   case PIPE_PRIM_POINTS:
+      brw_emit_point_clip( &c );
+      break;
+   default:
+      assert(0);
+      return;
+   }
+
+
+
+   /* get the program
+    */
+   program = brw_get_program(&c.func, &program_size);
+
+   /* Upload
+    */
+   brw->clip.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_CLIP_PROG],
+						&c.key,
+						sizeof(c.key),
+						program,
+						program_size,
+						&c.prog_data,
+						&brw->clip.prog_data );
+}
+
+
+static boolean search_cache( struct brw_context *brw,
+			       struct brw_clip_prog_key *key )
+{
+   return brw_search_cache(&brw->cache[BRW_CLIP_PROG],
+			   key, sizeof(*key),
+			   &brw->clip.prog_data,
+			   &brw->clip.prog_gs_offset);
+}
+
+
+
+
+/* Calculate interpolants for triangle and line rasterization.
+ */
+static void upload_clip_prog(struct brw_context *brw)
+{
+   struct brw_clip_prog_key key;
+
+   memset(&key, 0, sizeof(key));
+
+   /* Populate the key:
+    */
+   /* BRW_NEW_REDUCED_PRIMITIVE */
+   key.primitive = brw->reduced_primitive;
+   /* CACHE_NEW_VS_PROG */
+   key.attrs = brw->vs.prog_data->outputs_written;
+   /* BRW_NEW_RASTER */
+   key.do_flat_shading = (brw->attribs.Raster->flatshade);
+   /* BRW_NEW_CLIP */
+   key.nr_userclip = brw->attribs.Clip.nr; /* XXX */
+
+#if 0
+   key.clip_mode = BRW_CLIPMODE_NORMAL;
+
+   if (key.primitive == PIPE_PRIM_TRIANGLES) {
+      if (brw->attribs.Raster->cull_mode == PIPE_WINDING_BOTH)
+	 key.clip_mode = BRW_CLIPMODE_REJECT_ALL;
+      else {
+         if (brw->attribs.Raster->fill_cw != PIPE_POLYGON_MODE_FILL ||
+             brw->attribs.Raster->fill_ccw != PIPE_POLYGON_MODE_FILL)
+            key.do_unfilled = 1;
+
+	 /* Most cases the fixed function units will handle.  Cases where
+	  * one or more polygon faces are unfilled will require help:
+	  */
+	 if (key.do_unfilled) {
+	    key.clip_mode = BRW_CLIPMODE_CLIP_NON_REJECTED;
+
+	    if (brw->attribs.Raster->offset_cw ||
+                brw->attribs.Raster->offset_ccw) {
+	       key.offset_units = brw->attribs.Raster->offset_units;
+	       key.offset_factor = brw->attribs.Raster->offset_scale;
+	    }
+            key.fill_ccw = brw->attribs.Raster->fill_ccw;
+            key.fill_cw = brw->attribs.Raster->fill_cw;
+            key.offset_ccw = brw->attribs.Raster->offset_ccw;
+            key.offset_cw = brw->attribs.Raster->offset_cw;
+            if (brw->attribs.Raster->light_twoside &&
+                key.fill_cw != CLIP_CULL)
+               key.copy_bfc_cw = 1;
+	 }
+      }
+   }
+#else
+   key.clip_mode = BRW_CLIPMODE_ACCEPT_ALL;
+#endif
+
+   if (!search_cache(brw, &key))
+      compile_clip_prog( brw, &key );
+}
+
+const struct brw_tracked_state brw_clip_prog = {
+   .dirty = {
+      .brw   = (BRW_NEW_RASTERIZER |
+		BRW_NEW_CLIP |
+		BRW_NEW_REDUCED_PRIMITIVE),
+      .cache = CACHE_NEW_VS_PROG
+   },
+   .update = upload_clip_prog
+};
diff --git a/src/gallium/drivers/i965simple/brw_clip.h b/src/gallium/drivers/i965simple/brw_clip.h
new file mode 100644
index 0000000000..d70fc094ff
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_clip.h
@@ -0,0 +1,170 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#ifndef BRW_CLIP_H
+#define BRW_CLIP_H
+
+
+#include "brw_context.h"
+#include "brw_eu.h"
+
+#define MAX_VERTS (3+6+6)	
+
+/* Note that if unfilled primitives are being emitted, we have to fix
+ * up polygon offset and flatshading at this point:
+ */
+struct brw_clip_prog_key {
+   unsigned attrs:32;		
+   unsigned primitive:4;
+   unsigned nr_userclip:3;
+   unsigned do_flat_shading:1;
+   unsigned do_unfilled:1;
+   unsigned fill_cw:2;		/* includes cull information */
+   unsigned fill_ccw:2;		/* includes cull information */
+   unsigned offset_cw:1;
+   unsigned offset_ccw:1;
+   unsigned pad0:17;
+
+   unsigned copy_bfc_cw:1;
+   unsigned copy_bfc_ccw:1;
+   unsigned clip_mode:3;
+   unsigned pad1:27;
+   
+   float offset_factor;
+   float offset_units;
+};
+
+
+#define CLIP_LINE   0
+#define CLIP_POINT  1
+#define CLIP_FILL   2
+#define CLIP_CULL   3
+
+
+#define PRIM_MASK  (0x1f)
+
+struct brw_clip_compile {
+   struct brw_compile func;
+   struct brw_clip_prog_key key;
+   struct brw_clip_prog_data prog_data;
+   
+   struct {
+      struct brw_reg R0;
+      struct brw_reg vertex[MAX_VERTS];
+
+      struct brw_reg t;
+      struct brw_reg t0, t1;
+      struct brw_reg dp0, dp1;
+
+      struct brw_reg dpPrev;
+      struct brw_reg dp;
+      struct brw_reg loopcount;
+      struct brw_reg nr_verts;
+      struct brw_reg planemask;
+
+      struct brw_reg inlist;
+      struct brw_reg outlist;
+      struct brw_reg freelist;
+
+      struct brw_reg dir;
+      struct brw_reg tmp0, tmp1;
+      struct brw_reg offset;
+      
+      struct brw_reg fixed_planes;
+      struct brw_reg plane_equation;
+   } reg;
+
+   /* 3 different ways of expressing vertex size:
+    */
+   unsigned nr_attrs;
+   unsigned nr_regs;
+   unsigned nr_bytes;
+
+   unsigned first_tmp;
+   unsigned last_tmp;
+
+   boolean need_direction;
+
+   unsigned last_mrf;
+
+   unsigned header_position_offset;
+   unsigned offset[PIPE_MAX_ATTRIBS];
+};
+
+#define ATTR_SIZE  (4*4)
+
+/* Points are only culled, so no need for a clip routine, however it
+ * works out easier to have a dummy one.
+ */
+void brw_emit_unfilled_clip( struct brw_clip_compile *c );
+void brw_emit_tri_clip( struct brw_clip_compile *c );
+void brw_emit_line_clip( struct brw_clip_compile *c );
+void brw_emit_point_clip( struct brw_clip_compile *c );
+
+/* brw_clip_tri.c, for use by the unfilled clip routine:
+ */
+void brw_clip_tri_init_vertices( struct brw_clip_compile *c );
+void brw_clip_tri_flat_shade( struct brw_clip_compile *c );
+void brw_clip_tri( struct brw_clip_compile *c );
+void brw_clip_tri_emit_polygon( struct brw_clip_compile *c );
+void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, 
+			      unsigned nr_verts );
+
+
+/* Utils:
+ */
+
+void brw_clip_interp_vertex( struct brw_clip_compile *c,
+			     struct brw_indirect dest_ptr,
+			     struct brw_indirect v0_ptr, /* from */
+			     struct brw_indirect v1_ptr, /* to */
+			     struct brw_reg t0,
+			     boolean force_edgeflag );
+
+void brw_clip_init_planes( struct brw_clip_compile *c );
+
+void brw_clip_emit_vue(struct brw_clip_compile *c, 
+		       struct brw_indirect vert,
+		       boolean allocate,
+		       boolean eot,
+		       unsigned header);
+
+void brw_clip_kill_thread(struct brw_clip_compile *c);
+
+struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c );
+struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c );
+
+void brw_clip_copy_colors( struct brw_clip_compile *c,
+			   unsigned to, unsigned from );
+
+void brw_clip_init_clipmask( struct brw_clip_compile *c );
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_clip_line.c b/src/gallium/drivers/i965simple/brw_clip_line.c
new file mode 100644
index 0000000000..75d9e5fcda
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_clip_line.c
@@ -0,0 +1,245 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_util.h"
+#include "brw_clip.h"
+
+
+
+static void brw_clip_line_alloc_regs( struct brw_clip_compile *c )
+{
+   unsigned i = 0,j;
+
+   /* Register usage is static, precompute here:
+    */
+   c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++;
+
+   if (c->key.nr_userclip) {
+      c->reg.fixed_planes = brw_vec4_grf(i, 0);
+      i += (6 + c->key.nr_userclip + 1) / 2;
+
+      c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2;
+   }
+   else
+      c->prog_data.curb_read_length = 0;
+
+
+   /* Payload vertices plus space for more generated vertices:
+    */
+   for (j = 0; j < 4; j++) {
+      c->reg.vertex[j] = brw_vec4_grf(i, 0);
+      i += c->nr_regs;
+   }
+
+   c->reg.t           = brw_vec1_grf(i, 0);
+   c->reg.t0          = brw_vec1_grf(i, 1);
+   c->reg.t1          = brw_vec1_grf(i, 2);
+   c->reg.planemask   = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD);
+   c->reg.plane_equation = brw_vec4_grf(i, 4);
+   i++;
+
+   c->reg.dp0         = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */
+   c->reg.dp1         = brw_vec1_grf(i, 4);
+   i++;
+
+   if (!c->key.nr_userclip) {
+      c->reg.fixed_planes = brw_vec8_grf(i, 0);
+      i++;
+   }
+
+
+   c->first_tmp = i;
+   c->last_tmp = i;
+
+   c->prog_data.urb_read_length = c->nr_regs; /* ? */
+   c->prog_data.total_grf = i;
+}
+
+
+
+/* Line clipping, more or less following the following algorithm:
+ *
+ *  for (p=0;p<MAX_PLANES;p++) {
+ *     if (clipmask & (1 << p)) {
+ *        float dp0 = DOTPROD( vtx0, plane[p] );
+ *        float dp1 = DOTPROD( vtx1, plane[p] );
+ *
+ *        if (IS_NEGATIVE(dp1)) {
+ *           float t = dp1 / (dp1 - dp0);
+ *           if (t > t1) t1 = t;
+ *        } else {
+ *           float t = dp0 / (dp0 - dp1);
+ *           if (t > t0) t0 = t;
+ *        }
+ *
+ *        if (t0 + t1 >= 1.0)
+ *           return;
+ *     }
+ *  }
+ *
+ *  interp( ctx, newvtx0, vtx0, vtx1, t0 );
+ *  interp( ctx, newvtx1, vtx1, vtx0, t1 );
+ *
+ */
+static void clip_and_emit_line( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_indirect vtx0     = brw_indirect(0, 0);
+   struct brw_indirect vtx1      = brw_indirect(1, 0);
+   struct brw_indirect newvtx0   = brw_indirect(2, 0);
+   struct brw_indirect newvtx1   = brw_indirect(3, 0);
+   struct brw_indirect plane_ptr = brw_indirect(4, 0);
+   struct brw_instruction *plane_loop;
+   struct brw_instruction *plane_active;
+   struct brw_instruction *is_negative;
+   struct brw_instruction *is_neg2;
+   struct brw_instruction *not_culled;
+   struct brw_reg v1_null_ud = retype(vec1(brw_null_reg()), BRW_REGISTER_TYPE_UD);
+
+   brw_MOV(p, get_addr_reg(vtx0),      brw_address(c->reg.vertex[0]));
+   brw_MOV(p, get_addr_reg(vtx1),      brw_address(c->reg.vertex[1]));
+   brw_MOV(p, get_addr_reg(newvtx0),   brw_address(c->reg.vertex[2]));
+   brw_MOV(p, get_addr_reg(newvtx1),   brw_address(c->reg.vertex[3]));
+   brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c));
+
+   /* Note: init t0, t1 together:
+    */
+   brw_MOV(p, vec2(c->reg.t0), brw_imm_f(0));
+
+   brw_clip_init_planes(c);
+   brw_clip_init_clipmask(c);
+
+   /* -ve rhw workaround */
+   brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+   brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2),
+	   brw_imm_ud(1<<20));
+   brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(0x3f));
+   brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+
+   plane_loop = brw_DO(p, BRW_EXECUTE_1);
+   {
+      /* if (planemask & 1)
+       */
+      brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+      brw_AND(p, v1_null_ud, c->reg.planemask, brw_imm_ud(1));
+
+      plane_active = brw_IF(p, BRW_EXECUTE_1);
+      {
+	 if (c->key.nr_userclip)
+	    brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0));
+	 else
+	    brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0));
+
+#if 0
+	 /* dp = DP4(vtx->position, plane)
+	  */
+	 brw_DP4(p, vec4(c->reg.dp0), deref_4f(vtx0, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation);
+
+	 /* if (IS_NEGATIVE(dp1))
+	  */
+	 brw_set_conditionalmod(p, BRW_CONDITIONAL_L);
+	 brw_DP4(p, vec4(c->reg.dp1), deref_4f(vtx1, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation);
+#else
+         #warning "disabled"
+#endif
+	 is_negative = brw_IF(p, BRW_EXECUTE_1);
+	 {
+	    brw_ADD(p, c->reg.t, c->reg.dp1, negate(c->reg.dp0));
+	    brw_math_invert(p, c->reg.t, c->reg.t);
+	    brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp1);
+
+	    brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t1 );
+	    brw_MOV(p, c->reg.t1, c->reg.t);
+	    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+	 }
+	 is_negative = brw_ELSE(p, is_negative);
+	 {
+	    /* Coming back in.  We know that both cannot be negative
+	     * because the line would have been culled in that case.
+	     */
+
+	    /* If both are positive, do nothing */
+             brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.dp0, brw_imm_f(0.0));
+             is_neg2 = brw_IF(p, BRW_EXECUTE_1);
+             {
+		brw_ADD(p, c->reg.t, c->reg.dp0, negate(c->reg.dp1));
+		brw_math_invert(p, c->reg.t, c->reg.t);
+		brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp0);
+
+		brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t0 );
+		brw_MOV(p, c->reg.t0, c->reg.t);
+		brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+	     }
+	     brw_ENDIF(p, is_neg2);
+	 }
+	 brw_ENDIF(p, is_negative);
+      }
+      brw_ENDIF(p, plane_active);
+
+      /* plane_ptr++;
+       */
+      brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c));
+
+      /* while (planemask>>=1) != 0
+       */
+      brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+      brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1));
+   }
+   brw_WHILE(p, plane_loop);
+
+   brw_ADD(p, c->reg.t, c->reg.t0, c->reg.t1);
+   brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.t, brw_imm_f(1.0));
+   not_culled = brw_IF(p, BRW_EXECUTE_1);
+   {
+      brw_clip_interp_vertex(c, newvtx0, vtx0, vtx1, c->reg.t0, FALSE);
+      brw_clip_interp_vertex(c, newvtx1, vtx1, vtx0, c->reg.t1, FALSE);
+
+      brw_clip_emit_vue(c, newvtx0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START);
+      brw_clip_emit_vue(c, newvtx1, 0, 1, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END);
+   }
+   brw_ENDIF(p, not_culled);
+   brw_clip_kill_thread(c);
+}
+
+
+
+void brw_emit_line_clip( struct brw_clip_compile *c )
+{
+   brw_clip_line_alloc_regs(c);
+
+   if (c->key.do_flat_shading)
+      brw_clip_copy_colors(c, 0, 1);
+
+   clip_and_emit_line(c);
+}
diff --git a/src/gallium/drivers/i965simple/brw_clip_point.c b/src/gallium/drivers/i965simple/brw_clip_point.c
new file mode 100644
index 0000000000..6fce7210d1
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_clip_point.c
@@ -0,0 +1,47 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_util.h"
+#include "brw_clip.h"
+
+
+/* Point clipping, nothing to do?
+ */
+void brw_emit_point_clip( struct brw_clip_compile *c )
+{
+   /* Send an empty message to kill the thread:
+    */
+   brw_clip_tri_alloc_regs(c, 0);
+   brw_clip_kill_thread(c);
+}
diff --git a/src/gallium/drivers/i965simple/brw_clip_state.c b/src/gallium/drivers/i965simple/brw_clip_state.c
new file mode 100644
index 0000000000..8e78dd51be
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_clip_state.c
@@ -0,0 +1,93 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+
+static void upload_clip_unit( struct brw_context *brw )
+{
+   struct brw_clip_unit_state clip;
+
+   memset(&clip, 0, sizeof(clip));
+
+   /* CACHE_NEW_CLIP_PROG */
+   clip.thread0.grf_reg_count =
+      align(brw->clip.prog_data->total_grf, 16) / 16 - 1;
+   clip.thread0.kernel_start_pointer = brw->clip.prog_gs_offset >> 6;
+   clip.thread3.urb_entry_read_length = brw->clip.prog_data->urb_read_length;
+   clip.thread3.const_urb_entry_read_length = brw->clip.prog_data->curb_read_length;
+   clip.clip5.clip_mode = brw->clip.prog_data->clip_mode;
+
+   /* BRW_NEW_CURBE_OFFSETS */
+   clip.thread3.const_urb_entry_read_offset = brw->curbe.clip_start * 2;
+
+   /* BRW_NEW_URB_FENCE */
+   clip.thread4.nr_urb_entries = brw->urb.nr_clip_entries; 
+   clip.thread4.urb_entry_allocation_size = brw->urb.vsize - 1;
+   clip.thread4.max_threads = 1; /* 2 threads */
+
+   if (BRW_DEBUG & DEBUG_STATS)
+      clip.thread4.stats_enable = 1; 
+
+   /* CONSTANT */
+   clip.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
+   clip.thread1.single_program_flow = 1;
+   clip.thread3.dispatch_grf_start_reg = 1;
+   clip.thread3.urb_entry_read_offset = 0;
+   clip.clip5.userclip_enable_flags = 0x7f;
+   clip.clip5.userclip_must_clip = 1;
+   clip.clip5.guard_band_enable = 0;
+   clip.clip5.viewport_z_clip_enable = 1;
+   clip.clip5.viewport_xy_clip_enable = 1;
+   clip.clip5.vertex_position_space = BRW_CLIP_NDCSPACE;
+   clip.clip5.api_mode = BRW_CLIP_API_OGL;   
+   clip.clip6.clipper_viewport_state_ptr = 0;
+   clip.viewport_xmin = -1;
+   clip.viewport_xmax = 1;
+   clip.viewport_ymin = -1;
+   clip.viewport_ymax = 1;
+
+   brw->clip.state_gs_offset = brw_cache_data( &brw->cache[BRW_CLIP_UNIT], &clip );
+}
+
+
+const struct brw_tracked_state brw_clip_unit = {
+   .dirty = {
+      .brw   = (BRW_NEW_CURBE_OFFSETS |
+		BRW_NEW_URB_FENCE),
+      .cache = CACHE_NEW_CLIP_PROG
+   },
+   .update = upload_clip_unit
+};
diff --git a/src/gallium/drivers/i965simple/brw_clip_tri.c b/src/gallium/drivers/i965simple/brw_clip_tri.c
new file mode 100644
index 0000000000..c5da7b825e
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_clip_tri.c
@@ -0,0 +1,566 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_util.h"
+#include "brw_clip.h"
+
+static struct brw_reg get_tmp( struct brw_clip_compile *c )
+{
+   struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0);
+
+   if (++c->last_tmp > c->prog_data.total_grf)
+      c->prog_data.total_grf = c->last_tmp;
+
+   return tmp;
+}
+
+static void release_tmps( struct brw_clip_compile *c )
+{
+   c->last_tmp = c->first_tmp;
+}
+
+
+void brw_clip_tri_alloc_regs( struct brw_clip_compile *c,
+			      unsigned nr_verts )
+{
+   unsigned i = 0,j;
+
+   /* Register usage is static, precompute here:
+    */
+   c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++;
+
+   if (c->key.nr_userclip) {
+      c->reg.fixed_planes = brw_vec4_grf(i, 0);
+      i += (6 + c->key.nr_userclip + 1) / 2;
+
+      c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2;
+   }
+   else
+      c->prog_data.curb_read_length = 0;
+
+
+   /* Payload vertices plus space for more generated vertices:
+    */
+   for (j = 0; j < nr_verts; j++) {
+      c->reg.vertex[j] = brw_vec4_grf(i, 0);
+      i += c->nr_regs;
+   }
+
+   if (c->nr_attrs & 1) {
+      for (j = 0; j < 3; j++) {
+	 unsigned delta = c->nr_attrs*16 + 32;
+	 brw_MOV(&c->func, byte_offset(c->reg.vertex[j], delta), brw_imm_f(0));
+      }
+   }
+
+   c->reg.t          = brw_vec1_grf(i, 0);
+   c->reg.loopcount  = retype(brw_vec1_grf(i, 1), BRW_REGISTER_TYPE_UD);
+   c->reg.nr_verts   = retype(brw_vec1_grf(i, 2), BRW_REGISTER_TYPE_UD);
+   c->reg.planemask  = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD);
+   c->reg.plane_equation = brw_vec4_grf(i, 4);
+   i++;
+
+   c->reg.dpPrev     = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */
+   c->reg.dp         = brw_vec1_grf(i, 4);
+   i++;
+
+   c->reg.inlist     = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0);
+   i++;
+
+   c->reg.outlist    = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0);
+   i++;
+
+   c->reg.freelist   = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0);
+   i++;
+
+   if (!c->key.nr_userclip) {
+      c->reg.fixed_planes = brw_vec8_grf(i, 0);
+      i++;
+   }
+
+   if (c->key.do_unfilled) {
+      c->reg.dir     = brw_vec4_grf(i, 0);
+      c->reg.offset  = brw_vec4_grf(i, 4);
+      i++;
+      c->reg.tmp0    = brw_vec4_grf(i, 0);
+      c->reg.tmp1    = brw_vec4_grf(i, 4);
+      i++;
+   }
+
+   c->first_tmp = i;
+   c->last_tmp = i;
+
+   c->prog_data.urb_read_length = c->nr_regs; /* ? */
+   c->prog_data.total_grf = i;
+}
+
+
+
+void brw_clip_tri_init_vertices( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */
+   struct brw_instruction *is_rev;
+
+   /* Initial list of indices for incoming vertexes:
+    */
+   brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK));
+   brw_CMP(p,
+	   vec1(brw_null_reg()),
+	   BRW_CONDITIONAL_EQ,
+	   tmp0,
+	   brw_imm_ud(_3DPRIM_TRISTRIP_REVERSE));
+
+   /* XXX: Is there an easier way to do this?  Need to reverse every
+    * second tristrip element:  Can ignore sometimes?
+    */
+   is_rev = brw_IF(p, BRW_EXECUTE_1);
+   {
+      brw_MOV(p, get_element(c->reg.inlist, 0),  brw_address(c->reg.vertex[1]) );
+      brw_MOV(p, get_element(c->reg.inlist, 1),  brw_address(c->reg.vertex[0]) );
+      if (c->need_direction)
+	 brw_MOV(p, c->reg.dir, brw_imm_f(-1));
+   }
+   is_rev = brw_ELSE(p, is_rev);
+   {
+      brw_MOV(p, get_element(c->reg.inlist, 0),  brw_address(c->reg.vertex[0]) );
+      brw_MOV(p, get_element(c->reg.inlist, 1),  brw_address(c->reg.vertex[1]) );
+      if (c->need_direction)
+	 brw_MOV(p, c->reg.dir, brw_imm_f(1));
+   }
+   brw_ENDIF(p, is_rev);
+
+   brw_MOV(p, get_element(c->reg.inlist, 2),  brw_address(c->reg.vertex[2]) );
+   brw_MOV(p, brw_vec8_grf(c->reg.outlist.nr, 0), brw_imm_f(0));
+   brw_MOV(p, c->reg.nr_verts, brw_imm_ud(3));
+}
+
+
+
+void brw_clip_tri_flat_shade( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *is_poly;
+   struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */
+
+   brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK));
+   brw_CMP(p,
+	   vec1(brw_null_reg()),
+	   BRW_CONDITIONAL_EQ,
+	   tmp0,
+	   brw_imm_ud(_3DPRIM_POLYGON));
+
+   is_poly = brw_IF(p, BRW_EXECUTE_1);
+   {
+      brw_clip_copy_colors(c, 1, 0);
+      brw_clip_copy_colors(c, 2, 0);
+   }
+   is_poly = brw_ELSE(p, is_poly);
+   {
+      brw_clip_copy_colors(c, 0, 2);
+      brw_clip_copy_colors(c, 1, 2);
+   }
+   brw_ENDIF(p, is_poly);
+}
+
+
+
+/* Use mesa's clipping algorithms, translated to GEN4 assembly.
+ */
+void brw_clip_tri( struct brw_clip_compile *c )
+{
+#if 0
+   struct brw_compile *p = &c->func;
+   struct brw_indirect vtx = brw_indirect(0, 0);
+   struct brw_indirect vtxPrev = brw_indirect(1, 0);
+   struct brw_indirect vtxOut = brw_indirect(2, 0);
+   struct brw_indirect plane_ptr = brw_indirect(3, 0);
+   struct brw_indirect inlist_ptr = brw_indirect(4, 0);
+   struct brw_indirect outlist_ptr = brw_indirect(5, 0);
+   struct brw_indirect freelist_ptr = brw_indirect(6, 0);
+   struct brw_instruction *plane_loop;
+   struct brw_instruction *plane_active;
+   struct brw_instruction *vertex_loop;
+   struct brw_instruction *next_test;
+   struct brw_instruction *prev_test;
+
+   brw_MOV(p, get_addr_reg(vtxPrev),     brw_address(c->reg.vertex[2]) );
+   brw_MOV(p, get_addr_reg(plane_ptr),   brw_clip_plane0_address(c));
+   brw_MOV(p, get_addr_reg(inlist_ptr),  brw_address(c->reg.inlist));
+   brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist));
+
+   brw_MOV(p, get_addr_reg(freelist_ptr), brw_address(c->reg.vertex[3]) );
+
+   plane_loop = brw_DO(p, BRW_EXECUTE_1);
+   {
+      /* if (planemask & 1)
+       */
+      brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+      brw_AND(p, vec1(brw_null_reg()), c->reg.planemask, brw_imm_ud(1));
+
+      plane_active = brw_IF(p, BRW_EXECUTE_1);
+      {
+	 /* vtxOut = freelist_ptr++
+	  */
+	 brw_MOV(p, get_addr_reg(vtxOut),       get_addr_reg(freelist_ptr) );
+	 brw_ADD(p, get_addr_reg(freelist_ptr), get_addr_reg(freelist_ptr), brw_imm_uw(c->nr_regs * REG_SIZE));
+
+	 if (c->key.nr_userclip)
+	    brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0));
+	 else
+	    brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0));
+
+	 brw_MOV(p, c->reg.loopcount, c->reg.nr_verts);
+	 brw_MOV(p, c->reg.nr_verts, brw_imm_ud(0));
+
+	 vertex_loop = brw_DO(p, BRW_EXECUTE_1);
+	 {
+	    /* vtx = *input_ptr;
+	     */
+	    brw_MOV(p, get_addr_reg(vtx), deref_1uw(inlist_ptr, 0));
+
+	    /* IS_NEGATIVE(prev) */
+	    brw_set_conditionalmod(p, BRW_CONDITIONAL_L);
+	    brw_DP4(p, vec4(c->reg.dpPrev), deref_4f(vtxPrev, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation);
+	    prev_test = brw_IF(p, BRW_EXECUTE_1);
+	    {
+	       /* IS_POSITIVE(next)
+		*/
+	       brw_set_conditionalmod(p, BRW_CONDITIONAL_GE);
+	       brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation);
+	       next_test = brw_IF(p, BRW_EXECUTE_1);
+	       {
+
+		  /* Coming back in.
+		   */
+		  brw_ADD(p, c->reg.t, c->reg.dpPrev, negate(c->reg.dp));
+		  brw_math_invert(p, c->reg.t, c->reg.t);
+		  brw_MUL(p, c->reg.t, c->reg.t, c->reg.dpPrev);
+
+		  /* If (vtxOut == 0) vtxOut = vtxPrev
+		   */
+		  brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_EQ, get_addr_reg(vtxOut), brw_imm_uw(0) );
+		  brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtxPrev) );
+		  brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+
+		  brw_clip_interp_vertex(c, vtxOut, vtxPrev, vtx, c->reg.t, FALSE);
+
+		  /* *outlist_ptr++ = vtxOut;
+		   * nr_verts++;
+		   * vtxOut = 0;
+		   */
+		  brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxOut));
+		  brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short)));
+		  brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1));
+		  brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) );
+	       }
+	       brw_ENDIF(p, next_test);
+
+	    }
+	    prev_test = brw_ELSE(p, prev_test);
+	    {
+	       /* *outlist_ptr++ = vtxPrev;
+		* nr_verts++;
+		*/
+	       brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxPrev));
+	       brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short)));
+	       brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1));
+
+	       /* IS_NEGATIVE(next)
+		*/
+	       brw_set_conditionalmod(p, BRW_CONDITIONAL_L);
+	       brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation);
+	       next_test = brw_IF(p, BRW_EXECUTE_1);
+	       {
+		  /* Going out of bounds.  Avoid division by zero as we
+		   * know dp != dpPrev from DIFFERENT_SIGNS, above.
+		   */
+		  brw_ADD(p, c->reg.t, c->reg.dp, negate(c->reg.dpPrev));
+		  brw_math_invert(p, c->reg.t, c->reg.t);
+		  brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp);
+
+		  /* If (vtxOut == 0) vtxOut = vtx
+		   */
+		  brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_EQ, get_addr_reg(vtxOut), brw_imm_uw(0) );
+		  brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtx) );
+		  brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+
+		  brw_clip_interp_vertex(c, vtxOut, vtx, vtxPrev, c->reg.t, TRUE);
+
+		  /* *outlist_ptr++ = vtxOut;
+		   * nr_verts++;
+		   * vtxOut = 0;
+		   */
+		  brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxOut));
+		  brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short)));
+		  brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1));
+		  brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) );
+	       }
+	       brw_ENDIF(p, next_test);
+	    }
+	    brw_ENDIF(p, prev_test);
+
+	    /* vtxPrev = vtx;
+	     * inlist_ptr++;
+	     */
+	    brw_MOV(p, get_addr_reg(vtxPrev), get_addr_reg(vtx));
+	    brw_ADD(p, get_addr_reg(inlist_ptr), get_addr_reg(inlist_ptr), brw_imm_uw(sizeof(short)));
+
+	    /* while (--loopcount != 0)
+	     */
+	    brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+	    brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1));
+	 }
+	 brw_WHILE(p, vertex_loop);
+
+	 /* vtxPrev = *(outlist_ptr-1)  OR: outlist[nr_verts-1]
+	  * inlist = outlist
+	  * inlist_ptr = &inlist[0]
+	  * outlist_ptr = &outlist[0]
+	  */
+	 brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_w(-2));
+	 brw_MOV(p, get_addr_reg(vtxPrev), deref_1uw(outlist_ptr, 0));
+	 brw_MOV(p, brw_vec8_grf(c->reg.inlist.nr, 0), brw_vec8_grf(c->reg.outlist.nr, 0));
+	 brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist));
+	 brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist));
+      }
+      brw_ENDIF(p, plane_active);
+
+      /* plane_ptr++;
+       */
+      brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c));
+
+      /* nr_verts >= 3
+       */
+      brw_CMP(p,
+	      vec1(brw_null_reg()),
+	      BRW_CONDITIONAL_GE,
+	      c->reg.nr_verts,
+	      brw_imm_ud(3));
+
+      /* && (planemask>>=1) != 0
+       */
+      brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+      brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1));
+   }
+   brw_WHILE(p, plane_loop);
+#else
+         #warning "disabled"
+#endif
+}
+
+
+
+void brw_clip_tri_emit_polygon(struct brw_clip_compile *c)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *loop, *if_insn;
+
+   /* for (loopcount = nr_verts-2; loopcount > 0; loopcount--)
+    */
+   brw_set_conditionalmod(p, BRW_CONDITIONAL_G);
+   brw_ADD(p,
+	   c->reg.loopcount,
+	   c->reg.nr_verts,
+	   brw_imm_d(-2));
+
+   if_insn = brw_IF(p, BRW_EXECUTE_1);
+   {
+      struct brw_indirect v0 = brw_indirect(0, 0);
+      struct brw_indirect vptr = brw_indirect(1, 0);
+
+      brw_MOV(p, get_addr_reg(vptr), brw_address(c->reg.inlist));
+      brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0));
+
+      brw_clip_emit_vue(c, v0, 1, 0, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_START));
+
+      brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2));
+      brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0));
+
+      loop = brw_DO(p, BRW_EXECUTE_1);
+      {
+	 brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_TRIFAN << 2));
+
+	 brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2));
+	 brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0));
+
+	 brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+	 brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1));
+      }
+      brw_WHILE(p, loop);
+
+      brw_clip_emit_vue(c, v0, 0, 1, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_END));
+   }
+   brw_ENDIF(p, if_insn);
+}
+
+static void do_clip_tri( struct brw_clip_compile *c )
+{
+   brw_clip_init_planes(c);
+
+   brw_clip_tri(c);
+}
+
+
+static void maybe_do_clip_tri( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *do_clip;
+
+   brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0));
+   do_clip = brw_IF(p, BRW_EXECUTE_1);
+   {
+      do_clip_tri(c);
+   }
+   brw_ENDIF(p, do_clip);
+}
+
+static void brw_clip_test( struct brw_clip_compile *c )
+{
+#if 0
+    struct brw_reg t = retype(get_tmp(c), BRW_REGISTER_TYPE_UD);
+    struct brw_reg t1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD);
+    struct brw_reg t2 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD);
+    struct brw_reg t3 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD);
+
+    struct brw_reg v0 = get_tmp(c);
+    struct brw_reg v1 = get_tmp(c);
+    struct brw_reg v2 = get_tmp(c);
+
+    struct brw_indirect vt0 = brw_indirect(0, 0);
+    struct brw_indirect vt1 = brw_indirect(1, 0);
+    struct brw_indirect vt2 = brw_indirect(2, 0);
+
+    struct brw_compile *p = &c->func;
+
+    brw_MOV(p, get_addr_reg(vt0), brw_address(c->reg.vertex[0]));
+    brw_MOV(p, get_addr_reg(vt1), brw_address(c->reg.vertex[1]));
+    brw_MOV(p, get_addr_reg(vt2), brw_address(c->reg.vertex[2]));
+    brw_MOV(p, v0, deref_4f(vt0, c->offset[VERT_RESULT_HPOS]));
+    brw_MOV(p, v1, deref_4f(vt1, c->offset[VERT_RESULT_HPOS]));
+    brw_MOV(p, v2, deref_4f(vt2, c->offset[VERT_RESULT_HPOS]));
+
+    /* test nearz, xmin, ymin plane */
+    brw_CMP(p, t1, BRW_CONDITIONAL_LE, negate(v0), get_element(v0, 3));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    brw_CMP(p, t2, BRW_CONDITIONAL_LE, negate(v1), get_element(v1, 3));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    brw_CMP(p, t3, BRW_CONDITIONAL_LE, negate(v2), get_element(v2, 3));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    brw_XOR(p, t, t1, t2);
+    brw_XOR(p, t1, t2, t3);
+    brw_OR(p, t, t, t1);
+
+    brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ,
+	    get_element(t, 0), brw_imm_ud(0));
+    brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<5)));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ,
+	    get_element(t, 1), brw_imm_ud(0));
+    brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<3)));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ,
+	    get_element(t, 2), brw_imm_ud(0));
+    brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<1)));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+
+    /* test farz, xmax, ymax plane */
+    brw_CMP(p, t1, BRW_CONDITIONAL_L, v0, get_element(v0, 3));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    brw_CMP(p, t2, BRW_CONDITIONAL_L, v1, get_element(v1, 3));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    brw_CMP(p, t3, BRW_CONDITIONAL_L, v2, get_element(v2, 3));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+
+    brw_XOR(p, t, t1, t2);
+    brw_XOR(p, t1, t2, t3);
+    brw_OR(p, t, t, t1);
+
+    brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ,
+	    get_element(t, 0), brw_imm_ud(0));
+    brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<4)));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ,
+	    get_element(t, 1), brw_imm_ud(0));
+    brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<2)));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+    brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ,
+	    get_element(t, 2), brw_imm_ud(0));
+    brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<0)));
+    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+
+    release_tmps(c);
+#else
+         #warning "disabled"
+#endif
+}
+
+
+void brw_emit_tri_clip( struct brw_clip_compile *c )
+{
+   struct brw_instruction *neg_rhw;
+   struct brw_compile *p = &c->func;
+   brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6);
+   brw_clip_tri_init_vertices(c);
+   brw_clip_init_clipmask(c);
+
+   /* if -ve rhw workaround bit is set,
+      do cliptest */
+   brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+   brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2),
+	   brw_imm_ud(1<<20));
+   neg_rhw = brw_IF(p, BRW_EXECUTE_1);
+   {
+       brw_clip_test(c);
+   }
+   brw_ENDIF(p, neg_rhw);
+
+   /* Can't push into do_clip_tri because with polygon (or quad)
+    * flatshading, need to apply the flatshade here because we don't
+    * respect the PV when converting to trifan for emit:
+    */
+   if (c->key.do_flat_shading)
+      brw_clip_tri_flat_shade(c);
+
+   if (c->key.clip_mode == BRW_CLIPMODE_NORMAL)
+      do_clip_tri(c);
+   else
+      maybe_do_clip_tri(c);
+
+   brw_clip_tri_emit_polygon(c);
+
+   /* Send an empty message to kill the thread:
+    */
+   brw_clip_kill_thread(c);
+}
diff --git a/src/gallium/drivers/i965simple/brw_clip_unfilled.c b/src/gallium/drivers/i965simple/brw_clip_unfilled.c
new file mode 100644
index 0000000000..b774a76dd6
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_clip_unfilled.c
@@ -0,0 +1,477 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_util.h"
+#include "brw_clip.h"
+
+
+
+/* This is performed against the original triangles, so no indirection
+ * required:
+BZZZT!
+ */
+static void compute_tri_direction( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg e = c->reg.tmp0;
+   struct brw_reg f = c->reg.tmp1;
+   struct brw_reg v0 = byte_offset(c->reg.vertex[0], c->offset[VERT_RESULT_HPOS]); 
+   struct brw_reg v1 = byte_offset(c->reg.vertex[1], c->offset[VERT_RESULT_HPOS]); 
+   struct brw_reg v2 = byte_offset(c->reg.vertex[2], c->offset[VERT_RESULT_HPOS]); 
+
+
+   /* Calculate the vectors of two edges of the triangle:
+    */
+   brw_ADD(p, e, v0, negate(v2)); 
+   brw_ADD(p, f, v1, negate(v2)); 
+
+   /* Take their crossproduct:
+    */
+   brw_set_access_mode(p, BRW_ALIGN_16);
+   brw_MUL(p, vec4(brw_null_reg()), brw_swizzle(e, 1,2,0,3),  brw_swizzle(f,2,0,1,3));
+   brw_MAC(p, vec4(e),  negate(brw_swizzle(e, 2,0,1,3)), brw_swizzle(f,1,2,0,3));
+   brw_set_access_mode(p, BRW_ALIGN_1);
+
+   brw_MUL(p, c->reg.dir, c->reg.dir, vec4(e));
+}
+
+
+static void cull_direction( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *ccw;
+   unsigned conditional;
+
+   assert (!(c->key.fill_ccw == CLIP_CULL &&
+	     c->key.fill_cw == CLIP_CULL));
+
+   if (c->key.fill_ccw == CLIP_CULL)
+      conditional = BRW_CONDITIONAL_GE;
+   else
+      conditional = BRW_CONDITIONAL_L;
+
+   brw_CMP(p,
+	   vec1(brw_null_reg()),
+	   conditional,
+	   get_element(c->reg.dir, 2),
+	   brw_imm_f(0));
+   
+   ccw = brw_IF(p, BRW_EXECUTE_1);
+   {
+      brw_clip_kill_thread(c);
+   }
+   brw_ENDIF(p, ccw);
+}
+
+
+
+static void copy_bfc( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *ccw;
+   unsigned conditional;
+
+   /* Do we have any colors to copy? 
+    */
+   if (!(c->offset[VERT_RESULT_COL0] && c->offset[VERT_RESULT_BFC0]) &&
+       !(c->offset[VERT_RESULT_COL1] && c->offset[VERT_RESULT_BFC1]))
+      return;
+
+   /* In some wierd degnerate cases we can end up testing the
+    * direction twice, once for culling and once for bfc copying.  Oh
+    * well, that's what you get for setting wierd GL state.
+    */
+   if (c->key.copy_bfc_ccw)
+      conditional = BRW_CONDITIONAL_GE;
+   else
+      conditional = BRW_CONDITIONAL_L;
+
+   brw_CMP(p,
+	   vec1(brw_null_reg()),
+	   conditional,
+	   get_element(c->reg.dir, 2),
+	   brw_imm_f(0));
+   
+   ccw = brw_IF(p, BRW_EXECUTE_1);
+   {
+      unsigned i;
+
+      for (i = 0; i < 3; i++) {
+	 if (c->offset[VERT_RESULT_COL0] && c->offset[VERT_RESULT_BFC0])
+	    brw_MOV(p, 
+		    byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_COL0]),
+		    byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_BFC0]));
+
+	 if (c->offset[VERT_RESULT_COL1] && c->offset[VERT_RESULT_BFC1])
+	    brw_MOV(p, 
+		    byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_COL1]),
+		    byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_BFC1]));
+      }
+   }
+   brw_ENDIF(p, ccw);
+}
+
+
+
+
+/*
+  float iz	= 1.0 / dir.z;
+  float ac	= dir.x * iz;
+  float bc	= dir.y * iz;
+  offset = ctx->Polygon.OffsetUnits * DEPTH_SCALE;
+  offset += MAX2( abs(ac), abs(bc) ) * ctx->Polygon.OffsetFactor;
+  offset *= MRD;
+*/
+static void compute_offset( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg off = c->reg.offset;
+   struct brw_reg dir = c->reg.dir;
+   
+   brw_math_invert(p, get_element(off, 2), get_element(dir, 2));
+   brw_MUL(p, vec2(off), dir, get_element(off, 2));
+
+   brw_CMP(p, 
+	   vec1(brw_null_reg()), 
+	   BRW_CONDITIONAL_GE,
+	   brw_abs(get_element(off, 0)), 
+	   brw_abs(get_element(off, 1)));
+
+   brw_SEL(p, vec1(off), brw_abs(get_element(off, 0)), brw_abs(get_element(off, 1)));
+   brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+
+   brw_MUL(p, vec1(off), off, brw_imm_f(c->key.offset_factor));
+   brw_ADD(p, vec1(off), off, brw_imm_f(c->key.offset_units));
+}
+
+
+static void merge_edgeflags( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *is_poly;
+   struct brw_reg tmp0 = get_element_ud(c->reg.tmp0, 0);
+
+   brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); 
+   brw_CMP(p, 
+	   vec1(brw_null_reg()), 
+	   BRW_CONDITIONAL_EQ, 
+	   tmp0,
+	   brw_imm_ud(_3DPRIM_POLYGON));
+
+   /* Get away with using reg.vertex because we know that this is not
+    * a _3DPRIM_TRISTRIP_REVERSE:
+    */
+   is_poly = brw_IF(p, BRW_EXECUTE_1);
+   {   
+      brw_set_conditionalmod(p, BRW_CONDITIONAL_EQ);
+      brw_AND(p, vec1(brw_null_reg()), get_element_ud(c->reg.R0, 2), brw_imm_ud(1<<8));
+      brw_MOV(p, byte_offset(c->reg.vertex[0], c->offset[VERT_RESULT_EDGE]), brw_imm_f(0));
+      brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+
+      brw_set_conditionalmod(p, BRW_CONDITIONAL_EQ);
+      brw_AND(p, vec1(brw_null_reg()), get_element_ud(c->reg.R0, 2), brw_imm_ud(1<<9));
+      brw_MOV(p, byte_offset(c->reg.vertex[2], c->offset[VERT_RESULT_EDGE]), brw_imm_f(0));
+      brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+   }
+   brw_ENDIF(p, is_poly);
+}
+
+
+
+static void apply_one_offset( struct brw_clip_compile *c,
+			  struct brw_indirect vert )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg pos = deref_4f(vert, c->offset[VERT_RESULT_HPOS]);
+   struct brw_reg z = get_element(pos, 2);
+
+   brw_ADD(p, z, z, vec1(c->reg.offset));
+}
+
+
+
+/***********************************************************************
+ * Output clipped polygon as an unfilled primitive:
+ */
+static void emit_lines(struct brw_clip_compile *c,
+		       boolean do_offset)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *loop;
+   struct brw_instruction *draw_edge;
+   struct brw_indirect v0 = brw_indirect(0, 0);
+   struct brw_indirect v1 = brw_indirect(1, 0);
+   struct brw_indirect v0ptr = brw_indirect(2, 0);
+   struct brw_indirect v1ptr = brw_indirect(3, 0);
+
+   /* Need a seperate loop for offset:
+    */
+   if (do_offset) {
+      brw_MOV(p, c->reg.loopcount, c->reg.nr_verts);
+      brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist));
+
+      loop = brw_DO(p, BRW_EXECUTE_1);
+      {
+	 brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0));
+	 brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2));
+	    
+	 apply_one_offset(c, v0);
+	    
+	 brw_set_conditionalmod(p, BRW_CONDITIONAL_G);
+	 brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1));
+      }
+      brw_WHILE(p, loop);
+   }
+
+   /* v1ptr = &inlist[nr_verts]
+    * *v1ptr = v0
+    */
+   brw_MOV(p, c->reg.loopcount, c->reg.nr_verts);
+   brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist));
+   brw_ADD(p, get_addr_reg(v1ptr), get_addr_reg(v0ptr), retype(c->reg.nr_verts, BRW_REGISTER_TYPE_UW));
+   brw_ADD(p, get_addr_reg(v1ptr), get_addr_reg(v1ptr), retype(c->reg.nr_verts, BRW_REGISTER_TYPE_UW));
+   brw_MOV(p, deref_1uw(v1ptr, 0), deref_1uw(v0ptr, 0));
+
+   loop = brw_DO(p, BRW_EXECUTE_1);
+   {
+      brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0));
+      brw_MOV(p, get_addr_reg(v1), deref_1uw(v0ptr, 2));
+      brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2));
+
+      /* draw edge if edgeflag != 0 */
+      brw_CMP(p, 
+	      vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, 
+	      deref_1f(v0, c->offset[VERT_RESULT_EDGE]),
+	      brw_imm_f(0));
+      draw_edge = brw_IF(p, BRW_EXECUTE_1);
+      {
+	 brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START);
+	 brw_clip_emit_vue(c, v1, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END);
+      }
+      brw_ENDIF(p, draw_edge);
+
+      brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+      brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1));
+   }
+   brw_WHILE(p, loop);
+}
+
+
+
+static void emit_points(struct brw_clip_compile *c,
+			boolean do_offset )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *loop;
+   struct brw_instruction *draw_point;
+
+   struct brw_indirect v0 = brw_indirect(0, 0);
+   struct brw_indirect v0ptr = brw_indirect(2, 0);
+
+   brw_MOV(p, c->reg.loopcount, c->reg.nr_verts);
+   brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist));
+
+   loop = brw_DO(p, BRW_EXECUTE_1);
+   {
+      brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0));
+      brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2));
+
+      /* draw if edgeflag != 0 
+       */
+      brw_CMP(p, 
+	      vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, 
+	      deref_1f(v0, c->offset[VERT_RESULT_EDGE]),
+	      brw_imm_f(0));
+      draw_point = brw_IF(p, BRW_EXECUTE_1);
+      {
+	 if (do_offset)
+	    apply_one_offset(c, v0);
+
+	 brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END);
+      }
+      brw_ENDIF(p, draw_point);
+
+      brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+      brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1));
+   }
+   brw_WHILE(p, loop);
+}
+
+
+
+
+
+
+
+static void emit_primitives( struct brw_clip_compile *c,
+			     unsigned mode, 
+			     boolean do_offset )
+{
+   switch (mode) {
+   case CLIP_FILL:
+      brw_clip_tri_emit_polygon(c);
+      break;
+
+   case CLIP_LINE:
+      emit_lines(c, do_offset);
+      break;
+
+   case CLIP_POINT:
+      emit_points(c, do_offset);
+      break;
+
+   case CLIP_CULL:
+      assert(0);
+      break;
+   }
+} 
+
+
+
+static void emit_unfilled_primitives( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *ccw;
+
+   /* Direction culling has already been done.
+    */
+   if (c->key.fill_ccw != c->key.fill_cw &&
+       c->key.fill_ccw != CLIP_CULL &&
+       c->key.fill_cw != CLIP_CULL)
+   {
+      brw_CMP(p,
+	      vec1(brw_null_reg()),
+	      BRW_CONDITIONAL_GE,
+	      get_element(c->reg.dir, 2),
+	      brw_imm_f(0));
+   
+      ccw = brw_IF(p, BRW_EXECUTE_1);
+      {
+	 emit_primitives(c, c->key.fill_ccw, c->key.offset_ccw);
+      }
+      ccw = brw_ELSE(p, ccw);
+      {
+	 emit_primitives(c, c->key.fill_cw, c->key.offset_cw);
+      }
+      brw_ENDIF(p, ccw);
+   }
+   else if (c->key.fill_cw != CLIP_CULL) {
+      emit_primitives(c, c->key.fill_cw, c->key.offset_cw);
+   }
+   else if (c->key.fill_ccw != CLIP_CULL) { 
+      emit_primitives(c, c->key.fill_ccw, c->key.offset_ccw);
+   }
+}
+
+
+
+
+static void check_nr_verts( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *if_insn;
+
+   brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.nr_verts, brw_imm_d(3));      
+   if_insn = brw_IF(p, BRW_EXECUTE_1);
+   {
+      brw_clip_kill_thread(c);
+   }
+   brw_ENDIF(p, if_insn);
+}
+
+
+void brw_emit_unfilled_clip( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *do_clip;
+   
+
+   c->need_direction = ((c->key.offset_ccw || c->key.offset_cw) ||
+			(c->key.fill_ccw != c->key.fill_cw) ||
+			c->key.fill_ccw == CLIP_CULL ||
+			c->key.fill_cw == CLIP_CULL ||
+			c->key.copy_bfc_cw ||
+			c->key.copy_bfc_ccw);
+
+   brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6);
+   brw_clip_tri_init_vertices(c);
+
+   assert(c->offset[VERT_RESULT_EDGE]);
+
+   if (c->key.fill_ccw == CLIP_CULL &&
+       c->key.fill_cw == CLIP_CULL) {
+      brw_clip_kill_thread(c);
+      return;
+   }
+
+   merge_edgeflags(c);
+
+   /* Need to use the inlist indirection here: 
+    */
+   if (c->need_direction) 
+      compute_tri_direction(c);
+   
+   if (c->key.fill_ccw == CLIP_CULL ||
+       c->key.fill_cw == CLIP_CULL)
+      cull_direction(c);
+
+   if (c->key.offset_ccw ||
+       c->key.offset_cw)
+      compute_offset(c);
+
+   if (c->key.copy_bfc_ccw ||
+       c->key.copy_bfc_cw)
+      copy_bfc(c);
+
+   /* Need to do this whether we clip or not:
+    */
+   if (c->key.do_flat_shading)
+      brw_clip_tri_flat_shade(c);
+   
+   brw_clip_init_clipmask(c);
+   brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0));
+   do_clip = brw_IF(p, BRW_EXECUTE_1);
+   {
+      brw_clip_init_planes(c);
+      brw_clip_tri(c);
+      check_nr_verts(c);
+   }
+   brw_ENDIF(p, do_clip);
+   
+   emit_unfilled_primitives(c);
+   brw_clip_kill_thread(c);
+}
+
+
+
diff --git a/src/gallium/drivers/i965simple/brw_clip_util.c b/src/gallium/drivers/i965simple/brw_clip_util.c
new file mode 100644
index 0000000000..6d58ceafff
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_clip_util.c
@@ -0,0 +1,351 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_util.h"
+#include "brw_clip.h"
+
+
+
+
+
+static struct brw_reg get_tmp( struct brw_clip_compile *c )
+{
+   struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0);
+
+   if (++c->last_tmp > c->prog_data.total_grf)
+      c->prog_data.total_grf = c->last_tmp;
+
+   return tmp;
+}
+
+static void release_tmp( struct brw_clip_compile *c, struct brw_reg tmp )
+{
+   if (tmp.nr == c->last_tmp-1)
+      c->last_tmp--;
+}
+
+
+static struct brw_reg make_plane_ud(unsigned x, unsigned y, unsigned z, unsigned w)
+{
+   return brw_imm_ud((w<<24) | (z<<16) | (y<<8) | x);
+}
+
+
+void brw_clip_init_planes( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+
+   if (!c->key.nr_userclip) {
+      brw_MOV(p, get_element_ud(c->reg.fixed_planes, 0), make_plane_ud( 0,    0, 0xff, 1));
+      brw_MOV(p, get_element_ud(c->reg.fixed_planes, 1), make_plane_ud( 0,    0,    1, 1));
+      brw_MOV(p, get_element_ud(c->reg.fixed_planes, 2), make_plane_ud( 0, 0xff,    0, 1));
+      brw_MOV(p, get_element_ud(c->reg.fixed_planes, 3), make_plane_ud( 0,    1,    0, 1));
+      brw_MOV(p, get_element_ud(c->reg.fixed_planes, 4), make_plane_ud(0xff,  0,    0, 1));
+      brw_MOV(p, get_element_ud(c->reg.fixed_planes, 5), make_plane_ud( 1,    0,    0, 1));
+   }
+}
+
+
+
+#define W 3
+
+/* Project 'pos' to screen space (or back again), overwrite with results:
+ */
+static void brw_clip_project_position(struct brw_clip_compile *c, struct brw_reg pos )
+{
+   struct brw_compile *p = &c->func;
+
+   /* calc rhw
+    */
+   brw_math_invert(p, get_element(pos, W), get_element(pos, W));
+
+   /* value.xyz *= value.rhw
+    */
+   brw_set_access_mode(p, BRW_ALIGN_16);
+   brw_MUL(p, brw_writemask(pos, TGSI_WRITEMASK_XYZ), pos, brw_swizzle1(pos, W));
+   brw_set_access_mode(p, BRW_ALIGN_1);
+}
+
+
+static void brw_clip_project_vertex( struct brw_clip_compile *c,
+				     struct brw_indirect vert_addr )
+{
+#if 0
+   struct brw_compile *p = &c->func;
+   struct brw_reg tmp = get_tmp(c);
+
+   /* Fixup position.  Extract from the original vertex and re-project
+    * to screen space:
+    */
+   brw_MOV(p, tmp, deref_4f(vert_addr, c->offset[VERT_RESULT_HPOS]));
+   brw_clip_project_position(c, tmp);
+   brw_MOV(p, deref_4f(vert_addr, c->header_position_offset), tmp);
+
+   release_tmp(c, tmp);
+#else
+         #warning "disabled"
+#endif
+}
+
+
+
+
+/* Interpolate between two vertices and put the result into a0.0.
+ * Increment a0.0 accordingly.
+ */
+void brw_clip_interp_vertex( struct brw_clip_compile *c,
+			     struct brw_indirect dest_ptr,
+			     struct brw_indirect v0_ptr, /* from */
+			     struct brw_indirect v1_ptr, /* to */
+			     struct brw_reg t0,
+			     boolean force_edgeflag)
+{
+#if 0
+   struct brw_compile *p = &c->func;
+   struct brw_reg tmp = get_tmp(c);
+   unsigned i;
+
+   /* Just copy the vertex header:
+    */
+   brw_copy_indirect_to_indirect(p, dest_ptr, v0_ptr, 1);
+
+   /* Iterate over each attribute (could be done in pairs?)
+    */
+   for (i = 0; i < c->nr_attrs; i++) {
+      unsigned delta = i*16 + 32;
+
+      if (delta == c->offset[VERT_RESULT_EDGE]) {
+	 if (force_edgeflag)
+	    brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(1));
+	 else
+	    brw_MOV(p, deref_4f(dest_ptr, delta), deref_4f(v0_ptr, delta));
+      }
+      else {
+	 /* Interpolate:
+	  *
+	  *        New = attr0 + t*attr1 - t*attr0
+	  */
+	 brw_MUL(p,
+		 vec4(brw_null_reg()),
+		 deref_4f(v1_ptr, delta),
+		 t0);
+
+	 brw_MAC(p,
+		 tmp,
+		 negate(deref_4f(v0_ptr, delta)),
+		 t0);
+
+	 brw_ADD(p,
+		 deref_4f(dest_ptr, delta),
+		 deref_4f(v0_ptr, delta),
+		 tmp);
+      }
+   }
+
+   if (i & 1) {
+      unsigned delta = i*16 + 32;
+      brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(0));
+   }
+
+   release_tmp(c, tmp);
+
+   /* Recreate the projected (NDC) coordinate in the new vertex
+    * header:
+    */
+   brw_clip_project_vertex(c, dest_ptr );
+#else
+         #warning "disabled"
+#endif
+}
+
+
+
+
+#define MAX_MRF 16
+
+void brw_clip_emit_vue(struct brw_clip_compile *c,
+		       struct brw_indirect vert,
+		       boolean allocate,
+		       boolean eot,
+		       unsigned header)
+{
+   struct brw_compile *p = &c->func;
+   unsigned start = c->last_mrf;
+
+   assert(!(allocate && eot));
+
+   /* Cycle through mrf regs - probably futile as we have to wait for
+    * the allocation response anyway.  Also, the order this function
+    * is invoked doesn't correspond to the order the instructions will
+    * be executed, so it won't have any effect in many cases.
+    */
+#if 0
+   if (start + c->nr_regs + 1 >= MAX_MRF)
+      start = 0;
+
+   c->last_mrf = start + c->nr_regs + 1;
+#endif
+
+   /* Copy the vertex from vertn into m1..mN+1:
+    */
+   brw_copy_from_indirect(p, brw_message_reg(start+1), vert, c->nr_regs);
+
+   /* Overwrite PrimType and PrimStart in the message header, for
+    * each vertex in turn:
+    */
+   brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header));
+
+
+   /* Send each vertex as a seperate write to the urb.  This
+    * is different to the concept in brw_sf_emit.c, where
+    * subsequent writes are used to build up a single urb
+    * entry.  Each of these writes instantiates a seperate
+    * urb entry - (I think... what about 'allocate'?)
+    */
+   brw_urb_WRITE(p,
+		 allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD),
+		 start,
+		 c->reg.R0,
+		 allocate,
+		 1,		/* used */
+		 c->nr_regs + 1, /* msg length */
+		 allocate ? 1 : 0, /* response_length */
+		 eot,		/* eot */
+		 1,		/* writes_complete */
+		 0,		/* urb offset */
+		 BRW_URB_SWIZZLE_NONE);
+}
+
+
+
+void brw_clip_kill_thread(struct brw_clip_compile *c)
+{
+   struct brw_compile *p = &c->func;
+
+   /* Send an empty message to kill the thread and release any
+    * allocated urb entry:
+    */
+   brw_urb_WRITE(p,
+		 retype(brw_null_reg(), BRW_REGISTER_TYPE_UD),
+		 0,
+		 c->reg.R0,
+		 0,		/* allocate */
+		 0,		/* used */
+		 0, 		/* msg len */
+		 0, 		/* response len */
+		 1, 		/* eot */
+		 1,		/* writes complete */
+		 0,
+		 BRW_URB_SWIZZLE_NONE);
+}
+
+
+
+
+struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c )
+{
+   return brw_address(c->reg.fixed_planes);
+}
+
+
+struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c )
+{
+   if (c->key.nr_userclip) {
+      return brw_imm_uw(16);
+   }
+   else {
+      return brw_imm_uw(4);
+   }
+}
+
+
+/* If flatshading, distribute color from provoking vertex prior to
+ * clipping.
+ */
+void brw_clip_copy_colors( struct brw_clip_compile *c,
+			   unsigned to, unsigned from )
+{
+#if 0
+   struct brw_compile *p = &c->func;
+
+   if (c->offset[VERT_RESULT_COL0])
+      brw_MOV(p,
+	      byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_COL0]),
+	      byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_COL0]));
+
+   if (c->offset[VERT_RESULT_COL1])
+      brw_MOV(p,
+	      byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_COL1]),
+	      byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_COL1]));
+
+   if (c->offset[VERT_RESULT_BFC0])
+      brw_MOV(p,
+	      byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_BFC0]),
+	      byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_BFC0]));
+
+   if (c->offset[VERT_RESULT_BFC1])
+      brw_MOV(p,
+	      byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_BFC1]),
+	      byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_BFC1]));
+#else
+         #warning "disabled"
+#endif
+}
+
+
+
+void brw_clip_init_clipmask( struct brw_clip_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg incoming = get_element_ud(c->reg.R0, 2);
+
+   /* Shift so that lowest outcode bit is rightmost:
+    */
+   brw_SHR(p, c->reg.planemask, incoming, brw_imm_ud(26));
+
+   if (c->key.nr_userclip) {
+      struct brw_reg tmp = retype(vec1(get_tmp(c)), BRW_REGISTER_TYPE_UD);
+
+      /* Rearrange userclip outcodes so that they come directly after
+       * the fixed plane bits.
+       */
+      brw_AND(p, tmp, incoming, brw_imm_ud(0x3f<<14));
+      brw_SHR(p, tmp, tmp, brw_imm_ud(8));
+      brw_OR(p, c->reg.planemask, c->reg.planemask, tmp);
+
+      release_tmp(c, tmp);
+   }
+}
+
diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c
new file mode 100644
index 0000000000..c74cbf8d73
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_context.c
@@ -0,0 +1,114 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_context.h"
+#include "brw_defines.h"
+#include "brw_draw.h"
+#include "brw_vs.h"
+#include "brw_tex_layout.h"
+#include "brw_winsys.h"
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_context.h"
+#include "util/u_memory.h"
+#include "pipe/p_screen.h"
+
+
+#ifndef BRW_DEBUG
+int BRW_DEBUG = (0);
+#endif
+
+
+static void brw_destroy(struct pipe_context *pipe)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   if(brw->winsys->destroy)
+      brw->winsys->destroy(brw->winsys);
+   
+   FREE(brw);
+}
+
+
+static void brw_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+                      unsigned clearValue)
+{
+   int x, y, w, h;
+   /* FIXME: corny... */
+
+   x = 0;
+   y = 0;
+   w = ps->width;
+   h = ps->height;
+
+   pipe->surface_fill(pipe, ps, x, y, w, h, clearValue);
+}
+
+
+struct pipe_context *brw_create(struct pipe_screen *screen,
+                                struct brw_winsys *brw_winsys,
+                                unsigned pci_id)
+{
+   struct brw_context *brw;
+
+   debug_printf("%s: creating brw_context with pci id 0x%x\n",
+                __FUNCTION__, pci_id);
+
+   brw = CALLOC_STRUCT(brw_context);
+   if (brw == NULL)
+      return NULL;
+
+   brw->winsys = brw_winsys;
+   brw->pipe.winsys = screen->winsys;
+   brw->pipe.screen = screen;
+
+   brw->pipe.destroy = brw_destroy;
+   brw->pipe.clear = brw_clear;
+
+   brw_init_surface_functions(brw);
+   brw_init_texture_functions(brw);
+   brw_init_state_functions(brw);
+   brw_init_flush_functions(brw);
+   brw_init_draw_functions( brw );
+
+
+   brw_init_state( brw );
+
+   brw->pci_id = pci_id;
+   brw->dirty = ~0;
+   brw->hardware_dirty = ~0;
+
+   memset(&brw->wm.bind, ~0, sizeof(brw->wm.bind));
+
+   return &brw->pipe;
+}
+
diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h
new file mode 100644
index 0000000000..3079485180
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_context.h
@@ -0,0 +1,684 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#ifndef BRWCONTEXT_INC
+#define BRWCONTEXT_INC
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "tgsi/tgsi_scan.h"
+
+#include "brw_structs.h"
+#include "brw_winsys.h"
+
+
+/* Glossary:
+ *
+ * URB - uniform resource buffer.  A mid-sized buffer which is
+ * partitioned between the fixed function units and used for passing
+ * values (vertices, primitives, constants) between them.
+ *
+ * CURBE - constant URB entry.  An urb region (entry) used to hold
+ * constant values which the fixed function units can be instructed to
+ * preload into the GRF when spawining a thread.
+ *
+ * VUE - vertex URB entry.  An urb entry holding a vertex and usually
+ * a vertex header.  The header contains control information and
+ * things like primitive type, Begin/end flags and clip codes.
+ *
+ * PUE - primitive URB entry.  An urb entry produced by the setup (SF)
+ * unit holding rasterization and interpolation parameters.
+ *
+ * GRF - general register file.  One of several register files
+ * addressable by programmed threads.  The inputs (r0, payload, curbe,
+ * urb) of the thread are preloaded to this area before the thread is
+ * spawned.  The registers are individually 8 dwords wide and suitable
+ * for general usage.  Registers holding thread input values are not
+ * special and may be overwritten.
+ *
+ * MRF - message register file.  Threads communicate (and terminate)
+ * by sending messages.  Message parameters are placed in contigous
+ * MRF registers.  All program output is via these messages.  URB
+ * entries are populated by sending a message to the shared URB
+ * function containing the new data, together with a control word,
+ * often an unmodified copy of R0.
+ *
+ * R0 - GRF register 0.  Typically holds control information used when
+ * sending messages to other threads.
+ *
+ * EU or GEN4 EU: The name of the programmable subsystem of the
+ * i965 hardware.  Threads are executed by the EU, the registers
+ * described above are part of the EU architecture.
+ *
+ * Fixed function units:
+ *
+ * CS - Command streamer.  Notional first unit, little software
+ * interaction.  Holds the URB entries used for constant data, ie the
+ * CURBEs.
+ *
+ * VF/VS - Vertex Fetch / Vertex Shader.  The fixed function part of
+ * this unit is responsible for pulling vertices out of vertex buffers
+ * in vram and injecting them into the processing pipe as VUEs.  If
+ * enabled, it first passes them to a VS thread which is a good place
+ * for the driver to implement any active vertex shader.
+ *
+ * GS - Geometry Shader.  This corresponds to a new DX10 concept.  If
+ * enabled, incoming strips etc are passed to GS threads in individual
+ * line/triangle/point units.  The GS thread may perform arbitary
+ * computation and emit whatever primtives with whatever vertices it
+ * chooses.  This makes GS an excellent place to implement GL's
+ * unfilled polygon modes, though of course it is capable of much
+ * more.  Additionally, GS is used to translate away primitives not
+ * handled by latter units, including Quads and Lineloops.
+ *
+ * CS - Clipper.  Mesa's clipping algorithms are imported to run on
+ * this unit.  The fixed function part performs cliptesting against
+ * the 6 fixed clipplanes and makes descisions on whether or not the
+ * incoming primitive needs to be passed to a thread for clipping.
+ * User clip planes are handled via cooperation with the VS thread.
+ *
+ * SF - Strips Fans or Setup: Triangles are prepared for
+ * rasterization.  Interpolation coefficients are calculated.
+ * Flatshading and two-side lighting usually performed here.
+ *
+ * WM - Windower.  Interpolation of vertex attributes performed here.
+ * Fragment shader implemented here.  SIMD aspects of EU taken full
+ * advantage of, as pixels are processed in blocks of 16.
+ *
+ * CC - Color Calculator.  No EU threads associated with this unit.
+ * Handles blending and (presumably) depth and stencil testing.
+ */
+
+#define BRW_MAX_CURBE                    (32*16)
+
+struct brw_context;
+struct brw_winsys;
+
+
+/* Raised when we receive new state across the pipe interface:
+ */
+#define BRW_NEW_VIEWPORT                0x1
+#define BRW_NEW_RASTERIZER              0x2
+#define BRW_NEW_FS                      0x4
+#define BRW_NEW_BLEND                   0x8
+#define BRW_NEW_CLIP                    0x10
+#define BRW_NEW_SCISSOR                 0x20
+#define BRW_NEW_STIPPLE                 0x40
+#define BRW_NEW_FRAMEBUFFER             0x80
+#define BRW_NEW_ALPHA_TEST              0x100
+#define BRW_NEW_DEPTH_STENCIL           0x200
+#define BRW_NEW_SAMPLER                 0x400
+#define BRW_NEW_TEXTURE                 0x800
+#define BRW_NEW_CONSTANTS               0x1000
+#define BRW_NEW_VBO                     0x2000
+#define BRW_NEW_VS                      0x4000
+
+/* Raised for other internal events:
+ */
+#define BRW_NEW_URB_FENCE               0x10000
+#define BRW_NEW_PSP                     0x20000
+#define BRW_NEW_CURBE_OFFSETS           0x40000
+#define BRW_NEW_REDUCED_PRIMITIVE       0x80000
+#define BRW_NEW_PRIMITIVE               0x100000
+#define BRW_NEW_SCENE                 0x200000
+#define BRW_NEW_SF_LINKAGE              0x400000
+
+extern int BRW_DEBUG;
+
+#define DEBUG_TEXTURE	0x1
+#define DEBUG_STATE	0x2
+#define DEBUG_IOCTL	0x4
+#define DEBUG_PRIMS	0x8
+#define DEBUG_VERTS	0x10
+#define DEBUG_FALLBACKS	0x20
+#define DEBUG_VERBOSE	0x40
+#define DEBUG_DRI       0x80
+#define DEBUG_DMA       0x100
+#define DEBUG_SANITY    0x200
+#define DEBUG_SYNC      0x400
+#define DEBUG_SLEEP     0x800
+#define DEBUG_PIXEL     0x1000
+#define DEBUG_STATS     0x2000
+#define DEBUG_TILE      0x4000
+#define DEBUG_SINGLE_THREAD   0x8000
+#define DEBUG_WM        0x10000
+#define DEBUG_URB       0x20000
+#define DEBUG_VS        0x40000
+#define DEBUG_BATCH	0x80000
+#define DEBUG_BUFMGR	0x100000
+#define DEBUG_BLIT	0x200000
+#define DEBUG_REGION	0x400000
+#define DEBUG_MIPTREE	0x800000
+
+#define DBG(...) do {						\
+   if (BRW_DEBUG & FILE_DEBUG_FLAG)				\
+      debug_printf(__VA_ARGS__);				\
+} while(0)
+
+#define PRINT(...) do {						\
+   debug_printf(__VA_ARGS__);			                \
+} while(0)
+
+struct brw_state_flags {
+   unsigned cache;
+   unsigned brw;
+};
+
+
+struct brw_vertex_program {
+   struct pipe_shader_state program;
+   struct tgsi_shader_info info;
+   int id;
+};
+
+
+struct brw_fragment_program {
+   struct pipe_shader_state program;
+   struct tgsi_shader_info info;
+   
+   boolean UsesDepth; /* XXX add this to tgsi_shader_info? */
+   int id;
+};
+
+
+struct pipe_setup_linkage {
+   struct {
+      unsigned vp_output:5;
+      unsigned interp_mode:4;
+      unsigned bf_vp_output:5;
+   } fp_input[PIPE_MAX_SHADER_INPUTS];
+
+   unsigned fp_input_count:5;
+   unsigned max_vp_output:5;
+};
+   
+
+
+struct brw_texture {
+   struct pipe_texture base;
+
+   /* Derived from the above:
+    */
+   unsigned stride;
+   unsigned depth_pitch;          /* per-image on i945? */
+   unsigned total_nblocksy;
+
+   unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS];
+
+   /* Explicitly store the offset of each image for each cube face or
+    * depth value.  Pretty much have to accept that hardware formats
+    * are going to be so diverse that there is no unified way to
+    * compute the offsets of depth/cube images within a mipmap level,
+    * so have to store them as a lookup table:
+    */
+   unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS];   /**< array [depth] of offsets */
+
+   /* Includes image offset tables:
+    */
+   unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS];
+
+   /* The data is held here:
+    */
+   struct pipe_buffer *buffer;
+};
+
+/* Data about a particular attempt to compile a program.  Note that
+ * there can be many of these, each in a different GL state
+ * corresponding to a different brw_wm_prog_key struct, with different
+ * compiled programs:
+ */
+/* Data about a particular attempt to compile a program.  Note that
+ * there can be many of these, each in a different GL state
+ * corresponding to a different brw_wm_prog_key struct, with different
+ * compiled programs:
+ */
+
+struct brw_wm_prog_data {
+   unsigned curb_read_length;
+   unsigned urb_read_length;
+
+   unsigned first_curbe_grf;
+   unsigned total_grf;
+   unsigned total_scratch;
+
+   /* Internally generated constants for the CURBE.  These are loaded
+    * ahead of the data from the constant buffer.
+    */
+   const float internal_const[8];
+   unsigned nr_internal_consts;
+   unsigned max_const;
+
+   boolean error;
+};
+
+struct brw_sf_prog_data {
+   unsigned urb_read_length;
+   unsigned total_grf;
+
+   /* Each vertex may have upto 12 attributes, 4 components each,
+    * except WPOS which requires only 2.  (11*4 + 2) == 44 ==> 11
+    * rows.
+    *
+    * Actually we use 4 for each, so call it 12 rows.
+    */
+   unsigned urb_entry_size;
+};
+
+struct brw_clip_prog_data {
+   unsigned curb_read_length;	/* user planes? */
+   unsigned clip_mode;
+   unsigned urb_read_length;
+   unsigned total_grf;
+};
+
+struct brw_gs_prog_data {
+   unsigned urb_read_length;
+   unsigned total_grf;
+};
+
+struct brw_vs_prog_data {
+   unsigned curb_read_length;
+   unsigned urb_read_length;
+   unsigned total_grf;
+   unsigned outputs_written;
+
+   unsigned inputs_read;
+
+   unsigned max_const;
+
+   float    imm_buf[PIPE_MAX_CONSTANT][4];
+   unsigned num_imm;
+   unsigned num_consts;
+
+   /* Used for calculating urb partitions:
+    */
+   unsigned urb_entry_size;
+};
+
+
+#define BRW_MAX_TEX_UNIT 8
+#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1
+
+/* Create a fixed sized struct for caching binding tables:
+ */
+struct brw_surface_binding_table {
+   unsigned surf_ss_offset[BRW_WM_MAX_SURF];
+};
+
+
+struct brw_cache;
+
+struct brw_mem_pool {
+   struct pipe_buffer *buffer;
+
+   unsigned size;
+   unsigned offset;		/* offset of first free byte */
+
+   struct brw_context *brw;
+};
+
+struct brw_cache_item {
+   unsigned hash;
+   unsigned key_size;		/* for variable-sized keys */
+   const void *key;
+
+   unsigned offset;		/* offset within pool's buffer */
+   unsigned data_size;
+
+   struct brw_cache_item *next;
+};
+
+
+
+struct brw_cache {
+   unsigned id;
+
+   const char *name;
+
+   struct brw_context *brw;
+   struct brw_mem_pool *pool;
+
+   struct brw_cache_item **items;
+   unsigned size, n_items;
+
+   unsigned key_size;		/* for fixed-size keys */
+   unsigned aux_size;
+
+   unsigned last_addr;			/* offset of active item */
+};
+
+
+
+
+/* Considered adding a member to this struct to document which flags
+ * an update might raise so that ordering of the state atoms can be
+ * checked or derived at runtime.  Dropped the idea in favor of having
+ * a debug mode where the state is monitored for flags which are
+ * raised that have already been tested against.
+ */
+struct brw_tracked_state {
+   struct brw_state_flags dirty;
+   void (*update)( struct brw_context *brw );
+};
+
+
+/* Flags for brw->state.cache.
+ */
+#define CACHE_NEW_CC_VP                  (1<<BRW_CC_VP)
+#define CACHE_NEW_CC_UNIT                (1<<BRW_CC_UNIT)
+#define CACHE_NEW_WM_PROG                (1<<BRW_WM_PROG)
+#define CACHE_NEW_SAMPLER_DEFAULT_COLOR  (1<<BRW_SAMPLER_DEFAULT_COLOR)
+#define CACHE_NEW_SAMPLER                (1<<BRW_SAMPLER)
+#define CACHE_NEW_WM_UNIT                (1<<BRW_WM_UNIT)
+#define CACHE_NEW_SF_PROG                (1<<BRW_SF_PROG)
+#define CACHE_NEW_SF_VP                  (1<<BRW_SF_VP)
+#define CACHE_NEW_SF_UNIT                (1<<BRW_SF_UNIT)
+#define CACHE_NEW_VS_UNIT                (1<<BRW_VS_UNIT)
+#define CACHE_NEW_VS_PROG                (1<<BRW_VS_PROG)
+#define CACHE_NEW_GS_UNIT                (1<<BRW_GS_UNIT)
+#define CACHE_NEW_GS_PROG                (1<<BRW_GS_PROG)
+#define CACHE_NEW_CLIP_VP                (1<<BRW_CLIP_VP)
+#define CACHE_NEW_CLIP_UNIT              (1<<BRW_CLIP_UNIT)
+#define CACHE_NEW_CLIP_PROG              (1<<BRW_CLIP_PROG)
+#define CACHE_NEW_SURFACE                (1<<BRW_SS_SURFACE)
+#define CACHE_NEW_SURF_BIND              (1<<BRW_SS_SURF_BIND)
+
+
+
+
+enum brw_mempool_id {
+   BRW_GS_POOL,
+   BRW_SS_POOL,
+   BRW_MAX_POOL
+};
+
+
+struct brw_cached_batch_item {
+   struct header *header;
+   unsigned sz;
+   struct brw_cached_batch_item *next;
+};
+
+
+
+/* Protect against a future where PIPE_MAX_ATTRIBS > 32.  Wouldn't life
+ * be easier if C allowed arrays of packed elements?
+ */
+#define ATTRIB_BIT_DWORDS  ((PIPE_MAX_ATTRIBS+31)/32)
+
+
+
+
+struct brw_vertex_info {
+   unsigned varying;  /* varying:1[PIPE_MAX_ATTRIBS] */
+   unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_MAX_ATTRIBS] */
+};
+
+
+
+
+
+struct brw_context
+{
+   struct pipe_context pipe;
+   struct brw_winsys *winsys;
+
+   unsigned primitive;
+   unsigned reduced_primitive;
+
+   boolean emit_state_always;
+
+   struct {
+      struct brw_state_flags dirty;
+   } state;
+
+
+   struct {
+      const struct pipe_blend_state         *Blend;
+      const struct pipe_depth_stencil_alpha_state *DepthStencil;
+      const struct pipe_poly_stipple        *PolygonStipple;
+      const struct pipe_rasterizer_state    *Raster;
+      const struct pipe_sampler_state       *Samplers[PIPE_MAX_SAMPLERS];
+      const struct brw_vertex_program       *VertexProgram;
+      const struct brw_fragment_program     *FragmentProgram;
+
+      struct pipe_clip_state          Clip;
+      struct pipe_blend_color         BlendColor;
+      struct pipe_scissor_state       Scissor;
+      struct pipe_viewport_state      Viewport;
+      struct pipe_framebuffer_state   FrameBuffer;
+
+      const struct pipe_constant_buffer *Constants[2];
+      const struct brw_texture          *Texture[PIPE_MAX_SAMPLERS];
+   } attribs;
+
+   unsigned num_samplers;
+   unsigned num_textures;
+
+   struct brw_mem_pool pool[BRW_MAX_POOL];
+   struct brw_cache cache[BRW_MAX_CACHE];
+   struct brw_cached_batch_item *cached_batch_items;
+
+   struct {
+
+      /* Arrays with buffer objects to copy non-bufferobj arrays into
+       * for upload:
+       */
+      const struct pipe_vertex_buffer *vbo_array[PIPE_MAX_ATTRIBS];
+
+      struct brw_vertex_element_state inputs[PIPE_MAX_ATTRIBS];
+
+#define BRW_NR_UPLOAD_BUFS 17
+#define BRW_UPLOAD_INIT_SIZE (128*1024)
+
+      /* Summary of size and varying of active arrays, so we can check
+       * for changes to this state:
+       */
+      struct brw_vertex_info info;
+   } vb;
+
+
+   unsigned hardware_dirty;
+   unsigned dirty;
+   unsigned pci_id;
+   /* BRW_NEW_URB_ALLOCATIONS:
+    */
+   struct {
+      unsigned vsize;		/* vertex size plus header in urb registers */
+      unsigned csize;		/* constant buffer size in urb registers */
+      unsigned sfsize;		/* setup data size in urb registers */
+
+      boolean constrained;
+
+      unsigned nr_vs_entries;
+      unsigned nr_gs_entries;
+      unsigned nr_clip_entries;
+      unsigned nr_sf_entries;
+      unsigned nr_cs_entries;
+
+/*       unsigned vs_size; */
+/*       unsigned gs_size; */
+/*       unsigned clip_size; */
+/*       unsigned sf_size; */
+/*       unsigned cs_size; */
+
+      unsigned vs_start;
+      unsigned gs_start;
+      unsigned clip_start;
+      unsigned sf_start;
+      unsigned cs_start;
+   } urb;
+
+
+   /* BRW_NEW_CURBE_OFFSETS:
+    */
+   struct {
+      unsigned wm_start;
+      unsigned wm_size;
+      unsigned clip_start;
+      unsigned clip_size;
+      unsigned vs_start;
+      unsigned vs_size;
+      unsigned total_size;
+
+      unsigned gs_offset;
+
+      float *last_buf;
+      unsigned last_bufsz;
+   } curbe;
+
+   struct {
+      struct brw_vs_prog_data *prog_data;
+
+      unsigned prog_gs_offset;
+      unsigned state_gs_offset;
+   } vs;
+
+   struct {
+      struct brw_gs_prog_data *prog_data;
+
+      boolean prog_active;
+      unsigned prog_gs_offset;
+      unsigned state_gs_offset;
+   } gs;
+
+   struct {
+      struct brw_clip_prog_data *prog_data;
+
+      unsigned prog_gs_offset;
+      unsigned vp_gs_offset;
+      unsigned state_gs_offset;
+   } clip;
+
+
+   struct {
+      struct brw_sf_prog_data *prog_data;
+
+      struct pipe_setup_linkage linkage;
+
+      unsigned prog_gs_offset;
+      unsigned vp_gs_offset;
+      unsigned state_gs_offset;
+   } sf;
+
+   struct {
+      struct brw_wm_prog_data *prog_data;
+
+//      struct brw_wm_compiler *compile_data;
+
+
+      /**
+       * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER
+       * cache
+       */
+      struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT];
+
+      unsigned render_surf;
+      unsigned nr_surfaces;
+
+      unsigned max_threads;
+      struct pipe_buffer *scratch_buffer;
+      unsigned scratch_buffer_size;
+
+      unsigned sampler_count;
+      unsigned sampler_gs_offset;
+
+      struct brw_surface_binding_table bind;
+      unsigned bind_ss_offset;
+
+      unsigned prog_gs_offset;
+      unsigned state_gs_offset;
+   } wm;
+
+
+   struct {
+      unsigned vp_gs_offset;
+      unsigned state_gs_offset;
+   } cc;
+
+
+   /* Used to give every program string a unique id
+    */
+   unsigned program_id;
+};
+
+
+#define BRW_PACKCOLOR8888(r,g,b,a)  ((r<<24) | (g<<16) | (b<<8) | a)
+
+
+/*======================================================================
+ * brw_vtbl.c
+ */
+void brw_do_flush( struct brw_context *brw,
+		   unsigned flags );
+
+
+/*======================================================================
+ * brw_state.c
+ */
+void brw_validate_state(struct brw_context *brw);
+void brw_init_state(struct brw_context *brw);
+void brw_destroy_state(struct brw_context *brw);
+
+
+/*======================================================================
+ * brw_tex.c
+ */
+void brwUpdateTextureState( struct brw_context *brw );
+
+
+/* brw_urb.c
+ */
+void brw_upload_urb_fence(struct brw_context *brw);
+
+void brw_upload_constant_buffer_state(struct brw_context *brw);
+
+void brw_init_surface_functions(struct brw_context *brw);
+void brw_init_state_functions(struct brw_context *brw);
+void brw_init_flush_functions(struct brw_context *brw);
+void brw_init_string_functions(struct brw_context *brw);
+
+/*======================================================================
+ * Inline conversion functions.  These are better-typed than the
+ * macros used previously:
+ */
+static inline struct brw_context *
+brw_context( struct pipe_context *ctx )
+{
+   return (struct brw_context *)ctx;
+}
+
+#endif
+
diff --git a/src/gallium/drivers/i965simple/brw_curbe.c b/src/gallium/drivers/i965simple/brw_curbe.c
new file mode 100644
index 0000000000..904cde8e30
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_curbe.c
@@ -0,0 +1,369 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+
+#include "brw_context.h"
+#include "brw_defines.h"
+#include "brw_state.h"
+#include "brw_batch.h"
+#include "brw_util.h"
+#include "brw_wm.h"
+#include "pipe/p_state.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#define FILE_DEBUG_FLAG DEBUG_FALLBACKS
+
+/* Partition the CURBE between the various users of constant values:
+ */
+static void calculate_curbe_offsets( struct brw_context *brw )
+{
+   /* CACHE_NEW_WM_PROG */
+   unsigned nr_fp_regs = align(brw->wm.prog_data->max_const, 16);
+
+   /* BRW_NEW_VERTEX_PROGRAM */
+   unsigned nr_vp_regs = align(brw->vs.prog_data->max_const, 16);
+   unsigned nr_clip_regs = 0;
+   unsigned total_regs;
+
+#if 0
+   /* BRW_NEW_CLIP ? */
+   if (brw->attribs.Transform->ClipPlanesEnabled) {
+      unsigned nr_planes = 6 + brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled);
+      nr_clip_regs = align(nr_planes * 4, 16);
+   }
+#endif
+
+
+   total_regs = nr_fp_regs + nr_vp_regs + nr_clip_regs;
+
+   /* This can happen - what to do?  Probably rather than falling
+    * back, the best thing to do is emit programs which code the
+    * constants as immediate values.  Could do this either as a static
+    * cap on WM and VS, or adaptively.
+    *
+    * Unfortunately, this is currently dependent on the results of the
+    * program generation process (in the case of wm), so this would
+    * introduce the need to re-generate programs in the event of a
+    * curbe allocation failure.
+    */
+   /* Max size is 32 - just large enough to
+    * hold the 128 parameters allowed by
+    * the fragment and vertex program
+    * api's.  It's not clear what happens
+    * when both VP and FP want to use 128
+    * parameters, though.
+    */
+   assert(total_regs <= 32);
+
+   /* Lazy resize:
+    */
+   if (nr_fp_regs > brw->curbe.wm_size ||
+       nr_vp_regs > brw->curbe.vs_size ||
+       nr_clip_regs != brw->curbe.clip_size ||
+       (total_regs < brw->curbe.total_size / 4 &&
+	brw->curbe.total_size > 16)) {
+
+      unsigned reg = 0;
+
+      /* Calculate a new layout:
+       */
+      reg = 0;
+      brw->curbe.wm_start = reg;
+      brw->curbe.wm_size = nr_fp_regs; reg += nr_fp_regs;
+      brw->curbe.clip_start = reg;
+      brw->curbe.clip_size = nr_clip_regs; reg += nr_clip_regs;
+      brw->curbe.vs_start = reg;
+      brw->curbe.vs_size = nr_vp_regs; reg += nr_vp_regs;
+      brw->curbe.total_size = reg;
+
+#if 0
+      if (0)
+	 DBG("curbe wm %d+%d clip %d+%d vs %d+%d\n",
+		      brw->curbe.wm_start,
+		      brw->curbe.wm_size,
+		      brw->curbe.clip_start,
+		      brw->curbe.clip_size,
+		      brw->curbe.vs_start,
+		      brw->curbe.vs_size );
+#endif
+
+      brw->state.dirty.brw |= BRW_NEW_CURBE_OFFSETS;
+   }
+}
+
+
+const struct brw_tracked_state brw_curbe_offsets = {
+   .dirty = {
+      .brw  = (BRW_NEW_CLIP |
+	       BRW_NEW_VS),
+      .cache = CACHE_NEW_WM_PROG
+   },
+   .update = calculate_curbe_offsets
+};
+
+
+
+/* Define the number of curbes within CS's urb allocation.  Multiple
+ * urb entries -> multiple curbes.  These will be used by
+ * fixed-function hardware in a double-buffering scheme to avoid a
+ * pipeline stall each time the contents of the curbe is changed.
+ */
+void brw_upload_constant_buffer_state(struct brw_context *brw)
+{
+   struct brw_constant_buffer_state cbs;
+   memset(&cbs, 0, sizeof(cbs));
+
+   /* It appears that this is the state packet for the CS unit, ie. the
+    * urb entries detailed here are housed in the CS range from the
+    * URB_FENCE command.
+    */
+   cbs.header.opcode = CMD_CONST_BUFFER_STATE;
+   cbs.header.length = sizeof(cbs)/4 - 2;
+
+   /* BRW_NEW_URB_FENCE */
+   cbs.bits0.nr_urb_entries = brw->urb.nr_cs_entries;
+   cbs.bits0.urb_entry_size = brw->urb.csize - 1;
+
+   assert(brw->urb.nr_cs_entries);
+   BRW_CACHED_BATCH_STRUCT(brw, &cbs);
+}
+
+
+static float fixed_plane[6][4] = {
+   { 0,    0,   -1, 1 },
+   { 0,    0,    1, 1 },
+   { 0,   -1,    0, 1 },
+   { 0,    1,    0, 1 },
+   {-1,    0,    0, 1 },
+   { 1,    0,    0, 1 }
+};
+
+/* Upload a new set of constants.  Too much variability to go into the
+ * cache mechanism, but maybe would benefit from a comparison against
+ * the current uploaded set of constants.
+ */
+static void upload_constant_buffer(struct brw_context *brw)
+{
+   struct brw_mem_pool *pool = &brw->pool[BRW_GS_POOL];
+   unsigned sz = brw->curbe.total_size;
+   unsigned bufsz = sz * sizeof(float);
+   float *buf;
+   unsigned i;
+
+
+   if (sz == 0) {
+      struct brw_constant_buffer cb;
+      cb.header.opcode = CMD_CONST_BUFFER;
+      cb.header.length = sizeof(cb)/4 - 2;
+      cb.header.valid = 0;
+      cb.bits0.buffer_length = 0;
+      cb.bits0.buffer_address = 0;
+      BRW_BATCH_STRUCT(brw, &cb);
+
+      if (brw->curbe.last_buf) {
+	 free(brw->curbe.last_buf);
+	 brw->curbe.last_buf = NULL;
+	 brw->curbe.last_bufsz  = 0;
+      }
+
+      return;
+   }
+
+   buf = (float *)malloc(bufsz);
+
+   memset(buf, 0, bufsz);
+
+   if (brw->curbe.wm_size) {
+      unsigned offset = brw->curbe.wm_start * 16;
+
+      /* First the constant buffer constants:
+       */
+      
+      /* Then any internally generated constants: 
+       */
+      for (i = 0; i < brw->wm.prog_data->nr_internal_consts; i++)
+	 buf[offset + i] = brw->wm.prog_data->internal_const[i];
+
+      assert(brw->wm.prog_data->max_const == 
+	     brw->wm.prog_data->nr_internal_consts);
+   }
+
+
+   /* The clipplanes are actually delivered to both CLIP and VS units.
+    * VS uses them to calculate the outcode bitmasks.
+    */
+   if (brw->curbe.clip_size) {
+      unsigned offset = brw->curbe.clip_start * 16;
+      unsigned j;
+
+      /* If any planes are going this way, send them all this way:
+       */
+      for (i = 0; i < 6; i++) {
+	 buf[offset + i * 4 + 0] = fixed_plane[i][0];
+	 buf[offset + i * 4 + 1] = fixed_plane[i][1];
+	 buf[offset + i * 4 + 2] = fixed_plane[i][2];
+	 buf[offset + i * 4 + 3] = fixed_plane[i][3];
+      }
+
+      /* Clip planes: BRW_NEW_CLIP:
+       */
+      for (j = 0; j < brw->attribs.Clip.nr; j++) {
+	 buf[offset + i * 4 + 0] = brw->attribs.Clip.ucp[j][0];
+	 buf[offset + i * 4 + 1] = brw->attribs.Clip.ucp[j][1];
+	 buf[offset + i * 4 + 2] = brw->attribs.Clip.ucp[j][2];
+	 buf[offset + i * 4 + 3] = brw->attribs.Clip.ucp[j][3];
+	 i++;
+      }
+   }
+
+
+   if (brw->curbe.vs_size) {
+      unsigned offset = brw->curbe.vs_start * 16;
+      /*unsigned nr = vp->max_const;*/
+      const struct pipe_constant_buffer *cbuffer = brw->attribs.Constants[0];
+      struct pipe_winsys *ws = brw->pipe.winsys;
+      /* FIXME: buffer size is num_consts + num_immediates */
+      if (brw->vs.prog_data->num_consts) {
+         /* map the vertex constant buffer and copy to curbe: */
+         void *data = ws->buffer_map(ws, cbuffer->buffer, 0);
+         /* FIXME: this is wrong. the cbuffer->buffer->size currently
+          * represents size of consts + immediates. so if we'll
+          * have both we'll copy over the end of the buffer
+          * with the subsequent memcpy */
+         memcpy(&buf[offset], data, cbuffer->buffer->size);
+         ws->buffer_unmap(ws, cbuffer->buffer);
+         offset += cbuffer->buffer->size;
+      }
+      /*immediates*/
+      if (brw->vs.prog_data->num_imm) {
+         memcpy(&buf[offset], brw->vs.prog_data->imm_buf,
+                brw->vs.prog_data->num_imm * 4 * sizeof(float));
+      }
+   }
+
+   if (1) {
+      for (i = 0; i < sz; i+=4)
+	 debug_printf("curbe %d.%d: %f %f %f %f\n", i/8, i&4,
+		      buf[i+0], buf[i+1], buf[i+2], buf[i+3]);
+
+      debug_printf("last_buf %p buf %p sz %d/%d cmp %d\n",
+		   brw->curbe.last_buf, buf,
+		   bufsz, brw->curbe.last_bufsz,
+		   brw->curbe.last_buf ? memcmp(buf, brw->curbe.last_buf, bufsz) : -1);
+   }
+
+   if (brw->curbe.last_buf &&
+       bufsz == brw->curbe.last_bufsz &&
+       memcmp(buf, brw->curbe.last_buf, bufsz) == 0) {
+      free(buf);
+/*       return; */
+   }
+   else {
+      if (brw->curbe.last_buf)
+	 free(brw->curbe.last_buf);
+      brw->curbe.last_buf = buf;
+      brw->curbe.last_bufsz = bufsz;
+
+
+      if (!brw_pool_alloc(pool,
+			  bufsz,
+			  1 << 6,
+			  &brw->curbe.gs_offset)) {
+	 debug_printf("out of GS memory for curbe\n");
+	 assert(0);
+	 return;
+      }
+
+
+      /* Copy data to the buffer:
+       */
+      brw->winsys->buffer_subdata_typed(brw->winsys,
+					pool->buffer, 
+					brw->curbe.gs_offset, 
+					bufsz, 
+					buf,
+					BRW_CONSTANT_BUFFER );
+   }
+
+   /* TODO: only emit the constant_buffer packet when necessary, ie:
+      - contents have changed
+      - offset has changed
+      - hw requirements due to other packets emitted.
+   */
+   {
+      struct brw_constant_buffer cb;
+
+      memset(&cb, 0, sizeof(cb));
+
+      cb.header.opcode = CMD_CONST_BUFFER;
+      cb.header.length = sizeof(cb)/4 - 2;
+      cb.header.valid = 1;
+      cb.bits0.buffer_length = sz - 1;
+      cb.bits0.buffer_address = brw->curbe.gs_offset >> 6;
+
+      /* Because this provokes an action (ie copy the constants into the
+       * URB), it shouldn't be shortcircuited if identical to the
+       * previous time - because eg. the urb destination may have
+       * changed, or the urb contents different to last time.
+       *
+       * Note that the data referred to is actually copied internally,
+       * not just used in place according to passed pointer.
+       *
+       * It appears that the CS unit takes care of using each available
+       * URB entry (Const URB Entry == CURBE) in turn, and issuing
+       * flushes as necessary when doublebuffering of CURBEs isn't
+       * possible.
+       */
+      BRW_BATCH_STRUCT(brw, &cb);
+   }
+}
+
+/* This tracked state is unique in that the state it monitors varies
+ * dynamically depending on the parameters tracked by the fragment and
+ * vertex programs.  This is the template used as a starting point,
+ * each context will maintain a copy of this internally and update as
+ * required.
+ */
+const struct brw_tracked_state brw_constant_buffer = {
+   .dirty = {
+      .brw  = (BRW_NEW_CLIP |
+	       BRW_NEW_CONSTANTS |
+	       BRW_NEW_URB_FENCE | /* Implicit - hardware requires this, not used above */
+	       BRW_NEW_PSP | /* Implicit - hardware requires this, not used above */
+	       BRW_NEW_CURBE_OFFSETS),
+      .cache = (CACHE_NEW_WM_PROG)
+   },
+   .update = upload_constant_buffer
+};
+
diff --git a/src/gallium/drivers/i965simple/brw_defines.h b/src/gallium/drivers/i965simple/brw_defines.h
new file mode 100644
index 0000000000..9379a397f6
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_defines.h
@@ -0,0 +1,852 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#ifndef BRW_DEFINES_H
+#define BRW_DEFINES_H
+
+/*
+ */
+#define MI_NOOP                              0x00
+#define MI_USER_INTERRUPT                    0x02
+#define MI_WAIT_FOR_EVENT                    0x03
+#define MI_FLUSH                             0x04
+#define MI_REPORT_HEAD                       0x07
+#define MI_ARB_ON_OFF                        0x08
+#define MI_BATCH_BUFFER_END                  0x0A
+#define MI_OVERLAY_FLIP                      0x11
+#define MI_LOAD_SCAN_LINES_INCL              0x12
+#define MI_LOAD_SCAN_LINES_EXCL              0x13
+#define MI_DISPLAY_BUFFER_INFO               0x14
+#define MI_SET_CONTEXT                       0x18
+#define MI_STORE_DATA_IMM                    0x20
+#define MI_STORE_DATA_INDEX                  0x21
+#define MI_LOAD_REGISTER_IMM                 0x22
+#define MI_STORE_REGISTER_MEM                0x24
+#define MI_BATCH_BUFFER_START                0x31
+
+#define MI_SYNCHRONOUS_FLIP                  0x0
+#define MI_ASYNCHRONOUS_FLIP                 0x1
+
+#define MI_BUFFER_SECURE                     0x0
+#define MI_BUFFER_NONSECURE                  0x1
+
+#define MI_ARBITRATE_AT_CHAIN_POINTS         0x0
+#define MI_ARBITRATE_BETWEEN_INSTS           0x1
+#define MI_NO_ARBITRATION                    0x3
+
+#define MI_CONDITION_CODE_WAIT_DISABLED      0x0
+#define MI_CONDITION_CODE_WAIT_0             0x1
+#define MI_CONDITION_CODE_WAIT_1             0x2
+#define MI_CONDITION_CODE_WAIT_2             0x3
+#define MI_CONDITION_CODE_WAIT_3             0x4
+#define MI_CONDITION_CODE_WAIT_4             0x5
+
+#define MI_DISPLAY_PIPE_A                    0x0
+#define MI_DISPLAY_PIPE_B                    0x1
+
+#define MI_DISPLAY_PLANE_A                   0x0
+#define MI_DISPLAY_PLANE_B                   0x1
+#define MI_DISPLAY_PLANE_C                   0x2
+
+#define MI_STANDARD_FLIP                                 0x0
+#define MI_ENQUEUE_FLIP_PERFORM_BASE_FRAME_NUMBER_LOAD   0x1
+#define MI_ENQUEUE_FLIP_TARGET_FRAME_NUMBER_RELATIVE     0x2
+#define MI_ENQUEUE_FLIP_ABSOLUTE_TARGET_FRAME_NUMBER     0x3
+
+#define MI_PHYSICAL_ADDRESS                  0x0
+#define MI_VIRTUAL_ADDRESS                   0x1
+
+#define MI_BUFFER_MEMORY_MAIN                0x0
+#define MI_BUFFER_MEMORY_GTT                 0x2
+#define MI_BUFFER_MEMORY_PER_PROCESS_GTT     0x3
+
+#define MI_FLIP_CONTINUE                     0x0
+#define MI_FLIP_ON                           0x1
+#define MI_FLIP_OFF                          0x2
+
+#define MI_UNTRUSTED_REGISTER_SPACE          0x0
+#define MI_TRUSTED_REGISTER_SPACE            0x1
+
+/* 3D state:
+ */
+#define _3DOP_3DSTATE_PIPELINED       0x0
+#define _3DOP_3DSTATE_NONPIPELINED    0x1
+#define _3DOP_3DCONTROL               0x2
+#define _3DOP_3DPRIMITIVE             0x3
+
+#define _3DSTATE_PIPELINED_POINTERS       0x00
+#define _3DSTATE_BINDING_TABLE_POINTERS   0x01
+#define _3DSTATE_VERTEX_BUFFERS           0x08
+#define _3DSTATE_VERTEX_ELEMENTS          0x09
+#define _3DSTATE_INDEX_BUFFER             0x0A
+#define _3DSTATE_VF_STATISTICS            0x0B
+#define _3DSTATE_DRAWING_RECTANGLE            0x00
+#define _3DSTATE_CONSTANT_COLOR               0x01
+#define _3DSTATE_SAMPLER_PALETTE_LOAD         0x02
+#define _3DSTATE_CHROMA_KEY                   0x04
+#define _3DSTATE_DEPTH_BUFFER                 0x05
+#define _3DSTATE_POLY_STIPPLE_OFFSET          0x06
+#define _3DSTATE_POLY_STIPPLE_PATTERN         0x07
+#define _3DSTATE_LINE_STIPPLE                 0x08
+#define _3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP    0x09
+#define _3DCONTROL    0x00
+#define _3DPRIMITIVE  0x00
+
+#define PIPE_CONTROL_NOWRITE          0x00
+#define PIPE_CONTROL_WRITEIMMEDIATE   0x01
+#define PIPE_CONTROL_WRITEDEPTH       0x02
+#define PIPE_CONTROL_WRITETIMESTAMP   0x03
+
+#define PIPE_CONTROL_GTTWRITE_PROCESS_LOCAL 0x00
+#define PIPE_CONTROL_GTTWRITE_GLOBAL        0x01
+
+#define _3DPRIM_POINTLIST         0x01
+#define _3DPRIM_LINELIST          0x02
+#define _3DPRIM_LINESTRIP         0x03
+#define _3DPRIM_TRILIST           0x04
+#define _3DPRIM_TRISTRIP          0x05
+#define _3DPRIM_TRIFAN            0x06
+#define _3DPRIM_QUADLIST          0x07
+#define _3DPRIM_QUADSTRIP         0x08
+#define _3DPRIM_LINELIST_ADJ      0x09
+#define _3DPRIM_LINESTRIP_ADJ     0x0A
+#define _3DPRIM_TRILIST_ADJ       0x0B
+#define _3DPRIM_TRISTRIP_ADJ      0x0C
+#define _3DPRIM_TRISTRIP_REVERSE  0x0D
+#define _3DPRIM_POLYGON           0x0E
+#define _3DPRIM_RECTLIST          0x0F
+#define _3DPRIM_LINELOOP          0x10
+#define _3DPRIM_POINTLIST_BF      0x11
+#define _3DPRIM_LINESTRIP_CONT    0x12
+#define _3DPRIM_LINESTRIP_BF      0x13
+#define _3DPRIM_LINESTRIP_CONT_BF 0x14
+#define _3DPRIM_TRIFAN_NOSTIPPLE  0x15
+
+#define _3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL 0
+#define _3DPRIM_VERTEXBUFFER_ACCESS_RANDOM     1
+
+#define BRW_ANISORATIO_2     0
+#define BRW_ANISORATIO_4     1
+#define BRW_ANISORATIO_6     2
+#define BRW_ANISORATIO_8     3
+#define BRW_ANISORATIO_10    4
+#define BRW_ANISORATIO_12    5
+#define BRW_ANISORATIO_14    6
+#define BRW_ANISORATIO_16    7
+
+#define BRW_BLENDFACTOR_ONE                 0x1
+#define BRW_BLENDFACTOR_SRC_COLOR           0x2
+#define BRW_BLENDFACTOR_SRC_ALPHA           0x3
+#define BRW_BLENDFACTOR_DST_ALPHA           0x4
+#define BRW_BLENDFACTOR_DST_COLOR           0x5
+#define BRW_BLENDFACTOR_SRC_ALPHA_SATURATE  0x6
+#define BRW_BLENDFACTOR_CONST_COLOR         0x7
+#define BRW_BLENDFACTOR_CONST_ALPHA         0x8
+#define BRW_BLENDFACTOR_SRC1_COLOR          0x9
+#define BRW_BLENDFACTOR_SRC1_ALPHA          0x0A
+#define BRW_BLENDFACTOR_ZERO                0x11
+#define BRW_BLENDFACTOR_INV_SRC_COLOR       0x12
+#define BRW_BLENDFACTOR_INV_SRC_ALPHA       0x13
+#define BRW_BLENDFACTOR_INV_DST_ALPHA       0x14
+#define BRW_BLENDFACTOR_INV_DST_COLOR       0x15
+#define BRW_BLENDFACTOR_INV_CONST_COLOR     0x17
+#define BRW_BLENDFACTOR_INV_CONST_ALPHA     0x18
+#define BRW_BLENDFACTOR_INV_SRC1_COLOR      0x19
+#define BRW_BLENDFACTOR_INV_SRC1_ALPHA      0x1A
+
+#define BRW_BLENDFUNCTION_ADD               0
+#define BRW_BLENDFUNCTION_SUBTRACT          1
+#define BRW_BLENDFUNCTION_REVERSE_SUBTRACT  2
+#define BRW_BLENDFUNCTION_MIN               3
+#define BRW_BLENDFUNCTION_MAX               4
+
+#define BRW_ALPHATEST_FORMAT_UNORM8         0
+#define BRW_ALPHATEST_FORMAT_FLOAT32        1
+
+#define BRW_CHROMAKEY_KILL_ON_ANY_MATCH  0
+#define BRW_CHROMAKEY_REPLACE_BLACK      1
+
+#define BRW_CLIP_API_OGL     0
+#define BRW_CLIP_API_DX      1
+
+#define BRW_CLIPMODE_NORMAL              0
+#define BRW_CLIPMODE_CLIP_ALL            1
+#define BRW_CLIPMODE_CLIP_NON_REJECTED   2
+#define BRW_CLIPMODE_REJECT_ALL          3
+#define BRW_CLIPMODE_ACCEPT_ALL          4
+
+#define BRW_CLIP_NDCSPACE     0
+#define BRW_CLIP_SCREENSPACE  1
+
+#define BRW_COMPAREFUNCTION_ALWAYS       0
+#define BRW_COMPAREFUNCTION_NEVER        1
+#define BRW_COMPAREFUNCTION_LESS         2
+#define BRW_COMPAREFUNCTION_EQUAL        3
+#define BRW_COMPAREFUNCTION_LEQUAL       4
+#define BRW_COMPAREFUNCTION_GREATER      5
+#define BRW_COMPAREFUNCTION_NOTEQUAL     6
+#define BRW_COMPAREFUNCTION_GEQUAL       7
+
+#define BRW_COVERAGE_PIXELS_HALF     0
+#define BRW_COVERAGE_PIXELS_1        1
+#define BRW_COVERAGE_PIXELS_2        2
+#define BRW_COVERAGE_PIXELS_4        3
+
+#define BRW_CULLMODE_BOTH        0
+#define BRW_CULLMODE_NONE        1
+#define BRW_CULLMODE_FRONT       2
+#define BRW_CULLMODE_BACK        3
+
+#define BRW_DEFAULTCOLOR_R8G8B8A8_UNORM      0
+#define BRW_DEFAULTCOLOR_R32G32B32A32_FLOAT  1
+
+#define BRW_DEPTHFORMAT_D32_FLOAT_S8X24_UINT     0
+#define BRW_DEPTHFORMAT_D32_FLOAT                1
+#define BRW_DEPTHFORMAT_D24_UNORM_S8_UINT        2
+#define BRW_DEPTHFORMAT_D16_UNORM                5
+
+#define BRW_FLOATING_POINT_IEEE_754        0
+#define BRW_FLOATING_POINT_NON_IEEE_754    1
+
+#define BRW_FRONTWINDING_CW      0
+#define BRW_FRONTWINDING_CCW     1
+
+#define BRW_SPRITE_POINT_ENABLE  16
+
+#define BRW_INDEX_BYTE     0
+#define BRW_INDEX_WORD     1
+#define BRW_INDEX_DWORD    2
+
+#define BRW_LOGICOPFUNCTION_CLEAR            0
+#define BRW_LOGICOPFUNCTION_NOR              1
+#define BRW_LOGICOPFUNCTION_AND_INVERTED     2
+#define BRW_LOGICOPFUNCTION_COPY_INVERTED    3
+#define BRW_LOGICOPFUNCTION_AND_REVERSE      4
+#define BRW_LOGICOPFUNCTION_INVERT           5
+#define BRW_LOGICOPFUNCTION_XOR              6
+#define BRW_LOGICOPFUNCTION_NAND             7
+#define BRW_LOGICOPFUNCTION_AND              8
+#define BRW_LOGICOPFUNCTION_EQUIV            9
+#define BRW_LOGICOPFUNCTION_NOOP             10
+#define BRW_LOGICOPFUNCTION_OR_INVERTED      11
+#define BRW_LOGICOPFUNCTION_COPY             12
+#define BRW_LOGICOPFUNCTION_OR_REVERSE       13
+#define BRW_LOGICOPFUNCTION_OR               14
+#define BRW_LOGICOPFUNCTION_SET              15
+
+#define BRW_MAPFILTER_NEAREST        0x0
+#define BRW_MAPFILTER_LINEAR         0x1
+#define BRW_MAPFILTER_ANISOTROPIC    0x2
+
+#define BRW_MIPFILTER_NONE        0
+#define BRW_MIPFILTER_NEAREST     1
+#define BRW_MIPFILTER_LINEAR      3
+
+#define BRW_POLYGON_FRONT_FACING     0
+#define BRW_POLYGON_BACK_FACING      1
+
+#define BRW_PREFILTER_ALWAYS     0x0
+#define BRW_PREFILTER_NEVER      0x1
+#define BRW_PREFILTER_LESS       0x2
+#define BRW_PREFILTER_EQUAL      0x3
+#define BRW_PREFILTER_LEQUAL     0x4
+#define BRW_PREFILTER_GREATER    0x5
+#define BRW_PREFILTER_NOTEQUAL   0x6
+#define BRW_PREFILTER_GEQUAL     0x7
+
+#define BRW_PROVOKING_VERTEX_0    0
+#define BRW_PROVOKING_VERTEX_1    1
+#define BRW_PROVOKING_VERTEX_2    2
+
+#define BRW_RASTRULE_UPPER_LEFT  0
+#define BRW_RASTRULE_UPPER_RIGHT 1
+
+#define BRW_RENDERTARGET_CLAMPRANGE_UNORM    0
+#define BRW_RENDERTARGET_CLAMPRANGE_SNORM    1
+#define BRW_RENDERTARGET_CLAMPRANGE_FORMAT   2
+
+#define BRW_STENCILOP_KEEP               0
+#define BRW_STENCILOP_ZERO               1
+#define BRW_STENCILOP_REPLACE            2
+#define BRW_STENCILOP_INCRSAT            3
+#define BRW_STENCILOP_DECRSAT            4
+#define BRW_STENCILOP_INCR               5
+#define BRW_STENCILOP_DECR               6
+#define BRW_STENCILOP_INVERT             7
+
+#define BRW_SURFACE_MIPMAPLAYOUT_BELOW   0
+#define BRW_SURFACE_MIPMAPLAYOUT_RIGHT   1
+
+#define BRW_SURFACEFORMAT_R32G32B32A32_FLOAT             0x000
+#define BRW_SURFACEFORMAT_R32G32B32A32_SINT              0x001
+#define BRW_SURFACEFORMAT_R32G32B32A32_UINT              0x002
+#define BRW_SURFACEFORMAT_R32G32B32A32_UNORM             0x003
+#define BRW_SURFACEFORMAT_R32G32B32A32_SNORM             0x004
+#define BRW_SURFACEFORMAT_R64G64_FLOAT                   0x005
+#define BRW_SURFACEFORMAT_R32G32B32X32_FLOAT             0x006
+#define BRW_SURFACEFORMAT_R32G32B32A32_SSCALED           0x007
+#define BRW_SURFACEFORMAT_R32G32B32A32_USCALED           0x008
+#define BRW_SURFACEFORMAT_R32G32B32_FLOAT                0x040
+#define BRW_SURFACEFORMAT_R32G32B32_SINT                 0x041
+#define BRW_SURFACEFORMAT_R32G32B32_UINT                 0x042
+#define BRW_SURFACEFORMAT_R32G32B32_UNORM                0x043
+#define BRW_SURFACEFORMAT_R32G32B32_SNORM                0x044
+#define BRW_SURFACEFORMAT_R32G32B32_SSCALED              0x045
+#define BRW_SURFACEFORMAT_R32G32B32_USCALED              0x046
+#define BRW_SURFACEFORMAT_R16G16B16A16_UNORM             0x080
+#define BRW_SURFACEFORMAT_R16G16B16A16_SNORM             0x081
+#define BRW_SURFACEFORMAT_R16G16B16A16_SINT              0x082
+#define BRW_SURFACEFORMAT_R16G16B16A16_UINT              0x083
+#define BRW_SURFACEFORMAT_R16G16B16A16_FLOAT             0x084
+#define BRW_SURFACEFORMAT_R32G32_FLOAT                   0x085
+#define BRW_SURFACEFORMAT_R32G32_SINT                    0x086
+#define BRW_SURFACEFORMAT_R32G32_UINT                    0x087
+#define BRW_SURFACEFORMAT_R32_FLOAT_X8X24_TYPELESS       0x088
+#define BRW_SURFACEFORMAT_X32_TYPELESS_G8X24_UINT        0x089
+#define BRW_SURFACEFORMAT_L32A32_FLOAT                   0x08A
+#define BRW_SURFACEFORMAT_R32G32_UNORM                   0x08B
+#define BRW_SURFACEFORMAT_R32G32_SNORM                   0x08C
+#define BRW_SURFACEFORMAT_R64_FLOAT                      0x08D
+#define BRW_SURFACEFORMAT_R16G16B16X16_UNORM             0x08E
+#define BRW_SURFACEFORMAT_R16G16B16X16_FLOAT             0x08F
+#define BRW_SURFACEFORMAT_A32X32_FLOAT                   0x090
+#define BRW_SURFACEFORMAT_L32X32_FLOAT                   0x091
+#define BRW_SURFACEFORMAT_I32X32_FLOAT                   0x092
+#define BRW_SURFACEFORMAT_R16G16B16A16_SSCALED           0x093
+#define BRW_SURFACEFORMAT_R16G16B16A16_USCALED           0x094
+#define BRW_SURFACEFORMAT_R32G32_SSCALED                 0x095
+#define BRW_SURFACEFORMAT_R32G32_USCALED                 0x096
+#define BRW_SURFACEFORMAT_B8G8R8A8_UNORM                 0x0C0
+#define BRW_SURFACEFORMAT_B8G8R8A8_UNORM_SRGB            0x0C1
+#define BRW_SURFACEFORMAT_R10G10B10A2_UNORM              0x0C2
+#define BRW_SURFACEFORMAT_R10G10B10A2_UNORM_SRGB         0x0C3
+#define BRW_SURFACEFORMAT_R10G10B10A2_UINT               0x0C4
+#define BRW_SURFACEFORMAT_R10G10B10_SNORM_A2_UNORM       0x0C5
+#define BRW_SURFACEFORMAT_R8G8B8A8_UNORM                 0x0C7
+#define BRW_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB            0x0C8
+#define BRW_SURFACEFORMAT_R8G8B8A8_SNORM                 0x0C9
+#define BRW_SURFACEFORMAT_R8G8B8A8_SINT                  0x0CA
+#define BRW_SURFACEFORMAT_R8G8B8A8_UINT                  0x0CB
+#define BRW_SURFACEFORMAT_R16G16_UNORM                   0x0CC
+#define BRW_SURFACEFORMAT_R16G16_SNORM                   0x0CD
+#define BRW_SURFACEFORMAT_R16G16_SINT                    0x0CE
+#define BRW_SURFACEFORMAT_R16G16_UINT                    0x0CF
+#define BRW_SURFACEFORMAT_R16G16_FLOAT                   0x0D0
+#define BRW_SURFACEFORMAT_B10G10R10A2_UNORM              0x0D1
+#define BRW_SURFACEFORMAT_B10G10R10A2_UNORM_SRGB         0x0D2
+#define BRW_SURFACEFORMAT_R11G11B10_FLOAT                0x0D3
+#define BRW_SURFACEFORMAT_R32_SINT                       0x0D6
+#define BRW_SURFACEFORMAT_R32_UINT                       0x0D7
+#define BRW_SURFACEFORMAT_R32_FLOAT                      0x0D8
+#define BRW_SURFACEFORMAT_R24_UNORM_X8_TYPELESS          0x0D9
+#define BRW_SURFACEFORMAT_X24_TYPELESS_G8_UINT           0x0DA
+#define BRW_SURFACEFORMAT_L16A16_UNORM                   0x0DF
+#define BRW_SURFACEFORMAT_I24X8_UNORM                    0x0E0
+#define BRW_SURFACEFORMAT_L24X8_UNORM                    0x0E1
+#define BRW_SURFACEFORMAT_A24X8_UNORM                    0x0E2
+#define BRW_SURFACEFORMAT_I32_FLOAT                      0x0E3
+#define BRW_SURFACEFORMAT_L32_FLOAT                      0x0E4
+#define BRW_SURFACEFORMAT_A32_FLOAT                      0x0E5
+#define BRW_SURFACEFORMAT_B8G8R8X8_UNORM                 0x0E9
+#define BRW_SURFACEFORMAT_B8G8R8X8_UNORM_SRGB            0x0EA
+#define BRW_SURFACEFORMAT_R8G8B8X8_UNORM                 0x0EB
+#define BRW_SURFACEFORMAT_R8G8B8X8_UNORM_SRGB            0x0EC
+#define BRW_SURFACEFORMAT_R9G9B9E5_SHAREDEXP             0x0ED
+#define BRW_SURFACEFORMAT_B10G10R10X2_UNORM              0x0EE
+#define BRW_SURFACEFORMAT_L16A16_FLOAT                   0x0F0
+#define BRW_SURFACEFORMAT_R32_UNORM                      0x0F1
+#define BRW_SURFACEFORMAT_R32_SNORM                      0x0F2
+#define BRW_SURFACEFORMAT_R10G10B10X2_USCALED            0x0F3
+#define BRW_SURFACEFORMAT_R8G8B8A8_SSCALED               0x0F4
+#define BRW_SURFACEFORMAT_R8G8B8A8_USCALED               0x0F5
+#define BRW_SURFACEFORMAT_R16G16_SSCALED                 0x0F6
+#define BRW_SURFACEFORMAT_R16G16_USCALED                 0x0F7
+#define BRW_SURFACEFORMAT_R32_SSCALED                    0x0F8
+#define BRW_SURFACEFORMAT_R32_USCALED                    0x0F9
+#define BRW_SURFACEFORMAT_B5G6R5_UNORM                   0x100
+#define BRW_SURFACEFORMAT_B5G6R5_UNORM_SRGB              0x101
+#define BRW_SURFACEFORMAT_B5G5R5A1_UNORM                 0x102
+#define BRW_SURFACEFORMAT_B5G5R5A1_UNORM_SRGB            0x103
+#define BRW_SURFACEFORMAT_B4G4R4A4_UNORM                 0x104
+#define BRW_SURFACEFORMAT_B4G4R4A4_UNORM_SRGB            0x105
+#define BRW_SURFACEFORMAT_R8G8_UNORM                     0x106
+#define BRW_SURFACEFORMAT_R8G8_SNORM                     0x107
+#define BRW_SURFACEFORMAT_R8G8_SINT                      0x108
+#define BRW_SURFACEFORMAT_R8G8_UINT                      0x109
+#define BRW_SURFACEFORMAT_R16_UNORM                      0x10A
+#define BRW_SURFACEFORMAT_R16_SNORM                      0x10B
+#define BRW_SURFACEFORMAT_R16_SINT                       0x10C
+#define BRW_SURFACEFORMAT_R16_UINT                       0x10D
+#define BRW_SURFACEFORMAT_R16_FLOAT                      0x10E
+#define BRW_SURFACEFORMAT_I16_UNORM                      0x111
+#define BRW_SURFACEFORMAT_L16_UNORM                      0x112
+#define BRW_SURFACEFORMAT_A16_UNORM                      0x113
+#define BRW_SURFACEFORMAT_L8A8_UNORM                     0x114
+#define BRW_SURFACEFORMAT_I16_FLOAT                      0x115
+#define BRW_SURFACEFORMAT_L16_FLOAT                      0x116
+#define BRW_SURFACEFORMAT_A16_FLOAT                      0x117
+#define BRW_SURFACEFORMAT_R5G5_SNORM_B6_UNORM            0x119
+#define BRW_SURFACEFORMAT_B5G5R5X1_UNORM                 0x11A
+#define BRW_SURFACEFORMAT_B5G5R5X1_UNORM_SRGB            0x11B
+#define BRW_SURFACEFORMAT_R8G8_SSCALED                   0x11C
+#define BRW_SURFACEFORMAT_R8G8_USCALED                   0x11D
+#define BRW_SURFACEFORMAT_R16_SSCALED                    0x11E
+#define BRW_SURFACEFORMAT_R16_USCALED                    0x11F
+#define BRW_SURFACEFORMAT_R8_UNORM                       0x140
+#define BRW_SURFACEFORMAT_R8_SNORM                       0x141
+#define BRW_SURFACEFORMAT_R8_SINT                        0x142
+#define BRW_SURFACEFORMAT_R8_UINT                        0x143
+#define BRW_SURFACEFORMAT_A8_UNORM                       0x144
+#define BRW_SURFACEFORMAT_I8_UNORM                       0x145
+#define BRW_SURFACEFORMAT_L8_UNORM                       0x146
+#define BRW_SURFACEFORMAT_P4A4_UNORM                     0x147
+#define BRW_SURFACEFORMAT_A4P4_UNORM                     0x148
+#define BRW_SURFACEFORMAT_R8_SSCALED                     0x149
+#define BRW_SURFACEFORMAT_R8_USCALED                     0x14A
+#define BRW_SURFACEFORMAT_R1_UINT                        0x181
+#define BRW_SURFACEFORMAT_YCRCB_NORMAL                   0x182
+#define BRW_SURFACEFORMAT_YCRCB_SWAPUVY                  0x183
+#define BRW_SURFACEFORMAT_BC1_UNORM                      0x186
+#define BRW_SURFACEFORMAT_BC2_UNORM                      0x187
+#define BRW_SURFACEFORMAT_BC3_UNORM                      0x188
+#define BRW_SURFACEFORMAT_BC4_UNORM                      0x189
+#define BRW_SURFACEFORMAT_BC5_UNORM                      0x18A
+#define BRW_SURFACEFORMAT_BC1_UNORM_SRGB                 0x18B
+#define BRW_SURFACEFORMAT_BC2_UNORM_SRGB                 0x18C
+#define BRW_SURFACEFORMAT_BC3_UNORM_SRGB                 0x18D
+#define BRW_SURFACEFORMAT_MONO8                          0x18E
+#define BRW_SURFACEFORMAT_YCRCB_SWAPUV                   0x18F
+#define BRW_SURFACEFORMAT_YCRCB_SWAPY                    0x190
+#define BRW_SURFACEFORMAT_DXT1_RGB                       0x191
+#define BRW_SURFACEFORMAT_FXT1                           0x192
+#define BRW_SURFACEFORMAT_R8G8B8_UNORM                   0x193
+#define BRW_SURFACEFORMAT_R8G8B8_SNORM                   0x194
+#define BRW_SURFACEFORMAT_R8G8B8_SSCALED                 0x195
+#define BRW_SURFACEFORMAT_R8G8B8_USCALED                 0x196
+#define BRW_SURFACEFORMAT_R64G64B64A64_FLOAT             0x197
+#define BRW_SURFACEFORMAT_R64G64B64_FLOAT                0x198
+#define BRW_SURFACEFORMAT_BC4_SNORM                      0x199
+#define BRW_SURFACEFORMAT_BC5_SNORM                      0x19A
+#define BRW_SURFACEFORMAT_R16G16B16_UNORM                0x19C
+#define BRW_SURFACEFORMAT_R16G16B16_SNORM                0x19D
+#define BRW_SURFACEFORMAT_R16G16B16_SSCALED              0x19E
+#define BRW_SURFACEFORMAT_R16G16B16_USCALED              0x19F
+
+#define BRW_SURFACERETURNFORMAT_FLOAT32  0
+#define BRW_SURFACERETURNFORMAT_S1       1
+
+#define BRW_SURFACE_1D      0
+#define BRW_SURFACE_2D      1
+#define BRW_SURFACE_3D      2
+#define BRW_SURFACE_CUBE    3
+#define BRW_SURFACE_BUFFER  4
+#define BRW_SURFACE_NULL    7
+
+#define BRW_TEXCOORDMODE_WRAP            0
+#define BRW_TEXCOORDMODE_MIRROR          1
+#define BRW_TEXCOORDMODE_CLAMP           2
+#define BRW_TEXCOORDMODE_CUBE            3
+#define BRW_TEXCOORDMODE_CLAMP_BORDER    4
+#define BRW_TEXCOORDMODE_MIRROR_ONCE     5
+
+#define BRW_THREAD_PRIORITY_NORMAL   0
+#define BRW_THREAD_PRIORITY_HIGH     1
+
+#define BRW_TILEWALK_XMAJOR                 0
+#define BRW_TILEWALK_YMAJOR                 1
+
+#define BRW_VERTEX_SUBPIXEL_PRECISION_8BITS  0
+#define BRW_VERTEX_SUBPIXEL_PRECISION_4BITS  1
+
+#define BRW_VERTEXBUFFER_ACCESS_VERTEXDATA     0
+#define BRW_VERTEXBUFFER_ACCESS_INSTANCEDATA   1
+
+#define BRW_VFCOMPONENT_NOSTORE      0
+#define BRW_VFCOMPONENT_STORE_SRC    1
+#define BRW_VFCOMPONENT_STORE_0      2
+#define BRW_VFCOMPONENT_STORE_1_FLT  3
+#define BRW_VFCOMPONENT_STORE_1_INT  4
+#define BRW_VFCOMPONENT_STORE_VID    5
+#define BRW_VFCOMPONENT_STORE_IID    6
+#define BRW_VFCOMPONENT_STORE_PID    7
+
+
+
+/* Execution Unit (EU) defines
+ */
+
+#define BRW_ALIGN_1   0
+#define BRW_ALIGN_16  1
+
+#define BRW_ADDRESS_DIRECT                        0
+#define BRW_ADDRESS_REGISTER_INDIRECT_REGISTER    1
+
+#define BRW_CHANNEL_X     0
+#define BRW_CHANNEL_Y     1
+#define BRW_CHANNEL_Z     2
+#define BRW_CHANNEL_W     3
+
+#define BRW_COMPRESSION_NONE          0
+#define BRW_COMPRESSION_2NDHALF       1
+#define BRW_COMPRESSION_COMPRESSED    2
+
+#define BRW_CONDITIONAL_NONE  0
+#define BRW_CONDITIONAL_Z     1
+#define BRW_CONDITIONAL_NZ    2
+#define BRW_CONDITIONAL_EQ    1	/* Z */
+#define BRW_CONDITIONAL_NEQ   2	/* NZ */
+#define BRW_CONDITIONAL_G     3
+#define BRW_CONDITIONAL_GE    4
+#define BRW_CONDITIONAL_L     5
+#define BRW_CONDITIONAL_LE    6
+#define BRW_CONDITIONAL_C     7
+#define BRW_CONDITIONAL_O     8
+
+#define BRW_DEBUG_NONE        0
+#define BRW_DEBUG_BREAKPOINT  1
+
+#define BRW_DEPENDENCY_NORMAL         0
+#define BRW_DEPENDENCY_NOTCLEARED     1
+#define BRW_DEPENDENCY_NOTCHECKED     2
+#define BRW_DEPENDENCY_DISABLE        3
+
+#define BRW_EXECUTE_1     0
+#define BRW_EXECUTE_2     1
+#define BRW_EXECUTE_4     2
+#define BRW_EXECUTE_8     3
+#define BRW_EXECUTE_16    4
+#define BRW_EXECUTE_32    5
+
+#define BRW_HORIZONTAL_STRIDE_0   0
+#define BRW_HORIZONTAL_STRIDE_1   1
+#define BRW_HORIZONTAL_STRIDE_2   2
+#define BRW_HORIZONTAL_STRIDE_4   3
+
+#define BRW_INSTRUCTION_NORMAL    0
+#define BRW_INSTRUCTION_SATURATE  1
+
+#define BRW_MASK_ENABLE   0
+#define BRW_MASK_DISABLE  1
+
+#define BRW_OPCODE_MOV        1
+#define BRW_OPCODE_SEL        2
+#define BRW_OPCODE_NOT        4
+#define BRW_OPCODE_AND        5
+#define BRW_OPCODE_OR         6
+#define BRW_OPCODE_XOR        7
+#define BRW_OPCODE_SHR        8
+#define BRW_OPCODE_SHL        9
+#define BRW_OPCODE_RSR        10
+#define BRW_OPCODE_RSL        11
+#define BRW_OPCODE_ASR        12
+#define BRW_OPCODE_CMP        16
+#define BRW_OPCODE_JMPI       32
+#define BRW_OPCODE_IF         34
+#define BRW_OPCODE_IFF        35
+#define BRW_OPCODE_ELSE       36
+#define BRW_OPCODE_ENDIF      37
+#define BRW_OPCODE_DO         38
+#define BRW_OPCODE_WHILE      39
+#define BRW_OPCODE_BREAK      40
+#define BRW_OPCODE_CONTINUE   41
+#define BRW_OPCODE_HALT       42
+#define BRW_OPCODE_MSAVE      44
+#define BRW_OPCODE_MRESTORE   45
+#define BRW_OPCODE_PUSH       46
+#define BRW_OPCODE_POP        47
+#define BRW_OPCODE_WAIT       48
+#define BRW_OPCODE_SEND       49
+#define BRW_OPCODE_ADD        64
+#define BRW_OPCODE_MUL        65
+#define BRW_OPCODE_AVG        66
+#define BRW_OPCODE_FRC        67
+#define BRW_OPCODE_RNDU       68
+#define BRW_OPCODE_RNDD       69
+#define BRW_OPCODE_RNDE       70
+#define BRW_OPCODE_RNDZ       71
+#define BRW_OPCODE_MAC        72
+#define BRW_OPCODE_MACH       73
+#define BRW_OPCODE_LZD        74
+#define BRW_OPCODE_SAD2       80
+#define BRW_OPCODE_SADA2      81
+#define BRW_OPCODE_DP4        84
+#define BRW_OPCODE_DPH        85
+#define BRW_OPCODE_DP3        86
+#define BRW_OPCODE_DP2        87
+#define BRW_OPCODE_DPA2       88
+#define BRW_OPCODE_LINE       89
+#define BRW_OPCODE_NOP        126
+
+#define BRW_PREDICATE_NONE             0
+#define BRW_PREDICATE_NORMAL           1
+#define BRW_PREDICATE_ALIGN1_ANYV             2
+#define BRW_PREDICATE_ALIGN1_ALLV             3
+#define BRW_PREDICATE_ALIGN1_ANY2H            4
+#define BRW_PREDICATE_ALIGN1_ALL2H            5
+#define BRW_PREDICATE_ALIGN1_ANY4H            6
+#define BRW_PREDICATE_ALIGN1_ALL4H            7
+#define BRW_PREDICATE_ALIGN1_ANY8H            8
+#define BRW_PREDICATE_ALIGN1_ALL8H            9
+#define BRW_PREDICATE_ALIGN1_ANY16H           10
+#define BRW_PREDICATE_ALIGN1_ALL16H           11
+#define BRW_PREDICATE_ALIGN16_REPLICATE_X     2
+#define BRW_PREDICATE_ALIGN16_REPLICATE_Y     3
+#define BRW_PREDICATE_ALIGN16_REPLICATE_Z     4
+#define BRW_PREDICATE_ALIGN16_REPLICATE_W     5
+#define BRW_PREDICATE_ALIGN16_ANY4H           6
+#define BRW_PREDICATE_ALIGN16_ALL4H           7
+
+#define BRW_ARCHITECTURE_REGISTER_FILE    0
+#define BRW_GENERAL_REGISTER_FILE         1
+#define BRW_MESSAGE_REGISTER_FILE         2
+#define BRW_IMMEDIATE_VALUE               3
+
+#define BRW_REGISTER_TYPE_UD  0
+#define BRW_REGISTER_TYPE_D   1
+#define BRW_REGISTER_TYPE_UW  2
+#define BRW_REGISTER_TYPE_W   3
+#define BRW_REGISTER_TYPE_UB  4
+#define BRW_REGISTER_TYPE_B   5
+#define BRW_REGISTER_TYPE_VF  5	/* packed float vector, immediates only? */
+#define BRW_REGISTER_TYPE_HF  6
+#define BRW_REGISTER_TYPE_V   6	/* packed int vector, immediates only, uword dest only */
+#define BRW_REGISTER_TYPE_F   7
+
+#define BRW_ARF_NULL                  0x00
+#define BRW_ARF_ADDRESS               0x10
+#define BRW_ARF_ACCUMULATOR           0x20
+#define BRW_ARF_FLAG                  0x30
+#define BRW_ARF_MASK                  0x40
+#define BRW_ARF_MASK_STACK            0x50
+#define BRW_ARF_MASK_STACK_DEPTH      0x60
+#define BRW_ARF_STATE                 0x70
+#define BRW_ARF_CONTROL               0x80
+#define BRW_ARF_NOTIFICATION_COUNT    0x90
+#define BRW_ARF_IP                    0xA0
+
+#define BRW_AMASK   0
+#define BRW_IMASK   1
+#define BRW_LMASK   2
+#define BRW_CMASK   3
+
+
+
+#define BRW_THREAD_NORMAL     0
+#define BRW_THREAD_ATOMIC     1
+#define BRW_THREAD_SWITCH     2
+
+#define BRW_VERTICAL_STRIDE_0                 0
+#define BRW_VERTICAL_STRIDE_1                 1
+#define BRW_VERTICAL_STRIDE_2                 2
+#define BRW_VERTICAL_STRIDE_4                 3
+#define BRW_VERTICAL_STRIDE_8                 4
+#define BRW_VERTICAL_STRIDE_16                5
+#define BRW_VERTICAL_STRIDE_32                6
+#define BRW_VERTICAL_STRIDE_64                7
+#define BRW_VERTICAL_STRIDE_128               8
+#define BRW_VERTICAL_STRIDE_256               9
+#define BRW_VERTICAL_STRIDE_ONE_DIMENSIONAL   0xF
+
+#define BRW_WIDTH_1       0
+#define BRW_WIDTH_2       1
+#define BRW_WIDTH_4       2
+#define BRW_WIDTH_8       3
+#define BRW_WIDTH_16      4
+
+#define BRW_STATELESS_BUFFER_BOUNDARY_1K      0
+#define BRW_STATELESS_BUFFER_BOUNDARY_2K      1
+#define BRW_STATELESS_BUFFER_BOUNDARY_4K      2
+#define BRW_STATELESS_BUFFER_BOUNDARY_8K      3
+#define BRW_STATELESS_BUFFER_BOUNDARY_16K     4
+#define BRW_STATELESS_BUFFER_BOUNDARY_32K     5
+#define BRW_STATELESS_BUFFER_BOUNDARY_64K     6
+#define BRW_STATELESS_BUFFER_BOUNDARY_128K    7
+#define BRW_STATELESS_BUFFER_BOUNDARY_256K    8
+#define BRW_STATELESS_BUFFER_BOUNDARY_512K    9
+#define BRW_STATELESS_BUFFER_BOUNDARY_1M      10
+#define BRW_STATELESS_BUFFER_BOUNDARY_2M      11
+
+#define BRW_POLYGON_FACING_FRONT      0
+#define BRW_POLYGON_FACING_BACK       1
+
+#define BRW_MESSAGE_TARGET_NULL               0
+#define BRW_MESSAGE_TARGET_MATH               1
+#define BRW_MESSAGE_TARGET_SAMPLER            2
+#define BRW_MESSAGE_TARGET_GATEWAY            3
+#define BRW_MESSAGE_TARGET_DATAPORT_READ      4
+#define BRW_MESSAGE_TARGET_DATAPORT_WRITE     5
+#define BRW_MESSAGE_TARGET_URB                6
+#define BRW_MESSAGE_TARGET_THREAD_SPAWNER     7
+
+#define BRW_SAMPLER_RETURN_FORMAT_FLOAT32     0
+#define BRW_SAMPLER_RETURN_FORMAT_UINT32      2
+#define BRW_SAMPLER_RETURN_FORMAT_SINT32      3
+
+#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE              0
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE             0
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS        0
+#define BRW_SAMPLER_MESSAGE_SIMD8_KILLPIX             1
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD        1
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD         1
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_GRADIENTS  2
+#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS    2
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE    0
+#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE     2
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_RESINFO           2
+#define BRW_SAMPLER_MESSAGE_SIMD8_RESINFO             2
+#define BRW_SAMPLER_MESSAGE_SIMD16_RESINFO            2
+#define BRW_SAMPLER_MESSAGE_SIMD4X2_LD                3
+#define BRW_SAMPLER_MESSAGE_SIMD8_LD                  3
+#define BRW_SAMPLER_MESSAGE_SIMD16_LD                 3
+
+#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDLOW   0
+#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDHIGH  1
+#define BRW_DATAPORT_OWORD_BLOCK_2_OWORDS     2
+#define BRW_DATAPORT_OWORD_BLOCK_4_OWORDS     3
+#define BRW_DATAPORT_OWORD_BLOCK_8_OWORDS     4
+
+#define BRW_DATAPORT_OWORD_DUAL_BLOCK_1OWORD     0
+#define BRW_DATAPORT_OWORD_DUAL_BLOCK_4OWORDS    2
+
+#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS   2
+#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_16DWORDS  3
+
+#define BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ          0
+#define BRW_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ     1
+#define BRW_DATAPORT_READ_MESSAGE_DWORD_BLOCK_READ          2
+#define BRW_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ      3
+
+#define BRW_DATAPORT_READ_TARGET_DATA_CACHE      0
+#define BRW_DATAPORT_READ_TARGET_RENDER_CACHE    1
+#define BRW_DATAPORT_READ_TARGET_SAMPLER_CACHE   2
+
+#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE                0
+#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE_REPLICATED     1
+#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01         2
+#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN23         3
+#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01       4
+
+#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE                0
+#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE           1
+#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_BLOCK_WRITE                2
+#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_SCATTERED_WRITE            3
+#define BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE              4
+#define BRW_DATAPORT_WRITE_MESSAGE_STREAMED_VERTEX_BUFFER_WRITE     5
+#define BRW_DATAPORT_WRITE_MESSAGE_FLUSH_RENDER_CACHE               7
+
+#define BRW_MATH_FUNCTION_INV                              1
+#define BRW_MATH_FUNCTION_LOG                              2
+#define BRW_MATH_FUNCTION_EXP                              3
+#define BRW_MATH_FUNCTION_SQRT                             4
+#define BRW_MATH_FUNCTION_RSQ                              5
+#define BRW_MATH_FUNCTION_SIN                              6 /* was 7 */
+#define BRW_MATH_FUNCTION_COS                              7 /* was 8 */
+#define BRW_MATH_FUNCTION_SINCOS                           8 /* was 6 */
+#define BRW_MATH_FUNCTION_TAN                              9
+#define BRW_MATH_FUNCTION_POW                              10
+#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER   11
+#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT                 12
+#define BRW_MATH_FUNCTION_INT_DIV_REMAINDER                13
+
+#define BRW_MATH_INTEGER_UNSIGNED     0
+#define BRW_MATH_INTEGER_SIGNED       1
+
+#define BRW_MATH_PRECISION_FULL        0
+#define BRW_MATH_PRECISION_PARTIAL     1
+
+#define BRW_MATH_SATURATE_NONE         0
+#define BRW_MATH_SATURATE_SATURATE     1
+
+#define BRW_MATH_DATA_VECTOR  0
+#define BRW_MATH_DATA_SCALAR  1
+
+#define BRW_URB_OPCODE_WRITE  0
+
+#define BRW_URB_SWIZZLE_NONE          0
+#define BRW_URB_SWIZZLE_INTERLEAVE    1
+#define BRW_URB_SWIZZLE_TRANSPOSE     2
+
+#define BRW_SCRATCH_SPACE_SIZE_1K     0
+#define BRW_SCRATCH_SPACE_SIZE_2K     1
+#define BRW_SCRATCH_SPACE_SIZE_4K     2
+#define BRW_SCRATCH_SPACE_SIZE_8K     3
+#define BRW_SCRATCH_SPACE_SIZE_16K    4
+#define BRW_SCRATCH_SPACE_SIZE_32K    5
+#define BRW_SCRATCH_SPACE_SIZE_64K    6
+#define BRW_SCRATCH_SPACE_SIZE_128K   7
+#define BRW_SCRATCH_SPACE_SIZE_256K   8
+#define BRW_SCRATCH_SPACE_SIZE_512K   9
+#define BRW_SCRATCH_SPACE_SIZE_1M     10
+#define BRW_SCRATCH_SPACE_SIZE_2M     11
+
+
+
+
+#define CMD_URB_FENCE                 0x6000
+#define CMD_CONST_BUFFER_STATE        0x6001
+#define CMD_CONST_BUFFER              0x6002
+
+#define CMD_STATE_BASE_ADDRESS        0x6101
+#define CMD_STATE_INSN_POINTER        0x6102
+#define CMD_PIPELINE_SELECT           0x6104
+
+#define CMD_PIPELINED_STATE_POINTERS  0x7800
+#define CMD_BINDING_TABLE_PTRS        0x7801
+#define CMD_VERTEX_BUFFER             0x7808
+#define CMD_VERTEX_ELEMENT            0x7809
+#define CMD_INDEX_BUFFER              0x780a
+#define CMD_VF_STATISTICS             0x780b
+
+#define CMD_DRAW_RECT                 0x7900
+#define CMD_BLEND_CONSTANT_COLOR      0x7901
+#define CMD_CHROMA_KEY                0x7904
+#define CMD_DEPTH_BUFFER              0x7905
+#define CMD_POLY_STIPPLE_OFFSET       0x7906
+#define CMD_POLY_STIPPLE_PATTERN      0x7907
+#define CMD_LINE_STIPPLE_PATTERN      0x7908
+#define CMD_GLOBAL_DEPTH_OFFSET_CLAMP 0x7909
+
+#define CMD_PIPE_CONTROL              0x7a00
+
+#define CMD_3D_PRIM                   0x7b00
+
+#define CMD_MI_FLUSH                  0x0200
+
+
+/* Various values from the R0 vertex header:
+ */
+#define R02_PRIM_END    0x1
+#define R02_PRIM_START  0x2
+
+
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_draw.c b/src/gallium/drivers/i965simple/brw_draw.c
new file mode 100644
index 0000000000..648aaa0da5
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_draw.c
@@ -0,0 +1,239 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <stdlib.h>
+
+#include "brw_batch.h"
+#include "brw_draw.h"
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_state.h"
+
+#include "pipe/p_context.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+static unsigned hw_prim[PIPE_PRIM_POLYGON+1] = {
+   _3DPRIM_POINTLIST,
+   _3DPRIM_LINELIST,
+   _3DPRIM_LINELOOP,
+   _3DPRIM_LINESTRIP,
+   _3DPRIM_TRILIST,
+   _3DPRIM_TRISTRIP,
+   _3DPRIM_TRIFAN,
+   _3DPRIM_QUADLIST,
+   _3DPRIM_QUADSTRIP,
+   _3DPRIM_POLYGON
+};
+
+
+static const int reduced_prim[PIPE_PRIM_POLYGON+1] = {
+   PIPE_PRIM_POINTS,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES
+};
+
+
+/* When the primitive changes, set a state bit and re-validate.  Not
+ * the nicest and would rather deal with this by having all the
+ * programs be immune to the active primitive (ie. cope with all
+ * possibilities).  That may not be realistic however.
+ */
+static void brw_set_prim(struct brw_context *brw, int prim)
+{
+   PRINT("PRIM: %d\n", prim);
+
+   /* Slight optimization to avoid the GS program when not needed:
+    */
+   if (prim == PIPE_PRIM_QUAD_STRIP &&
+       brw->attribs.Raster->flatshade &&
+       brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_FILL &&
+       brw->attribs.Raster->fill_ccw == PIPE_POLYGON_MODE_FILL)
+      prim = PIPE_PRIM_TRIANGLE_STRIP;
+
+   if (prim != brw->primitive) {
+      brw->primitive = prim;
+      brw->state.dirty.brw |= BRW_NEW_PRIMITIVE;
+
+      if (reduced_prim[prim] != brw->reduced_primitive) {
+	 brw->reduced_primitive = reduced_prim[prim];
+	 brw->state.dirty.brw |= BRW_NEW_REDUCED_PRIMITIVE;
+      }
+
+      brw_validate_state(brw);
+   }
+
+}
+
+
+static unsigned trim(int prim, unsigned length)
+{
+   if (prim == PIPE_PRIM_QUAD_STRIP)
+      return length > 3 ? (length - length % 2) : 0;
+   else if (prim == PIPE_PRIM_QUADS)
+      return length - length % 4;
+   else
+      return length;
+}
+
+
+
+static boolean brw_emit_prim( struct brw_context *brw,
+			      boolean indexed,
+			      unsigned start,
+			      unsigned count )
+
+{
+   struct brw_3d_primitive prim_packet;
+
+   if (BRW_DEBUG & DEBUG_PRIMS)
+      PRINT("PRIM: %d %d %d\n",  brw->primitive, start, count);
+
+   prim_packet.header.opcode = CMD_3D_PRIM;
+   prim_packet.header.length = sizeof(prim_packet)/4 - 2;
+   prim_packet.header.pad = 0;
+   prim_packet.header.topology = hw_prim[brw->primitive];
+   prim_packet.header.indexed = indexed;
+
+   prim_packet.verts_per_instance = trim(brw->primitive, count);
+   prim_packet.start_vert_location = start;
+   prim_packet.instance_count = 1;
+   prim_packet.start_instance_location = 0;
+   prim_packet.base_vert_location = 0;
+
+   if (prim_packet.verts_per_instance == 0)
+      return TRUE;
+
+   return brw_batchbuffer_data( brw->winsys,
+                                &prim_packet,
+                                sizeof(prim_packet) );
+}
+
+
+/* May fail if out of video memory for texture or vbo upload, or on
+ * fallback conditions.
+ */
+static boolean brw_try_draw_elements( struct pipe_context *pipe,
+				      struct pipe_buffer *index_buffer,
+				      unsigned index_size,
+				      unsigned mode,
+				      unsigned start,
+				      unsigned count )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   /* Set the first primitive ahead of validate_state:
+    */
+   brw_set_prim(brw, mode);
+
+   /* Upload index, vertex data:
+    */
+   if (index_buffer &&
+       !brw_upload_indices( brw, index_buffer, index_size, start, count ))
+      return FALSE;
+
+   if (!brw_upload_vertex_buffers(brw))
+      return FALSE;
+
+   if (!brw_upload_vertex_elements( brw ))
+      return FALSE;
+
+   /* XXX:  Need to separate validate and upload of state.
+    */
+   if (brw->state.dirty.brw)
+      brw_validate_state( brw );
+
+   if (!brw_emit_prim(brw, index_buffer != NULL,
+                      start, count))
+      return FALSE;
+
+   return TRUE;
+}
+
+
+
+static boolean brw_draw_elements( struct pipe_context *pipe,
+				  struct pipe_buffer *indexBuffer,
+				  unsigned indexSize,
+				  unsigned mode,
+				  unsigned start,
+				  unsigned count )
+{
+   if (!brw_try_draw_elements( pipe,
+			       indexBuffer,
+			       indexSize,
+			       mode, start, count ))
+   {
+      /* flush ? */
+
+      if (!brw_try_draw_elements( pipe,
+				  indexBuffer,
+				  indexSize,
+				  mode, start,
+				  count )) {
+	 assert(0);
+	 return FALSE;
+      }
+   }
+
+   return TRUE;
+}
+
+
+
+static boolean brw_draw_arrays( struct pipe_context *pipe,
+				    unsigned mode,
+				    unsigned start,
+				    unsigned count )
+{
+   if (!brw_try_draw_elements( pipe, NULL, 0, mode, start, count )) {
+      /* flush ? */
+
+      if (!brw_try_draw_elements( pipe, NULL, 0, mode, start, count )) {
+	 assert(0);
+	 return FALSE;
+      }
+   }
+   
+   return TRUE;
+}
+
+
+
+void brw_init_draw_functions( struct brw_context *brw )
+{
+   brw->pipe.draw_arrays = brw_draw_arrays;
+   brw->pipe.draw_elements = brw_draw_elements;
+}
+
+
diff --git a/src/gallium/drivers/i965simple/brw_draw.h b/src/gallium/drivers/i965simple/brw_draw.h
new file mode 100644
index 0000000000..62fe0d5d0e
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_draw.h
@@ -0,0 +1,55 @@
+ /**************************************************************************
+ * 
+ * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef BRW_DRAW_H
+#define BRW_DRAW_H
+
+#include "pipe/p_context.h"
+
+struct brw_context;
+
+
+
+void brw_init_draw_functions( struct brw_context *brw );
+
+
+boolean brw_upload_vertices( struct brw_context *brw,
+			       unsigned min_index,
+			       unsigned max_index );
+
+boolean brw_upload_indices(struct brw_context *brw,
+                           const struct pipe_buffer *index_buffer,
+                           int ib_size, int start, int count);
+
+boolean brw_upload_vertex_buffers( struct brw_context *brw );
+boolean brw_upload_vertex_elements( struct brw_context *brw );
+
+unsigned brw_translate_surface_format( unsigned id );
+
+
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_draw_upload.c b/src/gallium/drivers/i965simple/brw_draw_upload.c
new file mode 100644
index 0000000000..2d9ca3f2ea
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_draw_upload.c
@@ -0,0 +1,300 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <stdlib.h>
+
+#include "brw_batch.h"
+#include "brw_draw.h"
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_state.h"
+
+
+struct brw_array_state {
+   union header_union header;
+
+   struct {
+      union {
+	 struct {
+	    unsigned pitch:11;
+	    unsigned pad:15;
+	    unsigned access_type:1;
+	    unsigned vb_index:5;
+	 } bits;
+	 unsigned dword;
+      } vb0;
+
+      struct pipe_buffer *buffer;
+      unsigned offset;
+
+      unsigned max_index;
+      unsigned instance_data_step_rate;
+
+   } vb[BRW_VBP_MAX];
+};
+
+
+
+unsigned brw_translate_surface_format( unsigned id )
+{
+   switch (id) {
+   case PIPE_FORMAT_R64_FLOAT:
+      return BRW_SURFACEFORMAT_R64_FLOAT;
+   case PIPE_FORMAT_R64G64_FLOAT:
+      return BRW_SURFACEFORMAT_R64G64_FLOAT;
+   case PIPE_FORMAT_R64G64B64_FLOAT:
+      return BRW_SURFACEFORMAT_R64G64B64_FLOAT;
+   case PIPE_FORMAT_R64G64B64A64_FLOAT:
+      return BRW_SURFACEFORMAT_R64G64B64A64_FLOAT;
+
+   case PIPE_FORMAT_R32_FLOAT:
+      return BRW_SURFACEFORMAT_R32_FLOAT;
+   case PIPE_FORMAT_R32G32_FLOAT:
+      return BRW_SURFACEFORMAT_R32G32_FLOAT;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      return BRW_SURFACEFORMAT_R32G32B32_FLOAT;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      return BRW_SURFACEFORMAT_R32G32B32A32_FLOAT;
+
+   case PIPE_FORMAT_R32_UNORM:
+      return BRW_SURFACEFORMAT_R32_UNORM;
+   case PIPE_FORMAT_R32G32_UNORM:
+      return BRW_SURFACEFORMAT_R32G32_UNORM;
+   case PIPE_FORMAT_R32G32B32_UNORM:
+      return BRW_SURFACEFORMAT_R32G32B32_UNORM;
+   case PIPE_FORMAT_R32G32B32A32_UNORM:
+      return BRW_SURFACEFORMAT_R32G32B32A32_UNORM;
+
+   case PIPE_FORMAT_R32_USCALED:
+      return BRW_SURFACEFORMAT_R32_USCALED;
+   case PIPE_FORMAT_R32G32_USCALED:
+      return BRW_SURFACEFORMAT_R32G32_USCALED;
+   case PIPE_FORMAT_R32G32B32_USCALED:
+      return BRW_SURFACEFORMAT_R32G32B32_USCALED;
+   case PIPE_FORMAT_R32G32B32A32_USCALED:
+      return BRW_SURFACEFORMAT_R32G32B32A32_USCALED;
+
+   case PIPE_FORMAT_R32_SNORM:
+      return BRW_SURFACEFORMAT_R32_SNORM;
+   case PIPE_FORMAT_R32G32_SNORM:
+      return BRW_SURFACEFORMAT_R32G32_SNORM;
+   case PIPE_FORMAT_R32G32B32_SNORM:
+      return BRW_SURFACEFORMAT_R32G32B32_SNORM;
+   case PIPE_FORMAT_R32G32B32A32_SNORM:
+      return BRW_SURFACEFORMAT_R32G32B32A32_SNORM;
+
+   case PIPE_FORMAT_R32_SSCALED:
+      return BRW_SURFACEFORMAT_R32_SSCALED;
+   case PIPE_FORMAT_R32G32_SSCALED:
+      return BRW_SURFACEFORMAT_R32G32_SSCALED;
+   case PIPE_FORMAT_R32G32B32_SSCALED:
+      return BRW_SURFACEFORMAT_R32G32B32_SSCALED;
+   case PIPE_FORMAT_R32G32B32A32_SSCALED:
+      return BRW_SURFACEFORMAT_R32G32B32A32_SSCALED;
+
+   case PIPE_FORMAT_R16_UNORM:
+      return BRW_SURFACEFORMAT_R16_UNORM;
+   case PIPE_FORMAT_R16G16_UNORM:
+      return BRW_SURFACEFORMAT_R16G16_UNORM;
+   case PIPE_FORMAT_R16G16B16_UNORM:
+      return BRW_SURFACEFORMAT_R16G16B16_UNORM;
+   case PIPE_FORMAT_R16G16B16A16_UNORM:
+      return BRW_SURFACEFORMAT_R16G16B16A16_UNORM;
+
+   case PIPE_FORMAT_R16_USCALED:
+      return BRW_SURFACEFORMAT_R16_USCALED;
+   case PIPE_FORMAT_R16G16_USCALED:
+      return BRW_SURFACEFORMAT_R16G16_USCALED;
+   case PIPE_FORMAT_R16G16B16_USCALED:
+      return BRW_SURFACEFORMAT_R16G16B16_USCALED;
+   case PIPE_FORMAT_R16G16B16A16_USCALED:
+      return BRW_SURFACEFORMAT_R16G16B16A16_USCALED;
+
+   case PIPE_FORMAT_R16_SNORM:
+      return BRW_SURFACEFORMAT_R16_SNORM;
+   case PIPE_FORMAT_R16G16_SNORM:
+      return BRW_SURFACEFORMAT_R16G16_SNORM;
+   case PIPE_FORMAT_R16G16B16_SNORM:
+      return BRW_SURFACEFORMAT_R16G16B16_SNORM;
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      return BRW_SURFACEFORMAT_R16G16B16A16_SNORM;
+
+   case PIPE_FORMAT_R16_SSCALED:
+      return BRW_SURFACEFORMAT_R16_SSCALED;
+   case PIPE_FORMAT_R16G16_SSCALED:
+      return BRW_SURFACEFORMAT_R16G16_SSCALED;
+   case PIPE_FORMAT_R16G16B16_SSCALED:
+      return BRW_SURFACEFORMAT_R16G16B16_SSCALED;
+   case PIPE_FORMAT_R16G16B16A16_SSCALED:
+      return BRW_SURFACEFORMAT_R16G16B16A16_SSCALED;
+
+   case PIPE_FORMAT_R8_UNORM:
+      return BRW_SURFACEFORMAT_R8_UNORM;
+   case PIPE_FORMAT_R8G8_UNORM:
+      return BRW_SURFACEFORMAT_R8G8_UNORM;
+   case PIPE_FORMAT_R8G8B8_UNORM:
+      return BRW_SURFACEFORMAT_R8G8B8_UNORM;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      return BRW_SURFACEFORMAT_R8G8B8A8_UNORM;
+
+   case PIPE_FORMAT_R8_USCALED:
+      return BRW_SURFACEFORMAT_R8_USCALED;
+   case PIPE_FORMAT_R8G8_USCALED:
+      return BRW_SURFACEFORMAT_R8G8_USCALED;
+   case PIPE_FORMAT_R8G8B8_USCALED:
+      return BRW_SURFACEFORMAT_R8G8B8_USCALED;
+   case PIPE_FORMAT_R8G8B8A8_USCALED:
+      return BRW_SURFACEFORMAT_R8G8B8A8_USCALED;
+
+   case PIPE_FORMAT_R8_SNORM:
+      return BRW_SURFACEFORMAT_R8_SNORM;
+   case PIPE_FORMAT_R8G8_SNORM:
+      return BRW_SURFACEFORMAT_R8G8_SNORM;
+   case PIPE_FORMAT_R8G8B8_SNORM:
+      return BRW_SURFACEFORMAT_R8G8B8_SNORM;
+   case PIPE_FORMAT_R8G8B8A8_SNORM:
+      return BRW_SURFACEFORMAT_R8G8B8A8_SNORM;
+
+   case PIPE_FORMAT_R8_SSCALED:
+      return BRW_SURFACEFORMAT_R8_SSCALED;
+   case PIPE_FORMAT_R8G8_SSCALED:
+      return BRW_SURFACEFORMAT_R8G8_SSCALED;
+   case PIPE_FORMAT_R8G8B8_SSCALED:
+      return BRW_SURFACEFORMAT_R8G8B8_SSCALED;
+   case PIPE_FORMAT_R8G8B8A8_SSCALED:
+      return BRW_SURFACEFORMAT_R8G8B8A8_SSCALED;
+
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+static unsigned get_index_type(int type)
+{
+   switch (type) {
+   case 1: return BRW_INDEX_BYTE;
+   case 2: return BRW_INDEX_WORD;
+   case 4: return BRW_INDEX_DWORD;
+   default: assert(0); return 0;
+   }
+}
+
+
+boolean brw_upload_vertex_buffers( struct brw_context *brw )
+{
+   struct brw_array_state vbp;
+   unsigned nr_enabled = 0;
+   unsigned i;
+
+   memset(&vbp, 0, sizeof(vbp));
+
+   /* This is a hardware limit:
+    */
+
+   for (i = 0; i < BRW_VEP_MAX; i++)
+   {
+      if (brw->vb.vbo_array[i] == NULL) {
+	 nr_enabled = i;
+	 break;
+      }
+
+      vbp.vb[i].vb0.bits.pitch = brw->vb.vbo_array[i]->stride;
+      vbp.vb[i].vb0.bits.pad = 0;
+      vbp.vb[i].vb0.bits.access_type = BRW_VERTEXBUFFER_ACCESS_VERTEXDATA;
+      vbp.vb[i].vb0.bits.vb_index = i;
+      vbp.vb[i].offset = brw->vb.vbo_array[i]->buffer_offset;
+      vbp.vb[i].buffer = brw->vb.vbo_array[i]->buffer;
+      vbp.vb[i].max_index = brw->vb.vbo_array[i]->max_index;
+   }
+
+
+   vbp.header.bits.length = (1 + nr_enabled * 4) - 2;
+   vbp.header.bits.opcode = CMD_VERTEX_BUFFER;
+
+   BEGIN_BATCH(vbp.header.bits.length+2, 0);
+   OUT_BATCH( vbp.header.dword );
+
+   for (i = 0; i < nr_enabled; i++) {
+      OUT_BATCH( vbp.vb[i].vb0.dword );
+      OUT_RELOC( vbp.vb[i].buffer,  PIPE_BUFFER_USAGE_GPU_READ,
+		 vbp.vb[i].offset);
+      OUT_BATCH( vbp.vb[i].max_index );
+      OUT_BATCH( vbp.vb[i].instance_data_step_rate );
+   }
+   ADVANCE_BATCH();
+   return TRUE;
+}
+
+
+
+boolean brw_upload_vertex_elements( struct brw_context *brw )
+{
+   struct brw_vertex_element_packet vep;
+
+   unsigned i;
+   unsigned nr_enabled = brw->attribs.VertexProgram->info.num_inputs;
+
+   memset(&vep, 0, sizeof(vep));
+
+   for (i = 0; i < nr_enabled; i++) 
+      vep.ve[i] = brw->vb.inputs[i];
+
+
+   vep.header.length = (1 + nr_enabled * sizeof(vep.ve[0])/4) - 2;
+   vep.header.opcode = CMD_VERTEX_ELEMENT;
+   brw_cached_batch_struct(brw, &vep, 4 + nr_enabled * sizeof(vep.ve[0]));
+
+   return TRUE;
+}
+
+boolean brw_upload_indices( struct brw_context *brw,
+                            const struct pipe_buffer *index_buffer,
+                            int ib_size, int start, int count)
+{
+   /* Emit the indexbuffer packet:
+    */
+   {
+      struct brw_indexbuffer ib;
+
+      memset(&ib, 0, sizeof(ib));
+
+      ib.header.bits.opcode = CMD_INDEX_BUFFER;
+      ib.header.bits.length = sizeof(ib)/4 - 2;
+      ib.header.bits.index_format = get_index_type(ib_size);
+      ib.header.bits.cut_index_enable = 0;
+
+
+      BEGIN_BATCH(4, 0);
+      OUT_BATCH( ib.header.dword );
+      OUT_RELOC( index_buffer, PIPE_BUFFER_USAGE_GPU_READ, start);
+      OUT_RELOC( index_buffer, PIPE_BUFFER_USAGE_GPU_READ, start + count);
+      OUT_BATCH( 0 );
+      ADVANCE_BATCH();
+   }
+   return TRUE;
+}
diff --git a/src/gallium/drivers/i965simple/brw_eu.c b/src/gallium/drivers/i965simple/brw_eu.c
new file mode 100644
index 0000000000..e2002d1821
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_eu.c
@@ -0,0 +1,130 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+  
+
+#include "brw_context.h"
+#include "brw_defines.h"
+#include "brw_eu.h"
+
+
+
+/* How does predicate control work when execution_size != 8?  Do I
+ * need to test/set for 0xffff when execution_size is 16?
+ */
+void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value )
+{
+   p->current->header.predicate_control = BRW_PREDICATE_NONE;
+
+   if (value != 0xff) {
+      if (value != p->flag_value) {
+	 brw_push_insn_state(p);
+	 brw_MOV(p, brw_flag_reg(), brw_imm_uw(value));
+	 p->flag_value = value;
+	 brw_pop_insn_state(p);
+      }
+
+      p->current->header.predicate_control = BRW_PREDICATE_NORMAL;
+   }   
+}
+
+void brw_set_predicate_control( struct brw_compile *p, unsigned pc )
+{
+   p->current->header.predicate_control = pc;
+}
+
+void brw_set_conditionalmod( struct brw_compile *p, unsigned conditional )
+{
+   p->current->header.destreg__conditonalmod = conditional;
+}
+
+void brw_set_access_mode( struct brw_compile *p, unsigned access_mode )
+{
+   p->current->header.access_mode = access_mode;
+}
+
+void brw_set_compression_control( struct brw_compile *p, boolean compression_control )
+{
+   p->current->header.compression_control = compression_control;
+}
+
+void brw_set_mask_control( struct brw_compile *p, unsigned value )
+{
+   p->current->header.mask_control = value;
+}
+
+void brw_set_saturate( struct brw_compile *p, unsigned value )
+{
+   p->current->header.saturate = value;
+}
+
+void brw_push_insn_state( struct brw_compile *p )
+{
+   assert(p->current != &p->stack[BRW_EU_MAX_INSN_STACK-1]);
+   memcpy(p->current+1, p->current, sizeof(struct brw_instruction));
+   p->current++;   
+}
+
+void brw_pop_insn_state( struct brw_compile *p )
+{
+   assert(p->current != p->stack);
+   p->current--;
+}
+
+
+/***********************************************************************
+ */
+void brw_init_compile( struct brw_compile *p )
+{
+   p->nr_insn = 0;
+   p->current = p->stack;
+   memset(p->current, 0, sizeof(p->current[0]));
+
+   /* Some defaults?
+    */
+   brw_set_mask_control(p, BRW_MASK_ENABLE); /* what does this do? */
+   brw_set_saturate(p, 0);
+   brw_set_compression_control(p, BRW_COMPRESSION_NONE);
+   brw_set_predicate_control_flag_value(p, 0xff); 
+}
+
+
+const unsigned *brw_get_program( struct brw_compile *p,
+			       unsigned *sz )
+{
+   unsigned i;
+
+   for (i = 0; i < 8; i++)
+      brw_NOP(p);
+
+   *sz = p->nr_insn * sizeof(struct brw_instruction);
+   return (const unsigned *)p->store;
+}
+
diff --git a/src/gallium/drivers/i965simple/brw_eu.h b/src/gallium/drivers/i965simple/brw_eu.h
new file mode 100644
index 0000000000..23151ae9ed
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_eu.h
@@ -0,0 +1,888 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#ifndef BRW_EU_H
+#define BRW_EU_H
+
+#include "brw_structs.h"
+#include "brw_defines.h"
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_shader_tokens.h"
+
+#define BRW_SWIZZLE4(a,b,c,d) (((a)<<0) | ((b)<<2) | ((c)<<4) | ((d)<<6))
+#define BRW_GET_SWZ(swz, idx) (((swz) >> ((idx)*2)) & 0x3)
+
+#define BRW_SWIZZLE_NOOP      BRW_SWIZZLE4(0,1,2,3)
+#define BRW_SWIZZLE_XYZW      BRW_SWIZZLE4(0,1,2,3)
+#define BRW_SWIZZLE_XXXX      BRW_SWIZZLE4(0,0,0,0)
+#define BRW_SWIZZLE_XYXY      BRW_SWIZZLE4(0,1,0,1)
+
+
+#define REG_SIZE (8*4)
+
+
+/* These aren't hardware structs, just something useful for us to pass around:
+ *
+ * Align1 operation has a lot of control over input ranges.  Used in
+ * WM programs to implement shaders decomposed into "channel serial"
+ * or "structure of array" form:
+ */
+struct brw_reg
+{
+   unsigned type:4;
+   unsigned file:2;
+   unsigned nr:8;
+   unsigned subnr:5;		/* :1 in align16 */
+   unsigned negate:1;		/* source only */
+   unsigned abs:1;		/* source only */
+   unsigned vstride:4;		/* source only */
+   unsigned width:3;		/* src only, align1 only */
+   unsigned hstride:2;   		/* src only, align1 only */
+   unsigned address_mode:1;	/* relative addressing, hopefully! */
+   unsigned pad0:1;
+
+   union {
+      struct {
+	 unsigned swizzle:8;		/* src only, align16 only */
+	 unsigned writemask:4;		/* dest only, align16 only */
+	 int  indirect_offset:10;	/* relative addressing offset */
+	 unsigned pad1:10;		/* two dwords total */
+      } bits;
+
+      float f;
+      int   d;
+      unsigned ud;
+   } dw1;
+};
+
+
+struct brw_indirect {
+   unsigned addr_subnr:4;
+   int addr_offset:10;
+   unsigned pad:18;
+};
+
+
+#define BRW_EU_MAX_INSN_STACK 5
+#define BRW_EU_MAX_INSN 1200
+
+struct brw_compile {
+   struct brw_instruction store[BRW_EU_MAX_INSN];
+   unsigned nr_insn;
+
+   /* Allow clients to push/pop instruction state:
+    */
+   struct brw_instruction stack[BRW_EU_MAX_INSN_STACK];
+   struct brw_instruction *current;
+
+   unsigned flag_value;
+   boolean single_program_flow;
+};
+
+
+
+static __inline int type_sz( unsigned type )
+{
+   switch( type ) {
+   case BRW_REGISTER_TYPE_UD:
+   case BRW_REGISTER_TYPE_D:
+   case BRW_REGISTER_TYPE_F:
+      return 4;
+   case BRW_REGISTER_TYPE_HF:
+   case BRW_REGISTER_TYPE_UW:
+   case BRW_REGISTER_TYPE_W:
+      return 2;
+   case BRW_REGISTER_TYPE_UB:
+   case BRW_REGISTER_TYPE_B:
+      return 1;
+   default:
+      return 0;
+   }
+}
+
+static __inline struct brw_reg brw_reg( unsigned file,
+					unsigned nr,
+					unsigned subnr,
+					unsigned type,
+					unsigned vstride,
+					unsigned width,
+					unsigned hstride,
+					unsigned swizzle,
+					unsigned writemask)
+{
+
+   struct brw_reg reg;
+   reg.type = type;
+   reg.file = file;
+   reg.nr = nr;
+   reg.subnr = subnr * type_sz(type);
+   reg.negate = 0;
+   reg.abs = 0;
+   reg.vstride = vstride;
+   reg.width = width;
+   reg.hstride = hstride;
+   reg.address_mode = BRW_ADDRESS_DIRECT;
+   reg.pad0 = 0;
+
+   /* Could do better: If the reg is r5.3<0;1,0>, we probably want to
+    * set swizzle and writemask to W, as the lower bits of subnr will
+    * be lost when converted to align16.  This is probably too much to
+    * keep track of as you'd want it adjusted by suboffset(), etc.
+    * Perhaps fix up when converting to align16?
+    */
+   reg.dw1.bits.swizzle = swizzle;
+   reg.dw1.bits.writemask = writemask;
+   reg.dw1.bits.indirect_offset = 0;
+   reg.dw1.bits.pad1 = 0;
+   return reg;
+}
+
+static __inline struct brw_reg brw_vec16_reg( unsigned file,
+					      unsigned nr,
+					      unsigned subnr )
+{
+   return brw_reg(file,
+		  nr,
+		  subnr,
+		  BRW_REGISTER_TYPE_F,
+		  BRW_VERTICAL_STRIDE_16,
+		  BRW_WIDTH_16,
+		  BRW_HORIZONTAL_STRIDE_1,
+		  BRW_SWIZZLE_XYZW,
+		  TGSI_WRITEMASK_XYZW);
+}
+
+static __inline struct brw_reg brw_vec8_reg( unsigned file,
+					     unsigned nr,
+					     unsigned subnr )
+{
+   return brw_reg(file,
+		  nr,
+		  subnr,
+		  BRW_REGISTER_TYPE_F,
+		  BRW_VERTICAL_STRIDE_8,
+		  BRW_WIDTH_8,
+		  BRW_HORIZONTAL_STRIDE_1,
+		  BRW_SWIZZLE_XYZW,
+		  TGSI_WRITEMASK_XYZW);
+}
+
+
+static __inline struct brw_reg brw_vec4_reg( unsigned file,
+					      unsigned nr,
+					      unsigned subnr )
+{
+   return brw_reg(file,
+		  nr,
+		  subnr,
+		  BRW_REGISTER_TYPE_F,
+		  BRW_VERTICAL_STRIDE_4,
+		  BRW_WIDTH_4,
+		  BRW_HORIZONTAL_STRIDE_1,
+		  BRW_SWIZZLE_XYZW,
+		  TGSI_WRITEMASK_XYZW);
+}
+
+
+static __inline struct brw_reg brw_vec2_reg( unsigned file,
+					      unsigned nr,
+					      unsigned subnr )
+{
+   return brw_reg(file,
+		  nr,
+		  subnr,
+		  BRW_REGISTER_TYPE_F,
+		  BRW_VERTICAL_STRIDE_2,
+		  BRW_WIDTH_2,
+		  BRW_HORIZONTAL_STRIDE_1,
+		  BRW_SWIZZLE_XYXY,
+		  TGSI_WRITEMASK_XY);
+}
+
+static __inline struct brw_reg brw_vec1_reg( unsigned file,
+					     unsigned nr,
+					     unsigned subnr )
+{
+   return brw_reg(file,
+		  nr,
+		  subnr,
+		  BRW_REGISTER_TYPE_F,
+		  BRW_VERTICAL_STRIDE_0,
+		  BRW_WIDTH_1,
+		  BRW_HORIZONTAL_STRIDE_0,
+		  BRW_SWIZZLE_XXXX,
+		  TGSI_WRITEMASK_X);
+}
+
+
+static __inline struct brw_reg retype( struct brw_reg reg,
+				       unsigned type )
+{
+   reg.type = type;
+   return reg;
+}
+
+static __inline struct brw_reg suboffset( struct brw_reg reg,
+					  unsigned delta )
+{
+   reg.subnr += delta * type_sz(reg.type);
+   return reg;
+}
+
+
+static __inline struct brw_reg offset( struct brw_reg reg,
+				       unsigned delta )
+{
+   reg.nr += delta;
+   return reg;
+}
+
+
+static __inline struct brw_reg byte_offset( struct brw_reg reg,
+					    unsigned bytes )
+{
+   unsigned newoffset = reg.nr * REG_SIZE + reg.subnr + bytes;
+   reg.nr = newoffset / REG_SIZE;
+   reg.subnr = newoffset % REG_SIZE;
+   return reg;
+}
+
+
+static __inline struct brw_reg brw_uw16_reg( unsigned file,
+					     unsigned nr,
+					     unsigned subnr )
+{
+   return suboffset(retype(brw_vec16_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr);
+}
+
+static __inline struct brw_reg brw_uw8_reg( unsigned file,
+					    unsigned nr,
+					    unsigned subnr )
+{
+   return suboffset(retype(brw_vec8_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr);
+}
+
+static __inline struct brw_reg brw_uw1_reg( unsigned file,
+					    unsigned nr,
+					    unsigned subnr )
+{
+   return suboffset(retype(brw_vec1_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr);
+}
+
+static __inline struct brw_reg brw_imm_reg( unsigned type )
+{
+   return brw_reg( BRW_IMMEDIATE_VALUE,
+		   0,
+		   0,
+		   type,
+		   BRW_VERTICAL_STRIDE_0,
+		   BRW_WIDTH_1,
+		   BRW_HORIZONTAL_STRIDE_0,
+		   0,
+		   0);
+}
+
+static __inline struct brw_reg brw_imm_f( float f )
+{
+   struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_F);
+   imm.dw1.f = f;
+   return imm;
+}
+
+static __inline struct brw_reg brw_imm_d( int d )
+{
+   struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_D);
+   imm.dw1.d = d;
+   return imm;
+}
+
+static __inline struct brw_reg brw_imm_ud( unsigned ud )
+{
+   struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UD);
+   imm.dw1.ud = ud;
+   return imm;
+}
+
+static __inline struct brw_reg brw_imm_uw( ushort uw )
+{
+   struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UW);
+   imm.dw1.ud = uw;
+   return imm;
+}
+
+static __inline struct brw_reg brw_imm_w( short w )
+{
+   struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_W);
+   imm.dw1.d = w;
+   return imm;
+}
+
+/* brw_imm_b and brw_imm_ub aren't supported by hardware - the type
+ * numbers alias with _V and _VF below:
+ */
+
+/* Vector of eight signed half-byte values:
+ */
+static __inline struct brw_reg brw_imm_v( unsigned v )
+{
+   struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_V);
+   imm.vstride = BRW_VERTICAL_STRIDE_0;
+   imm.width = BRW_WIDTH_8;
+   imm.hstride = BRW_HORIZONTAL_STRIDE_1;
+   imm.dw1.ud = v;
+   return imm;
+}
+
+/* Vector of four 8-bit float values:
+ */
+static __inline struct brw_reg brw_imm_vf( unsigned v )
+{
+   struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF);
+   imm.vstride = BRW_VERTICAL_STRIDE_0;
+   imm.width = BRW_WIDTH_4;
+   imm.hstride = BRW_HORIZONTAL_STRIDE_1;
+   imm.dw1.ud = v;
+   return imm;
+}
+
+#define VF_ZERO 0x0
+#define VF_ONE  0x30
+#define VF_NEG  (1<<7)
+
+static __inline struct brw_reg brw_imm_vf4( unsigned v0,
+					    unsigned v1,
+					    unsigned v2,
+					    unsigned v3)
+{
+   struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF);
+   imm.vstride = BRW_VERTICAL_STRIDE_0;
+   imm.width = BRW_WIDTH_4;
+   imm.hstride = BRW_HORIZONTAL_STRIDE_1;
+   imm.dw1.ud = ((v0 << 0) |
+		 (v1 << 8) |
+		 (v2 << 16) |
+		 (v3 << 24));
+   return imm;
+}
+
+
+static __inline struct brw_reg brw_address( struct brw_reg reg )
+{
+   return brw_imm_uw(reg.nr * REG_SIZE + reg.subnr);
+}
+
+
+static __inline struct brw_reg brw_vec1_grf( unsigned nr,
+					       unsigned subnr )
+{
+   return brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
+}
+
+static __inline struct brw_reg brw_vec8_grf( unsigned nr,
+					     unsigned subnr )
+{
+   return brw_vec8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
+}
+
+static __inline struct brw_reg brw_vec4_grf( unsigned nr,
+					     unsigned subnr )
+{
+   return brw_vec4_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
+}
+
+
+static __inline struct brw_reg brw_vec2_grf( unsigned nr,
+					     unsigned subnr )
+{
+   return brw_vec2_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
+}
+
+static __inline struct brw_reg brw_uw8_grf( unsigned nr,
+					    unsigned subnr )
+{
+   return brw_uw8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
+}
+
+static __inline struct brw_reg brw_null_reg( void )
+{
+   return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE,
+		       BRW_ARF_NULL,
+		       0);
+}
+
+static __inline struct brw_reg brw_address_reg( unsigned subnr )
+{
+   return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE,
+		      BRW_ARF_ADDRESS,
+		      subnr);
+}
+
+/* If/else instructions break in align16 mode if writemask & swizzle
+ * aren't xyzw.  This goes against the convention for other scalar
+ * regs:
+ */
+static __inline struct brw_reg brw_ip_reg( void )
+{
+   return brw_reg(BRW_ARCHITECTURE_REGISTER_FILE,
+		  BRW_ARF_IP,
+		  0,
+		  BRW_REGISTER_TYPE_UD,
+		  BRW_VERTICAL_STRIDE_4, /* ? */
+		  BRW_WIDTH_1,
+		  BRW_HORIZONTAL_STRIDE_0,
+		  BRW_SWIZZLE_XYZW, /* NOTE! */
+		  TGSI_WRITEMASK_XYZW); /* NOTE! */
+}
+
+static __inline struct brw_reg brw_acc_reg( void )
+{
+   return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE,
+		       BRW_ARF_ACCUMULATOR,
+		       0);
+}
+
+
+static __inline struct brw_reg brw_flag_reg( void )
+{
+   return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE,
+		      BRW_ARF_FLAG,
+		      0);
+}
+
+
+static __inline struct brw_reg brw_mask_reg( unsigned subnr )
+{
+   return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE,
+		      BRW_ARF_MASK,
+		      subnr);
+}
+
+static __inline struct brw_reg brw_message_reg( unsigned nr )
+{
+   return brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE,
+		       nr,
+		       0);
+}
+
+
+
+
+/* This is almost always called with a numeric constant argument, so
+ * make things easy to evaluate at compile time:
+ */
+static __inline unsigned cvt( unsigned val )
+{
+   switch (val) {
+   case 0: return 0;
+   case 1: return 1;
+   case 2: return 2;
+   case 4: return 3;
+   case 8: return 4;
+   case 16: return 5;
+   case 32: return 6;
+   }
+   return 0;
+}
+
+static __inline struct brw_reg stride( struct brw_reg reg,
+				       unsigned vstride,
+				       unsigned width,
+				       unsigned hstride )
+{
+
+   reg.vstride = cvt(vstride);
+   reg.width = cvt(width) - 1;
+   reg.hstride = cvt(hstride);
+   return reg;
+}
+
+static __inline struct brw_reg vec16( struct brw_reg reg )
+{
+   return stride(reg, 16,16,1);
+}
+
+static __inline struct brw_reg vec8( struct brw_reg reg )
+{
+   return stride(reg, 8,8,1);
+}
+
+static __inline struct brw_reg vec4( struct brw_reg reg )
+{
+   return stride(reg, 4,4,1);
+}
+
+static __inline struct brw_reg vec2( struct brw_reg reg )
+{
+   return stride(reg, 2,2,1);
+}
+
+static __inline struct brw_reg vec1( struct brw_reg reg )
+{
+   return stride(reg, 0,1,0);
+}
+
+static __inline struct brw_reg get_element( struct brw_reg reg, unsigned elt )
+{
+   return vec1(suboffset(reg, elt));
+}
+
+static __inline struct brw_reg get_element_ud( struct brw_reg reg, unsigned elt )
+{
+   return vec1(suboffset(retype(reg, BRW_REGISTER_TYPE_UD), elt));
+}
+
+
+static __inline struct brw_reg brw_swizzle( struct brw_reg reg,
+					    unsigned x,
+					    unsigned y,
+					    unsigned z,
+					    unsigned w)
+{
+   reg.dw1.bits.swizzle = BRW_SWIZZLE4(BRW_GET_SWZ(reg.dw1.bits.swizzle, x),
+				       BRW_GET_SWZ(reg.dw1.bits.swizzle, y),
+				       BRW_GET_SWZ(reg.dw1.bits.swizzle, z),
+				       BRW_GET_SWZ(reg.dw1.bits.swizzle, w));
+   return reg;
+}
+
+
+static __inline struct brw_reg brw_swizzle1( struct brw_reg reg,
+					     unsigned x )
+{
+   return brw_swizzle(reg, x, x, x, x);
+}
+
+static __inline struct brw_reg brw_writemask( struct brw_reg reg,
+					      unsigned mask )
+{
+   reg.dw1.bits.writemask &= mask;
+   return reg;
+}
+
+static __inline struct brw_reg brw_set_writemask( struct brw_reg reg,
+						  unsigned mask )
+{
+   reg.dw1.bits.writemask = mask;
+   return reg;
+}
+
+static __inline struct brw_reg negate( struct brw_reg reg )
+{
+   reg.negate ^= 1;
+   return reg;
+}
+
+static __inline struct brw_reg brw_abs( struct brw_reg reg )
+{
+   reg.abs = 1;
+   return reg;
+}
+
+/***********************************************************************
+ */
+static __inline struct brw_reg brw_vec4_indirect( unsigned subnr,
+						  int offset )
+{
+   struct brw_reg reg =  brw_vec4_grf(0, 0);
+   reg.subnr = subnr;
+   reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER;
+   reg.dw1.bits.indirect_offset = offset;
+   return reg;
+}
+
+static __inline struct brw_reg brw_vec1_indirect( unsigned subnr,
+						  int offset )
+{
+   struct brw_reg reg =  brw_vec1_grf(0, 0);
+   reg.subnr = subnr;
+   reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER;
+   reg.dw1.bits.indirect_offset = offset;
+   return reg;
+}
+
+static __inline struct brw_reg deref_4f(struct brw_indirect ptr, int offset)
+{
+   return brw_vec4_indirect(ptr.addr_subnr, ptr.addr_offset + offset);
+}
+
+static __inline struct brw_reg deref_1f(struct brw_indirect ptr, int offset)
+{
+   return brw_vec1_indirect(ptr.addr_subnr, ptr.addr_offset + offset);
+}
+
+static __inline struct brw_reg deref_4b(struct brw_indirect ptr, int offset)
+{
+   return retype(deref_4f(ptr, offset), BRW_REGISTER_TYPE_B);
+}
+
+static __inline struct brw_reg deref_1uw(struct brw_indirect ptr, int offset)
+{
+   return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UW);
+}
+
+static __inline struct brw_reg deref_1ud(struct brw_indirect ptr, int offset)
+{
+   return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UD);
+}
+
+static __inline struct brw_reg get_addr_reg(struct brw_indirect ptr)
+{
+   return brw_address_reg(ptr.addr_subnr);
+}
+
+static __inline struct brw_indirect brw_indirect_offset( struct brw_indirect ptr, int offset )
+{
+   ptr.addr_offset += offset;
+   return ptr;
+}
+
+static __inline struct brw_indirect brw_indirect( unsigned addr_subnr, int offset )
+{
+   struct brw_indirect ptr;
+   ptr.addr_subnr = addr_subnr;
+   ptr.addr_offset = offset;
+   ptr.pad = 0;
+   return ptr;
+}
+
+static __inline struct brw_instruction *current_insn( struct brw_compile *p)
+{
+	return &p->store[p->nr_insn];
+}
+
+void brw_pop_insn_state( struct brw_compile *p );
+void brw_push_insn_state( struct brw_compile *p );
+void brw_set_mask_control( struct brw_compile *p, unsigned value );
+void brw_set_saturate( struct brw_compile *p, unsigned value );
+void brw_set_access_mode( struct brw_compile *p, unsigned access_mode );
+void brw_set_compression_control( struct brw_compile *p, boolean control );
+void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value );
+void brw_set_predicate_control( struct brw_compile *p, unsigned pc );
+void brw_set_conditionalmod( struct brw_compile *p, unsigned conditional );
+
+void brw_init_compile( struct brw_compile *p );
+const unsigned *brw_get_program( struct brw_compile *p, unsigned *sz );
+
+
+struct brw_instruction *brw_alu1( struct brw_compile *p,
+				  unsigned opcode,
+				  struct brw_reg dest,
+				  struct brw_reg src );
+
+struct brw_instruction *brw_alu2(struct brw_compile *p,
+				 unsigned opcode,
+				 struct brw_reg dest,
+				 struct brw_reg src0,
+				 struct brw_reg src1 );
+
+/* Helpers for regular instructions:
+ */
+#define ALU1(OP)					\
+struct brw_instruction *brw_##OP(struct brw_compile *p,	\
+	      struct brw_reg dest,			\
+	      struct brw_reg src0);
+
+#define ALU2(OP)					\
+struct brw_instruction *brw_##OP(struct brw_compile *p,	\
+	      struct brw_reg dest,			\
+	      struct brw_reg src0,			\
+	      struct brw_reg src1);
+
+ALU1(MOV)
+ALU2(SEL)
+ALU1(NOT)
+ALU2(AND)
+ALU2(OR)
+ALU2(XOR)
+ALU2(SHR)
+ALU2(SHL)
+ALU2(RSR)
+ALU2(RSL)
+ALU2(ASR)
+ALU2(JMPI)
+ALU2(ADD)
+ALU2(MUL)
+ALU1(FRC)
+ALU1(RNDD)
+ALU2(MAC)
+ALU2(MACH)
+ALU1(LZD)
+ALU2(DP4)
+ALU2(DPH)
+ALU2(DP3)
+ALU2(DP2)
+ALU2(LINE)
+
+#undef ALU1
+#undef ALU2
+
+
+
+/* Helpers for SEND instruction:
+ */
+void brw_urb_WRITE(struct brw_compile *p,
+		   struct brw_reg dest,
+		   unsigned msg_reg_nr,
+		   struct brw_reg src0,
+		   boolean allocate,
+		   boolean used,
+		   unsigned msg_length,
+		   unsigned response_length,
+		   boolean eot,
+		   boolean writes_complete,
+		   unsigned offset,
+		   unsigned swizzle);
+
+void brw_fb_WRITE(struct brw_compile *p,
+		   struct brw_reg dest,
+		   unsigned msg_reg_nr,
+		   struct brw_reg src0,
+		   unsigned binding_table_index,
+		   unsigned msg_length,
+		   unsigned response_length,
+		   boolean eot);
+
+void brw_SAMPLE(struct brw_compile *p,
+		struct brw_reg dest,
+		unsigned msg_reg_nr,
+		struct brw_reg src0,
+		unsigned binding_table_index,
+		unsigned sampler,
+		unsigned writemask,
+		unsigned msg_type,
+		unsigned response_length,
+		unsigned msg_length,
+		boolean eot);
+
+void brw_math_16( struct brw_compile *p,
+		  struct brw_reg dest,
+		  unsigned function,
+		  unsigned saturate,
+		  unsigned msg_reg_nr,
+		  struct brw_reg src,
+		  unsigned precision );
+
+void brw_math( struct brw_compile *p,
+	       struct brw_reg dest,
+	       unsigned function,
+	       unsigned saturate,
+	       unsigned msg_reg_nr,
+	       struct brw_reg src,
+	       unsigned data_type,
+	       unsigned precision );
+
+void brw_dp_READ_16( struct brw_compile *p,
+		     struct brw_reg dest,
+		     unsigned msg_reg_nr,
+		     unsigned scratch_offset );
+
+void brw_dp_WRITE_16( struct brw_compile *p,
+		      struct brw_reg src,
+		      unsigned msg_reg_nr,
+		      unsigned scratch_offset );
+
+/* If/else/endif.  Works by manipulating the execution flags on each
+ * channel.
+ */
+struct brw_instruction *brw_IF(struct brw_compile *p,
+			       unsigned execute_size);
+
+struct brw_instruction *brw_ELSE(struct brw_compile *p,
+				 struct brw_instruction *if_insn);
+
+void brw_ENDIF(struct brw_compile *p,
+	       struct brw_instruction *if_or_else_insn);
+
+
+/* DO/WHILE loops:
+ */
+struct brw_instruction *brw_DO(struct brw_compile *p,
+			       unsigned execute_size);
+
+struct brw_instruction *brw_WHILE(struct brw_compile *p,
+	       struct brw_instruction *patch_insn);
+
+struct brw_instruction *brw_BREAK(struct brw_compile *p);
+struct brw_instruction *brw_CONT(struct brw_compile *p);
+/* Forward jumps:
+ */
+void brw_land_fwd_jump(struct brw_compile *p,
+		       struct brw_instruction *jmp_insn);
+
+
+
+void brw_NOP(struct brw_compile *p);
+
+/* Special case: there is never a destination, execution size will be
+ * taken from src0:
+ */
+void brw_CMP(struct brw_compile *p,
+	     struct brw_reg dest,
+	     unsigned conditional,
+	     struct brw_reg src0,
+	     struct brw_reg src1);
+
+void brw_print_reg( struct brw_reg reg );
+
+
+/***********************************************************************
+ * brw_eu_util.c:
+ */
+
+void brw_copy_indirect_to_indirect(struct brw_compile *p,
+				   struct brw_indirect dst_ptr,
+				   struct brw_indirect src_ptr,
+				   unsigned count);
+
+void brw_copy_from_indirect(struct brw_compile *p,
+			    struct brw_reg dst,
+			    struct brw_indirect ptr,
+			    unsigned count);
+
+void brw_copy4(struct brw_compile *p,
+	       struct brw_reg dst,
+	       struct brw_reg src,
+	       unsigned count);
+
+void brw_copy8(struct brw_compile *p,
+	       struct brw_reg dst,
+	       struct brw_reg src,
+	       unsigned count);
+
+void brw_math_invert( struct brw_compile *p,
+		      struct brw_reg dst,
+		      struct brw_reg src);
+
+void brw_set_src1( struct brw_instruction *insn,
+                          struct brw_reg reg );
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_eu_debug.c b/src/gallium/drivers/i965simple/brw_eu_debug.c
new file mode 100644
index 0000000000..4a94ddefa6
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_eu_debug.c
@@ -0,0 +1,90 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+    
+
+#include "pipe/p_debug.h"
+
+#include "brw_eu.h"
+
+void brw_print_reg( struct brw_reg hwreg )
+{
+   static const char *file[] = {
+      "arf",
+      "grf",
+      "msg",
+      "imm"
+   };
+
+   static const char *type[] = {
+      "ud",
+      "d",
+      "uw",
+      "w",
+      "ub",
+      "vf",
+      "hf",
+      "f"
+   };
+
+   debug_printf("%s%s", 
+		hwreg.abs ? "abs/" : "",
+		hwreg.negate ? "-" : "");
+     
+   if (hwreg.file == BRW_GENERAL_REGISTER_FILE &&
+       hwreg.nr % 2 == 0 &&
+       hwreg.subnr == 0 &&
+       hwreg.vstride == BRW_VERTICAL_STRIDE_8 &&
+       hwreg.width == BRW_WIDTH_8 &&
+       hwreg.hstride == BRW_HORIZONTAL_STRIDE_1 &&
+       hwreg.type == BRW_REGISTER_TYPE_F) {
+      debug_printf("vec%d", hwreg.nr);
+   }
+   else if (hwreg.file == BRW_GENERAL_REGISTER_FILE &&
+	    hwreg.vstride == BRW_VERTICAL_STRIDE_0 &&
+	    hwreg.width == BRW_WIDTH_1 &&
+	    hwreg.hstride == BRW_HORIZONTAL_STRIDE_0 &&
+	    hwreg.type == BRW_REGISTER_TYPE_F) {      
+      debug_printf("scl%d.%d", hwreg.nr, hwreg.subnr / 4);
+   }
+   else {
+      debug_printf("%s%d.%d<%d;%d,%d>:%s", 
+		   file[hwreg.file],
+		   hwreg.nr,
+		   hwreg.subnr / type_sz(hwreg.type),
+		   hwreg.vstride ? (1<<(hwreg.vstride-1)) : 0,
+		   1<<hwreg.width,
+		   hwreg.hstride ? (1<<(hwreg.hstride-1)) : 0,		
+		   type[hwreg.type]);
+   }
+}
+
+
+
diff --git a/src/gallium/drivers/i965simple/brw_eu_emit.c b/src/gallium/drivers/i965simple/brw_eu_emit.c
new file mode 100644
index 0000000000..400a80b6fb
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_eu_emit.c
@@ -0,0 +1,1080 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_context.h"
+#include "brw_defines.h"
+#include "brw_eu.h"
+
+
+
+
+/***********************************************************************
+ * Internal helper for constructing instructions
+ */
+
+static void guess_execution_size( struct brw_instruction *insn,
+				  struct brw_reg reg )
+{
+   if (reg.width == BRW_WIDTH_8 &&
+       insn->header.compression_control == BRW_COMPRESSION_COMPRESSED)
+      insn->header.execution_size = BRW_EXECUTE_16;
+   else
+      insn->header.execution_size = reg.width;	/* note - definitions are compatible */
+}
+
+
+static void brw_set_dest( struct brw_instruction *insn,
+			  struct brw_reg dest )
+{
+   insn->bits1.da1.dest_reg_file = dest.file;
+   insn->bits1.da1.dest_reg_type = dest.type;
+   insn->bits1.da1.dest_address_mode = dest.address_mode;
+
+   if (dest.address_mode == BRW_ADDRESS_DIRECT) {
+      insn->bits1.da1.dest_reg_nr = dest.nr;
+
+      if (insn->header.access_mode == BRW_ALIGN_1) {
+	 insn->bits1.da1.dest_subreg_nr = dest.subnr;
+	 insn->bits1.da1.dest_horiz_stride = BRW_HORIZONTAL_STRIDE_1;
+      }
+      else {
+	 insn->bits1.da16.dest_subreg_nr = dest.subnr / 16;
+	 insn->bits1.da16.dest_writemask = dest.dw1.bits.writemask;
+      }
+   }
+   else {
+      insn->bits1.ia1.dest_subreg_nr = dest.subnr;
+
+      /* These are different sizes in align1 vs align16:
+       */
+      if (insn->header.access_mode == BRW_ALIGN_1) {
+	 insn->bits1.ia1.dest_indirect_offset = dest.dw1.bits.indirect_offset;
+	 insn->bits1.ia1.dest_horiz_stride = BRW_HORIZONTAL_STRIDE_1;
+      }
+      else {
+	 insn->bits1.ia16.dest_indirect_offset = dest.dw1.bits.indirect_offset;
+      }
+   }
+
+   /* NEW: Set the execution size based on dest.width and
+    * insn->compression_control:
+    */
+   guess_execution_size(insn, dest);
+}
+
+static void brw_set_src0( struct brw_instruction *insn,
+		      struct brw_reg reg )
+{
+   assert(reg.file != BRW_MESSAGE_REGISTER_FILE);
+
+   insn->bits1.da1.src0_reg_file = reg.file;
+   insn->bits1.da1.src0_reg_type = reg.type;
+   insn->bits2.da1.src0_abs = reg.abs;
+   insn->bits2.da1.src0_negate = reg.negate;
+   insn->bits2.da1.src0_address_mode = reg.address_mode;
+
+   if (reg.file == BRW_IMMEDIATE_VALUE) {
+      insn->bits3.ud = reg.dw1.ud;
+
+      /* Required to set some fields in src1 as well:
+       */
+      insn->bits1.da1.src1_reg_file = 0; /* arf */
+      insn->bits1.da1.src1_reg_type = reg.type;
+   }
+   else
+   {
+      if (reg.address_mode == BRW_ADDRESS_DIRECT) {
+	 if (insn->header.access_mode == BRW_ALIGN_1) {
+	    insn->bits2.da1.src0_subreg_nr = reg.subnr;
+	    insn->bits2.da1.src0_reg_nr = reg.nr;
+	 }
+	 else {
+	    insn->bits2.da16.src0_subreg_nr = reg.subnr / 16;
+	    insn->bits2.da16.src0_reg_nr = reg.nr;
+	 }
+      }
+      else {
+	 insn->bits2.ia1.src0_subreg_nr = reg.subnr;
+
+	 if (insn->header.access_mode == BRW_ALIGN_1) {
+	    insn->bits2.ia1.src0_indirect_offset = reg.dw1.bits.indirect_offset;
+	 }
+	 else {
+	    insn->bits2.ia16.src0_subreg_nr = reg.dw1.bits.indirect_offset;
+	 }
+      }
+
+      if (insn->header.access_mode == BRW_ALIGN_1) {
+	 if (reg.width == BRW_WIDTH_1 &&
+	     insn->header.execution_size == BRW_EXECUTE_1) {
+	    insn->bits2.da1.src0_horiz_stride = BRW_HORIZONTAL_STRIDE_0;
+	    insn->bits2.da1.src0_width = BRW_WIDTH_1;
+	    insn->bits2.da1.src0_vert_stride = BRW_VERTICAL_STRIDE_0;
+	 }
+	 else {
+	    insn->bits2.da1.src0_horiz_stride = reg.hstride;
+	    insn->bits2.da1.src0_width = reg.width;
+	    insn->bits2.da1.src0_vert_stride = reg.vstride;
+	 }
+      }
+      else {
+	 insn->bits2.da16.src0_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X);
+	 insn->bits2.da16.src0_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y);
+	 insn->bits2.da16.src0_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z);
+	 insn->bits2.da16.src0_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W);
+
+	 /* This is an oddity of the fact we're using the same
+	  * descriptions for registers in align_16 as align_1:
+	  */
+	 if (reg.vstride == BRW_VERTICAL_STRIDE_8)
+	    insn->bits2.da16.src0_vert_stride = BRW_VERTICAL_STRIDE_4;
+	 else
+	    insn->bits2.da16.src0_vert_stride = reg.vstride;
+      }
+   }
+}
+
+
+void brw_set_src1( struct brw_instruction *insn,
+			  struct brw_reg reg )
+{
+   assert(reg.file != BRW_MESSAGE_REGISTER_FILE);
+
+   insn->bits1.da1.src1_reg_file = reg.file;
+   insn->bits1.da1.src1_reg_type = reg.type;
+   insn->bits3.da1.src1_abs = reg.abs;
+   insn->bits3.da1.src1_negate = reg.negate;
+
+   /* Only src1 can be immediate in two-argument instructions.
+    */
+   assert(insn->bits1.da1.src0_reg_file != BRW_IMMEDIATE_VALUE);
+
+   if (reg.file == BRW_IMMEDIATE_VALUE) {
+      insn->bits3.ud = reg.dw1.ud;
+   }
+   else {
+      /* This is a hardware restriction, which may or may not be lifted
+       * in the future:
+       */
+      assert (reg.address_mode == BRW_ADDRESS_DIRECT);
+      //assert (reg.file == BRW_GENERAL_REGISTER_FILE);
+
+      if (insn->header.access_mode == BRW_ALIGN_1) {
+	 insn->bits3.da1.src1_subreg_nr = reg.subnr;
+	 insn->bits3.da1.src1_reg_nr = reg.nr;
+      }
+      else {
+	 insn->bits3.da16.src1_subreg_nr = reg.subnr / 16;
+	 insn->bits3.da16.src1_reg_nr = reg.nr;
+      }
+
+      if (insn->header.access_mode == BRW_ALIGN_1) {
+	 if (reg.width == BRW_WIDTH_1 &&
+	     insn->header.execution_size == BRW_EXECUTE_1) {
+	    insn->bits3.da1.src1_horiz_stride = BRW_HORIZONTAL_STRIDE_0;
+	    insn->bits3.da1.src1_width = BRW_WIDTH_1;
+	    insn->bits3.da1.src1_vert_stride = BRW_VERTICAL_STRIDE_0;
+	 }
+	 else {
+	    insn->bits3.da1.src1_horiz_stride = reg.hstride;
+	    insn->bits3.da1.src1_width = reg.width;
+	    insn->bits3.da1.src1_vert_stride = reg.vstride;
+	 }
+      }
+      else {
+	 insn->bits3.da16.src1_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X);
+	 insn->bits3.da16.src1_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y);
+	 insn->bits3.da16.src1_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z);
+	 insn->bits3.da16.src1_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W);
+
+	 /* This is an oddity of the fact we're using the same
+	  * descriptions for registers in align_16 as align_1:
+	  */
+	 if (reg.vstride == BRW_VERTICAL_STRIDE_8)
+	    insn->bits3.da16.src1_vert_stride = BRW_VERTICAL_STRIDE_4;
+	 else
+	    insn->bits3.da16.src1_vert_stride = reg.vstride;
+      }
+   }
+}
+
+
+
+static void brw_set_math_message( struct brw_instruction *insn,
+				  unsigned msg_length,
+				  unsigned response_length,
+				  unsigned function,
+				  unsigned integer_type,
+				  boolean low_precision,
+				  boolean saturate,
+				  unsigned dataType )
+{
+   brw_set_src1(insn, brw_imm_d(0));
+
+   insn->bits3.math.function = function;
+   insn->bits3.math.int_type = integer_type;
+   insn->bits3.math.precision = low_precision;
+   insn->bits3.math.saturate = saturate;
+   insn->bits3.math.data_type = dataType;
+   insn->bits3.math.response_length = response_length;
+   insn->bits3.math.msg_length = msg_length;
+   insn->bits3.math.msg_target = BRW_MESSAGE_TARGET_MATH;
+   insn->bits3.math.end_of_thread = 0;
+}
+
+static void brw_set_urb_message( struct brw_instruction *insn,
+				 boolean allocate,
+				 boolean used,
+				 unsigned msg_length,
+				 unsigned response_length,
+				 boolean end_of_thread,
+				 boolean complete,
+				 unsigned offset,
+				 unsigned swizzle_control )
+{
+   brw_set_src1(insn, brw_imm_d(0));
+
+   insn->bits3.urb.opcode = 0;	/* ? */
+   insn->bits3.urb.offset = offset;
+   insn->bits3.urb.swizzle_control = swizzle_control;
+   insn->bits3.urb.allocate = allocate;
+   insn->bits3.urb.used = used;	/* ? */
+   insn->bits3.urb.complete = complete;
+   insn->bits3.urb.response_length = response_length;
+   insn->bits3.urb.msg_length = msg_length;
+   insn->bits3.urb.msg_target = BRW_MESSAGE_TARGET_URB;
+   insn->bits3.urb.end_of_thread = end_of_thread;
+}
+
+static void brw_set_dp_write_message( struct brw_instruction *insn,
+				      unsigned binding_table_index,
+				      unsigned msg_control,
+				      unsigned msg_type,
+				      unsigned msg_length,
+				      unsigned pixel_scoreboard_clear,
+				      unsigned response_length,
+				      unsigned end_of_thread )
+{
+   brw_set_src1(insn, brw_imm_d(0));
+
+   insn->bits3.dp_write.binding_table_index = binding_table_index;
+   insn->bits3.dp_write.msg_control = msg_control;
+   insn->bits3.dp_write.pixel_scoreboard_clear = pixel_scoreboard_clear;
+   insn->bits3.dp_write.msg_type = msg_type;
+   insn->bits3.dp_write.send_commit_msg = 0;
+   insn->bits3.dp_write.response_length = response_length;
+   insn->bits3.dp_write.msg_length = msg_length;
+   insn->bits3.dp_write.msg_target = BRW_MESSAGE_TARGET_DATAPORT_WRITE;
+   insn->bits3.urb.end_of_thread = end_of_thread;
+}
+
+static void brw_set_dp_read_message( struct brw_instruction *insn,
+				      unsigned binding_table_index,
+				      unsigned msg_control,
+				      unsigned msg_type,
+				      unsigned target_cache,
+				      unsigned msg_length,
+				      unsigned response_length,
+				      unsigned end_of_thread )
+{
+   brw_set_src1(insn, brw_imm_d(0));
+
+   insn->bits3.dp_read.binding_table_index = binding_table_index;
+   insn->bits3.dp_read.msg_control = msg_control;
+   insn->bits3.dp_read.msg_type = msg_type;
+   insn->bits3.dp_read.target_cache = target_cache;
+   insn->bits3.dp_read.response_length = response_length;
+   insn->bits3.dp_read.msg_length = msg_length;
+   insn->bits3.dp_read.msg_target = BRW_MESSAGE_TARGET_DATAPORT_READ;
+   insn->bits3.dp_read.end_of_thread = end_of_thread;
+}
+
+static void brw_set_sampler_message( struct brw_instruction *insn,
+				     unsigned binding_table_index,
+				     unsigned sampler,
+				     unsigned msg_type,
+				     unsigned response_length,
+				     unsigned msg_length,
+				     boolean eot)
+{
+   brw_set_src1(insn, brw_imm_d(0));
+
+   insn->bits3.sampler.binding_table_index = binding_table_index;
+   insn->bits3.sampler.sampler = sampler;
+   insn->bits3.sampler.msg_type = msg_type;
+   insn->bits3.sampler.return_format = BRW_SAMPLER_RETURN_FORMAT_FLOAT32;
+   insn->bits3.sampler.response_length = response_length;
+   insn->bits3.sampler.msg_length = msg_length;
+   insn->bits3.sampler.end_of_thread = eot;
+   insn->bits3.sampler.msg_target = BRW_MESSAGE_TARGET_SAMPLER;
+}
+
+
+
+static struct brw_instruction *next_insn( struct brw_compile *p,
+					  unsigned opcode )
+{
+   struct brw_instruction *insn;
+
+   assert(p->nr_insn + 1 < BRW_EU_MAX_INSN);
+
+   insn = &p->store[p->nr_insn++];
+   memcpy(insn, p->current, sizeof(*insn));
+
+   /* Reset this one-shot flag:
+    */
+
+   if (p->current->header.destreg__conditonalmod) {
+      p->current->header.destreg__conditonalmod = 0;
+      p->current->header.predicate_control = BRW_PREDICATE_NORMAL;
+   }
+
+   insn->header.opcode = opcode;
+   return insn;
+}
+
+
+struct brw_instruction *brw_alu1( struct brw_compile *p,
+				  unsigned opcode,
+				  struct brw_reg dest,
+				  struct brw_reg src )
+{
+   struct brw_instruction *insn = next_insn(p, opcode);
+   brw_set_dest(insn, dest);
+   brw_set_src0(insn, src);
+   return insn;
+}
+
+struct brw_instruction *brw_alu2(struct brw_compile *p,
+				 unsigned opcode,
+				 struct brw_reg dest,
+				 struct brw_reg src0,
+				 struct brw_reg src1 )
+{
+   struct brw_instruction *insn = next_insn(p, opcode);
+   brw_set_dest(insn, dest);
+   brw_set_src0(insn, src0);
+   brw_set_src1(insn, src1);
+   return insn;
+}
+
+
+/***********************************************************************
+ * Convenience routines.
+ */
+#define ALU1(OP)					\
+struct brw_instruction *brw_##OP(struct brw_compile *p,			\
+	      struct brw_reg dest,			\
+	      struct brw_reg src0)   			\
+{							\
+   return brw_alu1(p, BRW_OPCODE_##OP, dest, src0);    	\
+}
+
+#define ALU2(OP)					\
+struct brw_instruction *brw_##OP(struct brw_compile *p,			\
+	      struct brw_reg dest,			\
+	      struct brw_reg src0,			\
+	      struct brw_reg src1)   			\
+{							\
+   return brw_alu2(p, BRW_OPCODE_##OP, dest, src0, src1);	\
+}
+
+
+ALU1(MOV)
+ALU2(SEL)
+ALU1(NOT)
+ALU2(AND)
+ALU2(OR)
+ALU2(XOR)
+ALU2(SHR)
+ALU2(SHL)
+ALU2(RSR)
+ALU2(RSL)
+ALU2(ASR)
+ALU2(ADD)
+ALU2(MUL)
+ALU1(FRC)
+ALU1(RNDD)
+ALU2(MAC)
+ALU2(MACH)
+ALU1(LZD)
+ALU2(DP4)
+ALU2(DPH)
+ALU2(DP3)
+ALU2(DP2)
+ALU2(LINE)
+
+
+
+
+void brw_NOP(struct brw_compile *p)
+{
+   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_NOP);
+   brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
+   brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
+   brw_set_src1(insn, brw_imm_ud(0x0));
+}
+
+
+
+
+
+/***********************************************************************
+ * Comparisons, if/else/endif
+ */
+
+struct brw_instruction *brw_JMPI(struct brw_compile *p,
+	      struct brw_reg dest,
+	      struct brw_reg src0,
+	      struct brw_reg src1)
+{
+   struct brw_instruction *insn = brw_alu2(p, BRW_OPCODE_JMPI, dest, src0, src1);
+
+   p->current->header.predicate_control = BRW_PREDICATE_NONE;
+
+   return insn;
+}
+
+/* EU takes the value from the flag register and pushes it onto some
+ * sort of a stack (presumably merging with any flag value already on
+ * the stack).  Within an if block, the flags at the top of the stack
+ * control execution on each channel of the unit, eg. on each of the
+ * 16 pixel values in our wm programs.
+ *
+ * When the matching 'else' instruction is reached (presumably by
+ * countdown of the instruction count patched in by our ELSE/ENDIF
+ * functions), the relevent flags are inverted.
+ *
+ * When the matching 'endif' instruction is reached, the flags are
+ * popped off.  If the stack is now empty, normal execution resumes.
+ *
+ * No attempt is made to deal with stack overflow (14 elements?).
+ */
+struct brw_instruction *brw_IF(struct brw_compile *p, unsigned execute_size)
+{
+   struct brw_instruction *insn;
+
+   if (p->single_program_flow) {
+      assert(execute_size == BRW_EXECUTE_1);
+
+      insn = next_insn(p, BRW_OPCODE_ADD);
+      insn->header.predicate_inverse = 1;
+   } else {
+      insn = next_insn(p, BRW_OPCODE_IF);
+   }
+
+   /* Override the defaults for this instruction:
+    */
+   brw_set_dest(insn, brw_ip_reg());
+   brw_set_src0(insn, brw_ip_reg());
+   brw_set_src1(insn, brw_imm_d(0x0));
+
+   insn->header.execution_size = execute_size;
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+   insn->header.predicate_control = BRW_PREDICATE_NORMAL;
+   insn->header.mask_control = BRW_MASK_ENABLE;
+
+   p->current->header.predicate_control = BRW_PREDICATE_NONE;
+
+   return insn;
+}
+
+
+struct brw_instruction *brw_ELSE(struct brw_compile *p,
+				 struct brw_instruction *if_insn)
+{
+   struct brw_instruction *insn;
+
+   if (p->single_program_flow) {
+      insn = next_insn(p, BRW_OPCODE_ADD);
+   } else {
+      insn = next_insn(p, BRW_OPCODE_ELSE);
+   }
+
+   brw_set_dest(insn, brw_ip_reg());
+   brw_set_src0(insn, brw_ip_reg());
+   brw_set_src1(insn, brw_imm_d(0x0));
+
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+   insn->header.execution_size = if_insn->header.execution_size;
+   insn->header.mask_control = BRW_MASK_ENABLE;
+
+   /* Patch the if instruction to point at this instruction.
+    */
+   if (p->single_program_flow) {
+      assert(if_insn->header.opcode == BRW_OPCODE_ADD);
+
+      if_insn->bits3.ud = (insn - if_insn + 1) * 16;
+   } else {
+      assert(if_insn->header.opcode == BRW_OPCODE_IF);
+
+      if_insn->bits3.if_else.jump_count = insn - if_insn;
+      if_insn->bits3.if_else.pop_count = 1;
+      if_insn->bits3.if_else.pad0 = 0;
+   }
+
+   return insn;
+}
+
+void brw_ENDIF(struct brw_compile *p,
+	       struct brw_instruction *patch_insn)
+{
+   if (p->single_program_flow) {
+      /* In single program flow mode, there's no need to execute an ENDIF,
+       * since we don't need to do any stack operations, and if we're executing
+       * currently, we want to just continue executing.
+       */
+      struct brw_instruction *next = &p->store[p->nr_insn];
+
+      assert(patch_insn->header.opcode == BRW_OPCODE_ADD);
+
+      patch_insn->bits3.ud = (next - patch_insn) * 16;
+   } else {
+      struct brw_instruction *insn = next_insn(p, BRW_OPCODE_ENDIF);
+
+      brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
+      brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
+      brw_set_src1(insn, brw_imm_d(0x0));
+
+      insn->header.compression_control = BRW_COMPRESSION_NONE;
+      insn->header.execution_size = patch_insn->header.execution_size;
+      insn->header.mask_control = BRW_MASK_ENABLE;
+
+      assert(patch_insn->bits3.if_else.jump_count == 0);
+
+      /* Patch the if or else instructions to point at this or the next
+       * instruction respectively.
+       */
+      if (patch_insn->header.opcode == BRW_OPCODE_IF) {
+	 /* Automagically turn it into an IFF:
+	  */
+	 patch_insn->header.opcode = BRW_OPCODE_IFF;
+	 patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1;
+	 patch_insn->bits3.if_else.pop_count = 0;
+	 patch_insn->bits3.if_else.pad0 = 0;
+      } else if (patch_insn->header.opcode == BRW_OPCODE_ELSE) {
+	 patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1;
+	 patch_insn->bits3.if_else.pop_count = 1;
+	 patch_insn->bits3.if_else.pad0 = 0;
+      } else {
+	 assert(0);
+      }
+
+      /* Also pop item off the stack in the endif instruction:
+       */
+      insn->bits3.if_else.jump_count = 0;
+      insn->bits3.if_else.pop_count = 1;
+      insn->bits3.if_else.pad0 = 0;
+   }
+}
+
+struct brw_instruction *brw_BREAK(struct brw_compile *p)
+{
+   struct brw_instruction *insn;
+   insn = next_insn(p, BRW_OPCODE_BREAK);
+   brw_set_dest(insn, brw_ip_reg());
+   brw_set_src0(insn, brw_ip_reg());
+   brw_set_src1(insn, brw_imm_d(0x0));
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+   insn->header.execution_size = BRW_EXECUTE_8;
+   insn->header.mask_control = BRW_MASK_DISABLE;
+   insn->bits3.if_else.pad0 = 0;
+   return insn;
+}
+
+struct brw_instruction *brw_CONT(struct brw_compile *p)
+{
+   struct brw_instruction *insn;
+   insn = next_insn(p, BRW_OPCODE_CONTINUE);
+   brw_set_dest(insn, brw_ip_reg());
+   brw_set_src0(insn, brw_ip_reg());
+   brw_set_src1(insn, brw_imm_d(0x0));
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+   insn->header.execution_size = BRW_EXECUTE_8;
+   insn->header.mask_control = BRW_MASK_DISABLE;
+   insn->bits3.if_else.pad0 = 0;
+   return insn;
+}
+
+/* DO/WHILE loop:
+ */
+struct brw_instruction *brw_DO(struct brw_compile *p, unsigned execute_size)
+{
+   if (p->single_program_flow) {
+      return &p->store[p->nr_insn];
+   } else {
+      struct brw_instruction *insn = next_insn(p, BRW_OPCODE_DO);
+
+      /* Override the defaults for this instruction:
+       */
+      brw_set_dest(insn, brw_null_reg());
+      brw_set_src0(insn, brw_null_reg());
+      brw_set_src1(insn, brw_null_reg());
+
+      insn->header.compression_control = BRW_COMPRESSION_NONE;
+      insn->header.execution_size = execute_size;
+      insn->header.predicate_control = BRW_PREDICATE_NONE;
+      /* insn->header.mask_control = BRW_MASK_ENABLE; */
+      insn->header.mask_control = BRW_MASK_DISABLE;
+
+      return insn;
+   }
+}
+
+
+
+struct brw_instruction *brw_WHILE(struct brw_compile *p,
+	       struct brw_instruction *do_insn)
+{
+   struct brw_instruction *insn;
+
+   if (p->single_program_flow)
+      insn = next_insn(p, BRW_OPCODE_ADD);
+   else
+      insn = next_insn(p, BRW_OPCODE_WHILE);
+
+   brw_set_dest(insn, brw_ip_reg());
+   brw_set_src0(insn, brw_ip_reg());
+   brw_set_src1(insn, brw_imm_d(0x0));
+
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+
+   if (p->single_program_flow) {
+      insn->header.execution_size = BRW_EXECUTE_1;
+
+      insn->bits3.d = (do_insn - insn) * 16;
+   } else {
+      insn->header.execution_size = do_insn->header.execution_size;
+
+      assert(do_insn->header.opcode == BRW_OPCODE_DO);
+      insn->bits3.if_else.jump_count = do_insn - insn;
+      insn->bits3.if_else.pop_count = 0;
+      insn->bits3.if_else.pad0 = 0;
+   }
+
+/*    insn->header.mask_control = BRW_MASK_ENABLE; */
+
+   insn->header.mask_control = BRW_MASK_DISABLE;
+   p->current->header.predicate_control = BRW_PREDICATE_NONE;
+   return insn;
+}
+
+
+/* FORWARD JUMPS:
+ */
+void brw_land_fwd_jump(struct brw_compile *p,
+		       struct brw_instruction *jmp_insn)
+{
+   struct brw_instruction *landing = &p->store[p->nr_insn];
+
+   assert(jmp_insn->header.opcode == BRW_OPCODE_JMPI);
+   assert(jmp_insn->bits1.da1.src1_reg_file = BRW_IMMEDIATE_VALUE);
+
+   jmp_insn->bits3.ud = (landing - jmp_insn) - 1;
+}
+
+
+
+/* To integrate with the above, it makes sense that the comparison
+ * instruction should populate the flag register.  It might be simpler
+ * just to use the flag reg for most WM tasks?
+ */
+void brw_CMP(struct brw_compile *p,
+	     struct brw_reg dest,
+	     unsigned conditional,
+	     struct brw_reg src0,
+	     struct brw_reg src1)
+{
+   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_CMP);
+
+   insn->header.destreg__conditonalmod = conditional;
+   brw_set_dest(insn, dest);
+   brw_set_src0(insn, src0);
+   brw_set_src1(insn, src1);
+
+/*    guess_execution_size(insn, src0); */
+
+
+   /* Make it so that future instructions will use the computed flag
+    * value until brw_set_predicate_control_flag_value() is called
+    * again.
+    */
+   if (dest.file == BRW_ARCHITECTURE_REGISTER_FILE &&
+       dest.nr == 0) {
+      p->current->header.predicate_control = BRW_PREDICATE_NORMAL;
+      p->flag_value = 0xff;
+   }
+}
+
+
+
+/***********************************************************************
+ * Helpers for the various SEND message types:
+ */
+
+/* Invert 8 values
+ */
+void brw_math( struct brw_compile *p,
+	       struct brw_reg dest,
+	       unsigned function,
+	       unsigned saturate,
+	       unsigned msg_reg_nr,
+	       struct brw_reg src,
+	       unsigned data_type,
+	       unsigned precision )
+{
+   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
+   unsigned msg_length = (function == BRW_MATH_FUNCTION_POW) ? 2 : 1;
+   unsigned response_length = (function == BRW_MATH_FUNCTION_SINCOS) ? 2 : 1;
+
+   /* Example code doesn't set predicate_control for send
+    * instructions.
+    */
+   insn->header.predicate_control = 0;
+   insn->header.destreg__conditonalmod = msg_reg_nr;
+
+   brw_set_dest(insn, dest);
+   brw_set_src0(insn, src);
+   brw_set_math_message(insn,
+			msg_length, response_length,
+			function,
+			BRW_MATH_INTEGER_UNSIGNED,
+			precision,
+			saturate,
+			data_type);
+}
+
+/* Use 2 send instructions to invert 16 elements
+ */
+void brw_math_16( struct brw_compile *p,
+		  struct brw_reg dest,
+		  unsigned function,
+		  unsigned saturate,
+		  unsigned msg_reg_nr,
+		  struct brw_reg src,
+		  unsigned precision )
+{
+   struct brw_instruction *insn;
+   unsigned msg_length = (function == BRW_MATH_FUNCTION_POW) ? 2 : 1;
+   unsigned response_length = (function == BRW_MATH_FUNCTION_SINCOS) ? 2 : 1;
+
+   /* First instruction:
+    */
+   brw_push_insn_state(p);
+   brw_set_predicate_control_flag_value(p, 0xff);
+   brw_set_compression_control(p, BRW_COMPRESSION_NONE);
+
+   insn = next_insn(p, BRW_OPCODE_SEND);
+   insn->header.destreg__conditonalmod = msg_reg_nr;
+
+   brw_set_dest(insn, dest);
+   brw_set_src0(insn, src);
+   brw_set_math_message(insn,
+			msg_length, response_length,
+			function,
+			BRW_MATH_INTEGER_UNSIGNED,
+			precision,
+			saturate,
+			BRW_MATH_DATA_VECTOR);
+
+   /* Second instruction:
+    */
+   insn = next_insn(p, BRW_OPCODE_SEND);
+   insn->header.compression_control = BRW_COMPRESSION_2NDHALF;
+   insn->header.destreg__conditonalmod = msg_reg_nr+1;
+
+   brw_set_dest(insn, offset(dest,1));
+   brw_set_src0(insn, src);
+   brw_set_math_message(insn,
+			msg_length, response_length,
+			function,
+			BRW_MATH_INTEGER_UNSIGNED,
+			precision,
+			saturate,
+			BRW_MATH_DATA_VECTOR);
+
+   brw_pop_insn_state(p);
+}
+
+
+
+
+void brw_dp_WRITE_16( struct brw_compile *p,
+		      struct brw_reg src,
+		      unsigned msg_reg_nr,
+		      unsigned scratch_offset )
+{
+   {
+      brw_push_insn_state(p);
+      brw_set_mask_control(p, BRW_MASK_DISABLE);
+      brw_set_compression_control(p, BRW_COMPRESSION_NONE);
+
+      brw_MOV(p,
+	      retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_D),
+	      brw_imm_d(scratch_offset));
+
+      brw_pop_insn_state(p);
+   }
+
+   {
+      unsigned msg_length = 3;
+      struct brw_reg dest = retype(brw_null_reg(), BRW_REGISTER_TYPE_UW);
+      struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
+
+      insn->header.predicate_control = 0; /* XXX */
+      insn->header.compression_control = BRW_COMPRESSION_NONE;
+      insn->header.destreg__conditonalmod = msg_reg_nr;
+
+      brw_set_dest(insn, dest);
+      brw_set_src0(insn, src);
+
+      brw_set_dp_write_message(insn,
+			       255, /* bti */
+			       BRW_DATAPORT_OWORD_BLOCK_4_OWORDS, /* msg_control */
+			       BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE, /* msg_type */
+			       msg_length,
+			       0, /* pixel scoreboard */
+			       0, /* response_length */
+			       0); /* eot */
+   }
+
+}
+
+
+void brw_dp_READ_16( struct brw_compile *p,
+		      struct brw_reg dest,
+		      unsigned msg_reg_nr,
+		      unsigned scratch_offset )
+{
+   {
+      brw_push_insn_state(p);
+      brw_set_compression_control(p, BRW_COMPRESSION_NONE);
+      brw_set_mask_control(p, BRW_MASK_DISABLE);
+
+      brw_MOV(p,
+	      retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_D),
+	      brw_imm_d(scratch_offset));
+
+      brw_pop_insn_state(p);
+   }
+
+   {
+      struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
+
+      insn->header.predicate_control = 0; /* XXX */
+      insn->header.compression_control = BRW_COMPRESSION_NONE;
+      insn->header.destreg__conditonalmod = msg_reg_nr;
+
+      brw_set_dest(insn, dest);	/* UW? */
+      brw_set_src0(insn, retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW));
+
+      brw_set_dp_read_message(insn,
+			      255, /* bti */
+			      3,  /* msg_control */
+			      BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */
+			      1, /* target cache */
+			      1, /* msg_length */
+			      2, /* response_length */
+			      0); /* eot */
+   }
+}
+
+
+void brw_fb_WRITE(struct brw_compile *p,
+		   struct brw_reg dest,
+		   unsigned msg_reg_nr,
+		   struct brw_reg src0,
+		   unsigned binding_table_index,
+		   unsigned msg_length,
+		   unsigned response_length,
+		   boolean eot)
+{
+   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
+
+   insn->header.predicate_control = 0; /* XXX */
+   insn->header.compression_control = BRW_COMPRESSION_NONE;
+   insn->header.destreg__conditonalmod = msg_reg_nr;
+
+   brw_set_dest(insn, dest);
+   brw_set_src0(insn, src0);
+   brw_set_dp_write_message(insn,
+			    binding_table_index,
+			    BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE, /* msg_control */
+			    BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE, /* msg_type */
+			    msg_length,
+			    1,	/* pixel scoreboard */
+			    response_length,
+			    eot);
+}
+
+
+
+void brw_SAMPLE(struct brw_compile *p,
+		struct brw_reg dest,
+		unsigned msg_reg_nr,
+		struct brw_reg src0,
+		unsigned binding_table_index,
+		unsigned sampler,
+		unsigned writemask,
+		unsigned msg_type,
+		unsigned response_length,
+		unsigned msg_length,
+		boolean eot)
+{
+   boolean need_stall = 0;
+
+   if(writemask == 0) {
+/*       debug_printf("%s: zero writemask??\n", __FUNCTION__); */
+      return;
+   }
+
+   /* Hardware doesn't do destination dependency checking on send
+    * instructions properly.  Add a workaround which generates the
+    * dependency by other means.  In practice it seems like this bug
+    * only crops up for texture samples, and only where registers are
+    * written by the send and then written again later without being
+    * read in between.  Luckily for us, we already track that
+    * information and use it to modify the writemask for the
+    * instruction, so that is a guide for whether a workaround is
+    * needed.
+    */
+   if (writemask != TGSI_WRITEMASK_XYZW) {
+      unsigned dst_offset = 0;
+      unsigned i, newmask = 0, len = 0;
+
+      for (i = 0; i < 4; i++) {
+	 if (writemask & (1<<i))
+	    break;
+	 dst_offset += 2;
+      }
+      for (; i < 4; i++) {
+	 if (!(writemask & (1<<i)))
+	    break;
+	 newmask |= 1<<i;
+	 len++;
+      }
+
+      if (newmask != writemask) {
+	 need_stall = 1;
+/* 	 debug_printf("need stall %x %x\n", newmask , writemask); */
+      }
+      else {
+	 struct brw_reg m1 = brw_message_reg(msg_reg_nr);
+
+	 newmask = ~newmask & TGSI_WRITEMASK_XYZW;
+
+	 brw_push_insn_state(p);
+
+	 brw_set_compression_control(p, BRW_COMPRESSION_NONE);
+	 brw_set_mask_control(p, BRW_MASK_DISABLE);
+
+	 brw_MOV(p, m1, brw_vec8_grf(0,0));
+  	 brw_MOV(p, get_element_ud(m1, 2), brw_imm_ud(newmask << 12));
+
+	 brw_pop_insn_state(p);
+
+  	 src0 = retype(brw_null_reg(), BRW_REGISTER_TYPE_UW);
+	 dest = offset(dest, dst_offset);
+	 response_length = len * 2;
+      }
+   }
+
+   {
+      struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
+
+      insn->header.predicate_control = 0; /* XXX */
+      insn->header.compression_control = BRW_COMPRESSION_NONE;
+      insn->header.destreg__conditonalmod = msg_reg_nr;
+
+      brw_set_dest(insn, dest);
+      brw_set_src0(insn, src0);
+      brw_set_sampler_message(insn,
+			      binding_table_index,
+			      sampler,
+			      msg_type,
+			      response_length,
+			      msg_length,
+			      eot);
+   }
+
+   if (need_stall)
+   {
+      struct brw_reg reg = vec8(offset(dest, response_length-1));
+
+      /*  mov (8) r9.0<1>:f    r9.0<8;8,1>:f    { Align1 }
+       */
+      brw_push_insn_state(p);
+      brw_set_compression_control(p, FALSE);
+      brw_MOV(p, reg, reg);
+      brw_pop_insn_state(p);
+   }
+
+}
+
+/* All these variables are pretty confusing - we might be better off
+ * using bitmasks and macros for this, in the old style.  Or perhaps
+ * just having the caller instantiate the fields in dword3 itself.
+ */
+void brw_urb_WRITE(struct brw_compile *p,
+		   struct brw_reg dest,
+		   unsigned msg_reg_nr,
+		   struct brw_reg src0,
+		   boolean allocate,
+		   boolean used,
+		   unsigned msg_length,
+		   unsigned response_length,
+		   boolean eot,
+		   boolean writes_complete,
+		   unsigned offset,
+		   unsigned swizzle)
+{
+   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND);
+
+   assert(msg_length < 16);
+
+   brw_set_dest(insn, dest);
+   brw_set_src0(insn, src0);
+   brw_set_src1(insn, brw_imm_d(0));
+
+   insn->header.destreg__conditonalmod = msg_reg_nr;
+
+   brw_set_urb_message(insn,
+		       allocate,
+		       used,
+		       msg_length,
+		       response_length,
+		       eot,
+		       writes_complete,
+		       offset,
+		       swizzle);
+}
+
diff --git a/src/gallium/drivers/i965simple/brw_eu_util.c b/src/gallium/drivers/i965simple/brw_eu_util.c
new file mode 100644
index 0000000000..3a65b141f0
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_eu_util.c
@@ -0,0 +1,126 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+      
+
+#include "brw_context.h"
+#include "brw_defines.h"
+#include "brw_eu.h"
+
+
+void brw_math_invert( struct brw_compile *p, 
+			     struct brw_reg dst,
+			     struct brw_reg src)
+{
+   brw_math( p, 
+	     dst,
+	     BRW_MATH_FUNCTION_INV, 
+	     BRW_MATH_SATURATE_NONE,
+	     0,
+	     src,
+	     BRW_MATH_PRECISION_FULL, 
+	     BRW_MATH_DATA_VECTOR );
+}
+
+
+
+void brw_copy4(struct brw_compile *p,
+	       struct brw_reg dst,
+	       struct brw_reg src,
+	       unsigned count)
+{
+   unsigned i;
+
+   dst = vec4(dst);
+   src = vec4(src);
+
+   for (i = 0; i < count; i++)
+   {
+      unsigned delta = i*32;
+      brw_MOV(p, byte_offset(dst, delta),    byte_offset(src, delta));
+      brw_MOV(p, byte_offset(dst, delta+16), byte_offset(src, delta+16));
+   }
+}
+
+
+void brw_copy8(struct brw_compile *p,
+	       struct brw_reg dst,
+	       struct brw_reg src,
+	       unsigned count)
+{
+   unsigned i;
+
+   dst = vec8(dst);
+   src = vec8(src);
+
+   for (i = 0; i < count; i++)
+   {
+      unsigned delta = i*32;
+      brw_MOV(p, byte_offset(dst, delta),    byte_offset(src, delta));
+   }
+}
+
+
+void brw_copy_indirect_to_indirect(struct brw_compile *p,
+				   struct brw_indirect dst_ptr,
+				   struct brw_indirect src_ptr,
+				   unsigned count)
+{
+   unsigned i;
+
+   for (i = 0; i < count; i++)
+   {
+      unsigned delta = i*32;
+      brw_MOV(p, deref_4f(dst_ptr, delta),    deref_4f(src_ptr, delta));
+      brw_MOV(p, deref_4f(dst_ptr, delta+16), deref_4f(src_ptr, delta+16));
+   }
+}
+
+
+void brw_copy_from_indirect(struct brw_compile *p,
+			    struct brw_reg dst,
+			    struct brw_indirect ptr,
+			    unsigned count)
+{
+   unsigned i;
+
+   dst = vec4(dst);
+
+   for (i = 0; i < count; i++)
+   {
+      unsigned delta = i*32;
+      brw_MOV(p, byte_offset(dst, delta),    deref_4f(ptr, delta));
+      brw_MOV(p, byte_offset(dst, delta+16), deref_4f(ptr, delta+16));
+   }
+}
+
+
+
+
diff --git a/src/gallium/drivers/i965simple/brw_flush.c b/src/gallium/drivers/i965simple/brw_flush.c
new file mode 100644
index 0000000000..e6001c30d9
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_flush.c
@@ -0,0 +1,73 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_defines.h"
+#include "brw_context.h"
+#include "brw_defines.h"
+#include "brw_batch.h"
+
+
+static void brw_flush( struct pipe_context *pipe,
+                       unsigned flags,
+                       struct pipe_fence_handle **fence )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   /* Do we need to emit an MI_FLUSH command to flush the hardware
+    * caches?
+    */
+   if (flags & (PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE)) {
+      struct brw_mi_flush flush;
+
+      memset(&flush, 0, sizeof(flush));      
+      flush.opcode = CMD_MI_FLUSH;
+
+      if (!(flags & PIPE_FLUSH_RENDER_CACHE))
+	 flush.flags |= BRW_INHIBIT_FLUSH_RENDER_CACHE;
+
+      if (flags & PIPE_FLUSH_TEXTURE_CACHE)
+	 flush.flags |= BRW_FLUSH_READ_CACHE;
+
+      BRW_BATCH_STRUCT(brw, &flush);
+   }
+
+   /* If there are no flags, just flush pending commands to hardware:
+    */
+   FLUSH_BATCH( fence );
+}
+
+
+
+void brw_init_flush_functions( struct brw_context *brw )
+{
+   brw->pipe.flush = brw_flush;
+}
diff --git a/src/gallium/drivers/i965simple/brw_gs.c b/src/gallium/drivers/i965simple/brw_gs.c
new file mode 100644
index 0000000000..de60868ccc
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_gs.c
@@ -0,0 +1,196 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_util.h"
+#include "brw_state.h"
+#include "brw_gs.h"
+
+
+
+static void compile_gs_prog( struct brw_context *brw,
+			     struct brw_gs_prog_key *key )
+{
+   struct brw_gs_compile c;
+   const unsigned *program;
+   unsigned program_size;
+
+   memset(&c, 0, sizeof(c));
+
+   c.key = *key;
+
+   /* Need to locate the two positions present in vertex + header.
+    * These are currently hardcoded:
+    */
+   c.nr_attrs = brw_count_bits(c.key.attrs);
+   c.nr_regs = (c.nr_attrs + 1) / 2 + 1;  /* are vertices packed, or reg-aligned? */
+   c.nr_bytes = c.nr_regs * REG_SIZE;
+
+
+   /* Begin the compilation:
+    */
+   brw_init_compile(&c.func);
+
+   c.func.single_program_flow = 1;
+
+   /* For some reason the thread is spawned with only 4 channels
+    * unmasked.
+    */
+   brw_set_mask_control(&c.func, BRW_MASK_DISABLE);
+
+
+   /* Note that primitives which don't require a GS program have
+    * already been weeded out by this stage:
+    */
+   switch (key->primitive) {
+   case PIPE_PRIM_QUADS:
+      brw_gs_quads( &c );
+      break;
+   case PIPE_PRIM_QUAD_STRIP:
+      brw_gs_quad_strip( &c );
+      break;
+   case PIPE_PRIM_LINE_LOOP:
+      brw_gs_lines( &c );
+      break;
+   case PIPE_PRIM_LINES:
+      if (key->hint_gs_always)
+	 brw_gs_lines( &c );
+      else {
+	 return;
+      }
+      break;
+   case PIPE_PRIM_TRIANGLES:
+      if (key->hint_gs_always)
+	 brw_gs_tris( &c );
+      else {
+	 return;
+      }
+      break;
+   case PIPE_PRIM_POINTS:
+      if (key->hint_gs_always)
+	 brw_gs_points( &c );
+      else {
+	 return;
+      }
+      break;
+   default:
+      return;
+   }
+
+   /* get the program
+    */
+   program = brw_get_program(&c.func, &program_size);
+
+   /* Upload
+    */
+   brw->gs.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_GS_PROG],
+					      &c.key,
+					      sizeof(c.key),
+					      program,
+					      program_size,
+					      &c.prog_data,
+					      &brw->gs.prog_data );
+}
+
+
+static boolean search_cache( struct brw_context *brw,
+			       struct brw_gs_prog_key *key )
+{
+   return brw_search_cache(&brw->cache[BRW_GS_PROG],
+			   key, sizeof(*key),
+			   &brw->gs.prog_data,
+			   &brw->gs.prog_gs_offset);
+}
+
+
+static const int gs_prim[PIPE_PRIM_POLYGON+1] = {
+   PIPE_PRIM_POINTS,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_LINE_LOOP,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_QUADS,
+   PIPE_PRIM_QUAD_STRIP,
+   PIPE_PRIM_TRIANGLES
+};
+
+static void populate_key( struct brw_context *brw,
+			  struct brw_gs_prog_key *key )
+{
+   memset(key, 0, sizeof(*key));
+
+   /* CACHE_NEW_VS_PROG */
+   key->attrs = brw->vs.prog_data->outputs_written;
+
+   /* BRW_NEW_PRIMITIVE */
+   key->primitive = gs_prim[brw->primitive];
+
+   key->hint_gs_always = 0;	/* debug code? */
+
+   key->need_gs_prog = (key->hint_gs_always ||
+			brw->primitive == PIPE_PRIM_QUADS ||
+			brw->primitive == PIPE_PRIM_QUAD_STRIP ||
+			brw->primitive == PIPE_PRIM_LINE_LOOP);
+}
+
+/* Calculate interpolants for triangle and line rasterization.
+ */
+static void upload_gs_prog( struct brw_context *brw )
+{
+   struct brw_gs_prog_key key;
+
+   /* Populate the key:
+    */
+   populate_key(brw, &key);
+
+   if (brw->gs.prog_active != key.need_gs_prog) {
+      brw->state.dirty.cache |= CACHE_NEW_GS_PROG;
+      brw->gs.prog_active = key.need_gs_prog;
+   }
+
+   if (brw->gs.prog_active) {
+      if (!search_cache(brw, &key))
+	 compile_gs_prog( brw, &key );
+   }
+}
+
+
+const struct brw_tracked_state brw_gs_prog = {
+   .dirty = {
+      .brw   = BRW_NEW_PRIMITIVE,
+      .cache = CACHE_NEW_VS_PROG
+   },
+   .update = upload_gs_prog
+};
diff --git a/src/gallium/drivers/i965simple/brw_gs.h b/src/gallium/drivers/i965simple/brw_gs.h
new file mode 100644
index 0000000000..f09141c6aa
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_gs.h
@@ -0,0 +1,75 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+ 
+
+#ifndef BRW_GS_H
+#define BRW_GS_H
+
+
+#include "brw_context.h"
+#include "brw_eu.h"
+
+#define MAX_GS_VERTS (4)	     
+
+struct brw_gs_prog_key {
+   unsigned attrs:32;
+   unsigned primitive:4;
+   unsigned hint_gs_always:1;
+   unsigned need_gs_prog:1;
+   unsigned pad:26;
+};
+
+struct brw_gs_compile {
+   struct brw_compile func;
+   struct brw_gs_prog_key key;
+   struct brw_gs_prog_data prog_data;
+   
+   struct {
+      struct brw_reg R0;
+      struct brw_reg vertex[MAX_GS_VERTS];
+   } reg;
+
+   /* 3 different ways of expressing vertex size:
+    */
+   unsigned nr_attrs;
+   unsigned nr_regs;
+   unsigned nr_bytes;
+};
+
+#define ATTR_SIZE  (4*4)
+
+void brw_gs_quads( struct brw_gs_compile *c );
+void brw_gs_quad_strip( struct brw_gs_compile *c );
+void brw_gs_tris( struct brw_gs_compile *c );
+void brw_gs_lines( struct brw_gs_compile *c );
+void brw_gs_points( struct brw_gs_compile *c );
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_gs_emit.c b/src/gallium/drivers/i965simple/brw_gs_emit.c
new file mode 100644
index 0000000000..c3cc90b10f
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_gs_emit.c
@@ -0,0 +1,148 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_util.h"
+#include "brw_gs.h"
+
+static void brw_gs_alloc_regs( struct brw_gs_compile *c,
+			       unsigned nr_verts )
+{
+   unsigned i = 0,j;
+
+   /* Register usage is static, precompute here:
+    */
+   c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++;
+
+   /* Payload vertices plus space for more generated vertices:
+    */
+   for (j = 0; j < nr_verts; j++) {
+      c->reg.vertex[j] = brw_vec4_grf(i, 0);
+      i += c->nr_regs;
+   }
+
+   c->prog_data.urb_read_length = c->nr_regs; 
+   c->prog_data.total_grf = i;
+}
+
+
+static void brw_gs_emit_vue(struct brw_gs_compile *c, 
+			    struct brw_reg vert,
+			    boolean last,
+			    unsigned header)
+{
+   struct brw_compile *p = &c->func;
+   boolean allocate = !last;
+
+   /* Overwrite PrimType and PrimStart in the message header, for
+    * each vertex in turn:
+    */
+   brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header));
+
+   /* Copy the vertex from vertn into m1..mN+1:
+    */
+   brw_copy8(p, brw_message_reg(1), vert, c->nr_regs);
+
+   /* Send each vertex as a seperate write to the urb.  This is
+    * different to the concept in brw_sf_emit.c, where subsequent
+    * writes are used to build up a single urb entry.  Each of these
+    * writes instantiates a seperate urb entry, and a new one must be
+    * allocated each time.
+    */
+   brw_urb_WRITE(p, 
+		 allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD),
+		 0,
+		 c->reg.R0,
+		 allocate,
+		 1,		/* used */
+		 c->nr_regs + 1, /* msg length */
+		 allocate ? 1 : 0, /* response length */
+		 allocate ? 0 : 1, /* eot */
+		 1,		/* writes_complete */
+		 0,		/* urb offset */
+		 BRW_URB_SWIZZLE_NONE);
+}
+
+
+
+void brw_gs_quads( struct brw_gs_compile *c )
+{
+   brw_gs_alloc_regs(c, 4);
+   
+   /* Use polygons for correct edgeflag behaviour. Note that vertex 3
+    * is the PV for quads, but vertex 0 for polygons:
+    */
+   brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START));
+   brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2));
+   brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); 
+   brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END));
+}
+
+void brw_gs_quad_strip( struct brw_gs_compile *c )
+{
+   brw_gs_alloc_regs(c, 4);
+   
+   brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START));
+   brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2));
+   brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); 
+   brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END));
+}
+
+void brw_gs_tris( struct brw_gs_compile *c )
+{
+   brw_gs_alloc_regs(c, 3);
+   brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_TRILIST << 2) | R02_PRIM_START));
+   brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_TRILIST << 2));
+   brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_TRILIST << 2) | R02_PRIM_END));
+}
+
+void brw_gs_lines( struct brw_gs_compile *c )
+{
+   brw_gs_alloc_regs(c, 2);
+   brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_START));
+   brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_END));
+}
+
+void brw_gs_points( struct brw_gs_compile *c )
+{
+   brw_gs_alloc_regs(c, 1);
+   brw_gs_emit_vue(c, c->reg.vertex[0], 1, ((_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END));
+}
+
+
+
+
+
+
+
+
diff --git a/src/gallium/drivers/i965simple/brw_gs_state.c b/src/gallium/drivers/i965simple/brw_gs_state.c
new file mode 100644
index 0000000000..5b8016b2e9
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_gs_state.c
@@ -0,0 +1,90 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+
+
+static void upload_gs_unit( struct brw_context *brw )
+{
+   struct brw_gs_unit_state gs;
+
+   memset(&gs, 0, sizeof(gs));
+
+   /* CACHE_NEW_GS_PROG */
+   if (brw->gs.prog_active) {
+      gs.thread0.grf_reg_count =
+	 align(brw->gs.prog_data->total_grf, 16) / 16 - 1;
+      gs.thread0.kernel_start_pointer = brw->gs.prog_gs_offset >> 6;
+      gs.thread3.urb_entry_read_length = brw->gs.prog_data->urb_read_length;
+   }
+   else {
+      gs.thread0.grf_reg_count = 0;
+      gs.thread0.kernel_start_pointer = 0;
+      gs.thread3.urb_entry_read_length = 1;
+   }
+
+   /* BRW_NEW_URB_FENCE */
+   gs.thread4.nr_urb_entries = brw->urb.nr_gs_entries;
+   gs.thread4.urb_entry_allocation_size = brw->urb.vsize - 1;
+
+   gs.thread4.max_threads = 0; /* Hardware requirement */
+
+   if (BRW_DEBUG & DEBUG_STATS)
+      gs.thread4.stats_enable = 1;
+
+   /* CONSTANT */
+   gs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
+   gs.thread1.single_program_flow = 1;
+   gs.thread3.dispatch_grf_start_reg = 1;
+   gs.thread3.const_urb_entry_read_offset = 0;
+   gs.thread3.const_urb_entry_read_length = 0;
+   gs.thread3.urb_entry_read_offset = 0;
+
+
+   brw->gs.state_gs_offset = brw_cache_data( &brw->cache[BRW_GS_UNIT], &gs );
+}
+
+
+const struct brw_tracked_state brw_gs_unit = {
+   .dirty = {
+      .brw   = (BRW_NEW_CURBE_OFFSETS |
+		BRW_NEW_URB_FENCE),
+      .cache = CACHE_NEW_GS_PROG
+   },
+   .update = upload_gs_unit
+};
diff --git a/src/gallium/drivers/i965simple/brw_misc_state.c b/src/gallium/drivers/i965simple/brw_misc_state.c
new file mode 100644
index 0000000000..99ff4403a5
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_misc_state.c
@@ -0,0 +1,488 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_batch.h"
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+
+
+
+
+
+/***********************************************************************
+ * Blend color
+ */
+
+static void upload_blend_constant_color(struct brw_context *brw)
+{
+   struct brw_blend_constant_color bcc;
+
+   memset(&bcc, 0, sizeof(bcc));
+   bcc.header.opcode = CMD_BLEND_CONSTANT_COLOR;
+   bcc.header.length = sizeof(bcc)/4-2;
+   bcc.blend_constant_color[0] = brw->attribs.BlendColor.color[0];
+   bcc.blend_constant_color[1] = brw->attribs.BlendColor.color[1];
+   bcc.blend_constant_color[2] = brw->attribs.BlendColor.color[2];
+   bcc.blend_constant_color[3] = brw->attribs.BlendColor.color[3];
+
+   BRW_CACHED_BATCH_STRUCT(brw, &bcc);
+}
+
+
+const struct brw_tracked_state brw_blend_constant_color = {
+   .dirty = {
+      .brw = BRW_NEW_BLEND,
+      .cache = 0
+   },
+   .update = upload_blend_constant_color
+};
+
+
+/***********************************************************************
+ * Drawing rectangle 
+ */
+static void upload_drawing_rect(struct brw_context *brw)
+{
+   struct brw_drawrect bdr;
+
+   memset(&bdr, 0, sizeof(bdr));
+   bdr.header.opcode = CMD_DRAW_RECT;
+   bdr.header.length = sizeof(bdr)/4 - 2;
+   bdr.xmin = 0;
+   bdr.ymin = 0;
+   bdr.xmax = brw->attribs.FrameBuffer.cbufs[0]->width;
+   bdr.ymax = brw->attribs.FrameBuffer.cbufs[0]->height;
+   bdr.xorg = 0;
+   bdr.yorg = 0;
+
+   /* Can't use BRW_CACHED_BATCH_STRUCT because this is also emitted
+    * uncached in brw_draw.c:
+    */
+   BRW_BATCH_STRUCT(brw, &bdr);
+}
+
+const struct brw_tracked_state brw_drawing_rect = {
+   .dirty = {
+      .brw = BRW_NEW_SCENE,
+      .cache = 0
+   },
+   .update = upload_drawing_rect
+};
+
+/**
+ * Upload the binding table pointers, which point each stage's array of surface
+ * state pointers.
+ *
+ * The binding table pointers are relative to the surface state base address,
+ * which is the BRW_SS_POOL cache buffer.
+ */
+static void upload_binding_table_pointers(struct brw_context *brw)
+{
+   struct brw_binding_table_pointers btp;
+   memset(&btp, 0, sizeof(btp));
+
+   btp.header.opcode = CMD_BINDING_TABLE_PTRS;
+   btp.header.length = sizeof(btp)/4 - 2;
+   btp.vs = 0;
+   btp.gs = 0;
+   btp.clp = 0;
+   btp.sf = 0;
+   btp.wm = brw->wm.bind_ss_offset;
+
+   BRW_CACHED_BATCH_STRUCT(brw, &btp);
+}
+
+const struct brw_tracked_state brw_binding_table_pointers = {
+   .dirty = {
+      .brw = 0,
+      .cache = CACHE_NEW_SURF_BIND
+   },
+   .update = upload_binding_table_pointers,
+};
+
+
+/**
+ * Upload pointers to the per-stage state.
+ *
+ * The state pointers in this packet are all relative to the general state
+ * base address set by CMD_STATE_BASE_ADDRESS, which is the BRW_GS_POOL buffer.
+ */
+static void upload_pipelined_state_pointers(struct brw_context *brw )
+{
+   struct brw_pipelined_state_pointers psp;
+   memset(&psp, 0, sizeof(psp));
+
+   psp.header.opcode = CMD_PIPELINED_STATE_POINTERS;
+   psp.header.length = sizeof(psp)/4 - 2;
+
+   psp.vs.offset = brw->vs.state_gs_offset >> 5;
+   psp.sf.offset = brw->sf.state_gs_offset >> 5;
+   psp.wm.offset = brw->wm.state_gs_offset >> 5;
+   psp.cc.offset = brw->cc.state_gs_offset >> 5;
+
+   /* GS gets turned on and off regularly.  Need to re-emit URB fence
+    * after this occurs.
+    */
+   if (brw->gs.prog_active) {
+      psp.gs.offset = brw->gs.state_gs_offset >> 5;
+      psp.gs.enable = 1;
+   }
+
+   if (0) {
+      psp.clp.offset = brw->clip.state_gs_offset >> 5;
+      psp.clp.enable = 1;
+   }
+
+
+   if (BRW_CACHED_BATCH_STRUCT(brw, &psp))
+      brw->state.dirty.brw |= BRW_NEW_PSP;
+}
+
+const struct brw_tracked_state brw_pipelined_state_pointers = {
+   .dirty = {
+      .brw = 0,
+      .cache = (CACHE_NEW_VS_UNIT |
+		CACHE_NEW_GS_UNIT |
+		CACHE_NEW_GS_PROG |
+		CACHE_NEW_CLIP_UNIT |
+		CACHE_NEW_SF_UNIT |
+		CACHE_NEW_WM_UNIT |
+		CACHE_NEW_CC_UNIT)
+   },
+   .update = upload_pipelined_state_pointers
+};
+
+static void upload_psp_urb_cbs(struct brw_context *brw )
+{
+   upload_pipelined_state_pointers(brw);
+   brw_upload_urb_fence(brw);
+   brw_upload_constant_buffer_state(brw);
+}
+
+
+const struct brw_tracked_state brw_psp_urb_cbs = {
+   .dirty = {
+      .brw = BRW_NEW_URB_FENCE,
+      .cache = (CACHE_NEW_VS_UNIT |
+		CACHE_NEW_GS_UNIT |
+		CACHE_NEW_GS_PROG |
+		CACHE_NEW_CLIP_UNIT |
+		CACHE_NEW_SF_UNIT |
+		CACHE_NEW_WM_UNIT |
+		CACHE_NEW_CC_UNIT)
+   },
+   .update = upload_psp_urb_cbs
+};
+
+/**
+ * Upload the depthbuffer offset and format.
+ *
+ * We have to do this per state validation as we need to emit the relocation
+ * in the batch buffer.
+ */
+static void upload_depthbuffer(struct brw_context *brw)
+{
+   struct pipe_surface *depth_surface = brw->attribs.FrameBuffer.zsbuf;
+
+   BEGIN_BATCH(5, INTEL_BATCH_NO_CLIPRECTS);
+   OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (5 - 2));
+   if (depth_surface == NULL) {
+      OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) |
+		(BRW_SURFACE_NULL << 29));
+      OUT_BATCH(0);
+      OUT_BATCH(0);
+      OUT_BATCH(0);
+   } else {
+      unsigned int format;
+      struct brw_texture *tex = (struct brw_texture *)depth_surface->texture;
+      assert(depth_surface->block.width == 1);
+      assert(depth_surface->block.height == 1);
+      switch (depth_surface->block.size) {
+      case 2:
+	 format = BRW_DEPTHFORMAT_D16_UNORM;
+	 break;
+      case 4:
+	 if (depth_surface->format == PIPE_FORMAT_Z32_FLOAT)
+	    format = BRW_DEPTHFORMAT_D32_FLOAT;
+	 else
+	    format = BRW_DEPTHFORMAT_D24_UNORM_S8_UINT;
+	 break;
+      default:
+	 assert(0);
+	 return;
+      }
+
+      OUT_BATCH((depth_surface->stride - 1) |
+		(format << 18) |
+		(BRW_TILEWALK_YMAJOR << 26) |
+//		(depth_surface->region->tiled << 27) |
+		(BRW_SURFACE_2D << 29));
+      OUT_RELOC(tex->buffer,
+		PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE, 0);
+      OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) |
+		((depth_surface->stride/depth_surface->block.size - 1) << 6) |
+		((depth_surface->height - 1) << 19));
+      OUT_BATCH(0);
+   }
+   ADVANCE_BATCH();
+}
+
+const struct brw_tracked_state brw_depthbuffer = {
+   .dirty = {
+      .brw = BRW_NEW_SCENE,
+      .cache = 0
+   },
+   .update = upload_depthbuffer,
+};
+
+
+
+
+/***********************************************************************
+ * Polygon stipple packet
+ */
+
+static void upload_polygon_stipple(struct brw_context *brw)
+{
+   struct brw_polygon_stipple bps;
+   unsigned i;
+
+   memset(&bps, 0, sizeof(bps));
+   bps.header.opcode = CMD_POLY_STIPPLE_PATTERN;
+   bps.header.length = sizeof(bps)/4-2;
+
+   /* XXX: state tracker should send *all* state down initially!
+    */
+   if (brw->attribs.PolygonStipple)
+      for (i = 0; i < 32; i++)
+	 bps.stipple[i] = brw->attribs.PolygonStipple->stipple[31 - i]; /* invert */
+
+   BRW_CACHED_BATCH_STRUCT(brw, &bps);
+}
+
+const struct brw_tracked_state brw_polygon_stipple = {
+   .dirty = {
+      .brw = BRW_NEW_STIPPLE,
+      .cache = 0
+   },
+   .update = upload_polygon_stipple
+};
+
+
+/***********************************************************************
+ * Line stipple packet
+ */
+
+static void upload_line_stipple(struct brw_context *brw)
+{
+   struct brw_line_stipple bls;
+   float tmp;
+   int tmpi;
+
+   memset(&bls, 0, sizeof(bls));
+   bls.header.opcode = CMD_LINE_STIPPLE_PATTERN;
+   bls.header.length = sizeof(bls)/4 - 2;
+
+   bls.bits0.pattern = brw->attribs.Raster->line_stipple_pattern;
+   bls.bits1.repeat_count = brw->attribs.Raster->line_stipple_factor;
+
+   tmp = 1.0 / (float) brw->attribs.Raster->line_stipple_factor;
+   tmpi = tmp * (1<<13);
+
+
+   bls.bits1.inverse_repeat_count = tmpi;
+
+   BRW_CACHED_BATCH_STRUCT(brw, &bls);
+}
+
+const struct brw_tracked_state brw_line_stipple = {
+   .dirty = {
+      .brw = BRW_NEW_STIPPLE,
+      .cache = 0
+   },
+   .update = upload_line_stipple
+};
+
+
+/***********************************************************************
+ * Misc constant state packets
+ */
+
+static void upload_pipe_control(struct brw_context *brw)
+{
+   struct brw_pipe_control pc;
+
+   return;
+
+   memset(&pc, 0, sizeof(pc));
+
+   pc.header.opcode = CMD_PIPE_CONTROL;
+   pc.header.length = sizeof(pc)/4 - 2;
+   pc.header.post_sync_operation = PIPE_CONTROL_NOWRITE;
+
+   pc.header.instruction_state_cache_flush_enable = 1;
+
+   pc.bits1.dest_addr_type = PIPE_CONTROL_GTTWRITE_GLOBAL;
+
+   BRW_BATCH_STRUCT(brw, &pc);
+}
+
+const struct brw_tracked_state brw_pipe_control = {
+   .dirty = {
+      .brw = BRW_NEW_SCENE,
+      .cache = 0
+   },
+   .update = upload_pipe_control
+};
+
+
+/***********************************************************************
+ * Misc invarient state packets
+ */
+
+static void upload_invarient_state( struct brw_context *brw )
+{
+   {
+      struct brw_mi_flush flush;
+
+      memset(&flush, 0, sizeof(flush));      
+      flush.opcode = CMD_MI_FLUSH;
+      flush.flags = BRW_FLUSH_STATE_CACHE | BRW_FLUSH_READ_CACHE;
+      BRW_BATCH_STRUCT(brw, &flush);
+   }
+
+   {
+      /* 0x61040000  Pipeline Select */
+      /*     PipelineSelect            : 0 */
+      struct brw_pipeline_select ps;
+
+      memset(&ps, 0, sizeof(ps));
+      ps.header.opcode = CMD_PIPELINE_SELECT;
+      ps.header.pipeline_select = 0;
+      BRW_BATCH_STRUCT(brw, &ps);
+   }
+
+   {
+      struct brw_global_depth_offset_clamp gdo;
+      memset(&gdo, 0, sizeof(gdo));
+
+      /* Disable depth offset clamping.
+       */
+      gdo.header.opcode = CMD_GLOBAL_DEPTH_OFFSET_CLAMP;
+      gdo.header.length = sizeof(gdo)/4 - 2;
+      gdo.depth_offset_clamp = 0.0;
+
+      BRW_BATCH_STRUCT(brw, &gdo);
+   }
+
+
+   /* 0x61020000  State Instruction Pointer */
+   {
+      struct brw_system_instruction_pointer sip;
+      memset(&sip, 0, sizeof(sip));
+
+      sip.header.opcode = CMD_STATE_INSN_POINTER;
+      sip.header.length = 0;
+      sip.bits0.pad = 0;
+      sip.bits0.system_instruction_pointer = 0;
+      BRW_BATCH_STRUCT(brw, &sip);
+   }
+
+
+   {
+      struct brw_vf_statistics vfs;
+      memset(&vfs, 0, sizeof(vfs));
+
+      vfs.opcode = CMD_VF_STATISTICS;
+      if (BRW_DEBUG & DEBUG_STATS)
+	 vfs.statistics_enable = 1;
+
+      BRW_BATCH_STRUCT(brw, &vfs);
+   }
+
+   
+   {
+      struct brw_polygon_stipple_offset bpso;
+      
+      memset(&bpso, 0, sizeof(bpso));
+      bpso.header.opcode = CMD_POLY_STIPPLE_OFFSET;
+      bpso.header.length = sizeof(bpso)/4-2;      
+      bpso.bits0.x_offset = 0;
+      bpso.bits0.y_offset = 0;
+
+      BRW_BATCH_STRUCT(brw, &bpso);
+   }
+}
+
+const struct brw_tracked_state brw_invarient_state = {
+   .dirty = {
+      .brw = BRW_NEW_SCENE,
+      .cache = 0
+   },
+   .update = upload_invarient_state
+};
+
+/**
+ * Define the base addresses which some state is referenced from.
+ *
+ * This allows us to avoid having to emit relocations in many places for
+ * cached state, and instead emit pointers inside of large, mostly-static
+ * state pools.  This comes at the expense of memory, and more expensive cache
+ * misses.
+ */
+static void upload_state_base_address( struct brw_context *brw )
+{
+   /* Output the structure (brw_state_base_address) directly to the
+    * batchbuffer, so we can emit relocations inline.
+    */
+   BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS);
+   OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2));
+   OUT_RELOC(brw->pool[BRW_GS_POOL].buffer,
+	     PIPE_BUFFER_USAGE_GPU_READ,
+	     1); /* General state base address */
+   OUT_RELOC(brw->pool[BRW_SS_POOL].buffer,
+	     PIPE_BUFFER_USAGE_GPU_READ,
+	     1); /* Surface state base address */
+   OUT_BATCH(1); /* Indirect object base address */
+   OUT_BATCH(1); /* General state upper bound */
+   OUT_BATCH(1); /* Indirect object upper bound */
+   ADVANCE_BATCH();
+}
+
+
+const struct brw_tracked_state brw_state_base_address = {
+   .dirty = {
+      .brw = BRW_NEW_SCENE,
+      .cache = 0
+   },
+   .update = upload_state_base_address
+};
diff --git a/src/gallium/drivers/i965simple/brw_reg.h b/src/gallium/drivers/i965simple/brw_reg.h
new file mode 100644
index 0000000000..9e885c3b3b
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_reg.h
@@ -0,0 +1,76 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#define CMD_MI				(0x0 << 29)
+#define CMD_2D				(0x2 << 29)
+#define CMD_3D				(0x3 << 29)
+
+#define MI_BATCH_BUFFER_END		(CMD_MI | 0xA << 23)
+
+/* Stalls command execution waiting for the given events to have occurred. */
+#define MI_WAIT_FOR_EVENT               (CMD_MI | (0x3 << 23))
+#define MI_WAIT_FOR_PLANE_B_FLIP        (1<<6)
+#define MI_WAIT_FOR_PLANE_A_FLIP        (1<<2)
+
+/* Primitive dispatch on 830-945 */
+#define _3DPRIMITIVE			(CMD_3D | (0x1f << 24))
+#define PRIM_INDIRECT            (1<<23)
+#define PRIM_INLINE              (0<<23)
+#define PRIM_INDIRECT_SEQUENTIAL (0<<17)
+#define PRIM_INDIRECT_ELTS       (1<<17)
+
+#define PRIM3D_TRILIST		(0x0<<18)
+#define PRIM3D_TRISTRIP 	(0x1<<18)
+#define PRIM3D_TRISTRIP_RVRSE	(0x2<<18)
+#define PRIM3D_TRIFAN		(0x3<<18)
+#define PRIM3D_POLY		(0x4<<18)
+#define PRIM3D_LINELIST 	(0x5<<18)
+#define PRIM3D_LINESTRIP	(0x6<<18)
+#define PRIM3D_RECTLIST 	(0x7<<18)
+#define PRIM3D_POINTLIST	(0x8<<18)
+#define PRIM3D_DIB		(0x9<<18)
+#define PRIM3D_MASK		(0x1f<<18)
+
+#define XY_SETUP_BLT_CMD		(CMD_2D | (0x01 << 22) | 6)
+
+#define XY_COLOR_BLT_CMD		(CMD_2D | (0x50 << 22) | 4)
+
+#define XY_SRC_COPY_BLT_CMD             (CMD_2D | (0x53 << 22) | 6)
+
+/* BR00 */
+#define XY_BLT_WRITE_ALPHA	(1 << 21)
+#define XY_BLT_WRITE_RGB	(1 << 20)
+#define XY_SRC_TILED		(1 << 15)
+#define XY_DST_TILED		(1 << 11)
+
+/* BR13 */
+#define BR13_565		(0x1 << 24)
+#define BR13_8888		(0x3 << 24)
+
+#define FENCE_LINEAR 0
+#define FENCE_XMAJOR 1
+#define FENCE_YMAJOR 2
diff --git a/src/gallium/drivers/i965simple/brw_screen.c b/src/gallium/drivers/i965simple/brw_screen.c
new file mode 100644
index 0000000000..b22e105f10
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_screen.c
@@ -0,0 +1,246 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "util/u_memory.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_string.h"
+#include "util/u_simple_screen.h"
+
+#include "brw_context.h"
+#include "brw_screen.h"
+#include "brw_tex_layout.h"
+
+
+static const char *
+brw_get_vendor( struct pipe_screen *screen )
+{
+   return "Tungsten Graphics, Inc.";
+}
+
+
+static const char *
+brw_get_name( struct pipe_screen *screen )
+{
+   static char buffer[128];
+   const char *chipset;
+
+   switch (brw_screen(screen)->pci_id) {
+   case PCI_CHIP_I965_Q:
+      chipset = "Intel(R) 965Q";
+      break;
+   case PCI_CHIP_I965_G:
+   case PCI_CHIP_I965_G_1:
+      chipset = "Intel(R) 965G";
+      break;
+   case PCI_CHIP_I965_GM:
+      chipset = "Intel(R) 965GM";
+      break;
+   case PCI_CHIP_I965_GME:
+      chipset = "Intel(R) 965GME/GLE";
+      break;
+   default:
+      chipset = "unknown";
+      break;
+   }
+
+   util_snprintf(buffer, sizeof(buffer), "i965 (chipset: %s)", chipset);
+   return buffer;
+}
+
+
+static int
+brw_get_param(struct pipe_screen *screen, int param)
+{
+   switch (param) {
+   case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+      return 8;
+   case PIPE_CAP_NPOT_TEXTURES:
+      return 1;
+   case PIPE_CAP_TWO_SIDED_STENCIL:
+      return 1;
+   case PIPE_CAP_GLSL:
+      return 0;
+   case PIPE_CAP_S3TC:
+      return 0;
+   case PIPE_CAP_ANISOTROPIC_FILTER:
+      return 0;
+   case PIPE_CAP_POINT_SPRITE:
+      return 0;
+   case PIPE_CAP_MAX_RENDER_TARGETS:
+      return 1;
+   case PIPE_CAP_OCCLUSION_QUERY:
+      return 0;
+   case PIPE_CAP_TEXTURE_SHADOW_MAP:
+      return 1;
+   case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+      return 11; /* max 1024x1024 */
+   case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+      return 8;  /* max 128x128x128 */
+   case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+      return 11; /* max 1024x1024 */
+   default:
+      return 0;
+   }
+}
+
+
+static float
+brw_get_paramf(struct pipe_screen *screen, int param)
+{
+   switch (param) {
+   case PIPE_CAP_MAX_LINE_WIDTH:
+      /* fall-through */
+   case PIPE_CAP_MAX_LINE_WIDTH_AA:
+      return 7.5;
+
+   case PIPE_CAP_MAX_POINT_WIDTH:
+      /* fall-through */
+   case PIPE_CAP_MAX_POINT_WIDTH_AA:
+      return 255.0;
+
+   case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+      return 4.0;
+
+   case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+      return 16.0;
+
+   default:
+      return 0;
+   }
+}
+
+
+static boolean
+brw_is_format_supported( struct pipe_screen *screen,
+                         enum pipe_format format, 
+                         enum pipe_texture_target target,
+                         unsigned tex_usage, 
+                         unsigned geom_flags )
+{
+#if 0
+   /* XXX: This is broken -- rewrite if still needed. */
+   static const unsigned tex_supported[] = {
+      PIPE_FORMAT_R8G8B8A8_UNORM,
+      PIPE_FORMAT_A8R8G8B8_UNORM,
+      PIPE_FORMAT_R5G6B5_UNORM,
+      PIPE_FORMAT_L8_UNORM,
+      PIPE_FORMAT_A8_UNORM,
+      PIPE_FORMAT_I8_UNORM,
+      PIPE_FORMAT_L8A8_UNORM,
+      PIPE_FORMAT_YCBCR,
+      PIPE_FORMAT_YCBCR_REV,
+      PIPE_FORMAT_S8_Z24,
+   };
+
+
+   /* Actually a lot more than this - add later:
+    */
+   static const unsigned render_supported[] = {
+      PIPE_FORMAT_A8R8G8B8_UNORM,
+      PIPE_FORMAT_R5G6B5_UNORM,
+   };
+
+   /*
+    */
+   static const unsigned z_stencil_supported[] = {
+      PIPE_FORMAT_Z16_UNORM,
+      PIPE_FORMAT_Z32_UNORM,
+      PIPE_FORMAT_S8Z24_UNORM,
+   };
+
+   switch (type) {
+   case PIPE_RENDER_FORMAT:
+      *numFormats = Elements(render_supported);
+      return render_supported;
+
+   case PIPE_TEX_FORMAT:
+      *numFormats = Elements(tex_supported);
+      return render_supported;
+
+   case PIPE_Z_STENCIL_FORMAT:
+      *numFormats = Elements(render_supported);
+      return render_supported;
+
+   default:
+      *numFormats = 0;
+      return NULL;
+   }
+#else
+   switch (format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+   case PIPE_FORMAT_R5G6B5_UNORM:
+   case PIPE_FORMAT_S8Z24_UNORM:
+      return TRUE;
+   default:
+      return FALSE;
+   };
+   return FALSE;
+#endif
+}
+
+
+static void
+brw_destroy_screen( struct pipe_screen *screen )
+{
+   struct pipe_winsys *winsys = screen->winsys;
+
+   if(winsys->destroy)
+      winsys->destroy(winsys);
+
+   FREE(screen);
+}
+
+
+/**
+ * Create a new brw_screen object
+ */
+struct pipe_screen *
+brw_create_screen(struct pipe_winsys *winsys, uint pci_id)
+{
+   struct brw_screen *brwscreen = CALLOC_STRUCT(brw_screen);
+
+   if (!brwscreen)
+      return NULL;
+
+   brwscreen->pci_id = pci_id;
+
+   brwscreen->screen.winsys = winsys;
+
+   brwscreen->screen.destroy = brw_destroy_screen;
+
+   brwscreen->screen.get_name = brw_get_name;
+   brwscreen->screen.get_vendor = brw_get_vendor;
+   brwscreen->screen.get_param = brw_get_param;
+   brwscreen->screen.get_paramf = brw_get_paramf;
+   brwscreen->screen.is_format_supported = brw_is_format_supported;
+
+   brw_init_screen_texture_funcs(&brwscreen->screen);
+   u_simple_screen_init(&brwscreen->screen);
+
+   return &brwscreen->screen;
+}
diff --git a/src/gallium/drivers/i965simple/brw_screen.h b/src/gallium/drivers/i965simple/brw_screen.h
new file mode 100644
index 0000000000..d3c70387e6
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_screen.h
@@ -0,0 +1,68 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef BRW_SCREEN_H
+#define BRW_SCREEN_H
+
+
+#include "pipe/p_screen.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/**
+ * Subclass of pipe_screen
+ */
+struct brw_screen
+{
+   struct pipe_screen screen;
+
+   uint pci_id;
+};
+
+
+/** cast wrapper */
+static INLINE struct brw_screen *
+brw_screen(struct pipe_screen *pscreen)
+{
+   return (struct brw_screen *) pscreen;
+}
+
+
+extern struct pipe_screen *
+brw_create_screen(struct pipe_winsys *winsys, uint pci_id);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* BRW_SCREEN_H */
diff --git a/src/gallium/drivers/i965simple/brw_sf.c b/src/gallium/drivers/i965simple/brw_sf.c
new file mode 100644
index 0000000000..b82a2e143b
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_sf.c
@@ -0,0 +1,351 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_util.h"
+#include "brw_sf.h"
+#include "brw_state.h"
+#include "tgsi/tgsi_parse.h"
+
+
+static void compile_sf_prog( struct brw_context *brw,
+			     struct brw_sf_prog_key *key )
+{
+   struct brw_sf_compile c;
+   const unsigned *program;
+   unsigned program_size;
+
+   memset(&c, 0, sizeof(c));
+
+   /* Begin the compilation:
+    */
+   brw_init_compile(&c.func);
+
+   c.key = *key;
+
+
+   c.nr_attrs = c.key.vp_output_count;
+   c.nr_attr_regs = (c.nr_attrs+1)/2;
+
+   c.nr_setup_attrs = c.key.fp_input_count + 1; /* +1 for position */
+   c.nr_setup_regs = (c.nr_setup_attrs+1)/2;
+
+   c.prog_data.urb_read_length = c.nr_attr_regs;
+   c.prog_data.urb_entry_size = c.nr_setup_regs * 2;
+
+
+   /* Which primitive?  Or all three?
+    */
+   switch (key->primitive) {
+   case SF_TRIANGLES:
+      c.nr_verts = 3;
+      brw_emit_tri_setup( &c );
+      break;
+   case SF_LINES:
+      c.nr_verts = 2;
+      brw_emit_line_setup( &c );
+      break;
+   case SF_POINTS:
+      c.nr_verts = 1;
+      brw_emit_point_setup( &c );
+      break;
+
+   case SF_UNFILLED_TRIS:
+   default:
+      assert(0);
+      return;
+   }
+
+
+
+   /* get the program
+    */
+   program = brw_get_program(&c.func, &program_size);
+
+   /* Upload
+    */
+   brw->sf.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_SF_PROG],
+					      &c.key,
+					      sizeof(c.key),
+					      program,
+					      program_size,
+					      &c.prog_data,
+					      &brw->sf.prog_data );
+}
+
+
+static boolean search_cache( struct brw_context *brw,
+			       struct brw_sf_prog_key *key )
+{
+   return brw_search_cache(&brw->cache[BRW_SF_PROG],
+			   key, sizeof(*key),
+			   &brw->sf.prog_data,
+			   &brw->sf.prog_gs_offset);
+}
+
+
+/* Calculate interpolants for triangle and line rasterization.
+ */
+static void upload_sf_prog( struct brw_context *brw )
+{
+   const struct brw_fragment_program *fs = brw->attribs.FragmentProgram;
+   struct brw_sf_prog_key key;
+   struct tgsi_parse_context parse;
+   int i, done = 0;
+
+
+   memset(&key, 0, sizeof(key));
+
+   /* Populate the key, noting state dependencies:
+    */
+   /* CACHE_NEW_VS_PROG */
+   key.vp_output_count = brw->vs.prog_data->outputs_written;
+
+   /* BRW_NEW_FS */
+   key.fp_input_count = brw->attribs.FragmentProgram->info.file_max[TGSI_FILE_INPUT] + 1;
+
+
+   /* BRW_NEW_REDUCED_PRIMITIVE */
+   switch (brw->reduced_primitive) {
+   case PIPE_PRIM_TRIANGLES:
+//      if (key.attrs & (1<<VERT_RESULT_EDGE))
+//	 key.primitive = SF_UNFILLED_TRIS;
+//      else
+      key.primitive = SF_TRIANGLES;
+      break;
+   case PIPE_PRIM_LINES:
+      key.primitive = SF_LINES;
+      break;
+   case PIPE_PRIM_POINTS:
+      key.primitive = SF_POINTS;
+      break;
+   }
+
+
+
+   /* Scan fp inputs to figure out what interpolation modes are
+    * required for each incoming vp output.  There is an assumption
+    * that the state tracker makes sure there is a 1:1 linkage between
+    * these sets of attributes (XXX: position??)
+    */
+   tgsi_parse_init( &parse, fs->program.tokens );
+   while( !done &&
+	  !tgsi_parse_end_of_tokens( &parse ) ) 
+   {
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+	 if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) 
+	 {
+	    int first = parse.FullToken.FullDeclaration.DeclarationRange.First;
+	    int last = parse.FullToken.FullDeclaration.DeclarationRange.Last;
+	    int interp_mode = parse.FullToken.FullDeclaration.Declaration.Interpolate;
+	    //int semantic = parse.FullToken.FullDeclaration.Semantic.SemanticName;
+	    //int semantic_index = parse.FullToken.FullDeclaration.Semantic.SemanticIndex;
+
+	    debug_printf("fs input %d..%d interp mode %d\n", first, last, interp_mode);
+	    
+	    switch (interp_mode) {
+	    case TGSI_INTERPOLATE_CONSTANT:
+	       for (i = first; i <= last; i++) 
+		  key.const_mask |= (1 << i);
+	       break;
+	    case TGSI_INTERPOLATE_LINEAR:
+	       for (i = first; i <= last; i++) 
+		  key.linear_mask |= (1 << i);
+	       break;
+	    case TGSI_INTERPOLATE_PERSPECTIVE:
+	       for (i = first; i <= last; i++) 
+		  key.persp_mask |= (1 << i);
+	       break;
+	    default:
+	       break;
+	    }
+
+	    /* Also need stuff for flat shading, twosided color.
+	     */
+
+	 }
+	 break;
+      default:
+	 done = 1;
+	 break;
+      }
+   }
+
+   /* Hack: Adjust for position.  Optimize away when not required (ie
+    * for perspective interpolation).
+    */
+   key.persp_mask <<= 1;
+   key.linear_mask <<= 1; 
+   key.linear_mask |= 1;
+   key.const_mask <<= 1;
+
+   debug_printf("key.persp_mask: %x\n", key.persp_mask);
+   debug_printf("key.linear_mask: %x\n", key.linear_mask);
+   debug_printf("key.const_mask: %x\n", key.const_mask);
+
+
+//   key.do_point_sprite = brw->attribs.Point->PointSprite;
+//   key.SpriteOrigin = brw->attribs.Point->SpriteOrigin;
+
+//   key.do_flat_shading = (brw->attribs.Raster->flatshade);
+//   key.do_twoside_color = (brw->attribs.Light->Enabled && brw->attribs.Light->Model.TwoSide);
+
+//   if (key.do_twoside_color)
+//      key.frontface_ccw = (brw->attribs.Polygon->FrontFace == GL_CCW);
+
+
+   if (!search_cache(brw, &key))
+      compile_sf_prog( brw, &key );
+}
+
+
+const struct brw_tracked_state brw_sf_prog = {
+   .dirty = {
+      .brw   = (BRW_NEW_RASTERIZER |
+		BRW_NEW_REDUCED_PRIMITIVE |
+		BRW_NEW_VS |
+		BRW_NEW_FS),
+      .cache = 0,
+   },
+   .update = upload_sf_prog
+};
+
+
+
+#if 0
+/* Build a struct like the one we'd like the state tracker to pass to
+ * us.
+ */
+static void update_sf_linkage( struct brw_context *brw )
+{
+   const struct brw_vertex_program *vs = brw->attribs.VertexProgram;
+   const struct brw_fragment_program *fs = brw->attribs.FragmentProgram;
+   struct pipe_setup_linkage state;
+   struct tgsi_parse_context parse;
+
+   int i, j;
+   int nr_vp_outputs = 0;
+   int done = 0;
+
+   struct { 
+      unsigned semantic:8;
+      unsigned semantic_index:16;
+   } fp_semantic[32], vp_semantic[32];
+
+   memset(&state, 0, sizeof(state));
+
+   state.fp_input_count = 0;
+
+
+
+   
+
+
+   assert(state.fp_input_count == fs->program.num_inputs);
+
+      
+   /* Then scan vp outputs
+    */
+   done = 0;
+   tgsi_parse_init( &parse, vs->program.tokens );
+   while( !done &&
+	  !tgsi_parse_end_of_tokens( &parse ) ) 
+   {
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+	 if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) 
+	 {
+	    int first = parse.FullToken.FullDeclaration.DeclarationRange.First;
+	    int last = parse.FullToken.FullDeclaration.DeclarationRange.Last;
+
+	    for (i = first; i < last; i++) {
+	       vp_semantic[i].semantic = 
+		  parse.FullToken.FullDeclaration.Semantic.SemanticName;
+	       vp_semantic[i].semantic_index = 
+		  parse.FullToken.FullDeclaration.Semantic.SemanticIndex;
+	    }
+	    
+	    assert(last > nr_vp_outputs);
+	    nr_vp_outputs = last;
+	 }
+	 break;
+      default:
+	 done = 1;
+	 break;
+      }
+   }
+
+
+   /* Now match based on semantic information.
+    */
+   for (i = 0; i< state.fp_input_count; i++) {
+      for (j = 0; j < nr_vp_outputs; j++) {
+	 if (fp_semantic[i].semantic == vp_semantic[j].semantic &&
+	     fp_semantic[i].semantic_index == vp_semantic[j].semantic_index) {
+	    state.fp_input[i].vp_output = j;
+	 }
+      }
+      if (fp_semantic[i].semantic == TGSI_SEMANTIC_COLOR) {
+	 for (j = 0; j < nr_vp_outputs; j++) {
+	    if (TGSI_SEMANTIC_BCOLOR == vp_semantic[j].semantic &&
+		fp_semantic[i].semantic_index == vp_semantic[j].semantic_index) {
+	       state.fp_input[i].bf_vp_output = j;
+	    }
+	 }
+      }
+   }
+
+   if (memcmp(&brw->sf.linkage, &state, sizeof(state)) != 0) {
+      brw->sf.linkage = state;
+      brw->state.dirty.brw |= BRW_NEW_SF_LINKAGE;
+   }
+}
+
+
+const struct brw_tracked_state brw_sf_linkage = {
+   .dirty = {
+      .brw   = (BRW_NEW_VS |
+		BRW_NEW_FS),
+      .cache = 0,
+   },
+   .update = update_sf_linkage
+};
+
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_sf.h b/src/gallium/drivers/i965simple/brw_sf.h
new file mode 100644
index 0000000000..b7ada47560
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_sf.h
@@ -0,0 +1,122 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#ifndef BRW_SF_H
+#define BRW_SF_H
+
+#include "brw_context.h"
+#include "brw_eu.h"
+
+
+#define SF_POINTS    0
+#define SF_LINES     1
+#define SF_TRIANGLES 2
+#define SF_UNFILLED_TRIS   3
+
+
+
+struct brw_sf_prog_key {
+   unsigned vp_output_count:5;
+   unsigned fp_input_count:5;
+
+   unsigned primitive:2;
+   unsigned do_twoside_color:1;
+   unsigned do_flat_shading:1;
+   unsigned frontface_ccw:1;
+   unsigned do_point_sprite:1;
+
+   /* Interpolation masks;
+    */
+   unsigned linear_mask;
+   unsigned persp_mask;
+   unsigned const_mask;
+
+
+//   int SpriteOrigin;
+};
+
+struct brw_sf_point_tex {
+	boolean CoordReplace;
+};
+
+struct brw_sf_compile {
+   struct brw_compile func;
+   struct brw_sf_prog_key key;
+   struct brw_sf_prog_data prog_data;
+
+   struct brw_reg pv;
+   struct brw_reg det;
+   struct brw_reg dx0;
+   struct brw_reg dx2;
+   struct brw_reg dy0;
+   struct brw_reg dy2;
+
+   /* z and 1/w passed in seperately:
+    */
+   struct brw_reg z[3];
+   struct brw_reg inv_w[3];
+
+   /* The vertices:
+    */
+   struct brw_reg vert[3];
+
+    /* Temporaries, allocated after last vertex reg.
+    */
+   struct brw_reg inv_det;
+   struct brw_reg a1_sub_a0;
+   struct brw_reg a2_sub_a0;
+   struct brw_reg tmp;
+
+   struct brw_reg m1Cx;
+   struct brw_reg m2Cy;
+   struct brw_reg m3C0;
+
+   unsigned nr_verts;
+   unsigned nr_attrs;
+   unsigned nr_attr_regs;
+   unsigned nr_setup_attrs;
+   unsigned nr_setup_regs;
+#if 0
+   ubyte attr_to_idx[VERT_RESULT_MAX];
+   ubyte idx_to_attr[VERT_RESULT_MAX];
+   struct brw_sf_point_tex point_attrs[VERT_RESULT_MAX];
+#endif
+};
+
+
+void brw_emit_tri_setup( struct brw_sf_compile *c );
+void brw_emit_line_setup( struct brw_sf_compile *c );
+void brw_emit_point_setup( struct brw_sf_compile *c );
+void brw_emit_point_sprite_setup( struct brw_sf_compile *c );
+void brw_emit_anyprim_setup( struct brw_sf_compile *c );
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_sf_emit.c b/src/gallium/drivers/i965simple/brw_sf_emit.c
new file mode 100644
index 0000000000..78d6fa5e9e
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_sf_emit.c
@@ -0,0 +1,382 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_defines.h"
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_util.h"
+#include "brw_sf.h"
+
+
+
+/***********************************************************************
+ * Triangle setup.
+ */
+
+
+static void alloc_regs( struct brw_sf_compile *c )
+{
+   unsigned reg, i;
+
+   /* Values computed by fixed function unit:
+    */
+   c->pv  = retype(brw_vec1_grf(1, 1), BRW_REGISTER_TYPE_UD);
+   c->det = brw_vec1_grf(1, 2);
+   c->dx0 = brw_vec1_grf(1, 3);
+   c->dx2 = brw_vec1_grf(1, 4);
+   c->dy0 = brw_vec1_grf(1, 5);
+   c->dy2 = brw_vec1_grf(1, 6);
+
+   /* z and 1/w passed in seperately:
+    */
+   c->z[0]     = brw_vec1_grf(2, 0);
+   c->inv_w[0] = brw_vec1_grf(2, 1);
+   c->z[1]     = brw_vec1_grf(2, 2);
+   c->inv_w[1] = brw_vec1_grf(2, 3);
+   c->z[2]     = brw_vec1_grf(2, 4);
+   c->inv_w[2] = brw_vec1_grf(2, 5);
+
+   /* The vertices:
+    */
+   reg = 3;
+   for (i = 0; i < c->nr_verts; i++) {
+      c->vert[i] = brw_vec8_grf(reg, 0);
+      reg += c->nr_attr_regs;
+   }
+
+   /* Temporaries, allocated after last vertex reg.
+    */
+   c->inv_det = brw_vec1_grf(reg, 0);  reg++;
+   c->a1_sub_a0 = brw_vec8_grf(reg, 0);  reg++;
+   c->a2_sub_a0 = brw_vec8_grf(reg, 0);  reg++;
+   c->tmp = brw_vec8_grf(reg, 0);  reg++;
+
+   /* Note grf allocation:
+    */
+   c->prog_data.total_grf = reg;
+
+
+   /* Outputs of this program - interpolation coefficients for
+    * rasterization:
+    */
+   c->m1Cx = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 1, 0);
+   c->m2Cy = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 2, 0);
+   c->m3C0 = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 3, 0);
+}
+
+
+static void copy_z_inv_w( struct brw_sf_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   unsigned i;
+
+   brw_push_insn_state(p);
+
+   /* Copy both scalars with a single MOV:
+    */
+   for (i = 0; i < c->nr_verts; i++)
+      brw_MOV(p, vec2(suboffset(c->vert[i], 2)), vec2(c->z[i]));
+
+   brw_pop_insn_state(p);
+}
+
+
+static void invert_det( struct brw_sf_compile *c)
+{
+   brw_math(&c->func,
+	    c->inv_det,
+	    BRW_MATH_FUNCTION_INV,
+	    BRW_MATH_SATURATE_NONE,
+	    0,
+	    c->det,
+	    BRW_MATH_DATA_SCALAR,
+	    BRW_MATH_PRECISION_FULL);
+
+}
+
+#define NON_PERPECTIVE_ATTRS  (FRAG_BIT_WPOS | \
+                               FRAG_BIT_COL0 | \
+			       FRAG_BIT_COL1)
+
+static boolean calculate_masks( struct brw_sf_compile *c,
+				  unsigned reg,
+				  ushort *pc,
+				  ushort *pc_persp,
+				  ushort *pc_linear)
+{
+   boolean is_last_attr = (reg == c->nr_setup_regs - 1);
+   unsigned persp_mask = c->key.persp_mask;
+   unsigned linear_mask = c->key.linear_mask;
+
+   debug_printf("persp_mask: %x\n", persp_mask);
+   debug_printf("linear_mask: %x\n", linear_mask);
+
+   *pc_persp = 0;
+   *pc_linear = 0;
+   *pc = 0xf;
+
+   if (persp_mask & (1 << (reg*2)))
+      *pc_persp = 0xf;
+
+   if (linear_mask & (1 << (reg*2)))
+      *pc_linear = 0xf;
+
+   /* Maybe only processs one attribute on the final round:
+    */
+   if (reg*2+1 < c->nr_setup_attrs) {
+      *pc |= 0xf0;
+
+      if (persp_mask & (1 << (reg*2+1)))
+	 *pc_persp |= 0xf0;
+
+      if (linear_mask & (1 << (reg*2+1)))
+	 *pc_linear |= 0xf0;
+   }
+
+   debug_printf("pc: %x\n", *pc);
+   debug_printf("pc_persp: %x\n", *pc_persp);
+   debug_printf("pc_linear: %x\n", *pc_linear);
+   
+
+   return is_last_attr;
+}
+
+
+
+void brw_emit_tri_setup( struct brw_sf_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   unsigned i;
+
+   debug_printf("%s START ==============\n", __FUNCTION__);
+
+   c->nr_verts = 3;
+   alloc_regs(c);
+   invert_det(c);
+   copy_z_inv_w(c);
+
+
+   for (i = 0; i < c->nr_setup_regs; i++)
+   {
+      /* Pair of incoming attributes:
+       */
+      struct brw_reg a0 = offset(c->vert[0], i);
+      struct brw_reg a1 = offset(c->vert[1], i);
+      struct brw_reg a2 = offset(c->vert[2], i);
+      ushort pc = 0, pc_persp = 0, pc_linear = 0;
+      boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear);
+
+      if (pc_persp)
+      {
+	 brw_set_predicate_control_flag_value(p, pc_persp);
+	 brw_MUL(p, a0, a0, c->inv_w[0]);
+	 brw_MUL(p, a1, a1, c->inv_w[1]);
+	 brw_MUL(p, a2, a2, c->inv_w[2]);
+      }
+
+
+      /* Calculate coefficients for interpolated values:
+       */
+      if (pc_linear)
+      {
+	 brw_set_predicate_control_flag_value(p, pc_linear);
+
+	 brw_ADD(p, c->a1_sub_a0, a1, negate(a0));
+	 brw_ADD(p, c->a2_sub_a0, a2, negate(a0));
+
+	 /* calculate dA/dx
+	  */
+	 brw_MUL(p, brw_null_reg(), c->a1_sub_a0, c->dy2);
+	 brw_MAC(p, c->tmp, c->a2_sub_a0, negate(c->dy0));
+	 brw_MUL(p, c->m1Cx, c->tmp, c->inv_det);
+
+	 /* calculate dA/dy
+	  */
+	 brw_MUL(p, brw_null_reg(), c->a2_sub_a0, c->dx0);
+	 brw_MAC(p, c->tmp, c->a1_sub_a0, negate(c->dx2));
+	 brw_MUL(p, c->m2Cy, c->tmp, c->inv_det);
+      }
+
+      {
+	 brw_set_predicate_control_flag_value(p, pc);
+	 /* start point for interpolation
+	  */
+	 brw_MOV(p, c->m3C0, a0);
+
+	 /* Copy m0..m3 to URB.  m0 is implicitly copied from r0 in
+	  * the send instruction:
+	  */
+	 brw_urb_WRITE(p,
+		       brw_null_reg(),
+		       0,
+		       brw_vec8_grf(0, 0), /* r0, will be copied to m0 */
+		       0, 	/* allocate */
+		       1,	/* used */
+		       4, 	/* msg len */
+		       0,	/* response len */
+		       last,	/* eot */
+		       last, 	/* writes complete */
+		       i*4,	/* offset */
+		       BRW_URB_SWIZZLE_TRANSPOSE); /* XXX: Swizzle control "SF to windower" */
+      }
+   }
+
+   debug_printf("%s DONE ==============\n", __FUNCTION__);
+
+}
+
+
+
+void brw_emit_line_setup( struct brw_sf_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   unsigned i;
+
+
+   c->nr_verts = 2;
+   alloc_regs(c);
+   invert_det(c);
+   copy_z_inv_w(c);
+
+   for (i = 0; i < c->nr_setup_regs; i++)
+   {
+      /* Pair of incoming attributes:
+       */
+      struct brw_reg a0 = offset(c->vert[0], i);
+      struct brw_reg a1 = offset(c->vert[1], i);
+      ushort pc, pc_persp, pc_linear;
+      boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear);
+
+      if (pc_persp)
+      {
+	 brw_set_predicate_control_flag_value(p, pc_persp);
+	 brw_MUL(p, a0, a0, c->inv_w[0]);
+	 brw_MUL(p, a1, a1, c->inv_w[1]);
+      }
+
+      /* Calculate coefficients for position, color:
+       */
+      if (pc_linear) {
+	 brw_set_predicate_control_flag_value(p, pc_linear);
+
+	 brw_ADD(p, c->a1_sub_a0, a1, negate(a0));
+
+ 	 brw_MUL(p, c->tmp, c->a1_sub_a0, c->dx0);
+	 brw_MUL(p, c->m1Cx, c->tmp, c->inv_det);
+
+	 brw_MUL(p, c->tmp, c->a1_sub_a0, c->dy0);
+	 brw_MUL(p, c->m2Cy, c->tmp, c->inv_det);
+      }
+
+      {
+	 brw_set_predicate_control_flag_value(p, pc);
+
+	 /* start point for interpolation
+	  */
+	 brw_MOV(p, c->m3C0, a0);
+
+	 /* Copy m0..m3 to URB.
+	  */
+	 brw_urb_WRITE(p,
+		       brw_null_reg(),
+		       0,
+		       brw_vec8_grf(0, 0),
+		       0, 	/* allocate */
+		       1, 	/* used */
+		       4, 	/* msg len */
+		       0,	/* response len */
+		       last, 	/* eot */
+		       last, 	/* writes complete */
+		       i*4,	/* urb destination offset */
+		       BRW_URB_SWIZZLE_TRANSPOSE);
+      }
+   }
+}
+
+
+/* Points setup - several simplifications as all attributes are
+ * constant across the face of the point (point sprites excluded!)
+ */
+void brw_emit_point_setup( struct brw_sf_compile *c )
+{
+   struct brw_compile *p = &c->func;
+   unsigned i;
+
+   c->nr_verts = 1;
+   alloc_regs(c);
+   copy_z_inv_w(c);
+
+   brw_MOV(p, c->m1Cx, brw_imm_ud(0)); /* zero - move out of loop */
+   brw_MOV(p, c->m2Cy, brw_imm_ud(0)); /* zero - move out of loop */
+
+   for (i = 0; i < c->nr_setup_regs; i++)
+   {
+      struct brw_reg a0 = offset(c->vert[0], i);
+      ushort pc, pc_persp, pc_linear;
+      boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear);
+
+      if (pc_persp)
+      {
+	 /* This seems odd as the values are all constant, but the
+	  * fragment shader will be expecting it:
+	  */
+	 brw_set_predicate_control_flag_value(p, pc_persp);
+	 brw_MUL(p, a0, a0, c->inv_w[0]);
+      }
+
+
+      /* The delta values are always zero, just send the starting
+       * coordinate.  Again, this is to fit in with the interpolation
+       * code in the fragment shader.
+       */
+      {
+	 brw_set_predicate_control_flag_value(p, pc);
+
+	 brw_MOV(p, c->m3C0, a0); /* constant value */
+
+	 /* Copy m0..m3 to URB.
+	  */
+	 brw_urb_WRITE(p,
+		       brw_null_reg(),
+		       0,
+		       brw_vec8_grf(0, 0),
+		       0, 	/* allocate */
+		       1,	/* used */
+		       4, 	/* msg len */
+		       0,	/* response len */
+		       last, 	/* eot */
+		       last, 	/* writes complete */
+		       i*4,	/* urb destination offset */
+		       BRW_URB_SWIZZLE_TRANSPOSE);
+      }
+   }
+}
diff --git a/src/gallium/drivers/i965simple/brw_sf_state.c b/src/gallium/drivers/i965simple/brw_sf_state.c
new file mode 100644
index 0000000000..2a5de61c21
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_sf_state.c
@@ -0,0 +1,181 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+
+static void upload_sf_vp(struct brw_context *brw)
+{
+   struct brw_sf_viewport sfv;
+
+   memset(&sfv, 0, sizeof(sfv));
+
+
+   /* BRW_NEW_VIEWPORT */
+   {
+      const float *scale = brw->attribs.Viewport.scale;
+      const float *trans = brw->attribs.Viewport.translate;
+
+      sfv.viewport.m00 = scale[0];
+      sfv.viewport.m11 = scale[1];
+      sfv.viewport.m22 = scale[2]; 
+      sfv.viewport.m30 = trans[0];
+      sfv.viewport.m31 = trans[1];
+      sfv.viewport.m32 = trans[2];
+   }
+
+   /* _NEW_SCISSOR */
+   sfv.scissor.xmin = brw->attribs.Scissor.minx;
+   sfv.scissor.xmax = brw->attribs.Scissor.maxx - 1;
+   sfv.scissor.ymin = brw->attribs.Scissor.miny;
+   sfv.scissor.ymax = brw->attribs.Scissor.maxy - 1;
+
+   brw->sf.vp_gs_offset = brw_cache_data( &brw->cache[BRW_SF_VP], &sfv );
+}
+
+const struct brw_tracked_state brw_sf_vp = {
+   .dirty = {
+      .brw   = (BRW_NEW_SCISSOR |
+		BRW_NEW_VIEWPORT),
+      .cache = 0
+   },
+   .update = upload_sf_vp
+};
+
+static void upload_sf_unit( struct brw_context *brw )
+{
+   struct brw_sf_unit_state sf;
+   memset(&sf, 0, sizeof(sf));
+
+   /* CACHE_NEW_SF_PROG */
+   sf.thread0.grf_reg_count = align(brw->sf.prog_data->total_grf, 16) / 16 - 1;
+   sf.thread0.kernel_start_pointer = brw->sf.prog_gs_offset >> 6;
+   sf.thread3.urb_entry_read_length = brw->sf.prog_data->urb_read_length;
+
+   sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
+   sf.thread3.dispatch_grf_start_reg = 3;
+   sf.thread3.urb_entry_read_offset = 1;
+
+   /* BRW_NEW_URB_FENCE */
+   sf.thread4.nr_urb_entries = brw->urb.nr_sf_entries;
+   sf.thread4.urb_entry_allocation_size = brw->urb.sfsize - 1;
+   sf.thread4.max_threads = MIN2(12, brw->urb.nr_sf_entries / 2) - 1;
+
+   if (BRW_DEBUG & DEBUG_SINGLE_THREAD)
+      sf.thread4.max_threads = 0;
+
+   if (BRW_DEBUG & DEBUG_STATS)
+      sf.thread4.stats_enable = 1;
+
+   /* CACHE_NEW_SF_VP */
+   sf.sf5.sf_viewport_state_offset = brw->sf.vp_gs_offset >> 5;
+   sf.sf5.viewport_transform = 1;
+
+   /* BRW_NEW_RASTER */
+   if (brw->attribs.Raster->scissor)
+      sf.sf6.scissor = 1;
+
+#if 0
+   if (brw->attribs.Polygon->FrontFace == GL_CCW)
+      sf.sf5.front_winding = BRW_FRONTWINDING_CCW;
+   else
+      sf.sf5.front_winding = BRW_FRONTWINDING_CW;
+
+
+   if (brw->attribs.Polygon->CullFlag) {
+      switch (brw->attribs.Polygon->CullFaceMode) {
+      case GL_FRONT:
+	 sf.sf6.cull_mode = BRW_CULLMODE_FRONT;
+	 break;
+      case GL_BACK:
+	 sf.sf6.cull_mode = BRW_CULLMODE_BACK;
+	 break;
+      case GL_FRONT_AND_BACK:
+	 sf.sf6.cull_mode = BRW_CULLMODE_BOTH;
+	 break;
+      default:
+	 assert(0);
+	 break;
+      }
+   }
+   else
+      sf.sf6.cull_mode = BRW_CULLMODE_NONE;
+#else
+   sf.sf5.front_winding = BRW_FRONTWINDING_CCW;
+   sf.sf6.cull_mode = BRW_CULLMODE_NONE;
+#endif
+
+   sf.sf6.line_width = CLAMP(brw->attribs.Raster->line_width, 1.0, 5.0) * (1<<1);
+
+   sf.sf6.line_endcap_aa_region_width = 1;
+   if (brw->attribs.Raster->line_smooth)
+      sf.sf6.aa_enable = 1;
+   else if (sf.sf6.line_width <= 0x2)
+       sf.sf6.line_width = 0;
+
+   sf.sf6.point_rast_rule = 1;	/* opengl conventions */
+
+   sf.sf7.sprite_point = brw->attribs.Raster->point_sprite;
+   sf.sf7.point_size = CLAMP(brw->attribs.Raster->line_width, 1.0, 255.0) * (1<<3);
+   sf.sf7.use_point_size_state = !brw->attribs.Raster->point_size_per_vertex;
+
+   /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons:
+    */
+   sf.sf7.trifan_pv = 2;
+   sf.sf7.linestrip_pv = 1;
+   sf.sf7.tristrip_pv = 2;
+   sf.sf7.line_last_pixel_enable = 0;
+
+   /* Set bias for OpenGL rasterization rules:
+    */
+   sf.sf6.dest_org_vbias = 0x8;
+   sf.sf6.dest_org_hbias = 0x8;
+
+   brw->sf.state_gs_offset = brw_cache_data( &brw->cache[BRW_SF_UNIT], &sf );
+}
+
+
+const struct brw_tracked_state brw_sf_unit = {
+   .dirty = {
+      .brw   = (BRW_NEW_RASTERIZER |
+		BRW_NEW_URB_FENCE),
+      .cache = (CACHE_NEW_SF_VP |
+		CACHE_NEW_SF_PROG)
+   },
+   .update = upload_sf_unit
+};
+
+
diff --git a/src/gallium/drivers/i965simple/brw_shader_info.c b/src/gallium/drivers/i965simple/brw_shader_info.c
new file mode 100644
index 0000000000..86d877d7ef
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_shader_info.c
@@ -0,0 +1,48 @@
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+
+
+/**
+ * XXX this obsolete new and no longer compiled.
+ */
+void brw_shader_info(const struct tgsi_token *tokens,
+		     struct brw_shader_info *info )
+{
+   struct tgsi_parse_context parse;
+   int done = 0;
+
+   tgsi_parse_init( &parse, tokens );
+
+   while( !done &&
+	  !tgsi_parse_end_of_tokens( &parse ) ) 
+   {
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+      {
+	 const struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration;
+	 unsigned last = decl->DeclarationRange.Last;
+      
+	 // Broken by crazy wpos init:
+	 //assert( info->nr_regs[decl->Declaration.File] <= last);
+
+	 info->nr_regs[decl->Declaration.File] = MAX2(info->nr_regs[decl->Declaration.File],
+						      last+1);
+	 break;
+      }
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+      default:
+	 done = 1;
+	 break;
+      }
+   }
+
+   tgsi_parse_free (&parse);
+   
+}
diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c
new file mode 100644
index 0000000000..b47f5373f3
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_state.c
@@ -0,0 +1,469 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/* Authors:  Zack Rusin <zack@tungstengraphics.com>
+ *           Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_parse.h"
+
+#include "brw_context.h"
+#include "brw_defines.h"
+#include "brw_state.h"
+#include "brw_draw.h"
+
+
+#define DUP( TYPE, VAL )                        \
+do {                                            \
+   struct TYPE *x = malloc(sizeof(*x));         \
+   memcpy(x, VAL, sizeof(*x) );                 \
+   return x;                                    \
+} while (0)
+
+/************************************************************************
+ * Blend 
+ */
+static void *
+brw_create_blend_state(struct pipe_context *pipe,
+                        const struct pipe_blend_state *blend)
+{   
+   DUP( pipe_blend_state, blend );
+}
+
+static void brw_bind_blend_state(struct pipe_context *pipe,
+                                 void *blend)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.Blend = (struct pipe_blend_state*)blend;
+   brw->state.dirty.brw |= BRW_NEW_BLEND;
+}
+
+
+static void brw_delete_blend_state(struct pipe_context *pipe, void *blend)
+{
+   free(blend);
+}
+
+static void brw_set_blend_color( struct pipe_context *pipe,
+			     const struct pipe_blend_color *blend_color )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.BlendColor = *blend_color;
+
+   brw->state.dirty.brw |= BRW_NEW_BLEND;
+}
+
+/************************************************************************
+ * Sampler 
+ */
+
+static void *
+brw_create_sampler_state(struct pipe_context *pipe,
+                          const struct pipe_sampler_state *sampler)
+{
+   DUP( pipe_sampler_state, sampler );
+}
+
+static void brw_bind_sampler_states(struct pipe_context *pipe,
+                                    unsigned num, void **sampler)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == brw->num_samplers &&
+       !memcmp(brw->attribs.Samplers, sampler, num * sizeof(void *)))
+      return;
+
+   memcpy(brw->attribs.Samplers, sampler, num * sizeof(void *));
+   memset(&brw->attribs.Samplers[num], 0, (PIPE_MAX_SAMPLERS - num) *
+          sizeof(void *));
+
+   brw->num_samplers = num;
+
+   brw->state.dirty.brw |= BRW_NEW_SAMPLER;
+}
+
+static void brw_delete_sampler_state(struct pipe_context *pipe,
+                                      void *sampler)
+{
+   free(sampler);
+}
+
+
+/************************************************************************
+ * Depth stencil 
+ */
+
+static void *
+brw_create_depth_stencil_state(struct pipe_context *pipe,
+                           const struct pipe_depth_stencil_alpha_state *depth_stencil)
+{
+   DUP( pipe_depth_stencil_alpha_state, depth_stencil );
+}
+
+static void brw_bind_depth_stencil_state(struct pipe_context *pipe,
+                                         void *depth_stencil)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil;
+
+   brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL;
+}
+
+static void brw_delete_depth_stencil_state(struct pipe_context *pipe,
+                                           void *depth_stencil)
+{
+   free(depth_stencil);
+}
+
+/************************************************************************
+ * Scissor
+ */
+static void brw_set_scissor_state( struct pipe_context *pipe,
+                                 const struct pipe_scissor_state *scissor )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) );
+   brw->state.dirty.brw |= BRW_NEW_SCISSOR;
+}
+
+
+/************************************************************************
+ * Stipple
+ */
+
+static void brw_set_polygon_stipple( struct pipe_context *pipe,
+                                   const struct pipe_poly_stipple *stipple )
+{
+}
+
+
+/************************************************************************
+ * Fragment shader
+ */
+
+static void * brw_create_fs_state(struct pipe_context *pipe,
+                                   const struct pipe_shader_state *shader)
+{
+   struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program);
+
+   brw_fp->program.tokens = tgsi_dup_tokens(shader->tokens);
+   brw_fp->id = brw_context(pipe)->program_id++;
+
+   tgsi_scan_shader(shader->tokens, &brw_fp->info);
+
+#if 0
+   brw_shader_info(shader->tokens,
+		   &brw_fp->info2);
+#endif
+
+   tgsi_dump(shader->tokens, 0);
+
+
+   return (void *)brw_fp;
+}
+
+static void brw_bind_fs_state(struct pipe_context *pipe, void *shader)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader;
+   brw->state.dirty.brw |= BRW_NEW_FS;
+}
+
+static void brw_delete_fs_state(struct pipe_context *pipe, void *shader)
+{
+   struct brw_fragment_program *brw_fp = (struct brw_fragment_program *) shader;
+
+   FREE((void *) brw_fp->program.tokens);
+   FREE(brw_fp);
+}
+
+
+/************************************************************************
+ * Vertex shader and other TNL state 
+ */
+
+static void *brw_create_vs_state(struct pipe_context *pipe,
+                                 const struct pipe_shader_state *shader)
+{
+   struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program);
+
+   brw_vp->program.tokens = tgsi_dup_tokens(shader->tokens);
+   brw_vp->id = brw_context(pipe)->program_id++;
+
+   tgsi_scan_shader(shader->tokens, &brw_vp->info);
+
+#if 0
+   brw_shader_info(shader->tokens,
+		   &brw_vp->info2);
+#endif
+   tgsi_dump(shader->tokens, 0);
+
+   return (void *)brw_vp;
+}
+
+static void brw_bind_vs_state(struct pipe_context *pipe, void *vs)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.VertexProgram = (struct brw_vertex_program *)vs;
+   brw->state.dirty.brw |= BRW_NEW_VS;
+
+   debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n");
+}
+
+static void brw_delete_vs_state(struct pipe_context *pipe, void *shader)
+{
+   struct brw_vertex_program *brw_vp = (struct brw_vertex_program *) shader;
+
+   FREE((void *) brw_vp->program.tokens);
+   FREE(brw_vp);
+}
+
+
+static void brw_set_clip_state( struct pipe_context *pipe,
+                                const struct pipe_clip_state *clip )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.Clip = *clip;
+}
+
+
+static void brw_set_viewport_state( struct pipe_context *pipe,
+				     const struct pipe_viewport_state *viewport )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.Viewport = *viewport; /* struct copy */
+   brw->state.dirty.brw |= BRW_NEW_VIEWPORT;
+
+   /* pass the viewport info to the draw module */
+   //draw_set_viewport_state(brw->draw, viewport);
+}
+
+
+static void brw_set_vertex_buffers(struct pipe_context *pipe,
+				   unsigned count,
+				   const struct pipe_vertex_buffer *buffers)
+{
+   struct brw_context *brw = brw_context(pipe);
+   memcpy(brw->vb.vbo_array, buffers, count * sizeof(buffers[0]));
+}
+
+static void brw_set_vertex_elements(struct pipe_context *pipe,
+                                    unsigned count,
+                                    const struct pipe_vertex_element *elements)
+{
+   /* flush ? */
+   struct brw_context *brw = brw_context(pipe);
+   uint i;
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   for (i = 0; i < count; i++) {
+      struct brw_vertex_element_state el;
+      memset(&el, 0, sizeof(el));
+
+      el.ve0.src_offset = elements[i].src_offset;
+      el.ve0.src_format = brw_translate_surface_format(elements[i].src_format);
+      el.ve0.valid = 1;
+      el.ve0.vertex_buffer_index = elements[i].vertex_buffer_index;
+
+      el.ve1.dst_offset   = i * 4;
+
+      el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC;
+      el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC;
+      el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC;
+      el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC;
+
+      switch (elements[i].nr_components) {
+      case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0;
+      case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0;
+      case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT;
+         break;
+      }
+
+      brw->vb.inputs[i] = el;
+   }
+}
+
+
+
+/************************************************************************
+ * Constant buffers
+ */
+
+static void brw_set_constant_buffer(struct pipe_context *pipe,
+                                     uint shader, uint index,
+                                     const struct pipe_constant_buffer *buf)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   assert(buf == 0 || index == 0);
+
+   brw->attribs.Constants[shader] = buf;
+   brw->state.dirty.brw |= BRW_NEW_CONSTANTS;
+}
+
+
+/************************************************************************
+ * Texture surfaces
+ */
+
+
+static void brw_set_sampler_textures(struct pipe_context *pipe,
+                                     unsigned num,
+                                     struct pipe_texture **texture)
+{
+   struct brw_context *brw = brw_context(pipe);
+   uint i;
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == brw->num_textures &&
+       !memcmp(brw->attribs.Texture, texture, num *
+               sizeof(struct pipe_texture *)))
+      return;
+
+   for (i = 0; i < num; i++)
+      pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i],
+                             texture[i]);
+
+   for (i = num; i < brw->num_textures; i++)
+      pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i],
+                             NULL);
+
+   brw->num_textures = num;
+
+   brw->state.dirty.brw |= BRW_NEW_TEXTURE;
+}
+
+
+/************************************************************************
+ * Render targets, etc
+ */
+
+static void brw_set_framebuffer_state(struct pipe_context *pipe,
+				       const struct pipe_framebuffer_state *fb)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.FrameBuffer = *fb; /* struct copy */
+
+   brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER;
+}
+
+
+
+/************************************************************************
+ * Rasterizer state
+ */
+
+static void *
+brw_create_rasterizer_state(struct pipe_context *pipe,
+                             const struct pipe_rasterizer_state *rasterizer)
+{
+   DUP(pipe_rasterizer_state, rasterizer);
+}
+
+static void brw_bind_rasterizer_state( struct pipe_context *pipe,
+                                        void *setup )
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   brw->attribs.Raster = (struct pipe_rasterizer_state *)setup;
+
+   /* Also pass-through to draw module:
+    */
+   //draw_set_rasterizer_state(brw->draw, setup);
+
+   brw->state.dirty.brw |= BRW_NEW_RASTERIZER;
+}
+
+static void brw_delete_rasterizer_state(struct pipe_context *pipe,
+                                         void *setup)
+{
+   free(setup);
+}
+
+
+
+void
+brw_init_state_functions( struct brw_context *brw )
+{
+   brw->pipe.create_blend_state = brw_create_blend_state;
+   brw->pipe.bind_blend_state = brw_bind_blend_state;
+   brw->pipe.delete_blend_state = brw_delete_blend_state;
+
+   brw->pipe.create_sampler_state = brw_create_sampler_state;
+   brw->pipe.bind_sampler_states = brw_bind_sampler_states;
+   brw->pipe.delete_sampler_state = brw_delete_sampler_state;
+
+   brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state;
+   brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state;
+   brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state;
+
+   brw->pipe.create_rasterizer_state = brw_create_rasterizer_state;
+   brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state;
+   brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state;
+   brw->pipe.create_fs_state = brw_create_fs_state;
+   brw->pipe.bind_fs_state = brw_bind_fs_state;
+   brw->pipe.delete_fs_state = brw_delete_fs_state;
+   brw->pipe.create_vs_state = brw_create_vs_state;
+   brw->pipe.bind_vs_state = brw_bind_vs_state;
+   brw->pipe.delete_vs_state = brw_delete_vs_state;
+
+   brw->pipe.set_blend_color = brw_set_blend_color;
+   brw->pipe.set_clip_state = brw_set_clip_state;
+   brw->pipe.set_constant_buffer = brw_set_constant_buffer;
+   brw->pipe.set_framebuffer_state = brw_set_framebuffer_state;
+
+//   brw->pipe.set_feedback_state = brw_set_feedback_state;
+//   brw->pipe.set_feedback_buffer = brw_set_feedback_buffer;
+
+   brw->pipe.set_polygon_stipple = brw_set_polygon_stipple;
+   brw->pipe.set_scissor_state = brw_set_scissor_state;
+   brw->pipe.set_sampler_textures = brw_set_sampler_textures;
+   brw->pipe.set_viewport_state = brw_set_viewport_state;
+   brw->pipe.set_vertex_buffers = brw_set_vertex_buffers;
+   brw->pipe.set_vertex_elements = brw_set_vertex_elements;
+}
diff --git a/src/gallium/drivers/i965simple/brw_state.h b/src/gallium/drivers/i965simple/brw_state.h
new file mode 100644
index 0000000000..de0a6371b8
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_state.h
@@ -0,0 +1,151 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+    
+
+#ifndef BRW_STATE_H
+#define BRW_STATE_H
+
+#include "brw_context.h"
+#include "brw_winsys.h"
+
+
+const struct brw_tracked_state brw_blend_constant_color;
+const struct brw_tracked_state brw_cc_unit;
+const struct brw_tracked_state brw_cc_vp;
+const struct brw_tracked_state brw_clip_prog;
+const struct brw_tracked_state brw_clip_unit;
+const struct brw_tracked_state brw_constant_buffer_state;
+const struct brw_tracked_state brw_constant_buffer;
+const struct brw_tracked_state brw_curbe_offsets;
+const struct brw_tracked_state brw_invarient_state;
+const struct brw_tracked_state brw_gs_prog;
+const struct brw_tracked_state brw_gs_unit;
+const struct brw_tracked_state brw_drawing_rect;
+const struct brw_tracked_state brw_line_stipple;
+const struct brw_tracked_state brw_pipelined_state_pointers;
+const struct brw_tracked_state brw_binding_table_pointers;
+const struct brw_tracked_state brw_depthbuffer;
+const struct brw_tracked_state brw_polygon_stipple_offset;
+const struct brw_tracked_state brw_polygon_stipple;
+const struct brw_tracked_state brw_program_parameters;
+const struct brw_tracked_state brw_recalculate_urb_fence;
+const struct brw_tracked_state brw_sf_prog;
+const struct brw_tracked_state brw_sf_unit;
+const struct brw_tracked_state brw_sf_vp;
+const struct brw_tracked_state brw_state_base_address;
+const struct brw_tracked_state brw_urb_fence;
+const struct brw_tracked_state brw_vertex_state;
+const struct brw_tracked_state brw_vs_prog;
+const struct brw_tracked_state brw_vs_unit;
+const struct brw_tracked_state brw_wm_prog;
+const struct brw_tracked_state brw_wm_samplers;
+const struct brw_tracked_state brw_wm_surfaces;
+const struct brw_tracked_state brw_wm_unit;
+
+const struct brw_tracked_state brw_psp_urb_cbs;
+
+const struct brw_tracked_state brw_active_vertprog;
+const struct brw_tracked_state brw_tnl_vertprog;
+const struct brw_tracked_state brw_pipe_control;
+
+const struct brw_tracked_state brw_clear_surface_cache;
+const struct brw_tracked_state brw_clear_batch_cache;
+
+/***********************************************************************
+ * brw_state_cache.c
+ */
+unsigned brw_cache_data(struct brw_cache *cache,
+		      const void *data );
+
+unsigned brw_cache_data_sz(struct brw_cache *cache,
+			 const void *data,
+			 unsigned data_sz);
+
+unsigned brw_upload_cache( struct brw_cache *cache,
+			 const void *key,
+			 unsigned key_sz,
+			 const void *data,
+			 unsigned data_sz,
+			 const void *aux,
+			 void *aux_return );
+
+boolean brw_search_cache( struct brw_cache *cache,
+			    const void *key,
+			    unsigned key_size,
+			    void *aux_return,
+			    unsigned *offset_return);
+
+void brw_init_caches( struct brw_context *brw );
+void brw_destroy_caches( struct brw_context *brw );
+
+static inline struct pipe_buffer *brw_cache_buffer(struct brw_context *brw,
+                                                          enum brw_cache_id id)
+{
+   return brw->cache[id].pool->buffer;
+}
+
+/***********************************************************************
+ * brw_state_batch.c
+ */
+#define BRW_CACHED_BATCH_STRUCT(brw, s) brw_cached_batch_struct( brw, (s), sizeof(*(s)) )
+
+boolean brw_cached_batch_struct( struct brw_context *brw,
+				   const void *data,
+				   unsigned sz );
+
+void brw_destroy_batch_cache( struct brw_context *brw );
+
+
+/***********************************************************************
+ * brw_state_pool.c
+ */
+void brw_init_pools( struct brw_context *brw );
+void brw_destroy_pools( struct brw_context *brw );
+
+boolean brw_pool_alloc( struct brw_mem_pool *pool,
+			  unsigned size,
+			  unsigned alignment,
+			  unsigned *offset_return);
+
+void brw_pool_fence( struct brw_context *brw,
+		     struct brw_mem_pool *pool,
+		     unsigned fence );
+
+
+void brw_pool_check_wrap( struct brw_context *brw,
+			  struct brw_mem_pool *pool );
+
+void brw_clear_all_caches( struct brw_context *brw );
+void brw_invalidate_pools( struct brw_context *brw );
+void brw_clear_batch_cache_flush( struct brw_context *brw );
+
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_state_batch.c b/src/gallium/drivers/i965simple/brw_state_batch.c
new file mode 100644
index 0000000000..43a1c89fc4
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_state_batch.c
@@ -0,0 +1,113 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_state.h"
+#include "brw_winsys.h"
+
+#include "util/u_memory.h"
+
+/* A facility similar to the data caching code above, which aims to
+ * prevent identical commands being issued repeatedly.
+ */
+boolean brw_cached_batch_struct( struct brw_context *brw,
+                                 const void *data,
+                                 unsigned sz )
+{
+   struct brw_cached_batch_item *item = brw->cached_batch_items;
+   struct header *newheader = (struct header *)data;
+
+   if (brw->emit_state_always) {
+      brw_batchbuffer_data(brw->winsys, data, sz);
+      return TRUE;
+   }
+
+   while (item) {
+      if (item->header->opcode == newheader->opcode) {
+	 if (item->sz == sz && memcmp(item->header, newheader, sz) == 0)
+	    return FALSE;
+	 if (item->sz != sz) {
+	    FREE(item->header);
+	    item->header = MALLOC(sz);
+	    item->sz = sz;
+	 }
+	 goto emit;
+      }
+      item = item->next;
+   }
+
+   assert(!item);
+   item = CALLOC_STRUCT(brw_cached_batch_item);
+   item->header = MALLOC(sz);
+   item->sz = sz;
+   item->next = brw->cached_batch_items;
+   brw->cached_batch_items = item;
+
+emit:
+   memcpy(item->header, newheader, sz);
+   brw_batchbuffer_data(brw->winsys, data, sz);
+   return TRUE;
+}
+
+static void clear_batch_cache( struct brw_context *brw )
+{
+   struct brw_cached_batch_item *item = brw->cached_batch_items;
+
+   while (item) {
+      struct brw_cached_batch_item *next = item->next;
+      free((void *)item->header);
+      free(item);
+      item = next;
+   }
+
+   brw->cached_batch_items = NULL;
+
+
+   brw_clear_all_caches(brw);
+
+   brw_invalidate_pools(brw);
+}
+
+void brw_clear_batch_cache_flush( struct brw_context *brw )
+{
+   clear_batch_cache(brw);
+
+/*    brw_do_flush(brw, BRW_FLUSH_STATE_CACHE|BRW_FLUSH_READ_CACHE); */
+
+   brw->state.dirty.brw |= ~0;
+   brw->state.dirty.cache |= ~0;
+}
+
+
+
+void brw_destroy_batch_cache( struct brw_context *brw )
+{
+   clear_batch_cache(brw);
+}
diff --git a/src/gallium/drivers/i965simple/brw_state_cache.c b/src/gallium/drivers/i965simple/brw_state_cache.c
new file mode 100644
index 0000000000..094248fa69
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_state_cache.c
@@ -0,0 +1,443 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_state.h"
+
+#include "brw_wm.h"
+#include "brw_vs.h"
+#include "brw_clip.h"
+#include "brw_sf.h"
+#include "brw_gs.h"
+
+#include "util/u_memory.h"
+
+
+
+/***********************************************************************
+ * Check cache for uploaded version of struct, else upload new one.
+ * Fail when memory is exhausted.
+ *
+ * XXX: FIXME: Currently search is so slow it would be quicker to
+ * regenerate the data every time...
+ */
+
+static unsigned hash_key( const void *key, unsigned key_size )
+{
+   unsigned *ikey = (unsigned *)key;
+   unsigned hash = 0, i;
+
+   assert(key_size % 4 == 0);
+
+   /* I'm sure this can be improved on:
+    */
+   for (i = 0; i < key_size/4; i++)
+      hash ^= ikey[i];
+
+   return hash;
+}
+
+static struct brw_cache_item *search_cache( struct brw_cache *cache,
+					     unsigned hash,
+					     const void *key,
+					     unsigned key_size)
+{
+   struct brw_cache_item *c;
+
+   for (c = cache->items[hash % cache->size]; c; c = c->next) {
+      if (c->hash == hash &&
+	  c->key_size == key_size &&
+	  memcmp(c->key, key, key_size) == 0)
+	 return c;
+   }
+
+   return NULL;
+}
+
+
+static void rehash( struct brw_cache *cache )
+{
+   struct brw_cache_item **items;
+   struct brw_cache_item *c, *next;
+   unsigned size, i;
+
+   size = cache->size * 3;
+   items = (struct brw_cache_item**) MALLOC(size * sizeof(*items));
+   memset(items, 0, size * sizeof(*items));
+
+   for (i = 0; i < cache->size; i++)
+      for (c = cache->items[i]; c; c = next) {
+	 next = c->next;
+	 c->next = items[c->hash % size];
+	 items[c->hash % size] = c;
+      }
+
+   FREE(cache->items);
+   cache->items = items;
+   cache->size = size;
+}
+
+
+boolean brw_search_cache( struct brw_cache *cache,
+			    const void *key,
+			    unsigned key_size,
+			    void *aux_return,
+			    unsigned *offset_return)
+{
+   struct brw_cache_item *item;
+   unsigned addr = 0;
+   unsigned hash = hash_key(key, key_size);
+
+   item = search_cache(cache, hash, key, key_size);
+
+   if (item) {
+      if (aux_return)
+	 *(void **)aux_return = (void *)((char *)item->key + item->key_size);
+
+      *offset_return = addr = item->offset;
+   }
+
+   if (item == NULL || addr != cache->last_addr) {
+      cache->brw->state.dirty.cache |= 1<<cache->id;
+      cache->last_addr = addr;
+   }
+
+   return item != NULL;
+}
+
+unsigned brw_upload_cache( struct brw_cache *cache,
+			 const void *key,
+			 unsigned key_size,
+			 const void *data,
+			 unsigned data_size,
+			 const void *aux,
+			 void *aux_return )
+{
+   unsigned offset;
+   struct brw_cache_item *item = CALLOC_STRUCT(brw_cache_item);
+   unsigned hash = hash_key(key, key_size);
+   void *tmp = MALLOC(key_size + cache->aux_size);
+
+   if (!brw_pool_alloc(cache->pool, data_size, 1 << 6, &offset)) {
+      /* Should not be possible:
+       */
+      debug_printf("brw_pool_alloc failed\n");
+      exit(1);
+   }
+
+   memcpy(tmp, key, key_size);
+
+   if (cache->aux_size)
+      memcpy(tmp+key_size, aux, cache->aux_size);
+
+   item->key = tmp;
+   item->hash = hash;
+   item->key_size = key_size;
+   item->offset = offset;
+   item->data_size = data_size;
+
+   if (++cache->n_items > cache->size * 1.5)
+      rehash(cache);
+
+   hash %= cache->size;
+   item->next = cache->items[hash];
+   cache->items[hash] = item;
+
+   if (aux_return) {
+      assert(cache->aux_size);
+      *(void **)aux_return = (void *)((char *)item->key + item->key_size);
+   }
+
+   if (BRW_DEBUG & DEBUG_STATE)
+      debug_printf("upload %s: %d bytes to pool buffer %p offset %x\n",
+             cache->name, 
+	     data_size,
+             (void*)cache->pool->buffer,
+             offset);
+
+   /* Copy data to the buffer:
+    */
+   cache->brw->winsys->buffer_subdata_typed(cache->brw->winsys,
+					    cache->pool->buffer, 
+					    offset, 
+					    data_size, 
+					    data,
+					    cache->id);
+
+   cache->brw->state.dirty.cache |= 1<<cache->id;
+   cache->last_addr = offset;
+
+   return offset;
+}
+
+/* This doesn't really work with aux data.  Use search/upload instead
+ */
+unsigned brw_cache_data_sz(struct brw_cache *cache,
+			 const void *data,
+			 unsigned data_size)
+{
+   unsigned addr;
+
+   if (!brw_search_cache(cache, data, data_size, NULL, &addr)) {
+      addr = brw_upload_cache(cache,
+			      data, data_size,
+			      data, data_size,
+			      NULL, NULL);
+   }
+
+   return addr;
+}
+
+unsigned brw_cache_data(struct brw_cache *cache,
+		      const void *data)
+{
+   return brw_cache_data_sz(cache, data, cache->key_size);
+}
+
+enum pool_type {
+   DW_SURFACE_STATE,
+   DW_GENERAL_STATE
+};
+
+static void brw_init_cache( struct brw_context *brw,
+			    const char *name,
+			    unsigned id,
+			    unsigned key_size,
+			    unsigned aux_size,
+			    enum pool_type pool_type)
+{
+   struct brw_cache *cache = &brw->cache[id];
+   cache->brw = brw;
+   cache->id = id;
+   cache->name = name;
+   cache->items = NULL;
+
+   cache->size = 7;
+   cache->n_items = 0;
+   cache->items = (struct brw_cache_item **)
+      CALLOC(cache->size, sizeof(struct brw_cache_item));
+
+
+   cache->key_size = key_size;
+   cache->aux_size = aux_size;
+   switch (pool_type) {
+   case DW_GENERAL_STATE: cache->pool = &brw->pool[BRW_GS_POOL]; break;
+   case DW_SURFACE_STATE: cache->pool = &brw->pool[BRW_SS_POOL]; break;
+   default: assert(0); break;
+   }
+}
+
+void brw_init_caches( struct brw_context *brw )
+{
+
+   brw_init_cache(brw,
+		  "CC_VP",
+		  BRW_CC_VP,
+		  sizeof(struct brw_cc_viewport),
+		  0,
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "CC_UNIT",
+		  BRW_CC_UNIT,
+		  sizeof(struct brw_cc_unit_state),
+		  0,
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "WM_PROG",
+		  BRW_WM_PROG,
+		  sizeof(struct brw_wm_prog_key),
+		  sizeof(struct brw_wm_prog_data),
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "SAMPLER_DEFAULT_COLOR",
+		  BRW_SAMPLER_DEFAULT_COLOR,
+		  sizeof(struct brw_sampler_default_color),
+		  0,
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "SAMPLER",
+		  BRW_SAMPLER,
+		  0,		/* variable key/data size */
+		  0,
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "WM_UNIT",
+		  BRW_WM_UNIT,
+		  sizeof(struct brw_wm_unit_state),
+		  0,
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "SF_PROG",
+		  BRW_SF_PROG,
+		  sizeof(struct brw_sf_prog_key),
+		  sizeof(struct brw_sf_prog_data),
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "SF_VP",
+		  BRW_SF_VP,
+		  sizeof(struct brw_sf_viewport),
+		  0,
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "SF_UNIT",
+		  BRW_SF_UNIT,
+		  sizeof(struct brw_sf_unit_state),
+		  0,
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "VS_UNIT",
+		  BRW_VS_UNIT,
+		  sizeof(struct brw_vs_unit_state),
+		  0,
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "VS_PROG",
+		  BRW_VS_PROG,
+		  sizeof(struct brw_vs_prog_key),
+		  sizeof(struct brw_vs_prog_data),
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "CLIP_UNIT",
+		  BRW_CLIP_UNIT,
+		  sizeof(struct brw_clip_unit_state),
+		  0,
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "CLIP_PROG",
+		  BRW_CLIP_PROG,
+		  sizeof(struct brw_clip_prog_key),
+		  sizeof(struct brw_clip_prog_data),
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "GS_UNIT",
+		  BRW_GS_UNIT,
+		  sizeof(struct brw_gs_unit_state),
+		  0,
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "GS_PROG",
+		  BRW_GS_PROG,
+		  sizeof(struct brw_gs_prog_key),
+		  sizeof(struct brw_gs_prog_data),
+		  DW_GENERAL_STATE);
+
+   brw_init_cache(brw,
+		  "SS_SURFACE",
+		  BRW_SS_SURFACE,
+		  sizeof(struct brw_surface_state),
+		  0,
+		  DW_SURFACE_STATE);
+
+   brw_init_cache(brw,
+		  "SS_SURF_BIND",
+		  BRW_SS_SURF_BIND,
+		  sizeof(struct brw_surface_binding_table),
+		  0,
+		  DW_SURFACE_STATE);
+}
+
+
+/* When we lose hardware context, need to invalidate the surface cache
+ * as these structs must be explicitly re-uploaded.  They are subject
+ * to fixup by the memory manager as they contain absolute agp
+ * offsets, so we need to ensure there is a fresh version of the
+ * struct available to receive the fixup.
+ *
+ * XXX: Need to ensure that there aren't two versions of a surface or
+ * bufferobj with different backing data active in the same buffer at
+ * once?  Otherwise the cache could confuse them.  Maybe better not to
+ * cache at all?
+ *
+ * --> Isn't this the same as saying need to ensure batch is flushed
+ *         before new data is uploaded to an existing buffer?  We
+ *         already try to make sure of that.
+ */
+static void clear_cache( struct brw_cache *cache )
+{
+   struct brw_cache_item *c, *next;
+   unsigned i;
+
+   for (i = 0; i < cache->size; i++) {
+      for (c = cache->items[i]; c; c = next) {
+	 next = c->next;
+	 free((void *)c->key);
+	 free(c);
+      }
+      cache->items[i] = NULL;
+   }
+
+   cache->n_items = 0;
+}
+
+void brw_clear_all_caches( struct brw_context *brw )
+{
+   int i;
+
+   if (BRW_DEBUG & DEBUG_STATE)
+      debug_printf("%s\n", __FUNCTION__);
+
+   for (i = 0; i < BRW_MAX_CACHE; i++)
+      clear_cache(&brw->cache[i]);
+
+   if (brw->curbe.last_buf) {
+      FREE(brw->curbe.last_buf);
+      brw->curbe.last_buf = NULL;
+   }
+
+   brw->state.dirty.brw |= ~0;
+   brw->state.dirty.cache |= ~0;
+}
+
+
+
+
+
+void brw_destroy_caches( struct brw_context *brw )
+{
+   unsigned i;
+
+   for (i = 0; i < BRW_MAX_CACHE; i++)
+      clear_cache(&brw->cache[i]);
+}
diff --git a/src/gallium/drivers/i965simple/brw_state_pool.c b/src/gallium/drivers/i965simple/brw_state_pool.c
new file mode 100644
index 0000000000..e91263cb1f
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_state_pool.c
@@ -0,0 +1,138 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+/** @file brw_state_pool.c
+ * Implements the state pool allocator.
+ *
+ * For the 965, we create two state pools for state cache entries.  Objects
+ * will be allocated into the pools depending on which state base address
+ * their pointer is relative to in other 965 state.
+ *
+ * The state pools are relatively simple: As objects are allocated, increment
+ * the offset to allocate space.  When the pool is "full" (rather, close to
+ * full), we reset the pool and reset the state cache entries that point into
+ * the pool.
+ */
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "brw_context.h"
+#include "brw_state.h"
+
+boolean brw_pool_alloc( struct brw_mem_pool *pool,
+			  unsigned size,
+			  unsigned alignment,
+			  unsigned *offset_return)
+{
+   unsigned fixup = align(pool->offset, alignment) - pool->offset;
+
+   size = align(size, 4);
+
+   if (pool->offset + fixup + size >= pool->size) {
+      debug_printf("%s failed\n", __FUNCTION__);
+      assert(0);
+      exit(0);
+   }
+
+   pool->offset += fixup;
+   *offset_return = pool->offset;
+   pool->offset += size;
+
+   return TRUE;
+}
+
+static
+void brw_invalidate_pool( struct brw_mem_pool *pool )
+{
+   if (BRW_DEBUG & DEBUG_STATE)
+      debug_printf("\n\n\n %s \n\n\n", __FUNCTION__);
+
+   pool->offset = 0;
+
+   brw_clear_all_caches(pool->brw);
+}
+
+
+static void brw_init_pool( struct brw_context *brw,
+			   unsigned pool_id,
+			   unsigned size )
+{
+   struct brw_mem_pool *pool = &brw->pool[pool_id];
+
+   pool->size = size;
+   pool->brw = brw;
+
+   pool->buffer = pipe_buffer_create(brw->pipe.screen,
+                                     4096,
+                                     0 /*  DRM_BO_FLAG_MEM_TT */,
+                                     size);
+}
+
+static void brw_destroy_pool( struct brw_context *brw,
+			      unsigned pool_id )
+{
+   struct brw_mem_pool *pool = &brw->pool[pool_id];
+
+   pipe_buffer_reference( pool->brw->pipe.screen,
+			  &pool->buffer,
+			  NULL );
+}
+
+
+void brw_pool_check_wrap( struct brw_context *brw,
+			  struct brw_mem_pool *pool )
+{
+   if (pool->offset > (pool->size * 3) / 4) {
+      brw->state.dirty.brw |= BRW_NEW_SCENE;
+   }
+
+}
+
+void brw_init_pools( struct brw_context *brw )
+{
+   brw_init_pool(brw, BRW_GS_POOL, 0x80000);
+   brw_init_pool(brw, BRW_SS_POOL, 0x80000);
+}
+
+void brw_destroy_pools( struct brw_context *brw )
+{
+   brw_destroy_pool(brw, BRW_GS_POOL);
+   brw_destroy_pool(brw, BRW_SS_POOL);
+}
+
+
+void brw_invalidate_pools( struct brw_context *brw )
+{
+   brw_invalidate_pool(&brw->pool[BRW_GS_POOL]);
+   brw_invalidate_pool(&brw->pool[BRW_SS_POOL]);
+}
diff --git a/src/gallium/drivers/i965simple/brw_state_upload.c b/src/gallium/drivers/i965simple/brw_state_upload.c
new file mode 100644
index 0000000000..bac9161b5f
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_state_upload.c
@@ -0,0 +1,202 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_context.h"
+#include "brw_state.h"
+
+#include "util/u_memory.h"
+
+/* This is used to initialize brw->state.atoms[].  We could use this
+ * list directly except for a single atom, brw_constant_buffer, which
+ * has a .dirty value which changes according to the parameters of the
+ * current fragment and vertex programs, and so cannot be a static
+ * value.
+ */
+const struct brw_tracked_state *atoms[] =
+{
+   &brw_vs_prog,
+   &brw_gs_prog,
+   &brw_clip_prog,
+   &brw_sf_prog,
+   &brw_wm_prog,
+
+   /* Once all the programs are done, we know how large urb entry
+    * sizes need to be and can decide if we need to change the urb
+    * layout.
+    */
+   &brw_curbe_offsets,
+   &brw_recalculate_urb_fence,
+
+
+   &brw_cc_vp,
+   &brw_cc_unit,
+
+   &brw_wm_surfaces,		/* must do before samplers */
+   &brw_wm_samplers,
+
+   &brw_wm_unit,
+   &brw_sf_vp,
+   &brw_sf_unit,
+   &brw_vs_unit,		/* always required, enabled or not */
+   &brw_clip_unit,
+   &brw_gs_unit,
+
+   /* Command packets:
+    */
+   &brw_invarient_state,
+   &brw_state_base_address,
+   &brw_pipe_control,
+
+   &brw_binding_table_pointers,
+   &brw_blend_constant_color,
+
+   &brw_drawing_rect,
+   &brw_depthbuffer,
+
+   &brw_polygon_stipple,
+   &brw_line_stipple,
+
+   &brw_psp_urb_cbs,
+
+   &brw_constant_buffer
+};
+
+
+void brw_init_state( struct brw_context *brw )
+{
+   brw_init_pools(brw);
+   brw_init_caches(brw);
+
+   brw->state.dirty.brw = ~0;
+   brw->emit_state_always = 0;
+}
+
+
+void brw_destroy_state( struct brw_context *brw )
+{
+   brw_destroy_caches(brw);
+   brw_destroy_batch_cache(brw);
+   brw_destroy_pools(brw);
+}
+
+/***********************************************************************
+ */
+
+static boolean check_state( const struct brw_state_flags *a,
+			      const struct brw_state_flags *b )
+{
+   return ((a->brw & b->brw) ||
+	   (a->cache & b->cache));
+}
+
+static void accumulate_state( struct brw_state_flags *a,
+			      const struct brw_state_flags *b )
+{
+   a->brw |= b->brw;
+   a->cache |= b->cache;
+}
+
+
+static void xor_states( struct brw_state_flags *result,
+			     const struct brw_state_flags *a,
+			      const struct brw_state_flags *b )
+{
+   result->brw = a->brw ^ b->brw;
+   result->cache = a->cache ^ b->cache;
+}
+
+
+/***********************************************************************
+ * Emit all state:
+ */
+void brw_validate_state( struct brw_context *brw )
+{
+   struct brw_state_flags *state = &brw->state.dirty;
+   unsigned i;
+
+   if (brw->emit_state_always) 
+      state->brw |= ~0;
+
+   if (state->cache == 0 &&
+       state->brw == 0)
+      return;
+
+   if (brw->state.dirty.brw & BRW_NEW_SCENE)
+      brw_clear_batch_cache_flush(brw);
+
+   if (BRW_DEBUG) {
+      /* Debug version which enforces various sanity checks on the
+       * state flags which are generated and checked to help ensure
+       * state atoms are ordered correctly in the list.
+       */
+      struct brw_state_flags examined, prev;
+      memset(&examined, 0, sizeof(examined));
+      prev = *state;
+
+      for (i = 0; i < Elements(atoms); i++) {
+	 const struct brw_tracked_state *atom = atoms[i];
+	 struct brw_state_flags generated;
+
+	 assert(atom->dirty.brw ||
+		atom->dirty.cache);
+	 assert(atom->update);
+
+	 if (check_state(state, &atom->dirty)) {
+	    atom->update( brw );
+	 }
+
+	 accumulate_state(&examined, &atom->dirty);
+
+	 /* generated = (prev ^ state)
+	  * if (examined & generated)
+	  *     fail;
+	  */
+	 xor_states(&generated, &prev, state);
+	 assert(!check_state(&examined, &generated));
+	 prev = *state;
+      }
+   }
+   else {
+      for (i = 0; i < Elements(atoms); i++) {
+	 const struct brw_tracked_state *atom = atoms[i];
+
+	 assert(atom->dirty.brw ||
+		atom->dirty.cache);
+	 assert(atom->update);
+
+	 if (check_state(state, &atom->dirty))
+	    atom->update( brw );
+      }
+   }
+
+   memset(state, 0, sizeof(*state));
+}
diff --git a/src/gallium/drivers/i965simple/brw_structs.h b/src/gallium/drivers/i965simple/brw_structs.h
new file mode 100644
index 0000000000..bbb087e95d
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_structs.h
@@ -0,0 +1,1348 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#ifndef BRW_STRUCTS_H
+#define BRW_STRUCTS_H
+
+#include "pipe/p_compiler.h"
+
+/* Command packets:
+ */
+struct header
+{
+   unsigned length:16;
+   unsigned opcode:16;
+};
+
+
+union header_union
+{
+   struct header bits;
+   unsigned dword;
+};
+
+struct brw_3d_control
+{
+   struct
+   {
+      unsigned length:8;
+      unsigned notify_enable:1;
+      unsigned pad:3;
+      unsigned wc_flush_enable:1;
+      unsigned depth_stall_enable:1;
+      unsigned operation:2;
+      unsigned opcode:16;
+   } header;
+
+   struct
+   {
+      unsigned pad:2;
+      unsigned dest_addr_type:1;
+      unsigned dest_addr:29;
+   } dest;
+
+   unsigned dword2;
+   unsigned dword3;
+};
+
+
+struct brw_3d_primitive
+{
+   struct
+   {
+      unsigned length:8;
+      unsigned pad:2;
+      unsigned topology:5;
+      unsigned indexed:1;
+      unsigned opcode:16;
+   } header;
+
+   unsigned verts_per_instance;
+   unsigned start_vert_location;
+   unsigned instance_count;
+   unsigned start_instance_location;
+   unsigned base_vert_location;
+};
+
+/* These seem to be passed around as function args, so it works out
+ * better to keep them as #defines:
+ */
+#define BRW_FLUSH_READ_CACHE           0x1
+#define BRW_FLUSH_STATE_CACHE          0x2
+#define BRW_INHIBIT_FLUSH_RENDER_CACHE 0x4
+#define BRW_FLUSH_SNAPSHOT_COUNTERS    0x8
+
+struct brw_mi_flush
+{
+   unsigned flags:4;
+   unsigned pad:12;
+   unsigned opcode:16;
+};
+
+struct brw_vf_statistics
+{
+   unsigned statistics_enable:1;
+   unsigned pad:15;
+   unsigned opcode:16;
+};
+
+
+
+struct brw_binding_table_pointers
+{
+   struct header header;
+   unsigned vs;
+   unsigned gs;
+   unsigned clp;
+   unsigned sf;
+   unsigned wm;
+};
+
+
+struct brw_blend_constant_color
+{
+   struct header header;
+   float blend_constant_color[4];
+};
+
+
+struct brw_depthbuffer
+{
+   union header_union header;
+
+   union {
+      struct {
+	 unsigned pitch:18;
+	 unsigned format:3;
+	 unsigned pad:4;
+	 unsigned depth_offset_disable:1;
+	 unsigned tile_walk:1;
+	 unsigned tiled_surface:1;
+	 unsigned pad2:1;
+	 unsigned surface_type:3;
+      } bits;
+      unsigned dword;
+   } dword1;
+
+   unsigned dword2_base_addr;
+
+   union {
+      struct {
+	 unsigned pad:1;
+	 unsigned mipmap_layout:1;
+	 unsigned lod:4;
+	 unsigned width:13;
+	 unsigned height:13;
+      } bits;
+      unsigned dword;
+   } dword3;
+
+   union {
+      struct {
+	 unsigned pad:12;
+	 unsigned min_array_element:9;
+	 unsigned depth:11;
+      } bits;
+      unsigned dword;
+   } dword4;
+};
+
+struct brw_drawrect
+{
+   struct header header;
+   unsigned xmin:16;
+   unsigned ymin:16;
+   unsigned xmax:16;
+   unsigned ymax:16;
+   unsigned xorg:16;
+   unsigned yorg:16;
+};
+
+
+
+
+struct brw_global_depth_offset_clamp
+{
+   struct header header;
+   float depth_offset_clamp;
+};
+
+struct brw_indexbuffer
+{
+   union {
+      struct
+      {
+	 unsigned length:8;
+	 unsigned index_format:2;
+	 unsigned cut_index_enable:1;
+	 unsigned pad:5;
+	 unsigned opcode:16;
+      } bits;
+      unsigned dword;
+
+   } header;
+
+   unsigned buffer_start;
+   unsigned buffer_end;
+};
+
+
+struct brw_line_stipple
+{
+   struct header header;
+
+   struct
+   {
+      unsigned pattern:16;
+      unsigned pad:16;
+   } bits0;
+
+   struct
+   {
+      unsigned repeat_count:9;
+      unsigned pad:7;
+      unsigned inverse_repeat_count:16;
+   } bits1;
+};
+
+
+struct brw_pipelined_state_pointers
+{
+   struct header header;
+
+   struct {
+      unsigned pad:5;
+      unsigned offset:27;
+   } vs;
+
+   struct
+   {
+      unsigned enable:1;
+      unsigned pad:4;
+      unsigned offset:27;
+   } gs;
+
+   struct
+   {
+      unsigned enable:1;
+      unsigned pad:4;
+      unsigned offset:27;
+   } clp;
+
+   struct
+   {
+      unsigned pad:5;
+      unsigned offset:27;
+   } sf;
+
+   struct
+   {
+      unsigned pad:5;
+      unsigned offset:27;
+   } wm;
+
+   struct
+   {
+      unsigned pad:5;
+      unsigned offset:27; /* KW: check me! */
+   } cc;
+};
+
+
+struct brw_polygon_stipple_offset
+{
+   struct header header;
+
+   struct {
+      unsigned y_offset:5;
+      unsigned pad:3;
+      unsigned x_offset:5;
+      unsigned pad0:19;
+   } bits0;
+};
+
+
+
+struct brw_polygon_stipple
+{
+   struct header header;
+   unsigned stipple[32];
+};
+
+
+
+struct brw_pipeline_select
+{
+   struct
+   {
+      unsigned pipeline_select:1;
+      unsigned pad:15;
+      unsigned opcode:16;
+   } header;
+};
+
+
+struct brw_pipe_control
+{
+   struct
+   {
+      unsigned length:8;
+      unsigned notify_enable:1;
+      unsigned pad:2;
+      unsigned instruction_state_cache_flush_enable:1;
+      unsigned write_cache_flush_enable:1;
+      unsigned depth_stall_enable:1;
+      unsigned post_sync_operation:2;
+
+      unsigned opcode:16;
+   } header;
+
+   struct
+   {
+      unsigned pad:2;
+      unsigned dest_addr_type:1;
+      unsigned dest_addr:29;
+   } bits1;
+
+   unsigned data0;
+   unsigned data1;
+};
+
+
+struct brw_urb_fence
+{
+   struct
+   {
+      unsigned length:8;
+      unsigned vs_realloc:1;
+      unsigned gs_realloc:1;
+      unsigned clp_realloc:1;
+      unsigned sf_realloc:1;
+      unsigned vfe_realloc:1;
+      unsigned cs_realloc:1;
+      unsigned pad:2;
+      unsigned opcode:16;
+   } header;
+
+   struct
+   {
+      unsigned vs_fence:10;
+      unsigned gs_fence:10;
+      unsigned clp_fence:10;
+      unsigned pad:2;
+   } bits0;
+
+   struct
+   {
+      unsigned sf_fence:10;
+      unsigned vf_fence:10;
+      unsigned cs_fence:10;
+      unsigned pad:2;
+   } bits1;
+};
+
+struct brw_constant_buffer_state /* previously brw_command_streamer */
+{
+   struct header header;
+
+   struct
+   {
+      unsigned nr_urb_entries:3;
+      unsigned pad:1;
+      unsigned urb_entry_size:5;
+      unsigned pad0:23;
+   } bits0;
+};
+
+struct brw_constant_buffer
+{
+   struct
+   {
+      unsigned length:8;
+      unsigned valid:1;
+      unsigned pad:7;
+      unsigned opcode:16;
+   } header;
+
+   struct
+   {
+      unsigned buffer_length:6;
+      unsigned buffer_address:26;
+   } bits0;
+};
+
+struct brw_state_base_address
+{
+   struct header header;
+
+   struct
+   {
+      unsigned modify_enable:1;
+      unsigned pad:4;
+      unsigned general_state_address:27;
+   } bits0;
+
+   struct
+   {
+      unsigned modify_enable:1;
+      unsigned pad:4;
+      unsigned surface_state_address:27;
+   } bits1;
+
+   struct
+   {
+      unsigned modify_enable:1;
+      unsigned pad:4;
+      unsigned indirect_object_state_address:27;
+   } bits2;
+
+   struct
+   {
+      unsigned modify_enable:1;
+      unsigned pad:11;
+      unsigned general_state_upper_bound:20;
+   } bits3;
+
+   struct
+   {
+      unsigned modify_enable:1;
+      unsigned pad:11;
+      unsigned indirect_object_state_upper_bound:20;
+   } bits4;
+};
+
+struct brw_state_prefetch
+{
+   struct header header;
+
+   struct
+   {
+      unsigned prefetch_count:3;
+      unsigned pad:3;
+      unsigned prefetch_pointer:26;
+   } bits0;
+};
+
+struct brw_system_instruction_pointer
+{
+   struct header header;
+
+   struct
+   {
+      unsigned pad:4;
+      unsigned system_instruction_pointer:28;
+   } bits0;
+};
+
+
+
+
+/* State structs for the various fixed function units:
+ */
+
+
+struct thread0
+{
+   unsigned pad0:1;
+   unsigned grf_reg_count:3;
+   unsigned pad1:2;
+   unsigned kernel_start_pointer:26;
+};
+
+struct thread1
+{
+   unsigned ext_halt_exception_enable:1;
+   unsigned sw_exception_enable:1;
+   unsigned mask_stack_exception_enable:1;
+   unsigned timeout_exception_enable:1;
+   unsigned illegal_op_exception_enable:1;
+   unsigned pad0:3;
+   unsigned depth_coef_urb_read_offset:6;	/* WM only */
+   unsigned pad1:2;
+   unsigned floating_point_mode:1;
+   unsigned thread_priority:1;
+   unsigned binding_table_entry_count:8;
+   unsigned pad3:5;
+   unsigned single_program_flow:1;
+};
+
+struct thread2
+{
+   unsigned per_thread_scratch_space:4;
+   unsigned pad0:6;
+   unsigned scratch_space_base_pointer:22;
+};
+
+
+struct thread3
+{
+   unsigned dispatch_grf_start_reg:4;
+   unsigned urb_entry_read_offset:6;
+   unsigned pad0:1;
+   unsigned urb_entry_read_length:6;
+   unsigned pad1:1;
+   unsigned const_urb_entry_read_offset:6;
+   unsigned pad2:1;
+   unsigned const_urb_entry_read_length:6;
+   unsigned pad3:1;
+};
+
+
+
+struct brw_clip_unit_state
+{
+   struct thread0 thread0;
+   struct
+   {
+      unsigned pad0:7;
+      unsigned sw_exception_enable:1;
+      unsigned pad1:3;
+      unsigned mask_stack_exception_enable:1;
+      unsigned pad2:1;
+      unsigned illegal_op_exception_enable:1;
+      unsigned pad3:2;
+      unsigned floating_point_mode:1;
+      unsigned thread_priority:1;
+      unsigned binding_table_entry_count:8;
+      unsigned pad4:5;
+      unsigned single_program_flow:1;
+   } thread1;
+
+   struct thread2 thread2;
+   struct thread3 thread3;
+
+   struct
+   {
+      unsigned pad0:9;
+      unsigned gs_output_stats:1; /* not always */
+      unsigned stats_enable:1;
+      unsigned nr_urb_entries:7;
+      unsigned pad1:1;
+      unsigned urb_entry_allocation_size:5;
+      unsigned pad2:1;
+      unsigned max_threads:1; 	/* may be less */
+      unsigned pad3:6;
+   } thread4;
+
+   struct
+   {
+      unsigned pad0:13;
+      unsigned clip_mode:3;
+      unsigned userclip_enable_flags:8;
+      unsigned userclip_must_clip:1;
+      unsigned pad1:1;
+      unsigned guard_band_enable:1;
+      unsigned viewport_z_clip_enable:1;
+      unsigned viewport_xy_clip_enable:1;
+      unsigned vertex_position_space:1;
+      unsigned api_mode:1;
+      unsigned pad2:1;
+   } clip5;
+
+   struct
+   {
+      unsigned pad0:5;
+      unsigned clipper_viewport_state_ptr:27;
+   } clip6;
+
+
+   float viewport_xmin;
+   float viewport_xmax;
+   float viewport_ymin;
+   float viewport_ymax;
+};
+
+
+
+struct brw_cc_unit_state
+{
+   struct
+   {
+      unsigned pad0:3;
+      unsigned bf_stencil_pass_depth_pass_op:3;
+      unsigned bf_stencil_pass_depth_fail_op:3;
+      unsigned bf_stencil_fail_op:3;
+      unsigned bf_stencil_func:3;
+      unsigned bf_stencil_enable:1;
+      unsigned pad1:2;
+      unsigned stencil_write_enable:1;
+      unsigned stencil_pass_depth_pass_op:3;
+      unsigned stencil_pass_depth_fail_op:3;
+      unsigned stencil_fail_op:3;
+      unsigned stencil_func:3;
+      unsigned stencil_enable:1;
+   } cc0;
+
+
+   struct
+   {
+      unsigned bf_stencil_ref:8;
+      unsigned stencil_write_mask:8;
+      unsigned stencil_test_mask:8;
+      unsigned stencil_ref:8;
+   } cc1;
+
+
+   struct
+   {
+      unsigned logicop_enable:1;
+      unsigned pad0:10;
+      unsigned depth_write_enable:1;
+      unsigned depth_test_function:3;
+      unsigned depth_test:1;
+      unsigned bf_stencil_write_mask:8;
+      unsigned bf_stencil_test_mask:8;
+   } cc2;
+
+
+   struct
+   {
+      unsigned pad0:8;
+      unsigned alpha_test_func:3;
+      unsigned alpha_test:1;
+      unsigned blend_enable:1;
+      unsigned ia_blend_enable:1;
+      unsigned pad1:1;
+      unsigned alpha_test_format:1;
+      unsigned pad2:16;
+   } cc3;
+
+   struct
+   {
+      unsigned pad0:5;
+      unsigned cc_viewport_state_offset:27;
+   } cc4;
+
+   struct
+   {
+      unsigned pad0:2;
+      unsigned ia_dest_blend_factor:5;
+      unsigned ia_src_blend_factor:5;
+      unsigned ia_blend_function:3;
+      unsigned statistics_enable:1;
+      unsigned logicop_func:4;
+      unsigned pad1:11;
+      unsigned dither_enable:1;
+   } cc5;
+
+   struct
+   {
+      unsigned clamp_post_alpha_blend:1;
+      unsigned clamp_pre_alpha_blend:1;
+      unsigned clamp_range:2;
+      unsigned pad0:11;
+      unsigned y_dither_offset:2;
+      unsigned x_dither_offset:2;
+      unsigned dest_blend_factor:5;
+      unsigned src_blend_factor:5;
+      unsigned blend_function:3;
+   } cc6;
+
+   struct {
+      union {
+	 float f;
+	 ubyte ub[4];
+      } alpha_ref;
+   } cc7;
+};
+
+
+
+struct brw_sf_unit_state
+{
+   struct thread0 thread0;
+   struct thread1 thread1;
+   struct thread2 thread2;
+   struct thread3 thread3;
+
+   struct
+   {
+      unsigned pad0:10;
+      unsigned stats_enable:1;
+      unsigned nr_urb_entries:7;
+      unsigned pad1:1;
+      unsigned urb_entry_allocation_size:5;
+      unsigned pad2:1;
+      unsigned max_threads:6;
+      unsigned pad3:1;
+   } thread4;
+
+   struct
+   {
+      unsigned front_winding:1;
+      unsigned viewport_transform:1;
+      unsigned pad0:3;
+      unsigned sf_viewport_state_offset:27;
+   } sf5;
+
+   struct
+   {
+      unsigned pad0:9;
+      unsigned dest_org_vbias:4;
+      unsigned dest_org_hbias:4;
+      unsigned scissor:1;
+      unsigned disable_2x2_trifilter:1;
+      unsigned disable_zero_pix_trifilter:1;
+      unsigned point_rast_rule:2;
+      unsigned line_endcap_aa_region_width:2;
+      unsigned line_width:4;
+      unsigned fast_scissor_disable:1;
+      unsigned cull_mode:2;
+      unsigned aa_enable:1;
+   } sf6;
+
+   struct
+   {
+      unsigned point_size:11;
+      unsigned use_point_size_state:1;
+      unsigned subpixel_precision:1;
+      unsigned sprite_point:1;
+      unsigned pad0:11;
+      unsigned trifan_pv:2;
+      unsigned linestrip_pv:2;
+      unsigned tristrip_pv:2;
+      unsigned line_last_pixel_enable:1;
+   } sf7;
+
+};
+
+
+struct brw_gs_unit_state
+{
+   struct thread0 thread0;
+   struct thread1 thread1;
+   struct thread2 thread2;
+   struct thread3 thread3;
+
+   struct
+   {
+      unsigned pad0:10;
+      unsigned stats_enable:1;
+      unsigned nr_urb_entries:7;
+      unsigned pad1:1;
+      unsigned urb_entry_allocation_size:5;
+      unsigned pad2:1;
+      unsigned max_threads:1;
+      unsigned pad3:6;
+   } thread4;
+
+   struct
+   {
+      unsigned sampler_count:3;
+      unsigned pad0:2;
+      unsigned sampler_state_pointer:27;
+   } gs5;
+
+
+   struct
+   {
+      unsigned max_vp_index:4;
+      unsigned pad0:26;
+      unsigned reorder_enable:1;
+      unsigned pad1:1;
+   } gs6;
+};
+
+
+struct brw_vs_unit_state
+{
+   struct thread0 thread0;
+   struct thread1 thread1;
+   struct thread2 thread2;
+   struct thread3 thread3;
+
+   struct
+   {
+      unsigned pad0:10;
+      unsigned stats_enable:1;
+      unsigned nr_urb_entries:7;
+      unsigned pad1:1;
+      unsigned urb_entry_allocation_size:5;
+      unsigned pad2:1;
+      unsigned max_threads:4;
+      unsigned pad3:3;
+   } thread4;
+
+   struct
+   {
+      unsigned sampler_count:3;
+      unsigned pad0:2;
+      unsigned sampler_state_pointer:27;
+   } vs5;
+
+   struct
+   {
+      unsigned vs_enable:1;
+      unsigned vert_cache_disable:1;
+      unsigned pad0:30;
+   } vs6;
+};
+
+
+struct brw_wm_unit_state
+{
+   struct thread0 thread0;
+   struct thread1 thread1;
+   struct thread2 thread2;
+   struct thread3 thread3;
+
+   struct {
+      unsigned stats_enable:1;
+      unsigned pad0:1;
+      unsigned sampler_count:3;
+      unsigned sampler_state_pointer:27;
+   } wm4;
+
+   struct
+   {
+      unsigned enable_8_pix:1;
+      unsigned enable_16_pix:1;
+      unsigned enable_32_pix:1;
+      unsigned pad0:7;
+      unsigned legacy_global_depth_bias:1;
+      unsigned line_stipple:1;
+      unsigned depth_offset:1;
+      unsigned polygon_stipple:1;
+      unsigned line_aa_region_width:2;
+      unsigned line_endcap_aa_region_width:2;
+      unsigned early_depth_test:1;
+      unsigned thread_dispatch_enable:1;
+      unsigned program_uses_depth:1;
+      unsigned program_computes_depth:1;
+      unsigned program_uses_killpixel:1;
+      unsigned legacy_line_rast: 1;
+      unsigned pad1:1;
+      unsigned max_threads:6;
+      unsigned pad2:1;
+   } wm5;
+
+   float global_depth_offset_constant;
+   float global_depth_offset_scale;
+};
+
+struct brw_sampler_default_color {
+   float color[4];
+};
+
+struct brw_sampler_state
+{
+
+   struct
+   {
+      unsigned shadow_function:3;
+      unsigned lod_bias:11;
+      unsigned min_filter:3;
+      unsigned mag_filter:3;
+      unsigned mip_filter:2;
+      unsigned base_level:5;
+      unsigned pad:1;
+      unsigned lod_preclamp:1;
+      unsigned default_color_mode:1;
+      unsigned pad0:1;
+      unsigned disable:1;
+   } ss0;
+
+   struct
+   {
+      unsigned r_wrap_mode:3;
+      unsigned t_wrap_mode:3;
+      unsigned s_wrap_mode:3;
+      unsigned pad:3;
+      unsigned max_lod:10;
+      unsigned min_lod:10;
+   } ss1;
+
+
+   struct
+   {
+      unsigned pad:5;
+      unsigned default_color_pointer:27;
+   } ss2;
+
+   struct
+   {
+      unsigned pad:19;
+      unsigned max_aniso:3;
+      unsigned chroma_key_mode:1;
+      unsigned chroma_key_index:2;
+      unsigned chroma_key_enable:1;
+      unsigned monochrome_filter_width:3;
+      unsigned monochrome_filter_height:3;
+   } ss3;
+};
+
+
+struct brw_clipper_viewport
+{
+   float xmin;
+   float xmax;
+   float ymin;
+   float ymax;
+};
+
+struct brw_cc_viewport
+{
+   float min_depth;
+   float max_depth;
+};
+
+struct brw_sf_viewport
+{
+   struct {
+      float m00;
+      float m11;
+      float m22;
+      float m30;
+      float m31;
+      float m32;
+   } viewport;
+
+   struct {
+      short xmin;
+      short ymin;
+      short xmax;
+      short ymax;
+   } scissor;
+};
+
+/* Documented in the subsystem/shared-functions/sampler chapter...
+ */
+struct brw_surface_state
+{
+   struct {
+      unsigned cube_pos_z:1;
+      unsigned cube_neg_z:1;
+      unsigned cube_pos_y:1;
+      unsigned cube_neg_y:1;
+      unsigned cube_pos_x:1;
+      unsigned cube_neg_x:1;
+      unsigned pad:4;
+      unsigned mipmap_layout_mode:1;
+      unsigned vert_line_stride_ofs:1;
+      unsigned vert_line_stride:1;
+      unsigned color_blend:1;
+      unsigned writedisable_blue:1;
+      unsigned writedisable_green:1;
+      unsigned writedisable_red:1;
+      unsigned writedisable_alpha:1;
+      unsigned surface_format:9;
+      unsigned data_return_format:1;
+      unsigned pad0:1;
+      unsigned surface_type:3;
+   } ss0;
+
+   struct {
+      unsigned base_addr;
+   } ss1;
+
+   struct {
+      unsigned pad:2;
+      unsigned mip_count:4;
+      unsigned width:13;
+      unsigned height:13;
+   } ss2;
+
+   struct {
+      unsigned tile_walk:1;
+      unsigned tiled_surface:1;
+      unsigned pad:1;
+      unsigned pitch:18;
+      unsigned depth:11;
+   } ss3;
+
+   struct {
+      unsigned pad:19;
+      unsigned min_array_elt:9;
+      unsigned min_lod:4;
+   } ss4;
+};
+
+
+
+struct brw_vertex_buffer_state
+{
+   struct {
+      unsigned pitch:11;
+      unsigned pad:15;
+      unsigned access_type:1;
+      unsigned vb_index:5;
+   } vb0;
+
+   unsigned start_addr;
+   unsigned max_index;
+#if 1
+   unsigned instance_data_step_rate; /* not included for sequential/random vertices? */
+#endif
+};
+
+#define BRW_VBP_MAX 17
+
+struct brw_vb_array_state {
+   struct header header;
+   struct brw_vertex_buffer_state vb[BRW_VBP_MAX];
+};
+
+
+struct brw_vertex_element_state
+{
+   struct
+   {
+      unsigned src_offset:11;
+      unsigned pad:5;
+      unsigned src_format:9;
+      unsigned pad0:1;
+      unsigned valid:1;
+      unsigned vertex_buffer_index:5;
+   } ve0;
+
+   struct
+   {
+      unsigned dst_offset:8;
+      unsigned pad:8;
+      unsigned vfcomponent3:4;
+      unsigned vfcomponent2:4;
+      unsigned vfcomponent1:4;
+      unsigned vfcomponent0:4;
+   } ve1;
+};
+
+#define BRW_VEP_MAX 18
+
+struct brw_vertex_element_packet {
+   struct header header;
+   struct brw_vertex_element_state ve[BRW_VEP_MAX]; /* note: less than _TNL_ATTRIB_MAX */
+};
+
+
+struct brw_urb_immediate {
+   unsigned opcode:4;
+   unsigned offset:6;
+   unsigned swizzle_control:2;
+   unsigned pad:1;
+   unsigned allocate:1;
+   unsigned used:1;
+   unsigned complete:1;
+   unsigned response_length:4;
+   unsigned msg_length:4;
+   unsigned msg_target:4;
+   unsigned pad1:3;
+   unsigned end_of_thread:1;
+};
+
+/* Instruction format for the execution units:
+ */
+
+struct brw_instruction
+{
+   struct
+   {
+      unsigned opcode:7;
+      unsigned pad:1;
+      unsigned access_mode:1;
+      unsigned mask_control:1;
+      unsigned dependency_control:2;
+      unsigned compression_control:2;
+      unsigned thread_control:2;
+      unsigned predicate_control:4;
+      unsigned predicate_inverse:1;
+      unsigned execution_size:3;
+      unsigned destreg__conditonalmod:4; /* destreg - send, conditionalmod - others */
+      unsigned pad0:2;
+      unsigned debug_control:1;
+      unsigned saturate:1;
+   } header;
+
+   union {
+      struct
+      {
+	 unsigned dest_reg_file:2;
+	 unsigned dest_reg_type:3;
+	 unsigned src0_reg_file:2;
+	 unsigned src0_reg_type:3;
+	 unsigned src1_reg_file:2;
+	 unsigned src1_reg_type:3;
+	 unsigned pad:1;
+	 unsigned dest_subreg_nr:5;
+	 unsigned dest_reg_nr:8;
+	 unsigned dest_horiz_stride:2;
+	 unsigned dest_address_mode:1;
+      } da1;
+
+      struct
+      {
+	 unsigned dest_reg_file:2;
+	 unsigned dest_reg_type:3;
+	 unsigned src0_reg_file:2;
+	 unsigned src0_reg_type:3;
+	 unsigned pad:6;
+	 int dest_indirect_offset:10;	/* offset against the deref'd address reg */
+	 unsigned dest_subreg_nr:3; /* subnr for the address reg a0.x */
+	 unsigned dest_horiz_stride:2;
+	 unsigned dest_address_mode:1;
+      } ia1;
+
+      struct
+      {
+	 unsigned dest_reg_file:2;
+	 unsigned dest_reg_type:3;
+	 unsigned src0_reg_file:2;
+	 unsigned src0_reg_type:3;
+	 unsigned src1_reg_file:2;
+	 unsigned src1_reg_type:3;
+	 unsigned pad0:1;
+	 unsigned dest_writemask:4;
+	 unsigned dest_subreg_nr:1;
+	 unsigned dest_reg_nr:8;
+	 unsigned pad1:2;
+	 unsigned dest_address_mode:1;
+      } da16;
+
+      struct
+      {
+	 unsigned dest_reg_file:2;
+	 unsigned dest_reg_type:3;
+	 unsigned src0_reg_file:2;
+	 unsigned src0_reg_type:3;
+	 unsigned pad0:6;
+	 unsigned dest_writemask:4;
+	 int dest_indirect_offset:6;
+	 unsigned dest_subreg_nr:3;
+	 unsigned pad1:2;
+	 unsigned dest_address_mode:1;
+      } ia16;
+   } bits1;
+
+
+   union {
+      struct
+      {
+	 unsigned src0_subreg_nr:5;
+	 unsigned src0_reg_nr:8;
+	 unsigned src0_abs:1;
+	 unsigned src0_negate:1;
+	 unsigned src0_address_mode:1;
+	 unsigned src0_horiz_stride:2;
+	 unsigned src0_width:3;
+	 unsigned src0_vert_stride:4;
+	 unsigned flag_reg_nr:1;
+	 unsigned pad:6;
+      } da1;
+
+      struct
+      {
+	 int src0_indirect_offset:10;
+	 unsigned src0_subreg_nr:3;
+	 unsigned src0_abs:1;
+	 unsigned src0_negate:1;
+	 unsigned src0_address_mode:1;
+	 unsigned src0_horiz_stride:2;
+	 unsigned src0_width:3;
+	 unsigned src0_vert_stride:4;
+	 unsigned flag_reg_nr:1;
+	 unsigned pad:6;
+      } ia1;
+
+      struct
+      {
+	 unsigned src0_swz_x:2;
+	 unsigned src0_swz_y:2;
+	 unsigned src0_subreg_nr:1;
+	 unsigned src0_reg_nr:8;
+	 unsigned src0_abs:1;
+	 unsigned src0_negate:1;
+	 unsigned src0_address_mode:1;
+	 unsigned src0_swz_z:2;
+	 unsigned src0_swz_w:2;
+	 unsigned pad0:1;
+	 unsigned src0_vert_stride:4;
+	 unsigned flag_reg_nr:1;
+	 unsigned pad1:6;
+      } da16;
+
+      struct
+      {
+	 unsigned src0_swz_x:2;
+	 unsigned src0_swz_y:2;
+	 int src0_indirect_offset:6;
+	 unsigned src0_subreg_nr:3;
+	 unsigned src0_abs:1;
+	 unsigned src0_negate:1;
+	 unsigned src0_address_mode:1;
+	 unsigned src0_swz_z:2;
+	 unsigned src0_swz_w:2;
+	 unsigned pad0:1;
+	 unsigned src0_vert_stride:4;
+	 unsigned flag_reg_nr:1;
+	 unsigned pad1:6;
+      } ia16;
+
+   } bits2;
+
+   union
+   {
+      struct
+      {
+	 unsigned src1_subreg_nr:5;
+	 unsigned src1_reg_nr:8;
+	 unsigned src1_abs:1;
+	 unsigned src1_negate:1;
+	 unsigned pad:1;
+	 unsigned src1_horiz_stride:2;
+	 unsigned src1_width:3;
+	 unsigned src1_vert_stride:4;
+	 unsigned pad0:7;
+      } da1;
+
+      struct
+      {
+	 unsigned src1_swz_x:2;
+	 unsigned src1_swz_y:2;
+	 unsigned src1_subreg_nr:1;
+	 unsigned src1_reg_nr:8;
+	 unsigned src1_abs:1;
+	 unsigned src1_negate:1;
+	 unsigned pad0:1;
+	 unsigned src1_swz_z:2;
+	 unsigned src1_swz_w:2;
+	 unsigned pad1:1;
+	 unsigned src1_vert_stride:4;
+	 unsigned pad2:7;
+      } da16;
+
+      struct
+      {
+	 int  src1_indirect_offset:10;
+	 unsigned src1_subreg_nr:3;
+	 unsigned src1_abs:1;
+	 unsigned src1_negate:1;
+	 unsigned pad0:1;
+	 unsigned src1_horiz_stride:2;
+	 unsigned src1_width:3;
+	 unsigned src1_vert_stride:4;
+	 unsigned flag_reg_nr:1;
+	 unsigned pad1:6;
+      } ia1;
+
+      struct
+      {
+	 unsigned src1_swz_x:2;
+	 unsigned src1_swz_y:2;
+	 int  src1_indirect_offset:6;
+	 unsigned src1_subreg_nr:3;
+	 unsigned src1_abs:1;
+	 unsigned src1_negate:1;
+	 unsigned pad0:1;
+	 unsigned src1_swz_z:2;
+	 unsigned src1_swz_w:2;
+	 unsigned pad1:1;
+	 unsigned src1_vert_stride:4;
+	 unsigned flag_reg_nr:1;
+	 unsigned pad2:6;
+      } ia16;
+
+
+      struct
+      {
+	 int  jump_count:16;	/* note: signed */
+	 unsigned  pop_count:4;
+	 unsigned  pad0:12;
+      } if_else;
+
+      struct {
+	 unsigned function:4;
+	 unsigned int_type:1;
+	 unsigned precision:1;
+	 unsigned saturate:1;
+	 unsigned data_type:1;
+	 unsigned pad0:8;
+	 unsigned response_length:4;
+	 unsigned msg_length:4;
+	 unsigned msg_target:4;
+	 unsigned pad1:3;
+	 unsigned end_of_thread:1;
+      } math;
+
+      struct {
+	 unsigned binding_table_index:8;
+	 unsigned sampler:4;
+	 unsigned return_format:2;
+	 unsigned msg_type:2;
+	 unsigned response_length:4;
+	 unsigned msg_length:4;
+	 unsigned msg_target:4;
+	 unsigned pad1:3;
+	 unsigned end_of_thread:1;
+      } sampler;
+
+      struct brw_urb_immediate urb;
+
+      struct {
+	 unsigned binding_table_index:8;
+	 unsigned msg_control:4;
+	 unsigned msg_type:2;
+	 unsigned target_cache:2;
+	 unsigned response_length:4;
+	 unsigned msg_length:4;
+	 unsigned msg_target:4;
+	 unsigned pad1:3;
+	 unsigned end_of_thread:1;
+      } dp_read;
+
+      struct {
+	 unsigned binding_table_index:8;
+	 unsigned msg_control:3;
+	 unsigned pixel_scoreboard_clear:1;
+	 unsigned msg_type:3;
+	 unsigned send_commit_msg:1;
+	 unsigned response_length:4;
+	 unsigned msg_length:4;
+	 unsigned msg_target:4;
+	 unsigned pad1:3;
+	 unsigned end_of_thread:1;
+      } dp_write;
+
+      struct {
+	 unsigned pad:16;
+	 unsigned response_length:4;
+	 unsigned msg_length:4;
+	 unsigned msg_target:4;
+	 unsigned pad1:3;
+	 unsigned end_of_thread:1;
+      } generic;
+
+      int d;
+      unsigned ud;
+   } bits3;
+};
+
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c
new file mode 100644
index 0000000000..0a95dce194
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_surface.c
@@ -0,0 +1,127 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "brw_blit.h"
+#include "brw_context.h"
+#include "brw_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_tile.h"
+#include "util/u_rect.h"
+
+
+
+/* Assumes all values are within bounds -- no checking at this level -
+ * do it higher up if required.
+ */
+static void
+brw_surface_copy(struct pipe_context *pipe,
+                 boolean do_flip,
+                 struct pipe_surface *dst,
+                 unsigned dstx, unsigned dsty,
+                 struct pipe_surface *src,
+                 unsigned srcx, unsigned srcy, unsigned width, unsigned height)
+{
+   assert( dst != src );
+   assert( dst->block.size == src->block.size );
+   assert( dst->block.width == src->block.height );
+   assert( dst->block.height == src->block.height );
+
+   if (0) {
+      void *dst_map = pipe->screen->surface_map( pipe->screen,
+                                                 dst,
+                                                 PIPE_BUFFER_USAGE_CPU_WRITE );
+      
+      const void *src_map = pipe->screen->surface_map( pipe->screen,
+                                                       src,
+                                                       PIPE_BUFFER_USAGE_CPU_READ );
+      
+      pipe_copy_rect(dst_map,
+                     &dst->block,
+                     dst->stride,
+                     dstx, dsty, 
+                     width, height, 
+                     src_map, 
+                     do_flip ? -(int) src->stride : src->stride, 
+                     srcx, do_flip ? height - 1 - srcy : srcy);
+
+      pipe->screen->surface_unmap(pipe->screen, src);
+      pipe->screen->surface_unmap(pipe->screen, dst);
+   }
+   else {
+      struct brw_texture *dst_tex = (struct brw_texture *)dst->texture;
+      struct brw_texture *src_tex = (struct brw_texture *)src->texture;
+      assert(dst->block.width == 1);
+      assert(dst->block.height == 1);
+      brw_copy_blit(brw_context(pipe),
+                    do_flip,
+                    dst->block.size,
+                    (short) src->stride/src->block.size, src_tex->buffer, src->offset, FALSE,
+                    (short) dst->stride/dst->block.size, dst_tex->buffer, dst->offset, FALSE,
+                    (short) srcx, (short) srcy, (short) dstx, (short) dsty,
+                    (short) width, (short) height, PIPE_LOGICOP_COPY);
+   }
+}
+
+
+static void
+brw_surface_fill(struct pipe_context *pipe,
+                 struct pipe_surface *dst,
+                 unsigned dstx, unsigned dsty,
+                 unsigned width, unsigned height, unsigned value)
+{
+   if (0) {
+      void *dst_map = pipe->screen->surface_map( pipe->screen,
+                                                 dst,
+                                                 PIPE_BUFFER_USAGE_CPU_WRITE );
+
+      pipe_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value);
+
+      pipe->screen->surface_unmap(pipe->screen, dst);
+   }
+   else {
+      struct brw_texture *tex = (struct brw_texture *)dst->texture;
+      assert(dst->block.width == 1);
+      assert(dst->block.height == 1);
+      brw_fill_blit(brw_context(pipe),
+                    dst->block.size,
+                    (short) dst->stride/dst->block.size, 
+                    tex->buffer, dst->offset, FALSE,
+                    (short) dstx, (short) dsty,
+                    (short) width, (short) height,
+                    value);
+   }
+}
+
+
+void
+brw_init_surface_functions(struct brw_context *brw)
+{
+   brw->pipe.surface_copy  = brw_surface_copy;
+   brw->pipe.surface_fill  = brw_surface_fill;
+}
diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c
new file mode 100644
index 0000000000..448229ed4e
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_tex_layout.c
@@ -0,0 +1,398 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+/* Code to layout images in a mipmap tree for i965.
+ */
+
+#include "pipe/p_state.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "brw_context.h"
+#include "brw_tex_layout.h"
+
+
+#define FILE_DEBUG_FLAG DEBUG_TEXTURE
+
+#if 0
+unsigned intel_compressed_alignment(unsigned internalFormat)
+{
+    unsigned alignment = 4;
+
+    switch (internalFormat) {
+    case GL_COMPRESSED_RGB_FXT1_3DFX:
+    case GL_COMPRESSED_RGBA_FXT1_3DFX:
+        alignment = 8;
+        break;
+
+    default:
+        break;
+    }
+
+    return alignment;
+}
+#endif
+
+static unsigned minify( unsigned d )
+{
+   return MAX2(1, d>>1);
+}
+
+
+static void intel_miptree_set_image_offset(struct brw_texture *tex,
+                                           unsigned level,
+                                           unsigned img,
+                                           unsigned x, unsigned y)
+{
+   struct pipe_texture *pt = &tex->base;
+   if (img == 0 && level == 0)
+      assert(x == 0 && y == 0);
+   assert(img < tex->nr_images[level]);
+
+   tex->image_offset[level][img] = y * tex->stride + x * pt->block.size;
+}
+
+static void intel_miptree_set_level_info(struct brw_texture *tex,
+                                         unsigned level,
+                                         unsigned nr_images,
+                                         unsigned x, unsigned y,
+                                         unsigned w, unsigned h, unsigned d)
+{
+   struct pipe_texture *pt = &tex->base;
+
+   assert(level < PIPE_MAX_TEXTURE_LEVELS);
+
+   pt->width[level] = w;
+   pt->height[level] = h;
+   pt->depth[level] = d;
+   
+   pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w);
+   pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h);
+
+   tex->level_offset[level] = y * tex->stride + x * tex->base.block.size;
+   tex->nr_images[level] = nr_images;
+
+   /*
+   DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
+       level, w, h, d, x, y, tex->level_offset[level]);
+   */
+
+   /* Not sure when this would happen, but anyway: 
+    */
+   if (tex->image_offset[level]) {
+      FREE(tex->image_offset[level]);
+      tex->image_offset[level] = NULL;
+   }
+
+   assert(nr_images);
+   assert(!tex->image_offset[level]);
+
+   tex->image_offset[level] = (unsigned *) MALLOC(nr_images * sizeof(unsigned));
+   tex->image_offset[level][0] = 0;
+}
+
+static void i945_miptree_layout_2d(struct brw_texture *tex)
+{
+   struct pipe_texture *pt = &tex->base;
+   const int align_x = 2, align_y = 4;
+   unsigned level;
+   unsigned x = 0;
+   unsigned y = 0;
+   unsigned width = pt->width[0];
+   unsigned height = pt->height[0];
+   unsigned nblocksx = pt->nblocksx[0];
+   unsigned nblocksy = pt->nblocksy[0];
+
+   tex->stride = align(pt->nblocksx[0] * pt->block.size, 4);
+
+   /* May need to adjust pitch to accomodate the placement of
+    * the 2nd mipmap level.  This occurs when the alignment
+    * constraints of mipmap placement push the right edge of the
+    * 2nd mipmap level out past the width of its parent.
+    */
+   if (pt->last_level > 0) {
+      unsigned mip1_nblocksx 
+	 = align(pf_get_nblocksx(&pt->block, minify(width)), align_x)
+         + pf_get_nblocksx(&pt->block, minify(minify(width)));
+
+      if (mip1_nblocksx > nblocksx)
+	 tex->stride = mip1_nblocksx * pt->block.size;
+   }
+
+   /* Pitch must be a whole number of dwords
+    */
+   tex->stride = align(tex->stride, 64);
+   tex->total_nblocksy = 0;
+
+   for (level = 0; level <= pt->last_level; level++) {
+      intel_miptree_set_level_info(tex, level, 1, x, y, width,
+				   height, 1);
+
+      nblocksy = align(nblocksy, align_y);
+
+      /* Because the images are packed better, the final offset
+       * might not be the maximal one:
+       */
+      tex->total_nblocksy = MAX2(tex->total_nblocksy, y + nblocksy);
+
+      /* Layout_below: step right after second mipmap level.
+       */
+      if (level == 1) {
+	 x += align(nblocksx, align_x);
+      }
+      else {
+	 y += nblocksy;
+      }
+
+      width  = minify(width);
+      height = minify(height);
+      nblocksx = pf_get_nblocksx(&pt->block, width);
+      nblocksy = pf_get_nblocksy(&pt->block, height);
+   }
+}
+
+static boolean brw_miptree_layout(struct brw_texture *tex)
+{
+   struct pipe_texture *pt = &tex->base;
+   /* XXX: these vary depending on image format:
+    */
+/*    int align_w = 4; */
+
+   switch (pt->target) {
+   case PIPE_TEXTURE_CUBE:
+   case PIPE_TEXTURE_3D: {
+      unsigned width  = pt->width[0];
+      unsigned height = pt->height[0];
+      unsigned depth = pt->depth[0];
+      unsigned nblocksx = pt->nblocksx[0];
+      unsigned nblocksy = pt->nblocksy[0];
+      unsigned pack_x_pitch, pack_x_nr;
+      unsigned pack_y_pitch;
+      unsigned level;
+      unsigned align_h = 2;
+      unsigned align_w = 4;
+
+      tex->total_nblocksy = 0;
+
+      tex->stride = align(pt->nblocksx[0], 4);
+      pack_y_pitch = align(pt->nblocksy[0], align_h);
+
+      pack_x_pitch = tex->stride / pt->block.size;
+      pack_x_nr = 1;
+
+      for (level = 0; level <= pt->last_level; level++) {
+	 unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6;
+	 int x = 0;
+	 int y = 0;
+	 uint q, j;
+
+	 intel_miptree_set_level_info(tex, level, nr_images,
+				      0, tex->total_nblocksy,
+				      width, height, depth);
+
+	 for (q = 0; q < nr_images;) {
+	    for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) {
+	       intel_miptree_set_image_offset(tex, level, q, x, y);
+	       x += pack_x_pitch;
+	    }
+
+	    x = 0;
+	    y += pack_y_pitch;
+	 }
+
+
+	 tex->total_nblocksy += y;
+	 width  = minify(width);
+	 height = minify(height);
+	 depth  = minify(depth);
+         nblocksx = pf_get_nblocksx(&pt->block, width);
+         nblocksy = pf_get_nblocksy(&pt->block, height);
+
+         if (pt->compressed) {
+            pack_y_pitch = (height + 3) / 4;
+
+            if (pack_x_pitch > align(width, align_w)) {
+               pack_x_pitch = align(width, align_w);
+               pack_x_nr <<= 1;
+            }
+         } else {
+            if (pack_x_pitch > 4) {
+               pack_x_pitch >>= 1;
+               pack_x_nr <<= 1;
+               assert(pack_x_pitch * pack_x_nr * pt->block.size <= tex->stride);
+            }
+
+            if (pack_y_pitch > 2) {
+               pack_y_pitch >>= 1;
+               pack_y_pitch = align(pack_y_pitch, align_h);
+            }
+         }
+
+      }
+      break;
+   }
+
+   default:
+      i945_miptree_layout_2d(tex);
+      break;
+   }
+#if 0
+   PRINT("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__,
+       pt->pitch,
+       pt->total_nblocksy,
+       pt->block.size,
+       pt->stride * pt->total_nblocksy );
+#endif
+
+   return TRUE;
+}
+
+
+static struct pipe_texture *
+brw_texture_create_screen(struct pipe_screen *screen,
+                          const struct pipe_texture *templat)
+{
+   struct pipe_winsys *ws = screen->winsys;
+   struct brw_texture *tex = CALLOC_STRUCT(brw_texture);
+
+   if (tex) {
+      tex->base = *templat;
+      tex->base.refcount = 1;
+
+      tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width[0]);
+      tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height[0]);
+   
+      if (brw_miptree_layout(tex))
+	 tex->buffer = ws->buffer_create(ws, 64,
+                                          PIPE_BUFFER_USAGE_PIXEL,
+                                          tex->stride *
+                                          tex->total_nblocksy);
+
+      if (!tex->buffer) {
+	 FREE(tex);
+         return NULL;
+      }
+   }
+
+   return &tex->base;
+}
+
+
+static void
+brw_texture_release_screen(struct pipe_screen *screen,
+                           struct pipe_texture **pt)
+{
+   if (!*pt)
+      return;
+
+   /*
+   DBG("%s %p refcount will be %d\n",
+       __FUNCTION__, (void *) *pt, (*pt)->refcount - 1);
+   */
+   if (--(*pt)->refcount <= 0) {
+      struct brw_texture *tex = (struct brw_texture *)*pt;
+      uint i;
+
+      /*
+      DBG("%s deleting %p\n", __FUNCTION__, (void *) tex);
+      */
+
+      pipe_buffer_reference(screen, &tex->buffer, NULL);
+
+      for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++)
+         if (tex->image_offset[i])
+            free(tex->image_offset[i]);
+
+      free(tex);
+   }
+   *pt = NULL;
+}
+
+
+static struct pipe_surface *
+brw_get_tex_surface_screen(struct pipe_screen *screen,
+                           struct pipe_texture *pt,
+                           unsigned face, unsigned level, unsigned zslice)
+{
+   struct brw_texture *tex = (struct brw_texture *)pt;
+   struct pipe_surface *ps;
+   unsigned offset;  /* in bytes */
+
+   offset = tex->level_offset[level];
+
+   if (pt->target == PIPE_TEXTURE_CUBE) {
+      offset += tex->image_offset[level][face];
+   }
+   else if (pt->target == PIPE_TEXTURE_3D) {
+      offset += tex->image_offset[level][zslice];
+   }
+   else {
+      assert(face == 0);
+      assert(zslice == 0);
+   }
+
+   ps = CALLOC_STRUCT(pipe_surface);
+   if (ps) {
+      ps->refcount = 1;
+      pipe_texture_reference(&ps->texture, pt);
+      ps->format = pt->format;
+      ps->width = pt->width[level];
+      ps->height = pt->height[level];
+      ps->block = pt->block;
+      ps->nblocksx = pt->nblocksx[level];
+      ps->nblocksy = pt->nblocksy[level];
+      ps->stride = tex->stride;
+      ps->offset = offset;
+      ps->status = PIPE_SURFACE_STATUS_DEFINED;
+   }
+   return ps;
+}
+
+
+void
+brw_init_texture_functions(struct brw_context *brw)
+{
+//   brw->pipe.texture_update = brw_texture_update;
+}
+
+
+void
+brw_init_screen_texture_funcs(struct pipe_screen *screen)
+{
+   screen->texture_create  = brw_texture_create_screen;
+   screen->texture_release = brw_texture_release_screen;
+   screen->get_tex_surface = brw_get_tex_surface_screen;
+}
+
diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.h b/src/gallium/drivers/i965simple/brw_tex_layout.h
new file mode 100644
index 0000000000..a6b6ba8146
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_tex_layout.h
@@ -0,0 +1,44 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+
+
+#ifndef BRW_TEX_LAYOUT_H
+#define BRW_TEX_LAYOUT_H
+
+
+struct brw_context;
+struct pipe_screen;
+
+
+extern void
+brw_init_texture_functions(struct brw_context *brw);
+
+extern void
+brw_init_screen_texture_funcs(struct pipe_screen *screen);
+
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_urb.c b/src/gallium/drivers/i965simple/brw_urb.c
new file mode 100644
index 0000000000..101a4367b9
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_urb.c
@@ -0,0 +1,186 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_context.h"
+//#include "brw_state.h"
+#include "brw_batch.h"
+#include "brw_defines.h"
+
+#define VS 0
+#define GS 1
+#define CLP 2
+#define SF 3
+#define CS 4
+
+/* XXX: Are the min_entry_size numbers useful?
+ * XXX: Verify min_nr_entries, esp for VS.
+ * XXX: Verify SF min_entry_size.
+ */
+static const struct {
+   unsigned min_nr_entries;
+   unsigned preferred_nr_entries;
+   unsigned min_entry_size;
+   unsigned max_entry_size;
+} limits[CS+1] = {
+   { 8, 32, 1, 5 },			/* vs */
+   { 4, 8,  1, 5 },			/* gs */
+   { 6, 8,  1, 5 },			/* clp */
+   { 1, 8,  1, 12 },		        /* sf */
+   { 1, 4,  1, 32 }			/* cs */
+};
+
+
+static boolean check_urb_layout( struct brw_context *brw )
+{
+   brw->urb.vs_start = 0;
+   brw->urb.gs_start = brw->urb.nr_vs_entries * brw->urb.vsize;
+   brw->urb.clip_start = brw->urb.gs_start + brw->urb.nr_gs_entries * brw->urb.vsize;
+   brw->urb.sf_start = brw->urb.clip_start + brw->urb.nr_clip_entries * brw->urb.vsize;
+   brw->urb.cs_start = brw->urb.sf_start + brw->urb.nr_sf_entries * brw->urb.sfsize;
+
+   return brw->urb.cs_start + brw->urb.nr_cs_entries * brw->urb.csize <= 256;
+}
+
+/* Most minimal update, forces re-emit of URB fence packet after GS
+ * unit turned on/off.
+ */
+static void recalculate_urb_fence( struct brw_context *brw )
+{
+   unsigned csize = brw->curbe.total_size;
+   unsigned vsize = brw->vs.prog_data->urb_entry_size;
+   unsigned sfsize = brw->sf.prog_data->urb_entry_size;
+
+   if (csize < limits[CS].min_entry_size)
+      csize = limits[CS].min_entry_size;
+
+   if (vsize < limits[VS].min_entry_size)
+      vsize = limits[VS].min_entry_size;
+
+   if (sfsize < limits[SF].min_entry_size)
+      sfsize = limits[SF].min_entry_size;
+
+   if (brw->urb.vsize < vsize ||
+       brw->urb.sfsize < sfsize ||
+       brw->urb.csize < csize ||
+       (brw->urb.constrained && (brw->urb.vsize > brw->urb.vsize ||
+				 brw->urb.sfsize > brw->urb.sfsize ||
+				 brw->urb.csize > brw->urb.csize))) {
+
+
+      brw->urb.csize = csize;
+      brw->urb.sfsize = sfsize;
+      brw->urb.vsize = vsize;
+
+      brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries;
+      brw->urb.nr_gs_entries = limits[GS].preferred_nr_entries;
+      brw->urb.nr_clip_entries = limits[CLP].preferred_nr_entries;
+      brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries;
+      brw->urb.nr_cs_entries = limits[CS].preferred_nr_entries;
+
+      if (!check_urb_layout(brw)) {
+	 brw->urb.nr_vs_entries = limits[VS].min_nr_entries;
+	 brw->urb.nr_gs_entries = limits[GS].min_nr_entries;
+	 brw->urb.nr_clip_entries = limits[CLP].min_nr_entries;
+	 brw->urb.nr_sf_entries = limits[SF].min_nr_entries;
+	 brw->urb.nr_cs_entries = limits[CS].min_nr_entries;
+
+	 brw->urb.constrained = 1;
+
+	 if (!check_urb_layout(brw)) {
+	    /* This is impossible, given the maximal sizes of urb
+	     * entries and the values for minimum nr of entries
+	     * provided above.
+	     */
+	    debug_printf("couldn't calculate URB layout!\n");
+	    exit(1);
+	 }
+
+	 if (BRW_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS))
+	    debug_printf("URB CONSTRAINED\n");
+      }
+      else
+	 brw->urb.constrained = 0;
+
+      if (BRW_DEBUG & DEBUG_URB)
+	 debug_printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n",
+		      brw->urb.vs_start,
+		      brw->urb.gs_start,
+		      brw->urb.clip_start,
+		      brw->urb.sf_start,
+		      brw->urb.cs_start,
+		      256);
+
+      brw->state.dirty.brw |= BRW_NEW_URB_FENCE;
+   }
+}
+
+
+const struct brw_tracked_state brw_recalculate_urb_fence = {
+   .dirty = {
+      .brw = BRW_NEW_CURBE_OFFSETS,
+      .cache = (CACHE_NEW_VS_PROG |
+		CACHE_NEW_SF_PROG)
+   },
+   .update = recalculate_urb_fence
+};
+
+
+
+
+
+void brw_upload_urb_fence(struct brw_context *brw)
+{
+   struct brw_urb_fence uf;
+   memset(&uf, 0, sizeof(uf));
+
+   uf.header.opcode = CMD_URB_FENCE;
+   uf.header.length = sizeof(uf)/4-2;
+   uf.header.vs_realloc = 1;
+   uf.header.gs_realloc = 1;
+   uf.header.clp_realloc = 1;
+   uf.header.sf_realloc = 1;
+   uf.header.vfe_realloc = 1;
+   uf.header.cs_realloc = 1;
+
+   /* The ordering below is correct, not the layout in the
+    * instruction.
+    *
+    * There are 256 urb reg pairs in total.
+    */
+   uf.bits0.vs_fence  = brw->urb.gs_start;
+   uf.bits0.gs_fence  = brw->urb.clip_start;
+   uf.bits0.clp_fence = brw->urb.sf_start;
+   uf.bits1.sf_fence  = brw->urb.cs_start;
+   uf.bits1.cs_fence  = 256;
+
+   BRW_BATCH_STRUCT(brw, &uf);
+}
diff --git a/src/gallium/drivers/i965simple/brw_util.c b/src/gallium/drivers/i965simple/brw_util.c
new file mode 100644
index 0000000000..42391d7c8c
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_util.c
@@ -0,0 +1,104 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_util.h"
+#include "brw_defines.h"
+
+#include "pipe/p_defines.h"
+
+unsigned brw_count_bits( unsigned val )
+{
+   unsigned i;
+   for (i = 0; val ; val >>= 1)
+      if (val & 1)
+	 i++;
+   return i;
+}
+
+
+unsigned brw_translate_blend_equation( int mode )
+{
+   switch (mode) {
+   case PIPE_BLEND_ADD:
+      return BRW_BLENDFUNCTION_ADD;
+   case PIPE_BLEND_MIN:
+      return BRW_BLENDFUNCTION_MIN;
+   case PIPE_BLEND_MAX:
+      return BRW_BLENDFUNCTION_MAX;
+   case PIPE_BLEND_SUBTRACT:
+      return BRW_BLENDFUNCTION_SUBTRACT;
+   case PIPE_BLEND_REVERSE_SUBTRACT:
+      return BRW_BLENDFUNCTION_REVERSE_SUBTRACT;
+   default:
+      assert(0);
+      return BRW_BLENDFUNCTION_ADD;
+   }
+}
+
+unsigned brw_translate_blend_factor( int factor )
+{
+   switch(factor) {
+   case PIPE_BLENDFACTOR_ZERO:
+      return BRW_BLENDFACTOR_ZERO;
+   case PIPE_BLENDFACTOR_SRC_ALPHA:
+      return BRW_BLENDFACTOR_SRC_ALPHA;
+   case PIPE_BLENDFACTOR_ONE:
+      return BRW_BLENDFACTOR_ONE;
+   case PIPE_BLENDFACTOR_SRC_COLOR:
+      return BRW_BLENDFACTOR_SRC_COLOR;
+   case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+      return BRW_BLENDFACTOR_INV_SRC_COLOR;
+   case PIPE_BLENDFACTOR_DST_COLOR:
+      return BRW_BLENDFACTOR_DST_COLOR;
+   case PIPE_BLENDFACTOR_INV_DST_COLOR:
+      return BRW_BLENDFACTOR_INV_DST_COLOR;
+   case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+      return BRW_BLENDFACTOR_INV_SRC_ALPHA;
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      return BRW_BLENDFACTOR_DST_ALPHA;
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      return BRW_BLENDFACTOR_INV_DST_ALPHA;
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+      return BRW_BLENDFACTOR_SRC_ALPHA_SATURATE;
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+      return BRW_BLENDFACTOR_CONST_COLOR;
+   case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+      return BRW_BLENDFACTOR_INV_CONST_COLOR;
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      return BRW_BLENDFACTOR_CONST_ALPHA;
+   case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+      return BRW_BLENDFACTOR_INV_CONST_ALPHA;
+   default:
+      assert(0);
+      return BRW_BLENDFACTOR_ZERO;
+   }
+}
diff --git a/src/gallium/drivers/i965simple/brw_util.h b/src/gallium/drivers/i965simple/brw_util.h
new file mode 100644
index 0000000000..d60e5934db
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_util.h
@@ -0,0 +1,43 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+          
+
+#ifndef BRW_UTIL_H
+#define BRW_UTIL_H
+
+#include "pipe/p_state.h"
+
+extern unsigned brw_count_bits( unsigned val );
+extern unsigned brw_translate_blend_factor( int factor );
+extern unsigned brw_translate_blend_equation( int mode );
+
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_vs.c b/src/gallium/drivers/i965simple/brw_vs.c
new file mode 100644
index 0000000000..92327e896d
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_vs.c
@@ -0,0 +1,120 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_context.h"
+#include "brw_vs.h"
+#include "brw_util.h"
+#include "brw_state.h"
+
+
+static void do_vs_prog( struct brw_context *brw,
+			const struct brw_vertex_program *vp,
+			struct brw_vs_prog_key *key )
+{
+   unsigned program_size;
+   const unsigned *program;
+   struct brw_vs_compile c;
+
+   memset(&c, 0, sizeof(c));
+   memcpy(&c.key, key, sizeof(*key));
+
+   brw_init_compile(&c.func);
+   c.vp = vp;
+
+   c.prog_data.outputs_written = vp->info.num_outputs;
+   c.prog_data.inputs_read = vp->info.num_inputs;
+
+#if 0
+   if (c.key.copy_edgeflag) {
+      c.prog_data.outputs_written |= 1<<VERT_RESULT_EDGE;
+      c.prog_data.inputs_read |= 1<<VERT_ATTRIB_EDGEFLAG;
+   }
+#endif
+
+   /* Emit GEN4 code.
+    */
+   brw_vs_emit(&c);
+
+   /* get the program
+    */
+   program = brw_get_program(&c.func, &program_size);
+
+   /*
+    */
+   brw->vs.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_VS_PROG],
+					      &c.key,
+					      sizeof(c.key),
+					      program,
+					      program_size,
+					      &c.prog_data,
+					      &brw->vs.prog_data);
+}
+
+
+static void brw_upload_vs_prog( struct brw_context *brw )
+{
+   struct brw_vs_prog_key key;
+   const struct brw_vertex_program *vp = brw->attribs.VertexProgram;
+
+   assert(vp);
+
+   memset(&key, 0, sizeof(key));
+
+   /* Just upload the program verbatim for now.  Always send it all
+    * the inputs it asks for, whether they are varying or not.
+    */
+   key.program_string_id = vp->id;
+   key.nr_userclip = brw->attribs.Clip.nr;
+   key.copy_edgeflag = (brw->attribs.Raster->fill_cw != PIPE_POLYGON_MODE_FILL ||
+			brw->attribs.Raster->fill_ccw != PIPE_POLYGON_MODE_FILL);
+
+   /* Make an early check for the key.
+    */
+   if (brw_search_cache(&brw->cache[BRW_VS_PROG],
+			&key, sizeof(key),
+			&brw->vs.prog_data,
+			&brw->vs.prog_gs_offset))
+       return;
+
+   do_vs_prog(brw, vp, &key);
+}
+
+
+/* See brw_vs.c:
+ */
+const struct brw_tracked_state brw_vs_prog = {
+   .dirty = {
+      .brw   = BRW_NEW_VS,
+      .cache = 0
+   },
+   .update = brw_upload_vs_prog
+};
diff --git a/src/gallium/drivers/i965simple/brw_vs.h b/src/gallium/drivers/i965simple/brw_vs.h
new file mode 100644
index 0000000000..070f9dfcae
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_vs.h
@@ -0,0 +1,82 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#ifndef BRW_VS_H
+#define BRW_VS_H
+
+
+#include "brw_context.h"
+#include "brw_eu.h"
+
+
+struct brw_vs_prog_key {
+   unsigned program_string_id;
+   unsigned nr_userclip:4;
+   unsigned copy_edgeflag:1;
+   unsigned know_w_is_one:1;
+   unsigned pad:26;
+};
+
+
+struct brw_vs_compile {
+   struct brw_compile func;
+   struct brw_vs_prog_key key;
+   struct brw_vs_prog_data prog_data;
+
+   const struct brw_vertex_program *vp;
+
+   unsigned nr_inputs;
+
+   unsigned first_output;
+   unsigned nr_outputs;
+
+   unsigned first_tmp;
+   unsigned last_tmp;
+
+   struct brw_reg r0;
+   struct brw_reg r1;
+   struct brw_reg regs[12][128];
+   struct brw_reg tmp;
+   struct brw_reg stack;
+
+   struct {
+       boolean used_in_src;
+       struct brw_reg reg;
+   } output_regs[128];
+
+   struct brw_reg userplane[6];
+
+};
+
+void brw_vs_emit( struct brw_vs_compile *c );
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c
new file mode 100644
index 0000000000..e03d653482
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_vs_emit.c
@@ -0,0 +1,1330 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_context.h"
+#include "brw_vs.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+
+struct brw_prog_info {
+   unsigned num_temps;
+   unsigned num_addrs;
+   unsigned num_consts;
+
+   unsigned writes_psize;
+
+   unsigned pos_idx;
+   unsigned result_edge_idx;
+   unsigned edge_flag_idx;
+   unsigned psize_idx;
+};
+
+/* Do things as simply as possible.  Allocate and populate all regs
+ * ahead of time.
+ */
+static void brw_vs_alloc_regs( struct brw_vs_compile *c,
+                               struct brw_prog_info *info )
+{
+   unsigned i, reg = 0, mrf;
+   unsigned nr_params;
+
+   /* r0 -- reserved as usual
+    */
+   c->r0 = brw_vec8_grf(reg, 0); reg++;
+
+   /* User clip planes from curbe:
+    */
+   if (c->key.nr_userclip) {
+      for (i = 0; i < c->key.nr_userclip; i++) {
+	 c->userplane[i] = stride( brw_vec4_grf(reg+3+i/2, (i%2) * 4), 0, 4, 1);
+      }
+
+      /* Deal with curbe alignment:
+       */
+      reg += ((6+c->key.nr_userclip+3)/4)*2;
+   }
+
+   /* Vertex program parameters from curbe:
+    */
+   nr_params = c->prog_data.max_const;
+   for (i = 0; i < nr_params; i++) {
+      c->regs[TGSI_FILE_CONSTANT][i] = stride(brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1);
+   }
+   reg += (nr_params+1)/2;
+   c->prog_data.curb_read_length = reg - 1;
+
+
+
+   /* Allocate input regs:
+    */
+   c->nr_inputs = c->vp->info.num_inputs;
+   for (i = 0; i < c->nr_inputs; i++) {
+	 c->regs[TGSI_FILE_INPUT][i] = brw_vec8_grf(reg, 0);
+	 reg++;
+   }
+
+
+   /* Allocate outputs: TODO: could organize the non-position outputs
+    * to go straight into message regs.
+    */
+   c->nr_outputs = 0;
+   c->first_output = reg;
+   mrf = 4;
+   for (i = 0; i < c->vp->info.num_outputs; i++) {
+      c->nr_outputs++;
+#if 0
+      if (i == VERT_RESULT_HPOS) {
+         c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0);
+         reg++;
+      }
+      else if (i == VERT_RESULT_PSIZ) {
+         c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0);
+         reg++;
+         mrf++;		/* just a placeholder?  XXX fix later stages & remove this */
+      }
+      else {
+         c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(mrf);
+         mrf++;
+      }
+#else
+      /*treat pos differently for now */
+      if (i == info->pos_idx) {
+         c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0);
+         reg++;
+      } else {
+         c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(mrf);
+         mrf++;
+      }
+#endif
+   }
+
+   /* Allocate program temporaries:
+    */
+   for (i = 0; i < info->num_temps; i++) {
+      c->regs[TGSI_FILE_TEMPORARY][i] = brw_vec8_grf(reg, 0);
+      reg++;
+   }
+
+   /* Address reg(s).  Don't try to use the internal address reg until
+    * deref time.
+    */
+   for (i = 0; i < info->num_addrs; i++) {
+      c->regs[TGSI_FILE_ADDRESS][i] =  brw_reg(BRW_GENERAL_REGISTER_FILE,
+                                               reg,
+                                               0,
+                                               BRW_REGISTER_TYPE_D,
+                                               BRW_VERTICAL_STRIDE_8,
+                                               BRW_WIDTH_8,
+                                               BRW_HORIZONTAL_STRIDE_1,
+                                               BRW_SWIZZLE_XXXX,
+                                               TGSI_WRITEMASK_X);
+      reg++;
+   }
+
+   for (i = 0; i < 128; i++) {
+      if (c->output_regs[i].used_in_src) {
+         c->output_regs[i].reg = brw_vec8_grf(reg, 0);
+         reg++;
+      }
+   }
+
+   c->stack =  brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, reg, 0);
+   reg += 2;
+
+
+   /* Some opcodes need an internal temporary:
+    */
+   c->first_tmp = reg;
+   c->last_tmp = reg;		/* for allocation purposes */
+
+   /* Each input reg holds data from two vertices.  The
+    * urb_read_length is the number of registers read from *each*
+    * vertex urb, so is half the amount:
+    */
+   c->prog_data.urb_read_length = (c->nr_inputs+1)/2;
+
+   c->prog_data.urb_entry_size = (c->nr_outputs+2+3)/4;
+   c->prog_data.total_grf = reg;
+}
+
+
+static struct brw_reg get_tmp( struct brw_vs_compile *c )
+{
+   struct brw_reg tmp = brw_vec8_grf(c->last_tmp, 0);
+
+   if (++c->last_tmp > c->prog_data.total_grf)
+      c->prog_data.total_grf = c->last_tmp;
+
+   return tmp;
+}
+
+static void release_tmp( struct brw_vs_compile *c, struct brw_reg tmp )
+{
+   if (tmp.nr == c->last_tmp-1)
+      c->last_tmp--;
+}
+
+static void release_tmps( struct brw_vs_compile *c )
+{
+   c->last_tmp = c->first_tmp;
+}
+
+
+static void unalias1( struct brw_vs_compile *c,
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      void (*func)( struct brw_vs_compile *,
+				    struct brw_reg,
+				    struct brw_reg ))
+{
+   if (dst.file == arg0.file && dst.nr == arg0.nr) {
+      struct brw_compile *p = &c->func;
+      struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask);
+      func(c, tmp, arg0);
+      brw_MOV(p, dst, tmp);
+   }
+   else {
+      func(c, dst, arg0);
+   }
+}
+
+static void unalias2( struct brw_vs_compile *c,
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      struct brw_reg arg1,
+		      void (*func)( struct brw_vs_compile *,
+				    struct brw_reg,
+				    struct brw_reg,
+				    struct brw_reg ))
+{
+   if ((dst.file == arg0.file && dst.nr == arg0.nr) ||
+       (dst.file == arg1.file && dst.nr == arg1.nr)) {
+      struct brw_compile *p = &c->func;
+      struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask);
+      func(c, tmp, arg0, arg1);
+      brw_MOV(p, dst, tmp);
+   }
+   else {
+      func(c, dst, arg0, arg1);
+   }
+}
+
+static void emit_sop( struct brw_compile *p,
+                      struct brw_reg dst,
+                      struct brw_reg arg0,
+                      struct brw_reg arg1,
+		      unsigned cond)
+{
+   brw_push_insn_state(p);
+   brw_CMP(p, brw_null_reg(), cond, arg0, arg1);
+   brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+   brw_MOV(p, dst, brw_imm_f(1.0f));
+   brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+   brw_MOV(p, dst, brw_imm_f(0.0f));
+   brw_pop_insn_state(p);
+}
+
+static void emit_seq( struct brw_compile *p,
+                      struct brw_reg dst,
+                      struct brw_reg arg0,
+                      struct brw_reg arg1 )
+{
+   emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_EQ);
+}
+
+static void emit_sne( struct brw_compile *p,
+                      struct brw_reg dst,
+                      struct brw_reg arg0,
+                      struct brw_reg arg1 )
+{
+   emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_NEQ);
+}
+static void emit_slt( struct brw_compile *p,
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      struct brw_reg arg1 )
+{
+   emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_L);
+}
+
+static void emit_sle( struct brw_compile *p,
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      struct brw_reg arg1 )
+{
+   emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_LE);
+}
+
+static void emit_sgt( struct brw_compile *p,
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      struct brw_reg arg1 )
+{
+   emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_G);
+}
+
+static void emit_sge( struct brw_compile *p,
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      struct brw_reg arg1 )
+{
+  emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_GE);
+}
+
+static void emit_max( struct brw_compile *p,
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      struct brw_reg arg1 )
+{
+   brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, arg1);
+   brw_SEL(p, dst, arg1, arg0);
+   brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+}
+
+static void emit_min( struct brw_compile *p,
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      struct brw_reg arg1 )
+{
+   brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, arg1);
+   brw_SEL(p, dst, arg0, arg1);
+   brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+}
+
+
+static void emit_math1( struct brw_vs_compile *c,
+			unsigned function,
+			struct brw_reg dst,
+			struct brw_reg arg0,
+			unsigned precision)
+{
+   /* There are various odd behaviours with SEND on the simulator.  In
+    * addition there are documented issues with the fact that the GEN4
+    * processor doesn't do dependency control properly on SEND
+    * results.  So, on balance, this kludge to get around failures
+    * with writemasked math results looks like it might be necessary
+    * whether that turns out to be a simulator bug or not:
+    */
+   struct brw_compile *p = &c->func;
+   struct brw_reg tmp = dst;
+   boolean need_tmp = (dst.dw1.bits.writemask != 0xf ||
+			 dst.file != BRW_GENERAL_REGISTER_FILE);
+
+   if (need_tmp)
+      tmp = get_tmp(c);
+
+   brw_math(p,
+	    tmp,
+	    function,
+	    BRW_MATH_SATURATE_NONE,
+	    2,
+	    arg0,
+	    BRW_MATH_DATA_SCALAR,
+	    precision);
+
+   if (need_tmp) {
+      brw_MOV(p, dst, tmp);
+      release_tmp(c, tmp);
+   }
+}
+
+static void emit_math2( struct brw_vs_compile *c,
+			unsigned function,
+			struct brw_reg dst,
+			struct brw_reg arg0,
+			struct brw_reg arg1,
+			unsigned precision)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg tmp = dst;
+   boolean need_tmp = (dst.dw1.bits.writemask != 0xf ||
+			 dst.file != BRW_GENERAL_REGISTER_FILE);
+
+   if (need_tmp)
+      tmp = get_tmp(c);
+
+   brw_MOV(p, brw_message_reg(3), arg1);
+
+   brw_math(p,
+	    tmp,
+	    function,
+	    BRW_MATH_SATURATE_NONE,
+	    2,
+ 	    arg0,
+	    BRW_MATH_DATA_SCALAR,
+	    precision);
+
+   if (need_tmp) {
+      brw_MOV(p, dst, tmp);
+      release_tmp(c, tmp);
+   }
+}
+
+
+
+static void emit_exp_noalias( struct brw_vs_compile *c,
+			      struct brw_reg dst,
+			      struct brw_reg arg0 )
+{
+   struct brw_compile *p = &c->func;
+
+
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_X) {
+      struct brw_reg tmp = get_tmp(c);
+      struct brw_reg tmp_d = retype(tmp, BRW_REGISTER_TYPE_D);
+
+      /* tmp_d = floor(arg0.x) */
+      brw_RNDD(p, tmp_d, brw_swizzle1(arg0, 0));
+
+      /* result[0] = 2.0 ^ tmp */
+
+      /* Adjust exponent for floating point:
+       * exp += 127
+       */
+      brw_ADD(p, brw_writemask(tmp_d, TGSI_WRITEMASK_X), tmp_d, brw_imm_d(127));
+
+      /* Install exponent and sign.
+       * Excess drops off the edge:
+       */
+      brw_SHL(p, brw_writemask(retype(dst, BRW_REGISTER_TYPE_D), TGSI_WRITEMASK_X),
+	      tmp_d, brw_imm_d(23));
+
+      release_tmp(c, tmp);
+   }
+
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Y) {
+      /* result[1] = arg0.x - floor(arg0.x) */
+      brw_FRC(p, brw_writemask(dst, TGSI_WRITEMASK_Y), brw_swizzle1(arg0, 0));
+   }
+
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) {
+      /* As with the LOG instruction, we might be better off just
+       * doing a taylor expansion here, seeing as we have to do all
+       * the prep work.
+       *
+       * If mathbox partial precision is too low, consider also:
+       * result[3] = result[0] * EXP(result[1])
+       */
+      emit_math1(c,
+		 BRW_MATH_FUNCTION_EXP,
+		 brw_writemask(dst, TGSI_WRITEMASK_Z),
+		 brw_swizzle1(arg0, 0),
+		 BRW_MATH_PRECISION_PARTIAL);
+   }
+
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) {
+      /* result[3] = 1.0; */
+      brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_W), brw_imm_f(1));
+   }
+}
+
+
+static void emit_log_noalias( struct brw_vs_compile *c,
+			      struct brw_reg dst,
+			      struct brw_reg arg0 )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg tmp = dst;
+   struct brw_reg tmp_ud = retype(tmp, BRW_REGISTER_TYPE_UD);
+   struct brw_reg arg0_ud = retype(arg0, BRW_REGISTER_TYPE_UD);
+   boolean need_tmp = (dst.dw1.bits.writemask != 0xf ||
+			 dst.file != BRW_GENERAL_REGISTER_FILE);
+
+   if (need_tmp) {
+      tmp = get_tmp(c);
+      tmp_ud = retype(tmp, BRW_REGISTER_TYPE_UD);
+   }
+
+   /* Perform mant = frexpf(fabsf(x), &exp), adjust exp and mnt
+    * according to spec:
+    *
+    * These almost look likey they could be joined up, but not really
+    * practical:
+    *
+    * result[0].f = (x.i & ((1<<31)-1) >> 23) - 127
+    * result[1].i = (x.i & ((1<<23)-1)        + (127<<23)
+    */
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_XZ) {
+      brw_AND(p,
+	      brw_writemask(tmp_ud, TGSI_WRITEMASK_X),
+	      brw_swizzle1(arg0_ud, 0),
+	      brw_imm_ud((1U<<31)-1));
+
+      brw_SHR(p,
+	      brw_writemask(tmp_ud, TGSI_WRITEMASK_X),
+	      tmp_ud,
+	      brw_imm_ud(23));
+
+      brw_ADD(p,
+	      brw_writemask(tmp, TGSI_WRITEMASK_X),
+	      retype(tmp_ud, BRW_REGISTER_TYPE_D),	/* does it matter? */
+	      brw_imm_d(-127));
+   }
+
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_YZ) {
+      brw_AND(p,
+	      brw_writemask(tmp_ud, TGSI_WRITEMASK_Y),
+	      brw_swizzle1(arg0_ud, 0),
+	      brw_imm_ud((1<<23)-1));
+
+      brw_OR(p,
+	     brw_writemask(tmp_ud, TGSI_WRITEMASK_Y),
+	     tmp_ud,
+	     brw_imm_ud(127<<23));
+   }
+
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) {
+      /* result[2] = result[0] + LOG2(result[1]); */
+
+      /* Why bother?  The above is just a hint how to do this with a
+       * taylor series.  Maybe we *should* use a taylor series as by
+       * the time all the above has been done it's almost certainly
+       * quicker than calling the mathbox, even with low precision.
+       *
+       * Options are:
+       *    - result[0] + mathbox.LOG2(result[1])
+       *    - mathbox.LOG2(arg0.x)
+       *    - result[0] + inline_taylor_approx(result[1])
+       */
+      emit_math1(c,
+		 BRW_MATH_FUNCTION_LOG,
+		 brw_writemask(tmp, TGSI_WRITEMASK_Z),
+		 brw_swizzle1(tmp, 1),
+		 BRW_MATH_PRECISION_FULL);
+
+      brw_ADD(p,
+	      brw_writemask(tmp, TGSI_WRITEMASK_Z),
+	      brw_swizzle1(tmp, 2),
+	      brw_swizzle1(tmp, 0));
+   }
+
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) {
+      /* result[3] = 1.0; */
+      brw_MOV(p, brw_writemask(tmp, TGSI_WRITEMASK_W), brw_imm_f(1));
+   }
+
+   if (need_tmp) {
+      brw_MOV(p, dst, tmp);
+      release_tmp(c, tmp);
+   }
+}
+
+
+
+
+/* Need to unalias - consider swizzles:   r0 = DST r0.xxxx r1
+ */
+static void emit_dst_noalias( struct brw_vs_compile *c,
+			      struct brw_reg dst,
+			      struct brw_reg arg0,
+			      struct brw_reg arg1)
+{
+   struct brw_compile *p = &c->func;
+
+   /* There must be a better way to do this:
+    */
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_X)
+      brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_X), brw_imm_f(1.0));
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Y)
+      brw_MUL(p, brw_writemask(dst, TGSI_WRITEMASK_Y), arg0, arg1);
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z)
+      brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_Z), arg0);
+   if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W)
+      brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_W), arg1);
+}
+
+static void emit_xpd( struct brw_compile *p,
+		      struct brw_reg dst,
+		      struct brw_reg t,
+		      struct brw_reg u)
+{
+   brw_MUL(p, brw_null_reg(), brw_swizzle(t, 1,2,0,3),  brw_swizzle(u,2,0,1,3));
+   brw_MAC(p, dst,     negate(brw_swizzle(t, 2,0,1,3)), brw_swizzle(u,1,2,0,3));
+}
+
+
+
+static void emit_lit_noalias( struct brw_vs_compile *c,
+			      struct brw_reg dst,
+			      struct brw_reg arg0 )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *if_insn;
+   struct brw_reg tmp = dst;
+   boolean need_tmp = (dst.file != BRW_GENERAL_REGISTER_FILE);
+
+   if (need_tmp)
+      tmp = get_tmp(c);
+
+   brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_YZ), brw_imm_f(0));
+   brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_XW), brw_imm_f(1));
+
+   /* Need to use BRW_EXECUTE_8 and also do an 8-wide compare in order
+    * to get all channels active inside the IF.  In the clipping code
+    * we run with NoMask, so it's not an option and we can use
+    * BRW_EXECUTE_1 for all comparisions.
+    */
+   brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, brw_swizzle1(arg0,0), brw_imm_f(0));
+   if_insn = brw_IF(p, BRW_EXECUTE_8);
+   {
+      brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_Y), brw_swizzle1(arg0,0));
+
+      brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, brw_swizzle1(arg0,1), brw_imm_f(0));
+      brw_MOV(p, brw_writemask(tmp, TGSI_WRITEMASK_Z),  brw_swizzle1(arg0,1));
+      brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+
+      emit_math2(c,
+		 BRW_MATH_FUNCTION_POW,
+		 brw_writemask(dst, TGSI_WRITEMASK_Z),
+		 brw_swizzle1(tmp, 2),
+		 brw_swizzle1(arg0, 3),
+		 BRW_MATH_PRECISION_PARTIAL);
+   }
+
+   brw_ENDIF(p, if_insn);
+}
+
+
+
+
+
+/* TODO: relative addressing!
+ */
+static struct brw_reg get_reg( struct brw_vs_compile *c,
+			       unsigned file,
+			       unsigned index )
+{
+   switch (file) {
+   case TGSI_FILE_TEMPORARY:
+   case TGSI_FILE_INPUT:
+   case TGSI_FILE_OUTPUT:
+      assert(c->regs[file][index].nr != 0);
+      return c->regs[file][index];
+   case TGSI_FILE_CONSTANT:
+      assert(c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_imm].nr != 0);
+      return c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_imm];
+   case TGSI_FILE_IMMEDIATE:
+      assert(c->regs[TGSI_FILE_CONSTANT][index].nr != 0);
+      return c->regs[TGSI_FILE_CONSTANT][index];
+   case TGSI_FILE_ADDRESS:
+      assert(index == 0);
+      return c->regs[file][index];
+
+   case TGSI_FILE_NULL:			/* undef values */
+      return brw_null_reg();
+
+   default:
+      assert(0);
+      return brw_null_reg();
+   }
+}
+
+
+
+static struct brw_reg deref( struct brw_vs_compile *c,
+			     struct brw_reg arg,
+			     int offset)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg tmp = vec4(get_tmp(c));
+   struct brw_reg vp_address = retype(vec1(get_reg(c, TGSI_FILE_ADDRESS, 0)), BRW_REGISTER_TYPE_UW);
+   unsigned byte_offset = arg.nr * 32 + arg.subnr + offset * 16;
+   struct brw_reg indirect = brw_vec4_indirect(0,0);
+
+   {
+      brw_push_insn_state(p);
+      brw_set_access_mode(p, BRW_ALIGN_1);
+
+      /* This is pretty clunky - load the address register twice and
+       * fetch each 4-dword value in turn.  There must be a way to do
+       * this in a single pass, but I couldn't get it to work.
+       */
+      brw_ADD(p, brw_address_reg(0), vp_address, brw_imm_d(byte_offset));
+      brw_MOV(p, tmp, indirect);
+
+      brw_ADD(p, brw_address_reg(0), suboffset(vp_address, 8), brw_imm_d(byte_offset));
+      brw_MOV(p, suboffset(tmp, 4), indirect);
+
+      brw_pop_insn_state(p);
+   }
+
+   return vec8(tmp);
+}
+
+
+static void emit_arl( struct brw_vs_compile *c,
+		      struct brw_reg dst,
+		      struct brw_reg arg0 )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg tmp = dst;
+   boolean need_tmp = (dst.file != BRW_GENERAL_REGISTER_FILE);
+
+   if (need_tmp)
+      tmp = get_tmp(c);
+
+   brw_RNDD(p, tmp, arg0);
+   brw_MUL(p, dst, tmp, brw_imm_d(16));
+
+   if (need_tmp)
+      release_tmp(c, tmp);
+}
+
+
+/* Will return mangled results for SWZ op.  The emit_swz() function
+ * ignores this result and recalculates taking extended swizzles into
+ * account.
+ */
+static struct brw_reg get_arg( struct brw_vs_compile *c,
+			       struct tgsi_src_register *src )
+{
+   struct brw_reg reg;
+
+   if (src->File == TGSI_FILE_NULL)
+      return brw_null_reg();
+
+#if 0
+   if (src->RelAddr)
+      reg = deref(c, c->regs[PROGRAM_STATE_VAR][0], src->Index);
+   else
+#endif
+      reg = get_reg(c, src->File, src->Index);
+
+   /* Convert 3-bit swizzle to 2-bit.
+    */
+   reg.dw1.bits.swizzle = BRW_SWIZZLE4(src->SwizzleX,
+				       src->SwizzleY,
+				       src->SwizzleZ,
+				       src->SwizzleW);
+
+   /* Note this is ok for non-swizzle instructions:
+    */
+   reg.negate = src->Negate ? 1 : 0;
+
+   return reg;
+}
+
+
+static struct brw_reg get_dst( struct brw_vs_compile *c,
+			       const struct tgsi_dst_register *dst )
+{
+   struct brw_reg reg = get_reg(c, dst->File, dst->Index);
+
+   reg.dw1.bits.writemask = dst->WriteMask;
+
+   return reg;
+}
+
+
+
+
+static void emit_swz( struct brw_vs_compile *c,
+		      struct brw_reg dst,
+		      struct tgsi_src_register src )
+{
+   struct brw_compile *p = &c->func;
+   unsigned zeros_mask = 0;
+   unsigned ones_mask = 0;
+   unsigned src_mask = 0;
+   ubyte src_swz[4];
+   boolean need_tmp = (src.Negate &&
+			 dst.file != BRW_GENERAL_REGISTER_FILE);
+   struct brw_reg tmp = dst;
+   unsigned i;
+
+   if (need_tmp)
+      tmp = get_tmp(c);
+
+   for (i = 0; i < 4; i++) {
+      if (dst.dw1.bits.writemask & (1<<i)) {
+	 ubyte s = 0;
+         switch(i) {
+         case 0:
+            s = src.SwizzleX;
+            break;
+            s = src.SwizzleY;
+         case 1:
+            break;
+            s = src.SwizzleZ;
+         case 2:
+            break;
+            s = src.SwizzleW;
+         case 3:
+            break;
+         }
+	 switch (s) {
+	 case TGSI_SWIZZLE_X:
+	 case TGSI_SWIZZLE_Y:
+	 case TGSI_SWIZZLE_Z:
+	 case TGSI_SWIZZLE_W:
+	    src_mask |= 1<<i;
+	    src_swz[i] = s;
+	    break;
+	 case TGSI_EXTSWIZZLE_ZERO:
+	    zeros_mask |= 1<<i;
+	    break;
+	 case TGSI_EXTSWIZZLE_ONE:
+	    ones_mask |= 1<<i;
+	    break;
+	 }
+      }
+   }
+
+   /* Do src first, in case dst aliases src:
+    */
+   if (src_mask) {
+      struct brw_reg arg0;
+
+#if 0
+      if (src.RelAddr)
+	 arg0 = deref(c, c->regs[PROGRAM_STATE_VAR][0], src.Index);
+      else
+#endif
+	 arg0 = get_reg(c, src.File, src.Index);
+
+      arg0 = brw_swizzle(arg0,
+			 src_swz[0], src_swz[1],
+			 src_swz[2], src_swz[3]);
+
+      brw_MOV(p, brw_writemask(tmp, src_mask), arg0);
+   }
+
+   if (zeros_mask)
+      brw_MOV(p, brw_writemask(tmp, zeros_mask), brw_imm_f(0));
+
+   if (ones_mask)
+      brw_MOV(p, brw_writemask(tmp, ones_mask), brw_imm_f(1));
+
+   if (src.Negate)
+      brw_MOV(p, brw_writemask(tmp, src.Negate), negate(tmp));
+
+   if (need_tmp) {
+      brw_MOV(p, dst, tmp);
+      release_tmp(c, tmp);
+   }
+}
+
+
+
+/* Post-vertex-program processing.  Send the results to the URB.
+ */
+static void emit_vertex_write( struct brw_vs_compile *c, struct brw_prog_info *info)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg m0 = brw_message_reg(0);
+   struct brw_reg pos = c->regs[TGSI_FILE_OUTPUT][info->pos_idx];
+   struct brw_reg ndc;
+
+   if (c->key.copy_edgeflag) {
+      brw_MOV(p,
+	      get_reg(c, TGSI_FILE_OUTPUT, info->result_edge_idx),
+	      get_reg(c, TGSI_FILE_INPUT, info->edge_flag_idx));
+   }
+
+
+   /* Build ndc coords?   TODO: Shortcircuit when w is known to be one.
+    */
+   if (!c->key.know_w_is_one) {
+      ndc = get_tmp(c);
+      emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL);
+      brw_MUL(p, brw_writemask(ndc, TGSI_WRITEMASK_XYZ), pos, ndc);
+   }
+   else {
+      ndc = pos;
+   }
+
+   /* This includes the workaround for -ve rhw, so is no longer an
+    * optional step:
+    */
+   if (info->writes_psize ||
+       c->key.nr_userclip ||
+       !c->key.know_w_is_one)
+   {
+      struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD);
+      unsigned i;
+
+      brw_MOV(p, header1, brw_imm_ud(0));
+
+      brw_set_access_mode(p, BRW_ALIGN_16);
+
+      if (info->writes_psize) {
+	 struct brw_reg psiz = c->regs[TGSI_FILE_OUTPUT][info->psize_idx];
+	 brw_MUL(p, brw_writemask(header1, TGSI_WRITEMASK_W),
+                 brw_swizzle1(psiz, 0), brw_imm_f(1<<11));
+	 brw_AND(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1,
+                 brw_imm_ud(0x7ff<<8));
+      }
+
+
+      for (i = 0; i < c->key.nr_userclip; i++) {
+	 brw_set_conditionalmod(p, BRW_CONDITIONAL_L);
+	 brw_DP4(p, brw_null_reg(), pos, c->userplane[i]);
+	 brw_OR(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, brw_imm_ud(1<<i));
+	 brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+      }
+
+
+      /* i965 clipping workaround:
+       * 1) Test for -ve rhw
+       * 2) If set,
+       *      set ndc = (0,0,0,0)
+       *      set ucp[6] = 1
+       *
+       * Later, clipping will detect ucp[6] and ensure the primitive is
+       * clipped against all fixed planes.
+       */
+      if (!c->key.know_w_is_one) {
+	 brw_CMP(p,
+		 vec8(brw_null_reg()),
+		 BRW_CONDITIONAL_L,
+		 brw_swizzle1(ndc, 3),
+		 brw_imm_f(0));
+
+	 brw_OR(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, brw_imm_ud(1<<6));
+	 brw_MOV(p, ndc, brw_imm_f(0));
+	 brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+      }
+
+      brw_set_access_mode(p, BRW_ALIGN_1);	/* why? */
+      brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), header1);
+      brw_set_access_mode(p, BRW_ALIGN_16);
+
+      release_tmp(c, header1);
+   }
+   else {
+      brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), brw_imm_ud(0));
+   }
+
+
+   /* Emit the (interleaved) headers for the two vertices - an 8-reg
+    * of zeros followed by two sets of NDC coordinates:
+    */
+   brw_set_access_mode(p, BRW_ALIGN_1);
+   brw_MOV(p, offset(m0, 2), ndc);
+   brw_MOV(p, offset(m0, 3), pos);
+
+
+   brw_urb_WRITE(p,
+		 brw_null_reg(), /* dest */
+		 0,		/* starting mrf reg nr */
+		 c->r0,		/* src */
+		 0,		/* allocate */
+		 1,		/* used */
+		 c->nr_outputs + 3, /* msg len */
+		 0,		/* response len */
+		 1, 		/* eot */
+		 1, 		/* writes complete */
+		 0, 		/* urb destination offset */
+		 BRW_URB_SWIZZLE_INTERLEAVE);
+
+}
+
+static void
+post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst )
+{
+   struct tgsi_parse_context parse;
+   const struct tgsi_token *tokens = c->vp->program.tokens;
+   tgsi_parse_init(&parse, tokens);
+   while (!tgsi_parse_end_of_tokens(&parse)) {
+      tgsi_parse_token(&parse);
+      if (parse.FullToken.Token.Type == TGSI_TOKEN_TYPE_INSTRUCTION) {
+#if 0
+         struct brw_instruction *brw_inst1, *brw_inst2;
+         const struct tgsi_full_instruction *inst1, *inst2;
+         int offset;
+         inst1 = &parse.FullToken.FullInstruction;
+         brw_inst1 = inst1->Data;
+         switch (inst1->Opcode) {
+	 case TGSI_OPCODE_CAL:
+	 case TGSI_OPCODE_BRA:
+	    target_insn = inst1->BranchTarget;
+	    inst2 = &c->vp->program.Base.Instructions[target_insn];
+	    brw_inst2 = inst2->Data;
+	    offset = brw_inst2 - brw_inst1;
+	    brw_set_src1(brw_inst1, brw_imm_d(offset*16));
+	    break;
+	 case TGSI_OPCODE_END:
+	    offset = end_inst - brw_inst1;
+	    brw_set_src1(brw_inst1, brw_imm_d(offset*16));
+	    break;
+	 default:
+	    break;
+         }
+#endif
+      }
+   }
+   tgsi_parse_free(&parse);
+}
+
+static void process_declaration(const struct tgsi_full_declaration *decl,
+                                struct brw_prog_info *info)
+{
+   int first = decl->DeclarationRange.First;
+   int last = decl->DeclarationRange.Last;
+   
+   switch(decl->Declaration.File) {
+   case TGSI_FILE_CONSTANT: 
+      info->num_consts += last - first + 1;
+      break;
+   case TGSI_FILE_INPUT: {
+   }
+      break;
+   case TGSI_FILE_OUTPUT: {
+      assert(last == first);	/* for now */
+      if (decl->Declaration.Semantic) {
+         switch (decl->Semantic.SemanticName) {
+         case TGSI_SEMANTIC_POSITION: {
+            info->pos_idx = first;
+         }
+            break;
+         case TGSI_SEMANTIC_COLOR:
+            break;
+         case TGSI_SEMANTIC_BCOLOR:
+            break;
+         case TGSI_SEMANTIC_FOG:
+            break;
+         case TGSI_SEMANTIC_PSIZE: {
+            info->writes_psize = TRUE;
+            info->psize_idx = first;
+         }
+            break;
+         case TGSI_SEMANTIC_GENERIC:
+            break;
+         }
+      }
+   }
+      break;
+   case TGSI_FILE_TEMPORARY: {
+      info->num_temps += (last - first) + 1;
+   }
+      break;
+   case TGSI_FILE_SAMPLER: {
+   }
+      break;
+   case TGSI_FILE_ADDRESS: {
+      info->num_addrs += (last - first) + 1;
+   }
+      break;
+   case TGSI_FILE_IMMEDIATE: {
+   }
+      break;
+   case TGSI_FILE_NULL: {
+   }
+      break;
+   }
+}
+
+static void process_instruction(struct brw_vs_compile *c,
+                                struct tgsi_full_instruction *inst,
+                                struct brw_prog_info *info)
+{
+   struct brw_reg args[3], dst;
+   struct brw_compile *p = &c->func;
+   /*struct brw_indirect stack_index = brw_indirect(0, 0);*/
+   unsigned i;
+   unsigned index;
+   unsigned file;
+   /*FIXME: might not be the only one*/
+   const struct tgsi_dst_register *dst_reg = &inst->FullDstRegisters[0].DstRegister;
+   /*
+   struct brw_instruction *if_inst[MAX_IFSN];
+   unsigned insn, if_insn = 0;
+   */
+
+   for (i = 0; i < 3; i++) {
+      struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i];
+      index = src->SrcRegister.Index;
+      file = src->SrcRegister.File;
+      if (file == TGSI_FILE_OUTPUT && c->output_regs[index].used_in_src)
+         args[i] = c->output_regs[index].reg;
+      else
+         args[i] = get_arg(c, &src->SrcRegister);
+   }
+
+   /* Get dest regs.  Note that it is possible for a reg to be both
+    * dst and arg, given the static allocation of registers.  So
+    * care needs to be taken emitting multi-operation instructions.
+    */
+   index = dst_reg->Index;
+   file = dst_reg->File;
+   if (file == TGSI_FILE_OUTPUT && c->output_regs[index].used_in_src)
+      dst = c->output_regs[index].reg;
+   else
+      dst = get_dst(c, dst_reg);
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_ABS:
+      brw_MOV(p, dst, brw_abs(args[0]));
+      break;
+   case TGSI_OPCODE_ADD:
+      brw_ADD(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_DP3:
+      brw_DP3(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_DP4:
+      brw_DP4(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_DPH:
+      brw_DPH(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_DST:
+      unalias2(c, dst, args[0], args[1], emit_dst_noalias);
+      break;
+   case TGSI_OPCODE_EXP:
+      unalias1(c, dst, args[0], emit_exp_noalias);
+      break;
+   case TGSI_OPCODE_EX2:
+      emit_math1(c, BRW_MATH_FUNCTION_EXP, dst, args[0], BRW_MATH_PRECISION_FULL);
+      break;
+   case TGSI_OPCODE_ARL:
+      emit_arl(c, dst, args[0]);
+      break;
+   case TGSI_OPCODE_FLR:
+      brw_RNDD(p, dst, args[0]);
+      break;
+   case TGSI_OPCODE_FRC:
+      brw_FRC(p, dst, args[0]);
+      break;
+   case TGSI_OPCODE_LOG:
+      unalias1(c, dst, args[0], emit_log_noalias);
+      break;
+   case TGSI_OPCODE_LG2:
+      emit_math1(c, BRW_MATH_FUNCTION_LOG, dst, args[0], BRW_MATH_PRECISION_FULL);
+      break;
+   case TGSI_OPCODE_LIT:
+      unalias1(c, dst, args[0], emit_lit_noalias);
+      break;
+   case TGSI_OPCODE_MAD:
+      brw_MOV(p, brw_acc_reg(), args[2]);
+      brw_MAC(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_MAX:
+      emit_max(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_MIN:
+      emit_min(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_MOV:
+   case TGSI_OPCODE_SWZ:
+#if 0
+      /* The args[0] value can't be used here as it won't have
+       * correctly encoded the full swizzle:
+       */
+      emit_swz(c, dst, inst->SrcReg[0] );
+#endif
+      brw_MOV(p, dst, args[0]);
+      break;
+   case TGSI_OPCODE_MUL:
+      brw_MUL(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_POW:
+      emit_math2(c, BRW_MATH_FUNCTION_POW, dst, args[0], args[1], BRW_MATH_PRECISION_FULL);
+      break;
+   case TGSI_OPCODE_RCP:
+      emit_math1(c, BRW_MATH_FUNCTION_INV, dst, args[0], BRW_MATH_PRECISION_FULL);
+      break;
+   case TGSI_OPCODE_RSQ:
+      emit_math1(c, BRW_MATH_FUNCTION_RSQ, dst, args[0], BRW_MATH_PRECISION_FULL);
+      break;
+
+   case TGSI_OPCODE_SEQ:
+      emit_seq(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_SNE:
+      emit_sne(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_SGE:
+      emit_sge(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_SGT:
+      emit_sgt(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_SLT:
+      emit_slt(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_SLE:
+      emit_sle(p, dst, args[0], args[1]);
+      break;
+   case TGSI_OPCODE_SUB:
+      brw_ADD(p, dst, args[0], negate(args[1]));
+      break;
+   case TGSI_OPCODE_XPD:
+      emit_xpd(p, dst, args[0], args[1]);
+      break;
+#if 0
+   case TGSI_OPCODE_IF:
+      assert(if_insn < MAX_IFSN);
+      if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8);
+      break;
+   case TGSI_OPCODE_ELSE:
+      if_inst[if_insn-1] = brw_ELSE(p, if_inst[if_insn-1]);
+      break;
+   case TGSI_OPCODE_ENDIF:
+      assert(if_insn > 0);
+      brw_ENDIF(p, if_inst[--if_insn]);
+      break;
+   case TGSI_OPCODE_BRA:
+      brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+      brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
+      brw_set_predicate_control_flag_value(p, 0xff);
+      break;
+   case TGSI_OPCODE_CAL:
+      brw_set_access_mode(p, BRW_ALIGN_1);
+      brw_ADD(p, deref_1uw(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16));
+      brw_set_access_mode(p, BRW_ALIGN_16);
+      brw_ADD(p, get_addr_reg(stack_index),
+              get_addr_reg(stack_index), brw_imm_d(4));
+      inst->Data = &p->store[p->nr_insn];
+      brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
+      break;
+#endif
+   case TGSI_OPCODE_RET:
+#if 0
+      brw_ADD(p, get_addr_reg(stack_index),
+              get_addr_reg(stack_index), brw_imm_d(-4));
+      brw_set_access_mode(p, BRW_ALIGN_1);
+      brw_MOV(p, brw_ip_reg(), deref_1uw(stack_index, 0));
+      brw_set_access_mode(p, BRW_ALIGN_16);
+#else
+      /*brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));*/
+#endif
+      break;
+   case TGSI_OPCODE_END:
+      brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
+      break;
+   case TGSI_OPCODE_BGNSUB:
+   case TGSI_OPCODE_ENDSUB:
+      break;
+   default:
+      debug_printf("Unsupport opcode %d in vertex shader\n", inst->Instruction.Opcode);
+      break;
+   }
+
+   if (dst_reg->File == TGSI_FILE_OUTPUT
+       && dst_reg->Index != info->pos_idx
+       && c->output_regs[dst_reg->Index].used_in_src)
+      brw_MOV(p, get_dst(c, dst_reg), dst);
+
+   release_tmps(c);
+}
+
+/* Emit the fragment program instructions here.
+ */
+void brw_vs_emit(struct brw_vs_compile *c)
+{
+#define MAX_IFSN 32
+   struct brw_compile *p = &c->func;
+   struct brw_instruction *end_inst;
+   struct tgsi_parse_context parse;
+   struct brw_indirect stack_index = brw_indirect(0, 0);
+   const struct tgsi_token *tokens = c->vp->program.tokens;
+   struct brw_prog_info prog_info;
+   unsigned allocated_registers = 0;
+   memset(&prog_info, 0, sizeof(struct brw_prog_info));
+
+   brw_set_compression_control(p, BRW_COMPRESSION_NONE);
+   brw_set_access_mode(p, BRW_ALIGN_16);
+
+   tgsi_parse_init(&parse, tokens);
+   /* Message registers can't be read, so copy the output into GRF register
+      if they are used in source registers */
+   while (!tgsi_parse_end_of_tokens(&parse)) {
+      tgsi_parse_token(&parse);
+      unsigned i;
+      switch (parse.FullToken.Token.Type) {
+      case TGSI_TOKEN_TYPE_INSTRUCTION: {
+         const struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction;
+         for (i = 0; i < 3; ++i) {
+            const struct tgsi_src_register *src = &inst->FullSrcRegisters[i].SrcRegister;
+            unsigned index = src->Index;
+            unsigned file = src->File;
+            if (file == TGSI_FILE_OUTPUT)
+               c->output_regs[index].used_in_src = TRUE;
+         }
+      }
+         break;
+      default:
+         /* nothing */
+         break;
+      }
+   }
+   tgsi_parse_free(&parse);
+
+   tgsi_parse_init(&parse, tokens);
+
+   while (!tgsi_parse_end_of_tokens(&parse)) {
+      tgsi_parse_token(&parse);
+
+      switch (parse.FullToken.Token.Type) {
+      case TGSI_TOKEN_TYPE_DECLARATION: {
+         struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration;
+         process_declaration(decl, &prog_info);
+      }
+         break;
+      case TGSI_TOKEN_TYPE_IMMEDIATE: {
+         struct tgsi_full_immediate *imm = &parse.FullToken.FullImmediate;
+         assert(imm->Immediate.NrTokens == 4 + 1);
+         c->prog_data.imm_buf[c->prog_data.num_imm][0] = imm->u.ImmediateFloat32[0].Float;
+         c->prog_data.imm_buf[c->prog_data.num_imm][1] = imm->u.ImmediateFloat32[1].Float;
+         c->prog_data.imm_buf[c->prog_data.num_imm][2] = imm->u.ImmediateFloat32[2].Float;
+         c->prog_data.imm_buf[c->prog_data.num_imm][3] = imm->u.ImmediateFloat32[3].Float;
+         c->prog_data.num_imm++;
+      }
+         break;
+      case TGSI_TOKEN_TYPE_INSTRUCTION: {
+         struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction;
+         if (!allocated_registers) {
+            /* first instruction (declerations finished).
+             * now that we know what vars are being used allocate
+             * registers for them.*/
+            c->prog_data.num_consts = prog_info.num_consts;
+            c->prog_data.max_const = prog_info.num_consts + c->prog_data.num_imm;
+            brw_vs_alloc_regs(c, &prog_info);
+
+	    brw_set_access_mode(p, BRW_ALIGN_1);
+            brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack));
+	    brw_set_access_mode(p, BRW_ALIGN_16);
+            allocated_registers = 1;
+         }
+         process_instruction(c, inst, &prog_info);
+      }
+         break;
+      }
+   }
+
+   end_inst = &p->store[p->nr_insn];
+   emit_vertex_write(c, &prog_info);
+   post_vs_emit(c, end_inst);
+   tgsi_parse_free(&parse);
+
+}
diff --git a/src/gallium/drivers/i965simple/brw_vs_state.c b/src/gallium/drivers/i965simple/brw_vs_state.c
new file mode 100644
index 0000000000..1eaff87892
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_vs_state.c
@@ -0,0 +1,103 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+static void upload_vs_unit( struct brw_context *brw )
+{
+   struct brw_vs_unit_state vs;
+
+   memset(&vs, 0, sizeof(vs));
+
+   /* CACHE_NEW_VS_PROG */
+   vs.thread0.kernel_start_pointer = brw->vs.prog_gs_offset >> 6;
+   vs.thread0.grf_reg_count = align(brw->vs.prog_data->total_grf, 16) / 16 - 1;
+   vs.thread3.urb_entry_read_length = brw->vs.prog_data->urb_read_length;
+   vs.thread3.const_urb_entry_read_length = brw->vs.prog_data->curb_read_length;
+   vs.thread3.dispatch_grf_start_reg = 1;
+
+
+   /* BRW_NEW_URB_FENCE  */
+   vs.thread4.nr_urb_entries = brw->urb.nr_vs_entries;
+   vs.thread4.urb_entry_allocation_size = brw->urb.vsize - 1;
+   vs.thread4.max_threads = MIN2(
+      MAX2(0, (brw->urb.nr_vs_entries - 6) / 2 - 1),
+      15);
+
+
+
+   if (BRW_DEBUG & DEBUG_SINGLE_THREAD)
+      vs.thread4.max_threads = 0;
+
+   /* BRW_NEW_CURBE_OFFSETS, _NEW_TRANSFORM */
+   if (0 /*brw->attribs.Clip->ClipPlanesEnabled*/) {
+      /* Note that we read in the userclip planes as well, hence
+       * clip_start:
+       */
+      vs.thread3.const_urb_entry_read_offset = brw->curbe.clip_start * 2;
+   }
+   else {
+      vs.thread3.const_urb_entry_read_offset = brw->curbe.vs_start * 2;
+   }
+
+   vs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
+   vs.thread3.urb_entry_read_offset = 0;
+
+   /* No samplers for ARB_vp programs:
+    */
+   vs.vs5.sampler_count = 0;
+
+   if (BRW_DEBUG & DEBUG_STATS)
+      vs.thread4.stats_enable = 1;
+
+   /* Vertex program always enabled:
+    */
+   vs.vs6.vs_enable = 1;
+
+   brw->vs.state_gs_offset = brw_cache_data( &brw->cache[BRW_VS_UNIT], &vs );
+}
+
+
+const struct brw_tracked_state brw_vs_unit = {
+   .dirty = {
+      .brw   = (BRW_NEW_CLIP |
+		BRW_NEW_CURBE_OFFSETS |
+		BRW_NEW_URB_FENCE),
+      .cache = CACHE_NEW_VS_PROG
+   },
+   .update = upload_vs_unit
+};
diff --git a/src/gallium/drivers/i965simple/brw_winsys.h b/src/gallium/drivers/i965simple/brw_winsys.h
new file mode 100644
index 0000000000..ec1e400418
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_winsys.h
@@ -0,0 +1,209 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \file
+ * This is the interface that i965simple requires any window system
+ * hosting it to implement.  This is the only include file in i965simple
+ * which is public.
+ *
+ */
+
+#ifndef BRW_WINSYS_H
+#define BRW_WINSYS_H
+
+
+#include "pipe/p_defines.h"
+
+
+/* Pipe drivers are (meant to be!) independent of both GL and the
+ * window system.  The window system provides a buffer manager and a
+ * set of additional hooks for things like command buffer submission,
+ * etc.
+ *
+ * There clearly has to be some agreement between the window system
+ * driver and the hardware driver about the format of command buffers,
+ * etc.
+ */
+
+struct pipe_buffer;
+struct pipe_fence_handle;
+struct pipe_winsys;
+struct pipe_screen;
+
+
+/* The pipe driver currently understands the following chipsets:
+ */
+#define PCI_CHIP_I965_G			0x29A2
+#define PCI_CHIP_I965_Q			0x2992
+#define PCI_CHIP_I965_G_1		0x2982
+#define PCI_CHIP_I965_GM                0x2A02
+#define PCI_CHIP_I965_GME               0x2A12
+
+
+/* These are the names of all the state caches managed by the driver.
+ * 
+ * When data is uploaded to a buffer with buffer_subdata, we use the
+ * special version of that function below so that information about
+ * what type of data this is can be passed to the winsys backend.
+ * That in turn allows the correct flags to be set in the aub file
+ * dump to allow human-readable file dumps later on.
+ */
+
+enum brw_cache_id {
+   BRW_CC_VP,
+   BRW_CC_UNIT,
+   BRW_WM_PROG,
+   BRW_SAMPLER_DEFAULT_COLOR,
+   BRW_SAMPLER,
+   BRW_WM_UNIT,
+   BRW_SF_PROG,
+   BRW_SF_VP,
+   BRW_SF_UNIT,
+   BRW_VS_UNIT,
+   BRW_VS_PROG,
+   BRW_GS_UNIT,
+   BRW_GS_PROG,
+   BRW_CLIP_VP,
+   BRW_CLIP_UNIT,
+   BRW_CLIP_PROG,
+   BRW_SS_SURFACE,
+   BRW_SS_SURF_BIND,
+
+   BRW_MAX_CACHE
+};
+
+#define BRW_CONSTANT_BUFFER BRW_MAX_CACHE
+
+/**
+ * Additional winsys interface for i965simple.
+ *
+ * It is an over-simple batchbuffer mechanism.  Will want to improve the
+ * performance of this, perhaps based on the cmdstream stuff.  It
+ * would be pretty impossible to implement swz on top of this
+ * interface.
+ *
+ * Will also need additions/changes to implement static/dynamic
+ * indirect state.
+ */
+struct brw_winsys {
+
+   void (*destroy)(struct brw_winsys *);
+   
+   /**
+    * Reserve space on batch buffer.
+    *
+    * Returns a null pointer if there is insufficient space in the batch buffer
+    * to hold the requested number of dwords and relocations.
+    *
+    * The number of dwords should also include the number of relocations.
+    */
+   unsigned *(*batch_start)(struct brw_winsys *sws,
+                            unsigned dwords,
+                            unsigned relocs);
+
+   void (*batch_dword)(struct brw_winsys *sws,
+                       unsigned dword);
+
+   /**
+    * Emit a relocation to a buffer.
+    *
+    * Used not only when the buffer addresses are not pinned, but also to
+    * ensure refered buffers will not be destroyed until the current batch
+    * buffer execution is finished.
+    *
+    * The access flags is a combination of I915_BUFFER_ACCESS_WRITE and
+    * I915_BUFFER_ACCESS_READ macros.
+    */
+   void (*batch_reloc)(struct brw_winsys *sws,
+                       struct pipe_buffer *buf,
+                       unsigned access_flags,
+                       unsigned delta);
+
+
+   /* Not used yet, but really want this:
+    */
+   void (*batch_end)( struct brw_winsys *sws );
+
+   /**
+    * Flush the batch buffer.
+    *
+    * Fence argument must point to NULL or to a previous fence, and the caller
+    * must call fence_reference when done with the fence.
+    */
+   void (*batch_flush)(struct brw_winsys *sws,
+                       struct pipe_fence_handle **fence);
+
+
+   /* A version of buffer_subdata that includes information for the
+    * simulator:
+    */
+   void (*buffer_subdata_typed)(struct brw_winsys *sws, 
+				struct pipe_buffer *buf,
+				unsigned long offset, 
+				unsigned long size, 
+				const void *data,
+				unsigned data_type);
+   
+
+   /* A cheat so we don't have to think about relocations in a couple
+    * of places yet:
+    */
+   unsigned (*get_buffer_offset)( struct brw_winsys *sws,
+				  struct pipe_buffer *buf,
+				  unsigned flags );
+
+};
+
+#define BRW_BUFFER_ACCESS_WRITE   0x1
+#define BRW_BUFFER_ACCESS_READ    0x2
+
+#define BRW_BUFFER_USAGE_LIT_VERTEX  (PIPE_BUFFER_USAGE_CUSTOM << 0)
+
+
+struct pipe_context *brw_create(struct pipe_screen *,
+                                struct brw_winsys *,
+                                unsigned pci_id);
+
+static inline boolean brw_batchbuffer_data(struct brw_winsys *winsys,
+                                           const void *data,
+                                           unsigned bytes)
+{
+   static const unsigned incr = sizeof(unsigned);
+   uint i;
+   const unsigned *udata = (const unsigned*)(data);
+   unsigned size = bytes/incr;
+
+   winsys->batch_start(winsys, size, 0);
+   for (i = 0; i < size; ++i) {
+      winsys->batch_dword(winsys, udata[i]);
+   }
+   winsys->batch_end(winsys);
+
+   return (i == size);
+}
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_wm.c b/src/gallium/drivers/i965simple/brw_wm.c
new file mode 100644
index 0000000000..10161f2d2f
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_wm.c
@@ -0,0 +1,209 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_context.h"
+#include "brw_util.h"
+#include "brw_wm.h"
+#include "brw_eu.h"
+#include "brw_state.h"
+#include "util/u_memory.h"
+
+
+
+static void do_wm_prog( struct brw_context *brw,
+			struct brw_fragment_program *fp,
+			struct brw_wm_prog_key *key)
+{
+   struct brw_wm_compile *c = CALLOC_STRUCT(brw_wm_compile);
+   const unsigned *program;
+   unsigned program_size;
+
+   c->key = *key;
+   c->fp = fp;
+   
+   c->delta_xy[0] = brw_null_reg();
+   c->delta_xy[1] = brw_null_reg();
+   c->pixel_xy[0] = brw_null_reg();
+   c->pixel_xy[1] = brw_null_reg();
+   c->pixel_w = brw_null_reg();
+
+
+   debug_printf("XXXXXXXX FP\n");
+   
+   brw_wm_glsl_emit(c);
+
+   /* get the program
+    */
+   program = brw_get_program(&c->func, &program_size);
+
+   /*
+    */
+   brw->wm.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_WM_PROG],
+					      &c->key,
+					      sizeof(c->key),
+					      program,
+					      program_size,
+					      &c->prog_data,
+					      &brw->wm.prog_data );
+
+   FREE(c);
+}
+
+
+
+static void brw_wm_populate_key( struct brw_context *brw,
+				 struct brw_wm_prog_key *key )
+{
+   /* BRW_NEW_FRAGMENT_PROGRAM */
+   struct brw_fragment_program *fp =
+      (struct brw_fragment_program *)brw->attribs.FragmentProgram;
+   unsigned lookup = 0;
+   unsigned line_aa;
+   
+   memset(key, 0, sizeof(*key));
+
+   /* Build the index for table lookup
+    */
+   /* BRW_NEW_DEPTH_STENCIL */
+   if (fp->info.uses_kill ||
+       brw->attribs.DepthStencil->alpha.enabled)
+      lookup |= IZ_PS_KILL_ALPHATEST_BIT;
+
+   if (fp->info.writes_z)
+      lookup |= IZ_PS_COMPUTES_DEPTH_BIT;
+
+   if (brw->attribs.DepthStencil->depth.enabled)
+      lookup |= IZ_DEPTH_TEST_ENABLE_BIT;
+
+   if (brw->attribs.DepthStencil->depth.enabled &&
+       brw->attribs.DepthStencil->depth.writemask) /* ?? */
+      lookup |= IZ_DEPTH_WRITE_ENABLE_BIT;
+
+   if (brw->attribs.DepthStencil->stencil[0].enabled) {
+      lookup |= IZ_STENCIL_TEST_ENABLE_BIT;
+
+      if (brw->attribs.DepthStencil->stencil[0].writemask ||
+	  brw->attribs.DepthStencil->stencil[1].writemask)
+	 lookup |= IZ_STENCIL_WRITE_ENABLE_BIT;
+   }
+
+   /* XXX: when should this be disabled?
+    */
+   if (1)
+      lookup |= IZ_EARLY_DEPTH_TEST_BIT;
+
+
+   line_aa = AA_NEVER;
+
+   /* _NEW_LINE, _NEW_POLYGON, BRW_NEW_REDUCED_PRIMITIVE */
+   if (brw->attribs.Raster->line_smooth) {
+      if (brw->reduced_primitive == PIPE_PRIM_LINES) {
+	 line_aa = AA_ALWAYS;
+      }
+      else if (brw->reduced_primitive == PIPE_PRIM_TRIANGLES) {
+	 if (brw->attribs.Raster->fill_ccw == PIPE_POLYGON_MODE_LINE) {
+	    line_aa = AA_SOMETIMES;
+
+	    if (brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_LINE ||
+		(brw->attribs.Raster->cull_mode == PIPE_WINDING_CW))
+	       line_aa = AA_ALWAYS;
+	 }
+	 else if (brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_LINE) {
+	    line_aa = AA_SOMETIMES;
+
+	    if (brw->attribs.Raster->cull_mode == PIPE_WINDING_CCW)
+	       line_aa = AA_ALWAYS;
+	 }
+      }
+   }
+
+   brw_wm_lookup_iz(line_aa,
+		    lookup,
+		    key);
+
+
+#if 0
+   /* BRW_NEW_SAMPLER 
+    *
+    * Not doing any of this at the moment:
+    */
+   for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
+      const struct pipe_sampler_state *unit = brw->attribs.Samplers[i];
+
+      if (unit) {
+
+	 if (unit->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+	    key->shadowtex_mask |= 1<<i;
+	 }
+	 if (t->Image[0][t->BaseLevel]->InternalFormat == GL_YCBCR_MESA)
+	    key->yuvtex_mask |= 1<<i;
+      }
+   }
+#endif
+
+
+   /* Extra info:
+    */
+   key->program_string_id = fp->id;
+
+}
+
+
+static void brw_upload_wm_prog( struct brw_context *brw )
+{
+   struct brw_wm_prog_key key;
+   struct brw_fragment_program *fp = (struct brw_fragment_program *)
+      brw->attribs.FragmentProgram;
+
+   brw_wm_populate_key(brw, &key);
+
+   /* Make an early check for the key.
+    */
+   if (brw_search_cache(&brw->cache[BRW_WM_PROG],
+			&key, sizeof(key),
+			&brw->wm.prog_data,
+			&brw->wm.prog_gs_offset))
+      return;
+
+   do_wm_prog(brw, fp, &key);
+}
+
+
+const struct brw_tracked_state brw_wm_prog = {
+   .dirty = {
+      .brw   = (BRW_NEW_FS |
+		BRW_NEW_REDUCED_PRIMITIVE),
+      .cache = 0
+   },
+   .update = brw_upload_wm_prog
+};
+
diff --git a/src/gallium/drivers/i965simple/brw_wm.h b/src/gallium/drivers/i965simple/brw_wm.h
new file mode 100644
index 0000000000..b29c4393f0
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_wm.h
@@ -0,0 +1,142 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+              
+
+#ifndef BRW_WM_H
+#define BRW_WM_H
+
+
+#include "brw_context.h"
+#include "brw_eu.h"
+
+/* A big lookup table is used to figure out which and how many
+ * additional regs will inserted before the main payload in the WM
+ * program execution.  These mainly relate to depth and stencil
+ * processing and the early-depth-test optimization.
+ */
+#define IZ_PS_KILL_ALPHATEST_BIT    0x1
+#define IZ_PS_COMPUTES_DEPTH_BIT    0x2
+#define IZ_DEPTH_WRITE_ENABLE_BIT   0x4
+#define IZ_DEPTH_TEST_ENABLE_BIT    0x8
+#define IZ_STENCIL_WRITE_ENABLE_BIT 0x10
+#define IZ_STENCIL_TEST_ENABLE_BIT  0x20
+#define IZ_EARLY_DEPTH_TEST_BIT     0x40
+#define IZ_BIT_MAX                  0x80
+
+#define AA_NEVER     0
+#define AA_SOMETIMES 1
+#define AA_ALWAYS    2
+
+struct brw_wm_prog_key {
+   unsigned source_depth_reg:3;
+   unsigned aa_dest_stencil_reg:3;
+   unsigned dest_depth_reg:3;
+   unsigned nr_depth_regs:3;
+   unsigned shadowtex_mask:8;
+   unsigned computes_depth:1;	/* could be derived from program string */
+   unsigned source_depth_to_render_target:1;
+   unsigned runtime_check_aads_emit:1;
+
+   unsigned yuvtex_mask:8;
+
+   unsigned program_string_id;
+};
+
+
+
+
+
+#define PROGRAM_INTERNAL_PARAM
+#define MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS 1024 /* 72 for GL_ARB_f_p */
+#define BRW_WM_MAX_INSN  (MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS*3 + PIPE_MAX_ATTRIBS + 3)
+#define BRW_WM_MAX_GRF   128		/* hardware limit */
+#define BRW_WM_MAX_VREG  (BRW_WM_MAX_INSN * 4)
+#define BRW_WM_MAX_REF   (BRW_WM_MAX_INSN * 12)
+#define BRW_WM_MAX_PARAM 256
+#define BRW_WM_MAX_CONST 256
+#define BRW_WM_MAX_KILLS MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS
+
+#define PAYLOAD_DEPTH     (PIPE_MAX_ATTRIBS)
+
+#define MAX_IFSN 32
+#define MAX_LOOP_DEPTH 32
+
+struct brw_wm_compile {
+   struct brw_compile func;
+   struct brw_wm_prog_key key;
+   struct brw_wm_prog_data prog_data; /* result */
+
+   struct brw_fragment_program *fp;
+
+   unsigned grf_limit;
+   unsigned max_wm_grf;
+
+
+   struct brw_reg pixel_xy[2];
+   struct brw_reg delta_xy[2];
+   struct brw_reg pixel_w;
+
+
+   struct brw_reg wm_regs[8][32][4];
+
+   struct brw_reg payload_depth[4];
+   struct brw_reg payload_coef[16];
+
+   struct brw_reg emit_mask_reg;
+
+   struct brw_instruction *if_inst[MAX_IFSN];
+   int if_insn;
+
+   struct brw_instruction *loop_inst[MAX_LOOP_DEPTH];
+   int loop_insn;
+
+   struct brw_instruction *inst0;
+   struct brw_instruction *inst1;
+
+   struct brw_reg stack;
+   struct brw_indirect stack_index;
+
+   unsigned reg_index;
+
+   unsigned tmp_start;
+   unsigned tmp_index;
+};
+
+
+
+void brw_wm_lookup_iz( unsigned line_aa,
+		       unsigned lookup,
+		       struct brw_wm_prog_key *key );
+
+void brw_wm_glsl_emit(struct brw_wm_compile *c);
+void brw_wm_emit_decls(struct brw_wm_compile *c);
+
+#endif
diff --git a/src/gallium/drivers/i965simple/brw_wm_decl.c b/src/gallium/drivers/i965simple/brw_wm_decl.c
new file mode 100644
index 0000000000..d50e66f613
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_wm_decl.c
@@ -0,0 +1,392 @@
+
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_wm.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+
+static struct brw_reg alloc_tmp(struct brw_wm_compile *c)
+{
+   c->tmp_index++;
+   c->reg_index = MAX2(c->reg_index, c->tmp_start + c->tmp_index);
+   return brw_vec8_grf(c->tmp_start + c->tmp_index, 0);
+}
+
+static void release_tmps(struct brw_wm_compile *c)
+{
+   c->tmp_index = 0;
+}
+
+
+
+static int is_null( struct brw_reg reg )
+{
+   return (reg.file == BRW_ARCHITECTURE_REGISTER_FILE &&
+	   reg.nr == BRW_ARF_NULL);
+}
+
+static void emit_pixel_xy( struct brw_wm_compile *c )
+{
+   if (is_null(c->pixel_xy[0])) {
+
+      struct brw_compile *p = &c->func;
+      struct brw_reg r1_uw = retype(brw_vec1_grf(1, 0), BRW_REGISTER_TYPE_UW);
+
+      c->pixel_xy[0] = vec8(retype(alloc_tmp(c), BRW_REGISTER_TYPE_UW));
+      c->pixel_xy[1] = vec8(retype(alloc_tmp(c), BRW_REGISTER_TYPE_UW));
+
+      /* Calculate pixel centers by adding 1 or 0 to each of the
+       * micro-tile coordinates passed in r1.
+       */
+      brw_ADD(p,
+	      c->pixel_xy[0],
+	      stride(suboffset(r1_uw, 4), 2, 4, 0),
+	      brw_imm_v(0x10101010));
+
+      brw_ADD(p,
+	      c->pixel_xy[1],
+	      stride(suboffset(r1_uw, 5), 2, 4, 0),
+	      brw_imm_v(0x11001100));
+   }
+}
+
+
+
+
+
+
+static void emit_delta_xy( struct brw_wm_compile *c )
+{
+   if (is_null(c->delta_xy[0])) {
+      struct brw_compile *p = &c->func;
+      struct brw_reg r1 = brw_vec1_grf(1, 0);
+
+      emit_pixel_xy(c);
+
+      c->delta_xy[0] = alloc_tmp(c);
+      c->delta_xy[1] = alloc_tmp(c);
+
+      /* Calc delta X,Y by subtracting origin in r1 from the pixel
+       * centers.
+       */
+      brw_ADD(p,
+	      c->delta_xy[0],
+	      retype(c->pixel_xy[0], BRW_REGISTER_TYPE_UW),
+	      negate(r1));
+
+      brw_ADD(p,
+	      c->delta_xy[1],
+	      retype(c->pixel_xy[1], BRW_REGISTER_TYPE_UW),
+	      negate(suboffset(r1,1)));
+   }
+}
+
+
+
+#if 0
+static void emit_pixel_w( struct brw_wm_compile *c )
+{
+   if (is_null(c->pixel_w)) {
+      struct brw_compile *p = &c->func;
+
+      struct brw_reg interp_wpos = c->coef_wpos;
+      
+      c->pixel_w = alloc_tmp(c);
+
+      emit_delta_xy(c);
+
+      /* Calc 1/w - just linterp wpos[3] optimized by putting the
+       * result straight into a message reg.
+       */
+      struct brw_reg interp3 = brw_vec1_grf(interp_wpos.nr+1, 4);
+      brw_LINE(p, brw_null_reg(), interp3, c->delta_xy[0]);
+      brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), c->delta_xy[1]);
+
+      /* Calc w */
+      brw_math_16( p, 
+		   c->pixel_w,
+		   BRW_MATH_FUNCTION_INV,
+		   BRW_MATH_SATURATE_NONE,
+		   2, 
+		   brw_null_reg(),
+		   BRW_MATH_PRECISION_FULL);
+   }
+}
+#endif
+
+
+static void emit_cinterp(struct brw_wm_compile *c,
+			 int idx,
+			 int mask )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg interp[4];
+   struct brw_reg coef = c->payload_coef[idx];
+   int i;
+
+   interp[0] = brw_vec1_grf(coef.nr, 0);
+   interp[1] = brw_vec1_grf(coef.nr, 4);
+   interp[2] = brw_vec1_grf(coef.nr+1, 0);
+   interp[3] = brw_vec1_grf(coef.nr+1, 4);
+
+   for(i = 0; i < 4; i++ ) {
+      if (mask & (1<<i)) {
+	 struct brw_reg dst = c->wm_regs[TGSI_FILE_INPUT][idx][i];
+	 brw_MOV(p, dst, suboffset(interp[i],3));
+      }
+   }
+}
+
+static void emit_linterp(struct brw_wm_compile *c,
+			 int idx,
+			 int mask )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg interp[4];
+   struct brw_reg coef = c->payload_coef[idx];
+   int i;
+
+   emit_delta_xy(c);
+
+   interp[0] = brw_vec1_grf(coef.nr, 0);
+   interp[1] = brw_vec1_grf(coef.nr, 4);
+   interp[2] = brw_vec1_grf(coef.nr+1, 0);
+   interp[3] = brw_vec1_grf(coef.nr+1, 4);
+
+   for(i = 0; i < 4; i++ ) {
+      if (mask & (1<<i)) {
+	 struct brw_reg dst = c->wm_regs[TGSI_FILE_INPUT][idx][i];
+	 brw_LINE(p, brw_null_reg(), interp[i], c->delta_xy[0]);
+	 brw_MAC(p, dst, suboffset(interp[i],1), c->delta_xy[1]);
+      }
+   }
+}
+
+#if 0
+static void emit_pinterp(struct brw_wm_compile *c,
+			 int idx,
+			 int mask )
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg interp[4];
+   struct brw_reg coef = c->payload_coef[idx];
+   int i;
+
+   get_delta_xy(c);
+   get_pixel_w(c);
+
+   interp[0] = brw_vec1_grf(coef.nr, 0);
+   interp[1] = brw_vec1_grf(coef.nr, 4);
+   interp[2] = brw_vec1_grf(coef.nr+1, 0);
+   interp[3] = brw_vec1_grf(coef.nr+1, 4);
+
+   for(i = 0; i < 4; i++ ) {
+      if (mask & (1<<i)) {
+	 struct brw_reg dst = allocate_reg(c, TGSI_FILE_INPUT, idx, i);
+	 brw_LINE(p, brw_null_reg(), interp[i], c->delta_xy[0]);
+	 brw_MAC(p, dst, suboffset(interp[i],1), c->delta_xy[1]);
+	 brw_MUL(p, dst, dst, c->pixel_w);
+      }
+   }
+}
+#endif
+
+
+
+#if 0
+static void emit_wpos( )
+{ 
+   struct prog_dst_register dst = dst_reg(PROGRAM_INPUT, idx);
+   struct tgsi_full_src_register interp = src_reg(PROGRAM_PAYLOAD, idx);
+   struct tgsi_full_src_register deltas = get_delta_xy(c);
+   struct tgsi_full_src_register arg2;
+   unsigned opcode;
+
+   opcode = WM_LINTERP;
+   arg2 = src_undef();
+
+   /* Have to treat wpos.xy specially:
+    */
+   emit_op(c,
+	   WM_WPOSXY,
+	   dst_mask(dst, WRITEMASK_XY),
+	   0, 0, 0,
+	   get_pixel_xy(c),
+	   src_undef(),
+	   src_undef());
+      
+   dst = dst_mask(dst, WRITEMASK_ZW);
+
+   /* PROGRAM_INPUT.attr.xyzw = INTERP payload.interp[attr].x, deltas.xyw
+    */
+   emit_op(c,
+	   WM_LINTERP,
+	   dst,
+	   0, 0, 0,
+	   interp,
+	   deltas,
+	   arg2);
+}
+#endif
+
+
+
+
+/* Perform register allocation:
+ * 
+ *  -- r0???
+ *  -- passthrough depth regs (and stencil/aa??)
+ *  -- curbe ??
+ *  -- inputs (coefficients)
+ *
+ * Use a totally static register allocation.  This will perform poorly
+ * but is an easy way to get started (again).
+ */
+static void prealloc_reg(struct brw_wm_compile *c)
+{
+   int i, j;
+   int nr_curbe_regs = 0;
+
+   /* R0, then some depth related regs:
+    */
+   for (i = 0; i < c->key.nr_depth_regs; i++) {
+      c->payload_depth[i] =  brw_vec8_grf(i*2, 0);
+      c->reg_index += 2;
+   }
+
+
+   /* Then a copy of our part of the CURBE entry:
+    */
+   {
+      int nr_constants = c->fp->info.file_max[TGSI_FILE_CONSTANT] + 1;
+      int index = 0;
+
+      /* XXX number of constants, or highest numbered constant? */
+      assert(nr_constants == c->fp->info.file_count[TGSI_FILE_CONSTANT]);
+
+      c->prog_data.max_const = 4*nr_constants;
+      for (i = 0; i < nr_constants; i++) {
+	 for (j = 0; j < 4; j++, index++) 
+	    c->wm_regs[TGSI_FILE_CONSTANT][i][j] = brw_vec1_grf(c->reg_index + index/8,
+								index%8);
+      }
+
+      nr_curbe_regs = 2*((4*nr_constants+15)/16);
+      c->reg_index += nr_curbe_regs;
+   }
+
+   /* Adjust for parameter coefficients for position, which are
+    * currently always provided.
+    */
+//   c->position_coef[i] = brw_vec8_grf(c->reg_index, 0);
+   c->reg_index += 2;
+
+   /* Next we receive the plane coefficients for parameter
+    * interpolation:
+    */
+   assert(c->fp->info.file_max[TGSI_FILE_INPUT] == c->fp->info.num_inputs);
+   for (i = 0; i < c->fp->info.file_max[TGSI_FILE_INPUT] + 1; i++) {
+      c->payload_coef[i] = brw_vec8_grf(c->reg_index, 0);
+      c->reg_index += 2;
+   }
+
+   c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2;
+   c->prog_data.urb_read_length = (c->fp->info.num_inputs + 1) * 2;
+   c->prog_data.curb_read_length = nr_curbe_regs;
+
+   /* That's the end of the payload, now we can start allocating registers.
+    */
+   c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0);
+   c->reg_index++;
+
+   c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0);
+   c->reg_index += 2;
+
+   /* Now allocate room for the interpolated inputs and staging
+    * registers for the outputs:
+    */
+   /* XXX do we want to loop over the _number_ of inputs/outputs or loop
+    * to the highest input/output index that's used?
+    *  Probably the same, actually.
+    */
+   assert(c->fp->info.file_max[TGSI_FILE_INPUT] + 1 == c->fp->info.num_inputs);
+   assert(c->fp->info.file_max[TGSI_FILE_OUTPUT] + 1 == c->fp->info.num_outputs);
+   for (i = 0; i < c->fp->info.file_max[TGSI_FILE_INPUT] + 1; i++) 
+      for (j = 0; j < 4; j++)
+	 c->wm_regs[TGSI_FILE_INPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 );
+
+   for (i = 0; i < c->fp->info.file_max[TGSI_FILE_OUTPUT] + 1; i++) 
+      for (j = 0; j < 4; j++)
+	 c->wm_regs[TGSI_FILE_OUTPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 );
+
+   /* Beyond this we should only need registers for internal temporaries:
+    */
+   c->tmp_start = c->reg_index;
+}
+
+
+
+
+
+/* Need to interpolate fragment program inputs in as a preamble to the
+ * shader.  A more sophisticated compiler would do this on demand, but
+ * we'll do it up front:
+ */
+void brw_wm_emit_decls(struct brw_wm_compile *c)
+{
+   struct tgsi_parse_context parse;
+   int done = 0;
+
+   prealloc_reg(c);
+
+   tgsi_parse_init( &parse, c->fp->program.tokens );
+
+   while( !done &&
+	  !tgsi_parse_end_of_tokens( &parse ) ) 
+   {
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+      {
+	 const struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration;
+	 unsigned first = decl->DeclarationRange.First;
+	 unsigned last = decl->DeclarationRange.Last;
+	 unsigned mask = decl->Declaration.UsageMask; /* ? */
+	 unsigned i;
+
+	 if (decl->Declaration.File != TGSI_FILE_INPUT)
+	    break;
+
+	 for( i = first; i <= last; i++ ) {
+	    switch (decl->Declaration.Interpolate) {
+	    case TGSI_INTERPOLATE_CONSTANT:
+	       emit_cinterp(c, i, mask);
+	       break;
+
+	    case TGSI_INTERPOLATE_LINEAR:
+	       emit_linterp(c, i, mask);
+	       break;
+
+	    case TGSI_INTERPOLATE_PERSPECTIVE:
+	       //emit_pinterp(c, i, mask);
+	       emit_linterp(c, i, mask);
+	       break;
+	    }
+	 }
+	 break;
+      }
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+      default:
+         done = 1;
+	 break;
+      }
+   }
+
+   tgsi_parse_free (&parse);
+   
+   release_tmps(c);
+}
diff --git a/src/gallium/drivers/i965simple/brw_wm_glsl.c b/src/gallium/drivers/i965simple/brw_wm_glsl.c
new file mode 100644
index 0000000000..ab6410aa60
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_wm_glsl.c
@@ -0,0 +1,1076 @@
+
+#include "brw_context.h"
+#include "brw_eu.h"
+#include "brw_wm.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+
+
+
+static int get_scalar_dst_index(struct tgsi_full_instruction *inst)
+{
+   struct tgsi_dst_register dst = inst->FullDstRegisters[0].DstRegister;
+   int i;
+   for (i = 0; i < 4; i++)
+      if (dst.WriteMask & (1<<i))
+	 break;
+   return i;
+}
+
+static struct brw_reg alloc_tmp(struct brw_wm_compile *c)
+{
+   c->tmp_index++;
+   c->reg_index = MAX2(c->reg_index, c->tmp_index);
+   return brw_vec8_grf(c->tmp_start + c->tmp_index, 0);
+}
+
+static void release_tmps(struct brw_wm_compile *c)
+{
+   c->tmp_index = 0;
+}
+
+
+static struct brw_reg
+get_reg(struct brw_wm_compile *c, int file, int index, int component )
+{
+   switch (file) {
+   case TGSI_FILE_NULL:
+      return brw_null_reg();
+
+   case TGSI_FILE_SAMPLER:
+      /* Should never get here:
+       */
+      assert (0);	       
+      return brw_null_reg();
+
+   case TGSI_FILE_IMMEDIATE:
+      /* These need a different path:
+       */
+      assert(0);
+      return brw_null_reg();
+
+       
+   case TGSI_FILE_CONSTANT:
+   case TGSI_FILE_INPUT:
+   case TGSI_FILE_OUTPUT:
+   case TGSI_FILE_TEMPORARY:
+   case TGSI_FILE_ADDRESS:
+      return c->wm_regs[file][index][component];
+
+   default:
+      assert(0);
+      return brw_null_reg();
+   }
+}
+
+
+static struct brw_reg get_dst_reg(struct brw_wm_compile *c,
+				  struct tgsi_full_instruction *inst, 
+				  int component)
+{
+   return get_reg(c, 
+		  inst->FullDstRegisters[0].DstRegister.File, 
+		  inst->FullDstRegisters[0].DstRegister.Index,
+		  component);
+}
+
+static int get_swz( struct tgsi_src_register src, int index )
+{
+   switch (index & 3) {
+   case 0: return src.SwizzleX;
+   case 1: return src.SwizzleY;
+   case 2: return src.SwizzleZ;
+   case 3: return src.SwizzleW;
+   default: return 0;
+   }
+}
+
+static int get_ext_swz( struct tgsi_src_register_ext_swz src, int index )
+{
+   switch (index & 3) {
+   case 0: return src.ExtSwizzleX;
+   case 1: return src.ExtSwizzleY;
+   case 2: return src.ExtSwizzleZ;
+   case 3: return src.ExtSwizzleW;
+   default: return 0;
+   }
+}
+
+static struct brw_reg get_src_reg(struct brw_wm_compile *c,
+				  struct tgsi_full_src_register *src, 
+				  int index)
+{
+   struct brw_reg reg;
+   int component = index;
+   int neg = 0;
+   int abs = 0;
+
+   if (src->SrcRegister.Negate)
+      neg = 1;
+
+   component = get_swz(src->SrcRegister, component);
+
+   /* Yes, there are multiple negates:
+    */
+   switch (component & 3) {
+   case 0: neg ^= src->SrcRegisterExtSwz.NegateX; break;
+   case 1: neg ^= src->SrcRegisterExtSwz.NegateY; break;
+   case 2: neg ^= src->SrcRegisterExtSwz.NegateZ; break;
+   case 3: neg ^= src->SrcRegisterExtSwz.NegateW; break;
+   }
+
+   /* And multiple swizzles, fun isn't it:
+    */
+   component = get_ext_swz(src->SrcRegisterExtSwz, component);
+
+   /* Not handling indirect lookups yet:
+    */
+   assert(src->SrcRegister.Indirect == 0);
+
+   /* Don't know what dimension means:
+    */
+   assert(src->SrcRegister.Dimension == 0);
+
+   /* Will never handle any of this stuff: 
+    */
+   assert(src->SrcRegisterExtMod.Complement == 0);
+   assert(src->SrcRegisterExtMod.Bias == 0);
+   assert(src->SrcRegisterExtMod.Scale2X == 0);
+
+   if (src->SrcRegisterExtMod.Absolute)
+      abs = 1;
+
+   /* Another negate!  This is a post-absolute negate, which we
+    * can't do.  Need to clean the crap out of tgsi somehow.
+    */
+   assert(src->SrcRegisterExtMod.Negate == 0);
+
+   switch( component ) {
+   case TGSI_EXTSWIZZLE_X:
+   case TGSI_EXTSWIZZLE_Y:
+   case TGSI_EXTSWIZZLE_Z:
+   case TGSI_EXTSWIZZLE_W:
+      reg = get_reg(c, 
+		    src->SrcRegister.File, 
+		    src->SrcRegister.Index, 
+		    component );
+
+      if (neg) 
+	 reg = negate(reg);
+   
+      if (abs)
+	 reg = brw_abs(reg);
+
+      break;
+
+      /* XXX: this won't really work in the general case, but we know
+       * that the extended swizzle is only allowed in the SWZ
+       * instruction (right??), in which case using an immediate
+       * directly will work.
+       */
+   case TGSI_EXTSWIZZLE_ZERO:
+      reg = brw_imm_f(0);
+      break;
+
+   case TGSI_EXTSWIZZLE_ONE:
+      if (neg && !abs)
+	 reg = brw_imm_f(-1.0);
+      else
+	 reg = brw_imm_f(1.0);
+      break;
+
+   default:
+      assert(0);
+      break;
+   }
+
+    
+   return reg;
+}
+
+static void emit_abs( struct brw_wm_compile *c,
+		      struct tgsi_full_instruction *inst)
+{
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+
+   int i;
+   struct brw_compile *p = &c->func;
+   brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE);
+   for (i = 0; i < 4; i++) {
+      if (mask & (1<<i)) {
+	 struct brw_reg src, dst;
+	 dst = get_dst_reg(c, inst, i);
+	 src = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+	 brw_MOV(p, dst, brw_abs(src)); /* NOTE */
+      }
+   }
+   brw_set_saturate(p, 0);
+}
+
+
+static void emit_xpd(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   int i;
+   struct brw_compile *p = &c->func;
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   for (i = 0; i < 4; i++) {
+      unsigned i2 = (i+2)%3;
+      unsigned i1 = (i+1)%3;
+      if (mask & (1<<i)) {
+	 struct brw_reg src0, src1, dst;
+	 dst = get_dst_reg(c, inst, i);
+	 src0 = negate(get_src_reg(c, &inst->FullSrcRegisters[0], i2));
+	 src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i1);
+	 brw_MUL(p, brw_null_reg(), src0, src1);
+	 src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i1);
+	 src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i2);
+	 brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE);
+	 brw_MAC(p, dst, src0, src1);
+	 brw_set_saturate(p, 0);
+      }
+   }
+   brw_set_saturate(p, 0);
+}
+
+static void emit_dp3(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   struct brw_reg src0[3], src1[3], dst;
+   int i;
+   struct brw_compile *p = &c->func;
+   for (i = 0; i < 3; i++) {
+      src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+      src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i);
+   }
+
+   dst = get_dst_reg(c, inst, get_scalar_dst_index(inst));
+   brw_MUL(p, brw_null_reg(), src0[0], src1[0]);
+   brw_MAC(p, brw_null_reg(), src0[1], src1[1]);
+   brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+   brw_MAC(p, dst, src0[2], src1[2]);
+   brw_set_saturate(p, 0);
+}
+
+static void emit_dp4(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   struct brw_reg src0[4], src1[4], dst;
+   int i;
+   struct brw_compile *p = &c->func;
+   for (i = 0; i < 4; i++) {
+      src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+      src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i);
+   }
+   dst = get_dst_reg(c, inst, get_scalar_dst_index(inst));
+   brw_MUL(p, brw_null_reg(), src0[0], src1[0]);
+   brw_MAC(p, brw_null_reg(), src0[1], src1[1]);
+   brw_MAC(p, brw_null_reg(), src0[2], src1[2]);
+   brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+   brw_MAC(p, dst, src0[3], src1[3]);
+   brw_set_saturate(p, 0);
+}
+
+static void emit_dph(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   struct brw_reg src0[4], src1[4], dst;
+   int i;
+   struct brw_compile *p = &c->func;
+   for (i = 0; i < 4; i++) {
+      src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+      src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i);
+   }
+   dst = get_dst_reg(c, inst, get_scalar_dst_index(inst));
+   brw_MUL(p, brw_null_reg(), src0[0], src1[0]);
+   brw_MAC(p, brw_null_reg(), src0[1], src1[1]);
+   brw_MAC(p, dst, src0[2], src1[2]);
+   brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+   brw_ADD(p, dst, src0[3], src1[3]);
+   brw_set_saturate(p, 0);
+}
+
+static void emit_math1(struct brw_wm_compile *c,
+		       struct tgsi_full_instruction *inst, unsigned func)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg src0, dst;
+
+   src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0);
+   dst = get_dst_reg(c, inst, get_scalar_dst_index(inst));
+   brw_MOV(p, brw_message_reg(2), src0);
+   brw_math(p,
+	    dst,
+	    func,
+	    ((inst->Instruction.Saturate != TGSI_SAT_NONE) 
+	     ? BRW_MATH_SATURATE_SATURATE 
+	     : BRW_MATH_SATURATE_NONE),
+	    2,
+	    brw_null_reg(),
+	    BRW_MATH_DATA_VECTOR,
+	    BRW_MATH_PRECISION_FULL);
+}
+
+
+static void emit_alu2(struct brw_wm_compile *c,		      
+		      struct tgsi_full_instruction *inst,
+		      unsigned opcode)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg src0, src1, dst;
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   int i;
+   brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+   for (i = 0 ; i < 4; i++) {
+      if (mask & (1<<i)) {
+	 dst = get_dst_reg(c, inst, i);
+	 src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+	 src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i);
+	 brw_alu2(p, opcode, dst, src0, src1);
+      }
+   }
+   brw_set_saturate(p, 0);
+}
+
+
+static void emit_alu1(struct brw_wm_compile *c,
+		      struct tgsi_full_instruction *inst,
+		      unsigned opcode)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg src0, dst;
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   int i;
+   brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+   for (i = 0 ; i < 4; i++) {
+      if (mask & (1<<i)) {
+	 dst = get_dst_reg(c, inst, i);
+	 src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+	 brw_alu1(p, opcode, dst, src0);
+      }
+   }
+   if (inst->Instruction.Saturate != TGSI_SAT_NONE)
+      brw_set_saturate(p, 0);
+}
+
+
+static void emit_max(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   struct brw_compile *p = &c->func;
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   struct brw_reg src0, src1, dst;
+   int i;
+   brw_push_insn_state(p);
+   for (i = 0; i < 4; i++) {
+      if (mask & (1<<i)) {
+	 dst = get_dst_reg(c, inst, i);
+	 src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+	 src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i);
+	 brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+	 brw_MOV(p, dst, src0);
+	 brw_set_saturate(p, 0);
+
+	 brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src0, src1);
+	 brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+	 brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+	 brw_MOV(p, dst, src1);
+	 brw_set_saturate(p, 0);
+	 brw_set_predicate_control_flag_value(p, 0xff);
+      }
+   }
+   brw_pop_insn_state(p);
+}
+
+static void emit_min(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   struct brw_compile *p = &c->func;
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   struct brw_reg src0, src1, dst;
+   int i;
+   brw_push_insn_state(p);
+   for (i = 0; i < 4; i++) {
+      if (mask & (1<<i)) {
+	 dst = get_dst_reg(c, inst, i);
+	 src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+	 src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i);
+	 brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+	 brw_MOV(p, dst, src0);
+	 brw_set_saturate(p, 0);
+
+	 brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src1, src0);
+	 brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+	 brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+	 brw_MOV(p, dst, src1);
+	 brw_set_saturate(p, 0);
+	 brw_set_predicate_control_flag_value(p, 0xff);
+      }
+   }
+   brw_pop_insn_state(p);
+}
+
+static void emit_pow(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg dst, src0, src1;
+   dst = get_dst_reg(c, inst, get_scalar_dst_index(inst));
+   src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0);
+   src1 = get_src_reg(c, &inst->FullSrcRegisters[1], 0);
+
+   brw_MOV(p, brw_message_reg(2), src0);
+   brw_MOV(p, brw_message_reg(3), src1);
+
+   brw_math(p,
+	    dst,
+	    BRW_MATH_FUNCTION_POW,
+	    (inst->Instruction.Saturate != TGSI_SAT_NONE 
+	     ? BRW_MATH_SATURATE_SATURATE 
+	     : BRW_MATH_SATURATE_NONE),
+	    2,
+	    brw_null_reg(),
+	    BRW_MATH_DATA_VECTOR,
+	    BRW_MATH_PRECISION_FULL);
+}
+
+static void emit_lrp(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   struct brw_compile *p = &c->func;
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   struct brw_reg dst, tmp1, tmp2, src0, src1, src2;
+   int i;
+   for (i = 0; i < 4; i++) {
+      if (mask & (1<<i)) {
+	 dst = get_dst_reg(c, inst, i);
+	 src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+
+	 src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i);
+
+	 if (src1.nr == dst.nr) {
+	    tmp1 = alloc_tmp(c);
+	    brw_MOV(p, tmp1, src1);
+	 } else
+	    tmp1 = src1;
+
+	 src2 = get_src_reg(c, &inst->FullSrcRegisters[2], i);
+	 if (src2.nr == dst.nr) {
+	    tmp2 = alloc_tmp(c);
+	    brw_MOV(p, tmp2, src2);
+	 } else
+	    tmp2 = src2;
+
+	 brw_ADD(p, dst, negate(src0), brw_imm_f(1.0));
+	 brw_MUL(p, brw_null_reg(), dst, tmp2);
+	 brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+	 brw_MAC(p, dst, src0, tmp1);
+	 brw_set_saturate(p, 0);
+      }
+      release_tmps(c);
+   }
+}
+
+static void emit_kil(struct brw_wm_compile *c)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg depth = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW);
+   brw_push_insn_state(p);
+   brw_set_mask_control(p, BRW_MASK_DISABLE);
+   brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); //IMASK
+   brw_AND(p, depth, c->emit_mask_reg, depth);
+   brw_pop_insn_state(p);
+}
+
+static void emit_mad(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   struct brw_compile *p = &c->func;
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   struct brw_reg dst, src0, src1, src2;
+   int i;
+
+   for (i = 0; i < 4; i++) {
+      if (mask & (1<<i)) {
+	 dst = get_dst_reg(c, inst, i);
+	 src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+	 src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i);
+	 src2 = get_src_reg(c, &inst->FullSrcRegisters[2], i);
+	 brw_MUL(p, dst, src0, src1);
+
+	 brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0);
+	 brw_ADD(p, dst, dst, src2);
+	 brw_set_saturate(p, 0);
+      }
+   }
+}
+
+static void emit_sop(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst, unsigned cond)
+{
+   struct brw_compile *p = &c->func;
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   struct brw_reg dst, src0, src1;
+   int i;
+
+   brw_push_insn_state(p);
+   for (i = 0; i < 4; i++) {
+      if (mask & (1<<i)) {
+	 dst = get_dst_reg(c, inst, i);
+	 src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+	 src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i);
+	 brw_CMP(p, brw_null_reg(), cond, src0, src1);
+	 brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+	 brw_MOV(p, dst, brw_imm_f(0.0));
+	 brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+	 brw_MOV(p, dst, brw_imm_f(1.0));
+      }
+   }
+   brw_pop_insn_state(p);
+}
+
+
+static void emit_ddx(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   struct brw_compile *p = &c->func;
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   struct brw_reg interp[4];
+   struct brw_reg dst;
+   struct brw_reg src0, w;
+   unsigned nr, i;
+   src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0);
+   w = get_src_reg(c, &inst->FullSrcRegisters[1], 3);
+   nr = src0.nr;
+   interp[0] = brw_vec1_grf(nr, 0);
+   interp[1] = brw_vec1_grf(nr, 4);
+   interp[2] = brw_vec1_grf(nr+1, 0);
+   interp[3] = brw_vec1_grf(nr+1, 4);
+   brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE);
+   for(i = 0; i < 4; i++ ) {
+      if (mask & (1<<i)) {
+	 dst = get_dst_reg(c, inst, i);
+	 brw_MOV(p, dst, interp[i]);
+	 brw_MUL(p, dst, dst, w);
+      }
+   }
+   brw_set_saturate(p, 0);
+}
+
+static void emit_ddy(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+   struct brw_compile *p = &c->func;
+   unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+   struct brw_reg interp[4];
+   struct brw_reg dst;
+   struct brw_reg src0, w;
+   unsigned nr, i;
+
+   src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0);
+   nr = src0.nr;
+   w = get_src_reg(c, &inst->FullSrcRegisters[1], 3);
+   interp[0] = brw_vec1_grf(nr, 0);
+   interp[1] = brw_vec1_grf(nr, 4);
+   interp[2] = brw_vec1_grf(nr+1, 0);
+   interp[3] = brw_vec1_grf(nr+1, 4);
+   brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE);
+   for(i = 0; i < 4; i++ ) {
+      if (mask & (1<<i)) {
+	 dst = get_dst_reg(c, inst, i);
+	 brw_MOV(p, dst, suboffset(interp[i], 1));
+	 brw_MUL(p, dst, dst, w);
+      }
+   }
+   brw_set_saturate(p, 0);
+}
+
+/* TODO
+   BIAS on SIMD8 not workind yet...
+*/
+static void emit_txb(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+#if 0
+   struct brw_compile *p = &c->func;
+   struct brw_reg payload_reg = c->payload_depth[0];
+   struct brw_reg dst[4], src[4];
+   unsigned i;
+   for (i = 0; i < 4; i++)
+      dst[i] = get_dst_reg(c, inst, i);
+   for (i = 0; i < 4; i++)
+      src[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+
+#if 0
+   switch (inst->TexSrcTarget) {
+   case TEXTURE_1D_INDEX:
+      brw_MOV(p, brw_message_reg(2), src[0]);
+      brw_MOV(p, brw_message_reg(3), brw_imm_f(0));
+      brw_MOV(p, brw_message_reg(4), brw_imm_f(0));
+      break;
+   case TEXTURE_2D_INDEX:
+   case TEXTURE_RECT_INDEX:
+      brw_MOV(p, brw_message_reg(2), src[0]);
+      brw_MOV(p, brw_message_reg(3), src[1]);
+      brw_MOV(p, brw_message_reg(4), brw_imm_f(0));
+      break;
+   default:
+      brw_MOV(p, brw_message_reg(2), src[0]);
+      brw_MOV(p, brw_message_reg(3), src[1]);
+      brw_MOV(p, brw_message_reg(4), src[2]);
+      break;
+   }
+#else
+   brw_MOV(p, brw_message_reg(2), src[0]);
+   brw_MOV(p, brw_message_reg(3), src[1]);
+   brw_MOV(p, brw_message_reg(4), brw_imm_f(0));
+#endif
+
+   brw_MOV(p, brw_message_reg(5), src[3]);
+   brw_MOV(p, brw_message_reg(6), brw_imm_f(0));
+   brw_SAMPLE(p,
+	      retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW),
+	      1,
+	      retype(payload_reg, BRW_REGISTER_TYPE_UW),
+	      inst->TexSrcUnit + 1, /* surface */
+	      inst->TexSrcUnit,     /* sampler */
+	      inst->FullDstRegisters[0].DstRegister.WriteMask,
+	      BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS,
+	      4,
+	      4,
+	      0);
+#endif
+}
+
+static void emit_tex(struct brw_wm_compile *c,
+		     struct tgsi_full_instruction *inst)
+{
+#if 0
+   struct brw_compile *p = &c->func;
+   struct brw_reg payload_reg = c->payload_depth[0];
+   struct brw_reg dst[4], src[4];
+   unsigned msg_len;
+   unsigned i, nr;
+   unsigned emit;
+   boolean shadow = (c->key.shadowtex_mask & (1<<inst->TexSrcUnit)) ? 1 : 0;
+
+   for (i = 0; i < 4; i++)
+      dst[i] = get_dst_reg(c, inst, i);
+   for (i = 0; i < 4; i++)
+      src[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i);
+
+#if 0
+   switch (inst->TexSrcTarget) {
+   case TEXTURE_1D_INDEX:
+      emit = WRITEMASK_X;
+      nr = 1;
+      break;
+   case TEXTURE_2D_INDEX:
+   case TEXTURE_RECT_INDEX:
+      emit = WRITEMASK_XY;
+      nr = 2;
+      break;
+   default:
+      emit = WRITEMASK_XYZ;
+      nr = 3;
+      break;
+   }
+#else
+   emit = WRITEMASK_XY;
+   nr = 2;
+#endif
+
+   msg_len = 1;
+
+   for (i = 0; i < nr; i++) {
+      static const unsigned swz[4] = {0,1,2,2};
+      if (emit & (1<<i))
+	 brw_MOV(p, brw_message_reg(msg_len+1), src[swz[i]]);
+      else
+	 brw_MOV(p, brw_message_reg(msg_len+1), brw_imm_f(0));
+      msg_len += 1;
+   }
+
+   if (shadow) {
+      brw_MOV(p, brw_message_reg(5), brw_imm_f(0));
+      brw_MOV(p, brw_message_reg(6), src[2]);
+   }
+
+   brw_SAMPLE(p,
+	      retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW),
+	      1,
+	      retype(payload_reg, BRW_REGISTER_TYPE_UW),
+	      inst->TexSrcUnit + 1, /* surface */
+	      inst->TexSrcUnit,     /* sampler */
+	      inst->FullDstRegisters[0].DstRegister.WriteMask,
+	      BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE,
+	      4,
+	      shadow ? 6 : 4,
+	      0);
+
+   if (shadow)
+      brw_MOV(p, dst[3], brw_imm_f(1.0));
+#endif
+}
+
+
+
+
+
+
+
+
+static void emit_fb_write(struct brw_wm_compile *c,
+			  struct tgsi_full_instruction *inst)
+{
+   struct brw_compile *p = &c->func;
+   int nr = 2;
+   int channel;
+   int base_reg = 0;
+
+   // src0 = output color
+   // src1 = payload_depth[0]
+   // src2 = output depth
+   // dst = ???
+
+
+
+   /* Reserve a space for AA - may not be needed:
+    */
+   if (c->key.aa_dest_stencil_reg)
+      nr += 1;
+
+   {
+      brw_push_insn_state(p);
+      for (channel = 0; channel < 4; channel++) {
+	 struct brw_reg src0 = c->wm_regs[TGSI_FILE_OUTPUT][0][channel];
+
+	 /*  mov (8) m2.0<1>:ud   r28.0<8;8,1>:ud  { Align1 } */
+	 /*  mov (8) m6.0<1>:ud   r29.0<8;8,1>:ud  { Align1 SecHalf } */
+	 brw_MOV(p, brw_message_reg(nr + channel), src0);
+      }
+      /* skip over the regs populated above: */
+      nr += 8;
+      brw_pop_insn_state(p);
+   }
+    
+
+   /* Pass through control information:
+    */
+   /*  mov (8) m1.0<1>:ud   r1.0<8;8,1>:ud   { Align1 NoMask } */
+   {
+      brw_push_insn_state(p);
+      brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */
+      brw_MOV(p,
+	      brw_message_reg(base_reg + 1),
+	      brw_vec8_grf(1, 0));
+      brw_pop_insn_state(p);
+   }
+
+   /* Send framebuffer write message: */
+   brw_fb_WRITE(p,
+		retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW),
+		base_reg,
+		retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW),
+		0,              /* render surface always 0 */
+		nr,
+		0,
+		1);
+
+}
+
+
+static void brw_wm_emit_instruction( struct brw_wm_compile *c,
+				     struct tgsi_full_instruction *inst )
+{
+   struct brw_compile *p = &c->func;
+
+#if 0   
+   if (inst->CondUpdate)
+      brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ);
+   else
+      brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE);
+#else
+   brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE);
+#endif
+
+   switch (inst->Instruction.Opcode) {
+   case TGSI_OPCODE_ABS:
+      emit_abs(c, inst);
+      break;
+   case TGSI_OPCODE_ADD:
+      emit_alu2(c, inst, BRW_OPCODE_ADD);
+      break;
+   case TGSI_OPCODE_SUB:
+      assert(0);
+//      emit_alu2(c, inst, BRW_OPCODE_SUB);
+      break;
+   case TGSI_OPCODE_FRC:
+      emit_alu1(c, inst, BRW_OPCODE_FRC);
+      break;
+   case TGSI_OPCODE_FLR:
+      assert(0);
+//      emit_alu1(c, inst, BRW_OPCODE_FLR);
+      break;
+   case TGSI_OPCODE_LRP:
+      emit_lrp(c, inst);
+      break;
+   case TGSI_OPCODE_INT:
+      emit_alu1(c, inst, BRW_OPCODE_RNDD);
+      break;
+   case TGSI_OPCODE_MOV:
+      emit_alu1(c, inst, BRW_OPCODE_MOV);
+      break;
+   case TGSI_OPCODE_DP3:
+      emit_dp3(c, inst);
+      break;
+   case TGSI_OPCODE_DP4:
+      emit_dp4(c, inst);
+      break;
+   case TGSI_OPCODE_XPD:
+      emit_xpd(c, inst);
+      break;
+   case TGSI_OPCODE_DPH:
+      emit_dph(c, inst);
+      break;
+   case TGSI_OPCODE_RCP:
+      emit_math1(c, inst, BRW_MATH_FUNCTION_INV);
+      break;
+   case TGSI_OPCODE_RSQ:
+      emit_math1(c, inst, BRW_MATH_FUNCTION_RSQ);
+      break;
+   case TGSI_OPCODE_SIN:
+      emit_math1(c, inst, BRW_MATH_FUNCTION_SIN);
+      break;
+   case TGSI_OPCODE_COS:
+      emit_math1(c, inst, BRW_MATH_FUNCTION_COS);
+      break;
+   case TGSI_OPCODE_EX2:
+      emit_math1(c, inst, BRW_MATH_FUNCTION_EXP);
+      break;
+   case TGSI_OPCODE_LG2:
+      emit_math1(c, inst, BRW_MATH_FUNCTION_LOG);
+      break;
+   case TGSI_OPCODE_MAX:
+      emit_max(c, inst);
+      break;
+   case TGSI_OPCODE_MIN:
+      emit_min(c, inst);
+      break;
+   case TGSI_OPCODE_DDX:
+      emit_ddx(c, inst);
+      break;
+   case TGSI_OPCODE_DDY:
+      emit_ddy(c, inst);
+      break;
+   case TGSI_OPCODE_SLT:
+      emit_sop(c, inst, BRW_CONDITIONAL_L);
+      break;
+   case TGSI_OPCODE_SLE:
+      emit_sop(c, inst, BRW_CONDITIONAL_LE);
+      break;
+   case TGSI_OPCODE_SGT:
+      emit_sop(c, inst, BRW_CONDITIONAL_G);
+      break;
+   case TGSI_OPCODE_SGE:
+      emit_sop(c, inst, BRW_CONDITIONAL_GE);
+      break;
+   case TGSI_OPCODE_SEQ:
+      emit_sop(c, inst, BRW_CONDITIONAL_EQ);
+      break;
+   case TGSI_OPCODE_SNE:
+      emit_sop(c, inst, BRW_CONDITIONAL_NEQ);
+      break;
+   case TGSI_OPCODE_MUL:
+      emit_alu2(c, inst, BRW_OPCODE_MUL);
+      break;
+   case TGSI_OPCODE_POW:
+      emit_pow(c, inst);
+      break;
+   case TGSI_OPCODE_MAD:
+      emit_mad(c, inst);
+      break;
+   case TGSI_OPCODE_TEX:
+      emit_tex(c, inst);
+      break;
+   case TGSI_OPCODE_TXB:
+      emit_txb(c, inst);
+      break;
+   case TGSI_OPCODE_TEXKILL:
+      emit_kil(c);
+      break;
+   case TGSI_OPCODE_IF:
+      assert(c->if_insn < MAX_IFSN);
+      c->if_inst[c->if_insn++] = brw_IF(p, BRW_EXECUTE_8);
+      break;
+   case TGSI_OPCODE_ELSE:
+      c->if_inst[c->if_insn-1]  = brw_ELSE(p, c->if_inst[c->if_insn-1]);
+      break;
+   case TGSI_OPCODE_ENDIF:
+      assert(c->if_insn > 0);
+      brw_ENDIF(p, c->if_inst[--c->if_insn]);
+      break;
+   case TGSI_OPCODE_BGNSUB:
+   case TGSI_OPCODE_ENDSUB:
+      break;
+   case TGSI_OPCODE_CAL:
+      brw_push_insn_state(p);
+      brw_set_mask_control(p, BRW_MASK_DISABLE);
+      brw_set_access_mode(p, BRW_ALIGN_1);
+      brw_ADD(p, deref_1ud(c->stack_index, 0), brw_ip_reg(), brw_imm_d(3*16));
+      brw_set_access_mode(p, BRW_ALIGN_16);
+      brw_ADD(p, 
+	      get_addr_reg(c->stack_index),
+	      get_addr_reg(c->stack_index), brw_imm_d(4));
+//      orig_inst = inst->Data;
+//      orig_inst->Data = &p->store[p->nr_insn];
+      assert(0);
+      brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
+      brw_pop_insn_state(p);
+      break;
+
+   case TGSI_OPCODE_RET:
+#if 0
+      brw_push_insn_state(p);
+      brw_set_mask_control(p, BRW_MASK_DISABLE);
+      brw_ADD(p, 
+	      get_addr_reg(c->stack_index),
+	      get_addr_reg(c->stack_index), brw_imm_d(-4));
+      brw_set_access_mode(p, BRW_ALIGN_1);
+      brw_MOV(p, brw_ip_reg(), deref_1ud(c->stack_index, 0));
+      brw_set_access_mode(p, BRW_ALIGN_16);
+      brw_pop_insn_state(p);
+#else
+      emit_fb_write(c, inst);
+#endif
+
+      break;
+   case TGSI_OPCODE_LOOP:
+      c->loop_inst[c->loop_insn++] = brw_DO(p, BRW_EXECUTE_8);
+      break;
+   case TGSI_OPCODE_BRK:
+      brw_BREAK(p);
+      brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+      break;
+   case TGSI_OPCODE_CONT:
+      brw_CONT(p);
+      brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+      break;
+   case TGSI_OPCODE_ENDLOOP:
+      c->loop_insn--;
+      c->inst0 = c->inst1 = brw_WHILE(p, c->loop_inst[c->loop_insn]);
+      /* patch all the BREAK instructions from
+	 last BEGINLOOP */
+      while (c->inst0 > c->loop_inst[c->loop_insn]) {
+	 c->inst0--;
+	 if (c->inst0->header.opcode == BRW_OPCODE_BREAK) {
+	    c->inst0->bits3.if_else.jump_count = c->inst1 - c->inst0 + 1;
+	    c->inst0->bits3.if_else.pop_count = 0;
+	 } else if (c->inst0->header.opcode == BRW_OPCODE_CONTINUE) {
+	    c->inst0->bits3.if_else.jump_count = c->inst1 - c->inst0;
+	    c->inst0->bits3.if_else.pop_count = 0;
+	 }
+      }
+      break;
+   case TGSI_OPCODE_END:
+      emit_fb_write(c, inst);
+      break;
+
+   default:
+      debug_printf("unsupported IR in fragment shader %d\n",
+		   inst->Instruction.Opcode);
+   }
+#if 0
+   if (inst->CondUpdate)
+      brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+   else
+      brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+#endif
+}
+
+
+
+
+
+
+void brw_wm_glsl_emit(struct brw_wm_compile *c)
+{
+   struct tgsi_parse_context parse;
+   struct brw_compile *p = &c->func;
+
+   brw_init_compile(&c->func);
+   brw_set_compression_control(p, BRW_COMPRESSION_NONE);
+
+   c->reg_index = 0;
+   c->if_insn = 0;
+   c->loop_insn = 0;
+   c->stack_index = brw_indirect(0,0);
+
+   /* Do static register allocation and parameter interpolation:
+    */
+   brw_wm_emit_decls( c );
+
+   /* Emit the actual program.  All done with very direct translation,
+    * hopefully we can improve on this shortly...
+    */
+   brw_MOV(p, get_addr_reg(c->stack_index), brw_address(c->stack));
+
+   tgsi_parse_init( &parse, c->fp->program.tokens );
+
+   while( !tgsi_parse_end_of_tokens( &parse ) ) 
+   {
+      tgsi_parse_token( &parse );
+
+      switch( parse.FullToken.Token.Type ) {
+      case TGSI_TOKEN_TYPE_DECLARATION:
+	 /* already done */
+	 break;
+
+      case TGSI_TOKEN_TYPE_IMMEDIATE:
+         /* not handled yet */
+	 assert(0);
+         break;
+
+      case TGSI_TOKEN_TYPE_INSTRUCTION:
+         brw_wm_emit_instruction(c, &parse.FullToken.FullInstruction);
+         break;
+
+      default:
+         assert( 0 );
+      }
+   }
+
+   tgsi_parse_free (&parse);
+   
+   /* Fix up call targets:
+    */
+#if 0
+   {
+      unsigned nr_insns = c->fp->program.Base.NumInstructions;
+      unsigned insn, target_insn;
+      struct tgsi_full_instruction *inst1, *inst2;
+      struct brw_instruction *brw_inst1, *brw_inst2;
+      int offset;
+      for (insn = 0; insn < nr_insns; insn++) {
+	 inst1 = &c->fp->program.Base.Instructions[insn];
+	 brw_inst1 = inst1->Data;
+	 switch (inst1->Opcode) {
+	 case TGSI_OPCODE_CAL:
+	    target_insn = inst1->BranchTarget;
+	    inst2 = &c->fp->program.Base.Instructions[target_insn];
+	    brw_inst2 = inst2->Data;
+	    offset = brw_inst2 - brw_inst1;
+	    brw_set_src1(brw_inst1, brw_imm_d(offset*16));
+	    break;
+	 default:
+	    break;
+	 }
+      }
+   }
+#endif
+
+   c->prog_data.total_grf = c->reg_index;
+   c->prog_data.total_scratch = 0;
+}
diff --git a/src/gallium/drivers/i965simple/brw_wm_iz.c b/src/gallium/drivers/i965simple/brw_wm_iz.c
new file mode 100644
index 0000000000..6c5f25bf39
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_wm_iz.c
@@ -0,0 +1,214 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_wm.h"
+
+
+#undef P			/* prompted depth */
+#undef C			/* computed */
+#undef N			/* non-promoted? */
+
+#define P 0
+#define C 1
+#define N 2
+
+const struct {
+   unsigned mode:2;
+   unsigned sd_present:1;
+   unsigned sd_to_rt:1;
+   unsigned dd_present:1;
+   unsigned ds_present:1;
+} wm_iz_table[IZ_BIT_MAX] =
+{
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { C, 0, 1, 0, 0 }, 
+ { C, 0, 1, 0, 0 }, 
+ { C, 1, 1, 0, 0 }, 
+ { C, 1, 1, 0, 0 }, 
+ { C, 0, 1, 0, 0 }, 
+ { C, 0, 1, 0, 0 }, 
+ { C, 1, 1, 1, 0 }, 
+ { C, 1, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 1, 1, 1, 0 }, 
+ { C, 1, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { C, 0, 1, 0, 0 }, 
+ { C, 0, 1, 0, 0 }, 
+ { C, 1, 1, 0, 0 }, 
+ { C, 1, 1, 0, 0 }, 
+ { C, 0, 1, 0, 0 }, 
+ { C, 0, 1, 0, 0 }, 
+ { C, 1, 1, 1, 0 }, 
+ { C, 1, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 1, 1, 1, 0 }, 
+ { C, 1, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 0, 0, 0, 1 }, 
+ { C, 0, 0, 0, 1 }, 
+ { C, 0, 1, 0, 1 }, 
+ { C, 0, 1, 0, 1 }, 
+ { C, 1, 1, 0, 1 }, 
+ { C, 1, 1, 0, 1 }, 
+ { C, 0, 1, 0, 1 }, 
+ { C, 0, 1, 0, 1 }, 
+ { C, 1, 1, 1, 1 }, 
+ { C, 1, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 1, 1, 1, 1 }, 
+ { C, 1, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 0, 0, 0, 1 }, 
+ { C, 0, 0, 0, 1 }, 
+ { C, 0, 1, 0, 1 }, 
+ { C, 0, 1, 0, 1 }, 
+ { C, 1, 1, 0, 1 }, 
+ { C, 1, 1, 0, 1 }, 
+ { C, 0, 1, 0, 1 }, 
+ { C, 0, 1, 0, 1 }, 
+ { C, 1, 1, 1, 1 }, 
+ { C, 1, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 1, 1, 1, 1 }, 
+ { C, 1, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { N, 1, 1, 0, 0 }, 
+ { N, 0, 1, 0, 0 }, 
+ { N, 0, 1, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { N, 1, 1, 0, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { N, 1, 1, 0, 0 }, 
+ { N, 0, 1, 0, 0 }, 
+ { N, 0, 1, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { N, 1, 1, 0, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { C, 0, 1, 1, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { N, 1, 1, 0, 1 }, 
+ { N, 0, 1, 0, 1 }, 
+ { N, 0, 1, 0, 1 }, 
+ { P, 0, 0, 0, 0 }, 
+ { P, 0, 0, 0, 0 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { P, 0, 0, 0, 0 }, 
+ { N, 1, 1, 0, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { P, 0, 0, 0, 0 }, 
+ { C, 0, 0, 0, 1 }, 
+ { P, 0, 0, 0, 0 }, 
+ { C, 0, 1, 0, 1 }, 
+ { P, 0, 0, 0, 0 }, 
+ { C, 1, 1, 0, 1 }, 
+ { C, 0, 1, 0, 1 }, 
+ { C, 0, 1, 0, 1 }, 
+ { P, 0, 0, 0, 0 }, 
+ { C, 1, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { P, 0, 0, 0, 0 }, 
+ { C, 1, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 }, 
+ { C, 0, 1, 1, 1 } 
+};
+
+void brw_wm_lookup_iz( unsigned line_aa,
+		       unsigned lookup,
+		       struct brw_wm_prog_key *key )
+{
+   unsigned reg = 2;
+
+   assert (lookup < IZ_BIT_MAX);
+      
+   if (lookup & IZ_PS_COMPUTES_DEPTH_BIT)
+      key->computes_depth = 1;
+
+   if (wm_iz_table[lookup].sd_present) {
+      key->source_depth_reg = reg;
+      reg += 2;
+   }
+
+   if (wm_iz_table[lookup].sd_to_rt)
+      key->source_depth_to_render_target = 1;
+
+   if (wm_iz_table[lookup].ds_present || line_aa != AA_NEVER) {
+      key->aa_dest_stencil_reg = reg;
+      key->runtime_check_aads_emit = (!wm_iz_table[lookup].ds_present &&
+				      line_aa == AA_SOMETIMES);
+      reg++;
+   }
+
+   if (wm_iz_table[lookup].dd_present) {
+      key->dest_depth_reg = reg;
+      reg+=2;
+   }
+
+   key->nr_depth_regs = (reg+1)/2;
+}
+
diff --git a/src/gallium/drivers/i965simple/brw_wm_sampler_state.c b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c
new file mode 100644
index 0000000000..52b2909a65
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c
@@ -0,0 +1,275 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+
+#define COMPAREFUNC_ALWAYS		0
+#define COMPAREFUNC_NEVER		0x1
+#define COMPAREFUNC_LESS		0x2
+#define COMPAREFUNC_EQUAL		0x3
+#define COMPAREFUNC_LEQUAL		0x4
+#define COMPAREFUNC_GREATER		0x5
+#define COMPAREFUNC_NOTEQUAL		0x6
+#define COMPAREFUNC_GEQUAL		0x7
+
+/* Samplers aren't strictly wm state from the hardware's perspective,
+ * but that is the only situation in which we use them in this driver.
+ */
+
+static int intel_translate_shadow_compare_func(unsigned func)
+{
+   switch(func) {
+   case PIPE_FUNC_NEVER:
+       return COMPAREFUNC_ALWAYS;
+   case PIPE_FUNC_LESS:
+       return COMPAREFUNC_LEQUAL;
+   case PIPE_FUNC_LEQUAL:
+       return COMPAREFUNC_LESS;
+   case PIPE_FUNC_GREATER:
+       return COMPAREFUNC_GEQUAL;
+   case PIPE_FUNC_GEQUAL:
+      return COMPAREFUNC_GREATER;
+   case PIPE_FUNC_NOTEQUAL:
+      return COMPAREFUNC_EQUAL;
+   case PIPE_FUNC_EQUAL:
+      return COMPAREFUNC_NOTEQUAL;
+   case PIPE_FUNC_ALWAYS:
+       return COMPAREFUNC_NEVER;
+   }
+
+   debug_printf("Unknown value in %s: %x\n", __FUNCTION__, func);
+   return COMPAREFUNC_NEVER;
+}
+
+/* The brw (and related graphics cores) do not support GL_CLAMP.  The
+ * Intel drivers for "other operating systems" implement GL_CLAMP as
+ * GL_CLAMP_TO_EDGE, so the same is done here.
+ */
+static unsigned translate_wrap_mode( int wrap )
+{
+   switch( wrap ) {
+   case PIPE_TEX_WRAP_REPEAT:
+      return BRW_TEXCOORDMODE_WRAP;
+   case PIPE_TEX_WRAP_CLAMP:
+      return BRW_TEXCOORDMODE_CLAMP;
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      return BRW_TEXCOORDMODE_CLAMP; /* conform likes it this way */
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      return BRW_TEXCOORDMODE_CLAMP_BORDER;
+   case PIPE_TEX_WRAP_MIRROR_REPEAT:
+      return BRW_TEXCOORDMODE_MIRROR;
+   default:
+      return BRW_TEXCOORDMODE_WRAP;
+   }
+}
+
+
+static unsigned U_FIXED(float value, unsigned frac_bits)
+{
+   value *= (1<<frac_bits);
+   return value < 0 ? 0 : value;
+}
+
+static int S_FIXED(float value, unsigned frac_bits)
+{
+   return value * (1<<frac_bits);
+}
+
+
+static unsigned upload_default_color( struct brw_context *brw,
+                                      const float *color )
+{
+   struct brw_sampler_default_color sdc;
+
+   COPY_4V(sdc.color, color);
+
+   return brw_cache_data( &brw->cache[BRW_SAMPLER_DEFAULT_COLOR], &sdc );
+}
+
+
+/*
+ */
+static void brw_update_sampler_state( const struct pipe_sampler_state *pipe_sampler,
+				      unsigned sdc_gs_offset,
+				      struct brw_sampler_state *sampler)
+{
+   memset(sampler, 0, sizeof(*sampler));
+
+   switch (pipe_sampler->min_mip_filter) {
+   case PIPE_TEX_FILTER_NEAREST:
+      sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST;
+      break;
+   case PIPE_TEX_FILTER_LINEAR:
+      sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR;
+      break;
+   case PIPE_TEX_FILTER_ANISO:
+      sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC;
+      break;
+   default:
+      break;
+   }
+
+   switch (pipe_sampler->min_mip_filter) {
+   case PIPE_TEX_MIPFILTER_NEAREST:
+      sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST;
+      break;
+   case PIPE_TEX_MIPFILTER_LINEAR:
+      sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR;
+      break;
+   case PIPE_TEX_MIPFILTER_NONE:
+      sampler->ss0.mip_filter = BRW_MIPFILTER_NONE;
+      break;
+   default:
+      break;
+   }
+   /* Set Anisotropy:
+    */
+   switch (pipe_sampler->mag_img_filter) {
+   case PIPE_TEX_FILTER_NEAREST:
+      sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST;
+      break;
+   case PIPE_TEX_FILTER_LINEAR:
+      sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR;
+      break;
+   case PIPE_TEX_FILTER_ANISO:
+      sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR;
+      break;
+   default:
+      break;
+   }
+
+   if (pipe_sampler->max_anisotropy > 2.0) {
+      sampler->ss3.max_aniso = MAX2((pipe_sampler->max_anisotropy - 2) / 2,
+                                    BRW_ANISORATIO_16);
+   }
+
+   sampler->ss1.s_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_s);
+   sampler->ss1.r_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_r);
+   sampler->ss1.t_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_t);
+
+   /* Fulsim complains if I don't do this.  Hardware doesn't mind:
+    */
+#if 0
+   if (texObj->Target == GL_TEXTURE_CUBE_MAP_ARB) {
+      sampler->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CUBE;
+      sampler->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CUBE;
+      sampler->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CUBE;
+   }
+#endif
+
+   /* Set shadow function:
+    */
+   if (pipe_sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+      /* Shadowing is "enabled" by emitting a particular sampler
+       * message (sample_c).  So need to recompile WM program when
+       * shadow comparison is enabled on each/any texture unit.
+       */
+      sampler->ss0.shadow_function = intel_translate_shadow_compare_func(pipe_sampler->compare_func);
+   }
+
+   /* Set LOD bias:
+    */
+   sampler->ss0.lod_bias = S_FIXED(CLAMP(pipe_sampler->lod_bias, -16, 15), 6);
+
+   sampler->ss0.lod_preclamp = 1; /* OpenGL mode */
+   sampler->ss0.default_color_mode = 0; /* OpenGL/DX10 mode */
+
+   /* Set BaseMipLevel, MaxLOD, MinLOD:
+    *
+    * XXX: I don't think that using firstLevel, lastLevel works,
+    * because we always setup the surface state as if firstLevel ==
+    * level zero.  Probably have to subtract firstLevel from each of
+    * these:
+    */
+   sampler->ss0.base_level = U_FIXED(0, 1);
+
+   sampler->ss1.max_lod = U_FIXED(MIN2(MAX2(pipe_sampler->max_lod, 0), 13), 6);
+   sampler->ss1.min_lod = U_FIXED(MIN2(MAX2(pipe_sampler->min_lod, 0), 13), 6);
+
+   sampler->ss2.default_color_pointer = sdc_gs_offset >> 5;
+}
+
+
+
+/* All samplers must be uploaded in a single contiguous array, which
+ * complicates various things.  However, this is still too confusing -
+ * FIXME: simplify all the different new texture state flags.
+ */
+static void upload_wm_samplers(struct brw_context *brw)
+{
+   unsigned unit;
+   unsigned sampler_count = 0;
+
+   /* BRW_NEW_SAMPLER */
+   for (unit = 0; unit < brw->num_textures && unit < brw->num_samplers;
+        unit++) {
+      /* determine unit enable/disable by looking for a bound texture */
+      if (brw->attribs.Texture[unit]) {
+         const struct pipe_sampler_state *sampler = brw->attribs.Samplers[unit];
+	 unsigned sdc_gs_offset = upload_default_color(brw, sampler->border_color);
+
+	 brw_update_sampler_state(sampler,
+				  sdc_gs_offset,
+				  &brw->wm.sampler[unit]);
+
+	 sampler_count = unit + 1;
+      }
+   }
+
+   if (brw->wm.sampler_count != sampler_count) {
+      brw->wm.sampler_count = sampler_count;
+      brw->state.dirty.cache |= CACHE_NEW_SAMPLER;
+   }
+
+   brw->wm.sampler_gs_offset = 0;
+
+   if (brw->wm.sampler_count)
+      brw->wm.sampler_gs_offset =
+	 brw_cache_data_sz(&brw->cache[BRW_SAMPLER],
+			   brw->wm.sampler,
+			   sizeof(struct brw_sampler_state) * brw->wm.sampler_count);
+}
+
+const struct brw_tracked_state brw_wm_samplers = {
+   .dirty = {
+      .brw = BRW_NEW_SAMPLER,
+      .cache = 0
+   },
+   .update = upload_wm_samplers
+};
+
diff --git a/src/gallium/drivers/i965simple/brw_wm_state.c b/src/gallium/drivers/i965simple/brw_wm_state.c
new file mode 100644
index 0000000000..37a9bf919c
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_wm_state.c
@@ -0,0 +1,195 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+#include "brw_wm.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+/***********************************************************************
+ * WM unit - fragment programs and rasterization
+ */
+static void upload_wm_unit(struct brw_context *brw )
+{
+   struct brw_wm_unit_state wm;
+   unsigned max_threads;
+   unsigned per_thread;
+
+   if (BRW_DEBUG & DEBUG_SINGLE_THREAD)
+      max_threads = 0;
+   else
+      max_threads = 31;
+
+
+   memset(&wm, 0, sizeof(wm));
+
+   /* CACHE_NEW_WM_PROG */
+   wm.thread0.grf_reg_count = align(brw->wm.prog_data->total_grf, 16) / 16 - 1;
+   wm.thread0.kernel_start_pointer = brw->wm.prog_gs_offset >> 6;
+   wm.thread3.dispatch_grf_start_reg = brw->wm.prog_data->first_curbe_grf;
+   wm.thread3.urb_entry_read_length = brw->wm.prog_data->urb_read_length;
+   wm.thread3.const_urb_entry_read_length = brw->wm.prog_data->curb_read_length;
+
+   wm.wm5.max_threads = max_threads;
+
+   per_thread = align(brw->wm.prog_data->total_scratch, 1024);
+   assert(per_thread <= 12 * 1024);
+
+#if 0
+   if (brw->wm.prog_data->total_scratch) {
+      unsigned total = per_thread * (max_threads + 1);
+
+      /* Scratch space -- just have to make sure there is sufficient
+       * allocated for the active program and current number of threads.
+       */
+      brw->wm.scratch_buffer_size = total;
+      if (brw->wm.scratch_buffer &&
+	  brw->wm.scratch_buffer_size > brw->wm.scratch_buffer->size) {
+	 dri_bo_unreference(brw->wm.scratch_buffer);
+	 brw->wm.scratch_buffer = NULL;
+      }
+      if (!brw->wm.scratch_buffer) {
+	 brw->wm.scratch_buffer = dri_bo_alloc(intel->intelScreen->bufmgr,
+					       "wm scratch",
+					       brw->wm.scratch_buffer_size,
+					       4096, DRM_BO_FLAG_MEM_TT);
+      }
+   }
+   /* XXX: Scratch buffers are not implemented correectly.
+    *
+    * The scratch offset to be programmed into wm is relative to the general
+    * state base address.  However, using dri_bo_alloc/dri_bo_emit_reloc (or
+    * the previous bmGenBuffers scheme), we get an offset relative to the
+    * start of framebuffer.  Even before then, it was broken in other ways,
+    * so just fail for now if we hit that path.
+    */
+   assert(brw->wm.prog_data->total_scratch == 0);
+#endif
+
+   /* CACHE_NEW_SURFACE */
+   wm.thread1.binding_table_entry_count = brw->wm.nr_surfaces;
+
+   /* BRW_NEW_CURBE_OFFSETS */
+   wm.thread3.const_urb_entry_read_offset = brw->curbe.wm_start * 2;
+
+   wm.thread3.urb_entry_read_offset = 0;
+   wm.thread1.depth_coef_urb_read_offset = 1;
+   wm.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
+
+   /* CACHE_NEW_SAMPLER */
+   wm.wm4.sampler_count = (brw->wm.sampler_count + 1) / 4;
+   wm.wm4.sampler_state_pointer = brw->wm.sampler_gs_offset >> 5;
+
+   /* BRW_NEW_FRAGMENT_PROGRAM */
+   {
+      const struct brw_fragment_program *fp = brw->attribs.FragmentProgram;
+
+      if (fp->UsesDepth)
+	 wm.wm5.program_uses_depth = 1; /* as far as we can tell */
+
+      if (fp->info.writes_z)
+	 wm.wm5.program_computes_depth = 1;
+
+      /* BRW_NEW_ALPHA_TEST */
+      if (fp->info.uses_kill ||
+	  brw->attribs.DepthStencil->alpha.enabled)
+	 wm.wm5.program_uses_killpixel = 1;
+
+      wm.wm5.enable_8_pix = 1;
+   }
+
+   wm.wm5.thread_dispatch_enable = 1;	/* AKA: color_write */
+   wm.wm5.legacy_line_rast = 0;
+   wm.wm5.legacy_global_depth_bias = 0;
+   wm.wm5.early_depth_test = 1;	        /* never need to disable */
+   wm.wm5.line_aa_region_width = 0;
+   wm.wm5.line_endcap_aa_region_width = 1;
+
+   /* BRW_NEW_RASTERIZER */
+   if (brw->attribs.Raster->poly_stipple_enable)
+      wm.wm5.polygon_stipple = 1;
+
+#if 0
+   if (brw->attribs.Polygon->OffsetFill) {
+      wm.wm5.depth_offset = 1;
+      /* Something wierd going on with legacy_global_depth_bias,
+       * offset_constant, scaling and MRD.  This value passes glean
+       * but gives some odd results elsewere (eg. the
+       * quad-offset-units test).
+       */
+      wm.global_depth_offset_constant = brw->attribs.Polygon->OffsetUnits * 2;
+
+      /* This is the only value that passes glean:
+       */
+      wm.global_depth_offset_scale = brw->attribs.Polygon->OffsetFactor;
+   }
+#endif
+
+   if (brw->attribs.Raster->line_stipple_enable) {
+      wm.wm5.line_stipple = 1;
+   }
+
+   if (BRW_DEBUG & DEBUG_STATS)
+      wm.wm4.stats_enable = 1;
+
+   brw->wm.state_gs_offset = brw_cache_data( &brw->cache[BRW_WM_UNIT], &wm );
+
+   if (brw->wm.prog_data->total_scratch) {
+      /*
+      dri_emit_reloc(brw->cache[BRW_WM_UNIT].pool->buffer,
+		     DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE,
+		     (per_thread / 1024) - 1,
+		     brw->wm.state_gs_offset +
+		     ((char *)&wm.thread2 - (char *)&wm),
+		     brw->wm.scratch_buffer);
+      */
+   } else {
+      wm.thread2.scratch_space_base_pointer = 0;
+   }
+}
+
+const struct brw_tracked_state brw_wm_unit = {
+   .dirty = {
+      .brw = (BRW_NEW_RASTERIZER |
+	      BRW_NEW_ALPHA_TEST |
+	      BRW_NEW_FS |
+	      BRW_NEW_CURBE_OFFSETS),
+
+      .cache = (CACHE_NEW_SURFACE |
+		CACHE_NEW_WM_PROG |
+		CACHE_NEW_SAMPLER)
+   },
+   .update = upload_wm_unit
+};
+
diff --git a/src/gallium/drivers/i965simple/brw_wm_surface_state.c b/src/gallium/drivers/i965simple/brw_wm_surface_state.c
new file mode 100644
index 0000000000..1bab5bfdb3
--- /dev/null
+++ b/src/gallium/drivers/i965simple/brw_wm_surface_state.c
@@ -0,0 +1,305 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "brw_context.h"
+#include "brw_state.h"
+#include "brw_defines.h"
+
+static unsigned translate_tex_target( enum pipe_texture_target target )
+{
+   switch (target) {
+   case PIPE_TEXTURE_1D:
+      return BRW_SURFACE_1D;
+
+   case PIPE_TEXTURE_2D:
+      return BRW_SURFACE_2D;
+
+   case PIPE_TEXTURE_3D:
+      return BRW_SURFACE_3D;
+
+   case PIPE_TEXTURE_CUBE:
+      return BRW_SURFACE_CUBE;
+
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+static unsigned translate_tex_format( enum pipe_format pipe_format )
+{
+   switch( pipe_format ) {
+   case PIPE_FORMAT_L8_UNORM:
+      return BRW_SURFACEFORMAT_L8_UNORM;
+
+   case PIPE_FORMAT_I8_UNORM:
+      return BRW_SURFACEFORMAT_I8_UNORM;
+
+   case PIPE_FORMAT_A8_UNORM:
+      return BRW_SURFACEFORMAT_A8_UNORM;
+
+   case PIPE_FORMAT_A8L8_UNORM:
+      return BRW_SURFACEFORMAT_L8A8_UNORM;
+
+   case PIPE_FORMAT_R8G8B8_UNORM:
+      assert(0);		/* not supported for sampling */
+      return BRW_SURFACEFORMAT_R8G8B8_UNORM;
+
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      return BRW_SURFACEFORMAT_B8G8R8A8_UNORM;
+
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      return BRW_SURFACEFORMAT_R8G8B8A8_UNORM;
+
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      return BRW_SURFACEFORMAT_B5G6R5_UNORM;
+
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      return BRW_SURFACEFORMAT_B5G5R5A1_UNORM;
+
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      return BRW_SURFACEFORMAT_B4G4R4A4_UNORM;
+
+   case PIPE_FORMAT_YCBCR_REV:
+      return BRW_SURFACEFORMAT_YCRCB_NORMAL;
+
+   case PIPE_FORMAT_YCBCR:
+      return BRW_SURFACEFORMAT_YCRCB_SWAPUVY;
+#if 0
+   case PIPE_FORMAT_RGB_FXT1:
+   case PIPE_FORMAT_RGBA_FXT1:
+      return BRW_SURFACEFORMAT_FXT1;
+#endif
+
+   case PIPE_FORMAT_Z16_UNORM:
+      return BRW_SURFACEFORMAT_I16_UNORM;
+#if 0
+   case PIPE_FORMAT_RGB_DXT1:
+       return BRW_SURFACEFORMAT_DXT1_RGB;
+
+   case PIPE_FORMAT_RGBA_DXT1:
+       return BRW_SURFACEFORMAT_BC1_UNORM;
+
+   case PIPE_FORMAT_RGBA_DXT3:
+       return BRW_SURFACEFORMAT_BC2_UNORM;
+
+   case PIPE_FORMAT_RGBA_DXT5:
+       return BRW_SURFACEFORMAT_BC3_UNORM;
+
+   case PIPE_FORMAT_SRGBA8:
+      return BRW_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB;
+   case PIPE_FORMAT_SRGB_DXT1:
+      return BRW_SURFACEFORMAT_BC1_UNORM_SRGB;
+#endif
+
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+static unsigned brw_buffer_offset(struct brw_context *brw,
+                                  struct pipe_buffer *buffer)
+{
+   return brw->winsys->get_buffer_offset(brw->winsys,
+                                         buffer,
+                                         0);
+}
+
+static
+void brw_update_texture_surface( struct brw_context *brw,
+				 unsigned unit )
+{
+   const struct brw_texture *tObj = brw->attribs.Texture[unit];
+   struct brw_surface_state surf;
+
+   memset(&surf, 0, sizeof(surf));
+
+   surf.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW;
+   surf.ss0.surface_type = translate_tex_target(tObj->base.target);
+   surf.ss0.surface_format = translate_tex_format(tObj->base.format);
+
+   /* This is ok for all textures with channel width 8bit or less:
+    */
+/*    surf.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */
+
+   /* Updated in emit_reloc */
+   surf.ss1.base_addr = brw_buffer_offset( brw, tObj->buffer );
+
+   surf.ss2.mip_count = tObj->base.last_level;
+   surf.ss2.width = tObj->base.width[0] - 1;
+   surf.ss2.height = tObj->base.height[0] - 1;
+
+   surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR;
+   surf.ss3.tiled_surface = 0; /* always zero */
+   surf.ss3.pitch = tObj->stride - 1;
+   surf.ss3.depth = tObj->base.depth[0] - 1;
+
+   surf.ss4.min_lod = 0;
+
+   if (tObj->base.target == PIPE_TEXTURE_CUBE) {
+      surf.ss0.cube_pos_x = 1;
+      surf.ss0.cube_pos_y = 1;
+      surf.ss0.cube_pos_z = 1;
+      surf.ss0.cube_neg_x = 1;
+      surf.ss0.cube_neg_y = 1;
+      surf.ss0.cube_neg_z = 1;
+   }
+
+   brw->wm.bind.surf_ss_offset[unit + 1] =
+      brw_cache_data( &brw->cache[BRW_SS_SURFACE], &surf );
+}
+
+
+
+#define OFFSET(TYPE, FIELD) ( (unsigned)&(((TYPE *)0)->FIELD) )
+
+
+static void upload_wm_surfaces(struct brw_context *brw )
+{
+   unsigned i;
+
+   {
+      struct brw_surface_state surf;
+
+      /* BRW_NEW_FRAMEBUFFER
+       */
+      struct pipe_surface *pipe_surface = brw->attribs.FrameBuffer.cbufs[0];/*fixme*/
+      struct brw_texture *tex = (struct brw_texture *)pipe_surface->texture;
+
+      memset(&surf, 0, sizeof(surf));
+
+      if (pipe_surface != NULL) {
+	 if (pipe_surface->block.size == 4)
+	    surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM;
+	 else
+	    surf.ss0.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM;
+
+	 surf.ss0.surface_type = BRW_SURFACE_2D;
+
+	 surf.ss1.base_addr = brw_buffer_offset( brw, tex->buffer );
+
+	 surf.ss2.width = pipe_surface->width - 1;
+	 surf.ss2.height = pipe_surface->height - 1;
+	 surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR;
+	 surf.ss3.tiled_surface = 0;
+	 surf.ss3.pitch = pipe_surface->stride - 1;
+      } else {
+	 surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM;
+	 surf.ss0.surface_type = BRW_SURFACE_NULL;
+      }
+
+      /* BRW_NEW_BLEND */
+      surf.ss0.color_blend = (!brw->attribs.Blend->logicop_enable &&
+			      brw->attribs.Blend->blend_enable);
+
+
+      surf.ss0.writedisable_red =   !(brw->attribs.Blend->colormask & PIPE_MASK_R);
+      surf.ss0.writedisable_green = !(brw->attribs.Blend->colormask & PIPE_MASK_G);
+      surf.ss0.writedisable_blue =  !(brw->attribs.Blend->colormask & PIPE_MASK_B);
+      surf.ss0.writedisable_alpha = !(brw->attribs.Blend->colormask & PIPE_MASK_A);
+
+
+
+
+      brw->wm.bind.surf_ss_offset[0] = brw_cache_data( &brw->cache[BRW_SS_SURFACE], &surf );
+
+      brw->wm.nr_surfaces = 1;
+   }
+
+
+   /* BRW_NEW_TEXTURE
+    */
+   for (i = 0; i < brw->num_textures && i < brw->num_samplers; i++) {
+      const struct brw_texture *texUnit = brw->attribs.Texture[i];
+
+      if (texUnit &&
+	  texUnit->base.refcount/*(texUnit->refcount > 0) == really used */) {
+
+	 brw_update_texture_surface(brw, i);
+
+	 brw->wm.nr_surfaces = i+2;
+      }
+      else {
+	 brw->wm.bind.surf_ss_offset[i+1] = 0;
+      }
+   }
+
+   brw->wm.bind_ss_offset = brw_cache_data( &brw->cache[BRW_SS_SURF_BIND],
+					    &brw->wm.bind );
+}
+
+
+/* KW: Will find a different way to acheive this, see for example the
+ * state caches with relocs in the i915 swz driver.
+ */
+#if 0
+static void emit_reloc_wm_surfaces(struct brw_context *brw)
+{
+   int unit;
+
+   if (brw->state.draw_region != NULL) {
+      /* Emit framebuffer relocation */
+      dri_emit_reloc(brw_cache_buffer(brw, BRW_SS_SURFACE),
+		     DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE,
+		     0,
+		     brw->wm.bind.surf_ss_offset[0] +
+		     offsetof(struct brw_surface_state, ss1),
+		     brw->state.draw_region->buffer);
+   }
+
+   /* Emit relocations for texture buffers */
+   for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) {
+      struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[unit];
+      struct gl_texture_object *tObj = texUnit->_Current;
+      struct intel_texture_object *intelObj = intel_texture_object(tObj);
+
+      if (texUnit->_ReallyEnabled && intelObj->mt != NULL) {
+	 dri_emit_reloc(brw_cache_buffer(brw, BRW_SS_SURFACE),
+			DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ,
+			0,
+			brw->wm.bind.surf_ss_offset[unit + 1] +
+			offsetof(struct brw_surface_state, ss1),
+			intelObj->mt->region->buffer);
+      }
+   }
+}
+#endif
+
+const struct brw_tracked_state brw_wm_surfaces = {
+   .dirty = {
+      .brw = (BRW_NEW_FRAMEBUFFER |
+	      BRW_NEW_BLEND |
+	      BRW_NEW_TEXTURE),
+      .cache = 0
+   },
+   .update = upload_wm_surfaces,
+};
diff --git a/src/gallium/drivers/nouveau/nouveau_gldefs.h b/src/gallium/drivers/nouveau/nouveau_gldefs.h
new file mode 100644
index 0000000000..ff97aaa9af
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nouveau_gldefs.h
@@ -0,0 +1,196 @@
+#ifndef __NOUVEAU_GLDEFS_H__
+#define __NOUVEAU_GLDEFS_H__
+
+static INLINE unsigned
+nvgl_blend_func(unsigned factor)
+{
+	switch (factor) {
+	case PIPE_BLENDFACTOR_ZERO:
+		return 0x0000;
+	case PIPE_BLENDFACTOR_ONE:
+		return 0x0001;
+	case PIPE_BLENDFACTOR_SRC_COLOR:
+		return 0x0300;
+	case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+		return 0x0301;
+	case PIPE_BLENDFACTOR_SRC_ALPHA:
+		return 0x0302;
+	case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+		return 0x0303;
+	case PIPE_BLENDFACTOR_DST_ALPHA:
+		return 0x0304;
+	case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+		return 0x0305;
+	case PIPE_BLENDFACTOR_DST_COLOR:
+		return 0x0306;
+	case PIPE_BLENDFACTOR_INV_DST_COLOR:
+		return 0x0307;
+	case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+		return 0x0308;
+	case PIPE_BLENDFACTOR_CONST_COLOR:
+		return 0x8001;
+	case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+		return 0x8002;
+	case PIPE_BLENDFACTOR_CONST_ALPHA:
+		return 0x8003;
+	case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+		return 0x8004;
+	default:
+		return 0x0000;
+	}
+}
+
+static INLINE unsigned
+nvgl_blend_eqn(unsigned func)
+{
+	switch (func) {
+	case PIPE_BLEND_ADD:
+		return 0x8006;
+	case PIPE_BLEND_MIN:
+		return 0x8007;
+	case PIPE_BLEND_MAX:
+		return 0x8008;
+	case PIPE_BLEND_SUBTRACT:
+		return 0x800a;
+	case PIPE_BLEND_REVERSE_SUBTRACT:
+		return 0x800b;
+	default:
+		return 0x8006;
+	}
+}
+
+static INLINE unsigned
+nvgl_logicop_func(unsigned func)
+{
+	switch (func) {
+	case PIPE_LOGICOP_CLEAR:
+		return 0x1500;
+	case PIPE_LOGICOP_NOR:
+		return 0x1508;
+	case PIPE_LOGICOP_AND_INVERTED:
+		return 0x1504;
+	case PIPE_LOGICOP_COPY_INVERTED:
+		return 0x150c;
+	case PIPE_LOGICOP_AND_REVERSE:
+		return 0x1502;
+	case PIPE_LOGICOP_INVERT:
+		return 0x150a;
+	case PIPE_LOGICOP_XOR:
+		return 0x1506;
+	case PIPE_LOGICOP_NAND:
+		return 0x150e;
+	case PIPE_LOGICOP_AND:
+		return 0x1501;
+	case PIPE_LOGICOP_EQUIV:
+		return 0x1509;
+	case PIPE_LOGICOP_NOOP:
+		return 0x1505;
+	case PIPE_LOGICOP_OR_INVERTED:
+		return 0x150d;
+	case PIPE_LOGICOP_COPY:
+		return 0x1503;
+	case PIPE_LOGICOP_OR_REVERSE:
+		return 0x150b;
+	case PIPE_LOGICOP_OR:
+		return 0x1507;
+	case PIPE_LOGICOP_SET:
+		return 0x150f;
+	default:
+		return 0x1505;
+	}
+}
+
+static INLINE unsigned
+nvgl_comparison_op(unsigned op)
+{
+	switch (op) {
+	case PIPE_FUNC_NEVER:
+		return 0x0200;
+	case PIPE_FUNC_LESS:
+		return 0x0201;
+	case PIPE_FUNC_EQUAL:
+		return 0x0202;
+	case PIPE_FUNC_LEQUAL:
+		return 0x0203;
+	case PIPE_FUNC_GREATER:
+		return 0x0204;
+	case PIPE_FUNC_NOTEQUAL:
+		return 0x0205;
+	case PIPE_FUNC_GEQUAL:
+		return 0x0206;
+	case PIPE_FUNC_ALWAYS:
+		return 0x0207;
+	default:
+		return 0x0207;
+	}
+}
+
+static INLINE unsigned
+nvgl_polygon_mode(unsigned mode)
+{
+	switch (mode) {
+	case PIPE_POLYGON_MODE_POINT:
+		return 0x1b00;
+	case PIPE_POLYGON_MODE_LINE:
+		return 0x1b01;
+	case PIPE_POLYGON_MODE_FILL:
+		return 0x1b02;
+	default:
+		return 0x1b02;
+	}
+}
+
+static INLINE unsigned
+nvgl_stencil_op(unsigned op)
+{
+	switch (op) {
+	case PIPE_STENCIL_OP_ZERO:
+		return 0x0000;
+	case PIPE_STENCIL_OP_INVERT:
+		return 0x150a;
+	case PIPE_STENCIL_OP_KEEP:
+		return 0x1e00;
+	case PIPE_STENCIL_OP_REPLACE:
+		return 0x1e01;
+	case PIPE_STENCIL_OP_INCR:
+		return 0x1e02;
+	case PIPE_STENCIL_OP_DECR:
+		return 0x1e03;
+	case PIPE_STENCIL_OP_INCR_WRAP:
+		return 0x8507;
+	case PIPE_STENCIL_OP_DECR_WRAP:
+		return 0x8508;
+	default:
+		return 0x1e00;
+	}
+}
+
+static INLINE unsigned
+nvgl_primitive(unsigned prim) {
+	switch (prim) {
+	case PIPE_PRIM_POINTS:
+		return 0x0001;
+	case PIPE_PRIM_LINES:
+		return 0x0002;
+	case PIPE_PRIM_LINE_LOOP:
+		return 0x0003;
+	case PIPE_PRIM_LINE_STRIP:
+		return 0x0004;
+	case PIPE_PRIM_TRIANGLES:
+		return 0x0005;
+	case PIPE_PRIM_TRIANGLE_STRIP:
+		return 0x0006;
+	case PIPE_PRIM_TRIANGLE_FAN:
+		return 0x0007;
+	case PIPE_PRIM_QUADS:
+		return 0x0008;
+	case PIPE_PRIM_QUAD_STRIP:
+		return 0x0009;
+	case PIPE_PRIM_POLYGON:
+		return 0x000a;
+	default:
+		return 0;
+	}
+}
+
+#endif
diff --git a/src/gallium/drivers/nouveau/nouveau_push.h b/src/gallium/drivers/nouveau/nouveau_push.h
new file mode 100644
index 0000000000..54ef1c1291
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nouveau_push.h
@@ -0,0 +1,82 @@
+#ifndef __NOUVEAU_PUSH_H__
+#define __NOUVEAU_PUSH_H__
+
+#include "nouveau/nouveau_winsys.h"
+
+#ifndef NOUVEAU_PUSH_CONTEXT
+#error undefined push context
+#endif
+
+#define OUT_RING(data) do {                                                    \
+	NOUVEAU_PUSH_CONTEXT(pc);                                              \
+	(*pc->nvws->channel->pushbuf->cur++) = (data);                         \
+} while(0)
+
+#define OUT_RINGp(src,size) do {                                               \
+	NOUVEAU_PUSH_CONTEXT(pc);                                              \
+	memcpy(pc->nvws->channel->pushbuf->cur, (src), (size) * 4);            \
+	pc->nvws->channel->pushbuf->cur += (size);                             \
+} while(0)
+
+#define OUT_RINGf(data) do {                                                   \
+	union { float v; uint32_t u; } c;                                      \
+	c.v = (data);                                                          \
+	OUT_RING(c.u);                                                         \
+} while(0)
+
+#define BEGIN_RING(obj,mthd,size) do {                                         \
+	NOUVEAU_PUSH_CONTEXT(pc);                                              \
+	if (pc->nvws->channel->pushbuf->remaining < ((size) + 1))              \
+		pc->nvws->push_flush(pc->nvws, ((size) + 1), NULL);            \
+	OUT_RING((pc->obj->subc << 13) | ((size) << 18) | (mthd));             \
+	pc->nvws->channel->pushbuf->remaining -= ((size) + 1);                 \
+} while(0)
+
+#define BEGIN_RING_NI(obj,mthd,size) do {                                      \
+	BEGIN_RING(obj, (mthd) | 0x40000000, (size));                          \
+} while(0)
+
+#define FIRE_RING(fence) do {                                                  \
+	NOUVEAU_PUSH_CONTEXT(pc);                                              \
+	pc->nvws->push_flush(pc->nvws, 0, fence);                              \
+} while(0)
+
+#define OUT_RELOC(bo,data,flags,vor,tor) do {                                  \
+	NOUVEAU_PUSH_CONTEXT(pc);                                              \
+	pc->nvws->push_reloc(pc->nvws, pc->nvws->channel->pushbuf->cur++,      \
+			     (bo), (data), (flags), (vor), (tor));             \
+} while(0)
+
+/* Raw data + flags depending on FB/TT buffer */
+#define OUT_RELOCd(bo,data,flags,vor,tor) do {                                 \
+	OUT_RELOC((bo), (data), (flags) | NOUVEAU_BO_OR, (vor), (tor));        \
+} while(0)
+
+/* FB/TT object handle */
+#define OUT_RELOCo(bo,flags) do {                                              \
+	OUT_RELOC((bo), 0, (flags) | NOUVEAU_BO_OR,                            \
+		  pc->nvws->channel->vram->handle,                             \
+		  pc->nvws->channel->gart->handle);                            \
+} while(0)
+
+/* Low 32-bits of offset */
+#define OUT_RELOCl(bo,delta,flags) do {                                        \
+	OUT_RELOC((bo), (delta), (flags) | NOUVEAU_BO_LOW, 0, 0);              \
+} while(0)
+
+/* High 32-bits of offset */
+#define OUT_RELOCh(bo,delta,flags) do {                                        \
+	OUT_RELOC((bo), (delta), (flags) | NOUVEAU_BO_HIGH, 0, 0);             \
+} while(0)
+
+/* A reloc which'll recombine into a NV_DMA_METHOD packet header */
+#define OUT_RELOCm(bo, flags, obj, mthd, size) do {                            \
+	NOUVEAU_PUSH_CONTEXT(pc);                                              \
+	if (pc->nvws->channel->pushbuf->remaining < ((size) + 1))              \
+		pc->nvws->push_flush(pc->nvws->channel, ((size) + 1), NULL);   \
+	OUT_RELOCd((bo), (pc->obj->subc << 13) | ((size) << 18) | (mthd),      \
+		   (flags), 0, 0);                                             \
+	pc->nvws->channel->pushbuf->remaining -= ((size) + 1);                 \
+} while(0)
+
+#endif
diff --git a/src/gallium/drivers/nouveau/nouveau_stateobj.h b/src/gallium/drivers/nouveau/nouveau_stateobj.h
new file mode 100644
index 0000000000..4ae4ff4940
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nouveau_stateobj.h
@@ -0,0 +1,159 @@
+#ifndef __NOUVEAU_STATEOBJ_H__
+#define __NOUVEAU_STATEOBJ_H__
+
+#include "pipe/p_debug.h"
+
+struct nouveau_stateobj_reloc {
+	struct pipe_buffer *bo;
+
+	unsigned offset;
+	unsigned packet;
+
+	unsigned data;
+	unsigned flags;
+	unsigned vor;
+	unsigned tor;
+};
+
+struct nouveau_stateobj {
+	int refcount;
+
+	unsigned *push;
+	struct nouveau_stateobj_reloc *reloc;
+
+	unsigned *cur;
+	unsigned cur_packet;
+	unsigned cur_reloc;
+};
+
+static INLINE struct nouveau_stateobj *
+so_new(unsigned push, unsigned reloc)
+{
+	struct nouveau_stateobj *so;
+
+	so = MALLOC(sizeof(struct nouveau_stateobj));
+	so->refcount = 0;
+	so->push = MALLOC(sizeof(unsigned) * push);
+	so->reloc = MALLOC(sizeof(struct nouveau_stateobj_reloc) * reloc);
+
+	so->cur = so->push;
+	so->cur_reloc = so->cur_packet = 0;
+
+	return so;
+}
+
+static INLINE void
+so_ref(struct nouveau_stateobj *ref, struct nouveau_stateobj **pso)
+{
+	struct nouveau_stateobj *so = *pso;
+
+	if (ref) {
+		ref->refcount++;
+	}
+
+	if (so && --so->refcount <= 0) {
+		free(so->push);
+		free(so->reloc);
+		free(so);
+	}
+
+	*pso = ref;
+}
+
+static INLINE void
+so_data(struct nouveau_stateobj *so, unsigned data)
+{
+	(*so->cur++) = (data);
+	so->cur_packet += 4;
+}
+
+static INLINE void
+so_datap(struct nouveau_stateobj *so, unsigned *data, unsigned size)
+{
+	so->cur_packet += (4 * size);
+	while (size--)
+		(*so->cur++) = (*data++);
+}
+
+static INLINE void
+so_method(struct nouveau_stateobj *so, struct nouveau_grobj *gr,
+	  unsigned mthd, unsigned size)
+{
+	so->cur_packet = (gr->subc << 13) | (1 << 18) | (mthd - 4);
+	so_data(so, (gr->subc << 13) | (size << 18) | mthd);
+}
+
+static INLINE void
+so_reloc(struct nouveau_stateobj *so, struct pipe_buffer *bo,
+	 unsigned data, unsigned flags, unsigned vor, unsigned tor)
+{
+	struct nouveau_stateobj_reloc *r = &so->reloc[so->cur_reloc++];
+	
+	r->bo = bo;
+	r->offset = so->cur - so->push;
+	r->packet = so->cur_packet;
+	r->data = data;
+	r->flags = flags;
+	r->vor = vor;
+	r->tor = tor;
+	so_data(so, data);
+}
+
+static INLINE void
+so_dump(struct nouveau_stateobj *so)
+{
+	unsigned i, nr = so->cur - so->push;
+
+	for (i = 0; i < nr; i++)
+		debug_printf("+0x%04x: 0x%08x\n", i, so->push[i]);
+}
+
+static INLINE void
+so_emit(struct nouveau_winsys *nvws, struct nouveau_stateobj *so)
+{
+	struct nouveau_pushbuf *pb = nvws->channel->pushbuf;
+	unsigned nr, i;
+
+	nr = so->cur - so->push;
+	if (pb->remaining < nr)
+		nvws->push_flush(nvws, nr, NULL);
+	pb->remaining -= nr;
+
+	memcpy(pb->cur, so->push, nr * 4);
+	for (i = 0; i < so->cur_reloc; i++) {
+		struct nouveau_stateobj_reloc *r = &so->reloc[i];
+
+		nvws->push_reloc(nvws, pb->cur + r->offset, r->bo,
+				 r->data, r->flags, r->vor, r->tor);
+	}
+	pb->cur += nr;
+}
+
+static INLINE void
+so_emit_reloc_markers(struct nouveau_winsys *nvws, struct nouveau_stateobj *so)
+{
+	struct nouveau_pushbuf *pb = nvws->channel->pushbuf;
+	unsigned i;
+
+	if (!so)
+		return;
+
+	i = so->cur_reloc << 1;
+	if (nvws->channel->pushbuf->remaining < i)
+		nvws->push_flush(nvws, i, NULL);
+	nvws->channel->pushbuf->remaining -= i;
+
+	for (i = 0; i < so->cur_reloc; i++) {
+		struct nouveau_stateobj_reloc *r = &so->reloc[i];
+
+		nvws->push_reloc(nvws, pb->cur++, r->bo, r->packet,
+				 (r->flags & (NOUVEAU_BO_VRAM |
+					      NOUVEAU_BO_GART |
+					      NOUVEAU_BO_RDWR)) |
+				 NOUVEAU_BO_DUMMY, 0, 0);
+		nvws->push_reloc(nvws, pb->cur++, r->bo, r->data,
+				 r->flags | NOUVEAU_BO_DUMMY, r->vor, r->tor);
+	}
+}
+
+#endif
diff --git a/src/gallium/drivers/nouveau/nouveau_util.h b/src/gallium/drivers/nouveau/nouveau_util.h
new file mode 100644
index 0000000000..a10114beab
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nouveau_util.h
@@ -0,0 +1,91 @@
+#ifndef __NOUVEAU_UTIL_H__
+#define __NOUVEAU_UTIL_H__
+
+/* Determine how many vertices can be pushed into the command stream.
+ * Where the remaining space isn't large enough to represent all verices,
+ * split the buffer at primitive boundaries.
+ *
+ * Returns a count of vertices that can be rendered, and an index to
+ * restart drawing at after a flush.
+ */
+static INLINE unsigned
+nouveau_vbuf_split(unsigned remaining, unsigned overhead, unsigned vpp,
+		   unsigned mode, unsigned start, unsigned count,
+		   unsigned *restart)
+{
+	int max, adj = 0;
+
+	max  = remaining - overhead;
+	if (max < 0)
+		return 0;
+
+	max *= vpp;
+	if (max >= count)
+		return count;
+
+	switch (mode) {
+	case PIPE_PRIM_POINTS:
+		break;
+	case PIPE_PRIM_LINES:
+		max = max & 1;
+		break;
+	case PIPE_PRIM_TRIANGLES:
+		max = max - (max % 3);
+		break;
+	case PIPE_PRIM_QUADS:
+		max = max & 3;
+		break;
+	case PIPE_PRIM_LINE_LOOP:
+	case PIPE_PRIM_LINE_STRIP:
+		if (max < 2)
+			max = 0;
+		adj = 1;
+		break;
+	case PIPE_PRIM_POLYGON:
+	case PIPE_PRIM_TRIANGLE_STRIP:
+	case PIPE_PRIM_TRIANGLE_FAN:
+		if (max < 3)
+			max = 0;
+		adj = 2;
+		break;
+	case PIPE_PRIM_QUAD_STRIP:
+		if (max < 4)
+			max = 0;
+		adj = 3;
+		break;
+	default:
+		assert(0);
+	}
+
+	*restart = start + max - adj;
+	return max;
+}
+
+/* Integer base-2 logarithm, rounded towards zero. */
+static INLINE unsigned log2i(unsigned i)
+{
+	unsigned r = 0;
+
+	if (i & 0xffff0000) {
+		i >>= 16;
+		r += 16;
+	}
+	if (i & 0x0000ff00) {
+		i >>= 8;
+		r += 8;
+	}
+	if (i & 0x000000f0) {
+		i >>= 4;
+		r += 4;
+	}
+	if (i & 0x0000000c) {
+		i >>= 2;
+		r += 2;
+	}
+	if (i & 0x00000002) {
+		r += 1;
+	}
+	return r;
+}
+
+#endif
diff --git a/src/gallium/drivers/nouveau/nouveau_winsys.h b/src/gallium/drivers/nouveau/nouveau_winsys.h
new file mode 100644
index 0000000000..b86c4b9338
--- /dev/null
+++ b/src/gallium/drivers/nouveau/nouveau_winsys.h
@@ -0,0 +1,100 @@
+#ifndef NOUVEAU_WINSYS_H
+#define NOUVEAU_WINSYS_H
+
+#include <stdint.h>
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_defines.h"
+
+#include "nouveau/nouveau_bo.h"
+#include "nouveau/nouveau_channel.h"
+#include "nouveau/nouveau_class.h"
+#include "nouveau/nouveau_device.h"
+#include "nouveau/nouveau_grobj.h"
+#include "nouveau/nouveau_notifier.h"
+#include "nouveau/nouveau_resource.h"
+#include "nouveau/nouveau_pushbuf.h"
+
+#define NOUVEAU_CAP_HW_VTXBUF (0xbeef0000)
+#define NOUVEAU_CAP_HW_IDXBUF (0xbeef0001)
+
+#define NOUVEAU_TEXTURE_USAGE_LINEAR (1 << 16)
+
+#define NOUVEAU_BUFFER_USAGE_TEXTURE (1 << 16)
+#define NOUVEAU_BUFFER_USAGE_ZETA    (1 << 17)
+
+struct nouveau_winsys {
+	struct nouveau_context *nv;
+
+	struct nouveau_channel *channel;
+
+	int  (*res_init)(struct nouveau_resource **heap, unsigned start,
+			 unsigned size);
+	int  (*res_alloc)(struct nouveau_resource *heap, int size, void *priv,
+			  struct nouveau_resource **);
+	void (*res_free)(struct nouveau_resource **);
+
+	int  (*push_reloc)(struct nouveau_winsys *, void *ptr,
+			   struct pipe_buffer *, uint32_t data,
+			   uint32_t flags, uint32_t vor, uint32_t tor);
+	int  (*push_flush)(struct nouveau_winsys *, unsigned size,
+			   struct pipe_fence_handle **fence);
+			       
+	int       (*grobj_alloc)(struct nouveau_winsys *, int grclass,
+				 struct nouveau_grobj **);
+	void      (*grobj_free)(struct nouveau_grobj **);
+
+	int       (*notifier_alloc)(struct nouveau_winsys *, int count,
+				    struct nouveau_notifier **);
+	void      (*notifier_free)(struct nouveau_notifier **);
+	void      (*notifier_reset)(struct nouveau_notifier *, int id);
+	uint32_t  (*notifier_status)(struct nouveau_notifier *, int id);
+	uint32_t  (*notifier_retval)(struct nouveau_notifier *, int id);
+	int       (*notifier_wait)(struct nouveau_notifier *, int id,
+				   int status, double timeout);
+
+	int (*surface_copy)(struct nouveau_winsys *, struct pipe_surface *,
+			    unsigned, unsigned, struct pipe_surface *,
+			    unsigned, unsigned, unsigned, unsigned);
+	int (*surface_fill)(struct nouveau_winsys *, struct pipe_surface *,
+			    unsigned, unsigned, unsigned, unsigned, unsigned);
+
+	struct nouveau_bo *(*get_bo)(struct pipe_buffer *);
+};
+
+extern struct pipe_screen *
+nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+
+extern struct pipe_context *
+nv04_create(struct pipe_screen *, unsigned pctx_id);
+
+extern struct pipe_screen *
+nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+
+extern struct pipe_context *
+nv10_create(struct pipe_screen *, unsigned pctx_id);
+
+extern struct pipe_screen *
+nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+
+extern struct pipe_context *
+nv20_create(struct pipe_screen *, unsigned pctx_id);
+
+extern struct pipe_screen *
+nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+
+extern struct pipe_context *
+nv30_create(struct pipe_screen *, unsigned pctx_id);
+
+extern struct pipe_screen *
+nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+
+extern struct pipe_context *
+nv40_create(struct pipe_screen *, unsigned pctx_id);
+
+extern struct pipe_screen *
+nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *);
+
+extern struct pipe_context *
+nv50_create(struct pipe_screen *, unsigned pctx_id);
+
+#endif
diff --git a/src/gallium/drivers/nv04/Makefile b/src/gallium/drivers/nv04/Makefile
new file mode 100644
index 0000000000..4ed62dae95
--- /dev/null
+++ b/src/gallium/drivers/nv04/Makefile
@@ -0,0 +1,29 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nv04
+
+DRIVER_SOURCES = \
+	nv04_surface_2d.c \
+	nv04_clear.c \
+	nv04_context.c \
+	nv04_fragprog.c \
+	nv04_fragtex.c \
+	nv04_miptree.c \
+	nv04_prim_vbuf.c \
+	nv04_screen.c \
+	nv04_state.c \
+	nv04_state_emit.c \
+	nv04_surface.c \
+	nv04_vbo.c
+
+C_SOURCES = \
+	$(COMMON_SOURCES) \
+	$(DRIVER_SOURCES)
+
+ASM_SOURCES = 
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/drivers/nv04/nv04_clear.c b/src/gallium/drivers/nv04/nv04_clear.c
new file mode 100644
index 0000000000..01cacd36fe
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_clear.c
@@ -0,0 +1,12 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "nv04_context.h"
+
+void
+nv04_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+	   unsigned clearValue)
+{
+	pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue);
+}
diff --git a/src/gallium/drivers/nv04/nv04_context.c b/src/gallium/drivers/nv04/nv04_context.c
new file mode 100644
index 0000000000..d6710cd892
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_context.c
@@ -0,0 +1,107 @@
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+#include "nv04_context.h"
+#include "nv04_screen.h"
+
+static void
+nv04_flush(struct pipe_context *pipe, unsigned flags,
+	   struct pipe_fence_handle **fence)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	draw_flush(nv04->draw);
+
+	FIRE_RING(fence);
+}
+
+static void
+nv04_destroy(struct pipe_context *pipe)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	if (nv04->draw)
+		draw_destroy(nv04->draw);
+
+	FREE(nv04);
+}
+
+static void
+nv04_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield)
+{
+}
+
+static boolean
+nv04_init_hwctx(struct nv04_context *nv04)
+{
+	// requires a valid handle
+//	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_NOTIFY, 1);
+//	OUT_RING(0);
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_NOP, 1);
+	OUT_RING(0);
+
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_CONTROL, 1);
+	OUT_RING(0x40182800);
+//	OUT_RING(1<<20/*no cull*/);
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_BLEND, 1);
+//	OUT_RING(0x24|(1<<6)|(1<<8));
+	OUT_RING(0x120001a4);
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_FORMAT, 1);
+	OUT_RING(0x332213a1);
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_FILTER, 1);
+	OUT_RING(0x11001010);
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_COLORKEY, 1);
+	OUT_RING(0x0);
+//	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_OFFSET, 1);
+//	OUT_RING(SCREEN_OFFSET);
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_FOGCOLOR, 1);
+	OUT_RING(0xff000000);
+
+
+
+	FIRE_RING (NULL);
+	return TRUE;
+}
+
+struct pipe_context *
+nv04_create(struct pipe_screen *pscreen, unsigned pctx_id)
+{
+	struct nv04_screen *screen = nv04_screen(pscreen);
+	struct pipe_winsys *ws = pscreen->winsys;
+	struct nv04_context *nv04;
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nv04 = CALLOC(1, sizeof(struct nv04_context));
+	if (!nv04)
+		return NULL;
+	nv04->screen = screen;
+	nv04->pctx_id = pctx_id;
+
+	nv04->nvws = nvws;
+
+	nv04->pipe.winsys = ws;
+	nv04->pipe.screen = pscreen;
+	nv04->pipe.destroy = nv04_destroy;
+	nv04->pipe.set_edgeflags = nv04_set_edgeflags;
+	nv04->pipe.draw_arrays = nv04_draw_arrays;
+	nv04->pipe.draw_elements = nv04_draw_elements;
+	nv04->pipe.clear = nv04_clear;
+	nv04->pipe.flush = nv04_flush;
+
+	nv04_init_surface_functions(nv04);
+	nv04_init_state_functions(nv04);
+
+	nv04->draw = draw_create();
+	assert(nv04->draw);
+	draw_wide_point_threshold(nv04->draw, 0.0);
+	draw_wide_line_threshold(nv04->draw, 0.0);
+	draw_enable_line_stipple(nv04->draw, FALSE);
+	draw_enable_point_sprites(nv04->draw, FALSE);
+	draw_set_rasterize_stage(nv04->draw, nv04_draw_vbuf_stage(nv04));
+
+	nv04_init_hwctx(nv04);
+
+	return &nv04->pipe;
+}
+
diff --git a/src/gallium/drivers/nv04/nv04_context.h b/src/gallium/drivers/nv04/nv04_context.h
new file mode 100644
index 0000000000..2842b2c90d
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_context.h
@@ -0,0 +1,151 @@
+#ifndef __NV04_CONTEXT_H__
+#define __NV04_CONTEXT_H__
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_compiler.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "draw/draw_vertex.h"
+
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau/nouveau_gldefs.h"
+
+#define NOUVEAU_PUSH_CONTEXT(ctx)                                              \
+	struct nv04_screen *ctx = nv04->screen
+#include "nouveau/nouveau_push.h"
+
+#include "nv04_state.h"
+
+#define NOUVEAU_ERR(fmt, args...) \
+	fprintf(stderr, "%s:%d -  "fmt, __func__, __LINE__, ##args);
+#define NOUVEAU_MSG(fmt, args...) \
+	fprintf(stderr, "nouveau: "fmt, ##args);
+
+#include "nv04_screen.h"
+
+#define NV04_NEW_VERTPROG	(1 << 1)
+#define NV04_NEW_FRAGPROG	(1 << 2)
+#define NV04_NEW_BLEND		(1 << 3)
+#define NV04_NEW_RAST		(1 << 4)
+#define NV04_NEW_CONTROL	(1 << 5)
+#define NV04_NEW_VIEWPORT	(1 << 6)
+#define NV04_NEW_SAMPLER	(1 << 7)
+#define NV04_NEW_FRAMEBUFFER	(1 << 8)
+#define NV04_NEW_VTXARRAYS	(1 << 9)
+
+struct nv04_context {
+	struct pipe_context pipe;
+
+	struct nouveau_winsys *nvws;
+	struct nv04_screen *screen;
+	unsigned pctx_id;
+
+	struct draw_context *draw;
+
+	int chipset;
+	struct nouveau_notifier *sync;
+
+	uint32_t dirty;
+
+	struct nv04_blend_state *blend;
+	struct nv04_sampler_state *sampler[PIPE_MAX_SAMPLERS];
+	struct nv04_fragtex_state fragtex;
+	struct nv04_rasterizer_state *rast;
+	struct nv04_depth_stencil_alpha_state *dsa;
+
+	struct nv04_miptree *tex_miptree[PIPE_MAX_SAMPLERS];
+	unsigned dirty_samplers;
+	unsigned fp_samplers;
+	unsigned vp_samplers;
+
+	uint32_t rt_enable;
+	struct pipe_framebuffer_state *framebuffer;
+	struct pipe_surface *rt;
+	struct pipe_surface *zeta;
+
+	struct {
+		struct pipe_buffer *buffer;
+		uint32_t format;
+	} tex[16];
+
+	unsigned vb_enable;
+	struct {
+		struct pipe_buffer *buffer;
+		unsigned delta;
+	} vb[16];
+
+	float *constbuf[PIPE_SHADER_TYPES][32][4];
+	unsigned constbuf_nr[PIPE_SHADER_TYPES];
+
+	struct vertex_info vertex_info;
+	struct {
+	
+		struct nouveau_resource *exec_heap;
+		struct nouveau_resource *data_heap;
+
+		struct nv04_vertex_program *active;
+
+		struct nv04_vertex_program *current;
+		struct pipe_buffer *constant_buf;
+	} vertprog;
+
+	struct {
+		struct nv04_fragment_program *active;
+
+		struct nv04_fragment_program *current;
+		struct pipe_buffer *constant_buf;
+	} fragprog;
+
+	struct pipe_vertex_buffer  vtxbuf[PIPE_MAX_ATTRIBS];
+	struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS];
+
+	struct pipe_viewport_state viewport;
+};
+
+static INLINE struct nv04_context *
+nv04_context(struct pipe_context *pipe)
+{
+	return (struct nv04_context *)pipe;
+}
+
+extern void nv04_init_state_functions(struct nv04_context *nv04);
+extern void nv04_init_surface_functions(struct nv04_context *nv04);
+extern void nv04_screen_init_miptree_functions(struct pipe_screen *screen);
+
+/* nv04_clear.c */
+extern void nv04_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+		       unsigned clearValue);
+
+/* nv04_draw.c */
+extern struct draw_stage *nv04_draw_render_stage(struct nv04_context *nv04);
+
+/* nv04_fragprog.c */
+extern void nv04_fragprog_bind(struct nv04_context *,
+			       struct nv04_fragment_program *);
+extern void nv04_fragprog_destroy(struct nv04_context *,
+				  struct nv04_fragment_program *);
+
+/* nv04_fragtex.c */
+extern void nv04_fragtex_bind(struct nv04_context *);
+
+/* nv04_prim_vbuf.c */
+struct draw_stage *nv04_draw_vbuf_stage( struct nv04_context *nv04 );
+
+/* nv04_state.c and friends */
+extern void nv04_emit_hw_state(struct nv04_context *nv04);
+extern void nv04_state_tex_update(struct nv04_context *nv04);
+
+/* nv04_vbo.c */
+extern boolean nv04_draw_arrays(struct pipe_context *, unsigned mode,
+				unsigned start, unsigned count);
+extern boolean nv04_draw_elements( struct pipe_context *pipe,
+                    struct pipe_buffer *indexBuffer,
+                    unsigned indexSize,
+                    unsigned prim, unsigned start, unsigned count);
+
+
+#endif
diff --git a/src/gallium/drivers/nv04/nv04_fragprog.c b/src/gallium/drivers/nv04/nv04_fragprog.c
new file mode 100644
index 0000000000..8a2af41fe0
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_fragprog.c
@@ -0,0 +1,21 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+
+#include "nv04_context.h"
+
+void
+nv04_fragprog_bind(struct nv04_context *nv04, struct nv04_fragment_program *fp)
+{
+}
+
+void
+nv04_fragprog_destroy(struct nv04_context *nv04,
+		      struct nv04_fragment_program *fp)
+{
+}
+
diff --git a/src/gallium/drivers/nv04/nv04_fragtex.c b/src/gallium/drivers/nv04/nv04_fragtex.c
new file mode 100644
index 0000000000..21f990fd53
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_fragtex.c
@@ -0,0 +1,73 @@
+#include "nv04_context.h"
+#include "nouveau/nouveau_util.h"
+
+#define _(m,tf)                                                                \
+{                                                                              \
+  PIPE_FORMAT_##m,                                                             \
+  NV04_DX5_TEXTURED_TRIANGLE_FORMAT_COLOR_##tf,                                               \
+}
+
+struct nv04_texture_format {
+	uint	pipe;
+	int     format;
+};
+
+static struct nv04_texture_format
+nv04_texture_formats[] = {
+	_(A8R8G8B8_UNORM, A8R8G8B8),
+	_(X8R8G8B8_UNORM, X8R8G8B8),
+	_(A1R5G5B5_UNORM, A1R5G5B5),
+	_(A4R4G4B4_UNORM, A4R4G4B4),
+	_(L8_UNORM,       Y8      ),
+	_(A8_UNORM,       Y8      ),
+};
+
+static uint32_t
+nv04_fragtex_format(uint pipe_format)
+{
+	struct nv04_texture_format *tf = nv04_texture_formats;
+	int i;
+
+	for (i=0; i< sizeof(nv04_texture_formats)/sizeof(nv04_texture_formats[0]); i++) {
+		if (tf->pipe == pipe_format)
+			return tf->format;
+		tf++;
+	}
+
+	NOUVEAU_ERR("unknown texture format %s\n", pf_name(pipe_format));
+	return 0;
+}
+
+
+static void
+nv04_fragtex_build(struct nv04_context *nv04, int unit)
+{
+	struct nv04_miptree *nv04mt = nv04->tex_miptree[unit];
+	struct pipe_texture *pt = &nv04mt->base;
+
+	switch (pt->target) {
+	case PIPE_TEXTURE_2D:
+		break;
+	default:
+		NOUVEAU_ERR("Unknown target %d\n", pt->target);
+		return;
+	}
+
+	nv04->fragtex.format = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_ZOH_CORNER 
+		| NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ORIGIN_FOH_CORNER
+		| nv04_fragtex_format(pt->format)
+		| ( (pt->last_level + 1) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_MIPMAP_LEVELS_SHIFT )
+		| ( log2i(pt->width[0]) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_U_SHIFT )
+		| ( log2i(pt->height[0]) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_BASE_SIZE_V_SHIFT )
+		| NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP_TO_EDGE
+		| NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_CLAMP_TO_EDGE
+		;
+}
+
+
+void
+nv04_fragtex_bind(struct nv04_context *nv04)
+{
+	nv04_fragtex_build(nv04, 0);
+}
+
diff --git a/src/gallium/drivers/nv04/nv04_miptree.c b/src/gallium/drivers/nv04/nv04_miptree.c
new file mode 100644
index 0000000000..993c5ef5dd
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_miptree.c
@@ -0,0 +1,177 @@
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "nv04_context.h"
+#include "nv04_screen.h"
+
+static void
+nv04_miptree_layout(struct nv04_miptree *nv04mt)
+{
+	struct pipe_texture *pt = &nv04mt->base;
+	uint width = pt->width[0], height = pt->height[0];
+	uint offset = 0;
+	int nr_faces, l;
+
+	nr_faces = 1;
+
+	for (l = 0; l <= pt->last_level; l++) {
+		pt->width[l] = width;
+		pt->height[l] = height;
+
+		pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width);
+		pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height);
+		
+		nv04mt->level[l].pitch = pt->width[0];
+		nv04mt->level[l].pitch = (nv04mt->level[l].pitch + 63) & ~63;
+
+		width  = MAX2(1, width  >> 1);
+		height = MAX2(1, height >> 1);
+	}
+
+	for (l = 0; l <= pt->last_level; l++) {
+
+		nv04mt->level[l].image_offset = offset;
+		offset += nv04mt->level[l].pitch * pt->height[l];
+	}
+
+	nv04mt->total_size = offset;
+}
+
+static struct pipe_texture *
+nv04_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
+{
+	struct pipe_winsys *ws = pscreen->winsys;
+	struct nv04_miptree *mt;
+
+	mt = MALLOC(sizeof(struct nv04_miptree));
+	if (!mt)
+		return NULL;
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = pscreen;
+	mt->shadow_tex = NULL;
+	mt->shadow_surface = NULL;
+
+	//mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+
+	nv04_miptree_layout(mt);
+
+	mt->buffer = ws->buffer_create(ws, 256, PIPE_BUFFER_USAGE_PIXEL |
+						NOUVEAU_BUFFER_USAGE_TEXTURE,
+						mt->total_size);
+	if (!mt->buffer) {
+		printf("failed %d byte alloc\n",mt->total_size);
+		FREE(mt);
+		return NULL;
+	}
+	
+	return &mt->base;
+}
+
+static struct pipe_texture *
+nv04_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
+		     const unsigned *stride, struct pipe_buffer *pb)
+{
+	struct nv04_miptree *mt;
+
+	/* Only supports 2D, non-mipmapped textures for the moment */
+	if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 ||
+	    pt->depth[0] != 1)
+		return NULL;
+
+	mt = CALLOC_STRUCT(nv04_miptree);
+	if (!mt)
+		return NULL;
+
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = pscreen;
+	mt->level[0].pitch = stride[0];
+	mt->level[0].image_offset = CALLOC(1, sizeof(unsigned));
+
+	pipe_buffer_reference(pscreen, &mt->buffer, pb);
+	return &mt->base;
+}
+
+static void
+nv04_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt)
+{
+	struct pipe_texture *pt = *ppt;
+	struct nv04_miptree *mt = (struct nv04_miptree *)pt;
+	int l;
+
+	*ppt = NULL;
+	if (--pt->refcount)
+		return;
+
+	pipe_buffer_reference(pscreen, &mt->buffer, NULL);
+	for (l = 0; l <= pt->last_level; l++) {
+		if (mt->level[l].image_offset)
+			FREE(mt->level[l].image_offset);
+	}
+
+	if (mt->shadow_tex) {
+		assert(mt->shadow_surface);
+		pscreen->tex_surface_release(pscreen, &mt->shadow_surface);
+		nv04_miptree_release(pscreen, &mt->shadow_tex);
+	}
+
+	FREE(mt);
+}
+
+static struct pipe_surface *
+nv04_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
+			 unsigned face, unsigned level, unsigned zslice,
+			 unsigned flags)
+{
+	struct nv04_miptree *nv04mt = (struct nv04_miptree *)pt;
+	struct pipe_surface *ps;
+
+	ps = CALLOC_STRUCT(pipe_surface);
+	if (!ps)
+		return NULL;
+	pipe_texture_reference(&ps->texture, pt);
+	ps->format = pt->format;
+	ps->width = pt->width[level];
+	ps->height = pt->height[level];
+	ps->block = pt->block;
+	ps->nblocksx = pt->nblocksx[level];
+	ps->nblocksy = pt->nblocksy[level];
+	ps->stride = nv04mt->level[level].pitch;
+	ps->usage = flags;
+	ps->status = PIPE_SURFACE_STATUS_DEFINED;
+	ps->refcount = 1;
+	ps->face = face;
+	ps->level = level;
+	ps->zslice = zslice;
+
+	ps->offset = nv04mt->level[level].image_offset;
+
+	return ps;
+}
+
+static void
+nv04_miptree_surface_del(struct pipe_screen *pscreen,
+			 struct pipe_surface **psurface)
+{
+	struct pipe_surface *ps = *psurface;
+
+	*psurface = NULL;
+	if (--ps->refcount > 0)
+		return;
+
+	pipe_texture_reference(&ps->texture, NULL);
+	FREE(ps);
+}
+
+void
+nv04_screen_init_miptree_functions(struct pipe_screen *pscreen)
+{
+	pscreen->texture_create = nv04_miptree_create;
+	pscreen->texture_blanket = nv04_miptree_blanket;
+	pscreen->texture_release = nv04_miptree_release;
+	pscreen->get_tex_surface = nv04_miptree_surface_new;
+	pscreen->tex_surface_release = nv04_miptree_surface_del;
+}
+
diff --git a/src/gallium/drivers/nv04/nv04_prim_vbuf.c b/src/gallium/drivers/nv04/nv04_prim_vbuf.c
new file mode 100644
index 0000000000..18a8872ae3
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_prim_vbuf.c
@@ -0,0 +1,309 @@
+
+#include "pipe/p_debug.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_compiler.h"
+
+#include "draw/draw_vbuf.h"
+
+#include "nv04_context.h"
+#include "nv04_state.h"
+
+#define VERTEX_SIZE 40
+#define VERTEX_BUFFER_SIZE (4096*VERTEX_SIZE) // 4096 vertices of 40 bytes each
+
+/**
+ * Primitive renderer for nv04.
+ */
+struct nv04_vbuf_render {
+	struct vbuf_render base;
+
+	struct nv04_context *nv04;   
+
+	/** Vertex buffer */
+	unsigned char* buffer;
+
+	/** Vertex size in bytes */
+	unsigned vertex_size;
+
+	/** Current primitive */
+	unsigned prim;
+};
+
+
+/**
+ * Basically a cast wrapper.
+ */
+static INLINE struct nv04_vbuf_render *
+nv04_vbuf_render( struct vbuf_render *render )
+{
+	assert(render);
+	return (struct nv04_vbuf_render *)render;
+}
+
+
+static const struct vertex_info *
+nv04_vbuf_render_get_vertex_info( struct vbuf_render *render )
+{
+	struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render);
+	struct nv04_context *nv04 = nv04_render->nv04;
+	return &nv04->vertex_info;
+}
+
+
+static void *
+nv04_vbuf_render_allocate_vertices( struct vbuf_render *render,
+		ushort vertex_size,
+		ushort nr_vertices )
+{
+	struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render);
+
+	nv04_render->buffer = (unsigned char*) MALLOC(VERTEX_BUFFER_SIZE);
+	assert(!nv04_render->buffer);
+
+	return nv04_render->buffer;
+}
+
+
+static boolean 
+nv04_vbuf_render_set_primitive( struct vbuf_render *render, 
+		unsigned prim )
+{
+	struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render);
+
+	if (prim <= PIPE_PRIM_LINE_STRIP)
+		return FALSE;
+
+	nv04_render->prim = prim;
+	return TRUE;
+}
+
+static INLINE void nv04_2triangles(struct nv04_context* nv04, unsigned char* buffer, ushort v0, ushort v1, ushort v2, ushort v3, ushort v4, ushort v5)
+{
+	BEGIN_RING(fahrenheit,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0xA),49);
+	OUT_RINGp(buffer + VERTEX_SIZE * v0,8);
+	OUT_RINGp(buffer + VERTEX_SIZE * v1,8);
+	OUT_RINGp(buffer + VERTEX_SIZE * v2,8);
+	OUT_RINGp(buffer + VERTEX_SIZE * v3,8);
+	OUT_RINGp(buffer + VERTEX_SIZE * v4,8);
+	OUT_RINGp(buffer + VERTEX_SIZE * v5,8);
+	OUT_RING(0xFEDCBA);
+}
+
+static INLINE void nv04_1triangle(struct nv04_context* nv04, unsigned char* buffer, ushort v0, ushort v1, ushort v2)
+{
+	BEGIN_RING(fahrenheit,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0xD),25);
+	OUT_RINGp(buffer + VERTEX_SIZE * v0,8);
+	OUT_RINGp(buffer + VERTEX_SIZE * v1,8);
+	OUT_RINGp(buffer + VERTEX_SIZE * v2,8);
+	OUT_RING(0xFED);
+}
+
+static INLINE void nv04_1quad(struct nv04_context* nv04, unsigned char* buffer, ushort v0, ushort v1, ushort v2, ushort v3)
+{
+	BEGIN_RING(fahrenheit,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0xC),33);
+	OUT_RINGp(buffer + VERTEX_SIZE * v0,8);
+	OUT_RINGp(buffer + VERTEX_SIZE * v1,8);
+	OUT_RINGp(buffer + VERTEX_SIZE * v2,8);
+	OUT_RINGp(buffer + VERTEX_SIZE * v3,8);
+	OUT_RING(0xFECEDC);
+}
+
+static void nv04_vbuf_render_triangles_elts(struct nv04_vbuf_render * render, const ushort * indices, uint nr_indices)
+{
+	unsigned char* buffer = render->buffer;
+	struct nv04_context* nv04 = render->nv04;
+	int i;
+
+	for( i=0; i< nr_indices-5; i+=6)
+		nv04_2triangles(nv04,
+				buffer,
+				indices[i+0],
+				indices[i+1],
+				indices[i+2],
+				indices[i+3],
+				indices[i+4],
+				indices[i+5]
+			       );
+	if (i != nr_indices)
+	{
+		nv04_1triangle(nv04,
+				buffer,
+				indices[i+0],
+				indices[i+1],
+				indices[i+2]
+			       );
+		i+=3;
+	}
+	if (i != nr_indices)
+		NOUVEAU_ERR("Houston, we have lost some vertices\n");
+}
+
+static void nv04_vbuf_render_tri_strip_elts(struct nv04_vbuf_render* render, const ushort* indices, uint nr_indices)
+{
+	const uint32_t striptbl[]={0x321210,0x543432,0x765654,0x987876,0xBA9A98,0xDCBCBA,0xFEDEDC};
+	unsigned char* buffer = render->buffer;
+	struct nv04_context* nv04 = render->nv04;
+	int i,j;
+
+	for(i = 0; i<nr_indices; i+=14) 
+	{
+		int numvert = MIN2(16, nr_indices - i);
+		int numtri = numvert - 2;
+		if (numvert<3)
+			break;
+
+		BEGIN_RING( fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x0), numvert*8 );
+		for(j = 0; j<numvert; j++)
+			OUT_RINGp( buffer + VERTEX_SIZE * indices [i+j], 8 );
+
+		BEGIN_RING_NI( fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE(0), (numtri+1)/2 );
+		for(j = 0; j<numtri/2; j++ )
+			OUT_RING(striptbl[j]);
+		if (numtri%2)
+			OUT_RING(striptbl[numtri/2]&0xFFF);
+	}
+}
+
+static void nv04_vbuf_render_tri_fan_elts(struct nv04_vbuf_render* render, const ushort* indices, uint nr_indices)
+{
+	const uint32_t fantbl[]={0x320210,0x540430,0x760650,0x980870,0xBA0A90,0xDC0CB0,0xFE0ED0};
+	unsigned char* buffer = render->buffer;
+	struct nv04_context* nv04 = render->nv04;
+	int i,j;
+
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x0), 8);
+	OUT_RINGp(buffer + VERTEX_SIZE * indices[0], 8);
+
+	for(i = 1; i<nr_indices; i+=14)
+	{
+		int numvert=MIN2(15, nr_indices - i);
+		int numtri=numvert-2;
+		if (numvert < 3)
+			break;
+
+		BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x1), numvert*8);
+
+		for(j=0;j<numvert;j++)
+			OUT_RINGp( buffer + VERTEX_SIZE * indices[ i+j ], 8 );
+
+		BEGIN_RING_NI(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_DRAWPRIMITIVE(0), (numtri+1)/2);
+		for(j = 0; j<numtri/2; j++)
+			OUT_RING(fantbl[j]);
+		if (numtri%2)
+			OUT_RING(fantbl[numtri/2]&0xFFF);
+	}
+}
+
+static void nv04_vbuf_render_quads_elts(struct nv04_vbuf_render* render, const ushort* indices, uint nr_indices)
+{
+	unsigned char* buffer = render->buffer;
+	struct nv04_context* nv04 = render->nv04;
+	int i;
+
+	for(i = 0; i < nr_indices; i += 4)
+		nv04_1quad(nv04,
+				buffer,
+				indices[i+0],
+				indices[i+1],
+				indices[i+2],
+				indices[i+3]
+			       );
+}
+
+
+static void 
+nv04_vbuf_render_draw( struct vbuf_render *render,
+		const ushort *indices,
+		uint nr_indices)
+{
+	struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render);
+
+	// emit the indices
+	switch( nv04_render->prim )
+	{
+		case PIPE_PRIM_TRIANGLES:
+			nv04_vbuf_render_triangles_elts(nv04_render, indices, nr_indices);
+			break;
+		case PIPE_PRIM_QUAD_STRIP:
+		case PIPE_PRIM_TRIANGLE_STRIP:
+			nv04_vbuf_render_tri_strip_elts(nv04_render, indices, nr_indices);
+			break;
+		case PIPE_PRIM_TRIANGLE_FAN:
+		case PIPE_PRIM_POLYGON:
+			nv04_vbuf_render_tri_fan_elts(nv04_render, indices, nr_indices);
+			break;
+		case PIPE_PRIM_QUADS:
+			nv04_vbuf_render_quads_elts(nv04_render, indices, nr_indices);
+			break;
+		default:
+			NOUVEAU_ERR("You have to implement primitive %d, young padawan\n", nv04_render->prim);
+			break;
+	}
+}
+
+
+static void
+nv04_vbuf_render_release_vertices( struct vbuf_render *render,
+		void *vertices, 
+		unsigned vertex_size,
+		unsigned vertices_used )
+{
+	struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render);
+
+	free(nv04_render->buffer);
+	nv04_render->buffer = NULL;
+}
+
+
+static void
+nv04_vbuf_render_destroy( struct vbuf_render *render )
+{
+	struct nv04_vbuf_render *nv04_render = nv04_vbuf_render(render);
+	FREE(nv04_render);
+}
+
+
+/**
+ * Create a new primitive render.
+ */
+static struct vbuf_render *
+nv04_vbuf_render_create( struct nv04_context *nv04 )
+{
+	struct nv04_vbuf_render *nv04_render = CALLOC_STRUCT(nv04_vbuf_render);
+
+	nv04_render->nv04 = nv04;
+
+	nv04_render->base.max_vertex_buffer_bytes = VERTEX_BUFFER_SIZE;
+	nv04_render->base.max_indices = 65536; 
+	nv04_render->base.get_vertex_info = nv04_vbuf_render_get_vertex_info;
+	nv04_render->base.allocate_vertices = nv04_vbuf_render_allocate_vertices;
+	nv04_render->base.set_primitive = nv04_vbuf_render_set_primitive;
+	nv04_render->base.draw = nv04_vbuf_render_draw;
+	nv04_render->base.release_vertices = nv04_vbuf_render_release_vertices;
+	nv04_render->base.destroy = nv04_vbuf_render_destroy;
+
+	return &nv04_render->base;
+}
+
+
+/**
+ * Create a new primitive vbuf/render stage.
+ */
+struct draw_stage *nv04_draw_vbuf_stage( struct nv04_context *nv04 )
+{
+	struct vbuf_render *render;
+	struct draw_stage *stage;
+
+	render = nv04_vbuf_render_create(nv04);
+	if(!render)
+		return NULL;
+
+	stage = draw_vbuf_stage( nv04->draw, render );
+	if(!stage) {
+		render->destroy(render);
+		return NULL;
+	}
+
+	return stage;
+}
diff --git a/src/gallium/drivers/nv04/nv04_screen.c b/src/gallium/drivers/nv04/nv04_screen.c
new file mode 100644
index 0000000000..9ef38bc244
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_screen.c
@@ -0,0 +1,237 @@
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_simple_screen.h"
+
+#include "nv04_context.h"
+#include "nv04_screen.h"
+
+static const char *
+nv04_screen_get_name(struct pipe_screen *screen)
+{
+	struct nv04_screen *nv04screen = nv04_screen(screen);
+	struct nouveau_device *dev = nv04screen->nvws->channel->device;
+	static char buffer[128];
+
+	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
+	return buffer;
+}
+
+static const char *
+nv04_screen_get_vendor(struct pipe_screen *screen)
+{
+	return "nouveau";
+}
+
+static int
+nv04_screen_get_param(struct pipe_screen *screen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+		return 1;
+	case PIPE_CAP_NPOT_TEXTURES:
+		return 0;
+	case PIPE_CAP_TWO_SIDED_STENCIL:
+		return 0;
+	case PIPE_CAP_GLSL:
+		return 0;
+	case PIPE_CAP_S3TC:
+		return 0;
+	case PIPE_CAP_ANISOTROPIC_FILTER:
+		return 0;
+	case PIPE_CAP_POINT_SPRITE:
+		return 0;
+	case PIPE_CAP_MAX_RENDER_TARGETS:
+		return 1;
+	case PIPE_CAP_OCCLUSION_QUERY:
+		return 0;
+	case PIPE_CAP_TEXTURE_SHADOW_MAP:
+		return 0;
+	case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+		return 10;
+	case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+		return 0;
+	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+		return 0;
+	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
+		return 0;
+	case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
+		return 0;
+	case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
+		return 1;
+	case NOUVEAU_CAP_HW_VTXBUF:
+	case NOUVEAU_CAP_HW_IDXBUF:
+		return 0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0;
+	}
+}
+
+static float
+nv04_screen_get_paramf(struct pipe_screen *screen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_LINE_WIDTH:
+	case PIPE_CAP_MAX_LINE_WIDTH_AA:
+		return 0.0;
+	case PIPE_CAP_MAX_POINT_WIDTH:
+	case PIPE_CAP_MAX_POINT_WIDTH_AA:
+		return 0.0;
+	case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+		return 0.0;
+	case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+		return 0.0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0.0;
+	}
+}
+
+static boolean
+nv04_screen_is_format_supported(struct pipe_screen *screen,
+				enum pipe_format format,
+				enum pipe_texture_target target,
+				unsigned tex_usage, unsigned geom_flags)
+{
+	if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM: 
+		case PIPE_FORMAT_Z16_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	} else {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_X8R8G8B8_UNORM:
+		case PIPE_FORMAT_A1R5G5B5_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM: 
+		case PIPE_FORMAT_L8_UNORM:
+		case PIPE_FORMAT_A8_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	}
+
+	return FALSE;
+}
+
+static void *
+nv04_surface_map(struct pipe_screen *screen, struct pipe_surface *surface,
+		 unsigned flags )
+{
+	void *map;
+	struct nv04_miptree *nv04mt = (struct nv04_miptree *)surface->texture;
+
+	map = pipe_buffer_map(screen, nv04mt->buffer, flags);
+	if (!map)
+		return NULL;
+
+	return map + surface->offset;
+}
+
+static void
+nv04_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface)
+{
+	struct nv04_miptree *nv04mt = (struct nv04_miptree *)surface->texture;
+
+	pipe_buffer_unmap(screen, nv04mt->buffer);
+}
+
+static void
+nv04_screen_destroy(struct pipe_screen *pscreen)
+{
+	struct nv04_screen *screen = nv04_screen(pscreen);
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nvws->notifier_free(&screen->sync);
+	nvws->grobj_free(&screen->fahrenheit);
+	nv04_surface_2d_takedown(&screen->eng2d);
+
+	FREE(pscreen);
+}
+
+static struct pipe_buffer *
+nv04_surface_buffer(struct pipe_surface *surf)
+{
+	struct nv04_miptree *mt = (struct nv04_miptree *)surf->texture;
+
+	return mt->buffer;
+}
+
+struct pipe_screen *
+nv04_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+{
+	struct nv04_screen *screen = CALLOC_STRUCT(nv04_screen);
+	unsigned fahrenheit_class = 0, sub3d_class = 0;
+	unsigned chipset = nvws->channel->device->chipset;
+	int ret;
+
+	if (!screen)
+		return NULL;
+	screen->nvws = nvws;
+
+	if (chipset>=0x20) {
+		fahrenheit_class = 0;
+		sub3d_class = 0;
+	} else if (chipset>=0x10) {
+		fahrenheit_class = NV10_DX5_TEXTURED_TRIANGLE;
+		sub3d_class = NV10_CONTEXT_SURFACES_3D;
+	} else {
+		fahrenheit_class=NV04_DX5_TEXTURED_TRIANGLE;
+		sub3d_class = NV04_CONTEXT_SURFACES_3D;
+	}
+
+	if (!fahrenheit_class) {
+		NOUVEAU_ERR("Unknown nv04 chipset: nv%02x\n", chipset);
+		return NULL;
+	}
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(nvws);
+	screen->eng2d->buf = nv04_surface_buffer;
+
+	/* 3D object */
+	ret = nvws->grobj_alloc(nvws, fahrenheit_class, &screen->fahrenheit);
+	if (ret) {
+		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
+		return NULL;
+	}
+
+	/* 3D surface object */
+	ret = nvws->grobj_alloc(nvws, sub3d_class, &screen->context_surfaces_3d);
+	if (ret) {
+		NOUVEAU_ERR("Error creating 3D surface object: %d\n", ret);
+		return NULL;
+	}
+
+	/* Notifier for sync purposes */
+	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	if (ret) {
+		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
+		nv04_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	screen->pipe.winsys = ws;
+	screen->pipe.destroy = nv04_screen_destroy;
+
+	screen->pipe.get_name = nv04_screen_get_name;
+	screen->pipe.get_vendor = nv04_screen_get_vendor;
+	screen->pipe.get_param = nv04_screen_get_param;
+	screen->pipe.get_paramf = nv04_screen_get_paramf;
+
+	screen->pipe.is_format_supported = nv04_screen_is_format_supported;
+
+	screen->pipe.surface_map = nv04_surface_map;
+	screen->pipe.surface_unmap = nv04_surface_unmap;
+
+	nv04_screen_init_miptree_functions(&screen->pipe);
+	u_simple_screen_init(&screen->pipe);
+
+	return &screen->pipe;
+}
+
diff --git a/src/gallium/drivers/nv04/nv04_screen.h b/src/gallium/drivers/nv04/nv04_screen.h
new file mode 100644
index 0000000000..540aec907b
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_screen.h
@@ -0,0 +1,27 @@
+#ifndef __NV04_SCREEN_H__
+#define __NV04_SCREEN_H__
+
+#include "pipe/p_screen.h"
+#include "nv04_surface_2d.h"
+
+struct nv04_screen {
+	struct pipe_screen pipe;
+
+	struct nouveau_winsys *nvws;
+	unsigned chipset;
+
+	/* HW graphics objects */
+	struct nv04_surface_2d *eng2d;
+	struct nouveau_grobj *fahrenheit;
+	struct nouveau_grobj *context_surfaces_3d;
+	struct nouveau_notifier *sync;
+
+};
+
+static INLINE struct nv04_screen *
+nv04_screen(struct pipe_screen *screen)
+{
+	return (struct nv04_screen *)screen;
+}
+
+#endif
diff --git a/src/gallium/drivers/nv04/nv04_state.c b/src/gallium/drivers/nv04/nv04_state.c
new file mode 100644
index 0000000000..87c635f962
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_state.c
@@ -0,0 +1,458 @@
+#include "draw/draw_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/tgsi_parse.h"
+
+#include "nv04_context.h"
+#include "nv04_state.h"
+
+static void *
+nv04_blend_state_create(struct pipe_context *pipe,
+			const struct pipe_blend_state *cso)
+{
+	struct nv04_blend_state *cb;
+
+	cb = MALLOC(sizeof(struct nv04_blend_state));
+
+	cb->b_enable = cso->blend_enable ? 1 : 0;
+	cb->b_src = ((nvgl_blend_func(cso->alpha_src_factor)<<16) |
+			 (nvgl_blend_func(cso->rgb_src_factor)));
+	cb->b_dst = ((nvgl_blend_func(cso->alpha_dst_factor)<<16) |
+			 (nvgl_blend_func(cso->rgb_dst_factor)));
+	
+
+	return (void *)cb;
+}
+
+static void
+nv04_blend_state_bind(struct pipe_context *pipe, void *blend)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	nv04->blend = (struct nv04_blend_state*)blend;
+
+	nv04->dirty |= NV04_NEW_BLEND;
+}
+
+static void
+nv04_blend_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	free(hwcso);
+}
+
+
+static INLINE unsigned
+wrap_mode(unsigned wrap) {
+	unsigned ret;
+
+	switch (wrap) {
+	case PIPE_TEX_WRAP_REPEAT:
+		ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_REPEAT:
+		ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_MIRRORED_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+		ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP_TO_EDGE;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+		ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP_TO_BORDER;
+		break;
+	case PIPE_TEX_WRAP_CLAMP:
+		ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+	case PIPE_TEX_WRAP_MIRROR_CLAMP:
+	default:
+		NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
+		ret = NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_CLAMP;
+	}
+	return ret >> NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_SHIFT;
+}
+
+static void *
+nv04_sampler_state_create(struct pipe_context *pipe,
+			  const struct pipe_sampler_state *cso)
+{
+
+	struct nv04_sampler_state *ss;
+	uint32_t filter = 0;
+
+	ss = MALLOC(sizeof(struct nv04_sampler_state));
+
+	ss->format = ((wrap_mode(cso->wrap_s) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSU_SHIFT) |
+		    (wrap_mode(cso->wrap_t) << NV04_DX5_TEXTURED_TRIANGLE_FORMAT_ADDRESSV_SHIFT));
+
+	if (cso->max_anisotropy > 1.0) {
+		filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_ANISOTROPIC_MINIFY_ENABLE | NV04_DX5_TEXTURED_TRIANGLE_FILTER_ANISOTROPIC_MAGNIFY_ENABLE;
+	}
+
+	switch (cso->mag_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MAGNIFY_LINEAR;
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MAGNIFY_NEAREST;
+		break;
+	}
+
+	switch (cso->min_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST;
+			break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_LINEAR;
+			break;
+		}
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST;
+		break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV04_DX5_TEXTURED_TRIANGLE_FILTER_MINIFY_NEAREST;
+			break;
+		}
+		break;
+	}
+
+	ss->filter = filter;
+
+	return (void *)ss;
+}
+
+static void
+nv04_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		nv04->sampler[unit] = sampler[unit];
+		nv04->dirty_samplers |= (1 << unit);
+	}
+}
+
+static void
+nv04_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	free(hwcso);
+}
+
+static void
+nv04_set_sampler_texture(struct pipe_context *pipe, unsigned nr,
+			 struct pipe_texture **miptree)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		nv04->tex_miptree[unit] = (struct nv04_miptree *)miptree[unit];
+		nv04->dirty_samplers |= (1 << unit);
+	}
+}
+
+static void *
+nv04_rasterizer_state_create(struct pipe_context *pipe,
+			     const struct pipe_rasterizer_state *cso)
+{
+	struct nv04_rasterizer_state *rs;
+
+	/*XXX: ignored:
+	 * 	scissor
+	 * 	points/lines (no hw support, emulated with tris in gallium)
+	 */
+	rs = MALLOC(sizeof(struct nv04_rasterizer_state));
+
+	rs->blend = cso->flatshade ? NV04_DX5_TEXTURED_TRIANGLE_BLEND_SHADE_MODE_FLAT : NV04_DX5_TEXTURED_TRIANGLE_BLEND_SHADE_MODE_GOURAUD;
+
+	return (void *)rs;
+}
+
+static void
+nv04_rasterizer_state_bind(struct pipe_context *pipe, void *rast)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	nv04->rast = (struct nv04_rasterizer_state*)rast;
+
+	draw_set_rasterizer_state(nv04->draw, (nv04->rast ? nv04->rast->templ : NULL));
+
+	nv04->dirty |= NV04_NEW_RAST | NV04_NEW_BLEND;
+}
+
+static void
+nv04_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	free(hwcso);
+}
+
+static INLINE uint32_t nv04_compare_func(uint32_t f)
+{
+	switch ( f ) {
+		case PIPE_FUNC_NEVER:		return 1;
+		case PIPE_FUNC_LESS:		return 2;
+		case PIPE_FUNC_EQUAL:		return 3;
+		case PIPE_FUNC_LEQUAL:		return 4;
+		case PIPE_FUNC_GREATER:		return 5;
+		case PIPE_FUNC_NOTEQUAL:	return 6;
+		case PIPE_FUNC_GEQUAL:		return 7;
+		case PIPE_FUNC_ALWAYS:		return 8;
+	}
+	NOUVEAU_MSG("Unable to find the function\n");
+	return 0;
+}
+
+static void *
+nv04_depth_stencil_alpha_state_create(struct pipe_context *pipe,
+			const struct pipe_depth_stencil_alpha_state *cso)
+{
+	struct nv04_depth_stencil_alpha_state *hw;
+
+	hw = MALLOC(sizeof(struct nv04_depth_stencil_alpha_state));
+
+	hw->control = float_to_ubyte(cso->alpha.ref_value);
+	hw->control |= ( nv04_compare_func(cso->alpha.func) << NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ALPHA_FUNC_SHIFT );
+	hw->control |= cso->alpha.enabled ? NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ALPHA_TEST_ENABLE : 0;
+	hw->control |= NV04_DX5_TEXTURED_TRIANGLE_CONTROL_ORIGIN;
+	hw->control |= cso->depth.enabled ? (1 << NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_ENABLE_SHIFT) : 0;
+	hw->control |= ( nv04_compare_func(cso->depth.func)<< NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_FUNC_SHIFT );
+	hw->control |= 1 << NV04_DX5_TEXTURED_TRIANGLE_CONTROL_CULL_MODE_SHIFT; // no culling, handled by the draw module
+	hw->control |= NV04_DX5_TEXTURED_TRIANGLE_CONTROL_DITHER_ENABLE;
+	hw->control |= NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_PERSPECTIVE_ENABLE;
+	hw->control |= cso->depth.writemask ? (1 << NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_WRITE_ENABLE_SHIFT) : 0;
+	hw->control |= 1 << NV04_DX5_TEXTURED_TRIANGLE_CONTROL_Z_FORMAT_SHIFT; // integer zbuffer format
+
+	return (void *)hw;
+}
+
+static void
+nv04_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	nv04->dsa = hwcso;
+	nv04->dirty |= NV04_NEW_CONTROL;
+}
+
+static void
+nv04_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	free(hwcso);
+}
+
+static void *
+nv04_vp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *templ)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	return draw_create_vertex_shader(nv04->draw, templ);
+}
+
+static void
+nv04_vp_state_bind(struct pipe_context *pipe, void *shader)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	draw_bind_vertex_shader(nv04->draw, (struct draw_vertex_shader *) shader);
+
+	nv04->dirty |= NV04_NEW_VERTPROG;
+}
+
+static void
+nv04_vp_state_delete(struct pipe_context *pipe, void *shader)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	draw_delete_vertex_shader(nv04->draw, (struct draw_vertex_shader *) shader);
+}
+
+static void *
+nv04_fp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	struct nv04_fragment_program *fp;
+
+	fp = CALLOC(1, sizeof(struct nv04_fragment_program));
+	fp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+
+	return (void *)fp;
+}
+
+static void
+nv04_fp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+	struct nv04_fragment_program *fp = hwcso;
+
+	nv04->fragprog.current = fp;
+	nv04->dirty |= NV04_NEW_FRAGPROG;
+}
+
+static void
+nv04_fp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+	struct nv04_fragment_program *fp = hwcso;
+
+	nv04_fragprog_destroy(nv04, fp);
+	free((void*)fp->pipe.tokens);
+	free(fp);
+}
+
+static void
+nv04_set_blend_color(struct pipe_context *pipe,
+		     const struct pipe_blend_color *bcol)
+{
+}
+
+static void
+nv04_set_clip_state(struct pipe_context *pipe,
+		    const struct pipe_clip_state *clip)
+{
+}
+
+static void
+nv04_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
+			 const struct pipe_constant_buffer *buf )
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+	struct pipe_winsys *ws = pipe->winsys;
+
+	assert(shader < PIPE_SHADER_TYPES);
+	assert(index == 0);
+
+	if (buf) {
+		void *mapped;
+		if (buf->buffer && buf->buffer->size &&
+                    (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
+		{
+			memcpy(nv04->constbuf[shader], mapped, buf->buffer->size);
+			nv04->constbuf_nr[shader] =
+				buf->buffer->size / (4 * sizeof(float));
+			ws->buffer_unmap(ws, buf->buffer);
+		}
+	}
+}
+
+static void
+nv04_set_framebuffer_state(struct pipe_context *pipe,
+			   const struct pipe_framebuffer_state *fb)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+	
+	nv04->framebuffer = (struct pipe_framebuffer_state*)fb;
+
+	nv04->dirty |= NV04_NEW_FRAMEBUFFER;
+}
+static void
+nv04_set_polygon_stipple(struct pipe_context *pipe,
+			 const struct pipe_poly_stipple *stipple)
+{
+	NOUVEAU_ERR("line stipple hahaha\n");
+}
+
+static void
+nv04_set_scissor_state(struct pipe_context *pipe,
+		       const struct pipe_scissor_state *s)
+{
+/*	struct nv04_context *nv04 = nv04_context(pipe);
+
+	// XXX
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_SCISSOR_HORIZ, 2);
+	OUT_RING  (((s->maxx - s->minx) << 16) | s->minx);
+	OUT_RING  (((s->maxy - s->miny) << 16) | s->miny);*/
+}
+
+static void
+nv04_set_viewport_state(struct pipe_context *pipe,
+			const struct pipe_viewport_state *viewport)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	nv04->viewport = *viewport;
+
+	draw_set_viewport_state(nv04->draw, &nv04->viewport);
+}
+
+static void
+nv04_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
+		       const struct pipe_vertex_buffer *buffers)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	memcpy(nv04->vtxbuf, buffers, count * sizeof(buffers[0]));
+	nv04->dirty |= NV04_NEW_VTXARRAYS;
+
+	draw_set_vertex_buffers(nv04->draw, count, buffers);
+}
+
+static void
+nv04_set_vertex_elements(struct pipe_context *pipe, unsigned count,
+			const struct pipe_vertex_element *elements)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+
+	memcpy(nv04->vtxelt, elements, sizeof(*elements) * count);
+	nv04->dirty |= NV04_NEW_VTXARRAYS;
+
+	draw_set_vertex_elements(nv04->draw, count, elements);
+}
+
+void
+nv04_init_state_functions(struct nv04_context *nv04)
+{
+	nv04->pipe.create_blend_state = nv04_blend_state_create;
+	nv04->pipe.bind_blend_state = nv04_blend_state_bind;
+	nv04->pipe.delete_blend_state = nv04_blend_state_delete;
+
+	nv04->pipe.create_sampler_state = nv04_sampler_state_create;
+	nv04->pipe.bind_sampler_states = nv04_sampler_state_bind;
+	nv04->pipe.delete_sampler_state = nv04_sampler_state_delete;
+	nv04->pipe.set_sampler_textures = nv04_set_sampler_texture;
+
+	nv04->pipe.create_rasterizer_state = nv04_rasterizer_state_create;
+	nv04->pipe.bind_rasterizer_state = nv04_rasterizer_state_bind;
+	nv04->pipe.delete_rasterizer_state = nv04_rasterizer_state_delete;
+
+	nv04->pipe.create_depth_stencil_alpha_state = nv04_depth_stencil_alpha_state_create;
+	nv04->pipe.bind_depth_stencil_alpha_state = nv04_depth_stencil_alpha_state_bind;
+	nv04->pipe.delete_depth_stencil_alpha_state = nv04_depth_stencil_alpha_state_delete;
+
+	nv04->pipe.create_vs_state = nv04_vp_state_create;
+	nv04->pipe.bind_vs_state = nv04_vp_state_bind;
+	nv04->pipe.delete_vs_state = nv04_vp_state_delete;
+
+	nv04->pipe.create_fs_state = nv04_fp_state_create;
+	nv04->pipe.bind_fs_state = nv04_fp_state_bind;
+	nv04->pipe.delete_fs_state = nv04_fp_state_delete;
+
+	nv04->pipe.set_blend_color = nv04_set_blend_color;
+	nv04->pipe.set_clip_state = nv04_set_clip_state;
+	nv04->pipe.set_constant_buffer = nv04_set_constant_buffer;
+	nv04->pipe.set_framebuffer_state = nv04_set_framebuffer_state;
+	nv04->pipe.set_polygon_stipple = nv04_set_polygon_stipple;
+	nv04->pipe.set_scissor_state = nv04_set_scissor_state;
+	nv04->pipe.set_viewport_state = nv04_set_viewport_state;
+
+	nv04->pipe.set_vertex_buffers = nv04_set_vertex_buffers;
+	nv04->pipe.set_vertex_elements = nv04_set_vertex_elements;
+}
+
diff --git a/src/gallium/drivers/nv04/nv04_state.h b/src/gallium/drivers/nv04/nv04_state.h
new file mode 100644
index 0000000000..15d4685ec1
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_state.h
@@ -0,0 +1,74 @@
+#ifndef __NV04_STATE_H__
+#define __NV04_STATE_H__
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_scan.h"
+
+struct nv04_blend_state {
+	uint32_t b_enable;
+	uint32_t b_src;
+	uint32_t b_dst;
+};
+
+struct nv04_fragtex_state {
+	uint32_t format;
+};
+
+struct nv04_sampler_state {
+	uint32_t filter;
+	uint32_t format;
+};
+
+struct nv04_depth_stencil_alpha_state {
+	uint32_t control;
+};
+
+struct nv04_rasterizer_state {
+	uint32_t blend;
+
+	const struct pipe_rasterizer_state *templ;
+};
+
+struct nv04_miptree {
+	struct pipe_texture base;
+
+	struct pipe_buffer *buffer;
+	uint total_size;
+
+	struct pipe_texture *shadow_tex;
+	struct pipe_surface *shadow_surface;
+
+	struct {
+		uint pitch;
+		uint image_offset;
+	} level[PIPE_MAX_TEXTURE_LEVELS];
+};
+
+struct nv04_fragment_program_data {
+	unsigned offset;
+	unsigned index;
+};
+
+struct nv04_fragment_program {
+	struct pipe_shader_state pipe;
+	struct tgsi_shader_info info;
+
+	boolean translated;
+	boolean on_hw;
+	unsigned samplers;
+
+	uint32_t *insn;
+	int       insn_len;
+
+	struct nv04_fragment_program_data *consts;
+	unsigned nr_consts;
+
+	struct pipe_buffer *buffer;
+
+	uint32_t fp_control;
+	uint32_t fp_reg_control;
+};
+
+
+
+#endif
diff --git a/src/gallium/drivers/nv04/nv04_state_emit.c b/src/gallium/drivers/nv04/nv04_state_emit.c
new file mode 100644
index 0000000000..bd8ef1adbf
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_state_emit.c
@@ -0,0 +1,223 @@
+#include "nv04_context.h"
+#include "nv04_state.h"
+
+static void nv04_vertex_layout(struct pipe_context* pipe)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+	struct nv04_fragment_program *fp = nv04->fragprog.current;
+	uint32_t src = 0;
+	int i;
+	struct vertex_info vinfo;
+
+	memset(&vinfo, 0, sizeof(vinfo));
+
+	for (i = 0; i < fp->info.num_inputs; i++) {
+		int isn = fp->info.input_semantic_name[i];
+		int isi = fp->info.input_semantic_index[i];
+		switch (isn) {
+			case TGSI_SEMANTIC_POSITION:
+				draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src++);
+				break;
+			case TGSI_SEMANTIC_COLOR:
+				draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src++);
+				break;
+			default:
+			case TGSI_SEMANTIC_GENERIC:
+				draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++);
+				break;
+			case TGSI_SEMANTIC_FOG:
+				draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++);
+				break;
+		}
+	}
+
+	printf("%d vertex input\n",fp->info.num_inputs);
+	draw_compute_vertex_size(&vinfo);
+}
+
+static uint32_t nv04_blend_func(uint32_t f)
+{
+	switch ( f ) {
+		case PIPE_BLENDFACTOR_ZERO:			return 0x1;
+		case PIPE_BLENDFACTOR_ONE:			return 0x2;
+		case PIPE_BLENDFACTOR_SRC_COLOR:		return 0x3;
+		case PIPE_BLENDFACTOR_INV_SRC_COLOR:		return 0x4;
+		case PIPE_BLENDFACTOR_SRC_ALPHA:		return 0x5;
+		case PIPE_BLENDFACTOR_INV_SRC_ALPHA:		return 0x6;
+		case PIPE_BLENDFACTOR_DST_ALPHA:		return 0x7;
+		case PIPE_BLENDFACTOR_INV_DST_ALPHA:		return 0x8;
+		case PIPE_BLENDFACTOR_DST_COLOR:		return 0x9;
+		case PIPE_BLENDFACTOR_INV_DST_COLOR:		return 0xA;
+		case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:	return 0xB;
+	}
+	NOUVEAU_MSG("Unable to find the blend function 0x%x\n",f);
+	return 0;
+}
+
+static void nv04_emit_control(struct nv04_context* nv04)
+{
+	uint32_t control = nv04->dsa->control;
+
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_CONTROL, 1);
+	OUT_RING(control);
+}
+
+static void nv04_emit_blend(struct nv04_context* nv04)
+{
+	uint32_t blend;
+
+	blend=0x4; // texture MODULATE_ALPHA
+	blend|=0x20; // alpha is MSB
+	blend|=(2<<6); // flat shading
+	blend|=(1<<8); // persp correct
+	blend|=(0<<16); // no fog
+	blend|=(nv04->blend->b_enable<<20);
+	blend|=(nv04_blend_func(nv04->blend->b_src)<<24);
+	blend|=(nv04_blend_func(nv04->blend->b_dst)<<28);
+
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_BLEND, 1);
+	OUT_RING(blend);
+}
+
+static void nv04_emit_sampler(struct nv04_context *nv04, int unit)
+{
+	struct nv04_miptree *nv04mt = nv04->tex_miptree[unit];
+	struct pipe_texture *pt = &nv04mt->base;
+
+	BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_OFFSET, 3);
+	OUT_RELOCl(nv04mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD);
+	OUT_RELOCd(nv04mt->buffer, (nv04->fragtex.format | nv04->sampler[unit]->format), NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_OR | NOUVEAU_BO_RD, 1/*VRAM*/,2/*TT*/);
+	OUT_RING(nv04->sampler[unit]->filter);
+}
+
+static void nv04_state_emit_framebuffer(struct nv04_context* nv04)
+{
+	struct pipe_framebuffer_state* fb = nv04->framebuffer;
+	struct pipe_surface *rt, *zeta;
+	uint32_t rt_format, w, h;
+	int colour_format = 0, zeta_format = 0;
+	struct nv04_miptree *nv04mt = 0;
+
+	w = fb->cbufs[0]->width;
+	h = fb->cbufs[0]->height;
+	colour_format = fb->cbufs[0]->format;
+	rt = fb->cbufs[0];
+
+	if (fb->zsbuf) {
+		if (colour_format) {
+			assert(w == fb->zsbuf->width);
+			assert(h == fb->zsbuf->height);
+		} else {
+			w = fb->zsbuf->width;
+			h = fb->zsbuf->height;
+		}
+
+		zeta_format = fb->zsbuf->format;
+		zeta = fb->zsbuf;
+	}
+
+	switch (colour_format) {
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case 0:
+		rt_format = 0x108;
+		break;
+	case PIPE_FORMAT_R5G6B5_UNORM:
+		rt_format = 0x103;
+		break;
+	default:
+		assert(0);
+	}
+
+	BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_FORMAT, 1);
+	OUT_RING(rt_format);
+
+	nv04mt = (struct nv04_miptree *)rt->texture;
+	/* FIXME pitches have to be aligned ! */
+	BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_PITCH, 2);
+	OUT_RING(rt->stride|(zeta->stride<<16));
+	OUT_RELOCl(nv04mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	if (fb->zsbuf) {
+		nv04mt = (struct nv04_miptree *)zeta->texture;
+		BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_OFFSET_ZETA, 1);
+		OUT_RELOCl(nv04mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	}
+}
+
+void
+nv04_emit_hw_state(struct nv04_context *nv04)
+{
+	int i;
+
+	if (nv04->dirty & NV04_NEW_VERTPROG) {
+		//nv04_vertprog_bind(nv04, nv04->vertprog.current);
+		nv04->dirty &= ~NV04_NEW_VERTPROG;
+	}
+
+	if (nv04->dirty & NV04_NEW_FRAGPROG) {
+		nv04_fragprog_bind(nv04, nv04->fragprog.current);
+		nv04->dirty &= ~NV04_NEW_FRAGPROG;
+		nv04->dirty_samplers |= (1<<10);
+		nv04->dirty_samplers = 0;
+	}
+
+	if (nv04->dirty & NV04_NEW_CONTROL) {
+		nv04->dirty &= ~NV04_NEW_CONTROL;
+
+		BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_CONTROL, 1);
+		OUT_RING(nv04->dsa->control);
+	}
+
+	if (nv04->dirty & NV04_NEW_BLEND) {
+		nv04->dirty &= ~NV04_NEW_BLEND;
+
+		nv04_emit_blend(nv04);
+	}
+
+	if (nv04->dirty & NV04_NEW_VTXARRAYS) {
+		nv04->dirty &= ~NV04_NEW_VTXARRAYS;
+		nv04_vertex_layout(nv04);
+	}
+
+	if (nv04->dirty & NV04_NEW_SAMPLER) {
+		nv04->dirty &= ~NV04_NEW_SAMPLER;
+
+		nv04_emit_sampler(nv04, 0);
+	}
+
+	if (nv04->dirty & NV04_NEW_VIEWPORT) {
+		nv04->dirty &= ~NV04_NEW_VIEWPORT;
+//		nv04_state_emit_viewport(nv04);
+	}
+
+ 	if (nv04->dirty & NV04_NEW_FRAMEBUFFER) {
+		nv04->dirty &= ~NV04_NEW_FRAMEBUFFER;
+		nv04_state_emit_framebuffer(nv04);
+	}
+
+	/* Emit relocs for every referenced buffer.
+	 * This is to ensure the bufmgr has an accurate idea of how
+	 * the buffer is used.  This isn't very efficient, but we don't
+	 * seem to take a significant performance hit.  Will be improved
+	 * at some point.  Vertex arrays are emitted by nv04_vbo.c
+	 */
+
+	/* Render target */
+	BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_PITCH, 2);
+	OUT_RING(nv04->rt->stride|(nv04->zeta->stride<<16));
+	OUT_RELOCl(nv04->rt, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	if (nv04->zeta) {
+		BEGIN_RING(context_surfaces_3d, NV04_CONTEXT_SURFACES_3D_OFFSET_ZETA, 1);
+		OUT_RELOCl(nv04->zeta, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	}
+
+	/* Texture images */
+	for (i = 0; i < 1; i++) {
+		if (!(nv04->fp_samplers & (1 << i)))
+			continue;
+		struct nv04_miptree *nv04mt = nv04->tex_miptree[i];
+		BEGIN_RING(fahrenheit, NV04_DX5_TEXTURED_TRIANGLE_OFFSET, 2);
+		OUT_RELOCl(nv04mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD);
+		OUT_RELOCd(nv04mt->buffer, (nv04->fragtex.format | nv04->sampler[i]->format), NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_OR | NOUVEAU_BO_RD, 1/*VRAM*/,2/*TT*/);
+	}
+}
+
diff --git a/src/gallium/drivers/nv04/nv04_surface.c b/src/gallium/drivers/nv04/nv04_surface.c
new file mode 100644
index 0000000000..14abf16679
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_surface.c
@@ -0,0 +1,72 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "nv04_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+
+static void
+nv04_surface_copy(struct pipe_context *pipe, boolean do_flip,
+		  struct pipe_surface *dest, unsigned destx, unsigned desty,
+		  struct pipe_surface *src, unsigned srcx, unsigned srcy,
+		  unsigned width, unsigned height)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+	struct nv04_surface_2d *eng2d = nv04->screen->eng2d;
+
+	if (do_flip) {
+		desty += height;
+		while (height--) {
+			eng2d->copy(eng2d, dest, destx, desty--, src,
+				    srcx, srcy++, width, 1);
+		}
+		return;
+	}
+
+	eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height);
+}
+
+static void
+nv04_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
+		  unsigned destx, unsigned desty, unsigned width,
+		  unsigned height, unsigned value)
+{
+	struct nv04_context *nv04 = nv04_context(pipe);
+	struct nv04_surface_2d *eng2d = nv04->screen->eng2d;
+
+	eng2d->fill(eng2d, dest, destx, desty, width, height, value);
+}
+
+void
+nv04_init_surface_functions(struct nv04_context *nv04)
+{
+	nv04->pipe.surface_copy = nv04_surface_copy;
+	nv04->pipe.surface_fill = nv04_surface_fill;
+}
diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.c b/src/gallium/drivers/nv04/nv04_surface_2d.c
new file mode 100644
index 0000000000..230cfd17dd
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_surface_2d.c
@@ -0,0 +1,448 @@
+#include "pipe/p_context.h"
+#include "pipe/p_format.h"
+#include "util/u_memory.h"
+
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau/nouveau_util.h"
+#include "nv04_surface_2d.h"
+
+static INLINE int
+nv04_surface_format(enum pipe_format format)
+{
+	switch (format) {
+	case PIPE_FORMAT_A8_UNORM:
+		return NV04_CONTEXT_SURFACES_2D_FORMAT_Y8;
+	case PIPE_FORMAT_R16_SNORM:
+	case PIPE_FORMAT_R5G6B5_UNORM:
+		return NV04_CONTEXT_SURFACES_2D_FORMAT_R5G6B5;
+	case PIPE_FORMAT_X8R8G8B8_UNORM:
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+		return NV04_CONTEXT_SURFACES_2D_FORMAT_A8R8G8B8;
+	case PIPE_FORMAT_Z24S8_UNORM:
+		return NV04_CONTEXT_SURFACES_2D_FORMAT_Y32;
+	default:
+		return -1;
+	}
+}
+
+static INLINE int
+nv04_rect_format(enum pipe_format format)
+{
+	switch (format) {
+	case PIPE_FORMAT_A8_UNORM:
+		return NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A8R8G8B8;
+	case PIPE_FORMAT_R5G6B5_UNORM:
+		return NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A16R5G6B5;
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case PIPE_FORMAT_Z24S8_UNORM:
+		return NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT_A8R8G8B8;
+	default:
+		return -1;
+	}
+}
+
+static INLINE int
+nv04_scaled_image_format(enum pipe_format format)
+{
+	switch (format) {
+	case PIPE_FORMAT_A1R5G5B5_UNORM:
+		return NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_A1R5G5B5;
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+		return NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_A8R8G8B8;
+	case PIPE_FORMAT_X8R8G8B8_UNORM:
+		return NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_X8R8G8B8;
+	case PIPE_FORMAT_R5G6B5_UNORM:
+	case PIPE_FORMAT_R16_SNORM:
+		return NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT_R5G6B5;
+	default:
+		return -1;
+	}
+}
+
+static INLINE unsigned
+nv04_swizzle_bits(unsigned x, unsigned y)
+{
+	unsigned u = (x & 0x001) << 0 |
+	             (x & 0x002) << 1 |
+	             (x & 0x004) << 2 |
+	             (x & 0x008) << 3 |
+	             (x & 0x010) << 4 |
+	             (x & 0x020) << 5 |
+	             (x & 0x040) << 6 |
+	             (x & 0x080) << 7 |
+	             (x & 0x100) << 8 |
+	             (x & 0x200) << 9 |
+	             (x & 0x400) << 10 |
+	             (x & 0x800) << 11;
+
+	unsigned v = (y & 0x001) << 1 |
+	             (y & 0x002) << 2 |
+	             (y & 0x004) << 3 |
+	             (y & 0x008) << 4 |
+	             (y & 0x010) << 5 |
+	             (y & 0x020) << 6 |
+	             (y & 0x040) << 7 |
+	             (y & 0x080) << 8 |
+	             (y & 0x100) << 9 |
+	             (y & 0x200) << 10 |
+	             (y & 0x400) << 11 |
+	             (y & 0x800) << 12;
+	return v | u;
+}
+
+static int
+nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx,
+			  struct pipe_surface *dst, int dx, int dy,
+			  struct pipe_surface *src, int sx, int sy,
+			  int w, int h)
+{
+	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_grobj *swzsurf = ctx->swzsurf;
+	struct nouveau_grobj *sifm = ctx->sifm;
+	struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src));
+	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	const unsigned max_w = 1024;
+	const unsigned max_h = 1024;
+	const unsigned sub_w = w > max_w ? max_w : w;
+	const unsigned sub_h = h > max_h ? max_h : h;
+	unsigned cx;
+	unsigned cy;
+
+	/* POT or GTFO */
+	assert(!(w & (w - 1)) && !(h & (h - 1)));
+
+	BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_DMA_IMAGE, 1);
+	OUT_RELOCo(chan, dst_bo,
+	                 NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+
+	BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_FORMAT, 1);
+	OUT_RING  (chan, nv04_surface_format(dst->format) |
+	                 log2i(w) << NV04_SWIZZLED_SURFACE_FORMAT_BASE_SIZE_U_SHIFT |
+	                 log2i(h) << NV04_SWIZZLED_SURFACE_FORMAT_BASE_SIZE_V_SHIFT);
+ 
+	BEGIN_RING(chan, sifm, NV04_SCALED_IMAGE_FROM_MEMORY_DMA_IMAGE, 1);
+	OUT_RELOCo(chan, src_bo,
+	                 NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+	BEGIN_RING(chan, sifm, NV04_SCALED_IMAGE_FROM_MEMORY_SURFACE, 1);
+	OUT_RING  (chan, swzsurf->handle);
+
+	for (cy = 0; cy < h; cy += sub_h) {
+	  for (cx = 0; cx < w; cx += sub_w) {
+	    BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_OFFSET, 1);
+	    OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(cx, cy) *
+			     dst->block.size, NOUVEAU_BO_GART |
+			     NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+
+	    BEGIN_RING(chan, sifm, NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION, 9);
+	    OUT_RING  (chan, NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION_TRUNCATE);
+	    OUT_RING  (chan, nv04_scaled_image_format(src->format));
+	    OUT_RING  (chan, NV04_SCALED_IMAGE_FROM_MEMORY_OPERATION_SRCCOPY);
+	    OUT_RING  (chan, 0);
+	    OUT_RING  (chan, sub_h << 16 | sub_w);
+	    OUT_RING  (chan, 0);
+	    OUT_RING  (chan, sub_h << 16 | sub_w);
+	    OUT_RING  (chan, 1 << 20);
+	    OUT_RING  (chan, 1 << 20);
+
+	    BEGIN_RING(chan, sifm, NV04_SCALED_IMAGE_FROM_MEMORY_SIZE, 4);
+	    OUT_RING  (chan, sub_h << 16 | sub_w);
+	    OUT_RING  (chan, src->stride |
+			     NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_CENTER |
+			     NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_POINT_SAMPLE);
+	    OUT_RELOCl(chan, src_bo, src->offset + cy * src->stride +
+			     cx * src->block.size, NOUVEAU_BO_GART |
+			     NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+	    OUT_RING  (chan, 0);
+	  }
+	}
+
+	return 0;
+}
+
+static int
+nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx,
+		       struct pipe_surface *dst, int dx, int dy,
+		       struct pipe_surface *src, int sx, int sy, int w, int h)
+{
+	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_grobj *m2mf = ctx->m2mf;
+	struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src));
+	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	unsigned dst_offset, src_offset;
+
+	dst_offset = dst->offset + (dy * dst->stride) + (dx * dst->block.size);
+	src_offset = src->offset + (sy * src->stride) + (sx * src->block.size);
+
+	WAIT_RING (chan, 3 + ((h / 2047) + 1) * 9);
+	BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_BUFFER_IN, 2);
+	OUT_RELOCo(chan, src_bo,
+		   NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+	OUT_RELOCo(chan, dst_bo,
+		   NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+
+	while (h) {
+		int count = (h > 2047) ? 2047 : h;
+
+		BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 8);
+		OUT_RELOCl(chan, src_bo, src_offset,
+			   NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD);
+		OUT_RELOCl(chan, dst_bo, dst_offset,
+			   NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_WR);
+		OUT_RING  (chan, src->stride);
+		OUT_RING  (chan, dst->stride);
+		OUT_RING  (chan, w * src->block.size);
+		OUT_RING  (chan, count);
+		OUT_RING  (chan, 0x0101);
+		OUT_RING  (chan, 0);
+
+		h -= count;
+		src_offset += src->stride * count;
+		dst_offset += dst->stride * count;
+	}
+
+	return 0;
+}
+
+static int
+nv04_surface_copy_blit(struct nv04_surface_2d *ctx, struct pipe_surface *dst,
+		       int dx, int dy, struct pipe_surface *src, int sx, int sy,
+		       int w, int h)
+{
+	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_grobj *surf2d = ctx->surf2d;
+	struct nouveau_grobj *blit = ctx->blit;
+	struct nouveau_bo *src_bo = ctx->nvws->get_bo(ctx->buf(src));
+	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	int format;
+
+	format = nv04_surface_format(dst->format);
+	if (format < 0)
+		return 1;
+
+	WAIT_RING (chan, 12);
+	BEGIN_RING(chan, surf2d, NV04_CONTEXT_SURFACES_2D_DMA_IMAGE_SOURCE, 2);
+	OUT_RELOCo(chan, src_bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+	OUT_RELOCo(chan, dst_bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	BEGIN_RING(chan, surf2d, NV04_CONTEXT_SURFACES_2D_FORMAT, 4);
+	OUT_RING  (chan, format);
+	OUT_RING  (chan, (dst->stride << 16) | src->stride);
+	OUT_RELOCl(chan, src_bo, src->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
+	OUT_RELOCl(chan, dst_bo, dst->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+
+	BEGIN_RING(chan, blit, 0x0300, 3);
+	OUT_RING  (chan, (sy << 16) | sx);
+	OUT_RING  (chan, (dy << 16) | dx);
+	OUT_RING  (chan, ( h << 16) |  w);
+
+	return 0;
+}
+
+static void
+nv04_surface_copy(struct nv04_surface_2d *ctx, struct pipe_surface *dst,
+		  int dx, int dy, struct pipe_surface *src, int sx, int sy,
+		  int w, int h)
+{
+	int src_linear = src->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR;
+	int dst_linear = dst->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR;
+
+	assert(src->format == dst->format);
+
+	/* Setup transfer to swizzle the texture to vram if needed */
+	if (src_linear && !dst_linear && w > 1 && h > 1) {
+		nv04_surface_copy_swizzle(ctx, dst, dx, dy, src, sx, sy, w, h);
+		return;
+	}
+
+	/* NV_CONTEXT_SURFACES_2D has buffer alignment restrictions, fallback
+	 * to NV_MEMORY_TO_MEMORY_FORMAT in this case.
+	 */
+	if ((src->offset & 63) || (dst->offset & 63) ||
+	    (src->stride & 63) || (dst->stride & 63)) {
+		nv04_surface_copy_m2mf(ctx, dst, dx, dy, src, sx, sy, w, h);
+		return;
+	}
+
+	nv04_surface_copy_blit(ctx, dst, dx, dy, src, sx, sy, w, h);
+}
+
+static void
+nv04_surface_fill(struct nv04_surface_2d *ctx, struct pipe_surface *dst,
+		  int dx, int dy, int w, int h, unsigned value)
+{
+	struct nouveau_channel *chan = ctx->nvws->channel;
+	struct nouveau_grobj *surf2d = ctx->surf2d;
+	struct nouveau_grobj *rect = ctx->rect;
+	struct nouveau_bo *dst_bo = ctx->nvws->get_bo(ctx->buf(dst));
+	int cs2d_format, gdirect_format;
+
+	cs2d_format = nv04_surface_format(dst->format);
+	assert(cs2d_format >= 0);
+
+	gdirect_format = nv04_rect_format(dst->format);
+	assert(gdirect_format >= 0);
+
+	WAIT_RING (chan, 16);
+	BEGIN_RING(chan, surf2d, NV04_CONTEXT_SURFACES_2D_DMA_IMAGE_SOURCE, 2);
+	OUT_RELOCo(chan, dst_bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	OUT_RELOCo(chan, dst_bo, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	BEGIN_RING(chan, surf2d, NV04_CONTEXT_SURFACES_2D_FORMAT, 4);
+	OUT_RING  (chan, cs2d_format);
+	OUT_RING  (chan, (dst->stride << 16) | dst->stride);
+	OUT_RELOCl(chan, dst_bo, dst->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	OUT_RELOCl(chan, dst_bo, dst->offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+
+	BEGIN_RING(chan, rect, NV04_GDI_RECTANGLE_TEXT_COLOR_FORMAT, 1);
+	OUT_RING  (chan, gdirect_format);
+	BEGIN_RING(chan, rect, NV04_GDI_RECTANGLE_TEXT_COLOR1_A, 1);
+	OUT_RING  (chan, value);
+	BEGIN_RING(chan, rect,
+		   NV04_GDI_RECTANGLE_TEXT_UNCLIPPED_RECTANGLE_POINT(0), 2);
+	OUT_RING  (chan, (dx << 16) | dy);
+	OUT_RING  (chan, ( w << 16) |  h);
+}
+
+void
+nv04_surface_2d_takedown(struct nv04_surface_2d **pctx)
+{
+	struct nv04_surface_2d *ctx;
+
+	if (!pctx || !*pctx)
+		return;
+	ctx = *pctx;
+	*pctx = NULL;
+
+	nouveau_notifier_free(&ctx->ntfy);
+	nouveau_grobj_free(&ctx->m2mf);
+	nouveau_grobj_free(&ctx->surf2d);
+	nouveau_grobj_free(&ctx->swzsurf);
+	nouveau_grobj_free(&ctx->rect);
+	nouveau_grobj_free(&ctx->blit);
+	nouveau_grobj_free(&ctx->sifm);
+
+	FREE(ctx);
+}
+
+struct nv04_surface_2d *
+nv04_surface_2d_init(struct nouveau_winsys *nvws)
+{
+	struct nv04_surface_2d *ctx = CALLOC_STRUCT(nv04_surface_2d);
+	struct nouveau_channel *chan = nvws->channel;
+	unsigned handle = 0x88000000, class;
+	int ret;
+
+	if (!ctx)
+		return NULL;
+
+	ret = nouveau_notifier_alloc(chan, handle++, 1, &ctx->ntfy);
+	if (ret) {
+		nv04_surface_2d_takedown(&ctx);
+		return NULL;
+	}
+
+	ret = nouveau_grobj_alloc(chan, handle++, 0x0039, &ctx->m2mf);
+	if (ret) {
+		nv04_surface_2d_takedown(&ctx);
+		return NULL;
+	}
+
+	BEGIN_RING(chan, ctx->m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY, 1);
+	OUT_RING  (chan, ctx->ntfy->handle);
+
+	if (chan->device->chipset < 0x10)
+		class = NV04_CONTEXT_SURFACES_2D;
+	else
+		class = NV10_CONTEXT_SURFACES_2D;
+
+	ret = nouveau_grobj_alloc(chan, handle++, class, &ctx->surf2d);
+	if (ret) {
+		nv04_surface_2d_takedown(&ctx);
+		return NULL;
+	}
+
+	BEGIN_RING(chan, ctx->surf2d,
+			 NV04_CONTEXT_SURFACES_2D_DMA_IMAGE_SOURCE, 2);
+	OUT_RING  (chan, chan->vram->handle);
+	OUT_RING  (chan, chan->vram->handle);
+
+	if (chan->device->chipset < 0x10)
+		class = NV04_IMAGE_BLIT;
+	else
+		class = NV12_IMAGE_BLIT;
+
+	ret = nouveau_grobj_alloc(chan, handle++, class, &ctx->blit);
+	if (ret) {
+		nv04_surface_2d_takedown(&ctx);
+		return NULL;
+	}
+
+	BEGIN_RING(chan, ctx->blit, NV04_IMAGE_BLIT_DMA_NOTIFY, 1);
+	OUT_RING  (chan, ctx->ntfy->handle);
+	BEGIN_RING(chan, ctx->blit, NV04_IMAGE_BLIT_SURFACE, 1);
+	OUT_RING  (chan, ctx->surf2d->handle);
+	BEGIN_RING(chan, ctx->blit, NV04_IMAGE_BLIT_OPERATION, 1);
+	OUT_RING  (chan, NV04_IMAGE_BLIT_OPERATION_SRCCOPY);
+
+	ret = nouveau_grobj_alloc(chan, handle++, NV04_GDI_RECTANGLE_TEXT,
+				  &ctx->rect);
+	if (ret) {
+		nv04_surface_2d_takedown(&ctx);
+		return NULL;
+	}
+
+	BEGIN_RING(chan, ctx->rect, NV04_GDI_RECTANGLE_TEXT_DMA_NOTIFY, 1);
+	OUT_RING  (chan, ctx->ntfy->handle);
+	BEGIN_RING(chan, ctx->rect, NV04_GDI_RECTANGLE_TEXT_SURFACE, 1);
+	OUT_RING  (chan, ctx->surf2d->handle);
+	BEGIN_RING(chan, ctx->rect, NV04_GDI_RECTANGLE_TEXT_OPERATION, 1);
+	OUT_RING  (chan, NV04_GDI_RECTANGLE_TEXT_OPERATION_SRCCOPY);
+	BEGIN_RING(chan, ctx->rect,
+			 NV04_GDI_RECTANGLE_TEXT_MONOCHROME_FORMAT, 1);
+	OUT_RING  (chan, NV04_GDI_RECTANGLE_TEXT_MONOCHROME_FORMAT_LE);
+
+	switch (chan->device->chipset & 0xf0) {
+	case 0x00:
+	case 0x10:
+		class = NV04_SWIZZLED_SURFACE;
+		break;
+	case 0x20:
+		class = NV20_SWIZZLED_SURFACE;
+		break;
+	case 0x30:
+		class = NV30_SWIZZLED_SURFACE;
+		break;
+	case 0x40:
+	case 0x60:
+		class = NV40_SWIZZLED_SURFACE;
+		break;
+	default:
+		/* Famous last words: this really can't happen.. */
+		assert(0);
+		break;
+	}
+
+	ret = nouveau_grobj_alloc(chan, handle++, class, &ctx->swzsurf);
+	if (ret) {
+		nv04_surface_2d_takedown(&ctx);
+		return NULL;
+	}
+
+	if (chan->device->chipset < 0x10) {
+		class = NV04_SCALED_IMAGE_FROM_MEMORY;
+	} else
+	if (chan->device->chipset < 0x40) {
+		class = NV10_SCALED_IMAGE_FROM_MEMORY;
+	} else {
+		class = NV40_SCALED_IMAGE_FROM_MEMORY;
+	}
+
+	ret = nouveau_grobj_alloc(chan, handle++, class, &ctx->sifm);
+	if (ret) {
+		nv04_surface_2d_takedown(&ctx);
+		return NULL;
+	}
+
+	ctx->nvws = nvws;
+	ctx->copy = nv04_surface_copy;
+	ctx->fill = nv04_surface_fill;
+	return ctx;
+}
diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.h b/src/gallium/drivers/nv04/nv04_surface_2d.h
new file mode 100644
index 0000000000..21b8f86960
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_surface_2d.h
@@ -0,0 +1,29 @@
+#ifndef __NV04_SURFACE_2D_H__
+#define __NV04_SURFACE_2D_H__
+
+struct nv04_surface_2d {
+	struct nouveau_winsys *nvws;
+	struct nouveau_notifier *ntfy;
+	struct nouveau_grobj *surf2d;
+	struct nouveau_grobj *swzsurf;
+	struct nouveau_grobj *m2mf;
+	struct nouveau_grobj *rect;
+	struct nouveau_grobj *blit;
+	struct nouveau_grobj *sifm;
+
+	struct pipe_buffer *(*buf)(struct pipe_surface *);
+
+	void (*copy)(struct nv04_surface_2d *, struct pipe_surface *dst,
+		     int dx, int dy, struct pipe_surface *src, int sx, int sy,
+		     int w, int h);
+	void (*fill)(struct nv04_surface_2d *, struct pipe_surface *dst,
+		     int dx, int dy, int w, int h, unsigned value);
+};
+
+struct nv04_surface_2d *
+nv04_surface_2d_init(struct nouveau_winsys *nvws);
+
+void
+nv04_surface_2d_takedown(struct nv04_surface_2d **);
+
+#endif
diff --git a/src/gallium/drivers/nv04/nv04_vbo.c b/src/gallium/drivers/nv04/nv04_vbo.c
new file mode 100644
index 0000000000..d21a0e34f7
--- /dev/null
+++ b/src/gallium/drivers/nv04/nv04_vbo.c
@@ -0,0 +1,78 @@
+#include "draw/draw_context.h"
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+
+#include "nv04_context.h"
+#include "nv04_state.h"
+
+#include "nouveau/nouveau_channel.h"
+#include "nouveau/nouveau_pushbuf.h"
+
+boolean nv04_draw_elements( struct pipe_context *pipe,
+                    struct pipe_buffer *indexBuffer,
+                    unsigned indexSize,
+                    unsigned prim, unsigned start, unsigned count)
+{
+	struct nv04_context *nv04 = nv04_context( pipe );
+	struct draw_context *draw = nv04->draw;
+	unsigned i;
+
+	nv04_emit_hw_state(nv04);
+
+	/*
+	 * Map vertex buffers
+	 */
+	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
+		if (nv04->vtxbuf[i].buffer) {
+			void *buf
+				= pipe->winsys->buffer_map(pipe->winsys,
+						nv04->vtxbuf[i].buffer,
+						PIPE_BUFFER_USAGE_CPU_READ);
+			draw_set_mapped_vertex_buffer(draw, i, buf);
+		}
+	}
+	/* Map index buffer, if present */
+	if (indexBuffer) {
+		void *mapped_indexes
+			= pipe->winsys->buffer_map(pipe->winsys, indexBuffer,
+					PIPE_BUFFER_USAGE_CPU_READ);
+		draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
+	}
+	else {
+		/* no index/element buffer */
+		draw_set_mapped_element_buffer(draw, 0, NULL);
+	}
+
+	draw_set_mapped_constant_buffer(draw,
+					nv04->constbuf[PIPE_SHADER_VERTEX],
+					nv04->constbuf_nr[PIPE_SHADER_VERTEX]);
+
+	/* draw! */
+	draw_arrays(nv04->draw, prim, start, count);
+
+	/*
+	 * unmap vertex/index buffers
+	 */
+	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
+		if (nv04->vtxbuf[i].buffer) {
+			pipe->winsys->buffer_unmap(pipe->winsys, nv04->vtxbuf[i].buffer);
+			draw_set_mapped_vertex_buffer(draw, i, NULL);
+		}
+	}
+	if (indexBuffer) {
+		pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer);
+		draw_set_mapped_element_buffer(draw, 0, NULL);
+	}
+
+	return TRUE;
+}
+
+boolean nv04_draw_arrays( struct pipe_context *pipe,
+				 unsigned prim, unsigned start, unsigned count)
+{
+	printf("coucou in draw arrays\n");
+	return nv04_draw_elements(pipe, NULL, 0, prim, start, count);
+}
+
+
+
diff --git a/src/gallium/drivers/nv10/Makefile b/src/gallium/drivers/nv10/Makefile
new file mode 100644
index 0000000000..4ba7ce586d
--- /dev/null
+++ b/src/gallium/drivers/nv10/Makefile
@@ -0,0 +1,28 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nv10
+
+DRIVER_SOURCES = \
+	nv10_clear.c \
+	nv10_context.c \
+	nv10_fragprog.c \
+	nv10_fragtex.c \
+	nv10_miptree.c \
+	nv10_prim_vbuf.c \
+	nv10_screen.c \
+	nv10_state.c \
+	nv10_state_emit.c \
+	nv10_surface.c \
+	nv10_vbo.c
+
+C_SOURCES = \
+	$(COMMON_SOURCES) \
+	$(DRIVER_SOURCES)
+
+ASM_SOURCES = 
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/drivers/nv10/nv10_clear.c b/src/gallium/drivers/nv10/nv10_clear.c
new file mode 100644
index 0000000000..be7e09cf4b
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_clear.c
@@ -0,0 +1,12 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "nv10_context.h"
+
+void
+nv10_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+	   unsigned clearValue)
+{
+	pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue);
+}
diff --git a/src/gallium/drivers/nv10/nv10_context.c b/src/gallium/drivers/nv10/nv10_context.c
new file mode 100644
index 0000000000..ef2c0c5d9f
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_context.c
@@ -0,0 +1,296 @@
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+#include "nv10_context.h"
+#include "nv10_screen.h"
+
+static void
+nv10_flush(struct pipe_context *pipe, unsigned flags,
+	   struct pipe_fence_handle **fence)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	draw_flush(nv10->draw);
+
+	FIRE_RING(fence);
+}
+
+static void
+nv10_destroy(struct pipe_context *pipe)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	if (nv10->draw)
+		draw_destroy(nv10->draw);
+
+	FREE(nv10);
+}
+
+static void nv10_init_hwctx(struct nv10_context *nv10)
+{
+	struct nv10_screen *screen = nv10->screen;
+	struct nouveau_winsys *nvws = screen->nvws;
+	int i;
+	float projectionmatrix[16];
+
+	BEGIN_RING(celsius, NV10TCL_DMA_NOTIFY, 1);
+	OUT_RING  (screen->sync->handle);
+	BEGIN_RING(celsius, NV10TCL_DMA_IN_MEMORY0, 2);
+	OUT_RING  (nvws->channel->vram->handle);
+	OUT_RING  (nvws->channel->gart->handle);
+	BEGIN_RING(celsius, NV10TCL_DMA_IN_MEMORY2, 2);
+	OUT_RING  (nvws->channel->vram->handle);
+	OUT_RING  (nvws->channel->vram->handle);
+
+	BEGIN_RING(celsius, NV10TCL_NOP, 1);
+	OUT_RING  (0);
+
+	BEGIN_RING(celsius, NV10TCL_RT_HORIZ, 2);
+	OUT_RING  (0);
+	OUT_RING  (0);
+
+	BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_HORIZ(0), 1);
+	OUT_RING  ((0x7ff<<16)|0x800);
+	BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_VERT(0), 1);
+	OUT_RING  ((0x7ff<<16)|0x800);
+
+	for (i=1;i<8;i++) {
+		BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_HORIZ(i), 1);
+		OUT_RING  (0);
+		BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_VERT(i), 1);
+		OUT_RING  (0);
+	}
+
+	BEGIN_RING(celsius, 0x290, 1);
+	OUT_RING  ((0x10<<16)|1);
+	BEGIN_RING(celsius, 0x3f4, 1);
+	OUT_RING  (0);
+
+	BEGIN_RING(celsius, NV10TCL_NOP, 1);
+	OUT_RING  (0);
+
+	if (nv10->screen->celsius->grclass != NV10TCL) {
+		/* For nv11, nv17 */
+		BEGIN_RING(celsius, 0x120, 3);
+		OUT_RING  (0);
+		OUT_RING  (1);
+		OUT_RING  (2);
+
+		BEGIN_RING(celsius, NV10TCL_NOP, 1);
+		OUT_RING  (0);
+	}
+
+	BEGIN_RING(celsius, NV10TCL_NOP, 1);
+	OUT_RING  (0);
+
+	/* Set state */
+	BEGIN_RING(celsius, NV10TCL_FOG_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_FUNC, 2);
+	OUT_RING  (0x207);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_TX_ENABLE(0), 2);
+	OUT_RING  (0);
+	OUT_RING  (0);
+
+	BEGIN_RING(celsius, NV10TCL_RC_IN_ALPHA(0), 12);
+	OUT_RING  (0x30141010);
+	OUT_RING  (0);
+	OUT_RING  (0x20040000);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	OUT_RING  (0x00000c00);
+	OUT_RING  (0);
+	OUT_RING  (0x00000c00);
+	OUT_RING  (0x18000000);
+	OUT_RING  (0x300e0300);
+	OUT_RING  (0x0c091c80);
+
+	BEGIN_RING(celsius, NV10TCL_BLEND_FUNC_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_DITHER_ENABLE, 2);
+	OUT_RING  (1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_LINE_SMOOTH_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_VERTEX_WEIGHT_ENABLE, 2);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_BLEND_FUNC_SRC, 4);
+	OUT_RING  (1);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	OUT_RING  (0x8006);
+	BEGIN_RING(celsius, NV10TCL_STENCIL_MASK, 8);
+	OUT_RING  (0xff);
+	OUT_RING  (0x207);
+	OUT_RING  (0);
+	OUT_RING  (0xff);
+	OUT_RING  (0x1e00);
+	OUT_RING  (0x1e00);
+	OUT_RING  (0x1e00);
+	OUT_RING  (0x1d01);
+	BEGIN_RING(celsius, NV10TCL_NORMALIZE_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_FOG_ENABLE, 2);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_LIGHT_MODEL, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_COLOR_CONTROL, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_ENABLED_LIGHTS, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_POLYGON_OFFSET_POINT_ENABLE, 3);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_DEPTH_FUNC, 1);
+	OUT_RING  (0x201);
+	BEGIN_RING(celsius, NV10TCL_DEPTH_WRITE_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_DEPTH_TEST_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_POLYGON_OFFSET_FACTOR, 2);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_POINT_SIZE, 1);
+	OUT_RING  (8);
+	BEGIN_RING(celsius, NV10TCL_POINT_PARAMETERS_ENABLE, 2);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_LINE_WIDTH, 1);
+	OUT_RING  (8);
+	BEGIN_RING(celsius, NV10TCL_LINE_SMOOTH_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_POLYGON_MODE_FRONT, 2);
+	OUT_RING  (0x1b02);
+	OUT_RING  (0x1b02);
+	BEGIN_RING(celsius, NV10TCL_CULL_FACE, 2);
+	OUT_RING  (0x405);
+	OUT_RING  (0x901);
+	BEGIN_RING(celsius, NV10TCL_POLYGON_SMOOTH_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_CULL_FACE_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_TX_GEN_S(0), 8);
+	for (i=0;i<8;i++) {
+		OUT_RING  (0);
+	}
+	BEGIN_RING(celsius, NV10TCL_FOG_EQUATION_CONSTANT, 3);
+	OUT_RING  (0x3fc00000);	/* -1.50 */
+	OUT_RING  (0xbdb8aa0a);	/* -0.09 */
+	OUT_RING  (0);		/*  0.00 */
+
+	BEGIN_RING(celsius, NV10TCL_NOP, 1);
+	OUT_RING  (0);
+
+	BEGIN_RING(celsius, NV10TCL_FOG_MODE, 2);
+	OUT_RING  (0x802);
+	OUT_RING  (2);
+	/* for some reason VIEW_MATRIX_ENABLE need to be 6 instead of 4 when
+	 * using texturing, except when using the texture matrix
+	 */
+	BEGIN_RING(celsius, NV10TCL_VIEW_MATRIX_ENABLE, 1);
+	OUT_RING  (6);
+	BEGIN_RING(celsius, NV10TCL_COLOR_MASK, 1);
+	OUT_RING  (0x01010101);
+
+	/* Set vertex component */
+	BEGIN_RING(celsius, NV10TCL_VERTEX_COL_4F_R, 4);
+	OUT_RINGf (1.0);
+	OUT_RINGf (1.0);
+	OUT_RINGf (1.0);
+	OUT_RINGf (1.0);
+	BEGIN_RING(celsius, NV10TCL_VERTEX_COL2_3F_R, 3);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(celsius, NV10TCL_VERTEX_NOR_3F_X, 3);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	OUT_RINGf (1.0);
+	BEGIN_RING(celsius, NV10TCL_VERTEX_TX0_4F_S, 4);
+	OUT_RINGf (0.0);
+	OUT_RINGf (0.0);
+	OUT_RINGf (0.0);
+	OUT_RINGf (1.0);
+	BEGIN_RING(celsius, NV10TCL_VERTEX_TX1_4F_S, 4);
+	OUT_RINGf (0.0);
+	OUT_RINGf (0.0);
+	OUT_RINGf (0.0);
+	OUT_RINGf (1.0);
+	BEGIN_RING(celsius, NV10TCL_VERTEX_FOG_1F, 1);
+	OUT_RINGf (0.0);
+	BEGIN_RING(celsius, NV10TCL_EDGEFLAG_ENABLE, 1);
+	OUT_RING  (1);
+
+	memset(projectionmatrix, 0, sizeof(projectionmatrix));
+	BEGIN_RING(celsius, NV10TCL_PROJECTION_MATRIX(0), 16);
+	projectionmatrix[0*4+0] = 1.0;
+	projectionmatrix[1*4+1] = 1.0;
+	projectionmatrix[2*4+2] = 1.0;
+	projectionmatrix[3*4+3] = 1.0;
+	for (i=0;i<16;i++) {
+		OUT_RINGf  (projectionmatrix[i]);
+	}
+
+	BEGIN_RING(celsius, NV10TCL_DEPTH_RANGE_NEAR, 2);
+	OUT_RING  (0.0);
+	OUT_RINGf  (16777216.0);
+
+	BEGIN_RING(celsius, NV10TCL_VIEWPORT_SCALE_X, 4);
+	OUT_RINGf  (-2048.0);
+	OUT_RINGf  (-2048.0);
+	OUT_RINGf  (16777215.0 * 0.5);
+	OUT_RING  (0);
+
+	FIRE_RING (NULL);
+}
+
+static void
+nv10_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield)
+{
+}
+
+struct pipe_context *
+nv10_create(struct pipe_screen *pscreen, unsigned pctx_id)
+{
+	struct nv10_screen *screen = nv10_screen(pscreen);
+	struct pipe_winsys *ws = pscreen->winsys;
+	struct nv10_context *nv10;
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nv10 = CALLOC(1, sizeof(struct nv10_context));
+	if (!nv10)
+		return NULL;
+	nv10->screen = screen;
+	nv10->pctx_id = pctx_id;
+
+	nv10->nvws = nvws;
+
+	nv10->pipe.winsys = ws;
+	nv10->pipe.screen = pscreen;
+	nv10->pipe.destroy = nv10_destroy;
+	nv10->pipe.set_edgeflags = nv10_set_edgeflags;
+	nv10->pipe.draw_arrays = nv10_draw_arrays;
+	nv10->pipe.draw_elements = nv10_draw_elements;
+	nv10->pipe.clear = nv10_clear;
+	nv10->pipe.flush = nv10_flush;
+
+	nv10_init_surface_functions(nv10);
+	nv10_init_state_functions(nv10);
+
+	nv10->draw = draw_create();
+	assert(nv10->draw);
+	draw_set_rasterize_stage(nv10->draw, nv10_draw_vbuf_stage(nv10));
+
+	nv10_init_hwctx(nv10);
+
+	return &nv10->pipe;
+}
+
diff --git a/src/gallium/drivers/nv10/nv10_context.h b/src/gallium/drivers/nv10/nv10_context.h
new file mode 100644
index 0000000000..f3b56de25a
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_context.h
@@ -0,0 +1,153 @@
+#ifndef __NV10_CONTEXT_H__
+#define __NV10_CONTEXT_H__
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_compiler.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "draw/draw_vertex.h"
+
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau/nouveau_gldefs.h"
+
+#define NOUVEAU_PUSH_CONTEXT(ctx)                                              \
+	struct nv10_screen *ctx = nv10->screen
+#include "nouveau/nouveau_push.h"
+
+#include "nv10_state.h"
+
+#define NOUVEAU_ERR(fmt, args...) \
+	fprintf(stderr, "%s:%d -  "fmt, __func__, __LINE__, ##args);
+#define NOUVEAU_MSG(fmt, args...) \
+	fprintf(stderr, "nouveau: "fmt, ##args);
+
+#define NV10_NEW_VERTPROG	(1 << 0)
+#define NV10_NEW_FRAGPROG	(1 << 1)
+#define NV10_NEW_VTXARRAYS	(1 << 2)
+#define NV10_NEW_BLEND		(1 << 3)
+#define NV10_NEW_BLENDCOL	(1 << 4)
+#define NV10_NEW_RAST 		(1 << 5)
+#define NV10_NEW_DSA  		(1 << 6)
+#define NV10_NEW_VIEWPORT	(1 << 7)
+#define NV10_NEW_SCISSOR	(1 << 8)
+#define NV10_NEW_FRAMEBUFFER	(1 << 9)
+
+#include "nv10_screen.h"
+
+struct nv10_context {
+	struct pipe_context pipe;
+
+	struct nouveau_winsys *nvws;
+	struct nv10_screen *screen;
+	unsigned pctx_id;
+
+	struct draw_context *draw;
+
+	uint32_t dirty;
+
+	struct nv10_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS];
+	struct nv10_miptree *tex_miptree[PIPE_MAX_SAMPLERS];
+	unsigned dirty_samplers;
+	unsigned fp_samplers;
+	unsigned vp_samplers;
+
+	uint32_t rt_enable;
+	struct pipe_buffer *rt[4];
+	struct pipe_buffer *zeta;
+	uint32_t lma_offset;
+
+	struct nv10_blend_state *blend;
+	struct pipe_blend_color *blend_color;
+	struct nv10_rasterizer_state *rast;
+	struct nv10_depth_stencil_alpha_state *dsa;
+	struct pipe_viewport_state *viewport;
+	struct pipe_scissor_state *scissor;
+	struct pipe_framebuffer_state *framebuffer;
+
+	//struct pipe_buffer *constbuf[PIPE_SHADER_TYPES];
+	float *constbuf[PIPE_SHADER_TYPES][32][4];
+	unsigned constbuf_nr[PIPE_SHADER_TYPES];
+
+	struct vertex_info vertex_info;
+
+	struct {
+		struct pipe_buffer *buffer;
+		uint32_t format;
+	} tex[2];
+
+	unsigned vb_enable;
+	struct {
+		struct pipe_buffer *buffer;
+		unsigned delta;
+	} vb[16];
+
+/*	struct {
+	
+		struct nouveau_resource *exec_heap;
+		struct nouveau_resource *data_heap;
+
+		struct nv10_vertex_program *active;
+
+		struct nv10_vertex_program *current;
+	} vertprog;
+*/
+	struct {
+		struct nv10_fragment_program *active;
+
+		struct nv10_fragment_program *current;
+		struct pipe_buffer *constant_buf;
+	} fragprog;
+
+	struct pipe_vertex_buffer  vtxbuf[PIPE_MAX_ATTRIBS];
+	struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS];
+};
+
+static INLINE struct nv10_context *
+nv10_context(struct pipe_context *pipe)
+{
+	return (struct nv10_context *)pipe;
+}
+
+extern void nv10_init_state_functions(struct nv10_context *nv10);
+extern void nv10_init_surface_functions(struct nv10_context *nv10);
+
+extern void nv10_screen_init_miptree_functions(struct pipe_screen *pscreen);
+
+/* nv10_clear.c */
+extern void nv10_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+		       unsigned clearValue);
+
+/* nv10_draw.c */
+extern struct draw_stage *nv10_draw_render_stage(struct nv10_context *nv10);
+
+/* nv10_fragprog.c */
+extern void nv10_fragprog_bind(struct nv10_context *,
+			       struct nv10_fragment_program *);
+extern void nv10_fragprog_destroy(struct nv10_context *,
+				  struct nv10_fragment_program *);
+
+/* nv10_fragtex.c */
+extern void nv10_fragtex_bind(struct nv10_context *);
+
+/* nv10_prim_vbuf.c */
+struct draw_stage *nv10_draw_vbuf_stage( struct nv10_context *nv10 );
+extern void nv10_vtxbuf_bind(struct nv10_context* nv10);
+
+/* nv10_state.c and friends */
+extern void nv10_emit_hw_state(struct nv10_context *nv10);
+extern void nv10_state_tex_update(struct nv10_context *nv10);
+
+/* nv10_vbo.c */
+extern boolean nv10_draw_arrays(struct pipe_context *, unsigned mode,
+				unsigned start, unsigned count);
+extern boolean nv10_draw_elements( struct pipe_context *pipe,
+                    struct pipe_buffer *indexBuffer,
+                    unsigned indexSize,
+                    unsigned prim, unsigned start, unsigned count);
+
+
+#endif
diff --git a/src/gallium/drivers/nv10/nv10_fragprog.c b/src/gallium/drivers/nv10/nv10_fragprog.c
new file mode 100644
index 0000000000..698db5a16a
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_fragprog.c
@@ -0,0 +1,21 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+
+#include "nv10_context.h"
+
+void
+nv10_fragprog_bind(struct nv10_context *nv10, struct nv10_fragment_program *fp)
+{
+}
+
+void
+nv10_fragprog_destroy(struct nv10_context *nv10,
+		      struct nv10_fragment_program *fp)
+{
+}
+
diff --git a/src/gallium/drivers/nv10/nv10_fragtex.c b/src/gallium/drivers/nv10/nv10_fragtex.c
new file mode 100644
index 0000000000..27f2f87584
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_fragtex.c
@@ -0,0 +1,124 @@
+#include "nv10_context.h"
+#include "nouveau/nouveau_util.h"
+
+#define _(m,tf)                                                                \
+{                                                                              \
+  TRUE,                                                                        \
+  PIPE_FORMAT_##m,                                                             \
+  NV10TCL_TX_FORMAT_FORMAT_##tf,                                               \
+}
+
+struct nv10_texture_format {
+	boolean defined;
+	uint	pipe;
+	int     format;
+};
+
+static struct nv10_texture_format
+nv10_texture_formats[] = {
+	_(A8R8G8B8_UNORM, A8R8G8B8),
+	_(A1R5G5B5_UNORM, A1R5G5B5),
+	_(A4R4G4B4_UNORM, A4R4G4B4),
+	_(L8_UNORM      , L8      ),
+	_(A8_UNORM      , A8      ),
+	_(A8L8_UNORM    , A8L8    ),
+//	_(RGB_DXT1      , DXT1,   ),
+//	_(RGBA_DXT1     , DXT1,   ),
+//	_(RGBA_DXT3     , DXT3,   ),
+//	_(RGBA_DXT5     , DXT5,   ),
+	{},
+};
+
+static struct nv10_texture_format *
+nv10_fragtex_format(uint pipe_format)
+{
+	struct nv10_texture_format *tf = nv10_texture_formats;
+
+	while (tf->defined) {
+		if (tf->pipe == pipe_format)
+			return tf;
+		tf++;
+	}
+
+	return NULL;
+}
+
+
+static void
+nv10_fragtex_build(struct nv10_context *nv10, int unit)
+{
+#if 0
+	struct nv10_sampler_state *ps = nv10->tex_sampler[unit];
+	struct nv10_miptree *nv10mt = nv10->tex_miptree[unit];
+	struct pipe_texture *pt = &nv10mt->base;
+	struct nv10_texture_format *tf;
+	uint32_t txf, txs, txp;
+
+	tf = nv10_fragtex_format(pt->format);
+	if (!tf || !tf->defined) {
+		NOUVEAU_ERR("Unsupported texture format: 0x%x\n", pt->format);
+		return;
+	}
+
+	txf  = tf->format << 8;
+	txf |= (pt->last_level + 1) << 16;
+	txf |= log2i(pt->width[0]) << 20;
+	txf |= log2i(pt->height[0]) << 24;
+	txf |= log2i(pt->depth[0]) << 28;
+	txf |= 8;
+
+	switch (pt->target) {
+	case PIPE_TEXTURE_CUBE:
+		txf |= NV10TCL_TX_FORMAT_CUBE_MAP;
+		/* fall-through */
+	case PIPE_TEXTURE_2D:
+		txf |= (2<<4);
+		break;
+	case PIPE_TEXTURE_1D:
+		txf |= (1<<4);
+		break;
+	default:
+		NOUVEAU_ERR("Unknown target %d\n", pt->target);
+		return;
+	}
+
+	BEGIN_RING(celsius, NV10TCL_TX_OFFSET(unit), 8);
+	OUT_RELOCl(nv10mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD);
+	OUT_RELOCd(nv10mt->buffer,txf,NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_OR | NOUVEAU_BO_RD, 1/*VRAM*/,2/*TT*/);
+	OUT_RING  (ps->wrap);
+	OUT_RING  (0x40000000); /* enable */
+	OUT_RING  (txs);
+	OUT_RING  (ps->filt | 0x2000 /* magic */);
+	OUT_RING  ((pt->width[0] << 16) | pt->height[0]);
+	OUT_RING  (ps->bcol);
+#endif
+}
+
+void
+nv10_fragtex_bind(struct nv10_context *nv10)
+{
+#if 0
+	struct nv10_fragment_program *fp = nv10->fragprog.active;
+	unsigned samplers, unit;
+
+	samplers = nv10->fp_samplers & ~fp->samplers;
+	while (samplers) {
+		unit = ffs(samplers) - 1;
+		samplers &= ~(1 << unit);
+
+		BEGIN_RING(celsius, NV10TCL_TX_ENABLE(unit), 1);
+		OUT_RING  (0);
+	}
+
+	samplers = nv10->dirty_samplers & fp->samplers;
+	while (samplers) {
+		unit = ffs(samplers) - 1;
+		samplers &= ~(1 << unit);
+
+		nv10_fragtex_build(nv10, unit);
+	}
+
+	nv10->fp_samplers = fp->samplers;
+#endif
+}
+
diff --git a/src/gallium/drivers/nv10/nv10_miptree.c b/src/gallium/drivers/nv10/nv10_miptree.c
new file mode 100644
index 0000000000..9616135461
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_miptree.c
@@ -0,0 +1,174 @@
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "nv10_context.h"
+#include "nv10_screen.h"
+
+static void
+nv10_miptree_layout(struct nv10_miptree *nv10mt)
+{
+	struct pipe_texture *pt = &nv10mt->base;
+	boolean swizzled = FALSE;
+	uint width = pt->width[0], height = pt->height[0];
+	uint offset = 0;
+	int nr_faces, l, f;
+
+	if (pt->target == PIPE_TEXTURE_CUBE) {
+		nr_faces = 6;
+	} else {
+		nr_faces = 1;
+	}
+	
+	for (l = 0; l <= pt->last_level; l++) {
+		pt->width[l] = width;
+		pt->height[l] = height;
+		pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width);
+		pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height);
+
+		if (swizzled)
+			nv10mt->level[l].pitch = pt->nblocksx[l] * pt->block.size;
+		else
+			nv10mt->level[l].pitch = pt->nblocksx[0] * pt->block.size;
+		nv10mt->level[l].pitch = (nv10mt->level[l].pitch + 63) & ~63;
+
+		nv10mt->level[l].image_offset =
+			CALLOC(nr_faces, sizeof(unsigned));
+
+		width  = MAX2(1, width  >> 1);
+		height = MAX2(1, height >> 1);
+
+	}
+
+	for (f = 0; f < nr_faces; f++) {
+		for (l = 0; l <= pt->last_level; l++) {
+			nv10mt->level[l].image_offset[f] = offset;
+			offset += nv10mt->level[l].pitch * pt->height[l];
+		}
+	}
+
+	nv10mt->total_size = offset;
+}
+
+static struct pipe_texture *
+nv10_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
+		     const unsigned *stride, struct pipe_buffer *pb)
+{
+	struct nv10_miptree *mt;
+
+	/* Only supports 2D, non-mipmapped textures for the moment */
+	if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 ||
+	    pt->depth[0] != 1)
+		return NULL;
+
+	mt = CALLOC_STRUCT(nv10_miptree);
+	if (!mt)
+		return NULL;
+
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = pscreen;
+	mt->level[0].pitch = stride[0];
+	mt->level[0].image_offset = CALLOC(1, sizeof(unsigned));
+
+	pipe_buffer_reference(pscreen, &mt->buffer, pb);
+	return &mt->base;
+}
+
+static struct pipe_texture *
+nv10_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt)
+{
+	struct pipe_winsys *ws = screen->winsys;
+	struct nv10_miptree *mt;
+
+	mt = MALLOC(sizeof(struct nv10_miptree));
+	if (!mt)
+		return NULL;
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = screen;
+
+	nv10_miptree_layout(mt);
+
+	mt->buffer = ws->buffer_create(ws, 256, PIPE_BUFFER_USAGE_PIXEL,
+					   mt->total_size);
+	if (!mt->buffer) {
+		FREE(mt);
+		return NULL;
+	}
+	
+	return &mt->base;
+}
+
+static void
+nv10_miptree_release(struct pipe_screen *screen, struct pipe_texture **pt)
+{
+	struct pipe_texture *mt = *pt;
+
+	*pt = NULL;
+	if (--mt->refcount <= 0) {
+		struct nv10_miptree *nv10mt = (struct nv10_miptree *)mt;
+		int l;
+
+		pipe_buffer_reference(screen, &nv10mt->buffer, NULL);
+		for (l = 0; l <= mt->last_level; l++) {
+			if (nv10mt->level[l].image_offset)
+				FREE(nv10mt->level[l].image_offset);
+		}
+		FREE(nv10mt);
+	}
+}
+
+static void
+nv10_miptree_update(struct pipe_context *pipe, struct pipe_texture *mt,
+		    uint face, uint levels)
+{
+}
+
+
+static struct pipe_surface *
+nv10_miptree_surface_get(struct pipe_screen *screen, struct pipe_texture *pt,
+			 unsigned face, unsigned level, unsigned zslice,
+			 unsigned flags)
+{
+	struct pipe_winsys *ws = screen->winsys;
+	struct nv10_miptree *nv10mt = (struct nv10_miptree *)pt;
+	struct pipe_surface *ps;
+
+	ps = CALLOC_STRUCT(pipe_surface);
+	if (!ps)
+		return NULL;
+	pipe_texture_reference(&ps->texture, pt);
+	ps->format = pt->format;
+	ps->width = pt->width[level];
+	ps->height = pt->height[level];
+	ps->block = pt->block;
+	ps->nblocksx = pt->nblocksx[level];
+	ps->nblocksy = pt->nblocksy[level];
+	ps->stride = nv10mt->level[level].pitch;
+	ps->refcount = 1;
+
+	if (pt->target == PIPE_TEXTURE_CUBE) {
+		ps->offset = nv10mt->level[level].image_offset[face];
+	} else {
+		ps->offset = nv10mt->level[level].image_offset[0];
+	}
+
+	return ps;
+}
+
+static void
+nv10_miptree_surface_release(struct pipe_screen *screen,
+			     struct pipe_surface **surface)
+{
+}
+
+void nv10_screen_init_miptree_functions(struct pipe_screen *pscreen)
+{
+	pscreen->texture_create = nv10_miptree_create;
+	pscreen->texture_blanket = nv10_miptree_blanket;
+	pscreen->texture_release = nv10_miptree_release;
+	pscreen->get_tex_surface = nv10_miptree_surface_get;
+	pscreen->tex_surface_release = nv10_miptree_surface_release;
+}
+
diff --git a/src/gallium/drivers/nv10/nv10_prim_vbuf.c b/src/gallium/drivers/nv10/nv10_prim_vbuf.c
new file mode 100644
index 0000000000..7435d87315
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_prim_vbuf.c
@@ -0,0 +1,245 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \file
+ * Build post-transformation, post-clipping vertex buffers and element
+ * lists by hooking into the end of the primitive pipeline and
+ * manipulating the vertex_id field in the vertex headers.
+ *
+ * XXX: work in progress 
+ * 
+ * \author José Fonseca <jrfonseca@tungstengraphics.com>
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_debug.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+#include "nv10_context.h"
+#include "nv10_state.h"
+
+#include "draw/draw_vbuf.h"
+
+/**
+ * Primitive renderer for nv10.
+ */
+struct nv10_vbuf_render {
+	struct vbuf_render base;
+
+	struct nv10_context *nv10;   
+
+	/** Vertex buffer */
+	struct pipe_buffer* buffer;
+
+	/** Vertex size in bytes */
+	unsigned vertex_size;
+
+	/** Hardware primitive */
+	unsigned hwprim;
+};
+
+
+void nv10_vtxbuf_bind( struct nv10_context* nv10 )
+{
+	int i;
+	for(i = 0; i < 8; i++) {
+		BEGIN_RING(celsius, NV10TCL_VERTEX_ARRAY_ATTRIB_OFFSET(i), 1);
+		OUT_RING(0/*nv10->vtxbuf*/);
+		BEGIN_RING(celsius, NV10TCL_VERTEX_ARRAY_ATTRIB_FORMAT(i) ,1);
+		OUT_RING(0/*XXX*/);
+	}
+}
+
+/**
+ * Basically a cast wrapper.
+ */
+static INLINE struct nv10_vbuf_render *
+nv10_vbuf_render( struct vbuf_render *render )
+{
+	assert(render);
+	return (struct nv10_vbuf_render *)render;
+}
+
+
+static const struct vertex_info *
+nv10_vbuf_render_get_vertex_info( struct vbuf_render *render )
+{
+	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
+	struct nv10_context *nv10 = nv10_render->nv10;
+
+	nv10_emit_hw_state(nv10);
+
+	return &nv10->vertex_info;
+}
+
+
+static void *
+nv10_vbuf_render_allocate_vertices( struct vbuf_render *render,
+		ushort vertex_size,
+		ushort nr_vertices )
+{
+	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
+	struct nv10_context *nv10 = nv10_render->nv10;
+	struct pipe_winsys *winsys = nv10->pipe.winsys;
+	size_t size = (size_t)vertex_size * (size_t)nr_vertices;
+
+	assert(!nv10_render->buffer);
+	nv10_render->buffer = winsys->buffer_create(winsys, 64, PIPE_BUFFER_USAGE_VERTEX, size);
+
+	nv10->dirty |= NV10_NEW_VTXARRAYS;
+
+	return winsys->buffer_map(winsys, 
+			nv10_render->buffer, 
+			PIPE_BUFFER_USAGE_CPU_WRITE);
+}
+
+
+static boolean
+nv10_vbuf_render_set_primitive( struct vbuf_render *render, 
+		unsigned prim )
+{
+	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
+	unsigned hwp = nvgl_primitive(prim);
+	if (hwp == 0)
+		return FALSE;
+
+	nv10_render->hwprim = hwp;
+	return TRUE;
+}
+
+
+static void 
+nv10_vbuf_render_draw( struct vbuf_render *render,
+		const ushort *indices,
+		uint nr_indices)
+{
+	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
+	struct nv10_context *nv10 = nv10_render->nv10;
+	int push, i;
+
+	nv10_emit_hw_state(nv10);
+
+	BEGIN_RING(celsius, NV10TCL_VERTEX_ARRAY_OFFSET_POS, 1);
+	OUT_RELOCl(nv10_render->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD);
+
+	BEGIN_RING(celsius, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1);
+	OUT_RING(nv10_render->hwprim);
+
+	if (nr_indices & 1) {
+		BEGIN_RING(celsius, NV10TCL_VB_ELEMENT_U32, 1);
+		OUT_RING  (indices[0]);
+		indices++; nr_indices--;
+	}
+
+	while (nr_indices) {
+		// XXX too big/small ? check the size
+		push = MIN2(nr_indices, 1200 * 2);
+
+		BEGIN_RING_NI(celsius, NV10TCL_VB_ELEMENT_U16, push >> 1);
+		for (i = 0; i < push; i+=2)
+			OUT_RING((indices[i+1] << 16) | indices[i]);
+
+		nr_indices -= push;
+		indices  += push;
+	}
+
+	BEGIN_RING(celsius, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1);
+	OUT_RING  (0);
+}
+
+
+static void
+nv10_vbuf_render_release_vertices( struct vbuf_render *render,
+		void *vertices, 
+		unsigned vertex_size,
+		unsigned vertices_used )
+{
+	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
+	struct nv10_context *nv10 = nv10_render->nv10;
+	struct pipe_winsys *winsys = nv10->pipe.winsys;
+	struct pipe_screen *pscreen = &nv10->screen->pipe;
+
+	assert(nv10_render->buffer);
+	winsys->buffer_unmap(winsys, nv10_render->buffer);
+	pipe_buffer_reference(pscreen, &nv10_render->buffer, NULL);
+}
+
+
+static void
+nv10_vbuf_render_destroy( struct vbuf_render *render )
+{
+	struct nv10_vbuf_render *nv10_render = nv10_vbuf_render(render);
+	FREE(nv10_render);
+}
+
+
+/**
+ * Create a new primitive render.
+ */
+static struct vbuf_render *
+nv10_vbuf_render_create( struct nv10_context *nv10 )
+{
+	struct nv10_vbuf_render *nv10_render = CALLOC_STRUCT(nv10_vbuf_render);
+
+	nv10_render->nv10 = nv10;
+
+	nv10_render->base.max_vertex_buffer_bytes = 16*1024;
+	nv10_render->base.max_indices = 1024;
+	nv10_render->base.get_vertex_info = nv10_vbuf_render_get_vertex_info;
+	nv10_render->base.allocate_vertices = nv10_vbuf_render_allocate_vertices;
+	nv10_render->base.set_primitive = nv10_vbuf_render_set_primitive;
+	nv10_render->base.draw = nv10_vbuf_render_draw;
+	nv10_render->base.release_vertices = nv10_vbuf_render_release_vertices;
+	nv10_render->base.destroy = nv10_vbuf_render_destroy;
+
+	return &nv10_render->base;
+}
+
+
+/**
+ * Create a new primitive vbuf/render stage.
+ */
+struct draw_stage *nv10_draw_vbuf_stage( struct nv10_context *nv10 )
+{
+	struct vbuf_render *render;
+	struct draw_stage *stage;
+
+	render = nv10_vbuf_render_create(nv10);
+	if(!render)
+		return NULL;
+
+	stage = draw_vbuf_stage( nv10->draw, render );
+	if(!stage) {
+		render->destroy(render);
+		return NULL;
+	}
+
+	return stage;
+}
diff --git a/src/gallium/drivers/nv10/nv10_screen.c b/src/gallium/drivers/nv10/nv10_screen.c
new file mode 100644
index 0000000000..f417b06c94
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_screen.c
@@ -0,0 +1,226 @@
+#include "pipe/p_screen.h"
+#include "util/u_simple_screen.h"
+
+#include "nv10_context.h"
+#include "nv10_screen.h"
+
+static const char *
+nv10_screen_get_name(struct pipe_screen *screen)
+{
+	struct nv10_screen *nv10screen = nv10_screen(screen);
+	struct nouveau_device *dev = nv10screen->nvws->channel->device;
+	static char buffer[128];
+
+	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
+	return buffer;
+}
+
+static const char *
+nv10_screen_get_vendor(struct pipe_screen *screen)
+{
+	return "nouveau";
+}
+
+static int
+nv10_screen_get_param(struct pipe_screen *screen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+		return 2;
+	case PIPE_CAP_NPOT_TEXTURES:
+		return 0;
+	case PIPE_CAP_TWO_SIDED_STENCIL:
+		return 0;
+	case PIPE_CAP_GLSL:
+		return 0;
+	case PIPE_CAP_S3TC:
+		return 0;
+	case PIPE_CAP_ANISOTROPIC_FILTER:
+		return 1;
+	case PIPE_CAP_POINT_SPRITE:
+		return 0;
+	case PIPE_CAP_MAX_RENDER_TARGETS:
+		return 1;
+	case PIPE_CAP_OCCLUSION_QUERY:
+		return 0;
+	case PIPE_CAP_TEXTURE_SHADOW_MAP:
+		return 0;
+	case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+		return 12;
+	case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+		return 0;
+	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+		return 12;
+	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
+		return 0;
+	case NOUVEAU_CAP_HW_VTXBUF:
+	case NOUVEAU_CAP_HW_IDXBUF:
+		return 0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0;
+	}
+}
+
+static float
+nv10_screen_get_paramf(struct pipe_screen *screen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_LINE_WIDTH:
+	case PIPE_CAP_MAX_LINE_WIDTH_AA:
+		return 10.0;
+	case PIPE_CAP_MAX_POINT_WIDTH:
+	case PIPE_CAP_MAX_POINT_WIDTH_AA:
+		return 64.0;
+	case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+		return 2.0;
+	case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+		return 4.0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0.0;
+	}
+}
+
+static boolean
+nv10_screen_is_format_supported(struct pipe_screen *screen,
+				enum pipe_format format,
+				enum pipe_texture_target target,
+				unsigned tex_usage, unsigned geom_flags)
+{
+	if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM: 
+		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z16_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	} else {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_A1R5G5B5_UNORM:
+		case PIPE_FORMAT_A4R4G4B4_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM: 
+		case PIPE_FORMAT_L8_UNORM:
+		case PIPE_FORMAT_A8_UNORM:
+		case PIPE_FORMAT_I8_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	}
+
+	return FALSE;
+}
+
+static void *
+nv10_surface_map(struct pipe_screen *screen, struct pipe_surface *surface,
+		 unsigned flags )
+{
+	struct pipe_winsys *ws = screen->winsys;
+	void *map;
+        struct nv10_miptree *nv10mt = (struct nv10_miptree *)surface->texture;
+
+	map = ws->buffer_map(ws, nv10mt->buffer, flags);
+	if (!map)
+		return NULL;
+
+	return map + surface->offset;
+}
+
+static void
+nv10_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface)
+{
+	struct pipe_winsys *ws = screen->winsys;
+        struct nv10_miptree *nv10mt = (struct nv10_miptree *)surface->texture;
+
+	ws->buffer_unmap(ws, nv10mt->buffer);
+}
+
+static void
+nv10_screen_destroy(struct pipe_screen *pscreen)
+{
+	struct nv10_screen *screen = nv10_screen(pscreen);
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nvws->notifier_free(&screen->sync);
+	nvws->grobj_free(&screen->celsius);
+
+	FREE(pscreen);
+}
+
+static struct pipe_buffer *
+nv10_surface_buffer(struct pipe_surface *surf)
+{
+	struct nv10_miptree *mt = (struct nv10_miptree *)surf->texture;
+
+	return mt->buffer;
+}
+
+struct pipe_screen *
+nv10_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+{
+	struct nv10_screen *screen = CALLOC_STRUCT(nv10_screen);
+	unsigned celsius_class;
+	unsigned chipset = nvws->channel->device->chipset;
+	int ret;
+
+	if (!screen)
+		return NULL;
+	screen->nvws = nvws;
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(nvws);
+	screen->eng2d->buf = nv10_surface_buffer;
+
+	/* 3D object */
+	if (chipset>=0x20)
+		celsius_class=NV11TCL;
+	else if (chipset>=0x17)
+		celsius_class=NV17TCL;
+	else if (chipset>=0x11)
+		celsius_class=NV11TCL;
+	else
+		celsius_class=NV10TCL;
+
+	if (!celsius_class) {
+		NOUVEAU_ERR("Unknown nv1x chipset: nv%02x\n", chipset);
+		return NULL;
+	}
+
+	ret = nvws->grobj_alloc(nvws, celsius_class, &screen->celsius);
+	if (ret) {
+		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
+		return FALSE;
+	}
+
+	/* Notifier for sync purposes */
+	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	if (ret) {
+		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
+		nv10_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	screen->pipe.winsys = ws;
+	screen->pipe.destroy = nv10_screen_destroy;
+
+	screen->pipe.get_name = nv10_screen_get_name;
+	screen->pipe.get_vendor = nv10_screen_get_vendor;
+	screen->pipe.get_param = nv10_screen_get_param;
+	screen->pipe.get_paramf = nv10_screen_get_paramf;
+
+	screen->pipe.is_format_supported = nv10_screen_is_format_supported;
+
+	screen->pipe.surface_map = nv10_surface_map;
+	screen->pipe.surface_unmap = nv10_surface_unmap;
+
+	nv10_screen_init_miptree_functions(&screen->pipe);
+	u_simple_screen_init(&screen->pipe);
+
+	return &screen->pipe;
+}
+
diff --git a/src/gallium/drivers/nv10/nv10_screen.h b/src/gallium/drivers/nv10/nv10_screen.h
new file mode 100644
index 0000000000..60102a369a
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_screen.h
@@ -0,0 +1,24 @@
+#ifndef __NV10_SCREEN_H__
+#define __NV10_SCREEN_H__
+
+#include "pipe/p_screen.h"
+#include "nv04/nv04_surface_2d.h"
+
+struct nv10_screen {
+	struct pipe_screen pipe;
+
+	struct nouveau_winsys *nvws;
+
+	/* HW graphics objects */
+	struct nv04_surface_2d *eng2d;
+	struct nouveau_grobj *celsius;
+	struct nouveau_notifier *sync;
+};
+
+static INLINE struct nv10_screen *
+nv10_screen(struct pipe_screen *screen)
+{
+	return (struct nv10_screen *)screen;
+}
+
+#endif
diff --git a/src/gallium/drivers/nv10/nv10_state.c b/src/gallium/drivers/nv10/nv10_state.c
new file mode 100644
index 0000000000..119af66dfd
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_state.c
@@ -0,0 +1,589 @@
+#include "draw/draw_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/tgsi_parse.h"
+
+#include "nv10_context.h"
+#include "nv10_state.h"
+
+static void *
+nv10_blend_state_create(struct pipe_context *pipe,
+			const struct pipe_blend_state *cso)
+{
+	struct nv10_blend_state *cb;
+
+	cb = MALLOC(sizeof(struct nv10_blend_state));
+
+	cb->b_enable = cso->blend_enable ? 1 : 0;
+	cb->b_srcfunc = ((nvgl_blend_func(cso->alpha_src_factor)<<16) |
+			 (nvgl_blend_func(cso->rgb_src_factor)));
+	cb->b_dstfunc = ((nvgl_blend_func(cso->alpha_dst_factor)<<16) |
+			 (nvgl_blend_func(cso->rgb_dst_factor)));
+
+	cb->c_mask = (((cso->colormask & PIPE_MASK_A) ? (0x01<<24) : 0) |
+		      ((cso->colormask & PIPE_MASK_R) ? (0x01<<16) : 0) |
+		      ((cso->colormask & PIPE_MASK_G) ? (0x01<< 8) : 0) |
+		      ((cso->colormask & PIPE_MASK_B) ? (0x01<< 0) : 0));
+
+	cb->d_enable = cso->dither ? 1 : 0;
+
+	return (void *)cb;
+}
+
+static void
+nv10_blend_state_bind(struct pipe_context *pipe, void *blend)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	nv10->blend = (struct nv10_blend_state*)blend;
+
+	nv10->dirty |= NV10_NEW_BLEND;
+}
+
+static void
+nv10_blend_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+
+static INLINE unsigned
+wrap_mode(unsigned wrap) {
+	unsigned ret;
+
+	switch (wrap) {
+	case PIPE_TEX_WRAP_REPEAT:
+		ret = NV10TCL_TX_FORMAT_WRAP_S_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_REPEAT:
+		ret = NV10TCL_TX_FORMAT_WRAP_S_MIRRORED_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+		ret = NV10TCL_TX_FORMAT_WRAP_S_CLAMP_TO_EDGE;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+		ret = NV10TCL_TX_FORMAT_WRAP_S_CLAMP_TO_BORDER;
+		break;
+	case PIPE_TEX_WRAP_CLAMP:
+		ret = NV10TCL_TX_FORMAT_WRAP_S_CLAMP;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+	case PIPE_TEX_WRAP_MIRROR_CLAMP:
+	default:
+		NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
+		ret = NV10TCL_TX_FORMAT_WRAP_S_REPEAT;
+		break;
+	}
+
+	return ret >> NV10TCL_TX_FORMAT_WRAP_S_SHIFT;
+}
+
+static void *
+nv10_sampler_state_create(struct pipe_context *pipe,
+			  const struct pipe_sampler_state *cso)
+{
+	struct nv10_sampler_state *ps;
+	uint32_t filter = 0;
+
+	ps = MALLOC(sizeof(struct nv10_sampler_state));
+
+	ps->wrap = ((wrap_mode(cso->wrap_s) << NV10TCL_TX_FORMAT_WRAP_S_SHIFT) |
+		    (wrap_mode(cso->wrap_t) << NV10TCL_TX_FORMAT_WRAP_T_SHIFT));
+
+	ps->en = 0;
+	if (cso->max_anisotropy > 1.0) {
+		/* no idea, binary driver sets it, works without it.. meh.. */
+		ps->wrap |= (1 << 5);
+
+/*		if (cso->max_anisotropy >= 16.0) {
+			ps->en |= NV10TCL_TX_ENABLE_ANISO_16X;
+		} else
+		if (cso->max_anisotropy >= 12.0) {
+			ps->en |= NV10TCL_TX_ENABLE_ANISO_12X;
+		} else
+		if (cso->max_anisotropy >= 10.0) {
+			ps->en |= NV10TCL_TX_ENABLE_ANISO_10X;
+		} else
+		if (cso->max_anisotropy >= 8.0) {
+			ps->en |= NV10TCL_TX_ENABLE_ANISO_8X;
+		} else
+		if (cso->max_anisotropy >= 6.0) {
+			ps->en |= NV10TCL_TX_ENABLE_ANISO_6X;
+		} else
+		if (cso->max_anisotropy >= 4.0) {
+			ps->en |= NV10TCL_TX_ENABLE_ANISO_4X;
+		} else {
+			ps->en |= NV10TCL_TX_ENABLE_ANISO_2X;
+		}*/
+	}
+
+	switch (cso->mag_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		filter |= NV10TCL_TX_FILTER_MAGNIFY_LINEAR;
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		filter |= NV10TCL_TX_FILTER_MAGNIFY_NEAREST;
+		break;
+	}
+
+	switch (cso->min_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |= NV10TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST;
+			break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV10TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV10TCL_TX_FILTER_MINIFY_LINEAR;
+			break;
+		}
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |= NV10TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST;
+		break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV10TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV10TCL_TX_FILTER_MINIFY_NEAREST;
+			break;
+		}
+		break;
+	}
+
+	ps->filt = filter;
+
+/*	if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+		switch (cso->compare_func) {
+		case PIPE_FUNC_NEVER:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_NEVER;
+			break;
+		case PIPE_FUNC_GREATER:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_GREATER;
+			break;
+		case PIPE_FUNC_EQUAL:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_EQUAL;
+			break;
+		case PIPE_FUNC_GEQUAL:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_GEQUAL;
+			break;
+		case PIPE_FUNC_LESS:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_LESS;
+			break;
+		case PIPE_FUNC_NOTEQUAL:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_NOTEQUAL;
+			break;
+		case PIPE_FUNC_LEQUAL:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_LEQUAL;
+			break;
+		case PIPE_FUNC_ALWAYS:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_ALWAYS;
+			break;
+		default:
+			break;
+		}
+	}*/
+
+	ps->bcol = ((float_to_ubyte(cso->border_color[3]) << 24) |
+		    (float_to_ubyte(cso->border_color[0]) << 16) |
+		    (float_to_ubyte(cso->border_color[1]) <<  8) |
+		    (float_to_ubyte(cso->border_color[2]) <<  0));
+
+	return (void *)ps;
+}
+
+static void
+nv10_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		nv10->tex_sampler[unit] = sampler[unit];
+		nv10->dirty_samplers |= (1 << unit);
+	}
+}
+
+static void
+nv10_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void
+nv10_set_sampler_texture(struct pipe_context *pipe, unsigned nr,
+			 struct pipe_texture **miptree)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		nv10->tex_miptree[unit] = (struct nv10_miptree *)miptree[unit];
+		nv10->dirty_samplers |= (1 << unit);
+	}
+}
+
+static void *
+nv10_rasterizer_state_create(struct pipe_context *pipe,
+			     const struct pipe_rasterizer_state *cso)
+{
+	struct nv10_rasterizer_state *rs;
+	int i;
+
+	/*XXX: ignored:
+	 * 	light_twoside
+	 * 	offset_cw/ccw -nohw
+	 * 	scissor
+	 * 	point_smooth -nohw
+	 * 	multisample
+	 * 	offset_units / offset_scale
+	 */
+	rs = MALLOC(sizeof(struct nv10_rasterizer_state));
+
+	rs->templ = cso;
+	
+	rs->shade_model = cso->flatshade ? 0x1d00 : 0x1d01;
+
+	rs->line_width = (unsigned char)(cso->line_width * 8.0) & 0xff;
+	rs->line_smooth_en = cso->line_smooth ? 1 : 0;
+
+	rs->point_size = *(uint32_t*)&cso->point_size;
+
+	rs->poly_smooth_en = cso->poly_smooth ? 1 : 0;
+
+	if (cso->front_winding == PIPE_WINDING_CCW) {
+		rs->front_face = NV10TCL_FRONT_FACE_CCW;
+		rs->poly_mode_front = nvgl_polygon_mode(cso->fill_ccw);
+		rs->poly_mode_back  = nvgl_polygon_mode(cso->fill_cw);
+	} else {
+		rs->front_face = NV10TCL_FRONT_FACE_CW;
+		rs->poly_mode_front = nvgl_polygon_mode(cso->fill_cw);
+		rs->poly_mode_back  = nvgl_polygon_mode(cso->fill_ccw);
+	}
+
+	switch (cso->cull_mode) {
+	case PIPE_WINDING_CCW:
+		rs->cull_face_en = 1;
+		if (cso->front_winding == PIPE_WINDING_CCW)
+			rs->cull_face    = NV10TCL_CULL_FACE_FRONT;
+		else
+			rs->cull_face    = NV10TCL_CULL_FACE_BACK;
+		break;
+	case PIPE_WINDING_CW:
+		rs->cull_face_en = 1;
+		if (cso->front_winding == PIPE_WINDING_CW)
+			rs->cull_face    = NV10TCL_CULL_FACE_FRONT;
+		else
+			rs->cull_face    = NV10TCL_CULL_FACE_BACK;
+		break;
+	case PIPE_WINDING_BOTH:
+		rs->cull_face_en = 1;
+		rs->cull_face    = NV10TCL_CULL_FACE_FRONT_AND_BACK;
+		break;
+	case PIPE_WINDING_NONE:
+	default:
+		rs->cull_face_en = 0;
+		rs->cull_face    = 0;
+		break;
+	}
+
+	if (cso->point_sprite) {
+		rs->point_sprite = (1 << 0);
+		for (i = 0; i < 8; i++) {
+			if (cso->sprite_coord_mode[i] != PIPE_SPRITE_COORD_NONE)
+				rs->point_sprite |= (1 << (8 + i));
+		}
+	} else {
+		rs->point_sprite = 0;
+	}
+
+	return (void *)rs;
+}
+
+static void
+nv10_rasterizer_state_bind(struct pipe_context *pipe, void *rast)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	nv10->rast = (struct nv10_rasterizer_state*)rast;
+
+	draw_set_rasterizer_state(nv10->draw, (nv10->rast ? nv10->rast->templ : NULL));
+
+	nv10->dirty |= NV10_NEW_RAST;
+}
+
+static void
+nv10_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void *
+nv10_depth_stencil_alpha_state_create(struct pipe_context *pipe,
+			const struct pipe_depth_stencil_alpha_state *cso)
+{
+	struct nv10_depth_stencil_alpha_state *hw;
+
+	hw = MALLOC(sizeof(struct nv10_depth_stencil_alpha_state));
+
+	hw->depth.func		= nvgl_comparison_op(cso->depth.func);
+	hw->depth.write_enable	= cso->depth.writemask ? 1 : 0;
+	hw->depth.test_enable	= cso->depth.enabled ? 1 : 0;
+
+	hw->stencil.enable = cso->stencil[0].enabled ? 1 : 0;
+	hw->stencil.wmask = cso->stencil[0].writemask;
+	hw->stencil.func = nvgl_comparison_op(cso->stencil[0].func);
+	hw->stencil.ref	= cso->stencil[0].ref_value;
+	hw->stencil.vmask = cso->stencil[0].valuemask;
+	hw->stencil.fail = nvgl_stencil_op(cso->stencil[0].fail_op);
+	hw->stencil.zfail = nvgl_stencil_op(cso->stencil[0].zfail_op);
+	hw->stencil.zpass = nvgl_stencil_op(cso->stencil[0].zpass_op);
+
+	hw->alpha.enabled = cso->alpha.enabled ? 1 : 0;
+	hw->alpha.func = nvgl_comparison_op(cso->alpha.func);
+	hw->alpha.ref  = float_to_ubyte(cso->alpha.ref_value);
+
+	return (void *)hw;
+}
+
+static void
+nv10_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *dsa)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	nv10->dsa = (struct nv10_depth_stencil_alpha_state*)dsa;
+
+	nv10->dirty |= NV10_NEW_DSA;
+}
+
+static void
+nv10_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void *
+nv10_vp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *templ)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	return draw_create_vertex_shader(nv10->draw, templ);
+}
+
+static void
+nv10_vp_state_bind(struct pipe_context *pipe, void *shader)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	draw_bind_vertex_shader(nv10->draw, (struct draw_vertex_shader *) shader);
+
+	nv10->dirty |= NV10_NEW_VERTPROG;
+}
+
+static void
+nv10_vp_state_delete(struct pipe_context *pipe, void *shader)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	draw_delete_vertex_shader(nv10->draw, (struct draw_vertex_shader *) shader);
+}
+
+static void *
+nv10_fp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	struct nv10_fragment_program *fp;
+
+	fp = CALLOC(1, sizeof(struct nv10_fragment_program));
+	fp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+	
+	tgsi_scan_shader(cso->tokens, &fp->info);
+
+	return (void *)fp;
+}
+
+static void
+nv10_fp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+	struct nv10_fragment_program *fp = hwcso;
+
+	nv10->fragprog.current = fp;
+	nv10->dirty |= NV10_NEW_FRAGPROG;
+}
+
+static void
+nv10_fp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+	struct nv10_fragment_program *fp = hwcso;
+
+	nv10_fragprog_destroy(nv10, fp);
+	FREE((void*)fp->pipe.tokens);
+	FREE(fp);
+}
+
+static void
+nv10_set_blend_color(struct pipe_context *pipe,
+		     const struct pipe_blend_color *bcol)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	nv10->blend_color = (struct pipe_blend_color*)bcol;
+
+	nv10->dirty |= NV10_NEW_BLENDCOL;
+}
+
+static void
+nv10_set_clip_state(struct pipe_context *pipe,
+		    const struct pipe_clip_state *clip)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	draw_set_clip_state(nv10->draw, clip);
+}
+
+static void
+nv10_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
+			 const struct pipe_constant_buffer *buf )
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+	struct pipe_winsys *ws = pipe->winsys;
+
+	assert(shader < PIPE_SHADER_TYPES);
+	assert(index == 0);
+
+	if (buf) {
+		void *mapped;
+		if (buf->buffer && buf->buffer->size &&
+                    (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
+		{
+			memcpy(nv10->constbuf[shader], mapped, buf->buffer->size);
+			nv10->constbuf_nr[shader] =
+				buf->buffer->size / (4 * sizeof(float));
+			ws->buffer_unmap(ws, buf->buffer);
+		}
+	}
+}
+
+static void
+nv10_set_framebuffer_state(struct pipe_context *pipe,
+			   const struct pipe_framebuffer_state *fb)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	nv10->framebuffer = (struct pipe_framebuffer_state*)fb;
+
+	nv10->dirty |= NV10_NEW_FRAMEBUFFER;
+}
+
+static void
+nv10_set_polygon_stipple(struct pipe_context *pipe,
+			 const struct pipe_poly_stipple *stipple)
+{
+	NOUVEAU_ERR("line stipple hahaha\n");
+}
+
+static void
+nv10_set_scissor_state(struct pipe_context *pipe,
+		       const struct pipe_scissor_state *s)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	nv10->scissor = (struct pipe_scissor_state*)s;
+
+	nv10->dirty |= NV10_NEW_SCISSOR;
+}
+
+static void
+nv10_set_viewport_state(struct pipe_context *pipe,
+			const struct pipe_viewport_state *vpt)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	nv10->viewport = (struct pipe_viewport_state*)vpt;
+
+	draw_set_viewport_state(nv10->draw, nv10->viewport);
+
+	nv10->dirty |= NV10_NEW_VIEWPORT;
+}
+
+static void
+nv10_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
+			const struct pipe_vertex_buffer *vb)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	memcpy(nv10->vtxbuf, vb, sizeof(*vb) * count);
+	nv10->dirty |= NV10_NEW_VTXARRAYS;
+
+	draw_set_vertex_buffers(nv10->draw, count, vb);
+}
+
+static void
+nv10_set_vertex_elements(struct pipe_context *pipe, unsigned count,
+			 const struct pipe_vertex_element *ve)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+
+	memcpy(nv10->vtxelt, ve, sizeof(*ve) * count);
+	nv10->dirty |= NV10_NEW_VTXARRAYS;
+
+	draw_set_vertex_elements(nv10->draw, count, ve);
+}
+
+void
+nv10_init_state_functions(struct nv10_context *nv10)
+{
+	nv10->pipe.create_blend_state = nv10_blend_state_create;
+	nv10->pipe.bind_blend_state = nv10_blend_state_bind;
+	nv10->pipe.delete_blend_state = nv10_blend_state_delete;
+
+	nv10->pipe.create_sampler_state = nv10_sampler_state_create;
+	nv10->pipe.bind_sampler_states = nv10_sampler_state_bind;
+	nv10->pipe.delete_sampler_state = nv10_sampler_state_delete;
+	nv10->pipe.set_sampler_textures = nv10_set_sampler_texture;
+
+	nv10->pipe.create_rasterizer_state = nv10_rasterizer_state_create;
+	nv10->pipe.bind_rasterizer_state = nv10_rasterizer_state_bind;
+	nv10->pipe.delete_rasterizer_state = nv10_rasterizer_state_delete;
+
+	nv10->pipe.create_depth_stencil_alpha_state =
+		nv10_depth_stencil_alpha_state_create;
+	nv10->pipe.bind_depth_stencil_alpha_state =
+		nv10_depth_stencil_alpha_state_bind;
+	nv10->pipe.delete_depth_stencil_alpha_state =
+		nv10_depth_stencil_alpha_state_delete;
+
+	nv10->pipe.create_vs_state = nv10_vp_state_create;
+	nv10->pipe.bind_vs_state = nv10_vp_state_bind;
+	nv10->pipe.delete_vs_state = nv10_vp_state_delete;
+
+	nv10->pipe.create_fs_state = nv10_fp_state_create;
+	nv10->pipe.bind_fs_state = nv10_fp_state_bind;
+	nv10->pipe.delete_fs_state = nv10_fp_state_delete;
+
+	nv10->pipe.set_blend_color = nv10_set_blend_color;
+	nv10->pipe.set_clip_state = nv10_set_clip_state;
+	nv10->pipe.set_constant_buffer = nv10_set_constant_buffer;
+	nv10->pipe.set_framebuffer_state = nv10_set_framebuffer_state;
+	nv10->pipe.set_polygon_stipple = nv10_set_polygon_stipple;
+	nv10->pipe.set_scissor_state = nv10_set_scissor_state;
+	nv10->pipe.set_viewport_state = nv10_set_viewport_state;
+
+	nv10->pipe.set_vertex_buffers = nv10_set_vertex_buffers;
+	nv10->pipe.set_vertex_elements = nv10_set_vertex_elements;
+}
+
diff --git a/src/gallium/drivers/nv10/nv10_state.h b/src/gallium/drivers/nv10/nv10_state.h
new file mode 100644
index 0000000000..3a3fd0d4f4
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_state.h
@@ -0,0 +1,139 @@
+#ifndef __NV10_STATE_H__
+#define __NV10_STATE_H__
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_scan.h"
+
+struct nv10_blend_state {
+	uint32_t b_enable;
+	uint32_t b_srcfunc;
+	uint32_t b_dstfunc;
+
+	uint32_t c_mask;
+
+	uint32_t d_enable;
+};
+
+struct nv10_sampler_state {
+	uint32_t wrap;
+	uint32_t en;
+	uint32_t filt;
+	uint32_t bcol;
+};
+
+struct nv10_rasterizer_state {
+	uint32_t shade_model;
+
+	uint32_t line_width;
+	uint32_t line_smooth_en;
+
+	uint32_t point_size;
+
+	uint32_t poly_smooth_en;
+	
+	uint32_t poly_mode_front;
+	uint32_t poly_mode_back;
+
+	uint32_t front_face;
+	uint32_t cull_face;
+	uint32_t cull_face_en;
+
+	uint32_t point_sprite;
+
+	const struct pipe_rasterizer_state *templ;
+};
+
+struct nv10_vertex_program_exec {
+	uint32_t data[4];
+	boolean has_branch_offset;
+	int const_index;
+};
+
+struct nv10_vertex_program_data {
+	int index; /* immediates == -1 */
+	float value[4];
+};
+
+struct nv10_vertex_program {
+	const struct pipe_shader_state *pipe;
+
+	boolean translated;
+	struct nv10_vertex_program_exec *insns;
+	unsigned nr_insns;
+	struct nv10_vertex_program_data *consts;
+	unsigned nr_consts;
+
+	struct nouveau_resource *exec;
+	unsigned exec_start;
+	struct nouveau_resource *data;
+	unsigned data_start;
+	unsigned data_start_min;
+
+	uint32_t ir;
+	uint32_t or;
+};
+
+struct nv10_fragment_program_data {
+	unsigned offset;
+	unsigned index;
+};
+
+struct nv10_fragment_program {
+	struct pipe_shader_state pipe;
+	struct tgsi_shader_info info;
+
+	boolean translated;
+	boolean on_hw;
+	unsigned samplers;
+
+	uint32_t *insn;
+	int       insn_len;
+
+	struct nv10_fragment_program_data *consts;
+	unsigned nr_consts;
+
+	struct pipe_buffer *buffer;
+
+	uint32_t fp_control;
+	uint32_t fp_reg_control;
+};
+
+
+struct nv10_depth_stencil_alpha_state {
+	struct {
+		uint32_t func;
+		uint32_t write_enable;
+		uint32_t test_enable;
+	} depth;
+
+	struct {
+		uint32_t enable;
+		uint32_t wmask;
+		uint32_t func;
+		uint32_t ref;
+		uint32_t vmask;
+		uint32_t fail;
+		uint32_t zfail;
+		uint32_t zpass;
+	} stencil;
+
+	struct {
+		uint32_t enabled;
+		uint32_t func;
+		uint32_t ref;
+	} alpha;
+};
+
+struct nv10_miptree {
+	struct pipe_texture base;
+
+	struct pipe_buffer *buffer;
+	uint total_size;
+
+	struct {
+		uint pitch;
+		uint *image_offset;
+	} level[PIPE_MAX_TEXTURE_LEVELS];
+};
+
+#endif
diff --git a/src/gallium/drivers/nv10/nv10_state_emit.c b/src/gallium/drivers/nv10/nv10_state_emit.c
new file mode 100644
index 0000000000..5dec618b93
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_state_emit.c
@@ -0,0 +1,306 @@
+#include "nv10_context.h"
+#include "nv10_state.h"
+
+static void nv10_state_emit_blend(struct nv10_context* nv10)
+{
+	struct nv10_blend_state *b = nv10->blend;
+
+	BEGIN_RING(celsius, NV10TCL_DITHER_ENABLE, 1);
+	OUT_RING  (b->d_enable);
+
+	BEGIN_RING(celsius, NV10TCL_BLEND_FUNC_ENABLE, 3);
+	OUT_RING  (b->b_enable);
+	OUT_RING  (b->b_srcfunc);
+	OUT_RING  (b->b_dstfunc);
+
+	BEGIN_RING(celsius, NV10TCL_COLOR_MASK, 1);
+	OUT_RING  (b->c_mask);
+}
+
+static void nv10_state_emit_blend_color(struct nv10_context* nv10)
+{
+	struct pipe_blend_color *c = nv10->blend_color;
+
+	BEGIN_RING(celsius, NV10TCL_BLEND_COLOR, 1);
+	OUT_RING  ((float_to_ubyte(c->color[3]) << 24)|
+		   (float_to_ubyte(c->color[0]) << 16)|
+		   (float_to_ubyte(c->color[1]) << 8) |
+		   (float_to_ubyte(c->color[2]) << 0));
+}
+
+static void nv10_state_emit_rast(struct nv10_context* nv10)
+{
+	struct nv10_rasterizer_state *r = nv10->rast;
+
+	BEGIN_RING(celsius, NV10TCL_SHADE_MODEL, 2);
+	OUT_RING  (r->shade_model);
+	OUT_RING  (r->line_width);
+
+
+	BEGIN_RING(celsius, NV10TCL_POINT_SIZE, 1);
+	OUT_RING  (r->point_size);
+
+	BEGIN_RING(celsius, NV10TCL_POLYGON_MODE_FRONT, 2);
+	OUT_RING  (r->poly_mode_front);
+	OUT_RING  (r->poly_mode_back);
+
+
+	BEGIN_RING(celsius, NV10TCL_CULL_FACE, 2);
+	OUT_RING  (r->cull_face);
+	OUT_RING  (r->front_face);
+
+	BEGIN_RING(celsius, NV10TCL_LINE_SMOOTH_ENABLE, 2);
+	OUT_RING  (r->line_smooth_en);
+	OUT_RING  (r->poly_smooth_en);
+
+	BEGIN_RING(celsius, NV10TCL_CULL_FACE_ENABLE, 1);
+	OUT_RING  (r->cull_face_en);
+}
+
+static void nv10_state_emit_dsa(struct nv10_context* nv10)
+{
+	struct nv10_depth_stencil_alpha_state *d = nv10->dsa;
+
+	BEGIN_RING(celsius, NV10TCL_DEPTH_FUNC, 1);
+	OUT_RING (d->depth.func);
+
+	BEGIN_RING(celsius, NV10TCL_DEPTH_WRITE_ENABLE, 1);
+	OUT_RING (d->depth.write_enable);
+
+	BEGIN_RING(celsius, NV10TCL_DEPTH_TEST_ENABLE, 1);
+	OUT_RING (d->depth.test_enable);
+
+#if 0
+	BEGIN_RING(celsius, NV10TCL_STENCIL_ENABLE, 1);
+	OUT_RING (d->stencil.enable);
+	BEGIN_RING(celsius, NV10TCL_STENCIL_MASK, 7);
+	OUT_RINGp ((uint32_t *)&(d->stencil.wmask), 7);
+#endif
+
+	BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_ENABLE, 1);
+	OUT_RING (d->alpha.enabled);
+
+	BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_FUNC, 1);
+	OUT_RING (d->alpha.func);
+
+	BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_REF, 1);
+	OUT_RING (d->alpha.ref);
+}
+
+static void nv10_state_emit_viewport(struct nv10_context* nv10)
+{
+}
+
+static void nv10_state_emit_scissor(struct nv10_context* nv10)
+{
+	// XXX this is so not working
+/*	struct pipe_scissor_state *s = nv10->scissor;
+	BEGIN_RING(celsius, NV10TCL_SCISSOR_HORIZ, 2);
+	OUT_RING  (((s->maxx - s->minx) << 16) | s->minx);
+	OUT_RING  (((s->maxy - s->miny) << 16) | s->miny);*/
+}
+
+static void nv10_state_emit_framebuffer(struct nv10_context* nv10)
+{
+	struct pipe_framebuffer_state* fb = nv10->framebuffer;
+	struct pipe_surface *rt, *zeta = NULL;
+	uint32_t rt_format, w, h;
+	int colour_format = 0, zeta_format = 0;
+        struct nv10_miptree *nv10mt = 0;
+
+	w = fb->cbufs[0]->width;
+	h = fb->cbufs[0]->height;
+	colour_format = fb->cbufs[0]->format;
+	rt = fb->cbufs[0];
+
+	if (fb->zsbuf) {
+		if (colour_format) {
+			assert(w == fb->zsbuf->width);
+			assert(h == fb->zsbuf->height);
+		} else {
+			w = fb->zsbuf->width;
+			h = fb->zsbuf->height;
+		}
+
+		zeta_format = fb->zsbuf->format;
+		zeta = fb->zsbuf;
+	}
+
+	rt_format = NV10TCL_RT_FORMAT_TYPE_LINEAR;
+
+	switch (colour_format) {
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case 0:
+		rt_format |= NV10TCL_RT_FORMAT_COLOR_A8R8G8B8;
+		break;
+	case PIPE_FORMAT_R5G6B5_UNORM:
+		rt_format |= NV10TCL_RT_FORMAT_COLOR_R5G6B5;
+		break;
+	default:
+		assert(0);
+	}
+
+	if (zeta) {
+		BEGIN_RING(celsius, NV10TCL_RT_PITCH, 1);
+		OUT_RING  (rt->stride | (zeta->stride << 16));
+	} else {
+		BEGIN_RING(celsius, NV10TCL_RT_PITCH, 1);
+		OUT_RING  (rt->stride | (rt->stride << 16));
+	}
+
+	nv10mt = (struct nv10_miptree *)rt->texture;
+	nv10->rt[0] = nv10mt->buffer;
+
+	if (zeta_format)
+	{
+		nv10mt = (struct nv10_miptree *)zeta->texture;
+		nv10->zeta = nv10mt->buffer;
+	}
+
+	BEGIN_RING(celsius, NV10TCL_RT_HORIZ, 3);
+	OUT_RING  ((w << 16) | 0);
+	OUT_RING  ((h << 16) | 0);
+	OUT_RING  (rt_format);
+	BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_HORIZ(0), 2);
+	OUT_RING  (((w - 1) << 16) | 0 | 0x08000800);
+	OUT_RING  (((h - 1) << 16) | 0 | 0x08000800);
+}
+
+static void nv10_vertex_layout(struct nv10_context *nv10)
+{
+	struct nv10_fragment_program *fp = nv10->fragprog.current;
+	uint32_t src = 0;
+	int i;
+	struct vertex_info vinfo;
+
+	memset(&vinfo, 0, sizeof(vinfo));
+
+	for (i = 0; i < fp->info.num_inputs; i++) {
+		switch (fp->info.input_semantic_name[i]) {
+			case TGSI_SEMANTIC_POSITION:
+				draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src++);
+				break;
+			case TGSI_SEMANTIC_COLOR:
+				draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src++);
+				break;
+			default:
+			case TGSI_SEMANTIC_GENERIC:
+				draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++);
+				break;
+			case TGSI_SEMANTIC_FOG:
+				draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++);
+				break;
+		}
+	}
+	draw_compute_vertex_size(&vinfo);
+}
+
+void
+nv10_emit_hw_state(struct nv10_context *nv10)
+{
+	int i;
+
+	if (nv10->dirty & NV10_NEW_VERTPROG) {
+		//nv10_vertprog_bind(nv10, nv10->vertprog.current);
+		nv10->dirty &= ~NV10_NEW_VERTPROG;
+	}
+
+	if (nv10->dirty & NV10_NEW_FRAGPROG) {
+		nv10_fragprog_bind(nv10, nv10->fragprog.current);
+		/*XXX: clear NV10_NEW_FRAGPROG if no new program uploaded */
+		nv10->dirty_samplers |= (1<<10);
+		nv10->dirty_samplers = 0;
+	}
+
+	if (nv10->dirty_samplers || (nv10->dirty & NV10_NEW_FRAGPROG)) {
+		nv10_fragtex_bind(nv10);
+		nv10->dirty &= ~NV10_NEW_FRAGPROG;
+	}
+
+	if (nv10->dirty & NV10_NEW_VTXARRAYS) {
+		nv10->dirty &= ~NV10_NEW_VTXARRAYS;
+		nv10_vertex_layout(nv10);
+		nv10_vtxbuf_bind(nv10);
+	}
+
+	if (nv10->dirty & NV10_NEW_BLEND) {
+		nv10->dirty &= ~NV10_NEW_BLEND;
+		nv10_state_emit_blend(nv10);
+	}
+
+	if (nv10->dirty & NV10_NEW_BLENDCOL) {
+		nv10->dirty &= ~NV10_NEW_BLENDCOL;
+		nv10_state_emit_blend_color(nv10);
+	}
+
+	if (nv10->dirty & NV10_NEW_RAST) {
+		nv10->dirty &= ~NV10_NEW_RAST;
+		nv10_state_emit_rast(nv10);
+	}
+
+	if (nv10->dirty & NV10_NEW_DSA) {
+		nv10->dirty &= ~NV10_NEW_DSA;
+		nv10_state_emit_dsa(nv10);
+	}
+
+ 	if (nv10->dirty & NV10_NEW_VIEWPORT) {
+		nv10->dirty &= ~NV10_NEW_VIEWPORT;
+		nv10_state_emit_viewport(nv10);
+	}
+
+ 	if (nv10->dirty & NV10_NEW_SCISSOR) {
+		nv10->dirty &= ~NV10_NEW_SCISSOR;
+		nv10_state_emit_scissor(nv10);
+	}
+
+ 	if (nv10->dirty & NV10_NEW_FRAMEBUFFER) {
+		nv10->dirty &= ~NV10_NEW_FRAMEBUFFER;
+		nv10_state_emit_framebuffer(nv10);
+	}
+
+	/* Emit relocs for every referenced buffer.
+	 * This is to ensure the bufmgr has an accurate idea of how
+	 * the buffer is used.  This isn't very efficient, but we don't
+	 * seem to take a significant performance hit.  Will be improved
+	 * at some point.  Vertex arrays are emitted by nv10_vbo.c
+	 */
+
+	/* Render target */
+// XXX figre out who's who for NV10TCL_DMA_* and fill accordingly
+//	BEGIN_RING(celsius, NV10TCL_DMA_COLOR0, 1);
+//	OUT_RELOCo(nv10->rt[0], NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	BEGIN_RING(celsius, NV10TCL_COLOR_OFFSET, 1);
+	OUT_RELOCl(nv10->rt[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+
+	if (nv10->zeta) {
+// XXX
+//		BEGIN_RING(celsius, NV10TCL_DMA_ZETA, 1);
+//		OUT_RELOCo(nv10->zeta, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+		BEGIN_RING(celsius, NV10TCL_ZETA_OFFSET, 1);
+		OUT_RELOCl(nv10->zeta, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+		/* XXX for when we allocate LMA on nv17 */
+/*		BEGIN_RING(celsius, NV10TCL_LMA_DEPTH_BUFFER_OFFSET, 1);
+		OUT_RELOCl(nv10->zeta + lma_offset);*/
+	}
+
+	/* Vertex buffer */
+	BEGIN_RING(celsius, NV10TCL_DMA_VTXBUF0, 1);
+	OUT_RELOCo(nv10->rt[0], NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	BEGIN_RING(celsius, NV10TCL_COLOR_OFFSET, 1);
+	OUT_RELOCl(nv10->rt[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+
+	/* Texture images */
+	for (i = 0; i < 2; i++) {
+		if (!(nv10->fp_samplers & (1 << i)))
+			continue;
+		BEGIN_RING(celsius, NV10TCL_TX_OFFSET(i), 1);
+		OUT_RELOCl(nv10->tex[i].buffer, 0, NOUVEAU_BO_VRAM |
+			   NOUVEAU_BO_GART | NOUVEAU_BO_RD);
+		BEGIN_RING(celsius, NV10TCL_TX_FORMAT(i), 1);
+		OUT_RELOCd(nv10->tex[i].buffer, nv10->tex[i].format,
+			   NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD |
+			   NOUVEAU_BO_OR, NV10TCL_TX_FORMAT_DMA0,
+			   NV10TCL_TX_FORMAT_DMA1);
+	}
+}
+
diff --git a/src/gallium/drivers/nv10/nv10_surface.c b/src/gallium/drivers/nv10/nv10_surface.c
new file mode 100644
index 0000000000..2538151063
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_surface.c
@@ -0,0 +1,72 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "nv10_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+
+static void
+nv10_surface_copy(struct pipe_context *pipe, boolean do_flip,
+		  struct pipe_surface *dest, unsigned destx, unsigned desty,
+		  struct pipe_surface *src, unsigned srcx, unsigned srcy,
+		  unsigned width, unsigned height)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+	struct nv04_surface_2d *eng2d = nv10->screen->eng2d;
+
+	if (do_flip) {
+		desty += height;
+		while (height--) {
+			eng2d->copy(eng2d, dest, destx, desty--, src,
+				    srcx, srcy++, width, 1);
+		}
+		return;
+	}
+
+	eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height);
+}
+
+static void
+nv10_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
+		  unsigned destx, unsigned desty, unsigned width,
+		  unsigned height, unsigned value)
+{
+	struct nv10_context *nv10 = nv10_context(pipe);
+	struct nv04_surface_2d *eng2d = nv10->screen->eng2d;
+
+	eng2d->fill(eng2d, dest, destx, desty, width, height, value);
+}
+
+void
+nv10_init_surface_functions(struct nv10_context *nv10)
+{
+	nv10->pipe.surface_copy = nv10_surface_copy;
+	nv10->pipe.surface_fill = nv10_surface_fill;
+}
diff --git a/src/gallium/drivers/nv10/nv10_vbo.c b/src/gallium/drivers/nv10/nv10_vbo.c
new file mode 100644
index 0000000000..d0e788ac03
--- /dev/null
+++ b/src/gallium/drivers/nv10/nv10_vbo.c
@@ -0,0 +1,77 @@
+#include "draw/draw_context.h"
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+
+#include "nv10_context.h"
+#include "nv10_state.h"
+
+#include "nouveau/nouveau_channel.h"
+#include "nouveau/nouveau_pushbuf.h"
+
+boolean nv10_draw_elements( struct pipe_context *pipe,
+                    struct pipe_buffer *indexBuffer,
+                    unsigned indexSize,
+                    unsigned prim, unsigned start, unsigned count)
+{
+	struct nv10_context *nv10 = nv10_context( pipe );
+	struct draw_context *draw = nv10->draw;
+	unsigned i;
+
+	nv10_emit_hw_state(nv10);
+
+	/*
+	 * Map vertex buffers
+	 */
+	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
+		if (nv10->vtxbuf[i].buffer) {
+			void *buf
+				= pipe->winsys->buffer_map(pipe->winsys,
+						nv10->vtxbuf[i].buffer,
+						PIPE_BUFFER_USAGE_CPU_READ);
+			draw_set_mapped_vertex_buffer(draw, i, buf);
+		}
+	}
+	/* Map index buffer, if present */
+	if (indexBuffer) {
+		void *mapped_indexes
+			= pipe->winsys->buffer_map(pipe->winsys, indexBuffer,
+					PIPE_BUFFER_USAGE_CPU_READ);
+		draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
+	}
+	else {
+		/* no index/element buffer */
+		draw_set_mapped_element_buffer(draw, 0, NULL);
+	}
+
+	draw_set_mapped_constant_buffer(draw,
+					nv10->constbuf[PIPE_SHADER_VERTEX],
+					nv10->constbuf_nr[PIPE_SHADER_VERTEX]);
+
+	/* draw! */
+	draw_arrays(nv10->draw, prim, start, count);
+
+	/*
+	 * unmap vertex/index buffers
+	 */
+	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
+		if (nv10->vtxbuf[i].buffer) {
+			pipe->winsys->buffer_unmap(pipe->winsys, nv10->vtxbuf[i].buffer);
+			draw_set_mapped_vertex_buffer(draw, i, NULL);
+		}
+	}
+	if (indexBuffer) {
+		pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer);
+		draw_set_mapped_element_buffer(draw, 0, NULL);
+	}
+
+	return TRUE;
+}
+
+boolean nv10_draw_arrays( struct pipe_context *pipe,
+				 unsigned prim, unsigned start, unsigned count)
+{
+	return nv10_draw_elements(pipe, NULL, 0, prim, start, count);
+}
+
+
+
diff --git a/src/gallium/drivers/nv20/Makefile b/src/gallium/drivers/nv20/Makefile
new file mode 100644
index 0000000000..d777fd3d8b
--- /dev/null
+++ b/src/gallium/drivers/nv20/Makefile
@@ -0,0 +1,29 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nv20
+
+DRIVER_SOURCES = \
+	nv20_clear.c \
+	nv20_context.c \
+	nv20_fragprog.c \
+	nv20_fragtex.c \
+	nv20_miptree.c \
+	nv20_prim_vbuf.c \
+	nv20_screen.c \
+	nv20_state.c \
+	nv20_state_emit.c \
+	nv20_surface.c \
+	nv20_vbo.c
+#	nv20_vertprog.c
+
+C_SOURCES = \
+	$(COMMON_SOURCES) \
+	$(DRIVER_SOURCES)
+
+ASM_SOURCES = 
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/drivers/nv20/nv20_clear.c b/src/gallium/drivers/nv20/nv20_clear.c
new file mode 100644
index 0000000000..29f4afd87c
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_clear.c
@@ -0,0 +1,13 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "nv20_context.h"
+
+void
+nv20_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+	   unsigned clearValue)
+{
+	pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue);
+	ps->status = PIPE_SURFACE_STATUS_CLEAR;
+}
diff --git a/src/gallium/drivers/nv20/nv20_context.c b/src/gallium/drivers/nv20/nv20_context.c
new file mode 100644
index 0000000000..1659aec8fa
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_context.c
@@ -0,0 +1,419 @@
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+#include "nv20_context.h"
+#include "nv20_screen.h"
+
+static void
+nv20_flush(struct pipe_context *pipe, unsigned flags,
+	   struct pipe_fence_handle **fence)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	draw_flush(nv20->draw);
+
+	FIRE_RING(fence);
+}
+
+static void
+nv20_destroy(struct pipe_context *pipe)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	if (nv20->draw)
+		draw_destroy(nv20->draw);
+
+	FREE(nv20);
+}
+
+static void nv20_init_hwctx(struct nv20_context *nv20)
+{
+	struct nv20_screen *screen = nv20->screen;
+	struct nouveau_winsys *nvws = screen->nvws;
+	int i;
+	float projectionmatrix[16];
+	const boolean is_nv25tcl = (nv20->screen->kelvin->grclass == NV25TCL);
+
+	BEGIN_RING(kelvin, NV20TCL_DMA_NOTIFY, 1);
+	OUT_RING  (screen->sync->handle);
+	BEGIN_RING(kelvin, NV20TCL_DMA_TEXTURE0, 2);
+	OUT_RING  (nvws->channel->vram->handle);
+	OUT_RING  (nvws->channel->gart->handle); /* TEXTURE1 */
+	BEGIN_RING(kelvin, NV20TCL_DMA_COLOR, 2);
+	OUT_RING  (nvws->channel->vram->handle);
+	OUT_RING  (nvws->channel->vram->handle); /* ZETA */
+
+	BEGIN_RING(kelvin, NV20TCL_DMA_QUERY, 1);
+	OUT_RING  (0); /* renouveau: beef0351, unique */
+
+	BEGIN_RING(kelvin, NV20TCL_RT_HORIZ, 2);
+	OUT_RING  (0);
+	OUT_RING  (0);
+
+	BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_HORIZ(0), 1);
+	OUT_RING  ((0xfff << 16) | 0x0);
+	BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_VERT(0), 1);
+	OUT_RING  ((0xfff << 16) | 0x0);
+
+	for (i = 1; i < NV20TCL_VIEWPORT_CLIP_HORIZ__SIZE; i++) {
+		BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_HORIZ(i), 1);
+		OUT_RING  (0);
+		BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_VERT(i), 1);
+		OUT_RING  (0);
+	}
+
+	BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_MODE, 1);
+	OUT_RING  (0);
+
+	BEGIN_RING(kelvin, 0x17e0, 3);
+	OUT_RINGf (0.0);
+	OUT_RINGf (0.0);
+	OUT_RINGf (1.0);
+
+	if (is_nv25tcl) {
+		BEGIN_RING(kelvin, NV20TCL_TX_RCOMP, 1);
+		OUT_RING  (NV20TCL_TX_RCOMP_LEQUAL | 0xdb0);
+	} else {
+		BEGIN_RING(kelvin, 0x1e68, 1);
+		OUT_RING  (0x4b800000); /* 16777216.000000 */
+		BEGIN_RING(kelvin, NV20TCL_TX_RCOMP, 1);
+		OUT_RING  (NV20TCL_TX_RCOMP_LEQUAL);
+	}
+
+	BEGIN_RING(kelvin, 0x290, 1);
+	OUT_RING  ((0x10 << 16) | 1);
+	BEGIN_RING(kelvin, 0x9fc, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, 0x1d80, 1);
+	OUT_RING  (1);
+	BEGIN_RING(kelvin, 0x9f8, 1);
+	OUT_RING  (4);
+	BEGIN_RING(kelvin, 0x17ec, 3);
+	OUT_RINGf (0.0);
+	OUT_RINGf (1.0);
+	OUT_RINGf (0.0);
+
+	if (is_nv25tcl) {
+		BEGIN_RING(kelvin, 0x1d88, 1);
+		OUT_RING  (3);
+
+		BEGIN_RING(kelvin, NV25TCL_DMA_IN_MEMORY9, 1);
+		OUT_RING  (nvws->channel->vram->handle);
+		BEGIN_RING(kelvin, NV25TCL_DMA_IN_MEMORY8, 1);
+		OUT_RING  (nvws->channel->vram->handle);
+	}
+	BEGIN_RING(kelvin, NV20TCL_DMA_FENCE, 1);
+	OUT_RING  (0);	/* renouveau: beef1e10 */
+
+	BEGIN_RING(kelvin, 0x1e98, 1);
+	OUT_RING  (0);
+#if 0
+	if (is_nv25tcl) {
+		BEGIN_RING(NvSub3D, NV25TCL_DMA_IN_MEMORY4, 2);
+		OUT_RING  (NvDmaTT);	/* renouveau: beef0202 */
+		OUT_RING  (NvDmaFB);	/* renouveau: beef0201 */
+
+		BEGIN_RING(NvSub3D, NV20TCL_DMA_TEXTURE1, 1);
+		OUT_RING  (NvDmaTT);	/* renouveau: beef0202 */
+	}
+#endif
+	BEGIN_RING(kelvin, NV20TCL_NOTIFY, 1);
+	OUT_RING  (0);
+
+	BEGIN_RING(kelvin, 0x120, 3);
+	OUT_RING  (0);
+	OUT_RING  (1);
+	OUT_RING  (2);
+
+/* error: ILLEGAL_MTHD, PROTECTION_FAULT
+	BEGIN_RING(kelvin, NV20TCL_VIEWPORT_TRANSLATE_X, 4);
+	OUT_RINGf (0.0);
+	OUT_RINGf (512.0);
+	OUT_RINGf (0.0);
+	OUT_RINGf (0.0);
+*/
+
+	if (is_nv25tcl) {
+		BEGIN_RING(kelvin, 0x022c, 2);
+		OUT_RING  (0x280);
+		OUT_RING  (0x07d28000);
+	}
+
+/* * illegal method, protection fault
+	BEGIN_RING(NvSub3D, 0x1c2c, 1);
+	OUT_RING  (0); */
+
+	if (is_nv25tcl) {
+		BEGIN_RING(kelvin, 0x1da4, 1);
+		OUT_RING  (0);
+	}
+
+/* * crashes with illegal method, protection fault
+	BEGIN_RING(NvSub3D, 0x1c18, 1);
+	OUT_RING  (0x200); */
+
+	BEGIN_RING(kelvin, NV20TCL_RT_HORIZ, 2);
+	OUT_RING  ((0 << 16) | 0);
+	OUT_RING  ((0 << 16) | 0);
+
+	/* *** Set state *** */
+
+	BEGIN_RING(kelvin, NV20TCL_ALPHA_FUNC_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_ALPHA_FUNC_FUNC, 2);
+	OUT_RING  (NV20TCL_ALPHA_FUNC_FUNC_ALWAYS);
+	OUT_RING  (0);			/* NV20TCL_ALPHA_FUNC_REF */
+
+	for (i = 0; i < NV20TCL_TX_ENABLE__SIZE; ++i) {
+		BEGIN_RING(kelvin, NV20TCL_TX_ENABLE(i), 1);
+		OUT_RING  (0);
+	}
+	BEGIN_RING(kelvin, NV20TCL_TX_SHADER_OP, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_TX_SHADER_CULL_MODE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_RC_IN_ALPHA(0), 4);
+	OUT_RING  (0x30d410d0);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_RC_OUT_RGB(0), 4);
+	OUT_RING  (0x00000c00);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_RC_ENABLE, 1);
+	OUT_RING  (0x00011101);
+	BEGIN_RING(kelvin, NV20TCL_RC_FINAL0, 2);
+	OUT_RING  (0x130e0300);
+	OUT_RING  (0x0c091c80);
+	BEGIN_RING(kelvin, NV20TCL_RC_OUT_ALPHA(0), 4);
+	OUT_RING  (0x00000c00);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_RC_IN_RGB(0), 4);
+	OUT_RING  (0x20c400c0);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_RC_COLOR0, 2);
+	OUT_RING  (0);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_RC_CONSTANT_COLOR0(0), 4);
+	OUT_RING  (0x035125a0);
+	OUT_RING  (0);
+	OUT_RING  (0x40002000);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_MULTISAMPLE_CONTROL, 1);
+	OUT_RING  (0xffff0000);
+
+	BEGIN_RING(kelvin, NV20TCL_BLEND_FUNC_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_DITHER_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_STENCIL_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_BLEND_FUNC_SRC, 4);
+	OUT_RING  (NV20TCL_BLEND_FUNC_SRC_ONE);
+	OUT_RING  (NV20TCL_BLEND_FUNC_DST_ZERO);
+	OUT_RING  (0);			/* NV20TCL_BLEND_COLOR */
+	OUT_RING  (NV20TCL_BLEND_EQUATION_FUNC_ADD);
+	BEGIN_RING(kelvin, NV20TCL_STENCIL_MASK, 7);
+	OUT_RING  (0xff);
+	OUT_RING  (NV20TCL_STENCIL_FUNC_FUNC_ALWAYS);
+	OUT_RING  (0);			/* NV20TCL_STENCIL_FUNC_REF */
+	OUT_RING  (0xff);		/* NV20TCL_STENCIL_FUNC_MASK */
+	OUT_RING  (NV20TCL_STENCIL_OP_FAIL_KEEP);
+	OUT_RING  (NV20TCL_STENCIL_OP_ZFAIL_KEEP);
+	OUT_RING  (NV20TCL_STENCIL_OP_ZPASS_KEEP);
+
+	BEGIN_RING(kelvin, NV20TCL_COLOR_LOGIC_OP_ENABLE, 2);
+	OUT_RING  (0);
+	OUT_RING  (NV20TCL_COLOR_LOGIC_OP_OP_COPY);
+	BEGIN_RING(kelvin, 0x17cc, 1);
+	OUT_RING  (0);
+	if (is_nv25tcl) {
+		BEGIN_RING(kelvin, 0x1d84, 1);
+		OUT_RING  (1);
+	}
+	BEGIN_RING(kelvin, NV20TCL_LIGHTING_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_LIGHT_CONTROL, 1);
+	OUT_RING  (0x00020000);
+	BEGIN_RING(kelvin, NV20TCL_SEPARATE_SPECULAR_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_LIGHT_MODEL_TWO_SIDE_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_ENABLED_LIGHTS, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_NORMALIZE_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_POLYGON_STIPPLE_PATTERN(0),
+					NV20TCL_POLYGON_STIPPLE_PATTERN__SIZE);
+	for (i = 0; i < NV20TCL_POLYGON_STIPPLE_PATTERN__SIZE; ++i) {
+		OUT_RING(0xffffffff);
+	}
+
+	BEGIN_RING(kelvin, NV20TCL_POLYGON_OFFSET_POINT_ENABLE, 3);
+	OUT_RING  (0);
+	OUT_RING  (0);		/* NV20TCL.POLYGON_OFFSET_LINE_ENABLE */
+	OUT_RING  (0);		/* NV20TCL.POLYGON_OFFSET_FILL_ENABLE */
+	BEGIN_RING(kelvin, NV20TCL_DEPTH_FUNC, 1);
+	OUT_RING  (NV20TCL_DEPTH_FUNC_LESS);
+	BEGIN_RING(kelvin, NV20TCL_DEPTH_WRITE_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_DEPTH_TEST_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_POLYGON_OFFSET_FACTOR, 2);
+	OUT_RINGf (0.0);
+	OUT_RINGf (0.0);	/* NV20TCL.POLYGON_OFFSET_UNITS */
+	BEGIN_RING(kelvin, NV20TCL_DEPTH_UNK17D8, 1);
+	OUT_RING  (1);
+	if (!is_nv25tcl) {
+		BEGIN_RING(kelvin, 0x1d84, 1);
+		OUT_RING  (3);
+	}
+	BEGIN_RING(kelvin, NV20TCL_POINT_SIZE, 1);
+	if (!is_nv25tcl) {
+		OUT_RING  (8);
+	} else {
+		OUT_RINGf (1.0);
+	}
+	if (!is_nv25tcl) {
+		BEGIN_RING(kelvin, NV20TCL_POINT_PARAMETERS_ENABLE, 2);
+		OUT_RING  (0);
+		OUT_RING  (0);		/* NV20TCL.POINT_SMOOTH_ENABLE */
+	} else {
+		BEGIN_RING(kelvin, NV20TCL_POINT_PARAMETERS_ENABLE, 1);
+		OUT_RING  (0);
+		BEGIN_RING(kelvin, 0x0a1c, 1);
+		OUT_RING  (0x800);
+	}
+	BEGIN_RING(kelvin, NV20TCL_LINE_WIDTH, 1);
+	OUT_RING  (8);
+	BEGIN_RING(kelvin, NV20TCL_LINE_SMOOTH_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_POLYGON_MODE_FRONT, 2);
+	OUT_RING  (NV20TCL_POLYGON_MODE_FRONT_FILL);
+	OUT_RING  (NV20TCL_POLYGON_MODE_BACK_FILL);
+	BEGIN_RING(kelvin, NV20TCL_CULL_FACE, 2);
+	OUT_RING  (NV20TCL_CULL_FACE_BACK);
+	OUT_RING  (NV20TCL_FRONT_FACE_CCW);
+	BEGIN_RING(kelvin, NV20TCL_POLYGON_SMOOTH_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_CULL_FACE_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_SHADE_MODEL, 1);
+	OUT_RING  (NV20TCL_SHADE_MODEL_SMOOTH);
+	BEGIN_RING(kelvin, NV20TCL_POLYGON_STIPPLE_ENABLE, 1);
+	OUT_RING  (0);
+	BEGIN_RING(kelvin, NV20TCL_TX_GEN_S(0), 4 * NV20TCL_TX_GEN_S__SIZE);
+	for (i=0; i < 4 * NV20TCL_TX_GEN_S__SIZE; ++i) {
+		OUT_RING(0);
+	}
+	BEGIN_RING(kelvin, NV20TCL_FOG_EQUATION_CONSTANT, 3);
+	OUT_RINGf (1.5);
+	OUT_RINGf (-0.090168);		/* NV20TCL.FOG_EQUATION_LINEAR */
+	OUT_RINGf (0.0);		/* NV20TCL.FOG_EQUATION_QUADRATIC */
+	BEGIN_RING(kelvin, NV20TCL_FOG_MODE, 2);
+	OUT_RING  (NV20TCL_FOG_MODE_EXP_2);
+	OUT_RING  (NV20TCL_FOG_COORD_DIST_COORD_FOG);
+	BEGIN_RING(kelvin, NV20TCL_FOG_ENABLE, 2);
+	OUT_RING  (0);
+	OUT_RING  (0);			/* NV20TCL.FOG_COLOR */
+	BEGIN_RING(kelvin, NV20TCL_ENGINE, 1);
+	OUT_RING  (NV20TCL_ENGINE_FIXED);
+
+	for (i = 0; i < NV20TCL_TX_MATRIX_ENABLE__SIZE; ++i) {
+		BEGIN_RING(kelvin, NV20TCL_TX_MATRIX_ENABLE(i), 1);
+		OUT_RING  (0);
+	}
+
+	BEGIN_RING(kelvin, NV20TCL_VTX_ATTR_4F_X(1), 4 * 15);
+	OUT_RINGf(1.0); OUT_RINGf(0.0); OUT_RINGf(0.0); OUT_RINGf(1.0);
+	OUT_RINGf(0.0); OUT_RINGf(0.0); OUT_RINGf(1.0); OUT_RINGf(1.0);
+	OUT_RINGf(1.0); OUT_RINGf(1.0); OUT_RINGf(1.0); OUT_RINGf(1.0);
+	for (i = 4; i < 16; ++i) {
+		OUT_RINGf(0.0); OUT_RINGf(0.0); OUT_RINGf(0.0);	OUT_RINGf(1.0);
+	}
+
+	BEGIN_RING(kelvin, NV20TCL_EDGEFLAG_ENABLE, 1);
+	OUT_RING  (1);
+	BEGIN_RING(kelvin, NV20TCL_COLOR_MASK, 1);
+	OUT_RING (0x00010101);
+	BEGIN_RING(kelvin, NV20TCL_CLEAR_VALUE, 1);
+	OUT_RING (0);
+
+	memset(projectionmatrix, 0, sizeof(projectionmatrix));
+	projectionmatrix[0*4+0] = 1.0;
+	projectionmatrix[1*4+1] = 1.0;
+	projectionmatrix[2*4+2] = 16777215.0;
+	projectionmatrix[3*4+3] = 1.0;
+	BEGIN_RING(kelvin, NV20TCL_PROJECTION_MATRIX(0), 16);
+	for (i = 0; i < 16; i++) {
+		OUT_RINGf  (projectionmatrix[i]);
+	}
+
+	BEGIN_RING(kelvin, NV20TCL_DEPTH_RANGE_NEAR, 2);
+	OUT_RINGf (0.0);
+	OUT_RINGf (16777216.0); /* [0, 1] scaled approx to [0, 2^24] */
+
+	BEGIN_RING(kelvin, NV20TCL_VIEWPORT_SCALE0_X, 4);
+	OUT_RINGf (0.0); /* x-offset, w/2 + 1.031250 */
+	OUT_RINGf (0.0); /* y-offset, h/2 + 0.030762 */
+	OUT_RINGf (0.0);
+	OUT_RINGf (16777215.0);
+
+	BEGIN_RING(kelvin, NV20TCL_VIEWPORT_SCALE1_X, 4);
+	OUT_RINGf (0.0); /* no effect?, w/2 */
+	OUT_RINGf (0.0); /* no effect?, h/2 */
+	OUT_RINGf (16777215.0 * 0.5);
+	OUT_RINGf (65535.0);
+
+	FIRE_RING (NULL);
+}
+
+static void
+nv20_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield)
+{
+}
+
+struct pipe_context *
+nv20_create(struct pipe_screen *pscreen, unsigned pctx_id)
+{
+	struct nv20_screen *screen = nv20_screen(pscreen);
+	struct pipe_winsys *ws = pscreen->winsys;
+	struct nv20_context *nv20;
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nv20 = CALLOC(1, sizeof(struct nv20_context));
+	if (!nv20)
+		return NULL;
+	nv20->screen = screen;
+	nv20->pctx_id = pctx_id;
+
+	nv20->nvws = nvws;
+
+	nv20->pipe.winsys = ws;
+	nv20->pipe.screen = pscreen;
+	nv20->pipe.destroy = nv20_destroy;
+	nv20->pipe.set_edgeflags = nv20_set_edgeflags;
+	nv20->pipe.draw_arrays = nv20_draw_arrays;
+	nv20->pipe.draw_elements = nv20_draw_elements;
+	nv20->pipe.clear = nv20_clear;
+	nv20->pipe.flush = nv20_flush;
+
+	nv20_init_surface_functions(nv20);
+	nv20_init_state_functions(nv20);
+
+	nv20->draw = draw_create();
+	assert(nv20->draw);
+	draw_set_rasterize_stage(nv20->draw, nv20_draw_vbuf_stage(nv20));
+
+	nv20_init_hwctx(nv20);
+
+	return &nv20->pipe;
+}
+
diff --git a/src/gallium/drivers/nv20/nv20_context.h b/src/gallium/drivers/nv20/nv20_context.h
new file mode 100644
index 0000000000..8ad926db20
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_context.h
@@ -0,0 +1,153 @@
+#ifndef __NV20_CONTEXT_H__
+#define __NV20_CONTEXT_H__
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_compiler.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "draw/draw_vertex.h"
+
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau/nouveau_gldefs.h"
+
+#define NOUVEAU_PUSH_CONTEXT(ctx)                                              \
+	struct nv20_screen *ctx = nv20->screen
+#include "nouveau/nouveau_push.h"
+
+#include "nv20_state.h"
+
+#define NOUVEAU_ERR(fmt, args...) \
+	fprintf(stderr, "%s:%d -  "fmt, __func__, __LINE__, ##args);
+#define NOUVEAU_MSG(fmt, args...) \
+	fprintf(stderr, "nouveau: "fmt, ##args);
+
+#define NV20_NEW_VERTPROG	(1 << 0)
+#define NV20_NEW_FRAGPROG	(1 << 1)
+#define NV20_NEW_VTXARRAYS	(1 << 2)
+#define NV20_NEW_BLEND		(1 << 3)
+#define NV20_NEW_BLENDCOL	(1 << 4)
+#define NV20_NEW_RAST 		(1 << 5)
+#define NV20_NEW_DSA  		(1 << 6)
+#define NV20_NEW_VIEWPORT	(1 << 7)
+#define NV20_NEW_SCISSOR	(1 << 8)
+#define NV20_NEW_FRAMEBUFFER	(1 << 9)
+
+#include "nv20_screen.h"
+
+struct nv20_context {
+	struct pipe_context pipe;
+
+	struct nouveau_winsys *nvws;
+	struct nv20_screen *screen;
+	unsigned pctx_id;
+
+	struct draw_context *draw;
+
+	uint32_t dirty;
+
+	struct nv20_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS];
+	struct nv20_miptree *tex_miptree[PIPE_MAX_SAMPLERS];
+	unsigned dirty_samplers;
+	unsigned fp_samplers;
+	unsigned vp_samplers;
+
+	uint32_t rt_enable;
+	struct pipe_buffer *rt[4];
+	struct pipe_buffer *zeta;
+	uint32_t lma_offset;
+
+	struct nv20_blend_state *blend;
+	struct pipe_blend_color *blend_color;
+	struct nv20_rasterizer_state *rast;
+	struct nv20_depth_stencil_alpha_state *dsa;
+	struct pipe_viewport_state *viewport;
+	struct pipe_scissor_state *scissor;
+	struct pipe_framebuffer_state *framebuffer;
+
+	//struct pipe_buffer *constbuf[PIPE_SHADER_TYPES];
+	float *constbuf[PIPE_SHADER_TYPES][32][4];
+	unsigned constbuf_nr[PIPE_SHADER_TYPES];
+
+	struct vertex_info vertex_info;
+
+	struct {
+		struct pipe_buffer *buffer;
+		uint32_t format;
+	} tex[2];
+
+	unsigned vb_enable;
+	struct {
+		struct pipe_buffer *buffer;
+		unsigned delta;
+	} vb[16];
+
+/*	struct {
+	
+		struct nouveau_resource *exec_heap;
+		struct nouveau_resource *data_heap;
+
+		struct nv20_vertex_program *active;
+
+		struct nv20_vertex_program *current;
+	} vertprog;
+*/
+	struct {
+		struct nv20_fragment_program *active;
+
+		struct nv20_fragment_program *current;
+		struct pipe_buffer *constant_buf;
+	} fragprog;
+
+	struct pipe_vertex_buffer  vtxbuf[PIPE_MAX_ATTRIBS];
+	struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS];
+};
+
+static INLINE struct nv20_context *
+nv20_context(struct pipe_context *pipe)
+{
+	return (struct nv20_context *)pipe;
+}
+
+extern void nv20_init_state_functions(struct nv20_context *nv20);
+extern void nv20_init_surface_functions(struct nv20_context *nv20);
+
+extern void nv20_screen_init_miptree_functions(struct pipe_screen *pscreen);
+
+/* nv20_clear.c */
+extern void nv20_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+		       unsigned clearValue);
+
+/* nv20_draw.c */
+extern struct draw_stage *nv20_draw_render_stage(struct nv20_context *nv20);
+
+/* nv20_fragprog.c */
+extern void nv20_fragprog_bind(struct nv20_context *,
+			       struct nv20_fragment_program *);
+extern void nv20_fragprog_destroy(struct nv20_context *,
+				  struct nv20_fragment_program *);
+
+/* nv20_fragtex.c */
+extern void nv20_fragtex_bind(struct nv20_context *);
+
+/* nv20_prim_vbuf.c */
+struct draw_stage *nv20_draw_vbuf_stage( struct nv20_context *nv20 );
+extern void nv20_vtxbuf_bind(struct nv20_context* nv20);
+
+/* nv20_state.c and friends */
+extern void nv20_emit_hw_state(struct nv20_context *nv20);
+extern void nv20_state_tex_update(struct nv20_context *nv20);
+
+/* nv20_vbo.c */
+extern boolean nv20_draw_arrays(struct pipe_context *, unsigned mode,
+				unsigned start, unsigned count);
+extern boolean nv20_draw_elements( struct pipe_context *pipe,
+                    struct pipe_buffer *indexBuffer,
+                    unsigned indexSize,
+                    unsigned prim, unsigned start, unsigned count);
+
+
+#endif
diff --git a/src/gallium/drivers/nv20/nv20_fragprog.c b/src/gallium/drivers/nv20/nv20_fragprog.c
new file mode 100644
index 0000000000..4f496369dd
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_fragprog.c
@@ -0,0 +1,21 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+
+#include "nv20_context.h"
+
+void
+nv20_fragprog_bind(struct nv20_context *nv20, struct nv20_fragment_program *fp)
+{
+}
+
+void
+nv20_fragprog_destroy(struct nv20_context *nv20,
+		      struct nv20_fragment_program *fp)
+{
+}
+
diff --git a/src/gallium/drivers/nv20/nv20_fragtex.c b/src/gallium/drivers/nv20/nv20_fragtex.c
new file mode 100644
index 0000000000..495a7be912
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_fragtex.c
@@ -0,0 +1,124 @@
+#include "nv20_context.h"
+#include "nouveau/nouveau_util.h"
+
+#define _(m,tf)                                                                \
+{                                                                              \
+  TRUE,                                                                        \
+  PIPE_FORMAT_##m,                                                             \
+  NV20TCL_TX_FORMAT_FORMAT_##tf,                                               \
+}
+
+struct nv20_texture_format {
+	boolean defined;
+	uint	pipe;
+	int     format;
+};
+
+static struct nv20_texture_format
+nv20_texture_formats[] = {
+	_(A8R8G8B8_UNORM, A8R8G8B8),
+	_(A1R5G5B5_UNORM, A1R5G5B5),
+	_(A4R4G4B4_UNORM, A4R4G4B4),
+	_(L8_UNORM      , L8      ),
+	_(A8_UNORM      , A8      ),
+	_(A8L8_UNORM    , A8L8    ),
+/*	_(RGB_DXT1      , DXT1,   ), */
+/*	_(RGBA_DXT1     , DXT1,   ), */
+/*	_(RGBA_DXT3     , DXT3,   ), */
+/*	_(RGBA_DXT5     , DXT5,   ), */
+	{},
+};
+
+static struct nv20_texture_format *
+nv20_fragtex_format(uint pipe_format)
+{
+	struct nv20_texture_format *tf = nv20_texture_formats;
+
+	while (tf->defined) {
+		if (tf->pipe == pipe_format)
+			return tf;
+		tf++;
+	}
+
+	return NULL;
+}
+
+
+static void
+nv20_fragtex_build(struct nv20_context *nv20, int unit)
+{
+#if 0
+	struct nv20_sampler_state *ps = nv20->tex_sampler[unit];
+	struct nv20_miptree *nv20mt = nv20->tex_miptree[unit];
+	struct pipe_texture *pt = &nv20mt->base;
+	struct nv20_texture_format *tf;
+	uint32_t txf, txs, txp;
+
+	tf = nv20_fragtex_format(pt->format);
+	if (!tf || !tf->defined) {
+		NOUVEAU_ERR("Unsupported texture format: 0x%x\n", pt->format);
+		return;
+	}
+
+	txf  = tf->format << 8;
+	txf |= (pt->last_level + 1) << 16;
+	txf |= log2i(pt->width[0]) << 20;
+	txf |= log2i(pt->height[0]) << 24;
+	txf |= log2i(pt->depth[0]) << 28;
+	txf |= 8;
+
+	switch (pt->target) {
+	case PIPE_TEXTURE_CUBE:
+		txf |= NV10TCL_TX_FORMAT_CUBE_MAP;
+		/* fall-through */
+	case PIPE_TEXTURE_2D:
+		txf |= (2<<4);
+		break;
+	case PIPE_TEXTURE_1D:
+		txf |= (1<<4);
+		break;
+	default:
+		NOUVEAU_ERR("Unknown target %d\n", pt->target);
+		return;
+	}
+
+	BEGIN_RING(kelvin, NV10TCL_TX_OFFSET(unit), 8);
+	OUT_RELOCl(nv20mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD);
+	OUT_RELOCd(nv20mt->buffer,txf,NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_OR | NOUVEAU_BO_RD, 1/*VRAM*/,2/*TT*/);
+	OUT_RING  (ps->wrap);
+	OUT_RING  (0x40000000); /* enable */
+	OUT_RING  (txs);
+	OUT_RING  (ps->filt | 0x2000 /* magic */);
+	OUT_RING  ((pt->width[0] << 16) | pt->height[0]);
+	OUT_RING  (ps->bcol);
+#endif
+}
+
+void
+nv20_fragtex_bind(struct nv20_context *nv20)
+{
+#if 0
+	struct nv20_fragment_program *fp = nv20->fragprog.active;
+	unsigned samplers, unit;
+
+	samplers = nv20->fp_samplers & ~fp->samplers;
+	while (samplers) {
+		unit = ffs(samplers) - 1;
+		samplers &= ~(1 << unit);
+
+		BEGIN_RING(kelvin, NV10TCL_TX_ENABLE(unit), 1);
+		OUT_RING  (0);
+	}
+
+	samplers = nv20->dirty_samplers & fp->samplers;
+	while (samplers) {
+		unit = ffs(samplers) - 1;
+		samplers &= ~(1 << unit);
+
+		nv20_fragtex_build(nv20, unit);
+	}
+
+	nv20->fp_samplers = fp->samplers;
+#endif
+}
+
diff --git a/src/gallium/drivers/nv20/nv20_miptree.c b/src/gallium/drivers/nv20/nv20_miptree.c
new file mode 100644
index 0000000000..ef7e9c5428
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_miptree.c
@@ -0,0 +1,206 @@
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "nv20_context.h"
+#include "nv20_screen.h"
+
+static void
+nv20_miptree_layout(struct nv20_miptree *nv20mt)
+{
+	struct pipe_texture *pt = &nv20mt->base;
+	boolean swizzled = FALSE;
+	uint width = pt->width[0], height = pt->height[0];
+	uint offset = 0;
+	int nr_faces, l, f;
+
+	if (pt->target == PIPE_TEXTURE_CUBE) {
+		nr_faces = 6;
+	} else {
+		nr_faces = 1;
+	}
+	
+	for (l = 0; l <= pt->last_level; l++) {
+		pt->width[l] = width;
+		pt->height[l] = height;
+		pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width);
+		pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height);
+
+		if (swizzled)
+			nv20mt->level[l].pitch = pt->nblocksx[l] * pt->block.size;
+		else
+			nv20mt->level[l].pitch = pt->nblocksx[0] * pt->block.size;
+		nv20mt->level[l].pitch = (nv20mt->level[l].pitch + 63) & ~63;
+
+		nv20mt->level[l].image_offset =
+			CALLOC(nr_faces, sizeof(unsigned));
+
+		width  = MAX2(1, width  >> 1);
+		height = MAX2(1, height >> 1);
+
+	}
+
+	for (f = 0; f < nr_faces; f++) {
+		for (l = 0; l <= pt->last_level; l++) {
+			nv20mt->level[l].image_offset[f] = offset;
+			offset += nv20mt->level[l].pitch * pt->height[l];
+		}
+	}
+
+	nv20mt->total_size = offset;
+}
+
+static struct pipe_texture *
+nv20_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
+		     const unsigned *stride, struct pipe_buffer *pb)
+{
+	struct nv20_miptree *mt;
+
+	/* Only supports 2D, non-mipmapped textures for the moment */
+	if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 ||
+	    pt->depth[0] != 1)
+		return NULL;
+
+	mt = CALLOC_STRUCT(nv20_miptree);
+	if (!mt)
+		return NULL;
+
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = pscreen;
+	mt->level[0].pitch = stride[0];
+	mt->level[0].image_offset = CALLOC(1, sizeof(unsigned));
+
+	pipe_buffer_reference(pscreen, &mt->buffer, pb);
+	return &mt->base;
+}
+
+static struct pipe_texture *
+nv20_miptree_create(struct pipe_screen *screen, const struct pipe_texture *pt)
+{
+	struct pipe_winsys *ws = screen->winsys;
+	struct nv20_miptree *mt;
+	unsigned buf_usage = PIPE_BUFFER_USAGE_PIXEL |
+	                     NOUVEAU_BUFFER_USAGE_TEXTURE;
+
+	mt = MALLOC(sizeof(struct nv20_miptree));
+	if (!mt)
+		return NULL;
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = screen;
+
+	/* Swizzled textures must be POT */
+	if (pt->width[0] & (pt->width[0] - 1) ||
+	    pt->height[0] & (pt->height[0] - 1))
+		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+	else
+	if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY |
+	                     PIPE_TEXTURE_USAGE_DISPLAY_TARGET))
+		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+	else
+	if (pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC)
+		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+	else {
+		switch (pt->format) {
+		/* TODO: Figure out which formats can be swizzled */
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_X8R8G8B8_UNORM:
+		case PIPE_FORMAT_R16_SNORM:
+			break;
+		default:
+			mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+		}
+	}
+
+	if (pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC)
+		buf_usage |= PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+
+	nv20_miptree_layout(mt);
+
+	mt->buffer = ws->buffer_create(ws, 256, buf_usage, mt->total_size);
+	if (!mt->buffer) {
+		FREE(mt);
+		return NULL;
+	}
+	
+	return &mt->base;
+}
+
+static void
+nv20_miptree_release(struct pipe_screen *screen, struct pipe_texture **pt)
+{
+	struct pipe_texture *mt = *pt;
+
+	*pt = NULL;
+	if (--mt->refcount <= 0) {
+		struct nv20_miptree *nv20mt = (struct nv20_miptree *)mt;
+		int l;
+
+		pipe_buffer_reference(screen, &nv20mt->buffer, NULL);
+		for (l = 0; l <= mt->last_level; l++) {
+			if (nv20mt->level[l].image_offset)
+				FREE(nv20mt->level[l].image_offset);
+		}
+		FREE(nv20mt);
+	}
+}
+
+static struct pipe_surface *
+nv20_miptree_surface_get(struct pipe_screen *screen, struct pipe_texture *pt,
+			 unsigned face, unsigned level, unsigned zslice,
+			 unsigned flags)
+{
+	struct nv20_miptree *nv20mt = (struct nv20_miptree *)pt;
+	struct pipe_surface *ps;
+
+	ps = CALLOC_STRUCT(pipe_surface);
+	if (!ps)
+		return NULL;
+	pipe_texture_reference(&ps->texture, pt);
+	ps->format = pt->format;
+	ps->width = pt->width[level];
+	ps->height = pt->height[level];
+	ps->block = pt->block;
+	ps->nblocksx = pt->nblocksx[level];
+	ps->nblocksy = pt->nblocksy[level];
+	ps->stride = nv20mt->level[level].pitch;
+	ps->usage = flags;
+	ps->status = PIPE_SURFACE_STATUS_DEFINED;
+	ps->refcount = 1;
+
+	if (pt->target == PIPE_TEXTURE_CUBE) {
+		ps->offset = nv20mt->level[level].image_offset[face];
+	} else
+	if (pt->target == PIPE_TEXTURE_3D) {
+		ps->offset = nv20mt->level[level].image_offset[zslice];
+	} else {
+		ps->offset = nv20mt->level[level].image_offset[0];
+	}
+
+	return ps;
+}
+
+static void
+nv20_miptree_surface_release(struct pipe_screen *pscreen,
+			     struct pipe_surface **psurface)
+{
+	struct pipe_surface *ps = *psurface;
+
+	*psurface = NULL;
+	if (--ps->refcount > 0)
+		return;
+
+	pipe_texture_reference(&ps->texture, NULL);
+	FREE(ps);
+}
+
+void nv20_screen_init_miptree_functions(struct pipe_screen *pscreen)
+{
+	pscreen->texture_create = nv20_miptree_create;
+	pscreen->texture_blanket = nv20_miptree_blanket;
+	pscreen->texture_release = nv20_miptree_release;
+	pscreen->get_tex_surface = nv20_miptree_surface_get;
+	pscreen->tex_surface_release = nv20_miptree_surface_release;
+}
+
diff --git a/src/gallium/drivers/nv20/nv20_prim_vbuf.c b/src/gallium/drivers/nv20/nv20_prim_vbuf.c
new file mode 100644
index 0000000000..4dd7052814
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_prim_vbuf.c
@@ -0,0 +1,402 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \file
+ * Build post-transformation, post-clipping vertex buffers and element
+ * lists by hooking into the end of the primitive pipeline and
+ * manipulating the vertex_id field in the vertex headers.
+ *
+ * XXX: work in progress 
+ * 
+ * \author José Fonseca <jrfonseca@tungstengraphics.com>
+ * \author Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_debug.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+#include "nv20_context.h"
+#include "nv20_state.h"
+
+#include "draw/draw_vbuf.h"
+
+/**
+ * Primitive renderer for nv20.
+ */
+struct nv20_vbuf_render {
+	struct vbuf_render base;
+
+	struct nv20_context *nv20;   
+
+	/** Vertex buffer in VRAM */
+	struct pipe_buffer *pbuffer;
+
+	/** Vertex buffer in normal memory */
+	void *mbuffer;
+
+	/** Vertex size in bytes */
+	/*unsigned vertex_size;*/
+
+	/** Hardware primitive */
+	unsigned hwprim;
+};
+
+/**
+ * Basically a cast wrapper.
+ */
+static INLINE struct nv20_vbuf_render *
+nv20_vbuf_render(struct vbuf_render *render)
+{
+	assert(render);
+	return (struct nv20_vbuf_render *)render;
+}
+
+void nv20_vtxbuf_bind( struct nv20_context* nv20 )
+{
+#if 0
+	int i;
+	for(i = 0; i < NV20TCL_VTXBUF_ADDRESS__SIZE; i++) {
+		BEGIN_RING(kelvin, NV20TCL_VTXBUF_ADDRESS(i), 1);
+		OUT_RING(0/*nv20->vtxbuf*/);
+		BEGIN_RING(kelvin, NV20TCL_VTXFMT(i) ,1);
+		OUT_RING(0/*XXX*/);
+	}
+#endif
+}
+
+static const struct vertex_info *
+nv20_vbuf_render_get_vertex_info( struct vbuf_render *render )
+{
+	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
+	struct nv20_context *nv20 = nv20_render->nv20;
+
+	nv20_emit_hw_state(nv20);
+
+	return &nv20->vertex_info;
+}
+
+static void *
+nv20__allocate_mbuffer(struct nv20_vbuf_render *nv20_render, size_t size)
+{
+	nv20_render->mbuffer = MALLOC(size);
+	return nv20_render->mbuffer;
+}
+
+static void *
+nv20__allocate_pbuffer(struct nv20_vbuf_render *nv20_render, size_t size)
+{
+	struct pipe_winsys *winsys = nv20_render->nv20->pipe.winsys;
+	nv20_render->pbuffer = winsys->buffer_create(winsys, 64,
+					PIPE_BUFFER_USAGE_VERTEX, size);
+	return winsys->buffer_map(winsys,
+			nv20_render->pbuffer,
+			PIPE_BUFFER_USAGE_CPU_WRITE);
+}
+
+static void *
+nv20_vbuf_render_allocate_vertices( struct vbuf_render *render,
+		ushort vertex_size,
+		ushort nr_vertices )
+{
+	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
+	size_t size = (size_t)vertex_size * (size_t)nr_vertices;
+	void *buf;
+
+	assert(!nv20_render->pbuffer);
+	assert(!nv20_render->mbuffer);
+
+	/*
+	 * For small amount of vertices, don't bother with pipe vertex
+	 * buffer, the data will be passed directly via the fifo.
+	 */
+	/* XXX: Pipe vertex buffers don't work. */
+	if (0 && size > 16 * 1024)
+		buf = nv20__allocate_pbuffer(nv20_render, size);
+	else
+		buf = nv20__allocate_mbuffer(nv20_render, size);
+
+	if (buf)
+		nv20_render->nv20->dirty |= NV20_NEW_VTXARRAYS;
+
+	return buf;
+}
+
+static boolean
+nv20_vbuf_render_set_primitive( struct vbuf_render *render, 
+		unsigned prim )
+{
+	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
+	unsigned hwp = nvgl_primitive(prim);
+	if (hwp == 0)
+		return FALSE;
+
+	nv20_render->hwprim = hwp;
+	return TRUE;
+}
+
+static uint32_t
+nv20__vtxhwformat(unsigned stride, unsigned fields, unsigned type)
+{
+	return (stride << NV20TCL_VTXFMT_STRIDE_SHIFT) |
+		(fields << NV20TCL_VTXFMT_SIZE_SHIFT) |
+		(type << NV20TCL_VTXFMT_TYPE_SHIFT);
+}
+
+static unsigned
+nv20__emit_format(struct nv20_context *nv20, enum attrib_emit type, int hwattr)
+{
+	uint32_t hwfmt = 0;
+	unsigned fields;
+
+	switch (type) {
+	case EMIT_OMIT:
+		hwfmt = nv20__vtxhwformat(0, 0, 2);
+		fields = 0;
+		break;
+	case EMIT_1F:
+		hwfmt = nv20__vtxhwformat(4, 1, 2);
+		fields = 1;
+		break;
+	case EMIT_2F:
+		hwfmt = nv20__vtxhwformat(8, 2, 2);
+		fields = 2;
+		break;
+	case EMIT_3F:
+		hwfmt = nv20__vtxhwformat(12, 3, 2);
+		fields = 3;
+		break;
+	case EMIT_4F:
+		hwfmt = nv20__vtxhwformat(16, 4, 2);
+		fields = 4;
+		break;
+	default:
+		NOUVEAU_ERR("unhandled attrib_emit %d\n", type);
+		return 0;
+	}
+
+	BEGIN_RING(kelvin, NV20TCL_VTXFMT(hwattr), 1);
+	OUT_RING(hwfmt);
+	return fields;
+}
+
+static unsigned
+nv20__emit_vertex_array_format(struct nv20_context *nv20)
+{
+	struct vertex_info *vinfo = &nv20->vertex_info;
+	int hwattr = NV20TCL_VTXFMT__SIZE;
+	int attr = 0;
+	unsigned nr_fields = 0;
+
+	while (hwattr-- > 0) {
+		if (vinfo->hwfmt[0] & (1 << hwattr)) {
+			nr_fields += nv20__emit_format(nv20,
+					vinfo->attrib[attr].emit, hwattr);
+			attr++;
+		} else
+			nv20__emit_format(nv20, EMIT_OMIT, hwattr);
+	}
+
+	return nr_fields;
+}
+
+static void
+nv20__draw_mbuffer(struct nv20_vbuf_render *nv20_render,
+		const ushort *indices,
+		uint nr_indices)
+{
+	struct nv20_context *nv20 = nv20_render->nv20;
+	struct vertex_info *vinfo = &nv20->vertex_info;
+	unsigned nr_fields;
+	int max_push;
+	ubyte *data = nv20_render->mbuffer;
+	int vsz = 4 * vinfo->size;
+
+	nr_fields = nv20__emit_vertex_array_format(nv20);
+
+	BEGIN_RING(kelvin, NV20TCL_VERTEX_BEGIN_END, 1);
+	OUT_RING(nv20_render->hwprim);
+
+	max_push = 1200 / nr_fields;
+	while (nr_indices) {
+		int i;
+		int push = MIN2(nr_indices, max_push);
+
+		BEGIN_RING_NI(kelvin, NV20TCL_VERTEX_DATA, push * nr_fields);
+		for (i = 0; i < push; i++) {
+			/* XXX: fixme to handle other than floats? */
+			int f = nr_fields;
+			float *attrv = (float*)&data[indices[i] * vsz];
+			while (f-- > 0)
+				OUT_RINGf(*attrv++);
+		}
+
+		nr_indices -= push;
+		indices += push;
+	}
+
+	BEGIN_RING(kelvin, NV20TCL_VERTEX_BEGIN_END, 1);
+	OUT_RING(NV20TCL_VERTEX_BEGIN_END_STOP);
+}
+
+static void
+nv20__draw_pbuffer(struct nv20_vbuf_render *nv20_render,
+		const ushort *indices,
+		uint nr_indices)
+{
+	struct nv20_context *nv20 = nv20_render->nv20;
+	int push, i;
+
+	NOUVEAU_ERR("nv20__draw_pbuffer: this path is broken.\n");
+
+	BEGIN_RING(kelvin, NV10TCL_VERTEX_ARRAY_OFFSET_POS, 1);
+	OUT_RELOCl(nv20_render->pbuffer, 0,
+			NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD);
+
+	BEGIN_RING(kelvin, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1);
+	OUT_RING(nv20_render->hwprim);
+
+	if (nr_indices & 1) {
+		BEGIN_RING(kelvin, NV10TCL_VB_ELEMENT_U32, 1);
+		OUT_RING  (indices[0]);
+		indices++; nr_indices--;
+	}
+
+	while (nr_indices) {
+		// XXX too big/small ? check the size
+		push = MIN2(nr_indices, 1200 * 2);
+
+		BEGIN_RING_NI(kelvin, NV10TCL_VB_ELEMENT_U16, push >> 1);
+		for (i = 0; i < push; i+=2)
+			OUT_RING((indices[i+1] << 16) | indices[i]);
+
+		nr_indices -= push;
+		indices  += push;
+	}
+
+	BEGIN_RING(kelvin, NV10TCL_VERTEX_BUFFER_BEGIN_END, 1);
+	OUT_RING  (0);
+}
+
+static void
+nv20_vbuf_render_draw( struct vbuf_render *render,
+		const ushort *indices,
+		uint nr_indices)
+{
+	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
+
+	nv20_emit_hw_state(nv20_render->nv20);
+
+	if (nv20_render->pbuffer)
+		nv20__draw_pbuffer(nv20_render, indices, nr_indices);
+	else if (nv20_render->mbuffer)
+		nv20__draw_mbuffer(nv20_render, indices, nr_indices);
+	else
+		assert(0);
+}
+
+
+static void
+nv20_vbuf_render_release_vertices( struct vbuf_render *render,
+		void *vertices, 
+		unsigned vertex_size,
+		unsigned vertices_used )
+{
+	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
+	struct nv20_context *nv20 = nv20_render->nv20;
+	struct pipe_winsys *winsys = nv20->pipe.winsys;
+	struct pipe_screen *pscreen = &nv20->screen->pipe;
+
+	if (nv20_render->pbuffer) {
+		winsys->buffer_unmap(winsys, nv20_render->pbuffer);
+		pipe_buffer_reference(pscreen, &nv20_render->pbuffer, NULL);
+	} else if (nv20_render->mbuffer) {
+		FREE(nv20_render->mbuffer);
+		nv20_render->mbuffer = NULL;
+	} else
+		assert(0);
+}
+
+
+static void
+nv20_vbuf_render_destroy( struct vbuf_render *render )
+{
+	struct nv20_vbuf_render *nv20_render = nv20_vbuf_render(render);
+
+	assert(!nv20_render->pbuffer);
+	assert(!nv20_render->mbuffer);
+
+	FREE(nv20_render);
+}
+
+
+/**
+ * Create a new primitive render.
+ */
+static struct vbuf_render *
+nv20_vbuf_render_create( struct nv20_context *nv20 )
+{
+	struct nv20_vbuf_render *nv20_render = CALLOC_STRUCT(nv20_vbuf_render);
+
+	nv20_render->nv20 = nv20;
+
+	nv20_render->base.max_vertex_buffer_bytes = 16*1024;
+	nv20_render->base.max_indices = 1024;
+	nv20_render->base.get_vertex_info = nv20_vbuf_render_get_vertex_info;
+	nv20_render->base.allocate_vertices =
+					nv20_vbuf_render_allocate_vertices;
+	nv20_render->base.set_primitive = nv20_vbuf_render_set_primitive;
+	nv20_render->base.draw = nv20_vbuf_render_draw;
+	nv20_render->base.release_vertices = nv20_vbuf_render_release_vertices;
+	nv20_render->base.destroy = nv20_vbuf_render_destroy;
+
+	return &nv20_render->base;
+}
+
+
+/**
+ * Create a new primitive vbuf/render stage.
+ */
+struct draw_stage *nv20_draw_vbuf_stage( struct nv20_context *nv20 )
+{
+	struct vbuf_render *render;
+	struct draw_stage *stage;
+
+	render = nv20_vbuf_render_create(nv20);
+	if(!render)
+		return NULL;
+
+	stage = draw_vbuf_stage( nv20->draw, render );
+	if(!stage) {
+		render->destroy(render);
+		return NULL;
+	}
+
+	return stage;
+}
diff --git a/src/gallium/drivers/nv20/nv20_screen.c b/src/gallium/drivers/nv20/nv20_screen.c
new file mode 100644
index 0000000000..5f2b7b4f71
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_screen.c
@@ -0,0 +1,222 @@
+#include "pipe/p_screen.h"
+#include "util/u_simple_screen.h"
+
+#include "nv20_context.h"
+#include "nv20_screen.h"
+
+static const char *
+nv20_screen_get_name(struct pipe_screen *screen)
+{
+	struct nv20_screen *nv20screen = nv20_screen(screen);
+	struct nouveau_device *dev = nv20screen->nvws->channel->device;
+	static char buffer[128];
+
+	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
+	return buffer;
+}
+
+static const char *
+nv20_screen_get_vendor(struct pipe_screen *screen)
+{
+	return "nouveau";
+}
+
+static int
+nv20_screen_get_param(struct pipe_screen *screen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+		return 2;
+	case PIPE_CAP_NPOT_TEXTURES:
+		return 0;
+	case PIPE_CAP_TWO_SIDED_STENCIL:
+		return 0;
+	case PIPE_CAP_GLSL:
+		return 0;
+	case PIPE_CAP_S3TC:
+		return 0;
+	case PIPE_CAP_ANISOTROPIC_FILTER:
+		return 1;
+	case PIPE_CAP_POINT_SPRITE:
+		return 0;
+	case PIPE_CAP_MAX_RENDER_TARGETS:
+		return 1;
+	case PIPE_CAP_OCCLUSION_QUERY:
+		return 0;
+	case PIPE_CAP_TEXTURE_SHADOW_MAP:
+		return 0;
+	case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+		return 12;
+	case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+		return 0;
+	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+		return 12;
+	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
+		return 0;
+	case NOUVEAU_CAP_HW_VTXBUF:
+	case NOUVEAU_CAP_HW_IDXBUF:
+		return 0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0;
+	}
+}
+
+static float
+nv20_screen_get_paramf(struct pipe_screen *screen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_LINE_WIDTH:
+	case PIPE_CAP_MAX_LINE_WIDTH_AA:
+		return 10.0;
+	case PIPE_CAP_MAX_POINT_WIDTH:
+	case PIPE_CAP_MAX_POINT_WIDTH_AA:
+		return 64.0;
+	case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+		return 2.0;
+	case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+		return 4.0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0.0;
+	}
+}
+
+static boolean
+nv20_screen_is_format_supported(struct pipe_screen *screen,
+				enum pipe_format format,
+				enum pipe_texture_target target,
+				unsigned tex_usage, unsigned geom_flags)
+{
+	if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM: 
+		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z16_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	} else {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_A1R5G5B5_UNORM:
+		case PIPE_FORMAT_A4R4G4B4_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM: 
+		case PIPE_FORMAT_L8_UNORM:
+		case PIPE_FORMAT_A8_UNORM:
+		case PIPE_FORMAT_I8_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	}
+
+	return FALSE;
+}
+
+static void *
+nv20_surface_map(struct pipe_screen *screen, struct pipe_surface *surface,
+		 unsigned flags )
+{
+	struct pipe_winsys *ws = screen->winsys;
+	void *map;
+	struct nv20_miptree *nv20mt = (struct nv20_miptree *)surface->texture;
+
+	map = ws->buffer_map(ws, nv20mt->buffer, flags);
+	if (!map)
+		return NULL;
+
+	return map + surface->offset;
+}
+
+static void
+nv20_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface)
+{
+	struct pipe_winsys *ws = screen->winsys;
+	struct nv20_miptree *nv20mt = (struct nv20_miptree *)surface->texture;
+
+	ws->buffer_unmap(ws, nv20mt->buffer);
+}
+
+static void
+nv20_screen_destroy(struct pipe_screen *pscreen)
+{
+	struct nv20_screen *screen = nv20_screen(pscreen);
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nvws->notifier_free(&screen->sync);
+	nvws->grobj_free(&screen->kelvin);
+
+	FREE(pscreen);
+}
+
+static struct pipe_buffer *
+nv20_surface_buffer(struct pipe_surface *surf)
+{
+	struct nv20_miptree *mt = (struct nv20_miptree *)surf->texture;
+
+	return mt->buffer;
+}
+
+struct pipe_screen *
+nv20_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+{
+	struct nv20_screen *screen = CALLOC_STRUCT(nv20_screen);
+	unsigned kelvin_class = 0;
+	unsigned chipset = nvws->channel->device->chipset;
+	int ret;
+
+	if (!screen)
+		return NULL;
+	screen->nvws = nvws;
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(nvws);
+	screen->eng2d->buf = nv20_surface_buffer;
+
+	/* 3D object */
+	if (chipset >= 0x25)
+		kelvin_class = NV25TCL;
+	else if (chipset >= 0x20)
+		kelvin_class = NV20TCL;
+
+	if (!kelvin_class || chipset >= 0x30) {
+		NOUVEAU_ERR("Unknown nv2x chipset: nv%02x\n", chipset);
+		return NULL;
+	}
+
+	ret = nvws->grobj_alloc(nvws, kelvin_class, &screen->kelvin);
+	if (ret) {
+		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
+		return FALSE;
+	}
+
+	/* Notifier for sync purposes */
+	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	if (ret) {
+		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
+		nv20_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	screen->pipe.winsys = ws;
+	screen->pipe.destroy = nv20_screen_destroy;
+
+	screen->pipe.get_name = nv20_screen_get_name;
+	screen->pipe.get_vendor = nv20_screen_get_vendor;
+	screen->pipe.get_param = nv20_screen_get_param;
+	screen->pipe.get_paramf = nv20_screen_get_paramf;
+
+	screen->pipe.is_format_supported = nv20_screen_is_format_supported;
+
+	screen->pipe.surface_map = nv20_surface_map;
+	screen->pipe.surface_unmap = nv20_surface_unmap;
+
+	nv20_screen_init_miptree_functions(&screen->pipe);
+	u_simple_screen_init(&screen->pipe);
+
+	return &screen->pipe;
+}
+
diff --git a/src/gallium/drivers/nv20/nv20_screen.h b/src/gallium/drivers/nv20/nv20_screen.h
new file mode 100644
index 0000000000..bf2f2c0d9f
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_screen.h
@@ -0,0 +1,24 @@
+#ifndef __NV20_SCREEN_H__
+#define __NV20_SCREEN_H__
+
+#include "pipe/p_screen.h"
+#include "nv04/nv04_surface_2d.h"
+
+struct nv20_screen {
+	struct pipe_screen pipe;
+
+	struct nouveau_winsys *nvws;
+
+	/* HW graphics objects */
+	struct nv04_surface_2d *eng2d;
+	struct nouveau_grobj *kelvin;
+	struct nouveau_notifier *sync;
+};
+
+static INLINE struct nv20_screen *
+nv20_screen(struct pipe_screen *screen)
+{
+	return (struct nv20_screen *)screen;
+}
+
+#endif
diff --git a/src/gallium/drivers/nv20/nv20_state.c b/src/gallium/drivers/nv20/nv20_state.c
new file mode 100644
index 0000000000..ecec4f49a0
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_state.c
@@ -0,0 +1,582 @@
+#include "draw/draw_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/tgsi_parse.h"
+
+#include "nv20_context.h"
+#include "nv20_state.h"
+
+static void *
+nv20_blend_state_create(struct pipe_context *pipe,
+			const struct pipe_blend_state *cso)
+{
+	struct nv20_blend_state *cb;
+
+	cb = MALLOC(sizeof(struct nv20_blend_state));
+
+	cb->b_enable = cso->blend_enable ? 1 : 0;
+	cb->b_srcfunc = ((nvgl_blend_func(cso->alpha_src_factor)<<16) |
+			 (nvgl_blend_func(cso->rgb_src_factor)));
+	cb->b_dstfunc = ((nvgl_blend_func(cso->alpha_dst_factor)<<16) |
+			 (nvgl_blend_func(cso->rgb_dst_factor)));
+
+	cb->c_mask = (((cso->colormask & PIPE_MASK_A) ? (0x01<<24) : 0) |
+		      ((cso->colormask & PIPE_MASK_R) ? (0x01<<16) : 0) |
+		      ((cso->colormask & PIPE_MASK_G) ? (0x01<< 8) : 0) |
+		      ((cso->colormask & PIPE_MASK_B) ? (0x01<< 0) : 0));
+
+	cb->d_enable = cso->dither ? 1 : 0;
+
+	return (void *)cb;
+}
+
+static void
+nv20_blend_state_bind(struct pipe_context *pipe, void *blend)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	nv20->blend = (struct nv20_blend_state*)blend;
+
+	nv20->dirty |= NV20_NEW_BLEND;
+}
+
+static void
+nv20_blend_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+
+static INLINE unsigned
+wrap_mode(unsigned wrap) {
+	unsigned ret;
+
+	switch (wrap) {
+	case PIPE_TEX_WRAP_REPEAT:
+		ret = NV20TCL_TX_WRAP_S_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_REPEAT:
+		ret = NV20TCL_TX_WRAP_S_MIRRORED_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+		ret = NV20TCL_TX_WRAP_S_CLAMP_TO_EDGE;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+		ret = NV20TCL_TX_WRAP_S_CLAMP_TO_BORDER;
+		break;
+	case PIPE_TEX_WRAP_CLAMP:
+		ret = NV20TCL_TX_WRAP_S_CLAMP;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+	case PIPE_TEX_WRAP_MIRROR_CLAMP:
+	default:
+		NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
+		ret = NV20TCL_TX_WRAP_S_REPEAT;
+		break;
+	}
+
+	return (ret >> NV20TCL_TX_WRAP_S_SHIFT);
+}
+
+static void *
+nv20_sampler_state_create(struct pipe_context *pipe,
+			  const struct pipe_sampler_state *cso)
+{
+	struct nv20_sampler_state *ps;
+	uint32_t filter = 0;
+
+	ps = MALLOC(sizeof(struct nv20_sampler_state));
+
+	ps->wrap = ((wrap_mode(cso->wrap_s) << NV20TCL_TX_WRAP_S_SHIFT) |
+		    (wrap_mode(cso->wrap_t) << NV20TCL_TX_WRAP_T_SHIFT));
+
+	ps->en = 0;
+	if (cso->max_anisotropy > 1.0) {
+		/* no idea, binary driver sets it, works without it.. meh.. */
+		ps->wrap |= (1 << 5);
+
+/*		if (cso->max_anisotropy >= 8.0) {
+			ps->en |= NV20TCL_TX_ENABLE_ANISO_8X;
+		} else
+		if (cso->max_anisotropy >= 4.0) {
+			ps->en |= NV20TCL_TX_ENABLE_ANISO_4X;
+		} else {
+			ps->en |= NV20TCL_TX_ENABLE_ANISO_2X;
+		}*/
+	}
+
+	switch (cso->mag_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		filter |= NV20TCL_TX_FILTER_MAGNIFY_LINEAR;
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		filter |= NV20TCL_TX_FILTER_MAGNIFY_NEAREST;
+		break;
+	}
+
+	switch (cso->min_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |=
+				NV20TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST;
+			break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV20TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV20TCL_TX_FILTER_MINIFY_LINEAR;
+			break;
+		}
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |=
+				NV20TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST;
+		break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |=
+				NV20TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV20TCL_TX_FILTER_MINIFY_NEAREST;
+			break;
+		}
+		break;
+	}
+
+	ps->filt = filter;
+
+/*	if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+		switch (cso->compare_func) {
+		case PIPE_FUNC_NEVER:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_NEVER;
+			break;
+		case PIPE_FUNC_GREATER:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_GREATER;
+			break;
+		case PIPE_FUNC_EQUAL:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_EQUAL;
+			break;
+		case PIPE_FUNC_GEQUAL:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_GEQUAL;
+			break;
+		case PIPE_FUNC_LESS:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_LESS;
+			break;
+		case PIPE_FUNC_NOTEQUAL:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_NOTEQUAL;
+			break;
+		case PIPE_FUNC_LEQUAL:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_LEQUAL;
+			break;
+		case PIPE_FUNC_ALWAYS:
+			ps->wrap |= NV10TCL_TX_WRAP_RCOMP_ALWAYS;
+			break;
+		default:
+			break;
+		}
+	}*/
+
+	ps->bcol = ((float_to_ubyte(cso->border_color[3]) << 24) |
+		    (float_to_ubyte(cso->border_color[0]) << 16) |
+		    (float_to_ubyte(cso->border_color[1]) <<  8) |
+		    (float_to_ubyte(cso->border_color[2]) <<  0));
+
+	return (void *)ps;
+}
+
+static void
+nv20_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		nv20->tex_sampler[unit] = sampler[unit];
+		nv20->dirty_samplers |= (1 << unit);
+	}
+}
+
+static void
+nv20_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void
+nv20_set_sampler_texture(struct pipe_context *pipe, unsigned nr,
+			 struct pipe_texture **miptree)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		nv20->tex_miptree[unit] = (struct nv20_miptree *)miptree[unit];
+		nv20->dirty_samplers |= (1 << unit);
+	}
+}
+
+static void *
+nv20_rasterizer_state_create(struct pipe_context *pipe,
+			     const struct pipe_rasterizer_state *cso)
+{
+	struct nv20_rasterizer_state *rs;
+	int i;
+
+	/*XXX: ignored:
+	 * 	light_twoside
+	 * 	offset_cw/ccw -nohw
+	 * 	scissor
+	 * 	point_smooth -nohw
+	 * 	multisample
+	 * 	offset_units / offset_scale
+	 */
+	rs = MALLOC(sizeof(struct nv20_rasterizer_state));
+
+	rs->templ = cso;
+	
+	rs->shade_model = cso->flatshade ? NV20TCL_SHADE_MODEL_FLAT :
+						NV20TCL_SHADE_MODEL_SMOOTH;
+
+	rs->line_width = (unsigned char)(cso->line_width * 8.0) & 0xff;
+	rs->line_smooth_en = cso->line_smooth ? 1 : 0;
+
+	/* XXX: nv20 and nv25 different! */
+	rs->point_size = *(uint32_t*)&cso->point_size;
+
+	rs->poly_smooth_en = cso->poly_smooth ? 1 : 0;
+
+	if (cso->front_winding == PIPE_WINDING_CCW) {
+		rs->front_face = NV20TCL_FRONT_FACE_CCW;
+		rs->poly_mode_front = nvgl_polygon_mode(cso->fill_ccw);
+		rs->poly_mode_back  = nvgl_polygon_mode(cso->fill_cw);
+	} else {
+		rs->front_face = NV20TCL_FRONT_FACE_CW;
+		rs->poly_mode_front = nvgl_polygon_mode(cso->fill_cw);
+		rs->poly_mode_back  = nvgl_polygon_mode(cso->fill_ccw);
+	}
+
+	switch (cso->cull_mode) {
+	case PIPE_WINDING_CCW:
+		rs->cull_face_en = 1;
+		if (cso->front_winding == PIPE_WINDING_CCW)
+			rs->cull_face    = NV20TCL_CULL_FACE_FRONT;
+		else
+			rs->cull_face    = NV20TCL_CULL_FACE_BACK;
+		break;
+	case PIPE_WINDING_CW:
+		rs->cull_face_en = 1;
+		if (cso->front_winding == PIPE_WINDING_CW)
+			rs->cull_face    = NV20TCL_CULL_FACE_FRONT;
+		else
+			rs->cull_face    = NV20TCL_CULL_FACE_BACK;
+		break;
+	case PIPE_WINDING_BOTH:
+		rs->cull_face_en = 1;
+		rs->cull_face    = NV20TCL_CULL_FACE_FRONT_AND_BACK;
+		break;
+	case PIPE_WINDING_NONE:
+	default:
+		rs->cull_face_en = 0;
+		rs->cull_face    = 0;
+		break;
+	}
+
+	if (cso->point_sprite) {
+		rs->point_sprite = (1 << 0);
+		for (i = 0; i < 8; i++) {
+			if (cso->sprite_coord_mode[i] != PIPE_SPRITE_COORD_NONE)
+				rs->point_sprite |= (1 << (8 + i));
+		}
+	} else {
+		rs->point_sprite = 0;
+	}
+
+	return (void *)rs;
+}
+
+static void
+nv20_rasterizer_state_bind(struct pipe_context *pipe, void *rast)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	nv20->rast = (struct nv20_rasterizer_state*)rast;
+
+	draw_set_rasterizer_state(nv20->draw, (nv20->rast ? nv20->rast->templ : NULL));
+
+	nv20->dirty |= NV20_NEW_RAST;
+}
+
+static void
+nv20_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void *
+nv20_depth_stencil_alpha_state_create(struct pipe_context *pipe,
+			const struct pipe_depth_stencil_alpha_state *cso)
+{
+	struct nv20_depth_stencil_alpha_state *hw;
+
+	hw = MALLOC(sizeof(struct nv20_depth_stencil_alpha_state));
+
+	hw->depth.func		= nvgl_comparison_op(cso->depth.func);
+	hw->depth.write_enable	= cso->depth.writemask ? 1 : 0;
+	hw->depth.test_enable	= cso->depth.enabled ? 1 : 0;
+
+	hw->stencil.enable = cso->stencil[0].enabled ? 1 : 0;
+	hw->stencil.wmask = cso->stencil[0].writemask;
+	hw->stencil.func = nvgl_comparison_op(cso->stencil[0].func);
+	hw->stencil.ref	= cso->stencil[0].ref_value;
+	hw->stencil.vmask = cso->stencil[0].valuemask;
+	hw->stencil.fail = nvgl_stencil_op(cso->stencil[0].fail_op);
+	hw->stencil.zfail = nvgl_stencil_op(cso->stencil[0].zfail_op);
+	hw->stencil.zpass = nvgl_stencil_op(cso->stencil[0].zpass_op);
+
+	hw->alpha.enabled = cso->alpha.enabled ? 1 : 0;
+	hw->alpha.func = nvgl_comparison_op(cso->alpha.func);
+	hw->alpha.ref  = float_to_ubyte(cso->alpha.ref_value);
+
+	return (void *)hw;
+}
+
+static void
+nv20_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *dsa)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	nv20->dsa = (struct nv20_depth_stencil_alpha_state*)dsa;
+
+	nv20->dirty |= NV20_NEW_DSA;
+}
+
+static void
+nv20_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void *
+nv20_vp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *templ)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	return draw_create_vertex_shader(nv20->draw, templ);
+}
+
+static void
+nv20_vp_state_bind(struct pipe_context *pipe, void *shader)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	draw_bind_vertex_shader(nv20->draw, (struct draw_vertex_shader *) shader);
+
+	nv20->dirty |= NV20_NEW_VERTPROG;
+}
+
+static void
+nv20_vp_state_delete(struct pipe_context *pipe, void *shader)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	draw_delete_vertex_shader(nv20->draw, (struct draw_vertex_shader *) shader);
+}
+
+static void *
+nv20_fp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	struct nv20_fragment_program *fp;
+
+	fp = CALLOC(1, sizeof(struct nv20_fragment_program));
+	fp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+	
+	tgsi_scan_shader(cso->tokens, &fp->info);
+
+	return (void *)fp;
+}
+
+static void
+nv20_fp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+	struct nv20_fragment_program *fp = hwcso;
+
+	nv20->fragprog.current = fp;
+	nv20->dirty |= NV20_NEW_FRAGPROG;
+}
+
+static void
+nv20_fp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+	struct nv20_fragment_program *fp = hwcso;
+
+	nv20_fragprog_destroy(nv20, fp);
+	FREE((void*)fp->pipe.tokens);
+	FREE(fp);
+}
+
+static void
+nv20_set_blend_color(struct pipe_context *pipe,
+		     const struct pipe_blend_color *bcol)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	nv20->blend_color = (struct pipe_blend_color*)bcol;
+
+	nv20->dirty |= NV20_NEW_BLENDCOL;
+}
+
+static void
+nv20_set_clip_state(struct pipe_context *pipe,
+		    const struct pipe_clip_state *clip)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	draw_set_clip_state(nv20->draw, clip);
+}
+
+static void
+nv20_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
+			 const struct pipe_constant_buffer *buf )
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+	struct pipe_winsys *ws = pipe->winsys;
+
+	assert(shader < PIPE_SHADER_TYPES);
+	assert(index == 0);
+
+	if (buf) {
+		void *mapped;
+		if (buf->buffer && buf->buffer->size &&
+                    (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)))
+		{
+			memcpy(nv20->constbuf[shader], mapped, buf->buffer->size);
+			nv20->constbuf_nr[shader] =
+				buf->buffer->size / (4 * sizeof(float));
+			ws->buffer_unmap(ws, buf->buffer);
+		}
+	}
+}
+
+static void
+nv20_set_framebuffer_state(struct pipe_context *pipe,
+			   const struct pipe_framebuffer_state *fb)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	nv20->framebuffer = (struct pipe_framebuffer_state*)fb;
+
+	nv20->dirty |= NV20_NEW_FRAMEBUFFER;
+}
+
+static void
+nv20_set_polygon_stipple(struct pipe_context *pipe,
+			 const struct pipe_poly_stipple *stipple)
+{
+	NOUVEAU_ERR("line stipple hahaha\n");
+}
+
+static void
+nv20_set_scissor_state(struct pipe_context *pipe,
+		       const struct pipe_scissor_state *s)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	nv20->scissor = (struct pipe_scissor_state*)s;
+
+	nv20->dirty |= NV20_NEW_SCISSOR;
+}
+
+static void
+nv20_set_viewport_state(struct pipe_context *pipe,
+			const struct pipe_viewport_state *vpt)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	nv20->viewport = (struct pipe_viewport_state*)vpt;
+
+	draw_set_viewport_state(nv20->draw, nv20->viewport);
+
+	nv20->dirty |= NV20_NEW_VIEWPORT;
+}
+
+static void
+nv20_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
+			const struct pipe_vertex_buffer *vb)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	memcpy(nv20->vtxbuf, vb, sizeof(*vb) * count);
+	nv20->dirty |= NV20_NEW_VTXARRAYS;
+
+	draw_set_vertex_buffers(nv20->draw, count, vb);
+}
+
+static void
+nv20_set_vertex_elements(struct pipe_context *pipe, unsigned count,
+			 const struct pipe_vertex_element *ve)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+
+	memcpy(nv20->vtxelt, ve, sizeof(*ve) * count);
+	nv20->dirty |= NV20_NEW_VTXARRAYS;
+
+	draw_set_vertex_elements(nv20->draw, count, ve);
+}
+
+void
+nv20_init_state_functions(struct nv20_context *nv20)
+{
+	nv20->pipe.create_blend_state = nv20_blend_state_create;
+	nv20->pipe.bind_blend_state = nv20_blend_state_bind;
+	nv20->pipe.delete_blend_state = nv20_blend_state_delete;
+
+	nv20->pipe.create_sampler_state = nv20_sampler_state_create;
+	nv20->pipe.bind_sampler_states = nv20_sampler_state_bind;
+	nv20->pipe.delete_sampler_state = nv20_sampler_state_delete;
+	nv20->pipe.set_sampler_textures = nv20_set_sampler_texture;
+
+	nv20->pipe.create_rasterizer_state = nv20_rasterizer_state_create;
+	nv20->pipe.bind_rasterizer_state = nv20_rasterizer_state_bind;
+	nv20->pipe.delete_rasterizer_state = nv20_rasterizer_state_delete;
+
+	nv20->pipe.create_depth_stencil_alpha_state =
+		nv20_depth_stencil_alpha_state_create;
+	nv20->pipe.bind_depth_stencil_alpha_state =
+		nv20_depth_stencil_alpha_state_bind;
+	nv20->pipe.delete_depth_stencil_alpha_state =
+		nv20_depth_stencil_alpha_state_delete;
+
+	nv20->pipe.create_vs_state = nv20_vp_state_create;
+	nv20->pipe.bind_vs_state = nv20_vp_state_bind;
+	nv20->pipe.delete_vs_state = nv20_vp_state_delete;
+
+	nv20->pipe.create_fs_state = nv20_fp_state_create;
+	nv20->pipe.bind_fs_state = nv20_fp_state_bind;
+	nv20->pipe.delete_fs_state = nv20_fp_state_delete;
+
+	nv20->pipe.set_blend_color = nv20_set_blend_color;
+	nv20->pipe.set_clip_state = nv20_set_clip_state;
+	nv20->pipe.set_constant_buffer = nv20_set_constant_buffer;
+	nv20->pipe.set_framebuffer_state = nv20_set_framebuffer_state;
+	nv20->pipe.set_polygon_stipple = nv20_set_polygon_stipple;
+	nv20->pipe.set_scissor_state = nv20_set_scissor_state;
+	nv20->pipe.set_viewport_state = nv20_set_viewport_state;
+
+	nv20->pipe.set_vertex_buffers = nv20_set_vertex_buffers;
+	nv20->pipe.set_vertex_elements = nv20_set_vertex_elements;
+}
+
diff --git a/src/gallium/drivers/nv20/nv20_state.h b/src/gallium/drivers/nv20/nv20_state.h
new file mode 100644
index 0000000000..34f402fdcb
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_state.h
@@ -0,0 +1,139 @@
+#ifndef __NV20_STATE_H__
+#define __NV20_STATE_H__
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_scan.h"
+
+struct nv20_blend_state {
+	uint32_t b_enable;
+	uint32_t b_srcfunc;
+	uint32_t b_dstfunc;
+
+	uint32_t c_mask;
+
+	uint32_t d_enable;
+};
+
+struct nv20_sampler_state {
+	uint32_t wrap;
+	uint32_t en;
+	uint32_t filt;
+	uint32_t bcol;
+};
+
+struct nv20_rasterizer_state {
+	uint32_t shade_model;
+
+	uint32_t line_width;
+	uint32_t line_smooth_en;
+
+	uint32_t point_size;
+
+	uint32_t poly_smooth_en;
+	
+	uint32_t poly_mode_front;
+	uint32_t poly_mode_back;
+
+	uint32_t front_face;
+	uint32_t cull_face;
+	uint32_t cull_face_en;
+
+	uint32_t point_sprite;
+
+	const struct pipe_rasterizer_state *templ;
+};
+
+struct nv20_vertex_program_exec {
+	uint32_t data[4];
+	boolean has_branch_offset;
+	int const_index;
+};
+
+struct nv20_vertex_program_data {
+	int index; /* immediates == -1 */
+	float value[4];
+};
+
+struct nv20_vertex_program {
+	const struct pipe_shader_state *pipe;
+
+	boolean translated;
+	struct nv20_vertex_program_exec *insns;
+	unsigned nr_insns;
+	struct nv20_vertex_program_data *consts;
+	unsigned nr_consts;
+
+	struct nouveau_resource *exec;
+	unsigned exec_start;
+	struct nouveau_resource *data;
+	unsigned data_start;
+	unsigned data_start_min;
+
+	uint32_t ir;
+	uint32_t or;
+};
+
+struct nv20_fragment_program_data {
+	unsigned offset;
+	unsigned index;
+};
+
+struct nv20_fragment_program {
+	struct pipe_shader_state pipe;
+	struct tgsi_shader_info info;
+
+	boolean translated;
+	boolean on_hw;
+	unsigned samplers;
+
+	uint32_t *insn;
+	int       insn_len;
+
+	struct nv20_fragment_program_data *consts;
+	unsigned nr_consts;
+
+	struct pipe_buffer *buffer;
+
+	uint32_t fp_control;
+	uint32_t fp_reg_control;
+};
+
+
+struct nv20_depth_stencil_alpha_state {
+	struct {
+		uint32_t func;
+		uint32_t write_enable;
+		uint32_t test_enable;
+	} depth;
+
+	struct {
+		uint32_t enable;
+		uint32_t wmask;
+		uint32_t func;
+		uint32_t ref;
+		uint32_t vmask;
+		uint32_t fail;
+		uint32_t zfail;
+		uint32_t zpass;
+	} stencil;
+
+	struct {
+		uint32_t enabled;
+		uint32_t func;
+		uint32_t ref;
+	} alpha;
+};
+
+struct nv20_miptree {
+	struct pipe_texture base;
+
+	struct pipe_buffer *buffer;
+	uint total_size;
+
+	struct {
+		uint pitch;
+		uint *image_offset;
+	} level[PIPE_MAX_TEXTURE_LEVELS];
+};
+
+#endif
diff --git a/src/gallium/drivers/nv20/nv20_state_emit.c b/src/gallium/drivers/nv20/nv20_state_emit.c
new file mode 100644
index 0000000000..0f4df9ca31
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_state_emit.c
@@ -0,0 +1,396 @@
+#include "nv20_context.h"
+#include "nv20_state.h"
+#include "draw/draw_context.h"
+
+static void nv20_state_emit_blend(struct nv20_context* nv20)
+{
+	struct nv20_blend_state *b = nv20->blend;
+
+	BEGIN_RING(kelvin, NV20TCL_DITHER_ENABLE, 1);
+	OUT_RING  (b->d_enable);
+
+	BEGIN_RING(kelvin, NV20TCL_BLEND_FUNC_ENABLE, 1);
+	OUT_RING  (b->b_enable);
+
+	BEGIN_RING(kelvin, NV20TCL_BLEND_FUNC_SRC, 2);
+	OUT_RING  (b->b_srcfunc);
+	OUT_RING  (b->b_dstfunc);
+
+	BEGIN_RING(kelvin, NV20TCL_COLOR_MASK, 1);
+	OUT_RING  (b->c_mask);
+}
+
+static void nv20_state_emit_blend_color(struct nv20_context* nv20)
+{
+	struct pipe_blend_color *c = nv20->blend_color;
+
+	BEGIN_RING(kelvin, NV20TCL_BLEND_COLOR, 1);
+	OUT_RING  ((float_to_ubyte(c->color[3]) << 24)|
+		   (float_to_ubyte(c->color[0]) << 16)|
+		   (float_to_ubyte(c->color[1]) << 8) |
+		   (float_to_ubyte(c->color[2]) << 0));
+}
+
+static void nv20_state_emit_rast(struct nv20_context* nv20)
+{
+	struct nv20_rasterizer_state *r = nv20->rast;
+
+	BEGIN_RING(kelvin, NV20TCL_SHADE_MODEL, 2);
+	OUT_RING  (r->shade_model);
+	OUT_RING  (r->line_width);
+
+
+	BEGIN_RING(kelvin, NV20TCL_POINT_SIZE, 1);
+	OUT_RING  (r->point_size);
+
+	BEGIN_RING(kelvin, NV20TCL_POLYGON_MODE_FRONT, 2);
+	OUT_RING  (r->poly_mode_front);
+	OUT_RING  (r->poly_mode_back);
+
+
+	BEGIN_RING(kelvin, NV20TCL_CULL_FACE, 2);
+	OUT_RING  (r->cull_face);
+	OUT_RING  (r->front_face);
+
+	BEGIN_RING(kelvin, NV20TCL_LINE_SMOOTH_ENABLE, 2);
+	OUT_RING  (r->line_smooth_en);
+	OUT_RING  (r->poly_smooth_en);
+
+	BEGIN_RING(kelvin, NV20TCL_CULL_FACE_ENABLE, 1);
+	OUT_RING  (r->cull_face_en);
+}
+
+static void nv20_state_emit_dsa(struct nv20_context* nv20)
+{
+	struct nv20_depth_stencil_alpha_state *d = nv20->dsa;
+
+	BEGIN_RING(kelvin, NV20TCL_DEPTH_FUNC, 1);
+	OUT_RING (d->depth.func);
+
+	BEGIN_RING(kelvin, NV20TCL_DEPTH_WRITE_ENABLE, 1);
+	OUT_RING (d->depth.write_enable);
+
+	BEGIN_RING(kelvin, NV20TCL_DEPTH_TEST_ENABLE, 1);
+	OUT_RING (d->depth.test_enable);
+
+	BEGIN_RING(kelvin, NV20TCL_DEPTH_UNK17D8, 1);
+	OUT_RING (1);
+
+#if 0
+	BEGIN_RING(kelvin, NV20TCL_STENCIL_ENABLE, 1);
+	OUT_RING (d->stencil.enable);
+	BEGIN_RING(kelvin, NV20TCL_STENCIL_MASK, 7);
+	OUT_RINGp ((uint32_t *)&(d->stencil.wmask), 7);
+#endif
+
+	BEGIN_RING(kelvin, NV20TCL_ALPHA_FUNC_ENABLE, 1);
+	OUT_RING (d->alpha.enabled);
+
+	BEGIN_RING(kelvin, NV20TCL_ALPHA_FUNC_FUNC, 1);
+	OUT_RING (d->alpha.func);
+
+	BEGIN_RING(kelvin, NV20TCL_ALPHA_FUNC_REF, 1);
+	OUT_RING (d->alpha.ref);
+}
+
+static void nv20_state_emit_viewport(struct nv20_context* nv20)
+{
+}
+
+static void nv20_state_emit_scissor(struct nv20_context* nv20)
+{
+	/* NV20TCL_SCISSOR_* is probably a software method */
+/*	struct pipe_scissor_state *s = nv20->scissor;
+	BEGIN_RING(kelvin, NV20TCL_SCISSOR_HORIZ, 2);
+	OUT_RING  (((s->maxx - s->minx) << 16) | s->minx);
+	OUT_RING  (((s->maxy - s->miny) << 16) | s->miny);*/
+}
+
+static void nv20_state_emit_framebuffer(struct nv20_context* nv20)
+{
+	struct pipe_framebuffer_state* fb = nv20->framebuffer;
+	struct pipe_surface *rt, *zeta = NULL;
+	uint32_t rt_format, w, h;
+	int colour_format = 0, zeta_format = 0;
+	struct nv20_miptree *nv20mt = 0;
+
+	w = fb->cbufs[0]->width;
+	h = fb->cbufs[0]->height;
+	colour_format = fb->cbufs[0]->format;
+	rt = fb->cbufs[0];
+
+	if (fb->zsbuf) {
+		if (colour_format) {
+			assert(w == fb->zsbuf->width);
+			assert(h == fb->zsbuf->height);
+		} else {
+			w = fb->zsbuf->width;
+			h = fb->zsbuf->height;
+		}
+
+		zeta_format = fb->zsbuf->format;
+		zeta = fb->zsbuf;
+	}
+
+	rt_format = NV20TCL_RT_FORMAT_TYPE_LINEAR | 0x20;
+
+	switch (colour_format) {
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case 0:
+		rt_format |= NV20TCL_RT_FORMAT_COLOR_A8R8G8B8;
+		break;
+	case PIPE_FORMAT_R5G6B5_UNORM:
+		rt_format |= NV20TCL_RT_FORMAT_COLOR_R5G6B5;
+		break;
+	default:
+		assert(0);
+	}
+
+	if (zeta) {
+		BEGIN_RING(kelvin, NV20TCL_RT_PITCH, 1);
+		OUT_RING  (rt->stride | (zeta->stride << 16));
+	} else {
+		BEGIN_RING(kelvin, NV20TCL_RT_PITCH, 1);
+		OUT_RING  (rt->stride | (rt->stride << 16));
+	}
+
+	nv20mt = (struct nv20_miptree *)rt->texture;
+	nv20->rt[0] = nv20mt->buffer;
+
+	if (zeta_format)
+	{
+		nv20mt = (struct nv20_miptree *)zeta->texture;
+		nv20->zeta = nv20mt->buffer;
+	}
+
+	BEGIN_RING(kelvin, NV20TCL_RT_HORIZ, 3);
+	OUT_RING  ((w << 16) | 0);
+	OUT_RING  ((h << 16) | 0); /*NV20TCL_RT_VERT */
+	OUT_RING  (rt_format); /* NV20TCL_RT_FORMAT */
+	BEGIN_RING(kelvin, NV20TCL_VIEWPORT_CLIP_HORIZ(0), 2);
+	OUT_RING  (((w - 1) << 16) | 0);
+	OUT_RING  (((h - 1) << 16) | 0);
+}
+
+static void nv20_vertex_layout(struct nv20_context *nv20)
+{
+	struct nv20_fragment_program *fp = nv20->fragprog.current;
+	struct draw_context *dc = nv20->draw;
+	int src;
+	int i;
+	struct vertex_info *vinfo = &nv20->vertex_info;
+	const enum interp_mode colorInterp = INTERP_LINEAR;
+	boolean colors[2] = { FALSE };
+	boolean generics[12] = { FALSE };
+	boolean fog = FALSE;
+
+	memset(vinfo, 0, sizeof(*vinfo));
+
+	/*
+	 * Assumed NV20 hardware vertex attribute order:
+	 * 0 position, 1 ?, 2 ?, 3 col0,
+	 * 4 col1?, 5 ?, 6 ?, 7 ?,
+	 * 8 ?, 9 tex0, 10 tex1, 11 tex2,
+	 * 12 tex3, 13 ?, 14 ?, 15 ?
+	 * unaccounted: wgh, nor, fog
+	 * There are total 16 attrs.
+	 * vinfo->hwfmt[0] has a used-bit corresponding to each of these.
+	 * relation to TGSI_SEMANTIC_*:
+	 * - POSITION: position (always used)
+	 * - COLOR: col1, col0
+	 * - GENERIC: tex3, tex2, tex1, tex0, normal, weight
+	 * - FOG: fog
+	 */
+
+	for (i = 0; i < fp->info.num_inputs; i++) {
+		int isn = fp->info.input_semantic_name[i];
+		int isi = fp->info.input_semantic_index[i];
+		switch (isn) {
+		case TGSI_SEMANTIC_POSITION:
+			break;
+		case TGSI_SEMANTIC_COLOR:
+			assert(isi < 2);
+			colors[isi] = TRUE;
+			break;
+		case TGSI_SEMANTIC_GENERIC:
+			assert(isi < 12);
+			generics[isi] = TRUE;
+			break;
+		case TGSI_SEMANTIC_FOG:
+			fog = TRUE;
+			break;
+		default:
+			assert(0 && "unknown input_semantic_name");
+		}
+	}
+
+	/* always do position */ {
+		src = draw_find_vs_output(dc, TGSI_SEMANTIC_POSITION, 0);
+		draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_LINEAR, src);
+		vinfo->hwfmt[0] |= (1 << 0);
+	}
+
+	/* two unnamed generics */
+	for (i = 4; i < 6; i++) {
+		if (!generics[i])
+			continue;
+		src = draw_find_vs_output(dc, TGSI_SEMANTIC_GENERIC, i);
+		draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+		vinfo->hwfmt[0] |= (1 << (i - 3));
+	}
+
+	if (colors[0]) {
+		src = draw_find_vs_output(dc, TGSI_SEMANTIC_COLOR, 0);
+		draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
+		vinfo->hwfmt[0] |= (1 << 3);
+	}
+
+	if (colors[1]) {
+		src = draw_find_vs_output(dc, TGSI_SEMANTIC_COLOR, 1);
+		draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
+		vinfo->hwfmt[0] |= (1 << 4);
+	}
+
+	/* four unnamed generics */
+	for (i = 6; i < 10; i++) {
+		if (!generics[i])
+			continue;
+		src = draw_find_vs_output(dc, TGSI_SEMANTIC_GENERIC, i);
+		draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+		vinfo->hwfmt[0] |= (1 << (i - 1));
+	}
+
+	/* tex0, tex1, tex2, tex3 */
+	for (i = 0; i < 4; i++) {
+		if (!generics[i])
+			continue;
+		src = draw_find_vs_output(dc, TGSI_SEMANTIC_GENERIC, i);
+		draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+		vinfo->hwfmt[0] |= (1 << (i + 9));
+	}
+
+	/* two unnamed generics */
+	for (i = 10; i < 12; i++) {
+		if (!generics[i])
+			continue;
+		src = draw_find_vs_output(dc, TGSI_SEMANTIC_GENERIC, i);
+		draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+		vinfo->hwfmt[0] |= (1 << (i + 3));
+	}
+
+	if (fog) {
+		src = draw_find_vs_output(dc, TGSI_SEMANTIC_FOG, 0);
+		draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_PERSPECTIVE, src);
+		vinfo->hwfmt[0] |= (1 << 15);
+	}
+
+	draw_compute_vertex_size(vinfo);
+}
+
+void
+nv20_emit_hw_state(struct nv20_context *nv20)
+{
+	int i;
+
+	if (nv20->dirty & NV20_NEW_VERTPROG) {
+		//nv20_vertprog_bind(nv20, nv20->vertprog.current);
+		nv20->dirty &= ~NV20_NEW_VERTPROG;
+	}
+
+	if (nv20->dirty & NV20_NEW_FRAGPROG) {
+		nv20_fragprog_bind(nv20, nv20->fragprog.current);
+		/*XXX: clear NV20_NEW_FRAGPROG if no new program uploaded */
+		nv20->dirty_samplers |= (1<<10);
+		nv20->dirty_samplers = 0;
+	}
+
+	if (nv20->dirty_samplers || (nv20->dirty & NV20_NEW_FRAGPROG)) {
+		nv20_fragtex_bind(nv20);
+		nv20->dirty &= ~NV20_NEW_FRAGPROG;
+	}
+
+	if (nv20->dirty & NV20_NEW_VTXARRAYS) {
+		nv20->dirty &= ~NV20_NEW_VTXARRAYS;
+		nv20_vertex_layout(nv20);
+		nv20_vtxbuf_bind(nv20);
+	}
+
+	if (nv20->dirty & NV20_NEW_BLEND) {
+		nv20->dirty &= ~NV20_NEW_BLEND;
+		nv20_state_emit_blend(nv20);
+	}
+
+	if (nv20->dirty & NV20_NEW_BLENDCOL) {
+		nv20->dirty &= ~NV20_NEW_BLENDCOL;
+		nv20_state_emit_blend_color(nv20);
+	}
+
+	if (nv20->dirty & NV20_NEW_RAST) {
+		nv20->dirty &= ~NV20_NEW_RAST;
+		nv20_state_emit_rast(nv20);
+	}
+
+	if (nv20->dirty & NV20_NEW_DSA) {
+		nv20->dirty &= ~NV20_NEW_DSA;
+		nv20_state_emit_dsa(nv20);
+	}
+
+ 	if (nv20->dirty & NV20_NEW_VIEWPORT) {
+		nv20->dirty &= ~NV20_NEW_VIEWPORT;
+		nv20_state_emit_viewport(nv20);
+	}
+
+ 	if (nv20->dirty & NV20_NEW_SCISSOR) {
+		nv20->dirty &= ~NV20_NEW_SCISSOR;
+		nv20_state_emit_scissor(nv20);
+	}
+
+ 	if (nv20->dirty & NV20_NEW_FRAMEBUFFER) {
+		nv20->dirty &= ~NV20_NEW_FRAMEBUFFER;
+		nv20_state_emit_framebuffer(nv20);
+	}
+
+	/* Emit relocs for every referenced buffer.
+	 * This is to ensure the bufmgr has an accurate idea of how
+	 * the buffer is used.  This isn't very efficient, but we don't
+	 * seem to take a significant performance hit.  Will be improved
+	 * at some point.  Vertex arrays are emitted by nv20_vbo.c
+	 */
+
+	/* Render target */
+	BEGIN_RING(kelvin, NV20TCL_DMA_COLOR, 1);
+	OUT_RELOCo(nv20->rt[0], NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	BEGIN_RING(kelvin, NV20TCL_COLOR_OFFSET, 1);
+	OUT_RELOCl(nv20->rt[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+
+	if (nv20->zeta) {
+		BEGIN_RING(kelvin, NV20TCL_DMA_ZETA, 1);
+		OUT_RELOCo(nv20->zeta, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+		BEGIN_RING(kelvin, NV20TCL_ZETA_OFFSET, 1);
+		OUT_RELOCl(nv20->zeta, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+		/* XXX for when we allocate LMA on nv17 */
+/*		BEGIN_RING(kelvin, NV10TCL_LMA_DEPTH_BUFFER_OFFSET, 1);
+		OUT_RELOCl(nv20->zeta + lma_offset);*/
+	}
+
+	/* Vertex buffer */
+	BEGIN_RING(kelvin, NV20TCL_DMA_VTXBUF0, 1);
+	OUT_RELOCo(nv20->rt[0], NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	BEGIN_RING(kelvin, NV20TCL_COLOR_OFFSET, 1);
+	OUT_RELOCl(nv20->rt[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+
+	/* Texture images */
+	for (i = 0; i < 2; i++) {
+		if (!(nv20->fp_samplers & (1 << i)))
+			continue;
+		BEGIN_RING(kelvin, NV20TCL_TX_OFFSET(i), 1);
+		OUT_RELOCl(nv20->tex[i].buffer, 0, NOUVEAU_BO_VRAM |
+			   NOUVEAU_BO_GART | NOUVEAU_BO_RD);
+		BEGIN_RING(kelvin, NV20TCL_TX_FORMAT(i), 1);
+		OUT_RELOCd(nv20->tex[i].buffer, nv20->tex[i].format,
+			   NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD |
+			   NOUVEAU_BO_OR, NV20TCL_TX_FORMAT_DMA0,
+			   NV20TCL_TX_FORMAT_DMA1);
+	}
+}
+
diff --git a/src/gallium/drivers/nv20/nv20_surface.c b/src/gallium/drivers/nv20/nv20_surface.c
new file mode 100644
index 0000000000..6cd607583c
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_surface.c
@@ -0,0 +1,72 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "nv20_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+
+static void
+nv20_surface_copy(struct pipe_context *pipe, boolean do_flip,
+		  struct pipe_surface *dest, unsigned destx, unsigned desty,
+		  struct pipe_surface *src, unsigned srcx, unsigned srcy,
+		  unsigned width, unsigned height)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+	struct nv04_surface_2d *eng2d = nv20->screen->eng2d;
+
+	if (do_flip) {
+		desty += height;
+		while (height--) {
+			eng2d->copy(eng2d, dest, destx, desty--, src,
+				    srcx, srcy++, width, 1);
+		}
+		return;
+	}
+
+	eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height);
+}
+
+static void
+nv20_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
+		  unsigned destx, unsigned desty, unsigned width,
+		  unsigned height, unsigned value)
+{
+	struct nv20_context *nv20 = nv20_context(pipe);
+	struct nv04_surface_2d *eng2d = nv20->screen->eng2d;
+
+	eng2d->fill(eng2d, dest, destx, desty, width, height, value);
+}
+
+void
+nv20_init_surface_functions(struct nv20_context *nv20)
+{
+	nv20->pipe.surface_copy = nv20_surface_copy;
+	nv20->pipe.surface_fill = nv20_surface_fill;
+}
diff --git a/src/gallium/drivers/nv20/nv20_vbo.c b/src/gallium/drivers/nv20/nv20_vbo.c
new file mode 100644
index 0000000000..24d8f4bef0
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_vbo.c
@@ -0,0 +1,78 @@
+#include "draw/draw_context.h"
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+
+#include "nv20_context.h"
+#include "nv20_state.h"
+
+#include "nouveau/nouveau_channel.h"
+#include "nouveau/nouveau_pushbuf.h"
+
+boolean nv20_draw_elements( struct pipe_context *pipe,
+                    struct pipe_buffer *indexBuffer,
+                    unsigned indexSize,
+                    unsigned prim, unsigned start, unsigned count)
+{
+	struct nv20_context *nv20 = nv20_context( pipe );
+	struct draw_context *draw = nv20->draw;
+	unsigned i;
+
+	nv20_emit_hw_state(nv20);
+
+	/*
+	 * Map vertex buffers
+	 */
+	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
+		if (nv20->vtxbuf[i].buffer) {
+			void *buf
+				= pipe->winsys->buffer_map(pipe->winsys,
+						nv20->vtxbuf[i].buffer,
+						PIPE_BUFFER_USAGE_CPU_READ);
+			draw_set_mapped_vertex_buffer(draw, i, buf);
+		}
+	}
+	/* Map index buffer, if present */
+	if (indexBuffer) {
+		void *mapped_indexes
+			= pipe->winsys->buffer_map(pipe->winsys, indexBuffer,
+					PIPE_BUFFER_USAGE_CPU_READ);
+		draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes);
+	}
+	else {
+		/* no index/element buffer */
+		draw_set_mapped_element_buffer(draw, 0, NULL);
+	}
+
+	draw_set_mapped_constant_buffer(draw,
+					nv20->constbuf[PIPE_SHADER_VERTEX],
+					nv20->constbuf_nr[PIPE_SHADER_VERTEX]);
+
+	/* draw! */
+	draw_arrays(nv20->draw, prim, start, count);
+
+	/*
+	 * unmap vertex/index buffers
+	 */
+	for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
+		if (nv20->vtxbuf[i].buffer) {
+			pipe->winsys->buffer_unmap(pipe->winsys, nv20->vtxbuf[i].buffer);
+			draw_set_mapped_vertex_buffer(draw, i, NULL);
+		}
+	}
+	if (indexBuffer) {
+		pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer);
+		draw_set_mapped_element_buffer(draw, 0, NULL);
+	}
+
+	draw_flush(nv20->draw);
+	return TRUE;
+}
+
+boolean nv20_draw_arrays( struct pipe_context *pipe,
+				 unsigned prim, unsigned start, unsigned count)
+{
+	return nv20_draw_elements(pipe, NULL, 0, prim, start, count);
+}
+
+
+
diff --git a/src/gallium/drivers/nv20/nv20_vertprog.c b/src/gallium/drivers/nv20/nv20_vertprog.c
new file mode 100644
index 0000000000..5db0e807ff
--- /dev/null
+++ b/src/gallium/drivers/nv20/nv20_vertprog.c
@@ -0,0 +1,838 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "nv20_context.h"
+#include "nv20_state.h"
+
+/* TODO (at least...):
+ *  1. Indexed consts  + ARL
+ *  2. Arb. swz/negation
+ *  3. NV_vp11, NV_vp2, NV_vp3 features
+ *       - extra arith opcodes
+ *       - branching
+ *       - texture sampling
+ *       - indexed attribs
+ *       - indexed results
+ *  4. bugs
+ */
+
+#define SWZ_X 0
+#define SWZ_Y 1
+#define SWZ_Z 2
+#define SWZ_W 3
+#define MASK_X 8
+#define MASK_Y 4
+#define MASK_Z 2
+#define MASK_W 1
+#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W)
+#define DEF_SCALE 0
+#define DEF_CTEST 0
+#include "nv20_shader.h"
+
+#define swz(s,x,y,z,w) nv20_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w)
+#define neg(s) nv20_sr_neg((s))
+#define abs(s) nv20_sr_abs((s))
+
+struct nv20_vpc {
+	struct nv20_vertex_program *vp;
+
+	struct nv20_vertex_program_exec *vpi;
+
+	unsigned output_map[PIPE_MAX_SHADER_OUTPUTS];
+
+	int high_temp;
+	int temp_temp_count;
+
+	struct nv20_sreg *imm;
+	unsigned nr_imm;
+};
+
+static struct nv20_sreg
+temp(struct nv20_vpc *vpc)
+{
+	int idx;
+
+	idx  = vpc->temp_temp_count++;
+	idx += vpc->high_temp + 1;
+	return nv20_sr(NV30SR_TEMP, idx);
+}
+
+static struct nv20_sreg
+constant(struct nv20_vpc *vpc, int pipe, float x, float y, float z, float w)
+{
+	struct nv20_vertex_program *vp = vpc->vp;
+	struct nv20_vertex_program_data *vpd;
+	int idx;
+
+	if (pipe >= 0) {
+		for (idx = 0; idx < vp->nr_consts; idx++) {
+			if (vp->consts[idx].index == pipe)
+				return nv20_sr(NV30SR_CONST, idx);
+		}
+	}
+
+	idx = vp->nr_consts++;
+	vp->consts = realloc(vp->consts, sizeof(*vpd) * vp->nr_consts);
+	vpd = &vp->consts[idx];
+
+	vpd->index = pipe;
+	vpd->value[0] = x;
+	vpd->value[1] = y;
+	vpd->value[2] = z;
+	vpd->value[3] = w;
+	return nv20_sr(NV30SR_CONST, idx);
+}
+
+#define arith(cc,s,o,d,m,s0,s1,s2) \
+	nv20_vp_arith((cc), (s), NV30_VP_INST_##o, (d), (m), (s0), (s1), (s2))
+
+static void
+emit_src(struct nv20_vpc *vpc, uint32_t *hw, int pos, struct nv20_sreg src)
+{
+	struct nv20_vertex_program *vp = vpc->vp;
+	uint32_t sr = 0;
+
+	switch (src.type) {
+	case NV30SR_TEMP:
+		sr |= (NV30_VP_SRC_REG_TYPE_TEMP << NV30_VP_SRC_REG_TYPE_SHIFT);
+		sr |= (src.index << NV30_VP_SRC_TEMP_SRC_SHIFT);
+		break;
+	case NV30SR_INPUT:
+		sr |= (NV30_VP_SRC_REG_TYPE_INPUT <<
+		       NV30_VP_SRC_REG_TYPE_SHIFT);
+		vp->ir |= (1 << src.index);
+		hw[1] |= (src.index << NV30_VP_INST_INPUT_SRC_SHIFT);
+		break;
+	case NV30SR_CONST:
+		sr |= (NV30_VP_SRC_REG_TYPE_CONST <<
+		       NV30_VP_SRC_REG_TYPE_SHIFT);
+		assert(vpc->vpi->const_index == -1 ||
+		       vpc->vpi->const_index == src.index);
+		vpc->vpi->const_index = src.index;
+		break;
+	case NV30SR_NONE:
+		sr |= (NV30_VP_SRC_REG_TYPE_INPUT <<
+		       NV30_VP_SRC_REG_TYPE_SHIFT);
+		break;
+	default:
+		assert(0);
+	}
+
+	if (src.negate)
+		sr |= NV30_VP_SRC_NEGATE;
+
+	if (src.abs)
+		hw[0] |= (1 << (21 + pos));
+
+	sr |= ((src.swz[0] << NV30_VP_SRC_SWZ_X_SHIFT) |
+	       (src.swz[1] << NV30_VP_SRC_SWZ_Y_SHIFT) |
+	       (src.swz[2] << NV30_VP_SRC_SWZ_Z_SHIFT) |
+	       (src.swz[3] << NV30_VP_SRC_SWZ_W_SHIFT));
+
+/*
+ * |VVV|
+ * d�.�b
+ *  \u/
+ *
+ */
+
+	switch (pos) {
+	case 0:
+		hw[1] |= ((sr & NV30_VP_SRC0_HIGH_MASK) >>
+			  NV30_VP_SRC0_HIGH_SHIFT) << NV30_VP_INST_SRC0H_SHIFT;
+		hw[2] |= (sr & NV30_VP_SRC0_LOW_MASK) <<
+			  NV30_VP_INST_SRC0L_SHIFT;
+		break;
+	case 1:
+		hw[2] |= sr << NV30_VP_INST_SRC1_SHIFT;
+		break;
+	case 2:
+		hw[2] |= ((sr & NV30_VP_SRC2_HIGH_MASK) >>
+			  NV30_VP_SRC2_HIGH_SHIFT) << NV30_VP_INST_SRC2H_SHIFT;
+		hw[3] |= (sr & NV30_VP_SRC2_LOW_MASK) <<
+			  NV30_VP_INST_SRC2L_SHIFT;
+		break;
+	default:
+		assert(0);
+	}
+}
+
+static void
+emit_dst(struct nv20_vpc *vpc, uint32_t *hw, int slot, struct nv20_sreg dst)
+{
+	struct nv20_vertex_program *vp = vpc->vp;
+
+	switch (dst.type) {
+	case NV30SR_TEMP:
+		hw[0] |= (dst.index << NV30_VP_INST_DEST_TEMP_ID_SHIFT);
+		break;
+	case NV30SR_OUTPUT:
+		switch (dst.index) {
+		case NV30_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break;
+		case NV30_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break;
+		case NV30_VP_INST_DEST_BFC0 : vp->or |= (1 << 2); break;
+		case NV30_VP_INST_DEST_BFC1 : vp->or |= (1 << 3); break;
+		case NV30_VP_INST_DEST_FOGC : vp->or |= (1 << 4); break;
+		case NV30_VP_INST_DEST_PSZ  : vp->or |= (1 << 5); break;
+		case NV30_VP_INST_DEST_TC(0): vp->or |= (1 << 14); break;
+		case NV30_VP_INST_DEST_TC(1): vp->or |= (1 << 15); break;
+		case NV30_VP_INST_DEST_TC(2): vp->or |= (1 << 16); break;
+		case NV30_VP_INST_DEST_TC(3): vp->or |= (1 << 17); break;
+		case NV30_VP_INST_DEST_TC(4): vp->or |= (1 << 18); break;
+		case NV30_VP_INST_DEST_TC(5): vp->or |= (1 << 19); break;
+		case NV30_VP_INST_DEST_TC(6): vp->or |= (1 << 20); break;
+		case NV30_VP_INST_DEST_TC(7): vp->or |= (1 << 21); break;
+		default:
+			break;
+		}
+
+		hw[3] |= (dst.index << NV30_VP_INST_DEST_SHIFT);
+		hw[0] |= NV30_VP_INST_VEC_DEST_TEMP_MASK | (1<<20);
+
+		/*XXX: no way this is entirely correct, someone needs to
+		 *     figure out what exactly it is.
+		 */
+		hw[3] |= 0x800;
+		break;
+	default:
+		assert(0);
+	}
+}
+
+static void
+nv20_vp_arith(struct nv20_vpc *vpc, int slot, int op,
+	      struct nv20_sreg dst, int mask,
+	      struct nv20_sreg s0, struct nv20_sreg s1,
+	      struct nv20_sreg s2)
+{
+	struct nv20_vertex_program *vp = vpc->vp;
+	uint32_t *hw;
+
+	vp->insns = realloc(vp->insns, ++vp->nr_insns * sizeof(*vpc->vpi));
+	vpc->vpi = &vp->insns[vp->nr_insns - 1];
+	memset(vpc->vpi, 0, sizeof(*vpc->vpi));
+	vpc->vpi->const_index = -1;
+
+	hw = vpc->vpi->data;
+
+	hw[0] |= (NV30_VP_INST_COND_TR << NV30_VP_INST_COND_SHIFT);
+	hw[0] |= ((0 << NV30_VP_INST_COND_SWZ_X_SHIFT) |
+		  (1 << NV30_VP_INST_COND_SWZ_Y_SHIFT) |
+		  (2 << NV30_VP_INST_COND_SWZ_Z_SHIFT) |
+		  (3 << NV30_VP_INST_COND_SWZ_W_SHIFT));
+
+	hw[1] |= (op << NV30_VP_INST_VEC_OPCODE_SHIFT);
+//	hw[3] |= NV30_VP_INST_SCA_DEST_TEMP_MASK;
+//	hw[3] |= (mask << NV30_VP_INST_VEC_WRITEMASK_SHIFT);
+
+	if (dst.type == NV30SR_OUTPUT) {
+		if (slot)
+			hw[3] |= (mask << NV30_VP_INST_SDEST_WRITEMASK_SHIFT);
+		else
+			hw[3] |= (mask << NV30_VP_INST_VDEST_WRITEMASK_SHIFT);
+	} else {
+		if (slot)
+			hw[3] |= (mask << NV30_VP_INST_STEMP_WRITEMASK_SHIFT);
+		else
+			hw[3] |= (mask << NV30_VP_INST_VTEMP_WRITEMASK_SHIFT);
+	}
+
+	emit_dst(vpc, hw, slot, dst);
+	emit_src(vpc, hw, 0, s0);
+	emit_src(vpc, hw, 1, s1);
+	emit_src(vpc, hw, 2, s2);
+}
+
+static INLINE struct nv20_sreg
+tgsi_src(struct nv20_vpc *vpc, const struct tgsi_full_src_register *fsrc) {
+	struct nv20_sreg src;
+
+	switch (fsrc->SrcRegister.File) {
+	case TGSI_FILE_INPUT:
+		src = nv20_sr(NV30SR_INPUT, fsrc->SrcRegister.Index);
+		break;
+	case TGSI_FILE_CONSTANT:
+		src = constant(vpc, fsrc->SrcRegister.Index, 0, 0, 0, 0);
+		break;
+	case TGSI_FILE_IMMEDIATE:
+		src = vpc->imm[fsrc->SrcRegister.Index];
+		break;
+	case TGSI_FILE_TEMPORARY:
+		if (vpc->high_temp < fsrc->SrcRegister.Index)
+			vpc->high_temp = fsrc->SrcRegister.Index;
+		src = nv20_sr(NV30SR_TEMP, fsrc->SrcRegister.Index);
+		break;
+	default:
+		NOUVEAU_ERR("bad src file\n");
+		break;
+	}
+
+	src.abs = fsrc->SrcRegisterExtMod.Absolute;
+	src.negate = fsrc->SrcRegister.Negate;
+	src.swz[0] = fsrc->SrcRegister.SwizzleX;
+	src.swz[1] = fsrc->SrcRegister.SwizzleY;
+	src.swz[2] = fsrc->SrcRegister.SwizzleZ;
+	src.swz[3] = fsrc->SrcRegister.SwizzleW;
+	return src;
+}
+
+static INLINE struct nv20_sreg
+tgsi_dst(struct nv20_vpc *vpc, const struct tgsi_full_dst_register *fdst) {
+	struct nv20_sreg dst;
+
+	switch (fdst->DstRegister.File) {
+	case TGSI_FILE_OUTPUT:
+		dst = nv20_sr(NV30SR_OUTPUT,
+			      vpc->output_map[fdst->DstRegister.Index]);
+
+		break;
+	case TGSI_FILE_TEMPORARY:
+		dst = nv20_sr(NV30SR_TEMP, fdst->DstRegister.Index);
+		if (vpc->high_temp < dst.index)
+			vpc->high_temp = dst.index;
+		break;
+	default:
+		NOUVEAU_ERR("bad dst file\n");
+		break;
+	}
+
+	return dst;
+}
+
+static INLINE int
+tgsi_mask(uint tgsi)
+{
+	int mask = 0;
+
+	if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X;
+	if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y;
+	if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z;
+	if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W;
+	return mask;
+}
+
+static boolean
+nv20_vertprog_parse_instruction(struct nv20_vpc *vpc,
+				const struct tgsi_full_instruction *finst)
+{
+	struct nv20_sreg src[3], dst, tmp;
+	struct nv20_sreg none = nv20_sr(NV30SR_NONE, 0);
+	int mask;
+	int ai = -1, ci = -1;
+	int i;
+
+	if (finst->Instruction.Opcode == TGSI_OPCODE_END)
+		return TRUE;
+
+	vpc->temp_temp_count = 0;
+	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fsrc;
+
+		fsrc = &finst->FullSrcRegisters[i];
+		if (fsrc->SrcRegister.File == TGSI_FILE_TEMPORARY) {
+			src[i] = tgsi_src(vpc, fsrc);
+		}
+	}
+
+	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fsrc;
+
+		fsrc = &finst->FullSrcRegisters[i];
+		switch (fsrc->SrcRegister.File) {
+		case TGSI_FILE_INPUT:
+			if (ai == -1 || ai == fsrc->SrcRegister.Index) {
+				ai = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(vpc, fsrc);
+			} else {
+				src[i] = temp(vpc);
+				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
+				      tgsi_src(vpc, fsrc), none, none);
+			}
+			break;
+		/*XXX: index comparison is broken now that consts come from
+		 *     two different register files.
+		 */
+		case TGSI_FILE_CONSTANT:
+		case TGSI_FILE_IMMEDIATE:
+			if (ci == -1 || ci == fsrc->SrcRegister.Index) {
+				ci = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(vpc, fsrc);
+			} else {
+				src[i] = temp(vpc);
+				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
+				      tgsi_src(vpc, fsrc), none, none);
+			}
+			break;
+		case TGSI_FILE_TEMPORARY:
+			/* handled above */
+			break;
+		default:
+			NOUVEAU_ERR("bad src file\n");
+			return FALSE;
+		}
+	}
+
+	dst  = tgsi_dst(vpc, &finst->FullDstRegisters[0]);
+	mask = tgsi_mask(finst->FullDstRegisters[0].DstRegister.WriteMask);
+
+	switch (finst->Instruction.Opcode) {
+	case TGSI_OPCODE_ABS:
+		arith(vpc, 0, OP_MOV, dst, mask, abs(src[0]), none, none);
+		break;
+	case TGSI_OPCODE_ADD:
+		arith(vpc, 0, OP_ADD, dst, mask, src[0], none, src[1]);
+		break;
+	case TGSI_OPCODE_ARL:
+		arith(vpc, 0, OP_ARL, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_DP3:
+		arith(vpc, 0, OP_DP3, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DP4:
+		arith(vpc, 0, OP_DP4, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DPH:
+		arith(vpc, 0, OP_DPH, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DST:
+		arith(vpc, 0, OP_DST, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_EX2:
+		arith(vpc, 1, OP_EX2, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_EXP:
+		arith(vpc, 1, OP_EXP, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_FLR:
+		arith(vpc, 0, OP_FLR, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_FRC:
+		arith(vpc, 0, OP_FRC, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_LG2:
+		arith(vpc, 1, OP_LG2, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_LIT:
+		arith(vpc, 1, OP_LIT, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_LOG:
+		arith(vpc, 1, OP_LOG, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_MAD:
+		arith(vpc, 0, OP_MAD, dst, mask, src[0], src[1], src[2]);
+		break;
+	case TGSI_OPCODE_MAX:
+		arith(vpc, 0, OP_MAX, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_MIN:
+		arith(vpc, 0, OP_MIN, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_MOV:
+		arith(vpc, 0, OP_MOV, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_MUL:
+		arith(vpc, 0, OP_MUL, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_POW:
+		tmp = temp(vpc);
+		arith(vpc, 1, OP_LG2, tmp, MASK_X, none, none,
+		      swz(src[0], X, X, X, X));
+		arith(vpc, 0, OP_MUL, tmp, MASK_X, swz(tmp, X, X, X, X),
+		      swz(src[1], X, X, X, X), none);
+		arith(vpc, 1, OP_EX2, dst, mask, none, none,
+		      swz(tmp, X, X, X, X));
+		break;
+	case TGSI_OPCODE_RCP:
+		arith(vpc, 1, OP_RCP, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_RET:
+		break;
+	case TGSI_OPCODE_RSQ:
+		arith(vpc, 1, OP_RSQ, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_SGE:
+		arith(vpc, 0, OP_SGE, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SGT:
+		arith(vpc, 0, OP_SGT, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SLT:
+		arith(vpc, 0, OP_SLT, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SUB:
+		arith(vpc, 0, OP_ADD, dst, mask, src[0], none, neg(src[1]));
+		break;
+	case TGSI_OPCODE_XPD:
+		tmp = temp(vpc);
+		arith(vpc, 0, OP_MUL, tmp, mask,
+		      swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none);
+		arith(vpc, 0, OP_MAD, dst, (mask & ~MASK_W),
+		      swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y),
+		      neg(tmp));
+		break;
+	default:
+		NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode);
+		return FALSE;
+	}
+
+	return TRUE;
+}
+
+static boolean
+nv20_vertprog_parse_decl_output(struct nv20_vpc *vpc,
+				const struct tgsi_full_declaration *fdec)
+{
+	int hw;
+
+	switch (fdec->Semantic.SemanticName) {
+	case TGSI_SEMANTIC_POSITION:
+		hw = NV30_VP_INST_DEST_POS;
+		break;
+	case TGSI_SEMANTIC_COLOR:
+		if (fdec->Semantic.SemanticIndex == 0) {
+			hw = NV30_VP_INST_DEST_COL0;
+		} else
+		if (fdec->Semantic.SemanticIndex == 1) {
+			hw = NV30_VP_INST_DEST_COL1;
+		} else {
+			NOUVEAU_ERR("bad colour semantic index\n");
+			return FALSE;
+		}
+		break;
+	case TGSI_SEMANTIC_BCOLOR:
+		if (fdec->Semantic.SemanticIndex == 0) {
+			hw = NV30_VP_INST_DEST_BFC0;
+		} else
+		if (fdec->Semantic.SemanticIndex == 1) {
+			hw = NV30_VP_INST_DEST_BFC1;
+		} else {
+			NOUVEAU_ERR("bad bcolour semantic index\n");
+			return FALSE;
+		}
+		break;
+	case TGSI_SEMANTIC_FOG:
+		hw = NV30_VP_INST_DEST_FOGC;
+		break;
+	case TGSI_SEMANTIC_PSIZE:
+		hw = NV30_VP_INST_DEST_PSZ;
+		break;
+	case TGSI_SEMANTIC_GENERIC:
+		if (fdec->Semantic.SemanticIndex <= 7) {
+			hw = NV30_VP_INST_DEST_TC(fdec->Semantic.SemanticIndex);
+		} else {
+			NOUVEAU_ERR("bad generic semantic index\n");
+			return FALSE;
+		}
+		break;
+	default:
+		NOUVEAU_ERR("bad output semantic\n");
+		return FALSE;
+	}
+
+	vpc->output_map[fdec->DeclarationRange.First] = hw;
+	return TRUE;
+}
+
+static boolean
+nv20_vertprog_prepare(struct nv20_vpc *vpc)
+{
+	struct tgsi_parse_context p;
+	int nr_imm = 0;
+
+	tgsi_parse_init(&p, vpc->vp->pipe.tokens);
+	while (!tgsi_parse_end_of_tokens(&p)) {
+		const union tgsi_full_token *tok = &p.FullToken;
+
+		tgsi_parse_token(&p);
+		switch(tok->Token.Type) {
+		case TGSI_TOKEN_TYPE_IMMEDIATE:
+			nr_imm++;
+			break;
+		default:
+			break;
+		}
+	}
+	tgsi_parse_free(&p);
+
+	if (nr_imm) {
+		vpc->imm = CALLOC(nr_imm, sizeof(struct nv20_sreg));
+		assert(vpc->imm);
+	}
+
+	return TRUE;
+}
+
+static void
+nv20_vertprog_translate(struct nv20_context *nv20,
+			struct nv20_vertex_program *vp)
+{
+	struct tgsi_parse_context parse;
+	struct nv20_vpc *vpc = NULL;
+
+	tgsi_dump(vp->pipe.tokens,0);
+
+	vpc = CALLOC(1, sizeof(struct nv20_vpc));
+	if (!vpc)
+		return;
+	vpc->vp = vp;
+	vpc->high_temp = -1;
+
+	if (!nv20_vertprog_prepare(vpc)) {
+		FREE(vpc);
+		return;
+	}
+
+	tgsi_parse_init(&parse, vp->pipe.tokens);
+
+	while (!tgsi_parse_end_of_tokens(&parse)) {
+		tgsi_parse_token(&parse);
+
+		switch (parse.FullToken.Token.Type) {
+		case TGSI_TOKEN_TYPE_DECLARATION:
+		{
+			const struct tgsi_full_declaration *fdec;
+			fdec = &parse.FullToken.FullDeclaration;
+			switch (fdec->Declaration.File) {
+			case TGSI_FILE_OUTPUT:
+				if (!nv20_vertprog_parse_decl_output(vpc, fdec))
+					goto out_err;
+				break;
+			default:
+				break;
+			}
+		}
+			break;
+		case TGSI_TOKEN_TYPE_IMMEDIATE:
+		{
+			const struct tgsi_full_immediate *imm;
+
+			imm = &parse.FullToken.FullImmediate;
+			assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32);
+			assert(imm->Immediate.NrTokens == 4 + 1);
+			vpc->imm[vpc->nr_imm++] =
+				constant(vpc, -1,
+					 imm->u.ImmediateFloat32[0].Float,
+					 imm->u.ImmediateFloat32[1].Float,
+					 imm->u.ImmediateFloat32[2].Float,
+					 imm->u.ImmediateFloat32[3].Float);
+		}
+			break;
+		case TGSI_TOKEN_TYPE_INSTRUCTION:
+		{
+			const struct tgsi_full_instruction *finst;
+			finst = &parse.FullToken.FullInstruction;
+			if (!nv20_vertprog_parse_instruction(vpc, finst))
+				goto out_err;
+		}
+			break;
+		default:
+			break;
+		}
+	}
+
+	vp->insns[vp->nr_insns - 1].data[3] |= NV30_VP_INST_LAST;
+	vp->translated = TRUE;
+out_err:
+	tgsi_parse_free(&parse);
+	FREE(vpc);
+}
+
+static boolean
+nv20_vertprog_validate(struct nv20_context *nv20)
+{ 
+	struct nouveau_winsys *nvws = nv20->nvws;
+	struct pipe_winsys *ws = nv20->pipe.winsys;
+	struct nouveau_grobj *rankine = nv20->screen->rankine;
+	struct nv20_vertex_program *vp;
+	struct pipe_buffer *constbuf;
+	boolean upload_code = FALSE, upload_data = FALSE;
+	int i;
+
+	vp = nv20->vertprog;
+	constbuf = nv20->constbuf[PIPE_SHADER_VERTEX];
+
+	/* Translate TGSI shader into hw bytecode */
+	if (!vp->translated) {
+		nv20_vertprog_translate(nv20, vp);
+		if (!vp->translated)
+			return FALSE;
+	}
+
+	/* Allocate hw vtxprog exec slots */
+	if (!vp->exec) {
+		struct nouveau_resource *heap = nv20->screen->vp_exec_heap;
+		struct nouveau_stateobj *so;
+		uint vplen = vp->nr_insns;
+
+		if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) {
+			while (heap->next && heap->size < vplen) {
+				struct nv20_vertex_program *evict;
+				
+				evict = heap->next->priv;
+				nvws->res_free(&evict->exec);
+			}
+
+			if (nvws->res_alloc(heap, vplen, vp, &vp->exec))
+				assert(0);
+		}
+
+		so = so_new(2, 0);
+		so_method(so, rankine, NV34TCL_VP_START_FROM_ID, 1);
+		so_data  (so, vp->exec->start);
+		so_ref(so, &vp->so);
+
+		upload_code = TRUE;
+	}
+
+	/* Allocate hw vtxprog const slots */
+	if (vp->nr_consts && !vp->data) {
+		struct nouveau_resource *heap = nv20->screen->vp_data_heap;
+
+		if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) {
+			while (heap->next && heap->size < vp->nr_consts) {
+				struct nv20_vertex_program *evict;
+				
+				evict = heap->next->priv;
+				nvws->res_free(&evict->data);
+			}
+
+			if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data))
+				assert(0);
+		}
+
+		/*XXX: handle this some day */
+		assert(vp->data->start >= vp->data_start_min);
+
+		upload_data = TRUE;
+		if (vp->data_start != vp->data->start)
+			upload_code = TRUE;
+	}
+
+	/* If exec or data segments moved we need to patch the program to
+	 * fixup offsets and register IDs.
+	 */
+	if (vp->exec_start != vp->exec->start) {
+		for (i = 0; i < vp->nr_insns; i++) {
+			struct nv20_vertex_program_exec *vpi = &vp->insns[i];
+
+			if (vpi->has_branch_offset) {
+				assert(0);
+			}
+		}
+
+		vp->exec_start = vp->exec->start;
+	}
+
+	if (vp->nr_consts && vp->data_start != vp->data->start) {
+		for (i = 0; i < vp->nr_insns; i++) {
+			struct nv20_vertex_program_exec *vpi = &vp->insns[i];
+
+			if (vpi->const_index >= 0) {
+				vpi->data[1] &= ~NV30_VP_INST_CONST_SRC_MASK;
+				vpi->data[1] |=
+					(vpi->const_index + vp->data->start) <<
+					NV30_VP_INST_CONST_SRC_SHIFT;
+
+			}
+		}
+
+		vp->data_start = vp->data->start;
+	}
+
+	/* Update + Upload constant values */
+	if (vp->nr_consts) {
+		float *map = NULL;
+
+		if (constbuf) {
+			map = ws->buffer_map(ws, constbuf,
+					     PIPE_BUFFER_USAGE_CPU_READ);
+		}
+
+		for (i = 0; i < vp->nr_consts; i++) {
+			struct nv20_vertex_program_data *vpd = &vp->consts[i];
+
+			if (vpd->index >= 0) {
+				if (!upload_data &&
+				    !memcmp(vpd->value, &map[vpd->index * 4],
+					    4 * sizeof(float)))
+					continue;
+				memcpy(vpd->value, &map[vpd->index * 4],
+				       4 * sizeof(float));
+			}
+
+			BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_CONST_ID, 5);
+			OUT_RING  (i + vp->data->start);
+			OUT_RINGp ((uint32_t *)vpd->value, 4);
+		}
+
+		if (constbuf) {
+			ws->buffer_unmap(ws, constbuf);
+		}
+	}
+
+	/* Upload vtxprog */
+	if (upload_code) {
+#if 0
+		for (i = 0; i < vp->nr_insns; i++) {
+			NOUVEAU_MSG("VP inst %d: 0x%08x 0x%08x 0x%08x 0x%08x\n",
+				i, vp->insns[i].data[0], vp->insns[i].data[1],
+				vp->insns[i].data[2], vp->insns[i].data[3]);
+		}
+#endif
+		BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_FROM_ID, 1);
+		OUT_RING  (vp->exec->start);
+		for (i = 0; i < vp->nr_insns; i++) {
+			BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_INST(0), 4);
+			OUT_RINGp (vp->insns[i].data, 4);
+		}
+	}
+
+	if (vp->so != nv20->state.hw[NV30_STATE_VERTPROG]) {
+		so_ref(vp->so, &nv20->state.hw[NV30_STATE_VERTPROG]);
+		return TRUE;
+	}
+
+	return FALSE;
+}
+
+void
+nv20_vertprog_destroy(struct nv20_context *nv20, struct nv20_vertex_program *vp)
+{
+	struct nouveau_winsys *nvws = nv20->screen->nvws;
+
+	vp->translated = FALSE;
+
+	if (vp->nr_insns) {
+		FREE(vp->insns);
+		vp->insns = NULL;
+		vp->nr_insns = 0;
+	}
+
+	if (vp->nr_consts) {
+		FREE(vp->consts);
+		vp->consts = NULL;
+		vp->nr_consts = 0;
+	}
+
+	nvws->res_free(&vp->exec);
+	vp->exec_start = 0;
+	nvws->res_free(&vp->data);
+	vp->data_start = 0;
+	vp->data_start_min = 0;
+
+	vp->ir = vp->or = 0;
+	so_ref(NULL, &vp->so);
+}
+
+struct nv20_state_entry nv20_state_vertprog = {
+	.validate = nv20_vertprog_validate,
+	.dirty = {
+		.pipe = NV30_NEW_VERTPROG /*| NV30_NEW_UCP*/,
+		.hw = NV30_STATE_VERTPROG,
+	}
+};
diff --git a/src/gallium/drivers/nv30/Makefile b/src/gallium/drivers/nv30/Makefile
new file mode 100644
index 0000000000..69f2790dfe
--- /dev/null
+++ b/src/gallium/drivers/nv30/Makefile
@@ -0,0 +1,37 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nv30
+
+DRIVER_SOURCES = \
+	nv30_clear.c \
+	nv30_context.c \
+	nv30_draw.c \
+	nv30_fragprog.c \
+	nv30_fragtex.c \
+	nv30_miptree.c \
+	nv30_query.c \
+	nv30_screen.c \
+	nv30_state.c \
+	nv30_state_blend.c \
+	nv30_state_emit.c \
+	nv30_state_fb.c \
+	nv30_state_rasterizer.c \
+	nv30_state_scissor.c \
+	nv30_state_stipple.c \
+	nv30_state_viewport.c \
+	nv30_state_zsa.c \
+	nv30_surface.c \
+	nv30_vbo.c \
+	nv30_vertprog.c
+
+C_SOURCES = \
+	$(COMMON_SOURCES) \
+	$(DRIVER_SOURCES)
+
+ASM_SOURCES = 
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/drivers/nv30/nv30_clear.c b/src/gallium/drivers/nv30/nv30_clear.c
new file mode 100644
index 0000000000..8c3ca204d5
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_clear.c
@@ -0,0 +1,13 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "nv30_context.h"
+
+void
+nv30_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+	   unsigned clearValue)
+{
+	pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue);
+	ps->status = PIPE_SURFACE_STATUS_CLEAR;
+}
diff --git a/src/gallium/drivers/nv30/nv30_context.c b/src/gallium/drivers/nv30/nv30_context.c
new file mode 100644
index 0000000000..61654f8756
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_context.c
@@ -0,0 +1,72 @@
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+#include "nv30_context.h"
+#include "nv30_screen.h"
+
+static void
+nv30_flush(struct pipe_context *pipe, unsigned flags,
+	   struct pipe_fence_handle **fence)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	
+	if (flags & PIPE_FLUSH_TEXTURE_CACHE) {
+		BEGIN_RING(rankine, 0x1fd8, 1);
+		OUT_RING  (2);
+		BEGIN_RING(rankine, 0x1fd8, 1);
+		OUT_RING  (1);
+	}
+
+	FIRE_RING(fence);
+}
+
+static void
+nv30_destroy(struct pipe_context *pipe)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	if (nv30->draw)
+		draw_destroy(nv30->draw);
+	FREE(nv30);
+}
+
+struct pipe_context *
+nv30_create(struct pipe_screen *pscreen, unsigned pctx_id)
+{
+	struct nv30_screen *screen = nv30_screen(pscreen);
+	struct pipe_winsys *ws = pscreen->winsys;
+	struct nv30_context *nv30;
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nv30 = CALLOC(1, sizeof(struct nv30_context));
+	if (!nv30)
+		return NULL;
+	nv30->screen = screen;
+	nv30->pctx_id = pctx_id;
+
+	nv30->nvws = nvws;
+
+	nv30->pipe.winsys = ws;
+	nv30->pipe.screen = pscreen;
+	nv30->pipe.destroy = nv30_destroy;
+	nv30->pipe.draw_arrays = nv30_draw_arrays;
+	nv30->pipe.draw_elements = nv30_draw_elements;
+	nv30->pipe.clear = nv30_clear;
+	nv30->pipe.flush = nv30_flush;
+
+	nv30_init_query_functions(nv30);
+	nv30_init_surface_functions(nv30);
+	nv30_init_state_functions(nv30);
+
+	/* Create, configure, and install fallback swtnl path */
+	nv30->draw = draw_create();
+	draw_wide_point_threshold(nv30->draw, 9999999.0);
+	draw_wide_line_threshold(nv30->draw, 9999999.0);
+	draw_enable_line_stipple(nv30->draw, FALSE);
+	draw_enable_point_sprites(nv30->draw, FALSE);
+	draw_set_rasterize_stage(nv30->draw, nv30_draw_render_stage(nv30));
+
+	return &nv30->pipe;
+}
+	
diff --git a/src/gallium/drivers/nv30/nv30_context.h b/src/gallium/drivers/nv30/nv30_context.h
new file mode 100644
index 0000000000..b933769700
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_context.h
@@ -0,0 +1,212 @@
+#ifndef __NV30_CONTEXT_H__
+#define __NV30_CONTEXT_H__
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_compiler.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "draw/draw_vertex.h"
+
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau/nouveau_gldefs.h"
+
+#define NOUVEAU_PUSH_CONTEXT(ctx)                                              \
+	struct nv30_screen *ctx = nv30->screen
+#include "nouveau/nouveau_push.h"
+#include "nouveau/nouveau_stateobj.h"
+
+#include "nv30_state.h"
+
+#define NOUVEAU_ERR(fmt, args...) \
+	fprintf(stderr, "%s:%d -  "fmt, __func__, __LINE__, ##args);
+#define NOUVEAU_MSG(fmt, args...) \
+	fprintf(stderr, "nouveau: "fmt, ##args);
+
+enum nv30_state_index {
+	NV30_STATE_FB = 0,
+	NV30_STATE_VIEWPORT = 1,
+	NV30_STATE_BLEND = 2,
+	NV30_STATE_RAST = 3,
+	NV30_STATE_ZSA = 4,
+	NV30_STATE_BCOL = 5,
+	NV30_STATE_CLIP = 6,
+	NV30_STATE_SCISSOR = 7,
+	NV30_STATE_STIPPLE = 8,
+	NV30_STATE_FRAGPROG = 9,
+	NV30_STATE_VERTPROG = 10,
+	NV30_STATE_FRAGTEX0 = 11,
+	NV30_STATE_FRAGTEX1 = 12,
+	NV30_STATE_FRAGTEX2 = 13,
+	NV30_STATE_FRAGTEX3 = 14,
+	NV30_STATE_FRAGTEX4 = 15,
+	NV30_STATE_FRAGTEX5 = 16,
+	NV30_STATE_FRAGTEX6 = 17,
+	NV30_STATE_FRAGTEX7 = 18,
+	NV30_STATE_FRAGTEX8 = 19,
+	NV30_STATE_FRAGTEX9 = 20,
+	NV30_STATE_FRAGTEX10 = 21,
+	NV30_STATE_FRAGTEX11 = 22,
+	NV30_STATE_FRAGTEX12 = 23,
+	NV30_STATE_FRAGTEX13 = 24,
+	NV30_STATE_FRAGTEX14 = 25,
+	NV30_STATE_FRAGTEX15 = 26,
+	NV30_STATE_VERTTEX0 = 27,
+	NV30_STATE_VERTTEX1 = 28,
+	NV30_STATE_VERTTEX2 = 29,
+	NV30_STATE_VERTTEX3 = 30,
+	NV30_STATE_VTXBUF = 31,
+	NV30_STATE_VTXFMT = 32,
+	NV30_STATE_VTXATTR = 33,
+	NV30_STATE_MAX = 34
+};
+
+#include "nv30_screen.h"
+
+#define NV30_NEW_BLEND		(1 <<  0)
+#define NV30_NEW_RAST		(1 <<  1)
+#define NV30_NEW_ZSA		(1 <<  2)
+#define NV30_NEW_SAMPLER	(1 <<  3)
+#define NV30_NEW_FB		(1 <<  4)
+#define NV30_NEW_STIPPLE	(1 <<  5)
+#define NV30_NEW_SCISSOR	(1 <<  6)
+#define NV30_NEW_VIEWPORT	(1 <<  7)
+#define NV30_NEW_BCOL		(1 <<  8)
+#define NV30_NEW_VERTPROG	(1 <<  9)
+#define NV30_NEW_FRAGPROG	(1 << 10)
+#define NV30_NEW_ARRAYS		(1 << 11)
+#define NV30_NEW_UCP		(1 << 12)
+
+struct nv30_rasterizer_state {
+	struct pipe_rasterizer_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+struct nv30_zsa_state {
+	struct pipe_depth_stencil_alpha_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+struct nv30_blend_state {
+	struct pipe_blend_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+
+struct nv30_state {
+	unsigned scissor_enabled;
+	unsigned stipple_enabled;
+	unsigned viewport_bypass;
+	unsigned fp_samplers;
+
+	uint64_t dirty;
+	struct nouveau_stateobj *hw[NV30_STATE_MAX];
+};
+
+struct nv30_context {
+	struct pipe_context pipe;
+
+	struct nouveau_winsys *nvws;
+	struct nv30_screen *screen;
+	unsigned pctx_id;
+
+	struct draw_context *draw;
+
+	/* HW state derived from pipe states */
+	struct nv30_state state;
+
+	/* Context state */
+	unsigned dirty;
+	struct pipe_scissor_state scissor;
+	unsigned stipple[32];
+	struct nv30_vertex_program *vertprog;
+	struct nv30_fragment_program *fragprog;
+	struct pipe_buffer *constbuf[PIPE_SHADER_TYPES];
+	unsigned constbuf_nr[PIPE_SHADER_TYPES];
+	struct nv30_rasterizer_state *rasterizer;
+	struct nv30_zsa_state *zsa;
+	struct nv30_blend_state *blend;
+	struct pipe_blend_color blend_colour;
+	struct pipe_viewport_state viewport;
+	struct pipe_framebuffer_state framebuffer;
+	struct pipe_buffer *idxbuf;
+	unsigned idxbuf_format;
+	struct nv30_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS];
+	struct nv30_miptree *tex_miptree[PIPE_MAX_SAMPLERS];
+	unsigned nr_samplers;
+	unsigned nr_textures;
+	unsigned dirty_samplers;
+	struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
+	unsigned vtxbuf_nr;
+	struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS];
+	unsigned vtxelt_nr;
+	const unsigned *edgeflags;
+};
+
+static INLINE struct nv30_context *
+nv30_context(struct pipe_context *pipe)
+{
+	return (struct nv30_context *)pipe;
+}
+
+struct nv30_state_entry {
+	boolean (*validate)(struct nv30_context *nv30);
+	struct {
+		unsigned pipe;
+		unsigned hw;
+	} dirty;
+};
+
+extern void nv30_init_state_functions(struct nv30_context *nv30);
+extern void nv30_init_surface_functions(struct nv30_context *nv30);
+extern void nv30_init_query_functions(struct nv30_context *nv30);
+
+extern void nv30_screen_init_miptree_functions(struct pipe_screen *pscreen);
+
+/* nv30_draw.c */
+extern struct draw_stage *nv30_draw_render_stage(struct nv30_context *nv30);
+
+/* nv30_vertprog.c */
+extern void nv30_vertprog_destroy(struct nv30_context *,
+				  struct nv30_vertex_program *);
+
+/* nv30_fragprog.c */
+extern void nv30_fragprog_destroy(struct nv30_context *,
+				  struct nv30_fragment_program *);
+
+/* nv30_fragtex.c */
+extern void nv30_fragtex_bind(struct nv30_context *);
+
+/* nv30_state.c and friends */
+extern boolean nv30_state_validate(struct nv30_context *nv30);
+extern void nv30_state_emit(struct nv30_context *nv30);
+extern struct nv30_state_entry nv30_state_rasterizer;
+extern struct nv30_state_entry nv30_state_scissor;
+extern struct nv30_state_entry nv30_state_stipple;
+extern struct nv30_state_entry nv30_state_fragprog;
+extern struct nv30_state_entry nv30_state_vertprog;
+extern struct nv30_state_entry nv30_state_blend;
+extern struct nv30_state_entry nv30_state_blend_colour;
+extern struct nv30_state_entry nv30_state_zsa;
+extern struct nv30_state_entry nv30_state_viewport;
+extern struct nv30_state_entry nv30_state_framebuffer;
+extern struct nv30_state_entry nv30_state_fragtex;
+extern struct nv30_state_entry nv30_state_vbo;
+
+/* nv30_vbo.c */
+extern boolean nv30_draw_arrays(struct pipe_context *, unsigned mode,
+				unsigned start, unsigned count);
+extern boolean nv30_draw_elements(struct pipe_context *pipe,
+				  struct pipe_buffer *indexBuffer,
+				  unsigned indexSize,
+				  unsigned mode, unsigned start,
+				  unsigned count);
+
+/* nv30_clear.c */
+extern void nv30_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+		       unsigned clearValue);
+
+#endif
diff --git a/src/gallium/drivers/nv30/nv30_draw.c b/src/gallium/drivers/nv30/nv30_draw.c
new file mode 100644
index 0000000000..74fc138c05
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_draw.c
@@ -0,0 +1,61 @@
+#include "draw/draw_pipe.h"
+
+#include "nv30_context.h"
+
+struct nv30_draw_stage {
+	struct draw_stage draw;
+	struct nv30_context *nv30;
+};
+
+static void
+nv30_draw_point(struct draw_stage *draw, struct prim_header *prim)
+{
+	NOUVEAU_ERR("\n");
+}
+
+static void
+nv30_draw_line(struct draw_stage *draw, struct prim_header *prim)
+{
+	NOUVEAU_ERR("\n");
+}
+
+static void
+nv30_draw_tri(struct draw_stage *draw, struct prim_header *prim)
+{
+	NOUVEAU_ERR("\n");
+}
+
+static void
+nv30_draw_flush(struct draw_stage *draw, unsigned flags)
+{
+}
+
+static void
+nv30_draw_reset_stipple_counter(struct draw_stage *draw)
+{
+	NOUVEAU_ERR("\n");
+}
+
+static void
+nv30_draw_destroy(struct draw_stage *draw)
+{
+	FREE(draw);
+}
+
+struct draw_stage *
+nv30_draw_render_stage(struct nv30_context *nv30)
+{
+	struct nv30_draw_stage *nv30draw = CALLOC_STRUCT(nv30_draw_stage);
+
+	nv30draw->nv30 = nv30;
+	nv30draw->draw.draw = nv30->draw;
+	nv30draw->draw.point = nv30_draw_point;
+	nv30draw->draw.line = nv30_draw_line;
+	nv30draw->draw.tri = nv30_draw_tri;
+	nv30draw->draw.flush = nv30_draw_flush;
+	nv30draw->draw.reset_stipple_counter = nv30_draw_reset_stipple_counter;
+	nv30draw->draw.destroy = nv30_draw_destroy;
+
+	return &nv30draw->draw;
+}
+
diff --git a/src/gallium/drivers/nv30/nv30_fragprog.c b/src/gallium/drivers/nv30/nv30_fragprog.c
new file mode 100644
index 0000000000..320ba3f4bf
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_fragprog.c
@@ -0,0 +1,911 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+
+#include "nv30_context.h"
+
+#define SWZ_X 0
+#define SWZ_Y 1
+#define SWZ_Z 2
+#define SWZ_W 3
+#define MASK_X 1
+#define MASK_Y 2
+#define MASK_Z 4
+#define MASK_W 8
+#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W)
+#define DEF_SCALE NV30_FP_OP_DST_SCALE_1X
+#define DEF_CTEST NV30_FP_OP_COND_TR
+#include "nv30_shader.h"
+
+#define swz(s,x,y,z,w) nv30_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w)
+#define neg(s) nv30_sr_neg((s))
+#define abs(s) nv30_sr_abs((s))
+#define scale(s,v) nv30_sr_scale((s), NV30_FP_OP_DST_SCALE_##v)
+
+#define MAX_CONSTS 128
+#define MAX_IMM 32
+struct nv30_fpc {
+	struct nv30_fragment_program *fp;
+
+	uint attrib_map[PIPE_MAX_SHADER_INPUTS];
+
+	int high_temp;
+	int temp_temp_count;
+	int num_regs;
+
+	uint depth_id;
+	uint colour_id;
+
+	unsigned inst_offset;
+
+	struct {
+		int pipe;
+		float vals[4];
+	} consts[MAX_CONSTS];
+	int nr_consts;
+
+	struct nv30_sreg imm[MAX_IMM];
+	unsigned nr_imm;
+};
+
+static INLINE struct nv30_sreg
+temp(struct nv30_fpc *fpc)
+{
+	int idx;
+
+	idx  = fpc->temp_temp_count++;
+	idx += fpc->high_temp + 1;
+	return nv30_sr(NV30SR_TEMP, idx);
+}
+
+static INLINE struct nv30_sreg
+constant(struct nv30_fpc *fpc, int pipe, float vals[4])
+{
+	int idx;
+
+	if (fpc->nr_consts == MAX_CONSTS)
+		assert(0);
+	idx = fpc->nr_consts++;
+
+	fpc->consts[idx].pipe = pipe;
+	if (pipe == -1)
+		memcpy(fpc->consts[idx].vals, vals, 4 * sizeof(float));
+	return nv30_sr(NV30SR_CONST, idx);
+}
+
+#define arith(cc,s,o,d,m,s0,s1,s2) \
+	nv30_fp_arith((cc), (s), NV30_FP_OP_OPCODE_##o, \
+			(d), (m), (s0), (s1), (s2))
+#define tex(cc,s,o,u,d,m,s0,s1,s2) \
+	nv30_fp_tex((cc), (s), NV30_FP_OP_OPCODE_##o, (u), \
+		    (d), (m), (s0), none, none)
+
+static void
+grow_insns(struct nv30_fpc *fpc, int size)
+{
+	struct nv30_fragment_program *fp = fpc->fp;
+
+	fp->insn_len += size;
+	fp->insn = realloc(fp->insn, sizeof(uint32_t) * fp->insn_len);
+}
+
+static void
+emit_src(struct nv30_fpc *fpc, int pos, struct nv30_sreg src)
+{
+	struct nv30_fragment_program *fp = fpc->fp;
+	uint32_t *hw = &fp->insn[fpc->inst_offset];
+	uint32_t sr = 0;
+
+	switch (src.type) {
+	case NV30SR_INPUT:
+		sr |= (NV30_FP_REG_TYPE_INPUT << NV30_FP_REG_TYPE_SHIFT);
+		hw[0] |= (src.index << NV30_FP_OP_INPUT_SRC_SHIFT);
+		break;
+	case NV30SR_OUTPUT:
+		sr |= NV30_FP_REG_SRC_HALF;
+		/* fall-through */
+	case NV30SR_TEMP:
+		sr |= (NV30_FP_REG_TYPE_TEMP << NV30_FP_REG_TYPE_SHIFT);
+		sr |= (src.index << NV30_FP_REG_SRC_SHIFT);
+		break;
+	case NV30SR_CONST:
+		grow_insns(fpc, 4);
+		hw = &fp->insn[fpc->inst_offset];
+		if (fpc->consts[src.index].pipe >= 0) {
+			struct nv30_fragment_program_data *fpd;
+
+			fp->consts = realloc(fp->consts, ++fp->nr_consts *
+					     sizeof(*fpd));
+			fpd = &fp->consts[fp->nr_consts - 1];
+			fpd->offset = fpc->inst_offset + 4;
+			fpd->index = fpc->consts[src.index].pipe;
+			memset(&fp->insn[fpd->offset], 0, sizeof(uint32_t) * 4);
+		} else {
+			memcpy(&fp->insn[fpc->inst_offset + 4],
+				fpc->consts[src.index].vals,
+				sizeof(uint32_t) * 4);
+		}
+
+		sr |= (NV30_FP_REG_TYPE_CONST << NV30_FP_REG_TYPE_SHIFT);	
+		break;
+	case NV30SR_NONE:
+		sr |= (NV30_FP_REG_TYPE_INPUT << NV30_FP_REG_TYPE_SHIFT);
+		break;
+	default:
+		assert(0);
+	}
+
+	if (src.negate)
+		sr |= NV30_FP_REG_NEGATE;
+
+	if (src.abs)
+		hw[1] |= (1 << (29 + pos));
+
+	sr |= ((src.swz[0] << NV30_FP_REG_SWZ_X_SHIFT) |
+	       (src.swz[1] << NV30_FP_REG_SWZ_Y_SHIFT) |
+	       (src.swz[2] << NV30_FP_REG_SWZ_Z_SHIFT) |
+	       (src.swz[3] << NV30_FP_REG_SWZ_W_SHIFT));
+
+	hw[pos + 1] |= sr;
+}
+
+static void
+emit_dst(struct nv30_fpc *fpc, struct nv30_sreg dst)
+{
+	struct nv30_fragment_program *fp = fpc->fp;
+	uint32_t *hw = &fp->insn[fpc->inst_offset];
+
+	switch (dst.type) {
+	case NV30SR_TEMP:
+		if (fpc->num_regs < (dst.index + 1))
+			fpc->num_regs = dst.index + 1;
+		break;
+	case NV30SR_OUTPUT:
+		if (dst.index == 1) {
+			fp->fp_control |= 0xe;
+		} else {
+			hw[0] |= NV30_FP_OP_OUT_REG_HALF;
+		}
+		break;
+	case NV30SR_NONE:
+		hw[0] |= (1 << 30);
+		break;
+	default:
+		assert(0);
+	}
+
+	hw[0] |= (dst.index << NV30_FP_OP_OUT_REG_SHIFT);
+}
+
+static void
+nv30_fp_arith(struct nv30_fpc *fpc, int sat, int op,
+	      struct nv30_sreg dst, int mask,
+	      struct nv30_sreg s0, struct nv30_sreg s1, struct nv30_sreg s2)
+{
+	struct nv30_fragment_program *fp = fpc->fp;
+	uint32_t *hw;
+
+	fpc->inst_offset = fp->insn_len;
+	grow_insns(fpc, 4);
+	hw = &fp->insn[fpc->inst_offset];
+	memset(hw, 0, sizeof(uint32_t) * 4);
+
+	if (op == NV30_FP_OP_OPCODE_KIL)
+		fp->fp_control |= NV34TCL_FP_CONTROL_USES_KIL;
+	hw[0] |= (op << NV30_FP_OP_OPCODE_SHIFT);
+	hw[0] |= (mask << NV30_FP_OP_OUTMASK_SHIFT);
+	hw[2] |= (dst.dst_scale << NV30_FP_OP_DST_SCALE_SHIFT);
+
+	if (sat)
+		hw[0] |= NV30_FP_OP_OUT_SAT;
+
+	if (dst.cc_update)
+		hw[0] |= NV30_FP_OP_COND_WRITE_ENABLE;
+	hw[1] |= (dst.cc_test << NV30_FP_OP_COND_SHIFT);
+	hw[1] |= ((dst.cc_swz[0] << NV30_FP_OP_COND_SWZ_X_SHIFT) |
+		  (dst.cc_swz[1] << NV30_FP_OP_COND_SWZ_Y_SHIFT) |
+		  (dst.cc_swz[2] << NV30_FP_OP_COND_SWZ_Z_SHIFT) |
+		  (dst.cc_swz[3] << NV30_FP_OP_COND_SWZ_W_SHIFT));
+
+	emit_dst(fpc, dst);
+	emit_src(fpc, 0, s0);
+	emit_src(fpc, 1, s1);
+	emit_src(fpc, 2, s2);
+}
+
+static void
+nv30_fp_tex(struct nv30_fpc *fpc, int sat, int op, int unit,
+	    struct nv30_sreg dst, int mask,
+	    struct nv30_sreg s0, struct nv30_sreg s1, struct nv30_sreg s2)
+{
+	struct nv30_fragment_program *fp = fpc->fp;
+
+	nv30_fp_arith(fpc, sat, op, dst, mask, s0, s1, s2);
+
+	fp->insn[fpc->inst_offset] |= (unit << NV30_FP_OP_TEX_UNIT_SHIFT);
+	fp->samplers |= (1 << unit);
+}
+
+static INLINE struct nv30_sreg
+tgsi_src(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc)
+{
+	struct nv30_sreg src;
+
+	switch (fsrc->SrcRegister.File) {
+	case TGSI_FILE_INPUT:
+		src = nv30_sr(NV30SR_INPUT,
+			      fpc->attrib_map[fsrc->SrcRegister.Index]);
+		break;
+	case TGSI_FILE_CONSTANT:
+		src = constant(fpc, fsrc->SrcRegister.Index, NULL);
+		break;
+	case TGSI_FILE_IMMEDIATE:
+		assert(fsrc->SrcRegister.Index < fpc->nr_imm);
+		src = fpc->imm[fsrc->SrcRegister.Index];
+		break;
+	case TGSI_FILE_TEMPORARY:
+		src = nv30_sr(NV30SR_TEMP, fsrc->SrcRegister.Index + 1);
+		if (fpc->high_temp < src.index)
+			fpc->high_temp = src.index;
+		break;
+	/* This is clearly insane, but gallium hands us shaders like this.
+	 * Luckily fragprog results are just temp regs..
+	 */
+	case TGSI_FILE_OUTPUT:
+		if (fsrc->SrcRegister.Index == fpc->colour_id)
+			return nv30_sr(NV30SR_OUTPUT, 0);
+		else
+			return nv30_sr(NV30SR_OUTPUT, 1);
+		break;
+	default:
+		NOUVEAU_ERR("bad src file\n");
+		break;
+	}
+
+	src.abs = fsrc->SrcRegisterExtMod.Absolute;
+	src.negate = fsrc->SrcRegister.Negate;
+	src.swz[0] = fsrc->SrcRegister.SwizzleX;
+	src.swz[1] = fsrc->SrcRegister.SwizzleY;
+	src.swz[2] = fsrc->SrcRegister.SwizzleZ;
+	src.swz[3] = fsrc->SrcRegister.SwizzleW;
+	return src;
+}
+
+static INLINE struct nv30_sreg
+tgsi_dst(struct nv30_fpc *fpc, const struct tgsi_full_dst_register *fdst) {
+	int idx;
+
+	switch (fdst->DstRegister.File) {
+	case TGSI_FILE_OUTPUT:
+		if (fdst->DstRegister.Index == fpc->colour_id)
+			return nv30_sr(NV30SR_OUTPUT, 0);
+		else
+			return nv30_sr(NV30SR_OUTPUT, 1);
+		break;
+	case TGSI_FILE_TEMPORARY:
+		idx = fdst->DstRegister.Index + 1;
+		if (fpc->high_temp < idx)
+			fpc->high_temp = idx;
+		return nv30_sr(NV30SR_TEMP, idx);
+	case TGSI_FILE_NULL:
+		return nv30_sr(NV30SR_NONE, 0);
+	default:
+		NOUVEAU_ERR("bad dst file %d\n", fdst->DstRegister.File);
+		return nv30_sr(NV30SR_NONE, 0);
+	}
+}
+
+static INLINE int
+tgsi_mask(uint tgsi)
+{
+	int mask = 0;
+
+	if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X;
+	if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y;
+	if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z;
+	if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W;
+	return mask;
+}
+
+static boolean
+src_native_swz(struct nv30_fpc *fpc, const struct tgsi_full_src_register *fsrc,
+	       struct nv30_sreg *src)
+{
+	const struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0);
+	struct nv30_sreg tgsi = tgsi_src(fpc, fsrc);
+	uint mask = 0, zero_mask = 0, one_mask = 0, neg_mask = 0;
+	uint neg[4] = { fsrc->SrcRegisterExtSwz.NegateX,
+			fsrc->SrcRegisterExtSwz.NegateY,
+			fsrc->SrcRegisterExtSwz.NegateZ,
+			fsrc->SrcRegisterExtSwz.NegateW };
+	uint c;
+
+	for (c = 0; c < 4; c++) {
+		switch (tgsi_util_get_full_src_register_extswizzle(fsrc, c)) {
+		case TGSI_EXTSWIZZLE_X:
+		case TGSI_EXTSWIZZLE_Y:
+		case TGSI_EXTSWIZZLE_Z:
+		case TGSI_EXTSWIZZLE_W:
+			mask |= (1 << c);
+			break;
+		case TGSI_EXTSWIZZLE_ZERO:
+			zero_mask |= (1 << c);
+			tgsi.swz[c] = SWZ_X;
+			break;
+		case TGSI_EXTSWIZZLE_ONE:
+			one_mask |= (1 << c);
+			tgsi.swz[c] = SWZ_X;
+			break;
+		default:
+			assert(0);
+		}
+
+		if (!tgsi.negate && neg[c])
+			neg_mask |= (1 << c);
+	}
+
+	if (mask == MASK_ALL && !neg_mask)
+		return TRUE;
+
+	*src = temp(fpc);
+
+	if (mask)
+		arith(fpc, 0, MOV, *src, mask, tgsi, none, none);
+
+	if (zero_mask)
+		arith(fpc, 0, SFL, *src, zero_mask, *src, none, none);
+
+	if (one_mask)
+		arith(fpc, 0, STR, *src, one_mask, *src, none, none);
+
+	if (neg_mask) {
+		struct nv30_sreg one = temp(fpc);
+		arith(fpc, 0, STR, one, neg_mask, one, none, none);
+		arith(fpc, 0, MUL, *src, neg_mask, *src, neg(one), none);
+	}
+
+	return FALSE;
+}
+
+static boolean
+nv30_fragprog_parse_instruction(struct nv30_fpc *fpc,
+				const struct tgsi_full_instruction *finst)
+{
+	const struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0);
+	struct nv30_sreg src[3], dst, tmp;
+	int mask, sat, unit = 0;
+	int ai = -1, ci = -1;
+	int i;
+
+	if (finst->Instruction.Opcode == TGSI_OPCODE_END)
+		return TRUE;
+
+	fpc->temp_temp_count = 0;
+	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fsrc;
+
+		fsrc = &finst->FullSrcRegisters[i];
+		if (fsrc->SrcRegister.File == TGSI_FILE_TEMPORARY) {
+			src[i] = tgsi_src(fpc, fsrc);
+		}
+	}
+
+	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fsrc;
+
+		fsrc = &finst->FullSrcRegisters[i];
+
+		switch (fsrc->SrcRegister.File) {
+		case TGSI_FILE_INPUT:
+		case TGSI_FILE_CONSTANT:
+		case TGSI_FILE_TEMPORARY:
+			if (!src_native_swz(fpc, fsrc, &src[i]))
+				continue;
+			break;
+		default:
+			break;
+		}
+
+		switch (fsrc->SrcRegister.File) {
+		case TGSI_FILE_INPUT:
+			if (ai == -1 || ai == fsrc->SrcRegister.Index) {
+				ai = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(fpc, fsrc);
+			} else {
+				NOUVEAU_MSG("extra src attr %d\n",
+					 fsrc->SrcRegister.Index);
+				src[i] = temp(fpc);
+				arith(fpc, 0, MOV, src[i], MASK_ALL,
+				      tgsi_src(fpc, fsrc), none, none);
+			}
+			break;
+		case TGSI_FILE_CONSTANT:
+		case TGSI_FILE_IMMEDIATE:
+			if (ci == -1 || ci == fsrc->SrcRegister.Index) {
+				ci = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(fpc, fsrc);
+			} else {
+				src[i] = temp(fpc);
+				arith(fpc, 0, MOV, src[i], MASK_ALL,
+				      tgsi_src(fpc, fsrc), none, none);
+			}
+			break;
+		case TGSI_FILE_TEMPORARY:
+			/* handled above */
+			break;
+		case TGSI_FILE_SAMPLER:
+			unit = fsrc->SrcRegister.Index;
+			break;
+		case TGSI_FILE_OUTPUT:
+			break;
+		default:
+			NOUVEAU_ERR("bad src file\n");
+			return FALSE;
+		}
+	}
+
+	dst  = tgsi_dst(fpc, &finst->FullDstRegisters[0]);
+	mask = tgsi_mask(finst->FullDstRegisters[0].DstRegister.WriteMask);
+	sat  = (finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE);
+
+	switch (finst->Instruction.Opcode) {
+	case TGSI_OPCODE_ABS:
+		arith(fpc, sat, MOV, dst, mask, abs(src[0]), none, none);
+		break;
+	case TGSI_OPCODE_ADD:
+		arith(fpc, sat, ADD, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_CMP:
+		tmp = temp(fpc);
+		arith(fpc, sat, MOV, dst, mask, src[2], none, none);
+		tmp.cc_update = 1;
+		arith(fpc, 0, MOV, tmp, 0xf, src[0], none, none);
+		dst.cc_test = NV30_VP_INST_COND_LT;
+		arith(fpc, sat, MOV, dst, mask, src[1], none, none);
+		break;
+	case TGSI_OPCODE_COS:
+		arith(fpc, sat, COS, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_DP3:
+		arith(fpc, sat, DP3, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DP4:
+		arith(fpc, sat, DP4, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DPH:
+		tmp = temp(fpc);
+		arith(fpc, 0, DP3, tmp, MASK_X, src[0], src[1], none);
+		arith(fpc, sat, ADD, dst, mask, swz(tmp, X, X, X, X),
+		      swz(src[1], W, W, W, W), none);
+		break;
+	case TGSI_OPCODE_DST:
+		arith(fpc, sat, DST, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_EX2:
+		arith(fpc, sat, EX2, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_FLR:
+		arith(fpc, sat, FLR, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_FRC:
+		arith(fpc, sat, FRC, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_KILP:
+		arith(fpc, 0, KIL, none, 0, none, none, none);
+		break;
+	case TGSI_OPCODE_KIL:
+		dst = nv30_sr(NV30SR_NONE, 0);
+		dst.cc_update = 1;
+		arith(fpc, 0, MOV, dst, MASK_ALL, src[0], none, none);
+		dst.cc_update = 0; dst.cc_test = NV30_FP_OP_COND_LT;
+		arith(fpc, 0, KIL, dst, 0, none, none, none);
+		break;
+	case TGSI_OPCODE_LG2:
+		arith(fpc, sat, LG2, dst, mask, src[0], none, none);
+		break;
+//	case TGSI_OPCODE_LIT:
+	case TGSI_OPCODE_LRP:
+		arith(fpc, sat, LRP, dst, mask, src[0], src[1], src[2]);
+		break;
+	case TGSI_OPCODE_MAD:
+		arith(fpc, sat, MAD, dst, mask, src[0], src[1], src[2]);
+		break;
+	case TGSI_OPCODE_MAX:
+		arith(fpc, sat, MAX, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_MIN:
+		arith(fpc, sat, MIN, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_MOV:
+		arith(fpc, sat, MOV, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_MUL:
+		arith(fpc, sat, MUL, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_NOISE1:
+	case TGSI_OPCODE_NOISE2:
+	case TGSI_OPCODE_NOISE3:
+	case TGSI_OPCODE_NOISE4:
+		arith(fpc, sat, SFL, dst, mask, none, none, none);
+		break;
+	case TGSI_OPCODE_POW:
+		arith(fpc, sat, POW, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_RCP:
+		arith(fpc, sat, RCP, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_RET:
+		assert(0);
+		break;
+	case TGSI_OPCODE_RFL:
+		arith(fpc, 0, RFL, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_RSQ:
+		arith(fpc, sat, RSQ, dst, mask, abs(swz(src[0], X, X, X, X)), none, none);
+		break;
+	case TGSI_OPCODE_SCS:
+		if (mask & MASK_X) {
+			arith(fpc, sat, COS, dst, MASK_X,
+			      swz(src[0], X, X, X, X), none, none);
+		}
+		if (mask & MASK_Y) {
+			arith(fpc, sat, SIN, dst, MASK_Y,
+			      swz(src[0], X, X, X, X), none, none);
+		}
+		break;
+	case TGSI_OPCODE_SIN:
+		arith(fpc, sat, SIN, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_SGE:
+		arith(fpc, sat, SGE, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SGT:
+		arith(fpc, sat, SGT, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SLT:
+		arith(fpc, sat, SLT, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SUB:
+		arith(fpc, sat, ADD, dst, mask, src[0], neg(src[1]), none);
+		break;
+	case TGSI_OPCODE_TEX:
+		tex(fpc, sat, TEX, unit, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_TXB:
+		tex(fpc, sat, TXB, unit, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_TXP:
+		tex(fpc, sat, TXP, unit, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_XPD:
+		tmp = temp(fpc);
+		arith(fpc, 0, MUL, tmp, mask,
+		      swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none);
+		arith(fpc, sat, MAD, dst, (mask & ~MASK_W),
+		      swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y),
+		      neg(tmp));
+		break;
+	default:
+		NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode);
+		return FALSE;
+	}
+
+	return TRUE;
+}
+
+static boolean
+nv30_fragprog_parse_decl_attrib(struct nv30_fpc *fpc,
+				const struct tgsi_full_declaration *fdec)
+{
+	int hw;
+
+	switch (fdec->Semantic.SemanticName) {
+	case TGSI_SEMANTIC_POSITION:
+		hw = NV30_FP_OP_INPUT_SRC_POSITION;
+		break;
+	case TGSI_SEMANTIC_COLOR:
+		if (fdec->Semantic.SemanticIndex == 0) {
+			hw = NV30_FP_OP_INPUT_SRC_COL0;
+		} else
+		if (fdec->Semantic.SemanticIndex == 1) {
+			hw = NV30_FP_OP_INPUT_SRC_COL1;
+		} else {
+			NOUVEAU_ERR("bad colour semantic index\n");
+			return FALSE;
+		}
+		break;
+	case TGSI_SEMANTIC_FOG:
+		hw = NV30_FP_OP_INPUT_SRC_FOGC;
+		break;
+	case TGSI_SEMANTIC_GENERIC:
+		if (fdec->Semantic.SemanticIndex <= 7) {
+			hw = NV30_FP_OP_INPUT_SRC_TC(fdec->Semantic.
+						     SemanticIndex);
+		} else {
+			NOUVEAU_ERR("bad generic semantic index\n");
+			return FALSE;
+		}
+		break;
+	default:
+		NOUVEAU_ERR("bad input semantic\n");
+		return FALSE;
+	}
+
+	fpc->attrib_map[fdec->DeclarationRange.First] = hw;
+	return TRUE;
+}
+
+static boolean
+nv30_fragprog_parse_decl_output(struct nv30_fpc *fpc,
+				const struct tgsi_full_declaration *fdec)
+{
+	switch (fdec->Semantic.SemanticName) {
+	case TGSI_SEMANTIC_POSITION:
+		fpc->depth_id = fdec->DeclarationRange.First;
+		break;
+	case TGSI_SEMANTIC_COLOR:
+		fpc->colour_id = fdec->DeclarationRange.First;
+		break;
+	default:
+		NOUVEAU_ERR("bad output semantic\n");
+		return FALSE;
+	}
+
+	return TRUE;
+}
+
+static boolean
+nv30_fragprog_prepare(struct nv30_fpc *fpc)
+{
+	struct tgsi_parse_context p;
+	/*int high_temp = -1, i;*/
+
+	tgsi_parse_init(&p, fpc->fp->pipe.tokens);
+	while (!tgsi_parse_end_of_tokens(&p)) {
+		const union tgsi_full_token *tok = &p.FullToken;
+
+		tgsi_parse_token(&p);
+		switch(tok->Token.Type) {
+		case TGSI_TOKEN_TYPE_DECLARATION:
+		{
+			const struct tgsi_full_declaration *fdec;
+			fdec = &p.FullToken.FullDeclaration;
+			switch (fdec->Declaration.File) {
+			case TGSI_FILE_INPUT:
+				if (!nv30_fragprog_parse_decl_attrib(fpc, fdec))
+					goto out_err;
+				break;
+			case TGSI_FILE_OUTPUT:
+				if (!nv30_fragprog_parse_decl_output(fpc, fdec))
+					goto out_err;
+				break;
+			/*case TGSI_FILE_TEMPORARY:
+				if (fdec->DeclarationRange.Last > high_temp) {
+					high_temp =
+						fdec->DeclarationRange.Last;
+				}
+				break;*/
+			default:
+				break;
+			}
+		}
+			break;
+		case TGSI_TOKEN_TYPE_IMMEDIATE:
+		{
+			struct tgsi_full_immediate *imm;
+			float vals[4];
+			
+			imm = &p.FullToken.FullImmediate;
+			assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32);
+			assert(fpc->nr_imm < MAX_IMM);
+
+			vals[0] = imm->u.ImmediateFloat32[0].Float;
+			vals[1] = imm->u.ImmediateFloat32[1].Float;
+			vals[2] = imm->u.ImmediateFloat32[2].Float;
+			vals[3] = imm->u.ImmediateFloat32[3].Float;
+			fpc->imm[fpc->nr_imm++] = constant(fpc, -1, vals);
+		}
+			break;
+		default:
+			break;
+		}
+	}
+	tgsi_parse_free(&p);
+
+	/*if (++high_temp) {
+		fpc->r_temp = CALLOC(high_temp, sizeof(struct nv30_sreg));
+		for (i = 0; i < high_temp; i++)
+			fpc->r_temp[i] = temp(fpc);
+		fpc->r_temps_discard = 0;
+	}*/
+
+	return TRUE;
+
+out_err:
+	/*if (fpc->r_temp)
+		FREE(fpc->r_temp);*/
+	tgsi_parse_free(&p);
+	return FALSE;
+}
+
+static void
+nv30_fragprog_translate(struct nv30_context *nv30,
+			struct nv30_fragment_program *fp)
+{
+	struct tgsi_parse_context parse;
+	struct nv30_fpc *fpc = NULL;
+
+	tgsi_dump(fp->pipe.tokens,0);
+
+	fpc = CALLOC(1, sizeof(struct nv30_fpc));
+	if (!fpc)
+		return;
+	fpc->fp = fp;
+	fpc->high_temp = -1;
+	fpc->num_regs = 2;
+
+	if (!nv30_fragprog_prepare(fpc)) {
+		FREE(fpc);
+		return;
+	}
+
+	tgsi_parse_init(&parse, fp->pipe.tokens);
+
+	while (!tgsi_parse_end_of_tokens(&parse)) {
+		tgsi_parse_token(&parse);
+
+		switch (parse.FullToken.Token.Type) {
+		case TGSI_TOKEN_TYPE_INSTRUCTION:
+		{
+			const struct tgsi_full_instruction *finst;
+
+			finst = &parse.FullToken.FullInstruction;
+			if (!nv30_fragprog_parse_instruction(fpc, finst))
+				goto out_err;
+		}
+			break;
+		default:
+			break;
+		}
+	}
+
+	fp->fp_control |= (fpc->num_regs-1)/2;
+	fp->fp_reg_control = (1<<16)|0x4;
+
+	/* Terminate final instruction */
+	fp->insn[fpc->inst_offset] |= 0x00000001;
+
+	/* Append NOP + END instruction, may or may not be necessary. */
+	fpc->inst_offset = fp->insn_len;
+	grow_insns(fpc, 4);
+	fp->insn[fpc->inst_offset + 0] = 0x00000001;
+	fp->insn[fpc->inst_offset + 1] = 0x00000000;
+	fp->insn[fpc->inst_offset + 2] = 0x00000000;
+	fp->insn[fpc->inst_offset + 3] = 0x00000000;
+	
+	fp->translated = TRUE;
+	fp->on_hw = FALSE;
+out_err:
+	tgsi_parse_free(&parse);
+	FREE(fpc);
+}
+
+static void
+nv30_fragprog_upload(struct nv30_context *nv30,
+		     struct nv30_fragment_program *fp)
+{
+	struct pipe_winsys *ws = nv30->pipe.winsys;
+	const uint32_t le = 1;
+	uint32_t *map;
+	int i;
+
+	map = ws->buffer_map(ws, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+#if 0
+	for (i = 0; i < fp->insn_len; i++) {
+		fflush(stdout); fflush(stderr);
+		NOUVEAU_ERR("%d 0x%08x\n", i, fp->insn[i]);
+		fflush(stdout); fflush(stderr);
+	}
+#endif
+
+	if ((*(const uint8_t *)&le)) {
+		for (i = 0; i < fp->insn_len; i++) {
+			map[i] = fp->insn[i];
+		}
+	} else {
+		/* Weird swapping for big-endian chips */
+		for (i = 0; i < fp->insn_len; i++) {
+			map[i] = ((fp->insn[i] & 0xffff) << 16) |
+				  ((fp->insn[i] >> 16) & 0xffff);
+		}
+	}
+
+	ws->buffer_unmap(ws, fp->buffer);
+}
+
+static boolean
+nv30_fragprog_validate(struct nv30_context *nv30)
+{
+	struct nv30_fragment_program *fp = nv30->fragprog;
+	struct pipe_buffer *constbuf =
+		nv30->constbuf[PIPE_SHADER_FRAGMENT];
+	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct nouveau_stateobj *so;
+	boolean new_consts = FALSE;
+	int i;
+
+	if (fp->translated)
+		goto update_constants;
+
+	/*nv30->fallback_swrast &= ~NV30_NEW_FRAGPROG;*/
+	nv30_fragprog_translate(nv30, fp);
+	if (!fp->translated) {
+		/*nv30->fallback_swrast |= NV30_NEW_FRAGPROG;*/
+		return FALSE;
+	}
+
+	fp->buffer = ws->buffer_create(ws, 0x100, 0, fp->insn_len * 4);
+	nv30_fragprog_upload(nv30, fp);
+
+	so = so_new(8, 1);
+	so_method(so, nv30->screen->rankine, NV34TCL_FP_ACTIVE_PROGRAM, 1);
+	so_reloc (so, fp->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
+		  NV34TCL_FP_ACTIVE_PROGRAM_DMA0, NV34TCL_FP_ACTIVE_PROGRAM_DMA1);
+	so_method(so, nv30->screen->rankine, NV34TCL_FP_CONTROL, 1);
+	so_data  (so, fp->fp_control);
+	so_method(so, nv30->screen->rankine, NV34TCL_FP_REG_CONTROL, 1);
+	so_data  (so, fp->fp_reg_control);
+	so_method(so, nv30->screen->rankine, NV34TCL_TX_UNITS_ENABLE, 1);
+	so_data  (so, fp->samplers);
+	so_ref(so, &fp->so);
+
+update_constants:
+	if (fp->nr_consts) {
+		float *map;
+		
+		map = ws->buffer_map(ws, constbuf, PIPE_BUFFER_USAGE_CPU_READ);
+		for (i = 0; i < fp->nr_consts; i++) {
+			struct nv30_fragment_program_data *fpd = &fp->consts[i];
+			uint32_t *p = &fp->insn[fpd->offset];
+			uint32_t *cb = (uint32_t *)&map[fpd->index * 4];
+
+			if (!memcmp(p, cb, 4 * sizeof(float)))
+				continue;
+			memcpy(p, cb, 4 * sizeof(float));
+			new_consts = TRUE;
+		}
+		ws->buffer_unmap(ws, constbuf);
+
+		if (new_consts)
+			nv30_fragprog_upload(nv30, fp);
+	}
+
+	if (new_consts || fp->so != nv30->state.hw[NV30_STATE_FRAGPROG]) {
+		so_ref(fp->so, &nv30->state.hw[NV30_STATE_FRAGPROG]);
+		return TRUE;
+	}
+
+	return FALSE;
+}
+
+void
+nv30_fragprog_destroy(struct nv30_context *nv30,
+		      struct nv30_fragment_program *fp)
+{
+	if (fp->insn_len)
+		FREE(fp->insn);
+}
+
+struct nv30_state_entry nv30_state_fragprog = {
+	.validate = nv30_fragprog_validate,
+	.dirty = {
+		.pipe = NV30_NEW_FRAGPROG,
+		.hw = NV30_STATE_FRAGPROG
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_fragtex.c b/src/gallium/drivers/nv30/nv30_fragtex.c
new file mode 100644
index 0000000000..b1d2663af3
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_fragtex.c
@@ -0,0 +1,163 @@
+#include "nv30_context.h"
+#include "nouveau/nouveau_util.h"
+
+#define _(m,tf,ts0x,ts0y,ts0z,ts0w,ts1x,ts1y,ts1z,ts1w)                        \
+{                                                                              \
+  TRUE,                                                                        \
+  PIPE_FORMAT_##m,                                                             \
+  NV34TCL_TX_FORMAT_FORMAT_##tf,                                               \
+  (NV34TCL_TX_SWIZZLE_S0_X_##ts0x | NV34TCL_TX_SWIZZLE_S0_Y_##ts0y |           \
+   NV34TCL_TX_SWIZZLE_S0_Z_##ts0z | NV34TCL_TX_SWIZZLE_S0_W_##ts0w |           \
+   NV34TCL_TX_SWIZZLE_S1_X_##ts1x | NV34TCL_TX_SWIZZLE_S1_Y_##ts1y |           \
+   NV34TCL_TX_SWIZZLE_S1_Z_##ts1z | NV34TCL_TX_SWIZZLE_S1_W_##ts1w)            \
+}
+
+struct nv30_texture_format {
+	boolean defined;
+	uint	pipe;
+	int     format;
+	int     swizzle;
+};
+
+static struct nv30_texture_format
+nv30_texture_formats[] = {
+	_(A8R8G8B8_UNORM, A8R8G8B8,   S1,   S1,   S1,   S1, X, Y, Z, W),
+	_(A1R5G5B5_UNORM, A1R5G5B5,   S1,   S1,   S1,   S1, X, Y, Z, W),
+	_(A4R4G4B4_UNORM, A4R4G4B4,   S1,   S1,   S1,   S1, X, Y, Z, W),
+	_(R5G6B5_UNORM  , R5G6B5  ,   S1,   S1,   S1,  ONE, X, Y, Z, W),
+	_(L8_UNORM      , L8      ,   S1,   S1,   S1,  ONE, X, X, X, X),
+	_(A8_UNORM      , L8      , ZERO, ZERO, ZERO,   S1, X, X, X, X),
+	_(I8_UNORM      , L8      ,   S1,   S1,   S1,   S1, X, X, X, X),
+	_(A8L8_UNORM    , A8L8    ,   S1,   S1,   S1,   S1, X, X, X, Y),
+//	_(Z16_UNORM     , Z16     ,   S1,   S1,   S1,  ONE, X, X, X, X),
+//	_(Z24S8_UNORM   , Z24     ,   S1,   S1,   S1,  ONE, X, X, X, X),
+	_(DXT1_RGB      , DXT1    ,   S1,   S1,   S1,  ONE, X, Y, Z, W),
+	_(DXT1_RGBA     , DXT1    ,   S1,   S1,   S1,   S1, X, Y, Z, W),
+	_(DXT3_RGBA     , DXT3    ,   S1,   S1,   S1,   S1, X, Y, Z, W),
+	_(DXT5_RGBA     , DXT5    ,   S1,   S1,   S1,   S1, X, Y, Z, W),
+	{},
+};
+
+static struct nv30_texture_format *
+nv30_fragtex_format(uint pipe_format)
+{
+	struct nv30_texture_format *tf = nv30_texture_formats;
+	char fs[128];
+
+	while (tf->defined) {
+		if (tf->pipe == pipe_format)
+			return tf;
+		tf++;
+	}
+
+	NOUVEAU_ERR("unknown texture format %s\n", pf_name(pipe_format));
+	return NULL;
+}
+
+
+static struct nouveau_stateobj *
+nv30_fragtex_build(struct nv30_context *nv30, int unit)
+{
+	struct nv30_sampler_state *ps = nv30->tex_sampler[unit];
+	struct nv30_miptree *nv30mt = nv30->tex_miptree[unit];
+	struct pipe_texture *pt = &nv30mt->base;
+	struct nv30_texture_format *tf;
+	struct nouveau_stateobj *so;
+	uint32_t txf, txs , txp;
+	unsigned tex_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD;
+
+	tf = nv30_fragtex_format(pt->format);
+	if (!tf)
+		assert(0);
+
+	txf  = tf->format;
+	txf |= ((pt->last_level>0) ? NV34TCL_TX_FORMAT_MIPMAP : 0);
+	txf |= log2i(pt->width[0]) << 20;
+	txf |= log2i(pt->height[0]) << 24;
+	txf |= log2i(pt->depth[0]) << 28;
+	txf |= NV34TCL_TX_FORMAT_NO_BORDER | 0x10000;
+
+	switch (pt->target) {
+	case PIPE_TEXTURE_CUBE:
+		txf |= NV34TCL_TX_FORMAT_CUBIC;
+		/* fall-through */
+	case PIPE_TEXTURE_2D:
+		txf |= NV34TCL_TX_FORMAT_DIMS_2D;
+		break;
+	case PIPE_TEXTURE_3D:
+		txf |= NV34TCL_TX_FORMAT_DIMS_3D;
+		break;
+	case PIPE_TEXTURE_1D:
+		txf |= NV34TCL_TX_FORMAT_DIMS_1D;
+		break;
+	default:
+		NOUVEAU_ERR("Unknown target %d\n", pt->target);
+		return NULL;
+	}
+
+	if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
+		txp = 0;
+	} else {
+		txp  = nv30mt->level[0].pitch;
+		txf |= (1<<13) /*FIXME: NV34TCL_TX_FORMAT_LINEAR ? */;
+	}
+
+	txs = tf->swizzle;
+
+	so = so_new(16, 2);
+	so_method(so, nv30->screen->rankine, NV34TCL_TX_OFFSET(unit), 8);
+	so_reloc (so, nv30mt->buffer, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
+	so_reloc (so, nv30mt->buffer, txf, tex_flags | NOUVEAU_BO_OR,
+		  NV34TCL_TX_FORMAT_DMA0, NV34TCL_TX_FORMAT_DMA1);
+	so_data  (so, ps->wrap);
+	so_data  (so, NV34TCL_TX_ENABLE_ENABLE | ps->en);
+	so_data  (so, txs);
+	so_data  (so, ps->filt | 0x2000 /*voodoo*/);
+	so_data  (so, (pt->width[0] << NV34TCL_TX_NPOT_SIZE_W_SHIFT) |
+		       pt->height[0]);
+	so_data  (so, ps->bcol);
+
+	return so;
+}
+
+static boolean
+nv30_fragtex_validate(struct nv30_context *nv30)
+{
+	struct nv30_fragment_program *fp = nv30->fragprog;
+	struct nv30_state *state = &nv30->state;
+	struct nouveau_stateobj *so;
+	unsigned samplers, unit;
+
+	samplers = state->fp_samplers & ~fp->samplers;
+	while (samplers) {
+		unit = ffs(samplers) - 1;
+		samplers &= ~(1 << unit);
+
+		so = so_new(2, 0);
+		so_method(so, nv30->screen->rankine, NV34TCL_TX_ENABLE(unit), 1);
+		so_data  (so, 0);
+		so_ref(so, &nv30->state.hw[NV30_STATE_FRAGTEX0 + unit]);
+		state->dirty |= (1ULL << (NV30_STATE_FRAGTEX0 + unit));
+	}
+
+	samplers = nv30->dirty_samplers & fp->samplers;
+	while (samplers) {
+		unit = ffs(samplers) - 1;
+		samplers &= ~(1 << unit);
+
+		so = nv30_fragtex_build(nv30, unit);
+		so_ref(so, &nv30->state.hw[NV30_STATE_FRAGTEX0 + unit]);
+		state->dirty |= (1ULL << (NV30_STATE_FRAGTEX0 + unit));
+	}
+
+	nv30->state.fp_samplers = fp->samplers;
+	return FALSE;
+}
+
+struct nv30_state_entry nv30_state_fragtex = {
+	.validate = nv30_fragtex_validate,
+	.dirty = {
+		.pipe = NV30_NEW_SAMPLER | NV30_NEW_FRAGPROG,
+		.hw = 0
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_miptree.c b/src/gallium/drivers/nv30/nv30_miptree.c
new file mode 100644
index 0000000000..510c94d4e6
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_miptree.c
@@ -0,0 +1,235 @@
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "nv30_context.h"
+
+static void
+nv30_miptree_layout(struct nv30_miptree *nv30mt)
+{
+	struct pipe_texture *pt = &nv30mt->base;
+	uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0];
+	uint offset = 0;
+	int nr_faces, l, f;
+	uint wide_pitch = pt->tex_usage & (PIPE_TEXTURE_USAGE_SAMPLER |
+		                           PIPE_TEXTURE_USAGE_DEPTH_STENCIL |
+		                           PIPE_TEXTURE_USAGE_RENDER_TARGET |
+		                           PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+		                           PIPE_TEXTURE_USAGE_PRIMARY);
+
+	if (pt->target == PIPE_TEXTURE_CUBE) {
+		nr_faces = 6;
+	} else
+	if (pt->target == PIPE_TEXTURE_3D) {
+		nr_faces = pt->depth[0];
+	} else {
+		nr_faces = 1;
+	}
+
+	for (l = 0; l <= pt->last_level; l++) {
+		pt->width[l] = width;
+		pt->height[l] = height;
+		pt->depth[l] = depth;
+		pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width);
+		pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height);
+
+		if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR))
+			nv30mt->level[l].pitch = align(pt->width[0] * pt->block.size, 64);
+		else
+			nv30mt->level[l].pitch = pt->width[l] * pt->block.size;
+
+		nv30mt->level[l].image_offset =
+			CALLOC(nr_faces, sizeof(unsigned));
+
+		width  = MAX2(1, width  >> 1);
+		height = MAX2(1, height >> 1);
+		depth  = MAX2(1, depth  >> 1);
+	}
+
+	for (f = 0; f < nr_faces; f++) {
+		for (l = 0; l < pt->last_level; l++) {
+			nv30mt->level[l].image_offset[f] = offset;
+
+			if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) &&
+			    pt->width[l + 1] > 1 && pt->height[l + 1] > 1)
+				offset += align(nv30mt->level[l].pitch * pt->height[l], 64);
+			else
+				offset += nv30mt->level[l].pitch * pt->height[l];
+		}
+
+		nv30mt->level[l].image_offset[f] = offset;
+		offset += nv30mt->level[l].pitch * pt->height[l];
+	}
+
+	nv30mt->total_size = offset;
+}
+
+static struct pipe_texture *
+nv30_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
+{
+	struct pipe_winsys *ws = pscreen->winsys;
+	struct nv30_miptree *mt;
+
+	mt = MALLOC(sizeof(struct nv30_miptree));
+	if (!mt)
+		return NULL;
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = pscreen;
+	mt->shadow_tex = NULL;
+	mt->shadow_surface = NULL;
+
+	/* Swizzled textures must be POT */
+	if (pt->width[0] & (pt->width[0] - 1) ||
+	    pt->height[0] & (pt->height[0] - 1))
+		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+	else
+	if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY |
++	                     PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
++	                     PIPE_TEXTURE_USAGE_DEPTH_STENCIL))
+		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+	else
+	if (pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC)
+		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+	else {
+		switch (pt->format) {
+		/* TODO: Figure out which formats can be swizzled */
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_X8R8G8B8_UNORM:
+		case PIPE_FORMAT_R16_SNORM:
+		{
+			if (debug_get_bool_option("NOUVEAU_NO_SWIZZLE", FALSE))
+				mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+ 			break;
+		}
+		default:
+			mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+		}
+	}
+
+	nv30_miptree_layout(mt);
+
+	mt->buffer = ws->buffer_create(ws, 256,
+				       PIPE_BUFFER_USAGE_PIXEL |
+				       NOUVEAU_BUFFER_USAGE_TEXTURE,
+				       mt->total_size);
+	if (!mt->buffer) {
+		FREE(mt);
+		return NULL;
+	}
+
+	return &mt->base;
+}
+
+static struct pipe_texture *
+nv30_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
+		     const unsigned *stride, struct pipe_buffer *pb)
+{
+	struct nv30_miptree *mt;
+
+	/* Only supports 2D, non-mipmapped textures for the moment */
+	if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 ||
+	    pt->depth[0] != 1)
+		return NULL;
+
+	mt = CALLOC_STRUCT(nv30_miptree);
+	if (!mt)
+		return NULL;
+
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = pscreen;
+	mt->level[0].pitch = stride[0];
+	mt->level[0].image_offset = CALLOC(1, sizeof(unsigned));
+
+	pipe_buffer_reference(pscreen, &mt->buffer, pb);
+	return &mt->base;
+}
+
+static void
+nv30_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt)
+{
+	struct pipe_texture *pt = *ppt;
+	struct nv30_miptree *mt = (struct nv30_miptree *)pt;
+	int l;
+
+	*ppt = NULL;
+	if (--pt->refcount)
+		return;
+
+	pipe_buffer_reference(pscreen, &mt->buffer, NULL);
+	for (l = 0; l <= pt->last_level; l++) {
+		if (mt->level[l].image_offset)
+			FREE(mt->level[l].image_offset);
+	}
+
+	if (mt->shadow_tex) {
+		if (mt->shadow_surface)
+			pscreen->tex_surface_release(pscreen, &mt->shadow_surface);
+		nv30_miptree_release(pscreen, &mt->shadow_tex);
+	}
+
+	FREE(mt);
+}
+
+static struct pipe_surface *
+nv30_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
+			 unsigned face, unsigned level, unsigned zslice,
+			 unsigned flags)
+{
+	struct nv30_miptree *nv30mt = (struct nv30_miptree *)pt;
+	struct pipe_surface *ps;
+
+	ps = CALLOC_STRUCT(pipe_surface);
+	if (!ps)
+		return NULL;
+	pipe_texture_reference(&ps->texture, pt);
+	ps->format = pt->format;
+	ps->width = pt->width[level];
+	ps->height = pt->height[level];
+	ps->block = pt->block;
+	ps->nblocksx = pt->nblocksx[level];
+	ps->nblocksy = pt->nblocksy[level];
+	ps->stride = nv30mt->level[level].pitch;
+	ps->usage = flags;
+	ps->status = PIPE_SURFACE_STATUS_DEFINED;
+	ps->refcount = 1;
+	ps->face = face;
+	ps->level = level;
+	ps->zslice = zslice;
+
+	if (pt->target == PIPE_TEXTURE_CUBE) {
+		ps->offset = nv30mt->level[level].image_offset[face];
+	} else
+	if (pt->target == PIPE_TEXTURE_3D) {
+		ps->offset = nv30mt->level[level].image_offset[zslice];
+	} else {
+		ps->offset = nv30mt->level[level].image_offset[0];
+	}
+
+	return ps;
+}
+
+static void
+nv30_miptree_surface_del(struct pipe_screen *pscreen,
+			 struct pipe_surface **psurface)
+{
+	struct pipe_surface *ps = *psurface;
+
+	*psurface = NULL;
+	if (--ps->refcount > 0)
+		return;
+
+	pipe_texture_reference(&ps->texture, NULL);
+	FREE(ps);
+}
+
+void
+nv30_screen_init_miptree_functions(struct pipe_screen *pscreen)
+{
+	pscreen->texture_create = nv30_miptree_create;
+	pscreen->texture_blanket = nv30_miptree_blanket;
+	pscreen->texture_release = nv30_miptree_release;
+	pscreen->get_tex_surface = nv30_miptree_surface_new;
+	pscreen->tex_surface_release = nv30_miptree_surface_del;
+}
diff --git a/src/gallium/drivers/nv30/nv30_query.c b/src/gallium/drivers/nv30/nv30_query.c
new file mode 100644
index 0000000000..2f974cf5c4
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_query.c
@@ -0,0 +1,122 @@
+#include "pipe/p_context.h"
+
+#include "nv30_context.h"
+
+struct nv30_query {
+	struct nouveau_resource *object;
+	unsigned type;
+	boolean ready;
+	uint64_t result;
+};
+
+static INLINE struct nv30_query *
+nv30_query(struct pipe_query *pipe)
+{
+	return (struct nv30_query *)pipe;
+}
+
+static struct pipe_query *
+nv30_query_create(struct pipe_context *pipe, unsigned query_type)
+{
+	struct nv30_query *q;
+
+	q = CALLOC(1, sizeof(struct nv30_query));
+	q->type = query_type;
+
+	return (struct pipe_query *)q;
+}
+
+static void
+nv30_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nv30_query *q = nv30_query(pq);
+
+	if (q->object)
+		nv30->nvws->res_free(&q->object);
+	FREE(q);
+}
+
+static void
+nv30_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nv30_query *q = nv30_query(pq);
+
+	assert(q->type == PIPE_QUERY_OCCLUSION_COUNTER);
+
+	/* Happens when end_query() is called, then another begin_query()
+	 * without querying the result in-between.  For now we'll wait for
+	 * the existing query to notify completion, but it could be better.
+	 */
+	if (q->object) {
+		uint64_t tmp;
+		pipe->get_query_result(pipe, pq, 1, &tmp);
+	}
+
+	if (nv30->nvws->res_alloc(nv30->screen->query_heap, 1, NULL, &q->object))
+		assert(0);
+	nv30->nvws->notifier_reset(nv30->screen->query, q->object->start);
+
+	BEGIN_RING(rankine, NV34TCL_QUERY_RESET, 1);
+	OUT_RING  (1);
+	BEGIN_RING(rankine, NV34TCL_QUERY_UNK17CC, 1);
+	OUT_RING  (1);
+
+	q->ready = FALSE;
+}
+
+static void
+nv30_query_end(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nv30_query *q = nv30_query(pq);
+
+	BEGIN_RING(rankine, NV34TCL_QUERY_GET, 1);
+	OUT_RING  ((0x01 << NV34TCL_QUERY_GET_UNK24_SHIFT) |
+		   ((q->object->start * 32) << NV34TCL_QUERY_GET_OFFSET_SHIFT));
+	FIRE_RING(NULL);
+}
+
+static boolean
+nv30_query_result(struct pipe_context *pipe, struct pipe_query *pq,
+		  boolean wait, uint64_t *result)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nv30_query *q = nv30_query(pq);
+	struct nouveau_winsys *nvws = nv30->nvws;
+
+	assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER);
+
+	if (!q->ready) {
+		unsigned status;
+
+		status = nvws->notifier_status(nv30->screen->query,
+					       q->object->start);
+		if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) {
+			if (wait == FALSE)
+				return FALSE;
+			nvws->notifier_wait(nv30->screen->query, q->object->start,
+					    NV_NOTIFY_STATE_STATUS_COMPLETED,
+					    0);
+		}
+
+		q->result = nvws->notifier_retval(nv30->screen->query,
+						  q->object->start);
+		q->ready = TRUE;
+		nvws->res_free(&q->object);
+	}
+
+	*result = q->result;
+	return TRUE;
+}
+
+void
+nv30_init_query_functions(struct nv30_context *nv30)
+{
+	nv30->pipe.create_query = nv30_query_create;
+	nv30->pipe.destroy_query = nv30_query_destroy;
+	nv30->pipe.begin_query = nv30_query_begin;
+	nv30->pipe.end_query = nv30_query_end;
+	nv30->pipe.get_query_result = nv30_query_result;
+}
diff --git a/src/gallium/drivers/nv30/nv30_screen.c b/src/gallium/drivers/nv30/nv30_screen.c
new file mode 100644
index 0000000000..c97a73f0b1
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_screen.c
@@ -0,0 +1,401 @@
+#include "pipe/p_screen.h"
+#include "util/u_simple_screen.h"
+
+#include "nv30_context.h"
+#include "nv30_screen.h"
+
+#define NV30TCL_CHIPSET_3X_MASK 0x00000003
+#define NV34TCL_CHIPSET_3X_MASK 0x00000010
+#define NV35TCL_CHIPSET_3X_MASK 0x000001e0
+
+static const char *
+nv30_screen_get_name(struct pipe_screen *pscreen)
+{
+	struct nv30_screen *screen = nv30_screen(pscreen);
+	struct nouveau_device *dev = screen->nvws->channel->device;
+	static char buffer[128];
+
+	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
+	return buffer;
+}
+
+static const char *
+nv30_screen_get_vendor(struct pipe_screen *pscreen)
+{
+	return "nouveau";
+}
+
+static int
+nv30_screen_get_param(struct pipe_screen *pscreen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+		return 16;
+	case PIPE_CAP_NPOT_TEXTURES:
+		return 0;
+	case PIPE_CAP_TWO_SIDED_STENCIL:
+		return 1;
+	case PIPE_CAP_GLSL:
+		return 0;
+	case PIPE_CAP_S3TC:
+		return 0;
+	case PIPE_CAP_ANISOTROPIC_FILTER:
+		return 1;
+	case PIPE_CAP_POINT_SPRITE:
+		return 1;
+	case PIPE_CAP_MAX_RENDER_TARGETS:
+		return 2;
+	case PIPE_CAP_OCCLUSION_QUERY:
+		return 1;
+	case PIPE_CAP_TEXTURE_SHADOW_MAP:
+		return 1;
+	case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+		return 13;
+	case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+		return 10;
+	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+		return 13;
+	case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
+		return 0;
+	case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
+		return 1;
+	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
+		return 0;
+	case NOUVEAU_CAP_HW_VTXBUF:
+	case NOUVEAU_CAP_HW_IDXBUF:
+		return 1;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0;
+	}
+}
+
+static float
+nv30_screen_get_paramf(struct pipe_screen *pscreen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_LINE_WIDTH:
+	case PIPE_CAP_MAX_LINE_WIDTH_AA:
+		return 10.0;
+	case PIPE_CAP_MAX_POINT_WIDTH:
+	case PIPE_CAP_MAX_POINT_WIDTH_AA:
+		return 64.0;
+	case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+		return 8.0;
+	case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+		return 4.0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0.0;
+	}
+}
+
+static boolean
+nv30_screen_surface_format_supported(struct pipe_screen *pscreen,
+				     enum pipe_format format,
+				     enum pipe_texture_target target,
+				     unsigned tex_usage, unsigned geom_flags)
+{
+	if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM:
+		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z16_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	} else {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_A1R5G5B5_UNORM:
+		case PIPE_FORMAT_A4R4G4B4_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM:
+		case PIPE_FORMAT_L8_UNORM:
+		case PIPE_FORMAT_A8_UNORM:
+		case PIPE_FORMAT_I8_UNORM:
+		case PIPE_FORMAT_A8L8_UNORM:
+		case PIPE_FORMAT_Z16_UNORM:
+		case PIPE_FORMAT_Z24S8_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	}
+
+	return FALSE;
+}
+
+static struct pipe_buffer *
+nv30_surface_buffer(struct pipe_surface *surf)
+{
+	struct nv30_miptree *mt = (struct nv30_miptree *)surf->texture;
+
+	return mt->buffer;
+}
+
+static void *
+nv30_surface_map(struct pipe_screen *screen, struct pipe_surface *surface,
+		 unsigned flags )
+{
+	struct pipe_winsys	*ws = screen->winsys;
+	struct pipe_surface	*surface_to_map;
+	void			*map;
+
+	if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
+		struct nv30_miptree *mt = (struct nv30_miptree *)surface->texture;
+
+		if (!mt->shadow_tex) {
+			unsigned old_tex_usage = surface->texture->tex_usage;
+			surface->texture->tex_usage = NOUVEAU_TEXTURE_USAGE_LINEAR |
+			                              PIPE_TEXTURE_USAGE_DYNAMIC;
+			mt->shadow_tex = screen->texture_create(screen, surface->texture);
+			surface->texture->tex_usage = old_tex_usage;
+
+			assert(mt->shadow_tex->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR);
+		}
+
+		mt->shadow_surface = screen->get_tex_surface
+		(
+			screen, mt->shadow_tex,
+			surface->face, surface->level, surface->zslice,
+			surface->usage
+		);
+
+		surface_to_map = mt->shadow_surface;
+	}
+	else
+		surface_to_map = surface;
+
+	assert(surface_to_map);
+
+	map = ws->buffer_map(ws, nv30_surface_buffer(surface_to_map), flags);
+	if (!map)
+		return NULL;
+
+	return map + surface_to_map->offset;
+}
+
+static void
+nv30_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface)
+{
+	struct pipe_winsys	*ws = screen->winsys;
+	struct pipe_surface	*surface_to_unmap;
+
+	/* TODO: Copy from shadow just before push buffer is flushed instead.
+	         There are probably some programs that map/unmap excessively
+	         before rendering. */
+	if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
+		struct nv30_miptree *mt = (struct nv30_miptree *)surface->texture;
+
+		assert(mt->shadow_tex);
+
+		surface_to_unmap = mt->shadow_surface;
+	}
+	else
+		surface_to_unmap = surface;
+
+	assert(surface_to_unmap);
+
+	ws->buffer_unmap(ws, nv30_surface_buffer(surface_to_unmap));
+
+	if (surface_to_unmap != surface) {
+		struct nv30_screen *nvscreen = nv30_screen(screen);
+
+		nvscreen->eng2d->copy(nvscreen->eng2d, surface, 0, 0,
+		                      surface_to_unmap, 0, 0,
+		                      surface->width, surface->height);
+
+		screen->tex_surface_release(screen, &surface_to_unmap);
+	}
+}
+
+static void
+nv30_screen_destroy(struct pipe_screen *pscreen)
+{
+	struct nv30_screen *screen = nv30_screen(pscreen);
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nvws->res_free(&screen->vp_exec_heap);
+	nvws->res_free(&screen->vp_data_heap);
+	nvws->res_free(&screen->query_heap);
+	nvws->notifier_free(&screen->query);
+	nvws->notifier_free(&screen->sync);
+	nvws->grobj_free(&screen->rankine);
+
+	FREE(pscreen);
+}
+
+struct pipe_screen *
+nv30_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+{
+	struct nv30_screen *screen = CALLOC_STRUCT(nv30_screen);
+	struct nouveau_stateobj *so;
+	unsigned rankine_class = 0;
+	unsigned chipset = nvws->channel->device->chipset;
+	int ret, i;
+
+	if (!screen)
+		return NULL;
+	screen->nvws = nvws;
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(nvws);
+	screen->eng2d->buf = nv30_surface_buffer;
+
+	/* 3D object */
+	switch (chipset & 0xf0) {
+	case 0x30:
+		if (NV30TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f)))
+			rankine_class = 0x0397;
+		else
+		if (NV34TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f)))
+			rankine_class = 0x0697;
+		else
+		if (NV35TCL_CHIPSET_3X_MASK & (1 << (chipset & 0x0f)))
+			rankine_class = 0x0497;
+		break;
+	default:
+		break;
+	}
+
+	if (!rankine_class) {
+		NOUVEAU_ERR("Unknown nv3x chipset: nv%02x\n", chipset);
+		return NULL;
+	}
+
+	ret = nvws->grobj_alloc(nvws, rankine_class, &screen->rankine);
+	if (ret) {
+		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
+		return FALSE;
+	}
+
+	/* Notifier for sync purposes */
+	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	if (ret) {
+		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
+		nv30_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	/* Query objects */
+	ret = nvws->notifier_alloc(nvws, 32, &screen->query);
+	if (ret) {
+		NOUVEAU_ERR("Error initialising query objects: %d\n", ret);
+		nv30_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	ret = nvws->res_init(&screen->query_heap, 0, 32);
+	if (ret) {
+		NOUVEAU_ERR("Error initialising query object heap: %d\n", ret);
+		nv30_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	/* Vtxprog resources */
+	if (nvws->res_init(&screen->vp_exec_heap, 0, 256) ||
+	    nvws->res_init(&screen->vp_data_heap, 0, 256)) {
+		nv30_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	/* Static rankine initialisation */
+	so = so_new(128, 0);
+	so_method(so, screen->rankine, NV34TCL_DMA_NOTIFY, 1);
+	so_data  (so, screen->sync->handle);
+	so_method(so, screen->rankine, NV34TCL_DMA_TEXTURE0, 2);
+	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, nvws->channel->gart->handle);
+	so_method(so, screen->rankine, NV34TCL_DMA_COLOR1, 1);
+	so_data  (so, nvws->channel->vram->handle);
+	so_method(so, screen->rankine, NV34TCL_DMA_COLOR0, 2);
+	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, nvws->channel->vram->handle);
+	so_method(so, screen->rankine, NV34TCL_DMA_VTXBUF0, 2);
+	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, nvws->channel->gart->handle);
+/*	so_method(so, screen->rankine, NV34TCL_DMA_FENCE, 2);
+	so_data  (so, 0);
+	so_data  (so, screen->query->handle);*/
+	so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY7, 1);
+	so_data  (so, nvws->channel->vram->handle);
+	so_method(so, screen->rankine, NV34TCL_DMA_IN_MEMORY8, 1);
+	so_data  (so, nvws->channel->vram->handle);
+
+	for (i=1; i<8; i++) {
+		so_method(so, screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(i), 1);
+		so_data  (so, 0);
+		so_method(so, screen->rankine, NV34TCL_VIEWPORT_CLIP_VERT(i), 1);
+		so_data  (so, 0);
+	}
+
+	so_method(so, screen->rankine, 0x220, 1);
+	so_data  (so, 1);
+
+	so_method(so, screen->rankine, 0x03b0, 1);
+	so_data  (so, 0x00100000);
+	so_method(so, screen->rankine, 0x1454, 1);
+	so_data  (so, 0);
+	so_method(so, screen->rankine, 0x1d80, 1);
+	so_data  (so, 3);
+	so_method(so, screen->rankine, 0x1450, 1);
+	so_data  (so, 0x00030004);
+
+	/* NEW */
+	so_method(so, screen->rankine, 0x1e98, 1);
+	so_data  (so, 0);
+	so_method(so, screen->rankine, 0x17e0, 3);
+	so_data  (so, fui(0.0));
+	so_data  (so, fui(0.0));
+	so_data  (so, fui(1.0));
+	so_method(so, screen->rankine, 0x1f80, 16);
+	for (i=0; i<16; i++) {
+		so_data  (so, (i==8) ? 0x0000ffff : 0);
+	}
+
+	so_method(so, screen->rankine, 0x120, 3);
+	so_data  (so, 0);
+	so_data  (so, 1);
+	so_data  (so, 2);
+
+	so_method(so, screen->rankine, 0x1d88, 1);
+	so_data  (so, 0x00001200);
+
+	so_method(so, screen->rankine, NV34TCL_RC_ENABLE, 1);
+	so_data  (so, 0);
+
+	so_method(so, screen->rankine, NV34TCL_DEPTH_RANGE_NEAR, 2);
+	so_data  (so, fui(0.0));
+	so_data  (so, fui(1.0));
+
+	so_method(so, screen->rankine, NV34TCL_MULTISAMPLE_CONTROL, 1);
+	so_data  (so, 0xffff0000);
+
+	/* enables use of vp rather than fixed-function somehow */
+	so_method(so, screen->rankine, 0x1e94, 1);
+	so_data  (so, 0x13);
+
+	so_emit(nvws, so);
+	so_ref(NULL, &so);
+	nvws->push_flush(nvws, 0, NULL);
+
+	screen->pipe.winsys = ws;
+	screen->pipe.destroy = nv30_screen_destroy;
+
+	screen->pipe.get_name = nv30_screen_get_name;
+	screen->pipe.get_vendor = nv30_screen_get_vendor;
+	screen->pipe.get_param = nv30_screen_get_param;
+	screen->pipe.get_paramf = nv30_screen_get_paramf;
+
+	screen->pipe.is_format_supported = nv30_screen_surface_format_supported;
+
+	screen->pipe.surface_map = nv30_surface_map;
+	screen->pipe.surface_unmap = nv30_surface_unmap;
+
+	nv30_screen_init_miptree_functions(&screen->pipe);
+	u_simple_screen_init(&screen->pipe);
+
+	return &screen->pipe;
+}
diff --git a/src/gallium/drivers/nv30/nv30_screen.h b/src/gallium/drivers/nv30/nv30_screen.h
new file mode 100644
index 0000000000..b11e470f94
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_screen.h
@@ -0,0 +1,37 @@
+#ifndef __NV30_SCREEN_H__
+#define __NV30_SCREEN_H__
+
+#include "pipe/p_screen.h"
+#include "nv04/nv04_surface_2d.h"
+
+struct nv30_screen {
+	struct pipe_screen pipe;
+
+	struct nouveau_winsys *nvws;
+
+	unsigned cur_pctx;
+
+	/* HW graphics objects */
+	struct nv04_surface_2d *eng2d;
+	struct nouveau_grobj *rankine;
+	struct nouveau_notifier *sync;
+
+	/* Query object resources */
+	struct nouveau_notifier *query;
+	struct nouveau_resource *query_heap;
+
+	/* Vtxprog resources */
+	struct nouveau_resource *vp_exec_heap;
+	struct nouveau_resource *vp_data_heap;
+
+	/* Current 3D state of channel */
+	struct nouveau_stateobj *state[NV30_STATE_MAX];
+};
+
+static INLINE struct nv30_screen *
+nv30_screen(struct pipe_screen *screen)
+{
+	return (struct nv30_screen *)screen;
+}
+
+#endif
diff --git a/src/gallium/drivers/nv30/nv30_shader.h b/src/gallium/drivers/nv30/nv30_shader.h
new file mode 100644
index 0000000000..dd3a36f78f
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_shader.h
@@ -0,0 +1,490 @@
+#ifndef __NV30_SHADER_H__
+#define __NV30_SHADER_H__
+
+/* Vertex programs instruction set
+ *
+ * 128bit opcodes, split into 4 32-bit ones for ease of use.
+ *
+ * Non-native instructions
+ *   ABS - MOV + NV40_VP_INST0_DEST_ABS
+ *   POW - EX2 + MUL + LG2
+ *   SUB - ADD, second source negated
+ *   SWZ - MOV
+ *   XPD -  
+ *
+ * Register access
+ *   - Only one INPUT can be accessed per-instruction (move extras into TEMPs)
+ *   - Only one CONST can be accessed per-instruction (move extras into TEMPs)
+ *
+ * Relative Addressing
+ *   According to the value returned for
+ *   MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB
+ *
+ *   there are only two address registers available.  The destination in the
+ *   ARL instruction is set to TEMP <n> (The temp isn't actually written).
+ *
+ *   When using vanilla ARB_v_p, the proprietary driver will squish both the
+ *   available ADDRESS regs into the first hardware reg in the X and Y
+ *   components.
+ *
+ *   To use an address reg as an index into consts, the CONST_SRC is set to
+ *   (const_base + offset) and INDEX_CONST is set.
+ *
+ *   To access the second address reg use ADDR_REG_SELECT_1. A particular
+ *   component of the address regs is selected with ADDR_SWZ.
+ *
+ *   Only one address register can be accessed per instruction.
+ *
+ * Conditional execution (see NV_vertex_program{2,3} for details) Conditional
+ * execution of an instruction is enabled by setting COND_TEST_ENABLE, and
+ * selecting the condition which will allow the test to pass with
+ * COND_{FL,LT,...}.  It is possible to swizzle the values in the condition
+ * register, which allows for testing against an individual component.
+ *
+ * Branching:
+ *
+ *   The BRA/CAL instructions seem to follow a slightly different opcode
+ *   layout.  The destination instruction ID (IADDR) overlaps a source field.
+ *   Instruction ID's seem to be numbered based on the UPLOAD_FROM_ID FIFO
+ *   command, and is incremented automatically on each UPLOAD_INST FIFO
+ *   command.
+ *
+ *   Conditional branching is achieved by using the condition tests described
+ *   above.  There doesn't appear to be dedicated looping instructions, but
+ *   this can be done using a temp reg + conditional branching.
+ *
+ *   Subroutines may be uploaded before the main program itself, but the first
+ *   executed instruction is determined by the PROGRAM_START_ID FIFO command.
+ *
+ */
+
+/* DWORD 0 */
+
+#define NV30_VP_INST_ADDR_REG_SELECT_1        (1 << 24)
+#define NV30_VP_INST_SRC2_ABS           (1 << 23) /* guess */
+#define NV30_VP_INST_SRC1_ABS           (1 << 22) /* guess */
+#define NV30_VP_INST_SRC0_ABS           (1 << 21) /* guess */
+#define NV30_VP_INST_VEC_RESULT         (1 << 20)
+#define NV30_VP_INST_DEST_TEMP_ID_SHIFT        16
+#define NV30_VP_INST_DEST_TEMP_ID_MASK        (0x0F << 16)
+#define NV30_VP_INST_COND_UPDATE_ENABLE        (1<<15)
+#define NV30_VP_INST_VEC_DEST_TEMP_MASK      (0xF << 16)
+#define NV30_VP_INST_COND_TEST_ENABLE        (1<<14)
+#define NV30_VP_INST_COND_SHIFT          11
+#define NV30_VP_INST_COND_MASK          (0x07 << 11)
+#  define NV30_VP_INST_COND_FL  0 /* guess */  
+#  define NV30_VP_INST_COND_LT  1  
+#  define NV30_VP_INST_COND_EQ  2
+#  define NV30_VP_INST_COND_LE  3
+#  define NV30_VP_INST_COND_GT  4
+#  define NV30_VP_INST_COND_NE  5
+#  define NV30_VP_INST_COND_GE  6
+#  define NV30_VP_INST_COND_TR  7 /* guess */
+#define NV30_VP_INST_COND_SWZ_X_SHIFT        9
+#define NV30_VP_INST_COND_SWZ_X_MASK        (0x03 <<  9)
+#define NV30_VP_INST_COND_SWZ_Y_SHIFT        7
+#define NV30_VP_INST_COND_SWZ_Y_MASK        (0x03 <<  7)
+#define NV30_VP_INST_COND_SWZ_Z_SHIFT        5
+#define NV30_VP_INST_COND_SWZ_Z_MASK        (0x03 <<  5)
+#define NV30_VP_INST_COND_SWZ_W_SHIFT        3
+#define NV30_VP_INST_COND_SWZ_W_MASK        (0x03 <<  3)
+#define NV30_VP_INST_COND_SWZ_ALL_SHIFT        3
+#define NV30_VP_INST_COND_SWZ_ALL_MASK        (0xFF <<  3)
+#define NV30_VP_INST_ADDR_SWZ_SHIFT        1
+#define NV30_VP_INST_ADDR_SWZ_MASK        (0x03 <<  1)
+#define NV30_VP_INST_SCA_OPCODEH_SHIFT        0
+#define NV30_VP_INST_SCA_OPCODEH_MASK        (0x01 <<  0)
+
+/* DWORD 1 */
+#define NV30_VP_INST_SCA_OPCODEL_SHIFT        28
+#define NV30_VP_INST_SCA_OPCODEL_MASK        (0x0F << 28)
+#  define NV30_VP_INST_OP_NOP  0x00
+#  define NV30_VP_INST_OP_RCP  0x02
+#  define NV30_VP_INST_OP_RCC  0x03
+#  define NV30_VP_INST_OP_RSQ  0x04
+#  define NV30_VP_INST_OP_EXP  0x05
+#  define NV30_VP_INST_OP_LOG  0x06
+#  define NV30_VP_INST_OP_LIT  0x07
+#  define NV30_VP_INST_OP_BRA  0x09
+#  define NV30_VP_INST_OP_CAL  0x0B
+#  define NV30_VP_INST_OP_RET  0x0C
+#  define NV30_VP_INST_OP_LG2  0x0D
+#  define NV30_VP_INST_OP_EX2  0x0E
+#  define NV30_VP_INST_OP_SIN  0x0F
+#  define NV30_VP_INST_OP_COS  0x10
+#define NV30_VP_INST_VEC_OPCODE_SHIFT        23
+#define NV30_VP_INST_VEC_OPCODE_MASK        (0x1F << 23)
+#  define NV30_VP_INST_OP_NOPV  0x00
+#  define NV30_VP_INST_OP_MOV  0x01
+#  define NV30_VP_INST_OP_MUL  0x02
+#  define NV30_VP_INST_OP_ADD  0x03
+#  define NV30_VP_INST_OP_MAD  0x04
+#  define NV30_VP_INST_OP_DP3  0x05
+#  define NV30_VP_INST_OP_DP4  0x07
+#  define NV30_VP_INST_OP_DPH  0x06
+#  define NV30_VP_INST_OP_DST  0x08
+#  define NV30_VP_INST_OP_MIN  0x09
+#  define NV30_VP_INST_OP_MAX  0x0A
+#  define NV30_VP_INST_OP_SLT  0x0B
+#  define NV30_VP_INST_OP_SGE  0x0C
+#  define NV30_VP_INST_OP_ARL  0x0D
+#  define NV30_VP_INST_OP_FRC  0x0E
+#  define NV30_VP_INST_OP_FLR  0x0F
+#  define NV30_VP_INST_OP_SEQ  0x10
+#  define NV30_VP_INST_OP_SFL  0x11
+#  define NV30_VP_INST_OP_SGT  0x12
+#  define NV30_VP_INST_OP_SLE  0x13
+#  define NV30_VP_INST_OP_SNE  0x14
+#  define NV30_VP_INST_OP_STR  0x15
+#  define NV30_VP_INST_OP_SSG  0x16
+#  define NV30_VP_INST_OP_ARR  0x17
+#  define NV30_VP_INST_OP_ARA  0x18
+#define NV30_VP_INST_CONST_SRC_SHIFT        14
+#define NV30_VP_INST_CONST_SRC_MASK        (0xFF << 14)
+#define NV30_VP_INST_INPUT_SRC_SHIFT        9    /*NV20*/
+#define NV30_VP_INST_INPUT_SRC_MASK        (0x0F <<  9)  /*NV20*/
+#  define NV30_VP_INST_IN_POS  0    /* These seem to match the bindings specified in */
+#  define NV30_VP_INST_IN_WEIGHT  1    /* the ARB_v_p spec (2.14.3.1) */
+#  define NV30_VP_INST_IN_NORMAL  2    
+#  define NV30_VP_INST_IN_COL0  3    /* Should probably confirm them all though */
+#  define NV30_VP_INST_IN_COL1  4
+#  define NV30_VP_INST_IN_FOGC  5
+#  define NV30_VP_INST_IN_TC0  8
+#  define NV30_VP_INST_IN_TC(n)  (8+n)
+#define NV30_VP_INST_SRC0H_SHIFT        0    /*NV20*/
+#define NV30_VP_INST_SRC0H_MASK          (0x1FF << 0)  /*NV20*/
+
+/* Please note: the IADDR fields overlap other fields because they are used
+ * only for branch instructions.  See Branching: label above
+ *
+ * DWORD 2
+ */
+#define NV30_VP_INST_SRC0L_SHIFT        26    /*NV20*/
+#define NV30_VP_INST_SRC0L_MASK         (0x3F  <<26)  /* NV30_VP_SRC0_LOW_MASK << 26 */
+#define NV30_VP_INST_SRC1_SHIFT         11    /*NV20*/
+#define NV30_VP_INST_SRC1_MASK          (0x7FFF<<11)  /*NV20*/
+#define NV30_VP_INST_SRC2H_SHIFT        0    /*NV20*/
+#define NV30_VP_INST_SRC2H_MASK          (0x7FF << 0)  /* NV30_VP_SRC2_HIGH_MASK >> 4*/
+#define NV30_VP_INST_IADDR_SHIFT        2
+#define NV30_VP_INST_IADDR_MASK          (0xF <<  28)   /* NV30_VP_SRC2_LOW_MASK << 28 */
+
+/* DWORD 3 */
+#define NV30_VP_INST_SRC2L_SHIFT        28    /*NV20*/
+#define NV30_VP_INST_SRC2L_MASK          (0x0F  <<28)  /*NV20*/
+#define NV30_VP_INST_STEMP_WRITEMASK_SHIFT      24
+#define NV30_VP_INST_STEMP_WRITEMASK_MASK      (0x0F << 24)
+#define NV30_VP_INST_VTEMP_WRITEMASK_SHIFT      20
+#define NV30_VP_INST_VTEMP_WRITEMASK_MASK      (0x0F << 20)
+#define NV30_VP_INST_SDEST_WRITEMASK_SHIFT      16
+#define NV30_VP_INST_SDEST_WRITEMASK_MASK      (0x0F << 16)
+#define NV30_VP_INST_VDEST_WRITEMASK_SHIFT      12    /*NV20*/
+#define NV30_VP_INST_VDEST_WRITEMASK_MASK      (0x0F << 12)  /*NV20*/
+#define NV30_VP_INST_DEST_SHIFT        2
+#define NV30_VP_INST_DEST_MASK        (0x0F <<  2)
+#  define NV30_VP_INST_DEST_POS  0
+#  define NV30_VP_INST_DEST_BFC0  1
+#  define NV30_VP_INST_DEST_BFC1  2
+#  define NV30_VP_INST_DEST_COL0  3
+#  define NV30_VP_INST_DEST_COL1  4
+#  define NV30_VP_INST_DEST_FOGC  5
+#  define NV30_VP_INST_DEST_PSZ   6
+#  define NV30_VP_INST_DEST_TC(n)  (8+n)
+
+#define NV30_VP_INST_LAST                           (1 << 0)
+
+/* Useful to split the source selection regs into their pieces */
+#define NV30_VP_SRC0_HIGH_SHIFT                                                6
+#define NV30_VP_SRC0_HIGH_MASK                                        0x00007FC0
+#define NV30_VP_SRC0_LOW_MASK                                         0x0000003F
+#define NV30_VP_SRC2_HIGH_SHIFT                                                4
+#define NV30_VP_SRC2_HIGH_MASK                                        0x00007FF0
+#define NV30_VP_SRC2_LOW_MASK                                         0x0000000F
+
+
+/* Source-register definition - matches NV20 exactly */
+#define NV30_VP_SRC_NEGATE          (1<<14)
+#define NV30_VP_SRC_SWZ_X_SHIFT        12
+#define NV30_VP_SRC_REG_SWZ_X_MASK        (0x03  <<12)
+#define NV30_VP_SRC_SWZ_Y_SHIFT        10
+#define NV30_VP_SRC_REG_SWZ_Y_MASK        (0x03  <<10)
+#define NV30_VP_SRC_SWZ_Z_SHIFT        8
+#define NV30_VP_SRC_REG_SWZ_Z_MASK        (0x03  << 8)
+#define NV30_VP_SRC_SWZ_W_SHIFT        6
+#define NV30_VP_SRC_REG_SWZ_W_MASK        (0x03  << 6)
+#define NV30_VP_SRC_REG_SWZ_ALL_SHIFT        6
+#define NV30_VP_SRC_REG_SWZ_ALL_MASK        (0xFF  << 6)
+#define NV30_VP_SRC_TEMP_SRC_SHIFT        2
+#define NV30_VP_SRC_REG_TEMP_ID_MASK        (0x0F  << 0)
+#define NV30_VP_SRC_REG_TYPE_SHIFT        0
+#define NV30_VP_SRC_REG_TYPE_MASK        (0x03  << 0)
+#define NV30_VP_SRC_REG_TYPE_TEMP  1
+#define NV30_VP_SRC_REG_TYPE_INPUT  2
+#define NV30_VP_SRC_REG_TYPE_CONST  3 /* guess */
+
+/*
+ * Each fragment program opcode appears to be comprised of 4 32-bit values.
+ *
+ *   0 - Opcode, output reg/mask, ATTRIB source
+ *   1 - Source 0
+ *   2 - Source 1
+ *   3 - Source 2
+ *
+ * There appears to be no special difference between result regs and temp regs.
+ *     result.color == R0.xyzw
+ *     result.depth == R1.z
+ * When the fragprog contains instructions to write depth, NV30_TCL_PRIMITIVE_3D_UNK1D78=0
+ * otherwise it is set to 1.
+ *
+ * Constants are inserted directly after the instruction that uses them.
+ * 
+ * It appears that it's not possible to use two input registers in one
+ * instruction as the input sourcing is done in the instruction dword
+ * and not the source selection dwords.  As such instructions such as:
+ * 
+ *     ADD result.color, fragment.color, fragment.texcoord[0];
+ *
+ * must be split into two MOV's and then an ADD (nvidia does this) but
+ * I'm not sure why it's not just one MOV and then source the second input
+ * in the ADD instruction..
+ *
+ * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary
+ * negation requires multiplication with a const.
+ *
+ * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO/SWIZZLE_ONE
+ * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as SWIZZLE_ZERO
+ * is implemented simply by not writing to the relevant components of the destination.
+ *
+ * Conditional execution
+ *   TODO
+ * 
+ * Non-native instructions:
+ *   LIT
+ *   LRP - MAD+MAD
+ *   SUB - ADD, negate second source
+ *   RSQ - LG2 + EX2
+ *   POW - LG2 + MUL + EX2
+ *   SCS - COS + SIN
+ *   XPD
+ */
+
+//== Opcode / Destination selection ==
+#define NV30_FP_OP_PROGRAM_END          (1 << 0)
+#define NV30_FP_OP_OUT_REG_SHIFT        1
+#define NV30_FP_OP_OUT_REG_MASK          (31 << 1)  /* uncertain */
+/* Needs to be set when writing outputs to get expected result.. */
+#define NV30_FP_OP_OUT_REG_HALF          (1 << 7)
+#define NV30_FP_OP_COND_WRITE_ENABLE        (1 << 8)
+#define NV30_FP_OP_OUTMASK_SHIFT        9
+#define NV30_FP_OP_OUTMASK_MASK          (0xF << 9)
+#  define NV30_FP_OP_OUT_X  (1<<9)
+#  define NV30_FP_OP_OUT_Y  (1<<10)
+#  define NV30_FP_OP_OUT_Z  (1<<11)
+#  define NV30_FP_OP_OUT_W  (1<<12)
+/* Uncertain about these, especially the input_src values.. it's possible that
+ * they can be dynamically changed.
+ */
+#define NV30_FP_OP_INPUT_SRC_SHIFT        13
+#define NV30_FP_OP_INPUT_SRC_MASK        (15 << 13)
+#  define NV30_FP_OP_INPUT_SRC_POSITION  0x0
+#  define NV30_FP_OP_INPUT_SRC_COL0  0x1
+#  define NV30_FP_OP_INPUT_SRC_COL1  0x2
+#  define NV30_FP_OP_INPUT_SRC_FOGC  0x3
+#  define NV30_FP_OP_INPUT_SRC_TC0    0x4
+#  define NV30_FP_OP_INPUT_SRC_TC(n)  (0x4 + n)
+#define NV30_FP_OP_TEX_UNIT_SHIFT        17
+#define NV30_FP_OP_TEX_UNIT_MASK        (0xF << 17) /* guess */
+#define NV30_FP_OP_PRECISION_SHIFT        22
+#define NV30_FP_OP_PRECISION_MASK        (3 << 22)
+#   define NV30_FP_PRECISION_FP32  0
+#   define NV30_FP_PRECISION_FP16  1
+#   define NV30_FP_PRECISION_FX12  2
+#define NV30_FP_OP_OPCODE_SHIFT          24
+#define NV30_FP_OP_OPCODE_MASK          (0x3F << 24)
+#  define NV30_FP_OP_OPCODE_NOP  0x00
+#  define NV30_FP_OP_OPCODE_MOV  0x01
+#  define NV30_FP_OP_OPCODE_MUL  0x02
+#  define NV30_FP_OP_OPCODE_ADD  0x03
+#  define NV30_FP_OP_OPCODE_MAD  0x04
+#  define NV30_FP_OP_OPCODE_DP3  0x05
+#  define NV30_FP_OP_OPCODE_DP4  0x06
+#  define NV30_FP_OP_OPCODE_DST  0x07
+#  define NV30_FP_OP_OPCODE_MIN  0x08
+#  define NV30_FP_OP_OPCODE_MAX  0x09
+#  define NV30_FP_OP_OPCODE_SLT  0x0A
+#  define NV30_FP_OP_OPCODE_SGE  0x0B
+#  define NV30_FP_OP_OPCODE_SLE  0x0C
+#  define NV30_FP_OP_OPCODE_SGT  0x0D
+#  define NV30_FP_OP_OPCODE_SNE  0x0E
+#  define NV30_FP_OP_OPCODE_SEQ  0x0F
+#  define NV30_FP_OP_OPCODE_FRC  0x10
+#  define NV30_FP_OP_OPCODE_FLR  0x11
+#  define NV30_FP_OP_OPCODE_KIL  0x12
+#  define NV30_FP_OP_OPCODE_PK4B   0x13
+#  define NV30_FP_OP_OPCODE_UP4B   0x14
+#  define NV30_FP_OP_OPCODE_DDX  0x15 /* can only write XY */
+#  define NV30_FP_OP_OPCODE_DDY  0x16 /* can only write XY */
+#  define NV30_FP_OP_OPCODE_TEX  0x17
+#  define NV30_FP_OP_OPCODE_TXP  0x18
+#  define NV30_FP_OP_OPCODE_TXD  0x19
+#  define NV30_FP_OP_OPCODE_RCP  0x1A
+#  define NV30_FP_OP_OPCODE_RSQ  0x1B
+#  define NV30_FP_OP_OPCODE_EX2  0x1C
+#  define NV30_FP_OP_OPCODE_LG2  0x1D
+#  define NV30_FP_OP_OPCODE_LIT  0x1E
+#  define NV30_FP_OP_OPCODE_LRP  0x1F
+#  define NV30_FP_OP_OPCODE_STR  0x20 
+#  define NV30_FP_OP_OPCODE_SFL  0x21
+#  define NV30_FP_OP_OPCODE_COS  0x22
+#  define NV30_FP_OP_OPCODE_SIN  0x23
+#  define NV30_FP_OP_OPCODE_PK2H   0x24
+#  define NV30_FP_OP_OPCODE_UP2H   0x25
+#  define NV30_FP_OP_OPCODE_POW  0x26
+#  define NV30_FP_OP_OPCODE_PK4UB  0x27
+#  define NV30_FP_OP_OPCODE_UP4UB  0x28
+#  define NV30_FP_OP_OPCODE_PK2US  0x29
+#  define NV30_FP_OP_OPCODE_UP2US  0x2A
+#  define NV30_FP_OP_OPCODE_DP2A   0x2E
+#  define NV30_FP_OP_OPCODE_TXB  0x31
+#  define NV30_FP_OP_OPCODE_RFL  0x36
+#  define NV30_FP_OP_OPCODE_DIV  0x3A
+#define NV30_FP_OP_OUT_SAT          (1 << 31)
+
+/* high order bits of SRC0 */
+#define NV30_FP_OP_OUT_ABS          (1 << 29)
+#define NV30_FP_OP_COND_SWZ_W_SHIFT        27
+#define NV30_FP_OP_COND_SWZ_W_MASK        (3 << 27)
+#define NV30_FP_OP_COND_SWZ_Z_SHIFT        25
+#define NV30_FP_OP_COND_SWZ_Z_MASK        (3 << 25)
+#define NV30_FP_OP_COND_SWZ_Y_SHIFT        23
+#define NV30_FP_OP_COND_SWZ_Y_MASK        (3 << 23)
+#define NV30_FP_OP_COND_SWZ_X_SHIFT        21
+#define NV30_FP_OP_COND_SWZ_X_MASK        (3 << 21)
+#define NV30_FP_OP_COND_SWZ_ALL_SHIFT        21
+#define NV30_FP_OP_COND_SWZ_ALL_MASK        (0xFF << 21)
+#define NV30_FP_OP_COND_SHIFT          18
+#define NV30_FP_OP_COND_MASK          (0x07 << 18)
+#  define NV30_FP_OP_COND_FL  0
+#  define NV30_FP_OP_COND_LT  1
+#  define NV30_FP_OP_COND_EQ  2
+#  define NV30_FP_OP_COND_LE  3
+#  define NV30_FP_OP_COND_GT  4
+#  define NV30_FP_OP_COND_NE  5
+#  define NV30_FP_OP_COND_GE  6
+#  define NV30_FP_OP_COND_TR  7
+
+/* high order bits of SRC1 */
+#define NV30_FP_OP_DST_SCALE_SHIFT        28
+#define NV30_FP_OP_DST_SCALE_MASK        (3 << 28)
+#define NV30_FP_OP_DST_SCALE_1X                                                0
+#define NV30_FP_OP_DST_SCALE_2X                                                1
+#define NV30_FP_OP_DST_SCALE_4X                                                2
+#define NV30_FP_OP_DST_SCALE_8X                                                3
+#define NV30_FP_OP_DST_SCALE_INV_2X                                            5
+#define NV30_FP_OP_DST_SCALE_INV_4X                                            6
+#define NV30_FP_OP_DST_SCALE_INV_8X                                            7
+
+
+/* high order bits of SRC2 */
+#define NV30_FP_OP_INDEX_INPUT          (1 << 30)
+
+//== Register selection ==
+#define NV30_FP_REG_TYPE_SHIFT          0
+#define NV30_FP_REG_TYPE_MASK          (3 << 0)
+#  define NV30_FP_REG_TYPE_TEMP  0
+#  define NV30_FP_REG_TYPE_INPUT  1
+#  define NV30_FP_REG_TYPE_CONST  2
+#define NV30_FP_REG_SRC_SHIFT          2 /* uncertain */
+#define NV30_FP_REG_SRC_MASK          (31 << 2)
+#define NV30_FP_REG_SRC_HALF          (1 << 8)
+#define NV30_FP_REG_SWZ_ALL_SHIFT        9
+#define NV30_FP_REG_SWZ_ALL_MASK        (255 << 9)
+#define NV30_FP_REG_SWZ_X_SHIFT          9
+#define NV30_FP_REG_SWZ_X_MASK          (3 << 9)
+#define NV30_FP_REG_SWZ_Y_SHIFT          11
+#define NV30_FP_REG_SWZ_Y_MASK          (3 << 11)
+#define NV30_FP_REG_SWZ_Z_SHIFT          13
+#define NV30_FP_REG_SWZ_Z_MASK          (3 << 13)
+#define NV30_FP_REG_SWZ_W_SHIFT          15
+#define NV30_FP_REG_SWZ_W_MASK          (3 << 15)
+#  define NV30_FP_SWIZZLE_X  0
+#  define NV30_FP_SWIZZLE_Y  1
+#  define NV30_FP_SWIZZLE_Z  2
+#  define NV30_FP_SWIZZLE_W  3
+#define NV30_FP_REG_NEGATE          (1 << 17)
+
+#define NV30SR_NONE	0
+#define NV30SR_OUTPUT	1
+#define NV30SR_INPUT	2
+#define NV30SR_TEMP	3
+#define NV30SR_CONST	4
+
+struct nv30_sreg {
+	int type;
+	int index;
+
+	int dst_scale;
+
+	int negate;
+	int abs;
+	int swz[4];
+
+	int cc_update;
+	int cc_update_reg;
+	int cc_test;
+	int cc_test_reg;
+	int cc_swz[4];
+};
+
+static INLINE struct nv30_sreg
+nv30_sr(int type, int index)
+{
+	struct nv30_sreg temp = {
+		.type = type,
+		.index = index,
+		.dst_scale = DEF_SCALE,
+		.abs = 0,
+		.negate = 0,
+		.swz = { 0, 1, 2, 3 },
+		.cc_update = 0,
+		.cc_update_reg = 0,
+		.cc_test = DEF_CTEST,
+		.cc_test_reg = 0,
+		.cc_swz = { 0, 1, 2, 3 },
+	};
+	return temp;
+}
+
+static INLINE struct nv30_sreg
+nv30_sr_swz(struct nv30_sreg src, int x, int y, int z, int w)
+{
+	struct nv30_sreg dst = src;
+
+	dst.swz[SWZ_X] = src.swz[x];
+	dst.swz[SWZ_Y] = src.swz[y];
+	dst.swz[SWZ_Z] = src.swz[z];
+	dst.swz[SWZ_W] = src.swz[w];
+	return dst;
+}
+
+static INLINE struct nv30_sreg
+nv30_sr_neg(struct nv30_sreg src)
+{
+	src.negate = !src.negate;
+	return src;
+}
+
+static INLINE struct nv30_sreg
+nv30_sr_abs(struct nv30_sreg src)
+{
+	src.abs = 1;
+	return src;
+}
+
+static INLINE struct nv30_sreg
+nv30_sr_scale(struct nv30_sreg src, int scale)
+{
+	src.dst_scale = scale;
+	return src;
+}
+
+#endif
diff --git a/src/gallium/drivers/nv30/nv30_state.c b/src/gallium/drivers/nv30/nv30_state.c
new file mode 100644
index 0000000000..26147565a5
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_state.c
@@ -0,0 +1,725 @@
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "tgsi/tgsi_parse.h"
+
+#include "nv30_context.h"
+#include "nv30_state.h"
+
+static void *
+nv30_blend_state_create(struct pipe_context *pipe,
+			const struct pipe_blend_state *cso)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nouveau_grobj *rankine = nv30->screen->rankine;
+	struct nv30_blend_state *bso = CALLOC(1, sizeof(*bso));
+	struct nouveau_stateobj *so = so_new(16, 0);
+
+	if (cso->blend_enable) {
+		so_method(so, rankine, NV34TCL_BLEND_FUNC_ENABLE, 3);
+		so_data  (so, 1);
+		so_data  (so, (nvgl_blend_func(cso->alpha_src_factor) << 16) |
+			       nvgl_blend_func(cso->rgb_src_factor));
+		so_data  (so, nvgl_blend_func(cso->alpha_dst_factor) << 16 |
+			      nvgl_blend_func(cso->rgb_dst_factor));
+		/* FIXME: Gallium assumes GL_EXT_blend_func_separate.
+		   It is not the case for NV30 */
+		so_method(so, rankine, NV34TCL_BLEND_EQUATION, 1);
+		so_data  (so, nvgl_blend_eqn(cso->rgb_func));
+	} else {
+		so_method(so, rankine, NV34TCL_BLEND_FUNC_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_method(so, rankine, NV34TCL_COLOR_MASK, 1);
+	so_data  (so, (((cso->colormask & PIPE_MASK_A) ? (0x01 << 24) : 0) |
+		       ((cso->colormask & PIPE_MASK_R) ? (0x01 << 16) : 0) |
+		       ((cso->colormask & PIPE_MASK_G) ? (0x01 <<  8) : 0) |
+		       ((cso->colormask & PIPE_MASK_B) ? (0x01 <<  0) : 0)));
+
+	if (cso->logicop_enable) {
+		so_method(so, rankine, NV34TCL_COLOR_LOGIC_OP_ENABLE, 2);
+		so_data  (so, 1);
+		so_data  (so, nvgl_logicop_func(cso->logicop_func));
+	} else {
+		so_method(so, rankine, NV34TCL_COLOR_LOGIC_OP_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_method(so, rankine, NV34TCL_DITHER_ENABLE, 1);
+	so_data  (so, cso->dither ? 1 : 0);
+
+	so_ref(so, &bso->so);
+	bso->pipe = *cso;
+	return (void *)bso;
+}
+
+static void
+nv30_blend_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->blend = hwcso;
+	nv30->dirty |= NV30_NEW_BLEND;
+}
+
+static void
+nv30_blend_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv30_blend_state *bso = hwcso;
+
+	so_ref(NULL, &bso->so);
+	FREE(bso);
+}
+
+
+static INLINE unsigned
+wrap_mode(unsigned wrap) {
+	unsigned ret;
+
+	switch (wrap) {
+	case PIPE_TEX_WRAP_REPEAT:
+		ret = NV34TCL_TX_WRAP_S_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_REPEAT:
+		ret = NV34TCL_TX_WRAP_S_MIRRORED_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+		ret = NV34TCL_TX_WRAP_S_CLAMP_TO_EDGE;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+		ret = NV34TCL_TX_WRAP_S_CLAMP_TO_BORDER;
+		break;
+	case PIPE_TEX_WRAP_CLAMP:
+		ret = NV34TCL_TX_WRAP_S_CLAMP;
+		break;
+/*	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+		ret = NV34TCL_TX_WRAP_S_MIRROR_CLAMP_TO_EDGE;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+		ret = NV34TCL_TX_WRAP_S_MIRROR_CLAMP_TO_BORDER;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP:
+		ret = NV34TCL_TX_WRAP_S_MIRROR_CLAMP;
+		break;*/
+	default:
+		NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
+		ret = NV34TCL_TX_WRAP_S_REPEAT;
+		break;
+	}
+
+	return ret >> NV34TCL_TX_WRAP_S_SHIFT;
+}
+
+static void *
+nv30_sampler_state_create(struct pipe_context *pipe,
+			  const struct pipe_sampler_state *cso)
+{
+	struct nv30_sampler_state *ps;
+	uint32_t filter = 0;
+
+	ps = MALLOC(sizeof(struct nv30_sampler_state));
+
+	ps->fmt = 0;
+	/* TODO: Not all RECTs formats have this bit set, bits 15-8 of format
+	   are the tx format to use. We should store normalized coord flag
+	   in sampler state structure, and set appropriate format in
+	   nvxx_fragtex_build()
+	 */
+	/*NV34TCL_TX_FORMAT_RECT*/
+	/*if (!cso->normalized_coords) {
+		ps->fmt |= (1<<14) ;
+	}*/
+
+	ps->wrap = ((wrap_mode(cso->wrap_s) << NV34TCL_TX_WRAP_S_SHIFT) |
+		    (wrap_mode(cso->wrap_t) << NV34TCL_TX_WRAP_T_SHIFT) |
+		    (wrap_mode(cso->wrap_r) << NV34TCL_TX_WRAP_R_SHIFT));
+
+	ps->en = 0;
+
+	if (cso->max_anisotropy >= 8.0) {
+		ps->en |= NV34TCL_TX_ENABLE_ANISO_8X;
+	} else
+	if (cso->max_anisotropy >= 4.0) {
+		ps->en |= NV34TCL_TX_ENABLE_ANISO_4X;
+	} else
+	if (cso->max_anisotropy >= 2.0) {
+		ps->en |= NV34TCL_TX_ENABLE_ANISO_2X;
+	}
+
+	switch (cso->mag_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		filter |= NV34TCL_TX_FILTER_MAGNIFY_LINEAR;
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		filter |= NV34TCL_TX_FILTER_MAGNIFY_NEAREST;
+		break;
+	}
+
+	switch (cso->min_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_NEAREST;
+			break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV34TCL_TX_FILTER_MINIFY_LINEAR;
+			break;
+		}
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_NEAREST;
+		break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV34TCL_TX_FILTER_MINIFY_NEAREST;
+			break;
+		}
+		break;
+	}
+
+	ps->filt = filter;
+
+	{
+		float limit;
+
+		limit = CLAMP(cso->lod_bias, -16.0, 15.0);
+		ps->filt |= (int)(cso->lod_bias * 256.0) & 0x1fff;
+
+		limit = CLAMP(cso->max_lod, 0.0, 15.0);
+		ps->en |= (int)(limit) << 14 /*NV34TCL_TX_ENABLE_MIPMAP_MAX_LOD_SHIFT*/;
+
+		limit = CLAMP(cso->min_lod, 0.0, 15.0);
+		ps->en |= (int)(limit) << 26 /*NV34TCL_TX_ENABLE_MIPMAP_MIN_LOD_SHIFT*/;
+	}
+
+	if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+		switch (cso->compare_func) {
+		case PIPE_FUNC_NEVER:
+			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_NEVER;
+			break;
+		case PIPE_FUNC_GREATER:
+			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_GREATER;
+			break;
+		case PIPE_FUNC_EQUAL:
+			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_EQUAL;
+			break;
+		case PIPE_FUNC_GEQUAL:
+			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_GEQUAL;
+			break;
+		case PIPE_FUNC_LESS:
+			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_LESS;
+			break;
+		case PIPE_FUNC_NOTEQUAL:
+			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_NOTEQUAL;
+			break;
+		case PIPE_FUNC_LEQUAL:
+			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_LEQUAL;
+			break;
+		case PIPE_FUNC_ALWAYS:
+			ps->wrap |= NV34TCL_TX_WRAP_RCOMP_ALWAYS;
+			break;
+		default:
+			break;
+		}
+	}
+
+	ps->bcol = ((float_to_ubyte(cso->border_color[3]) << 24) |
+		    (float_to_ubyte(cso->border_color[0]) << 16) |
+		    (float_to_ubyte(cso->border_color[1]) <<  8) |
+		    (float_to_ubyte(cso->border_color[2]) <<  0));
+
+	return (void *)ps;
+}
+
+static void
+nv30_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		nv30->tex_sampler[unit] = sampler[unit];
+		nv30->dirty_samplers |= (1 << unit);
+	}
+
+	for (unit = nr; unit < nv30->nr_samplers; unit++) {
+		nv30->tex_sampler[unit] = NULL;
+		nv30->dirty_samplers |= (1 << unit);
+	}
+
+	nv30->nr_samplers = nr;
+	nv30->dirty |= NV30_NEW_SAMPLER;
+}
+
+static void
+nv30_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void
+nv30_set_sampler_texture(struct pipe_context *pipe, unsigned nr,
+			 struct pipe_texture **miptree)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		pipe_texture_reference((struct pipe_texture **)
+				       &nv30->tex_miptree[unit], miptree[unit]);
+		nv30->dirty_samplers |= (1 << unit);
+	}
+
+	for (unit = nr; unit < nv30->nr_textures; unit++) {
+		pipe_texture_reference((struct pipe_texture **)
+				       &nv30->tex_miptree[unit], NULL);
+		nv30->dirty_samplers |= (1 << unit);
+	}
+
+	nv30->nr_textures = nr;
+	nv30->dirty |= NV30_NEW_SAMPLER;
+}
+
+static void *
+nv30_rasterizer_state_create(struct pipe_context *pipe,
+			     const struct pipe_rasterizer_state *cso)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nv30_rasterizer_state *rsso = CALLOC(1, sizeof(*rsso));
+	struct nouveau_stateobj *so = so_new(32, 0);
+	struct nouveau_grobj *rankine = nv30->screen->rankine;
+
+	/*XXX: ignored:
+	 * 	light_twoside
+	 * 	point_smooth -nohw
+	 * 	multisample
+	 */
+
+	so_method(so, rankine, NV34TCL_SHADE_MODEL, 1);
+	so_data  (so, cso->flatshade ? NV34TCL_SHADE_MODEL_FLAT :
+				       NV34TCL_SHADE_MODEL_SMOOTH);
+
+	so_method(so, rankine, NV34TCL_LINE_WIDTH, 2);
+	so_data  (so, (unsigned char)(cso->line_width * 8.0) & 0xff);
+	so_data  (so, cso->line_smooth ? 1 : 0);
+	so_method(so, rankine, NV34TCL_LINE_STIPPLE_ENABLE, 2);
+	so_data  (so, cso->line_stipple_enable ? 1 : 0);
+	so_data  (so, (cso->line_stipple_pattern << 16) |
+		       cso->line_stipple_factor);
+
+	so_method(so, rankine, NV34TCL_POINT_SIZE, 1);
+	so_data  (so, fui(cso->point_size));
+
+	so_method(so, rankine, NV34TCL_POLYGON_MODE_FRONT, 6);
+	if (cso->front_winding == PIPE_WINDING_CCW) {
+		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
+		so_data(so, nvgl_polygon_mode(cso->fill_cw));
+		switch (cso->cull_mode) {
+		case PIPE_WINDING_CCW:
+			so_data(so, NV34TCL_CULL_FACE_FRONT);
+			break;
+		case PIPE_WINDING_CW:
+			so_data(so, NV34TCL_CULL_FACE_BACK);
+			break;
+		case PIPE_WINDING_BOTH:
+			so_data(so, NV34TCL_CULL_FACE_FRONT_AND_BACK);
+			break;
+		default:
+			so_data(so, NV34TCL_CULL_FACE_BACK);
+			break;
+		}
+		so_data(so, NV34TCL_FRONT_FACE_CCW);
+	} else {
+		so_data(so, nvgl_polygon_mode(cso->fill_cw));
+		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
+		switch (cso->cull_mode) {
+		case PIPE_WINDING_CCW:
+			so_data(so, NV34TCL_CULL_FACE_BACK);
+			break;
+		case PIPE_WINDING_CW:
+			so_data(so, NV34TCL_CULL_FACE_FRONT);
+			break;
+		case PIPE_WINDING_BOTH:
+			so_data(so, NV34TCL_CULL_FACE_FRONT_AND_BACK);
+			break;
+		default:
+			so_data(so, NV34TCL_CULL_FACE_BACK);
+			break;
+		}
+		so_data(so, NV34TCL_FRONT_FACE_CW);
+	}
+	so_data(so, cso->poly_smooth ? 1 : 0);
+	so_data(so, (cso->cull_mode != PIPE_WINDING_NONE) ? 1 : 0);
+
+	so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1);
+	so_data  (so, cso->poly_stipple_enable ? 1 : 0);
+
+	so_method(so, rankine, NV34TCL_POLYGON_OFFSET_POINT_ENABLE, 3);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if (cso->offset_cw || cso->offset_ccw) {
+		so_method(so, rankine, NV34TCL_POLYGON_OFFSET_FACTOR, 2);
+		so_data  (so, fui(cso->offset_scale));
+		so_data  (so, fui(cso->offset_units * 2));
+	}
+
+	so_method(so, rankine, NV34TCL_POINT_SPRITE, 1);
+	if (cso->point_sprite) {
+		unsigned psctl = (1 << 0), i;
+
+		for (i = 0; i < 8; i++) {
+			if (cso->sprite_coord_mode[i] != PIPE_SPRITE_COORD_NONE)
+				psctl |= (1 << (8 + i));
+		}
+
+		so_data(so, psctl);
+	} else {
+		so_data(so, 0);
+	}
+
+	so_ref(so, &rsso->so);
+	rsso->pipe = *cso;
+	return (void *)rsso;
+}
+
+static void
+nv30_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->rasterizer = hwcso;
+	nv30->dirty |= NV30_NEW_RAST;
+	/*nv30->draw_dirty |= NV30_NEW_RAST;*/
+}
+
+static void
+nv30_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv30_rasterizer_state *rsso = hwcso;
+
+	so_ref(NULL, &rsso->so);
+	FREE(rsso);
+}
+
+static void *
+nv30_depth_stencil_alpha_state_create(struct pipe_context *pipe,
+			const struct pipe_depth_stencil_alpha_state *cso)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nv30_zsa_state *zsaso = CALLOC(1, sizeof(*zsaso));
+	struct nouveau_stateobj *so = so_new(32, 0);
+	struct nouveau_grobj *rankine = nv30->screen->rankine;
+
+	so_method(so, rankine, NV34TCL_DEPTH_FUNC, 3);
+	so_data  (so, nvgl_comparison_op(cso->depth.func));
+	so_data  (so, cso->depth.writemask ? 1 : 0);
+	so_data  (so, cso->depth.enabled ? 1 : 0);
+
+	so_method(so, rankine, NV34TCL_ALPHA_FUNC_ENABLE, 3);
+	so_data  (so, cso->alpha.enabled ? 1 : 0);
+	so_data  (so, nvgl_comparison_op(cso->alpha.func));
+	so_data  (so, float_to_ubyte(cso->alpha.ref_value));
+
+	if (cso->stencil[0].enabled) {
+		so_method(so, rankine, NV34TCL_STENCIL_FRONT_ENABLE, 8);
+		so_data  (so, cso->stencil[0].enabled ? 1 : 0);
+		so_data  (so, cso->stencil[0].writemask);
+		so_data  (so, nvgl_comparison_op(cso->stencil[0].func));
+		so_data  (so, cso->stencil[0].ref_value);
+		so_data  (so, cso->stencil[0].valuemask);
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].fail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].zfail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].zpass_op));
+	} else {
+		so_method(so, rankine, NV34TCL_STENCIL_FRONT_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	if (cso->stencil[1].enabled) {
+		so_method(so, rankine, NV34TCL_STENCIL_BACK_ENABLE, 8);
+		so_data  (so, cso->stencil[1].enabled ? 1 : 0);
+		so_data  (so, cso->stencil[1].writemask);
+		so_data  (so, nvgl_comparison_op(cso->stencil[1].func));
+		so_data  (so, cso->stencil[1].ref_value);
+		so_data  (so, cso->stencil[1].valuemask);
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].fail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].zfail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].zpass_op));
+	} else {
+		so_method(so, rankine, NV34TCL_STENCIL_BACK_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_ref(so, &zsaso->so);
+	zsaso->pipe = *cso;
+	return (void *)zsaso;
+}
+
+static void
+nv30_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->zsa = hwcso;
+	nv30->dirty |= NV30_NEW_ZSA;
+}
+
+static void
+nv30_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv30_zsa_state *zsaso = hwcso;
+
+	so_ref(NULL, &zsaso->so);
+	FREE(zsaso);
+}
+
+static void *
+nv30_vp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	/*struct nv30_context *nv30 = nv30_context(pipe);*/
+	struct nv30_vertex_program *vp;
+
+	vp = CALLOC(1, sizeof(struct nv30_vertex_program));
+	vp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+	/*vp->draw = draw_create_vertex_shader(nv30->draw, &vp->pipe);*/
+
+	return (void *)vp;
+}
+
+static void
+nv30_vp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->vertprog = hwcso;
+	nv30->dirty |= NV30_NEW_VERTPROG;
+	/*nv30->draw_dirty |= NV30_NEW_VERTPROG;*/
+}
+
+static void
+nv30_vp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nv30_vertex_program *vp = hwcso;
+
+	/*draw_delete_vertex_shader(nv30->draw, vp->draw);*/
+	nv30_vertprog_destroy(nv30, vp);
+	FREE((void*)vp->pipe.tokens);
+	FREE(vp);
+}
+
+static void *
+nv30_fp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	struct nv30_fragment_program *fp;
+
+	fp = CALLOC(1, sizeof(struct nv30_fragment_program));
+	fp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+
+	tgsi_scan_shader(fp->pipe.tokens, &fp->info);
+
+	return (void *)fp;
+}
+
+static void
+nv30_fp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->fragprog = hwcso;
+	nv30->dirty |= NV30_NEW_FRAGPROG;
+}
+
+static void
+nv30_fp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nv30_fragment_program *fp = hwcso;
+
+	nv30_fragprog_destroy(nv30, fp);
+	FREE((void*)fp->pipe.tokens);
+	FREE(fp);
+}
+
+static void
+nv30_set_blend_color(struct pipe_context *pipe,
+		     const struct pipe_blend_color *bcol)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->blend_colour = *bcol;
+	nv30->dirty |= NV30_NEW_BCOL;
+}
+
+static void
+nv30_set_clip_state(struct pipe_context *pipe,
+		    const struct pipe_clip_state *clip)
+{
+}
+
+static void
+nv30_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
+			 const struct pipe_constant_buffer *buf )
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->constbuf[shader] = buf->buffer;
+	nv30->constbuf_nr[shader] = buf->buffer->size / (4 * sizeof(float));
+
+	if (shader == PIPE_SHADER_VERTEX) {
+		nv30->dirty |= NV30_NEW_VERTPROG;
+	} else
+	if (shader == PIPE_SHADER_FRAGMENT) {
+		nv30->dirty |= NV30_NEW_FRAGPROG;
+	}
+}
+
+static void
+nv30_set_framebuffer_state(struct pipe_context *pipe,
+			   const struct pipe_framebuffer_state *fb)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->framebuffer = *fb;
+	nv30->dirty |= NV30_NEW_FB;
+}
+
+static void
+nv30_set_polygon_stipple(struct pipe_context *pipe,
+			 const struct pipe_poly_stipple *stipple)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	memcpy(nv30->stipple, stipple->stipple, 4 * 32);
+	nv30->dirty |= NV30_NEW_STIPPLE;
+}
+
+static void
+nv30_set_scissor_state(struct pipe_context *pipe,
+		       const struct pipe_scissor_state *s)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->scissor = *s;
+	nv30->dirty |= NV30_NEW_SCISSOR;
+}
+
+static void
+nv30_set_viewport_state(struct pipe_context *pipe,
+			const struct pipe_viewport_state *vpt)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->viewport = *vpt;
+	nv30->dirty |= NV30_NEW_VIEWPORT;
+	/*nv30->draw_dirty |= NV30_NEW_VIEWPORT;*/
+}
+
+static void
+nv30_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
+			const struct pipe_vertex_buffer *vb)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	memcpy(nv30->vtxbuf, vb, sizeof(*vb) * count);
+	nv30->vtxbuf_nr = count;
+
+	nv30->dirty |= NV30_NEW_ARRAYS;
+	/*nv30->draw_dirty |= NV30_NEW_ARRAYS;*/
+}
+
+static void
+nv30_set_vertex_elements(struct pipe_context *pipe, unsigned count,
+			 const struct pipe_vertex_element *ve)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	memcpy(nv30->vtxelt, ve, sizeof(*ve) * count);
+	nv30->vtxelt_nr = count;
+
+	nv30->dirty |= NV30_NEW_ARRAYS;
+	/*nv30->draw_dirty |= NV30_NEW_ARRAYS;*/
+}
+
+static void
+nv30_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+
+	nv30->edgeflags = bitfield;
+	nv30->dirty |= NV30_NEW_ARRAYS;
+	/*nv30->draw_dirty |= NV30_NEW_ARRAYS;*/
+}
+
+void
+nv30_init_state_functions(struct nv30_context *nv30)
+{
+	nv30->pipe.create_blend_state = nv30_blend_state_create;
+	nv30->pipe.bind_blend_state = nv30_blend_state_bind;
+	nv30->pipe.delete_blend_state = nv30_blend_state_delete;
+
+	nv30->pipe.create_sampler_state = nv30_sampler_state_create;
+	nv30->pipe.bind_sampler_states = nv30_sampler_state_bind;
+	nv30->pipe.delete_sampler_state = nv30_sampler_state_delete;
+	nv30->pipe.set_sampler_textures = nv30_set_sampler_texture;
+
+	nv30->pipe.create_rasterizer_state = nv30_rasterizer_state_create;
+	nv30->pipe.bind_rasterizer_state = nv30_rasterizer_state_bind;
+	nv30->pipe.delete_rasterizer_state = nv30_rasterizer_state_delete;
+
+	nv30->pipe.create_depth_stencil_alpha_state =
+		nv30_depth_stencil_alpha_state_create;
+	nv30->pipe.bind_depth_stencil_alpha_state =
+		nv30_depth_stencil_alpha_state_bind;
+	nv30->pipe.delete_depth_stencil_alpha_state =
+		nv30_depth_stencil_alpha_state_delete;
+
+	nv30->pipe.create_vs_state = nv30_vp_state_create;
+	nv30->pipe.bind_vs_state = nv30_vp_state_bind;
+	nv30->pipe.delete_vs_state = nv30_vp_state_delete;
+
+	nv30->pipe.create_fs_state = nv30_fp_state_create;
+	nv30->pipe.bind_fs_state = nv30_fp_state_bind;
+	nv30->pipe.delete_fs_state = nv30_fp_state_delete;
+
+	nv30->pipe.set_blend_color = nv30_set_blend_color;
+	nv30->pipe.set_clip_state = nv30_set_clip_state;
+	nv30->pipe.set_constant_buffer = nv30_set_constant_buffer;
+	nv30->pipe.set_framebuffer_state = nv30_set_framebuffer_state;
+	nv30->pipe.set_polygon_stipple = nv30_set_polygon_stipple;
+	nv30->pipe.set_scissor_state = nv30_set_scissor_state;
+	nv30->pipe.set_viewport_state = nv30_set_viewport_state;
+
+	nv30->pipe.set_edgeflags = nv30_set_edgeflags;
+	nv30->pipe.set_vertex_buffers = nv30_set_vertex_buffers;
+	nv30->pipe.set_vertex_elements = nv30_set_vertex_elements;
+}
+
diff --git a/src/gallium/drivers/nv30/nv30_state.h b/src/gallium/drivers/nv30/nv30_state.h
new file mode 100644
index 0000000000..2023278e37
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_state.h
@@ -0,0 +1,88 @@
+#ifndef __NV30_STATE_H__
+#define __NV30_STATE_H__
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_scan.h"
+
+struct nv30_sampler_state {
+	uint32_t fmt;
+	uint32_t wrap;
+	uint32_t en;
+	uint32_t filt;
+	uint32_t bcol;
+};
+
+struct nv30_vertex_program_exec {
+	uint32_t data[4];
+	boolean has_branch_offset;
+	int const_index;
+};
+
+struct nv30_vertex_program_data {
+	int index; /* immediates == -1 */
+	float value[4];
+};
+
+struct nv30_vertex_program {
+	struct pipe_shader_state pipe;
+
+	boolean translated;
+
+	struct nv30_vertex_program_exec *insns;
+	unsigned nr_insns;
+	struct nv30_vertex_program_data *consts;
+	unsigned nr_consts;
+
+	struct nouveau_resource *exec;
+	unsigned exec_start;
+	struct nouveau_resource *data;
+	unsigned data_start;
+	unsigned data_start_min;
+
+	uint32_t ir;
+	uint32_t or;
+	struct nouveau_stateobj *so;
+};
+
+struct nv30_fragment_program_data {
+	unsigned offset;
+	unsigned index;
+};
+
+struct nv30_fragment_program {
+	struct pipe_shader_state pipe;
+	struct tgsi_shader_info info;
+
+	boolean translated;
+	boolean on_hw;
+	unsigned samplers;
+
+	uint32_t *insn;
+	int       insn_len;
+
+	struct nv30_fragment_program_data *consts;
+	unsigned nr_consts;
+
+	struct pipe_buffer *buffer;
+
+	uint32_t fp_control;
+	uint32_t fp_reg_control;
+	struct nouveau_stateobj *so;
+};
+
+struct nv30_miptree {
+	struct pipe_texture base;
+
+	struct pipe_buffer *buffer;
+	uint total_size;
+
+	struct pipe_texture *shadow_tex;
+	struct pipe_surface *shadow_surface;
+
+	struct {
+		uint pitch;
+		uint *image_offset;
+	} level[PIPE_MAX_TEXTURE_LEVELS];
+};
+
+#endif
diff --git a/src/gallium/drivers/nv30/nv30_state_blend.c b/src/gallium/drivers/nv30/nv30_state_blend.c
new file mode 100644
index 0000000000..44d43e132a
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_state_blend.c
@@ -0,0 +1,40 @@
+#include "nv30_context.h"
+
+static boolean
+nv30_state_blend_validate(struct nv30_context *nv30)
+{
+	so_ref(nv30->blend->so, &nv30->state.hw[NV30_STATE_BLEND]);
+	return TRUE;
+}
+
+struct nv30_state_entry nv30_state_blend = {
+	.validate = nv30_state_blend_validate,
+	.dirty = {
+		.pipe = NV30_NEW_BLEND,
+		.hw = NV30_STATE_BLEND
+	}
+};
+
+static boolean
+nv30_state_blend_colour_validate(struct nv30_context *nv30)
+{
+	struct nouveau_stateobj *so = so_new(2, 0);
+	struct pipe_blend_color *bcol = &nv30->blend_colour;
+
+	so_method(so, nv30->screen->rankine, NV34TCL_BLEND_COLOR, 1);
+	so_data  (so, ((float_to_ubyte(bcol->color[3]) << 24) |
+		       (float_to_ubyte(bcol->color[0]) << 16) |
+		       (float_to_ubyte(bcol->color[1]) <<  8) |
+		       (float_to_ubyte(bcol->color[2]) <<  0)));
+
+	so_ref(so, &nv30->state.hw[NV30_STATE_BCOL]);
+	return TRUE;
+}
+
+struct nv30_state_entry nv30_state_blend_colour = {
+	.validate = nv30_state_blend_colour_validate,
+	.dirty = {
+		.pipe = NV30_NEW_BCOL,
+		.hw = NV30_STATE_BCOL
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_state_emit.c b/src/gallium/drivers/nv30/nv30_state_emit.c
new file mode 100644
index 0000000000..f77b08ff69
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_state_emit.c
@@ -0,0 +1,118 @@
+#include "nv30_context.h"
+#include "nv30_state.h"
+
+static struct nv30_state_entry *render_states[] = {
+	&nv30_state_framebuffer,
+	&nv30_state_rasterizer,
+	&nv30_state_scissor,
+	&nv30_state_stipple,
+	&nv30_state_fragprog,
+	&nv30_state_fragtex,
+	&nv30_state_vertprog,
+	&nv30_state_blend,
+	&nv30_state_blend_colour,
+	&nv30_state_zsa,
+	&nv30_state_viewport,
+	&nv30_state_vbo,
+	NULL
+};
+
+static void
+nv30_state_do_validate(struct nv30_context *nv30,
+		       struct nv30_state_entry **states)
+{
+	const struct pipe_framebuffer_state *fb = &nv30->framebuffer;
+	unsigned i;
+
+	for (i = 0; i < fb->nr_cbufs; i++)
+		fb->cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED;
+	if (fb->zsbuf)
+		fb->zsbuf->status = PIPE_SURFACE_STATUS_DEFINED;
+
+	while (*states) {
+		struct nv30_state_entry *e = *states;
+
+		if (nv30->dirty & e->dirty.pipe) {
+			if (e->validate(nv30)) {
+				nv30->state.dirty |= (1ULL << e->dirty.hw);
+			}
+		}
+
+		states++;
+	}
+	nv30->dirty = 0;
+}
+
+void
+nv30_state_emit(struct nv30_context *nv30)
+{
+	struct nv30_state *state = &nv30->state;
+	struct nv30_screen *screen = nv30->screen;
+	unsigned i, samplers;
+	uint64_t states;
+
+	if (nv30->pctx_id != screen->cur_pctx) {
+		for (i = 0; i < NV30_STATE_MAX; i++) {
+			if (state->hw[i] && screen->state[i] != state->hw[i])
+				state->dirty |= (1ULL << i);
+		}
+
+		screen->cur_pctx = nv30->pctx_id;
+	}
+
+	for (i = 0, states = state->dirty; states; i++) {
+		if (!(states & (1ULL << i)))
+			continue;
+		so_ref (state->hw[i], &nv30->screen->state[i]);
+		if (state->hw[i])
+			so_emit(nv30->nvws, nv30->screen->state[i]);
+		states &= ~(1ULL << i);
+	}
+
+	state->dirty = 0;
+
+	so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_FB]);
+	for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) {
+		if (!(samplers & (1 << i)))
+			continue;
+		so_emit_reloc_markers(nv30->nvws,
+				      state->hw[NV30_STATE_FRAGTEX0+i]);
+		samplers &= ~(1ULL << i);
+	}
+	so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_FRAGPROG]);
+	if (state->hw[NV30_STATE_VTXBUF] /*&& nv30->render_mode == HW*/)
+		so_emit_reloc_markers(nv30->nvws, state->hw[NV30_STATE_VTXBUF]);
+}
+
+boolean
+nv30_state_validate(struct nv30_context *nv30)
+{
+#if 0
+	boolean was_sw = nv30->fallback_swtnl ? TRUE : FALSE;
+
+	if (nv30->render_mode != HW) {
+		/* Don't even bother trying to go back to hw if none
+		 * of the states that caused swtnl previously have changed.
+		 */
+		if ((nv30->fallback_swtnl & nv30->dirty)
+				!= nv30->fallback_swtnl)
+			return FALSE;
+
+		/* Attempt to go to hwtnl again */
+		nv30->pipe.flush(&nv30->pipe, 0, NULL);
+		nv30->dirty |= (NV30_NEW_VIEWPORT |
+				NV30_NEW_VERTPROG |
+				NV30_NEW_ARRAYS);
+		nv30->render_mode = HW;
+	}
+#endif
+	nv30_state_do_validate(nv30, render_states);
+#if 0
+	if (nv30->fallback_swtnl || nv30->fallback_swrast)
+		return FALSE;
+	
+	if (was_sw)
+		NOUVEAU_ERR("swtnl->hw\n");
+#endif
+	return TRUE;
+}
diff --git a/src/gallium/drivers/nv30/nv30_state_fb.c b/src/gallium/drivers/nv30/nv30_state_fb.c
new file mode 100644
index 0000000000..77368cb205
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_state_fb.c
@@ -0,0 +1,144 @@
+#include "nv30_context.h"
+#include "nouveau/nouveau_util.h"
+
+static boolean
+nv30_state_framebuffer_validate(struct nv30_context *nv30)
+{
+	struct pipe_framebuffer_state *fb = &nv30->framebuffer;
+	struct pipe_surface *rt[2], *zeta = NULL;
+	uint32_t rt_enable, rt_format;
+	int i, colour_format = 0, zeta_format = 0;
+	struct nouveau_stateobj *so = so_new(64, 10);
+	unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM;
+	unsigned w = fb->width;
+	unsigned h = fb->height;
+	struct nv30_miptree *nv30mt;
+
+	rt_enable = 0;
+	for (i = 0; i < fb->nr_cbufs; i++) {
+		if (colour_format) {
+			assert(colour_format == fb->cbufs[i]->format);
+		} else {
+			colour_format = fb->cbufs[i]->format;
+			rt_enable |= (NV34TCL_RT_ENABLE_COLOR0 << i);
+			rt[i] = fb->cbufs[i];
+		}
+	}
+
+	if (rt_enable & NV34TCL_RT_ENABLE_COLOR1)
+		rt_enable |= NV34TCL_RT_ENABLE_MRT;
+
+	if (fb->zsbuf) {
+		zeta_format = fb->zsbuf->format;
+		zeta = fb->zsbuf;
+	}
+
+	if (!(rt[0]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
+		assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1)));
+		for (i = 1; i < fb->nr_cbufs; i++)
+			assert(!(rt[i]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR));
+
+		/* FIXME: NV34TCL_RT_FORMAT_LOG2_[WIDTH/HEIGHT] */
+		rt_format = NV34TCL_RT_FORMAT_TYPE_SWIZZLED |
+		log2i(fb->width) << 16 /*NV34TCL_RT_FORMAT_LOG2_WIDTH_SHIFT*/ |
+		log2i(fb->height) << 24 /*NV34TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT*/;
+	}
+	else
+		rt_format = NV34TCL_RT_FORMAT_TYPE_LINEAR;
+
+	switch (colour_format) {
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case 0:
+		rt_format |= NV34TCL_RT_FORMAT_COLOR_A8R8G8B8;
+		break;
+	case PIPE_FORMAT_R5G6B5_UNORM:
+		rt_format |= NV34TCL_RT_FORMAT_COLOR_R5G6B5;
+		break;
+	default:
+		assert(0);
+	}
+
+	switch (zeta_format) {
+	case PIPE_FORMAT_Z16_UNORM:
+		rt_format |= NV34TCL_RT_FORMAT_ZETA_Z16;
+		break;
+	case PIPE_FORMAT_Z24S8_UNORM:
+	case 0:
+		rt_format |= NV34TCL_RT_FORMAT_ZETA_Z24S8;
+		break;
+	default:
+		assert(0);
+	}
+
+	if (rt_enable & NV34TCL_RT_ENABLE_COLOR0) {
+		uint32_t pitch = rt[0]->stride;
+		if (zeta) {
+			pitch |= (zeta->stride << 16);
+		} else {
+			pitch |= (pitch << 16);
+		}
+
+		nv30mt = (struct nv30_miptree *)rt[0]->texture;
+		so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR0, 1);
+		so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR,
+			  nv30->nvws->channel->vram->handle,
+			  nv30->nvws->channel->gart->handle);
+		so_method(so, nv30->screen->rankine, NV34TCL_COLOR0_PITCH, 2);
+		so_data  (so, pitch);
+		so_reloc (so, nv30mt->buffer, rt[0]->offset, rt_flags |
+			  NOUVEAU_BO_LOW, 0, 0);
+	}
+
+	if (rt_enable & NV34TCL_RT_ENABLE_COLOR1) {
+		nv30mt = (struct nv30_miptree *)rt[1]->texture;
+		so_method(so, nv30->screen->rankine, NV34TCL_DMA_COLOR1, 1);
+		so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR,
+			  nv30->nvws->channel->vram->handle,
+			  nv30->nvws->channel->gart->handle);
+		so_method(so, nv30->screen->rankine, NV34TCL_COLOR1_OFFSET, 2);
+		so_reloc (so, nv30mt->buffer, rt[1]->offset, rt_flags |
+			  NOUVEAU_BO_LOW, 0, 0);
+		so_data  (so, rt[1]->stride);
+	}
+
+	if (zeta_format) {
+		nv30mt = (struct nv30_miptree *)zeta->texture;
+		so_method(so, nv30->screen->rankine, NV34TCL_DMA_ZETA, 1);
+		so_reloc (so, nv30mt->buffer, 0, rt_flags | NOUVEAU_BO_OR,
+			  nv30->nvws->channel->vram->handle,
+			  nv30->nvws->channel->gart->handle);
+		so_method(so, nv30->screen->rankine, NV34TCL_ZETA_OFFSET, 1);
+		so_reloc (so, nv30mt->buffer, zeta->offset, rt_flags |
+			  NOUVEAU_BO_LOW, 0, 0);
+		/* TODO: allocate LMA depth buffer */
+	}
+
+	so_method(so, nv30->screen->rankine, NV34TCL_RT_ENABLE, 1);
+	so_data  (so, rt_enable);
+	so_method(so, nv30->screen->rankine, NV34TCL_RT_HORIZ, 3);
+	so_data  (so, (w << 16) | 0);
+	so_data  (so, (h << 16) | 0);
+	so_data  (so, rt_format);
+	so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_HORIZ, 2);
+	so_data  (so, (w << 16) | 0);
+	so_data  (so, (h << 16) | 0);
+	so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2);
+	so_data  (so, ((w - 1) << 16) | 0);
+	so_data  (so, ((h - 1) << 16) | 0);
+	so_method(so, nv30->screen->rankine, 0x1d88, 1);
+	so_data  (so, (1 << 12) | h);
+	/* Wonder why this is needed, context should all be set to zero on init */
+	so_method(so, nv30->screen->rankine, NV34TCL_VIEWPORT_TX_ORIGIN, 1);
+	so_data  (so, 0);
+
+	so_ref(so, &nv30->state.hw[NV30_STATE_FB]);
+	return TRUE;
+}
+
+struct nv30_state_entry nv30_state_framebuffer = {
+	.validate = nv30_state_framebuffer_validate,
+	.dirty = {
+		.pipe = NV30_NEW_FB,
+		.hw = NV30_STATE_FB
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_state_rasterizer.c b/src/gallium/drivers/nv30/nv30_state_rasterizer.c
new file mode 100644
index 0000000000..6d1b60e043
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_state_rasterizer.c
@@ -0,0 +1,17 @@
+#include "nv30_context.h"
+
+static boolean
+nv30_state_rasterizer_validate(struct nv30_context *nv30)
+{
+	so_ref(nv30->rasterizer->so,
+	       &nv30->state.hw[NV30_STATE_RAST]);
+	return TRUE;
+}
+
+struct nv30_state_entry nv30_state_rasterizer = {
+	.validate = nv30_state_rasterizer_validate,
+	.dirty = {
+		.pipe = NV30_NEW_RAST,
+		.hw = NV30_STATE_RAST
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_state_scissor.c b/src/gallium/drivers/nv30/nv30_state_scissor.c
new file mode 100644
index 0000000000..1db9bc1795
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_state_scissor.c
@@ -0,0 +1,35 @@
+#include "nv30_context.h"
+
+static boolean
+nv30_state_scissor_validate(struct nv30_context *nv30)
+{
+	struct pipe_rasterizer_state *rast = &nv30->rasterizer->pipe;
+	struct pipe_scissor_state *s = &nv30->scissor;
+	struct nouveau_stateobj *so;
+
+	if (nv30->state.hw[NV30_STATE_SCISSOR] &&
+	    (rast->scissor == 0 && nv30->state.scissor_enabled == 0))
+		return FALSE;
+	nv30->state.scissor_enabled = rast->scissor;
+
+	so = so_new(3, 0);
+	so_method(so, nv30->screen->rankine, NV34TCL_SCISSOR_HORIZ, 2);
+	if (nv30->state.scissor_enabled) {
+		so_data  (so, ((s->maxx - s->minx) << 16) | s->minx);
+		so_data  (so, ((s->maxy - s->miny) << 16) | s->miny);
+	} else {
+		so_data  (so, 4096 << 16);
+		so_data  (so, 4096 << 16);
+	}
+
+	so_ref(so, &nv30->state.hw[NV30_STATE_SCISSOR]);
+	return TRUE;
+}
+
+struct nv30_state_entry nv30_state_scissor = {
+	.validate = nv30_state_scissor_validate,
+	.dirty = {
+		.pipe = NV30_NEW_SCISSOR | NV30_NEW_RAST,
+		.hw = NV30_STATE_SCISSOR
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_state_stipple.c b/src/gallium/drivers/nv30/nv30_state_stipple.c
new file mode 100644
index 0000000000..41b42813b4
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_state_stipple.c
@@ -0,0 +1,39 @@
+#include "nv30_context.h"
+
+static boolean
+nv30_state_stipple_validate(struct nv30_context *nv30)
+{
+	struct pipe_rasterizer_state *rast = &nv30->rasterizer->pipe;
+	struct nouveau_grobj *rankine = nv30->screen->rankine;
+	struct nouveau_stateobj *so;
+
+	if (nv30->state.hw[NV30_STATE_STIPPLE] &&
+	   (rast->poly_stipple_enable == 0 && nv30->state.stipple_enabled == 0))
+		return FALSE;
+
+	if (rast->poly_stipple_enable) {
+		unsigned i;
+
+		so = so_new(35, 0);
+		so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1);
+		so_data  (so, 1);
+		so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_PATTERN(0), 32);
+		for (i = 0; i < 32; i++)
+			so_data(so, nv30->stipple[i]);
+	} else {
+		so = so_new(2, 0);
+		so_method(so, rankine, NV34TCL_POLYGON_STIPPLE_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_ref(so, &nv30->state.hw[NV30_STATE_STIPPLE]);
+	return TRUE;
+}
+
+struct nv30_state_entry nv30_state_stipple = {
+	.validate = nv30_state_stipple_validate,
+	.dirty = {
+		.pipe = NV30_NEW_STIPPLE | NV30_NEW_RAST,
+		.hw = NV30_STATE_STIPPLE,
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_state_viewport.c b/src/gallium/drivers/nv30/nv30_state_viewport.c
new file mode 100644
index 0000000000..951d40ebfd
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_state_viewport.c
@@ -0,0 +1,70 @@
+#include "nv30_context.h"
+
+static boolean
+nv30_state_viewport_validate(struct nv30_context *nv30)
+{
+	struct pipe_viewport_state *vpt = &nv30->viewport;
+	struct nouveau_stateobj *so;
+	unsigned bypass;
+
+	if (/*nv30->render_mode == HW &&*/ !nv30->rasterizer->pipe.bypass_clipping)
+		bypass = 0;
+	else
+		bypass = 1;
+
+	if (nv30->state.hw[NV30_STATE_VIEWPORT] &&
+	    (bypass || !(nv30->dirty & NV30_NEW_VIEWPORT)) &&
+	    nv30->state.viewport_bypass == bypass)
+		return FALSE;
+	nv30->state.viewport_bypass = bypass;
+
+	so = so_new(11, 0);
+	if (!bypass) {
+		so_method(so, nv30->screen->rankine,
+			  NV34TCL_VIEWPORT_TRANSLATE_X, 8);
+		so_data  (so, fui(vpt->translate[0]));
+		so_data  (so, fui(vpt->translate[1]));
+		so_data  (so, fui(vpt->translate[2]));
+		so_data  (so, fui(vpt->translate[3]));
+		so_data  (so, fui(vpt->scale[0]));
+		so_data  (so, fui(vpt->scale[1]));
+		so_data  (so, fui(vpt->scale[2]));
+		so_data  (so, fui(vpt->scale[3]));
+/*		so_method(so, nv30->screen->rankine, 0x1d78, 1);
+		so_data  (so, 1);
+*/	} else {
+		so_method(so, nv30->screen->rankine,
+			  NV34TCL_VIEWPORT_TRANSLATE_X, 8);
+		so_data  (so, fui(0.0));
+		so_data  (so, fui(0.0));
+		so_data  (so, fui(0.0));
+		so_data  (so, fui(0.0));
+		so_data  (so, fui(1.0));
+		so_data  (so, fui(1.0));
+		so_data  (so, fui(1.0));
+		so_data  (so, fui(0.0));
+		/* Not entirely certain what this is yet.  The DDX uses this
+		 * value also as it fixes rendering when you pass
+		 * pre-transformed vertices to the GPU.  My best gusss is that
+		 * this bypasses some culling/clipping stage.  Might be worth
+		 * noting that points/lines are uneffected by whatever this
+		 * value fixes, only filled polygons are effected.
+		 */
+/*		so_method(so, nv30->screen->rankine, 0x1d78, 1);
+		so_data  (so, 0x110);
+*/	}
+	/* TODO/FIXME: never saw value 0x0110 in renouveau dumps, only 0x0001 */
+	so_method(so, nv30->screen->rankine, 0x1d78, 1);
+	so_data  (so, 1);
+
+	so_ref(so, &nv30->state.hw[NV30_STATE_VIEWPORT]);
+	return TRUE;
+}
+
+struct nv30_state_entry nv30_state_viewport = {
+	.validate = nv30_state_viewport_validate,
+	.dirty = {
+		.pipe = NV30_NEW_VIEWPORT | NV30_NEW_RAST,
+		.hw = NV30_STATE_VIEWPORT
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_state_zsa.c b/src/gallium/drivers/nv30/nv30_state_zsa.c
new file mode 100644
index 0000000000..0940b7269b
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_state_zsa.c
@@ -0,0 +1,17 @@
+#include "nv30_context.h"
+
+static boolean
+nv30_state_zsa_validate(struct nv30_context *nv30)
+{
+	so_ref(nv30->zsa->so,
+	       &nv30->state.hw[NV30_STATE_ZSA]);
+	return TRUE;
+}
+
+struct nv30_state_entry nv30_state_zsa = {
+	.validate = nv30_state_zsa_validate,
+	.dirty = {
+		.pipe = NV30_NEW_ZSA,
+		.hw = NV30_STATE_ZSA
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_surface.c b/src/gallium/drivers/nv30/nv30_surface.c
new file mode 100644
index 0000000000..0f8dc12045
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_surface.c
@@ -0,0 +1,72 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "nv30_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+
+static void
+nv30_surface_copy(struct pipe_context *pipe, boolean do_flip,
+		  struct pipe_surface *dest, unsigned destx, unsigned desty,
+		  struct pipe_surface *src, unsigned srcx, unsigned srcy,
+		  unsigned width, unsigned height)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nv04_surface_2d *eng2d = nv30->screen->eng2d;
+
+	if (do_flip) {
+		desty += height;
+		while (height--) {
+			eng2d->copy(eng2d, dest, destx, desty--, src,
+				    srcx, srcy++, width, 1);
+		}
+		return;
+	}
+
+	eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height);
+}
+
+static void
+nv30_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
+		  unsigned destx, unsigned desty, unsigned width,
+		  unsigned height, unsigned value)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nv04_surface_2d *eng2d = nv30->screen->eng2d;
+
+	eng2d->fill(eng2d, dest, destx, desty, width, height, value);
+}
+
+void
+nv30_init_surface_functions(struct nv30_context *nv30)
+{
+	nv30->pipe.surface_copy = nv30_surface_copy;
+	nv30->pipe.surface_fill = nv30_surface_fill;
+}
diff --git a/src/gallium/drivers/nv30/nv30_vbo.c b/src/gallium/drivers/nv30/nv30_vbo.c
new file mode 100644
index 0000000000..2d6d48ac16
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_vbo.c
@@ -0,0 +1,556 @@
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+
+#include "nv30_context.h"
+#include "nv30_state.h"
+
+#include "nouveau/nouveau_channel.h"
+#include "nouveau/nouveau_pushbuf.h"
+#include "nouveau/nouveau_util.h"
+
+#define FORCE_SWTNL 0
+
+static INLINE int
+nv30_vbo_format_to_hw(enum pipe_format pipe, unsigned *fmt, unsigned *ncomp)
+{
+	switch (pipe) {
+	case PIPE_FORMAT_R32_FLOAT:
+	case PIPE_FORMAT_R32G32_FLOAT:
+	case PIPE_FORMAT_R32G32B32_FLOAT:
+	case PIPE_FORMAT_R32G32B32A32_FLOAT:
+		*fmt = NV34TCL_VTXFMT_TYPE_FLOAT;
+		break;
+	case PIPE_FORMAT_R8_UNORM:
+	case PIPE_FORMAT_R8G8_UNORM:
+	case PIPE_FORMAT_R8G8B8_UNORM:
+	case PIPE_FORMAT_R8G8B8A8_UNORM:
+		*fmt = NV34TCL_VTXFMT_TYPE_UBYTE;
+		break;
+	case PIPE_FORMAT_R16_SSCALED:
+	case PIPE_FORMAT_R16G16_SSCALED:
+	case PIPE_FORMAT_R16G16B16_SSCALED:
+	case PIPE_FORMAT_R16G16B16A16_SSCALED:
+		*fmt = NV34TCL_VTXFMT_TYPE_USHORT;
+		break;
+	default:
+		NOUVEAU_ERR("Unknown format %s\n", pf_name(pipe));
+		return 1;
+	}
+
+	switch (pipe) {
+	case PIPE_FORMAT_R8_UNORM:
+	case PIPE_FORMAT_R32_FLOAT:
+	case PIPE_FORMAT_R16_SSCALED:
+		*ncomp = 1;
+		break;
+	case PIPE_FORMAT_R8G8_UNORM:
+	case PIPE_FORMAT_R32G32_FLOAT:
+	case PIPE_FORMAT_R16G16_SSCALED:
+		*ncomp = 2;
+		break;
+	case PIPE_FORMAT_R8G8B8_UNORM:
+	case PIPE_FORMAT_R32G32B32_FLOAT:
+	case PIPE_FORMAT_R16G16B16_SSCALED:
+		*ncomp = 3;
+		break;
+	case PIPE_FORMAT_R8G8B8A8_UNORM:
+	case PIPE_FORMAT_R32G32B32A32_FLOAT:
+	case PIPE_FORMAT_R16G16B16A16_SSCALED:
+		*ncomp = 4;
+		break;
+	default:
+		NOUVEAU_ERR("Unknown format %s\n", pf_name(pipe));
+		return 1;
+	}
+
+	return 0;
+}
+
+static boolean
+nv30_vbo_set_idxbuf(struct nv30_context *nv30, struct pipe_buffer *ib,
+		    unsigned ib_size)
+{
+	struct pipe_screen *pscreen = &nv30->screen->pipe;
+	unsigned type;
+
+	if (!ib) {
+		nv30->idxbuf = NULL;
+		nv30->idxbuf_format = 0xdeadbeef;
+		return FALSE;
+	}
+
+	if (!pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF) || ib_size == 1)
+		return FALSE;
+
+	switch (ib_size) {
+	case 2:
+		type = NV34TCL_IDXBUF_FORMAT_TYPE_U16;
+		break;
+	case 4:
+		type = NV34TCL_IDXBUF_FORMAT_TYPE_U32;
+		break;
+	default:
+		return FALSE;
+	}
+
+	if (ib != nv30->idxbuf ||
+	    type != nv30->idxbuf_format) {
+		nv30->dirty |= NV30_NEW_ARRAYS;
+		nv30->idxbuf = ib;
+		nv30->idxbuf_format = type;
+	}
+
+	return TRUE;
+}
+
+static boolean
+nv30_vbo_static_attrib(struct nv30_context *nv30, struct nouveau_stateobj *so,
+		       int attrib, struct pipe_vertex_element *ve,
+		       struct pipe_vertex_buffer *vb)
+{
+	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct nouveau_grobj *rankine = nv30->screen->rankine;
+	unsigned type, ncomp;
+	void *map;
+
+	if (nv30_vbo_format_to_hw(ve->src_format, &type, &ncomp))
+		return FALSE;
+
+	map  = ws->buffer_map(ws, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+	map += vb->buffer_offset + ve->src_offset;
+
+	switch (type) {
+	case NV34TCL_VTXFMT_TYPE_FLOAT:
+	{
+		float *v = map;
+
+		switch (ncomp) {
+		case 4:
+			so_method(so, rankine, NV34TCL_VTX_ATTR_4F_X(attrib), 4);
+			so_data  (so, fui(v[0]));
+			so_data  (so, fui(v[1]));
+			so_data  (so, fui(v[2]));
+			so_data  (so, fui(v[3]));
+			break;
+		case 3:
+			so_method(so, rankine, NV34TCL_VTX_ATTR_3F_X(attrib), 3);
+			so_data  (so, fui(v[0]));
+			so_data  (so, fui(v[1]));
+			so_data  (so, fui(v[2]));
+			break;
+		case 2:
+			so_method(so, rankine, NV34TCL_VTX_ATTR_2F_X(attrib), 2);
+			so_data  (so, fui(v[0]));
+			so_data  (so, fui(v[1]));
+			break;
+		case 1:
+			so_method(so, rankine, NV34TCL_VTX_ATTR_1F(attrib), 1);
+			so_data  (so, fui(v[0]));
+			break;
+		default:
+			ws->buffer_unmap(ws, vb->buffer);
+			return FALSE;
+		}
+	}
+		break;
+	default:
+		ws->buffer_unmap(ws, vb->buffer);
+		return FALSE;
+	}
+
+	ws->buffer_unmap(ws, vb->buffer);
+
+	return TRUE;
+}
+
+boolean
+nv30_draw_arrays(struct pipe_context *pipe,
+		 unsigned mode, unsigned start, unsigned count)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nouveau_channel *chan = nv30->nvws->channel;
+	unsigned restart = 0;
+
+	nv30_vbo_set_idxbuf(nv30, NULL, 0);
+	if (FORCE_SWTNL || !nv30_state_validate(nv30)) {
+		/*return nv30_draw_elements_swtnl(pipe, NULL, 0,
+						mode, start, count);*/
+		return FALSE;
+	}
+
+	while (count) {
+		unsigned vc, nr;
+
+		nv30_state_emit(nv30);
+
+		vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 256,
+					mode, start, count, &restart);
+		if (!vc) {
+			FIRE_RING(NULL);
+			continue;
+		}
+
+		BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (nvgl_primitive(mode));
+
+		nr = (vc & 0xff);
+		if (nr) {
+			BEGIN_RING(rankine, NV34TCL_VB_VERTEX_BATCH, 1);
+			OUT_RING  (((nr - 1) << 24) | start);
+			start += nr;
+		}
+
+		nr = vc >> 8;
+		while (nr) {
+			unsigned push = nr > 2047 ? 2047 : nr;
+
+			nr -= push;
+
+			BEGIN_RING_NI(rankine, NV34TCL_VB_VERTEX_BATCH, push);
+			while (push--) {
+				OUT_RING(((0x100 - 1) << 24) | start);
+				start += 0x100;
+			}
+		}
+
+		BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (0);
+
+		count -= vc;
+		start = restart;
+	}
+
+	pipe->flush(pipe, 0, NULL);
+	return TRUE;
+}
+
+static INLINE void
+nv30_draw_elements_u08(struct nv30_context *nv30, void *ib,
+		       unsigned mode, unsigned start, unsigned count)
+{
+	struct nouveau_channel *chan = nv30->nvws->channel;
+
+	while (count) {
+		uint8_t *elts = (uint8_t *)ib + start;
+		unsigned vc, push, restart = 0;
+
+		nv30_state_emit(nv30);
+
+		vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 2,
+					mode, start, count, &restart);
+		if (vc == 0) {
+			FIRE_RING(NULL);
+			continue;
+		}
+		count -= vc;
+
+		BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (nvgl_primitive(mode));
+
+		if (vc & 1) {
+			BEGIN_RING(rankine, NV34TCL_VB_ELEMENT_U32, 1);
+			OUT_RING  (elts[0]);
+			elts++; vc--;
+		}
+
+		while (vc) {
+			unsigned i;
+
+			push = MIN2(vc, 2047 * 2);
+
+			BEGIN_RING_NI(rankine, NV34TCL_VB_ELEMENT_U16, push >> 1);
+			for (i = 0; i < push; i+=2)
+				OUT_RING((elts[i+1] << 16) | elts[i]);
+
+			vc -= push;
+			elts += push;
+		}
+
+		BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (0);
+
+		start = restart;
+	}
+}
+
+static INLINE void
+nv30_draw_elements_u16(struct nv30_context *nv30, void *ib,
+		       unsigned mode, unsigned start, unsigned count)
+{
+	struct nouveau_channel *chan = nv30->nvws->channel;
+
+	while (count) {
+		uint16_t *elts = (uint16_t *)ib + start;
+		unsigned vc, push, restart = 0;
+
+		nv30_state_emit(nv30);
+
+		vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 2,
+					mode, start, count, &restart);
+		if (vc == 0) {
+			FIRE_RING(NULL);
+			continue;
+		}
+		count -= vc;
+
+		BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (nvgl_primitive(mode));
+
+		if (vc & 1) {
+			BEGIN_RING(rankine, NV34TCL_VB_ELEMENT_U32, 1);
+			OUT_RING  (elts[0]);
+			elts++; vc--;
+		}
+
+		while (vc) {
+			unsigned i;
+
+			push = MIN2(vc, 2047 * 2);
+
+			BEGIN_RING_NI(rankine, NV34TCL_VB_ELEMENT_U16, push >> 1);
+			for (i = 0; i < push; i+=2)
+				OUT_RING((elts[i+1] << 16) | elts[i]);
+
+			vc -= push;
+			elts += push;
+		}
+
+		BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (0);
+
+		start = restart;
+	}
+}
+
+static INLINE void
+nv30_draw_elements_u32(struct nv30_context *nv30, void *ib,
+		       unsigned mode, unsigned start, unsigned count)
+{
+	struct nouveau_channel *chan = nv30->nvws->channel;
+
+	while (count) {
+		uint32_t *elts = (uint32_t *)ib + start;
+		unsigned vc, push, restart = 0;
+
+		nv30_state_emit(nv30);
+
+		vc = nouveau_vbuf_split(chan->pushbuf->remaining, 5, 1,
+					mode, start, count, &restart);
+		if (vc == 0) {
+			FIRE_RING(NULL);
+			continue;
+		}
+		count -= vc;
+
+		BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (nvgl_primitive(mode));
+
+		while (vc) {
+			push = MIN2(vc, 2047);
+
+			BEGIN_RING_NI(rankine, NV34TCL_VB_ELEMENT_U32, push);
+			OUT_RINGp    (elts, push);
+
+			vc -= push;
+			elts += push;
+		}
+
+		BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (0);
+
+		start = restart;
+	}
+}
+
+static boolean
+nv30_draw_elements_inline(struct pipe_context *pipe,
+			  struct pipe_buffer *ib, unsigned ib_size,
+			  unsigned mode, unsigned start, unsigned count)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct pipe_winsys *ws = pipe->winsys;
+	void *map;
+
+	map = ws->buffer_map(ws, ib, PIPE_BUFFER_USAGE_CPU_READ);
+	if (!ib) {
+		NOUVEAU_ERR("failed mapping ib\n");
+		return FALSE;
+	}
+
+	switch (ib_size) {
+	case 1:
+		nv30_draw_elements_u08(nv30, map, mode, start, count);
+		break;
+	case 2:
+		nv30_draw_elements_u16(nv30, map, mode, start, count);
+		break;
+	case 4:
+		nv30_draw_elements_u32(nv30, map, mode, start, count);
+		break;
+	default:
+		NOUVEAU_ERR("invalid idxbuf fmt %d\n", ib_size);
+		break;
+	}
+
+	ws->buffer_unmap(ws, ib);
+	return TRUE;
+}
+
+static boolean
+nv30_draw_elements_vbo(struct pipe_context *pipe,
+		       unsigned mode, unsigned start, unsigned count)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	struct nouveau_channel *chan = nv30->nvws->channel;
+	unsigned restart = 0;
+
+	while (count) {
+		unsigned nr, vc;
+
+		nv30_state_emit(nv30);
+
+		vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 256,
+					mode, start, count, &restart);
+		if (!vc) {
+			FIRE_RING(NULL);
+			continue;
+		}
+		
+		BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (nvgl_primitive(mode));
+
+		nr = (vc & 0xff);
+		if (nr) {
+			BEGIN_RING(rankine, NV34TCL_VB_INDEX_BATCH, 1);
+			OUT_RING  (((nr - 1) << 24) | start);
+			start += nr;
+		}
+
+		nr = vc >> 8;
+		while (nr) {
+			unsigned push = nr > 2047 ? 2047 : nr;
+
+			nr -= push;
+
+			BEGIN_RING_NI(rankine, NV34TCL_VB_INDEX_BATCH, push);
+			while (push--) {
+				OUT_RING(((0x100 - 1) << 24) | start);
+				start += 0x100;
+			}
+		}
+
+		BEGIN_RING(rankine, NV34TCL_VERTEX_BEGIN_END, 1);
+		OUT_RING  (0);
+
+		count -= vc;
+		start = restart;
+	}
+
+	return TRUE;
+}
+
+boolean
+nv30_draw_elements(struct pipe_context *pipe,
+		   struct pipe_buffer *indexBuffer, unsigned indexSize,
+		   unsigned mode, unsigned start, unsigned count)
+{
+	struct nv30_context *nv30 = nv30_context(pipe);
+	boolean idxbuf;
+
+	idxbuf = nv30_vbo_set_idxbuf(nv30, indexBuffer, indexSize);
+	if (FORCE_SWTNL || !nv30_state_validate(nv30)) {
+		/*return nv30_draw_elements_swtnl(pipe, NULL, 0,
+						mode, start, count);*/
+		return FALSE;	
+	}
+
+	if (idxbuf) {
+		nv30_draw_elements_vbo(pipe, mode, start, count);
+	} else {
+		nv30_draw_elements_inline(pipe, indexBuffer, indexSize,
+					  mode, start, count);
+	}
+
+	pipe->flush(pipe, 0, NULL);
+	return TRUE;
+}
+
+static boolean
+nv30_vbo_validate(struct nv30_context *nv30)
+{
+	struct nouveau_stateobj *vtxbuf, *vtxfmt, *sattr = NULL;
+	struct nouveau_grobj *rankine = nv30->screen->rankine;
+	struct pipe_buffer *ib = nv30->idxbuf;
+	unsigned ib_format = nv30->idxbuf_format;
+	unsigned vb_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD;
+	int hw;
+
+	if (nv30->edgeflags) {
+		/*nv30->fallback_swtnl |= NV30_NEW_ARRAYS;*/
+		return FALSE;
+	}
+
+	vtxbuf = so_new(20, 18);
+	so_method(vtxbuf, rankine, NV34TCL_VTXBUF_ADDRESS(0), nv30->vtxelt_nr);
+	vtxfmt = so_new(17, 0);
+	so_method(vtxfmt, rankine, NV34TCL_VTXFMT(0), nv30->vtxelt_nr);
+
+	for (hw = 0; hw < nv30->vtxelt_nr; hw++) {
+		struct pipe_vertex_element *ve;
+		struct pipe_vertex_buffer *vb;
+		unsigned type, ncomp;
+
+		ve = &nv30->vtxelt[hw];
+		vb = &nv30->vtxbuf[ve->vertex_buffer_index];
+
+		if (!vb->stride) {
+			if (!sattr)
+				sattr = so_new(16 * 5, 0);
+
+			if (nv30_vbo_static_attrib(nv30, sattr, hw, ve, vb)) {
+				so_data(vtxbuf, 0);
+				so_data(vtxfmt, NV34TCL_VTXFMT_TYPE_FLOAT);
+				continue;
+			}
+		}
+
+		if (nv30_vbo_format_to_hw(ve->src_format, &type, &ncomp)) {
+			/*nv30->fallback_swtnl |= NV30_NEW_ARRAYS;*/
+			so_ref(NULL, &vtxbuf);
+			so_ref(NULL, &vtxfmt);
+			return FALSE;
+		}
+
+		so_reloc(vtxbuf, vb->buffer, vb->buffer_offset + ve->src_offset,
+			 vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
+			 0, NV34TCL_VTXBUF_ADDRESS_DMA1);
+		so_data (vtxfmt, ((vb->stride << NV34TCL_VTXFMT_STRIDE_SHIFT) |
+				  (ncomp << NV34TCL_VTXFMT_SIZE_SHIFT) | type));
+	}
+
+	if (ib) {
+		so_method(vtxbuf, rankine, NV34TCL_IDXBUF_ADDRESS, 2);
+		so_reloc (vtxbuf, ib, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
+		so_reloc (vtxbuf, ib, ib_format, vb_flags | NOUVEAU_BO_OR,
+			  0, NV34TCL_IDXBUF_FORMAT_DMA1);
+	}
+
+	so_method(vtxbuf, rankine, 0x1710, 1);
+	so_data  (vtxbuf, 0);
+
+	so_ref(vtxbuf, &nv30->state.hw[NV30_STATE_VTXBUF]);
+	nv30->state.dirty |= (1ULL << NV30_STATE_VTXBUF);
+	so_ref(vtxfmt, &nv30->state.hw[NV30_STATE_VTXFMT]);
+	nv30->state.dirty |= (1ULL << NV30_STATE_VTXFMT);
+	so_ref(sattr, &nv30->state.hw[NV30_STATE_VTXATTR]);
+	nv30->state.dirty |= (1ULL << NV30_STATE_VTXATTR);
+	return FALSE;
+}
+
+struct nv30_state_entry nv30_state_vbo = {
+	.validate = nv30_vbo_validate,
+	.dirty = {
+		.pipe = NV30_NEW_ARRAYS,
+		.hw = 0,
+	}
+};
diff --git a/src/gallium/drivers/nv30/nv30_vertprog.c b/src/gallium/drivers/nv30/nv30_vertprog.c
new file mode 100644
index 0000000000..d262725057
--- /dev/null
+++ b/src/gallium/drivers/nv30/nv30_vertprog.c
@@ -0,0 +1,838 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "nv30_context.h"
+#include "nv30_state.h"
+
+/* TODO (at least...):
+ *  1. Indexed consts  + ARL
+ *  2. Arb. swz/negation
+ *  3. NV_vp11, NV_vp2, NV_vp3 features
+ *       - extra arith opcodes
+ *       - branching
+ *       - texture sampling
+ *       - indexed attribs
+ *       - indexed results
+ *  4. bugs
+ */
+
+#define SWZ_X 0
+#define SWZ_Y 1
+#define SWZ_Z 2
+#define SWZ_W 3
+#define MASK_X 8
+#define MASK_Y 4
+#define MASK_Z 2
+#define MASK_W 1
+#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W)
+#define DEF_SCALE 0
+#define DEF_CTEST 0
+#include "nv30_shader.h"
+
+#define swz(s,x,y,z,w) nv30_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w)
+#define neg(s) nv30_sr_neg((s))
+#define abs(s) nv30_sr_abs((s))
+
+struct nv30_vpc {
+	struct nv30_vertex_program *vp;
+
+	struct nv30_vertex_program_exec *vpi;
+
+	unsigned output_map[PIPE_MAX_SHADER_OUTPUTS];
+
+	int high_temp;
+	int temp_temp_count;
+
+	struct nv30_sreg *imm;
+	unsigned nr_imm;
+};
+
+static struct nv30_sreg
+temp(struct nv30_vpc *vpc)
+{
+	int idx;
+
+	idx  = vpc->temp_temp_count++;
+	idx += vpc->high_temp + 1;
+	return nv30_sr(NV30SR_TEMP, idx);
+}
+
+static struct nv30_sreg
+constant(struct nv30_vpc *vpc, int pipe, float x, float y, float z, float w)
+{
+	struct nv30_vertex_program *vp = vpc->vp;
+	struct nv30_vertex_program_data *vpd;
+	int idx;
+
+	if (pipe >= 0) {
+		for (idx = 0; idx < vp->nr_consts; idx++) {
+			if (vp->consts[idx].index == pipe)
+				return nv30_sr(NV30SR_CONST, idx);
+		}
+	}
+
+	idx = vp->nr_consts++;
+	vp->consts = realloc(vp->consts, sizeof(*vpd) * vp->nr_consts);
+	vpd = &vp->consts[idx];
+
+	vpd->index = pipe;
+	vpd->value[0] = x;
+	vpd->value[1] = y;
+	vpd->value[2] = z;
+	vpd->value[3] = w;
+	return nv30_sr(NV30SR_CONST, idx);
+}
+
+#define arith(cc,s,o,d,m,s0,s1,s2) \
+	nv30_vp_arith((cc), (s), NV30_VP_INST_##o, (d), (m), (s0), (s1), (s2))
+
+static void
+emit_src(struct nv30_vpc *vpc, uint32_t *hw, int pos, struct nv30_sreg src)
+{
+	struct nv30_vertex_program *vp = vpc->vp;
+	uint32_t sr = 0;
+
+	switch (src.type) {
+	case NV30SR_TEMP:
+		sr |= (NV30_VP_SRC_REG_TYPE_TEMP << NV30_VP_SRC_REG_TYPE_SHIFT);
+		sr |= (src.index << NV30_VP_SRC_TEMP_SRC_SHIFT);
+		break;
+	case NV30SR_INPUT:
+		sr |= (NV30_VP_SRC_REG_TYPE_INPUT <<
+		       NV30_VP_SRC_REG_TYPE_SHIFT);
+		vp->ir |= (1 << src.index);
+		hw[1] |= (src.index << NV30_VP_INST_INPUT_SRC_SHIFT);
+		break;
+	case NV30SR_CONST:
+		sr |= (NV30_VP_SRC_REG_TYPE_CONST <<
+		       NV30_VP_SRC_REG_TYPE_SHIFT);
+		assert(vpc->vpi->const_index == -1 ||
+		       vpc->vpi->const_index == src.index);
+		vpc->vpi->const_index = src.index;
+		break;
+	case NV30SR_NONE:
+		sr |= (NV30_VP_SRC_REG_TYPE_INPUT <<
+		       NV30_VP_SRC_REG_TYPE_SHIFT);
+		break;
+	default:
+		assert(0);
+	}
+
+	if (src.negate)
+		sr |= NV30_VP_SRC_NEGATE;
+
+	if (src.abs)
+		hw[0] |= (1 << (21 + pos));
+
+	sr |= ((src.swz[0] << NV30_VP_SRC_SWZ_X_SHIFT) |
+	       (src.swz[1] << NV30_VP_SRC_SWZ_Y_SHIFT) |
+	       (src.swz[2] << NV30_VP_SRC_SWZ_Z_SHIFT) |
+	       (src.swz[3] << NV30_VP_SRC_SWZ_W_SHIFT));
+
+/*
+ * |VVV|
+ * d�.�b
+ *  \u/
+ *
+ */
+
+	switch (pos) {
+	case 0:
+		hw[1] |= ((sr & NV30_VP_SRC0_HIGH_MASK) >>
+			  NV30_VP_SRC0_HIGH_SHIFT) << NV30_VP_INST_SRC0H_SHIFT;
+		hw[2] |= (sr & NV30_VP_SRC0_LOW_MASK) <<
+			  NV30_VP_INST_SRC0L_SHIFT;
+		break;
+	case 1:
+		hw[2] |= sr << NV30_VP_INST_SRC1_SHIFT;
+		break;
+	case 2:
+		hw[2] |= ((sr & NV30_VP_SRC2_HIGH_MASK) >>
+			  NV30_VP_SRC2_HIGH_SHIFT) << NV30_VP_INST_SRC2H_SHIFT;
+		hw[3] |= (sr & NV30_VP_SRC2_LOW_MASK) <<
+			  NV30_VP_INST_SRC2L_SHIFT;
+		break;
+	default:
+		assert(0);
+	}
+}
+
+static void
+emit_dst(struct nv30_vpc *vpc, uint32_t *hw, int slot, struct nv30_sreg dst)
+{
+	struct nv30_vertex_program *vp = vpc->vp;
+
+	switch (dst.type) {
+	case NV30SR_TEMP:
+		hw[0] |= (dst.index << NV30_VP_INST_DEST_TEMP_ID_SHIFT);
+		break;
+	case NV30SR_OUTPUT:
+		switch (dst.index) {
+		case NV30_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break;
+		case NV30_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break;
+		case NV30_VP_INST_DEST_BFC0 : vp->or |= (1 << 2); break;
+		case NV30_VP_INST_DEST_BFC1 : vp->or |= (1 << 3); break;
+		case NV30_VP_INST_DEST_FOGC : vp->or |= (1 << 4); break;
+		case NV30_VP_INST_DEST_PSZ  : vp->or |= (1 << 5); break;
+		case NV30_VP_INST_DEST_TC(0): vp->or |= (1 << 14); break;
+		case NV30_VP_INST_DEST_TC(1): vp->or |= (1 << 15); break;
+		case NV30_VP_INST_DEST_TC(2): vp->or |= (1 << 16); break;
+		case NV30_VP_INST_DEST_TC(3): vp->or |= (1 << 17); break;
+		case NV30_VP_INST_DEST_TC(4): vp->or |= (1 << 18); break;
+		case NV30_VP_INST_DEST_TC(5): vp->or |= (1 << 19); break;
+		case NV30_VP_INST_DEST_TC(6): vp->or |= (1 << 20); break;
+		case NV30_VP_INST_DEST_TC(7): vp->or |= (1 << 21); break;
+		default:
+			break;
+		}
+
+		hw[3] |= (dst.index << NV30_VP_INST_DEST_SHIFT);
+		hw[0] |= NV30_VP_INST_VEC_DEST_TEMP_MASK | (1<<20);
+
+		/*XXX: no way this is entirely correct, someone needs to
+		 *     figure out what exactly it is.
+		 */
+		hw[3] |= 0x800;
+		break;
+	default:
+		assert(0);
+	}
+}
+
+static void
+nv30_vp_arith(struct nv30_vpc *vpc, int slot, int op,
+	      struct nv30_sreg dst, int mask,
+	      struct nv30_sreg s0, struct nv30_sreg s1,
+	      struct nv30_sreg s2)
+{
+	struct nv30_vertex_program *vp = vpc->vp;
+	uint32_t *hw;
+
+	vp->insns = realloc(vp->insns, ++vp->nr_insns * sizeof(*vpc->vpi));
+	vpc->vpi = &vp->insns[vp->nr_insns - 1];
+	memset(vpc->vpi, 0, sizeof(*vpc->vpi));
+	vpc->vpi->const_index = -1;
+
+	hw = vpc->vpi->data;
+
+	hw[0] |= (NV30_VP_INST_COND_TR << NV30_VP_INST_COND_SHIFT);
+	hw[0] |= ((0 << NV30_VP_INST_COND_SWZ_X_SHIFT) |
+		  (1 << NV30_VP_INST_COND_SWZ_Y_SHIFT) |
+		  (2 << NV30_VP_INST_COND_SWZ_Z_SHIFT) |
+		  (3 << NV30_VP_INST_COND_SWZ_W_SHIFT));
+
+	hw[1] |= (op << NV30_VP_INST_VEC_OPCODE_SHIFT);
+//	hw[3] |= NV30_VP_INST_SCA_DEST_TEMP_MASK;
+//	hw[3] |= (mask << NV30_VP_INST_VEC_WRITEMASK_SHIFT);
+
+	if (dst.type == NV30SR_OUTPUT) {
+		if (slot)
+			hw[3] |= (mask << NV30_VP_INST_SDEST_WRITEMASK_SHIFT);
+		else
+			hw[3] |= (mask << NV30_VP_INST_VDEST_WRITEMASK_SHIFT);
+	} else {
+		if (slot)
+			hw[3] |= (mask << NV30_VP_INST_STEMP_WRITEMASK_SHIFT);
+		else
+			hw[3] |= (mask << NV30_VP_INST_VTEMP_WRITEMASK_SHIFT);
+	}
+
+	emit_dst(vpc, hw, slot, dst);
+	emit_src(vpc, hw, 0, s0);
+	emit_src(vpc, hw, 1, s1);
+	emit_src(vpc, hw, 2, s2);
+}
+
+static INLINE struct nv30_sreg
+tgsi_src(struct nv30_vpc *vpc, const struct tgsi_full_src_register *fsrc) {
+	struct nv30_sreg src;
+
+	switch (fsrc->SrcRegister.File) {
+	case TGSI_FILE_INPUT:
+		src = nv30_sr(NV30SR_INPUT, fsrc->SrcRegister.Index);
+		break;
+	case TGSI_FILE_CONSTANT:
+		src = constant(vpc, fsrc->SrcRegister.Index, 0, 0, 0, 0);
+		break;
+	case TGSI_FILE_IMMEDIATE:
+		src = vpc->imm[fsrc->SrcRegister.Index];
+		break;
+	case TGSI_FILE_TEMPORARY:
+		if (vpc->high_temp < fsrc->SrcRegister.Index)
+			vpc->high_temp = fsrc->SrcRegister.Index;
+		src = nv30_sr(NV30SR_TEMP, fsrc->SrcRegister.Index);
+		break;
+	default:
+		NOUVEAU_ERR("bad src file\n");
+		break;
+	}
+
+	src.abs = fsrc->SrcRegisterExtMod.Absolute;
+	src.negate = fsrc->SrcRegister.Negate;
+	src.swz[0] = fsrc->SrcRegister.SwizzleX;
+	src.swz[1] = fsrc->SrcRegister.SwizzleY;
+	src.swz[2] = fsrc->SrcRegister.SwizzleZ;
+	src.swz[3] = fsrc->SrcRegister.SwizzleW;
+	return src;
+}
+
+static INLINE struct nv30_sreg
+tgsi_dst(struct nv30_vpc *vpc, const struct tgsi_full_dst_register *fdst) {
+	struct nv30_sreg dst;
+
+	switch (fdst->DstRegister.File) {
+	case TGSI_FILE_OUTPUT:
+		dst = nv30_sr(NV30SR_OUTPUT,
+			      vpc->output_map[fdst->DstRegister.Index]);
+
+		break;
+	case TGSI_FILE_TEMPORARY:
+		dst = nv30_sr(NV30SR_TEMP, fdst->DstRegister.Index);
+		if (vpc->high_temp < dst.index)
+			vpc->high_temp = dst.index;
+		break;
+	default:
+		NOUVEAU_ERR("bad dst file\n");
+		break;
+	}
+
+	return dst;
+}
+
+static INLINE int
+tgsi_mask(uint tgsi)
+{
+	int mask = 0;
+
+	if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X;
+	if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y;
+	if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z;
+	if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W;
+	return mask;
+}
+
+static boolean
+nv30_vertprog_parse_instruction(struct nv30_vpc *vpc,
+				const struct tgsi_full_instruction *finst)
+{
+	struct nv30_sreg src[3], dst, tmp;
+	struct nv30_sreg none = nv30_sr(NV30SR_NONE, 0);
+	int mask;
+	int ai = -1, ci = -1;
+	int i;
+
+	if (finst->Instruction.Opcode == TGSI_OPCODE_END)
+		return TRUE;
+
+	vpc->temp_temp_count = 0;
+	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fsrc;
+
+		fsrc = &finst->FullSrcRegisters[i];
+		if (fsrc->SrcRegister.File == TGSI_FILE_TEMPORARY) {
+			src[i] = tgsi_src(vpc, fsrc);
+		}
+	}
+
+	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fsrc;
+
+		fsrc = &finst->FullSrcRegisters[i];
+		switch (fsrc->SrcRegister.File) {
+		case TGSI_FILE_INPUT:
+			if (ai == -1 || ai == fsrc->SrcRegister.Index) {
+				ai = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(vpc, fsrc);
+			} else {
+				src[i] = temp(vpc);
+				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
+				      tgsi_src(vpc, fsrc), none, none);
+			}
+			break;
+		/*XXX: index comparison is broken now that consts come from
+		 *     two different register files.
+		 */
+		case TGSI_FILE_CONSTANT:
+		case TGSI_FILE_IMMEDIATE:
+			if (ci == -1 || ci == fsrc->SrcRegister.Index) {
+				ci = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(vpc, fsrc);
+			} else {
+				src[i] = temp(vpc);
+				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
+				      tgsi_src(vpc, fsrc), none, none);
+			}
+			break;
+		case TGSI_FILE_TEMPORARY:
+			/* handled above */
+			break;
+		default:
+			NOUVEAU_ERR("bad src file\n");
+			return FALSE;
+		}
+	}
+
+	dst  = tgsi_dst(vpc, &finst->FullDstRegisters[0]);
+	mask = tgsi_mask(finst->FullDstRegisters[0].DstRegister.WriteMask);
+
+	switch (finst->Instruction.Opcode) {
+	case TGSI_OPCODE_ABS:
+		arith(vpc, 0, OP_MOV, dst, mask, abs(src[0]), none, none);
+		break;
+	case TGSI_OPCODE_ADD:
+		arith(vpc, 0, OP_ADD, dst, mask, src[0], none, src[1]);
+		break;
+	case TGSI_OPCODE_ARL:
+		arith(vpc, 0, OP_ARL, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_DP3:
+		arith(vpc, 0, OP_DP3, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DP4:
+		arith(vpc, 0, OP_DP4, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DPH:
+		arith(vpc, 0, OP_DPH, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DST:
+		arith(vpc, 0, OP_DST, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_EX2:
+		arith(vpc, 1, OP_EX2, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_EXP:
+		arith(vpc, 1, OP_EXP, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_FLR:
+		arith(vpc, 0, OP_FLR, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_FRC:
+		arith(vpc, 0, OP_FRC, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_LG2:
+		arith(vpc, 1, OP_LG2, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_LIT:
+		arith(vpc, 1, OP_LIT, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_LOG:
+		arith(vpc, 1, OP_LOG, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_MAD:
+		arith(vpc, 0, OP_MAD, dst, mask, src[0], src[1], src[2]);
+		break;
+	case TGSI_OPCODE_MAX:
+		arith(vpc, 0, OP_MAX, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_MIN:
+		arith(vpc, 0, OP_MIN, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_MOV:
+		arith(vpc, 0, OP_MOV, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_MUL:
+		arith(vpc, 0, OP_MUL, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_POW:
+		tmp = temp(vpc);
+		arith(vpc, 1, OP_LG2, tmp, MASK_X, none, none,
+		      swz(src[0], X, X, X, X));
+		arith(vpc, 0, OP_MUL, tmp, MASK_X, swz(tmp, X, X, X, X),
+		      swz(src[1], X, X, X, X), none);
+		arith(vpc, 1, OP_EX2, dst, mask, none, none,
+		      swz(tmp, X, X, X, X));
+		break;
+	case TGSI_OPCODE_RCP:
+		arith(vpc, 1, OP_RCP, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_RET:
+		break;
+	case TGSI_OPCODE_RSQ:
+		arith(vpc, 1, OP_RSQ, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_SGE:
+		arith(vpc, 0, OP_SGE, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SGT:
+		arith(vpc, 0, OP_SGT, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SLT:
+		arith(vpc, 0, OP_SLT, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SUB:
+		arith(vpc, 0, OP_ADD, dst, mask, src[0], none, neg(src[1]));
+		break;
+	case TGSI_OPCODE_XPD:
+		tmp = temp(vpc);
+		arith(vpc, 0, OP_MUL, tmp, mask,
+		      swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none);
+		arith(vpc, 0, OP_MAD, dst, (mask & ~MASK_W),
+		      swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y),
+		      neg(tmp));
+		break;
+	default:
+		NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode);
+		return FALSE;
+	}
+
+	return TRUE;
+}
+
+static boolean
+nv30_vertprog_parse_decl_output(struct nv30_vpc *vpc,
+				const struct tgsi_full_declaration *fdec)
+{
+	int hw;
+
+	switch (fdec->Semantic.SemanticName) {
+	case TGSI_SEMANTIC_POSITION:
+		hw = NV30_VP_INST_DEST_POS;
+		break;
+	case TGSI_SEMANTIC_COLOR:
+		if (fdec->Semantic.SemanticIndex == 0) {
+			hw = NV30_VP_INST_DEST_COL0;
+		} else
+		if (fdec->Semantic.SemanticIndex == 1) {
+			hw = NV30_VP_INST_DEST_COL1;
+		} else {
+			NOUVEAU_ERR("bad colour semantic index\n");
+			return FALSE;
+		}
+		break;
+	case TGSI_SEMANTIC_BCOLOR:
+		if (fdec->Semantic.SemanticIndex == 0) {
+			hw = NV30_VP_INST_DEST_BFC0;
+		} else
+		if (fdec->Semantic.SemanticIndex == 1) {
+			hw = NV30_VP_INST_DEST_BFC1;
+		} else {
+			NOUVEAU_ERR("bad bcolour semantic index\n");
+			return FALSE;
+		}
+		break;
+	case TGSI_SEMANTIC_FOG:
+		hw = NV30_VP_INST_DEST_FOGC;
+		break;
+	case TGSI_SEMANTIC_PSIZE:
+		hw = NV30_VP_INST_DEST_PSZ;
+		break;
+	case TGSI_SEMANTIC_GENERIC:
+		if (fdec->Semantic.SemanticIndex <= 7) {
+			hw = NV30_VP_INST_DEST_TC(fdec->Semantic.SemanticIndex);
+		} else {
+			NOUVEAU_ERR("bad generic semantic index\n");
+			return FALSE;
+		}
+		break;
+	default:
+		NOUVEAU_ERR("bad output semantic\n");
+		return FALSE;
+	}
+
+	vpc->output_map[fdec->DeclarationRange.First] = hw;
+	return TRUE;
+}
+
+static boolean
+nv30_vertprog_prepare(struct nv30_vpc *vpc)
+{
+	struct tgsi_parse_context p;
+	int nr_imm = 0;
+
+	tgsi_parse_init(&p, vpc->vp->pipe.tokens);
+	while (!tgsi_parse_end_of_tokens(&p)) {
+		const union tgsi_full_token *tok = &p.FullToken;
+
+		tgsi_parse_token(&p);
+		switch(tok->Token.Type) {
+		case TGSI_TOKEN_TYPE_IMMEDIATE:
+			nr_imm++;
+			break;
+		default:
+			break;
+		}
+	}
+	tgsi_parse_free(&p);
+
+	if (nr_imm) {
+		vpc->imm = CALLOC(nr_imm, sizeof(struct nv30_sreg));
+		assert(vpc->imm);
+	}
+
+	return TRUE;
+}
+
+static void
+nv30_vertprog_translate(struct nv30_context *nv30,
+			struct nv30_vertex_program *vp)
+{
+	struct tgsi_parse_context parse;
+	struct nv30_vpc *vpc = NULL;
+
+	tgsi_dump(vp->pipe.tokens,0);
+
+	vpc = CALLOC(1, sizeof(struct nv30_vpc));
+	if (!vpc)
+		return;
+	vpc->vp = vp;
+	vpc->high_temp = -1;
+
+	if (!nv30_vertprog_prepare(vpc)) {
+		FREE(vpc);
+		return;
+	}
+
+	tgsi_parse_init(&parse, vp->pipe.tokens);
+
+	while (!tgsi_parse_end_of_tokens(&parse)) {
+		tgsi_parse_token(&parse);
+
+		switch (parse.FullToken.Token.Type) {
+		case TGSI_TOKEN_TYPE_DECLARATION:
+		{
+			const struct tgsi_full_declaration *fdec;
+			fdec = &parse.FullToken.FullDeclaration;
+			switch (fdec->Declaration.File) {
+			case TGSI_FILE_OUTPUT:
+				if (!nv30_vertprog_parse_decl_output(vpc, fdec))
+					goto out_err;
+				break;
+			default:
+				break;
+			}
+		}
+			break;
+		case TGSI_TOKEN_TYPE_IMMEDIATE:
+		{
+			const struct tgsi_full_immediate *imm;
+
+			imm = &parse.FullToken.FullImmediate;
+			assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32);
+			assert(imm->Immediate.NrTokens == 4 + 1);
+			vpc->imm[vpc->nr_imm++] =
+				constant(vpc, -1,
+					 imm->u.ImmediateFloat32[0].Float,
+					 imm->u.ImmediateFloat32[1].Float,
+					 imm->u.ImmediateFloat32[2].Float,
+					 imm->u.ImmediateFloat32[3].Float);
+		}
+			break;
+		case TGSI_TOKEN_TYPE_INSTRUCTION:
+		{
+			const struct tgsi_full_instruction *finst;
+			finst = &parse.FullToken.FullInstruction;
+			if (!nv30_vertprog_parse_instruction(vpc, finst))
+				goto out_err;
+		}
+			break;
+		default:
+			break;
+		}
+	}
+
+	vp->insns[vp->nr_insns - 1].data[3] |= NV30_VP_INST_LAST;
+	vp->translated = TRUE;
+out_err:
+	tgsi_parse_free(&parse);
+	FREE(vpc);
+}
+
+static boolean
+nv30_vertprog_validate(struct nv30_context *nv30)
+{ 
+	struct nouveau_winsys *nvws = nv30->nvws;
+	struct pipe_winsys *ws = nv30->pipe.winsys;
+	struct nouveau_grobj *rankine = nv30->screen->rankine;
+	struct nv30_vertex_program *vp;
+	struct pipe_buffer *constbuf;
+	boolean upload_code = FALSE, upload_data = FALSE;
+	int i;
+
+	vp = nv30->vertprog;
+	constbuf = nv30->constbuf[PIPE_SHADER_VERTEX];
+
+	/* Translate TGSI shader into hw bytecode */
+	if (!vp->translated) {
+		nv30_vertprog_translate(nv30, vp);
+		if (!vp->translated)
+			return FALSE;
+	}
+
+	/* Allocate hw vtxprog exec slots */
+	if (!vp->exec) {
+		struct nouveau_resource *heap = nv30->screen->vp_exec_heap;
+		struct nouveau_stateobj *so;
+		uint vplen = vp->nr_insns;
+
+		if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) {
+			while (heap->next && heap->size < vplen) {
+				struct nv30_vertex_program *evict;
+				
+				evict = heap->next->priv;
+				nvws->res_free(&evict->exec);
+			}
+
+			if (nvws->res_alloc(heap, vplen, vp, &vp->exec))
+				assert(0);
+		}
+
+		so = so_new(2, 0);
+		so_method(so, rankine, NV34TCL_VP_START_FROM_ID, 1);
+		so_data  (so, vp->exec->start);
+		so_ref(so, &vp->so);
+
+		upload_code = TRUE;
+	}
+
+	/* Allocate hw vtxprog const slots */
+	if (vp->nr_consts && !vp->data) {
+		struct nouveau_resource *heap = nv30->screen->vp_data_heap;
+
+		if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) {
+			while (heap->next && heap->size < vp->nr_consts) {
+				struct nv30_vertex_program *evict;
+				
+				evict = heap->next->priv;
+				nvws->res_free(&evict->data);
+			}
+
+			if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data))
+				assert(0);
+		}
+
+		/*XXX: handle this some day */
+		assert(vp->data->start >= vp->data_start_min);
+
+		upload_data = TRUE;
+		if (vp->data_start != vp->data->start)
+			upload_code = TRUE;
+	}
+
+	/* If exec or data segments moved we need to patch the program to
+	 * fixup offsets and register IDs.
+	 */
+	if (vp->exec_start != vp->exec->start) {
+		for (i = 0; i < vp->nr_insns; i++) {
+			struct nv30_vertex_program_exec *vpi = &vp->insns[i];
+
+			if (vpi->has_branch_offset) {
+				assert(0);
+			}
+		}
+
+		vp->exec_start = vp->exec->start;
+	}
+
+	if (vp->nr_consts && vp->data_start != vp->data->start) {
+		for (i = 0; i < vp->nr_insns; i++) {
+			struct nv30_vertex_program_exec *vpi = &vp->insns[i];
+
+			if (vpi->const_index >= 0) {
+				vpi->data[1] &= ~NV30_VP_INST_CONST_SRC_MASK;
+				vpi->data[1] |=
+					(vpi->const_index + vp->data->start) <<
+					NV30_VP_INST_CONST_SRC_SHIFT;
+
+			}
+		}
+
+		vp->data_start = vp->data->start;
+	}
+
+	/* Update + Upload constant values */
+	if (vp->nr_consts) {
+		float *map = NULL;
+
+		if (constbuf) {
+			map = ws->buffer_map(ws, constbuf,
+					     PIPE_BUFFER_USAGE_CPU_READ);
+		}
+
+		for (i = 0; i < vp->nr_consts; i++) {
+			struct nv30_vertex_program_data *vpd = &vp->consts[i];
+
+			if (vpd->index >= 0) {
+				if (!upload_data &&
+				    !memcmp(vpd->value, &map[vpd->index * 4],
+					    4 * sizeof(float)))
+					continue;
+				memcpy(vpd->value, &map[vpd->index * 4],
+				       4 * sizeof(float));
+			}
+
+			BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_CONST_ID, 5);
+			OUT_RING  (i + vp->data->start);
+			OUT_RINGp ((uint32_t *)vpd->value, 4);
+		}
+
+		if (constbuf) {
+			ws->buffer_unmap(ws, constbuf);
+		}
+	}
+
+	/* Upload vtxprog */
+	if (upload_code) {
+#if 0
+		for (i = 0; i < vp->nr_insns; i++) {
+			NOUVEAU_MSG("VP inst %d: 0x%08x 0x%08x 0x%08x 0x%08x\n",
+				i, vp->insns[i].data[0], vp->insns[i].data[1],
+				vp->insns[i].data[2], vp->insns[i].data[3]);
+		}
+#endif
+		BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_FROM_ID, 1);
+		OUT_RING  (vp->exec->start);
+		for (i = 0; i < vp->nr_insns; i++) {
+			BEGIN_RING(rankine, NV34TCL_VP_UPLOAD_INST(0), 4);
+			OUT_RINGp (vp->insns[i].data, 4);
+		}
+	}
+
+	if (vp->so != nv30->state.hw[NV30_STATE_VERTPROG]) {
+		so_ref(vp->so, &nv30->state.hw[NV30_STATE_VERTPROG]);
+		return TRUE;
+	}
+
+	return FALSE;
+}
+
+void
+nv30_vertprog_destroy(struct nv30_context *nv30, struct nv30_vertex_program *vp)
+{
+	struct nouveau_winsys *nvws = nv30->screen->nvws;
+
+	vp->translated = FALSE;
+
+	if (vp->nr_insns) {
+		FREE(vp->insns);
+		vp->insns = NULL;
+		vp->nr_insns = 0;
+	}
+
+	if (vp->nr_consts) {
+		FREE(vp->consts);
+		vp->consts = NULL;
+		vp->nr_consts = 0;
+	}
+
+	nvws->res_free(&vp->exec);
+	vp->exec_start = 0;
+	nvws->res_free(&vp->data);
+	vp->data_start = 0;
+	vp->data_start_min = 0;
+
+	vp->ir = vp->or = 0;
+	so_ref(NULL, &vp->so);
+}
+
+struct nv30_state_entry nv30_state_vertprog = {
+	.validate = nv30_vertprog_validate,
+	.dirty = {
+		.pipe = NV30_NEW_VERTPROG /*| NV30_NEW_UCP*/,
+		.hw = NV30_STATE_VERTPROG,
+	}
+};
diff --git a/src/gallium/drivers/nv40/Makefile b/src/gallium/drivers/nv40/Makefile
new file mode 100644
index 0000000000..9c8eadf7e4
--- /dev/null
+++ b/src/gallium/drivers/nv40/Makefile
@@ -0,0 +1,37 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nv40
+
+DRIVER_SOURCES = \
+	nv40_clear.c \
+	nv40_context.c \
+	nv40_draw.c \
+	nv40_fragprog.c \
+	nv40_fragtex.c \
+	nv40_miptree.c \
+	nv40_query.c \
+	nv40_screen.c \
+	nv40_state.c \
+	nv40_state_blend.c \
+	nv40_state_emit.c \
+	nv40_state_fb.c \
+	nv40_state_rasterizer.c \
+	nv40_state_scissor.c \
+	nv40_state_stipple.c \
+	nv40_state_viewport.c \
+	nv40_state_zsa.c \
+	nv40_surface.c \
+	nv40_vbo.c \
+	nv40_vertprog.c
+
+C_SOURCES = \
+	$(COMMON_SOURCES) \
+	$(DRIVER_SOURCES)
+
+ASM_SOURCES = 
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/drivers/nv40/nv40_clear.c b/src/gallium/drivers/nv40/nv40_clear.c
new file mode 100644
index 0000000000..59efd620e3
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_clear.c
@@ -0,0 +1,13 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "nv40_context.h"
+
+void
+nv40_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+	   unsigned clearValue)
+{
+	pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue);
+	ps->status = PIPE_SURFACE_STATUS_CLEAR;
+}
diff --git a/src/gallium/drivers/nv40/nv40_context.c b/src/gallium/drivers/nv40/nv40_context.c
new file mode 100644
index 0000000000..5d325f5067
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_context.c
@@ -0,0 +1,72 @@
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+#include "nv40_context.h"
+#include "nv40_screen.h"
+
+static void
+nv40_flush(struct pipe_context *pipe, unsigned flags,
+	   struct pipe_fence_handle **fence)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	
+	if (flags & PIPE_FLUSH_TEXTURE_CACHE) {
+		BEGIN_RING(curie, 0x1fd8, 1);
+		OUT_RING  (2);
+		BEGIN_RING(curie, 0x1fd8, 1);
+		OUT_RING  (1);
+	}
+
+	FIRE_RING(fence);
+}
+
+static void
+nv40_destroy(struct pipe_context *pipe)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	if (nv40->draw)
+		draw_destroy(nv40->draw);
+	FREE(nv40);
+}
+
+struct pipe_context *
+nv40_create(struct pipe_screen *pscreen, unsigned pctx_id)
+{
+	struct nv40_screen *screen = nv40_screen(pscreen);
+	struct pipe_winsys *ws = pscreen->winsys;
+	struct nv40_context *nv40;
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nv40 = CALLOC(1, sizeof(struct nv40_context));
+	if (!nv40)
+		return NULL;
+	nv40->screen = screen;
+	nv40->pctx_id = pctx_id;
+
+	nv40->nvws = nvws;
+
+	nv40->pipe.winsys = ws;
+	nv40->pipe.screen = pscreen;
+	nv40->pipe.destroy = nv40_destroy;
+	nv40->pipe.draw_arrays = nv40_draw_arrays;
+	nv40->pipe.draw_elements = nv40_draw_elements;
+	nv40->pipe.clear = nv40_clear;
+	nv40->pipe.flush = nv40_flush;
+
+	nv40_init_query_functions(nv40);
+	nv40_init_surface_functions(nv40);
+	nv40_init_state_functions(nv40);
+
+	/* Create, configure, and install fallback swtnl path */
+	nv40->draw = draw_create();
+	draw_wide_point_threshold(nv40->draw, 9999999.0);
+	draw_wide_line_threshold(nv40->draw, 9999999.0);
+	draw_enable_line_stipple(nv40->draw, FALSE);
+	draw_enable_point_sprites(nv40->draw, FALSE);
+	draw_set_rasterize_stage(nv40->draw, nv40_draw_render_stage(nv40));
+
+	return &nv40->pipe;
+}
+	
diff --git a/src/gallium/drivers/nv40/nv40_context.h b/src/gallium/drivers/nv40/nv40_context.h
new file mode 100644
index 0000000000..adcfbdd85a
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_context.h
@@ -0,0 +1,233 @@
+#ifndef __NV40_CONTEXT_H__
+#define __NV40_CONTEXT_H__
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_compiler.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "draw/draw_vertex.h"
+
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau/nouveau_gldefs.h"
+
+#define NOUVEAU_PUSH_CONTEXT(ctx)                                              \
+	struct nv40_screen *ctx = nv40->screen
+#include "nouveau/nouveau_push.h"
+#include "nouveau/nouveau_stateobj.h"
+
+#include "nv40_state.h"
+
+#define NOUVEAU_ERR(fmt, args...) \
+	fprintf(stderr, "%s:%d -  "fmt, __func__, __LINE__, ##args);
+#define NOUVEAU_MSG(fmt, args...) \
+	fprintf(stderr, "nouveau: "fmt, ##args);
+
+enum nv40_state_index {
+	NV40_STATE_FB = 0,
+	NV40_STATE_VIEWPORT = 1,
+	NV40_STATE_BLEND = 2,
+	NV40_STATE_RAST = 3,
+	NV40_STATE_ZSA = 4,
+	NV40_STATE_BCOL = 5,
+	NV40_STATE_CLIP = 6,
+	NV40_STATE_SCISSOR = 7,
+	NV40_STATE_STIPPLE = 8,
+	NV40_STATE_FRAGPROG = 9,
+	NV40_STATE_VERTPROG = 10,
+	NV40_STATE_FRAGTEX0 = 11,
+	NV40_STATE_FRAGTEX1 = 12,
+	NV40_STATE_FRAGTEX2 = 13,
+	NV40_STATE_FRAGTEX3 = 14,
+	NV40_STATE_FRAGTEX4 = 15,
+	NV40_STATE_FRAGTEX5 = 16,
+	NV40_STATE_FRAGTEX6 = 17,
+	NV40_STATE_FRAGTEX7 = 18,
+	NV40_STATE_FRAGTEX8 = 19,
+	NV40_STATE_FRAGTEX9 = 20,
+	NV40_STATE_FRAGTEX10 = 21,
+	NV40_STATE_FRAGTEX11 = 22,
+	NV40_STATE_FRAGTEX12 = 23,
+	NV40_STATE_FRAGTEX13 = 24,
+	NV40_STATE_FRAGTEX14 = 25,
+	NV40_STATE_FRAGTEX15 = 26,
+	NV40_STATE_VERTTEX0 = 27,
+	NV40_STATE_VERTTEX1 = 28,
+	NV40_STATE_VERTTEX2 = 29,
+	NV40_STATE_VERTTEX3 = 30,
+	NV40_STATE_VTXBUF = 31,
+	NV40_STATE_VTXFMT = 32,
+	NV40_STATE_VTXATTR = 33,
+	NV40_STATE_MAX = 34
+};
+
+#include "nv40_screen.h"
+
+#define NV40_NEW_BLEND		(1 <<  0)
+#define NV40_NEW_RAST		(1 <<  1)
+#define NV40_NEW_ZSA		(1 <<  2)
+#define NV40_NEW_SAMPLER	(1 <<  3)
+#define NV40_NEW_FB		(1 <<  4)
+#define NV40_NEW_STIPPLE	(1 <<  5)
+#define NV40_NEW_SCISSOR	(1 <<  6)
+#define NV40_NEW_VIEWPORT	(1 <<  7)
+#define NV40_NEW_BCOL		(1 <<  8)
+#define NV40_NEW_VERTPROG	(1 <<  9)
+#define NV40_NEW_FRAGPROG	(1 << 10)
+#define NV40_NEW_ARRAYS		(1 << 11)
+#define NV40_NEW_UCP		(1 << 12)
+
+struct nv40_rasterizer_state {
+	struct pipe_rasterizer_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+struct nv40_zsa_state {
+	struct pipe_depth_stencil_alpha_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+struct nv40_blend_state {
+	struct pipe_blend_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+
+struct nv40_state {
+	unsigned scissor_enabled;
+	unsigned stipple_enabled;
+	unsigned viewport_bypass;
+	unsigned fp_samplers;
+
+	uint64_t dirty;
+	struct nouveau_stateobj *hw[NV40_STATE_MAX];
+};
+
+struct nv40_context {
+	struct pipe_context pipe;
+
+	struct nouveau_winsys *nvws;
+	struct nv40_screen *screen;
+	unsigned pctx_id;
+
+	struct draw_context *draw;
+
+	/* HW state derived from pipe states */
+	struct nv40_state state;
+	struct {
+		struct nv40_vertex_program *vertprog;
+
+		unsigned nr_attribs;
+		unsigned hw[PIPE_MAX_SHADER_INPUTS];
+		unsigned draw[PIPE_MAX_SHADER_INPUTS];
+		unsigned emit[PIPE_MAX_SHADER_INPUTS];
+	} swtnl;
+
+	enum {
+		HW, SWTNL, SWRAST
+	} render_mode;
+	unsigned fallback_swtnl;
+	unsigned fallback_swrast;
+
+	/* Context state */
+	unsigned dirty, draw_dirty;
+	struct pipe_scissor_state scissor;
+	unsigned stipple[32];
+	struct pipe_clip_state clip;
+	struct nv40_vertex_program *vertprog;
+	struct nv40_fragment_program *fragprog;
+	struct pipe_buffer *constbuf[PIPE_SHADER_TYPES];
+	unsigned constbuf_nr[PIPE_SHADER_TYPES];
+	struct nv40_rasterizer_state *rasterizer;
+	struct nv40_zsa_state *zsa;
+	struct nv40_blend_state *blend;
+	struct pipe_blend_color blend_colour;
+	struct pipe_viewport_state viewport;
+	struct pipe_framebuffer_state framebuffer;
+	struct pipe_buffer *idxbuf;
+	unsigned idxbuf_format;
+	struct nv40_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS];
+	struct nv40_miptree *tex_miptree[PIPE_MAX_SAMPLERS];
+	unsigned nr_samplers;
+	unsigned nr_textures;
+	unsigned dirty_samplers;
+	struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
+	unsigned vtxbuf_nr;
+	struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS];
+	unsigned vtxelt_nr;
+	const unsigned *edgeflags;
+};
+
+static INLINE struct nv40_context *
+nv40_context(struct pipe_context *pipe)
+{
+	return (struct nv40_context *)pipe;
+}
+
+struct nv40_state_entry {
+	boolean (*validate)(struct nv40_context *nv40);
+	struct {
+		unsigned pipe;
+		unsigned hw;
+	} dirty;
+};
+
+extern void nv40_init_state_functions(struct nv40_context *nv40);
+extern void nv40_init_surface_functions(struct nv40_context *nv40);
+extern void nv40_init_query_functions(struct nv40_context *nv40);
+
+extern void nv40_screen_init_miptree_functions(struct pipe_screen *pscreen);
+
+/* nv40_draw.c */
+extern struct draw_stage *nv40_draw_render_stage(struct nv40_context *nv40);
+extern boolean nv40_draw_elements_swtnl(struct pipe_context *pipe,
+					struct pipe_buffer *idxbuf,
+					unsigned ib_size, unsigned mode,
+					unsigned start, unsigned count);
+
+/* nv40_vertprog.c */
+extern void nv40_vertprog_destroy(struct nv40_context *,
+				  struct nv40_vertex_program *);
+
+/* nv40_fragprog.c */
+extern void nv40_fragprog_destroy(struct nv40_context *,
+				  struct nv40_fragment_program *);
+
+/* nv40_fragtex.c */
+extern void nv40_fragtex_bind(struct nv40_context *);
+
+/* nv40_state.c and friends */
+extern boolean nv40_state_validate(struct nv40_context *nv40);
+extern boolean nv40_state_validate_swtnl(struct nv40_context *nv40);
+extern void nv40_state_emit(struct nv40_context *nv40);
+extern struct nv40_state_entry nv40_state_rasterizer;
+extern struct nv40_state_entry nv40_state_scissor;
+extern struct nv40_state_entry nv40_state_stipple;
+extern struct nv40_state_entry nv40_state_fragprog;
+extern struct nv40_state_entry nv40_state_vertprog;
+extern struct nv40_state_entry nv40_state_blend;
+extern struct nv40_state_entry nv40_state_blend_colour;
+extern struct nv40_state_entry nv40_state_zsa;
+extern struct nv40_state_entry nv40_state_viewport;
+extern struct nv40_state_entry nv40_state_framebuffer;
+extern struct nv40_state_entry nv40_state_fragtex;
+extern struct nv40_state_entry nv40_state_vbo;
+extern struct nv40_state_entry nv40_state_vtxfmt;
+
+/* nv40_vbo.c */
+extern boolean nv40_draw_arrays(struct pipe_context *, unsigned mode,
+				unsigned start, unsigned count);
+extern boolean nv40_draw_elements(struct pipe_context *pipe,
+				  struct pipe_buffer *indexBuffer,
+				  unsigned indexSize,
+				  unsigned mode, unsigned start,
+				  unsigned count);
+
+/* nv40_clear.c */
+extern void nv40_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+		       unsigned clearValue);
+
+#endif
diff --git a/src/gallium/drivers/nv40/nv40_draw.c b/src/gallium/drivers/nv40/nv40_draw.c
new file mode 100644
index 0000000000..c83ff91d7e
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_draw.c
@@ -0,0 +1,349 @@
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_pack_color.h"
+
+#include "draw/draw_context.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_pipe.h"
+
+#include "nv40_context.h"
+#define NV40_SHADER_NO_FUCKEDNESS
+#include "nv40_shader.h"
+
+/* Simple, but crappy, swtnl path, hopefully we wont need to hit this very
+ * often at all.  Uses "quadro style" vertex submission + a fixed vertex
+ * layout to avoid the need to generate a vertex program or vtxfmt.
+ */
+
+struct nv40_render_stage {
+	struct draw_stage stage;
+	struct nv40_context *nv40;
+	unsigned prim;
+};
+
+static INLINE struct nv40_render_stage *
+nv40_render_stage(struct draw_stage *stage)
+{
+	return (struct nv40_render_stage *)stage;
+}
+
+static INLINE void
+nv40_render_vertex(struct nv40_context *nv40, const struct vertex_header *v)
+{
+	unsigned i;
+
+	for (i = 0; i < nv40->swtnl.nr_attribs; i++) {
+		unsigned idx = nv40->swtnl.draw[i];
+		unsigned hw = nv40->swtnl.hw[i];
+
+		switch (nv40->swtnl.emit[i]) {
+		case EMIT_OMIT:
+			break;
+		case EMIT_1F:
+			BEGIN_RING(curie, NV40TCL_VTX_ATTR_1F(hw), 1);
+			OUT_RING  (fui(v->data[idx][0]));
+			break;
+		case EMIT_2F:
+			BEGIN_RING(curie, NV40TCL_VTX_ATTR_2F_X(hw), 2);
+			OUT_RING  (fui(v->data[idx][0]));
+			OUT_RING  (fui(v->data[idx][1]));
+			break;
+		case EMIT_3F:
+			BEGIN_RING(curie, NV40TCL_VTX_ATTR_3F_X(hw), 3);
+			OUT_RING  (fui(v->data[idx][0]));
+			OUT_RING  (fui(v->data[idx][1]));
+			OUT_RING  (fui(v->data[idx][2]));
+			break;
+		case EMIT_4F:
+			BEGIN_RING(curie, NV40TCL_VTX_ATTR_4F_X(hw), 4);
+			OUT_RING  (fui(v->data[idx][0]));
+			OUT_RING  (fui(v->data[idx][1]));
+			OUT_RING  (fui(v->data[idx][2]));
+			OUT_RING  (fui(v->data[idx][3]));
+			break;
+		case EMIT_4UB:
+			BEGIN_RING(curie, NV40TCL_VTX_ATTR_4UB(hw), 1);
+			OUT_RING  (pack_ub4(float_to_ubyte(v->data[idx][0]),
+					    float_to_ubyte(v->data[idx][1]),
+					    float_to_ubyte(v->data[idx][2]),
+					    float_to_ubyte(v->data[idx][3])));
+			break;
+		default:
+			assert(0);
+			break;
+		}
+	}
+}
+
+static INLINE void
+nv40_render_prim(struct draw_stage *stage, struct prim_header *prim,
+	       unsigned mode, unsigned count)
+{
+	struct nv40_render_stage *rs = nv40_render_stage(stage);
+	struct nv40_context *nv40 = rs->nv40;
+	struct nouveau_pushbuf *pb = nv40->nvws->channel->pushbuf;
+	unsigned i;
+
+	/* Ensure there's room for 4xfloat32 + potentially 3 begin/end */
+	if (pb->remaining < ((count * 20) + 6)) {
+		if (rs->prim != NV40TCL_BEGIN_END_STOP) {
+			NOUVEAU_ERR("AIII, missed flush\n");
+			assert(0);
+		}
+		FIRE_RING(NULL);
+		nv40_state_emit(nv40);
+	}
+
+	/* Switch primitive modes if necessary */
+	if (rs->prim != mode) {
+		if (rs->prim != NV40TCL_BEGIN_END_STOP) {
+			BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+			OUT_RING  (NV40TCL_BEGIN_END_STOP);	
+		}
+
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (mode);
+		rs->prim = mode;
+	}
+
+	/* Emit vertex data */
+	for (i = 0; i < count; i++)
+		nv40_render_vertex(nv40, prim->v[i]);
+
+	/* If it's likely we'll need to empty the push buffer soon, finish
+	 * off the primitive now.
+	 */
+	if (pb->remaining < ((count * 20) + 6)) {
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (NV40TCL_BEGIN_END_STOP);
+		rs->prim = NV40TCL_BEGIN_END_STOP;
+	}
+}
+
+static void
+nv40_render_point(struct draw_stage *draw, struct prim_header *prim)
+{
+	nv40_render_prim(draw, prim, NV40TCL_BEGIN_END_POINTS, 1);
+}
+
+static void
+nv40_render_line(struct draw_stage *draw, struct prim_header *prim)
+{
+	nv40_render_prim(draw, prim, NV40TCL_BEGIN_END_LINES, 2);
+}
+
+static void
+nv40_render_tri(struct draw_stage *draw, struct prim_header *prim)
+{
+	nv40_render_prim(draw, prim, NV40TCL_BEGIN_END_TRIANGLES, 3);
+}
+
+static void
+nv40_render_flush(struct draw_stage *draw, unsigned flags)
+{
+	struct nv40_render_stage *rs = nv40_render_stage(draw);
+	struct nv40_context *nv40 = rs->nv40;
+
+	if (rs->prim != NV40TCL_BEGIN_END_STOP) {
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (NV40TCL_BEGIN_END_STOP);
+		rs->prim = NV40TCL_BEGIN_END_STOP;
+	}
+}
+
+static void
+nv40_render_reset_stipple_counter(struct draw_stage *draw)
+{
+}
+
+static void
+nv40_render_destroy(struct draw_stage *draw)
+{
+	FREE(draw);
+}
+
+static INLINE void
+emit_mov(struct nv40_vertex_program *vp,
+	 unsigned dst, unsigned src, unsigned vor, unsigned mask)
+{
+	struct nv40_vertex_program_exec *inst;
+
+	vp->insns = realloc(vp->insns,
+			    sizeof(struct nv40_vertex_program_exec) *
+			    ++vp->nr_insns);
+	inst = &vp->insns[vp->nr_insns - 1];
+
+	inst->data[0] = 0x401f9c6c;
+	inst->data[1] = 0x0040000d | (src << 8);
+	inst->data[2] = 0x8106c083;
+	inst->data[3] = 0x6041ff80 | (dst << 2) | (mask << 13);
+	inst->const_index = -1;
+	inst->has_branch_offset = FALSE;
+
+	vp->ir |= (1 << src);
+	if (vor != ~0)
+		vp->or |= (1 << vor);
+}
+
+static struct nv40_vertex_program *
+create_drawvp(struct nv40_context *nv40)
+{
+	struct nv40_vertex_program *vp = CALLOC_STRUCT(nv40_vertex_program);
+	unsigned i;
+
+	emit_mov(vp, NV40_VP_INST_DEST_POS, 0, ~0, 0xf);
+	emit_mov(vp, NV40_VP_INST_DEST_COL0, 3, 0, 0xf);
+	emit_mov(vp, NV40_VP_INST_DEST_COL1, 4, 1, 0xf);
+	emit_mov(vp, NV40_VP_INST_DEST_BFC0, 3, 2, 0xf);
+	emit_mov(vp, NV40_VP_INST_DEST_BFC1, 4, 3, 0xf);
+	emit_mov(vp, NV40_VP_INST_DEST_FOGC, 5, 4, 0x8);
+	for (i = 0; i < 8; i++)
+		emit_mov(vp, NV40_VP_INST_DEST_TC(i), 8 + i, 14 + i, 0xf);
+
+	vp->insns[vp->nr_insns - 1].data[3] |= 1;
+	vp->translated = TRUE;
+	return vp;
+}
+
+struct draw_stage *
+nv40_draw_render_stage(struct nv40_context *nv40)
+{
+	struct nv40_render_stage *render = CALLOC_STRUCT(nv40_render_stage);
+
+	if (!nv40->swtnl.vertprog)
+		nv40->swtnl.vertprog = create_drawvp(nv40);
+
+	render->nv40 = nv40;
+	render->stage.draw = nv40->draw;
+	render->stage.point = nv40_render_point;
+	render->stage.line = nv40_render_line;
+	render->stage.tri = nv40_render_tri;
+	render->stage.flush = nv40_render_flush;
+	render->stage.reset_stipple_counter = nv40_render_reset_stipple_counter;
+	render->stage.destroy = nv40_render_destroy;
+
+	return &render->stage;
+}
+
+boolean
+nv40_draw_elements_swtnl(struct pipe_context *pipe,
+			 struct pipe_buffer *idxbuf, unsigned idxbuf_size,
+			 unsigned mode, unsigned start, unsigned count)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct pipe_winsys *ws = pipe->winsys;
+	unsigned i;
+	void *map;
+
+	if (!nv40_state_validate_swtnl(nv40))
+		return FALSE;
+	nv40->state.dirty &= ~(1ULL << NV40_STATE_VTXBUF);
+	nv40_state_emit(nv40);
+
+	for (i = 0; i < nv40->vtxbuf_nr; i++) {
+		map = ws->buffer_map(ws, nv40->vtxbuf[i].buffer,
+                                      PIPE_BUFFER_USAGE_CPU_READ);
+		draw_set_mapped_vertex_buffer(nv40->draw, i, map);
+	}
+
+	if (idxbuf) {
+		map = ws->buffer_map(ws, idxbuf, PIPE_BUFFER_USAGE_CPU_READ);
+		draw_set_mapped_element_buffer(nv40->draw, idxbuf_size, map);
+	} else {
+		draw_set_mapped_element_buffer(nv40->draw, 0, NULL);
+	}
+
+	if (nv40->constbuf[PIPE_SHADER_VERTEX]) {
+		const unsigned nr = nv40->constbuf_nr[PIPE_SHADER_VERTEX];
+
+		map = ws->buffer_map(ws, nv40->constbuf[PIPE_SHADER_VERTEX],
+				     PIPE_BUFFER_USAGE_CPU_READ);
+		draw_set_mapped_constant_buffer(nv40->draw, map, nr);
+	}
+
+	draw_arrays(nv40->draw, mode, start, count);
+
+	for (i = 0; i < nv40->vtxbuf_nr; i++)
+		ws->buffer_unmap(ws, nv40->vtxbuf[i].buffer);
+
+	if (idxbuf)
+		ws->buffer_unmap(ws, idxbuf);
+
+	if (nv40->constbuf[PIPE_SHADER_VERTEX])
+		ws->buffer_unmap(ws, nv40->constbuf[PIPE_SHADER_VERTEX]);
+
+	draw_flush(nv40->draw);
+	pipe->flush(pipe, 0, NULL);
+
+	return TRUE;
+}
+
+static INLINE void
+emit_attrib(struct nv40_context *nv40, unsigned hw, unsigned emit,
+	    unsigned semantic, unsigned index)
+{
+	unsigned draw_out = draw_find_vs_output(nv40->draw, semantic, index);
+	unsigned a = nv40->swtnl.nr_attribs++;
+
+	nv40->swtnl.hw[a] = hw;
+	nv40->swtnl.emit[a] = emit;
+	nv40->swtnl.draw[a] = draw_out;
+}
+
+static boolean
+nv40_state_vtxfmt_validate(struct nv40_context *nv40)
+{
+	struct nv40_fragment_program *fp = nv40->fragprog;
+	unsigned colour = 0, texcoords = 0, fog = 0, i;
+
+	/* Determine needed fragprog inputs */
+	for (i = 0; i < fp->info.num_inputs; i++) {
+		switch (fp->info.input_semantic_name[i]) {
+		case TGSI_SEMANTIC_POSITION:
+			break;
+		case TGSI_SEMANTIC_COLOR:
+			colour |= (1 << fp->info.input_semantic_index[i]);
+			break;
+		case TGSI_SEMANTIC_GENERIC:
+			texcoords |= (1 << fp->info.input_semantic_index[i]);
+			break;
+		case TGSI_SEMANTIC_FOG:
+			fog = 1;
+			break;
+		default:
+			assert(0);
+		}
+	}
+
+	nv40->swtnl.nr_attribs = 0;
+
+	/* Map draw vtxprog output to hw attribute IDs */
+	for (i = 0; i < 2; i++) {
+		if (!(colour & (1 << i)))
+			continue;
+		emit_attrib(nv40, 3 + i, EMIT_4UB, TGSI_SEMANTIC_COLOR, i);
+	}
+
+	for (i = 0; i < 8; i++) {
+		if (!(texcoords & (1 << i)))
+			continue;
+		emit_attrib(nv40, 8 + i, EMIT_4F, TGSI_SEMANTIC_GENERIC, i);
+	}
+
+	if (fog) {
+		emit_attrib(nv40, 5, EMIT_1F, TGSI_SEMANTIC_FOG, 0);
+	}
+
+	emit_attrib(nv40, 0, EMIT_3F, TGSI_SEMANTIC_POSITION, 0);
+
+	return FALSE;
+}
+
+struct nv40_state_entry nv40_state_vtxfmt = {
+	.validate = nv40_state_vtxfmt_validate,
+	.dirty = {
+		.pipe = NV40_NEW_ARRAYS | NV40_NEW_FRAGPROG,
+		.hw = 0
+	}
+};
+
diff --git a/src/gallium/drivers/nv40/nv40_fragprog.c b/src/gallium/drivers/nv40/nv40_fragprog.c
new file mode 100644
index 0000000000..91dcbebda0
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_fragprog.c
@@ -0,0 +1,991 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+
+#include "nv40_context.h"
+
+#define SWZ_X 0
+#define SWZ_Y 1
+#define SWZ_Z 2
+#define SWZ_W 3
+#define MASK_X 1
+#define MASK_Y 2
+#define MASK_Z 4
+#define MASK_W 8
+#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W)
+#define DEF_SCALE NV40_FP_OP_DST_SCALE_1X
+#define DEF_CTEST NV40_FP_OP_COND_TR
+#include "nv40_shader.h"
+
+#define swz(s,x,y,z,w) nv40_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w)
+#define neg(s) nv40_sr_neg((s))
+#define abs(s) nv40_sr_abs((s))
+#define scale(s,v) nv40_sr_scale((s), NV40_FP_OP_DST_SCALE_##v)
+
+#define MAX_CONSTS 128
+#define MAX_IMM 32
+struct nv40_fpc {
+	struct nv40_fragment_program *fp;
+
+	uint attrib_map[PIPE_MAX_SHADER_INPUTS];
+
+	unsigned r_temps;
+	unsigned r_temps_discard;
+	struct nv40_sreg r_result[PIPE_MAX_SHADER_OUTPUTS];
+	struct nv40_sreg *r_temp;
+
+	int num_regs;
+
+	unsigned inst_offset;
+	unsigned have_const;
+
+	struct {
+		int pipe;
+		float vals[4];
+	} consts[MAX_CONSTS];
+	int nr_consts;
+
+	struct nv40_sreg imm[MAX_IMM];
+	unsigned nr_imm;
+};
+
+static INLINE struct nv40_sreg
+temp(struct nv40_fpc *fpc)
+{
+	int idx = ffs(~fpc->r_temps) - 1;
+
+	if (idx < 0) {
+		NOUVEAU_ERR("out of temps!!\n");
+		assert(0);
+		return nv40_sr(NV40SR_TEMP, 0);
+	}
+
+	fpc->r_temps |= (1 << idx);
+	fpc->r_temps_discard |= (1 << idx);
+	return nv40_sr(NV40SR_TEMP, idx);
+}
+
+static INLINE void
+release_temps(struct nv40_fpc *fpc)
+{
+	fpc->r_temps &= ~fpc->r_temps_discard;
+	fpc->r_temps_discard = 0;
+}
+
+static INLINE struct nv40_sreg
+constant(struct nv40_fpc *fpc, int pipe, float vals[4])
+{
+	int idx;
+
+	if (fpc->nr_consts == MAX_CONSTS)
+		assert(0);
+	idx = fpc->nr_consts++;
+
+	fpc->consts[idx].pipe = pipe;
+	if (pipe == -1)
+		memcpy(fpc->consts[idx].vals, vals, 4 * sizeof(float));
+	return nv40_sr(NV40SR_CONST, idx);
+}
+
+#define arith(cc,s,o,d,m,s0,s1,s2) \
+	nv40_fp_arith((cc), (s), NV40_FP_OP_OPCODE_##o, \
+			(d), (m), (s0), (s1), (s2))
+#define tex(cc,s,o,u,d,m,s0,s1,s2) \
+	nv40_fp_tex((cc), (s), NV40_FP_OP_OPCODE_##o, (u), \
+		    (d), (m), (s0), none, none)
+
+static void
+grow_insns(struct nv40_fpc *fpc, int size)
+{
+	struct nv40_fragment_program *fp = fpc->fp;
+
+	fp->insn_len += size;
+	fp->insn = realloc(fp->insn, sizeof(uint32_t) * fp->insn_len);
+}
+
+static void
+emit_src(struct nv40_fpc *fpc, int pos, struct nv40_sreg src)
+{
+	struct nv40_fragment_program *fp = fpc->fp;
+	uint32_t *hw = &fp->insn[fpc->inst_offset];
+	uint32_t sr = 0;
+
+	switch (src.type) {
+	case NV40SR_INPUT:
+		sr |= (NV40_FP_REG_TYPE_INPUT << NV40_FP_REG_TYPE_SHIFT);
+		hw[0] |= (src.index << NV40_FP_OP_INPUT_SRC_SHIFT);
+		break;
+	case NV40SR_OUTPUT:
+		sr |= NV40_FP_REG_SRC_HALF;
+		/* fall-through */
+	case NV40SR_TEMP:
+		sr |= (NV40_FP_REG_TYPE_TEMP << NV40_FP_REG_TYPE_SHIFT);
+		sr |= (src.index << NV40_FP_REG_SRC_SHIFT);
+		break;
+	case NV40SR_CONST:
+		if (!fpc->have_const) {
+			grow_insns(fpc, 4);
+			fpc->have_const = 1;
+		}
+
+		hw = &fp->insn[fpc->inst_offset];
+		if (fpc->consts[src.index].pipe >= 0) {
+			struct nv40_fragment_program_data *fpd;
+
+			fp->consts = realloc(fp->consts, ++fp->nr_consts *
+					     sizeof(*fpd));
+			fpd = &fp->consts[fp->nr_consts - 1];
+			fpd->offset = fpc->inst_offset + 4;
+			fpd->index = fpc->consts[src.index].pipe;
+			memset(&fp->insn[fpd->offset], 0, sizeof(uint32_t) * 4);
+		} else {
+			memcpy(&fp->insn[fpc->inst_offset + 4],
+				fpc->consts[src.index].vals,
+				sizeof(uint32_t) * 4);
+		}
+
+		sr |= (NV40_FP_REG_TYPE_CONST << NV40_FP_REG_TYPE_SHIFT);	
+		break;
+	case NV40SR_NONE:
+		sr |= (NV40_FP_REG_TYPE_INPUT << NV40_FP_REG_TYPE_SHIFT);
+		break;
+	default:
+		assert(0);
+	}
+
+	if (src.negate)
+		sr |= NV40_FP_REG_NEGATE;
+
+	if (src.abs)
+		hw[1] |= (1 << (29 + pos));
+
+	sr |= ((src.swz[0] << NV40_FP_REG_SWZ_X_SHIFT) |
+	       (src.swz[1] << NV40_FP_REG_SWZ_Y_SHIFT) |
+	       (src.swz[2] << NV40_FP_REG_SWZ_Z_SHIFT) |
+	       (src.swz[3] << NV40_FP_REG_SWZ_W_SHIFT));
+
+	hw[pos + 1] |= sr;
+}
+
+static void
+emit_dst(struct nv40_fpc *fpc, struct nv40_sreg dst)
+{
+	struct nv40_fragment_program *fp = fpc->fp;
+	uint32_t *hw = &fp->insn[fpc->inst_offset];
+
+	switch (dst.type) {
+	case NV40SR_TEMP:
+		if (fpc->num_regs < (dst.index + 1))
+			fpc->num_regs = dst.index + 1;
+		break;
+	case NV40SR_OUTPUT:
+		if (dst.index == 1) {
+			fp->fp_control |= 0xe;
+		} else {
+			hw[0] |= NV40_FP_OP_OUT_REG_HALF;
+		}
+		break;
+	case NV40SR_NONE:
+		hw[0] |= (1 << 30);
+		break;
+	default:
+		assert(0);
+	}
+
+	hw[0] |= (dst.index << NV40_FP_OP_OUT_REG_SHIFT);
+}
+
+static void
+nv40_fp_arith(struct nv40_fpc *fpc, int sat, int op,
+	      struct nv40_sreg dst, int mask,
+	      struct nv40_sreg s0, struct nv40_sreg s1, struct nv40_sreg s2)
+{
+	struct nv40_fragment_program *fp = fpc->fp;
+	uint32_t *hw;
+
+	fpc->inst_offset = fp->insn_len;
+	fpc->have_const = 0;
+	grow_insns(fpc, 4);
+	hw = &fp->insn[fpc->inst_offset];
+	memset(hw, 0, sizeof(uint32_t) * 4);
+
+	if (op == NV40_FP_OP_OPCODE_KIL)
+		fp->fp_control |= NV40TCL_FP_CONTROL_KIL;
+	hw[0] |= (op << NV40_FP_OP_OPCODE_SHIFT);
+	hw[0] |= (mask << NV40_FP_OP_OUTMASK_SHIFT);
+	hw[2] |= (dst.dst_scale << NV40_FP_OP_DST_SCALE_SHIFT);
+
+	if (sat)
+		hw[0] |= NV40_FP_OP_OUT_SAT;
+
+	if (dst.cc_update)
+		hw[0] |= NV40_FP_OP_COND_WRITE_ENABLE;
+	hw[1] |= (dst.cc_test << NV40_FP_OP_COND_SHIFT);
+	hw[1] |= ((dst.cc_swz[0] << NV40_FP_OP_COND_SWZ_X_SHIFT) |
+		  (dst.cc_swz[1] << NV40_FP_OP_COND_SWZ_Y_SHIFT) |
+		  (dst.cc_swz[2] << NV40_FP_OP_COND_SWZ_Z_SHIFT) |
+		  (dst.cc_swz[3] << NV40_FP_OP_COND_SWZ_W_SHIFT));
+
+	emit_dst(fpc, dst);
+	emit_src(fpc, 0, s0);
+	emit_src(fpc, 1, s1);
+	emit_src(fpc, 2, s2);
+}
+
+static void
+nv40_fp_tex(struct nv40_fpc *fpc, int sat, int op, int unit,
+	    struct nv40_sreg dst, int mask,
+	    struct nv40_sreg s0, struct nv40_sreg s1, struct nv40_sreg s2)
+{
+	struct nv40_fragment_program *fp = fpc->fp;
+
+	nv40_fp_arith(fpc, sat, op, dst, mask, s0, s1, s2);
+
+	fp->insn[fpc->inst_offset] |= (unit << NV40_FP_OP_TEX_UNIT_SHIFT);
+	fp->samplers |= (1 << unit);
+}
+
+static INLINE struct nv40_sreg
+tgsi_src(struct nv40_fpc *fpc, const struct tgsi_full_src_register *fsrc)
+{
+	struct nv40_sreg src;
+
+	switch (fsrc->SrcRegister.File) {
+	case TGSI_FILE_INPUT:
+		src = nv40_sr(NV40SR_INPUT,
+			      fpc->attrib_map[fsrc->SrcRegister.Index]);
+		break;
+	case TGSI_FILE_CONSTANT:
+		src = constant(fpc, fsrc->SrcRegister.Index, NULL);
+		break;
+	case TGSI_FILE_IMMEDIATE:
+		assert(fsrc->SrcRegister.Index < fpc->nr_imm);
+		src = fpc->imm[fsrc->SrcRegister.Index];
+		break;
+	case TGSI_FILE_TEMPORARY:
+		src = fpc->r_temp[fsrc->SrcRegister.Index];
+		break;
+	/* NV40 fragprog result regs are just temps, so this is simple */
+	case TGSI_FILE_OUTPUT:
+		src = fpc->r_result[fsrc->SrcRegister.Index];
+		break;
+	default:
+		NOUVEAU_ERR("bad src file\n");
+		break;
+	}
+
+	src.abs = fsrc->SrcRegisterExtMod.Absolute;
+	src.negate = fsrc->SrcRegister.Negate;
+	src.swz[0] = fsrc->SrcRegister.SwizzleX;
+	src.swz[1] = fsrc->SrcRegister.SwizzleY;
+	src.swz[2] = fsrc->SrcRegister.SwizzleZ;
+	src.swz[3] = fsrc->SrcRegister.SwizzleW;
+	return src;
+}
+
+static INLINE struct nv40_sreg
+tgsi_dst(struct nv40_fpc *fpc, const struct tgsi_full_dst_register *fdst) {
+	switch (fdst->DstRegister.File) {
+	case TGSI_FILE_OUTPUT:
+		return fpc->r_result[fdst->DstRegister.Index];
+	case TGSI_FILE_TEMPORARY:
+		return fpc->r_temp[fdst->DstRegister.Index];
+	case TGSI_FILE_NULL:
+		return nv40_sr(NV40SR_NONE, 0);
+	default:
+		NOUVEAU_ERR("bad dst file %d\n", fdst->DstRegister.File);
+		return nv40_sr(NV40SR_NONE, 0);
+	}
+}
+
+static INLINE int
+tgsi_mask(uint tgsi)
+{
+	int mask = 0;
+
+	if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X;
+	if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y;
+	if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z;
+	if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W;
+	return mask;
+}
+
+static boolean
+src_native_swz(struct nv40_fpc *fpc, const struct tgsi_full_src_register *fsrc,
+	       struct nv40_sreg *src)
+{
+	const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0);
+	struct nv40_sreg tgsi = tgsi_src(fpc, fsrc);
+	uint mask = 0, zero_mask = 0, one_mask = 0, neg_mask = 0;
+	uint neg[4] = { fsrc->SrcRegisterExtSwz.NegateX,
+			fsrc->SrcRegisterExtSwz.NegateY,
+			fsrc->SrcRegisterExtSwz.NegateZ,
+			fsrc->SrcRegisterExtSwz.NegateW };
+	uint c;
+
+	for (c = 0; c < 4; c++) {
+		switch (tgsi_util_get_full_src_register_extswizzle(fsrc, c)) {
+		case TGSI_EXTSWIZZLE_X:
+		case TGSI_EXTSWIZZLE_Y:
+		case TGSI_EXTSWIZZLE_Z:
+		case TGSI_EXTSWIZZLE_W:
+			mask |= (1 << c);
+			break;
+		case TGSI_EXTSWIZZLE_ZERO:
+			zero_mask |= (1 << c);
+			tgsi.swz[c] = SWZ_X;
+			break;
+		case TGSI_EXTSWIZZLE_ONE:
+			one_mask |= (1 << c);
+			tgsi.swz[c] = SWZ_X;
+			break;
+		default:
+			assert(0);
+		}
+
+		if (!tgsi.negate && neg[c])
+			neg_mask |= (1 << c);
+	}
+
+	if (mask == MASK_ALL && !neg_mask)
+		return TRUE;
+
+	*src = temp(fpc);
+
+	if (mask)
+		arith(fpc, 0, MOV, *src, mask, tgsi, none, none);
+
+	if (zero_mask)
+		arith(fpc, 0, SFL, *src, zero_mask, *src, none, none);
+
+	if (one_mask)
+		arith(fpc, 0, STR, *src, one_mask, *src, none, none);
+
+	if (neg_mask) {
+		struct nv40_sreg one = temp(fpc);
+		arith(fpc, 0, STR, one, neg_mask, one, none, none);
+		arith(fpc, 0, MUL, *src, neg_mask, *src, neg(one), none);
+	}
+
+	return FALSE;
+}
+
+static boolean
+nv40_fragprog_parse_instruction(struct nv40_fpc *fpc,
+				const struct tgsi_full_instruction *finst)
+{
+	const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0);
+	struct nv40_sreg src[3], dst, tmp;
+	int mask, sat, unit;
+	int ai = -1, ci = -1, ii = -1;
+	int i;
+
+	if (finst->Instruction.Opcode == TGSI_OPCODE_END)
+		return TRUE;
+
+	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fsrc;
+
+		fsrc = &finst->FullSrcRegisters[i];
+		if (fsrc->SrcRegister.File == TGSI_FILE_TEMPORARY) {
+			src[i] = tgsi_src(fpc, fsrc);
+		}
+	}
+
+	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fsrc;
+
+		fsrc = &finst->FullSrcRegisters[i];
+
+		switch (fsrc->SrcRegister.File) {
+		case TGSI_FILE_INPUT:
+		case TGSI_FILE_CONSTANT:
+		case TGSI_FILE_TEMPORARY:
+			if (!src_native_swz(fpc, fsrc, &src[i]))
+				continue;
+			break;
+		default:
+			break;
+		}
+
+		switch (fsrc->SrcRegister.File) {
+		case TGSI_FILE_INPUT:
+			if (ai == -1 || ai == fsrc->SrcRegister.Index) {
+				ai = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(fpc, fsrc);
+			} else {
+				src[i] = temp(fpc);
+				arith(fpc, 0, MOV, src[i], MASK_ALL,
+				      tgsi_src(fpc, fsrc), none, none);
+			}
+			break;
+		case TGSI_FILE_CONSTANT:
+			if ((ci == -1 && ii == -1) ||
+			    ci == fsrc->SrcRegister.Index) {
+				ci = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(fpc, fsrc);
+			} else {
+				src[i] = temp(fpc);
+				arith(fpc, 0, MOV, src[i], MASK_ALL,
+				      tgsi_src(fpc, fsrc), none, none);
+			}
+			break;
+		case TGSI_FILE_IMMEDIATE:
+			if ((ci == -1 && ii == -1) ||
+			    ii == fsrc->SrcRegister.Index) {
+				ii = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(fpc, fsrc);
+			} else {
+				src[i] = temp(fpc);
+				arith(fpc, 0, MOV, src[i], MASK_ALL,
+				      tgsi_src(fpc, fsrc), none, none);
+			}
+			break;
+		case TGSI_FILE_TEMPORARY:
+			/* handled above */
+			break;
+		case TGSI_FILE_SAMPLER:
+			unit = fsrc->SrcRegister.Index;
+			break;
+		case TGSI_FILE_OUTPUT:
+			break;
+		default:
+			NOUVEAU_ERR("bad src file\n");
+			return FALSE;
+		}
+	}
+
+	dst  = tgsi_dst(fpc, &finst->FullDstRegisters[0]);
+	mask = tgsi_mask(finst->FullDstRegisters[0].DstRegister.WriteMask);
+	sat  = (finst->Instruction.Saturate == TGSI_SAT_ZERO_ONE);
+
+	switch (finst->Instruction.Opcode) {
+	case TGSI_OPCODE_ABS:
+		arith(fpc, sat, MOV, dst, mask, abs(src[0]), none, none);
+		break;
+	case TGSI_OPCODE_ADD:
+		arith(fpc, sat, ADD, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_CMP:
+		tmp = temp(fpc);
+		arith(fpc, sat, MOV, dst, mask, src[2], none, none);
+		tmp.cc_update = 1;
+		arith(fpc, 0, MOV, tmp, 0xf, src[0], none, none);
+		dst.cc_test = NV40_VP_INST_COND_LT;
+		arith(fpc, sat, MOV, dst, mask, src[1], none, none);
+		break;
+	case TGSI_OPCODE_COS:
+		arith(fpc, sat, COS, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_DDX:
+		if (mask & (MASK_Z | MASK_W)) {
+			tmp = temp(fpc);
+			arith(fpc, sat, DDX, tmp, MASK_X | MASK_Y,
+			      swz(src[0], Z, W, Z, W), none, none);
+			arith(fpc, 0, MOV, tmp, MASK_Z | MASK_W,
+			      swz(tmp, X, Y, X, Y), none, none);
+			arith(fpc, sat, DDX, tmp, MASK_X | MASK_Y, src[0],
+			      none, none);
+			arith(fpc, 0, MOV, dst, mask, tmp, none, none);
+		} else {
+			arith(fpc, sat, DDX, dst, mask, src[0], none, none);
+		}
+		break;
+	case TGSI_OPCODE_DDY:
+		if (mask & (MASK_Z | MASK_W)) {
+			tmp = temp(fpc);
+			arith(fpc, sat, DDY, tmp, MASK_X | MASK_Y,
+			      swz(src[0], Z, W, Z, W), none, none);
+			arith(fpc, 0, MOV, tmp, MASK_Z | MASK_W,
+			      swz(tmp, X, Y, X, Y), none, none);
+			arith(fpc, sat, DDY, tmp, MASK_X | MASK_Y, src[0],
+			      none, none);
+			arith(fpc, 0, MOV, dst, mask, tmp, none, none);
+		} else {
+			arith(fpc, sat, DDY, dst, mask, src[0], none, none);
+		}
+		break;
+	case TGSI_OPCODE_DP3:
+		arith(fpc, sat, DP3, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DP4:
+		arith(fpc, sat, DP4, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DPH:
+		tmp = temp(fpc);
+		arith(fpc, 0, DP3, tmp, MASK_X, src[0], src[1], none);
+		arith(fpc, sat, ADD, dst, mask, swz(tmp, X, X, X, X),
+		      swz(src[1], W, W, W, W), none);
+		break;
+	case TGSI_OPCODE_DST:
+		arith(fpc, sat, DST, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_EX2:
+		arith(fpc, sat, EX2, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_FLR:
+		arith(fpc, sat, FLR, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_FRC:
+		arith(fpc, sat, FRC, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_KILP:
+		arith(fpc, 0, KIL, none, 0, none, none, none);
+		break;
+	case TGSI_OPCODE_KIL:
+		dst = nv40_sr(NV40SR_NONE, 0);
+		dst.cc_update = 1;
+		arith(fpc, 0, MOV, dst, MASK_ALL, src[0], none, none);
+		dst.cc_update = 0; dst.cc_test = NV40_FP_OP_COND_LT;
+		arith(fpc, 0, KIL, dst, 0, none, none, none);
+		break;
+	case TGSI_OPCODE_LG2:
+		arith(fpc, sat, LG2, dst, mask, src[0], none, none);
+		break;
+//	case TGSI_OPCODE_LIT:
+	case TGSI_OPCODE_LRP:
+		tmp = temp(fpc);
+		arith(fpc, 0, MAD, tmp, mask, neg(src[0]), src[2], src[2]);
+		arith(fpc, sat, MAD, dst, mask, src[0], src[1], tmp);
+		break;
+	case TGSI_OPCODE_MAD:
+		arith(fpc, sat, MAD, dst, mask, src[0], src[1], src[2]);
+		break;
+	case TGSI_OPCODE_MAX:
+		arith(fpc, sat, MAX, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_MIN:
+		arith(fpc, sat, MIN, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_MOV:
+		arith(fpc, sat, MOV, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_MUL:
+		arith(fpc, sat, MUL, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_NOISE1:
+	case TGSI_OPCODE_NOISE2:
+	case TGSI_OPCODE_NOISE3:
+	case TGSI_OPCODE_NOISE4:
+		arith(fpc, sat, SFL, dst, mask, none, none, none);
+		break;
+	case TGSI_OPCODE_POW:
+		tmp = temp(fpc);
+		arith(fpc, 0, LG2, tmp, MASK_X,
+		      swz(src[0], X, X, X, X), none, none);
+		arith(fpc, 0, MUL, tmp, MASK_X, swz(tmp, X, X, X, X),
+		      swz(src[1], X, X, X, X), none);
+		arith(fpc, sat, EX2, dst, mask,
+		      swz(tmp, X, X, X, X), none, none);
+		break;
+	case TGSI_OPCODE_RCP:
+		arith(fpc, sat, RCP, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_RET:
+		assert(0);
+		break;
+	case TGSI_OPCODE_RFL:
+		tmp = temp(fpc);
+		arith(fpc, 0, DP3, tmp, MASK_X, src[0], src[0], none);
+		arith(fpc, 0, DP3, tmp, MASK_Y, src[0], src[1], none);
+		arith(fpc, 0, DIV, scale(tmp, 2X), MASK_Z,
+		      swz(tmp, Y, Y, Y, Y), swz(tmp, X, X, X, X), none);
+		arith(fpc, sat, MAD, dst, mask,
+		      swz(tmp, Z, Z, Z, Z), src[0], neg(src[1]));
+		break;
+	case TGSI_OPCODE_RSQ:
+		tmp = temp(fpc);
+		arith(fpc, 0, LG2, scale(tmp, INV_2X), MASK_X,
+		      abs(swz(src[0], X, X, X, X)), none, none);
+		arith(fpc, sat, EX2, dst, mask,
+		      neg(swz(tmp, X, X, X, X)), none, none);
+		break;
+	case TGSI_OPCODE_SCS:
+		if (mask & MASK_X) {
+			arith(fpc, sat, COS, dst, MASK_X,
+			      swz(src[0], X, X, X, X), none, none);
+		}
+		if (mask & MASK_Y) {
+			arith(fpc, sat, SIN, dst, MASK_Y,
+			      swz(src[0], X, X, X, X), none, none);
+		}
+		break;
+	case TGSI_OPCODE_SEQ:
+		arith(fpc, sat, SEQ, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SFL:
+		arith(fpc, sat, SFL, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SGE:
+		arith(fpc, sat, SGE, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SGT:
+		arith(fpc, sat, SGT, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SIN:
+		arith(fpc, sat, SIN, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_SLE:
+		arith(fpc, sat, SLE, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SLT:
+		arith(fpc, sat, SLT, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SNE:
+		arith(fpc, sat, SNE, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_STR:
+		arith(fpc, sat, STR, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SUB:
+		arith(fpc, sat, ADD, dst, mask, src[0], neg(src[1]), none);
+		break;
+	case TGSI_OPCODE_TEX:
+		tex(fpc, sat, TEX, unit, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_TXB:
+		tex(fpc, sat, TXB, unit, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_TXP:
+		tex(fpc, sat, TXP, unit, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_XPD:
+		tmp = temp(fpc);
+		arith(fpc, 0, MUL, tmp, mask,
+		      swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none);
+		arith(fpc, sat, MAD, dst, (mask & ~MASK_W),
+		      swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y),
+		      neg(tmp));
+		break;
+	default:
+		NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode);
+		return FALSE;
+	}
+
+	release_temps(fpc);
+	return TRUE;
+}
+
+static boolean
+nv40_fragprog_parse_decl_attrib(struct nv40_fpc *fpc,
+				const struct tgsi_full_declaration *fdec)
+{
+	int hw;
+
+	switch (fdec->Semantic.SemanticName) {
+	case TGSI_SEMANTIC_POSITION:
+		hw = NV40_FP_OP_INPUT_SRC_POSITION;
+		break;
+	case TGSI_SEMANTIC_COLOR:
+		if (fdec->Semantic.SemanticIndex == 0) {
+			hw = NV40_FP_OP_INPUT_SRC_COL0;
+		} else
+		if (fdec->Semantic.SemanticIndex == 1) {
+			hw = NV40_FP_OP_INPUT_SRC_COL1;
+		} else {
+			NOUVEAU_ERR("bad colour semantic index\n");
+			return FALSE;
+		}
+		break;
+	case TGSI_SEMANTIC_FOG:
+		hw = NV40_FP_OP_INPUT_SRC_FOGC;
+		break;
+	case TGSI_SEMANTIC_GENERIC:
+		if (fdec->Semantic.SemanticIndex <= 7) {
+			hw = NV40_FP_OP_INPUT_SRC_TC(fdec->Semantic.
+						     SemanticIndex);
+		} else {
+			NOUVEAU_ERR("bad generic semantic index\n");
+			return FALSE;
+		}
+		break;
+	default:
+		NOUVEAU_ERR("bad input semantic\n");
+		return FALSE;
+	}
+
+	fpc->attrib_map[fdec->DeclarationRange.First] = hw;
+	return TRUE;
+}
+
+static boolean
+nv40_fragprog_parse_decl_output(struct nv40_fpc *fpc,
+				const struct tgsi_full_declaration *fdec)
+{
+	unsigned idx = fdec->DeclarationRange.First;
+	unsigned hw;
+
+	switch (fdec->Semantic.SemanticName) {
+	case TGSI_SEMANTIC_POSITION:
+		hw = 1;
+		break;
+	case TGSI_SEMANTIC_COLOR:
+		switch (fdec->Semantic.SemanticIndex) {
+		case 0: hw = 0; break;
+		case 1: hw = 2; break;
+		case 2: hw = 3; break;
+		case 3: hw = 4; break;
+		default:
+			NOUVEAU_ERR("bad rcol index\n");
+			return FALSE;
+		}
+		break;
+	default:
+		NOUVEAU_ERR("bad output semantic\n");
+		return FALSE;
+	}
+
+	fpc->r_result[idx] = nv40_sr(NV40SR_OUTPUT, hw);
+	fpc->r_temps |= (1 << hw);
+	return TRUE;
+}
+
+static boolean
+nv40_fragprog_prepare(struct nv40_fpc *fpc)
+{
+	struct tgsi_parse_context p;
+	int high_temp = -1, i;
+
+	tgsi_parse_init(&p, fpc->fp->pipe.tokens);
+	while (!tgsi_parse_end_of_tokens(&p)) {
+		const union tgsi_full_token *tok = &p.FullToken;
+
+		tgsi_parse_token(&p);
+		switch(tok->Token.Type) {
+		case TGSI_TOKEN_TYPE_DECLARATION:
+		{
+			const struct tgsi_full_declaration *fdec;
+			fdec = &p.FullToken.FullDeclaration;
+			switch (fdec->Declaration.File) {
+			case TGSI_FILE_INPUT:
+				if (!nv40_fragprog_parse_decl_attrib(fpc, fdec))
+					goto out_err;
+				break;
+			case TGSI_FILE_OUTPUT:
+				if (!nv40_fragprog_parse_decl_output(fpc, fdec))
+					goto out_err;
+				break;
+			case TGSI_FILE_TEMPORARY:
+				if (fdec->DeclarationRange.Last > high_temp) {
+					high_temp =
+						fdec->DeclarationRange.Last;
+				}
+				break;
+			default:
+				break;
+			}
+		}
+			break;
+		case TGSI_TOKEN_TYPE_IMMEDIATE:
+		{
+			struct tgsi_full_immediate *imm;
+			float vals[4];
+			
+			imm = &p.FullToken.FullImmediate;
+			assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32);
+			assert(fpc->nr_imm < MAX_IMM);
+
+			vals[0] = imm->u.ImmediateFloat32[0].Float;
+			vals[1] = imm->u.ImmediateFloat32[1].Float;
+			vals[2] = imm->u.ImmediateFloat32[2].Float;
+			vals[3] = imm->u.ImmediateFloat32[3].Float;
+			fpc->imm[fpc->nr_imm++] = constant(fpc, -1, vals);
+		}
+			break;
+		default:
+			break;
+		}
+	}
+	tgsi_parse_free(&p);
+
+	if (++high_temp) {
+		fpc->r_temp = CALLOC(high_temp, sizeof(struct nv40_sreg));
+		for (i = 0; i < high_temp; i++)
+			fpc->r_temp[i] = temp(fpc);
+		fpc->r_temps_discard = 0;
+	}
+
+	return TRUE;
+
+out_err:
+	if (fpc->r_temp)
+		FREE(fpc->r_temp);
+	tgsi_parse_free(&p);
+	return FALSE;
+}
+
+static void
+nv40_fragprog_translate(struct nv40_context *nv40,
+			struct nv40_fragment_program *fp)
+{
+	struct tgsi_parse_context parse;
+	struct nv40_fpc *fpc = NULL;
+
+	fpc = CALLOC(1, sizeof(struct nv40_fpc));
+	if (!fpc)
+		return;
+	fpc->fp = fp;
+	fpc->num_regs = 2;
+
+	if (!nv40_fragprog_prepare(fpc)) {
+		FREE(fpc);
+		return;
+	}
+
+	tgsi_parse_init(&parse, fp->pipe.tokens);
+
+	while (!tgsi_parse_end_of_tokens(&parse)) {
+		tgsi_parse_token(&parse);
+
+		switch (parse.FullToken.Token.Type) {
+		case TGSI_TOKEN_TYPE_INSTRUCTION:
+		{
+			const struct tgsi_full_instruction *finst;
+
+			finst = &parse.FullToken.FullInstruction;
+			if (!nv40_fragprog_parse_instruction(fpc, finst))
+				goto out_err;
+		}
+			break;
+		default:
+			break;
+		}
+	}
+
+	fp->fp_control |= fpc->num_regs << NV40TCL_FP_CONTROL_TEMP_COUNT_SHIFT;
+
+	/* Terminate final instruction */
+	fp->insn[fpc->inst_offset] |= 0x00000001;
+
+	/* Append NOP + END instruction, may or may not be necessary. */
+	fpc->inst_offset = fp->insn_len;
+	grow_insns(fpc, 4);
+	fp->insn[fpc->inst_offset + 0] = 0x00000001;
+	fp->insn[fpc->inst_offset + 1] = 0x00000000;
+	fp->insn[fpc->inst_offset + 2] = 0x00000000;
+	fp->insn[fpc->inst_offset + 3] = 0x00000000;
+	
+	fp->translated = TRUE;
+out_err:
+	tgsi_parse_free(&parse);
+	if (fpc->r_temp)
+		FREE(fpc->r_temp);
+	FREE(fpc);
+}
+
+static void
+nv40_fragprog_upload(struct nv40_context *nv40,
+		     struct nv40_fragment_program *fp)
+{
+	struct pipe_winsys *ws = nv40->pipe.winsys;
+	const uint32_t le = 1;
+	uint32_t *map;
+	int i;
+
+	map = ws->buffer_map(ws, fp->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+#if 0
+	for (i = 0; i < fp->insn_len; i++) {
+		fflush(stdout); fflush(stderr);
+		NOUVEAU_ERR("%d 0x%08x\n", i, fp->insn[i]);
+		fflush(stdout); fflush(stderr);
+	}
+#endif
+
+	if ((*(const uint8_t *)&le)) {
+		for (i = 0; i < fp->insn_len; i++) {
+			map[i] = fp->insn[i];
+		}
+	} else {
+		/* Weird swapping for big-endian chips */
+		for (i = 0; i < fp->insn_len; i++) {
+			map[i] = ((fp->insn[i] & 0xffff) << 16) |
+				  ((fp->insn[i] >> 16) & 0xffff);
+		}
+	}
+
+	ws->buffer_unmap(ws, fp->buffer);
+}
+
+static boolean
+nv40_fragprog_validate(struct nv40_context *nv40)
+{
+	struct nv40_fragment_program *fp = nv40->fragprog;
+	struct pipe_buffer *constbuf =
+		nv40->constbuf[PIPE_SHADER_FRAGMENT];
+	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct nouveau_stateobj *so;
+	boolean new_consts = FALSE;
+	int i;
+
+	if (fp->translated)
+		goto update_constants;
+
+	nv40->fallback_swrast &= ~NV40_NEW_FRAGPROG;
+	nv40_fragprog_translate(nv40, fp);
+	if (!fp->translated) {
+		nv40->fallback_swrast |= NV40_NEW_FRAGPROG;
+		return FALSE;
+	}
+
+	fp->buffer = ws->buffer_create(ws, 0x100, 0, fp->insn_len * 4);
+	nv40_fragprog_upload(nv40, fp);
+
+	so = so_new(4, 1);
+	so_method(so, nv40->screen->curie, NV40TCL_FP_ADDRESS, 1);
+	so_reloc (so, fp->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
+		  NV40TCL_FP_ADDRESS_DMA0, NV40TCL_FP_ADDRESS_DMA1);
+	so_method(so, nv40->screen->curie, NV40TCL_FP_CONTROL, 1);
+	so_data  (so, fp->fp_control);
+	so_ref(so, &fp->so);
+
+update_constants:
+	if (fp->nr_consts) {
+		float *map;
+		
+		map = ws->buffer_map(ws, constbuf, PIPE_BUFFER_USAGE_CPU_READ);
+		for (i = 0; i < fp->nr_consts; i++) {
+			struct nv40_fragment_program_data *fpd = &fp->consts[i];
+			uint32_t *p = &fp->insn[fpd->offset];
+			uint32_t *cb = (uint32_t *)&map[fpd->index * 4];
+
+			if (!memcmp(p, cb, 4 * sizeof(float)))
+				continue;
+			memcpy(p, cb, 4 * sizeof(float));
+			new_consts = TRUE;
+		}
+		ws->buffer_unmap(ws, constbuf);
+
+		if (new_consts)
+			nv40_fragprog_upload(nv40, fp);
+	}
+
+	if (new_consts || fp->so != nv40->state.hw[NV40_STATE_FRAGPROG]) {
+		so_ref(fp->so, &nv40->state.hw[NV40_STATE_FRAGPROG]);
+		return TRUE;
+	}
+
+	return FALSE;
+}
+
+void
+nv40_fragprog_destroy(struct nv40_context *nv40,
+		      struct nv40_fragment_program *fp)
+{
+	if (fp->insn_len)
+		FREE(fp->insn);
+}
+
+struct nv40_state_entry nv40_state_fragprog = {
+	.validate = nv40_fragprog_validate,
+	.dirty = {
+		.pipe = NV40_NEW_FRAGPROG,
+		.hw = NV40_STATE_FRAGPROG
+	}
+};
+
diff --git a/src/gallium/drivers/nv40/nv40_fragtex.c b/src/gallium/drivers/nv40/nv40_fragtex.c
new file mode 100644
index 0000000000..0227d22620
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_fragtex.c
@@ -0,0 +1,168 @@
+#include "nv40_context.h"
+
+#define _(m,tf,ts0x,ts0y,ts0z,ts0w,ts1x,ts1y,ts1z,ts1w,sx,sy,sz,sw)            \
+{                                                                              \
+  TRUE,                                                                        \
+  PIPE_FORMAT_##m,                                                             \
+  NV40TCL_TEX_FORMAT_FORMAT_##tf,                                              \
+  (NV40TCL_TEX_SWIZZLE_S0_X_##ts0x | NV40TCL_TEX_SWIZZLE_S0_Y_##ts0y |         \
+   NV40TCL_TEX_SWIZZLE_S0_Z_##ts0z | NV40TCL_TEX_SWIZZLE_S0_W_##ts0w |         \
+   NV40TCL_TEX_SWIZZLE_S1_X_##ts1x | NV40TCL_TEX_SWIZZLE_S1_Y_##ts1y |         \
+   NV40TCL_TEX_SWIZZLE_S1_Z_##ts1z | NV40TCL_TEX_SWIZZLE_S1_W_##ts1w),         \
+  ((NV40TCL_TEX_FILTER_SIGNED_RED*sx) | (NV40TCL_TEX_FILTER_SIGNED_GREEN*sy) |       \
+   (NV40TCL_TEX_FILTER_SIGNED_BLUE*sz) | (NV40TCL_TEX_FILTER_SIGNED_ALPHA*sw))       \
+}
+
+struct nv40_texture_format {
+	boolean defined;
+	uint	pipe;
+	int     format;
+	int     swizzle;
+	int     sign;
+};
+
+static struct nv40_texture_format
+nv40_texture_formats[] = {
+	_(A8R8G8B8_UNORM, A8R8G8B8,   S1,   S1,   S1,   S1, X, Y, Z, W, 0, 0, 0, 0),
+	_(A1R5G5B5_UNORM, A1R5G5B5,   S1,   S1,   S1,   S1, X, Y, Z, W, 0, 0, 0, 0),
+	_(A4R4G4B4_UNORM, A4R4G4B4,   S1,   S1,   S1,   S1, X, Y, Z, W, 0, 0, 0, 0),
+	_(R5G6B5_UNORM  , R5G6B5  ,   S1,   S1,   S1,  ONE, X, Y, Z, W, 0, 0, 0, 0),
+	_(L8_UNORM      , L8      ,   S1,   S1,   S1,  ONE, X, X, X, X, 0, 0, 0, 0),
+	_(A8_UNORM      , L8      , ZERO, ZERO, ZERO,   S1, X, X, X, X, 0, 0, 0, 0),
+	_(R16_SNORM     , A16     , ZERO, ZERO,   S1,  ONE, X, X, X, Y, 1, 1, 1, 1),
+	_(I8_UNORM      , L8      ,   S1,   S1,   S1,   S1, X, X, X, X, 0, 0, 0, 0),
+	_(A8L8_UNORM    , A8L8    ,   S1,   S1,   S1,   S1, X, X, X, Y, 0, 0, 0, 0),
+	_(Z16_UNORM     , Z16     ,   S1,   S1,   S1,  ONE, X, X, X, X, 0, 0, 0, 0),
+	_(Z24S8_UNORM   , Z24     ,   S1,   S1,   S1,  ONE, X, X, X, X, 0, 0, 0, 0),
+	_(DXT1_RGB      , DXT1    ,   S1,   S1,   S1,  ONE, X, Y, Z, W, 0, 0, 0, 0),
+	_(DXT1_RGBA     , DXT1    ,   S1,   S1,   S1,   S1, X, Y, Z, W, 0, 0, 0, 0),
+	_(DXT3_RGBA     , DXT3    ,   S1,   S1,   S1,   S1, X, Y, Z, W, 0, 0, 0, 0),
+	_(DXT5_RGBA     , DXT5    ,   S1,   S1,   S1,   S1, X, Y, Z, W, 0, 0, 0, 0),
+	{},
+};
+
+static struct nv40_texture_format *
+nv40_fragtex_format(uint pipe_format)
+{
+	struct nv40_texture_format *tf = nv40_texture_formats;
+
+	while (tf->defined) {
+		if (tf->pipe == pipe_format)
+			return tf;
+		tf++;
+	}
+
+	NOUVEAU_ERR("unknown texture format %s\n", pf_name(pipe_format));
+	return NULL;
+}
+
+
+static struct nouveau_stateobj *
+nv40_fragtex_build(struct nv40_context *nv40, int unit)
+{
+	struct nv40_sampler_state *ps = nv40->tex_sampler[unit];
+	struct nv40_miptree *nv40mt = nv40->tex_miptree[unit];
+	struct pipe_texture *pt = &nv40mt->base;
+	struct nv40_texture_format *tf;
+	struct nouveau_stateobj *so;
+	uint32_t txf, txs, txp;
+	unsigned tex_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD;
+
+	tf = nv40_fragtex_format(pt->format);
+	if (!tf)
+		assert(0);
+
+	txf  = ps->fmt;
+	txf |= tf->format | 0x8000;
+	txf |= ((pt->last_level + 1) << NV40TCL_TEX_FORMAT_MIPMAP_COUNT_SHIFT);
+
+	if (1) /* XXX */
+		txf |= NV40TCL_TEX_FORMAT_NO_BORDER;
+
+	switch (pt->target) {
+	case PIPE_TEXTURE_CUBE:
+		txf |= NV40TCL_TEX_FORMAT_CUBIC;
+		/* fall-through */
+	case PIPE_TEXTURE_2D:
+		txf |= NV40TCL_TEX_FORMAT_DIMS_2D;
+		break;
+	case PIPE_TEXTURE_3D:
+		txf |= NV40TCL_TEX_FORMAT_DIMS_3D;
+		break;
+	case PIPE_TEXTURE_1D:
+		txf |= NV40TCL_TEX_FORMAT_DIMS_1D;
+		break;
+	default:
+		NOUVEAU_ERR("Unknown target %d\n", pt->target);
+		return NULL;
+	}
+
+	if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
+		txp = 0;
+	} else {
+		txp  = nv40mt->level[0].pitch;
+		txf |= NV40TCL_TEX_FORMAT_LINEAR;
+	}
+
+	txs = tf->swizzle;
+
+	so = so_new(16, 2);
+	so_method(so, nv40->screen->curie, NV40TCL_TEX_OFFSET(unit), 8);
+	so_reloc (so, nv40mt->buffer, 0, tex_flags | NOUVEAU_BO_LOW, 0, 0);
+	so_reloc (so, nv40mt->buffer, txf, tex_flags | NOUVEAU_BO_OR,
+		  NV40TCL_TEX_FORMAT_DMA0, NV40TCL_TEX_FORMAT_DMA1);
+	so_data  (so, ps->wrap);
+	so_data  (so, NV40TCL_TEX_ENABLE_ENABLE | ps->en);
+	so_data  (so, txs);
+	so_data  (so, ps->filt | tf->sign | 0x2000 /*voodoo*/);
+	so_data  (so, (pt->width[0] << NV40TCL_TEX_SIZE0_W_SHIFT) |
+		       pt->height[0]);
+	so_data  (so, ps->bcol);
+	so_method(so, nv40->screen->curie, NV40TCL_TEX_SIZE1(unit), 1);
+	so_data  (so, (pt->depth[0] << NV40TCL_TEX_SIZE1_DEPTH_SHIFT) | txp);
+
+	return so;
+}
+
+static boolean
+nv40_fragtex_validate(struct nv40_context *nv40)
+{
+	struct nv40_fragment_program *fp = nv40->fragprog;
+	struct nv40_state *state = &nv40->state;
+	struct nouveau_stateobj *so;
+	unsigned samplers, unit;
+
+	samplers = state->fp_samplers & ~fp->samplers;
+	while (samplers) {
+		unit = ffs(samplers) - 1;
+		samplers &= ~(1 << unit);
+
+		so = so_new(2, 0);
+		so_method(so, nv40->screen->curie, NV40TCL_TEX_ENABLE(unit), 1);
+		so_data  (so, 0);
+		so_ref(so, &nv40->state.hw[NV40_STATE_FRAGTEX0 + unit]);
+		state->dirty |= (1ULL << (NV40_STATE_FRAGTEX0 + unit));
+	}
+
+	samplers = nv40->dirty_samplers & fp->samplers;
+	while (samplers) {
+		unit = ffs(samplers) - 1;
+		samplers &= ~(1 << unit);
+
+		so = nv40_fragtex_build(nv40, unit);
+		so_ref(so, &nv40->state.hw[NV40_STATE_FRAGTEX0 + unit]);
+		state->dirty |= (1ULL << (NV40_STATE_FRAGTEX0 + unit));
+	}
+
+	nv40->state.fp_samplers = fp->samplers;
+	return FALSE;
+}
+
+struct nv40_state_entry nv40_state_fragtex = {
+	.validate = nv40_fragtex_validate,
+	.dirty = {
+		.pipe = NV40_NEW_SAMPLER | NV40_NEW_FRAGPROG,
+		.hw = 0
+	}
+};
+
diff --git a/src/gallium/drivers/nv40/nv40_miptree.c b/src/gallium/drivers/nv40/nv40_miptree.c
new file mode 100644
index 0000000000..e38b1e7f5c
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_miptree.c
@@ -0,0 +1,238 @@
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "nv40_context.h"
+
+static void
+nv40_miptree_layout(struct nv40_miptree *mt)
+{
+	struct pipe_texture *pt = &mt->base;
+	uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0];
+	uint offset = 0;
+	int nr_faces, l, f;
+	uint wide_pitch = pt->tex_usage & (PIPE_TEXTURE_USAGE_SAMPLER |
+		                           PIPE_TEXTURE_USAGE_DEPTH_STENCIL |
+		                           PIPE_TEXTURE_USAGE_RENDER_TARGET |
+		                           PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+		                           PIPE_TEXTURE_USAGE_PRIMARY);
+
+	if (pt->target == PIPE_TEXTURE_CUBE) {
+		nr_faces = 6;
+	} else
+	if (pt->target == PIPE_TEXTURE_3D) {
+		nr_faces = pt->depth[0];
+	} else {
+		nr_faces = 1;
+	}
+
+	for (l = 0; l <= pt->last_level; l++) {
+		pt->width[l] = width;
+		pt->height[l] = height;
+		pt->depth[l] = depth;
+		pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width);
+		pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height);
+
+		if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR))
+			mt->level[l].pitch = align(pt->width[0] * pt->block.size, 64);
+		else
+			mt->level[l].pitch = pt->width[l] * pt->block.size;
+
+		mt->level[l].image_offset =
+			CALLOC(nr_faces, sizeof(unsigned));
+
+		width  = MAX2(1, width  >> 1);
+		height = MAX2(1, height >> 1);
+		depth  = MAX2(1, depth  >> 1);
+	}
+
+	for (f = 0; f < nr_faces; f++) {
+		for (l = 0; l < pt->last_level; l++) {
+			mt->level[l].image_offset[f] = offset;
+
+			if (!(pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR) &&
+			    pt->width[l + 1] > 1 && pt->height[l + 1] > 1)
+				offset += align(mt->level[l].pitch * pt->height[l], 64);
+			else
+				offset += mt->level[l].pitch * pt->height[l];
+		}
+
+		mt->level[l].image_offset[f] = offset;
+		offset += mt->level[l].pitch * pt->height[l];
+	}
+
+	mt->total_size = offset;
+}
+
+static struct pipe_texture *
+nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
+{
+	struct pipe_winsys *ws = pscreen->winsys;
+	struct nv40_miptree *mt;
+	unsigned buf_usage = PIPE_BUFFER_USAGE_PIXEL |
+	                     NOUVEAU_BUFFER_USAGE_TEXTURE;
+
+	mt = MALLOC(sizeof(struct nv40_miptree));
+	if (!mt)
+		return NULL;
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = pscreen;
+	mt->shadow_tex = NULL;
+	mt->shadow_surface = NULL;
+
+	/* Swizzled textures must be POT */
+	if (pt->width[0] & (pt->width[0] - 1) ||
+	    pt->height[0] & (pt->height[0] - 1))
+		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+	else
+	if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY |
+	                     PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+	                     PIPE_TEXTURE_USAGE_DEPTH_STENCIL))
+		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+	else
+	if (pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC)
+		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+	else {
+		switch (pt->format) {
+		/* TODO: Figure out which formats can be swizzled */
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_X8R8G8B8_UNORM:
+		case PIPE_FORMAT_R16_SNORM:
+		{
+			if (debug_get_bool_option("NOUVEAU_NO_SWIZZLE", FALSE))
+				mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+			break;
+		}
+		default:
+			mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
+		}
+	}
+
+	if (pt->tex_usage & PIPE_TEXTURE_USAGE_DYNAMIC)
+		buf_usage |= PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+
+	nv40_miptree_layout(mt);
+
+	mt->buffer = ws->buffer_create(ws, 256, buf_usage, mt->total_size);
+	if (!mt->buffer) {
+		FREE(mt);
+		return NULL;
+	}
+
+	return &mt->base;
+}
+
+static struct pipe_texture *
+nv40_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
+		     const unsigned *stride, struct pipe_buffer *pb)
+{
+	struct nv40_miptree *mt;
+
+	/* Only supports 2D, non-mipmapped textures for the moment */
+	if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 ||
+	    pt->depth[0] != 1)
+		return NULL;
+
+	mt = CALLOC_STRUCT(nv40_miptree);
+	if (!mt)
+		return NULL;
+
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = pscreen;
+	mt->level[0].pitch = stride[0];
+	mt->level[0].image_offset = CALLOC(1, sizeof(unsigned));
+
+	pipe_buffer_reference(pscreen, &mt->buffer, pb);
+	return &mt->base;
+}
+
+static void
+nv40_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt)
+{
+	struct pipe_texture *pt = *ppt;
+	struct nv40_miptree *mt = (struct nv40_miptree *)pt;
+	int l;
+
+	*ppt = NULL;
+	if (--pt->refcount)
+		return;
+
+	pipe_buffer_reference(pscreen, &mt->buffer, NULL);
+	for (l = 0; l <= pt->last_level; l++) {
+		if (mt->level[l].image_offset)
+			FREE(mt->level[l].image_offset);
+	}
+
+	if (mt->shadow_tex) {
+		if (mt->shadow_surface)
+			pscreen->tex_surface_release(pscreen, &mt->shadow_surface);
+		nv40_miptree_release(pscreen, &mt->shadow_tex);
+	}
+
+	FREE(mt);
+}
+
+static struct pipe_surface *
+nv40_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
+			 unsigned face, unsigned level, unsigned zslice,
+			 unsigned flags)
+{
+	struct nv40_miptree *mt = (struct nv40_miptree *)pt;
+	struct pipe_surface *ps;
+
+	ps = CALLOC_STRUCT(pipe_surface);
+	if (!ps)
+		return NULL;
+	pipe_texture_reference(&ps->texture, pt);
+	ps->format = pt->format;
+	ps->width = pt->width[level];
+	ps->height = pt->height[level];
+	ps->block = pt->block;
+	ps->nblocksx = pt->nblocksx[level];
+	ps->nblocksy = pt->nblocksy[level];
+	ps->stride = mt->level[level].pitch;
+	ps->usage = flags;
+	ps->status = PIPE_SURFACE_STATUS_DEFINED;
+	ps->refcount = 1;
+	ps->face = face;
+	ps->level = level;
+	ps->zslice = zslice;
+
+	if (pt->target == PIPE_TEXTURE_CUBE) {
+		ps->offset = mt->level[level].image_offset[face];
+	} else
+	if (pt->target == PIPE_TEXTURE_3D) {
+		ps->offset = mt->level[level].image_offset[zslice];
+	} else {
+		ps->offset = mt->level[level].image_offset[0];
+	}
+
+	return ps;
+}
+
+static void
+nv40_miptree_surface_del(struct pipe_screen *pscreen,
+			 struct pipe_surface **psurface)
+{
+	struct pipe_surface *ps = *psurface;
+
+	*psurface = NULL;
+	if (--ps->refcount > 0)
+		return;
+
+	pipe_texture_reference(&ps->texture, NULL);
+	FREE(ps);
+}
+
+void
+nv40_screen_init_miptree_functions(struct pipe_screen *pscreen)
+{
+	pscreen->texture_create = nv40_miptree_create;
+	pscreen->texture_blanket = nv40_miptree_blanket;
+	pscreen->texture_release = nv40_miptree_release;
+	pscreen->get_tex_surface = nv40_miptree_surface_new;
+	pscreen->tex_surface_release = nv40_miptree_surface_del;
+}
+
diff --git a/src/gallium/drivers/nv40/nv40_query.c b/src/gallium/drivers/nv40/nv40_query.c
new file mode 100644
index 0000000000..9b9a43f49d
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_query.c
@@ -0,0 +1,122 @@
+#include "pipe/p_context.h"
+
+#include "nv40_context.h"
+
+struct nv40_query {
+	struct nouveau_resource *object;
+	unsigned type;
+	boolean ready;
+	uint64_t result;
+};
+
+static INLINE struct nv40_query *
+nv40_query(struct pipe_query *pipe)
+{
+	return (struct nv40_query *)pipe;
+}
+
+static struct pipe_query *
+nv40_query_create(struct pipe_context *pipe, unsigned query_type)
+{
+	struct nv40_query *q;
+
+	q = CALLOC(1, sizeof(struct nv40_query));
+	q->type = query_type;
+
+	return (struct pipe_query *)q;
+}
+
+static void
+nv40_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv40_query *q = nv40_query(pq);
+
+	if (q->object)
+		nv40->nvws->res_free(&q->object);
+	FREE(q);
+}
+
+static void
+nv40_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv40_query *q = nv40_query(pq);
+
+	assert(q->type == PIPE_QUERY_OCCLUSION_COUNTER);
+
+	/* Happens when end_query() is called, then another begin_query()
+	 * without querying the result in-between.  For now we'll wait for
+	 * the existing query to notify completion, but it could be better.
+	 */
+	if (q->object) {
+		uint64_t tmp;
+		pipe->get_query_result(pipe, pq, 1, &tmp);
+	}
+
+	if (nv40->nvws->res_alloc(nv40->screen->query_heap, 1, NULL, &q->object))
+		assert(0);
+	nv40->nvws->notifier_reset(nv40->screen->query, q->object->start);
+
+	BEGIN_RING(curie, NV40TCL_QUERY_RESET, 1);
+	OUT_RING  (1);
+	BEGIN_RING(curie, NV40TCL_QUERY_UNK17CC, 1);
+	OUT_RING  (1);
+
+	q->ready = FALSE;
+}
+
+static void
+nv40_query_end(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv40_query *q = nv40_query(pq);
+
+	BEGIN_RING(curie, NV40TCL_QUERY_GET, 1);
+	OUT_RING  ((0x01 << NV40TCL_QUERY_GET_UNK24_SHIFT) |
+		   ((q->object->start * 32) << NV40TCL_QUERY_GET_OFFSET_SHIFT));
+	FIRE_RING(NULL);
+}
+
+static boolean
+nv40_query_result(struct pipe_context *pipe, struct pipe_query *pq,
+		  boolean wait, uint64_t *result)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv40_query *q = nv40_query(pq);
+	struct nouveau_winsys *nvws = nv40->nvws;
+
+	assert(q->object && q->type == PIPE_QUERY_OCCLUSION_COUNTER);
+
+	if (!q->ready) {
+		unsigned status;
+
+		status = nvws->notifier_status(nv40->screen->query,
+					       q->object->start);
+		if (status != NV_NOTIFY_STATE_STATUS_COMPLETED) {
+			if (wait == FALSE)
+				return FALSE;
+			nvws->notifier_wait(nv40->screen->query, q->object->start,
+					    NV_NOTIFY_STATE_STATUS_COMPLETED,
+					    0);
+		}
+
+		q->result = nvws->notifier_retval(nv40->screen->query,
+						  q->object->start);
+		q->ready = TRUE;
+		nvws->res_free(&q->object);
+	}
+
+	*result = q->result;
+	return TRUE;
+}
+
+void
+nv40_init_query_functions(struct nv40_context *nv40)
+{
+	nv40->pipe.create_query = nv40_query_create;
+	nv40->pipe.destroy_query = nv40_query_destroy;
+	nv40->pipe.begin_query = nv40_query_begin;
+	nv40->pipe.end_query = nv40_query_end;
+	nv40->pipe.get_query_result = nv40_query_result;
+}
diff --git a/src/gallium/drivers/nv40/nv40_screen.c b/src/gallium/drivers/nv40/nv40_screen.c
new file mode 100644
index 0000000000..2372bc8441
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_screen.c
@@ -0,0 +1,383 @@
+#include "pipe/p_screen.h"
+#include "util/u_simple_screen.h"
+
+#include "nv40_context.h"
+#include "nv40_screen.h"
+
+#define NV4X_GRCLASS4097_CHIPSETS 0x00000baf
+#define NV4X_GRCLASS4497_CHIPSETS 0x00005450
+#define NV6X_GRCLASS4497_CHIPSETS 0x00000088
+
+static const char *
+nv40_screen_get_name(struct pipe_screen *pscreen)
+{
+	struct nv40_screen *screen = nv40_screen(pscreen);
+	struct nouveau_device *dev = screen->nvws->channel->device;
+	static char buffer[128];
+
+	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
+	return buffer;
+}
+
+static const char *
+nv40_screen_get_vendor(struct pipe_screen *pscreen)
+{
+	return "nouveau";
+}
+
+static int
+nv40_screen_get_param(struct pipe_screen *pscreen, int param)
+{
+	struct nv40_screen *screen = nv40_screen(pscreen);
+
+	switch (param) {
+	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+		return 16;
+	case PIPE_CAP_NPOT_TEXTURES:
+		return 1;
+	case PIPE_CAP_TWO_SIDED_STENCIL:
+		return 1;
+	case PIPE_CAP_GLSL:
+		return 0;
+	case PIPE_CAP_S3TC:
+		return 1;
+	case PIPE_CAP_ANISOTROPIC_FILTER:
+		return 1;
+	case PIPE_CAP_POINT_SPRITE:
+		return 1;
+	case PIPE_CAP_MAX_RENDER_TARGETS:
+		return 4;
+	case PIPE_CAP_OCCLUSION_QUERY:
+		return 1;
+	case PIPE_CAP_TEXTURE_SHADOW_MAP:
+		return 1;
+	case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+		return 13;
+	case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+		return 10;
+	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+		return 13;
+	case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
+	case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
+		return 1;
+	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
+		return 0; /* We have 4 - but unsupported currently */
+	case NOUVEAU_CAP_HW_VTXBUF:
+		return 1;
+	case NOUVEAU_CAP_HW_IDXBUF:
+		if (screen->curie->grclass == NV40TCL)
+			return 1;
+		return 0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0;
+	}
+}
+
+static float
+nv40_screen_get_paramf(struct pipe_screen *pscreen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_LINE_WIDTH:
+	case PIPE_CAP_MAX_LINE_WIDTH_AA:
+		return 10.0;
+	case PIPE_CAP_MAX_POINT_WIDTH:
+	case PIPE_CAP_MAX_POINT_WIDTH_AA:
+		return 64.0;
+	case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+		return 16.0;
+	case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+		return 16.0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0.0;
+	}
+}
+
+static boolean
+nv40_screen_surface_format_supported(struct pipe_screen *pscreen,
+				     enum pipe_format format,
+				     enum pipe_texture_target target,
+				     unsigned tex_usage, unsigned geom_flags)
+{
+	if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM: 
+		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z16_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	} else {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_A1R5G5B5_UNORM:
+		case PIPE_FORMAT_A4R4G4B4_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM:
+		case PIPE_FORMAT_R16_SNORM:
+		case PIPE_FORMAT_L8_UNORM:
+		case PIPE_FORMAT_A8_UNORM:
+		case PIPE_FORMAT_I8_UNORM:
+		case PIPE_FORMAT_A8L8_UNORM:
+		case PIPE_FORMAT_Z16_UNORM:
+		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_DXT1_RGB:
+		case PIPE_FORMAT_DXT1_RGBA:
+		case PIPE_FORMAT_DXT3_RGBA:
+		case PIPE_FORMAT_DXT5_RGBA:
+			return TRUE;
+		default:
+			break;
+		}
+	}
+
+	return FALSE;
+}
+
+static struct pipe_buffer *
+nv40_surface_buffer(struct pipe_surface *surf)
+{
+	struct nv40_miptree *mt = (struct nv40_miptree *)surf->texture;
+
+	return mt->buffer;
+}
+
+static void *
+nv40_surface_map(struct pipe_screen *screen, struct pipe_surface *surface,
+		 unsigned flags )
+{
+	struct pipe_winsys	*ws = screen->winsys;
+	struct pipe_surface	*surface_to_map;
+	void			*map;
+
+	if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
+		struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture;
+
+		if (!mt->shadow_tex) {
+			unsigned old_tex_usage = surface->texture->tex_usage;
+			surface->texture->tex_usage = NOUVEAU_TEXTURE_USAGE_LINEAR |
+			                              PIPE_TEXTURE_USAGE_DYNAMIC;
+			mt->shadow_tex = screen->texture_create(screen, surface->texture);
+			surface->texture->tex_usage = old_tex_usage;
+
+			assert(mt->shadow_tex->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR);
+		}
+
+		mt->shadow_surface = screen->get_tex_surface
+		(
+			screen, mt->shadow_tex,
+			surface->face, surface->level, surface->zslice,
+			surface->usage
+		);
+
+		surface_to_map = mt->shadow_surface;
+	}
+	else
+		surface_to_map = surface;
+
+	assert(surface_to_map);
+	map = ws->buffer_map(ws, nv40_surface_buffer(surface_to_map), flags);
+	if (!map)
+		return NULL;
+
+	return map + surface_to_map->offset;
+}
+
+static void
+nv40_surface_unmap(struct pipe_screen *screen, struct pipe_surface *surface)
+{
+	struct pipe_winsys	*ws = screen->winsys;
+	struct pipe_surface	*surface_to_unmap;
+
+	/* TODO: Copy from shadow just before push buffer is flushed instead.
+	         There are probably some programs that map/unmap excessively
+	         before rendering. */
+	if (!(surface->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
+		struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture;
+
+		assert(mt->shadow_tex);
+
+		surface_to_unmap = mt->shadow_surface;
+	}
+	else
+		surface_to_unmap = surface;
+
+	assert(surface_to_unmap);
+
+	ws->buffer_unmap(ws, nv40_surface_buffer(surface_to_unmap));
+
+	if (surface_to_unmap != surface) {
+		struct nv40_screen *nvscreen = nv40_screen(screen);
+
+		nvscreen->eng2d->copy(nvscreen->eng2d, surface, 0, 0,
+		                      surface_to_unmap, 0, 0,
+		                      surface->width, surface->height);
+
+		screen->tex_surface_release(screen, &surface_to_unmap);
+	}
+}
+
+static void
+nv40_screen_destroy(struct pipe_screen *pscreen)
+{
+	struct nv40_screen *screen = nv40_screen(pscreen);
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	nvws->res_free(&screen->vp_exec_heap);
+	nvws->res_free(&screen->vp_data_heap);
+	nvws->res_free(&screen->query_heap);
+	nvws->notifier_free(&screen->query);
+	nvws->notifier_free(&screen->sync);
+	nvws->grobj_free(&screen->curie);
+
+	FREE(pscreen);
+}
+
+struct pipe_screen *
+nv40_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+{
+	struct nv40_screen *screen = CALLOC_STRUCT(nv40_screen);
+	struct nouveau_stateobj *so;
+	unsigned curie_class;
+	unsigned chipset = nvws->channel->device->chipset;
+	int ret;
+
+	if (!screen)
+		return NULL;
+	screen->nvws = nvws;
+
+	/* 2D engine setup */
+	screen->eng2d = nv04_surface_2d_init(nvws);
+	screen->eng2d->buf = nv40_surface_buffer;
+
+	/* 3D object */
+	switch (chipset & 0xf0) {
+	case 0x40:
+		if (NV4X_GRCLASS4097_CHIPSETS & (1 << (chipset & 0x0f)))
+			curie_class = NV40TCL;
+		else
+		if (NV4X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f)))
+			curie_class = NV44TCL;
+		break;
+	case 0x60:
+		if (NV6X_GRCLASS4497_CHIPSETS & (1 << (chipset & 0x0f)))
+			curie_class = NV44TCL;
+		break;
+	default:
+		break;
+	}
+
+	if (!curie_class) {
+		NOUVEAU_ERR("Unknown nv4x chipset: nv%02x\n", chipset);
+		return NULL;
+	}
+
+	ret = nvws->grobj_alloc(nvws, curie_class, &screen->curie);
+	if (ret) {
+		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
+		return FALSE;
+	}
+
+	/* Notifier for sync purposes */
+	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	if (ret) {
+		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
+		nv40_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	/* Query objects */
+	ret = nvws->notifier_alloc(nvws, 32, &screen->query);
+	if (ret) {
+		NOUVEAU_ERR("Error initialising query objects: %d\n", ret);
+		nv40_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	ret = nvws->res_init(&screen->query_heap, 0, 32);
+	if (ret) {
+		NOUVEAU_ERR("Error initialising query object heap: %d\n", ret);
+		nv40_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	/* Vtxprog resources */
+	if (nvws->res_init(&screen->vp_exec_heap, 0, 512) ||
+	    nvws->res_init(&screen->vp_data_heap, 0, 256)) {
+		nv40_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	/* Static curie initialisation */
+	so = so_new(128, 0);
+	so_method(so, screen->curie, NV40TCL_DMA_NOTIFY, 1);
+	so_data  (so, screen->sync->handle);
+	so_method(so, screen->curie, NV40TCL_DMA_TEXTURE0, 2);
+	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, nvws->channel->gart->handle);
+	so_method(so, screen->curie, NV40TCL_DMA_COLOR1, 1);
+	so_data  (so, nvws->channel->vram->handle);
+	so_method(so, screen->curie, NV40TCL_DMA_COLOR0, 2);
+	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, nvws->channel->vram->handle);
+	so_method(so, screen->curie, NV40TCL_DMA_VTXBUF0, 2);
+	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, nvws->channel->gart->handle);
+	so_method(so, screen->curie, NV40TCL_DMA_FENCE, 2);
+	so_data  (so, 0);
+	so_data  (so, screen->query->handle);
+	so_method(so, screen->curie, NV40TCL_DMA_UNK01AC, 2);
+	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, nvws->channel->vram->handle);
+	so_method(so, screen->curie, NV40TCL_DMA_COLOR2, 2);
+	so_data  (so, nvws->channel->vram->handle);
+	so_data  (so, nvws->channel->vram->handle);
+
+	so_method(so, screen->curie, 0x1ea4, 3);
+	so_data  (so, 0x00000010);
+	so_data  (so, 0x01000100);
+	so_data  (so, 0xff800006);
+
+	/* vtxprog output routing */
+	so_method(so, screen->curie, 0x1fc4, 1);
+	so_data  (so, 0x06144321);
+	so_method(so, screen->curie, 0x1fc8, 2);
+	so_data  (so, 0xedcba987);
+	so_data  (so, 0x00000021);
+	so_method(so, screen->curie, 0x1fd0, 1);
+	so_data  (so, 0x00171615);
+	so_method(so, screen->curie, 0x1fd4, 1);
+	so_data  (so, 0x001b1a19);
+
+	so_method(so, screen->curie, 0x1ef8, 1);
+	so_data  (so, 0x0020ffff);
+	so_method(so, screen->curie, 0x1d64, 1);
+	so_data  (so, 0x00d30000);
+	so_method(so, screen->curie, 0x1e94, 1);
+	so_data  (so, 0x00000001);
+
+	so_emit(nvws, so);
+	so_ref(NULL, &so);
+	nvws->push_flush(nvws, 0, NULL);
+
+	screen->pipe.winsys = ws;
+	screen->pipe.destroy = nv40_screen_destroy;
+
+	screen->pipe.get_name = nv40_screen_get_name;
+	screen->pipe.get_vendor = nv40_screen_get_vendor;
+	screen->pipe.get_param = nv40_screen_get_param;
+	screen->pipe.get_paramf = nv40_screen_get_paramf;
+
+	screen->pipe.is_format_supported = nv40_screen_surface_format_supported;
+
+	screen->pipe.surface_map = nv40_surface_map;
+	screen->pipe.surface_unmap = nv40_surface_unmap;
+
+	nv40_screen_init_miptree_functions(&screen->pipe);
+	u_simple_screen_init(&screen->pipe);
+
+	return &screen->pipe;
+}
+
diff --git a/src/gallium/drivers/nv40/nv40_screen.h b/src/gallium/drivers/nv40/nv40_screen.h
new file mode 100644
index 0000000000..4500aa0e5c
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_screen.h
@@ -0,0 +1,37 @@
+#ifndef __NV40_SCREEN_H__
+#define __NV40_SCREEN_H__
+
+#include "pipe/p_screen.h"
+#include "nv04/nv04_surface_2d.h"
+
+struct nv40_screen {
+	struct pipe_screen pipe;
+
+	struct nouveau_winsys *nvws;
+
+	unsigned cur_pctx;
+
+	/* HW graphics objects */
+	struct nv04_surface_2d *eng2d;
+	struct nouveau_grobj *curie;
+	struct nouveau_notifier *sync;
+
+	/* Query object resources */
+	struct nouveau_notifier *query;
+	struct nouveau_resource *query_heap;
+
+	/* Vtxprog resources */
+	struct nouveau_resource *vp_exec_heap;
+	struct nouveau_resource *vp_data_heap;
+
+	/* Current 3D state of channel */
+	struct nouveau_stateobj *state[NV40_STATE_MAX];
+};
+
+static INLINE struct nv40_screen *
+nv40_screen(struct pipe_screen *screen)
+{
+	return (struct nv40_screen *)screen;
+}
+
+#endif
diff --git a/src/gallium/drivers/nv40/nv40_shader.h b/src/gallium/drivers/nv40/nv40_shader.h
new file mode 100644
index 0000000000..854dccf548
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_shader.h
@@ -0,0 +1,556 @@
+#ifndef __NV40_SHADER_H__
+#define __NV40_SHADER_H__
+
+/* Vertex programs instruction set
+ *
+ * The NV40 instruction set is very similar to NV30.  Most fields are in
+ * a slightly different position in the instruction however.
+ *
+ * Merged instructions
+ *     In some cases it is possible to put two instructions into one opcode
+ *     slot.  The rules for when this is OK is not entirely clear to me yet.
+ *
+ *     There are separate writemasks and dest temp register fields for each
+ *     grouping of instructions.  There is however only one field with the
+ *     ID of a result register.  Writing to temp/result regs is selected by
+ *     setting VEC_RESULT/SCA_RESULT.
+ *
+ * Temporary registers
+ *     The source/dest temp register fields have been extended by 1 bit, to
+ *     give a total of 32 temporary registers.
+ *
+ * Relative Addressing
+ *     NV40 can use an address register to index into vertex attribute regs.
+ *     This is done by putting the offset value into INPUT_SRC and setting
+ *     the INDEX_INPUT flag.
+ *
+ * Conditional execution (see NV_vertex_program{2,3} for details)
+ *     There is a second condition code register on NV40, it's use is enabled
+ *     by setting the COND_REG_SELECT_1 flag.
+ *
+ * Texture lookup
+ *     TODO
+ */
+
+/* ---- OPCODE BITS 127:96 / data DWORD 0 --- */
+#define NV40_VP_INST_VEC_RESULT                                        (1 << 30)
+/* uncertain.. */
+#define NV40_VP_INST_COND_UPDATE_ENABLE                        ((1 << 14)|1<<29)
+/* use address reg as index into attribs */
+#define NV40_VP_INST_INDEX_INPUT                                       (1 << 27)
+#define NV40_VP_INST_COND_REG_SELECT_1                                 (1 << 25)
+#define NV40_VP_INST_ADDR_REG_SELECT_1                                 (1 << 24)
+#define NV40_VP_INST_SRC2_ABS                                          (1 << 23)
+#define NV40_VP_INST_SRC1_ABS                                          (1 << 22)
+#define NV40_VP_INST_SRC0_ABS                                          (1 << 21)
+#define NV40_VP_INST_VEC_DEST_TEMP_SHIFT                                      15
+#define NV40_VP_INST_VEC_DEST_TEMP_MASK                             (0x1F << 15)
+#define NV40_VP_INST_COND_TEST_ENABLE                                  (1 << 13)
+#define NV40_VP_INST_COND_SHIFT                                               10
+#define NV40_VP_INST_COND_MASK                                       (0x7 << 10)
+#    define NV40_VP_INST_COND_FL                                               0
+#    define NV40_VP_INST_COND_LT                                               1
+#    define NV40_VP_INST_COND_EQ                                               2
+#    define NV40_VP_INST_COND_LE                                               3
+#    define NV40_VP_INST_COND_GT                                               4
+#    define NV40_VP_INST_COND_NE                                               5
+#    define NV40_VP_INST_COND_GE                                               6
+#    define NV40_VP_INST_COND_TR                                               7
+#define NV40_VP_INST_COND_SWZ_X_SHIFT                                          8
+#define NV40_VP_INST_COND_SWZ_X_MASK                                    (3 << 8)
+#define NV40_VP_INST_COND_SWZ_Y_SHIFT                                          6
+#define NV40_VP_INST_COND_SWZ_Y_MASK                                    (3 << 6)
+#define NV40_VP_INST_COND_SWZ_Z_SHIFT                                          4
+#define NV40_VP_INST_COND_SWZ_Z_MASK                                    (3 << 4)
+#define NV40_VP_INST_COND_SWZ_W_SHIFT                                          2
+#define NV40_VP_INST_COND_SWZ_W_MASK                                    (3 << 2)
+#define NV40_VP_INST_COND_SWZ_ALL_SHIFT                                        2
+#define NV40_VP_INST_COND_SWZ_ALL_MASK                               (0xFF << 2)
+#define NV40_VP_INST_ADDR_SWZ_SHIFT                                            0
+#define NV40_VP_INST_ADDR_SWZ_MASK                                   (0x03 << 0)
+#define NV40_VP_INST0_KNOWN ( \
+                NV40_VP_INST_INDEX_INPUT | \
+                NV40_VP_INST_COND_REG_SELECT_1 | \
+                NV40_VP_INST_ADDR_REG_SELECT_1 | \
+                NV40_VP_INST_SRC2_ABS | \
+                NV40_VP_INST_SRC1_ABS | \
+                NV40_VP_INST_SRC0_ABS | \
+                NV40_VP_INST_VEC_DEST_TEMP_MASK | \
+                NV40_VP_INST_COND_TEST_ENABLE | \
+                NV40_VP_INST_COND_MASK | \
+                NV40_VP_INST_COND_SWZ_ALL_MASK | \
+                NV40_VP_INST_ADDR_SWZ_MASK)
+
+/* ---- OPCODE BITS 95:64 / data DWORD 1 --- */
+#define NV40_VP_INST_VEC_OPCODE_SHIFT                                         22
+#define NV40_VP_INST_VEC_OPCODE_MASK                                (0x1F << 22)
+#    define NV40_VP_INST_OP_NOP                                             0x00
+#    define NV40_VP_INST_OP_MOV                                             0x01
+#    define NV40_VP_INST_OP_MUL                                             0x02
+#    define NV40_VP_INST_OP_ADD                                             0x03
+#    define NV40_VP_INST_OP_MAD                                             0x04
+#    define NV40_VP_INST_OP_DP3                                             0x05
+#    define NV40_VP_INST_OP_DPH                                             0x06
+#    define NV40_VP_INST_OP_DP4                                             0x07
+#    define NV40_VP_INST_OP_DST                                             0x08
+#    define NV40_VP_INST_OP_MIN                                             0x09
+#    define NV40_VP_INST_OP_MAX                                             0x0A
+#    define NV40_VP_INST_OP_SLT                                             0x0B
+#    define NV40_VP_INST_OP_SGE                                             0x0C
+#    define NV40_VP_INST_OP_ARL                                             0x0D
+#    define NV40_VP_INST_OP_FRC                                             0x0E
+#    define NV40_VP_INST_OP_FLR                                             0x0F
+#    define NV40_VP_INST_OP_SEQ                                             0x10
+#    define NV40_VP_INST_OP_SFL                                             0x11
+#    define NV40_VP_INST_OP_SGT                                             0x12
+#    define NV40_VP_INST_OP_SLE                                             0x13
+#    define NV40_VP_INST_OP_SNE                                             0x14
+#    define NV40_VP_INST_OP_STR                                             0x15
+#    define NV40_VP_INST_OP_SSG                                             0x16
+#    define NV40_VP_INST_OP_ARR                                             0x17
+#    define NV40_VP_INST_OP_ARA                                             0x18
+#    define NV40_VP_INST_OP_TXL                                             0x19
+#define NV40_VP_INST_SCA_OPCODE_SHIFT                                         27
+#define NV40_VP_INST_SCA_OPCODE_MASK                                (0x1F << 27)
+#    define NV40_VP_INST_OP_NOP                                             0x00
+#    define NV40_VP_INST_OP_MOV                                             0x01
+#    define NV40_VP_INST_OP_RCP                                             0x02
+#    define NV40_VP_INST_OP_RCC                                             0x03
+#    define NV40_VP_INST_OP_RSQ                                             0x04
+#    define NV40_VP_INST_OP_EXP                                             0x05
+#    define NV40_VP_INST_OP_LOG                                             0x06
+#    define NV40_VP_INST_OP_LIT                                             0x07
+#    define NV40_VP_INST_OP_BRA                                             0x09
+#    define NV40_VP_INST_OP_CAL                                             0x0B
+#    define NV40_VP_INST_OP_RET                                             0x0C
+#    define NV40_VP_INST_OP_LG2                                             0x0D
+#    define NV40_VP_INST_OP_EX2                                             0x0E
+#    define NV40_VP_INST_OP_SIN                                             0x0F
+#    define NV40_VP_INST_OP_COS                                             0x10
+#    define NV40_VP_INST_OP_PUSHA                                           0x13
+#    define NV40_VP_INST_OP_POPA                                            0x14
+#define NV40_VP_INST_CONST_SRC_SHIFT                                          12
+#define NV40_VP_INST_CONST_SRC_MASK                                 (0xFF << 12)
+#define NV40_VP_INST_INPUT_SRC_SHIFT                                           8
+#define NV40_VP_INST_INPUT_SRC_MASK                                  (0x0F << 8)
+#    define NV40_VP_INST_IN_POS                                                0
+#    define NV40_VP_INST_IN_WEIGHT                                             1
+#    define NV40_VP_INST_IN_NORMAL                                             2
+#    define NV40_VP_INST_IN_COL0                                               3
+#    define NV40_VP_INST_IN_COL1                                               4
+#    define NV40_VP_INST_IN_FOGC                                               5
+#    define NV40_VP_INST_IN_TC0                                                8
+#    define NV40_VP_INST_IN_TC(n)                                          (8+n)
+#define NV40_VP_INST_SRC0H_SHIFT                                               0
+#define NV40_VP_INST_SRC0H_MASK                                      (0xFF << 0)
+#define NV40_VP_INST1_KNOWN ( \
+                NV40_VP_INST_VEC_OPCODE_MASK | \
+                NV40_VP_INST_SCA_OPCODE_MASK | \
+                NV40_VP_INST_CONST_SRC_MASK  | \
+                NV40_VP_INST_INPUT_SRC_MASK  | \
+                NV40_VP_INST_SRC0H_MASK \
+                )
+
+/* ---- OPCODE BITS 63:32 / data DWORD 2 --- */
+#define NV40_VP_INST_SRC0L_SHIFT                                              23
+#define NV40_VP_INST_SRC0L_MASK                                    (0x1FF << 23)
+#define NV40_VP_INST_SRC1_SHIFT                                                6
+#define NV40_VP_INST_SRC1_MASK                                    (0x1FFFF << 6)
+#define NV40_VP_INST_SRC2H_SHIFT                                               0
+#define NV40_VP_INST_SRC2H_MASK                                      (0x3F << 0)
+#define NV40_VP_INST_IADDRH_SHIFT                                              0
+#define NV40_VP_INST_IADDRH_MASK                                     (0x1F << 0)
+
+/* ---- OPCODE BITS 31:0 / data DWORD 3 --- */
+#define NV40_VP_INST_IADDRL_SHIFT                                             29
+#define NV40_VP_INST_IADDRL_MASK                                       (7 << 29)
+#define NV40_VP_INST_SRC2L_SHIFT                                              21
+#define NV40_VP_INST_SRC2L_MASK                                    (0x7FF << 21)
+#define NV40_VP_INST_SCA_WRITEMASK_SHIFT                                      17
+#define NV40_VP_INST_SCA_WRITEMASK_MASK                              (0xF << 17)
+#    define NV40_VP_INST_SCA_WRITEMASK_X                               (1 << 20)
+#    define NV40_VP_INST_SCA_WRITEMASK_Y                               (1 << 19)
+#    define NV40_VP_INST_SCA_WRITEMASK_Z                               (1 << 18)
+#    define NV40_VP_INST_SCA_WRITEMASK_W                               (1 << 17)
+#define NV40_VP_INST_VEC_WRITEMASK_SHIFT                                      13
+#define NV40_VP_INST_VEC_WRITEMASK_MASK                              (0xF << 13)
+#    define NV40_VP_INST_VEC_WRITEMASK_X                               (1 << 16)
+#    define NV40_VP_INST_VEC_WRITEMASK_Y                               (1 << 15)
+#    define NV40_VP_INST_VEC_WRITEMASK_Z                               (1 << 14)
+#    define NV40_VP_INST_VEC_WRITEMASK_W                               (1 << 13)
+#define NV40_VP_INST_SCA_RESULT                                        (1 << 12)
+#define NV40_VP_INST_SCA_DEST_TEMP_SHIFT                                       7
+#define NV40_VP_INST_SCA_DEST_TEMP_MASK                              (0x1F << 7)
+#define NV40_VP_INST_DEST_SHIFT                                                2
+#define NV40_VP_INST_DEST_MASK                                         (31 << 2)
+#    define NV40_VP_INST_DEST_POS                                              0
+#    define NV40_VP_INST_DEST_COL0                                             1
+#    define NV40_VP_INST_DEST_COL1                                             2
+#    define NV40_VP_INST_DEST_BFC0                                             3
+#    define NV40_VP_INST_DEST_BFC1                                             4
+#    define NV40_VP_INST_DEST_FOGC                                             5
+#    define NV40_VP_INST_DEST_PSZ                                              6
+#    define NV40_VP_INST_DEST_TC0                                              7
+#    define NV40_VP_INST_DEST_TC(n)                                        (7+n)
+#    define NV40_VP_INST_DEST_TEMP                                          0x1F
+#define NV40_VP_INST_INDEX_CONST                                        (1 << 1)
+#define NV40_VP_INST_LAST                                               (1 << 0)
+#define NV40_VP_INST3_KNOWN ( \
+                NV40_VP_INST_SRC2L_MASK |\
+                NV40_VP_INST_SCA_WRITEMASK_MASK |\
+                NV40_VP_INST_VEC_WRITEMASK_MASK |\
+                NV40_VP_INST_SCA_DEST_TEMP_MASK |\
+                NV40_VP_INST_DEST_MASK |\
+                NV40_VP_INST_INDEX_CONST)
+
+/* Useful to split the source selection regs into their pieces */
+#define NV40_VP_SRC0_HIGH_SHIFT                                                9
+#define NV40_VP_SRC0_HIGH_MASK                                        0x0001FE00
+#define NV40_VP_SRC0_LOW_MASK                                         0x000001FF
+#define NV40_VP_SRC2_HIGH_SHIFT                                               11
+#define NV40_VP_SRC2_HIGH_MASK                                        0x0001F800
+#define NV40_VP_SRC2_LOW_MASK                                         0x000007FF
+
+/* Source selection - these are the bits you fill NV40_VP_INST_SRCn with */
+#define NV40_VP_SRC_NEGATE                                             (1 << 16)
+#define NV40_VP_SRC_SWZ_X_SHIFT                                               14
+#define NV40_VP_SRC_SWZ_X_MASK                                         (3 << 14)
+#define NV40_VP_SRC_SWZ_Y_SHIFT                                               12
+#define NV40_VP_SRC_SWZ_Y_MASK                                         (3 << 12)
+#define NV40_VP_SRC_SWZ_Z_SHIFT                                               10
+#define NV40_VP_SRC_SWZ_Z_MASK                                         (3 << 10)
+#define NV40_VP_SRC_SWZ_W_SHIFT                                                8
+#define NV40_VP_SRC_SWZ_W_MASK                                          (3 << 8)
+#define NV40_VP_SRC_SWZ_ALL_SHIFT                                              8
+#define NV40_VP_SRC_SWZ_ALL_MASK                                     (0xFF << 8)
+#define NV40_VP_SRC_TEMP_SRC_SHIFT                                             2
+#define NV40_VP_SRC_TEMP_SRC_MASK                                    (0x1F << 2)
+#define NV40_VP_SRC_REG_TYPE_SHIFT                                             0
+#define NV40_VP_SRC_REG_TYPE_MASK                                       (3 << 0)
+#    define NV40_VP_SRC_REG_TYPE_UNK0                                          0
+#    define NV40_VP_SRC_REG_TYPE_TEMP                                          1
+#    define NV40_VP_SRC_REG_TYPE_INPUT                                         2
+#    define NV40_VP_SRC_REG_TYPE_CONST                                         3
+
+
+/*
+ * Each fragment program opcode appears to be comprised of 4 32-bit values.
+ *
+ *         0 - Opcode, output reg/mask, ATTRIB source
+ *         1 - Source 0
+ *         2 - Source 1
+ *         3 - Source 2
+ *
+ * There appears to be no special difference between result regs and temp regs.
+ *                 result.color == R0.xyzw
+ *                 result.depth == R1.z
+ * When the fragprog contains instructions to write depth,
+ * NV30_TCL_PRIMITIVE_3D_UNK1D78=0 otherwise it is set to 1.
+ *
+ * Constants are inserted directly after the instruction that uses them.
+ * 
+ * It appears that it's not possible to use two input registers in one
+ * instruction as the input sourcing is done in the instruction dword
+ * and not the source selection dwords.  As such instructions such as:
+ * 
+ *                 ADD result.color, fragment.color, fragment.texcoord[0];
+ *
+ * must be split into two MOV's and then an ADD (nvidia does this) but
+ * I'm not sure why it's not just one MOV and then source the second input
+ * in the ADD instruction..
+ *
+ * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary
+ * negation requires multiplication with a const.
+ *
+ * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO and
+ * SWIZZLE_ONE.
+ *
+ * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as
+ * SWIZZLE_ZERO is implemented simply by not writing to the relevant components
+ * of the destination.
+ *
+ * Looping
+ *   Loops appear to be fairly expensive on NV40 at least, the proprietary
+ *   driver goes to a lot of effort to avoid using the native looping
+ *   instructions.  If the total number of *executed* instructions between
+ *   REP/ENDREP or LOOP/ENDLOOP is <=500, the driver will unroll the loop.
+ *   The maximum loop count is 255.
+ *
+ * Conditional execution
+ *   TODO
+ * 
+ * Non-native instructions:
+ *         LIT
+ *         LRP - MAD+MAD
+ *         SUB - ADD, negate second source
+ *         RSQ - LG2 + EX2
+ *         POW - LG2 + MUL + EX2
+ *         SCS - COS + SIN
+ *         XPD
+ *         DP2 - MUL + ADD
+ *         NRM
+ */
+
+//== Opcode / Destination selection ==
+#define NV40_FP_OP_PROGRAM_END                                          (1 << 0)
+#define NV40_FP_OP_OUT_REG_SHIFT                                               1
+#define NV40_FP_OP_OUT_REG_MASK                                        (63 << 1)
+/* Needs to be set when writing outputs to get expected result.. */
+#define NV40_FP_OP_OUT_REG_HALF                                         (1 << 7)
+#define NV40_FP_OP_COND_WRITE_ENABLE                                    (1 << 8)
+#define NV40_FP_OP_OUTMASK_SHIFT                                               9
+#define NV40_FP_OP_OUTMASK_MASK                                       (0xF << 9)
+#    define NV40_FP_OP_OUT_X                                            (1 << 9)
+#    define NV40_FP_OP_OUT_Y                                            (1 <<10)
+#    define NV40_FP_OP_OUT_Z                                            (1 <<11)
+#    define NV40_FP_OP_OUT_W                                            (1 <<12)
+/* Uncertain about these, especially the input_src values.. it's possible that
+ * they can be dynamically changed.
+ */
+#define NV40_FP_OP_INPUT_SRC_SHIFT                                            13
+#define NV40_FP_OP_INPUT_SRC_MASK                                     (15 << 13)
+#    define NV40_FP_OP_INPUT_SRC_POSITION                                    0x0
+#    define NV40_FP_OP_INPUT_SRC_COL0                                        0x1
+#    define NV40_FP_OP_INPUT_SRC_COL1                                        0x2
+#    define NV40_FP_OP_INPUT_SRC_FOGC                                        0x3
+#    define NV40_FP_OP_INPUT_SRC_TC0                                         0x4
+#    define NV40_FP_OP_INPUT_SRC_TC(n)                                 (0x4 + n)
+#    define NV40_FP_OP_INPUT_SRC_FACING                                      0xE
+#define NV40_FP_OP_TEX_UNIT_SHIFT                                             17
+#define NV40_FP_OP_TEX_UNIT_MASK                                     (0xF << 17)
+#define NV40_FP_OP_PRECISION_SHIFT                                            22
+#define NV40_FP_OP_PRECISION_MASK                                      (3 << 22)
+#   define NV40_FP_PRECISION_FP32                                              0
+#   define NV40_FP_PRECISION_FP16                                              1
+#   define NV40_FP_PRECISION_FX12                                              2
+#define NV40_FP_OP_OPCODE_SHIFT                                               24
+#define NV40_FP_OP_OPCODE_MASK                                      (0x3F << 24)
+#        define NV40_FP_OP_OPCODE_NOP                                       0x00
+#        define NV40_FP_OP_OPCODE_MOV                                       0x01
+#        define NV40_FP_OP_OPCODE_MUL                                       0x02
+#        define NV40_FP_OP_OPCODE_ADD                                       0x03
+#        define NV40_FP_OP_OPCODE_MAD                                       0x04
+#        define NV40_FP_OP_OPCODE_DP3                                       0x05
+#        define NV40_FP_OP_OPCODE_DP4                                       0x06
+#        define NV40_FP_OP_OPCODE_DST                                       0x07
+#        define NV40_FP_OP_OPCODE_MIN                                       0x08
+#        define NV40_FP_OP_OPCODE_MAX                                       0x09
+#        define NV40_FP_OP_OPCODE_SLT                                       0x0A
+#        define NV40_FP_OP_OPCODE_SGE                                       0x0B
+#        define NV40_FP_OP_OPCODE_SLE                                       0x0C
+#        define NV40_FP_OP_OPCODE_SGT                                       0x0D
+#        define NV40_FP_OP_OPCODE_SNE                                       0x0E
+#        define NV40_FP_OP_OPCODE_SEQ                                       0x0F
+#        define NV40_FP_OP_OPCODE_FRC                                       0x10
+#        define NV40_FP_OP_OPCODE_FLR                                       0x11
+#        define NV40_FP_OP_OPCODE_KIL                                       0x12
+#        define NV40_FP_OP_OPCODE_PK4B                                      0x13
+#        define NV40_FP_OP_OPCODE_UP4B                                      0x14
+/* DDX/DDY can only write to XY */
+#        define NV40_FP_OP_OPCODE_DDX                                       0x15
+#        define NV40_FP_OP_OPCODE_DDY                                       0x16
+#        define NV40_FP_OP_OPCODE_TEX                                       0x17
+#        define NV40_FP_OP_OPCODE_TXP                                       0x18
+#        define NV40_FP_OP_OPCODE_TXD                                       0x19
+#        define NV40_FP_OP_OPCODE_RCP                                       0x1A
+#        define NV40_FP_OP_OPCODE_EX2                                       0x1C
+#        define NV40_FP_OP_OPCODE_LG2                                       0x1D
+#        define NV40_FP_OP_OPCODE_STR                                       0x20
+#        define NV40_FP_OP_OPCODE_SFL                                       0x21
+#        define NV40_FP_OP_OPCODE_COS                                       0x22
+#        define NV40_FP_OP_OPCODE_SIN                                       0x23
+#        define NV40_FP_OP_OPCODE_PK2H                                      0x24
+#        define NV40_FP_OP_OPCODE_UP2H                                      0x25
+#        define NV40_FP_OP_OPCODE_PK4UB                                     0x27
+#        define NV40_FP_OP_OPCODE_UP4UB                                     0x28
+#        define NV40_FP_OP_OPCODE_PK2US                                     0x29
+#        define NV40_FP_OP_OPCODE_UP2US                                     0x2A
+#        define NV40_FP_OP_OPCODE_DP2A                                      0x2E
+#        define NV40_FP_OP_OPCODE_TXL                                       0x2F
+#        define NV40_FP_OP_OPCODE_TXB                                       0x31
+#        define NV40_FP_OP_OPCODE_DIV                                       0x3A
+#        define NV40_FP_OP_OPCODE_UNK_LIT                                   0x3C
+/* The use of these instructions appears to be indicated by bit 31 of DWORD 2.*/
+#        define NV40_FP_OP_BRA_OPCODE_BRK                                    0x0
+#        define NV40_FP_OP_BRA_OPCODE_CAL                                    0x1
+#        define NV40_FP_OP_BRA_OPCODE_IF                                     0x2
+#        define NV40_FP_OP_BRA_OPCODE_LOOP                                   0x3
+#        define NV40_FP_OP_BRA_OPCODE_REP                                    0x4
+#        define NV40_FP_OP_BRA_OPCODE_RET                                    0x5
+#define NV40_FP_OP_OUT_SAT                                             (1 << 31)
+
+/* high order bits of SRC0 */
+#define NV40_FP_OP_OUT_ABS                                             (1 << 29)
+#define NV40_FP_OP_COND_SWZ_W_SHIFT                                           27
+#define NV40_FP_OP_COND_SWZ_W_MASK                                     (3 << 27)
+#define NV40_FP_OP_COND_SWZ_Z_SHIFT                                           25
+#define NV40_FP_OP_COND_SWZ_Z_MASK                                     (3 << 25)
+#define NV40_FP_OP_COND_SWZ_Y_SHIFT                                           23
+#define NV40_FP_OP_COND_SWZ_Y_MASK                                     (3 << 23)
+#define NV40_FP_OP_COND_SWZ_X_SHIFT                                           21
+#define NV40_FP_OP_COND_SWZ_X_MASK                                     (3 << 21)
+#define NV40_FP_OP_COND_SWZ_ALL_SHIFT                                         21
+#define NV40_FP_OP_COND_SWZ_ALL_MASK                                (0xFF << 21)
+#define NV40_FP_OP_COND_SHIFT                                                 18
+#define NV40_FP_OP_COND_MASK                                        (0x07 << 18)
+#        define NV40_FP_OP_COND_FL                                             0
+#        define NV40_FP_OP_COND_LT                                             1
+#        define NV40_FP_OP_COND_EQ                                             2
+#        define NV40_FP_OP_COND_LE                                             3
+#        define NV40_FP_OP_COND_GT                                             4
+#        define NV40_FP_OP_COND_NE                                             5
+#        define NV40_FP_OP_COND_GE                                             6
+#        define NV40_FP_OP_COND_TR                                             7
+
+/* high order bits of SRC1 */
+#define NV40_FP_OP_OPCODE_IS_BRANCH                                      (1<<31)
+#define NV40_FP_OP_DST_SCALE_SHIFT                                            28
+#define NV40_FP_OP_DST_SCALE_MASK                                      (3 << 28)
+#define NV40_FP_OP_DST_SCALE_1X                                                0
+#define NV40_FP_OP_DST_SCALE_2X                                                1
+#define NV40_FP_OP_DST_SCALE_4X                                                2
+#define NV40_FP_OP_DST_SCALE_8X                                                3
+#define NV40_FP_OP_DST_SCALE_INV_2X                                            5
+#define NV40_FP_OP_DST_SCALE_INV_4X                                            6
+#define NV40_FP_OP_DST_SCALE_INV_8X                                            7
+
+/* SRC1 LOOP */
+#define NV40_FP_OP_LOOP_INCR_SHIFT                                            19
+#define NV40_FP_OP_LOOP_INCR_MASK                                   (0xFF << 19)
+#define NV40_FP_OP_LOOP_INDEX_SHIFT                                           10
+#define NV40_FP_OP_LOOP_INDEX_MASK                                  (0xFF << 10)
+#define NV40_FP_OP_LOOP_COUNT_SHIFT                                            2
+#define NV40_FP_OP_LOOP_COUNT_MASK                                   (0xFF << 2)
+
+/* SRC1 IF */
+#define NV40_FP_OP_ELSE_ID_SHIFT                                               2
+#define NV40_FP_OP_ELSE_ID_MASK                                      (0xFF << 2)
+
+/* SRC1 CAL */
+#define NV40_FP_OP_IADDR_SHIFT                                                 2
+#define NV40_FP_OP_IADDR_MASK                                        (0xFF << 2)
+
+/* SRC1 REP
+ *   I have no idea why there are 3 count values here..  but they
+ *   have always been filled with the same value in my tests so
+ *   far..
+ */
+#define NV40_FP_OP_REP_COUNT1_SHIFT                                            2
+#define NV40_FP_OP_REP_COUNT1_MASK                                   (0xFF << 2)
+#define NV40_FP_OP_REP_COUNT2_SHIFT                                           10
+#define NV40_FP_OP_REP_COUNT2_MASK                                  (0xFF << 10)
+#define NV40_FP_OP_REP_COUNT3_SHIFT                                           19
+#define NV40_FP_OP_REP_COUNT3_MASK                                  (0xFF << 19)
+
+/* SRC2 REP/IF */
+#define NV40_FP_OP_END_ID_SHIFT                                                2
+#define NV40_FP_OP_END_ID_MASK                                       (0xFF << 2)
+
+// SRC2 high-order
+#define NV40_FP_OP_INDEX_INPUT                                         (1 << 30)
+#define NV40_FP_OP_ADDR_INDEX_SHIFT                                           19
+#define NV40_FP_OP_ADDR_INDEX_MASK                                   (0xF << 19)
+
+//== Register selection ==
+#define NV40_FP_REG_TYPE_SHIFT                                                 0
+#define NV40_FP_REG_TYPE_MASK                                           (3 << 0)
+#        define NV40_FP_REG_TYPE_TEMP                                          0
+#        define NV40_FP_REG_TYPE_INPUT                                         1
+#        define NV40_FP_REG_TYPE_CONST                                         2
+#define NV40_FP_REG_SRC_SHIFT                                                  2
+#define NV40_FP_REG_SRC_MASK                                           (63 << 2)
+#define NV40_FP_REG_SRC_HALF                                            (1 << 8)
+#define NV40_FP_REG_SWZ_ALL_SHIFT                                              9
+#define NV40_FP_REG_SWZ_ALL_MASK                                      (255 << 9)
+#define NV40_FP_REG_SWZ_X_SHIFT                                                9
+#define NV40_FP_REG_SWZ_X_MASK                                          (3 << 9)
+#define NV40_FP_REG_SWZ_Y_SHIFT                                               11
+#define NV40_FP_REG_SWZ_Y_MASK                                         (3 << 11)
+#define NV40_FP_REG_SWZ_Z_SHIFT                                               13
+#define NV40_FP_REG_SWZ_Z_MASK                                         (3 << 13)
+#define NV40_FP_REG_SWZ_W_SHIFT                                               15
+#define NV40_FP_REG_SWZ_W_MASK                                         (3 << 15)
+#        define NV40_FP_SWIZZLE_X                                              0
+#        define NV40_FP_SWIZZLE_Y                                              1
+#        define NV40_FP_SWIZZLE_Z                                              2
+#        define NV40_FP_SWIZZLE_W                                              3
+#define NV40_FP_REG_NEGATE                                             (1 << 17)
+
+#ifndef NV40_SHADER_NO_FUCKEDNESS
+#define NV40SR_NONE	0
+#define NV40SR_OUTPUT	1
+#define NV40SR_INPUT	2
+#define NV40SR_TEMP	3
+#define NV40SR_CONST	4
+
+struct nv40_sreg {
+	int type;
+	int index;
+
+	int dst_scale;
+
+	int negate;
+	int abs;
+	int swz[4];
+
+	int cc_update;
+	int cc_update_reg;
+	int cc_test;
+	int cc_test_reg;
+	int cc_swz[4];
+};
+
+static INLINE struct nv40_sreg
+nv40_sr(int type, int index)
+{
+	struct nv40_sreg temp = {
+		.type = type,
+		.index = index,
+		.dst_scale = DEF_SCALE,
+		.abs = 0,
+		.negate = 0,
+		.swz = { 0, 1, 2, 3 },
+		.cc_update = 0,
+		.cc_update_reg = 0,
+		.cc_test = DEF_CTEST,
+		.cc_test_reg = 0,
+		.cc_swz = { 0, 1, 2, 3 },
+	};
+	return temp;
+}
+
+static INLINE struct nv40_sreg
+nv40_sr_swz(struct nv40_sreg src, int x, int y, int z, int w)
+{
+	struct nv40_sreg dst = src;
+
+	dst.swz[SWZ_X] = src.swz[x];
+	dst.swz[SWZ_Y] = src.swz[y];
+	dst.swz[SWZ_Z] = src.swz[z];
+	dst.swz[SWZ_W] = src.swz[w];
+	return dst;
+}
+
+static INLINE struct nv40_sreg
+nv40_sr_neg(struct nv40_sreg src)
+{
+	src.negate = !src.negate;
+	return src;
+}
+
+static INLINE struct nv40_sreg
+nv40_sr_abs(struct nv40_sreg src)
+{
+	src.abs = 1;
+	return src;
+}
+
+static INLINE struct nv40_sreg
+nv40_sr_scale(struct nv40_sreg src, int scale)
+{
+	src.dst_scale = scale;
+	return src;
+}
+#endif
+
+#endif
diff --git a/src/gallium/drivers/nv40/nv40_state.c b/src/gallium/drivers/nv40/nv40_state.c
new file mode 100644
index 0000000000..2eff25aa83
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_state.c
@@ -0,0 +1,740 @@
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "draw/draw_context.h"
+
+#include "tgsi/tgsi_parse.h"
+
+#include "nv40_context.h"
+#include "nv40_state.h"
+
+static void *
+nv40_blend_state_create(struct pipe_context *pipe,
+			const struct pipe_blend_state *cso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nouveau_grobj *curie = nv40->screen->curie;
+	struct nv40_blend_state *bso = CALLOC(1, sizeof(*bso));
+	struct nouveau_stateobj *so = so_new(16, 0);
+
+	if (cso->blend_enable) {
+		so_method(so, curie, NV40TCL_BLEND_ENABLE, 3);
+		so_data  (so, 1);
+		so_data  (so, (nvgl_blend_func(cso->alpha_src_factor) << 16) |
+			       nvgl_blend_func(cso->rgb_src_factor));
+		so_data  (so, nvgl_blend_func(cso->alpha_dst_factor) << 16 |
+			      nvgl_blend_func(cso->rgb_dst_factor));
+		so_method(so, curie, NV40TCL_BLEND_EQUATION, 1);
+		so_data  (so, nvgl_blend_eqn(cso->alpha_func) << 16 |
+			      nvgl_blend_eqn(cso->rgb_func));
+	} else {
+		so_method(so, curie, NV40TCL_BLEND_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_method(so, curie, NV40TCL_COLOR_MASK, 1);
+	so_data  (so, (((cso->colormask & PIPE_MASK_A) ? (0x01 << 24) : 0) |
+		       ((cso->colormask & PIPE_MASK_R) ? (0x01 << 16) : 0) |
+		       ((cso->colormask & PIPE_MASK_G) ? (0x01 <<  8) : 0) |
+		       ((cso->colormask & PIPE_MASK_B) ? (0x01 <<  0) : 0)));
+
+	if (cso->logicop_enable) {
+		so_method(so, curie, NV40TCL_COLOR_LOGIC_OP_ENABLE, 2);
+		so_data  (so, 1);
+		so_data  (so, nvgl_logicop_func(cso->logicop_func));
+	} else {
+		so_method(so, curie, NV40TCL_COLOR_LOGIC_OP_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_method(so, curie, NV40TCL_DITHER_ENABLE, 1);
+	so_data  (so, cso->dither ? 1 : 0);
+
+	so_ref(so, &bso->so);
+	bso->pipe = *cso;
+	return (void *)bso;
+}
+
+static void
+nv40_blend_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->blend = hwcso;
+	nv40->dirty |= NV40_NEW_BLEND;
+}
+
+static void
+nv40_blend_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv40_blend_state *bso = hwcso;
+
+	so_ref(NULL, &bso->so);
+	FREE(bso);
+}
+
+
+static INLINE unsigned
+wrap_mode(unsigned wrap) {
+	unsigned ret;
+
+	switch (wrap) {
+	case PIPE_TEX_WRAP_REPEAT:
+		ret = NV40TCL_TEX_WRAP_S_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_REPEAT:
+		ret = NV40TCL_TEX_WRAP_S_MIRRORED_REPEAT;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+		ret = NV40TCL_TEX_WRAP_S_CLAMP_TO_EDGE;
+		break;
+	case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+		ret = NV40TCL_TEX_WRAP_S_CLAMP_TO_BORDER;
+		break;
+	case PIPE_TEX_WRAP_CLAMP:
+		ret = NV40TCL_TEX_WRAP_S_CLAMP;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+		ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP_TO_EDGE;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+		ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP_TO_BORDER;
+		break;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP:
+		ret = NV40TCL_TEX_WRAP_S_MIRROR_CLAMP;
+		break;
+	default:
+		NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
+		ret = NV40TCL_TEX_WRAP_S_REPEAT;
+		break;
+	}
+
+	return ret >> NV40TCL_TEX_WRAP_S_SHIFT;
+}
+
+static void *
+nv40_sampler_state_create(struct pipe_context *pipe,
+			  const struct pipe_sampler_state *cso)
+{
+	struct nv40_sampler_state *ps;
+	uint32_t filter = 0;
+
+	ps = MALLOC(sizeof(struct nv40_sampler_state));
+
+	ps->fmt = 0;
+	if (!cso->normalized_coords)
+		ps->fmt |= NV40TCL_TEX_FORMAT_RECT;
+
+	ps->wrap = ((wrap_mode(cso->wrap_s) << NV40TCL_TEX_WRAP_S_SHIFT) |
+		    (wrap_mode(cso->wrap_t) << NV40TCL_TEX_WRAP_T_SHIFT) |
+		    (wrap_mode(cso->wrap_r) << NV40TCL_TEX_WRAP_R_SHIFT));
+
+	ps->en = 0;
+	if (cso->max_anisotropy >= 2.0) {
+		/* no idea, binary driver sets it, works without it.. meh.. */
+		ps->wrap |= (1 << 5);
+
+		if (cso->max_anisotropy >= 16.0) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_16X;
+		} else
+		if (cso->max_anisotropy >= 12.0) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_12X;
+		} else
+		if (cso->max_anisotropy >= 10.0) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_10X;
+		} else
+		if (cso->max_anisotropy >= 8.0) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_8X;
+		} else
+		if (cso->max_anisotropy >= 6.0) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_6X;
+		} else
+		if (cso->max_anisotropy >= 4.0) {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_4X;
+		} else {
+			ps->en |= NV40TCL_TEX_ENABLE_ANISO_2X;
+		}
+	}
+
+	switch (cso->mag_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		filter |= NV40TCL_TEX_FILTER_MAG_LINEAR;
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		filter |= NV40TCL_TEX_FILTER_MAG_NEAREST;
+		break;
+	}
+
+	switch (cso->min_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |= NV40TCL_TEX_FILTER_MIN_LINEAR_MIPMAP_NEAREST;
+			break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV40TCL_TEX_FILTER_MIN_LINEAR_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV40TCL_TEX_FILTER_MIN_LINEAR;
+			break;
+		}
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		switch (cso->min_mip_filter) {
+		case PIPE_TEX_MIPFILTER_NEAREST:
+			filter |= NV40TCL_TEX_FILTER_MIN_NEAREST_MIPMAP_NEAREST;
+		break;
+		case PIPE_TEX_MIPFILTER_LINEAR:
+			filter |= NV40TCL_TEX_FILTER_MIN_NEAREST_MIPMAP_LINEAR;
+			break;
+		case PIPE_TEX_MIPFILTER_NONE:
+		default:
+			filter |= NV40TCL_TEX_FILTER_MIN_NEAREST;
+			break;
+		}
+		break;
+	}
+
+	ps->filt = filter;
+
+	{
+		float limit;
+
+		limit = CLAMP(cso->lod_bias, -16.0, 15.0);
+		ps->filt |= (int)(cso->lod_bias * 256.0) & 0x1fff;
+
+		limit = CLAMP(cso->max_lod, 0.0, 15.0);
+		ps->en |= (int)(limit * 256.0) << 7;
+
+		limit = CLAMP(cso->min_lod, 0.0, 15.0);
+		ps->en |= (int)(limit * 256.0) << 19;
+	}
+
+
+	if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+		switch (cso->compare_func) {
+		case PIPE_FUNC_NEVER:
+			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_NEVER;
+			break;
+		case PIPE_FUNC_GREATER:
+			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_GREATER;
+			break;
+		case PIPE_FUNC_EQUAL:
+			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_EQUAL;
+			break;
+		case PIPE_FUNC_GEQUAL:
+			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_GEQUAL;
+			break;
+		case PIPE_FUNC_LESS:
+			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_LESS;
+			break;
+		case PIPE_FUNC_NOTEQUAL:
+			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_NOTEQUAL;
+			break;
+		case PIPE_FUNC_LEQUAL:
+			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_LEQUAL;
+			break;
+		case PIPE_FUNC_ALWAYS:
+			ps->wrap |= NV40TCL_TEX_WRAP_RCOMP_ALWAYS;
+			break;
+		default:
+			break;
+		}
+	}
+
+	ps->bcol = ((float_to_ubyte(cso->border_color[3]) << 24) |
+		    (float_to_ubyte(cso->border_color[0]) << 16) |
+		    (float_to_ubyte(cso->border_color[1]) <<  8) |
+		    (float_to_ubyte(cso->border_color[2]) <<  0));
+
+	return (void *)ps;
+}
+
+static void
+nv40_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		nv40->tex_sampler[unit] = sampler[unit];
+		nv40->dirty_samplers |= (1 << unit);
+	}
+
+	for (unit = nr; unit < nv40->nr_samplers; unit++) {
+		nv40->tex_sampler[unit] = NULL;
+		nv40->dirty_samplers |= (1 << unit);
+	}
+
+	nv40->nr_samplers = nr;
+	nv40->dirty |= NV40_NEW_SAMPLER;
+}
+
+static void
+nv40_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void
+nv40_set_sampler_texture(struct pipe_context *pipe, unsigned nr,
+			 struct pipe_texture **miptree)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	unsigned unit;
+
+	for (unit = 0; unit < nr; unit++) {
+		pipe_texture_reference((struct pipe_texture **)
+				       &nv40->tex_miptree[unit], miptree[unit]);
+		nv40->dirty_samplers |= (1 << unit);
+	}
+
+	for (unit = nr; unit < nv40->nr_textures; unit++) {
+		pipe_texture_reference((struct pipe_texture **)
+				       &nv40->tex_miptree[unit], NULL);
+		nv40->dirty_samplers |= (1 << unit);
+	}
+
+	nv40->nr_textures = nr;
+	nv40->dirty |= NV40_NEW_SAMPLER;
+}
+
+static void *
+nv40_rasterizer_state_create(struct pipe_context *pipe,
+			     const struct pipe_rasterizer_state *cso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv40_rasterizer_state *rsso = CALLOC(1, sizeof(*rsso));
+	struct nouveau_stateobj *so = so_new(32, 0);
+	struct nouveau_grobj *curie = nv40->screen->curie;
+
+	/*XXX: ignored:
+	 * 	light_twoside
+	 * 	point_smooth -nohw
+	 * 	multisample
+	 */
+
+	so_method(so, curie, NV40TCL_SHADE_MODEL, 1);
+	so_data  (so, cso->flatshade ? NV40TCL_SHADE_MODEL_FLAT :
+				       NV40TCL_SHADE_MODEL_SMOOTH);
+
+	so_method(so, curie, NV40TCL_LINE_WIDTH, 2);
+	so_data  (so, (unsigned char)(cso->line_width * 8.0) & 0xff);
+	so_data  (so, cso->line_smooth ? 1 : 0);
+	so_method(so, curie, NV40TCL_LINE_STIPPLE_ENABLE, 2);
+	so_data  (so, cso->line_stipple_enable ? 1 : 0);
+	so_data  (so, (cso->line_stipple_pattern << 16) |
+		       cso->line_stipple_factor);
+
+	so_method(so, curie, NV40TCL_POINT_SIZE, 1);
+	so_data  (so, fui(cso->point_size));
+
+	so_method(so, curie, NV40TCL_POLYGON_MODE_FRONT, 6);
+	if (cso->front_winding == PIPE_WINDING_CCW) {
+		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
+		so_data(so, nvgl_polygon_mode(cso->fill_cw));
+		switch (cso->cull_mode) {
+		case PIPE_WINDING_CCW:
+			so_data(so, NV40TCL_CULL_FACE_FRONT);
+			break;
+		case PIPE_WINDING_CW:
+			so_data(so, NV40TCL_CULL_FACE_BACK);
+			break;
+		case PIPE_WINDING_BOTH:
+			so_data(so, NV40TCL_CULL_FACE_FRONT_AND_BACK);
+			break;
+		default:
+			so_data(so, NV40TCL_CULL_FACE_BACK);
+			break;
+		}
+		so_data(so, NV40TCL_FRONT_FACE_CCW);
+	} else {
+		so_data(so, nvgl_polygon_mode(cso->fill_cw));
+		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
+		switch (cso->cull_mode) {
+		case PIPE_WINDING_CCW:
+			so_data(so, NV40TCL_CULL_FACE_BACK);
+			break;
+		case PIPE_WINDING_CW:
+			so_data(so, NV40TCL_CULL_FACE_FRONT);
+			break;
+		case PIPE_WINDING_BOTH:
+			so_data(so, NV40TCL_CULL_FACE_FRONT_AND_BACK);
+			break;
+		default:
+			so_data(so, NV40TCL_CULL_FACE_BACK);
+			break;
+		}
+		so_data(so, NV40TCL_FRONT_FACE_CW);
+	}
+	so_data(so, cso->poly_smooth ? 1 : 0);
+	so_data(so, (cso->cull_mode != PIPE_WINDING_NONE) ? 1 : 0);
+
+	so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1);
+	so_data  (so, cso->poly_stipple_enable ? 1 : 0);
+
+	so_method(so, curie, NV40TCL_POLYGON_OFFSET_POINT_ENABLE, 3);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if (cso->offset_cw || cso->offset_ccw) {
+		so_method(so, curie, NV40TCL_POLYGON_OFFSET_FACTOR, 2);
+		so_data  (so, fui(cso->offset_scale));
+		so_data  (so, fui(cso->offset_units * 2));
+	}
+
+	so_method(so, curie, NV40TCL_POINT_SPRITE, 1);
+	if (cso->point_sprite) {
+		unsigned psctl = (1 << 0), i;
+
+		for (i = 0; i < 8; i++) {
+			if (cso->sprite_coord_mode[i] != PIPE_SPRITE_COORD_NONE)
+				psctl |= (1 << (8 + i));
+		}
+
+		so_data(so, psctl);
+	} else {
+		so_data(so, 0);
+	}
+
+	so_ref(so, &rsso->so);
+	rsso->pipe = *cso;
+	return (void *)rsso;
+}
+
+static void
+nv40_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->rasterizer = hwcso;
+	nv40->dirty |= NV40_NEW_RAST;
+	nv40->draw_dirty |= NV40_NEW_RAST;
+}
+
+static void
+nv40_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv40_rasterizer_state *rsso = hwcso;
+
+	so_ref(NULL, &rsso->so);
+	FREE(rsso);
+}
+
+static void *
+nv40_depth_stencil_alpha_state_create(struct pipe_context *pipe,
+			const struct pipe_depth_stencil_alpha_state *cso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv40_zsa_state *zsaso = CALLOC(1, sizeof(*zsaso));
+	struct nouveau_stateobj *so = so_new(32, 0);
+	struct nouveau_grobj *curie = nv40->screen->curie;
+
+	so_method(so, curie, NV40TCL_DEPTH_FUNC, 3);
+	so_data  (so, nvgl_comparison_op(cso->depth.func));
+	so_data  (so, cso->depth.writemask ? 1 : 0);
+	so_data  (so, cso->depth.enabled ? 1 : 0);
+
+	so_method(so, curie, NV40TCL_ALPHA_TEST_ENABLE, 3);
+	so_data  (so, cso->alpha.enabled ? 1 : 0);
+	so_data  (so, nvgl_comparison_op(cso->alpha.func));
+	so_data  (so, float_to_ubyte(cso->alpha.ref_value));
+
+	if (cso->stencil[0].enabled) {
+		so_method(so, curie, NV40TCL_STENCIL_FRONT_ENABLE, 8);
+		so_data  (so, cso->stencil[0].enabled ? 1 : 0);
+		so_data  (so, cso->stencil[0].writemask);
+		so_data  (so, nvgl_comparison_op(cso->stencil[0].func));
+		so_data  (so, cso->stencil[0].ref_value);
+		so_data  (so, cso->stencil[0].valuemask);
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].fail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].zfail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].zpass_op));
+	} else {
+		so_method(so, curie, NV40TCL_STENCIL_FRONT_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	if (cso->stencil[1].enabled) {
+		so_method(so, curie, NV40TCL_STENCIL_BACK_ENABLE, 8);
+		so_data  (so, cso->stencil[1].enabled ? 1 : 0);
+		so_data  (so, cso->stencil[1].writemask);
+		so_data  (so, nvgl_comparison_op(cso->stencil[1].func));
+		so_data  (so, cso->stencil[1].ref_value);
+		so_data  (so, cso->stencil[1].valuemask);
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].fail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].zfail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].zpass_op));
+	} else {
+		so_method(so, curie, NV40TCL_STENCIL_BACK_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_ref(so, &zsaso->so);
+	zsaso->pipe = *cso;
+	return (void *)zsaso;
+}
+
+static void
+nv40_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->zsa = hwcso;
+	nv40->dirty |= NV40_NEW_ZSA;
+}
+
+static void
+nv40_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv40_zsa_state *zsaso = hwcso;
+
+	so_ref(NULL, &zsaso->so);
+	FREE(zsaso);
+}
+
+static void *
+nv40_vp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv40_vertex_program *vp;
+
+	vp = CALLOC(1, sizeof(struct nv40_vertex_program));
+	vp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+	vp->draw = draw_create_vertex_shader(nv40->draw, &vp->pipe);
+
+	return (void *)vp;
+}
+
+static void
+nv40_vp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->vertprog = hwcso;
+	nv40->dirty |= NV40_NEW_VERTPROG;
+	nv40->draw_dirty |= NV40_NEW_VERTPROG;
+}
+
+static void
+nv40_vp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv40_vertex_program *vp = hwcso;
+
+	draw_delete_vertex_shader(nv40->draw, vp->draw);
+	nv40_vertprog_destroy(nv40, vp);
+	FREE((void*)vp->pipe.tokens);
+	FREE(vp);
+}
+
+static void *
+nv40_fp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	struct nv40_fragment_program *fp;
+
+	fp = CALLOC(1, sizeof(struct nv40_fragment_program));
+	fp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+
+	tgsi_scan_shader(fp->pipe.tokens, &fp->info);
+
+	return (void *)fp;
+}
+
+static void
+nv40_fp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->fragprog = hwcso;
+	nv40->dirty |= NV40_NEW_FRAGPROG;
+}
+
+static void
+nv40_fp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv40_fragment_program *fp = hwcso;
+
+	nv40_fragprog_destroy(nv40, fp);
+	FREE((void*)fp->pipe.tokens);
+	FREE(fp);
+}
+
+static void
+nv40_set_blend_color(struct pipe_context *pipe,
+		     const struct pipe_blend_color *bcol)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->blend_colour = *bcol;
+	nv40->dirty |= NV40_NEW_BCOL;
+}
+
+static void
+nv40_set_clip_state(struct pipe_context *pipe,
+		    const struct pipe_clip_state *clip)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->clip = *clip;
+	nv40->dirty |= NV40_NEW_UCP;
+	nv40->draw_dirty |= NV40_NEW_UCP;
+}
+
+static void
+nv40_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
+			 const struct pipe_constant_buffer *buf )
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->constbuf[shader] = buf->buffer;
+	nv40->constbuf_nr[shader] = buf->buffer->size / (4 * sizeof(float));
+
+	if (shader == PIPE_SHADER_VERTEX) {
+		nv40->dirty |= NV40_NEW_VERTPROG;
+	} else
+	if (shader == PIPE_SHADER_FRAGMENT) {
+		nv40->dirty |= NV40_NEW_FRAGPROG;
+	}
+}
+
+static void
+nv40_set_framebuffer_state(struct pipe_context *pipe,
+			   const struct pipe_framebuffer_state *fb)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->framebuffer = *fb;
+	nv40->dirty |= NV40_NEW_FB;
+}
+
+static void
+nv40_set_polygon_stipple(struct pipe_context *pipe,
+			 const struct pipe_poly_stipple *stipple)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	memcpy(nv40->stipple, stipple->stipple, 4 * 32);
+	nv40->dirty |= NV40_NEW_STIPPLE;
+}
+
+static void
+nv40_set_scissor_state(struct pipe_context *pipe,
+		       const struct pipe_scissor_state *s)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->scissor = *s;
+	nv40->dirty |= NV40_NEW_SCISSOR;
+}
+
+static void
+nv40_set_viewport_state(struct pipe_context *pipe,
+			const struct pipe_viewport_state *vpt)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->viewport = *vpt;
+	nv40->dirty |= NV40_NEW_VIEWPORT;
+	nv40->draw_dirty |= NV40_NEW_VIEWPORT;
+}
+
+static void
+nv40_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
+			const struct pipe_vertex_buffer *vb)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	memcpy(nv40->vtxbuf, vb, sizeof(*vb) * count);
+	nv40->vtxbuf_nr = count;
+
+	nv40->dirty |= NV40_NEW_ARRAYS;
+	nv40->draw_dirty |= NV40_NEW_ARRAYS;
+}
+
+static void
+nv40_set_vertex_elements(struct pipe_context *pipe, unsigned count,
+			 const struct pipe_vertex_element *ve)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	memcpy(nv40->vtxelt, ve, sizeof(*ve) * count);
+	nv40->vtxelt_nr = count;
+
+	nv40->dirty |= NV40_NEW_ARRAYS;
+	nv40->draw_dirty |= NV40_NEW_ARRAYS;
+}
+
+static void
+nv40_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+
+	nv40->edgeflags = bitfield;
+	nv40->dirty |= NV40_NEW_ARRAYS;
+	nv40->draw_dirty |= NV40_NEW_ARRAYS;
+}
+
+void
+nv40_init_state_functions(struct nv40_context *nv40)
+{
+	nv40->pipe.create_blend_state = nv40_blend_state_create;
+	nv40->pipe.bind_blend_state = nv40_blend_state_bind;
+	nv40->pipe.delete_blend_state = nv40_blend_state_delete;
+
+	nv40->pipe.create_sampler_state = nv40_sampler_state_create;
+	nv40->pipe.bind_sampler_states = nv40_sampler_state_bind;
+	nv40->pipe.delete_sampler_state = nv40_sampler_state_delete;
+	nv40->pipe.set_sampler_textures = nv40_set_sampler_texture;
+
+	nv40->pipe.create_rasterizer_state = nv40_rasterizer_state_create;
+	nv40->pipe.bind_rasterizer_state = nv40_rasterizer_state_bind;
+	nv40->pipe.delete_rasterizer_state = nv40_rasterizer_state_delete;
+
+	nv40->pipe.create_depth_stencil_alpha_state =
+		nv40_depth_stencil_alpha_state_create;
+	nv40->pipe.bind_depth_stencil_alpha_state =
+		nv40_depth_stencil_alpha_state_bind;
+	nv40->pipe.delete_depth_stencil_alpha_state =
+		nv40_depth_stencil_alpha_state_delete;
+
+	nv40->pipe.create_vs_state = nv40_vp_state_create;
+	nv40->pipe.bind_vs_state = nv40_vp_state_bind;
+	nv40->pipe.delete_vs_state = nv40_vp_state_delete;
+
+	nv40->pipe.create_fs_state = nv40_fp_state_create;
+	nv40->pipe.bind_fs_state = nv40_fp_state_bind;
+	nv40->pipe.delete_fs_state = nv40_fp_state_delete;
+
+	nv40->pipe.set_blend_color = nv40_set_blend_color;
+	nv40->pipe.set_clip_state = nv40_set_clip_state;
+	nv40->pipe.set_constant_buffer = nv40_set_constant_buffer;
+	nv40->pipe.set_framebuffer_state = nv40_set_framebuffer_state;
+	nv40->pipe.set_polygon_stipple = nv40_set_polygon_stipple;
+	nv40->pipe.set_scissor_state = nv40_set_scissor_state;
+	nv40->pipe.set_viewport_state = nv40_set_viewport_state;
+
+	nv40->pipe.set_edgeflags = nv40_set_edgeflags;
+	nv40->pipe.set_vertex_buffers = nv40_set_vertex_buffers;
+	nv40->pipe.set_vertex_elements = nv40_set_vertex_elements;
+}
+
diff --git a/src/gallium/drivers/nv40/nv40_state.h b/src/gallium/drivers/nv40/nv40_state.h
new file mode 100644
index 0000000000..9c55903ae3
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_state.h
@@ -0,0 +1,91 @@
+#ifndef __NV40_STATE_H__
+#define __NV40_STATE_H__
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_scan.h"
+
+struct nv40_sampler_state {
+	uint32_t fmt;
+	uint32_t wrap;
+	uint32_t en;
+	uint32_t filt;
+	uint32_t bcol;
+};
+
+struct nv40_vertex_program_exec {
+	uint32_t data[4];
+	boolean has_branch_offset;
+	int const_index;
+};
+
+struct nv40_vertex_program_data {
+	int index; /* immediates == -1 */
+	float value[4];
+};
+
+struct nv40_vertex_program {
+	struct pipe_shader_state pipe;
+
+	struct draw_vertex_shader *draw;
+
+	boolean translated;
+
+	struct pipe_clip_state ucp;
+
+	struct nv40_vertex_program_exec *insns;
+	unsigned nr_insns;
+	struct nv40_vertex_program_data *consts;
+	unsigned nr_consts;
+
+	struct nouveau_resource *exec;
+	unsigned exec_start;
+	struct nouveau_resource *data;
+	unsigned data_start;
+	unsigned data_start_min;
+
+	uint32_t ir;
+	uint32_t or;
+	uint32_t clip_ctrl;
+	struct nouveau_stateobj *so;
+};
+
+struct nv40_fragment_program_data {
+	unsigned offset;
+	unsigned index;
+};
+
+struct nv40_fragment_program {
+	struct pipe_shader_state pipe;
+	struct tgsi_shader_info info;
+
+	boolean translated;
+	unsigned samplers;
+
+	uint32_t *insn;
+	int       insn_len;
+
+	struct nv40_fragment_program_data *consts;
+	unsigned nr_consts;
+
+	struct pipe_buffer *buffer;
+
+	uint32_t fp_control;
+	struct nouveau_stateobj *so;
+};
+
+struct nv40_miptree {
+	struct pipe_texture base;
+
+	struct pipe_buffer *buffer;
+	uint total_size;
+
+	struct pipe_texture *shadow_tex;
+	struct pipe_surface *shadow_surface;
+
+	struct {
+		uint pitch;
+		uint *image_offset;
+	} level[PIPE_MAX_TEXTURE_LEVELS];
+};
+
+#endif
diff --git a/src/gallium/drivers/nv40/nv40_state_blend.c b/src/gallium/drivers/nv40/nv40_state_blend.c
new file mode 100644
index 0000000000..95e6d7394f
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_state_blend.c
@@ -0,0 +1,40 @@
+#include "nv40_context.h"
+
+static boolean
+nv40_state_blend_validate(struct nv40_context *nv40)
+{
+	so_ref(nv40->blend->so, &nv40->state.hw[NV40_STATE_BLEND]);
+	return TRUE;
+}
+
+struct nv40_state_entry nv40_state_blend = {
+	.validate = nv40_state_blend_validate,
+	.dirty = {
+		.pipe = NV40_NEW_BLEND,
+		.hw = NV40_STATE_BLEND
+	}
+};
+
+static boolean
+nv40_state_blend_colour_validate(struct nv40_context *nv40)
+{
+	struct nouveau_stateobj *so = so_new(2, 0);
+	struct pipe_blend_color *bcol = &nv40->blend_colour;
+
+	so_method(so, nv40->screen->curie, NV40TCL_BLEND_COLOR, 1);
+	so_data  (so, ((float_to_ubyte(bcol->color[3]) << 24) |
+		       (float_to_ubyte(bcol->color[0]) << 16) |
+		       (float_to_ubyte(bcol->color[1]) <<  8) |
+		       (float_to_ubyte(bcol->color[2]) <<  0)));
+
+	so_ref(so, &nv40->state.hw[NV40_STATE_BCOL]);
+	return TRUE;
+}
+
+struct nv40_state_entry nv40_state_blend_colour = {
+	.validate = nv40_state_blend_colour_validate,
+	.dirty = {
+		.pipe = NV40_NEW_BCOL,
+		.hw = NV40_STATE_BCOL
+	}
+};
diff --git a/src/gallium/drivers/nv40/nv40_state_emit.c b/src/gallium/drivers/nv40/nv40_state_emit.c
new file mode 100644
index 0000000000..ce859def10
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_state_emit.c
@@ -0,0 +1,184 @@
+#include "nv40_context.h"
+#include "nv40_state.h"
+#include "draw/draw_context.h"
+
+static struct nv40_state_entry *render_states[] = {
+	&nv40_state_framebuffer,
+	&nv40_state_rasterizer,
+	&nv40_state_scissor,
+	&nv40_state_stipple,
+	&nv40_state_fragprog,
+	&nv40_state_fragtex,
+	&nv40_state_vertprog,
+	&nv40_state_blend,
+	&nv40_state_blend_colour,
+	&nv40_state_zsa,
+	&nv40_state_viewport,
+	&nv40_state_vbo,
+	NULL
+};
+
+static struct nv40_state_entry *swtnl_states[] = {
+	&nv40_state_framebuffer,
+	&nv40_state_rasterizer,
+	&nv40_state_scissor,
+	&nv40_state_stipple,
+	&nv40_state_fragprog,
+	&nv40_state_fragtex,
+	&nv40_state_vertprog,
+	&nv40_state_blend,
+	&nv40_state_blend_colour,
+	&nv40_state_zsa,
+	&nv40_state_viewport,
+	&nv40_state_vtxfmt,
+	NULL
+};
+
+static void
+nv40_state_do_validate(struct nv40_context *nv40,
+		       struct nv40_state_entry **states)
+{
+	const struct pipe_framebuffer_state *fb = &nv40->framebuffer;
+	unsigned i;
+
+	for (i = 0; i < fb->nr_cbufs; i++)
+		fb->cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED;
+	if (fb->zsbuf)
+		fb->zsbuf->status = PIPE_SURFACE_STATUS_DEFINED;
+
+	while (*states) {
+		struct nv40_state_entry *e = *states;
+
+		if (nv40->dirty & e->dirty.pipe) {
+			if (e->validate(nv40))
+				nv40->state.dirty |= (1ULL << e->dirty.hw);
+		}
+
+		states++;
+	}
+	nv40->dirty = 0;
+}
+
+void
+nv40_state_emit(struct nv40_context *nv40)
+{
+	struct nv40_state *state = &nv40->state;
+	struct nv40_screen *screen = nv40->screen;
+	unsigned i, samplers;
+	uint64_t states;
+
+	if (nv40->pctx_id != screen->cur_pctx) {
+		for (i = 0; i < NV40_STATE_MAX; i++) {
+			if (state->hw[i] && screen->state[i] != state->hw[i])
+				state->dirty |= (1ULL << i);
+		}
+
+		screen->cur_pctx = nv40->pctx_id;
+	}
+
+	for (i = 0, states = state->dirty; states; i++) {
+		if (!(states & (1ULL << i)))
+			continue;
+		so_ref (state->hw[i], &nv40->screen->state[i]);
+		if (state->hw[i])
+			so_emit(nv40->nvws, nv40->screen->state[i]);
+		states &= ~(1ULL << i);
+	}
+
+	if (state->dirty & ((1ULL << NV40_STATE_FRAGPROG) |
+			    (1ULL << NV40_STATE_FRAGTEX0))) {
+		BEGIN_RING(curie, NV40TCL_TEX_CACHE_CTL, 1);
+		OUT_RING  (2);
+		BEGIN_RING(curie, NV40TCL_TEX_CACHE_CTL, 1);
+		OUT_RING  (1);
+	}
+
+	state->dirty = 0;
+
+	so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_FB]);
+	for (i = 0, samplers = state->fp_samplers; i < 16 && samplers; i++) {
+		if (!(samplers & (1 << i)))
+			continue;
+		so_emit_reloc_markers(nv40->nvws,
+				      state->hw[NV40_STATE_FRAGTEX0+i]);
+		samplers &= ~(1ULL << i);
+	}
+	so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_FRAGPROG]);
+	if (state->hw[NV40_STATE_VTXBUF] && nv40->render_mode == HW)
+		so_emit_reloc_markers(nv40->nvws, state->hw[NV40_STATE_VTXBUF]);
+}
+
+boolean
+nv40_state_validate(struct nv40_context *nv40)
+{
+	boolean was_sw = nv40->fallback_swtnl ? TRUE : FALSE;
+
+	if (nv40->render_mode != HW) {
+		/* Don't even bother trying to go back to hw if none
+		 * of the states that caused swtnl previously have changed.
+		 */
+		if ((nv40->fallback_swtnl & nv40->dirty)
+				!= nv40->fallback_swtnl)
+			return FALSE;
+
+		/* Attempt to go to hwtnl again */
+		nv40->pipe.flush(&nv40->pipe, 0, NULL);
+		nv40->dirty |= (NV40_NEW_VIEWPORT |
+				NV40_NEW_VERTPROG |
+				NV40_NEW_ARRAYS);
+		nv40->render_mode = HW;
+	}
+
+	nv40_state_do_validate(nv40, render_states);
+	if (nv40->fallback_swtnl || nv40->fallback_swrast)
+		return FALSE;
+	
+	if (was_sw)
+		NOUVEAU_ERR("swtnl->hw\n");
+
+	return TRUE;
+}
+
+boolean
+nv40_state_validate_swtnl(struct nv40_context *nv40)
+{
+	struct draw_context *draw = nv40->draw;
+
+	/* Setup for swtnl */
+	if (nv40->render_mode == HW) {
+		NOUVEAU_ERR("hw->swtnl 0x%08x\n", nv40->fallback_swtnl);
+		nv40->pipe.flush(&nv40->pipe, 0, NULL);
+		nv40->dirty |= (NV40_NEW_VIEWPORT |
+				NV40_NEW_VERTPROG |
+				NV40_NEW_ARRAYS);
+		nv40->render_mode = SWTNL;
+	}
+
+	if (nv40->draw_dirty & NV40_NEW_VERTPROG)
+		draw_bind_vertex_shader(draw, nv40->vertprog->draw);
+
+	if (nv40->draw_dirty & NV40_NEW_RAST)
+		draw_set_rasterizer_state(draw, &nv40->rasterizer->pipe);
+
+	if (nv40->draw_dirty & NV40_NEW_UCP)
+		draw_set_clip_state(draw, &nv40->clip);
+
+	if (nv40->draw_dirty & NV40_NEW_VIEWPORT)
+		draw_set_viewport_state(draw, &nv40->viewport);
+
+	if (nv40->draw_dirty & NV40_NEW_ARRAYS) {
+		draw_set_edgeflags(draw, nv40->edgeflags);
+		draw_set_vertex_buffers(draw, nv40->vtxbuf_nr, nv40->vtxbuf);
+		draw_set_vertex_elements(draw, nv40->vtxelt_nr, nv40->vtxelt);	
+	}
+
+	nv40_state_do_validate(nv40, swtnl_states);
+	if (nv40->fallback_swrast) {
+		NOUVEAU_ERR("swtnl->swrast 0x%08x\n", nv40->fallback_swrast);
+		return FALSE;
+	}
+
+	nv40->draw_dirty = 0;
+	return TRUE;
+}
+
diff --git a/src/gallium/drivers/nv40/nv40_state_fb.c b/src/gallium/drivers/nv40/nv40_state_fb.c
new file mode 100644
index 0000000000..454abad31f
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_state_fb.c
@@ -0,0 +1,162 @@
+#include "nv40_context.h"
+#include "nouveau/nouveau_util.h"
+
+static struct pipe_buffer *
+nv40_surface_buffer(struct pipe_surface *surface)
+{
+	struct nv40_miptree *mt = (struct nv40_miptree *)surface->texture;
+	return mt->buffer;
+}
+
+static boolean
+nv40_state_framebuffer_validate(struct nv40_context *nv40)
+{
+	struct pipe_framebuffer_state *fb = &nv40->framebuffer;
+	struct pipe_surface *rt[4], *zeta;
+	uint32_t rt_enable, rt_format;
+	int i, colour_format = 0, zeta_format = 0;
+	struct nouveau_stateobj *so = so_new(64, 10);
+	unsigned rt_flags = NOUVEAU_BO_RDWR | NOUVEAU_BO_VRAM;
+	unsigned w = fb->width;
+	unsigned h = fb->height;
+
+	rt_enable = 0;
+	for (i = 0; i < fb->nr_cbufs; i++) {
+		if (colour_format) {
+			assert(colour_format == fb->cbufs[i]->format);
+		} else {
+			colour_format = fb->cbufs[i]->format;
+			rt_enable |= (NV40TCL_RT_ENABLE_COLOR0 << i);
+			rt[i] = fb->cbufs[i];
+		}
+	}
+
+	if (rt_enable & (NV40TCL_RT_ENABLE_COLOR1 | NV40TCL_RT_ENABLE_COLOR2 |
+			 NV40TCL_RT_ENABLE_COLOR3))
+		rt_enable |= NV40TCL_RT_ENABLE_MRT;
+
+	if (fb->zsbuf) {
+		zeta_format = fb->zsbuf->format;
+		zeta = fb->zsbuf;
+	}
+
+	if (!(rt[0]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) {
+		assert(!(fb->width & (fb->width - 1)) && !(fb->height & (fb->height - 1)));
+		for (i = 1; i < fb->nr_cbufs; i++)
+			assert(!(rt[i]->texture->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR));
+
+		rt_format = NV40TCL_RT_FORMAT_TYPE_SWIZZLED |
+		            log2i(fb->width) << NV40TCL_RT_FORMAT_LOG2_WIDTH_SHIFT |
+		            log2i(fb->height) << NV40TCL_RT_FORMAT_LOG2_HEIGHT_SHIFT;
+	}
+	else
+		rt_format = NV40TCL_RT_FORMAT_TYPE_LINEAR;
+
+	switch (colour_format) {
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case 0:
+		rt_format |= NV40TCL_RT_FORMAT_COLOR_A8R8G8B8;
+		break;
+	case PIPE_FORMAT_R5G6B5_UNORM:
+		rt_format |= NV40TCL_RT_FORMAT_COLOR_R5G6B5;
+		break;
+	default:
+		assert(0);
+	}
+
+	switch (zeta_format) {
+	case PIPE_FORMAT_Z16_UNORM:
+		rt_format |= NV40TCL_RT_FORMAT_ZETA_Z16;
+		break;
+	case PIPE_FORMAT_Z24S8_UNORM:
+	case 0:
+		rt_format |= NV40TCL_RT_FORMAT_ZETA_Z24S8;
+		break;
+	default:
+		assert(0);
+	}
+
+	if (rt_enable & NV40TCL_RT_ENABLE_COLOR0) {
+		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR0, 1);
+		so_reloc (so, nv40_surface_buffer(rt[0]), 0, rt_flags | NOUVEAU_BO_OR,
+			  nv40->nvws->channel->vram->handle,
+			  nv40->nvws->channel->gart->handle);
+		so_method(so, nv40->screen->curie, NV40TCL_COLOR0_PITCH, 2);
+		so_data  (so, rt[0]->stride);
+		so_reloc (so, nv40_surface_buffer(rt[0]), rt[0]->offset, rt_flags |
+			  NOUVEAU_BO_LOW, 0, 0);
+	}
+
+	if (rt_enable & NV40TCL_RT_ENABLE_COLOR1) {
+		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR1, 1);
+		so_reloc (so, nv40_surface_buffer(rt[1]), 0, rt_flags | NOUVEAU_BO_OR,
+			  nv40->nvws->channel->vram->handle,
+			  nv40->nvws->channel->gart->handle);
+		so_method(so, nv40->screen->curie, NV40TCL_COLOR1_OFFSET, 2);
+		so_reloc (so, nv40_surface_buffer(rt[1]), rt[1]->offset, rt_flags |
+			  NOUVEAU_BO_LOW, 0, 0);
+		so_data  (so, rt[1]->stride);
+	}
+
+	if (rt_enable & NV40TCL_RT_ENABLE_COLOR2) {
+		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR2, 1);
+		so_reloc (so, nv40_surface_buffer(rt[2]), 0, rt_flags | NOUVEAU_BO_OR,
+			  nv40->nvws->channel->vram->handle,
+			  nv40->nvws->channel->gart->handle);
+		so_method(so, nv40->screen->curie, NV40TCL_COLOR2_OFFSET, 1);
+		so_reloc (so, nv40_surface_buffer(rt[2]), rt[2]->offset, rt_flags |
+			  NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, nv40->screen->curie, NV40TCL_COLOR2_PITCH, 1);
+		so_data  (so, rt[2]->stride);
+	}
+
+	if (rt_enable & NV40TCL_RT_ENABLE_COLOR3) {
+		so_method(so, nv40->screen->curie, NV40TCL_DMA_COLOR3, 1);
+		so_reloc (so, nv40_surface_buffer(rt[3]), 0, rt_flags | NOUVEAU_BO_OR,
+			  nv40->nvws->channel->vram->handle,
+			  nv40->nvws->channel->gart->handle);
+		so_method(so, nv40->screen->curie, NV40TCL_COLOR3_OFFSET, 1);
+		so_reloc (so, nv40_surface_buffer(rt[3]), rt[3]->offset, rt_flags |
+			  NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, nv40->screen->curie, NV40TCL_COLOR3_PITCH, 1);
+		so_data  (so, rt[3]->stride);
+	}
+
+	if (zeta_format) {
+		so_method(so, nv40->screen->curie, NV40TCL_DMA_ZETA, 1);
+		so_reloc (so, nv40_surface_buffer(zeta), 0, rt_flags | NOUVEAU_BO_OR,
+			  nv40->nvws->channel->vram->handle,
+			  nv40->nvws->channel->gart->handle);
+		so_method(so, nv40->screen->curie, NV40TCL_ZETA_OFFSET, 1);
+		so_reloc (so, nv40_surface_buffer(zeta), zeta->offset, rt_flags |
+			  NOUVEAU_BO_LOW, 0, 0);
+		so_method(so, nv40->screen->curie, NV40TCL_ZETA_PITCH, 1);
+		so_data  (so, zeta->stride);
+	}
+
+	so_method(so, nv40->screen->curie, NV40TCL_RT_ENABLE, 1);
+	so_data  (so, rt_enable);
+	so_method(so, nv40->screen->curie, NV40TCL_RT_HORIZ, 3);
+	so_data  (so, (w << 16) | 0);
+	so_data  (so, (h << 16) | 0);
+	so_data  (so, rt_format);
+	so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_HORIZ, 2);
+	so_data  (so, (w << 16) | 0);
+	so_data  (so, (h << 16) | 0);
+	so_method(so, nv40->screen->curie, NV40TCL_VIEWPORT_CLIP_HORIZ(0), 2);
+	so_data  (so, ((w - 1) << 16) | 0);
+	so_data  (so, ((h - 1) << 16) | 0);
+	so_method(so, nv40->screen->curie, 0x1d88, 1);
+	so_data  (so, (1 << 12) | h);
+
+	so_ref(so, &nv40->state.hw[NV40_STATE_FB]);
+	return TRUE;
+}
+
+struct nv40_state_entry nv40_state_framebuffer = {
+	.validate = nv40_state_framebuffer_validate,
+	.dirty = {
+		.pipe = NV40_NEW_FB,
+		.hw = NV40_STATE_FB
+	}
+};
diff --git a/src/gallium/drivers/nv40/nv40_state_rasterizer.c b/src/gallium/drivers/nv40/nv40_state_rasterizer.c
new file mode 100644
index 0000000000..9ecda5990f
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_state_rasterizer.c
@@ -0,0 +1,17 @@
+#include "nv40_context.h"
+
+static boolean
+nv40_state_rasterizer_validate(struct nv40_context *nv40)
+{
+	so_ref(nv40->rasterizer->so,
+	       &nv40->state.hw[NV40_STATE_RAST]);
+	return TRUE;
+}
+
+struct nv40_state_entry nv40_state_rasterizer = {
+	.validate = nv40_state_rasterizer_validate,
+	.dirty = {
+		.pipe = NV40_NEW_RAST,
+		.hw = NV40_STATE_RAST
+	}
+};
diff --git a/src/gallium/drivers/nv40/nv40_state_scissor.c b/src/gallium/drivers/nv40/nv40_state_scissor.c
new file mode 100644
index 0000000000..285239ef41
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_state_scissor.c
@@ -0,0 +1,35 @@
+#include "nv40_context.h"
+
+static boolean
+nv40_state_scissor_validate(struct nv40_context *nv40)
+{
+	struct pipe_rasterizer_state *rast = &nv40->rasterizer->pipe;
+	struct pipe_scissor_state *s = &nv40->scissor;
+	struct nouveau_stateobj *so;
+
+	if (nv40->state.hw[NV40_STATE_SCISSOR] &&
+	    (rast->scissor == 0 && nv40->state.scissor_enabled == 0))
+		return FALSE;
+	nv40->state.scissor_enabled = rast->scissor;
+
+	so = so_new(3, 0);
+	so_method(so, nv40->screen->curie, NV40TCL_SCISSOR_HORIZ, 2);
+	if (nv40->state.scissor_enabled) {
+		so_data  (so, ((s->maxx - s->minx) << 16) | s->minx);
+		so_data  (so, ((s->maxy - s->miny) << 16) | s->miny);
+	} else {
+		so_data  (so, 4096 << 16);
+		so_data  (so, 4096 << 16);
+	}
+
+	so_ref(so, &nv40->state.hw[NV40_STATE_SCISSOR]);
+	return TRUE;
+}
+
+struct nv40_state_entry nv40_state_scissor = {
+	.validate = nv40_state_scissor_validate,
+	.dirty = {
+		.pipe = NV40_NEW_SCISSOR | NV40_NEW_RAST,
+		.hw = NV40_STATE_SCISSOR
+	}
+};
diff --git a/src/gallium/drivers/nv40/nv40_state_stipple.c b/src/gallium/drivers/nv40/nv40_state_stipple.c
new file mode 100644
index 0000000000..b51024ad9b
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_state_stipple.c
@@ -0,0 +1,39 @@
+#include "nv40_context.h"
+
+static boolean
+nv40_state_stipple_validate(struct nv40_context *nv40)
+{
+	struct pipe_rasterizer_state *rast = &nv40->rasterizer->pipe;
+	struct nouveau_grobj *curie = nv40->screen->curie;
+	struct nouveau_stateobj *so;
+
+	if (nv40->state.hw[NV40_STATE_STIPPLE] &&
+	   (rast->poly_stipple_enable == 0 && nv40->state.stipple_enabled == 0))
+		return FALSE;
+
+	if (rast->poly_stipple_enable) {
+		unsigned i;
+
+		so = so_new(35, 0);
+		so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1);
+		so_data  (so, 1);
+		so_method(so, curie, NV40TCL_POLYGON_STIPPLE_PATTERN(0), 32);
+		for (i = 0; i < 32; i++)
+			so_data(so, nv40->stipple[i]);
+	} else {
+		so = so_new(2, 0);
+		so_method(so, curie, NV40TCL_POLYGON_STIPPLE_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_ref(so, &nv40->state.hw[NV40_STATE_STIPPLE]);
+	return TRUE;
+}
+
+struct nv40_state_entry nv40_state_stipple = {
+	.validate = nv40_state_stipple_validate,
+	.dirty = {
+		.pipe = NV40_NEW_STIPPLE | NV40_NEW_RAST,
+		.hw = NV40_STATE_STIPPLE,
+	}
+};
diff --git a/src/gallium/drivers/nv40/nv40_state_viewport.c b/src/gallium/drivers/nv40/nv40_state_viewport.c
new file mode 100644
index 0000000000..869a55b405
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_state_viewport.c
@@ -0,0 +1,67 @@
+#include "nv40_context.h"
+
+static boolean
+nv40_state_viewport_validate(struct nv40_context *nv40)
+{
+	struct pipe_viewport_state *vpt = &nv40->viewport;
+	struct nouveau_stateobj *so;
+	unsigned bypass;
+
+	if (nv40->render_mode == HW && !nv40->rasterizer->pipe.bypass_clipping)
+		bypass = 0;
+	else
+		bypass = 1;
+
+	if (nv40->state.hw[NV40_STATE_VIEWPORT] &&
+	    (bypass || !(nv40->dirty & NV40_NEW_VIEWPORT)) &&
+	    nv40->state.viewport_bypass == bypass)
+		return FALSE;
+	nv40->state.viewport_bypass = bypass;
+
+	so = so_new(11, 0);
+	if (!bypass) {
+		so_method(so, nv40->screen->curie,
+			  NV40TCL_VIEWPORT_TRANSLATE_X, 8);
+		so_data  (so, fui(vpt->translate[0]));
+		so_data  (so, fui(vpt->translate[1]));
+		so_data  (so, fui(vpt->translate[2]));
+		so_data  (so, fui(vpt->translate[3]));
+		so_data  (so, fui(vpt->scale[0]));
+		so_data  (so, fui(vpt->scale[1]));
+		so_data  (so, fui(vpt->scale[2]));
+		so_data  (so, fui(vpt->scale[3]));
+		so_method(so, nv40->screen->curie, 0x1d78, 1);
+		so_data  (so, 1);
+	} else {
+		so_method(so, nv40->screen->curie,
+			  NV40TCL_VIEWPORT_TRANSLATE_X, 8);
+		so_data  (so, fui(0.0));
+		so_data  (so, fui(0.0));
+		so_data  (so, fui(0.0));
+		so_data  (so, fui(0.0));
+		so_data  (so, fui(1.0));
+		so_data  (so, fui(1.0));
+		so_data  (so, fui(1.0));
+		so_data  (so, fui(0.0));
+		/* Not entirely certain what this is yet.  The DDX uses this
+		 * value also as it fixes rendering when you pass
+		 * pre-transformed vertices to the GPU.  My best gusss is that
+		 * this bypasses some culling/clipping stage.  Might be worth
+		 * noting that points/lines are uneffected by whatever this
+		 * value fixes, only filled polygons are effected.
+		 */
+		so_method(so, nv40->screen->curie, 0x1d78, 1);
+		so_data  (so, 0x110);
+	}
+
+	so_ref(so, &nv40->state.hw[NV40_STATE_VIEWPORT]);
+	return TRUE;
+}
+
+struct nv40_state_entry nv40_state_viewport = {
+	.validate = nv40_state_viewport_validate,
+	.dirty = {
+		.pipe = NV40_NEW_VIEWPORT | NV40_NEW_RAST,
+		.hw = NV40_STATE_VIEWPORT
+	}
+};
diff --git a/src/gallium/drivers/nv40/nv40_state_zsa.c b/src/gallium/drivers/nv40/nv40_state_zsa.c
new file mode 100644
index 0000000000..fb760677c8
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_state_zsa.c
@@ -0,0 +1,17 @@
+#include "nv40_context.h"
+
+static boolean
+nv40_state_zsa_validate(struct nv40_context *nv40)
+{
+	so_ref(nv40->zsa->so,
+	       &nv40->state.hw[NV40_STATE_ZSA]);
+	return TRUE;
+}
+
+struct nv40_state_entry nv40_state_zsa = {
+	.validate = nv40_state_zsa_validate,
+	.dirty = {
+		.pipe = NV40_NEW_ZSA,
+		.hw = NV40_STATE_ZSA
+	}
+};
diff --git a/src/gallium/drivers/nv40/nv40_surface.c b/src/gallium/drivers/nv40/nv40_surface.c
new file mode 100644
index 0000000000..c4a5fb20d9
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_surface.c
@@ -0,0 +1,72 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "nv40_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+
+static void
+nv40_surface_copy(struct pipe_context *pipe, boolean do_flip,
+		  struct pipe_surface *dest, unsigned destx, unsigned desty,
+		  struct pipe_surface *src, unsigned srcx, unsigned srcy,
+		  unsigned width, unsigned height)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv04_surface_2d *eng2d = nv40->screen->eng2d;
+
+	if (do_flip) {
+		desty += height;
+		while (height--) {
+			eng2d->copy(eng2d, dest, destx, desty--, src,
+				    srcx, srcy++, width, 1);
+		}
+		return;
+	}
+
+	eng2d->copy(eng2d, dest, destx, desty, src, srcx, srcy, width, height);
+}
+
+static void
+nv40_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
+		  unsigned destx, unsigned desty, unsigned width,
+		  unsigned height, unsigned value)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nv04_surface_2d *eng2d = nv40->screen->eng2d;
+
+	eng2d->fill(eng2d, dest, destx, desty, width, height, value);
+}
+
+void
+nv40_init_surface_functions(struct nv40_context *nv40)
+{
+	nv40->pipe.surface_copy = nv40_surface_copy;
+	nv40->pipe.surface_fill = nv40_surface_fill;
+}
diff --git a/src/gallium/drivers/nv40/nv40_vbo.c b/src/gallium/drivers/nv40/nv40_vbo.c
new file mode 100644
index 0000000000..8f1834628f
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_vbo.c
@@ -0,0 +1,555 @@
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+
+#include "nv40_context.h"
+#include "nv40_state.h"
+
+#include "nouveau/nouveau_channel.h"
+#include "nouveau/nouveau_pushbuf.h"
+#include "nouveau/nouveau_util.h"
+
+#define FORCE_SWTNL 0
+
+static INLINE int
+nv40_vbo_format_to_hw(enum pipe_format pipe, unsigned *fmt, unsigned *ncomp)
+{
+	switch (pipe) {
+	case PIPE_FORMAT_R32_FLOAT:
+	case PIPE_FORMAT_R32G32_FLOAT:
+	case PIPE_FORMAT_R32G32B32_FLOAT:
+	case PIPE_FORMAT_R32G32B32A32_FLOAT:
+		*fmt = NV40TCL_VTXFMT_TYPE_FLOAT;
+		break;
+	case PIPE_FORMAT_R8_UNORM:
+	case PIPE_FORMAT_R8G8_UNORM:
+	case PIPE_FORMAT_R8G8B8_UNORM:
+	case PIPE_FORMAT_R8G8B8A8_UNORM:
+		*fmt = NV40TCL_VTXFMT_TYPE_UBYTE;
+		break;
+	case PIPE_FORMAT_R16_SSCALED:
+	case PIPE_FORMAT_R16G16_SSCALED:
+	case PIPE_FORMAT_R16G16B16_SSCALED:
+	case PIPE_FORMAT_R16G16B16A16_SSCALED:
+		*fmt = NV40TCL_VTXFMT_TYPE_USHORT;
+		break;
+	default:
+		NOUVEAU_ERR("Unknown format %s\n", pf_name(pipe));
+		return 1;
+	}
+
+	switch (pipe) {
+	case PIPE_FORMAT_R8_UNORM:
+	case PIPE_FORMAT_R32_FLOAT:
+	case PIPE_FORMAT_R16_SSCALED:
+		*ncomp = 1;
+		break;
+	case PIPE_FORMAT_R8G8_UNORM:
+	case PIPE_FORMAT_R32G32_FLOAT:
+	case PIPE_FORMAT_R16G16_SSCALED:
+		*ncomp = 2;
+		break;
+	case PIPE_FORMAT_R8G8B8_UNORM:
+	case PIPE_FORMAT_R32G32B32_FLOAT:
+	case PIPE_FORMAT_R16G16B16_SSCALED:
+		*ncomp = 3;
+		break;
+	case PIPE_FORMAT_R8G8B8A8_UNORM:
+	case PIPE_FORMAT_R32G32B32A32_FLOAT:
+	case PIPE_FORMAT_R16G16B16A16_SSCALED:
+		*ncomp = 4;
+		break;
+	default:
+		NOUVEAU_ERR("Unknown format %s\n", pf_name(pipe));
+		return 1;
+	}
+
+	return 0;
+}
+
+static boolean
+nv40_vbo_set_idxbuf(struct nv40_context *nv40, struct pipe_buffer *ib,
+		    unsigned ib_size)
+{
+	struct pipe_screen *pscreen = &nv40->screen->pipe;
+	unsigned type;
+
+	if (!ib) {
+		nv40->idxbuf = NULL;
+		nv40->idxbuf_format = 0xdeadbeef;
+		return FALSE;
+	}
+
+	if (!pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF) || ib_size == 1)
+		return FALSE;
+
+	switch (ib_size) {
+	case 2:
+		type = NV40TCL_IDXBUF_FORMAT_TYPE_U16;
+		break;
+	case 4:
+		type = NV40TCL_IDXBUF_FORMAT_TYPE_U32;
+		break;
+	default:
+		return FALSE;
+	}
+
+	if (ib != nv40->idxbuf ||
+	    type != nv40->idxbuf_format) {
+		nv40->dirty |= NV40_NEW_ARRAYS;
+		nv40->idxbuf = ib;
+		nv40->idxbuf_format = type;
+	}
+
+	return TRUE;
+}
+
+static boolean
+nv40_vbo_static_attrib(struct nv40_context *nv40, struct nouveau_stateobj *so,
+		       int attrib, struct pipe_vertex_element *ve,
+		       struct pipe_vertex_buffer *vb)
+{
+	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct nouveau_grobj *curie = nv40->screen->curie;
+	unsigned type, ncomp;
+	void *map;
+
+	if (nv40_vbo_format_to_hw(ve->src_format, &type, &ncomp))
+		return FALSE;
+
+	map  = ws->buffer_map(ws, vb->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+	map += vb->buffer_offset + ve->src_offset;
+
+	switch (type) {
+	case NV40TCL_VTXFMT_TYPE_FLOAT:
+	{
+		float *v = map;
+
+		switch (ncomp) {
+		case 4:
+			so_method(so, curie, NV40TCL_VTX_ATTR_4F_X(attrib), 4);
+			so_data  (so, fui(v[0]));
+			so_data  (so, fui(v[1]));
+			so_data  (so, fui(v[2]));
+			so_data  (so, fui(v[3]));
+			break;
+		case 3:
+			so_method(so, curie, NV40TCL_VTX_ATTR_3F_X(attrib), 3);
+			so_data  (so, fui(v[0]));
+			so_data  (so, fui(v[1]));
+			so_data  (so, fui(v[2]));
+			break;
+		case 2:
+			so_method(so, curie, NV40TCL_VTX_ATTR_2F_X(attrib), 2);
+			so_data  (so, fui(v[0]));
+			so_data  (so, fui(v[1]));
+			break;
+		case 1:
+			so_method(so, curie, NV40TCL_VTX_ATTR_1F(attrib), 1);
+			so_data  (so, fui(v[0]));
+			break;
+		default:
+			ws->buffer_unmap(ws, vb->buffer);
+			return FALSE;
+		}
+	}
+		break;
+	default:
+		ws->buffer_unmap(ws, vb->buffer);
+		return FALSE;
+	}
+
+	ws->buffer_unmap(ws, vb->buffer);
+
+	return TRUE;
+}
+
+boolean
+nv40_draw_arrays(struct pipe_context *pipe,
+		 unsigned mode, unsigned start, unsigned count)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nouveau_channel *chan = nv40->nvws->channel;
+	unsigned restart;
+
+	nv40_vbo_set_idxbuf(nv40, NULL, 0);
+	if (FORCE_SWTNL || !nv40_state_validate(nv40)) {
+		return nv40_draw_elements_swtnl(pipe, NULL, 0,
+						mode, start, count);
+	}
+
+	while (count) {
+		unsigned vc, nr;
+
+		nv40_state_emit(nv40);
+
+		vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 256,
+					mode, start, count, &restart);
+		if (!vc) {
+			FIRE_RING(NULL);
+			continue;
+		}
+
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (nvgl_primitive(mode));
+
+		nr = (vc & 0xff);
+		if (nr) {
+			BEGIN_RING(curie, NV40TCL_VB_VERTEX_BATCH, 1);
+			OUT_RING  (((nr - 1) << 24) | start);
+			start += nr;
+		}
+
+		nr = vc >> 8;
+		while (nr) {
+			unsigned push = nr > 2047 ? 2047 : nr;
+
+			nr -= push;
+
+			BEGIN_RING_NI(curie, NV40TCL_VB_VERTEX_BATCH, push);
+			while (push--) {
+				OUT_RING(((0x100 - 1) << 24) | start);
+				start += 0x100;
+			}
+		}
+
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (0);
+
+		count -= vc;
+		start = restart;
+	}
+
+	pipe->flush(pipe, 0, NULL);
+	return TRUE;
+}
+
+static INLINE void
+nv40_draw_elements_u08(struct nv40_context *nv40, void *ib,
+		       unsigned mode, unsigned start, unsigned count)
+{
+	struct nouveau_channel *chan = nv40->nvws->channel;
+
+	while (count) {
+		uint8_t *elts = (uint8_t *)ib + start;
+		unsigned vc, push, restart;
+
+		nv40_state_emit(nv40);
+
+		vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 2,
+					mode, start, count, &restart);
+		if (vc == 0) {
+			FIRE_RING(NULL);
+			continue;
+		}
+		count -= vc;
+
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (nvgl_primitive(mode));
+
+		if (vc & 1) {
+			BEGIN_RING(curie, NV40TCL_VB_ELEMENT_U32, 1);
+			OUT_RING  (elts[0]);
+			elts++; vc--;
+		}
+
+		while (vc) {
+			unsigned i;
+
+			push = MIN2(vc, 2047 * 2);
+
+			BEGIN_RING_NI(curie, NV40TCL_VB_ELEMENT_U16, push >> 1);
+			for (i = 0; i < push; i+=2)
+				OUT_RING((elts[i+1] << 16) | elts[i]);
+
+			vc -= push;
+			elts += push;
+		}
+
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (0);
+
+		start = restart;
+	}
+}
+
+static INLINE void
+nv40_draw_elements_u16(struct nv40_context *nv40, void *ib,
+		       unsigned mode, unsigned start, unsigned count)
+{
+	struct nouveau_channel *chan = nv40->nvws->channel;
+
+	while (count) {
+		uint16_t *elts = (uint16_t *)ib + start;
+		unsigned vc, push, restart;
+
+		nv40_state_emit(nv40);
+
+		vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 2,
+					mode, start, count, &restart);
+		if (vc == 0) {
+			FIRE_RING(NULL);
+			continue;
+		}
+		count -= vc;
+
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (nvgl_primitive(mode));
+
+		if (vc & 1) {
+			BEGIN_RING(curie, NV40TCL_VB_ELEMENT_U32, 1);
+			OUT_RING  (elts[0]);
+			elts++; vc--;
+		}
+
+		while (vc) {
+			unsigned i;
+
+			push = MIN2(vc, 2047 * 2);
+
+			BEGIN_RING_NI(curie, NV40TCL_VB_ELEMENT_U16, push >> 1);
+			for (i = 0; i < push; i+=2)
+				OUT_RING((elts[i+1] << 16) | elts[i]);
+
+			vc -= push;
+			elts += push;
+		}
+
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (0);
+
+		start = restart;
+	}
+}
+
+static INLINE void
+nv40_draw_elements_u32(struct nv40_context *nv40, void *ib,
+		       unsigned mode, unsigned start, unsigned count)
+{
+	struct nouveau_channel *chan = nv40->nvws->channel;
+
+	while (count) {
+		uint32_t *elts = (uint32_t *)ib + start;
+		unsigned vc, push, restart;
+
+		nv40_state_emit(nv40);
+
+		vc = nouveau_vbuf_split(chan->pushbuf->remaining, 5, 1,
+					mode, start, count, &restart);
+		if (vc == 0) {
+			FIRE_RING(NULL);
+			continue;
+		}
+		count -= vc;
+
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (nvgl_primitive(mode));
+
+		while (vc) {
+			push = MIN2(vc, 2047);
+
+			BEGIN_RING_NI(curie, NV40TCL_VB_ELEMENT_U32, push);
+			OUT_RINGp    (elts, push);
+
+			vc -= push;
+			elts += push;
+		}
+
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (0);
+
+		start = restart;
+	}
+}
+
+static boolean
+nv40_draw_elements_inline(struct pipe_context *pipe,
+			  struct pipe_buffer *ib, unsigned ib_size,
+			  unsigned mode, unsigned start, unsigned count)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct pipe_winsys *ws = pipe->winsys;
+	void *map;
+
+	map = ws->buffer_map(ws, ib, PIPE_BUFFER_USAGE_CPU_READ);
+	if (!ib) {
+		NOUVEAU_ERR("failed mapping ib\n");
+		return FALSE;
+	}
+
+	switch (ib_size) {
+	case 1:
+		nv40_draw_elements_u08(nv40, map, mode, start, count);
+		break;
+	case 2:
+		nv40_draw_elements_u16(nv40, map, mode, start, count);
+		break;
+	case 4:
+		nv40_draw_elements_u32(nv40, map, mode, start, count);
+		break;
+	default:
+		NOUVEAU_ERR("invalid idxbuf fmt %d\n", ib_size);
+		break;
+	}
+
+	ws->buffer_unmap(ws, ib);
+	return TRUE;
+}
+
+static boolean
+nv40_draw_elements_vbo(struct pipe_context *pipe,
+		       unsigned mode, unsigned start, unsigned count)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	struct nouveau_channel *chan = nv40->nvws->channel;
+	unsigned restart;
+
+	while (count) {
+		unsigned nr, vc;
+
+		nv40_state_emit(nv40);
+
+		vc = nouveau_vbuf_split(chan->pushbuf->remaining, 6, 256,
+					mode, start, count, &restart);
+		if (!vc) {
+			FIRE_RING(NULL);
+			continue;
+		}
+		
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (nvgl_primitive(mode));
+
+		nr = (vc & 0xff);
+		if (nr) {
+			BEGIN_RING(curie, NV40TCL_VB_INDEX_BATCH, 1);
+			OUT_RING  (((nr - 1) << 24) | start);
+			start += nr;
+		}
+
+		nr = vc >> 8;
+		while (nr) {
+			unsigned push = nr > 2047 ? 2047 : nr;
+
+			nr -= push;
+
+			BEGIN_RING_NI(curie, NV40TCL_VB_INDEX_BATCH, push);
+			while (push--) {
+				OUT_RING(((0x100 - 1) << 24) | start);
+				start += 0x100;
+			}
+		}
+
+		BEGIN_RING(curie, NV40TCL_BEGIN_END, 1);
+		OUT_RING  (0);
+
+		count -= vc;
+		start = restart;
+	}
+
+	return TRUE;
+}
+
+boolean
+nv40_draw_elements(struct pipe_context *pipe,
+		   struct pipe_buffer *indexBuffer, unsigned indexSize,
+		   unsigned mode, unsigned start, unsigned count)
+{
+	struct nv40_context *nv40 = nv40_context(pipe);
+	boolean idxbuf;
+
+	idxbuf = nv40_vbo_set_idxbuf(nv40, indexBuffer, indexSize);
+	if (FORCE_SWTNL || !nv40_state_validate(nv40)) {
+		return nv40_draw_elements_swtnl(pipe, NULL, 0,
+						mode, start, count);
+	}
+
+	if (idxbuf) {
+		nv40_draw_elements_vbo(pipe, mode, start, count);
+	} else {
+		nv40_draw_elements_inline(pipe, indexBuffer, indexSize,
+					  mode, start, count);
+	}
+
+	pipe->flush(pipe, 0, NULL);
+	return TRUE;
+}
+
+static boolean
+nv40_vbo_validate(struct nv40_context *nv40)
+{
+	struct nouveau_stateobj *vtxbuf, *vtxfmt, *sattr = NULL;
+	struct nouveau_grobj *curie = nv40->screen->curie;
+	struct pipe_buffer *ib = nv40->idxbuf;
+	unsigned ib_format = nv40->idxbuf_format;
+	unsigned vb_flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD;
+	int hw;
+
+	if (nv40->edgeflags) {
+		nv40->fallback_swtnl |= NV40_NEW_ARRAYS;
+		return FALSE;
+	}
+
+	vtxbuf = so_new(20, 18);
+	so_method(vtxbuf, curie, NV40TCL_VTXBUF_ADDRESS(0), nv40->vtxelt_nr);
+	vtxfmt = so_new(17, 0);
+	so_method(vtxfmt, curie, NV40TCL_VTXFMT(0), nv40->vtxelt_nr);
+
+	for (hw = 0; hw < nv40->vtxelt_nr; hw++) {
+		struct pipe_vertex_element *ve;
+		struct pipe_vertex_buffer *vb;
+		unsigned type, ncomp;
+
+		ve = &nv40->vtxelt[hw];
+		vb = &nv40->vtxbuf[ve->vertex_buffer_index];
+
+		if (!vb->stride) {
+			if (!sattr)
+				sattr = so_new(16 * 5, 0);
+
+			if (nv40_vbo_static_attrib(nv40, sattr, hw, ve, vb)) {
+				so_data(vtxbuf, 0);
+				so_data(vtxfmt, NV40TCL_VTXFMT_TYPE_FLOAT);
+				continue;
+			}
+		}
+
+		if (nv40_vbo_format_to_hw(ve->src_format, &type, &ncomp)) {
+			nv40->fallback_swtnl |= NV40_NEW_ARRAYS;
+			so_ref(NULL, &vtxbuf);
+			so_ref(NULL, &vtxfmt);
+			return FALSE;
+		}
+
+		so_reloc(vtxbuf, vb->buffer, vb->buffer_offset + ve->src_offset,
+			 vb_flags | NOUVEAU_BO_LOW | NOUVEAU_BO_OR,
+			 0, NV40TCL_VTXBUF_ADDRESS_DMA1);
+		so_data (vtxfmt, ((vb->stride << NV40TCL_VTXFMT_STRIDE_SHIFT) |
+				  (ncomp << NV40TCL_VTXFMT_SIZE_SHIFT) | type));
+	}
+
+	if (ib) {
+		so_method(vtxbuf, curie, NV40TCL_IDXBUF_ADDRESS, 2);
+		so_reloc (vtxbuf, ib, 0, vb_flags | NOUVEAU_BO_LOW, 0, 0);
+		so_reloc (vtxbuf, ib, ib_format, vb_flags | NOUVEAU_BO_OR,
+			  0, NV40TCL_IDXBUF_FORMAT_DMA1);
+	}
+
+	so_method(vtxbuf, curie, 0x1710, 1);
+	so_data  (vtxbuf, 0);
+
+	so_ref(vtxbuf, &nv40->state.hw[NV40_STATE_VTXBUF]);
+	nv40->state.dirty |= (1ULL << NV40_STATE_VTXBUF);
+	so_ref(vtxfmt, &nv40->state.hw[NV40_STATE_VTXFMT]);
+	nv40->state.dirty |= (1ULL << NV40_STATE_VTXFMT);
+	so_ref(sattr, &nv40->state.hw[NV40_STATE_VTXATTR]);
+	nv40->state.dirty |= (1ULL << NV40_STATE_VTXATTR);
+	return FALSE;
+}
+
+struct nv40_state_entry nv40_state_vbo = {
+	.validate = nv40_vbo_validate,
+	.dirty = {
+		.pipe = NV40_NEW_ARRAYS,
+		.hw = 0,
+	}
+};
+
diff --git a/src/gallium/drivers/nv40/nv40_vertprog.c b/src/gallium/drivers/nv40/nv40_vertprog.c
new file mode 100644
index 0000000000..0862386638
--- /dev/null
+++ b/src/gallium/drivers/nv40/nv40_vertprog.c
@@ -0,0 +1,1070 @@
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+
+#include "nv40_context.h"
+#include "nv40_state.h"
+
+/* TODO (at least...):
+ *  1. Indexed consts  + ARL
+ *  3. NV_vp11, NV_vp2, NV_vp3 features
+ *       - extra arith opcodes
+ *       - branching
+ *       - texture sampling
+ *       - indexed attribs
+ *       - indexed results
+ *  4. bugs
+ */
+
+#define SWZ_X 0
+#define SWZ_Y 1
+#define SWZ_Z 2
+#define SWZ_W 3
+#define MASK_X 8
+#define MASK_Y 4
+#define MASK_Z 2
+#define MASK_W 1
+#define MASK_ALL (MASK_X|MASK_Y|MASK_Z|MASK_W)
+#define DEF_SCALE 0
+#define DEF_CTEST 0
+#include "nv40_shader.h"
+
+#define swz(s,x,y,z,w) nv40_sr_swz((s), SWZ_##x, SWZ_##y, SWZ_##z, SWZ_##w)
+#define neg(s) nv40_sr_neg((s))
+#define abs(s) nv40_sr_abs((s))
+
+#define NV40_VP_INST_DEST_CLIP(n) ((~0 - 6) + (n))
+
+struct nv40_vpc {
+	struct nv40_vertex_program *vp;
+
+	struct nv40_vertex_program_exec *vpi;
+
+	unsigned r_temps;
+	unsigned r_temps_discard;
+	struct nv40_sreg r_result[PIPE_MAX_SHADER_OUTPUTS];
+	struct nv40_sreg *r_address;
+	struct nv40_sreg *r_temp;
+
+	struct nv40_sreg *imm;
+	unsigned nr_imm;
+
+	unsigned hpos_idx;
+};
+
+static struct nv40_sreg
+temp(struct nv40_vpc *vpc)
+{
+	int idx = ffs(~vpc->r_temps) - 1;
+
+	if (idx < 0) {
+		NOUVEAU_ERR("out of temps!!\n");
+		assert(0);
+		return nv40_sr(NV40SR_TEMP, 0);
+	}
+
+	vpc->r_temps |= (1 << idx);
+	vpc->r_temps_discard |= (1 << idx);
+	return nv40_sr(NV40SR_TEMP, idx);
+}
+
+static INLINE void
+release_temps(struct nv40_vpc *vpc)
+{
+	vpc->r_temps &= ~vpc->r_temps_discard;
+	vpc->r_temps_discard = 0;
+}
+
+static struct nv40_sreg
+constant(struct nv40_vpc *vpc, int pipe, float x, float y, float z, float w)
+{
+	struct nv40_vertex_program *vp = vpc->vp;
+	struct nv40_vertex_program_data *vpd;
+	int idx;
+
+	if (pipe >= 0) {
+		for (idx = 0; idx < vp->nr_consts; idx++) {
+			if (vp->consts[idx].index == pipe)
+				return nv40_sr(NV40SR_CONST, idx);
+		}
+	}
+
+	idx = vp->nr_consts++;
+	vp->consts = realloc(vp->consts, sizeof(*vpd) * vp->nr_consts);
+	vpd = &vp->consts[idx];
+
+	vpd->index = pipe;
+	vpd->value[0] = x;
+	vpd->value[1] = y;
+	vpd->value[2] = z;
+	vpd->value[3] = w;
+	return nv40_sr(NV40SR_CONST, idx);
+}
+
+#define arith(cc,s,o,d,m,s0,s1,s2) \
+	nv40_vp_arith((cc), (s), NV40_VP_INST_##o, (d), (m), (s0), (s1), (s2))
+
+static void
+emit_src(struct nv40_vpc *vpc, uint32_t *hw, int pos, struct nv40_sreg src)
+{
+	struct nv40_vertex_program *vp = vpc->vp;
+	uint32_t sr = 0;
+
+	switch (src.type) {
+	case NV40SR_TEMP:
+		sr |= (NV40_VP_SRC_REG_TYPE_TEMP << NV40_VP_SRC_REG_TYPE_SHIFT);
+		sr |= (src.index << NV40_VP_SRC_TEMP_SRC_SHIFT);
+		break;
+	case NV40SR_INPUT:
+		sr |= (NV40_VP_SRC_REG_TYPE_INPUT <<
+		       NV40_VP_SRC_REG_TYPE_SHIFT);
+		vp->ir |= (1 << src.index);
+		hw[1] |= (src.index << NV40_VP_INST_INPUT_SRC_SHIFT);
+		break;
+	case NV40SR_CONST:
+		sr |= (NV40_VP_SRC_REG_TYPE_CONST <<
+		       NV40_VP_SRC_REG_TYPE_SHIFT);
+		assert(vpc->vpi->const_index == -1 ||
+		       vpc->vpi->const_index == src.index);
+		vpc->vpi->const_index = src.index;
+		break;
+	case NV40SR_NONE:
+		sr |= (NV40_VP_SRC_REG_TYPE_INPUT <<
+		       NV40_VP_SRC_REG_TYPE_SHIFT);
+		break;
+	default:
+		assert(0);
+	}
+
+	if (src.negate)
+		sr |= NV40_VP_SRC_NEGATE;
+
+	if (src.abs)
+		hw[0] |= (1 << (21 + pos));
+
+	sr |= ((src.swz[0] << NV40_VP_SRC_SWZ_X_SHIFT) |
+	       (src.swz[1] << NV40_VP_SRC_SWZ_Y_SHIFT) |
+	       (src.swz[2] << NV40_VP_SRC_SWZ_Z_SHIFT) |
+	       (src.swz[3] << NV40_VP_SRC_SWZ_W_SHIFT));
+
+	switch (pos) {
+	case 0:
+		hw[1] |= ((sr & NV40_VP_SRC0_HIGH_MASK) >>
+			  NV40_VP_SRC0_HIGH_SHIFT) << NV40_VP_INST_SRC0H_SHIFT;
+		hw[2] |= (sr & NV40_VP_SRC0_LOW_MASK) <<
+			  NV40_VP_INST_SRC0L_SHIFT;
+		break;
+	case 1:
+		hw[2] |= sr << NV40_VP_INST_SRC1_SHIFT;
+		break;
+	case 2:
+		hw[2] |= ((sr & NV40_VP_SRC2_HIGH_MASK) >>
+			  NV40_VP_SRC2_HIGH_SHIFT) << NV40_VP_INST_SRC2H_SHIFT;
+		hw[3] |= (sr & NV40_VP_SRC2_LOW_MASK) <<
+			  NV40_VP_INST_SRC2L_SHIFT;
+		break;
+	default:
+		assert(0);
+	}
+}
+
+static void
+emit_dst(struct nv40_vpc *vpc, uint32_t *hw, int slot, struct nv40_sreg dst)
+{
+	struct nv40_vertex_program *vp = vpc->vp;
+
+	switch (dst.type) {
+	case NV40SR_TEMP:
+		hw[3] |= NV40_VP_INST_DEST_MASK;
+		if (slot == 0) {
+			hw[0] |= (dst.index <<
+				  NV40_VP_INST_VEC_DEST_TEMP_SHIFT);
+		} else {
+			hw[3] |= (dst.index << 
+				  NV40_VP_INST_SCA_DEST_TEMP_SHIFT);
+		}
+		break;
+	case NV40SR_OUTPUT:
+		switch (dst.index) {
+		case NV40_VP_INST_DEST_COL0 : vp->or |= (1 << 0); break;
+		case NV40_VP_INST_DEST_COL1 : vp->or |= (1 << 1); break;
+		case NV40_VP_INST_DEST_BFC0 : vp->or |= (1 << 2); break;
+		case NV40_VP_INST_DEST_BFC1 : vp->or |= (1 << 3); break;
+		case NV40_VP_INST_DEST_FOGC : vp->or |= (1 << 4); break;
+		case NV40_VP_INST_DEST_PSZ  : vp->or |= (1 << 5); break;
+		case NV40_VP_INST_DEST_TC(0): vp->or |= (1 << 14); break;
+		case NV40_VP_INST_DEST_TC(1): vp->or |= (1 << 15); break;
+		case NV40_VP_INST_DEST_TC(2): vp->or |= (1 << 16); break;
+		case NV40_VP_INST_DEST_TC(3): vp->or |= (1 << 17); break;
+		case NV40_VP_INST_DEST_TC(4): vp->or |= (1 << 18); break;
+		case NV40_VP_INST_DEST_TC(5): vp->or |= (1 << 19); break;
+		case NV40_VP_INST_DEST_TC(6): vp->or |= (1 << 20); break;
+		case NV40_VP_INST_DEST_TC(7): vp->or |= (1 << 21); break;
+		case NV40_VP_INST_DEST_CLIP(0):
+			vp->or |= (1 << 6);
+			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE0;
+			dst.index = NV40_VP_INST_DEST_FOGC;
+			break;
+		case NV40_VP_INST_DEST_CLIP(1):
+			vp->or |= (1 << 7);
+			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE1;
+			dst.index = NV40_VP_INST_DEST_FOGC;
+			break;
+		case NV40_VP_INST_DEST_CLIP(2):
+			vp->or |= (1 << 8);
+			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE2;
+			dst.index = NV40_VP_INST_DEST_FOGC;
+			break;
+		case NV40_VP_INST_DEST_CLIP(3):
+			vp->or |= (1 << 9);
+			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE3;
+			dst.index = NV40_VP_INST_DEST_PSZ;
+			break;
+		case NV40_VP_INST_DEST_CLIP(4):
+			vp->or |= (1 << 10);
+			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE4;
+			dst.index = NV40_VP_INST_DEST_PSZ;
+			break;
+		case NV40_VP_INST_DEST_CLIP(5):
+			vp->or |= (1 << 11);
+			vp->clip_ctrl |= NV40TCL_CLIP_PLANE_ENABLE_PLANE5;
+			dst.index = NV40_VP_INST_DEST_PSZ;
+			break;
+		default:
+			break;
+		}
+
+		hw[3] |= (dst.index << NV40_VP_INST_DEST_SHIFT);
+		if (slot == 0) {
+			hw[0] |= NV40_VP_INST_VEC_RESULT;
+			hw[0] |= NV40_VP_INST_VEC_DEST_TEMP_MASK | (1<<20);
+		} else {
+			hw[3] |= NV40_VP_INST_SCA_RESULT;
+			hw[3] |= NV40_VP_INST_SCA_DEST_TEMP_MASK;
+		}
+		break;
+	default:
+		assert(0);
+	}
+}
+
+static void
+nv40_vp_arith(struct nv40_vpc *vpc, int slot, int op,
+	      struct nv40_sreg dst, int mask,
+	      struct nv40_sreg s0, struct nv40_sreg s1,
+	      struct nv40_sreg s2)
+{
+	struct nv40_vertex_program *vp = vpc->vp;
+	uint32_t *hw;
+
+	vp->insns = realloc(vp->insns, ++vp->nr_insns * sizeof(*vpc->vpi));
+	vpc->vpi = &vp->insns[vp->nr_insns - 1];
+	memset(vpc->vpi, 0, sizeof(*vpc->vpi));
+	vpc->vpi->const_index = -1;
+
+	hw = vpc->vpi->data;
+
+	hw[0] |= (NV40_VP_INST_COND_TR << NV40_VP_INST_COND_SHIFT);
+	hw[0] |= ((0 << NV40_VP_INST_COND_SWZ_X_SHIFT) |
+		  (1 << NV40_VP_INST_COND_SWZ_Y_SHIFT) |
+		  (2 << NV40_VP_INST_COND_SWZ_Z_SHIFT) |
+		  (3 << NV40_VP_INST_COND_SWZ_W_SHIFT));
+
+	if (slot == 0) {
+		hw[1] |= (op << NV40_VP_INST_VEC_OPCODE_SHIFT);
+		hw[3] |= NV40_VP_INST_SCA_DEST_TEMP_MASK;
+		hw[3] |= (mask << NV40_VP_INST_VEC_WRITEMASK_SHIFT);
+	} else {
+		hw[1] |= (op << NV40_VP_INST_SCA_OPCODE_SHIFT);
+		hw[0] |= (NV40_VP_INST_VEC_DEST_TEMP_MASK | (1 << 20));
+		hw[3] |= (mask << NV40_VP_INST_SCA_WRITEMASK_SHIFT);
+	}
+
+	emit_dst(vpc, hw, slot, dst);
+	emit_src(vpc, hw, 0, s0);
+	emit_src(vpc, hw, 1, s1);
+	emit_src(vpc, hw, 2, s2);
+}
+
+static INLINE struct nv40_sreg
+tgsi_src(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc) {
+	struct nv40_sreg src;
+
+	switch (fsrc->SrcRegister.File) {
+	case TGSI_FILE_INPUT:
+		src = nv40_sr(NV40SR_INPUT, fsrc->SrcRegister.Index);
+		break;
+	case TGSI_FILE_CONSTANT:
+		src = constant(vpc, fsrc->SrcRegister.Index, 0, 0, 0, 0);
+		break;
+	case TGSI_FILE_IMMEDIATE:
+		src = vpc->imm[fsrc->SrcRegister.Index];
+		break;
+	case TGSI_FILE_TEMPORARY:
+		src = vpc->r_temp[fsrc->SrcRegister.Index];
+		break;
+	default:
+		NOUVEAU_ERR("bad src file\n");
+		break;
+	}
+
+	src.abs = fsrc->SrcRegisterExtMod.Absolute;
+	src.negate = fsrc->SrcRegister.Negate;
+	src.swz[0] = fsrc->SrcRegister.SwizzleX;
+	src.swz[1] = fsrc->SrcRegister.SwizzleY;
+	src.swz[2] = fsrc->SrcRegister.SwizzleZ;
+	src.swz[3] = fsrc->SrcRegister.SwizzleW;
+	return src;
+}
+
+static INLINE struct nv40_sreg
+tgsi_dst(struct nv40_vpc *vpc, const struct tgsi_full_dst_register *fdst) {
+	struct nv40_sreg dst;
+
+	switch (fdst->DstRegister.File) {
+	case TGSI_FILE_OUTPUT:
+		dst = vpc->r_result[fdst->DstRegister.Index];
+		break;
+	case TGSI_FILE_TEMPORARY:
+		dst = vpc->r_temp[fdst->DstRegister.Index];
+		break;
+	case TGSI_FILE_ADDRESS:
+		dst = vpc->r_address[fdst->DstRegister.Index];
+		break;
+	default:
+		NOUVEAU_ERR("bad dst file\n");
+		break;
+	}
+
+	return dst;
+}
+
+static INLINE int
+tgsi_mask(uint tgsi)
+{
+	int mask = 0;
+
+	if (tgsi & TGSI_WRITEMASK_X) mask |= MASK_X;
+	if (tgsi & TGSI_WRITEMASK_Y) mask |= MASK_Y;
+	if (tgsi & TGSI_WRITEMASK_Z) mask |= MASK_Z;
+	if (tgsi & TGSI_WRITEMASK_W) mask |= MASK_W;
+	return mask;
+}
+
+static boolean
+src_native_swz(struct nv40_vpc *vpc, const struct tgsi_full_src_register *fsrc,
+	       struct nv40_sreg *src)
+{
+	const struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0);
+	struct nv40_sreg tgsi = tgsi_src(vpc, fsrc);
+	uint mask = 0, zero_mask = 0, one_mask = 0, neg_mask = 0;
+	uint neg[4] = { fsrc->SrcRegisterExtSwz.NegateX,
+			fsrc->SrcRegisterExtSwz.NegateY,
+			fsrc->SrcRegisterExtSwz.NegateZ,
+			fsrc->SrcRegisterExtSwz.NegateW };
+	uint c;
+
+	for (c = 0; c < 4; c++) {
+		switch (tgsi_util_get_full_src_register_extswizzle(fsrc, c)) {
+		case TGSI_EXTSWIZZLE_X:
+		case TGSI_EXTSWIZZLE_Y:
+		case TGSI_EXTSWIZZLE_Z:
+		case TGSI_EXTSWIZZLE_W:
+			mask |= tgsi_mask(1 << c);
+			break;
+		case TGSI_EXTSWIZZLE_ZERO:
+			zero_mask |= tgsi_mask(1 << c);
+			tgsi.swz[c] = SWZ_X;
+			break;
+		case TGSI_EXTSWIZZLE_ONE:
+			one_mask |= tgsi_mask(1 << c);
+			tgsi.swz[c] = SWZ_X;
+			break;
+		default:
+			assert(0);
+		}
+
+		if (!tgsi.negate && neg[c])
+			neg_mask |= tgsi_mask(1 << c);
+	}
+
+	if (mask == MASK_ALL && !neg_mask)
+		return TRUE;
+
+	*src = temp(vpc);
+
+	if (mask)
+		arith(vpc, 0, OP_MOV, *src, mask, tgsi, none, none);
+
+	if (zero_mask)
+		arith(vpc, 0, OP_SFL, *src, zero_mask, *src, none, none);
+
+	if (one_mask)
+		arith(vpc, 0, OP_STR, *src, one_mask, *src, none, none);
+
+	if (neg_mask) {
+		struct nv40_sreg one = temp(vpc);
+		arith(vpc, 0, OP_STR, one, neg_mask, one, none, none);
+		arith(vpc, 0, OP_MUL, *src, neg_mask, *src, neg(one), none);
+	}
+
+	return FALSE;
+}
+
+static boolean
+nv40_vertprog_parse_instruction(struct nv40_vpc *vpc,
+				const struct tgsi_full_instruction *finst)
+{
+	struct nv40_sreg src[3], dst, tmp;
+	struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0);
+	int mask;
+	int ai = -1, ci = -1, ii = -1;
+	int i;
+
+	if (finst->Instruction.Opcode == TGSI_OPCODE_END)
+		return TRUE;
+
+	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fsrc;
+
+		fsrc = &finst->FullSrcRegisters[i];
+		if (fsrc->SrcRegister.File == TGSI_FILE_TEMPORARY) {
+			src[i] = tgsi_src(vpc, fsrc);
+		}
+	}
+
+	for (i = 0; i < finst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fsrc;
+
+		fsrc = &finst->FullSrcRegisters[i];
+
+		switch (fsrc->SrcRegister.File) {
+		case TGSI_FILE_INPUT:
+		case TGSI_FILE_CONSTANT:
+		case TGSI_FILE_TEMPORARY:
+			if (!src_native_swz(vpc, fsrc, &src[i]))
+				continue;
+			break;
+		default:
+			break;
+		}
+
+		switch (fsrc->SrcRegister.File) {
+		case TGSI_FILE_INPUT:
+			if (ai == -1 || ai == fsrc->SrcRegister.Index) {
+				ai = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(vpc, fsrc);
+			} else {
+				src[i] = temp(vpc);
+				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
+				      tgsi_src(vpc, fsrc), none, none);
+			}
+			break;
+		case TGSI_FILE_CONSTANT:
+			if ((ci == -1 && ii == -1) ||
+			    ci == fsrc->SrcRegister.Index) {
+				ci = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(vpc, fsrc);
+			} else {
+				src[i] = temp(vpc);
+				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
+				      tgsi_src(vpc, fsrc), none, none);
+			}
+			break;
+		case TGSI_FILE_IMMEDIATE:
+			if ((ci == -1 && ii == -1) ||
+			    ii == fsrc->SrcRegister.Index) {
+				ii = fsrc->SrcRegister.Index;
+				src[i] = tgsi_src(vpc, fsrc);
+			} else {
+				src[i] = temp(vpc);
+				arith(vpc, 0, OP_MOV, src[i], MASK_ALL,
+				      tgsi_src(vpc, fsrc), none, none);
+			}
+			break;
+		case TGSI_FILE_TEMPORARY:
+			/* handled above */
+			break;
+		default:
+			NOUVEAU_ERR("bad src file\n");
+			return FALSE;
+		}
+	}
+
+	dst  = tgsi_dst(vpc, &finst->FullDstRegisters[0]);
+	mask = tgsi_mask(finst->FullDstRegisters[0].DstRegister.WriteMask);
+
+	switch (finst->Instruction.Opcode) {
+	case TGSI_OPCODE_ABS:
+		arith(vpc, 0, OP_MOV, dst, mask, abs(src[0]), none, none);
+		break;
+	case TGSI_OPCODE_ADD:
+		arith(vpc, 0, OP_ADD, dst, mask, src[0], none, src[1]);
+		break;
+	case TGSI_OPCODE_ARL:
+		arith(vpc, 0, OP_ARL, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_DP3:
+		arith(vpc, 0, OP_DP3, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DP4:
+		arith(vpc, 0, OP_DP4, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DPH:
+		arith(vpc, 0, OP_DPH, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_DST:
+		arith(vpc, 0, OP_DST, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_EX2:
+		arith(vpc, 1, OP_EX2, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_EXP:
+		arith(vpc, 1, OP_EXP, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_FLR:
+		arith(vpc, 0, OP_FLR, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_FRC:
+		arith(vpc, 0, OP_FRC, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_LG2:
+		arith(vpc, 1, OP_LG2, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_LIT:
+		arith(vpc, 1, OP_LIT, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_LOG:
+		arith(vpc, 1, OP_LOG, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_MAD:
+		arith(vpc, 0, OP_MAD, dst, mask, src[0], src[1], src[2]);
+		break;
+	case TGSI_OPCODE_MAX:
+		arith(vpc, 0, OP_MAX, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_MIN:
+		arith(vpc, 0, OP_MIN, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_MOV:
+		arith(vpc, 0, OP_MOV, dst, mask, src[0], none, none);
+		break;
+	case TGSI_OPCODE_MUL:
+		arith(vpc, 0, OP_MUL, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_POW:
+		tmp = temp(vpc);
+		arith(vpc, 1, OP_LG2, tmp, MASK_X, none, none,
+		      swz(src[0], X, X, X, X));
+		arith(vpc, 0, OP_MUL, tmp, MASK_X, swz(tmp, X, X, X, X),
+		      swz(src[1], X, X, X, X), none);
+		arith(vpc, 1, OP_EX2, dst, mask, none, none,
+		      swz(tmp, X, X, X, X));
+		break;
+	case TGSI_OPCODE_RCP:
+		arith(vpc, 1, OP_RCP, dst, mask, none, none, src[0]);
+		break;
+	case TGSI_OPCODE_RET:
+		break;
+	case TGSI_OPCODE_RSQ:
+		arith(vpc, 1, OP_RSQ, dst, mask, none, none, abs(src[0]));
+		break;
+	case TGSI_OPCODE_SGE:
+		arith(vpc, 0, OP_SGE, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SLT:
+		arith(vpc, 0, OP_SLT, dst, mask, src[0], src[1], none);
+		break;
+	case TGSI_OPCODE_SUB:
+		arith(vpc, 0, OP_ADD, dst, mask, src[0], none, neg(src[1]));
+		break;
+	case TGSI_OPCODE_XPD:
+		tmp = temp(vpc);
+		arith(vpc, 0, OP_MUL, tmp, mask,
+		      swz(src[0], Z, X, Y, Y), swz(src[1], Y, Z, X, X), none);
+		arith(vpc, 0, OP_MAD, dst, (mask & ~MASK_W),
+		      swz(src[0], Y, Z, X, X), swz(src[1], Z, X, Y, Y),
+		      neg(tmp));
+		break;
+	default:
+		NOUVEAU_ERR("invalid opcode %d\n", finst->Instruction.Opcode);
+		return FALSE;
+	}
+
+	release_temps(vpc);
+	return TRUE;
+}
+
+static boolean
+nv40_vertprog_parse_decl_output(struct nv40_vpc *vpc,
+				const struct tgsi_full_declaration *fdec)
+{
+	unsigned idx = fdec->DeclarationRange.First;
+	int hw;
+
+	switch (fdec->Semantic.SemanticName) {
+	case TGSI_SEMANTIC_POSITION:
+		hw = NV40_VP_INST_DEST_POS;
+		vpc->hpos_idx = idx;
+		break;
+	case TGSI_SEMANTIC_COLOR:
+		if (fdec->Semantic.SemanticIndex == 0) {
+			hw = NV40_VP_INST_DEST_COL0;
+		} else
+		if (fdec->Semantic.SemanticIndex == 1) {
+			hw = NV40_VP_INST_DEST_COL1;
+		} else {
+			NOUVEAU_ERR("bad colour semantic index\n");
+			return FALSE;
+		}
+		break;
+	case TGSI_SEMANTIC_BCOLOR:
+		if (fdec->Semantic.SemanticIndex == 0) {
+			hw = NV40_VP_INST_DEST_BFC0;
+		} else
+		if (fdec->Semantic.SemanticIndex == 1) {
+			hw = NV40_VP_INST_DEST_BFC1;
+		} else {
+			NOUVEAU_ERR("bad bcolour semantic index\n");
+			return FALSE;
+		}
+		break;
+	case TGSI_SEMANTIC_FOG:
+		hw = NV40_VP_INST_DEST_FOGC;
+		break;
+	case TGSI_SEMANTIC_PSIZE:
+		hw = NV40_VP_INST_DEST_PSZ;
+		break;
+	case TGSI_SEMANTIC_GENERIC:
+		if (fdec->Semantic.SemanticIndex <= 7) {
+			hw = NV40_VP_INST_DEST_TC(fdec->Semantic.SemanticIndex);
+		} else {
+			NOUVEAU_ERR("bad generic semantic index\n");
+			return FALSE;
+		}
+		break;
+	default:
+		NOUVEAU_ERR("bad output semantic\n");
+		return FALSE;
+	}
+
+	vpc->r_result[idx] = nv40_sr(NV40SR_OUTPUT, hw);
+	return TRUE;
+}
+
+static boolean
+nv40_vertprog_prepare(struct nv40_vpc *vpc)
+{
+	struct tgsi_parse_context p;
+	int high_temp = -1, high_addr = -1, nr_imm = 0, i;
+
+	tgsi_parse_init(&p, vpc->vp->pipe.tokens);
+	while (!tgsi_parse_end_of_tokens(&p)) {
+		const union tgsi_full_token *tok = &p.FullToken;
+
+		tgsi_parse_token(&p);
+		switch(tok->Token.Type) {
+		case TGSI_TOKEN_TYPE_IMMEDIATE:
+			nr_imm++;
+			break;
+		case TGSI_TOKEN_TYPE_DECLARATION:
+		{
+			const struct tgsi_full_declaration *fdec;
+
+			fdec = &p.FullToken.FullDeclaration;
+			switch (fdec->Declaration.File) {
+			case TGSI_FILE_TEMPORARY:
+				if (fdec->DeclarationRange.Last > high_temp) {
+					high_temp =
+						fdec->DeclarationRange.Last;
+				}
+				break;
+#if 0 /* this would be nice.. except gallium doesn't track it */
+			case TGSI_FILE_ADDRESS:
+				if (fdec->DeclarationRange.Last > high_addr) {
+					high_addr =
+						fdec->DeclarationRange.Last;
+				}
+				break;
+#endif
+			case TGSI_FILE_OUTPUT:
+				if (!nv40_vertprog_parse_decl_output(vpc, fdec))
+					return FALSE;
+				break;
+			default:
+				break;
+			}
+		}
+			break;
+#if 1 /* yay, parse instructions looking for address regs instead */
+		case TGSI_TOKEN_TYPE_INSTRUCTION:
+		{
+			const struct tgsi_full_instruction *finst;
+			const struct tgsi_full_dst_register *fdst;
+
+			finst = &p.FullToken.FullInstruction;
+			fdst = &finst->FullDstRegisters[0];
+
+			if (fdst->DstRegister.File == TGSI_FILE_ADDRESS) {
+				if (fdst->DstRegister.Index > high_addr)
+					high_addr = fdst->DstRegister.Index;
+			}
+		
+		}
+			break;
+#endif
+		default:
+			break;
+		}
+	}
+	tgsi_parse_free(&p);
+
+	if (nr_imm) {
+		vpc->imm = CALLOC(nr_imm, sizeof(struct nv40_sreg));
+		assert(vpc->imm);
+	}
+
+	if (++high_temp) {
+		vpc->r_temp = CALLOC(high_temp, sizeof(struct nv40_sreg));
+		for (i = 0; i < high_temp; i++)
+			vpc->r_temp[i] = temp(vpc);
+	}
+
+	if (++high_addr) {
+		vpc->r_address = CALLOC(high_addr, sizeof(struct nv40_sreg));
+		for (i = 0; i < high_addr; i++)
+			vpc->r_address[i] = temp(vpc);
+	}
+
+	vpc->r_temps_discard = 0;
+	return TRUE;
+}
+
+static void
+nv40_vertprog_translate(struct nv40_context *nv40,
+			struct nv40_vertex_program *vp)
+{
+	struct tgsi_parse_context parse;
+	struct nv40_vpc *vpc = NULL;
+	struct nv40_sreg none = nv40_sr(NV40SR_NONE, 0);
+	int i;
+
+	vpc = CALLOC(1, sizeof(struct nv40_vpc));
+	if (!vpc)
+		return;
+	vpc->vp = vp;
+
+	if (!nv40_vertprog_prepare(vpc)) {
+		FREE(vpc);
+		return;
+	}
+
+	/* Redirect post-transform vertex position to a temp if user clip
+	 * planes are enabled.  We need to append code the the vtxprog
+	 * to handle clip planes later.
+	 */
+	if (vp->ucp.nr)  {
+		vpc->r_result[vpc->hpos_idx] = temp(vpc);
+		vpc->r_temps_discard = 0;
+	}
+
+	tgsi_parse_init(&parse, vp->pipe.tokens);
+
+	while (!tgsi_parse_end_of_tokens(&parse)) {
+		tgsi_parse_token(&parse);
+
+		switch (parse.FullToken.Token.Type) {
+		case TGSI_TOKEN_TYPE_IMMEDIATE:
+		{
+			const struct tgsi_full_immediate *imm;
+
+			imm = &parse.FullToken.FullImmediate;
+			assert(imm->Immediate.DataType == TGSI_IMM_FLOAT32);
+			assert(imm->Immediate.NrTokens == 4 + 1);
+			vpc->imm[vpc->nr_imm++] =
+				constant(vpc, -1,
+					 imm->u.ImmediateFloat32[0].Float,
+					 imm->u.ImmediateFloat32[1].Float,
+					 imm->u.ImmediateFloat32[2].Float,
+					 imm->u.ImmediateFloat32[3].Float);
+		}
+			break;
+		case TGSI_TOKEN_TYPE_INSTRUCTION:
+		{
+			const struct tgsi_full_instruction *finst;
+			finst = &parse.FullToken.FullInstruction;
+			if (!nv40_vertprog_parse_instruction(vpc, finst))
+				goto out_err;
+		}
+			break;
+		default:
+			break;
+		}
+	}
+
+	/* Write out HPOS if it was redirected to a temp earlier */
+	if (vpc->r_result[vpc->hpos_idx].type != NV40SR_OUTPUT) {
+		struct nv40_sreg hpos = nv40_sr(NV40SR_OUTPUT,
+						NV40_VP_INST_DEST_POS);
+		struct nv40_sreg htmp = vpc->r_result[vpc->hpos_idx];
+
+		arith(vpc, 0, OP_MOV, hpos, MASK_ALL, htmp, none, none);
+	}
+
+	/* Insert code to handle user clip planes */
+	for (i = 0; i < vp->ucp.nr; i++) {
+		struct nv40_sreg cdst = nv40_sr(NV40SR_OUTPUT,
+						NV40_VP_INST_DEST_CLIP(i));
+		struct nv40_sreg ceqn = constant(vpc, -1,
+						 nv40->clip.ucp[i][0],
+						 nv40->clip.ucp[i][1],
+						 nv40->clip.ucp[i][2],
+						 nv40->clip.ucp[i][3]);
+		struct nv40_sreg htmp = vpc->r_result[vpc->hpos_idx];
+		unsigned mask;
+
+		switch (i) {
+		case 0: case 3: mask = MASK_Y; break;
+		case 1: case 4: mask = MASK_Z; break;
+		case 2: case 5: mask = MASK_W; break;
+		default:
+			NOUVEAU_ERR("invalid clip dist #%d\n", i);
+			goto out_err;
+		}
+
+		arith(vpc, 0, OP_DP4, cdst, mask, htmp, ceqn, none);
+	}
+
+	vp->insns[vp->nr_insns - 1].data[3] |= NV40_VP_INST_LAST;
+	vp->translated = TRUE;
+out_err:
+	tgsi_parse_free(&parse);
+	if (vpc->r_temp)
+		FREE(vpc->r_temp); 
+	if (vpc->r_address)
+		FREE(vpc->r_address); 
+	if (vpc->imm)	
+		FREE(vpc->imm); 
+	FREE(vpc);
+}
+
+static boolean
+nv40_vertprog_validate(struct nv40_context *nv40)
+{ 
+	struct nouveau_winsys *nvws = nv40->nvws;
+	struct pipe_winsys *ws = nv40->pipe.winsys;
+	struct nouveau_grobj *curie = nv40->screen->curie;
+	struct nv40_vertex_program *vp;
+	struct pipe_buffer *constbuf;
+	boolean upload_code = FALSE, upload_data = FALSE;
+	int i;
+
+	if (nv40->render_mode == HW) {
+		vp = nv40->vertprog;
+		constbuf = nv40->constbuf[PIPE_SHADER_VERTEX];
+
+		if ((nv40->dirty & NV40_NEW_UCP) ||
+		    memcmp(&nv40->clip, &vp->ucp, sizeof(vp->ucp))) {
+			nv40_vertprog_destroy(nv40, vp);
+			memcpy(&vp->ucp, &nv40->clip, sizeof(vp->ucp));
+		}
+	} else {
+		vp = nv40->swtnl.vertprog;
+		constbuf = NULL;
+	}
+
+	/* Translate TGSI shader into hw bytecode */
+	if (vp->translated)
+		goto check_gpu_resources;
+
+	nv40->fallback_swtnl &= ~NV40_NEW_VERTPROG;
+	nv40_vertprog_translate(nv40, vp);
+	if (!vp->translated) {
+		nv40->fallback_swtnl |= NV40_NEW_VERTPROG;
+		return FALSE;
+	}
+
+check_gpu_resources:
+	/* Allocate hw vtxprog exec slots */
+	if (!vp->exec) {
+		struct nouveau_resource *heap = nv40->screen->vp_exec_heap;
+		struct nouveau_stateobj *so;
+		uint vplen = vp->nr_insns;
+
+		if (nvws->res_alloc(heap, vplen, vp, &vp->exec)) {
+			while (heap->next && heap->size < vplen) {
+				struct nv40_vertex_program *evict;
+				
+				evict = heap->next->priv;
+				nvws->res_free(&evict->exec);
+			}
+
+			if (nvws->res_alloc(heap, vplen, vp, &vp->exec))
+				assert(0);
+		}
+
+		so = so_new(7, 0);
+		so_method(so, curie, NV40TCL_VP_START_FROM_ID, 1);
+		so_data  (so, vp->exec->start);
+		so_method(so, curie, NV40TCL_VP_ATTRIB_EN, 2);
+		so_data  (so, vp->ir);
+		so_data  (so, vp->or);
+		so_method(so, curie,  NV40TCL_CLIP_PLANE_ENABLE, 1);
+		so_data  (so, vp->clip_ctrl);
+		so_ref(so, &vp->so);
+
+		upload_code = TRUE;
+	}
+
+	/* Allocate hw vtxprog const slots */
+	if (vp->nr_consts && !vp->data) {
+		struct nouveau_resource *heap = nv40->screen->vp_data_heap;
+
+		if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data)) {
+			while (heap->next && heap->size < vp->nr_consts) {
+				struct nv40_vertex_program *evict;
+				
+				evict = heap->next->priv;
+				nvws->res_free(&evict->data);
+			}
+
+			if (nvws->res_alloc(heap, vp->nr_consts, vp, &vp->data))
+				assert(0);
+		}
+
+		/*XXX: handle this some day */
+		assert(vp->data->start >= vp->data_start_min);
+
+		upload_data = TRUE;
+		if (vp->data_start != vp->data->start)
+			upload_code = TRUE;
+	}
+
+	/* If exec or data segments moved we need to patch the program to
+	 * fixup offsets and register IDs.
+	 */
+	if (vp->exec_start != vp->exec->start) {
+		for (i = 0; i < vp->nr_insns; i++) {
+			struct nv40_vertex_program_exec *vpi = &vp->insns[i];
+
+			if (vpi->has_branch_offset) {
+				assert(0);
+			}
+		}
+
+		vp->exec_start = vp->exec->start;
+	}
+
+	if (vp->nr_consts && vp->data_start != vp->data->start) {
+		for (i = 0; i < vp->nr_insns; i++) {
+			struct nv40_vertex_program_exec *vpi = &vp->insns[i];
+
+			if (vpi->const_index >= 0) {
+				vpi->data[1] &= ~NV40_VP_INST_CONST_SRC_MASK;
+				vpi->data[1] |=
+					(vpi->const_index + vp->data->start) <<
+					NV40_VP_INST_CONST_SRC_SHIFT;
+
+			}
+		}
+
+		vp->data_start = vp->data->start;
+	}
+
+	/* Update + Upload constant values */
+	if (vp->nr_consts) {
+		float *map = NULL;
+
+		if (constbuf) {
+			map = ws->buffer_map(ws, constbuf,
+					     PIPE_BUFFER_USAGE_CPU_READ);
+		}
+
+		for (i = 0; i < vp->nr_consts; i++) {
+			struct nv40_vertex_program_data *vpd = &vp->consts[i];
+
+			if (vpd->index >= 0) {
+				if (!upload_data &&
+				    !memcmp(vpd->value, &map[vpd->index * 4],
+					    4 * sizeof(float)))
+					continue;
+				memcpy(vpd->value, &map[vpd->index * 4],
+				       4 * sizeof(float));
+			}
+
+			BEGIN_RING(curie, NV40TCL_VP_UPLOAD_CONST_ID, 5);
+			OUT_RING  (i + vp->data->start);
+			OUT_RINGp ((uint32_t *)vpd->value, 4);
+		}
+
+		if (constbuf)
+			ws->buffer_unmap(ws, constbuf);
+	}
+
+	/* Upload vtxprog */
+	if (upload_code) {
+#if 0
+		for (i = 0; i < vp->nr_insns; i++) {
+			NOUVEAU_MSG("VP %d: 0x%08x\n", i, vp->insns[i].data[0]);
+			NOUVEAU_MSG("VP %d: 0x%08x\n", i, vp->insns[i].data[1]);
+			NOUVEAU_MSG("VP %d: 0x%08x\n", i, vp->insns[i].data[2]);
+			NOUVEAU_MSG("VP %d: 0x%08x\n", i, vp->insns[i].data[3]);
+		}
+#endif
+		BEGIN_RING(curie, NV40TCL_VP_UPLOAD_FROM_ID, 1);
+		OUT_RING  (vp->exec->start);
+		for (i = 0; i < vp->nr_insns; i++) {
+			BEGIN_RING(curie, NV40TCL_VP_UPLOAD_INST(0), 4);
+			OUT_RINGp (vp->insns[i].data, 4);
+		}
+	}
+
+	if (vp->so != nv40->state.hw[NV40_STATE_VERTPROG]) {
+		so_ref(vp->so, &nv40->state.hw[NV40_STATE_VERTPROG]);
+		return TRUE;
+	}
+
+	return FALSE;
+}
+
+void
+nv40_vertprog_destroy(struct nv40_context *nv40, struct nv40_vertex_program *vp)
+{
+	struct nouveau_winsys *nvws = nv40->screen->nvws;
+
+	vp->translated = FALSE;
+
+	if (vp->nr_insns) {
+		FREE(vp->insns);
+		vp->insns = NULL;
+		vp->nr_insns = 0;
+	}
+
+	if (vp->nr_consts) {
+		FREE(vp->consts);
+		vp->consts = NULL;
+		vp->nr_consts = 0;
+	}
+
+	nvws->res_free(&vp->exec);
+	vp->exec_start = 0;
+	nvws->res_free(&vp->data);
+	vp->data_start = 0;
+	vp->data_start_min = 0;
+
+	vp->ir = vp->or = vp->clip_ctrl = 0;
+	so_ref(NULL, &vp->so);
+}
+
+struct nv40_state_entry nv40_state_vertprog = {
+	.validate = nv40_vertprog_validate,
+	.dirty = {
+		.pipe = NV40_NEW_VERTPROG | NV40_NEW_UCP,
+		.hw = NV40_STATE_VERTPROG,
+	}
+};
+
diff --git a/src/gallium/drivers/nv50/Makefile b/src/gallium/drivers/nv50/Makefile
new file mode 100644
index 0000000000..be30400c03
--- /dev/null
+++ b/src/gallium/drivers/nv50/Makefile
@@ -0,0 +1,29 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nv50
+
+DRIVER_SOURCES = \
+	nv50_clear.c \
+	nv50_context.c \
+	nv50_draw.c \
+	nv50_miptree.c \
+	nv50_query.c \
+	nv50_program.c \
+	nv50_screen.c \
+	nv50_state.c \
+	nv50_state_validate.c \
+	nv50_surface.c \
+	nv50_tex.c \
+	nv50_vbo.c
+
+C_SOURCES = \
+	$(COMMON_SOURCES) \
+	$(DRIVER_SOURCES)
+
+ASM_SOURCES = 
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/drivers/nv50/nv50_clear.c b/src/gallium/drivers/nv50/nv50_clear.c
new file mode 100644
index 0000000000..f9bc3b53ca
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_clear.c
@@ -0,0 +1,92 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+#include "nv50_context.h"
+
+void
+nv50_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+	   unsigned clearValue)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct pipe_framebuffer_state fb, s_fb = nv50->framebuffer;
+	struct pipe_scissor_state sc, s_sc = nv50->scissor;
+	unsigned dirty = nv50->dirty;
+
+	nv50->dirty = 0;
+
+	if (ps->format == PIPE_FORMAT_Z24S8_UNORM ||
+	    ps->format == PIPE_FORMAT_Z16_UNORM) {
+		fb.nr_cbufs = 0;
+		fb.zsbuf = ps;
+	} else {
+		fb.nr_cbufs = 1;
+		fb.cbufs[0] = ps;
+		fb.zsbuf = NULL;
+	}
+	fb.width = ps->width;
+	fb.height = ps->height;
+	pipe->set_framebuffer_state(pipe, &fb);
+
+	sc.minx = sc.miny = 0;
+	sc.maxx = fb.width;
+	sc.maxy = fb.height;
+	pipe->set_scissor_state(pipe, &sc);
+
+	nv50_state_validate(nv50);
+
+	switch (ps->format) {
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+		BEGIN_RING(chan, tesla, 0x0d80, 4);
+		OUT_RINGf (chan, ubyte_to_float((clearValue >> 16) & 0xff));
+		OUT_RINGf (chan, ubyte_to_float((clearValue >>  8) & 0xff));
+		OUT_RINGf (chan, ubyte_to_float((clearValue >>  0) & 0xff));
+		OUT_RINGf (chan, ubyte_to_float((clearValue >> 24) & 0xff));
+		BEGIN_RING(chan, tesla, 0x19d0, 1);
+		OUT_RING  (chan, 0x3c);
+		break;
+	case PIPE_FORMAT_Z24S8_UNORM:
+		BEGIN_RING(chan, tesla, 0x0d90, 1);
+		OUT_RINGf (chan, (float)(clearValue >> 8) * (1.0 / 16777215.0));
+		BEGIN_RING(chan, tesla, 0x0da0, 1);
+		OUT_RING  (chan, clearValue & 0xff);
+		BEGIN_RING(chan, tesla, 0x19d0, 1);
+		OUT_RING  (chan, 0x03);
+		break;
+	default:
+		pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height,
+				   clearValue);
+		break;
+	}
+
+	pipe->set_framebuffer_state(pipe, &s_fb);
+	pipe->set_scissor_state(pipe, &s_sc);
+	nv50->dirty |= dirty;
+
+	ps->status = PIPE_SURFACE_STATUS_CLEAR;
+}
+
diff --git a/src/gallium/drivers/nv50/nv50_context.c b/src/gallium/drivers/nv50/nv50_context.c
new file mode 100644
index 0000000000..565a5da668
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_context.c
@@ -0,0 +1,90 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+#include "nv50_context.h"
+#include "nv50_screen.h"
+
+static void
+nv50_flush(struct pipe_context *pipe, unsigned flags,
+	   struct pipe_fence_handle **fence)
+{
+	struct nv50_context *nv50 = (struct nv50_context *)pipe;
+	
+	FIRE_RING(nv50->screen->nvws->channel);
+}
+
+static void
+nv50_destroy(struct pipe_context *pipe)
+{
+	struct nv50_context *nv50 = (struct nv50_context *)pipe;
+
+	draw_destroy(nv50->draw);
+	FREE(nv50);
+}
+
+
+static void
+nv50_set_edgeflags(struct pipe_context *pipe, const unsigned *bitfield)
+{
+}
+
+struct pipe_context *
+nv50_create(struct pipe_screen *pscreen, unsigned pctx_id)
+{
+	struct pipe_winsys *pipe_winsys = pscreen->winsys;
+	struct nv50_screen *screen = nv50_screen(pscreen);
+	struct nv50_context *nv50;
+
+	nv50 = CALLOC_STRUCT(nv50_context);
+	if (!nv50)
+		return NULL;
+	nv50->screen = screen;
+	nv50->pctx_id = pctx_id;
+
+	nv50->pipe.winsys = pipe_winsys;
+	nv50->pipe.screen = pscreen;
+
+	nv50->pipe.destroy = nv50_destroy;
+
+	nv50->pipe.set_edgeflags = nv50_set_edgeflags;
+	nv50->pipe.draw_arrays = nv50_draw_arrays;
+	nv50->pipe.draw_elements = nv50_draw_elements;
+	nv50->pipe.clear = nv50_clear;
+
+	nv50->pipe.flush = nv50_flush;
+
+	nv50_init_surface_functions(nv50);
+	nv50_init_state_functions(nv50);
+	nv50_init_query_functions(nv50);
+
+	nv50->draw = draw_create();
+	assert(nv50->draw);
+	draw_set_rasterize_stage(nv50->draw, nv50_draw_render_stage(nv50));
+
+	return &nv50->pipe;
+}
+
+		
diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h
new file mode 100644
index 0000000000..1e9d45cb34
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_context.h
@@ -0,0 +1,207 @@
+#ifndef __NV50_CONTEXT_H__
+#define __NV50_CONTEXT_H__
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_compiler.h"
+
+#include "util/u_memory.h"
+#include "util/u_math.h"
+
+#include "draw/draw_vertex.h"
+
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau/nouveau_gldefs.h"
+#include "nouveau/nouveau_stateobj.h"
+
+#include "nv50_screen.h"
+#include "nv50_program.h"
+
+#define NOUVEAU_ERR(fmt, args...) \
+	fprintf(stderr, "%s:%d -  "fmt, __func__, __LINE__, ##args);
+#define NOUVEAU_MSG(fmt, args...) \
+	fprintf(stderr, "nouveau: "fmt, ##args);
+
+/* Constant buffer assignment */
+#define NV50_CB_PMISC		0
+#define NV50_CB_PVP		1
+#define NV50_CB_PFP		2
+#define NV50_CB_PGP		3
+#define NV50_CB_TIC		4
+#define NV50_CB_TSC		5
+#define NV50_CB_PUPLOAD         6
+
+#define NV50_NEW_BLEND		(1 << 0)
+#define NV50_NEW_ZSA		(1 << 1)
+#define NV50_NEW_BLEND_COLOUR	(1 << 2)
+#define NV50_NEW_STIPPLE	(1 << 3)
+#define NV50_NEW_SCISSOR	(1 << 4)
+#define NV50_NEW_VIEWPORT	(1 << 5)
+#define NV50_NEW_RASTERIZER	(1 << 6)
+#define NV50_NEW_FRAMEBUFFER	(1 << 7)
+#define NV50_NEW_VERTPROG	(1 << 8)
+#define NV50_NEW_VERTPROG_CB	(1 << 9)
+#define NV50_NEW_FRAGPROG	(1 << 10)
+#define NV50_NEW_FRAGPROG_CB	(1 << 11)
+#define NV50_NEW_ARRAYS		(1 << 12)
+#define NV50_NEW_SAMPLER	(1 << 13)
+#define NV50_NEW_TEXTURE	(1 << 14)
+
+struct nv50_blend_stateobj {
+	struct pipe_blend_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+struct nv50_zsa_stateobj {
+	struct pipe_depth_stencil_alpha_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+struct nv50_rasterizer_stateobj {
+	struct pipe_rasterizer_state pipe;
+	struct nouveau_stateobj *so;
+};
+
+struct nv50_miptree_level {
+	struct pipe_buffer **image;
+	int *image_offset;
+	unsigned image_dirty_cpu[512/32];
+	unsigned image_dirty_gpu[512/32];
+};
+
+struct nv50_miptree {
+	struct pipe_texture base;
+	struct pipe_buffer *buffer;
+
+	struct nv50_miptree_level level[PIPE_MAX_TEXTURE_LEVELS];
+	int image_nr;
+	int total_size;
+};
+
+static INLINE struct nv50_miptree *
+nv50_miptree(struct pipe_texture *pt)
+{
+	return (struct nv50_miptree *)pt;
+}
+
+struct nv50_surface {
+	struct pipe_surface base;
+};
+
+static INLINE struct nv50_surface *
+nv50_surface(struct pipe_surface *pt)
+{
+	return (struct nv50_surface *)pt;
+}
+
+static INLINE struct pipe_buffer *
+nv50_surface_buffer(struct pipe_surface *surface)
+{
+	struct nv50_miptree *mt = (struct nv50_miptree *)surface->texture;
+	return mt->buffer;
+}
+
+struct nv50_state {
+	unsigned dirty;
+
+	struct nouveau_stateobj *fb;
+	struct nouveau_stateobj *blend;
+	struct nouveau_stateobj *blend_colour;
+	struct nouveau_stateobj *zsa;
+	struct nouveau_stateobj *rast;
+	struct nouveau_stateobj *stipple;
+	struct nouveau_stateobj *scissor;
+	unsigned scissor_enabled;
+	struct nouveau_stateobj *viewport;
+	unsigned viewport_bypass;
+	struct nouveau_stateobj *tsc_upload;
+	struct nouveau_stateobj *tic_upload;
+	struct nouveau_stateobj *vertprog;
+	struct nouveau_stateobj *fragprog;
+	struct nouveau_stateobj *vtxfmt;
+	struct nouveau_stateobj *vtxbuf;
+};
+
+struct nv50_context {
+	struct pipe_context pipe;
+
+	struct nv50_screen *screen;
+	unsigned pctx_id;
+
+	struct draw_context *draw;
+
+	struct nv50_state state;
+
+	unsigned dirty;
+	struct nv50_blend_stateobj *blend;
+	struct nv50_zsa_stateobj *zsa;
+	struct nv50_rasterizer_stateobj *rasterizer;
+	struct pipe_blend_color blend_colour;
+	struct pipe_poly_stipple stipple;
+	struct pipe_scissor_state scissor;
+	struct pipe_viewport_state viewport;
+	struct pipe_framebuffer_state framebuffer;
+	struct nv50_program *vertprog;
+	struct nv50_program *fragprog;
+	struct pipe_buffer *constbuf[PIPE_SHADER_TYPES];
+	struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
+	unsigned vtxbuf_nr;
+	struct pipe_vertex_element vtxelt[PIPE_MAX_ATTRIBS];
+	unsigned vtxelt_nr;
+	unsigned *sampler[PIPE_MAX_SAMPLERS];
+	unsigned sampler_nr;
+	struct nv50_miptree *miptree[PIPE_MAX_SAMPLERS];
+	unsigned miptree_nr;
+};
+
+static INLINE struct nv50_context *
+nv50_context(struct pipe_context *pipe)
+{
+	return (struct nv50_context *)pipe;
+}
+
+extern void nv50_init_surface_functions(struct nv50_context *nv50);
+extern void nv50_init_state_functions(struct nv50_context *nv50);
+extern void nv50_init_query_functions(struct nv50_context *nv50);
+
+extern void nv50_screen_init_miptree_functions(struct pipe_screen *pscreen);
+
+extern int
+nv50_surface_do_copy(struct nv50_screen *screen, struct pipe_surface *dst,
+		     int dx, int dy, struct pipe_surface *src, int sx, int sy,
+		     int w, int h);
+
+/* nv50_draw.c */
+extern struct draw_stage *nv50_draw_render_stage(struct nv50_context *nv50);
+
+/* nv50_vbo.c */
+extern boolean nv50_draw_arrays(struct pipe_context *, unsigned mode,
+				unsigned start, unsigned count);
+extern boolean nv50_draw_elements(struct pipe_context *pipe,
+				  struct pipe_buffer *indexBuffer,
+				  unsigned indexSize,
+				  unsigned mode, unsigned start,
+				  unsigned count);
+extern void nv50_vbo_validate(struct nv50_context *nv50);
+
+/* nv50_clear.c */
+extern void nv50_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+		       unsigned clearValue);
+
+/* nv50_program.c */
+extern void nv50_vertprog_validate(struct nv50_context *nv50);
+extern void nv50_fragprog_validate(struct nv50_context *nv50);
+extern void nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p);
+
+/* nv50_state_validate.c */
+extern boolean nv50_state_validate(struct nv50_context *nv50);
+
+/* nv50_tex.c */
+extern void nv50_tex_validate(struct nv50_context *);
+
+/* nv50_miptree.c */
+extern void nv50_miptree_sync(struct pipe_screen *, struct nv50_miptree *,
+			      unsigned level, unsigned image);
+
+#endif
diff --git a/src/gallium/drivers/nv50/nv50_draw.c b/src/gallium/drivers/nv50/nv50_draw.c
new file mode 100644
index 0000000000..2f6f607261
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_draw.c
@@ -0,0 +1,89 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "draw/draw_pipe.h"
+
+#include "nv50_context.h"
+
+struct nv50_render_stage {
+	struct draw_stage stage;
+	struct nv50_context *nv50;
+};
+
+static INLINE struct nv50_render_stage *
+nv50_render_stage(struct draw_stage *stage)
+{
+	return (struct nv50_render_stage *)stage;
+}
+
+static void
+nv50_render_point(struct draw_stage *stage, struct prim_header *prim)
+{
+	NOUVEAU_ERR("\n");
+}
+
+static void
+nv50_render_line(struct draw_stage *stage, struct prim_header *prim)
+{
+	NOUVEAU_ERR("\n");
+}
+
+static void
+nv50_render_tri(struct draw_stage *stage, struct prim_header *prim)
+{
+	NOUVEAU_ERR("\n");
+}
+
+static void
+nv50_render_flush(struct draw_stage *stage, unsigned flags)
+{
+}
+
+static void
+nv50_render_reset_stipple_counter(struct draw_stage *stage)
+{
+	NOUVEAU_ERR("\n");
+}
+
+static void
+nv50_render_destroy(struct draw_stage *stage)
+{
+	FREE(stage);
+}
+
+struct draw_stage *
+nv50_draw_render_stage(struct nv50_context *nv50)
+{
+	struct nv50_render_stage *rs = CALLOC_STRUCT(nv50_render_stage);
+
+	rs->nv50 = nv50;
+	rs->stage.draw = nv50->draw;
+	rs->stage.destroy = nv50_render_destroy;
+	rs->stage.point = nv50_render_point;
+	rs->stage.line = nv50_render_line;
+	rs->stage.tri = nv50_render_tri;
+	rs->stage.flush = nv50_render_flush;
+	rs->stage.reset_stipple_counter = nv50_render_reset_stipple_counter;
+
+	return &rs->stage;
+}
+
diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c
new file mode 100644
index 0000000000..91091d53f5
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_miptree.c
@@ -0,0 +1,320 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "nv50_context.h"
+
+static struct pipe_texture *
+nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp)
+{
+	struct pipe_winsys *ws = pscreen->winsys;
+	struct nv50_miptree *mt = CALLOC_STRUCT(nv50_miptree);
+	struct pipe_texture *pt = &mt->base;
+	unsigned usage, width = tmp->width[0], height = tmp->height[0];
+	unsigned depth = tmp->depth[0];
+	int i, l;
+
+	mt->base = *tmp;
+	mt->base.refcount = 1;
+	mt->base.screen = pscreen;
+
+	usage = PIPE_BUFFER_USAGE_PIXEL;
+	switch (pt->format) {
+	case PIPE_FORMAT_Z24S8_UNORM:
+	case PIPE_FORMAT_Z16_UNORM:
+		usage |= NOUVEAU_BUFFER_USAGE_ZETA;
+		break;
+	default:
+		break;
+	}
+
+	switch (pt->target) {
+	case PIPE_TEXTURE_3D:
+		mt->image_nr = pt->depth[0];
+		break;
+	case PIPE_TEXTURE_CUBE:
+		mt->image_nr = 6;
+		break;
+	default:
+		mt->image_nr = 1;
+		break;
+	}
+
+	for (l = 0; l <= pt->last_level; l++) {
+		struct nv50_miptree_level *lvl = &mt->level[l];
+
+		pt->width[l] = width;
+		pt->height[l] = height;
+		pt->depth[l] = depth;
+		pt->nblocksx[l] = pf_get_nblocksx(&pt->block, width);
+		pt->nblocksy[l] = pf_get_nblocksy(&pt->block, height);
+
+		lvl->image_offset = CALLOC(mt->image_nr, sizeof(int));
+		lvl->image = CALLOC(mt->image_nr, sizeof(struct pipe_buffer *));
+
+		width = MAX2(1, width >> 1);
+		height = MAX2(1, height >> 1);
+		depth = MAX2(1, depth >> 1);
+	}
+
+	for (i = 0; i < mt->image_nr; i++) {
+		for (l = 0; l <= pt->last_level; l++) {
+			struct nv50_miptree_level *lvl = &mt->level[l];
+			int size;
+
+			size  = align(pt->width[l], 8) * pt->block.size;
+			size  = align(size, 64);
+			size *= align(pt->height[l], 8) * pt->block.size;
+
+			lvl->image[i] = ws->buffer_create(ws, 256, 0, size);
+			lvl->image_offset[i] = mt->total_size;
+
+			mt->total_size += size;
+		}
+	}
+
+	mt->buffer = ws->buffer_create(ws, 256, usage, mt->total_size);
+	if (!mt->buffer) {
+		FREE(mt);
+		return NULL;
+	}
+
+	return &mt->base;
+}
+
+static struct pipe_texture *
+nv50_miptree_blanket(struct pipe_screen *pscreen, const struct pipe_texture *pt,
+		     const unsigned *stride, struct pipe_buffer *pb)
+{
+	struct nv50_miptree *mt;
+
+	/* Only supports 2D, non-mipmapped textures for the moment */
+	if (pt->target != PIPE_TEXTURE_2D || pt->last_level != 0 ||
+	    pt->depth[0] != 1)
+		return NULL;
+
+	mt = CALLOC_STRUCT(nv50_miptree);
+	if (!mt)
+		return NULL;
+
+	mt->base = *pt;
+	mt->base.refcount = 1;
+	mt->base.screen = pscreen;
+	mt->image_nr = 1;
+	mt->level[0].image_offset = CALLOC(1, sizeof(unsigned));
+
+	pipe_buffer_reference(pscreen, &mt->buffer, pb);
+	return &mt->base;
+}
+
+static INLINE void
+mark_dirty(uint32_t *flags, unsigned image)
+{
+	flags[image / 32] |= (1 << (image % 32));
+}
+
+static INLINE void
+mark_clean(uint32_t *flags, unsigned image)
+{
+	flags[image / 32] &= ~(1 << (image % 32));
+}
+
+static INLINE int
+is_dirty(uint32_t *flags, unsigned image)
+{
+	return !!(flags[image / 32] & (1 << (image % 32)));
+}
+
+static void
+nv50_miptree_release(struct pipe_screen *pscreen, struct pipe_texture **ppt)
+{
+	struct pipe_texture *pt = *ppt;
+
+	*ppt = NULL;
+
+	if (--pt->refcount <= 0) {
+		struct nv50_miptree *mt = nv50_miptree(pt);
+
+		pipe_buffer_reference(pscreen, &mt->buffer, NULL);
+		FREE(mt);
+	}
+}
+
+void
+nv50_miptree_sync(struct pipe_screen *pscreen, struct nv50_miptree *mt,
+		  unsigned level, unsigned image)
+{
+	struct nv50_screen *nvscreen = nv50_screen(pscreen);
+	struct nv50_miptree_level *lvl = &mt->level[level];
+	struct pipe_surface *dst, *src;
+	unsigned face = 0, zslice = 0;
+
+	if (!is_dirty(lvl->image_dirty_cpu, image))
+		return;
+
+	if (mt->base.target == PIPE_TEXTURE_CUBE)
+		face = image;
+	else
+	if (mt->base.target == PIPE_TEXTURE_3D)
+		zslice = image;
+
+	/* Mark as clean already - so we don't continually call this function
+	 * trying to get a GPU_WRITE pipe_surface!
+	 */
+	mark_clean(lvl->image_dirty_cpu, image);
+
+	/* Pretend we're doing CPU access so we get the backing pipe_surface
+	 * and not a view into the larger miptree.
+	 */
+	src = pscreen->get_tex_surface(pscreen, &mt->base, face, level, zslice,
+				       PIPE_BUFFER_USAGE_CPU_READ);
+
+	/* Pretend we're only reading with the GPU so surface doesn't get marked
+	 * as dirtied by the GPU.
+	 */
+	dst = pscreen->get_tex_surface(pscreen, &mt->base, face, level, zslice,
+				       PIPE_BUFFER_USAGE_GPU_READ);
+
+	nv50_surface_do_copy(nvscreen, dst, 0, 0, src, 0, 0, dst->width, dst->height);
+
+	pscreen->tex_surface_release(pscreen, &dst);
+	pscreen->tex_surface_release(pscreen, &src);
+}
+
+/* The reverse of the above */
+static void
+nv50_miptree_sync_cpu(struct pipe_screen *pscreen, struct nv50_miptree *mt,
+		      unsigned level, unsigned image)
+{
+	struct nv50_screen *nvscreen = nv50_screen(pscreen);
+	struct nv50_miptree_level *lvl = &mt->level[level];
+	struct pipe_surface *dst, *src;
+	unsigned face = 0, zslice = 0;
+
+	if (!is_dirty(lvl->image_dirty_gpu, image))
+		return;
+
+	if (mt->base.target == PIPE_TEXTURE_CUBE)
+		face = image;
+	else
+	if (mt->base.target == PIPE_TEXTURE_3D)
+		zslice = image;
+
+	mark_clean(lvl->image_dirty_gpu, image);
+
+	src = pscreen->get_tex_surface(pscreen, &mt->base, face, level, zslice,
+				       PIPE_BUFFER_USAGE_GPU_READ);
+	dst = pscreen->get_tex_surface(pscreen, &mt->base, face, level, zslice,
+				       PIPE_BUFFER_USAGE_CPU_READ);
+
+	nv50_surface_do_copy(nvscreen, dst, 0, 0, src, 0, 0, dst->width, dst->height);
+
+	pscreen->tex_surface_release(pscreen, &dst);
+	pscreen->tex_surface_release(pscreen, &src);
+}
+
+static struct pipe_surface *
+nv50_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt,
+			 unsigned face, unsigned level, unsigned zslice,
+			 unsigned flags)
+{
+	struct nv50_miptree *mt = nv50_miptree(pt);
+	struct nv50_miptree_level *lvl = &mt->level[level];
+	struct pipe_surface *ps;
+	int img;
+
+	if (pt->target == PIPE_TEXTURE_CUBE)
+		img = face;
+	else
+	if (pt->target == PIPE_TEXTURE_3D)
+		img = zslice;
+	else
+		img = 0;
+
+	ps = CALLOC_STRUCT(pipe_surface);
+	if (!ps)
+		return NULL;
+	pipe_texture_reference(&ps->texture, pt);
+	ps->format = pt->format;
+	ps->width = pt->width[level];
+	ps->height = pt->height[level];
+	ps->block = pt->block;
+	ps->nblocksx = pt->nblocksx[level];
+	ps->nblocksy = pt->nblocksy[level];
+	ps->stride = ps->width * ps->block.size;
+	ps->usage = flags;
+	ps->status = PIPE_SURFACE_STATUS_DEFINED;
+	ps->refcount = 1;
+	ps->face = face;
+	ps->level = level;
+	ps->zslice = zslice;
+
+	if (flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) {
+		assert(!(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE));
+		nv50_miptree_sync_cpu(pscreen, mt, level, img);
+
+		ps->offset = 0;
+		pipe_texture_reference(&ps->texture, pt);
+
+		if (flags & PIPE_BUFFER_USAGE_CPU_WRITE)
+			mark_dirty(lvl->image_dirty_cpu, img);
+	} else {
+		nv50_miptree_sync(pscreen, mt, level, img);
+
+		ps->offset = lvl->image_offset[img];
+		pipe_texture_reference(&ps->texture, pt);
+
+		if (flags & PIPE_BUFFER_USAGE_GPU_WRITE)
+			mark_dirty(lvl->image_dirty_gpu, img);
+	}
+
+	return ps;
+}
+
+static void
+nv50_miptree_surface_del(struct pipe_screen *pscreen,
+			 struct pipe_surface **psurface)
+{
+	struct pipe_surface *ps = *psurface;
+	struct nv50_surface *s = nv50_surface(ps);
+
+	*psurface = NULL;
+
+	if (--ps->refcount <= 0) {
+		pipe_texture_reference(&ps->texture, NULL);
+		FREE(s);
+	}
+}
+
+void
+nv50_screen_init_miptree_functions(struct pipe_screen *pscreen)
+{
+	pscreen->texture_create = nv50_miptree_create;
+	pscreen->texture_blanket = nv50_miptree_blanket;
+	pscreen->texture_release = nv50_miptree_release;
+	pscreen->get_tex_surface = nv50_miptree_surface_new;
+	pscreen->tex_surface_release = nv50_miptree_surface_del;
+}
+
diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c
new file mode 100644
index 0000000000..14c5d47e79
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_program.c
@@ -0,0 +1,1784 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_util.h"
+
+#include "nv50_context.h"
+
+#define NV50_SU_MAX_TEMP 64
+//#define NV50_PROGRAM_DUMP
+
+/* ARL - gallium craps itself on progs/vp/arl.txt
+ *
+ * MSB - Like MAD, but MUL+SUB
+ * 	- Fuck it off, introduce a way to negate args for ops that
+ * 	  support it.
+ *
+ * Look into inlining IMMD for ops other than MOV (make it general?)
+ * 	- Maybe even relax restrictions a bit, can't do P_RESULT + P_IMMD,
+ * 	  but can emit to P_TEMP first - then MOV later. NVIDIA does this
+ *
+ * In ops such as ADD it's possible to construct a bad opcode in the !is_long()
+ * case, if the emit_src() causes the inst to suddenly become long.
+ *
+ * Verify half-insns work where expected - and force disable them where they
+ * don't work - MUL has it forcibly disabled atm as it fixes POW..
+ *
+ * FUCK! watch dst==src vectors, can overwrite components that are needed.
+ * 	ie. SUB R0, R0.yzxw, R0
+ *
+ * Things to check with renouveau:
+ * 	FP attr/result assignment - how?
+ * 		attrib
+ * 			- 0x16bc maps vp output onto fp hpos
+ * 			- 0x16c0 maps vp output onto fp col0
+ * 		result
+ * 			- colr always 0-3
+ * 			- depr always 4
+ * 0x16bc->0x16e8 --> some binding between vp/fp regs
+ * 0x16b8 --> VP output count
+ *
+ * 0x1298 --> "MOV rcol.x, fcol.y" "MOV depr, fcol.y" = 0x00000005
+ * 	      "MOV rcol.x, fcol.y" = 0x00000004
+ * 0x19a8 --> as above but 0x00000100 and 0x00000000
+ * 	- 0x00100000 used when KIL used
+ * 0x196c --> as above but 0x00000011 and 0x00000000
+ *
+ * 0x1988 --> 0xXXNNNNNN
+ * 	- XX == FP high something
+ */
+struct nv50_reg {
+	enum {
+		P_TEMP,
+		P_ATTR,
+		P_RESULT,
+		P_CONST,
+		P_IMMD
+	} type;
+	int index;
+
+	int hw;
+	int neg;
+};
+
+struct nv50_pc {
+	struct nv50_program *p;
+
+	/* hw resources */
+	struct nv50_reg *r_temp[NV50_SU_MAX_TEMP];
+
+	/* tgsi resources */
+	struct nv50_reg *temp;
+	int temp_nr;
+	struct nv50_reg *attr;
+	int attr_nr;
+	struct nv50_reg *result;
+	int result_nr;
+	struct nv50_reg *param;
+	int param_nr;
+	struct nv50_reg *immd;
+	float *immd_buf;
+	int immd_nr;
+
+	struct nv50_reg *temp_temp[16];
+	unsigned temp_temp_nr;
+};
+
+static void
+alloc_reg(struct nv50_pc *pc, struct nv50_reg *reg)
+{
+	int i;
+
+	if (reg->type == P_RESULT) {
+		if (pc->p->cfg.high_result < (reg->hw + 1))
+			pc->p->cfg.high_result = reg->hw + 1;
+	}
+
+	if (reg->type != P_TEMP)
+		return;
+
+	if (reg->hw >= 0) {
+		/*XXX: do this here too to catch FP temp-as-attr usage..
+		 *     not clean, but works */
+		if (pc->p->cfg.high_temp < (reg->hw + 1))
+			pc->p->cfg.high_temp = reg->hw + 1;
+		return;
+	}
+
+	for (i = 0; i < NV50_SU_MAX_TEMP; i++) {
+		if (!(pc->r_temp[i])) {
+			pc->r_temp[i] = reg;
+			reg->hw = i;
+			if (pc->p->cfg.high_temp < (i + 1))
+				pc->p->cfg.high_temp = i + 1;
+			return;
+		}
+	}
+
+	assert(0);
+}
+
+static struct nv50_reg *
+alloc_temp(struct nv50_pc *pc, struct nv50_reg *dst)
+{
+	struct nv50_reg *r;
+	int i;
+
+	if (dst && dst->type == P_TEMP && dst->hw == -1)
+		return dst;
+
+	for (i = 0; i < NV50_SU_MAX_TEMP; i++) {
+		if (!pc->r_temp[i]) {
+			r = CALLOC_STRUCT(nv50_reg);
+			r->type = P_TEMP;
+			r->index = -1;
+			r->hw = i;
+			pc->r_temp[i] = r;
+			return r;
+		}
+	}
+
+	assert(0);
+	return NULL;
+}
+
+static void
+free_temp(struct nv50_pc *pc, struct nv50_reg *r)
+{
+	if (r->index == -1) {
+		unsigned hw = r->hw;
+
+		FREE(pc->r_temp[hw]);
+		pc->r_temp[hw] = NULL;
+	}
+}
+
+static int
+alloc_temp4(struct nv50_pc *pc, struct nv50_reg *dst[4], int idx)
+{
+	int i;
+
+	if ((idx + 4) >= NV50_SU_MAX_TEMP)
+		return 1;
+
+	if (pc->r_temp[idx] || pc->r_temp[idx + 1] ||
+	    pc->r_temp[idx + 2] || pc->r_temp[idx + 3])
+		return alloc_temp4(pc, dst, idx + 1);
+
+	for (i = 0; i < 4; i++) {
+		dst[i] = CALLOC_STRUCT(nv50_reg);
+		dst[i]->type = P_TEMP;
+		dst[i]->index = -1;
+		dst[i]->hw = idx + i;
+		pc->r_temp[idx + i] = dst[i];
+	}
+
+	return 0;
+}
+
+static void
+free_temp4(struct nv50_pc *pc, struct nv50_reg *reg[4])
+{
+	int i;
+
+	for (i = 0; i < 4; i++)
+		free_temp(pc, reg[i]);
+}
+
+static struct nv50_reg *
+temp_temp(struct nv50_pc *pc)
+{
+	if (pc->temp_temp_nr >= 16)
+		assert(0);
+
+	pc->temp_temp[pc->temp_temp_nr] = alloc_temp(pc, NULL);
+	return pc->temp_temp[pc->temp_temp_nr++];
+}
+
+static void
+kill_temp_temp(struct nv50_pc *pc)
+{
+	int i;
+	
+	for (i = 0; i < pc->temp_temp_nr; i++)
+		free_temp(pc, pc->temp_temp[i]);
+	pc->temp_temp_nr = 0;
+}
+
+static int
+ctor_immd(struct nv50_pc *pc, float x, float y, float z, float w)
+{
+	pc->immd_buf = REALLOC(pc->immd_buf, (pc->immd_nr * r * sizeof(float)),
+			       (pc->immd_nr + 1) * 4 * sizeof(float));
+	pc->immd_buf[(pc->immd_nr * 4) + 0] = x;
+	pc->immd_buf[(pc->immd_nr * 4) + 1] = y;
+	pc->immd_buf[(pc->immd_nr * 4) + 2] = z;
+	pc->immd_buf[(pc->immd_nr * 4) + 3] = w;
+	
+	return pc->immd_nr++;
+}
+
+static struct nv50_reg *
+alloc_immd(struct nv50_pc *pc, float f)
+{
+	struct nv50_reg *r = CALLOC_STRUCT(nv50_reg);
+	unsigned hw;
+
+	hw = ctor_immd(pc, f, 0, 0, 0) * 4;
+	r->type = P_IMMD;
+	r->hw = hw;
+	r->index = -1;
+	return r;
+}
+
+static struct nv50_program_exec *
+exec(struct nv50_pc *pc)
+{
+	struct nv50_program_exec *e = CALLOC_STRUCT(nv50_program_exec);
+
+	e->param.index = -1;
+	return e;
+}
+
+static void
+emit(struct nv50_pc *pc, struct nv50_program_exec *e)
+{
+	struct nv50_program *p = pc->p;
+
+	if (p->exec_tail)
+		p->exec_tail->next = e;
+	if (!p->exec_head)
+		p->exec_head = e;
+	p->exec_tail = e;
+	p->exec_size += (e->inst[0] & 1) ? 2 : 1;
+}
+
+static INLINE void set_long(struct nv50_pc *, struct nv50_program_exec *);
+
+static boolean
+is_long(struct nv50_program_exec *e)
+{
+	if (e->inst[0] & 1)
+		return TRUE;
+	return FALSE;
+}
+
+static boolean
+is_immd(struct nv50_program_exec *e)
+{
+	if (is_long(e) && (e->inst[1] & 3) == 3)
+		return TRUE;
+	return FALSE;
+}
+
+static INLINE void
+set_pred(struct nv50_pc *pc, unsigned pred, unsigned idx,
+	 struct nv50_program_exec *e)
+{
+	set_long(pc, e);
+	e->inst[1] &= ~((0x1f << 7) | (0x3 << 12));
+	e->inst[1] |= (pred << 7) | (idx << 12);
+}
+
+static INLINE void
+set_pred_wr(struct nv50_pc *pc, unsigned on, unsigned idx,
+	    struct nv50_program_exec *e)
+{
+	set_long(pc, e);
+	e->inst[1] &= ~((0x3 << 4) | (1 << 6));
+	e->inst[1] |= (idx << 4) | (on << 6);
+}
+
+static INLINE void
+set_long(struct nv50_pc *pc, struct nv50_program_exec *e)
+{
+	if (is_long(e))
+		return;
+
+	e->inst[0] |= 1;
+	set_pred(pc, 0xf, 0, e);
+	set_pred_wr(pc, 0, 0, e);
+}
+
+static INLINE void
+set_dst(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_program_exec *e)
+{
+	if (dst->type == P_RESULT) {
+		set_long(pc, e);
+		e->inst[1] |= 0x00000008;
+	}
+
+	alloc_reg(pc, dst);
+	e->inst[0] |= (dst->hw << 2);
+}
+
+static INLINE void
+set_immd(struct nv50_pc *pc, struct nv50_reg *imm, struct nv50_program_exec *e)
+{
+	unsigned val = fui(pc->immd_buf[imm->hw]); /* XXX */
+
+	set_long(pc, e);
+	/*XXX: can't be predicated - bits overlap.. catch cases where both
+	 *     are required and avoid them. */
+	set_pred(pc, 0, 0, e);
+	set_pred_wr(pc, 0, 0, e);
+
+	e->inst[1] |= 0x00000002 | 0x00000001;
+	e->inst[0] |= (val & 0x3f) << 16;
+	e->inst[1] |= (val >> 6) << 2;
+}
+
+static void
+emit_interp(struct nv50_pc *pc, struct nv50_reg *dst,
+	    struct nv50_reg *src, struct nv50_reg *iv)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] |= 0x80000000;
+	set_dst(pc, dst, e);
+	alloc_reg(pc, src);
+	e->inst[0] |= (src->hw << 16);
+	if (iv) {
+		e->inst[0] |= (1 << 25);
+		alloc_reg(pc, iv);
+		e->inst[0] |= (iv->hw << 9);
+	}
+
+	emit(pc, e);
+}
+
+static void
+set_data(struct nv50_pc *pc, struct nv50_reg *src, unsigned m, unsigned s,
+	 struct nv50_program_exec *e)
+{
+	set_long(pc, e);
+#if 1
+	e->inst[1] |= (1 << 22);
+#else
+	if (src->type == P_IMMD) {
+		e->inst[1] |= (NV50_CB_PMISC << 22);
+	} else {
+		if (pc->p->type == PIPE_SHADER_VERTEX)
+			e->inst[1] |= (NV50_CB_PVP << 22);
+		else
+			e->inst[1] |= (NV50_CB_PFP << 22);
+	}
+#endif
+
+	e->param.index = src->hw;
+	e->param.shift = s;
+	e->param.mask = m << (s % 32);
+}
+
+static void
+emit_mov(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] |= 0x10000000;
+
+	set_dst(pc, dst, e);
+
+	if (0 && dst->type != P_RESULT && src->type == P_IMMD) {
+		set_immd(pc, src, e);
+		/*XXX: 32-bit, but steals part of "half" reg space - need to
+		 *     catch and handle this case if/when we do half-regs
+		 */
+		e->inst[0] |= 0x00008000;
+	} else
+	if (src->type == P_IMMD || src->type == P_CONST) {
+		set_long(pc, e);
+		set_data(pc, src, 0x7f, 9, e);
+		e->inst[1] |= 0x20000000; /* src0 const? */
+	} else {
+		if (src->type == P_ATTR) {
+			set_long(pc, e);
+			e->inst[1] |= 0x00200000;
+		}
+
+		alloc_reg(pc, src);
+		e->inst[0] |= (src->hw << 9);
+	}
+
+	/* We really should support "half" instructions here at some point,
+	 * but I don't feel confident enough about them yet.
+	 */
+	set_long(pc, e);
+	if (is_long(e) && !is_immd(e)) {
+		e->inst[1] |= 0x04000000; /* 32-bit */
+		e->inst[1] |= 0x0003c000; /* "subsubop" 0xf == mov */
+	}
+
+	emit(pc, e);
+}
+
+static boolean
+check_swap_src_0_1(struct nv50_pc *pc,
+		   struct nv50_reg **s0, struct nv50_reg **s1)
+{
+	struct nv50_reg *src0 = *s0, *src1 = *s1;
+
+	if (src0->type == P_CONST) {
+		if (src1->type != P_CONST) {
+			*s0 = src1;
+			*s1 = src0;
+			return TRUE;
+		}
+	} else
+	if (src1->type == P_ATTR) {
+		if (src0->type != P_ATTR) {
+			*s0 = src1;
+			*s1 = src0;
+			return TRUE;
+		}
+	}
+
+	return FALSE;
+}
+
+static void
+set_src_0(struct nv50_pc *pc, struct nv50_reg *src, struct nv50_program_exec *e)
+{
+	if (src->type == P_ATTR) {
+		set_long(pc, e);
+		e->inst[1] |= 0x00200000;
+	} else
+	if (src->type == P_CONST || src->type == P_IMMD) {
+		struct nv50_reg *temp = temp_temp(pc);
+
+		emit_mov(pc, temp, src);
+		src = temp;
+	}
+
+	alloc_reg(pc, src);
+	e->inst[0] |= (src->hw << 9);
+}
+
+static void
+set_src_1(struct nv50_pc *pc, struct nv50_reg *src, struct nv50_program_exec *e)
+{
+	if (src->type == P_ATTR) {
+		struct nv50_reg *temp = temp_temp(pc);
+
+		emit_mov(pc, temp, src);
+		src = temp;
+	} else
+	if (src->type == P_CONST || src->type == P_IMMD) {
+		assert(!(e->inst[0] & 0x00800000));
+		if (e->inst[0] & 0x01000000) {
+			struct nv50_reg *temp = temp_temp(pc);
+
+			emit_mov(pc, temp, src);
+			src = temp;
+		} else {
+			set_data(pc, src, 0x7f, 16, e);
+			e->inst[0] |= 0x00800000;
+		}
+	}
+
+	alloc_reg(pc, src);
+	e->inst[0] |= (src->hw << 16);
+}
+
+static void
+set_src_2(struct nv50_pc *pc, struct nv50_reg *src, struct nv50_program_exec *e)
+{
+	set_long(pc, e);
+
+	if (src->type == P_ATTR) {
+		struct nv50_reg *temp = temp_temp(pc);
+
+		emit_mov(pc, temp, src);
+		src = temp;
+	} else
+	if (src->type == P_CONST || src->type == P_IMMD) {
+		assert(!(e->inst[0] & 0x01000000));
+		if (e->inst[0] & 0x00800000) {
+			struct nv50_reg *temp = temp_temp(pc);
+
+			emit_mov(pc, temp, src);
+			src = temp;
+		} else {
+			set_data(pc, src, 0x7f, 32+14, e);
+			e->inst[0] |= 0x01000000;
+		}
+	}
+
+	alloc_reg(pc, src);
+	e->inst[1] |= (src->hw << 14);
+}
+
+static void
+emit_mul(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
+	 struct nv50_reg *src1)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] |= 0xc0000000;
+	set_long(pc, e);
+
+	check_swap_src_0_1(pc, &src0, &src1);
+	set_dst(pc, dst, e);
+	set_src_0(pc, src0, e);
+	set_src_1(pc, src1, e);
+
+	emit(pc, e);
+}
+
+static void
+emit_add(struct nv50_pc *pc, struct nv50_reg *dst,
+	 struct nv50_reg *src0, struct nv50_reg *src1)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] |= 0xb0000000;
+
+	check_swap_src_0_1(pc, &src0, &src1);
+	set_dst(pc, dst, e);
+	set_src_0(pc, src0, e);
+	if (is_long(e))
+		set_src_2(pc, src1, e);
+	else
+		set_src_1(pc, src1, e);
+
+	emit(pc, e);
+}
+
+static void
+emit_minmax(struct nv50_pc *pc, unsigned sub, struct nv50_reg *dst,
+	    struct nv50_reg *src0, struct nv50_reg *src1)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	set_long(pc, e);
+	e->inst[0] |= 0xb0000000;
+	e->inst[1] |= (sub << 29);
+
+	check_swap_src_0_1(pc, &src0, &src1);
+	set_dst(pc, dst, e);
+	set_src_0(pc, src0, e);
+	set_src_1(pc, src1, e);
+
+	emit(pc, e);
+}
+
+static void
+emit_sub(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
+	 struct nv50_reg *src1)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] |= 0xb0000000;
+
+	set_long(pc, e);
+	if (check_swap_src_0_1(pc, &src0, &src1))
+		e->inst[1] |= 0x04000000;
+	else
+		e->inst[1] |= 0x08000000;
+
+	set_dst(pc, dst, e);
+	set_src_0(pc, src0, e);
+	set_src_2(pc, src1, e);
+
+	emit(pc, e);
+}
+
+static void
+emit_mad(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
+	 struct nv50_reg *src1, struct nv50_reg *src2)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] |= 0xe0000000;
+
+	check_swap_src_0_1(pc, &src0, &src1);
+	set_dst(pc, dst, e);
+	set_src_0(pc, src0, e);
+	set_src_1(pc, src1, e);
+	set_src_2(pc, src2, e);
+
+	emit(pc, e);
+}
+
+static void
+emit_msb(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src0,
+	 struct nv50_reg *src1, struct nv50_reg *src2)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] |= 0xe0000000;
+	set_long(pc, e);
+	e->inst[1] |= 0x08000000; /* src0 * src1 - src2 */
+
+	check_swap_src_0_1(pc, &src0, &src1);
+	set_dst(pc, dst, e);
+	set_src_0(pc, src0, e);
+	set_src_1(pc, src1, e);
+	set_src_2(pc, src2, e);
+
+	emit(pc, e);
+}
+
+static void
+emit_flop(struct nv50_pc *pc, unsigned sub,
+	  struct nv50_reg *dst, struct nv50_reg *src)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] |= 0x90000000;
+	if (sub) {
+		set_long(pc, e);
+		e->inst[1] |= (sub << 29);
+	}
+
+	set_dst(pc, dst, e);
+	set_src_0(pc, src, e);
+
+	emit(pc, e);
+}
+
+static void
+emit_preex2(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] |= 0xb0000000;
+
+	set_dst(pc, dst, e);
+	set_src_0(pc, src, e);
+	set_long(pc, e);
+	e->inst[1] |= (6 << 29) | 0x00004000;
+
+	emit(pc, e);
+}
+
+static void
+emit_precossin(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] |= 0xb0000000;
+
+	set_dst(pc, dst, e);
+	set_src_0(pc, src, e);
+	set_long(pc, e);
+	e->inst[1] |= (6 << 29);
+
+	emit(pc, e);
+}
+
+static void
+emit_set(struct nv50_pc *pc, unsigned c_op, struct nv50_reg *dst,
+	 struct nv50_reg *src0, struct nv50_reg *src1)
+{
+	struct nv50_program_exec *e = exec(pc);
+	unsigned inv_cop[8] = { 0, 4, 2, 6, 1, 5, 3, 7 };
+	struct nv50_reg *rdst;
+
+	assert(c_op <= 7);
+	if (check_swap_src_0_1(pc, &src0, &src1))
+		c_op = inv_cop[c_op];
+
+	rdst = dst;
+	if (dst->type != P_TEMP)
+		dst = alloc_temp(pc, NULL);
+
+	/* set.u32 */
+	set_long(pc, e);
+	e->inst[0] |= 0xb0000000;
+	e->inst[1] |= (3 << 29);
+	e->inst[1] |= (c_op << 14);
+	/*XXX: breaks things, .u32 by default?
+	 *     decuda will disasm as .u16 and use .lo/.hi regs, but this
+	 *     doesn't seem to match what the hw actually does.
+	inst[1] |= 0x04000000; << breaks things.. .u32 by default?
+	 */
+	set_dst(pc, dst, e);
+	set_src_0(pc, src0, e);
+	set_src_1(pc, src1, e);
+	emit(pc, e);
+
+	/* cvt.f32.u32 */
+	e = exec(pc);
+	e->inst[0] = 0xa0000001;
+	e->inst[1] = 0x64014780;
+	set_dst(pc, rdst, e);
+	set_src_0(pc, dst, e);
+	emit(pc, e);
+
+	if (dst != rdst)
+		free_temp(pc, dst);
+}
+
+static void
+emit_flr(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] = 0xa0000000; /* cvt */
+	set_long(pc, e);
+	e->inst[1] |= (6 << 29); /* cvt */
+	e->inst[1] |= 0x08000000; /* integer mode */
+	e->inst[1] |= 0x04000000; /* 32 bit */
+	e->inst[1] |= ((0x1 << 3)) << 14; /* .rn */
+	e->inst[1] |= (1 << 14); /* src .f32 */
+	set_dst(pc, dst, e);
+	set_src_0(pc, src, e);
+
+	emit(pc, e);
+}
+
+static void
+emit_pow(struct nv50_pc *pc, struct nv50_reg *dst,
+	 struct nv50_reg *v, struct nv50_reg *e)
+{
+	struct nv50_reg *temp = alloc_temp(pc, NULL);
+
+	emit_flop(pc, 3, temp, v);
+	emit_mul(pc, temp, temp, e);
+	emit_preex2(pc, temp, temp);
+	emit_flop(pc, 6, dst, temp);
+
+	free_temp(pc, temp);
+}
+
+static void
+emit_abs(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	e->inst[0] = 0xa0000000; /* cvt */
+	set_long(pc, e);
+	e->inst[1] |= (6 << 29); /* cvt */
+	e->inst[1] |= 0x04000000; /* 32 bit */
+	e->inst[1] |= (1 << 14); /* src .f32 */
+	e->inst[1] |= ((1 << 6) << 14); /* .abs */
+	set_dst(pc, dst, e);
+	set_src_0(pc, src, e);
+
+	emit(pc, e);
+}
+
+static void
+emit_lit(struct nv50_pc *pc, struct nv50_reg **dst, unsigned mask,
+	 struct nv50_reg **src)
+{
+	struct nv50_reg *one = alloc_immd(pc, 1.0);
+	struct nv50_reg *zero = alloc_immd(pc, 0.0);
+	struct nv50_reg *neg128 = alloc_immd(pc, -127.999999);
+	struct nv50_reg *pos128 = alloc_immd(pc,  127.999999);
+	struct nv50_reg *tmp[4];
+
+	if (mask & (1 << 0))
+		emit_mov(pc, dst[0], one);
+
+	if (mask & (1 << 3))
+		emit_mov(pc, dst[3], one);
+
+	if (mask & (3 << 1)) {
+		if (mask & (1 << 1))
+			tmp[0] = dst[1];
+		else
+			tmp[0] = temp_temp(pc);
+		emit_minmax(pc, 4, tmp[0], src[0], zero);
+	}
+
+	if (mask & (1 << 2)) {
+		set_pred_wr(pc, 1, 0, pc->p->exec_tail);
+
+		tmp[1] = temp_temp(pc);
+		emit_minmax(pc, 4, tmp[1], src[1], zero);
+
+		tmp[3] = temp_temp(pc);
+		emit_minmax(pc, 4, tmp[3], src[3], neg128);
+		emit_minmax(pc, 5, tmp[3], tmp[3], pos128);
+
+		emit_pow(pc, dst[2], tmp[1], tmp[3]);
+		emit_mov(pc, dst[2], zero);
+		set_pred(pc, 3, 0, pc->p->exec_tail);
+	}
+}
+
+static void
+emit_neg(struct nv50_pc *pc, struct nv50_reg *dst, struct nv50_reg *src)
+{
+	struct nv50_program_exec *e = exec(pc);
+
+	set_long(pc, e);
+	e->inst[0] |= 0xa0000000; /* delta */
+	e->inst[1] |= (7 << 29); /* delta */
+	e->inst[1] |= 0x04000000; /* negate arg0? probably not */
+	e->inst[1] |= (1 << 14); /* src .f32 */
+	set_dst(pc, dst, e);
+	set_src_0(pc, src, e);
+
+	emit(pc, e);
+}
+
+static void
+emit_kil(struct nv50_pc *pc, struct nv50_reg *src)
+{
+	struct nv50_program_exec *e;
+	const int r_pred = 1;
+
+	/* Sets predicate reg ? */
+	e = exec(pc);
+	e->inst[0] = 0xa00001fd;
+	e->inst[1] = 0xc4014788;
+	set_src_0(pc, src, e);
+	set_pred_wr(pc, 1, r_pred, e);
+	emit(pc, e);
+
+	/* This is probably KILP */
+	e = exec(pc);
+	e->inst[0] = 0x000001fe;
+	set_long(pc, e);
+	set_pred(pc, 1 /* LT? */, r_pred, e);
+	emit(pc, e);
+}
+
+static struct nv50_reg *
+tgsi_dst(struct nv50_pc *pc, int c, const struct tgsi_full_dst_register *dst)
+{
+	switch (dst->DstRegister.File) {
+	case TGSI_FILE_TEMPORARY:
+		return &pc->temp[dst->DstRegister.Index * 4 + c];
+	case TGSI_FILE_OUTPUT:
+		return &pc->result[dst->DstRegister.Index * 4 + c];
+	case TGSI_FILE_NULL:
+		return NULL;
+	default:
+		break;
+	}
+
+	return NULL;
+}
+
+static struct nv50_reg *
+tgsi_src(struct nv50_pc *pc, int chan, const struct tgsi_full_src_register *src)
+{
+	struct nv50_reg *r = NULL;
+	struct nv50_reg *temp;
+	unsigned c;
+
+	c = tgsi_util_get_full_src_register_extswizzle(src, chan);
+	switch (c) {
+	case TGSI_EXTSWIZZLE_X:
+	case TGSI_EXTSWIZZLE_Y:
+	case TGSI_EXTSWIZZLE_Z:
+	case TGSI_EXTSWIZZLE_W:
+		switch (src->SrcRegister.File) {
+		case TGSI_FILE_INPUT:
+			r = &pc->attr[src->SrcRegister.Index * 4 + c];
+			break;
+		case TGSI_FILE_TEMPORARY:
+			r = &pc->temp[src->SrcRegister.Index * 4 + c];
+			break;
+		case TGSI_FILE_CONSTANT:
+			r = &pc->param[src->SrcRegister.Index * 4 + c];
+			break;
+		case TGSI_FILE_IMMEDIATE:
+			r = &pc->immd[src->SrcRegister.Index * 4 + c];
+			break;
+		case TGSI_FILE_SAMPLER:
+			break;
+		default:
+			assert(0);
+			break;
+		}
+		break;
+	case TGSI_EXTSWIZZLE_ZERO:
+		r = alloc_immd(pc, 0.0);
+		break;
+	case TGSI_EXTSWIZZLE_ONE:
+		r = alloc_immd(pc, 1.0);
+		break;
+	default:
+		assert(0);
+		break;
+	}
+
+	switch (tgsi_util_get_full_src_register_sign_mode(src, chan)) {
+	case TGSI_UTIL_SIGN_KEEP:
+		break;
+	case TGSI_UTIL_SIGN_CLEAR:
+		temp = temp_temp(pc);
+		emit_abs(pc, temp, r);
+		r = temp;
+		break;
+	case TGSI_UTIL_SIGN_TOGGLE:
+		temp = temp_temp(pc);
+		emit_neg(pc, temp, r);
+		r = temp;
+		break;
+	case TGSI_UTIL_SIGN_SET:
+		temp = temp_temp(pc);
+		emit_abs(pc, temp, r);
+		emit_neg(pc, temp, r);
+		r = temp;
+		break;
+	default:
+		assert(0);
+		break;
+	}
+
+	return r;
+}
+
+static boolean
+nv50_program_tx_insn(struct nv50_pc *pc, const union tgsi_full_token *tok)
+{
+	const struct tgsi_full_instruction *inst = &tok->FullInstruction;
+	struct nv50_reg *rdst[4], *dst[4], *src[3][4], *temp;
+	unsigned mask, sat, unit;
+	int i, c;
+
+	mask = inst->FullDstRegisters[0].DstRegister.WriteMask;
+	sat = inst->Instruction.Saturate == TGSI_SAT_ZERO_ONE;
+
+	for (c = 0; c < 4; c++) {
+		if (mask & (1 << c))
+			dst[c] = tgsi_dst(pc, c, &inst->FullDstRegisters[0]);
+		else
+			dst[c] = NULL;
+	}
+
+	for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
+		const struct tgsi_full_src_register *fs = &inst->FullSrcRegisters[i];
+
+		if (fs->SrcRegister.File == TGSI_FILE_SAMPLER)
+			unit = fs->SrcRegister.Index;
+
+		for (c = 0; c < 4; c++)
+			src[i][c] = tgsi_src(pc, c, fs);
+	}
+
+	if (sat) {
+		for (c = 0; c < 4; c++) {
+			rdst[c] = dst[c];
+			dst[c] = temp_temp(pc);
+		}
+	}
+
+	switch (inst->Instruction.Opcode) {
+	case TGSI_OPCODE_ABS:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_abs(pc, dst[c], src[0][c]);
+		}
+		break;
+	case TGSI_OPCODE_ADD:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_add(pc, dst[c], src[0][c], src[1][c]);
+		}
+		break;
+	case TGSI_OPCODE_COS:
+		temp = alloc_temp(pc, NULL);
+		emit_precossin(pc, temp, src[0][0]);
+		emit_flop(pc, 5, temp, temp);
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mov(pc, dst[c], temp);
+		}
+		break;
+	case TGSI_OPCODE_DP3:
+		temp = alloc_temp(pc, NULL);
+		emit_mul(pc, temp, src[0][0], src[1][0]);
+		emit_mad(pc, temp, src[0][1], src[1][1], temp);
+		emit_mad(pc, temp, src[0][2], src[1][2], temp);
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mov(pc, dst[c], temp);
+		}
+		free_temp(pc, temp);
+		break;
+	case TGSI_OPCODE_DP4:
+		temp = alloc_temp(pc, NULL);
+		emit_mul(pc, temp, src[0][0], src[1][0]);
+		emit_mad(pc, temp, src[0][1], src[1][1], temp);
+		emit_mad(pc, temp, src[0][2], src[1][2], temp);
+		emit_mad(pc, temp, src[0][3], src[1][3], temp);
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mov(pc, dst[c], temp);
+		}
+		free_temp(pc, temp);
+		break;
+	case TGSI_OPCODE_DPH:
+		temp = alloc_temp(pc, NULL);
+		emit_mul(pc, temp, src[0][0], src[1][0]);
+		emit_mad(pc, temp, src[0][1], src[1][1], temp);
+		emit_mad(pc, temp, src[0][2], src[1][2], temp);
+		emit_add(pc, temp, src[1][3], temp);
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mov(pc, dst[c], temp);
+		}
+		free_temp(pc, temp);
+		break;
+	case TGSI_OPCODE_DST:
+	{
+		struct nv50_reg *one = alloc_immd(pc, 1.0);
+		if (mask & (1 << 0))
+			emit_mov(pc, dst[0], one);
+		if (mask & (1 << 1))
+			emit_mul(pc, dst[1], src[0][1], src[1][1]);
+		if (mask & (1 << 2))
+			emit_mov(pc, dst[2], src[0][2]);
+		if (mask & (1 << 3))
+			emit_mov(pc, dst[3], src[1][3]);
+		FREE(one);
+	}
+		break;
+	case TGSI_OPCODE_EX2:
+		temp = alloc_temp(pc, NULL);
+		emit_preex2(pc, temp, src[0][0]);
+		emit_flop(pc, 6, temp, temp);
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mov(pc, dst[c], temp);
+		}
+		free_temp(pc, temp);
+		break;
+	case TGSI_OPCODE_FLR:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_flr(pc, dst[c], src[0][c]);
+		}
+		break;
+	case TGSI_OPCODE_FRC:
+		temp = alloc_temp(pc, NULL);
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_flr(pc, temp, src[0][c]);
+			emit_sub(pc, dst[c], src[0][c], temp);
+		}
+		free_temp(pc, temp);
+		break;
+	case TGSI_OPCODE_KIL:
+		emit_kil(pc, src[0][0]);
+		emit_kil(pc, src[0][1]);
+		emit_kil(pc, src[0][2]);
+		emit_kil(pc, src[0][3]);
+		break;
+	case TGSI_OPCODE_LIT:
+		emit_lit(pc, &dst[0], mask, &src[0][0]);
+		break;
+	case TGSI_OPCODE_LG2:
+		temp = alloc_temp(pc, NULL);
+		emit_flop(pc, 3, temp, src[0][0]);
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mov(pc, dst[c], temp);
+		}
+		break;
+	case TGSI_OPCODE_LRP:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			/*XXX: we can do better than this */
+			temp = alloc_temp(pc, NULL);
+			emit_neg(pc, temp, src[0][c]);
+			emit_mad(pc, temp, temp, src[2][c], src[2][c]);
+			emit_mad(pc, dst[c], src[0][c], src[1][c], temp);
+			free_temp(pc, temp);
+		}
+		break;
+	case TGSI_OPCODE_MAD:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mad(pc, dst[c], src[0][c], src[1][c], src[2][c]);
+		}
+		break;
+	case TGSI_OPCODE_MAX:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_minmax(pc, 4, dst[c], src[0][c], src[1][c]);
+		}
+		break;
+	case TGSI_OPCODE_MIN:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_minmax(pc, 5, dst[c], src[0][c], src[1][c]);
+		}
+		break;
+	case TGSI_OPCODE_MOV:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mov(pc, dst[c], src[0][c]);
+		}
+		break;
+	case TGSI_OPCODE_MUL:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mul(pc, dst[c], src[0][c], src[1][c]);
+		}
+		break;
+	case TGSI_OPCODE_POW:
+		temp = alloc_temp(pc, NULL);
+		emit_pow(pc, temp, src[0][0], src[1][0]);
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mov(pc, dst[c], temp);
+		}
+		free_temp(pc, temp);
+		break;
+	case TGSI_OPCODE_RCP:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_flop(pc, 0, dst[c], src[0][0]);
+		}
+		break;
+	case TGSI_OPCODE_RSQ:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_flop(pc, 2, dst[c], src[0][0]);
+		}
+		break;
+	case TGSI_OPCODE_SCS:
+		temp = alloc_temp(pc, NULL);
+		emit_precossin(pc, temp, src[0][0]);
+		if (mask & (1 << 0))
+			emit_flop(pc, 5, dst[0], temp);
+		if (mask & (1 << 1))
+			emit_flop(pc, 4, dst[1], temp);
+		break;
+	case TGSI_OPCODE_SGE:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_set(pc, 6, dst[c], src[0][c], src[1][c]);
+		}
+		break;
+	case TGSI_OPCODE_SIN:
+		temp = alloc_temp(pc, NULL);
+		emit_precossin(pc, temp, src[0][0]);
+		emit_flop(pc, 4, temp, temp);
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_mov(pc, dst[c], temp);
+		}
+		break;
+	case TGSI_OPCODE_SLT:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_set(pc, 1, dst[c], src[0][c], src[1][c]);
+		}
+		break;
+	case TGSI_OPCODE_SUB:
+		for (c = 0; c < 4; c++) {
+			if (!(mask & (1 << c)))
+				continue;
+			emit_sub(pc, dst[c], src[0][c], src[1][c]);
+		}
+		break;
+	case TGSI_OPCODE_TEX:
+	case TGSI_OPCODE_TXP:
+	{
+		struct nv50_reg *t[4];
+		struct nv50_program_exec *e;
+
+		alloc_temp4(pc, t, 0);
+		emit_mov(pc, t[0], src[0][0]);
+		emit_mov(pc, t[1], src[0][1]);
+
+		e = exec(pc);
+		e->inst[0] = 0xf6400000;
+		e->inst[0] |= (unit << 9);
+		set_long(pc, e);
+		e->inst[1] |= 0x0000c004;
+		set_dst(pc, t[0], e);
+		emit(pc, e);
+
+		if (mask & (1 << 0)) emit_mov(pc, dst[0], t[0]);
+		if (mask & (1 << 1)) emit_mov(pc, dst[1], t[1]);
+		if (mask & (1 << 2)) emit_mov(pc, dst[2], t[2]);
+		if (mask & (1 << 3)) emit_mov(pc, dst[3], t[3]);
+
+		free_temp4(pc, t);
+	}
+		break;
+	case TGSI_OPCODE_XPD:
+		temp = alloc_temp(pc, NULL);
+		if (mask & (1 << 0)) {
+			emit_mul(pc, temp, src[0][2], src[1][1]);
+			emit_msb(pc, dst[0], src[0][1], src[1][2], temp);
+		}
+		if (mask & (1 << 1)) {
+			emit_mul(pc, temp, src[0][0], src[1][2]);
+			emit_msb(pc, dst[1], src[0][2], src[1][0], temp);
+		}
+		if (mask & (1 << 2)) {
+			emit_mul(pc, temp, src[0][1], src[1][0]);
+			emit_msb(pc, dst[2], src[0][0], src[1][1], temp);
+		}
+		free_temp(pc, temp);
+		break;
+	case TGSI_OPCODE_END:
+		break;
+	default:
+		NOUVEAU_ERR("invalid opcode %d\n", inst->Instruction.Opcode);
+		return FALSE;
+	}
+
+	if (sat) {
+		for (c = 0; c < 4; c++) {
+			struct nv50_program_exec *e;
+
+			if (!(mask & (1 << c)))
+				continue;
+			e = exec(pc);
+
+			e->inst[0] = 0xa0000000; /* cvt */
+			set_long(pc, e);
+			e->inst[1] |= (6 << 29); /* cvt */
+			e->inst[1] |= 0x04000000; /* 32 bit */
+			e->inst[1] |= (1 << 14); /* src .f32 */
+			e->inst[1] |= ((1 << 5) << 14); /* .sat */
+			set_dst(pc, rdst[c], e);
+			set_src_0(pc, dst[c], e);
+			emit(pc, e);
+		}
+	}
+
+	kill_temp_temp(pc);
+	return TRUE;
+}
+
+static boolean
+nv50_program_tx_prep(struct nv50_pc *pc)
+{
+	struct tgsi_parse_context p;
+	boolean ret = FALSE;
+	unsigned i, c;
+
+	tgsi_parse_init(&p, pc->p->pipe.tokens);
+	while (!tgsi_parse_end_of_tokens(&p)) {
+		const union tgsi_full_token *tok = &p.FullToken;
+
+		tgsi_parse_token(&p);
+		switch (tok->Token.Type) {
+		case TGSI_TOKEN_TYPE_IMMEDIATE:
+		{
+			const struct tgsi_full_immediate *imm =
+				&p.FullToken.FullImmediate;
+
+			ctor_immd(pc, imm->u.ImmediateFloat32[0].Float,
+				      imm->u.ImmediateFloat32[1].Float,
+				      imm->u.ImmediateFloat32[2].Float,
+				      imm->u.ImmediateFloat32[3].Float);
+		}
+			break;
+		case TGSI_TOKEN_TYPE_DECLARATION:
+		{
+			const struct tgsi_full_declaration *d;
+			unsigned last;
+
+			d = &p.FullToken.FullDeclaration;
+			last = d->DeclarationRange.Last;
+
+			switch (d->Declaration.File) {
+			case TGSI_FILE_TEMPORARY:
+				if (pc->temp_nr < (last + 1))
+					pc->temp_nr = last + 1;
+				break;
+			case TGSI_FILE_OUTPUT:
+				if (pc->result_nr < (last + 1))
+					pc->result_nr = last + 1;
+				break;
+			case TGSI_FILE_INPUT:
+				if (pc->attr_nr < (last + 1))
+					pc->attr_nr = last + 1;
+				break;
+			case TGSI_FILE_CONSTANT:
+				if (pc->param_nr < (last + 1))
+					pc->param_nr = last + 1;
+				break;
+			case TGSI_FILE_SAMPLER:
+				break;
+			default:
+				NOUVEAU_ERR("bad decl file %d\n",
+					    d->Declaration.File);
+				goto out_err;
+			}
+		}
+			break;
+		case TGSI_TOKEN_TYPE_INSTRUCTION:
+			break;
+		default:
+			break;
+		}
+	}
+
+	if (pc->temp_nr) {
+		pc->temp = CALLOC(pc->temp_nr * 4, sizeof(struct nv50_reg));
+		if (!pc->temp)
+			goto out_err;
+
+		for (i = 0; i < pc->temp_nr; i++) {
+			for (c = 0; c < 4; c++) {
+				pc->temp[i*4+c].type = P_TEMP;
+				pc->temp[i*4+c].hw = -1;
+				pc->temp[i*4+c].index = i;
+			}
+		}
+	}
+
+	if (pc->attr_nr) {
+		struct nv50_reg *iv = NULL;
+		int aid = 0;
+
+		pc->attr = CALLOC(pc->attr_nr * 4, sizeof(struct nv50_reg));
+		if (!pc->attr)
+			goto out_err;
+
+		if (pc->p->type == PIPE_SHADER_FRAGMENT) {
+			iv = alloc_temp(pc, NULL);
+			emit_interp(pc, iv, iv, NULL);
+			emit_flop(pc, 0, iv, iv);
+			aid++;
+		}
+
+		for (i = 0; i < pc->attr_nr; i++) {
+			struct nv50_reg *a = &pc->attr[i*4];
+
+			for (c = 0; c < 4; c++) {
+				if (pc->p->type == PIPE_SHADER_FRAGMENT) {
+					struct nv50_reg *at =
+						alloc_temp(pc, NULL);
+					pc->attr[i*4+c].type = at->type;
+					pc->attr[i*4+c].hw = at->hw;
+					pc->attr[i*4+c].index = at->index;
+				} else {
+					pc->p->cfg.vp.attr[aid/32] |=
+						(1 << (aid % 32));
+					pc->attr[i*4+c].type = P_ATTR;
+					pc->attr[i*4+c].hw = aid++;
+					pc->attr[i*4+c].index = i;
+				}
+			}
+
+			if (pc->p->type != PIPE_SHADER_FRAGMENT)
+				continue;
+
+			emit_interp(pc, &a[0], &a[0], iv);
+			emit_interp(pc, &a[1], &a[1], iv);
+			emit_interp(pc, &a[2], &a[2], iv);
+			emit_interp(pc, &a[3], &a[3], iv);
+		}
+
+		if (iv)
+			free_temp(pc, iv);
+	}
+
+	if (pc->result_nr) {
+		int rid = 0;
+
+		pc->result = CALLOC(pc->result_nr * 4, sizeof(struct nv50_reg));
+		if (!pc->result)
+			goto out_err;
+
+		for (i = 0; i < pc->result_nr; i++) {
+			for (c = 0; c < 4; c++) {
+				if (pc->p->type == PIPE_SHADER_FRAGMENT) {
+					pc->result[i*4+c].type = P_TEMP;
+					pc->result[i*4+c].hw = -1;
+				} else {
+					pc->result[i*4+c].type = P_RESULT;
+					pc->result[i*4+c].hw = rid++;
+				}
+				pc->result[i*4+c].index = i;
+			}
+		}
+	}
+
+	if (pc->param_nr) {
+		int rid = 0;
+
+		pc->param = CALLOC(pc->param_nr * 4, sizeof(struct nv50_reg));
+		if (!pc->param)
+			goto out_err;
+
+		for (i = 0; i < pc->param_nr; i++) {
+			for (c = 0; c < 4; c++) {
+				pc->param[i*4+c].type = P_CONST;
+				pc->param[i*4+c].hw = rid++;
+				pc->param[i*4+c].index = i;
+			}
+		}
+	}
+
+	if (pc->immd_nr) {
+		int rid = pc->param_nr * 4;
+
+		pc->immd = CALLOC(pc->immd_nr * 4, sizeof(struct nv50_reg));
+		if (!pc->immd)
+			goto out_err;
+
+		for (i = 0; i < pc->immd_nr; i++) {
+			for (c = 0; c < 4; c++) {
+				pc->immd[i*4+c].type = P_IMMD;
+				pc->immd[i*4+c].hw = rid++;
+				pc->immd[i*4+c].index = i;
+			}
+		}
+	}
+
+	ret = TRUE;
+out_err:
+	tgsi_parse_free(&p);
+	return ret;
+}
+
+static boolean
+nv50_program_tx(struct nv50_program *p)
+{
+	struct tgsi_parse_context parse;
+	struct nv50_pc *pc;
+	boolean ret;
+
+	pc = CALLOC_STRUCT(nv50_pc);
+	if (!pc)
+		return FALSE;
+	pc->p = p;
+	pc->p->cfg.high_temp = 4;
+
+	ret = nv50_program_tx_prep(pc);
+	if (ret == FALSE)
+		goto out_cleanup;
+
+	tgsi_parse_init(&parse, pc->p->pipe.tokens);
+	while (!tgsi_parse_end_of_tokens(&parse)) {
+		const union tgsi_full_token *tok = &parse.FullToken;
+
+		tgsi_parse_token(&parse);
+
+		switch (tok->Token.Type) {
+		case TGSI_TOKEN_TYPE_INSTRUCTION:
+			ret = nv50_program_tx_insn(pc, tok);
+			if (ret == FALSE)
+				goto out_err;
+			break;
+		default:
+			break;
+		}
+	}
+
+	if (p->type == PIPE_SHADER_FRAGMENT) {
+		struct nv50_reg out;
+
+		out.type = P_TEMP;
+		for (out.hw = 0; out.hw < pc->result_nr * 4; out.hw++)
+			emit_mov(pc, &out, &pc->result[out.hw]);
+	}
+
+	assert(is_long(pc->p->exec_tail) && !is_immd(pc->p->exec_head));
+	pc->p->exec_tail->inst[1] |= 0x00000001;
+
+	p->param_nr = pc->param_nr * 4;
+	p->immd_nr = pc->immd_nr * 4;
+	p->immd = pc->immd_buf;
+
+out_err:
+	tgsi_parse_free(&parse);
+
+out_cleanup:
+	return ret;
+}
+
+static void
+nv50_program_validate(struct nv50_context *nv50, struct nv50_program *p)
+{
+	if (nv50_program_tx(p) == FALSE)
+		assert(0);
+	p->translated = TRUE;
+}
+
+static void
+nv50_program_upload_data(struct nv50_context *nv50, float *map,
+			 unsigned start, unsigned count)
+{
+	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+
+	while (count) {
+		unsigned nr = count > 2047 ? 2047 : count;
+
+		BEGIN_RING(chan, tesla, 0x00000f00, 1);
+		OUT_RING  (chan, (NV50_CB_PMISC << 0) | (start << 8));
+		BEGIN_RING(chan, tesla, 0x40000f04, nr);
+		OUT_RINGp (chan, map, nr);
+
+		map += nr;
+		start += nr;
+		count -= nr;
+	}
+}
+
+static void
+nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p)
+{
+	struct nouveau_winsys *nvws = nv50->screen->nvws;
+	struct pipe_winsys *ws = nv50->pipe.winsys;
+	unsigned nr = p->param_nr + p->immd_nr;
+
+	if (!p->data && nr) {
+		struct nouveau_resource *heap = nv50->screen->vp_data_heap;
+
+		if (nvws->res_alloc(heap, nr, p, &p->data)) {
+			while (heap->next && heap->size < nr) {
+				struct nv50_program *evict = heap->next->priv;
+				nvws->res_free(&evict->data);
+			}
+
+			if (nvws->res_alloc(heap, nr, p, &p->data))
+				assert(0);
+		}
+	}
+
+	if (p->param_nr) {
+		float *map = ws->buffer_map(ws, nv50->constbuf[p->type],
+					    PIPE_BUFFER_USAGE_CPU_READ);
+		nv50_program_upload_data(nv50, map, p->data->start,
+					 p->param_nr);
+		ws->buffer_unmap(ws, nv50->constbuf[p->type]);
+	}
+
+	if (p->immd_nr) {
+		nv50_program_upload_data(nv50, p->immd,
+					 p->data->start + p->param_nr,
+					 p->immd_nr);
+	}
+}
+
+static void
+nv50_program_validate_code(struct nv50_context *nv50, struct nv50_program *p)
+{
+	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct pipe_winsys *ws = nv50->pipe.winsys;
+	struct nv50_program_exec *e;
+	struct nouveau_stateobj *so;
+	const unsigned flags = NOUVEAU_BO_VRAM | NOUVEAU_BO_WR;
+	unsigned start, count, *up, *ptr;
+	boolean upload = FALSE;
+
+	if (!p->buffer) {
+		p->buffer = ws->buffer_create(ws, 0x100, 0, p->exec_size * 4);
+		upload = TRUE;
+	}
+
+	if (p->data && p->data->start != p->data_start) {
+		for (e = p->exec_head; e; e = e->next) {
+			unsigned ei, ci;
+
+			if (e->param.index < 0)
+				continue;
+			ei = e->param.shift >> 5;
+			ci = e->param.index + p->data->start;
+
+			e->inst[ei] &= ~e->param.mask;
+			e->inst[ei] |= (ci << e->param.shift);
+		}
+
+		p->data_start = p->data->start;
+		upload = TRUE;
+	}
+
+	if (!upload)
+		return;
+
+#ifdef NV50_PROGRAM_DUMP
+	NOUVEAU_ERR("-------\n");
+	up = ptr = MALLOC(p->exec_size * 4);
+	for (e = p->exec_head; e; e = e->next) {
+		NOUVEAU_ERR("0x%08x\n", e->inst[0]);
+		if (is_long(e))
+			NOUVEAU_ERR("0x%08x\n", e->inst[1]);
+	}
+
+#endif
+
+	up = ptr = MALLOC(p->exec_size * 4);
+	for (e = p->exec_head; e; e = e->next) {
+		*(ptr++) = e->inst[0];
+		if (is_long(e))
+			*(ptr++) = e->inst[1];
+	}
+
+	so = so_new(4,2);
+	so_method(so, nv50->screen->tesla, 0x1280, 3);
+	so_reloc (so, p->buffer, 0, flags | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, p->buffer, 0, flags | NOUVEAU_BO_LOW, 0, 0);
+	so_data  (so, (NV50_CB_PUPLOAD << 16) | 0x0800); //(p->exec_size * 4));
+
+	start = 0; count = p->exec_size;
+	while (count) {
+		struct nouveau_winsys *nvws = nv50->screen->nvws;
+		unsigned nr;
+
+		so_emit(nvws, so);
+
+		nr = MIN2(count, 2047);
+		nr = MIN2(nvws->channel->pushbuf->remaining, nr);
+		if (nvws->channel->pushbuf->remaining < (nr + 3)) {
+			FIRE_RING(chan);
+			continue;
+		}
+
+		BEGIN_RING(chan, tesla, 0x0f00, 1);
+		OUT_RING  (chan, (start << 8) | NV50_CB_PUPLOAD);
+		BEGIN_RING(chan, tesla, 0x40000f04, nr);	
+		OUT_RINGp (chan, up + start, nr);
+
+		start += nr;
+		count -= nr;
+	}
+
+	FREE(up);
+	so_ref(NULL, &so);
+}
+
+void
+nv50_vertprog_validate(struct nv50_context *nv50)
+{
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct nv50_program *p = nv50->vertprog;
+	struct nouveau_stateobj *so;
+
+	if (!p->translated) {
+		nv50_program_validate(nv50, p);
+		if (!p->translated)
+			assert(0);
+	}
+
+	nv50_program_validate_data(nv50, p);
+	nv50_program_validate_code(nv50, p);
+
+	so = so_new(13, 2);
+	so_method(so, tesla, NV50TCL_VP_ADDRESS_HIGH, 2);
+	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		  NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		  NOUVEAU_BO_LOW, 0, 0);
+	so_method(so, tesla, 0x1650, 2);
+	so_data  (so, p->cfg.vp.attr[0]);
+	so_data  (so, p->cfg.vp.attr[1]);
+	so_method(so, tesla, 0x16b8, 1);
+	so_data  (so, p->cfg.high_result);
+	so_method(so, tesla, 0x16ac, 2);
+	so_data  (so, p->cfg.high_result); //8);
+	so_data  (so, p->cfg.high_temp);
+	so_method(so, tesla, 0x140c, 1);
+	so_data  (so, 0); /* program start offset */
+	so_ref(so, &nv50->state.vertprog);
+}
+
+void
+nv50_fragprog_validate(struct nv50_context *nv50)
+{
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct nv50_program *p = nv50->fragprog;
+	struct nouveau_stateobj *so;
+
+	if (!p->translated) {
+		nv50_program_validate(nv50, p);
+		if (!p->translated)
+			assert(0);
+	}
+
+	nv50_program_validate_data(nv50, p);
+	nv50_program_validate_code(nv50, p);
+
+	so = so_new(64, 2);
+	so_method(so, tesla, NV50TCL_FP_ADDRESS_HIGH, 2);
+	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		  NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, p->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD |
+		  NOUVEAU_BO_LOW, 0, 0);
+	so_method(so, tesla, 0x1904, 4);
+	so_data  (so, 0x00040404); /* p: 0x01000404 */
+	so_data  (so, 0x00000004);
+	so_data  (so, 0x00000000);
+	so_data  (so, 0x00000000);
+	so_method(so, tesla, 0x16bc, 3); /*XXX: fixme */
+	so_data  (so, 0x03020100);
+	so_data  (so, 0x07060504);
+	so_data  (so, 0x0b0a0908);
+	so_method(so, tesla, 0x1988, 2);
+	so_data  (so, 0x08080408); //0x08040404); /* p: 0x0f000401 */
+	so_data  (so, p->cfg.high_temp);
+	so_method(so, tesla, 0x1414, 1);
+	so_data  (so, 0); /* program start offset */
+	so_ref(so, &nv50->state.fragprog);
+}
+
+void
+nv50_program_destroy(struct nv50_context *nv50, struct nv50_program *p)
+{
+	struct pipe_screen *pscreen = nv50->pipe.screen;
+
+	while (p->exec_head) {
+		struct nv50_program_exec *e = p->exec_head;
+
+		p->exec_head = e->next;
+		FREE(e);
+	}
+	p->exec_tail = NULL;
+	p->exec_size = 0;
+
+	if (p->buffer)
+		pipe_buffer_reference(pscreen, &p->buffer, NULL);
+
+	nv50->screen->nvws->res_free(&p->data);
+
+	p->translated = 0;
+}
+
diff --git a/src/gallium/drivers/nv50/nv50_program.h b/src/gallium/drivers/nv50/nv50_program.h
new file mode 100644
index 0000000000..78deed6a38
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_program.h
@@ -0,0 +1,45 @@
+#ifndef __NV50_PROGRAM_H__
+#define __NV50_PROGRAM_H__
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_scan.h"
+
+struct nv50_program_exec {
+	struct nv50_program_exec *next;
+
+	unsigned inst[2];
+	struct {
+		int index;
+		unsigned mask;
+		unsigned shift;
+	} param;
+};
+
+struct nv50_program {
+	struct pipe_shader_state pipe;
+	struct tgsi_shader_info info;
+	boolean translated;
+
+	unsigned type;
+	struct nv50_program_exec *exec_head;
+	struct nv50_program_exec *exec_tail;
+	unsigned exec_size;
+	struct nouveau_resource *data;
+	unsigned data_start;
+
+	struct pipe_buffer *buffer;
+
+	float *immd;
+	unsigned immd_nr;
+	unsigned param_nr;
+
+	struct {
+		unsigned high_temp;
+		unsigned high_result;
+		struct {
+			unsigned attr[2];
+		} vp;
+	} cfg;
+};
+
+#endif
diff --git a/src/gallium/drivers/nv50/nv50_query.c b/src/gallium/drivers/nv50/nv50_query.c
new file mode 100644
index 0000000000..20745ceab8
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_query.c
@@ -0,0 +1,134 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+
+#include "nv50_context.h"
+
+struct nv50_query {
+	struct pipe_buffer *buffer;
+	unsigned type;
+	boolean ready;
+	uint64_t result;
+};
+
+static INLINE struct nv50_query *
+nv50_query(struct pipe_query *pipe)
+{
+	return (struct nv50_query *)pipe;
+}
+
+static struct pipe_query *
+nv50_query_create(struct pipe_context *pipe, unsigned type)
+{
+	struct pipe_winsys *ws = pipe->winsys;
+	struct nv50_query *q = CALLOC_STRUCT(nv50_query);
+
+	assert (q->type == PIPE_QUERY_OCCLUSION_COUNTER);
+	q->type = type;
+
+	q->buffer = ws->buffer_create(ws, 256, 0, 16);
+	if (!q->buffer) {
+		FREE(q);
+		return NULL;
+	}
+
+	return (struct pipe_query *)q;
+}
+
+static void
+nv50_query_destroy(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nv50_query *q = nv50_query(pq);
+
+	if (q) {
+		pipe_buffer_reference(pipe->screen, &q->buffer, NULL);
+		FREE(q);
+	}
+}
+
+static void
+nv50_query_begin(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct nv50_query *q = nv50_query(pq);
+
+	BEGIN_RING(chan, tesla, 0x1530, 1);
+	OUT_RING  (chan, 1);
+	BEGIN_RING(chan, tesla, 0x1514, 1);
+	OUT_RING  (chan, 1);
+
+	q->ready = FALSE;
+}
+
+static void
+nv50_query_end(struct pipe_context *pipe, struct pipe_query *pq)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct nv50_query *q = nv50_query(pq);
+
+	WAIT_RING (chan, 5);
+	BEGIN_RING(chan, tesla, 0x1b00, 4);
+	OUT_RELOCh(chan, q->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	OUT_RELOCl(chan, q->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
+	OUT_RING  (chan, 0x00000000);
+	OUT_RING  (chan, 0x0100f002);
+	FIRE_RING (chan);
+}
+
+static boolean
+nv50_query_result(struct pipe_context *pipe, struct pipe_query *pq,
+		  boolean wait, uint64_t *result)
+{
+	struct pipe_winsys *ws = pipe->winsys;
+	struct nv50_query *q = nv50_query(pq);
+
+	/*XXX: Want to be able to return FALSE here instead of blocking
+	 *     until the result is available..
+	 */
+
+	if (!q->ready) {
+		uint32_t *map = ws->buffer_map(ws, q->buffer,
+					       PIPE_BUFFER_USAGE_CPU_READ);
+		q->result = map[1];
+		q->ready = TRUE;
+		ws->buffer_unmap(ws, q->buffer);
+	}
+
+	*result = q->result;
+	return q->ready;
+}
+
+void
+nv50_init_query_functions(struct nv50_context *nv50)
+{
+	nv50->pipe.create_query = nv50_query_create;
+	nv50->pipe.destroy_query = nv50_query_destroy;
+	nv50->pipe.begin_query = nv50_query_begin;
+	nv50->pipe.end_query = nv50_query_end;
+	nv50->pipe.get_query_result = nv50_query_result;
+}
diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c
new file mode 100644
index 0000000000..58d7a621a8
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_screen.c
@@ -0,0 +1,356 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "pipe/p_screen.h"
+
+#include "util/u_simple_screen.h"
+
+#include "nv50_context.h"
+#include "nv50_screen.h"
+
+#include "nouveau/nouveau_stateobj.h"
+
+#define NV5X_GRCLASS5097_CHIPSETS 0x00000001
+#define NV8X_GRCLASS8297_CHIPSETS 0x00000050
+#define NV9X_GRCLASS8297_CHIPSETS 0x00000014
+
+static boolean
+nv50_screen_is_format_supported(struct pipe_screen *pscreen,
+				enum pipe_format format,
+				enum pipe_texture_target target,
+				unsigned tex_usage, unsigned geom_flags)
+{
+	if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM:
+		case PIPE_FORMAT_Z24S8_UNORM:
+		case PIPE_FORMAT_Z16_UNORM:
+			return TRUE;
+		default:
+			break;
+		}
+	} else {
+		switch (format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+		case PIPE_FORMAT_A1R5G5B5_UNORM:
+		case PIPE_FORMAT_A4R4G4B4_UNORM:
+		case PIPE_FORMAT_R5G6B5_UNORM:
+		case PIPE_FORMAT_L8_UNORM:
+		case PIPE_FORMAT_A8_UNORM:
+		case PIPE_FORMAT_I8_UNORM:
+		case PIPE_FORMAT_A8L8_UNORM:
+		case PIPE_FORMAT_DXT1_RGB:
+		case PIPE_FORMAT_DXT1_RGBA:
+		case PIPE_FORMAT_DXT3_RGBA:
+		case PIPE_FORMAT_DXT5_RGBA:
+			return TRUE;
+		default:
+			break;
+		}
+	}
+
+	return FALSE;
+}
+
+static const char *
+nv50_screen_get_name(struct pipe_screen *pscreen)
+{
+	struct nv50_screen *screen = nv50_screen(pscreen);
+	struct nouveau_device *dev = screen->nvws->channel->device;
+	static char buffer[128];
+
+	snprintf(buffer, sizeof(buffer), "NV%02X", dev->chipset);
+	return buffer;
+}
+
+static const char *
+nv50_screen_get_vendor(struct pipe_screen *pscreen)
+{
+	return "nouveau";
+}
+
+static int
+nv50_screen_get_param(struct pipe_screen *pscreen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+		return 32;
+	case PIPE_CAP_NPOT_TEXTURES:
+		return 1;
+	case PIPE_CAP_TWO_SIDED_STENCIL:
+		return 1;
+	case PIPE_CAP_GLSL:
+		return 0;
+	case PIPE_CAP_S3TC:
+		return 1;
+	case PIPE_CAP_ANISOTROPIC_FILTER:
+		return 1;
+	case PIPE_CAP_POINT_SPRITE:
+		return 0;
+	case PIPE_CAP_MAX_RENDER_TARGETS:
+		return 8;
+	case PIPE_CAP_OCCLUSION_QUERY:
+		return 1;
+	case PIPE_CAP_TEXTURE_SHADOW_MAP:
+		return 1;
+	case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+		return 13;
+	case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+		return 10;
+	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+		return 13;
+	case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
+	case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
+		return 1;
+	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
+		return 0;
+	case NOUVEAU_CAP_HW_VTXBUF:	
+		return 1;
+	case NOUVEAU_CAP_HW_IDXBUF:	
+		return 0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0;
+	}
+}
+
+static float
+nv50_screen_get_paramf(struct pipe_screen *pscreen, int param)
+{
+	switch (param) {
+	case PIPE_CAP_MAX_LINE_WIDTH:
+	case PIPE_CAP_MAX_LINE_WIDTH_AA:
+		return 10.0;
+	case PIPE_CAP_MAX_POINT_WIDTH:
+	case PIPE_CAP_MAX_POINT_WIDTH_AA:
+		return 64.0;
+	case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+		return 16.0;
+	case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+		return 4.0;
+	default:
+		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
+		return 0.0;
+	}
+}
+
+static void
+nv50_screen_destroy(struct pipe_screen *pscreen)
+{
+	FREE(pscreen);
+}
+
+struct pipe_screen *
+nv50_screen_create(struct pipe_winsys *ws, struct nouveau_winsys *nvws)
+{
+	struct nv50_screen *screen = CALLOC_STRUCT(nv50_screen);
+	struct nouveau_stateobj *so;
+	unsigned tesla_class = 0, ret;
+	unsigned chipset = nvws->channel->device->chipset;
+	int i;
+
+	if (!screen)
+		return NULL;
+	screen->nvws = nvws;
+
+	/* 2D object */
+	ret = nvws->grobj_alloc(nvws, NV50_2D, &screen->eng2d);
+	if (ret) {
+		NOUVEAU_ERR("Error creating 2D object: %d\n", ret);
+		nv50_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	/* 3D object */
+	if ((chipset & 0xf0) != 0x50 && (chipset & 0xf0) != 0x80) {
+		NOUVEAU_ERR("Not a G8x chipset\n");
+		nv50_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	switch (chipset & 0xf0) {
+	case 0x50:
+		if (NV5X_GRCLASS5097_CHIPSETS & (1 << (chipset & 0x0f)))
+			tesla_class = 0x5097;
+		break;
+	case 0x80:
+		if (NV8X_GRCLASS8297_CHIPSETS & (1 << (chipset & 0x0f)))
+			tesla_class = 0x8297;
+		break;
+	case 0x90:
+		if (NV9X_GRCLASS8297_CHIPSETS & (1 << (chipset & 0x0f)))
+			tesla_class = 0x8297;
+		break;
+	default:
+		break;
+	}
+
+	if (tesla_class == 0) {
+		NOUVEAU_ERR("Unknown G8x chipset: NV%02x\n", chipset);
+		nv50_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	ret = nvws->grobj_alloc(nvws, tesla_class, &screen->tesla);
+	if (ret) {
+		NOUVEAU_ERR("Error creating 3D object: %d\n", ret);
+		nv50_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	/* Sync notifier */
+	ret = nvws->notifier_alloc(nvws, 1, &screen->sync);
+	if (ret) {
+		NOUVEAU_ERR("Error creating notifier object: %d\n", ret);
+		nv50_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	/* Static 2D init */
+	so = so_new(64, 0);
+	so_method(so, screen->eng2d, NV50_2D_DMA_NOTIFY, 4);
+	so_data  (so, screen->sync->handle);
+	so_data  (so, screen->nvws->channel->vram->handle);
+	so_data  (so, screen->nvws->channel->vram->handle);
+	so_data  (so, screen->nvws->channel->vram->handle);
+	so_method(so, screen->eng2d, NV50_2D_OPERATION, 1);
+	so_data  (so, NV50_2D_OPERATION_SRCCOPY);
+	so_method(so, screen->eng2d, 0x0290, 1);
+	so_data  (so, 0);
+	so_method(so, screen->eng2d, 0x0888, 1);
+	so_data  (so, 1);
+	so_emit(nvws, so);
+	so_ref(NULL, &so);
+
+	/* Static tesla init */
+	so = so_new(256, 20);
+
+	so_method(so, screen->tesla, 0x1558, 1);
+	so_data  (so, 1);
+	so_method(so, screen->tesla, NV50TCL_DMA_NOTIFY, 1);
+	so_data  (so, screen->sync->handle);
+	so_method(so, screen->tesla, NV50TCL_DMA_UNK0(0),
+				     NV50TCL_DMA_UNK0__SIZE);
+	for (i = 0; i < NV50TCL_DMA_UNK0__SIZE; i++)
+		so_data(so, nvws->channel->vram->handle);
+	so_method(so, screen->tesla, NV50TCL_DMA_UNK1(0),
+				     NV50TCL_DMA_UNK1__SIZE);
+	for (i = 0; i < NV50TCL_DMA_UNK1__SIZE; i++)
+		so_data(so, nvws->channel->vram->handle);
+	so_method(so, screen->tesla, 0x121c, 1);
+	so_data  (so, 1);
+
+	so_method(so, screen->tesla, 0x13bc, 1);
+	so_data  (so, 0x54);
+	so_method(so, screen->tesla, 0x13ac, 1);
+	so_data  (so, 1);
+	so_method(so, screen->tesla, 0x16b8, 1);
+	so_data  (so, 8);
+
+	/* Shared constant buffer */
+	screen->constbuf = ws->buffer_create(ws, 0, 0, 128 * 4 * 4);
+	if (nvws->res_init(&screen->vp_data_heap, 0, 128)) {
+		NOUVEAU_ERR("Error initialising constant buffer\n");
+		nv50_screen_destroy(&screen->pipe);
+		return NULL;
+	}
+
+	so_method(so, screen->tesla, 0x1280, 3);
+	so_reloc (so, screen->constbuf, 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, screen->constbuf, 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+	so_data  (so, (NV50_CB_PMISC << 16) | 0x00001000);
+
+	/* Texture sampler/image unit setup - we abuse the constant buffer
+	 * upload mechanism for the moment to upload data to the tex config
+	 * blocks.  At some point we *may* want to go the NVIDIA way of doing
+	 * things?
+	 */
+	screen->tic = ws->buffer_create(ws, 0, 0, 32 * 8 * 4);
+	so_method(so, screen->tesla, 0x1280, 3);
+	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+	so_data  (so, (NV50_CB_TIC << 16) | 0x0800);
+	so_method(so, screen->tesla, 0x1574, 3);
+	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, screen->tic, 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+	so_data  (so, 0x00000800);
+
+	screen->tsc = ws->buffer_create(ws, 0, 0, 32 * 8 * 4);
+	so_method(so, screen->tesla, 0x1280, 3);
+	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+	so_data  (so, (NV50_CB_TSC << 16) | 0x0800);
+	so_method(so, screen->tesla, 0x155c, 3);
+	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
+	so_reloc (so, screen->tsc, 0, NOUVEAU_BO_VRAM |
+		  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+	so_data  (so, 0x00000800);
+
+
+	/* Vertex array limits - max them out */
+	for (i = 0; i < 16; i++) {
+		so_method(so, screen->tesla, 0x1080 + (i * 8), 2);
+		so_data  (so, 0x000000ff);
+		so_data  (so, 0xffffffff);
+	}
+
+	so_method(so, screen->tesla, NV50TCL_DEPTH_RANGE_NEAR, 2);
+	so_data  (so, fui(0.0));
+	so_data  (so, fui(1.0));
+
+	so_method(so, screen->tesla, 0x1234, 1);
+	so_data  (so, 1);
+	so_method(so, screen->tesla, 0x1458, 1);
+	so_data  (so, 1);
+
+	so_emit(nvws, so);
+	so_ref(so, &screen->static_init);
+	nvws->push_flush(nvws, 0, NULL);
+
+	screen->pipe.winsys = ws;
+
+	screen->pipe.destroy = nv50_screen_destroy;
+
+	screen->pipe.get_name = nv50_screen_get_name;
+	screen->pipe.get_vendor = nv50_screen_get_vendor;
+	screen->pipe.get_param = nv50_screen_get_param;
+	screen->pipe.get_paramf = nv50_screen_get_paramf;
+
+	screen->pipe.is_format_supported = nv50_screen_is_format_supported;
+
+	nv50_screen_init_miptree_functions(&screen->pipe);
+	nv50_surface_init_screen_functions(&screen->pipe);
+	u_simple_screen_init(&screen->pipe);
+
+	return &screen->pipe;
+}
+
diff --git a/src/gallium/drivers/nv50/nv50_screen.h b/src/gallium/drivers/nv50/nv50_screen.h
new file mode 100644
index 0000000000..c888ca071c
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_screen.h
@@ -0,0 +1,34 @@
+#ifndef __NV50_SCREEN_H__
+#define __NV50_SCREEN_H__
+
+#include "pipe/p_screen.h"
+
+struct nv50_screen {
+	struct pipe_screen pipe;
+
+	struct nouveau_winsys *nvws;
+
+	unsigned cur_pctx;
+
+	struct nouveau_grobj *tesla;
+	struct nouveau_grobj *eng2d;
+	struct nouveau_notifier *sync;
+
+	struct pipe_buffer *constbuf;
+	struct nouveau_resource *vp_data_heap;
+
+	struct pipe_buffer *tic;
+	struct pipe_buffer *tsc;
+
+	struct nouveau_stateobj *static_init;
+};
+
+static INLINE struct nv50_screen *
+nv50_screen(struct pipe_screen *screen)
+{
+	return (struct nv50_screen *)screen;
+}
+
+void nv50_surface_init_screen_functions(struct pipe_screen *);
+
+#endif
diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c
new file mode 100644
index 0000000000..787ff958ec
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_state.c
@@ -0,0 +1,664 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "tgsi/tgsi_parse.h"
+
+#include "nv50_context.h"
+#include "nv50_texture.h"
+
+#include "nouveau/nouveau_stateobj.h"
+
+static void *
+nv50_blend_state_create(struct pipe_context *pipe,
+			const struct pipe_blend_state *cso)
+{
+	struct nouveau_stateobj *so = so_new(64, 0);
+	struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla;
+	struct nv50_blend_stateobj *bso = CALLOC_STRUCT(nv50_blend_stateobj);
+	unsigned cmask = 0, i;
+
+	/*XXX ignored:
+	 * 	- dither
+	 */
+
+	if (cso->blend_enable == 0) {
+		so_method(so, tesla, NV50TCL_BLEND_ENABLE(0), 8);
+		for (i = 0; i < 8; i++)
+			so_data(so, 0);
+	} else {
+		so_method(so, tesla, NV50TCL_BLEND_ENABLE(0), 8);
+		for (i = 0; i < 8; i++)
+			so_data(so, 1);
+		so_method(so, tesla, NV50TCL_BLEND_EQUATION_RGB, 5);
+		so_data  (so, nvgl_blend_eqn(cso->rgb_func));
+		so_data  (so, 0x4000 | nvgl_blend_func(cso->rgb_src_factor));
+		so_data  (so, 0x4000 | nvgl_blend_func(cso->rgb_dst_factor));
+		so_data  (so, nvgl_blend_eqn(cso->alpha_func));
+		so_data  (so, 0x4000 | nvgl_blend_func(cso->alpha_src_factor));
+		so_method(so, tesla, NV50TCL_BLEND_FUNC_DST_ALPHA, 1);
+		so_data  (so, 0x4000 | nvgl_blend_func(cso->alpha_dst_factor));
+	}
+
+	if (cso->logicop_enable == 0 ) {
+		so_method(so, tesla, NV50TCL_LOGIC_OP_ENABLE, 1);
+		so_data  (so, 0);
+	} else {
+		so_method(so, tesla, NV50TCL_LOGIC_OP_ENABLE, 2);
+		so_data  (so, 1);
+		so_data  (so, nvgl_logicop_func(cso->logicop_func));
+	}
+
+	if (cso->colormask & PIPE_MASK_R)
+		cmask |= (1 << 0);
+	if (cso->colormask & PIPE_MASK_G)
+		cmask |= (1 << 4);
+	if (cso->colormask & PIPE_MASK_B)
+		cmask |= (1 << 8);
+	if (cso->colormask & PIPE_MASK_A)
+		cmask |= (1 << 12);
+	so_method(so, tesla, NV50TCL_COLOR_MASK(0), 8);
+	for (i = 0; i < 8; i++)
+		so_data(so, cmask);
+
+	bso->pipe = *cso;
+	so_ref(so, &bso->so);
+	return (void *)bso;
+}
+
+static void
+nv50_blend_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	nv50->blend = hwcso;
+	nv50->dirty |= NV50_NEW_BLEND;
+}
+
+static void
+nv50_blend_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv50_blend_stateobj *bso = hwcso;
+
+	so_ref(NULL, &bso->so);
+	FREE(bso);
+}
+
+static INLINE unsigned
+wrap_mode(unsigned wrap)
+{
+	switch (wrap) {
+	case PIPE_TEX_WRAP_REPEAT:
+		return NV50TSC_1_0_WRAPS_REPEAT;
+	case PIPE_TEX_WRAP_MIRROR_REPEAT:
+		return NV50TSC_1_0_WRAPS_MIRROR_REPEAT;
+	case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+		return NV50TSC_1_0_WRAPS_CLAMP_TO_EDGE;
+	case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+		return NV50TSC_1_0_WRAPS_CLAMP_TO_BORDER;
+	case PIPE_TEX_WRAP_CLAMP:
+		return NV50TSC_1_0_WRAPS_CLAMP;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+		return NV50TSC_1_0_WRAPS_MIRROR_CLAMP_TO_EDGE;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+		return NV50TSC_1_0_WRAPS_MIRROR_CLAMP_TO_BORDER;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP:
+		return NV50TSC_1_0_WRAPS_MIRROR_CLAMP;
+	default:
+		NOUVEAU_ERR("unknown wrap mode: %d\n", wrap);
+		return NV50TSC_1_0_WRAPS_REPEAT;
+	}
+}
+static void *
+nv50_sampler_state_create(struct pipe_context *pipe,
+			  const struct pipe_sampler_state *cso)
+{
+	unsigned *tsc = CALLOC(8, sizeof(unsigned));
+
+	tsc[0] = (0x00024000 |
+		  (wrap_mode(cso->wrap_s) << 0) |
+		  (wrap_mode(cso->wrap_t) << 3) |
+		  (wrap_mode(cso->wrap_r) << 6));
+
+	switch (cso->mag_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		tsc[1] |= NV50TSC_1_1_MAGF_LINEAR;
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		tsc[1] |= NV50TSC_1_1_MAGF_NEAREST;
+		break;
+	}
+
+	switch (cso->min_img_filter) {
+	case PIPE_TEX_FILTER_LINEAR:
+		tsc[1] |= NV50TSC_1_1_MINF_LINEAR;
+		break;
+	case PIPE_TEX_FILTER_NEAREST:
+	default:
+		tsc[1] |= NV50TSC_1_1_MINF_NEAREST;
+		break;
+	}
+
+	switch (cso->min_mip_filter) {
+	case PIPE_TEX_MIPFILTER_LINEAR:
+		tsc[1] |= NV50TSC_1_1_MIPF_LINEAR;
+		break;
+	case PIPE_TEX_MIPFILTER_NEAREST:
+		tsc[1] |= NV50TSC_1_1_MIPF_NEAREST;
+		break;
+	case PIPE_TEX_MIPFILTER_NONE:
+	default:
+		tsc[1] |= NV50TSC_1_1_MIPF_NONE;
+		break;
+	}
+
+	if (cso->max_anisotropy >= 16.0)
+		tsc[0] |= (7 << 20);
+	else
+	if (cso->max_anisotropy >= 12.0)
+		tsc[0] |= (6 << 20);
+	else
+	if (cso->max_anisotropy >= 10.0)
+		tsc[0] |= (5 << 20);
+	else
+	if (cso->max_anisotropy >= 8.0)
+		tsc[0] |= (4 << 20);
+	else
+	if (cso->max_anisotropy >= 6.0)
+		tsc[0] |= (3 << 20);
+	else
+	if (cso->max_anisotropy >= 4.0)
+		tsc[0] |= (2 << 20);
+	else
+	if (cso->max_anisotropy >= 2.0)
+		tsc[0] |= (1 << 20);
+
+	if (cso->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+		tsc[0] |= (1 << 8);
+		tsc[0] |= (nvgl_comparison_op(cso->compare_func) & 0x7);
+	}
+
+	return (void *)tsc;
+}
+
+static void
+nv50_sampler_state_bind(struct pipe_context *pipe, unsigned nr, void **sampler)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+	int i;
+
+	nv50->sampler_nr = nr;
+	for (i = 0; i < nv50->sampler_nr; i++)
+		nv50->sampler[i] = sampler[i];
+
+	nv50->dirty |= NV50_NEW_SAMPLER;
+}
+
+static void
+nv50_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	FREE(hwcso);
+}
+
+static void
+nv50_set_sampler_texture(struct pipe_context *pipe, unsigned nr,
+			 struct pipe_texture **pt)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+	int i;
+
+	for (i = 0; i < nr; i++)
+		pipe_texture_reference((void *)&nv50->miptree[i], pt[i]);
+	for (i = nr; i < nv50->miptree_nr; i++)
+		pipe_texture_reference((void *)&nv50->miptree[i], NULL);
+
+	nv50->miptree_nr = nr;
+	nv50->dirty |= NV50_NEW_TEXTURE;
+}
+
+static void *
+nv50_rasterizer_state_create(struct pipe_context *pipe,
+			     const struct pipe_rasterizer_state *cso)
+{
+	struct nouveau_stateobj *so = so_new(64, 0);
+	struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla;
+	struct nv50_rasterizer_stateobj *rso =
+		CALLOC_STRUCT(nv50_rasterizer_stateobj);
+
+	/*XXX: ignored
+	 * 	- light_twosize
+	 * 	- point_smooth
+	 * 	- multisample
+	 * 	- point_sprite / sprite_coord_mode
+	 */
+
+	so_method(so, tesla, NV50TCL_SHADE_MODEL, 1);
+	so_data  (so, cso->flatshade ? NV50TCL_SHADE_MODEL_FLAT :
+				       NV50TCL_SHADE_MODEL_SMOOTH);
+
+	so_method(so, tesla, NV50TCL_LINE_WIDTH, 1);
+	so_data  (so, fui(cso->line_width));
+	so_method(so, tesla, NV50TCL_LINE_SMOOTH_ENABLE, 1);
+	so_data  (so, cso->line_smooth ? 1 : 0);
+	if (cso->line_stipple_enable) {
+		so_method(so, tesla, NV50TCL_LINE_STIPPLE_ENABLE, 1);
+		so_data  (so, 1);
+		so_method(so, tesla, NV50TCL_LINE_STIPPLE_PATTERN, 1);
+		so_data  (so, (cso->line_stipple_pattern << 8) |
+			       cso->line_stipple_factor);
+	} else {
+		so_method(so, tesla, NV50TCL_LINE_STIPPLE_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	so_method(so, tesla, NV50TCL_POINT_SIZE, 1);
+	so_data  (so, fui(cso->point_size));
+
+	so_method(so, tesla, NV50TCL_POLYGON_MODE_FRONT, 3);
+	if (cso->front_winding == PIPE_WINDING_CCW) {
+		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
+		so_data(so, nvgl_polygon_mode(cso->fill_cw));
+	} else {
+		so_data(so, nvgl_polygon_mode(cso->fill_cw));
+		so_data(so, nvgl_polygon_mode(cso->fill_ccw));
+	}
+	so_data(so, cso->poly_smooth ? 1 : 0);
+
+	so_method(so, tesla, NV50TCL_CULL_FACE_ENABLE, 3);
+	so_data  (so, cso->cull_mode != PIPE_WINDING_NONE);
+	if (cso->front_winding == PIPE_WINDING_CCW) {
+		so_data(so, NV50TCL_FRONT_FACE_CCW);
+		switch (cso->cull_mode) {
+		case PIPE_WINDING_CCW:
+			so_data(so, NV50TCL_CULL_FACE_FRONT);
+			break;
+		case PIPE_WINDING_CW:
+			so_data(so, NV50TCL_CULL_FACE_BACK);
+			break;
+		case PIPE_WINDING_BOTH:
+			so_data(so, NV50TCL_CULL_FACE_FRONT_AND_BACK);
+			break;
+		default:
+			so_data(so, NV50TCL_CULL_FACE_BACK);
+			break;
+		}
+	} else {
+		so_data(so, NV50TCL_FRONT_FACE_CW);
+		switch (cso->cull_mode) {
+		case PIPE_WINDING_CCW:
+			so_data(so, NV50TCL_CULL_FACE_BACK);
+			break;
+		case PIPE_WINDING_CW:
+			so_data(so, NV50TCL_CULL_FACE_FRONT);
+			break;
+		case PIPE_WINDING_BOTH:
+			so_data(so, NV50TCL_CULL_FACE_FRONT_AND_BACK);
+			break;
+		default:
+			so_data(so, NV50TCL_CULL_FACE_BACK);
+			break;
+		}
+	}
+
+	so_method(so, tesla, NV50TCL_POLYGON_STIPPLE_ENABLE, 1);
+	so_data  (so, cso->poly_stipple_enable ? 1 : 0);
+
+	so_method(so, tesla, NV50TCL_POLYGON_OFFSET_POINT_ENABLE, 3);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_POINT) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_POINT))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_LINE) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_LINE))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+	if ((cso->offset_cw && cso->fill_cw == PIPE_POLYGON_MODE_FILL) ||
+	    (cso->offset_ccw && cso->fill_ccw == PIPE_POLYGON_MODE_FILL))
+		so_data(so, 1);
+	else
+		so_data(so, 0);
+
+	if (cso->offset_cw || cso->offset_ccw) {
+		so_method(so, tesla, NV50TCL_POLYGON_OFFSET_FACTOR, 1);
+		so_data  (so, fui(cso->offset_scale));
+		so_method(so, tesla, NV50TCL_POLYGON_OFFSET_UNITS, 1);
+		so_data  (so, fui(cso->offset_units));
+	}
+
+	rso->pipe = *cso;
+	so_ref(so, &rso->so);
+	return (void *)rso;
+}
+
+static void
+nv50_rasterizer_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	nv50->rasterizer = hwcso;
+	nv50->dirty |= NV50_NEW_RASTERIZER;
+}
+
+static void
+nv50_rasterizer_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv50_rasterizer_stateobj *rso = hwcso;
+
+	so_ref(NULL, &rso->so);
+	FREE(rso);
+}
+
+static void *
+nv50_depth_stencil_alpha_state_create(struct pipe_context *pipe,
+			const struct pipe_depth_stencil_alpha_state *cso)
+{
+	struct nouveau_grobj *tesla = nv50_context(pipe)->screen->tesla;
+	struct nv50_zsa_stateobj *zsa = CALLOC_STRUCT(nv50_zsa_stateobj);
+	struct nouveau_stateobj *so = so_new(64, 0);
+
+	so_method(so, tesla, NV50TCL_DEPTH_WRITE_ENABLE, 1);
+	so_data  (so, cso->depth.writemask ? 1 : 0);
+	if (cso->depth.enabled) {
+		so_method(so, tesla, NV50TCL_DEPTH_TEST_ENABLE, 1);
+		so_data  (so, 1);
+		so_method(so, tesla, NV50TCL_DEPTH_TEST_FUNC, 1);
+		so_data  (so, nvgl_comparison_op(cso->depth.func));
+	} else {
+		so_method(so, tesla, NV50TCL_DEPTH_TEST_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	/*XXX: yes, I know they're backwards.. header needs fixing */
+	if (cso->stencil[0].enabled) {
+		so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 5);
+		so_data  (so, 1);
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].fail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].zfail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[0].zpass_op));
+		so_data  (so, nvgl_comparison_op(cso->stencil[0].func));
+		so_method(so, tesla, NV50TCL_STENCIL_BACK_FUNC_REF, 3);
+		so_data  (so, cso->stencil[0].ref_value);
+		so_data  (so, cso->stencil[0].writemask);
+		so_data  (so, cso->stencil[0].valuemask);
+	} else {
+		so_method(so, tesla, NV50TCL_STENCIL_BACK_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	if (cso->stencil[1].enabled) {
+		so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 8);
+		so_data  (so, 1);
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].fail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].zfail_op));
+		so_data  (so, nvgl_stencil_op(cso->stencil[1].zpass_op));
+		so_data  (so, nvgl_comparison_op(cso->stencil[1].func));
+		so_data  (so, cso->stencil[1].ref_value);
+		so_data  (so, cso->stencil[1].writemask);
+		so_data  (so, cso->stencil[1].valuemask);
+	} else {
+		so_method(so, tesla, NV50TCL_STENCIL_FRONT_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	if (cso->alpha.enabled) {
+		so_method(so, tesla, NV50TCL_ALPHA_TEST_ENABLE, 1);
+		so_data  (so, 1);
+		so_method(so, tesla, NV50TCL_ALPHA_TEST_REF, 2);
+		so_data  (so, fui(cso->alpha.ref_value));
+		so_data  (so, nvgl_comparison_op(cso->alpha.func));
+	} else {
+		so_method(so, tesla, NV50TCL_ALPHA_TEST_ENABLE, 1);
+		so_data  (so, 0);
+	}
+
+	zsa->pipe = *cso;
+	so_ref(so, &zsa->so);
+	return (void *)zsa;
+}
+
+static void
+nv50_depth_stencil_alpha_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	nv50->zsa = hwcso;
+	nv50->dirty |= NV50_NEW_ZSA;
+}
+
+static void
+nv50_depth_stencil_alpha_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv50_zsa_stateobj *zsa = hwcso;
+
+	so_ref(NULL, &zsa->so);
+	FREE(zsa);
+}
+
+static void *
+nv50_vp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	struct nv50_program *p = CALLOC_STRUCT(nv50_program);
+
+	p->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+	p->type = PIPE_SHADER_VERTEX;
+	tgsi_scan_shader(p->pipe.tokens, &p->info);
+	return (void *)p;
+}
+
+static void
+nv50_vp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	nv50->vertprog = hwcso;
+	nv50->dirty |= NV50_NEW_VERTPROG;
+}
+
+static void
+nv50_vp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+	struct nv50_program *p = hwcso;
+
+	nv50_program_destroy(nv50, p);
+	FREE((void*)p->pipe.tokens);
+	FREE(p);
+}
+
+static void *
+nv50_fp_state_create(struct pipe_context *pipe,
+		     const struct pipe_shader_state *cso)
+{
+	struct nv50_program *p = CALLOC_STRUCT(nv50_program);
+
+	p->pipe.tokens = tgsi_dup_tokens(cso->tokens);
+	p->type = PIPE_SHADER_FRAGMENT;
+	tgsi_scan_shader(p->pipe.tokens, &p->info);
+	return (void *)p;
+}
+
+static void
+nv50_fp_state_bind(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	nv50->fragprog = hwcso;
+	nv50->dirty |= NV50_NEW_FRAGPROG;
+}
+
+static void
+nv50_fp_state_delete(struct pipe_context *pipe, void *hwcso)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+	struct nv50_program *p = hwcso;
+
+	nv50_program_destroy(nv50, p);
+	FREE((void*)p->pipe.tokens);
+	FREE(p);
+}
+
+static void
+nv50_set_blend_color(struct pipe_context *pipe,
+		     const struct pipe_blend_color *bcol)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	nv50->blend_colour = *bcol;
+	nv50->dirty |= NV50_NEW_BLEND_COLOUR;
+}
+
+static void
+nv50_set_clip_state(struct pipe_context *pipe,
+		    const struct pipe_clip_state *clip)
+{
+}
+
+static void
+nv50_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
+			 const struct pipe_constant_buffer *buf )
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	if (shader == PIPE_SHADER_VERTEX) {
+		nv50->constbuf[PIPE_SHADER_VERTEX] = buf->buffer;
+		nv50->dirty |= NV50_NEW_VERTPROG_CB;
+	} else
+	if (shader == PIPE_SHADER_FRAGMENT) {
+		nv50->constbuf[PIPE_SHADER_FRAGMENT] = buf->buffer;
+		nv50->dirty |= NV50_NEW_FRAGPROG_CB;
+	}
+}
+
+static void
+nv50_set_framebuffer_state(struct pipe_context *pipe,
+			   const struct pipe_framebuffer_state *fb)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	nv50->framebuffer = *fb;
+	nv50->dirty |= NV50_NEW_FRAMEBUFFER;
+}
+
+static void
+nv50_set_polygon_stipple(struct pipe_context *pipe,
+			 const struct pipe_poly_stipple *stipple)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	nv50->stipple = *stipple;
+	nv50->dirty |= NV50_NEW_STIPPLE;
+}
+
+static void
+nv50_set_scissor_state(struct pipe_context *pipe,
+		       const struct pipe_scissor_state *s)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	nv50->scissor = *s;
+	nv50->dirty |= NV50_NEW_SCISSOR;
+}
+
+static void
+nv50_set_viewport_state(struct pipe_context *pipe,
+			const struct pipe_viewport_state *vpt)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	nv50->viewport = *vpt;
+	nv50->dirty |= NV50_NEW_VIEWPORT;
+}
+
+static void
+nv50_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
+			const struct pipe_vertex_buffer *vb)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	memcpy(nv50->vtxbuf, vb, sizeof(*vb) * count);
+	nv50->vtxbuf_nr = count;
+
+	nv50->dirty |= NV50_NEW_ARRAYS;
+}
+
+static void
+nv50_set_vertex_elements(struct pipe_context *pipe, unsigned count,
+			 const struct pipe_vertex_element *ve)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	memcpy(nv50->vtxelt, ve, sizeof(*ve) * count);
+	nv50->vtxelt_nr = count;
+
+	nv50->dirty |= NV50_NEW_ARRAYS;
+}
+
+void
+nv50_init_state_functions(struct nv50_context *nv50)
+{
+	nv50->pipe.create_blend_state = nv50_blend_state_create;
+	nv50->pipe.bind_blend_state = nv50_blend_state_bind;
+	nv50->pipe.delete_blend_state = nv50_blend_state_delete;
+
+	nv50->pipe.create_sampler_state = nv50_sampler_state_create;
+	nv50->pipe.bind_sampler_states = nv50_sampler_state_bind;
+	nv50->pipe.delete_sampler_state = nv50_sampler_state_delete;
+	nv50->pipe.set_sampler_textures = nv50_set_sampler_texture;
+
+	nv50->pipe.create_rasterizer_state = nv50_rasterizer_state_create;
+	nv50->pipe.bind_rasterizer_state = nv50_rasterizer_state_bind;
+	nv50->pipe.delete_rasterizer_state = nv50_rasterizer_state_delete;
+
+	nv50->pipe.create_depth_stencil_alpha_state =
+		nv50_depth_stencil_alpha_state_create;
+	nv50->pipe.bind_depth_stencil_alpha_state =
+		nv50_depth_stencil_alpha_state_bind;
+	nv50->pipe.delete_depth_stencil_alpha_state =
+		nv50_depth_stencil_alpha_state_delete;
+
+	nv50->pipe.create_vs_state = nv50_vp_state_create;
+	nv50->pipe.bind_vs_state = nv50_vp_state_bind;
+	nv50->pipe.delete_vs_state = nv50_vp_state_delete;
+
+	nv50->pipe.create_fs_state = nv50_fp_state_create;
+	nv50->pipe.bind_fs_state = nv50_fp_state_bind;
+	nv50->pipe.delete_fs_state = nv50_fp_state_delete;
+
+	nv50->pipe.set_blend_color = nv50_set_blend_color;
+	nv50->pipe.set_clip_state = nv50_set_clip_state;
+	nv50->pipe.set_constant_buffer = nv50_set_constant_buffer;
+	nv50->pipe.set_framebuffer_state = nv50_set_framebuffer_state;
+	nv50->pipe.set_polygon_stipple = nv50_set_polygon_stipple;
+	nv50->pipe.set_scissor_state = nv50_set_scissor_state;
+	nv50->pipe.set_viewport_state = nv50_set_viewport_state;
+
+	nv50->pipe.set_vertex_buffers = nv50_set_vertex_buffers;
+	nv50->pipe.set_vertex_elements = nv50_set_vertex_elements;
+}
+
diff --git a/src/gallium/drivers/nv50/nv50_state_validate.c b/src/gallium/drivers/nv50/nv50_state_validate.c
new file mode 100644
index 0000000000..948112ffa9
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_state_validate.c
@@ -0,0 +1,313 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "nv50_context.h"
+#include "nouveau/nouveau_stateobj.h"
+
+static void
+nv50_state_validate_fb(struct nv50_context *nv50)
+{
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct nouveau_stateobj *so = so_new(128, 18);
+	struct pipe_framebuffer_state *fb = &nv50->framebuffer;
+	unsigned i, w, h, gw = 0;
+
+	for (i = 0; i < fb->nr_cbufs; i++) {
+		if (!gw) {
+			w = fb->cbufs[i]->width;
+			h = fb->cbufs[i]->height;
+			gw = 1;
+		} else {
+			assert(w == fb->cbufs[i]->width);
+			assert(h == fb->cbufs[i]->height);
+		}
+
+		so_method(so, tesla, NV50TCL_RT_HORIZ(i), 2);
+		so_data  (so, fb->cbufs[i]->width);
+		so_data  (so, fb->cbufs[i]->height);
+
+		so_method(so, tesla, NV50TCL_RT_ADDRESS_HIGH(i), 5);
+		so_reloc (so, nv50_surface_buffer(fb->cbufs[i]), fb->cbufs[i]->offset,
+			  NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH |
+			  NOUVEAU_BO_RDWR, 0, 0);
+		so_reloc (so, nv50_surface_buffer(fb->cbufs[i]), fb->cbufs[i]->offset,
+			  NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
+			  NOUVEAU_BO_RDWR, 0, 0);
+		switch (fb->cbufs[i]->format) {
+		case PIPE_FORMAT_A8R8G8B8_UNORM:
+			so_data(so, 0xcf);
+			break;
+		case PIPE_FORMAT_R5G6B5_UNORM:
+			so_data(so, 0xe8);
+			break;
+		default:
+			NOUVEAU_ERR("AIIII unknown format %s\n",
+				    pf_name(fb->cbufs[i]->format));
+			so_data(so, 0xe6);
+			break;
+		}
+		so_data(so, 0x00000000);
+		so_data(so, 0x00000000);
+
+		so_method(so, tesla, 0x1224, 1);
+		so_data  (so, 1);
+	}
+
+	if (fb->zsbuf) {
+		if (!gw) {
+			w = fb->zsbuf->width;
+			h = fb->zsbuf->height;
+			gw = 1;
+		} else {
+			assert(w == fb->zsbuf->width);
+			assert(h == fb->zsbuf->height);
+		}
+
+		so_method(so, tesla, NV50TCL_ZETA_ADDRESS_HIGH, 5);
+		so_reloc (so, nv50_surface_buffer(fb->zsbuf), fb->zsbuf->offset,
+			  NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH |
+			  NOUVEAU_BO_RDWR, 0, 0);
+		so_reloc (so, nv50_surface_buffer(fb->zsbuf), fb->zsbuf->offset,
+			  NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
+			  NOUVEAU_BO_RDWR, 0, 0);
+		switch (fb->zsbuf->format) {
+		case PIPE_FORMAT_Z24S8_UNORM:
+			so_data(so, 0x16);
+			break;
+		case PIPE_FORMAT_Z16_UNORM:
+			so_data(so, 0x15);
+			break;
+		default:
+			NOUVEAU_ERR("AIIII unknown format %s\n",
+				    pf_name(fb->zsbuf->format));
+			so_data(so, 0x16);
+			break;
+		}
+		so_data(so, 0x00000000);
+		so_data(so, 0x00000000);
+
+		so_method(so, tesla, 0x1538, 1);
+		so_data  (so, 1);
+		so_method(so, tesla, 0x1228, 3);
+		so_data  (so, fb->zsbuf->width);
+		so_data  (so, fb->zsbuf->height);
+		so_data  (so, 0x00010001);
+	}
+
+	so_method(so, tesla, NV50TCL_VIEWPORT_HORIZ, 2);
+	so_data  (so, w << 16);
+	so_data  (so, h << 16);
+	so_method(so, tesla, 0x0e04, 2);
+	so_data  (so, w << 16);
+	so_data  (so, h << 16);
+	so_method(so, tesla, 0xdf8, 2);
+	so_data  (so, 0);
+	so_data  (so, h);
+
+	so_ref(so, &nv50->state.fb);
+}
+
+static void
+nv50_state_emit(struct nv50_context *nv50)
+{
+	struct nv50_screen *screen = nv50->screen;
+	struct nouveau_winsys *nvws = screen->nvws;
+
+	if (nv50->pctx_id != screen->cur_pctx) {
+		nv50->state.dirty |= 0xffffffff;
+		screen->cur_pctx = nv50->pctx_id;
+	}
+
+	if (nv50->state.dirty & NV50_NEW_FRAMEBUFFER)
+		so_emit(nvws, nv50->state.fb);
+	if (nv50->state.dirty & NV50_NEW_BLEND)
+		so_emit(nvws, nv50->state.blend);
+	if (nv50->state.dirty & NV50_NEW_ZSA)
+		so_emit(nvws, nv50->state.zsa);
+	if (nv50->state.dirty & NV50_NEW_VERTPROG)
+		so_emit(nvws, nv50->state.vertprog);
+	if (nv50->state.dirty & NV50_NEW_FRAGPROG)
+		so_emit(nvws, nv50->state.fragprog);
+	if (nv50->state.dirty & NV50_NEW_RASTERIZER)
+		so_emit(nvws, nv50->state.rast);
+	if (nv50->state.dirty & NV50_NEW_BLEND_COLOUR)
+		so_emit(nvws, nv50->state.blend_colour);
+	if (nv50->state.dirty & NV50_NEW_STIPPLE)
+		so_emit(nvws, nv50->state.stipple);
+	if (nv50->state.dirty & NV50_NEW_SCISSOR)
+		so_emit(nvws, nv50->state.scissor);
+	if (nv50->state.dirty & NV50_NEW_VIEWPORT)
+		so_emit(nvws, nv50->state.viewport);
+	if (nv50->state.dirty & NV50_NEW_SAMPLER)
+		so_emit(nvws, nv50->state.tsc_upload);
+	if (nv50->state.dirty & NV50_NEW_TEXTURE)
+		so_emit(nvws, nv50->state.tic_upload);
+	if (nv50->state.dirty & NV50_NEW_ARRAYS) {
+		so_emit(nvws, nv50->state.vtxfmt);
+		so_emit(nvws, nv50->state.vtxbuf);
+	}
+	nv50->state.dirty = 0;
+
+	so_emit_reloc_markers(nvws, nv50->state.fb);
+	so_emit_reloc_markers(nvws, nv50->state.vertprog);
+	so_emit_reloc_markers(nvws, nv50->state.fragprog);
+	so_emit_reloc_markers(nvws, nv50->state.vtxbuf);
+	so_emit_reloc_markers(nvws, nv50->screen->static_init);
+}
+
+boolean
+nv50_state_validate(struct nv50_context *nv50)
+{
+	const struct pipe_framebuffer_state *fb = &nv50->framebuffer;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct nouveau_stateobj *so;
+	unsigned i;
+
+	for (i = 0; i < fb->nr_cbufs; i++)
+		fb->cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED;
+
+	if (fb->zsbuf)
+		fb->zsbuf->status = PIPE_SURFACE_STATUS_DEFINED;
+
+	if (nv50->dirty & NV50_NEW_FRAMEBUFFER)
+		nv50_state_validate_fb(nv50);
+
+	if (nv50->dirty & NV50_NEW_BLEND)
+		so_ref(nv50->blend->so, &nv50->state.blend);
+
+	if (nv50->dirty & NV50_NEW_ZSA)
+		so_ref(nv50->zsa->so, &nv50->state.zsa);
+
+	if (nv50->dirty & (NV50_NEW_VERTPROG | NV50_NEW_VERTPROG_CB))
+		nv50_vertprog_validate(nv50);
+
+	if (nv50->dirty & (NV50_NEW_FRAGPROG | NV50_NEW_FRAGPROG_CB))
+		nv50_fragprog_validate(nv50);
+
+	if (nv50->dirty & NV50_NEW_RASTERIZER)
+		so_ref(nv50->rasterizer->so, &nv50->state.rast);
+
+	if (nv50->dirty & NV50_NEW_BLEND_COLOUR) {
+		so = so_new(5, 0);
+		so_method(so, tesla, NV50TCL_BLEND_COLOR(0), 4);
+		so_data  (so, fui(nv50->blend_colour.color[0]));
+		so_data  (so, fui(nv50->blend_colour.color[1]));
+		so_data  (so, fui(nv50->blend_colour.color[2]));
+		so_data  (so, fui(nv50->blend_colour.color[3]));
+		so_ref(so, &nv50->state.blend_colour);
+	}
+
+	if (nv50->dirty & NV50_NEW_STIPPLE) {
+		so = so_new(33, 0);
+		so_method(so, tesla, NV50TCL_POLYGON_STIPPLE_PATTERN(0), 32);
+		for (i = 0; i < 32; i++)
+			so_data(so, nv50->stipple.stipple[i]);
+		so_ref(so, &nv50->state.stipple);
+	}
+
+	if (nv50->dirty & (NV50_NEW_SCISSOR | NV50_NEW_RASTERIZER)) {
+		struct pipe_rasterizer_state *rast = &nv50->rasterizer->pipe;
+		struct pipe_scissor_state *s = &nv50->scissor;
+
+		if (nv50->state.scissor &&
+		    (rast->scissor == 0 && nv50->state.scissor_enabled == 0))
+			goto scissor_uptodate;
+		nv50->state.scissor_enabled = rast->scissor;
+
+		so = so_new(3, 0);
+		so_method(so, tesla, 0x0ff4, 2);
+		if (nv50->state.scissor_enabled) {
+			so_data(so, ((s->maxx - s->minx) << 16) | s->minx);
+			so_data(so, ((s->maxy - s->miny) << 16) | s->miny);
+		} else {
+			so_data(so, (8192 << 16));
+			so_data(so, (8192 << 16));
+		}
+		so_ref(so, &nv50->state.scissor);
+		nv50->state.dirty |= NV50_NEW_SCISSOR;
+	}
+scissor_uptodate:
+
+	if (nv50->dirty & NV50_NEW_VIEWPORT) {
+		unsigned bypass;
+
+		if (!nv50->rasterizer->pipe.bypass_clipping)
+			bypass = 0;
+		else
+			bypass = 1;
+
+		if (nv50->state.viewport &&
+		    (bypass || !(nv50->dirty & NV50_NEW_VIEWPORT)) &&
+		    nv50->state.viewport_bypass == bypass)
+			goto viewport_uptodate;
+		nv50->state.viewport_bypass = bypass;
+
+		so = so_new(12, 0);
+		if (!bypass) {
+			so_method(so, tesla, NV50TCL_VIEWPORT_UNK1(0), 3);
+			so_data  (so, fui(nv50->viewport.translate[0]));
+			so_data  (so, fui(nv50->viewport.translate[1]));
+			so_data  (so, fui(nv50->viewport.translate[2]));
+			so_method(so, tesla, NV50TCL_VIEWPORT_UNK0(0), 3);
+			so_data  (so, fui(nv50->viewport.scale[0]));
+			so_data  (so, fui(-nv50->viewport.scale[1]));
+			so_data  (so, fui(nv50->viewport.scale[2]));
+			so_method(so, tesla, 0x192c, 1);
+			so_data  (so, 1);
+			so_method(so, tesla, 0x0f90, 1);
+			so_data  (so, 0);
+		} else {
+			so_method(so, tesla, 0x192c, 1);
+			so_data  (so, 0);
+			so_method(so, tesla, 0x0f90, 1);
+			so_data  (so, 1);
+		}
+
+		so_ref(so, &nv50->state.viewport);
+	}
+viewport_uptodate:
+
+	if (nv50->dirty & NV50_NEW_SAMPLER) {
+		int i;
+
+		so = so_new(nv50->sampler_nr * 8 + 3, 0);
+		so_method(so, tesla, 0x0f00, 1);
+		so_data  (so, NV50_CB_TSC);
+		so_method(so, tesla, 0x40000f04, nv50->sampler_nr * 8);
+		for (i = 0; i < nv50->sampler_nr; i++)
+			so_datap (so, nv50->sampler[i], 8);
+		so_ref(so, &nv50->state.tsc_upload);
+	}
+
+	if (nv50->dirty & NV50_NEW_TEXTURE)
+		nv50_tex_validate(nv50);
+
+	if (nv50->dirty & NV50_NEW_ARRAYS)
+		nv50_vbo_validate(nv50);
+
+	nv50->state.dirty |= nv50->dirty;
+	nv50->dirty = 0;
+	nv50_state_emit(nv50);
+
+	return TRUE;
+}
+
diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c
new file mode 100644
index 0000000000..f2dd2eb30b
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_surface.c
@@ -0,0 +1,230 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#define __NOUVEAU_PUSH_H__
+#include <stdint.h>
+#include "nouveau/nouveau_pushbuf.h"
+#include "nv50_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+
+#include "util/u_tile.h"
+
+static INLINE int
+nv50_format(enum pipe_format format)
+{
+	switch (format) {
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case PIPE_FORMAT_Z24S8_UNORM:
+		return NV50_2D_DST_FORMAT_32BPP;
+	case PIPE_FORMAT_X8R8G8B8_UNORM:
+		return NV50_2D_DST_FORMAT_24BPP;
+	case PIPE_FORMAT_R5G6B5_UNORM:
+		return NV50_2D_DST_FORMAT_16BPP;
+	case PIPE_FORMAT_A8_UNORM:
+		return NV50_2D_DST_FORMAT_8BPP;
+	default:
+		return -1;
+	}
+}
+
+static int
+nv50_surface_set(struct nv50_screen *screen, struct pipe_surface *ps, int dst)
+{
+	struct nouveau_channel *chan = screen->nvws->channel;
+	struct nouveau_grobj *eng2d = screen->eng2d;
+	struct nouveau_bo *bo;
+ 	int format, mthd = dst ? NV50_2D_DST_FORMAT : NV50_2D_SRC_FORMAT;
+ 	int flags = NOUVEAU_BO_VRAM | (dst ? NOUVEAU_BO_WR : NOUVEAU_BO_RD);
+ 
+	bo = screen->nvws->get_bo(nv50_miptree(ps->texture)->buffer);
+	if (!bo)
+		return 1;
+
+ 	format = nv50_format(ps->format);
+ 	if (format < 0)
+ 		return 1;
+  
+ 	if (!bo->tiled) {
+ 		BEGIN_RING(chan, eng2d, mthd, 2);
+ 		OUT_RING  (chan, format);
+ 		OUT_RING  (chan, 1);
+ 		BEGIN_RING(chan, eng2d, mthd + 0x14, 5);
+ 		OUT_RING  (chan, ps->stride);
+ 		OUT_RING  (chan, ps->width);
+ 		OUT_RING  (chan, ps->height);
+ 		OUT_RELOCh(chan, bo, ps->offset, flags);
+ 		OUT_RELOCl(chan, bo, ps->offset, flags);
+ 	} else {
+ 		BEGIN_RING(chan, eng2d, mthd, 5);
+ 		OUT_RING  (chan, format);
+ 		OUT_RING  (chan, 0);
+ 		OUT_RING  (chan, 0);
+ 		OUT_RING  (chan, 1);
+ 		OUT_RING  (chan, 0);
+ 		BEGIN_RING(chan, eng2d, mthd + 0x18, 4);
+ 		OUT_RING  (chan, ps->width);
+ 		OUT_RING  (chan, ps->height);
+ 		OUT_RELOCh(chan, bo, ps->offset, flags);
+ 		OUT_RELOCl(chan, bo, ps->offset, flags);
+ 	}
+ 
+#if 0
+ 	if (dst) {
+ 		BEGIN_RING(chan, eng2d, NV50_2D_CLIP_X, 4);
+ 		OUT_RING  (chan, 0);
+ 		OUT_RING  (chan, 0);
+ 		OUT_RING  (chan, surf->width);
+ 		OUT_RING  (chan, surf->height);
+ 	}
+#endif
+  
+ 	return 0;
+}
+
+int
+nv50_surface_do_copy(struct nv50_screen *screen, struct pipe_surface *dst,
+		     int dx, int dy, struct pipe_surface *src, int sx, int sy,
+		     int w, int h)
+{
+	struct nouveau_channel *chan = screen->nvws->channel;
+	struct nouveau_grobj *eng2d = screen->eng2d;
+	int ret;
+
+	WAIT_RING (chan, 32);
+
+	ret = nv50_surface_set(screen, dst, 1);
+	if (ret)
+		return ret;
+
+	ret = nv50_surface_set(screen, src, 0);
+	if (ret)
+		return ret;
+
+	BEGIN_RING(chan, eng2d, 0x088c, 1);
+	OUT_RING  (chan, 0);
+	BEGIN_RING(chan, eng2d, NV50_2D_BLIT_DST_X, 4);
+	OUT_RING  (chan, dx);
+	OUT_RING  (chan, dy);
+	OUT_RING  (chan, w);
+	OUT_RING  (chan, h);
+	BEGIN_RING(chan, eng2d, 0x08c0, 4);
+	OUT_RING  (chan, 0);
+	OUT_RING  (chan, 1);
+	OUT_RING  (chan, 0);
+	OUT_RING  (chan, 1);
+	BEGIN_RING(chan, eng2d, 0x08d0, 4);
+	OUT_RING  (chan, 0);
+	OUT_RING  (chan, sx);
+	OUT_RING  (chan, 0);
+	OUT_RING  (chan, sy);
+
+	return 0;
+}
+
+static void
+nv50_surface_copy(struct pipe_context *pipe, boolean flip,
+		  struct pipe_surface *dest, unsigned destx, unsigned desty,
+		  struct pipe_surface *src, unsigned srcx, unsigned srcy,
+		  unsigned width, unsigned height)
+{
+	struct nv50_context *nv50 = (struct nv50_context *)pipe;
+	struct nv50_screen *screen = nv50->screen;
+
+	assert(src->format == dest->format);
+
+	if (flip) {
+		desty += height;
+		while (height--) {
+			nv50_surface_do_copy(screen, dest, destx, desty--, src,
+					     srcx, srcy++, width, 1);
+		}
+	} else {
+		nv50_surface_do_copy(screen, dest, destx, desty, src, srcx,
+				     srcy, width, height);
+	}
+}
+
+static void
+nv50_surface_fill(struct pipe_context *pipe, struct pipe_surface *dest,
+		  unsigned destx, unsigned desty, unsigned width,
+		  unsigned height, unsigned value)
+{
+	struct nv50_context *nv50 = (struct nv50_context *)pipe;
+	struct nv50_screen *screen = nv50->screen;
+	struct nouveau_channel *chan = screen->nvws->channel;
+	struct nouveau_grobj *eng2d = screen->eng2d;
+	int format, ret;
+
+	format = nv50_format(dest->format);
+	if (format < 0)
+		return;
+
+	WAIT_RING (chan, 32);
+
+	ret = nv50_surface_set(screen, dest, 1);
+	if (ret)
+		return;
+
+	BEGIN_RING(chan, eng2d, 0x0580, 3);
+	OUT_RING  (chan, 4);
+	OUT_RING  (chan, format);
+	OUT_RING  (chan, value);
+	BEGIN_RING(chan, eng2d, NV50_2D_RECT_X1, 4);
+	OUT_RING  (chan, destx);
+	OUT_RING  (chan, desty);
+	OUT_RING  (chan, width);
+	OUT_RING  (chan, height);
+}
+
+static void *
+nv50_surface_map(struct pipe_screen *screen, struct pipe_surface *ps,
+		 unsigned flags )
+{
+	struct pipe_winsys *ws = screen->winsys;
+
+	return ws->buffer_map(ws, nv50_surface_buffer(ps), flags);
+}
+
+static void
+nv50_surface_unmap(struct pipe_screen *pscreen, struct pipe_surface *ps)
+{
+	struct pipe_winsys *ws = pscreen->winsys;
+
+	ws->buffer_unmap(ws, nv50_surface_buffer(ps));
+}
+
+void
+nv50_init_surface_functions(struct nv50_context *nv50)
+{
+	nv50->pipe.surface_copy = nv50_surface_copy;
+	nv50->pipe.surface_fill = nv50_surface_fill;
+}
+
+void
+nv50_surface_init_screen_functions(struct pipe_screen *pscreen)
+{
+	pscreen->surface_map = nv50_surface_map;
+	pscreen->surface_unmap = nv50_surface_unmap;
+}
+
diff --git a/src/gallium/drivers/nv50/nv50_tex.c b/src/gallium/drivers/nv50/nv50_tex.c
new file mode 100644
index 0000000000..675f9b20cb
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_tex.c
@@ -0,0 +1,163 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "nv50_context.h"
+#include "nv50_texture.h"
+
+#include "nouveau/nouveau_stateobj.h"
+
+static int
+nv50_tex_construct(struct nouveau_stateobj *so, struct nv50_miptree *mt)
+{
+	switch (mt->base.format) {
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+		so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_8_8_8_8);
+		break;
+	case PIPE_FORMAT_A1R5G5B5_UNORM:
+		so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_1_5_5_5);
+		break;
+	case PIPE_FORMAT_A4R4G4B4_UNORM:
+		so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_4_4_4_4);
+		break;
+	case PIPE_FORMAT_R5G6B5_UNORM:
+		so_data(so, NV50TIC_0_0_MAPA_ONE | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_5_6_5);
+		break;
+	case PIPE_FORMAT_L8_UNORM:
+		so_data(so, NV50TIC_0_0_MAPA_ONE | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C0 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_8);
+		break;
+	case PIPE_FORMAT_A8_UNORM:
+		so_data(so, NV50TIC_0_0_MAPA_C0 | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_ZERO | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_ZERO | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_ZERO | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_8);
+		break;
+	case PIPE_FORMAT_I8_UNORM:
+		so_data(so, NV50TIC_0_0_MAPA_C0 | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C0 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_8);
+		break;
+	case PIPE_FORMAT_A8L8_UNORM:
+		so_data(so, NV50TIC_0_0_MAPA_C1 | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C0 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C0 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_8_8);
+		break;
+	case PIPE_FORMAT_DXT1_RGB:
+		so_data(so, NV50TIC_0_0_MAPA_ONE | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_DXT1);
+		break;
+	case PIPE_FORMAT_DXT1_RGBA:
+		so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_DXT1);
+		break;
+	case PIPE_FORMAT_DXT3_RGBA:
+		so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_DXT3);
+		break;
+	case PIPE_FORMAT_DXT5_RGBA:
+		so_data(so, NV50TIC_0_0_MAPA_C3 | NV50TIC_0_0_TYPEA_UNORM |
+			    NV50TIC_0_0_MAPR_C0 | NV50TIC_0_0_TYPER_UNORM |
+			    NV50TIC_0_0_MAPG_C1 | NV50TIC_0_0_TYPEG_UNORM |
+			    NV50TIC_0_0_MAPB_C2 | NV50TIC_0_0_TYPEB_UNORM |
+			    NV50TIC_0_0_FMT_DXT5);
+		break;
+	default:
+		return 1;
+	}
+
+	so_reloc(so, mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_LOW |
+		     NOUVEAU_BO_RD, 0, 0);
+	so_data (so, 0xd0005000);
+	so_data (so, 0x00300000);
+	so_data (so, mt->base.width[0]);
+	so_data (so, (mt->base.depth[0] << 16) | mt->base.height[0]);
+	so_data (so, 0x03000000);
+	so_reloc(so, mt->buffer, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_HIGH |
+		     NOUVEAU_BO_RD, 0, 0);
+
+	return 0;
+}
+
+void
+nv50_tex_validate(struct nv50_context *nv50)
+{
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct nouveau_stateobj *so;
+	int unit, level, image;
+
+	so = so_new(nv50->miptree_nr * 8 + 3, nv50->miptree_nr * 2);
+	so_method(so, tesla, 0x0f00, 1);
+	so_data  (so, NV50_CB_TIC);
+	so_method(so, tesla, 0x40000f04, nv50->miptree_nr * 8);
+	for (unit = 0; unit < nv50->miptree_nr; unit++) {
+		struct nv50_miptree *mt = nv50->miptree[unit];
+
+		for (level = 0; level <= mt->base.last_level; level++) {
+			for (image = 0; image < mt->image_nr; image++) {
+				nv50_miptree_sync(&nv50->screen->pipe, mt,
+						  level, image);
+			}
+		}
+
+		if (nv50_tex_construct(so, mt)) {
+			NOUVEAU_ERR("failed tex validate\n");
+			so_ref(NULL, &so);
+			return;
+		}
+	}
+
+	so_ref(so, &nv50->state.tic_upload);
+}
+
diff --git a/src/gallium/drivers/nv50/nv50_texture.h b/src/gallium/drivers/nv50/nv50_texture.h
new file mode 100644
index 0000000000..aca622c73b
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_texture.h
@@ -0,0 +1,129 @@
+#ifndef __NV50_TEXTURE_H__
+#define __NV50_TEXTURE_H__
+
+/* It'd be really nice to have these in nouveau_class.h generated by
+ * renouveau like the rest of the object header - but not sure it can
+ * handle non-object stuff nicely - need to look into it.
+ */
+
+/* Texture image control block */
+#define NV50TIC_0_0_MAPA_MASK                                     0x38000000
+#define NV50TIC_0_0_MAPA_ZERO                                     0x00000000
+#define NV50TIC_0_0_MAPA_C0                                       0x10000000
+#define NV50TIC_0_0_MAPA_C1                                       0x18000000
+#define NV50TIC_0_0_MAPA_C2                                       0x20000000
+#define NV50TIC_0_0_MAPA_C3                                       0x28000000
+#define NV50TIC_0_0_MAPA_ONE                                      0x38000000
+#define NV50TIC_0_0_MAPR_MASK                                     0x07000000
+#define NV50TIC_0_0_MAPR_ZERO                                     0x00000000
+#define NV50TIC_0_0_MAPR_C0                                       0x02000000
+#define NV50TIC_0_0_MAPR_C1                                       0x03000000
+#define NV50TIC_0_0_MAPR_C2                                       0x04000000
+#define NV50TIC_0_0_MAPR_C3                                       0x05000000
+#define NV50TIC_0_0_MAPR_ONE                                      0x07000000
+#define NV50TIC_0_0_MAPG_MASK                                     0x00e00000
+#define NV50TIC_0_0_MAPG_ZERO                                     0x00000000
+#define NV50TIC_0_0_MAPG_C0                                       0x00400000
+#define NV50TIC_0_0_MAPG_C1                                       0x00600000
+#define NV50TIC_0_0_MAPG_C2                                       0x00800000
+#define NV50TIC_0_0_MAPG_C3                                       0x00a00000
+#define NV50TIC_0_0_MAPG_ONE                                      0x00e00000
+#define NV50TIC_0_0_MAPB_MASK                                     0x001c0000
+#define NV50TIC_0_0_MAPB_ZERO                                     0x00000000
+#define NV50TIC_0_0_MAPB_C0                                       0x00080000
+#define NV50TIC_0_0_MAPB_C1                                       0x000c0000
+#define NV50TIC_0_0_MAPB_C2                                       0x00100000
+#define NV50TIC_0_0_MAPB_C3                                       0x00140000
+#define NV50TIC_0_0_MAPB_ONE                                      0x001c0000
+#define NV50TIC_0_0_TYPEA_MASK                                    0x00038000
+#define NV50TIC_0_0_TYPEA_UNORM                                   0x00010000
+#define NV50TIC_0_0_TYPER_MASK                                    0x00007000
+#define NV50TIC_0_0_TYPER_UNORM                                   0x00002000
+#define NV50TIC_0_0_TYPEG_MASK                                    0x00000e00
+#define NV50TIC_0_0_TYPEG_UNORM                                   0x00000400
+#define NV50TIC_0_0_TYPEB_MASK                                    0x000001c0
+#define NV50TIC_0_0_TYPEB_UNORM                                   0x00000080
+#define NV50TIC_0_0_FMT_MASK                                      0x0000003c
+#define NV50TIC_0_0_FMT_8_8_8_8                                   0x00000008
+#define NV50TIC_0_0_FMT_4_4_4_4                                   0x00000012
+#define NV50TIC_0_0_FMT_1_5_5_5                                   0x00000013
+#define NV50TIC_0_0_FMT_5_6_5                                     0x00000015
+#define NV50TIC_0_0_FMT_8_8                                       0x00000018
+#define NV50TIC_0_0_FMT_8                                         0x0000001d
+#define NV50TIC_0_0_FMT_DXT1                                      0x00000024
+#define NV50TIC_0_0_FMT_DXT3                                      0x00000025
+#define NV50TIC_0_0_FMT_DXT5                                      0x00000026
+
+#define NV50TIC_0_1_OFFSET_LOW_MASK                               0xffffffff
+#define NV50TIC_0_1_OFFSET_LOW_SHIFT                                       0
+
+#define NV50TIC_0_2_UNKNOWN_MASK                                  0xffffffff
+
+#define NV50TIC_0_3_UNKNOWN_MASK                                  0xffffffff
+
+#define NV50TIC_0_4_WIDTH_MASK                                    0x0000ffff
+#define NV50TIC_0_4_WIDTH_SHIFT                                            0
+
+#define NV50TIC_0_5_DEPTH_MASK                                    0xffff0000
+#define NV50TIC_0_5_DEPTH_SHIFT                                           16
+#define NV50TIC_0_5_HEIGHT_MASK                                   0x0000ffff
+#define NV50TIC_0_5_HEIGHT_SHIFT                                           0
+
+#define NV50TIC_0_6_UNKNOWN_MASK                                  0xffffffff
+
+#define NV50TIC_0_7_OFFSET_HIGH_MASK                              0xffffffff
+#define NV50TIC_0_7_OFFSET_HIGH_SHIFT                                      0
+
+/* Texture sampler control block */
+#define NV50TSC_1_0_WRAPS_MASK                                   0x00000007
+#define NV50TSC_1_0_WRAPS_REPEAT                                 0x00000000
+#define NV50TSC_1_0_WRAPS_MIRROR_REPEAT                          0x00000001
+#define NV50TSC_1_0_WRAPS_CLAMP_TO_EDGE                          0x00000002
+#define NV50TSC_1_0_WRAPS_CLAMP_TO_BORDER                        0x00000003
+#define NV50TSC_1_0_WRAPS_CLAMP                                  0x00000004
+#define NV50TSC_1_0_WRAPS_MIRROR_CLAMP_TO_EDGE                   0x00000005
+#define NV50TSC_1_0_WRAPS_MIRROR_CLAMP_TO_BORDER                 0x00000006
+#define NV50TSC_1_0_WRAPS_MIRROR_CLAMP                           0x00000007
+#define NV50TSC_1_0_WRAPT_MASK                                   0x00000038
+#define NV50TSC_1_0_WRAPT_REPEAT                                 0x00000000
+#define NV50TSC_1_0_WRAPT_MIRROR_REPEAT                          0x00000008
+#define NV50TSC_1_0_WRAPT_CLAMP_TO_EDGE                          0x00000010
+#define NV50TSC_1_0_WRAPT_CLAMP_TO_BORDER                        0x00000018
+#define NV50TSC_1_0_WRAPT_CLAMP                                  0x00000020
+#define NV50TSC_1_0_WRAPT_MIRROR_CLAMP_TO_EDGE                   0x00000028
+#define NV50TSC_1_0_WRAPT_MIRROR_CLAMP_TO_BORDER                 0x00000030
+#define NV50TSC_1_0_WRAPT_MIRROR_CLAMP                           0x00000038
+#define NV50TSC_1_0_WRAPR_MASK                                   0x000001c0
+#define NV50TSC_1_0_WRAPR_REPEAT                                 0x00000000
+#define NV50TSC_1_0_WRAPR_MIRROR_REPEAT                          0x00000040
+#define NV50TSC_1_0_WRAPR_CLAMP_TO_EDGE                          0x00000080
+#define NV50TSC_1_0_WRAPR_CLAMP_TO_BORDER                        0x000000c0
+#define NV50TSC_1_0_WRAPR_CLAMP                                  0x00000100
+#define NV50TSC_1_0_WRAPR_MIRROR_CLAMP_TO_EDGE                   0x00000140
+#define NV50TSC_1_0_WRAPR_MIRROR_CLAMP_TO_BORDER                 0x00000180
+#define NV50TSC_1_0_WRAPR_MIRROR_CLAMP                           0x000001c0
+
+#define NV50TSC_1_1_MAGF_MASK                                    0x00000003
+#define NV50TSC_1_1_MAGF_NEAREST                                 0x00000001
+#define NV50TSC_1_1_MAGF_LINEAR                                  0x00000002
+#define NV50TSC_1_1_MINF_MASK                                    0x00000030
+#define NV50TSC_1_1_MINF_NEAREST                                 0x00000010
+#define NV50TSC_1_1_MINF_LINEAR                                  0x00000020
+#define NV50TSC_1_1_MIPF_MASK                                    0x000000c0
+#define NV50TSC_1_1_MIPF_NONE                                    0x00000040
+#define NV50TSC_1_1_MIPF_NEAREST                                 0x00000080
+#define NV50TSC_1_1_MIPF_LINEAR                                  0x000000c0
+
+#define NV50TSC_1_2_UNKNOWN_MASK                                 0xffffffff
+
+#define NV50TSC_1_3_UNKNOWN_MASK                                 0xffffffff
+
+#define NV50TSC_1_4_UNKNOWN_MASK                                 0xffffffff
+
+#define NV50TSC_1_5_UNKNOWN_MASK                                 0xffffffff
+
+#define NV50TSC_1_6_UNKNOWN_MASK                                 0xffffffff
+
+#define NV50TSC_1_7_UNKNOWN_MASK                                 0xffffffff
+
+#endif
diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c
new file mode 100644
index 0000000000..0c970adb03
--- /dev/null
+++ b/src/gallium/drivers/nv50/nv50_vbo.c
@@ -0,0 +1,254 @@
+/*
+ * Copyright 2008 Ben Skeggs
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
+ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+
+#include "nv50_context.h"
+
+static INLINE unsigned
+nv50_prim(unsigned mode)
+{
+	switch (mode) {
+	case PIPE_PRIM_POINTS: return NV50TCL_VERTEX_BEGIN_POINTS;
+	case PIPE_PRIM_LINES: return NV50TCL_VERTEX_BEGIN_LINES;
+	case PIPE_PRIM_LINE_LOOP: return NV50TCL_VERTEX_BEGIN_LINE_LOOP;
+	case PIPE_PRIM_LINE_STRIP: return NV50TCL_VERTEX_BEGIN_LINE_STRIP;
+	case PIPE_PRIM_TRIANGLES: return NV50TCL_VERTEX_BEGIN_TRIANGLES;
+	case PIPE_PRIM_TRIANGLE_STRIP:
+		return NV50TCL_VERTEX_BEGIN_TRIANGLE_STRIP;
+	case PIPE_PRIM_TRIANGLE_FAN: return NV50TCL_VERTEX_BEGIN_TRIANGLE_FAN;
+	case PIPE_PRIM_QUADS: return NV50TCL_VERTEX_BEGIN_QUADS;
+	case PIPE_PRIM_QUAD_STRIP: return NV50TCL_VERTEX_BEGIN_QUAD_STRIP;
+	case PIPE_PRIM_POLYGON: return NV50TCL_VERTEX_BEGIN_POLYGON;
+	default:
+		break;
+	}
+
+	NOUVEAU_ERR("invalid primitive type %d\n", mode);
+	return NV50TCL_VERTEX_BEGIN_POINTS;
+}
+
+boolean
+nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start,
+		 unsigned count)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+
+	nv50_state_validate(nv50);
+
+	BEGIN_RING(chan, tesla, 0x142c, 1);
+	OUT_RING  (chan, 0);
+	BEGIN_RING(chan, tesla, 0x142c, 1);
+	OUT_RING  (chan, 0);
+	BEGIN_RING(chan, tesla, 0x1440, 1);
+	OUT_RING  (chan, 0);
+	BEGIN_RING(chan, tesla, 0x1334, 1);
+	OUT_RING  (chan, 0);
+
+	BEGIN_RING(chan, tesla, NV50TCL_VERTEX_BEGIN, 1);
+	OUT_RING  (chan, nv50_prim(mode));
+	BEGIN_RING(chan, tesla, NV50TCL_VERTEX_BUFFER_FIRST, 2);
+	OUT_RING  (chan, start);
+	OUT_RING  (chan, count);
+	BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1);
+	OUT_RING  (chan, 0);
+
+	pipe->flush(pipe, 0, NULL);
+	return TRUE;
+}
+
+static INLINE void
+nv50_draw_elements_inline_u08(struct nv50_context *nv50, uint8_t *map,
+			      unsigned start, unsigned count)
+{
+	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+
+	map += start;
+
+	if (count & 1) {
+		BEGIN_RING(chan, tesla, 0x15e8, 1);
+		OUT_RING  (chan, map[0]);
+		map++;
+		count--;
+	}
+
+	while (count) {
+		unsigned nr = count > 2046 ? 2046 : count;
+		int i;
+
+		BEGIN_RING(chan, tesla, 0x400015f0, nr >> 1);
+		for (i = 0; i < nr; i += 2)
+			OUT_RING  (chan, (map[1] << 16) | map[0]);
+
+		count -= nr;
+		map += nr;
+	}
+}
+
+static INLINE void
+nv50_draw_elements_inline_u16(struct nv50_context *nv50, uint16_t *map,
+			      unsigned start, unsigned count)
+{
+	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+
+	map += start;
+
+	if (count & 1) {
+		BEGIN_RING(chan, tesla, 0x15e8, 1);
+		OUT_RING  (chan, map[0]);
+		map++;
+		count--;
+	}
+
+	while (count) {
+		unsigned nr = count > 2046 ? 2046 : count;
+		int i;
+
+		BEGIN_RING(chan, tesla, 0x400015f0, nr >> 1);
+		for (i = 0; i < nr; i += 2)
+			OUT_RING  (chan, (map[1] << 16) | map[0]);
+
+		count -= nr;
+		map += nr;
+	}
+}
+
+static INLINE void
+nv50_draw_elements_inline_u32(struct nv50_context *nv50, uint8_t *map,
+			      unsigned start, unsigned count)
+{
+	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+
+	map += start;
+
+	while (count) {
+		unsigned nr = count > 2047 ? 2047 : count;
+
+		BEGIN_RING(chan, tesla, 0x400015e8, nr);
+		OUT_RINGp (chan, map, nr);
+
+		count -= nr;
+		map += nr;
+	}
+}
+
+boolean
+nv50_draw_elements(struct pipe_context *pipe,
+		   struct pipe_buffer *indexBuffer, unsigned indexSize,
+		   unsigned mode, unsigned start, unsigned count)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+	struct nouveau_channel *chan = nv50->screen->nvws->channel;
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct pipe_winsys *ws = pipe->winsys;
+	void *map = ws->buffer_map(ws, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ);
+
+	nv50_state_validate(nv50);
+
+	BEGIN_RING(chan, tesla, 0x142c, 1);
+	OUT_RING  (chan, 0);
+	BEGIN_RING(chan, tesla, 0x142c, 1);
+	OUT_RING  (chan, 0);
+
+	BEGIN_RING(chan, tesla, NV50TCL_VERTEX_BEGIN, 1);
+	OUT_RING  (chan, nv50_prim(mode));
+	switch (indexSize) {
+	case 1:
+		nv50_draw_elements_inline_u08(nv50, map, start, count);
+		break;
+	case 2:
+		nv50_draw_elements_inline_u16(nv50, map, start, count);
+		break;
+	case 4:
+		nv50_draw_elements_inline_u32(nv50, map, start, count);
+		break;
+	default:
+		assert(0);
+	}
+	BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1);
+	OUT_RING  (chan, 0);
+
+	pipe->flush(pipe, 0, NULL);
+	return TRUE;
+}
+
+void
+nv50_vbo_validate(struct nv50_context *nv50)
+{
+	struct nouveau_grobj *tesla = nv50->screen->tesla;
+	struct nouveau_stateobj *vtxbuf, *vtxfmt;
+	int i, vpi = 0;
+
+	vtxbuf = so_new(nv50->vtxelt_nr * 4, nv50->vtxelt_nr * 2);
+	vtxfmt = so_new(nv50->vtxelt_nr + 1, 0);
+	so_method(vtxfmt, tesla, 0x1ac0, nv50->vtxelt_nr);
+
+	for (i = 0; i < nv50->vtxelt_nr; i++) {
+		struct pipe_vertex_element *ve = &nv50->vtxelt[i];
+		struct pipe_vertex_buffer *vb =
+			&nv50->vtxbuf[ve->vertex_buffer_index];
+
+		switch (ve->src_format) {
+		case PIPE_FORMAT_R32G32B32A32_FLOAT:
+			so_data(vtxfmt, 0x7e080000 | i);
+			break;
+		case PIPE_FORMAT_R32G32B32_FLOAT:
+			so_data(vtxfmt, 0x7e100000 | i);
+			break;
+		case PIPE_FORMAT_R32G32_FLOAT:
+			so_data(vtxfmt, 0x7e200000 | i);
+			break;
+		case PIPE_FORMAT_R32_FLOAT:
+			so_data(vtxfmt, 0x7e900000 | i);
+			break;
+		case PIPE_FORMAT_R8G8B8A8_UNORM:
+			so_data(vtxfmt, 0x24500000 | i);
+			break;
+		default:
+		{
+			NOUVEAU_ERR("invalid vbo format %s\n",
+				    pf_name(ve->src_format));
+			assert(0);
+			return;
+		}
+		}
+
+		so_method(vtxbuf, tesla, 0x900 + (i * 16), 3);
+		so_data  (vtxbuf, 0x20000000 | vb->stride);
+		so_reloc (vtxbuf, vb->buffer, vb->buffer_offset +
+			  ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
+			  NOUVEAU_BO_RD | NOUVEAU_BO_HIGH, 0, 0);
+		so_reloc (vtxbuf, vb->buffer, vb->buffer_offset +
+			  ve->src_offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART |
+			  NOUVEAU_BO_RD | NOUVEAU_BO_LOW, 0, 0);
+	}
+
+	so_ref (vtxfmt, &nv50->state.vtxfmt);
+	so_ref (vtxbuf, &nv50->state.vtxbuf);
+}
+
diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile
new file mode 100644
index 0000000000..e83d943cd8
--- /dev/null
+++ b/src/gallium/drivers/r300/Makefile
@@ -0,0 +1,21 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = r300
+
+C_SOURCES = \
+	r300_chipset.c \
+	r300_clear.c \
+	r300_context.c \
+	r300_emit.c \
+	r300_flush.c \
+	r300_screen.c \
+	r300_state.c \
+	r300_state_shader.c \
+	r300_surface.c \
+	r300_swtcl_emit.c \
+	r300_texture.c
+
+include ../../Makefile.template
+
+symlinks:
diff --git a/src/gallium/drivers/r300/SConscript b/src/gallium/drivers/r300/SConscript
new file mode 100644
index 0000000000..18684c3e7f
--- /dev/null
+++ b/src/gallium/drivers/r300/SConscript
@@ -0,0 +1,17 @@
+Import('*')
+
+env = env.Clone()
+
+r300 = env.ConvenienceLibrary(
+	target = 'r300',
+	source = [
+		'r300_blit.c',
+		'r300_clear.c',
+		'r300_context.c',
+		'r300_screen.c',
+		'r300_state.c',
+		'r300_surface.c',
+	])
+
+Export('r300')
+
diff --git a/src/gallium/drivers/r300/r300_chipset.c b/src/gallium/drivers/r300/r300_chipset.c
new file mode 100644
index 0000000000..7def62422a
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_chipset.c
@@ -0,0 +1,348 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_chipset.h"
+#include "pipe/p_debug.h"
+
+/* r300_chipset: A file all to itself for deducing the various properties of
+ * Radeons. */
+
+/* Parse a PCI ID and fill an r300_capabilities struct with information. */
+void r300_parse_chipset(struct r300_capabilities* caps)
+{
+    /* Reasonable defaults */
+    caps->has_tcl = TRUE;
+    caps->is_r500 = FALSE;
+    caps->num_vert_fpus = 4;
+
+
+    /* Note: These are not ordered by PCI ID. I leave that task to GCC,
+     * which will perform the ordering while collating jump tables. Instead,
+     * I've tried to group them according to capabilities and age. */
+    switch (caps->pci_id) {
+        case 0x4144:
+            caps->family = CHIP_FAMILY_R300;
+            break;
+
+        case 0x4145:
+        case 0x4146:
+        case 0x4147:
+        case 0x4E44:
+        case 0x4E45:
+        case 0x4E46:
+        case 0x4E47:
+            caps->family = CHIP_FAMILY_R300;
+            break;
+
+        case 0x4150:
+        case 0x4151:
+        case 0x4152:
+        case 0x4153:
+        case 0x4154:
+        case 0x4155:
+        case 0x4156:
+        case 0x4E50:
+        case 0x4E51:
+        case 0x4E52:
+        case 0x4E53:
+        case 0x4E54:
+        case 0x4E56:
+            caps->family = CHIP_FAMILY_RV350;
+            break;
+
+        case 0x4148:
+        case 0x4149:
+        case 0x414A:
+        case 0x414B:
+        case 0x4E48:
+        case 0x4E49:
+        case 0x4E4B:
+            caps->family = CHIP_FAMILY_R350;
+            break;
+
+        case 0x4E4A:
+            caps->family = CHIP_FAMILY_R360;
+            break;
+
+        case 0x5460:
+        case 0x5462:
+        case 0x5464:
+        case 0x5B60:
+        case 0x5B62:
+        case 0x5B63:
+        case 0x5B64:
+        case 0x5B65:
+            caps->family = CHIP_FAMILY_RV370;
+            break;
+
+        case 0x3150:
+        case 0x3152:
+        case 0x3154:
+        case 0x3E50:
+        case 0x3E54:
+            caps->family = CHIP_FAMILY_RV380;
+            break;
+
+        case 0x4A48:
+        case 0x4A49:
+        case 0x4A4A:
+        case 0x4A4B:
+        case 0x4A4C:
+        case 0x4A4D:
+        case 0x4A4E:
+        case 0x4A4F:
+        case 0x4A50:
+        case 0x4A54:
+            caps->family = CHIP_FAMILY_R420;
+            caps->num_vert_fpus = 6;
+            break;
+
+        case 0x5548:
+        case 0x5549:
+        case 0x554A:
+        case 0x554B:
+        case 0x5550:
+        case 0x5551:
+        case 0x5552:
+        case 0x5554:
+        case 0x5D57:
+            caps->family = CHIP_FAMILY_R423;
+            caps->num_vert_fpus = 6;
+            break;
+
+        case 0x554C:
+        case 0x554D:
+        case 0x554E:
+        case 0x554F:
+        case 0x5D48:
+        case 0x5D49:
+        case 0x5D4A:
+            caps->family = CHIP_FAMILY_R430;
+            caps->num_vert_fpus = 6;
+            break;
+
+        case 0x5D4C:
+        case 0x5D4D:
+        case 0x5D4E:
+        case 0x5D4F:
+        case 0x5D50:
+        case 0x5D52:
+            caps->family = CHIP_FAMILY_R480;
+            caps->num_vert_fpus = 6;
+            break;
+
+        case 0x4B49:
+        case 0x4B4A:
+        case 0x4B4B:
+        case 0x4B4C:
+            caps->family = CHIP_FAMILY_R481;
+            caps->num_vert_fpus = 6;
+            break;
+
+        case 0x5E4C:
+        case 0x5E4F:
+        case 0x564A:
+        case 0x564B:
+        case 0x564F:
+        case 0x5652:
+        case 0x5653:
+        case 0x5657:
+        case 0x5E48:
+        case 0x5E4A:
+        case 0x5E4B:
+        case 0x5E4D:
+            caps->family = CHIP_FAMILY_RV410;
+            caps->num_vert_fpus = 6;
+            break;
+
+        case 0x5954:
+        case 0x5955:
+            caps->family = CHIP_FAMILY_RS480;
+            caps->has_tcl = FALSE;
+            break;
+
+        case 0x5974:
+        case 0x5975:
+            caps->family = CHIP_FAMILY_RS482;
+            caps->has_tcl = FALSE;
+            break;
+
+        case 0x5A41:
+        case 0x5A42:
+            caps->family = CHIP_FAMILY_RS400;
+            caps->has_tcl = FALSE;
+            break;
+
+        case 0x5A61:
+        case 0x5A62:
+            caps->family = CHIP_FAMILY_RC410;
+            caps->has_tcl = FALSE;
+            break;
+
+        case 0x791E:
+        case 0x791F:
+            caps->family = CHIP_FAMILY_RS690;
+            caps->has_tcl = FALSE;
+            break;
+
+        case 0x796C:
+        case 0x796D:
+        case 0x796E:
+        case 0x796F:
+            caps->family = CHIP_FAMILY_RS740;
+            caps->has_tcl = FALSE;
+            break;
+
+        case 0x7100:
+        case 0x7101:
+        case 0x7102:
+        case 0x7103:
+        case 0x7104:
+        case 0x7105:
+        case 0x7106:
+        case 0x7108:
+        case 0x7109:
+        case 0x710A:
+        case 0x710B:
+        case 0x710C:
+        case 0x710E:
+        case 0x710F:
+            caps->family = CHIP_FAMILY_R520;
+            caps->num_vert_fpus = 8;
+            caps->is_r500 = TRUE;
+            break;
+
+        case 0x7140:
+        case 0x7141:
+        case 0x7142:
+        case 0x7143:
+        case 0x7144:
+        case 0x7145:
+        case 0x7146:
+        case 0x7147:
+        case 0x7149:
+        case 0x714A:
+        case 0x714B:
+        case 0x714C:
+        case 0x714D:
+        case 0x714E:
+        case 0x714F:
+        case 0x7151:
+        case 0x7152:
+        case 0x7153:
+        case 0x715E:
+        case 0x715F:
+        case 0x7180:
+        case 0x7181:
+        case 0x7183:
+        case 0x7186:
+        case 0x7187:
+        case 0x7188:
+        case 0x718A:
+        case 0x718B:
+        case 0x718C:
+        case 0x718D:
+        case 0x718F:
+        case 0x7193:
+        case 0x7196:
+        case 0x719B:
+        case 0x719F:
+        case 0x7200:
+        case 0x7210:
+        case 0x7211:
+            caps->family = CHIP_FAMILY_RV515;
+            caps->num_vert_fpus = 2;
+            caps->is_r500 = TRUE;
+            break;
+
+        case 0x71C0:
+        case 0x71C1:
+        case 0x71C2:
+        case 0x71C3:
+        case 0x71C4:
+        case 0x71C5:
+        case 0x71C6:
+        case 0x71C7:
+        case 0x71CD:
+        case 0x71CE:
+        case 0x71D2:
+        case 0x71D4:
+        case 0x71D5:
+        case 0x71D6:
+        case 0x71DA:
+        case 0x71DE:
+            caps->family = CHIP_FAMILY_RV530;
+            caps->num_vert_fpus = 5;
+            caps->is_r500 = TRUE;
+            break;
+
+        case 0x7240:
+        case 0x7243:
+        case 0x7244:
+        case 0x7245:
+        case 0x7246:
+        case 0x7247:
+        case 0x7248:
+        case 0x7249:
+        case 0x724A:
+        case 0x724B:
+        case 0x724C:
+        case 0x724D:
+        case 0x724E:
+        case 0x724F:
+        case 0x7284:
+            caps->family = CHIP_FAMILY_R580;
+            caps->num_vert_fpus = 8;
+            caps->is_r500 = TRUE;
+            break;
+
+        case 0x7280:
+            caps->family = CHIP_FAMILY_RV570;
+            caps->num_vert_fpus = 5;
+            caps->is_r500 = TRUE;
+            break;
+
+        case 0x7281:
+        case 0x7283:
+        case 0x7287:
+        case 0x7288:
+        case 0x7289:
+        case 0x728B:
+        case 0x728C:
+        case 0x7290:
+        case 0x7291:
+        case 0x7293:
+        case 0x7297:
+            caps->family = CHIP_FAMILY_RV560;
+            caps->num_vert_fpus = 5;
+            caps->is_r500 = TRUE;
+            break;
+
+        default:
+            debug_printf("r300: Warning: Unknown chipset 0x%x\n",
+                caps->pci_id);
+            break;
+    }
+
+    /* Force off TCL for now */
+    caps->has_tcl = FALSE;
+}
diff --git a/src/gallium/drivers/r300/r300_chipset.h b/src/gallium/drivers/r300/r300_chipset.h
new file mode 100644
index 0000000000..a9cd372ec5
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_chipset.h
@@ -0,0 +1,79 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_CHIPSET_H
+#define R300_CHIPSET_H
+
+#include "pipe/p_compiler.h"
+
+/* Structure containing all the possible information about a specific Radeon
+ * in the R3xx, R4xx, and R5xx families. */
+struct r300_capabilities {
+    /* PCI ID */
+    uint32_t pci_id;
+    /* Chipset family */
+    int family;
+    /* The number of vertex floating-point units */
+    int num_vert_fpus;
+    /* The number of fragment pipes */
+    int num_frag_pipes;
+    /* Whether or not TCL is physically present */
+    boolean has_tcl;
+    /* Whether or not this is an RV515 or newer; R500s have many differences
+     * that require extra consideration, compared to their R3xx cousins:
+     * - Extra bit of width and height on texture sizes
+     * - Blend color is split across two registers
+     * - Universal Shader (US) block used for fragment shaders */
+    boolean is_r500;
+};
+
+/* Enumerations for legibility and telling which card we're running on. */
+enum {
+    CHIP_FAMILY_R300 = 0,
+    CHIP_FAMILY_R350,
+    CHIP_FAMILY_R360,
+    CHIP_FAMILY_RV350,
+    CHIP_FAMILY_RV370,
+    CHIP_FAMILY_RV380,
+    CHIP_FAMILY_R420,
+    CHIP_FAMILY_R423,
+    CHIP_FAMILY_R430,
+    CHIP_FAMILY_R480,
+    CHIP_FAMILY_R481,
+    CHIP_FAMILY_RV410,
+    CHIP_FAMILY_RS400,
+    CHIP_FAMILY_RC410,
+    CHIP_FAMILY_RS480,
+    CHIP_FAMILY_RS482,
+    CHIP_FAMILY_RS690,
+    CHIP_FAMILY_RS740,
+    CHIP_FAMILY_RV515,
+    CHIP_FAMILY_R520,
+    CHIP_FAMILY_RV530,
+    CHIP_FAMILY_R580,
+    CHIP_FAMILY_RV560,
+    CHIP_FAMILY_RV570
+};
+
+void r300_parse_chipset(struct r300_capabilities* caps);
+
+#endif /* R300_CHIPSET_H */
diff --git a/src/gallium/drivers/r300/r300_clear.c b/src/gallium/drivers/r300/r300_clear.c
new file mode 100644
index 0000000000..fd28437aaa
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_clear.c
@@ -0,0 +1,33 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_clear.h"
+
+/* This gets its own file because Intel's is in its own file.
+ * I assume there's a good reason. */
+void r300_clear(struct pipe_context* pipe,
+                struct pipe_surface* ps,
+                unsigned color)
+{
+    pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, color);
+    ps->status = PIPE_SURFACE_STATUS_DEFINED;
+}
\ No newline at end of file
diff --git a/src/gallium/drivers/r300/r300_clear.h b/src/gallium/drivers/r300/r300_clear.h
new file mode 100644
index 0000000000..e24a0690c9
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_clear.h
@@ -0,0 +1,27 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "pipe/p_context.h"
+
+void r300_clear(struct pipe_context* pipe,
+                struct pipe_surface* ps,
+                unsigned color);
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
new file mode 100644
index 0000000000..7b605ae87a
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -0,0 +1,68 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_context.h"
+
+static void r300_destroy_context(struct pipe_context* context) {
+    struct r300_context* r300 = r300_context(context);
+
+    draw_destroy(r300->draw);
+
+    FREE(r300->blend_color_state);
+    FREE(r300->scissor_state);
+    FREE(r300);
+}
+
+struct pipe_context* r300_create_context(struct pipe_screen* screen,
+                                         struct pipe_winsys* winsys,
+                                         struct r300_winsys* r300_winsys)
+{
+    struct r300_context* r300 = CALLOC_STRUCT(r300_context);
+
+    if (!r300)
+        return NULL;
+
+    r300->winsys = r300_winsys;
+    r300->context.winsys = winsys;
+    r300->context.screen = r300_create_screen(winsys, r300_winsys);
+
+    r300->context.destroy = r300_destroy_context;
+
+    r300->context.clear = r300_clear;
+
+    r300->draw = draw_create();
+    /*XXX draw_set_rasterize_stage(r300->draw, r300_draw_swtcl_stage(r300));*/
+
+    r300->blend_color_state = CALLOC_STRUCT(r300_blend_color_state);
+    r300->scissor_state = CALLOC_STRUCT(r300_scissor_state);
+
+    r300_init_flush_functions(r300);
+
+    r300_init_surface_functions(r300);
+
+    r300_init_state_functions(r300);
+
+    r300->dirty_state = R300_NEW_KITCHEN_SINK;
+    r300->dirty_hw++;
+
+    return &r300->context;
+}
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
new file mode 100644
index 0000000000..376c57639d
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -0,0 +1,190 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_CONTEXT_H
+#define R300_CONTEXT_H
+
+#include "draw/draw_context.h"
+#include "pipe/p_context.h"
+#include "tgsi/tgsi_scan.h"
+#include "util/u_memory.h"
+
+#include "r300_clear.h"
+#include "r300_screen.h"
+#include "r300_winsys.h"
+
+struct r300_blend_state {
+    uint32_t blend_control;       /* R300_RB3D_CBLEND: 0x4e04 */
+    uint32_t alpha_blend_control; /* R300_RB3D_ABLEND: 0x4e08 */
+    uint32_t rop;                 /* R300_RB3D_ROPCNTL: 0x4e18 */
+    uint32_t dither;              /* R300_RB3D_DITHER_CTL: 0x4e50 */
+};
+
+struct r300_blend_color_state {
+    /* RV515 and earlier */
+    uint32_t blend_color;            /* R300_RB3D_BLEND_COLOR: 0x4e10 */
+    /* R520 and newer */
+    uint32_t blend_color_red_alpha;  /* R500_RB3D_CONSTANT_COLOR_AR: 0x4ef8 */
+    uint32_t blend_color_green_blue; /* R500_RB3D_CONSTANT_COLOR_GB: 0x4efc */
+};
+
+struct r300_dsa_state {
+    uint32_t alpha_function;    /* R300_FG_ALPHA_FUNC: 0x4bd4 */
+    uint32_t alpha_reference;   /* R500_FG_ALPHA_VALUE: 0x4be0 */
+    uint32_t z_buffer_control;  /* R300_ZB_CNTL: 0x4f00 */
+    uint32_t z_stencil_control; /* R300_ZB_ZSTENCILCNTL: 0x4f04 */
+    uint32_t stencil_ref_mask;  /* R300_ZB_STENCILREFMASK: 0x4f08 */
+    uint32_t z_buffer_top;      /* R300_ZB_ZTOP: 0x4f14 */
+    uint32_t stencil_ref_bf;    /* R500_ZB_STENCILREFMASK_BF: 0x4fd4 */
+};
+
+struct r300_rs_state {
+    uint32_t vap_control_status;    /* R300_VAP_CNTL_STATUS: 0x2140 */
+    uint32_t point_size;            /* R300_GA_POINT_SIZE: 0x421c */
+    uint32_t line_control;          /* R300_GA_LINE_CNTL: 0x4234 */
+    uint32_t depth_scale_front;  /* R300_SU_POLY_OFFSET_FRONT_SCALE: 0x42a4 */
+    uint32_t depth_offset_front;/* R300_SU_POLY_OFFSET_FRONT_OFFSET: 0x42a8 */
+    uint32_t depth_scale_back;    /* R300_SU_POLY_OFFSET_BACK_SCALE: 0x42ac */
+    uint32_t depth_offset_back;  /* R300_SU_POLY_OFFSET_BACK_OFFSET: 0x42b0 */
+    uint32_t polygon_offset_enable; /* R300_SU_POLY_OFFSET_ENABLE: 0x42b4 */
+    uint32_t cull_mode;             /* R300_SU_CULL_MODE: 0x42b8 */
+    uint32_t line_stipple_config;   /* R300_GA_LINE_STIPPLE_CONFIG: 0x4328 */
+    uint32_t line_stipple_value;    /* R300_GA_LINE_STIPPLE_VALUE: 0x4260 */
+};
+
+struct r300_sampler_state {
+    uint32_t filter0;      /* R300_TX_FILTER0: 0x4400 */
+    uint32_t filter1;      /* R300_TX_FILTER1: 0x4440 */
+    uint32_t border_color; /* R300_TX_BORDER_COLOR: 0x45c0 */
+};
+
+struct r300_scissor_state {
+    uint32_t scissor_top_left;     /* R300_SC_SCISSORS_TL: 0x43e0 */
+    uint32_t scissor_bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */
+};
+
+struct r300_texture_state {
+};
+
+#define R300_NEW_BLEND           0x000001
+#define R300_NEW_BLEND_COLOR     0x000002
+#define R300_NEW_DSA             0x000004
+#define R300_NEW_FRAMEBUFFERS    0x000008
+#define R300_NEW_FRAGMENT_SHADER 0x000010
+#define R300_NEW_RASTERIZER      0x000020
+#define R300_NEW_SAMPLER         0x000040
+#define R300_NEW_SCISSOR         0x004000
+#define R300_NEW_TEXTURE         0x008000
+#define R300_NEW_VERTEX_SHADER   0x800000
+#define R300_NEW_KITCHEN_SINK    0xffffff
+
+/* The next several objects are not pure Radeon state; they inherit from
+ * various Gallium classes. */
+
+struct r3xx_fragment_shader {
+    /* Parent class */
+    struct pipe_shader_state state;
+    struct tgsi_shader_info info;
+
+    /* Has this shader been translated yet? */
+    boolean translated;
+};
+
+struct r300_fragment_shader {
+    /* Parent class */
+    struct r3xx_fragment_shader shader;
+};
+
+struct r500_fragment_shader {
+    /* Parent class */
+    struct r3xx_fragment_shader shader;
+};
+
+struct r300_texture {
+    /* Parent class */
+    struct pipe_texture tex;
+
+    /* Offsets into the buffer. */
+    unsigned offset[PIPE_MAX_TEXTURE_LEVELS];
+
+    /* Total size of this texture, in bytes. */
+    unsigned size;
+
+    /* Pipe buffer backing this texture. */
+    struct pipe_buffer* buffer;
+};
+
+struct r300_context {
+    /* Parent class */
+    struct pipe_context context;
+
+    /* The interface to the windowing system, etc. */
+    struct r300_winsys* winsys;
+    /* Draw module. Used mostly for SW TCL. */
+    struct draw_context* draw;
+
+    /* Various CSO state objects. */
+    /* Blend state. */
+    struct r300_blend_state* blend_state;
+    /* Blend color state. */
+    struct r300_blend_color_state* blend_color_state;
+    /* Depth, stencil, and alpha state. */
+    struct r300_dsa_state* dsa_state;
+    /* Fragment shader. */
+    struct r3xx_fragment_shader* fs;
+    /* Framebuffer state. We currently don't need our own version of this. */
+    struct pipe_framebuffer_state framebuffer_state;
+    /* Rasterizer state. */
+    struct r300_rs_state* rs_state;
+    /* Sampler states. */
+    struct r300_sampler_state* sampler_states[8];
+    int sampler_count;
+    /* Scissor state. */
+    struct r300_scissor_state* scissor_state;
+    /* Texture states. */
+    struct r300_texture* textures[8];
+    struct r300_texture_state* texture_states[8];
+    int texture_count;
+    /* Bitmask of dirty state objects. */
+    uint32_t dirty_state;
+    /* Flag indicating whether or not the HW is dirty. */
+    uint32_t dirty_hw;
+};
+
+/* Convenience cast wrapper. */
+static struct r300_context* r300_context(struct pipe_context* context) {
+    return (struct r300_context*)context;
+}
+
+/* Context initialization. */
+void r300_init_state_functions(struct r300_context* r300);
+void r300_init_surface_functions(struct r300_context* r300);
+
+/* Fun with includes: r300_winsys also declares this prototype.
+ * We'll just step out in that case... */
+#ifndef R300_WINSYS_H
+struct pipe_context* r300_create_context(struct pipe_screen* screen,
+                                         struct pipe_winsys* winsys,
+                                         struct r300_winsys* r300_winsys);
+#endif
+
+#endif /* R300_CONTEXT_H */
diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h
new file mode 100644
index 0000000000..385b61a096
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_cs.h
@@ -0,0 +1,129 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_CS_H
+#define R300_CS_H
+
+#include "r300_reg.h"
+#include "r300_winsys.h"
+
+/* Pack a 32-bit float into a dword. */
+static uint32_t pack_float_32(float f)
+{
+    union {
+        float f;
+        uint32_t u;
+    } u;
+
+    u.f = f;
+    return u.u;
+}
+
+/* Yes, I know macros are ugly. However, they are much prettier than the code
+ * that they neatly hide away, and don't have the cost of function setup,so
+ * we're going to use them. */
+
+#define MAX_CS_SIZE 64 * 1024 / 4
+
+/* XXX stolen from radeon_drm.h */
+#define RADEON_GEM_DOMAIN_CPU  0x1
+#define RADEON_GEM_DOMAIN_GTT  0x2
+#define RADEON_GEM_DOMAIN_VRAM 0x4
+
+/* XXX stolen from radeon_reg.h */
+#define RADEON_CP_PACKET0 0x0
+
+#define CP_PACKET0(register, count) \
+    (RADEON_CP_PACKET0 | ((count) << 16) | ((register) >> 2))
+
+#define CP_PACKET3(op, count) \
+    (RADEON_CP_PACKET3 | (op) | ((count) << 16))
+
+#define CS_LOCALS(context) \
+    struct r300_winsys* cs_winsys = context->winsys; \
+    struct radeon_cs* cs = cs_winsys->cs; \
+    int cs_count = 0;
+
+#define CHECK_CS(size) \
+    cs_winsys->check_cs(cs, (size))
+
+#define BEGIN_CS(size) do { \
+    CHECK_CS(size); \
+    debug_printf("r300: BEGIN_CS, count %d, in %s (%s:%d)\n", \
+        size, __FUNCTION__, __FILE__, __LINE__); \
+    cs_winsys->begin_cs(cs, (size), __FILE__, __FUNCTION__, __LINE__); \
+    cs_count = size; \
+} while (0)
+
+#define OUT_CS(value) do { \
+    cs_winsys->write_cs_dword(cs, (value)); \
+    cs_count--; \
+} while (0)
+
+#define OUT_CS_32F(value) do { \
+    cs_winsys->write_cs_dword(cs, pack_float_32(value)); \
+    cs_count--; \
+} while (0)
+
+#define OUT_CS_REG(register, value) do { \
+    debug_printf("r300: writing 0x%08X to register 0x%04X\n", \
+        value, register); \
+    assert(register); \
+    OUT_CS(CP_PACKET0(register, 0)); \
+    OUT_CS(value); \
+} while (0)
+
+/* Note: This expects count to be the number of registers,
+ * not the actual packet0 count! */
+#define OUT_CS_REG_SEQ(register, count) do { \
+    debug_printf("r300: writing register sequence of %d to 0x%04X\n", \
+        count, register); \
+    assert(register); \
+    OUT_CS(CP_PACKET0(register, ((count) - 1))); \
+} while (0)
+
+#define OUT_CS_RELOC(bo, offset, rd, wd, flags) do { \
+    debug_printf("r300: writing relocation for buffer %p, offset %d\n", \
+        bo, offset); \
+    assert(bo); \
+    OUT_CS(offset); \
+    cs_winsys->write_cs_reloc(cs, bo, rd, wd, flags); \
+    cs_count -= 2; \
+} while (0)
+
+#define END_CS do { \
+    debug_printf("r300: END_CS in %s (%s:%d)\n", __FUNCTION__, __FILE__, \
+        __LINE__); \
+    if (cs_count != 0) \
+        debug_printf("r300: Warning: cs_count off by %d\n", cs_count); \
+    cs_winsys->end_cs(cs, __FILE__, __FUNCTION__, __LINE__); \
+} while (0)
+
+#define FLUSH_CS do { \
+    debug_printf("r300: FLUSH_CS in %s (%s:%d)\n", __FUNCTION__, __FILE__, \
+        __LINE__); \
+    cs_winsys->flush_cs(cs); \
+} while (0)
+
+#include "r300_cs_inlines.h"
+
+#endif /* R300_CS_H */
diff --git a/src/gallium/drivers/r300/r300_cs_inlines.h b/src/gallium/drivers/r300/r300_cs_inlines.h
new file mode 100644
index 0000000000..71e6623699
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_cs_inlines.h
@@ -0,0 +1,37 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+/* r300_cs_inlines: This is just a handful of useful inlines for sending
+ * (very) common instructions to the CS buffer. Should only be included from
+ * r300_cs.h, probably. */
+
+#ifdef R300_CS_H
+
+#define R300_PACIFY do { \
+    OUT_CS_REG(R300_SC_SCREENDOOR, 0x0); \
+    OUT_CS_REG(RADEON_WAIT_UNTIL, (1 << 15) | (1 << 17) | \
+        (1 << 18) | (1 << 31)); \
+    OUT_CS_REG(R300_SC_SCREENDOOR, 0xffffff); \
+} while (0)
+
+
+#endif /* R300_CS_H */
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
new file mode 100644
index 0000000000..585a9e729d
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -0,0 +1,155 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+/* r300_emit: Functions for emitting state. */
+
+#include "r300_emit.h"
+
+void r300_emit_blend_state(struct r300_context* r300,
+                           struct r300_blend_state* blend)
+{
+    CS_LOCALS(r300);
+    BEGIN_CS(7);
+    OUT_CS_REG_SEQ(R300_RB3D_CBLEND, 2);
+    OUT_CS(blend->blend_control);
+    OUT_CS(blend->alpha_blend_control);
+    OUT_CS_REG(R300_RB3D_ROPCNTL, blend->rop);
+    OUT_CS_REG(R300_RB3D_DITHER_CTL, blend->dither);
+    END_CS;
+}
+
+void r300_emit_blend_color_state(struct r300_context* r300,
+                                 struct r300_blend_color_state* bc)
+{
+    struct r300_screen* r300screen =
+        (struct r300_screen*)r300->context.screen;
+    CS_LOCALS(r300);
+    if (r300screen->caps->is_r500) {
+        BEGIN_CS(3);
+        OUT_CS_REG_SEQ(R500_RB3D_CONSTANT_COLOR_AR, 2);
+        OUT_CS(bc->blend_color_red_alpha);
+        OUT_CS(bc->blend_color_green_blue);
+        END_CS;
+    } else {
+        BEGIN_CS(2);
+        OUT_CS_REG(R300_RB3D_BLEND_COLOR, bc->blend_color);
+        END_CS;
+    }
+}
+
+void r300_emit_dsa_state(struct r300_context* r300,
+                           struct r300_dsa_state* dsa)
+{
+    struct r300_screen* r300screen =
+        (struct r300_screen*)r300->context.screen;
+    CS_LOCALS(r300);
+    BEGIN_CS(r300screen->caps->is_r500 ? 8 : 8);
+    OUT_CS_REG(R300_FG_ALPHA_FUNC, dsa->alpha_function);
+    /* XXX figure out the r300 counterpart for this */
+    if (r300screen->caps->is_r500) {
+        /* OUT_CS_REG(R500_FG_ALPHA_VALUE, dsa->alpha_reference); */
+    }
+    OUT_CS_REG_SEQ(R300_ZB_CNTL, 3);
+    OUT_CS(dsa->z_buffer_control);
+    OUT_CS(dsa->z_stencil_control);
+    OUT_CS(dsa->stencil_ref_mask);
+    OUT_CS_REG(R300_ZB_ZTOP, dsa->z_buffer_top);
+    if (r300screen->caps->is_r500) {
+        /* OUT_CS_REG(R500_ZB_STENCILREFMASK_BF, dsa->stencil_ref_bf); */
+    }
+    END_CS;
+}
+
+/* XXX add pitch, stride, z/stencil buf */
+void r300_emit_fb_state(struct r300_context* r300,
+                        struct pipe_framebuffer_state* fb)
+{
+    CS_LOCALS(r300);
+    struct r300_texture* tex;
+    int i;
+
+    BEGIN_CS((3 * fb->nr_cbufs) + 6);
+    for (i = 0; i < fb->nr_cbufs; i++) {
+        tex = (struct r300_texture*)fb->cbufs[i]->texture;
+        OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1);
+        OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+    }
+    R300_PACIFY;
+    END_CS;
+}
+
+void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs)
+{
+    struct r300_screen* r300screen =
+        (struct r300_screen*)r300->context.screen;
+    CS_LOCALS(r300);
+    BEGIN_CS(14);
+    OUT_CS_REG(R300_VAP_CNTL_STATUS, rs->vap_control_status);
+    OUT_CS_REG_SEQ(R300_SU_POLY_OFFSET_FRONT_SCALE, 6);
+    OUT_CS(rs->depth_scale_front);
+    OUT_CS(rs->depth_offset_front);
+    OUT_CS(rs->depth_scale_back);
+    OUT_CS(rs->depth_offset_back);
+    OUT_CS(rs->polygon_offset_enable);
+    OUT_CS(rs->cull_mode);
+    OUT_CS_REG(R300_GA_LINE_STIPPLE_CONFIG, rs->line_stipple_config);
+    OUT_CS_REG(R300_GA_LINE_STIPPLE_VALUE, rs->line_stipple_value);
+    END_CS;
+}
+
+static void r300_emit_dirty_state(struct r300_context* r300)
+{
+    struct r300_screen* r300screen =
+        (struct r300_screen*)r300->context.screen;
+    CS_LOCALS(r300);
+
+    if (!(r300->dirty_state) && !(r300->dirty_hw)) {
+        return;
+    }
+
+    /* XXX check size */
+
+    if (r300->dirty_state & R300_NEW_BLEND) {
+        r300_emit_blend_state(r300, r300->blend_state);
+    }
+
+    if (r300->dirty_state & R300_NEW_BLEND_COLOR) {
+        r300_emit_blend_color_state(r300, r300->blend_color_state);
+    }
+
+    if (r300->dirty_state & R300_NEW_DSA) {
+        r300_emit_dsa_state(r300, r300->dsa_state);
+    }
+
+    if (r300->dirty_state & R300_NEW_RASTERIZER) {
+        r300_emit_rs_state(r300, r300->rs_state);
+    }
+
+    if (r300->dirty_state & R300_NEW_SCISSOR) {
+        struct r300_scissor_state* scissor = r300->scissor_state;
+        /* XXX next two are contiguous regs */
+        OUT_CS_REG(R300_SC_SCISSORS_TL, scissor->scissor_top_left);
+        OUT_CS_REG(R300_SC_SCISSORS_BR, scissor->scissor_bottom_right);
+    }
+
+    r300->dirty_state = 0;
+}
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
new file mode 100644
index 0000000000..b6e69386f9
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -0,0 +1,36 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_context.h"
+#include "r300_cs.h"
+#include "r300_screen.h"
+
+void r300_emit_blend_state(struct r300_context* r300,
+                           struct r300_blend_state* blend);
+
+void r300_emit_blend_color_state(struct r300_context* r300,
+                                 struct r300_blend_color_state* bc);
+
+void r300_emit_dsa_state(struct r300_context* r300,
+                         struct r300_dsa_state* dsa);
+
+void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs);
diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c
new file mode 100644
index 0000000000..3766f0a0a7
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_flush.c
@@ -0,0 +1,42 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_flush.h"
+
+static void r300_flush(struct pipe_context* pipe,
+                       unsigned flags,
+                       struct pipe_fence_handle** fence)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    CS_LOCALS(r300);
+
+    if (r300->dirty_hw) {
+        FLUSH_CS;
+        r300->dirty_state = R300_NEW_KITCHEN_SINK;
+        r300->dirty_hw = 0;
+    }
+}
+
+void r300_init_flush_functions(struct r300_context* r300)
+{
+    r300->context.flush = r300_flush;
+}
diff --git a/src/gallium/drivers/r300/r300_flush.h b/src/gallium/drivers/r300/r300_flush.h
new file mode 100644
index 0000000000..a1b224b39c
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_flush.h
@@ -0,0 +1,33 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_FLUSH_H
+#define R300_FLUSH_H
+
+#include "pipe/p_context.h"
+
+#include "r300_context.h"
+#include "r300_cs.h"
+
+void r300_init_flush_functions(struct r300_context* r300);
+
+#endif /* R300_FLUSH_H */
diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h
new file mode 100644
index 0000000000..dbd0cc28e2
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_reg.h
@@ -0,0 +1,3290 @@
+/**************************************************************************
+
+Copyright (C) 2004-2005 Nicolai Haehnle et al.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, and/or sell copies of the Software, and to permit persons to whom
+the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice (including the next
+paragraph) shall be included in all copies or substantial portions of the
+Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/* *INDENT-OFF* */
+
+#ifndef _R300_REG_H
+#define _R300_REG_H
+
+#define R300_MC_INIT_MISC_LAT_TIMER	0x180
+#	define R300_MC_MISC__MC_CPR_INIT_LAT_SHIFT	0
+#	define R300_MC_MISC__MC_VF_INIT_LAT_SHIFT	4
+#	define R300_MC_MISC__MC_DISP0R_INIT_LAT_SHIFT	8
+#	define R300_MC_MISC__MC_DISP1R_INIT_LAT_SHIFT	12
+#	define R300_MC_MISC__MC_FIXED_INIT_LAT_SHIFT	16
+#	define R300_MC_MISC__MC_E2R_INIT_LAT_SHIFT	20
+#	define R300_MC_MISC__MC_SAME_PAGE_PRIO_SHIFT	24
+#	define R300_MC_MISC__MC_GLOBW_INIT_LAT_SHIFT	28
+
+
+#define R300_MC_INIT_GFX_LAT_TIMER	0x154
+#	define R300_MC_MISC__MC_G3D0R_INIT_LAT_SHIFT	0
+#	define R300_MC_MISC__MC_G3D1R_INIT_LAT_SHIFT	4
+#	define R300_MC_MISC__MC_G3D2R_INIT_LAT_SHIFT	8
+#	define R300_MC_MISC__MC_G3D3R_INIT_LAT_SHIFT	12
+#	define R300_MC_MISC__MC_TX0R_INIT_LAT_SHIFT	16
+#	define R300_MC_MISC__MC_TX1R_INIT_LAT_SHIFT	20
+#	define R300_MC_MISC__MC_GLOBR_INIT_LAT_SHIFT	24
+#	define R300_MC_MISC__MC_GLOBW_FULL_LAT_SHIFT	28
+
+/*
+ * This file contains registers and constants for the R300. They have been
+ * found mostly by examining command buffers captured using glxtest, as well
+ * as by extrapolating some known registers and constants from the R200.
+ * I am fairly certain that they are correct unless stated otherwise
+ * in comments.
+ */
+
+#define R300_SE_VPORT_XSCALE                0x1D98
+#define R300_SE_VPORT_XOFFSET               0x1D9C
+#define R300_SE_VPORT_YSCALE                0x1DA0
+#define R300_SE_VPORT_YOFFSET               0x1DA4
+#define R300_SE_VPORT_ZSCALE                0x1DA8
+#define R300_SE_VPORT_ZOFFSET               0x1DAC
+
+
+/*
+ * Vertex Array Processing (VAP) Control
+ */
+#define R300_VAP_CNTL	0x2080
+#       define R300_PVS_NUM_SLOTS_SHIFT                 0
+#       define R300_PVS_NUM_CNTLRS_SHIFT                4
+#       define R300_PVS_NUM_FPUS_SHIFT                  8
+#       define R300_VF_MAX_VTX_NUM_SHIFT                18
+#       define R300_GL_CLIP_SPACE_DEF                   (0 << 22)
+#       define R300_DX_CLIP_SPACE_DEF                   (1 << 22)
+#       define R500_TCL_STATE_OPTIMIZATION              (1 << 23)
+
+/* This register is written directly and also starts data section
+ * in many 3d CP_PACKET3's
+ */
+#define R300_VAP_VF_CNTL	0x2084
+#	define	R300_VAP_VF_CNTL__PRIM_TYPE__SHIFT              0
+#	define  R300_VAP_VF_CNTL__PRIM_NONE                     (0<<0)
+#	define  R300_VAP_VF_CNTL__PRIM_POINTS                   (1<<0)
+#	define  R300_VAP_VF_CNTL__PRIM_LINES                    (2<<0)
+#	define  R300_VAP_VF_CNTL__PRIM_LINE_STRIP               (3<<0)
+#	define  R300_VAP_VF_CNTL__PRIM_TRIANGLES                (4<<0)
+#	define  R300_VAP_VF_CNTL__PRIM_TRIANGLE_FAN             (5<<0)
+#	define  R300_VAP_VF_CNTL__PRIM_TRIANGLE_STRIP           (6<<0)
+#	define  R300_VAP_VF_CNTL__PRIM_LINE_LOOP                (12<<0)
+#	define  R300_VAP_VF_CNTL__PRIM_QUADS                    (13<<0)
+#	define  R300_VAP_VF_CNTL__PRIM_QUAD_STRIP               (14<<0)
+#	define  R300_VAP_VF_CNTL__PRIM_POLYGON                  (15<<0)
+
+#	define	R300_VAP_VF_CNTL__PRIM_WALK__SHIFT              4
+	/* State based - direct writes to registers trigger vertex
+           generation */
+#	define	R300_VAP_VF_CNTL__PRIM_WALK_STATE_BASED         (0<<4)
+#	define	R300_VAP_VF_CNTL__PRIM_WALK_INDICES             (1<<4)
+#	define	R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST         (2<<4)
+#	define	R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_EMBEDDED     (3<<4)
+
+	/* I don't think I saw these three used.. */
+#	define	R300_VAP_VF_CNTL__COLOR_ORDER__SHIFT            6
+#	define	R300_VAP_VF_CNTL__TCL_OUTPUT_CTL_ENA__SHIFT     9
+#	define	R300_VAP_VF_CNTL__PROG_STREAM_ENA__SHIFT        10
+
+	/* index size - when not set the indices are assumed to be 16 bit */
+#	define	R300_VAP_VF_CNTL__INDEX_SIZE_32bit              (1<<11)
+	/* number of vertices */
+#	define	R300_VAP_VF_CNTL__NUM_VERTICES__SHIFT           16
+
+#define R500_VAP_INDEX_OFFSET		    0x208c
+
+#define R300_VAP_OUTPUT_VTX_FMT_0           0x2090
+#       define R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT     (1<<0)
+#       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT (1<<1)
+#       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT (1<<2)
+#       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT (1<<3)
+#       define R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT (1<<4)
+#       define R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT (1<<16)
+
+#define R300_VAP_OUTPUT_VTX_FMT_1           0x2094
+	/* each of the following is 3 bits wide, specifies number
+	   of components */
+#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT 0
+#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT 3
+#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT 6
+#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT 9
+#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT 12
+#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT 15
+#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT 18
+#       define R300_VAP_OUTPUT_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT 21
+#	define R300_VAP_OUTPUT_VTX_FMT_1__NOT_PRESENT  0
+#	define R300_VAP_OUTPUT_VTX_FMT_1__1_COMPONENT  1
+#	define R300_VAP_OUTPUT_VTX_FMT_1__2_COMPONENTS 2
+#	define R300_VAP_OUTPUT_VTX_FMT_1__3_COMPONENTS 3
+#	define R300_VAP_OUTPUT_VTX_FMT_1__4_COMPONENTS 4
+
+#define R300_SE_VTE_CNTL                  0x20b0
+#	define     R300_VPORT_X_SCALE_ENA                (1 << 0)
+#	define     R300_VPORT_X_OFFSET_ENA               (1 << 1)
+#	define     R300_VPORT_Y_SCALE_ENA                (1 << 2)
+#	define     R300_VPORT_Y_OFFSET_ENA               (1 << 3)
+#	define     R300_VPORT_Z_SCALE_ENA                (1 << 4)
+#	define     R300_VPORT_Z_OFFSET_ENA               (1 << 5)
+#	define     R300_VTX_XY_FMT                       (1 << 8)
+#	define     R300_VTX_Z_FMT                        (1 << 9)
+#	define     R300_VTX_W0_FMT                       (1 << 10)
+#	define     R300_SERIAL_PROC_ENA                  (1 << 11)
+
+#define R300_VAP_VTX_SIZE               0x20b4
+
+/* BEGIN: Vertex data assembly - lots of uncertainties */
+
+/* gap */
+
+/* Maximum Vertex Indx Clamp */
+#define R300_VAP_VF_MAX_VTX_INDX         0x2134
+/* Minimum Vertex Indx Clamp */
+#define R300_VAP_VF_MIN_VTX_INDX         0x2138
+
+/** Vertex assembler/processor control status */
+#define R300_VAP_CNTL_STATUS              0x2140
+/* No swap at all (default) */
+#	define R300_VC_NO_SWAP                  (0 << 0)
+/* 16-bit swap: 0xAABBCCDD becomes 0xBBAADDCC */
+#	define R300_VC_16BIT_SWAP               (1 << 0)
+/* 32-bit swap: 0xAABBCCDD becomes 0xDDCCBBAA */
+#	define R300_VC_32BIT_SWAP               (2 << 0)
+/* Half-dword swap: 0xAABBCCDD becomes 0xCCDDAABB */
+#	define R300_VC_HALF_DWORD_SWAP          (3 << 0)
+/* The TCL engine will not be used (as it is logically or even physically removed) */
+#	define R300_VAP_TCL_BYPASS		(1 << 8)
+/* Read only flag if TCL engine is busy. */
+#	define R300_VAP_PVS_BUSY                (1 << 11)
+/* TODO: gap for MAX_MPS */
+/* Read only flag if the vertex store is busy. */
+#	define R300_VAP_VS_BUSY                 (1 << 24)
+/* Read only flag if the reciprocal engine is busy. */
+#	define R300_VAP_RCP_BUSY                (1 << 25)
+/* Read only flag if the viewport transform engine is busy. */
+#	define R300_VAP_VTE_BUSY                (1 << 26)
+/* Read only flag if the memory interface unit is busy. */
+#	define R300_VAP_MUI_BUSY                (1 << 27)
+/* Read only flag if the vertex cache is busy. */
+#	define R300_VAP_VC_BUSY                 (1 << 28)
+/* Read only flag if the vertex fetcher is busy. */
+#	define R300_VAP_VF_BUSY                 (1 << 29)
+/* Read only flag if the register pipeline is busy. */
+#	define R300_VAP_REGPIPE_BUSY            (1 << 30)
+/* Read only flag if the VAP engine is busy. */
+#	define R300_VAP_VAP_BUSY                (1 << 31)
+
+/* gap */
+
+/* Where do we get our vertex data?
+ *
+ * Vertex data either comes either from immediate mode registers or from
+ * vertex arrays.
+ * There appears to be no mixed mode (though we can force the pitch of
+ * vertex arrays to 0, effectively reusing the same element over and over
+ * again).
+ *
+ * Immediate mode is controlled by the INPUT_CNTL registers. I am not sure
+ * if these registers influence vertex array processing.
+ *
+ * Vertex arrays are controlled via the 3D_LOAD_VBPNTR packet3.
+ *
+ * In both cases, vertex attributes are then passed through INPUT_ROUTE.
+ *
+ * Beginning with INPUT_ROUTE_0_0 is a list of WORDs that route vertex data
+ * into the vertex processor's input registers.
+ * The first word routes the first input, the second word the second, etc.
+ * The corresponding input is routed into the register with the given index.
+ * The list is ended by a word with INPUT_ROUTE_END set.
+ *
+ * Always set COMPONENTS_4 in immediate mode.
+ */
+
+#define R300_VAP_PROG_STREAM_CNTL_0                     0x2150
+#       define R300_DATA_TYPE_0_SHIFT                   0
+#       define R300_DATA_TYPE_FLOAT_1                   0
+#       define R300_DATA_TYPE_FLOAT_2                   1
+#       define R300_DATA_TYPE_FLOAT_3                   2
+#       define R300_DATA_TYPE_FLOAT_4                   3
+#       define R300_DATA_TYPE_BYTE                      4
+#       define R300_DATA_TYPE_D3DCOLOR                  5
+#       define R300_DATA_TYPE_SHORT_2                   6
+#       define R300_DATA_TYPE_SHORT_4                   7
+#       define R300_DATA_TYPE_VECTOR_3_TTT              8
+#       define R300_DATA_TYPE_VECTOR_3_EET              9
+#       define R300_SKIP_DWORDS_SHIFT                   4
+#       define R300_DST_VEC_LOC_SHIFT                   8
+#       define R300_LAST_VEC                            (1 << 13)
+#       define R300_SIGNED                              (1 << 14)
+#       define R300_NORMALIZE                           (1 << 15)
+#       define R300_DATA_TYPE_1_SHIFT                   16
+#define R300_VAP_PROG_STREAM_CNTL_1                     0x2154
+#define R300_VAP_PROG_STREAM_CNTL_2                     0x2158
+#define R300_VAP_PROG_STREAM_CNTL_3                     0x215C
+#define R300_VAP_PROG_STREAM_CNTL_4                     0x2160
+#define R300_VAP_PROG_STREAM_CNTL_5                     0x2164
+#define R300_VAP_PROG_STREAM_CNTL_6                     0x2168
+#define R300_VAP_PROG_STREAM_CNTL_7                     0x216C
+/* gap */
+
+/* Notes:
+ *  - always set up to produce at least two attributes:
+ *    if vertex program uses only position, fglrx will set normal, too
+ *  - INPUT_CNTL_0_COLOR and INPUT_CNTL_COLOR bits are always equal.
+ */
+#define R300_VAP_VTX_STATE_CNTL               0x2180
+#       define R300_COLOR_0_ASSEMBLY_SHIFT    0
+#       define R300_SEL_COLOR                 0
+#       define R300_SEL_USER_COLOR_0          1
+#       define R300_SEL_USER_COLOR_1          2
+#       define R300_COLOR_1_ASSEMBLY_SHIFT    2
+#       define R300_COLOR_2_ASSEMBLY_SHIFT    4
+#       define R300_COLOR_3_ASSEMBLY_SHIFT    6
+#       define R300_COLOR_4_ASSEMBLY_SHIFT    8
+#       define R300_COLOR_5_ASSEMBLY_SHIFT    10
+#       define R300_COLOR_6_ASSEMBLY_SHIFT    12
+#       define R300_COLOR_7_ASSEMBLY_SHIFT    14
+#       define R300_UPDATE_USER_COLOR_0_ENA   (1 << 16)
+
+/*
+ * Each bit in this field applies to the corresponding vector in the VSM
+ * memory (i.e. Bit 0 applies to VECTOR_0 (POSITION), etc.). If the bit
+ * is set, then the corresponding 4-Dword Vector is output into the Vertex Stream.
+ */
+#define R300_VAP_VSM_VTX_ASSM               0x2184
+#       define R300_INPUT_CNTL_POS               0x00000001
+#       define R300_INPUT_CNTL_NORMAL            0x00000002
+#       define R300_INPUT_CNTL_COLOR             0x00000004
+#       define R300_INPUT_CNTL_TC0               0x00000400
+#       define R300_INPUT_CNTL_TC1               0x00000800
+#       define R300_INPUT_CNTL_TC2               0x00001000 /* GUESS */
+#       define R300_INPUT_CNTL_TC3               0x00002000 /* GUESS */
+#       define R300_INPUT_CNTL_TC4               0x00004000 /* GUESS */
+#       define R300_INPUT_CNTL_TC5               0x00008000 /* GUESS */
+#       define R300_INPUT_CNTL_TC6               0x00010000 /* GUESS */
+#       define R300_INPUT_CNTL_TC7               0x00020000 /* GUESS */
+
+/* Programmable Stream Control Signed Normalize Control */
+#define R300_VAP_PSC_SGN_NORM_CNTL         0x21dc
+#	define SGN_NORM_ZERO                 0
+#	define SGN_NORM_ZERO_CLAMP_MINUS_ONE 1
+#	define SGN_NORM_NO_ZERO              2
+
+/* gap */
+
+/* Words parallel to INPUT_ROUTE_0; All words that are active in INPUT_ROUTE_0
+ * are set to a swizzling bit pattern, other words are 0.
+ *
+ * In immediate mode, the pattern is always set to xyzw. In vertex array
+ * mode, the swizzling pattern is e.g. used to set zw components in texture
+ * coordinates with only tweo components.
+ */
+#define R300_VAP_PROG_STREAM_CNTL_EXT_0                 0x21e0
+#       define R300_SWIZZLE0_SHIFT                      0
+#       define R300_SWIZZLE_SELECT_X_SHIFT              0
+#       define R300_SWIZZLE_SELECT_Y_SHIFT              3
+#       define R300_SWIZZLE_SELECT_Z_SHIFT              6
+#       define R300_SWIZZLE_SELECT_W_SHIFT              9
+
+#       define R300_SWIZZLE_SELECT_X                    0
+#       define R300_SWIZZLE_SELECT_Y                    1
+#       define R300_SWIZZLE_SELECT_Z                    2
+#       define R300_SWIZZLE_SELECT_W                    3
+#       define R300_SWIZZLE_SELECT_FP_ZERO              4
+#       define R300_SWIZZLE_SELECT_FP_ONE               5
+/* alternate forms for r300_emit.c */
+#       define R300_INPUT_ROUTE_SELECT_X    0
+#       define R300_INPUT_ROUTE_SELECT_Y    1
+#       define R300_INPUT_ROUTE_SELECT_Z    2
+#       define R300_INPUT_ROUTE_SELECT_W    3
+#       define R300_INPUT_ROUTE_SELECT_ZERO 4
+#       define R300_INPUT_ROUTE_SELECT_ONE  5
+
+#       define R300_WRITE_ENA_SHIFT                     12
+#       define R300_WRITE_ENA_X                         1
+#       define R300_WRITE_ENA_Y                         2
+#       define R300_WRITE_ENA_Z                         4
+#       define R300_WRITE_ENA_W                         8
+#       define R300_SWIZZLE1_SHIFT                      16
+#define R300_VAP_PROG_STREAM_CNTL_EXT_1                 0x21e4
+#define R300_VAP_PROG_STREAM_CNTL_EXT_2                 0x21e8
+#define R300_VAP_PROG_STREAM_CNTL_EXT_3                 0x21ec
+#define R300_VAP_PROG_STREAM_CNTL_EXT_4                 0x21f0
+#define R300_VAP_PROG_STREAM_CNTL_EXT_5                 0x21f4
+#define R300_VAP_PROG_STREAM_CNTL_EXT_6                 0x21f8
+#define R300_VAP_PROG_STREAM_CNTL_EXT_7                 0x21fc
+
+/* END: Vertex data assembly */
+
+/* gap */
+
+/* BEGIN: Upload vertex program and data */
+
+/*
+ * The programmable vertex shader unit has a memory bank of unknown size
+ * that can be written to in 16 byte units by writing the address into
+ * UPLOAD_ADDRESS, followed by data in UPLOAD_DATA (multiples of 4 DWORDs).
+ *
+ * Pointers into the memory bank are always in multiples of 16 bytes.
+ *
+ * The memory bank is divided into areas with fixed meaning.
+ *
+ * Starting at address UPLOAD_PROGRAM: Vertex program instructions.
+ * Native limits reported by drivers from ATI suggest size 256 (i.e. 4KB),
+ * whereas the difference between known addresses suggests size 512.
+ *
+ * Starting at address UPLOAD_PARAMETERS: Vertex program parameters.
+ * Native reported limits and the VPI layout suggest size 256, whereas
+ * difference between known addresses suggests size 512.
+ *
+ * At address UPLOAD_POINTSIZE is a vector (0, 0, ps, 0), where ps is the
+ * floating point pointsize. The exact purpose of this state is uncertain,
+ * as there is also the R300_RE_POINTSIZE register.
+ *
+ * Multiple vertex programs and parameter sets can be loaded at once,
+ * which could explain the size discrepancy.
+ */
+#define R300_VAP_PVS_VECTOR_INDX_REG         0x2200
+#       define R300_PVS_CODE_START           0
+#       define R300_MAX_PVS_CODE_LINES       256
+#       define R500_MAX_PVS_CODE_LINES       1024
+#       define R300_PVS_CONST_START          512
+#       define R500_PVS_CONST_START          1024
+#       define R300_MAX_PVS_CONST_VECS       256
+#       define R500_MAX_PVS_CONST_VECS       1024
+#       define R300_PVS_UCP_START            1024
+#       define R500_PVS_UCP_START            1536
+#       define R300_POINT_VPORT_SCALE_OFFSET 1030
+#       define R500_POINT_VPORT_SCALE_OFFSET 1542
+#       define R300_POINT_GEN_TEX_OFFSET     1031
+#       define R500_POINT_GEN_TEX_OFFSET     1543
+
+/*
+ * These are obsolete defines form r300_context.h, but they might give some
+ * clues when investigating the addresses further...
+ */
+#if 0
+#define VSF_DEST_PROGRAM        0x0
+#define VSF_DEST_MATRIX0        0x200
+#define VSF_DEST_MATRIX1        0x204
+#define VSF_DEST_MATRIX2        0x208
+#define VSF_DEST_VECTOR0        0x20c
+#define VSF_DEST_VECTOR1        0x20d
+#define VSF_DEST_UNKNOWN1       0x400
+#define VSF_DEST_UNKNOWN2       0x406
+#endif
+
+/* gap */
+
+#define R300_VAP_PVS_UPLOAD_DATA            0x2208
+
+/* END: Upload vertex program and data */
+
+/* gap */
+
+/* I do not know the purpose of this register. However, I do know that
+ * it is set to 221C_CLEAR for clear operations and to 221C_NORMAL
+ * for normal rendering.
+ *
+ * 2007-11-05: This register is the user clip plane control register, but there
+ * also seems to be a rendering mode control; the NORMAL/CLEAR defines.
+ *
+ * See bug #9871. http://bugs.freedesktop.org/attachment.cgi?id=10672&action=view
+ */
+#define R300_VAP_CLIP_CNTL                       0x221C
+#       define R300_VAP_UCP_ENABLE_0             (1 << 0)
+#       define R300_VAP_UCP_ENABLE_1             (1 << 1)
+#       define R300_VAP_UCP_ENABLE_2             (1 << 2)
+#       define R300_VAP_UCP_ENABLE_3             (1 << 3)
+#       define R300_VAP_UCP_ENABLE_4             (1 << 4)
+#       define R300_VAP_UCP_ENABLE_5             (1 << 5)
+#       define R300_PS_UCP_MODE_DIST_COP         (0 << 14)
+#       define R300_PS_UCP_MODE_RADIUS_COP       (1 << 14)
+#       define R300_PS_UCP_MODE_RADIUS_COP_CLIP  (2 << 14)
+#       define R300_PS_UCP_MODE_CLIP_AS_TRIFAN   (3 << 14)
+#       define R300_CLIP_DISABLE                 (1 << 16)
+#       define R300_UCP_CULL_ONLY_ENABLE         (1 << 17)
+#       define R300_BOUNDARY_EDGE_FLAG_ENABLE    (1 << 18)
+#       define R500_COLOR2_IS_TEXTURE            (1 << 20)
+#       define R500_COLOR3_IS_TEXTURE            (1 << 21)
+
+/* These seem to be per-pixel and per-vertex X and Y clipping planes. The first
+ * plane is per-pixel and the second plane is per-vertex.
+ *
+ * This was determined by experimentation alone but I believe it is correct.
+ *
+ * These registers are called X_QUAD0_1_FL to X_QUAD0_4_FL by glxtest.
+ */
+#define R300_VAP_GB_VERT_CLIP_ADJ                   0x2220
+#define R300_VAP_GB_VERT_DISC_ADJ                   0x2224
+#define R300_VAP_GB_HORZ_CLIP_ADJ                   0x2228
+#define R300_VAP_GB_HORZ_DISC_ADJ                   0x222c
+
+/* gap */
+
+/* Sometimes, END_OF_PKT and 0x2284=0 are the only commands sent between
+ * rendering commands and overwriting vertex program parameters.
+ * Therefore, I suspect writing zero to 0x2284 synchronizes the engine and
+ * avoids bugs caused by still running shaders reading bad data from memory.
+ */
+#define R300_VAP_PVS_STATE_FLUSH_REG        0x2284
+
+/* This register is used to define the number of core clocks to wait for a
+ * vertex to be received by the VAP input controller (while the primitive
+ * path is backed up) before forcing any accumulated vertices to be submitted
+ * to the vertex processing path.
+ */
+#define VAP_PVS_VTX_TIMEOUT_REG             0x2288
+#       define R300_2288_R300                    0x00750000 /* -- nh */
+#       define R300_2288_RV350                   0x0000FFFF /* -- Vladimir */
+
+/* gap */
+
+/* Addresses are relative to the vertex program instruction area of the
+ * memory bank. PROGRAM_END points to the last instruction of the active
+ * program
+ *
+ * The meaning of the two UNKNOWN fields is obviously not known. However,
+ * experiments so far have shown that both *must* point to an instruction
+ * inside the vertex program, otherwise the GPU locks up.
+ *
+ * fglrx usually sets CNTL_3_UNKNOWN to the end of the program and
+ * R300_PVS_CNTL_1_POS_END_SHIFT points to instruction where last write to
+ * position takes place.
+ *
+ * Most likely this is used to ignore rest of the program in cases
+ * where group of verts arent visible. For some reason this "section"
+ * is sometimes accepted other instruction that have no relationship with
+ * position calculations.
+ */
+#define R300_VAP_PVS_CODE_CNTL_0            0x22D0
+#       define R300_PVS_FIRST_INST_SHIFT         0
+#       define R300_PVS_XYZW_VALID_INST_SHIFT    10
+#       define R300_PVS_LAST_INST_SHIFT          20
+/* Addresses are relative the the vertex program parameters area. */
+#define R300_VAP_PVS_CONST_CNTL             0x22D4
+#       define R300_PVS_CONST_BASE_OFFSET_SHIFT  0
+#       define R300_PVS_MAX_CONST_ADDR_SHIFT     16
+#define R300_VAP_PVS_CODE_CNTL_1	    0x22D8
+#       define R300_PVS_LAST_VTX_SRC_INST_SHIFT  0
+#define R300_VAP_PVS_FLOW_CNTL_OPC          0x22DC
+
+/* The entire range from 0x2300 to 0x2AC inclusive seems to be used for
+ * immediate vertices
+ */
+#define R300_VAP_VTX_COLOR_R                0x2464
+#define R300_VAP_VTX_COLOR_G                0x2468
+#define R300_VAP_VTX_COLOR_B                0x246C
+#define R300_VAP_VTX_POS_0_X_1              0x2490 /* used for glVertex2*() */
+#define R300_VAP_VTX_POS_0_Y_1              0x2494
+#define R300_VAP_VTX_COLOR_PKD              0x249C /* RGBA */
+#define R300_VAP_VTX_POS_0_X_2              0x24A0 /* used for glVertex3*() */
+#define R300_VAP_VTX_POS_0_Y_2              0x24A4
+#define R300_VAP_VTX_POS_0_Z_2              0x24A8
+/* write 0 to indicate end of packet? */
+#define R300_VAP_VTX_END_OF_PKT             0x24AC
+
+/* gap */
+
+/* These are values from r300_reg/r300_reg.h - they are known to be correct
+ * and are here so we can use one register file instead of several
+ * - Vladimir
+ */
+#define R300_GB_VAP_RASTER_VTX_FMT_0	0x4000
+#	define R300_GB_VAP_RASTER_VTX_FMT_0__POS_PRESENT	(1<<0)
+#	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_0_PRESENT	(1<<1)
+#	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_1_PRESENT	(1<<2)
+#	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_2_PRESENT	(1<<3)
+#	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_3_PRESENT	(1<<4)
+#	define R300_GB_VAP_RASTER_VTX_FMT_0__COLOR_SPACE	(0xf<<5)
+#	define R300_GB_VAP_RASTER_VTX_FMT_0__PT_SIZE_PRESENT	(0x1<<16)
+
+#define R300_GB_VAP_RASTER_VTX_FMT_1	0x4004
+	/* each of the following is 3 bits wide, specifies number
+	   of components */
+#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_0_COMP_CNT_SHIFT	0
+#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_1_COMP_CNT_SHIFT	3
+#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_2_COMP_CNT_SHIFT	6
+#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_3_COMP_CNT_SHIFT	9
+#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_4_COMP_CNT_SHIFT	12
+#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_5_COMP_CNT_SHIFT	15
+#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_6_COMP_CNT_SHIFT	18
+#	define R300_GB_VAP_RASTER_VTX_FMT_1__TEX_7_COMP_CNT_SHIFT	21
+
+/* UNK30 seems to enables point to quad transformation on textures
+ * (or something closely related to that).
+ * This bit is rather fatal at the time being due to lackings at pixel
+ * shader side
+ * Specifies top of Raster pipe specific enable controls.
+ */
+#define R300_GB_ENABLE	0x4008
+#	define R300_GB_POINT_STUFF_DISABLE     (0 << 0)
+#	define R300_GB_POINT_STUFF_ENABLE      (1 << 0) /* Specifies if points will have stuffed texture coordinates. */
+#	define R300_GB_LINE_STUFF_DISABLE      (0 << 1)
+#	define R300_GB_LINE_STUFF_ENABLE       (1 << 1) /* Specifies if lines will have stuffed texture coordinates. */
+#	define R300_GB_TRIANGLE_STUFF_DISABLE  (0 << 2)
+#	define R300_GB_TRIANGLE_STUFF_ENABLE   (1 << 2) /* Specifies if triangles will have stuffed texture coordinates. */
+#	define R300_GB_STENCIL_AUTO_DISABLE    (0 << 4)
+#	define R300_GB_STENCIL_AUTO_ENABLE     (1 << 4) /* Enable stencil auto inc/dec based on triangle cw/ccw, force into dzy low bit. */
+#	define R300_GB_STENCIL_AUTO_FORCE      (2 << 4) /* Force 0 into dzy low bit. */
+
+	/* each of the following is 2 bits wide */
+#define R300_GB_TEX_REPLICATE	0 /* Replicate VAP source texture coordinates (S,T,[R,Q]). */
+#define R300_GB_TEX_ST		1 /* Stuff with source texture coordinates (S,T). */
+#define R300_GB_TEX_STR		2 /* Stuff with source texture coordinates (S,T,R). */
+#	define R300_GB_TEX0_SOURCE_SHIFT	16
+#	define R300_GB_TEX1_SOURCE_SHIFT	18
+#	define R300_GB_TEX2_SOURCE_SHIFT	20
+#	define R300_GB_TEX3_SOURCE_SHIFT	22
+#	define R300_GB_TEX4_SOURCE_SHIFT	24
+#	define R300_GB_TEX5_SOURCE_SHIFT	26
+#	define R300_GB_TEX6_SOURCE_SHIFT	28
+#	define R300_GB_TEX7_SOURCE_SHIFT	30
+
+/* MSPOS - positions for multisample antialiasing (?) */
+#define R300_GB_MSPOS0                           0x4010
+	/* shifts - each of the fields is 4 bits */
+#	define R300_GB_MSPOS0__MS_X0_SHIFT	0
+#	define R300_GB_MSPOS0__MS_Y0_SHIFT	4
+#	define R300_GB_MSPOS0__MS_X1_SHIFT	8
+#	define R300_GB_MSPOS0__MS_Y1_SHIFT	12
+#	define R300_GB_MSPOS0__MS_X2_SHIFT	16
+#	define R300_GB_MSPOS0__MS_Y2_SHIFT	20
+#	define R300_GB_MSPOS0__MSBD0_Y		24
+#	define R300_GB_MSPOS0__MSBD0_X		28
+
+#define R300_GB_MSPOS1                           0x4014
+#	define R300_GB_MSPOS1__MS_X3_SHIFT	0
+#	define R300_GB_MSPOS1__MS_Y3_SHIFT	4
+#	define R300_GB_MSPOS1__MS_X4_SHIFT	8
+#	define R300_GB_MSPOS1__MS_Y4_SHIFT	12
+#	define R300_GB_MSPOS1__MS_X5_SHIFT	16
+#	define R300_GB_MSPOS1__MS_Y5_SHIFT	20
+#	define R300_GB_MSPOS1__MSBD1		24
+
+/* Specifies the graphics pipeline configuration for rasterization. */
+#define R300_GB_TILE_CONFIG                      0x4018
+#	define R300_GB_TILE_DISABLE             (0 << 0)
+#	define R300_GB_TILE_ENABLE              (1 << 0)
+#	define R300_GB_TILE_PIPE_COUNT_RV300	(0 << 1) /* RV350 (1 pipe, 1 ctx) */
+#	define R300_GB_TILE_PIPE_COUNT_R300	(3 << 1) /* R300 (2 pipes, 1 ctx) */
+#	define R300_GB_TILE_PIPE_COUNT_R420_3P  (6 << 1) /* R420-3P (3 pipes, 1 ctx) */
+#	define R300_GB_TILE_PIPE_COUNT_R420	(7 << 1) /* R420 (4 pipes, 1 ctx) */
+#	define R300_GB_TILE_SIZE_8		(0 << 4)
+#	define R300_GB_TILE_SIZE_16		(1 << 4)
+#	define R300_GB_TILE_SIZE_32		(2 << 4)
+#	define R300_GB_SUPER_SIZE_1		(0 << 6)
+#	define R300_GB_SUPER_SIZE_2		(1 << 6)
+#	define R300_GB_SUPER_SIZE_4		(2 << 6)
+#	define R300_GB_SUPER_SIZE_8		(3 << 6)
+#	define R300_GB_SUPER_SIZE_16		(4 << 6)
+#	define R300_GB_SUPER_SIZE_32		(5 << 6)
+#	define R300_GB_SUPER_SIZE_64		(6 << 6)
+#	define R300_GB_SUPER_SIZE_128		(7 << 6)
+#	define R300_GB_SUPER_X_SHIFT		9	/* 3 bits wide */
+#	define R300_GB_SUPER_Y_SHIFT		12	/* 3 bits wide */
+#	define R300_GB_SUPER_TILE_A		(0 << 15)
+#	define R300_GB_SUPER_TILE_B		(1 << 15)
+#	define R300_GB_SUBPIXEL_1_12		(0 << 16)
+#	define R300_GB_SUBPIXEL_1_16		(1 << 16)
+#	define GB_TILE_CONFIG_QUADS_PER_RAS_4   (0 << 17)
+#	define GB_TILE_CONFIG_QUADS_PER_RAS_8   (1 << 17)
+#	define GB_TILE_CONFIG_QUADS_PER_RAS_16  (2 << 17)
+#	define GB_TILE_CONFIG_QUADS_PER_RAS_32  (3 << 17)
+#	define GB_TILE_CONFIG_BB_SCAN_INTERCEPT (0 << 19)
+#	define GB_TILE_CONFIG_BB_SCAN_BOUND_BOX (1 << 19)
+#	define GB_TILE_CONFIG_ALT_SCAN_EN_LR    (0 << 20)
+#	define GB_TILE_CONFIG_ALT_SCAN_EN_LRL   (1 << 20)
+#	define GB_TILE_CONFIG_ALT_OFFSET        (0 << 21)
+#	define GB_TILE_CONFIG_SUBPRECISION      (0 << 22)
+#	define GB_TILE_CONFIG_ALT_TILING_DEF    (0 << 23)
+#	define GB_TILE_CONFIG_ALT_TILING_3_2    (1 << 23)
+#	define GB_TILE_CONFIG_Z_EXTENDED_24_1   (0 << 24)
+#	define GB_TILE_CONFIG_Z_EXTENDED_S25_1  (1 << 24)
+
+/* Specifies the sizes of the various FIFO`s in the sc/rs/us. This register must be the first one written */
+#define R300_GB_FIFO_SIZE	0x4024
+	/* each of the following is 2 bits wide */
+#define R300_GB_FIFO_SIZE_32	0
+#define R300_GB_FIFO_SIZE_64	1
+#define R300_GB_FIFO_SIZE_128	2
+#define R300_GB_FIFO_SIZE_256	3
+#	define R300_SC_IFIFO_SIZE_SHIFT	0
+#	define R300_SC_TZFIFO_SIZE_SHIFT	2
+#	define R300_SC_BFIFO_SIZE_SHIFT	4
+
+#	define R300_US_OFIFO_SIZE_SHIFT	12
+#	define R300_US_WFIFO_SIZE_SHIFT	14
+	/* the following use the same constants as above, but meaning is
+	   is times 2 (i.e. instead of 32 words it means 64 */
+#	define R300_RS_TFIFO_SIZE_SHIFT	6
+#	define R300_RS_CFIFO_SIZE_SHIFT	8
+#	define R300_US_RAM_SIZE_SHIFT		10
+	/* watermarks, 3 bits wide */
+#	define R300_RS_HIGHWATER_COL_SHIFT	16
+#	define R300_RS_HIGHWATER_TEX_SHIFT	19
+#	define R300_OFIFO_HIGHWATER_SHIFT	22	/* two bits only */
+#	define R300_CUBE_FIFO_HIGHWATER_COL_SHIFT	24
+
+#define GB_Z_PEQ_CONFIG                          0x4028
+#	define GB_Z_PEQ_CONFIG_Z_PEQ_SIZE_4_4    (0 << 0)
+#	define GB_Z_PEQ_CONFIG_Z_PEQ_SIZE_8_8    (1 << 0)
+
+/* Specifies various polygon specific selects (fog, depth, perspective). */
+#define R300_GB_SELECT                           0x401c
+#	define R300_GB_FOG_SELECT_C0A		(0 << 0)
+#	define R300_GB_FOG_SELECT_C1A           (1 << 0)
+#	define R300_GB_FOG_SELECT_C2A           (2 << 0)
+#	define R300_GB_FOG_SELECT_C3A           (3 << 0)
+#	define R300_GB_FOG_SELECT_1_1_W         (4 << 0)
+#	define R300_GB_FOG_SELECT_Z		(5 << 0)
+#	define R300_GB_DEPTH_SELECT_Z		(0 << 3)
+#	define R300_GB_DEPTH_SELECT_1_1_W	(1 << 3)
+#	define R300_GB_W_SELECT_1_W		(0 << 4)
+#	define R300_GB_W_SELECT_1		(1 << 4)
+#	define R300_GB_FOG_STUFF_DISABLE        (0 << 5)
+#	define R300_GB_FOG_STUFF_ENABLE         (1 << 5)
+#	define R300_GB_FOG_STUFF_TEX_SHIFT      6
+#	define R300_GB_FOG_STUFF_TEX_MASK       0x000003c0
+#	define R300_GB_FOG_STUFF_COMP_SHIFT     10
+#	define R300_GB_FOG_STUFF_COMP_MASK      0x00000c00
+
+/* Specifies the graphics pipeline configuration for antialiasing. */
+#define R300_GB_AA_CONFIG                         0x4020
+#	define GB_AA_CONFIG_AA_DISABLE           (0 << 0)
+#	define GB_AA_CONFIG_AA_ENABLE            (1 << 0)
+#	define GB_AA_CONFIG_NUM_AA_SUBSAMPLES_2  (0 << 1)
+#	define GB_AA_CONFIG_NUM_AA_SUBSAMPLES_3  (1 << 1)
+#	define GB_AA_CONFIG_NUM_AA_SUBSAMPLES_4  (2 << 1)
+#	define GB_AA_CONFIG_NUM_AA_SUBSAMPLES_6  (3 << 1)
+
+/* Selects which of 4 pipes are active. */
+#define GB_PIPE_SELECT                           0x402c
+#	define GB_PIPE_SELECT_PIPE0_ID_SHIFT  0
+#	define GB_PIPE_SELECT_PIPE1_ID_SHIFT  2
+#	define GB_PIPE_SELECT_PIPE2_ID_SHIFT  4
+#	define GB_PIPE_SELECT_PIPE3_ID_SHIFT  6
+#	define GB_PIPE_SELECT_PIPE_MASK_SHIFT 8
+#	define GB_PIPE_SELECT_MAX_PIPE        12
+#	define GB_PIPE_SELECT_BAD_PIPES       14
+#	define GB_PIPE_SELECT_CONFIG_PIPES    18
+
+
+/* Specifies the sizes of the various FIFO`s in the sc/rs. */
+#define GB_FIFO_SIZE1                            0x4070
+/* High water mark for SC input fifo */
+#	define GB_FIFO_SIZE1_SC_HIGHWATER_IFIFO_SHIFT 0
+#	define GB_FIFO_SIZE1_SC_HIGHWATER_IFIFO_MASK  0x0000003f
+/* High water mark for SC input fifo (B) */
+#	define GB_FIFO_SIZE1_SC_HIGHWATER_BFIFO_SHIFT 6
+#	define GB_FIFO_SIZE1_SC_HIGHWATER_BFIFO_MASK  0x00000fc0
+/* High water mark for RS colors' fifo */
+#	define GB_FIFO_SIZE1_SC_HIGHWATER_COL_SHIFT   12
+#	define GB_FIFO_SIZE1_SC_HIGHWATER_COL_MASK    0x0003f000
+/* High water mark for RS textures' fifo */
+#	define GB_FIFO_SIZE1_SC_HIGHWATER_TEX_SHIFT   18
+#	define GB_FIFO_SIZE1_SC_HIGHWATER_TEX_MASK    0x00fc0000
+
+/* This table specifies the source location and format for up to 16 texture
+ * addresses (i[0]:i[15]) and four colors (c[0]:c[3])
+ */
+#define R500_RS_IP_0					0x4074
+#define R500_RS_IP_1					0x4078
+#define R500_RS_IP_2					0x407C
+#define R500_RS_IP_3					0x4080
+#define R500_RS_IP_4					0x4084
+#define R500_RS_IP_5					0x4088
+#define R500_RS_IP_6					0x408C
+#define R500_RS_IP_7					0x4090
+#define R500_RS_IP_8					0x4094
+#define R500_RS_IP_9					0x4098
+#define R500_RS_IP_10					0x409C
+#define R500_RS_IP_11					0x40A0
+#define R500_RS_IP_12					0x40A4
+#define R500_RS_IP_13					0x40A8
+#define R500_RS_IP_14					0x40AC
+#define R500_RS_IP_15					0x40B0
+#define R500_RS_IP_PTR_K0                               62
+#define R500_RS_IP_PTR_K1                               63
+#define R500_RS_IP_TEX_PTR_S_SHIFT 			0
+#define R500_RS_IP_TEX_PTR_T_SHIFT 			6
+#define R500_RS_IP_TEX_PTR_R_SHIFT 			12
+#define R500_RS_IP_TEX_PTR_Q_SHIFT 			18
+#define R500_RS_IP_COL_PTR_SHIFT 			24
+#define R500_RS_IP_COL_FMT_SHIFT 			27
+#	define R500_RS_COL_PTR(x)		        (x << 24)
+#       define R500_RS_COL_FMT(x)                       (x << 27)
+/* gap */
+#define R500_RS_IP_OFFSET_DIS 				(0 << 31)
+#define R500_RS_IP_OFFSET_EN 				(1 << 31)
+
+/* gap */
+
+/* Zero to flush caches. */
+#define R300_TX_INVALTAGS                   0x4100
+#define R300_TX_FLUSH                       0x0
+
+/* The upper enable bits are guessed, based on fglrx reported limits. */
+#define R300_TX_ENABLE                      0x4104
+#       define R300_TX_ENABLE_0                  (1 << 0)
+#       define R300_TX_ENABLE_1                  (1 << 1)
+#       define R300_TX_ENABLE_2                  (1 << 2)
+#       define R300_TX_ENABLE_3                  (1 << 3)
+#       define R300_TX_ENABLE_4                  (1 << 4)
+#       define R300_TX_ENABLE_5                  (1 << 5)
+#       define R300_TX_ENABLE_6                  (1 << 6)
+#       define R300_TX_ENABLE_7                  (1 << 7)
+#       define R300_TX_ENABLE_8                  (1 << 8)
+#       define R300_TX_ENABLE_9                  (1 << 9)
+#       define R300_TX_ENABLE_10                 (1 << 10)
+#       define R300_TX_ENABLE_11                 (1 << 11)
+#       define R300_TX_ENABLE_12                 (1 << 12)
+#       define R300_TX_ENABLE_13                 (1 << 13)
+#       define R300_TX_ENABLE_14                 (1 << 14)
+#       define R300_TX_ENABLE_15                 (1 << 15)
+
+#define R500_TX_FILTER_4		    0x4110
+#	define R500_TX_WEIGHT_1_SHIFT            (0)
+#	define R500_TX_WEIGHT_0_SHIFT            (11)
+#	define R500_TX_WEIGHT_PAIR               (1<<22)
+#	define R500_TX_PHASE_SHIFT               (23)
+#	define R500_TX_DIRECTION_HORIZONTAL	 (0<<27)
+#	define R500_TX_DIRECTION_VERITCAL	 (1<<27)
+
+/* S Texture Coordinate of Vertex 0 for Point texture stuffing (LLC) */
+#define R300_GA_POINT_S0                              0x4200
+
+/* T Texture Coordinate of Vertex 0 for Point texture stuffing (LLC) */
+#define R300_GA_POINT_T0                              0x4204
+
+/* S Texture Coordinate of Vertex 2 for Point texture stuffing (URC) */
+#define R300_GA_POINT_S1                              0x4208
+
+/* T Texture Coordinate of Vertex 2 for Point texture stuffing (URC) */
+#define R300_GA_POINT_T1                              0x420c
+
+/* Specifies amount to shift integer position of vertex (screen space) before
+ * converting to float for triangle stipple.
+ */
+#define R300_GA_TRIANGLE_STIPPLE            0x4214
+#	define R300_GA_TRIANGLE_STIPPLE_X_SHIFT_SHIFT 0
+#	define R300_GA_TRIANGLE_STIPPLE_X_SHIFT_MASK  0x0000000f
+#	define R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_SHIFT 16
+#	define R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_MASK  0x000f0000
+
+/* The pointsize is given in multiples of 6. The pointsize can be enormous:
+ * Clear() renders a single point that fills the entire framebuffer.
+ * 1/2 Height of point; fixed (16.0), subpixel format (1/12 or 1/16, even if in
+ * 8b precision).
+ */
+#define R300_GA_POINT_SIZE                   0x421C
+#       define R300_POINTSIZE_Y_SHIFT         0
+#       define R300_POINTSIZE_Y_MASK          0x0000ffff
+#       define R300_POINTSIZE_X_SHIFT         16
+#       define R300_POINTSIZE_X_MASK          0xffff0000
+#       define R300_POINTSIZE_MAX             (R300_POINTSIZE_Y_MASK / 6)
+
+/* Blue fill color */
+#define R500_GA_FILL_R                                0x4220
+
+/* Blue fill color */
+#define R500_GA_FILL_G                                0x4224
+
+/* Blue fill color */
+#define R500_GA_FILL_B                                0x4228
+
+/* Alpha fill color */
+#define R500_GA_FILL_A                                0x422c
+
+
+/* Specifies maximum and minimum point & sprite sizes for per vertex size
+ * specification. The lower part (15:0) is MIN and (31:16) is max.
+ */
+#define R300_GA_POINT_MINMAX                0x4230
+#       define R300_GA_POINT_MINMAX_MIN_SHIFT          0
+#       define R300_GA_POINT_MINMAX_MIN_MASK           (0xFFFF << 0)
+#       define R300_GA_POINT_MINMAX_MAX_SHIFT          16
+#       define R300_GA_POINT_MINMAX_MAX_MASK           (0xFFFF << 16)
+
+/* 1/2 width of line, in subpixels (1/12 or 1/16 only, even in 8b
+ * subprecision); (16.0) fixed format.
+ *
+ * The line width is given in multiples of 6.
+ * In default mode lines are classified as vertical lines.
+ * HO: horizontal
+ * VE: vertical or horizontal
+ * HO & VE: no classification
+ */
+#define R300_GA_LINE_CNTL                             0x4234
+#       define R300_GA_LINE_CNTL_WIDTH_SHIFT       0
+#       define R300_GA_LINE_CNTL_WIDTH_MASK        0x0000ffff
+#	define R300_GA_LINE_CNTL_END_TYPE_HOR      (0 << 16)
+#	define R300_GA_LINE_CNTL_END_TYPE_VER      (1 << 16)
+#	define R300_GA_LINE_CNTL_END_TYPE_SQR      (2 << 16) /* horizontal or vertical depending upon slope */
+#	define R300_GA_LINE_CNTL_END_TYPE_COMP     (3 << 16) /* Computed (perpendicular to slope) */
+#	define R500_GA_LINE_CNTL_SORT_NO           (0 << 18)
+#	define R500_GA_LINE_CNTL_SORT_MINX_MINY    (1 << 18)
+/** TODO: looks wrong */
+#       define R300_LINESIZE_MAX              (R300_GA_LINE_CNTL_WIDTH_MASK / 6)
+/** TODO: looks wrong */
+#       define R300_LINE_CNT_HO               (1 << 16)
+/** TODO: looks wrong */
+#       define R300_LINE_CNT_VE               (1 << 17)
+
+/* Line Stipple configuration information. */
+#define R300_GA_LINE_STIPPLE_CONFIG                   0x4238
+#	define R300_GA_LINE_STIPPLE_CONFIG_LINE_RESET_NO     (0 << 0)
+#	define R300_GA_LINE_STIPPLE_CONFIG_LINE_RESET_LINE   (1 << 0)
+#	define R300_GA_LINE_STIPPLE_CONFIG_LINE_RESET_PACKET (2 << 0)
+#	define R300_GA_LINE_STIPPLE_CONFIG_STIPPLE_SCALE_SHIFT 2
+#	define R300_GA_LINE_STIPPLE_CONFIG_STIPPLE_SCALE_MASK  0xfffffffc
+
+/* Used to load US instructions and constants */
+#define R500_GA_US_VECTOR_INDEX               0x4250
+#	define R500_GA_US_VECTOR_INDEX_SHIFT       0
+#	define R500_GA_US_VECTOR_INDEX_MASK        0x000000ff
+#	define R500_GA_US_VECTOR_INDEX_TYPE_INSTR  (0 << 16)
+#	define R500_GA_US_VECTOR_INDEX_TYPE_CONST  (1 << 16)
+#	define R500_GA_US_VECTOR_INDEX_CLAMP_NO    (0 << 17)
+#	define R500_GA_US_VECTOR_INDEX_CLAMP_CONST (1 << 17)
+
+/* Data register for loading US instructions and constants */
+#define R500_GA_US_VECTOR_DATA                0x4254
+
+/* Specifies color properties and mappings of textures. */
+#define R500_GA_COLOR_CONTROL_PS3                     0x4258
+#	define R500_TEX0_SHADING_PS3_SOLID       (0 << 0)
+#	define R500_TEX0_SHADING_PS3_FLAT        (1 << 0)
+#	define R500_TEX0_SHADING_PS3_GOURAUD     (2 << 0)
+#	define R500_TEX1_SHADING_PS3_SOLID       (0 << 2)
+#	define R500_TEX1_SHADING_PS3_FLAT        (1 << 2)
+#	define R500_TEX1_SHADING_PS3_GOURAUD     (2 << 2)
+#	define R500_TEX2_SHADING_PS3_SOLID       (0 << 4)
+#	define R500_TEX2_SHADING_PS3_FLAT        (1 << 4)
+#	define R500_TEX2_SHADING_PS3_GOURAUD     (2 << 4)
+#	define R500_TEX3_SHADING_PS3_SOLID       (0 << 6)
+#	define R500_TEX3_SHADING_PS3_FLAT        (1 << 6)
+#	define R500_TEX3_SHADING_PS3_GOURAUD     (2 << 6)
+#	define R500_TEX4_SHADING_PS3_SOLID       (0 << 8)
+#	define R500_TEX4_SHADING_PS3_FLAT        (1 << 8)
+#	define R500_TEX4_SHADING_PS3_GOURAUD     (2 << 8)
+#	define R500_TEX5_SHADING_PS3_SOLID       (0 << 10)
+#	define R500_TEX5_SHADING_PS3_FLAT        (1 << 10)
+#	define R500_TEX5_SHADING_PS3_GOURAUD     (2 << 10)
+#	define R500_TEX6_SHADING_PS3_SOLID       (0 << 12)
+#	define R500_TEX6_SHADING_PS3_FLAT        (1 << 12)
+#	define R500_TEX6_SHADING_PS3_GOURAUD     (2 << 12)
+#	define R500_TEX7_SHADING_PS3_SOLID       (0 << 14)
+#	define R500_TEX7_SHADING_PS3_FLAT        (1 << 14)
+#	define R500_TEX7_SHADING_PS3_GOURAUD     (2 << 14)
+#	define R500_TEX8_SHADING_PS3_SOLID       (0 << 16)
+#	define R500_TEX8_SHADING_PS3_FLAT        (1 << 16)
+#	define R500_TEX8_SHADING_PS3_GOURAUD     (2 << 16)
+#	define R500_TEX9_SHADING_PS3_SOLID       (0 << 18)
+#	define R500_TEX9_SHADING_PS3_FLAT        (1 << 18)
+#	define R500_TEX9_SHADING_PS3_GOURAUD     (2 << 18)
+#	define R500_TEX10_SHADING_PS3_SOLID      (0 << 20)
+#	define R500_TEX10_SHADING_PS3_FLAT       (1 << 20)
+#	define R500_TEX10_SHADING_PS3_GOURAUD    (2 << 20)
+#	define R500_COLOR0_TEX_OVERRIDE_NO       (0 << 22)
+#	define R500_COLOR0_TEX_OVERRIDE_TEX_0    (1 << 22)
+#	define R500_COLOR0_TEX_OVERRIDE_TEX_1    (2 << 22)
+#	define R500_COLOR0_TEX_OVERRIDE_TEX_2    (3 << 22)
+#	define R500_COLOR0_TEX_OVERRIDE_TEX_3    (4 << 22)
+#	define R500_COLOR0_TEX_OVERRIDE_TEX_4    (5 << 22)
+#	define R500_COLOR0_TEX_OVERRIDE_TEX_5    (6 << 22)
+#	define R500_COLOR0_TEX_OVERRIDE_TEX_6    (7 << 22)
+#	define R500_COLOR0_TEX_OVERRIDE_TEX_7    (8 << 22)
+#	define R500_COLOR0_TEX_OVERRIDE_TEX_8_C2 (9 << 22)
+#	define R500_COLOR0_TEX_OVERRIDE_TEX_9_C3 (10 << 22)
+#	define R500_COLOR1_TEX_OVERRIDE_NO       (0 << 26)
+#	define R500_COLOR1_TEX_OVERRIDE_TEX_0    (1 << 26)
+#	define R500_COLOR1_TEX_OVERRIDE_TEX_1    (2 << 26)
+#	define R500_COLOR1_TEX_OVERRIDE_TEX_2    (3 << 26)
+#	define R500_COLOR1_TEX_OVERRIDE_TEX_3    (4 << 26)
+#	define R500_COLOR1_TEX_OVERRIDE_TEX_4    (5 << 26)
+#	define R500_COLOR1_TEX_OVERRIDE_TEX_5    (6 << 26)
+#	define R500_COLOR1_TEX_OVERRIDE_TEX_6    (7 << 26)
+#	define R500_COLOR1_TEX_OVERRIDE_TEX_7    (8 << 26)
+#	define R500_COLOR1_TEX_OVERRIDE_TEX_8_C2 (9 << 26)
+#	define R500_COLOR1_TEX_OVERRIDE_TEX_9_C3 (10 << 26)
+
+/* Returns idle status of various G3D block, captured when GA_IDLE written or
+ * when hard or soft reset asserted.
+ */
+#define R500_GA_IDLE                                  0x425c
+#	define R500_GA_IDLE_PIPE3_Z_IDLE  (0 << 0)
+#	define R500_GA_IDLE_PIPE2_Z_IDLE  (0 << 1)
+#	define R500_GA_IDLE_PIPE3_CD_IDLE (0 << 2)
+#	define R500_GA_IDLE_PIPE2_CD_IDLE (0 << 3)
+#	define R500_GA_IDLE_PIPE3_FG_IDLE (0 << 4)
+#	define R500_GA_IDLE_PIPE2_FG_IDLE (0 << 5)
+#	define R500_GA_IDLE_PIPE3_US_IDLE (0 << 6)
+#	define R500_GA_IDLE_PIPE2_US_IDLE (0 << 7)
+#	define R500_GA_IDLE_PIPE3_SC_IDLE (0 << 8)
+#	define R500_GA_IDLE_PIPE2_SC_IDLE (0 << 9)
+#	define R500_GA_IDLE_PIPE3_RS_IDLE (0 << 10)
+#	define R500_GA_IDLE_PIPE2_RS_IDLE (0 << 11)
+#	define R500_GA_IDLE_PIPE1_Z_IDLE  (0 << 12)
+#	define R500_GA_IDLE_PIPE0_Z_IDLE  (0 << 13)
+#	define R500_GA_IDLE_PIPE1_CD_IDLE (0 << 14)
+#	define R500_GA_IDLE_PIPE0_CD_IDLE (0 << 15)
+#	define R500_GA_IDLE_PIPE1_FG_IDLE (0 << 16)
+#	define R500_GA_IDLE_PIPE0_FG_IDLE (0 << 17)
+#	define R500_GA_IDLE_PIPE1_US_IDLE (0 << 18)
+#	define R500_GA_IDLE_PIPE0_US_IDLE (0 << 19)
+#	define R500_GA_IDLE_PIPE1_SC_IDLE (0 << 20)
+#	define R500_GA_IDLE_PIPE0_SC_IDLE (0 << 21)
+#	define R500_GA_IDLE_PIPE1_RS_IDLE (0 << 22)
+#	define R500_GA_IDLE_PIPE0_RS_IDLE (0 << 23)
+#	define R500_GA_IDLE_SU_IDLE       (0 << 24)
+#	define R500_GA_IDLE_GA_IDLE       (0 << 25)
+#	define R500_GA_IDLE_GA_UNIT2_IDLE (0 << 26)
+
+/* Current value of stipple accumulator. */
+#define R300_GA_LINE_STIPPLE_VALUE            0x4260
+
+/* S Texture Coordinate Value for Vertex 0 of Line (stuff textures -- i.e. AA) */
+#define R300_GA_LINE_S0                               0x4264
+/* S Texture Coordinate Value for Vertex 1 of Lines (V2 of parallelogram -- stuff textures -- i.e. AA) */
+#define R300_GA_LINE_S1                               0x4268
+
+/* GA Input fifo high water marks */
+#define R500_GA_FIFO_CNTL                             0x4270
+#	define R500_GA_FIFO_CNTL_VERTEX_FIFO_MASK   0x00000007
+#	define R500_GA_FIFO_CNTL_VERTEX_FIFO_SHIFT  0
+#	define R500_GA_FIFO_CNTL_VERTEX_INDEX_MASK  0x00000038
+#	define R500_GA_FIFO_CNTL_VERTEX_INDEX_SHIFT 3
+#	define R500_GA_FIFO_CNTL_VERTEX_REG_MASK    0x00003fc0
+#	define R500_GA_FIFO_CNTL_VERTEX_REG_SHIFT   6
+
+/* GA enhance/tweaks */
+#define R300_GA_ENHANCE                               0x4274
+#	define R300_GA_ENHANCE_DEADLOCK_CNTL_NO_EFFECT   (0 << 0)
+#	define R300_GA_ENHANCE_DEADLOCK_CNTL_PREVENT_TCL (1 << 0) /* Prevents TCL interface from deadlocking on GA side. */
+#	define R300_GA_ENHANCE_FASTSYNC_CNTL_NO_EFFECT   (0 << 1)
+#	define R300_GA_ENHANCE_FASTSYNC_CNTL_ENABLE      (1 << 1) /* Enables high-performance register/primitive switching. */
+#	define R500_GA_ENHANCE_REG_READWRITE_NO_EFFECT   (0 << 2) /* R520+ only */
+#	define R500_GA_ENHANCE_REG_READWRITE_ENABLE      (1 << 2) /* R520+ only, Enables GA support of simultaneous register reads and writes. */
+#	define R500_GA_ENHANCE_REG_NOSTALL_NO_EFFECT     (0 << 3)
+#	define R500_GA_ENHANCE_REG_NOSTALL_ENABLE        (1 << 3) /* Enables GA support of no-stall reads for register read back. */
+
+#define R300_GA_COLOR_CONTROL                   0x4278
+#	define R300_GA_COLOR_CONTROL_RGB0_SHADING_SOLID      (0 << 0)
+#	define R300_GA_COLOR_CONTROL_RGB0_SHADING_FLAT       (1 << 0)
+#	define R300_GA_COLOR_CONTROL_RGB0_SHADING_GOURAUD    (2 << 0)
+#	define R300_GA_COLOR_CONTROL_ALPHA0_SHADING_SOLID    (0 << 2)
+#	define R300_GA_COLOR_CONTROL_ALPHA0_SHADING_FLAT     (1 << 2)
+#	define R300_GA_COLOR_CONTROL_ALPHA0_SHADING_GOURAUD  (2 << 2)
+#	define R300_GA_COLOR_CONTROL_RGB1_SHADING_SOLID      (0 << 4)
+#	define R300_GA_COLOR_CONTROL_RGB1_SHADING_FLAT       (1 << 4)
+#	define R300_GA_COLOR_CONTROL_RGB1_SHADING_GOURAUD    (2 << 4)
+#	define R300_GA_COLOR_CONTROL_ALPHA1_SHADING_SOLID    (0 << 6)
+#	define R300_GA_COLOR_CONTROL_ALPHA1_SHADING_FLAT     (1 << 6)
+#	define R300_GA_COLOR_CONTROL_ALPHA1_SHADING_GOURAUD  (2 << 6)
+#	define R300_GA_COLOR_CONTROL_RGB2_SHADING_SOLID      (0 << 8)
+#	define R300_GA_COLOR_CONTROL_RGB2_SHADING_FLAT       (1 << 8)
+#	define R300_GA_COLOR_CONTROL_RGB2_SHADING_GOURAUD    (2 << 8)
+#	define R300_GA_COLOR_CONTROL_ALPHA2_SHADING_SOLID    (0 << 10)
+#	define R300_GA_COLOR_CONTROL_ALPHA2_SHADING_FLAT     (1 << 10)
+#	define R300_GA_COLOR_CONTROL_ALPHA2_SHADING_GOURAUD  (2 << 10)
+#	define R300_GA_COLOR_CONTROL_RGB3_SHADING_SOLID      (0 << 12)
+#	define R300_GA_COLOR_CONTROL_RGB3_SHADING_FLAT       (1 << 12)
+#	define R300_GA_COLOR_CONTROL_RGB3_SHADING_GOURAUD    (2 << 12)
+#	define R300_GA_COLOR_CONTROL_ALPHA3_SHADING_SOLID    (0 << 14)
+#	define R300_GA_COLOR_CONTROL_ALPHA3_SHADING_FLAT     (1 << 14)
+#	define R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD  (2 << 14)
+#	define R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_FIRST  (0 << 16)
+#	define R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_SECOND (1 << 16)
+#	define R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_THIRD  (2 << 16)
+#	define R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST   (3 << 16)
+
+/** TODO: might be candidate for removal */
+#	define R300_RE_SHADE_MODEL_SMOOTH     ( \
+	R300_GA_COLOR_CONTROL_RGB0_SHADING_GOURAUD | R300_GA_COLOR_CONTROL_ALPHA0_SHADING_GOURAUD | \
+	R300_GA_COLOR_CONTROL_RGB1_SHADING_GOURAUD | R300_GA_COLOR_CONTROL_ALPHA1_SHADING_GOURAUD | \
+	R300_GA_COLOR_CONTROL_RGB2_SHADING_GOURAUD | R300_GA_COLOR_CONTROL_ALPHA2_SHADING_GOURAUD | \
+	R300_GA_COLOR_CONTROL_RGB3_SHADING_GOURAUD | R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD | \
+	R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST )
+/** TODO: might be candidate for removal, the GOURAUD stuff also looks buggy to me */
+#	define R300_RE_SHADE_MODEL_FLAT     ( \
+	R300_GA_COLOR_CONTROL_RGB0_SHADING_FLAT | R300_GA_COLOR_CONTROL_ALPHA0_SHADING_FLAT | \
+	R300_GA_COLOR_CONTROL_RGB1_SHADING_FLAT | R300_GA_COLOR_CONTROL_ALPHA1_SHADING_GOURAUD | \
+	R300_GA_COLOR_CONTROL_RGB2_SHADING_FLAT | R300_GA_COLOR_CONTROL_ALPHA2_SHADING_FLAT | \
+	R300_GA_COLOR_CONTROL_RGB3_SHADING_FLAT | R300_GA_COLOR_CONTROL_ALPHA3_SHADING_GOURAUD | \
+	R300_GA_COLOR_CONTROL_PROVOKING_VERTEX_LAST )
+
+/* Specifies red & green components of fill color -- S312 format -- Backwards comp. */
+#define R300_GA_SOLID_RG                         0x427c
+#	define GA_SOLID_RG_COLOR_GREEN_SHIFT 0
+#	define GA_SOLID_RG_COLOR_GREEN_MASK  0x0000ffff
+#	define GA_SOLID_RG_COLOR_RED_SHIFT   16
+#	define GA_SOLID_RG_COLOR_RED_MASK    0xffff0000
+/* Specifies blue & alpha components of fill color -- S312 format -- Backwards comp. */
+#define R300_GA_SOLID_BA                         0x4280
+#	define GA_SOLID_BA_COLOR_ALPHA_SHIFT 0
+#	define GA_SOLID_BA_COLOR_ALPHA_MASK  0x0000ffff
+#	define GA_SOLID_BA_COLOR_BLUE_SHIFT  16
+#	define GA_SOLID_BA_COLOR_BLUE_MASK   0xffff0000
+
+/* Polygon Mode
+ * Dangerous
+ */
+#define R300_GA_POLY_MODE                             0x4288
+#	define R300_GA_POLY_MODE_DISABLE           (0 << 0)
+#	define R300_GA_POLY_MODE_DUAL              (1 << 0) /* send 2 sets of 3 polys with specified poly type */
+/* reserved */
+#	define R300_GA_POLY_MODE_FRONT_PTYPE_POINT (0 << 4)
+#	define R300_GA_POLY_MODE_FRONT_PTYPE_LINE  (1 << 4)
+#	define R300_GA_POLY_MODE_FRONT_PTYPE_TRI   (2 << 4)
+/* reserved */
+#	define R300_GA_POLY_MODE_BACK_PTYPE_POINT  (0 << 7)
+#	define R300_GA_POLY_MODE_BACK_PTYPE_LINE   (1 << 7)
+#	define R300_GA_POLY_MODE_BACK_PTYPE_TRI    (2 << 7)
+/* reserved */
+
+/* Specifies the rouding mode for geometry & color SPFP to FP conversions. */
+#define R300_GA_ROUND_MODE                            0x428c
+#	define R300_GA_ROUND_MODE_GEOMETRY_ROUND_TRUNC   (0 << 0)
+#	define R300_GA_ROUND_MODE_GEOMETRY_ROUND_NEAREST (1 << 0)
+#	define R300_GA_ROUND_MODE_COLOR_ROUND_TRUNC      (0 << 2)
+#	define R300_GA_ROUND_MODE_COLOR_ROUND_NEAREST    (1 << 2)
+#	define R300_GA_ROUND_MODE_RGB_CLAMP_RGB          (0 << 4)
+#	define R300_GA_ROUND_MODE_RGB_CLAMP_FP20         (1 << 4)
+#	define R300_GA_ROUND_MODE_ALPHA_CLAMP_RGB        (0 << 5)
+#	define R300_GA_ROUND_MODE_ALPHA_CLAMP_FP20       (1 << 5)
+#	define R500_GA_ROUND_MODE_GEOMETRY_MASK_SHIFT    6
+#	define R500_GA_ROUND_MODE_GEOMETRY_MASK_MASK     0x000003c0
+
+/* Specifies x & y offsets for vertex data after conversion to FP.
+ * Offsets are in S15 format (subpixels -- 1/12 or 1/16, even in 8b
+ * subprecision).
+ */
+#define R300_GA_OFFSET                                0x4290
+#	define R300_GA_OFFSET_X_OFFSET_SHIFT 0
+#	define R300_GA_OFFSET_X_OFFSET_MASK  0x0000ffff
+#	define R300_GA_OFFSET_Y_OFFSET_SHIFT 16
+#	define R300_GA_OFFSET_Y_OFFSET_MASK  0xffff0000
+
+/* Specifies the scale to apply to fog. */
+#define R300_GA_FOG_SCALE                     0x4294
+/* Specifies the offset to apply to fog. */
+#define R300_GA_FOG_OFFSET                    0x4298
+/* Specifies number of cycles to assert reset, and also causes RB3D soft reset to assert. */
+#define R300_GA_SOFT_RESET                    0x429c
+
+/* Not sure why there are duplicate of factor and constant values.
+ * My best guess so far is that there are seperate zbiases for test and write.
+ * Ordering might be wrong.
+ * Some of the tests indicate that fgl has a fallback implementation of zbias
+ * via pixel shaders.
+ */
+#define R300_SU_TEX_WRAP                      0x42A0
+#define R300_SU_POLY_OFFSET_FRONT_SCALE       0x42A4
+#define R300_SU_POLY_OFFSET_FRONT_OFFSET      0x42A8
+#define R300_SU_POLY_OFFSET_BACK_SCALE        0x42AC
+#define R300_SU_POLY_OFFSET_BACK_OFFSET       0x42B0
+
+/* This register needs to be set to (1<<1) for RV350 to correctly
+ * perform depth test (see --vb-triangles in r300_demo)
+ * Don't know about other chips. - Vladimir
+ * This is set to 3 when GL_POLYGON_OFFSET_FILL is on.
+ * My guess is that there are two bits for each zbias primitive
+ * (FILL, LINE, POINT).
+ *  One to enable depth test and one for depth write.
+ * Yet this doesnt explain why depth writes work ...
+ */
+#define R300_SU_POLY_OFFSET_ENABLE	       0x42B4
+#	define R300_FRONT_ENABLE	       (1 << 0)
+#	define R300_BACK_ENABLE 	       (1 << 1)
+#	define R300_PARA_ENABLE 	       (1 << 2)
+
+#define R300_SU_CULL_MODE                      0x42B8
+#       define R300_CULL_FRONT                   (1 << 0)
+#       define R300_CULL_BACK                    (1 << 1)
+#       define R300_FRONT_FACE_CCW               (0 << 2)
+#       define R300_FRONT_FACE_CW                (1 << 2)
+
+/* SU Depth Scale value */
+#define R300_SU_DEPTH_SCALE                 0x42c0
+/* SU Depth Offset value */
+#define R300_SU_DEPTH_OFFSET                0x42c4
+
+
+/* BEGIN: Rasterization / Interpolators - many guesses */
+
+/*
+ * TC_CNT is the number of incoming texture coordinate sets (i.e. it depends
+ * on the vertex program, *not* the fragment program)
+ */
+#define R300_RS_COUNT                      0x4300
+#       define R300_IT_COUNT_SHIFT               0
+#       define R300_IT_COUNT_MASK                0x0000007f
+#       define R300_IC_COUNT_SHIFT               7
+#       define R300_IC_COUNT_MASK                0x00000780
+#       define R300_W_ADDR_SHIFT                 12
+#       define R300_W_ADDR_MASK                  0x0003f000
+#       define R300_HIRES_DIS                    (0 << 18)
+#       define R300_HIRES_EN                     (1 << 18)
+
+#define R300_RS_INST_COUNT                       0x4304
+#       define R300_RS_INST_COUNT_SHIFT          0
+#       define R300_RS_INST_COUNT_MASK           0x0000000f
+#       define R300_RS_TX_OFFSET_SHIFT           5
+#	define R300_RS_TX_OFFSET_MASK            0x000000e0
+
+/* gap */
+
+/* Only used for texture coordinates.
+ * Use the source field to route texture coordinate input from the
+ * vertex program to the desired interpolator. Note that the source
+ * field is relative to the outputs the vertex program *actually*
+ * writes. If a vertex program only writes texcoord[1], this will
+ * be source index 0.
+ * Set INTERP_USED on all interpolators that produce data used by
+ * the fragment program. INTERP_USED looks like a swizzling mask,
+ * but I haven't seen it used that way.
+ *
+ * Note: The _UNKNOWN constants are always set in their respective
+ * register. I don't know if this is necessary.
+ */
+#define R300_RS_IP_0				        0x4310
+#define R300_RS_IP_1				        0x4314
+#define R300_RS_IP_2				        0x4318
+#define R300_RS_IP_3				        0x431C
+#       define R300_RS_INTERP_SRC_SHIFT          2 /* TODO: check for removal */
+#       define R300_RS_INTERP_SRC_MASK           (7 << 2) /* TODO: check for removal */
+#	define R300_RS_TEX_PTR(x)		        (x << 0)
+#	define R300_RS_COL_PTR(x)		        (x << 6)
+#	define R300_RS_COL_FMT(x)		        (x << 9)
+#	define R300_RS_COL_FMT_RGBA		        0
+#	define R300_RS_COL_FMT_RGB0		        1
+#	define R300_RS_COL_FMT_RGB1		        2
+#	define R300_RS_COL_FMT_000A		        4
+#	define R300_RS_COL_FMT_0000		        5
+#	define R300_RS_COL_FMT_0001		        6
+#	define R300_RS_COL_FMT_111A		        8
+#	define R300_RS_COL_FMT_1110		        9
+#	define R300_RS_COL_FMT_1111		        10
+#	define R300_RS_SEL_S(x)		                (x << 13)
+#	define R300_RS_SEL_T(x)		                (x << 16)
+#	define R300_RS_SEL_R(x)		                (x << 19)
+#	define R300_RS_SEL_Q(x)		                (x << 22)
+#	define R300_RS_SEL_C0		                0
+#	define R300_RS_SEL_C1		                1
+#	define R300_RS_SEL_C2		                2
+#	define R300_RS_SEL_C3		                3
+#	define R300_RS_SEL_K0		                4
+#	define R300_RS_SEL_K1		                5
+
+
+/*  */
+#define R500_RS_INST_0					0x4320
+#define R500_RS_INST_1					0x4324
+#define R500_RS_INST_2					0x4328
+#define R500_RS_INST_3					0x432c
+#define R500_RS_INST_4					0x4330
+#define R500_RS_INST_5					0x4334
+#define R500_RS_INST_6					0x4338
+#define R500_RS_INST_7					0x433c
+#define R500_RS_INST_8					0x4340
+#define R500_RS_INST_9					0x4344
+#define R500_RS_INST_10					0x4348
+#define R500_RS_INST_11					0x434c
+#define R500_RS_INST_12					0x4350
+#define R500_RS_INST_13					0x4354
+#define R500_RS_INST_14					0x4358
+#define R500_RS_INST_15					0x435c
+#define R500_RS_INST_TEX_ID_SHIFT			0
+#define R500_RS_INST_TEX_CN_WRITE			(1 << 4)
+#define R500_RS_INST_TEX_ADDR_SHIFT			5
+#define R500_RS_INST_COL_ID_SHIFT			12
+#define R500_RS_INST_COL_CN_NO_WRITE			(0 << 16)
+#define R500_RS_INST_COL_CN_WRITE			(1 << 16)
+#define R500_RS_INST_COL_CN_WRITE_FBUFFER		(2 << 16)
+#define R500_RS_INST_COL_CN_WRITE_BACKFACE		(3 << 16)
+#define R500_RS_INST_COL_ADDR_SHIFT			18
+#define R500_RS_INST_TEX_ADJ				(1 << 25)
+#define R500_RS_INST_W_CN				(1 << 26)
+
+/* These DWORDs control how vertex data is routed into fragment program
+ * registers, after interpolators.
+ */
+#define R300_RS_INST_0                     0x4330
+#define R300_RS_INST_1                     0x4334
+#define R300_RS_INST_2                     0x4338
+#define R300_RS_INST_3                     0x433C
+#define R300_RS_INST_4                     0x4340
+#define R300_RS_INST_5                     0x4344
+#define R300_RS_INST_6                     0x4348
+#define R300_RS_INST_7                     0x434C
+#	define R300_RS_INST_TEX_ID(x)  		((x) << 0)
+#	define R300_RS_INST_TEX_CN_WRITE 	(1 << 3)
+#	define R300_RS_INST_TEX_ADDR_SHIFT 	6
+#	define R300_RS_INST_COL_ID(x)		((x) << 11)
+#	define R300_RS_INST_COL_CN_WRITE	(1 << 14)
+#	define R300_RS_INST_COL_ADDR_SHIFT	17
+#	define R300_RS_INST_TEX_ADJ		(1 << 22)
+#	define R300_RS_COL_BIAS_UNUSED_SHIFT    23
+
+/* END: Rasterization / Interpolators - many guesses */
+
+/* Hierarchical Z Enable */
+#define R300_SC_HYPERZ                   0x43a4
+#	define R300_SC_HYPERZ_DISABLE     (0 << 0)
+#	define R300_SC_HYPERZ_ENABLE      (1 << 0)
+#	define R300_SC_HYPERZ_MIN         (0 << 1)
+#	define R300_SC_HYPERZ_MAX         (1 << 1)
+#	define R300_SC_HYPERZ_ADJ_256     (0 << 2)
+#	define R300_SC_HYPERZ_ADJ_128     (1 << 2)
+#	define R300_SC_HYPERZ_ADJ_64      (2 << 2)
+#	define R300_SC_HYPERZ_ADJ_32      (3 << 2)
+#	define R300_SC_HYPERZ_ADJ_16      (4 << 2)
+#	define R300_SC_HYPERZ_ADJ_8       (5 << 2)
+#	define R300_SC_HYPERZ_ADJ_4       (6 << 2)
+#	define R300_SC_HYPERZ_ADJ_2       (7 << 2)
+#	define R300_SC_HYPERZ_HZ_Z0MIN_NO (0 << 5)
+#	define R300_SC_HYPERZ_HZ_Z0MIN    (1 << 5)
+#	define R300_SC_HYPERZ_HZ_Z0MAX_NO (0 << 6)
+#	define R300_SC_HYPERZ_HZ_Z0MAX    (1 << 6)
+
+#define R300_SC_EDGERULE                 0x43a8
+
+/* BEGIN: Scissors and cliprects */
+
+/* There are four clipping rectangles. Their corner coordinates are inclusive.
+ * Every pixel is assigned a number from 0 and 15 by setting bits 0-3 depending
+ * on whether the pixel is inside cliprects 0-3, respectively. For example,
+ * if a pixel is inside cliprects 0 and 1, but outside 2 and 3, it is assigned
+ * the number 3 (binary 0011).
+ * Iff the bit corresponding to the pixel's number in RE_CLIPRECT_CNTL is set,
+ * the pixel is rasterized.
+ *
+ * In addition to this, there is a scissors rectangle. Only pixels inside the
+ * scissors rectangle are drawn. (coordinates are inclusive)
+ *
+ * For some reason, the top-left corner of the framebuffer is at (1440, 1440)
+ * for the purpose of clipping and scissors.
+ */
+#define R300_SC_CLIPRECT_TL_0               0x43B0
+#define R300_SC_CLIPRECT_BR_0               0x43B4
+#define R300_SC_CLIPRECT_TL_1               0x43B8
+#define R300_SC_CLIPRECT_BR_1               0x43BC
+#define R300_SC_CLIPRECT_TL_2               0x43C0
+#define R300_SC_CLIPRECT_BR_2               0x43C4
+#define R300_SC_CLIPRECT_TL_3               0x43C8
+#define R300_SC_CLIPRECT_BR_3               0x43CC
+#       define R300_CLIPRECT_OFFSET              1440
+#       define R300_CLIPRECT_MASK                0x1FFF
+#       define R300_CLIPRECT_X_SHIFT             0
+#       define R300_CLIPRECT_X_MASK              (0x1FFF << 0)
+#       define R300_CLIPRECT_Y_SHIFT             13
+#       define R300_CLIPRECT_Y_MASK              (0x1FFF << 13)
+#define R300_SC_CLIP_RULE                   0x43D0
+#       define R300_CLIP_OUT                     (1 << 0)
+#       define R300_CLIP_0                       (1 << 1)
+#       define R300_CLIP_1                       (1 << 2)
+#       define R300_CLIP_10                      (1 << 3)
+#       define R300_CLIP_2                       (1 << 4)
+#       define R300_CLIP_20                      (1 << 5)
+#       define R300_CLIP_21                      (1 << 6)
+#       define R300_CLIP_210                     (1 << 7)
+#       define R300_CLIP_3                       (1 << 8)
+#       define R300_CLIP_30                      (1 << 9)
+#       define R300_CLIP_31                      (1 << 10)
+#       define R300_CLIP_310                     (1 << 11)
+#       define R300_CLIP_32                      (1 << 12)
+#       define R300_CLIP_320                     (1 << 13)
+#       define R300_CLIP_321                     (1 << 14)
+#       define R300_CLIP_3210                    (1 << 15)
+
+/* gap */
+
+#define R300_SC_SCISSORS_TL                 0x43E0
+#define R300_SC_SCISSORS_BR                 0x43E4
+#       define R300_SCISSORS_OFFSET              1440
+#       define R300_SCISSORS_X_SHIFT             0
+#       define R300_SCISSORS_X_MASK              (0x1FFF << 0)
+#       define R300_SCISSORS_Y_SHIFT             13
+#       define R300_SCISSORS_Y_MASK              (0x1FFF << 13)
+
+/* Screen door sample mask */
+#define R300_SC_SCREENDOOR                 0x43e8
+
+/* END: Scissors and cliprects */
+
+/* BEGIN: Texture specification */
+
+/*
+ * The texture specification dwords are grouped by meaning and not by texture
+ * unit. This means that e.g. the offset for texture image unit N is found in
+ * register TX_OFFSET_0 + (4*N)
+ */
+#define R300_TX_FILTER0_0                        0x4400
+#define R300_TX_FILTER0_1                        0x4404
+#define R300_TX_FILTER0_2                        0x4408
+#define R300_TX_FILTER0_3                        0x440c
+#define R300_TX_FILTER0_4                        0x4410
+#define R300_TX_FILTER0_5                        0x4414
+#define R300_TX_FILTER0_6                        0x4418
+#define R300_TX_FILTER0_7                        0x441c
+#define R300_TX_FILTER0_8                        0x4420
+#define R300_TX_FILTER0_9                        0x4424
+#define R300_TX_FILTER0_10                       0x4428
+#define R300_TX_FILTER0_11                       0x442c
+#define R300_TX_FILTER0_12                       0x4430
+#define R300_TX_FILTER0_13                       0x4434
+#define R300_TX_FILTER0_14                       0x4438
+#define R300_TX_FILTER0_15                       0x443c
+#       define R300_TX_REPEAT                    0
+#       define R300_TX_MIRRORED                  1
+#       define R300_TX_CLAMP_TO_EDGE             2
+#	define R300_TX_MIRROR_ONCE_TO_EDGE       3
+#       define R300_TX_CLAMP                     4
+#	define R300_TX_MIRROR_ONCE               5
+#       define R300_TX_CLAMP_TO_BORDER           6
+#	define R300_TX_MIRROR_ONCE_TO_BORDER     7
+#       define R300_TX_WRAP_S_SHIFT              0
+#       define R300_TX_WRAP_S_MASK               (7 << 0)
+#       define R300_TX_WRAP_T_SHIFT              3
+#       define R300_TX_WRAP_T_MASK               (7 << 3)
+#       define R300_TX_WRAP_R_SHIFT              6
+#       define R300_TX_WRAP_R_MASK               (7 << 6)
+#	define R300_TX_MAG_FILTER_4              (0 << 9)
+#       define R300_TX_MAG_FILTER_NEAREST        (1 << 9)
+#       define R300_TX_MAG_FILTER_LINEAR         (2 << 9)
+#       define R300_TX_MAG_FILTER_ANISO          (3 << 9)
+#       define R300_TX_MAG_FILTER_MASK           (3 << 9)
+#       define R300_TX_MIN_FILTER_NEAREST        (1 << 11)
+#       define R300_TX_MIN_FILTER_LINEAR         (2 << 11)
+#	define R300_TX_MIN_FILTER_ANISO          (3 << 11)
+#	define R300_TX_MIN_FILTER_MASK           (3 << 11)
+#	define R300_TX_MIN_FILTER_MIP_NONE       (0 << 13)
+#	define R300_TX_MIN_FILTER_MIP_NEAREST    (1 << 13)
+#	define R300_TX_MIN_FILTER_MIP_LINEAR     (2 << 13)
+#	define R300_TX_MIN_FILTER_MIP_MASK       (3 << 13)
+#	define R300_TX_MAX_ANISO_1_TO_1          (0 << 21)
+#	define R300_TX_MAX_ANISO_2_TO_1          (1 << 21)
+#	define R300_TX_MAX_ANISO_4_TO_1          (2 << 21)
+#	define R300_TX_MAX_ANISO_8_TO_1          (3 << 21)
+#	define R300_TX_MAX_ANISO_16_TO_1         (4 << 21)
+#	define R300_TX_MAX_ANISO_MASK            (7 << 21)
+
+#define R300_TX_FILTER1_0                      0x4440
+#	define R300_CHROMA_KEY_MODE_DISABLE    0
+#	define R300_CHROMA_KEY_FORCE	       1
+#	define R300_CHROMA_KEY_BLEND           2
+#	define R300_MC_ROUND_NORMAL            (0<<2)
+#	define R300_MC_ROUND_MPEG4             (1<<2)
+#	define R300_LOD_BIAS_SHIFT             3
+#	define R300_LOD_BIAS_MASK	       0x1ff8
+#	define R300_EDGE_ANISO_EDGE_DIAG       (0<<13)
+#	define R300_EDGE_ANISO_EDGE_ONLY       (1<<13)
+#	define R300_MC_COORD_TRUNCATE_DISABLE  (0<<14)
+#	define R300_MC_COORD_TRUNCATE_MPEG     (1<<14)
+#	define R300_TX_TRI_PERF_0_8            (0<<15)
+#	define R300_TX_TRI_PERF_1_8            (1<<15)
+#	define R300_TX_TRI_PERF_1_4            (2<<15)
+#	define R300_TX_TRI_PERF_3_8            (3<<15)
+#	define R300_ANISO_THRESHOLD_MASK       (7<<17)
+
+#	define R500_MACRO_SWITCH               (1<<22)
+#	define R500_BORDER_FIX                 (1<<31)
+
+#define R300_TX_SIZE_0                      0x4480
+#       define R300_TX_WIDTHMASK_SHIFT           0
+#       define R300_TX_WIDTHMASK_MASK            (2047 << 0)
+#       define R300_TX_HEIGHTMASK_SHIFT          11
+#       define R300_TX_HEIGHTMASK_MASK           (2047 << 11)
+#	define R300_TX_DEPTHMASK_SHIFT		 22
+#	define R300_TX_DEPTHMASK_MASK		 (0xf << 22)
+#       define R300_TX_MAX_MIP_LEVEL_SHIFT       26
+#       define R300_TX_MAX_MIP_LEVEL_MASK        (0xf << 26)
+#       define R300_TX_SIZE_PROJECTED            (1<<30)
+#       define R300_TX_SIZE_TXPITCH_EN           (1<<31)
+#define R300_TX_FORMAT_0                    0x44C0
+	/* The interpretation of the format word by Wladimir van der Laan */
+	/* The X, Y, Z and W refer to the layout of the components.
+	   They are given meanings as R, G, B and Alpha by the swizzle
+	   specification */
+#	define R300_TX_FORMAT_X8		    0x0
+#	define R500_TX_FORMAT_X1		    0x0 // bit set in format 2
+#	define R300_TX_FORMAT_X16		    0x1
+#	define R500_TX_FORMAT_X1_REV		    0x0 // bit set in format 2
+#	define R300_TX_FORMAT_Y4X4		    0x2
+#	define R300_TX_FORMAT_Y8X8		    0x3
+#	define R300_TX_FORMAT_Y16X16		    0x4
+#	define R300_TX_FORMAT_Z3Y3X2		    0x5
+#	define R300_TX_FORMAT_Z5Y6X5		    0x6
+#	define R300_TX_FORMAT_Z6Y5X5		    0x7
+#	define R300_TX_FORMAT_Z11Y11X10		    0x8
+#	define R300_TX_FORMAT_Z10Y11X11		    0x9
+#	define R300_TX_FORMAT_W4Z4Y4X4		    0xA
+#	define R300_TX_FORMAT_W1Z5Y5X5		    0xB
+#	define R300_TX_FORMAT_W8Z8Y8X8		    0xC
+#	define R300_TX_FORMAT_W2Z10Y10X10	    0xD
+#	define R300_TX_FORMAT_W16Z16Y16X16	    0xE
+#	define R300_TX_FORMAT_DXT1	    	    0xF
+#	define R300_TX_FORMAT_DXT3	    	    0x10
+#	define R300_TX_FORMAT_DXT5	    	    0x11
+#	define R300_TX_FORMAT_D3DMFT_CxV8U8	    0x12     /* no swizzle */
+#	define R300_TX_FORMAT_A8R8G8B8	    	    0x13     /* no swizzle */
+#	define R300_TX_FORMAT_B8G8_B8G8	    	    0x14     /* no swizzle */
+#	define R300_TX_FORMAT_G8R8_G8B8	    	    0x15     /* no swizzle */
+
+	/* These two values are wrong, but they're the only values that
+	 * produce any even vaguely correct results.  Can r300 only do 16-bit
+	 * depth textures?
+	 */
+#	define R300_TX_FORMAT_X24_Y8	    	    0x1e
+#	define R300_TX_FORMAT_X32	    	    0x1e
+
+	/* 0x16 - some 16 bit green format.. ?? */
+#	define R300_TX_FORMAT_3D		   (1 << 25)
+#	define R300_TX_FORMAT_CUBIC_MAP		   (2 << 25)
+
+	/* gap */
+	/* Floating point formats */
+	/* Note - hardware supports both 16 and 32 bit floating point */
+#	define R300_TX_FORMAT_FL_I16	    	    0x18
+#	define R300_TX_FORMAT_FL_I16A16	    	    0x19
+#	define R300_TX_FORMAT_FL_R16G16B16A16	    0x1A
+#	define R300_TX_FORMAT_FL_I32	    	    0x1B
+#	define R300_TX_FORMAT_FL_I32A32	    	    0x1C
+#	define R300_TX_FORMAT_FL_R32G32B32A32	    0x1D
+	/* alpha modes, convenience mostly */
+	/* if you have alpha, pick constant appropriate to the
+	   number of channels (1 for I8, 2 for I8A8, 4 for R8G8B8A8, etc */
+# 	define R300_TX_FORMAT_ALPHA_1CH		    0x000
+# 	define R300_TX_FORMAT_ALPHA_2CH		    0x200
+# 	define R300_TX_FORMAT_ALPHA_4CH		    0x600
+# 	define R300_TX_FORMAT_ALPHA_NONE	    0xA00
+	/* Swizzling */
+	/* constants */
+#	define R300_TX_FORMAT_X		0
+#	define R300_TX_FORMAT_Y		1
+#	define R300_TX_FORMAT_Z		2
+#	define R300_TX_FORMAT_W		3
+#	define R300_TX_FORMAT_ZERO	4
+#	define R300_TX_FORMAT_ONE	5
+	/* 2.0*Z, everything above 1.0 is set to 0.0 */
+#	define R300_TX_FORMAT_CUT_Z	6
+	/* 2.0*W, everything above 1.0 is set to 0.0 */
+#	define R300_TX_FORMAT_CUT_W	7
+
+#	define R300_TX_FORMAT_B_SHIFT	18
+#	define R300_TX_FORMAT_G_SHIFT	15
+#	define R300_TX_FORMAT_R_SHIFT	12
+#	define R300_TX_FORMAT_A_SHIFT	9
+	/* Convenience macro to take care of layout and swizzling */
+#	define R300_EASY_TX_FORMAT(B, G, R, A, FMT)	(		\
+		((R300_TX_FORMAT_##B)<<R300_TX_FORMAT_B_SHIFT)		\
+		| ((R300_TX_FORMAT_##G)<<R300_TX_FORMAT_G_SHIFT)	\
+		| ((R300_TX_FORMAT_##R)<<R300_TX_FORMAT_R_SHIFT)	\
+		| ((R300_TX_FORMAT_##A)<<R300_TX_FORMAT_A_SHIFT)	\
+		| (R300_TX_FORMAT_##FMT)				\
+		)
+	/* These can be ORed with result of R300_EASY_TX_FORMAT()
+	   We don't really know what they do. Take values from a
+           constant color ? */
+#	define R300_TX_FORMAT_CONST_X		(1<<5)
+#	define R300_TX_FORMAT_CONST_Y		(2<<5)
+#	define R300_TX_FORMAT_CONST_Z		(4<<5)
+#	define R300_TX_FORMAT_CONST_W		(8<<5)
+
+#	define R300_TX_FORMAT_YUV_MODE		0x00800000
+
+#define R300_TX_FORMAT2_0		    0x4500 /* obvious missing in gap */
+#       define R300_TX_PITCHMASK_SHIFT           0
+#       define R300_TX_PITCHMASK_MASK            (2047 << 0)
+#	define R500_TXFORMAT_MSB		 (1 << 14)
+#	define R500_TXWIDTH_BIT11	         (1 << 15)
+#	define R500_TXHEIGHT_BIT11	         (1 << 16)
+#	define R500_POW2FIX2FLT			 (1 << 17)
+#	define R500_SEL_FILTER4_TC0		 (0 << 18)
+#	define R500_SEL_FILTER4_TC1		 (1 << 18)
+#	define R500_SEL_FILTER4_TC2		 (2 << 18)
+#	define R500_SEL_FILTER4_TC3		 (3 << 18)
+
+#define R300_TX_OFFSET_0                    0x4540
+#define R300_TX_OFFSET_1                    0x4544
+#define R300_TX_OFFSET_2                    0x4548
+#define R300_TX_OFFSET_3                    0x454C
+#define R300_TX_OFFSET_4                    0x4550
+#define R300_TX_OFFSET_5                    0x4554
+#define R300_TX_OFFSET_6                    0x4558
+#define R300_TX_OFFSET_7                    0x455C
+	/* BEGIN: Guess from R200 */
+#       define R300_TXO_ENDIAN_NO_SWAP           (0 << 0)
+#       define R300_TXO_ENDIAN_BYTE_SWAP         (1 << 0)
+#       define R300_TXO_ENDIAN_WORD_SWAP         (2 << 0)
+#       define R300_TXO_ENDIAN_HALFDW_SWAP       (3 << 0)
+#       define R300_TXO_MACRO_TILE               (1 << 2)
+#       define R300_TXO_MICRO_TILE_LINEAR        (0 << 3)
+#       define R300_TXO_MICRO_TILE               (1 << 3)
+#       define R300_TXO_MICRO_TILE_SQUARE        (2 << 3)
+#       define R300_TXO_OFFSET_MASK              0xffffffe0
+#       define R300_TXO_OFFSET_SHIFT             5
+	/* END: Guess from R200 */
+
+/* 32 bit chroma key */
+#define R300_TX_CHROMA_KEY_0                      0x4580
+#define R300_TX_CHROMA_KEY_1                      0x4584
+#define R300_TX_CHROMA_KEY_2                      0x4588
+#define R300_TX_CHROMA_KEY_3                      0x458c
+#define R300_TX_CHROMA_KEY_4                      0x4590
+#define R300_TX_CHROMA_KEY_5                      0x4594
+#define R300_TX_CHROMA_KEY_6                      0x4598
+#define R300_TX_CHROMA_KEY_7                      0x459c
+#define R300_TX_CHROMA_KEY_8                      0x45a0
+#define R300_TX_CHROMA_KEY_9                      0x45a4
+#define R300_TX_CHROMA_KEY_10                     0x45a8
+#define R300_TX_CHROMA_KEY_11                     0x45ac
+#define R300_TX_CHROMA_KEY_12                     0x45b0
+#define R300_TX_CHROMA_KEY_13                     0x45b4
+#define R300_TX_CHROMA_KEY_14                     0x45b8
+#define R300_TX_CHROMA_KEY_15                     0x45bc
+/* ff00ff00 == { 0, 1.0, 0, 1.0 } */
+
+/* Border Color */
+#define R300_TX_BORDER_COLOR_0              0x45c0
+#define R300_TX_BORDER_COLOR_1              0x45c4
+#define R300_TX_BORDER_COLOR_2              0x45c8
+#define R300_TX_BORDER_COLOR_3              0x45cc
+#define R300_TX_BORDER_COLOR_4              0x45d0
+#define R300_TX_BORDER_COLOR_5              0x45d4
+#define R300_TX_BORDER_COLOR_6              0x45d8
+#define R300_TX_BORDER_COLOR_7              0x45dc
+#define R300_TX_BORDER_COLOR_8              0x45e0
+#define R300_TX_BORDER_COLOR_9              0x45e4
+#define R300_TX_BORDER_COLOR_10             0x45e8
+#define R300_TX_BORDER_COLOR_11             0x45ec
+#define R300_TX_BORDER_COLOR_12             0x45f0
+#define R300_TX_BORDER_COLOR_13             0x45f4
+#define R300_TX_BORDER_COLOR_14             0x45f8
+#define R300_TX_BORDER_COLOR_15             0x45fc
+
+
+/* END: Texture specification */
+
+/* BEGIN: Fragment program instruction set */
+
+/* Fragment programs are written directly into register space.
+ * There are separate instruction streams for texture instructions and ALU
+ * instructions.
+ * In order to synchronize these streams, the program is divided into up
+ * to 4 nodes. Each node begins with a number of TEX operations, followed
+ * by a number of ALU operations.
+ * The first node can have zero TEX ops, all subsequent nodes must have at
+ * least
+ * one TEX ops.
+ * All nodes must have at least one ALU op.
+ *
+ * The index of the last node is stored in PFS_CNTL_0: A value of 0 means
+ * 1 node, a value of 3 means 4 nodes.
+ * The total amount of instructions is defined in PFS_CNTL_2. The offsets are
+ * offsets into the respective instruction streams, while *_END points to the
+ * last instruction relative to this offset.
+ */
+#define R300_US_CONFIG                      0x4600
+#       define R300_PFS_CNTL_LAST_NODES_SHIFT    0
+#       define R300_PFS_CNTL_LAST_NODES_MASK     (3 << 0)
+#       define R300_PFS_CNTL_FIRST_NODE_HAS_TEX  (1 << 3)
+#define R300_US_PIXSIZE                     0x4604
+/* There is an unshifted value here which has so far always been equal to the
+ * index of the highest used temporary register.
+ */
+#define R300_US_CODE_OFFSET                 0x4608
+#       define R300_PFS_CNTL_ALU_OFFSET_SHIFT    0
+#       define R300_PFS_CNTL_ALU_OFFSET_MASK     (63 << 0)
+#       define R300_PFS_CNTL_ALU_END_SHIFT       6
+#       define R300_PFS_CNTL_ALU_END_MASK        (63 << 6)
+#       define R300_PFS_CNTL_TEX_OFFSET_SHIFT    13
+#       define R300_PFS_CNTL_TEX_OFFSET_MASK     (31 << 13)
+#       define R300_PFS_CNTL_TEX_END_SHIFT       18
+#       define R300_PFS_CNTL_TEX_END_MASK        (31 << 18)
+
+/* gap */
+
+/* Nodes are stored backwards. The last active node is always stored in
+ * PFS_NODE_3.
+ * Example: In a 2-node program, NODE_0 and NODE_1 are set to 0. The
+ * first node is stored in NODE_2, the second node is stored in NODE_3.
+ *
+ * Offsets are relative to the master offset from PFS_CNTL_2.
+ */
+#define R300_US_CODE_ADDR_0                 0x4610
+#define R300_US_CODE_ADDR_1                 0x4614
+#define R300_US_CODE_ADDR_2                 0x4618
+#define R300_US_CODE_ADDR_3                 0x461C
+#       define R300_ALU_START_SHIFT         0
+#       define R300_ALU_START_MASK          (63 << 0)
+#       define R300_ALU_SIZE_SHIFT          6
+#       define R300_ALU_SIZE_MASK           (63 << 6)
+#       define R300_TEX_START_SHIFT         12
+#       define R300_TEX_START_MASK          (31 << 12)
+#       define R300_TEX_SIZE_SHIFT          17
+#       define R300_TEX_SIZE_MASK           (31 << 17)
+#	define R300_RGBA_OUT                (1 << 22)
+#	define R300_W_OUT                   (1 << 23)
+
+/* TEX
+ * As far as I can tell, texture instructions cannot write into output
+ * registers directly. A subsequent ALU instruction is always necessary,
+ * even if it's just MAD o0, r0, 1, 0
+ */
+#define R300_US_TEX_INST_0                  0x4620
+#	define R300_SRC_ADDR_SHIFT          0
+#	define R300_SRC_ADDR_MASK           (31 << 0)
+#	define R300_DST_ADDR_SHIFT          6
+#	define R300_DST_ADDR_MASK           (31 << 6)
+#	define R300_TEX_ID_SHIFT            11
+#       define R300_TEX_ID_MASK             (15 << 11)
+#	define R300_TEX_INST_SHIFT		15
+#		define R300_TEX_OP_NOP	        0
+#		define R300_TEX_OP_LD	        1
+#		define R300_TEX_OP_KIL	        2
+#		define R300_TEX_OP_TXP	        3
+#		define R300_TEX_OP_TXB	        4
+#	define R300_TEX_INST_MASK               (7 << 15)
+
+/* Output format from the unfied shader */
+#define R300_US_OUT_FMT_0                   0x46A4
+#	define R300_US_OUT_FMT_C4_8         (0 << 0)
+#	define R300_US_OUT_FMT_C4_10        (1 << 0)
+#	define R300_US_OUT_FMT_C4_10_GAMMA  (2 << 0)
+#	define R300_US_OUT_FMT_C_16         (3 << 0)
+#	define R300_US_OUT_FMT_C2_16        (4 << 0)
+#	define R300_US_OUT_FMT_C4_16        (5 << 0)
+#	define R300_US_OUT_FMT_C_16_MPEG    (6 << 0)
+#	define R300_US_OUT_FMT_C2_16_MPEG   (7 << 0)
+#	define R300_US_OUT_FMT_C2_4         (8 << 0)
+#	define R300_US_OUT_FMT_C_3_3_2      (9 << 0)
+#	define R300_US_OUT_FMT_C_6_5_6      (10 << 0)
+#	define R300_US_OUT_FMT_C_11_11_10   (11 << 0)
+#	define R300_US_OUT_FMT_C_10_11_11   (12 << 0)
+#	define R300_US_OUT_FMT_C_2_10_10_10 (13 << 0)
+/* reserved */
+#	define R300_US_OUT_FMT_UNUSED       (15 << 0)
+#	define R300_US_OUT_FMT_C_16_FP      (16 << 0)
+#	define R300_US_OUT_FMT_C2_16_FP     (17 << 0)
+#	define R300_US_OUT_FMT_C4_16_FP     (18 << 0)
+#	define R300_US_OUT_FMT_C_32_FP      (19 << 0)
+#	define R300_US_OUT_FMT_C2_32_FP     (20 << 0)
+#	define R300_US_OUT_FMT_C4_32_FP     (21 << 0)
+#   define R300_C0_SEL_A				(0 << 8)
+#   define R300_C0_SEL_R				(1 << 8)
+#   define R300_C0_SEL_G				(2 << 8)
+#   define R300_C0_SEL_B				(3 << 8)
+#   define R300_C1_SEL_A				(0 << 10)
+#   define R300_C1_SEL_R				(1 << 10)
+#   define R300_C1_SEL_G				(2 << 10)
+#   define R300_C1_SEL_B				(3 << 10)
+#   define R300_C2_SEL_A				(0 << 12)
+#   define R300_C2_SEL_R				(1 << 12)
+#   define R300_C2_SEL_G				(2 << 12)
+#   define R300_C2_SEL_B				(3 << 12)
+#   define R300_C3_SEL_A				(0 << 14)
+#   define R300_C3_SEL_R				(1 << 14)
+#   define R300_C3_SEL_G				(2 << 14)
+#   define R300_C3_SEL_B				(3 << 14)
+#   define R300_OUT_SIGN(x)				(x << 16)
+
+/* ALU
+ * The ALU instructions register blocks are enumerated according to the order
+ * in which fglrx. I assume there is space for 64 instructions, since
+ * each block has space for a maximum of 64 DWORDs, and this matches reported
+ * native limits.
+ *
+ * The basic functional block seems to be one MAD for each color and alpha,
+ * and an adder that adds all components after the MUL.
+ *  - ADD, MUL, MAD etc.: use MAD with appropriate neutral operands
+ *  - DP4: Use OUTC_DP4, OUTA_DP4
+ *  - DP3: Use OUTC_DP3, OUTA_DP4, appropriate alpha operands
+ *  - DPH: Use OUTC_DP4, OUTA_DP4, appropriate alpha operands
+ *  - CMPH: If ARG2 > 0.5, return ARG0, else return ARG1
+ *  - CMP: If ARG2 < 0, return ARG1, else return ARG0
+ *  - FLR: use FRC+MAD
+ *  - XPD: use MAD+MAD
+ *  - SGE, SLT: use MAD+CMP
+ *  - RSQ: use ABS modifier for argument
+ *  - Use OUTC_REPL_ALPHA to write results of an alpha-only operation
+ *    (e.g. RCP) into color register
+ *  - apparently, there's no quick DST operation
+ *  - fglrx set FPI2_UNKNOWN_31 on a "MAD fragment.color, tmp0, tmp1, tmp2"
+ *  - fglrx set FPI2_UNKNOWN_31 on a "MAX r2, r1, c0"
+ *  - fglrx once set FPI0_UNKNOWN_31 on a "FRC r1, r1"
+ *
+ * Operand selection
+ * First stage selects three sources from the available registers and
+ * constant parameters. This is defined in INSTR1 (color) and INSTR3 (alpha).
+ * fglrx sorts the three source fields: Registers before constants,
+ * lower indices before higher indices; I do not know whether this is
+ * necessary.
+ *
+ * fglrx fills unused sources with "read constant 0"
+ * According to specs, you cannot select more than two different constants.
+ *
+ * Second stage selects the operands from the sources. This is defined in
+ * INSTR0 (color) and INSTR2 (alpha). You can also select the special constants
+ * zero and one.
+ * Swizzling and negation happens in this stage, as well.
+ *
+ * Important: Color and alpha seem to be mostly separate, i.e. their sources
+ * selection appears to be fully independent (the register storage is probably
+ * physically split into a color and an alpha section).
+ * However (because of the apparent physical split), there is some interaction
+ * WRT swizzling. If, for example, you want to load an R component into an
+ * Alpha operand, this R component is taken from a *color* source, not from
+ * an alpha source. The corresponding register doesn't even have to appear in
+ * the alpha sources list. (I hope this all makes sense to you)
+ *
+ * Destination selection
+ * The destination register index is in FPI1 (color) and FPI3 (alpha)
+ * together with enable bits.
+ * There are separate enable bits for writing into temporary registers
+ * (DSTC_REG_* /DSTA_REG) and and program output registers (DSTC_OUTPUT_*
+ * /DSTA_OUTPUT). You can write to both at once, or not write at all (the
+ * same index must be used for both).
+ *
+ * Note: There is a special form for LRP
+ *  - Argument order is the same as in ARB_fragment_program.
+ *  - Operation is MAD
+ *  - ARG1 is set to ARGC_SRC1C_LRP/ARGC_SRC1A_LRP
+ *  - Set FPI0/FPI2_SPECIAL_LRP
+ * Arbitrary LRP (including support for swizzling) requires vanilla MAD+MAD
+ */
+#define R300_US_ALU_RGB_ADDR_0                   0x46C0
+#       define R300_ALU_SRC0C_SHIFT             0
+#       define R300_ALU_SRC0C_MASK              (31 << 0)
+#       define R300_ALU_SRC0C_CONST             (1 << 5)
+#       define R300_ALU_SRC1C_SHIFT             6
+#       define R300_ALU_SRC1C_MASK              (31 << 6)
+#       define R300_ALU_SRC1C_CONST             (1 << 11)
+#       define R300_ALU_SRC2C_SHIFT             12
+#       define R300_ALU_SRC2C_MASK              (31 << 12)
+#       define R300_ALU_SRC2C_CONST             (1 << 17)
+#       define R300_ALU_SRC_MASK                0x0003ffff
+#       define R300_ALU_DSTC_SHIFT              18
+#       define R300_ALU_DSTC_MASK               (31 << 18)
+#		define R300_ALU_DSTC_REG_MASK_SHIFT     23
+#       define R300_ALU_DSTC_REG_X              (1 << 23)
+#       define R300_ALU_DSTC_REG_Y              (1 << 24)
+#       define R300_ALU_DSTC_REG_Z              (1 << 25)
+#		define R300_ALU_DSTC_OUTPUT_MASK_SHIFT  26
+#       define R300_ALU_DSTC_OUTPUT_X           (1 << 26)
+#       define R300_ALU_DSTC_OUTPUT_Y           (1 << 27)
+#       define R300_ALU_DSTC_OUTPUT_Z           (1 << 28)
+
+#define R300_US_ALU_ALPHA_ADDR_0                 0x47C0
+#       define R300_ALU_SRC0A_SHIFT             0
+#       define R300_ALU_SRC0A_MASK              (31 << 0)
+#       define R300_ALU_SRC0A_CONST             (1 << 5)
+#       define R300_ALU_SRC1A_SHIFT             6
+#       define R300_ALU_SRC1A_MASK              (31 << 6)
+#       define R300_ALU_SRC1A_CONST             (1 << 11)
+#       define R300_ALU_SRC2A_SHIFT             12
+#       define R300_ALU_SRC2A_MASK              (31 << 12)
+#       define R300_ALU_SRC2A_CONST             (1 << 17)
+#       define R300_ALU_SRC_MASK                0x0003ffff
+#       define R300_ALU_DSTA_SHIFT              18
+#       define R300_ALU_DSTA_MASK               (31 << 18)
+#       define R300_ALU_DSTA_REG                (1 << 23)
+#       define R300_ALU_DSTA_OUTPUT             (1 << 24)
+#		define R300_ALU_DSTA_DEPTH              (1 << 27)
+
+#define R300_US_ALU_RGB_INST_0                   0x48C0
+#       define R300_ALU_ARGC_SRC0C_XYZ          0
+#       define R300_ALU_ARGC_SRC0C_XXX          1
+#       define R300_ALU_ARGC_SRC0C_YYY          2
+#       define R300_ALU_ARGC_SRC0C_ZZZ          3
+#       define R300_ALU_ARGC_SRC1C_XYZ          4
+#       define R300_ALU_ARGC_SRC1C_XXX          5
+#       define R300_ALU_ARGC_SRC1C_YYY          6
+#       define R300_ALU_ARGC_SRC1C_ZZZ          7
+#       define R300_ALU_ARGC_SRC2C_XYZ          8
+#       define R300_ALU_ARGC_SRC2C_XXX          9
+#       define R300_ALU_ARGC_SRC2C_YYY          10
+#       define R300_ALU_ARGC_SRC2C_ZZZ          11
+#       define R300_ALU_ARGC_SRC0A              12
+#       define R300_ALU_ARGC_SRC1A              13
+#       define R300_ALU_ARGC_SRC2A              14
+#       define R300_ALU_ARGC_SRCP_XYZ           15
+#       define R300_ALU_ARGC_SRCP_XXX           16
+#       define R300_ALU_ARGC_SRCP_YYY           17
+#       define R300_ALU_ARGC_SRCP_ZZZ           18
+#       define R300_ALU_ARGC_SRCP_WWW           19
+#       define R300_ALU_ARGC_ZERO               20
+#       define R300_ALU_ARGC_ONE                21
+#       define R300_ALU_ARGC_HALF               22
+#       define R300_ALU_ARGC_SRC0C_YZX          23
+#       define R300_ALU_ARGC_SRC1C_YZX          24
+#       define R300_ALU_ARGC_SRC2C_YZX          25
+#       define R300_ALU_ARGC_SRC0C_ZXY          26
+#       define R300_ALU_ARGC_SRC1C_ZXY          27
+#       define R300_ALU_ARGC_SRC2C_ZXY          28
+#       define R300_ALU_ARGC_SRC0CA_WZY         29
+#       define R300_ALU_ARGC_SRC1CA_WZY         30
+#       define R300_ALU_ARGC_SRC2CA_WZY         31
+
+#       define R300_ALU_ARG0C_SHIFT             0
+#       define R300_ALU_ARG0C_MASK              (31 << 0)
+#       define R300_ALU_ARG0C_NOP               (0 << 5)
+#       define R300_ALU_ARG0C_NEG               (1 << 5)
+#       define R300_ALU_ARG0C_ABS               (2 << 5)
+#       define R300_ALU_ARG0C_NAB               (3 << 5)
+#       define R300_ALU_ARG1C_SHIFT             7
+#       define R300_ALU_ARG1C_MASK              (31 << 7)
+#       define R300_ALU_ARG1C_NOP               (0 << 12)
+#       define R300_ALU_ARG1C_NEG               (1 << 12)
+#       define R300_ALU_ARG1C_ABS               (2 << 12)
+#       define R300_ALU_ARG1C_NAB               (3 << 12)
+#       define R300_ALU_ARG2C_SHIFT             14
+#       define R300_ALU_ARG2C_MASK              (31 << 14)
+#       define R300_ALU_ARG2C_NOP               (0 << 19)
+#       define R300_ALU_ARG2C_NEG               (1 << 19)
+#       define R300_ALU_ARG2C_ABS               (2 << 19)
+#       define R300_ALU_ARG2C_NAB               (3 << 19)
+#       define R300_ALU_SRCP_1_MINUS_2_SRC0     (0 << 21)
+#       define R300_ALU_SRCP_SRC1_MINUS_SRC0    (1 << 21)
+#       define R300_ALU_SRCP_SRC1_PLUS_SRC0     (2 << 21)
+#       define R300_ALU_SRCP_1_MINUS_SRC0       (3 << 21)
+
+#       define R300_ALU_OUTC_MAD                (0 << 23)
+#       define R300_ALU_OUTC_DP3                (1 << 23)
+#       define R300_ALU_OUTC_DP4                (2 << 23)
+#       define R300_ALU_OUTC_D2A                (3 << 23)
+#       define R300_ALU_OUTC_MIN                (4 << 23)
+#       define R300_ALU_OUTC_MAX                (5 << 23)
+#       define R300_ALU_OUTC_CMPH               (7 << 23)
+#       define R300_ALU_OUTC_CMP                (8 << 23)
+#       define R300_ALU_OUTC_FRC                (9 << 23)
+#       define R300_ALU_OUTC_REPL_ALPHA         (10 << 23)
+
+#       define R300_ALU_OUTC_MOD_NOP            (0 << 27)
+#       define R300_ALU_OUTC_MOD_MUL2           (1 << 27)
+#       define R300_ALU_OUTC_MOD_MUL4           (2 << 27)
+#       define R300_ALU_OUTC_MOD_MUL8           (3 << 27)
+#       define R300_ALU_OUTC_MOD_DIV2           (4 << 27)
+#       define R300_ALU_OUTC_MOD_DIV4           (5 << 27)
+#       define R300_ALU_OUTC_MOD_DIV8           (6 << 27)
+
+#       define R300_ALU_OUTC_CLAMP              (1 << 30)
+#       define R300_ALU_INSERT_NOP              (1 << 31)
+
+#define R300_US_ALU_ALPHA_INST_0                 0x49C0
+#       define R300_ALU_ARGA_SRC0C_X            0
+#       define R300_ALU_ARGA_SRC0C_Y            1
+#       define R300_ALU_ARGA_SRC0C_Z            2
+#       define R300_ALU_ARGA_SRC1C_X            3
+#       define R300_ALU_ARGA_SRC1C_Y            4
+#       define R300_ALU_ARGA_SRC1C_Z            5
+#       define R300_ALU_ARGA_SRC2C_X            6
+#       define R300_ALU_ARGA_SRC2C_Y            7
+#       define R300_ALU_ARGA_SRC2C_Z            8
+#       define R300_ALU_ARGA_SRC0A              9
+#       define R300_ALU_ARGA_SRC1A              10
+#       define R300_ALU_ARGA_SRC2A              11
+#       define R300_ALU_ARGA_SRCP_X             12
+#       define R300_ALU_ARGA_SRCP_Y             13
+#       define R300_ALU_ARGA_SRCP_Z             14
+#       define R300_ALU_ARGA_SRCP_W             15
+
+#       define R300_ALU_ARGA_ZERO               16
+#       define R300_ALU_ARGA_ONE                17
+#       define R300_ALU_ARGA_HALF               18
+#       define R300_ALU_ARG0A_SHIFT             0
+#       define R300_ALU_ARG0A_MASK              (31 << 0)
+#       define R300_ALU_ARG0A_NOP               (0 << 5)
+#       define R300_ALU_ARG0A_NEG               (1 << 5)
+#	define R300_ALU_ARG0A_ABS		 (2 << 5)
+#	define R300_ALU_ARG0A_NAB		 (3 << 5)
+#       define R300_ALU_ARG1A_SHIFT             7
+#       define R300_ALU_ARG1A_MASK              (31 << 7)
+#       define R300_ALU_ARG1A_NOP               (0 << 12)
+#       define R300_ALU_ARG1A_NEG               (1 << 12)
+#	define R300_ALU_ARG1A_ABS		 (2 << 12)
+#	define R300_ALU_ARG1A_NAB		 (3 << 12)
+#       define R300_ALU_ARG2A_SHIFT             14
+#       define R300_ALU_ARG2A_MASK              (31 << 14)
+#       define R300_ALU_ARG2A_NOP               (0 << 19)
+#       define R300_ALU_ARG2A_NEG               (1 << 19)
+#	define R300_ALU_ARG2A_ABS		 (2 << 19)
+#	define R300_ALU_ARG2A_NAB		 (3 << 19)
+#       define R300_ALU_SRCP_1_MINUS_2_SRC0     (0 << 21)
+#       define R300_ALU_SRCP_SRC1_MINUS_SRC0    (1 << 21)
+#       define R300_ALU_SRCP_SRC1_PLUS_SRC0     (2 << 21)
+#       define R300_ALU_SRCP_1_MINUS_SRC0       (3 << 21)
+
+#       define R300_ALU_OUTA_MAD                (0 << 23)
+#       define R300_ALU_OUTA_DP4                (1 << 23)
+#       define R300_ALU_OUTA_MIN                (2 << 23)
+#       define R300_ALU_OUTA_MAX                (3 << 23)
+#       define R300_ALU_OUTA_CND                (5 << 23)
+#       define R300_ALU_OUTA_CMP                (6 << 23)
+#       define R300_ALU_OUTA_FRC                (7 << 23)
+#       define R300_ALU_OUTA_EX2                (8 << 23)
+#       define R300_ALU_OUTA_LG2                (9 << 23)
+#       define R300_ALU_OUTA_RCP                (10 << 23)
+#       define R300_ALU_OUTA_RSQ                (11 << 23)
+
+#       define R300_ALU_OUTA_MOD_NOP            (0 << 27)
+#       define R300_ALU_OUTA_MOD_MUL2           (1 << 27)
+#       define R300_ALU_OUTA_MOD_MUL4           (2 << 27)
+#       define R300_ALU_OUTA_MOD_MUL8           (3 << 27)
+#       define R300_ALU_OUTA_MOD_DIV2           (4 << 27)
+#       define R300_ALU_OUTA_MOD_DIV4           (5 << 27)
+#       define R300_ALU_OUTA_MOD_DIV8           (6 << 27)
+
+#       define R300_ALU_OUTA_CLAMP              (1 << 30)
+/* END: Fragment program instruction set */
+
+/* Fog: Fog Blending Enable */
+#define R300_FG_FOG_BLEND                             0x4bc0
+#       define R300_FG_FOG_BLEND_DISABLE              (0 << 0)
+#       define R300_FG_FOG_BLEND_ENABLE               (1 << 0)
+#	define R300_FG_FOG_BLEND_FN_LINEAR            (0 << 1)
+#	define R300_FG_FOG_BLEND_FN_EXP               (1 << 1)
+#	define R300_FG_FOG_BLEND_FN_EXP2              (2 << 1)
+#	define R300_FG_FOG_BLEND_FN_CONSTANT          (3 << 1)
+#	define R300_FG_FOG_BLEND_FN_MASK              (3 << 1)
+
+/* Fog: Red Component of Fog Color */
+#define R300_FG_FOG_COLOR_R                           0x4bc8
+/* Fog: Green Component of Fog Color */
+#define R300_FG_FOG_COLOR_G                           0x4bcc
+/* Fog: Blue Component of Fog Color */
+#define R300_FG_FOG_COLOR_B                           0x4bd0
+#	define R300_FG_FOG_COLOR_MASK 0x000003ff
+
+/* Fog: Constant Factor for Fog Blending */
+#define R300_FG_FOG_FACTOR                            0x4bc4
+#	define FG_FOG_FACTOR_MASK 0x000003ff
+
+/* Fog: Alpha function */
+#define R300_FG_ALPHA_FUNC                            0x4bd4
+#       define R300_FG_ALPHA_FUNC_VAL_MASK               0x000000ff
+#       define R300_FG_ALPHA_FUNC_NEVER                     (0 << 8)
+#       define R300_FG_ALPHA_FUNC_LESS                      (1 << 8)
+#       define R300_FG_ALPHA_FUNC_EQUAL                     (2 << 8)
+#       define R300_FG_ALPHA_FUNC_LE                        (3 << 8)
+#       define R300_FG_ALPHA_FUNC_GREATER                   (4 << 8)
+#       define R300_FG_ALPHA_FUNC_NOTEQUAL                  (5 << 8)
+#       define R300_FG_ALPHA_FUNC_GE                        (6 << 8)
+#       define R300_FG_ALPHA_FUNC_ALWAYS                    (7 << 8)
+#       define R300_ALPHA_TEST_OP_MASK                      (7 << 8)
+#       define R300_FG_ALPHA_FUNC_DISABLE                   (0 << 11)
+#       define R300_FG_ALPHA_FUNC_ENABLE                    (1 << 11)
+
+#       define R500_FG_ALPHA_FUNC_10BIT                     (0 << 12)
+#       define R500_FG_ALPHA_FUNC_8BIT                      (1 << 12)
+
+#       define R300_FG_ALPHA_FUNC_MASK_DISABLE              (0 << 16)
+#       define R300_FG_ALPHA_FUNC_MASK_ENABLE               (1 << 16)
+#       define R300_FG_ALPHA_FUNC_CFG_2_OF_4                (0 << 17)
+#       define R300_FG_ALPHA_FUNC_CFG_3_OF_6                (1 << 17)
+
+#       define R300_FG_ALPHA_FUNC_DITH_DISABLE              (0 << 20)
+#       define R300_FG_ALPHA_FUNC_DITH_ENABLE               (1 << 20)
+
+#       define R500_FG_ALPHA_FUNC_OFFSET_DISABLE            (0 << 24)
+#       define R500_FG_ALPHA_FUNC_OFFSET_ENABLE             (1 << 24) /* Not supported in R520 */
+#       define R500_FG_ALPHA_FUNC_DISC_ZERO_MASK_DISABLE    (0 << 25)
+#       define R500_FG_ALPHA_FUNC_DISC_ZERO_MASK_ENABLE     (1 << 25)
+
+#       define R500_FG_ALPHA_FUNC_FP16_DISABLE              (0 << 28)
+#       define R500_FG_ALPHA_FUNC_FP16_ENABLE               (1 << 28)
+
+
+/* Fog: Where does the depth come from? */
+#define R300_FG_DEPTH_SRC                  0x4bd8
+#	define R300_FG_DEPTH_SRC_SCAN   (0 << 0)
+#	define R300_FG_DEPTH_SRC_SHADER (1 << 0)
+
+/* Fog: Alpha Compare Value */
+#define R500_FG_ALPHA_VALUE                0x4be0
+#	define R500_FG_ALPHA_VALUE_MASK 0x0000ffff
+
+/* gap */
+
+/* Fragment program parameters in 7.16 floating point */
+#define R300_PFS_PARAM_0_X                  0x4C00
+#define R300_PFS_PARAM_0_Y                  0x4C04
+#define R300_PFS_PARAM_0_Z                  0x4C08
+#define R300_PFS_PARAM_0_W                  0x4C0C
+/* last consts */
+#define R300_PFS_PARAM_31_X                 0x4DF0
+#define R300_PFS_PARAM_31_Y                 0x4DF4
+#define R300_PFS_PARAM_31_Z                 0x4DF8
+#define R300_PFS_PARAM_31_W                 0x4DFC
+
+/* Unpipelined. */
+#define R300_RB3D_CCTL                      0x4e00
+#	define R300_RB3D_CCTL_NUM_MULTIWRITES_1_BUFFER                (0 << 5)
+#	define R300_RB3D_CCTL_NUM_MULTIWRITES_2_BUFFERS               (1 << 5)
+#	define R300_RB3D_CCTL_NUM_MULTIWRITES_3_BUFFERS               (2 << 5)
+#	define R300_RB3D_CCTL_NUM_MULTIWRITES_4_BUFFERS               (3 << 5)
+#	define R300_RB3D_CCTL_CLRCMP_FLIPE_DISABLE                    (0 << 7)
+#	define R300_RB3D_CCTL_CLRCMP_FLIPE_ENABLE                     (1 << 7)
+#	define R300_RB3D_CCTL_AA_COMPRESSION_DISABLE                  (0 << 9)
+#	define R300_RB3D_CCTL_AA_COMPRESSION_ENABLE                   (1 << 9)
+#	define R300_RB3D_CCTL_CMASK_DISABLE                           (0 << 10)
+#	define R300_RB3D_CCTL_CMASK_ENABLE                            (1 << 10)
+/* reserved */
+#	define R300_RB3D_CCTL_INDEPENDENT_COLOR_CHANNEL_MASK_DISABLE  (0 << 12)
+#	define R300_RB3D_CCTL_INDEPENDENT_COLOR_CHANNEL_MASK_ENABLE   (1 << 12)
+#	define R300_RB3D_CCTL_WRITE_COMPRESSION_ENABLE                (0 << 13)
+#	define R300_RB3D_CCTL_WRITE_COMPRESSION_DISABLE               (1 << 13)
+#	define R300_RB3D_CCTL_INDEPENDENT_COLORFORMAT_ENABLE_DISABLE  (0 << 14)
+#	define R300_RB3D_CCTL_INDEPENDENT_COLORFORMAT_ENABLE_ENABLE   (1 << 14)
+
+
+/* Notes:
+ * - AFAIK fglrx always sets BLEND_UNKNOWN when blending is used in
+ *   the application
+ * - AFAIK fglrx always sets BLEND_NO_SEPARATE when CBLEND and ABLEND
+ *    are set to the same
+ *   function (both registers are always set up completely in any case)
+ * - Most blend flags are simply copied from R200 and not tested yet
+ */
+#define R300_RB3D_CBLEND                    0x4E04
+#define R300_RB3D_ABLEND                    0x4E08
+/* the following only appear in CBLEND */
+#       define R300_ALPHA_BLEND_ENABLE         (1 << 0)
+#       define R300_SEPARATE_ALPHA_ENABLE      (1 << 1)
+#       define R300_READ_ENABLE                (1 << 2)
+#       define R300_DISCARD_SRC_PIXELS_DIS     (0 << 3)
+#       define R300_DISCARD_SRC_PIXELS_SRC_ALPHA_0     (1 << 3)
+#       define R300_DISCARD_SRC_PIXELS_SRC_COLOR_0     (2 << 3)
+#       define R300_DISCARD_SRC_PIXELS_SRC_ALPHA_COLOR_0     (3 << 3)
+#       define R300_DISCARD_SRC_PIXELS_SRC_ALPHA_1     (4 << 3)
+#       define R300_DISCARD_SRC_PIXELS_SRC_COLOR_1     (5 << 3)
+#       define R300_DISCARD_SRC_PIXELS_SRC_ALPHA_COLOR_1     (6 << 3)
+
+/* the following are shared between CBLEND and ABLEND */
+#       define R300_FCN_MASK                         (3  << 12)
+#       define R300_COMB_FCN_ADD_CLAMP               (0  << 12)
+#       define R300_COMB_FCN_ADD_NOCLAMP             (1  << 12)
+#       define R300_COMB_FCN_SUB_CLAMP               (2  << 12)
+#       define R300_COMB_FCN_SUB_NOCLAMP             (3  << 12)
+#       define R300_COMB_FCN_MIN                     (4  << 12)
+#       define R300_COMB_FCN_MAX                     (5  << 12)
+#       define R300_COMB_FCN_RSUB_CLAMP              (6  << 12)
+#       define R300_COMB_FCN_RSUB_NOCLAMP            (7  << 12)
+#       define R300_BLEND_GL_ZERO                    (32)
+#       define R300_BLEND_GL_ONE                     (33)
+#       define R300_BLEND_GL_SRC_COLOR               (34)
+#       define R300_BLEND_GL_ONE_MINUS_SRC_COLOR     (35)
+#       define R300_BLEND_GL_DST_COLOR               (36)
+#       define R300_BLEND_GL_ONE_MINUS_DST_COLOR     (37)
+#       define R300_BLEND_GL_SRC_ALPHA               (38)
+#       define R300_BLEND_GL_ONE_MINUS_SRC_ALPHA     (39)
+#       define R300_BLEND_GL_DST_ALPHA               (40)
+#       define R300_BLEND_GL_ONE_MINUS_DST_ALPHA     (41)
+#       define R300_BLEND_GL_SRC_ALPHA_SATURATE      (42)
+#       define R300_BLEND_GL_CONST_COLOR             (43)
+#       define R300_BLEND_GL_ONE_MINUS_CONST_COLOR   (44)
+#       define R300_BLEND_GL_CONST_ALPHA             (45)
+#       define R300_BLEND_GL_ONE_MINUS_CONST_ALPHA   (46)
+#       define R300_BLEND_MASK                       (63)
+#       define R300_SRC_BLEND_SHIFT                  (16)
+#       define R300_DST_BLEND_SHIFT                  (24)
+
+/* Constant color used by the blender. Pipelined through the blender.
+ * Note: For R520, this field is ignored, use RB3D_CONSTANT_COLOR_GB__BLUE,
+ * RB3D_CONSTANT_COLOR_GB__GREEN, etc. instead.
+ */
+#define R300_RB3D_BLEND_COLOR               0x4E10
+
+
+/* 3D Color Channel Mask. If all the channels used in the current color format
+ * are disabled, then the cb will discard all the incoming quads. Pipelined
+ * through the blender.
+ */
+#define RB3D_COLOR_CHANNEL_MASK                  0x4E0C
+#	define RB3D_COLOR_CHANNEL_MASK_BLUE_MASK0  (1 << 0)
+#	define RB3D_COLOR_CHANNEL_MASK_GREEN_MASK0 (1 << 1)
+#	define RB3D_COLOR_CHANNEL_MASK_RED_MASK0   (1 << 2)
+#	define RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK0 (1 << 3)
+#	define RB3D_COLOR_CHANNEL_MASK_BLUE_MASK1  (1 << 4)
+#	define RB3D_COLOR_CHANNEL_MASK_GREEN_MASK1 (1 << 5)
+#	define RB3D_COLOR_CHANNEL_MASK_RED_MASK1   (1 << 6)
+#	define RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK1 (1 << 7)
+#	define RB3D_COLOR_CHANNEL_MASK_BLUE_MASK2  (1 << 8)
+#	define RB3D_COLOR_CHANNEL_MASK_GREEN_MASK2 (1 << 9)
+#	define RB3D_COLOR_CHANNEL_MASK_RED_MASK2   (1 << 10)
+#	define RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK2 (1 << 11)
+#	define RB3D_COLOR_CHANNEL_MASK_BLUE_MASK3  (1 << 12)
+#	define RB3D_COLOR_CHANNEL_MASK_GREEN_MASK3 (1 << 13)
+#	define RB3D_COLOR_CHANNEL_MASK_RED_MASK3   (1 << 14)
+#	define RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK3 (1 << 15)
+
+/* Clear color that is used when the color mask is set to 00. Unpipelined.
+ * Program this register with a 32-bit value in ARGB8888 or ARGB2101010
+ * formats, ignoring the fields.
+ */
+#define RB3D_COLOR_CLEAR_VALUE                   0x4e14
+
+/* gap */
+
+/* Color Compare Color. Stalls the 2d/3d datapath until it is idle. */
+#define RB3D_CLRCMP_CLR                     0x4e20
+
+/* Color Compare Mask. Stalls the 2d/3d datapath until it is idle. */
+#define RB3D_CLRCMP_MSK                     0x4e24
+
+/* Color Buffer Address Offset of multibuffer 0. Unpipelined. */
+#define R300_RB3D_COLOROFFSET0              0x4E28
+#       define R300_COLOROFFSET_MASK             0xFFFFFFE0
+/* Color Buffer Address Offset of multibuffer 1. Unpipelined. */
+#define R300_RB3D_COLOROFFSET1              0x4E2C
+/* Color Buffer Address Offset of multibuffer 2. Unpipelined. */
+#define R300_RB3D_COLOROFFSET2              0x4E30
+/* Color Buffer Address Offset of multibuffer 3. Unpipelined. */
+#define R300_RB3D_COLOROFFSET3              0x4E34
+
+/* Color buffer format and tiling control for all the multibuffers and the
+ * pitch of multibuffer 0 to 3. Unpipelined. The cache must be empty before any
+ * of the registers are changed.
+ *
+ * Bit 16: Larger tiles
+ * Bit 17: 4x2 tiles
+ * Bit 18: Extremely weird tile like, but some pixels duplicated?
+ */
+#define R300_RB3D_COLORPITCH0               0x4E38
+#       define R300_COLORPITCH_MASK              0x00003FFE
+#       define R300_COLOR_TILE_DISABLE            (0 << 16)
+#       define R300_COLOR_TILE_ENABLE             (1 << 16)
+#       define R300_COLOR_MICROTILE_DISABLE       (0 << 17)
+#       define R300_COLOR_MICROTILE_ENABLE        (1 << 17)
+#       define R300_COLOR_MICROTILE_ENABLE_SQUARE (2 << 17) /* Only available in 16-bit */
+#       define R300_COLOR_ENDIAN_NO_SWAP          (0 << 19)
+#       define R300_COLOR_ENDIAN_WORD_SWAP        (1 << 19)
+#       define R300_COLOR_ENDIAN_DWORD_SWAP       (2 << 19)
+#       define R300_COLOR_ENDIAN_HALF_DWORD_SWAP  (3 << 19)
+#	define R500_COLOR_FORMAT_ARGB10101010     (0 << 21)
+#	define R500_COLOR_FORMAT_UV1010           (1 << 21)
+#	define R500_COLOR_FORMAT_CI8              (2 << 21) /* 2D only */
+#	define R300_COLOR_FORMAT_ARGB1555         (3 << 21)
+#       define R300_COLOR_FORMAT_RGB565           (4 << 21)
+#       define R500_COLOR_FORMAT_ARGB2101010      (5 << 21)
+#       define R300_COLOR_FORMAT_ARGB8888         (6 << 21)
+#       define R300_COLOR_FORMAT_ARGB32323232     (7 << 21)
+/* reserved */
+#       define R300_COLOR_FORMAT_I8               (9 << 21)
+#       define R300_COLOR_FORMAT_ARGB16161616     (10 << 21)
+#       define R300_COLOR_FORMAT_VYUY             (11 << 21)
+#       define R300_COLOR_FORMAT_YVYU             (12 << 21)
+#       define R300_COLOR_FORMAT_UV88             (13 << 21)
+#       define R500_COLOR_FORMAT_I10              (14 << 21)
+#       define R300_COLOR_FORMAT_ARGB4444         (15 << 21)
+#define R300_RB3D_COLORPITCH1               0x4E3C
+#define R300_RB3D_COLORPITCH2               0x4E40
+#define R300_RB3D_COLORPITCH3               0x4E44
+
+/* gap */
+
+/* Destination Color Buffer Cache Control/Status. If the cb is in e2 mode, then
+ * a flush or free will not occur upon a write to this register, but a sync
+ * will be immediately sent if one is requested. If both DC_FLUSH and DC_FREE
+ * are zero but DC_FINISH is one, then a sync will be sent immediately -- the
+ * cb will not wait for all the previous operations to complete before sending
+ * the sync. Unpipelined except when DC_FINISH and DC_FREE are both set to
+ * zero.
+ *
+ * Set to 0A before 3D operations, set to 02 afterwards.
+ */
+#define R300_RB3D_DSTCACHE_CTLSTAT               0x4e4c
+#	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_NO_EFFECT         (0 << 0)
+#	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_NO_EFFECT_1       (1 << 0)
+#	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D    (2 << 0)
+#	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D_1  (3 << 0)
+#	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_NO_EFFECT          (0 << 2)
+#	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_NO_EFFECT_1        (1 << 2)
+#	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS       (2 << 2)
+#	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS_1     (3 << 2)
+#	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FINISH_NO_SIGNAL        (0 << 4)
+#	define R300_RB3D_DSTCACHE_CTLSTAT_DC_FINISH_SIGNAL           (1 << 4)
+
+#define R300_RB3D_DITHER_CTL 0x4E50
+#	define R300_RB3D_DITHER_CTL_DITHER_MODE_TRUNCATE         (0 << 0)
+#	define R300_RB3D_DITHER_CTL_DITHER_MODE_ROUND            (1 << 0)
+#	define R300_RB3D_DITHER_CTL_DITHER_MODE_LUT              (2 << 0)
+/* reserved */
+#	define R300_RB3D_DITHER_CTL_ALPHA_DITHER_MODE_TRUNCATE   (0 << 2)
+#	define R300_RB3D_DITHER_CTL_ALPHA_DITHER_MODE_ROUND      (1 << 2)
+#	define R300_RB3D_DITHER_CTL_ALPHA_DITHER_MODE_LUT        (2 << 2)
+/* reserved */
+
+/* Resolve buffer destination address. The cache must be empty before changing
+ * this register if the cb is in resolve mode. Unpipelined
+ */
+#define R300_RB3D_AARESOLVE_OFFSET        0x4e80
+#	define R300_RB3D_AARESOLVE_OFFSET_SHIFT 5
+#	define R300_RB3D_AARESOLVE_OFFSET_MASK 0xffffffe0 /* At least according to the calculations of Christoph Brill */
+
+/* Resolve Buffer Pitch and Tiling Control. The cache must be empty before
+ * changing this register if the cb is in resolve mode. Unpipelined
+ */
+#define R300_RB3D_AARESOLVE_PITCH         0x4e84
+#	define R300_RB3D_AARESOLVE_PITCH_SHIFT 1
+#	define R300_RB3D_AARESOLVE_PITCH_MASK  0x00003ffe /* At least according to the calculations of Christoph Brill */
+
+/* Resolve Buffer Control. Unpipelined */
+#define R300_RB3D_AARESOLVE_CTL           0x4e88
+#	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_MODE_NORMAL   (0 << 0)
+#	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_MODE_RESOLVE  (1 << 0)
+#	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_GAMMA_10      (0 << 1)
+#	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_GAMMA_22      (1 << 1)
+#	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_ALPHA_SAMPLE0 (0 << 2)
+#	define R300_RB3D_AARESOLVE_CTL_AARESOLVE_ALPHA_AVERAGE (1 << 2)
+
+
+/* Discard src pixels less than or equal to threshold. */
+#define R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD 0x4ea0
+/* Discard src pixels greater than or equal to threshold. */
+#define R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD 0x4ea4
+#	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_BLUE_SHIFT 0
+#	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_BLUE_MASK 0x000000ff
+#	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_GREEN_SHIFT 8
+#	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_GREEN_MASK 0x0000ff00
+#	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_RED_SHIFT 16
+#	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_RED_MASK 0x00ff0000
+#	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_ALPHA_SHIFT 24
+#	define R500_RB3D_DISCARD_SRC_PIXEL_THRESHOLD_ALPHA_MASK 0xff000000
+
+/* 3D ROP Control. Stalls the 2d/3d datapath until it is idle. */
+#define R300_RB3D_ROPCNTL                             0x4e18
+#	define R300_RB3D_ROPCNTL_ROP_ENABLE            0x00000004
+#	define R300_RB3D_ROPCNTL_ROP_MASK              (15 << 8)
+#	define R300_RB3D_ROPCNTL_ROP_SHIFT             8
+
+/* Color Compare Flip. Stalls the 2d/3d datapath until it is idle. */
+#define R300_RB3D_CLRCMP_FLIPE                        0x4e1c
+
+/* Sets the fifo sizes */
+#define R500_RB3D_FIFO_SIZE                           0x4ef4
+#	define R500_RB3D_FIFO_SIZE_OP_FIFO_SIZE_FULL   (0 << 0)
+#	define R500_RB3D_FIFO_SIZE_OP_FIFO_SIZE_HALF   (1 << 0)
+#	define R500_RB3D_FIFO_SIZE_OP_FIFO_SIZE_QUATER (2 << 0)
+#	define R500_RB3D_FIFO_SIZE_OP_FIFO_SIZE_EIGTHS (3 << 0)
+
+/* Constant color used by the blender. Pipelined through the blender. */
+#define R500_RB3D_CONSTANT_COLOR_AR                   0x4ef8
+#	define R500_RB3D_CONSTANT_COLOR_AR_RED_MASK    0x0000ffff
+#	define R500_RB3D_CONSTANT_COLOR_AR_RED_SHIFT   0
+#	define R500_RB3D_CONSTANT_COLOR_AR_ALPHA_MASK  0xffff0000
+#	define R500_RB3D_CONSTANT_COLOR_AR_ALPHA_SHIFT 16
+
+/* Constant color used by the blender. Pipelined through the blender. */
+#define R500_RB3D_CONSTANT_COLOR_GB                   0x4efc
+#	define R500_RB3D_CONSTANT_COLOR_AR_BLUE_MASK   0x0000ffff
+#	define R500_RB3D_CONSTANT_COLOR_AR_BLUE_SHIFT  0
+#	define R500_RB3D_CONSTANT_COLOR_AR_GREEN_MASK  0xffff0000
+#	define R500_RB3D_CONSTANT_COLOR_AR_GREEN_SHIFT 16
+
+/* gap */
+/* There seems to be no "write only" setting, so use Z-test = ALWAYS
+ * for this.
+ * Bit (1<<8) is the "test" bit. so plain write is 6  - vd
+ */
+#define R300_ZB_CNTL                             0x4F00
+#	define R300_STENCIL_ENABLE		 (1 << 0)
+#	define R300_Z_ENABLE		         (1 << 1)
+#	define R300_Z_WRITE_ENABLE		 (1 << 2)
+#	define R300_Z_SIGNED_COMPARE		 (1 << 3)
+#	define R300_STENCIL_FRONT_BACK		 (1 << 4)
+
+#define R300_ZB_ZSTENCILCNTL                   0x4f04
+	/* functions */
+#	define R300_ZS_NEVER			0
+#	define R300_ZS_LESS			1
+#	define R300_ZS_LEQUAL			2
+#	define R300_ZS_EQUAL			3
+#	define R300_ZS_GEQUAL			4
+#	define R300_ZS_GREATER			5
+#	define R300_ZS_NOTEQUAL			6
+#	define R300_ZS_ALWAYS			7
+#       define R300_ZS_MASK                     7
+	/* operations */
+#	define R300_ZS_KEEP			0
+#	define R300_ZS_ZERO			1
+#	define R300_ZS_REPLACE			2
+#	define R300_ZS_INCR			3
+#	define R300_ZS_DECR			4
+#	define R300_ZS_INVERT			5
+#	define R300_ZS_INCR_WRAP		6
+#	define R300_ZS_DECR_WRAP		7
+#	define R300_Z_FUNC_SHIFT		0
+	/* front and back refer to operations done for front
+	   and back faces, i.e. separate stencil function support */
+#	define R300_S_FRONT_FUNC_SHIFT	        3
+#	define R300_S_FRONT_SFAIL_OP_SHIFT	6
+#	define R300_S_FRONT_ZPASS_OP_SHIFT	9
+#	define R300_S_FRONT_ZFAIL_OP_SHIFT      12
+#	define R300_S_BACK_FUNC_SHIFT           15
+#	define R300_S_BACK_SFAIL_OP_SHIFT       18
+#	define R300_S_BACK_ZPASS_OP_SHIFT       21
+#	define R300_S_BACK_ZFAIL_OP_SHIFT       24
+
+#define R300_ZB_STENCILREFMASK                        0x4f08
+#	define R300_STENCILREF_SHIFT       0
+#	define R300_STENCILREF_MASK        0x000000ff
+#	define R300_STENCILMASK_SHIFT      8
+#	define R300_STENCILMASK_MASK       0x0000ff00
+#	define R300_STENCILWRITEMASK_SHIFT 16
+#	define R300_STENCILWRITEMASK_MASK  0x00ff0000
+
+/* gap */
+
+#define R300_ZB_FORMAT                             0x4f10
+#	define R300_DEPTHFORMAT_16BIT_INT_Z   (0 << 0)
+#	define R300_DEPTHFORMAT_16BIT_13E3    (1 << 0)
+#	define R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL   (2 << 0)
+/* reserved up to (15 << 0) */
+#	define R300_INVERT_13E3_LEADING_ONES  (0 << 4)
+#	define R300_INVERT_13E3_LEADING_ZEROS (1 << 4)
+
+#define R300_ZB_ZTOP                             0x4F14
+#	define R300_ZTOP_DISABLE                 (0 << 0)
+#	define R300_ZTOP_ENABLE                  (1 << 0)
+
+/* gap */
+
+#define R300_ZB_ZCACHE_CTLSTAT            0x4f18
+#       define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_NO_EFFECT      (0 << 0)
+#       define R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE (1 << 0)
+#       define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_NO_EFFECT       (0 << 1)
+#       define R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE            (1 << 1)
+#       define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_IDLE            (0 << 31)
+#       define R300_ZB_ZCACHE_CTLSTAT_ZC_BUSY_BUSY            (1 << 31)
+
+#define R300_ZB_BW_CNTL                     0x4f1c
+#	define R300_HIZ_DISABLE                              (0 << 0)
+#	define R300_HIZ_ENABLE                               (1 << 0)
+#	define R300_HIZ_MIN                                  (0 << 1)
+#	define R300_HIZ_MAX                                  (1 << 1)
+#	define R300_FAST_FILL_DISABLE                        (0 << 2)
+#	define R300_FAST_FILL_ENABLE                         (1 << 2)
+#	define R300_RD_COMP_DISABLE                          (0 << 3)
+#	define R300_RD_COMP_ENABLE                           (1 << 3)
+#	define R300_WR_COMP_DISABLE                          (0 << 4)
+#	define R300_WR_COMP_ENABLE                           (1 << 4)
+#	define R300_ZB_CB_CLEAR_RMW                          (0 << 5)
+#	define R300_ZB_CB_CLEAR_CACHE_LINEAR                 (1 << 5)
+#	define R300_FORCE_COMPRESSED_STENCIL_VALUE_DISABLE   (0 << 6)
+#	define R300_FORCE_COMPRESSED_STENCIL_VALUE_ENABLE    (1 << 6)
+
+#	define R500_ZEQUAL_OPTIMIZE_ENABLE                   (0 << 7)
+#	define R500_ZEQUAL_OPTIMIZE_DISABLE                  (1 << 7)
+#	define R500_SEQUAL_OPTIMIZE_ENABLE                   (0 << 8)
+#	define R500_SEQUAL_OPTIMIZE_DISABLE                  (1 << 8)
+
+#	define R500_BMASK_ENABLE                             (0 << 10)
+#	define R500_BMASK_DISABLE                            (1 << 10)
+#	define R500_HIZ_EQUAL_REJECT_DISABLE                 (0 << 11)
+#	define R500_HIZ_EQUAL_REJECT_ENABLE                  (1 << 11)
+#	define R500_HIZ_FP_EXP_BITS_DISABLE                  (0 << 12)
+#	define R500_HIZ_FP_EXP_BITS_1                        (1 << 12)
+#	define R500_HIZ_FP_EXP_BITS_2                        (2 << 12)
+#	define R500_HIZ_FP_EXP_BITS_3                        (3 << 12)
+#	define R500_HIZ_FP_EXP_BITS_4                        (4 << 12)
+#	define R500_HIZ_FP_EXP_BITS_5                        (5 << 12)
+#	define R500_HIZ_FP_INVERT_LEADING_ONES               (0 << 15)
+#	define R500_HIZ_FP_INVERT_LEADING_ZEROS              (1 << 15)
+#	define R500_TILE_OVERWRITE_RECOMPRESSION_ENABLE      (0 << 16)
+#	define R500_TILE_OVERWRITE_RECOMPRESSION_DISABLE     (1 << 16)
+#	define R500_CONTIGUOUS_6XAA_SAMPLES_ENABLE           (0 << 17)
+#	define R500_CONTIGUOUS_6XAA_SAMPLES_DISABLE          (1 << 17)
+#	define R500_PEQ_PACKING_DISABLE                      (0 << 18)
+#	define R500_PEQ_PACKING_ENABLE                       (1 << 18)
+#	define R500_COVERED_PTR_MASKING_DISABLE              (0 << 18)
+#	define R500_COVERED_PTR_MASKING_ENABLE               (1 << 18)
+
+
+/* gap */
+
+/* Z Buffer Address Offset.
+ * Bits 31 to 5 are used for aligned Z buffer address offset for macro tiles.
+ */
+#define R300_ZB_DEPTHOFFSET               0x4f20
+
+/* Z Buffer Pitch and Endian Control */
+#define R300_ZB_DEPTHPITCH                0x4f24
+#       define R300_DEPTHPITCH_MASK              0x00003FFC
+#       define R300_DEPTHMACROTILE_DISABLE      (0 << 16)
+#       define R300_DEPTHMACROTILE_ENABLE       (1 << 16)
+#       define R300_DEPTHMICROTILE_LINEAR       (0 << 17)
+#       define R300_DEPTHMICROTILE_TILED        (1 << 17)
+#       define R300_DEPTHMICROTILE_TILED_SQUARE (2 << 17)
+#       define R300_DEPTHENDIAN_NO_SWAP         (0 << 18)
+#       define R300_DEPTHENDIAN_WORD_SWAP       (1 << 18)
+#       define R300_DEPTHENDIAN_DWORD_SWAP      (2 << 18)
+#       define R300_DEPTHENDIAN_HALF_DWORD_SWAP (3 << 18)
+
+/* Z Buffer Clear Value */
+#define R300_ZB_DEPTHCLEARVALUE                  0x4f28
+
+/* Hierarchical Z Memory Offset */
+#define R300_ZB_HIZ_OFFSET                       0x4f44
+
+/* Hierarchical Z Write Index */
+#define R300_ZB_HIZ_WRINDEX                      0x4f48
+
+/* Hierarchical Z Data */
+#define R300_ZB_HIZ_DWORD                        0x4f4c
+
+/* Hierarchical Z Read Index */
+#define R300_ZB_HIZ_RDINDEX                      0x4f50
+
+/* Hierarchical Z Pitch */
+#define R300_ZB_HIZ_PITCH                        0x4f54
+
+/* Z Buffer Z Pass Counter Data */
+#define R300_ZB_ZPASS_DATA                       0x4f58
+
+/* Z Buffer Z Pass Counter Address */
+#define R300_ZB_ZPASS_ADDR                       0x4f5c
+
+/* Depth buffer X and Y coordinate offset */
+#define R300_ZB_DEPTHXY_OFFSET                   0x4f60
+#	define R300_DEPTHX_OFFSET_SHIFT  1
+#	define R300_DEPTHX_OFFSET_MASK   0x000007FE
+#	define R300_DEPTHY_OFFSET_SHIFT  17
+#	define R300_DEPTHY_OFFSET_MASK   0x07FE0000
+
+/* Sets the fifo sizes */
+#define R500_ZB_FIFO_SIZE                        0x4fd0
+#	define R500_OP_FIFO_SIZE_FULL   (0 << 0)
+#	define R500_OP_FIFO_SIZE_HALF   (1 << 0)
+#	define R500_OP_FIFO_SIZE_QUATER (2 << 0)
+#	define R500_OP_FIFO_SIZE_EIGTHS (4 << 0)
+
+/* Stencil Reference Value and Mask for backfacing quads */
+/* R300_ZB_STENCILREFMASK handles front face */
+#define R500_ZB_STENCILREFMASK_BF                0x4fd4
+#	define R500_STENCILREF_SHIFT       0
+#	define R500_STENCILREF_MASK        0x000000ff
+#	define R500_STENCILMASK_SHIFT      8
+#	define R500_STENCILMASK_MASK       0x0000ff00
+#	define R500_STENCILWRITEMASK_SHIFT 16
+#	define R500_STENCILWRITEMASK_MASK  0x00ff0000
+
+/**
+ * \defgroup R3XX_R5XX_PROGRAMMABLE_VERTEX_SHADER_DESCRIPTION R3XX-R5XX PROGRAMMABLE VERTEX SHADER DESCRIPTION
+ *
+ * The PVS_DST_MATH_INST is used to identify whether the instruction is a Vector
+ * Engine instruction or a Math Engine instruction.
+ */
+
+/*\{*/
+
+enum {
+	/* R3XX */
+	VECTOR_NO_OP			= 0,
+	VE_DOT_PRODUCT			= 1,
+	VE_MULTIPLY			= 2,
+	VE_ADD				= 3,
+	VE_MULTIPLY_ADD			= 4,
+	VE_DISTANCE_VECTOR		= 5,
+	VE_FRACTION			= 6,
+	VE_MAXIMUM			= 7,
+	VE_MINIMUM			= 8,
+	VE_SET_GREATER_THAN_EQUAL	= 9,
+	VE_SET_LESS_THAN		= 10,
+	VE_MULTIPLYX2_ADD		= 11,
+	VE_MULTIPLY_CLAMP		= 12,
+	VE_FLT2FIX_DX			= 13,
+	VE_FLT2FIX_DX_RND		= 14,
+	/* R5XX */
+	VE_PRED_SET_EQ_PUSH		= 15,
+	VE_PRED_SET_GT_PUSH		= 16,
+	VE_PRED_SET_GTE_PUSH		= 17,
+	VE_PRED_SET_NEQ_PUSH		= 18,
+	VE_COND_WRITE_EQ		= 19,
+	VE_COND_WRITE_GT		= 20,
+	VE_COND_WRITE_GTE		= 21,
+	VE_COND_WRITE_NEQ		= 22,
+	VE_COND_MUX_EQ			= 23,
+	VE_COND_MUX_GT			= 24,
+	VE_COND_MUX_GTE			= 25,
+	VE_SET_GREATER_THAN		= 26,
+	VE_SET_EQUAL			= 27,
+	VE_SET_NOT_EQUAL		= 28,
+};
+
+enum {
+	/* R3XX */
+	MATH_NO_OP			= 0,
+	ME_EXP_BASE2_DX			= 1,
+	ME_LOG_BASE2_DX			= 2,
+	ME_EXP_BASEE_FF			= 3,
+	ME_LIGHT_COEFF_DX		= 4,
+	ME_POWER_FUNC_FF		= 5,
+	ME_RECIP_DX			= 6,
+	ME_RECIP_FF			= 7,
+	ME_RECIP_SQRT_DX		= 8,
+	ME_RECIP_SQRT_FF		= 9,
+	ME_MULTIPLY			= 10,
+	ME_EXP_BASE2_FULL_DX		= 11,
+	ME_LOG_BASE2_FULL_DX		= 12,
+	ME_POWER_FUNC_FF_CLAMP_B	= 13,
+	ME_POWER_FUNC_FF_CLAMP_B1	= 14,
+	ME_POWER_FUNC_FF_CLAMP_01	= 15,
+	ME_SIN				= 16,
+	ME_COS				= 17,
+	/* R5XX */
+	ME_LOG_BASE2_IEEE		= 18,
+	ME_RECIP_IEEE			= 19,
+	ME_RECIP_SQRT_IEEE		= 20,
+	ME_PRED_SET_EQ			= 21,
+	ME_PRED_SET_GT			= 22,
+	ME_PRED_SET_GTE			= 23,
+	ME_PRED_SET_NEQ			= 24,
+	ME_PRED_SET_CLR			= 25,
+	ME_PRED_SET_INV			= 26,
+	ME_PRED_SET_POP			= 27,
+	ME_PRED_SET_RESTORE		= 28,
+};
+
+enum {
+	/* R3XX */
+	PVS_MACRO_OP_2CLK_MADD		= 0,
+	PVS_MACRO_OP_2CLK_M2X_ADD	= 1,
+};
+
+enum {
+	PVS_SRC_REG_TEMPORARY		= 0,	/* Intermediate Storage */
+	PVS_SRC_REG_INPUT		= 1,	/* Input Vertex Storage */
+	PVS_SRC_REG_CONSTANT		= 2,	/* Constant State Storage */
+	PVS_SRC_REG_ALT_TEMPORARY	= 3,	/* Alternate Intermediate Storage */
+};
+
+enum {
+	PVS_DST_REG_TEMPORARY		= 0,	/* Intermediate Storage */
+	PVS_DST_REG_A0			= 1,	/* Address Register Storage */
+	PVS_DST_REG_OUT			= 2,	/* Output Memory. Used for all outputs */
+	PVS_DST_REG_OUT_REPL_X		= 3,	/* Output Memory & Replicate X to all channels */
+	PVS_DST_REG_ALT_TEMPORARY	= 4,	/* Alternate Intermediate Storage */
+	PVS_DST_REG_INPUT		= 5,	/* Output Memory & Replicate X to all channels */
+};
+
+enum {
+	PVS_SRC_SELECT_X		= 0,	/* Select X Component */
+	PVS_SRC_SELECT_Y		= 1,	/* Select Y Component */
+	PVS_SRC_SELECT_Z		= 2,	/* Select Z Component */
+	PVS_SRC_SELECT_W		= 3,	/* Select W Component */
+	PVS_SRC_SELECT_FORCE_0		= 4,	/* Force Component to 0.0 */
+	PVS_SRC_SELECT_FORCE_1		= 5,	/* Force Component to 1.0 */
+};
+
+/* PVS Opcode & Destination Operand Description */
+
+enum {
+	PVS_DST_OPCODE_MASK		= 0x3f,
+	PVS_DST_OPCODE_SHIFT		= 0,
+	PVS_DST_MATH_INST_MASK		= 0x1,
+	PVS_DST_MATH_INST_SHIFT		= 6,
+	PVS_DST_MACRO_INST_MASK		= 0x1,
+	PVS_DST_MACRO_INST_SHIFT	= 7,
+	PVS_DST_REG_TYPE_MASK		= 0xf,
+	PVS_DST_REG_TYPE_SHIFT		= 8,
+	PVS_DST_ADDR_MODE_1_MASK	= 0x1,
+	PVS_DST_ADDR_MODE_1_SHIFT	= 12,
+	PVS_DST_OFFSET_MASK		= 0x7f,
+	PVS_DST_OFFSET_SHIFT		= 13,
+	PVS_DST_WE_X_MASK		= 0x1,
+	PVS_DST_WE_X_SHIFT		= 20,
+	PVS_DST_WE_Y_MASK		= 0x1,
+	PVS_DST_WE_Y_SHIFT		= 21,
+	PVS_DST_WE_Z_MASK		= 0x1,
+	PVS_DST_WE_Z_SHIFT		= 22,
+	PVS_DST_WE_W_MASK		= 0x1,
+	PVS_DST_WE_W_SHIFT		= 23,
+	PVS_DST_VE_SAT_MASK		= 0x1,
+	PVS_DST_VE_SAT_SHIFT		= 24,
+	PVS_DST_ME_SAT_MASK		= 0x1,
+	PVS_DST_ME_SAT_SHIFT		= 25,
+	PVS_DST_PRED_ENABLE_MASK	= 0x1,
+	PVS_DST_PRED_ENABLE_SHIFT	= 26,
+	PVS_DST_PRED_SENSE_MASK		= 0x1,
+	PVS_DST_PRED_SENSE_SHIFT	= 27,
+	PVS_DST_DUAL_MATH_OP_MASK	= 0x3,
+	PVS_DST_DUAL_MATH_OP_SHIFT	= 27,
+	PVS_DST_ADDR_SEL_MASK		= 0x3,
+	PVS_DST_ADDR_SEL_SHIFT		= 29,
+	PVS_DST_ADDR_MODE_0_MASK	= 0x1,
+	PVS_DST_ADDR_MODE_0_SHIFT	= 31,
+};
+
+/* PVS Source Operand Description */
+
+enum {
+	PVS_SRC_REG_TYPE_MASK		= 0x3,
+	PVS_SRC_REG_TYPE_SHIFT		= 0,
+	SPARE_0_MASK			= 0x1,
+	SPARE_0_SHIFT			= 2,
+	PVS_SRC_ABS_XYZW_MASK		= 0x1,
+	PVS_SRC_ABS_XYZW_SHIFT		= 3,
+	PVS_SRC_ADDR_MODE_0_MASK	= 0x1,
+	PVS_SRC_ADDR_MODE_0_SHIFT	= 4,
+	PVS_SRC_OFFSET_MASK		= 0xff,
+	PVS_SRC_OFFSET_SHIFT		= 5,
+	PVS_SRC_SWIZZLE_X_MASK		= 0x7,
+	PVS_SRC_SWIZZLE_X_SHIFT		= 13,
+	PVS_SRC_SWIZZLE_Y_MASK		= 0x7,
+	PVS_SRC_SWIZZLE_Y_SHIFT		= 16,
+	PVS_SRC_SWIZZLE_Z_MASK		= 0x7,
+	PVS_SRC_SWIZZLE_Z_SHIFT		= 19,
+	PVS_SRC_SWIZZLE_W_MASK		= 0x7,
+	PVS_SRC_SWIZZLE_W_SHIFT		= 22,
+	PVS_SRC_MODIFIER_X_MASK		= 0x1,
+	PVS_SRC_MODIFIER_X_SHIFT	= 25,
+	PVS_SRC_MODIFIER_Y_MASK		= 0x1,
+	PVS_SRC_MODIFIER_Y_SHIFT	= 26,
+	PVS_SRC_MODIFIER_Z_MASK		= 0x1,
+	PVS_SRC_MODIFIER_Z_SHIFT	= 27,
+	PVS_SRC_MODIFIER_W_MASK		= 0x1,
+	PVS_SRC_MODIFIER_W_SHIFT	= 28,
+	PVS_SRC_ADDR_SEL_MASK		= 0x3,
+	PVS_SRC_ADDR_SEL_SHIFT		= 29,
+	PVS_SRC_ADDR_MODE_1_MASK	= 0x0,
+	PVS_SRC_ADDR_MODE_1_SHIFT	= 32,
+};
+
+/*\}*/
+
+/* BEGIN: Packet 3 commands */
+
+/* A primitive emission dword. */
+#define R300_PRIM_TYPE_NONE                     (0 << 0)
+#define R300_PRIM_TYPE_POINT                    (1 << 0)
+#define R300_PRIM_TYPE_LINE                     (2 << 0)
+#define R300_PRIM_TYPE_LINE_STRIP               (3 << 0)
+#define R300_PRIM_TYPE_TRI_LIST                 (4 << 0)
+#define R300_PRIM_TYPE_TRI_FAN                  (5 << 0)
+#define R300_PRIM_TYPE_TRI_STRIP                (6 << 0)
+#define R300_PRIM_TYPE_TRI_TYPE2                (7 << 0)
+#define R300_PRIM_TYPE_RECT_LIST                (8 << 0)
+#define R300_PRIM_TYPE_3VRT_POINT_LIST          (9 << 0)
+#define R300_PRIM_TYPE_3VRT_LINE_LIST           (10 << 0)
+	/* GUESS (based on r200) */
+#define R300_PRIM_TYPE_POINT_SPRITES            (11 << 0)
+#define R300_PRIM_TYPE_LINE_LOOP                (12 << 0)
+#define R300_PRIM_TYPE_QUADS                    (13 << 0)
+#define R300_PRIM_TYPE_QUAD_STRIP               (14 << 0)
+#define R300_PRIM_TYPE_POLYGON                  (15 << 0)
+#define R300_PRIM_TYPE_MASK                     0xF
+#define R300_PRIM_WALK_IND                      (1 << 4)
+#define R300_PRIM_WALK_LIST                     (2 << 4)
+#define R300_PRIM_WALK_RING                     (3 << 4)
+#define R300_PRIM_WALK_MASK                     (3 << 4)
+	/* GUESS (based on r200) */
+#define R300_PRIM_COLOR_ORDER_BGRA              (0 << 6)
+#define R300_PRIM_COLOR_ORDER_RGBA              (1 << 6)
+#define R300_PRIM_NUM_VERTICES_SHIFT            16
+#define R300_PRIM_NUM_VERTICES_MASK             0xffff
+
+
+
+/*
+ * The R500 unified shader (US) registers come in banks of 512 each, one
+ * for each instruction slot in the shader.  You can't touch them directly.
+ * R500_US_VECTOR_INDEX() sets the base instruction to modify; successive
+ * writes to R500_GA_US_VECTOR_DATA autoincrement the index after the
+ * instruction is fully specified.
+ */
+#define R500_US_ALU_ALPHA_INST_0			0xa800
+#   define R500_ALPHA_OP_MAD				0
+#   define R500_ALPHA_OP_DP				1
+#   define R500_ALPHA_OP_MIN				2
+#   define R500_ALPHA_OP_MAX				3
+/* #define R500_ALPHA_OP_RESERVED			4 */
+#   define R500_ALPHA_OP_CND				5
+#   define R500_ALPHA_OP_CMP				6
+#   define R500_ALPHA_OP_FRC				7
+#   define R500_ALPHA_OP_EX2				8
+#   define R500_ALPHA_OP_LN2				9
+#   define R500_ALPHA_OP_RCP				10
+#   define R500_ALPHA_OP_RSQ				11
+#   define R500_ALPHA_OP_SIN				12
+#   define R500_ALPHA_OP_COS				13
+#   define R500_ALPHA_OP_MDH				14
+#   define R500_ALPHA_OP_MDV				15
+#   define R500_ALPHA_ADDRD(x)				(x << 4)
+#   define R500_ALPHA_ADDRD_REL				(1 << 11)
+#  define R500_ALPHA_SEL_A_SHIFT			12
+#   define R500_ALPHA_SEL_A_SRC0			(0 << 12)
+#   define R500_ALPHA_SEL_A_SRC1			(1 << 12)
+#   define R500_ALPHA_SEL_A_SRC2			(2 << 12)
+#   define R500_ALPHA_SEL_A_SRCP			(3 << 12)
+#   define R500_ALPHA_SWIZ_A_R				(0 << 14)
+#   define R500_ALPHA_SWIZ_A_G				(1 << 14)
+#   define R500_ALPHA_SWIZ_A_B				(2 << 14)
+#   define R500_ALPHA_SWIZ_A_A				(3 << 14)
+#   define R500_ALPHA_SWIZ_A_0				(4 << 14)
+#   define R500_ALPHA_SWIZ_A_HALF			(5 << 14)
+#   define R500_ALPHA_SWIZ_A_1				(6 << 14)
+/* #define R500_ALPHA_SWIZ_A_UNUSED			(7 << 14) */
+#   define R500_ALPHA_MOD_A_NOP				(0 << 17)
+#   define R500_ALPHA_MOD_A_NEG				(1 << 17)
+#   define R500_ALPHA_MOD_A_ABS				(2 << 17)
+#   define R500_ALPHA_MOD_A_NAB				(3 << 17)
+#  define R500_ALPHA_SEL_B_SHIFT			19
+#   define R500_ALPHA_SEL_B_SRC0			(0 << 19)
+#   define R500_ALPHA_SEL_B_SRC1			(1 << 19)
+#   define R500_ALPHA_SEL_B_SRC2			(2 << 19)
+#   define R500_ALPHA_SEL_B_SRCP			(3 << 19)
+#   define R500_ALPHA_SWIZ_B_R				(0 << 21)
+#   define R500_ALPHA_SWIZ_B_G				(1 << 21)
+#   define R500_ALPHA_SWIZ_B_B				(2 << 21)
+#   define R500_ALPHA_SWIZ_B_A				(3 << 21)
+#   define R500_ALPHA_SWIZ_B_0				(4 << 21)
+#   define R500_ALPHA_SWIZ_B_HALF			(5 << 21)
+#   define R500_ALPHA_SWIZ_B_1				(6 << 21)
+/* #define R500_ALPHA_SWIZ_B_UNUSED			(7 << 21) */
+#   define R500_ALPHA_MOD_B_NOP				(0 << 24)
+#   define R500_ALPHA_MOD_B_NEG				(1 << 24)
+#   define R500_ALPHA_MOD_B_ABS				(2 << 24)
+#   define R500_ALPHA_MOD_B_NAB				(3 << 24)
+#   define R500_ALPHA_OMOD_IDENTITY			(0 << 26)
+#   define R500_ALPHA_OMOD_MUL_2			(1 << 26)
+#   define R500_ALPHA_OMOD_MUL_4			(2 << 26)
+#   define R500_ALPHA_OMOD_MUL_8			(3 << 26)
+#   define R500_ALPHA_OMOD_DIV_2			(4 << 26)
+#   define R500_ALPHA_OMOD_DIV_4			(5 << 26)
+#   define R500_ALPHA_OMOD_DIV_8			(6 << 26)
+#   define R500_ALPHA_OMOD_DISABLE			(7 << 26)
+#   define R500_ALPHA_TARGET(x)				(x << 29)
+#   define R500_ALPHA_W_OMASK				(1 << 31)
+#define R500_US_ALU_ALPHA_ADDR_0			0x9800
+#   define R500_ALPHA_ADDR0(x)				(x << 0)
+#   define R500_ALPHA_ADDR0_CONST			(1 << 8)
+#   define R500_ALPHA_ADDR0_REL				(1 << 9)
+#   define R500_ALPHA_ADDR1(x)				(x << 10)
+#   define R500_ALPHA_ADDR1_CONST			(1 << 18)
+#   define R500_ALPHA_ADDR1_REL				(1 << 19)
+#   define R500_ALPHA_ADDR2(x)				(x << 20)
+#   define R500_ALPHA_ADDR2_CONST			(1 << 28)
+#   define R500_ALPHA_ADDR2_REL				(1 << 29)
+#   define R500_ALPHA_SRCP_OP_1_MINUS_2A0		(0 << 30)
+#   define R500_ALPHA_SRCP_OP_A1_MINUS_A0		(1 << 30)
+#   define R500_ALPHA_SRCP_OP_A1_PLUS_A0		(2 << 30)
+#   define R500_ALPHA_SRCP_OP_1_MINUS_A0		(3 << 30)
+#define R500_US_ALU_RGBA_INST_0				0xb000
+#   define R500_ALU_RGBA_OP_MAD				(0 << 0)
+#   define R500_ALU_RGBA_OP_DP3				(1 << 0)
+#   define R500_ALU_RGBA_OP_DP4				(2 << 0)
+#   define R500_ALU_RGBA_OP_D2A				(3 << 0)
+#   define R500_ALU_RGBA_OP_MIN				(4 << 0)
+#   define R500_ALU_RGBA_OP_MAX				(5 << 0)
+/* #define R500_ALU_RGBA_OP_RESERVED			(6 << 0) */
+#   define R500_ALU_RGBA_OP_CND				(7 << 0)
+#   define R500_ALU_RGBA_OP_CMP				(8 << 0)
+#   define R500_ALU_RGBA_OP_FRC				(9 << 0)
+#   define R500_ALU_RGBA_OP_SOP				(10 << 0)
+#   define R500_ALU_RGBA_OP_MDH				(11 << 0)
+#   define R500_ALU_RGBA_OP_MDV				(12 << 0)
+#   define R500_ALU_RGBA_ADDRD(x)			(x << 4)
+#   define R500_ALU_RGBA_ADDRD_REL			(1 << 11)
+#  define R500_ALU_RGBA_SEL_C_SHIFT			12
+#   define R500_ALU_RGBA_SEL_C_SRC0			(0 << 12)
+#   define R500_ALU_RGBA_SEL_C_SRC1			(1 << 12)
+#   define R500_ALU_RGBA_SEL_C_SRC2			(2 << 12)
+#   define R500_ALU_RGBA_SEL_C_SRCP			(3 << 12)
+#   define R500_ALU_RGBA_R_SWIZ_R			(0 << 14)
+#   define R500_ALU_RGBA_R_SWIZ_G			(1 << 14)
+#   define R500_ALU_RGBA_R_SWIZ_B			(2 << 14)
+#   define R500_ALU_RGBA_R_SWIZ_A			(3 << 14)
+#   define R500_ALU_RGBA_R_SWIZ_0			(4 << 14)
+#   define R500_ALU_RGBA_R_SWIZ_HALF			(5 << 14)
+#   define R500_ALU_RGBA_R_SWIZ_1			(6 << 14)
+/* #define R500_ALU_RGBA_R_SWIZ_UNUSED			(7 << 14) */
+#   define R500_ALU_RGBA_G_SWIZ_R			(0 << 17)
+#   define R500_ALU_RGBA_G_SWIZ_G			(1 << 17)
+#   define R500_ALU_RGBA_G_SWIZ_B			(2 << 17)
+#   define R500_ALU_RGBA_G_SWIZ_A			(3 << 17)
+#   define R500_ALU_RGBA_G_SWIZ_0			(4 << 17)
+#   define R500_ALU_RGBA_G_SWIZ_HALF			(5 << 17)
+#   define R500_ALU_RGBA_G_SWIZ_1			(6 << 17)
+/* #define R500_ALU_RGBA_G_SWIZ_UNUSED			(7 << 17) */
+#   define R500_ALU_RGBA_B_SWIZ_R			(0 << 20)
+#   define R500_ALU_RGBA_B_SWIZ_G			(1 << 20)
+#   define R500_ALU_RGBA_B_SWIZ_B			(2 << 20)
+#   define R500_ALU_RGBA_B_SWIZ_A			(3 << 20)
+#   define R500_ALU_RGBA_B_SWIZ_0			(4 << 20)
+#   define R500_ALU_RGBA_B_SWIZ_HALF			(5 << 20)
+#   define R500_ALU_RGBA_B_SWIZ_1			(6 << 20)
+/* #define R500_ALU_RGBA_B_SWIZ_UNUSED			(7 << 20) */
+#   define R500_ALU_RGBA_MOD_C_NOP			(0 << 23)
+#   define R500_ALU_RGBA_MOD_C_NEG			(1 << 23)
+#   define R500_ALU_RGBA_MOD_C_ABS			(2 << 23)
+#   define R500_ALU_RGBA_MOD_C_NAB			(3 << 23)
+#  define R500_ALU_RGBA_ALPHA_SEL_C_SHIFT		25
+#   define R500_ALU_RGBA_ALPHA_SEL_C_SRC0		(0 << 25)
+#   define R500_ALU_RGBA_ALPHA_SEL_C_SRC1		(1 << 25)
+#   define R500_ALU_RGBA_ALPHA_SEL_C_SRC2		(2 << 25)
+#   define R500_ALU_RGBA_ALPHA_SEL_C_SRCP		(3 << 25)
+#   define R500_ALU_RGBA_A_SWIZ_R			(0 << 27)
+#   define R500_ALU_RGBA_A_SWIZ_G			(1 << 27)
+#   define R500_ALU_RGBA_A_SWIZ_B			(2 << 27)
+#   define R500_ALU_RGBA_A_SWIZ_A			(3 << 27)
+#   define R500_ALU_RGBA_A_SWIZ_0			(4 << 27)
+#   define R500_ALU_RGBA_A_SWIZ_HALF			(5 << 27)
+#   define R500_ALU_RGBA_A_SWIZ_1			(6 << 27)
+/* #define R500_ALU_RGBA_A_SWIZ_UNUSED			(7 << 27) */
+#   define R500_ALU_RGBA_ALPHA_MOD_C_NOP		(0 << 30)
+#   define R500_ALU_RGBA_ALPHA_MOD_C_NEG		(1 << 30)
+#   define R500_ALU_RGBA_ALPHA_MOD_C_ABS		(2 << 30)
+#   define R500_ALU_RGBA_ALPHA_MOD_C_NAB		(3 << 30)
+#define R500_US_ALU_RGB_INST_0				0xa000
+#  define R500_ALU_RGB_SEL_A_SHIFT			0
+#   define R500_ALU_RGB_SEL_A_SRC0			(0 << 0)
+#   define R500_ALU_RGB_SEL_A_SRC1			(1 << 0)
+#   define R500_ALU_RGB_SEL_A_SRC2			(2 << 0)
+#   define R500_ALU_RGB_SEL_A_SRCP			(3 << 0)
+#   define R500_ALU_RGB_R_SWIZ_A_R			(0 << 2)
+#   define R500_ALU_RGB_R_SWIZ_A_G			(1 << 2)
+#   define R500_ALU_RGB_R_SWIZ_A_B			(2 << 2)
+#   define R500_ALU_RGB_R_SWIZ_A_A			(3 << 2)
+#   define R500_ALU_RGB_R_SWIZ_A_0			(4 << 2)
+#   define R500_ALU_RGB_R_SWIZ_A_HALF			(5 << 2)
+#   define R500_ALU_RGB_R_SWIZ_A_1			(6 << 2)
+/* #define R500_ALU_RGB_R_SWIZ_A_UNUSED			(7 << 2) */
+#   define R500_ALU_RGB_G_SWIZ_A_R			(0 << 5)
+#   define R500_ALU_RGB_G_SWIZ_A_G			(1 << 5)
+#   define R500_ALU_RGB_G_SWIZ_A_B			(2 << 5)
+#   define R500_ALU_RGB_G_SWIZ_A_A			(3 << 5)
+#   define R500_ALU_RGB_G_SWIZ_A_0			(4 << 5)
+#   define R500_ALU_RGB_G_SWIZ_A_HALF			(5 << 5)
+#   define R500_ALU_RGB_G_SWIZ_A_1			(6 << 5)
+/* #define R500_ALU_RGB_G_SWIZ_A_UNUSED			(7 << 5) */
+#   define R500_ALU_RGB_B_SWIZ_A_R			(0 << 8)
+#   define R500_ALU_RGB_B_SWIZ_A_G			(1 << 8)
+#   define R500_ALU_RGB_B_SWIZ_A_B			(2 << 8)
+#   define R500_ALU_RGB_B_SWIZ_A_A			(3 << 8)
+#   define R500_ALU_RGB_B_SWIZ_A_0			(4 << 8)
+#   define R500_ALU_RGB_B_SWIZ_A_HALF			(5 << 8)
+#   define R500_ALU_RGB_B_SWIZ_A_1			(6 << 8)
+/* #define R500_ALU_RGB_B_SWIZ_A_UNUSED			(7 << 8) */
+#   define R500_ALU_RGB_MOD_A_NOP			(0 << 11)
+#   define R500_ALU_RGB_MOD_A_NEG			(1 << 11)
+#   define R500_ALU_RGB_MOD_A_ABS			(2 << 11)
+#   define R500_ALU_RGB_MOD_A_NAB			(3 << 11)
+#  define R500_ALU_RGB_SEL_B_SHIFT			13
+#   define R500_ALU_RGB_SEL_B_SRC0			(0 << 13)
+#   define R500_ALU_RGB_SEL_B_SRC1			(1 << 13)
+#   define R500_ALU_RGB_SEL_B_SRC2			(2 << 13)
+#   define R500_ALU_RGB_SEL_B_SRCP			(3 << 13)
+#   define R500_ALU_RGB_R_SWIZ_B_R			(0 << 15)
+#   define R500_ALU_RGB_R_SWIZ_B_G			(1 << 15)
+#   define R500_ALU_RGB_R_SWIZ_B_B			(2 << 15)
+#   define R500_ALU_RGB_R_SWIZ_B_A			(3 << 15)
+#   define R500_ALU_RGB_R_SWIZ_B_0			(4 << 15)
+#   define R500_ALU_RGB_R_SWIZ_B_HALF			(5 << 15)
+#   define R500_ALU_RGB_R_SWIZ_B_1			(6 << 15)
+/* #define R500_ALU_RGB_R_SWIZ_B_UNUSED			(7 << 15) */
+#   define R500_ALU_RGB_G_SWIZ_B_R			(0 << 18)
+#   define R500_ALU_RGB_G_SWIZ_B_G			(1 << 18)
+#   define R500_ALU_RGB_G_SWIZ_B_B			(2 << 18)
+#   define R500_ALU_RGB_G_SWIZ_B_A			(3 << 18)
+#   define R500_ALU_RGB_G_SWIZ_B_0			(4 << 18)
+#   define R500_ALU_RGB_G_SWIZ_B_HALF			(5 << 18)
+#   define R500_ALU_RGB_G_SWIZ_B_1			(6 << 18)
+/* #define R500_ALU_RGB_G_SWIZ_B_UNUSED			(7 << 18) */
+#   define R500_ALU_RGB_B_SWIZ_B_R			(0 << 21)
+#   define R500_ALU_RGB_B_SWIZ_B_G			(1 << 21)
+#   define R500_ALU_RGB_B_SWIZ_B_B			(2 << 21)
+#   define R500_ALU_RGB_B_SWIZ_B_A			(3 << 21)
+#   define R500_ALU_RGB_B_SWIZ_B_0			(4 << 21)
+#   define R500_ALU_RGB_B_SWIZ_B_HALF			(5 << 21)
+#   define R500_ALU_RGB_B_SWIZ_B_1			(6 << 21)
+/* #define R500_ALU_RGB_B_SWIZ_B_UNUSED			(7 << 21) */
+#   define R500_ALU_RGB_MOD_B_NOP			(0 << 24)
+#   define R500_ALU_RGB_MOD_B_NEG			(1 << 24)
+#   define R500_ALU_RGB_MOD_B_ABS			(2 << 24)
+#   define R500_ALU_RGB_MOD_B_NAB			(3 << 24)
+#   define R500_ALU_RGB_OMOD_IDENTITY			(0 << 26)
+#   define R500_ALU_RGB_OMOD_MUL_2			(1 << 26)
+#   define R500_ALU_RGB_OMOD_MUL_4			(2 << 26)
+#   define R500_ALU_RGB_OMOD_MUL_8			(3 << 26)
+#   define R500_ALU_RGB_OMOD_DIV_2			(4 << 26)
+#   define R500_ALU_RGB_OMOD_DIV_4			(5 << 26)
+#   define R500_ALU_RGB_OMOD_DIV_8			(6 << 26)
+#   define R500_ALU_RGB_OMOD_DISABLE			(7 << 26)
+#   define R500_ALU_RGB_TARGET(x)			(x << 29)
+#   define R500_ALU_RGB_WMASK				(1 << 31)
+#define R500_US_ALU_RGB_ADDR_0				0x9000
+#   define R500_RGB_ADDR0(x)				(x << 0)
+#   define R500_RGB_ADDR0_CONST				(1 << 8)
+#   define R500_RGB_ADDR0_REL				(1 << 9)
+#   define R500_RGB_ADDR1(x)				(x << 10)
+#   define R500_RGB_ADDR1_CONST				(1 << 18)
+#   define R500_RGB_ADDR1_REL				(1 << 19)
+#   define R500_RGB_ADDR2(x)				(x << 20)
+#   define R500_RGB_ADDR2_CONST				(1 << 28)
+#   define R500_RGB_ADDR2_REL				(1 << 29)
+#   define R500_RGB_SRCP_OP_1_MINUS_2RGB0		(0 << 30)
+#   define R500_RGB_SRCP_OP_RGB1_MINUS_RGB0		(1 << 30)
+#   define R500_RGB_SRCP_OP_RGB1_PLUS_RGB0		(2 << 30)
+#   define R500_RGB_SRCP_OP_1_MINUS_RGB0		(3 << 30)
+#define R500_US_CMN_INST_0				0xb800
+#  define R500_INST_TYPE_MASK				(3 << 0)
+#   define R500_INST_TYPE_ALU				(0 << 0)
+#   define R500_INST_TYPE_OUT				(1 << 0)
+#   define R500_INST_TYPE_FC				(2 << 0)
+#   define R500_INST_TYPE_TEX				(3 << 0)
+#   define R500_INST_TEX_SEM_WAIT			(1 << 2)
+#   define R500_INST_RGB_PRED_SEL_NONE			(0 << 3)
+#   define R500_INST_RGB_PRED_SEL_RGBA			(1 << 3)
+#   define R500_INST_RGB_PRED_SEL_RRRR			(2 << 3)
+#   define R500_INST_RGB_PRED_SEL_GGGG			(3 << 3)
+#   define R500_INST_RGB_PRED_SEL_BBBB			(4 << 3)
+#   define R500_INST_RGB_PRED_SEL_AAAA			(5 << 3)
+#   define R500_INST_RGB_PRED_INV			(1 << 6)
+#   define R500_INST_WRITE_INACTIVE			(1 << 7)
+#   define R500_INST_LAST				(1 << 8)
+#   define R500_INST_NOP				(1 << 9)
+#   define R500_INST_ALU_WAIT				(1 << 10)
+#   define R500_INST_RGB_WMASK_R			(1 << 11)
+#   define R500_INST_RGB_WMASK_G			(1 << 12)
+#   define R500_INST_RGB_WMASK_B			(1 << 13)
+#   define R500_INST_ALPHA_WMASK			(1 << 14)
+#   define R500_INST_RGB_OMASK_R			(1 << 15)
+#   define R500_INST_RGB_OMASK_G			(1 << 16)
+#   define R500_INST_RGB_OMASK_B			(1 << 17)
+#   define R500_INST_ALPHA_OMASK			(1 << 18)
+#   define R500_INST_RGB_CLAMP				(1 << 19)
+#   define R500_INST_ALPHA_CLAMP			(1 << 20)
+#   define R500_INST_ALU_RESULT_SEL			(1 << 21)
+#   define R500_INST_ALPHA_PRED_INV			(1 << 22)
+#   define R500_INST_ALU_RESULT_OP_EQ			(0 << 23)
+#   define R500_INST_ALU_RESULT_OP_LT			(1 << 23)
+#   define R500_INST_ALU_RESULT_OP_GE			(2 << 23)
+#   define R500_INST_ALU_RESULT_OP_NE			(3 << 23)
+#   define R500_INST_ALPHA_PRED_SEL_NONE		(0 << 25)
+#   define R500_INST_ALPHA_PRED_SEL_RGBA		(1 << 25)
+#   define R500_INST_ALPHA_PRED_SEL_RRRR		(2 << 25)
+#   define R500_INST_ALPHA_PRED_SEL_GGGG		(3 << 25)
+#   define R500_INST_ALPHA_PRED_SEL_BBBB		(4 << 25)
+#   define R500_INST_ALPHA_PRED_SEL_AAAA		(5 << 25)
+/* XXX next four are kind of guessed */
+#   define R500_INST_STAT_WE_R				(1 << 28)
+#   define R500_INST_STAT_WE_G				(1 << 29)
+#   define R500_INST_STAT_WE_B				(1 << 30)
+#   define R500_INST_STAT_WE_A				(1 << 31)
+
+/* note that these are 8 bit lengths, despite the offsets, at least for R500 */
+#define R500_US_CODE_ADDR				0x4630
+#   define R500_US_CODE_START_ADDR(x)			(x << 0)
+#   define R500_US_CODE_END_ADDR(x)			(x << 16)
+#define R500_US_CODE_OFFSET				0x4638
+#   define R500_US_CODE_OFFSET_ADDR(x)			(x << 0)
+#define R500_US_CODE_RANGE				0x4634
+#   define R500_US_CODE_RANGE_ADDR(x)			(x << 0)
+#   define R500_US_CODE_RANGE_SIZE(x)			(x << 16)
+#define R500_US_CONFIG					0x4600
+#   define R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO		(1 << 1)
+#define R500_US_FC_ADDR_0				0xa000
+#   define R500_FC_BOOL_ADDR(x)				(x << 0)
+#   define R500_FC_INT_ADDR(x)				(x << 8)
+#   define R500_FC_JUMP_ADDR(x)				(x << 16)
+#   define R500_FC_JUMP_GLOBAL				(1 << 31)
+#define R500_US_FC_BOOL_CONST				0x4620
+#   define R500_FC_KBOOL(x)				(x)
+#define R500_US_FC_CTRL					0x4624
+#   define R500_FC_TEST_EN				(1 << 30)
+#   define R500_FC_FULL_FC_EN				(1 << 31)
+#define R500_US_FC_INST_0				0x9800
+#   define R500_FC_OP_JUMP				(0 << 0)
+#   define R500_FC_OP_LOOP				(1 << 0)
+#   define R500_FC_OP_ENDLOOP				(2 << 0)
+#   define R500_FC_OP_REP				(3 << 0)
+#   define R500_FC_OP_ENDREP				(4 << 0)
+#   define R500_FC_OP_BREAKLOOP				(5 << 0)
+#   define R500_FC_OP_BREAKREP				(6 << 0)
+#   define R500_FC_OP_CONTINUE				(7 << 0)
+#   define R500_FC_B_ELSE				(1 << 4)
+#   define R500_FC_JUMP_ANY				(1 << 5)
+#   define R500_FC_A_OP_NONE				(0 << 6)
+#   define R500_FC_A_OP_POP				(1 << 6)
+#   define R500_FC_A_OP_PUSH				(2 << 6)
+#   define R500_FC_JUMP_FUNC(x)				(x << 8)
+#   define R500_FC_B_POP_CNT(x)				(x << 16)
+#   define R500_FC_B_OP0_NONE				(0 << 24)
+#   define R500_FC_B_OP0_DECR				(1 << 24)
+#   define R500_FC_B_OP0_INCR				(2 << 24)
+#   define R500_FC_B_OP1_DECR				(0 << 26)
+#   define R500_FC_B_OP1_NONE				(1 << 26)
+#   define R500_FC_B_OP1_INCR				(2 << 26)
+#   define R500_FC_IGNORE_UNCOVERED			(1 << 28)
+#define R500_US_FC_INT_CONST_0				0x4c00
+#   define R500_FC_INT_CONST_KR(x)			(x << 0)
+#   define R500_FC_INT_CONST_KG(x)			(x << 8)
+#   define R500_FC_INT_CONST_KB(x)			(x << 16)
+/* _0 through _15 */
+#define R500_US_FORMAT0_0				0x4640
+#   define R500_FORMAT_TXWIDTH(x)			(x << 0)
+#   define R500_FORMAT_TXHEIGHT(x)			(x << 11)
+#   define R500_FORMAT_TXDEPTH(x)			(x << 22)
+/* _0 through _3 */
+#define R500_US_OUT_FMT_0				0x46A4
+#   define R500_OUT_FMT_C4_8				(0 << 0)
+#   define R500_OUT_FMT_C4_10				(1 << 0)
+#   define R500_OUT_FMT_C4_10_GAMMA			(2 << 0)
+#   define R500_OUT_FMT_C_16				(3 << 0)
+#   define R500_OUT_FMT_C2_16				(4 << 0)
+#   define R500_OUT_FMT_C4_16				(5 << 0)
+#   define R500_OUT_FMT_C_16_MPEG			(6 << 0)
+#   define R500_OUT_FMT_C2_16_MPEG			(7 << 0)
+#   define R500_OUT_FMT_C2_4				(8 << 0)
+#   define R500_OUT_FMT_C_3_3_2				(9 << 0)
+#   define R500_OUT_FMT_C_6_5_6				(10 << 0)
+#   define R500_OUT_FMT_C_11_11_10			(11 << 0)
+#   define R500_OUT_FMT_C_10_11_11			(12 << 0)
+#   define R500_OUT_FMT_C_2_10_10_10			(13 << 0)
+/* #define R500_OUT_FMT_RESERVED			(14 << 0) */
+#   define R500_OUT_FMT_UNUSED				(15 << 0)
+#   define R500_OUT_FMT_C_16_FP				(16 << 0)
+#   define R500_OUT_FMT_C2_16_FP			(17 << 0)
+#   define R500_OUT_FMT_C4_16_FP			(18 << 0)
+#   define R500_OUT_FMT_C_32_FP				(19 << 0)
+#   define R500_OUT_FMT_C2_32_FP			(20 << 0)
+#   define R500_OUT_FMT_C4_32_FP			(21 << 0)
+#   define R500_C0_SEL_A				(0 << 8)
+#   define R500_C0_SEL_R				(1 << 8)
+#   define R500_C0_SEL_G				(2 << 8)
+#   define R500_C0_SEL_B				(3 << 8)
+#   define R500_C1_SEL_A				(0 << 10)
+#   define R500_C1_SEL_R				(1 << 10)
+#   define R500_C1_SEL_G				(2 << 10)
+#   define R500_C1_SEL_B				(3 << 10)
+#   define R500_C2_SEL_A				(0 << 12)
+#   define R500_C2_SEL_R				(1 << 12)
+#   define R500_C2_SEL_G				(2 << 12)
+#   define R500_C2_SEL_B				(3 << 12)
+#   define R500_C3_SEL_A				(0 << 14)
+#   define R500_C3_SEL_R				(1 << 14)
+#   define R500_C3_SEL_G				(2 << 14)
+#   define R500_C3_SEL_B				(3 << 14)
+#   define R500_OUT_SIGN(x)				(x << 16)
+#   define R500_ROUND_ADJ				(1 << 20)
+#define R500_US_PIXSIZE					0x4604
+#   define R500_PIX_SIZE(x)				(x)
+#define R500_US_TEX_ADDR_0				0x9800
+#   define R500_TEX_SRC_ADDR(x)				(x << 0)
+#   define R500_TEX_SRC_ADDR_REL			(1 << 7)
+#   define R500_TEX_SRC_S_SWIZ_R			(0 << 8)
+#   define R500_TEX_SRC_S_SWIZ_G			(1 << 8)
+#   define R500_TEX_SRC_S_SWIZ_B			(2 << 8)
+#   define R500_TEX_SRC_S_SWIZ_A			(3 << 8)
+#   define R500_TEX_SRC_T_SWIZ_R			(0 << 10)
+#   define R500_TEX_SRC_T_SWIZ_G			(1 << 10)
+#   define R500_TEX_SRC_T_SWIZ_B			(2 << 10)
+#   define R500_TEX_SRC_T_SWIZ_A			(3 << 10)
+#   define R500_TEX_SRC_R_SWIZ_R			(0 << 12)
+#   define R500_TEX_SRC_R_SWIZ_G			(1 << 12)
+#   define R500_TEX_SRC_R_SWIZ_B			(2 << 12)
+#   define R500_TEX_SRC_R_SWIZ_A			(3 << 12)
+#   define R500_TEX_SRC_Q_SWIZ_R			(0 << 14)
+#   define R500_TEX_SRC_Q_SWIZ_G			(1 << 14)
+#   define R500_TEX_SRC_Q_SWIZ_B			(2 << 14)
+#   define R500_TEX_SRC_Q_SWIZ_A			(3 << 14)
+#   define R500_TEX_DST_ADDR(x)				(x << 16)
+#   define R500_TEX_DST_ADDR_REL			(1 << 23)
+#   define R500_TEX_DST_R_SWIZ_R			(0 << 24)
+#   define R500_TEX_DST_R_SWIZ_G			(1 << 24)
+#   define R500_TEX_DST_R_SWIZ_B			(2 << 24)
+#   define R500_TEX_DST_R_SWIZ_A			(3 << 24)
+#   define R500_TEX_DST_G_SWIZ_R			(0 << 26)
+#   define R500_TEX_DST_G_SWIZ_G			(1 << 26)
+#   define R500_TEX_DST_G_SWIZ_B			(2 << 26)
+#   define R500_TEX_DST_G_SWIZ_A			(3 << 26)
+#   define R500_TEX_DST_B_SWIZ_R			(0 << 28)
+#   define R500_TEX_DST_B_SWIZ_G			(1 << 28)
+#   define R500_TEX_DST_B_SWIZ_B			(2 << 28)
+#   define R500_TEX_DST_B_SWIZ_A			(3 << 28)
+#   define R500_TEX_DST_A_SWIZ_R			(0 << 30)
+#   define R500_TEX_DST_A_SWIZ_G			(1 << 30)
+#   define R500_TEX_DST_A_SWIZ_B			(2 << 30)
+#   define R500_TEX_DST_A_SWIZ_A			(3 << 30)
+#define R500_US_TEX_ADDR_DXDY_0				0xa000
+#   define R500_DX_ADDR(x)				(x << 0)
+#   define R500_DX_ADDR_REL				(1 << 7)
+#   define R500_DX_S_SWIZ_R				(0 << 8)
+#   define R500_DX_S_SWIZ_G				(1 << 8)
+#   define R500_DX_S_SWIZ_B				(2 << 8)
+#   define R500_DX_S_SWIZ_A				(3 << 8)
+#   define R500_DX_T_SWIZ_R				(0 << 10)
+#   define R500_DX_T_SWIZ_G				(1 << 10)
+#   define R500_DX_T_SWIZ_B				(2 << 10)
+#   define R500_DX_T_SWIZ_A				(3 << 10)
+#   define R500_DX_R_SWIZ_R				(0 << 12)
+#   define R500_DX_R_SWIZ_G				(1 << 12)
+#   define R500_DX_R_SWIZ_B				(2 << 12)
+#   define R500_DX_R_SWIZ_A				(3 << 12)
+#   define R500_DX_Q_SWIZ_R				(0 << 14)
+#   define R500_DX_Q_SWIZ_G				(1 << 14)
+#   define R500_DX_Q_SWIZ_B				(2 << 14)
+#   define R500_DX_Q_SWIZ_A				(3 << 14)
+#   define R500_DY_ADDR(x)				(x << 16)
+#   define R500_DY_ADDR_REL				(1 << 17)
+#   define R500_DY_S_SWIZ_R				(0 << 24)
+#   define R500_DY_S_SWIZ_G				(1 << 24)
+#   define R500_DY_S_SWIZ_B				(2 << 24)
+#   define R500_DY_S_SWIZ_A				(3 << 24)
+#   define R500_DY_T_SWIZ_R				(0 << 26)
+#   define R500_DY_T_SWIZ_G				(1 << 26)
+#   define R500_DY_T_SWIZ_B				(2 << 26)
+#   define R500_DY_T_SWIZ_A				(3 << 26)
+#   define R500_DY_R_SWIZ_R				(0 << 28)
+#   define R500_DY_R_SWIZ_G				(1 << 28)
+#   define R500_DY_R_SWIZ_B				(2 << 28)
+#   define R500_DY_R_SWIZ_A				(3 << 28)
+#   define R500_DY_Q_SWIZ_R				(0 << 30)
+#   define R500_DY_Q_SWIZ_G				(1 << 30)
+#   define R500_DY_Q_SWIZ_B				(2 << 30)
+#   define R500_DY_Q_SWIZ_A				(3 << 30)
+#define R500_US_TEX_INST_0				0x9000
+#   define R500_TEX_ID(x)				(x << 16)
+#   define R500_TEX_INST_NOP				(0 << 22)
+#   define R500_TEX_INST_LD				(1 << 22)
+#   define R500_TEX_INST_TEXKILL			(2 << 22)
+#   define R500_TEX_INST_PROJ				(3 << 22)
+#   define R500_TEX_INST_LODBIAS			(4 << 22)
+#   define R500_TEX_INST_LOD				(5 << 22)
+#   define R500_TEX_INST_DXDY				(6 << 22)
+#   define R500_TEX_SEM_ACQUIRE				(1 << 25)
+#   define R500_TEX_IGNORE_UNCOVERED			(1 << 26)
+#   define R500_TEX_UNSCALED				(1 << 27)
+#define R300_US_W_FMT					0x46b4
+#   define R300_W_FMT_W0				(0 << 0)
+#   define R300_W_FMT_W24				(1 << 0)
+#   define R300_W_FMT_W24FP				(2 << 0)
+#   define R300_W_SRC_US				(0 << 2)
+#   define R300_W_SRC_RAS				(1 << 2)
+
+
+/* Draw a primitive from vertex data in arrays loaded via 3D_LOAD_VBPNTR.
+ * Two parameter dwords:
+ * 0. VAP_VTX_FMT: The first parameter is not written to hardware
+ * 1. VAP_VF_CTL: The second parameter is a standard primitive emission dword.
+ */
+#define R300_PACKET3_3D_DRAW_VBUF           0x00002800
+
+/* Draw a primitive from immediate vertices in this packet
+ * Up to 16382 dwords:
+ * 0. VAP_VTX_FMT: The first parameter is not written to hardware
+ * 1. VAP_VF_CTL: The second parameter is a standard primitive emission dword.
+ * 2 to end: Up to 16380 dwords of vertex data.
+ */
+#define R300_PACKET3_3D_DRAW_IMMD           0x00002900
+
+/* Draw a primitive from vertex data in arrays loaded via 3D_LOAD_VBPNTR and
+ * immediate vertices in this packet
+ * Up to 16382 dwords:
+ * 0. VAP_VTX_FMT: The first parameter is not written to hardware
+ * 1. VAP_VF_CTL: The second parameter is a standard primitive emission dword.
+ * 2 to end: Up to 16380 dwords of vertex data.
+ */
+#define R300_PACKET3_3D_DRAW_INDX           0x00002A00
+
+
+/* Specify the full set of vertex arrays as (address, stride).
+ * The first parameter is the number of vertex arrays specified.
+ * The rest of the command is a variable length list of blocks, where
+ * each block is three dwords long and specifies two arrays.
+ * The first dword of a block is split into two words, the lower significant
+ * word refers to the first array, the more significant word to the second
+ * array in the block.
+ * The low byte of each word contains the size of an array entry in dwords,
+ * the high byte contains the stride of the array.
+ * The second dword of a block contains the pointer to the first array,
+ * the third dword of a block contains the pointer to the second array.
+ * Note that if the total number of arrays is odd, the third dword of
+ * the last block is omitted.
+ */
+#define R300_PACKET3_3D_LOAD_VBPNTR         0x00002F00
+
+#define R300_PACKET3_INDX_BUFFER            0x00003300
+#    define R300_EB_UNK1_SHIFT                      24
+#    define R300_EB_UNK1                    (0x80<<24)
+#    define R300_EB_UNK2                        0x0810
+
+/* Same as R300_PACKET3_3D_DRAW_VBUF but without VAP_VTX_FMT */
+#define R300_PACKET3_3D_DRAW_VBUF_2         0x00003400
+/* Same as R300_PACKET3_3D_DRAW_IMMD but without VAP_VTX_FMT */
+#define R300_PACKET3_3D_DRAW_IMMD_2         0x00003500
+/* Same as R300_PACKET3_3D_DRAW_INDX but without VAP_VTX_FMT */
+#define R300_PACKET3_3D_DRAW_INDX_2         0x00003600
+
+/* Clears a portion of hierachical Z RAM
+ * 3 dword parameters
+ * 0. START
+ * 1. COUNT: 13:0 (max is 0x3FFF)
+ * 2. CLEAR_VALUE: Value to write into HIZ RAM.
+ */
+#define R300_PACKET3_3D_CLEAR_HIZ           0x00003700
+
+/* Draws a set of primitives using vertex buffers pointed by the state data.
+ * At least 2 Parameters:
+ * 0. VAP_VF_CNTL: The first parameter is a standard primitive emission dword.
+ * 2 to end: Data or indices (see other 3D_DRAW_* packets for details)
+ */
+#define R300_PACKET3_3D_DRAW_128            0x00003900
+
+/* END: Packet 3 commands */
+
+
+/* Color formats for 2d packets
+ */
+#define R300_CP_COLOR_FORMAT_CI8	2
+#define R300_CP_COLOR_FORMAT_ARGB1555	3
+#define R300_CP_COLOR_FORMAT_RGB565	4
+#define R300_CP_COLOR_FORMAT_ARGB8888	6
+#define R300_CP_COLOR_FORMAT_RGB332	7
+#define R300_CP_COLOR_FORMAT_RGB8	9
+#define R300_CP_COLOR_FORMAT_ARGB4444	15
+
+/*
+ * CP type-3 packets
+ */
+#define R300_CP_CMD_BITBLT_MULTI	0xC0009B00
+
+/* XXX Corbin's stuff from radeon and r200 */
+
+#define RADEON_WAIT_UNTIL                   0x1720
+#       define RADEON_WAIT_CRTC_PFLIP       (1 << 0)
+#       define RADEON_WAIT_2D_IDLECLEAN     (1 << 16)
+#       define RADEON_WAIT_3D_IDLECLEAN     (1 << 17)
+#       define RADEON_WAIT_HOST_IDLECLEAN   (1 << 18)
+
+#define RADEON_CP_PACKET3                           0xC0000000
+
+#define R200_3D_DRAW_IMMD_2      0xC0003500
+
+#endif /* _R300_REG_H */
+
+/* *INDENT-ON* */
+
+/* vim: set foldenable foldmarker=\\{,\\} foldmethod=marker : */
diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c
new file mode 100644
index 0000000000..8ed66a1660
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_screen.c
@@ -0,0 +1,271 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_screen.h"
+
+/* Return the identifier behind whom the brave coders responsible for this
+ * amalgamation of code, sweat, and duct tape, routinely obscure their names.
+ *
+ * ...I should have just put "Corbin Simpson", but I'm not that cool.
+ *
+ * (Or egotistical. Yet.) */
+static const char* r300_get_vendor(struct pipe_screen* pscreen)
+{
+    return "X.Org R300 Project";
+}
+
+static const char* chip_families[] = {
+    "R300",
+    "R350",
+    "R360",
+    "RV350",
+    "RV370",
+    "RV380",
+    "R420",
+    "R423",
+    "R430",
+    "R480",
+    "R481",
+    "RV410",
+    "RS400",
+    "RC410",
+    "RS480",
+    "RS482",
+    "RS690",
+    "RS740",
+    "RV515",
+    "R520",
+    "RV530",
+    "R580",
+    "RV560",
+    "RV570"
+};
+
+static const char* r300_get_name(struct pipe_screen* pscreen)
+{
+    struct r300_screen* r300screen = r300_screen(pscreen);
+
+    return chip_families[r300screen->caps->family];
+}
+
+static int r300_get_param(struct pipe_screen* pscreen, int param)
+{
+    struct r300_screen* r300screen = r300_screen(pscreen);
+
+    switch (param) {
+        /* XXX cases marked "IN THEORY" are possible on the hardware,
+         * but haven't been implemented yet. */
+        case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+            /* XXX I'm told this goes up to 16 */
+            return 8;
+        case PIPE_CAP_NPOT_TEXTURES:
+            /* IN THEORY */
+            return 0;
+        case PIPE_CAP_TWO_SIDED_STENCIL:
+            if (r300screen->caps->is_r500) {
+                return 1;
+            } else {
+                return 0;
+            }
+            return 0;
+        case PIPE_CAP_GLSL:
+            /* IN THEORY */
+            return 0;
+        case PIPE_CAP_S3TC:
+            /* IN THEORY */
+            return 0;
+        case PIPE_CAP_ANISOTROPIC_FILTER:
+            /* IN THEORY */
+            return 0;
+        case PIPE_CAP_POINT_SPRITE:
+            /* IN THEORY */
+            return 0;
+        case PIPE_CAP_MAX_RENDER_TARGETS:
+            /* XXX 4 eventually */
+            return 1;
+        case PIPE_CAP_OCCLUSION_QUERY:
+            /* IN THEORY */
+            return 0;
+        case PIPE_CAP_TEXTURE_SHADOW_MAP:
+            /* IN THEORY */
+            return 0;
+        case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+            if (r300screen->caps->is_r500) {
+                /* 13 == 4096x4096 */
+                return 13;
+            } else {
+                /* 12 == 2048x2048 */
+                return 12;
+            }
+        case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+            /* So, technically, the limit is the same as above, but some math
+             * shows why this is silly. Assuming RGBA, 4cpp, we can see that
+             * 4096*4096*4096 = 64.0 GiB exactly, so it's not exactly
+             * practical. However, if at some point a game really wants this,
+             * then we can remove this limit. */
+            if (r300screen->caps->is_r500) {
+                /* 9 == 256x256x256 */
+                return 9;
+            } else {
+                /* 8 == 128*128*128 */
+                return 8;
+            }
+        case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+            if (r300screen->caps->is_r500) {
+                /* 13 == 4096x4096 */
+                return 13;
+            } else {
+                /* 12 == 2048x2048 */
+                return 12;
+            }
+        case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
+            return 1;
+        case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
+            return 1;
+        case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
+            /* XXX guessing */
+            return 2;
+        default:
+            debug_printf("r300: Implementation error: Bad param %d\n",
+                param);
+            return 0;
+    }
+}
+
+static float r300_get_paramf(struct pipe_screen* pscreen, int param)
+{
+    switch (param) {
+        case PIPE_CAP_MAX_LINE_WIDTH:
+        case PIPE_CAP_MAX_LINE_WIDTH_AA:
+            /* XXX this is the biggest thing that will fit in that register.
+            * Perhaps the actual rendering limits are less? */
+            return 10922.0f;
+        case PIPE_CAP_MAX_POINT_WIDTH:
+        case PIPE_CAP_MAX_POINT_WIDTH_AA:
+            /* XXX this is the biggest thing that will fit in that register.
+             * Perhaps the actual rendering limits are less? */
+            return 10922.0f;
+        case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+            return 16.0f;
+        case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+            return 16.0f;
+        default:
+            debug_printf("r300: Implementation error: Bad paramf %d\n",
+                param);
+            return 0.0f;
+    }
+}
+
+/* XXX moar formats */
+static boolean check_tex_2d_format(enum pipe_format format)
+{
+    switch (format) {
+        case PIPE_FORMAT_A8R8G8B8_UNORM:
+        case PIPE_FORMAT_I8_UNORM:
+            return TRUE;
+        default:
+            debug_printf("r300: Warning: Got unknown format: %s, in %s\n",
+                pf_name(format), __FUNCTION__);
+            break;
+    }
+
+    return FALSE;
+}
+
+/* XXX moar targets */
+static boolean r300_is_format_supported(struct pipe_screen* pscreen,
+                                        enum pipe_format format,
+                                        enum pipe_texture_target target,
+                                        unsigned tex_usage,
+                                        unsigned geom_flags)
+{
+    switch (target) {
+        case PIPE_TEXTURE_2D:
+            return check_tex_2d_format(format);
+        default:
+            debug_printf("r300: Warning: Got unknown format target: %d\n",
+                format);
+            break;
+    }
+
+    return FALSE;
+}
+
+static void* r300_surface_map(struct pipe_screen* screen,
+                              struct pipe_surface* surface,
+                              unsigned flags)
+{
+    struct r300_texture* tex = (struct r300_texture*)surface->texture;
+    char* map = pipe_buffer_map(screen, tex->buffer, flags);
+
+    if (!map) {
+        return NULL;
+    }
+
+    return map + surface->offset;
+}
+
+static void r300_surface_unmap(struct pipe_screen* screen,
+                               struct pipe_surface* surface)
+{
+    struct r300_texture* tex = (struct r300_texture*)surface->texture;
+    pipe_buffer_unmap(screen, tex->buffer);
+}
+
+static void r300_destroy_screen(struct pipe_screen* pscreen)
+{
+    struct r300_screen* r300screen = r300_screen(pscreen);
+
+    FREE(r300screen->caps);
+    FREE(r300screen);
+}
+
+struct pipe_screen* r300_create_screen(struct pipe_winsys* winsys,
+                                       struct r300_winsys* r300_winsys)
+{
+    struct r300_screen* r300screen = CALLOC_STRUCT(r300_screen);
+    struct r300_capabilities* caps = CALLOC_STRUCT(r300_capabilities);
+
+    if (!r300screen || !caps)
+        return NULL;
+
+    caps->pci_id = r300_winsys->pci_id;
+    caps->num_frag_pipes = r300_winsys->gb_pipes;
+
+    r300_parse_chipset(caps);
+
+    r300screen->caps = caps;
+    r300screen->screen.winsys = winsys;
+    r300screen->screen.destroy = r300_destroy_screen;
+    r300screen->screen.get_name = r300_get_name;
+    r300screen->screen.get_vendor = r300_get_vendor;
+    r300screen->screen.get_param = r300_get_param;
+    r300screen->screen.get_paramf = r300_get_paramf;
+    r300screen->screen.is_format_supported = r300_is_format_supported;
+    r300screen->screen.surface_map = r300_surface_map;
+    r300screen->screen.surface_unmap = r300_surface_unmap;
+
+    r300_init_screen_texture_functions(&r300screen->screen);
+    u_simple_screen_init(&r300screen->screen);
+
+    return &r300screen->screen;
+}
diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h
new file mode 100644
index 0000000000..2e25f61dbf
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_screen.h
@@ -0,0 +1,52 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_SCREEN_H
+#define R300_SCREEN_H
+
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "util/u_memory.h"
+#include "util/u_simple_screen.h"
+
+#include "r300_chipset.h"
+#include "r300_texture.h"
+#include "r300_winsys.h"
+
+struct r300_screen {
+    /* Parent class */
+    struct pipe_screen screen;
+
+    /* Chipset capabilities */
+    struct r300_capabilities* caps;
+};
+
+/* Convenience cast wrapper. */
+static struct r300_screen* r300_screen(struct pipe_screen* screen) {
+    return (struct r300_screen*)screen;
+}
+
+/* Creates a new r300 screen. */
+struct pipe_screen* r300_create_screen(struct pipe_winsys* winsys,
+                                       struct r300_winsys* r300_winsys);
+
+#endif /* R300_SCREEN_H */
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
new file mode 100644
index 0000000000..9392d72342
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -0,0 +1,826 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "util/u_math.h"
+#include "util/u_pack_color.h"
+#include "pipe/p_debug.h"
+
+#include "r300_context.h"
+#include "r300_reg.h"
+
+/* r300_state: Functions used to intialize state context by translating
+ * Gallium state objects into semi-native r300 state objects.
+ *
+ * XXX break this file up into pieces if it gets too big! */
+
+/* Pack a float into a dword. */
+static uint32_t pack_float_32(float f)
+{
+    union {
+        float f;
+        uint32_t u;
+    } u;
+
+    u.f = f;
+    return u.u;
+}
+
+static uint32_t translate_blend_function(int blend_func) {
+    switch (blend_func) {
+        case PIPE_BLEND_ADD:
+            return R300_COMB_FCN_ADD_CLAMP;
+        case PIPE_BLEND_SUBTRACT:
+            return R300_COMB_FCN_SUB_CLAMP;
+        case PIPE_BLEND_REVERSE_SUBTRACT:
+            return R300_COMB_FCN_RSUB_CLAMP;
+        case PIPE_BLEND_MIN:
+            return R300_COMB_FCN_MIN;
+        case PIPE_BLEND_MAX:
+            return R300_COMB_FCN_MAX;
+        default:
+            debug_printf("r300: Unknown blend function %d\n", blend_func);
+            break;
+    }
+    return 0;
+}
+
+/* XXX we can also offer the D3D versions of some of these... */
+static uint32_t translate_blend_factor(int blend_fact) {
+    switch (blend_fact) {
+        case PIPE_BLENDFACTOR_ONE:
+            return R300_BLEND_GL_ONE;
+        case PIPE_BLENDFACTOR_SRC_COLOR:
+            return R300_BLEND_GL_SRC_COLOR;
+        case PIPE_BLENDFACTOR_SRC_ALPHA:
+            return R300_BLEND_GL_SRC_ALPHA;
+        case PIPE_BLENDFACTOR_DST_ALPHA:
+            return R300_BLEND_GL_DST_ALPHA;
+        case PIPE_BLENDFACTOR_DST_COLOR:
+            return R300_BLEND_GL_DST_COLOR;
+        case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+            return R300_BLEND_GL_SRC_ALPHA_SATURATE;
+        case PIPE_BLENDFACTOR_CONST_COLOR:
+            return R300_BLEND_GL_CONST_COLOR;
+        case PIPE_BLENDFACTOR_CONST_ALPHA:
+            return R300_BLEND_GL_CONST_ALPHA;
+        /* XXX WTF are these?
+        case PIPE_BLENDFACTOR_SRC1_COLOR:
+        case PIPE_BLENDFACTOR_SRC1_ALPHA: */
+        case PIPE_BLENDFACTOR_ZERO:
+            return R300_BLEND_GL_ZERO;
+        case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+            return R300_BLEND_GL_ONE_MINUS_SRC_COLOR;
+        case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+            return R300_BLEND_GL_ONE_MINUS_SRC_ALPHA;
+        case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+            return R300_BLEND_GL_ONE_MINUS_DST_ALPHA;
+        case PIPE_BLENDFACTOR_INV_DST_COLOR:
+            return R300_BLEND_GL_ONE_MINUS_DST_COLOR;
+        case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+            return R300_BLEND_GL_ONE_MINUS_CONST_COLOR;
+        case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+            return R300_BLEND_GL_ONE_MINUS_CONST_ALPHA;
+        /* XXX see above
+        case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+        case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: */
+        default:
+            debug_printf("r300: Unknown blend factor %d\n", blend_fact);
+            break;
+    }
+    return 0;
+}
+
+/* Create a new blend state based on the CSO blend state.
+ *
+ * This encompasses alpha blending, logic/raster ops, and blend dithering. */
+static void* r300_create_blend_state(struct pipe_context* pipe,
+                                     const struct pipe_blend_state* state)
+{
+    struct r300_blend_state* blend = CALLOC_STRUCT(r300_blend_state);
+
+    if (state->blend_enable) {
+        /* XXX for now, always do separate alpha...
+         * is it faster to do it with one reg? */
+        blend->blend_control = R300_ALPHA_BLEND_ENABLE |
+                R300_SEPARATE_ALPHA_ENABLE |
+                R300_READ_ENABLE |
+                translate_blend_function(state->rgb_func) |
+                (translate_blend_factor(state->rgb_src_factor) <<
+                    R300_SRC_BLEND_SHIFT) |
+                (translate_blend_factor(state->rgb_dst_factor) <<
+                    R300_DST_BLEND_SHIFT);
+        blend->alpha_blend_control =
+                translate_blend_function(state->alpha_func) |
+                (translate_blend_factor(state->alpha_src_factor) <<
+                    R300_SRC_BLEND_SHIFT) |
+                (translate_blend_factor(state->alpha_dst_factor) <<
+                    R300_DST_BLEND_SHIFT);
+    }
+
+    /* PIPE_LOGICOP_* don't need to be translated, fortunately. */
+    /* XXX are logicops still allowed if blending's disabled?
+     * Does Gallium take care of it for us? */
+    if (state->logicop_enable) {
+        blend->rop = R300_RB3D_ROPCNTL_ROP_ENABLE |
+                (state->logicop_func) << R300_RB3D_ROPCNTL_ROP_SHIFT;
+    }
+
+    if (state->dither) {
+        blend->dither = R300_RB3D_DITHER_CTL_DITHER_MODE_LUT |
+                R300_RB3D_DITHER_CTL_ALPHA_DITHER_MODE_LUT;
+    }
+
+    return (void*)blend;
+}
+
+/* Bind blend state. */
+static void r300_bind_blend_state(struct pipe_context* pipe,
+                                  void* state)
+{
+    struct r300_context* r300 = r300_context(pipe);
+
+    r300->blend_state = (struct r300_blend_state*)state;
+    r300->dirty_state |= R300_NEW_BLEND;
+}
+
+/* Free blend state. */
+static void r300_delete_blend_state(struct pipe_context* pipe,
+                                    void* state)
+{
+    FREE(state);
+}
+
+/* Set blend color.
+ * Setup both R300 and R500 registers, figure out later which one to write. */
+static void r300_set_blend_color(struct pipe_context* pipe,
+                                 const struct pipe_blend_color* color)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    uint32_t r, g, b, a;
+    ubyte ur, ug, ub, ua;
+
+    r = util_iround(color->color[0] * 1023.0f);
+    g = util_iround(color->color[1] * 1023.0f);
+    b = util_iround(color->color[2] * 1023.0f);
+    a = util_iround(color->color[3] * 1023.0f);
+
+    ur = float_to_ubyte(color->color[0]);
+    ug = float_to_ubyte(color->color[1]);
+    ub = float_to_ubyte(color->color[2]);
+    ua = float_to_ubyte(color->color[3]);
+
+    r300->blend_color_state->blend_color = (a << 24) | (r << 16) | (g << 8) | b;
+
+    r300->blend_color_state->blend_color_red_alpha = ur | (ua << 16);
+    r300->blend_color_state->blend_color_green_blue = ub | (ug << 16);
+
+    r300->dirty_state |= R300_NEW_BLEND_COLOR;
+}
+
+static void r300_set_clip_state(struct pipe_context* pipe,
+                                const struct pipe_clip_state* state)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    /* XXX Draw */
+    draw_flush(r300->draw);
+    draw_set_clip_state(r300->draw, state);
+}
+
+static void
+    r300_set_constant_buffer(struct pipe_context* pipe,
+                             uint shader, uint index,
+                             const struct pipe_constant_buffer* buffer)
+{
+    /* XXX */
+}
+
+static uint32_t translate_depth_stencil_function(int zs_func) {
+    switch (zs_func) {
+        case PIPE_FUNC_NEVER:
+            return R300_ZS_NEVER;
+        case PIPE_FUNC_LESS:
+            return R300_ZS_LESS;
+        case PIPE_FUNC_EQUAL:
+            return R300_ZS_EQUAL;
+        case PIPE_FUNC_LEQUAL:
+            return R300_ZS_LEQUAL;
+        case PIPE_FUNC_GREATER:
+            return R300_ZS_GREATER;
+        case PIPE_FUNC_NOTEQUAL:
+            return R300_ZS_NOTEQUAL;
+        case PIPE_FUNC_GEQUAL:
+            return R300_ZS_GEQUAL;
+        case PIPE_FUNC_ALWAYS:
+            return R300_ZS_ALWAYS;
+        default:
+            debug_printf("r300: Unknown depth/stencil function %d\n",
+                zs_func);
+            break;
+    }
+    return 0;
+}
+
+static uint32_t translate_stencil_op(int s_op) {
+    switch (s_op) {
+        case PIPE_STENCIL_OP_KEEP:
+            return R300_ZS_KEEP;
+        case PIPE_STENCIL_OP_ZERO:
+            return R300_ZS_ZERO;
+        case PIPE_STENCIL_OP_REPLACE:
+            return R300_ZS_REPLACE;
+        case PIPE_STENCIL_OP_INCR:
+            return R300_ZS_INCR;
+        case PIPE_STENCIL_OP_DECR:
+            return R300_ZS_DECR;
+        case PIPE_STENCIL_OP_INCR_WRAP:
+            return R300_ZS_INCR_WRAP;
+        case PIPE_STENCIL_OP_DECR_WRAP:
+            return R300_ZS_DECR_WRAP;
+        case PIPE_STENCIL_OP_INVERT:
+            return R300_ZS_INVERT;
+        default:
+            debug_printf("r300: Unknown stencil op %d", s_op);
+            break;
+    }
+    return 0;
+}
+
+static uint32_t translate_alpha_function(int alpha_func) {
+    switch (alpha_func) {
+        case PIPE_FUNC_NEVER:
+            return R300_FG_ALPHA_FUNC_NEVER;
+        case PIPE_FUNC_LESS:
+            return R300_FG_ALPHA_FUNC_LESS;
+        case PIPE_FUNC_EQUAL:
+            return R300_FG_ALPHA_FUNC_EQUAL;
+        case PIPE_FUNC_LEQUAL:
+            return R300_FG_ALPHA_FUNC_LE;
+        case PIPE_FUNC_GREATER:
+            return R300_FG_ALPHA_FUNC_GREATER;
+        case PIPE_FUNC_NOTEQUAL:
+            return R300_FG_ALPHA_FUNC_NOTEQUAL;
+        case PIPE_FUNC_GEQUAL:
+            return R300_FG_ALPHA_FUNC_GE;
+        case PIPE_FUNC_ALWAYS:
+            return R300_FG_ALPHA_FUNC_ALWAYS;
+        default:
+            debug_printf("r300: Unknown alpha function %d", alpha_func);
+            break;
+    }
+    return 0;
+}
+
+/* Create a new depth, stencil, and alpha state based on the CSO dsa state.
+ *
+ * This contains the depth buffer, stencil buffer, alpha test, and such.
+ * On the Radeon, depth and stencil buffer setup are intertwined, which is
+ * the reason for some of the strange-looking assignments across registers. */
+static void*
+        r300_create_dsa_state(struct pipe_context* pipe,
+                              const struct pipe_depth_stencil_alpha_state* state)
+{
+    struct r300_dsa_state* dsa = CALLOC_STRUCT(r300_dsa_state);
+
+    /* Depth test setup. */
+    if (state->depth.enabled) {
+        dsa->z_buffer_control |= R300_Z_ENABLE;
+
+        if (state->depth.writemask) {
+            dsa->z_buffer_control |= R300_Z_WRITE_ENABLE;
+        }
+
+        dsa->z_stencil_control |=
+            (translate_depth_stencil_function(state->depth.func) <<
+                R300_Z_FUNC_SHIFT);
+    }
+
+    /* Stencil buffer setup. */
+    if (state->stencil[0].enabled) {
+        dsa->z_buffer_control |= R300_STENCIL_ENABLE;
+        dsa->z_stencil_control |=
+                (translate_depth_stencil_function(state->stencil[0].func) <<
+                    R300_S_FRONT_FUNC_SHIFT) |
+                (translate_stencil_op(state->stencil[0].fail_op) <<
+                    R300_S_FRONT_SFAIL_OP_SHIFT) |
+                (translate_stencil_op(state->stencil[0].zpass_op) <<
+                    R300_S_FRONT_ZPASS_OP_SHIFT) |
+                (translate_stencil_op(state->stencil[0].zfail_op) <<
+                    R300_S_FRONT_ZFAIL_OP_SHIFT);
+
+        dsa->stencil_ref_mask = (state->stencil[0].ref_value) |
+                (state->stencil[0].valuemask << R300_STENCILMASK_SHIFT) |
+                (state->stencil[0].writemask << R300_STENCILWRITEMASK_SHIFT);
+
+        if (state->stencil[1].enabled) {
+            dsa->z_buffer_control |= R300_STENCIL_FRONT_BACK;
+            dsa->z_stencil_control |=
+                (translate_depth_stencil_function(state->stencil[1].func) <<
+                    R300_S_BACK_FUNC_SHIFT) |
+                (translate_stencil_op(state->stencil[1].fail_op) <<
+                    R300_S_BACK_SFAIL_OP_SHIFT) |
+                (translate_stencil_op(state->stencil[1].zpass_op) <<
+                    R300_S_BACK_ZPASS_OP_SHIFT) |
+                (translate_stencil_op(state->stencil[1].zfail_op) <<
+                    R300_S_BACK_ZFAIL_OP_SHIFT);
+
+            dsa->stencil_ref_bf = (state->stencil[1].ref_value) |
+                (state->stencil[1].valuemask << R300_STENCILMASK_SHIFT) |
+                (state->stencil[1].writemask << R300_STENCILWRITEMASK_SHIFT);
+        }
+    }
+
+    /* Alpha test setup. */
+    if (state->alpha.enabled) {
+        dsa->alpha_function = translate_alpha_function(state->alpha.func) |
+            R300_FG_ALPHA_FUNC_ENABLE;
+        dsa->alpha_reference = CLAMP(state->alpha.ref_value * 1023.0f,
+                                     0, 1023);
+    } else {
+        dsa->z_buffer_top = R300_ZTOP_ENABLE;
+    }
+
+    return (void*)dsa;
+}
+
+/* Bind DSA state. */
+static void r300_bind_dsa_state(struct pipe_context* pipe,
+                                void* state)
+{
+    struct r300_context* r300 = r300_context(pipe);
+
+    r300->dsa_state = (struct r300_dsa_state*)state;
+    r300->dirty_state |= R300_NEW_DSA;
+}
+
+/* Free DSA state. */
+static void r300_delete_dsa_state(struct pipe_context* pipe,
+                                  void* state)
+{
+    FREE(state);
+}
+
+static void r300_set_edgeflags(struct pipe_context* pipe,
+                               const unsigned* bitfield)
+{
+    /* XXX you know it's bad when i915 has this blank too */
+}
+
+static void
+    r300_set_framebuffer_state(struct pipe_context* pipe,
+                               const struct pipe_framebuffer_state* state)
+{
+    struct r300_context* r300 = r300_context(pipe);
+
+    draw_flush(r300->draw);
+
+    r300->framebuffer_state = *state;
+
+    r300->dirty_state |= R300_NEW_FRAMEBUFFERS;
+}
+
+/* Create fragment shader state. */
+static void* r300_create_fs_state(struct pipe_context* pipe,
+                                  const struct pipe_shader_state* shader)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    struct r3xx_fragment_shader* fs = NULL;
+
+    if (r300_screen(r300->context.screen)->caps->is_r500) {
+        fs =
+            (struct r3xx_fragment_shader*)CALLOC_STRUCT(r500_fragment_shader);
+    } else {
+        fs =
+            (struct r3xx_fragment_shader*)CALLOC_STRUCT(r300_fragment_shader);
+    }
+
+    /* Copy state directly into shader. */
+    fs->state = *shader;
+
+    return (void*)fs;
+}
+
+/* Bind fragment shader state. */
+static void r300_bind_fs_state(struct pipe_context* pipe, void* shader)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    struct r3xx_fragment_shader* fs = (struct r3xx_fragment_shader*)shader;
+
+    if (!fs->translated) {
+        if (r300_screen(r300->context.screen)->caps->is_r500) {
+            r500_translate_shader(r300, fs);
+        } else {
+            r300_translate_shader(r300, fs);
+        }
+    }
+
+    r300->fs = fs;
+
+    r300->dirty_state |= R300_NEW_FRAGMENT_SHADER;
+}
+
+/* Delete fragment shader state. */
+static void r300_delete_fs_state(struct pipe_context* pipe, void* shader)
+{
+    FREE(shader);
+}
+
+static void r300_set_polygon_stipple(struct pipe_context* pipe,
+                                     const struct pipe_poly_stipple* state)
+{
+    /* XXX */
+}
+
+static INLINE int pack_float_16_6x(float f) {
+    return ((int)(f * 6.0) & 0xffff);
+}
+
+/* Create a new rasterizer state based on the CSO rasterizer state.
+ *
+ * This is a very large chunk of state, and covers most of the graphics
+ * backend (GB), geometry assembly (GA), and setup unit (SU) blocks.
+ *
+ * In a not entirely unironic sidenote, this state has nearly nothing to do
+ * with the actual block on the Radeon called the rasterizer (RS). */
+static void* r300_create_rs_state(struct pipe_context* pipe,
+                                  const struct pipe_rasterizer_state* state)
+{
+    struct r300_rs_state* rs = CALLOC_STRUCT(r300_rs_state);
+
+    /* XXX this is part of HW TCL */
+    /* XXX endian control */
+    rs->vap_control_status = R300_VAP_TCL_BYPASS;
+
+    rs->point_size = pack_float_16_6x(state->point_size) |
+        (pack_float_16_6x(state->point_size) << R300_POINTSIZE_X_SHIFT);
+
+    rs->line_control = pack_float_16_6x(state->line_width) |
+        R300_GA_LINE_CNTL_END_TYPE_COMP;
+
+    /* Radeons don't think in "CW/CCW", they think in "front/back". */
+    if (state->front_winding == PIPE_WINDING_CW) {
+        rs->cull_mode = R300_FRONT_FACE_CW;
+
+        if (state->offset_cw) {
+            rs->polygon_offset_enable |= R300_FRONT_ENABLE;
+        }
+        if (state->offset_ccw) {
+            rs->polygon_offset_enable |= R300_BACK_ENABLE;
+        }
+    } else {
+        rs->cull_mode = R300_FRONT_FACE_CCW;
+
+        if (state->offset_ccw) {
+            rs->polygon_offset_enable |= R300_FRONT_ENABLE;
+        }
+        if (state->offset_cw) {
+            rs->polygon_offset_enable |= R300_BACK_ENABLE;
+        }
+    }
+    if (state->front_winding & state->cull_mode) {
+        rs->cull_mode |= R300_CULL_FRONT;
+    }
+    if (~(state->front_winding) & state->cull_mode) {
+        rs->cull_mode |= R300_CULL_BACK;
+    }
+
+    if (rs->polygon_offset_enable) {
+        rs->depth_offset_front = rs->depth_offset_back =
+                pack_float_32(state->offset_units);
+        rs->depth_scale_front = rs->depth_scale_back =
+                pack_float_32(state->offset_scale);
+    }
+
+    if (state->line_stipple_enable) {
+        rs->line_stipple_config =
+            R300_GA_LINE_STIPPLE_CONFIG_LINE_RESET_LINE |
+            (pack_float_32((float)state->line_stipple_factor) &
+                R300_GA_LINE_STIPPLE_CONFIG_STIPPLE_SCALE_MASK);
+        /* XXX this might need to be scaled up */
+        rs->line_stipple_value = state->line_stipple_pattern;
+    }
+
+    return (void*)rs;
+}
+
+/* Bind rasterizer state. */
+static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
+{
+    struct r300_context* r300 = r300_context(pipe);
+
+    r300->rs_state = (struct r300_rs_state*)state;
+    r300->dirty_state |= R300_NEW_RASTERIZER;
+}
+
+/* Free rasterizer state. */
+static void r300_delete_rs_state(struct pipe_context* pipe, void* state)
+{
+    FREE(state);
+}
+
+static uint32_t translate_wrap(int wrap) {
+    switch (wrap) {
+        case PIPE_TEX_WRAP_REPEAT:
+            return R300_TX_REPEAT;
+        case PIPE_TEX_WRAP_CLAMP:
+            return R300_TX_CLAMP;
+        case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+            return R300_TX_CLAMP_TO_EDGE;
+        case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+            return R300_TX_CLAMP_TO_BORDER;
+        case PIPE_TEX_WRAP_MIRROR_REPEAT:
+            return R300_TX_REPEAT | R300_TX_MIRRORED;
+        case PIPE_TEX_WRAP_MIRROR_CLAMP:
+            return R300_TX_CLAMP | R300_TX_MIRRORED;
+        case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+            return R300_TX_CLAMP_TO_EDGE | R300_TX_MIRRORED;
+        case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+            return R300_TX_CLAMP_TO_EDGE | R300_TX_MIRRORED;
+        default:
+            debug_printf("r300: Unknown texture wrap %d", wrap);
+            return 0;
+    }
+}
+
+static uint32_t translate_tex_filters(int min, int mag, int mip) {
+    uint32_t retval = 0;
+    switch (min) {
+        case PIPE_TEX_FILTER_NEAREST:
+            retval |= R300_TX_MIN_FILTER_NEAREST;
+        case PIPE_TEX_FILTER_LINEAR:
+            retval |= R300_TX_MIN_FILTER_LINEAR;
+        case PIPE_TEX_FILTER_ANISO:
+            retval |= R300_TX_MIN_FILTER_ANISO;
+        default:
+            debug_printf("r300: Unknown texture filter %d", min);
+            break;
+    }
+    switch (mag) {
+        case PIPE_TEX_FILTER_NEAREST:
+            retval |= R300_TX_MAG_FILTER_NEAREST;
+        case PIPE_TEX_FILTER_LINEAR:
+            retval |= R300_TX_MAG_FILTER_LINEAR;
+        case PIPE_TEX_FILTER_ANISO:
+            retval |= R300_TX_MAG_FILTER_ANISO;
+        default:
+            debug_printf("r300: Unknown texture filter %d", mag);
+            break;
+    }
+    switch (mip) {
+        case PIPE_TEX_MIPFILTER_NONE:
+            retval |= R300_TX_MIN_FILTER_MIP_NONE;
+        case PIPE_TEX_MIPFILTER_NEAREST:
+            retval |= R300_TX_MIN_FILTER_MIP_NEAREST;
+        case PIPE_TEX_MIPFILTER_LINEAR:
+            retval |= R300_TX_MIN_FILTER_MIP_LINEAR;
+        default:
+            debug_printf("r300: Unknown texture filter %d", mip);
+            break;
+    }
+
+    return retval;
+}
+
+static uint32_t anisotropy(float max_aniso) {
+    if (max_aniso >= 16.0f) {
+        return R300_TX_MAX_ANISO_16_TO_1;
+    } else if (max_aniso >= 8.0f) {
+        return R300_TX_MAX_ANISO_8_TO_1;
+    } else if (max_aniso >= 4.0f) {
+        return R300_TX_MAX_ANISO_4_TO_1;
+    } else if (max_aniso >= 2.0f) {
+        return R300_TX_MAX_ANISO_2_TO_1;
+    } else {
+        return R300_TX_MAX_ANISO_1_TO_1;
+    }
+}
+
+static void*
+        r300_create_sampler_state(struct pipe_context* pipe,
+                                  const struct pipe_sampler_state* state)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    struct r300_sampler_state* sampler = CALLOC_STRUCT(r300_sampler_state);
+    int lod_bias;
+
+    sampler->filter0 |=
+        (translate_wrap(state->wrap_s) << R300_TX_WRAP_S_SHIFT) |
+        (translate_wrap(state->wrap_t) << R300_TX_WRAP_T_SHIFT) |
+        (translate_wrap(state->wrap_r) << R300_TX_WRAP_R_SHIFT);
+
+    sampler->filter0 |= translate_tex_filters(state->min_img_filter,
+                                              state->mag_img_filter,
+                                              state->min_mip_filter);
+
+    lod_bias = CLAMP((int)(state->lod_bias * 32), -(1 << 9), (1 << 9) - 1);
+
+    sampler->filter1 |= lod_bias << R300_LOD_BIAS_SHIFT;
+
+    sampler->filter1 |= anisotropy(state->max_anisotropy);
+
+    util_pack_color(state->border_color, PIPE_FORMAT_A8R8G8B8_UNORM,
+                    &sampler->border_color);
+
+    /* R500-specific fixups and optimizations */
+    if (r300_screen(r300->context.screen)->caps->is_r500) {
+        sampler->filter1 |= R500_BORDER_FIX;
+    }
+
+    return (void*)sampler;
+}
+
+static void r300_bind_sampler_states(struct pipe_context* pipe,
+                                     unsigned count,
+                                     void** states)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    int i;
+
+    if (count > 8) {
+        return;
+    }
+
+    for (i = 0; i < count; i++) {
+        if (r300->sampler_states[i] != states[i]) {
+            r300->sampler_states[i] = (struct r300_sampler_state*)states[i];
+            r300->dirty_state |= (R300_NEW_SAMPLER << i);
+        }
+    }
+
+    r300->sampler_count = count;
+}
+
+static void r300_delete_sampler_state(struct pipe_context* pipe, void* state)
+{
+    FREE(state);
+}
+
+static void r300_set_sampler_textures(struct pipe_context* pipe,
+                                      unsigned count,
+                                      struct pipe_texture** texture)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    int i;
+
+    /* XXX magic num */
+    if (count > 8) {
+        return;
+    }
+
+    for (i = 0; i < count; i++) {
+        if (r300->textures[i] != (struct r300_texture*)texture[i]) {
+            pipe_texture_reference((struct pipe_texture**)&r300->textures[i],
+                texture[i]);
+            r300->dirty_state |= (R300_NEW_TEXTURE << i);
+        }
+    }
+
+    for (i = count; i < 8; i++) {
+        if (r300->textures[i]) {
+            pipe_texture_reference((struct pipe_texture**)&r300->textures[i],
+                NULL);
+            r300->dirty_state |= (R300_NEW_TEXTURE << i);
+        }
+    }
+
+    r300->texture_count = count;
+}
+
+static void r300_set_scissor_state(struct pipe_context* pipe,
+                                   const struct pipe_scissor_state* state)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    draw_flush(r300->draw);
+
+    uint32_t left, top, right, bottom;
+
+    /* So, a bit of info. The scissors are offset by R300_SCISSORS_OFFSET in
+     * both directions for all values, and can only be 13 bits wide. Why?
+     * We may never know. */
+    left = (state->minx + R300_SCISSORS_OFFSET) & 0x1fff;
+    top = (state->miny + R300_SCISSORS_OFFSET) & 0x1fff;
+    right = (state->maxx + R300_SCISSORS_OFFSET) & 0x1fff;
+    bottom = (state->maxy + R300_SCISSORS_OFFSET) & 0x1fff;
+
+    r300->scissor_state->scissor_top_left = (left << R300_SCISSORS_X_SHIFT) |
+            (top << R300_SCISSORS_Y_SHIFT);
+    r300->scissor_state->scissor_bottom_right =
+        (right << R300_SCISSORS_X_SHIFT) | (bottom << R300_SCISSORS_Y_SHIFT);
+
+    r300->dirty_state |= R300_NEW_SCISSOR;
+}
+
+static void r300_set_viewport_state(struct pipe_context* pipe,
+                                    const struct pipe_viewport_state* state)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    /* XXX handing this off to Draw for now */
+    draw_set_viewport_state(r300->draw, state);
+}
+
+static void r300_set_vertex_buffers(struct pipe_context* pipe,
+                                    unsigned count,
+                                    const struct pipe_vertex_buffer* buffers)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    /* XXX Draw */
+    draw_flush(r300->draw);
+    draw_set_vertex_buffers(r300->draw, count, buffers);
+}
+
+static void r300_set_vertex_elements(struct pipe_context* pipe,
+                                    unsigned count,
+                                    const struct pipe_vertex_element* elements)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    /* XXX Draw */
+    draw_flush(r300->draw);
+    draw_set_vertex_elements(r300->draw, count, elements);
+}
+
+static void* r300_create_vs_state(struct pipe_context* pipe,
+                                  const struct pipe_shader_state* state)
+{
+    struct r300_context* context = r300_context(pipe);
+    /* XXX handing this off to Draw for now */
+    return draw_create_vertex_shader(context->draw, state);
+}
+
+static void r300_bind_vs_state(struct pipe_context* pipe, void* state) {
+    struct r300_context* context = r300_context(pipe);
+    /* XXX handing this off to Draw for now */
+    draw_bind_vertex_shader(context->draw, (struct draw_vertex_shader*)state);
+}
+
+static void r300_delete_vs_state(struct pipe_context* pipe, void* state)
+{
+    struct r300_context* context = r300_context(pipe);
+    /* XXX handing this off to Draw for now */
+    draw_delete_vertex_shader(context->draw, (struct draw_vertex_shader*)state);
+}
+
+void r300_init_state_functions(struct r300_context* r300)
+{
+    r300->context.create_blend_state = r300_create_blend_state;
+    r300->context.bind_blend_state = r300_bind_blend_state;
+    r300->context.delete_blend_state = r300_delete_blend_state;
+
+    r300->context.set_blend_color = r300_set_blend_color;
+
+    r300->context.set_clip_state = r300_set_clip_state;
+
+    r300->context.set_constant_buffer = r300_set_constant_buffer;
+
+    r300->context.create_depth_stencil_alpha_state = r300_create_dsa_state;
+    r300->context.bind_depth_stencil_alpha_state = r300_bind_dsa_state;
+    r300->context.delete_depth_stencil_alpha_state = r300_delete_dsa_state;
+
+    r300->context.set_edgeflags = r300_set_edgeflags;
+
+    r300->context.set_framebuffer_state = r300_set_framebuffer_state;
+
+    r300->context.create_fs_state = r300_create_fs_state;
+    r300->context.bind_fs_state = r300_bind_fs_state;
+    r300->context.delete_fs_state = r300_delete_fs_state;
+
+    r300->context.set_polygon_stipple = r300_set_polygon_stipple;
+
+    r300->context.create_rasterizer_state = r300_create_rs_state;
+    r300->context.bind_rasterizer_state = r300_bind_rs_state;
+    r300->context.delete_rasterizer_state = r300_delete_rs_state;
+
+    r300->context.create_sampler_state = r300_create_sampler_state;
+    r300->context.bind_sampler_states = r300_bind_sampler_states;
+    r300->context.delete_sampler_state = r300_delete_sampler_state;
+
+    r300->context.set_sampler_textures = r300_set_sampler_textures;
+
+    r300->context.set_scissor_state = r300_set_scissor_state;
+
+    r300->context.set_viewport_state = r300_set_viewport_state;
+
+    r300->context.set_vertex_buffers = r300_set_vertex_buffers;
+    r300->context.set_vertex_elements = r300_set_vertex_elements;
+
+    r300->context.create_vs_state = r300_create_vs_state;
+    r300->context.bind_vs_state = r300_bind_vs_state;
+    r300->context.delete_vs_state = r300_delete_vs_state;
+}
diff --git a/src/gallium/drivers/r300/r300_state_shader.c b/src/gallium/drivers/r300/r300_state_shader.c
new file mode 100644
index 0000000000..e87172128f
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_state_shader.c
@@ -0,0 +1,33 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_state_shader.h"
+
+void r300_translate_shader(struct r300_context* r300,
+                           struct r300_fragment_shader* fs)
+{
+}
+
+void r500_translate_shader(struct r300_context* r300,
+                           struct r500_fragment_shader* fs)
+{
+}
diff --git a/src/gallium/drivers/r300/r300_state_shader.h b/src/gallium/drivers/r300/r300_state_shader.h
new file mode 100644
index 0000000000..a20bd4276c
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_state_shader.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_STATE_SHADER_H
+#define R300_STATE_SHADER_H
+
+#include "r300_context.h"
+#include "r300_screen.h"
+
+void r300_translate_shader(struct r300_context* r300,
+                           struct r300_fragment_shader* fs);
+
+void r500_translate_shader(struct r300_context* r300,
+                           struct r500_fragment_shader* fs);
+
+#endif /* R300_STATE_SHADER_H */
diff --git a/src/gallium/drivers/r300/r300_surface.c b/src/gallium/drivers/r300/r300_surface.c
new file mode 100644
index 0000000000..1e1f96a7f9
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_surface.c
@@ -0,0 +1,352 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_surface.h"
+
+/* Provides pipe_context's "surface_fill". Commonly used for clearing
+ * buffers. */
+static void r300_surface_fill(struct pipe_context* pipe,
+                              struct pipe_surface* dest,
+                              unsigned x, unsigned y,
+                              unsigned w, unsigned h,
+                              unsigned color)
+{
+    struct r300_context* r300 = r300_context(pipe);
+    CS_LOCALS(r300);
+    struct r300_capabilities* caps = ((struct r300_screen*)pipe->screen)->caps;
+    struct r300_texture* tex = (struct r300_texture*)dest->texture;
+    int i;
+    float r, g, b, a;
+    r = (float)((color >> 16) & 0xff) / 255.0f;
+    g = (float)((color >>  8) & 0xff) / 255.0f;
+    b = (float)((color >>  0) & 0xff) / 255.0f;
+    debug_printf("r300: Filling surface %p at (%d,%d),"
+        " dimensions %dx%d (stride %d), color 0x%x\n",
+        dest, x, y, w, h, dest->stride, color);
+
+    /* Fallback? */
+    if (0) {
+        debug_printf("r300: Falling back on surface clear...");
+        void* map = pipe->screen->surface_map(pipe->screen, dest,
+            PIPE_BUFFER_USAGE_CPU_WRITE);
+        pipe_fill_rect(map, &dest->block, &dest->stride, x, y, w, h, color);
+        pipe->screen->surface_unmap(pipe->screen, dest);
+        return;
+    }
+
+BEGIN_CS((caps->is_r500) ? 309 : 280);
+R300_PACIFY;
+OUT_CS_REG(R300_TX_INVALTAGS, 0x0);
+R300_PACIFY;
+/* Flush PVS. */
+OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0);
+
+OUT_CS_REG(R300_SE_VTE_CNTL, R300_VPORT_X_SCALE_ENA |
+    R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA |
+    R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA |
+    R300_VPORT_Z_OFFSET_ENA | R300_VTX_W0_FMT);
+/* Vertex size. */
+OUT_CS_REG(R300_VAP_VTX_SIZE, 0x8);
+/* Max and min vertex index clamp. */
+OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, 0xFFFFFF);
+OUT_CS_REG(R300_VAP_VF_MIN_VTX_INDX, 0x0);
+/* XXX endian */
+OUT_CS_REG(R300_VAP_CNTL_STATUS, R300_VC_NO_SWAP);
+OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0, 0x0);
+/* XXX magic number not in r300_reg */
+OUT_CS_REG(R300_VAP_PSC_SGN_NORM_CNTL, 0xAAAAAAAA);
+OUT_CS_REG(R300_VAP_CLIP_CNTL, 0x0);
+OUT_CS_REG_SEQ(R300_VAP_GB_VERT_CLIP_ADJ, 4);
+OUT_CS_32F(1.0);
+OUT_CS_32F(1.0);
+OUT_CS_32F(1.0);
+OUT_CS_32F(1.0);
+/* XXX is this too long? */
+OUT_CS_REG(VAP_PVS_VTX_TIMEOUT_REG, 0xFFFF);
+OUT_CS_REG(R300_GB_ENABLE, R300_GB_POINT_STUFF_ENABLE |
+    R300_GB_LINE_STUFF_ENABLE | R300_GB_TRIANGLE_STUFF_ENABLE);
+/* XXX more magic numbers */
+OUT_CS_REG(R300_GB_MSPOS0, 0x66666666);
+OUT_CS_REG(R300_GB_MSPOS1, 0x66666666);
+/* XXX why doesn't classic Mesa write the number of pipes, too? */
+OUT_CS_REG(R300_GB_TILE_CONFIG, R300_GB_TILE_ENABLE | R300_GB_TILE_SIZE_16);
+OUT_CS_REG(R300_GB_SELECT, R300_GB_FOG_SELECT_1_1_W);
+OUT_CS_REG(R300_GB_AA_CONFIG, 0x0);
+/* XXX point tex stuffing */
+OUT_CS_REG_SEQ(R300_GA_POINT_S0, 1);
+OUT_CS_32F(0.0);
+OUT_CS_REG_SEQ(R300_GA_POINT_S1, 1);
+OUT_CS_32F(1.0);
+OUT_CS_REG(R300_GA_TRIANGLE_STIPPLE, 0x5 |
+    (0x5 << R300_GA_TRIANGLE_STIPPLE_Y_SHIFT_SHIFT));
+/* XXX should this be related to the actual point size? */
+OUT_CS_REG(R300_GA_POINT_MINMAX, 0x6 |
+    (0x1800 << R300_GA_POINT_MINMAX_MAX_SHIFT));
+/* XXX this big chunk should be refactored into rs_state */
+OUT_CS_REG(R300_GA_LINE_CNTL, 0x00030006);
+OUT_CS_REG(R300_GA_LINE_STIPPLE_CONFIG, 0x3BAAAAAB);
+OUT_CS_REG(R300_GA_LINE_STIPPLE_VALUE, 0x00000000);
+OUT_CS_REG(R300_GA_LINE_S0, 0x00000000);
+OUT_CS_REG(R300_GA_LINE_S1, 0x3F800000);
+OUT_CS_REG(R300_GA_ENHANCE, 0x00000002);
+OUT_CS_REG(R300_GA_COLOR_CONTROL, 0x0003AAAA);
+OUT_CS_REG(R300_GA_SOLID_RG, 0x00000000);
+OUT_CS_REG(R300_GA_SOLID_BA, 0x00000000);
+OUT_CS_REG(R300_GA_POLY_MODE, 0x00000000);
+OUT_CS_REG(R300_GA_ROUND_MODE, 0x00000001);
+OUT_CS_REG(R300_GA_OFFSET, 0x00000000);
+OUT_CS_REG(R300_GA_FOG_SCALE, 0x3DBF1412);
+OUT_CS_REG(R300_GA_FOG_OFFSET, 0x00000000);
+OUT_CS_REG(R300_SU_TEX_WRAP, 0x00000000);
+OUT_CS_REG(R300_SU_POLY_OFFSET_FRONT_SCALE, 0x00000000);
+OUT_CS_REG(R300_SU_POLY_OFFSET_FRONT_OFFSET, 0x00000000);
+OUT_CS_REG(R300_SU_POLY_OFFSET_BACK_SCALE, 0x00000000);
+OUT_CS_REG(R300_SU_POLY_OFFSET_BACK_OFFSET, 0x00000000);
+OUT_CS_REG(R300_SU_POLY_OFFSET_ENABLE, 0x00000000);
+OUT_CS_REG(R300_SU_CULL_MODE, 0x00000000);
+OUT_CS_REG(R300_SU_DEPTH_SCALE, 0x4B7FFFFF);
+OUT_CS_REG(R300_SU_DEPTH_OFFSET, 0x00000000);
+OUT_CS_REG(R300_SC_HYPERZ, 0x0000001C);
+OUT_CS_REG(R300_SC_EDGERULE, 0x2DA49525);
+OUT_CS_REG(R300_SC_SCREENDOOR, 0x00FFFFFF);
+OUT_CS_REG(R300_FG_FOG_BLEND, 0x00000002);
+OUT_CS_REG(R300_FG_FOG_COLOR_R, 0x00000000);
+OUT_CS_REG(R300_FG_FOG_COLOR_G, 0x00000000);
+OUT_CS_REG(R300_FG_FOG_COLOR_B, 0x00000000);
+OUT_CS_REG(R300_FG_DEPTH_SRC, 0x00000000);
+OUT_CS_REG(R300_FG_DEPTH_SRC, 0x00000000);
+OUT_CS_REG(R300_RB3D_CCTL, 0x00000000);
+OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F);
+
+/* XXX: Oh the wonderful unknown */
+OUT_CS_REG_SEQ(0x4E54, 8);
+for (i = 0; i < 8; i++)
+    OUT_CS(0x00000000);
+OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, 0x00000000);
+OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000);
+OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_GTE_THRESHOLD, 0xFFFFFFFF);
+OUT_CS_REG(R300_ZB_FORMAT, 0x00000002);
+OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, 0x00000003);
+OUT_CS_REG(R300_ZB_BW_CNTL, 0x00000000);
+OUT_CS_REG(R300_ZB_DEPTHCLEARVALUE, 0x00000000);
+OUT_CS_REG(0x4F30, 0x00000000);
+OUT_CS_REG(0x4F34, 0x00000000);
+OUT_CS_REG(R300_ZB_HIZ_OFFSET, 0x00000000);
+OUT_CS_REG(R300_ZB_HIZ_PITCH, 0x00000000);
+R300_PACIFY;
+if (caps->has_tcl) {
+    OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0,
+        (R300_DATA_TYPE_FLOAT_4 << R300_DATA_TYPE_0_SHIFT) |
+        ((R300_LAST_VEC | (1 << R300_DST_VEC_LOC_SHIFT) |
+            R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT));
+} else {
+    OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_0,
+        (R300_DATA_TYPE_FLOAT_4 << R300_DATA_TYPE_0_SHIFT) |
+        ((R300_LAST_VEC | (2 << R300_DST_VEC_LOC_SHIFT) |
+            R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT));
+}
+OUT_CS_REG(R300_FG_FOG_BLEND, 0x00000000);
+OUT_CS_REG(R300_VAP_PROG_STREAM_CNTL_EXT_0, 0xF688F688);
+OUT_CS_REG(R300_VAP_VTX_STATE_CNTL, 0x1);
+OUT_CS_REG(R300_VAP_VSM_VTX_ASSM, 0x405);
+OUT_CS_REG(R300_SE_VTE_CNTL, 0x0000043F);
+OUT_CS_REG(R300_VAP_VTX_SIZE, 0x00000008);
+OUT_CS_REG(R300_VAP_PSC_SGN_NORM_CNTL, 0xAAAAAAAA);
+OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_0, 0x00000003);
+OUT_CS_REG(R300_VAP_OUTPUT_VTX_FMT_1, 0x00000000);
+OUT_CS_REG(R300_TX_ENABLE, 0x0);
+/* XXX viewport setup */
+OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6);
+OUT_CS_32F(1.0);
+OUT_CS_32F((float)x);
+OUT_CS_32F(1.0);
+OUT_CS_32F((float)y);
+OUT_CS_32F(1.0);
+OUT_CS_32F(0.0);
+
+if (caps->has_tcl) {
+    OUT_CS_REG(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE |
+        R300_PS_UCP_MODE_CLIP_AS_TRIFAN);
+}
+
+OUT_CS_REG(R300_GA_POINT_SIZE, ((h * 6) & R300_POINTSIZE_Y_MASK) |
+    ((w * 6) << R300_POINTSIZE_X_SHIFT));
+
+/* XXX RS block and fp setup */
+if (caps->is_r500) {
+    OUT_CS_REG_SEQ(R500_RS_IP_0, 8);
+    for (i = 0; i < 8; i++) {
+        /* I like the operator macros more than the shift macros... */
+        OUT_CS((R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) |
+            (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
+            (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
+            (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT));
+    }
+    /* XXX */
+    OUT_CS_REG_SEQ(R300_RS_COUNT, 2);
+    OUT_CS((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
+    OUT_CS(0x0);
+    OUT_CS_REG(R500_RS_INST_0, R500_RS_INST_COL_CN_WRITE);
+
+    OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
+    OUT_CS_REG(R500_US_PIXSIZE, 0x00000000);
+    OUT_CS_REG(R500_US_CODE_ADDR, R500_US_CODE_START_ADDR(0) |
+        R500_US_CODE_END_ADDR(1));
+    OUT_CS_REG(R500_US_CODE_RANGE, R500_US_CODE_RANGE_ADDR(0) |
+        R500_US_CODE_RANGE_SIZE(1));
+    OUT_CS_REG(R500_US_CODE_OFFSET, R500_US_CODE_OFFSET_ADDR(0));
+    R300_PACIFY;
+    OUT_CS_REG(R500_GA_US_VECTOR_INDEX,
+        0 | R500_GA_US_VECTOR_INDEX_TYPE_INSTR);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_INST_TYPE_OUT | R500_INST_TEX_SEM_WAIT | R500_INST_LAST |
+        R500_INST_RGB_OMASK_R | R500_INST_RGB_OMASK_G | R500_INST_RGB_OMASK_B |
+        R500_INST_ALPHA_OMASK | R500_INST_RGB_CLAMP | R500_INST_ALPHA_CLAMP);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_RGB_ADDR0(0) | R500_RGB_ADDR1(0) | R500_RGB_ADDR1_CONST |
+        R500_RGB_ADDR2(0) | R500_RGB_ADDR2_CONST);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_ALPHA_ADDR0(0) | R500_ALPHA_ADDR1(0) | R500_ALPHA_ADDR1_CONST |
+        R500_ALPHA_ADDR2(0) | R500_ALPHA_ADDR2_CONST);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_ALU_RGB_SEL_A_SRC0 | R500_ALU_RGB_R_SWIZ_A_R |
+        R500_ALU_RGB_G_SWIZ_A_G | R500_ALU_RGB_B_SWIZ_A_B |
+        R500_ALU_RGB_SEL_B_SRC0 | R500_ALU_RGB_R_SWIZ_B_R |
+        R500_ALU_RGB_B_SWIZ_B_G | R500_ALU_RGB_G_SWIZ_B_B);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_ALPHA_OP_CMP | R500_ALPHA_SWIZ_A_A | R500_ALPHA_SWIZ_B_A);
+    OUT_CS_REG(R500_GA_US_VECTOR_DATA,
+        R500_ALU_RGBA_OP_CMP | R500_ALU_RGBA_R_SWIZ_0 |
+        R500_ALU_RGBA_G_SWIZ_0 | R500_ALU_RGBA_B_SWIZ_0 |
+        R500_ALU_RGBA_A_SWIZ_0);
+} else {
+    OUT_CS_REG_SEQ(R300_RS_IP_0, 8);
+    for (i = 0; i < 8; i++) {
+        OUT_CS(R300_RS_SEL_T(R300_RS_SEL_K0) |
+            R300_RS_SEL_R(R300_RS_SEL_K0) | R300_RS_SEL_Q(R300_RS_SEL_K1));
+    }
+    /* XXX */
+    OUT_CS_REG_SEQ(R300_RS_COUNT, 2);
+    OUT_CS((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN);
+    OUT_CS(1);
+    OUT_CS_REG(R300_RS_INST_0, R300_RS_INST_COL_CN_WRITE);
+
+    /* XXX magic numbers */
+    OUT_CS_REG(R300_US_CONFIG, 0);
+    OUT_CS_REG(R300_US_PIXSIZE, 2);
+    OUT_CS_REG(R300_US_CODE_OFFSET, 0x0);
+    OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0);
+    OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0);
+    OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0);
+    OUT_CS_REG(R300_US_CODE_ADDR_3, 0x400000);
+    OUT_CS_REG(R300_US_ALU_RGB_INST_0, 0x50A80);
+    OUT_CS_REG(R300_US_ALU_RGB_ADDR_0, 0x1C000000);
+    OUT_CS_REG(R300_US_ALU_ALPHA_INST_0, 0x40889);
+    OUT_CS_REG(R300_US_ALU_ALPHA_ADDR_0, 0x1000000);
+    OUT_CS_REG_SEQ(R300_US_OUT_FMT_0, 4);
+    OUT_CS(R300_C0_SEL_B | R300_C1_SEL_G | R300_C2_SEL_R | R300_C3_SEL_A);
+    OUT_CS(R300_US_OUT_FMT_UNUSED);
+    OUT_CS(R300_US_OUT_FMT_UNUSED);
+    OUT_CS(R300_US_OUT_FMT_UNUSED);
+    OUT_CS_REG(R300_US_W_FMT, R300_W_FMT_W0);
+}
+/* XXX these magic numbers should be explained when
+ * this becomes a cached state object */
+if (caps->has_tcl) {
+    OUT_CS_REG(R300_VAP_CNTL, 0xA |
+        (0x5 << R300_PVS_NUM_CNTLRS_SHIFT) |
+        (0xB << R300_VF_MAX_VTX_NUM_SHIFT) |
+        (caps->num_vert_fpus << R300_PVS_NUM_FPUS_SHIFT));
+    OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_0, 0x00100000);
+    OUT_CS_REG(R300_VAP_PVS_CONST_CNTL, 0x00000000);
+    OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_1, 0x00000001);
+    R300_PACIFY;
+    /* XXX translate these back into normal instructions */
+    OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x1);
+    OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0x0);
+    OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xF00203);
+    OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xD10001);
+    OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x1248001);
+    OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x0);
+    OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xF02203);
+    OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0xD10021);
+    OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x1248021);
+    OUT_CS_REG(R300_VAP_PVS_UPLOAD_DATA, 0x0);
+} else {
+    OUT_CS_REG(R300_VAP_CNTL, 0xA |
+        (0x5 << R300_PVS_NUM_CNTLRS_SHIFT) |
+        (0x5 << R300_VF_MAX_VTX_NUM_SHIFT) |
+        (caps->num_vert_fpus << R300_PVS_NUM_FPUS_SHIFT));
+}
+R300_PACIFY;
+END_CS;
+
+r300_emit_blend_state(r300, &blend_clear_state);
+r300_emit_blend_color_state(r300, &blend_color_clear_state);
+r300_emit_dsa_state(r300, &dsa_clear_state);
+
+BEGIN_CS(36);
+R300_PACIFY;
+/* Flush colorbuffer and blend caches. */
+OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT,
+    R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D |
+    R300_RB3D_DSTCACHE_CTLSTAT_DC_FINISH_SIGNAL);
+OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT,
+    R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
+    R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
+
+OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0, 1);
+OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+/* XXX this should not be so rigid and it still doesn't work right */
+OUT_CS_REG(R300_RB3D_COLORPITCH0, (dest->stride >> 2) | R300_COLOR_FORMAT_ARGB8888);
+OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F);
+/* XXX Packet3 */
+OUT_CS(CP_PACKET3(R200_3D_DRAW_IMMD_2, 8));
+OUT_CS(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING |
+(1 << R300_PRIM_NUM_VERTICES_SHIFT));
+OUT_CS_32F(w / 2.0);
+OUT_CS_32F(h / 2.0);
+/* XXX this should be the depth value to clear to */
+OUT_CS_32F(1.0);
+OUT_CS_32F(1.0);
+OUT_CS_32F(r);
+OUT_CS_32F(g);
+OUT_CS_32F(b);
+OUT_CS_32F(1.0);
+
+/* XXX figure out why this is 0xA and not 0x2 */
+OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, 0xA);
+/* XXX OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT,
+    R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
+    R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); */
+R300_PACIFY;
+
+END_CS;
+FLUSH_CS;
+
+    r300->dirty_state = R300_NEW_KITCHEN_SINK;
+}
+
+void r300_init_surface_functions(struct r300_context* r300)
+{
+    r300->context.surface_fill = r300_surface_fill;
+}
diff --git a/src/gallium/drivers/r300/r300_surface.h b/src/gallium/drivers/r300/r300_surface.h
new file mode 100644
index 0000000000..e1d53116a1
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_surface.h
@@ -0,0 +1,58 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_SURFACE_H
+#define R300_SURFACE_H
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+
+#include "util/u_rect.h"
+
+#include "r300_context.h"
+#include "r300_cs.h"
+#include "r300_emit.h"
+
+const struct r300_blend_state blend_clear_state = {
+    .blend_control = 0x0,
+    .alpha_blend_control = 0x0,
+    .rop = 0x0,
+    .dither = 0x0,
+};
+
+const struct r300_blend_color_state blend_color_clear_state = {
+    .blend_color = 0x0,
+    .blend_color_red_alpha = 0x0,
+    .blend_color_green_blue = 0x0,
+};
+
+const struct r300_dsa_state dsa_clear_state = {
+    .alpha_function = 0x0,
+    .alpha_reference = 0x0,
+    .z_buffer_control = 0x0,
+    .z_stencil_control = 0x0,
+    .stencil_ref_mask = R300_STENCILWRITEMASK_MASK,
+    .z_buffer_top = R300_ZTOP_ENABLE,
+    .stencil_ref_bf = 0x0,
+};
+
+#endif /* R300_SURFACE_H */
diff --git a/src/gallium/drivers/r300/r300_swtcl_emit.c b/src/gallium/drivers/r300/r300_swtcl_emit.c
new file mode 100644
index 0000000000..f6e98d23e9
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_swtcl_emit.c
@@ -0,0 +1,129 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "draw/draw_pipe.h"
+#include "util/u_memory.h"
+
+#include "r300_cs.h"
+#include "r300_context.h"
+#include "r300_reg.h"
+
+/* r300_swtcl_emit: Primitive vertex emission using an immediate
+ * vertex buffer and no HW TCL. */
+
+struct swtcl_stage {
+    /* Parent class */
+    struct draw_stage draw;
+
+    struct r300_context* r300;
+};
+
+static INLINE struct swtcl_stage* swtcl_stage(struct draw_stage* draw) {
+    return (struct swtcl_stage*)draw;
+}
+
+static void r300_emit_vertex(struct r300_context* r300,
+                             const struct vertex_header* vertex)
+{
+    /* XXX */
+}
+
+static INLINE void r300_emit_prim(struct draw_stage* draw,
+                                  struct prim_header* prim,
+                                  unsigned hwprim,
+                                  unsigned count)
+{
+    struct r300_context* r300 = swtcl_stage(draw)->r300;
+    CS_LOCALS(r300);
+    int i;
+
+    r300_emit_dirty_state(r300);
+
+    /* XXX should be count * vtx size */
+    BEGIN_CS(2 + count + 6);
+    OUT_CS(CP_PACKET3(R200_3D_DRAW_IMMD_2, count));
+    OUT_CS(hwprim | R300_PRIM_WALK_RING |
+        (count << R300_PRIM_NUM_VERTICES_SHIFT));
+
+    for (i = 0; i < count; i++) {
+        r300_emit_vertex(r300, prim->v[i]);
+    }
+    R300_PACIFY;
+    END_CS;
+}
+
+/* Just as an aside...
+ *
+ * Radeons can do many more primitives:
+ * - Line strip
+ * - Triangle fan
+ * - Triangle strip
+ * - Line loop
+ * - Quads
+ * - Quad strip
+ * - Polygons
+ *
+ * The following were just the only ones in Draw. */
+
+static void r300_emit_point(struct draw_stage* draw, struct prim_header* prim)
+{
+    r300_emit_prim(draw, prim, R300_PRIM_TYPE_POINT, 1);
+}
+
+static void r300_emit_line(struct draw_stage* draw, struct prim_header* prim)
+{
+    r300_emit_prim(draw, prim, R300_PRIM_TYPE_LINE, 2);
+}
+
+static void r300_emit_tri(struct draw_stage* draw, struct prim_header* prim)
+{
+    r300_emit_prim(draw, prim, R300_PRIM_TYPE_TRI_LIST, 3);
+}
+
+static void r300_swtcl_flush(struct draw_stage* draw, unsigned flags)
+{
+}
+
+static void r300_reset_stipple(struct draw_stage* draw)
+{
+    /* XXX */
+}
+
+static void r300_swtcl_destroy(struct draw_stage* draw)
+{
+    FREE(draw);
+}
+
+struct draw_stage* r300_draw_swtcl_stage(struct r300_context* r300)
+{
+    struct swtcl_stage* swtcl = CALLOC_STRUCT(swtcl_stage);
+
+    swtcl->r300 = r300;
+    swtcl->draw.point = r300_emit_point;
+    swtcl->draw.line = r300_emit_line;
+    swtcl->draw.tri = r300_emit_tri;
+    swtcl->draw.flush = r300_swtcl_flush;
+    swtcl->draw.reset_stipple_counter = r300_reset_stipple;
+    swtcl->draw.destroy = r300_swtcl_destroy;
+
+    return &swtcl->draw;
+}
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
new file mode 100644
index 0000000000..ff812c09f8
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -0,0 +1,193 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_texture.h"
+
+static int minify(int i)
+{
+    return MAX2(1, i >> 1);
+}
+
+static void r300_setup_miptree(struct r300_texture* tex)
+{
+    struct pipe_texture* base = &tex->tex;
+    int stride, size, offset;
+
+    for (int i = 0; i <= base->last_level; i++) {
+        if (i > 0) {
+            base->width[i] = minify(base->width[i-1]);
+            base->height[i] = minify(base->height[i-1]);
+            base->depth[i] = minify(base->depth[i-1]);
+        }
+
+        base->nblocksx[i] = pf_get_nblocksx(&base->block, base->width[i]);
+        base->nblocksy[i] = pf_get_nblocksy(&base->block, base->width[i]);
+
+        /* Radeons enjoy things in multiples of 32. */
+        /* XXX NPOT -> 64, not 32 */
+        stride = (base->nblocksx[i] * base->block.size + 63) & ~63;
+        size = stride * base->nblocksy[i] * base->depth[i];
+
+        /* XXX 64 for NPOT */
+        tex->offset[i] = (tex->size + 63) & ~63;
+        tex->size = tex->offset[i] + size;
+    }
+}
+
+/* Create a new texture. */
+static struct pipe_texture*
+    r300_texture_create(struct pipe_screen* screen,
+                        const struct pipe_texture* template)
+{
+    /* XXX struct r300_screen* r300screen = r300_screen(screen); */
+
+    struct r300_texture* tex = CALLOC_STRUCT(r300_texture);
+
+    if (!tex) {
+        return NULL;
+    }
+
+    tex->tex = *template;
+    tex->tex.refcount = 1;
+    tex->tex.screen = screen;
+
+    r300_setup_miptree(tex);
+
+    tex->buffer = screen->buffer_create(screen, 63,
+                                        PIPE_BUFFER_USAGE_PIXEL,
+                                        tex->size);
+
+    if (!tex->buffer) {
+        FREE(tex);
+        return NULL;
+    }
+
+    return (struct pipe_texture*)tex;
+}
+
+static void r300_texture_release(struct pipe_screen* screen,
+                                 struct pipe_texture** texture)
+{
+    if (!*texture) {
+        return;
+    }
+
+    (*texture)->refcount--;
+
+    if ((*texture)->refcount <= 0) {
+        struct r300_texture* tex = (struct r300_texture*)*texture;
+
+        pipe_buffer_reference(screen, &tex->buffer, NULL);
+
+        FREE(tex);
+    }
+
+    *texture = NULL;
+}
+
+static struct pipe_surface* r300_get_tex_surface(struct pipe_screen* screen,
+                                                 struct pipe_texture* texture,
+                                                 unsigned face,
+                                                 unsigned level,
+                                                 unsigned zslice,
+                                                 unsigned flags)
+{
+    struct r300_texture* tex = (struct r300_texture*)texture;
+    struct pipe_surface* surface = CALLOC_STRUCT(pipe_surface);
+    unsigned offset;
+
+    /* XXX this is certainly dependent on tex target */
+    offset = tex->offset[level];
+
+    if (surface) {
+        surface->refcount = 1;
+        pipe_texture_reference(&surface->texture, texture);
+        surface->format = texture->format;
+        surface->width = texture->width[level];
+        surface->height = texture->height[level];
+        surface->block = texture->block;
+        surface->nblocksx = texture->nblocksx[level];
+        surface->nblocksy = texture->nblocksy[level];
+        /* XXX save the actual stride instead plz kthnxbai */
+        surface->stride =
+            (texture->nblocksx[level] * texture->block.size + 63) & ~63;
+        surface->offset = offset;
+        surface->usage = flags;
+        surface->status = PIPE_SURFACE_STATUS_DEFINED;
+    }
+
+    return surface;
+}
+
+static void r300_tex_surface_release(struct pipe_screen* screen,
+                                     struct pipe_surface** surface)
+{
+    struct pipe_surface* s = *surface;
+
+    s->refcount--;
+
+    if (s->refcount <= 0) {
+        pipe_texture_reference(&s->texture, NULL);
+        FREE(s);
+    }
+
+    *surface = NULL;
+}
+
+static struct pipe_texture*
+    r300_texture_blanket(struct pipe_screen* screen,
+                         const struct pipe_texture* base,
+                         const unsigned* stride,
+                         struct pipe_buffer* buffer)
+{
+    struct r300_texture* tex;
+
+    if (base->target != PIPE_TEXTURE_2D ||
+        base->last_level != 0 ||
+        base->depth[0] != 1) {
+        return NULL;
+    }
+
+    tex = CALLOC_STRUCT(r300_texture);
+    if (!tex) {
+        return NULL;
+    }
+
+    tex->tex = *base;
+    tex->tex.refcount = 1;
+    tex->tex.screen = screen;
+
+    /* XXX tex->stride = *stride; */
+
+    pipe_buffer_reference(screen, &tex->buffer, buffer);
+
+    return (struct pipe_texture*)tex;
+}
+
+void r300_init_screen_texture_functions(struct pipe_screen* screen)
+{
+    screen->texture_create = r300_texture_create;
+    screen->texture_release = r300_texture_release;
+    screen->get_tex_surface = r300_get_tex_surface;
+    screen->tex_surface_release = r300_tex_surface_release;
+    screen->texture_blanket = r300_texture_blanket;
+}
diff --git a/src/gallium/drivers/r300/r300_texture.h b/src/gallium/drivers/r300/r300_texture.h
new file mode 100644
index 0000000000..7964229a94
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_texture.h
@@ -0,0 +1,34 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_TEXTURE_H
+#define R300_TEXTURE_H
+
+#include "pipe/p_screen.h"
+
+#include "util/u_math.h"
+
+#include "r300_context.h"
+
+void r300_init_screen_texture_functions(struct pipe_screen* screen);
+
+#endif /* R300_TEXTURE_H */
diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h
new file mode 100644
index 0000000000..5a3a212892
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_winsys.h
@@ -0,0 +1,94 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_WINSYS_H
+#define R300_WINSYS_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* The public interface header for the r300 pipe driver.
+ * Any winsys hosting this pipe needs to implement r300_winsys and then
+ * call r300_create_context to start things. */
+
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+
+struct radeon_cs;
+
+struct r300_winsys {
+
+    /* PCI ID */
+    uint32_t pci_id;
+
+    /* GB pipe count */
+    uint32_t gb_pipes;
+
+    /* CS object. This is very much like Intel's batchbuffer.
+     * Fill it full of dwords and relocs and then submit.
+     * Repeat as needed. */
+    /* Note: Unlike Mesa's version of this, we don't keep a copy of the CSM
+     * that was used to create this CS. Is this a good idea? */
+    /* Note: The pipe driver doesn't know how to use this. This is purely
+     * for the winsys. */
+    struct radeon_cs* cs;
+
+    /* Check to see if there's room for commands. */
+    boolean (*check_cs)(struct radeon_cs* cs, int size);
+
+    /* Start a command emit. */
+    void (*begin_cs)(struct radeon_cs* cs,
+           int size,
+           const char* file,
+           const char* function,
+           int line);
+
+    /* Write a dword to the command buffer. */
+    void (*write_cs_dword)(struct radeon_cs* cs, uint32_t dword);
+
+    /* Write a relocated dword to the command buffer. */
+    void (*write_cs_reloc)(struct radeon_cs* cs,
+           struct pipe_buffer* bo,
+           uint32_t rd,
+           uint32_t wd,
+           uint32_t flags);
+
+    /* Finish a command emit. */
+    void (*end_cs)(struct radeon_cs* cs,
+           const char* file,
+           const char* function,
+           int line);
+
+    /* Flush the CS. */
+    void (*flush_cs)(struct radeon_cs* cs);
+};
+
+struct pipe_context* r300_create_context(struct pipe_screen* screen,
+                                         struct pipe_winsys* winsys,
+                                         struct r300_winsys* r300_winsys);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* R300_WINSYS_H */
diff --git a/src/gallium/drivers/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile
new file mode 100644
index 0000000000..120bdfd9dd
--- /dev/null
+++ b/src/gallium/drivers/softpipe/Makefile
@@ -0,0 +1,47 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = softpipe
+
+C_SOURCES = \
+	sp_fs_exec.c \
+	sp_fs_sse.c \
+	sp_fs_llvm.c \
+	sp_clear.c \
+	sp_flush.c \
+	sp_query.c \
+	sp_context.c \
+	sp_draw_arrays.c \
+	sp_prim_setup.c \
+	sp_prim_vbuf.c \
+	sp_quad.c \
+	sp_quad_alpha_test.c \
+	sp_quad_blend.c \
+	sp_quad_colormask.c \
+	sp_quad_coverage.c \
+	sp_quad_depth_test.c \
+	sp_quad_earlyz.c \
+	sp_quad_fs.c \
+	sp_quad_occlusion.c \
+	sp_quad_output.c \
+	sp_quad_stencil.c \
+	sp_quad_stipple.c \
+	sp_screen.c \
+        sp_setup.c \
+	sp_state_blend.c \
+	sp_state_clip.c \
+	sp_state_derived.c \
+	sp_state_fs.c \
+	sp_state_sampler.c \
+	sp_state_rasterizer.c \
+	sp_state_surface.c \
+	sp_state_vertex.c \
+	sp_texture.c \
+	sp_tex_sample.c \
+	sp_tile_cache.c \
+	sp_surface.c 
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/drivers/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript
new file mode 100644
index 0000000000..c1f7daa8ab
--- /dev/null
+++ b/src/gallium/drivers/softpipe/SConscript
@@ -0,0 +1,46 @@
+Import('*')
+
+env = env.Clone()
+
+softpipe = env.ConvenienceLibrary(
+	target = 'softpipe',
+	source = [
+		'sp_fs_exec.c',
+		'sp_fs_sse.c',
+		'sp_fs_llvm.c',
+		'sp_clear.c',
+		'sp_context.c',
+		'sp_draw_arrays.c',
+		'sp_flush.c',
+		'sp_prim_setup.c',
+		'sp_prim_vbuf.c',
+		'sp_setup.c',
+		'sp_quad_alpha_test.c',
+		'sp_quad_blend.c',
+		'sp_quad.c',
+		'sp_quad_colormask.c',
+		'sp_quad_coverage.c',
+		'sp_quad_depth_test.c',
+		'sp_quad_earlyz.c',
+		'sp_quad_fs.c',
+		'sp_quad_occlusion.c',
+		'sp_quad_output.c',
+		'sp_quad_stencil.c',
+		'sp_quad_stipple.c',
+		'sp_query.c',
+		'sp_screen.c',
+		'sp_state_blend.c',
+		'sp_state_clip.c',
+		'sp_state_derived.c',
+		'sp_state_fs.c',
+		'sp_state_rasterizer.c',
+		'sp_state_sampler.c',
+		'sp_state_surface.c',
+		'sp_state_vertex.c',
+		'sp_surface.c',
+		'sp_tex_sample.c',
+		'sp_texture.c',
+		'sp_tile_cache.c',
+	])
+
+Export('softpipe')
\ No newline at end of file
diff --git a/src/gallium/drivers/softpipe/sp_clear.c b/src/gallium/drivers/softpipe/sp_clear.c
new file mode 100644
index 0000000000..ad108ec446
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_clear.c
@@ -0,0 +1,107 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Brian Paul
+ */
+
+
+#include "pipe/p_defines.h"
+#include "util/u_pack_color.h"
+#include "sp_clear.h"
+#include "sp_context.h"
+#include "sp_surface.h"
+#include "sp_state.h"
+#include "sp_tile_cache.h"
+
+
+/**
+ * Convert packed pixel from one format to another.
+ */
+static unsigned
+convert_color(enum pipe_format srcFormat, unsigned srcColor,
+              enum pipe_format dstFormat)
+{
+   ubyte r, g, b, a;
+   unsigned dstColor;
+
+   util_unpack_color_ub(srcFormat, &srcColor, &r, &g, &b, &a);
+   util_pack_color_ub(r, g, b, a, dstFormat, &dstColor);
+
+   return dstColor;
+}
+
+
+
+/**
+ * Clear the given surface to the specified value.
+ * No masking, no scissor (clear entire buffer).
+ * Note: when clearing a color buffer, the clearValue is always
+ * encoded as PIPE_FORMAT_A8R8G8B8_UNORM.
+ */
+void
+softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+               unsigned clearValue)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+   uint i;
+
+   if (softpipe->no_rast)
+      return;
+
+#if 0
+   softpipe_update_derived(softpipe); /* not needed?? */
+#endif
+
+   if (ps == sp_tile_cache_get_surface(softpipe->zsbuf_cache)) {
+      sp_tile_cache_clear(softpipe->zsbuf_cache, clearValue);
+      softpipe->framebuffer.zsbuf->status = PIPE_SURFACE_STATUS_CLEAR;
+#if TILE_CLEAR_OPTIMIZATION
+      return;
+#endif
+   }
+
+   for (i = 0; i < softpipe->framebuffer.nr_cbufs; i++) {
+      if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[i])) {
+         unsigned cv;
+         if (ps->format != PIPE_FORMAT_A8R8G8B8_UNORM) {
+            cv = convert_color(PIPE_FORMAT_A8R8G8B8_UNORM, clearValue,
+                               ps->format);
+         }
+         else {
+            cv = clearValue;
+         }
+         sp_tile_cache_clear(softpipe->cbuf_cache[i], cv);
+         softpipe->framebuffer.cbufs[i]->status = PIPE_SURFACE_STATUS_CLEAR;
+      }
+   }
+
+#if !TILE_CLEAR_OPTIMIZATION
+   /* non-cached surface */
+   pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue);
+#endif
+}
diff --git a/src/gallium/drivers/softpipe/sp_clear.h b/src/gallium/drivers/softpipe/sp_clear.h
new file mode 100644
index 0000000000..a8ed1c4ecc
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_clear.h
@@ -0,0 +1,43 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Brian Paul
+ */
+
+#ifndef SP_CLEAR_H
+#define SP_CLEAR_H
+
+#include "pipe/p_state.h"
+struct pipe_context;
+
+extern void
+softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps,
+               unsigned clearValue);
+
+
+#endif /* SP_CLEAR_H */
diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
new file mode 100644
index 0000000000..c2d882a819
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -0,0 +1,288 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2008 VMware, Inc.  All rights reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "sp_clear.h"
+#include "sp_context.h"
+#include "sp_flush.h"
+#include "sp_prim_setup.h"
+#include "sp_prim_vbuf.h"
+#include "sp_state.h"
+#include "sp_surface.h"
+#include "sp_tile_cache.h"
+#include "sp_texture.h"
+#include "sp_winsys.h"
+#include "sp_query.h"
+
+
+
+/**
+ * Map any drawing surfaces which aren't already mapped
+ */
+void
+softpipe_map_surfaces(struct softpipe_context *sp)
+{
+   unsigned i;
+
+   for (i = 0; i < sp->framebuffer.nr_cbufs; i++) {
+      sp_tile_cache_map_surfaces(sp->cbuf_cache[i]);
+   }
+
+   sp_tile_cache_map_surfaces(sp->zsbuf_cache);
+}
+
+
+/**
+ * Unmap any mapped drawing surfaces
+ */
+void
+softpipe_unmap_surfaces(struct softpipe_context *sp)
+{
+   uint i;
+
+   for (i = 0; i < sp->framebuffer.nr_cbufs; i++)
+      sp_flush_tile_cache(sp, sp->cbuf_cache[i]);
+   sp_flush_tile_cache(sp, sp->zsbuf_cache);
+
+   for (i = 0; i < sp->framebuffer.nr_cbufs; i++) {
+      sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]);
+   }
+   sp_tile_cache_unmap_surfaces(sp->zsbuf_cache);
+}
+
+
+static void softpipe_destroy( struct pipe_context *pipe )
+{
+   struct softpipe_context *softpipe = softpipe_context( pipe );
+   struct pipe_screen *screen = pipe->screen;
+   uint i;
+
+   if (softpipe->draw)
+      draw_destroy( softpipe->draw );
+
+   for (i = 0; i < SP_NUM_QUAD_THREADS; i++) {
+      softpipe->quad[i].polygon_stipple->destroy( softpipe->quad[i].polygon_stipple );
+      softpipe->quad[i].earlyz->destroy( softpipe->quad[i].earlyz );
+      softpipe->quad[i].shade->destroy( softpipe->quad[i].shade );
+      softpipe->quad[i].alpha_test->destroy( softpipe->quad[i].alpha_test );
+      softpipe->quad[i].depth_test->destroy( softpipe->quad[i].depth_test );
+      softpipe->quad[i].stencil_test->destroy( softpipe->quad[i].stencil_test );
+      softpipe->quad[i].occlusion->destroy( softpipe->quad[i].occlusion );
+      softpipe->quad[i].coverage->destroy( softpipe->quad[i].coverage );
+      softpipe->quad[i].blend->destroy( softpipe->quad[i].blend );
+      softpipe->quad[i].colormask->destroy( softpipe->quad[i].colormask );
+      softpipe->quad[i].output->destroy( softpipe->quad[i].output );
+   }
+
+   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++)
+      sp_destroy_tile_cache(softpipe->cbuf_cache[i]);
+   sp_destroy_tile_cache(softpipe->zsbuf_cache);
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++)
+      sp_destroy_tile_cache(softpipe->tex_cache[i]);
+
+   for (i = 0; i < Elements(softpipe->constants); i++) {
+      if (softpipe->constants[i].buffer) {
+         pipe_buffer_reference(screen, &softpipe->constants[i].buffer, NULL);
+      }
+   }
+
+   FREE( softpipe );
+}
+
+
+struct pipe_context *
+softpipe_create( struct pipe_screen *screen,
+                 struct pipe_winsys *pipe_winsys,
+                 void *unused )
+{
+   struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context);
+   uint i;
+
+   util_init_math();
+
+#ifdef PIPE_ARCH_X86
+   softpipe->use_sse = !debug_get_bool_option( "GALLIUM_NOSSE", FALSE );
+#else
+   softpipe->use_sse = FALSE;
+#endif
+
+   softpipe->dump_fs = debug_get_bool_option( "GALLIUM_DUMP_FS", FALSE );
+
+   softpipe->pipe.winsys = pipe_winsys;
+   softpipe->pipe.screen = screen;
+   softpipe->pipe.destroy = softpipe_destroy;
+
+   /* state setters */
+   softpipe->pipe.create_blend_state = softpipe_create_blend_state;
+   softpipe->pipe.bind_blend_state   = softpipe_bind_blend_state;
+   softpipe->pipe.delete_blend_state = softpipe_delete_blend_state;
+
+   softpipe->pipe.create_sampler_state = softpipe_create_sampler_state;
+   softpipe->pipe.bind_sampler_states  = softpipe_bind_sampler_states;
+   softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state;
+
+   softpipe->pipe.create_depth_stencil_alpha_state = softpipe_create_depth_stencil_state;
+   softpipe->pipe.bind_depth_stencil_alpha_state   = softpipe_bind_depth_stencil_state;
+   softpipe->pipe.delete_depth_stencil_alpha_state = softpipe_delete_depth_stencil_state;
+
+   softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state;
+   softpipe->pipe.bind_rasterizer_state   = softpipe_bind_rasterizer_state;
+   softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state;
+
+   softpipe->pipe.create_fs_state = softpipe_create_fs_state;
+   softpipe->pipe.bind_fs_state   = softpipe_bind_fs_state;
+   softpipe->pipe.delete_fs_state = softpipe_delete_fs_state;
+
+   softpipe->pipe.create_vs_state = softpipe_create_vs_state;
+   softpipe->pipe.bind_vs_state   = softpipe_bind_vs_state;
+   softpipe->pipe.delete_vs_state = softpipe_delete_vs_state;
+
+   softpipe->pipe.set_blend_color = softpipe_set_blend_color;
+   softpipe->pipe.set_clip_state = softpipe_set_clip_state;
+   softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer;
+   softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state;
+   softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple;
+   softpipe->pipe.set_scissor_state = softpipe_set_scissor_state;
+   softpipe->pipe.set_sampler_textures = softpipe_set_sampler_textures;
+   softpipe->pipe.set_viewport_state = softpipe_set_viewport_state;
+
+   softpipe->pipe.set_vertex_buffers = softpipe_set_vertex_buffers;
+   softpipe->pipe.set_vertex_elements = softpipe_set_vertex_elements;
+
+   softpipe->pipe.draw_arrays = softpipe_draw_arrays;
+   softpipe->pipe.draw_elements = softpipe_draw_elements;
+   softpipe->pipe.draw_range_elements = softpipe_draw_range_elements;
+   softpipe->pipe.set_edgeflags = softpipe_set_edgeflags;
+
+
+   softpipe->pipe.clear = softpipe_clear;
+   softpipe->pipe.flush = softpipe_flush;
+
+   softpipe_init_query_funcs( softpipe );
+   softpipe_init_texture_funcs( softpipe );
+
+   /*
+    * Alloc caches for accessing drawing surfaces and textures.
+    * Must be before quad stage setup!
+    */
+   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++)
+      softpipe->cbuf_cache[i] = sp_create_tile_cache( screen );
+   softpipe->zsbuf_cache = sp_create_tile_cache( screen );
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++)
+      softpipe->tex_cache[i] = sp_create_tile_cache( screen );
+
+
+   /* setup quad rendering stages */
+   for (i = 0; i < SP_NUM_QUAD_THREADS; i++) {
+      softpipe->quad[i].polygon_stipple = sp_quad_polygon_stipple_stage(softpipe);
+      softpipe->quad[i].earlyz = sp_quad_earlyz_stage(softpipe);
+      softpipe->quad[i].shade = sp_quad_shade_stage(softpipe);
+      softpipe->quad[i].alpha_test = sp_quad_alpha_test_stage(softpipe);
+      softpipe->quad[i].depth_test = sp_quad_depth_test_stage(softpipe);
+      softpipe->quad[i].stencil_test = sp_quad_stencil_test_stage(softpipe);
+      softpipe->quad[i].occlusion = sp_quad_occlusion_stage(softpipe);
+      softpipe->quad[i].coverage = sp_quad_coverage_stage(softpipe);
+      softpipe->quad[i].blend = sp_quad_blend_stage(softpipe);
+      softpipe->quad[i].colormask = sp_quad_colormask_stage(softpipe);
+      softpipe->quad[i].output = sp_quad_output_stage(softpipe);
+   }
+
+   /* vertex shader samplers */
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      softpipe->tgsi.vert_samplers[i].base.get_samples = sp_get_samples_vertex;
+      softpipe->tgsi.vert_samplers[i].unit = i;
+      softpipe->tgsi.vert_samplers[i].sp = softpipe;
+      softpipe->tgsi.vert_samplers[i].cache = softpipe->tex_cache[i];
+      softpipe->tgsi.vert_samplers_list[i] = &softpipe->tgsi.vert_samplers[i];
+   }
+
+   /* fragment shader samplers */
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      softpipe->tgsi.frag_samplers[i].base.get_samples = sp_get_samples_fragment;
+      softpipe->tgsi.frag_samplers[i].unit = i;
+      softpipe->tgsi.frag_samplers[i].sp = softpipe;
+      softpipe->tgsi.frag_samplers[i].cache = softpipe->tex_cache[i];
+      softpipe->tgsi.frag_samplers_list[i] = &softpipe->tgsi.frag_samplers[i];
+   }
+
+   /*
+    * Create drawing context and plug our rendering stage into it.
+    */
+   softpipe->draw = draw_create();
+   if (!softpipe->draw) 
+      goto fail;
+
+   draw_texture_samplers(softpipe->draw,
+                         PIPE_MAX_SAMPLERS,
+                         (struct tgsi_sampler **)
+                            softpipe->tgsi.vert_samplers_list);
+
+   softpipe->setup = sp_draw_render_stage(softpipe);
+   if (!softpipe->setup)
+      goto fail;
+
+   if (debug_get_bool_option( "SP_NO_RAST", FALSE ))
+      softpipe->no_rast = TRUE;
+
+   if (debug_get_bool_option( "SP_NO_VBUF", FALSE )) {
+      /* Deprecated path -- vbuf is the intended interface to the draw module:
+       */
+      draw_set_rasterize_stage(softpipe->draw, softpipe->setup);
+   }
+   else {
+      sp_init_vbuf(softpipe);
+   }
+
+   /* plug in AA line/point stages */
+   draw_install_aaline_stage(softpipe->draw, &softpipe->pipe);
+   draw_install_aapoint_stage(softpipe->draw, &softpipe->pipe);
+
+#if USE_DRAW_STAGE_PSTIPPLE
+   /* Do polygon stipple w/ texture map + frag prog? */
+   draw_install_pstipple_stage(softpipe->draw, &softpipe->pipe);
+#endif
+
+   sp_init_surface_functions(softpipe);
+
+   return &softpipe->pipe;
+
+ fail:
+   softpipe_destroy(&softpipe->pipe);
+   return NULL;
+}
+
diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
new file mode 100644
index 0000000000..e2451c6ecb
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -0,0 +1,174 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef SP_CONTEXT_H
+#define SP_CONTEXT_H
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+
+#include "draw/draw_vertex.h"
+
+#include "sp_quad.h"
+#include "sp_tex_sample.h"
+
+
+/**
+ * This is a temporary variable for testing draw-stage polygon stipple.
+ * If zero, do stipple in sp_quad_stipple.c
+ */
+#define USE_DRAW_STAGE_PSTIPPLE 1
+
+/* Number of threads working on individual quads.
+ * Setting to 1 disables this feature.
+ */
+#define SP_NUM_QUAD_THREADS 1
+
+struct softpipe_winsys;
+struct softpipe_vbuf_render;
+struct draw_context;
+struct draw_stage;
+struct softpipe_tile_cache;
+struct sp_fragment_shader;
+struct sp_vertex_shader;
+
+
+struct softpipe_context {
+   struct pipe_context pipe;  /**< base class */
+
+   /* The most recent drawing state as set by the driver:
+    */
+   const struct pipe_blend_state   *blend;
+   const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS];
+   const struct pipe_depth_stencil_alpha_state   *depth_stencil;
+   const struct pipe_rasterizer_state *rasterizer;
+   const struct sp_fragment_shader *fs;
+   const struct sp_vertex_shader *vs;
+
+   struct pipe_blend_color blend_color;
+   struct pipe_clip_state clip;
+   struct pipe_constant_buffer constants[PIPE_SHADER_TYPES];
+   struct pipe_framebuffer_state framebuffer;
+   struct pipe_poly_stipple poly_stipple;
+   struct pipe_scissor_state scissor;
+   struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
+   struct pipe_viewport_state viewport;
+   struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+   struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
+   unsigned dirty;
+
+   unsigned num_samplers;
+   unsigned num_textures;
+   unsigned num_vertex_elements;
+   unsigned num_vertex_buffers;
+
+   boolean no_rast;
+
+   /* Counter for occlusion queries.  Note this supports overlapping
+    * queries.
+    */
+   uint64_t occlusion_count;
+
+   /*
+    * Mapped vertex buffers
+    */
+   ubyte *mapped_vbuffer[PIPE_MAX_ATTRIBS];
+   
+   /** Mapped constant buffers */
+   void *mapped_constants[PIPE_SHADER_TYPES];
+
+   /** Vertex format */
+   struct vertex_info vertex_info;
+   struct vertex_info vertex_info_vbuf;
+
+   int psize_slot;
+
+   unsigned reduced_api_prim;  /**< PIPE_PRIM_POINTS, _LINES or _TRIANGLES */
+
+#if 0
+   /* Stipple derived state:
+    */
+   ubyte stipple_masks[16][16];
+#endif
+
+   /** Derived from scissor and surface bounds: */
+   struct pipe_scissor_state cliprect;
+
+   unsigned line_stipple_counter;
+
+   /** Software quad rendering pipeline */
+   struct {
+      struct quad_stage *polygon_stipple;
+      struct quad_stage *earlyz;
+      struct quad_stage *shade;
+      struct quad_stage *alpha_test;
+      struct quad_stage *stencil_test;
+      struct quad_stage *depth_test;
+      struct quad_stage *occlusion;
+      struct quad_stage *coverage;
+      struct quad_stage *blend;
+      struct quad_stage *colormask;
+      struct quad_stage *output;
+
+      struct quad_stage *first; /**< points to one of the above stages */
+   } quad[SP_NUM_QUAD_THREADS];
+
+   /** TGSI exec things */
+   struct {
+      struct sp_shader_sampler vert_samplers[PIPE_MAX_SAMPLERS];
+      struct sp_shader_sampler *vert_samplers_list[PIPE_MAX_SAMPLERS];
+      struct sp_shader_sampler frag_samplers[PIPE_MAX_SAMPLERS];
+      struct sp_shader_sampler *frag_samplers_list[PIPE_MAX_SAMPLERS];
+   } tgsi;
+
+   /** The primitive drawing context */
+   struct draw_context *draw;
+   struct draw_stage *setup;
+   struct draw_stage *vbuf;
+   struct softpipe_vbuf_render *vbuf_render;
+
+   struct softpipe_tile_cache *cbuf_cache[PIPE_MAX_COLOR_BUFS];
+   struct softpipe_tile_cache *zsbuf_cache;
+
+   struct softpipe_tile_cache *tex_cache[PIPE_MAX_SAMPLERS];
+
+   int use_sse : 1;
+   int dump_fs : 1;
+};
+
+
+static INLINE struct softpipe_context *
+softpipe_context( struct pipe_context *pipe )
+{
+   return (struct softpipe_context *)pipe;
+}
+
+#endif /* SP_CONTEXT_H */
+
diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c
new file mode 100644
index 0000000000..7e3a25e34b
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c
@@ -0,0 +1,208 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Brian Paul
+ *    Keith Whitwell
+ */
+
+
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+
+#include "sp_context.h"
+#include "sp_state.h"
+
+#include "draw/draw_context.h"
+
+
+
+static void
+softpipe_map_constant_buffers(struct softpipe_context *sp)
+{
+   struct pipe_winsys *ws = sp->pipe.winsys;
+   uint i, size;
+
+   for (i = 0; i < PIPE_SHADER_TYPES; i++) {
+      if (sp->constants[i].buffer && sp->constants[i].buffer->size)
+         sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer,
+                                                  PIPE_BUFFER_USAGE_CPU_READ);
+   }
+
+   if (sp->constants[PIPE_SHADER_VERTEX].buffer)
+      size = sp->constants[PIPE_SHADER_VERTEX].buffer->size;
+   else
+      size = 0;
+
+   draw_set_mapped_constant_buffer(sp->draw,
+                                   sp->mapped_constants[PIPE_SHADER_VERTEX],
+                                   size);
+}
+
+static void
+softpipe_unmap_constant_buffers(struct softpipe_context *sp)
+{
+   struct pipe_winsys *ws = sp->pipe.winsys;
+   uint i;
+
+   /* really need to flush all prims since the vert/frag shaders const buffers
+    * are going away now.
+    */
+   draw_flush(sp->draw);
+
+   draw_set_mapped_constant_buffer(sp->draw, NULL, 0);
+
+   for (i = 0; i < 2; i++) {
+      if (sp->constants[i].buffer && sp->constants[i].buffer->size)
+         ws->buffer_unmap(ws, sp->constants[i].buffer);
+      sp->mapped_constants[i] = NULL;
+   }
+}
+
+
+static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = {
+   PIPE_PRIM_POINTS,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_LINES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES,
+   PIPE_PRIM_TRIANGLES
+};
+
+
+boolean
+softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
+                     unsigned start, unsigned count)
+{
+   return softpipe_draw_elements(pipe, NULL, 0, mode, start, count);
+}
+
+
+
+/**
+ * Draw vertex arrays, with optional indexing.
+ * Basically, map the vertex buffers (and drawing surfaces), then hand off
+ * the drawing to the 'draw' module.
+ *
+ * XXX should the element buffer be specified/bound with a separate function?
+ */
+
+boolean
+softpipe_draw_range_elements(struct pipe_context *pipe,
+                             struct pipe_buffer *indexBuffer,
+                             unsigned indexSize,
+                             unsigned min_index,
+                             unsigned max_index,
+                             unsigned mode, unsigned start, unsigned count)
+{
+   struct softpipe_context *sp = softpipe_context(pipe);
+   struct draw_context *draw = sp->draw;
+   unsigned i;
+
+   sp->reduced_api_prim = reduced_prim[mode];
+
+   if (sp->dirty)
+      softpipe_update_derived( sp );
+
+   softpipe_map_surfaces(sp);
+   softpipe_map_constant_buffers(sp);
+
+   /*
+    * Map vertex buffers
+    */
+   for (i = 0; i < sp->num_vertex_buffers; i++) {
+      void *buf
+         = pipe_buffer_map(pipe->screen,
+                                    sp->vertex_buffer[i].buffer,
+                                    PIPE_BUFFER_USAGE_CPU_READ);
+      draw_set_mapped_vertex_buffer(draw, i, buf);
+   }
+   /* Map index buffer, if present */
+   if (indexBuffer) {
+      void *mapped_indexes
+         = pipe_buffer_map(pipe->screen, indexBuffer,
+                                    PIPE_BUFFER_USAGE_CPU_READ);
+      draw_set_mapped_element_buffer_range(draw, indexSize,
+                                           min_index,
+                                           max_index,
+                                           mapped_indexes);
+   }
+   else {
+      /* no index/element buffer */
+      draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL);
+   }
+
+
+   /* draw! */
+   draw_arrays(draw, mode, start, count);
+
+   /*
+    * unmap vertex/index buffers - will cause draw module to flush
+    */
+   for (i = 0; i < sp->num_vertex_buffers; i++) {
+      draw_set_mapped_vertex_buffer(draw, i, NULL);
+      pipe_buffer_unmap(pipe->screen, sp->vertex_buffer[i].buffer);
+   }
+   if (indexBuffer) {
+      draw_set_mapped_element_buffer(draw, 0, NULL);
+      pipe_buffer_unmap(pipe->screen, indexBuffer);
+   }
+
+
+   /* Note: leave drawing surfaces mapped */
+   softpipe_unmap_constant_buffers(sp);
+
+   return TRUE;
+}
+
+boolean
+softpipe_draw_elements(struct pipe_context *pipe,
+                       struct pipe_buffer *indexBuffer,
+                       unsigned indexSize,
+                       unsigned mode, unsigned start, unsigned count)
+{
+   return softpipe_draw_range_elements( pipe, indexBuffer,
+                                        indexSize,
+                                        0, 0xffffffff,
+                                        mode, start, count );
+}
+
+
+
+void
+softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags)
+{
+   struct softpipe_context *sp = softpipe_context(pipe);
+   draw_set_edgeflags(sp->draw, edgeflags);
+}
+
diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c
new file mode 100644
index 0000000000..c21faf57f3
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_flush.c
@@ -0,0 +1,92 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "pipe/p_defines.h"
+#include "draw/draw_context.h"
+#include "sp_flush.h"
+#include "sp_context.h"
+#include "sp_surface.h"
+#include "sp_state.h"
+#include "sp_tile_cache.h"
+#include "sp_winsys.h"
+
+
+void
+softpipe_flush( struct pipe_context *pipe,
+		unsigned flags,
+                struct pipe_fence_handle **fence )
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+   uint i;
+
+   draw_flush(softpipe->draw);
+
+   if (flags & PIPE_FLUSH_TEXTURE_CACHE) {
+      for (i = 0; i < softpipe->num_textures; i++) {
+         sp_flush_tile_cache(softpipe, softpipe->tex_cache[i]);
+      }
+   }
+
+   if (flags & PIPE_FLUSH_RENDER_CACHE) {
+      for (i = 0; i < softpipe->framebuffer.nr_cbufs; i++)
+         if (softpipe->cbuf_cache[i])
+            sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]);
+
+      if (softpipe->zsbuf_cache)
+         sp_flush_tile_cache(softpipe, softpipe->zsbuf_cache);
+
+      /* Need this call for hardware buffers before swapbuffers.
+       *
+       * there should probably be another/different flush-type function
+       * that's called before swapbuffers because we don't always want
+       * to unmap surfaces when flushing.
+       */
+      softpipe_unmap_surfaces(softpipe);
+   }
+
+   /* Enable to dump BMPs of the color/depth buffers each frame */
+#if 0
+   if(flags & PIPE_FLUSH_FRAME) {
+      static unsigned frame_no = 1;
+      static char filename[256];
+      util_snprintf(filename, sizeof(filename), "cbuf_%u.bmp", frame_no);
+      debug_dump_surface_bmp(filename, softpipe->framebuffer.cbufs[0]);
+      util_snprintf(filename, sizeof(filename), "zsbuf_%u.bmp", frame_no);
+      debug_dump_surface_bmp(filename, softpipe->framebuffer.zsbuf);
+      ++frame_no;
+   }
+#endif
+   
+   if (fence)
+      *fence = NULL;
+}
+
diff --git a/src/gallium/drivers/softpipe/sp_flush.h b/src/gallium/drivers/softpipe/sp_flush.h
new file mode 100644
index 0000000000..68d9b5fa83
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_flush.h
@@ -0,0 +1,37 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef SP_FLUSH_H
+#define SP_FLUSH_H
+
+struct pipe_context;
+struct pipe_fence_handle;
+
+void softpipe_flush(struct pipe_context *pipe, unsigned flags,
+                    struct pipe_fence_handle **fence);
+
+#endif
diff --git a/src/gallium/drivers/softpipe/sp_fs.h b/src/gallium/drivers/softpipe/sp_fs.h
new file mode 100644
index 0000000000..4792ace3a3
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_fs.h
@@ -0,0 +1,54 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef SP_FS_H
+#define SP_FS_H
+
+struct sp_fragment_shader *
+softpipe_create_fs_exec(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ);
+
+struct sp_fragment_shader *
+softpipe_create_fs_sse(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ);
+
+struct sp_fragment_shader *
+softpipe_create_fs_llvm(struct softpipe_context *softpipe,
+			const struct pipe_shader_state *templ);
+
+struct tgsi_interp_coef;
+struct tgsi_exec_vector;
+
+void sp_setup_pos_vector(const struct tgsi_interp_coef *coef,
+			 float x, float y,
+			 struct tgsi_exec_vector *quadpos);
+
+
+#endif
diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c
new file mode 100644
index 0000000000..453b0373f0
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_fs_exec.c
@@ -0,0 +1,164 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_fs.h"
+#include "sp_headers.h"
+
+
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "tgsi/tgsi_exec.h"
+#include "tgsi/tgsi_parse.h"
+
+struct sp_exec_fragment_shader
+{
+   struct sp_fragment_shader base;
+};
+
+
+/** cast wrapper */
+static INLINE struct sp_exec_fragment_shader *
+sp_exec_fragment_shader(const struct sp_fragment_shader *base)
+{
+   return (struct sp_exec_fragment_shader *) base;
+}
+
+
+/**
+ * Compute quad X,Y,Z,W for the four fragments in a quad.
+ *
+ * This should really be part of the compiled shader.
+ */
+void
+sp_setup_pos_vector(const struct tgsi_interp_coef *coef,
+		    float x, float y,
+		    struct tgsi_exec_vector *quadpos)
+{
+   uint chan;
+   /* do X */
+   quadpos->xyzw[0].f[0] = x;
+   quadpos->xyzw[0].f[1] = x + 1;
+   quadpos->xyzw[0].f[2] = x;
+   quadpos->xyzw[0].f[3] = x + 1;
+
+   /* do Y */
+   quadpos->xyzw[1].f[0] = y;
+   quadpos->xyzw[1].f[1] = y;
+   quadpos->xyzw[1].f[2] = y + 1;
+   quadpos->xyzw[1].f[3] = y + 1;
+
+   /* do Z and W for all fragments in the quad */
+   for (chan = 2; chan < 4; chan++) {
+      const float dadx = coef->dadx[chan];
+      const float dady = coef->dady[chan];
+      const float a0 = coef->a0[chan] + dadx * x + dady * y;
+      quadpos->xyzw[chan].f[0] = a0;
+      quadpos->xyzw[chan].f[1] = a0 + dadx;
+      quadpos->xyzw[chan].f[2] = a0 + dady;
+      quadpos->xyzw[chan].f[3] = a0 + dadx + dady;
+   }
+}
+
+
+static void
+exec_prepare( const struct sp_fragment_shader *base,
+	      struct tgsi_exec_machine *machine,
+	      struct tgsi_sampler **samplers )
+{
+   /*
+    * Bind tokens/shader to the interpreter's machine state.
+    * Avoid redundant binding.
+    */
+   if (machine->Tokens != base->shader.tokens) {
+      tgsi_exec_machine_bind_shader( machine,
+                                     base->shader.tokens,
+                                     PIPE_MAX_SAMPLERS,
+                                     samplers );
+   }
+}
+
+
+
+
+/* TODO: hide the machine struct in here somewhere, remove from this
+ * interface:
+ */
+static unsigned 
+exec_run( const struct sp_fragment_shader *base,
+	  struct tgsi_exec_machine *machine,
+	  struct quad_header *quad )
+{
+
+   /* Compute X, Y, Z, W vals for this quad */
+   sp_setup_pos_vector(quad->posCoef, 
+		       (float)quad->input.x0, (float)quad->input.y0, 
+		       &machine->QuadPos);
+   
+   return tgsi_exec_machine_run( machine );
+}
+
+
+
+static void 
+exec_delete( struct sp_fragment_shader *base )
+{
+   FREE((void *) base->shader.tokens);
+   FREE(base);
+}
+
+
+
+
+
+struct sp_fragment_shader *
+softpipe_create_fs_exec(struct softpipe_context *softpipe,
+			const struct pipe_shader_state *templ)
+{
+   struct sp_exec_fragment_shader *shader;
+
+   /* Decide whether we'll be codegenerating this shader and if so do
+    * that now.
+    */
+
+   shader = CALLOC_STRUCT(sp_exec_fragment_shader);
+   if (!shader)
+      return NULL;
+
+   /* we need to keep a local copy of the tokens */
+   shader->base.shader.tokens = tgsi_dup_tokens(templ->tokens);
+   shader->base.prepare = exec_prepare;
+   shader->base.run = exec_run;
+   shader->base.delete = exec_delete;
+
+   return &shader->base;
+}
+
diff --git a/src/gallium/drivers/softpipe/sp_fs_llvm.c b/src/gallium/drivers/softpipe/sp_fs_llvm.c
new file mode 100644
index 0000000000..34adac5226
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_fs_llvm.c
@@ -0,0 +1,200 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:
+ *   Zack Rusin
+ */
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_fs.h"
+
+
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "tgsi/tgsi_sse2.h"
+
+#if 0
+
+struct sp_llvm_fragment_shader {
+   struct sp_fragment_shader base;
+   struct gallivm_prog *llvm_prog;
+};
+
+static void
+shade_quad_llvm(struct quad_stage *qs,
+                struct quad_header *quad)
+{
+   struct quad_shade_stage *qss = quad_shade_stage(qs);
+   struct softpipe_context *softpipe = qs->softpipe;
+   float dests[4][16][4] ALIGN16_ATTRIB;
+   float inputs[4][16][4] ALIGN16_ATTRIB;
+   const float fx = (float) quad->x0;
+   const float fy = (float) quad->y0;
+   struct gallivm_prog *llvm = qss->llvm_prog;
+
+   inputs[0][0][0] = fx;
+   inputs[1][0][0] = fx + 1.0f;
+   inputs[2][0][0] = fx;
+   inputs[3][0][0] = fx + 1.0f;
+
+   inputs[0][0][1] = fy;
+   inputs[1][0][1] = fy;
+   inputs[2][0][1] = fy + 1.0f;
+   inputs[3][0][1] = fy + 1.0f;
+
+
+   gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef);
+
+#if DLLVM
+   debug_printf("MASK = %d\n", quad->mask);
+   for (int i = 0; i < 4; ++i) {
+      for (int j = 0; j < 2; ++j) {
+         debug_printf("IN(%d,%d) [%f %f %f %f]\n", i, j, 
+                inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]);
+      }
+   }
+#endif
+
+   quad->mask &=
+      gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs,
+                                   softpipe->mapped_constants[PIPE_SHADER_FRAGMENT],
+                                   qss->samplers);
+#if DLLVM
+   debug_printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n",
+          dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], 
+          dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]);
+#endif
+
+   /* store result color */
+   if (qss->colorOutSlot >= 0) {
+      unsigned i;
+      /* XXX need to handle multiple color outputs someday */
+      allvmrt(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot]
+             == TGSI_SEMANTIC_COLOR);
+      for (i = 0; i < QUAD_SIZE; ++i) {
+         quad->outputs.color[0][0][i] = dests[i][qss->colorOutSlot][0];
+         quad->outputs.color[0][1][i] = dests[i][qss->colorOutSlot][1];
+         quad->outputs.color[0][2][i] = dests[i][qss->colorOutSlot][2];
+         quad->outputs.color[0][3][i] = dests[i][qss->colorOutSlot][3];
+      }
+   }
+#if DLLVM
+   for (int i = 0; i < QUAD_SIZE; ++i) {
+      debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot,
+             quad->outputs.color[0][0][i],
+             quad->outputs.color[0][1][i],
+             quad->outputs.color[0][2][i],
+             quad->outputs.color[0][3][i]);
+   }
+#endif
+
+   /* store result Z */
+   if (qss->depthOutSlot >= 0) {
+      /* output[slot] is new Z */
+      uint i;
+      for (i = 0; i < 4; i++) {
+         quad->outputs.depth[i] = dests[i][0][2];
+      }
+   }
+   else {
+      /* copy input Z (which was interpolated by the executor) to output Z */
+      uint i;
+      for (i = 0; i < 4; i++) {
+         quad->outputs.depth[i] = inputs[i][0][2];
+      }
+   }
+#if DLLVM
+   debug_printf("D [%f, %f, %f, %f] mask = %d\n",
+             quad->outputs.depth[0],
+             quad->outputs.depth[1],
+             quad->outputs.depth[2],
+             quad->outputs.depth[3], quad->mask);
+#endif
+
+   /* shader may cull fragments */
+   if( quad->mask ) {
+      qs->next->run( qs->next, quad );
+   }
+}
+
+
+unsigned 
+run_llvm_fs( const struct sp_fragment_shader *base,
+	     struct foo *machine )
+{
+}
+
+
+void 
+delete_llvm_fs( struct sp_fragment_shader *base )
+{
+   FREE(base);
+}
+
+
+struct sp_fragment_shader *
+softpipe_create_fs_llvm(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ)
+{
+   struct sp_llvm_fragment_shader *shader = NULL;
+
+   /* LLVM fragment shaders currently disabled:
+    */
+   state = CALLOC_STRUCT(sp_llvm_shader_state);
+   if (!state)
+      return NULL;
+
+   state->llvm_prog = 0;
+
+   if (!gallivm_global_cpu_engine()) {
+      gallivm_cpu_engine_create(state->llvm_prog);
+   }
+   else
+      gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog);
+   
+   if (shader) {
+      shader->base.run = run_llvm_fs;
+      shader->base.delete = delete_llvm_fs;
+   }
+
+   return shader;
+}
+
+
+#else
+
+struct sp_fragment_shader *
+softpipe_create_fs_llvm(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ)
+{
+   return NULL;
+}
+
+#endif
diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c
new file mode 100644
index 0000000000..9a273c8764
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_fs_sse.c
@@ -0,0 +1,169 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_fs.h"
+#include "sp_headers.h"
+
+
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "tgsi/tgsi_exec.h"
+#include "tgsi/tgsi_sse2.h"
+
+
+#if defined(PIPE_ARCH_X86)
+
+#include "rtasm/rtasm_x86sse.h"
+
+/* Surely this should be defined somewhere in a tgsi header:
+ */
+typedef void (PIPE_CDECL *codegen_function)(
+   const struct tgsi_exec_vector *input,
+   struct tgsi_exec_vector *output,
+   const float (*constant)[4],
+   struct tgsi_exec_vector *temporary,
+   const struct tgsi_interp_coef *coef,
+   float (*immediates)[4]
+   //, const struct tgsi_exec_vector *quadPos
+ );
+
+
+struct sp_sse_fragment_shader {
+   struct sp_fragment_shader base;
+   struct x86_function             sse2_program;
+   codegen_function func;
+   float immediates[TGSI_EXEC_NUM_IMMEDIATES][4];
+};
+
+
+
+static void
+fs_sse_prepare( const struct sp_fragment_shader *base,
+		struct tgsi_exec_machine *machine,
+		struct tgsi_sampler **samplers )
+{
+}
+
+
+/* TODO: codegenerate the whole run function, skip this wrapper.
+ * TODO: break dependency on tgsi_exec_machine struct
+ * TODO: push Position calculation into the generated shader
+ * TODO: process >1 quad at a time
+ */
+static unsigned 
+fs_sse_run( const struct sp_fragment_shader *base,
+	    struct tgsi_exec_machine *machine,
+	    struct quad_header *quad )
+{
+   struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base;
+
+   /* Compute X, Y, Z, W vals for this quad -- place in temp[0] for now */
+   sp_setup_pos_vector(quad->posCoef, 
+		       (float)quad->input.x0, (float)quad->input.y0, 
+		       machine->Temps);
+
+   /* init kill mask */
+   tgsi_set_kill_mask(machine, 0x0);
+   tgsi_set_exec_mask(machine, 1, 1, 1, 1);
+
+   shader->func( machine->Inputs,
+		 machine->Outputs,
+		 machine->Consts,
+		 machine->Temps,
+		 machine->InterpCoefs,
+                 shader->immediates
+		 //	 , &machine->QuadPos
+      );
+
+   return ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]);
+}
+
+
+static void 
+fs_sse_delete( struct sp_fragment_shader *base )
+{
+   struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base;
+
+   x86_release_func( &shader->sse2_program );
+   FREE(shader);
+}
+
+
+struct sp_fragment_shader *
+softpipe_create_fs_sse(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ)
+{
+   struct sp_sse_fragment_shader *shader;
+
+   if (!softpipe->use_sse)
+      return NULL;
+
+   shader = CALLOC_STRUCT(sp_sse_fragment_shader);
+   if (!shader)
+      return NULL;
+
+   x86_init_func( &shader->sse2_program );
+   
+   if (!tgsi_emit_sse2( templ->tokens, &shader->sse2_program,
+                        shader->immediates, FALSE )) {
+      FREE(shader);
+      return NULL;
+   }
+
+   shader->func = (codegen_function) x86_get_func( &shader->sse2_program );
+   if (!shader->func) {
+      x86_release_func( &shader->sse2_program );
+      FREE(shader);
+      return NULL;
+   }
+
+   shader->base.shader = *templ;
+   shader->base.prepare = fs_sse_prepare;
+   shader->base.run = fs_sse_run;
+   shader->base.delete = fs_sse_delete;
+
+   return &shader->base;
+}
+
+
+#else
+
+/* Maybe put this varient in the header file.
+ */
+struct sp_fragment_shader *
+softpipe_create_fs_sse(struct softpipe_context *softpipe,
+		       const struct pipe_shader_state *templ)
+{
+   return NULL;
+}
+
+#endif
diff --git a/src/gallium/drivers/softpipe/sp_headers.h b/src/gallium/drivers/softpipe/sp_headers.h
new file mode 100644
index 0000000000..4a42cb3c19
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_headers.h
@@ -0,0 +1,95 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef SP_HEADERS_H
+#define SP_HEADERS_H
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_exec.h"
+
+#define PRIM_POINT 1
+#define PRIM_LINE  2
+#define PRIM_TRI   3
+
+
+/* The rasterizer generates 2x2 quads of fragment and feeds them to
+ * the current fp_machine (see below).
+ * Remember that Y=0=top with Y increasing down the window.
+ */
+#define QUAD_TOP_LEFT     0
+#define QUAD_TOP_RIGHT    1
+#define QUAD_BOTTOM_LEFT  2
+#define QUAD_BOTTOM_RIGHT 3
+
+#define MASK_TOP_LEFT     (1 << QUAD_TOP_LEFT)
+#define MASK_TOP_RIGHT    (1 << QUAD_TOP_RIGHT)
+#define MASK_BOTTOM_LEFT  (1 << QUAD_BOTTOM_LEFT)
+#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT)
+#define MASK_ALL          0xf
+
+
+/**
+ * Encodes everything we need to know about a 2x2 pixel block.  Uses
+ * "Channel-Serial" or "SoA" layout.  
+ */
+struct quad_header_input
+{
+   int x0;
+   int y0;
+   float coverage[QUAD_SIZE];    /** fragment coverage for antialiasing */
+   unsigned facing:1;   /**< Front (0) or back (1) facing? */
+   unsigned prim:2;     /**< PRIM_POINT, LINE, TRI */
+};
+
+struct quad_header_inout
+{
+   unsigned mask:4;
+};
+
+struct quad_header_output
+{
+   /** colors in SOA format (rrrr, gggg, bbbb, aaaa) */
+   float color[PIPE_MAX_COLOR_BUFS][NUM_CHANNELS][QUAD_SIZE];
+   float depth[QUAD_SIZE];
+};
+
+struct quad_header {
+   struct quad_header_input input;
+   struct quad_header_inout inout;
+   struct quad_header_output output;
+
+   const struct tgsi_interp_coef *coef;
+   const struct tgsi_interp_coef *posCoef;
+
+   unsigned nr_attrs;
+};
+
+#endif /* SP_HEADERS_H */
+
diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c
new file mode 100644
index 0000000000..038ff04d4f
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_prim_setup.c
@@ -0,0 +1,190 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief A draw stage that drives our triangle setup routines from
+ * within the draw pipeline.  One of two ways to drive setup, the
+ * other being in sp_prim_vbuf.c.
+ *
+ * \author  Keith Whitwell <keith@tungstengraphics.com>
+ * \author  Brian Paul
+ */
+
+
+#include "sp_context.h"
+#include "sp_setup.h"
+#include "sp_state.h"
+#include "sp_prim_setup.h"
+#include "draw/draw_pipe.h"
+#include "draw/draw_vertex.h"
+#include "util/u_memory.h"
+
+/**
+ * Triangle setup info (derived from draw_stage).
+ * Also used for line drawing (taking some liberties).
+ */
+struct setup_stage {
+   struct draw_stage stage; /**< This must be first (base class) */
+
+   struct setup_context *setup;
+};
+
+
+
+/**
+ * Basically a cast wrapper.
+ */
+static INLINE struct setup_stage *setup_stage( struct draw_stage *stage )
+{
+   return (struct setup_stage *)stage;
+}
+
+
+typedef const float (*cptrf4)[4];
+
+static void
+do_tri(struct draw_stage *stage, struct prim_header *prim)
+{
+   struct setup_stage *setup = setup_stage( stage );
+   
+   setup_tri( setup->setup,
+              (cptrf4)prim->v[0]->data,
+              (cptrf4)prim->v[1]->data,
+              (cptrf4)prim->v[2]->data );
+}
+
+static void
+do_line(struct draw_stage *stage, struct prim_header *prim)
+{
+   struct setup_stage *setup = setup_stage( stage );
+
+   setup_line( setup->setup,
+               (cptrf4)prim->v[0]->data,
+               (cptrf4)prim->v[1]->data );
+}
+
+static void
+do_point(struct draw_stage *stage, struct prim_header *prim)
+{
+   struct setup_stage *setup = setup_stage( stage );
+
+   setup_point( setup->setup,
+                (cptrf4)prim->v[0]->data );
+}
+
+
+
+
+static void setup_begin( struct draw_stage *stage )
+{
+   struct setup_stage *setup = setup_stage(stage);
+
+   setup_prepare( setup->setup );
+
+   stage->point = do_point;
+   stage->line = do_line;
+   stage->tri = do_tri;
+}
+
+
+static void setup_first_point( struct draw_stage *stage,
+			       struct prim_header *header )
+{
+   setup_begin(stage);
+   stage->point( stage, header );
+}
+
+static void setup_first_line( struct draw_stage *stage,
+			       struct prim_header *header )
+{
+   setup_begin(stage);
+   stage->line( stage, header );
+}
+
+
+static void setup_first_tri( struct draw_stage *stage,
+			       struct prim_header *header )
+{
+   setup_begin(stage);
+   stage->tri( stage, header );
+}
+
+
+
+static void setup_flush( struct draw_stage *stage,
+			 unsigned flags )
+{
+   stage->point = setup_first_point;
+   stage->line = setup_first_line;
+   stage->tri = setup_first_tri;
+}
+
+
+static void reset_stipple_counter( struct draw_stage *stage )
+{
+}
+
+
+static void render_destroy( struct draw_stage *stage )
+{
+   struct setup_stage *ssetup = setup_stage(stage);
+   setup_destroy_context(ssetup->setup);
+   FREE( stage );
+}
+
+
+/**
+ * Create a new primitive setup/render stage.
+ */
+struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe )
+{
+   struct setup_stage *sstage = CALLOC_STRUCT(setup_stage);
+
+   sstage->setup = setup_create_context(softpipe);
+   sstage->stage.draw = softpipe->draw;
+   sstage->stage.point = setup_first_point;
+   sstage->stage.line = setup_first_line;
+   sstage->stage.tri = setup_first_tri;
+   sstage->stage.flush = setup_flush;
+   sstage->stage.reset_stipple_counter = reset_stipple_counter;
+   sstage->stage.destroy = render_destroy;
+
+   return (struct draw_stage *)sstage;
+}
+
+struct setup_context *
+sp_draw_setup_context( struct draw_stage *stage )
+{
+   struct setup_stage *ssetup = setup_stage(stage);
+   return ssetup->setup;
+}
+
+void
+sp_draw_flush( struct draw_stage *stage )
+{
+   stage->flush( stage, 0 );
+}
diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.h b/src/gallium/drivers/softpipe/sp_prim_setup.h
new file mode 100644
index 0000000000..49bdd98ed8
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_prim_setup.h
@@ -0,0 +1,85 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef SP_PRIM_SETUP_H
+#define SP_PRIM_SETUP_H
+
+
+/**
+ * vbuf is a special stage to gather the stream of triangles, lines, points
+ * together and reconstruct vertex buffers for hardware upload.
+ *
+ * First attempt, work in progress.
+ * 
+ * TODO:
+ *    - separate out vertex buffer building and primitive emit, ie >1 draw per vb.
+ *    - tell vbuf stage how to build hw vertices directly
+ *    - pass vbuf stage a buffer pointer for direct emit to agp/vram.
+ *
+ *
+ *
+ * Vertices are just an array of floats, with all the attributes
+ * packed.  We currently assume a layout like:
+ *
+ * attr[0][0..3] - window position
+ * attr[1..n][0..3] - remaining attributes.
+ *
+ * Attributes are assumed to be 4 floats wide but are packed so that
+ * all the enabled attributes run contiguously.
+ */
+
+
+struct draw_stage;
+struct softpipe_context;
+
+
+typedef void (*vbuf_draw_func)( struct pipe_context *pipe,
+                                unsigned prim,
+                                const ushort *elements,
+                                unsigned nr_elements,
+                                const void *vertex_buffer,
+                                unsigned nr_vertices );
+
+
+extern struct draw_stage *
+sp_draw_render_stage( struct softpipe_context *softpipe );
+
+extern struct setup_context *
+sp_draw_setup_context( struct draw_stage * );
+
+extern void
+sp_draw_flush( struct draw_stage * );
+
+
+extern struct draw_stage *
+sp_draw_vbuf_stage( struct draw_context *draw_context,
+                    struct pipe_context *pipe,
+                    vbuf_draw_func draw );
+
+
+#endif /* SP_PRIM_SETUP_H */
diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
new file mode 100644
index 0000000000..9cd5784e5b
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
@@ -0,0 +1,410 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * Interface between 'draw' module's output and the softpipe rasterizer/setup
+ * code.  When the 'draw' module has finished filling a vertex buffer, the
+ * draw_arrays() functions below will be called.  Loop over the vertices and
+ * call the point/line/tri setup functions.
+ *
+ * Authors
+ *  Brian Paul
+ */
+
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_prim_vbuf.h"
+#include "sp_prim_setup.h"
+#include "sp_setup.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vbuf.h"
+#include "util/u_memory.h"
+
+
+#define SP_MAX_VBUF_INDEXES 1024
+#define SP_MAX_VBUF_SIZE    4096
+
+typedef const float (*cptrf4)[4];
+
+/**
+ * Subclass of vbuf_render.
+ */
+struct softpipe_vbuf_render
+{
+   struct vbuf_render base;
+   struct softpipe_context *softpipe;
+   uint prim;
+   uint vertex_size;
+   void *vertex_buffer;
+};
+
+
+/** cast wrapper */
+static struct softpipe_vbuf_render *
+softpipe_vbuf_render(struct vbuf_render *vbr)
+{
+   return (struct softpipe_vbuf_render *) vbr;
+}
+
+
+static const struct vertex_info *
+sp_vbuf_get_vertex_info(struct vbuf_render *vbr)
+{
+   struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
+   return softpipe_get_vbuf_vertex_info(cvbr->softpipe);
+}
+
+
+static void *
+sp_vbuf_allocate_vertices(struct vbuf_render *vbr,
+                            ushort vertex_size, ushort nr_vertices)
+{
+   struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
+   assert(!cvbr->vertex_buffer);
+   cvbr->vertex_buffer = align_malloc(vertex_size * nr_vertices, 16);
+   cvbr->vertex_size = vertex_size;
+   return cvbr->vertex_buffer;
+}
+
+
+static void
+sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices,
+                           unsigned vertex_size, unsigned vertices_used)
+{
+   struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
+   align_free(vertices);
+   assert(vertices == cvbr->vertex_buffer);
+   cvbr->vertex_buffer = NULL;
+}
+
+
+static boolean
+sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim)
+{
+   struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
+
+   /* XXX: break this dependency - make setup_context live under
+    * softpipe, rename the old "setup" draw stage to something else.
+    */
+   struct setup_context *setup_ctx = sp_draw_setup_context(cvbr->softpipe->setup);
+   
+   setup_prepare( setup_ctx );
+
+
+
+   cvbr->prim = prim;
+   return TRUE;
+
+}
+
+
+static INLINE cptrf4 get_vert( const void *vertex_buffer,
+                               int index,
+                               int stride )
+{
+   return (cptrf4)((char *)vertex_buffer + index * stride);
+}
+
+
+/**
+ * draw elements / indexed primitives
+ */
+static void
+sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr)
+{
+   struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
+   struct softpipe_context *softpipe = cvbr->softpipe;
+   const unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float);
+   const void *vertex_buffer = cvbr->vertex_buffer;
+   unsigned i;
+
+   /* XXX: break this dependency - make setup_context live under
+    * softpipe, rename the old "setup" draw stage to something else.
+    */
+   struct draw_stage *setup = softpipe->setup;
+   struct setup_context *setup_ctx = sp_draw_setup_context(setup);
+
+   switch (cvbr->prim) {
+   case PIPE_PRIM_POINTS:
+      for (i = 0; i < nr; i++) {
+         setup_point( setup_ctx,
+                      get_vert(vertex_buffer, indices[i-0], stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINES:
+      for (i = 1; i < nr; i += 2) {
+         setup_line( setup_ctx,
+                     get_vert(vertex_buffer, indices[i-1], stride),
+                     get_vert(vertex_buffer, indices[i-0], stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINE_STRIP:
+      for (i = 1; i < nr; i ++) {
+         setup_line( setup_ctx,
+                     get_vert(vertex_buffer, indices[i-1], stride),
+                     get_vert(vertex_buffer, indices[i-0], stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINE_LOOP:
+      for (i = 1; i < nr; i ++) {
+         setup_line( setup_ctx,
+                     get_vert(vertex_buffer, indices[i-1], stride),
+                     get_vert(vertex_buffer, indices[i-0], stride) );
+      }
+      if (nr) {
+         setup_line( setup_ctx,
+                     get_vert(vertex_buffer, indices[nr-1], stride),
+                     get_vert(vertex_buffer, indices[0], stride) );
+      }
+      break;
+
+
+   case PIPE_PRIM_TRIANGLES:
+      for (i = 2; i < nr; i += 3) {
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, indices[i-2], stride),
+                    get_vert(vertex_buffer, indices[i-1], stride),
+                    get_vert(vertex_buffer, indices[i-0], stride));
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      for (i = 2; i < nr; i += 1) {
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, indices[i+(i&1)-2], stride),
+                    get_vert(vertex_buffer, indices[i-(i&1)-1], stride),
+                    get_vert(vertex_buffer, indices[i-0], stride));
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_FAN:
+   case PIPE_PRIM_POLYGON:
+      for (i = 2; i < nr; i += 1) {
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, indices[0], stride),
+                    get_vert(vertex_buffer, indices[i-1], stride),
+                    get_vert(vertex_buffer, indices[i-0], stride));
+      }
+      break;
+   case PIPE_PRIM_QUADS:
+      for (i = 3; i < nr; i += 4) {
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, indices[i-3], stride),
+                    get_vert(vertex_buffer, indices[i-2], stride),
+                    get_vert(vertex_buffer, indices[i-0], stride));
+
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, indices[i-2], stride),
+                    get_vert(vertex_buffer, indices[i-1], stride),
+                    get_vert(vertex_buffer, indices[i-0], stride));
+      }
+      break;
+   case PIPE_PRIM_QUAD_STRIP:
+      for (i = 3; i < nr; i += 2) {
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, indices[i-3], stride),
+                    get_vert(vertex_buffer, indices[i-2], stride),
+                    get_vert(vertex_buffer, indices[i-0], stride));
+
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, indices[i-1], stride),
+                    get_vert(vertex_buffer, indices[i-3], stride),
+                    get_vert(vertex_buffer, indices[i-0], stride));
+      }
+      break;
+   default:
+      assert(0);
+   }
+
+   /* XXX: why are we calling this???  If we had to call something, it
+    * would be a function in sp_setup.c:
+    */
+   sp_draw_flush( setup );
+}
+
+
+/**
+ * This function is hit when the draw module is working in pass-through mode.
+ * It's up to us to convert the vertex array into point/line/tri prims.
+ */
+static void
+sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr)
+{
+   struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
+   struct softpipe_context *softpipe = cvbr->softpipe;
+   const unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float);
+   const void *vertex_buffer =
+      (void *) get_vert(cvbr->vertex_buffer, start, stride);
+   unsigned i;
+
+   /* XXX: break this dependency - make setup_context live under
+    * softpipe, rename the old "setup" draw stage to something else.
+    */
+   struct draw_stage *setup = softpipe->setup;
+   struct setup_context *setup_ctx = sp_draw_setup_context(setup);
+
+   switch (cvbr->prim) {
+   case PIPE_PRIM_POINTS:
+      for (i = 0; i < nr; i++) {
+         setup_point( setup_ctx,
+                      get_vert(vertex_buffer, i-0, stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINES:
+      for (i = 1; i < nr; i += 2) {
+         setup_line( setup_ctx,
+                     get_vert(vertex_buffer, i-1, stride),
+                     get_vert(vertex_buffer, i-0, stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINE_STRIP:
+      for (i = 1; i < nr; i ++) {
+         setup_line( setup_ctx,
+                     get_vert(vertex_buffer, i-1, stride),
+                     get_vert(vertex_buffer, i-0, stride) );
+      }
+      break;
+
+   case PIPE_PRIM_LINE_LOOP:
+      for (i = 1; i < nr; i ++) {
+         setup_line( setup_ctx,
+                     get_vert(vertex_buffer, i-1, stride),
+                     get_vert(vertex_buffer, i-0, stride) );
+      }
+      if (nr) {
+         setup_line( setup_ctx,
+                     get_vert(vertex_buffer, nr-1, stride),
+                     get_vert(vertex_buffer, 0, stride) );
+      }
+      break;
+
+
+   case PIPE_PRIM_TRIANGLES:
+      for (i = 2; i < nr; i += 3) {
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, i-2, stride),
+                    get_vert(vertex_buffer, i-1, stride),
+                    get_vert(vertex_buffer, i-0, stride));
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      for (i = 2; i < nr; i += 1) {
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, i+(i&1)-2, stride),
+                    get_vert(vertex_buffer, i-(i&1)-1, stride),
+                    get_vert(vertex_buffer, i-0, stride));
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_FAN:
+   case PIPE_PRIM_POLYGON:
+      for (i = 2; i < nr; i += 1) {
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, 0, stride),
+                    get_vert(vertex_buffer, i-1, stride),
+                    get_vert(vertex_buffer, i-0, stride));
+      }
+      break;
+   case PIPE_PRIM_QUADS:
+      for (i = 3; i < nr; i += 4) {
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, i-3, stride),
+                    get_vert(vertex_buffer, i-2, stride),
+                    get_vert(vertex_buffer, i-0, stride));
+
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, i-2, stride),
+                    get_vert(vertex_buffer, i-1, stride),
+                    get_vert(vertex_buffer, i-0, stride));
+      }
+      break;
+   case PIPE_PRIM_QUAD_STRIP:
+      for (i = 3; i < nr; i += 2) {
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, i-3, stride),
+                    get_vert(vertex_buffer, i-2, stride),
+                    get_vert(vertex_buffer, i-0, stride));
+
+         setup_tri( setup_ctx,
+                    get_vert(vertex_buffer, i-1, stride),
+                    get_vert(vertex_buffer, i-3, stride),
+                    get_vert(vertex_buffer, i-0, stride));
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+
+static void
+sp_vbuf_destroy(struct vbuf_render *vbr)
+{
+   struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
+   cvbr->softpipe->vbuf_render = NULL;
+   FREE(cvbr);
+}
+
+
+/**
+ * Initialize the post-transform vertex buffer information for the given
+ * context.
+ */
+void
+sp_init_vbuf(struct softpipe_context *sp)
+{
+   assert(sp->draw);
+
+   sp->vbuf_render = CALLOC_STRUCT(softpipe_vbuf_render);
+
+   sp->vbuf_render->base.max_indices = SP_MAX_VBUF_INDEXES;
+   sp->vbuf_render->base.max_vertex_buffer_bytes = SP_MAX_VBUF_SIZE;
+
+   sp->vbuf_render->base.get_vertex_info = sp_vbuf_get_vertex_info;
+   sp->vbuf_render->base.allocate_vertices = sp_vbuf_allocate_vertices;
+   sp->vbuf_render->base.set_primitive = sp_vbuf_set_primitive;
+   sp->vbuf_render->base.draw = sp_vbuf_draw;
+   sp->vbuf_render->base.draw_arrays = sp_vbuf_draw_arrays;
+   sp->vbuf_render->base.release_vertices = sp_vbuf_release_vertices;
+   sp->vbuf_render->base.destroy = sp_vbuf_destroy;
+
+   sp->vbuf_render->softpipe = sp;
+
+   sp->vbuf = draw_vbuf_stage(sp->draw, &sp->vbuf_render->base);
+
+   draw_set_rasterize_stage(sp->draw, sp->vbuf);
+
+   draw_set_render(sp->draw, &sp->vbuf_render->base);
+}
diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.h b/src/gallium/drivers/softpipe/sp_prim_vbuf.h
new file mode 100644
index 0000000000..1de9cc2a89
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef SP_VBUF_H
+#define SP_VBUF_H
+
+
+struct softpipe_context;
+
+extern void
+sp_init_vbuf(struct softpipe_context *softpipe);
+
+
+#endif /* SP_VBUF_H */
diff --git a/src/gallium/drivers/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad.c
new file mode 100644
index 0000000000..892ef87ee9
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad.c
@@ -0,0 +1,118 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "pipe/p_shader_tokens.h"
+
+static void
+sp_push_quad_first(
+   struct softpipe_context *sp,
+   struct quad_stage *quad,
+   uint i )
+{
+   quad->next = sp->quad[i].first;
+   sp->quad[i].first = quad;
+}
+
+static void
+sp_build_depth_stencil(
+   struct softpipe_context *sp,
+   uint i )
+{
+   if (sp->depth_stencil->stencil[0].enabled ||
+       sp->depth_stencil->stencil[1].enabled) {
+      sp_push_quad_first( sp, sp->quad[i].stencil_test, i );
+   }
+   else if (sp->depth_stencil->depth.enabled &&
+            sp->framebuffer.zsbuf) {
+      sp_push_quad_first( sp, sp->quad[i].depth_test, i );
+   }
+}
+
+void
+sp_build_quad_pipeline(struct softpipe_context *sp)
+{
+   uint i;
+
+   boolean early_depth_test =
+               sp->depth_stencil->depth.enabled &&
+               sp->framebuffer.zsbuf &&
+               !sp->depth_stencil->alpha.enabled &&
+               !sp->fs->info.uses_kill &&
+               !sp->fs->info.writes_z;
+
+   /* build up the pipeline in reverse order... */
+   for (i = 0; i < SP_NUM_QUAD_THREADS; i++) {
+      sp->quad[i].first = sp->quad[i].output;
+
+      if (sp->blend->colormask != 0xf) {
+         sp_push_quad_first( sp, sp->quad[i].colormask, i );
+      }
+
+      if (sp->blend->blend_enable ||
+          sp->blend->logicop_enable) {
+         sp_push_quad_first( sp, sp->quad[i].blend, i );
+      }
+
+      if (sp->depth_stencil->depth.occlusion_count) {
+         sp_push_quad_first( sp, sp->quad[i].occlusion, i );
+      }
+
+      if (sp->rasterizer->poly_smooth ||
+          sp->rasterizer->line_smooth ||
+          sp->rasterizer->point_smooth) {
+         sp_push_quad_first( sp, sp->quad[i].coverage, i );
+      }
+
+      if (!early_depth_test) {
+         sp_build_depth_stencil( sp, i );
+      }
+
+      if (sp->depth_stencil->alpha.enabled) {
+         sp_push_quad_first( sp, sp->quad[i].alpha_test, i );
+      }
+
+      /* XXX always enable shader? */
+      if (1) {
+         sp_push_quad_first( sp, sp->quad[i].shade, i );
+      }
+
+      if (early_depth_test) {
+         sp_build_depth_stencil( sp, i );
+         sp_push_quad_first( sp, sp->quad[i].earlyz, i );
+      }
+
+#if !USE_DRAW_STAGE_PSTIPPLE
+      if (sp->rasterizer->poly_stipple_enable) {
+         sp_push_quad_first( sp, sp->quad[i].polygon_stipple, i );
+      }
+#endif
+   }
+}
+
diff --git a/src/gallium/drivers/softpipe/sp_quad.h b/src/gallium/drivers/softpipe/sp_quad.h
new file mode 100644
index 0000000000..08513cb95f
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad.h
@@ -0,0 +1,69 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef SP_QUAD_H
+#define SP_QUAD_H
+
+
+struct softpipe_context;
+struct quad_header;
+
+
+struct quad_stage {
+   struct softpipe_context *softpipe;
+
+   struct quad_stage *next;
+
+   void (*begin)(struct quad_stage *qs);
+
+   /** the stage action */
+   void (*run)(struct quad_stage *qs, struct quad_header *quad);
+
+   void (*destroy)(struct quad_stage *qs);
+};
+
+
+struct quad_stage *sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe );
+struct quad_stage *sp_quad_earlyz_stage( struct softpipe_context *softpipe );
+struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe );
+struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe );
+struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe );
+struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe );
+struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe );
+struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe );
+struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe );
+struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe );
+struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe );
+
+void sp_build_quad_pipeline(struct softpipe_context *sp);
+
+void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad);
+
+#endif /* SP_QUAD_H */
diff --git a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c
new file mode 100644
index 0000000000..85c9f037a3
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c
@@ -0,0 +1,108 @@
+
+/**
+ * quad alpha test
+ */
+
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_quad.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+
+
+static void
+alpha_test_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   struct softpipe_context *softpipe = qs->softpipe;
+   const float ref = softpipe->depth_stencil->alpha.ref_value;
+   unsigned passMask = 0x0, j;
+   const uint cbuf = 0; /* only output[0].alpha is tested */
+   const float *aaaa = quad->output.color[cbuf][3];
+
+   switch (softpipe->depth_stencil->alpha.func) {
+   case PIPE_FUNC_NEVER:
+      break;
+   case PIPE_FUNC_LESS:
+      /*
+       * If mask were an array [4] we could do this SIMD-style:
+       * passMask = (quad->outputs.color[0][3] <= vec4(ref));
+       */
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (aaaa[j] < ref) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_EQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (aaaa[j] == ref) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_LEQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (aaaa[j] <= ref) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_GREATER:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (aaaa[j] > ref) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_NOTEQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (aaaa[j] != ref) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_GEQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (aaaa[j] >= ref) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_ALWAYS:
+      passMask = MASK_ALL;
+      break;
+   default:
+      assert(0);
+   }
+
+   quad->inout.mask &= passMask;
+
+   if (quad->inout.mask)
+      qs->next->run(qs->next, quad);
+}
+
+
+static void alpha_test_begin(struct quad_stage *qs)
+{
+   qs->next->begin(qs->next);
+}
+
+
+static void alpha_test_destroy(struct quad_stage *qs)
+{
+   FREE( qs );
+}
+
+
+struct quad_stage *
+sp_quad_alpha_test_stage( struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
+
+   stage->softpipe = softpipe;
+   stage->begin = alpha_test_begin;
+   stage->run = alpha_test_quad;
+   stage->destroy = alpha_test_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c
new file mode 100644
index 0000000000..fb1d430a4f
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_blend.c
@@ -0,0 +1,759 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * quad blending
+ * \author Brian Paul
+ */
+
+#include "pipe/p_defines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_surface.h"
+#include "sp_tile_cache.h"
+#include "sp_quad.h"
+
+
+#define VEC4_COPY(DST, SRC) \
+do { \
+    DST[0] = SRC[0]; \
+    DST[1] = SRC[1]; \
+    DST[2] = SRC[2]; \
+    DST[3] = SRC[3]; \
+} while(0)
+
+#define VEC4_SCALAR(DST, SRC) \
+do { \
+    DST[0] = SRC; \
+    DST[1] = SRC; \
+    DST[2] = SRC; \
+    DST[3] = SRC; \
+} while(0)
+
+#define VEC4_ADD(R, A, B) \
+do { \
+   R[0] = A[0] + B[0]; \
+   R[1] = A[1] + B[1]; \
+   R[2] = A[2] + B[2]; \
+   R[3] = A[3] + B[3]; \
+} while (0)
+
+#define VEC4_SUB(R, A, B) \
+do { \
+   R[0] = A[0] - B[0]; \
+   R[1] = A[1] - B[1]; \
+   R[2] = A[2] - B[2]; \
+   R[3] = A[3] - B[3]; \
+} while (0)
+
+#define VEC4_MUL(R, A, B) \
+do { \
+   R[0] = A[0] * B[0]; \
+   R[1] = A[1] * B[1]; \
+   R[2] = A[2] * B[2]; \
+   R[3] = A[3] * B[3]; \
+} while (0)
+
+#define VEC4_MIN(R, A, B) \
+do { \
+   R[0] = (A[0] < B[0]) ? A[0] : B[0]; \
+   R[1] = (A[1] < B[1]) ? A[1] : B[1]; \
+   R[2] = (A[2] < B[2]) ? A[2] : B[2]; \
+   R[3] = (A[3] < B[3]) ? A[3] : B[3]; \
+} while (0)
+
+#define VEC4_MAX(R, A, B) \
+do { \
+   R[0] = (A[0] > B[0]) ? A[0] : B[0]; \
+   R[1] = (A[1] > B[1]) ? A[1] : B[1]; \
+   R[2] = (A[2] > B[2]) ? A[2] : B[2]; \
+   R[3] = (A[3] > B[3]) ? A[3] : B[3]; \
+} while (0)
+
+
+
+static void
+logicop_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   struct softpipe_context *softpipe = qs->softpipe;
+   uint cbuf;
+
+   /* loop over colorbuffer outputs */
+   for (cbuf = 0; cbuf < softpipe->framebuffer.nr_cbufs; cbuf++) {
+      float dest[4][QUAD_SIZE];
+      ubyte src[4][4], dst[4][4], res[4][4];
+      uint *src4 = (uint *) src;
+      uint *dst4 = (uint *) dst;
+      uint *res4 = (uint *) res;
+      struct softpipe_cached_tile *
+         tile = sp_get_cached_tile(softpipe,
+                                   softpipe->cbuf_cache[cbuf],
+                                   quad->input.x0, quad->input.y0);
+      float (*quadColor)[4] = quad->output.color[cbuf];
+      uint i, j;
+
+      /* get/swizzle dest colors */
+      for (j = 0; j < QUAD_SIZE; j++) {
+         int x = (quad->input.x0 & (TILE_SIZE-1)) + (j & 1);
+         int y = (quad->input.y0 & (TILE_SIZE-1)) + (j >> 1);
+         for (i = 0; i < 4; i++) {
+            dest[i][j] = tile->data.color[y][x][i];
+         }
+      }
+
+      /* convert to ubyte */
+      for (j = 0; j < 4; j++) { /* loop over R,G,B,A channels */
+         dst[j][0] = float_to_ubyte(dest[j][0]); /* P0 */
+         dst[j][1] = float_to_ubyte(dest[j][1]); /* P1 */
+         dst[j][2] = float_to_ubyte(dest[j][2]); /* P2 */
+         dst[j][3] = float_to_ubyte(dest[j][3]); /* P3 */
+
+         src[j][0] = float_to_ubyte(quadColor[j][0]); /* P0 */
+         src[j][1] = float_to_ubyte(quadColor[j][1]); /* P1 */
+         src[j][2] = float_to_ubyte(quadColor[j][2]); /* P2 */
+         src[j][3] = float_to_ubyte(quadColor[j][3]); /* P3 */
+      }
+
+      switch (softpipe->blend->logicop_func) {
+      case PIPE_LOGICOP_CLEAR:
+         for (j = 0; j < 4; j++)
+            res4[j] = 0;
+         break;
+      case PIPE_LOGICOP_NOR:
+         for (j = 0; j < 4; j++)
+            res4[j] = ~(src4[j] | dst4[j]);
+         break;
+      case PIPE_LOGICOP_AND_INVERTED:
+         for (j = 0; j < 4; j++)
+            res4[j] = ~src4[j] & dst4[j];
+         break;
+      case PIPE_LOGICOP_COPY_INVERTED:
+         for (j = 0; j < 4; j++)
+            res4[j] = ~src4[j];
+         break;
+      case PIPE_LOGICOP_AND_REVERSE:
+         for (j = 0; j < 4; j++)
+            res4[j] = src4[j] & ~dst4[j];
+         break;
+      case PIPE_LOGICOP_INVERT:
+         for (j = 0; j < 4; j++)
+            res4[j] = ~dst4[j];
+         break;
+      case PIPE_LOGICOP_XOR:
+         for (j = 0; j < 4; j++)
+            res4[j] = dst4[j] ^ src4[j];
+         break;
+      case PIPE_LOGICOP_NAND:
+         for (j = 0; j < 4; j++)
+            res4[j] = ~(src4[j] & dst4[j]);
+         break;
+      case PIPE_LOGICOP_AND:
+         for (j = 0; j < 4; j++)
+            res4[j] = src4[j] & dst4[j];
+         break;
+      case PIPE_LOGICOP_EQUIV:
+         for (j = 0; j < 4; j++)
+            res4[j] = ~(src4[j] ^ dst4[j]);
+         break;
+      case PIPE_LOGICOP_NOOP:
+         for (j = 0; j < 4; j++)
+            res4[j] = dst4[j];
+         break;
+      case PIPE_LOGICOP_OR_INVERTED:
+         for (j = 0; j < 4; j++)
+            res4[j] = ~src4[j] | dst4[j];
+         break;
+      case PIPE_LOGICOP_COPY:
+         for (j = 0; j < 4; j++)
+            res4[j] = src4[j];
+         break;
+      case PIPE_LOGICOP_OR_REVERSE:
+         for (j = 0; j < 4; j++)
+            res4[j] = src4[j] | ~dst4[j];
+         break;
+      case PIPE_LOGICOP_OR:
+         for (j = 0; j < 4; j++)
+            res4[j] = src4[j] | dst4[j];
+         break;
+      case PIPE_LOGICOP_SET:
+         for (j = 0; j < 4; j++)
+            res4[j] = ~0;
+         break;
+      default:
+         assert(0);
+      }
+
+      for (j = 0; j < 4; j++) {
+         quadColor[j][0] = ubyte_to_float(res[j][0]);
+         quadColor[j][1] = ubyte_to_float(res[j][1]);
+         quadColor[j][2] = ubyte_to_float(res[j][2]);
+         quadColor[j][3] = ubyte_to_float(res[j][3]);
+      }
+   }
+
+   /* pass quad to next stage */
+   qs->next->run(qs->next, quad);
+}
+
+
+
+
+static void
+blend_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   static const float zero[4] = { 0, 0, 0, 0 };
+   static const float one[4] = { 1, 1, 1, 1 };
+
+   struct softpipe_context *softpipe = qs->softpipe;
+   uint cbuf;
+
+   if (softpipe->blend->logicop_enable) {
+      logicop_quad(qs, quad);
+      return;
+   }
+
+   /* loop over colorbuffer outputs */
+   for (cbuf = 0; cbuf < softpipe->framebuffer.nr_cbufs; cbuf++) {
+      float source[4][QUAD_SIZE], dest[4][QUAD_SIZE];
+      struct softpipe_cached_tile *tile
+         = sp_get_cached_tile(softpipe,
+                              softpipe->cbuf_cache[cbuf],
+                              quad->input.x0, quad->input.y0);
+      float (*quadColor)[4] = quad->output.color[cbuf];
+      uint i, j;
+
+      /* get/swizzle dest colors */
+      for (j = 0; j < QUAD_SIZE; j++) {
+         int x = (quad->input.x0 & (TILE_SIZE-1)) + (j & 1);
+         int y = (quad->input.y0 & (TILE_SIZE-1)) + (j >> 1);
+         for (i = 0; i < 4; i++) {
+            dest[i][j] = tile->data.color[y][x][i];
+         }
+      }
+
+      /*
+       * Compute src/first term RGB
+       */
+      switch (softpipe->blend->rgb_src_factor) {
+      case PIPE_BLENDFACTOR_ONE:
+         VEC4_COPY(source[0], quadColor[0]); /* R */
+         VEC4_COPY(source[1], quadColor[1]); /* G */
+         VEC4_COPY(source[2], quadColor[2]); /* B */
+         break;
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+         VEC4_MUL(source[0], quadColor[0], quadColor[0]); /* R */
+         VEC4_MUL(source[1], quadColor[1], quadColor[1]); /* G */
+         VEC4_MUL(source[2], quadColor[2], quadColor[2]); /* B */
+         break;
+      case PIPE_BLENDFACTOR_SRC_ALPHA:
+         {
+            const float *alpha = quadColor[3];
+            VEC4_MUL(source[0], quadColor[0], alpha); /* R */
+            VEC4_MUL(source[1], quadColor[1], alpha); /* G */
+            VEC4_MUL(source[2], quadColor[2], alpha); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_DST_COLOR:
+         VEC4_MUL(source[0], quadColor[0], dest[0]); /* R */
+         VEC4_MUL(source[1], quadColor[1], dest[1]); /* G */
+         VEC4_MUL(source[2], quadColor[2], dest[2]); /* B */
+         break;
+      case PIPE_BLENDFACTOR_DST_ALPHA:
+         {
+            const float *alpha = dest[3];
+            VEC4_MUL(source[0], quadColor[0], alpha); /* R */
+            VEC4_MUL(source[1], quadColor[1], alpha); /* G */
+            VEC4_MUL(source[2], quadColor[2], alpha); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+         {
+            const float *alpha = quadColor[3];
+            float diff[4], temp[4];
+            VEC4_SUB(diff, one, dest[3]);
+            VEC4_MIN(temp, alpha, diff);
+            VEC4_MUL(source[0], quadColor[0], temp); /* R */
+            VEC4_MUL(source[1], quadColor[1], temp); /* G */
+            VEC4_MUL(source[2], quadColor[2], temp); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+         {
+            float comp[4];
+            VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */
+            VEC4_MUL(source[0], quadColor[0], comp); /* R */
+            VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */
+            VEC4_MUL(source[1], quadColor[1], comp); /* G */
+            VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */
+            VEC4_MUL(source[2], quadColor[2], comp); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_CONST_ALPHA:
+         {
+            float alpha[4];
+            VEC4_SCALAR(alpha, softpipe->blend_color.color[3]);
+            VEC4_MUL(source[0], quadColor[0], alpha); /* R */
+            VEC4_MUL(source[1], quadColor[1], alpha); /* G */
+            VEC4_MUL(source[2], quadColor[2], alpha); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_SRC1_COLOR:
+         assert(0); /* to do */
+         break;
+      case PIPE_BLENDFACTOR_SRC1_ALPHA:
+         assert(0); /* to do */
+         break;
+      case PIPE_BLENDFACTOR_ZERO:
+         VEC4_COPY(source[0], zero); /* R */
+         VEC4_COPY(source[1], zero); /* G */
+         VEC4_COPY(source[2], zero); /* B */
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+         {
+            float inv_comp[4];
+            VEC4_SUB(inv_comp, one, quadColor[0]); /* R */
+            VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */
+            VEC4_SUB(inv_comp, one, quadColor[1]); /* G */
+            VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */
+            VEC4_SUB(inv_comp, one, quadColor[2]); /* B */
+            VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+         {
+            float inv_alpha[4];
+            VEC4_SUB(inv_alpha, one, quadColor[3]);
+            VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */
+            VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */
+            VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+         {
+            float inv_alpha[4];
+            VEC4_SUB(inv_alpha, one, dest[3]);
+            VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */
+            VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */
+            VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_DST_COLOR:
+         {
+            float inv_comp[4];
+            VEC4_SUB(inv_comp, one, dest[0]); /* R */
+            VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */
+            VEC4_SUB(inv_comp, one, dest[1]); /* G */
+            VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */
+            VEC4_SUB(inv_comp, one, dest[2]); /* B */
+            VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+         {
+            float inv_comp[4];
+            /* R */
+            VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]);
+            VEC4_MUL(source[0], quadColor[0], inv_comp);
+            /* G */
+            VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]);
+            VEC4_MUL(source[1], quadColor[1], inv_comp);
+            /* B */
+            VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]);
+            VEC4_MUL(source[2], quadColor[2], inv_comp);
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+         {
+            float inv_alpha[4];
+            VEC4_SCALAR(inv_alpha, 1.0f - softpipe->blend_color.color[3]);
+            VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */
+            VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */
+            VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+         assert(0); /* to do */
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+         assert(0); /* to do */
+         break;
+      default:
+         assert(0);
+      }
+
+      /*
+       * Compute src/first term A
+       */
+      switch (softpipe->blend->alpha_src_factor) {
+      case PIPE_BLENDFACTOR_ONE:
+         VEC4_COPY(source[3], quadColor[3]); /* A */
+         break;
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_SRC_ALPHA:
+         {
+            const float *alpha = quadColor[3];
+            VEC4_MUL(source[3], quadColor[3], alpha); /* A */
+         }
+         break;
+      case PIPE_BLENDFACTOR_DST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_DST_ALPHA:
+         VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */
+         break;
+      case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+         /* multiply alpha by 1.0 */
+         VEC4_COPY(source[3], quadColor[3]); /* A */
+         break;
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_CONST_ALPHA:
+         {
+            float comp[4];
+            VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */
+            VEC4_MUL(source[3], quadColor[3], comp); /* A */
+         }
+         break;
+      case PIPE_BLENDFACTOR_ZERO:
+         VEC4_COPY(source[3], zero); /* A */
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+         {
+            float inv_alpha[4];
+            VEC4_SUB(inv_alpha, one, quadColor[3]);
+            VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_DST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+         {
+            float inv_alpha[4];
+            VEC4_SUB(inv_alpha, one, dest[3]);
+            VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+         {
+            float inv_comp[4];
+            /* A */
+            VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]);
+            VEC4_MUL(source[3], quadColor[3], inv_comp);
+         }
+         break;
+      default:
+         assert(0);
+      }
+
+
+      /*
+       * Compute dest/second term RGB
+       */
+      switch (softpipe->blend->rgb_dst_factor) {
+      case PIPE_BLENDFACTOR_ONE:
+         /* dest = dest * 1   NO-OP, leave dest as-is */
+         break;
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+         VEC4_MUL(dest[0], dest[0], quadColor[0]); /* R */
+         VEC4_MUL(dest[1], dest[1], quadColor[1]); /* G */
+         VEC4_MUL(dest[2], dest[2], quadColor[2]); /* B */
+         break;
+      case PIPE_BLENDFACTOR_SRC_ALPHA:
+         VEC4_MUL(dest[0], dest[0], quadColor[3]); /* R * A */
+         VEC4_MUL(dest[1], dest[1], quadColor[3]); /* G * A */
+         VEC4_MUL(dest[2], dest[2], quadColor[3]); /* B * A */
+         break;
+      case PIPE_BLENDFACTOR_DST_ALPHA:
+         VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */
+         VEC4_MUL(dest[1], dest[1], dest[3]); /* G * A */
+         VEC4_MUL(dest[2], dest[2], dest[3]); /* B * A */
+         break;
+      case PIPE_BLENDFACTOR_DST_COLOR:
+         VEC4_MUL(dest[0], dest[0], dest[0]); /* R */
+         VEC4_MUL(dest[1], dest[1], dest[1]); /* G */
+         VEC4_MUL(dest[2], dest[2], dest[2]); /* B */
+         break;
+      case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+         assert(0); /* illegal */
+         break;
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+         {
+            float comp[4];
+            VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */
+            VEC4_MUL(dest[0], dest[0], comp); /* R */
+            VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */
+            VEC4_MUL(dest[1], dest[1], comp); /* G */
+            VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */
+            VEC4_MUL(dest[2], dest[2], comp); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_CONST_ALPHA:
+         {
+            float comp[4];
+            VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */
+            VEC4_MUL(dest[0], dest[0], comp); /* R */
+            VEC4_MUL(dest[1], dest[1], comp); /* G */
+            VEC4_MUL(dest[2], dest[2], comp); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_ZERO:
+         VEC4_COPY(dest[0], zero); /* R */
+         VEC4_COPY(dest[1], zero); /* G */
+         VEC4_COPY(dest[2], zero); /* B */
+         break;
+      case PIPE_BLENDFACTOR_SRC1_COLOR:
+      case PIPE_BLENDFACTOR_SRC1_ALPHA:
+         /* XXX what are these? */
+         assert(0);
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+         {
+            float inv_comp[4];
+            VEC4_SUB(inv_comp, one, quadColor[0]); /* R */
+            VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */
+            VEC4_SUB(inv_comp, one, quadColor[1]); /* G */
+            VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */
+            VEC4_SUB(inv_comp, one, quadColor[2]); /* B */
+            VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+         {
+            float one_minus_alpha[QUAD_SIZE];
+            VEC4_SUB(one_minus_alpha, one, quadColor[3]);
+            VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */
+            VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */
+            VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+         {
+            float inv_comp[4];
+            VEC4_SUB(inv_comp, one, dest[3]); /* A */
+            VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */
+            VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */
+            VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_DST_COLOR:
+         {
+            float inv_comp[4];
+            VEC4_SUB(inv_comp, one, dest[0]); /* R */
+            VEC4_MUL(dest[0], dest[0], inv_comp); /* R */
+            VEC4_SUB(inv_comp, one, dest[1]); /* G */
+            VEC4_MUL(dest[1], dest[1], inv_comp); /* G */
+            VEC4_SUB(inv_comp, one, dest[2]); /* B */
+            VEC4_MUL(dest[2], dest[2], inv_comp); /* B */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+         {
+            float inv_comp[4];
+            /* R */
+            VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]);
+            VEC4_MUL(dest[0], dest[0], inv_comp);
+            /* G */
+            VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]);
+            VEC4_MUL(dest[1], dest[1], inv_comp);
+            /* B */
+            VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]);
+            VEC4_MUL(dest[2], dest[2], inv_comp);
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+         {
+            float inv_comp[4];
+            VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]);
+            VEC4_MUL(dest[0], dest[0], inv_comp);
+            VEC4_MUL(dest[1], dest[1], inv_comp);
+            VEC4_MUL(dest[2], dest[2], inv_comp);
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+      case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+         /* XXX what are these? */
+         assert(0);
+         break;
+      default:
+         assert(0);
+      }
+
+      /*
+       * Compute dest/second term A
+       */
+      switch (softpipe->blend->alpha_dst_factor) {
+      case PIPE_BLENDFACTOR_ONE:
+         /* dest = dest * 1   NO-OP, leave dest as-is */
+         break;
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_SRC_ALPHA:
+         VEC4_MUL(dest[3], dest[3], quadColor[3]); /* A * A */
+         break;
+      case PIPE_BLENDFACTOR_DST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_DST_ALPHA:
+         VEC4_MUL(dest[3], dest[3], dest[3]); /* A */
+         break;
+      case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+         assert(0); /* illegal */
+         break;
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_CONST_ALPHA:
+         {
+            float comp[4];
+            VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */
+            VEC4_MUL(dest[3], dest[3], comp); /* A */
+         }
+         break;
+      case PIPE_BLENDFACTOR_ZERO:
+         VEC4_COPY(dest[3], zero); /* A */
+         break;
+      case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+         {
+            float one_minus_alpha[QUAD_SIZE];
+            VEC4_SUB(one_minus_alpha, one, quadColor[3]);
+            VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_DST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+         {
+            float inv_comp[4];
+            VEC4_SUB(inv_comp, one, dest[3]); /* A */
+            VEC4_MUL(dest[3], inv_comp, dest[3]); /* A */
+         }
+         break;
+      case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+         /* fall-through */
+      case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+         {
+            float inv_comp[4];
+            VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]);
+            VEC4_MUL(dest[3], dest[3], inv_comp);
+         }
+         break;
+      default:
+         assert(0);
+      }
+
+      /*
+       * Combine RGB terms
+       */
+      switch (softpipe->blend->rgb_func) {
+      case PIPE_BLEND_ADD:
+         VEC4_ADD(quadColor[0], source[0], dest[0]); /* R */
+         VEC4_ADD(quadColor[1], source[1], dest[1]); /* G */
+         VEC4_ADD(quadColor[2], source[2], dest[2]); /* B */
+         break;
+      case PIPE_BLEND_SUBTRACT:
+         VEC4_SUB(quadColor[0], source[0], dest[0]); /* R */
+         VEC4_SUB(quadColor[1], source[1], dest[1]); /* G */
+         VEC4_SUB(quadColor[2], source[2], dest[2]); /* B */
+         break;
+      case PIPE_BLEND_REVERSE_SUBTRACT:
+         VEC4_SUB(quadColor[0], dest[0], source[0]); /* R */
+         VEC4_SUB(quadColor[1], dest[1], source[1]); /* G */
+         VEC4_SUB(quadColor[2], dest[2], source[2]); /* B */
+         break;
+      case PIPE_BLEND_MIN:
+         VEC4_MIN(quadColor[0], source[0], dest[0]); /* R */
+         VEC4_MIN(quadColor[1], source[1], dest[1]); /* G */
+         VEC4_MIN(quadColor[2], source[2], dest[2]); /* B */
+         break;
+      case PIPE_BLEND_MAX:
+         VEC4_MAX(quadColor[0], source[0], dest[0]); /* R */
+         VEC4_MAX(quadColor[1], source[1], dest[1]); /* G */
+         VEC4_MAX(quadColor[2], source[2], dest[2]); /* B */
+         break;
+      default:
+         assert(0);
+      }
+
+      /*
+       * Combine A terms
+       */
+      switch (softpipe->blend->alpha_func) {
+      case PIPE_BLEND_ADD:
+         VEC4_ADD(quadColor[3], source[3], dest[3]); /* A */
+         break;
+      case PIPE_BLEND_SUBTRACT:
+         VEC4_SUB(quadColor[3], source[3], dest[3]); /* A */
+         break;
+      case PIPE_BLEND_REVERSE_SUBTRACT:
+         VEC4_SUB(quadColor[3], dest[3], source[3]); /* A */
+         break;
+      case PIPE_BLEND_MIN:
+         VEC4_MIN(quadColor[3], source[3], dest[3]); /* A */
+         break;
+      case PIPE_BLEND_MAX:
+         VEC4_MAX(quadColor[3], source[3], dest[3]); /* A */
+         break;
+      default:
+         assert(0);
+      }
+
+   } /* cbuf loop */
+
+   /* pass blended quad to next stage */
+   qs->next->run(qs->next, quad);
+}
+
+
+static void blend_begin(struct quad_stage *qs)
+{
+   qs->next->begin(qs->next);
+}
+
+
+static void blend_destroy(struct quad_stage *qs)
+{
+   FREE( qs );
+}
+
+
+struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
+
+   stage->softpipe = softpipe;
+   stage->begin = blend_begin;
+   stage->run = blend_quad;
+   stage->destroy = blend_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_quad_bufloop.c b/src/gallium/drivers/softpipe/sp_quad_bufloop.c
new file mode 100644
index 0000000000..d7d6a6974d
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_bufloop.c
@@ -0,0 +1,74 @@
+
+#include "util/u_memory.h"
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_surface.h"
+#include "sp_quad.h"
+
+
+/**
+ * Loop over colorbuffers, passing quad to next stage each time.
+ */
+static void
+cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   struct softpipe_context *softpipe = qs->softpipe;
+   float tmp[PIPE_MAX_COLOR_BUFS][4][QUAD_SIZE];
+   unsigned i;
+
+   assert(sizeof(quad->outputs.color) == sizeof(tmp));
+   assert(softpipe->framebuffer.nr_cbufs <= PIPE_MAX_COLOR_BUFS);
+
+   /* make copy of original colors since they can get modified
+    * by blending and masking.
+    * XXX we won't have to do this if the fragment program actually emits
+    * N separate colors and we're drawing to N color buffers (MRT).
+    * But if we emitted one color and glDrawBuffer(GL_FRONT_AND_BACK) is
+    * in effect, we need to save/restore colors like this.
+    */
+   memcpy(tmp, quad->outputs.color, sizeof(tmp));
+
+   for (i = 0; i < softpipe->framebuffer.nr_cbufs; i++) {
+      /* set current cbuffer */
+#if 0 /* obsolete & going away */
+      softpipe->current_cbuf = i;
+#endif
+
+      /* pass blended quad to next stage */
+      qs->next->run(qs->next, quad);
+
+      /* restore quad's colors for next buffer */
+      memcpy(quad->outputs.color, tmp, sizeof(tmp));
+   }
+}
+
+
+static void cbuf_loop_begin(struct quad_stage *qs)
+{
+   qs->next->begin(qs->next);
+}
+
+
+static void cbuf_loop_destroy(struct quad_stage *qs)
+{
+   FREE( qs );
+}
+
+
+/**
+ * Create the colorbuffer loop stage.
+ * This is used to implement multiple render targets and GL_FRONT_AND_BACK
+ * rendering.
+ */
+struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
+
+   stage->softpipe = softpipe;
+   stage->begin = cbuf_loop_begin;
+   stage->run = cbuf_loop_quad;
+   stage->destroy = cbuf_loop_destroy;
+
+   return stage;
+}
+
diff --git a/src/gallium/drivers/softpipe/sp_quad_colormask.c b/src/gallium/drivers/softpipe/sp_quad_colormask.c
new file mode 100644
index 0000000000..563c2fc739
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_colormask.c
@@ -0,0 +1,116 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief  quad colormask stage
+ * \author Brian Paul
+ */
+
+#include "pipe/p_defines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_surface.h"
+#include "sp_quad.h"
+#include "sp_tile_cache.h"
+
+
+
+/**
+ * XXX colormask could be rolled into blending...
+ */
+static void
+colormask_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   struct softpipe_context *softpipe = qs->softpipe;
+   uint cbuf;
+
+   /* loop over colorbuffer outputs */
+   for (cbuf = 0; cbuf < softpipe->framebuffer.nr_cbufs; cbuf++) {
+      float dest[4][QUAD_SIZE];
+      struct softpipe_cached_tile *tile
+         = sp_get_cached_tile(softpipe,
+                              softpipe->cbuf_cache[cbuf],
+                              quad->input.x0, quad->input.y0);
+      float (*quadColor)[4] = quad->output.color[cbuf];
+      uint i, j;
+
+      /* get/swizzle dest colors */
+      for (j = 0; j < QUAD_SIZE; j++) {
+         int x = (quad->input.x0 & (TILE_SIZE-1)) + (j & 1);
+         int y = (quad->input.y0 & (TILE_SIZE-1)) + (j >> 1);
+         for (i = 0; i < 4; i++) {
+            dest[i][j] = tile->data.color[y][x][i];
+         }
+      }
+
+      /* R */
+      if (!(softpipe->blend->colormask & PIPE_MASK_R))
+          COPY_4V(quadColor[0], dest[0]);
+
+      /* G */
+      if (!(softpipe->blend->colormask & PIPE_MASK_G))
+          COPY_4V(quadColor[1], dest[1]);
+
+      /* B */
+      if (!(softpipe->blend->colormask & PIPE_MASK_B))
+          COPY_4V(quadColor[2], dest[2]);
+
+      /* A */
+      if (!(softpipe->blend->colormask & PIPE_MASK_A))
+          COPY_4V(quadColor[3], dest[3]);
+   }
+
+   /* pass quad to next stage */
+   qs->next->run(qs->next, quad);
+}
+
+
+static void colormask_begin(struct quad_stage *qs)
+{
+   qs->next->begin(qs->next);
+}
+
+
+static void colormask_destroy(struct quad_stage *qs)
+{
+   FREE( qs );
+}
+
+
+struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
+
+   stage->softpipe = softpipe;
+   stage->begin = colormask_begin;
+   stage->run = colormask_quad;
+   stage->destroy = colormask_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_quad_coverage.c b/src/gallium/drivers/softpipe/sp_quad_coverage.c
new file mode 100644
index 0000000000..c27fd1482d
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_coverage.c
@@ -0,0 +1,93 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * \brief  Apply AA coverage to quad alpha valus
+ * \author  Brian Paul
+ */
+
+
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_quad.h"
+
+
+/**
+ * Multiply quad's alpha values by the fragment coverage.
+ */
+static void
+coverage_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   struct softpipe_context *softpipe = qs->softpipe;
+
+   if ((softpipe->rasterizer->poly_smooth && quad->input.prim == PRIM_TRI) ||
+       (softpipe->rasterizer->line_smooth && quad->input.prim == PRIM_LINE) ||
+       (softpipe->rasterizer->point_smooth && quad->input.prim == PRIM_POINT)) {
+      uint cbuf;
+
+      /* loop over colorbuffer outputs */
+      for (cbuf = 0; cbuf < softpipe->framebuffer.nr_cbufs; cbuf++) {
+         float (*quadColor)[4] = quad->output.color[cbuf];
+         unsigned j;
+         for (j = 0; j < QUAD_SIZE; j++) {
+            assert(quad->input.coverage[j] >= 0.0);
+            assert(quad->input.coverage[j] <= 1.0);
+         quadColor[3][j] *= quad->input.coverage[j];
+         }
+      }
+   }
+
+   qs->next->run(qs->next, quad);
+}
+
+
+static void coverage_begin(struct quad_stage *qs)
+{
+   qs->next->begin(qs->next);
+}
+
+
+static void coverage_destroy(struct quad_stage *qs)
+{
+   FREE( qs );
+}
+
+
+struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
+
+   stage->softpipe = softpipe;
+   stage->begin = coverage_begin;
+   stage->run = coverage_quad;
+   stage->destroy = coverage_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c
new file mode 100644
index 0000000000..523bd3e080
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c
@@ -0,0 +1,290 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief  Quad depth testing
+ */
+
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_surface.h"
+#include "sp_quad.h"
+#include "sp_tile_cache.h"
+
+
+/**
+ * Do depth testing for a quad.
+ * Not static since it's used by the stencil code.
+ */
+
+/*
+ * To increase efficiency, we should probably have multiple versions
+ * of this function that are specifically for Z16, Z32 and FP Z buffers.
+ * Try to effectively do that with codegen...
+ */
+
+void
+sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   struct softpipe_context *softpipe = qs->softpipe;
+   struct pipe_surface *ps = softpipe->framebuffer.zsbuf;
+   const enum pipe_format format = ps->format;
+   unsigned bzzzz[QUAD_SIZE];  /**< Z values fetched from depth buffer */
+   unsigned qzzzz[QUAD_SIZE];  /**< Z values from the quad */
+   unsigned zmask = 0;
+   unsigned j;
+   struct softpipe_cached_tile *tile
+      = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->input.x0, quad->input.y0);
+
+   assert(ps); /* shouldn't get here if there's no zbuffer */
+
+   /*
+    * Convert quad's float depth values to int depth values (qzzzz).
+    * If the Z buffer stores integer values, we _have_ to do the depth
+    * compares with integers (not floats).  Otherwise, the float->int->float
+    * conversion of Z values (which isn't an identity function) will cause
+    * Z-fighting errors.
+    *
+    * Also, get the zbuffer values (bzzzz) from the cached tile.
+    */
+   switch (format) {
+   case PIPE_FORMAT_Z16_UNORM:
+      {
+         float scale = 65535.0;
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            qzzzz[j] = (unsigned) (quad->output.depth[j] * scale);
+         }
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+            bzzzz[j] = tile->data.depth16[y][x];
+         }
+      }
+      break;
+   case PIPE_FORMAT_Z32_UNORM:
+      {
+         double scale = (double) (uint) ~0UL;
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            qzzzz[j] = (unsigned) (quad->output.depth[j] * scale);
+         }
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+            bzzzz[j] = tile->data.depth32[y][x];
+         }
+      }
+      break;
+   case PIPE_FORMAT_X8Z24_UNORM:
+      /* fall-through */
+   case PIPE_FORMAT_S8Z24_UNORM:
+      {
+         float scale = (float) ((1 << 24) - 1);
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            qzzzz[j] = (unsigned) (quad->output.depth[j] * scale);
+         }
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+            bzzzz[j] = tile->data.depth32[y][x] & 0xffffff;
+         }
+      }
+      break;
+   case PIPE_FORMAT_Z24X8_UNORM:
+      /* fall-through */
+   case PIPE_FORMAT_Z24S8_UNORM:
+      {
+         float scale = (float) ((1 << 24) - 1);
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            qzzzz[j] = (unsigned) (quad->output.depth[j] * scale);
+         }
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+            bzzzz[j] = tile->data.depth32[y][x] >> 8;
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+
+   switch (softpipe->depth_stencil->depth.func) {
+   case PIPE_FUNC_NEVER:
+      /* zmask = 0 */
+      break;
+   case PIPE_FUNC_LESS:
+      /* Note this is pretty much a single sse or cell instruction.  
+       * Like this:  quad->mask &= (quad->outputs.depth < zzzz);
+       */
+      for (j = 0; j < QUAD_SIZE; j++) {
+	 if (qzzzz[j] < bzzzz[j]) 
+	    zmask |= 1 << j;
+      }
+      break;
+   case PIPE_FUNC_EQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+	 if (qzzzz[j] == bzzzz[j]) 
+	    zmask |= 1 << j;
+      }
+      break;
+   case PIPE_FUNC_LEQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+	 if (qzzzz[j] <= bzzzz[j]) 
+	    zmask |= (1 << j);
+      }
+      break;
+   case PIPE_FUNC_GREATER:
+      for (j = 0; j < QUAD_SIZE; j++) {
+	 if (qzzzz[j] > bzzzz[j]) 
+	    zmask |= (1 << j);
+      }
+      break;
+   case PIPE_FUNC_NOTEQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+	 if (qzzzz[j] != bzzzz[j]) 
+	    zmask |= (1 << j);
+      }
+      break;
+   case PIPE_FUNC_GEQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+	 if (qzzzz[j] >= bzzzz[j]) 
+	    zmask |= (1 << j);
+      }
+      break;
+   case PIPE_FUNC_ALWAYS:
+      zmask = MASK_ALL;
+      break;
+   default:
+      assert(0);
+   }
+
+   quad->inout.mask &= zmask;
+
+   if (softpipe->depth_stencil->depth.writemask) {
+      
+      /* This is also efficient with sse / spe instructions: 
+       */
+      for (j = 0; j < QUAD_SIZE; j++) {
+	 if (quad->inout.mask & (1 << j)) {
+	    bzzzz[j] = qzzzz[j];
+	 }
+      }
+
+      /* put updated Z values back into cached tile */
+      switch (format) {
+      case PIPE_FORMAT_Z16_UNORM:
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+            tile->data.depth16[y][x] = (ushort) bzzzz[j];
+         }
+         break;
+      case PIPE_FORMAT_X8Z24_UNORM:
+         /* fall-through */
+         /* (yes, this falls through to a different case than above) */
+      case PIPE_FORMAT_Z32_UNORM:
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+            tile->data.depth32[y][x] = bzzzz[j];
+         }
+         break;
+      case PIPE_FORMAT_S8Z24_UNORM:
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+            uint s8z24 = tile->data.depth32[y][x];
+            s8z24 = (s8z24 & 0xff000000) | bzzzz[j];
+            tile->data.depth32[y][x] = s8z24;
+         }
+         break;
+      case PIPE_FORMAT_Z24S8_UNORM:
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+            uint z24s8 = tile->data.depth32[y][x];
+            z24s8 = (z24s8 & 0xff) | (bzzzz[j] << 8);
+            tile->data.depth32[y][x] = z24s8;
+         }
+         break;
+      case PIPE_FORMAT_Z24X8_UNORM:
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+            tile->data.depth32[y][x] = bzzzz[j] << 8;
+         }
+         break;
+      default:
+         assert(0);
+      }
+   }
+}
+
+
+static void
+depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   sp_depth_test_quad(qs, quad);
+
+   if (quad->inout.mask)
+      qs->next->run(qs->next, quad);
+}
+
+
+static void depth_test_begin(struct quad_stage *qs)
+{
+   qs->next->begin(qs->next);
+}
+
+
+static void depth_test_destroy(struct quad_stage *qs)
+{
+   FREE( qs );
+}
+
+
+struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
+
+   stage->softpipe = softpipe;
+   stage->begin = depth_test_begin;
+   stage->run = depth_test_quad;
+   stage->destroy = depth_test_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_quad_earlyz.c b/src/gallium/drivers/softpipe/sp_quad_earlyz.c
new file mode 100644
index 0000000000..6e2dde304e
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_earlyz.c
@@ -0,0 +1,88 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \brief  Quad early-z testing
+ */
+
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "sp_headers.h"
+#include "sp_quad.h"
+
+
+/**
+ * All this stage does is compute the quad's Z values (which is normally
+ * done by the shading stage).
+ * The next stage will do the actual depth test.
+ */
+static void
+earlyz_quad(
+   struct quad_stage    *qs,
+   struct quad_header   *quad )
+{
+   const float fx = (float) quad->input.x0;
+   const float fy = (float) quad->input.y0;
+   const float dzdx = quad->posCoef->dadx[2];
+   const float dzdy = quad->posCoef->dady[2];
+   const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy;
+
+   quad->output.depth[0] = z0;
+   quad->output.depth[1] = z0 + dzdx;
+   quad->output.depth[2] = z0 + dzdy;
+   quad->output.depth[3] = z0 + dzdx + dzdy;
+
+   qs->next->run( qs->next, quad );
+}
+
+static void
+earlyz_begin(
+   struct quad_stage *qs )
+{
+   qs->next->begin( qs->next );
+}
+
+static void
+earlyz_destroy(
+   struct quad_stage *qs )
+{
+   FREE( qs );
+}
+
+struct quad_stage *
+sp_quad_earlyz_stage(
+   struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT( quad_stage );
+
+   stage->softpipe = softpipe;
+   stage->begin = earlyz_begin;
+   stage->run = earlyz_quad;
+   stage->destroy = earlyz_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c
new file mode 100644
index 0000000000..5dacbbe55f
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_fs.c
@@ -0,0 +1,189 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2008 VMware, Inc.  All rights reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Vertices are just an array of floats, with all the attributes
+ * packed.  We currently assume a layout like:
+ *
+ * attr[0][0..3] - window position
+ * attr[1..n][0..3] - remaining attributes.
+ *
+ * Attributes are assumed to be 4 floats wide but are packed so that
+ * all the enabled attributes run contiguously.
+ */
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_headers.h"
+#include "sp_quad.h"
+#include "sp_texture.h"
+#include "sp_tex_sample.h"
+
+
+struct quad_shade_stage
+{
+   struct quad_stage stage;  /**< base class */
+   struct tgsi_exec_machine machine;
+   struct tgsi_exec_vector *inputs, *outputs;
+};
+
+
+/** cast wrapper */
+static INLINE struct quad_shade_stage *
+quad_shade_stage(struct quad_stage *qs)
+{
+   return (struct quad_shade_stage *) qs;
+}
+
+
+
+/**
+ * Execute fragment shader for the four fragments in the quad.
+ */
+static void
+shade_quad(
+   struct quad_stage *qs,
+   struct quad_header *quad )
+{
+   struct quad_shade_stage *qss = quad_shade_stage( qs );
+   struct softpipe_context *softpipe = qs->softpipe;
+   struct tgsi_exec_machine *machine = &qss->machine;
+   boolean z_written;
+   
+   /* Consts do not require 16 byte alignment. */
+   machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT];
+
+   machine->InterpCoefs = quad->coef;
+
+   /* run shader */
+   quad->inout.mask &= softpipe->fs->run( softpipe->fs, 
+				    &qss->machine,
+				    quad );
+
+   /* store outputs */
+   z_written = FALSE;
+   {
+      const ubyte *sem_name = softpipe->fs->info.output_semantic_name;
+      const ubyte *sem_index = softpipe->fs->info.output_semantic_index;
+      const uint n = qss->stage.softpipe->fs->info.num_outputs;
+      uint i;
+      for (i = 0; i < n; i++) {
+         switch (sem_name[i]) {
+         case TGSI_SEMANTIC_COLOR:
+            {
+               uint cbuf = sem_index[i];
+               memcpy(quad->output.color[cbuf],
+                      &machine->Outputs[i].xyzw[0].f[0],
+                      sizeof(quad->output.color[0]) );
+            }
+            break;
+         case TGSI_SEMANTIC_POSITION:
+            {
+               uint j;
+               for (j = 0; j < 4; j++) {
+                  quad->output.depth[j] = machine->Outputs[0].xyzw[2].f[j];
+               }
+               z_written = TRUE;
+            }
+            break;
+         }
+      }
+   }
+
+   if (!z_written) {
+      /* compute Z values now, as in the quad earlyz stage */
+      /* XXX we should really only do this if the earlyz stage is not used */
+      const float fx = (float) quad->input.x0;
+      const float fy = (float) quad->input.y0;
+      const float dzdx = quad->posCoef->dadx[2];
+      const float dzdy = quad->posCoef->dady[2];
+      const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy;
+
+      quad->output.depth[0] = z0;
+      quad->output.depth[1] = z0 + dzdx;
+      quad->output.depth[2] = z0 + dzdy;
+      quad->output.depth[3] = z0 + dzdx + dzdy;
+   }
+
+   /* shader may cull fragments */
+   if( quad->inout.mask ) {
+      qs->next->run( qs->next, quad );
+   }
+}
+
+/**
+ * Per-primitive (or per-begin?) setup
+ */
+static void shade_begin(struct quad_stage *qs)
+{
+   struct quad_shade_stage *qss = quad_shade_stage(qs);
+   struct softpipe_context *softpipe = qs->softpipe;
+
+   softpipe->fs->prepare( softpipe->fs, 
+			  &qss->machine,
+			  (struct tgsi_sampler **)
+                             softpipe->tgsi.frag_samplers_list );
+
+   qs->next->begin(qs->next);
+}
+
+
+static void shade_destroy(struct quad_stage *qs)
+{
+   struct quad_shade_stage *qss = (struct quad_shade_stage *) qs;
+
+   tgsi_exec_machine_free_data(&qss->machine);
+   FREE( qss->inputs );
+   FREE( qss->outputs );
+   FREE( qs );
+}
+
+
+struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe )
+{
+   struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage);
+
+   /* allocate storage for program inputs/outputs, aligned to 16 bytes */
+   qss->inputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->inputs) + 16);
+   qss->outputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->outputs) + 16);
+   qss->machine.Inputs = align16(qss->inputs);
+   qss->machine.Outputs = align16(qss->outputs);
+
+   qss->stage.softpipe = softpipe;
+   qss->stage.begin = shade_begin;
+   qss->stage.run = shade_quad;
+   qss->stage.destroy = shade_destroy;
+
+   tgsi_exec_machine_init( &qss->machine );
+
+   return &qss->stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_quad_occlusion.c b/src/gallium/drivers/softpipe/sp_quad_occlusion.c
new file mode 100644
index 0000000000..169bd82876
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_occlusion.c
@@ -0,0 +1,85 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * \brief  Quad occlusion counter stage
+ * \author  Brian Paul
+ */
+
+
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_surface.h"
+#include "sp_quad.h"
+
+static unsigned count_bits( unsigned val )
+{
+   unsigned i;
+
+   for (i = 0; val ; val >>= 1)
+      i += (val & 1);
+
+   return i;
+}
+
+static void
+occlusion_count_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   struct softpipe_context *softpipe = qs->softpipe;
+
+   softpipe->occlusion_count += count_bits(quad->inout.mask);
+
+   qs->next->run(qs->next, quad);
+}
+
+
+static void occlusion_begin(struct quad_stage *qs)
+{
+   qs->next->begin(qs->next);
+}
+
+
+static void occlusion_destroy(struct quad_stage *qs)
+{
+   FREE( qs );
+}
+
+
+struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
+
+   stage->softpipe = softpipe;
+   stage->begin = occlusion_begin;
+   stage->run = occlusion_count_quad;
+   stage->destroy = occlusion_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_quad_output.c b/src/gallium/drivers/softpipe/sp_quad_output.c
new file mode 100644
index 0000000000..a37c8b4c39
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_output.c
@@ -0,0 +1,103 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "util/u_memory.h"
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_surface.h"
+#include "sp_quad.h"
+#include "sp_tile_cache.h"
+
+
+/**
+ * Last step of quad processing: write quad colors to the framebuffer,
+ * taking mask into account.
+ */
+static void
+output_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   /* in-tile pos: */
+   const int itx = quad->input.x0 % TILE_SIZE;
+   const int ity = quad->input.y0 % TILE_SIZE;
+
+   struct softpipe_context *softpipe = qs->softpipe;
+   uint cbuf;
+
+   /* loop over colorbuffer outputs */
+   for (cbuf = 0; cbuf < softpipe->framebuffer.nr_cbufs; cbuf++) {
+      struct softpipe_cached_tile *tile
+         = sp_get_cached_tile(softpipe,
+                              softpipe->cbuf_cache[cbuf],
+                              quad->input.x0, quad->input.y0);
+      float (*quadColor)[4] = quad->output.color[cbuf];
+      int i, j;
+
+      /* get/swizzle dest colors */
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (quad->inout.mask & (1 << j)) {
+            int x = itx + (j & 1);
+            int y = ity + (j >> 1);
+            for (i = 0; i < 4; i++) { /* loop over color chans */
+               tile->data.color[y][x][i] = quadColor[i][j];
+            }
+            if (0) {
+               debug_printf("sp write pixel %d,%d: %g, %g, %g\n",
+                            quad->input.x0 + x,
+                            quad->input.y0 + y,
+                            quadColor[0][j],
+                            quadColor[1][j],
+                            quadColor[2][j]);
+            }
+         }
+      }
+   }
+}
+
+
+static void output_begin(struct quad_stage *qs)
+{
+   assert(qs->next == NULL);
+}
+
+
+static void output_destroy(struct quad_stage *qs)
+{
+   FREE( qs );
+}
+
+
+struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
+
+   stage->softpipe = softpipe;
+   stage->begin = output_begin;
+   stage->run = output_quad;
+   stage->destroy = output_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_quad_stencil.c b/src/gallium/drivers/softpipe/sp_quad_stencil.c
new file mode 100644
index 0000000000..7495515764
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_stencil.c
@@ -0,0 +1,352 @@
+
+/**
+ * \brief Quad stencil testing
+ */
+
+
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_surface.h"
+#include "sp_tile_cache.h"
+#include "sp_quad.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+
+
+/** Only 8-bit stencil supported */
+#define STENCIL_MAX 0xff
+
+
+/**
+ * Do the basic stencil test (compare stencil buffer values against the
+ * reference value.
+ *
+ * \param stencilVals  the stencil values from the stencil buffer
+ * \param func  the stencil func (PIPE_FUNC_x)
+ * \param ref  the stencil reference value
+ * \param valMask  the stencil value mask indicating which bits of the stencil
+ *                 values and ref value are to be used.
+ * \return mask indicating which pixels passed the stencil test
+ */
+static unsigned
+do_stencil_test(const ubyte stencilVals[QUAD_SIZE], unsigned func,
+                unsigned ref, unsigned valMask)
+{
+   unsigned passMask = 0x0;
+   unsigned j;
+
+   ref &= valMask;
+
+   switch (func) {
+   case PIPE_FUNC_NEVER:
+      /* passMask = 0x0 */
+      break;
+   case PIPE_FUNC_LESS:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (ref < (stencilVals[j] & valMask)) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_EQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (ref == (stencilVals[j] & valMask)) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_LEQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (ref <= (stencilVals[j] & valMask)) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_GREATER:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (ref > (stencilVals[j] & valMask)) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_NOTEQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (ref != (stencilVals[j] & valMask)) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_GEQUAL:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (ref >= (stencilVals[j] & valMask)) {
+            passMask |= (1 << j);
+         }
+      }
+      break;
+   case PIPE_FUNC_ALWAYS:
+      passMask = MASK_ALL;
+      break;
+   default:
+      assert(0);
+   }
+
+   return passMask;
+}
+
+
+/**
+ * Apply the stencil operator to stencil values.
+ *
+ * \param stencilVals  the stencil buffer values (read and written)
+ * \param mask  indicates which pixels to update
+ * \param op  the stencil operator (PIPE_STENCIL_OP_x)
+ * \param ref  the stencil reference value
+ * \param wrtMask  writemask controlling which bits are changed in the
+ *                 stencil values
+ */
+static void
+apply_stencil_op(ubyte stencilVals[QUAD_SIZE],
+                 unsigned mask, unsigned op, ubyte ref, ubyte wrtMask)
+{
+   unsigned j;
+   ubyte newstencil[QUAD_SIZE];
+
+   for (j = 0; j < QUAD_SIZE; j++) {
+      newstencil[j] = stencilVals[j];
+   }
+
+   switch (op) {
+   case PIPE_STENCIL_OP_KEEP:
+      /* no-op */
+      break;
+   case PIPE_STENCIL_OP_ZERO:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (mask & (1 << j)) {
+            newstencil[j] = 0;
+         }
+      }
+      break;
+   case PIPE_STENCIL_OP_REPLACE:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (mask & (1 << j)) {
+            newstencil[j] = ref;
+         }
+      }
+      break;
+   case PIPE_STENCIL_OP_INCR:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (mask & (1 << j)) {
+            if (stencilVals[j] < STENCIL_MAX) {
+               newstencil[j] = stencilVals[j] + 1;
+            }
+         }
+      }
+      break;
+   case PIPE_STENCIL_OP_DECR:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (mask & (1 << j)) {
+            if (stencilVals[j] > 0) {
+               newstencil[j] = stencilVals[j] - 1;
+            }
+         }
+      }
+      break;
+   case PIPE_STENCIL_OP_INCR_WRAP:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (mask & (1 << j)) {
+            newstencil[j] = stencilVals[j] + 1;
+         }
+      }
+      break;
+   case PIPE_STENCIL_OP_DECR_WRAP:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (mask & (1 << j)) {
+            newstencil[j] = stencilVals[j] - 1;
+         }
+      }
+      break;
+   case PIPE_STENCIL_OP_INVERT:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         if (mask & (1 << j)) {
+            newstencil[j] = ~stencilVals[j];
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+
+   /*
+    * update the stencil values
+    */
+   if (wrtMask != STENCIL_MAX) {
+      /* apply bit-wise stencil buffer writemask */
+      for (j = 0; j < QUAD_SIZE; j++) {
+         stencilVals[j] = (wrtMask & newstencil[j]) | (~wrtMask & stencilVals[j]);
+      }
+   }
+   else {
+      for (j = 0; j < QUAD_SIZE; j++) {
+         stencilVals[j] = newstencil[j];
+      }
+   }
+}
+
+
+/**
+ * Do stencil (and depth) testing.  Stenciling depends on the outcome of
+ * depth testing.
+ */
+static void
+stencil_test_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   struct softpipe_context *softpipe = qs->softpipe;
+   struct pipe_surface *ps = softpipe->framebuffer.zsbuf;
+   unsigned func, zFailOp, zPassOp, failOp;
+   ubyte ref, wrtMask, valMask;
+   ubyte stencilVals[QUAD_SIZE];
+   struct softpipe_cached_tile *tile
+      = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->input.x0, quad->input.y0);
+   uint j;
+   uint face = quad->input.facing;
+
+   if (!softpipe->depth_stencil->stencil[1].enabled) {
+      /* single-sided stencil test, use front (face=0) state */
+      face = 0;
+   }
+
+   /* choose front or back face function, operator, etc */
+   /* XXX we could do these initializations once per primitive */
+   func    = softpipe->depth_stencil->stencil[face].func;
+   failOp  = softpipe->depth_stencil->stencil[face].fail_op;
+   zFailOp = softpipe->depth_stencil->stencil[face].zfail_op;
+   zPassOp = softpipe->depth_stencil->stencil[face].zpass_op;
+   ref     = softpipe->depth_stencil->stencil[face].ref_value;
+   wrtMask = softpipe->depth_stencil->stencil[face].writemask;
+   valMask = softpipe->depth_stencil->stencil[face].valuemask;
+
+   assert(ps); /* shouldn't get here if there's no stencil buffer */
+
+   /* get stencil values from cached tile */
+   switch (ps->format) {
+   case PIPE_FORMAT_S8Z24_UNORM:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+         stencilVals[j] = tile->data.depth32[y][x] >> 24;
+      }
+      break;
+   case PIPE_FORMAT_Z24S8_UNORM:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+         stencilVals[j] = tile->data.depth32[y][x] & 0xff;
+      }
+      break;
+   case PIPE_FORMAT_S8_UNORM:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+         stencilVals[j] = tile->data.stencil8[y][x];
+      }
+      break;
+   default:
+      assert(0);
+   }
+
+   /* do the stencil test first */
+   {
+      unsigned passMask, failMask;
+      passMask = do_stencil_test(stencilVals, func, ref, valMask);
+      failMask = quad->inout.mask & ~passMask;
+      quad->inout.mask &= passMask;
+
+      if (failOp != PIPE_STENCIL_OP_KEEP) {
+         apply_stencil_op(stencilVals, failMask, failOp, ref, wrtMask);
+      }
+   }
+
+   if (quad->inout.mask) {
+      /* now the pixels that passed the stencil test are depth tested */
+      if (softpipe->depth_stencil->depth.enabled) {
+         const unsigned origMask = quad->inout.mask;
+
+         sp_depth_test_quad(qs, quad);  /* quad->mask is updated */
+
+         /* update stencil buffer values according to z pass/fail result */
+         if (zFailOp != PIPE_STENCIL_OP_KEEP) {
+            const unsigned failMask = origMask & ~quad->inout.mask;
+            apply_stencil_op(stencilVals, failMask, zFailOp, ref, wrtMask);
+         }
+
+         if (zPassOp != PIPE_STENCIL_OP_KEEP) {
+            const unsigned passMask = origMask & quad->inout.mask;
+            apply_stencil_op(stencilVals, passMask, zPassOp, ref, wrtMask);
+         }
+      }
+      else {
+         /* no depth test, apply Zpass operator to stencil buffer values */
+         apply_stencil_op(stencilVals, quad->inout.mask, zPassOp, ref, wrtMask);
+      }
+
+   }
+
+   /* put new stencil values into cached tile */
+   switch (ps->format) {
+   case PIPE_FORMAT_S8Z24_UNORM:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+         uint s8z24 = tile->data.depth32[y][x];
+         s8z24 = (stencilVals[j] << 24) | (s8z24 & 0xffffff);
+         tile->data.depth32[y][x] = s8z24;
+      }
+      break;
+   case PIPE_FORMAT_Z24S8_UNORM:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+         uint z24s8 = tile->data.depth32[y][x];
+         z24s8 = (z24s8 & 0xffffff00) | stencilVals[j];
+         tile->data.depth32[y][x] = z24s8;
+      }
+      break;
+   case PIPE_FORMAT_S8_UNORM:
+      for (j = 0; j < QUAD_SIZE; j++) {
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
+         tile->data.stencil8[y][x] = stencilVals[j];
+      }
+      break;
+   default:
+      assert(0);
+   }
+
+   if (quad->inout.mask)
+      qs->next->run(qs->next, quad);
+}
+
+
+static void stencil_begin(struct quad_stage *qs)
+{
+   qs->next->begin(qs->next);
+}
+
+
+static void stencil_destroy(struct quad_stage *qs)
+{
+   FREE( qs );
+}
+
+
+struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
+
+   stage->softpipe = softpipe;
+   stage->begin = stencil_begin;
+   stage->run = stencil_test_quad;
+   stage->destroy = stencil_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_quad_stipple.c b/src/gallium/drivers/softpipe/sp_quad_stipple.c
new file mode 100644
index 0000000000..ccf37f6be5
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_quad_stipple.c
@@ -0,0 +1,94 @@
+
+/**
+ * quad polygon stipple stage
+ */
+
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_quad.h"
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+
+
+/**
+ * Apply polygon stipple to quads produced by triangle rasterization
+ */
+static void
+stipple_quad(struct quad_stage *qs, struct quad_header *quad)
+{
+   static const uint bit31 = 1 << 31;
+   static const uint bit30 = 1 << 30;
+
+   if (quad->input.prim == PRIM_TRI) {
+      struct softpipe_context *softpipe = qs->softpipe;
+      /* need to invert Y to index into OpenGL's stipple pattern */
+      int y0, y1;
+      uint stipple0, stipple1;
+      if (softpipe->rasterizer->origin_lower_left) {
+         y0 = softpipe->framebuffer.height - 1 - quad->input.y0;
+         y1 = y0 - 1;
+      }
+      else {
+         y0 = quad->input.y0;
+         y1 = y0 + 1;
+      }
+      stipple0 = softpipe->poly_stipple.stipple[y0 % 32];
+      stipple1 = softpipe->poly_stipple.stipple[y1 % 32];
+
+#if 1
+      {
+      const int col0 = quad->input.x0 % 32;
+      if ((stipple0 & (bit31 >> col0)) == 0)
+         quad->inout.mask &= ~MASK_TOP_LEFT;
+
+      if ((stipple0 & (bit30 >> col0)) == 0)
+         quad->inout.mask &= ~MASK_TOP_RIGHT;
+
+      if ((stipple1 & (bit31 >> col0)) == 0)
+         quad->inout.mask &= ~MASK_BOTTOM_LEFT;
+
+      if ((stipple1 & (bit30 >> col0)) == 0)
+         quad->inout.mask &= ~MASK_BOTTOM_RIGHT;
+      }
+#else
+      /* We'd like to use this code, but we'd need to redefine
+       * MASK_TOP_LEFT to be (1 << 1) and MASK_TOP_RIGHT to be (1 << 0),
+       * and similarly for the BOTTOM bits.  But that may have undesirable
+       * side effects elsewhere.
+       */
+      const int col0 = 30 - (quad->input.x0 % 32);
+      quad->inout.mask &= (((stipple0 >> col0) & 0x3) | 
+                     (((stipple1 >> col0) & 0x3) << 2));
+#endif
+      if (!quad->inout.mask)
+         return;
+   }
+
+   qs->next->run(qs->next, quad);
+}
+
+
+static void stipple_begin(struct quad_stage *qs)
+{
+   qs->next->begin(qs->next);
+}
+
+
+static void stipple_destroy(struct quad_stage *qs)
+{
+   FREE( qs );
+}
+
+
+struct quad_stage *
+sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe )
+{
+   struct quad_stage *stage = CALLOC_STRUCT(quad_stage);
+
+   stage->softpipe = softpipe;
+   stage->begin = stipple_begin;
+   stage->run = stipple_quad;
+   stage->destroy = stipple_destroy;
+
+   return stage;
+}
diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c
new file mode 100644
index 0000000000..b0d8e01426
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_query.c
@@ -0,0 +1,107 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "draw/draw_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+#include "sp_context.h"
+#include "sp_query.h"
+
+struct softpipe_query {
+   uint64_t start;
+   uint64_t end;
+};
+
+
+static struct softpipe_query *softpipe_query( struct pipe_query *p )
+{
+   return (struct softpipe_query *)p;
+}
+
+static struct pipe_query *
+softpipe_create_query(struct pipe_context *pipe, 
+		      unsigned type)
+{
+   assert(type == PIPE_QUERY_OCCLUSION_COUNTER);
+   return (struct pipe_query *)CALLOC_STRUCT( softpipe_query );
+}
+
+
+static void
+softpipe_destroy_query(struct pipe_context *pipe, struct pipe_query *q)
+{
+   FREE(q);
+}
+
+
+static void
+softpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q)
+{
+   struct softpipe_context *softpipe = softpipe_context( pipe );
+   struct softpipe_query *sq = softpipe_query(q);
+   
+   sq->start = softpipe->occlusion_count;
+}
+
+
+static void
+softpipe_end_query(struct pipe_context *pipe, struct pipe_query *q)
+{
+   struct softpipe_context *softpipe = softpipe_context( pipe );
+   struct softpipe_query *sq = softpipe_query(q);
+
+   sq->end = softpipe->occlusion_count;
+}
+
+
+static boolean
+softpipe_get_query_result(struct pipe_context *pipe, 
+			  struct pipe_query *q,
+			  boolean wait,
+			  uint64_t *result )
+{
+   struct softpipe_query *sq = softpipe_query(q);
+   *result = sq->end - sq->start;
+   return TRUE;
+}
+
+
+void softpipe_init_query_funcs(struct softpipe_context *softpipe )
+{
+   softpipe->pipe.create_query = softpipe_create_query;
+   softpipe->pipe.destroy_query = softpipe_destroy_query;
+   softpipe->pipe.begin_query = softpipe_begin_query;
+   softpipe->pipe.end_query = softpipe_end_query;
+   softpipe->pipe.get_query_result = softpipe_get_query_result;
+}
+
+
diff --git a/src/gallium/drivers/softpipe/sp_query.h b/src/gallium/drivers/softpipe/sp_query.h
new file mode 100644
index 0000000000..05060a4575
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_query.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Author:
+ *    Keith Whitwell
+ */
+
+#ifndef SP_QUERY_H
+#define SP_QUERY_H
+
+struct softpipe_context;
+extern void softpipe_init_query_funcs(struct softpipe_context * );
+
+
+#endif /* SP_QUERY_H */
diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
new file mode 100644
index 0000000000..7380a6ae2b
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -0,0 +1,181 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "util/u_memory.h"
+#include "util/u_simple_screen.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+
+#include "sp_texture.h"
+#include "sp_winsys.h"
+#include "sp_screen.h"
+
+
+static const char *
+softpipe_get_vendor(struct pipe_screen *screen)
+{
+   return "Tungsten Graphics, Inc.";
+}
+
+
+static const char *
+softpipe_get_name(struct pipe_screen *screen)
+{
+   return "softpipe";
+}
+
+
+static int
+softpipe_get_param(struct pipe_screen *screen, int param)
+{
+   switch (param) {
+   case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+      return PIPE_MAX_SAMPLERS;
+   case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
+      return PIPE_MAX_SAMPLERS;
+   case PIPE_CAP_NPOT_TEXTURES:
+      return 1;
+   case PIPE_CAP_TWO_SIDED_STENCIL:
+      return 1;
+   case PIPE_CAP_GLSL:
+      return 1;
+   case PIPE_CAP_S3TC:
+      return 0;
+   case PIPE_CAP_ANISOTROPIC_FILTER:
+      return 0;
+   case PIPE_CAP_POINT_SPRITE:
+      return 1;
+   case PIPE_CAP_MAX_RENDER_TARGETS:
+      return PIPE_MAX_COLOR_BUFS;
+   case PIPE_CAP_OCCLUSION_QUERY:
+      return 1;
+   case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
+      return 1;
+   case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
+      return 1;
+   case PIPE_CAP_TEXTURE_SHADOW_MAP:
+      return 1;
+   case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+      return 12; /* max 2Kx2K */
+   case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+      return 8;  /* max 128x128x128 */
+   case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+      return 12; /* max 2Kx2K */
+   default:
+      return 0;
+   }
+}
+
+
+static float
+softpipe_get_paramf(struct pipe_screen *screen, int param)
+{
+   switch (param) {
+   case PIPE_CAP_MAX_LINE_WIDTH:
+      /* fall-through */
+   case PIPE_CAP_MAX_LINE_WIDTH_AA:
+      return 255.0; /* arbitrary */
+   case PIPE_CAP_MAX_POINT_WIDTH:
+      /* fall-through */
+   case PIPE_CAP_MAX_POINT_WIDTH_AA:
+      return 255.0; /* arbitrary */
+   case PIPE_CAP_MAX_TEXTURE_ANISOTROPY:
+      return 0.0;
+   case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
+      return 16.0; /* arbitrary */
+   default:
+      return 0;
+   }
+}
+
+
+/**
+ * Query format support for creating a texture, drawing surface, etc.
+ * \param format  the format to test
+ * \param type  one of PIPE_TEXTURE, PIPE_SURFACE
+ */
+static boolean
+softpipe_is_format_supported( struct pipe_screen *screen,
+                              enum pipe_format format, 
+                              enum pipe_texture_target target,
+                              unsigned tex_usage, 
+                              unsigned geom_flags )
+{
+   switch(format) {
+   case PIPE_FORMAT_DXT1_RGB:
+   case PIPE_FORMAT_DXT1_RGBA:
+   case PIPE_FORMAT_DXT3_RGBA:
+   case PIPE_FORMAT_DXT5_RGBA:
+      return FALSE;
+   default:
+      return TRUE;
+   }
+}
+
+
+static void
+softpipe_destroy_screen( struct pipe_screen *screen )
+{
+   struct pipe_winsys *winsys = screen->winsys;
+
+   if(winsys->destroy)
+      winsys->destroy(winsys);
+
+   FREE(screen);
+}
+
+
+
+/**
+ * Create a new pipe_screen object
+ * Note: we're not presently subclassing pipe_screen (no softpipe_screen).
+ */
+struct pipe_screen *
+softpipe_create_screen(struct pipe_winsys *winsys)
+{
+   struct softpipe_screen *screen = CALLOC_STRUCT(softpipe_screen);
+
+   if (!screen)
+      return NULL;
+
+   screen->base.winsys = winsys;
+
+   screen->base.destroy = softpipe_destroy_screen;
+
+   screen->base.get_name = softpipe_get_name;
+   screen->base.get_vendor = softpipe_get_vendor;
+   screen->base.get_param = softpipe_get_param;
+   screen->base.get_paramf = softpipe_get_paramf;
+   screen->base.is_format_supported = softpipe_is_format_supported;
+
+   softpipe_init_screen_texture_funcs(&screen->base);
+   u_simple_screen_init(&screen->base);
+
+   return &screen->base;
+}
diff --git a/src/gallium/drivers/softpipe/sp_screen.h b/src/gallium/drivers/softpipe/sp_screen.h
new file mode 100644
index 0000000000..3d4bfd3e84
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_screen.h
@@ -0,0 +1,58 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef SP_SCREEN_H
+#define SP_SCREEN_H
+
+#include "pipe/p_screen.h"
+#include "pipe/p_defines.h"
+
+
+
+struct softpipe_screen {
+   struct pipe_screen base;
+
+   /* Increments whenever textures are modified.  Contexts can track
+    * this.
+    */
+   unsigned timestamp;          
+};
+
+
+
+
+static INLINE struct softpipe_screen *
+softpipe_screen( struct pipe_screen *pipe )
+{
+   return (struct softpipe_screen *)pipe;
+}
+
+
+#endif /* SP_SCREEN_H */
diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c
new file mode 100644
index 0000000000..b1adb9cb7a
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_setup.c
@@ -0,0 +1,1569 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * \brief  Primitive rasterization/rendering (points, lines, triangles)
+ *
+ * \author  Keith Whitwell <keith@tungstengraphics.com>
+ * \author  Brian Paul
+ */
+
+#include "sp_setup.h"
+
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_quad.h"
+#include "sp_state.h"
+#include "sp_prim_setup.h"
+#include "draw/draw_context.h"
+#include "draw/draw_private.h"
+#include "draw/draw_vertex.h"
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_thread.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+
+#define DEBUG_VERTS 0
+#define DEBUG_FRAGS 0
+
+/**
+ * Triangle edge info
+ */
+struct edge {
+   float dx;		/**< X(v1) - X(v0), used only during setup */
+   float dy;		/**< Y(v1) - Y(v0), used only during setup */
+   float dxdy;		/**< dx/dy */
+   float sx, sy;	/**< first sample point coord */
+   int lines;		/**< number of lines on this edge */
+};
+
+#if SP_NUM_QUAD_THREADS > 1
+
+/* Set to 1 if you want other threads to be instantly
+ * notified of pending jobs.
+ */
+#define INSTANT_NOTEMPTY_NOTIFY 0
+
+struct thread_info
+{
+   struct setup_context *setup;
+   uint id;
+   pipe_thread handle;
+};
+
+struct quad_job;
+
+typedef void (* quad_job_routine)( struct setup_context *setup, uint thread, struct quad_job *job );
+
+struct quad_job
+{
+   struct quad_header_input input;
+   struct quad_header_inout inout;
+   quad_job_routine routine;
+};
+
+#define NUM_QUAD_JOBS 64
+
+struct quad_job_que
+{
+   struct quad_job jobs[NUM_QUAD_JOBS];
+   uint first;
+   uint last;
+   pipe_mutex que_mutex;
+   pipe_condvar que_notfull_condvar;
+   pipe_condvar que_notempty_condvar;
+   uint jobs_added;
+   uint jobs_done;
+   pipe_condvar que_done_condvar;
+};
+
+static void
+add_quad_job( struct quad_job_que *que, struct quad_header *quad, quad_job_routine routine )
+{
+#if INSTANT_NOTEMPTY_NOTIFY
+   boolean empty;
+#endif
+
+   /* Wait for empty slot, see if the que is empty.
+    */
+   pipe_mutex_lock( que->que_mutex );
+   while ((que->last + 1) % NUM_QUAD_JOBS == que->first) {
+#if !INSTANT_NOTEMPTY_NOTIFY
+      pipe_condvar_broadcast( que->que_notempty_condvar );
+#endif
+      pipe_condvar_wait( que->que_notfull_condvar, que->que_mutex );
+   }
+#if INSTANT_NOTEMPTY_NOTIFY
+   empty = que->last == que->first;
+#endif
+   que->jobs_added++;
+   pipe_mutex_unlock( que->que_mutex );
+
+   /* Submit new job.
+    */
+   que->jobs[que->last].input = quad->input;
+   que->jobs[que->last].inout = quad->inout;
+   que->jobs[que->last].routine = routine;
+   que->last = (que->last + 1) % NUM_QUAD_JOBS;
+
+#if INSTANT_NOTEMPTY_NOTIFY
+   /* If the que was empty, notify consumers there's a job to be done.
+    */
+   if (empty) {
+      pipe_mutex_lock( que->que_mutex );
+      pipe_condvar_broadcast( que->que_notempty_condvar );
+      pipe_mutex_unlock( que->que_mutex );
+   }
+#endif
+}
+
+#endif
+
+/**
+ * Triangle setup info (derived from draw_stage).
+ * Also used for line drawing (taking some liberties).
+ */
+struct setup_context {
+   struct softpipe_context *softpipe;
+
+   /* Vertices are just an array of floats making up each attribute in
+    * turn.  Currently fixed at 4 floats, but should change in time.
+    * Codegen will help cope with this.
+    */
+   const float (*vmax)[4];
+   const float (*vmid)[4];
+   const float (*vmin)[4];
+   const float (*vprovoke)[4];
+
+   struct edge ebot;
+   struct edge etop;
+   struct edge emaj;
+
+   float oneoverarea;
+
+   struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS];
+   struct tgsi_interp_coef posCoef;  /* For Z, W */
+   struct quad_header quad;
+
+#if SP_NUM_QUAD_THREADS > 1
+   struct quad_job_que que;
+   struct thread_info threads[SP_NUM_QUAD_THREADS];
+#endif
+
+   struct {
+      int left[2];   /**< [0] = row0, [1] = row1 */
+      int right[2];
+      int y;
+      unsigned y_flags;
+      unsigned mask;     /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */
+   } span;
+
+#if DEBUG_FRAGS
+   uint numFragsEmitted;  /**< per primitive */
+   uint numFragsWritten;  /**< per primitive */
+#endif
+
+   unsigned winding;		/* which winding to cull */
+};
+
+#if SP_NUM_QUAD_THREADS > 1
+
+static PIPE_THREAD_ROUTINE( quad_thread, param )
+{
+   struct thread_info *info = (struct thread_info *) param;
+   struct quad_job_que *que = &info->setup->que;
+
+   for (;;) {
+      struct quad_job job;
+      boolean full;
+
+      /* Wait for an available job.
+       */
+      pipe_mutex_lock( que->que_mutex );
+      while (que->last == que->first)
+         pipe_condvar_wait( que->que_notempty_condvar, que->que_mutex );
+
+      /* See if the que is full.
+       */
+      full = (que->last + 1) % NUM_QUAD_JOBS == que->first;
+
+      /* Take a job and remove it from que.
+       */
+      job = que->jobs[que->first];
+      que->first = (que->first + 1) % NUM_QUAD_JOBS;
+
+      /* Notify the producer if the que is not full.
+       */
+      if (full)
+         pipe_condvar_signal( que->que_notfull_condvar );
+      pipe_mutex_unlock( que->que_mutex );
+
+      job.routine( info->setup, info->id, &job );
+
+      /* Notify the producer if that's the last finished job.
+       */
+      pipe_mutex_lock( que->que_mutex );
+      que->jobs_done++;
+      if (que->jobs_added == que->jobs_done)
+         pipe_condvar_signal( que->que_done_condvar );
+      pipe_mutex_unlock( que->que_mutex );
+   }
+
+   return NULL;
+}
+
+#define WAIT_FOR_COMPLETION(setup) \
+   do {\
+      pipe_mutex_lock( setup->que.que_mutex );\
+      if (!INSTANT_NOTEMPTY_NOTIFY)\
+         pipe_condvar_broadcast( setup->que.que_notempty_condvar );\
+      while (setup->que.jobs_added != setup->que.jobs_done)\
+         pipe_condvar_wait( setup->que.que_done_condvar, setup->que.que_mutex );\
+      pipe_mutex_unlock( setup->que.que_mutex );\
+   } while (0)
+
+#else
+
+#define WAIT_FOR_COMPLETION(setup) ((void) 0)
+
+#endif
+
+/**
+ * Test if x is NaN or +/- infinity.
+ */
+static INLINE boolean
+is_inf_or_nan(float x)
+{
+   union fi tmp;
+   tmp.f = x;
+   return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31);
+}
+
+
+static boolean cull_tri( struct setup_context *setup,
+		      float det )
+{
+   if (det != 0) 
+   {   
+      /* if (det < 0 then Z points toward camera and triangle is 
+       * counter-clockwise winding.
+       */
+      unsigned winding = (det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW;
+      
+      if ((winding & setup->winding) == 0) 
+	 return FALSE;
+   }
+
+   /* Culled:
+    */
+   return TRUE;
+}
+
+
+
+/**
+ * Clip setup->quad against the scissor/surface bounds.
+ */
+static INLINE void
+quad_clip( struct setup_context *setup, struct quad_header *quad )
+{
+   const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect;
+   const int minx = (int) cliprect->minx;
+   const int maxx = (int) cliprect->maxx;
+   const int miny = (int) cliprect->miny;
+   const int maxy = (int) cliprect->maxy;
+
+   if (quad->input.x0 >= maxx ||
+       quad->input.y0 >= maxy ||
+       quad->input.x0 + 1 < minx ||
+       quad->input.y0 + 1 < miny) {
+      /* totally clipped */
+      quad->inout.mask = 0x0;
+      return;
+   }
+   if (quad->input.x0 < minx)
+      quad->inout.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
+   if (quad->input.y0 < miny)
+      quad->inout.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
+   if (quad->input.x0 == maxx - 1)
+      quad->inout.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
+   if (quad->input.y0 == maxy - 1)
+      quad->inout.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
+}
+
+
+/**
+ * Emit a quad (pass to next stage) with clipping.
+ */
+static INLINE void
+clip_emit_quad( struct setup_context *setup, struct quad_header *quad, uint thread )
+{
+   quad_clip( setup, quad );
+   if (quad->inout.mask) {
+      struct softpipe_context *sp = setup->softpipe;
+
+      sp->quad[thread].first->run( sp->quad[thread].first, quad );
+   }
+}
+
+#if SP_NUM_QUAD_THREADS > 1
+
+static void
+clip_emit_quad_job( struct setup_context *setup, uint thread, struct quad_job *job )
+{
+   struct quad_header quad;
+
+   quad.input = job->input;
+   quad.inout = job->inout;
+   quad.coef = setup->quad.coef;
+   quad.posCoef = setup->quad.posCoef;
+   quad.nr_attrs = setup->quad.nr_attrs;
+   clip_emit_quad( setup, &quad, thread );
+}
+
+#define CLIP_EMIT_QUAD(setup) add_quad_job( &setup->que, &setup->quad, clip_emit_quad_job )
+
+#else
+
+#define CLIP_EMIT_QUAD(setup) clip_emit_quad( setup, &setup->quad, 0 )
+
+#endif
+
+/**
+ * Emit a quad (pass to next stage).  No clipping is done.
+ */
+static INLINE void
+emit_quad( struct setup_context *setup, struct quad_header *quad, uint thread )
+{
+   struct softpipe_context *sp = setup->softpipe;
+#if DEBUG_FRAGS
+   uint mask = quad->inout.mask;
+#endif
+
+#if DEBUG_FRAGS
+   if (mask & 1) setup->numFragsEmitted++;
+   if (mask & 2) setup->numFragsEmitted++;
+   if (mask & 4) setup->numFragsEmitted++;
+   if (mask & 8) setup->numFragsEmitted++;
+#endif
+   sp->quad[thread].first->run( sp->quad[thread].first, quad );
+#if DEBUG_FRAGS
+   mask = quad->inout.mask;
+   if (mask & 1) setup->numFragsWritten++;
+   if (mask & 2) setup->numFragsWritten++;
+   if (mask & 4) setup->numFragsWritten++;
+   if (mask & 8) setup->numFragsWritten++;
+#endif
+}
+
+#if SP_NUM_QUAD_THREADS > 1
+
+static void
+emit_quad_job( struct setup_context *setup, uint thread, struct quad_job *job )
+{
+   struct quad_header quad;
+
+   quad.input = job->input;
+   quad.inout = job->inout;
+   quad.coef = setup->quad.coef;
+   quad.posCoef = setup->quad.posCoef;
+   quad.nr_attrs = setup->quad.nr_attrs;
+   emit_quad( setup, &quad, thread );
+}
+
+#define EMIT_QUAD(setup,x,y,mask) do {\
+      setup->quad.input.x0 = x;\
+      setup->quad.input.y0 = y;\
+      setup->quad.inout.mask = mask;\
+      add_quad_job( &setup->que, &setup->quad, emit_quad_job );\
+   } while (0)
+
+#else
+
+#define EMIT_QUAD(setup,x,y,mask) do {\
+      setup->quad.input.x0 = x;\
+      setup->quad.input.y0 = y;\
+      setup->quad.inout.mask = mask;\
+      emit_quad( setup, &setup->quad, 0 );\
+   } while (0)
+
+#endif
+
+/**
+ * Given an X or Y coordinate, return the block/quad coordinate that it
+ * belongs to.
+ */
+static INLINE int block( int x )
+{
+   return x & ~1;
+}
+
+
+/**
+ * Render a horizontal span of quads
+ */
+static void flush_spans( struct setup_context *setup )
+{
+   const int xleft0 = setup->span.left[0];
+   const int xleft1 = setup->span.left[1];
+   const int xright0 = setup->span.right[0];
+   const int xright1 = setup->span.right[1];
+   int minleft, maxright;
+   int x;
+
+   switch (setup->span.y_flags) {
+   case 0x3:
+      /* both odd and even lines written (both quad rows) */
+      minleft = block(MIN2(xleft0, xleft1));
+      maxright = block(MAX2(xright0, xright1));
+      for (x = minleft; x <= maxright; x += 2) {
+         /* determine which of the four pixels is inside the span bounds */
+         uint mask = 0x0;
+         if (x >= xleft0 && x < xright0)
+            mask |= MASK_TOP_LEFT;
+         if (x >= xleft1 && x < xright1)
+            mask |= MASK_BOTTOM_LEFT;
+         if (x+1 >= xleft0 && x+1 < xright0)
+            mask |= MASK_TOP_RIGHT;
+         if (x+1 >= xleft1 && x+1 < xright1)
+            mask |= MASK_BOTTOM_RIGHT;
+         EMIT_QUAD( setup, x, setup->span.y, mask );
+      }
+      break;
+
+   case 0x1:
+      /* only even line written (quad top row) */
+      minleft = block(xleft0);
+      maxright = block(xright0);
+      for (x = minleft; x <= maxright; x += 2) {
+         uint mask = 0x0;
+         if (x >= xleft0 && x < xright0)
+            mask |= MASK_TOP_LEFT;
+         if (x+1 >= xleft0 && x+1 < xright0)
+            mask |= MASK_TOP_RIGHT;
+         EMIT_QUAD( setup, x, setup->span.y, mask );
+      }
+      break;
+
+   case 0x2:
+      /* only odd line written (quad bottom row) */
+      minleft = block(xleft1);
+      maxright = block(xright1);
+      for (x = minleft; x <= maxright; x += 2) {
+         uint mask = 0x0;
+         if (x >= xleft1 && x < xright1)
+            mask |= MASK_BOTTOM_LEFT;
+         if (x+1 >= xleft1 && x+1 < xright1)
+            mask |= MASK_BOTTOM_RIGHT;
+         EMIT_QUAD( setup, x, setup->span.y, mask );
+      }
+      break;
+
+   default:
+      return;
+   }
+
+   setup->span.y = 0;
+   setup->span.y_flags = 0;
+   setup->span.right[0] = 0;
+   setup->span.right[1] = 0;
+}
+
+
+#if DEBUG_VERTS
+static void print_vertex(const struct setup_context *setup,
+                         const float (*v)[4])
+{
+   int i;
+   debug_printf("   Vertex: (%p)\n", v);
+   for (i = 0; i < setup->quad.nr_attrs; i++) {
+      debug_printf("     %d: %f %f %f %f\n",  i,
+              v[i][0], v[i][1], v[i][2], v[i][3]);
+   }
+}
+#endif
+
+/**
+ * \return FALSE if coords are inf/nan (cull the tri), TRUE otherwise
+ */
+static boolean setup_sort_vertices( struct setup_context *setup,
+                                    float det,
+                                    const float (*v0)[4],
+                                    const float (*v1)[4],
+                                    const float (*v2)[4] )
+{
+   setup->vprovoke = v2;
+
+   /* determine bottom to top order of vertices */
+   {
+      float y0 = v0[0][1];
+      float y1 = v1[0][1];
+      float y2 = v2[0][1];
+      if (y0 <= y1) {
+	 if (y1 <= y2) {
+	    /* y0<=y1<=y2 */
+	    setup->vmin = v0;
+	    setup->vmid = v1;
+	    setup->vmax = v2;
+	 }
+	 else if (y2 <= y0) {
+	    /* y2<=y0<=y1 */
+	    setup->vmin = v2;
+	    setup->vmid = v0;
+	    setup->vmax = v1;
+	 }
+	 else {
+	    /* y0<=y2<=y1 */
+	    setup->vmin = v0;
+	    setup->vmid = v2;
+	    setup->vmax = v1;
+	 }
+      }
+      else {
+	 if (y0 <= y2) {
+	    /* y1<=y0<=y2 */
+	    setup->vmin = v1;
+	    setup->vmid = v0;
+	    setup->vmax = v2;
+	 }
+	 else if (y2 <= y1) {
+	    /* y2<=y1<=y0 */
+	    setup->vmin = v2;
+	    setup->vmid = v1;
+	    setup->vmax = v0;
+	 }
+	 else {
+	    /* y1<=y2<=y0 */
+	    setup->vmin = v1;
+	    setup->vmid = v2;
+	    setup->vmax = v0;
+	 }
+      }
+   }
+
+   setup->ebot.dx = setup->vmid[0][0] - setup->vmin[0][0];
+   setup->ebot.dy = setup->vmid[0][1] - setup->vmin[0][1];
+   setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0];
+   setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1];
+   setup->etop.dx = setup->vmax[0][0] - setup->vmid[0][0];
+   setup->etop.dy = setup->vmax[0][1] - setup->vmid[0][1];
+
+   /*
+    * Compute triangle's area.  Use 1/area to compute partial
+    * derivatives of attributes later.
+    *
+    * The area will be the same as prim->det, but the sign may be
+    * different depending on how the vertices get sorted above.
+    *
+    * To determine whether the primitive is front or back facing we
+    * use the prim->det value because its sign is correct.
+    */
+   {
+      const float area = (setup->emaj.dx * setup->ebot.dy -
+			    setup->ebot.dx * setup->emaj.dy);
+
+      setup->oneoverarea = 1.0f / area;
+
+      /*
+      debug_printf("%s one-over-area %f  area %f  det %f\n",
+                   __FUNCTION__, setup->oneoverarea, area, det );
+      */
+      if (is_inf_or_nan(setup->oneoverarea))
+         return FALSE;
+   }
+
+   /* We need to know if this is a front or back-facing triangle for:
+    *  - the GLSL gl_FrontFacing fragment attribute (bool)
+    *  - two-sided stencil test
+    */
+   setup->quad.input.facing = (det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW);
+
+   return TRUE;
+}
+
+
+/**
+ * Compute a0 for a constant-valued coefficient (GL_FLAT shading).
+ * The value value comes from vertex[slot][i].
+ * The result will be put into setup->coef[slot].a0[i].
+ * \param slot  which attribute slot
+ * \param i  which component of the slot (0..3)
+ */
+static void const_coeff( struct setup_context *setup,
+                         struct tgsi_interp_coef *coef,
+                         uint vertSlot, uint i)
+{
+   assert(i <= 3);
+
+   coef->dadx[i] = 0;
+   coef->dady[i] = 0;
+
+   /* need provoking vertex info!
+    */
+   coef->a0[i] = setup->vprovoke[vertSlot][i];
+}
+
+
+/**
+ * Compute a0, dadx and dady for a linearly interpolated coefficient,
+ * for a triangle.
+ */
+static void tri_linear_coeff( struct setup_context *setup,
+                              struct tgsi_interp_coef *coef,
+                              uint vertSlot, uint i)
+{
+   float botda = setup->vmid[vertSlot][i] - setup->vmin[vertSlot][i];
+   float majda = setup->vmax[vertSlot][i] - setup->vmin[vertSlot][i];
+   float a = setup->ebot.dy * majda - botda * setup->emaj.dy;
+   float b = setup->emaj.dx * botda - majda * setup->ebot.dx;
+   float dadx = a * setup->oneoverarea;
+   float dady = b * setup->oneoverarea;
+
+   assert(i <= 3);
+
+   coef->dadx[i] = dadx;
+   coef->dady[i] = dady;
+
+   /* calculate a0 as the value which would be sampled for the
+    * fragment at (0,0), taking into account that we want to sample at
+    * pixel centers, in other words (0.5, 0.5).
+    *
+    * this is neat but unfortunately not a good way to do things for
+    * triangles with very large values of dadx or dady as it will
+    * result in the subtraction and re-addition from a0 of a very
+    * large number, which means we'll end up loosing a lot of the
+    * fractional bits and precision from a0.  the way to fix this is
+    * to define a0 as the sample at a pixel center somewhere near vmin
+    * instead - i'll switch to this later.
+    */
+   coef->a0[i] = (setup->vmin[vertSlot][i] -
+                  (dadx * (setup->vmin[0][0] - 0.5f) +
+                   dady * (setup->vmin[0][1] - 0.5f)));
+
+   /*
+   debug_printf("attr[%d].%c: %f dx:%f dy:%f\n",
+		slot, "xyzw"[i],
+		setup->coef[slot].a0[i],
+		setup->coef[slot].dadx[i],
+		setup->coef[slot].dady[i]);
+   */
+}
+
+
+/**
+ * Compute a0, dadx and dady for a perspective-corrected interpolant,
+ * for a triangle.
+ * We basically multiply the vertex value by 1/w before computing
+ * the plane coefficients (a0, dadx, dady).
+ * Later, when we compute the value at a particular fragment position we'll
+ * divide the interpolated value by the interpolated W at that fragment.
+ */
+static void tri_persp_coeff( struct setup_context *setup,
+                             struct tgsi_interp_coef *coef,
+                             uint vertSlot, uint i)
+{
+   /* premultiply by 1/w  (v[0][3] is always W):
+    */
+   float mina = setup->vmin[vertSlot][i] * setup->vmin[0][3];
+   float mida = setup->vmid[vertSlot][i] * setup->vmid[0][3];
+   float maxa = setup->vmax[vertSlot][i] * setup->vmax[0][3];
+   float botda = mida - mina;
+   float majda = maxa - mina;
+   float a = setup->ebot.dy * majda - botda * setup->emaj.dy;
+   float b = setup->emaj.dx * botda - majda * setup->ebot.dx;
+   float dadx = a * setup->oneoverarea;
+   float dady = b * setup->oneoverarea;
+
+   /*
+   debug_printf("tri persp %d,%d: %f %f %f\n", vertSlot, i,
+          	setup->vmin[vertSlot][i],
+          	setup->vmid[vertSlot][i],
+       		setup->vmax[vertSlot][i]
+          );
+   */
+   assert(i <= 3);
+
+   coef->dadx[i] = dadx;
+   coef->dady[i] = dady;
+   coef->a0[i] = (mina -
+                  (dadx * (setup->vmin[0][0] - 0.5f) +
+                   dady * (setup->vmin[0][1] - 0.5f)));
+}
+
+
+/**
+ * Special coefficient setup for gl_FragCoord.
+ * X and Y are trivial, though Y has to be inverted for OpenGL.
+ * Z and W are copied from posCoef which should have already been computed.
+ * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask.
+ */
+static void
+setup_fragcoord_coeff(struct setup_context *setup, uint slot)
+{
+   /*X*/
+   setup->coef[slot].a0[0] = 0;
+   setup->coef[slot].dadx[0] = 1.0;
+   setup->coef[slot].dady[0] = 0.0;
+   /*Y*/
+   if (setup->softpipe->rasterizer->origin_lower_left) {
+      /* y=0=bottom */
+      const int winHeight = setup->softpipe->framebuffer.height;
+      setup->coef[slot].a0[1] = (float) (winHeight - 1);
+      setup->coef[slot].dady[1] = -1.0;
+   }
+   else {
+      /* y=0=top */
+      setup->coef[slot].a0[1] = 0.0;
+      setup->coef[slot].dady[1] = 1.0;
+   }
+   setup->coef[slot].dadx[1] = 0.0;
+   /*Z*/
+   setup->coef[slot].a0[2] = setup->posCoef.a0[2];
+   setup->coef[slot].dadx[2] = setup->posCoef.dadx[2];
+   setup->coef[slot].dady[2] = setup->posCoef.dady[2];
+   /*W*/
+   setup->coef[slot].a0[3] = setup->posCoef.a0[3];
+   setup->coef[slot].dadx[3] = setup->posCoef.dadx[3];
+   setup->coef[slot].dady[3] = setup->posCoef.dady[3];
+}
+
+
+
+/**
+ * Compute the setup->coef[] array dadx, dady, a0 values.
+ * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized.
+ */
+static void setup_tri_coefficients( struct setup_context *setup )
+{
+   struct softpipe_context *softpipe = setup->softpipe;
+   const struct sp_fragment_shader *spfs = softpipe->fs;
+   const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
+   uint fragSlot;
+
+   /* z and w are done by linear interpolation:
+    */
+   tri_linear_coeff(setup, &setup->posCoef, 0, 2);
+   tri_linear_coeff(setup, &setup->posCoef, 0, 3);
+
+   /* setup interpolation for all the remaining attributes:
+    */
+   for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) {
+      const uint vertSlot = vinfo->attrib[fragSlot].src_index;
+      uint j;
+
+      switch (vinfo->attrib[fragSlot].interp_mode) {
+      case INTERP_CONSTANT:
+         for (j = 0; j < NUM_CHANNELS; j++)
+            const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
+         break;
+      case INTERP_LINEAR:
+         for (j = 0; j < NUM_CHANNELS; j++)
+            tri_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
+         break;
+      case INTERP_PERSPECTIVE:
+         for (j = 0; j < NUM_CHANNELS; j++)
+            tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
+         break;
+      case INTERP_POS:
+         setup_fragcoord_coeff(setup, fragSlot);
+         break;
+      default:
+         assert(0);
+      }
+
+      if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) {
+         /* FOG.y = front/back facing  XXX fix this */
+         setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing;
+         setup->coef[fragSlot].dadx[1] = 0.0;
+         setup->coef[fragSlot].dady[1] = 0.0;
+      }
+   }
+}
+
+
+
+static void setup_tri_edges( struct setup_context *setup )
+{
+   float vmin_x = setup->vmin[0][0] + 0.5f;
+   float vmid_x = setup->vmid[0][0] + 0.5f;
+
+   float vmin_y = setup->vmin[0][1] - 0.5f;
+   float vmid_y = setup->vmid[0][1] - 0.5f;
+   float vmax_y = setup->vmax[0][1] - 0.5f;
+
+   setup->emaj.sy = ceilf(vmin_y);
+   setup->emaj.lines = (int) ceilf(vmax_y - setup->emaj.sy);
+   setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy;
+   setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy;
+
+   setup->etop.sy = ceilf(vmid_y);
+   setup->etop.lines = (int) ceilf(vmax_y - setup->etop.sy);
+   setup->etop.dxdy = setup->etop.dx / setup->etop.dy;
+   setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy;
+
+   setup->ebot.sy = ceilf(vmin_y);
+   setup->ebot.lines = (int) ceilf(vmid_y - setup->ebot.sy);
+   setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy;
+   setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy;
+}
+
+
+/**
+ * Render the upper or lower half of a triangle.
+ * Scissoring/cliprect is applied here too.
+ */
+static void subtriangle( struct setup_context *setup,
+			 struct edge *eleft,
+			 struct edge *eright,
+			 unsigned lines )
+{
+   const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect;
+   const int minx = (int) cliprect->minx;
+   const int maxx = (int) cliprect->maxx;
+   const int miny = (int) cliprect->miny;
+   const int maxy = (int) cliprect->maxy;
+   int y, start_y, finish_y;
+   int sy = (int)eleft->sy;
+
+   assert((int)eleft->sy == (int) eright->sy);
+
+   /* clip top/bottom */
+   start_y = sy;
+   finish_y = sy + lines;
+
+   if (start_y < miny)
+      start_y = miny;
+
+   if (finish_y > maxy)
+      finish_y = maxy;
+
+   start_y -= sy;
+   finish_y -= sy;
+
+   /*
+   debug_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y);
+   */
+
+   for (y = start_y; y < finish_y; y++) {
+
+      /* avoid accumulating adds as floats don't have the precision to
+       * accurately iterate large triangle edges that way.  luckily we
+       * can just multiply these days.
+       *
+       * this is all drowned out by the attribute interpolation anyway.
+       */
+      int left = (int)(eleft->sx + y * eleft->dxdy);
+      int right = (int)(eright->sx + y * eright->dxdy);
+
+      /* clip left/right */
+      if (left < minx)
+         left = minx;
+      if (right > maxx)
+         right = maxx;
+
+      if (left < right) {
+         int _y = sy + y;
+         if (block(_y) != setup->span.y) {
+            flush_spans(setup);
+            setup->span.y = block(_y);
+         }
+
+         setup->span.left[_y&1] = left;
+         setup->span.right[_y&1] = right;
+         setup->span.y_flags |= 1<<(_y&1);
+      }
+   }
+
+
+   /* save the values so that emaj can be restarted:
+    */
+   eleft->sx += lines * eleft->dxdy;
+   eright->sx += lines * eright->dxdy;
+   eleft->sy += lines;
+   eright->sy += lines;
+}
+
+
+/**
+ * Recalculate prim's determinant.  This is needed as we don't have
+ * get this information through the vbuf_render interface & we must
+ * calculate it here.
+ */
+static float
+calc_det( const float (*v0)[4],
+          const float (*v1)[4],
+          const float (*v2)[4] )
+{
+   /* edge vectors e = v0 - v2, f = v1 - v2 */
+   const float ex = v0[0][0] - v2[0][0];
+   const float ey = v0[0][1] - v2[0][1];
+   const float fx = v1[0][0] - v2[0][0];
+   const float fy = v1[0][1] - v2[0][1];
+
+   /* det = cross(e,f).z */
+   return ex * fy - ey * fx;
+}
+
+
+/**
+ * Do setup for triangle rasterization, then render the triangle.
+ */
+void setup_tri( struct setup_context *setup,
+                const float (*v0)[4],
+                const float (*v1)[4],
+                const float (*v2)[4] )
+{
+   float det;
+
+#if DEBUG_VERTS
+   debug_printf("Setup triangle:\n");
+   print_vertex(setup, v0);
+   print_vertex(setup, v1);
+   print_vertex(setup, v2);
+#endif
+
+   if (setup->softpipe->no_rast)
+      return;
+   
+   det = calc_det(v0, v1, v2);
+   /*
+   debug_printf("%s\n", __FUNCTION__ );
+   */
+
+#if DEBUG_FRAGS
+   setup->numFragsEmitted = 0;
+   setup->numFragsWritten = 0;
+#endif
+
+   if (cull_tri( setup, det ))
+      return;
+
+   if (!setup_sort_vertices( setup, det, v0, v1, v2 ))
+      return;
+   setup_tri_coefficients( setup );
+   setup_tri_edges( setup );
+
+   setup->quad.input.prim = PRIM_TRI;
+
+   setup->span.y = 0;
+   setup->span.y_flags = 0;
+   setup->span.right[0] = 0;
+   setup->span.right[1] = 0;
+   /*   setup->span.z_mode = tri_z_mode( setup->ctx ); */
+
+   /*   init_constant_attribs( setup ); */
+
+   if (setup->oneoverarea < 0.0) {
+      /* emaj on left:
+       */
+      subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines );
+      subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines );
+   }
+   else {
+      /* emaj on right:
+       */
+      subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines );
+      subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines );
+   }
+
+   flush_spans( setup );
+
+   WAIT_FOR_COMPLETION(setup);
+
+#if DEBUG_FRAGS
+   printf("Tri: %u frags emitted, %u written\n",
+          setup->numFragsEmitted,
+          setup->numFragsWritten);
+#endif
+}
+
+
+
+/**
+ * Compute a0, dadx and dady for a linearly interpolated coefficient,
+ * for a line.
+ */
+static void
+line_linear_coeff(struct setup_context *setup,
+                  struct tgsi_interp_coef *coef,
+                  uint vertSlot, uint i)
+{
+   const float da = setup->vmax[vertSlot][i] - setup->vmin[vertSlot][i];
+   const float dadx = da * setup->emaj.dx * setup->oneoverarea;
+   const float dady = da * setup->emaj.dy * setup->oneoverarea;
+   coef->dadx[i] = dadx;
+   coef->dady[i] = dady;
+   coef->a0[i] = (setup->vmin[vertSlot][i] -
+                  (dadx * (setup->vmin[0][0] - 0.5f) +
+                   dady * (setup->vmin[0][1] - 0.5f)));
+}
+
+
+/**
+ * Compute a0, dadx and dady for a perspective-corrected interpolant,
+ * for a line.
+ */
+static void
+line_persp_coeff(struct setup_context *setup,
+                  struct tgsi_interp_coef *coef,
+                  uint vertSlot, uint i)
+{
+   /* XXX double-check/verify this arithmetic */
+   const float a0 = setup->vmin[vertSlot][i] * setup->vmin[0][3];
+   const float a1 = setup->vmax[vertSlot][i] * setup->vmax[0][3];
+   const float da = a1 - a0;
+   const float dadx = da * setup->emaj.dx * setup->oneoverarea;
+   const float dady = da * setup->emaj.dy * setup->oneoverarea;
+   coef->dadx[i] = dadx;
+   coef->dady[i] = dady;
+   coef->a0[i] = (setup->vmin[vertSlot][i] -
+                  (dadx * (setup->vmin[0][0] - 0.5f) +
+                   dady * (setup->vmin[0][1] - 0.5f)));
+}
+
+
+/**
+ * Compute the setup->coef[] array dadx, dady, a0 values.
+ * Must be called after setup->vmin,vmax are initialized.
+ */
+static INLINE boolean
+setup_line_coefficients(struct setup_context *setup,
+                        const float (*v0)[4],
+                        const float (*v1)[4])
+{
+   struct softpipe_context *softpipe = setup->softpipe;
+   const struct sp_fragment_shader *spfs = softpipe->fs;
+   const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
+   uint fragSlot;
+   float area;
+
+   /* use setup->vmin, vmax to point to vertices */
+   setup->vprovoke = v1;
+   setup->vmin = v0;
+   setup->vmax = v1;
+
+   setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0];
+   setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1];
+
+   /* NOTE: this is not really area but something proportional to it */
+   area = setup->emaj.dx * setup->emaj.dx + setup->emaj.dy * setup->emaj.dy;
+   if (area == 0.0f || is_inf_or_nan(area))
+      return FALSE;
+   setup->oneoverarea = 1.0f / area;
+
+   /* z and w are done by linear interpolation:
+    */
+   line_linear_coeff(setup, &setup->posCoef, 0, 2);
+   line_linear_coeff(setup, &setup->posCoef, 0, 3);
+
+   /* setup interpolation for all the remaining attributes:
+    */
+   for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) {
+      const uint vertSlot = vinfo->attrib[fragSlot].src_index;
+      uint j;
+
+      switch (vinfo->attrib[fragSlot].interp_mode) {
+      case INTERP_CONSTANT:
+         for (j = 0; j < NUM_CHANNELS; j++)
+            const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
+         break;
+      case INTERP_LINEAR:
+         for (j = 0; j < NUM_CHANNELS; j++)
+            line_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
+         break;
+      case INTERP_PERSPECTIVE:
+         for (j = 0; j < NUM_CHANNELS; j++)
+            line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
+         break;
+      case INTERP_POS:
+         setup_fragcoord_coeff(setup, fragSlot);
+         break;
+      default:
+         assert(0);
+      }
+
+      if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) {
+         /* FOG.y = front/back facing  XXX fix this */
+         setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing;
+         setup->coef[fragSlot].dadx[1] = 0.0;
+         setup->coef[fragSlot].dady[1] = 0.0;
+      }
+   }
+   return TRUE;
+}
+
+
+/**
+ * Plot a pixel in a line segment.
+ */
+static INLINE void
+plot(struct setup_context *setup, int x, int y)
+{
+   const int iy = y & 1;
+   const int ix = x & 1;
+   const int quadX = x - ix;
+   const int quadY = y - iy;
+   const int mask = (1 << ix) << (2 * iy);
+
+   if (quadX != setup->quad.input.x0 ||
+       quadY != setup->quad.input.y0)
+   {
+      /* flush prev quad, start new quad */
+
+      if (setup->quad.input.x0 != -1)
+         CLIP_EMIT_QUAD(setup);
+
+      setup->quad.input.x0 = quadX;
+      setup->quad.input.y0 = quadY;
+      setup->quad.inout.mask = 0x0;
+   }
+
+   setup->quad.inout.mask |= mask;
+}
+
+
+/**
+ * Do setup for line rasterization, then render the line.
+ * Single-pixel width, no stipple, etc.  We rely on the 'draw' module
+ * to handle stippling and wide lines.
+ */
+void
+setup_line(struct setup_context *setup,
+           const float (*v0)[4],
+           const float (*v1)[4])
+{
+   int x0 = (int) v0[0][0];
+   int x1 = (int) v1[0][0];
+   int y0 = (int) v0[0][1];
+   int y1 = (int) v1[0][1];
+   int dx = x1 - x0;
+   int dy = y1 - y0;
+   int xstep, ystep;
+
+#if DEBUG_VERTS
+   debug_printf("Setup line:\n");
+   print_vertex(setup, v0);
+   print_vertex(setup, v1);
+#endif
+
+   if (setup->softpipe->no_rast)
+      return;
+
+   if (dx == 0 && dy == 0)
+      return;
+
+   if (!setup_line_coefficients(setup, v0, v1))
+      return;
+
+   assert(v0[0][0] < 1.0e9);
+   assert(v0[0][1] < 1.0e9);
+   assert(v1[0][0] < 1.0e9);
+   assert(v1[0][1] < 1.0e9);
+
+   if (dx < 0) {
+      dx = -dx;   /* make positive */
+      xstep = -1;
+   }
+   else {
+      xstep = 1;
+   }
+
+   if (dy < 0) {
+      dy = -dy;   /* make positive */
+      ystep = -1;
+   }
+   else {
+      ystep = 1;
+   }
+
+   assert(dx >= 0);
+   assert(dy >= 0);
+
+   setup->quad.input.x0 = setup->quad.input.y0 = -1;
+   setup->quad.inout.mask = 0x0;
+   setup->quad.input.prim = PRIM_LINE;
+   /* XXX temporary: set coverage to 1.0 so the line appears
+    * if AA mode happens to be enabled.
+    */
+   setup->quad.input.coverage[0] =
+   setup->quad.input.coverage[1] =
+   setup->quad.input.coverage[2] =
+   setup->quad.input.coverage[3] = 1.0;
+
+   if (dx > dy) {
+      /*** X-major line ***/
+      int i;
+      const int errorInc = dy + dy;
+      int error = errorInc - dx;
+      const int errorDec = error - dx;
+
+      for (i = 0; i < dx; i++) {
+         plot(setup, x0, y0);
+
+         x0 += xstep;
+         if (error < 0) {
+            error += errorInc;
+         }
+         else {
+            error += errorDec;
+            y0 += ystep;
+         }
+      }
+   }
+   else {
+      /*** Y-major line ***/
+      int i;
+      const int errorInc = dx + dx;
+      int error = errorInc - dy;
+      const int errorDec = error - dy;
+
+      for (i = 0; i < dy; i++) {
+         plot(setup, x0, y0);
+
+         y0 += ystep;
+         if (error < 0) {
+            error += errorInc;
+         }
+         else {
+            error += errorDec;
+            x0 += xstep;
+         }
+      }
+   }
+
+   /* draw final quad */
+   if (setup->quad.inout.mask) {
+      CLIP_EMIT_QUAD(setup);
+   }
+
+   WAIT_FOR_COMPLETION(setup);
+}
+
+
+static void
+point_persp_coeff(struct setup_context *setup,
+                  const float (*vert)[4],
+                  struct tgsi_interp_coef *coef,
+                  uint vertSlot, uint i)
+{
+   assert(i <= 3);
+   coef->dadx[i] = 0.0F;
+   coef->dady[i] = 0.0F;
+   coef->a0[i] = vert[vertSlot][i] * vert[0][3];
+}
+
+
+/**
+ * Do setup for point rasterization, then render the point.
+ * Round or square points...
+ * XXX could optimize a lot for 1-pixel points.
+ */
+void
+setup_point( struct setup_context *setup,
+             const float (*v0)[4] )
+{
+   struct softpipe_context *softpipe = setup->softpipe;
+   const struct sp_fragment_shader *spfs = softpipe->fs;
+   const int sizeAttr = setup->softpipe->psize_slot;
+   const float size
+      = sizeAttr > 0 ? v0[sizeAttr][0]
+      : setup->softpipe->rasterizer->point_size;
+   const float halfSize = 0.5F * size;
+   const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth;
+   const float x = v0[0][0];  /* Note: data[0] is always position */
+   const float y = v0[0][1];
+   const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
+   uint fragSlot;
+
+#if DEBUG_VERTS
+   debug_printf("Setup point:\n");
+   print_vertex(setup, v0);
+#endif
+
+   if (softpipe->no_rast)
+      return;
+
+   /* For points, all interpolants are constant-valued.
+    * However, for point sprites, we'll need to setup texcoords appropriately.
+    * XXX: which coefficients are the texcoords???
+    * We may do point sprites as textured quads...
+    *
+    * KW: We don't know which coefficients are texcoords - ultimately
+    * the choice of what interpolation mode to use for each attribute
+    * should be determined by the fragment program, using
+    * per-attribute declaration statements that include interpolation
+    * mode as a parameter.  So either the fragment program will have
+    * to be adjusted for pointsprite vs normal point behaviour, or
+    * otherwise a special interpolation mode will have to be defined
+    * which matches the required behaviour for point sprites.  But -
+    * the latter is not a feature of normal hardware, and as such
+    * probably should be ruled out on that basis.
+    */
+   setup->vprovoke = v0;
+
+   /* setup Z, W */
+   const_coeff(setup, &setup->posCoef, 0, 2);
+   const_coeff(setup, &setup->posCoef, 0, 3);
+
+   for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) {
+      const uint vertSlot = vinfo->attrib[fragSlot].src_index;
+      uint j;
+
+      switch (vinfo->attrib[fragSlot].interp_mode) {
+      case INTERP_CONSTANT:
+         /* fall-through */
+      case INTERP_LINEAR:
+         for (j = 0; j < NUM_CHANNELS; j++)
+            const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
+         break;
+      case INTERP_PERSPECTIVE:
+         for (j = 0; j < NUM_CHANNELS; j++)
+            point_persp_coeff(setup, setup->vprovoke,
+                              &setup->coef[fragSlot], vertSlot, j);
+         break;
+      case INTERP_POS:
+         setup_fragcoord_coeff(setup, fragSlot);
+         break;
+      default:
+         assert(0);
+      }
+
+      if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) {
+         /* FOG.y = front/back facing  XXX fix this */
+         setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing;
+         setup->coef[fragSlot].dadx[1] = 0.0;
+         setup->coef[fragSlot].dady[1] = 0.0;
+      }
+   }
+
+   setup->quad.input.prim = PRIM_POINT;
+
+   if (halfSize <= 0.5 && !round) {
+      /* special case for 1-pixel points */
+      const int ix = ((int) x) & 1;
+      const int iy = ((int) y) & 1;
+      setup->quad.input.x0 = (int) x - ix;
+      setup->quad.input.y0 = (int) y - iy;
+      setup->quad.inout.mask = (1 << ix) << (2 * iy);
+      CLIP_EMIT_QUAD(setup);
+   }
+   else {
+      if (round) {
+         /* rounded points */
+         const int ixmin = block((int) (x - halfSize));
+         const int ixmax = block((int) (x + halfSize));
+         const int iymin = block((int) (y - halfSize));
+         const int iymax = block((int) (y + halfSize));
+         const float rmin = halfSize - 0.7071F;  /* 0.7071 = sqrt(2)/2 */
+         const float rmax = halfSize + 0.7071F;
+         const float rmin2 = MAX2(0.0F, rmin * rmin);
+         const float rmax2 = rmax * rmax;
+         const float cscale = 1.0F / (rmax2 - rmin2);
+         int ix, iy;
+
+         for (iy = iymin; iy <= iymax; iy += 2) {
+            for (ix = ixmin; ix <= ixmax; ix += 2) {
+               float dx, dy, dist2, cover;
+
+               setup->quad.inout.mask = 0x0;
+
+               dx = (ix + 0.5f) - x;
+               dy = (iy + 0.5f) - y;
+               dist2 = dx * dx + dy * dy;
+               if (dist2 <= rmax2) {
+                  cover = 1.0F - (dist2 - rmin2) * cscale;
+                  setup->quad.input.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f);
+                  setup->quad.inout.mask |= MASK_TOP_LEFT;
+               }
+
+               dx = (ix + 1.5f) - x;
+               dy = (iy + 0.5f) - y;
+               dist2 = dx * dx + dy * dy;
+               if (dist2 <= rmax2) {
+                  cover = 1.0F - (dist2 - rmin2) * cscale;
+                  setup->quad.input.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f);
+                  setup->quad.inout.mask |= MASK_TOP_RIGHT;
+               }
+
+               dx = (ix + 0.5f) - x;
+               dy = (iy + 1.5f) - y;
+               dist2 = dx * dx + dy * dy;
+               if (dist2 <= rmax2) {
+                  cover = 1.0F - (dist2 - rmin2) * cscale;
+                  setup->quad.input.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f);
+                  setup->quad.inout.mask |= MASK_BOTTOM_LEFT;
+               }
+
+               dx = (ix + 1.5f) - x;
+               dy = (iy + 1.5f) - y;
+               dist2 = dx * dx + dy * dy;
+               if (dist2 <= rmax2) {
+                  cover = 1.0F - (dist2 - rmin2) * cscale;
+                  setup->quad.input.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f);
+                  setup->quad.inout.mask |= MASK_BOTTOM_RIGHT;
+               }
+
+               if (setup->quad.inout.mask) {
+                  setup->quad.input.x0 = ix;
+                  setup->quad.input.y0 = iy;
+                  CLIP_EMIT_QUAD(setup);
+               }
+            }
+         }
+      }
+      else {
+         /* square points */
+         const int xmin = (int) (x + 0.75 - halfSize);
+         const int ymin = (int) (y + 0.25 - halfSize);
+         const int xmax = xmin + (int) size;
+         const int ymax = ymin + (int) size;
+         /* XXX could apply scissor to xmin,ymin,xmax,ymax now */
+         const int ixmin = block(xmin);
+         const int ixmax = block(xmax - 1);
+         const int iymin = block(ymin);
+         const int iymax = block(ymax - 1);
+         int ix, iy;
+
+         /*
+         debug_printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax);
+         */
+         for (iy = iymin; iy <= iymax; iy += 2) {
+            uint rowMask = 0xf;
+            if (iy < ymin) {
+               /* above the top edge */
+               rowMask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
+            }
+            if (iy + 1 >= ymax) {
+               /* below the bottom edge */
+               rowMask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
+            }
+
+            for (ix = ixmin; ix <= ixmax; ix += 2) {
+               uint mask = rowMask;
+
+               if (ix < xmin) {
+                  /* fragment is past left edge of point, turn off left bits */
+                  mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
+               }
+               if (ix + 1 >= xmax) {
+                  /* past the right edge */
+                  mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
+               }
+
+               setup->quad.inout.mask = mask;
+               setup->quad.input.x0 = ix;
+               setup->quad.input.y0 = iy;
+               CLIP_EMIT_QUAD(setup);
+            }
+         }
+      }
+   }
+
+   WAIT_FOR_COMPLETION(setup);
+}
+
+void setup_prepare( struct setup_context *setup )
+{
+   struct softpipe_context *sp = setup->softpipe;
+   unsigned i;
+
+   if (sp->dirty) {
+      softpipe_update_derived(sp);
+   }
+
+   /* Mark surfaces as defined now */
+   for (i = 0; i < sp->framebuffer.nr_cbufs; i++){
+      if (sp->framebuffer.cbufs[i]) {
+         sp->framebuffer.cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED;
+      }
+   }
+   if (sp->framebuffer.zsbuf) {
+      sp->framebuffer.zsbuf->status = PIPE_SURFACE_STATUS_DEFINED;
+   }
+
+   /* Note: nr_attrs is only used for debugging (vertex printing) */
+   setup->quad.nr_attrs = draw_num_vs_outputs(sp->draw);
+
+   for (i = 0; i < SP_NUM_QUAD_THREADS; i++) {
+      sp->quad[i].first->begin( sp->quad[i].first );
+   }
+
+   if (sp->reduced_api_prim == PIPE_PRIM_TRIANGLES &&
+       sp->rasterizer->fill_cw == PIPE_POLYGON_MODE_FILL &&
+       sp->rasterizer->fill_ccw == PIPE_POLYGON_MODE_FILL) {
+      /* we'll do culling */
+      setup->winding = sp->rasterizer->cull_mode;
+   }
+   else {
+      /* 'draw' will do culling */
+      setup->winding = PIPE_WINDING_NONE;
+   }
+}
+
+
+
+void setup_destroy_context( struct setup_context *setup )
+{
+   FREE( setup );
+}
+
+
+/**
+ * Create a new primitive setup/render stage.
+ */
+struct setup_context *setup_create_context( struct softpipe_context *softpipe )
+{
+   struct setup_context *setup = CALLOC_STRUCT(setup_context);
+#if SP_NUM_QUAD_THREADS > 1
+   uint i;
+#endif
+
+   setup->softpipe = softpipe;
+
+   setup->quad.coef = setup->coef;
+   setup->quad.posCoef = &setup->posCoef;
+
+#if SP_NUM_QUAD_THREADS > 1
+   setup->que.first = 0;
+   setup->que.last = 0;
+   pipe_mutex_init( setup->que.que_mutex );
+   pipe_condvar_init( setup->que.que_notfull_condvar );
+   pipe_condvar_init( setup->que.que_notempty_condvar );
+   setup->que.jobs_added = 0;
+   setup->que.jobs_done = 0;
+   pipe_condvar_init( setup->que.que_done_condvar );
+   for (i = 0; i < SP_NUM_QUAD_THREADS; i++) {
+      setup->threads[i].setup = setup;
+      setup->threads[i].id = i;
+      setup->threads[i].handle = pipe_thread_create( quad_thread, &setup->threads[i] );
+   }
+#endif
+
+   return setup;
+}
+
diff --git a/src/gallium/drivers/softpipe/sp_setup.h b/src/gallium/drivers/softpipe/sp_setup.h
new file mode 100644
index 0000000000..d54f334428
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_setup.h
@@ -0,0 +1,53 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+#ifndef SP_SETUP_H
+#define SP_SETUP_H
+
+struct setup_context;
+struct softpipe_context;
+
+void 
+setup_tri( struct setup_context *setup,
+	   const float (*v0)[4],
+	   const float (*v1)[4],
+	   const float (*v2)[4] );
+
+void
+setup_line(struct setup_context *setup,
+           const float (*v0)[4],
+           const float (*v1)[4]);
+
+void
+setup_point( struct setup_context *setup,
+             const float (*v0)[4] );
+
+
+struct setup_context *setup_create_context( struct softpipe_context *softpipe );
+void setup_prepare( struct setup_context *setup );
+void setup_destroy_context( struct setup_context *setup );
+
+#endif
diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h
new file mode 100644
index 0000000000..3eff41ffa5
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_state.h
@@ -0,0 +1,206 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef SP_STATE_H
+#define SP_STATE_H
+
+#include "pipe/p_state.h"
+#include "tgsi/tgsi_scan.h"
+
+
+#define SP_NEW_VIEWPORT      0x1
+#define SP_NEW_RASTERIZER    0x2
+#define SP_NEW_FS            0x4
+#define SP_NEW_BLEND         0x8
+#define SP_NEW_CLIP          0x10
+#define SP_NEW_SCISSOR       0x20
+#define SP_NEW_STIPPLE       0x40
+#define SP_NEW_FRAMEBUFFER   0x80
+#define SP_NEW_DEPTH_STENCIL_ALPHA 0x100
+#define SP_NEW_CONSTANTS     0x200
+#define SP_NEW_SAMPLER       0x400
+#define SP_NEW_TEXTURE       0x800
+#define SP_NEW_VERTEX        0x1000
+#define SP_NEW_VS            0x2000
+#define SP_NEW_QUERY         0x4000
+
+
+struct tgsi_sampler;
+struct tgsi_exec_machine;
+struct vertex_info;
+
+
+/**
+ * Subclass of pipe_shader_state (though it doesn't really need to be).
+ *
+ * This is starting to look an awful lot like a quad pipeline stage...
+ */
+struct sp_fragment_shader {
+   struct pipe_shader_state shader;
+
+   struct tgsi_shader_info info;
+
+   void (*prepare)( const struct sp_fragment_shader *shader,
+		    struct tgsi_exec_machine *machine,
+		    struct tgsi_sampler **samplers);
+
+   /* Run the shader - this interface will get cleaned up in the
+    * future:
+    */
+   unsigned (*run)( const struct sp_fragment_shader *shader,
+		    struct tgsi_exec_machine *machine,
+		    struct quad_header *quad );
+
+
+   void (*delete)( struct sp_fragment_shader * );
+};
+
+
+/** Subclass of pipe_shader_state */
+struct sp_vertex_shader {
+   struct pipe_shader_state shader;  /* Note: this field not actually used */
+   struct draw_vertex_shader *draw_data;
+};
+
+
+
+void *
+softpipe_create_blend_state(struct pipe_context *,
+                            const struct pipe_blend_state *);
+void softpipe_bind_blend_state(struct pipe_context *,
+                               void *);
+void softpipe_delete_blend_state(struct pipe_context *,
+                                 void *);
+
+void *
+softpipe_create_sampler_state(struct pipe_context *,
+                              const struct pipe_sampler_state *);
+void softpipe_bind_sampler_states(struct pipe_context *, unsigned, void **);
+void softpipe_delete_sampler_state(struct pipe_context *, void *);
+
+void *
+softpipe_create_depth_stencil_state(struct pipe_context *,
+                                    const struct pipe_depth_stencil_alpha_state *);
+void softpipe_bind_depth_stencil_state(struct pipe_context *, void *);
+void softpipe_delete_depth_stencil_state(struct pipe_context *, void *);
+
+void *
+softpipe_create_rasterizer_state(struct pipe_context *,
+                                 const struct pipe_rasterizer_state *);
+void softpipe_bind_rasterizer_state(struct pipe_context *, void *);
+void softpipe_delete_rasterizer_state(struct pipe_context *, void *);
+
+void softpipe_set_framebuffer_state( struct pipe_context *,
+			     const struct pipe_framebuffer_state * );
+
+void softpipe_set_blend_color( struct pipe_context *pipe,
+                               const struct pipe_blend_color *blend_color );
+
+void softpipe_set_clip_state( struct pipe_context *,
+			     const struct pipe_clip_state * );
+
+void softpipe_set_constant_buffer(struct pipe_context *,
+                                  uint shader, uint index,
+                                  const struct pipe_constant_buffer *buf);
+
+void *softpipe_create_fs_state(struct pipe_context *,
+                               const struct pipe_shader_state *);
+void softpipe_bind_fs_state(struct pipe_context *, void *);
+void softpipe_delete_fs_state(struct pipe_context *, void *);
+void *softpipe_create_vs_state(struct pipe_context *,
+                               const struct pipe_shader_state *);
+void softpipe_bind_vs_state(struct pipe_context *, void *);
+void softpipe_delete_vs_state(struct pipe_context *, void *);
+
+void softpipe_set_polygon_stipple( struct pipe_context *,
+				  const struct pipe_poly_stipple * );
+
+void softpipe_set_scissor_state( struct pipe_context *,
+                                 const struct pipe_scissor_state * );
+
+void softpipe_set_sampler_textures( struct pipe_context *,
+                                    unsigned num,
+                                    struct pipe_texture ** );
+
+void softpipe_set_viewport_state( struct pipe_context *,
+                                  const struct pipe_viewport_state * );
+
+void softpipe_set_vertex_elements(struct pipe_context *,
+                                  unsigned count,
+                                  const struct pipe_vertex_element *);
+
+void softpipe_set_vertex_buffers(struct pipe_context *,
+                                 unsigned count,
+                                 const struct pipe_vertex_buffer *);
+
+
+void softpipe_update_derived( struct softpipe_context *softpipe );
+
+
+boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
+			     unsigned start, unsigned count);
+
+boolean softpipe_draw_elements(struct pipe_context *pipe,
+			       struct pipe_buffer *indexBuffer,
+			       unsigned indexSize,
+			       unsigned mode, unsigned start, unsigned count);
+boolean
+softpipe_draw_range_elements(struct pipe_context *pipe,
+                             struct pipe_buffer *indexBuffer,
+                             unsigned indexSize,
+                             unsigned min_index,
+                             unsigned max_index,
+                             unsigned mode, unsigned start, unsigned count);
+
+void
+softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags);
+
+
+void
+softpipe_map_surfaces(struct softpipe_context *sp);
+
+void
+softpipe_unmap_surfaces(struct softpipe_context *sp);
+
+void
+softpipe_map_texture_surfaces(struct softpipe_context *sp);
+
+void
+softpipe_unmap_texture_surfaces(struct softpipe_context *sp);
+
+
+struct vertex_info *
+softpipe_get_vertex_info(struct softpipe_context *softpipe);
+
+struct vertex_info *
+softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe);
+
+
+#endif
diff --git a/src/gallium/drivers/softpipe/sp_state_blend.c b/src/gallium/drivers/softpipe/sp_state_blend.c
new file mode 100644
index 0000000000..384fe559af
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_state_blend.c
@@ -0,0 +1,98 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "util/u_memory.h"
+#include "sp_context.h"
+#include "sp_state.h"
+
+
+void *
+softpipe_create_blend_state(struct pipe_context *pipe,
+                            const struct pipe_blend_state *blend)
+{
+   return mem_dup(blend, sizeof(*blend));
+}
+
+void softpipe_bind_blend_state( struct pipe_context *pipe,
+                                void *blend )
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   softpipe->blend = (const struct pipe_blend_state *)blend;
+
+   softpipe->dirty |= SP_NEW_BLEND;
+}
+
+void softpipe_delete_blend_state(struct pipe_context *pipe,
+                                 void *blend)
+{
+   FREE( blend );
+}
+
+
+void softpipe_set_blend_color( struct pipe_context *pipe,
+			     const struct pipe_blend_color *blend_color )
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   softpipe->blend_color = *blend_color;
+
+   softpipe->dirty |= SP_NEW_BLEND;
+}
+
+
+/** XXX move someday?  Or consolidate all these simple state setters
+ * into one file.
+ */
+
+
+void *
+softpipe_create_depth_stencil_state(struct pipe_context *pipe,
+				    const struct pipe_depth_stencil_alpha_state *depth_stencil)
+{
+   return mem_dup(depth_stencil, sizeof(*depth_stencil));
+}
+
+void
+softpipe_bind_depth_stencil_state(struct pipe_context *pipe,
+                                  void *depth_stencil)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   softpipe->depth_stencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil;
+
+   softpipe->dirty |= SP_NEW_DEPTH_STENCIL_ALPHA;
+}
+
+void
+softpipe_delete_depth_stencil_state(struct pipe_context *pipe, void *depth)
+{
+   FREE( depth );
+}
diff --git a/src/gallium/drivers/softpipe/sp_state_clip.c b/src/gallium/drivers/softpipe/sp_state_clip.c
new file mode 100644
index 0000000000..4946c776e3
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_state_clip.c
@@ -0,0 +1,79 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+#include "sp_context.h"
+#include "sp_state.h"
+#include "draw/draw_context.h"
+
+
+void softpipe_set_clip_state( struct pipe_context *pipe,
+			     const struct pipe_clip_state *clip )
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   /* pass the clip state to the draw module */
+   draw_set_clip_state(softpipe->draw, clip);
+}
+
+
+void softpipe_set_viewport_state( struct pipe_context *pipe,
+                                  const struct pipe_viewport_state *viewport )
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   /* pass the viewport info to the draw module */
+   draw_set_viewport_state(softpipe->draw, viewport);
+
+   softpipe->viewport = *viewport; /* struct copy */
+   softpipe->dirty |= SP_NEW_VIEWPORT;
+}
+
+
+void softpipe_set_scissor_state( struct pipe_context *pipe,
+                                 const struct pipe_scissor_state *scissor )
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   draw_flush(softpipe->draw);
+
+   softpipe->scissor = *scissor; /* struct copy */
+   softpipe->dirty |= SP_NEW_SCISSOR;
+}
+
+
+void softpipe_set_polygon_stipple( struct pipe_context *pipe,
+                                   const struct pipe_poly_stipple *stipple )
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   draw_flush(softpipe->draw);
+
+   softpipe->poly_stipple = *stipple; /* struct copy */
+   softpipe->dirty |= SP_NEW_STIPPLE;
+}
diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c
new file mode 100644
index 0000000000..6b6a4c3ff3
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_state_derived.c
@@ -0,0 +1,210 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw/draw_context.h"
+#include "draw/draw_vertex.h"
+#include "draw/draw_private.h"
+#include "sp_context.h"
+#include "sp_state.h"
+
+
+/**
+ * Mark the current vertex layout as "invalid".
+ * We'll validate the vertex layout later, when we start to actually
+ * render a point or line or tri.
+ */
+static void
+invalidate_vertex_layout(struct softpipe_context *softpipe)
+{
+   softpipe->vertex_info.num_attribs =  0;
+}
+
+
+/**
+ * The vertex info describes how to convert the post-transformed vertices
+ * (simple float[][4]) used by the 'draw' module into vertices for
+ * rasterization.
+ *
+ * This function validates the vertex layout and returns a pointer to a
+ * vertex_info object.
+ */
+struct vertex_info *
+softpipe_get_vertex_info(struct softpipe_context *softpipe)
+{
+   struct vertex_info *vinfo = &softpipe->vertex_info;
+
+   if (vinfo->num_attribs == 0) {
+      /* compute vertex layout now */
+      const struct sp_fragment_shader *spfs = softpipe->fs;
+      const enum interp_mode colorInterp
+         = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR;
+      uint i;
+
+      if (softpipe->vbuf) {
+         /* if using the post-transform vertex buffer, tell draw_vbuf to
+          * simply emit the whole post-xform vertex as-is:
+          */
+         struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf;
+         const uint num = draw_num_vs_outputs(softpipe->draw);
+         uint i;
+
+         /* No longer any need to try and emit draw vertex_header info.
+          */
+         vinfo_vbuf->num_attribs = 0;
+         for (i = 0; i < num; i++) {
+            draw_emit_vertex_attr(vinfo_vbuf, EMIT_4F, INTERP_PERSPECTIVE, i);
+         }
+         draw_compute_vertex_size(vinfo_vbuf);
+      }
+
+      /*
+       * Loop over fragment shader inputs, searching for the matching output
+       * from the vertex shader.
+       */
+      vinfo->num_attribs = 0;
+      for (i = 0; i < spfs->info.num_inputs; i++) {
+         int src;
+         switch (spfs->info.input_semantic_name[i]) {
+         case TGSI_SEMANTIC_POSITION:
+            src = draw_find_vs_output(softpipe->draw,
+                                      TGSI_SEMANTIC_POSITION, 0);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src);
+            break;
+
+         case TGSI_SEMANTIC_COLOR:
+            src = draw_find_vs_output(softpipe->draw, TGSI_SEMANTIC_COLOR, 
+                                 spfs->info.input_semantic_index[i]);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
+            break;
+
+         case TGSI_SEMANTIC_FOG:
+            src = draw_find_vs_output(softpipe->draw, TGSI_SEMANTIC_FOG, 0);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+            break;
+
+         case TGSI_SEMANTIC_GENERIC:
+            /* this includes texcoords and varying vars */
+            src = draw_find_vs_output(softpipe->draw, TGSI_SEMANTIC_GENERIC,
+                                      spfs->info.input_semantic_index[i]);
+            draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
+            break;
+
+         default:
+            assert(0);
+         }
+      }
+
+      softpipe->psize_slot = draw_find_vs_output(softpipe->draw,
+                                                 TGSI_SEMANTIC_PSIZE, 0);
+      if (softpipe->psize_slot > 0) {
+         draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT,
+                               softpipe->psize_slot);
+      }
+
+      draw_compute_vertex_size(vinfo);
+   }
+
+   return vinfo;
+}
+
+
+/**
+ * Called from vbuf module.
+ *
+ * Note that there's actually two different vertex layouts in softpipe.
+ *
+ * The normal one is computed in softpipe_get_vertex_info() above and is
+ * used by the point/line/tri "setup" code.
+ *
+ * The other one (this one) is only used by the vbuf module (which is
+ * not normally used by default but used in testing).  For the vbuf module,
+ * we basically want to pass-through the draw module's vertex layout as-is.
+ * When the softpipe vbuf code begins drawing, the normal vertex layout
+ * will come into play again.
+ */
+struct vertex_info *
+softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe)
+{
+   (void) softpipe_get_vertex_info(softpipe);
+   return &softpipe->vertex_info_vbuf;
+}
+
+
+/**
+ * Recompute cliprect from scissor bounds, scissor enable and surface size.
+ */
+static void
+compute_cliprect(struct softpipe_context *sp)
+{
+   uint surfWidth = sp->framebuffer.width;
+   uint surfHeight = sp->framebuffer.height;
+
+   if (sp->rasterizer->scissor) {
+      /* clip to scissor rect */
+      sp->cliprect.minx = MAX2(sp->scissor.minx, 0);
+      sp->cliprect.miny = MAX2(sp->scissor.miny, 0);
+      sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth);
+      sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight);
+   }
+   else {
+      /* clip to surface bounds */
+      sp->cliprect.minx = 0;
+      sp->cliprect.miny = 0;
+      sp->cliprect.maxx = surfWidth;
+      sp->cliprect.maxy = surfHeight;
+   }
+}
+
+
+/* Hopefully this will remain quite simple, otherwise need to pull in
+ * something like the state tracker mechanism.
+ */
+void softpipe_update_derived( struct softpipe_context *softpipe )
+{
+   if (softpipe->dirty & (SP_NEW_RASTERIZER |
+                          SP_NEW_FS |
+                          SP_NEW_VS))
+      invalidate_vertex_layout( softpipe );
+
+   if (softpipe->dirty & (SP_NEW_SCISSOR |
+                          SP_NEW_DEPTH_STENCIL_ALPHA |
+                          SP_NEW_FRAMEBUFFER))
+      compute_cliprect(softpipe);
+
+   if (softpipe->dirty & (SP_NEW_BLEND |
+                          SP_NEW_DEPTH_STENCIL_ALPHA |
+                          SP_NEW_FRAMEBUFFER |
+                          SP_NEW_RASTERIZER |
+                          SP_NEW_FS | 
+			  SP_NEW_QUERY))
+      sp_build_quad_pipeline(softpipe);
+
+   softpipe->dirty = 0;
+}
diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c
new file mode 100644
index 0000000000..4d01a9dbe1
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_state_fs.c
@@ -0,0 +1,160 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_fs.h"
+
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw/draw_context.h"
+#include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_scan.h"
+
+
+void *
+softpipe_create_fs_state(struct pipe_context *pipe,
+                         const struct pipe_shader_state *templ)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+   struct sp_fragment_shader *state;
+
+   /* debug */
+   if (softpipe->dump_fs) 
+      tgsi_dump(templ->tokens, 0);
+
+   /* codegen */
+   state = softpipe_create_fs_llvm( softpipe, templ );
+   if (!state) {
+      state = softpipe_create_fs_sse( softpipe, templ );
+      if (!state) {
+         state = softpipe_create_fs_exec( softpipe, templ );
+      }
+   }
+
+   assert(state);
+
+   /* get/save the summary info for this shader */
+   tgsi_scan_shader(templ->tokens, &state->info);
+
+   return state;
+}
+
+
+void
+softpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   softpipe->fs = (struct sp_fragment_shader *) fs;
+
+   softpipe->dirty |= SP_NEW_FS;
+}
+
+
+void
+softpipe_delete_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct sp_fragment_shader *state = fs;
+
+   assert(fs != softpipe_context(pipe)->fs);
+   
+   state->delete( state );
+}
+
+
+void *
+softpipe_create_vs_state(struct pipe_context *pipe,
+                         const struct pipe_shader_state *templ)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+   struct sp_vertex_shader *state;
+
+   state = CALLOC_STRUCT(sp_vertex_shader);
+   if (state == NULL ) {
+      return NULL;
+   }
+
+   state->draw_data = draw_create_vertex_shader(softpipe->draw, templ);
+   if (state->draw_data == NULL) {
+      FREE( state );
+      return NULL;
+   }
+
+   return state;
+}
+
+
+void
+softpipe_bind_vs_state(struct pipe_context *pipe, void *vs)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   softpipe->vs = (const struct sp_vertex_shader *)vs;
+
+   draw_bind_vertex_shader(softpipe->draw,
+                           (softpipe->vs ? softpipe->vs->draw_data : NULL));
+
+   softpipe->dirty |= SP_NEW_VS;
+}
+
+
+void
+softpipe_delete_vs_state(struct pipe_context *pipe, void *vs)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   struct sp_vertex_shader *state =
+      (struct sp_vertex_shader *)vs;
+
+   draw_delete_vertex_shader(softpipe->draw, state->draw_data);
+   FREE( state );
+}
+
+
+
+void
+softpipe_set_constant_buffer(struct pipe_context *pipe,
+                             uint shader, uint index,
+                             const struct pipe_constant_buffer *buf)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+   struct pipe_screen *screen = pipe->screen;
+
+   assert(shader < PIPE_SHADER_TYPES);
+   assert(index == 0);
+
+   /* note: reference counting */
+   pipe_buffer_reference(screen,
+			 &softpipe->constants[shader].buffer,
+			 buf ? buf->buffer : NULL);
+
+   softpipe->dirty |= SP_NEW_CONSTANTS;
+}
diff --git a/src/gallium/drivers/softpipe/sp_state_rasterizer.c b/src/gallium/drivers/softpipe/sp_state_rasterizer.c
new file mode 100644
index 0000000000..87b7219683
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_state_rasterizer.c
@@ -0,0 +1,62 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_defines.h"
+#include "util/u_memory.h"
+#include "sp_context.h"
+#include "sp_state.h"
+#include "draw/draw_context.h"
+
+
+
+void *
+softpipe_create_rasterizer_state(struct pipe_context *pipe,
+                                 const struct pipe_rasterizer_state *rast)
+{
+   return mem_dup(rast, sizeof(*rast));
+}
+
+void softpipe_bind_rasterizer_state(struct pipe_context *pipe,
+                                    void *setup)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   /* pass-through to draw module */
+   draw_set_rasterizer_state(softpipe->draw, setup);
+
+   softpipe->rasterizer = (struct pipe_rasterizer_state *)setup;
+
+   softpipe->dirty |= SP_NEW_RASTERIZER;
+}
+
+void softpipe_delete_rasterizer_state(struct pipe_context *pipe,
+                                      void *rasterizer)
+{
+   FREE( rasterizer );
+}
+
+
diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c
new file mode 100644
index 0000000000..99a28c0d7e
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_state_sampler.c
@@ -0,0 +1,118 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:
+ *  Brian Paul
+ */
+
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+
+#include "draw/draw_context.h"
+
+#include "sp_context.h"
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_texture.h"
+#include "sp_tile_cache.h"
+#include "draw/draw_context.h"
+
+
+
+void *
+softpipe_create_sampler_state(struct pipe_context *pipe,
+                              const struct pipe_sampler_state *sampler)
+{
+   return mem_dup(sampler, sizeof(*sampler));
+}
+
+
+void
+softpipe_bind_sampler_states(struct pipe_context *pipe,
+                             unsigned num, void **sampler)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+   unsigned i;
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == softpipe->num_samplers &&
+       !memcmp(softpipe->sampler, sampler, num * sizeof(void *)))
+      return;
+
+   draw_flush(softpipe->draw);
+
+   for (i = 0; i < num; ++i)
+      softpipe->sampler[i] = sampler[i];
+   for (i = num; i < PIPE_MAX_SAMPLERS; ++i)
+      softpipe->sampler[i] = NULL;
+
+   softpipe->num_samplers = num;
+
+   softpipe->dirty |= SP_NEW_SAMPLER;
+}
+
+
+void
+softpipe_set_sampler_textures(struct pipe_context *pipe,
+                              unsigned num, struct pipe_texture **texture)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+   uint i;
+
+   assert(num <= PIPE_MAX_SAMPLERS);
+
+   /* Check for no-op */
+   if (num == softpipe->num_textures &&
+       !memcmp(softpipe->texture, texture, num * sizeof(struct pipe_texture *)))
+      return;
+
+   draw_flush(softpipe->draw);
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++) {
+      struct pipe_texture *tex = i < num ? texture[i] : NULL;
+
+      pipe_texture_reference(&softpipe->texture[i], tex);
+      sp_tile_cache_set_texture(pipe, softpipe->tex_cache[i], tex);
+   }
+
+   softpipe->num_textures = num;
+
+   softpipe->dirty |= SP_NEW_TEXTURE;
+}
+
+
+void
+softpipe_delete_sampler_state(struct pipe_context *pipe,
+                              void *sampler)
+{
+   FREE( sampler );
+}
+
+
+
diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c
new file mode 100644
index 0000000000..1493c65884
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_state_surface.c
@@ -0,0 +1,130 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+#include "pipe/p_inlines.h"
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_surface.h"
+#include "sp_tile_cache.h"
+
+#include "draw/draw_context.h"
+
+
+/**
+ * XXX this might get moved someday
+ * Set the framebuffer surface info: color buffers, zbuffer, stencil buffer.
+ * Here, we flush the old surfaces and update the tile cache to point to the new
+ * surfaces.
+ */
+void
+softpipe_set_framebuffer_state(struct pipe_context *pipe,
+                               const struct pipe_framebuffer_state *fb)
+{
+   struct softpipe_context *sp = softpipe_context(pipe);
+   uint i;
+
+   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
+      /* check if changing cbuf */
+      if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) {
+         /* flush old */
+         sp_flush_tile_cache(sp, sp->cbuf_cache[i]);
+
+         /* assign new */
+         sp->framebuffer.cbufs[i] = fb->cbufs[i];
+
+         /* update cache */
+         sp_tile_cache_set_surface(sp->cbuf_cache[i], fb->cbufs[i]);
+      }
+   }
+
+   sp->framebuffer.nr_cbufs = fb->nr_cbufs;
+
+   /* zbuf changing? */
+   if (sp->framebuffer.zsbuf != fb->zsbuf) {
+      /* flush old */
+      sp_flush_tile_cache(sp, sp->zsbuf_cache);
+
+      /* assign new */
+      sp->framebuffer.zsbuf = fb->zsbuf;
+
+      /* update cache */
+      sp_tile_cache_set_surface(sp->zsbuf_cache, fb->zsbuf);
+   }
+
+#if 0
+   /* XXX combined depth/stencil here */
+
+   /* sbuf changing? */
+   if (sp->framebuffer.sbuf != fb->sbuf) {
+      /* flush old */
+      sp_flush_tile_cache(sp, sp->sbuf_cache_sep);
+
+      /* assign new */
+      sp->framebuffer.sbuf = fb->sbuf;
+
+      /* update cache */
+      if (fb->sbuf != fb->zbuf) {
+         /* separate stencil buf */
+         sp->sbuf_cache = sp->sbuf_cache_sep;
+         sp_tile_cache_set_surface(sp->sbuf_cache, fb->sbuf);
+      }
+      else {
+         /* combined depth/stencil */
+         sp->sbuf_cache = sp->zbuf_cache;
+         sp_tile_cache_set_surface(sp->sbuf_cache, fb->sbuf);
+      }
+   }
+#endif
+
+   /* Tell draw module how deep the Z/depth buffer is */
+   {
+      int depth_bits;
+      double mrd;
+      if (sp->framebuffer.zsbuf) {
+         depth_bits = pf_get_component_bits(sp->framebuffer.zsbuf->format,
+                                            PIPE_FORMAT_COMP_Z);
+      }
+      else {
+         depth_bits = 0;
+      }
+      if (depth_bits > 16) {
+         mrd = 0.0000001;
+      }
+      else {
+         mrd = 0.00002;
+      }
+      draw_set_mrd(sp->draw, mrd);
+   }
+
+   sp->framebuffer.width = fb->width;
+   sp->framebuffer.height = fb->height;
+
+   sp->dirty |= SP_NEW_FRAMEBUFFER;
+}
diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c
new file mode 100644
index 0000000000..46b6991195
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_state_vertex.c
@@ -0,0 +1,73 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_surface.h"
+
+#include "draw/draw_context.h"
+
+
+void
+softpipe_set_vertex_elements(struct pipe_context *pipe,
+                             unsigned count,
+                             const struct pipe_vertex_element *attribs)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(softpipe->vertex_element, attribs,
+          count * sizeof(struct pipe_vertex_element));
+   softpipe->num_vertex_elements = count;
+
+   softpipe->dirty |= SP_NEW_VERTEX;
+
+   draw_set_vertex_elements(softpipe->draw, count, attribs);
+}
+
+
+void
+softpipe_set_vertex_buffers(struct pipe_context *pipe,
+                            unsigned count,
+                            const struct pipe_vertex_buffer *buffers)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(softpipe->vertex_buffer, buffers, count * sizeof(buffers[0]));
+   softpipe->num_vertex_buffers = count;
+
+   softpipe->dirty |= SP_NEW_VERTEX;
+
+   draw_set_vertex_buffers(softpipe->draw, count, buffers);
+}
diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c
new file mode 100644
index 0000000000..6ade732698
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_surface.c
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "util/u_rect.h"
+#include "sp_context.h"
+
+
+
+void
+sp_init_surface_functions(struct softpipe_context *sp)
+{
+   sp->pipe.surface_copy = util_surface_copy;
+   sp->pipe.surface_fill = util_surface_fill;
+}
diff --git a/src/gallium/drivers/softpipe/sp_surface.h b/src/gallium/drivers/softpipe/sp_surface.h
new file mode 100644
index 0000000000..22de3ba43f
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_surface.h
@@ -0,0 +1,42 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* Authors:  Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#ifndef SP_SURFACE_H
+#define SP_SURFACE_H
+
+
+struct softpipe_context;
+
+
+extern void
+sp_init_surface_functions(struct softpipe_context *sp);
+
+
+#endif /* SP_SURFACE_H */
diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c
new file mode 100644
index 0000000000..32aa5025e4
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_tex_sample.c
@@ -0,0 +1,1229 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2008 VMware, Inc.  All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Texture sampling
+ *
+ * Authors:
+ *   Brian Paul
+ */
+
+#include "sp_context.h"
+#include "sp_headers.h"
+#include "sp_surface.h"
+#include "sp_texture.h"
+#include "sp_tex_sample.h"
+#include "sp_tile_cache.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+
+
+/*
+ * Note, the FRAC macro has to work perfectly.  Otherwise you'll sometimes
+ * see 1-pixel bands of improperly weighted linear-filtered textures.
+ * The tests/texwrap.c demo is a good test.
+ * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0.
+ * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x).
+ */
+#define FRAC(f)  ((f) - util_ifloor(f))
+
+
+/**
+ * Linear interpolation macro
+ */
+static INLINE float
+lerp(float a, float v0, float v1)
+{
+   return v0 + a * (v1 - v0);
+}
+
+
+/**
+ * Do 2D/biliner interpolation of float values.
+ * v00, v10, v01 and v11 are typically four texture samples in a square/box.
+ * a and b are the horizontal and vertical interpolants.
+ * It's important that this function is inlined when compiled with
+ * optimization!  If we find that's not true on some systems, convert
+ * to a macro.
+ */
+static INLINE float
+lerp_2d(float a, float b,
+        float v00, float v10, float v01, float v11)
+{
+   const float temp0 = lerp(a, v00, v10);
+   const float temp1 = lerp(a, v01, v11);
+   return lerp(b, temp0, temp1);
+}
+
+
+/**
+ * As above, but 3D interpolation of 8 values.
+ */
+static INLINE float
+lerp_3d(float a, float b, float c,
+        float v000, float v100, float v010, float v110,
+        float v001, float v101, float v011, float v111)
+{
+   const float temp0 = lerp_2d(a, b, v000, v100, v010, v110);
+   const float temp1 = lerp_2d(a, b, v001, v101, v011, v111);
+   return lerp(c, temp0, temp1);
+}
+
+
+
+/**
+ * If A is a signed integer, A % B doesn't give the right value for A < 0
+ * (in terms of texture repeat).  Just casting to unsigned fixes that.
+ */
+#define REMAINDER(A, B) ((unsigned) (A) % (unsigned) (B))
+
+
+/**
+ * Apply texture coord wrapping mode and return integer texture indexes
+ * for a vector of four texcoords (S or T or P).
+ * \param wrapMode  PIPE_TEX_WRAP_x
+ * \param s  the incoming texcoords
+ * \param size  the texture image size
+ * \param icoord  returns the integer texcoords
+ * \return  integer texture index
+ */
+static INLINE void
+nearest_texcoord_4(unsigned wrapMode, const float s[4], unsigned size,
+                   int icoord[4])
+{
+   uint ch;
+   switch (wrapMode) {
+   case PIPE_TEX_WRAP_REPEAT:
+      /* s limited to [0,1) */
+      /* i limited to [0,size-1] */
+      for (ch = 0; ch < 4; ch++) {
+         int i = util_ifloor(s[ch] * size);
+         icoord[ch] = REMAINDER(i, size);
+      }
+      return;
+   case PIPE_TEX_WRAP_CLAMP:
+      /* s limited to [0,1] */
+      /* i limited to [0,size-1] */
+      for (ch = 0; ch < 4; ch++) {
+         if (s[ch] <= 0.0F)
+            icoord[ch] = 0;
+         else if (s[ch] >= 1.0F)
+            icoord[ch] = size - 1;
+         else
+            icoord[ch] = util_ifloor(s[ch] * size);
+      }
+      return;
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const float min = 1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            if (s[ch] < min)
+               icoord[ch] = 0;
+            else if (s[ch] > max)
+               icoord[ch] = size - 1;
+            else
+               icoord[ch] = util_ifloor(s[ch] * size);
+         }
+      }
+      return;
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [-1, size] */
+         const float min = -1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            if (s[ch] <= min)
+               icoord[ch] = -1;
+            else if (s[ch] >= max)
+               icoord[ch] = size;
+            else
+               icoord[ch] = util_ifloor(s[ch] * size);
+         }
+      }
+      return;
+   case PIPE_TEX_WRAP_MIRROR_REPEAT:
+      {
+         const float min = 1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            const int flr = util_ifloor(s[ch]);
+            float u;
+            if (flr & 1)
+               u = 1.0F - (s[ch] - (float) flr);
+            else
+               u = s[ch] - (float) flr;
+            if (u < min)
+               icoord[ch] = 0;
+            else if (u > max)
+               icoord[ch] = size - 1;
+            else
+               icoord[ch] = util_ifloor(u * size);
+         }
+      }
+      return;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP:
+      for (ch = 0; ch < 4; ch++) {
+         /* s limited to [0,1] */
+         /* i limited to [0,size-1] */
+         const float u = fabsf(s[ch]);
+         if (u <= 0.0F)
+            icoord[ch] = 0;
+         else if (u >= 1.0F)
+            icoord[ch] = size - 1;
+         else
+            icoord[ch] = util_ifloor(u * size);
+      }
+      return;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const float min = 1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            const float u = fabsf(s[ch]);
+            if (u < min)
+               icoord[ch] = 0;
+            else if (u > max)
+               icoord[ch] = size - 1;
+            else
+               icoord[ch] = util_ifloor(u * size);
+         }
+      }
+      return;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const float min = -1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            const float u = fabsf(s[ch]);
+            if (u < min)
+               icoord[ch] = -1;
+            else if (u > max)
+               icoord[ch] = size;
+            else
+               icoord[ch] = util_ifloor(u * size);
+         }
+      }
+      return;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * Used to compute texel locations for linear sampling for four texcoords.
+ * \param wrapMode  PIPE_TEX_WRAP_x
+ * \param s  the texcoords
+ * \param size  the texture image size
+ * \param icoord0  returns first texture indexes
+ * \param icoord1  returns second texture indexes (usually icoord0 + 1)
+ * \param w  returns blend factor/weight between texture indexes
+ * \param icoord  returns the computed integer texture coords
+ */
+static INLINE void
+linear_texcoord_4(unsigned wrapMode, const float s[4], unsigned size,
+                  int icoord0[4], int icoord1[4], float w[4])
+{
+   uint ch;
+
+   switch (wrapMode) {
+   case PIPE_TEX_WRAP_REPEAT:
+      for (ch = 0; ch < 4; ch++) {
+         float u = s[ch] * size - 0.5F;
+         icoord0[ch] = REMAINDER(util_ifloor(u), size);
+         icoord1[ch] = REMAINDER(icoord0[ch] + 1, size);
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_CLAMP:
+      for (ch = 0; ch < 4; ch++) {
+         float u = CLAMP(s[ch], 0.0F, 1.0F);
+         u = u * size - 0.5f;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      for (ch = 0; ch < 4; ch++) {
+         float u = CLAMP(s[ch], 0.0F, 1.0F);
+         u = u * size - 0.5f;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         if (icoord0[ch] < 0)
+            icoord0[ch] = 0;
+         if (icoord1[ch] >= (int) size)
+            icoord1[ch] = size - 1;
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      {
+         const float min = -1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            float u = CLAMP(s[ch], min, max);
+            u = u * size - 0.5f;
+            icoord0[ch] = util_ifloor(u);
+            icoord1[ch] = icoord0[ch] + 1;
+            w[ch] = FRAC(u);
+         }
+      }
+      break;;
+   case PIPE_TEX_WRAP_MIRROR_REPEAT:
+      for (ch = 0; ch < 4; ch++) {
+         const int flr = util_ifloor(s[ch]);
+         float u;
+         if (flr & 1)
+            u = 1.0F - (s[ch] - (float) flr);
+         else
+            u = s[ch] - (float) flr;
+         u = u * size - 0.5F;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         if (icoord0[ch] < 0)
+            icoord0[ch] = 0;
+         if (icoord1[ch] >= (int) size)
+            icoord1[ch] = size - 1;
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP:
+      for (ch = 0; ch < 4; ch++) {
+         float u = fabsf(s[ch]);
+         if (u >= 1.0F)
+            u = (float) size;
+         else
+            u *= size;
+         u -= 0.5F;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+      for (ch = 0; ch < 4; ch++) {
+         float u = fabsf(s[ch]);
+         if (u >= 1.0F)
+            u = (float) size;
+         else
+            u *= size;
+         u -= 0.5F;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         if (icoord0[ch] < 0)
+            icoord0[ch] = 0;
+         if (icoord1[ch] >= (int) size)
+            icoord1[ch] = size - 1;
+         w[ch] = FRAC(u);
+      }
+      break;;
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+      {
+         const float min = -1.0F / (2.0F * size);
+         const float max = 1.0F - min;
+         for (ch = 0; ch < 4; ch++) {
+            float u = fabsf(s[ch]);
+            if (u <= min)
+               u = min * size;
+            else if (u >= max)
+               u = max * size;
+            else
+               u *= size;
+            u -= 0.5F;
+            icoord0[ch] = util_ifloor(u);
+            icoord1[ch] = icoord0[ch] + 1;
+            w[ch] = FRAC(u);
+         }
+      }
+      break;;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * For RECT textures / unnormalized texcoords
+ * Only a subset of wrap modes supported.
+ */
+static INLINE void
+nearest_texcoord_unnorm_4(unsigned wrapMode, const float s[4], unsigned size,
+                          int icoord[4])
+{
+   uint ch;
+   switch (wrapMode) {
+   case PIPE_TEX_WRAP_CLAMP:
+      for (ch = 0; ch < 4; ch++) {
+         int i = util_ifloor(s[ch]);
+         icoord[ch]= CLAMP(i, 0, (int) size-1);
+      }
+      return;
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      /* fall-through */
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      for (ch = 0; ch < 4; ch++) {
+         icoord[ch]= util_ifloor( CLAMP(s[ch], 0.5F, (float) size - 0.5F) );
+      }
+      return;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * For RECT textures / unnormalized texcoords.
+ * Only a subset of wrap modes supported.
+ */
+static INLINE void
+linear_texcoord_unnorm_4(unsigned wrapMode, const float s[4], unsigned size,
+                         int icoord0[4], int icoord1[4], float w[4])
+{
+   uint ch;
+   switch (wrapMode) {
+   case PIPE_TEX_WRAP_CLAMP:
+      for (ch = 0; ch < 4; ch++) {
+         /* Not exactly what the spec says, but it matches NVIDIA output */
+         float u = CLAMP(s[ch] - 0.5F, 0.0f, (float) size - 1.0f);
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         w[ch] = FRAC(u);
+      }
+      return;
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      /* fall-through */
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      for (ch = 0; ch < 4; ch++) {
+         float u = CLAMP(s[ch], 0.5F, (float) size - 0.5F);
+         u -= 0.5F;
+         icoord0[ch] = util_ifloor(u);
+         icoord1[ch] = icoord0[ch] + 1;
+         if (icoord1[ch] > (int) size - 1)
+            icoord1[ch] = size - 1;
+         w[ch] = FRAC(u);
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+static unsigned
+choose_cube_face(float rx, float ry, float rz, float *newS, float *newT)
+{
+   /*
+      major axis
+      direction     target                             sc     tc    ma
+      ----------    -------------------------------    ---    ---   ---
+       +rx          TEXTURE_CUBE_MAP_POSITIVE_X_EXT    -rz    -ry   rx
+       -rx          TEXTURE_CUBE_MAP_NEGATIVE_X_EXT    +rz    -ry   rx
+       +ry          TEXTURE_CUBE_MAP_POSITIVE_Y_EXT    +rx    +rz   ry
+       -ry          TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT    +rx    -rz   ry
+       +rz          TEXTURE_CUBE_MAP_POSITIVE_Z_EXT    +rx    -ry   rz
+       -rz          TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT    -rx    -ry   rz
+   */
+   const float arx = fabsf(rx), ary = fabsf(ry), arz = fabsf(rz);
+   unsigned face;
+   float sc, tc, ma;
+
+   if (arx > ary && arx > arz) {
+      if (rx >= 0.0F) {
+         face = PIPE_TEX_FACE_POS_X;
+         sc = -rz;
+         tc = -ry;
+         ma = arx;
+      }
+      else {
+         face = PIPE_TEX_FACE_NEG_X;
+         sc = rz;
+         tc = -ry;
+         ma = arx;
+      }
+   }
+   else if (ary > arx && ary > arz) {
+      if (ry >= 0.0F) {
+         face = PIPE_TEX_FACE_POS_Y;
+         sc = rx;
+         tc = rz;
+         ma = ary;
+      }
+      else {
+         face = PIPE_TEX_FACE_NEG_Y;
+         sc = rx;
+         tc = -rz;
+         ma = ary;
+      }
+   }
+   else {
+      if (rz > 0.0F) {
+         face = PIPE_TEX_FACE_POS_Z;
+         sc = rx;
+         tc = -ry;
+         ma = arz;
+      }
+      else {
+         face = PIPE_TEX_FACE_NEG_Z;
+         sc = -rx;
+         tc = -ry;
+         ma = arz;
+      }
+   }
+
+   *newS = ( sc / ma + 1.0F ) * 0.5F;
+   *newT = ( tc / ma + 1.0F ) * 0.5F;
+
+   return face;
+}
+
+
+/**
+ * Examine the quad's texture coordinates to compute the partial
+ * derivatives w.r.t X and Y, then compute lambda (level of detail).
+ *
+ * This is only done for fragment shaders, not vertex shaders.
+ */
+static float
+compute_lambda(const struct pipe_texture *tex,
+               const struct pipe_sampler_state *sampler,
+               const float s[QUAD_SIZE],
+               const float t[QUAD_SIZE],
+               const float p[QUAD_SIZE],
+               float lodbias)
+{
+   float rho, lambda;
+
+   assert(sampler->normalized_coords);
+
+   assert(s);
+   {
+      float dsdx = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT];
+      float dsdy = s[QUAD_TOP_LEFT]     - s[QUAD_BOTTOM_LEFT];
+      dsdx = fabsf(dsdx);
+      dsdy = fabsf(dsdy);
+      rho = MAX2(dsdx, dsdy) * tex->width[0];
+   }
+   if (t) {
+      float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT];
+      float dtdy = t[QUAD_TOP_LEFT]     - t[QUAD_BOTTOM_LEFT];
+      float max;
+      dtdx = fabsf(dtdx);
+      dtdy = fabsf(dtdy);
+      max = MAX2(dtdx, dtdy) * tex->height[0];
+      rho = MAX2(rho, max);
+   }
+   if (p) {
+      float dpdx = p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT];
+      float dpdy = p[QUAD_TOP_LEFT]     - p[QUAD_BOTTOM_LEFT];
+      float max;
+      dpdx = fabsf(dpdx);
+      dpdy = fabsf(dpdy);
+      max = MAX2(dpdx, dpdy) * tex->depth[0];
+      rho = MAX2(rho, max);
+   }
+
+   lambda = util_fast_log2(rho);
+   lambda += lodbias + sampler->lod_bias;
+   lambda = CLAMP(lambda, sampler->min_lod, sampler->max_lod);
+
+   return lambda;
+}
+
+
+/**
+ * Do several things here:
+ * 1. Compute lambda from the texcoords, if needed
+ * 2. Determine if we're minifying or magnifying
+ * 3. If minifying, choose mipmap levels
+ * 4. Return image filter to use within mipmap images
+ * \param level0  Returns first mipmap level to sample from
+ * \param level1  Returns second mipmap level to sample from
+ * \param levelBlend  Returns blend factor between levels, in [0,1]
+ * \param imgFilter  Returns either the min or mag filter, depending on lambda
+ */
+static void
+choose_mipmap_levels(const struct pipe_texture *texture,
+                     const struct pipe_sampler_state *sampler,
+                     const float s[QUAD_SIZE],
+                     const float t[QUAD_SIZE],
+                     const float p[QUAD_SIZE],
+                     boolean computeLambda,
+                     float lodbias,
+                     unsigned *level0, unsigned *level1, float *levelBlend,
+                     unsigned *imgFilter)
+{
+   if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) {
+      /* no mipmap selection needed */
+      *level0 = *level1 = CLAMP((int) sampler->min_lod,
+                                0, (int) texture->last_level);
+
+      if (sampler->min_img_filter != sampler->mag_img_filter) {
+         /* non-mipmapped texture, but still need to determine if doing
+          * minification or magnification.
+          */
+         float lambda = compute_lambda(texture, sampler, s, t, p, lodbias);
+         if (lambda <= 0.0) {
+            *imgFilter = sampler->mag_img_filter;
+         }
+         else {
+            *imgFilter = sampler->min_img_filter;
+         }
+      }
+      else {
+         *imgFilter = sampler->mag_img_filter;
+      }
+   }
+   else {
+      float lambda;
+
+      if (computeLambda)
+         /* fragment shader */
+         lambda = compute_lambda(texture, sampler, s, t, p, lodbias);
+      else
+         /* vertex shader */
+         lambda = lodbias; /* not really a bias, but absolute LOD */
+
+      if (lambda <= 0.0) { /* XXX threshold depends on the filter */
+         /* magnifying */
+         *imgFilter = sampler->mag_img_filter;
+         *level0 = *level1 = 0;
+      }
+      else {
+         /* minifying */
+         *imgFilter = sampler->min_img_filter;
+
+         /* choose mipmap level(s) and compute the blend factor between them */
+         if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NEAREST) {
+            /* Nearest mipmap level */
+            const int lvl = (int) (lambda + 0.5);
+            *level0 =
+            *level1 = CLAMP(lvl, 0, (int) texture->last_level);
+         }
+         else {
+            /* Linear interpolation between mipmap levels */
+            const int lvl = (int) lambda;
+            *level0 = CLAMP(lvl,     0, (int) texture->last_level);
+            *level1 = CLAMP(lvl + 1, 0, (int) texture->last_level);
+            *levelBlend = FRAC(lambda);  /* blending weight between levels */
+         }
+      }
+   }
+}
+
+
+/**
+ * Get a texel from a texture, using the texture tile cache.
+ *
+ * \param face  the cube face in 0..5
+ * \param level  the mipmap level
+ * \param x  the x coord of texel within 2D image
+ * \param y  the y coord of texel within 2D image
+ * \param z  which slice of a 3D texture
+ * \param rgba  the quad to put the texel/color into
+ * \param j  which element of the rgba quad to write to
+ *
+ * XXX maybe move this into sp_tile_cache.c and merge with the
+ * sp_get_cached_tile_tex() function.  Also, get 4 texels instead of 1...
+ */
+static void
+get_texel(const struct tgsi_sampler *tgsi_sampler,
+          unsigned face, unsigned level, int x, int y, int z,
+          float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j)
+{
+   const struct sp_shader_sampler *samp = sp_shader_sampler(tgsi_sampler);
+   struct softpipe_context *sp = samp->sp;
+   const uint unit = samp->unit;
+   const struct pipe_texture *texture = sp->texture[unit];
+   const struct pipe_sampler_state *sampler = sp->sampler[unit];
+
+   if (x < 0 || x >= (int) texture->width[level] ||
+       y < 0 || y >= (int) texture->height[level] ||
+       z < 0 || z >= (int) texture->depth[level]) {
+      rgba[0][j] = sampler->border_color[0];
+      rgba[1][j] = sampler->border_color[1];
+      rgba[2][j] = sampler->border_color[2];
+      rgba[3][j] = sampler->border_color[3];
+   }
+   else {
+      const int tx = x % TILE_SIZE;
+      const int ty = y % TILE_SIZE;
+      const struct softpipe_cached_tile *tile
+         = sp_get_cached_tile_tex(sp, samp->cache,
+                                  x, y, z, face, level);
+      rgba[0][j] = tile->data.color[ty][tx][0];
+      rgba[1][j] = tile->data.color[ty][tx][1];
+      rgba[2][j] = tile->data.color[ty][tx][2];
+      rgba[3][j] = tile->data.color[ty][tx][3];
+      if (0)
+      {
+         debug_printf("Get texel %f %f %f %f from %s\n",
+                      rgba[0][j], rgba[1][j], rgba[2][j], rgba[3][j],
+                      pf_name(texture->format));
+      }
+   }
+}
+
+
+/**
+ * Compare texcoord 'p' (aka R) against texture value 'rgba[0]'
+ * When we sampled the depth texture, the depth value was put into all
+ * RGBA channels.  We look at the red channel here.
+ */
+static INLINE void
+shadow_compare(uint compare_func,
+               float rgba[NUM_CHANNELS][QUAD_SIZE],
+               const float p[QUAD_SIZE],
+               uint j)
+{
+   int k;
+   switch (compare_func) {
+   case PIPE_FUNC_LESS:
+      k = p[j] < rgba[0][j];
+      break;
+   case PIPE_FUNC_LEQUAL:
+      k = p[j] <= rgba[0][j];
+      break;
+   case PIPE_FUNC_GREATER:
+      k = p[j] > rgba[0][j];
+      break;
+   case PIPE_FUNC_GEQUAL:
+      k = p[j] >= rgba[0][j];
+      break;
+   case PIPE_FUNC_EQUAL:
+      k = p[j] == rgba[0][j];
+      break;
+   case PIPE_FUNC_NOTEQUAL:
+      k = p[j] != rgba[0][j];
+      break;
+   case PIPE_FUNC_ALWAYS:
+      k = 1;
+      break;
+   case PIPE_FUNC_NEVER:
+      k = 0;
+      break;
+   default:
+      k = 0;
+      assert(0);
+      break;
+   }
+
+   rgba[0][j] = rgba[1][j] = rgba[2][j] = (float) k;
+}
+
+
+/**
+ * Common code for sampling 1D/2D/cube textures.
+ * Could probably extend for 3D...
+ */
+static void
+sp_get_samples_2d_common(const struct tgsi_sampler *tgsi_sampler,
+                         const float s[QUAD_SIZE],
+                         const float t[QUAD_SIZE],
+                         const float p[QUAD_SIZE],
+                         boolean computeLambda,
+                         float lodbias,
+                         float rgba[NUM_CHANNELS][QUAD_SIZE],
+                         const unsigned faces[4])
+{
+   const struct sp_shader_sampler *samp = sp_shader_sampler(tgsi_sampler);
+   const struct softpipe_context *sp = samp->sp;
+   const uint unit = samp->unit;
+   const struct pipe_texture *texture = sp->texture[unit];
+   const struct pipe_sampler_state *sampler = sp->sampler[unit];
+   const uint compare_func = sampler->compare_func;
+   unsigned level0, level1, j, imgFilter;
+   int width, height;
+   float levelBlend;
+
+   choose_mipmap_levels(texture, sampler, s, t, p, computeLambda, lodbias,
+                        &level0, &level1, &levelBlend, &imgFilter);
+
+   assert(sampler->normalized_coords);
+
+   width = texture->width[level0];
+   height = texture->height[level0];
+
+   assert(width > 0);
+
+   switch (imgFilter) {
+   case PIPE_TEX_FILTER_NEAREST:
+      {
+         int x[4], y[4];
+         nearest_texcoord_4(sampler->wrap_s, s, width, x);
+         nearest_texcoord_4(sampler->wrap_t, t, height, y);
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            get_texel(tgsi_sampler, faces[j], level0, x[j], y[j], 0, rgba, j);
+            if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+               shadow_compare(compare_func, rgba, p, j);
+            }
+
+            if (level0 != level1) {
+               /* get texels from second mipmap level and blend */
+               float rgba2[4][4];
+               unsigned c;
+               x[j] /= 2;
+               y[j] /= 2;
+               get_texel(tgsi_sampler, faces[j], level1, x[j], y[j], 0,
+                         rgba2, j);
+               if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){
+                  shadow_compare(compare_func, rgba2, p, j);
+               }
+
+               for (c = 0; c < NUM_CHANNELS; c++) {
+                  rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]);
+               }
+            }
+         }
+      }
+      break;
+   case PIPE_TEX_FILTER_LINEAR:
+   case PIPE_TEX_FILTER_ANISO:
+      {
+         int x0[4], y0[4], x1[4], y1[4];
+         float xw[4], yw[4]; /* weights */
+
+         linear_texcoord_4(sampler->wrap_s, s, width, x0, x1, xw);
+         linear_texcoord_4(sampler->wrap_t, t, height, y0, y1, yw);
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            float tx[4][4]; /* texels */
+            int c;
+            get_texel(tgsi_sampler, faces[j], level0, x0[j], y0[j], 0, tx, 0);
+            get_texel(tgsi_sampler, faces[j], level0, x1[j], y0[j], 0, tx, 1);
+            get_texel(tgsi_sampler, faces[j], level0, x0[j], y1[j], 0, tx, 2);
+            get_texel(tgsi_sampler, faces[j], level0, x1[j], y1[j], 0, tx, 3);
+            if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+               shadow_compare(compare_func, tx, p, 0);
+               shadow_compare(compare_func, tx, p, 1);
+               shadow_compare(compare_func, tx, p, 2);
+               shadow_compare(compare_func, tx, p, 3);
+            }
+
+            /* interpolate R, G, B, A */
+            for (c = 0; c < 4; c++) {
+               rgba[c][j] = lerp_2d(xw[j], yw[j],
+                                    tx[c][0], tx[c][1],
+                                    tx[c][2], tx[c][3]);
+            }
+
+            if (level0 != level1) {
+               /* get texels from second mipmap level and blend */
+               float rgba2[4][4];
+               x0[j] /= 2;
+               y0[j] /= 2;
+               x1[j] /= 2;
+               y1[j] /= 2;
+               get_texel(tgsi_sampler, faces[j], level1, x0[j], y0[j], 0, tx, 0);
+               get_texel(tgsi_sampler, faces[j], level1, x1[j], y0[j], 0, tx, 1);
+               get_texel(tgsi_sampler, faces[j], level1, x0[j], y1[j], 0, tx, 2);
+               get_texel(tgsi_sampler, faces[j], level1, x1[j], y1[j], 0, tx, 3);
+               if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){
+                  shadow_compare(compare_func, tx, p, 0);
+                  shadow_compare(compare_func, tx, p, 1);
+                  shadow_compare(compare_func, tx, p, 2);
+                  shadow_compare(compare_func, tx, p, 3);
+               }
+
+               /* interpolate R, G, B, A */
+               for (c = 0; c < 4; c++) {
+                  rgba2[c][j] = lerp_2d(xw[j], yw[j],
+                                        tx[c][0], tx[c][1], tx[c][2], tx[c][3]);
+               }
+
+               for (c = 0; c < NUM_CHANNELS; c++) {
+                  rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]);
+               }
+            }
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+static INLINE void
+sp_get_samples_1d(const struct tgsi_sampler *sampler,
+                  const float s[QUAD_SIZE],
+                  const float t[QUAD_SIZE],
+                  const float p[QUAD_SIZE],
+                  boolean computeLambda,
+                  float lodbias,
+                  float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   static const unsigned faces[4] = {0, 0, 0, 0};
+   static const float tzero[4] = {0, 0, 0, 0};
+   sp_get_samples_2d_common(sampler, s, tzero, NULL,
+                            computeLambda, lodbias, rgba, faces);
+}
+
+
+static INLINE void
+sp_get_samples_2d(const struct tgsi_sampler *sampler,
+                  const float s[QUAD_SIZE],
+                  const float t[QUAD_SIZE],
+                  const float p[QUAD_SIZE],
+                  boolean computeLambda,
+                  float lodbias,
+                  float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   static const unsigned faces[4] = {0, 0, 0, 0};
+   sp_get_samples_2d_common(sampler, s, t, p,
+                            computeLambda, lodbias, rgba, faces);
+}
+
+
+static INLINE void
+sp_get_samples_3d(const struct tgsi_sampler *tgsi_sampler,
+                  const float s[QUAD_SIZE],
+                  const float t[QUAD_SIZE],
+                  const float p[QUAD_SIZE],
+                  boolean computeLambda,
+                  float lodbias,
+                  float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   const struct sp_shader_sampler *samp = sp_shader_sampler(tgsi_sampler);
+   const struct softpipe_context *sp = samp->sp;
+   const uint unit = samp->unit;
+   const struct pipe_texture *texture = sp->texture[unit];
+   const struct pipe_sampler_state *sampler = sp->sampler[unit];
+   /* get/map pipe_surfaces corresponding to 3D tex slices */
+   unsigned level0, level1, j, imgFilter;
+   int width, height, depth;
+   float levelBlend;
+   const uint face = 0;
+
+   choose_mipmap_levels(texture, sampler, s, t, p, computeLambda, lodbias,
+                        &level0, &level1, &levelBlend, &imgFilter);
+
+   assert(sampler->normalized_coords);
+
+   width = texture->width[level0];
+   height = texture->height[level0];
+   depth = texture->depth[level0];
+
+   assert(width > 0);
+   assert(height > 0);
+   assert(depth > 0);
+
+   switch (imgFilter) {
+   case PIPE_TEX_FILTER_NEAREST:
+      {
+         int x[4], y[4], z[4];
+         nearest_texcoord_4(sampler->wrap_s, s, width, x);
+         nearest_texcoord_4(sampler->wrap_t, t, height, y);
+         nearest_texcoord_4(sampler->wrap_r, p, depth, z);
+         for (j = 0; j < QUAD_SIZE; j++) {
+            get_texel(tgsi_sampler, face, level0, x[j], y[j], z[j], rgba, j);
+            if (level0 != level1) {
+               /* get texels from second mipmap level and blend */
+               float rgba2[4][4];
+               unsigned c;
+               x[j] /= 2;
+               y[j] /= 2;
+               z[j] /= 2;
+               get_texel(tgsi_sampler, face, level1, x[j], y[j], z[j], rgba2, j);
+               for (c = 0; c < NUM_CHANNELS; c++) {
+                  rgba[c][j] = lerp(levelBlend, rgba2[c][j], rgba[c][j]);
+               }
+            }
+         }
+      }
+      break;
+   case PIPE_TEX_FILTER_LINEAR:
+   case PIPE_TEX_FILTER_ANISO:
+      {
+         int x0[4], x1[4], y0[4], y1[4], z0[4], z1[4];
+         float xw[4], yw[4], zw[4]; /* interpolation weights */
+         linear_texcoord_4(sampler->wrap_s, s, width,  x0, x1, xw);
+         linear_texcoord_4(sampler->wrap_t, t, height, y0, y1, yw);
+         linear_texcoord_4(sampler->wrap_r, p, depth,  z0, z1, zw);
+
+         for (j = 0; j < QUAD_SIZE; j++) {
+            int c;
+            float tx0[4][4], tx1[4][4];
+            get_texel(tgsi_sampler, face, level0, x0[j], y0[j], z0[j], tx0, 0);
+            get_texel(tgsi_sampler, face, level0, x1[j], y0[j], z0[j], tx0, 1);
+            get_texel(tgsi_sampler, face, level0, x0[j], y1[j], z0[j], tx0, 2);
+            get_texel(tgsi_sampler, face, level0, x1[j], y1[j], z0[j], tx0, 3);
+            get_texel(tgsi_sampler, face, level0, x0[j], y0[j], z1[j], tx1, 0);
+            get_texel(tgsi_sampler, face, level0, x1[j], y0[j], z1[j], tx1, 1);
+            get_texel(tgsi_sampler, face, level0, x0[j], y1[j], z1[j], tx1, 2);
+            get_texel(tgsi_sampler, face, level0, x1[j], y1[j], z1[j], tx1, 3);
+
+            /* interpolate R, G, B, A */
+            for (c = 0; c < 4; c++) {
+               rgba[c][j] = lerp_3d(xw[j], yw[j], zw[j],
+                                    tx0[c][0], tx0[c][1],
+                                    tx0[c][2], tx0[c][3],
+                                    tx1[c][0], tx1[c][1],
+                                    tx1[c][2], tx1[c][3]);
+            }
+
+            if (level0 != level1) {
+               /* get texels from second mipmap level and blend */
+               float rgba2[4][4];
+               x0[j] /= 2;
+               y0[j] /= 2;
+               z0[j] /= 2;
+               x1[j] /= 2;
+               y1[j] /= 2;
+               z1[j] /= 2;
+               get_texel(tgsi_sampler, face, level1, x0[j], y0[j], z0[j], tx0, 0);
+               get_texel(tgsi_sampler, face, level1, x1[j], y0[j], z0[j], tx0, 1);
+               get_texel(tgsi_sampler, face, level1, x0[j], y1[j], z0[j], tx0, 2);
+               get_texel(tgsi_sampler, face, level1, x1[j], y1[j], z0[j], tx0, 3);
+               get_texel(tgsi_sampler, face, level1, x0[j], y0[j], z1[j], tx1, 0);
+               get_texel(tgsi_sampler, face, level1, x1[j], y0[j], z1[j], tx1, 1);
+               get_texel(tgsi_sampler, face, level1, x0[j], y1[j], z1[j], tx1, 2);
+               get_texel(tgsi_sampler, face, level1, x1[j], y1[j], z1[j], tx1, 3);
+
+               /* interpolate R, G, B, A */
+               for (c = 0; c < 4; c++) {
+                  rgba2[c][j] = lerp_3d(xw[j], yw[j], zw[j],
+                                        tx0[c][0], tx0[c][1],
+                                        tx0[c][2], tx0[c][3],
+                                        tx1[c][0], tx1[c][1],
+                                        tx1[c][2], tx1[c][3]);
+               }
+
+               /* blend mipmap levels */
+               for (c = 0; c < NUM_CHANNELS; c++) {
+                  rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]);
+               }
+            }
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+static void
+sp_get_samples_cube(const struct tgsi_sampler *sampler,
+                    const float s[QUAD_SIZE],
+                    const float t[QUAD_SIZE],
+                    const float p[QUAD_SIZE],
+                    boolean computeLambda,
+                    float lodbias,
+                    float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   unsigned faces[QUAD_SIZE], j;
+   float ssss[4], tttt[4];
+   for (j = 0; j < QUAD_SIZE; j++) {
+      faces[j] = choose_cube_face(s[j], t[j], p[j], ssss + j, tttt + j);
+   }
+   sp_get_samples_2d_common(sampler, ssss, tttt, NULL,
+                            computeLambda, lodbias, rgba, faces);
+}
+
+
+static void
+sp_get_samples_rect(const struct tgsi_sampler *tgsi_sampler,
+                    const float s[QUAD_SIZE],
+                    const float t[QUAD_SIZE],
+                    const float p[QUAD_SIZE],
+                    boolean computeLambda,
+                    float lodbias,
+                    float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   const struct sp_shader_sampler *samp = sp_shader_sampler(tgsi_sampler);
+   const struct softpipe_context *sp = samp->sp;
+   const uint unit = samp->unit;
+   const struct pipe_texture *texture = sp->texture[unit];
+   const struct pipe_sampler_state *sampler = sp->sampler[unit];
+   const uint face = 0;
+   const uint compare_func = sampler->compare_func;
+   unsigned level0, level1, j, imgFilter;
+   int width, height;
+   float levelBlend;
+
+   choose_mipmap_levels(texture, sampler, s, t, p, computeLambda, lodbias,
+                        &level0, &level1, &levelBlend, &imgFilter);
+
+   /* texture RECTS cannot be mipmapped */
+   assert(level0 == level1);
+
+   width = texture->width[level0];
+   height = texture->height[level0];
+
+   assert(width > 0);
+
+   switch (imgFilter) {
+   case PIPE_TEX_FILTER_NEAREST:
+      {
+         int x[4], y[4];
+         nearest_texcoord_unnorm_4(sampler->wrap_s, s, width, x);
+         nearest_texcoord_unnorm_4(sampler->wrap_t, t, height, y);
+         for (j = 0; j < QUAD_SIZE; j++) {
+            get_texel(tgsi_sampler, face, level0, x[j], y[j], 0, rgba, j);
+            if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+               shadow_compare(compare_func, rgba, p, j);
+            }
+         }
+      }
+      break;
+   case PIPE_TEX_FILTER_LINEAR:
+   case PIPE_TEX_FILTER_ANISO:
+      {
+         int x0[4], y0[4], x1[4], y1[4];
+         float xw[4], yw[4]; /* weights */
+         linear_texcoord_unnorm_4(sampler->wrap_s, s, width,  x0, x1, xw);
+         linear_texcoord_unnorm_4(sampler->wrap_t, t, height, y0, y1, yw);
+         for (j = 0; j < QUAD_SIZE; j++) {
+            float tx[4][4]; /* texels */
+            int c;
+            get_texel(tgsi_sampler, face, level0, x0[j], y0[j], 0, tx, 0);
+            get_texel(tgsi_sampler, face, level0, x1[j], y0[j], 0, tx, 1);
+            get_texel(tgsi_sampler, face, level0, x0[j], y1[j], 0, tx, 2);
+            get_texel(tgsi_sampler, face, level0, x1[j], y1[j], 0, tx, 3);
+            if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+               shadow_compare(compare_func, tx, p, 0);
+               shadow_compare(compare_func, tx, p, 1);
+               shadow_compare(compare_func, tx, p, 2);
+               shadow_compare(compare_func, tx, p, 3);
+            }
+            for (c = 0; c < 4; c++) {
+               rgba[c][j] = lerp_2d(xw[j], yw[j],
+                                    tx[c][0], tx[c][1], tx[c][2], tx[c][3]);
+            }
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * Common code for vertex/fragment program texture sampling.
+ */
+static INLINE void
+sp_get_samples(struct tgsi_sampler *tgsi_sampler,
+               const float s[QUAD_SIZE],
+               const float t[QUAD_SIZE],
+               const float p[QUAD_SIZE],
+               boolean computeLambda,
+               float lodbias,
+               float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   const struct sp_shader_sampler *samp = sp_shader_sampler(tgsi_sampler);
+   const struct softpipe_context *sp = samp->sp;
+   const uint unit = samp->unit;
+   const struct pipe_texture *texture = sp->texture[unit];
+   const struct pipe_sampler_state *sampler = sp->sampler[unit];
+
+   if (!texture)
+      return;
+
+   switch (texture->target) {
+   case PIPE_TEXTURE_1D:
+      assert(sampler->normalized_coords);
+      sp_get_samples_1d(tgsi_sampler, s, t, p, computeLambda, lodbias, rgba);
+      break;
+   case PIPE_TEXTURE_2D:
+      if (sampler->normalized_coords)
+         sp_get_samples_2d(tgsi_sampler, s, t, p, computeLambda, lodbias, rgba);
+      else
+         sp_get_samples_rect(tgsi_sampler, s, t, p, computeLambda, lodbias, rgba);
+      break;
+   case PIPE_TEXTURE_3D:
+      assert(sampler->normalized_coords);
+      sp_get_samples_3d(tgsi_sampler, s, t, p, computeLambda, lodbias, rgba);
+      break;
+   case PIPE_TEXTURE_CUBE:
+      assert(sampler->normalized_coords);
+      sp_get_samples_cube(tgsi_sampler, s, t, p, computeLambda, lodbias, rgba);
+      break;
+   default:
+      assert(0);
+   }
+
+#if 0 /* DEBUG */
+   {
+      int i;
+      printf("Sampled at %f, %f, %f:\n", s[0], t[0], p[0]);
+      for (i = 0; i < 4; i++) {
+         printf("Frag %d: %f %f %f %f\n", i,
+                rgba[0][i],
+                rgba[1][i],
+                rgba[2][i],
+                rgba[3][i]);
+      }
+   }
+#endif
+}
+
+
+/**
+ * Called via tgsi_sampler::get_samples() when running a fragment shader.
+ * Get four filtered RGBA values from the sampler's texture.
+ */
+void
+sp_get_samples_fragment(struct tgsi_sampler *tgsi_sampler,
+                        const float s[QUAD_SIZE],
+                        const float t[QUAD_SIZE],
+                        const float p[QUAD_SIZE],
+                        float lodbias,
+                        float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   sp_get_samples(tgsi_sampler, s, t, p, TRUE, lodbias, rgba);
+}
+
+
+/**
+ * Called via tgsi_sampler::get_samples() when running a vertex shader.
+ * Get four filtered RGBA values from the sampler's texture.
+ */
+void
+sp_get_samples_vertex(struct tgsi_sampler *tgsi_sampler,
+                      const float s[QUAD_SIZE],
+                      const float t[QUAD_SIZE],
+                      const float p[QUAD_SIZE],
+                      float lodbias,
+                      float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+   sp_get_samples(tgsi_sampler, s, t, p, FALSE, lodbias, rgba);
+}
diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.h b/src/gallium/drivers/softpipe/sp_tex_sample.h
new file mode 100644
index 0000000000..40d8eb2c2a
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_tex_sample.h
@@ -0,0 +1,73 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef SP_TEX_SAMPLE_H
+#define SP_TEX_SAMPLE_H
+
+
+#include "tgsi/tgsi_exec.h"
+
+
+/**
+ * Subclass of tgsi_sampler
+ */
+struct sp_shader_sampler
+{
+   struct tgsi_sampler base;  /**< base class */
+
+   uint unit;
+   struct softpipe_context *sp;
+   struct softpipe_tile_cache *cache;
+};
+
+
+
+static INLINE const struct sp_shader_sampler *
+sp_shader_sampler(const struct tgsi_sampler *sampler)
+{
+   return (const struct sp_shader_sampler *) sampler;
+}
+
+
+extern void
+sp_get_samples_fragment(struct tgsi_sampler *tgsi_sampler,
+                        const float s[QUAD_SIZE],
+                        const float t[QUAD_SIZE],
+                        const float p[QUAD_SIZE],
+                        float lodbias,
+                        float rgba[NUM_CHANNELS][QUAD_SIZE]);
+
+extern void
+sp_get_samples_vertex(struct tgsi_sampler *tgsi_sampler,
+                      const float s[QUAD_SIZE],
+                      const float t[QUAD_SIZE],
+                      const float p[QUAD_SIZE],
+                      float lodbias,
+                      float rgba[NUM_CHANNELS][QUAD_SIZE]);
+
+
+#endif /* SP_TEX_SAMPLE_H */
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
new file mode 100644
index 0000000000..3eed0d0d29
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -0,0 +1,348 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Michel Dänzer <michel@tungstengraphics.com>
+  */
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "sp_context.h"
+#include "sp_state.h"
+#include "sp_texture.h"
+#include "sp_tile_cache.h"
+#include "sp_screen.h"
+
+
+/* Simple, maximally packed layout.
+ */
+
+static unsigned minify( unsigned d )
+{
+   return MAX2(1, d>>1);
+}
+
+
+/* Conventional allocation path for non-display textures:
+ */
+static boolean
+softpipe_texture_layout(struct pipe_screen *screen,
+                        struct softpipe_texture * spt)
+{
+   struct pipe_winsys *ws = screen->winsys;
+   struct pipe_texture *pt = &spt->base;
+   unsigned level;
+   unsigned width = pt->width[0];
+   unsigned height = pt->height[0];
+   unsigned depth = pt->depth[0];
+
+   unsigned buffer_size = 0;
+
+   for (level = 0; level <= pt->last_level; level++) {
+      pt->width[level] = width;
+      pt->height[level] = height;
+      pt->depth[level] = depth;
+      pt->nblocksx[level] = pf_get_nblocksx(&pt->block, width);  
+      pt->nblocksy[level] = pf_get_nblocksy(&pt->block, height);  
+      spt->stride[level] = pt->nblocksx[level]*pt->block.size;
+
+      spt->level_offset[level] = buffer_size;
+
+      buffer_size += (pt->nblocksy[level] *
+                      ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) *
+                      spt->stride[level]);
+
+      width  = minify(width);
+      height = minify(height);
+      depth = minify(depth);
+   }
+
+   spt->buffer = ws->buffer_create(ws, 32,
+                                   PIPE_BUFFER_USAGE_PIXEL,
+                                   buffer_size);
+
+   return spt->buffer != NULL;
+}
+
+static boolean
+softpipe_displaytarget_layout(struct pipe_screen *screen,
+                              struct softpipe_texture * spt)
+{
+   struct pipe_winsys *ws = screen->winsys;
+   unsigned usage = (PIPE_BUFFER_USAGE_CPU_READ_WRITE |
+                     PIPE_BUFFER_USAGE_GPU_READ_WRITE);
+
+   spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]);  
+   spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]);  
+
+   spt->buffer = ws->surface_buffer_create( ws, 
+                                            spt->base.width[0], 
+                                            spt->base.height[0],
+                                            spt->base.format,
+                                            usage,
+                                            &spt->stride[0]);
+
+   return spt->buffer != NULL;
+}
+
+
+
+
+
+static struct pipe_texture *
+softpipe_texture_create(struct pipe_screen *screen,
+                        const struct pipe_texture *templat)
+{
+   struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture);
+   if (!spt)
+      return NULL;
+
+   spt->base = *templat;
+   spt->base.refcount = 1;
+   spt->base.screen = screen;
+
+   if (spt->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) {
+      if (!softpipe_displaytarget_layout(screen, spt))
+         goto fail;
+   }
+   else {
+      if (!softpipe_texture_layout(screen, spt))
+         goto fail;
+   }
+    
+   assert(spt->base.refcount == 1);
+   return &spt->base;
+
+ fail:
+   FREE(spt);
+   return NULL;
+}
+
+
+static struct pipe_texture *
+softpipe_texture_blanket(struct pipe_screen * screen,
+                         const struct pipe_texture *base,
+                         const unsigned *stride,
+                         struct pipe_buffer *buffer)
+{
+   struct softpipe_texture *spt;
+   assert(screen);
+
+   /* Only supports one type */
+   if (base->target != PIPE_TEXTURE_2D ||
+       base->last_level != 0 ||
+       base->depth[0] != 1) {
+      return NULL;
+   }
+
+   spt = CALLOC_STRUCT(softpipe_texture);
+   if (!spt)
+      return NULL;
+
+   spt->base = *base;
+   spt->base.refcount = 1;
+   spt->base.screen = screen;
+   spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]);  
+   spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]);  
+   spt->stride[0] = stride[0];
+
+   pipe_buffer_reference(screen, &spt->buffer, buffer);
+
+   return &spt->base;
+}
+
+
+static void
+softpipe_texture_release(struct pipe_screen *screen,
+                         struct pipe_texture **pt)
+{
+   if (!*pt)
+      return;
+
+   if (--(*pt)->refcount <= 0) {
+      struct softpipe_texture *spt = softpipe_texture(*pt);
+
+      pipe_buffer_reference(screen, &spt->buffer, NULL);
+      FREE(spt);
+   }
+   *pt = NULL;
+}
+
+
+static struct pipe_surface *
+softpipe_get_tex_surface(struct pipe_screen *screen,
+                         struct pipe_texture *pt,
+                         unsigned face, unsigned level, unsigned zslice,
+                         unsigned usage)
+{
+   struct softpipe_texture *spt = softpipe_texture(pt);
+   struct pipe_surface *ps;
+
+   assert(level <= pt->last_level);
+
+   ps = CALLOC_STRUCT(pipe_surface);
+   if (ps) {
+      ps->refcount = 1;
+      pipe_texture_reference(&ps->texture, pt);
+      ps->format = pt->format;
+      ps->block = pt->block;
+      ps->width = pt->width[level];
+      ps->height = pt->height[level];
+      ps->nblocksx = pt->nblocksx[level];
+      ps->nblocksy = pt->nblocksy[level];
+      ps->stride = spt->stride[level];
+      ps->offset = spt->level_offset[level];
+      ps->usage = usage;
+
+      /* Because we are softpipe, anything that the state tracker
+       * thought was going to be done with the GPU will actually get
+       * done with the CPU.  Let's adjust the flags to take that into
+       * account.
+       */
+      if (ps->usage & PIPE_BUFFER_USAGE_GPU_WRITE) {
+         /* GPU_WRITE means "render" and that can involve reads (blending) */
+         ps->usage |= PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_CPU_READ;
+      }
+
+      if (ps->usage & PIPE_BUFFER_USAGE_GPU_READ)
+         ps->usage |= PIPE_BUFFER_USAGE_CPU_READ;
+
+      if (ps->usage & (PIPE_BUFFER_USAGE_CPU_WRITE |
+                       PIPE_BUFFER_USAGE_GPU_WRITE)) {
+         /* Mark the surface as dirty.  The tile cache will look for this. */
+         spt->modified = TRUE;
+      }
+
+      ps->face = face;
+      ps->level = level;
+      ps->zslice = zslice;
+
+      if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) {
+         ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) *
+            ps->nblocksy *
+            ps->stride;
+      }
+      else {
+         assert(face == 0);
+         assert(zslice == 0);
+      }
+   }
+   return ps;
+}
+
+
+static void 
+softpipe_tex_surface_release(struct pipe_screen *screen, 
+                             struct pipe_surface **s)
+{
+   struct pipe_surface *surf = *s;
+   /* Effectively do the texture_update work here - if texture images
+    * needed post-processing to put them into hardware layout, this is
+    * where it would happen.  For softpipe, nothing to do.
+    */
+   assert(surf->texture);
+   if (--surf->refcount == 0) {
+      pipe_texture_reference(&surf->texture, NULL);
+      FREE(surf);
+   }
+   *s = NULL;
+}
+
+
+static void *
+softpipe_surface_map( struct pipe_screen *screen,
+                      struct pipe_surface *surface,
+                      unsigned flags )
+{
+   ubyte *map;
+   struct softpipe_texture *spt;
+
+   if (flags & ~surface->usage) {
+      assert(0);
+      return NULL;
+   }
+
+   assert(surface->texture);
+   spt = softpipe_texture(surface->texture);
+   map = pipe_buffer_map(screen, spt->buffer, flags);
+   if (map == NULL)
+      return NULL;
+
+   /* May want to different things here depending on read/write nature
+    * of the map:
+    */
+   if (surface->texture &&
+       (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) 
+   {
+      /* Do something to notify sharing contexts of a texture change.
+       * In softpipe, that would mean flushing the texture cache.
+       */
+      softpipe_screen(screen)->timestamp++;
+   }
+   
+   return map + surface->offset;
+}
+
+
+static void
+softpipe_surface_unmap(struct pipe_screen *screen,
+                       struct pipe_surface *surface)
+{
+   struct softpipe_texture *spt;
+
+   assert(surface->texture);
+   spt = softpipe_texture(surface->texture);
+
+   pipe_buffer_unmap( screen, spt->buffer );
+}
+
+
+void
+softpipe_init_texture_funcs(struct softpipe_context *sp)
+{
+}
+
+
+void
+softpipe_init_screen_texture_funcs(struct pipe_screen *screen)
+{
+   screen->texture_create = softpipe_texture_create;
+   screen->texture_blanket = softpipe_texture_blanket;
+   screen->texture_release = softpipe_texture_release;
+
+   screen->get_tex_surface = softpipe_get_tex_surface;
+   screen->tex_surface_release = softpipe_tex_surface_release;
+
+   screen->surface_map = softpipe_surface_map;
+   screen->surface_unmap = softpipe_surface_unmap;
+}
diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h
new file mode 100644
index 0000000000..c1636920cd
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_texture.h
@@ -0,0 +1,70 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef SP_TEXTURE_H
+#define SP_TEXTURE_H
+
+
+#include "pipe/p_state.h"
+
+
+struct pipe_context;
+struct pipe_screen;
+struct softpipe_context;
+
+
+struct softpipe_texture
+{
+   struct pipe_texture base;
+
+   unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS];
+   unsigned stride[PIPE_MAX_TEXTURE_LEVELS];
+
+   /* The data is held here:
+    */
+   struct pipe_buffer *buffer;
+
+   boolean modified;
+};
+
+
+/** cast wrapper */
+static INLINE struct softpipe_texture *
+softpipe_texture(struct pipe_texture *pt)
+{
+   return (struct softpipe_texture *) pt;
+}
+
+
+extern void
+softpipe_init_texture_funcs( struct softpipe_context *softpipe );
+
+extern void
+softpipe_init_screen_texture_funcs(struct pipe_screen *screen);
+
+
+#endif /* SP_TEXTURE */
diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c
new file mode 100644
index 0000000000..ab76009375
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.c
@@ -0,0 +1,614 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Framebuffer/surface tile caching.
+ *
+ * Author:
+ *    Brian Paul
+ */
+
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+#include "util/u_tile.h"
+#include "sp_context.h"
+#include "sp_surface.h"
+#include "sp_texture.h"
+#include "sp_tile_cache.h"
+
+#define NUM_ENTRIES 32
+
+
+/** XXX move these */
+#define MAX_WIDTH 2048
+#define MAX_HEIGHT 2048
+
+
+struct softpipe_tile_cache
+{
+   struct pipe_screen *screen;
+   struct pipe_surface *surface;  /**< the surface we're caching */
+   void *surface_map;
+   struct pipe_texture *texture;  /**< if caching a texture */
+   struct softpipe_cached_tile entries[NUM_ENTRIES];
+   uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32];
+   float clear_color[4];
+   uint clear_val;
+   boolean depth_stencil; /** Is the surface a depth/stencil format? */
+
+   struct pipe_surface *tex_surf;
+   void *tex_surf_map;
+   int tex_face, tex_level, tex_z;
+
+   struct softpipe_cached_tile tile;  /**< scratch tile for clears */
+};
+
+
+/**
+ * Return the position in the cache for the tile that contains win pos (x,y).
+ * We currently use a direct mapped cache so this is like a hack key.
+ * At some point we should investige something more sophisticated, like
+ * a LRU replacement policy.
+ */
+#define CACHE_POS(x, y) \
+   (((x) / TILE_SIZE + ((y) / TILE_SIZE) * 5) % NUM_ENTRIES)
+
+
+
+/**
+ * Is the tile at (x,y) in cleared state?
+ */
+static INLINE uint
+is_clear_flag_set(const uint *bitvec, int x, int y)
+{
+   int pos, bit;
+   x /= TILE_SIZE;
+   y /= TILE_SIZE;
+   pos = y * (MAX_WIDTH / TILE_SIZE) + x;
+   assert(pos / 32 < (MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32);
+   bit = bitvec[pos / 32] & (1 << (pos & 31));
+   return bit;
+}
+   
+
+/**
+ * Mark the tile at (x,y) as not cleared.
+ */
+static INLINE void
+clear_clear_flag(uint *bitvec, int x, int y)
+{
+   int pos;
+   x /= TILE_SIZE;
+   y /= TILE_SIZE;
+   pos = y * (MAX_WIDTH / TILE_SIZE) + x;
+   assert(pos / 32 < (MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32);
+   bitvec[pos / 32] &= ~(1 << (pos & 31));
+}
+   
+
+struct softpipe_tile_cache *
+sp_create_tile_cache( struct pipe_screen *screen )
+{
+   struct softpipe_tile_cache *tc;
+   uint pos;
+
+   tc = CALLOC_STRUCT( softpipe_tile_cache );
+   if (tc) {
+      tc->screen = screen;
+      for (pos = 0; pos < NUM_ENTRIES; pos++) {
+         tc->entries[pos].x =
+         tc->entries[pos].y = -1;
+      }
+   }
+   return tc;
+}
+
+
+void
+sp_destroy_tile_cache(struct softpipe_tile_cache *tc)
+{
+   uint pos;
+
+   for (pos = 0; pos < NUM_ENTRIES; pos++) {
+      /*assert(tc->entries[pos].x < 0);*/
+   }
+   if (tc->surface) {
+      pipe_surface_reference(&tc->surface, NULL);
+   }
+   if (tc->tex_surf) {
+      pipe_surface_reference(&tc->tex_surf, NULL);
+   }
+
+   FREE( tc );
+}
+
+
+/**
+ * Specify the surface to cache.
+ */
+void
+sp_tile_cache_set_surface(struct softpipe_tile_cache *tc,
+                          struct pipe_surface *ps)
+{
+   assert(!tc->texture);
+
+   if (tc->surface_map) {
+      tc->screen->surface_unmap(tc->screen, tc->surface);
+      tc->surface_map = NULL;
+   }
+
+   pipe_surface_reference(&tc->surface, ps);
+
+   if (tc->surface) {
+      if (tc->surface_map) /* XXX: this is always NULL!? */
+	 tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface,
+                                                   PIPE_BUFFER_USAGE_CPU_READ | 
+                                                   PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM ||
+                           ps->format == PIPE_FORMAT_X8Z24_UNORM ||
+                           ps->format == PIPE_FORMAT_Z24S8_UNORM ||
+                           ps->format == PIPE_FORMAT_Z24X8_UNORM ||
+                           ps->format == PIPE_FORMAT_Z16_UNORM ||
+                           ps->format == PIPE_FORMAT_Z32_UNORM ||
+                           ps->format == PIPE_FORMAT_S8_UNORM);
+   }
+}
+
+
+/**
+ * Return the surface being cached.
+ */
+struct pipe_surface *
+sp_tile_cache_get_surface(struct softpipe_tile_cache *tc)
+{
+   return tc->surface;
+}
+
+
+void
+sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc)
+{
+   if (tc->surface && !tc->surface_map)
+      tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface,
+                                                PIPE_BUFFER_USAGE_CPU_WRITE |
+                                                PIPE_BUFFER_USAGE_CPU_READ);
+
+   if (tc->tex_surf && !tc->tex_surf_map)
+      tc->tex_surf_map = tc->screen->surface_map(tc->screen, tc->tex_surf,
+                                                 PIPE_BUFFER_USAGE_CPU_READ);
+}
+
+
+void
+sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc)
+{
+   if (tc->surface_map) {
+      tc->screen->surface_unmap(tc->screen, tc->surface);
+      tc->surface_map = NULL;
+   }
+
+   if (tc->tex_surf_map) {
+      tc->screen->surface_unmap(tc->screen, tc->tex_surf);
+      tc->tex_surf_map = NULL;
+   }
+}
+
+
+/**
+ * Specify the texture to cache.
+ */
+void
+sp_tile_cache_set_texture(struct pipe_context *pipe,
+                          struct softpipe_tile_cache *tc,
+                          struct pipe_texture *texture)
+{
+   uint i;
+
+   assert(!tc->surface);
+
+   pipe_texture_reference(&tc->texture, texture);
+
+   if (tc->tex_surf_map) {
+      tc->screen->surface_unmap(tc->screen, tc->tex_surf);
+      tc->tex_surf_map = NULL;
+   }
+   pipe_surface_reference(&tc->tex_surf, NULL);
+
+   /* mark as entries as invalid/empty */
+   /* XXX we should try to avoid this when the teximage hasn't changed */
+   for (i = 0; i < NUM_ENTRIES; i++) {
+      tc->entries[i].x = -1;
+   }
+
+   tc->tex_face = -1; /* any invalid value here */
+}
+
+
+/**
+ * Set pixels in a tile to the given clear color/value, float.
+ */
+static void
+clear_tile_rgba(struct softpipe_cached_tile *tile,
+                enum pipe_format format,
+                const float clear_value[4])
+{
+   if (clear_value[0] == 0.0 &&
+       clear_value[1] == 0.0 &&
+       clear_value[2] == 0.0 &&
+       clear_value[3] == 0.0) {
+      memset(tile->data.color, 0, sizeof(tile->data.color));
+   }
+   else {
+      uint i, j;
+      for (i = 0; i < TILE_SIZE; i++) {
+         for (j = 0; j < TILE_SIZE; j++) {
+            tile->data.color[i][j][0] = clear_value[0];
+            tile->data.color[i][j][1] = clear_value[1];
+            tile->data.color[i][j][2] = clear_value[2];
+            tile->data.color[i][j][3] = clear_value[3];
+         }
+      }
+   }
+}
+
+
+/**
+ * Set a tile to a solid value/color.
+ */
+static void
+clear_tile(struct softpipe_cached_tile *tile,
+           enum pipe_format format,
+           uint clear_value)
+{
+   uint i, j;
+
+   switch (pf_get_size(format)) {
+   case 1:
+      memset(tile->data.any, 0, TILE_SIZE * TILE_SIZE);
+      break;
+   case 2:
+      if (clear_value == 0) {
+         memset(tile->data.any, 0, 2 * TILE_SIZE * TILE_SIZE);
+      }
+      else {
+         for (i = 0; i < TILE_SIZE; i++) {
+            for (j = 0; j < TILE_SIZE; j++) {
+               tile->data.depth16[i][j] = (ushort) clear_value;
+            }
+         }
+      }
+      break;
+   case 4:
+      if (clear_value == 0) {
+         memset(tile->data.any, 0, 4 * TILE_SIZE * TILE_SIZE);
+      }
+      else {
+         for (i = 0; i < TILE_SIZE; i++) {
+            for (j = 0; j < TILE_SIZE; j++) {
+               tile->data.color32[i][j] = clear_value;
+            }
+         }
+      }
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+/**
+ * Actually clear the tiles which were flagged as being in a clear state.
+ */
+static void
+sp_tile_cache_flush_clear(struct pipe_context *pipe,
+                          struct softpipe_tile_cache *tc)
+{
+   struct pipe_surface *ps = tc->surface;
+   const uint w = tc->surface->width;
+   const uint h = tc->surface->height;
+   uint x, y;
+   uint numCleared = 0;
+
+   /* clear the scratch tile to the clear value */
+   clear_tile(&tc->tile, ps->format, tc->clear_val);
+
+   /* push the tile to all positions marked as clear */
+   for (y = 0; y < h; y += TILE_SIZE) {
+      for (x = 0; x < w; x += TILE_SIZE) {
+         if (is_clear_flag_set(tc->clear_flags, x, y)) {
+            pipe_put_tile_raw(ps,
+                              x, y, TILE_SIZE, TILE_SIZE,
+                              tc->tile.data.color32, 0/*STRIDE*/);
+
+            /* do this? */
+            clear_clear_flag(tc->clear_flags, x, y);
+
+            numCleared++;
+         }
+      }
+   }
+#if 0
+   debug_printf("num cleared: %u\n", numCleared);
+#endif
+}
+
+
+/**
+ * Flush the tile cache: write all dirty tiles back to the surface.
+ * any tiles "flagged" as cleared will be "really" cleared.
+ */
+void
+sp_flush_tile_cache(struct softpipe_context *softpipe,
+                    struct softpipe_tile_cache *tc)
+{
+   struct pipe_surface *ps = tc->surface;
+   int inuse = 0, pos;
+
+   if (ps) {
+      /* caching a drawing surface */
+      for (pos = 0; pos < NUM_ENTRIES; pos++) {
+         struct softpipe_cached_tile *tile = tc->entries + pos;
+         if (tile->x >= 0) {
+            if (tc->depth_stencil) {
+               pipe_put_tile_raw(ps,
+                                 tile->x, tile->y, TILE_SIZE, TILE_SIZE,
+                                 tile->data.depth32, 0/*STRIDE*/);
+            }
+            else {
+               pipe_put_tile_rgba(ps,
+                                  tile->x, tile->y, TILE_SIZE, TILE_SIZE,
+                                  (float *) tile->data.color);
+            }
+            tile->x = tile->y = -1;  /* mark as empty */
+            inuse++;
+         }
+      }
+
+#if TILE_CLEAR_OPTIMIZATION
+      sp_tile_cache_flush_clear(&softpipe->pipe, tc);
+#endif
+   }
+   else if (tc->texture) {
+      /* caching a texture, mark all entries as empty */
+      for (pos = 0; pos < NUM_ENTRIES; pos++) {
+         tc->entries[pos].x = -1;
+      }
+      tc->tex_face = -1;
+   }
+
+#if 0
+   debug_printf("flushed tiles in use: %d\n", inuse);
+#endif
+}
+
+
+/**
+ * Get a tile from the cache.
+ * \param x, y  position of tile, in pixels
+ */
+struct softpipe_cached_tile *
+sp_get_cached_tile(struct softpipe_context *softpipe,
+                   struct softpipe_tile_cache *tc, int x, int y)
+{
+   struct pipe_surface *ps = tc->surface;
+
+   /* tile pos in framebuffer: */
+   const int tile_x = x & ~(TILE_SIZE - 1);
+   const int tile_y = y & ~(TILE_SIZE - 1);
+
+   /* cache pos/entry: */
+   const int pos = CACHE_POS(x, y);
+   struct softpipe_cached_tile *tile = tc->entries + pos;
+
+   if (tile_x != tile->x ||
+       tile_y != tile->y) {
+
+      if (tile->x != -1) {
+         /* put dirty tile back in framebuffer */
+         if (tc->depth_stencil) {
+            pipe_put_tile_raw(ps,
+                              tile->x, tile->y, TILE_SIZE, TILE_SIZE,
+                              tile->data.depth32, 0/*STRIDE*/);
+         }
+         else {
+            pipe_put_tile_rgba(ps,
+                               tile->x, tile->y, TILE_SIZE, TILE_SIZE,
+                               (float *) tile->data.color);
+         }
+      }
+
+      tile->x = tile_x;
+      tile->y = tile_y;
+
+      if (is_clear_flag_set(tc->clear_flags, x, y)) {
+         /* don't get tile from framebuffer, just clear it */
+         if (tc->depth_stencil) {
+            clear_tile(tile, ps->format, tc->clear_val);
+         }
+         else {
+            clear_tile_rgba(tile, ps->format, tc->clear_color);
+         }
+         clear_clear_flag(tc->clear_flags, x, y);
+      }
+      else {
+         /* get new tile data from surface */
+         if (tc->depth_stencil) {
+            pipe_get_tile_raw(ps,
+                              tile->x, tile->y, TILE_SIZE, TILE_SIZE,
+                              tile->data.depth32, 0/*STRIDE*/);
+         }
+         else {
+            pipe_get_tile_rgba(ps,
+                               tile->x, tile->y, TILE_SIZE, TILE_SIZE,
+                               (float *) tile->data.color);
+         }
+      }
+   }
+
+   return tile;
+}
+
+
+/**
+ * Given the texture face, level, zslice, x and y values, compute
+ * the cache entry position/index where we'd hope to find the
+ * cached texture tile.
+ * This is basically a direct-map cache.
+ * XXX There's probably lots of ways in which we can improve this.
+ */
+static INLINE uint
+tex_cache_pos(int x, int y, int z, int face, int level)
+{
+   uint entry = x + y * 5 + z * 4 + face + level;
+   return entry % NUM_ENTRIES;
+}
+
+
+/**
+ * Similar to sp_get_cached_tile() but for textures.
+ * Tiles are read-only and indexed with more params.
+ */
+const struct softpipe_cached_tile *
+sp_get_cached_tile_tex(struct softpipe_context *sp,
+                       struct softpipe_tile_cache *tc, int x, int y, int z,
+                       int face, int level)
+{
+   struct pipe_screen *screen = sp->pipe.screen;
+   /* tile pos in framebuffer: */
+   const int tile_x = x & ~(TILE_SIZE - 1);
+   const int tile_y = y & ~(TILE_SIZE - 1);
+   /* cache pos/entry: */
+   const uint pos = tex_cache_pos(x / TILE_SIZE, y / TILE_SIZE, z,
+                                  face, level);
+   struct softpipe_cached_tile *tile = tc->entries + pos;
+
+   if (tc->texture) {
+      struct softpipe_texture *spt = softpipe_texture(tc->texture);
+      if (spt->modified) {
+         /* texture was modified, force a cache reload */
+         tile->x = -1;
+         spt->modified = FALSE;
+      }
+   }
+
+   if (tile_x != tile->x ||
+       tile_y != tile->y ||
+       z != tile->z ||
+       face != tile->face ||
+       level != tile->level) {
+      /* cache miss */
+
+      /* check if we need to get a new surface */
+      if (!tc->tex_surf ||
+          tc->tex_face != face ||
+          tc->tex_level != level ||
+          tc->tex_z != z) {
+         /* get new surface (view into texture) */
+
+	 if (tc->tex_surf_map)
+            tc->screen->surface_unmap(tc->screen, tc->tex_surf);
+
+         tc->tex_surf = screen->get_tex_surface(screen, tc->texture, face, level, z, 
+                                                PIPE_BUFFER_USAGE_CPU_READ);
+         tc->tex_surf_map = screen->surface_map(screen, tc->tex_surf,
+                                                PIPE_BUFFER_USAGE_CPU_READ);
+
+         tc->tex_face = face;
+         tc->tex_level = level;
+         tc->tex_z = z;
+      }
+
+      /* get tile from the surface (view into texture) */
+      pipe_get_tile_rgba(tc->tex_surf,
+                         tile_x, tile_y, TILE_SIZE, TILE_SIZE,
+                         (float *) tile->data.color);
+      tile->x = tile_x;
+      tile->y = tile_y;
+      tile->z = z;
+      tile->face = face;
+      tile->level = level;
+   }
+
+   return tile;
+}
+
+
+/**
+ * When a whole surface is being cleared to a value we can avoid
+ * fetching tiles above.
+ * Save the color and set a 'clearflag' for each tile of the screen.
+ */
+void
+sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue)
+{
+   uint r, g, b, a;
+   uint pos;
+
+   tc->clear_val = clearValue;
+
+   switch (tc->surface->format) {
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      r = (clearValue >> 24) & 0xff;
+      g = (clearValue >> 16) & 0xff;
+      b = (clearValue >>  8) & 0xff;
+      a = (clearValue      ) & 0xff;
+      break;
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      r = (clearValue >> 16) & 0xff;
+      g = (clearValue >>  8) & 0xff;
+      b = (clearValue      ) & 0xff;
+      a = (clearValue >> 24) & 0xff;
+      break;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      r = (clearValue >>  8) & 0xff;
+      g = (clearValue >> 16) & 0xff;
+      b = (clearValue >> 24) & 0xff;
+      a = (clearValue      ) & 0xff;
+      break;
+   default:
+      r = g = b = a = 0;
+   }
+
+   tc->clear_color[0] = r / 255.0f;
+   tc->clear_color[1] = g / 255.0f;
+   tc->clear_color[2] = b / 255.0f;
+   tc->clear_color[3] = a / 255.0f;
+
+#if TILE_CLEAR_OPTIMIZATION
+   /* set flags to indicate all the tiles are cleared */
+   memset(tc->clear_flags, 255, sizeof(tc->clear_flags));
+#else
+   /* disable the optimization */
+   memset(tc->clear_flags, 0, sizeof(tc->clear_flags));
+#endif
+
+   for (pos = 0; pos < NUM_ENTRIES; pos++) {
+      struct softpipe_cached_tile *tile = tc->entries + pos;
+      tile->x = tile->y = -1;
+   }
+}
diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.h b/src/gallium/drivers/softpipe/sp_tile_cache.h
new file mode 100644
index 0000000000..a66bb50bcc
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.h
@@ -0,0 +1,105 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef SP_TILE_CACHE_H
+#define SP_TILE_CACHE_H
+
+#define TILE_CLEAR_OPTIMIZATION 1
+
+
+#include "pipe/p_compiler.h"
+
+
+struct softpipe_context;
+struct softpipe_tile_cache;
+
+
+/**
+ * Cache tile size (width and height). This needs to be a power of two.
+ */
+#define TILE_SIZE 64
+
+
+
+struct softpipe_cached_tile
+{
+   int x, y;           /**< pos of tile in window coords */
+   int z, face, level; /**< Extra texture indexes */
+   union {
+      float color[TILE_SIZE][TILE_SIZE][4];
+      uint color32[TILE_SIZE][TILE_SIZE];
+      uint depth32[TILE_SIZE][TILE_SIZE];
+      ushort depth16[TILE_SIZE][TILE_SIZE];
+      ubyte stencil8[TILE_SIZE][TILE_SIZE];
+      ubyte any[1];
+   } data;
+};
+
+
+extern struct softpipe_tile_cache *
+sp_create_tile_cache( struct pipe_screen *screen );
+
+extern void
+sp_destroy_tile_cache(struct softpipe_tile_cache *tc);
+
+extern void
+sp_tile_cache_set_surface(struct softpipe_tile_cache *tc,
+                          struct pipe_surface *sps);
+
+extern struct pipe_surface *
+sp_tile_cache_get_surface(struct softpipe_tile_cache *tc);
+
+extern void
+sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc);
+
+extern void
+sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc);
+
+extern void
+sp_tile_cache_set_texture(struct pipe_context *pipe,
+                          struct softpipe_tile_cache *tc,
+                          struct pipe_texture *texture);
+
+extern void
+sp_flush_tile_cache(struct softpipe_context *softpipe,
+                    struct softpipe_tile_cache *tc);
+
+extern void
+sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue);
+
+extern struct softpipe_cached_tile *
+sp_get_cached_tile(struct softpipe_context *softpipe,
+                   struct softpipe_tile_cache *tc, int x, int y);
+
+extern const struct softpipe_cached_tile *
+sp_get_cached_tile_tex(struct softpipe_context *softpipe,
+                       struct softpipe_tile_cache *tc, int x, int y, int z,
+                       int face, int level);
+
+
+#endif /* SP_TILE_CACHE_H */
+
diff --git a/src/gallium/drivers/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h
new file mode 100644
index 0000000000..4ab666486c
--- /dev/null
+++ b/src/gallium/drivers/softpipe/sp_winsys.h
@@ -0,0 +1,73 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* This is the interface that softpipe requires any window system
+ * hosting it to implement.  This is the only include file in softpipe
+ * which is public.
+ */
+
+
+#ifndef SP_WINSYS_H
+#define SP_WINSYS_H
+
+
+#include "pipe/p_compiler.h" /* for boolean */
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+enum pipe_format;
+
+struct softpipe_winsys {
+   /** test if the given format is supported for front/back color bufs */
+   boolean (*is_format_supported)( struct softpipe_winsys *sws,
+                                   enum pipe_format format );
+
+};
+
+struct pipe_screen;
+struct pipe_winsys;
+struct pipe_context;
+
+
+struct pipe_context *softpipe_create( struct pipe_screen *,
+                                      struct pipe_winsys *,
+				      void *unused );
+
+
+struct pipe_screen *
+softpipe_create_screen(struct pipe_winsys *);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* SP_WINSYS_H */
diff --git a/src/gallium/drivers/trace/Makefile b/src/gallium/drivers/trace/Makefile
new file mode 100644
index 0000000000..3859b8acb0
--- /dev/null
+++ b/src/gallium/drivers/trace/Makefile
@@ -0,0 +1,18 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = trace
+
+C_SOURCES = \
+	tr_context.c \
+	tr_dump.c \
+	tr_screen.c \
+	tr_state.c \
+	tr_texture.c \
+	tr_winsys.c 
+
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README
new file mode 100644
index 0000000000..f0e1cd596d
--- /dev/null
+++ b/src/gallium/drivers/trace/README
@@ -0,0 +1,64 @@
+                             TRACE PIPE DRIVER
+
+
+= About =
+
+This directory contains a Gallium3D pipe driver which traces all incoming calls.
+
+
+= Build Instructions =
+
+To build, invoke scons on the top dir as
+ 
+ scons statetrackers=mesa drivers=softpipe,i965simple,trace winsys=xlib
+
+
+= Usage =
+
+To use do
+
+ ln -s libGL.so build/linux-x86-debug/gallium/winsys/xlib/libGL.so.1
+ export LD_LIBRARY_PATH=$PWD/build/linux-x86-debug/gallium/winsys/xlib
+
+ensure the right libGL.so is being picked by doing
+
+ ldd progs/trivial/tri 
+
+and then try running
+
+ GALLIUM_TRACE=tri.trace progs/trivial/tri
+
+which should create a tri.trace file, which is an XML file. You can view copying 
+trace.xsl to the same directory, and opening with a XSLT capable browser such as 
+Firefox or Internet Explorer.
+
+
+= Integrating =
+
+You can integrate the trace pipe driver either inside the state tracker or the 
+winsys. The procedure on both cases is the same. Let's assume you have a 
+pipe_screen and a pipe_context pair obtained by the usual means (variable and 
+function names are just for illustration purposes):
+
+  real_screen = real_screen_create(...);
+  
+  real_context = real_context_create(...);
+  
+The trace screen and pipe_context is then created by doing
+
+  trace_screen = trace_screen_create(real_screen);
+  
+  trace_context = trace_context_create(trace_screen, real_context);
+  
+You can then simply use trace_screen and trace_context instead of real_screen
+and real_context.
+
+Do not call trace_winsys_create. Simply pass trace_screen->winsys or 
+trace_context->winsys in places you would pass winsys.
+
+You can create as many contexts you wish. Just ensure that you don't mistake 
+trace_screen with real_screen when creating them.
+
+
+--
+Jose Fonseca <jrfonseca@tungstengraphics.com>
diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript
new file mode 100644
index 0000000000..0a6bfb8f4c
--- /dev/null
+++ b/src/gallium/drivers/trace/SConscript
@@ -0,0 +1,16 @@
+Import('*')
+
+env = env.Clone()
+
+trace = env.ConvenienceLibrary(
+    target = 'trace',
+    source = [
+        'tr_context.c',
+        'tr_dump.c',
+        'tr_screen.c',
+        'tr_state.c',
+        'tr_texture.c',
+        'tr_winsys.c',
+    ])
+
+Export('trace')
\ No newline at end of file
diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c
new file mode 100644
index 0000000000..ec8be27077
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_context.c
@@ -0,0 +1,1072 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_memory.h"
+#include "pipe/p_screen.h"
+
+#include "tr_dump.h"
+#include "tr_state.h"
+#include "tr_screen.h"
+#include "tr_texture.h"
+#include "tr_winsys.h"
+#include "tr_context.h"
+
+
+static INLINE struct pipe_texture * 
+trace_texture_unwrap(struct trace_context *tr_ctx,
+                     struct pipe_texture *texture)
+{
+   struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen); 
+   struct trace_texture *tr_tex;
+   
+   if(!texture)
+      return NULL;
+   
+   tr_tex = trace_texture(tr_scr, texture);
+   
+   assert(tr_tex->texture);
+   assert(tr_tex->texture->screen == tr_scr->screen);
+   return tr_tex->texture;
+}
+
+
+static INLINE struct pipe_surface * 
+trace_surface_unwrap(struct trace_context *tr_ctx,
+                     struct pipe_surface *surface)
+{
+   struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen); 
+   struct trace_texture *tr_tex;
+   struct trace_surface *tr_surf;
+   
+   if(!surface)
+      return NULL;
+
+   assert(surface->texture);
+   if(!surface->texture)
+      return surface;
+   
+   tr_tex = trace_texture(tr_scr, surface->texture);
+   tr_surf = trace_surface(tr_tex, surface);
+   
+   assert(tr_surf->surface);
+   assert(tr_surf->surface->texture->screen == tr_scr->screen);
+   return tr_surf->surface;
+}
+
+
+static INLINE void
+trace_context_set_edgeflags(struct pipe_context *_pipe,
+                            const unsigned *bitfield)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "set_edgeflags");
+   
+   trace_dump_arg(ptr, pipe);
+   /* FIXME: we don't know how big this array is */
+   trace_dump_arg(ptr, bitfield);
+
+   pipe->set_edgeflags(pipe, bitfield);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE boolean
+trace_context_draw_arrays(struct pipe_context *_pipe,
+                          unsigned mode, unsigned start, unsigned count)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   boolean result;
+
+   trace_dump_call_begin("pipe_context", "draw_arrays");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(uint, mode);
+   trace_dump_arg(uint, start);
+   trace_dump_arg(uint, count);
+
+   result = pipe->draw_arrays(pipe, mode, start, count);;
+
+   trace_dump_ret(bool, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static INLINE boolean
+trace_context_draw_elements(struct pipe_context *_pipe,
+                          struct pipe_buffer *indexBuffer,
+                          unsigned indexSize,
+                          unsigned mode, unsigned start, unsigned count)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   boolean result;
+
+   trace_winsys_user_buffer_update(_pipe->winsys, indexBuffer);
+
+   trace_dump_call_begin("pipe_context", "draw_elements");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, indexBuffer);
+   trace_dump_arg(uint, indexSize);
+   trace_dump_arg(uint, mode);
+   trace_dump_arg(uint, start);
+   trace_dump_arg(uint, count);
+
+   result = pipe->draw_elements(pipe, indexBuffer, indexSize, mode, start, count);;
+
+   trace_dump_ret(bool, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static INLINE boolean
+trace_context_draw_range_elements(struct pipe_context *_pipe,
+                                  struct pipe_buffer *indexBuffer,
+                                  unsigned indexSize,
+                                  unsigned minIndex,
+                                  unsigned maxIndex,
+                                  unsigned mode, 
+                                  unsigned start, 
+                                  unsigned count)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   boolean result;
+
+   trace_winsys_user_buffer_update(_pipe->winsys, indexBuffer);
+
+   trace_dump_call_begin("pipe_context", "draw_range_elements");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, indexBuffer);
+   trace_dump_arg(uint, indexSize);
+   trace_dump_arg(uint, minIndex);
+   trace_dump_arg(uint, maxIndex);
+   trace_dump_arg(uint, mode);
+   trace_dump_arg(uint, start);
+   trace_dump_arg(uint, count);
+
+   result = pipe->draw_range_elements(pipe, 
+                                      indexBuffer, 
+                                      indexSize, minIndex, maxIndex, 
+                                      mode, start, count);
+   
+   trace_dump_ret(bool, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static INLINE struct pipe_query *
+trace_context_create_query(struct pipe_context *_pipe,
+                           unsigned query_type)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   struct pipe_query *result;
+
+   trace_dump_call_begin("pipe_context", "create_query");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(uint, query_type);
+
+   result = pipe->create_query(pipe, query_type);;
+
+   trace_dump_ret(ptr, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static INLINE void
+trace_context_destroy_query(struct pipe_context *_pipe,
+                            struct pipe_query *query)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "destroy_query");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, query);
+
+   pipe->destroy_query(pipe, query);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_begin_query(struct pipe_context *_pipe, 
+                          struct pipe_query *query)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "begin_query");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, query);
+
+   pipe->begin_query(pipe, query);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_end_query(struct pipe_context *_pipe, 
+                        struct pipe_query *query)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "end_query");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, query);
+
+   pipe->end_query(pipe, query);
+
+   trace_dump_call_end();
+}
+
+
+static INLINE boolean
+trace_context_get_query_result(struct pipe_context *_pipe, 
+                               struct pipe_query *query,
+                               boolean wait,
+                               uint64_t *presult)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   uint64_t result;
+   boolean _result;
+
+   trace_dump_call_begin("pipe_context", "get_query_result");
+
+   trace_dump_arg(ptr, pipe);
+
+   _result = pipe->get_query_result(pipe, query, wait, presult);;
+   result = *presult;
+
+   trace_dump_arg(uint, result);
+   trace_dump_ret(bool, _result);
+   
+   trace_dump_call_end();
+   
+   return _result;
+}
+
+
+static INLINE void *
+trace_context_create_blend_state(struct pipe_context *_pipe,
+                                 const struct pipe_blend_state *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   void * result;
+
+   trace_dump_call_begin("pipe_context", "create_blend_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(blend_state, state);
+
+   result = pipe->create_blend_state(pipe, state);;
+
+   trace_dump_ret(ptr, result);
+
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static INLINE void
+trace_context_bind_blend_state(struct pipe_context *_pipe, 
+                               void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "bind_blend_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->bind_blend_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_delete_blend_state(struct pipe_context *_pipe, 
+                                 void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "delete_blend_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->delete_blend_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void *
+trace_context_create_sampler_state(struct pipe_context *_pipe,
+                                   const struct pipe_sampler_state *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   void * result;
+
+   trace_dump_call_begin("pipe_context", "create_sampler_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(sampler_state, state);
+
+   result = pipe->create_sampler_state(pipe, state);;
+
+   trace_dump_ret(ptr, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static INLINE void
+trace_context_bind_sampler_states(struct pipe_context *_pipe, 
+                                  unsigned num_states, void **states)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "bind_sampler_states");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(uint, num_states);
+   trace_dump_arg_array(ptr, states, num_states);
+
+   pipe->bind_sampler_states(pipe, num_states, states);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_delete_sampler_state(struct pipe_context *_pipe, 
+                                   void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "delete_sampler_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->delete_sampler_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void *
+trace_context_create_rasterizer_state(struct pipe_context *_pipe,
+                                      const struct pipe_rasterizer_state *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   void * result;
+
+   trace_dump_call_begin("pipe_context", "create_rasterizer_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(rasterizer_state, state);
+
+   result = pipe->create_rasterizer_state(pipe, state);;
+
+   trace_dump_ret(ptr, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static INLINE void
+trace_context_bind_rasterizer_state(struct pipe_context *_pipe, 
+                                    void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "bind_rasterizer_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->bind_rasterizer_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_delete_rasterizer_state(struct pipe_context *_pipe, 
+                                      void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "delete_rasterizer_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->delete_rasterizer_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void *
+trace_context_create_depth_stencil_alpha_state(struct pipe_context *_pipe,
+                                               const struct pipe_depth_stencil_alpha_state *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   void * result;
+
+   trace_dump_call_begin("pipe_context", "create_depth_stencil_alpha_state");
+
+   result = pipe->create_depth_stencil_alpha_state(pipe, state);;
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(depth_stencil_alpha_state, state);
+   
+   trace_dump_ret(ptr, result);
+
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static INLINE void
+trace_context_bind_depth_stencil_alpha_state(struct pipe_context *_pipe, 
+                                             void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "bind_depth_stencil_alpha_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->bind_depth_stencil_alpha_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_delete_depth_stencil_alpha_state(struct pipe_context *_pipe, 
+                                               void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "delete_depth_stencil_alpha_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->delete_depth_stencil_alpha_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void *
+trace_context_create_fs_state(struct pipe_context *_pipe,
+                              const struct pipe_shader_state *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   void * result;
+
+   trace_dump_call_begin("pipe_context", "create_fs_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(shader_state, state);
+
+   result = pipe->create_fs_state(pipe, state);;
+
+   trace_dump_ret(ptr, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static INLINE void
+trace_context_bind_fs_state(struct pipe_context *_pipe, 
+                            void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "bind_fs_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->bind_fs_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_delete_fs_state(struct pipe_context *_pipe, 
+                              void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "delete_fs_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->delete_fs_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void *
+trace_context_create_vs_state(struct pipe_context *_pipe,
+                              const struct pipe_shader_state *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   void * result;
+
+   trace_dump_call_begin("pipe_context", "create_vs_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(shader_state, state);
+
+   result = pipe->create_vs_state(pipe, state);;
+
+   trace_dump_ret(ptr, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static INLINE void
+trace_context_bind_vs_state(struct pipe_context *_pipe, 
+                            void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "bind_vs_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->bind_vs_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_delete_vs_state(struct pipe_context *_pipe, 
+                              void *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "delete_vs_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, state);
+
+   pipe->delete_vs_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_set_blend_color(struct pipe_context *_pipe,
+                              const struct pipe_blend_color *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "set_blend_color");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(blend_color, state);
+
+   pipe->set_blend_color(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_set_clip_state(struct pipe_context *_pipe,
+                             const struct pipe_clip_state *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "set_clip_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(clip_state, state);
+
+   pipe->set_clip_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_set_constant_buffer(struct pipe_context *_pipe,
+                                  uint shader, uint index,
+                                  const struct pipe_constant_buffer *buffer)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_winsys_user_buffer_update(_pipe->winsys, (struct pipe_buffer *)buffer);
+   
+   trace_dump_call_begin("pipe_context", "set_constant_buffer");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(uint, shader);
+   trace_dump_arg(uint, index);
+   trace_dump_arg(constant_buffer, buffer);
+
+   pipe->set_constant_buffer(pipe, shader, index, buffer);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_set_framebuffer_state(struct pipe_context *_pipe,
+                                    const struct pipe_framebuffer_state *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   struct pipe_framebuffer_state unwrapped_state;
+   unsigned i;
+   
+   /* Unwrap the input state */
+   memcpy(&unwrapped_state, state, sizeof(unwrapped_state));
+   for(i = 0; i < state->nr_cbufs; ++i)
+      unwrapped_state.cbufs[i] = trace_surface_unwrap(tr_ctx, state->cbufs[i]);
+   for(i = state->nr_cbufs; i < PIPE_MAX_COLOR_BUFS; ++i)
+      unwrapped_state.cbufs[i] = NULL;
+   unwrapped_state.zsbuf = trace_surface_unwrap(tr_ctx, state->zsbuf);
+   state = &unwrapped_state;
+   
+   trace_dump_call_begin("pipe_context", "set_framebuffer_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(framebuffer_state, state);
+
+   pipe->set_framebuffer_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_set_polygon_stipple(struct pipe_context *_pipe,
+                                  const struct pipe_poly_stipple *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "set_polygon_stipple");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(poly_stipple, state);
+
+   pipe->set_polygon_stipple(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_set_scissor_state(struct pipe_context *_pipe,
+                                const struct pipe_scissor_state *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "set_scissor_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(scissor_state, state);
+
+   pipe->set_scissor_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_set_viewport_state(struct pipe_context *_pipe,
+                                 const struct pipe_viewport_state *state)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "set_viewport_state");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(viewport_state, state);
+
+   pipe->set_viewport_state(pipe, state);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_set_sampler_textures(struct pipe_context *_pipe,
+                                   unsigned num_textures,
+                                   struct pipe_texture **textures)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS];
+   unsigned i;
+   
+   for(i = 0; i < num_textures; ++i)
+      unwrapped_textures[i] = trace_texture_unwrap(tr_ctx, textures[i]);
+   textures = unwrapped_textures;
+
+   trace_dump_call_begin("pipe_context", "set_sampler_textures");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(uint, num_textures);
+   trace_dump_arg_array(ptr, textures, num_textures);
+
+   pipe->set_sampler_textures(pipe, num_textures, textures);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_set_vertex_buffers(struct pipe_context *_pipe,
+                                 unsigned num_buffers,
+                                 const struct pipe_vertex_buffer *buffers)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   unsigned i;
+
+   for(i = 0; i < num_buffers; ++i)
+      trace_winsys_user_buffer_update(_pipe->winsys, buffers[i].buffer);
+
+   trace_dump_call_begin("pipe_context", "set_vertex_buffers");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(uint, num_buffers);
+   
+   trace_dump_arg_begin("buffers");
+   trace_dump_struct_array(vertex_buffer, buffers, num_buffers);
+   trace_dump_arg_end();
+
+   pipe->set_vertex_buffers(pipe, num_buffers, buffers);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_set_vertex_elements(struct pipe_context *_pipe,
+                                  unsigned num_elements,
+                                  const struct pipe_vertex_element *elements)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "set_vertex_elements");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(uint, num_elements);
+
+   trace_dump_arg_begin("elements");
+   trace_dump_struct_array(vertex_element, elements, num_elements);
+   trace_dump_arg_end();
+
+   pipe->set_vertex_elements(pipe, num_elements, elements);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_surface_copy(struct pipe_context *_pipe,
+                           boolean do_flip,
+                           struct pipe_surface *dest,
+                           unsigned destx, unsigned desty,
+                           struct pipe_surface *src,
+                           unsigned srcx, unsigned srcy,
+                           unsigned width, unsigned height)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   dest = trace_surface_unwrap(tr_ctx, dest);
+   src = trace_surface_unwrap(tr_ctx, src);
+   
+   trace_dump_call_begin("pipe_context", "surface_copy");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(bool, do_flip);
+   trace_dump_arg(ptr, dest);
+   trace_dump_arg(uint, destx);
+   trace_dump_arg(uint, desty);
+   trace_dump_arg(ptr, src);
+   trace_dump_arg(uint, srcx);
+   trace_dump_arg(uint, srcy);
+   trace_dump_arg(uint, width);
+   trace_dump_arg(uint, height);
+
+   pipe->surface_copy(pipe, do_flip, 
+                      dest, destx, desty, 
+                      src, srcx, srcy, width, height);
+   
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_surface_fill(struct pipe_context *_pipe,
+                           struct pipe_surface *dst,
+                           unsigned dstx, unsigned dsty,
+                           unsigned width, unsigned height,
+                           unsigned value)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   dst = trace_surface_unwrap(tr_ctx, dst);
+
+   trace_dump_call_begin("pipe_context", "surface_fill");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, dst);
+   trace_dump_arg(uint, dstx);
+   trace_dump_arg(uint, dsty);
+   trace_dump_arg(uint, width);
+   trace_dump_arg(uint, height);
+
+   pipe->surface_fill(pipe, dst, dstx, dsty, width, height, value);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_clear(struct pipe_context *_pipe, 
+                    struct pipe_surface *surface,
+                    unsigned clearValue)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   surface = trace_surface_unwrap(tr_ctx, surface);
+
+   trace_dump_call_begin("pipe_context", "clear");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(ptr, surface);
+   trace_dump_arg(uint, clearValue);
+
+   pipe->clear(pipe, surface, clearValue);;
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_flush(struct pipe_context *_pipe,
+                    unsigned flags,
+                    struct pipe_fence_handle **fence)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "flush");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(uint, flags);
+
+   pipe->flush(pipe, flags, fence);;
+
+   if(fence)
+      trace_dump_ret(ptr, *fence);
+
+   trace_dump_call_end();
+}
+
+
+static INLINE void
+trace_context_destroy(struct pipe_context *_pipe)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+
+   trace_dump_call_begin("pipe_context", "destroy");
+
+   trace_dump_arg(ptr, pipe);
+
+   pipe->destroy(pipe);
+   
+   trace_dump_call_end();
+
+   FREE(tr_ctx);
+}
+
+
+struct pipe_context *
+trace_context_create(struct pipe_screen *screen, 
+                     struct pipe_context *pipe)
+{
+   struct trace_context *tr_ctx;
+   
+   if(!pipe)
+      goto error1;
+   
+   if(!trace_dump_enabled())
+      goto error1;
+   
+   tr_ctx = CALLOC_STRUCT(trace_context);
+   if(!tr_ctx)
+      goto error1;
+
+   tr_ctx->base.winsys = screen->winsys;
+   tr_ctx->base.screen = screen;
+   tr_ctx->base.destroy = trace_context_destroy;
+   tr_ctx->base.set_edgeflags = trace_context_set_edgeflags;
+   tr_ctx->base.draw_arrays = trace_context_draw_arrays;
+   tr_ctx->base.draw_elements = trace_context_draw_elements;
+   tr_ctx->base.draw_range_elements = trace_context_draw_range_elements;
+   tr_ctx->base.create_query = trace_context_create_query;
+   tr_ctx->base.destroy_query = trace_context_destroy_query;
+   tr_ctx->base.begin_query = trace_context_begin_query;
+   tr_ctx->base.end_query = trace_context_end_query;
+   tr_ctx->base.get_query_result = trace_context_get_query_result;
+   tr_ctx->base.create_blend_state = trace_context_create_blend_state;
+   tr_ctx->base.bind_blend_state = trace_context_bind_blend_state;
+   tr_ctx->base.delete_blend_state = trace_context_delete_blend_state;
+   tr_ctx->base.create_sampler_state = trace_context_create_sampler_state;
+   tr_ctx->base.bind_sampler_states = trace_context_bind_sampler_states;
+   tr_ctx->base.delete_sampler_state = trace_context_delete_sampler_state;
+   tr_ctx->base.create_rasterizer_state = trace_context_create_rasterizer_state;
+   tr_ctx->base.bind_rasterizer_state = trace_context_bind_rasterizer_state;
+   tr_ctx->base.delete_rasterizer_state = trace_context_delete_rasterizer_state;
+   tr_ctx->base.create_depth_stencil_alpha_state = trace_context_create_depth_stencil_alpha_state;
+   tr_ctx->base.bind_depth_stencil_alpha_state = trace_context_bind_depth_stencil_alpha_state;
+   tr_ctx->base.delete_depth_stencil_alpha_state = trace_context_delete_depth_stencil_alpha_state;
+   tr_ctx->base.create_fs_state = trace_context_create_fs_state;
+   tr_ctx->base.bind_fs_state = trace_context_bind_fs_state;
+   tr_ctx->base.delete_fs_state = trace_context_delete_fs_state;
+   tr_ctx->base.create_vs_state = trace_context_create_vs_state;
+   tr_ctx->base.bind_vs_state = trace_context_bind_vs_state;
+   tr_ctx->base.delete_vs_state = trace_context_delete_vs_state;
+   tr_ctx->base.set_blend_color = trace_context_set_blend_color;
+   tr_ctx->base.set_clip_state = trace_context_set_clip_state;
+   tr_ctx->base.set_constant_buffer = trace_context_set_constant_buffer;
+   tr_ctx->base.set_framebuffer_state = trace_context_set_framebuffer_state;
+   tr_ctx->base.set_polygon_stipple = trace_context_set_polygon_stipple;
+   tr_ctx->base.set_scissor_state = trace_context_set_scissor_state;
+   tr_ctx->base.set_viewport_state = trace_context_set_viewport_state;
+   tr_ctx->base.set_sampler_textures = trace_context_set_sampler_textures;
+   tr_ctx->base.set_vertex_buffers = trace_context_set_vertex_buffers;
+   tr_ctx->base.set_vertex_elements = trace_context_set_vertex_elements;
+   tr_ctx->base.surface_copy = trace_context_surface_copy;
+   tr_ctx->base.surface_fill = trace_context_surface_fill;
+   tr_ctx->base.clear = trace_context_clear;
+   tr_ctx->base.flush = trace_context_flush;
+
+   tr_ctx->pipe = pipe;
+   
+   trace_dump_call_begin("", "pipe_context_create");
+   trace_dump_arg_begin("screen");
+   trace_dump_ptr(pipe->screen);
+   trace_dump_arg_end();
+   trace_dump_ret(ptr, pipe);
+   trace_dump_call_end();
+
+   return &tr_ctx->base;
+   
+error1:
+   return pipe;
+}
diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h
new file mode 100644
index 0000000000..7831900ec2
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_context.h
@@ -0,0 +1,68 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef TR_CONTEXT_H_
+#define TR_CONTEXT_H_
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "pipe/p_context.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+   
+struct trace_context
+{
+   struct pipe_context base;
+   
+   struct pipe_context *pipe;
+};
+
+
+static INLINE struct trace_context *
+trace_context(struct pipe_context *pipe)
+{
+   assert(pipe);
+   return (struct trace_context *)pipe;
+}
+
+
+
+struct pipe_context *
+trace_context_create(struct pipe_screen *screen,
+                     struct pipe_context *pipe);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* TR_CONTEXT_H_ */
diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c
new file mode 100644
index 0000000000..a0ead0ded3
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_dump.c
@@ -0,0 +1,404 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+/**
+ * @file
+ * Trace dumping functions.
+ * 
+ * For now we just use standard XML for dumping the trace calls, as this is
+ * simple to write, parse, and visually inspect, but the actual representation 
+ * is abstracted out of this file, so that we can switch to a binary 
+ * representation if/when it becomes justified.
+ * 
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>   
+ */
+
+#include "pipe/p_config.h"
+
+#if defined(PIPE_OS_LINUX)
+#include <stdlib.h>
+#endif
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "util/u_memory.h"
+#include "util/u_string.h"
+#include "util/u_stream.h"
+
+#include "tr_dump.h"
+
+
+static struct util_stream *stream = NULL;
+static unsigned refcount = 0;
+
+
+static INLINE void 
+trace_dump_write(const char *buf, size_t size)
+{
+   if(stream)
+      util_stream_write(stream, buf, size);
+}
+
+
+static INLINE void 
+trace_dump_writes(const char *s)
+{
+   trace_dump_write(s, strlen(s));
+}
+
+
+static INLINE void 
+trace_dump_writef(const char *format, ...)
+{
+   static char buf[1024];
+   unsigned len;
+   va_list ap;
+   va_start(ap, format);
+   len = util_vsnprintf(buf, sizeof(buf), format, ap);
+   va_end(ap);
+   trace_dump_write(buf, len);
+}
+
+
+static INLINE void 
+trace_dump_escape(const char *str) 
+{
+   const unsigned char *p = (const unsigned char *)str;
+   unsigned char c;
+   while((c = *p++) != 0) {
+      if(c == '<')
+         trace_dump_writes("&lt;");
+      else if(c == '>')
+         trace_dump_writes("&gt;");
+      else if(c == '&')
+         trace_dump_writes("&amp;");
+      else if(c == '\'')
+         trace_dump_writes("&apos;");
+      else if(c == '\"')
+         trace_dump_writes("&quot;");
+      else if(c >= 0x20 && c <= 0x7e)
+         trace_dump_writef("%c", c);
+      else
+         trace_dump_writef("&#%u;", c);
+   }
+}
+
+
+static INLINE void 
+trace_dump_indent(unsigned level)
+{
+   unsigned i;
+   for(i = 0; i < level; ++i)
+      trace_dump_writes("\t");
+}
+
+
+static INLINE void 
+trace_dump_newline(void) 
+{
+   trace_dump_writes("\n");
+}
+
+
+static INLINE void 
+trace_dump_tag(const char *name)
+{
+   trace_dump_writes("<");
+   trace_dump_writes(name);
+   trace_dump_writes("/>");
+}
+
+
+static INLINE void 
+trace_dump_tag_begin(const char *name)
+{
+   trace_dump_writes("<");
+   trace_dump_writes(name);
+   trace_dump_writes(">");
+}
+
+static INLINE void 
+trace_dump_tag_begin1(const char *name, 
+                      const char *attr1, const char *value1)
+{
+   trace_dump_writes("<");
+   trace_dump_writes(name);
+   trace_dump_writes(" ");
+   trace_dump_writes(attr1);
+   trace_dump_writes("='");
+   trace_dump_escape(value1);
+   trace_dump_writes("'>");
+}
+
+
+static INLINE void 
+trace_dump_tag_begin2(const char *name, 
+                      const char *attr1, const char *value1,
+                      const char *attr2, const char *value2)
+{
+   trace_dump_writes("<");
+   trace_dump_writes(name);
+   trace_dump_writes(" ");
+   trace_dump_writes(attr1);
+   trace_dump_writes("=\'");
+   trace_dump_escape(value1);
+   trace_dump_writes("\' ");
+   trace_dump_writes(attr2);
+   trace_dump_writes("=\'");
+   trace_dump_escape(value2);
+   trace_dump_writes("\'>");
+}
+
+
+static INLINE void 
+trace_dump_tag_begin3(const char *name, 
+                      const char *attr1, const char *value1,
+                      const char *attr2, const char *value2,
+                      const char *attr3, const char *value3)
+{
+   trace_dump_writes("<");
+   trace_dump_writes(name);
+   trace_dump_writes(" ");
+   trace_dump_writes(attr1);
+   trace_dump_writes("=\'");
+   trace_dump_escape(value1);
+   trace_dump_writes("\' ");
+   trace_dump_writes(attr2);
+   trace_dump_writes("=\'");
+   trace_dump_escape(value2);
+   trace_dump_writes("\' ");
+   trace_dump_writes(attr3);
+   trace_dump_writes("=\'");
+   trace_dump_escape(value3);
+   trace_dump_writes("\'>");
+}
+
+
+static INLINE void
+trace_dump_tag_end(const char *name)
+{
+   trace_dump_writes("</");
+   trace_dump_writes(name);
+   trace_dump_writes(">");
+}
+
+static void 
+trace_dump_trace_close(void)
+{
+   if(stream) {
+      trace_dump_writes("</trace>\n");
+      util_stream_close(stream);
+      stream = NULL;
+      refcount = 0;
+   }
+}
+
+boolean trace_dump_trace_begin()
+{
+   const char *filename;
+   
+   filename = debug_get_option("GALLIUM_TRACE", NULL);
+   if(!filename)
+      return FALSE;
+   
+   if(!stream) {
+   
+      stream = util_stream_create(filename, 0);
+      if(!stream)
+         return FALSE;
+      
+      trace_dump_writes("<?xml version='1.0' encoding='UTF-8'?>\n");
+      trace_dump_writes("<?xml-stylesheet type='text/xsl' href='trace.xsl'?>\n");
+      trace_dump_writes("<trace version='0.1'>\n");
+      
+#if defined(PIPE_OS_LINUX)
+      /* Linux applications rarely cleanup GL / Gallium resources so catch 
+       * application exit here */ 
+      atexit(trace_dump_trace_close);
+#endif
+   }
+   
+   ++refcount;
+   
+   return TRUE;
+}
+
+boolean trace_dump_enabled(void)
+{
+   return stream ? TRUE : FALSE;
+}
+
+void trace_dump_trace_end(void)
+{
+   if(stream)
+      if(!--refcount)
+         trace_dump_trace_close();
+}
+
+void trace_dump_call_begin(const char *klass, const char *method)
+{
+   trace_dump_indent(1);
+   trace_dump_tag_begin2("call", "class", klass, "method", method);
+   trace_dump_newline();
+}
+
+void trace_dump_call_end(void)
+{
+   trace_dump_indent(1);
+   trace_dump_tag_end("call");
+   trace_dump_newline();
+   util_stream_flush(stream);
+}
+
+void trace_dump_arg_begin(const char *name)
+{
+   trace_dump_indent(2);
+   trace_dump_tag_begin1("arg", "name", name);
+}
+
+void trace_dump_arg_end(void)
+{
+   trace_dump_tag_end("arg");
+   trace_dump_newline();
+}
+
+void trace_dump_ret_begin(void)
+{
+   trace_dump_indent(2);
+   trace_dump_tag_begin("ret");
+}
+
+void trace_dump_ret_end(void)
+{
+   trace_dump_tag_end("ret");
+   trace_dump_newline();
+}
+
+void trace_dump_bool(int value)
+{
+   trace_dump_writef("<bool>%c</bool>", value ? '1' : '0');
+}
+
+void trace_dump_int(long long int value)
+{
+   trace_dump_writef("<int>%lli</int>", value);
+}
+
+void trace_dump_uint(long long unsigned value)
+{
+   trace_dump_writef("<uint>%llu</uint>", value);
+}
+
+void trace_dump_float(double value)
+{
+   trace_dump_writef("<float>%g</float>", value);
+}
+
+void trace_dump_bytes(const void *data,
+                      long unsigned size)
+{
+   static const char hex_table[16] = "0123456789ABCDEF";
+   const uint8_t *p = data;
+   long unsigned i;
+   trace_dump_writes("<bytes>");
+   for(i = 0; i < size; ++i) {
+      uint8_t byte = *p++;
+      char hex[2];
+      hex[0] = hex_table[byte >> 4];
+      hex[1] = hex_table[byte & 0xf];
+      trace_dump_write(hex, 2);
+   }
+   trace_dump_writes("</bytes>");
+}
+
+void trace_dump_string(const char *str)
+{
+   trace_dump_writes("<string>");
+   trace_dump_escape(str);
+   trace_dump_writes("</string>");
+}
+
+void trace_dump_enum(const char *value)
+{
+   trace_dump_writes("<enum>");
+   trace_dump_escape(value);
+   trace_dump_writes("</enum>");
+}
+
+void trace_dump_array_begin(void)
+{
+   trace_dump_writes("<array>");
+}
+
+void trace_dump_array_end(void)
+{
+   trace_dump_writes("</array>");
+}
+
+void trace_dump_elem_begin(void)
+{
+   trace_dump_writes("<elem>");
+}
+
+void trace_dump_elem_end(void)
+{
+   trace_dump_writes("</elem>");
+}
+
+void trace_dump_struct_begin(const char *name)
+{
+   trace_dump_writef("<struct name='%s'>", name);
+}
+
+void trace_dump_struct_end(void)
+{
+   trace_dump_writes("</struct>");
+}
+
+void trace_dump_member_begin(const char *name)
+{
+   trace_dump_writef("<member name='%s'>", name);
+}
+
+void trace_dump_member_end(void)
+{
+   trace_dump_writes("</member>");
+}
+
+void trace_dump_null(void)
+{
+   trace_dump_writes("<null/>");
+}
+
+void trace_dump_ptr(const void *value)
+{
+   if(value)
+      trace_dump_writef("<ptr>0x%08lx</ptr>", (unsigned long)(uintptr_t)value);
+   else
+      trace_dump_null();
+}
diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h
new file mode 100644
index 0000000000..76a53731b3
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_dump.h
@@ -0,0 +1,132 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Trace data dumping primitives.
+ */
+
+#ifndef TR_DUMP_H
+#define TR_DUMP_H
+
+
+#include "pipe/p_compiler.h"
+
+
+boolean trace_dump_trace_begin(void);
+boolean trace_dump_enabled(void);
+void trace_dump_trace_end(void);
+void trace_dump_call_begin(const char *klass, const char *method);
+void trace_dump_call_end(void);
+void trace_dump_arg_begin(const char *name);
+void trace_dump_arg_end(void);
+void trace_dump_ret_begin(void);
+void trace_dump_ret_end(void);
+void trace_dump_bool(int value);
+void trace_dump_int(long long int value);
+void trace_dump_uint(long long unsigned value);
+void trace_dump_float(double value);
+void trace_dump_bytes(const void *data, long unsigned size);
+void trace_dump_string(const char *str);
+void trace_dump_enum(const char *value);
+void trace_dump_array_begin(void);
+void trace_dump_array_end(void);
+void trace_dump_elem_begin(void);
+void trace_dump_elem_end(void);
+void trace_dump_struct_begin(const char *name);
+void trace_dump_struct_end(void);
+void trace_dump_member_begin(const char *name);
+void trace_dump_member_end(void);
+void trace_dump_null(void);
+void trace_dump_ptr(const void *value);
+
+
+/*
+ * Code saving macros. 
+ */
+
+#define trace_dump_arg(_type, _arg) \
+   do { \
+      trace_dump_arg_begin(#_arg); \
+      trace_dump_##_type(_arg); \
+      trace_dump_arg_end(); \
+   } while(0)
+
+#define trace_dump_ret(_type, _arg) \
+   do { \
+      trace_dump_ret_begin(); \
+      trace_dump_##_type(_arg); \
+      trace_dump_ret_end(); \
+   } while(0)
+
+#define trace_dump_array(_type, _obj, _size) \
+   do { \
+      unsigned long idx; \
+      trace_dump_array_begin(); \
+      for(idx = 0; idx < (_size); ++idx) { \
+         trace_dump_elem_begin(); \
+         trace_dump_##_type((_obj)[idx]); \
+         trace_dump_elem_end(); \
+      } \
+      trace_dump_array_end(); \
+   } while(0)
+
+#define trace_dump_struct_array(_type, _obj, _size) \
+   do { \
+      unsigned long idx; \
+      trace_dump_array_begin(); \
+      for(idx = 0; idx < (_size); ++idx) { \
+         trace_dump_elem_begin(); \
+         trace_dump_##_type(&(_obj)[idx]); \
+         trace_dump_elem_end(); \
+      } \
+      trace_dump_array_end(); \
+   } while(0)
+
+#define trace_dump_member(_type, _obj, _member) \
+   do { \
+      trace_dump_member_begin(#_member); \
+      trace_dump_##_type((_obj)->_member); \
+      trace_dump_member_end(); \
+   } while(0)
+
+#define trace_dump_arg_array(_type, _arg, _size) \
+   do { \
+      trace_dump_arg_begin(#_arg); \
+      trace_dump_array(_type, _arg, _size); \
+      trace_dump_arg_end(); \
+   } while(0)
+
+#define trace_dump_member_array(_type, _obj, _member) \
+   do { \
+      trace_dump_member_begin(#_member); \
+      trace_dump_array(_type, (_obj)->_member, sizeof((_obj)->_member)/sizeof((_obj)->_member[0])); \
+      trace_dump_member_end(); \
+   } while(0)
+
+
+#endif /* TR_DUMP_H */
diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c
new file mode 100644
index 0000000000..8789f86b1a
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_screen.c
@@ -0,0 +1,469 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_memory.h"
+
+#include "tr_dump.h"
+#include "tr_state.h"
+#include "tr_winsys.h"
+#include "tr_texture.h"
+#include "tr_screen.h"
+
+
+static const char *
+trace_screen_get_name(struct pipe_screen *_screen)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   const char *result;
+   
+   trace_dump_call_begin("pipe_screen", "get_name");
+   
+   trace_dump_arg(ptr, screen);
+
+   result = screen->get_name(screen);
+   
+   trace_dump_ret(string, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static const char *
+trace_screen_get_vendor(struct pipe_screen *_screen)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   const char *result;
+   
+   trace_dump_call_begin("pipe_screen", "get_vendor");
+   
+   trace_dump_arg(ptr, screen);
+  
+   result = screen->get_vendor(screen);
+   
+   trace_dump_ret(string, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static int 
+trace_screen_get_param(struct pipe_screen *_screen, 
+                       int param)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   int result;
+   
+   trace_dump_call_begin("pipe_screen", "get_param");
+   
+   trace_dump_arg(ptr, screen);
+   trace_dump_arg(int, param);
+
+   result = screen->get_param(screen, param);
+   
+   trace_dump_ret(int, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static float 
+trace_screen_get_paramf(struct pipe_screen *_screen, 
+                        int param)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   float result;
+   
+   trace_dump_call_begin("pipe_screen", "get_paramf");
+   
+   trace_dump_arg(ptr, screen);
+   trace_dump_arg(int, param);
+
+   result = screen->get_paramf(screen, param);
+   
+   trace_dump_ret(float, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static boolean 
+trace_screen_is_format_supported(struct pipe_screen *_screen,
+                                 enum pipe_format format,
+                                 enum pipe_texture_target target,
+                                 unsigned tex_usage, 
+                                 unsigned geom_flags)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   boolean result;
+   
+   trace_dump_call_begin("pipe_screen", "is_format_supported");
+   
+   trace_dump_arg(ptr, screen);
+   trace_dump_arg(format, format);
+   trace_dump_arg(int, target);
+   trace_dump_arg(uint, tex_usage);
+   trace_dump_arg(uint, geom_flags);
+
+   result = screen->is_format_supported(screen, format, target, tex_usage, geom_flags);
+   
+   trace_dump_ret(bool, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static struct pipe_texture *
+trace_screen_texture_create(struct pipe_screen *_screen,
+                            const struct pipe_texture *templat)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   struct pipe_texture *result;
+   
+   trace_dump_call_begin("pipe_screen", "texture_create");
+
+   trace_dump_arg(ptr, screen);
+   trace_dump_arg(template, templat);
+
+   result = screen->texture_create(screen, templat);
+   
+   trace_dump_ret(ptr, result);
+   
+   trace_dump_call_end();
+   
+   result = trace_texture_create(tr_scr, result);
+   
+   return result;
+}
+
+
+static struct pipe_texture *
+trace_screen_texture_blanket(struct pipe_screen *_screen,
+                             const struct pipe_texture *templat,
+                             const unsigned *ppitch,
+                             struct pipe_buffer *buffer)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   unsigned pitch = *ppitch;
+   struct pipe_texture *result;
+
+   trace_dump_call_begin("pipe_screen", "texture_blanket");
+
+   trace_dump_arg(ptr, screen);
+   trace_dump_arg(template, templat);
+   trace_dump_arg(uint, pitch);
+   trace_dump_arg(ptr, buffer);
+
+   result = screen->texture_blanket(screen, templat, ppitch, buffer);
+   
+   trace_dump_ret(ptr, result);
+   
+   trace_dump_call_end();
+   
+   result = trace_texture_create(tr_scr, result);
+   
+   return result;
+}
+
+
+static void 
+trace_screen_texture_release(struct pipe_screen *_screen,
+                             struct pipe_texture **ptexture)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   struct trace_texture *tr_tex;
+   struct pipe_texture *texture;
+   
+   assert(ptexture);
+   if(*ptexture) {
+      tr_tex = trace_texture(tr_scr, *ptexture);
+      texture = tr_tex->texture;
+      assert(texture->screen == screen);
+   }
+   else
+      texture = NULL;
+   
+   if (*ptexture) {
+      if (!--(*ptexture)->refcount) {
+         trace_dump_call_begin("pipe_screen", "texture_destroy");
+         
+         trace_dump_arg(ptr, screen);
+         trace_dump_arg(ptr, texture);
+         
+         trace_texture_destroy(tr_scr, *ptexture);
+         
+         trace_dump_call_end();
+      }
+   
+      *ptexture = NULL;
+   }
+}
+
+
+static struct pipe_surface *
+trace_screen_get_tex_surface(struct pipe_screen *_screen,
+                             struct pipe_texture *texture,
+                             unsigned face, unsigned level,
+                             unsigned zslice,
+                             unsigned usage)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   struct trace_texture *tr_tex;
+   struct pipe_surface *result;
+   
+   assert(texture);
+   tr_tex = trace_texture(tr_scr, texture);
+   texture = tr_tex->texture;
+   assert(texture->screen == screen);
+   
+   trace_dump_call_begin("pipe_screen", "get_tex_surface");
+   
+   trace_dump_arg(ptr, screen);
+   trace_dump_arg(ptr, texture);
+   trace_dump_arg(uint, face);
+   trace_dump_arg(uint, level);
+   trace_dump_arg(uint, zslice);
+   trace_dump_arg(uint, usage);
+
+   result = screen->get_tex_surface(screen, texture, face, level, zslice, usage);
+
+   trace_dump_ret(ptr, result);
+   
+   trace_dump_call_end();
+   
+   result = trace_surface_create(tr_tex, result);
+
+   return result;
+}
+
+
+static void 
+trace_screen_tex_surface_release(struct pipe_screen *_screen,
+                                 struct pipe_surface **psurface)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   struct trace_texture *tr_tex;
+   struct trace_surface *tr_surf;
+   struct pipe_surface *surface;
+   
+   assert(psurface);
+   if(*psurface) {
+      tr_tex = trace_texture(tr_scr, (*psurface)->texture);
+      tr_surf = trace_surface(tr_tex, *psurface);
+      surface = tr_surf->surface;
+   }
+   else
+      surface = NULL;
+   
+   if (*psurface) {
+      if (!--(*psurface)->refcount) {
+         trace_dump_call_begin("pipe_screen", "tex_surface_destroy");
+         
+         trace_dump_arg(ptr, screen);
+         trace_dump_arg(ptr, surface);
+
+         trace_surface_destroy(tr_tex, *psurface);
+
+         trace_dump_call_end();
+      }
+   
+      *psurface = NULL;
+   }
+}
+
+
+static void *
+trace_screen_surface_map(struct pipe_screen *_screen,
+                         struct pipe_surface *surface,
+                         unsigned flags)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   struct trace_texture *tr_tex;
+   struct trace_surface *tr_surf;
+   void *map;
+   
+   tr_tex = trace_texture(tr_scr, surface->texture);
+   tr_surf = trace_surface(tr_tex, surface);
+   surface = tr_surf->surface;
+
+   map = screen->surface_map(screen, surface, flags);
+   if(map) {
+      if(flags & PIPE_BUFFER_USAGE_CPU_WRITE) {
+         assert(!tr_surf->map);
+         tr_surf->map = map;
+      }
+   }
+   
+   return map;
+}
+
+
+static void 
+trace_screen_surface_unmap(struct pipe_screen *_screen,
+                           struct pipe_surface *surface)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   struct trace_texture *tr_tex;
+   struct trace_surface *tr_surf;
+   
+   tr_tex = trace_texture(tr_scr, surface->texture);
+   tr_surf = trace_surface(tr_tex, surface);
+   surface = tr_surf->surface;
+   
+   if(tr_surf->map) {
+      size_t size = surface->nblocksy * surface->stride;
+      
+      trace_dump_call_begin("pipe_winsys", "surface_write");
+      
+      trace_dump_arg(ptr, screen);
+      
+      trace_dump_arg(ptr, surface);
+      
+      trace_dump_arg_begin("data");
+      trace_dump_bytes(tr_surf->map, size);
+      trace_dump_arg_end();
+
+      trace_dump_arg_begin("stride");
+      trace_dump_uint(surface->stride);
+      trace_dump_arg_end();
+
+      trace_dump_arg_begin("size");
+      trace_dump_uint(size);
+      trace_dump_arg_end();
+   
+      trace_dump_call_end();
+
+      tr_surf->map = NULL;
+   }
+
+   screen->surface_unmap(screen, surface);
+}
+
+
+static void
+trace_screen_destroy(struct pipe_screen *_screen)
+{
+   struct trace_screen *tr_scr = trace_screen(_screen);
+   struct pipe_screen *screen = tr_scr->screen;
+   
+   trace_dump_call_begin("pipe_screen", "destroy");
+   
+   trace_dump_arg(ptr, screen);
+
+   screen->destroy(screen);
+   
+   trace_dump_call_end();
+
+   trace_dump_trace_end();
+
+   FREE(tr_scr);
+}
+
+
+struct pipe_screen *
+trace_screen_create(struct pipe_screen *screen)
+{
+   struct trace_screen *tr_scr;
+   struct pipe_winsys *winsys;
+   
+   if(!screen)
+      goto error1;
+
+   if(!trace_dump_trace_begin())
+      goto error1;
+
+   tr_scr = CALLOC_STRUCT(trace_screen);
+   if(!tr_scr)
+      goto error2;
+
+   winsys = trace_winsys_create(screen->winsys);
+   if(!winsys)
+      goto error3;
+   
+   tr_scr->base.winsys = winsys;
+   tr_scr->base.destroy = trace_screen_destroy;
+   tr_scr->base.get_name = trace_screen_get_name;
+   tr_scr->base.get_vendor = trace_screen_get_vendor;
+   tr_scr->base.get_param = trace_screen_get_param;
+   tr_scr->base.get_paramf = trace_screen_get_paramf;
+   tr_scr->base.is_format_supported = trace_screen_is_format_supported;
+   tr_scr->base.texture_create = trace_screen_texture_create;
+   tr_scr->base.texture_blanket = trace_screen_texture_blanket;
+   tr_scr->base.texture_release = trace_screen_texture_release;
+   tr_scr->base.get_tex_surface = trace_screen_get_tex_surface;
+   tr_scr->base.tex_surface_release = trace_screen_tex_surface_release;
+   tr_scr->base.surface_map = trace_screen_surface_map;
+   tr_scr->base.surface_unmap = trace_screen_surface_unmap;
+   
+   tr_scr->screen = screen;
+
+   trace_dump_call_begin("", "pipe_screen_create");
+   trace_dump_arg_begin("winsys");
+   trace_dump_ptr(screen->winsys);
+   trace_dump_arg_end();
+   trace_dump_ret(ptr, screen);
+   trace_dump_call_end();
+
+   return &tr_scr->base;
+
+error3:
+   FREE(tr_scr);
+error2:
+   trace_dump_trace_end();
+error1:
+   return screen;
+}
+
+
+struct trace_screen *
+trace_screen(struct pipe_screen *screen)
+{
+   assert(screen);
+   assert(screen->destroy == trace_screen_destroy);
+   return (struct trace_screen *)screen;
+}
diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h
new file mode 100644
index 0000000000..93fefdb9a5
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_screen.h
@@ -0,0 +1,60 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef TR_SCREEN_H_
+#define TR_SCREEN_H_
+
+
+#include "pipe/p_screen.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+   
+struct trace_screen
+{
+   struct pipe_screen base;
+   
+   struct pipe_screen *screen;
+};
+
+
+struct trace_screen *
+trace_screen(struct pipe_screen *screen);
+
+
+struct pipe_screen *
+trace_screen_create(struct pipe_screen *screen);
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* TR_SCREEN_H_ */
diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c
new file mode 100644
index 0000000000..524f2d6194
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_state.c
@@ -0,0 +1,462 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_compiler.h"
+#include "util/u_memory.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "tr_dump.h"
+#include "tr_state.h"
+
+
+void trace_dump_format(enum pipe_format format)
+{
+   trace_dump_enum(pf_name(format) );
+}
+
+
+void trace_dump_block(const struct pipe_format_block *block)
+{
+   trace_dump_struct_begin("pipe_format_block");
+   trace_dump_member(uint, block, size);
+   trace_dump_member(uint, block, width);
+   trace_dump_member(uint, block, height);
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_template(const struct pipe_texture *templat)
+{
+   if(!templat) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_texture");
+   
+   trace_dump_member(int, templat, target);
+   trace_dump_member(format, templat, format);
+   
+   trace_dump_member_begin("width");
+   trace_dump_array(uint, templat->width, 1);
+   trace_dump_member_end();
+
+   trace_dump_member_begin("height");
+   trace_dump_array(uint, templat->height, 1);
+   trace_dump_member_end();
+
+   trace_dump_member_begin("depth");
+   trace_dump_array(uint, templat->depth, 1);
+   trace_dump_member_end();
+
+   trace_dump_member_begin("block");
+   trace_dump_block(&templat->block);
+   trace_dump_member_end();
+   
+   trace_dump_member(uint, templat, last_level);
+   trace_dump_member(uint, templat, tex_usage);
+   
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_rasterizer_state");
+
+   trace_dump_member(bool, state, flatshade);
+   trace_dump_member(bool, state, light_twoside);
+   trace_dump_member(uint, state, front_winding);
+   trace_dump_member(uint, state, cull_mode);
+   trace_dump_member(uint, state, fill_cw);
+   trace_dump_member(uint, state, fill_ccw);
+   trace_dump_member(bool, state, offset_cw);
+   trace_dump_member(bool, state, offset_ccw);
+   trace_dump_member(bool, state, scissor);
+   trace_dump_member(bool, state, poly_smooth);
+   trace_dump_member(bool, state, poly_stipple_enable);
+   trace_dump_member(bool, state, point_smooth);
+   trace_dump_member(bool, state, point_sprite);
+   trace_dump_member(bool, state, point_size_per_vertex);
+   trace_dump_member(bool, state, multisample);
+   trace_dump_member(bool, state, line_smooth);
+   trace_dump_member(bool, state, line_stipple_enable);
+   trace_dump_member(uint, state, line_stipple_factor);
+   trace_dump_member(uint, state, line_stipple_pattern);
+   trace_dump_member(bool, state, line_last_pixel);
+   trace_dump_member(bool, state, bypass_clipping);
+   trace_dump_member(bool, state, bypass_vs);
+   trace_dump_member(bool, state, origin_lower_left);
+   trace_dump_member(bool, state, flatshade_first);
+   trace_dump_member(bool, state, gl_rasterization_rules);
+
+   trace_dump_member(float, state, line_width);
+   trace_dump_member(float, state, point_size);
+   trace_dump_member(float, state, point_size_min);
+   trace_dump_member(float, state, point_size_max);
+   trace_dump_member(float, state, offset_units);
+   trace_dump_member(float, state, offset_scale);
+   
+   trace_dump_member_array(uint, state, sprite_coord_mode);
+   
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_poly_stipple(const struct pipe_poly_stipple *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_poly_stipple");
+
+   trace_dump_member_begin("stipple");
+   trace_dump_array(uint,
+                    state->stipple, 
+                    Elements(state->stipple));
+   trace_dump_member_end();
+   
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_viewport_state(const struct pipe_viewport_state *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_viewport_state");
+
+   trace_dump_member_array(float, state, scale);
+   trace_dump_member_array(float, state, translate);
+   
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_scissor_state(const struct pipe_scissor_state *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_scissor_state");
+
+   trace_dump_member(uint, state, minx);
+   trace_dump_member(uint, state, miny);
+   trace_dump_member(uint, state, maxx);
+   trace_dump_member(uint, state, maxy);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_clip_state(const struct pipe_clip_state *state)
+{
+   unsigned i;
+   
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_clip_state");
+
+   trace_dump_member_begin("ucp");
+   trace_dump_array_begin();
+   for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) {
+      trace_dump_elem_begin();
+      trace_dump_array(float, state->ucp[i], 4);
+      trace_dump_elem_end();
+   }
+   trace_dump_array_end();
+   trace_dump_member_end();
+
+   trace_dump_member(uint, state, nr);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_constant_buffer(const struct pipe_constant_buffer *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_constant_buffer");
+
+   trace_dump_member(ptr, state, buffer);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_shader_state(const struct pipe_shader_state *state)
+{
+   static char str[8192];
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   tgsi_dump_str(state->tokens, 0, str, sizeof(str));
+   
+   trace_dump_struct_begin("pipe_shader_state");
+
+   trace_dump_member_begin("tokens");
+   trace_dump_string(str);
+   trace_dump_member_end();
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state)
+{
+   unsigned i;
+   
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_depth_stencil_alpha_state");
+
+   trace_dump_member_begin("depth");
+   trace_dump_struct_begin("pipe_depth_state");
+   trace_dump_member(bool, &state->depth, enabled);
+   trace_dump_member(bool, &state->depth, writemask);
+   trace_dump_member(uint, &state->depth, func);
+   trace_dump_member(bool, &state->depth, occlusion_count);
+   trace_dump_struct_end();
+   trace_dump_member_end();
+   
+   trace_dump_member_begin("stencil");
+   trace_dump_array_begin();
+   for(i = 0; i < Elements(state->stencil); ++i) {
+      trace_dump_elem_begin();
+      trace_dump_struct_begin("pipe_stencil_state");
+      trace_dump_member(bool, &state->stencil[i], enabled);
+      trace_dump_member(uint, &state->stencil[i], func);
+      trace_dump_member(uint, &state->stencil[i], fail_op);
+      trace_dump_member(uint, &state->stencil[i], zpass_op);
+      trace_dump_member(uint, &state->stencil[i], zfail_op);
+      trace_dump_member(uint, &state->stencil[i], ref_value);
+      trace_dump_member(uint, &state->stencil[i], valuemask);
+      trace_dump_member(uint, &state->stencil[i], writemask);
+      trace_dump_struct_end();
+      trace_dump_elem_end();
+   }
+   trace_dump_array_end();
+   trace_dump_member_end();
+
+   trace_dump_member_begin("alpha");
+   trace_dump_struct_begin("pipe_alpha_state");
+   trace_dump_member(bool, &state->alpha, enabled);
+   trace_dump_member(uint, &state->alpha, func);
+   trace_dump_member(float, &state->alpha, ref_value);
+   trace_dump_struct_end();
+   trace_dump_member_end();
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_blend_state(const struct pipe_blend_state *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_blend_state");
+
+   trace_dump_member(bool, state, blend_enable);
+
+   trace_dump_member(uint, state, rgb_func);
+   trace_dump_member(uint, state, rgb_src_factor);
+   trace_dump_member(uint, state, rgb_dst_factor);
+
+   trace_dump_member(uint, state, alpha_func);
+   trace_dump_member(uint, state, alpha_src_factor);
+   trace_dump_member(uint, state, alpha_dst_factor);
+
+   trace_dump_member(bool, state, logicop_enable);
+   trace_dump_member(uint, state, logicop_func);
+
+   trace_dump_member(uint, state, colormask);
+   trace_dump_member(bool, state, dither);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_blend_color(const struct pipe_blend_color *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_blend_color");
+
+   trace_dump_member_array(float, state, color);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state)
+{
+   trace_dump_struct_begin("pipe_framebuffer_state");
+
+   trace_dump_member(uint, state, width);
+   trace_dump_member(uint, state, height);
+   trace_dump_member(uint, state, nr_cbufs);
+   trace_dump_member_array(ptr, state, cbufs);
+   trace_dump_member(ptr, state, zsbuf);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_sampler_state(const struct pipe_sampler_state *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_sampler_state");
+
+   trace_dump_member(uint, state, wrap_s);
+   trace_dump_member(uint, state, wrap_t);
+   trace_dump_member(uint, state, wrap_r);
+   trace_dump_member(uint, state, min_img_filter);
+   trace_dump_member(uint, state, min_mip_filter);
+   trace_dump_member(uint, state, mag_img_filter);
+   trace_dump_member(bool, state, compare_mode);
+   trace_dump_member(uint, state, compare_func);
+   trace_dump_member(bool, state, normalized_coords);
+   trace_dump_member(uint, state, prefilter);
+   trace_dump_member(float, state, shadow_ambient);
+   trace_dump_member(float, state, lod_bias);
+   trace_dump_member(float, state, min_lod);
+   trace_dump_member(float, state, max_lod);
+   trace_dump_member_array(float, state, border_color);
+   trace_dump_member(float, state, max_anisotropy);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_surface(const struct pipe_surface *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_surface");
+
+   trace_dump_member(format, state, format);
+   trace_dump_member(uint, state, status);
+   trace_dump_member(uint, state, clear_value);
+   trace_dump_member(uint, state, width);
+   trace_dump_member(uint, state, height);
+
+   trace_dump_member_begin("block");
+   trace_dump_block(&state->block);
+   trace_dump_member_end();
+   
+   trace_dump_member(uint, state, nblocksx);
+   trace_dump_member(uint, state, nblocksy);
+   trace_dump_member(uint, state, stride);
+   trace_dump_member(uint, state, layout);
+   trace_dump_member(uint, state, offset);
+   trace_dump_member(uint, state, refcount);
+   trace_dump_member(uint, state, usage);
+
+   trace_dump_member(ptr, state, texture);
+   trace_dump_member(uint, state, face);
+   trace_dump_member(uint, state, level);
+   trace_dump_member(uint, state, zslice);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_vertex_buffer");
+
+   trace_dump_member(uint, state, stride);
+   trace_dump_member(uint, state, max_index);
+   trace_dump_member(uint, state, buffer_offset);
+   trace_dump_member(ptr, state, buffer);
+
+   trace_dump_struct_end();
+}
+
+
+void trace_dump_vertex_element(const struct pipe_vertex_element *state)
+{
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_vertex_element");
+
+   trace_dump_member(uint, state, src_offset);
+
+   trace_dump_member(uint, state, vertex_buffer_index);
+   trace_dump_member(uint, state, nr_components);
+ 
+   trace_dump_member(format, state, src_format);
+
+   trace_dump_struct_end();
+}
diff --git a/src/gallium/drivers/trace/tr_state.h b/src/gallium/drivers/trace/tr_state.h
new file mode 100644
index 0000000000..5ae533dc66
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_state.h
@@ -0,0 +1,76 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef TR_STATE_H
+#define TR_STATE_H
+
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+
+
+void trace_dump_format(enum pipe_format format);
+
+void trace_dump_block(const struct pipe_format_block *block);
+
+void trace_dump_template(const struct pipe_texture *templat);
+
+
+void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state);
+
+void trace_dump_poly_stipple(const struct pipe_poly_stipple *state);
+
+void trace_dump_viewport_state(const struct pipe_viewport_state *state);
+
+void trace_dump_scissor_state(const struct pipe_scissor_state *state);
+
+void trace_dump_clip_state(const struct pipe_clip_state *state);
+
+void trace_dump_constant_buffer(const struct pipe_constant_buffer *state);
+
+void trace_dump_token(const struct tgsi_token *token);
+
+void trace_dump_shader_state(const struct pipe_shader_state *state);
+
+void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state);
+
+void trace_dump_blend_state(const struct pipe_blend_state *state);
+
+void trace_dump_blend_color(const struct pipe_blend_color *state);
+
+void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state);
+
+void trace_dump_sampler_state(const struct pipe_sampler_state *state);
+
+void trace_dump_surface(const struct pipe_surface *state);
+
+void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state);
+
+void trace_dump_vertex_element(const struct pipe_vertex_element *state);
+
+
+#endif /* TR_STATE_H */
diff --git a/src/gallium/drivers/trace/tr_texture.c b/src/gallium/drivers/trace/tr_texture.c
new file mode 100644
index 0000000000..1cc4f0bd43
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_texture.c
@@ -0,0 +1,111 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "pipe/p_inlines.h"
+#include "util/u_hash_table.h"
+#include "util/u_memory.h"
+
+#include "tr_screen.h"
+#include "tr_texture.h"
+
+
+struct pipe_texture *
+trace_texture_create(struct trace_screen *tr_scr, 
+                     struct pipe_texture *texture)
+{
+   struct trace_texture *tr_tex;
+   
+   if(!texture)
+      goto error;
+   
+   assert(texture->screen == tr_scr->screen);
+   
+   tr_tex = CALLOC_STRUCT(trace_texture);
+   if(!tr_tex)
+      goto error;
+   
+   memcpy(&tr_tex->base, texture, sizeof(struct pipe_texture));
+   tr_tex->base.screen = &tr_scr->base;
+   tr_tex->texture = texture;
+   
+   return &tr_tex->base;
+   
+error:
+   pipe_texture_reference(&texture, NULL);
+   return NULL;
+}
+
+
+void
+trace_texture_destroy(struct trace_screen *tr_scr, 
+                      struct pipe_texture *texture)
+{
+   struct trace_texture *tr_tex = trace_texture(tr_scr, texture); 
+   pipe_texture_reference(&tr_tex->texture, NULL);
+   FREE(tr_tex);
+}
+
+
+struct pipe_surface *
+trace_surface_create(struct trace_texture *tr_tex, 
+                     struct pipe_surface *surface)
+{
+   struct trace_surface *tr_surf;
+   
+   if(!surface)
+      goto error;
+   
+   assert(surface->texture == tr_tex->texture);
+   
+   tr_surf = CALLOC_STRUCT(trace_surface);
+   if(!tr_surf)
+      goto error;
+   
+   memcpy(&tr_surf->base, surface, sizeof(struct pipe_surface));
+   
+   tr_surf->base.texture = NULL;
+   pipe_texture_reference(&tr_surf->base.texture, &tr_tex->base);
+   tr_surf->surface = surface;
+
+   return &tr_surf->base;
+   
+error:
+   pipe_surface_reference(&surface, NULL);
+   return NULL;
+}
+
+
+void
+trace_surface_destroy(struct trace_texture *tr_tex, 
+                      struct pipe_surface *surface)
+{
+   struct trace_surface *tr_surf = trace_surface(tr_tex, surface);
+   pipe_texture_reference(&tr_surf->base.texture, NULL);
+   pipe_surface_reference(&tr_surf->surface, NULL);
+   FREE(tr_surf);
+}
+
diff --git a/src/gallium/drivers/trace/tr_texture.h b/src/gallium/drivers/trace/tr_texture.h
new file mode 100644
index 0000000000..9e72edb8a3
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_texture.h
@@ -0,0 +1,95 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef TR_TEXTURE_H_
+#define TR_TEXTURE_H_
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
+
+#include "tr_screen.h"
+
+
+struct trace_texture
+{
+   struct pipe_texture base;
+
+   struct pipe_texture *texture;
+};
+
+
+struct trace_surface
+{
+   struct pipe_surface base;
+
+   struct pipe_surface *surface;
+   
+   void *map;
+};
+
+
+static INLINE struct trace_texture *
+trace_texture(struct trace_screen *tr_scr, 
+              struct pipe_texture *texture)
+{
+   if(!texture)
+      return NULL;
+   assert(texture->screen == &tr_scr->base);
+   return (struct trace_texture *)texture;
+}
+
+
+static INLINE struct trace_surface *
+trace_surface(struct trace_texture *tr_tex, 
+              struct pipe_surface *surface)
+{
+   if(!surface)
+      return NULL;
+   assert(surface->texture == &tr_tex->base);
+   return (struct trace_surface *)surface;
+}
+
+
+struct pipe_texture *
+trace_texture_create(struct trace_screen *tr_scr, 
+                     struct pipe_texture *texture);
+
+void
+trace_texture_destroy(struct trace_screen *tr_scr, 
+                      struct pipe_texture *texture);
+
+struct pipe_surface *
+trace_surface_create(struct trace_texture *tr_tex, 
+                     struct pipe_surface *surface);
+
+void
+trace_surface_destroy(struct trace_texture *tr_tex,
+                      struct pipe_surface *surface);
+
+
+#endif /* TR_TEXTURE_H_ */
diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c
new file mode 100644
index 0000000000..c4148fe810
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_winsys.c
@@ -0,0 +1,450 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_memory.h"
+#include "util/u_hash_table.h"
+
+#include "tr_dump.h"
+#include "tr_state.h"
+#include "tr_screen.h"
+#include "tr_texture.h"
+#include "tr_winsys.h"
+
+
+static unsigned trace_buffer_hash(void *buffer)
+{
+   return (unsigned)(uintptr_t)buffer;
+}
+
+
+static int trace_buffer_compare(void *buffer1, void *buffer2)
+{
+   return (char *)buffer2 - (char *)buffer1;
+}
+
+                  
+static const char *
+trace_winsys_get_name(struct pipe_winsys *_winsys)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   const char *result;
+   
+   trace_dump_call_begin("pipe_winsys", "get_name");
+   
+   trace_dump_arg(ptr, winsys);
+
+   result = winsys->get_name(winsys);
+   
+   trace_dump_ret(string, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static void 
+trace_winsys_flush_frontbuffer(struct pipe_winsys *_winsys,
+                               struct pipe_surface *surface,
+                               void *context_private)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+
+   assert(surface);
+   if(surface->texture) {
+      struct trace_screen *tr_scr = trace_screen(surface->texture->screen);
+      struct trace_texture *tr_tex = trace_texture(tr_scr, surface->texture);
+      struct trace_surface *tr_surf = trace_surface(tr_tex, surface);
+      surface = tr_surf->surface;
+   }
+   
+   trace_dump_call_begin("pipe_winsys", "flush_frontbuffer");
+   
+   trace_dump_arg(ptr, winsys);
+   trace_dump_arg(ptr, surface);
+   /* XXX: hide, as there is nothing we can do with this
+   trace_dump_arg(ptr, context_private);
+   */
+
+   winsys->flush_frontbuffer(winsys, surface, context_private);
+   
+   trace_dump_call_end();
+}
+
+
+static struct pipe_buffer *
+trace_winsys_surface_buffer_create(struct pipe_winsys *_winsys,
+                                   unsigned width, unsigned height,
+                                   enum pipe_format format,
+                                   unsigned usage,
+                                   unsigned *pstride)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   unsigned stride;
+   struct pipe_buffer *result;
+   
+   trace_dump_call_begin("pipe_winsys", "surface_buffer_create");
+   
+   trace_dump_arg(ptr, winsys);
+   trace_dump_arg(uint, width);
+   trace_dump_arg(uint, height);
+   trace_dump_arg(format, format);
+   trace_dump_arg(uint, usage);
+
+   result = winsys->surface_buffer_create(winsys,
+                                          width, height,
+                                          format,
+                                          usage,
+                                          pstride);
+   
+   stride = *pstride;
+   
+   trace_dump_arg(uint, stride);
+   
+   trace_dump_ret(ptr, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static struct pipe_buffer *
+trace_winsys_buffer_create(struct pipe_winsys *_winsys, 
+                           unsigned alignment, 
+                           unsigned usage,
+                           unsigned size)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   struct pipe_buffer *buffer;
+   
+   trace_dump_call_begin("pipe_winsys", "buffer_create");
+   
+   trace_dump_arg(ptr, winsys);
+   trace_dump_arg(uint, alignment);
+   trace_dump_arg(uint, usage);
+   trace_dump_arg(uint, size);
+
+   buffer = winsys->buffer_create(winsys, alignment, usage, size);
+   
+   trace_dump_ret(ptr, buffer);
+   
+   trace_dump_call_end();
+
+   /* Zero the buffer to avoid dumping uninitialized memory */
+   if(buffer->usage & PIPE_BUFFER_USAGE_CPU_WRITE) {
+      void *map;
+      map = winsys->buffer_map(winsys, buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+      if(map) {
+         memset(map, 0, buffer->size);
+         winsys->buffer_unmap(winsys, buffer);
+      }
+   }
+   
+   return buffer;
+}
+
+
+static struct pipe_buffer *
+trace_winsys_user_buffer_create(struct pipe_winsys *_winsys, 
+                                void *data,
+                                unsigned size)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   struct pipe_buffer *result;
+   
+   trace_dump_call_begin("pipe_winsys", "user_buffer_create");
+   
+   trace_dump_arg(ptr, winsys);
+   trace_dump_arg_begin("data");
+   trace_dump_bytes(data, size);
+   trace_dump_arg_end();
+   trace_dump_arg(uint, size);
+
+   result = winsys->user_buffer_create(winsys, data, size);
+   
+   trace_dump_ret(ptr, result);
+   
+   trace_dump_call_end();
+   
+   /* XXX: Mark the user buffers. (we should wrap pipe_buffers, but is is 
+    * impossible to do so while texture-less surfaces are still around */
+   if(result) {
+      assert(!(result->usage & TRACE_BUFFER_USAGE_USER));
+      result->usage |= TRACE_BUFFER_USAGE_USER;
+   }
+   
+   return result;
+}
+
+
+void
+trace_winsys_user_buffer_update(struct pipe_winsys *_winsys, 
+                                struct pipe_buffer *buffer)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   const void *map;
+   
+   if(buffer && buffer->usage & TRACE_BUFFER_USAGE_USER) {
+      map = winsys->buffer_map(winsys, buffer, PIPE_BUFFER_USAGE_CPU_READ);
+      if(map) {
+         trace_dump_call_begin("pipe_winsys", "buffer_write");
+         
+         trace_dump_arg(ptr, winsys);
+         
+         trace_dump_arg(ptr, buffer);
+         
+         trace_dump_arg_begin("data");
+         trace_dump_bytes(map, buffer->size);
+         trace_dump_arg_end();
+      
+         trace_dump_arg_begin("size");
+         trace_dump_uint(buffer->size);
+         trace_dump_arg_end();
+      
+         trace_dump_call_end();
+         
+         winsys->buffer_unmap(winsys, buffer);
+      }
+   }
+}
+
+
+static void *
+trace_winsys_buffer_map(struct pipe_winsys *_winsys, 
+                        struct pipe_buffer *buffer,
+                        unsigned usage)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   void *map;
+   
+   map = winsys->buffer_map(winsys, buffer, usage);
+   if(map) {
+      if(usage & PIPE_BUFFER_USAGE_CPU_WRITE) {
+         assert(!hash_table_get(tr_ws->buffer_maps, buffer));
+         hash_table_set(tr_ws->buffer_maps, buffer, map);
+      }
+   }
+   
+   return map;
+}
+
+
+static void
+trace_winsys_buffer_unmap(struct pipe_winsys *_winsys, 
+                          struct pipe_buffer *buffer)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   const void *map;
+   
+   map = hash_table_get(tr_ws->buffer_maps, buffer);
+   if(map) {
+      trace_dump_call_begin("pipe_winsys", "buffer_write");
+      
+      trace_dump_arg(ptr, winsys);
+      
+      trace_dump_arg(ptr, buffer);
+      
+      trace_dump_arg_begin("data");
+      trace_dump_bytes(map, buffer->size);
+      trace_dump_arg_end();
+
+      trace_dump_arg_begin("size");
+      trace_dump_uint(buffer->size);
+      trace_dump_arg_end();
+   
+      trace_dump_call_end();
+
+      hash_table_remove(tr_ws->buffer_maps, buffer);
+   }
+   
+   winsys->buffer_unmap(winsys, buffer);
+}
+
+
+static void
+trace_winsys_buffer_destroy(struct pipe_winsys *_winsys,
+                            struct pipe_buffer *buffer)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   
+   trace_dump_call_begin("pipe_winsys", "buffer_destroy");
+   
+   trace_dump_arg(ptr, winsys);
+   trace_dump_arg(ptr, buffer);
+
+   winsys->buffer_destroy(winsys, buffer);
+   
+   trace_dump_call_end();
+}
+
+
+static void
+trace_winsys_fence_reference(struct pipe_winsys *_winsys,
+                             struct pipe_fence_handle **pdst,
+                             struct pipe_fence_handle *src)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   struct pipe_fence_handle *dst = *pdst;
+   
+   trace_dump_call_begin("pipe_winsys", "fence_reference");
+   
+   trace_dump_arg(ptr, winsys);
+   trace_dump_arg(ptr, dst);
+   trace_dump_arg(ptr, src);
+
+   winsys->fence_reference(winsys, pdst, src);
+   
+   trace_dump_call_end();
+}
+
+
+static int
+trace_winsys_fence_signalled(struct pipe_winsys *_winsys,
+                             struct pipe_fence_handle *fence,
+                             unsigned flag)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   int result;
+   
+   trace_dump_call_begin("pipe_winsys", "fence_signalled");
+   
+   trace_dump_arg(ptr, winsys);
+   trace_dump_arg(ptr, fence);
+   trace_dump_arg(uint, flag);
+
+   result = winsys->fence_signalled(winsys, fence, flag);
+   
+   trace_dump_ret(int, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static int
+trace_winsys_fence_finish(struct pipe_winsys *_winsys,
+                          struct pipe_fence_handle *fence,
+                          unsigned flag)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   int result;
+   
+   trace_dump_call_begin("pipe_winsys", "fence_finish");
+   
+   trace_dump_arg(ptr, winsys);
+   trace_dump_arg(ptr, fence);
+   trace_dump_arg(uint, flag);
+
+   result = winsys->fence_finish(winsys, fence, flag);
+   
+   trace_dump_ret(int, result);
+   
+   trace_dump_call_end();
+   
+   return result;
+}
+
+
+static void
+trace_winsys_destroy(struct pipe_winsys *_winsys)
+{
+   struct trace_winsys *tr_ws = trace_winsys(_winsys);
+   struct pipe_winsys *winsys = tr_ws->winsys;
+   
+   trace_dump_call_begin("pipe_winsys", "destroy");
+   
+   trace_dump_arg(ptr, winsys);
+
+   /* 
+   winsys->destroy(winsys); 
+   */
+   
+   trace_dump_call_end();
+   
+   hash_table_destroy(tr_ws->buffer_maps);
+
+   FREE(tr_ws);
+}
+
+
+struct pipe_winsys *
+trace_winsys_create(struct pipe_winsys *winsys)
+{
+   struct trace_winsys *tr_ws;
+   
+   if(!winsys)
+      goto error1;
+   
+   tr_ws = CALLOC_STRUCT(trace_winsys);
+   if(!tr_ws)
+      goto error1;
+
+   tr_ws->base.destroy = trace_winsys_destroy;
+   tr_ws->base.get_name = trace_winsys_get_name;
+   tr_ws->base.flush_frontbuffer = trace_winsys_flush_frontbuffer;
+   tr_ws->base.surface_buffer_create = trace_winsys_surface_buffer_create;
+   tr_ws->base.buffer_create = trace_winsys_buffer_create;
+   tr_ws->base.user_buffer_create = trace_winsys_user_buffer_create;
+   tr_ws->base.buffer_map = trace_winsys_buffer_map;
+   tr_ws->base.buffer_unmap = trace_winsys_buffer_unmap;
+   tr_ws->base.buffer_destroy = trace_winsys_buffer_destroy;
+   tr_ws->base.fence_reference = trace_winsys_fence_reference;
+   tr_ws->base.fence_signalled = trace_winsys_fence_signalled;
+   tr_ws->base.fence_finish = trace_winsys_fence_finish;
+   
+   tr_ws->winsys = winsys;
+
+   tr_ws->buffer_maps = hash_table_create(trace_buffer_hash, 
+                                          trace_buffer_compare);
+   if(!tr_ws->buffer_maps)
+      goto error2;
+   
+   trace_dump_call_begin("", "pipe_winsys_create");
+   trace_dump_ret(ptr, winsys);
+   trace_dump_call_end();
+
+   return &tr_ws->base;
+   
+error2:
+   FREE(tr_ws);
+error1:
+   return winsys;
+}
diff --git a/src/gallium/drivers/trace/tr_winsys.h b/src/gallium/drivers/trace/tr_winsys.h
new file mode 100644
index 0000000000..0fd2a40556
--- /dev/null
+++ b/src/gallium/drivers/trace/tr_winsys.h
@@ -0,0 +1,76 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef TR_WINSYS_H_
+#define TR_WINSYS_H_
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_debug.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+
+/**
+ * It often happens that new data is written directly to the user buffers 
+ * without mapping/unmapping. This flag marks user buffers, so that their 
+ * contents can be dumpped before being used by the pipe context.
+ */
+#define TRACE_BUFFER_USAGE_USER  (1 << 31)
+
+
+struct hash_table;
+
+
+struct trace_winsys
+{
+   struct pipe_winsys base;
+   
+   struct pipe_winsys *winsys;
+   
+   struct hash_table *buffer_maps;
+};
+
+
+static INLINE struct trace_winsys *
+trace_winsys(struct pipe_winsys *winsys)
+{
+   assert(winsys);
+   return (struct trace_winsys *)winsys;
+}
+
+
+
+struct pipe_winsys *
+trace_winsys_create(struct pipe_winsys *winsys);
+
+
+void
+trace_winsys_user_buffer_update(struct pipe_winsys *winsys, 
+                                struct pipe_buffer *buffer);
+
+
+#endif /* TR_WINSYS_H_ */
diff --git a/src/gallium/drivers/trace/trace.xsl b/src/gallium/drivers/trace/trace.xsl
new file mode 100644
index 0000000000..9cd621e7ab
--- /dev/null
+++ b/src/gallium/drivers/trace/trace.xsl
@@ -0,0 +1,185 @@
+<?xml version="1.0"?>
+
+<!--
+
+Copyright 2008 Tungsten Graphics, Inc.
+
+This program is free software: you can redistribute it and/or modify it
+under the terms of the GNU Lesser General Public License as published
+by the Free Software Foundation, either version 3 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU Lesser General Public License for more details.
+
+You should have received a copy of the GNU Lesser General Public License
+along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+!-->
+
+<xsl:transform version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
+
+	<xsl:output method="html" />
+
+	<xsl:strip-space elements="*" />
+
+	<xsl:template match="/trace">
+		<html>
+			<head>
+				<title>Gallium Trace</title>
+			</head>
+			<style>
+				body {
+					font-family: verdana, sans-serif;
+					font-size: 11px;
+					font-weight: normal;
+					text-align : left;
+				}
+
+				.fun {
+					font-weight: bold;
+				}
+
+				.var {
+					font-style: italic;
+				}
+
+				.typ {
+					display: none;
+				}
+
+				.lit {
+					color: #0000ff;
+				}
+
+				.ptr {
+					color: #008000;
+				}
+			</style>
+			<body>
+				<ol class="calls">
+					<xsl:apply-templates/>
+				</ol>
+			</body>
+		</html>
+	</xsl:template>
+
+	<xsl:template match="call">
+		<li>
+			<span class="fun">
+				<xsl:value-of select="@class"/>
+				<xsl:text>::</xsl:text>
+				<xsl:value-of select="@method"/>
+			</span>
+			<xsl:text>(</xsl:text>
+			<xsl:apply-templates select="arg"/>
+			<xsl:text>)</xsl:text>
+			<xsl:apply-templates select="ret"/>
+		</li>
+	</xsl:template>
+
+	<xsl:template match="arg|member">
+			<xsl:apply-templates select="@name"/>
+			<xsl:text> = </xsl:text>
+			<xsl:apply-templates />
+			<xsl:if test="position() != last()">
+				<xsl:text>, </xsl:text>
+			</xsl:if>
+	</xsl:template>
+
+	<xsl:template match="ret">
+		<xsl:text> = </xsl:text>
+		<xsl:apply-templates />
+	</xsl:template>
+
+	<xsl:template match="bool|int|uint|float|enum">
+		<span class="lit">
+			<xsl:value-of select="text()"/>
+		</span>
+	</xsl:template>
+
+	<xsl:template match="bytes">
+		<span class="lit">
+			<xsl:text>...</xsl:text>
+		</span>
+	</xsl:template>
+
+	<xsl:template match="string">
+		<span class="lit">
+			<xsl:text>"</xsl:text>
+			<xsl:call-template name="break">
+				<xsl:with-param name="text" select="text()"/>
+			</xsl:call-template>
+			<xsl:text>"</xsl:text>
+		</span>
+	</xsl:template>
+
+	<xsl:template match="array|struct">
+		<xsl:text>{</xsl:text>
+		<xsl:apply-templates />
+		<xsl:text>}</xsl:text>
+	</xsl:template>
+
+	<xsl:template match="elem">
+		<xsl:apply-templates />
+		<xsl:if test="position() != last()">
+			<xsl:text>, </xsl:text>
+		</xsl:if>
+	</xsl:template>
+
+	<xsl:template match="null">
+		<span class="ptr">
+			<xsl:text>NULL</xsl:text>
+		</span>
+	</xsl:template>
+
+	<xsl:template match="ptr">
+		<span class="ptr">
+			<xsl:value-of select="text()"/>
+		</span>
+	</xsl:template>
+
+	<xsl:template match="@name">
+		<span class="var">
+			<xsl:value-of select="."/>
+		</span>
+	</xsl:template>
+	
+	<xsl:template name="break">
+		<xsl:param name="text" select="."/>
+		<xsl:choose>
+			<xsl:when test="contains($text, '&#xa;')">
+				<xsl:value-of select="substring-before($text, '&#xa;')"/>
+				<br/>
+				<xsl:call-template name="break">
+					 <xsl:with-param name="text" select="substring-after($text, '&#xa;')"/>
+				</xsl:call-template>
+			</xsl:when>
+			<xsl:otherwise>
+				<xsl:value-of select="$text"/>
+			</xsl:otherwise>
+		</xsl:choose>
+	</xsl:template>
+
+	<xsl:template name="replace">
+		<xsl:param name="text"/>
+		<xsl:param name="from"/>
+		<xsl:param name="to"/>
+		<xsl:choose>
+			<xsl:when test="contains($text,$from)">
+				<xsl:value-of select="concat(substring-before($text,$from),$to)"/>
+				<xsl:call-template name="replace">
+					<xsl:with-param name="text" select="substring-after($text,$from)"/>
+					<xsl:with-param name="from" select="$from"/>
+					<xsl:with-param name="to" select="$to"/>
+				</xsl:call-template>
+			</xsl:when>
+			<xsl:otherwise>
+				<xsl:value-of select="$text"/>
+			</xsl:otherwise>
+		</xsl:choose>
+	</xsl:template>
+
+</xsl:transform>
diff --git a/src/gallium/include/pipe/internal/p_winsys_screen.h b/src/gallium/include/pipe/internal/p_winsys_screen.h
new file mode 100644
index 0000000000..ee835578b2
--- /dev/null
+++ b/src/gallium/include/pipe/internal/p_winsys_screen.h
@@ -0,0 +1,185 @@
+ /**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * \file
+ * This is the interface that Gallium3D requires any window system
+ * hosting it to implement.  This is the only include file in Gallium3D
+ * which is public.
+ */
+
+#ifndef P_WINSYS_H
+#define P_WINSYS_H
+
+
+#include "pipe/p_format.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/** Opaque type */
+struct pipe_fence_handle;
+
+struct pipe_surface;
+
+
+/**
+ * Gallium3D drivers are (meant to be!) independent of both GL and the
+ * window system.  The window system provides a buffer manager and a
+ * set of additional hooks for things like command buffer submission,
+ * etc.
+ *
+ * There clearly has to be some agreement between the window system
+ * driver and the hardware driver about the format of command buffers,
+ * etc.
+ */
+struct pipe_winsys
+{
+   void (*destroy)( struct pipe_winsys *ws );
+
+   /** Returns name of this winsys interface */
+   const char *(*get_name)( struct pipe_winsys *ws );
+
+   /**
+    * Do any special operations to ensure frontbuffer contents are
+    * displayed, eg copy fake frontbuffer.
+    */
+   void (*flush_frontbuffer)( struct pipe_winsys *ws,
+                              struct pipe_surface *surf,
+                              void *context_private );
+
+
+   /**
+    * Buffer management. Buffer attributes are mostly fixed over its lifetime.
+    *
+    * Remember that gallium gets to choose the interface it needs, and the
+    * window systems must then implement that interface (rather than the
+    * other way around...).
+    *
+    * usage is a bitmask of PIPE_BUFFER_USAGE_PIXEL/VERTEX/INDEX/CONSTANT. This
+    * usage argument is only an optimization hint, not a guarantee, therefore 
+    * proper behavior must be observed in all circumstances.
+    *
+    * alignment indicates the client's alignment requirements, eg for
+    * SSE instructions.
+    */
+   struct pipe_buffer *(*buffer_create)( struct pipe_winsys *ws, 
+                                         unsigned alignment, 
+                                         unsigned usage,
+                                         unsigned size );
+
+   /** 
+    * Create a buffer that wraps user-space data.
+    *
+    * Effectively this schedules a delayed call to buffer_create
+    * followed by an upload of the data at *some point in the future*,
+    * or perhaps never.  Basically the allocate/upload is delayed
+    * until the buffer is actually passed to hardware.
+    *
+    * The intention is to provide a quick way to turn regular data
+    * into a buffer, and secondly to avoid a copy operation if that
+    * data subsequently turns out to be only accessed by the CPU.  
+    *
+    * Common example is OpenGL vertex buffers that are subsequently
+    * processed either by software TNL in the driver or by passing to
+    * hardware.
+    *
+    * XXX: What happens if the delayed call to buffer_create() fails?
+    *
+    * Note that ptr may be accessed at any time upto the time when the
+    * buffer is destroyed, so the data must not be freed before then.
+    */
+   struct pipe_buffer *(*user_buffer_create)(struct pipe_winsys *ws, 
+                                                    void *ptr,
+                                                    unsigned bytes);
+
+   /**
+    * Allocate storage for a display target surface.
+    * 
+    * Often surfaces which are meant to be blitted to the front screen (i.e.,
+    * display targets) must be allocated with special characteristics, memory 
+    * pools, or obtained directly from the windowing system.
+    *  
+    * This callback is invoked by the pipe_screenwhen creating a texture marked
+    * with the PIPE_TEXTURE_USAGE_DISPLAY_TARGET flag  to get the underlying 
+    * buffer storage.
+    */
+   struct pipe_buffer *(*surface_buffer_create)(struct pipe_winsys *ws,
+						unsigned width, unsigned height,
+						enum pipe_format format,
+						unsigned usage,
+						unsigned *stride);
+
+
+   /** 
+    * Map the entire data store of a buffer object into the client's address.
+    * flags is bitmask of PIPE_BUFFER_USAGE_CPU_READ/WRITE flags. 
+    */
+   void *(*buffer_map)( struct pipe_winsys *ws, 
+			struct pipe_buffer *buf,
+			unsigned usage );
+   
+   void (*buffer_unmap)( struct pipe_winsys *ws, 
+			 struct pipe_buffer *buf );
+
+   void (*buffer_destroy)( struct pipe_winsys *ws,
+			   struct pipe_buffer *buf );
+
+
+   /** Set ptr = fence, with reference counting */
+   void (*fence_reference)( struct pipe_winsys *ws,
+                            struct pipe_fence_handle **ptr,
+                            struct pipe_fence_handle *fence );
+
+   /**
+    * Checks whether the fence has been signalled.
+    * \param flags  driver-specific meaning
+    * \return zero on success.
+    */
+   int (*fence_signalled)( struct pipe_winsys *ws,
+                           struct pipe_fence_handle *fence,
+                           unsigned flag );
+
+   /**
+    * Wait for the fence to finish.
+    * \param flags  driver-specific meaning
+    * \return zero on success.
+    */
+   int (*fence_finish)( struct pipe_winsys *ws,
+                        struct pipe_fence_handle *fence,
+                        unsigned flag );
+
+};
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* P_WINSYS_H */
diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h
new file mode 100644
index 0000000000..bc2a0a7ef3
--- /dev/null
+++ b/src/gallium/include/pipe/p_compiler.h
@@ -0,0 +1,158 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef P_COMPILER_H
+#define P_COMPILER_H
+
+
+#include "p_config.h"
+
+#ifndef XFree86Server
+#include <stdlib.h>
+#include <string.h>
+#else
+#include "xf86_ansic.h"
+#include "xf86_libc.h"
+#endif
+
+
+#if defined(_WIN32) && !defined(__WIN32__)
+#define __WIN32__
+#endif
+
+#if defined(_MSC_VER)
+
+/* Avoid 'expression is always true' warning */
+#pragma warning(disable: 4296)
+
+#endif /* _MSC_VER */
+
+
+#if defined(_MSC_VER)
+
+typedef __int8             int8_t;
+typedef unsigned __int8    uint8_t;
+typedef __int16            int16_t;
+typedef unsigned __int16   uint16_t;
+#ifndef __eglplatform_h_
+typedef __int32            int32_t;
+#endif
+typedef unsigned __int32   uint32_t;
+typedef __int64            int64_t;
+typedef unsigned __int64   uint64_t;
+
+#if defined(_WIN64)
+typedef __int64            intptr_t;
+typedef unsigned __int64   uintptr_t;
+#else
+typedef __int32            intptr_t;
+typedef unsigned __int32   uintptr_t;
+#endif
+
+#define INT64_C(__val) __val##i64
+#define UINT64_C(__val) __val##ui64
+
+#ifndef __cplusplus
+#define false   0
+#define true    1
+#define bool    _Bool
+typedef int     _Bool;
+#define __bool_true_false_are_defined   1
+#endif /* !__cplusplus */
+
+#else
+#ifndef __STDC_LIMIT_MACROS
+#define __STDC_LIMIT_MACROS 1
+#endif
+#include <stdint.h>
+#include <stdbool.h>
+#endif
+
+
+typedef unsigned int       uint;
+typedef unsigned char      ubyte;
+typedef unsigned short     ushort;
+
+#if 0
+#define boolean bool
+#else
+typedef unsigned char boolean;
+#endif
+#ifndef TRUE
+#define TRUE  true
+#endif
+#ifndef FALSE
+#define FALSE false
+#endif
+
+
+/* Function inlining */
+#ifndef INLINE
+#  ifdef __cplusplus
+#    define INLINE inline
+#  elif defined(__GNUC__)
+#    define INLINE __inline__
+#  elif defined(_MSC_VER)
+#    define INLINE __inline
+#  elif defined(__ICL)
+#    define INLINE __inline
+#  elif defined(__INTEL_COMPILER)
+#    define INLINE inline
+#  elif defined(__WATCOMC__) && (__WATCOMC__ >= 1100)
+#    define INLINE __inline
+#  else
+#    define INLINE
+#  endif
+#endif
+
+
+/* This should match linux gcc cdecl semantics everywhere, so that we
+ * just codegen one calling convention on all platforms.
+ */
+#ifdef _MSC_VER
+#define PIPE_CDECL __cdecl
+#else
+#define PIPE_CDECL
+#endif
+
+
+
+#if defined(__GNUC__)
+#define ALIGN16_DECL(TYPE, NAME, SIZE)  TYPE NAME##___aligned[SIZE] __attribute__(( aligned( 16 ) ))
+#define ALIGN16_ASSIGN(NAME) NAME##___aligned
+#define ALIGN16_ATTRIB  __attribute__(( aligned( 16 ) ))
+#define ALIGN8_ATTRIB  __attribute__(( aligned( 8 ) ))
+#else
+#define ALIGN16_DECL(TYPE, NAME, SIZE)  TYPE NAME##___unaligned[SIZE + 1]
+#define ALIGN16_ASSIGN(NAME) align16(NAME##___unaligned)
+#define ALIGN16_ATTRIB
+#define ALIGN8_ATTRIB
+#endif
+
+
+
+#endif /* P_COMPILER_H */
diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h
new file mode 100644
index 0000000000..05cbd2fc4d
--- /dev/null
+++ b/src/gallium/include/pipe/p_config.h
@@ -0,0 +1,148 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Gallium configuration defines.
+ * 
+ * This header file sets several defines based on the compiler, processor 
+ * architecture, and operating system being used. These defines should be used 
+ * throughout the code to facilitate porting to new platforms. It is likely that 
+ * this file is auto-generated by an autoconf-like tool at some point, as some 
+ * things cannot be determined by pre-defined environment alone. 
+ * 
+ * See also:
+ * - http://gcc.gnu.org/onlinedocs/cpp/Common-Predefined-Macros.html
+ * - echo | gcc -dM -E - | sort
+ * - http://msdn.microsoft.com/en-us/library/b0084kay.aspx
+ * 
+ * @author José Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef P_CONFIG_H_
+#define P_CONFIG_H_
+
+
+/*
+ * Compiler
+ */
+
+#if defined(__GNUC__)
+#define PIPE_CC_GCC
+#endif
+
+/*
+ * Meaning of _MSC_VER value:
+ * - 1400: Visual C++ 2005
+ * - 1310: Visual C++ .NET 2003
+ * - 1300: Visual C++ .NET 2002
+ * 
+ * __MSC__ seems to be an old macro -- it is not pre-defined on recent MSVC 
+ * versions.
+ */
+#if defined(_MSC_VER) || defined(__MSC__)
+#define PIPE_CC_MSVC
+#endif
+
+#if defined(__ICL)
+#define PIPE_CC_ICL
+#endif
+
+
+/*
+ * Processor architecture
+ */
+
+#if defined(__i386__) /* gcc */ || defined(_M_IX86) /* msvc */ || defined(_X86_) || defined(__386__) || defined(i386)
+#define PIPE_ARCH_X86
+#endif
+
+#if defined(__x86_64__) /* gcc */ || defined(_M_X64) /* msvc */ || defined(_M_AMD64) /* msvc */
+#define PIPE_ARCH_X86_64
+#endif
+
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
+#if defined(PIPE_CC_GCC) && !defined(__SSE2__)
+/* #warning SSE2 support requires -msse -msse2 compiler options */
+#else
+#define PIPE_ARCH_SSE
+#endif
+#endif
+
+#if defined(__PPC__)
+#define PIPE_ARCH_PPC
+#if defined(__PPC64__)
+#define PIPE_ARCH_PPC_64
+#endif
+#endif
+
+
+/*
+ * Operating system family.
+ * 
+ * See subsystem below for a more fine-grained distinction.
+ */
+
+#if defined(__linux__)
+#define PIPE_OS_LINUX
+#endif
+
+#if defined(_WIN32) || defined(WIN32)
+#define PIPE_OS_WINDOWS
+#endif
+
+
+/*
+ * Subsystem.
+ * 
+ * NOTE: There is no way to auto-detect most of these.
+ */
+
+#if defined(PIPE_OS_LINUX)
+#define PIPE_SUBSYSTEM_DRI
+#endif /* PIPE_OS_LINUX */
+
+#if defined(PIPE_OS_WINDOWS)
+#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+/* Windows 2000/XP Display Driver */ 
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT)
+/* Windows 2000/XP Miniport Driver */ 
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+/* Windows User-space Library */
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE)
+/* Windows CE 5.0/6.0 */
+#else
+#ifdef _WIN32_WCE
+#define PIPE_SUBSYSTEM_WINDOWS_CE
+#else /* !_WIN32_WCE */
+#error No PIPE_SUBSYSTEM_WINDOWS_xxx subsystem defined. 
+#endif /* !_WIN32_WCE */
+#endif
+#endif /* PIPE_OS_WINDOWS */
+
+
+#endif /* P_CONFIG_H_ */
diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h
new file mode 100644
index 0000000000..9454cc87db
--- /dev/null
+++ b/src/gallium/include/pipe/p_context.h
@@ -0,0 +1,227 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef PIPE_CONTEXT_H
+#define PIPE_CONTEXT_H
+
+#include "p_state.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+   
+struct pipe_screen;
+struct pipe_fence_handle;
+struct pipe_state_cache;
+struct pipe_query;
+struct pipe_winsys;
+
+
+/**
+ * Gallium rendering context.  Basically:
+ *  - state setting functions
+ *  - VBO drawing functions
+ *  - surface functions
+ */
+struct pipe_context {
+   struct pipe_winsys *winsys;
+   struct pipe_screen *screen;
+
+   void *priv;  /**< context private data (for DRI for example) */
+   void *draw;  /**< private, for draw module (temporary?) */
+
+   void (*destroy)( struct pipe_context * );
+
+   
+   /* Possible interface for setting edgeflags.  These aren't really
+    * vertex elements, so don't fit there.
+    */
+   void (*set_edgeflags)( struct pipe_context *,
+                          const unsigned *bitfield );
+
+
+   /**
+    * VBO drawing (return false on fallbacks (temporary??))
+    */
+   /*@{*/
+   boolean (*draw_arrays)( struct pipe_context *pipe,
+			   unsigned mode, unsigned start, unsigned count);
+
+   boolean (*draw_elements)( struct pipe_context *pipe,
+			     struct pipe_buffer *indexBuffer,
+			     unsigned indexSize,
+			     unsigned mode, unsigned start, unsigned count);
+
+   /* XXX: this is (probably) a temporary entrypoint, as the range
+    * information should be available from the vertex_buffer state.
+    * Using this to quickly evaluate a specialized path in the draw
+    * module.
+    */
+   boolean (*draw_range_elements)( struct pipe_context *pipe,
+                                   struct pipe_buffer *indexBuffer,
+                                   unsigned indexSize,
+                                   unsigned minIndex,
+                                   unsigned maxIndex,
+                                   unsigned mode, 
+                                   unsigned start, 
+                                   unsigned count);
+   /*@}*/
+
+
+   /**
+    * Query objects
+    */
+   /*@{*/
+   struct pipe_query *(*create_query)( struct pipe_context *pipe,
+                                              unsigned query_type );
+
+   void (*destroy_query)(struct pipe_context *pipe,
+                         struct pipe_query *q);
+
+   void (*begin_query)(struct pipe_context *pipe, struct pipe_query *q);
+   void (*end_query)(struct pipe_context *pipe, struct pipe_query *q);
+
+   boolean (*get_query_result)(struct pipe_context *pipe, 
+                               struct pipe_query *q,
+                               boolean wait,
+                               uint64_t *result);
+   /*@}*/
+
+   /**
+    * State functions (create/bind/destroy state objects)
+    */
+   /*@{*/
+   void * (*create_blend_state)(struct pipe_context *,
+                                const struct pipe_blend_state *);
+   void   (*bind_blend_state)(struct pipe_context *, void *);
+   void   (*delete_blend_state)(struct pipe_context *, void  *);
+
+   void * (*create_sampler_state)(struct pipe_context *,
+                                  const struct pipe_sampler_state *);
+   void   (*bind_sampler_states)(struct pipe_context *, unsigned num, void **);
+   void   (*delete_sampler_state)(struct pipe_context *, void *);
+
+   void * (*create_rasterizer_state)(struct pipe_context *,
+                                     const struct pipe_rasterizer_state *);
+   void   (*bind_rasterizer_state)(struct pipe_context *, void *);
+   void   (*delete_rasterizer_state)(struct pipe_context *, void *);
+
+   void * (*create_depth_stencil_alpha_state)(struct pipe_context *,
+                                        const struct pipe_depth_stencil_alpha_state *);
+   void   (*bind_depth_stencil_alpha_state)(struct pipe_context *, void *);
+   void   (*delete_depth_stencil_alpha_state)(struct pipe_context *, void *);
+
+   void * (*create_fs_state)(struct pipe_context *,
+                             const struct pipe_shader_state *);
+   void   (*bind_fs_state)(struct pipe_context *, void *);
+   void   (*delete_fs_state)(struct pipe_context *, void *);
+
+   void * (*create_vs_state)(struct pipe_context *,
+                             const struct pipe_shader_state *);
+   void   (*bind_vs_state)(struct pipe_context *, void *);
+   void   (*delete_vs_state)(struct pipe_context *, void *);
+   /*@}*/
+
+   /**
+    * Parameter-like state (or properties)
+    */
+   /*@{*/
+   void (*set_blend_color)( struct pipe_context *,
+                            const struct pipe_blend_color * );
+
+   void (*set_clip_state)( struct pipe_context *,
+			   const struct pipe_clip_state * );
+
+   void (*set_constant_buffer)( struct pipe_context *,
+                                uint shader, uint index,
+                                const struct pipe_constant_buffer *buf );
+
+   void (*set_framebuffer_state)( struct pipe_context *,
+                                  const struct pipe_framebuffer_state * );
+
+   void (*set_polygon_stipple)( struct pipe_context *,
+				const struct pipe_poly_stipple * );
+
+   void (*set_scissor_state)( struct pipe_context *,
+                              const struct pipe_scissor_state * );
+
+   void (*set_viewport_state)( struct pipe_context *,
+                               const struct pipe_viewport_state * );
+
+   void (*set_sampler_textures)( struct pipe_context *,
+                                 unsigned num_textures,
+                                 struct pipe_texture ** );
+
+   void (*set_vertex_buffers)( struct pipe_context *,
+                               unsigned num_buffers,
+                               const struct pipe_vertex_buffer * );
+
+   void (*set_vertex_elements)( struct pipe_context *,
+                                unsigned num_elements,
+                                const struct pipe_vertex_element * );
+   /*@}*/
+
+
+   /**
+    * Surface functions
+    */
+   /*@{*/
+   void (*surface_copy)(struct pipe_context *pipe,
+                        boolean do_flip,/**< flip surface contents vertically */
+			struct pipe_surface *dest,
+			unsigned destx, unsigned desty,
+			struct pipe_surface *src, /* don't make this const - 
+						     need to map/unmap */
+			unsigned srcx, unsigned srcy,
+			unsigned width, unsigned height);
+
+   void (*surface_fill)(struct pipe_context *pipe,
+			struct pipe_surface *dst,
+			unsigned dstx, unsigned dsty,
+			unsigned width, unsigned height,
+			unsigned value);
+
+   void (*clear)(struct pipe_context *pipe, 
+		 struct pipe_surface *ps,
+		 unsigned clearValue);
+   /*@}*/
+
+
+   /** Flush rendering (flags = bitmask of PIPE_FLUSH_x tokens) */
+   void (*flush)( struct pipe_context *pipe,
+                  unsigned flags,
+                  struct pipe_fence_handle **fence );
+};
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* PIPE_CONTEXT_H */
diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h
new file mode 100644
index 0000000000..e9c95982dd
--- /dev/null
+++ b/src/gallium/include/pipe/p_debug.h
@@ -0,0 +1,361 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Cross-platform debugging helpers.
+ * 
+ * For now it just has assert and printf replacements, but it might be extended 
+ * with stack trace reports and more advanced logging in the near future. 
+ * 
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef P_DEBUG_H_
+#define P_DEBUG_H_
+
+
+#include <stdarg.h>
+
+#include "p_compiler.h"
+
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+
+#if defined(DBG) || defined(DEBUG)
+#ifndef DEBUG
+#define DEBUG 1
+#endif
+#else
+#ifndef NDEBUG
+#define NDEBUG 1
+#endif
+#endif
+
+   
+/* MSVC bebore VC7 does not have the __FUNCTION__ macro */
+#if defined(_MSC_VER) && _MSC_VER < 1300
+#define __FUNCTION__ "???"
+#endif
+
+
+void _debug_vprintf(const char *format, va_list ap);
+   
+
+static INLINE void
+_debug_printf(const char *format, ...)
+{
+   va_list ap;
+   va_start(ap, format);
+   _debug_vprintf(format, ap);
+   va_end(ap);
+}
+
+
+/**
+ * Print debug messages.
+ *
+ * The actual channel used to output debug message is platform specific. To 
+ * avoid misformating or truncation, follow these rules of thumb:   
+ * - output whole lines
+ * - avoid outputing large strings (512 bytes is the current maximum length 
+ * that is guaranteed to be printed in all platforms)
+ */
+static INLINE void
+debug_printf(const char *format, ...)
+{
+#ifdef DEBUG
+   va_list ap;
+   va_start(ap, format);
+   _debug_vprintf(format, ap);
+   va_end(ap);
+#else
+   (void) format; /* silence warning */
+#endif
+}
+
+
+#ifdef DEBUG
+#define debug_vprintf(_format, _ap) _debug_vprintf(_format, _ap)
+#else
+#define debug_vprintf(_format, _ap) ((void)0)
+#endif
+
+
+#ifdef DEBUG
+/**
+ * Dump a blob in hex to the same place that debug_printf sends its
+ * messages.
+ */
+void debug_print_blob( const char *name, const void *blob, unsigned size );
+
+/* Print a message along with a prettified format string
+ */
+void debug_print_format(const char *msg, unsigned fmt );
+#else
+#define debug_print_blob(_name, _blob, _size) ((void)0)
+#define debug_print_format(_msg, _fmt) ((void)0)
+#endif
+
+
+void _debug_break(void);
+
+
+/**
+ * Hard-coded breakpoint.
+ */
+#ifdef DEBUG
+#if defined(PIPE_ARCH_X86) && defined(PIPE_CC_GCC)
+#define debug_break() __asm("int3")
+#elif defined(PIPE_ARCH_X86) && defined(PIPE_CC_MSVC)
+#define debug_break()  do { _asm {int 3} } while(0)
+#else
+#define debug_break() _debug_break()
+#endif
+#else /* !DEBUG */
+#define debug_break() ((void)0)
+#endif /* !DEBUG */
+
+
+long
+debug_get_num_option(const char *name, long dfault);
+
+void _debug_assert_fail(const char *expr, 
+                        const char *file, 
+                        unsigned line, 
+                        const char *function);
+
+
+/** 
+ * Assert macro
+ * 
+ * Do not expect that the assert call terminates -- errors must be handled 
+ * regardless of assert behavior.
+ */
+#ifdef DEBUG
+#define debug_assert(expr) ((expr) ? (void)0 : _debug_assert_fail(#expr, __FILE__, __LINE__, __FUNCTION__))
+#else
+#define debug_assert(expr) ((void)0)
+#endif
+
+
+/** Override standard assert macro */
+#ifdef assert
+#undef assert
+#endif
+#define assert(expr) debug_assert(expr)
+
+
+/**
+ * Output the current function name.
+ */
+#ifdef DEBUG
+#define debug_checkpoint() \
+   _debug_printf("%s\n", __FUNCTION__)
+#else
+#define debug_checkpoint() \
+   ((void)0) 
+#endif
+
+
+/**
+ * Output the full source code position.
+ */
+#ifdef DEBUG
+#define debug_checkpoint_full() \
+   _debug_printf("%s:%u:%s", __FILE__, __LINE__, __FUNCTION__) 
+#else
+#define debug_checkpoint_full() \
+   ((void)0) 
+#endif
+
+
+/**
+ * Output a warning message. Muted on release version.
+ */
+#ifdef DEBUG
+#define debug_warning(__msg) \
+   _debug_printf("%s:%u:%s: warning: %s\n", __FILE__, __LINE__, __FUNCTION__, __msg)
+#else
+#define debug_warning(__msg) \
+   ((void)0) 
+#endif
+
+
+/**
+ * Output an error message. Not muted on release version.
+ */
+#ifdef DEBUG
+#define debug_error(__msg) \
+   _debug_printf("%s:%u:%s: error: %s\n", __FILE__, __LINE__, __FUNCTION__, __msg) 
+#else
+#define debug_error(__msg) \
+   _debug_printf("error: %s\n", __msg)
+#endif
+
+
+/**
+ * Used by debug_dump_enum and debug_dump_flags to describe symbols.
+ */
+struct debug_named_value
+{
+   const char *name;
+   unsigned long value;
+};
+
+
+/**
+ * Some C pre-processor magic to simplify creating named values.
+ * 
+ * Example:
+ * @code
+ * static const debug_named_value my_names[] = {
+ *    DEBUG_NAMED_VALUE(MY_ENUM_VALUE_X),
+ *    DEBUG_NAMED_VALUE(MY_ENUM_VALUE_Y),
+ *    DEBUG_NAMED_VALUE(MY_ENUM_VALUE_Z),
+ *    DEBUG_NAMED_VALUE_END
+ * };
+ * 
+ *    ...
+ *    debug_printf("%s = %s\n", 
+ *                 name,
+ *                 debug_dump_enum(my_names, my_value));
+ *    ...
+ * @endcode
+ */
+#define DEBUG_NAMED_VALUE(__symbol) {#__symbol, (unsigned long)__symbol} 
+#define DEBUG_NAMED_VALUE_END {NULL, 0} 
+
+
+/**
+ * Convert a enum value to a string.
+ */
+const char *
+debug_dump_enum(const struct debug_named_value *names, 
+                unsigned long value);
+
+const char *
+debug_dump_enum_noprefix(const struct debug_named_value *names, 
+                         const char *prefix,
+                         unsigned long value);
+
+
+/**
+ * Convert binary flags value to a string.
+ */
+const char *
+debug_dump_flags(const struct debug_named_value *names, 
+                 unsigned long value);
+
+
+/**
+ * Get option.
+ * 
+ * It is an alias for getenv on Linux. 
+ * 
+ * On Windows it reads C:\gallium.cfg, which is a text file with CR+LF line 
+ * endings with one option per line as
+ *  
+ *   NAME=value
+ * 
+ * This file must be terminated with an extra empty line.
+ */
+const char *
+debug_get_option(const char *name, const char *dfault);
+
+boolean
+debug_get_bool_option(const char *name, boolean dfault);
+
+long
+debug_get_num_option(const char *name, long dfault);
+
+unsigned long
+debug_get_flags_option(const char *name, 
+                       const struct debug_named_value *flags,
+                       unsigned long dfault);
+
+
+void *
+debug_malloc(const char *file, unsigned line, const char *function,
+             size_t size);
+
+void
+debug_free(const char *file, unsigned line, const char *function,
+           void *ptr);
+
+void *
+debug_calloc(const char *file, unsigned line, const char *function,
+             size_t count, size_t size );
+
+void *
+debug_realloc(const char *file, unsigned line, const char *function,
+              void *old_ptr, size_t old_size, size_t new_size );
+
+unsigned long
+debug_memory_begin(void);
+
+void 
+debug_memory_end(unsigned long beginning);
+
+
+#if defined(PROFILE) && defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY)
+
+void
+debug_profile_start(void);
+
+void 
+debug_profile_stop(void);
+
+#endif
+
+
+#ifdef DEBUG
+struct pipe_surface;
+void debug_dump_image(const char *prefix,
+                      unsigned format, unsigned cpp,
+                      unsigned width, unsigned height,
+                      unsigned stride,
+                      const void *data);
+void debug_dump_surface(const char *prefix,
+                        struct pipe_surface *surface);   
+void debug_dump_surface_bmp(const char *filename,
+                            struct pipe_surface *surface);
+#else
+#define debug_dump_image(prefix, format, cpp, width, height, stride, data) ((void)0)
+#define debug_dump_surface(prefix, surface) ((void)0)
+#define debug_dump_surface_bmp(filename, surface) ((void)0)
+#endif
+
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif /* P_DEBUG_H_ */
diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
new file mode 100644
index 0000000000..4f0b301f31
--- /dev/null
+++ b/src/gallium/include/pipe/p_defines.h
@@ -0,0 +1,307 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef PIPE_DEFINES_H
+#define PIPE_DEFINES_H
+
+#include "p_format.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define PIPE_BLENDFACTOR_ONE                 0x1
+#define PIPE_BLENDFACTOR_SRC_COLOR           0x2
+#define PIPE_BLENDFACTOR_SRC_ALPHA           0x3
+#define PIPE_BLENDFACTOR_DST_ALPHA           0x4
+#define PIPE_BLENDFACTOR_DST_COLOR           0x5
+#define PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE  0x6
+#define PIPE_BLENDFACTOR_CONST_COLOR         0x7
+#define PIPE_BLENDFACTOR_CONST_ALPHA         0x8
+#define PIPE_BLENDFACTOR_SRC1_COLOR          0x9
+#define PIPE_BLENDFACTOR_SRC1_ALPHA          0x0A
+#define PIPE_BLENDFACTOR_ZERO                0x11
+#define PIPE_BLENDFACTOR_INV_SRC_COLOR       0x12
+#define PIPE_BLENDFACTOR_INV_SRC_ALPHA       0x13
+#define PIPE_BLENDFACTOR_INV_DST_ALPHA       0x14
+#define PIPE_BLENDFACTOR_INV_DST_COLOR       0x15
+#define PIPE_BLENDFACTOR_INV_CONST_COLOR     0x17
+#define PIPE_BLENDFACTOR_INV_CONST_ALPHA     0x18
+#define PIPE_BLENDFACTOR_INV_SRC1_COLOR      0x19
+#define PIPE_BLENDFACTOR_INV_SRC1_ALPHA      0x1A
+
+#define PIPE_BLEND_ADD               0
+#define PIPE_BLEND_SUBTRACT          1
+#define PIPE_BLEND_REVERSE_SUBTRACT  2
+#define PIPE_BLEND_MIN               3
+#define PIPE_BLEND_MAX               4
+
+#define PIPE_LOGICOP_CLEAR            0
+#define PIPE_LOGICOP_NOR              1
+#define PIPE_LOGICOP_AND_INVERTED     2
+#define PIPE_LOGICOP_COPY_INVERTED    3
+#define PIPE_LOGICOP_AND_REVERSE      4
+#define PIPE_LOGICOP_INVERT           5
+#define PIPE_LOGICOP_XOR              6
+#define PIPE_LOGICOP_NAND             7
+#define PIPE_LOGICOP_AND              8
+#define PIPE_LOGICOP_EQUIV            9
+#define PIPE_LOGICOP_NOOP             10
+#define PIPE_LOGICOP_OR_INVERTED      11
+#define PIPE_LOGICOP_COPY             12
+#define PIPE_LOGICOP_OR_REVERSE       13
+#define PIPE_LOGICOP_OR               14
+#define PIPE_LOGICOP_SET              15  
+
+#define PIPE_MASK_R  0x1
+#define PIPE_MASK_G  0x2
+#define PIPE_MASK_B  0x4
+#define PIPE_MASK_A  0x8
+#define PIPE_MASK_RGBA 0xf
+
+
+/**
+ * Inequality functions.  Used for depth test, stencil compare, alpha
+ * test, shadow compare, etc.
+ */
+#define PIPE_FUNC_NEVER    0
+#define PIPE_FUNC_LESS     1
+#define PIPE_FUNC_EQUAL    2
+#define PIPE_FUNC_LEQUAL   3
+#define PIPE_FUNC_GREATER  4
+#define PIPE_FUNC_NOTEQUAL 5
+#define PIPE_FUNC_GEQUAL   6
+#define PIPE_FUNC_ALWAYS   7
+
+/** Polygon fill mode */
+#define PIPE_POLYGON_MODE_FILL  0
+#define PIPE_POLYGON_MODE_LINE  1
+#define PIPE_POLYGON_MODE_POINT 2
+
+/** Polygon front/back window, also for culling */
+#define PIPE_WINDING_NONE 0
+#define PIPE_WINDING_CW   1
+#define PIPE_WINDING_CCW  2
+#define PIPE_WINDING_BOTH (PIPE_WINDING_CW | PIPE_WINDING_CCW)
+
+/** Stencil ops */
+#define PIPE_STENCIL_OP_KEEP       0
+#define PIPE_STENCIL_OP_ZERO       1
+#define PIPE_STENCIL_OP_REPLACE    2
+#define PIPE_STENCIL_OP_INCR       3
+#define PIPE_STENCIL_OP_DECR       4
+#define PIPE_STENCIL_OP_INCR_WRAP  5
+#define PIPE_STENCIL_OP_DECR_WRAP  6
+#define PIPE_STENCIL_OP_INVERT     7
+
+/** Texture types */
+enum pipe_texture_target {
+   PIPE_TEXTURE_1D   = 0,
+   PIPE_TEXTURE_2D   = 1,
+   PIPE_TEXTURE_3D   = 2,
+   PIPE_TEXTURE_CUBE = 3
+};
+
+#define PIPE_TEX_FACE_POS_X 0
+#define PIPE_TEX_FACE_NEG_X 1
+#define PIPE_TEX_FACE_POS_Y 2
+#define PIPE_TEX_FACE_NEG_Y 3
+#define PIPE_TEX_FACE_POS_Z 4
+#define PIPE_TEX_FACE_NEG_Z 5
+
+#define PIPE_TEX_WRAP_REPEAT                   0
+#define PIPE_TEX_WRAP_CLAMP                    1
+#define PIPE_TEX_WRAP_CLAMP_TO_EDGE            2
+#define PIPE_TEX_WRAP_CLAMP_TO_BORDER          3
+#define PIPE_TEX_WRAP_MIRROR_REPEAT            4
+#define PIPE_TEX_WRAP_MIRROR_CLAMP             5
+#define PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE     6
+#define PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER   7
+
+/* Between mipmaps, ie mipfilter
+ */
+#define PIPE_TEX_MIPFILTER_NEAREST  0
+#define PIPE_TEX_MIPFILTER_LINEAR   1
+#define PIPE_TEX_MIPFILTER_NONE     2
+
+/* Within a mipmap, ie min/mag filter 
+ */
+#define PIPE_TEX_FILTER_NEAREST      0
+#define PIPE_TEX_FILTER_LINEAR       1
+#define PIPE_TEX_FILTER_ANISO        2 
+
+
+#define PIPE_TEX_COMPARE_NONE          0
+#define PIPE_TEX_COMPARE_R_TO_TEXTURE  1
+
+#define PIPE_TEX_FACE_POS_X   0
+#define PIPE_TEX_FACE_NEG_X   1
+#define PIPE_TEX_FACE_POS_Y   2
+#define PIPE_TEX_FACE_NEG_Y   3
+#define PIPE_TEX_FACE_POS_Z   4
+#define PIPE_TEX_FACE_NEG_Z   5
+#define PIPE_TEX_FACE_MAX     6
+
+#define PIPE_TEXTURE_USAGE_RENDER_TARGET   0x1
+#define PIPE_TEXTURE_USAGE_DISPLAY_TARGET  0x2 /* ie a backbuffer */
+#define PIPE_TEXTURE_USAGE_PRIMARY         0x4 /* ie a frontbuffer */
+#define PIPE_TEXTURE_USAGE_DEPTH_STENCIL   0x8
+#define PIPE_TEXTURE_USAGE_SAMPLER         0x10
+#define PIPE_TEXTURE_USAGE_DYNAMIC         0x20
+/** Pipe driver custom usage flags should be greater or equal to this value */
+#define PIPE_TEXTURE_USAGE_CUSTOM          (1 << 16)
+
+#define PIPE_TEXTURE_GEOM_NON_SQUARE       0x1
+#define PIPE_TEXTURE_GEOM_NON_POWER_OF_TWO 0x2
+
+
+/**
+ * Surface layout
+ */
+#define PIPE_SURFACE_LAYOUT_LINEAR  0
+
+
+/**
+ * Surface status
+ */
+#define PIPE_SURFACE_STATUS_UNDEFINED  0
+#define PIPE_SURFACE_STATUS_DEFINED    1
+#define PIPE_SURFACE_STATUS_CLEAR      2
+
+
+/**
+ * Buffer usage flags
+ */
+#define PIPE_BUFFER_USAGE_CPU_READ  (1 << 0)
+#define PIPE_BUFFER_USAGE_CPU_WRITE (1 << 1)
+#define PIPE_BUFFER_USAGE_GPU_READ  (1 << 2)
+#define PIPE_BUFFER_USAGE_GPU_WRITE (1 << 3)
+#define PIPE_BUFFER_USAGE_PIXEL     (1 << 4)
+#define PIPE_BUFFER_USAGE_VERTEX    (1 << 5)
+#define PIPE_BUFFER_USAGE_INDEX     (1 << 6)
+#define PIPE_BUFFER_USAGE_CONSTANT  (1 << 7)
+#define PIPE_BUFFER_USAGE_DISCARD   (1 << 8)
+/** Pipe driver custom usage flags should be greater or equal to this value */
+#define PIPE_BUFFER_USAGE_CUSTOM    (1 << 16)
+
+/* Convenient shortcuts */
+#define PIPE_BUFFER_USAGE_CPU_READ_WRITE \
+   ( PIPE_BUFFER_USAGE_CPU_READ | PIPE_BUFFER_USAGE_CPU_WRITE )
+#define PIPE_BUFFER_USAGE_GPU_READ_WRITE \
+   ( PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE )
+#define PIPE_BUFFER_USAGE_WRITE \
+   ( PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_GPU_WRITE )
+
+
+/** 
+ * Flush types:
+ */
+#define PIPE_FLUSH_RENDER_CACHE   0x1
+#define PIPE_FLUSH_TEXTURE_CACHE  0x2
+#define PIPE_FLUSH_SWAPBUFFERS    0x4
+#define PIPE_FLUSH_FRAME          0x8 /**< Mark the end of a frame */
+
+
+/**
+ * Shaders
+ */
+#define PIPE_SHADER_VERTEX   0
+#define PIPE_SHADER_FRAGMENT 1
+#define PIPE_SHADER_TYPES    2
+
+
+/**
+ * Primitive types:
+ */
+#define PIPE_PRIM_POINTS          0
+#define PIPE_PRIM_LINES           1
+#define PIPE_PRIM_LINE_LOOP       2
+#define PIPE_PRIM_LINE_STRIP      3
+#define PIPE_PRIM_TRIANGLES       4
+#define PIPE_PRIM_TRIANGLE_STRIP  5
+#define PIPE_PRIM_TRIANGLE_FAN    6
+#define PIPE_PRIM_QUADS           7
+#define PIPE_PRIM_QUAD_STRIP      8
+#define PIPE_PRIM_POLYGON         9
+
+
+/**
+ * Query object types
+ */
+#define PIPE_QUERY_OCCLUSION_COUNTER     0
+#define PIPE_QUERY_PRIMITIVES_GENERATED  1
+#define PIPE_QUERY_PRIMITIVES_EMITTED    2
+#define PIPE_QUERY_TYPES                 3
+
+
+/**
+ * Point sprite coord modes
+ */
+#define PIPE_SPRITE_COORD_NONE       0
+#define PIPE_SPRITE_COORD_UPPER_LEFT 1
+#define PIPE_SPRITE_COORD_LOWER_LEFT 2
+
+
+/**
+ * Implementation capabilities/limits
+ * Passed to pipe->get_param()
+ * XXX this will need some fine tuning...
+ */
+#define PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS 1
+#define PIPE_CAP_NPOT_TEXTURES           2
+#define PIPE_CAP_TWO_SIDED_STENCIL       3
+#define PIPE_CAP_GLSL                    4  /* XXX need something better */
+#define PIPE_CAP_S3TC                    5
+#define PIPE_CAP_ANISOTROPIC_FILTER      6
+#define PIPE_CAP_POINT_SPRITE            7
+#define PIPE_CAP_MAX_RENDER_TARGETS      8
+#define PIPE_CAP_OCCLUSION_QUERY         9
+#define PIPE_CAP_TEXTURE_SHADOW_MAP      10
+#define PIPE_CAP_MAX_TEXTURE_2D_LEVELS   11
+#define PIPE_CAP_MAX_TEXTURE_3D_LEVELS   12
+#define PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS 13
+#define PIPE_CAP_MAX_LINE_WIDTH          14
+#define PIPE_CAP_MAX_LINE_WIDTH_AA       15
+#define PIPE_CAP_MAX_POINT_WIDTH         16
+#define PIPE_CAP_MAX_POINT_WIDTH_AA      17
+#define PIPE_CAP_MAX_TEXTURE_ANISOTROPY  18
+#define PIPE_CAP_MAX_TEXTURE_LOD_BIAS    19
+#define PIPE_CAP_GUARD_BAND_LEFT         20  /*< float */
+#define PIPE_CAP_GUARD_BAND_TOP          21  /*< float */
+#define PIPE_CAP_GUARD_BAND_RIGHT        22  /*< float */
+#define PIPE_CAP_GUARD_BAND_BOTTOM       23  /*< float */
+#define PIPE_CAP_TEXTURE_MIRROR_CLAMP    24
+#define PIPE_CAP_TEXTURE_MIRROR_REPEAT   25
+#define PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS 26
+
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/include/pipe/p_error.h b/src/gallium/include/pipe/p_error.h
new file mode 100644
index 0000000000..b865b22635
--- /dev/null
+++ b/src/gallium/include/pipe/p_error.h
@@ -0,0 +1,65 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Gallium error codes.
+ * 
+ * @author José Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+#ifndef P_ERROR_H_
+#define P_ERROR_H_
+
+
+#ifdef	__cplusplus
+extern "C" {
+#endif
+
+   
+/**
+ * Gallium error codes.
+ * 
+ * - A zero value always means success.
+ * - A negative value always means failure.
+ * - The meaning of a positive value is function dependent. 
+ */
+enum pipe_error {
+   PIPE_OK = 0,
+   PIPE_ERROR = -1,    /**< Generic error */
+   PIPE_ERROR_BAD_INPUT = -2, 
+   PIPE_ERROR_OUT_OF_MEMORY = -3,
+   PIPE_ERROR_RETRY = -4
+   /* TODO */
+};
+
+
+#ifdef	__cplusplus
+}
+#endif
+
+#endif /* P_ERROR_H_ */
diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h
new file mode 100644
index 0000000000..b42f98ceba
--- /dev/null
+++ b/src/gallium/include/pipe/p_format.h
@@ -0,0 +1,586 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright (c) 2008 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef PIPE_FORMAT_H
+#define PIPE_FORMAT_H
+
+#include "p_compiler.h"
+#include "p_debug.h"
+
+#include "util/u_string.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * The pipe_format enum is a 32-bit wide bitfield that encodes all the
+ * information needed to uniquely describe a pixel format.
+ */
+
+/**
+ * Possible format layouts are encoded in the first 2 bits.
+ * The interpretation of the remaining 30 bits depends on a particular
+ * format layout.
+ */
+#define PIPE_FORMAT_LAYOUT_RGBAZS   0
+#define PIPE_FORMAT_LAYOUT_YCBCR    1
+#define PIPE_FORMAT_LAYOUT_DXT      2  /**< XXX temporary? */
+#define PIPE_FORMAT_LAYOUT_MIXED    3
+
+static INLINE uint pf_layout(uint f)  /**< PIPE_FORMAT_LAYOUT_ */
+{
+   return f & 0x3;
+}
+
+/**
+ * RGBAZS Format Layout.
+ */
+
+/**
+ * Format component selectors for RGBAZS & MIXED layout.
+ */
+#define PIPE_FORMAT_COMP_R    0
+#define PIPE_FORMAT_COMP_G    1
+#define PIPE_FORMAT_COMP_B    2
+#define PIPE_FORMAT_COMP_A    3
+#define PIPE_FORMAT_COMP_0    4
+#define PIPE_FORMAT_COMP_1    5
+#define PIPE_FORMAT_COMP_Z    6
+#define PIPE_FORMAT_COMP_S    7
+
+/**
+ * Format types for RGBAZS layout.
+ */
+#define PIPE_FORMAT_TYPE_UNKNOWN 0
+#define PIPE_FORMAT_TYPE_FLOAT   1  /**< 16/32/64-bit/channel formats */
+#define PIPE_FORMAT_TYPE_UNORM   2  /**< uints, normalized to [0,1] */
+#define PIPE_FORMAT_TYPE_SNORM   3  /**< ints, normalized to [-1,1] */
+#define PIPE_FORMAT_TYPE_USCALED 4  /**< uints, not normalized */
+#define PIPE_FORMAT_TYPE_SSCALED 5  /**< ints, not normalized */
+#define PIPE_FORMAT_TYPE_SRGB    6  /**< sRGB colorspace */
+#define PIPE_FORMAT_TYPE_FIXED   7  /**< 16.16 fixed point */
+
+
+/**
+ * Because the destination vector is assumed to be RGBA FLOAT, we
+ * need to know how to swizzle and expand components from the source
+ * vector.
+ * Let's take U_A1_R5_G5_B5 as an example. X swizzle is A, X size
+ * is 1 bit and type is UNORM. So we take the most significant bit
+ * from source vector, convert 0 to 0.0 and 1 to 1.0 and save it
+ * in the last component of the destination RGBA component.
+ * Next, Y swizzle is R, Y size is 5 and type is UNORM. We normalize
+ * those 5 bits into [0.0; 1.0] range and put it into second
+ * component of the destination vector. Rinse and repeat for
+ * components Z and W.
+ * If any of size fields is zero, it means the source format contains
+ * less than four components.
+ * If any swizzle is 0 or 1, the corresponding destination component
+ * should be filled with 0.0 and 1.0, respectively.
+ */
+typedef uint pipe_format_rgbazs_t;
+
+static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask)
+{
+   return (f >> shift) & mask;
+}
+
+#define pf_swizzle_x(f)       pf_get(f, 2, 0x7)  /**< PIPE_FORMAT_COMP_ */
+#define pf_swizzle_y(f)       pf_get(f, 5, 0x7)  /**< PIPE_FORMAT_COMP_ */
+#define pf_swizzle_z(f)       pf_get(f, 8, 0x7)  /**< PIPE_FORMAT_COMP_ */
+#define pf_swizzle_w(f)       pf_get(f, 11, 0x7) /**< PIPE_FORMAT_COMP_ */
+#define pf_swizzle_xyzw(f,i)  pf_get(f, 2+((i)*3), 0x7)
+#define pf_size_x(f)          pf_get(f, 14, 0x7) /**< Size of X */
+#define pf_size_y(f)          pf_get(f, 17, 0x7) /**< Size of Y */
+#define pf_size_z(f)          pf_get(f, 20, 0x7) /**< Size of Z */
+#define pf_size_w(f)          pf_get(f, 23, 0x7) /**< Size of W */
+#define pf_size_xyzw(f,i)     pf_get(f, 14+((i)*3), 0x7)
+#define pf_exp2(f)            pf_get(f, 26, 0x7) /**< Scale size by 2 ^ exp2 */
+#define pf_type(f)            pf_get(f, 29, 0x7) /**< PIPE_FORMAT_TYPE_ */
+
+/**
+ * Helper macro to encode the above structure into a 32-bit value.
+ */
+#define _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, EXP2, TYPE ) (\
+   (PIPE_FORMAT_LAYOUT_RGBAZS << 0) |\
+   ((SWZ) << 2) |\
+   ((SIZEX) << 14) |\
+   ((SIZEY) << 17) |\
+   ((SIZEZ) << 20) |\
+   ((SIZEW) << 23) |\
+   ((EXP2) << 26) |\
+   ((TYPE) << 29) )
+
+/**
+ * Helper macro to encode the swizzle part of the structure above.
+ */
+#define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 0) | ((SWZY) << 3) | ((SWZZ) << 6) | ((SWZW) << 9))
+
+/**
+ * Shorthand macro for RGBAZS layout with component sizes in 1-bit units.
+ */
+#define _PIPE_FORMAT_RGBAZS_1( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\
+   _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 0, TYPE )
+
+/**
+ * Shorthand macro for RGBAZS layout with component sizes in 2-bit units.
+ */
+#define _PIPE_FORMAT_RGBAZS_2( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\
+   _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE )
+
+/**
+ * Shorthand macro for RGBAZS layout with component sizes in 8-bit units.
+ */
+#define _PIPE_FORMAT_RGBAZS_8( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\
+   _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 3, TYPE )
+
+/**
+ * Shorthand macro for RGBAZS layout with component sizes in 64-bit units.
+ */
+#define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\
+   _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 6, TYPE )
+
+typedef uint pipe_format_mixed_t;
+
+/* NOTE: Use pf_swizzle_* and pf_size_* macros for swizzles and sizes.
+ */
+
+#define pf_mixed_sign_x(f)       pf_get( f, 26, 0x1 )       /*< Sign of X */
+#define pf_mixed_sign_y(f)       pf_get( f, 27, 0x1 )       /*< Sign of Y */
+#define pf_mixed_sign_z(f)       pf_get( f, 28, 0x1 )       /*< Sign of Z */
+#define pf_mixed_sign_w(f)       pf_get( f, 29, 0x1 )       /*< Sign of W */
+#define pf_mixed_sign_xyzw(f, i) pf_get( f, 26 + (i), 0x1 )
+#define pf_mixed_normalized(f)   pf_get( f, 30, 0x1 )       /*< Type is NORM (1) or SCALED (0) */
+#define pf_mixed_scale8(f)       pf_get( f, 31, 0x1 )       /*< Scale size by either one (0) or eight (1) */
+
+/**
+ * Helper macro to encode the above structure into a 32-bit value.
+ */
+#define _PIPE_FORMAT_MIXED( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, SIGNX, SIGNY, SIGNZ, SIGNW, NORMALIZED, SCALE8 ) (\
+   (PIPE_FORMAT_LAYOUT_MIXED << 0) |\
+   ((SWZ) << 2) |\
+   ((SIZEX) << 14) |\
+   ((SIZEY) << 17) |\
+   ((SIZEZ) << 20) |\
+   ((SIZEW) << 23) |\
+   ((SIGNX) << 26) |\
+   ((SIGNY) << 27) |\
+   ((SIGNZ) << 28) |\
+   ((SIGNW) << 29) |\
+   ((NORMALIZED) << 30) |\
+   ((SCALE8) << 31) )
+
+/**
+ * Shorthand macro for common format swizzles.
+ */
+#define _PIPE_FORMAT_R001 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_1 )
+#define _PIPE_FORMAT_RG01 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_1 )
+#define _PIPE_FORMAT_RGB1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_1 )
+#define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A )
+#define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B )
+#define _PIPE_FORMAT_ABGR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R )
+#define _PIPE_FORMAT_BGRA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_A )
+#define _PIPE_FORMAT_1RGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_1, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B )
+#define _PIPE_FORMAT_1BGR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_1, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R )
+#define _PIPE_FORMAT_BGR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 )
+#define _PIPE_FORMAT_0000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 )
+#define _PIPE_FORMAT_000R _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_R )
+#define _PIPE_FORMAT_RRR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 )
+#define _PIPE_FORMAT_RRRR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R )
+#define _PIPE_FORMAT_RRRG _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G )
+#define _PIPE_FORMAT_Z000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 )
+#define _PIPE_FORMAT_0Z00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 )
+#define _PIPE_FORMAT_SZ00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 )
+#define _PIPE_FORMAT_ZS00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 )
+#define _PIPE_FORMAT_S000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 )
+
+/**
+ * YCBCR Format Layout.
+ */
+
+/**
+ * This only contains a flag that indicates whether the format is reversed or
+ * not.
+ */
+typedef uint pipe_format_ycbcr_t;
+
+/**
+ * Helper macro to encode the above structure into a 32-bit value.
+ */
+#define _PIPE_FORMAT_YCBCR( REV ) (\
+   (PIPE_FORMAT_LAYOUT_YCBCR << 0) |\
+   ((REV) << 2) )
+
+static INLINE uint pf_rev(pipe_format_ycbcr_t f)
+{
+   return (f >> 2) & 0x1;
+}
+
+
+/**
+  * Compresssed format layouts (this will probably change)
+  */
+#define _PIPE_FORMAT_DXT( LEVEL, RSIZE, GSIZE, BSIZE, ASIZE, TYPE ) \
+   ((PIPE_FORMAT_LAYOUT_DXT << 0) | \
+    ((LEVEL) << 2) | \
+    ((RSIZE) << 5) | \
+    ((GSIZE) << 8) | \
+    ((BSIZE) << 11) | \
+    ((ASIZE) << 14) | \
+    ((TYPE) << 29))
+
+
+
+/**
+ * Texture/surface image formats (preliminary)
+ */
+
+/* KW: Added lots of surface formats to support vertex element layout
+ * definitions, and eventually render-to-vertex-buffer.  Could
+ * consider making float/int/uint/scaled/normalized a separate
+ * parameter, but on the other hand there are special cases like
+ * z24s8, compressed textures, ycbcr, etc that won't fit that model.
+ */
+
+enum pipe_format {
+   PIPE_FORMAT_NONE                  = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_0000, 0, 0, 0, 0, PIPE_FORMAT_TYPE_UNKNOWN ),
+   PIPE_FORMAT_A8R8G8B8_UNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_X8R8G8B8_UNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_1RGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_B8G8R8A8_UNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_B8G8R8X8_UNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_A1R5G5B5_UNORM        = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_A4R4G4B4_UNORM        = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R5G6B5_UNORM          = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB1, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_A2B10G10R10_UNORM     = _PIPE_FORMAT_RGBAZS_2 ( _PIPE_FORMAT_ABGR, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_L8_UNORM              = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */
+   PIPE_FORMAT_A8_UNORM              = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 0, 0, 0, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */
+   PIPE_FORMAT_I8_UNORM              = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */
+   PIPE_FORMAT_A8L8_UNORM            = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha, luminance */
+   PIPE_FORMAT_L16_UNORM             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ushort luminance */
+   PIPE_FORMAT_YCBCR                 = _PIPE_FORMAT_YCBCR( 0 ),
+   PIPE_FORMAT_YCBCR_REV             = _PIPE_FORMAT_YCBCR( 1 ),
+   PIPE_FORMAT_Z16_UNORM             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_Z32_UNORM             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_Z32_FLOAT             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ),
+   PIPE_FORMAT_S8Z24_UNORM           = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_Z24S8_UNORM           = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ZS00, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_X8Z24_UNORM           = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_0Z00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_Z24X8_UNORM           = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_S8_UNORM              = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte stencil */
+   PIPE_FORMAT_R64_FLOAT             = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ),
+   PIPE_FORMAT_R64G64_FLOAT          = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ),
+   PIPE_FORMAT_R64G64B64_FLOAT       = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ),
+   PIPE_FORMAT_R64G64B64A64_FLOAT    = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ),
+   PIPE_FORMAT_R32_FLOAT             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ),
+   PIPE_FORMAT_R32G32_FLOAT          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ),
+   PIPE_FORMAT_R32G32B32_FLOAT       = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ),
+   PIPE_FORMAT_R32G32B32A32_FLOAT    = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ),
+   PIPE_FORMAT_R32_UNORM             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R32G32_UNORM          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R32G32B32_UNORM       = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R32G32B32A32_UNORM    = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R32_USCALED           = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R32G32_USCALED        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R32G32B32_USCALED     = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R32G32B32A32_USCALED  = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R32_SNORM             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R32G32_SNORM          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R32G32B32_SNORM       = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R32G32B32A32_SNORM    = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R32_SSCALED           = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R32G32_SSCALED        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R32G32B32_SSCALED     = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R32G32B32A32_SSCALED  = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R16_UNORM             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R16G16_UNORM          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R16G16B16_UNORM       = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R16G16B16A16_UNORM    = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R16_USCALED           = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R16G16_USCALED        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R16G16B16_USCALED     = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R16G16B16A16_USCALED  = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R16_SNORM             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R16G16_SNORM          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R16G16B16_SNORM       = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R16G16B16A16_SNORM    = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R16_SSCALED           = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R16G16_SSCALED        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R16G16B16_SSCALED     = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R16G16B16A16_SSCALED  = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R8_UNORM              = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R8G8_UNORM            = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R8G8B8_UNORM          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R8G8B8A8_UNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R8G8B8X8_UNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_R8_USCALED            = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R8G8_USCALED          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R8G8B8_USCALED        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R8G8B8A8_USCALED      = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R8G8B8X8_USCALED      = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ),
+   PIPE_FORMAT_R8_SNORM              = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R8G8_SNORM            = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R8G8B8_SNORM          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R8G8B8A8_SNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R8G8B8X8_SNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_B6G5R5_SNORM          = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_BGR1, 6, 5, 5, 0, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_A8B8G8R8_SNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_X8B8G8R8_SNORM        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ),
+   PIPE_FORMAT_R8_SSCALED            = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R8G8_SSCALED          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R8G8B8_SSCALED        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R8G8B8A8_SSCALED      = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R8G8B8X8_SSCALED      = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ),
+   PIPE_FORMAT_R32_FIXED             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FIXED ),
+   PIPE_FORMAT_R32G32_FIXED          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FIXED ),
+   PIPE_FORMAT_R32G32B32_FIXED       = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FIXED ),
+   PIPE_FORMAT_R32G32B32A32_FIXED    = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FIXED ),
+   /* sRGB formats */
+   PIPE_FORMAT_L8_SRGB               = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_A8L8_SRGB             = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_R8G8B8_SRGB           = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_R8G8B8A8_SRGB         = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_R8G8B8X8_SRGB         = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_A8R8G8B8_SRGB         = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_X8R8G8B8_SRGB         = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_1RGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_B8G8R8A8_SRGB         = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_B8G8R8X8_SRGB         = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ),
+
+   /* mixed formats */
+   PIPE_FORMAT_X8UB8UG8SR8S_NORM     = _PIPE_FORMAT_MIXED( _PIPE_FORMAT_1BGR, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1 ),
+   PIPE_FORMAT_B6UG5SR5S_NORM        = _PIPE_FORMAT_MIXED( _PIPE_FORMAT_BGR1, 6, 5, 5, 0, 0, 1, 1, 0, 1, 0 ),
+
+   /* compressed formats */
+   PIPE_FORMAT_DXT1_RGB              = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 0, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_DXT1_RGBA             = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 8, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_DXT3_RGBA             = _PIPE_FORMAT_DXT( 3, 8, 8, 8, 8, PIPE_FORMAT_TYPE_UNORM ),
+   PIPE_FORMAT_DXT5_RGBA             = _PIPE_FORMAT_DXT( 5, 8, 8, 8, 8, PIPE_FORMAT_TYPE_UNORM ),
+
+   /* sRGB, compressed */
+   PIPE_FORMAT_DXT1_SRGB             = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 0, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_DXT1_SRGBA            = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 8, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_DXT3_SRGBA            = _PIPE_FORMAT_DXT( 3, 8, 8, 8, 8, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_DXT5_SRGBA            = _PIPE_FORMAT_DXT( 5, 8, 8, 8, 8, PIPE_FORMAT_TYPE_SRGB )
+};
+
+/**
+ * Builds pipe format name from format token.
+ */
+extern const char *pf_name( enum pipe_format format );
+
+/**
+ * Return bits for a particular component.
+ * \param comp  component index, starting at 0
+ */
+static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp )
+{
+   uint size;
+
+   if (pf_swizzle_x(format) == comp) {
+      size = pf_size_x(format);
+   }
+   else if (pf_swizzle_y(format) == comp) {
+      size = pf_size_y(format);
+   }
+   else if (pf_swizzle_z(format) == comp) {
+      size = pf_size_z(format);
+   }
+   else if (pf_swizzle_w(format) == comp) {
+      size = pf_size_w(format);
+   }
+   else {
+      size = 0;
+   }
+   if (pf_layout( format ) == PIPE_FORMAT_LAYOUT_RGBAZS)
+      return size << pf_exp2( format );
+   return size << (pf_mixed_scale8( format ) * 3);
+}
+
+/**
+ * Return total bits needed for the pixel format.
+ */
+static INLINE uint pf_get_bits( enum pipe_format format )
+{
+   switch (pf_layout(format)) {
+   case PIPE_FORMAT_LAYOUT_RGBAZS:
+   case PIPE_FORMAT_LAYOUT_MIXED:
+      return
+         pf_get_component_bits( format, PIPE_FORMAT_COMP_0 ) +
+         pf_get_component_bits( format, PIPE_FORMAT_COMP_1 ) +
+         pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) +
+         pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) +
+         pf_get_component_bits( format, PIPE_FORMAT_COMP_B ) +
+         pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) +
+         pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) +
+         pf_get_component_bits( format, PIPE_FORMAT_COMP_S );
+   case PIPE_FORMAT_LAYOUT_YCBCR:
+      assert( format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV );
+      /* return effective bits per pixel */
+      return 16; 
+   default:
+      assert( 0 );
+      return 0;
+   }
+}
+
+/**
+ * Return bytes per pixel for the given format.
+ */
+static INLINE uint pf_get_size( enum pipe_format format )
+{
+   assert(pf_get_bits(format) % 8 == 0);
+   return pf_get_bits(format) / 8;
+}
+
+/**
+ * Describe accurately the pixel format.
+ * 
+ * The chars-per-pixel concept falls apart with compressed and yuv images, where
+ * more than one pixel are coded in a single data block. This structure 
+ * describes that block.
+ * 
+ * Simple pixel formats are effectively a 1x1xcpp block.
+ */
+struct pipe_format_block
+{
+   /** Block size in bytes */
+   unsigned size;
+   
+   /** Block width in pixels */
+   unsigned width;
+   
+   /** Block height in pixels */
+   unsigned height;
+};
+
+/**
+ * Describe pixel format's block.   
+ * 
+ * @sa http://msdn2.microsoft.com/en-us/library/ms796147.aspx
+ */
+static INLINE void 
+pf_get_block(enum pipe_format format, struct pipe_format_block *block)
+{
+   switch(format) {
+   case PIPE_FORMAT_DXT1_RGBA:
+   case PIPE_FORMAT_DXT1_RGB:
+   case PIPE_FORMAT_DXT1_SRGBA:
+   case PIPE_FORMAT_DXT1_SRGB:
+      block->size = 8;
+      block->width = 4;
+      block->height = 4;
+      break;
+   case PIPE_FORMAT_DXT3_RGBA:
+   case PIPE_FORMAT_DXT5_RGBA:
+   case PIPE_FORMAT_DXT3_SRGBA:
+   case PIPE_FORMAT_DXT5_SRGBA:
+      block->size = 16;
+      block->width = 4;
+      block->height = 4;
+      break;
+   case PIPE_FORMAT_YCBCR:
+   case PIPE_FORMAT_YCBCR_REV:
+      block->size = 4; /* 2*cpp */
+      block->width = 2;
+      block->height = 1;
+      break;
+   default:
+      block->size = pf_get_size(format);
+      block->width = 1;
+      block->height = 1;
+      break;
+   }
+}
+
+static INLINE unsigned
+pf_get_nblocksx(const struct pipe_format_block *block, unsigned x)
+{
+   return (x + block->width - 1)/block->width;
+}
+
+static INLINE unsigned
+pf_get_nblocksy(const struct pipe_format_block *block, unsigned y)
+{
+   return (y + block->height - 1)/block->height;
+}
+
+static INLINE unsigned
+pf_get_nblocks(const struct pipe_format_block *block, unsigned width, unsigned height)
+{
+   return pf_get_nblocksx(block, width)*pf_get_nblocksy(block, height);
+}
+
+static INLINE boolean 
+pf_is_compressed( enum pipe_format format )
+{
+   return pf_layout(format) == PIPE_FORMAT_LAYOUT_DXT ? TRUE : FALSE;
+}
+
+static INLINE boolean 
+pf_is_ycbcr( enum pipe_format format )
+{
+   return pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR ? TRUE : FALSE;
+}
+
+static INLINE boolean 
+pf_has_alpha( enum pipe_format format )
+{
+   switch (pf_layout(format)) {
+   case PIPE_FORMAT_LAYOUT_RGBAZS:
+   case PIPE_FORMAT_LAYOUT_MIXED:
+      /* FIXME: pf_get_component_bits( PIPE_FORMAT_A8L8_UNORM, PIPE_FORMAT_COMP_A ) should not return 0 right? */
+      if(format == PIPE_FORMAT_A8_UNORM || 
+         format == PIPE_FORMAT_A8L8_UNORM || 
+         format == PIPE_FORMAT_A8L8_SRGB)
+         return TRUE;
+      return pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) ? TRUE : FALSE;
+   case PIPE_FORMAT_LAYOUT_YCBCR:
+      return FALSE; 
+   case PIPE_FORMAT_LAYOUT_DXT:
+      switch (format) {
+      case PIPE_FORMAT_DXT1_RGBA:
+      case PIPE_FORMAT_DXT3_RGBA:
+      case PIPE_FORMAT_DXT5_RGBA:
+      case PIPE_FORMAT_DXT1_SRGBA:
+      case PIPE_FORMAT_DXT3_SRGBA:
+      case PIPE_FORMAT_DXT5_SRGBA:
+         return TRUE;
+      default:
+         return FALSE;
+      }
+   default:
+      assert( 0 );
+      return FALSE;
+   }
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h
new file mode 100644
index 0000000000..1219c817b4
--- /dev/null
+++ b/src/gallium/include/pipe/p_inlines.h
@@ -0,0 +1,189 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef P_INLINES_H
+#define P_INLINES_H
+
+#include "p_context.h"
+#include "p_defines.h"
+#include "p_screen.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/* XXX: these are a kludge.  will fix when all surfaces are views into
+ * textures, and free-floating winsys surfaces go away.
+ */
+static INLINE void *
+pipe_surface_map( struct pipe_surface *surf, unsigned flags )
+{
+   struct pipe_screen *screen;
+   assert(surf->texture);
+   screen = surf->texture->screen;
+   return screen->surface_map( screen, surf, flags );
+}
+
+static INLINE void
+pipe_surface_unmap( struct pipe_surface *surf )
+{
+   struct pipe_screen *screen;
+   assert(surf->texture);
+   screen = surf->texture->screen;
+   screen->surface_unmap( screen, surf );
+}
+
+
+
+/**
+ * Set 'ptr' to point to 'surf' and update reference counting.
+ * The old thing pointed to, if any, will be unreferenced first.
+ * 'surf' may be NULL.
+ */
+static INLINE void
+pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf)
+{
+   /* bump the refcount first */
+   if (surf) {
+      assert(surf->refcount);
+      surf->refcount++;
+   }
+
+   if (*ptr) {
+      struct pipe_screen *screen;
+      assert((*ptr)->refcount);
+      assert((*ptr)->texture);
+      screen = (*ptr)->texture->screen;
+      screen->tex_surface_release( screen, ptr );
+      assert(!*ptr);
+   }
+
+   *ptr = surf;
+}
+
+
+/**
+ * \sa pipe_surface_reference
+ */
+static INLINE void
+pipe_texture_reference(struct pipe_texture **ptr,
+		       struct pipe_texture *pt)
+{
+   assert(ptr);
+
+   if (pt) { 
+      assert(pt->refcount);
+      pt->refcount++;
+   }
+
+   if (*ptr) {
+      struct pipe_screen *screen = (*ptr)->screen;
+      assert(screen);
+      assert((*ptr)->refcount);
+      screen->texture_release(screen, ptr);
+
+      assert(!*ptr);
+   }
+
+   *ptr = pt;
+}
+
+
+static INLINE void
+pipe_texture_release(struct pipe_texture **ptr)
+{
+   struct pipe_screen *screen;
+   assert(ptr);
+   screen = (*ptr)->screen;
+   assert((*ptr)->refcount);
+   screen->texture_release(screen, ptr);
+   *ptr = NULL;
+}
+
+
+/**
+ * Convenience wrappers for screen buffer functions.
+ */
+
+static INLINE struct pipe_buffer *
+pipe_buffer_create( struct pipe_screen *screen,
+                    unsigned alignment, unsigned usage, unsigned size )
+{
+   return screen->buffer_create(screen, alignment, usage, size);
+}
+
+static INLINE struct pipe_buffer *
+pipe_user_buffer_create( struct pipe_screen *screen, void *ptr, unsigned size )
+{
+   return screen->user_buffer_create(screen, ptr, size);
+}
+
+static INLINE void *
+pipe_buffer_map(struct pipe_screen *screen,
+                struct pipe_buffer *buf,
+                unsigned usage)
+{
+   return screen->buffer_map(screen, buf, usage);
+}
+
+static INLINE void
+pipe_buffer_unmap(struct pipe_screen *screen,
+                  struct pipe_buffer *buf)
+{
+   screen->buffer_unmap(screen, buf);
+}
+
+/* XXX: thread safety issues!
+ */
+static INLINE void
+pipe_buffer_reference(struct pipe_screen *screen,
+		      struct pipe_buffer **ptr,
+		      struct pipe_buffer *buf)
+{
+   if (buf) {
+      assert(buf->refcount);
+      buf->refcount++;
+   }
+
+   if (*ptr) {
+      assert((*ptr)->refcount);
+      if(--(*ptr)->refcount == 0) {
+         screen->buffer_destroy( screen, *ptr );
+      }
+   }
+
+   *ptr = buf;
+}
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* P_INLINES_H */
diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h
new file mode 100644
index 0000000000..17d1548253
--- /dev/null
+++ b/src/gallium/include/pipe/p_screen.h
@@ -0,0 +1,246 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * 
+ * Screen, Adapter or GPU
+ *
+ * These are driver functions/facilities that are context independent.
+ */
+
+
+#ifndef P_SCREEN_H
+#define P_SCREEN_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "pipe/p_defines.h"
+
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/** Opaque type */
+struct pipe_fence_handle;
+struct pipe_winsys;
+struct pipe_buffer;
+
+
+/**
+ * Gallium screen/adapter context.  Basically everything
+ * hardware-specific that doesn't actually require a rendering
+ * context.
+ */
+struct pipe_screen {
+   struct pipe_winsys *winsys;
+
+   void (*destroy)( struct pipe_screen * );
+
+
+   const char *(*get_name)( struct pipe_screen * );
+
+   const char *(*get_vendor)( struct pipe_screen * );
+
+   /**
+    * Query an integer-valued capability/parameter/limit
+    * \param param  one of PIPE_CAP_x
+    */
+   int (*get_param)( struct pipe_screen *, int param );
+
+   /**
+    * Query a float-valued capability/parameter/limit
+    * \param param  one of PIPE_CAP_x
+    */
+   float (*get_paramf)( struct pipe_screen *, int param );
+
+   /**
+    * Check if the given pipe_format is supported as a texture or
+    * drawing surface.
+    * \param tex_usage  bitmask of PIPE_TEXTURE_USAGE_*
+    * \param flags  bitmask of PIPE_TEXTURE_GEOM_*
+    */
+   boolean (*is_format_supported)( struct pipe_screen *,
+                                   enum pipe_format format,
+                                   enum pipe_texture_target target,
+                                   unsigned tex_usage, 
+                                   unsigned geom_flags );
+
+   /**
+    * Create a new texture object, using the given template info.
+    */
+   struct pipe_texture * (*texture_create)(struct pipe_screen *,
+                                           const struct pipe_texture *templat);
+
+   /**
+    * Create a new texture object, using the given template info, but on top of 
+    * existing memory.
+    * 
+    * It is assumed that the buffer data is layed out according to the expected
+    * by the hardware. NULL will be returned if any inconsistency is found.  
+    */
+   struct pipe_texture * (*texture_blanket)(struct pipe_screen *,
+                                            const struct pipe_texture *templat,
+                                            const unsigned *stride,
+                                            struct pipe_buffer *buffer);
+
+   void (*texture_release)(struct pipe_screen *,
+                           struct pipe_texture **pt);
+
+   /** Get a surface which is a "view" into a texture */
+   struct pipe_surface *(*get_tex_surface)(struct pipe_screen *,
+                                           struct pipe_texture *texture,
+                                           unsigned face, unsigned level,
+                                           unsigned zslice,
+                                           unsigned usage );
+
+   /* Surfaces allocated by the above must be released here:
+    */
+   void (*tex_surface_release)( struct pipe_screen *,
+                                struct pipe_surface ** );
+   
+
+   void *(*surface_map)( struct pipe_screen *,
+                         struct pipe_surface *surface,
+                         unsigned flags );
+
+   void (*surface_unmap)( struct pipe_screen *,
+                          struct pipe_surface *surface );
+
+
+   /**
+    * Buffer management. Buffer attributes are mostly fixed over its lifetime.
+    *
+    */
+   struct pipe_buffer *(*buffer_create)( struct pipe_screen *screen,
+                                         unsigned alignment,
+                                         unsigned usage,
+                                         unsigned size );
+
+   /**
+    * Create a buffer that wraps user-space data.
+    *
+    * Effectively this schedules a delayed call to buffer_create
+    * followed by an upload of the data at *some point in the future*,
+    * or perhaps never.  Basically the allocate/upload is delayed
+    * until the buffer is actually passed to hardware.
+    *
+    * The intention is to provide a quick way to turn regular data
+    * into a buffer, and secondly to avoid a copy operation if that
+    * data subsequently turns out to be only accessed by the CPU.
+    *
+    * Common example is OpenGL vertex buffers that are subsequently
+    * processed either by software TNL in the driver or by passing to
+    * hardware.
+    *
+    * XXX: What happens if the delayed call to buffer_create() fails?
+    *
+    * Note that ptr may be accessed at any time upto the time when the
+    * buffer is destroyed, so the data must not be freed before then.
+    */
+   struct pipe_buffer *(*user_buffer_create)(struct pipe_screen *screen,
+                                             void *ptr,
+                                             unsigned bytes);
+
+   /**
+    * Allocate storage for a display target surface.
+    *
+    * Often surfaces which are meant to be blitted to the front screen (i.e.,
+    * display targets) must be allocated with special characteristics, memory
+    * pools, or obtained directly from the windowing system.
+    *
+    * This callback is invoked by the pipe_screenwhen creating a texture marked
+    * with the PIPE_TEXTURE_USAGE_DISPLAY_TARGET flag  to get the underlying
+    * buffer storage.
+    */
+   struct pipe_buffer *(*surface_buffer_create)(struct pipe_screen *screen,
+						unsigned width, unsigned height,
+						enum pipe_format format,
+						unsigned usage,
+						unsigned *stride);
+
+
+   /**
+    * Map the entire data store of a buffer object into the client's address.
+    * flags is bitmask of PIPE_BUFFER_USAGE_CPU_READ/WRITE flags.
+    */
+   void *(*buffer_map)( struct pipe_screen *screen,
+			struct pipe_buffer *buf,
+			unsigned usage );
+
+   void (*buffer_unmap)( struct pipe_screen *screen,
+			 struct pipe_buffer *buf );
+
+   void (*buffer_destroy)( struct pipe_screen *screen,
+			   struct pipe_buffer *buf );
+
+
+   /**
+    * Do any special operations to ensure frontbuffer contents are
+    * displayed, eg copy fake frontbuffer.
+    */
+   void (*flush_frontbuffer)( struct pipe_screen *screen,
+                              struct pipe_surface *surf,
+                              void *context_private );
+
+
+
+   /** Set ptr = fence, with reference counting */
+   void (*fence_reference)( struct pipe_screen *screen,
+                            struct pipe_fence_handle **ptr,
+                            struct pipe_fence_handle *fence );
+
+   /**
+    * Checks whether the fence has been signalled.
+    * \param flags  driver-specific meaning
+    * \return zero on success.
+    */
+   int (*fence_signalled)( struct pipe_screen *screen,
+                           struct pipe_fence_handle *fence,
+                           unsigned flag );
+
+   /**
+    * Wait for the fence to finish.
+    * \param flags  driver-specific meaning
+    * \return zero on success.
+    */
+   int (*fence_finish)( struct pipe_screen *screen,
+                        struct pipe_fence_handle *fence,
+                        unsigned flag );
+
+};
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* P_SCREEN_H */
diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h
new file mode 100644
index 0000000000..35df70e7b7
--- /dev/null
+++ b/src/gallium/include/pipe/p_shader_tokens.h
@@ -0,0 +1,811 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef TGSI_TOKEN_H
+#define TGSI_TOKEN_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include "p_compiler.h"
+
+struct tgsi_version
+{
+   unsigned MajorVersion  : 8;
+   unsigned MinorVersion  : 8;
+   unsigned Padding       : 16;
+};
+
+struct tgsi_header
+{
+   unsigned HeaderSize : 8;
+   unsigned BodySize   : 24;
+};
+
+#define TGSI_PROCESSOR_FRAGMENT  0
+#define TGSI_PROCESSOR_VERTEX    1
+#define TGSI_PROCESSOR_GEOMETRY  2
+
+struct tgsi_processor
+{
+   unsigned Processor  : 4;  /* TGSI_PROCESSOR_ */
+   unsigned Padding    : 28;
+};
+
+#define TGSI_TOKEN_TYPE_DECLARATION    0
+#define TGSI_TOKEN_TYPE_IMMEDIATE      1
+#define TGSI_TOKEN_TYPE_INSTRUCTION    2
+
+struct tgsi_token
+{
+   unsigned Type       : 4;  /**< TGSI_TOKEN_TYPE_x */
+   unsigned NrTokens   : 8;  /**< UINT */
+   unsigned Padding    : 19;
+   unsigned Extended   : 1;  /**< BOOL */
+};
+
+enum tgsi_file_type {
+   TGSI_FILE_NULL        =0,
+   TGSI_FILE_CONSTANT    =1,
+   TGSI_FILE_INPUT       =2,
+   TGSI_FILE_OUTPUT      =3,
+   TGSI_FILE_TEMPORARY   =4,
+   TGSI_FILE_SAMPLER     =5,
+   TGSI_FILE_ADDRESS     =6,
+   TGSI_FILE_IMMEDIATE   =7,
+   TGSI_FILE_COUNT      /**< how many TGSI_FILE_ types */
+};
+
+
+#define TGSI_WRITEMASK_NONE     0x00
+#define TGSI_WRITEMASK_X        0x01
+#define TGSI_WRITEMASK_Y        0x02
+#define TGSI_WRITEMASK_XY       0x03
+#define TGSI_WRITEMASK_Z        0x04
+#define TGSI_WRITEMASK_XZ       0x05
+#define TGSI_WRITEMASK_YZ       0x06
+#define TGSI_WRITEMASK_XYZ      0x07
+#define TGSI_WRITEMASK_W        0x08
+#define TGSI_WRITEMASK_XW       0x09
+#define TGSI_WRITEMASK_YW       0x0A
+#define TGSI_WRITEMASK_XYW      0x0B
+#define TGSI_WRITEMASK_ZW       0x0C
+#define TGSI_WRITEMASK_XZW      0x0D
+#define TGSI_WRITEMASK_YZW      0x0E
+#define TGSI_WRITEMASK_XYZW     0x0F
+
+#define TGSI_INTERPOLATE_CONSTANT      0
+#define TGSI_INTERPOLATE_LINEAR        1
+#define TGSI_INTERPOLATE_PERSPECTIVE   2
+#define TGSI_INTERPOLATE_COUNT         3
+
+struct tgsi_declaration
+{
+   unsigned Type        : 4;  /**< TGSI_TOKEN_TYPE_DECLARATION */
+   unsigned NrTokens    : 8;  /**< UINT */
+   unsigned File        : 4;  /**< one of TGSI_FILE_x */
+   unsigned UsageMask   : 4;  /**< bitmask of TGSI_WRITEMASK_x flags */
+   unsigned Interpolate : 4;  /**< one of TGSI_INTERPOLATE_x */
+   unsigned Semantic    : 1;  /**< BOOL, any semantic info? */
+   unsigned Centroid    : 1;  /**< centroid sampling? */
+   unsigned Invariant   : 1;  /**< invariant optimization? */
+   unsigned Padding     : 4;
+   unsigned Extended    : 1;  /**< BOOL */
+};
+
+struct tgsi_declaration_range
+{
+   unsigned First   : 16; /**< UINT */
+   unsigned Last    : 16; /**< UINT */
+};
+
+#define TGSI_SEMANTIC_POSITION 0
+#define TGSI_SEMANTIC_COLOR    1
+#define TGSI_SEMANTIC_BCOLOR   2 /**< back-face color */
+#define TGSI_SEMANTIC_FOG      3
+#define TGSI_SEMANTIC_PSIZE    4
+#define TGSI_SEMANTIC_GENERIC  5
+#define TGSI_SEMANTIC_NORMAL   6
+#define TGSI_SEMANTIC_COUNT    7 /**< number of semantic values */
+
+struct tgsi_declaration_semantic
+{
+   unsigned SemanticName   : 8;  /**< one of TGSI_SEMANTIC_x */
+   unsigned SemanticIndex  : 16; /**< UINT */
+   unsigned Padding        : 8;
+};
+
+#define TGSI_IMM_FLOAT32   0
+
+struct tgsi_immediate
+{
+   unsigned Type       : 4;  /**< TGSI_TOKEN_TYPE_IMMEDIATE */
+   unsigned NrTokens   : 8;  /**< UINT */
+   unsigned DataType   : 4;  /**< one of TGSI_IMM_x */
+   unsigned Padding    : 15;
+   unsigned Extended   : 1;  /**< BOOL */
+};
+
+struct tgsi_immediate_float32
+{
+   float Float;
+};
+
+/*
+ * GL_NV_vertex_program
+ */
+#define TGSI_OPCODE_ARL                 0
+#define TGSI_OPCODE_MOV                 1
+#define TGSI_OPCODE_LIT                 2
+#define TGSI_OPCODE_RCP                 3
+#define TGSI_OPCODE_RSQ                 4
+#define TGSI_OPCODE_EXP                 5
+#define TGSI_OPCODE_LOG                 6
+#define TGSI_OPCODE_MUL                 7
+#define TGSI_OPCODE_ADD                 8
+#define TGSI_OPCODE_DP3                 9
+#define TGSI_OPCODE_DP4                 10
+#define TGSI_OPCODE_DST                 11
+#define TGSI_OPCODE_MIN                 12
+#define TGSI_OPCODE_MAX                 13
+#define TGSI_OPCODE_SLT                 14
+#define TGSI_OPCODE_SGE                 15
+#define TGSI_OPCODE_MAD                 16
+
+/*
+ * GL_ATI_fragment_shader
+ */
+#define TGSI_OPCODE_SUB                 17
+#define TGSI_OPCODE_DOT3                TGSI_OPCODE_DP3
+#define TGSI_OPCODE_DOT4                TGSI_OPCODE_DP4
+#define TGSI_OPCODE_LERP                18
+#define TGSI_OPCODE_CND                 19
+#define TGSI_OPCODE_CND0                20
+#define TGSI_OPCODE_DOT2ADD             21
+
+/*
+ * GL_EXT_vertex_shader
+ */
+#define TGSI_OPCODE_INDEX               22
+#define TGSI_OPCODE_NEGATE              23
+#define TGSI_OPCODE_MADD                TGSI_OPCODE_MAD
+#define TGSI_OPCODE_FRAC                24
+#define TGSI_OPCODE_SETGE               TGSI_OPCODE_SGE
+#define TGSI_OPCODE_SETLT               TGSI_OPCODE_SLT
+#define TGSI_OPCODE_CLAMP               25
+#define TGSI_OPCODE_FLOOR               26
+#define TGSI_OPCODE_ROUND               27
+#define TGSI_OPCODE_EXPBASE2            28
+#define TGSI_OPCODE_LOGBASE2            29
+#define TGSI_OPCODE_POWER               30
+#define TGSI_OPCODE_RECIP               TGSI_OPCODE_RCP
+#define TGSI_OPCODE_RECIPSQRT           TGSI_OPCODE_RSQ
+#define TGSI_OPCODE_CROSSPRODUCT        31
+#define TGSI_OPCODE_MULTIPLYMATRIX      32
+
+/*
+ * GL_NV_vertex_program1_1
+ */
+#define TGSI_OPCODE_ABS                 33
+#define TGSI_OPCODE_RCC                 34
+#define TGSI_OPCODE_DPH                 35
+
+/*
+ * GL_NV_fragment_program
+ */
+#define TGSI_OPCODE_COS                 36
+#define TGSI_OPCODE_DDX                 37
+#define TGSI_OPCODE_DDY                 38
+#define TGSI_OPCODE_EX2                 TGSI_OPCODE_EXPBASE2
+#define TGSI_OPCODE_FLR                 TGSI_OPCODE_FLOOR
+#define TGSI_OPCODE_FRC                 TGSI_OPCODE_FRAC
+#define TGSI_OPCODE_KILP                39  /* predicated kill */
+#define TGSI_OPCODE_LG2                 TGSI_OPCODE_LOGBASE2
+#define TGSI_OPCODE_LRP                 TGSI_OPCODE_LERP
+#define TGSI_OPCODE_PK2H                40
+#define TGSI_OPCODE_PK2US               41
+#define TGSI_OPCODE_PK4B                42
+#define TGSI_OPCODE_PK4UB               43
+#define TGSI_OPCODE_POW                 TGSI_OPCODE_POWER
+#define TGSI_OPCODE_RFL                 44
+#define TGSI_OPCODE_SEQ                 45
+#define TGSI_OPCODE_SFL                 46
+#define TGSI_OPCODE_SGT                 47
+#define TGSI_OPCODE_SIN                 48
+#define TGSI_OPCODE_SLE                 49
+#define TGSI_OPCODE_SNE                 50
+#define TGSI_OPCODE_STR                 51
+#define TGSI_OPCODE_TEX                 52
+#define TGSI_OPCODE_TXD                 53
+#define TGSI_OPCODE_TXP                 54
+#define TGSI_OPCODE_UP2H                55
+#define TGSI_OPCODE_UP2US               56
+#define TGSI_OPCODE_UP4B                57
+#define TGSI_OPCODE_UP4UB               58
+#define TGSI_OPCODE_X2D                 59
+
+/*
+ * GL_NV_vertex_program2
+ */
+#define TGSI_OPCODE_ARA                 60
+#define TGSI_OPCODE_ARR                 61
+#define TGSI_OPCODE_BRA                 62
+#define TGSI_OPCODE_CAL                 63
+#define TGSI_OPCODE_RET                 64
+#define TGSI_OPCODE_SSG                 65
+
+/*
+ * GL_ARB_vertex_program
+ */
+#define TGSI_OPCODE_SWZ                 118
+#define TGSI_OPCODE_XPD                 TGSI_OPCODE_CROSSPRODUCT
+
+/*
+ * GL_ARB_fragment_program
+ */
+#define TGSI_OPCODE_CMP                 66
+#define TGSI_OPCODE_KIL                 116  /* conditional kill */
+#define TGSI_OPCODE_SCS                 67
+#define TGSI_OPCODE_TXB                 68
+
+/*
+ * GL_NV_fragment_program_option
+ */
+/* No new opcode */
+
+/*
+ * GL_NV_fragment_program2
+ */
+#define TGSI_OPCODE_NRM                 69
+#define TGSI_OPCODE_DIV                 70
+#define TGSI_OPCODE_DP2                 71
+#define TGSI_OPCODE_DP2A                TGSI_OPCODE_DOT2ADD
+#define TGSI_OPCODE_TXL                 72
+#define TGSI_OPCODE_BRK                 73
+#define TGSI_OPCODE_IF                  74
+#define TGSI_OPCODE_LOOP                75
+#define TGSI_OPCODE_REP                 76
+#define TGSI_OPCODE_ELSE                77
+#define TGSI_OPCODE_ENDIF               78
+#define TGSI_OPCODE_ENDLOOP             79
+#define TGSI_OPCODE_ENDREP              80
+
+/*
+ * GL_NV_vertex_program2_option
+ */
+
+/*
+ * GL_NV_vertex_program3
+ */
+#define TGSI_OPCODE_PUSHA               81
+#define TGSI_OPCODE_POPA                82
+
+/*
+ * GL_NV_gpu_program4
+ */
+#define TGSI_OPCODE_CEIL                83
+#define TGSI_OPCODE_I2F                 84
+#define TGSI_OPCODE_NOT                 85
+#define TGSI_OPCODE_TRUNC               86
+#define TGSI_OPCODE_SHL                 87
+#define TGSI_OPCODE_SHR                 88
+#define TGSI_OPCODE_AND                 89
+#define TGSI_OPCODE_OR                  90
+#define TGSI_OPCODE_MOD                 91
+#define TGSI_OPCODE_XOR                 92
+#define TGSI_OPCODE_SAD                 93
+#define TGSI_OPCODE_TXF                 94
+#define TGSI_OPCODE_TXQ                 95
+#define TGSI_OPCODE_CONT                96
+
+/*
+ * GL_NV_vertex_program4
+ */
+/* Same as GL_NV_gpu_program4 */
+
+/*
+ * GL_NV_fragment_program4
+ */
+/* Same as GL_NV_gpu_program4 */
+
+/*
+ * GL_NV_geometry_program4
+ */
+/* Same as GL_NV_gpu_program4 */
+#define TGSI_OPCODE_EMIT                97
+#define TGSI_OPCODE_ENDPRIM             98
+
+/*
+ * GLSL
+ */
+#define TGSI_OPCODE_BGNLOOP2            99
+#define TGSI_OPCODE_BGNSUB              100
+#define TGSI_OPCODE_ENDLOOP2            101
+#define TGSI_OPCODE_ENDSUB              102
+#define TGSI_OPCODE_INT                 TGSI_OPCODE_TRUNC
+#define TGSI_OPCODE_NOISE1              103
+#define TGSI_OPCODE_NOISE2              104
+#define TGSI_OPCODE_NOISE3              105
+#define TGSI_OPCODE_NOISE4              106
+#define TGSI_OPCODE_NOP                 107
+
+/*
+ * ps_1_1
+ */
+#define TGSI_OPCODE_TEXKILL             TGSI_OPCODE_KIL
+
+/*
+ * ps_1_2
+ */
+/* CMP - use TGSI_OPCODE_CND0 */
+
+/*
+ * ps_1_3
+ */
+/* CMP - use TGSI_OPCODE_CND0 */
+
+/*
+ * ps_1_4
+ */
+#define TGSI_OPCODE_TEXLD               TGSI_OPCODE_TEX
+
+/*
+ * ps_2_0
+ */
+#define TGSI_OPCODE_M4X4                TGSI_OPCODE_MULTIPLYMATRIX
+#define TGSI_OPCODE_M4X3                108
+#define TGSI_OPCODE_M3X4                109
+#define TGSI_OPCODE_M3X3                110
+#define TGSI_OPCODE_M3X2                111
+#define TGSI_OPCODE_CRS                 TGSI_OPCODE_XPD
+#define TGSI_OPCODE_NRM4                112
+#define TGSI_OPCODE_SINCOS              TGSI_OPCODE_SCS
+#define TGSI_OPCODE_TEXLDB              TGSI_OPCODE_TXB
+#define TGSI_OPCODE_DP2ADD              TGSI_OPCODE_DP2A
+
+/*
+ * ps_2_x
+ */
+#define TGSI_OPCODE_CALL                TGSI_OPCODE_CAL
+#define TGSI_OPCODE_CALLNZ              113
+#define TGSI_OPCODE_IFC                 114
+#define TGSI_OPCODE_BREAK               TGSI_OPCODE_BRK
+#define TGSI_OPCODE_BREAKC              115
+#define TGSI_OPCODE_DSX                 TGSI_OPCODE_DDX
+#define TGSI_OPCODE_DSY                 TGSI_OPCODE_DDY
+#define TGSI_OPCODE_TEXLDD              TGSI_OPCODE_TXD
+
+/*
+ * vs_1_1
+ */
+#define TGSI_OPCODE_EXPP                TGSI_OPCODE_EXP
+#define TGSI_OPCODE_LOGP                TGSI_OPCODE_LG2
+
+/*
+ * vs_2_0
+ */
+#define TGSI_OPCODE_SGN                 TGSI_OPCODE_SSG
+#define TGSI_OPCODE_MOVA                TGSI_OPCODE_ARR
+/* EXPP - use TGSI_OPCODE_EX2 */
+
+/*
+ * vs_2_x
+ */
+
+#define TGSI_OPCODE_END                 117  /* aka HALT */
+
+#define TGSI_OPCODE_LAST                119
+
+#define TGSI_SAT_NONE            0  /* do not saturate */
+#define TGSI_SAT_ZERO_ONE        1  /* clamp to [0,1] */
+#define TGSI_SAT_MINUS_PLUS_ONE  2  /* clamp to [-1,1] */
+
+/**
+ * Opcode is the operation code to execute. A given operation defines the
+ * semantics how the source registers (if any) are interpreted and what is
+ * written to the destination registers (if any) as a result of execution.
+ *
+ * NumDstRegs and NumSrcRegs is the number of destination and source registers,
+ * respectively. For a given operation code, those numbers are fixed and are
+ * present here only for convenience.
+ *
+ * If Extended is TRUE, it is now executed.
+ *
+ * Saturate controls how are final results in destination registers modified.
+ */
+
+struct tgsi_instruction
+{
+   unsigned Type       : 4;  /* TGSI_TOKEN_TYPE_INSTRUCTION */
+   unsigned NrTokens   : 8;  /* UINT */
+   unsigned Opcode     : 8;  /* TGSI_OPCODE_ */
+   unsigned Saturate   : 2;  /* TGSI_SAT_ */
+   unsigned NumDstRegs : 2;  /* UINT */
+   unsigned NumSrcRegs : 4;  /* UINT */
+   unsigned Padding    : 3;
+   unsigned Extended   : 1;  /* BOOL */
+};
+
+/*
+ * If tgsi_instruction::Extended is TRUE, tgsi_instruction_ext follows.
+ * 
+ * Then, tgsi_instruction::NumDstRegs of tgsi_dst_register follow.
+ * 
+ * Then, tgsi_instruction::NumSrcRegs of tgsi_src_register follow.
+ *
+ * tgsi_instruction::NrTokens contains the total number of words that make the
+ * instruction, including the instruction word.
+ */
+
+#define TGSI_INSTRUCTION_EXT_TYPE_NV        0
+#define TGSI_INSTRUCTION_EXT_TYPE_LABEL     1
+#define TGSI_INSTRUCTION_EXT_TYPE_TEXTURE   2
+#define TGSI_INSTRUCTION_EXT_TYPE_PREDICATE 3
+
+struct tgsi_instruction_ext
+{
+   unsigned Type       : 4;  /* TGSI_INSTRUCTION_EXT_TYPE_ */
+   unsigned Padding    : 27;
+   unsigned Extended   : 1;  /* BOOL */
+};
+
+/*
+ * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_NV, it should
+ * be cast to tgsi_instruction_ext_nv.
+ * 
+ * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_LABEL, it
+ * should be cast to tgsi_instruction_ext_label.
+ * 
+ * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_TEXTURE, it
+ * should be cast to tgsi_instruction_ext_texture.
+ * 
+ * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_PREDICATE, it
+ * should be cast to tgsi_instruction_ext_predicate.
+ * 
+ * If tgsi_instruction_ext::Extended is TRUE, another tgsi_instruction_ext
+ * follows.
+ */
+
+#define TGSI_PRECISION_DEFAULT      0
+#define TGSI_PRECISION_FLOAT32      1
+#define TGSI_PRECISION_FLOAT16      2
+#define TGSI_PRECISION_FIXED12      3
+
+#define TGSI_CC_GT      0
+#define TGSI_CC_EQ      1
+#define TGSI_CC_LT      2
+#define TGSI_CC_GE      3
+#define TGSI_CC_LE      4
+#define TGSI_CC_NE      5
+#define TGSI_CC_TR      6
+#define TGSI_CC_FL      7
+
+#define TGSI_SWIZZLE_X      0
+#define TGSI_SWIZZLE_Y      1
+#define TGSI_SWIZZLE_Z      2
+#define TGSI_SWIZZLE_W      3
+
+/**
+ * Precision controls the precision at which the operation should be executed.
+ *
+ * CondDstUpdate enables condition code register writes. When this field is
+ * TRUE, CondDstIndex specifies the index of the condition code register to
+ * update.
+ *
+ * CondFlowEnable enables conditional execution of the operation. When this
+ * field is TRUE, CondFlowIndex specifies the index of the condition code
+ * register to test against CondMask with component swizzle controled by
+ * CondSwizzleX, CondSwizzleY, CondSwizzleZ and CondSwizzleW. If the test fails,
+ * the operation is not executed.
+ */
+
+struct tgsi_instruction_ext_nv
+{
+   unsigned Type             : 4;    /* TGSI_INSTRUCTION_EXT_TYPE_NV */
+   unsigned Precision        : 4;    /* TGSI_PRECISION_ */
+   unsigned CondDstIndex     : 4;    /* UINT */
+   unsigned CondFlowIndex    : 4;    /* UINT */
+   unsigned CondMask         : 4;    /* TGSI_CC_ */
+   unsigned CondSwizzleX     : 2;    /* TGSI_SWIZZLE_ */
+   unsigned CondSwizzleY     : 2;    /* TGSI_SWIZZLE_ */
+   unsigned CondSwizzleZ     : 2;    /* TGSI_SWIZZLE_ */
+   unsigned CondSwizzleW     : 2;    /* TGSI_SWIZZLE_ */
+   unsigned CondDstUpdate    : 1;    /* BOOL */
+   unsigned CondFlowEnable   : 1;    /* BOOL */
+   unsigned Padding          : 1;
+   unsigned Extended         : 1;    /* BOOL */
+};
+
+struct tgsi_instruction_ext_label
+{
+   unsigned Type     : 4;    /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */
+   unsigned Label    : 24;   /* UINT */
+   unsigned Padding  : 3;
+   unsigned Extended : 1;    /* BOOL */
+};
+
+#define TGSI_TEXTURE_UNKNOWN        0
+#define TGSI_TEXTURE_1D             1
+#define TGSI_TEXTURE_2D             2
+#define TGSI_TEXTURE_3D             3
+#define TGSI_TEXTURE_CUBE           4
+#define TGSI_TEXTURE_RECT           5
+#define TGSI_TEXTURE_SHADOW1D       6
+#define TGSI_TEXTURE_SHADOW2D       7
+#define TGSI_TEXTURE_SHADOWRECT     8
+#define TGSI_TEXTURE_COUNT          9
+
+struct tgsi_instruction_ext_texture
+{
+   unsigned Type     : 4;    /* TGSI_INSTRUCTION_EXT_TYPE_TEXTURE */
+   unsigned Texture  : 8;    /* TGSI_TEXTURE_ */
+   unsigned Padding  : 19;
+   unsigned Extended : 1;    /* BOOL */
+};
+
+struct tgsi_instruction_ext_predicate
+{
+   unsigned Type             : 4;    /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */
+   unsigned PredDstIndex     : 4;    /* UINT */
+   unsigned PredWriteMask    : 4;    /* TGSI_WRITEMASK_ */
+   unsigned Padding          : 19;
+   unsigned Extended         : 1;    /* BOOL */
+};
+
+/**
+ * File specifies the register array to access.
+ *
+ * Index specifies the element number of a register in the register file.
+ *
+ * If Indirect is TRUE, Index should be offset by the X component of a source
+ * register that follows. The register can be now fetched into local storage
+ * for further processing.
+ *
+ * If Negate is TRUE, all components of the fetched register are negated.
+ *
+ * The fetched register components are swizzled according to SwizzleX, SwizzleY,
+ * SwizzleZ and SwizzleW.
+ *
+ * If Extended is TRUE, any further modifications to the source register are
+ * made to this temporary storage.
+ */
+
+struct tgsi_src_register
+{
+   unsigned File        : 4;  /* TGSI_FILE_ */
+   unsigned SwizzleX    : 2;  /* TGSI_SWIZZLE_ */
+   unsigned SwizzleY    : 2;  /* TGSI_SWIZZLE_ */
+   unsigned SwizzleZ    : 2;  /* TGSI_SWIZZLE_ */
+   unsigned SwizzleW    : 2;  /* TGSI_SWIZZLE_ */
+   unsigned Negate      : 1;  /* BOOL */
+   unsigned Indirect    : 1;  /* BOOL */
+   unsigned Dimension   : 1;  /* BOOL */
+   int      Index       : 16; /* SINT */
+   unsigned Extended    : 1;  /* BOOL */
+};
+
+/**
+ * If tgsi_src_register::Extended is TRUE, tgsi_src_register_ext follows.
+ * 
+ * Then, if tgsi_src_register::Indirect is TRUE, another tgsi_src_register
+ * follows.
+ *
+ * Then, if tgsi_src_register::Dimension is TRUE, tgsi_dimension follows.
+ */
+
+#define TGSI_SRC_REGISTER_EXT_TYPE_SWZ      0
+#define TGSI_SRC_REGISTER_EXT_TYPE_MOD      1
+
+struct tgsi_src_register_ext
+{
+   unsigned Type     : 4;    /* TGSI_SRC_REGISTER_EXT_TYPE_ */
+   unsigned Padding  : 27;
+   unsigned Extended : 1;    /* BOOL */
+};
+
+/**
+ * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_SWZ,
+ * it should be cast to tgsi_src_register_ext_swz.
+ * 
+ * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_MOD,
+ * it should be cast to tgsi_src_register_ext_mod.
+ * 
+ * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext
+ * follows.
+ */
+
+#define TGSI_EXTSWIZZLE_X       TGSI_SWIZZLE_X
+#define TGSI_EXTSWIZZLE_Y       TGSI_SWIZZLE_Y
+#define TGSI_EXTSWIZZLE_Z       TGSI_SWIZZLE_Z
+#define TGSI_EXTSWIZZLE_W       TGSI_SWIZZLE_W
+#define TGSI_EXTSWIZZLE_ZERO    4
+#define TGSI_EXTSWIZZLE_ONE     5
+
+/**
+ * ExtSwizzleX, ExtSwizzleY, ExtSwizzleZ and ExtSwizzleW swizzle the source
+ * register in an extended manner.
+ *
+ * NegateX, NegateY, NegateZ and NegateW negate individual components of the
+ * source register.
+ *
+ * NOTE: To simplify matter, if this token is present, the corresponding Swizzle
+ *       and Negate fields in tgsi_src_register should be set to X,Y,Z,W
+ *       and FALSE, respectively.
+ */
+
+struct tgsi_src_register_ext_swz
+{
+   unsigned Type         : 4;    /* TGSI_SRC_REGISTER_EXT_TYPE_SWZ */
+   unsigned ExtSwizzleX  : 4;    /* TGSI_EXTSWIZZLE_ */
+   unsigned ExtSwizzleY  : 4;    /* TGSI_EXTSWIZZLE_ */
+   unsigned ExtSwizzleZ  : 4;    /* TGSI_EXTSWIZZLE_ */
+   unsigned ExtSwizzleW  : 4;    /* TGSI_EXTSWIZZLE_ */
+   unsigned NegateX      : 1;    /* BOOL */
+   unsigned NegateY      : 1;    /* BOOL */
+   unsigned NegateZ      : 1;    /* BOOL */
+   unsigned NegateW      : 1;    /* BOOL */
+   unsigned Padding      : 7;
+   unsigned Extended     : 1;    /* BOOL */
+};
+
+/**
+ * Extra src register modifiers
+ *
+ * If Complement is TRUE, the source register is modified by subtracting it
+ * from 1.0.
+ *
+ * If Bias is TRUE, the source register is modified by subtracting 0.5 from it.
+ *
+ * If Scale2X is TRUE, the source register is modified by multiplying it by 2.0.
+ *
+ * If Absolute is TRUE, the source register is modified by removing the sign.
+ *
+ * If Negate is TRUE, the source register is modified by negating it.
+ */
+
+struct tgsi_src_register_ext_mod
+{
+   unsigned Type         : 4;    /* TGSI_SRC_REGISTER_EXT_TYPE_MOD */
+   unsigned Complement   : 1;    /* BOOL */
+   unsigned Bias         : 1;    /* BOOL */
+   unsigned Scale2X      : 1;    /* BOOL */
+   unsigned Absolute     : 1;    /* BOOL */
+   unsigned Negate       : 1;    /* BOOL */
+   unsigned Padding      : 22;
+   unsigned Extended     : 1;    /* BOOL */
+};
+
+struct tgsi_dimension
+{
+   unsigned Indirect    : 1;  /* BOOL */
+   unsigned Dimension   : 1;  /* BOOL */
+   unsigned Padding     : 13;
+   int      Index       : 16; /* SINT */
+   unsigned Extended    : 1;  /* BOOL */
+};
+
+struct tgsi_dst_register
+{
+   unsigned File        : 4;  /* TGSI_FILE_ */
+   unsigned WriteMask   : 4;  /* TGSI_WRITEMASK_ */
+   unsigned Indirect    : 1;  /* BOOL */
+   unsigned Dimension   : 1;  /* BOOL */
+   int      Index       : 16; /* SINT */
+   unsigned Padding     : 5;
+   unsigned Extended    : 1;  /* BOOL */
+};
+
+/*
+ * If tgsi_dst_register::Extended is TRUE, tgsi_dst_register_ext follows.
+ * 
+ * Then, if tgsi_dst_register::Indirect is TRUE, tgsi_src_register follows.
+ */
+
+#define TGSI_DST_REGISTER_EXT_TYPE_CONDCODE     0
+#define TGSI_DST_REGISTER_EXT_TYPE_MODULATE     1
+#define TGSI_DST_REGISTER_EXT_TYPE_PREDICATE    2
+
+struct tgsi_dst_register_ext
+{
+   unsigned Type     : 4;    /* TGSI_DST_REGISTER_EXT_TYPE_ */
+   unsigned Padding  : 27;
+   unsigned Extended : 1;    /* BOOL */
+};
+
+/**
+ * Extra destination register modifiers
+ *
+ * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_CONDCODE,
+ * it should be cast to tgsi_dst_register_ext_condcode.
+ * 
+ * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_MODULATE,
+ * it should be cast to tgsi_dst_register_ext_modulate.
+ * 
+ * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_PREDICATE,
+ * it should be cast to tgsi_dst_register_ext_predicate.
+ * 
+ * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext
+ * follows.
+ */
+struct tgsi_dst_register_ext_concode
+{
+   unsigned Type         : 4;    /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */
+   unsigned CondMask     : 4;    /* TGSI_CC_ */
+   unsigned CondSwizzleX : 2;    /* TGSI_SWIZZLE_ */
+   unsigned CondSwizzleY : 2;    /* TGSI_SWIZZLE_ */
+   unsigned CondSwizzleZ : 2;    /* TGSI_SWIZZLE_ */
+   unsigned CondSwizzleW : 2;    /* TGSI_SWIZZLE_ */
+   unsigned CondSrcIndex : 4;    /* UINT */
+   unsigned Padding      : 11;
+   unsigned Extended     : 1;    /* BOOL */
+};
+
+#define TGSI_MODULATE_1X        0
+#define TGSI_MODULATE_2X        1
+#define TGSI_MODULATE_4X        2
+#define TGSI_MODULATE_8X        3
+#define TGSI_MODULATE_HALF      4
+#define TGSI_MODULATE_QUARTER   5
+#define TGSI_MODULATE_EIGHTH    6
+#define TGSI_MODULATE_COUNT     7
+
+struct tgsi_dst_register_ext_modulate
+{
+   unsigned Type     : 4;    /* TGSI_DST_REGISTER_EXT_TYPE_MODULATE */
+   unsigned Modulate : 4;    /* TGSI_MODULATE_ */
+   unsigned Padding  : 23;
+   unsigned Extended : 1;    /* BOOL */
+};
+
+/*
+ * Currently, the following constraints apply.
+ *
+ * - PredSwizzleXYZW is either set to identity or replicate.
+ * - PredSrcIndex is 0.
+ */
+
+struct tgsi_dst_register_ext_predicate
+{
+   unsigned Type         : 4;    /* TGSI_DST_REGISTER_EXT_TYPE_PREDICATE */
+   unsigned PredSwizzleX : 2;    /* TGSI_SWIZZLE_ */
+   unsigned PredSwizzleY : 2;    /* TGSI_SWIZZLE_ */
+   unsigned PredSwizzleZ : 2;    /* TGSI_SWIZZLE_ */
+   unsigned PredSwizzleW : 2;    /* TGSI_SWIZZLE_ */
+   unsigned PredSrcIndex : 4;    /* UINT */
+   unsigned Negate       : 1;    /* BOOL */
+   unsigned Padding      : 14;
+   unsigned Extended     : 1;    /* BOOL */
+};
+
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* TGSI_TOKEN_H */
+
diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h
new file mode 100644
index 0000000000..9dc541630c
--- /dev/null
+++ b/src/gallium/include/pipe/p_state.h
@@ -0,0 +1,366 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * @file
+ * 
+ * Abstract graphics pipe state objects.
+ *
+ * Basic notes:
+ *   1. Want compact representations, so we use bitfields.
+ *   2. Put bitfields before other (GLfloat) fields.
+ */
+
+
+#ifndef PIPE_STATE_H
+#define PIPE_STATE_H
+
+#include "p_compiler.h"
+#include "p_defines.h"
+#include "p_format.h"
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/**
+ * Implementation limits
+ */
+#define PIPE_MAX_ATTRIBS          32
+#define PIPE_MAX_CLIP_PLANES       6
+#define PIPE_MAX_COLOR_BUFS        8
+#define PIPE_MAX_CONSTANT         32
+#define PIPE_MAX_SAMPLERS         16
+#define PIPE_MAX_SHADER_INPUTS    16
+#define PIPE_MAX_SHADER_OUTPUTS   16
+#define PIPE_MAX_TEXTURE_LEVELS   16
+
+
+/* fwd decls */
+struct pipe_screen;
+struct pipe_surface;
+
+
+
+/**
+ * The driver will certainly subclass this to include actual memory
+ * management information.
+ */
+struct pipe_buffer
+{
+   unsigned alignment;
+   unsigned usage;
+   unsigned size;
+
+   /** Reference count */
+   unsigned refcount;
+};
+
+
+/**
+ * Primitive (point/line/tri) rasterization info
+ */
+struct pipe_rasterizer_state
+{
+   unsigned flatshade:1;
+   unsigned light_twoside:1;
+   unsigned front_winding:2;  /**< PIPE_WINDING_x */
+   unsigned cull_mode:2;      /**< PIPE_WINDING_x */
+   unsigned fill_cw:2;        /**< PIPE_POLYGON_MODE_x */
+   unsigned fill_ccw:2;       /**< PIPE_POLYGON_MODE_x */
+   unsigned offset_cw:1;
+   unsigned offset_ccw:1;
+   unsigned scissor:1;
+   unsigned poly_smooth:1;
+   unsigned poly_stipple_enable:1;
+   unsigned point_smooth:1;
+   unsigned point_sprite:1;
+   unsigned point_size_per_vertex:1; /**< size computed in vertex shader */
+   unsigned multisample:1;         /* XXX maybe more ms state in future */
+   unsigned line_smooth:1;
+   unsigned line_stipple_enable:1;
+   unsigned line_stipple_factor:8;  /**< [1..256] actually */
+   unsigned line_stipple_pattern:16;
+   unsigned line_last_pixel:1;
+   unsigned bypass_clipping:1;
+   unsigned bypass_vs:1; /**< Skip the vertex shader.  Note that the shader is
+                            still needed though, to indicate inputs/outputs */
+   unsigned origin_lower_left:1;  /**< Is (0,0) the lower-left corner? */
+   unsigned flatshade_first:1;   /**< take color attribute from the first vertex of a primitive */
+   unsigned gl_rasterization_rules:1; /**< enable tweaks for GL rasterization?  */
+
+   float line_width;
+   float point_size;           /**< used when no per-vertex size */
+   float point_size_min;        /* XXX - temporary, will go away */
+   float point_size_max;        /* XXX - temporary, will go away */
+   float offset_units;
+   float offset_scale;
+   ubyte sprite_coord_mode[PIPE_MAX_SHADER_OUTPUTS]; /**< PIPE_SPRITE_COORD_ */
+};
+
+
+struct pipe_poly_stipple
+{
+   unsigned stipple[32];
+};
+
+
+struct pipe_viewport_state
+{
+   float scale[4];
+   float translate[4];
+};
+
+
+struct pipe_scissor_state
+{
+   unsigned minx:16;
+   unsigned miny:16;
+   unsigned maxx:16;
+   unsigned maxy:16;
+};
+
+
+struct pipe_clip_state
+{
+   float ucp[PIPE_MAX_CLIP_PLANES][4];
+   unsigned nr;
+};
+
+
+/**
+ * Constants for vertex/fragment shaders
+ */
+struct pipe_constant_buffer
+{
+   struct pipe_buffer *buffer;
+};
+
+
+struct pipe_shader_state
+{
+   const struct tgsi_token *tokens;
+};
+
+
+struct pipe_depth_state 
+{
+   unsigned enabled:1;         /**< depth test enabled? */
+   unsigned writemask:1;       /**< allow depth buffer writes? */
+   unsigned func:3;            /**< depth test func (PIPE_FUNC_x) */
+   unsigned occlusion_count:1; /**< do occlusion counting? */
+};
+
+
+struct pipe_stencil_state
+{
+   unsigned enabled:1;  /**< stencil[0]: stencil enabled, stencil[1]: two-side enabled */
+   unsigned func:3;     /**< PIPE_FUNC_x */
+   unsigned fail_op:3;  /**< PIPE_STENCIL_OP_x */
+   unsigned zpass_op:3; /**< PIPE_STENCIL_OP_x */
+   unsigned zfail_op:3; /**< PIPE_STENCIL_OP_x */
+   ubyte ref_value;
+   ubyte valuemask;
+   ubyte writemask;
+};
+
+
+struct pipe_alpha_state
+{
+   unsigned enabled:1;
+   unsigned func:3;     /**< PIPE_FUNC_x */
+   float ref_value;     /**< reference value */
+};
+
+
+struct pipe_depth_stencil_alpha_state
+{
+   struct pipe_depth_state depth;
+   struct pipe_stencil_state stencil[2]; /**< [0] = front, [1] = back */
+   struct pipe_alpha_state alpha;
+};
+
+
+struct pipe_blend_state
+{
+   unsigned blend_enable:1;
+
+   unsigned rgb_func:3;          /**< PIPE_BLEND_x */
+   unsigned rgb_src_factor:5;    /**< PIPE_BLENDFACTOR_x */
+   unsigned rgb_dst_factor:5;    /**< PIPE_BLENDFACTOR_x */
+
+   unsigned alpha_func:3;        /**< PIPE_BLEND_x */
+   unsigned alpha_src_factor:5;  /**< PIPE_BLENDFACTOR_x */
+   unsigned alpha_dst_factor:5;  /**< PIPE_BLENDFACTOR_x */
+
+   unsigned logicop_enable:1;
+   unsigned logicop_func:4;      /**< PIPE_LOGICOP_x */
+
+   unsigned colormask:4;         /**< bitmask of PIPE_MASK_R/G/B/A */
+   unsigned dither:1;
+};
+
+
+struct pipe_blend_color
+{
+   float color[4];
+};
+
+
+struct pipe_framebuffer_state
+{
+   unsigned width, height;
+
+   /** multiple colorbuffers for multiple render targets */
+   unsigned nr_cbufs;
+   struct pipe_surface *cbufs[PIPE_MAX_COLOR_BUFS];
+
+   struct pipe_surface *zsbuf;      /**< Z/stencil buffer */
+};
+
+
+/**
+ * Texture sampler state.
+ */
+struct pipe_sampler_state
+{
+   unsigned wrap_s:3;            /**< PIPE_TEX_WRAP_x */
+   unsigned wrap_t:3;            /**< PIPE_TEX_WRAP_x */
+   unsigned wrap_r:3;            /**< PIPE_TEX_WRAP_x */
+   unsigned min_img_filter:2;    /**< PIPE_TEX_FILTER_x */
+   unsigned min_mip_filter:2;    /**< PIPE_TEX_MIPFILTER_x */
+   unsigned mag_img_filter:2;    /**< PIPE_TEX_FILTER_x */
+   unsigned compare_mode:1;      /**< PIPE_TEX_COMPARE_x */
+   unsigned compare_func:3;      /**< PIPE_FUNC_x */
+   unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */
+   unsigned prefilter:4;         /**< Wierd sampling state exposed by some api's */
+   float shadow_ambient;         /**< shadow test fail color/intensity */
+   float lod_bias;               /**< LOD/lambda bias */
+   float min_lod, max_lod;       /**< LOD clamp range, after bias */
+   float border_color[4];
+   float max_anisotropy;
+};
+
+
+/**
+ * 2D surface.  This is basically a view into a memory buffer.
+ * May be a renderbuffer, texture mipmap level, etc.
+ */
+struct pipe_surface
+{
+   enum pipe_format format;      /**< PIPE_FORMAT_x */
+   unsigned status;              /**< PIPE_SURFACE_STATUS_x */
+   unsigned clear_value;         /**< XXX may be temporary */
+   unsigned width;               /**< logical width in pixels */
+   unsigned height;              /**< logical height in pixels */
+   struct pipe_format_block block;
+   unsigned nblocksx;            /**< allocated width in blocks */
+   unsigned nblocksy;            /**< allocated height in blocks */
+   unsigned stride;              /**< stride in bytes between rows of blocks */
+   unsigned layout;              /**< PIPE_SURFACE_LAYOUT_x */
+   unsigned offset;              /**< offset from start of buffer, in bytes */
+   unsigned refcount;
+   unsigned usage;               /**< PIPE_BUFFER_USAGE_*  */
+
+   struct pipe_texture *texture; /**< texture into which this is a view  */
+   unsigned face;
+   unsigned level;
+   unsigned zslice;
+};
+
+
+/**
+ * Texture object.
+ */
+struct pipe_texture
+{ 
+   enum pipe_texture_target target; /**< PIPE_TEXTURE_x */
+   enum pipe_format format;         /**< PIPE_FORMAT_x */
+
+   unsigned width[PIPE_MAX_TEXTURE_LEVELS];
+   unsigned height[PIPE_MAX_TEXTURE_LEVELS];
+   unsigned depth[PIPE_MAX_TEXTURE_LEVELS];
+
+   struct pipe_format_block block;
+   unsigned nblocksx[PIPE_MAX_TEXTURE_LEVELS]; /**< allocated width in blocks */
+   unsigned nblocksy[PIPE_MAX_TEXTURE_LEVELS]; /**< allocated height in blocks */
+
+   unsigned last_level:8;    /**< Index of last mipmap level present/defined */
+   unsigned compressed:1;
+
+   unsigned nr_samples:8;    /**< for multisampled surfaces, nr of samples */
+
+   unsigned tex_usage;       /* PIPE_TEXTURE_USAGE_* */
+
+   /* These are also refcounted:
+    */
+   unsigned refcount;
+
+   struct pipe_screen *screen; /**< screen that this texture belongs to */
+};
+
+
+/**
+ * A vertex buffer.  Typically, all the vertex data/attributes for
+ * drawing something will be in one buffer.  But it's also possible, for
+ * example, to put colors in one buffer and texcoords in another.
+ */
+struct pipe_vertex_buffer
+{
+   unsigned stride;    /**< stride to same attrib in next vertex, in bytes */
+   unsigned max_index;   /**< number of vertices in this buffer */
+   unsigned buffer_offset;  /**< offset to start of data in buffer, in bytes */
+   struct pipe_buffer *buffer;  /**< the actual buffer */
+};
+
+
+/**
+ * Information to describe a vertex attribute (position, color, etc)
+ */
+struct pipe_vertex_element
+{
+   /** Offset of this attribute, in bytes, from the start of the vertex */
+   unsigned src_offset;
+
+   /** Which vertex_buffer (as given to pipe->set_vertex_buffer()) does
+    * this attribute live in?
+    */
+   unsigned vertex_buffer_index:8;
+   unsigned nr_components:8;
+ 
+   enum pipe_format src_format; 	   /**< PIPE_FORMAT_* */
+};
+
+
+#ifdef __cplusplus
+}
+#endif
+   
+#endif
diff --git a/src/gallium/include/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h
new file mode 100644
index 0000000000..8af3cd958b
--- /dev/null
+++ b/src/gallium/include/pipe/p_thread.h
@@ -0,0 +1,276 @@
+/**************************************************************************
+ * 
+ * Copyright 1999-2006 Brian Paul
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * @file
+ * 
+ * Thread, mutex, condition var and thread-specific data functions.
+ */
+
+
+#ifndef _P_THREAD2_H_
+#define _P_THREAD2_H_
+
+
+#include "pipe/p_compiler.h"
+
+
+#if defined(PIPE_OS_LINUX)
+
+#include <pthread.h> /* POSIX threads headers */
+#include <stdio.h> /* for perror() */
+
+typedef pthread_t pipe_thread;
+
+#define PIPE_THREAD_ROUTINE( name, param ) \
+   void *name( void *param )
+
+static INLINE pipe_thread pipe_thread_create( void *(* routine)( void *), void *param )
+{
+   pipe_thread thread;
+   if (pthread_create( &thread, NULL, routine, param ))
+      return 0;
+   return thread;
+}
+
+static INLINE int pipe_thread_wait( pipe_thread thread )
+{
+   return pthread_join( thread, NULL );
+}
+
+static INLINE int pipe_thread_destroy( pipe_thread thread )
+{
+   return pthread_detach( thread );
+}
+
+typedef pthread_mutex_t pipe_mutex;
+typedef pthread_cond_t pipe_condvar;
+
+#define pipe_static_mutex(mutex) \
+   static pipe_mutex mutex = PTHREAD_MUTEX_INITIALIZER
+
+#define pipe_mutex_init(mutex) \
+   pthread_mutex_init(&(mutex), NULL)
+
+#define pipe_mutex_destroy(mutex) \
+   pthread_mutex_destroy(&(mutex))
+
+#define pipe_mutex_lock(mutex) \
+   (void) pthread_mutex_lock(&(mutex))
+
+#define pipe_mutex_unlock(mutex) \
+   (void) pthread_mutex_unlock(&(mutex))
+
+#define pipe_static_condvar(mutex) \
+   static pipe_condvar mutex = PTHREAD_COND_INITIALIZER
+
+#define pipe_condvar_init(cond)	\
+   pthread_cond_init(&(cond), NULL)
+
+#define pipe_condvar_destroy(cond) \
+   pthread_cond_destroy(&(cond))
+
+#define pipe_condvar_wait(cond, mutex) \
+  pthread_cond_wait(&(cond), &(mutex))
+
+#define pipe_condvar_signal(cond) \
+  pthread_cond_signal(&(cond))
+
+#define pipe_condvar_broadcast(cond) \
+  pthread_cond_broadcast(&(cond))
+
+
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+
+#include <windows.h>
+
+typedef HANDLE pipe_thread;
+
+#define PIPE_THREAD_ROUTINE( name, param ) \
+   void * WINAPI name( void *param )
+
+static INLINE pipe_thread pipe_thread_create( void *(WINAPI * routine)( void *), void *param )
+{
+   DWORD id;
+   return CreateThread( NULL, 0, (LPTHREAD_START_ROUTINE) routine, param, 0, &id );
+}
+
+static INLINE int pipe_thread_wait( pipe_thread thread )
+{
+   if (WaitForSingleObject( thread, INFINITE ) == WAIT_OBJECT_0)
+      return 0;
+   return -1;
+}
+
+static INLINE int pipe_thread_destroy( pipe_thread thread )
+{
+   if (CloseHandle( thread ))
+      return 0;
+   return -1;
+}
+
+typedef CRITICAL_SECTION pipe_mutex;
+
+#define pipe_static_mutex(name) \
+   /*static*/ pipe_mutex name = {0,0,0,0,0,0}
+
+#define pipe_mutex_init(name) \
+   InitializeCriticalSection(&name)
+
+#define pipe_mutex_destroy(name) \
+   DeleteCriticalSection(&name)
+
+#define pipe_mutex_lock(name) \
+   EnterCriticalSection(&name)
+
+#define pipe_mutex_unlock(name) \
+   LeaveCriticalSection(&name)
+
+/* XXX: dummy definitions, make it compile */
+
+typedef unsigned pipe_condvar;
+
+#define pipe_condvar_init(condvar) \
+   (void) condvar
+
+#define pipe_condvar_broadcast(condvar) \
+   (void) condvar
+
+#else
+
+/** Dummy definitions */
+
+typedef unsigned pipe_thread;
+typedef unsigned pipe_mutex;
+typedef unsigned pipe_condvar;
+
+#define pipe_static_mutex(mutex) \
+   static pipe_mutex mutex = 0
+
+#define pipe_mutex_init(mutex) \
+   (void) mutex
+
+#define pipe_mutex_destroy(mutex) \
+   (void) mutex
+
+#define pipe_mutex_lock(mutex) \
+   (void) mutex
+
+#define pipe_mutex_unlock(mutex) \
+   (void) mutex
+
+#define pipe_static_condvar(condvar) \
+   static unsigned condvar = 0
+
+#define pipe_condvar_init(condvar) \
+   (void) condvar
+
+#define pipe_condvar_destroy(condvar) \
+   (void) condvar
+
+#define pipe_condvar_wait(condvar, mutex) \
+   (void) condvar
+
+#define pipe_condvar_signal(condvar) \
+   (void) condvar
+
+#define pipe_condvar_broadcast(condvar) \
+   (void) condvar
+
+
+#endif  /* PIPE_OS_? */
+
+
+
+/*
+ * Thread-specific data.
+ */
+
+typedef struct {
+#if defined(PIPE_OS_LINUX)
+   pthread_key_t key;
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   DWORD key;
+#endif
+   int initMagic;
+} pipe_tsd;
+
+
+#define PIPE_TSD_INIT_MAGIC 0xff8adc98
+
+
+static INLINE void
+pipe_tsd_init(pipe_tsd *tsd)
+{
+#if defined(PIPE_OS_LINUX)
+   if (pthread_key_create(&tsd->key, NULL/*free*/) != 0) {
+      perror("pthread_key_create(): failed to allocate key for thread specific data");
+      exit(-1);
+   }
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   assert(0);
+#endif
+   tsd->initMagic = PIPE_TSD_INIT_MAGIC;
+}
+
+static INLINE void *
+pipe_tsd_get(pipe_tsd *tsd)
+{
+   if (tsd->initMagic != (int) PIPE_TSD_INIT_MAGIC) {
+      pipe_tsd_init(tsd);
+   }
+#if defined(PIPE_OS_LINUX)
+   return pthread_getspecific(tsd->key);
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   assert(0);
+   return NULL;
+#else
+   assert(0);
+   return NULL;
+#endif
+}
+
+static INLINE void
+pipe_tsd_set(pipe_tsd *tsd, void *value)
+{
+   if (tsd->initMagic != (int) PIPE_TSD_INIT_MAGIC) {
+      pipe_tsd_init(tsd);
+   }
+#if defined(PIPE_OS_LINUX)
+   if (pthread_setspecific(tsd->key, value) != 0) {
+      perror("pthread_set_specific() failed");
+      exit(-1);
+   }
+#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
+   assert(0);
+#else
+   assert(0);
+#endif
+}
+
+
+
+#endif /* _P_THREAD2_H_ */
diff --git a/src/gallium/include/state_tracker/drm_api.h b/src/gallium/include/state_tracker/drm_api.h
new file mode 100644
index 0000000000..54480fa047
--- /dev/null
+++ b/src/gallium/include/state_tracker/drm_api.h
@@ -0,0 +1,33 @@
+
+#ifndef _DRM_API_H_
+#define _DRM_API_H_
+
+struct pipe_screen;
+struct pipe_winsys;
+struct pipe_context;
+
+struct drm_api
+{
+	/**
+	 * Special buffer function
+	 */
+	/*@{*/
+	struct pipe_screen*  (*create_screen)(int drmFB, int pciID);
+	struct pipe_context* (*create_context)(struct pipe_screen *screen);
+	/*@}*/
+
+	/**
+	 * Special buffer function
+	 */
+	/*@{*/
+	struct pipe_buffer* (*buffer_from_handle)(struct pipe_winsys *winsys, const char *name, unsigned handle);
+	unsigned (*handle_from_buffer)(struct pipe_winsys *winsys, struct pipe_buffer *buffer);
+	/*@}*/
+};
+
+/**
+ * A driver needs to export this symbol
+ */
+extern struct drm_api drm_api_hocks;
+
+#endif
diff --git a/src/gallium/state_trackers/Makefile b/src/gallium/state_trackers/Makefile
new file mode 100644
index 0000000000..265ca468c2
--- /dev/null
+++ b/src/gallium/state_trackers/Makefile
@@ -0,0 +1,25 @@
+TOP = ../../..
+include $(TOP)/configs/current
+
+
+SUBDIRS = $(GALLIUM_STATE_TRACKERS_DIRS)
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	rm -f `find . -name \*.[oa]`
+	rm -f `find . -name depend`
+
+
+# Dummy install target
+install:
diff --git a/src/gallium/state_trackers/README b/src/gallium/state_trackers/README
new file mode 100644
index 0000000000..28dd27bbd5
--- /dev/null
+++ b/src/gallium/state_trackers/README
@@ -0,0 +1,2 @@
+This directory is a placeholder for incubating state-trackers. Mesa's
+state-tracker is in src/mesa.
diff --git a/src/gallium/state_trackers/egl/Makefile b/src/gallium/state_trackers/egl/Makefile
new file mode 100644
index 0000000000..c9124b1c90
--- /dev/null
+++ b/src/gallium/state_trackers/egl/Makefile
@@ -0,0 +1,29 @@
+TARGET     = libegldrm.a
+CFILES     = $(wildcard ./*.c)
+OBJECTS    = $(patsubst ./%.c,./%.o,$(CFILES))
+GALLIUMDIR = ../..
+TOP        = ../../../..
+
+include ${TOP}/configs/current
+
+CFLAGS += -g -Wall -Werror=implicit-function-declaration -fPIC \
+          -I${GALLIUMDIR}/include \
+          -I${GALLIUMDIR}/auxiliary \
+          -I${TOP}/src/mesa/drivers/dri/common \
+          -I${TOP}/src/mesa \
+          -I$(TOP)/include \
+          -I$(TOP)/src/egl/main \
+          $(shell pkg-config --cflags pixman-1 xorg-server) \
+          ${LIBDRM_CFLAGS}
+
+#############################################
+
+.PHONY	= all clean
+
+all: ${TARGET}
+
+${TARGET}: ${OBJECTS}
+	ar rcs $@ $^
+
+clean:
+	rm -rf ${OBJECTS} ${TARGET}
diff --git a/src/gallium/state_trackers/egl/egl_context.c b/src/gallium/state_trackers/egl/egl_context.c
new file mode 100644
index 0000000000..8564972b91
--- /dev/null
+++ b/src/gallium/state_trackers/egl/egl_context.c
@@ -0,0 +1,193 @@
+
+#include "utils.h"
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "egl_tracker.h"
+
+#include "egllog.h"
+
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+
+#include "state_tracker/st_public.h"
+#include "state_tracker/drm_api.h"
+
+#include "GL/internal/glcore.h"
+
+#define need_GL_ARB_multisample
+#define need_GL_ARB_point_parameters
+#define need_GL_ARB_texture_compression
+#define need_GL_ARB_vertex_buffer_object
+#define need_GL_ARB_vertex_program
+#define need_GL_ARB_window_pos
+#define need_GL_EXT_blend_color
+#define need_GL_EXT_blend_equation_separate
+#define need_GL_EXT_blend_func_separate
+#define need_GL_EXT_blend_minmax
+#define need_GL_EXT_cull_vertex
+#define need_GL_EXT_fog_coord
+#define need_GL_EXT_framebuffer_object
+#define need_GL_EXT_multi_draw_arrays
+#define need_GL_EXT_secondary_color
+#define need_GL_NV_vertex_program
+#include "extension_helper.h"
+
+/**
+ * TODO HACK! FUGLY!
+ * Copied for intel extentions.
+ */
+const struct dri_extension card_extensions[] = {
+	{"GL_ARB_multisample", GL_ARB_multisample_functions},
+	{"GL_ARB_multitexture", NULL},
+	{"GL_ARB_point_parameters", GL_ARB_point_parameters_functions},
+	{"GL_ARB_texture_border_clamp", NULL},
+	{"GL_ARB_texture_compression", GL_ARB_texture_compression_functions},
+	{"GL_ARB_texture_cube_map", NULL},
+	{"GL_ARB_texture_env_add", NULL},
+	{"GL_ARB_texture_env_combine", NULL},
+	{"GL_ARB_texture_env_dot3", NULL},
+	{"GL_ARB_texture_mirrored_repeat", NULL},
+	{"GL_ARB_texture_rectangle", NULL},
+	{"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions},
+	{"GL_ARB_pixel_buffer_object", NULL},
+	{"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
+	{"GL_ARB_window_pos", GL_ARB_window_pos_functions},
+	{"GL_EXT_blend_color", GL_EXT_blend_color_functions},
+	{"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions},
+	{"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
+	{"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
+	{"GL_EXT_blend_subtract", NULL},
+	{"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions},
+	{"GL_EXT_fog_coord", GL_EXT_fog_coord_functions},
+	{"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions},
+	{"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions},
+	{"GL_EXT_packed_depth_stencil", NULL},
+	{"GL_EXT_pixel_buffer_object", NULL},
+	{"GL_EXT_secondary_color", GL_EXT_secondary_color_functions},
+	{"GL_EXT_stencil_wrap", NULL},
+	{"GL_EXT_texture_edge_clamp", NULL},
+	{"GL_EXT_texture_env_combine", NULL},
+	{"GL_EXT_texture_env_dot3", NULL},
+	{"GL_EXT_texture_filter_anisotropic", NULL},
+	{"GL_EXT_texture_lod_bias", NULL},
+	{"GL_3DFX_texture_compression_FXT1", NULL},
+	{"GL_APPLE_client_storage", NULL},
+	{"GL_MESA_pack_invert", NULL},
+	{"GL_MESA_ycbcr_texture", NULL},
+	{"GL_NV_blend_square", NULL},
+	{"GL_NV_vertex_program", GL_NV_vertex_program_functions},
+	{"GL_NV_vertex_program1_1", NULL},
+	{"GL_SGIS_generate_mipmap", NULL },
+	{NULL, NULL}
+};
+
+EGLContext
+drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list)
+{
+	struct drm_device *dev = (struct drm_device *)drv;
+	struct drm_context *ctx;
+	struct drm_context *share = NULL;
+	struct st_context *st_share = NULL;
+	_EGLConfig *conf;
+	int i;
+	__GLcontextModes *visual;
+
+	conf = _eglLookupConfig(drv, dpy, config);
+	if (!conf) {
+		_eglError(EGL_BAD_CONFIG, "eglCreateContext");
+		return EGL_NO_CONTEXT;
+	}
+
+	for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
+		switch (attrib_list[i]) {
+			/* no attribs defined for now */
+			default:
+				_eglError(EGL_BAD_ATTRIBUTE, "eglCreateContext");
+				return EGL_NO_CONTEXT;
+		}
+	}
+
+	ctx = (struct drm_context *) calloc(1, sizeof(struct drm_context));
+	if (!ctx)
+		goto err_c;
+
+	_eglInitContext(drv, dpy, &ctx->base, config, attrib_list);
+
+	ctx->pipe = drm_api_hocks.create_context(dev->screen);
+	if (!ctx->pipe)
+		goto err_pipe;
+
+	if (share)
+		st_share = share->st;
+
+	visual = drm_visual_from_config(conf);
+	ctx->st = st_create_context(ctx->pipe, visual, st_share);
+	drm_visual_modes_destroy(visual);
+
+	if (!ctx->st)
+		goto err_gl;
+
+	/* generate handle and insert into hash table */
+	_eglSaveContext(&ctx->base);
+	assert(_eglGetContextHandle(&ctx->base));
+
+	return _eglGetContextHandle(&ctx->base);
+
+err_gl:
+	ctx->pipe->destroy(ctx->pipe);
+err_pipe:
+	free(ctx);
+err_c:
+	return EGL_NO_CONTEXT;
+}
+
+EGLBoolean
+drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context)
+{
+	struct drm_context *c = lookup_drm_context(context);
+	_eglRemoveContext(&c->base);
+	if (c->base.IsBound) {
+		c->base.DeletePending = EGL_TRUE;
+	} else {
+		st_destroy_context(c->st);
+		c->pipe->destroy(c->pipe);
+		free(c);
+	}
+	return EGL_TRUE;
+}
+
+EGLBoolean
+drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context)
+{
+	struct drm_surface *readSurf = lookup_drm_surface(read);
+	struct drm_surface *drawSurf = lookup_drm_surface(draw);
+	struct drm_context *ctx = lookup_drm_context(context);
+	EGLBoolean b;
+
+	b = _eglMakeCurrent(drv, dpy, draw, read, context);
+	if (!b)
+		return EGL_FALSE;
+
+	if (ctx) {
+		if (!drawSurf || !readSurf)
+			return EGL_FALSE;
+
+		drawSurf->user = ctx;
+		readSurf->user = ctx;
+
+		st_make_current(ctx->st, drawSurf->stfb, readSurf->stfb);
+
+		/* st_resize_framebuffer needs a bound context to work */
+		st_resize_framebuffer(drawSurf->stfb, drawSurf->w, drawSurf->h);
+		st_resize_framebuffer(readSurf->stfb, readSurf->w, readSurf->h);
+	} else {
+		drawSurf->user = NULL;
+		readSurf->user = NULL;
+
+		st_make_current(NULL, NULL, NULL);
+	}
+
+	return EGL_TRUE;
+}
diff --git a/src/gallium/state_trackers/egl/egl_surface.c b/src/gallium/state_trackers/egl/egl_surface.c
new file mode 100644
index 0000000000..091d437d81
--- /dev/null
+++ b/src/gallium/state_trackers/egl/egl_surface.c
@@ -0,0 +1,409 @@
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "egl_tracker.h"
+
+#include "egllog.h"
+
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_context.h"
+
+#include "state_tracker/drm_api.h"
+
+/*
+ * Util functions
+ */
+
+static struct drm_mode_modeinfo *
+drm_find_mode(drmModeConnectorPtr connector, _EGLMode *mode)
+{
+	int i;
+	struct drm_mode_modeinfo *m = NULL;
+
+	for (i = 0; i < connector->count_modes; i++) {
+		m = &connector->modes[i];
+		if (m->hdisplay == mode->Width && m->vdisplay == mode->Height && m->vrefresh == mode->RefreshRate)
+			break;
+		m = &connector->modes[0]; /* if we can't find one, return first */
+	}
+
+	return m;
+}
+
+static struct st_framebuffer *
+drm_create_framebuffer(const __GLcontextModes *visual,
+                       unsigned width,
+                       unsigned height,
+                       void *priv)
+{
+	enum pipe_format colorFormat, depthFormat, stencilFormat;
+
+	if (visual->redBits == 5)
+		colorFormat = PIPE_FORMAT_R5G6B5_UNORM;
+	else
+		colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+	if (visual->depthBits == 16)
+		depthFormat = PIPE_FORMAT_Z16_UNORM;
+	else if (visual->depthBits == 24)
+		depthFormat = PIPE_FORMAT_S8Z24_UNORM;
+	else
+		depthFormat = PIPE_FORMAT_NONE;
+
+	if (visual->stencilBits == 8)
+		stencilFormat = PIPE_FORMAT_S8Z24_UNORM;
+	else
+		stencilFormat = PIPE_FORMAT_NONE;
+
+	return st_create_framebuffer(visual,
+	                             colorFormat,
+	                             depthFormat,
+	                             stencilFormat,
+	                             width,
+	                             height,
+	                             priv);
+}
+
+static void
+drm_create_texture(_EGLDriver *drv,
+                   struct drm_screen *scrn,
+                   unsigned w, unsigned h)
+{
+	struct drm_device *dev = (struct drm_device *)drv;
+	struct pipe_screen *screen = dev->screen;
+	struct pipe_surface *surface;
+	struct pipe_texture *texture;
+	struct pipe_texture templat;
+	struct pipe_buffer *buf;
+	unsigned stride = 1024;
+	unsigned pitch = 0;
+	unsigned size = 0;
+
+	/* ugly */
+	if (stride < w)
+		stride = 2048;
+
+	pitch = stride * 4;
+	size = h * 2 * pitch;
+
+	buf = pipe_buffer_create(screen,
+	                         0, /* alignment */
+	                         PIPE_BUFFER_USAGE_GPU_READ_WRITE |
+	                         PIPE_BUFFER_USAGE_CPU_READ_WRITE,
+	                         size);
+
+	if (!buf)
+		goto err_buf;
+
+	memset(&templat, 0, sizeof(templat));
+	templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+	templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET;
+	templat.target = PIPE_TEXTURE_2D;
+	templat.last_level = 0;
+	templat.depth[0] = 1;
+	templat.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	templat.width[0] = w;
+	templat.height[0] = h;
+	pf_get_block(templat.format, &templat.block);
+
+	texture = screen->texture_blanket(dev->screen,
+	                                  &templat,
+	                                  &pitch,
+	                                  buf);
+	if (!texture)
+		goto err_tex;
+
+	surface = screen->get_tex_surface(screen,
+	                                  texture,
+	                                  0,
+	                                  0,
+	                                  0,
+	                                  PIPE_BUFFER_USAGE_GPU_WRITE);
+
+	if (!surface)
+		goto err_surf;
+
+
+	scrn->tex = texture;
+	scrn->surface = surface;
+	scrn->buffer = buf;
+	scrn->front.width = w;
+	scrn->front.height = h;
+	scrn->front.pitch = pitch;
+	scrn->front.handle = drm_api_hocks.handle_from_buffer(dev->winsys, scrn->buffer);
+	if (0)
+		goto err_handle;
+
+	return;
+
+err_handle:
+	pipe_surface_reference(&surface, NULL);
+err_surf:
+	pipe_texture_reference(&texture, NULL);
+err_tex:
+	pipe_buffer_reference(screen, &buf, NULL);
+err_buf:
+	return;
+}
+
+/*
+ * Exported functions
+ */
+
+void
+drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen)
+{
+	struct drm_device *dev = (struct drm_device *)drv;
+
+	screen->surf = NULL;
+
+	drmModeSetCrtc(
+		dev->drmFD,
+		screen->crtcID,
+		0, // FD
+		0, 0,
+		NULL, 0, // List of output ids
+		NULL);
+
+	drmModeRmFB(dev->drmFD, screen->fbID);
+	drmModeFreeFB(screen->fb);
+	screen->fb = NULL;
+
+	pipe_surface_reference(&screen->surface, NULL);
+	pipe_texture_reference(&screen->tex, NULL);
+	pipe_buffer_reference(dev->screen, &screen->buffer, NULL);
+
+	screen->shown = 0;
+}
+
+EGLSurface
+drm_create_window_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list)
+{
+	return EGL_NO_SURFACE;
+}
+
+
+EGLSurface
+drm_create_pixmap_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list)
+{
+	return EGL_NO_SURFACE;
+}
+
+
+EGLSurface
+drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+                           const EGLint *attrib_list)
+{
+	int i;
+	int width = -1;
+	int height = -1;
+	struct drm_surface *surf = NULL;
+	__GLcontextModes *visual;
+	_EGLConfig *conf;
+
+	conf = _eglLookupConfig(drv, dpy, config);
+	if (!conf) {
+		_eglError(EGL_BAD_CONFIG, "eglCreatePbufferSurface");
+		return EGL_NO_CONTEXT;
+	}
+
+	for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) {
+		switch (attrib_list[i]) {
+			case EGL_WIDTH:
+				width = attrib_list[++i];
+				break;
+			case EGL_HEIGHT:
+				height = attrib_list[++i];
+				break;
+			default:
+				_eglError(EGL_BAD_ATTRIBUTE, "eglCreatePbufferSurface");
+				return EGL_NO_SURFACE;
+		}
+	}
+
+	if (width < 1 || height < 1) {
+		_eglError(EGL_BAD_ATTRIBUTE, "eglCreatePbufferSurface");
+		return EGL_NO_SURFACE;
+	}
+
+	surf = (struct drm_surface *) calloc(1, sizeof(struct drm_surface));
+	if (!surf)
+		goto err;
+
+	if (!_eglInitSurface(drv, dpy, &surf->base, EGL_PBUFFER_BIT, config, attrib_list))
+		goto err_surf;
+
+	surf->w = width;
+	surf->h = height;
+
+	visual = drm_visual_from_config(conf);
+	surf->stfb = drm_create_framebuffer(visual,
+	                                    width,
+	                                    height,
+	                                    (void*)surf);
+	drm_visual_modes_destroy(visual);
+
+	_eglSaveSurface(&surf->base);
+	return surf->base.Handle;
+
+err_surf:
+	free(surf);
+err:
+	return EGL_NO_SURFACE;
+}
+
+EGLSurface
+drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg,
+                               const EGLint *attrib_list)
+{
+	EGLSurface surf = drm_create_pbuffer_surface(drv, dpy, cfg, attrib_list);
+
+	return surf;
+}
+
+EGLBoolean
+drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy,
+                             EGLScreenMESA screen,
+                             EGLSurface surface, EGLModeMESA m)
+{
+	struct drm_device *dev = (struct drm_device *)drv;
+	struct drm_surface *surf = lookup_drm_surface(surface);
+	struct drm_screen *scrn = lookup_drm_screen(dpy, screen);
+	struct pipe_context *pipe;
+	_EGLMode *mode = _eglLookupMode(dpy, m);
+	int ret;
+	unsigned int i, k;
+
+	if (scrn->shown)
+		drm_takedown_shown_screen(drv, scrn);
+
+
+	drm_create_texture(drv, scrn, mode->Width, mode->Height);
+	if (!scrn->buffer)
+		return EGL_FALSE;
+
+	ret = drmModeAddFB(dev->drmFD,
+	                   scrn->front.width, scrn->front.height,
+	                   32, 32, scrn->front.pitch,
+	                   scrn->front.handle,
+	                   &scrn->fbID);
+
+	if (ret)
+		goto err_bo;
+
+	scrn->fb = drmModeGetFB(dev->drmFD, scrn->fbID);
+	if (!scrn->fb)
+		goto err_bo;
+
+	/* find a fitting crtc */
+	{
+		drmModeConnector *con = scrn->connector;
+
+		scrn->mode = drm_find_mode(con, mode);
+		if (!scrn->mode)
+			goto err_fb;
+
+		for (k = 0; k < con->count_encoders; k++) {
+			drmModeEncoder *enc = drmModeGetEncoder(dev->drmFD, con->encoders[k]);
+			for (i = 0; i < dev->res->count_crtcs; i++) {
+				if (enc->possible_crtcs & (1<<i)) {
+					/* save the ID */
+					scrn->crtcID = dev->res->crtcs[i];
+
+					/* skip the rest */
+					i = dev->res->count_crtcs;
+					k = dev->res->count_encoders;
+				}
+			}
+			drmModeFreeEncoder(enc);
+		}
+	}
+
+	ret = drmModeSetCrtc(dev->drmFD,
+	                     scrn->crtcID,
+	                     scrn->fbID,
+	                     0, 0,
+	                     &scrn->connectorID, 1,
+	                     scrn->mode);
+
+	if (ret)
+		goto err_crtc;
+
+	surf->screen = scrn;
+
+	scrn->surf = surf;
+	scrn->shown = 1;
+
+	return EGL_TRUE;
+
+err_crtc:
+	scrn->crtcID = 0;
+
+err_fb:
+	drmModeRmFB(dev->drmFD, scrn->fbID);
+	drmModeFreeFB(scrn->fb);
+	scrn->fb = NULL;
+
+err_bo:
+	pipe_surface_reference(&scrn->surface, NULL);
+	pipe_texture_reference(&scrn->tex, NULL);
+	pipe_buffer_reference(dev->screen, &scrn->buffer, NULL);
+
+	return EGL_FALSE;
+}
+
+EGLBoolean
+drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
+{
+	struct drm_surface *surf = lookup_drm_surface(surface);
+	_eglRemoveSurface(&surf->base);
+	if (surf->base.IsBound) {
+		surf->base.DeletePending = EGL_TRUE;
+	} else {
+		if (surf->screen)
+			drm_takedown_shown_screen(drv, surf->screen);
+		st_unreference_framebuffer(surf->stfb);
+		free(surf);
+	}
+	return EGL_TRUE;
+}
+
+EGLBoolean
+drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
+{
+	struct drm_surface *surf = lookup_drm_surface(draw);
+	struct pipe_surface *back_surf;
+
+	if (!surf)
+		return EGL_FALSE;
+
+	/* error checking */
+	if (!_eglSwapBuffers(drv, dpy, draw))
+		return EGL_FALSE;
+
+	st_get_framebuffer_surface(surf->stfb, ST_SURFACE_BACK_LEFT, &back_surf);
+
+	if (back_surf) {
+
+		st_notify_swapbuffers(surf->stfb);
+
+		if (surf->screen) {
+			surf->user->pipe->flush(surf->user->pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE, NULL);
+			surf->user->pipe->surface_copy(surf->user->pipe,
+				0,
+				surf->screen->surface,
+				0, 0,
+				back_surf,
+				0, 0,
+				surf->w, surf->h);
+			surf->user->pipe->flush(surf->user->pipe, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE, NULL);
+			/* TODO stuff here */
+		}
+
+		st_notify_swapbuffers_complete(surf->stfb);
+	}
+
+	return EGL_TRUE;
+}
diff --git a/src/gallium/state_trackers/egl/egl_tracker.c b/src/gallium/state_trackers/egl/egl_tracker.c
new file mode 100644
index 0000000000..dec82c3a00
--- /dev/null
+++ b/src/gallium/state_trackers/egl/egl_tracker.c
@@ -0,0 +1,217 @@
+
+#include "utils.h"
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "egl_tracker.h"
+
+#include "egllog.h"
+#include "state_tracker/drm_api.h"
+
+#include "pipe/p_screen.h"
+#include "pipe/internal/p_winsys_screen.h"
+
+/** HACK */
+void* driDriverAPI;
+extern const struct dri_extension card_extensions[];
+
+
+/*
+ * Exported functions
+ */
+
+/**
+ * The bootstrap function.  Return a new drm_driver object and
+ * plug in API functions.
+ */
+_EGLDriver *
+_eglMain(_EGLDisplay *dpy, const char *args)
+{
+	struct drm_device *drm;
+
+	drm = (struct drm_device *) calloc(1, sizeof(struct drm_device));
+	if (!drm) {
+		return NULL;
+	}
+
+	/* First fill in the dispatch table with defaults */
+	_eglInitDriverFallbacks(&drm->base);
+	/* then plug in our Drm-specific functions */
+	drm->base.API.Initialize = drm_initialize;
+	drm->base.API.Terminate = drm_terminate;
+	drm->base.API.CreateContext = drm_create_context;
+	drm->base.API.MakeCurrent = drm_make_current;
+	drm->base.API.CreateWindowSurface = drm_create_window_surface;
+	drm->base.API.CreatePixmapSurface = drm_create_pixmap_surface;
+	drm->base.API.CreatePbufferSurface = drm_create_pbuffer_surface;
+	drm->base.API.DestroySurface = drm_destroy_surface;
+	drm->base.API.DestroyContext = drm_destroy_context;
+	drm->base.API.CreateScreenSurfaceMESA = drm_create_screen_surface_mesa;
+	drm->base.API.ShowScreenSurfaceMESA = drm_show_screen_surface_mesa;
+	drm->base.API.SwapBuffers = drm_swap_buffers;
+
+	drm->base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/;
+	drm->base.Name = "DRM/Gallium/Win";
+
+	/* enable supported extensions */
+	drm->base.Extensions.MESA_screen_surface = EGL_TRUE;
+	drm->base.Extensions.MESA_copy_context = EGL_TRUE;
+
+	return &drm->base;
+}
+
+static void
+drm_get_device_id(struct drm_device *device)
+{
+	char path[512];
+	FILE *file;
+
+	/* TODO get the real minor */
+	int minor = 0;
+
+	snprintf(path, sizeof(path), "/sys/class/drm/card%d/device/device", minor);
+	file = fopen(path, "r");
+	if (!file) {
+		_eglLog(_EGL_WARNING, "Could not retrive device ID\n");
+		return;
+	}
+
+	fgets(path, sizeof( path ), file);
+	sscanf(path, "%x", &device->deviceID);
+	fclose(file);
+}
+
+static void
+drm_update_res(struct drm_device *dev)
+{
+	drmModeFreeResources(dev->res);
+	dev->res = drmModeGetResources(dev->drmFD);
+}
+
+static void
+drm_add_modes_from_connector(_EGLScreen *screen, drmModeConnectorPtr connector)
+{
+	struct drm_mode_modeinfo *m;
+	int i;
+
+	for (i = 0; i < connector->count_modes; i++) {
+		m = &connector->modes[i];
+		_eglAddNewMode(screen, m->hdisplay, m->vdisplay, m->vrefresh, m->name);
+	}
+}
+
+EGLBoolean
+drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor)
+{
+	_EGLDisplay *disp = _eglLookupDisplay(dpy);
+	struct drm_device *dev = (struct drm_device *)drv;
+	struct drm_screen *screen = NULL;
+	drmModeConnectorPtr connector = NULL;
+	drmModeResPtr res = NULL;
+	unsigned count_connectors = 0;
+	int num_screens = 0;
+	EGLint i;
+	int fd;
+
+	fd = drmOpen("i915", NULL);
+	if (fd < 0)
+		goto err_fd;
+
+	dev->drmFD = fd;
+	drm_get_device_id(dev);
+
+	dev->screen = drm_api_hocks.create_screen(dev->drmFD, dev->deviceID);
+	if (!dev->screen)
+		goto err_screen;
+	dev->winsys = dev->screen->winsys;
+
+	/* TODO HACK */
+	driInitExtensions(NULL, card_extensions, GL_FALSE);
+
+	drm_update_res(dev);
+	res = dev->res;
+	if (res)
+		count_connectors = res->count_connectors;
+	else
+		_eglLog(_EGL_WARNING, "Could not retrive kms information\n");
+
+	for(i = 0; i < count_connectors && i < MAX_SCREENS; i++) {
+		connector = drmModeGetConnector(fd, res->connectors[i]);
+
+		if (!connector)
+			continue;
+
+		if (connector->connection != DRM_MODE_CONNECTED) {
+			drmModeFreeConnector(connector);
+			continue;
+		}
+
+		screen = malloc(sizeof(struct drm_screen));
+		memset(screen, 0, sizeof(*screen));
+		screen->connector = connector;
+		screen->connectorID = connector->connector_id;
+		_eglInitScreen(&screen->base);
+		_eglAddScreen(disp, &screen->base);
+		drm_add_modes_from_connector(&screen->base, connector);
+		dev->screens[num_screens++] = screen;
+	}
+	dev->count_screens = num_screens;
+
+	/* for now we only have one config */
+	_EGLConfig *config = calloc(1, sizeof(*config));
+	memset(config, 1, sizeof(*config));
+	_eglInitConfig(config, 1);
+	_eglSetConfigAttrib(config, EGL_RED_SIZE, 8);
+	_eglSetConfigAttrib(config, EGL_GREEN_SIZE, 8);
+	_eglSetConfigAttrib(config, EGL_BLUE_SIZE, 8);
+	_eglSetConfigAttrib(config, EGL_ALPHA_SIZE, 8);
+	_eglSetConfigAttrib(config, EGL_BUFFER_SIZE, 32);
+	_eglSetConfigAttrib(config, EGL_DEPTH_SIZE, 24);
+	_eglSetConfigAttrib(config, EGL_STENCIL_SIZE, 8);
+	_eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_PBUFFER_BIT);
+	_eglAddConfig(disp, config);
+
+	drv->Initialized = EGL_TRUE;
+
+	*major = 1;
+	*minor = 4;
+
+	return EGL_TRUE;
+
+err_screen:
+	drmClose(fd);
+err_fd:
+	return EGL_FALSE;
+}
+
+EGLBoolean
+drm_terminate(_EGLDriver *drv, EGLDisplay dpy)
+{
+	struct drm_device *dev = (struct drm_device *)drv;
+	struct drm_screen *screen;
+	int i = 0;
+
+	drmFreeVersion(dev->version);
+
+	for (i = 0; i < dev->count_screens; i++) {
+		screen = dev->screens[i];
+
+		if (screen->shown)
+			drm_takedown_shown_screen(drv, screen);
+
+		drmModeFreeConnector(screen->connector);
+		_eglDestroyScreen(&screen->base);
+		dev->screens[i] = NULL;
+	}
+
+	dev->screen->destroy(dev->screen);
+	dev->winsys = NULL;
+
+	drmClose(dev->drmFD);
+
+	_eglCleanupDisplay(_eglLookupDisplay(dpy));
+	free(dev);
+
+	return EGL_TRUE;
+}
diff --git a/src/gallium/state_trackers/egl/egl_tracker.h b/src/gallium/state_trackers/egl/egl_tracker.h
new file mode 100644
index 0000000000..0b4dd9797d
--- /dev/null
+++ b/src/gallium/state_trackers/egl/egl_tracker.h
@@ -0,0 +1,191 @@
+
+#ifndef _EGL_TRACKER_H_
+#define _EGL_TRACKER_H_
+
+#include <stdint.h>
+
+#include "eglconfig.h"
+#include "eglcontext.h"
+#include "egldisplay.h"
+#include "egldriver.h"
+#include "eglglobals.h"
+#include "eglmode.h"
+#include "eglscreen.h"
+#include "eglsurface.h"
+
+#include "xf86drm.h"
+#include "xf86drmMode.h"
+
+#include "pipe/p_compiler.h"
+
+#include "state_tracker/st_public.h"
+
+#define MAX_SCREENS 16
+
+struct pipe_winsys;
+struct pipe_screen;
+struct pipe_context;
+struct state_tracker;
+
+struct drm_screen;
+struct drm_context;
+
+struct drm_device
+{
+	_EGLDriver base; /* base class/object */
+
+	/*
+	 * pipe
+	 */
+
+	struct pipe_winsys *winsys;
+	struct pipe_screen *screen;
+
+	/*
+	 * drm
+	 */
+
+	int drmFD;
+	drmVersionPtr version;
+	int deviceID;
+
+	drmModeResPtr res;
+
+	struct drm_screen *screens[MAX_SCREENS];
+	size_t count_screens;
+};
+
+struct drm_surface
+{
+	_EGLSurface base; /* base class/object */
+
+	/*
+	 * pipe
+	 */
+
+
+	struct st_framebuffer *stfb;
+
+	/*
+	 * drm
+	 */
+
+	struct drm_context *user;
+	struct drm_screen *screen;
+
+	int w;
+	int h;
+};
+
+struct drm_context
+{
+	_EGLContext base; /* base class/object */
+
+	/* pipe */
+
+	struct pipe_context *pipe;
+	struct st_context *st;
+};
+
+struct drm_screen
+{
+	_EGLScreen base;
+
+	/*
+	 * pipe
+	 */
+
+	struct pipe_buffer *buffer;
+	struct pipe_texture *tex;
+	struct pipe_surface *surface;
+
+	/*
+	 * drm
+	 */
+
+	struct {
+		unsigned height;
+		unsigned width;
+		unsigned pitch;
+		unsigned handle;
+	} front;
+
+	/* currently only support one connector */
+	drmModeConnectorPtr connector;
+	uint32_t connectorID;
+
+	/* Has this screen been shown */
+	int shown;
+
+	/* Surface that is currently attached to this screen */
+	struct drm_surface *surf;
+
+	/* framebuffer */
+	drmModeFBPtr fb;
+	uint32_t fbID;
+
+	/* crtc and mode used */
+	/*drmModeCrtcPtr crtc;*/
+	uint32_t crtcID;
+
+	struct drm_mode_modeinfo *mode;
+};
+
+
+static INLINE struct drm_context *
+lookup_drm_context(EGLContext context)
+{
+	_EGLContext *c = _eglLookupContext(context);
+	return (struct drm_context *) c;
+}
+
+
+static INLINE struct drm_surface *
+lookup_drm_surface(EGLSurface surface)
+{
+	_EGLSurface *s = _eglLookupSurface(surface);
+	return (struct drm_surface *) s;
+}
+
+static INLINE struct drm_screen *
+lookup_drm_screen(EGLDisplay dpy, EGLScreenMESA screen)
+{
+	_EGLScreen *s = _eglLookupScreen(dpy, screen);
+	return (struct drm_screen *) s;
+}
+
+/**
+ * egl_visual.h
+ */
+/*@{*/
+void drm_visual_modes_destroy(__GLcontextModes *modes);
+__GLcontextModes* drm_visual_modes_create(unsigned count, size_t minimum_size);
+__GLcontextModes* drm_visual_from_config(_EGLConfig *conf);
+/*@}*/
+
+/**
+ * egl_surface.h
+ */
+/*@{*/
+void drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen);
+/*@}*/
+
+/**
+ * All function exported to the egl side.
+ */
+/*@{*/
+EGLBoolean drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor);
+EGLBoolean drm_terminate(_EGLDriver *drv, EGLDisplay dpy);
+EGLContext drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list);
+EGLBoolean drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context);
+EGLSurface drm_create_window_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list);
+EGLSurface drm_create_pixmap_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list);
+EGLSurface drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list);
+EGLSurface drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, const EGLint *attrib_list);
+EGLBoolean drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, EGLSurface surface, EGLModeMESA m);
+EGLBoolean drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface);
+EGLBoolean drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context);
+EGLBoolean drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw);
+/*@}*/
+
+#endif
diff --git a/src/gallium/state_trackers/egl/egl_visual.c b/src/gallium/state_trackers/egl/egl_visual.c
new file mode 100644
index 0000000000..e59f893851
--- /dev/null
+++ b/src/gallium/state_trackers/egl/egl_visual.c
@@ -0,0 +1,85 @@
+
+#include "egl_tracker.h"
+
+#include "egllog.h"
+
+void
+drm_visual_modes_destroy(__GLcontextModes *modes)
+{
+   _eglLog(_EGL_DEBUG, "%s", __FUNCTION__);
+
+   while (modes) {
+      __GLcontextModes * const next = modes->next;
+      free(modes);
+      modes = next;
+   }
+}
+
+__GLcontextModes *
+drm_visual_modes_create(unsigned count, size_t minimum_size)
+{
+	/* This code copied from libGLX, and modified */
+	const size_t size = (minimum_size > sizeof(__GLcontextModes))
+		? minimum_size : sizeof(__GLcontextModes);
+	__GLcontextModes * head = NULL;
+	__GLcontextModes ** next;
+	unsigned   i;
+
+	_eglLog(_EGL_DEBUG, "%s %d %d", __FUNCTION__, count, minimum_size);
+
+	next = & head;
+	for (i = 0 ; i < count ; i++) {
+		*next = (__GLcontextModes *) calloc(1, size);
+		if (*next == NULL) {
+			drm_visual_modes_destroy(head);
+			head = NULL;
+			break;
+		}
+
+		(*next)->doubleBufferMode = 1;
+		(*next)->visualID = GLX_DONT_CARE;
+		(*next)->visualType = GLX_DONT_CARE;
+		(*next)->visualRating = GLX_NONE;
+		(*next)->transparentPixel = GLX_NONE;
+		(*next)->transparentRed = GLX_DONT_CARE;
+		(*next)->transparentGreen = GLX_DONT_CARE;
+		(*next)->transparentBlue = GLX_DONT_CARE;
+		(*next)->transparentAlpha = GLX_DONT_CARE;
+		(*next)->transparentIndex = GLX_DONT_CARE;
+		(*next)->xRenderable = GLX_DONT_CARE;
+		(*next)->fbconfigID = GLX_DONT_CARE;
+		(*next)->swapMethod = GLX_SWAP_UNDEFINED_OML;
+		(*next)->bindToTextureRgb = GLX_DONT_CARE;
+		(*next)->bindToTextureRgba = GLX_DONT_CARE;
+		(*next)->bindToMipmapTexture = GLX_DONT_CARE;
+		(*next)->bindToTextureTargets = 0;
+		(*next)->yInverted = GLX_DONT_CARE;
+
+		next = & ((*next)->next);
+	}
+
+	return head;
+}
+
+__GLcontextModes *
+drm_visual_from_config(_EGLConfig *conf)
+{
+	__GLcontextModes *visual;
+	(void)conf;
+
+	visual = drm_visual_modes_create(1, sizeof(*visual));
+	visual->redBits = 8;
+	visual->greenBits = 8;
+	visual->blueBits = 8;
+	visual->alphaBits = 8;
+
+	visual->rgbBits = 32;
+	visual->doubleBufferMode = 1;
+
+	visual->depthBits = 24;
+	visual->haveDepthBuffer = visual->depthBits > 0;
+	visual->stencilBits = 8;
+	visual->haveStencilBuffer = visual->stencilBits > 0;
+
+	return visual;
+}
diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile
new file mode 100644
index 0000000000..cddfca54fe
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/Makefile
@@ -0,0 +1,21 @@
+TARGET		= libg3dvl.a
+OBJECTS		= vl_display.o vl_screen.o vl_context.o vl_surface.o vl_shader_build.o vl_util.o vl_basic_csc.o	\
+		  vl_r16snorm_mc_buf.o
+GALLIUMDIR	= ../..
+
+CFLAGS		+= -g -Wall -Werror=implicit-function-declaration -fPIC	\
+		   -I${GALLIUMDIR}/include				\
+		   -I${GALLIUMDIR}/auxiliary				\
+		   -I${GALLIUMDIR}/winsys/g3dvl				\
+
+#############################################
+
+.PHONY	= all clean
+
+all: ${TARGET}
+
+${TARGET}: ${OBJECTS}
+	ar rcs $@ $^
+
+clean:
+	rm -rf ${OBJECTS} ${TARGET}
diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
new file mode 100644
index 0000000000..187a13a560
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
@@ -0,0 +1,715 @@
+#define VL_INTERNAL
+#include "vl_basic_csc.h"
+#include <assert.h>
+#include <pipe/p_context.h>
+#include <pipe/p_state.h>
+#include <pipe/p_inlines.h>
+#include <tgsi/tgsi_parse.h>
+#include <tgsi/tgsi_build.h>
+#include <util/u_memory.h>
+#include "vl_csc.h"
+#include "vl_surface.h"
+#include "vl_shader_build.h"
+#include "vl_types.h"
+
+struct vlVertexShaderConsts
+{
+	struct vlVertex4f	dst_scale;
+	struct vlVertex4f	dst_trans;
+	struct vlVertex4f	src_scale;
+	struct vlVertex4f	src_trans;
+};
+
+struct vlFragmentShaderConsts
+{
+	struct vlVertex4f	bias;
+	float			matrix[16];
+};
+
+struct vlBasicCSC
+{
+	struct vlCSC				base;
+
+	struct pipe_context			*pipe;
+	struct pipe_viewport_state		viewport;
+	struct pipe_framebuffer_state		framebuffer;
+	struct pipe_texture			*framebuffer_tex;
+	void					*sampler;
+	void					*vertex_shader, *fragment_shader;
+	struct pipe_vertex_buffer 		vertex_bufs[2];
+	struct pipe_vertex_element		vertex_elems[2];
+	struct pipe_constant_buffer		vs_const_buf, fs_const_buf;
+};
+
+static int vlResizeFrameBuffer
+(
+	struct vlCSC *csc,
+	unsigned int width,
+	unsigned int height
+)
+{
+	struct vlBasicCSC	*basic_csc;
+	struct pipe_context	*pipe;
+	struct pipe_texture	template;
+
+	assert(csc);
+
+	basic_csc = (struct vlBasicCSC*)csc;
+	pipe = basic_csc->pipe;
+
+	if (basic_csc->framebuffer.width == width && basic_csc->framebuffer.height == height)
+		return 0;
+
+	basic_csc->viewport.scale[0] = width;
+	basic_csc->viewport.scale[1] = height;
+	basic_csc->viewport.scale[2] = 1;
+	basic_csc->viewport.scale[3] = 1;
+	basic_csc->viewport.translate[0] = 0;
+	basic_csc->viewport.translate[1] = 0;
+	basic_csc->viewport.translate[2] = 0;
+	basic_csc->viewport.translate[3] = 0;
+	
+	if (basic_csc->framebuffer_tex)
+	{
+		pipe_surface_reference(&basic_csc->framebuffer.cbufs[0], NULL);
+		pipe_texture_reference(&basic_csc->framebuffer_tex, NULL);
+	}
+	
+	memset(&template, 0, sizeof(struct pipe_texture));
+	template.target = PIPE_TEXTURE_2D;
+	template.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	template.last_level = 0;
+	template.width[0] = width;
+	template.height[0] = height;
+	template.depth[0] = 1;
+	template.compressed = 0;
+	pf_get_block(template.format, &template.block);
+	template.tex_usage = PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+
+	basic_csc->framebuffer_tex = pipe->screen->texture_create(pipe->screen, &template);
+
+	basic_csc->framebuffer.width = width;
+	basic_csc->framebuffer.height = height;
+	basic_csc->framebuffer.cbufs[0] = pipe->screen->get_tex_surface
+	(
+		pipe->screen,
+		basic_csc->framebuffer_tex,
+		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
+	);
+
+	/* Clear to black, in case video doesn't fill the entire window */
+	pipe->clear(pipe, basic_csc->framebuffer.cbufs[0], 0);
+
+	return 0;
+}
+
+static int vlBegin
+(
+	struct vlCSC *csc
+)
+{
+	struct vlBasicCSC	*basic_csc;
+	struct pipe_context	*pipe;
+
+	assert(csc);
+
+	basic_csc = (struct vlBasicCSC*)csc;
+	pipe = basic_csc->pipe;
+
+	pipe->set_framebuffer_state(pipe, &basic_csc->framebuffer);
+	pipe->set_viewport_state(pipe, &basic_csc->viewport);
+	pipe->bind_sampler_states(pipe, 1, (void**)&basic_csc->sampler);
+	/* Source texture set in vlPutPictureCSC() */
+	pipe->bind_vs_state(pipe, basic_csc->vertex_shader);
+	pipe->bind_fs_state(pipe, basic_csc->fragment_shader);
+	pipe->set_vertex_buffers(pipe, 2, basic_csc->vertex_bufs);
+	pipe->set_vertex_elements(pipe, 2, basic_csc->vertex_elems);
+	pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &basic_csc->vs_const_buf);
+	pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &basic_csc->fs_const_buf);
+
+	return 0;
+}
+
+static int vlPutPictureCSC
+(
+	struct vlCSC *csc,
+	struct vlSurface *surface,
+	int srcx,
+	int srcy,
+	int srcw,
+	int srch,
+	int destx,
+	int desty,
+	int destw,
+	int desth,
+	enum vlPictureType picture_type
+)
+{
+	struct vlBasicCSC		*basic_csc;
+	struct pipe_context		*pipe;
+	struct vlVertexShaderConsts	*vs_consts;
+
+	assert(csc);
+	assert(surface);
+
+	basic_csc = (struct vlBasicCSC*)csc;
+	pipe = basic_csc->pipe;
+
+	vs_consts = pipe_buffer_map
+	(
+		pipe->screen,
+		basic_csc->vs_const_buf.buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD
+	);
+
+	vs_consts->dst_scale.x = destw / (float)basic_csc->framebuffer.cbufs[0]->width;
+	vs_consts->dst_scale.y = desth / (float)basic_csc->framebuffer.cbufs[0]->height;
+	vs_consts->dst_scale.z = 1;
+	vs_consts->dst_scale.w = 1;
+	vs_consts->dst_trans.x = destx / (float)basic_csc->framebuffer.cbufs[0]->width;
+	vs_consts->dst_trans.y = desty / (float)basic_csc->framebuffer.cbufs[0]->height;
+	vs_consts->dst_trans.z = 0;
+	vs_consts->dst_trans.w = 0;
+
+	vs_consts->src_scale.x = srcw / (float)surface->texture->width[0];
+	vs_consts->src_scale.y = srch / (float)surface->texture->height[0];
+	vs_consts->src_scale.z = 1;
+	vs_consts->src_scale.w = 1;
+	vs_consts->src_trans.x = srcx / (float)surface->texture->width[0];
+	vs_consts->src_trans.y = srcy / (float)surface->texture->height[0];
+	vs_consts->src_trans.z = 0;
+	vs_consts->src_trans.w = 0;
+
+	pipe_buffer_unmap(pipe->screen, basic_csc->vs_const_buf.buffer);
+
+	pipe->set_sampler_textures(pipe, 1, &surface->texture);
+	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_STRIP, 0, 4);
+
+	return 0;
+}
+
+static int vlEnd
+(
+	struct vlCSC *csc
+)
+{
+	assert(csc);
+
+	return 0;
+}
+
+static struct pipe_surface* vlGetFrameBuffer
+(
+	struct vlCSC *csc
+)
+{
+	struct vlBasicCSC	*basic_csc;
+
+	assert(csc);
+
+	basic_csc = (struct vlBasicCSC*)csc;
+
+	return basic_csc->framebuffer.cbufs[0];
+}
+
+static int vlDestroy
+(
+	struct vlCSC *csc
+)
+{
+	struct vlBasicCSC	*basic_csc;
+	struct pipe_context	*pipe;
+	unsigned int		i;
+
+	assert(csc);
+
+	basic_csc = (struct vlBasicCSC*)csc;
+	pipe = basic_csc->pipe;
+
+	if (basic_csc->framebuffer_tex)
+	{
+		pipe_surface_reference(&basic_csc->framebuffer.cbufs[0], NULL);
+		pipe_texture_reference(&basic_csc->framebuffer_tex, NULL);
+	}
+
+	pipe->delete_sampler_state(pipe, basic_csc->sampler);
+	pipe->delete_vs_state(pipe, basic_csc->vertex_shader);
+	pipe->delete_fs_state(pipe, basic_csc->fragment_shader);
+
+	for (i = 0; i < 2; ++i)
+		pipe_buffer_reference(pipe->screen, &basic_csc->vertex_bufs[i].buffer, NULL);
+
+	pipe_buffer_reference(pipe->screen, &basic_csc->vs_const_buf.buffer, NULL);
+	pipe_buffer_reference(pipe->screen, &basic_csc->fs_const_buf.buffer, NULL);
+
+	FREE(basic_csc);
+
+	return 0;
+}
+
+/*
+ * Represents 2 triangles in a strip in normalized coords.
+ * Used to render the surface onto the frame buffer.
+ */
+static const struct vlVertex2f surface_verts[4] =
+{
+	{0.0f, 0.0f},
+	{0.0f, 1.0f},
+	{1.0f, 0.0f},
+	{1.0f, 1.0f}
+};
+
+/*
+ * Represents texcoords for the above. We can use the position values directly.
+ * TODO: Duplicate these in the shader, no need to create a buffer.
+ */
+static const struct vlVertex2f *surface_texcoords = surface_verts;
+
+/*
+ * Identity color conversion constants, for debugging
+ */
+static const struct vlFragmentShaderConsts identity =
+{
+	{
+		0.0f, 0.0f, 0.0f, 0.0f
+	},
+	{
+		1.0f, 0.0f, 0.0f, 0.0f,
+		0.0f, 1.0f, 0.0f, 0.0f,
+		0.0f, 0.0f, 1.0f, 0.0f,
+		0.0f, 0.0f, 0.0f, 1.0f
+	}
+};
+
+/*
+ * Converts ITU-R BT.601 YCbCr pixels to RGB pixels where:
+ * Y is in [16,235], Cb and Cr are in [16,240]
+ * R, G, and B are in [16,235]
+ */
+static const struct vlFragmentShaderConsts bt_601 =
+{
+	{
+		0.0f,		0.501960784f,	0.501960784f,	0.0f
+	},
+	{
+		1.0f,		0.0f,		1.371f,		0.0f,
+		1.0f,		-0.336f,	-0.698f,	0.0f,
+		1.0f,		1.732f,		0.0f,		0.0f,
+		0.0f,		0.0f,		0.0f,		1.0f
+	}
+};
+
+/*
+ * Converts ITU-R BT.601 YCbCr pixels to RGB pixels where:
+ * Y is in [16,235], Cb and Cr are in [16,240]
+ * R, G, and B are in [0,255]
+ */
+static const struct vlFragmentShaderConsts bt_601_full =
+{
+	{
+		0.062745098f,	0.501960784f,	0.501960784f,	0.0f
+	},
+	{
+		1.164f,		0.0f,		1.596f,		0.0f,
+		1.164f,		-0.391f,	-0.813f,	0.0f,
+		1.164f,		2.018f,		0.0f,		0.0f,
+		0.0f,		0.0f,		0.0f,		1.0f
+	}
+};
+
+/*
+ * Converts ITU-R BT.709 YCbCr pixels to RGB pixels where:
+ * Y is in [16,235], Cb and Cr are in [16,240]
+ * R, G, and B are in [16,235]
+ */
+static const struct vlFragmentShaderConsts bt_709 =
+{
+	{
+		0.0f,		0.501960784f,	0.501960784f,	0.0f
+	},
+	{
+		1.0f,		0.0f,		1.540f,		0.0f,
+		1.0f,		-0.183f,	-0.459f,	0.0f,
+		1.0f,		1.816f,		0.0f,		0.0f,
+		0.0f,		0.0f,		0.0f,		1.0f
+	}
+};
+
+/*
+ * Converts ITU-R BT.709 YCbCr pixels to RGB pixels where:
+ * Y is in [16,235], Cb and Cr are in [16,240]
+ * R, G, and B are in [0,255]
+ */
+const struct vlFragmentShaderConsts bt_709_full =
+{
+	{
+		0.062745098f,	0.501960784f,	0.501960784f,	0.0f
+	},
+	{
+		1.164f,		0.0f,		1.793f,		0.0f,
+		1.164f,		-0.213f,	-0.534f,	0.0f,
+		1.164f,		2.115f,		0.0f,		0.0f,
+		0.0f,		0.0f,		0.0f,		1.0f
+	}
+};
+
+static int vlCreateVertexShader
+(
+	struct vlBasicCSC *csc
+)
+{
+	const unsigned int		max_tokens = 50;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(context);
+
+	pipe = csc->pipe;
+	tokens = (struct tgsi_token*)MALLOC(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Vertex texcoords
+	 */
+	for (i = 0; i < 2; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0		; Scaling vector to scale vertex pos rect to destination size
+	 * decl c1		; Translation vector to move vertex pos rect into position
+	 * decl c2		; Scaling vector to scale texcoord rect to source size
+	 * decl c3		; Translation vector to move texcoord rect into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 3);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Vertex texcoords
+	 */
+	for (i = 0; i < 2; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * madd o0, i0, c0, c1	; Scale and translate unit output rect to destination size and pos
+	 * madd o1, i1, c2, c3	; Scale and translate unit texcoord rect to source size and pos
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_inst4(TGSI_OPCODE_MADD, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i, TGSI_FILE_CONSTANT, i * 2, TGSI_FILE_CONSTANT, i * 2 + 1);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	csc->vertex_shader = pipe->create_vs_state(pipe, &vs);
+	FREE(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShader
+(
+	struct vlBasicCSC *csc
+)
+{
+	const unsigned int		max_tokens = 50;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(context);
+
+	pipe = csc->pipe;
+	tokens = (struct tgsi_token*)MALLOC(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/* decl i0		; Texcoords for s0 */
+	decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, 1, 0, 0, TGSI_INTERPOLATE_LINEAR);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl c0		; Bias vector for CSC
+	 * decl c1-c4		; CSC matrix c1-c4
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 4);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0		; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl s0		; Sampler for tex containing picture to display */
+	decl = vl_decl_samplers(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* tex2d t0, i0, s0	; Read src pixel */
+	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* sub t0, t0, c0	; Subtract bias vector from pixel */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * dp4 o0.x, t0, c1	; Multiply pixel by the color conversion matrix
+	 * dp4 o0.y, t0, c2
+	 * dp4 o0.z, t0, c3
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_DP4, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 1);
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	csc->fragment_shader = pipe->create_fs_state(pipe, &fs);
+	FREE(tokens);
+
+	return 0;
+}
+
+static int vlCreateDataBufs
+(
+	struct vlBasicCSC *csc
+)
+{
+	struct pipe_context *pipe;
+
+	assert(csc);
+
+	pipe = csc->pipe;
+
+	/*
+	 * Create our vertex buffer and vertex buffer element
+	 * VB contains 4 vertices that render a quad covering the entire window
+	 * to display a rendered surface
+	 * Quad is rendered as a tri strip
+	 */
+	csc->vertex_bufs[0].stride = sizeof(struct vlVertex2f);
+	csc->vertex_bufs[0].max_index = 3;
+	csc->vertex_bufs[0].buffer_offset = 0;
+	csc->vertex_bufs[0].buffer = pipe_buffer_create
+	(
+		pipe->screen,
+		1,
+		PIPE_BUFFER_USAGE_VERTEX,
+		sizeof(struct vlVertex2f) * 4
+	);
+
+	memcpy
+	(
+		pipe_buffer_map(pipe->screen, csc->vertex_bufs[0].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		surface_verts,
+		sizeof(struct vlVertex2f) * 4
+	);
+
+	pipe_buffer_unmap(pipe->screen, csc->vertex_bufs[0].buffer);
+
+	csc->vertex_elems[0].src_offset = 0;
+	csc->vertex_elems[0].vertex_buffer_index = 0;
+	csc->vertex_elems[0].nr_components = 2;
+	csc->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/*
+	 * Create our texcoord buffer and texcoord buffer element
+	 * Texcoord buffer contains the TCs for mapping the rendered surface to the 4 vertices
+	 */
+	csc->vertex_bufs[1].stride = sizeof(struct vlVertex2f);
+	csc->vertex_bufs[1].max_index = 3;
+	csc->vertex_bufs[1].buffer_offset = 0;
+	csc->vertex_bufs[1].buffer = pipe_buffer_create
+	(
+		pipe->screen,
+		1,
+		PIPE_BUFFER_USAGE_VERTEX,
+		sizeof(struct vlVertex2f) * 4
+	);
+
+	memcpy
+	(
+		pipe_buffer_map(pipe->screen, csc->vertex_bufs[1].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		surface_texcoords,
+		sizeof(struct vlVertex2f) * 4
+	);
+
+	pipe_buffer_unmap(pipe->screen, csc->vertex_bufs[1].buffer);
+
+	csc->vertex_elems[1].src_offset = 0;
+	csc->vertex_elems[1].vertex_buffer_index = 1;
+	csc->vertex_elems[1].nr_components = 2;
+	csc->vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/*
+	 * Create our vertex shader's constant buffer
+	 * Const buffer contains scaling and translation vectors
+	 */
+	csc->vs_const_buf.buffer = pipe_buffer_create
+	(
+		pipe->screen,
+		1,
+		PIPE_BUFFER_USAGE_CONSTANT | PIPE_BUFFER_USAGE_DISCARD,
+		sizeof(struct vlVertexShaderConsts)
+	);
+
+	/*
+	 * Create our fragment shader's constant buffer
+	 * Const buffer contains the color conversion matrix and bias vectors
+	 */
+	csc->fs_const_buf.buffer = pipe_buffer_create
+	(
+		pipe->screen,
+		1,
+		PIPE_BUFFER_USAGE_CONSTANT,
+		sizeof(struct vlFragmentShaderConsts)
+	);
+
+	/*
+	 * TODO: Refactor this into a seperate function,
+	 * allow changing the CSC matrix at runtime to switch between regular & full versions
+	 */
+	memcpy
+	(
+		pipe_buffer_map(pipe->screen, csc->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		&bt_601_full,
+		sizeof(struct vlFragmentShaderConsts)
+	);
+
+	pipe_buffer_unmap(pipe->screen, csc->fs_const_buf.buffer);
+
+	return 0;
+}
+
+static int vlInit
+(
+	struct vlBasicCSC *csc
+)
+{
+	struct pipe_context		*pipe;
+	struct pipe_sampler_state	sampler;
+
+	assert(csc);
+
+	pipe = csc->pipe;
+
+	/* Delay creating the FB until vlPutPictureCSC() so we know window size */
+	csc->framebuffer_tex = NULL;
+	csc->framebuffer.width = 0;
+	csc->framebuffer.height = 0;
+	csc->framebuffer.nr_cbufs = 1;
+	csc->framebuffer.cbufs[0] = NULL;
+	csc->framebuffer.zsbuf = NULL;
+
+	sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+	sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+	sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+	sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR;
+	sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+	sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR;
+	sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
+	sampler.compare_func = PIPE_FUNC_ALWAYS;
+	sampler.normalized_coords = 1;
+	/*sampler.prefilter = ;*/
+	/*sampler.shadow_ambient = ;*/
+	/*sampler.lod_bias = ;*/
+	/*sampler.min_lod = ;*/
+	/*sampler.max_lod = ;*/
+	/*sampler.border_color[i] = ;*/
+	/*sampler.max_anisotropy = ;*/
+	csc->sampler = pipe->create_sampler_state(pipe, &sampler);
+
+	vlCreateVertexShader(csc);
+	vlCreateFragmentShader(csc);
+	vlCreateDataBufs(csc);
+
+	return 0;
+}
+
+int vlCreateBasicCSC
+(
+	struct pipe_context *pipe,
+	struct vlCSC **csc
+)
+{
+	struct vlBasicCSC *basic_csc;
+
+	assert(pipe);
+	assert(csc);
+
+	basic_csc = CALLOC_STRUCT(vlBasicCSC);
+
+	if (!basic_csc)
+		return 1;
+
+	basic_csc->base.vlResizeFrameBuffer = &vlResizeFrameBuffer;
+	basic_csc->base.vlBegin = &vlBegin;
+	basic_csc->base.vlPutPicture = &vlPutPictureCSC;
+	basic_csc->base.vlEnd = &vlEnd;
+	basic_csc->base.vlGetFrameBuffer = &vlGetFrameBuffer;
+	basic_csc->base.vlDestroy = &vlDestroy;
+	basic_csc->pipe = pipe;
+
+	vlInit(basic_csc);
+
+	*csc = &basic_csc->base;
+
+	return 0;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.h b/src/gallium/state_trackers/g3dvl/vl_basic_csc.h
new file mode 100644
index 0000000000..2e17f1d814
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.h
@@ -0,0 +1,13 @@
+#ifndef vl_basic_csc_h
+#define vl_basic_csc_h
+
+struct pipe_context;
+struct vlCSC;
+
+int vlCreateBasicCSC
+(
+	struct pipe_context *pipe,
+	struct vlCSC **csc
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c
new file mode 100644
index 0000000000..65ddb9f01e
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_context.c
@@ -0,0 +1,208 @@
+#define VL_INTERNAL
+#include "vl_context.h"
+#include <assert.h>
+#include <pipe/p_context.h>
+#include <pipe/p_state.h>
+#include <util/u_memory.h>
+#include "vl_render.h"
+#include "vl_r16snorm_mc_buf.h"
+#include "vl_csc.h"
+#include "vl_basic_csc.h"
+
+static int vlInitCommon(struct vlContext *context)
+{
+	struct pipe_context			*pipe;
+	struct pipe_rasterizer_state		rast;
+	struct pipe_blend_state			blend;
+	struct pipe_depth_stencil_alpha_state	dsa;
+	unsigned int				i;
+
+	assert(context);
+
+	pipe = context->pipe;
+
+	rast.flatshade = 1;
+	rast.flatshade_first = 0;
+	rast.light_twoside = 0;
+	rast.front_winding = PIPE_WINDING_CCW;
+	rast.cull_mode = PIPE_WINDING_CW;
+	rast.fill_cw = PIPE_POLYGON_MODE_FILL;
+	rast.fill_ccw = PIPE_POLYGON_MODE_FILL;
+	rast.offset_cw = 0;
+	rast.offset_ccw = 0;
+	rast.scissor = 0;
+	rast.poly_smooth = 0;
+	rast.poly_stipple_enable = 0;
+	rast.point_sprite = 0;
+	rast.point_size_per_vertex = 0;
+	rast.multisample = 0;
+	rast.line_smooth = 0;
+	rast.line_stipple_enable = 0;
+	rast.line_stipple_factor = 0;
+	rast.line_stipple_pattern = 0;
+	rast.line_last_pixel = 0;
+	/* Don't need clipping, but viewport mapping done here */
+	rast.bypass_clipping = 0;
+	rast.bypass_vs = 0;
+	rast.origin_lower_left = 0;
+	rast.line_width = 1;
+	rast.point_smooth = 0;
+	rast.point_size = 1;
+	rast.offset_units = 1;
+	rast.offset_scale = 1;
+	/*rast.sprite_coord_mode[i] = ;*/
+	context->raster = pipe->create_rasterizer_state(pipe, &rast);
+	pipe->bind_rasterizer_state(pipe, context->raster);
+
+	blend.blend_enable = 0;
+	blend.rgb_func = PIPE_BLEND_ADD;
+	blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+	blend.rgb_dst_factor = PIPE_BLENDFACTOR_ONE;
+	blend.alpha_func = PIPE_BLEND_ADD;
+	blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+	blend.alpha_dst_factor = PIPE_BLENDFACTOR_ONE;
+	blend.logicop_enable = 0;
+	blend.logicop_func = PIPE_LOGICOP_CLEAR;
+	/* Needed to allow color writes to FB, even if blending disabled */
+	blend.colormask = PIPE_MASK_RGBA;
+	blend.dither = 0;
+	context->blend = pipe->create_blend_state(pipe, &blend);
+	pipe->bind_blend_state(pipe, context->blend);
+
+	dsa.depth.enabled = 0;
+	dsa.depth.writemask = 0;
+	dsa.depth.func = PIPE_FUNC_ALWAYS;
+	dsa.depth.occlusion_count = 0;
+	for (i = 0; i < 2; ++i)
+	{
+		dsa.stencil[i].enabled = 0;
+		dsa.stencil[i].func = PIPE_FUNC_ALWAYS;
+		dsa.stencil[i].fail_op = PIPE_STENCIL_OP_KEEP;
+		dsa.stencil[i].zpass_op = PIPE_STENCIL_OP_KEEP;
+		dsa.stencil[i].zfail_op = PIPE_STENCIL_OP_KEEP;
+		dsa.stencil[i].ref_value = 0;
+		dsa.stencil[i].valuemask = 0;
+		dsa.stencil[i].writemask = 0;
+	}
+	dsa.alpha.enabled = 0;
+	dsa.alpha.func = PIPE_FUNC_ALWAYS;
+	dsa.alpha.ref_value = 0;
+	context->dsa = pipe->create_depth_stencil_alpha_state(pipe, &dsa);
+	pipe->bind_depth_stencil_alpha_state(pipe, context->dsa);
+
+	return 0;
+}
+
+int vlCreateContext
+(
+	struct vlScreen *screen,
+	struct pipe_context *pipe,
+	unsigned int picture_width,
+	unsigned int picture_height,
+	enum vlFormat picture_format,
+	enum vlProfile profile,
+	enum vlEntryPoint entry_point,
+	struct vlContext **context
+)
+{
+	struct vlContext *ctx;
+
+	assert(screen);
+	assert(context);
+	assert(pipe);
+
+	ctx = CALLOC_STRUCT(vlContext);
+
+	if (!ctx)
+		return 1;
+
+	ctx->screen = screen;
+	ctx->pipe = pipe;
+	ctx->picture_width = picture_width;
+	ctx->picture_height = picture_height;
+	ctx->picture_format = picture_format;
+	ctx->profile = profile;
+	ctx->entry_point = entry_point;
+
+	vlInitCommon(ctx);
+
+	vlCreateR16SNormBufferedMC(pipe, picture_width, picture_height, picture_format, &ctx->render);
+	vlCreateBasicCSC(pipe, &ctx->csc);
+
+	*context = ctx;
+
+	return 0;
+}
+
+int vlDestroyContext
+(
+	struct vlContext *context
+)
+{
+	assert(context);
+
+	/* XXX: Must unbind shaders before we can delete them for some reason */
+	context->pipe->bind_vs_state(context->pipe, NULL);
+	context->pipe->bind_fs_state(context->pipe, NULL);
+
+	context->render->vlDestroy(context->render);
+	context->csc->vlDestroy(context->csc);
+
+	context->pipe->delete_blend_state(context->pipe, context->blend);
+	context->pipe->delete_rasterizer_state(context->pipe, context->raster);
+	context->pipe->delete_depth_stencil_alpha_state(context->pipe, context->dsa);
+
+	FREE(context);
+
+	return 0;
+}
+
+struct vlScreen* vlContextGetScreen
+(
+	struct vlContext *context
+)
+{
+	assert(context);
+
+	return context->screen;
+}
+
+struct pipe_context* vlGetPipeContext
+(
+	struct vlContext *context
+)
+{
+	assert(context);
+
+	return context->pipe;
+}
+
+unsigned int vlGetPictureWidth
+(
+	struct vlContext *context
+)
+{
+	assert(context);
+
+	return context->picture_width;
+}
+
+unsigned int vlGetPictureHeight
+(
+	struct vlContext *context
+)
+{
+	assert(context);
+
+	return context->picture_height;
+}
+
+enum vlFormat vlGetPictureFormat
+(
+	struct vlContext *context
+)
+{
+	assert(context);
+
+	return context->picture_format;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.h b/src/gallium/state_trackers/g3dvl/vl_context.h
new file mode 100644
index 0000000000..3d14634c44
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_context.h
@@ -0,0 +1,73 @@
+#ifndef vl_context_h
+#define vl_context_h
+
+#include "vl_types.h"
+
+struct pipe_context;
+
+#ifdef VL_INTERNAL
+struct vlRender;
+struct vlCSC;
+
+struct vlContext
+{
+	struct vlScreen		*screen;
+	struct pipe_context	*pipe;
+	unsigned int		picture_width;
+	unsigned int		picture_height;
+	enum vlFormat		picture_format;
+	enum vlProfile		profile;
+	enum vlEntryPoint	entry_point;
+
+	void			*raster;
+	void			*dsa;
+	void			*blend;
+
+	struct vlRender		*render;
+	struct vlCSC		*csc;
+};
+#endif
+
+int vlCreateContext
+(
+	struct vlScreen *screen,
+	struct pipe_context *pipe,
+	unsigned int picture_width,
+	unsigned int picture_height,
+	enum vlFormat picture_format,
+	enum vlProfile profile,
+	enum vlEntryPoint entry_point,
+	struct vlContext **context
+);
+
+int vlDestroyContext
+(
+	struct vlContext *context
+);
+
+struct vlScreen* vlContextGetScreen
+(
+	struct vlContext *context
+);
+
+struct pipe_context* vlGetPipeContext
+(
+	struct vlContext *context
+);
+
+unsigned int vlGetPictureWidth
+(
+	struct vlContext *context
+);
+
+unsigned int vlGetPictureHeight
+(
+	struct vlContext *context
+);
+
+enum vlFormat vlGetPictureFormat
+(
+	struct vlContext *context
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_csc.h b/src/gallium/state_trackers/g3dvl/vl_csc.h
new file mode 100644
index 0000000000..36417a2792
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_csc.h
@@ -0,0 +1,53 @@
+#ifndef vl_csc_h
+#define vl_csc_h
+
+#include "vl_types.h"
+
+struct pipe_surface;
+
+struct vlCSC
+{
+	int (*vlResizeFrameBuffer)
+	(
+		struct vlCSC *csc,
+		unsigned int width,
+		unsigned int height
+	);
+
+	int (*vlBegin)
+	(
+		struct vlCSC *csc
+	);
+
+	int (*vlPutPicture)
+	(
+		struct vlCSC *csc,
+		struct vlSurface *surface,
+		int srcx,
+		int srcy,
+		int srcw,
+		int srch,
+		int destx,
+		int desty,
+		int destw,
+		int desth,
+		enum vlPictureType picture_type
+	);
+
+	int (*vlEnd)
+	(
+		struct vlCSC *csc
+	);
+
+	struct pipe_surface* (*vlGetFrameBuffer)
+	(
+		struct vlCSC *csc
+	);
+
+	int (*vlDestroy)
+	(
+		struct vlCSC *csc
+	);
+};
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_defs.h b/src/gallium/state_trackers/g3dvl/vl_defs.h
new file mode 100644
index 0000000000..d612d02502
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_defs.h
@@ -0,0 +1,11 @@
+#ifndef vl_defs_h
+#define vl_defs_h
+
+#define VL_BLOCK_WIDTH		8
+#define VL_BLOCK_HEIGHT		8
+#define VL_BLOCK_SIZE		(VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT)
+#define VL_MACROBLOCK_WIDTH	16
+#define VL_MACROBLOCK_HEIGHT	16
+#define VL_MACROBLOCK_SIZE	(VL_MACROBLOCK_WIDTH * VL_MACROBLOCK_HEIGHT)
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_display.c b/src/gallium/state_trackers/g3dvl/vl_display.c
new file mode 100644
index 0000000000..dce06de758
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_display.c
@@ -0,0 +1,48 @@
+#define VL_INTERNAL
+#include "vl_display.h"
+#include <assert.h>
+#include <util/u_memory.h>
+
+int vlCreateDisplay
+(
+	vlNativeDisplay native_display,
+	struct vlDisplay **display
+)
+{
+	struct vlDisplay *dpy;
+
+	assert(native_display);
+	assert(display);
+
+	dpy = CALLOC_STRUCT(vlDisplay);
+
+	if (!dpy)
+		return 1;
+
+	dpy->native = native_display;
+	*display = dpy;
+
+	return 0;
+}
+
+int vlDestroyDisplay
+(
+	struct vlDisplay *display
+)
+{
+	assert(display);
+
+	FREE(display);
+
+	return 0;
+}
+
+vlNativeDisplay vlGetNativeDisplay
+(
+	struct vlDisplay *display
+)
+{
+	assert(display);
+
+	return display->native;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_display.h b/src/gallium/state_trackers/g3dvl/vl_display.h
new file mode 100644
index 0000000000..e11fd40799
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_display.h
@@ -0,0 +1,29 @@
+#ifndef vl_display_h
+#define vl_display_h
+
+#include "vl_types.h"
+
+#ifdef VL_INTERNAL
+struct vlDisplay
+{
+	vlNativeDisplay native;
+};
+#endif
+
+int vlCreateDisplay
+(
+	vlNativeDisplay native_display,
+	struct vlDisplay **display
+);
+
+int vlDestroyDisplay
+(
+	struct vlDisplay *display
+);
+
+vlNativeDisplay vlGetNativeDisplay
+(
+	struct vlDisplay *display
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
new file mode 100644
index 0000000000..2176bb86d8
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
@@ -0,0 +1,1155 @@
+#define VL_INTERNAL
+#include "vl_r16snorm_mc_buf.h"
+#include <assert.h>
+#include <pipe/p_context.h>
+#include <pipe/p_screen.h>
+#include <pipe/p_state.h>
+#include <pipe/p_inlines.h>
+#include <tgsi/tgsi_parse.h>
+#include <tgsi/tgsi_build.h>
+#include <util/u_math.h>
+#include <util/u_memory.h>
+#include "vl_render.h"
+#include "vl_shader_build.h"
+#include "vl_surface.h"
+#include "vl_util.h"
+#include "vl_types.h"
+#include "vl_defs.h"
+
+const unsigned int DEFAULT_BUF_ALIGNMENT = 1;
+
+enum vlMacroBlockTypeEx
+{
+	vlMacroBlockExTypeIntra,
+	vlMacroBlockExTypeFwdPredictedFrame,
+	vlMacroBlockExTypeFwdPredictedField,
+	vlMacroBlockExTypeBkwdPredictedFrame,
+	vlMacroBlockExTypeBkwdPredictedField,
+	vlMacroBlockExTypeBiPredictedFrame,
+	vlMacroBlockExTypeBiPredictedField,
+
+	vlNumMacroBlockExTypes
+};
+
+struct vlVertexShaderConsts
+{
+	struct vlVertex4f denorm;
+};
+
+struct vlFragmentShaderConsts
+{
+	struct vlVertex4f multiplier;
+	struct vlVertex4f div;
+};
+
+struct vlMacroBlockVertexStream0
+{
+	struct vlVertex2f pos;
+	struct vlVertex2f luma_tc;
+	struct vlVertex2f cb_tc;
+	struct vlVertex2f cr_tc;
+};
+
+struct vlR16SnormBufferedMC
+{
+	struct vlRender				base;
+
+	unsigned int				picture_width;
+	unsigned int				picture_height;
+	enum vlFormat				picture_format;
+	unsigned int				macroblocks_per_picture;
+
+	struct vlSurface			*buffered_surface;
+	struct vlSurface			*past_surface;
+	struct vlSurface			*future_surface;
+	struct vlVertex2f			surface_tex_inv_size;
+	struct vlVertex2f			zero_block[3];
+	unsigned int				num_macroblocks;
+	struct vlMpeg2MacroBlock		*macroblocks;
+	struct pipe_surface			*tex_surface[3];
+	short					*texels[3];
+
+	struct pipe_context			*pipe;
+	struct pipe_viewport_state		viewport;
+	struct pipe_framebuffer_state		render_target;
+
+	union
+	{
+		void					*all[5];
+		struct
+		{
+			void				*y;
+			void				*cb;
+			void				*cr;
+			void				*ref[2];
+		};
+	} samplers;
+
+	union
+	{
+		struct pipe_texture			*all[5];
+		struct
+		{
+			struct pipe_texture		*y;
+			struct pipe_texture		*cb;
+			struct pipe_texture		*cr;
+			struct pipe_texture		*ref[2];
+		};
+	} textures;
+
+	union
+	{
+		struct pipe_vertex_buffer 		all[3];
+		struct
+		{
+			struct pipe_vertex_buffer	ycbcr;
+			struct pipe_vertex_buffer	ref[2];
+		};
+	} vertex_bufs;
+
+	void					*i_vs, *p_vs[2], *b_vs[2];
+	void					*i_fs, *p_fs[2], *b_fs[2];
+	struct pipe_vertex_element		vertex_elems[8];
+	struct pipe_constant_buffer		vs_const_buf;
+	struct pipe_constant_buffer		fs_const_buf;
+};
+
+static inline int vlBegin
+(
+	struct vlRender *render
+)
+{
+	assert(render);
+
+	return 0;
+}
+
+static inline int vlGrabFrameCodedBlock(short *src, short *dst, unsigned int dst_pitch)
+{
+	unsigned int y;
+
+	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		memcpy
+		(
+			dst + y * dst_pitch,
+			src + y * VL_BLOCK_WIDTH,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	return 0;
+}
+
+static inline int vlGrabFieldCodedBlock(short *src, short *dst, unsigned int dst_pitch)
+{
+	unsigned int y;
+
+	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		memcpy
+		(
+			dst + y * dst_pitch * 2,
+			src + y * VL_BLOCK_WIDTH,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	return 0;
+}
+
+static inline int vlGrabNoBlock(short *dst, unsigned int dst_pitch)
+{
+	unsigned int y;
+
+	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		memset
+		(
+			dst + y * dst_pitch,
+			0,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	return 0;
+}
+
+static inline int vlGrabBlocks
+(
+	struct vlR16SnormBufferedMC *mc,
+	unsigned int mbx,
+	unsigned int mby,
+	enum vlDCTType dct_type,
+	unsigned int coded_block_pattern,
+	short *blocks
+)
+{
+	short			*texels;
+	unsigned int		tex_pitch;
+	unsigned int		x, y, tb = 0, sb = 0;
+	unsigned int		mbpx = mbx * VL_MACROBLOCK_WIDTH, mbpy = mby * VL_MACROBLOCK_HEIGHT;
+
+	assert(mc);
+	assert(blocks);
+
+	tex_pitch = mc->tex_surface[0]->stride / mc->tex_surface[0]->block.size;
+	texels = mc->texels[0] + mbpy * tex_pitch + mbpx;
+
+	for (y = 0; y < 2; ++y)
+	{
+		for (x = 0; x < 2; ++x, ++tb)
+		{
+			if ((coded_block_pattern >> (5 - tb)) & 1)
+			{
+				short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT;
+
+				if (dct_type == vlDCTTypeFrameCoded)
+				{
+					vlGrabFrameCodedBlock
+					(
+						cur_block,
+						texels + y * tex_pitch * VL_BLOCK_HEIGHT + x * VL_BLOCK_WIDTH,
+						tex_pitch
+					);
+				}
+				else
+				{
+					vlGrabFieldCodedBlock
+					(
+						cur_block,
+						texels + y * tex_pitch + x * VL_BLOCK_WIDTH,
+						tex_pitch
+					);
+				}
+
+				++sb;
+			}
+			else if (mc->zero_block[0].x < 0.0f)
+			{
+				vlGrabNoBlock(texels + y * tex_pitch * VL_BLOCK_HEIGHT + x * VL_BLOCK_WIDTH, tex_pitch);
+
+				mc->zero_block[0].x = (mbpx + x * 8) * mc->surface_tex_inv_size.x;
+				mc->zero_block[0].y = (mbpy + y * 8) * mc->surface_tex_inv_size.y;
+			}
+		}
+	}
+
+	/* TODO: Implement 422, 444 */
+	mbpx >>= 1;
+	mbpy >>= 1;
+
+	for (tb = 0; tb < 2; ++tb)
+	{
+		tex_pitch = mc->tex_surface[tb + 1]->stride / mc->tex_surface[tb + 1]->block.size;
+		texels = mc->texels[tb + 1] + mbpy * tex_pitch + mbpx;
+
+		if ((coded_block_pattern >> (1 - tb)) & 1)
+		{
+			short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT;
+
+			vlGrabFrameCodedBlock
+			(
+				cur_block,
+				texels,
+				tex_pitch
+			);
+
+			++sb;
+		}
+		else if (mc->zero_block[tb + 1].x < 0.0f)
+		{
+			vlGrabNoBlock(texels, tex_pitch);
+
+			mc->zero_block[tb + 1].x = (mbpx << 1) * mc->surface_tex_inv_size.x;
+			mc->zero_block[tb + 1].y = (mbpy << 1) * mc->surface_tex_inv_size.y;
+		}
+	}
+
+	return 0;
+}
+
+static inline enum vlMacroBlockTypeEx vlGetMacroBlockTypeEx(struct vlMpeg2MacroBlock *mb)
+{
+	assert(mb);
+
+	switch (mb->mb_type)
+	{
+		case vlMacroBlockTypeIntra:
+			return vlMacroBlockExTypeIntra;
+		case vlMacroBlockTypeFwdPredicted:
+			return mb->mo_type == vlMotionTypeFrame ?
+				vlMacroBlockExTypeFwdPredictedFrame : vlMacroBlockExTypeFwdPredictedField;
+		case vlMacroBlockTypeBkwdPredicted:
+			return mb->mo_type == vlMotionTypeFrame ?
+				vlMacroBlockExTypeBkwdPredictedFrame : vlMacroBlockExTypeBkwdPredictedField;
+		case vlMacroBlockTypeBiPredicted:
+			return mb->mo_type == vlMotionTypeFrame ?
+				vlMacroBlockExTypeBiPredictedFrame : vlMacroBlockExTypeBiPredictedField;
+		default:
+			assert(0);
+	}
+
+	/* Unreachable */
+	return -1;
+}
+
+static inline int vlGrabMacroBlock
+(
+	struct vlR16SnormBufferedMC *mc,
+	struct vlMpeg2MacroBlock *macroblock
+)
+{
+	assert(mc);
+	assert(macroblock);
+	assert(mc->num_macroblocks < mc->macroblocks_per_picture);
+
+	mc->macroblocks[mc->num_macroblocks].mbx = macroblock->mbx;
+	mc->macroblocks[mc->num_macroblocks].mby = macroblock->mby;
+	mc->macroblocks[mc->num_macroblocks].mb_type = macroblock->mb_type;
+	mc->macroblocks[mc->num_macroblocks].mo_type = macroblock->mo_type;
+	mc->macroblocks[mc->num_macroblocks].dct_type = macroblock->dct_type;
+	mc->macroblocks[mc->num_macroblocks].PMV[0][0][0] = macroblock->PMV[0][0][0];
+	mc->macroblocks[mc->num_macroblocks].PMV[0][0][1] = macroblock->PMV[0][0][1];
+	mc->macroblocks[mc->num_macroblocks].PMV[0][1][0] = macroblock->PMV[0][1][0];
+	mc->macroblocks[mc->num_macroblocks].PMV[0][1][1] = macroblock->PMV[0][1][1];
+	mc->macroblocks[mc->num_macroblocks].PMV[1][0][0] = macroblock->PMV[1][0][0];
+	mc->macroblocks[mc->num_macroblocks].PMV[1][0][1] = macroblock->PMV[1][0][1];
+	mc->macroblocks[mc->num_macroblocks].PMV[1][1][0] = macroblock->PMV[1][1][0];
+	mc->macroblocks[mc->num_macroblocks].PMV[1][1][1] = macroblock->PMV[1][1][1];
+	mc->macroblocks[mc->num_macroblocks].cbp = macroblock->cbp;
+	mc->macroblocks[mc->num_macroblocks].blocks = macroblock->blocks;
+
+	vlGrabBlocks
+	(
+		mc,
+		macroblock->mbx,
+		macroblock->mby,
+		macroblock->dct_type,
+		macroblock->cbp,
+		macroblock->blocks
+	);
+
+	mc->num_macroblocks++;
+
+	return 0;
+}
+
+#define SET_BLOCK(vb, cbp, mbx, mby, unitx, unity, ofsx, ofsy, hx, hy, lm, cbm, crm, zb)					\
+	do {															\
+	(vb)[0].pos.x = (mbx) * (unitx) + (ofsx);		(vb)[0].pos.y = (mby) * (unity) + (ofsy);			\
+	(vb)[1].pos.x = (mbx) * (unitx) + (ofsx);		(vb)[1].pos.y = (mby) * (unity) + (ofsy) + (hy);		\
+	(vb)[2].pos.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[2].pos.y = (mby) * (unity) + (ofsy);			\
+	(vb)[3].pos.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[3].pos.y = (mby) * (unity) + (ofsy);			\
+	(vb)[4].pos.x = (mbx) * (unitx) + (ofsx);		(vb)[4].pos.y = (mby) * (unity) + (ofsy) + (hy);		\
+	(vb)[5].pos.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[5].pos.y = (mby) * (unity) + (ofsy) + (hy);		\
+																\
+	if ((cbp) & (lm))													\
+	{															\
+		(vb)[0].luma_tc.x = (mbx) * (unitx) + (ofsx);		(vb)[0].luma_tc.y = (mby) * (unity) + (ofsy);		\
+		(vb)[1].luma_tc.x = (mbx) * (unitx) + (ofsx);		(vb)[1].luma_tc.y = (mby) * (unity) + (ofsy) + (hy);	\
+		(vb)[2].luma_tc.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[2].luma_tc.y = (mby) * (unity) + (ofsy);		\
+		(vb)[3].luma_tc.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[3].luma_tc.y = (mby) * (unity) + (ofsy);		\
+		(vb)[4].luma_tc.x = (mbx) * (unitx) + (ofsx);		(vb)[4].luma_tc.y = (mby) * (unity) + (ofsy) + (hy);	\
+		(vb)[5].luma_tc.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[5].luma_tc.y = (mby) * (unity) + (ofsy) + (hy);	\
+	}															\
+	else															\
+	{															\
+		(vb)[0].luma_tc.x = (zb)[0].x;		(vb)[0].luma_tc.y = (zb)[0].y;						\
+		(vb)[1].luma_tc.x = (zb)[0].x;		(vb)[1].luma_tc.y = (zb)[0].y + (hy);					\
+		(vb)[2].luma_tc.x = (zb)[0].x + (hx);	(vb)[2].luma_tc.y = (zb)[0].y;						\
+		(vb)[3].luma_tc.x = (zb)[0].x + (hx);	(vb)[3].luma_tc.y = (zb)[0].y;						\
+		(vb)[4].luma_tc.x = (zb)[0].x;		(vb)[4].luma_tc.y = (zb)[0].y + (hy);					\
+		(vb)[5].luma_tc.x = (zb)[0].x + (hx);	(vb)[5].luma_tc.y = (zb)[0].y + (hy);					\
+	}															\
+																\
+	if ((cbp) & (cbm))													\
+	{															\
+		(vb)[0].cb_tc.x = (mbx) * (unitx) + (ofsx);		(vb)[0].cb_tc.y = (mby) * (unity) + (ofsy);		\
+		(vb)[1].cb_tc.x = (mbx) * (unitx) + (ofsx);		(vb)[1].cb_tc.y = (mby) * (unity) + (ofsy) + (hy);	\
+		(vb)[2].cb_tc.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[2].cb_tc.y = (mby) * (unity) + (ofsy);		\
+		(vb)[3].cb_tc.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[3].cb_tc.y = (mby) * (unity) + (ofsy);		\
+		(vb)[4].cb_tc.x = (mbx) * (unitx) + (ofsx);		(vb)[4].cb_tc.y = (mby) * (unity) + (ofsy) + (hy);	\
+		(vb)[5].cb_tc.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[5].cb_tc.y = (mby) * (unity) + (ofsy) + (hy);	\
+	}															\
+	else															\
+	{															\
+		(vb)[0].cb_tc.x = (zb)[1].x;		(vb)[0].cb_tc.y = (zb)[1].y;						\
+		(vb)[1].cb_tc.x = (zb)[1].x;		(vb)[1].cb_tc.y = (zb)[1].y + (hy);					\
+		(vb)[2].cb_tc.x = (zb)[1].x + (hx);	(vb)[2].cb_tc.y = (zb)[1].y;						\
+		(vb)[3].cb_tc.x = (zb)[1].x + (hx);	(vb)[3].cb_tc.y = (zb)[1].y;						\
+		(vb)[4].cb_tc.x = (zb)[1].x;		(vb)[4].cb_tc.y = (zb)[1].y + (hy);					\
+		(vb)[5].cb_tc.x = (zb)[1].x + (hx);	(vb)[5].cb_tc.y = (zb)[1].y + (hy);					\
+	}															\
+																\
+	if ((cbp) & (crm))													\
+	{															\
+		(vb)[0].cr_tc.x = (mbx) * (unitx) + (ofsx);		(vb)[0].cr_tc.y = (mby) * (unity) + (ofsy);		\
+		(vb)[1].cr_tc.x = (mbx) * (unitx) + (ofsx);		(vb)[1].cr_tc.y = (mby) * (unity) + (ofsy) + (hy);	\
+		(vb)[2].cr_tc.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[2].cr_tc.y = (mby) * (unity) + (ofsy);		\
+		(vb)[3].cr_tc.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[3].cr_tc.y = (mby) * (unity) + (ofsy);		\
+		(vb)[4].cr_tc.x = (mbx) * (unitx) + (ofsx);		(vb)[4].cr_tc.y = (mby) * (unity) + (ofsy) + (hy);	\
+		(vb)[5].cr_tc.x = (mbx) * (unitx) + (ofsx) + (hx);	(vb)[5].cr_tc.y = (mby) * (unity) + (ofsy) + (hy);	\
+	}															\
+	else															\
+	{															\
+		(vb)[0].cr_tc.x = (zb)[2].x;		(vb)[0].cr_tc.y = (zb)[2].y;						\
+		(vb)[1].cr_tc.x = (zb)[2].x;		(vb)[1].cr_tc.y = (zb)[2].y + (hy);					\
+		(vb)[2].cr_tc.x = (zb)[2].x + (hx);	(vb)[2].cr_tc.y = (zb)[2].y;						\
+		(vb)[3].cr_tc.x = (zb)[2].x + (hx);	(vb)[3].cr_tc.y = (zb)[2].y;						\
+		(vb)[4].cr_tc.x = (zb)[2].x;		(vb)[4].cr_tc.y = (zb)[2].y + (hy);					\
+		(vb)[5].cr_tc.x = (zb)[2].x + (hx);	(vb)[5].cr_tc.y = (zb)[2].y + (hy);					\
+	}															\
+	} while (0)
+
+static inline int vlGenMacroblockVerts
+(
+	struct vlR16SnormBufferedMC *mc,
+	struct vlMpeg2MacroBlock *macroblock,
+	unsigned int pos,
+	struct vlMacroBlockVertexStream0 *ycbcr_vb,
+	struct vlVertex2f **ref_vb
+)
+{
+	struct vlVertex2f	mo_vec[2];
+	unsigned int		i;
+
+	assert(mc);
+	assert(macroblock);
+	assert(ycbcr_vb);
+	assert(pos < mc->macroblocks_per_picture);
+
+	switch (macroblock->mb_type)
+	{
+		case vlMacroBlockTypeBiPredicted:
+		{
+			struct vlVertex2f *vb;
+
+			assert(ref_vb && ref_vb[1]);
+
+			vb = ref_vb[1] + pos * 2 * 24;
+
+			mo_vec[0].x = macroblock->PMV[0][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+			mo_vec[0].y = macroblock->PMV[0][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+			if (macroblock->mo_type == vlMotionTypeFrame)
+			{
+				for (i = 0; i < 24 * 2; i += 2)
+				{
+					vb[i].x = mo_vec[0].x;
+					vb[i].y = mo_vec[0].y;
+				}
+			}
+			else
+			{
+				mo_vec[1].x = macroblock->PMV[1][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+				mo_vec[1].y = macroblock->PMV[1][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+				for (i = 0; i < 24 * 2; i += 2)
+				{
+					vb[i].x = mo_vec[0].x;
+					vb[i].y = mo_vec[0].y;
+					vb[i + 1].x = mo_vec[1].x;
+					vb[i + 1].y = mo_vec[1].y;
+				}
+			}
+
+			/* fall-through */
+		}
+		case vlMacroBlockTypeFwdPredicted:
+		case vlMacroBlockTypeBkwdPredicted:
+		{
+			struct vlVertex2f *vb;
+
+			assert(ref_vb && ref_vb[0]);
+
+			vb = ref_vb[0] + pos * 2 * 24;
+
+			if (macroblock->mb_type == vlMacroBlockTypeBkwdPredicted)
+			{
+				mo_vec[0].x = macroblock->PMV[0][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+				mo_vec[0].y = macroblock->PMV[0][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+				if (macroblock->mo_type == vlMotionTypeField)
+				{
+					mo_vec[1].x = macroblock->PMV[1][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+					mo_vec[1].y = macroblock->PMV[1][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+				}
+			}
+			else
+			{
+				mo_vec[0].x = macroblock->PMV[0][0][0] * 0.5f * mc->surface_tex_inv_size.x;
+				mo_vec[0].y = macroblock->PMV[0][0][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+				if (macroblock->mo_type == vlMotionTypeField)
+				{
+					mo_vec[1].x = macroblock->PMV[1][0][0] * 0.5f * mc->surface_tex_inv_size.x;
+					mo_vec[1].y = macroblock->PMV[1][0][1] * 0.5f * mc->surface_tex_inv_size.y;
+				}
+			}
+
+			if (macroblock->mo_type == vlMotionTypeFrame)
+			{
+				for (i = 0; i < 24 * 2; i += 2)
+				{
+					vb[i].x = mo_vec[0].x;
+					vb[i].y = mo_vec[0].y;
+				}
+			}
+			else
+			{
+				for (i = 0; i < 24 * 2; i += 2)
+				{
+					vb[i].x = mo_vec[0].x;
+					vb[i].y = mo_vec[0].y;
+					vb[i + 1].x = mo_vec[1].x;
+					vb[i + 1].y = mo_vec[1].y;
+				}
+			}
+
+			/* fall-through */
+		}
+		case vlMacroBlockTypeIntra:
+		{
+			const struct vlVertex2f	unit =
+			{
+				mc->surface_tex_inv_size.x * VL_MACROBLOCK_WIDTH,
+				mc->surface_tex_inv_size.y * VL_MACROBLOCK_HEIGHT
+			};
+			const struct vlVertex2f half =
+			{
+				mc->surface_tex_inv_size.x * (VL_MACROBLOCK_WIDTH / 2),
+				mc->surface_tex_inv_size.y * (VL_MACROBLOCK_HEIGHT / 2)
+			};
+
+			struct vlMacroBlockVertexStream0 *vb;
+
+			vb = ycbcr_vb + pos * 24;
+
+			SET_BLOCK
+			(
+				vb,
+				macroblock->cbp, macroblock->mbx, macroblock->mby,
+				unit.x, unit.y, 0, 0, half.x, half.y,
+				32, 2, 1, mc->zero_block
+			);
+
+			SET_BLOCK
+			(
+				vb + 6,
+				macroblock->cbp, macroblock->mbx, macroblock->mby,
+				unit.x, unit.y, half.x, 0, half.x, half.y,
+				16, 2, 1, mc->zero_block
+			);
+
+			SET_BLOCK
+			(
+				vb + 12,
+				macroblock->cbp, macroblock->mbx, macroblock->mby,
+				unit.x, unit.y, 0, half.y, half.x, half.y,
+				8, 2, 1, mc->zero_block
+			);
+
+			SET_BLOCK
+			(
+				vb + 18,
+				macroblock->cbp, macroblock->mbx, macroblock->mby,
+				unit.x, unit.y, half.x, half.y, half.x, half.y,
+				4, 2, 1, mc->zero_block
+			);
+
+			break;
+		}
+		default:
+			assert(0);
+	}
+
+	return 0;
+}
+
+static int vlFlush
+(
+	struct vlRender *render
+)
+{
+	struct vlR16SnormBufferedMC	*mc;
+	struct pipe_context		*pipe;
+	struct vlVertexShaderConsts	*vs_consts;
+	unsigned int			num_macroblocks[vlNumMacroBlockExTypes] = {0};
+	unsigned int			offset[vlNumMacroBlockExTypes];
+	unsigned int			vb_start = 0;
+	unsigned int			i;
+
+	assert(render);
+
+	mc = (struct vlR16SnormBufferedMC*)render;
+
+	if (!mc->buffered_surface)
+		return 0;
+
+	if (mc->num_macroblocks < mc->macroblocks_per_picture)
+		return 0;
+
+	assert(mc->num_macroblocks <= mc->macroblocks_per_picture);
+
+	pipe = mc->pipe;
+
+	for (i = 0; i < mc->num_macroblocks; ++i)
+	{
+		enum vlMacroBlockTypeEx mb_type_ex = vlGetMacroBlockTypeEx(&mc->macroblocks[i]);
+
+		num_macroblocks[mb_type_ex]++;
+	}
+
+	offset[0] = 0;
+
+	for (i = 1; i < vlNumMacroBlockExTypes; ++i)
+		offset[i] = offset[i - 1] + num_macroblocks[i - 1];
+
+	{
+		struct vlMacroBlockVertexStream0	*ycbcr_vb;
+		struct vlVertex2f			*ref_vb[2];
+
+		ycbcr_vb = (struct vlMacroBlockVertexStream0*)pipe_buffer_map
+		(
+			pipe->screen,
+			mc->vertex_bufs.ycbcr.buffer,
+			PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD
+		);
+
+		for (i = 0; i < 2; ++i)
+			ref_vb[i] = (struct vlVertex2f*)pipe_buffer_map
+			(
+				pipe->screen,
+				mc->vertex_bufs.ref[i].buffer,
+				PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD
+			);
+
+		for (i = 0; i < mc->num_macroblocks; ++i)
+		{
+			enum vlMacroBlockTypeEx mb_type_ex = vlGetMacroBlockTypeEx(&mc->macroblocks[i]);
+
+			vlGenMacroblockVerts(mc, &mc->macroblocks[i], offset[mb_type_ex], ycbcr_vb, ref_vb);
+
+			offset[mb_type_ex]++;
+		}
+
+		pipe_buffer_unmap(pipe->screen, mc->vertex_bufs.ycbcr.buffer);
+		for (i = 0; i < 2; ++i)
+			pipe_buffer_unmap(pipe->screen, mc->vertex_bufs.ref[i].buffer);
+	}
+
+	for (i = 0; i < 3; ++i)
+	{
+		pipe_surface_unmap(mc->tex_surface[i]);
+		pipe_surface_reference(&mc->tex_surface[i], NULL);
+	}
+
+	mc->render_target.cbufs[0] = pipe->screen->get_tex_surface
+	(
+		pipe->screen,
+		mc->buffered_surface->texture,
+		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
+	);
+
+	pipe->set_framebuffer_state(pipe, &mc->render_target);
+	pipe->set_viewport_state(pipe, &mc->viewport);
+	vs_consts = pipe_buffer_map
+	(
+		pipe->screen,
+		mc->vs_const_buf.buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD
+	);
+
+	vs_consts->denorm.x = mc->buffered_surface->texture->width[0];
+	vs_consts->denorm.y = mc->buffered_surface->texture->height[0];
+
+	pipe_buffer_unmap(pipe->screen, mc->vs_const_buf.buffer);
+	pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &mc->vs_const_buf);
+	pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &mc->fs_const_buf);
+
+	if (num_macroblocks[vlMacroBlockExTypeIntra] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 1, mc->vertex_bufs.all);
+		pipe->set_vertex_elements(pipe, 4, mc->vertex_elems);
+		pipe->set_sampler_textures(pipe, 3, mc->textures.all);
+		pipe->bind_sampler_states(pipe, 3, mc->samplers.all);
+		pipe->bind_vs_state(pipe, mc->i_vs);
+		pipe->bind_fs_state(pipe, mc->i_fs);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeIntra] * 24);
+		vb_start += num_macroblocks[vlMacroBlockExTypeIntra] * 24;
+	}
+
+	if (num_macroblocks[vlMacroBlockExTypeFwdPredictedFrame] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs.all);
+		pipe->set_vertex_elements(pipe, 6, mc->vertex_elems);
+		mc->textures.ref[0] = mc->past_surface->texture;
+		pipe->set_sampler_textures(pipe, 4, mc->textures.all);
+		pipe->bind_sampler_states(pipe, 4, mc->samplers.all);
+		pipe->bind_vs_state(pipe, mc->p_vs[0]);
+		pipe->bind_fs_state(pipe, mc->p_fs[0]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeFwdPredictedFrame] * 24);
+		vb_start += num_macroblocks[vlMacroBlockExTypeFwdPredictedFrame] * 24;
+	}
+
+	if (num_macroblocks[vlMacroBlockExTypeFwdPredictedField] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs.all);
+		pipe->set_vertex_elements(pipe, 6, mc->vertex_elems);
+		mc->textures.ref[0] = mc->past_surface->texture;
+		pipe->set_sampler_textures(pipe, 4, mc->textures.all);
+		pipe->bind_sampler_states(pipe, 4, mc->samplers.all);
+		pipe->bind_vs_state(pipe, mc->p_vs[1]);
+		pipe->bind_fs_state(pipe, mc->p_fs[1]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeFwdPredictedField] * 24);
+		vb_start += num_macroblocks[vlMacroBlockExTypeFwdPredictedField] * 24;
+	}
+
+	if (num_macroblocks[vlMacroBlockExTypeBkwdPredictedFrame] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs.all);
+		pipe->set_vertex_elements(pipe, 6, mc->vertex_elems);
+		mc->textures.ref[0] = mc->future_surface->texture;
+		pipe->set_sampler_textures(pipe, 4, mc->textures.all);
+		pipe->bind_sampler_states(pipe, 4, mc->samplers.all);
+		pipe->bind_vs_state(pipe, mc->p_vs[0]);
+		pipe->bind_fs_state(pipe, mc->p_fs[0]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeBkwdPredictedFrame] * 24);
+		vb_start += num_macroblocks[vlMacroBlockExTypeBkwdPredictedFrame] * 24;
+	}
+
+	if (num_macroblocks[vlMacroBlockExTypeBkwdPredictedField] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs.all);
+		pipe->set_vertex_elements(pipe, 6, mc->vertex_elems);
+		mc->textures.ref[0] = mc->future_surface->texture;
+		pipe->set_sampler_textures(pipe, 4, mc->textures.all);
+		pipe->bind_sampler_states(pipe, 4, mc->samplers.all);
+		pipe->bind_vs_state(pipe, mc->p_vs[1]);
+		pipe->bind_fs_state(pipe, mc->p_fs[1]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeBkwdPredictedField] * 24);
+		vb_start += num_macroblocks[vlMacroBlockExTypeBkwdPredictedField] * 24;
+	}
+
+	if (num_macroblocks[vlMacroBlockExTypeBiPredictedFrame] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 3, mc->vertex_bufs.all);
+		pipe->set_vertex_elements(pipe, 8, mc->vertex_elems);
+		mc->textures.ref[0] = mc->past_surface->texture;
+		mc->textures.ref[1] = mc->future_surface->texture;
+		pipe->set_sampler_textures(pipe, 5, mc->textures.all);
+		pipe->bind_sampler_states(pipe, 5, mc->samplers.all);
+		pipe->bind_vs_state(pipe, mc->b_vs[0]);
+		pipe->bind_fs_state(pipe, mc->b_fs[0]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeBiPredictedFrame] * 24);
+		vb_start += num_macroblocks[vlMacroBlockExTypeBiPredictedFrame] * 24;
+	}
+
+	if (num_macroblocks[vlMacroBlockExTypeBiPredictedField] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 3, mc->vertex_bufs.all);
+		pipe->set_vertex_elements(pipe, 8, mc->vertex_elems);
+		mc->textures.ref[0] = mc->past_surface->texture;
+		mc->textures.ref[1] = mc->future_surface->texture;
+		pipe->set_sampler_textures(pipe, 5, mc->textures.all);
+		pipe->bind_sampler_states(pipe, 5, mc->samplers.all);
+		pipe->bind_vs_state(pipe, mc->b_vs[1]);
+		pipe->bind_fs_state(pipe, mc->b_fs[1]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, vb_start, num_macroblocks[vlMacroBlockExTypeBiPredictedField] * 24);
+		vb_start += num_macroblocks[vlMacroBlockExTypeBiPredictedField] * 24;
+	}
+
+	pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, &mc->buffered_surface->render_fence);
+	pipe_surface_reference(&mc->render_target.cbufs[0], NULL);
+
+	for (i = 0; i < 3; ++i)
+		mc->zero_block[i].x = -1.0f;
+
+	mc->buffered_surface = NULL;
+	mc->num_macroblocks = 0;
+
+	return 0;
+}
+
+static int vlRenderMacroBlocksMpeg2R16SnormBuffered
+(
+	struct vlRender *render,
+	struct vlMpeg2MacroBlockBatch *batch,
+	struct vlSurface *surface
+)
+{
+	struct vlR16SnormBufferedMC	*mc;
+	bool				new_surface = false;
+	unsigned int			i;
+
+	assert(render);
+
+	mc = (struct vlR16SnormBufferedMC*)render;
+
+	if (mc->buffered_surface)
+	{
+		if (mc->buffered_surface != surface)
+		{
+			vlFlush(&mc->base);
+			new_surface = true;
+		}
+	}
+	else
+		new_surface = true;
+
+	if (new_surface)
+	{
+		mc->buffered_surface = surface;
+		mc->past_surface = batch->past_surface;
+		mc->future_surface = batch->future_surface;
+		mc->surface_tex_inv_size.x = 1.0f / surface->texture->width[0];
+		mc->surface_tex_inv_size.y = 1.0f / surface->texture->height[0];
+
+		for (i = 0; i < 3; ++i)
+		{
+			mc->tex_surface[i] = mc->pipe->screen->get_tex_surface
+			(
+				mc->pipe->screen,
+				mc->textures.all[i],
+				0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD
+			);
+
+			mc->texels[i] = pipe_surface_map(mc->tex_surface[i], PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_DISCARD);
+		}
+	}
+
+	for (i = 0; i < batch->num_macroblocks; ++i)
+		vlGrabMacroBlock(mc, &batch->macroblocks[i]);
+
+	return 0;
+}
+
+static inline int vlEnd
+(
+	struct vlRender *render
+)
+{
+	assert(render);
+
+	return 0;
+}
+
+static int vlDestroy
+(
+	struct vlRender *render
+)
+{
+	struct vlR16SnormBufferedMC	*mc;
+	struct pipe_context		*pipe;
+	unsigned int			i;
+
+	assert(render);
+
+	mc = (struct vlR16SnormBufferedMC*)render;
+	pipe = mc->pipe;
+
+	for (i = 0; i < 5; ++i)
+		pipe->delete_sampler_state(pipe, mc->samplers.all[i]);
+
+	for (i = 0; i < 3; ++i)
+		pipe_buffer_reference(pipe->screen, &mc->vertex_bufs.all[i].buffer, NULL);
+
+	/* Textures 3 & 4 are not created directly, no need to release them here */
+	for (i = 0; i < 3; ++i)
+		pipe_texture_reference(&mc->textures.all[i], NULL);
+
+	pipe->delete_vs_state(pipe, mc->i_vs);
+	pipe->delete_fs_state(pipe, mc->i_fs);
+
+	for (i = 0; i < 2; ++i)
+	{
+		pipe->delete_vs_state(pipe, mc->p_vs[i]);
+		pipe->delete_fs_state(pipe, mc->p_fs[i]);
+		pipe->delete_vs_state(pipe, mc->b_vs[i]);
+		pipe->delete_fs_state(pipe, mc->b_fs[i]);
+	}
+
+	pipe_buffer_reference(pipe->screen, &mc->vs_const_buf.buffer, NULL);
+	pipe_buffer_reference(pipe->screen, &mc->fs_const_buf.buffer, NULL);
+
+	FREE(mc->macroblocks);
+	FREE(mc);
+
+	return 0;
+}
+
+/*
+ * Muliplier renormalizes block samples from 16 bits to 12 bits.
+ * Divider is used when calculating Y % 2 for choosing top or bottom
+ * field for P or B macroblocks.
+ * TODO: Use immediates.
+ */
+static const struct vlFragmentShaderConsts fs_consts =
+{
+	{32767.0f / 255.0f, 32767.0f / 255.0f, 32767.0f / 255.0f, 0.0f},
+	{0.5f, 2.0f, 0.0f, 0.0f}
+};
+
+#include "vl_r16snorm_mc_buf_shaders.inc"
+
+static int vlCreateDataBufs
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int	mbw = align(mc->picture_width, VL_MACROBLOCK_WIDTH) / VL_MACROBLOCK_WIDTH;
+	const unsigned int	mbh = align(mc->picture_height, VL_MACROBLOCK_HEIGHT) / VL_MACROBLOCK_HEIGHT;
+
+	struct pipe_context	*pipe;
+	unsigned int		i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	mc->macroblocks_per_picture = mbw * mbh;
+
+	/* Create our vertex buffers */
+	mc->vertex_bufs.ycbcr.stride = sizeof(struct vlVertex2f) * 4;
+	mc->vertex_bufs.ycbcr.max_index = 24 * mc->macroblocks_per_picture - 1;
+	mc->vertex_bufs.ycbcr.buffer_offset = 0;
+	mc->vertex_bufs.ycbcr.buffer = pipe_buffer_create
+	(
+		pipe->screen,
+		DEFAULT_BUF_ALIGNMENT,
+		PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_DISCARD,
+		sizeof(struct vlVertex2f) * 4 * 24 * mc->macroblocks_per_picture
+	);
+
+	for (i = 1; i < 3; ++i)
+	{
+		mc->vertex_bufs.all[i].stride = sizeof(struct vlVertex2f) * 2;
+		mc->vertex_bufs.all[i].max_index = 24 * mc->macroblocks_per_picture - 1;
+		mc->vertex_bufs.all[i].buffer_offset = 0;
+		mc->vertex_bufs.all[i].buffer = pipe_buffer_create
+		(
+			pipe->screen,
+			DEFAULT_BUF_ALIGNMENT,
+			PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_DISCARD,
+			sizeof(struct vlVertex2f) * 2 * 24 * mc->macroblocks_per_picture
+		);
+	}
+
+	/* Position element */
+	mc->vertex_elems[0].src_offset = 0;
+	mc->vertex_elems[0].vertex_buffer_index = 0;
+	mc->vertex_elems[0].nr_components = 2;
+	mc->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* Luma, texcoord element */
+	mc->vertex_elems[1].src_offset = sizeof(struct vlVertex2f);
+	mc->vertex_elems[1].vertex_buffer_index = 0;
+	mc->vertex_elems[1].nr_components = 2;
+	mc->vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* Chroma Cr texcoord element */
+	mc->vertex_elems[2].src_offset = sizeof(struct vlVertex2f) * 2;
+	mc->vertex_elems[2].vertex_buffer_index = 0;
+	mc->vertex_elems[2].nr_components = 2;
+	mc->vertex_elems[2].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* Chroma Cb texcoord element */
+	mc->vertex_elems[3].src_offset = sizeof(struct vlVertex2f) * 3;
+	mc->vertex_elems[3].vertex_buffer_index = 0;
+	mc->vertex_elems[3].nr_components = 2;
+	mc->vertex_elems[3].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* First ref surface top field texcoord element */
+	mc->vertex_elems[4].src_offset = 0;
+	mc->vertex_elems[4].vertex_buffer_index = 1;
+	mc->vertex_elems[4].nr_components = 2;
+	mc->vertex_elems[4].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* First ref surface bottom field texcoord element */
+	mc->vertex_elems[5].src_offset = sizeof(struct vlVertex2f);
+	mc->vertex_elems[5].vertex_buffer_index = 1;
+	mc->vertex_elems[5].nr_components = 2;
+	mc->vertex_elems[5].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* Second ref surface top field texcoord element */
+	mc->vertex_elems[6].src_offset = 0;
+	mc->vertex_elems[6].vertex_buffer_index = 2;
+	mc->vertex_elems[6].nr_components = 2;
+	mc->vertex_elems[6].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* Second ref surface bottom field texcoord element */
+	mc->vertex_elems[7].src_offset = sizeof(struct vlVertex2f);
+	mc->vertex_elems[7].vertex_buffer_index = 2;
+	mc->vertex_elems[7].nr_components = 2;
+	mc->vertex_elems[7].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* Create our constant buffer */
+	mc->vs_const_buf.buffer = pipe_buffer_create
+	(
+		pipe->screen,
+		DEFAULT_BUF_ALIGNMENT,
+		PIPE_BUFFER_USAGE_CONSTANT | PIPE_BUFFER_USAGE_DISCARD,
+		sizeof(struct vlVertexShaderConsts)
+	);
+
+	mc->fs_const_buf.buffer = pipe_buffer_create
+	(
+		pipe->screen,
+		DEFAULT_BUF_ALIGNMENT,
+		PIPE_BUFFER_USAGE_CONSTANT,
+		sizeof(struct vlFragmentShaderConsts)
+	);
+
+	memcpy
+	(
+		pipe_buffer_map(pipe->screen, mc->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		&fs_consts,
+		sizeof(struct vlFragmentShaderConsts)
+	);
+
+	pipe_buffer_unmap(pipe->screen, mc->fs_const_buf.buffer);
+
+	mc->macroblocks = MALLOC(sizeof(struct vlMpeg2MacroBlock) * mc->macroblocks_per_picture);
+
+	return 0;
+}
+
+static int vlInit
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	struct pipe_context		*pipe;
+	struct pipe_sampler_state	sampler;
+	struct pipe_texture		template;
+	unsigned int			filters[5];
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+
+	mc->buffered_surface = NULL;
+	mc->past_surface = NULL;
+	mc->future_surface = NULL;
+	for (i = 0; i < 3; ++i)
+		mc->zero_block[i].x = -1.0f;
+	mc->num_macroblocks = 0;
+
+	/* For MC we render to textures, which are rounded up to nearest POT */
+	mc->viewport.scale[0] = vlRoundUpPOT(mc->picture_width);
+	mc->viewport.scale[1] = vlRoundUpPOT(mc->picture_height);
+	mc->viewport.scale[2] = 1;
+	mc->viewport.scale[3] = 1;
+	mc->viewport.translate[0] = 0;
+	mc->viewport.translate[1] = 0;
+	mc->viewport.translate[2] = 0;
+	mc->viewport.translate[3] = 0;
+
+	mc->render_target.width = vlRoundUpPOT(mc->picture_width);
+	mc->render_target.height = vlRoundUpPOT(mc->picture_height);
+	mc->render_target.nr_cbufs = 1;
+	/* FB for MC stage is a vlSurface created by the user, set at render time */
+	mc->render_target.zsbuf = NULL;
+
+	filters[0] = PIPE_TEX_FILTER_NEAREST;
+	/* FIXME: Linear causes discoloration around block edges */
+	filters[1] = /*mc->picture_format == vlFormatYCbCr444 ?*/ PIPE_TEX_FILTER_NEAREST /*: PIPE_TEX_FILTER_LINEAR*/;
+	filters[2] = /*mc->picture_format == vlFormatYCbCr444 ?*/ PIPE_TEX_FILTER_NEAREST /*: PIPE_TEX_FILTER_LINEAR*/;
+	filters[3] = PIPE_TEX_FILTER_LINEAR;
+	filters[4] = PIPE_TEX_FILTER_LINEAR;
+
+	for (i = 0; i < 5; ++i)
+	{
+		sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+		sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+		sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+		sampler.min_img_filter = filters[i];
+		sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+		sampler.mag_img_filter = filters[i];
+		sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
+		sampler.compare_func = PIPE_FUNC_ALWAYS;
+		sampler.normalized_coords = 1;
+		/*sampler.prefilter = ;*/
+		/*sampler.shadow_ambient = ;*/
+		/*sampler.lod_bias = ;*/
+		sampler.min_lod = 0;
+		/*sampler.max_lod = ;*/
+		/*sampler.border_color[i] = ;*/
+		/*sampler.max_anisotropy = ;*/
+		mc->samplers.all[i] = pipe->create_sampler_state(pipe, &sampler);
+	}
+
+	memset(&template, 0, sizeof(struct pipe_texture));
+	template.target = PIPE_TEXTURE_2D;
+	template.format = PIPE_FORMAT_R16_SNORM;
+	template.last_level = 0;
+	template.width[0] = vlRoundUpPOT(mc->picture_width);
+	template.height[0] = vlRoundUpPOT(mc->picture_height);
+	template.depth[0] = 1;
+	template.compressed = 0;
+	pf_get_block(template.format, &template.block);
+	template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_DYNAMIC;
+
+	mc->textures.y = pipe->screen->texture_create(pipe->screen, &template);
+
+	if (mc->picture_format == vlFormatYCbCr420)
+	{
+		template.width[0] = vlRoundUpPOT(mc->picture_width / 2);
+		template.height[0] = vlRoundUpPOT(mc->picture_height / 2);
+	}
+	else if (mc->picture_format == vlFormatYCbCr422)
+		template.height[0] = vlRoundUpPOT(mc->picture_height / 2);
+
+	mc->textures.cb = pipe->screen->texture_create(pipe->screen, &template);
+	mc->textures.cr = pipe->screen->texture_create(pipe->screen, &template);
+
+	/* textures.all[3] & textures.all[4] are assigned from vlSurfaces for P and B macroblocks at render time */
+
+	vlCreateVertexShaderIMB(mc);
+	vlCreateFragmentShaderIMB(mc);
+	vlCreateVertexShaderFramePMB(mc);
+	vlCreateVertexShaderFieldPMB(mc);
+	vlCreateFragmentShaderFramePMB(mc);
+	vlCreateFragmentShaderFieldPMB(mc);
+	vlCreateVertexShaderFrameBMB(mc);
+	vlCreateVertexShaderFieldBMB(mc);
+	vlCreateFragmentShaderFrameBMB(mc);
+	vlCreateFragmentShaderFieldBMB(mc);
+	vlCreateDataBufs(mc);
+
+	return 0;
+}
+
+int vlCreateR16SNormBufferedMC
+(
+	struct pipe_context *pipe,
+	unsigned int picture_width,
+	unsigned int picture_height,
+	enum vlFormat picture_format,
+	struct vlRender **render
+)
+{
+	struct vlR16SnormBufferedMC *mc;
+
+	assert(pipe);
+	assert(render);
+
+	mc = CALLOC_STRUCT(vlR16SnormBufferedMC);
+
+	mc->base.vlBegin = &vlBegin;
+	mc->base.vlRenderMacroBlocksMpeg2 = &vlRenderMacroBlocksMpeg2R16SnormBuffered;
+	mc->base.vlEnd = &vlEnd;
+	mc->base.vlFlush = &vlFlush;
+	mc->base.vlDestroy = &vlDestroy;
+	mc->pipe = pipe;
+	mc->picture_width = picture_width;
+	mc->picture_height = picture_height;
+
+	vlInit(mc);
+
+	*render = &mc->base;
+
+	return 0;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.h b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.h
new file mode 100644
index 0000000000..27177d64ca
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.h
@@ -0,0 +1,18 @@
+#ifndef vl_r16snorm_mc_buf_h
+#define vl_r16snorm_mc_buf_h
+
+#include "vl_types.h"
+
+struct pipe_context;
+struct vlRender;
+
+int vlCreateR16SNormBufferedMC
+(
+	struct pipe_context *pipe,
+	unsigned int picture_width,
+	unsigned int picture_height,
+	enum vlFormat picture_format,
+	struct vlRender **render
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf_shaders.inc b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf_shaders.inc
new file mode 100644
index 0000000000..ef4a4b2add
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf_shaders.inc
@@ -0,0 +1,1185 @@
+static int vlCreateVertexShaderIMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 50;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma Cb texcoords
+	 * decl i3		; Chroma Cr texcoords
+	 */
+	for (i = 0; i < 4; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma Cb texcoords
+	 * decl o3		; Chroma Cr texcoords
+	 */
+	for (i = 0; i < 4; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * mov o0, i0		; Move input vertex pos to output
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma Cb texcoords to output
+	 * mov o3, i3		; Move input chroma Cr texcoords to output
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->i_vs = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderIMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Luma texcoords
+	 * decl i1			; Chroma Cb texcoords
+	 * decl i2			; Chroma Cr texcoords
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header,max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i2, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul o0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->i_fs = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFramePMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma Cb texcoords
+	 * decl i3		; Chroma Cr texcoords
+	 * decl i4		; Ref surface top field texcoords
+	 * decl i5		; Ref surface bottom field texcoords (unused, packed in the same stream)
+	 */
+	for (i = 0; i < 6; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma Cb texcoords
+	 * decl o3		; Chroma Cr texcoords
+	 * decl o4		; Ref macroblock texcoords
+	 */
+	for (i = 0; i < 5; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * mov o0, i0		; Move input vertex pos to output
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma Cb texcoords to output
+	 * mov o3, i3		; Move input chroma Cr texcoords to output
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* add o4, i0, i4	; Translate vertex pos by motion vec to form ref macroblock texcoords */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 4, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, 4);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->p_vs[0] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFieldPMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma Cb texcoords
+	 * decl i3		; Chroma Cr texcoords
+	 * decl i4              ; Ref macroblock top field texcoords
+	 * decl i5              ; Ref macroblock bottom field texcoords
+	 */
+	for (i = 0; i < 6; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl c0		; Render target dimensions */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma Cb texcoords
+	 * decl o3		; Chroma Cr texcoords
+	 * decl o4		; Ref macroblock top field texcoords
+	 * decl o5		; Ref macroblock bottom field texcoords
+	 * decl o6		; Denormalized vertex pos
+	 */
+	for (i = 0; i < 7; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * mov o0, i0		; Move input vertex pos to output
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma Cb texcoords to output
+	 * mov o3, i3		; Move input chroma Cr texcoords to output
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * add o4, i0, i4	; Translate vertex pos by motion vec to form top field macroblock texcoords
+	 * add o5, i0, i5	; Translate vertex pos by motion vec to form bottom field macroblock texcoords
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 4, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, i + 4);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul o6, i0, c0	; Denorm vertex pos */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 6, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->p_vs[1] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFramePMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Luma texcoords
+	 * decl i1			; Chroma Cb texcoords
+	 * decl i2			; Chroma Cr texcoords
+	 * decl i3			; Ref macroblock texcoords
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for ref surface texture
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i2, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* tex2d t1, i3, s3		; Read texel from ref macroblock */
+	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 3, TGSI_FILE_SAMPLER, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->p_fs[0] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFieldPMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 200;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Luma texcoords
+	 * decl i1			; Chroma Cb texcoords
+	 * decl i2			; Chroma Cr texcoords
+	 * decl i3			; Ref macroblock top field texcoords
+	 * decl i4			; Ref macroblock bottom field texcoords
+	 * decl i5			; Denormalized vertex pos
+	 */
+	for (i = 0; i < 6; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
+	 * decl c1			; Constants 1/2 & 2 in .x, .y channels for Y-mod-2 top/bottom field selection
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0-t4 */
+	decl = vl_decl_temps(0, 4);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for ref surface texture
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i2, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t1, i3, s3		; Read texel from ref macroblock top field
+	 * tex2d t2, i4, s3		; Read texel from ref macroblock bottom field
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 3, TGSI_FILE_SAMPLER, 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* XXX: Pos values off by 0.5? */
+	/* sub t4, i5.y, c1.x		; Sub 0.5 from denormalized pos */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 5, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t4, c1.x		; Multiply pos Y-coord by 1/2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* floor t3, t3			; Get rid of fractional part */
+	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t3, c1.y		; Multiply by 2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->p_fs[1] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFrameBMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma Cb texcoords
+	 * decl i3		; Chroma Cr texcoords
+	 * decl i4              ; First ref macroblock top field texcoords
+	 * decl i5              ; First ref macroblock bottom field texcoords (unused, packed in the same stream)
+	 * decl i6		; Second ref macroblock top field texcoords
+	 * decl i7		; Second ref macroblock bottom field texcoords (unused, packed in the same stream)
+	 */
+	for (i = 0; i < 8; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma Cb texcoords
+	 * decl o3		; Chroma Cr texcoords
+	 * decl o4		; First ref macroblock texcoords
+	 * decl o5		; Second ref macroblock texcoords
+	 */
+	for (i = 0; i < 6; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * mov o0, i0		; Move input vertex pos to output
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma Cb texcoords to output
+	 * mov o3, i3		; Move input chroma Cr texcoords to output
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * add o4, i0, i4	; Translate vertex pos by motion vec to form first ref macroblock texcoords
+	 * add o5, i0, i6	; Translate vertex pos by motion vec to form second ref macroblock texcoords
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 4, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, (i + 2) * 2);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->b_vs[0] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFieldBMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma Cb texcoords
+	 * decl i3		; Chroma Cr texcoords
+	 * decl i4              ; First ref macroblock top field texcoords
+	 * decl i5              ; First ref macroblock bottom field texcoords
+	 * decl i6              ; Second ref macroblock top field texcoords
+	 * decl i7              ; Second ref macroblock bottom field texcoords
+	 */
+	for (i = 0; i < 8; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl c0		; Render target dimensions */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma Cb texcoords
+	 * decl o3		; Chroma Cr texcoords
+	 * decl o4		; First ref macroblock top field texcoords
+	 * decl o5		; First ref macroblock Bottom field texcoords
+	 * decl o6		; Second ref macroblock top field texcoords
+	 * decl o7		; Second ref macroblock Bottom field texcoords
+	 * decl o8		; Denormalized vertex pos
+	 */
+	for (i = 0; i < 9; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * mov o0, i0		; Move input vertex pos to output
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma Cb texcoords to output
+	 * mov o3, i3		; Move input chroma Cr texcoords to output
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * add o4, i0, i4	; Translate vertex pos by motion vec to form first top field macroblock texcoords
+	 * add o5, i0, i5	; Translate vertex pos by motion vec to form first bottom field macroblock texcoords
+	 * add o6, i0, i6	; Translate vertex pos by motion vec to form second top field macroblock texcoords
+	 * add o7, i0, i7	; Translate vertex pos by motion vec to form second bottom field macroblock texcoords
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 4, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, i + 4);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul o8, i0, c0	; Denorm vertex pos */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 8, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->b_vs[1] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFrameBMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Luma texcoords
+	 * decl i1			; Chroma Cb texcoords
+	 * decl i2			; Chroma Cr texcoords
+	 * decl i3			; First ref macroblock texcoords
+	 * decl i4			; Second ref macroblock texcoords
+	 */
+	for (i = 0; i < 5; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
+	 * decl c1			; Constant 1/2 in .x channel to use as weight to blend past and future texels
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0-t2 */
+	decl = vl_decl_temps(0, 2);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for first ref surface texture
+	 * decl s4			; Sampler for second ref surface texture
+	 */
+	for (i = 0; i < 5; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i2, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t1, i3, s3		; Read texel from first ref macroblock
+	 * tex2d t2, i4, s4		; Read texel from second ref macroblock
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 3, TGSI_FILE_SAMPLER, i + 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* lerp t1, c1.x, t1, t2	; Blend past and future texels */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->b_fs[0] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFieldBMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 200;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Luma texcoords
+	 * decl i1			; Chroma Cb texcoords
+	 * decl i2			; Chroma Cr texcoords
+	 * decl i3			; First ref macroblock top field texcoords
+	 * decl i4			; First ref macroblock bottom field texcoords
+	 * decl i5			; Second ref macroblock top field texcoords
+	 * decl i6			; Second ref macroblock bottom field texcoords
+	 * decl i7			; Denormalized vertex pos
+	 */
+	for (i = 0; i < 8; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
+	 * decl c1			; Constants 1/2 & 2 in .x, .y channels to use as weight to blend past and future texels
+	 *				; and for Y-mod-2 top/bottom field selection
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0-t5 */
+	decl = vl_decl_temps(0, 5);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for first ref surface texture
+	 * decl s4			; Sampler for second ref surface texture
+	 */
+	for (i = 0; i < 5; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i2, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* XXX: Pos values off by 0.5? */
+	/* sub t4, i7.y, c1.x		; Sub 0.5 from denormalized pos */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 7, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t4, c1.x		; Multiply pos Y-coord by 1/2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* floor t3, t3			; Get rid of fractional part */
+	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t3, c1.y		; Multiply by 2 */
+	inst = vl_inst3( TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t1, i3, s3		; Read texel from past ref macroblock top field
+	 * tex2d t2, i4, s3		; Read texel from past ref macroblock bottom field
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 3, TGSI_FILE_SAMPLER, 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t4, i5, s4		; Read texel from future ref macroblock top field
+	 * tex2d t5, i6, s4		; Read texel from future ref macroblock bottom field
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 4, TGSI_FILE_INPUT, i + 5, TGSI_FILE_SAMPLER, 4);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t2, t3, t4, t5		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 2, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 5);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* lerp t1, c1.x, t1, t2	; Blend past and future texels */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->b_fs[1] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_render.h b/src/gallium/state_trackers/g3dvl/vl_render.h
new file mode 100644
index 0000000000..166030b498
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_render.h
@@ -0,0 +1,38 @@
+#ifndef vl_render_h
+#define vl_render_h
+
+#include "vl_types.h"
+
+struct pipe_surface;
+
+struct vlRender
+{
+	int (*vlBegin)
+	(
+		struct vlRender *render
+	);
+
+	int (*vlRenderMacroBlocksMpeg2)
+	(
+		struct vlRender *render,
+		struct vlMpeg2MacroBlockBatch *batch,
+		struct vlSurface *surface
+	);
+
+	int (*vlEnd)
+	(
+		struct vlRender *render
+	);
+
+	int (*vlFlush)
+	(
+		struct vlRender *render
+	);
+
+	int (*vlDestroy)
+	(
+		struct vlRender *render
+	);
+};
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_screen.c b/src/gallium/state_trackers/g3dvl/vl_screen.c
new file mode 100644
index 0000000000..ade8643a66
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_screen.c
@@ -0,0 +1,115 @@
+#define VL_INTERNAL
+#include "vl_screen.h"
+#include <assert.h>
+#include <util/u_memory.h>
+
+int vlCreateScreen
+(
+	struct vlDisplay *display,
+	int screen,
+	struct pipe_screen *pscreen,
+	struct vlScreen **vl_screen
+)
+{
+	struct vlScreen *scrn;
+
+	assert(display);
+	assert(pscreen);
+	assert(vl_screen);
+
+	scrn = CALLOC_STRUCT(vlScreen);
+
+	if (!scrn)
+		return 1;
+
+	scrn->display = display;
+	scrn->ordinal = screen;
+	scrn->pscreen = pscreen;
+	*vl_screen = scrn;
+
+	return 0;
+}
+
+int vlDestroyScreen
+(
+	struct vlScreen *screen
+)
+{
+	assert(screen);
+
+	FREE(screen);
+
+	return 0;
+}
+
+struct vlDisplay* vlGetDisplay
+(
+	struct vlScreen *screen
+)
+{
+	assert(screen);
+
+	return screen->display;
+}
+
+struct pipe_screen* vlGetPipeScreen
+(
+	struct vlScreen *screen
+)
+{
+	assert(screen);
+
+	return screen->pscreen;
+}
+
+unsigned int vlGetMaxProfiles
+(
+	struct vlScreen *screen
+)
+{
+	assert(screen);
+
+	return vlProfileCount;
+}
+
+int vlQueryProfiles
+(
+	struct vlScreen *screen,
+	enum vlProfile *profiles
+)
+{
+	assert(screen);
+	assert(profiles);
+
+	profiles[0] = vlProfileMpeg2Simple;
+	profiles[1] = vlProfileMpeg2Main;
+
+	return 0;
+}
+
+unsigned int vlGetMaxEntryPoints
+(
+	struct vlScreen *screen
+)
+{
+	assert(screen);
+
+	return vlEntryPointCount;
+}
+
+int vlQueryEntryPoints
+(
+	struct vlScreen *screen,
+	enum vlProfile profile,
+	enum vlEntryPoint *entry_points
+)
+{
+	assert(screen);
+	assert(entry_points);
+
+	entry_points[0] = vlEntryPointIDCT;
+	entry_points[1] = vlEntryPointMC;
+	entry_points[2] = vlEntryPointCSC;
+
+	return 0;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_screen.h b/src/gallium/state_trackers/g3dvl/vl_screen.h
new file mode 100644
index 0000000000..98f3d429b6
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_screen.h
@@ -0,0 +1,63 @@
+#ifndef vl_screen_h
+#define vl_screen_h
+
+#include "vl_types.h"
+
+struct pipe_screen;
+
+#ifdef VL_INTERNAL
+struct vlScreen
+{
+	struct vlDisplay	*display;
+	unsigned int		ordinal;
+	struct pipe_screen	*pscreen;
+};
+#endif
+
+int vlCreateScreen
+(
+	struct vlDisplay *display,
+	int screen,
+	struct pipe_screen *pscreen,
+	struct vlScreen **vl_screen
+);
+
+int vlDestroyScreen
+(
+	struct vlScreen *screen
+);
+
+struct vlDisplay* vlGetDisplay
+(
+	struct vlScreen *screen
+);
+
+struct pipe_screen* vlGetPipeScreen
+(
+	struct vlScreen *screen
+);
+
+unsigned int vlGetMaxProfiles
+(
+	struct vlScreen *screen
+);
+
+int vlQueryProfiles
+(
+	struct vlScreen *screen,
+	enum vlProfile *profiles
+);
+
+unsigned int vlGetMaxEntryPoints
+(
+	struct vlScreen *screen
+);
+
+int vlQueryEntryPoints
+(
+	struct vlScreen *screen,
+	enum vlProfile profile,
+	enum vlEntryPoint *entry_points
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_shader_build.c b/src/gallium/state_trackers/g3dvl/vl_shader_build.c
new file mode 100644
index 0000000000..51f1721a33
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_shader_build.c
@@ -0,0 +1,215 @@
+#include "vl_shader_build.h"
+#include <assert.h>
+#include <tgsi/tgsi_parse.h>
+#include <tgsi/tgsi_build.h>
+
+struct tgsi_full_declaration vl_decl_input(unsigned int name, unsigned int index, unsigned int first, unsigned int last)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+
+	decl.Declaration.File = TGSI_FILE_INPUT;
+	decl.Declaration.Semantic = 1;
+	decl.Semantic.SemanticName = name;
+	decl.Semantic.SemanticIndex = index;
+	decl.DeclarationRange.First = first;
+	decl.DeclarationRange.Last = last;
+
+	return decl;
+}
+
+struct tgsi_full_declaration vl_decl_interpolated_input
+(
+	unsigned int name,
+	unsigned int index,
+	unsigned int first,
+	unsigned int last,
+	int interpolation
+)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+
+	assert
+	(
+		interpolation == TGSI_INTERPOLATE_CONSTANT ||
+		interpolation == TGSI_INTERPOLATE_LINEAR ||
+		interpolation == TGSI_INTERPOLATE_PERSPECTIVE
+	);
+
+	decl.Declaration.File = TGSI_FILE_INPUT;
+	decl.Declaration.Semantic = 1;
+	decl.Semantic.SemanticName = name;
+	decl.Semantic.SemanticIndex = index;
+	decl.Declaration.Interpolate = interpolation;;
+	decl.DeclarationRange.First = first;
+	decl.DeclarationRange.Last = last;
+
+	return decl;
+}
+
+struct tgsi_full_declaration vl_decl_constants(unsigned int name, unsigned int index, unsigned int first, unsigned int last)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+
+	decl.Declaration.File = TGSI_FILE_CONSTANT;
+	decl.Declaration.Semantic = 1;
+	decl.Semantic.SemanticName = name;
+	decl.Semantic.SemanticIndex = index;
+	decl.DeclarationRange.First = first;
+	decl.DeclarationRange.Last = last;
+
+	return decl;
+}
+
+struct tgsi_full_declaration vl_decl_output(unsigned int name, unsigned int index, unsigned int first, unsigned int last)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+
+	decl.Declaration.File = TGSI_FILE_OUTPUT;
+	decl.Declaration.Semantic = 1;
+	decl.Semantic.SemanticName = name;
+	decl.Semantic.SemanticIndex = index;
+	decl.DeclarationRange.First = first;
+	decl.DeclarationRange.Last = last;
+
+	return decl;
+}
+
+struct tgsi_full_declaration vl_decl_temps(unsigned int first, unsigned int last)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+
+	decl = tgsi_default_full_declaration();
+	decl.Declaration.File = TGSI_FILE_TEMPORARY;
+	decl.DeclarationRange.First = first;
+	decl.DeclarationRange.Last = last;
+
+	return decl;
+}
+
+struct tgsi_full_declaration vl_decl_samplers(unsigned int first, unsigned int last)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+
+	decl = tgsi_default_full_declaration();
+	decl.Declaration.File = TGSI_FILE_SAMPLER;
+	decl.DeclarationRange.First = first;
+	decl.DeclarationRange.Last = last;
+
+	return decl;
+}
+
+struct tgsi_full_instruction vl_inst2
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src_file,
+	unsigned int src_index
+)
+{
+	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
+
+	inst.Instruction.Opcode = opcode;
+	inst.Instruction.NumDstRegs = 1;
+	inst.FullDstRegisters[0].DstRegister.File = dst_file;
+	inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+	inst.Instruction.NumSrcRegs = 1;
+	inst.FullSrcRegisters[0].SrcRegister.File = src_file;
+	inst.FullSrcRegisters[0].SrcRegister.Index = src_index;
+
+	return inst;
+}
+
+struct tgsi_full_instruction vl_inst3
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index
+)
+{
+	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
+
+	inst.Instruction.Opcode = opcode;
+	inst.Instruction.NumDstRegs = 1;
+	inst.FullDstRegisters[0].DstRegister.File = dst_file;
+	inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+	inst.Instruction.NumSrcRegs = 2;
+	inst.FullSrcRegisters[0].SrcRegister.File = src1_file;
+	inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
+	inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
+	inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
+
+	return inst;
+}
+
+struct tgsi_full_instruction vl_tex
+(
+	int tex,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index
+)
+{
+	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
+
+	inst.Instruction.Opcode = TGSI_OPCODE_TEX;
+	inst.Instruction.NumDstRegs = 1;
+	inst.FullDstRegisters[0].DstRegister.File = dst_file;
+	inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+	inst.Instruction.NumSrcRegs = 2;
+	inst.InstructionExtTexture.Texture = tex;
+	inst.FullSrcRegisters[0].SrcRegister.File = src1_file;
+	inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
+	inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
+	inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
+
+	return inst;
+}
+
+struct tgsi_full_instruction vl_inst4
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index,
+	enum tgsi_file_type src3_file,
+	unsigned int src3_index
+)
+{
+	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
+
+	inst.Instruction.Opcode = opcode;
+	inst.Instruction.NumDstRegs = 1;
+	inst.FullDstRegisters[0].DstRegister.File = dst_file;
+	inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+	inst.Instruction.NumSrcRegs = 3;
+	inst.FullSrcRegisters[0].SrcRegister.File = src1_file;
+	inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
+	inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
+	inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
+	inst.FullSrcRegisters[2].SrcRegister.File = src3_file;
+	inst.FullSrcRegisters[2].SrcRegister.Index = src3_index;
+
+	return inst;
+}
+
+struct tgsi_full_instruction vl_end(void)
+{
+	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
+
+	inst.Instruction.Opcode = TGSI_OPCODE_END;
+	inst.Instruction.NumDstRegs = 0;
+	inst.Instruction.NumSrcRegs = 0;
+
+	return inst;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_shader_build.h b/src/gallium/state_trackers/g3dvl/vl_shader_build.h
new file mode 100644
index 0000000000..dc615cb156
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_shader_build.h
@@ -0,0 +1,61 @@
+#ifndef vl_shader_build_h
+#define vl_shader_build_h
+
+#include <pipe/p_shader_tokens.h>
+
+struct tgsi_full_declaration vl_decl_input(unsigned int name, unsigned int index, unsigned int first, unsigned int last);
+struct tgsi_full_declaration vl_decl_interpolated_input
+(
+	unsigned int name,
+	unsigned int index,
+	unsigned int first,
+	unsigned int last,
+	int interpolation
+);
+struct tgsi_full_declaration vl_decl_constants(unsigned int name, unsigned int index, unsigned int first, unsigned int last);
+struct tgsi_full_declaration vl_decl_output(unsigned int name, unsigned int index, unsigned int first, unsigned int last);
+struct tgsi_full_declaration vl_decl_temps(unsigned int first, unsigned int last);
+struct tgsi_full_declaration vl_decl_samplers(unsigned int first, unsigned int last);
+struct tgsi_full_instruction vl_inst2
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src_file,
+	unsigned int src_index
+);
+struct tgsi_full_instruction vl_inst3
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index
+);
+struct tgsi_full_instruction vl_tex
+(
+	int tex,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index
+);
+struct tgsi_full_instruction vl_inst4
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index,
+	enum tgsi_file_type src3_file,
+	unsigned int src3_index
+);
+struct tgsi_full_instruction vl_end(void);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.c b/src/gallium/state_trackers/g3dvl/vl_surface.c
new file mode 100644
index 0000000000..92388f7978
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_surface.c
@@ -0,0 +1,243 @@
+#define VL_INTERNAL
+#include "vl_surface.h"
+#include <assert.h>
+#include <string.h>
+#include <pipe/p_screen.h>
+#include <pipe/p_state.h>
+#include <pipe/p_inlines.h>
+#include <util/u_memory.h>
+#include <vl_winsys.h>
+#include "vl_screen.h"
+#include "vl_context.h"
+#include "vl_render.h"
+#include "vl_csc.h"
+#include "vl_util.h"
+
+int vlCreateSurface
+(
+	struct vlScreen *screen,
+	unsigned int width,
+	unsigned int height,
+	enum vlFormat format,
+	struct vlSurface **surface
+)
+{
+	struct vlSurface	*sfc;
+	struct pipe_texture	template;
+
+	assert(screen);
+	assert(surface);
+
+	sfc = CALLOC_STRUCT(vlSurface);
+
+	if (!sfc)
+		return 1;
+
+	sfc->screen = screen;
+	sfc->width = width;
+	sfc->height = height;
+	sfc->format = format;
+
+	memset(&template, 0, sizeof(struct pipe_texture));
+	template.target = PIPE_TEXTURE_2D;
+	template.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+	template.last_level = 0;
+	template.width[0] = vlRoundUpPOT(sfc->width);
+	template.height[0] = vlRoundUpPOT(sfc->height);
+	template.depth[0] = 1;
+	template.compressed = 0;
+	pf_get_block(template.format, &template.block);
+	template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_RENDER_TARGET;
+
+	sfc->texture = vlGetPipeScreen(screen)->texture_create(vlGetPipeScreen(screen), &template);
+
+	if (!sfc->texture)
+	{
+		FREE(sfc);
+		return 1;
+	}
+
+	*surface = sfc;
+
+	return 0;
+}
+
+int vlDestroySurface
+(
+	struct vlSurface *surface
+)
+{
+	assert(surface);
+
+	pipe_texture_reference(&surface->texture, NULL);
+	FREE(surface);
+
+	return 0;
+}
+
+int vlRenderMacroBlocksMpeg2
+(
+	struct vlMpeg2MacroBlockBatch *batch,
+	struct vlSurface *surface
+)
+{
+	assert(batch);
+	assert(surface);
+	assert(surface->context);
+
+	surface->context->render->vlBegin(surface->context->render);
+
+	surface->context->render->vlRenderMacroBlocksMpeg2
+	(
+		surface->context->render,
+		batch,
+		surface
+	);
+
+	surface->context->render->vlEnd(surface->context->render);
+
+	return 0;
+}
+
+int vlPutPicture
+(
+	struct vlSurface *surface,
+	vlNativeDrawable drawable,
+	int srcx,
+	int srcy,
+	int srcw,
+	int srch,
+	int destx,
+	int desty,
+	int destw,
+	int desth,
+	int drawable_w,
+	int drawable_h,
+	enum vlPictureType picture_type
+)
+{
+	struct vlCSC		*csc;
+	struct pipe_context	*pipe;
+
+	assert(surface);
+	assert(surface->context);
+
+	surface->context->render->vlFlush(surface->context->render);
+
+	csc = surface->context->csc;
+	pipe = surface->context->pipe;
+
+	csc->vlResizeFrameBuffer(csc, drawable_w, drawable_h);
+
+	csc->vlBegin(csc);
+
+	csc->vlPutPicture
+	(
+		csc,
+		surface,
+		srcx,
+		srcy,
+		srcw,
+		srch,
+		destx,
+		desty,
+		destw,
+		desth,
+		picture_type
+	);
+
+	csc->vlEnd(csc);
+
+	pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, &surface->disp_fence);
+
+	bind_pipe_drawable(pipe, drawable);
+
+	pipe->screen->flush_frontbuffer
+	(
+		pipe->screen,
+		csc->vlGetFrameBuffer(csc),
+		pipe->priv
+	);
+
+	return 0;
+}
+
+int vlSurfaceGetStatus
+(
+	struct vlSurface *surface,
+	enum vlResourceStatus *status
+)
+{
+	assert(surface);
+	assert(surface->context);
+	assert(status);
+
+	if (surface->render_fence && !surface->context->pipe->screen->fence_signalled(surface->context->pipe->screen, surface->render_fence, 0))
+	{
+		*status = vlResourceStatusRendering;
+		return 0;
+	}
+
+	if (surface->disp_fence && !surface->context->pipe->screen->fence_signalled(surface->context->pipe->screen, surface->disp_fence, 0))
+	{
+		*status = vlResourceStatusDisplaying;
+		return 0;
+	}
+
+	*status = vlResourceStatusFree;
+
+	return 0;
+}
+
+int vlSurfaceFlush
+(
+	struct vlSurface *surface
+)
+{
+	assert(surface);
+	assert(surface->context);
+
+	surface->context->render->vlFlush(surface->context->render);
+
+	return 0;
+}
+
+int vlSurfaceSync
+(
+	struct vlSurface *surface
+)
+{
+	assert(surface);
+	assert(surface->context);
+	assert(surface->render_fence);
+
+	surface->context->pipe->screen->fence_finish(surface->context->pipe->screen, surface->render_fence, 0);
+
+	return 0;
+}
+
+struct vlScreen* vlSurfaceGetScreen
+(
+	struct vlSurface *surface
+)
+{
+	assert(surface);
+
+	return surface->screen;
+}
+
+struct vlContext* vlBindToContext
+(
+	struct vlSurface *surface,
+	struct vlContext *context
+)
+{
+	struct vlContext *old;
+
+	assert(surface);
+
+	old = surface->context;
+	surface->context = context;
+
+	return old;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.h b/src/gallium/state_trackers/g3dvl/vl_surface.h
new file mode 100644
index 0000000000..133e1515ef
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_surface.h
@@ -0,0 +1,86 @@
+#ifndef vl_surface_h
+#define vl_surface_h
+
+#include "vl_types.h"
+
+#ifdef VL_INTERNAL
+struct pipe_texture;
+
+struct vlSurface
+{
+	struct vlScreen			*screen;
+	struct vlContext		*context;
+	unsigned int			width;
+	unsigned int			height;
+	enum vlFormat			format;
+	struct pipe_texture		*texture;
+	struct pipe_fence_handle	*render_fence;
+	struct pipe_fence_handle	*disp_fence;
+};
+#endif
+
+int vlCreateSurface
+(
+	struct vlScreen *screen,
+	unsigned int width,
+	unsigned int height,
+	enum vlFormat format,
+	struct vlSurface **surface
+);
+
+int vlDestroySurface
+(
+	struct vlSurface *surface
+);
+
+int vlRenderMacroBlocksMpeg2
+(
+	struct vlMpeg2MacroBlockBatch *batch,
+	struct vlSurface *surface
+);
+
+int vlPutPicture
+(
+	struct vlSurface *surface,
+	vlNativeDrawable drawable,
+	int srcx,
+	int srcy,
+	int srcw,
+	int srch,
+	int destx,
+	int desty,
+	int destw,
+	int desth,
+	int drawable_w,
+	int drawable_h,
+	enum vlPictureType picture_type
+);
+
+int vlSurfaceGetStatus
+(
+	struct vlSurface *surface,
+	enum vlResourceStatus *status
+);
+
+int vlSurfaceFlush
+(
+	struct vlSurface *surface
+);
+
+int vlSurfaceSync
+(
+	struct vlSurface *surface
+);
+
+struct vlScreen* vlSurfaceGetScreen
+(
+	struct vlSurface *surface
+);
+
+struct vlContext* vlBindToContext
+(
+	struct vlSurface *surface,
+	struct vlContext *context
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_types.h b/src/gallium/state_trackers/g3dvl/vl_types.h
new file mode 100644
index 0000000000..274e1f7437
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_types.h
@@ -0,0 +1,115 @@
+#ifndef vl_types_h
+#define vl_types_h
+
+#if 1 /*#ifdef X11*/
+#include <X11/Xlib.h>
+
+typedef Display* vlNativeDisplay;
+typedef Drawable vlNativeDrawable;
+#endif
+
+struct vlDisplay;
+struct vlScreen;
+struct vlContext;
+struct vlSurface;
+
+enum vlResourceStatus
+{
+	vlResourceStatusFree,
+	vlResourceStatusRendering,
+	vlResourceStatusDisplaying
+};
+
+enum vlProfile
+{
+	vlProfileMpeg2Simple,
+	vlProfileMpeg2Main,
+
+	vlProfileCount
+};
+
+enum vlEntryPoint
+{
+	vlEntryPointIDCT,
+	vlEntryPointMC,
+	vlEntryPointCSC,
+
+	vlEntryPointCount
+};
+
+enum vlFormat
+{
+	vlFormatYCbCr420,
+	vlFormatYCbCr422,
+	vlFormatYCbCr444
+};
+
+enum vlPictureType
+{
+	vlPictureTypeTopField,
+	vlPictureTypeBottomField,
+	vlPictureTypeFrame
+};
+
+enum vlMotionType
+{
+	vlMotionTypeField,
+	vlMotionTypeFrame,
+	vlMotionTypeDualPrime,
+	vlMotionType16x8
+};
+
+enum vlFieldOrder
+{
+	vlFieldOrderFirst,
+	vlFieldOrderSecond
+};
+
+enum vlDCTType
+{
+	vlDCTTypeFrameCoded,
+	vlDCTTypeFieldCoded
+};
+
+struct vlVertex2f
+{
+	float x, y;
+};
+
+struct vlVertex4f
+{
+	float x, y, z, w;
+};
+
+enum vlMacroBlockType
+{
+	vlMacroBlockTypeIntra,
+	vlMacroBlockTypeFwdPredicted,
+	vlMacroBlockTypeBkwdPredicted,
+	vlMacroBlockTypeBiPredicted,
+
+	vlNumMacroBlockTypes
+};
+
+struct vlMpeg2MacroBlock
+{
+	unsigned int		mbx, mby;
+	enum vlMacroBlockType	mb_type;
+	enum vlMotionType	mo_type;
+	enum vlDCTType		dct_type;
+	int			PMV[2][2][2];
+	unsigned int		cbp;
+	short			*blocks;
+};
+
+struct vlMpeg2MacroBlockBatch
+{
+	struct vlSurface		*past_surface;
+	struct vlSurface		*future_surface;
+	enum vlPictureType		picture_type;
+	enum vlFieldOrder		field_order;
+	unsigned int			num_macroblocks;
+	struct vlMpeg2MacroBlock	*macroblocks;
+};
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_util.c b/src/gallium/state_trackers/g3dvl/vl_util.c
new file mode 100644
index 0000000000..50aa9af66f
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_util.c
@@ -0,0 +1,16 @@
+#include "vl_util.h"
+#include <assert.h>
+
+unsigned int vlRoundUpPOT(unsigned int x)
+{
+	unsigned int i;
+
+	assert(x > 0);
+
+	--x;
+
+	for (i = 1; i < sizeof(unsigned int) * 8; i <<= 1)
+		x |= x >> i;
+
+	return x + 1;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_util.h b/src/gallium/state_trackers/g3dvl/vl_util.h
new file mode 100644
index 0000000000..bc98e79df4
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_util.h
@@ -0,0 +1,6 @@
+#ifndef vl_util_h
+#define vl_util_h
+
+unsigned int vlRoundUpPOT(unsigned int x);
+
+#endif
diff --git a/src/gallium/state_trackers/glx/Makefile b/src/gallium/state_trackers/glx/Makefile
new file mode 100644
index 0000000000..f779035763
--- /dev/null
+++ b/src/gallium/state_trackers/glx/Makefile
@@ -0,0 +1,25 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+
+SUBDIRS = xlib
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	rm -f `find . -name \*.[oa]`
+	rm -f `find . -name depend`
+
+
+# Dummy install target
+install:
diff --git a/src/gallium/state_trackers/glx/dri/dri_context.c b/src/gallium/state_trackers/glx/dri/dri_context.c
new file mode 100644
index 0000000000..9424e18bee
--- /dev/null
+++ b/src/gallium/state_trackers/glx/dri/dri_context.c
@@ -0,0 +1,168 @@
+/**************************************************************************
+ *
+ * Copyright 2009, VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "dri_screen.h"
+#include "dri_context.h"
+#include "dri_winsys.h"
+
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_context.h"
+#include "pipe/p_context.h"
+
+#include "util/u_memory.h"
+
+
+GLboolean
+dri_create_context(const __GLcontextModes *visual,
+                   __DRIcontextPrivate *cPriv,
+                   void *sharedContextPrivate)
+{
+   __DRIscreenPrivate *sPriv = cPriv->driScreenPriv;
+   struct dri_screen *screen = dri_screen(sPriv);
+   struct dri_context *ctx = NULL;
+   struct st_context *st_share = NULL;
+
+   if (sharedContextPrivate) {
+      st_share = ((struct dri_context *) sharedContextPrivate)->st;
+   }
+
+   ctx = CALLOC_STRUCT(dri_context);
+   if (ctx == NULL)
+      goto fail;
+
+   cPriv->driverPrivate = ctx;
+   ctx->cPriv = cPriv;
+   ctx->sPriv = sPriv;
+
+   driParseConfigFiles(&ctx->optionCache, 
+                       &screen->optionCache,
+                       sPriv->myNum, 
+                       "dri");
+   
+   ctx->pipe = screen->pipe_screen->create_context(screen->pipe_screen,
+                                                   screen->pipe_winsys,
+                                                   hw_winsys );
+   if (ctx->pipe == NULL)
+      goto fail;
+
+   ctx->pipe->priv = ctx;       /* I guess */
+
+   ctx->st = st_create_context(ctx->pipe, visual, st_share);
+   if (ctx->st == NULL)
+      goto fail;
+
+   dri_init_extensions( ctx );
+
+   return GL_TRUE;
+
+fail:
+   if (ctx && ctx->st)
+      st_destroy_context( ctx->st );
+
+   if (ctx && ctx->pipe)
+      ctx->pipe->destroy( ctx->pipe );
+
+   FREE(ctx);
+   return FALSE;
+}
+
+
+void
+dri_destroy_context(__DRIcontextPrivate *cPriv)
+{
+   struct dri_context *ctx = dri_context(cPriv);
+   struct dri_screen *screen = dri_screen(cPriv->driScreenPriv);
+   struct pipe_winsys *winsys = screen->winsys;
+
+   /* No particular reason to wait for command completion before
+    * destroying a context, but it is probably worthwhile flushing it
+    * to avoid having to add code elsewhere to cope with flushing a
+    * partially destroyed context.
+    */
+   st_flush(ctx->st);
+
+   if (screen->dummyContext == ctx)
+      screen->dummyContext = NULL;
+
+   /* Also frees ctx->pipe?
+    */
+   st_destroy_context(ctx->st);
+
+   FREE(ctx);
+}
+
+
+GLboolean
+dri_unbind_context(__DRIcontextPrivate *cPriv)
+{
+   struct dri_context *ctx = dri_context(cPriv);
+   st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+   /* XXX make_current(NULL)? */
+   return GL_TRUE;
+}
+
+
+GLboolean
+dri_make_current(__DRIcontextPrivate *cPriv,
+                 __DRIdrawablePrivate *driDrawPriv,
+                 __DRIdrawablePrivate *driReadPriv)
+{
+   if (cPriv) {
+      struct dri_context *ctx = dri_context(cPriv);
+      struct dri_screen *screen = dri_screen(cPriv->driScreenPriv);
+      struct dri_drawable *draw = dri_drawable(driDrawPriv);
+      struct dri_drawable *read = dri_drawable(driReadPriv);
+
+      /* This is for situations in which we need a rendering context but
+       * there may not be any currently bound.
+       */
+      screen->dummyContext = ctx;
+
+      st_make_current( ctx->st, 
+                       draw->stfb, 
+                       read->stfb );
+
+      ctx->dPriv = driDrawPriv;
+
+      /* Update window sizes if necessary:
+       */
+      if (draw->stamp != driDrawPriv->lastStamp) {
+         dri_update_window_size( draw );
+      }
+
+      if (read->stamp != driReadPriv->lastStamp) {
+         dri_update_window_size( read );
+      }
+
+   }
+   else {
+      st_make_current(NULL, NULL, NULL);
+   }
+
+   return GL_TRUE;
+}
diff --git a/src/gallium/state_trackers/glx/dri/dri_context.h b/src/gallium/state_trackers/glx/dri/dri_context.h
new file mode 100644
index 0000000000..4e6a305abb
--- /dev/null
+++ b/src/gallium/state_trackers/glx/dri/dri_context.h
@@ -0,0 +1,95 @@
+/**************************************************************************
+ *
+ * Copyright (C) 2009 VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef DRI_CONTEXT_H
+#define DRI_CONTEXT_H
+
+#include "pipe/p_compiler.h"
+#include "drm.h"
+#include "dri_util.h"
+
+
+struct pipe_context;
+struct pipe_fence;
+struct st_context;
+
+
+struct dri_context
+{
+   __DRIcontextPrivate *cPriv;
+   __DRIdrawablePrivate *dPriv;
+
+   struct st_context *st;
+   struct pipe_context *pipe;
+
+   boolean locked;
+
+   /**
+    * Configuration cache
+    */
+   driOptionCache optionCache;
+};
+
+
+static INLINE struct dri_context *
+dri_context(__DRIcontextPrivate *driContextPriv)
+{
+   return (struct dri_context *) driContextPriv->driverPrivate;
+}
+
+/***********************************************************************
+ * dri_context.c
+ */
+void 
+dri_destroy_context(__DRIcontextPrivate * driContextPriv);
+
+boolean 
+dri_unbind_context(__DRIcontextPrivate * driContextPriv);
+
+boolean
+dri_make_current(__DRIcontextPrivate * driContextPriv,
+                 __DRIdrawablePrivate * driDrawPriv,
+                 __DRIdrawablePrivate * driReadPriv);
+
+boolean
+dri_create_context(const __GLcontextModes * visual,
+                   __DRIcontextPrivate * driContextPriv,
+                   void *sharedContextPrivate);
+
+
+
+/***********************************************************************
+ * dri_lock.c
+ */
+void dri_lock_hardware( struct dri_context *context, 
+                        struct dri_drawable *drawable );
+
+void dri_unlock_hardware( struct dri_context *dri );
+boolean dri_is_locked( struct dri_context *dri );
+
+
+
+#endif
diff --git a/src/gallium/state_trackers/glx/dri/dri_drawable.c b/src/gallium/state_trackers/glx/dri/dri_drawable.c
new file mode 100644
index 0000000000..b712acda88
--- /dev/null
+++ b/src/gallium/state_trackers/glx/dri/dri_drawable.c
@@ -0,0 +1,363 @@
+/**************************************************************************
+ *
+ * Copyright 2009, VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "dri_screen.h"
+#include "dri_context.h"
+#include "dri_swapbuffers.h"
+
+#include "pipe/p_context.h"
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_context.h"
+#include "state_tracker/st_cb_fbo.h"
+
+
+static void
+blit_swapbuffers(__DRIdrawablePrivate *dPriv,
+                 __DRIcontextPrivate *cPriv,
+		 struct pipe_surface *src,
+		 const drm_clip_rect_t *rect)
+{
+   struct dri_screen *screen = dri_screen(dPriv->driScreenPriv);
+   struct dri_drawable *fb = dri_drawable(dPriv);
+   struct dri_context *context = dri_context(cPriv);
+
+   const int nbox = dPriv->numClipRects;
+   const drm_clip_rect_t *pbox = dPriv->pClipRects;
+
+   struct pipe_surface *dest = fb->front_surface;
+   const int backWidth = fb->stfb->Base.Width;
+   const int backHeight = fb->stfb->Base.Height;
+   int i;
+
+   for (i = 0; i < nbox; i++, pbox++) {
+      drm_clip_rect_t box;
+      drm_clip_rect_t sbox;
+	 
+      if (pbox->x1 > pbox->x2 ||
+	  pbox->y1 > pbox->y2 ||
+	  (pbox->x2 - pbox->x1) > dest->width || 
+	  (pbox->y2 - pbox->y1) > dest->height) 
+	 continue;
+
+      box = *pbox;
+
+      if (rect) {
+	 drm_clip_rect_t rrect;
+
+	 rrect.x1 = dPriv->x + rect->x1;
+	 rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y;
+	 rrect.x2 = rect->x2 + rrect.x1;
+	 rrect.y2 = rect->y2 + rrect.y1;
+	 if (rrect.x1 > box.x1)
+	    box.x1 = rrect.x1;
+	 if (rrect.y1 > box.y1)
+	    box.y1 = rrect.y1;
+	 if (rrect.x2 < box.x2)
+	    box.x2 = rrect.x2;
+	 if (rrect.y2 < box.y2)
+	    box.y2 = rrect.y2;
+
+	 if (box.x1 > box.x2 || box.y1 > box.y2)
+	    continue;
+      }
+
+      /* restrict blit to size of actually rendered area */
+      if (box.x2 - box.x1 > backWidth)
+	 box.x2 = backWidth + box.x1;
+      if (box.y2 - box.y1 > backHeight)
+	 box.y2 = backHeight + box.y1;
+
+      debug_printf("%s: box %d,%d-%d,%d\n", __FUNCTION__,
+                   box.x1, box.y1, box.x2, box.y2);
+
+      sbox.x1 = box.x1 - dPriv->x;
+      sbox.y1 = box.y1 - dPriv->y;
+
+      ctx->st->pipe->surface_copy( ctx->st->pipe,
+                                   FALSE,
+                                   dest,
+                                   box.x1, box.y1,
+                                   src,
+                                   sbox.x1, sbox.y1,
+                                   box.x2 - box.x1, 
+                                   box.y2 - box.y1 );
+   }
+}
+
+/**
+ * Display a colorbuffer surface in an X window.
+ * Used for SwapBuffers and flushing front buffer rendering.
+ *
+ * \param dPriv  the window/drawable to display into
+ * \param surf  the surface to display
+ * \param rect  optional subrect of surface to display (may be NULL).
+ */
+static void
+dri_display_surface(__DRIdrawablePrivate *dPriv,
+                    struct pipe_surface *source,
+                    const drm_clip_rect_t *rect)
+{
+   struct dri_drawable *drawable = dri_drawable(dPriv);
+   struct dri_screen *screen = dri_screen(dPriv->driScreenPriv);
+   struct dri_context *context = screen->dummy_context;
+   struct pipe_winsys *winsys = screen->winsys;
+      
+   if (!context) 
+      return;
+
+   if (drawable->last_swap_fence) {
+      winsys->fence_finish( winsys,
+                            drawable->last_swap_fence,
+                            0 );
+
+      winsys->fence_reference( winsys,
+                               &drawable->last_swap_fence,
+                               NULL );
+   }
+
+   drawable->last_swap_fence = drawable->first_swap_fence;
+   drawable->first_swap_fence = NULL;
+
+   /* Call lock_hardware to update dPriv cliprects. 
+    */
+   dri_lock_hardware(context, drawable);
+   {
+      if (dPriv->numClipRects) {
+         blit_swapbuffers( context, dPriv, source, rect );
+      }
+   }
+   dri_unlock_hardware(context);
+
+   if (drawble->stamp != drawable->dPriv->lastStamp) {
+      dri_update_window_size( dpriv );
+   }
+}
+
+
+
+/**
+ * This will be called a drawable is known to have moved/resized.
+ */
+void
+dri_update_window_size(__DRIdrawablePrivate *dPriv)
+{
+   struct dri_drawable *drawable = dri_drawable(dPriv);
+   st_resize_framebuffer(drawable->stfb, dPriv->w, dPriv->h);
+   drawable->stamp = dPriv->lastStamp;
+}
+
+
+
+void
+dri_swap_buffers(__DRIdrawablePrivate * dPriv)
+{
+   struct dri_drawable *drawable = dri_drawable(dPriv);
+   struct pipe_surface *back_surf;
+
+   assert(drawable);
+   assert(drawable->stfb);
+
+   back_surf = st_get_framebuffer_surface(drawable->stfb,
+                                          ST_SURFACE_BACK_LEFT);
+   if (back_surf) {
+      st_notify_swapbuffers(drawable->stfb);
+      dri_display_surface(dPriv, back_surf, NULL);
+      st_notify_swapbuffers_complete(drawable->stfb);
+   }
+}
+
+
+/**
+ * Called via glXCopySubBufferMESA() to copy a subrect of the back
+ * buffer to the front buffer/screen.
+ */
+void
+dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h)
+{
+   struct dri_drawable *drawable = dri_drawable(dPriv);
+   struct pipe_surface *back_surf;
+
+   assert(drawable);
+   assert(drawable->stfb);
+
+   back_surf = st_get_framebuffer_surface(drawable->stfb,
+                                          ST_SURFACE_BACK_LEFT);
+   if (back_surf) {
+      drm_clip_rect_t rect;
+      rect.x1 = x;
+      rect.y1 = y;
+      rect.x2 = w;
+      rect.y2 = h;
+
+      st_notify_swapbuffers(drawable->stfb);
+      dri_display_surface(dPriv, back_surf, &rect);
+   }
+}
+
+
+
+/*
+ * The state tracker keeps track of whether the fake frontbuffer has
+ * been touched by any rendering since the last time we copied its
+ * contents to the real frontbuffer.  Our task is easy:
+ */
+static void
+dri_flush_frontbuffer( struct pipe_winsys *winsys,
+                       struct pipe_surface *surf,
+                       void *context_private)
+{
+   struct dri_context *dri = (struct dri_context *) context_private;
+   __DRIdrawablePrivate *dPriv = dri->driDrawable;
+
+   dri_display_surface(dPriv, surf, NULL);
+}
+
+
+
+/* Need to create a surface which wraps the front surface to support
+ * client-side swapbuffers.
+ */
+static void
+dri_create_front_surface(struct dri_screen *screen, 
+                         struct pipe_winsys *winsys, 
+                         unsigned handle)
+{
+   struct pipe_screen *pipe_screen = screen->pipe_screen;
+   struct pipe_texture *texture;
+   struct pipe_texture templat;
+   struct pipe_surface *surface;
+   struct pipe_buffer *buffer;
+   unsigned pitch;
+
+   assert(screen->front.cpp == 4);
+
+//   buffer = dri_buffer_from_handle(screen->winsys,
+//                                        "front", handle);
+
+   if (!buffer)
+      return;
+
+   screen->front.buffer = dri_bo(buffer);
+
+   memset(&templat, 0, sizeof(templat));
+   templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+   templat.target = PIPE_TEXTURE_2D;
+   templat.last_level = 0;
+   templat.depth[0] = 1;
+   templat.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+   templat.width[0] = screen->front.width;
+   templat.height[0] = screen->front.height;
+   pf_get_block(templat.format, &templat.block);
+   pitch = screen->front.pitch;
+
+   texture = pipe_screen->texture_blanket(pipe_screen,
+                                          &templat,
+                                          &pitch,
+                                          buffer);
+
+   /* Unref the buffer we don't need it anyways */
+   pipe_buffer_reference(screen, &buffer, NULL);
+
+   surface = pipe_screen->get_tex_surface(pipe_screen,
+                                          texture,
+                                          0,
+                                          0,
+                                          0,
+                                          PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   screen->front.texture = texture;
+   screen->front.surface = surface;
+}
+
+/**
+ * This is called when we need to set up GL rendering to a new X window.
+ */
+static boolean
+dri_create_buffer(__DRIscreenPrivate *sPriv,
+                  __DRIdrawablePrivate *dPriv,
+                  const __GLcontextModes *visual,
+                  boolean isPixmap)
+{
+   enum pipe_format colorFormat, depthFormat, stencilFormat;
+   struct dri_drawable *drawable;
+
+   if (isPixmap) 
+      goto fail;          /* not implemented */
+
+   drawable = CALLOC_STRUCT(dri_drawable);
+   if (drawable == NULL)
+      goto fail;
+
+   /* XXX: todo: use the pipe_screen queries to figure out which
+    * render targets are supportable.
+    */
+   if (visual->redBits == 5)
+      colorFormat = PIPE_FORMAT_R5G6B5_UNORM;
+   else
+      colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+   if (visual->depthBits == 16)
+      depthFormat = PIPE_FORMAT_Z16_UNORM;
+   else if (visual->depthBits == 24) {
+      if (visual->stencilBits == 8)
+         depthFormat = PIPE_FORMAT_S8Z24_UNORM;
+      else
+         depthFormat = PIPE_FORMAT_X8Z24_UNORM;
+   }
+
+   drawable->stfb = st_create_framebuffer(visual,
+                                    colorFormat,
+                                    depthFormat,
+                                    dPriv->w,
+                                    dPriv->h,
+                                    (void*) drawable);
+   if (drawable->stfb == NULL)
+      goto fail;
+
+   dPriv->driverPrivate = (void *) drawable;
+   return GL_TRUE;
+
+fail:
+   FREE(drawable);
+   return GL_FALSE;
+}
+
+static void
+dri_destroy_buffer(__DRIdrawablePrivate *dPriv)
+{
+   struct dri_drawable *drawable = dri_drawable(dPriv);
+
+   /* No particular need to wait on fences before dereferencing them:
+    */
+   winsys->fence_reference( winsys, &ctx->last_swap_fence, NULL );
+   winsys->fence_reference( winsys, &ctx->first_swap_fence, NULL );
+
+   st_unreference_framebuffer(drawable->stfb);
+
+   FREE(drawable);
+}
+
diff --git a/src/gallium/state_trackers/glx/dri/dri_drawable.h b/src/gallium/state_trackers/glx/dri/dri_drawable.h
new file mode 100644
index 0000000000..1001bb8c57
--- /dev/null
+++ b/src/gallium/state_trackers/glx/dri/dri_drawable.h
@@ -0,0 +1,73 @@
+/**************************************************************************
+ *
+ * Copyright 2009, VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef DRI_DRAWABLE_H
+#define DRI_DRAWABLE_H
+
+#include "pipe/p_compiler.h"
+
+struct pipe_surface;
+struct pipe_fence;
+struct st_framebuffer;
+
+
+struct dri_drawable
+{
+   __DRIdrawablePrivate *dPriv;
+   unsigned stamp;
+
+   struct pipe_fence *last_swap_fence;
+   struct pipe_fence *first_swap_fence;
+
+   struct st_framebuffer *stfb;
+};
+
+
+static INLINE struct dri_drawable *
+dri_drawable(__DRIdrawablePrivate * driDrawPriv)
+{
+   return (struct dri_drawable *) driDrawPriv->driverPrivate;
+}
+
+
+/***********************************************************************
+ * dri_drawable.c
+ */
+
+void 
+dri_swap_buffers(__DRIdrawablePrivate * dPriv);
+
+void 
+dri_copy_sub_buffer(__DRIdrawablePrivate * dPriv,
+                    int x, int y, 
+                    int w, int h);
+
+void 
+dri_update_window_size(__DRIdrawablePrivate *dPriv);
+
+
+#endif
diff --git a/src/gallium/state_trackers/glx/dri/dri_extensions.c b/src/gallium/state_trackers/glx/dri/dri_extensions.c
new file mode 100644
index 0000000000..126faf7601
--- /dev/null
+++ b/src/gallium/state_trackers/glx/dri/dri_extensions.c
@@ -0,0 +1,108 @@
+/**************************************************************************
+ *
+ * Copyright 2009, VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+
+
+#define need_GL_ARB_multisample
+#define need_GL_ARB_point_parameters
+#define need_GL_ARB_texture_compression
+#define need_GL_ARB_vertex_buffer_object
+#define need_GL_ARB_vertex_program
+#define need_GL_ARB_window_pos
+#define need_GL_EXT_blend_color
+#define need_GL_EXT_blend_equation_separate
+#define need_GL_EXT_blend_func_separate
+#define need_GL_EXT_blend_minmax
+#define need_GL_EXT_cull_vertex
+#define need_GL_EXT_fog_coord
+#define need_GL_EXT_framebuffer_object
+#define need_GL_EXT_multi_draw_arrays
+#define need_GL_EXT_secondary_color
+#define need_GL_NV_vertex_program
+#include "extension_helper.h"
+
+
+/**
+ * Extension strings exported by the driver.
+ */
+const struct dri_extension card_extensions[] = {
+   {"GL_ARB_multisample", GL_ARB_multisample_functions},
+   {"GL_ARB_multitexture", NULL},
+   {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions},
+   {"GL_ARB_texture_border_clamp", NULL},
+   {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions},
+   {"GL_ARB_texture_cube_map", NULL},
+   {"GL_ARB_texture_env_add", NULL},
+   {"GL_ARB_texture_env_combine", NULL},
+   {"GL_ARB_texture_env_dot3", NULL},
+   {"GL_ARB_texture_mirrored_repeat", NULL},
+   {"GL_ARB_texture_rectangle", NULL},
+   {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions},
+   {"GL_ARB_pixel_buffer_object", NULL},
+   {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
+   {"GL_ARB_window_pos", GL_ARB_window_pos_functions},
+   {"GL_EXT_blend_color", GL_EXT_blend_color_functions},
+   {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions},
+   {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
+   {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
+   {"GL_EXT_blend_subtract", NULL},
+   {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions},
+   {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions},
+   {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions},
+   {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions},
+   {"GL_EXT_packed_depth_stencil", NULL},
+   {"GL_EXT_pixel_buffer_object", NULL},
+   {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions},
+   {"GL_EXT_stencil_wrap", NULL},
+   {"GL_EXT_texture_edge_clamp", NULL},
+   {"GL_EXT_texture_env_combine", NULL},
+   {"GL_EXT_texture_env_dot3", NULL},
+   {"GL_EXT_texture_filter_anisotropic", NULL},
+   {"GL_EXT_texture_lod_bias", NULL},
+   {"GL_3DFX_texture_compression_FXT1", NULL},
+   {"GL_APPLE_client_storage", NULL},
+   {"GL_MESA_pack_invert", NULL},
+   {"GL_MESA_ycbcr_texture", NULL},
+   {"GL_NV_blend_square", NULL},
+   {"GL_NV_vertex_program", GL_NV_vertex_program_functions},
+   {"GL_NV_vertex_program1_1", NULL},
+   {"GL_SGIS_generate_mipmap", NULL },
+   {NULL, NULL}
+};
+
+
+
+void 
+dri_init_extensions( void )
+{
+   /* The card_extensions list should be pruned according to the
+    * capabilities of the pipe_screen.  This is actually something
+    * that can/should be done inside st_create_context().
+    */
+   driInitExtensions( ctx->st->ctx, card_extensions, GL_TRUE );
+}
diff --git a/src/gallium/state_trackers/glx/dri/dri_lock.c b/src/gallium/state_trackers/glx/dri/dri_lock.c
new file mode 100644
index 0000000000..b272ab55f3
--- /dev/null
+++ b/src/gallium/state_trackers/glx/dri/dri_lock.c
@@ -0,0 +1,90 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "pipe/p_thread.h"
+#include "dri_context.h"
+#include "xf86drm.h"
+
+pipe_static_mutex( lockMutex );
+
+static void
+dri_contended_lock(struct dri_context *ctx)
+{
+   __DRIdrawablePrivate *dPriv = ctx->dPriv;
+   __DRIcontextPrivate *cPriv = ctx->cPriv;
+   __DRIscreenPrivate *sPriv = cPriv->driScreenPriv;
+
+   drmGetLock(sPriv->fd, cPriv->hHWContext, 0);
+
+   /* Perform round trip communication with server (including dropping
+    * and retaking the above lock) to update window dimensions:
+    */
+   if (dPriv)
+      DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv);
+}
+
+
+/* Lock the hardware and validate our state.
+ */
+void dri_lock_hardware( struct dri_context *ctx )
+{
+   __DRIcontextPrivate *cPriv = ctx->cPriv;
+   __DRIscreenPrivate *sPriv = cPriv->driScreenPriv;
+   char __ret = 0;
+
+   pipe_mutex_lock(lockMutex);
+   assert(!ctx->locked);
+
+   DRM_CAS((drmLock *) &sPriv->pSAREA->lock,
+           cPriv->hHWContext,
+           (DRM_LOCK_HELD | cPriv->hHWContext),
+           __ret);
+
+   if (__ret)
+      dri_contended_lock( ctx );
+
+   ctx->locked = TRUE;
+}
+
+
+/* Unlock the hardware using the global current context
+ */
+void dri_unlock_hardware( struct dri_context *ctx )
+{
+   __DRIcontextPrivate *cPriv = ctx->cPriv;
+   __DRIscreenPrivate *sPriv = cPriv->driScreenPriv;
+
+   assert(ctx->locked);
+   ctx->locked = FALSE;
+
+   DRM_UNLOCK(sPriv->fd, 
+              (drmLock *) &sPriv->pSAREA->lock,
+              cPriv->hHWContext);
+
+   pipe_mutex_unlock(lockMutex);
+}
diff --git a/src/gallium/state_trackers/glx/dri/dri_screen.c b/src/gallium/state_trackers/glx/dri/dri_screen.c
new file mode 100644
index 0000000000..f7119b949a
--- /dev/null
+++ b/src/gallium/state_trackers/glx/dri/dri_screen.c
@@ -0,0 +1,255 @@
+/**************************************************************************
+ *
+ * Copyright 2009, VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "utils.h"
+#include "vblank.h"
+#include "xmlpool.h"
+
+#include "dri_context.h"
+#include "dri_screen.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_cb_fbo.h"
+
+
+PUBLIC const char __driConfigOptions[] =
+   DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE
+    DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS)
+    DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0)
+   DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY
+//   DRI_CONF_FORCE_S3TC_ENABLE(false)
+    DRI_CONF_ALLOW_LARGE_TEXTURES(1)
+   DRI_CONF_SECTION_END DRI_CONF_END;
+
+const uint __driNConfigOptions = 3;
+
+static PFNGLXCREATECONTEXTMODES create_context_modes = NULL;
+
+extern const struct dri_extension card_extensions[];
+
+
+
+static const __DRIextension *driScreenExtensions[] = {
+    &driReadDrawableExtension,
+    &driCopySubBufferExtension.base,
+    &driSwapControlExtension.base,
+    &driFrameTrackingExtension.base,
+    &driMediaStreamCounterExtension.base,
+    NULL
+};
+
+
+
+
+static const char *
+dri_get_name( struct pipe_winsys *winsys )
+{
+   return "dri";
+}
+
+
+
+static void
+dri_destroy_screen(__DRIscreenPrivate * sPriv)
+{
+   struct dri_screen *screen = dri_screen(sPriv);
+
+   screen->pipe_screen->destroy( screen->pipe_screen );
+   screen->pipe_winsys->destroy( screen->pipe_winsys );
+   FREE(screen);
+   sPriv->private = NULL;
+}
+
+
+/**
+ * Get information about previous buffer swaps.
+ */
+static int
+dri_get_swap_info(__DRIdrawablePrivate * dPriv, 
+                  __DRIswapInfo * sInfo)
+{
+   if (dPriv == NULL || 
+       dPriv->driverPrivate == NULL ||
+       sInfo == NULL) 
+      return -1;
+   else
+      return 0;
+}
+
+static const __DRIconfig **
+dri_fill_in_modes(__DRIscreenPrivate *psp,
+                  unsigned pixel_bits )
+{
+   __DRIconfig **configs;
+   __GLcontextModes *m;
+   unsigned num_modes;
+   uint8_t depth_bits_array[3];
+   uint8_t stencil_bits_array[3];
+   uint8_t msaa_samples_array[1];
+   unsigned depth_buffer_factor;
+   unsigned back_buffer_factor;
+   GLenum fb_format;
+   GLenum fb_type;
+   int i;
+
+   static const GLenum back_buffer_modes[] = {
+      GLX_NONE, GLX_SWAP_UNDEFINED_OML
+   };
+
+   depth_bits_array[0] = 0;
+   depth_bits_array[1] = depth_bits;
+   depth_bits_array[2] = depth_bits;
+
+   stencil_bits_array[0] = 0;   /* no depth or stencil */
+   stencil_bits_array[1] = 0;   /* z24x8 */
+   stencil_bits_array[2] = 8;   /* z24s8 */
+
+   msaa_samples_array[0] = 0;
+
+   depth_buffer_factor = 3;
+   back_buffer_factor = 1;
+
+   num_modes = depth_buffer_factor * back_buffer_factor * 4;
+
+   if (pixel_bits == 16) {
+      fb_format = GL_RGB;
+      fb_type = GL_UNSIGNED_SHORT_5_6_5;
+   }
+   else {
+      fb_format = GL_BGRA;
+      fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+   }
+
+   configs = driCreateConfigs(fb_format, fb_type,
+			      depth_bits_array, 
+                              stencil_bits_array, depth_buffer_factor, 
+                              back_buffer_modes, back_buffer_factor,
+                              msaa_samples_array, 1);
+   if (configs == NULL) {
+      debug_printf("%s: driCreateConfigs failed\n", __FUNCTION__);
+      return NULL;
+   }
+
+   return configs;
+}
+
+
+
+/* This is the driver specific part of the createNewScreen entry point.
+ * 
+ * Returns the __GLcontextModes supported by this driver.
+ */
+static const __DRIconfig **dri_init_screen(__DRIscreenPrivate *sPriv)
+{
+   static const __DRIversion ddx_expected = { 1, 6, 0 }; /* hw query */
+   static const __DRIversion dri_expected = { 4, 0, 0 };
+   static const __DRIversion drm_expected = { 1, 5, 0 }; /* hw query */
+   struct dri_screen *screen;
+
+   if (!driCheckDriDdxDrmVersions2("dri",
+                                   &sPriv->dri_version, &dri_expected,
+                                   &sPriv->ddx_version, &ddx_expected,
+                                   &sPriv->drm_version, &drm_expected)) {
+      return NULL;
+   }
+
+   /* Set up dispatch table to cope with all known extensions:
+    */
+   driInitExtensions( NULL, card_extensions, GL_FALSE );
+
+
+   screen = CALLOC_STRUCT(dri_screen);
+   if (!screen)
+      goto fail;
+
+   screen->sPriv = sPriv;
+   sPriv->private = (void *) screen;
+
+
+   /* Search the registered winsys' for one that likes this sPriv.
+    * This is required in situations where multiple devices speak to
+    * the same DDX and are built into the same binary.  
+    *
+    * Note that cases like Intel i915 vs i965 doesn't fall into this
+    * category because they are built into separate binaries.
+    *
+    * Nonetheless, it's healthy to keep that level of detail out of
+    * this state_tracker.
+    */
+   for (i = 0; 
+        i < dri1_winsys_count && 
+           screen->st_winsys == NULL; 
+        i++) 
+   {
+      screen->dri_winsys = 
+         dri_winsys[i]->check_dri_privates( sPriv->pDevPriv,
+                                            sPriv->pSAREA
+                                            /* versions, etc?? */));
+   }
+                                             
+
+   driParseOptionInfo(&screen->optionCache,
+                      __driConfigOptions, 
+                      __driNConfigOptions);
+
+
+   /* Plug our info back into the __DRIscreenPrivate:
+    */
+   sPriv->private = (void *) screen;
+   sPriv->extensions = driScreenExtensions;
+
+   return dri_fill_in_modes(sPriv, 
+                            dri_priv->cpp * 8,
+                            24,
+                            8,
+                            1);
+fail:
+   return NULL;
+}
+
+
+
+const struct __DriverAPIRec driDriverAPI = {
+   .InitScreen		 = dri_init_screen,
+   .DestroyScreen	 = dri_destroy_screen,
+   .CreateContext	 = dri_create_context,
+   .DestroyContext	 = dri_destroy_context,
+   .CreateBuffer	 = dri_create_buffer,
+   .DestroyBuffer	 = dri_destroy_buffer,
+   .SwapBuffers		 = dri_swap_buffers,
+   .MakeCurrent		 = dri_make_current,
+   .UnbindContext	 = dri_unbind_context,
+   .GetSwapInfo		 = dri_get_swap_info,
+   .GetDrawableMSC	 = driDrawableGetMSC32,
+   .WaitForMSC		 = driWaitForMSC32,
+   .CopySubBuffer	 = dri_copy_sub_buffer,
+
+   //.InitScreen2		 = dri_init_screen2,
+};
diff --git a/src/gallium/state_trackers/glx/dri/dri_screen.h b/src/gallium/state_trackers/glx/dri/dri_screen.h
new file mode 100644
index 0000000000..12ed86d22a
--- /dev/null
+++ b/src/gallium/state_trackers/glx/dri/dri_screen.h
@@ -0,0 +1,83 @@
+/**************************************************************************
+ *
+ * Copyright 2009, VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef DRI_SCREEN_H
+#define DRI_SCREEN_H
+
+#include "dri_util.h"
+#include "xmlconfig.h"
+
+#include "pipe/p_compiler.h"
+
+struct dri_screen
+{
+   __DRIScreenPrivate *sPriv;
+   struct pipe_winsys *pipe_winsys;
+   struct pipe_screen *pipe_screen;
+
+   struct {
+      /* Need a pipe_surface pointer to do client-side swapbuffers:
+       */
+      unsigned long buffer_handle;
+      struct pipe_surface *surface;
+      struct pipe_texture *texture;
+
+      int pitch;                   /* row stride, in bytes */
+      int width;
+      int height;
+      int size;
+      int cpp;                     /* for front and back buffers */
+   } front;
+
+   int deviceID;
+   int drmMinor;
+
+
+   /**
+    * Configuration cache with default values for all contexts
+    */
+   driOptionCache optionCache;
+
+   /**
+    * Temporary(?) context to use for SwapBuffers or other situations in
+    * which we need a rendering context, but none is currently bound.
+    */
+   struct dri_context *dummyContext;
+};
+
+
+
+/** cast wrapper */
+static INLINE struct dri_screen *
+dri_screen(__DRIscreenPrivate *sPriv)
+{
+   return (struct dri_screen *) sPriv->private;
+}
+
+
+
+#endif
diff --git a/src/gallium/state_trackers/glx/xlib/Makefile b/src/gallium/state_trackers/glx/xlib/Makefile
new file mode 100644
index 0000000000..1b63db1f0e
--- /dev/null
+++ b/src/gallium/state_trackers/glx/xlib/Makefile
@@ -0,0 +1,25 @@
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = xlib
+
+
+DRIVER_INCLUDES = \
+	-I$(TOP)/include \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/mesa/main \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/gallium/auxiliary
+
+C_SOURCES = \
+	glxapi.c	\
+	fakeglx.c	\
+	fakeglx_fonts.c	\
+	xm_api.c
+
+
+include ../../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/state_trackers/glx/xlib/SConscript b/src/gallium/state_trackers/glx/xlib/SConscript
new file mode 100644
index 0000000000..14cdad69cb
--- /dev/null
+++ b/src/gallium/state_trackers/glx/xlib/SConscript
@@ -0,0 +1,26 @@
+#######################################################################
+# SConscript for xlib state_tracker
+
+Import('*')
+
+if env['platform'] == 'linux' \
+        and 'mesa' in env['statetrackers'] \
+        and ('softpipe' or 'i915simple' or 'trace') in env['drivers'] \
+        and not env['dri']:
+
+    env = env.Clone()
+
+    env.Append(CPPPATH = [
+        '#/src/mesa',
+        '#/src/mesa/main',
+    ])
+
+    st_xlib = env.ConvenienceLibrary(
+	target = 'st_xlib',
+	source = [ 'glxapi.c',
+		'fakeglx.c',
+		'fakeglx_fonts.c',
+		'xm_api.c',
+		]
+    )
+    Export('st_xlib')
diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx.c b/src/gallium/state_trackers/glx/xlib/fakeglx.c
new file mode 100644
index 0000000000..65e7048188
--- /dev/null
+++ b/src/gallium/state_trackers/glx/xlib/fakeglx.c
@@ -0,0 +1,2743 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ *
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ * This is an emulation of the GLX API which allows Mesa/GLX-based programs
+ * to run on X servers which do not have the real GLX extension.
+ *
+ * Thanks to the contributors:
+ *
+ * Initial version:  Philip Brown (phil@bolthole.com)
+ * Better glXGetConfig() support: Armin Liebchen (liebchen@asylum.cs.utah.edu)
+ * Further visual-handling refinements: Wolfram Gloger
+ *    (wmglo@Dent.MED.Uni-Muenchen.DE).
+ *
+ * Notes:
+ *   Don't be fooled, stereo isn't supported yet.
+ */
+
+
+
+#include "glxapi.h"
+#include "xm_api.h"
+#include "context.h"
+#include "config.h"
+#include "macros.h"
+#include "imports.h"
+#include "version.h"
+#include "fakeglx.h"
+#include "state_tracker/st_context.h"
+#include "state_tracker/st_public.h"
+
+
+/* This indicates the client-side GLX API and GLX encoder version. */
+#define CLIENT_MAJOR_VERSION 1
+#define CLIENT_MINOR_VERSION 4  /* but don't have 1.3's pbuffers, etc yet */
+
+/* This indicates the server-side GLX decoder version.
+ * GLX 1.4 indicates OpenGL 1.3 support
+ */
+#define SERVER_MAJOR_VERSION 1
+#define SERVER_MINOR_VERSION 4
+
+/* This is appended onto the glXGetClient/ServerString version strings. */
+#define MESA_GLX_VERSION "Mesa " MESA_VERSION_STRING
+
+/* Who implemented this GLX? */
+#define VENDOR "Brian Paul"
+
+#define EXTENSIONS \
+   "GLX_MESA_copy_sub_buffer " \
+   "GLX_MESA_pixmap_colormap " \
+   "GLX_MESA_release_buffers " \
+   "GLX_ARB_get_proc_address " \
+   "GLX_EXT_texture_from_pixmap " \
+   "GLX_EXT_visual_info " \
+   "GLX_EXT_visual_rating " \
+   /*"GLX_SGI_video_sync "*/ \
+   "GLX_SGIX_fbconfig " \
+   "GLX_SGIX_pbuffer "
+
+/*
+ * Our fake GLX context will contain a "real" GLX context and an XMesa context.
+ *
+ * Note that a pointer to a __GLXcontext is a pointer to a fake_glx_context,
+ * and vice versa.
+ *
+ * We really just need this structure in order to make the libGL functions
+ * glXGetCurrentContext(), glXGetCurrentDrawable() and glXGetCurrentDisplay()
+ * work correctly.
+ */
+struct fake_glx_context {
+   __GLXcontext glxContext;   /* this MUST be first! */
+   XMesaContext xmesaContext;
+};
+
+
+
+/**********************************************************************/
+/***                       GLX Visual Code                          ***/
+/**********************************************************************/
+
+#define DONT_CARE -1
+
+
+static XMesaVisual *VisualTable = NULL;
+static int NumVisuals = 0;
+
+
+
+/* Macro to handle c_class vs class field name in XVisualInfo struct */
+#if defined(__cplusplus) || defined(c_plusplus)
+#define CLASS c_class
+#else
+#define CLASS class
+#endif
+
+
+
+/*
+ * Test if the given XVisualInfo is usable for Mesa rendering.
+ */
+static GLboolean
+is_usable_visual( XVisualInfo *vinfo )
+{
+   switch (vinfo->CLASS) {
+      case StaticGray:
+      case GrayScale:
+         /* Any StaticGray/GrayScale visual works in RGB or CI mode */
+         return GL_TRUE;
+      case StaticColor:
+      case PseudoColor:
+	 /* Any StaticColor/PseudoColor visual of at least 4 bits */
+	 if (vinfo->depth>=4) {
+	    return GL_TRUE;
+	 }
+	 else {
+	    return GL_FALSE;
+	 }
+      case TrueColor:
+      case DirectColor:
+	 /* Any depth of TrueColor or DirectColor works in RGB mode */
+	 return GL_TRUE;
+      default:
+	 /* This should never happen */
+	 return GL_FALSE;
+   }
+}
+
+
+/*
+ * Given an XVisualInfo and RGB, Double, and Depth buffer flags, save the
+ * configuration in our list of GLX visuals.
+ */
+static XMesaVisual
+save_glx_visual( Display *dpy, XVisualInfo *vinfo,
+                 GLboolean rgbFlag, GLboolean alphaFlag, GLboolean dbFlag,
+                 GLboolean stereoFlag,
+                 GLint depth_size, GLint stencil_size,
+                 GLint accumRedSize, GLint accumGreenSize,
+                 GLint accumBlueSize, GLint accumAlphaSize,
+                 GLint level, GLint numAuxBuffers )
+{
+   GLboolean ximageFlag = GL_TRUE;
+   XMesaVisual xmvis;
+   GLint i;
+   GLboolean comparePointers;
+
+   if (dbFlag) {
+      /* Check if the MESA_BACK_BUFFER env var is set */
+      char *backbuffer = _mesa_getenv("MESA_BACK_BUFFER");
+      if (backbuffer) {
+         if (backbuffer[0]=='p' || backbuffer[0]=='P') {
+            ximageFlag = GL_FALSE;
+         }
+         else if (backbuffer[0]=='x' || backbuffer[0]=='X') {
+            ximageFlag = GL_TRUE;
+         }
+         else {
+            _mesa_warning(NULL, "Mesa: invalid value for MESA_BACK_BUFFER environment variable, using an XImage.");
+         }
+      }
+   }
+
+   if (stereoFlag) {
+      /* stereo not supported */
+      return NULL;
+   }
+
+   if (stencil_size > 0 && depth_size > 0)
+      depth_size = 24;
+
+   /* Comparing IDs uses less memory but sometimes fails. */
+   /* XXX revisit this after 3.0 is finished. */
+   if (_mesa_getenv("MESA_GLX_VISUAL_HACK"))
+      comparePointers = GL_TRUE;
+   else
+      comparePointers = GL_FALSE;
+
+   /* Force the visual to have an alpha channel */
+   if (rgbFlag && _mesa_getenv("MESA_GLX_FORCE_ALPHA"))
+      alphaFlag = GL_TRUE;
+
+   /* First check if a matching visual is already in the list */
+   for (i=0; i<NumVisuals; i++) {
+      XMesaVisual v = VisualTable[i];
+      if (v->display == dpy
+          && v->mesa_visual.level == level
+          && v->mesa_visual.numAuxBuffers == numAuxBuffers
+          && v->ximage_flag == ximageFlag
+          && v->mesa_visual.rgbMode == rgbFlag
+          && v->mesa_visual.doubleBufferMode == dbFlag
+          && v->mesa_visual.stereoMode == stereoFlag
+          && (v->mesa_visual.alphaBits > 0) == alphaFlag
+          && (v->mesa_visual.depthBits >= depth_size || depth_size == 0)
+          && (v->mesa_visual.stencilBits >= stencil_size || stencil_size == 0)
+          && (v->mesa_visual.accumRedBits >= accumRedSize || accumRedSize == 0)
+          && (v->mesa_visual.accumGreenBits >= accumGreenSize || accumGreenSize == 0)
+          && (v->mesa_visual.accumBlueBits >= accumBlueSize || accumBlueSize == 0)
+          && (v->mesa_visual.accumAlphaBits >= accumAlphaSize || accumAlphaSize == 0)) {
+         /* now either compare XVisualInfo pointers or visual IDs */
+         if ((!comparePointers && v->visinfo->visualid == vinfo->visualid)
+             || (comparePointers && v->vishandle == vinfo)) {
+            return v;
+         }
+      }
+   }
+
+   /* Create a new visual and add it to the list. */
+
+   xmvis = XMesaCreateVisual( dpy, vinfo, rgbFlag, alphaFlag, dbFlag,
+                              stereoFlag, ximageFlag,
+                              depth_size, stencil_size,
+                              accumRedSize, accumBlueSize,
+                              accumBlueSize, accumAlphaSize, 0, level,
+                              GLX_NONE_EXT );
+   if (xmvis) {
+      /* Save a copy of the pointer now so we can find this visual again
+       * if we need to search for it in find_glx_visual().
+       */
+      xmvis->vishandle = vinfo;
+      /* Allocate more space for additional visual */
+      VisualTable = (XMesaVisual *) _mesa_realloc( VisualTable, 
+                                   sizeof(XMesaVisual) * NumVisuals, 
+                                   sizeof(XMesaVisual) * (NumVisuals + 1));
+      /* add xmvis to the list */
+      VisualTable[NumVisuals] = xmvis;
+      NumVisuals++;
+      /* XXX minor hack, because XMesaCreateVisual doesn't support an
+       * aux buffers parameter.
+       */
+      xmvis->mesa_visual.numAuxBuffers = numAuxBuffers;
+   }
+   return xmvis;
+}
+
+
+/**
+ * Return the default number of bits for the Z buffer.
+ * If defined, use the MESA_GLX_DEPTH_BITS env var value.
+ * Otherwise, use the DEFAULT_SOFTWARE_DEPTH_BITS constant.
+ * XXX probably do the same thing for stencil, accum, etc.
+ */
+static GLint
+default_depth_bits(void)
+{
+   int zBits;
+   const char *zEnv = _mesa_getenv("MESA_GLX_DEPTH_BITS");
+   if (zEnv)
+      zBits = _mesa_atoi(zEnv);
+   else
+      zBits = DEFAULT_SOFTWARE_DEPTH_BITS;
+   return zBits;
+}
+
+static GLint
+default_alpha_bits(void)
+{
+   int aBits;
+   const char *aEnv = _mesa_getenv("MESA_GLX_ALPHA_BITS");
+   if (aEnv)
+      aBits = _mesa_atoi(aEnv);
+   else
+      aBits = 0;
+   return aBits;
+}
+
+static GLint
+default_accum_bits(void)
+{
+   return 16;
+}
+
+
+
+/*
+ * Create a GLX visual from a regular XVisualInfo.
+ * This is called when Fake GLX is given an XVisualInfo which wasn't
+ * returned by glXChooseVisual.  Since this is the first time we're
+ * considering this visual we'll take a guess at reasonable values
+ * for depth buffer size, stencil size, accum size, etc.
+ * This is the best we can do with a client-side emulation of GLX.
+ */
+static XMesaVisual
+create_glx_visual( Display *dpy, XVisualInfo *visinfo )
+{
+   GLint zBits = default_depth_bits();
+   GLint accBits = default_accum_bits();
+   GLboolean alphaFlag = default_alpha_bits() > 0;
+
+   if (is_usable_visual( visinfo )) {
+      /* Configure this visual as RGB, double-buffered, depth-buffered. */
+      /* This is surely wrong for some people's needs but what else */
+      /* can be done?  They should use glXChooseVisual(). */
+      return save_glx_visual( dpy, visinfo,
+                              GL_TRUE,   /* rgb */
+                              alphaFlag, /* alpha */
+                              GL_TRUE,   /* double */
+                              GL_FALSE,  /* stereo */
+                              zBits,
+                              STENCIL_BITS,
+                              accBits, /* r */
+                              accBits, /* g */
+                              accBits, /* b */
+                              accBits, /* a */
+                              0,         /* level */
+                              0          /* numAux */
+         );
+   }
+   else {
+      _mesa_warning(NULL, "Mesa: error in glXCreateContext: bad visual\n");
+      return NULL;
+   }
+}
+
+
+
+/*
+ * Find the GLX visual associated with an XVisualInfo.
+ */
+static XMesaVisual
+find_glx_visual( Display *dpy, XVisualInfo *vinfo )
+{
+   int i;
+
+   /* try to match visual id */
+   for (i=0;i<NumVisuals;i++) {
+      if (VisualTable[i]->display==dpy
+          && VisualTable[i]->visinfo->visualid == vinfo->visualid) {
+         return VisualTable[i];
+      }
+   }
+
+   /* if that fails, try to match pointers */
+   for (i=0;i<NumVisuals;i++) {
+      if (VisualTable[i]->display==dpy && VisualTable[i]->vishandle==vinfo) {
+         return VisualTable[i];
+      }
+   }
+
+   return NULL;
+}
+
+
+
+
+
+
+/**
+ * Try to get an X visual which matches the given arguments.
+ */
+static XVisualInfo *
+get_visual( Display *dpy, int scr, unsigned int depth, int xclass )
+{
+   XVisualInfo temp, *vis;
+   long mask;
+   int n;
+   unsigned int default_depth;
+   int default_class;
+
+   mask = VisualScreenMask | VisualDepthMask | VisualClassMask;
+   temp.screen = scr;
+   temp.depth = depth;
+   temp.CLASS = xclass;
+
+   default_depth = DefaultDepth(dpy,scr);
+   default_class = DefaultVisual(dpy,scr)->CLASS;
+
+   if (depth==default_depth && xclass==default_class) {
+      /* try to get root window's visual */
+      temp.visualid = DefaultVisual(dpy,scr)->visualid;
+      mask |= VisualIDMask;
+   }
+
+   vis = XGetVisualInfo( dpy, mask, &temp, &n );
+
+   /* In case bits/pixel > 24, make sure color channels are still <=8 bits.
+    * An SGI Infinite Reality system, for example, can have 30bpp pixels:
+    * 10 bits per color channel.  Mesa's limited to a max of 8 bits/channel.
+    */
+   if (vis && depth > 24 && (xclass==TrueColor || xclass==DirectColor)) {
+      if (_mesa_bitcount((GLuint) vis->red_mask  ) <= 8 &&
+          _mesa_bitcount((GLuint) vis->green_mask) <= 8 &&
+          _mesa_bitcount((GLuint) vis->blue_mask ) <= 8) {
+         return vis;
+      }
+      else {
+         XFree((void *) vis);
+         return NULL;
+      }
+   }
+   
+   return vis;
+}
+
+
+
+/*
+ * Retrieve the value of the given environment variable and find
+ * the X visual which matches it.
+ * Input:  dpy - the display
+ *         screen - the screen number
+ *         varname - the name of the environment variable
+ * Return:  an XVisualInfo pointer to NULL if error.
+ */
+static XVisualInfo *
+get_env_visual(Display *dpy, int scr, const char *varname)
+{
+   char value[100], type[100];
+   int depth, xclass = -1;
+   XVisualInfo *vis;
+
+   if (!_mesa_getenv( varname )) {
+      return NULL;
+   }
+
+   _mesa_strncpy( value, _mesa_getenv(varname), 100 );
+   value[99] = 0;
+
+   sscanf( value, "%s %d", type, &depth );
+
+   if (_mesa_strcmp(type,"TrueColor")==0)          xclass = TrueColor;
+   else if (_mesa_strcmp(type,"DirectColor")==0)   xclass = DirectColor;
+   else if (_mesa_strcmp(type,"PseudoColor")==0)   xclass = PseudoColor;
+   else if (_mesa_strcmp(type,"StaticColor")==0)   xclass = StaticColor;
+   else if (_mesa_strcmp(type,"GrayScale")==0)     xclass = GrayScale;
+   else if (_mesa_strcmp(type,"StaticGray")==0)    xclass = StaticGray;
+
+   if (xclass>-1 && depth>0) {
+      vis = get_visual( dpy, scr, depth, xclass );
+      if (vis) {
+	 return vis;
+      }
+   }
+
+   _mesa_warning(NULL, "GLX unable to find visual class=%s, depth=%d.",
+                 type, depth);
+
+   return NULL;
+}
+
+
+
+/*
+ * Select an X visual which satisfies the RGBA flag and minimum depth.
+ * Input:  dpy, 
+ *         screen - X display and screen number
+ *         min_depth - minimum visual depth
+ *         preferred_class - preferred GLX visual class or DONT_CARE
+ * Return:  pointer to an XVisualInfo or NULL.
+ */
+static XVisualInfo *
+choose_x_visual( Display *dpy, int screen, int min_depth,
+                 int preferred_class )
+{
+   XVisualInfo *vis;
+   int xclass, visclass = 0;
+   int depth;
+
+   /* First see if the MESA_RGB_VISUAL env var is defined */
+   vis = get_env_visual( dpy, screen, "MESA_RGB_VISUAL" );
+   if (vis) {
+      return vis;
+   }
+   /* Otherwise, search for a suitable visual */
+   if (preferred_class==DONT_CARE) {
+      for (xclass=0;xclass<6;xclass++) {
+         switch (xclass) {
+         case 0:  visclass = TrueColor;    break;
+         case 1:  visclass = DirectColor;  break;
+         case 2:  visclass = PseudoColor;  break;
+         case 3:  visclass = StaticColor;  break;
+         case 4:  visclass = GrayScale;    break;
+         case 5:  visclass = StaticGray;   break;
+         }
+         if (min_depth==0) {
+            /* start with shallowest */
+            for (depth=0;depth<=32;depth++) {
+               if (visclass==TrueColor && depth==8) {
+                  /* Special case:  try to get 8-bit PseudoColor before */
+                  /* 8-bit TrueColor */
+                  vis = get_visual( dpy, screen, 8, PseudoColor );
+                  if (vis) {
+                     return vis;
+                  }
+               }
+               vis = get_visual( dpy, screen, depth, visclass );
+               if (vis) {
+                  return vis;
+               }
+            }
+         }
+         else {
+            /* start with deepest */
+            for (depth=32;depth>=min_depth;depth--) {
+               if (visclass==TrueColor && depth==8) {
+                  /* Special case:  try to get 8-bit PseudoColor before */
+                  /* 8-bit TrueColor */
+                  vis = get_visual( dpy, screen, 8, PseudoColor );
+                  if (vis) {
+                     return vis;
+                  }
+               }
+               vis = get_visual( dpy, screen, depth, visclass );
+               if (vis) {
+                  return vis;
+               }
+            }
+         }
+      }
+   }
+   else {
+      /* search for a specific visual class */
+      switch (preferred_class) {
+      case GLX_TRUE_COLOR_EXT:    visclass = TrueColor;    break;
+      case GLX_DIRECT_COLOR_EXT:  visclass = DirectColor;  break;
+      case GLX_PSEUDO_COLOR_EXT:  visclass = PseudoColor;  break;
+      case GLX_STATIC_COLOR_EXT:  visclass = StaticColor;  break;
+      case GLX_GRAY_SCALE_EXT:    visclass = GrayScale;    break;
+      case GLX_STATIC_GRAY_EXT:   visclass = StaticGray;   break;
+      default:   return NULL;
+      }
+      if (min_depth==0) {
+         /* start with shallowest */
+         for (depth=0;depth<=32;depth++) {
+            vis = get_visual( dpy, screen, depth, visclass );
+            if (vis) {
+               return vis;
+            }
+         }
+      }
+      else {
+         /* start with deepest */
+         for (depth=32;depth>=min_depth;depth--) {
+            vis = get_visual( dpy, screen, depth, visclass );
+            if (vis) {
+               return vis;
+            }
+         }
+      }
+   }
+
+   /* didn't find a visual */
+   return NULL;
+}
+
+
+
+
+/**********************************************************************/
+/***             Display-related functions                          ***/
+/**********************************************************************/
+
+
+/**
+ * Free all XMesaVisuals which are associated with the given display.
+ */
+static void
+destroy_visuals_on_display(Display *dpy)
+{
+   int i;
+   for (i = 0; i < NumVisuals; i++) {
+      if (VisualTable[i]->display == dpy) {
+         /* remove this visual */
+         int j;
+         free(VisualTable[i]);
+         for (j = i; j < NumVisuals - 1; j++)
+            VisualTable[j] = VisualTable[j + 1];
+         NumVisuals--;
+      }
+   }
+}
+
+
+/**
+ * Called from XCloseDisplay() to let us free our display-related data.
+ */
+static int
+close_display_callback(Display *dpy, XExtCodes *codes)
+{
+   destroy_visuals_on_display(dpy);
+   xmesa_destroy_buffers_on_display(dpy);
+   return 0;
+}
+
+
+/**
+ * Look for the named extension on given display and return a pointer
+ * to the _XExtension data, or NULL if extension not found.
+ */
+static _XExtension *
+lookup_extension(Display *dpy, const char *extName)
+{
+   _XExtension *ext;
+   for (ext = dpy->ext_procs; ext; ext = ext->next) {
+      if (ext->name && strcmp(ext->name, extName) == 0) {
+         return ext;
+      }
+   }
+   return NULL;
+}
+
+
+/**
+ * Whenever we're given a new Display pointer, call this function to
+ * register our close_display_callback function.
+ */
+static void
+register_with_display(Display *dpy)
+{
+   const char *extName = "MesaGLX";
+   _XExtension *ext;
+
+   ext = lookup_extension(dpy, extName);
+   if (!ext) {
+      XExtCodes *c = XAddExtension(dpy);
+      ext = dpy->ext_procs;  /* new extension is at head of list */
+      assert(c->extension == ext->codes.extension);
+      ext->name = _mesa_strdup(extName);
+      ext->close_display = close_display_callback;
+   }
+}
+
+
+/**********************************************************************/
+/***                  Begin Fake GLX API Functions                  ***/
+/**********************************************************************/
+
+
+/**
+ * Helper used by glXChooseVisual and glXChooseFBConfig.
+ * The fbConfig parameter must be GL_FALSE for the former and GL_TRUE for
+ * the later.
+ * In either case, the attribute list is terminated with the value 'None'.
+ */
+static XMesaVisual
+choose_visual( Display *dpy, int screen, const int *list, GLboolean fbConfig )
+{
+   const GLboolean rgbModeDefault = fbConfig;
+   const int *parselist;
+   XVisualInfo *vis;
+   int min_ci = 0;
+   int min_red=0, min_green=0, min_blue=0;
+   GLboolean rgb_flag = rgbModeDefault;
+   GLboolean alpha_flag = GL_FALSE;
+   GLboolean double_flag = GL_FALSE;
+   GLboolean stereo_flag = GL_FALSE;
+   GLint depth_size = 0;
+   GLint stencil_size = 0;
+   GLint accumRedSize = 0;
+   GLint accumGreenSize = 0;
+   GLint accumBlueSize = 0;
+   GLint accumAlphaSize = 0;
+   int level = 0;
+   int visual_type = DONT_CARE;
+   int trans_type = DONT_CARE;
+   int trans_value = DONT_CARE;
+   GLint caveat = DONT_CARE;
+   XMesaVisual xmvis = NULL;
+   int desiredVisualID = -1;
+   int numAux = 0;
+
+   parselist = list;
+
+   while (*parselist) {
+
+      switch (*parselist) {
+	 case GLX_USE_GL:
+            if (fbConfig) {
+               /* invalid token */
+               return NULL;
+            }
+            else {
+               /* skip */
+               parselist++;
+            }
+	    break;
+	 case GLX_BUFFER_SIZE:
+	    parselist++;
+	    min_ci = *parselist++;
+	    break;
+	 case GLX_LEVEL:
+	    parselist++;
+            level = *parselist++;
+	    break;
+	 case GLX_RGBA:
+            if (fbConfig) {
+               /* invalid token */
+               return NULL;
+            }
+            else {
+               rgb_flag = GL_TRUE;
+               parselist++;
+            }
+	    break;
+	 case GLX_DOUBLEBUFFER:
+            parselist++;
+            if (fbConfig) {
+               double_flag = *parselist++;
+            }
+            else {
+               double_flag = GL_TRUE;
+            }
+	    break;
+	 case GLX_STEREO:
+            parselist++;
+            if (fbConfig) {
+               stereo_flag = *parselist++;
+            }
+            else {
+               stereo_flag = GL_TRUE;
+            }
+            break;
+	 case GLX_AUX_BUFFERS:
+	    parselist++;
+            numAux = *parselist++;
+            if (numAux > MAX_AUX_BUFFERS)
+               return NULL;
+	    break;
+	 case GLX_RED_SIZE:
+	    parselist++;
+	    min_red = *parselist++;
+	    break;
+	 case GLX_GREEN_SIZE:
+	    parselist++;
+	    min_green = *parselist++;
+	    break;
+	 case GLX_BLUE_SIZE:
+	    parselist++;
+	    min_blue = *parselist++;
+	    break;
+	 case GLX_ALPHA_SIZE:
+	    parselist++;
+            {
+               GLint size = *parselist++;
+               alpha_flag = size ? GL_TRUE : GL_FALSE;
+            }
+	    break;
+	 case GLX_DEPTH_SIZE:
+	    parselist++;
+	    depth_size = *parselist++;
+	    break;
+	 case GLX_STENCIL_SIZE:
+	    parselist++;
+	    stencil_size = *parselist++;
+	    break;
+	 case GLX_ACCUM_RED_SIZE:
+	    parselist++;
+            {
+               GLint size = *parselist++;
+               accumRedSize = MAX2( accumRedSize, size );
+            }
+            break;
+	 case GLX_ACCUM_GREEN_SIZE:
+	    parselist++;
+            {
+               GLint size = *parselist++;
+               accumGreenSize = MAX2( accumGreenSize, size );
+            }
+            break;
+	 case GLX_ACCUM_BLUE_SIZE:
+	    parselist++;
+            {
+               GLint size = *parselist++;
+               accumBlueSize = MAX2( accumBlueSize, size );
+            }
+            break;
+	 case GLX_ACCUM_ALPHA_SIZE:
+	    parselist++;
+            {
+               GLint size = *parselist++;
+               accumAlphaSize = MAX2( accumAlphaSize, size );
+            }
+	    break;
+
+         /*
+          * GLX_EXT_visual_info extension
+          */
+         case GLX_X_VISUAL_TYPE_EXT:
+            parselist++;
+            visual_type = *parselist++;
+            break;
+         case GLX_TRANSPARENT_TYPE_EXT:
+            parselist++;
+            trans_type = *parselist++;
+            break;
+         case GLX_TRANSPARENT_INDEX_VALUE_EXT:
+            parselist++;
+            trans_value = *parselist++;
+            break;
+         case GLX_TRANSPARENT_RED_VALUE_EXT:
+         case GLX_TRANSPARENT_GREEN_VALUE_EXT:
+         case GLX_TRANSPARENT_BLUE_VALUE_EXT:
+         case GLX_TRANSPARENT_ALPHA_VALUE_EXT:
+	    /* ignore */
+	    parselist++;
+	    parselist++;
+	    break;
+
+         /*
+          * GLX_EXT_visual_info extension
+          */
+         case GLX_VISUAL_CAVEAT_EXT:
+            parselist++;
+            caveat = *parselist++; /* ignored for now */
+            break;
+
+         /*
+          * GLX_ARB_multisample
+          */
+         case GLX_SAMPLE_BUFFERS_ARB:
+            /* ms not supported */
+            return NULL;
+         case GLX_SAMPLES_ARB:
+            /* ms not supported */
+            return NULL;
+
+         /*
+          * FBConfig attribs.
+          */
+         case GLX_RENDER_TYPE:
+            if (!fbConfig)
+               return NULL;
+            parselist++;
+            if (*parselist == GLX_RGBA_BIT) {
+               rgb_flag = GL_TRUE;
+            }
+            else if (*parselist == GLX_COLOR_INDEX_BIT) {
+               rgb_flag = GL_FALSE;
+            }
+            else if (*parselist == 0) {
+               rgb_flag = GL_TRUE;
+            }
+            parselist++;
+            break;
+         case GLX_DRAWABLE_TYPE:
+            if (!fbConfig)
+               return NULL;
+            parselist++;
+            if (*parselist & ~(GLX_WINDOW_BIT | GLX_PIXMAP_BIT | GLX_PBUFFER_BIT)) {
+               return NULL; /* bad bit */
+            }
+            parselist++;
+            break;
+         case GLX_FBCONFIG_ID:
+            if (!fbConfig)
+               return NULL;
+            parselist++;
+            desiredVisualID = *parselist++;
+            break;
+         case GLX_X_RENDERABLE:
+            if (!fbConfig)
+               return NULL;
+            parselist += 2;
+            /* ignore */
+            break;
+
+#ifdef GLX_EXT_texture_from_pixmap
+         case GLX_BIND_TO_TEXTURE_RGB_EXT:
+            parselist++; /*skip*/
+            break;
+         case GLX_BIND_TO_TEXTURE_RGBA_EXT:
+            parselist++; /*skip*/
+            break;
+         case GLX_BIND_TO_MIPMAP_TEXTURE_EXT:
+            parselist++; /*skip*/
+            break;
+         case GLX_BIND_TO_TEXTURE_TARGETS_EXT:
+            parselist++;
+            if (*parselist & ~(GLX_TEXTURE_1D_BIT_EXT |
+                               GLX_TEXTURE_2D_BIT_EXT |
+                               GLX_TEXTURE_RECTANGLE_BIT_EXT)) {
+               /* invalid bit */
+               return NULL;
+            }
+            break;
+         case GLX_Y_INVERTED_EXT:
+            parselist++; /*skip*/
+            break;
+#endif
+
+	 case None:
+            /* end of list */
+	    break;
+
+	 default:
+	    /* undefined attribute */
+            _mesa_warning(NULL, "unexpected attrib 0x%x in choose_visual()",
+                          *parselist);
+	    return NULL;
+      }
+   }
+
+   (void) caveat;
+
+
+   /*
+    * Since we're only simulating the GLX extension this function will never
+    * find any real GL visuals.  Instead, all we can do is try to find an RGB
+    * or CI visual of appropriate depth.  Other requested attributes such as
+    * double buffering, depth buffer, etc. will be associated with the X
+    * visual and stored in the VisualTable[].
+    */
+   if (desiredVisualID != -1) {
+      /* try to get a specific visual, by visualID */
+      XVisualInfo temp;
+      int n;
+      temp.visualid = desiredVisualID;
+      temp.screen = screen;
+      vis = XGetVisualInfo(dpy, VisualIDMask | VisualScreenMask, &temp, &n);
+      if (vis) {
+         /* give the visual some useful GLX attributes */
+         double_flag = GL_TRUE;
+         rgb_flag = GL_TRUE;
+         depth_size = default_depth_bits();
+         stencil_size = STENCIL_BITS;
+         /* XXX accum??? */
+      }
+   }
+   else if (level==0) {
+      /* normal color planes */
+      /* Get an RGB visual */
+      int min_rgb = min_red + min_green + min_blue;
+      if (min_rgb>1 && min_rgb<8) {
+         /* a special case to be sure we can get a monochrome visual */
+         min_rgb = 1;
+      }
+      vis = choose_x_visual( dpy, screen, min_rgb, visual_type );
+   }
+   else {
+      _mesa_warning(NULL, "overlay not supported");
+      return NULL;
+   }
+
+   if (vis) {
+      /* Note: we're not exactly obeying the glXChooseVisual rules here.
+       * When GLX_DEPTH_SIZE = 1 is specified we're supposed to choose the
+       * largest depth buffer size, which is 32bits/value.  Instead, we
+       * return 16 to maintain performance with earlier versions of Mesa.
+       */
+      if (stencil_size > 0)
+         depth_size = 24;  /* if Z and stencil, always use 24+8 format */
+      else if (depth_size > 24)
+         depth_size = 32;
+      else if (depth_size > 16)
+         depth_size = 24;
+      else if (depth_size > 0) {
+         depth_size = default_depth_bits();
+      }
+
+      if (!alpha_flag) {
+         alpha_flag = default_alpha_bits() > 0;
+      }
+
+      /* we only support one size of stencil and accum buffers. */
+      if (stencil_size > 0)
+         stencil_size = STENCIL_BITS;
+
+      if (accumRedSize > 0 || 
+          accumGreenSize > 0 || 
+          accumBlueSize > 0 ||
+          accumAlphaSize > 0) {
+
+         accumRedSize = 
+            accumGreenSize = 
+            accumBlueSize = default_accum_bits();
+
+         accumAlphaSize = alpha_flag ? accumRedSize : 0;
+      }
+
+      xmvis = save_glx_visual( dpy, vis, rgb_flag, alpha_flag, double_flag,
+                               stereo_flag, depth_size, stencil_size,
+                               accumRedSize, accumGreenSize,
+                               accumBlueSize, accumAlphaSize, level, numAux );
+   }
+
+   return xmvis;
+}
+
+
+static XVisualInfo *
+Fake_glXChooseVisual( Display *dpy, int screen, int *list )
+{
+   XMesaVisual xmvis;
+
+   /* register ourselves as an extension on this display */
+   register_with_display(dpy);
+
+   xmvis = choose_visual(dpy, screen, list, GL_FALSE);
+   if (xmvis) {
+      /* create a new vishandle - the cached one may be stale */
+      xmvis->vishandle = (XVisualInfo *) _mesa_malloc(sizeof(XVisualInfo));
+      if (xmvis->vishandle) {
+         _mesa_memcpy(xmvis->vishandle, xmvis->visinfo, sizeof(XVisualInfo));
+      }
+      return xmvis->vishandle;
+   }
+   else
+      return NULL;
+}
+
+
+static GLXContext
+Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo,
+                       GLXContext share_list, Bool direct )
+{
+   XMesaVisual xmvis;
+   struct fake_glx_context *glxCtx;
+   struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list;
+
+   if (!dpy || !visinfo)
+      return 0;
+
+   glxCtx = CALLOC_STRUCT(fake_glx_context);
+   if (!glxCtx)
+      return 0;
+
+   /* deallocate unused windows/buffers */
+#if 0
+   XMesaGarbageCollect();
+#endif
+
+   xmvis = find_glx_visual( dpy, visinfo );
+   if (!xmvis) {
+      /* This visual wasn't found with glXChooseVisual() */
+      xmvis = create_glx_visual( dpy, visinfo );
+      if (!xmvis) {
+         /* unusable visual */
+         _mesa_free(glxCtx);
+         return NULL;
+      }
+   }
+
+   glxCtx->xmesaContext = XMesaCreateContext(xmvis,
+                                   shareCtx ? shareCtx->xmesaContext : NULL);
+   if (!glxCtx->xmesaContext) {
+      _mesa_free(glxCtx);
+      return NULL;
+   }
+
+   glxCtx->glxContext.isDirect = GL_FALSE;
+   glxCtx->glxContext.currentDpy = dpy;
+   glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
+
+   assert((void *) glxCtx == (void *) &(glxCtx->glxContext));
+
+   return (GLXContext) glxCtx;
+}
+
+
+/* XXX these may have to be removed due to thread-safety issues. */
+static GLXContext MakeCurrent_PrevContext = 0;
+static GLXDrawable MakeCurrent_PrevDrawable = 0;
+static GLXDrawable MakeCurrent_PrevReadable = 0;
+static XMesaBuffer MakeCurrent_PrevDrawBuffer = 0;
+static XMesaBuffer MakeCurrent_PrevReadBuffer = 0;
+
+
+/* GLX 1.3 and later */
+static Bool
+Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw,
+                            GLXDrawable read, GLXContext ctx )
+{
+   struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx;
+   static boolean firsttime = 1, no_rast = 0;
+
+   if (firsttime) {
+      no_rast = getenv("SP_NO_RAST") != NULL;
+      firsttime = 0;
+   }
+
+
+   if (ctx && draw && read) {
+      XMesaBuffer drawBuffer, readBuffer;
+      XMesaContext xmctx = glxCtx->xmesaContext;
+
+      /* Find the XMesaBuffer which corresponds to the GLXDrawable 'draw' */
+      if (ctx == MakeCurrent_PrevContext
+          && draw == MakeCurrent_PrevDrawable) {
+         drawBuffer = MakeCurrent_PrevDrawBuffer;
+      }
+      else {
+         drawBuffer = XMesaFindBuffer( dpy, draw );
+      }
+      if (!drawBuffer) {
+         /* drawable must be a new window! */
+         drawBuffer = XMesaCreateWindowBuffer( xmctx->xm_visual, draw );
+         if (!drawBuffer) {
+            /* Out of memory, or context/drawable depth mismatch */
+            return False;
+         }
+      }
+
+      /* Find the XMesaBuffer which corresponds to the GLXDrawable 'read' */
+      if (ctx == MakeCurrent_PrevContext
+          && read == MakeCurrent_PrevReadable) {
+         readBuffer = MakeCurrent_PrevReadBuffer;
+      }
+      else {
+         readBuffer = XMesaFindBuffer( dpy, read );
+      }
+      if (!readBuffer) {
+         /* drawable must be a new window! */
+         readBuffer = XMesaCreateWindowBuffer( xmctx->xm_visual, read );
+         if (!readBuffer) {
+            /* Out of memory, or context/drawable depth mismatch */
+            return False;
+         }
+      }
+
+      if (no_rast &&
+          MakeCurrent_PrevContext == ctx &&
+          MakeCurrent_PrevDrawable == draw &&
+          MakeCurrent_PrevReadable == read &&
+          MakeCurrent_PrevDrawBuffer == drawBuffer &&
+          MakeCurrent_PrevReadBuffer == readBuffer)
+         return True;
+          
+      MakeCurrent_PrevContext = ctx;
+      MakeCurrent_PrevDrawable = draw;
+      MakeCurrent_PrevReadable = read;
+      MakeCurrent_PrevDrawBuffer = drawBuffer;
+      MakeCurrent_PrevReadBuffer = readBuffer;
+
+      /* Now make current! */
+      if (XMesaMakeCurrent2(xmctx, drawBuffer, readBuffer)) {
+         ((__GLXcontext *) ctx)->currentDpy = dpy;
+         ((__GLXcontext *) ctx)->currentDrawable = draw;
+         ((__GLXcontext *) ctx)->currentReadable = read;
+         return True;
+      }
+      else {
+         return False;
+      }
+   }
+   else if (!ctx && !draw && !read) {
+      /* release current context w/out assigning new one. */
+      XMesaMakeCurrent2( NULL, NULL, NULL );
+      MakeCurrent_PrevContext = 0;
+      MakeCurrent_PrevDrawable = 0;
+      MakeCurrent_PrevReadable = 0;
+      MakeCurrent_PrevDrawBuffer = 0;
+      MakeCurrent_PrevReadBuffer = 0;
+      return True;
+   }
+   else {
+      /* The args must either all be non-zero or all zero.
+       * This is an error.
+       */
+      return False;
+   }
+}
+
+
+static Bool
+Fake_glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx )
+{
+   return Fake_glXMakeContextCurrent( dpy, drawable, drawable, ctx );
+}
+
+
+static GLXPixmap
+Fake_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap )
+{
+   XMesaVisual v;
+   XMesaBuffer b;
+
+   v = find_glx_visual( dpy, visinfo );
+   if (!v) {
+      v = create_glx_visual( dpy, visinfo );
+      if (!v) {
+         /* unusable visual */
+         return 0;
+      }
+   }
+
+   b = XMesaCreatePixmapBuffer( v, pixmap, 0 );
+   if (!b) {
+      return 0;
+   }
+   return b->drawable;
+}
+
+
+/*** GLX_MESA_pixmap_colormap ***/
+
+static GLXPixmap
+Fake_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo,
+                             Pixmap pixmap, Colormap cmap )
+{
+   XMesaVisual v;
+   XMesaBuffer b;
+
+   v = find_glx_visual( dpy, visinfo );
+   if (!v) {
+      v = create_glx_visual( dpy, visinfo );
+      if (!v) {
+         /* unusable visual */
+         return 0;
+      }
+   }
+
+   b = XMesaCreatePixmapBuffer( v, pixmap, cmap );
+   if (!b) {
+      return 0;
+   }
+   return b->drawable;
+}
+
+
+static void
+Fake_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap )
+{
+   XMesaBuffer b = XMesaFindBuffer(dpy, pixmap);
+   if (b) {
+      XMesaDestroyBuffer(b);
+   }
+   else if (_mesa_getenv("MESA_DEBUG")) {
+      _mesa_warning(NULL, "Mesa: glXDestroyGLXPixmap: invalid pixmap\n");
+   }
+}
+
+
+static void
+Fake_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst,
+                     unsigned long mask )
+{
+   struct fake_glx_context *fakeSrc = (struct fake_glx_context *) src;
+   struct fake_glx_context *fakeDst = (struct fake_glx_context *) dst;
+   XMesaContext xm_src = fakeSrc->xmesaContext;
+   XMesaContext xm_dst = fakeDst->xmesaContext;
+   (void) dpy;
+   if (MakeCurrent_PrevContext == src) {
+      _mesa_Flush();
+   }
+   st_copy_context_state( xm_src->st, xm_dst->st, (GLuint) mask );
+}
+
+
+static Bool
+Fake_glXQueryExtension( Display *dpy, int *errorb, int *event )
+{
+   /* Mesa's GLX isn't really an X extension but we try to act like one. */
+   (void) dpy;
+   (void) errorb;
+   (void) event;
+   return True;
+}
+
+
+extern void _kw_ungrab_all( Display *dpy );
+void _kw_ungrab_all( Display *dpy )
+{
+   XUngrabPointer( dpy, CurrentTime );
+   XUngrabKeyboard( dpy, CurrentTime );
+}
+
+
+static void
+Fake_glXDestroyContext( Display *dpy, GLXContext ctx )
+{
+   struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx;
+   (void) dpy;
+   MakeCurrent_PrevContext = 0;
+   MakeCurrent_PrevDrawable = 0;
+   MakeCurrent_PrevReadable = 0;
+   MakeCurrent_PrevDrawBuffer = 0;
+   MakeCurrent_PrevReadBuffer = 0;
+   XMesaDestroyContext( glxCtx->xmesaContext );
+   XMesaGarbageCollect();
+   _mesa_free(glxCtx);
+}
+
+
+static Bool
+Fake_glXIsDirect( Display *dpy, GLXContext ctx )
+{
+   (void) dpy;
+   (void) ctx;
+   return False;
+}
+
+
+
+static void
+Fake_glXSwapBuffers( Display *dpy, GLXDrawable drawable )
+{
+   XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable );
+   static boolean firsttime = 1, no_rast = 0;
+
+   if (firsttime) {
+      no_rast = getenv("SP_NO_RAST") != NULL;
+      firsttime = 0;
+   }
+
+   if (no_rast)
+      return;
+
+   if (buffer) {
+      XMesaSwapBuffers(buffer);
+   }
+   else if (_mesa_getenv("MESA_DEBUG")) {
+      _mesa_warning(NULL, "glXSwapBuffers: invalid drawable 0x%x\n",
+                    (int) drawable);
+   }
+}
+
+
+
+/*** GLX_MESA_copy_sub_buffer ***/
+
+static void
+Fake_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable,
+                           int x, int y, int width, int height )
+{
+   XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable );
+   if (buffer) {
+      XMesaCopySubBuffer(buffer, x, y, width, height);
+   }
+   else if (_mesa_getenv("MESA_DEBUG")) {
+      _mesa_warning(NULL, "Mesa: glXCopySubBufferMESA: invalid drawable\n");
+   }
+}
+
+
+static Bool
+Fake_glXQueryVersion( Display *dpy, int *maj, int *min )
+{
+   (void) dpy;
+   /* Return GLX version, not Mesa version */
+   assert(CLIENT_MAJOR_VERSION == SERVER_MAJOR_VERSION);
+   *maj = CLIENT_MAJOR_VERSION;
+   *min = MIN2( CLIENT_MINOR_VERSION, SERVER_MINOR_VERSION );
+   return True;
+}
+
+
+/*
+ * Query the GLX attributes of the given XVisualInfo.
+ */
+static int
+get_config( XMesaVisual xmvis, int attrib, int *value, GLboolean fbconfig )
+{
+   ASSERT(xmvis);
+   switch(attrib) {
+      case GLX_USE_GL:
+         if (fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+         *value = (int) True;
+	 return 0;
+      case GLX_BUFFER_SIZE:
+	 *value = xmvis->visinfo->depth;
+	 return 0;
+      case GLX_LEVEL:
+	 *value = xmvis->mesa_visual.level;
+	 return 0;
+      case GLX_RGBA:
+         if (fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+	 if (xmvis->mesa_visual.rgbMode) {
+	    *value = True;
+	 }
+	 else {
+	    *value = False;
+	 }
+	 return 0;
+      case GLX_DOUBLEBUFFER:
+	 *value = (int) xmvis->mesa_visual.doubleBufferMode;
+	 return 0;
+      case GLX_STEREO:
+	 *value = (int) xmvis->mesa_visual.stereoMode;
+	 return 0;
+      case GLX_AUX_BUFFERS:
+	 *value = xmvis->mesa_visual.numAuxBuffers;
+	 return 0;
+      case GLX_RED_SIZE:
+         *value = xmvis->mesa_visual.redBits;
+	 return 0;
+      case GLX_GREEN_SIZE:
+         *value = xmvis->mesa_visual.greenBits;
+	 return 0;
+      case GLX_BLUE_SIZE:
+         *value = xmvis->mesa_visual.blueBits;
+	 return 0;
+      case GLX_ALPHA_SIZE:
+         *value = xmvis->mesa_visual.alphaBits;
+	 return 0;
+      case GLX_DEPTH_SIZE:
+         *value = xmvis->mesa_visual.depthBits;
+	 return 0;
+      case GLX_STENCIL_SIZE:
+	 *value = xmvis->mesa_visual.stencilBits;
+	 return 0;
+      case GLX_ACCUM_RED_SIZE:
+	 *value = xmvis->mesa_visual.accumRedBits;
+	 return 0;
+      case GLX_ACCUM_GREEN_SIZE:
+	 *value = xmvis->mesa_visual.accumGreenBits;
+	 return 0;
+      case GLX_ACCUM_BLUE_SIZE:
+	 *value = xmvis->mesa_visual.accumBlueBits;
+	 return 0;
+      case GLX_ACCUM_ALPHA_SIZE:
+         *value = xmvis->mesa_visual.accumAlphaBits;
+	 return 0;
+
+      /*
+       * GLX_EXT_visual_info extension
+       */
+      case GLX_X_VISUAL_TYPE_EXT:
+         switch (xmvis->visinfo->CLASS) {
+            case StaticGray:   *value = GLX_STATIC_GRAY_EXT;   return 0;
+            case GrayScale:    *value = GLX_GRAY_SCALE_EXT;    return 0;
+            case StaticColor:  *value = GLX_STATIC_GRAY_EXT;   return 0;
+            case PseudoColor:  *value = GLX_PSEUDO_COLOR_EXT;  return 0;
+            case TrueColor:    *value = GLX_TRUE_COLOR_EXT;    return 0;
+            case DirectColor:  *value = GLX_DIRECT_COLOR_EXT;  return 0;
+         }
+         return 0;
+      case GLX_TRANSPARENT_TYPE_EXT:
+         /* normal planes */
+         *value = GLX_NONE_EXT;
+         return 0;
+      case GLX_TRANSPARENT_INDEX_VALUE_EXT:
+         /* undefined */
+         return 0;
+      case GLX_TRANSPARENT_RED_VALUE_EXT:
+         /* undefined */
+         return 0;
+      case GLX_TRANSPARENT_GREEN_VALUE_EXT:
+         /* undefined */
+         return 0;
+      case GLX_TRANSPARENT_BLUE_VALUE_EXT:
+         /* undefined */
+         return 0;
+      case GLX_TRANSPARENT_ALPHA_VALUE_EXT:
+         /* undefined */
+         return 0;
+
+      /*
+       * GLX_EXT_visual_info extension
+       */
+      case GLX_VISUAL_CAVEAT_EXT:
+         /* test for zero, just in case */
+         if (xmvis->mesa_visual.visualRating > 0)
+            *value = xmvis->mesa_visual.visualRating;
+         else
+            *value = GLX_NONE_EXT;
+         return 0;
+
+      /*
+       * GLX_ARB_multisample
+       */
+      case GLX_SAMPLE_BUFFERS_ARB:
+         *value = 0;
+         return 0;
+      case GLX_SAMPLES_ARB:
+         *value = 0;
+         return 0;
+
+      /*
+       * For FBConfigs:
+       */
+      case GLX_SCREEN_EXT:
+         if (!fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+         *value = xmvis->visinfo->screen;
+         break;
+      case GLX_DRAWABLE_TYPE: /*SGIX too */
+         if (!fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+         *value = GLX_WINDOW_BIT | GLX_PIXMAP_BIT | GLX_PBUFFER_BIT;
+         break;
+      case GLX_RENDER_TYPE_SGIX:
+         if (!fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+         if (xmvis->mesa_visual.rgbMode)
+            *value = GLX_RGBA_BIT;
+         else
+            *value = GLX_COLOR_INDEX_BIT;
+         break;
+      case GLX_X_RENDERABLE_SGIX:
+         if (!fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+         *value = True; /* XXX really? */
+         break;
+      case GLX_FBCONFIG_ID_SGIX:
+         if (!fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+         *value = xmvis->visinfo->visualid;
+         break;
+      case GLX_MAX_PBUFFER_WIDTH:
+         if (!fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+         /* XXX or MAX_WIDTH? */
+         *value = DisplayWidth(xmvis->display, xmvis->visinfo->screen);
+         break;
+      case GLX_MAX_PBUFFER_HEIGHT:
+         if (!fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+         *value = DisplayHeight(xmvis->display, xmvis->visinfo->screen);
+         break;
+      case GLX_MAX_PBUFFER_PIXELS:
+         if (!fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+         *value = DisplayWidth(xmvis->display, xmvis->visinfo->screen) *
+                  DisplayHeight(xmvis->display, xmvis->visinfo->screen);
+         break;
+      case GLX_VISUAL_ID:
+         if (!fbconfig)
+            return GLX_BAD_ATTRIBUTE;
+         *value = xmvis->visinfo->visualid;
+         break;
+
+#ifdef GLX_EXT_texture_from_pixmap
+      case GLX_BIND_TO_TEXTURE_RGB_EXT:
+         *value = True; /*XXX*/
+         break;
+      case GLX_BIND_TO_TEXTURE_RGBA_EXT:
+         /* XXX review */
+         *value = xmvis->mesa_visual.alphaBits > 0 ? True : False;
+         break;
+      case GLX_BIND_TO_MIPMAP_TEXTURE_EXT:
+         *value = True; /*XXX*/
+         break;
+      case GLX_BIND_TO_TEXTURE_TARGETS_EXT:
+         *value = (GLX_TEXTURE_1D_BIT_EXT |
+                   GLX_TEXTURE_2D_BIT_EXT |
+                   GLX_TEXTURE_RECTANGLE_BIT_EXT); /*XXX*/
+         break;
+      case GLX_Y_INVERTED_EXT:
+         *value = True; /*XXX*/
+         break;
+#endif
+
+      default:
+	 return GLX_BAD_ATTRIBUTE;
+   }
+   return Success;
+}
+
+
+static int
+Fake_glXGetConfig( Display *dpy, XVisualInfo *visinfo,
+                   int attrib, int *value )
+{
+   XMesaVisual xmvis;
+   int k;
+   if (!dpy || !visinfo)
+      return GLX_BAD_ATTRIBUTE;
+
+   xmvis = find_glx_visual( dpy, visinfo );
+   if (!xmvis) {
+      /* this visual wasn't obtained with glXChooseVisual */
+      xmvis = create_glx_visual( dpy, visinfo );
+      if (!xmvis) {
+	 /* this visual can't be used for GL rendering */
+	 if (attrib==GLX_USE_GL) {
+	    *value = (int) False;
+	    return 0;
+	 }
+	 else {
+	    return GLX_BAD_VISUAL;
+	 }
+      }
+   }
+
+   k = get_config(xmvis, attrib, value, GL_FALSE);
+   return k;
+}
+
+
+static void
+Fake_glXWaitGL( void )
+{
+   XMesaContext xmesa = XMesaGetCurrentContext();
+   XMesaFlush( xmesa );
+}
+
+
+
+static void
+Fake_glXWaitX( void )
+{
+   XMesaContext xmesa = XMesaGetCurrentContext();
+   XMesaFlush( xmesa );
+}
+
+
+static const char *
+get_extensions( void )
+{
+   return EXTENSIONS;
+}
+
+
+
+/* GLX 1.1 and later */
+static const char *
+Fake_glXQueryExtensionsString( Display *dpy, int screen )
+{
+   (void) dpy;
+   (void) screen;
+   return get_extensions();
+}
+
+
+
+/* GLX 1.1 and later */
+static const char *
+Fake_glXQueryServerString( Display *dpy, int screen, int name )
+{
+   static char version[1000];
+   _mesa_sprintf(version, "%d.%d %s",
+                 SERVER_MAJOR_VERSION, SERVER_MINOR_VERSION, MESA_GLX_VERSION);
+
+   (void) dpy;
+   (void) screen;
+
+   switch (name) {
+      case GLX_EXTENSIONS:
+         return get_extensions();
+      case GLX_VENDOR:
+	 return VENDOR;
+      case GLX_VERSION:
+	 return version;
+      default:
+         return NULL;
+   }
+}
+
+
+
+/* GLX 1.1 and later */
+static const char *
+Fake_glXGetClientString( Display *dpy, int name )
+{
+   static char version[1000];
+   _mesa_sprintf(version, "%d.%d %s", CLIENT_MAJOR_VERSION,
+                 CLIENT_MINOR_VERSION, MESA_GLX_VERSION);
+
+   (void) dpy;
+
+   switch (name) {
+      case GLX_EXTENSIONS:
+         return get_extensions();
+      case GLX_VENDOR:
+	 return VENDOR;
+      case GLX_VERSION:
+	 return version;
+      default:
+         return NULL;
+   }
+}
+
+
+
+/*
+ * GLX 1.3 and later
+ */
+
+
+static int
+Fake_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config,
+                           int attribute, int *value )
+{
+   XMesaVisual v = (XMesaVisual) config;
+   (void) dpy;
+   (void) config;
+
+   if (!dpy || !config || !value)
+      return -1;
+
+   return get_config(v, attribute, value, GL_TRUE);
+}
+
+
+static GLXFBConfig *
+Fake_glXGetFBConfigs( Display *dpy, int screen, int *nelements )
+{
+   XVisualInfo *visuals, visTemplate;
+   const long visMask = VisualScreenMask;
+   int i;
+
+   /* Get list of all X visuals */
+   visTemplate.screen = screen;
+   visuals = XGetVisualInfo(dpy, visMask, &visTemplate, nelements);
+   if (*nelements > 0) {
+      XMesaVisual *results;
+      results = (XMesaVisual *) _mesa_malloc(*nelements * sizeof(XMesaVisual));
+      if (!results) {
+         *nelements = 0;
+         return NULL;
+      }
+      for (i = 0; i < *nelements; i++) {
+         results[i] = create_glx_visual(dpy, visuals + i);
+      }
+      return (GLXFBConfig *) results;
+   }
+   return NULL;
+}
+
+
+static GLXFBConfig *
+Fake_glXChooseFBConfig( Display *dpy, int screen,
+                        const int *attribList, int *nitems )
+{
+   XMesaVisual xmvis;
+
+   if (!attribList || !attribList[0]) {
+      /* return list of all configs (per GLX_SGIX_fbconfig spec) */
+      return Fake_glXGetFBConfigs(dpy, screen, nitems);
+   }
+
+   xmvis = choose_visual(dpy, screen, attribList, GL_TRUE);
+   if (xmvis) {
+      GLXFBConfig *config = (GLXFBConfig *) _mesa_malloc(sizeof(XMesaVisual));
+      if (!config) {
+         *nitems = 0;
+         return NULL;
+      }
+      *nitems = 1;
+      config[0] = (GLXFBConfig) xmvis;
+      return (GLXFBConfig *) config;
+   }
+   else {
+      *nitems = 0;
+      return NULL;
+   }
+}
+
+
+static XVisualInfo *
+Fake_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config )
+{
+   if (dpy && config) {
+      XMesaVisual xmvis = (XMesaVisual) config;
+#if 0      
+      return xmvis->vishandle;
+#else
+      /* create a new vishandle - the cached one may be stale */
+      xmvis->vishandle = (XVisualInfo *) _mesa_malloc(sizeof(XVisualInfo));
+      if (xmvis->vishandle) {
+         _mesa_memcpy(xmvis->vishandle, xmvis->visinfo, sizeof(XVisualInfo));
+      }
+      return xmvis->vishandle;
+#endif
+   }
+   else {
+      return NULL;
+   }
+}
+
+
+static GLXWindow
+Fake_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win,
+                      const int *attribList )
+{
+   XMesaVisual xmvis = (XMesaVisual) config;
+   XMesaBuffer xmbuf;
+   if (!xmvis)
+      return 0;
+
+   xmbuf = XMesaCreateWindowBuffer(xmvis, win);
+   if (!xmbuf)
+      return 0;
+
+   (void) dpy;
+   (void) attribList;  /* Ignored in GLX 1.3 */
+
+   return win;  /* A hack for now */
+}
+
+
+static void
+Fake_glXDestroyWindow( Display *dpy, GLXWindow window )
+{
+   XMesaBuffer b = XMesaFindBuffer(dpy, (Drawable) window);
+   if (b)
+      XMesaDestroyBuffer(b);
+   /* don't destroy X window */
+}
+
+
+/* XXX untested */
+static GLXPixmap
+Fake_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap,
+                      const int *attribList )
+{
+   XMesaVisual v = (XMesaVisual) config;
+   XMesaBuffer b;
+   const int *attr;
+   int target = 0, format = 0, mipmap = 0;
+   int value;
+
+   if (!dpy || !config || !pixmap)
+      return 0;
+
+   for (attr = attribList; *attr; attr++) {
+      switch (*attr) {
+      case GLX_TEXTURE_FORMAT_EXT:
+         attr++;
+         switch (*attr) {
+         case GLX_TEXTURE_FORMAT_NONE_EXT:
+         case GLX_TEXTURE_FORMAT_RGB_EXT:
+         case GLX_TEXTURE_FORMAT_RGBA_EXT:
+            format = *attr;
+            break;
+         default:
+            /* error */
+            return 0;
+         }
+         break;
+      case GLX_TEXTURE_TARGET_EXT:
+         attr++;
+         switch (*attr) {
+         case GLX_TEXTURE_1D_EXT:
+         case GLX_TEXTURE_2D_EXT:
+         case GLX_TEXTURE_RECTANGLE_EXT:
+            target = *attr;
+            break;
+         default:
+            /* error */
+            return 0;
+         }
+         break;
+      case GLX_MIPMAP_TEXTURE_EXT:
+         attr++;
+         if (*attr)
+            mipmap = 1;
+         break;
+      default:
+         /* error */
+         return 0;
+      }
+   }
+
+   if (format == GLX_TEXTURE_FORMAT_RGB_EXT) {
+      if (get_config(v, GLX_BIND_TO_TEXTURE_RGB_EXT,
+                     &value, GL_TRUE) != Success
+          || !value) {
+         return 0; /* error! */
+      }
+   }
+   else if (format == GLX_TEXTURE_FORMAT_RGBA_EXT) {
+      if (get_config(v, GLX_BIND_TO_TEXTURE_RGBA_EXT,
+                     &value, GL_TRUE) != Success
+          || !value) {
+         return 0; /* error! */
+      }
+   }
+   if (mipmap) {
+      if (get_config(v, GLX_BIND_TO_MIPMAP_TEXTURE_EXT,
+                     &value, GL_TRUE) != Success
+          || !value) {
+         return 0; /* error! */
+      }
+   }
+   if (target == GLX_TEXTURE_1D_EXT) {
+      if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT,
+                     &value, GL_TRUE) != Success
+          || (value & GLX_TEXTURE_1D_BIT_EXT) == 0) {
+         return 0; /* error! */
+      }
+   }
+   else if (target == GLX_TEXTURE_2D_EXT) {
+      if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT,
+                     &value, GL_TRUE) != Success
+          || (value & GLX_TEXTURE_2D_BIT_EXT) == 0) {
+         return 0; /* error! */
+      }
+   }
+   if (target == GLX_TEXTURE_RECTANGLE_EXT) {
+      if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT,
+                     &value, GL_TRUE) != Success
+          || (value & GLX_TEXTURE_RECTANGLE_BIT_EXT) == 0) {
+         return 0; /* error! */
+      }
+   }
+
+   if (format || target || mipmap) {
+      /* texture from pixmap */
+      b = XMesaCreatePixmapTextureBuffer(v, pixmap, 0, format, target, mipmap);
+   }
+   else {
+      b = XMesaCreatePixmapBuffer( v, pixmap, 0 );
+   }
+   if (!b) {
+      return 0;
+   }
+
+   return pixmap;
+}
+
+
+static void
+Fake_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap )
+{
+   XMesaBuffer b = XMesaFindBuffer(dpy, (Drawable)pixmap);
+   if (b)
+      XMesaDestroyBuffer(b);
+   /* don't destroy X pixmap */
+}
+
+
+static GLXPbuffer
+Fake_glXCreatePbuffer( Display *dpy, GLXFBConfig config,
+                       const int *attribList )
+{
+   XMesaVisual xmvis = (XMesaVisual) config;
+   XMesaBuffer xmbuf;
+   const int *attrib;
+   int width = 0, height = 0;
+   GLboolean useLargest = GL_FALSE, preserveContents = GL_FALSE;
+
+   (void) dpy;
+
+   for (attrib = attribList; *attrib; attrib++) {
+      switch (*attrib) {
+         case GLX_PBUFFER_WIDTH:
+            attrib++;
+            width = *attrib;
+            break;
+         case GLX_PBUFFER_HEIGHT:
+            attrib++;
+            height = *attrib;
+            break;
+         case GLX_PRESERVED_CONTENTS:
+            attrib++;
+            preserveContents = *attrib; /* ignored */
+            break;
+         case GLX_LARGEST_PBUFFER:
+            attrib++;
+            useLargest = *attrib; /* ignored */
+            break;
+         default:
+            return 0;
+      }
+   }
+
+   /* not used at this time */
+   (void) useLargest;
+   (void) preserveContents;
+
+   if (width == 0 || height == 0)
+      return 0;
+
+   xmbuf = XMesaCreatePBuffer( xmvis, 0, width, height);
+   /* A GLXPbuffer handle must be an X Drawable because that's what
+    * glXMakeCurrent takes.
+    */
+   if (xmbuf)
+      return (GLXPbuffer) xmbuf->drawable;
+   else
+      return 0;
+}
+
+
+static void
+Fake_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf )
+{
+   XMesaBuffer b = XMesaFindBuffer(dpy, pbuf);
+   if (b) {
+      XMesaDestroyBuffer(b);
+   }
+}
+
+
+static void
+Fake_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute,
+                       unsigned int *value )
+{
+   XMesaBuffer xmbuf = XMesaFindBuffer(dpy, draw);
+   if (!xmbuf)
+      return;
+
+   switch (attribute) {
+      case GLX_WIDTH:
+         *value = xmesa_buffer_width(xmbuf);
+         break;
+      case GLX_HEIGHT:
+         *value = xmesa_buffer_width(xmbuf);
+         break;
+      case GLX_PRESERVED_CONTENTS:
+         *value = True;
+         break;
+      case GLX_LARGEST_PBUFFER:
+         *value = xmesa_buffer_width(xmbuf) * xmesa_buffer_height(xmbuf);
+         break;
+      case GLX_FBCONFIG_ID:
+         *value = xmbuf->xm_visual->visinfo->visualid;
+         return;
+#ifdef GLX_EXT_texture_from_pixmap
+      case GLX_TEXTURE_FORMAT_EXT:
+         *value = xmbuf->TextureFormat;
+         break;
+      case GLX_TEXTURE_TARGET_EXT:
+         *value = xmbuf->TextureTarget;
+         break;
+      case GLX_MIPMAP_TEXTURE_EXT:
+         *value = xmbuf->TextureMipmap;
+         break;
+#endif
+
+      default:
+         return; /* raise BadValue error */
+   }
+}
+
+
+static GLXContext
+Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config,
+                          int renderType, GLXContext shareList, Bool direct )
+{
+   struct fake_glx_context *glxCtx;
+   struct fake_glx_context *shareCtx = (struct fake_glx_context *) shareList;
+   XMesaVisual xmvis = (XMesaVisual) config;
+
+   if (!dpy || !config ||
+       (renderType != GLX_RGBA_TYPE && renderType != GLX_COLOR_INDEX_TYPE))
+      return 0;
+
+   glxCtx = CALLOC_STRUCT(fake_glx_context);
+   if (!glxCtx)
+      return 0;
+
+   /* deallocate unused windows/buffers */
+   XMesaGarbageCollect();
+
+   glxCtx->xmesaContext = XMesaCreateContext(xmvis,
+                                   shareCtx ? shareCtx->xmesaContext : NULL);
+   if (!glxCtx->xmesaContext) {
+      _mesa_free(glxCtx);
+      return NULL;
+   }
+
+   glxCtx->glxContext.isDirect = GL_FALSE;
+   glxCtx->glxContext.currentDpy = dpy;
+   glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
+
+   assert((void *) glxCtx == (void *) &(glxCtx->glxContext));
+
+   return (GLXContext) glxCtx;
+}
+
+
+static int
+Fake_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value )
+{
+   struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx;
+   XMesaContext xmctx = glxCtx->xmesaContext;
+
+   (void) dpy;
+   (void) ctx;
+
+   switch (attribute) {
+   case GLX_FBCONFIG_ID:
+      *value = xmctx->xm_visual->visinfo->visualid;
+      break;
+   case GLX_RENDER_TYPE:
+      if (xmctx->xm_visual->mesa_visual.rgbMode)
+         *value = GLX_RGBA_BIT;
+      else
+         *value = GLX_COLOR_INDEX_BIT;
+      break;
+   case GLX_SCREEN:
+      *value = 0;
+      return Success;
+   default:
+      return GLX_BAD_ATTRIBUTE;
+   }
+   return 0;
+}
+
+
+static void
+Fake_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask )
+{
+   XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable);
+   if (xmbuf)
+      xmbuf->selectedEvents = mask;
+}
+
+
+static void
+Fake_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable,
+                          unsigned long *mask )
+{
+   XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable);
+   if (xmbuf)
+      *mask = xmbuf->selectedEvents;
+   else
+      *mask = 0;
+}
+
+
+
+/*** GLX_SGI_swap_control ***/
+
+static int
+Fake_glXSwapIntervalSGI(int interval)
+{
+   (void) interval;
+   return 0;
+}
+
+
+
+/*** GLX_SGI_video_sync ***/
+
+static unsigned int FrameCounter = 0;
+
+static int
+Fake_glXGetVideoSyncSGI(unsigned int *count)
+{
+   /* this is a bogus implementation */
+   *count = FrameCounter++;
+   return 0;
+}
+
+static int
+Fake_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
+{
+   if (divisor <= 0 || remainder < 0)
+      return GLX_BAD_VALUE;
+   /* this is a bogus implementation */
+   FrameCounter++;
+   while (FrameCounter % divisor != remainder)
+      FrameCounter++;
+   *count = FrameCounter;
+   return 0;
+}
+
+
+
+/*** GLX_SGI_make_current_read ***/
+
+static Bool
+Fake_glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx)
+{
+   return Fake_glXMakeContextCurrent( dpy, draw, read, ctx );
+}
+
+/* not used
+static GLXDrawable
+Fake_glXGetCurrentReadDrawableSGI(void)
+{
+   return 0;
+}
+*/
+
+
+/*** GLX_SGIX_video_source ***/
+#if defined(_VL_H)
+
+static GLXVideoSourceSGIX
+Fake_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode)
+{
+   (void) dpy;
+   (void) screen;
+   (void) server;
+   (void) path;
+   (void) nodeClass;
+   (void) drainNode;
+   return 0;
+}
+
+static void
+Fake_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src)
+{
+   (void) dpy;
+   (void) src;
+}
+
+#endif
+
+
+/*** GLX_EXT_import_context ***/
+
+static void
+Fake_glXFreeContextEXT(Display *dpy, GLXContext context)
+{
+   (void) dpy;
+   (void) context;
+}
+
+static GLXContextID
+Fake_glXGetContextIDEXT(const GLXContext context)
+{
+   (void) context;
+   return 0;
+}
+
+static GLXContext
+Fake_glXImportContextEXT(Display *dpy, GLXContextID contextID)
+{
+   (void) dpy;
+   (void) contextID;
+   return 0;
+}
+
+static int
+Fake_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value)
+{
+   (void) dpy;
+   (void) context;
+   (void) attribute;
+   (void) value;
+   return 0;
+}
+
+
+
+/*** GLX_SGIX_fbconfig ***/
+
+static int
+Fake_glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value)
+{
+   return Fake_glXGetFBConfigAttrib(dpy, config, attribute, value);
+}
+
+static GLXFBConfigSGIX *
+Fake_glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements)
+{
+   return (GLXFBConfig *) Fake_glXChooseFBConfig(dpy, screen, attrib_list, nelements);
+}
+
+
+static GLXPixmap
+Fake_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap)
+{
+   XMesaVisual xmvis = (XMesaVisual) config;
+   XMesaBuffer xmbuf = XMesaCreatePixmapBuffer(xmvis, pixmap, 0);
+   return xmbuf->drawable; /* need to return an X ID */
+}
+
+
+static GLXContext
+Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct)
+{
+   XMesaVisual xmvis = (XMesaVisual) config;
+   struct fake_glx_context *glxCtx;
+   struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list;
+
+   glxCtx = CALLOC_STRUCT(fake_glx_context);
+   if (!glxCtx)
+      return 0;
+
+   /* deallocate unused windows/buffers */
+   XMesaGarbageCollect();
+
+   glxCtx->xmesaContext = XMesaCreateContext(xmvis,
+                                   shareCtx ? shareCtx->xmesaContext : NULL);
+   if (!glxCtx->xmesaContext) {
+      _mesa_free(glxCtx);
+      return NULL;
+   }
+
+   glxCtx->glxContext.isDirect = GL_FALSE;
+   glxCtx->glxContext.currentDpy = dpy;
+   glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
+
+   assert((void *) glxCtx == (void *) &(glxCtx->glxContext));
+
+   return (GLXContext) glxCtx;
+}
+
+
+static XVisualInfo *
+Fake_glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config)
+{
+   return Fake_glXGetVisualFromFBConfig(dpy, config);
+}
+
+
+static GLXFBConfigSGIX
+Fake_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis)
+{
+   XMesaVisual xmvis = find_glx_visual(dpy, vis);
+   if (!xmvis) {
+      /* This visual wasn't found with glXChooseVisual() */
+      xmvis = create_glx_visual(dpy, vis);
+   }
+
+   return (GLXFBConfigSGIX) xmvis;
+}
+
+
+
+/*** GLX_SGIX_pbuffer ***/
+
+static GLXPbufferSGIX
+Fake_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config,
+                             unsigned int width, unsigned int height,
+                             int *attribList)
+{
+   XMesaVisual xmvis = (XMesaVisual) config;
+   XMesaBuffer xmbuf;
+   const int *attrib;
+   GLboolean useLargest = GL_FALSE, preserveContents = GL_FALSE;
+
+   (void) dpy;
+
+   for (attrib = attribList; attrib && *attrib; attrib++) {
+      switch (*attrib) {
+         case GLX_PRESERVED_CONTENTS_SGIX:
+            attrib++;
+            preserveContents = *attrib; /* ignored */
+            break;
+         case GLX_LARGEST_PBUFFER_SGIX:
+            attrib++;
+            useLargest = *attrib; /* ignored */
+            break;
+         default:
+            return 0;
+      }
+   }
+
+   /* not used at this time */
+   (void) useLargest;
+   (void) preserveContents;
+
+   xmbuf = XMesaCreatePBuffer( xmvis, 0, width, height);
+   /* A GLXPbuffer handle must be an X Drawable because that's what
+    * glXMakeCurrent takes.
+    */
+   return (GLXPbuffer) xmbuf->drawable;
+}
+
+
+static void
+Fake_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf)
+{
+   XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf);
+   if (xmbuf) {
+      XMesaDestroyBuffer(xmbuf);
+   }
+}
+
+
+static int
+Fake_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value)
+{
+   const XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf);
+
+   if (!xmbuf) {
+      /* Generate GLXBadPbufferSGIX for bad pbuffer */
+      return 0;
+   }
+
+   switch (attribute) {
+      case GLX_PRESERVED_CONTENTS_SGIX:
+         *value = True;
+         break;
+      case GLX_LARGEST_PBUFFER_SGIX:
+         *value = xmesa_buffer_width(xmbuf) * xmesa_buffer_height(xmbuf);
+         break;
+      case GLX_WIDTH_SGIX:
+         *value = xmesa_buffer_width(xmbuf);
+         break;
+      case GLX_HEIGHT_SGIX:
+         *value = xmesa_buffer_height(xmbuf);
+         break;
+      case GLX_EVENT_MASK_SGIX:
+         *value = 0;  /* XXX might be wrong */
+         break;
+      default:
+         *value = 0;
+   }
+   return 0;
+}
+
+
+static void
+Fake_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask)
+{
+   XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable);
+   if (xmbuf) {
+      /* Note: we'll never generate clobber events */
+      xmbuf->selectedEvents = mask;
+   }
+}
+
+
+static void
+Fake_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask)
+{
+   XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable);
+   if (xmbuf) {
+      *mask = xmbuf->selectedEvents;
+   }
+   else {
+      *mask = 0;
+   }
+}
+
+
+
+/*** GLX_SGI_cushion ***/
+
+static void
+Fake_glXCushionSGI(Display *dpy, Window win, float cushion)
+{
+   (void) dpy;
+   (void) win;
+   (void) cushion;
+}
+
+
+
+/*** GLX_SGIX_video_resize ***/
+
+static int
+Fake_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window)
+{
+   (void) dpy;
+   (void) screen;
+   (void) channel;
+   (void) window;
+   return 0;
+}
+
+static int
+Fake_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h)
+{
+   (void) dpy;
+   (void) screen;
+   (void) channel;
+   (void) x;
+   (void) y;
+   (void) w;
+   (void) h;
+   return 0;
+}
+
+static int
+Fake_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h)
+{
+   (void) dpy;
+   (void) screen;
+   (void) channel;
+   (void) x;
+   (void) y;
+   (void) w;
+   (void) h;
+   return 0;
+}
+
+static int
+Fake_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh)
+{
+   (void) dpy;
+   (void) screen;
+   (void) channel;
+   (void) dx;
+   (void) dy;
+   (void) dw;
+   (void) dh;
+   return 0;
+}
+
+static int
+Fake_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype)
+{
+   (void) dpy;
+   (void) screen;
+   (void) channel;
+   (void) synctype;
+   return 0;
+}
+
+
+
+/*** GLX_SGIX_dmbuffer **/
+
+#if defined(_DM_BUFFER_H_)
+static Bool
+Fake_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer)
+{
+   (void) dpy;
+   (void) pbuffer;
+   (void) params;
+   (void) dmbuffer;
+   return False;
+}
+#endif
+
+
+/*** GLX_SGIX_swap_group ***/
+
+static void
+Fake_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member)
+{
+   (void) dpy;
+   (void) drawable;
+   (void) member;
+}
+
+
+
+/*** GLX_SGIX_swap_barrier ***/
+
+static void
+Fake_glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier)
+{
+   (void) dpy;
+   (void) drawable;
+   (void) barrier;
+}
+
+static Bool
+Fake_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max)
+{
+   (void) dpy;
+   (void) screen;
+   (void) max;
+   return False;
+}
+
+
+
+/*** GLX_SUN_get_transparent_index ***/
+
+static Status
+Fake_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent)
+{
+   (void) dpy;
+   (void) overlay;
+   (void) underlay;
+   (void) pTransparent;
+   return 0;
+}
+
+
+
+/*** GLX_MESA_release_buffers ***/
+
+/*
+ * Release the depth, stencil, accum buffers attached to a GLXDrawable
+ * (a window or pixmap) prior to destroying the GLXDrawable.
+ */
+static Bool
+Fake_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d )
+{
+   XMesaBuffer b = XMesaFindBuffer(dpy, d);
+   if (b) {
+      XMesaDestroyBuffer(b);
+      return True;
+   }
+   return False;
+}
+
+/*** GLX_EXT_texture_from_pixmap ***/
+
+static void
+Fake_glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer,
+                        const int *attrib_list)
+{
+   XMesaBuffer b = XMesaFindBuffer(dpy, drawable);
+   if (b)
+      XMesaBindTexImage(dpy, b, buffer, attrib_list);
+}
+
+static void
+Fake_glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer)
+{
+   XMesaBuffer b = XMesaFindBuffer(dpy, drawable);
+   if (b)
+      XMesaReleaseTexImage(dpy, b, buffer);
+}
+
+
+
+/**
+ * Create a new GLX API dispatch table with its function pointers
+ * initialized to point to Mesa's "fake" GLX API functions.
+ *
+ * Note: there used to be a similar function
+ * (_real_GetGLXDispatchTable) that returns a new dispatch table with
+ * all pointers initalized to point to "real" GLX functions (which
+ * understand GLX wire protocol, etc).
+ */
+struct _glxapi_table *
+_mesa_GetGLXDispatchTable(void)
+{
+   static struct _glxapi_table glx;
+
+   /* be sure our dispatch table size <= libGL's table */
+   {
+      GLuint size = sizeof(struct _glxapi_table) / sizeof(void *);
+      (void) size;
+      assert(_glxapi_get_dispatch_table_size() >= size);
+   }
+
+   /* initialize the whole table to no-ops */
+   _glxapi_set_no_op_table(&glx);
+
+   /* now initialize the table with the functions I implement */
+   glx.ChooseVisual = Fake_glXChooseVisual;
+   glx.CopyContext = Fake_glXCopyContext;
+   glx.CreateContext = Fake_glXCreateContext;
+   glx.CreateGLXPixmap = Fake_glXCreateGLXPixmap;
+   glx.DestroyContext = Fake_glXDestroyContext;
+   glx.DestroyGLXPixmap = Fake_glXDestroyGLXPixmap;
+   glx.GetConfig = Fake_glXGetConfig;
+   /*glx.GetCurrentContext = Fake_glXGetCurrentContext;*/
+   /*glx.GetCurrentDrawable = Fake_glXGetCurrentDrawable;*/
+   glx.IsDirect = Fake_glXIsDirect;
+   glx.MakeCurrent = Fake_glXMakeCurrent;
+   glx.QueryExtension = Fake_glXQueryExtension;
+   glx.QueryVersion = Fake_glXQueryVersion;
+   glx.SwapBuffers = Fake_glXSwapBuffers;
+   glx.UseXFont = Fake_glXUseXFont;
+   glx.WaitGL = Fake_glXWaitGL;
+   glx.WaitX = Fake_glXWaitX;
+
+   /*** GLX_VERSION_1_1 ***/
+   glx.GetClientString = Fake_glXGetClientString;
+   glx.QueryExtensionsString = Fake_glXQueryExtensionsString;
+   glx.QueryServerString = Fake_glXQueryServerString;
+
+   /*** GLX_VERSION_1_2 ***/
+   /*glx.GetCurrentDisplay = Fake_glXGetCurrentDisplay;*/
+
+   /*** GLX_VERSION_1_3 ***/
+   glx.ChooseFBConfig = Fake_glXChooseFBConfig;
+   glx.CreateNewContext = Fake_glXCreateNewContext;
+   glx.CreatePbuffer = Fake_glXCreatePbuffer;
+   glx.CreatePixmap = Fake_glXCreatePixmap;
+   glx.CreateWindow = Fake_glXCreateWindow;
+   glx.DestroyPbuffer = Fake_glXDestroyPbuffer;
+   glx.DestroyPixmap = Fake_glXDestroyPixmap;
+   glx.DestroyWindow = Fake_glXDestroyWindow;
+   /*glx.GetCurrentReadDrawable = Fake_glXGetCurrentReadDrawable;*/
+   glx.GetFBConfigAttrib = Fake_glXGetFBConfigAttrib;
+   glx.GetFBConfigs = Fake_glXGetFBConfigs;
+   glx.GetSelectedEvent = Fake_glXGetSelectedEvent;
+   glx.GetVisualFromFBConfig = Fake_glXGetVisualFromFBConfig;
+   glx.MakeContextCurrent = Fake_glXMakeContextCurrent;
+   glx.QueryContext = Fake_glXQueryContext;
+   glx.QueryDrawable = Fake_glXQueryDrawable;
+   glx.SelectEvent = Fake_glXSelectEvent;
+
+   /*** GLX_SGI_swap_control ***/
+   glx.SwapIntervalSGI = Fake_glXSwapIntervalSGI;
+
+   /*** GLX_SGI_video_sync ***/
+   glx.GetVideoSyncSGI = Fake_glXGetVideoSyncSGI;
+   glx.WaitVideoSyncSGI = Fake_glXWaitVideoSyncSGI;
+
+   /*** GLX_SGI_make_current_read ***/
+   glx.MakeCurrentReadSGI = Fake_glXMakeCurrentReadSGI;
+   /*glx.GetCurrentReadDrawableSGI = Fake_glXGetCurrentReadDrawableSGI;*/
+
+/*** GLX_SGIX_video_source ***/
+#if defined(_VL_H)
+   glx.CreateGLXVideoSourceSGIX = Fake_glXCreateGLXVideoSourceSGIX;
+   glx.DestroyGLXVideoSourceSGIX = Fake_glXDestroyGLXVideoSourceSGIX;
+#endif
+
+   /*** GLX_EXT_import_context ***/
+   glx.FreeContextEXT = Fake_glXFreeContextEXT;
+   glx.GetContextIDEXT = Fake_glXGetContextIDEXT;
+   /*glx.GetCurrentDisplayEXT = Fake_glXGetCurrentDisplayEXT;*/
+   glx.ImportContextEXT = Fake_glXImportContextEXT;
+   glx.QueryContextInfoEXT = Fake_glXQueryContextInfoEXT;
+
+   /*** GLX_SGIX_fbconfig ***/
+   glx.GetFBConfigAttribSGIX = Fake_glXGetFBConfigAttribSGIX;
+   glx.ChooseFBConfigSGIX = Fake_glXChooseFBConfigSGIX;
+   glx.CreateGLXPixmapWithConfigSGIX = Fake_glXCreateGLXPixmapWithConfigSGIX;
+   glx.CreateContextWithConfigSGIX = Fake_glXCreateContextWithConfigSGIX;
+   glx.GetVisualFromFBConfigSGIX = Fake_glXGetVisualFromFBConfigSGIX;
+   glx.GetFBConfigFromVisualSGIX = Fake_glXGetFBConfigFromVisualSGIX;
+
+   /*** GLX_SGIX_pbuffer ***/
+   glx.CreateGLXPbufferSGIX = Fake_glXCreateGLXPbufferSGIX;
+   glx.DestroyGLXPbufferSGIX = Fake_glXDestroyGLXPbufferSGIX;
+   glx.QueryGLXPbufferSGIX = Fake_glXQueryGLXPbufferSGIX;
+   glx.SelectEventSGIX = Fake_glXSelectEventSGIX;
+   glx.GetSelectedEventSGIX = Fake_glXGetSelectedEventSGIX;
+
+   /*** GLX_SGI_cushion ***/
+   glx.CushionSGI = Fake_glXCushionSGI;
+
+   /*** GLX_SGIX_video_resize ***/
+   glx.BindChannelToWindowSGIX = Fake_glXBindChannelToWindowSGIX;
+   glx.ChannelRectSGIX = Fake_glXChannelRectSGIX;
+   glx.QueryChannelRectSGIX = Fake_glXQueryChannelRectSGIX;
+   glx.QueryChannelDeltasSGIX = Fake_glXQueryChannelDeltasSGIX;
+   glx.ChannelRectSyncSGIX = Fake_glXChannelRectSyncSGIX;
+
+   /*** GLX_SGIX_dmbuffer **/
+#if defined(_DM_BUFFER_H_)
+   glx.AssociateDMPbufferSGIX = NULL;
+#endif
+
+   /*** GLX_SGIX_swap_group ***/
+   glx.JoinSwapGroupSGIX = Fake_glXJoinSwapGroupSGIX;
+
+   /*** GLX_SGIX_swap_barrier ***/
+   glx.BindSwapBarrierSGIX = Fake_glXBindSwapBarrierSGIX;
+   glx.QueryMaxSwapBarriersSGIX = Fake_glXQueryMaxSwapBarriersSGIX;
+
+   /*** GLX_SUN_get_transparent_index ***/
+   glx.GetTransparentIndexSUN = Fake_glXGetTransparentIndexSUN;
+
+   /*** GLX_MESA_copy_sub_buffer ***/
+   glx.CopySubBufferMESA = Fake_glXCopySubBufferMESA;
+
+   /*** GLX_MESA_release_buffers ***/
+   glx.ReleaseBuffersMESA = Fake_glXReleaseBuffersMESA;
+
+   /*** GLX_MESA_pixmap_colormap ***/
+   glx.CreateGLXPixmapMESA = Fake_glXCreateGLXPixmapMESA;
+
+   /*** GLX_EXT_texture_from_pixmap ***/
+   glx.BindTexImageEXT = Fake_glXBindTexImageEXT;
+   glx.ReleaseTexImageEXT = Fake_glXReleaseTexImageEXT;
+
+   return &glx;
+}
diff --git a/src/mesa/drivers/dri/i965/brw_vs_tnl.c b/src/gallium/state_trackers/glx/xlib/fakeglx.h
index eacc289f1f..e5fd960072 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_tnl.c
+++ b/src/gallium/state_trackers/glx/xlib/fakeglx.h
@@ -1,8 +1,9 @@
+
 /*
  * Mesa 3-D graphics library
- * Version:  6.3
+ * Version:  3.5
  *
- * Copyright (C) 2005  Tungsten Graphics   All Rights Reserved.
+ * Copyright (C) 1999-2000  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -17,43 +18,24 @@
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * TUNGSTEN GRAPHICS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
- * WHETHER IN
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
  * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
-/**
- * \file t_vp_build.c
- * Create a vertex program to execute the current fixed function T&L pipeline.
- * \author Keith Whitwell
- */
 
+#ifndef FAKEGLX_H
+#define FAKEGLX_H
 
-#include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
-#include "brw_vs.h"
-#include "brw_state.h"
 
+#include <X11/Xlib.h>
 
-static void prepare_active_vertprog( struct brw_context *brw )
-{
-   const struct gl_vertex_program *prev = brw->vertex_program;
+struct _glxapi_table;
 
-   brw->vertex_program = brw->attribs.VertexProgram->_Current;
+extern struct _glxapi_table *_mesa_GetGLXDispatchTable(void);
 
-   if (brw->vertex_program != prev) 
-      brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM;
-}
+extern void Fake_glXUseXFont( Font font, int first, int count, int listbase );
 
 
+#endif
 
-const struct brw_tracked_state brw_active_vertprog = {
-   .dirty = {
-      .mesa = _NEW_PROGRAM,
-      .brw = 0,
-      .cache = 0
-   },
-   .prepare = prepare_active_vertprog
-};
diff --git a/src/gallium/state_trackers/glx/xlib/fakeglx_fonts.c b/src/gallium/state_trackers/glx/xlib/fakeglx_fonts.c
new file mode 100644
index 0000000000..e359046756
--- /dev/null
+++ b/src/gallium/state_trackers/glx/xlib/fakeglx_fonts.c
@@ -0,0 +1,373 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2000  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/* xfonts.c -- glXUseXFont() for Mesa written by
+ * Copyright (C) 1995 Thorsten.Ohl @ Physik.TH-Darmstadt.de
+ */
+
+#include "context.h"
+#include "imports.h"
+#include "fakeglx.h"
+#include <GL/glx.h>
+
+
+/* Some debugging info.  */
+
+#ifdef DEBUG
+#undef _R
+#undef _G
+#undef _B
+#include <ctype.h>
+
+int debug_xfonts = 0;
+
+static void
+dump_char_struct(XCharStruct * ch, char *prefix)
+{
+   printf("%slbearing = %d, rbearing = %d, width = %d\n",
+	  prefix, ch->lbearing, ch->rbearing, ch->width);
+   printf("%sascent = %d, descent = %d, attributes = %u\n",
+	  prefix, ch->ascent, ch->descent, (unsigned int) ch->attributes);
+}
+
+static void
+dump_font_struct(XFontStruct * font)
+{
+   printf("ascent = %d, descent = %d\n", font->ascent, font->descent);
+   printf("char_or_byte2 = (%u,%u)\n",
+	  font->min_char_or_byte2, font->max_char_or_byte2);
+   printf("byte1 = (%u,%u)\n", font->min_byte1, font->max_byte1);
+   printf("all_chars_exist = %s\n", font->all_chars_exist ? "True" : "False");
+   printf("default_char = %c (\\%03o)\n",
+	  (char) (isprint(font->default_char) ? font->default_char : ' '),
+	  font->default_char);
+   dump_char_struct(&font->min_bounds, "min> ");
+   dump_char_struct(&font->max_bounds, "max> ");
+#if 0
+   for (c = font->min_char_or_byte2; c <= font->max_char_or_byte2; c++) {
+      char prefix[8];
+      sprintf(prefix, "%d> ", c);
+      dump_char_struct(&font->per_char[c], prefix);
+   }
+#endif
+}
+
+static void
+dump_bitmap(unsigned int width, unsigned int height, GLubyte * bitmap)
+{
+   unsigned int x, y;
+
+   printf("    ");
+   for (x = 0; x < 8 * width; x++)
+      printf("%o", 7 - (x % 8));
+   putchar('\n');
+   for (y = 0; y < height; y++) {
+      printf("%3o:", y);
+      for (x = 0; x < 8 * width; x++)
+	 putchar((bitmap[width * (height - y - 1) + x / 8] & (1 << (7 - (x %
+									 8))))
+		 ? '*' : '.');
+      printf("   ");
+      for (x = 0; x < width; x++)
+	 printf("0x%02x, ", bitmap[width * (height - y - 1) + x]);
+      putchar('\n');
+   }
+}
+#endif /* DEBUG */
+
+
+/* Implementation.  */
+
+/* Fill a BITMAP with a character C from thew current font
+   in the graphics context GC.  WIDTH is the width in bytes
+   and HEIGHT is the height in bits.
+
+   Note that the generated bitmaps must be used with
+
+        glPixelStorei (GL_UNPACK_SWAP_BYTES, GL_FALSE);
+        glPixelStorei (GL_UNPACK_LSB_FIRST, GL_FALSE);
+        glPixelStorei (GL_UNPACK_ROW_LENGTH, 0);
+        glPixelStorei (GL_UNPACK_SKIP_ROWS, 0);
+        glPixelStorei (GL_UNPACK_SKIP_PIXELS, 0);
+        glPixelStorei (GL_UNPACK_ALIGNMENT, 1);
+
+   Possible optimizations:
+
+     * use only one reusable pixmap with the maximum dimensions.
+     * draw the entire font into a single pixmap (careful with
+       proportional fonts!).
+*/
+
+
+/*
+ * Generate OpenGL-compatible bitmap.
+ */
+static void
+fill_bitmap(Display * dpy, Window win, GC gc,
+	    unsigned int width, unsigned int height,
+	    int x0, int y0, unsigned int c, GLubyte * bitmap)
+{
+   XImage *image;
+   unsigned int x, y;
+   Pixmap pixmap;
+   XChar2b char2b;
+
+   pixmap = XCreatePixmap(dpy, win, 8 * width, height, 1);
+   XSetForeground(dpy, gc, 0);
+   XFillRectangle(dpy, pixmap, gc, 0, 0, 8 * width, height);
+   XSetForeground(dpy, gc, 1);
+
+   char2b.byte1 = (c >> 8) & 0xff;
+   char2b.byte2 = (c & 0xff);
+
+   XDrawString16(dpy, pixmap, gc, x0, y0, &char2b, 1);
+
+   image = XGetImage(dpy, pixmap, 0, 0, 8 * width, height, 1, XYPixmap);
+   if (image) {
+      /* Fill the bitmap (X11 and OpenGL are upside down wrt each other).  */
+      for (y = 0; y < height; y++)
+	 for (x = 0; x < 8 * width; x++)
+	    if (XGetPixel(image, x, y))
+	       bitmap[width * (height - y - 1) + x / 8] |=
+		  (1 << (7 - (x % 8)));
+      XDestroyImage(image);
+   }
+
+   XFreePixmap(dpy, pixmap);
+}
+
+/*
+ * determine if a given glyph is valid and return the
+ * corresponding XCharStruct.
+ */
+static XCharStruct *
+isvalid(XFontStruct * fs, unsigned int which)
+{
+   unsigned int rows, pages;
+   unsigned int byte1 = 0, byte2 = 0;
+   int i, valid = 1;
+
+   rows = fs->max_byte1 - fs->min_byte1 + 1;
+   pages = fs->max_char_or_byte2 - fs->min_char_or_byte2 + 1;
+
+   if (rows == 1) {
+      /* "linear" fonts */
+      if ((fs->min_char_or_byte2 > which) || (fs->max_char_or_byte2 < which))
+	 valid = 0;
+   }
+   else {
+      /* "matrix" fonts */
+      byte2 = which & 0xff;
+      byte1 = which >> 8;
+      if ((fs->min_char_or_byte2 > byte2) ||
+	  (fs->max_char_or_byte2 < byte2) ||
+	  (fs->min_byte1 > byte1) || (fs->max_byte1 < byte1))
+	 valid = 0;
+   }
+
+   if (valid) {
+      if (fs->per_char) {
+	 if (rows == 1) {
+	    /* "linear" fonts */
+	    return (fs->per_char + (which - fs->min_char_or_byte2));
+	 }
+	 else {
+	    /* "matrix" fonts */
+	    i = ((byte1 - fs->min_byte1) * pages) +
+	       (byte2 - fs->min_char_or_byte2);
+	    return (fs->per_char + i);
+	 }
+      }
+      else {
+	 return (&fs->min_bounds);
+      }
+   }
+   return (NULL);
+}
+
+
+void
+Fake_glXUseXFont(Font font, int first, int count, int listbase)
+{
+   Display *dpy;
+   Window win;
+   Pixmap pixmap;
+   GC gc;
+   XGCValues values;
+   unsigned long valuemask;
+   XFontStruct *fs;
+   GLint swapbytes, lsbfirst, rowlength;
+   GLint skiprows, skippixels, alignment;
+   unsigned int max_width, max_height, max_bm_width, max_bm_height;
+   GLubyte *bm;
+   int i;
+
+   dpy = glXGetCurrentDisplay();
+   if (!dpy)
+      return;			/* I guess glXMakeCurrent wasn't called */
+   win = RootWindow(dpy, DefaultScreen(dpy));
+
+   fs = XQueryFont(dpy, font);
+   if (!fs) {
+      _mesa_error(NULL, GL_INVALID_VALUE,
+		  "Couldn't get font structure information");
+      return;
+   }
+
+   /* Allocate a bitmap that can fit all characters.  */
+   max_width = fs->max_bounds.rbearing - fs->min_bounds.lbearing;
+   max_height = fs->max_bounds.ascent + fs->max_bounds.descent;
+   max_bm_width = (max_width + 7) / 8;
+   max_bm_height = max_height;
+
+   bm = (GLubyte *) MALLOC((max_bm_width * max_bm_height) * sizeof(GLubyte));
+   if (!bm) {
+      XFreeFontInfo(NULL, fs, 1);
+      _mesa_error(NULL, GL_OUT_OF_MEMORY,
+		  "Couldn't allocate bitmap in glXUseXFont()");
+      return;
+   }
+
+#if 0
+   /* get the page info */
+   pages = fs->max_char_or_byte2 - fs->min_char_or_byte2 + 1;
+   firstchar = (fs->min_byte1 << 8) + fs->min_char_or_byte2;
+   lastchar = (fs->max_byte1 << 8) + fs->max_char_or_byte2;
+   rows = fs->max_byte1 - fs->min_byte1 + 1;
+   unsigned int first_char, last_char, pages, rows;
+#endif
+
+   /* Save the current packing mode for bitmaps.  */
+   glGetIntegerv(GL_UNPACK_SWAP_BYTES, &swapbytes);
+   glGetIntegerv(GL_UNPACK_LSB_FIRST, &lsbfirst);
+   glGetIntegerv(GL_UNPACK_ROW_LENGTH, &rowlength);
+   glGetIntegerv(GL_UNPACK_SKIP_ROWS, &skiprows);
+   glGetIntegerv(GL_UNPACK_SKIP_PIXELS, &skippixels);
+   glGetIntegerv(GL_UNPACK_ALIGNMENT, &alignment);
+
+   /* Enforce a standard packing mode which is compatible with
+      fill_bitmap() from above.  This is actually the default mode,
+      except for the (non)alignment.  */
+   glPixelStorei(GL_UNPACK_SWAP_BYTES, GL_FALSE);
+   glPixelStorei(GL_UNPACK_LSB_FIRST, GL_FALSE);
+   glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
+   glPixelStorei(GL_UNPACK_SKIP_ROWS, 0);
+   glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0);
+   glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
+
+   pixmap = XCreatePixmap(dpy, win, 10, 10, 1);
+   values.foreground = BlackPixel(dpy, DefaultScreen(dpy));
+   values.background = WhitePixel(dpy, DefaultScreen(dpy));
+   values.font = fs->fid;
+   valuemask = GCForeground | GCBackground | GCFont;
+   gc = XCreateGC(dpy, pixmap, valuemask, &values);
+   XFreePixmap(dpy, pixmap);
+
+#ifdef DEBUG
+   if (debug_xfonts)
+      dump_font_struct(fs);
+#endif
+
+   for (i = 0; i < count; i++) {
+      unsigned int width, height, bm_width, bm_height;
+      GLfloat x0, y0, dx, dy;
+      XCharStruct *ch;
+      int x, y;
+      unsigned int c = first + i;
+      int list = listbase + i;
+      int valid;
+
+      /* check on index validity and get the bounds */
+      ch = isvalid(fs, c);
+      if (!ch) {
+	 ch = &fs->max_bounds;
+	 valid = 0;
+      }
+      else {
+	 valid = 1;
+      }
+
+#ifdef DEBUG
+      if (debug_xfonts) {
+	 char s[7];
+	 sprintf(s, isprint(c) ? "%c> " : "\\%03o> ", c);
+	 dump_char_struct(ch, s);
+      }
+#endif
+
+      /* glBitmap()' parameters:
+         straight from the glXUseXFont(3) manpage.  */
+      width = ch->rbearing - ch->lbearing;
+      height = ch->ascent + ch->descent;
+      x0 = -ch->lbearing;
+      y0 = ch->descent - 0;	/* XXX used to subtract 1 here */
+      /* but that caused a conformace failure */
+      dx = ch->width;
+      dy = 0;
+
+      /* X11's starting point.  */
+      x = -ch->lbearing;
+      y = ch->ascent;
+
+      /* Round the width to a multiple of eight.  We will use this also
+         for the pixmap for capturing the X11 font.  This is slightly
+         inefficient, but it makes the OpenGL part real easy.  */
+      bm_width = (width + 7) / 8;
+      bm_height = height;
+
+      glNewList(list, GL_COMPILE);
+      if (valid && (bm_width > 0) && (bm_height > 0)) {
+
+	 MEMSET(bm, '\0', bm_width * bm_height);
+	 fill_bitmap(dpy, win, gc, bm_width, bm_height, x, y, c, bm);
+
+	 glBitmap(width, height, x0, y0, dx, dy, bm);
+#ifdef DEBUG
+	 if (debug_xfonts) {
+	    printf("width/height = %u/%u\n", width, height);
+	    printf("bm_width/bm_height = %u/%u\n", bm_width, bm_height);
+	    dump_bitmap(bm_width, bm_height, bm);
+	 }
+#endif
+      }
+      else {
+	 glBitmap(0, 0, 0.0, 0.0, dx, dy, NULL);
+      }
+      glEndList();
+   }
+
+   FREE(bm);
+   XFreeFontInfo(NULL, fs, 1);
+   XFreeGC(dpy, gc);
+
+   /* Restore saved packing modes.  */
+   glPixelStorei(GL_UNPACK_SWAP_BYTES, swapbytes);
+   glPixelStorei(GL_UNPACK_LSB_FIRST, lsbfirst);
+   glPixelStorei(GL_UNPACK_ROW_LENGTH, rowlength);
+   glPixelStorei(GL_UNPACK_SKIP_ROWS, skiprows);
+   glPixelStorei(GL_UNPACK_SKIP_PIXELS, skippixels);
+   glPixelStorei(GL_UNPACK_ALIGNMENT, alignment);
+}
diff --git a/src/gallium/state_trackers/glx/xlib/glxapi.c b/src/gallium/state_trackers/glx/xlib/glxapi.c
new file mode 100644
index 0000000000..1ff04804f1
--- /dev/null
+++ b/src/gallium/state_trackers/glx/xlib/glxapi.c
@@ -0,0 +1,1246 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ * 
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ * This is the GLX API dispatcher.  Calls to the glX* functions are
+ * either routed to the real GLX encoders or to Mesa's pseudo-GLX functions.
+ * See the glxapi.h file for more details.
+ */
+
+
+#include <assert.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include "main/glheader.h"
+#include "glapi/glapi.h"
+#include "glxapi.h"
+#include "fakeglx.h"
+#include "pipe/p_thread.h"
+
+
+struct display_dispatch {
+   Display *Dpy;
+   struct _glxapi_table *Table;
+   struct display_dispatch *Next;
+};
+
+static struct display_dispatch *DispatchList = NULL;
+
+
+/* Display -> Dispatch caching */
+static Display *prevDisplay = NULL;
+static struct _glxapi_table *prevTable = NULL;
+
+
+static struct _glxapi_table *
+get_dispatch(Display *dpy)
+{
+   if (!dpy)
+      return NULL;
+
+   /* search list of display/dispatch pairs for this display */
+   {
+      const struct display_dispatch *d = DispatchList;
+      while (d) {
+         if (d->Dpy == dpy) {
+            prevDisplay = dpy;
+            prevTable = d->Table;
+            return d->Table;  /* done! */
+         }
+         d = d->Next;
+      }
+   }
+
+   /* A new display, determine if we should use real GLX
+    * or Mesa's pseudo-GLX.
+    */
+   {
+      struct _glxapi_table *t = _mesa_GetGLXDispatchTable();
+
+      if (t) {
+         struct display_dispatch *d;
+         d = (struct display_dispatch *) malloc(sizeof(struct display_dispatch));
+         if (d) {
+            d->Dpy = dpy;
+            d->Table = t;
+            /* insert at head of list */
+            d->Next = DispatchList;
+            DispatchList = d;
+            /* update cache */
+            prevDisplay = dpy;
+            prevTable = t;
+            return t;
+         }
+      }
+   }
+
+   /* If we get here that means we can't use real GLX on this display
+    * and the Mesa pseudo-GLX software renderer wasn't compiled in.
+    * Or, we ran out of memory!
+    */
+   return NULL;
+}
+
+
+/* Don't use the GET_DISPATCH defined in glthread.h */
+#undef GET_DISPATCH
+
+#define GET_DISPATCH(DPY, TABLE)	\
+   if (DPY == prevDisplay) {		\
+      TABLE = prevTable;		\
+   }					\
+   else if (!DPY) {			\
+      TABLE = NULL;			\
+   }					\
+   else {				\
+      TABLE = get_dispatch(DPY);	\
+   }
+
+   
+
+
+/**
+ * GLX API current context.
+ */
+pipe_tsd ContextTSD;
+
+
+static void
+SetCurrentContext(GLXContext c)
+{
+   pipe_tsd_set(&ContextTSD, c);
+}
+
+
+/*
+ * GLX API entrypoints
+ */
+
+/*** GLX_VERSION_1_0 ***/
+
+XVisualInfo PUBLIC *
+glXChooseVisual(Display *dpy, int screen, int *list)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return NULL;
+   return (t->ChooseVisual)(dpy, screen, list);
+}
+
+
+void PUBLIC
+glXCopyContext(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->CopyContext)(dpy, src, dst, mask);
+}
+
+
+GLXContext PUBLIC
+glXCreateContext(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreateContext)(dpy, visinfo, shareList, direct);
+}
+
+
+GLXPixmap PUBLIC
+glXCreateGLXPixmap(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreateGLXPixmap)(dpy, visinfo, pixmap);
+}
+
+
+void PUBLIC
+glXDestroyContext(Display *dpy, GLXContext ctx)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   if (glXGetCurrentContext() == ctx)
+      SetCurrentContext(NULL);
+   (t->DestroyContext)(dpy, ctx);
+}
+
+
+void PUBLIC
+glXDestroyGLXPixmap(Display *dpy, GLXPixmap pixmap)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->DestroyGLXPixmap)(dpy, pixmap);
+}
+
+
+int PUBLIC
+glXGetConfig(Display *dpy, XVisualInfo *visinfo, int attrib, int *value)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return GLX_NO_EXTENSION;
+   return (t->GetConfig)(dpy, visinfo, attrib, value);
+}
+
+
+GLXContext PUBLIC
+glXGetCurrentContext(void)
+{
+   return (GLXContext) pipe_tsd_get(&ContextTSD);
+}
+
+
+GLXDrawable PUBLIC
+glXGetCurrentDrawable(void)
+{
+   __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext();
+   return gc ? gc->currentDrawable : 0;
+}
+
+
+Bool PUBLIC
+glXIsDirect(Display *dpy, GLXContext ctx)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return False;
+   return (t->IsDirect)(dpy, ctx);
+}
+
+
+Bool PUBLIC
+glXMakeCurrent(Display *dpy, GLXDrawable drawable, GLXContext ctx)
+{
+   Bool b;
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t) {
+      return False;
+   }
+   b = (*t->MakeCurrent)(dpy, drawable, ctx);
+   if (b) {
+      SetCurrentContext(ctx);
+   }
+   return b;
+}
+
+
+Bool PUBLIC
+glXQueryExtension(Display *dpy, int *errorb, int *event)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return False;
+   return (t->QueryExtension)(dpy, errorb, event);
+}
+
+
+Bool PUBLIC
+glXQueryVersion(Display *dpy, int *maj, int *min)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return False;
+   return (t->QueryVersion)(dpy, maj, min);
+}
+
+
+void PUBLIC
+glXSwapBuffers(Display *dpy, GLXDrawable drawable)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->SwapBuffers)(dpy, drawable);
+}
+
+
+void PUBLIC
+glXUseXFont(Font font, int first, int count, int listBase)
+{
+   struct _glxapi_table *t;
+   Display *dpy = glXGetCurrentDisplay();
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->UseXFont)(font, first, count, listBase);
+}
+
+
+void PUBLIC
+glXWaitGL(void)
+{
+   struct _glxapi_table *t;
+   Display *dpy = glXGetCurrentDisplay();
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->WaitGL)();
+}
+
+
+void PUBLIC
+glXWaitX(void)
+{
+   struct _glxapi_table *t;
+   Display *dpy = glXGetCurrentDisplay();
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->WaitX)();
+}
+
+
+
+/*** GLX_VERSION_1_1 ***/
+
+const char PUBLIC *
+glXGetClientString(Display *dpy, int name)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return NULL;
+   return (t->GetClientString)(dpy, name);
+}
+
+
+const char PUBLIC *
+glXQueryExtensionsString(Display *dpy, int screen)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return NULL;
+   return (t->QueryExtensionsString)(dpy, screen);
+}
+
+
+const char PUBLIC *
+glXQueryServerString(Display *dpy, int screen, int name)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return NULL;
+   return (t->QueryServerString)(dpy, screen, name);
+}
+
+
+/*** GLX_VERSION_1_2 ***/
+
+Display PUBLIC *
+glXGetCurrentDisplay(void)
+{
+   /* Same code as in libGL's glxext.c */
+   __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext();
+   if (NULL == gc) return NULL;
+   return gc->currentDpy;
+}
+
+
+
+/*** GLX_VERSION_1_3 ***/
+
+GLXFBConfig PUBLIC *
+glXChooseFBConfig(Display *dpy, int screen, const int *attribList, int *nitems)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->ChooseFBConfig)(dpy, screen, attribList, nitems);
+}
+
+
+GLXContext PUBLIC
+glXCreateNewContext(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreateNewContext)(dpy, config, renderType, shareList, direct);
+}
+
+
+GLXPbuffer PUBLIC
+glXCreatePbuffer(Display *dpy, GLXFBConfig config, const int *attribList)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreatePbuffer)(dpy, config, attribList);
+}
+
+
+GLXPixmap PUBLIC
+glXCreatePixmap(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreatePixmap)(dpy, config, pixmap, attribList);
+}
+
+
+GLXWindow PUBLIC
+glXCreateWindow(Display *dpy, GLXFBConfig config, Window win, const int *attribList)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreateWindow)(dpy, config, win, attribList);
+}
+
+
+void PUBLIC
+glXDestroyPbuffer(Display *dpy, GLXPbuffer pbuf)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->DestroyPbuffer)(dpy, pbuf);
+}
+
+
+void PUBLIC
+glXDestroyPixmap(Display *dpy, GLXPixmap pixmap)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->DestroyPixmap)(dpy, pixmap);
+}
+
+
+void PUBLIC
+glXDestroyWindow(Display *dpy, GLXWindow window)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->DestroyWindow)(dpy, window);
+}
+
+
+GLXDrawable PUBLIC
+glXGetCurrentReadDrawable(void)
+{
+   __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext();
+   return gc ? gc->currentReadable : 0;
+}
+
+
+int PUBLIC
+glXGetFBConfigAttrib(Display *dpy, GLXFBConfig config, int attribute, int *value)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return GLX_NO_EXTENSION;
+   return (t->GetFBConfigAttrib)(dpy, config, attribute, value);
+}
+
+
+GLXFBConfig PUBLIC *
+glXGetFBConfigs(Display *dpy, int screen, int *nelements)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->GetFBConfigs)(dpy, screen, nelements);
+}
+
+void PUBLIC
+glXGetSelectedEvent(Display *dpy, GLXDrawable drawable, unsigned long *mask)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->GetSelectedEvent)(dpy, drawable, mask);
+}
+
+
+XVisualInfo PUBLIC *
+glXGetVisualFromFBConfig(Display *dpy, GLXFBConfig config)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return NULL;
+   return (t->GetVisualFromFBConfig)(dpy, config);
+}
+
+
+Bool PUBLIC
+glXMakeContextCurrent(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx)
+{
+   Bool b;
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return False;
+   b = (t->MakeContextCurrent)(dpy, draw, read, ctx);
+   if (b) {
+      SetCurrentContext(ctx);
+   }
+   return b;
+}
+
+
+int PUBLIC
+glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   assert(t);
+   if (!t)
+      return 0; /* XXX correct? */
+   return (t->QueryContext)(dpy, ctx, attribute, value);
+}
+
+
+void PUBLIC
+glXQueryDrawable(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->QueryDrawable)(dpy, draw, attribute, value);
+}
+
+
+void PUBLIC
+glXSelectEvent(Display *dpy, GLXDrawable drawable, unsigned long mask)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->SelectEvent)(dpy, drawable, mask);
+}
+
+
+
+/*** GLX_SGI_swap_control ***/
+
+int PUBLIC
+glXSwapIntervalSGI(int interval)
+{
+   struct _glxapi_table *t;
+   Display *dpy = glXGetCurrentDisplay();
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->SwapIntervalSGI)(interval);
+}
+
+
+
+/*** GLX_SGI_video_sync ***/
+
+int PUBLIC
+glXGetVideoSyncSGI(unsigned int *count)
+{
+   struct _glxapi_table *t;
+   Display *dpy = glXGetCurrentDisplay();
+   GET_DISPATCH(dpy, t);
+   if (!t || !glXGetCurrentContext())
+      return GLX_BAD_CONTEXT;
+   return (t->GetVideoSyncSGI)(count);
+}
+
+int PUBLIC
+glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
+{
+   struct _glxapi_table *t;
+   Display *dpy = glXGetCurrentDisplay();
+   GET_DISPATCH(dpy, t);
+   if (!t || !glXGetCurrentContext())
+      return GLX_BAD_CONTEXT;
+   return (t->WaitVideoSyncSGI)(divisor, remainder, count);
+}
+
+
+
+/*** GLX_SGI_make_current_read ***/
+
+Bool PUBLIC
+glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return False;
+   return (t->MakeCurrentReadSGI)(dpy, draw, read, ctx);
+}
+
+GLXDrawable PUBLIC
+glXGetCurrentReadDrawableSGI(void)
+{
+   return glXGetCurrentReadDrawable();
+}
+
+
+#if defined(_VL_H)
+
+GLXVideoSourceSGIX PUBLIC
+glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreateGLXVideoSourceSGIX)(dpy, screen, server, path, nodeClass, drainNode);
+}
+
+void PUBLIC
+glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->DestroyGLXVideoSourceSGIX)(dpy, src);
+}
+
+#endif
+
+
+/*** GLX_EXT_import_context ***/
+
+void PUBLIC
+glXFreeContextEXT(Display *dpy, GLXContext context)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->FreeContextEXT)(dpy, context);
+}
+
+GLXContextID PUBLIC
+glXGetContextIDEXT(const GLXContext context)
+{
+   return ((__GLXcontext *) context)->xid;
+}
+
+Display PUBLIC *
+glXGetCurrentDisplayEXT(void)
+{
+   return glXGetCurrentDisplay();
+}
+
+GLXContext PUBLIC
+glXImportContextEXT(Display *dpy, GLXContextID contextID)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->ImportContextEXT)(dpy, contextID);
+}
+
+int PUBLIC
+glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute,int *value)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;  /* XXX ok? */
+   return (t->QueryContextInfoEXT)(dpy, context, attribute, value);
+}
+
+
+
+/*** GLX_SGIX_fbconfig ***/
+
+int PUBLIC
+glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->GetFBConfigAttribSGIX)(dpy, config, attribute, value);
+}
+
+GLXFBConfigSGIX PUBLIC *
+glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->ChooseFBConfigSGIX)(dpy, screen, attrib_list, nelements);
+}
+
+GLXPixmap PUBLIC
+glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreateGLXPixmapWithConfigSGIX)(dpy, config, pixmap);
+}
+
+GLXContext PUBLIC
+glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreateContextWithConfigSGIX)(dpy, config, render_type, share_list, direct);
+}
+
+XVisualInfo PUBLIC *
+glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->GetVisualFromFBConfigSGIX)(dpy, config);
+}
+
+GLXFBConfigSGIX PUBLIC
+glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->GetFBConfigFromVisualSGIX)(dpy, vis);
+}
+
+
+
+/*** GLX_SGIX_pbuffer ***/
+
+GLXPbufferSGIX PUBLIC
+glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, unsigned int width, unsigned int height, int *attrib_list)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreateGLXPbufferSGIX)(dpy, config, width, height, attrib_list);
+}
+
+void PUBLIC
+glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->DestroyGLXPbufferSGIX)(dpy, pbuf);
+}
+
+int PUBLIC
+glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->QueryGLXPbufferSGIX)(dpy, pbuf, attribute, value);
+}
+
+void PUBLIC
+glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->SelectEventSGIX)(dpy, drawable, mask);
+}
+
+void PUBLIC
+glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->GetSelectedEventSGIX)(dpy, drawable, mask);
+}
+
+
+
+/*** GLX_SGI_cushion ***/
+
+void PUBLIC
+glXCushionSGI(Display *dpy, Window win, float cushion)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->CushionSGI)(dpy, win, cushion);
+}
+
+
+
+/*** GLX_SGIX_video_resize ***/
+
+int PUBLIC
+glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->BindChannelToWindowSGIX)(dpy, screen, channel, window);
+}
+
+int PUBLIC
+glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->ChannelRectSGIX)(dpy, screen, channel, x, y, w, h);
+}
+
+int PUBLIC
+glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->QueryChannelRectSGIX)(dpy, screen, channel, x, y, w, h);
+}
+
+int PUBLIC
+glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->QueryChannelDeltasSGIX)(dpy, screen, channel, dx, dy, dw, dh);
+}
+
+int PUBLIC
+glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->ChannelRectSyncSGIX)(dpy, screen, channel, synctype);
+}
+
+
+
+#if defined(_DM_BUFFER_H_)
+
+Bool PUBLIC
+glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return False;
+   return (t->AssociateDMPbufferSGIX)(dpy, pbuffer, params, dmbuffer);
+}
+
+#endif
+
+
+/*** GLX_SGIX_swap_group ***/
+
+void PUBLIC
+glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (*t->JoinSwapGroupSGIX)(dpy, drawable, member);
+}
+
+
+/*** GLX_SGIX_swap_barrier ***/
+
+void PUBLIC
+glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (*t->BindSwapBarrierSGIX)(dpy, drawable, barrier);
+}
+
+Bool PUBLIC
+glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return False;
+   return (*t->QueryMaxSwapBarriersSGIX)(dpy, screen, max);
+}
+
+
+
+/*** GLX_SUN_get_transparent_index ***/
+
+Status PUBLIC
+glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return False;
+   return (*t->GetTransparentIndexSUN)(dpy, overlay, underlay, pTransparent);
+}
+
+
+
+/*** GLX_MESA_copy_sub_buffer ***/
+
+void PUBLIC
+glXCopySubBufferMESA(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return;
+   (t->CopySubBufferMESA)(dpy, drawable, x, y, width, height);
+}
+
+
+
+/*** GLX_MESA_release_buffers ***/
+
+Bool PUBLIC
+glXReleaseBuffersMESA(Display *dpy, Window w)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return False;
+   return (t->ReleaseBuffersMESA)(dpy, w);
+}
+
+
+
+/*** GLX_MESA_pixmap_colormap ***/
+
+GLXPixmap PUBLIC
+glXCreateGLXPixmapMESA(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (!t)
+      return 0;
+   return (t->CreateGLXPixmapMESA)(dpy, visinfo, pixmap, cmap);
+}
+
+/*** GLX_EXT_texture_from_pixmap */
+
+void
+glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer,
+                   const int *attrib_list)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (t)
+      t->BindTexImageEXT(dpy, drawable, buffer, attrib_list);
+}
+
+void
+glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer)
+{
+   struct _glxapi_table *t;
+   GET_DISPATCH(dpy, t);
+   if (t)
+      t->ReleaseTexImageEXT(dpy, drawable, buffer);
+}
+
+
+/**********************************************************************/
+/* GLX API management functions                                       */
+/**********************************************************************/
+
+
+const char *
+_glxapi_get_version(void)
+{
+   return "1.3";
+}
+
+
+
+/*
+ * Return size of the GLX dispatch table, in entries, not bytes.
+ */
+GLuint
+_glxapi_get_dispatch_table_size(void)
+{
+   return sizeof(struct _glxapi_table) / sizeof(void *);
+}
+
+
+static int
+generic_no_op_func(void)
+{
+   return 0;
+}
+
+
+/*
+ * Initialize all functions in given dispatch table to be no-ops
+ */
+void
+_glxapi_set_no_op_table(struct _glxapi_table *t)
+{
+   typedef int (*nop_func)(void);
+   nop_func *dispatch = (nop_func *) t;
+   GLuint n = _glxapi_get_dispatch_table_size();
+   GLuint i;
+   for (i = 0; i < n; i++) {
+      dispatch[i] = generic_no_op_func;
+   }
+}
+
+
+struct name_address_pair {
+   const char *Name;
+   __GLXextFuncPtr Address;
+};
+
+static struct name_address_pair GLX_functions[] = {
+   /*** GLX_VERSION_1_0 ***/
+   { "glXChooseVisual", (__GLXextFuncPtr) glXChooseVisual },
+   { "glXCopyContext", (__GLXextFuncPtr) glXCopyContext },
+   { "glXCreateContext", (__GLXextFuncPtr) glXCreateContext },
+   { "glXCreateGLXPixmap", (__GLXextFuncPtr) glXCreateGLXPixmap },
+   { "glXDestroyContext", (__GLXextFuncPtr) glXDestroyContext },
+   { "glXDestroyGLXPixmap", (__GLXextFuncPtr) glXDestroyGLXPixmap },
+   { "glXGetConfig", (__GLXextFuncPtr) glXGetConfig },
+   { "glXGetCurrentContext", (__GLXextFuncPtr) glXGetCurrentContext },
+   { "glXGetCurrentDrawable", (__GLXextFuncPtr) glXGetCurrentDrawable },
+   { "glXIsDirect", (__GLXextFuncPtr) glXIsDirect },
+   { "glXMakeCurrent", (__GLXextFuncPtr) glXMakeCurrent },
+   { "glXQueryExtension", (__GLXextFuncPtr) glXQueryExtension },
+   { "glXQueryVersion", (__GLXextFuncPtr) glXQueryVersion },
+   { "glXSwapBuffers", (__GLXextFuncPtr) glXSwapBuffers },
+   { "glXUseXFont", (__GLXextFuncPtr) glXUseXFont },
+   { "glXWaitGL", (__GLXextFuncPtr) glXWaitGL },
+   { "glXWaitX", (__GLXextFuncPtr) glXWaitX },
+
+   /*** GLX_VERSION_1_1 ***/
+   { "glXGetClientString", (__GLXextFuncPtr) glXGetClientString },
+   { "glXQueryExtensionsString", (__GLXextFuncPtr) glXQueryExtensionsString },
+   { "glXQueryServerString", (__GLXextFuncPtr) glXQueryServerString },
+
+   /*** GLX_VERSION_1_2 ***/
+   { "glXGetCurrentDisplay", (__GLXextFuncPtr) glXGetCurrentDisplay },
+
+   /*** GLX_VERSION_1_3 ***/
+   { "glXChooseFBConfig", (__GLXextFuncPtr) glXChooseFBConfig },
+   { "glXCreateNewContext", (__GLXextFuncPtr) glXCreateNewContext },
+   { "glXCreatePbuffer", (__GLXextFuncPtr) glXCreatePbuffer },
+   { "glXCreatePixmap", (__GLXextFuncPtr) glXCreatePixmap },
+   { "glXCreateWindow", (__GLXextFuncPtr) glXCreateWindow },
+   { "glXDestroyPbuffer", (__GLXextFuncPtr) glXDestroyPbuffer },
+   { "glXDestroyPixmap", (__GLXextFuncPtr) glXDestroyPixmap },
+   { "glXDestroyWindow", (__GLXextFuncPtr) glXDestroyWindow },
+   { "glXGetCurrentReadDrawable", (__GLXextFuncPtr) glXGetCurrentReadDrawable },
+   { "glXGetFBConfigAttrib", (__GLXextFuncPtr) glXGetFBConfigAttrib },
+   { "glXGetFBConfigs", (__GLXextFuncPtr) glXGetFBConfigs },
+   { "glXGetSelectedEvent", (__GLXextFuncPtr) glXGetSelectedEvent },
+   { "glXGetVisualFromFBConfig", (__GLXextFuncPtr) glXGetVisualFromFBConfig },
+   { "glXMakeContextCurrent", (__GLXextFuncPtr) glXMakeContextCurrent },
+   { "glXQueryContext", (__GLXextFuncPtr) glXQueryContext },
+   { "glXQueryDrawable", (__GLXextFuncPtr) glXQueryDrawable },
+   { "glXSelectEvent", (__GLXextFuncPtr) glXSelectEvent },
+
+   /*** GLX_VERSION_1_4 ***/
+   { "glXGetProcAddress", (__GLXextFuncPtr) glXGetProcAddress },
+
+   /*** GLX_SGI_swap_control ***/
+   { "glXSwapIntervalSGI", (__GLXextFuncPtr) glXSwapIntervalSGI },
+
+   /*** GLX_SGI_video_sync ***/
+   { "glXGetVideoSyncSGI", (__GLXextFuncPtr) glXGetVideoSyncSGI },
+   { "glXWaitVideoSyncSGI", (__GLXextFuncPtr) glXWaitVideoSyncSGI },
+
+   /*** GLX_SGI_make_current_read ***/
+   { "glXMakeCurrentReadSGI", (__GLXextFuncPtr) glXMakeCurrentReadSGI },
+   { "glXGetCurrentReadDrawableSGI", (__GLXextFuncPtr) glXGetCurrentReadDrawableSGI },
+
+   /*** GLX_SGIX_video_source ***/
+#if defined(_VL_H)
+   { "glXCreateGLXVideoSourceSGIX", (__GLXextFuncPtr) glXCreateGLXVideoSourceSGIX },
+   { "glXDestroyGLXVideoSourceSGIX", (__GLXextFuncPtr) glXDestroyGLXVideoSourceSGIX },
+#endif
+
+   /*** GLX_EXT_import_context ***/
+   { "glXFreeContextEXT", (__GLXextFuncPtr) glXFreeContextEXT },
+   { "glXGetContextIDEXT", (__GLXextFuncPtr) glXGetContextIDEXT },
+   { "glXGetCurrentDisplayEXT", (__GLXextFuncPtr) glXGetCurrentDisplayEXT },
+   { "glXImportContextEXT", (__GLXextFuncPtr) glXImportContextEXT },
+   { "glXQueryContextInfoEXT", (__GLXextFuncPtr) glXQueryContextInfoEXT },
+
+   /*** GLX_SGIX_fbconfig ***/
+   { "glXGetFBConfigAttribSGIX", (__GLXextFuncPtr) glXGetFBConfigAttribSGIX },
+   { "glXChooseFBConfigSGIX", (__GLXextFuncPtr) glXChooseFBConfigSGIX },
+   { "glXCreateGLXPixmapWithConfigSGIX", (__GLXextFuncPtr) glXCreateGLXPixmapWithConfigSGIX },
+   { "glXCreateContextWithConfigSGIX", (__GLXextFuncPtr) glXCreateContextWithConfigSGIX },
+   { "glXGetVisualFromFBConfigSGIX", (__GLXextFuncPtr) glXGetVisualFromFBConfigSGIX },
+   { "glXGetFBConfigFromVisualSGIX", (__GLXextFuncPtr) glXGetFBConfigFromVisualSGIX },
+
+   /*** GLX_SGIX_pbuffer ***/
+   { "glXCreateGLXPbufferSGIX", (__GLXextFuncPtr) glXCreateGLXPbufferSGIX },
+   { "glXDestroyGLXPbufferSGIX", (__GLXextFuncPtr) glXDestroyGLXPbufferSGIX },
+   { "glXQueryGLXPbufferSGIX", (__GLXextFuncPtr) glXQueryGLXPbufferSGIX },
+   { "glXSelectEventSGIX", (__GLXextFuncPtr) glXSelectEventSGIX },
+   { "glXGetSelectedEventSGIX", (__GLXextFuncPtr) glXGetSelectedEventSGIX },
+
+   /*** GLX_SGI_cushion ***/
+   { "glXCushionSGI", (__GLXextFuncPtr) glXCushionSGI },
+
+   /*** GLX_SGIX_video_resize ***/
+   { "glXBindChannelToWindowSGIX", (__GLXextFuncPtr) glXBindChannelToWindowSGIX },
+   { "glXChannelRectSGIX", (__GLXextFuncPtr) glXChannelRectSGIX },
+   { "glXQueryChannelRectSGIX", (__GLXextFuncPtr) glXQueryChannelRectSGIX },
+   { "glXQueryChannelDeltasSGIX", (__GLXextFuncPtr) glXQueryChannelDeltasSGIX },
+   { "glXChannelRectSyncSGIX", (__GLXextFuncPtr) glXChannelRectSyncSGIX },
+
+   /*** GLX_SGIX_dmbuffer **/
+#if defined(_DM_BUFFER_H_)
+   { "glXAssociateDMPbufferSGIX", (__GLXextFuncPtr) glXAssociateDMPbufferSGIX },
+#endif
+
+   /*** GLX_SGIX_swap_group ***/
+   { "glXJoinSwapGroupSGIX", (__GLXextFuncPtr) glXJoinSwapGroupSGIX },
+
+   /*** GLX_SGIX_swap_barrier ***/
+   { "glXBindSwapBarrierSGIX", (__GLXextFuncPtr) glXBindSwapBarrierSGIX },
+   { "glXQueryMaxSwapBarriersSGIX", (__GLXextFuncPtr) glXQueryMaxSwapBarriersSGIX },
+
+   /*** GLX_SUN_get_transparent_index ***/
+   { "glXGetTransparentIndexSUN", (__GLXextFuncPtr) glXGetTransparentIndexSUN },
+
+   /*** GLX_MESA_copy_sub_buffer ***/
+   { "glXCopySubBufferMESA", (__GLXextFuncPtr) glXCopySubBufferMESA },
+
+   /*** GLX_MESA_pixmap_colormap ***/
+   { "glXCreateGLXPixmapMESA", (__GLXextFuncPtr) glXCreateGLXPixmapMESA },
+
+   /*** GLX_MESA_release_buffers ***/
+   { "glXReleaseBuffersMESA", (__GLXextFuncPtr) glXReleaseBuffersMESA },
+
+   /*** GLX_ARB_get_proc_address ***/
+   { "glXGetProcAddressARB", (__GLXextFuncPtr) glXGetProcAddressARB },
+
+   /*** GLX_EXT_texture_from_pixmap ***/
+   { "glXBindTexImageEXT", (__GLXextFuncPtr) glXBindTexImageEXT },
+   { "glXReleaseTexImageEXT", (__GLXextFuncPtr) glXReleaseTexImageEXT },
+
+   { NULL, NULL }   /* end of list */
+};
+
+
+
+/*
+ * Return address of named glX function, or NULL if not found.
+ */
+__GLXextFuncPtr
+_glxapi_get_proc_address(const char *funcName)
+{
+   GLuint i;
+   for (i = 0; GLX_functions[i].Name; i++) {
+      if (strcmp(GLX_functions[i].Name, funcName) == 0)
+         return GLX_functions[i].Address;
+   }
+   return NULL;
+}
+
+
+
+/*
+ * This function does not get dispatched through the dispatch table
+ * since it's really a "meta" function.
+ */
+__GLXextFuncPtr
+glXGetProcAddressARB(const GLubyte *procName)
+{
+   __GLXextFuncPtr f;
+
+   f = _glxapi_get_proc_address((const char *) procName);
+   if (f) {
+      return f;
+   }
+
+   f = (__GLXextFuncPtr) _glapi_get_proc_address((const char *) procName);
+   return f;
+}
+
+
+/* GLX 1.4 */
+void (*glXGetProcAddress(const GLubyte *procName))()
+{
+   return glXGetProcAddressARB(procName);
+}
diff --git a/src/gallium/state_trackers/glx/xlib/glxapi.h b/src/gallium/state_trackers/glx/xlib/glxapi.h
new file mode 100644
index 0000000000..b4e12b4162
--- /dev/null
+++ b/src/gallium/state_trackers/glx/xlib/glxapi.h
@@ -0,0 +1,213 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.3
+ * 
+ * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef _glxapi_h_
+#define _glxapi_h_
+
+
+#define GLX_GLXEXT_PROTOTYPES
+#include "GL/glx.h"
+
+
+/* The GLX API dispatcher (i.e. this code) is being built into stand-alone
+ * Mesa.  We don't know anything about XFree86 or real GLX so we define a
+ * minimal __GLXContextRec here so some of the functions in this file can
+ * work properly.
+ */
+typedef struct __GLXcontextRec {
+   Display *currentDpy;
+   GLboolean isDirect;
+   GLXDrawable currentDrawable;
+   GLXDrawable currentReadable;
+   XID xid;
+} __GLXcontext;
+
+
+/*
+ * Almost all the GLX API functions get routed through this dispatch table.
+ * The exceptions are the glXGetCurrentXXX() functions.
+ *
+ * This dispatch table allows multiple GLX client-side modules to coexist.
+ * Specifically, a real GLX library (like SGI's or the Utah GLX) and Mesa's
+ * pseudo-GLX can be present at the same time.  The former being used on
+ * GLX-enabled X servers and the later on non-GLX X servers.
+ *
+ * Red Hat has been using this since Red Hat Linux 7.0 (I think).
+ * This'll be a standard feature in XFree86 4.3.  It basically allows one
+ * libGL to do both DRI-rendering and "fake GLX" rendering to X displays
+ * that lack the GLX extension.
+ */
+struct _glxapi_table {
+   /*** GLX_VERSION_1_0 ***/
+   XVisualInfo *(*ChooseVisual)(Display *dpy, int screen, int *list);
+   void (*CopyContext)(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask);
+   GLXContext (*CreateContext)(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct);
+   GLXPixmap (*CreateGLXPixmap)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap);
+   void (*DestroyContext)(Display *dpy, GLXContext ctx);
+   void (*DestroyGLXPixmap)(Display *dpy, GLXPixmap pixmap);
+   int (*GetConfig)(Display *dpy, XVisualInfo *visinfo, int attrib, int *value);
+   /*GLXContext (*GetCurrentContext)(void);*/
+   /*GLXDrawable (*GetCurrentDrawable)(void);*/
+   Bool (*IsDirect)(Display *dpy, GLXContext ctx);
+   Bool (*MakeCurrent)(Display *dpy, GLXDrawable drawable, GLXContext ctx);
+   Bool (*QueryExtension)(Display *dpy, int *errorb, int *event);
+   Bool (*QueryVersion)(Display *dpy, int *maj, int *min);
+   void (*SwapBuffers)(Display *dpy, GLXDrawable drawable);
+   void (*UseXFont)(Font font, int first, int count, int listBase);
+   void (*WaitGL)(void);
+   void (*WaitX)(void);
+
+   /*** GLX_VERSION_1_1 ***/
+   const char *(*GetClientString)(Display *dpy, int name);
+   const char *(*QueryExtensionsString)(Display *dpy, int screen);
+   const char *(*QueryServerString)(Display *dpy, int screen, int name);
+
+   /*** GLX_VERSION_1_2 ***/
+   /*Display *(*GetCurrentDisplay)(void);*/
+
+   /*** GLX_VERSION_1_3 ***/
+   GLXFBConfig *(*ChooseFBConfig)(Display *dpy, int screen, const int *attribList, int *nitems);
+   GLXContext (*CreateNewContext)(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct);
+   GLXPbuffer (*CreatePbuffer)(Display *dpy, GLXFBConfig config, const int *attribList);
+   GLXPixmap (*CreatePixmap)(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList);
+   GLXWindow (*CreateWindow)(Display *dpy, GLXFBConfig config, Window win, const int *attribList);
+   void (*DestroyPbuffer)(Display *dpy, GLXPbuffer pbuf);
+   void (*DestroyPixmap)(Display *dpy, GLXPixmap pixmap);
+   void (*DestroyWindow)(Display *dpy, GLXWindow window);
+   /*GLXDrawable (*GetCurrentReadDrawable)(void);*/
+   int (*GetFBConfigAttrib)(Display *dpy, GLXFBConfig config, int attribute, int *value);
+   GLXFBConfig *(*GetFBConfigs)(Display *dpy, int screen, int *nelements);
+   void (*GetSelectedEvent)(Display *dpy, GLXDrawable drawable, unsigned long *mask);
+   XVisualInfo *(*GetVisualFromFBConfig)(Display *dpy, GLXFBConfig config);
+   Bool (*MakeContextCurrent)(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx);
+   int (*QueryContext)(Display *dpy, GLXContext ctx, int attribute, int *value);
+   void (*QueryDrawable)(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value);
+   void (*SelectEvent)(Display *dpy, GLXDrawable drawable, unsigned long mask);
+
+   /*** GLX_SGI_swap_control ***/
+   int (*SwapIntervalSGI)(int);
+
+   /*** GLX_SGI_video_sync ***/
+   int (*GetVideoSyncSGI)(unsigned int *count);
+   int (*WaitVideoSyncSGI)(int divisor, int remainder, unsigned int *count);
+
+   /*** GLX_SGI_make_current_read ***/
+   Bool (*MakeCurrentReadSGI)(Display *, GLXDrawable, GLXDrawable, GLXContext);
+   /*GLXDrawable (*GetCurrentReadDrawableSGI)(void);*/
+
+   /*** GLX_SGIX_video_source (needs video library) ***/
+#if defined(_VL_H_)
+   GLXVideoSourceSGIX (*CreateGLXVideoSourceSGIX)(Display *, int, VLServer, VLPath, int, VLNode);
+   void (*DestroyGLXVideoSourceSGIX)(Display *, GLXVideoSourceSGIX);
+#else
+   void *CreateGLXVideoSourceSGIX;
+   void *DestroyGLXVideoSourceSGIX;
+#endif
+
+   /*** GLX_EXT_import_context ***/
+   void (*FreeContextEXT)(Display *dpy, GLXContext context);
+   GLXContextID (*GetContextIDEXT)(const GLXContext context);
+   /*Display *(*GetCurrentDisplayEXT)(void);*/
+   GLXContext (*ImportContextEXT)(Display *dpy, GLXContextID contextID);
+   int (*QueryContextInfoEXT)(Display *dpy, GLXContext context, int attribute,int *value);
+
+   /*** GLX_SGIX_fbconfig ***/
+   int (*GetFBConfigAttribSGIX)(Display *, GLXFBConfigSGIX, int, int *);
+   GLXFBConfigSGIX * (*ChooseFBConfigSGIX)(Display *, int, int *, int *);
+   GLXPixmap (*CreateGLXPixmapWithConfigSGIX)(Display *, GLXFBConfigSGIX, Pixmap);
+   GLXContext (*CreateContextWithConfigSGIX)(Display *, GLXFBConfigSGIX, int, GLXContext, Bool);
+   XVisualInfo * (*GetVisualFromFBConfigSGIX)(Display *, GLXFBConfigSGIX);
+   GLXFBConfigSGIX (*GetFBConfigFromVisualSGIX)(Display *, XVisualInfo *);
+
+   /*** GLX_SGIX_pbuffer ***/
+   GLXPbufferSGIX (*CreateGLXPbufferSGIX)(Display *, GLXFBConfigSGIX, unsigned int, unsigned int, int *);
+   void (*DestroyGLXPbufferSGIX)(Display *, GLXPbufferSGIX);
+   int (*QueryGLXPbufferSGIX)(Display *, GLXPbufferSGIX, int, unsigned int *);
+   void (*SelectEventSGIX)(Display *, GLXDrawable, unsigned long);
+   void (*GetSelectedEventSGIX)(Display *, GLXDrawable, unsigned long *);
+
+   /*** GLX_SGI_cushion ***/
+   void (*CushionSGI)(Display *, Window, float);
+
+   /*** GLX_SGIX_video_resize ***/
+   int (*BindChannelToWindowSGIX)(Display *, int, int, Window);
+   int (*ChannelRectSGIX)(Display *, int, int, int, int, int, int);
+   int (*QueryChannelRectSGIX)(Display *, int, int, int *, int *, int *, int *);
+   int (*QueryChannelDeltasSGIX)(Display *, int, int, int *, int *, int *, int *);
+   int (*ChannelRectSyncSGIX)(Display *, int, int, GLenum);
+
+   /*** GLX_SGIX_dmbuffer (needs dmedia library) ***/
+#if defined (_DM_BUFFER_H_)
+   Bool (*AssociateDMPbufferSGIX)(Display *, GLXPbufferSGIX, DMparams *, DMbuffer);
+#else
+   void *AssociciateDMPbufferSGIX;
+#endif
+
+   /*** GLX_SGIX_swap_group ***/
+   void (*JoinSwapGroupSGIX)(Display *, GLXDrawable, GLXDrawable);
+
+   /*** GLX_SGIX_swap_barrier ***/
+   void (*BindSwapBarrierSGIX)(Display *, GLXDrawable, int);
+   Bool (*QueryMaxSwapBarriersSGIX)(Display *, int, int *);
+
+   /*** GLX_SUN_get_transparent_index ***/
+   Status (*GetTransparentIndexSUN)(Display *, Window, Window, long *);
+
+   /*** GLX_MESA_copy_sub_buffer ***/
+   void (*CopySubBufferMESA)(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height);
+
+   /*** GLX_MESA_release_buffers ***/
+   Bool (*ReleaseBuffersMESA)(Display *dpy, Window w);
+
+   /*** GLX_MESA_pixmap_colormap ***/
+   GLXPixmap (*CreateGLXPixmapMESA)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap);
+
+   /*** GLX_EXT_texture_from_pixmap ***/
+   void (*BindTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer,
+                           const int *attrib_list);
+   void (*ReleaseTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer);
+};
+
+
+
+extern const char *
+_glxapi_get_version(void);
+
+
+
+
+extern GLuint
+_glxapi_get_dispatch_table_size(void);
+
+
+extern void
+_glxapi_set_no_op_table(struct _glxapi_table *t);
+
+
+extern __GLXextFuncPtr
+_glxapi_get_proc_address(const char *funcName);
+
+
+#endif
diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.c b/src/gallium/state_trackers/glx/xlib/xm_api.c
new file mode 100644
index 0000000000..33dc044ad5
--- /dev/null
+++ b/src/gallium/state_trackers/glx/xlib/xm_api.c
@@ -0,0 +1,1230 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ *
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file xm_api.c
+ *
+ * All the XMesa* API functions.
+ *
+ *
+ * NOTES:
+ *
+ * The window coordinate system origin (0,0) is in the lower-left corner
+ * of the window.  X11's window coordinate origin is in the upper-left
+ * corner of the window.  Therefore, most drawing functions in this
+ * file have to flip Y coordinates.
+ *
+ * Define USE_XSHM in the Makefile with -DUSE_XSHM if you want to compile
+ * in support for the MIT Shared Memory extension.  If enabled, when you
+ * use an Ximage for the back buffer in double buffered mode, the "swap"
+ * operation will be faster.  You must also link with -lXext.
+ *
+ * Byte swapping:  If the Mesa host and the X display use a different
+ * byte order then there's some trickiness to be aware of when using
+ * XImages.  The byte ordering used for the XImage is that of the X
+ * display, not the Mesa host.
+ * The color-to-pixel encoding for True/DirectColor must be done
+ * according to the display's visual red_mask, green_mask, and blue_mask.
+ * If XPutPixel is used to put a pixel into an XImage then XPutPixel will
+ * do byte swapping if needed.  If one wants to directly "poke" the pixel
+ * into the XImage's buffer then the pixel must be byte swapped first.
+ *
+ */
+
+#ifdef __CYGWIN__
+#undef WIN32
+#undef __WIN32__
+#endif
+
+#include "xm_api.h"
+#include "main/context.h"
+#include "main/framebuffer.h"
+
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_context.h"
+
+#include "xm_winsys.h"
+#include <GL/glx.h>
+
+
+/* Driver interface routines, set up by xlib backend on library
+ * _init().  These are global in the same way that function names are
+ * global.
+ */
+static struct xm_driver driver;
+
+void xmesa_set_driver( const struct xm_driver *templ )
+{
+   driver = *templ;
+}
+
+/**
+ * Global X driver lock
+ */
+pipe_mutex _xmesa_lock;
+
+
+
+/**********************************************************************/
+/*****                     X Utility Functions                    *****/
+/**********************************************************************/
+
+
+/**
+ * Return the host's byte order as LSBFirst or MSBFirst ala X.
+ */
+static int host_byte_order( void )
+{
+   int i = 1;
+   char *cptr = (char *) &i;
+   return (*cptr==1) ? LSBFirst : MSBFirst;
+}
+
+
+/**
+ * Check if the X Shared Memory extension is available.
+ * Return:  0 = not available
+ *          1 = shared XImage support available
+ *          2 = shared Pixmap support available also
+ */
+int xmesa_check_for_xshm( Display *display )
+{
+#if defined(USE_XSHM)
+   int major, minor, ignore;
+   Bool pixmaps;
+
+   if (getenv("SP_NO_RAST")) 
+      return 0;
+
+   if (getenv("MESA_NOSHM")) {
+      return 0;
+   }
+
+   if (XQueryExtension( display, "MIT-SHM", &ignore, &ignore, &ignore )) {
+      if (XShmQueryVersion( display, &major, &minor, &pixmaps )==True) {
+	 return (pixmaps==True) ? 2 : 1;
+      }
+      else {
+	 return 0;
+      }
+   }
+   else {
+      return 0;
+   }
+#else
+   /* No  XSHM support */
+   return 0;
+#endif
+}
+
+
+/**
+ * Return the true number of bits per pixel for XImages.
+ * For example, if we request a 24-bit deep visual we may actually need/get
+ * 32bpp XImages.  This function returns the appropriate bpp.
+ * Input:  dpy - the X display
+ *         visinfo - desribes the visual to be used for XImages
+ * Return:  true number of bits per pixel for XImages
+ */
+static int
+bits_per_pixel( XMesaVisual xmv )
+{
+   Display *dpy = xmv->display;
+   XVisualInfo * visinfo = xmv->visinfo;
+   XImage *img;
+   int bitsPerPixel;
+   /* Create a temporary XImage */
+   img = XCreateImage( dpy, visinfo->visual, visinfo->depth,
+		       ZPixmap, 0,           /*format, offset*/
+		       (char*) MALLOC(8),    /*data*/
+		       1, 1,                 /*width, height*/
+		       32,                   /*bitmap_pad*/
+		       0                     /*bytes_per_line*/
+                     );
+   assert(img);
+   /* grab the bits/pixel value */
+   bitsPerPixel = img->bits_per_pixel;
+   /* free the XImage */
+   _mesa_free( img->data );
+   img->data = NULL;
+   XDestroyImage( img );
+   return bitsPerPixel;
+}
+
+
+
+/*
+ * Determine if a given X window ID is valid (window exists).
+ * Do this by calling XGetWindowAttributes() for the window and
+ * checking if we catch an X error.
+ * Input:  dpy - the display
+ *         win - the window to check for existance
+ * Return:  GL_TRUE - window exists
+ *          GL_FALSE - window doesn't exist
+ */
+static GLboolean WindowExistsFlag;
+
+static int window_exists_err_handler( Display* dpy, XErrorEvent* xerr )
+{
+   (void) dpy;
+   if (xerr->error_code == BadWindow) {
+      WindowExistsFlag = GL_FALSE;
+   }
+   return 0;
+}
+
+static GLboolean window_exists( Display *dpy, Window win )
+{
+   XWindowAttributes wa;
+   int (*old_handler)( Display*, XErrorEvent* );
+   WindowExistsFlag = GL_TRUE;
+   old_handler = XSetErrorHandler(window_exists_err_handler);
+   XGetWindowAttributes( dpy, win, &wa ); /* dummy request */
+   XSetErrorHandler(old_handler);
+   return WindowExistsFlag;
+}
+
+static Status
+get_drawable_size( Display *dpy, Drawable d, uint *width, uint *height )
+{
+   Window root;
+   Status stat;
+   int xpos, ypos;
+   unsigned int w, h, bw, depth;
+   stat = XGetGeometry(dpy, d, &root, &xpos, &ypos, &w, &h, &bw, &depth);
+   *width = w;
+   *height = h;
+   return stat;
+}
+
+
+/**
+ * Return the size of the window (or pixmap) that corresponds to the
+ * given XMesaBuffer.
+ * \param width  returns width in pixels
+ * \param height  returns height in pixels
+ */
+static void
+xmesa_get_window_size(Display *dpy, XMesaBuffer b,
+                      GLuint *width, GLuint *height)
+{
+   Status stat;
+
+   pipe_mutex_lock(_xmesa_lock);
+   XSync(b->xm_visual->display, 0); /* added for Chromium */
+   stat = get_drawable_size(dpy, b->drawable, width, height);
+   pipe_mutex_unlock(_xmesa_lock);
+
+   if (!stat) {
+      /* probably querying a window that's recently been destroyed */
+      _mesa_warning(NULL, "XGetGeometry failed!\n");
+      *width = *height = 1;
+   }
+}
+
+#define GET_REDMASK(__v)        __v->mesa_visual.redMask
+#define GET_GREENMASK(__v)      __v->mesa_visual.greenMask
+#define GET_BLUEMASK(__v)       __v->mesa_visual.blueMask
+
+
+/**
+ * Choose the pixel format for the given visual.
+ * This will tell the gallium driver how to pack pixel data into
+ * drawing surfaces.
+ */
+static GLuint
+choose_pixel_format(XMesaVisual v)
+{
+   boolean native_byte_order = (host_byte_order() == 
+                                ImageByteOrder(v->display));
+
+   if (   GET_REDMASK(v)   == 0x0000ff
+       && GET_GREENMASK(v) == 0x00ff00
+       && GET_BLUEMASK(v)  == 0xff0000
+       && v->BitsPerPixel == 32) {
+      if (native_byte_order) {
+         /* no byteswapping needed */
+         return 0 /* PIXEL_FORMAT_U_A8_B8_G8_R8 */;
+      }
+      else {
+         return PIPE_FORMAT_R8G8B8A8_UNORM;
+      }
+   }
+   else if (   GET_REDMASK(v)   == 0xff0000
+            && GET_GREENMASK(v) == 0x00ff00
+            && GET_BLUEMASK(v)  == 0x0000ff
+            && v->BitsPerPixel == 32) {
+      if (native_byte_order) {
+         /* no byteswapping needed */
+         return PIPE_FORMAT_A8R8G8B8_UNORM;
+      }
+      else {
+         return PIPE_FORMAT_B8G8R8A8_UNORM;
+      }
+   }
+   else if (   GET_REDMASK(v)   == 0xf800
+            && GET_GREENMASK(v) == 0x07e0
+            && GET_BLUEMASK(v)  == 0x001f
+            && native_byte_order
+            && v->BitsPerPixel == 16) {
+      /* 5-6-5 RGB */
+      return PIPE_FORMAT_R5G6B5_UNORM;
+   }
+
+   assert(0);
+   return 0;
+}
+
+
+
+/**********************************************************************/
+/*****                Linked list of XMesaBuffers                 *****/
+/**********************************************************************/
+
+XMesaBuffer XMesaBufferList = NULL;
+
+
+/**
+ * Allocate a new XMesaBuffer object which corresponds to the given drawable.
+ * Note that XMesaBuffer is derived from GLframebuffer.
+ * The new XMesaBuffer will not have any size (Width=Height=0).
+ *
+ * \param d  the corresponding X drawable (window or pixmap)
+ * \param type  either WINDOW, PIXMAP or PBUFFER, describing d
+ * \param vis  the buffer's visual
+ * \param cmap  the window's colormap, if known.
+ * \return new XMesaBuffer or NULL if any problem
+ */
+static XMesaBuffer
+create_xmesa_buffer(Drawable d, BufferType type,
+                    XMesaVisual vis, Colormap cmap)
+{
+   XMesaBuffer b;
+   GLframebuffer *fb;
+   enum pipe_format colorFormat, depthFormat, stencilFormat;
+   uint width, height;
+
+   ASSERT(type == WINDOW || type == PIXMAP || type == PBUFFER);
+
+   b = (XMesaBuffer) CALLOC_STRUCT(xmesa_buffer);
+   if (!b)
+      return NULL;
+
+   b->drawable = d;
+
+   b->xm_visual = vis;
+   b->type = type;
+   b->cmap = cmap;
+
+   /* determine PIPE_FORMATs for buffers */
+   colorFormat = choose_pixel_format(vis);
+
+   if (vis->mesa_visual.depthBits == 0)
+      depthFormat = PIPE_FORMAT_NONE;
+#ifdef GALLIUM_CELL /* XXX temporary for Cell! */
+   else
+      depthFormat = PIPE_FORMAT_S8Z24_UNORM;
+#else
+   else if (vis->mesa_visual.depthBits <= 16)
+      depthFormat = PIPE_FORMAT_Z16_UNORM;
+   else if (vis->mesa_visual.depthBits <= 24)
+      depthFormat = PIPE_FORMAT_S8Z24_UNORM;
+   else
+      depthFormat = PIPE_FORMAT_Z32_UNORM;
+#endif
+
+   if (vis->mesa_visual.stencilBits == 8) {
+      if (depthFormat == PIPE_FORMAT_S8Z24_UNORM)
+         stencilFormat = depthFormat;
+      else
+         stencilFormat = PIPE_FORMAT_S8_UNORM;
+   }
+   else {
+      /* no stencil */
+      stencilFormat = PIPE_FORMAT_NONE;
+      if (depthFormat == PIPE_FORMAT_S8Z24_UNORM) {
+         /* use 24-bit Z, undefined stencil channel */
+         depthFormat = PIPE_FORMAT_X8Z24_UNORM;
+      }
+   }
+
+
+   get_drawable_size(vis->display, d, &width, &height);
+
+   /*
+    * Create framebuffer, but we'll plug in our own renderbuffers below.
+    */
+   b->stfb = st_create_framebuffer(&vis->mesa_visual,
+                                   colorFormat, depthFormat, stencilFormat,
+                                   width, height,
+                                   (void *) b);
+   fb = &b->stfb->Base;
+
+   /*
+    * Create scratch XImage for xmesa_display_surface()
+    */
+   b->tempImage = XCreateImage(vis->display,
+                               vis->visinfo->visual,
+                               vis->visinfo->depth,
+                               ZPixmap, 0,   /* format, offset */
+                               NULL,         /* data */
+                               0, 0,         /* size */
+                               32,           /* bitmap_pad */
+                               0);           /* bytes_per_line */
+
+   /* GLX_EXT_texture_from_pixmap */
+   b->TextureTarget = 0;
+   b->TextureFormat = GLX_TEXTURE_FORMAT_NONE_EXT;
+   b->TextureMipmap = 0;
+
+   /* insert buffer into linked list */
+   b->Next = XMesaBufferList;
+   XMesaBufferList = b;
+
+   return b;
+}
+
+
+/**
+ * Find an XMesaBuffer by matching X display and colormap but NOT matching
+ * the notThis buffer.
+ */
+XMesaBuffer
+xmesa_find_buffer(Display *dpy, Colormap cmap, XMesaBuffer notThis)
+{
+   XMesaBuffer b;
+   for (b = XMesaBufferList; b; b = b->Next) {
+      if (b->xm_visual->display == dpy &&
+          b->cmap == cmap &&
+          b != notThis) {
+         return b;
+      }
+   }
+   return NULL;
+}
+
+
+/**
+ * Remove buffer from linked list, delete if no longer referenced.
+ */
+static void
+xmesa_free_buffer(XMesaBuffer buffer)
+{
+   XMesaBuffer prev = NULL, b;
+
+   for (b = XMesaBufferList; b; b = b->Next) {
+      if (b == buffer) {
+         struct gl_framebuffer *fb = &buffer->stfb->Base;
+
+         /* unlink buffer from list */
+         if (prev)
+            prev->Next = buffer->Next;
+         else
+            XMesaBufferList = buffer->Next;
+
+         /* mark as delete pending */
+         fb->DeletePending = GL_TRUE;
+
+         /* Since the X window for the XMesaBuffer is going away, we don't
+          * want to dereference this pointer in the future.
+          */
+         b->drawable = 0;
+
+         buffer->tempImage->data = NULL;
+         XDestroyImage(buffer->tempImage);
+
+         /* Unreference.  If count = zero we'll really delete the buffer */
+         _mesa_unreference_framebuffer(&fb);
+
+         XFreeGC(b->xm_visual->display, b->gc);
+
+         free(buffer);
+
+         return;
+      }
+      /* continue search */
+      prev = b;
+   }
+   /* buffer not found in XMesaBufferList */
+   _mesa_problem(NULL,"xmesa_free_buffer() - buffer not found\n");
+}
+
+
+
+/**********************************************************************/
+/*****                   Misc Private Functions                   *****/
+/**********************************************************************/
+
+
+/**
+ * When a context is bound for the first time, we can finally finish
+ * initializing the context's visual and buffer information.
+ * \param v  the XMesaVisual to initialize
+ * \param b  the XMesaBuffer to initialize (may be NULL)
+ * \param rgb_flag  TRUE = RGBA mode, FALSE = color index mode
+ * \param window  the window/pixmap we're rendering into
+ * \param cmap  the colormap associated with the window/pixmap
+ * \return GL_TRUE=success, GL_FALSE=failure
+ */
+static GLboolean
+initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b,
+                             GLboolean rgb_flag, Drawable window,
+                             Colormap cmap)
+{
+   ASSERT(!b || b->xm_visual == v);
+
+   /* Save true bits/pixel */
+   v->BitsPerPixel = bits_per_pixel(v);
+   assert(v->BitsPerPixel > 0);
+
+   if (rgb_flag == GL_FALSE) {
+      /* COLOR-INDEXED WINDOW: not supported*/
+      return GL_FALSE;
+   }
+   else {
+      /* RGB WINDOW:
+       * We support RGB rendering into almost any kind of visual.
+       */
+      const int xclass = v->mesa_visual.visualType;
+      if (xclass != GLX_TRUE_COLOR && xclass == !GLX_DIRECT_COLOR) {
+	 _mesa_warning(NULL,
+            "XMesa: RGB mode rendering not supported in given visual.\n");
+	 return GL_FALSE;
+      }
+      v->mesa_visual.indexBits = 0;
+
+      if (v->BitsPerPixel == 32) {
+         /* We use XImages for all front/back buffers.  If an X Window or
+          * X Pixmap is 32bpp, there's no guarantee that the alpha channel
+          * will be preserved.  For XImages we're in luck.
+          */
+         v->mesa_visual.alphaBits = 8;
+      }
+   }
+
+   /*
+    * If MESA_INFO env var is set print out some debugging info
+    * which can help Brian figure out what's going on when a user
+    * reports bugs.
+    */
+   if (_mesa_getenv("MESA_INFO")) {
+      _mesa_printf("X/Mesa visual = %p\n", (void *) v);
+      _mesa_printf("X/Mesa level = %d\n", v->mesa_visual.level);
+      _mesa_printf("X/Mesa depth = %d\n", v->visinfo->depth);
+      _mesa_printf("X/Mesa bits per pixel = %d\n", v->BitsPerPixel);
+   }
+
+   if (b && window) {
+      /* these should have been set in create_xmesa_buffer */
+      ASSERT(b->drawable == window);
+
+      /* Setup for single/double buffering */
+      if (v->mesa_visual.doubleBufferMode) {
+         /* Double buffered */
+         b->shm = xmesa_check_for_xshm( v->display );
+      }
+
+      /* X11 graphics context */
+      b->gc = XCreateGC( v->display, window, 0, NULL );
+      XSetFunction( v->display, b->gc, GXcopy );
+   }
+
+   return GL_TRUE;
+}
+
+
+
+#define NUM_VISUAL_TYPES   6
+
+/**
+ * Convert an X visual type to a GLX visual type.
+ * 
+ * \param visualType X visual type (i.e., \c TrueColor, \c StaticGray, etc.)
+ *        to be converted.
+ * \return If \c visualType is a valid X visual type, a GLX visual type will
+ *         be returned.  Otherwise \c GLX_NONE will be returned.
+ * 
+ * \note
+ * This code was lifted directly from lib/GL/glx/glcontextmodes.c in the
+ * DRI CVS tree.
+ */
+static GLint
+xmesa_convert_from_x_visual_type( int visualType )
+{
+    static const int glx_visual_types[ NUM_VISUAL_TYPES ] = {
+	GLX_STATIC_GRAY,  GLX_GRAY_SCALE,
+	GLX_STATIC_COLOR, GLX_PSEUDO_COLOR,
+	GLX_TRUE_COLOR,   GLX_DIRECT_COLOR
+    };
+
+    return ( (unsigned) visualType < NUM_VISUAL_TYPES )
+	? glx_visual_types[ visualType ] : GLX_NONE;
+}
+
+
+/**********************************************************************/
+/*****                       Public Functions                     *****/
+/**********************************************************************/
+
+
+/*
+ * Create a new X/Mesa visual.
+ * Input:  display - X11 display
+ *         visinfo - an XVisualInfo pointer
+ *         rgb_flag - GL_TRUE = RGB mode,
+ *                    GL_FALSE = color index mode
+ *         alpha_flag - alpha buffer requested?
+ *         db_flag - GL_TRUE = double-buffered,
+ *                   GL_FALSE = single buffered
+ *         stereo_flag - stereo visual?
+ *         ximage_flag - GL_TRUE = use an XImage for back buffer,
+ *                       GL_FALSE = use an off-screen pixmap for back buffer
+ *         depth_size - requested bits/depth values, or zero
+ *         stencil_size - requested bits/stencil values, or zero
+ *         accum_red_size - requested bits/red accum values, or zero
+ *         accum_green_size - requested bits/green accum values, or zero
+ *         accum_blue_size - requested bits/blue accum values, or zero
+ *         accum_alpha_size - requested bits/alpha accum values, or zero
+ *         num_samples - number of samples/pixel if multisampling, or zero
+ *         level - visual level, usually 0
+ *         visualCaveat - ala the GLX extension, usually GLX_NONE
+ * Return;  a new XMesaVisual or 0 if error.
+ */
+PUBLIC
+XMesaVisual XMesaCreateVisual( Display *display,
+                               XVisualInfo * visinfo,
+                               GLboolean rgb_flag,
+                               GLboolean alpha_flag,
+                               GLboolean db_flag,
+                               GLboolean stereo_flag,
+                               GLboolean ximage_flag,
+                               GLint depth_size,
+                               GLint stencil_size,
+                               GLint accum_red_size,
+                               GLint accum_green_size,
+                               GLint accum_blue_size,
+                               GLint accum_alpha_size,
+                               GLint num_samples,
+                               GLint level,
+                               GLint visualCaveat )
+{
+   XMesaVisual v;
+   GLint red_bits, green_bits, blue_bits, alpha_bits;
+
+   /* For debugging only */
+   if (_mesa_getenv("MESA_XSYNC")) {
+      /* This makes debugging X easier.
+       * In your debugger, set a breakpoint on _XError to stop when an
+       * X protocol error is generated.
+       */
+      XSynchronize( display, 1 );
+   }
+
+   v = (XMesaVisual) CALLOC_STRUCT(xmesa_visual);
+   if (!v) {
+      return NULL;
+   }
+
+   v->display = display;
+
+   /* Save a copy of the XVisualInfo struct because the user may X_mesa_free()
+    * the struct but we may need some of the information contained in it
+    * at a later time.
+    */
+   v->visinfo = (XVisualInfo *) MALLOC(sizeof(*visinfo));
+   if(!v->visinfo) {
+      _mesa_free(v);
+      return NULL;
+   }
+   MEMCPY(v->visinfo, visinfo, sizeof(*visinfo));
+
+   v->ximage_flag = ximage_flag;
+
+   v->mesa_visual.redMask = visinfo->red_mask;
+   v->mesa_visual.greenMask = visinfo->green_mask;
+   v->mesa_visual.blueMask = visinfo->blue_mask;
+   v->mesa_visual.visualID = visinfo->visualid;
+   v->mesa_visual.screen = visinfo->screen;
+
+#if !(defined(__cplusplus) || defined(c_plusplus))
+   v->mesa_visual.visualType = xmesa_convert_from_x_visual_type(visinfo->class);
+#else
+   v->mesa_visual.visualType = xmesa_convert_from_x_visual_type(visinfo->c_class);
+#endif
+
+   v->mesa_visual.visualRating = visualCaveat;
+
+   if (alpha_flag)
+      v->mesa_visual.alphaBits = 8;
+
+   (void) initialize_visual_and_buffer( v, NULL, rgb_flag, 0, 0 );
+
+   {
+      const int xclass = v->mesa_visual.visualType;
+      if (xclass == GLX_TRUE_COLOR || xclass == GLX_DIRECT_COLOR) {
+         red_bits   = _mesa_bitcount(GET_REDMASK(v));
+         green_bits = _mesa_bitcount(GET_GREENMASK(v));
+         blue_bits  = _mesa_bitcount(GET_BLUEMASK(v));
+      }
+      else {
+         /* this is an approximation */
+         int depth;
+         depth = v->visinfo->depth;
+         red_bits = depth / 3;
+         depth -= red_bits;
+         green_bits = depth / 2;
+         depth -= green_bits;
+         blue_bits = depth;
+         alpha_bits = 0;
+         assert( red_bits + green_bits + blue_bits == v->visinfo->depth );
+      }
+      alpha_bits = v->mesa_visual.alphaBits;
+   }
+
+   _mesa_initialize_visual( &v->mesa_visual,
+                            rgb_flag, db_flag, stereo_flag,
+                            red_bits, green_bits,
+                            blue_bits, alpha_bits,
+                            v->mesa_visual.indexBits,
+                            depth_size,
+                            stencil_size,
+                            accum_red_size, accum_green_size,
+                            accum_blue_size, accum_alpha_size,
+                            0 );
+
+   /* XXX minor hack */
+   v->mesa_visual.level = level;
+   return v;
+}
+
+
+PUBLIC
+void XMesaDestroyVisual( XMesaVisual v )
+{
+   _mesa_free(v->visinfo);
+   _mesa_free(v);
+}
+
+
+
+/**
+ * Create a new XMesaContext.
+ * \param v  the XMesaVisual
+ * \param share_list  another XMesaContext with which to share display
+ *                    lists or NULL if no sharing is wanted.
+ * \return an XMesaContext or NULL if error.
+ */
+PUBLIC
+XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list )
+{
+   static GLboolean firstTime = GL_TRUE;
+   struct pipe_screen *screen;
+   struct pipe_context *pipe;
+   XMesaContext c;
+   GLcontext *mesaCtx;
+   uint pf;
+
+   if (firstTime) {
+      pipe_mutex_init(_xmesa_lock);
+      firstTime = GL_FALSE;
+   }
+
+   /* Note: the XMesaContext contains a Mesa GLcontext struct (inheritance) */
+   c = (XMesaContext) CALLOC_STRUCT(xmesa_context);
+   if (!c)
+      return NULL;
+
+   pf = choose_pixel_format(v);
+   assert(pf);
+
+   c->xm_visual = v;
+   c->xm_buffer = NULL;   /* set later by XMesaMakeCurrent */
+   
+   /* XXX: create once per Xlib Display.
+    */
+   screen = driver.create_pipe_screen();
+   if (screen == NULL)
+      goto fail;
+
+   pipe = driver.create_pipe_context( screen,
+                                     (void *)c );
+   if (pipe == NULL)
+      goto fail;
+
+   c->st = st_create_context(pipe, 
+                             &v->mesa_visual,
+                             share_list ? share_list->st : NULL);
+   if (c->st == NULL)
+      goto fail;
+
+   mesaCtx = c->st->ctx;
+   c->st->ctx->DriverCtx = c;
+
+#if 00
+   _mesa_enable_sw_extensions(mesaCtx);
+   _mesa_enable_1_3_extensions(mesaCtx);
+   _mesa_enable_1_4_extensions(mesaCtx);
+   _mesa_enable_1_5_extensions(mesaCtx);
+   _mesa_enable_2_0_extensions(mesaCtx);
+#endif
+
+   return c;
+
+ fail:
+   if (c->st)
+      st_destroy_context(c->st);
+   else if (pipe)
+      pipe->destroy(pipe);
+
+   if (screen)
+      screen->destroy( screen );
+
+   FREE(c);
+   return NULL;
+}
+
+
+
+PUBLIC
+void XMesaDestroyContext( XMesaContext c )
+{
+   st_destroy_context(c->st);
+
+   /* FIXME: We should destroy the screen here, but if we do so, surfaces may 
+    * outlive it, causing segfaults
+   struct pipe_screen *screen = c->st->pipe->screen; 
+   screen->destroy(screen);
+   */
+
+   _mesa_free(c);
+}
+
+
+
+/**
+ * Private function for creating an XMesaBuffer which corresponds to an
+ * X window or pixmap.
+ * \param v  the window's XMesaVisual
+ * \param w  the window we're wrapping
+ * \return  new XMesaBuffer or NULL if error
+ */
+PUBLIC XMesaBuffer
+XMesaCreateWindowBuffer(XMesaVisual v, Window w)
+{
+   XWindowAttributes attr;
+   XMesaBuffer b;
+   Colormap cmap;
+   int depth;
+
+   assert(v);
+   assert(w);
+
+   /* Check that window depth matches visual depth */
+   XGetWindowAttributes( v->display, w, &attr );
+   depth = attr.depth;
+   if (v->visinfo->depth != depth) {
+      _mesa_warning(NULL, "XMesaCreateWindowBuffer: depth mismatch between visual (%d) and window (%d)!\n",
+                    v->visinfo->depth, depth);
+      return NULL;
+   }
+
+   /* Find colormap */
+   if (attr.colormap) {
+      cmap = attr.colormap;
+   }
+   else {
+      _mesa_warning(NULL, "Window %u has no colormap!\n", (unsigned int) w);
+      /* this is weird, a window w/out a colormap!? */
+      /* OK, let's just allocate a new one and hope for the best */
+      cmap = XCreateColormap(v->display, w, attr.visual, AllocNone);
+   }
+
+   b = create_xmesa_buffer((Drawable) w, WINDOW, v, cmap);
+   if (!b)
+      return NULL;
+
+   if (!initialize_visual_and_buffer( v, b, v->mesa_visual.rgbMode,
+                                      (Drawable) w, cmap )) {
+      xmesa_free_buffer(b);
+      return NULL;
+   }
+
+   return b;
+}
+
+
+
+/**
+ * Create a new XMesaBuffer from an X pixmap.
+ *
+ * \param v    the XMesaVisual
+ * \param p    the pixmap
+ * \param cmap the colormap, may be 0 if using a \c GLX_TRUE_COLOR or
+ *             \c GLX_DIRECT_COLOR visual for the pixmap
+ * \returns new XMesaBuffer or NULL if error
+ */
+PUBLIC XMesaBuffer
+XMesaCreatePixmapBuffer(XMesaVisual v, Pixmap p, Colormap cmap)
+{
+   XMesaBuffer b;
+
+   assert(v);
+
+   b = create_xmesa_buffer((Drawable) p, PIXMAP, v, cmap);
+   if (!b)
+      return NULL;
+
+   if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode,
+				     (Drawable) p, cmap)) {
+      xmesa_free_buffer(b);
+      return NULL;
+   }
+
+   return b;
+}
+
+
+/**
+ * For GLX_EXT_texture_from_pixmap
+ */
+XMesaBuffer
+XMesaCreatePixmapTextureBuffer(XMesaVisual v, Pixmap p,
+                               Colormap cmap,
+                               int format, int target, int mipmap)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   XMesaBuffer b;
+   GLuint width, height;
+
+   assert(v);
+
+   b = create_xmesa_buffer((Drawable) p, PIXMAP, v, cmap);
+   if (!b)
+      return NULL;
+
+   /* get pixmap size, update framebuffer/renderbuffer dims */
+   xmesa_get_window_size(v->display, b, &width, &height);
+   _mesa_resize_framebuffer(NULL, &(b->stfb->Base), width, height);
+
+   if (target == 0) {
+      /* examine dims */
+      if (ctx->Extensions.ARB_texture_non_power_of_two) {
+         target = GLX_TEXTURE_2D_EXT;
+      }
+      else if (   _mesa_bitcount(width)  == 1
+               && _mesa_bitcount(height) == 1) {
+         /* power of two size */
+         if (height == 1) {
+            target = GLX_TEXTURE_1D_EXT;
+         }
+         else {
+            target = GLX_TEXTURE_2D_EXT;
+         }
+      }
+      else if (ctx->Extensions.NV_texture_rectangle) {
+         target = GLX_TEXTURE_RECTANGLE_EXT;
+      }
+      else {
+         /* non power of two textures not supported */
+         XMesaDestroyBuffer(b);
+         return 0;
+      }
+   }
+
+   b->TextureTarget = target;
+   b->TextureFormat = format;
+   b->TextureMipmap = mipmap;
+
+   if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode,
+				     (Drawable) p, cmap)) {
+      xmesa_free_buffer(b);
+      return NULL;
+   }
+
+   return b;
+}
+
+
+
+XMesaBuffer
+XMesaCreatePBuffer(XMesaVisual v, Colormap cmap,
+                   unsigned int width, unsigned int height)
+{
+   Window root;
+   Drawable drawable;  /* X Pixmap Drawable */
+   XMesaBuffer b;
+
+   /* allocate pixmap for front buffer */
+   root = RootWindow( v->display, v->visinfo->screen );
+   drawable = XCreatePixmap(v->display, root, width, height,
+                            v->visinfo->depth);
+   if (!drawable)
+      return NULL;
+
+   b = create_xmesa_buffer(drawable, PBUFFER, v, cmap);
+   if (!b)
+      return NULL;
+
+   if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode,
+				     drawable, cmap)) {
+      xmesa_free_buffer(b);
+      return NULL;
+   }
+
+   return b;
+}
+
+
+
+/*
+ * Deallocate an XMesaBuffer structure and all related info.
+ */
+PUBLIC void
+XMesaDestroyBuffer(XMesaBuffer b)
+{
+   xmesa_free_buffer(b);
+}
+
+
+/**
+ * Query the current window size and update the corresponding GLframebuffer
+ * and all attached renderbuffers.
+ * Called when:
+ *  1. the first time a buffer is bound to a context.
+ *  2. SwapBuffers.  XXX probabaly from xm_flush_frontbuffer() too...
+ * Note: it's possible (and legal) for xmctx to be NULL.  That can happen
+ * when resizing a buffer when no rendering context is bound.
+ */
+void
+xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer)
+{
+   GLuint width, height;
+   xmesa_get_window_size(drawBuffer->xm_visual->display, drawBuffer, &width, &height);
+   st_resize_framebuffer(drawBuffer->stfb, width, height);
+}
+
+
+
+
+/*
+ * Bind buffer b to context c and make c the current rendering context.
+ */
+PUBLIC
+GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer,
+                             XMesaBuffer readBuffer )
+{
+   if (c) {
+      if (!drawBuffer || !readBuffer)
+         return GL_FALSE;  /* must specify buffers! */
+
+#if 0
+      /* XXX restore this optimization */
+      if (&(c->mesa) == _mesa_get_current_context()
+          && c->mesa.DrawBuffer == &drawBuffer->mesa_buffer
+          && c->mesa.ReadBuffer == &readBuffer->mesa_buffer
+          && xmesa_buffer(c->mesa.DrawBuffer)->wasCurrent) {
+         /* same context and buffer, do nothing */
+         return GL_TRUE;
+      }
+#endif
+
+      c->xm_buffer = drawBuffer;
+
+      /* Call this periodically to detect when the user has begun using
+       * GL rendering from multiple threads.
+       */
+      _glapi_check_multithread();
+
+      st_make_current(c->st, drawBuffer->stfb, readBuffer->stfb);
+
+      xmesa_check_and_update_buffer_size(c, drawBuffer);
+      if (readBuffer != drawBuffer)
+         xmesa_check_and_update_buffer_size(c, readBuffer);
+
+      /* Solution to Stephane Rehel's problem with glXReleaseBuffersMESA(): */
+      drawBuffer->wasCurrent = GL_TRUE;
+   }
+   else {
+      /* Detach */
+      st_make_current( NULL, NULL, NULL );
+   }
+   return GL_TRUE;
+}
+
+
+/*
+ * Unbind the context c from its buffer.
+ */
+GLboolean XMesaUnbindContext( XMesaContext c )
+{
+   /* A no-op for XFree86 integration purposes */
+   return GL_TRUE;
+}
+
+
+XMesaContext XMesaGetCurrentContext( void )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   if (ctx) {
+      XMesaContext xmesa = xmesa_context(ctx);
+      return xmesa;
+   }
+   else {
+      return 0;
+   }
+}
+
+
+
+
+
+
+/*
+ * Copy the back buffer to the front buffer.  If there's no back buffer
+ * this is a no-op.
+ */
+PUBLIC
+void XMesaSwapBuffers( XMesaBuffer b )
+{
+   struct pipe_surface *surf;
+
+   /* If we're swapping the buffer associated with the current context
+    * we have to flush any pending rendering commands first.
+    */
+   st_notify_swapbuffers(b->stfb);
+
+   st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT, &surf);
+   if (surf) {
+      driver.display_surface(b, surf);
+   }
+
+   xmesa_check_and_update_buffer_size(NULL, b);
+}
+
+
+
+/*
+ * Copy sub-region of back buffer to front buffer
+ */
+void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height )
+{
+   struct pipe_surface *surf_front;
+   struct pipe_surface *surf_back;
+   struct pipe_context *pipe = NULL; /* XXX fix */
+
+   st_get_framebuffer_surface(b->stfb, ST_SURFACE_FRONT_LEFT, &surf_front);
+   st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT, &surf_back);
+
+   if (!surf_front || !surf_back)
+      return;
+
+   pipe->surface_copy(pipe,
+                      FALSE,
+                      surf_front, x, y,  /* dest */
+                      surf_back, x, y,   /* src */
+                      width, height);
+}
+
+
+
+void XMesaFlush( XMesaContext c )
+{
+   if (c && c->xm_visual->display) {
+      st_finish(c->st);
+      XSync( c->xm_visual->display, False );
+   }
+}
+
+
+
+
+
+XMesaBuffer XMesaFindBuffer( Display *dpy, Drawable d )
+{
+   XMesaBuffer b;
+   for (b=XMesaBufferList; b; b=b->Next) {
+      if (b->drawable == d && b->xm_visual->display == dpy) {
+         return b;
+      }
+   }
+   return NULL;
+}
+
+
+/**
+ * Free/destroy all XMesaBuffers associated with given display.
+ */
+void xmesa_destroy_buffers_on_display(Display *dpy)
+{
+   XMesaBuffer b, next;
+   for (b = XMesaBufferList; b; b = next) {
+      next = b->Next;
+      if (b->xm_visual->display == dpy) {
+         xmesa_free_buffer(b);
+      }
+   }
+}
+
+
+/*
+ * Look for XMesaBuffers whose X window has been destroyed.
+ * Deallocate any such XMesaBuffers.
+ */
+void XMesaGarbageCollect( void )
+{
+   XMesaBuffer b, next;
+   for (b=XMesaBufferList; b; b=next) {
+      next = b->Next;
+      if (b->xm_visual &&
+          b->xm_visual->display &&
+          b->drawable &&
+          b->type == WINDOW) {
+         XSync(b->xm_visual->display, False);
+         if (!window_exists( b->xm_visual->display, b->drawable )) {
+            /* found a dead window, free the ancillary info */
+            XMesaDestroyBuffer( b );
+         }
+      }
+   }
+}
+
+
+
+
+PUBLIC void
+XMesaBindTexImage(Display *dpy, XMesaBuffer drawable, int buffer,
+                  const int *attrib_list)
+{
+}
+
+
+
+PUBLIC void
+XMesaReleaseTexImage(Display *dpy, XMesaBuffer drawable, int buffer)
+{
+}
+
diff --git a/src/gallium/state_trackers/glx/xlib/xm_api.h b/src/gallium/state_trackers/glx/xlib/xm_api.h
new file mode 100644
index 0000000000..2b8302d174
--- /dev/null
+++ b/src/gallium/state_trackers/glx/xlib/xm_api.h
@@ -0,0 +1,393 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ * 
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+
+/* Sample Usage:
+
+In addition to the usual X calls to select a visual, create a colormap
+and create a window, you must do the following to use the X/Mesa interface:
+
+1. Call XMesaCreateVisual() to make an XMesaVisual from an XVisualInfo.
+
+2. Call XMesaCreateContext() to create an X/Mesa rendering context, given
+   the XMesaVisual.
+
+3. Call XMesaCreateWindowBuffer() to create an XMesaBuffer from an X window
+   and XMesaVisual.
+
+4. Call XMesaMakeCurrent() to bind the XMesaBuffer to an XMesaContext and
+   to make the context the current one.
+
+5. Make gl* calls to render your graphics.
+
+6. Use XMesaSwapBuffers() when double buffering to swap front/back buffers.
+
+7. Before the X window is destroyed, call XMesaDestroyBuffer().
+
+8. Before exiting, call XMesaDestroyVisual and XMesaDestroyContext.
+
+*/
+
+
+
+
+#ifndef XMESA_H
+#define XMESA_H
+
+
+#include "mtypes.h"
+#include "state_tracker/st_context.h"
+#include "state_tracker/st_public.h"
+#include "pipe/p_thread.h"
+
+
+# include <X11/Xlib.h>
+# include <X11/Xlibint.h>
+# include <X11/Xutil.h>
+# ifdef USE_XSHM  /* was SHM */
+#  include <sys/ipc.h>
+#  include <sys/shm.h>
+#  include <X11/extensions/XShm.h>
+# endif
+
+typedef struct xmesa_buffer *XMesaBuffer;
+typedef struct xmesa_context *XMesaContext;
+typedef struct xmesa_visual *XMesaVisual;
+
+
+
+/*
+ * Create a new X/Mesa visual.
+ * Input:  display - X11 display
+ *         visinfo - an XVisualInfo pointer
+ *         rgb_flag - GL_TRUE = RGB mode,
+ *                    GL_FALSE = color index mode
+ *         alpha_flag - alpha buffer requested?
+ *         db_flag - GL_TRUE = double-buffered,
+ *                   GL_FALSE = single buffered
+ *         stereo_flag - stereo visual?
+ *         ximage_flag - GL_TRUE = use an XImage for back buffer,
+ *                       GL_FALSE = use an off-screen pixmap for back buffer
+ *         depth_size - requested bits/depth values, or zero
+ *         stencil_size - requested bits/stencil values, or zero
+ *         accum_red_size - requested bits/red accum values, or zero
+ *         accum_green_size - requested bits/green accum values, or zero
+ *         accum_blue_size - requested bits/blue accum values, or zero
+ *         accum_alpha_size - requested bits/alpha accum values, or zero
+ *         num_samples - number of samples/pixel if multisampling, or zero
+ *         level - visual level, usually 0
+ *         visualCaveat - ala the GLX extension, usually GLX_NONE_EXT
+ * Return;  a new XMesaVisual or 0 if error.
+ */
+extern XMesaVisual XMesaCreateVisual( Display *display,
+                                      XVisualInfo * visinfo,
+                                      GLboolean rgb_flag,
+                                      GLboolean alpha_flag,
+                                      GLboolean db_flag,
+                                      GLboolean stereo_flag,
+                                      GLboolean ximage_flag,
+                                      GLint depth_size,
+                                      GLint stencil_size,
+                                      GLint accum_red_size,
+                                      GLint accum_green_size,
+                                      GLint accum_blue_size,
+                                      GLint accum_alpha_size,
+                                      GLint num_samples,
+                                      GLint level,
+                                      GLint visualCaveat );
+
+/*
+ * Destroy an XMesaVisual, but not the associated XVisualInfo.
+ */
+extern void XMesaDestroyVisual( XMesaVisual v );
+
+
+
+/*
+ * Create a new XMesaContext for rendering into an X11 window.
+ *
+ * Input:  visual - an XMesaVisual
+ *         share_list - another XMesaContext with which to share display
+ *                      lists or NULL if no sharing is wanted.
+ * Return:  an XMesaContext or NULL if error.
+ */
+extern XMesaContext XMesaCreateContext( XMesaVisual v,
+					XMesaContext share_list );
+
+
+/*
+ * Destroy a rendering context as returned by XMesaCreateContext()
+ */
+extern void XMesaDestroyContext( XMesaContext c );
+
+
+
+/*
+ * Create an XMesaBuffer from an X window.
+ */
+extern XMesaBuffer XMesaCreateWindowBuffer( XMesaVisual v, Window w );
+
+
+/*
+ * Create an XMesaBuffer from an X pixmap.
+ */
+extern XMesaBuffer XMesaCreatePixmapBuffer( XMesaVisual v,
+					    Pixmap p,
+					    Colormap cmap );
+
+
+/*
+ * Destroy an XMesaBuffer, but not the corresponding window or pixmap.
+ */
+extern void XMesaDestroyBuffer( XMesaBuffer b );
+
+
+/*
+ * Return the XMesaBuffer handle which corresponds to an X drawable, if any.
+ *
+ * New in Mesa 2.3.
+ */
+extern XMesaBuffer XMesaFindBuffer( Display *dpy,
+				    Drawable d );
+
+
+
+/*
+ * Bind two buffers (read and draw) to a context and make the
+ * context the current one.
+ * New in Mesa 3.3
+ */
+extern GLboolean XMesaMakeCurrent2( XMesaContext c,
+                                    XMesaBuffer drawBuffer,
+                                    XMesaBuffer readBuffer );
+
+
+/*
+ * Unbind the current context from its buffer.
+ */
+extern GLboolean XMesaUnbindContext( XMesaContext c );
+
+
+/*
+ * Return a handle to the current context.
+ */
+extern XMesaContext XMesaGetCurrentContext( void );
+
+
+/*
+ * Swap the front and back buffers for the given buffer.  No action is
+ * taken if the buffer is not double buffered.
+ */
+extern void XMesaSwapBuffers( XMesaBuffer b );
+
+
+/*
+ * Copy a sub-region of the back buffer to the front buffer.
+ *
+ * New in Mesa 2.6
+ */
+extern void XMesaCopySubBuffer( XMesaBuffer b,
+				int x,
+				int y,
+				int width,
+				int height );
+
+
+
+
+
+/*
+ * Flush/sync a context
+ */
+extern void XMesaFlush( XMesaContext c );
+
+
+
+/*
+ * Scan for XMesaBuffers whose window/pixmap has been destroyed, then free
+ * any memory used by that buffer.
+ *
+ * New in Mesa 2.3.
+ */
+extern void XMesaGarbageCollect( void );
+
+
+
+/*
+ * Create a pbuffer.
+ * New in Mesa 4.1
+ */
+extern XMesaBuffer XMesaCreatePBuffer(XMesaVisual v, Colormap cmap,
+                                      unsigned int width, unsigned int height);
+
+
+
+/*
+ * Texture from Pixmap
+ * New in Mesa 7.1
+ */
+extern void
+XMesaBindTexImage(Display *dpy, XMesaBuffer drawable, int buffer,
+                  const int *attrib_list);
+
+extern void
+XMesaReleaseTexImage(Display *dpy, XMesaBuffer drawable, int buffer);
+
+
+extern XMesaBuffer
+XMesaCreatePixmapTextureBuffer(XMesaVisual v, Pixmap p,
+                               Colormap cmap,
+                               int format, int target, int mipmap);
+
+
+
+
+/***********************************************************************
+ */
+
+extern pipe_mutex _xmesa_lock;
+
+extern struct xmesa_buffer *XMesaBufferList;
+
+
+/**
+ * Visual inforation, derived from GLvisual.
+ * Basically corresponds to an XVisualInfo.
+ */
+struct xmesa_visual {
+   GLvisual mesa_visual;	/* Device independent visual parameters */
+   Display *display;	/* The X11 display */
+   XVisualInfo * visinfo;	/* X's visual info (pointer to private copy) */
+   XVisualInfo *vishandle;	/* Only used in fakeglx.c */
+   GLint BitsPerPixel;		/* True bits per pixel for XImages */
+
+   GLboolean ximage_flag;	/* Use XImage for back buffer (not pixmap)? */
+};
+
+
+/**
+ * Context info, derived from st_context.
+ * Basically corresponds to a GLXContext.
+ */
+struct xmesa_context {
+   struct st_context *st;
+   XMesaVisual xm_visual;	/** pixel format info */
+   XMesaBuffer xm_buffer;	/** current drawbuffer */
+};
+
+
+/**
+ * Types of X/GLX drawables we might render into.
+ */
+typedef enum {
+   WINDOW,          /* An X window */
+   GLXWINDOW,       /* GLX window */
+   PIXMAP,          /* GLX pixmap */
+   PBUFFER          /* GLX Pbuffer */
+} BufferType;
+
+
+/**
+ * Framebuffer information, derived from.
+ * Basically corresponds to a GLXDrawable.
+ */
+struct xmesa_buffer {
+   struct st_framebuffer *stfb;
+
+   GLboolean wasCurrent;	/* was ever the current buffer? */
+   XMesaVisual xm_visual;	/* the X/Mesa visual */
+   Drawable drawable;	/* Usually the X window ID */
+   Colormap cmap;		/* the X colormap */
+   BufferType type;             /* window, pixmap, pbuffer or glxwindow */
+
+   XImage *tempImage;
+   unsigned long selectedEvents;/* for pbuffers only */
+
+   GLuint shm;			/* X Shared Memory extension status:	*/
+				/*    0 = not available			*/
+				/*    1 = XImage support available	*/
+				/*    2 = Pixmap support available too	*/
+#if defined(USE_XSHM)
+   XShmSegmentInfo shminfo;
+#endif
+
+   GC gc;			/* scratch GC for span, line, tri drawing */
+
+   /* GLX_EXT_texture_from_pixmap */
+   GLint TextureTarget; /** GLX_TEXTURE_1D_EXT, for example */
+   GLint TextureFormat; /** GLX_TEXTURE_FORMAT_RGB_EXT, for example */
+   GLint TextureMipmap; /** 0 or 1 */
+
+   struct xmesa_buffer *Next;	/* Linked list pointer: */
+};
+
+
+
+/** cast wrapper */
+static INLINE XMesaContext
+xmesa_context(GLcontext *ctx)
+{
+   return (XMesaContext) ctx->DriverCtx;
+}
+
+
+/** cast wrapper */
+static INLINE XMesaBuffer
+xmesa_buffer(GLframebuffer *fb)
+{
+   struct st_framebuffer *stfb = (struct st_framebuffer *) fb;
+   return (XMesaBuffer) st_framebuffer_private(stfb);
+}
+
+
+extern void
+xmesa_delete_framebuffer(struct gl_framebuffer *fb);
+
+extern XMesaBuffer
+xmesa_find_buffer(Display *dpy, Colormap cmap, XMesaBuffer notThis);
+
+extern void
+xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer);
+
+extern void
+xmesa_destroy_buffers_on_display(Display *dpy);
+
+static INLINE GLuint
+xmesa_buffer_width(XMesaBuffer b)
+{
+   return b->stfb->Base.Width;
+}
+
+static INLINE GLuint
+xmesa_buffer_height(XMesaBuffer b)
+{
+   return b->stfb->Base.Height;
+}
+
+extern int
+xmesa_check_for_xshm(Display *display);
+
+
+#endif
diff --git a/src/gallium/state_trackers/glx/xlib/xm_winsys.h b/src/gallium/state_trackers/glx/xlib/xm_winsys.h
new file mode 100644
index 0000000000..0e57605c34
--- /dev/null
+++ b/src/gallium/state_trackers/glx/xlib/xm_winsys.h
@@ -0,0 +1,59 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef XM_WINSYS_H
+#define XM_WINSYS_H
+
+struct pipe_context;
+struct pipe_screen;
+struct pipe_surface;
+struct xmesa_buffer;
+
+
+struct xm_driver {
+
+   struct pipe_screen *(*create_pipe_screen)( void );
+
+   /* The context_private argument needs to go away.  Is currently used
+    * in a round-about way to associate a display-target surface with its
+    * Xlib window.
+    */
+   struct pipe_context *(*create_pipe_context)( struct pipe_screen *,
+                                                void *context_private );
+
+   void (*display_surface)( struct xmesa_buffer *, 
+                            struct pipe_surface * );
+
+};
+
+
+extern void
+xmesa_set_driver( const struct xm_driver *driver );
+
+
+#endif
diff --git a/src/gallium/state_trackers/python/README b/src/gallium/state_trackers/python/README
new file mode 100644
index 0000000000..8f45fb6d1b
--- /dev/null
+++ b/src/gallium/state_trackers/python/README
@@ -0,0 +1,33 @@
+This directory contains Python bindings to Gallium3D. It looks like a state
+tracker from the pipe driver perspective, and it looks like a pipe driver from
+the python script perspective.
+
+
+To build you'll need:
+* Python (with development packages)
+* SCons
+* SWIG
+* Python Imaging Library with TK support (for the samples)
+
+Invoke scons on the top dir as
+ 
+ scons statetrackers=python
+
+To use do
+
+ export PYTHONPATH=build/XXXX-XXXX-XXXX/gallium/state_trackers/python
+
+and then try running
+
+  python src/gallium/state_trackers/python/samples/tri.py
+
+which should show a triangle.
+
+
+This is still work in progress:
+- errors are not handled properly and almost always result in crash
+- state atoms with array members are awkward to set
+- there no efficient way to view images
+
+--
+Jose Fonseca <jrfonseca@tungstengraphics.com>
diff --git a/src/gallium/state_trackers/python/SConscript b/src/gallium/state_trackers/python/SConscript
new file mode 100644
index 0000000000..1581182aec
--- /dev/null
+++ b/src/gallium/state_trackers/python/SConscript
@@ -0,0 +1,34 @@
+import sys
+import os.path
+
+Import('*')
+
+if 'python' in env['statetrackers']:
+
+    env = env.Clone()
+    
+    env.Tool('python')
+    
+    env.Tool('swig')
+    env.Append(SWIGPATH = ['#src/gallium/include', '#src/gallium/include/pipe'])
+    env.Append(SWIGFLAGS = ['-python', '-keyword'])
+    
+    env.Append(CPPPATH = '.')
+
+    pyst = env.ConvenienceLibrary(
+        target = 'pyst',
+        source = [
+            'gallium.i',
+            'st_device.c',
+            'st_sample.c',
+            'st_softpipe_winsys.c',
+        ],
+    )
+
+    env.SharedLibrary(
+        target = '_gallium',
+        source = [
+            'st_hardpipe_winsys.c',
+        ],
+        LIBS = [pyst, softpipe, trace] + auxiliaries + env['LIBS'],
+    )
diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i
new file mode 100644
index 0000000000..f4c4b36ea7
--- /dev/null
+++ b/src/gallium/state_trackers/python/gallium.i
@@ -0,0 +1,103 @@
+ /**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * SWIG interface definion for Gallium types.
+ *
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+%module gallium;
+
+%{
+
+#include <stdio.h>
+#include <Python.h>
+
+#include "pipe/p_screen.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+#include "pipe/p_shader_tokens.h" 
+#include "cso_cache/cso_context.h"
+#include "util/u_draw_quad.h" 
+#include "util/u_tile.h" 
+#include "tgsi/tgsi_text.h" 
+#include "tgsi/tgsi_dump.h" 
+
+#include "st_device.h"
+#include "st_sample.h"
+
+%}
+
+%include "typemaps.i"
+
+%include "cstring.i"
+
+%include "carrays.i"
+%array_class(unsigned char, ByteArray);
+%array_class(int, IntArray);
+%array_class(unsigned, UnsignedArray);
+%array_class(float, FloatArray);
+
+
+%rename(Device) st_device;
+%rename(Context) st_context;
+%rename(Texture) pipe_texture;
+%rename(Surface) pipe_surface;
+%rename(Buffer) st_buffer;
+
+%rename(BlendColor) pipe_blend_color;
+%rename(Blend) pipe_blend_state;
+%rename(Clip) pipe_clip_state;
+%rename(ConstantBuffer) pipe_constant_buffer;
+%rename(Depth) pipe_depth_state;
+%rename(Stencil) pipe_stencil_state;
+%rename(Alpha) pipe_alpha_state;
+%rename(DepthStencilAlpha) pipe_depth_stencil_alpha_state;
+%rename(FormatBlock) pipe_format_block;
+%rename(Framebuffer) pipe_framebuffer_state;
+%rename(PolyStipple) pipe_poly_stipple;
+%rename(Rasterizer) pipe_rasterizer_state;
+%rename(Sampler) pipe_sampler_state;
+%rename(Scissor) pipe_scissor_state;
+%rename(Shader) pipe_shader_state;
+%rename(VertexBuffer) pipe_vertex_buffer;
+%rename(VertexElement) pipe_vertex_element;
+%rename(Viewport) pipe_viewport_state;
+
+
+%include "p_compiler.i"
+%include "pipe/p_defines.h";
+%include "p_format.i"
+
+%include "p_device.i"
+%include "p_context.i"
+%include "p_texture.i"
+%include "p_state.i"
+
diff --git a/src/gallium/state_trackers/python/p_compiler.i b/src/gallium/state_trackers/python/p_compiler.i
new file mode 100644
index 0000000000..15f6ba5b9d
--- /dev/null
+++ b/src/gallium/state_trackers/python/p_compiler.i
@@ -0,0 +1,29 @@
+ /**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+typedef unsigned char ubyte;
diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i
new file mode 100644
index 0000000000..7b8b64b592
--- /dev/null
+++ b/src/gallium/state_trackers/python/p_context.i
@@ -0,0 +1,289 @@
+ /**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * SWIG interface definion for Gallium types.
+ *
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+%nodefaultctor st_context;
+%nodefaultdtor st_context;
+
+struct st_context {
+};
+
+%extend st_context {
+   
+   ~st_context() {
+      st_context_destroy($self);
+   }
+   
+   /*
+    * State functions (create/bind/destroy state objects)
+    */
+
+   void set_blend( const struct pipe_blend_state *state ) {
+      cso_set_blend($self->cso, state);
+   }
+   
+   void set_sampler( unsigned index, const struct pipe_sampler_state *state ) {
+      cso_single_sampler($self->cso, index, state);
+      cso_single_sampler_done($self->cso);
+   }
+
+   void set_rasterizer( const struct pipe_rasterizer_state *state ) {
+      cso_set_rasterizer($self->cso, state);
+   }
+
+   void set_depth_stencil_alpha(const struct pipe_depth_stencil_alpha_state *state) {
+      cso_set_depth_stencil_alpha($self->cso, state);
+   }
+
+   void set_fragment_shader( const struct pipe_shader_state *state ) {
+      void *fs;
+      
+      if(!state) {
+         cso_set_fragment_shader_handle($self->cso, NULL);
+         return;
+      }
+      
+      fs = $self->pipe->create_fs_state($self->pipe, state);
+      if(!fs)
+         return;
+      
+      if(cso_set_fragment_shader_handle($self->cso, fs) != PIPE_OK)
+         return;
+
+      cso_delete_fragment_shader($self->cso, $self->fs);
+      $self->fs = fs;
+   }
+
+   void set_vertex_shader( const struct pipe_shader_state *state ) {
+      void *vs;
+      
+      if(!state) {
+         cso_set_vertex_shader_handle($self->cso, NULL);
+         return;
+      }
+      
+      vs = $self->pipe->create_vs_state($self->pipe, state);
+      if(!vs)
+         return;
+      
+      if(cso_set_vertex_shader_handle($self->cso, vs) != PIPE_OK)
+         return;
+
+      cso_delete_vertex_shader($self->cso, $self->vs);
+      $self->vs = vs;
+   }
+
+   /*
+    * Parameter-like state (or properties)
+    */
+   
+   void set_blend_color(const struct pipe_blend_color *state ) {
+      cso_set_blend_color($self->cso, state);
+   }
+
+   void set_clip(const struct pipe_clip_state *state ) {
+      $self->pipe->set_clip_state($self->pipe, state);
+   }
+
+   void set_constant_buffer(unsigned shader, unsigned index,
+                            struct st_buffer *buffer ) 
+   {
+      struct pipe_constant_buffer state;
+      memset(&state, 0, sizeof(state));
+      state.buffer = buffer ? buffer->buffer : NULL;
+      state.size = buffer->buffer->size;
+      $self->pipe->set_constant_buffer($self->pipe, shader, index, &state);
+   }
+
+   void set_framebuffer(const struct pipe_framebuffer_state *state ) {
+      cso_set_framebuffer($self->cso, state);
+   }
+
+   void set_polygon_stipple(const struct pipe_poly_stipple *state ) {
+      $self->pipe->set_polygon_stipple($self->pipe, state);
+   }
+
+   void set_scissor(const struct pipe_scissor_state *state ) {
+      $self->pipe->set_scissor_state($self->pipe, state);
+   }
+
+   void set_viewport(const struct pipe_viewport_state *state) {
+      cso_set_viewport($self->cso, state);
+   }
+
+   void set_sampler_texture(unsigned index,
+                            struct pipe_texture *texture) {
+      if(!texture)
+         texture = $self->default_texture;
+      pipe_texture_reference(&$self->sampler_textures[index], texture);
+      $self->pipe->set_sampler_textures($self->pipe, 
+                                        PIPE_MAX_SAMPLERS,
+                                        $self->sampler_textures);
+   }
+
+   void set_vertex_buffer(unsigned index,
+                          unsigned pitch, 
+                          unsigned max_index,
+                          unsigned buffer_offset,
+                          struct st_buffer *buffer)
+   {
+      unsigned i;
+      struct pipe_vertex_buffer state;
+      
+      memset(&state, 0, sizeof(state));
+      state.pitch = pitch;
+      state.max_index = max_index;
+      state.buffer_offset = buffer_offset;
+      state.buffer = buffer ? buffer->buffer : NULL;
+
+      memcpy(&$self->vertex_buffers[index], &state, sizeof(state));
+      
+      for(i = 0; i < PIPE_MAX_ATTRIBS; ++i)
+         if(self->vertex_buffers[i].buffer)
+            $self->num_vertex_buffers = i + 1;
+      
+      $self->pipe->set_vertex_buffers($self->pipe, 
+                                      $self->num_vertex_buffers, 
+                                      $self->vertex_buffers);
+   }
+
+   void set_vertex_element(unsigned index,
+                           const struct pipe_vertex_element *element) 
+   {
+      memcpy(&$self->vertex_elements[index], element, sizeof(*element));
+   }
+
+   void set_vertex_elements(unsigned num) 
+   {
+      $self->num_vertex_elements = num;
+      $self->pipe->set_vertex_elements($self->pipe, 
+                                       $self->num_vertex_elements, 
+                                       $self->vertex_elements);
+   }
+
+   /*
+    * Draw functions
+    */
+   
+   void draw_arrays(unsigned mode, unsigned start, unsigned count) {
+      $self->pipe->draw_arrays($self->pipe, mode, start, count);
+   }
+
+   void draw_elements( struct st_buffer *indexBuffer,
+                       unsigned indexSize,
+                       unsigned mode, unsigned start, unsigned count) 
+   {
+      $self->pipe->draw_elements($self->pipe, 
+                                 indexBuffer->buffer, 
+                                 indexSize, 
+                                 mode, start, count);
+   }
+
+   void draw_range_elements( struct st_buffer *indexBuffer,
+                             unsigned indexSize, unsigned minIndex, unsigned maxIndex,
+                             unsigned mode, unsigned start, unsigned count)
+   {
+      $self->pipe->draw_range_elements($self->pipe, 
+                                       indexBuffer->buffer, 
+                                       indexSize, minIndex, maxIndex,
+                                       mode, start, count);
+   }
+
+   void draw_vertices(unsigned prim,
+                      unsigned num_verts,
+                      unsigned num_attribs,
+                      const float *vertices) 
+   {
+      struct pipe_context *pipe = $self->pipe;
+      struct pipe_screen *screen = pipe->screen;
+      struct pipe_buffer *vbuf;
+      float *map;
+      unsigned size;
+
+      size = num_verts * num_attribs * 4 * sizeof(float);
+
+      vbuf = pipe_buffer_create(screen,
+                                32,
+                                PIPE_BUFFER_USAGE_VERTEX, 
+                                size);
+      if(!vbuf)
+         goto error1;
+      
+      map = pipe_buffer_map(screen, vbuf, PIPE_BUFFER_USAGE_CPU_WRITE);
+      if (!map)
+         goto error2;
+      memcpy(map, vertices, size);
+      pipe_buffer_unmap(screen, vbuf);
+      
+      util_draw_vertex_buffer(pipe, vbuf, 0, prim, num_verts, num_attribs);
+      
+error2:
+      pipe_buffer_reference(screen, &vbuf, NULL);
+error1:
+      ;
+   }
+   
+   void
+   flush(unsigned flags = 0) {
+      struct pipe_fence_handle *fence = NULL; 
+      $self->pipe->flush($self->pipe, flags | PIPE_FLUSH_RENDER_CACHE, &fence);
+      /* TODO: allow asynchronous operation */ 
+      $self->pipe->winsys->fence_finish( $self->pipe->winsys, fence, 0 );
+      $self->pipe->winsys->fence_reference( $self->pipe->winsys, &fence, NULL );
+   }
+
+   /*
+    * Surface functions
+    */
+   
+   void surface_copy(int do_flip,
+                     struct pipe_surface *dest,
+                     unsigned destx, unsigned desty,
+                     struct pipe_surface *src,
+                     unsigned srcx, unsigned srcy,
+                     unsigned width, unsigned height) {
+      $self->pipe->surface_copy($self->pipe, do_flip, dest, destx, desty, src, srcx, srcy, width, height);
+   }
+
+   void surface_fill(struct pipe_surface *dst,
+                     unsigned x, unsigned y,
+                     unsigned width, unsigned height,
+                     unsigned value) {
+      $self->pipe->surface_fill($self->pipe, dst, x, y, width, height, value);
+   }
+
+   void surface_clear(struct pipe_surface *surface, unsigned value = 0) {
+      $self->pipe->clear($self->pipe, surface, value);
+   }
+
+};
diff --git a/src/gallium/state_trackers/python/p_device.i b/src/gallium/state_trackers/python/p_device.i
new file mode 100644
index 0000000000..84fd2e4349
--- /dev/null
+++ b/src/gallium/state_trackers/python/p_device.i
@@ -0,0 +1,130 @@
+ /**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * SWIG interface definion for Gallium types.
+ *
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+%nodefaultctor st_device;
+%nodefaultdtor st_device;
+
+
+struct st_device {
+};
+
+%newobject st_device::texture_create;
+%newobject st_device::context_create;
+%newobject st_device::buffer_create;
+
+%extend st_device {
+   
+   st_device(int hardware = 1) {
+      return st_device_create(hardware ? TRUE : FALSE);
+   }
+
+   ~st_device() {
+      st_device_destroy($self);
+   }
+   
+   const char * get_name( void ) {
+      return $self->screen->get_name($self->screen);
+   }
+
+   const char * get_vendor( void ) {
+      return $self->screen->get_vendor($self->screen);
+   }
+
+   /**
+    * Query an integer-valued capability/parameter/limit
+    * \param param  one of PIPE_CAP_x
+    */
+   int get_param( int param ) {
+      return $self->screen->get_param($self->screen, param);
+   }
+
+   /**
+    * Query a float-valued capability/parameter/limit
+    * \param param  one of PIPE_CAP_x
+    */
+   float get_paramf( int param ) {
+      return $self->screen->get_paramf($self->screen, param);
+   }
+
+   /**
+    * Check if the given pipe_format is supported as a texture or
+    * drawing surface.
+    * \param type  one of PIPE_TEXTURE, PIPE_SURFACE
+    */
+   int is_format_supported( enum pipe_format format, 
+                            enum pipe_texture_target target,
+                            unsigned tex_usage, 
+                            unsigned geom_flags ) {
+      return $self->screen->is_format_supported( $self->screen, 
+                                                 format, 
+                                                 target, 
+                                                 tex_usage, 
+                                                 geom_flags );
+   }
+
+   struct st_context *
+   context_create(void) {
+      return st_context_create($self);
+   }
+
+   struct pipe_texture * 
+   texture_create(
+         enum pipe_format format,
+         unsigned width,
+         unsigned height,
+         unsigned depth = 1,
+         unsigned last_level = 0,
+         enum pipe_texture_target target = PIPE_TEXTURE_2D,
+         unsigned tex_usage = 0
+      ) {
+      struct pipe_texture templat;
+      memset(&templat, 0, sizeof(templat));
+      templat.format = format;
+      pf_get_block(templat.format, &templat.block);
+      templat.width[0] = width;
+      templat.height[0] = height;
+      templat.depth[0] = depth;
+      templat.last_level = last_level;
+      templat.target = target;
+      templat.tex_usage = tex_usage;
+      return $self->screen->texture_create($self->screen, &templat);
+   }
+   
+   struct st_buffer *
+   buffer_create(unsigned size, unsigned alignment = 0, unsigned usage = 0) {
+      return st_buffer_create($self, alignment, usage, size);
+   }
+
+};
diff --git a/src/gallium/state_trackers/python/p_format.i b/src/gallium/state_trackers/python/p_format.i
new file mode 100644
index 0000000000..26fb12b387
--- /dev/null
+++ b/src/gallium/state_trackers/python/p_format.i
@@ -0,0 +1,162 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright (c) 2008 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/* 
+ * XXX: SWIG can't parse p_format.h, so we need to duplicate the relevant 
+ * declarations here 
+ */
+
+%{
+#include "pipe/p_format.h" 
+%}
+
+enum pipe_format {
+   PIPE_FORMAT_NONE,
+   PIPE_FORMAT_A8R8G8B8_UNORM,
+   PIPE_FORMAT_X8R8G8B8_UNORM,
+   PIPE_FORMAT_B8G8R8A8_UNORM,
+   PIPE_FORMAT_B8G8R8X8_UNORM,
+   PIPE_FORMAT_A1R5G5B5_UNORM,
+   PIPE_FORMAT_A4R4G4B4_UNORM,
+   PIPE_FORMAT_R5G6B5_UNORM,
+   PIPE_FORMAT_A2B10G10R10_UNORM,
+   PIPE_FORMAT_L8_UNORM,
+   PIPE_FORMAT_A8_UNORM,
+   PIPE_FORMAT_I8_UNORM,
+   PIPE_FORMAT_A8L8_UNORM,
+   PIPE_FORMAT_L16_UNORM,
+   PIPE_FORMAT_YCBCR,
+   PIPE_FORMAT_YCBCR_REV,
+   PIPE_FORMAT_Z16_UNORM,
+   PIPE_FORMAT_Z32_UNORM,
+   PIPE_FORMAT_Z32_FLOAT,
+   PIPE_FORMAT_S8Z24_UNORM,
+   PIPE_FORMAT_Z24S8_UNORM,
+   PIPE_FORMAT_X8Z24_UNORM,
+   PIPE_FORMAT_Z24X8_UNORM,
+   PIPE_FORMAT_S8_UNORM,
+   PIPE_FORMAT_R64_FLOAT,
+   PIPE_FORMAT_R64G64_FLOAT,
+   PIPE_FORMAT_R64G64B64_FLOAT,
+   PIPE_FORMAT_R64G64B64A64_FLOAT,
+   PIPE_FORMAT_R32_FLOAT,
+   PIPE_FORMAT_R32G32_FLOAT,
+   PIPE_FORMAT_R32G32B32_FLOAT,
+   PIPE_FORMAT_R32G32B32A32_FLOAT,
+   PIPE_FORMAT_R32_UNORM,
+   PIPE_FORMAT_R32G32_UNORM,
+   PIPE_FORMAT_R32G32B32_UNORM,
+   PIPE_FORMAT_R32G32B32A32_UNORM,
+   PIPE_FORMAT_R32_USCALED,
+   PIPE_FORMAT_R32G32_USCALED,
+   PIPE_FORMAT_R32G32B32_USCALED,
+   PIPE_FORMAT_R32G32B32A32_USCALED,
+   PIPE_FORMAT_R32_SNORM,
+   PIPE_FORMAT_R32G32_SNORM,
+   PIPE_FORMAT_R32G32B32_SNORM,
+   PIPE_FORMAT_R32G32B32A32_SNORM,
+   PIPE_FORMAT_R32_SSCALED,
+   PIPE_FORMAT_R32G32_SSCALED,
+   PIPE_FORMAT_R32G32B32_SSCALED,
+   PIPE_FORMAT_R32G32B32A32_SSCALED,
+   PIPE_FORMAT_R16_UNORM,
+   PIPE_FORMAT_R16G16_UNORM,
+   PIPE_FORMAT_R16G16B16_UNORM,
+   PIPE_FORMAT_R16G16B16A16_UNORM,
+   PIPE_FORMAT_R16_USCALED,
+   PIPE_FORMAT_R16G16_USCALED,
+   PIPE_FORMAT_R16G16B16_USCALED,
+   PIPE_FORMAT_R16G16B16A16_USCALED,
+   PIPE_FORMAT_R16_SNORM,
+   PIPE_FORMAT_R16G16_SNORM,
+   PIPE_FORMAT_R16G16B16_SNORM,
+   PIPE_FORMAT_R16G16B16A16_SNORM,
+   PIPE_FORMAT_R16_SSCALED,
+   PIPE_FORMAT_R16G16_SSCALED,
+   PIPE_FORMAT_R16G16B16_SSCALED,
+   PIPE_FORMAT_R16G16B16A16_SSCALED,
+   PIPE_FORMAT_R8_UNORM,
+   PIPE_FORMAT_R8G8_UNORM,
+   PIPE_FORMAT_R8G8B8_UNORM,
+   PIPE_FORMAT_R8G8B8A8_UNORM,
+   PIPE_FORMAT_R8G8B8X8_UNORM,
+   PIPE_FORMAT_R8_USCALED,
+   PIPE_FORMAT_R8G8_USCALED,
+   PIPE_FORMAT_R8G8B8_USCALED,
+   PIPE_FORMAT_R8G8B8A8_USCALED,
+   PIPE_FORMAT_R8G8B8X8_USCALED,
+   PIPE_FORMAT_R8_SNORM,
+   PIPE_FORMAT_R8G8_SNORM,
+   PIPE_FORMAT_R8G8B8_SNORM,
+   PIPE_FORMAT_R8G8B8A8_SNORM,
+   PIPE_FORMAT_R8G8B8X8_SNORM,
+   PIPE_FORMAT_B6G5R5_SNORM,
+   PIPE_FORMAT_A8B8G8R8_SNORM,
+   PIPE_FORMAT_X8B8G8R8_SNORM,
+   PIPE_FORMAT_R8_SSCALED,
+   PIPE_FORMAT_R8G8_SSCALED,
+   PIPE_FORMAT_R8G8B8_SSCALED,
+   PIPE_FORMAT_R8G8B8A8_SSCALED,
+   PIPE_FORMAT_R8G8B8X8_SSCALED,
+   PIPE_FORMAT_R32_FIXED,
+   PIPE_FORMAT_R32G32_FIXED,
+   PIPE_FORMAT_R32G32B32_FIXED,
+   PIPE_FORMAT_R32G32B32A32_FIXED,
+
+   PIPE_FORMAT_L8_SRGB,
+   PIPE_FORMAT_A8L8_SRGB,
+   PIPE_FORMAT_R8G8B8_SRGB,
+   PIPE_FORMAT_R8G8B8A8_SRGB,
+   PIPE_FORMAT_R8G8B8X8_SRGB,
+   PIPE_FORMAT_A8R8G8B8_SRGB,
+   PIPE_FORMAT_X8R8G8B8_SRGB,
+   PIPE_FORMAT_B8G8R8A8_SRGB,
+   PIPE_FORMAT_B8G8R8X8_SRGB,
+
+   PIPE_FORMAT_X8UB8UG8SR8S_NORM,
+   PIPE_FORMAT_B6UG5SR5S_NORM,
+
+   PIPE_FORMAT_DXT1_RGB,
+   PIPE_FORMAT_DXT1_RGBA,
+   PIPE_FORMAT_DXT3_RGBA,
+   PIPE_FORMAT_DXT5_RGBA,
+
+   PIPE_FORMAT_DXT1_SRGB,
+   PIPE_FORMAT_DXT1_SRGBA,
+   PIPE_FORMAT_DXT3_SRGBA,
+   PIPE_FORMAT_DXT5_SRGBA,
+};
+
+
+struct pipe_format_block
+{
+   unsigned size;
+   unsigned width;
+   unsigned height;
+};
+
diff --git a/src/gallium/state_trackers/python/p_state.i b/src/gallium/state_trackers/python/p_state.i
new file mode 100644
index 0000000000..7f5760b3b6
--- /dev/null
+++ b/src/gallium/state_trackers/python/p_state.i
@@ -0,0 +1,109 @@
+ /**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * SWIG interface definion for Gallium types.
+ *
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+%module gallium;
+
+%ignore winsys;
+%ignore pipe_vertex_buffer::buffer;
+
+%include "pipe/p_state.h";
+
+
+%array_class(struct pipe_stencil_state, StencilArray);
+
+
+%extend pipe_framebuffer_state {
+   
+   pipe_framebuffer_state(void) {
+      return CALLOC_STRUCT(pipe_framebuffer_state);
+   }
+   
+   ~pipe_framebuffer_state() {
+      unsigned index;
+      for(index = 0; index < PIPE_MAX_COLOR_BUFS; ++index)
+         pipe_surface_reference(&$self->cbufs[index], NULL);
+      pipe_surface_reference(&$self->zsbuf, NULL);
+      FREE($self);
+   }
+   
+   void
+   set_cbuf(unsigned index, struct pipe_surface *surface) {
+      pipe_surface_reference(&$self->cbufs[index], surface);
+   }
+   
+   void
+   set_zsbuf(struct pipe_surface *surface) {
+      pipe_surface_reference(&$self->zsbuf, surface);
+   }
+   
+};
+
+
+%extend pipe_shader_state {
+   
+   pipe_shader_state(const char *text, unsigned num_tokens = 1024) {
+      struct tgsi_token *tokens;
+      struct pipe_shader_state *shader;
+      
+      tokens = MALLOC(num_tokens * sizeof(struct tgsi_token));
+      if(!tokens)
+         goto error1;
+      
+      if(tgsi_text_translate(text, tokens, num_tokens ) != TRUE)
+         goto error2;
+      
+      shader = CALLOC_STRUCT(pipe_shader_state);
+      if(!shader)
+         goto error3;
+      
+      shader->tokens = tokens;
+      
+      return shader;
+      
+error3:
+error2:
+      FREE(tokens);
+error1:      
+      return NULL;
+   }
+   
+   ~pipe_shader_state() {
+      FREE((void*)$self->tokens);
+      FREE($self);
+   }
+
+   void dump(unsigned flags = 0) {
+      tgsi_dump($self->tokens, flags);
+   }
+}
diff --git a/src/gallium/state_trackers/python/p_texture.i b/src/gallium/state_trackers/python/p_texture.i
new file mode 100644
index 0000000000..08ba0ebe4d
--- /dev/null
+++ b/src/gallium/state_trackers/python/p_texture.i
@@ -0,0 +1,232 @@
+ /**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * SWIG interface definion for Gallium types.
+ *
+ * @author Jose Fonseca <jrfonseca@tungstengraphics.com>
+ */
+
+
+%nodefaultctor pipe_texture;
+%nodefaultctor pipe_surface;
+%nodefaultctor st_buffer;
+
+%nodefaultdtor pipe_texture;
+%nodefaultdtor pipe_surface;
+%nodefaultdtor st_buffer;
+
+%ignore pipe_texture::screen;
+
+%ignore pipe_surface::winsys;
+%immutable pipe_surface::texture;
+%immutable pipe_surface::buffer;
+
+%newobject pipe_texture::get_surface;
+
+
+%extend pipe_texture {
+   
+   ~pipe_texture() {
+      struct pipe_texture *ptr = $self;
+      pipe_texture_reference(&ptr, NULL);
+   }
+   
+   unsigned get_width(unsigned level=0) {
+      return $self->width[level];
+   }
+   
+   unsigned get_height(unsigned level=0) {
+      return $self->height[level];
+   }
+   
+   unsigned get_depth(unsigned level=0) {
+      return $self->depth[level];
+   }
+   
+   unsigned get_nblocksx(unsigned level=0) {
+      return $self->nblocksx[level];
+   }
+   
+   unsigned get_nblocksy(unsigned level=0) {
+      return $self->nblocksy[level];
+   }
+   
+   /** Get a surface which is a "view" into a texture */
+   struct pipe_surface *
+   get_surface(unsigned face=0, unsigned level=0, unsigned zslice=0, unsigned usage=0 )
+   {
+      struct pipe_screen *screen = $self->screen;
+      return screen->get_tex_surface(screen, $self, face, level, zslice, usage);
+   }
+   
+};
+
+
+%extend pipe_surface {
+   
+   ~pipe_surface() {
+      struct pipe_surface *ptr = $self;
+      pipe_surface_reference(&ptr, NULL);
+   }
+   
+   // gets mapped to pipe_surface_map automatically
+   void * map( unsigned flags );
+
+   // gets mapped to pipe_surface_unmap automatically
+   void unmap( void );
+
+   void
+   get_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, char *raw, unsigned stride) {
+      pipe_get_tile_raw($self, x, y, w, h, raw, stride);
+   }
+
+   void
+   put_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, const char *raw, unsigned stride) {
+      pipe_put_tile_raw($self, x, y, w, h, raw, stride);
+   }
+
+   void
+   get_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, float *rgba) {
+      pipe_get_tile_rgba($self, x, y, w, h, rgba);
+   }
+
+   void
+   put_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba) {
+      pipe_put_tile_rgba($self, x, y, w, h, rgba);
+   }
+
+   void
+   get_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, unsigned *z) {
+      pipe_get_tile_z($self, x, y, w, h, z);
+   }
+
+   void
+   put_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned *z) {
+      pipe_put_tile_z($self, x, y, w, h, z);
+   }
+   
+   void
+   sample_rgba(float *rgba) {
+      st_sample_surface($self, rgba);
+   }
+   
+   unsigned
+   compare_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba, float tol = 0.0) 
+   {
+      float *rgba2;
+      const float *p1;
+      const float *p2;
+      unsigned i, j, n;
+      
+      rgba2 = MALLOC(h*w*4*sizeof(float));
+      if(!rgba2)
+         return ~0;
+
+      pipe_get_tile_rgba($self, x, y, w, h, rgba2);
+
+      p1 = rgba;
+      p2 = rgba2;
+      n = 0;
+      for(i = h*w; i; --i) {
+         unsigned differs = 0;
+         for(j = 4; j; --j) {
+            float delta = *p2++ - *p1++;
+            if (delta < -tol || delta > tol)
+                differs = 1;
+         }
+         n += differs;
+      }
+      
+      FREE(rgba2);
+      
+      return n;
+   }
+
+};
+
+struct st_buffer {
+};
+
+%extend st_buffer {
+   
+   ~st_buffer() {
+      st_buffer_destroy($self);
+   }
+   
+   unsigned __len__(void) 
+   {
+      assert($self->buffer->refcount);
+      return $self->buffer->size;
+   }
+   
+   %cstring_output_allocate_size(char **STRING, int *LENGTH, free(*$1));
+   void read(char **STRING, int *LENGTH)
+   {
+      struct pipe_screen *screen = $self->st_dev->screen;
+      const char *map;
+      
+      assert($self->buffer->refcount);
+      
+      *LENGTH = $self->buffer->size;
+      *STRING = (char *) malloc($self->buffer->size);
+      if(!*STRING)
+         return;
+      
+      map = pipe_buffer_map(screen, $self->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+      if(map) {
+         memcpy(*STRING, map, $self->buffer->size);
+         pipe_buffer_unmap(screen, $self->buffer);
+      }
+   }
+   
+   %cstring_input_binary(const char *STRING, unsigned LENGTH);
+   void write(const char *STRING, unsigned LENGTH, unsigned offset = 0) 
+   {
+      struct pipe_screen *screen = $self->st_dev->screen;
+      char *map;
+      
+      assert($self->buffer->refcount);
+      
+      if(offset > $self->buffer->size) {
+         PyErr_SetString(PyExc_ValueError, "offset must be smaller than buffer size");
+         return;
+      }
+
+      if(offset + LENGTH > $self->buffer->size) {
+         PyErr_SetString(PyExc_ValueError, "data length must fit inside the buffer");
+         return;
+      }
+
+      map = pipe_buffer_map(screen, $self->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+      if(map) {
+         memcpy(map + offset, STRING, LENGTH);
+         pipe_buffer_unmap(screen, $self->buffer);
+      }
+   }
+};
diff --git a/src/gallium/state_trackers/python/retrace/README b/src/gallium/state_trackers/python/retrace/README
new file mode 100644
index 0000000000..822cd11404
--- /dev/null
+++ b/src/gallium/state_trackers/python/retrace/README
@@ -0,0 +1,17 @@
+This is an application written in python to replay the traces captured by the
+ trace pipe driver. 
+
+
+To use it follow the instructions in src/gallium/drivers/trace/README and
+src/gallium/state_trackers/python/README, and then do
+
+  python src/gallium/state_trackers/python/samples/retrace/interpreter.py filename.trace
+
+
+This is still work in progress:
+- not everything is captured/replayed
+  - surface/textures contents
+- any tiny error will result in a crash
+
+--
+Jose Fonseca <jrfonseca@tungstengraphics.com>
diff --git a/src/gallium/state_trackers/python/retrace/format.py b/src/gallium/state_trackers/python/retrace/format.py
new file mode 100755
index 0000000000..0bf6baf0b9
--- /dev/null
+++ b/src/gallium/state_trackers/python/retrace/format.py
@@ -0,0 +1,100 @@
+#!/usr/bin/env python
+##########################################################################
+#
+# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+# All Rights Reserved.
+#
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the
+# "Software"), to deal in the Software without restriction, including
+# without limitation the rights to use, copy, modify, merge, publish,
+# distribute, sub license, and/or sell copies of the Software, and to
+# permit persons to whom the Software is furnished to do so, subject to
+# the following conditions:
+#
+# The above copyright notice and this permission notice (including the
+# next paragraph) shall be included in all copies or substantial portions
+# of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+#
+##########################################################################
+
+
+class Formatter:
+    '''Plain formatter'''
+
+    def __init__(self, stream):
+        self.stream = stream
+
+    def text(self, text):
+        self.stream.write(text)
+
+    def newline(self):
+        self.text('\n')
+
+    def function(self, name):
+        self.text(name)
+
+    def variable(self, name):
+        self.text(name)
+
+    def literal(self, value):
+        self.text(str(value))
+
+    def address(self, addr):
+        self.text(str(addr))
+
+
+class AnsiFormatter(Formatter):
+    '''Formatter for plain-text files which outputs ANSI escape codes. See
+    http://en.wikipedia.org/wiki/ANSI_escape_code for more information
+    concerning ANSI escape codes.
+    '''
+
+    _csi = '\33['
+
+    _normal = '0m'
+    _bold = '1m'
+    _italic = '3m'
+    _red = '31m'
+    _green = '32m'
+    _blue = '34m'
+
+    def _escape(self, code):
+        self.text(self._csi + code)
+
+    def function(self, name):
+        self._escape(self._bold)
+        Formatter.function(self, name)
+        self._escape(self._normal)
+
+    def variable(self, name):
+        self._escape(self._italic)
+        Formatter.variable(self, name)
+        self._escape(self._normal)
+
+    def literal(self, value):
+        self._escape(self._blue)
+        Formatter.literal(self, value)
+        self._escape(self._normal)
+
+    def address(self, value):
+        self._escape(self._green)
+        Formatter.address(self, value)
+        self._escape(self._normal)
+
+
+
+def DefaultFormatter(stream):
+    if stream.isatty():
+        return AnsiFormatter(stream)
+    else:
+        return Formatter(stream)
+
diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py
new file mode 100755
index 0000000000..f418f80d7b
--- /dev/null
+++ b/src/gallium/state_trackers/python/retrace/interpreter.py
@@ -0,0 +1,578 @@
+#!/usr/bin/env python
+##########################################################################
+# 
+# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+# All Rights Reserved.
+# 
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the
+# "Software"), to deal in the Software without restriction, including
+# without limitation the rights to use, copy, modify, merge, publish,
+# distribute, sub license, and/or sell copies of the Software, and to
+# permit persons to whom the Software is furnished to do so, subject to
+# the following conditions:
+# 
+# The above copyright notice and this permission notice (including the
+# next paragraph) shall be included in all copies or substantial portions
+# of the Software.
+# 
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+# 
+##########################################################################
+
+
+import sys
+import struct
+
+import gallium
+import model
+import parser
+
+
+def make_image(surface):
+    pixels = gallium.FloatArray(surface.height*surface.width*4)
+    surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels)
+
+    import Image
+    outimage = Image.new(
+        mode='RGB',
+        size=(surface.width, surface.height),
+        color=(0,0,0))
+    outpixels = outimage.load()
+    for y in range(0, surface.height):
+        for x in range(0, surface.width):
+            offset = (y*surface.width + x)*4
+            r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)]
+            outpixels[x, y] = r, g, b
+    return outimage
+
+def save_image(filename, surface):
+    outimage = make_image(surface)
+    outimage.save(filename, "PNG")
+
+def show_image(surface):
+    outimage = make_image(surface)
+    
+    import Tkinter as tk
+    from PIL import Image, ImageTk
+    root = tk.Tk()
+    
+    root.title('background image')
+    
+    image1 = ImageTk.PhotoImage(outimage)
+    w = image1.width()
+    h = image1.height()
+    x = 100
+    y = 100
+    root.geometry("%dx%d+%d+%d" % (w, h, x, y))
+    panel1 = tk.Label(root, image=image1)
+    panel1.pack(side='top', fill='both', expand='yes')
+    panel1.image = image1
+    root.mainloop()
+
+
+
+
+class Struct:
+    """C-like struct"""
+
+    # A basic Python class can pass as a C-like structure
+    pass
+
+
+struct_factories = {
+    "pipe_blend_color": gallium.BlendColor,
+    "pipe_blend_state": gallium.Blend,
+    #"pipe_clip_state": gallium.Clip,
+    #"pipe_constant_buffer": gallium.ConstantBuffer,
+    "pipe_depth_state": gallium.Depth,
+    "pipe_stencil_state": gallium.Stencil,
+    "pipe_alpha_state": gallium.Alpha,
+    "pipe_depth_stencil_alpha_state": gallium.DepthStencilAlpha,
+    "pipe_format_block": gallium.FormatBlock,
+    #"pipe_framebuffer_state": gallium.Framebuffer,
+    "pipe_poly_stipple": gallium.PolyStipple,
+    "pipe_rasterizer_state": gallium.Rasterizer,
+    "pipe_sampler_state": gallium.Sampler,
+    "pipe_scissor_state": gallium.Scissor,
+    #"pipe_shader_state": gallium.Shader,
+    #"pipe_vertex_buffer": gallium.VertexBuffer,
+    "pipe_vertex_element": gallium.VertexElement,
+    "pipe_viewport_state": gallium.Viewport,
+    #"pipe_texture": gallium.Texture,
+}
+
+
+member_array_factories = {
+    "pipe_rasterizer_state": {"sprite_coord_mode": gallium.ByteArray},                          
+    "pipe_poly_stipple": {"stipple": gallium.UnsignedArray},                          
+    "pipe_viewport_state": {"scale": gallium.FloatArray, "translate": gallium.FloatArray},                          
+    #"pipe_clip_state": {"ucp": gallium.FloatArray},
+    "pipe_depth_stencil_alpha_state": {"stencil": gallium.StencilArray},
+    "pipe_blend_color": {"color": gallium.FloatArray},
+    "pipe_sampler_state": {"border_color": gallium.FloatArray},              
+}
+
+
+class Translator(model.Visitor):
+    """Translate model arguments into regular Python objects"""
+
+    def __init__(self, interpreter):
+        self.interpreter = interpreter
+        self.result = None
+
+    def visit(self, node):
+        self.result = None
+        node.visit(self)
+        return self.result
+        
+    def visit_literal(self, node):
+        self.result = node.value
+    
+    def visit_named_constant(self, node):
+        # lookup the named constant in the gallium module
+        self.result = getattr(gallium, node.name)
+    
+    def visit_array(self, node):
+        array = []
+        for element in node.elements:
+            array.append(self.visit(element))
+        self.result = array
+    
+    def visit_struct(self, node):
+        struct_factory = struct_factories.get(node.name, Struct)
+        struct = struct_factory()
+        for member_name, member_node in node.members:
+            member_value = self.visit(member_node)
+            try:
+                array_factory = member_array_factories[node.name][member_name]
+            except KeyError:
+                pass
+            else:
+                assert isinstance(member_value, list)
+                array = array_factory(len(member_value))
+                for i in range(len(member_value)):
+                    array[i] = member_value[i]
+                member_value = array
+            #print node.name, member_name, member_value
+            assert isinstance(struct, Struct) or hasattr(struct, member_name)
+            setattr(struct, member_name, member_value)
+        self.result = struct
+    
+    def visit_pointer(self, node):
+        self.result = self.interpreter.lookup_object(node.address)
+
+
+class Object:
+    
+    def __init__(self, interpreter, real):
+        self.interpreter = interpreter
+        self.real = real
+        
+
+class Global(Object):
+
+    def __init__(self, interpreter, real):
+        self.interpreter = interpreter
+        self.real = real
+        
+    def pipe_winsys_create(self):
+        return Winsys(self.interpreter, gallium.Device())
+
+    def pipe_screen_create(self, winsys):
+        return Screen(self.interpreter, winsys.real)
+    
+    def pipe_context_create(self, screen):
+        context = screen.real.context_create()
+        return Context(self.interpreter, context)
+
+    
+class Winsys(Object):
+    
+    def __init__(self, interpreter, real):
+        self.interpreter = interpreter
+        self.real = real
+
+    def get_name(self):
+        pass
+    
+    def user_buffer_create(self, data, size):
+        # We don't really care to distinguish between user and regular buffers
+        buffer = self.real.buffer_create(size, 
+                                         4, 
+                                         gallium.PIPE_BUFFER_USAGE_CPU_READ |
+                                         gallium.PIPE_BUFFER_USAGE_CPU_WRITE )
+        assert size == len(data)
+        buffer.write(data)
+        return buffer
+    
+    def buffer_create(self, alignment, usage, size):
+        return self.real.buffer_create(size, alignment, usage)
+    
+    def buffer_destroy(self, buffer):
+        pass
+    
+    def buffer_write(self, buffer, data, size):
+        assert size == len(data)
+        buffer.write(data)
+        
+    def fence_finish(self, fence, flags):
+        pass
+    
+    def fence_reference(self, dst, src):
+        pass
+    
+    def flush_frontbuffer(self, surface):
+        pass
+
+    def surface_alloc(self):
+        return None
+    
+    def surface_release(self, surface):
+        pass
+
+
+class Screen(Object):
+    
+    def destroy(self):
+        pass
+
+    def get_name(self):
+        pass
+    
+    def get_vendor(self):
+        pass
+    
+    def get_param(self, param):
+        pass
+    
+    def get_paramf(self, param):
+        pass
+    
+    def is_format_supported(self, format, target, tex_usage, geom_flags):
+        return self.real.is_format_supported(format, target, tex_usage, geom_flags)
+    
+    def texture_create(self, template):
+        return self.real.texture_create(
+            format = template.format,
+            width = template.width[0],
+            height = template.height[0],
+            depth = template.depth[0],
+            last_level = template.last_level,
+            target = template.target,
+            tex_usage = template.tex_usage,
+        )
+
+    def texture_destroy(self, texture):
+        self.interpreter.unregister_object(texture)
+
+    def texture_release(self, surface):
+        pass
+
+    def get_tex_surface(self, texture, face, level, zslice, usage):
+        return texture.get_surface(face, level, zslice, usage)
+    
+    def tex_surface_destroy(self, surface):
+        self.interpreter.unregister_object(surface)
+
+    def tex_surface_release(self, surface):
+        pass
+
+    def surface_write(self, surface, data, stride, size):
+        assert surface.nblocksy * stride == size 
+        surface.put_tile_raw(0, 0, surface.width, surface.height, data, stride)
+
+
+class Context(Object):
+    
+    def __init__(self, interpreter, real):
+        Object.__init__(self, interpreter, real)
+        self.cbufs = []
+        self.zsbuf = None
+        self.vbufs = []
+        self.velems = []
+
+    def destroy(self):
+        pass
+    
+    def create_blend_state(self, state):
+        return state
+
+    def bind_blend_state(self, state):
+        if state is not None:
+            self.real.set_blend(state)
+
+    def delete_blend_state(self, state):
+        pass
+    
+    def create_sampler_state(self, state):
+        return state
+
+    def delete_sampler_state(self, state):
+        pass
+
+    def bind_sampler_states(self, n, states):
+        for i in range(n):
+            self.real.set_sampler(i, states[i])
+        
+    def create_rasterizer_state(self, state):
+        return state
+
+    def bind_rasterizer_state(self, state):
+        if state is not None:
+            self.real.set_rasterizer(state)
+        
+    def delete_rasterizer_state(self, state):
+        pass
+    
+    def create_depth_stencil_alpha_state(self, state):
+        return state
+
+    def bind_depth_stencil_alpha_state(self, state):
+        if state is not None:
+            self.real.set_depth_stencil_alpha(state)
+            
+    def delete_depth_stencil_alpha_state(self, state):
+        pass
+
+    def create_fs_state(self, state):
+        tokens = str(state.tokens)
+        shader = gallium.Shader(tokens)
+        return shader
+
+    create_vs_state = create_fs_state
+    
+    def bind_fs_state(self, state):
+        self.real.set_fragment_shader(state)
+        
+    def bind_vs_state(self, state):
+        self.real.set_vertex_shader(state)
+
+    def delete_fs_state(self, state):
+        pass
+    
+    delete_vs_state = delete_fs_state
+    
+    def set_blend_color(self, state):
+        self.real.set_blend_color(state)
+
+    def set_clip_state(self, state):
+        _state = gallium.Clip()
+        _state.nr = state.nr
+        if state.nr:
+            # FIXME
+            ucp = gallium.FloatArray(gallium.PIPE_MAX_CLIP_PLANES*4)
+            for i in range(len(state.ucp)):
+                for j in range(len(state.ucp[i])):
+                    ucp[i*4 + j] = state.ucp[i][j]
+            _state.ucp = ucp
+        self.real.set_clip(_state)
+
+    def set_constant_buffer(self, shader, index, state):
+        if state is not None:
+            self.real.set_constant_buffer(shader, index, state.buffer)
+
+            if 1:
+                data = state.buffer.read()
+                format = '4f'
+                index = 0
+                for offset in range(0, len(data), struct.calcsize(format)):
+                    x, y, z, w = struct.unpack_from(format, data, offset)
+                    sys.stdout.write('\tCONST[%2u] = {%10.4f, %10.4f, %10.4f, %10.4f}\n' % (index, x, y, z, w))
+                    index += 1
+
+    def set_framebuffer_state(self, state):
+        _state = gallium.Framebuffer()
+        _state.width = state.width
+        _state.height = state.height
+        _state.num_cbufs = state.num_cbufs
+        for i in range(len(state.cbufs)):
+            _state.set_cbuf(i, state.cbufs[i])
+        _state.set_zsbuf(state.zsbuf)    
+        self.real.set_framebuffer(_state)
+        
+        self.cbufs = state.cbufs
+        self.zsbuf = state.zsbuf
+
+    def set_polygon_stipple(self, state):
+        self.real.set_polygon_stipple(state)
+
+    def set_scissor_state(self, state):
+        self.real.set_scissor(state)
+
+    def set_viewport_state(self, state):
+        self.real.set_viewport(state)
+
+    def set_sampler_textures(self, n, textures):
+        for i in range(n):
+            self.real.set_sampler_texture(i, textures[i])
+
+    def set_vertex_buffers(self, n, vbufs):
+        self.vbufs = vbufs[0:n]
+        for i in range(n):
+            vbuf = vbufs[i]
+            self.real.set_vertex_buffer(
+                i,
+                pitch = vbuf.pitch,
+                max_index = vbuf.max_index,
+                buffer_offset = vbuf.buffer_offset,
+                buffer = vbuf.buffer,
+            )
+
+    def set_vertex_elements(self, n, elements):
+        self.velems = elements[0:n]
+        for i in range(n):
+            self.real.set_vertex_element(i, elements[i])
+        self.real.set_vertex_elements(n)
+
+    def set_edgeflags(self, bitfield):
+        # FIXME
+        pass
+    
+    def dump_vertices(self, start, count):
+        for index in range(start, start + count):
+            if index >= start + 16:
+                sys.stdout.write('\t...\n')
+                break
+            sys.stdout.write('\t{\n')
+            for velem in self.velems:
+                vbuf = self.vbufs[velem.vertex_buffer_index]
+
+                offset = vbuf.buffer_offset + velem.src_offset + vbuf.pitch*index
+                format = {
+                    gallium.PIPE_FORMAT_R32_FLOAT: 'f',
+                    gallium.PIPE_FORMAT_R32G32_FLOAT: '2f',
+                    gallium.PIPE_FORMAT_R32G32B32_FLOAT: '3f',
+                    gallium.PIPE_FORMAT_R32G32B32A32_FLOAT: '4f',
+                    gallium.PIPE_FORMAT_B8G8R8A8_UNORM: '4B',
+                }[velem.src_format]
+
+                data = vbuf.buffer.read()
+                values = struct.unpack_from(format, data, offset)
+                sys.stdout.write('\t\t{' + ', '.join(map(str, values)) + '},\n')
+                assert len(values) == velem.nr_components
+            sys.stdout.write('\t},\n')
+
+    def dump_indices(self, ibuf, isize, start, count):
+        format = {
+            1: 'B',
+            2: 'H',
+            4: 'I',
+        }[isize]
+
+        assert struct.calcsize(format) == isize
+
+        data = ibuf.read()
+        maxindex, minindex = 0, 0xffffffff
+
+        sys.stdout.write('\t{\n')
+        for i in range(start, start + count):
+            if i >= start + 16:
+                sys.stdout.write('\t...\n')
+                break
+            offset = i*isize
+            index, = struct.unpack_from(format, data, offset)
+            sys.stdout.write('\t\t%u,\n' % index)
+            minindex = min(minindex, index)
+            maxindex = max(maxindex, index)
+        sys.stdout.write('\t},\n')
+
+        return minindex, maxindex
+
+    def draw_arrays(self, mode, start, count):
+        self.dump_vertices(start, count)
+            
+        self.real.draw_arrays(mode, start, count)
+    
+    def draw_elements(self, indexBuffer, indexSize, mode, start, count):
+        minindex, maxindex = self.dump_indices(indexBuffer, indexSize, start, count)
+        self.dump_vertices(minindex, maxindex - minindex)
+
+        self.real.draw_elements(indexBuffer, indexSize, mode, start, count)
+        
+    def draw_range_elements(self, indexBuffer, indexSize, minIndex, maxIndex, mode, start, count):
+        minindex, maxindex = self.dump_indices(indexBuffer, indexSize, start, count)
+        minindex = min(minindex, minIndex)
+        maxindex = min(maxindex, maxIndex)
+        self.dump_vertices(minindex, maxindex - minindex)
+
+        self.real.draw_range_elements(indexBuffer, indexSize, minIndex, maxIndex, mode, start, count)
+        
+    def flush(self, flags):
+        self.real.flush(flags)
+        if flags & gallium.PIPE_FLUSH_FRAME:
+            self._update()
+        return None
+
+    def clear(self, surface, value):
+        self.real.surface_clear(surface, value)
+        
+    def _update(self):
+        self.real.flush()
+    
+        if self.cbufs and self.cbufs[0]:
+            show_image(self.cbufs[0])
+    
+
+class Interpreter(parser.TraceDumper):
+    
+    ignore_calls = set((
+            ('pipe_screen', 'is_format_supported'),
+            ('pipe_screen', 'get_param'),
+            ('pipe_screen', 'get_paramf'),
+    ))
+
+    def __init__(self, stream):
+        parser.TraceDumper.__init__(self, stream)
+        self.objects = {}
+        self.result = None
+        self.globl = Global(self, None)
+
+    def register_object(self, address, object):
+        self.objects[address] = object
+        
+    def unregister_object(self, object):
+        # FIXME:
+        pass
+
+    def lookup_object(self, address):
+        return self.objects[address]
+    
+    def interpret(self, trace):
+        for call in trace.calls:
+            self.interpret_call(call)
+
+    def handle_call(self, call):
+
+        if (call.klass, call.method) in self.ignore_calls:
+            return
+
+        parser.TraceDumper.handle_call(self, call)
+        
+        args = [self.interpret_arg(arg) for name, arg in call.args] 
+        
+        if call.klass:
+            obj = args[0]
+            args = args[1:]
+        else:
+            obj = self.globl
+            
+        method = getattr(obj, call.method)
+        ret = method(*args)
+        
+        if call.ret and isinstance(call.ret, model.Pointer):
+            self.register_object(call.ret.address, ret)
+
+    def interpret_arg(self, node):
+        translator = Translator(self)
+        return translator.visit(node)
+    
+
+if __name__ == '__main__':
+    parser.main(Interpreter)
diff --git a/src/gallium/state_trackers/python/retrace/model.py b/src/gallium/state_trackers/python/retrace/model.py
new file mode 100755
index 0000000000..ae0f4327d7
--- /dev/null
+++ b/src/gallium/state_trackers/python/retrace/model.py
@@ -0,0 +1,211 @@
+#!/usr/bin/env python
+##########################################################################
+# 
+# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+# All Rights Reserved.
+# 
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the
+# "Software"), to deal in the Software without restriction, including
+# without limitation the rights to use, copy, modify, merge, publish,
+# distribute, sub license, and/or sell copies of the Software, and to
+# permit persons to whom the Software is furnished to do so, subject to
+# the following conditions:
+# 
+# The above copyright notice and this permission notice (including the
+# next paragraph) shall be included in all copies or substantial portions
+# of the Software.
+# 
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+# 
+##########################################################################
+
+
+'''Trace data model.'''
+
+
+import sys
+import string
+import format
+
+try:
+    from cStringIO import StringIO
+except ImportError:
+    from StringIO import StringIO
+
+
+class Node:
+    
+    def visit(self, visitor):
+        raise NotImplementedError
+
+    def __str__(self):
+        stream = StringIO()
+        formatter = format.DefaultFormatter(stream)
+        pretty_printer = PrettyPrinter(formatter)
+        self.visit(pretty_printer)
+        return stream.getvalue()
+
+
+class Literal(Node):
+    
+    def __init__(self, value):
+        self.value = value
+
+    def visit(self, visitor):
+        visitor.visit_literal(self)
+
+
+class NamedConstant(Node):
+    
+    def __init__(self, name):
+        self.name = name
+
+    def visit(self, visitor):
+        visitor.visit_named_constant(self)
+    
+
+class Array(Node):
+    
+    def __init__(self, elements):
+        self.elements = elements
+
+    def visit(self, visitor):
+        visitor.visit_array(self)
+
+
+class Struct(Node):
+    
+    def __init__(self, name, members):
+        self.name = name
+        self.members = members        
+
+    def visit(self, visitor):
+        visitor.visit_struct(self)
+
+        
+class Pointer(Node):
+    
+    def __init__(self, address):
+        self.address = address
+
+    def visit(self, visitor):
+        visitor.visit_pointer(self)
+
+
+class Call:
+    
+    def __init__(self, klass, method, args, ret):
+        self.klass = klass
+        self.method = method
+        self.args = args
+        self.ret = ret
+        
+    def visit(self, visitor):
+        visitor.visit_call(self)
+
+
+class Trace:
+    
+    def __init__(self, calls):
+        self.calls = calls
+        
+    def visit(self, visitor):
+        visitor.visit_trace(self)
+    
+    
+class Visitor:
+    
+    def visit_literal(self, node):
+        raise NotImplementedError
+    
+    def visit_named_constant(self, node):
+        raise NotImplementedError
+    
+    def visit_array(self, node):
+        raise NotImplementedError
+    
+    def visit_struct(self, node):
+        raise NotImplementedError
+    
+    def visit_pointer(self, node):
+        raise NotImplementedError
+    
+    def visit_call(self, node):
+        raise NotImplementedError
+    
+    def visit_trace(self, node):
+        raise NotImplementedError
+
+
+class PrettyPrinter:
+
+    def __init__(self, formatter):
+        self.formatter = formatter
+    
+    def visit_literal(self, node):
+        if isinstance(node.value, basestring):
+            if len(node.value) >= 4096 or node.value.strip(string.printable):
+                self.formatter.text('...')
+                return
+
+            self.formatter.literal('"' + node.value + '"')
+            return
+
+        self.formatter.literal(repr(node.value))
+    
+    def visit_named_constant(self, node):
+        self.formatter.literal(node.name)
+    
+    def visit_array(self, node):
+        self.formatter.text('{')
+        sep = ''
+        for value in node.elements:
+            self.formatter.text(sep)
+            value.visit(self) 
+            sep = ', '
+        self.formatter.text('}')
+    
+    def visit_struct(self, node):
+        self.formatter.text('{')
+        sep = ''
+        for name, value in node.members:
+            self.formatter.text(sep)
+            self.formatter.variable(name)
+            self.formatter.text(' = ')
+            value.visit(self) 
+            sep = ', '
+        self.formatter.text('}')
+    
+    def visit_pointer(self, node):
+        self.formatter.address(node.address)
+    
+    def visit_call(self, node):
+        if node.klass is not None:
+            self.formatter.function(node.klass + '::' + node.method)
+        else:
+            self.formatter.function(node.method)
+        self.formatter.text('(')
+        sep = ''
+        for name, value in node.args:
+            self.formatter.text(sep)
+            self.formatter.variable(name)
+            self.formatter.text(' = ')
+            value.visit(self) 
+            sep = ', '
+        self.formatter.text(')')
+        if node.ret is not None:
+            self.formatter.text(' = ')
+            node.ret.visit(self)
+    
+    def visit_trace(self, node):
+        for call in node.calls:
+            call.visit(self)
+            self.formatter.newline()
+
diff --git a/src/gallium/state_trackers/python/retrace/parser.py b/src/gallium/state_trackers/python/retrace/parser.py
new file mode 100755
index 0000000000..5205f2d8dd
--- /dev/null
+++ b/src/gallium/state_trackers/python/retrace/parser.py
@@ -0,0 +1,357 @@
+#!/usr/bin/env python
+##########################################################################
+# 
+# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+# All Rights Reserved.
+# 
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the
+# "Software"), to deal in the Software without restriction, including
+# without limitation the rights to use, copy, modify, merge, publish,
+# distribute, sub license, and/or sell copies of the Software, and to
+# permit persons to whom the Software is furnished to do so, subject to
+# the following conditions:
+# 
+# The above copyright notice and this permission notice (including the
+# next paragraph) shall be included in all copies or substantial portions
+# of the Software.
+# 
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+# 
+##########################################################################
+
+
+import sys
+import xml.parsers.expat
+import binascii
+
+from model import *
+
+
+ELEMENT_START, ELEMENT_END, CHARACTER_DATA, EOF = range(4)
+
+
+class XmlToken:
+
+    def __init__(self, type, name_or_data, attrs = None, line = None, column = None):
+        assert type in (ELEMENT_START, ELEMENT_END, CHARACTER_DATA, EOF)
+        self.type = type
+        self.name_or_data = name_or_data
+        self.attrs = attrs
+        self.line = line
+        self.column = column
+
+    def __str__(self):
+        if self.type == ELEMENT_START:
+            return '<' + self.name_or_data + ' ...>'
+        if self.type == ELEMENT_END:
+            return '</' + self.name_or_data + '>'
+        if self.type == CHARACTER_DATA:
+            return self.name_or_data
+        if self.type == EOF:
+            return 'end of file'
+        assert 0
+
+
+class XmlTokenizer:
+    """Expat based XML tokenizer."""
+
+    def __init__(self, fp, skip_ws = True):
+        self.fp = fp
+        self.tokens = []
+        self.index = 0
+        self.final = False
+        self.skip_ws = skip_ws
+        
+        self.character_pos = 0, 0
+        self.character_data = ''
+        
+        self.parser = xml.parsers.expat.ParserCreate()
+        self.parser.StartElementHandler  = self.handle_element_start
+        self.parser.EndElementHandler    = self.handle_element_end
+        self.parser.CharacterDataHandler = self.handle_character_data
+    
+    def handle_element_start(self, name, attributes):
+        self.finish_character_data()
+        line, column = self.pos()
+        token = XmlToken(ELEMENT_START, name, attributes, line, column)
+        self.tokens.append(token)
+    
+    def handle_element_end(self, name):
+        self.finish_character_data()
+        line, column = self.pos()
+        token = XmlToken(ELEMENT_END, name, None, line, column)
+        self.tokens.append(token)
+
+    def handle_character_data(self, data):
+        if not self.character_data:
+            self.character_pos = self.pos()
+        self.character_data += data
+    
+    def finish_character_data(self):
+        if self.character_data:
+            if not self.skip_ws or not self.character_data.isspace(): 
+                line, column = self.character_pos
+                token = XmlToken(CHARACTER_DATA, self.character_data, None, line, column)
+                self.tokens.append(token)
+            self.character_data = ''
+    
+    def next(self):
+        size = 16*1024
+        while self.index >= len(self.tokens) and not self.final:
+            self.tokens = []
+            self.index = 0
+            data = self.fp.read(size)
+            self.final = len(data) < size
+            data = data.rstrip('\0')
+            try:
+                self.parser.Parse(data, self.final)
+            except xml.parsers.expat.ExpatError, e:
+                #if e.code == xml.parsers.expat.errors.XML_ERROR_NO_ELEMENTS:
+                if e.code == 3:
+                    pass
+                else:
+                    raise e
+        if self.index >= len(self.tokens):
+            line, column = self.pos()
+            token = XmlToken(EOF, None, None, line, column)
+        else:
+            token = self.tokens[self.index]
+            self.index += 1
+        return token
+
+    def pos(self):
+        return self.parser.CurrentLineNumber, self.parser.CurrentColumnNumber
+
+
+class TokenMismatch(Exception):
+
+    def __init__(self, expected, found):
+        self.expected = expected
+        self.found = found
+
+    def __str__(self):
+        return '%u:%u: %s expected, %s found' % (self.found.line, self.found.column, str(self.expected), str(self.found))
+
+
+
+class XmlParser:
+    """Base XML document parser."""
+
+    def __init__(self, fp):
+        self.tokenizer = XmlTokenizer(fp)
+        self.consume()
+    
+    def consume(self):
+        self.token = self.tokenizer.next()
+
+    def match_element_start(self, name):
+        return self.token.type == ELEMENT_START and self.token.name_or_data == name
+    
+    def match_element_end(self, name):
+        return self.token.type == ELEMENT_END and self.token.name_or_data == name
+
+    def element_start(self, name):
+        while self.token.type == CHARACTER_DATA:
+            self.consume()
+        if self.token.type != ELEMENT_START:
+            raise TokenMismatch(XmlToken(ELEMENT_START, name), self.token)
+        if self.token.name_or_data != name:
+            raise TokenMismatch(XmlToken(ELEMENT_START, name), self.token)
+        attrs = self.token.attrs
+        self.consume()
+        return attrs
+    
+    def element_end(self, name):
+        while self.token.type == CHARACTER_DATA:
+            self.consume()
+        if self.token.type != ELEMENT_END:
+            raise TokenMismatch(XmlToken(ELEMENT_END, name), self.token)
+        if self.token.name_or_data != name:
+            raise TokenMismatch(XmlToken(ELEMENT_END, name), self.token)
+        self.consume()
+
+    def character_data(self, strip = True):
+        data = ''
+        while self.token.type == CHARACTER_DATA:
+            data += self.token.name_or_data
+            self.consume()
+        if strip:
+            data = data.strip()
+        return data
+
+
+class TraceParser(XmlParser):
+
+    def parse(self):
+        self.element_start('trace')
+        while self.token.type not in (ELEMENT_END, EOF):
+            call = self.parse_call()
+            self.handle_call(call)
+        if self.token.type != EOF:
+            self.element_end('trace')
+
+    def parse_call(self):
+        attrs = self.element_start('call')
+        klass = attrs['class']
+        method = attrs['method']
+        args = []
+        ret = None
+        while self.token.type == ELEMENT_START:
+            if self.token.name_or_data == 'arg':
+                arg = self.parse_arg()
+                args.append(arg)
+            elif self.token.name_or_data == 'ret':
+                ret = self.parse_ret()
+            elif self.token.name_or_data == 'call':
+                # ignore nested function calls
+                self.parse_call()
+            else:
+                raise TokenMismatch("<arg ...> or <ret ...>", self.token)
+        self.element_end('call')
+        
+        return Call(klass, method, args, ret)
+
+    def parse_arg(self):
+        attrs = self.element_start('arg')
+        name = attrs['name']
+        value = self.parse_value()
+        self.element_end('arg')
+
+        return name, value
+
+    def parse_ret(self):
+        attrs = self.element_start('ret')
+        value = self.parse_value()
+        self.element_end('ret')
+
+        return value
+
+    def parse_value(self):
+        expected_tokens = ('null', 'bool', 'int', 'uint', 'float', 'string', 'enum', 'array', 'struct', 'ptr', 'bytes')
+        if self.token.type == ELEMENT_START:
+            if self.token.name_or_data in expected_tokens:
+                method = getattr(self, 'parse_' +  self.token.name_or_data)
+                return method()
+        raise TokenMismatch(" or " .join(expected_tokens), self.token)
+
+    def parse_null(self):
+        self.element_start('null')
+        self.element_end('null')
+        return Literal(None)
+        
+    def parse_bool(self):
+        self.element_start('bool')
+        value = int(self.character_data())
+        self.element_end('bool')
+        return Literal(value)
+        
+    def parse_int(self):
+        self.element_start('int')
+        value = int(self.character_data())
+        self.element_end('int')
+        return Literal(value)
+        
+    def parse_uint(self):
+        self.element_start('uint')
+        value = int(self.character_data())
+        self.element_end('uint')
+        return Literal(value)
+        
+    def parse_float(self):
+        self.element_start('float')
+        value = float(self.character_data())
+        self.element_end('float')
+        return Literal(value)
+        
+    def parse_enum(self):
+        self.element_start('enum')
+        name = self.character_data()
+        self.element_end('enum')
+        return NamedConstant(name)
+        
+    def parse_string(self):
+        self.element_start('string')
+        value = self.character_data()
+        self.element_end('string')
+        return Literal(value)
+        
+    def parse_bytes(self):
+        self.element_start('bytes')
+        value = binascii.a2b_hex(self.character_data())
+        self.element_end('bytes')
+        return Literal(value)
+        
+    def parse_array(self):
+        self.element_start('array')
+        elems = []
+        while self.token.type != ELEMENT_END:
+            elems.append(self.parse_elem())
+        self.element_end('array')
+        return Array(elems)
+
+    def parse_elem(self):
+        self.element_start('elem')
+        value = self.parse_value()
+        self.element_end('elem')
+        return value
+
+    def parse_struct(self):
+        attrs = self.element_start('struct')
+        name = attrs['name']
+        members = []
+        while self.token.type != ELEMENT_END:
+            members.append(self.parse_member())
+        self.element_end('struct')
+        return Struct(name, members)
+
+    def parse_member(self):
+        attrs = self.element_start('member')
+        name = attrs['name']
+        value = self.parse_value()
+        self.element_end('member')
+
+        return name, value
+
+    def parse_ptr(self):
+        self.element_start('ptr')
+        address = self.character_data()
+        self.element_end('ptr')
+
+        return Pointer(address)
+
+    def handle_call(self, call):
+        pass
+    
+    
+class TraceDumper(TraceParser):
+    
+    def __init__(self, fp):
+        TraceParser.__init__(self, fp)
+        self.formatter = format.DefaultFormatter(sys.stdout)
+        self.pretty_printer = PrettyPrinter(self.formatter)
+
+    def handle_call(self, call):
+        call.visit(self.pretty_printer)
+        self.formatter.newline()
+        
+
+def main(ParserFactory):
+    for arg in sys.argv[1:]:
+        if arg.endswith('.gz'):
+            import gzip
+            stream = gzip.GzipFile(arg, 'rt')
+        else:
+            stream = open(arg, 'rt')
+        parser = ParserFactory(stream)
+        parser.parse()
+
+
+if __name__ == '__main__':
+    main(TraceDumper)
diff --git a/src/gallium/state_trackers/python/samples/tri.py b/src/gallium/state_trackers/python/samples/tri.py
new file mode 100644
index 0000000000..193479f7d6
--- /dev/null
+++ b/src/gallium/state_trackers/python/samples/tri.py
@@ -0,0 +1,229 @@
+#!/usr/bin/env python
+##########################################################################
+# 
+# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+# All Rights Reserved.
+# 
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the
+# "Software"), to deal in the Software without restriction, including
+# without limitation the rights to use, copy, modify, merge, publish,
+# distribute, sub license, and/or sell copies of the Software, and to
+# permit persons to whom the Software is furnished to do so, subject to
+# the following conditions:
+# 
+# The above copyright notice and this permission notice (including the
+# next paragraph) shall be included in all copies or substantial portions
+# of the Software.
+# 
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+# 
+##########################################################################
+
+
+from gallium import *
+
+
+def make_image(surface):
+    pixels = FloatArray(surface.height*surface.width*4)
+    surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels)
+
+    import Image
+    outimage = Image.new(
+        mode='RGB',
+        size=(surface.width, surface.height),
+        color=(0,0,0))
+    outpixels = outimage.load()
+    for y in range(0, surface.height):
+        for x in range(0, surface.width):
+            offset = (y*surface.width + x)*4
+            r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)]
+            outpixels[x, y] = r, g, b
+    return outimage
+
+def save_image(filename, surface):
+    outimage = make_image(surface)
+    outimage.save(filename, "PNG")
+
+def show_image(surface):
+    outimage = make_image(surface)
+    
+    import Tkinter as tk
+    from PIL import Image, ImageTk
+    root = tk.Tk()
+    
+    root.title('background image')
+    
+    image1 = ImageTk.PhotoImage(outimage)
+    w = image1.width()
+    h = image1.height()
+    x = 100
+    y = 100
+    root.geometry("%dx%d+%d+%d" % (w, h, x, y))
+    panel1 = tk.Label(root, image=image1)
+    panel1.pack(side='top', fill='both', expand='yes')
+    panel1.image = image1
+    root.mainloop()
+
+
+def test(dev):
+    ctx = dev.context_create()
+
+    width = 255
+    height = 255
+
+    # disabled blending/masking
+    blend = Blend()
+    blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE
+    blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE
+    blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO
+    blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO
+    blend.colormask = PIPE_MASK_RGBA
+    ctx.set_blend(blend)
+
+    # no-op depth/stencil/alpha
+    depth_stencil_alpha = DepthStencilAlpha()
+    ctx.set_depth_stencil_alpha(depth_stencil_alpha)
+
+    # rasterizer
+    rasterizer = Rasterizer()
+    rasterizer.front_winding = PIPE_WINDING_CW
+    rasterizer.cull_mode = PIPE_WINDING_NONE
+    rasterizer.bypass_clipping = 1
+    rasterizer.scissor = 1
+    #rasterizer.bypass_vs = 1
+    ctx.set_rasterizer(rasterizer)
+
+    # viewport (identity, we setup vertices in wincoords)
+    viewport = Viewport()
+    scale = FloatArray(4)
+    scale[0] = 1.0
+    scale[1] = 1.0
+    scale[2] = 1.0
+    scale[3] = 1.0
+    viewport.scale = scale
+    translate = FloatArray(4)
+    translate[0] = 0.0
+    translate[1] = 0.0
+    translate[2] = 0.0
+    translate[3] = 0.0
+    viewport.translate = translate
+    ctx.set_viewport(viewport)
+
+    # samplers
+    sampler = Sampler()
+    sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE
+    sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE
+    sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE
+    sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE
+    sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST
+    sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST
+    sampler.normalized_coords = 1
+    ctx.set_sampler(0, sampler)
+
+    # scissor
+    scissor = Scissor()
+    scissor.minx = 0
+    scissor.miny = 0
+    scissor.maxx = width
+    scissor.maxy = height
+    ctx.set_scissor(scissor)
+
+    clip = Clip()
+    clip.nr = 0
+    ctx.set_clip(clip)
+
+    # framebuffer
+    cbuf = dev.texture_create(
+        PIPE_FORMAT_X8R8G8B8_UNORM, 
+        width, height,
+        tex_usage=PIPE_TEXTURE_USAGE_DISPLAY_TARGET,
+    )
+    _cbuf = cbuf.get_surface(usage = PIPE_BUFFER_USAGE_GPU_READ|PIPE_BUFFER_USAGE_GPU_WRITE)
+    fb = Framebuffer()
+    fb.width = width
+    fb.height = height
+    fb.num_cbufs = 1
+    fb.set_cbuf(0, _cbuf)
+    ctx.set_framebuffer(fb)
+    _cbuf.clear_value = 0x00000000
+    ctx.surface_clear(_cbuf, _cbuf.clear_value)
+    del _cbuf
+    
+    # vertex shader
+    vs = Shader('''
+        VERT1.1
+        DCL IN[0], POSITION, CONSTANT
+        DCL IN[1], COLOR, CONSTANT
+        DCL OUT[0], POSITION, CONSTANT
+        DCL OUT[1], COLOR, CONSTANT
+        0:MOV OUT[0], IN[0]
+        1:MOV OUT[1], IN[1]
+        2:END
+    ''')
+    ctx.set_vertex_shader(vs)
+
+    # fragment shader
+    fs = Shader('''
+        FRAG1.1
+        DCL IN[0], COLOR, LINEAR
+        DCL OUT[0], COLOR, CONSTANT
+        0:MOV OUT[0], IN[0]
+        1:END
+    ''')
+    ctx.set_fragment_shader(fs)
+
+    nverts = 3
+    nattrs = 2
+    verts = FloatArray(nverts * nattrs * 4)
+
+    verts[ 0] = 128.0 # x1
+    verts[ 1] =  32.0 # y1
+    verts[ 2] =   0.0 # z1
+    verts[ 3] =   1.0 # w1
+    verts[ 4] =   1.0 # r1
+    verts[ 5] =   0.0 # g1
+    verts[ 6] =   0.0 # b1
+    verts[ 7] =   1.0 # a1
+    verts[ 8] =  32.0 # x2
+    verts[ 9] = 224.0 # y2
+    verts[10] =   0.0 # z2
+    verts[11] =   1.0 # w2
+    verts[12] =   0.0 # r2
+    verts[13] =   1.0 # g2
+    verts[14] =   0.0 # b2
+    verts[15] =   1.0 # a2
+    verts[16] = 224.0 # x3
+    verts[17] = 224.0 # y3
+    verts[18] =   0.0 # z3
+    verts[19] =   1.0 # w3
+    verts[20] =   0.0 # r3
+    verts[21] =   0.0 # g3
+    verts[22] =   1.0 # b3
+    verts[23] =   1.0 # a3
+
+    ctx.draw_vertices(PIPE_PRIM_TRIANGLES,
+                      nverts, 
+                      nattrs, 
+                      verts)
+
+    ctx.flush()
+    
+    show_image(cbuf.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE))
+    #save_image('tri.png', cbuf.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE))
+
+
+
+def main():
+    dev = Device()
+    test(dev)
+
+
+if __name__ == '__main__':
+    main()
diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c
new file mode 100644
index 0000000000..20dd8d269d
--- /dev/null
+++ b/src/gallium/state_trackers/python/st_device.c
@@ -0,0 +1,323 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "pipe/p_screen.h"
+#include "pipe/p_context.h"
+#include "pipe/p_shader_tokens.h"
+#include "pipe/p_inlines.h"
+#include "cso_cache/cso_context.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "util/u_simple_shaders.h"
+#include "trace/tr_screen.h"
+#include "trace/tr_context.h"
+
+#include "st_device.h"
+#include "st_winsys.h"
+
+
+static void
+st_device_really_destroy(struct st_device *st_dev) 
+{
+   if(st_dev->screen)
+      st_dev->screen->destroy(st_dev->screen);
+   
+   FREE(st_dev);
+}
+
+
+void
+st_device_destroy(struct st_device *st_dev) 
+{
+   if(!--st_dev->refcount)
+      st_device_really_destroy(st_dev);
+}
+
+
+static struct st_device *
+st_device_create_from_st_winsys(const struct st_winsys *st_ws) 
+{
+   struct st_device *st_dev;
+   
+   if(!st_ws->screen_create ||
+      !st_ws->context_create)
+      return NULL;
+   
+   st_dev = CALLOC_STRUCT(st_device);
+   if(!st_dev)
+      return NULL;
+   
+   st_dev->refcount = 1;
+   st_dev->st_ws = st_ws;
+   
+   st_dev->real_screen = st_ws->screen_create();
+   if(!st_dev->real_screen) {
+      st_device_destroy(st_dev);
+      return NULL;
+   }
+
+   st_dev->screen = trace_screen_create(st_dev->real_screen);
+   if(!st_dev->screen) {
+      st_device_destroy(st_dev);
+      return NULL;
+   }
+   
+   return st_dev;
+}
+
+
+struct st_device *
+st_device_create(boolean hardware) {
+   if(hardware)
+      return st_device_create_from_st_winsys(&st_hardpipe_winsys);
+   else
+      return st_device_create_from_st_winsys(&st_softpipe_winsys);
+}
+
+
+void
+st_context_destroy(struct st_context *st_ctx) 
+{
+   unsigned i;
+   
+   if(st_ctx) {
+      struct st_device *st_dev = st_ctx->st_dev;
+      
+      if(st_ctx->cso) {
+         cso_delete_vertex_shader(st_ctx->cso, st_ctx->vs);
+         cso_delete_fragment_shader(st_ctx->cso, st_ctx->fs);
+         
+         cso_destroy_context(st_ctx->cso);
+      }
+      
+      if(st_ctx->pipe)
+         st_ctx->pipe->destroy(st_ctx->pipe);
+      
+      for(i = 0; i < PIPE_MAX_SAMPLERS; ++i)
+         pipe_texture_reference(&st_ctx->sampler_textures[i], NULL);
+      pipe_texture_reference(&st_ctx->default_texture, NULL);
+
+      FREE(st_ctx);
+      
+      if(!--st_dev->refcount)
+         st_device_really_destroy(st_dev);
+   }
+}
+
+
+struct st_context *
+st_context_create(struct st_device *st_dev) 
+{
+   struct st_context *st_ctx;
+   
+   st_ctx = CALLOC_STRUCT(st_context);
+   if(!st_ctx)
+      return NULL;
+   
+   st_ctx->st_dev = st_dev;
+   ++st_dev->refcount;
+   
+   st_ctx->real_pipe = st_dev->st_ws->context_create(st_dev->real_screen);
+   if(!st_ctx->real_pipe) {
+      st_context_destroy(st_ctx);
+      return NULL;
+   }
+   
+   st_ctx->pipe = trace_context_create(st_dev->screen, st_ctx->real_pipe);
+   if(!st_ctx->pipe) {
+      st_context_destroy(st_ctx);
+      return NULL;
+   }
+
+   st_ctx->cso = cso_create_context(st_ctx->pipe);
+   if(!st_ctx->cso) {
+      st_context_destroy(st_ctx);
+      return NULL;
+   }
+   
+   /* disabled blending/masking */
+   {
+      struct pipe_blend_state blend;
+      memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.colormask = PIPE_MASK_RGBA;
+      cso_set_blend(st_ctx->cso, &blend);
+   }
+
+   /* no-op depth/stencil/alpha */
+   {
+      struct pipe_depth_stencil_alpha_state depthstencil;
+      memset(&depthstencil, 0, sizeof(depthstencil));
+      cso_set_depth_stencil_alpha(st_ctx->cso, &depthstencil);
+   }
+
+   /* rasterizer */
+   {
+      struct pipe_rasterizer_state rasterizer;
+      memset(&rasterizer, 0, sizeof(rasterizer));
+      rasterizer.front_winding = PIPE_WINDING_CW;
+      rasterizer.cull_mode = PIPE_WINDING_NONE;
+      rasterizer.bypass_clipping = 1;
+      /*rasterizer.bypass_vs = 1;*/
+      cso_set_rasterizer(st_ctx->cso, &rasterizer);
+   }
+
+   /* identity viewport */
+   {
+      struct pipe_viewport_state viewport;
+      viewport.scale[0] = 1.0;
+      viewport.scale[1] = 1.0;
+      viewport.scale[2] = 1.0;
+      viewport.scale[3] = 1.0;
+      viewport.translate[0] = 0.0;
+      viewport.translate[1] = 0.0;
+      viewport.translate[2] = 0.0;
+      viewport.translate[3] = 0.0;
+      cso_set_viewport(st_ctx->cso, &viewport);
+   }
+
+   /* samplers */
+   {
+      struct pipe_sampler_state sampler;
+      unsigned i;
+      memset(&sampler, 0, sizeof(sampler));
+      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST;
+      sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+      sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
+      sampler.normalized_coords = 1;
+      for (i = 0; i < PIPE_MAX_SAMPLERS; i++)
+         cso_single_sampler(st_ctx->cso, i, &sampler);
+      cso_single_sampler_done(st_ctx->cso);
+   }
+
+   /* default textures */
+   {
+      struct pipe_screen *screen = st_dev->screen;
+      struct pipe_texture templat;
+      struct pipe_surface *surface;
+      unsigned i;
+
+      memset( &templat, 0, sizeof( templat ) );
+      templat.target = PIPE_TEXTURE_2D;
+      templat.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+      templat.block.size = 4;
+      templat.block.width = 1;
+      templat.block.height = 1;
+      templat.width[0] = 1;
+      templat.height[0] = 1;
+      templat.depth[0] = 1;
+      templat.last_level = 0;
+   
+      st_ctx->default_texture = screen->texture_create( screen, &templat );
+      if(st_ctx->default_texture) {
+         surface = screen->get_tex_surface( screen, 
+                                            st_ctx->default_texture, 0, 0, 0,
+                                            PIPE_BUFFER_USAGE_CPU_WRITE );
+         if(surface) {
+            uint32_t *map;
+            map = (uint32_t *) pipe_surface_map(surface, PIPE_BUFFER_USAGE_CPU_WRITE );
+            if(map) {
+               *map = 0x00000000;
+               pipe_surface_unmap( surface );
+            }
+            pipe_surface_reference(&surface, NULL);
+         }
+      }
+   
+      for (i = 0; i < PIPE_MAX_SAMPLERS; i++)
+         pipe_texture_reference(&st_ctx->sampler_textures[i], st_ctx->default_texture);
+      
+      cso_set_sampler_textures(st_ctx->cso, PIPE_MAX_SAMPLERS, st_ctx->sampler_textures);
+   }
+   
+   /* vertex shader */
+   {
+      struct pipe_shader_state vert_shader;
+
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_GENERIC };
+      const uint semantic_indexes[] = { 0, 0 };
+      st_ctx->vs = util_make_vertex_passthrough_shader(st_ctx->pipe, 
+                                                       2, 
+                                                       semantic_names,
+                                                       semantic_indexes,
+                                                       &vert_shader);
+      cso_set_vertex_shader_handle(st_ctx->cso, st_ctx->vs);
+   }
+
+   /* fragment shader */
+   {
+      struct pipe_shader_state frag_shader;
+      st_ctx->fs = util_make_fragment_passthrough_shader(st_ctx->pipe, 
+                                                         &frag_shader);
+      cso_set_fragment_shader_handle(st_ctx->cso, st_ctx->fs);
+   }
+
+   return st_ctx;
+}
+
+
+void
+st_buffer_destroy(struct st_buffer *st_buf)
+{
+   if(st_buf) {
+      struct pipe_screen *screen = st_buf->st_dev->screen;
+      pipe_buffer_reference(screen, &st_buf->buffer, NULL);
+      FREE(st_buf);
+   }
+}
+
+
+struct st_buffer *
+st_buffer_create(struct st_device *st_dev,
+                 unsigned alignment, unsigned usage, unsigned size)
+{
+   struct pipe_screen *screen = st_dev->screen;
+   struct st_buffer *st_buf;
+   
+   st_buf = CALLOC_STRUCT(st_buffer);
+   if(!st_buf)
+      return NULL;
+
+   st_buf->st_dev = st_dev;
+   
+   st_buf->buffer = pipe_buffer_create(screen, alignment, usage, size);
+   if(!st_buf->buffer) {
+      st_buffer_destroy(st_buf);
+      return NULL;
+   }
+   
+   return st_buf;
+}
+
diff --git a/src/gallium/state_trackers/python/st_device.h b/src/gallium/state_trackers/python/st_device.h
new file mode 100644
index 0000000000..7cfe6de9f6
--- /dev/null
+++ b/src/gallium/state_trackers/python/st_device.h
@@ -0,0 +1,101 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_DEVICE_H_
+#define ST_DEVICE_H_
+
+
+#include "pipe/p_state.h"
+
+struct cso_context;
+struct pipe_screen;
+struct pipe_context;
+struct st_winsys; 
+
+
+struct st_buffer {
+   struct st_device *st_dev;
+   
+   struct pipe_buffer *buffer;
+};
+
+
+struct st_context {
+   struct st_device *st_dev;
+   
+   struct pipe_context *real_pipe;
+   struct pipe_context *pipe;
+   
+   struct cso_context *cso;
+   
+   void *vs;
+   void *fs;
+
+   struct pipe_texture *default_texture;
+   struct pipe_texture *sampler_textures[PIPE_MAX_SAMPLERS];
+   
+   unsigned num_vertex_buffers;
+   struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS];
+   
+   unsigned num_vertex_elements;
+   struct pipe_vertex_element vertex_elements[PIPE_MAX_ATTRIBS];
+};
+
+
+struct st_device {
+   const struct st_winsys *st_ws; 
+
+   struct pipe_screen *real_screen;
+   struct pipe_screen *screen;
+   
+   /* FIXME: we also need to refcount for textures and surfaces... */
+   unsigned refcount;
+};
+
+
+struct st_buffer *
+st_buffer_create(struct st_device *st_dev,
+                 unsigned alignment, unsigned usage, unsigned size);
+
+void
+st_buffer_destroy(struct st_buffer *st_buf);
+
+struct st_context *
+st_context_create(struct st_device *st_dev);
+
+void
+st_context_destroy(struct st_context *st_ctx);
+
+struct st_device *
+st_device_create(boolean hardware);
+
+void
+st_device_destroy(struct st_device *st_dev);
+
+
+#endif /* ST_DEVICE_H_ */
diff --git a/src/gallium/state_trackers/python/st_hardpipe_winsys.c b/src/gallium/state_trackers/python/st_hardpipe_winsys.c
new file mode 100644
index 0000000000..8b33c70fd7
--- /dev/null
+++ b/src/gallium/state_trackers/python/st_hardpipe_winsys.c
@@ -0,0 +1,62 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Stub for hardware pipe driver support.
+ */
+
+
+#include "pipe/p_compiler.h"
+
+#include "st_winsys.h"
+
+
+/* XXX: Force init_gallium symbol to be linked */
+extern void init_gallium(void);
+void (*force_init_gallium_linkage)(void) = &init_gallium;
+
+
+static struct pipe_screen *
+st_hardpipe_screen_create(void)
+{
+   return st_softpipe_winsys.screen_create();
+}
+
+
+static struct pipe_context *
+st_hardpipe_context_create(struct pipe_screen *screen)
+{
+   return st_softpipe_winsys.context_create(screen);
+}
+
+
+const struct st_winsys st_hardpipe_winsys = {
+   &st_hardpipe_screen_create,
+   &st_hardpipe_context_create
+};
diff --git a/src/gallium/state_trackers/python/st_sample.c b/src/gallium/state_trackers/python/st_sample.c
new file mode 100644
index 0000000000..7765df3c4a
--- /dev/null
+++ b/src/gallium/state_trackers/python/st_sample.c
@@ -0,0 +1,549 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "st_sample.h"
+
+
+/**
+ * Use our own pseudo random generator to ensure consistent runs among
+ * multiple runs and platforms.
+ * 
+ * @sa http://en.wikipedia.org/wiki/Linear_congruential_generator
+ */
+static uint32_t st_random(void) {
+   static uint64_t seed = UINT64_C(0xbb9a063afb0a739d);
+
+   seed = UINT64_C(134775813) * seed + UINT64_C(1);
+   
+   return (uint16_t)(seed >> 32); 
+}
+
+
+/**
+ * We don't want to include the patent-encumbered DXT code here, so instead
+ * we store several uncompressed/compressed data pairs for hardware testing
+ * purposes. 
+ */
+struct dxt_data
+{
+   uint8_t rgba[16*4];
+   uint8_t raw[16];
+};
+
+
+static const struct dxt_data 
+dxt1_rgb_data[] = {
+   {
+      {
+         0x99, 0xb0, 0x8e, 0xff,
+         0x5d, 0x62, 0x89, 0xff,
+         0x99, 0xb0, 0x8e, 0xff,
+         0x99, 0xb0, 0x8e, 0xff,
+         0xd6, 0xff, 0x94, 0xff,
+         0x5d, 0x62, 0x89, 0xff,
+         0x99, 0xb0, 0x8e, 0xff,
+         0xd6, 0xff, 0x94, 0xff,
+         0x5d, 0x62, 0x89, 0xff,
+         0x5d, 0x62, 0x89, 0xff,
+         0x99, 0xb0, 0x8e, 0xff,
+         0x21, 0x14, 0x84, 0xff,
+         0x5d, 0x62, 0x89, 0xff,
+         0x21, 0x14, 0x84, 0xff,
+         0x21, 0x14, 0x84, 0xff,
+         0x99, 0xb0, 0x8e, 0xff
+      },
+      {0xf2, 0xd7, 0xb0, 0x20, 0xae, 0x2c, 0x6f, 0x97}
+   },
+   {
+      {
+         0xb5, 0xcf, 0x9c, 0xff,
+         0x83, 0x8c, 0x8b, 0xff,
+         0x21, 0x08, 0x6b, 0xff,
+         0x83, 0x8c, 0x8b, 0xff,
+         0x52, 0x4a, 0x7b, 0xff,
+         0x83, 0x8c, 0x8b, 0xff,
+         0x83, 0x8c, 0x8b, 0xff,
+         0xb5, 0xcf, 0x9c, 0xff,
+         0x21, 0x08, 0x6b, 0xff,
+         0xb5, 0xcf, 0x9c, 0xff,
+         0x83, 0x8c, 0x8b, 0xff,
+         0x52, 0x4a, 0x7b, 0xff,
+         0xb5, 0xcf, 0x9c, 0xff,
+         0x83, 0x8c, 0x8b, 0xff,
+         0x52, 0x4a, 0x7b, 0xff,
+         0x83, 0x8c, 0x8b, 0xff
+      },
+      {0x73, 0xb6, 0x4d, 0x20, 0x98, 0x2b, 0xe1, 0xb8}
+   },
+   {
+      {
+         0x00, 0x2c, 0xff, 0xff,
+         0x94, 0x8d, 0x7b, 0xff,
+         0x4a, 0x5c, 0xbd, 0xff,
+         0x4a, 0x5c, 0xbd, 0xff,
+         0x4a, 0x5c, 0xbd, 0xff,
+         0x94, 0x8d, 0x7b, 0xff,
+         0x94, 0x8d, 0x7b, 0xff,
+         0x94, 0x8d, 0x7b, 0xff,
+         0xde, 0xbe, 0x39, 0xff,
+         0x94, 0x8d, 0x7b, 0xff,
+         0xde, 0xbe, 0x39, 0xff,
+         0xde, 0xbe, 0x39, 0xff,
+         0xde, 0xbe, 0x39, 0xff,
+         0xde, 0xbe, 0x39, 0xff,
+         0xde, 0xbe, 0x39, 0xff,
+         0x94, 0x8d, 0x7b, 0xff
+      },
+      {0xe7, 0xdd, 0x7f, 0x01, 0xf9, 0xab, 0x08, 0x80}
+   },
+   {
+      {
+         0x6b, 0x24, 0x21, 0xff,
+         0x7b, 0x4f, 0x5d, 0xff,
+         0x7b, 0x4f, 0x5d, 0xff,
+         0x8b, 0x7a, 0x99, 0xff,
+         0x7b, 0x4f, 0x5d, 0xff,
+         0x7b, 0x4f, 0x5d, 0xff,
+         0x6b, 0x24, 0x21, 0xff,
+         0x8b, 0x7a, 0x99, 0xff,
+         0x9c, 0xa6, 0xd6, 0xff,
+         0x6b, 0x24, 0x21, 0xff,
+         0x7b, 0x4f, 0x5d, 0xff,
+         0x8b, 0x7a, 0x99, 0xff,
+         0x6b, 0x24, 0x21, 0xff,
+         0x8b, 0x7a, 0x99, 0xff,
+         0x7b, 0x4f, 0x5d, 0xff,
+         0x9c, 0xa6, 0xd6, 0xff
+      },
+      {0x3a, 0x9d, 0x24, 0x69, 0xbd, 0x9f, 0xb4, 0x39}
+   }
+};
+
+
+static const struct dxt_data 
+dxt1_rgba_data[] = {
+   {
+      {
+         0x00, 0x00, 0x00, 0x00,
+         0x4e, 0xaa, 0x90, 0xff,
+         0x4e, 0xaa, 0x90, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0x4e, 0xaa, 0x90, 0xff,
+         0x29, 0xff, 0xff, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0x4e, 0xaa, 0x90, 0xff,
+         0x73, 0x55, 0x21, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x4e, 0xaa, 0x90, 0xff,
+         0x4e, 0xaa, 0x90, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x4e, 0xaa, 0x90, 0xff
+      },
+      {0xff, 0x2f, 0xa4, 0x72, 0xeb, 0xb2, 0xbd, 0xbe}
+   },
+   {
+      {
+         0xb5, 0xe3, 0x63, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0x6b, 0x24, 0x84, 0xff,
+         0xb5, 0xe3, 0x63, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0xb5, 0xe3, 0x63, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x6b, 0x24, 0x84, 0xff,
+         0x6b, 0x24, 0x84, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0xb5, 0xe3, 0x63, 0xff,
+         0x90, 0x83, 0x73, 0xff,
+         0xb5, 0xe3, 0x63, 0xff
+      },
+      {0x30, 0x69, 0x0c, 0xb7, 0x4d, 0xf7, 0x0f, 0x67}
+   },
+   {
+      {
+         0x00, 0x00, 0x00, 0x00,
+         0xc6, 0x86, 0x8c, 0xff,
+         0xc6, 0x86, 0x8c, 0xff,
+         0x21, 0x65, 0x42, 0xff,
+         0x21, 0x65, 0x42, 0xff,
+         0x21, 0x65, 0x42, 0xff,
+         0x21, 0x65, 0x42, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x21, 0x65, 0x42, 0xff,
+         0xc6, 0x86, 0x8c, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0xc6, 0x86, 0x8c, 0xff
+      },
+      {0x28, 0x23, 0x31, 0xc4, 0x17, 0xc0, 0xd3, 0x7f}
+   },
+   {
+      {
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0xc6, 0xe3, 0x9c, 0xff,
+         0x7b, 0x1c, 0x52, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x00, 0x00, 0x00, 0x00,
+         0x7b, 0x1c, 0x52, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0x7b, 0x1c, 0x52, 0xff,
+         0xa0, 0x7f, 0x77, 0xff,
+         0xc6, 0xe3, 0x9c, 0xff,
+         0x00, 0x00, 0x00, 0x00,
+         0xa0, 0x7f, 0x77, 0xff
+      },
+      {0xea, 0x78, 0x13, 0xc7, 0x7f, 0xfc, 0x33, 0xb6}
+   },
+};
+
+
+static const struct dxt_data 
+dxt3_rgba_data[] = {
+   {
+      {
+         0x6d, 0xc6, 0x96, 0x77,
+         0x6d, 0xc6, 0x96, 0xee,
+         0x6d, 0xc6, 0x96, 0xaa,
+         0x8c, 0xff, 0xb5, 0x44,
+         0x6d, 0xc6, 0x96, 0xff,
+         0x6d, 0xc6, 0x96, 0x88,
+         0x31, 0x55, 0x5a, 0x66,
+         0x6d, 0xc6, 0x96, 0x99,
+         0x31, 0x55, 0x5a, 0xbb,
+         0x31, 0x55, 0x5a, 0x55,
+         0x31, 0x55, 0x5a, 0x11,
+         0x6d, 0xc6, 0x96, 0xcc,
+         0x6d, 0xc6, 0x96, 0xcc,
+         0x6d, 0xc6, 0x96, 0x11,
+         0x31, 0x55, 0x5a, 0x44,
+         0x31, 0x55, 0x5a, 0x88
+      },
+      {0xe7, 0x4a, 0x8f, 0x96, 0x5b, 0xc1, 0x1c, 0x84, 0xf6, 0x8f, 0xab, 0x32, 0x2a, 0x9a, 0x95, 0x5a}
+   },
+   {
+      {
+         0xad, 0xeb, 0x73, 0x99,
+         0x97, 0xaa, 0x86, 0x66,
+         0x6b, 0x28, 0xad, 0x99,
+         0xad, 0xeb, 0x73, 0x99,
+         0x6b, 0x28, 0xad, 0x22,
+         0xad, 0xeb, 0x73, 0xff,
+         0x97, 0xaa, 0x86, 0x55,
+         0x6b, 0x28, 0xad, 0x55,
+         0x6b, 0x28, 0xad, 0x44,
+         0xad, 0xeb, 0x73, 0x33,
+         0x6b, 0x28, 0xad, 0xee,
+         0x6b, 0x28, 0xad, 0x99,
+         0x97, 0xaa, 0x86, 0x66,
+         0xad, 0xeb, 0x73, 0xbb,
+         0x97, 0xaa, 0x86, 0x99,
+         0xad, 0xeb, 0x73, 0xbb
+      },
+      {0x69, 0x99, 0xf2, 0x55, 0x34, 0x9e, 0xb6, 0xb9, 0x4e, 0xaf, 0x55, 0x69, 0x18, 0x61, 0x51, 0x22}
+   },
+   {
+      {
+         0x63, 0xd7, 0xd6, 0x00,
+         0x57, 0x62, 0x5d, 0xdd,
+         0x57, 0x62, 0x5d, 0xcc,
+         0x57, 0x62, 0x5d, 0xbb,
+         0x52, 0x28, 0x21, 0xaa,
+         0x57, 0x62, 0x5d, 0xcc,
+         0x57, 0x62, 0x5d, 0xcc,
+         0x57, 0x62, 0x5d, 0x66,
+         0x57, 0x62, 0x5d, 0x22,
+         0x57, 0x62, 0x5d, 0xdd,
+         0x63, 0xd7, 0xd6, 0xee,
+         0x57, 0x62, 0x5d, 0x33,
+         0x63, 0xd7, 0xd6, 0x55,
+         0x52, 0x28, 0x21, 0x55,
+         0x57, 0x62, 0x5d, 0x11,
+         0x5d, 0x9c, 0x99, 0xee
+      },
+      {0xd0, 0xbc, 0xca, 0x6c, 0xd2, 0x3e, 0x55, 0xe1, 0xba, 0x66, 0x44, 0x51, 0xfc, 0xfd, 0xcf, 0xb4}
+   },
+   {
+      {
+         0x94, 0x6f, 0x60, 0x22,
+         0x94, 0x6f, 0x60, 0x22,
+         0xc5, 0xab, 0x76, 0x11,
+         0xc5, 0xab, 0x76, 0xee,
+         0x63, 0x34, 0x4a, 0xdd,
+         0x63, 0x34, 0x4a, 0x33,
+         0x94, 0x6f, 0x60, 0x77,
+         0xf7, 0xe7, 0x8c, 0x00,
+         0x94, 0x6f, 0x60, 0x33,
+         0x63, 0x34, 0x4a, 0xaa,
+         0x94, 0x6f, 0x60, 0x77,
+         0x63, 0x34, 0x4a, 0xcc,
+         0x94, 0x6f, 0x60, 0xaa,
+         0xf7, 0xe7, 0x8c, 0x99,
+         0x63, 0x34, 0x4a, 0x44,
+         0xc5, 0xab, 0x76, 0xaa
+      },
+      {0x22, 0xe1, 0x3d, 0x07, 0xa3, 0xc7, 0x9a, 0xa4, 0x31, 0xf7, 0xa9, 0x61, 0xaf, 0x35, 0x77, 0x93}
+   },
+};
+
+
+static const struct dxt_data 
+dxt5_rgba_data[] = {
+   {
+      {
+         0x6d, 0xc6, 0x96, 0x74,
+         0x6d, 0xc6, 0x96, 0xf8,
+         0x6d, 0xc6, 0x96, 0xb6,
+         0x8c, 0xff, 0xb5, 0x53,
+         0x6d, 0xc6, 0x96, 0xf8,
+         0x6d, 0xc6, 0x96, 0x95,
+         0x31, 0x55, 0x5a, 0x53,
+         0x6d, 0xc6, 0x96, 0x95,
+         0x31, 0x55, 0x5a, 0xb6,
+         0x31, 0x55, 0x5a, 0x53,
+         0x31, 0x55, 0x5a, 0x11,
+         0x6d, 0xc6, 0x96, 0xd7,
+         0x6d, 0xc6, 0x96, 0xb6,
+         0x6d, 0xc6, 0x96, 0x11,
+         0x31, 0x55, 0x5a, 0x32,
+         0x31, 0x55, 0x5a, 0x95
+      },
+      {0xf8, 0x11, 0xc5, 0x0c, 0x9a, 0x73, 0xb4, 0x9c, 0xf6, 0x8f, 0xab, 0x32, 0x2a, 0x9a, 0x95, 0x5a}
+   },
+   {
+      {
+         0xad, 0xeb, 0x73, 0xa1,
+         0x97, 0xaa, 0x86, 0x65,
+         0x6b, 0x28, 0xad, 0xa1,
+         0xad, 0xeb, 0x73, 0xa1,
+         0x6b, 0x28, 0xad, 0x2a,
+         0xad, 0xeb, 0x73, 0xfb,
+         0x97, 0xaa, 0x86, 0x47,
+         0x6b, 0x28, 0xad, 0x65,
+         0x6b, 0x28, 0xad, 0x47,
+         0xad, 0xeb, 0x73, 0x47,
+         0x6b, 0x28, 0xad, 0xdd,
+         0x6b, 0x28, 0xad, 0xa1,
+         0x97, 0xaa, 0x86, 0x65,
+         0xad, 0xeb, 0x73, 0xbf,
+         0x97, 0xaa, 0x86, 0xa1,
+         0xad, 0xeb, 0x73, 0xbf
+      },
+      {0xfb, 0x2a, 0x34, 0x19, 0xdc, 0xbf, 0xe8, 0x71, 0x4e, 0xaf, 0x55, 0x69, 0x18, 0x61, 0x51, 0x22}
+   },
+   {
+      {
+         0x63, 0xd7, 0xd6, 0x00,
+         0x57, 0x62, 0x5d, 0xf5,
+         0x57, 0x62, 0x5d, 0xd2,
+         0x57, 0x62, 0x5d, 0xaf,
+         0x52, 0x28, 0x21, 0xaf,
+         0x57, 0x62, 0x5d, 0xd2,
+         0x57, 0x62, 0x5d, 0xd2,
+         0x57, 0x62, 0x5d, 0x69,
+         0x57, 0x62, 0x5d, 0x23,
+         0x57, 0x62, 0x5d, 0xd2,
+         0x63, 0xd7, 0xd6, 0xf5,
+         0x57, 0x62, 0x5d, 0x46,
+         0x63, 0xd7, 0xd6, 0x46,
+         0x52, 0x28, 0x21, 0x69,
+         0x57, 0x62, 0x5d, 0x23,
+         0x5d, 0x9c, 0x99, 0xf5
+      },
+      {0xf5, 0x00, 0x81, 0x36, 0xa9, 0x17, 0xec, 0x1e, 0xba, 0x66, 0x44, 0x51, 0xfc, 0xfd, 0xcf, 0xb4}
+   },
+   {
+      {
+         0x94, 0x6f, 0x60, 0x25,
+         0x94, 0x6f, 0x60, 0x25,
+         0xc5, 0xab, 0x76, 0x05,
+         0xc5, 0xab, 0x76, 0xe8,
+         0x63, 0x34, 0x4a, 0xe8,
+         0x63, 0x34, 0x4a, 0x25,
+         0x94, 0x6f, 0x60, 0x86,
+         0xf7, 0xe7, 0x8c, 0x05,
+         0x94, 0x6f, 0x60, 0x25,
+         0x63, 0x34, 0x4a, 0xa7,
+         0x94, 0x6f, 0x60, 0x66,
+         0x63, 0x34, 0x4a, 0xc7,
+         0x94, 0x6f, 0x60, 0xa7,
+         0xf7, 0xe7, 0x8c, 0xa7,
+         0x63, 0x34, 0x4a, 0x45,
+         0xc5, 0xab, 0x76, 0xa7
+      },
+      {0xe8, 0x05, 0x7f, 0x80, 0x33, 0x5f, 0xb5, 0x79, 0x31, 0xf7, 0xa9, 0x61, 0xaf, 0x35, 0x77, 0x93}
+   },
+};
+
+
+static INLINE void 
+st_sample_dxt_pixel_block(enum pipe_format format, 
+                          const struct pipe_format_block *block,
+                          uint8_t *raw,
+                          float *rgba, unsigned rgba_stride, 
+                          unsigned w, unsigned h)
+{
+   const struct dxt_data *data;
+   unsigned n;
+   unsigned i;
+   unsigned x, y, ch;
+   
+   switch(format) {
+   case PIPE_FORMAT_DXT1_RGB:
+      data = dxt1_rgb_data;
+      n = sizeof(dxt1_rgb_data)/sizeof(dxt1_rgb_data[0]);
+      break;
+   case PIPE_FORMAT_DXT1_RGBA:
+      data = dxt1_rgba_data;
+      n = sizeof(dxt1_rgba_data)/sizeof(dxt1_rgba_data[0]);
+      break;
+   case PIPE_FORMAT_DXT3_RGBA:
+      data = dxt3_rgba_data;
+      n = sizeof(dxt3_rgba_data)/sizeof(dxt3_rgba_data[0]);
+      break;
+   case PIPE_FORMAT_DXT5_RGBA:
+      data = dxt5_rgba_data;
+      n = sizeof(dxt5_rgba_data)/sizeof(dxt5_rgba_data[0]);
+      break;
+   default:
+      assert(0);
+   }
+   
+   i = st_random() % n;
+   
+   for(y = 0; y < h; ++y)
+      for(x = 0; x < w; ++x)
+         for(ch = 0; ch < 4; ++ch)
+            rgba[y*rgba_stride + x*4 + ch] = (float)(data[i].rgba[y*4*4 + x*4 + ch])/255.0f;
+   
+   memcpy(raw, data[i].raw, block->size);
+}
+
+
+static INLINE void 
+st_sample_generic_pixel_block(enum pipe_format format, 
+                              const struct pipe_format_block *block,
+                              uint8_t *raw,
+                              float *rgba, unsigned rgba_stride,
+                              unsigned w, unsigned h)
+{
+   unsigned i;
+   unsigned x, y, ch;
+   
+   for(i = 0; i < block->size; ++i)
+      raw[i] = (uint8_t)st_random();
+   
+   
+   pipe_tile_raw_to_rgba(format,
+                         raw,
+                         w, h,
+                         rgba, rgba_stride);
+ 
+   if(format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV) {
+      for(y = 0; y < h; ++y) {
+         for(x = 0; x < w; ++x) {
+            for(ch = 0; ch < 4; ++ch) {
+               unsigned offset = y*rgba_stride + x*4 + ch;
+               rgba[offset] = CLAMP(rgba[offset], 0.0f, 1.0f);
+            }
+         }
+      }
+   }
+}
+
+
+/**
+ * Randomly sample pixels.
+ */
+void 
+st_sample_pixel_block(enum pipe_format format,
+                      const struct pipe_format_block *block,
+                      void *raw,
+                      float *rgba, unsigned rgba_stride,
+                      unsigned w, unsigned h)
+{
+   switch(format) {
+   case PIPE_FORMAT_DXT1_RGB:
+   case PIPE_FORMAT_DXT1_RGBA:
+   case PIPE_FORMAT_DXT3_RGBA:
+   case PIPE_FORMAT_DXT5_RGBA:
+      st_sample_dxt_pixel_block(format, block, raw, rgba, rgba_stride, w, h);
+      break;
+
+   default:
+      st_sample_generic_pixel_block(format, block, raw, rgba, rgba_stride, w, h);
+      break;
+   }
+}
+
+
+void
+st_sample_surface(struct pipe_surface *surface, float *rgba) 
+{
+   const struct pipe_format_block *block = &surface->block;
+   unsigned rgba_stride = surface->width*4;
+   void *raw;
+   unsigned x, y;
+
+   raw = pipe_surface_map(surface, PIPE_BUFFER_USAGE_CPU_READ);
+   if(!raw)
+      return;
+
+   for (y = 0; y < surface->nblocksy; ++y) {
+      for(x = 0; x < surface->nblocksx; ++x) {
+         st_sample_pixel_block(surface->format,
+                               block,
+                               (uint8_t*)raw + y*surface->stride + x*block->size, 
+                               rgba + y*block->height*rgba_stride + x*block->width*4,
+                               rgba_stride,
+                               MIN2(block->width, surface->width - x*block->width), 
+                               MIN2(block->height, surface->height - y*block->height));
+       }
+   }
+   
+   pipe_surface_unmap(surface);
+}
diff --git a/src/gallium/state_trackers/python/st_sample.h b/src/gallium/state_trackers/python/st_sample.h
new file mode 100644
index 0000000000..ff04a12613
--- /dev/null
+++ b/src/gallium/state_trackers/python/st_sample.h
@@ -0,0 +1,47 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_SAMPLE_H_
+#define ST_SAMPLE_H_
+
+
+#include "pipe/p_format.h"
+
+
+void 
+st_sample_pixel_block(enum pipe_format format,
+                      const struct pipe_format_block *block,
+                      void *raw,
+                      float *rgba, unsigned rgba_stride,
+                      unsigned w, unsigned h);
+
+void
+st_sample_surface(struct pipe_surface *surface, float *rgba);
+
+
+#endif /* ST_SAMPLE_H_ */
diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c
new file mode 100644
index 0000000000..4d798df99b
--- /dev/null
+++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c
@@ -0,0 +1,271 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * Softpipe support. 
+ * 
+ * @author Keith Whitwell
+ * @author Brian Paul
+ * @author Jose Fonseca
+ */
+
+
+#include "pipe/internal/p_winsys_screen.h"/* port to just p_screen */
+#include "pipe/p_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "softpipe/sp_winsys.h"
+#include "st_winsys.h"
+
+
+struct st_softpipe_buffer
+{
+   struct pipe_buffer base;
+   boolean userBuffer;  /** Is this a user-space buffer? */
+   void *data;
+   void *mapped;
+};
+
+
+/** Cast wrapper */
+static INLINE struct st_softpipe_buffer *
+st_softpipe_buffer( struct pipe_buffer *buf )
+{
+   return (struct st_softpipe_buffer *)buf;
+}
+
+
+static void *
+st_softpipe_buffer_map(struct pipe_winsys *winsys, 
+                       struct pipe_buffer *buf,
+                       unsigned flags)
+{
+   struct st_softpipe_buffer *st_softpipe_buf = st_softpipe_buffer(buf);
+   st_softpipe_buf->mapped = st_softpipe_buf->data;
+   return st_softpipe_buf->mapped;
+}
+
+
+static void
+st_softpipe_buffer_unmap(struct pipe_winsys *winsys, 
+                         struct pipe_buffer *buf)
+{
+   struct st_softpipe_buffer *st_softpipe_buf = st_softpipe_buffer(buf);
+   st_softpipe_buf->mapped = NULL;
+}
+
+
+static void
+st_softpipe_buffer_destroy(struct pipe_winsys *winsys,
+                           struct pipe_buffer *buf)
+{
+   struct st_softpipe_buffer *oldBuf = st_softpipe_buffer(buf);
+
+   if (oldBuf->data) {
+      if (!oldBuf->userBuffer)
+         align_free(oldBuf->data);
+
+      oldBuf->data = NULL;
+   }
+
+   FREE(oldBuf);
+}
+
+
+static void
+st_softpipe_flush_frontbuffer(struct pipe_winsys *winsys,
+                              struct pipe_surface *surf,
+                              void *context_private)
+{
+}
+
+
+
+static const char *
+st_softpipe_get_name(struct pipe_winsys *winsys)
+{
+   return "softpipe";
+}
+
+
+static struct pipe_buffer *
+st_softpipe_buffer_create(struct pipe_winsys *winsys, 
+                          unsigned alignment, 
+                          unsigned usage,
+                          unsigned size)
+{
+   struct st_softpipe_buffer *buffer = CALLOC_STRUCT(st_softpipe_buffer);
+
+   buffer->base.refcount = 1;
+   buffer->base.alignment = alignment;
+   buffer->base.usage = usage;
+   buffer->base.size = size;
+
+   buffer->data = align_malloc(size, alignment);
+
+   return &buffer->base;
+}
+
+
+/**
+ * Create buffer which wraps user-space data.
+ */
+static struct pipe_buffer *
+st_softpipe_user_buffer_create(struct pipe_winsys *winsys, 
+                               void *ptr, 
+                               unsigned bytes)
+{
+   struct st_softpipe_buffer *buffer;
+   
+   buffer = CALLOC_STRUCT(st_softpipe_buffer);
+   if(!buffer)
+      return NULL;
+   
+   buffer->base.refcount = 1;
+   buffer->base.size = bytes;
+   buffer->userBuffer = TRUE;
+   buffer->data = ptr;
+
+   return &buffer->base;
+}
+
+
+/**
+ * Round n up to next multiple.
+ */
+static INLINE unsigned
+round_up(unsigned n, unsigned multiple)
+{
+   return (n + multiple - 1) & ~(multiple - 1);
+}
+
+
+static struct pipe_buffer *
+st_softpipe_surface_buffer_create(struct pipe_winsys *winsys,
+                                  unsigned width, unsigned height,
+                                  enum pipe_format format,
+                                  unsigned usage,
+                                  unsigned *stride)
+{
+   const unsigned alignment = 64;
+   struct pipe_format_block block;
+   unsigned nblocksx, nblocksy;
+
+   pf_get_block(format, &block);
+   nblocksx = pf_get_nblocksx(&block, width);
+   nblocksy = pf_get_nblocksy(&block, height);
+   *stride = round_up(nblocksx * block.size, alignment);
+
+   return winsys->buffer_create(winsys, alignment,
+                                usage,
+                                *stride * nblocksy);
+}
+
+
+static void
+st_softpipe_fence_reference(struct pipe_winsys *winsys, 
+                            struct pipe_fence_handle **ptr,
+                            struct pipe_fence_handle *fence)
+{
+}
+
+
+static int
+st_softpipe_fence_signalled(struct pipe_winsys *winsys, 
+                            struct pipe_fence_handle *fence,
+                            unsigned flag)
+{
+   return 0;
+}
+
+
+static int
+st_softpipe_fence_finish(struct pipe_winsys *winsys, 
+                         struct pipe_fence_handle *fence,
+                         unsigned flag)
+{
+   return 0;
+}
+
+
+static void
+st_softpipe_destroy(struct pipe_winsys *winsys)
+{
+   FREE(winsys);
+}
+
+
+static struct pipe_screen *
+st_softpipe_screen_create(void)
+{
+   static struct pipe_winsys *winsys;
+   struct pipe_screen *screen;
+
+   winsys = CALLOC_STRUCT(pipe_winsys);
+   if(!winsys)
+      return NULL;
+
+   winsys->destroy = st_softpipe_destroy;
+   
+   winsys->buffer_create = st_softpipe_buffer_create;
+   winsys->user_buffer_create = st_softpipe_user_buffer_create;
+   winsys->buffer_map = st_softpipe_buffer_map;
+   winsys->buffer_unmap = st_softpipe_buffer_unmap;
+   winsys->buffer_destroy = st_softpipe_buffer_destroy;
+
+   winsys->surface_buffer_create = st_softpipe_surface_buffer_create;
+
+   winsys->fence_reference = st_softpipe_fence_reference;
+   winsys->fence_signalled = st_softpipe_fence_signalled;
+   winsys->fence_finish = st_softpipe_fence_finish;
+
+   winsys->flush_frontbuffer = st_softpipe_flush_frontbuffer;
+   winsys->get_name = st_softpipe_get_name;
+
+   screen = softpipe_create_screen(winsys);
+   if(!screen)
+      st_softpipe_destroy(winsys);
+
+   return screen;
+}
+
+
+static struct pipe_context *
+st_softpipe_context_create(struct pipe_screen *screen)
+{
+   return softpipe_create(screen, screen->winsys, NULL);
+}
+
+
+const struct st_winsys st_softpipe_winsys = {
+   &st_softpipe_screen_create,
+   &st_softpipe_context_create,
+};
diff --git a/src/gallium/state_trackers/python/st_winsys.h b/src/gallium/state_trackers/python/st_winsys.h
new file mode 100644
index 0000000000..b8cb612d86
--- /dev/null
+++ b/src/gallium/state_trackers/python/st_winsys.h
@@ -0,0 +1,52 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_WINSYS_H_
+#define ST_WINSYS_H_
+
+
+struct pipe_screen;
+struct pipe_context;
+
+
+struct st_winsys 
+{
+   struct pipe_screen *
+   (*screen_create)(void);
+
+   struct pipe_context *
+   (*context_create)(struct pipe_screen *screen);
+};
+
+
+extern const struct st_winsys st_softpipe_winsys;
+
+extern const struct st_winsys st_hardpipe_winsys;
+
+
+#endif /* ST_WINSYS_H_ */
diff --git a/src/gallium/state_trackers/python/tests/base.py b/src/gallium/state_trackers/python/tests/base.py
new file mode 100644
index 0000000000..8477aa5fc9
--- /dev/null
+++ b/src/gallium/state_trackers/python/tests/base.py
@@ -0,0 +1,193 @@
+#!/usr/bin/env python
+##########################################################################
+# 
+# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+# All Rights Reserved.
+# 
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the
+# "Software"), to deal in the Software without restriction, including
+# without limitation the rights to use, copy, modify, merge, publish,
+# distribute, sub license, and/or sell copies of the Software, and to
+# permit persons to whom the Software is furnished to do so, subject to
+# the following conditions:
+# 
+# The above copyright notice and this permission notice (including the
+# next paragraph) shall be included in all copies or substantial portions
+# of the Software.
+# 
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+# 
+##########################################################################
+
+
+"""Base classes for tests.
+
+Loosely inspired on Python's unittest module.
+"""
+
+
+from gallium import *
+
+
+# Enumerate all pixel formats
+formats = {}
+for name, value in globals().items():
+    if name.startswith("PIPE_FORMAT_") and isinstance(value, int):
+        formats[value] = name
+
+
+def make_image(width, height, rgba):
+    import Image
+    outimage = Image.new(
+        mode='RGB',
+        size=(width, height),
+        color=(0,0,0))
+    outpixels = outimage.load()
+    for y in range(0, height):
+        for x in range(0, width):
+            offset = (y*width + x)*4
+            r, g, b, a = [int(min(max(rgba[offset + ch], 0.0), 1.0)*255) for ch in range(4)]
+            outpixels[x, y] = r, g, b
+    return outimage
+
+def save_image(width, height, rgba, filename):
+    outimage = make_image(width, height, rgba)
+    outimage.save(filename, "PNG")
+
+def show_image(width, height, **rgbas):
+    import Tkinter as tk
+    from PIL import Image, ImageTk
+
+    root = tk.Tk()
+    
+    x = 64
+    y = 64
+    
+    labels = rgbas.keys()
+    labels.sort() 
+    for i in range(len(labels)):
+        label = labels[i]
+        outimage = make_image(width, height, rgbas[label])
+        
+        if i:
+            window = tk.Toplevel(root)
+        else:
+            window = root    
+        window.title(label)
+        image1 = ImageTk.PhotoImage(outimage)
+        w = image1.width()
+        h = image1.height()
+        window.geometry("%dx%d+%d+%d" % (w, h, x, y))
+        panel1 = tk.Label(window, image=image1)
+        panel1.pack(side='top', fill='both', expand='yes')
+        panel1.image = image1
+        x += w + 2
+    
+    root.mainloop()
+
+
+class TestFailure(Exception):
+
+    pass
+
+class TestSkip(Exception):
+    
+    pass
+
+
+class Test:
+
+    def __init__(self):
+        pass
+
+    def _run(self, result):
+        raise NotImplementedError
+    
+    def run(self):
+        result = TestResult()
+        self._run(result)
+        result.summary()
+
+
+class TestCase(Test):
+    
+    def __init__(self, dev, **kargs):
+        Test.__init__(self)
+        self.dev = dev
+        self.__dict__.update(kargs)
+
+    def description(self):
+        raise NotImplementedError
+        
+    def test(self):
+        raise NotImplementedError
+    
+    def _run(self, result):
+        result.test_start(self)
+        try:
+            self.test()
+        except KeyboardInterrupt:
+            raise
+        except TestSkip:
+            result.test_skipped(self)
+        except TestFailure:
+            result.test_failed(self)
+        else:
+            result.test_passed(self)
+
+
+class TestSuite(Test):
+    
+    def __init__(self, tests = None):
+        Test.__init__(self)
+        if tests is None:
+            self.tests = []
+        else:
+            self.tests = tests
+
+    def add_test(self, test):
+        self.tests.append(test) 
+    
+    def _run(self, result):
+        for test in self.tests:
+            test._run(result)
+
+
+class TestResult:
+    
+    def __init__(self):
+        self.tests = 0
+        self.passed = 0
+        self.skipped = 0
+        self.failed = 0
+        self.failed_descriptions = []
+    
+    def test_start(self, test):
+        self.tests += 1
+        print "Running %s..." % test.description()
+    
+    def test_passed(self, test):
+        self.passed += 1
+        print "PASS"
+            
+    def test_skipped(self, test):
+        self.skipped += 1
+        print "SKIP"
+        
+    def test_failed(self, test):
+        self.failed += 1
+        self.failed_descriptions.append(test.description())
+        print "FAIL"
+
+    def summary(self):
+        print "%u tests, %u passed, %u skipped, %u failed" % (self.tests, self.passed, self.skipped, self.failed)
+        for description in self.failed_descriptions:
+            print "  %s" % description
+ 
\ No newline at end of file
diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py
new file mode 100644
index 0000000000..880a61306c
--- /dev/null
+++ b/src/gallium/state_trackers/python/tests/texture.py
@@ -0,0 +1,397 @@
+#!/usr/bin/env python
+##########################################################################
+# 
+# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+# All Rights Reserved.
+# 
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the
+# "Software"), to deal in the Software without restriction, including
+# without limitation the rights to use, copy, modify, merge, publish,
+# distribute, sub license, and/or sell copies of the Software, and to
+# permit persons to whom the Software is furnished to do so, subject to
+# the following conditions:
+# 
+# The above copyright notice and this permission notice (including the
+# next paragraph) shall be included in all copies or substantial portions
+# of the Software.
+# 
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+# 
+##########################################################################
+
+
+import sys
+from gallium import *
+from base import *
+
+
+def lods(*dims):
+    size = max(dims)
+    lods = 0
+    while size:
+        lods += 1
+        size >>= 1
+    return lods
+
+
+def minify(dims, level = 1):
+    return [max(dim>>level, 1) for dim in dims]
+
+
+def tex_coords(texture, face, level, zslice):
+    st = [ 
+        [0.0, 0.0], 
+        [1.0, 0.0], 
+        [1.0, 1.0], 
+        [0.0, 1.0],
+    ] 
+    
+    if texture.target == PIPE_TEXTURE_2D:
+        return [[s, t, 0.0] for s, t in st]
+    elif texture.target == PIPE_TEXTURE_3D:
+        depth = texture.get_depth(level)
+        if depth > 1:
+            r = float(zslice)/float(depth - 1)
+        else:
+            r = 0.0
+        return [[s, t, r] for s, t in st]
+    elif texture.target == PIPE_TEXTURE_CUBE:
+        result = []
+        for s, t in st:
+            # See http://developer.nvidia.com/object/cube_map_ogl_tutorial.html
+            sc = 2.0*s - 1.0
+            tc = 2.0*t - 1.0
+            if face == PIPE_TEX_FACE_POS_X:
+                rx = 1.0
+                ry = -tc
+                rz = -sc
+            if face == PIPE_TEX_FACE_NEG_X:
+                rx = -1.0
+                ry = -tc
+                rz = sc
+            if face == PIPE_TEX_FACE_POS_Y:
+                rx = sc
+                ry = 1.0
+                rz = tc
+            if face == PIPE_TEX_FACE_NEG_Y:
+                rx = sc
+                ry = -1.0
+                rz = -tc
+            if face == PIPE_TEX_FACE_POS_Z:
+                rx = sc
+                ry = -tc
+                rz = 1.0
+            if face == PIPE_TEX_FACE_NEG_Z:
+                rx = -sc
+                ry = -tc
+                rz = -1.0
+            result.append([rx, ry, rz])
+        return result
+
+def is_pot(n):
+    return n & (n - 1) == 0
+      
+                
+class TextureTest(TestCase):
+    
+    def description(self):
+        target = {
+            PIPE_TEXTURE_1D: "1d", 
+            PIPE_TEXTURE_2D: "2d", 
+            PIPE_TEXTURE_3D: "3d", 
+            PIPE_TEXTURE_CUBE: "cube",
+        }[self.target]
+        format = formats[self.format]
+        if self.target == PIPE_TEXTURE_CUBE:
+            face = {
+                PIPE_TEX_FACE_POS_X: "+x",
+                PIPE_TEX_FACE_NEG_X: "-x",
+                PIPE_TEX_FACE_POS_Y: "+y",
+                PIPE_TEX_FACE_NEG_Y: "-y", 
+                PIPE_TEX_FACE_POS_Z: "+z", 
+                PIPE_TEX_FACE_NEG_Z: "-z",
+            }[self.face]
+        else:
+            face = ""
+        return "%s %s %ux%ux%u last_level=%u face=%s level=%u zslice=%u" % (
+            target, format, 
+            self.width, self.height, self.depth, self.last_level, 
+            face, self.level, self.zslice, 
+        )
+    
+    def test(self):
+        dev = self.dev
+        
+        target = self.target
+        format = self.format
+        width = self.width
+        height = self.height
+        depth = self.depth
+        last_level = self.last_level
+        face = self.face
+        level = self.level
+        zslice = self.zslice
+        
+        tex_usage = PIPE_TEXTURE_USAGE_SAMPLER
+        geom_flags = 0
+        if width != height:
+            geom_flags |= PIPE_TEXTURE_GEOM_NON_SQUARE
+        if not is_pot(width) or not is_pot(height) or not is_pot(depth):
+            geom_flags |= PIPE_TEXTURE_GEOM_NON_POWER_OF_TWO
+        
+        if not dev.is_format_supported(format, target, tex_usage, geom_flags):
+            raise TestSkip
+        
+        ctx = self.dev.context_create()
+    
+        # disabled blending/masking
+        blend = Blend()
+        blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE
+        blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE
+        blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO
+        blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO
+        blend.colormask = PIPE_MASK_RGBA
+        ctx.set_blend(blend)
+    
+        # no-op depth/stencil/alpha
+        depth_stencil_alpha = DepthStencilAlpha()
+        ctx.set_depth_stencil_alpha(depth_stencil_alpha)
+    
+        # rasterizer
+        rasterizer = Rasterizer()
+        rasterizer.front_winding = PIPE_WINDING_CW
+        rasterizer.cull_mode = PIPE_WINDING_NONE
+        rasterizer.bypass_clipping = 1
+        #rasterizer.bypass_vs = 1
+        ctx.set_rasterizer(rasterizer)
+    
+        # viewport (identity, we setup vertices in wincoords)
+        viewport = Viewport()
+        scale = FloatArray(4)
+        scale[0] = 1.0
+        scale[1] = 1.0
+        scale[2] = 1.0
+        scale[3] = 1.0
+        viewport.scale = scale
+        translate = FloatArray(4)
+        translate[0] = 0.0
+        translate[1] = 0.0
+        translate[2] = 0.0
+        translate[3] = 0.0
+        viewport.translate = translate
+        ctx.set_viewport(viewport)
+    
+        # samplers
+        sampler = Sampler()
+        sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE
+        sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE
+        sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE
+        sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST
+        sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST
+        sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST
+        sampler.normalized_coords = 1
+        sampler.min_lod = 0
+        sampler.max_lod = PIPE_MAX_TEXTURE_LEVELS - 1
+        ctx.set_sampler(0, sampler)
+    
+        #  texture 
+        texture = dev.texture_create(
+            target = target,
+            format = format, 
+            width = width, 
+            height = height,
+            depth = depth, 
+            last_level = last_level,
+            tex_usage = tex_usage,
+        )
+        
+        expected_rgba = FloatArray(height*width*4) 
+        texture.get_surface(
+            usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE,
+            face = face,
+            level = level,
+            zslice = zslice,
+        ).sample_rgba(expected_rgba)
+        
+        ctx.set_sampler_texture(0, texture)
+
+        #  framebuffer 
+        cbuf_tex = dev.texture_create(
+            PIPE_FORMAT_A8R8G8B8_UNORM, 
+            width, 
+            height,
+            tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET,
+        )
+
+        cbuf = cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE|PIPE_BUFFER_USAGE_GPU_READ)
+        fb = Framebuffer()
+        fb.width = width
+        fb.height = height
+        fb.num_cbufs = 1
+        fb.set_cbuf(0, cbuf)
+        ctx.set_framebuffer(fb)
+        ctx.surface_clear(cbuf, 0x00000000)
+        del fb
+    
+        # vertex shader
+        vs = Shader('''
+            VERT1.1
+            DCL IN[0], POSITION, CONSTANT
+            DCL IN[1], GENERIC, CONSTANT
+            DCL OUT[0], POSITION, CONSTANT
+            DCL OUT[1], GENERIC, CONSTANT
+            0:MOV OUT[0], IN[0]
+            1:MOV OUT[1], IN[1]
+            2:END
+        ''')
+        #vs.dump()
+        ctx.set_vertex_shader(vs)
+    
+        # fragment shader
+        op = {
+            PIPE_TEXTURE_1D: "1D", 
+            PIPE_TEXTURE_2D: "2D", 
+            PIPE_TEXTURE_3D: "3D", 
+            PIPE_TEXTURE_CUBE: "CUBE",
+        }[target]
+        fs = Shader('''
+            FRAG1.1
+            DCL IN[0], GENERIC[0], LINEAR
+            DCL OUT[0], COLOR, CONSTANT
+            DCL SAMP[0], CONSTANT
+            0:TEX OUT[0], IN[0], SAMP[0], %s
+            1:END
+        ''' % op)
+        #fs.dump()
+        ctx.set_fragment_shader(fs)
+
+        nverts = 4
+        nattrs = 2
+        verts = FloatArray(nverts * nattrs * 4)
+    
+        x = 0
+        y = 0
+        w, h = minify((width, height), level)
+    
+        pos = [
+            [x, y],
+            [x+w, y],
+            [x+w, y+h],
+            [x, y+h],
+        ]
+    
+        tex = tex_coords(texture, face, level, zslice)
+    
+        for i in range(0, 4):
+            j = 8*i
+            verts[j + 0] = pos[i][0] # x
+            verts[j + 1] = pos[i][1] # y
+            verts[j + 2] = 0.0 # z
+            verts[j + 3] = 1.0 # w
+            verts[j + 4] = tex[i][0] # s
+            verts[j + 5] = tex[i][1] # r
+            verts[j + 6] = tex[i][2] # q
+            verts[j + 7] = 1.0
+    
+        ctx.draw_vertices(PIPE_PRIM_TRIANGLE_FAN,
+                          nverts, 
+                          nattrs, 
+                          verts)
+    
+        ctx.flush()
+    
+        cbuf = cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ)
+        
+        total = h*w
+        different = cbuf.compare_tile_rgba(x, y, w, h, expected_rgba, tol=4.0/256)
+        if different:
+            sys.stderr.write("%u out of %u pixels differ\n" % (different, total))
+
+        if float(total - different)/float(total) < 0.85:
+        
+            if 0:
+                rgba = FloatArray(h*w*4)
+                cbuf.get_tile_rgba(x, y, w, h, rgba)
+                show_image(w, h, Result=rgba, Expected=expected_rgba)
+                save_image(w, h, rgba, "result.png")
+                save_image(w, h, expected_rgba, "expected.png")
+            #sys.exit(0)
+            
+            raise TestFailure
+
+        del ctx
+        
+
+
+def main():
+    dev = Device()
+    suite = TestSuite()
+    
+    targets = []
+    targets += [PIPE_TEXTURE_2D]
+    targets += [PIPE_TEXTURE_CUBE]
+    targets += [PIPE_TEXTURE_3D]
+    
+    formats = []
+    formats += [PIPE_FORMAT_A8R8G8B8_UNORM]
+    formats += [PIPE_FORMAT_R5G6B5_UNORM]
+    formats += [PIPE_FORMAT_L8_UNORM]
+    formats += [PIPE_FORMAT_YCBCR]
+    formats += [PIPE_FORMAT_DXT1_RGB]
+    
+    sizes = [64, 32, 16, 8, 4, 2, 1]
+    #sizes = [1020, 508, 252, 62, 30, 14, 6, 3]
+    #sizes = [64]
+    #sizes = [63]
+    
+    for target in targets:
+        for format in formats:
+            for size in sizes:
+                if target == PIPE_TEXTURE_CUBE:
+                    faces = [
+                        PIPE_TEX_FACE_POS_X,
+                        PIPE_TEX_FACE_NEG_X,
+                        PIPE_TEX_FACE_POS_Y,
+                        PIPE_TEX_FACE_NEG_Y, 
+                        PIPE_TEX_FACE_POS_Z, 
+                        PIPE_TEX_FACE_NEG_Z,
+                    ]
+                    #faces = [PIPE_TEX_FACE_NEG_X]
+                else:
+                    faces = [0]
+                if target == PIPE_TEXTURE_3D:
+                    depth = size
+                else:
+                    depth = 1
+                for face in faces:
+                    levels = lods(size)
+                    for last_level in range(levels):
+                        for level in range(0, last_level + 1):
+                            zslice = 0
+                            while zslice < depth >> level:
+                                test = TextureTest(
+                                    dev = dev,
+                                    target = target,
+                                    format = format, 
+                                    width = size,
+                                    height = size,
+                                    depth = depth,
+                                    last_level = last_level,
+                                    face = face,
+                                    level = level,
+                                    zslice = zslice,
+                                )
+                                suite.add_test(test)
+                                zslice = (zslice + 1)*2 - 1
+    suite.run()
+
+
+if __name__ == '__main__':
+    main()
diff --git a/src/gallium/state_trackers/wgl/SConscript b/src/gallium/state_trackers/wgl/SConscript
new file mode 100644
index 0000000000..c72f495735
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/SConscript
@@ -0,0 +1,43 @@
+import os
+
+Import('*')
+
+if env['platform'] in ['windows']:
+     
+    env = env.Clone()
+     
+    env.Append(CPPPATH = [
+    	'#src/mesa',
+    	'.',
+    ])
+     
+    env.Append(CPPDEFINES = [
+        '_GDI32_', # prevent wgl* being declared __declspec(dllimport)
+        'BUILD_GL32', # declare gl* as __declspec(dllexport) in Mesa headers 
+        '__GL_EXPORTS',
+        '_GNU_H_WINDOWS32_DEFINES',
+    ])
+     
+    sources = [
+        'stw.c',
+
+        'icd/stw_icd.c',
+
+        'wgl/stw_wgl.c',
+
+        'shared/stw_context.c',
+        'shared/stw_device.c',
+        'shared/stw_framebuffer.c',
+        'shared/stw_pixelformat.c',
+        'shared/stw_quirks.c',
+        'shared/stw_arbextensionsstring.c',
+        'shared/stw_getprocaddress.c',
+        'shared/stw_arbpixelformat.c',
+    ]
+
+    wgl = env.ConvenienceLibrary(
+        target ='wgl',
+        source = sources,
+    )
+    
+    Export('wgl')
diff --git a/src/gallium/state_trackers/wgl/icd/stw_icd.c b/src/gallium/state_trackers/wgl/icd/stw_icd.c
new file mode 100644
index 0000000000..1aa4b8a6e2
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/icd/stw_icd.c
@@ -0,0 +1,773 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <windows.h>
+#include <stdio.h>
+
+#include "GL/gl.h"
+
+#include "pipe/p_debug.h"
+#include "pipe/p_thread.h"
+
+#include "shared/stw_public.h"
+#include "icd/stw_icd.h"
+#include "stw.h"
+
+
+#define DRV_CONTEXT_MAX 32
+
+struct stw_icd
+{
+   pipe_mutex mutex;
+
+   GLCLTPROCTABLE cpt;
+   boolean cpt_initialized;
+
+   struct {
+      struct stw_context *ctx;
+   } ctx_array[DRV_CONTEXT_MAX];
+};
+
+
+static struct stw_icd *stw_icd = NULL;
+
+
+boolean
+stw_icd_init( void )
+{
+   static struct stw_icd stw_icd_storage;
+
+   assert(!stw_icd);
+
+   stw_icd = &stw_icd_storage;
+   memset(stw_icd, 0, sizeof *stw_icd);
+
+   pipe_mutex_init( stw_icd->mutex );
+
+   return TRUE;
+}
+
+void
+stw_icd_cleanup(void)
+{
+   int i;
+
+   if (!stw_icd)
+      return;
+   
+   pipe_mutex_lock( stw_icd->mutex );
+   {
+      /* Ensure all contexts are destroyed */
+      for (i = 0; i < DRV_CONTEXT_MAX; i++)
+         if (stw_icd->ctx_array[i].ctx) 
+            stw_delete_context( stw_icd->ctx_array[i].ctx );
+   }
+   pipe_mutex_unlock( stw_icd->mutex );
+
+   pipe_mutex_init( stw_icd->mutex );
+   stw_icd = NULL;
+}
+
+
+static struct stw_context *
+lookup_context( struct stw_icd *icd, 
+                DHGLRC dhglrc )
+{
+   if (dhglrc == 0 || 
+       dhglrc >= DRV_CONTEXT_MAX)
+      return NULL;
+
+   if (icd == NULL)
+      return NULL;
+
+   return icd->ctx_array[dhglrc - 1].ctx;
+}
+
+BOOL APIENTRY
+DrvCopyContext(
+   DHGLRC dhrcSource,
+   DHGLRC dhrcDest,
+   UINT fuMask )
+{
+   BOOL ret = FALSE;
+
+   if (!stw_icd)
+      return FALSE;
+
+   pipe_mutex_lock( stw_icd->mutex );
+   {
+      struct stw_context *src = lookup_context( stw_icd, dhrcSource );
+      struct stw_context *dst = lookup_context( stw_icd, dhrcDest );
+   
+      if (src == NULL || dst == NULL) 
+         goto done;
+
+      ret = stw_copy_context( src, dst, fuMask );
+   }
+done:
+   pipe_mutex_unlock( stw_icd->mutex );
+
+   return ret;
+}
+
+DHGLRC APIENTRY
+DrvCreateLayerContext(
+   HDC hdc,
+   INT iLayerPlane )
+{
+   DHGLRC handle = 0;
+
+   if (!stw_icd)
+      return handle;
+
+   pipe_mutex_lock( stw_icd->mutex );
+   {
+      int i;
+
+      for (i = 0; i < DRV_CONTEXT_MAX; i++) {
+         if (stw_icd->ctx_array[i].ctx == NULL)
+            break;
+      }
+   
+      /* No slot available, fail:
+       */
+      if (i == DRV_CONTEXT_MAX)
+         goto done;
+
+      stw_icd->ctx_array[i].ctx = stw_create_context( hdc, iLayerPlane );
+      if (stw_icd->ctx_array[i].ctx == NULL) 
+         goto done;
+      
+      /* success:
+       */
+      handle = (DHGLRC) i + 1;
+   }
+done:
+   pipe_mutex_unlock( stw_icd->mutex );
+
+   return handle;
+}
+
+DHGLRC APIENTRY
+DrvCreateContext(
+   HDC hdc )
+{
+   return DrvCreateLayerContext( hdc, 0 );
+}
+
+BOOL APIENTRY
+DrvDeleteContext(
+   DHGLRC dhglrc )
+{
+   BOOL ret = FALSE;
+
+   if (!stw_icd)
+      return ret;
+
+   pipe_mutex_lock( stw_icd->mutex );
+   {
+      struct stw_context *ctx;
+
+      ctx = lookup_context( stw_icd, dhglrc );
+      if (ctx == NULL) 
+         goto done;
+      
+      if (stw_delete_context( ctx ) == FALSE)
+         goto done;
+      
+      stw_icd->ctx_array[dhglrc - 1].ctx = NULL;
+      ret = TRUE;
+
+   }
+done:
+   pipe_mutex_unlock( stw_icd->mutex );
+   
+   return ret;
+}
+
+BOOL APIENTRY
+DrvDescribeLayerPlane(
+   HDC hdc,
+   INT iPixelFormat,
+   INT iLayerPlane,
+   UINT nBytes,
+   LPLAYERPLANEDESCRIPTOR plpd )
+{
+   debug_printf( "%s\n", __FUNCTION__ );
+
+   return FALSE;
+}
+
+LONG APIENTRY
+DrvDescribePixelFormat(
+   HDC hdc,
+   INT iPixelFormat,
+   ULONG cjpfd,
+   PIXELFORMATDESCRIPTOR *ppfd )
+{
+   LONG r;
+
+   r = stw_pixelformat_describe( hdc, iPixelFormat, cjpfd, ppfd );
+
+   debug_printf( "%s( %p, %d, %u, %p ) = %d\n",
+                 __FUNCTION__, hdc, iPixelFormat, cjpfd, ppfd, r );
+
+   return r;
+}
+
+int APIENTRY
+DrvGetLayerPaletteEntries(
+   HDC hdc,
+   INT iLayerPlane,
+   INT iStart,
+   INT cEntries,
+   COLORREF *pcr )
+{
+   debug_printf( "%s\n", __FUNCTION__ );
+
+   return 0;
+}
+
+PROC APIENTRY
+DrvGetProcAddress(
+   LPCSTR lpszProc )
+{
+   PROC r;
+
+   r = stw_get_proc_address( lpszProc );
+
+   debug_printf( "%s( \", __FUNCTION__%s\" ) = %p\n", lpszProc, r );
+
+   return r;
+}
+
+BOOL APIENTRY
+DrvRealizeLayerPalette(
+   HDC hdc,
+   INT iLayerPlane,
+   BOOL bRealize )
+{
+   debug_printf( "%s\n", __FUNCTION__ );
+
+   return FALSE;
+}
+
+BOOL APIENTRY
+DrvReleaseContext(
+   DHGLRC dhglrc )
+{
+   BOOL ret = FALSE;
+
+   if (!stw_icd)
+      return ret;
+
+   pipe_mutex_lock( stw_icd->mutex );
+   {
+      struct stw_context *ctx;
+
+      /* XXX: The expectation is that ctx is the same context which is
+       * current for this thread.  We should check that and return False
+       * if not the case.
+       */
+      ctx = lookup_context( stw_icd, dhglrc );
+      if (ctx == NULL) 
+         goto done;
+
+      if (stw_make_current( NULL, NULL ) == FALSE)
+         goto done;
+
+      ret = TRUE;
+   }
+done:
+   pipe_mutex_unlock( stw_icd->mutex );
+
+   return ret;
+}
+
+void APIENTRY
+DrvSetCallbackProcs(
+   INT nProcs,
+   PROC *pProcs )
+{
+   debug_printf( "%s( %d, %p )\n", __FUNCTION__, nProcs, pProcs );
+
+   return;
+}
+
+
+static void init_proc_table( GLCLTPROCTABLE *cpt )
+{ 
+   GLDISPATCHTABLE *disp = &cpt->glDispatchTable;
+
+   memset( cpt, 0, sizeof *cpt );
+   cpt->cEntries = OPENGL_VERSION_110_ENTRIES;
+
+#define GPA_GL( NAME ) disp->NAME = gl##NAME
+   GPA_GL( NewList );
+   GPA_GL( EndList );
+   GPA_GL( CallList );
+   GPA_GL( CallLists );
+   GPA_GL( DeleteLists );
+   GPA_GL( GenLists );
+   GPA_GL( ListBase );
+   GPA_GL( Begin );
+   GPA_GL( Bitmap );
+   GPA_GL( Color3b );
+   GPA_GL( Color3bv );
+   GPA_GL( Color3d );
+   GPA_GL( Color3dv );
+   GPA_GL( Color3f );
+   GPA_GL( Color3fv );
+   GPA_GL( Color3i );
+   GPA_GL( Color3iv );
+   GPA_GL( Color3s );
+   GPA_GL( Color3sv );
+   GPA_GL( Color3ub );
+   GPA_GL( Color3ubv );
+   GPA_GL( Color3ui );
+   GPA_GL( Color3uiv );
+   GPA_GL( Color3us );
+   GPA_GL( Color3usv );
+   GPA_GL( Color4b );
+   GPA_GL( Color4bv );
+   GPA_GL( Color4d );
+   GPA_GL( Color4dv );
+   GPA_GL( Color4f );
+   GPA_GL( Color4fv );
+   GPA_GL( Color4i );
+   GPA_GL( Color4iv );
+   GPA_GL( Color4s );
+   GPA_GL( Color4sv );
+   GPA_GL( Color4ub );
+   GPA_GL( Color4ubv );
+   GPA_GL( Color4ui );
+   GPA_GL( Color4uiv );
+   GPA_GL( Color4us );
+   GPA_GL( Color4usv );
+   GPA_GL( EdgeFlag );
+   GPA_GL( EdgeFlagv );
+   GPA_GL( End );
+   GPA_GL( Indexd );
+   GPA_GL( Indexdv );
+   GPA_GL( Indexf );
+   GPA_GL( Indexfv );
+   GPA_GL( Indexi );
+   GPA_GL( Indexiv );
+   GPA_GL( Indexs );
+   GPA_GL( Indexsv );
+   GPA_GL( Normal3b );
+   GPA_GL( Normal3bv );
+   GPA_GL( Normal3d );
+   GPA_GL( Normal3dv );
+   GPA_GL( Normal3f );
+   GPA_GL( Normal3fv );
+   GPA_GL( Normal3i );
+   GPA_GL( Normal3iv );
+   GPA_GL( Normal3s );
+   GPA_GL( Normal3sv );
+   GPA_GL( RasterPos2d );
+   GPA_GL( RasterPos2dv );
+   GPA_GL( RasterPos2f );
+   GPA_GL( RasterPos2fv );
+   GPA_GL( RasterPos2i );
+   GPA_GL( RasterPos2iv );
+   GPA_GL( RasterPos2s );
+   GPA_GL( RasterPos2sv );
+   GPA_GL( RasterPos3d );
+   GPA_GL( RasterPos3dv );
+   GPA_GL( RasterPos3f );
+   GPA_GL( RasterPos3fv );
+   GPA_GL( RasterPos3i );
+   GPA_GL( RasterPos3iv );
+   GPA_GL( RasterPos3s );
+   GPA_GL( RasterPos3sv );
+   GPA_GL( RasterPos4d );
+   GPA_GL( RasterPos4dv );
+   GPA_GL( RasterPos4f );
+   GPA_GL( RasterPos4fv );
+   GPA_GL( RasterPos4i );
+   GPA_GL( RasterPos4iv );
+   GPA_GL( RasterPos4s );
+   GPA_GL( RasterPos4sv );
+   GPA_GL( Rectd );
+   GPA_GL( Rectdv );
+   GPA_GL( Rectf );
+   GPA_GL( Rectfv );
+   GPA_GL( Recti );
+   GPA_GL( Rectiv );
+   GPA_GL( Rects );
+   GPA_GL( Rectsv );
+   GPA_GL( TexCoord1d );
+   GPA_GL( TexCoord1dv );
+   GPA_GL( TexCoord1f );
+   GPA_GL( TexCoord1fv );
+   GPA_GL( TexCoord1i );
+   GPA_GL( TexCoord1iv );
+   GPA_GL( TexCoord1s );
+   GPA_GL( TexCoord1sv );
+   GPA_GL( TexCoord2d );
+   GPA_GL( TexCoord2dv );
+   GPA_GL( TexCoord2f );
+   GPA_GL( TexCoord2fv );
+   GPA_GL( TexCoord2i );
+   GPA_GL( TexCoord2iv );
+   GPA_GL( TexCoord2s );
+   GPA_GL( TexCoord2sv );
+   GPA_GL( TexCoord3d );
+   GPA_GL( TexCoord3dv );
+   GPA_GL( TexCoord3f );
+   GPA_GL( TexCoord3fv );
+   GPA_GL( TexCoord3i );
+   GPA_GL( TexCoord3iv );
+   GPA_GL( TexCoord3s );
+   GPA_GL( TexCoord3sv );
+   GPA_GL( TexCoord4d );
+   GPA_GL( TexCoord4dv );
+   GPA_GL( TexCoord4f );
+   GPA_GL( TexCoord4fv );
+   GPA_GL( TexCoord4i );
+   GPA_GL( TexCoord4iv );
+   GPA_GL( TexCoord4s );
+   GPA_GL( TexCoord4sv );
+   GPA_GL( Vertex2d );
+   GPA_GL( Vertex2dv );
+   GPA_GL( Vertex2f );
+   GPA_GL( Vertex2fv );
+   GPA_GL( Vertex2i );
+   GPA_GL( Vertex2iv );
+   GPA_GL( Vertex2s );
+   GPA_GL( Vertex2sv );
+   GPA_GL( Vertex3d );
+   GPA_GL( Vertex3dv );
+   GPA_GL( Vertex3f );
+   GPA_GL( Vertex3fv );
+   GPA_GL( Vertex3i );
+   GPA_GL( Vertex3iv );
+   GPA_GL( Vertex3s );
+   GPA_GL( Vertex3sv );
+   GPA_GL( Vertex4d );
+   GPA_GL( Vertex4dv );
+   GPA_GL( Vertex4f );
+   GPA_GL( Vertex4fv );
+   GPA_GL( Vertex4i );
+   GPA_GL( Vertex4iv );
+   GPA_GL( Vertex4s );
+   GPA_GL( Vertex4sv );
+   GPA_GL( ClipPlane );
+   GPA_GL( ColorMaterial );
+   GPA_GL( CullFace );
+   GPA_GL( Fogf );
+   GPA_GL( Fogfv );
+   GPA_GL( Fogi );
+   GPA_GL( Fogiv );
+   GPA_GL( FrontFace );
+   GPA_GL( Hint );
+   GPA_GL( Lightf );
+   GPA_GL( Lightfv );
+   GPA_GL( Lighti );
+   GPA_GL( Lightiv );
+   GPA_GL( LightModelf );
+   GPA_GL( LightModelfv );
+   GPA_GL( LightModeli );
+   GPA_GL( LightModeliv );
+   GPA_GL( LineStipple );
+   GPA_GL( LineWidth );
+   GPA_GL( Materialf );
+   GPA_GL( Materialfv );
+   GPA_GL( Materiali );
+   GPA_GL( Materialiv );
+   GPA_GL( PointSize );
+   GPA_GL( PolygonMode );
+   GPA_GL( PolygonStipple );
+   GPA_GL( Scissor );
+   GPA_GL( ShadeModel );
+   GPA_GL( TexParameterf );
+   GPA_GL( TexParameterfv );
+   GPA_GL( TexParameteri );
+   GPA_GL( TexParameteriv );
+   GPA_GL( TexImage1D );
+   GPA_GL( TexImage2D );
+   GPA_GL( TexEnvf );
+   GPA_GL( TexEnvfv );
+   GPA_GL( TexEnvi );
+   GPA_GL( TexEnviv );
+   GPA_GL( TexGend );
+   GPA_GL( TexGendv );
+   GPA_GL( TexGenf );
+   GPA_GL( TexGenfv );
+   GPA_GL( TexGeni );
+   GPA_GL( TexGeniv );
+   GPA_GL( FeedbackBuffer );
+   GPA_GL( SelectBuffer );
+   GPA_GL( RenderMode );
+   GPA_GL( InitNames );
+   GPA_GL( LoadName );
+   GPA_GL( PassThrough );
+   GPA_GL( PopName );
+   GPA_GL( PushName );
+   GPA_GL( DrawBuffer );
+   GPA_GL( Clear );
+   GPA_GL( ClearAccum );
+   GPA_GL( ClearIndex );
+   GPA_GL( ClearColor );
+   GPA_GL( ClearStencil );
+   GPA_GL( ClearDepth );
+   GPA_GL( StencilMask );
+   GPA_GL( ColorMask );
+   GPA_GL( DepthMask );
+   GPA_GL( IndexMask );
+   GPA_GL( Accum );
+   GPA_GL( Disable );
+   GPA_GL( Enable );
+   GPA_GL( Finish );
+   GPA_GL( Flush );
+   GPA_GL( PopAttrib );
+   GPA_GL( PushAttrib );
+   GPA_GL( Map1d );
+   GPA_GL( Map1f );
+   GPA_GL( Map2d );
+   GPA_GL( Map2f );
+   GPA_GL( MapGrid1d );
+   GPA_GL( MapGrid1f );
+   GPA_GL( MapGrid2d );
+   GPA_GL( MapGrid2f );
+   GPA_GL( EvalCoord1d );
+   GPA_GL( EvalCoord1dv );
+   GPA_GL( EvalCoord1f );
+   GPA_GL( EvalCoord1fv );
+   GPA_GL( EvalCoord2d );
+   GPA_GL( EvalCoord2dv );
+   GPA_GL( EvalCoord2f );
+   GPA_GL( EvalCoord2fv );
+   GPA_GL( EvalMesh1 );
+   GPA_GL( EvalPoint1 );
+   GPA_GL( EvalMesh2 );
+   GPA_GL( EvalPoint2 );
+   GPA_GL( AlphaFunc );
+   GPA_GL( BlendFunc );
+   GPA_GL( LogicOp );
+   GPA_GL( StencilFunc );
+   GPA_GL( StencilOp );
+   GPA_GL( DepthFunc );
+   GPA_GL( PixelZoom );
+   GPA_GL( PixelTransferf );
+   GPA_GL( PixelTransferi );
+   GPA_GL( PixelStoref );
+   GPA_GL( PixelStorei );
+   GPA_GL( PixelMapfv );
+   GPA_GL( PixelMapuiv );
+   GPA_GL( PixelMapusv );
+   GPA_GL( ReadBuffer );
+   GPA_GL( CopyPixels );
+   GPA_GL( ReadPixels );
+   GPA_GL( DrawPixels );
+   GPA_GL( GetBooleanv );
+   GPA_GL( GetClipPlane );
+   GPA_GL( GetDoublev );
+   GPA_GL( GetError );
+   GPA_GL( GetFloatv );
+   GPA_GL( GetIntegerv );
+   GPA_GL( GetLightfv );
+   GPA_GL( GetLightiv );
+   GPA_GL( GetMapdv );
+   GPA_GL( GetMapfv );
+   GPA_GL( GetMapiv );
+   GPA_GL( GetMaterialfv );
+   GPA_GL( GetMaterialiv );
+   GPA_GL( GetPixelMapfv );
+   GPA_GL( GetPixelMapuiv );
+   GPA_GL( GetPixelMapusv );
+   GPA_GL( GetPolygonStipple );
+   GPA_GL( GetString );
+   GPA_GL( GetTexEnvfv );
+   GPA_GL( GetTexEnviv );
+   GPA_GL( GetTexGendv );
+   GPA_GL( GetTexGenfv );
+   GPA_GL( GetTexGeniv );
+   GPA_GL( GetTexImage );
+   GPA_GL( GetTexParameterfv );
+   GPA_GL( GetTexParameteriv );
+   GPA_GL( GetTexLevelParameterfv );
+   GPA_GL( GetTexLevelParameteriv );
+   GPA_GL( IsEnabled );
+   GPA_GL( IsList );
+   GPA_GL( DepthRange );
+   GPA_GL( Frustum );
+   GPA_GL( LoadIdentity );
+   GPA_GL( LoadMatrixf );
+   GPA_GL( LoadMatrixd );
+   GPA_GL( MatrixMode );
+   GPA_GL( MultMatrixf );
+   GPA_GL( MultMatrixd );
+   GPA_GL( Ortho );
+   GPA_GL( PopMatrix );
+   GPA_GL( PushMatrix );
+   GPA_GL( Rotated );
+   GPA_GL( Rotatef );
+   GPA_GL( Scaled );
+   GPA_GL( Scalef );
+   GPA_GL( Translated );
+   GPA_GL( Translatef );
+   GPA_GL( Viewport );
+   GPA_GL( ArrayElement );
+   GPA_GL( BindTexture );
+   GPA_GL( ColorPointer );
+   GPA_GL( DisableClientState );
+   GPA_GL( DrawArrays );
+   GPA_GL( DrawElements );
+   GPA_GL( EdgeFlagPointer );
+   GPA_GL( EnableClientState );
+   GPA_GL( IndexPointer );
+   GPA_GL( Indexub );
+   GPA_GL( Indexubv );
+   GPA_GL( InterleavedArrays );
+   GPA_GL( NormalPointer );
+   GPA_GL( PolygonOffset );
+   GPA_GL( TexCoordPointer );
+   GPA_GL( VertexPointer );
+   GPA_GL( AreTexturesResident );
+   GPA_GL( CopyTexImage1D );
+   GPA_GL( CopyTexImage2D );
+   GPA_GL( CopyTexSubImage1D );
+   GPA_GL( CopyTexSubImage2D );
+   GPA_GL( DeleteTextures );
+   GPA_GL( GenTextures );
+   GPA_GL( GetPointerv );
+   GPA_GL( IsTexture );
+   GPA_GL( PrioritizeTextures );
+   GPA_GL( TexSubImage1D );
+   GPA_GL( TexSubImage2D );
+   GPA_GL( PopClientAttrib );
+   GPA_GL( PushClientAttrib );
+}
+
+PGLCLTPROCTABLE APIENTRY
+DrvSetContext(
+   HDC hdc,
+   DHGLRC dhglrc,
+   PFN_SETPROCTABLE pfnSetProcTable )
+{
+   PGLCLTPROCTABLE result = NULL;
+
+   if (!stw_icd)
+      return result;
+
+   pipe_mutex_lock( stw_icd->mutex ); 
+   {
+      struct stw_context *ctx;
+
+      debug_printf( "%s( 0x%p, %u, 0x%p )\n", 
+                    __FUNCTION__, hdc, dhglrc, pfnSetProcTable );
+
+      /* Although WGL allows different dispatch entrypoints per 
+       */
+      if (!stw_icd->cpt_initialized) {
+         init_proc_table( &stw_icd->cpt );
+         stw_icd->cpt_initialized = TRUE;
+      }
+
+      ctx = lookup_context( stw_icd, dhglrc );
+      if (ctx == NULL)
+         goto done;
+
+      if (!stw_make_current( hdc, ctx ))
+         goto done;
+
+      result = &stw_icd->cpt;
+   }
+done:
+   pipe_mutex_unlock( stw_icd->mutex );
+
+   return result;
+}
+
+int APIENTRY
+DrvSetLayerPaletteEntries(
+   HDC hdc,
+   INT iLayerPlane,
+   INT iStart,
+   INT cEntries,
+   CONST COLORREF *pcr )
+{
+   debug_printf( "%s\n", __FUNCTION__ );
+
+   return 0;
+}
+
+BOOL APIENTRY
+DrvSetPixelFormat(
+   HDC hdc,
+   LONG iPixelFormat )
+{
+   BOOL r;
+
+   r = stw_pixelformat_set( hdc, iPixelFormat );
+
+   debug_printf( "%s( %p, %d ) = %s\n", __FUNCTION__, hdc, iPixelFormat, r ? "TRUE" : "FALSE" );
+
+   return r;
+}
+
+BOOL APIENTRY
+DrvShareLists(
+   DHGLRC dhglrc1,
+   DHGLRC dhglrc2 )
+{
+   debug_printf( "%s\n", __FUNCTION__ );
+
+   return FALSE;
+}
+
+BOOL APIENTRY
+DrvSwapBuffers(
+   HDC hdc )
+{
+   debug_printf( "%s( %p )\n", __FUNCTION__, hdc );
+
+   return stw_swap_buffers( hdc );
+}
+
+BOOL APIENTRY
+DrvSwapLayerBuffers(
+   HDC hdc,
+   UINT fuPlanes )
+{
+   debug_printf( "%s\n", __FUNCTION__ );
+
+   return FALSE;
+}
+
+BOOL APIENTRY
+DrvValidateVersion(
+   ULONG ulVersion )
+{
+   debug_printf( "%s( %u )\n", __FUNCTION__, ulVersion );
+
+   /* TODO: get the expected version from the winsys */
+   
+   return ulVersion == 1;
+}
diff --git a/src/gallium/state_trackers/wgl/icd/stw_icd.h b/src/gallium/state_trackers/wgl/icd/stw_icd.h
new file mode 100644
index 0000000000..8e676fb5b7
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/icd/stw_icd.h
@@ -0,0 +1,489 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef DRV_H
+#define DRV_H
+
+
+#include <windows.h>
+
+#include "GL/gl.h"
+
+
+typedef ULONG DHGLRC;
+
+#define OPENGL_VERSION_110_ENTRIES  336
+
+struct __GLdispatchTableRec
+{
+   void (GLAPIENTRY * NewList)(GLuint, GLenum);
+   void (GLAPIENTRY * EndList)(void);
+   void (GLAPIENTRY * CallList)(GLuint);
+   void (GLAPIENTRY * CallLists)(GLsizei, GLenum, const GLvoid *);
+   void (GLAPIENTRY * DeleteLists)(GLuint, GLsizei);
+   GLuint (GLAPIENTRY * GenLists)(GLsizei);
+   void (GLAPIENTRY * ListBase)(GLuint);
+   void (GLAPIENTRY * Begin)(GLenum);
+   void (GLAPIENTRY * Bitmap)(GLsizei, GLsizei, GLfloat, GLfloat, GLfloat, GLfloat, const GLubyte *);
+   void (GLAPIENTRY * Color3b)(GLbyte, GLbyte, GLbyte);
+   void (GLAPIENTRY * Color3bv)(const GLbyte *);
+   void (GLAPIENTRY * Color3d)(GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * Color3dv)(const GLdouble *);
+   void (GLAPIENTRY * Color3f)(GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * Color3fv)(const GLfloat *);
+   void (GLAPIENTRY * Color3i)(GLint, GLint, GLint);
+   void (GLAPIENTRY * Color3iv)(const GLint *);
+   void (GLAPIENTRY * Color3s)(GLshort, GLshort, GLshort);
+   void (GLAPIENTRY * Color3sv)(const GLshort *);
+   void (GLAPIENTRY * Color3ub)(GLubyte, GLubyte, GLubyte);
+   void (GLAPIENTRY * Color3ubv)(const GLubyte *);
+   void (GLAPIENTRY * Color3ui)(GLuint, GLuint, GLuint);
+   void (GLAPIENTRY * Color3uiv)(const GLuint *);
+   void (GLAPIENTRY * Color3us)(GLushort, GLushort, GLushort);
+   void (GLAPIENTRY * Color3usv)(const GLushort *);
+   void (GLAPIENTRY * Color4b)(GLbyte, GLbyte, GLbyte, GLbyte);
+   void (GLAPIENTRY * Color4bv)(const GLbyte *);
+   void (GLAPIENTRY * Color4d)(GLdouble, GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * Color4dv)(const GLdouble *);
+   void (GLAPIENTRY * Color4f)(GLfloat, GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * Color4fv)(const GLfloat *);
+   void (GLAPIENTRY * Color4i)(GLint, GLint, GLint, GLint);
+   void (GLAPIENTRY * Color4iv)(const GLint *);
+   void (GLAPIENTRY * Color4s)(GLshort, GLshort, GLshort, GLshort);
+   void (GLAPIENTRY * Color4sv)(const GLshort *);
+   void (GLAPIENTRY * Color4ub)(GLubyte, GLubyte, GLubyte, GLubyte);
+   void (GLAPIENTRY * Color4ubv)(const GLubyte *);
+   void (GLAPIENTRY * Color4ui)(GLuint, GLuint, GLuint, GLuint);
+   void (GLAPIENTRY * Color4uiv)(const GLuint *);
+   void (GLAPIENTRY * Color4us)(GLushort, GLushort, GLushort, GLushort);
+   void (GLAPIENTRY * Color4usv)(const GLushort *);
+   void (GLAPIENTRY * EdgeFlag)(GLboolean);
+   void (GLAPIENTRY * EdgeFlagv)(const GLboolean *);
+   void (GLAPIENTRY * End)(void);
+   void (GLAPIENTRY * Indexd)(GLdouble);
+   void (GLAPIENTRY * Indexdv)(const GLdouble *);
+   void (GLAPIENTRY * Indexf)(GLfloat);
+   void (GLAPIENTRY * Indexfv)(const GLfloat *);
+   void (GLAPIENTRY * Indexi)(GLint);
+   void (GLAPIENTRY * Indexiv)(const GLint *);
+   void (GLAPIENTRY * Indexs)(GLshort);
+   void (GLAPIENTRY * Indexsv)(const GLshort *);
+   void (GLAPIENTRY * Normal3b)(GLbyte, GLbyte, GLbyte);
+   void (GLAPIENTRY * Normal3bv)(const GLbyte *);
+   void (GLAPIENTRY * Normal3d)(GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * Normal3dv)(const GLdouble *);
+   void (GLAPIENTRY * Normal3f)(GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * Normal3fv)(const GLfloat *);
+   void (GLAPIENTRY * Normal3i)(GLint, GLint, GLint);
+   void (GLAPIENTRY * Normal3iv)(const GLint *);
+   void (GLAPIENTRY * Normal3s)(GLshort, GLshort, GLshort);
+   void (GLAPIENTRY * Normal3sv)(const GLshort *);
+   void (GLAPIENTRY * RasterPos2d)(GLdouble, GLdouble);
+   void (GLAPIENTRY * RasterPos2dv)(const GLdouble *);
+   void (GLAPIENTRY * RasterPos2f)(GLfloat, GLfloat);
+   void (GLAPIENTRY * RasterPos2fv)(const GLfloat *);
+   void (GLAPIENTRY * RasterPos2i)(GLint, GLint);
+   void (GLAPIENTRY * RasterPos2iv)(const GLint *);
+   void (GLAPIENTRY * RasterPos2s)(GLshort, GLshort);
+   void (GLAPIENTRY * RasterPos2sv)(const GLshort *);
+   void (GLAPIENTRY * RasterPos3d)(GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * RasterPos3dv)(const GLdouble *);
+   void (GLAPIENTRY * RasterPos3f)(GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * RasterPos3fv)(const GLfloat *);
+   void (GLAPIENTRY * RasterPos3i)(GLint, GLint, GLint);
+   void (GLAPIENTRY * RasterPos3iv)(const GLint *);
+   void (GLAPIENTRY * RasterPos3s)(GLshort, GLshort, GLshort);
+   void (GLAPIENTRY * RasterPos3sv)(const GLshort *);
+   void (GLAPIENTRY * RasterPos4d)(GLdouble, GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * RasterPos4dv)(const GLdouble *);
+   void (GLAPIENTRY * RasterPos4f)(GLfloat, GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * RasterPos4fv)(const GLfloat *);
+   void (GLAPIENTRY * RasterPos4i)(GLint, GLint, GLint, GLint);
+   void (GLAPIENTRY * RasterPos4iv)(const GLint *);
+   void (GLAPIENTRY * RasterPos4s)(GLshort, GLshort, GLshort, GLshort);
+   void (GLAPIENTRY * RasterPos4sv)(const GLshort *);
+   void (GLAPIENTRY * Rectd)(GLdouble, GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * Rectdv)(const GLdouble *, const GLdouble *);
+   void (GLAPIENTRY * Rectf)(GLfloat, GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * Rectfv)(const GLfloat *, const GLfloat *);
+   void (GLAPIENTRY * Recti)(GLint, GLint, GLint, GLint);
+   void (GLAPIENTRY * Rectiv)(const GLint *, const GLint *);
+   void (GLAPIENTRY * Rects)(GLshort, GLshort, GLshort, GLshort);
+   void (GLAPIENTRY * Rectsv)(const GLshort *, const GLshort *);
+   void (GLAPIENTRY * TexCoord1d)(GLdouble);
+   void (GLAPIENTRY * TexCoord1dv)(const GLdouble *);
+   void (GLAPIENTRY * TexCoord1f)(GLfloat);
+   void (GLAPIENTRY * TexCoord1fv)(const GLfloat *);
+   void (GLAPIENTRY * TexCoord1i)(GLint);
+   void (GLAPIENTRY * TexCoord1iv)(const GLint *);
+   void (GLAPIENTRY * TexCoord1s)(GLshort);
+   void (GLAPIENTRY * TexCoord1sv)(const GLshort *);
+   void (GLAPIENTRY * TexCoord2d)(GLdouble, GLdouble);
+   void (GLAPIENTRY * TexCoord2dv)(const GLdouble *);
+   void (GLAPIENTRY * TexCoord2f)(GLfloat, GLfloat);
+   void (GLAPIENTRY * TexCoord2fv)(const GLfloat *);
+   void (GLAPIENTRY * TexCoord2i)(GLint, GLint);
+   void (GLAPIENTRY * TexCoord2iv)(const GLint *);
+   void (GLAPIENTRY * TexCoord2s)(GLshort, GLshort);
+   void (GLAPIENTRY * TexCoord2sv)(const GLshort *);
+   void (GLAPIENTRY * TexCoord3d)(GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * TexCoord3dv)(const GLdouble *);
+   void (GLAPIENTRY * TexCoord3f)(GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * TexCoord3fv)(const GLfloat *);
+   void (GLAPIENTRY * TexCoord3i)(GLint, GLint, GLint);
+   void (GLAPIENTRY * TexCoord3iv)(const GLint *);
+   void (GLAPIENTRY * TexCoord3s)(GLshort, GLshort, GLshort);
+   void (GLAPIENTRY * TexCoord3sv)(const GLshort *);
+   void (GLAPIENTRY * TexCoord4d)(GLdouble, GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * TexCoord4dv)(const GLdouble *);
+   void (GLAPIENTRY * TexCoord4f)(GLfloat, GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * TexCoord4fv)(const GLfloat *);
+   void (GLAPIENTRY * TexCoord4i)(GLint, GLint, GLint, GLint);
+   void (GLAPIENTRY * TexCoord4iv)(const GLint *);
+   void (GLAPIENTRY * TexCoord4s)(GLshort, GLshort, GLshort, GLshort);
+   void (GLAPIENTRY * TexCoord4sv)(const GLshort *);
+   void (GLAPIENTRY * Vertex2d)(GLdouble, GLdouble);
+   void (GLAPIENTRY * Vertex2dv)(const GLdouble *);
+   void (GLAPIENTRY * Vertex2f)(GLfloat, GLfloat);
+   void (GLAPIENTRY * Vertex2fv)(const GLfloat *);
+   void (GLAPIENTRY * Vertex2i)(GLint, GLint);
+   void (GLAPIENTRY * Vertex2iv)(const GLint *);
+   void (GLAPIENTRY * Vertex2s)(GLshort, GLshort);
+   void (GLAPIENTRY * Vertex2sv)(const GLshort *);
+   void (GLAPIENTRY * Vertex3d)(GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * Vertex3dv)(const GLdouble *);
+   void (GLAPIENTRY * Vertex3f)(GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * Vertex3fv)(const GLfloat *);
+   void (GLAPIENTRY * Vertex3i)(GLint, GLint, GLint);
+   void (GLAPIENTRY * Vertex3iv)(const GLint *);
+   void (GLAPIENTRY * Vertex3s)(GLshort, GLshort, GLshort);
+   void (GLAPIENTRY * Vertex3sv)(const GLshort *);
+   void (GLAPIENTRY * Vertex4d)(GLdouble, GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * Vertex4dv)(const GLdouble *);
+   void (GLAPIENTRY * Vertex4f)(GLfloat, GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * Vertex4fv)(const GLfloat *);
+   void (GLAPIENTRY * Vertex4i)(GLint, GLint, GLint, GLint);
+   void (GLAPIENTRY * Vertex4iv)(const GLint *);
+   void (GLAPIENTRY * Vertex4s)(GLshort, GLshort, GLshort, GLshort);
+   void (GLAPIENTRY * Vertex4sv)(const GLshort *);
+   void (GLAPIENTRY * ClipPlane)(GLenum, const GLdouble *);
+   void (GLAPIENTRY * ColorMaterial)(GLenum, GLenum);
+   void (GLAPIENTRY * CullFace)(GLenum);
+   void (GLAPIENTRY * Fogf)(GLenum, GLfloat);
+   void (GLAPIENTRY * Fogfv)(GLenum, const GLfloat *);
+   void (GLAPIENTRY * Fogi)(GLenum, GLint);
+   void (GLAPIENTRY * Fogiv)(GLenum, const GLint *);
+   void (GLAPIENTRY * FrontFace)(GLenum);
+   void (GLAPIENTRY * Hint)(GLenum, GLenum);
+   void (GLAPIENTRY * Lightf)(GLenum, GLenum, GLfloat);
+   void (GLAPIENTRY * Lightfv)(GLenum, GLenum, const GLfloat *);
+   void (GLAPIENTRY * Lighti)(GLenum, GLenum, GLint);
+   void (GLAPIENTRY * Lightiv)(GLenum, GLenum, const GLint *);
+   void (GLAPIENTRY * LightModelf)(GLenum, GLfloat);
+   void (GLAPIENTRY * LightModelfv)(GLenum, const GLfloat *);
+   void (GLAPIENTRY * LightModeli)(GLenum, GLint);
+   void (GLAPIENTRY * LightModeliv)(GLenum, const GLint *);
+   void (GLAPIENTRY * LineStipple)(GLint, GLushort);
+   void (GLAPIENTRY * LineWidth)(GLfloat);
+   void (GLAPIENTRY * Materialf)(GLenum, GLenum, GLfloat);
+   void (GLAPIENTRY * Materialfv)(GLenum, GLenum, const GLfloat *);
+   void (GLAPIENTRY * Materiali)(GLenum, GLenum, GLint);
+   void (GLAPIENTRY * Materialiv)(GLenum, GLenum, const GLint *);
+   void (GLAPIENTRY * PointSize)(GLfloat);
+   void (GLAPIENTRY * PolygonMode)(GLenum, GLenum);
+   void (GLAPIENTRY * PolygonStipple)(const GLubyte *);
+   void (GLAPIENTRY * Scissor)(GLint, GLint, GLsizei, GLsizei);
+   void (GLAPIENTRY * ShadeModel)(GLenum);
+   void (GLAPIENTRY * TexParameterf)(GLenum, GLenum, GLfloat);
+   void (GLAPIENTRY * TexParameterfv)(GLenum, GLenum, const GLfloat *);
+   void (GLAPIENTRY * TexParameteri)(GLenum, GLenum, GLint);
+   void (GLAPIENTRY * TexParameteriv)(GLenum, GLenum, const GLint *);
+   void (GLAPIENTRY * TexImage1D)(GLenum, GLint, GLint, GLsizei, GLint, GLenum, GLenum, const GLvoid *);
+   void (GLAPIENTRY * TexImage2D)(GLenum, GLint, GLint, GLsizei, GLsizei, GLint, GLenum, GLenum, const GLvoid *);
+   void (GLAPIENTRY * TexEnvf)(GLenum, GLenum, GLfloat);
+   void (GLAPIENTRY * TexEnvfv)(GLenum, GLenum, const GLfloat *);
+   void (GLAPIENTRY * TexEnvi)(GLenum, GLenum, GLint);
+   void (GLAPIENTRY * TexEnviv)(GLenum, GLenum, const GLint *);
+   void (GLAPIENTRY * TexGend)(GLenum, GLenum, GLdouble);
+   void (GLAPIENTRY * TexGendv)(GLenum, GLenum, const GLdouble *);
+   void (GLAPIENTRY * TexGenf)(GLenum, GLenum, GLfloat);
+   void (GLAPIENTRY * TexGenfv)(GLenum, GLenum, const GLfloat *);
+   void (GLAPIENTRY * TexGeni)(GLenum, GLenum, GLint);
+   void (GLAPIENTRY * TexGeniv)(GLenum, GLenum, const GLint *);
+   void (GLAPIENTRY * FeedbackBuffer)(GLsizei, GLenum, GLfloat *);
+   void (GLAPIENTRY * SelectBuffer)(GLsizei, GLuint *);
+   GLint (GLAPIENTRY * RenderMode)(GLenum);
+   void (GLAPIENTRY * InitNames)(void);
+   void (GLAPIENTRY * LoadName)(GLuint);
+   void (GLAPIENTRY * PassThrough)(GLfloat);
+   void (GLAPIENTRY * PopName)(void);
+   void (GLAPIENTRY * PushName)(GLuint);
+   void (GLAPIENTRY * DrawBuffer)(GLenum);
+   void (GLAPIENTRY * Clear)(GLbitfield);
+   void (GLAPIENTRY * ClearAccum)(GLfloat, GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * ClearIndex)(GLfloat);
+   void (GLAPIENTRY * ClearColor)(GLclampf, GLclampf, GLclampf, GLclampf);
+   void (GLAPIENTRY * ClearStencil)(GLint);
+   void (GLAPIENTRY * ClearDepth)(GLclampd);
+   void (GLAPIENTRY * StencilMask)(GLuint);
+   void (GLAPIENTRY * ColorMask)(GLboolean, GLboolean, GLboolean, GLboolean);
+   void (GLAPIENTRY * DepthMask)(GLboolean);
+   void (GLAPIENTRY * IndexMask)(GLuint);
+   void (GLAPIENTRY * Accum)(GLenum, GLfloat);
+   void (GLAPIENTRY * Disable)(GLenum);
+   void (GLAPIENTRY * Enable)(GLenum);
+   void (GLAPIENTRY * Finish)(void);
+   void (GLAPIENTRY * Flush)(void);
+   void (GLAPIENTRY * PopAttrib)(void);
+   void (GLAPIENTRY * PushAttrib)(GLbitfield);
+   void (GLAPIENTRY * Map1d)(GLenum, GLdouble, GLdouble, GLint, GLint, const GLdouble *);
+   void (GLAPIENTRY * Map1f)(GLenum, GLfloat, GLfloat, GLint, GLint, const GLfloat *);
+   void (GLAPIENTRY * Map2d)(GLenum, GLdouble, GLdouble, GLint, GLint, GLdouble, GLdouble, GLint, GLint, const GLdouble *);
+   void (GLAPIENTRY * Map2f)(GLenum, GLfloat, GLfloat, GLint, GLint, GLfloat, GLfloat, GLint, GLint, const GLfloat *);
+   void (GLAPIENTRY * MapGrid1d)(GLint, GLdouble, GLdouble);
+   void (GLAPIENTRY * MapGrid1f)(GLint, GLfloat, GLfloat);
+   void (GLAPIENTRY * MapGrid2d)(GLint, GLdouble, GLdouble, GLint, GLdouble, GLdouble);
+   void (GLAPIENTRY * MapGrid2f)(GLint, GLfloat, GLfloat, GLint, GLfloat, GLfloat);
+   void (GLAPIENTRY * EvalCoord1d)(GLdouble);
+   void (GLAPIENTRY * EvalCoord1dv)(const GLdouble *);
+   void (GLAPIENTRY * EvalCoord1f)(GLfloat);
+   void (GLAPIENTRY * EvalCoord1fv)(const GLfloat *);
+   void (GLAPIENTRY * EvalCoord2d)(GLdouble, GLdouble);
+   void (GLAPIENTRY * EvalCoord2dv)(const GLdouble *);
+   void (GLAPIENTRY * EvalCoord2f)(GLfloat, GLfloat);
+   void (GLAPIENTRY * EvalCoord2fv)(const GLfloat *);
+   void (GLAPIENTRY * EvalMesh1)(GLenum, GLint, GLint);
+   void (GLAPIENTRY * EvalPoint1)(GLint);
+   void (GLAPIENTRY * EvalMesh2)(GLenum, GLint, GLint, GLint, GLint);
+   void (GLAPIENTRY * EvalPoint2)(GLint, GLint);
+   void (GLAPIENTRY * AlphaFunc)(GLenum, GLclampf);
+   void (GLAPIENTRY * BlendFunc)(GLenum, GLenum);
+   void (GLAPIENTRY * LogicOp)(GLenum);
+   void (GLAPIENTRY * StencilFunc)(GLenum, GLint, GLuint);
+   void (GLAPIENTRY * StencilOp)(GLenum, GLenum, GLenum);
+   void (GLAPIENTRY * DepthFunc)(GLenum);
+   void (GLAPIENTRY * PixelZoom)(GLfloat, GLfloat);
+   void (GLAPIENTRY * PixelTransferf)(GLenum, GLfloat);
+   void (GLAPIENTRY * PixelTransferi)(GLenum, GLint);
+   void (GLAPIENTRY * PixelStoref)(GLenum, GLfloat);
+   void (GLAPIENTRY * PixelStorei)(GLenum, GLint);
+   void (GLAPIENTRY * PixelMapfv)(GLenum, GLint, const GLfloat *);
+   void (GLAPIENTRY * PixelMapuiv)(GLenum, GLint, const GLuint *);
+   void (GLAPIENTRY * PixelMapusv)(GLenum, GLint, const GLushort *);
+   void (GLAPIENTRY * ReadBuffer)(GLenum);
+   void (GLAPIENTRY * CopyPixels)(GLint, GLint, GLsizei, GLsizei, GLenum);
+   void (GLAPIENTRY * ReadPixels)(GLint, GLint, GLsizei, GLsizei, GLenum, GLenum, GLvoid *);
+   void (GLAPIENTRY * DrawPixels)(GLsizei, GLsizei, GLenum, GLenum, const GLvoid *);
+   void (GLAPIENTRY * GetBooleanv)(GLenum, GLboolean *);
+   void (GLAPIENTRY * GetClipPlane)(GLenum, GLdouble *);
+   void (GLAPIENTRY * GetDoublev)(GLenum, GLdouble *);
+   GLenum (GLAPIENTRY * GetError)(void);
+   void (GLAPIENTRY * GetFloatv)(GLenum, GLfloat *);
+   void (GLAPIENTRY * GetIntegerv)(GLenum, GLint *);
+   void (GLAPIENTRY * GetLightfv)(GLenum, GLenum, GLfloat *);
+   void (GLAPIENTRY * GetLightiv)(GLenum, GLenum, GLint *);
+   void (GLAPIENTRY * GetMapdv)(GLenum, GLenum, GLdouble *);
+   void (GLAPIENTRY * GetMapfv)(GLenum, GLenum, GLfloat *);
+   void (GLAPIENTRY * GetMapiv)(GLenum, GLenum, GLint *);
+   void (GLAPIENTRY * GetMaterialfv)(GLenum, GLenum, GLfloat *);
+   void (GLAPIENTRY * GetMaterialiv)(GLenum, GLenum, GLint *);
+   void (GLAPIENTRY * GetPixelMapfv)(GLenum, GLfloat *);
+   void (GLAPIENTRY * GetPixelMapuiv)(GLenum, GLuint *);
+   void (GLAPIENTRY * GetPixelMapusv)(GLenum, GLushort *);
+   void (GLAPIENTRY * GetPolygonStipple)(GLubyte *);
+   const GLubyte * (GLAPIENTRY * GetString)(GLenum);
+   void (GLAPIENTRY * GetTexEnvfv)(GLenum, GLenum, GLfloat *);
+   void (GLAPIENTRY * GetTexEnviv)(GLenum, GLenum, GLint *);
+   void (GLAPIENTRY * GetTexGendv)(GLenum, GLenum, GLdouble *);
+   void (GLAPIENTRY * GetTexGenfv)(GLenum, GLenum, GLfloat *);
+   void (GLAPIENTRY * GetTexGeniv)(GLenum, GLenum, GLint *);
+   void (GLAPIENTRY * GetTexImage)(GLenum, GLint, GLenum, GLenum, GLvoid *);
+   void (GLAPIENTRY * GetTexParameterfv)(GLenum, GLenum, GLfloat *);
+   void (GLAPIENTRY * GetTexParameteriv)(GLenum, GLenum, GLint *);
+   void (GLAPIENTRY * GetTexLevelParameterfv)(GLenum, GLint, GLenum, GLfloat *);
+   void (GLAPIENTRY * GetTexLevelParameteriv)(GLenum, GLint, GLenum, GLint *);
+   GLboolean (GLAPIENTRY * IsEnabled)(GLenum);
+   GLboolean (GLAPIENTRY * IsList)(GLuint);
+   void (GLAPIENTRY * DepthRange)(GLclampd, GLclampd);
+   void (GLAPIENTRY * Frustum)(GLdouble, GLdouble, GLdouble, GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * LoadIdentity)(void);
+   void (GLAPIENTRY * LoadMatrixf)(const GLfloat *);
+   void (GLAPIENTRY * LoadMatrixd)(const GLdouble *);
+   void (GLAPIENTRY * MatrixMode)(GLenum);
+   void (GLAPIENTRY * MultMatrixf)(const GLfloat *);
+   void (GLAPIENTRY * MultMatrixd)(const GLdouble *);
+   void (GLAPIENTRY * Ortho)(GLdouble, GLdouble, GLdouble, GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * PopMatrix)(void);
+   void (GLAPIENTRY * PushMatrix)(void);
+   void (GLAPIENTRY * Rotated)(GLdouble, GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * Rotatef)(GLfloat, GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * Scaled)(GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * Scalef)(GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * Translated)(GLdouble, GLdouble, GLdouble);
+   void (GLAPIENTRY * Translatef)(GLfloat, GLfloat, GLfloat);
+   void (GLAPIENTRY * Viewport)(GLint, GLint, GLsizei, GLsizei);
+   void (GLAPIENTRY * ArrayElement)(GLint);
+   void (GLAPIENTRY * BindTexture)(GLenum, GLuint);
+   void (GLAPIENTRY * ColorPointer)(GLint, GLenum, GLsizei, const GLvoid *);
+   void (GLAPIENTRY * DisableClientState)(GLenum);
+   void (GLAPIENTRY * DrawArrays)(GLenum, GLint, GLsizei);
+   void (GLAPIENTRY * DrawElements)(GLenum, GLsizei, GLenum, const GLvoid *);
+   void (GLAPIENTRY * EdgeFlagPointer)(GLsizei, const GLvoid *);
+   void (GLAPIENTRY * EnableClientState)(GLenum);
+   void (GLAPIENTRY * IndexPointer)(GLenum, GLsizei, const GLvoid *);
+   void (GLAPIENTRY * Indexub)(GLubyte);
+   void (GLAPIENTRY * Indexubv)(const GLubyte *);
+   void (GLAPIENTRY * InterleavedArrays)(GLenum, GLsizei, const GLvoid *);
+   void (GLAPIENTRY * NormalPointer)(GLenum, GLsizei, const GLvoid *);
+   void (GLAPIENTRY * PolygonOffset)(GLfloat, GLfloat);
+   void (GLAPIENTRY * TexCoordPointer)(GLint, GLenum, GLsizei, const GLvoid *);
+   void (GLAPIENTRY * VertexPointer)(GLint, GLenum, GLsizei, const GLvoid *);
+   GLboolean (GLAPIENTRY * AreTexturesResident)(GLsizei, const GLuint *, GLboolean *);
+   void (GLAPIENTRY * CopyTexImage1D)(GLenum, GLint, GLenum, GLint, GLint, GLsizei, GLint);
+   void (GLAPIENTRY * CopyTexImage2D)(GLenum, GLint, GLenum, GLint, GLint, GLsizei, GLsizei, GLint);
+   void (GLAPIENTRY * CopyTexSubImage1D)(GLenum, GLint, GLint, GLint, GLint, GLsizei);
+   void (GLAPIENTRY * CopyTexSubImage2D)(GLenum, GLint, GLint, GLint, GLint, GLint, GLsizei, GLsizei);
+   void (GLAPIENTRY * DeleteTextures)(GLsizei, const GLuint *);
+   void (GLAPIENTRY * GenTextures)(GLsizei, GLuint *);
+   void (GLAPIENTRY * GetPointerv)(GLenum, GLvoid **);
+   GLboolean (GLAPIENTRY * IsTexture)(GLuint);
+   void (GLAPIENTRY * PrioritizeTextures)(GLsizei, const GLuint *, const GLclampf *);
+   void (GLAPIENTRY * TexSubImage1D)(GLenum, GLint, GLint, GLsizei, GLenum, GLenum, const GLvoid *);
+   void (GLAPIENTRY * TexSubImage2D)(GLenum, GLint, GLint, GLint, GLsizei, GLsizei, GLenum, GLenum, const GLvoid *);
+   void (GLAPIENTRY * PopClientAttrib)(void);
+   void (GLAPIENTRY * PushClientAttrib)(GLbitfield);
+};
+
+typedef struct __GLdispatchTableRec GLDISPATCHTABLE;
+
+typedef struct _GLCLTPROCTABLE
+{
+   int cEntries;
+   GLDISPATCHTABLE glDispatchTable;
+} GLCLTPROCTABLE, * PGLCLTPROCTABLE;
+
+typedef VOID (APIENTRY * PFN_SETPROCTABLE)(PGLCLTPROCTABLE);
+
+BOOL APIENTRY
+DrvCopyContext(
+   DHGLRC dhrcSource,
+   DHGLRC dhrcDest,
+   UINT fuMask );
+
+DHGLRC APIENTRY
+DrvCreateLayerContext(
+   HDC hdc,
+   INT iLayerPlane );
+
+DHGLRC APIENTRY
+DrvCreateContext(
+   HDC hdc );
+
+BOOL APIENTRY
+DrvDeleteContext(
+   DHGLRC dhglrc );
+
+BOOL APIENTRY
+DrvDescribeLayerPlane(
+   HDC hdc,
+   INT iPixelFormat,
+   INT iLayerPlane,
+   UINT nBytes,
+   LPLAYERPLANEDESCRIPTOR plpd );
+
+LONG APIENTRY
+DrvDescribePixelFormat(
+   HDC hdc,
+   INT iPixelFormat,
+   ULONG cjpfd,
+   PIXELFORMATDESCRIPTOR *ppfd );
+
+int APIENTRY
+DrvGetLayerPaletteEntries(
+   HDC hdc,
+   INT iLayerPlane,
+   INT iStart,
+   INT cEntries,
+   COLORREF *pcr );
+
+PROC APIENTRY
+DrvGetProcAddress(
+   LPCSTR lpszProc );
+
+BOOL APIENTRY
+DrvRealizeLayerPalette(
+   HDC hdc,
+   INT iLayerPlane,
+   BOOL bRealize );
+
+BOOL APIENTRY
+DrvReleaseContext(
+   DHGLRC dhglrc );
+
+void APIENTRY
+DrvSetCallbackProcs(
+   INT nProcs,
+   PROC *pProcs );
+
+PGLCLTPROCTABLE APIENTRY
+DrvSetContext(
+   HDC hdc,
+   DHGLRC dhglrc,
+   PFN_SETPROCTABLE pfnSetProcTable );
+
+int APIENTRY
+DrvSetLayerPaletteEntries(
+   HDC hdc,
+   INT iLayerPlane,
+   INT iStart,
+   INT cEntries,
+   CONST COLORREF *pcr );
+
+BOOL APIENTRY
+DrvSetPixelFormat(
+   HDC hdc,
+   LONG iPixelFormat );
+
+BOOL APIENTRY
+DrvShareLists(
+   DHGLRC dhglrc1,
+   DHGLRC dhglrc2 );
+
+BOOL APIENTRY
+DrvSwapBuffers(
+   HDC hdc );
+
+BOOL APIENTRY
+DrvSwapLayerBuffers(
+   HDC hdc,
+   UINT fuPlanes );
+
+BOOL APIENTRY
+DrvValidateVersion(
+   ULONG ulVersion );
+
+#endif /* DRV_H */
diff --git a/src/gallium/state_trackers/wgl/opengl32.def b/src/gallium/state_trackers/wgl/opengl32.def
new file mode 100644
index 0000000000..596417ed84
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/opengl32.def
@@ -0,0 +1,388 @@
+EXPORTS
+;	GlmfBeginGlsBlock
+;	GlmfCloseMetaFile
+;	GlmfEndGlsBlock
+;	GlmfEndPlayback
+;	GlmfInitPlayback
+;	GlmfPlayGlsRecord
+	glAccum
+	glAlphaFunc
+	glAreTexturesResident
+	glArrayElement
+	glBegin
+	glBindTexture
+	glBitmap
+	glBlendFunc
+	glCallList
+	glCallLists
+	glClear
+	glClearAccum
+	glClearColor
+	glClearDepth
+	glClearIndex
+	glClearStencil
+	glClipPlane
+	glColor3b
+	glColor3bv
+	glColor3d
+	glColor3dv
+	glColor3f
+	glColor3fv
+	glColor3i
+	glColor3iv
+	glColor3s
+	glColor3sv
+	glColor3ub
+	glColor3ubv
+	glColor3ui
+	glColor3uiv
+	glColor3us
+	glColor3usv
+	glColor4b
+	glColor4bv
+	glColor4d
+	glColor4dv
+	glColor4f
+	glColor4fv
+	glColor4i
+	glColor4iv
+	glColor4s
+	glColor4sv
+	glColor4ub
+	glColor4ubv
+	glColor4ui
+	glColor4uiv
+	glColor4us
+	glColor4usv
+	glColorMask
+	glColorMaterial
+	glColorPointer
+	glCopyPixels
+	glCopyTexImage1D
+	glCopyTexImage2D
+	glCopyTexSubImage1D
+	glCopyTexSubImage2D
+	glCullFace
+;	glDebugEntry
+	glDeleteLists
+	glDeleteTextures
+	glDepthFunc
+	glDepthMask
+	glDepthRange
+	glDisable
+	glDisableClientState
+	glDrawArrays
+	glDrawBuffer
+	glDrawElements
+	glDrawPixels
+	glEdgeFlag
+	glEdgeFlagPointer
+	glEdgeFlagv
+	glEnable
+	glEnableClientState
+	glEnd
+	glEndList
+	glEvalCoord1d
+	glEvalCoord1dv
+	glEvalCoord1f
+	glEvalCoord1fv
+	glEvalCoord2d
+	glEvalCoord2dv
+	glEvalCoord2f
+	glEvalCoord2fv
+	glEvalMesh1
+	glEvalMesh2
+	glEvalPoint1
+	glEvalPoint2
+	glFeedbackBuffer
+	glFinish
+	glFlush
+	glFogf
+	glFogfv
+	glFogi
+	glFogiv
+	glFrontFace
+	glFrustum
+	glGenLists
+	glGenTextures
+	glGetBooleanv
+	glGetClipPlane
+	glGetDoublev
+	glGetError
+	glGetFloatv
+	glGetIntegerv
+	glGetLightfv
+	glGetLightiv
+	glGetMapdv
+	glGetMapfv
+	glGetMapiv
+	glGetMaterialfv
+	glGetMaterialiv
+	glGetPixelMapfv
+	glGetPixelMapuiv
+	glGetPixelMapusv
+	glGetPointerv
+	glGetPolygonStipple
+	glGetString
+	glGetTexEnvfv
+	glGetTexEnviv
+	glGetTexGendv
+	glGetTexGenfv
+	glGetTexGeniv
+	glGetTexImage
+	glGetTexLevelParameterfv
+	glGetTexLevelParameteriv
+	glGetTexParameterfv
+	glGetTexParameteriv
+	glHint
+	glIndexMask
+	glIndexPointer
+	glIndexd
+	glIndexdv
+	glIndexf
+	glIndexfv
+	glIndexi
+	glIndexiv
+	glIndexs
+	glIndexsv
+	glIndexub
+	glIndexubv
+	glInitNames
+	glInterleavedArrays
+	glIsEnabled
+	glIsList
+	glIsTexture
+	glLightModelf
+	glLightModelfv
+	glLightModeli
+	glLightModeliv
+	glLightf
+	glLightfv
+	glLighti
+	glLightiv
+	glLineStipple
+	glLineWidth
+	glListBase
+	glLoadIdentity
+	glLoadMatrixd
+	glLoadMatrixf
+	glLoadName
+	glLogicOp
+	glMap1d
+	glMap1f
+	glMap2d
+	glMap2f
+	glMapGrid1d
+	glMapGrid1f
+	glMapGrid2d
+	glMapGrid2f
+	glMaterialf
+	glMaterialfv
+	glMateriali
+	glMaterialiv
+	glMatrixMode
+	glMultMatrixd
+	glMultMatrixf
+	glNewList
+	glNormal3b
+	glNormal3bv
+	glNormal3d
+	glNormal3dv
+	glNormal3f
+	glNormal3fv
+	glNormal3i
+	glNormal3iv
+	glNormal3s
+	glNormal3sv
+	glNormalPointer
+	glOrtho
+	glPassThrough
+	glPixelMapfv
+	glPixelMapuiv
+	glPixelMapusv
+	glPixelStoref
+	glPixelStorei
+	glPixelTransferf
+	glPixelTransferi
+	glPixelZoom
+	glPointSize
+	glPolygonMode
+	glPolygonOffset
+	glPolygonStipple
+	glPopAttrib
+	glPopClientAttrib
+	glPopMatrix
+	glPopName
+	glPrioritizeTextures
+	glPushAttrib
+	glPushClientAttrib
+	glPushMatrix
+	glPushName
+	glRasterPos2d
+	glRasterPos2dv
+	glRasterPos2f
+	glRasterPos2fv
+	glRasterPos2i
+	glRasterPos2iv
+	glRasterPos2s
+	glRasterPos2sv
+	glRasterPos3d
+	glRasterPos3dv
+	glRasterPos3f
+	glRasterPos3fv
+	glRasterPos3i
+	glRasterPos3iv
+	glRasterPos3s
+	glRasterPos3sv
+	glRasterPos4d
+	glRasterPos4dv
+	glRasterPos4f
+	glRasterPos4fv
+	glRasterPos4i
+	glRasterPos4iv
+	glRasterPos4s
+	glRasterPos4sv
+	glReadBuffer
+	glReadPixels
+	glRectd
+	glRectdv
+	glRectf
+	glRectfv
+	glRecti
+	glRectiv
+	glRects
+	glRectsv
+	glRenderMode
+	glRotated
+	glRotatef
+	glScaled
+	glScalef
+	glScissor
+	glSelectBuffer
+	glShadeModel
+	glStencilFunc
+	glStencilMask
+	glStencilOp
+	glTexCoord1d
+	glTexCoord1dv
+	glTexCoord1f
+	glTexCoord1fv
+	glTexCoord1i
+	glTexCoord1iv
+	glTexCoord1s
+	glTexCoord1sv
+	glTexCoord2d
+	glTexCoord2dv
+	glTexCoord2f
+	glTexCoord2fv
+	glTexCoord2i
+	glTexCoord2iv
+	glTexCoord2s
+	glTexCoord2sv
+	glTexCoord3d
+	glTexCoord3dv
+	glTexCoord3f
+	glTexCoord3fv
+	glTexCoord3i
+	glTexCoord3iv
+	glTexCoord3s
+	glTexCoord3sv
+	glTexCoord4d
+	glTexCoord4dv
+	glTexCoord4f
+	glTexCoord4fv
+	glTexCoord4i
+	glTexCoord4iv
+	glTexCoord4s
+	glTexCoord4sv
+	glTexCoordPointer
+	glTexEnvf
+	glTexEnvfv
+	glTexEnvi
+	glTexEnviv
+	glTexGend
+	glTexGendv
+	glTexGenf
+	glTexGenfv
+	glTexGeni
+	glTexGeniv
+	glTexImage1D
+	glTexImage2D
+	glTexParameterf
+	glTexParameterfv
+	glTexParameteri
+	glTexParameteriv
+	glTexSubImage1D
+	glTexSubImage2D
+	glTranslated
+	glTranslatef
+	glVertex2d
+	glVertex2dv
+	glVertex2f
+	glVertex2fv
+	glVertex2i
+	glVertex2iv
+	glVertex2s
+	glVertex2sv
+	glVertex3d
+	glVertex3dv
+	glVertex3f
+	glVertex3fv
+	glVertex3i
+	glVertex3iv
+	glVertex3s
+	glVertex3sv
+	glVertex4d
+	glVertex4dv
+	glVertex4f
+	glVertex4fv
+	glVertex4i
+	glVertex4iv
+	glVertex4s
+	glVertex4sv
+	glVertexPointer
+	glViewport
+	wglChoosePixelFormat
+	wglCopyContext
+	wglCreateContext
+	wglCreateLayerContext
+	wglDeleteContext
+	wglDescribeLayerPlane
+	wglDescribePixelFormat
+	wglGetCurrentContext
+	wglGetCurrentDC
+;	wglGetDefaultProcAddress
+	wglGetLayerPaletteEntries
+	wglGetPixelFormat
+	wglGetProcAddress
+	wglMakeCurrent
+	wglRealizeLayerPalette
+	wglSetLayerPaletteEntries
+	wglSetPixelFormat
+	wglShareLists
+	wglSwapBuffers
+	wglSwapLayerBuffers
+;	wglSwapMultipleBuffers
+	wglUseFontBitmapsA
+	wglUseFontBitmapsW
+	wglUseFontOutlinesA
+	wglUseFontOutlinesW
+	wglGetExtensionsStringARB
+	DrvCopyContext
+	DrvCreateContext
+	DrvCreateLayerContext
+	DrvDeleteContext
+	DrvDescribeLayerPlane
+	DrvDescribePixelFormat
+	DrvGetLayerPaletteEntries
+	DrvGetProcAddress
+	DrvRealizeLayerPalette
+	DrvReleaseContext
+	DrvSetCallbackProcs
+	DrvSetContext
+	DrvSetLayerPaletteEntries
+	DrvSetPixelFormat
+	DrvShareLists
+	DrvSwapBuffers
+	DrvSwapLayerBuffers
+	DrvValidateVersion
diff --git a/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.c b/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.c
new file mode 100644
index 0000000000..b3934cb464
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.c
@@ -0,0 +1,42 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include <windows.h>
+
+#include "stw_arbextensionsstring.h"
+
+WINGDIAPI const char * APIENTRY
+wglGetExtensionsStringARB(
+   HDC hdc )
+{
+   (void) hdc;
+
+   return
+      "WGL_ARB_extensions_string "
+      "WGL_ARB_multisample "
+      "WGL_ARB_pixel_format";
+}
diff --git a/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.h b/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.h
new file mode 100644
index 0000000000..a0e4c5d98e
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_arbextensionsstring.h
@@ -0,0 +1,35 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef WGL_ARBEXTENSIONSSTRING_H
+#define WGL_ARBEXTENSIONSSTRING_H
+
+WINGDIAPI const char * APIENTRY
+wglGetExtensionsStringARB(
+   HDC hdc );
+
+#endif /* WGL_ARBEXTENSIONSSTRING_H */
diff --git a/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.c b/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.c
new file mode 100644
index 0000000000..f563635420
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.c
@@ -0,0 +1,519 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <windows.h>
+
+#include "pipe/p_compiler.h"
+#include "util/u_memory.h"
+#include "stw_public.h"
+#include "stw_pixelformat.h"
+#include "stw_arbpixelformat.h"
+
+#define WGL_NUMBER_PIXEL_FORMATS_ARB            0x2000
+#define WGL_DRAW_TO_WINDOW_ARB                  0x2001
+#define WGL_DRAW_TO_BITMAP_ARB                  0x2002
+#define WGL_ACCELERATION_ARB                    0x2003
+#define WGL_NEED_PALETTE_ARB                    0x2004
+#define WGL_NEED_SYSTEM_PALETTE_ARB             0x2005
+#define WGL_SWAP_LAYER_BUFFERS_ARB              0x2006
+#define WGL_SWAP_METHOD_ARB                     0x2007
+#define WGL_NUMBER_OVERLAYS_ARB                 0x2008
+#define WGL_NUMBER_UNDERLAYS_ARB                0x2009
+#define WGL_TRANSPARENT_ARB                     0x200A
+#define WGL_TRANSPARENT_RED_VALUE_ARB           0x2037
+#define WGL_TRANSPARENT_GREEN_VALUE_ARB         0x2038
+#define WGL_TRANSPARENT_BLUE_VALUE_ARB          0x2039
+#define WGL_TRANSPARENT_ALPHA_VALUE_ARB         0x203A
+#define WGL_TRANSPARENT_INDEX_VALUE_ARB         0x203B
+#define WGL_SHARE_DEPTH_ARB                     0x200C
+#define WGL_SHARE_STENCIL_ARB                   0x200D
+#define WGL_SHARE_ACCUM_ARB                     0x200E
+#define WGL_SUPPORT_GDI_ARB                     0x200F
+#define WGL_SUPPORT_OPENGL_ARB                  0x2010
+#define WGL_DOUBLE_BUFFER_ARB                   0x2011
+#define WGL_STEREO_ARB                          0x2012
+#define WGL_PIXEL_TYPE_ARB                      0x2013
+#define WGL_COLOR_BITS_ARB                      0x2014
+#define WGL_RED_BITS_ARB                        0x2015
+#define WGL_RED_SHIFT_ARB                       0x2016
+#define WGL_GREEN_BITS_ARB                      0x2017
+#define WGL_GREEN_SHIFT_ARB                     0x2018
+#define WGL_BLUE_BITS_ARB                       0x2019
+#define WGL_BLUE_SHIFT_ARB                      0x201A
+#define WGL_ALPHA_BITS_ARB                      0x201B
+#define WGL_ALPHA_SHIFT_ARB                     0x201C
+#define WGL_ACCUM_BITS_ARB                      0x201D
+#define WGL_ACCUM_RED_BITS_ARB                  0x201E
+#define WGL_ACCUM_GREEN_BITS_ARB                0x201F
+#define WGL_ACCUM_BLUE_BITS_ARB                 0x2020
+#define WGL_ACCUM_ALPHA_BITS_ARB                0x2021
+#define WGL_DEPTH_BITS_ARB                      0x2022
+#define WGL_STENCIL_BITS_ARB                    0x2023
+#define WGL_AUX_BUFFERS_ARB                     0x2024
+
+#define WGL_NO_ACCELERATION_ARB                 0x2025
+#define WGL_GENERIC_ACCELERATION_ARB            0x2026
+#define WGL_FULL_ACCELERATION_ARB               0x2027
+
+#define WGL_SWAP_EXCHANGE_ARB                   0x2028
+#define WGL_SWAP_COPY_ARB                       0x2029
+#define WGL_SWAP_UNDEFINED_ARB                  0x202A
+
+#define WGL_TYPE_RGBA_ARB                       0x202B
+#define WGL_TYPE_COLORINDEX_ARB                 0x202C
+
+/* From arb_multisample:
+ */
+#define WGL_SAMPLE_BUFFERS_ARB               0x2041
+#define WGL_SAMPLES_ARB                      0x2042
+
+
+static boolean
+query_attrib(
+   int iPixelFormat,
+   int iLayerPlane,
+   int attrib,
+   int *pvalue )
+{
+   uint count;
+   uint index;
+   const struct pixelformat_info *pf;
+
+   count = pixelformat_get_extended_count();
+
+   if (attrib == WGL_NUMBER_PIXEL_FORMATS_ARB) {
+      *pvalue = (int) count;
+      return TRUE;
+   }
+
+   index = (uint) iPixelFormat - 1;
+   if (index >= count)
+      return FALSE;
+
+   pf = pixelformat_get_info( index );
+
+   switch (attrib) {
+   case WGL_DRAW_TO_WINDOW_ARB:
+      *pvalue = TRUE;
+      return TRUE;
+
+   case WGL_DRAW_TO_BITMAP_ARB:
+      *pvalue = FALSE;
+      return TRUE;
+
+   case WGL_NEED_PALETTE_ARB:
+      *pvalue = FALSE;
+      return TRUE;
+
+   case WGL_NEED_SYSTEM_PALETTE_ARB:
+      *pvalue = FALSE;
+      return TRUE;
+
+   case WGL_SWAP_METHOD_ARB:
+      if (pf->flags & PF_FLAG_DOUBLEBUFFER)
+         *pvalue = WGL_SWAP_COPY_ARB;
+      else
+         *pvalue = WGL_SWAP_UNDEFINED_ARB;
+      return TRUE;
+
+   case WGL_SWAP_LAYER_BUFFERS_ARB:
+      *pvalue = FALSE;
+      return TRUE;
+
+   case WGL_NUMBER_OVERLAYS_ARB:
+      *pvalue = 0;
+      return TRUE;
+
+   case WGL_NUMBER_UNDERLAYS_ARB:
+      *pvalue = 0;
+      return TRUE;
+   }
+
+   if (iLayerPlane != 0)
+      return FALSE;
+
+   switch (attrib) {
+   case WGL_ACCELERATION_ARB:
+      *pvalue = WGL_FULL_ACCELERATION_ARB;
+      break;
+
+   case WGL_TRANSPARENT_ARB:
+      *pvalue = FALSE;
+      break;
+
+   case WGL_TRANSPARENT_RED_VALUE_ARB:
+   case WGL_TRANSPARENT_GREEN_VALUE_ARB:
+   case WGL_TRANSPARENT_BLUE_VALUE_ARB:
+   case WGL_TRANSPARENT_ALPHA_VALUE_ARB:
+   case WGL_TRANSPARENT_INDEX_VALUE_ARB:
+      break;
+
+   case WGL_SHARE_DEPTH_ARB:
+   case WGL_SHARE_STENCIL_ARB:
+   case WGL_SHARE_ACCUM_ARB:
+      *pvalue = TRUE;
+      break;
+
+   case WGL_SUPPORT_GDI_ARB:
+      *pvalue = FALSE;
+      break;
+
+   case WGL_SUPPORT_OPENGL_ARB:
+      *pvalue = TRUE;
+      break;
+
+   case WGL_DOUBLE_BUFFER_ARB:
+      if (pf->flags & PF_FLAG_DOUBLEBUFFER)
+         *pvalue = TRUE;
+      else
+         *pvalue = FALSE;
+      break;
+
+   case WGL_STEREO_ARB:
+      *pvalue = FALSE;
+      break;
+
+   case WGL_PIXEL_TYPE_ARB:
+      *pvalue = WGL_TYPE_RGBA_ARB;
+      break;
+
+   case WGL_COLOR_BITS_ARB:
+      *pvalue = (int) (pf->color.redbits + pf->color.greenbits + pf->color.bluebits);
+      break;
+
+   case WGL_RED_BITS_ARB:
+      *pvalue = (int) pf->color.redbits;
+      break;
+
+   case WGL_RED_SHIFT_ARB:
+      *pvalue = (int) pf->color.redshift;
+      break;
+
+   case WGL_GREEN_BITS_ARB:
+      *pvalue = (int) pf->color.greenbits;
+      break;
+
+   case WGL_GREEN_SHIFT_ARB:
+      *pvalue = (int) pf->color.greenshift;
+      break;
+
+   case WGL_BLUE_BITS_ARB:
+      *pvalue = (int) pf->color.bluebits;
+      break;
+
+   case WGL_BLUE_SHIFT_ARB:
+      *pvalue = (int) pf->color.blueshift;
+      break;
+
+   case WGL_ALPHA_BITS_ARB:
+      *pvalue = (int) pf->alpha.alphabits;
+      break;
+
+   case WGL_ALPHA_SHIFT_ARB:
+      *pvalue = (int) pf->alpha.alphashift;
+      break;
+
+   case WGL_ACCUM_BITS_ARB:
+   case WGL_ACCUM_RED_BITS_ARB:
+   case WGL_ACCUM_GREEN_BITS_ARB:
+   case WGL_ACCUM_BLUE_BITS_ARB:
+   case WGL_ACCUM_ALPHA_BITS_ARB:
+      *pvalue = 0;
+      break;
+
+   case WGL_DEPTH_BITS_ARB:
+      *pvalue = (int) pf->depth.depthbits;
+      break;
+
+   case WGL_STENCIL_BITS_ARB:
+      *pvalue = (int) pf->depth.stencilbits;
+      break;
+
+   case WGL_AUX_BUFFERS_ARB:
+      *pvalue = 0;
+      break;
+
+   case WGL_SAMPLE_BUFFERS_ARB:
+      if (pf->flags & PF_FLAG_MULTISAMPLED)
+         *pvalue = stw_query_sample_buffers();
+      else
+         *pvalue = 0;
+      break;
+
+   case WGL_SAMPLES_ARB:
+      if (pf->flags & PF_FLAG_MULTISAMPLED)
+         *pvalue = stw_query_samples();
+      else
+         *pvalue = 0;
+      break;
+
+   default:
+      return FALSE;
+   }
+
+   return TRUE;
+}
+
+struct attrib_match_info
+{
+   int attribute;
+   int weight;
+   BOOL exact;
+};
+
+static struct attrib_match_info attrib_match[] = {
+
+   /* WGL_ARB_pixel_format */
+   { WGL_DRAW_TO_WINDOW_ARB,      0, TRUE },
+   { WGL_DRAW_TO_BITMAP_ARB,      0, TRUE },
+   { WGL_ACCELERATION_ARB,        0, TRUE },
+   { WGL_NEED_PALETTE_ARB,        0, TRUE },
+   { WGL_NEED_SYSTEM_PALETTE_ARB, 0, TRUE },
+   { WGL_SWAP_LAYER_BUFFERS_ARB,  0, TRUE },
+   { WGL_SWAP_METHOD_ARB,         0, TRUE },
+   { WGL_NUMBER_OVERLAYS_ARB,     4, FALSE },
+   { WGL_NUMBER_UNDERLAYS_ARB,    4, FALSE },
+   /*{ WGL_SHARE_DEPTH_ARB,         0, TRUE },*/     /* no overlays -- ignore */
+   /*{ WGL_SHARE_STENCIL_ARB,       0, TRUE },*/   /* no overlays -- ignore */
+   /*{ WGL_SHARE_ACCUM_ARB,         0, TRUE },*/     /* no overlays -- ignore */
+   { WGL_SUPPORT_GDI_ARB,         0, TRUE },
+   { WGL_SUPPORT_OPENGL_ARB,      0, TRUE },
+   { WGL_DOUBLE_BUFFER_ARB,       0, TRUE },
+   { WGL_STEREO_ARB,              0, TRUE },
+   { WGL_PIXEL_TYPE_ARB,          0, TRUE },
+   { WGL_COLOR_BITS_ARB,          1, FALSE },
+   { WGL_RED_BITS_ARB,            1, FALSE },
+   { WGL_GREEN_BITS_ARB,          1, FALSE },
+   { WGL_BLUE_BITS_ARB,           1, FALSE },
+   { WGL_ALPHA_BITS_ARB,          1, FALSE },
+   { WGL_ACCUM_BITS_ARB,          1, FALSE },
+   { WGL_ACCUM_RED_BITS_ARB,      1, FALSE },
+   { WGL_ACCUM_GREEN_BITS_ARB,    1, FALSE },
+   { WGL_ACCUM_BLUE_BITS_ARB,     1, FALSE },
+   { WGL_ACCUM_ALPHA_BITS_ARB,    1, FALSE },
+   { WGL_DEPTH_BITS_ARB,          1, FALSE },
+   { WGL_STENCIL_BITS_ARB,        1, FALSE },
+   { WGL_AUX_BUFFERS_ARB,         2, FALSE },
+
+   /* WGL_ARB_multisample */
+   { WGL_SAMPLE_BUFFERS_ARB,      2, FALSE },
+   { WGL_SAMPLES_ARB,             2, FALSE }
+};
+
+struct pixelformat_score
+{
+   int points;
+   uint index;
+};
+
+static BOOL
+score_pixelformats(
+   struct pixelformat_score *scores,
+   uint count,
+   int attribute,
+   int expected_value )
+{
+   uint i;
+   struct attrib_match_info *ami = NULL;
+   uint index;
+
+   /* Find out if a given attribute should be considered for score calculation.
+    */
+   for (i = 0; i < sizeof( attrib_match ) / sizeof( attrib_match[0] ); i++) {
+      if (attrib_match[i].attribute == attribute) {
+         ami = &attrib_match[i];
+         break;
+      }
+   }
+   if (ami == NULL)
+      return TRUE;
+
+   /* Iterate all pixelformats, query the requested attribute and calculate
+    * score points.
+    */
+   for (index = 0; index < count; index++) {
+      int actual_value;
+
+      if (!query_attrib( index + 1, 0, attribute, &actual_value ))
+         return FALSE;
+
+      if (ami->exact) {
+         /* For an exact match criteria, if the actual and expected values differ,
+          * the score is set to 0 points, effectively removing the pixelformat
+          * from a list of matching pixelformats.
+          */
+         if (actual_value != expected_value)
+            scores[index].points = 0;
+      }
+      else {
+         /* For a minimum match criteria, if the actual value is smaller than the expected
+          * value, the pixelformat is rejected (score set to 0). However, if the actual
+          * value is bigger, the pixelformat is given a penalty to favour pixelformats that
+          * more closely match the expected values.
+          */
+         if (actual_value < expected_value)
+            scores[index].points = 0;
+         else if (actual_value > expected_value)
+            scores[index].points -= (actual_value - expected_value) * ami->weight;
+      }
+   }
+
+   return TRUE;
+}
+
+WINGDIAPI BOOL APIENTRY
+wglChoosePixelFormatARB(
+   HDC hdc,
+   const int *piAttribIList,
+   const FLOAT *pfAttribFList,
+   UINT nMaxFormats,
+   int *piFormats,
+   UINT *nNumFormats )
+{
+   uint count;
+   struct pixelformat_score *scores;
+   uint i;
+
+   *nNumFormats = 0;
+
+   /* Allocate and initialize pixelformat score table -- better matches
+    * have higher scores. Start with a high score and take out penalty
+    * points for a mismatch when the match does not have to be exact.
+    * Set a score to 0 if there is a mismatch for an exact match criteria.
+    */
+   count = pixelformat_get_extended_count();
+   scores = (struct pixelformat_score *) MALLOC( count * sizeof( struct pixelformat_score ) );
+   if (scores == NULL)
+      return FALSE;
+   for (i = 0; i < count; i++) {
+      scores[i].points = 0x7fffffff;
+      scores[i].index = i;
+   }
+
+   /* Given the attribute list calculate a score for each pixelformat.
+    */
+   if (piAttribIList != NULL) {
+      while (*piAttribIList != 0) {
+         if (!score_pixelformats( scores, count, piAttribIList[0], piAttribIList[1] )) {
+            FREE( scores );
+            return FALSE;
+         }
+         piAttribIList += 2;
+      }
+   }
+   if (pfAttribFList != NULL) {
+      while (*pfAttribFList != 0) {
+         if (!score_pixelformats( scores, count, (int) pfAttribFList[0], (int) pfAttribFList[1] )) {
+            FREE( scores );
+            return FALSE;
+         }
+         pfAttribFList += 2;
+      }
+   }
+
+   /* Bubble-sort the resulting scores. Pixelformats with higher scores go first.
+    * TODO: Find out if there are any patent issues with it.
+    */
+   if (count > 1) {
+      uint n = count;
+      boolean swapped;
+
+      do {
+         swapped = FALSE;
+         for (i = 1; i < n; i++) {
+            if (scores[i - 1].points < scores[i].points) {
+               struct pixelformat_score score = scores[i - 1];
+
+               scores[i - 1] = scores[i];
+               scores[i] = score;
+               swapped = TRUE;
+            }
+         }
+         n--;
+      }
+      while (swapped);
+   }
+
+   /* Return a list of pixelformats that are the best match.
+    * Reject pixelformats with non-positive scores.
+    */
+   for (i = 0; i < count; i++) {
+      if (scores[i].points > 0) {
+         if (*nNumFormats < nMaxFormats)
+            piFormats[*nNumFormats] = scores[i].index + 1;
+         (*nNumFormats)++;
+      }
+   }
+
+   FREE( scores );
+   return TRUE;
+}
+
+WINGDIAPI BOOL APIENTRY
+wglGetPixelFormatAttribfvARB(
+   HDC hdc,
+   int iPixelFormat,
+   int iLayerPlane,
+   UINT nAttributes,
+   const int *piAttributes,
+   FLOAT *pfValues )
+{
+   UINT i;
+
+   (void) hdc;
+
+   for (i = 0; i < nAttributes; i++) {
+      int value;
+
+      if (!query_attrib( iPixelFormat, iLayerPlane, piAttributes[i], &value ))
+         return FALSE;
+      pfValues[i] = (FLOAT) value;
+   }
+
+   return TRUE;
+}
+
+WINGDIAPI BOOL APIENTRY
+wglGetPixelFormatAttribivARB(
+   HDC hdc,
+   int iPixelFormat,
+   int iLayerPlane,
+   UINT nAttributes,
+   const int *piAttributes,
+   int *piValues )
+{
+   UINT i;
+
+   (void) hdc;
+
+   for (i = 0; i < nAttributes; i++) {
+      if (!query_attrib( iPixelFormat, iLayerPlane, piAttributes[i], &piValues[i] ))
+         return FALSE;
+   }
+
+   return TRUE;
+}
diff --git a/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.h b/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.h
new file mode 100644
index 0000000000..a6c4259942
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_arbpixelformat.h
@@ -0,0 +1,61 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef WGL_ARBPIXELFORMAT_H
+#define WGL_ARBPIXELFORMAT_H
+
+
+/* Extension functions for get_proc_address:
+ */
+WINGDIAPI BOOL APIENTRY
+wglChoosePixelFormatARB(
+   HDC hdc,
+   const int *piAttribIList,
+   const FLOAT *pfAttribFList,
+   UINT nMaxFormats,
+   int *piFormats,
+   UINT *nNumFormats );
+
+WINGDIAPI BOOL APIENTRY
+wglGetPixelFormatAttribfvARB(
+   HDC hdc,
+   int iPixelFormat,
+   int iLayerPlane,
+   UINT nAttributes,
+   const int *piAttributes,
+   FLOAT *pfValues );
+
+WINGDIAPI BOOL APIENTRY
+wglGetPixelFormatAttribivARB(
+   HDC hdc,
+   int iPixelFormat,
+   int iLayerPlane,
+   UINT nAttributes,
+   const int *piAttributes,
+   int *piValues );
+
+#endif /* WGL_ARBPIXELFORMAT_H */
diff --git a/src/gallium/state_trackers/wgl/shared/stw_context.c b/src/gallium/state_trackers/wgl/shared/stw_context.c
new file mode 100644
index 0000000000..2abf97b5ad
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_context.c
@@ -0,0 +1,292 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <windows.h>
+
+#include "main/mtypes.h"
+#include "main/context.h"
+#include "pipe/p_compiler.h"
+#include "pipe/p_context.h"
+#include "state_tracker/st_context.h"
+#include "state_tracker/st_public.h"
+#include "shared/stw_device.h"
+#include "shared/stw_winsys.h"
+#include "shared/stw_framebuffer.h"
+#include "shared/stw_pixelformat.h"
+#include "stw_public.h"
+#include "stw_context.h"
+
+static struct stw_context *ctx_head = NULL;
+
+static HDC current_hdc = NULL;
+static struct stw_context *current_hrc = NULL;
+
+BOOL
+stw_copy_context(
+   struct stw_context *src,
+   struct stw_context *dst,
+   UINT mask )
+{
+   (void) src;
+   (void) dst;
+   (void) mask;
+
+   return FALSE;
+}
+
+struct stw_context *
+stw_create_context(
+   HDC hdc,
+   int iLayerPlane )
+{
+   uint pfi;
+   const struct pixelformat_info *pf = NULL;
+   struct stw_context *ctx = NULL;
+   GLvisual *visual = NULL;
+   struct pipe_context *pipe = NULL;
+
+   if (iLayerPlane != 0)
+      return NULL;
+
+   pfi = stw_pixelformat_get( hdc );
+   if (pfi == 0)
+      return NULL;
+
+   pf = pixelformat_get_info( pfi - 1 );
+
+   ctx = CALLOC_STRUCT( stw_context );
+   if (ctx == NULL)
+      return NULL;
+
+   ctx->hdc = hdc;
+   ctx->color_bits = GetDeviceCaps( ctx->hdc, BITSPIXEL );
+
+   /* Create visual based on flags
+    */
+   visual = _mesa_create_visual(
+      GL_TRUE,
+      (pf->flags & PF_FLAG_DOUBLEBUFFER) ? GL_TRUE : GL_FALSE,
+      GL_FALSE,
+      pf->color.redbits,
+      pf->color.greenbits,
+      pf->color.bluebits,
+      pf->alpha.alphabits,
+      0,
+      pf->depth.depthbits,
+      pf->depth.stencilbits,
+      0,
+      0,
+      0,
+      0,
+      (pf->flags & PF_FLAG_MULTISAMPLED) ? stw_query_samples() : 0 );
+   if (visual == NULL) 
+      goto fail;
+
+   pipe = stw_dev->stw_winsys->create_context( stw_dev->screen );
+   if (pipe == NULL) 
+      goto fail;
+
+   assert(!pipe->priv);
+   pipe->priv = hdc;
+
+   ctx->st = st_create_context( pipe, visual, NULL );
+   if (ctx->st == NULL) 
+      goto fail;
+
+   ctx->st->ctx->DriverCtx = ctx;
+
+   ctx->next = ctx_head;
+   ctx_head = ctx;
+
+   return ctx;
+
+fail:
+   if (visual)
+      _mesa_destroy_visual( visual );
+   
+   if (pipe)
+      pipe->destroy( pipe );
+      
+   FREE( ctx );
+   return NULL;
+}
+
+
+BOOL
+stw_delete_context(
+   struct stw_context *hglrc )
+{
+   struct stw_context **link = &ctx_head;
+   struct stw_context *ctx = ctx_head;
+
+   while (ctx != NULL) {
+      if (ctx == hglrc) {
+         GLcontext *glctx = ctx->st->ctx;
+         GET_CURRENT_CONTEXT( glcurctx );
+         struct stw_framebuffer *fb;
+
+         /* Unbind current if deleting current context.
+          */
+         if (glcurctx == glctx)
+            st_make_current( NULL, NULL, NULL );
+
+         fb = framebuffer_from_hdc( ctx->hdc );
+         if (fb)
+            framebuffer_destroy( fb );
+
+         if (WindowFromDC( ctx->hdc ) != NULL)
+            ReleaseDC( WindowFromDC( ctx->hdc ), ctx->hdc );
+
+         st_destroy_context( ctx->st );
+
+         *link = ctx->next;
+         FREE( ctx );
+         return TRUE;
+      }
+
+      link = &ctx->next;
+      ctx = ctx->next;
+   }
+
+   return FALSE;
+}
+
+/* Find the width and height of the window named by hdc.
+ */
+static void
+get_window_size( HDC hdc, GLuint *width, GLuint *height )
+{
+   if (WindowFromDC( hdc )) {
+      RECT rect;
+
+      GetClientRect( WindowFromDC( hdc ), &rect );
+      *width = rect.right - rect.left;
+      *height = rect.bottom - rect.top;
+   }
+   else {
+      *width = GetDeviceCaps( hdc, HORZRES );
+      *height = GetDeviceCaps( hdc, VERTRES );
+   }
+}
+
+struct stw_context *
+stw_get_current_context( void )
+{
+   return current_hrc;
+}
+
+HDC
+stw_get_current_dc( void )
+{
+    return current_hdc;
+}
+
+BOOL
+stw_make_current(
+   HDC hdc,
+   struct stw_context *hglrc )
+{
+   struct stw_context *ctx = ctx_head;
+   GET_CURRENT_CONTEXT( glcurctx );
+   struct stw_framebuffer *fb;
+   GLuint width = 0;
+   GLuint height = 0;
+
+   current_hdc = hdc;
+   current_hrc = hglrc;
+
+   if (hdc == NULL || hglrc == NULL) {
+      st_make_current( NULL, NULL, NULL );
+      return TRUE;
+   }
+
+   while (ctx != NULL) {
+      if (ctx == hglrc)
+         break;
+      ctx = ctx->next;
+   }
+   if (ctx == NULL)
+      return FALSE;
+
+   /* Return if already current.
+    */
+   if (glcurctx != NULL) {
+      struct stw_context *curctx = (struct stw_context *) glcurctx->DriverCtx;
+
+      if (curctx != NULL && curctx == ctx && ctx->hdc == hdc)
+         return TRUE;
+   }
+
+   fb = framebuffer_from_hdc( hdc );
+
+   if (hdc != NULL)
+      get_window_size( hdc, &width, &height );
+
+   /* Lazy creation of framebuffers.
+    */
+   if (fb == NULL && ctx != NULL && hdc != NULL) {
+      GLvisual *visual = &ctx->st->ctx->Visual;
+
+      fb = framebuffer_create( hdc, visual, width, height );
+      if (fb == NULL)
+         return FALSE;
+
+      fb->dib_hDC = CreateCompatibleDC( hdc );
+      fb->hbmDIB = NULL;
+      fb->pbPixels = NULL;
+   }
+
+   if (ctx && fb) {
+      st_make_current( ctx->st, fb->stfb, fb->stfb );
+      framebuffer_resize( fb, width, height );
+      ctx->hdc = hdc;
+      ctx->st->pipe->priv = hdc;
+   }
+   else {
+      /* Detach */
+      st_make_current( NULL, NULL, NULL );
+   }
+
+   return TRUE;
+}
+
+struct stw_context *
+stw_context_from_hdc(
+   HDC hdc )
+{
+   struct stw_context *ctx = ctx_head;
+
+   while (ctx != NULL) {
+      if (ctx->hdc == hdc)
+         return ctx;
+      ctx = ctx->next;
+   }
+   return NULL;
+}
+
+
+
diff --git a/src/gallium/state_trackers/wgl/shared/stw_context.h b/src/gallium/state_trackers/wgl/shared/stw_context.h
new file mode 100644
index 0000000000..89a8f900d8
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_context.h
@@ -0,0 +1,53 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef STW_CONTEXT_H
+#define STW_CONTEXT_H
+
+#include <windows.h>
+
+struct st_context;
+
+struct stw_context
+{
+   struct st_context *st;
+   HDC hdc;
+   DWORD color_bits;
+   struct stw_context *next;
+};
+
+struct stw_context *
+stw_context_from_hdc(HDC hdc );
+
+
+
+
+
+
+
+
+#endif /* STW_CONTEXT_H */
diff --git a/src/gallium/state_trackers/wgl/shared/stw_device.c b/src/gallium/state_trackers/wgl/shared/stw_device.c
new file mode 100644
index 0000000000..6873e813ee
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_device.c
@@ -0,0 +1,91 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <windows.h>
+
+#include "pipe/p_debug.h"
+#include "pipe/p_screen.h"
+
+#include "shared/stw_device.h"
+#include "shared/stw_winsys.h"
+#include "shared/stw_pixelformat.h"
+#include "shared/stw_public.h"
+#include "stw.h"
+
+
+struct stw_device *stw_dev = NULL;
+
+
+/**
+ * XXX: Dispatch pipe_screen::flush_front_buffer to our 
+ * stw_winsys::flush_front_buffer.
+ */
+static void 
+st_flush_frontbuffer(struct pipe_screen *screen,
+                     struct pipe_surface *surf,
+                     void *context_private )
+{
+   const struct stw_winsys *stw_winsys = stw_dev->stw_winsys;
+   HDC hdc = (HDC)context_private;
+   
+   stw_winsys->flush_frontbuffer(screen, surf, hdc);
+}
+
+
+boolean
+stw_shared_init(const struct stw_winsys *stw_winsys)
+{
+   static struct stw_device stw_dev_storage;
+
+   assert(!stw_dev);
+
+   stw_dev = &stw_dev_storage;
+   memset(stw_dev, 0, sizeof(*stw_dev));
+
+   stw_dev->stw_winsys = stw_winsys;
+
+   stw_dev->screen = stw_winsys->create_screen();
+   if(!stw_dev->screen)
+      goto error1;
+
+   stw_dev->screen->flush_frontbuffer = st_flush_frontbuffer;
+   
+   pixelformat_init();
+
+   return TRUE;
+
+error1:
+   stw_dev = NULL;
+   return FALSE;
+}
+
+
+void
+stw_shared_cleanup(void)
+{
+   stw_dev = NULL;
+}
diff --git a/src/gallium/state_trackers/wgl/shared/stw_device.h b/src/gallium/state_trackers/wgl/shared/stw_device.h
new file mode 100644
index 0000000000..bc0bce37c6
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_device.h
@@ -0,0 +1,44 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ST_DEVICE_H_
+#define ST_DEVICE_H_
+
+
+struct pipe_screen;
+
+struct stw_device
+{
+   const struct stw_winsys *stw_winsys;
+   struct pipe_screen *screen;
+};
+
+
+extern struct stw_device *stw_dev;
+
+
+#endif /* ST_DEVICE_H_ */
diff --git a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c
new file mode 100644
index 0000000000..c70b31a488
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.c
@@ -0,0 +1,212 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include <windows.h>
+
+#include "main/context.h"
+#include "pipe/p_format.h"
+#include "pipe/p_screen.h"
+#include "state_tracker/st_context.h"
+#include "state_tracker/st_public.h"
+#include "stw_framebuffer.h"
+#include "stw_device.h"
+#include "stw_public.h"
+#include "stw_winsys.h"
+
+
+void
+framebuffer_resize(
+   struct stw_framebuffer *fb,
+   GLuint width,
+   GLuint height )
+{
+   if (fb->hbmDIB == NULL || fb->stfb->Base.Width != width || fb->stfb->Base.Height != height) {
+      if (fb->hbmDIB)
+         DeleteObject( fb->hbmDIB );
+
+      fb->hbmDIB = CreateCompatibleBitmap(
+         fb->hDC,
+         width,
+         height );
+   }
+
+   st_resize_framebuffer( fb->stfb, width, height );
+}
+
+static struct stw_framebuffer *fb_head = NULL;
+
+static LRESULT CALLBACK
+window_proc(
+   HWND hWnd,
+   UINT uMsg,
+   WPARAM wParam,
+   LPARAM lParam )
+{
+   struct stw_framebuffer *fb;
+
+   for (fb = fb_head; fb != NULL; fb = fb->next)
+      if (fb->hWnd == hWnd)
+         break;
+   assert( fb != NULL );
+
+   if (uMsg == WM_SIZE && wParam != SIZE_MINIMIZED)
+      framebuffer_resize( fb, LOWORD( lParam ), HIWORD( lParam ) );
+
+   return CallWindowProc( fb->WndProc, hWnd, uMsg, wParam, lParam );
+}
+
+/* Create a new framebuffer object which will correspond to the given HDC.
+ */
+struct stw_framebuffer *
+framebuffer_create(
+   HDC hdc,
+   GLvisual *visual,
+   GLuint width,
+   GLuint height )
+{
+   struct stw_framebuffer *fb;
+   enum pipe_format colorFormat, depthFormat, stencilFormat;
+
+   fb = CALLOC_STRUCT( stw_framebuffer );
+   if (fb == NULL)
+      return NULL;
+
+   /* Determine PIPE_FORMATs for buffers.
+    */
+   colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+   if (visual->depthBits == 0)
+      depthFormat = PIPE_FORMAT_NONE;
+   else if (visual->depthBits <= 16)
+      depthFormat = PIPE_FORMAT_Z16_UNORM;
+   else if (visual->depthBits <= 24)
+      depthFormat = PIPE_FORMAT_S8Z24_UNORM;
+   else
+      depthFormat = PIPE_FORMAT_Z32_UNORM;
+
+   if (visual->stencilBits == 8) {
+      if (depthFormat == PIPE_FORMAT_S8Z24_UNORM)
+         stencilFormat = depthFormat;
+      else
+         stencilFormat = PIPE_FORMAT_S8_UNORM;
+   }
+   else {
+      stencilFormat = PIPE_FORMAT_NONE;
+   }
+
+   fb->stfb = st_create_framebuffer(
+      visual,
+      colorFormat,
+      depthFormat,
+      stencilFormat,
+      width,
+      height,
+      (void *) fb );
+
+   fb->cColorBits = GetDeviceCaps( hdc, BITSPIXEL );
+   fb->hDC = hdc;
+
+   /* Subclass a window associated with the device context.
+    */
+   fb->hWnd = WindowFromDC( hdc );
+   if (fb->hWnd != NULL) {
+      fb->WndProc = (WNDPROC) SetWindowLong(
+         fb->hWnd,
+         GWL_WNDPROC,
+         (LONG) window_proc );
+   }
+
+   fb->next = fb_head;
+   fb_head = fb;
+   return fb;
+}
+
+void
+framebuffer_destroy(
+   struct stw_framebuffer *fb )
+{
+   struct stw_framebuffer **link = &fb_head;
+   struct stw_framebuffer *pfb = fb_head;
+
+   while (pfb != NULL) {
+      if (pfb == fb) {
+         if (fb->hWnd != NULL) {
+            SetWindowLong(
+               fb->hWnd,
+               GWL_WNDPROC,
+               (LONG) fb->WndProc );
+         }
+
+         *link = fb->next;
+         FREE( fb );
+         return;
+      }
+
+      link = &pfb->next;
+      pfb = pfb->next;
+   }
+}
+
+/* Given an hdc, return the corresponding stw_framebuffer.
+ */
+struct stw_framebuffer *
+framebuffer_from_hdc(
+   HDC hdc )
+{
+   struct stw_framebuffer *fb;
+
+   for (fb = fb_head; fb != NULL; fb = fb->next)
+      if (fb->hDC == hdc)
+         return fb;
+   return NULL;
+}
+
+
+BOOL
+stw_swap_buffers(
+   HDC hdc )
+{
+   struct stw_framebuffer *fb;
+   struct pipe_surface *surf;
+
+   fb = framebuffer_from_hdc( hdc );
+   if (fb == NULL)
+      return FALSE;
+
+   /* If we're swapping the buffer associated with the current context
+    * we have to flush any pending rendering commands first.
+    */
+   st_notify_swapbuffers( fb->stfb );
+
+   st_get_framebuffer_surface( fb->stfb, ST_SURFACE_BACK_LEFT, &surf );
+
+   stw_dev->stw_winsys->flush_frontbuffer(stw_dev->screen,
+                                          surf,
+                                          hdc );
+
+   return TRUE;
+}
diff --git a/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h
new file mode 100644
index 0000000000..2e16e421f2
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_framebuffer.h
@@ -0,0 +1,71 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef STW_FRAMEBUFFER_H
+#define STW_FRAMEBUFFER_H
+
+#include "main/mtypes.h"
+
+/* Windows framebuffer, derived from gl_framebuffer.
+ */
+struct stw_framebuffer
+{
+   struct st_framebuffer *stfb;
+   HDC hDC;
+   int pixelformat;
+   BYTE cColorBits;
+   HDC dib_hDC;
+   HBITMAP hbmDIB;
+   HBITMAP hOldBitmap;
+   PBYTE pbPixels;
+   HWND hWnd;
+   WNDPROC WndProc;
+   struct stw_framebuffer *next;
+};
+
+struct stw_framebuffer *
+framebuffer_create(
+   HDC hdc,
+   GLvisual *visual,
+   GLuint width,
+   GLuint height );
+
+void
+framebuffer_destroy(
+   struct stw_framebuffer *fb );
+
+void
+framebuffer_resize(
+   struct stw_framebuffer *fb,
+   GLuint width,
+   GLuint height );
+
+struct stw_framebuffer *
+framebuffer_from_hdc(
+   HDC hdc );
+
+#endif /* STW_FRAMEBUFFER_H */
diff --git a/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c b/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c
new file mode 100644
index 0000000000..ac2d6fc260
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_getprocaddress.c
@@ -0,0 +1,71 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <windows.h>
+
+#include "glapi/glapi.h"
+#include "stw_arbextensionsstring.h"
+#include "stw_arbpixelformat.h"
+#include "stw_public.h"
+
+struct extension_entry
+{
+   const char *name;
+   PROC proc;
+};
+
+#define EXTENTRY(P) { #P, (PROC) P }
+
+static struct extension_entry extension_entries[] = {
+
+   /* WGL_ARB_extensions_string */
+   EXTENTRY( wglGetExtensionsStringARB ),
+
+   /* WGL_ARB_pixel_format */
+   EXTENTRY( wglChoosePixelFormatARB ),
+   EXTENTRY( wglGetPixelFormatAttribfvARB ),
+   EXTENTRY( wglGetPixelFormatAttribivARB ),
+
+   { NULL, NULL }
+};
+
+PROC
+stw_get_proc_address(
+   LPCSTR lpszProc )
+{
+   struct extension_entry *entry;
+
+   PROC p = (PROC) _glapi_get_proc_address( (const char *) lpszProc );
+   if (p)
+      return p;
+
+   for (entry = extension_entries; entry->name; entry++)
+      if (strcmp( lpszProc, entry->name ) == 0)
+         return entry->proc;
+
+   return NULL;
+}
diff --git a/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c
new file mode 100644
index 0000000000..84b7b287b9
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.c
@@ -0,0 +1,286 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "pipe/p_debug.h"
+#include "stw_pixelformat.h"
+#include "stw_public.h"
+
+#define MAX_PIXELFORMATS   16
+
+static struct pixelformat_info pixelformats[MAX_PIXELFORMATS];
+static uint pixelformat_count = 0;
+static uint pixelformat_extended_count = 0;
+
+static uint currentpixelformat = 0;
+
+
+static void
+add_standard_pixelformats(
+   struct pixelformat_info **ppf,
+   uint flags )
+{
+   struct pixelformat_info *pf = *ppf;
+   struct pixelformat_color_info color24 = { 8, 0, 8, 8, 8, 16 };
+   struct pixelformat_alpha_info alpha8 = { 8, 24 };
+   struct pixelformat_alpha_info noalpha = { 0, 0 };
+   struct pixelformat_depth_info depth24s8 = { 24, 8 };
+   struct pixelformat_depth_info depth16 = { 16, 0 };
+
+   pf->flags = PF_FLAG_DOUBLEBUFFER | flags;
+   pf->color = color24;
+   pf->alpha = alpha8;
+   pf->depth = depth16;
+   pf++;
+
+   pf->flags = PF_FLAG_DOUBLEBUFFER | flags;
+   pf->color = color24;
+   pf->alpha = alpha8;
+   pf->depth = depth24s8;
+   pf++;
+
+   pf->flags = PF_FLAG_DOUBLEBUFFER | flags;
+   pf->color = color24;
+   pf->alpha = noalpha;
+   pf->depth = depth16;
+   pf++;
+
+   pf->flags = PF_FLAG_DOUBLEBUFFER | flags;
+   pf->color = color24;
+   pf->alpha = noalpha;
+   pf->depth = depth24s8;
+   pf++;
+
+   pf->flags = flags;
+   pf->color = color24;
+   pf->alpha = noalpha;
+   pf->depth = depth16;
+   pf++;
+
+   pf->flags = flags;
+   pf->color = color24;
+   pf->alpha = noalpha;
+   pf->depth = depth24s8;
+   pf++;
+
+   *ppf = pf;
+}
+
+void
+pixelformat_init( void )
+{
+   struct pixelformat_info *pf = pixelformats;
+
+   add_standard_pixelformats( &pf, 0 );
+   pixelformat_count = pf - pixelformats;
+
+   add_standard_pixelformats( &pf, PF_FLAG_MULTISAMPLED );
+   pixelformat_extended_count = pf - pixelformats;
+
+   assert( pixelformat_extended_count <= MAX_PIXELFORMATS );
+}
+
+uint
+pixelformat_get_count( void )
+{
+   return pixelformat_count;
+}
+
+uint
+pixelformat_get_extended_count( void )
+{
+   return pixelformat_extended_count;
+}
+
+const struct pixelformat_info *
+pixelformat_get_info( uint index )
+{
+   assert( index < pixelformat_extended_count );
+
+   return &pixelformats[index];
+}
+
+
+int
+stw_pixelformat_describe(
+   HDC hdc,
+   int iPixelFormat,
+   UINT nBytes,
+   LPPIXELFORMATDESCRIPTOR ppfd )
+{
+   uint count;
+   uint index;
+   const struct pixelformat_info *pf;
+
+   (void) hdc;
+
+   count = pixelformat_get_extended_count();
+   index = (uint) iPixelFormat - 1;
+
+   if (ppfd == NULL)
+      return count;
+   if (index >= count || nBytes != sizeof( PIXELFORMATDESCRIPTOR ))
+      return 0;
+
+   pf = pixelformat_get_info( index );
+
+   ppfd->nSize = sizeof( PIXELFORMATDESCRIPTOR );
+   ppfd->nVersion = 1;
+   ppfd->dwFlags = PFD_DRAW_TO_WINDOW | PFD_SUPPORT_OPENGL;
+   if (pf->flags & PF_FLAG_DOUBLEBUFFER)
+      ppfd->dwFlags |= PFD_DOUBLEBUFFER | PFD_SWAP_COPY;
+   ppfd->iPixelType = PFD_TYPE_RGBA;
+   ppfd->cColorBits = pf->color.redbits + pf->color.greenbits + pf->color.bluebits;
+   ppfd->cRedBits = pf->color.redbits;
+   ppfd->cRedShift = pf->color.redshift;
+   ppfd->cGreenBits = pf->color.greenbits;
+   ppfd->cGreenShift = pf->color.greenshift;
+   ppfd->cBlueBits = pf->color.bluebits;
+   ppfd->cBlueShift = pf->color.blueshift;
+   ppfd->cAlphaBits = pf->alpha.alphabits;
+   ppfd->cAlphaShift = pf->alpha.alphashift;
+   ppfd->cAccumBits = 0;
+   ppfd->cAccumRedBits = 0;
+   ppfd->cAccumGreenBits = 0;
+   ppfd->cAccumBlueBits = 0;
+   ppfd->cAccumAlphaBits = 0;
+   ppfd->cDepthBits = pf->depth.depthbits;
+   ppfd->cStencilBits = pf->depth.stencilbits;
+   ppfd->cAuxBuffers = 0;
+   ppfd->iLayerType = 0;
+   ppfd->bReserved = 0;
+   ppfd->dwLayerMask = 0;
+   ppfd->dwVisibleMask = 0;
+   ppfd->dwDamageMask = 0;
+
+   return count;
+}
+
+/* Only used by the wgl code, but have it here to avoid exporting the
+ * pixelformat.h functionality.
+ */
+int stw_pixelformat_choose( HDC hdc,
+                            CONST PIXELFORMATDESCRIPTOR *ppfd )
+{
+   uint count;
+   uint index;
+   uint bestindex;
+   uint bestdelta;
+
+   (void) hdc;
+
+   count = pixelformat_get_count();
+   bestindex = count;
+   bestdelta = 0xffffffff;
+
+   for (index = 0; index < count; index++) {
+      uint delta = 0;
+      const struct pixelformat_info *pf = pixelformat_get_info( index );
+
+      if (!(ppfd->dwFlags & PFD_DOUBLEBUFFER_DONTCARE) &&
+          !!(ppfd->dwFlags & PFD_DOUBLEBUFFER) !=
+          !!(pf->flags & PF_FLAG_DOUBLEBUFFER))
+         continue;
+
+      if (ppfd->cColorBits != pf->color.redbits + pf->color.greenbits + pf->color.bluebits)
+         delta += 8;
+
+      if (ppfd->cDepthBits != pf->depth.depthbits)
+         delta += 4;
+
+      if (ppfd->cStencilBits != pf->depth.stencilbits)
+         delta += 2;
+
+      if (ppfd->cAlphaBits != pf->alpha.alphabits)
+         delta++;
+
+      if (delta < bestdelta) {
+         bestindex = index;
+         bestdelta = delta;
+         if (bestdelta == 0)
+            break;
+      }
+   }
+
+   if (bestindex == count)
+      return 0;
+
+   return bestindex + 1;
+}
+
+
+int
+stw_pixelformat_get(
+   HDC hdc )
+{
+   return currentpixelformat;
+}
+
+
+BOOL
+stw_pixelformat_set(
+   HDC hdc,
+   int iPixelFormat )
+{
+   uint count;
+   uint index;
+
+   (void) hdc;
+
+   index = (uint) iPixelFormat - 1;
+   count = pixelformat_get_extended_count();
+   if (index >= count)
+      return FALSE;
+
+   currentpixelformat = iPixelFormat;
+   
+   /* Some applications mistakenly use the undocumented wglSetPixelFormat 
+    * function instead of SetPixelFormat, so we call SetPixelFormat here to 
+    * avoid opengl32.dll's wglCreateContext to fail */
+   if (GetPixelFormat(hdc) == 0) {
+        SetPixelFormat(hdc, iPixelFormat, NULL);
+   }
+   
+   return TRUE;
+}
+
+
+
+/* XXX: this needs to be turned into queries on pipe_screen or
+ * stw_winsys.
+ */
+int
+stw_query_sample_buffers( void )
+{
+   return 1;
+}
+
+int
+stw_query_samples( void )
+{
+   return 4;
+}
+
diff --git a/src/gallium/state_trackers/wgl/shared/stw_pixelformat.h b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.h
new file mode 100644
index 0000000000..7ca4194a2a
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_pixelformat.h
@@ -0,0 +1,83 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef PIXELFORMAT_H
+#define PIXELFORMAT_H
+
+#include <windows.h>
+#include "pipe/p_compiler.h"
+
+#define PF_FLAG_DOUBLEBUFFER  0x00000001
+#define PF_FLAG_MULTISAMPLED  0x00000002
+
+struct pixelformat_color_info
+{
+   uint redbits;
+   uint redshift;
+   uint greenbits;
+   uint greenshift;
+   uint bluebits;
+   uint blueshift;
+};
+
+struct pixelformat_alpha_info
+{
+   uint alphabits;
+   uint alphashift;
+};
+
+struct pixelformat_depth_info
+{
+   uint depthbits;
+   uint stencilbits;
+};
+
+struct pixelformat_info
+{
+   uint flags;
+   struct pixelformat_color_info color;
+   struct pixelformat_alpha_info alpha;
+   struct pixelformat_depth_info depth;
+};
+
+void
+pixelformat_init( void );
+
+uint
+pixelformat_get_count( void );
+
+uint
+pixelformat_get_extended_count( void );
+
+const struct pixelformat_info *
+pixelformat_get_info( uint index );
+
+int stw_query_sample_buffers( void );
+int stw_query_samples( void );
+
+
+#endif /* PIXELFORMAT_H */
diff --git a/src/gallium/state_trackers/wgl/shared/stw_public.h b/src/gallium/state_trackers/wgl/shared/stw_public.h
new file mode 100644
index 0000000000..75b504a50f
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_public.h
@@ -0,0 +1,75 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef STW_PUBLIC_H
+#define STW_PUBLIC_H
+
+#include <windows.h>
+#include "pipe/p_compiler.h"
+
+struct stw_winsys;
+struct stw_context;
+
+boolean
+st_shared_init(const struct stw_winsys *stw_winsys);
+
+void
+st_shared_cleanup(void);
+
+
+BOOL stw_copy_context( struct stw_context *src,
+                       struct stw_context *dst,
+                       UINT mask );
+
+struct stw_context *stw_create_context( HDC hdc, int iLayerPlane );
+
+BOOL stw_delete_context( struct stw_context *ctx );
+
+struct stw_context *stw_get_current_context( void );
+
+HDC stw_get_current_dc( void );
+
+BOOL stw_make_current( HDC hdc, struct stw_context *ctx );
+
+BOOL stw_swap_buffers( HDC hdc );
+
+PROC stw_get_proc_address( LPCSTR lpszProc );
+
+int stw_pixelformat_describe( HDC hdc,
+                              int iPixelFormat,
+                              UINT nBytes,
+                              LPPIXELFORMATDESCRIPTOR ppfd );
+
+int stw_pixelformat_get( HDC hdc );
+
+BOOL stw_pixelformat_set( HDC hdc,
+                          int iPixelFormat );
+
+int stw_pixelformat_choose( HDC hdc,
+                            CONST PIXELFORMATDESCRIPTOR *ppfd );
+
+#endif
diff --git a/src/gallium/state_trackers/wgl/shared/stw_quirks.c b/src/gallium/state_trackers/wgl/shared/stw_quirks.c
new file mode 100644
index 0000000000..0961ce3bb0
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_quirks.c
@@ -0,0 +1,113 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * @file
+ * 
+ * This is hopefully a temporary hack to define some needed dispatch
+ * table entries.  Hopefully, I'll find a better solution.  The
+ * dispatch table generation scripts ought to be making these dummy
+ * stubs as well.
+ */
+
+void gl_dispatch_stub_543(void){}
+void gl_dispatch_stub_544(void){}
+void gl_dispatch_stub_545(void){}
+void gl_dispatch_stub_546(void){}
+void gl_dispatch_stub_547(void){}
+void gl_dispatch_stub_548(void){}
+void gl_dispatch_stub_549(void){}
+void gl_dispatch_stub_550(void){}
+void gl_dispatch_stub_551(void){}
+void gl_dispatch_stub_552(void){}
+void gl_dispatch_stub_553(void){}
+void gl_dispatch_stub_554(void){}
+void gl_dispatch_stub_555(void){}
+void gl_dispatch_stub_556(void){}
+void gl_dispatch_stub_557(void){}
+void gl_dispatch_stub_558(void){}
+void gl_dispatch_stub_559(void){}
+void gl_dispatch_stub_560(void){}
+void gl_dispatch_stub_561(void){}
+void gl_dispatch_stub_565(void){}
+void gl_dispatch_stub_566(void){}
+void gl_dispatch_stub_570(void){}
+void gl_dispatch_stub_577(void){}
+void gl_dispatch_stub_578(void){}
+void gl_dispatch_stub_582(void){}
+void gl_dispatch_stub_603(void){}
+void gl_dispatch_stub_607(void){}
+void gl_dispatch_stub_645(void){}
+void gl_dispatch_stub_646(void){}
+void gl_dispatch_stub_647(void){}
+void gl_dispatch_stub_648(void){}
+void gl_dispatch_stub_649(void){}
+void gl_dispatch_stub_650(void){}
+void gl_dispatch_stub_651(void){}
+void gl_dispatch_stub_652(void){}
+void gl_dispatch_stub_653(void){}
+void gl_dispatch_stub_657(void){}
+void gl_dispatch_stub_733(void){}
+void gl_dispatch_stub_734(void){}
+void gl_dispatch_stub_735(void){}
+void gl_dispatch_stub_736(void){}
+void gl_dispatch_stub_737(void){}
+void gl_dispatch_stub_738(void){}
+void gl_dispatch_stub_744(void){}
+void gl_dispatch_stub_745(void){}
+void gl_dispatch_stub_746(void){}
+void gl_dispatch_stub_760(void){}
+void gl_dispatch_stub_761(void){}
+void gl_dispatch_stub_763(void){}
+void gl_dispatch_stub_764(void){}
+void gl_dispatch_stub_765(void){}
+void gl_dispatch_stub_766(void){}
+void gl_dispatch_stub_767(void){}
+void gl_dispatch_stub_768(void){}
+
+void gl_dispatch_stub_562(void){}
+void gl_dispatch_stub_563(void){}
+void gl_dispatch_stub_564(void){}
+void gl_dispatch_stub_567(void){}
+void gl_dispatch_stub_568(void){}
+void gl_dispatch_stub_569(void){}
+void gl_dispatch_stub_580(void){}
+void gl_dispatch_stub_581(void){}
+void gl_dispatch_stub_606(void){}
+void gl_dispatch_stub_654(void){}
+void gl_dispatch_stub_655(void){}
+void gl_dispatch_stub_656(void){}
+void gl_dispatch_stub_739(void){}
+void gl_dispatch_stub_740(void){}
+void gl_dispatch_stub_741(void){}
+void gl_dispatch_stub_748(void){}
+void gl_dispatch_stub_749(void){}
+void gl_dispatch_stub_769(void){}
+void gl_dispatch_stub_770(void){}
+void gl_dispatch_stub_771(void){}
+void gl_dispatch_stub_772(void){}
+void gl_dispatch_stub_773(void){}
diff --git a/src/gallium/state_trackers/wgl/shared/stw_winsys.h b/src/gallium/state_trackers/wgl/shared/stw_winsys.h
new file mode 100644
index 0000000000..a85a9a2257
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/shared/stw_winsys.h
@@ -0,0 +1,59 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef STW_WINSYS_H
+#define STW_WINSYS_H
+
+#include <windows.h> /* for HDC */
+
+#include "pipe/p_compiler.h"
+
+struct pipe_screen;
+struct pipe_context;
+struct pipe_surface;
+
+struct stw_winsys
+{
+   struct pipe_screen *
+   (*create_screen)( void );
+
+   struct pipe_context *
+   (*create_context)( struct pipe_screen *screen );
+
+   void
+   (*flush_frontbuffer)( struct pipe_screen *screen,
+                         struct pipe_surface *surf,
+                         HDC hDC );
+};
+
+boolean
+st_init(const struct stw_winsys *stw_winsys);
+
+void
+st_cleanup(void);
+
+#endif /* STW_WINSYS_H */
diff --git a/src/gallium/state_trackers/wgl/stw.c b/src/gallium/state_trackers/wgl/stw.c
new file mode 100644
index 0000000000..8bccdad221
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/stw.c
@@ -0,0 +1,57 @@
+/**************************************************************************
+ *
+ * Copyright 2009, VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "stw.h"
+#include "shared/stw_winsys.h"
+
+boolean
+st_init(const struct stw_winsys *stw_winsys)
+{
+   if (!stw_shared_init( stw_winsys ))
+      goto fail;
+
+   if (!stw_icd_init())
+      goto fail;
+
+   if (!stw_wgl_init())
+      goto fail;
+
+   return TRUE;
+
+fail:
+   st_cleanup();
+   return FALSE;
+}
+
+
+void
+st_cleanup(void)
+{
+   stw_icd_cleanup();
+   stw_shared_cleanup();
+   stw_wgl_cleanup();
+}
diff --git a/src/gallium/state_trackers/wgl/stw.h b/src/gallium/state_trackers/wgl/stw.h
new file mode 100644
index 0000000000..450af4ccb6
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/stw.h
@@ -0,0 +1,53 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef STW_H
+#define STW_H
+
+#include "pipe/p_compiler.h"
+
+struct stw_winsys;
+
+/* Public interface:
+ */
+boolean stw_init( const struct stw_winsys *stw_winsys );
+void stw_cleanup( void );
+
+
+
+/* Internal functions
+ */
+boolean stw_shared_init( const struct stw_winsys *stw_winsys );
+boolean stw_icd_init( void ); 
+boolean stw_wgl_init( void );
+
+void stw_shared_cleanup( void );
+void stw_icd_cleanup( void ); 
+void stw_wgl_cleanup( void );
+
+
+#endif /* STW_H */
diff --git a/src/gallium/state_trackers/wgl/wgl/stw_wgl.c b/src/gallium/state_trackers/wgl/wgl/stw_wgl.c
new file mode 100644
index 0000000000..f50b79b4e1
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/wgl/stw_wgl.c
@@ -0,0 +1,353 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <windows.h>
+
+#include "pipe/p_debug.h"
+#include "shared/stw_public.h"
+#include "stw_wgl.h"
+#include "stw.h"
+
+boolean stw_wgl_init( void )
+{
+   debug_printf("%s\n", __FUNCTION__);
+   return TRUE;
+}
+
+void stw_wgl_cleanup( void )
+{
+}
+
+static INLINE struct stw_context *stw_context( HGLRC hglrc )
+{
+   return (struct stw_context *)hglrc;
+}
+
+
+WINGDIAPI BOOL APIENTRY
+wglCopyContext(
+   HGLRC hglrcSrc,
+   HGLRC hglrcDst,
+   UINT mask )
+{
+   return stw_copy_context( stw_context(hglrcSrc), 
+                            stw_context(hglrcDst), 
+                            mask );
+}
+
+WINGDIAPI HGLRC APIENTRY
+wglCreateContext(
+   HDC hdc )
+{
+   return (HGLRC) stw_create_context( hdc, 0 );
+}
+
+WINGDIAPI HGLRC APIENTRY
+wglCreateLayerContext(
+   HDC hdc,
+   int iLayerPlane )
+{
+   return (HGLRC) stw_create_context( hdc, iLayerPlane );
+}
+
+WINGDIAPI BOOL APIENTRY
+wglDeleteContext(
+   HGLRC hglrc )
+{
+   return stw_delete_context( stw_context(hglrc) );
+}
+
+
+WINGDIAPI HGLRC APIENTRY
+wglGetCurrentContext( VOID )
+{
+   return (HGLRC) stw_get_current_context();
+}
+
+WINGDIAPI HDC APIENTRY
+wglGetCurrentDC( VOID )
+{
+   return stw_get_current_dc();
+}
+
+WINGDIAPI BOOL APIENTRY
+wglMakeCurrent(
+   HDC hdc,
+   HGLRC hglrc )
+{
+   return stw_make_current( hdc, stw_context(hglrc) );
+}
+
+
+WINGDIAPI BOOL APIENTRY
+wglSwapBuffers(
+   HDC hdc )
+{
+   return stw_swap_buffers( hdc );
+}
+
+
+WINGDIAPI BOOL APIENTRY
+wglSwapLayerBuffers(
+   HDC hdc,
+   UINT fuPlanes )
+{
+   (void) hdc;
+   (void) fuPlanes;
+
+   return FALSE;
+}
+
+WINGDIAPI PROC APIENTRY
+wglGetProcAddress(
+    LPCSTR lpszProc )
+{
+   return stw_get_proc_address( lpszProc );
+}
+
+
+WINGDIAPI int APIENTRY
+wglChoosePixelFormat(
+   HDC hdc,
+   CONST PIXELFORMATDESCRIPTOR *ppfd )
+{
+   if (ppfd->nSize != sizeof( PIXELFORMATDESCRIPTOR ) || ppfd->nVersion != 1)
+      return 0;
+   if (ppfd->iPixelType != PFD_TYPE_RGBA)
+      return 0;
+   if (!(ppfd->dwFlags & PFD_DRAW_TO_WINDOW))
+      return 0;
+   if (!(ppfd->dwFlags & PFD_SUPPORT_OPENGL))
+      return 0;
+   if (ppfd->dwFlags & PFD_DRAW_TO_BITMAP)
+      return 0;
+   if (ppfd->dwFlags & PFD_SUPPORT_GDI)
+      return 0;
+   if (!(ppfd->dwFlags & PFD_STEREO_DONTCARE) && (ppfd->dwFlags & PFD_STEREO))
+      return 0;
+
+   return stw_pixelformat_choose( hdc, ppfd );
+}
+
+WINGDIAPI int APIENTRY
+wglDescribePixelFormat(
+   HDC hdc,
+   int iPixelFormat,
+   UINT nBytes,
+   LPPIXELFORMATDESCRIPTOR ppfd )
+{
+   return stw_pixelformat_describe( hdc, iPixelFormat, nBytes, ppfd );
+}
+
+WINGDIAPI int APIENTRY
+wglGetPixelFormat(
+   HDC hdc )
+{
+   return stw_pixelformat_get( hdc );
+}
+
+WINGDIAPI BOOL APIENTRY
+wglSetPixelFormat(
+   HDC hdc,
+   int iPixelFormat,
+   const PIXELFORMATDESCRIPTOR *ppfd )
+{
+   if (ppfd->nSize != sizeof( PIXELFORMATDESCRIPTOR ))
+      return FALSE;
+
+   return stw_pixelformat_set( hdc, iPixelFormat );
+}
+
+
+WINGDIAPI BOOL APIENTRY
+wglUseFontBitmapsA(
+   HDC hdc,
+   DWORD first,
+   DWORD count,
+   DWORD listBase )
+{
+   (void) hdc;
+   (void) first;
+   (void) count;
+   (void) listBase;
+
+   assert( 0 );
+
+   return FALSE;
+}
+
+WINGDIAPI BOOL APIENTRY
+wglShareLists(
+   HGLRC hglrc1,
+   HGLRC hglrc2 )
+{
+   (void) hglrc1;
+   (void) hglrc2;
+
+   assert( 0 );
+
+   return FALSE;
+}
+
+WINGDIAPI BOOL APIENTRY
+wglUseFontBitmapsW(
+   HDC hdc,
+   DWORD first,
+   DWORD count,
+   DWORD listBase )
+{
+   (void) hdc;
+   (void) first;
+   (void) count;
+   (void) listBase;
+
+   assert( 0 );
+
+   return FALSE;
+}
+
+WINGDIAPI BOOL APIENTRY
+wglUseFontOutlinesA(
+   HDC hdc,
+   DWORD first,
+   DWORD count,
+   DWORD listBase,
+   FLOAT deviation,
+   FLOAT extrusion,
+   int format,
+   LPGLYPHMETRICSFLOAT lpgmf )
+{
+   (void) hdc;
+   (void) first;
+   (void) count;
+   (void) listBase;
+   (void) deviation;
+   (void) extrusion;
+   (void) format;
+   (void) lpgmf;
+
+   assert( 0 );
+
+   return FALSE;
+}
+
+WINGDIAPI BOOL APIENTRY
+wglUseFontOutlinesW(
+   HDC hdc,
+   DWORD first,
+   DWORD count,
+   DWORD listBase,
+   FLOAT deviation,
+   FLOAT extrusion,
+   int format,
+   LPGLYPHMETRICSFLOAT lpgmf )
+{
+   (void) hdc;
+   (void) first;
+   (void) count;
+   (void) listBase;
+   (void) deviation;
+   (void) extrusion;
+   (void) format;
+   (void) lpgmf;
+
+   assert( 0 );
+
+   return FALSE;
+}
+
+WINGDIAPI BOOL APIENTRY
+wglDescribeLayerPlane(
+   HDC hdc,
+   int iPixelFormat,
+   int iLayerPlane,
+   UINT nBytes,
+   LPLAYERPLANEDESCRIPTOR plpd )
+{
+   (void) hdc;
+   (void) iPixelFormat;
+   (void) iLayerPlane;
+   (void) nBytes;
+   (void) plpd;
+
+   assert( 0 );
+
+   return FALSE;
+}
+
+WINGDIAPI int APIENTRY
+wglSetLayerPaletteEntries(
+   HDC hdc,
+   int iLayerPlane,
+   int iStart,
+   int cEntries,
+   CONST COLORREF *pcr )
+{
+   (void) hdc;
+   (void) iLayerPlane;
+   (void) iStart;
+   (void) cEntries;
+   (void) pcr;
+
+   assert( 0 );
+
+   return 0;
+}
+
+WINGDIAPI int APIENTRY
+wglGetLayerPaletteEntries(
+   HDC hdc,
+   int iLayerPlane,
+   int iStart,
+   int cEntries,
+   COLORREF *pcr )
+{
+   (void) hdc;
+   (void) iLayerPlane;
+   (void) iStart;
+   (void) cEntries;
+   (void) pcr;
+
+   assert( 0 );
+
+   return 0;
+}
+
+WINGDIAPI BOOL APIENTRY
+wglRealizeLayerPalette(
+   HDC hdc,
+   int iLayerPlane,
+   BOOL bRealize )
+{
+   (void) hdc;
+   (void) iLayerPlane;
+   (void) bRealize;
+
+   assert( 0 );
+
+   return FALSE;
+}
diff --git a/src/gallium/state_trackers/wgl/wgl/stw_wgl.h b/src/gallium/state_trackers/wgl/wgl/stw_wgl.h
new file mode 100644
index 0000000000..b86cc240f2
--- /dev/null
+++ b/src/gallium/state_trackers/wgl/wgl/stw_wgl.h
@@ -0,0 +1,63 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef STW_WGL_H_
+#define STW_WGL_H_
+
+
+#include <windows.h>
+
+#include "GL/gl.h"
+
+
+/*
+ * Undeclared APIs exported by opengl32.dll
+ */
+
+WINGDIAPI BOOL WINAPI
+wglSwapBuffers(HDC hdc);
+
+WINGDIAPI int WINAPI
+wglChoosePixelFormat(HDC hdc,
+                     CONST PIXELFORMATDESCRIPTOR *ppfd);
+
+WINGDIAPI int WINAPI
+wglDescribePixelFormat(HDC hdc,
+                       int iPixelFormat,
+                       UINT nBytes,
+                       LPPIXELFORMATDESCRIPTOR ppfd);
+
+WINGDIAPI int WINAPI
+wglGetPixelFormat(HDC hdc);
+
+WINGDIAPI BOOL WINAPI
+wglSetPixelFormat(HDC hdc,
+                  int iPixelFormat,
+                  CONST PIXELFORMATDESCRIPTOR *ppfd);
+
+
+#endif /* STW_WGL_H_ */
diff --git a/src/gallium/winsys/Makefile b/src/gallium/winsys/Makefile
new file mode 100644
index 0000000000..2360a6a94a
--- /dev/null
+++ b/src/gallium/winsys/Makefile
@@ -0,0 +1,24 @@
+TOP = ../../..
+include $(TOP)/configs/current
+
+
+SUBDIRS = $(GALLIUM_WINSYS_DIRS)
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	rm -f `find . -name \*.[oa]`
+
+
+# Dummy install target
+install:
diff --git a/src/gallium/winsys/SConscript b/src/gallium/winsys/SConscript
new file mode 100644
index 0000000000..30c3378dff
--- /dev/null
+++ b/src/gallium/winsys/SConscript
@@ -0,0 +1,16 @@
+Import('*')
+
+if env['dri']:
+	SConscript([
+		'drm/SConscript',
+	])
+	
+if 'xlib' in env['winsys']:
+	SConscript([
+		'xlib/SConscript',
+	])
+
+if 'gdi' in env['winsys']:
+	SConscript([
+		'gdi/SConscript',
+	])
diff --git a/src/gallium/winsys/drm/Makefile b/src/gallium/winsys/drm/Makefile
new file mode 100644
index 0000000000..a84fcd8418
--- /dev/null
+++ b/src/gallium/winsys/drm/Makefile
@@ -0,0 +1,38 @@
+# src/gallium/winsys/drm/Makefile
+
+TOP = ../../../..
+
+include $(TOP)/configs/current
+
+SUBDIRS = $(GALLIUM_WINSYS_DRM_DIRS)
+
+default: $(TOP)/$(LIB_DIR)/gallium subdirs
+
+
+$(TOP)/$(LIB_DIR)/gallium:
+	-mkdir -p $(TOP)/$(LIB_DIR)/gallium
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+install:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE) install) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE) clean) ; \
+		fi \
+	done
+	-rm -f common/*.o
diff --git a/src/gallium/winsys/drm/Makefile.template b/src/gallium/winsys/drm/Makefile.template
new file mode 100644
index 0000000000..cccf8abdfd
--- /dev/null
+++ b/src/gallium/winsys/drm/Makefile.template
@@ -0,0 +1,125 @@
+# -*-makefile-*-
+
+MESA_MODULES = \
+	$(TOP)/src/mesa/libmesagallium.a \
+	$(GALLIUM_AUXILIARIES)
+
+COMMON_GALLIUM_SOURCES = \
+        $(TOP)/src/mesa/drivers/dri/common/utils.c \
+        $(TOP)/src/mesa/drivers/dri/common/vblank.c \
+        $(TOP)/src/mesa/drivers/dri/common/dri_util.c \
+        $(TOP)/src/mesa/drivers/dri/common/xmlconfig.c
+
+COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \
+        $(TOP)/src/mesa/drivers/common/driverfuncs.c \
+        $(TOP)/src/mesa/drivers/dri/common/texmem.c \
+        $(TOP)/src/mesa/drivers/dri/common/drirenderbuffer.c
+
+COMMON_BM_SOURCES = \
+	$(TOP)/src/mesa/drivers/dri/common/dri_bufmgr.c \
+	$(TOP)/src/mesa/drivers/dri/common/dri_drmpool.c
+
+
+ifeq ($(WINDOW_SYSTEM),dri)
+WINOBJ=
+WINLIB=
+INCLUDES = $(SHARED_INCLUDES) $(EXPAT_INCLUDES)
+
+OBJECTS = \
+	$(C_SOURCES:.c=.o) \
+	$(ASM_SOURCES:.S=.o) 
+
+else
+# miniglx
+WINOBJ=
+WINLIB=-L$(MESA)/src/glx/mini
+MINIGLX_INCLUDES = -I$(TOP)/src/glx/mini
+INCLUDES = $(MINIGLX_INCLUDES) \
+	   $(SHARED_INCLUDES) \
+	   $(PCIACCESS_CFLAGS)
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+	  $(MINIGLX_SOURCES:.c=.o) \
+	  $(ASM_SOURCES:.S=.o) 
+endif
+
+
+### Include directories
+SHARED_INCLUDES = \
+	-I. \
+	-I$(TOP)/src/mesa/drivers/dri/common \
+	-Iserver \
+	-I$(TOP)/include \
+	-I$(TOP)/include/GL/internal \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/gallium/winsys/common \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/mesa/main \
+	-I$(TOP)/src/mesa/glapi \
+	-I$(TOP)/src/mesa/math \
+	-I$(TOP)/src/mesa/transform \
+	-I$(TOP)/src/mesa/shader \
+	-I$(TOP)/src/mesa/swrast \
+	-I$(TOP)/src/mesa/swrast_setup \
+	-I$(TOP)/src/egl/main \
+	-I$(TOP)/src/egl/drivers/dri \
+	$(LIBDRM_CFLAGS)
+
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES)  $< -o $@
+
+
+##### TARGETS #####
+
+default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/gallium/$(LIBNAME) $(LIBNAME_EGL) $(TOP)/$(LIB_DIR)/gallium/$(LIBNAME_EGL)
+
+
+$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template
+	$(TOP)/bin/mklib -noprefix -o $@ \
+		$(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES)  $(WINOBJ) $(DRI_LIB_DEPS) $(DRIVER_EXTRAS)
+
+$(LIBNAME_EGL): $(WINSYS_OBJECTS) $(LIBS)
+	$(TOP)/bin/mklib -o $(LIBNAME_EGL) \
+		-linker "$(CC)" \
+		-noprefix \
+		$(OBJECTS) $(MKLIB_OPTIONS) $(WINSYS_OBJECTS) $(PIPE_DRIVERS) $(WINOBJ) $(DRI_LIB_DEPS) \
+		--whole-archive $(LIBS) $(GALLIUM_AUXILIARIES) --no-whole-archive $(DRIVER_EXTRAS)
+
+$(TOP)/$(LIB_DIR)/gallium/$(LIBNAME): $(LIBNAME)
+	$(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR)/gallium
+
+$(TOP)/$(LIB_DIR)/gallium/$(LIBNAME_EGL): $(LIBNAME_EGL)
+	$(INSTALL) $(LIBNAME_EGL) $(TOP)/$(LIB_DIR)/gallium
+
+depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \
+		$(ASM_SOURCES) 2> /dev/null
+
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` `find ../include`
+
+
+# Remove .o and backup files
+clean:
+	-rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS)
+	-rm -f depend depend.bak
+
+
+install: $(LIBNAME)
+	$(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR)
+	$(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR)
+
+
+include depend
diff --git a/src/gallium/winsys/drm/SConscript b/src/gallium/winsys/drm/SConscript
new file mode 100644
index 0000000000..aef5210a32
--- /dev/null
+++ b/src/gallium/winsys/drm/SConscript
@@ -0,0 +1,54 @@
+Import('*')
+
+if env['dri']:
+
+	drienv = env.Clone()
+
+	drienv.Replace(CPPPATH = [
+		'#src/mesa/drivers/dri/common',
+		'#include',
+		'#include/GL/internal',
+		'#src/gallium/include',
+		'#src/gallium/auxiliary',
+		'#src/gallium/drivers',
+		'#src/mesa',
+		'#src/mesa/main',
+		'#src/mesa/glapi',
+		'#src/mesa/math',
+		'#src/mesa/transform',
+		'#src/mesa/shader',
+		'#src/mesa/swrast',
+		'#src/mesa/swrast_setup',
+		'#src/egl/main',
+		'#src/egl/drivers/dri',
+	])
+
+	drienv.ParseConfig('pkg-config --cflags --libs libdrm')
+
+	COMMON_GALLIUM_SOURCES = [
+		'#src/mesa/drivers/dri/common/utils.c',
+		'#src/mesa/drivers/dri/common/vblank.c',
+		'#src/mesa/drivers/dri/common/dri_util.c',
+		'#src/mesa/drivers/dri/common/xmlconfig.c',
+	]
+
+	COMMON_BM_SOURCES = [
+		'#src/mesa/drivers/dri/common/dri_bufmgr.c',
+		'#src/mesa/drivers/dri/common/dri_drmpool.c',
+	]
+
+	Export([
+		'drienv',
+		'COMMON_GALLIUM_SOURCES',
+		'COMMON_BM_SOURCES',
+	])
+
+	# TODO: Installation
+	#install: $(LIBNAME)
+	#	$(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR)
+	#	$(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR)
+
+	if 'intel' in env['winsys']:
+		SConscript([
+			'intel/SConscript',
+		])
diff --git a/src/gallium/winsys/drm/intel/Makefile b/src/gallium/winsys/drm/intel/Makefile
new file mode 100644
index 0000000000..eede9fc866
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/Makefile
@@ -0,0 +1,25 @@
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+
+SUBDIRS = gem egl
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	rm -f `find . -name \*.[oa]`
+	rm -f `find . -name depend`
+
+
+# Dummy install target
+install:
diff --git a/src/gallium/winsys/drm/intel/common/Makefile b/src/gallium/winsys/drm/intel/common/Makefile
new file mode 100644
index 0000000000..bf1a7d691f
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/Makefile
@@ -0,0 +1,23 @@
+TOP = ../../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = inteldrm
+
+C_SOURCES = \
+	intel_be_batchbuffer.c \
+	intel_be_context.c \
+	intel_be_device.c \
+	ws_dri_bufmgr.c \
+	ws_dri_drmpool.c \
+	ws_dri_fencemgr.c \
+	ws_dri_mallocpool.c \
+	ws_dri_slabpool.c
+
+
+include ./Makefile.template
+
+DRIVER_DEFINES = $(shell pkg-config libdrm --cflags \
+                && pkg-config libdrm --atleast-version=2.3.1 \
+                && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP")
+symlinks:
+
diff --git a/src/gallium/winsys/drm/intel/common/Makefile.template b/src/gallium/winsys/drm/intel/common/Makefile.template
new file mode 100644
index 0000000000..02ed363a43
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/Makefile.template
@@ -0,0 +1,64 @@
+# -*-makefile-*-
+
+
+# We still have a dependency on the "dri" buffer manager.  Most likely
+# the interface can be reused in non-dri environments, and also as a
+# frontend to simpler memory managers.
+#
+COMMON_SOURCES = 
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+          $(CPP_SOURCES:.cpp=.o) \
+	  $(ASM_SOURCES:.S=.o) 
+
+
+### Include directories
+INCLUDES = \
+	-I. \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/include \
+        $(DRIVER_INCLUDES)
+
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.cpp.o:
+	$(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES)  $< -o $@
+
+
+##### TARGETS #####
+
+default: depend symlinks $(LIBNAME)
+
+
+$(LIBNAME): $(OBJECTS) Makefile Makefile.template
+	$(TOP)/bin/mklib -o $@ -static $(OBJECTS) $(DRIVER_LIBS)
+
+
+depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \
+		$(ASM_SOURCES) 2> /dev/null
+
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` `find ../include`
+
+
+# Remove .o and backup files
+clean::
+	-rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS)
+	-rm -f depend depend.bak
+
+
+include depend
diff --git a/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.c b/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.c
new file mode 100644
index 0000000000..bc13a5761e
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.c
@@ -0,0 +1,429 @@
+
+#include "intel_be_batchbuffer.h"
+#include "intel_be_context.h"
+#include "intel_be_device.h"
+#include <errno.h>
+
+#include "xf86drm.h"
+
+static void
+intel_realloc_relocs(struct intel_be_batchbuffer *batch, int num_relocs)
+{
+    unsigned long size = num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER;
+
+    size *= sizeof(uint32_t);
+    batch->reloc = realloc(batch->reloc, size);
+    batch->reloc_size = num_relocs;
+}
+
+
+void
+intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch)
+{
+   /*
+    * Get a new, free batchbuffer.
+    */
+    drmBO *bo;
+    struct drm_bo_info_req *req;
+
+   driBOUnrefUserList(batch->list);
+   driBOResetList(batch->list);
+
+   /* base.size is the size available to the i915simple driver */
+   batch->base.size = batch->device->max_batch_size - BATCH_RESERVED;
+   batch->base.actual_size = batch->device->max_batch_size;
+   driBOData(batch->buffer, batch->base.actual_size, NULL, NULL, 0);
+
+   /*
+    * Add the batchbuffer to the validate list.
+    */
+
+   driBOAddListItem(batch->list, batch->buffer,
+		    DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT,
+		    DRM_BO_FLAG_EXE | DRM_BO_MASK_MEM,
+		    &batch->dest_location, &batch->node);
+
+   req = &batch->node->bo_arg.d.req.bo_req;
+
+   /*
+    * Set up information needed for us to make relocations
+    * relative to the underlying drm buffer objects.
+    */
+
+   driReadLockKernelBO();
+   bo = driBOKernel(batch->buffer);
+   req->presumed_offset = (uint64_t) bo->offset;
+   req->hint = DRM_BO_HINT_PRESUMED_OFFSET;
+   batch->drmBOVirtual = (uint8_t *) bo->virtual;
+   driReadUnlockKernelBO();
+
+   /*
+    * Adjust the relocation buffer size.
+    */
+
+   if (batch->reloc_size > INTEL_MAX_RELOCS ||
+       batch->reloc == NULL)
+     intel_realloc_relocs(batch, INTEL_DEFAULT_RELOCS);
+
+   assert(batch->reloc != NULL);
+   batch->reloc[0] = 0; /* No relocs yet. */
+   batch->reloc[1] = 1; /* Reloc type 1 */
+   batch->reloc[2] = 0; /* Only a single relocation list. */
+   batch->reloc[3] = 0; /* Only a single relocation list. */
+
+   batch->base.map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0);
+   batch->poolOffset = driBOPoolOffset(batch->buffer);
+   batch->base.ptr = batch->base.map;
+   batch->dirty_state = ~0;
+   batch->nr_relocs = 0;
+   batch->flags = 0;
+   batch->id = 0;//batch->intel->intelScreen->batch_id++;
+}
+
+/*======================================================================
+ * Public functions
+ */
+struct intel_be_batchbuffer *
+intel_be_batchbuffer_alloc(struct intel_be_context *intel)
+{
+   struct intel_be_batchbuffer *batch = calloc(sizeof(*batch), 1);
+
+   batch->intel = intel;
+   batch->device = intel->device;
+
+   driGenBuffers(intel->device->batchPool, "batchbuffer", 1,
+                 &batch->buffer, 4096,
+                 DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0);
+   batch->last_fence = NULL;
+   batch->list = driBOCreateList(20);
+   batch->reloc = NULL;
+   intel_be_batchbuffer_reset(batch);
+   return batch;
+}
+
+void
+intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch)
+{
+   if (batch->last_fence) {
+      driFenceFinish(batch->last_fence,
+		     DRM_FENCE_TYPE_EXE, FALSE);
+      driFenceUnReference(&batch->last_fence);
+   }
+   if (batch->base.map) {
+      driBOUnmap(batch->buffer);
+      batch->base.map = NULL;
+   }
+   driBOUnReference(batch->buffer);
+   driBOFreeList(batch->list);
+   if (batch->reloc)
+       free(batch->reloc);
+   batch->buffer = NULL;
+   free(batch);
+}
+
+void
+intel_be_offset_relocation(struct intel_be_batchbuffer *batch,
+			unsigned pre_add,
+			struct _DriBufferObject *driBO,
+			uint64_t val_flags,
+			uint64_t val_mask)
+{
+    int itemLoc;
+    struct _drmBONode *node;
+    uint32_t *reloc;
+    struct drm_bo_info_req *req;
+
+    driBOAddListItem(batch->list, driBO, val_flags, val_mask,
+		     &itemLoc, &node);
+    req = &node->bo_arg.d.req.bo_req;
+
+    if (!(req->hint &  DRM_BO_HINT_PRESUMED_OFFSET)) {
+
+	/*
+	 * Stop other threads from tampering with the underlying
+	 * drmBO while we're reading its offset.
+	 */
+
+	driReadLockKernelBO();
+	req->presumed_offset = (uint64_t) driBOKernel(driBO)->offset;
+	driReadUnlockKernelBO();
+	req->hint = DRM_BO_HINT_PRESUMED_OFFSET;
+    }
+
+    pre_add += driBOPoolOffset(driBO);
+
+    if (batch->nr_relocs == batch->reloc_size)
+	intel_realloc_relocs(batch, batch->reloc_size * 2);
+
+    reloc = batch->reloc +
+	(I915_RELOC_HEADER + batch->nr_relocs * I915_RELOC0_STRIDE);
+
+    reloc[0] = ((uint8_t *)batch->base.ptr - batch->drmBOVirtual);
+    i915_batchbuffer_dword(&batch->base, req->presumed_offset + pre_add);
+    reloc[1] = pre_add;
+    reloc[2] = itemLoc;
+    reloc[3] = batch->dest_location;
+    batch->nr_relocs++;
+}
+
+static void
+i915_drm_copy_reply(const struct drm_bo_info_rep * rep, drmBO * buf)
+{
+    buf->handle = rep->handle;
+    buf->flags = rep->flags;
+    buf->size = rep->size;
+    buf->offset = rep->offset;
+    buf->mapHandle = rep->arg_handle;
+    buf->proposedFlags = rep->proposed_flags;
+    buf->start = rep->buffer_start;
+    buf->fenceFlags = rep->fence_flags;
+    buf->replyFlags = rep->rep_flags;
+    buf->pageAlignment = rep->page_alignment;
+}
+
+static int
+i915_execbuf(struct intel_be_batchbuffer *batch,
+	     unsigned int used,
+	     boolean ignore_cliprects,
+	     drmBOList *list,
+	     struct drm_i915_execbuffer *ea)
+{
+// struct intel_be_context *intel = batch->intel;
+   drmBONode *node;
+   drmMMListHead *l;
+   struct drm_i915_op_arg *arg, *first;
+   struct drm_bo_op_req *req;
+   struct drm_bo_info_rep *rep;
+   uint64_t *prevNext = NULL;
+   drmBO *buf;
+   int ret = 0;
+   uint32_t count = 0;
+
+   first = NULL;
+   for (l = list->list.next; l != &list->list; l = l->next) {
+      node = DRMLISTENTRY(drmBONode, l, head);
+
+      arg = &node->bo_arg;
+      req = &arg->d.req;
+
+      if (!first)
+	 first = arg;
+
+      if (prevNext)
+	 *prevNext = (unsigned long)arg;
+
+      prevNext = &arg->next;
+      req->bo_req.handle = node->buf->handle;
+      req->op = drm_bo_validate;
+      req->bo_req.flags = node->arg0;
+      req->bo_req.mask = node->arg1;
+      req->bo_req.hint |= 0;
+      count++;
+   }
+
+   memset(ea, 0, sizeof(*ea));
+   ea->num_buffers = count;
+   ea->batch.start = batch->poolOffset;
+   ea->batch.used = used;
+#if 0 /* ZZZ JB: no cliprects used */
+   ea->batch.cliprects = intel->pClipRects;
+   ea->batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects;
+   ea->batch.DR1 = 0;
+   ea->batch.DR4 = 0;((((GLuint) intel->drawX) & 0xffff) |
+		   (((GLuint) intel->drawY) << 16));
+#else
+   ea->batch.cliprects = NULL;
+   ea->batch.num_cliprects = 0;
+   ea->batch.DR1 = 0;
+   ea->batch.DR4 = 0;
+#endif
+   ea->fence_arg.flags = DRM_I915_FENCE_FLAG_FLUSHED;
+   ea->ops_list = (unsigned long) first;
+   first->reloc_ptr = (unsigned long) batch->reloc;
+   batch->reloc[0] = batch->nr_relocs;
+
+   //return -EFAULT;
+   do {
+      ret = drmCommandWriteRead(batch->device->fd, DRM_I915_EXECBUFFER, ea,
+				sizeof(*ea));
+   } while (ret == -EAGAIN);
+
+   if (ret != 0)
+      return ret;
+
+   for (l = list->list.next; l != &list->list; l = l->next) {
+      node = DRMLISTENTRY(drmBONode, l, head);
+      arg = &node->bo_arg;
+      rep = &arg->d.rep.bo_info;
+
+      if (!arg->handled) {
+	 return -EFAULT;
+      }
+      if (arg->d.rep.ret)
+	 return arg->d.rep.ret;
+
+      buf = node->buf;
+      i915_drm_copy_reply(rep, buf);
+   }
+   return 0;
+}
+
+/* TODO: Push this whole function into bufmgr.
+ */
+static struct _DriFenceObject *
+do_flush_locked(struct intel_be_batchbuffer *batch,
+                unsigned int used,
+                boolean ignore_cliprects, boolean allow_unlock)
+{
+   struct intel_be_context *intel = batch->intel;
+   struct _DriFenceObject *fo;
+   drmFence fence;
+   drmBOList *boList;
+   struct drm_i915_execbuffer ea;
+   int ret = 0;
+
+   driBOValidateUserList(batch->list);
+   boList = driGetdrmBOList(batch->list);
+
+#if 0 /* ZZZ JB Allways run */
+   if (!(intel->numClipRects == 0 && !ignore_cliprects)) {
+#else
+   if (1) {
+#endif
+      ret = i915_execbuf(batch, used, ignore_cliprects, boList, &ea);
+   } else {
+     driPutdrmBOList(batch->list);
+     fo = NULL;
+     goto out;
+   }
+   driPutdrmBOList(batch->list);
+   if (ret)
+      abort();
+
+   if (ea.fence_arg.error != 0) {
+
+     /*
+      * The hardware has been idled by the kernel.
+      * Don't fence the driBOs.
+      */
+
+       if (batch->last_fence)
+	   driFenceUnReference(&batch->last_fence);
+#if 0 /* ZZZ JB: no _mesa_* funcs in gallium */
+       _mesa_printf("fence error\n");
+#endif
+       batch->last_fence = NULL;
+       fo = NULL;
+       goto out;
+   }
+
+   fence.handle = ea.fence_arg.handle;
+   fence.fence_class = ea.fence_arg.fence_class;
+   fence.type = ea.fence_arg.type;
+   fence.flags = ea.fence_arg.flags;
+   fence.signaled = ea.fence_arg.signaled;
+
+   fo = driBOFenceUserList(batch->device->fenceMgr, batch->list,
+			   "SuperFence", &fence);
+
+   if (driFenceType(fo) & DRM_I915_FENCE_TYPE_RW) {
+       if (batch->last_fence)
+	   driFenceUnReference(&batch->last_fence);
+   /*
+	* FIXME: Context last fence??
+	*/
+       batch->last_fence = fo;
+       driFenceReference(fo);
+   }
+ out:
+#if 0 /* ZZZ JB: fix this */
+   intel->vtbl.lost_hardware(intel);
+#else
+   (void)intel;
+#endif
+   return fo;
+}
+
+
+struct _DriFenceObject *
+intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch)
+{
+   struct intel_be_context *intel = batch->intel;
+   unsigned int used = batch->base.ptr - batch->base.map;
+   boolean was_locked = batch->intel->hardware_locked(intel);
+   struct _DriFenceObject *fence;
+
+   if (used == 0) {
+      driFenceReference(batch->last_fence);
+      return batch->last_fence;
+   }
+
+   /* Add the MI_BATCH_BUFFER_END.  Always add an MI_FLUSH - this is a
+    * performance drain that we would like to avoid.
+    */
+#if 0 /* ZZZ JB: what should we do here? */
+   if (used & 4) {
+      ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd();
+      ((int *) batch->base.ptr)[1] = 0;
+      ((int *) batch->base.ptr)[2] = MI_BATCH_BUFFER_END;
+      used += 12;
+   }
+   else {
+      ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd();
+      ((int *) batch->base.ptr)[1] = MI_BATCH_BUFFER_END;
+      used += 8;
+   }
+#else
+   if (used & 4) {
+      ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH;
+      ((int *) batch->base.ptr)[1] = 0;
+      ((int *) batch->base.ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END;
+      used += 12;
+   }
+   else {
+      ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH;
+      ((int *) batch->base.ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END;
+      used += 8;
+   }
+#endif
+   driBOUnmap(batch->buffer);
+   batch->base.ptr = NULL;
+   batch->base.map = NULL;
+
+   /* TODO: Just pass the relocation list and dma buffer up to the
+    * kernel.
+    */
+   if (!was_locked)
+      intel->hardware_lock(intel);
+
+   fence = do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS),
+			   FALSE);
+
+   if (!was_locked)
+      intel->hardware_unlock(intel);
+
+   /* Reset the buffer:
+    */
+   intel_be_batchbuffer_reset(batch);
+   return fence;
+}
+
+void
+intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch)
+{
+   struct _DriFenceObject *fence = intel_be_batchbuffer_flush(batch);
+   driFenceFinish(fence, driFenceType(fence), FALSE);
+   driFenceUnReference(&fence);
+}
+
+#if 0
+void
+intel_be_batchbuffer_data(struct intel_be_batchbuffer *batch,
+                       const void *data, unsigned int bytes, unsigned int flags)
+{
+   assert((bytes & 3) == 0);
+   intel_batchbuffer_require_space(batch, bytes, flags);
+   memcpy(batch->base.ptr, data, bytes);
+   batch->base.ptr += bytes;
+}
+#endif
diff --git a/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.h b/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.h
new file mode 100644
index 0000000000..f150e3a674
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.h
@@ -0,0 +1,69 @@
+
+#ifndef INTEL_BE_BATCHBUFFER_H
+#define INTEL_BE_BATCHBUFFER_H
+
+#include "i915simple/i915_batch.h"
+
+#include "ws_dri_bufmgr.h"
+
+#define BATCH_RESERVED 16
+
+#define INTEL_DEFAULT_RELOCS 100
+#define INTEL_MAX_RELOCS 400
+
+#define INTEL_BATCH_NO_CLIPRECTS 0x1
+#define INTEL_BATCH_CLIPRECTS    0x2
+
+struct intel_be_context;
+struct intel_be_device;
+
+struct intel_be_batchbuffer
+{
+	struct i915_batchbuffer base;
+
+	struct intel_be_context *intel;
+	struct intel_be_device *device;
+
+	struct _DriBufferObject *buffer;
+	struct _DriFenceObject *last_fence;
+	uint32_t flags;
+
+	struct _DriBufferList *list;
+	size_t list_count;
+
+	uint32_t *reloc;
+	size_t reloc_size;
+	size_t nr_relocs;
+
+	uint32_t dirty_state;
+	uint32_t id;
+
+	uint32_t poolOffset;
+	uint8_t *drmBOVirtual;
+	struct _drmBONode *node; /* Validation list node for this buffer */
+	int dest_location;     /* Validation list sequence for this buffer */
+};
+
+struct intel_be_batchbuffer *
+intel_be_batchbuffer_alloc(struct intel_be_context *intel);
+
+void
+intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch);
+
+void
+intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch);
+
+struct _DriFenceObject *
+intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch);
+
+void
+intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch);
+
+void
+intel_be_offset_relocation(struct intel_be_batchbuffer *batch,
+			unsigned pre_add,
+			struct _DriBufferObject *driBO,
+			uint64_t val_flags,
+			uint64_t val_mask);
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/common/intel_be_context.c b/src/gallium/winsys/drm/intel/common/intel_be_context.c
new file mode 100644
index 0000000000..1af39674f4
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/intel_be_context.c
@@ -0,0 +1,107 @@
+
+/*
+ * Authors: Jakob Bornecrantz <jakob-at-tungstengraphics.com>
+ */
+
+#include "ws_dri_fencemgr.h"
+#include "intel_be_device.h"
+#include "intel_be_context.h"
+#include "intel_be_batchbuffer.h"
+
+static INLINE struct intel_be_context *
+intel_be_context(struct i915_winsys *sws)
+{
+	return (struct intel_be_context *)sws;
+}
+
+/* Simple batchbuffer interface:
+ */
+
+static struct i915_batchbuffer*
+intel_i915_batch_get(struct i915_winsys *sws)
+{
+	struct intel_be_context *intel = intel_be_context(sws);
+	return &intel->batch->base;
+}
+
+static void intel_i915_batch_reloc(struct i915_winsys *sws,
+				   struct pipe_buffer *buf,
+				   unsigned access_flags,
+				   unsigned delta)
+{
+	struct intel_be_context *intel = intel_be_context(sws);
+
+	unsigned flags = DRM_BO_FLAG_MEM_TT;
+	unsigned mask = DRM_BO_MASK_MEM;
+
+	if (access_flags & I915_BUFFER_ACCESS_WRITE) {
+		flags |= DRM_BO_FLAG_WRITE;
+		mask |= DRM_BO_FLAG_WRITE;
+	}
+
+	if (access_flags & I915_BUFFER_ACCESS_READ) {
+		flags |= DRM_BO_FLAG_READ;
+		mask |= DRM_BO_FLAG_READ;
+	}
+
+	intel_be_offset_relocation(intel->batch,
+				delta,
+				dri_bo(buf),
+				flags,
+				mask);
+}
+
+static void intel_i915_batch_flush(struct i915_winsys *sws,
+				   struct pipe_fence_handle **fence)
+{
+	struct intel_be_context *intel = intel_be_context(sws);
+
+	union {
+		struct _DriFenceObject *dri;
+		struct pipe_fence_handle *pipe;
+	} fu;
+
+	if (fence)
+		assert(!*fence);
+
+	fu.dri = intel_be_batchbuffer_flush(intel->batch);
+
+	if (!fu.dri) {
+		assert(0);
+		*fence = NULL;
+		return;
+	}
+
+	if (fu.dri) {
+		if (fence)
+			*fence = fu.pipe;
+		else
+			driFenceUnReference(&fu.dri);
+	}
+
+}
+
+boolean
+intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *device)
+{
+	assert(intel);
+	assert(device);
+
+	intel->device = device;
+
+	/* TODO move framebuffer createion to the driver */
+
+	intel->base.batch_get = intel_i915_batch_get;
+	intel->base.batch_reloc = intel_i915_batch_reloc;
+	intel->base.batch_flush = intel_i915_batch_flush;
+
+	intel->batch = intel_be_batchbuffer_alloc(intel);
+
+	return true;
+}
+
+void
+intel_be_destroy_context(struct intel_be_context *intel)
+{
+	intel_be_batchbuffer_free(intel->batch);
+}
diff --git a/src/gallium/winsys/drm/intel/common/intel_be_context.h b/src/gallium/winsys/drm/intel/common/intel_be_context.h
new file mode 100644
index 0000000000..d5cbc93594
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/intel_be_context.h
@@ -0,0 +1,40 @@
+/* These need to be diffrent from the intel winsys */
+#ifndef INTEL_BE_CONTEXT_H
+#define INTEL_BE_CONTEXT_H
+
+#include "i915simple/i915_winsys.h"
+
+struct intel_be_context
+{
+	/** Interface to i915simple driver */
+	struct i915_winsys base;
+
+	struct intel_be_device *device;
+	struct intel_be_batchbuffer *batch;
+
+	/*
+     * Hardware lock functions.
+     *
+     * Needs to be filled in by the winsys.
+     */
+	void (*hardware_lock)(struct intel_be_context *context);
+	void (*hardware_unlock)(struct intel_be_context *context);
+	boolean (*hardware_locked)(struct intel_be_context *context);
+};
+
+/**
+ * Intialize a allocated intel_be_context struct.
+ *
+ * Remember to set the hardware_* functions.
+ */
+boolean
+intel_be_init_context(struct intel_be_context *intel,
+		      struct intel_be_device *device);
+
+/**
+ * Destroy a intel_be_context.
+ * Does not free the struct that is up to the winsys.
+ */
+void
+intel_be_destroy_context(struct intel_be_context *intel);
+#endif
diff --git a/src/gallium/winsys/drm/intel/common/intel_be_device.c b/src/gallium/winsys/drm/intel/common/intel_be_device.c
new file mode 100644
index 0000000000..85ab1a2684
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/intel_be_device.c
@@ -0,0 +1,296 @@
+
+
+/*
+ * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com>
+ *			 Jakob Bornecrantz <jakob-at-tungstengraphics-dot-com>
+ */
+
+#include "intel_be_device.h"
+#include "ws_dri_bufmgr.h"
+#include "ws_dri_bufpool.h"
+#include "ws_dri_fencemgr.h"
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+
+#include "i915simple/i915_screen.h"
+
+/* Turn a pipe winsys into an intel/pipe winsys:
+ */
+static INLINE struct intel_be_device *
+intel_be_device( struct pipe_winsys *winsys )
+{
+	return (struct intel_be_device *)winsys;
+}
+
+
+/*
+ * Buffer functions.
+ *
+ * Most callbacks map direcly onto dri_bufmgr operations:
+ */
+
+static void *intel_be_buffer_map(struct pipe_winsys *winsys,
+					struct pipe_buffer *buf,
+					unsigned flags )
+{
+	unsigned drm_flags = 0;
+	
+	if (flags & PIPE_BUFFER_USAGE_CPU_WRITE)
+		drm_flags |= DRM_BO_FLAG_WRITE;
+
+	if (flags & PIPE_BUFFER_USAGE_CPU_READ)
+		drm_flags |= DRM_BO_FLAG_READ;
+
+	return driBOMap( dri_bo(buf), drm_flags, 0 );
+}
+
+static void intel_be_buffer_unmap(struct pipe_winsys *winsys,
+					 struct pipe_buffer *buf)
+{
+	driBOUnmap( dri_bo(buf) );
+}
+
+static void
+intel_be_buffer_destroy(struct pipe_winsys *winsys,
+			  struct pipe_buffer *buf)
+{
+	driBOUnReference( dri_bo(buf) );
+	FREE(buf);
+}
+
+static struct pipe_buffer *
+intel_be_buffer_create(struct pipe_winsys *winsys,
+						  unsigned alignment,
+						  unsigned usage,
+						  unsigned size )
+{
+	struct intel_be_buffer *buffer = CALLOC_STRUCT( intel_be_buffer );
+	struct intel_be_device *iws = intel_be_device(winsys);
+	unsigned flags = 0;
+	struct _DriBufferPool *pool;
+
+	buffer->base.refcount = 1;
+	buffer->base.alignment = alignment;
+	buffer->base.usage = usage;
+	buffer->base.size = size;
+
+	if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) {
+		flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED;
+		pool = iws->mallocPool;
+	} else if (usage & PIPE_BUFFER_USAGE_CUSTOM) {
+		/* For vertex buffers */
+		flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT;
+		pool = iws->vertexPool;
+	} else {
+		flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT;
+		pool = iws->regionPool;
+	}
+
+	if (usage & PIPE_BUFFER_USAGE_GPU_READ)
+		flags |= DRM_BO_FLAG_READ;
+
+	if (usage & PIPE_BUFFER_USAGE_GPU_WRITE)
+		flags |= DRM_BO_FLAG_WRITE;
+
+	/* drm complains if we don't set any read/write flags.
+	 */
+	if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0)
+		flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE;
+
+	buffer->pool = pool;
+	driGenBuffers( buffer->pool,
+		"pipe buffer", 1, &buffer->driBO, alignment, flags, 0 );
+
+	driBOData( buffer->driBO, size, NULL, buffer->pool, 0 );
+
+	return &buffer->base;
+}
+
+
+static struct pipe_buffer *
+intel_be_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes)
+{
+	struct intel_be_buffer *buffer = CALLOC_STRUCT( intel_be_buffer );
+	struct intel_be_device *iws = intel_be_device(winsys);
+
+	driGenUserBuffer( iws->regionPool,
+			  "pipe user buffer", &buffer->driBO, ptr, bytes );
+
+	buffer->base.refcount = 1;
+
+	return &buffer->base;
+}
+
+struct pipe_buffer *
+intel_be_buffer_from_handle(struct intel_be_device *device,
+                            const char* name, unsigned handle)
+{
+	struct intel_be_buffer *be_buf = malloc(sizeof(*be_buf));
+	struct pipe_buffer *buffer;
+
+	if (!be_buf)
+		goto err;
+
+	memset(be_buf, 0, sizeof(*be_buf));
+
+	driGenBuffers(device->staticPool, name, 1, &be_buf->driBO, 0, 0, 0);
+    driBOSetReferenced(be_buf->driBO, handle);
+
+	if (0) /** XXX TODO check error */
+		goto err_bo;
+	
+	buffer = &be_buf->base;
+	buffer->refcount = 1;
+	buffer->alignment = 0;
+	buffer->usage = 0;
+	buffer->size = driBOSize(be_buf->driBO);
+
+	return buffer;
+err_bo:
+	free(be_buf);
+err:
+	return NULL;
+}
+
+
+static struct pipe_buffer *
+intel_i915_surface_buffer_create(struct pipe_winsys *winsys,
+				 unsigned width, unsigned height,
+				 enum pipe_format format,
+				 unsigned usage,
+				 unsigned *stride)
+{
+	const unsigned alignment = 64;
+	struct pipe_format_block block;
+	unsigned nblocksx, nblocksy;
+
+	pf_get_block(format, &block);
+	nblocksx = pf_get_nblocksx(&block, width);
+	nblocksy = pf_get_nblocksy(&block, height);
+	*stride = round_up(nblocksx * block.size, alignment);
+
+	return winsys->buffer_create(winsys, alignment,
+				     usage,
+				     *stride * nblocksy);
+}
+
+
+/*
+ * Fence functions
+ */
+
+static void
+intel_be_fence_reference( struct pipe_winsys *sws,
+		       struct pipe_fence_handle **ptr,
+		       struct pipe_fence_handle *fence )
+{
+	if (*ptr)
+		driFenceUnReference((struct _DriFenceObject **)ptr);
+
+	if (fence)
+		*ptr = (struct pipe_fence_handle *)driFenceReference((struct _DriFenceObject *)fence);
+}
+
+static int
+intel_be_fence_signalled( struct pipe_winsys *sws,
+		       struct pipe_fence_handle *fence,
+		       unsigned flag )
+{
+	return driFenceSignaled((struct _DriFenceObject *)fence, flag);
+}
+
+static int
+intel_be_fence_finish( struct pipe_winsys *sws,
+		    struct pipe_fence_handle *fence,
+		    unsigned flag )
+{
+	return driFenceFinish((struct _DriFenceObject *)fence, flag, 0);
+}
+
+
+/*
+ * Misc functions
+ */
+
+boolean
+intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id)
+{
+	dev->fd = fd;
+	dev->max_batch_size = 16 * 4096;
+	dev->max_vertex_size = 128 * 4096;
+
+	dev->base.buffer_create = intel_be_buffer_create;
+	dev->base.user_buffer_create = intel_be_user_buffer_create;
+	dev->base.buffer_map = intel_be_buffer_map;
+	dev->base.buffer_unmap = intel_be_buffer_unmap;
+	dev->base.buffer_destroy = intel_be_buffer_destroy;
+	dev->base.surface_buffer_create = intel_i915_surface_buffer_create;
+	dev->base.fence_reference = intel_be_fence_reference;
+	dev->base.fence_signalled = intel_be_fence_signalled;
+	dev->base.fence_finish = intel_be_fence_finish;
+
+#if 0 /* Set by the winsys */
+	dev->base.flush_frontbuffer = intel_flush_frontbuffer;
+	dev->base.get_name = intel_get_name;
+#endif
+
+	dev->fMan = driInitFreeSlabManager(10, 10);
+	dev->fenceMgr = driFenceMgrTTMInit(dev->fd);
+
+	dev->mallocPool = driMallocPoolInit();
+	dev->staticPool = driDRMPoolInit(dev->fd);
+	/* Sizes: 64 128 256 512 1024 2048 4096 8192 16384 32768 */
+	dev->regionPool = driSlabPoolInit(dev->fd,
+					  DRM_BO_FLAG_READ |
+					  DRM_BO_FLAG_WRITE |
+					  DRM_BO_FLAG_MEM_TT,
+					  DRM_BO_FLAG_READ |
+					  DRM_BO_FLAG_WRITE |
+					  DRM_BO_FLAG_MEM_TT,
+					  64,
+					  10, 120, 4096 * 64, 0,
+					  dev->fMan);
+
+	dev->vertexPool = driSlabPoolInit(dev->fd,
+					  DRM_BO_FLAG_READ |
+					  DRM_BO_FLAG_WRITE |
+					  DRM_BO_FLAG_MEM_TT,
+					  DRM_BO_FLAG_READ |
+					  DRM_BO_FLAG_WRITE |
+					  DRM_BO_FLAG_MEM_TT,
+					  dev->max_vertex_size,
+					  1, 120, dev->max_vertex_size * 4, 0,
+					  dev->fMan);
+
+	dev->batchPool = driSlabPoolInit(dev->fd,
+					 DRM_BO_FLAG_EXE |
+					 DRM_BO_FLAG_MEM_TT,
+					 DRM_BO_FLAG_EXE |
+					 DRM_BO_FLAG_MEM_TT,
+					 dev->max_batch_size,
+					 1, 40, dev->max_batch_size * 16, 0,
+					 dev->fMan);
+
+	/* Fill in this struct with callbacks that i915simple will need to
+	 * communicate with the window system, buffer manager, etc.
+	 */
+	dev->screen = i915_create_screen(&dev->base, id);
+
+	return true;
+}
+
+void
+intel_be_destroy_device(struct intel_be_device *dev)
+{
+	driPoolTakeDown(dev->mallocPool);
+	driPoolTakeDown(dev->staticPool);
+	driPoolTakeDown(dev->regionPool);
+	driPoolTakeDown(dev->vertexPool);
+	driPoolTakeDown(dev->batchPool);
+
+	/** TODO takedown fenceMgr and fMan */
+}
diff --git a/src/gallium/winsys/drm/intel/common/intel_be_device.h b/src/gallium/winsys/drm/intel/common/intel_be_device.h
new file mode 100644
index 0000000000..534d638b6a
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/intel_be_device.h
@@ -0,0 +1,72 @@
+#ifndef INTEL_DRM_DEVICE_H
+#define INTEL_DRM_DEVICE_H
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_context.h"
+
+/*
+ * Device
+ */
+
+struct intel_be_device
+{
+	struct pipe_winsys base;
+
+	/**
+	 * Hw level screen
+	 */
+	struct pipe_screen *screen;
+
+	int fd; /**< Drm file discriptor */
+
+	size_t max_batch_size;
+	size_t max_vertex_size;
+
+	struct _DriFenceMgr *fenceMgr;
+
+	struct _DriBufferPool *batchPool;
+	struct _DriBufferPool *regionPool;
+	struct _DriBufferPool *mallocPool;
+	struct _DriBufferPool *vertexPool;
+	struct _DriBufferPool *staticPool;
+	struct _DriFreeSlabManager *fMan;
+};
+
+boolean
+intel_be_init_device(struct intel_be_device *device, int fd, unsigned id);
+
+void
+intel_be_destroy_device(struct intel_be_device *dev);
+
+/*
+ * Buffer
+ */
+
+struct intel_be_buffer {
+	struct pipe_buffer base;
+	struct _DriBufferPool *pool;
+	struct _DriBufferObject *driBO;
+};
+
+/**
+ * Create a be buffer from a drm bo handle
+ *
+ * Takes a reference
+ */
+struct pipe_buffer *
+intel_be_buffer_from_handle(struct intel_be_device *device,
+                            const char* name, unsigned handle);
+
+static INLINE struct intel_be_buffer *
+intel_be_buffer(struct pipe_buffer *buf)
+{
+	return (struct intel_be_buffer *)buf;
+}
+
+static INLINE struct _DriBufferObject *
+dri_bo(struct pipe_buffer *buf)
+{
+	return intel_be_buffer(buf)->driBO;
+}
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c
new file mode 100644
index 0000000000..517a97b3ee
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c
@@ -0,0 +1,949 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+/*
+ * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com>
+ *          Keith Whitwell <keithw-at-tungstengraphics-dot-com>
+ */
+
+#include <xf86drm.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include "pipe/p_thread.h"
+#include "errno.h"
+#include "ws_dri_bufmgr.h"
+#include "string.h"
+#include "pipe/p_debug.h"
+#include "ws_dri_bufpool.h"
+#include "ws_dri_fencemgr.h"
+
+
+/*
+ * This lock is here to protect drmBO structs changing underneath us during a
+ * validate list call, since validatelist cannot take individiual locks for
+ * each drmBO. Validatelist takes this lock in write mode. Any access to an
+ * individual drmBO should take this lock in read mode, since in that case, the
+ * driBufferObject mutex will protect the access. Locking order is
+ * driBufferObject mutex - > this rw lock.
+ */
+
+pipe_static_mutex(bmMutex);
+pipe_static_condvar(bmCond);
+
+static int kernelReaders = 0;
+static int num_buffers = 0;
+static int num_user_buffers = 0;
+
+static drmBO *drmBOListBuf(void *iterator)
+{
+    drmBONode *node;
+    drmMMListHead *l = (drmMMListHead *) iterator;
+    node = DRMLISTENTRY(drmBONode, l, head);
+    return node->buf;
+}
+
+static void *drmBOListIterator(drmBOList *list)
+{
+    void *ret = list->list.next;
+
+    if (ret == &list->list)
+	return NULL;
+    return ret;
+}
+
+static void *drmBOListNext(drmBOList *list, void *iterator)
+{
+    void *ret;
+
+    drmMMListHead *l = (drmMMListHead *) iterator;
+    ret = l->next;
+    if (ret == &list->list)
+	return NULL;
+    return ret;
+}
+
+static drmBONode *drmAddListItem(drmBOList *list, drmBO *item,
+				 uint64_t arg0,
+				 uint64_t arg1)
+{
+    drmBONode *node;
+    drmMMListHead *l;
+
+    l = list->free.next;
+    if (l == &list->free) {
+	node = (drmBONode *) malloc(sizeof(*node));
+	if (!node) {
+	    return NULL;
+	}
+	list->numCurrent++;
+    }
+    else {
+	DRMLISTDEL(l);
+	node = DRMLISTENTRY(drmBONode, l, head);
+    }
+    node->buf = item;
+    node->arg0 = arg0;
+    node->arg1 = arg1;
+    DRMLISTADD(&node->head, &list->list);
+    list->numOnList++;
+    return node;
+}
+
+static int drmAddValidateItem(drmBOList *list, drmBO *buf, uint64_t flags,
+			      uint64_t mask, int *newItem)
+{
+    drmBONode *node, *cur;
+    drmMMListHead *l;
+
+    *newItem = 0;
+    cur = NULL;
+
+    for (l = list->list.next; l != &list->list; l = l->next) {
+	node = DRMLISTENTRY(drmBONode, l, head);
+	if (node->buf == buf) {
+	    cur = node;
+	    break;
+	}
+    }
+    if (!cur) {
+	cur = drmAddListItem(list, buf, flags, mask);
+	if (!cur) {
+	    return -ENOMEM;
+	}
+	*newItem = 1;
+	cur->arg0 = flags;
+	cur->arg1 = mask;
+    }
+    else {
+        uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM;
+	uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM;
+
+	if (mask & cur->arg1 & ~DRM_BO_MASK_MEM  & (cur->arg0 ^ flags)) {
+	    return -EINVAL;
+	}
+
+	cur->arg1 |= mask;
+	cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask);
+
+	if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) &&
+	    (cur->arg0 & DRM_BO_MASK_MEM) == 0) {
+	    return -EINVAL;
+	}
+    }
+    return 0;
+}
+
+static void drmBOFreeList(drmBOList *list)
+{
+    drmBONode *node;
+    drmMMListHead *l;
+
+    l = list->list.next;
+    while(l != &list->list) {
+	DRMLISTDEL(l);
+	node = DRMLISTENTRY(drmBONode, l, head);
+	free(node);
+	l = list->list.next;
+	list->numCurrent--;
+	list->numOnList--;
+    }
+
+    l = list->free.next;
+    while(l != &list->free) {
+	DRMLISTDEL(l);
+	node = DRMLISTENTRY(drmBONode, l, head);
+	free(node);
+	l = list->free.next;
+	list->numCurrent--;
+    }
+}
+
+static int drmAdjustListNodes(drmBOList *list)
+{
+    drmBONode *node;
+    drmMMListHead *l;
+    int ret = 0;
+
+    while(list->numCurrent < list->numTarget) {
+	node = (drmBONode *) malloc(sizeof(*node));
+	if (!node) {
+	    ret = -ENOMEM;
+	    break;
+	}
+	list->numCurrent++;
+	DRMLISTADD(&node->head, &list->free);
+    }
+
+    while(list->numCurrent > list->numTarget) {
+	l = list->free.next;
+	if (l == &list->free)
+	    break;
+	DRMLISTDEL(l);
+	node = DRMLISTENTRY(drmBONode, l, head);
+	free(node);
+	list->numCurrent--;
+    }
+    return ret;
+}
+
+static int drmBOCreateList(int numTarget, drmBOList *list)
+{
+    DRMINITLISTHEAD(&list->list);
+    DRMINITLISTHEAD(&list->free);
+    list->numTarget = numTarget;
+    list->numCurrent = 0;
+    list->numOnList = 0;
+    return drmAdjustListNodes(list);
+}
+
+static int drmBOResetList(drmBOList *list)
+{
+    drmMMListHead *l;
+    int ret;
+
+    ret = drmAdjustListNodes(list);
+    if (ret)
+	return ret;
+
+    l = list->list.next;
+    while (l != &list->list) {
+	DRMLISTDEL(l);
+	DRMLISTADD(l, &list->free);
+	list->numOnList--;
+	l = list->list.next;
+    }
+    return drmAdjustListNodes(list);
+}
+
+void driWriteLockKernelBO(void)
+{
+    pipe_mutex_lock(bmMutex);
+    while(kernelReaders != 0)
+	pipe_condvar_wait(bmCond, bmMutex);
+}
+
+void driWriteUnlockKernelBO(void)
+{
+    pipe_mutex_unlock(bmMutex);
+}
+
+void driReadLockKernelBO(void)
+{
+    pipe_mutex_lock(bmMutex);
+    kernelReaders++;
+    pipe_mutex_unlock(bmMutex);
+}
+
+void driReadUnlockKernelBO(void)
+{
+    pipe_mutex_lock(bmMutex);
+    if (--kernelReaders == 0)
+       pipe_condvar_broadcast(bmCond);
+    pipe_mutex_unlock(bmMutex);
+}
+
+
+
+
+/*
+ * TODO: Introduce fence pools in the same way as
+ * buffer object pools.
+ */
+
+typedef struct _DriBufferObject
+{
+   DriBufferPool *pool;
+   pipe_mutex mutex;
+   int refCount;
+   const char *name;
+   uint64_t flags;
+   unsigned hint;
+   unsigned alignment;
+   unsigned createdByReference;
+   void *private;
+   /* user-space buffer: */
+   unsigned userBuffer;
+   void *userData;
+   unsigned userSize;
+} DriBufferObject;
+
+typedef struct _DriBufferList {
+    drmBOList drmBuffers;  /* List of kernel buffers needing validation */
+    drmBOList driBuffers;  /* List of user-space buffers needing validation */
+} DriBufferList;
+
+
+void
+bmError(int val, const char *file, const char *function, int line)
+{
+   printf("Fatal video memory manager error \"%s\".\n"
+          "Check kernel logs or set the LIBGL_DEBUG\n"
+          "environment variable to \"verbose\" for more info.\n"
+          "Detected in file %s, line %d, function %s.\n",
+          strerror(-val), file, line, function);
+#ifndef NDEBUG
+   abort();
+#else
+   abort();
+#endif
+}
+
+extern drmBO *
+driBOKernel(struct _DriBufferObject *buf)
+{
+   drmBO *ret;
+
+   driReadLockKernelBO();
+   pipe_mutex_lock(buf->mutex);
+   assert(buf->private != NULL);
+   ret = buf->pool->kernel(buf->pool, buf->private);
+   if (!ret)
+      BM_CKFATAL(-EINVAL);
+   pipe_mutex_unlock(buf->mutex);
+   driReadUnlockKernelBO();
+
+   return ret;
+}
+
+void
+driBOWaitIdle(struct _DriBufferObject *buf, int lazy)
+{
+
+  /*
+   * This function may block. Is it sane to keep the mutex held during
+   * that time??
+   */
+
+   pipe_mutex_lock(buf->mutex);
+   BM_CKFATAL(buf->pool->waitIdle(buf->pool, buf->private, &buf->mutex, lazy));
+   pipe_mutex_unlock(buf->mutex);
+}
+
+void *
+driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint)
+{
+   void *virtual;
+   int retval;
+
+   if (buf->userBuffer) {
+      return buf->userData;
+   }
+
+   pipe_mutex_lock(buf->mutex);
+   assert(buf->private != NULL);
+   retval = buf->pool->map(buf->pool, buf->private, flags, hint,
+			   &buf->mutex, &virtual);
+   pipe_mutex_unlock(buf->mutex);
+
+   return retval == 0 ? virtual : NULL;
+}
+
+void
+driBOUnmap(struct _DriBufferObject *buf)
+{
+   if (buf->userBuffer)
+      return;
+
+   assert(buf->private != NULL);
+   pipe_mutex_lock(buf->mutex);
+   BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private));
+   pipe_mutex_unlock(buf->mutex);
+}
+
+unsigned long
+driBOOffset(struct _DriBufferObject *buf)
+{
+   unsigned long ret;
+
+   assert(buf->private != NULL);
+
+   pipe_mutex_lock(buf->mutex);
+   ret = buf->pool->offset(buf->pool, buf->private);
+   pipe_mutex_unlock(buf->mutex);
+   return ret;
+}
+
+unsigned long
+driBOPoolOffset(struct _DriBufferObject *buf)
+{
+   unsigned long ret;
+
+   assert(buf->private != NULL);
+
+   pipe_mutex_lock(buf->mutex);
+   ret = buf->pool->poolOffset(buf->pool, buf->private);
+   pipe_mutex_unlock(buf->mutex);
+   return ret;
+}
+
+uint64_t
+driBOFlags(struct _DriBufferObject *buf)
+{
+   uint64_t ret;
+
+   assert(buf->private != NULL);
+
+   driReadLockKernelBO();
+   pipe_mutex_lock(buf->mutex);
+   ret = buf->pool->flags(buf->pool, buf->private);
+   pipe_mutex_unlock(buf->mutex);
+   driReadUnlockKernelBO();
+   return ret;
+}
+
+struct _DriBufferObject *
+driBOReference(struct _DriBufferObject *buf)
+{
+   pipe_mutex_lock(buf->mutex);
+   if (++buf->refCount == 1) {
+      pipe_mutex_unlock(buf->mutex);
+      BM_CKFATAL(-EINVAL);
+   }
+   pipe_mutex_unlock(buf->mutex);
+   return buf;
+}
+
+void
+driBOUnReference(struct _DriBufferObject *buf)
+{
+   int tmp;
+
+   if (!buf)
+      return;
+
+   pipe_mutex_lock(buf->mutex);
+   tmp = --buf->refCount;
+   if (!tmp) {
+      pipe_mutex_unlock(buf->mutex);
+      if (buf->private) {
+	 if (buf->createdByReference)
+	    buf->pool->unreference(buf->pool, buf->private);
+	 else
+	    buf->pool->destroy(buf->pool, buf->private);
+      }
+      if (buf->userBuffer)
+	 num_user_buffers--;
+      else
+	 num_buffers--;
+      free(buf);
+   } else
+     pipe_mutex_unlock(buf->mutex);
+
+}
+
+
+int
+driBOData(struct _DriBufferObject *buf,
+          unsigned size, const void *data,
+	  DriBufferPool *newPool,
+	  uint64_t flags)
+{
+   void *virtual = NULL;
+   int newBuffer;
+   int retval = 0;
+   struct _DriBufferPool *pool;
+
+   assert(!buf->userBuffer); /* XXX just do a memcpy? */
+
+   pipe_mutex_lock(buf->mutex);
+   pool = buf->pool;
+
+   if (pool == NULL && newPool != NULL) {
+       buf->pool = newPool;
+       pool = newPool;
+   }
+   if (newPool == NULL)
+       newPool = pool;
+
+   if (!pool->create) {
+      assert((size_t)"driBOData called on invalid buffer\n" & 0);
+      BM_CKFATAL(-EINVAL);
+   }
+
+   newBuffer = (!buf->private || pool != newPool ||
+ 		pool->size(pool, buf->private) < size);
+
+   if (!flags)
+       flags = buf->flags;
+
+   if (newBuffer) {
+
+       if (buf->createdByReference) {
+	  assert((size_t)"driBOData requiring resizing called on shared buffer.\n" & 0);
+	  BM_CKFATAL(-EINVAL);
+       }
+
+       if (buf->private)
+	   buf->pool->destroy(buf->pool, buf->private);
+
+       pool = newPool;
+       buf->pool = newPool;
+       buf->private = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE,
+				  buf->alignment);
+      if (!buf->private)
+	  retval = -ENOMEM;
+
+      if (retval == 0)
+	  retval = pool->map(pool, buf->private,
+			     DRM_BO_FLAG_WRITE,
+			     DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual);
+   } else if (pool->map(pool, buf->private, DRM_BO_FLAG_WRITE,
+			DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual)) {
+       /*
+	* Buffer is busy. need to create a new one.
+	*/
+
+       void *newBuf;
+
+       newBuf = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE,
+			     buf->alignment);
+       if (newBuf) {
+	   buf->pool->destroy(buf->pool, buf->private);
+	   buf->private = newBuf;
+       }
+
+       retval = pool->map(pool, buf->private,
+			  DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual);
+   } else {
+       uint64_t flag_diff = flags ^ buf->flags;
+
+       /*
+	* We might need to change buffer flags.
+	*/
+
+       if (flag_diff){
+	   assert(pool->setStatus != NULL);
+	   BM_CKFATAL(pool->unmap(pool, buf->private));
+	   BM_CKFATAL(pool->setStatus(pool, buf->private, flag_diff,
+				      buf->flags));
+	   if (!data)
+	     goto out;
+
+	   retval = pool->map(pool, buf->private,
+			      DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual);
+       }
+   }
+
+   if (retval == 0) {
+      if (data)
+	 memcpy(virtual, data, size);
+
+      BM_CKFATAL(pool->unmap(pool, buf->private));
+   }
+
+ out:
+   pipe_mutex_unlock(buf->mutex);
+
+   return retval;
+}
+
+void
+driBOSubData(struct _DriBufferObject *buf,
+             unsigned long offset, unsigned long size, const void *data)
+{
+   void *virtual;
+
+   assert(!buf->userBuffer); /* XXX just do a memcpy? */
+
+   pipe_mutex_lock(buf->mutex);
+   if (size && data) {
+      BM_CKFATAL(buf->pool->map(buf->pool, buf->private,
+                                DRM_BO_FLAG_WRITE, 0, &buf->mutex,
+				&virtual));
+      memcpy((unsigned char *) virtual + offset, data, size);
+      BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private));
+   }
+   pipe_mutex_unlock(buf->mutex);
+}
+
+void
+driBOGetSubData(struct _DriBufferObject *buf,
+                unsigned long offset, unsigned long size, void *data)
+{
+   void *virtual;
+
+   assert(!buf->userBuffer); /* XXX just do a memcpy? */
+
+   pipe_mutex_lock(buf->mutex);
+   if (size && data) {
+      BM_CKFATAL(buf->pool->map(buf->pool, buf->private,
+                                DRM_BO_FLAG_READ, 0, &buf->mutex, &virtual));
+      memcpy(data, (unsigned char *) virtual + offset, size);
+      BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private));
+   }
+   pipe_mutex_unlock(buf->mutex);
+}
+
+void
+driBOSetReferenced(struct _DriBufferObject *buf,
+		   unsigned long handle)
+{
+   pipe_mutex_lock(buf->mutex);
+   if (buf->private != NULL) {
+      assert((size_t)"Invalid buffer for setReferenced\n" & 0);
+      BM_CKFATAL(-EINVAL);
+
+   }
+   if (buf->pool->reference == NULL) {
+      assert((size_t)"Invalid buffer pool for setReferenced\n" & 0);
+      BM_CKFATAL(-EINVAL);
+   }
+   buf->private = buf->pool->reference(buf->pool, handle);
+   if (!buf->private) {
+      assert((size_t)"Invalid buffer pool for setStatic\n" & 0);
+      BM_CKFATAL(-ENOMEM);
+   }
+   buf->createdByReference = TRUE;
+   buf->flags = buf->pool->kernel(buf->pool, buf->private)->flags;
+   pipe_mutex_unlock(buf->mutex);
+}
+
+int
+driGenBuffers(struct _DriBufferPool *pool,
+              const char *name,
+              unsigned n,
+              struct _DriBufferObject *buffers[],
+              unsigned alignment, uint64_t flags, unsigned hint)
+{
+   struct _DriBufferObject *buf;
+   int i;
+
+   flags = (flags) ? flags : DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_VRAM |
+      DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE;
+
+   ++num_buffers;
+
+   assert(pool);
+
+   for (i = 0; i < n; ++i) {
+      buf = (struct _DriBufferObject *) calloc(1, sizeof(*buf));
+      if (!buf)
+	 return -ENOMEM;
+
+      pipe_mutex_init(buf->mutex);
+      pipe_mutex_lock(buf->mutex);
+      buf->refCount = 1;
+      buf->flags = flags;
+      buf->hint = hint;
+      buf->name = name;
+      buf->alignment = alignment;
+      buf->pool = pool;
+      buf->createdByReference = 0;
+      pipe_mutex_unlock(buf->mutex);
+      buffers[i] = buf;
+   }
+   return 0;
+}
+
+void
+driGenUserBuffer(struct _DriBufferPool *pool,
+                 const char *name,
+                 struct _DriBufferObject **buffers,
+                 void *ptr, unsigned bytes)
+{
+   const unsigned alignment = 1, flags = 0, hint = 0;
+
+   --num_buffers; /* JB: is inced in GenBuffes */
+   driGenBuffers(pool, name, 1, buffers, alignment, flags, hint);
+   ++num_user_buffers;
+
+   (*buffers)->userBuffer = 1;
+   (*buffers)->userData = ptr;
+   (*buffers)->userSize = bytes;
+}
+
+void
+driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[])
+{
+   int i;
+
+   for (i = 0; i < n; ++i) {
+      driBOUnReference(buffers[i]);
+   }
+}
+
+
+void
+driInitBufMgr(int fd)
+{
+   ;
+}
+
+/*
+ * Note that lists are per-context and don't need mutex protection.
+ */
+
+struct _DriBufferList *
+driBOCreateList(int target)
+{
+    struct _DriBufferList *list = calloc(sizeof(*list), 1);
+
+    BM_CKFATAL(drmBOCreateList(target, &list->drmBuffers));
+    BM_CKFATAL(drmBOCreateList(target, &list->driBuffers));
+    return list;
+}
+
+int
+driBOResetList(struct _DriBufferList * list)
+{
+    int ret;
+    ret = drmBOResetList(&list->drmBuffers);
+    if (ret)
+	return ret;
+    ret = drmBOResetList(&list->driBuffers);
+    return ret;
+}
+
+void
+driBOFreeList(struct _DriBufferList * list)
+{
+   drmBOFreeList(&list->drmBuffers);
+   drmBOFreeList(&list->driBuffers);
+   free(list);
+}
+
+
+/*
+ * Copied from libdrm, because it is needed by driAddValidateItem.
+ */
+
+static drmBONode *
+driAddListItem(drmBOList * list, drmBO * item,
+	       uint64_t arg0, uint64_t arg1)
+{
+    drmBONode *node;
+    drmMMListHead *l;
+
+    l = list->free.next;
+    if (l == &list->free) {
+	node = (drmBONode *) malloc(sizeof(*node));
+	if (!node) {
+	    return NULL;
+	}
+	list->numCurrent++;
+    } else {
+	DRMLISTDEL(l);
+	node = DRMLISTENTRY(drmBONode, l, head);
+    }
+    memset(&node->bo_arg, 0, sizeof(node->bo_arg));
+    node->buf = item;
+    node->arg0 = arg0;
+    node->arg1 = arg1;
+    DRMLISTADDTAIL(&node->head, &list->list);
+    list->numOnList++;
+    return node;
+}
+
+/*
+ * Slightly modified version compared to the libdrm version.
+ * This one returns the list index of the buffer put on the list.
+ */
+
+static int
+driAddValidateItem(drmBOList * list, drmBO * buf, uint64_t flags,
+		   uint64_t mask, int *itemLoc,
+		   struct _drmBONode **pnode)
+{
+    drmBONode *node, *cur;
+    drmMMListHead *l;
+    int count = 0;
+
+    cur = NULL;
+
+    for (l = list->list.next; l != &list->list; l = l->next) {
+	node = DRMLISTENTRY(drmBONode, l, head);
+	if (node->buf == buf) {
+	    cur = node;
+	    break;
+	}
+	count++;
+    }
+    if (!cur) {
+	cur = driAddListItem(list, buf, flags, mask);
+	if (!cur)
+	    return -ENOMEM;
+
+	cur->arg0 = flags;
+	cur->arg1 = mask;
+    } else {
+        uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM;
+	uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM;
+
+	if (mask & cur->arg1 & ~DRM_BO_MASK_MEM  & (cur->arg0 ^ flags)) {
+	    return -EINVAL;
+	}
+
+	cur->arg1 |= mask;
+	cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask);
+
+	if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) &&
+	    (cur->arg0 & DRM_BO_MASK_MEM) == 0) {
+	    return -EINVAL;
+	}
+    }
+    *itemLoc = count;
+    *pnode = cur;
+    return 0;
+}
+
+
+void
+driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf,
+                 uint64_t flags, uint64_t mask, int *itemLoc,
+		 struct _drmBONode **node)
+{
+   int newItem;
+
+   pipe_mutex_lock(buf->mutex);
+   BM_CKFATAL(driAddValidateItem(&list->drmBuffers,
+				 buf->pool->kernel(buf->pool, buf->private),
+                                 flags, mask, itemLoc, node));
+   BM_CKFATAL(drmAddValidateItem(&list->driBuffers, (drmBO *) buf,
+				 flags, mask, &newItem));
+   if (newItem)
+     buf->refCount++;
+
+   pipe_mutex_unlock(buf->mutex);
+}
+
+drmBOList *driGetdrmBOList(struct _DriBufferList *list)
+{
+	driWriteLockKernelBO();
+	return &list->drmBuffers;
+}
+
+void driPutdrmBOList(struct _DriBufferList *list)
+{
+	driWriteUnlockKernelBO();
+}
+
+
+void
+driBOFence(struct _DriBufferObject *buf, struct _DriFenceObject *fence)
+{
+   pipe_mutex_lock(buf->mutex);
+   if (buf->pool->fence)
+       BM_CKFATAL(buf->pool->fence(buf->pool, buf->private, fence));
+   pipe_mutex_unlock(buf->mutex);
+
+}
+
+void
+driBOUnrefUserList(struct _DriBufferList *list)
+{
+    struct _DriBufferObject *buf;
+    void *curBuf;
+
+    curBuf = drmBOListIterator(&list->driBuffers);
+    while (curBuf) {
+	buf = (struct _DriBufferObject *)drmBOListBuf(curBuf);
+	driBOUnReference(buf);
+	curBuf = drmBOListNext(&list->driBuffers, curBuf);
+    }
+}
+
+struct _DriFenceObject *
+driBOFenceUserList(struct _DriFenceMgr *mgr,
+		   struct _DriBufferList *list, const char *name,
+		   drmFence *kFence)
+{
+    struct _DriFenceObject *fence;
+    struct _DriBufferObject *buf;
+    void *curBuf;
+
+    fence = driFenceCreate(mgr, kFence->fence_class, kFence->type,
+			   kFence, sizeof(*kFence));
+    curBuf = drmBOListIterator(&list->driBuffers);
+
+   /*
+    * User-space fencing callbacks.
+    */
+
+   while (curBuf) {
+        buf = (struct _DriBufferObject *) drmBOListBuf(curBuf);
+	driBOFence(buf, fence);
+	driBOUnReference(buf);
+	curBuf = drmBOListNext(&list->driBuffers, curBuf);
+   }
+
+   driBOResetList(list);
+   return fence;
+}
+
+void
+driBOValidateUserList(struct _DriBufferList * list)
+{
+    void *curBuf;
+    struct _DriBufferObject *buf;
+
+    curBuf = drmBOListIterator(&list->driBuffers);
+
+    /*
+     * User-space validation callbacks.
+     */
+
+    while (curBuf) {
+	buf = (struct _DriBufferObject *) drmBOListBuf(curBuf);
+	pipe_mutex_lock(buf->mutex);
+	if (buf->pool->validate)
+	    BM_CKFATAL(buf->pool->validate(buf->pool, buf->private, &buf->mutex));
+	pipe_mutex_unlock(buf->mutex);
+	curBuf = drmBOListNext(&list->driBuffers, curBuf);
+    }
+}
+
+
+void
+driPoolTakeDown(struct _DriBufferPool *pool)
+{
+   pool->takeDown(pool);
+
+}
+
+unsigned long
+driBOSize(struct _DriBufferObject *buf)
+{
+  unsigned long size;
+
+   pipe_mutex_lock(buf->mutex);
+   size = buf->pool->size(buf->pool, buf->private);
+   pipe_mutex_unlock(buf->mutex);
+
+  return size;
+
+}
+
+drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list)
+{
+    return &list->drmBuffers;
+}
+
+drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list)
+{
+    return &list->driBuffers;
+}
+
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.h b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.h
new file mode 100644
index 0000000000..e6c0cff0a0
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.h
@@ -0,0 +1,138 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+/*
+ * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com>
+ *          Keith Whitwell <keithw-at-tungstengraphics-dot-com>
+ */
+
+#ifndef _PSB_BUFMGR_H_
+#define _PSB_BUFMGR_H_
+#include <xf86mm.h>
+#include "i915_drm.h"
+#include "ws_dri_fencemgr.h"
+
+typedef struct _drmBONode
+{
+    drmMMListHead head;
+    drmBO *buf;
+    struct drm_i915_op_arg bo_arg;
+    uint64_t arg0;
+    uint64_t arg1;
+} drmBONode;
+
+typedef struct _drmBOList {
+    unsigned numTarget;
+    unsigned numCurrent;
+    unsigned numOnList;
+    drmMMListHead list;
+    drmMMListHead free;
+} drmBOList;
+
+
+struct _DriFenceObject;
+struct _DriBufferObject;
+struct _DriBufferPool;
+struct _DriBufferList;
+
+/*
+ * Return a pointer to the libdrm buffer object this DriBufferObject
+ * uses.
+ */
+
+extern drmBO *driBOKernel(struct _DriBufferObject *buf);
+extern void *driBOMap(struct _DriBufferObject *buf, unsigned flags,
+                      unsigned hint);
+extern void driBOUnmap(struct _DriBufferObject *buf);
+extern unsigned long driBOOffset(struct _DriBufferObject *buf);
+extern unsigned long driBOPoolOffset(struct _DriBufferObject *buf);
+
+extern uint64_t driBOFlags(struct _DriBufferObject *buf);
+extern struct _DriBufferObject *driBOReference(struct _DriBufferObject *buf);
+extern void driBOUnReference(struct _DriBufferObject *buf);
+
+extern int driBOData(struct _DriBufferObject *r_buf,
+		     unsigned size, const void *data,
+		     struct _DriBufferPool *pool, uint64_t flags);
+
+extern void driBOSubData(struct _DriBufferObject *buf,
+                         unsigned long offset, unsigned long size,
+                         const void *data);
+extern void driBOGetSubData(struct _DriBufferObject *buf,
+                            unsigned long offset, unsigned long size,
+                            void *data);
+extern int driGenBuffers(struct _DriBufferPool *pool,
+			 const char *name,
+			 unsigned n,
+			 struct _DriBufferObject *buffers[],
+			 unsigned alignment, uint64_t flags, unsigned hint);
+extern void driGenUserBuffer(struct _DriBufferPool *pool,
+                             const char *name,
+                             struct _DriBufferObject *buffers[],
+                             void *ptr, unsigned bytes);
+extern void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]);
+extern void driInitBufMgr(int fd);
+extern struct _DriBufferList *driBOCreateList(int target);
+extern int driBOResetList(struct _DriBufferList * list);
+extern void driBOAddListItem(struct _DriBufferList * list,
+			     struct _DriBufferObject *buf,
+                             uint64_t flags, uint64_t mask, int *itemLoc,
+			     struct _drmBONode **node);
+
+extern void driBOValidateList(int fd, struct _DriBufferList * list);
+extern void driBOFreeList(struct _DriBufferList * list);
+extern struct _DriFenceObject *driBOFenceUserList(struct _DriFenceMgr *mgr,
+						  struct _DriBufferList *list,
+						  const char *name,
+						  drmFence *kFence);
+extern void driBOUnrefUserList(struct _DriBufferList *list);
+extern void driBOValidateUserList(struct _DriBufferList * list);
+extern drmBOList *driGetdrmBOList(struct _DriBufferList *list);
+extern void driPutdrmBOList(struct _DriBufferList *list);
+
+extern void driBOFence(struct _DriBufferObject *buf,
+                       struct _DriFenceObject *fence);
+
+extern void driPoolTakeDown(struct _DriBufferPool *pool);
+extern void driBOSetReferenced(struct _DriBufferObject *buf,
+			       unsigned long handle);
+unsigned long driBOSize(struct _DriBufferObject *buf);
+extern void driBOWaitIdle(struct _DriBufferObject *buf, int lazy);
+extern void driPoolTakeDown(struct _DriBufferPool *pool);
+
+extern void driReadLockKernelBO(void);
+extern void driReadUnlockKernelBO(void);
+extern void driWriteLockKernelBO(void);
+extern void driWriteUnlockKernelBO(void);
+
+/*
+ * For debugging purposes.
+ */
+
+extern drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list);
+extern drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list);
+#endif
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h b/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h
new file mode 100644
index 0000000000..ad3b6f3931
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h
@@ -0,0 +1,102 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+/*
+ * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com>
+ */
+
+#ifndef _PSB_BUFPOOL_H_
+#define _PSB_BUFPOOL_H_
+
+#include <xf86drm.h>
+#include "pipe/p_thread.h"
+struct _DriFenceObject;
+
+typedef struct _DriBufferPool
+{
+   int fd;
+   int (*map) (struct _DriBufferPool * pool, void *private,
+               unsigned flags, int hint, pipe_mutex *mutex,
+	       void **virtual);
+   int (*unmap) (struct _DriBufferPool * pool, void *private);
+   int (*destroy) (struct _DriBufferPool * pool, void *private);
+   unsigned long (*offset) (struct _DriBufferPool * pool, void *private);
+   unsigned long (*poolOffset) (struct _DriBufferPool * pool, void *private);
+   uint64_t (*flags) (struct _DriBufferPool * pool, void *private);
+   unsigned long (*size) (struct _DriBufferPool * pool, void *private);
+   void *(*create) (struct _DriBufferPool * pool, unsigned long size,
+                    uint64_t flags, unsigned hint, unsigned alignment);
+   void *(*reference) (struct _DriBufferPool * pool, unsigned handle);
+   int (*unreference) (struct _DriBufferPool * pool, void *private);
+   int (*fence) (struct _DriBufferPool * pool, void *private,
+                 struct _DriFenceObject * fence);
+   drmBO *(*kernel) (struct _DriBufferPool * pool, void *private);
+   int (*validate) (struct _DriBufferPool * pool, void *private, pipe_mutex *mutex);
+   int (*waitIdle) (struct _DriBufferPool *pool, void *private, pipe_mutex *mutex,
+		    int lazy);
+   int (*setStatus)  (struct _DriBufferPool *pool, void *private,
+		      uint64_t flag_diff, uint64_t old_flags);
+   void (*takeDown) (struct _DriBufferPool * pool);
+   void *data;
+} DriBufferPool;
+
+extern void bmError(int val, const char *file, const char *function,
+                    int line);
+#define BM_CKFATAL(val)					       \
+  do{							       \
+    int tstVal = (val);					       \
+    if (tstVal) 					       \
+      bmError(tstVal, __FILE__, __FUNCTION__, __LINE__);       \
+  } while(0);
+
+
+/*
+ * Builtin pools.
+ */
+
+/*
+ * Kernel buffer objects. Size in multiples of page size. Page size aligned.
+ */
+
+extern struct _DriBufferPool *driDRMPoolInit(int fd);
+extern struct _DriBufferPool *driMallocPoolInit(void);
+
+struct _DriFreeSlabManager;
+extern struct _DriBufferPool * driSlabPoolInit(int fd, uint64_t flags,
+					       uint64_t validMask,
+					       uint32_t smallestSize,
+					       uint32_t numSizes,
+					       uint32_t desiredNumBuffers,
+					       uint32_t maxSlabSize,
+					       uint32_t pageAlignment,
+					       struct _DriFreeSlabManager *fMan);
+extern void driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan);
+extern struct _DriFreeSlabManager *
+driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec);
+
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c
new file mode 100644
index 0000000000..54618b1c82
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c
@@ -0,0 +1,268 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+/*
+ * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com>
+ */
+
+#include <xf86drm.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include "ws_dri_bufpool.h"
+#include "ws_dri_bufmgr.h"
+#include "assert.h"
+
+/*
+ * Buffer pool implementation using DRM buffer objects as DRI buffer objects.
+ */
+
+static void *
+pool_create(struct _DriBufferPool *pool,
+            unsigned long size, uint64_t flags, unsigned hint,
+            unsigned alignment)
+{
+   drmBO *buf = (drmBO *) malloc(sizeof(*buf));
+   int ret;
+   unsigned pageSize = getpagesize();
+
+   if (!buf)
+      return NULL;
+
+   if ((alignment > pageSize) && (alignment % pageSize)) {
+      free(buf);
+      return NULL;
+   }
+
+   ret = drmBOCreate(pool->fd, size, alignment / pageSize,
+		     NULL,
+                     flags, hint, buf);
+   if (ret) {
+      free(buf);
+      return NULL;
+   }
+
+   return (void *) buf;
+}
+
+static void *
+pool_reference(struct _DriBufferPool *pool, unsigned handle)
+{
+   drmBO *buf = (drmBO *) malloc(sizeof(*buf));
+   int ret;
+
+   if (!buf)
+      return NULL;
+
+   ret = drmBOReference(pool->fd, handle, buf);
+
+   if (ret) {
+      free(buf);
+      return NULL;
+   }
+
+   return (void *) buf;
+}
+
+static int
+pool_destroy(struct _DriBufferPool *pool, void *private)
+{
+   int ret;
+   drmBO *buf = (drmBO *) private;
+   driReadLockKernelBO();
+   ret = drmBOUnreference(pool->fd, buf);
+   free(buf);
+   driReadUnlockKernelBO();
+   return ret;
+}
+
+static int
+pool_unreference(struct _DriBufferPool *pool, void *private)
+{
+   int ret;
+   drmBO *buf = (drmBO *) private;
+   driReadLockKernelBO();
+   ret = drmBOUnreference(pool->fd, buf);
+   free(buf);
+   driReadUnlockKernelBO();
+   return ret;
+}
+
+static int
+pool_map(struct _DriBufferPool *pool, void *private, unsigned flags,
+         int hint, pipe_mutex *mutex, void **virtual)
+{
+   drmBO *buf = (drmBO *) private;
+   int ret;
+
+   driReadLockKernelBO();
+   ret = drmBOMap(pool->fd, buf, flags, hint, virtual);
+   driReadUnlockKernelBO();
+   return ret;
+}
+
+static int
+pool_unmap(struct _DriBufferPool *pool, void *private)
+{
+   drmBO *buf = (drmBO *) private;
+   int ret;
+
+   driReadLockKernelBO();
+   ret = drmBOUnmap(pool->fd, buf);
+   driReadUnlockKernelBO();
+
+   return ret;
+}
+
+static unsigned long
+pool_offset(struct _DriBufferPool *pool, void *private)
+{
+   drmBO *buf = (drmBO *) private;
+   unsigned long offset;
+
+   driReadLockKernelBO();
+   assert(buf->flags & DRM_BO_FLAG_NO_MOVE);
+   offset = buf->offset;
+   driReadUnlockKernelBO();
+
+   return buf->offset;
+}
+
+static unsigned long
+pool_poolOffset(struct _DriBufferPool *pool, void *private)
+{
+   return 0;
+}
+
+static uint64_t
+pool_flags(struct _DriBufferPool *pool, void *private)
+{
+   drmBO *buf = (drmBO *) private;
+   uint64_t flags;
+
+   driReadLockKernelBO();
+   flags = buf->flags;
+   driReadUnlockKernelBO();
+
+   return flags;
+}
+
+
+static unsigned long
+pool_size(struct _DriBufferPool *pool, void *private)
+{
+   drmBO *buf = (drmBO *) private;
+   unsigned long size;
+
+   driReadLockKernelBO();
+   size = buf->size;
+   driReadUnlockKernelBO();
+
+   return buf->size;
+}
+
+static int
+pool_fence(struct _DriBufferPool *pool, void *private,
+           struct _DriFenceObject *fence)
+{
+   /*
+    * Noop. The kernel handles all fencing.
+    */
+
+   return 0;
+}
+
+static drmBO *
+pool_kernel(struct _DriBufferPool *pool, void *private)
+{
+   return (drmBO *) private;
+}
+
+static int
+pool_waitIdle(struct _DriBufferPool *pool, void *private, pipe_mutex *mutex,
+	      int lazy)
+{
+   drmBO *buf = (drmBO *) private;
+   int ret;
+
+   driReadLockKernelBO();
+   ret = drmBOWaitIdle(pool->fd, buf, (lazy) ? DRM_BO_HINT_WAIT_LAZY:0);
+   driReadUnlockKernelBO();
+
+   return ret;
+}
+
+
+static void
+pool_takedown(struct _DriBufferPool *pool)
+{
+   free(pool);
+}
+
+/*static int
+pool_setStatus(struct _DriBufferPool *pool, void *private,
+	       uint64_t flag_diff, uint64_t old_flags)
+{
+   drmBO *buf = (drmBO *) private;
+   uint64_t new_flags = old_flags ^ flag_diff;
+   int ret;
+
+   driReadLockKernelBO();
+   ret = drmBOSetStatus(pool->fd, buf, new_flags, flag_diff,
+			0, 0, 0);
+   driReadUnlockKernelBO();
+   return ret;
+}*/
+
+struct _DriBufferPool *
+driDRMPoolInit(int fd)
+{
+   struct _DriBufferPool *pool;
+
+   pool = (struct _DriBufferPool *) malloc(sizeof(*pool));
+
+   if (!pool)
+      return NULL;
+
+   pool->fd = fd;
+   pool->map = &pool_map;
+   pool->unmap = &pool_unmap;
+   pool->destroy = &pool_destroy;
+   pool->offset = &pool_offset;
+   pool->poolOffset = &pool_poolOffset;
+   pool->flags = &pool_flags;
+   pool->size = &pool_size;
+   pool->create = &pool_create;
+   pool->fence = &pool_fence;
+   pool->kernel = &pool_kernel;
+   pool->validate = NULL;
+   pool->waitIdle = &pool_waitIdle;
+   pool->takeDown = &pool_takedown;
+   pool->reference = &pool_reference;
+   pool->unreference = &pool_unreference;
+   pool->data = NULL;
+   return pool;
+}
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c
new file mode 100644
index 0000000000..831c75d30c
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c
@@ -0,0 +1,377 @@
+#include "ws_dri_fencemgr.h"
+#include "pipe/p_thread.h"
+#include <xf86mm.h>
+#include <string.h>
+#include <unistd.h>
+
+/*
+ * Note: Locking order is
+ * _DriFenceObject::mutex
+ * _DriFenceMgr::mutex
+ */
+
+struct _DriFenceMgr {
+    /*
+     * Constant members. Need no mutex protection.
+     */
+    struct _DriFenceMgrCreateInfo info;
+    void *private;
+
+    /*
+     * These members are protected by this->mutex
+     */
+    pipe_mutex mutex;
+    int refCount;
+    drmMMListHead *heads;
+    int num_fences;
+};
+
+struct _DriFenceObject {
+
+    /*
+     * These members are constant and need no mutex protection.
+     */
+    struct _DriFenceMgr *mgr;
+    uint32_t fence_class;
+    uint32_t fence_type;
+
+    /*
+     * These members are protected by mgr->mutex.
+     */
+    drmMMListHead head;
+    int refCount;
+
+    /*
+     * These members are protected by this->mutex.
+     */
+    pipe_mutex mutex;
+    uint32_t signaled_type;
+    void *private;
+};
+
+uint32_t
+driFenceType(struct _DriFenceObject *fence)
+{
+  return fence->fence_type;
+}
+
+struct _DriFenceMgr *
+driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info)
+{
+  struct _DriFenceMgr *tmp;
+  uint32_t i;
+
+  tmp = calloc(1, sizeof(*tmp));
+  if (!tmp)
+      return NULL;
+
+  pipe_mutex_init(tmp->mutex);
+  pipe_mutex_lock(tmp->mutex);
+  tmp->refCount = 1;
+  tmp->info = *info;
+  tmp->num_fences = 0;
+  tmp->heads = calloc(tmp->info.num_classes, sizeof(*tmp->heads));
+  if (!tmp->heads)
+      goto out_err;
+
+  for (i=0; i<tmp->info.num_classes; ++i) {
+      DRMINITLISTHEAD(&tmp->heads[i]);
+  }
+  pipe_mutex_unlock(tmp->mutex);
+  return tmp;
+
+  out_err:
+  if (tmp)
+      free(tmp);
+  return NULL;
+}
+
+static void
+driFenceMgrUnrefUnlock(struct _DriFenceMgr **pMgr)
+{
+    struct _DriFenceMgr *mgr = *pMgr;
+
+    *pMgr = NULL;
+    if (--mgr->refCount == 0)
+	free(mgr);
+    else
+	pipe_mutex_unlock(mgr->mutex);
+}
+
+void
+driFenceMgrUnReference(struct _DriFenceMgr **pMgr)
+{
+    pipe_mutex_lock((*pMgr)->mutex);
+    driFenceMgrUnrefUnlock(pMgr);
+}
+
+static void
+driFenceUnReferenceLocked(struct _DriFenceObject **pFence)
+{
+    struct _DriFenceObject *fence = *pFence;
+    struct _DriFenceMgr *mgr = fence->mgr;
+
+    *pFence = NULL;
+    if (--fence->refCount == 0) {
+	DRMLISTDELINIT(&fence->head);
+	if (fence->private)
+	    mgr->info.unreference(mgr, &fence->private);
+    --mgr->num_fences;
+	fence->mgr = NULL;
+	--mgr->refCount;
+	free(fence);
+
+    }
+}
+
+
+static void
+driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr,
+			      drmMMListHead *list,
+			      uint32_t fence_class,
+			      uint32_t fence_type)
+{
+    struct _DriFenceObject *entry;
+    drmMMListHead *prev;
+
+    while(list != &mgr->heads[fence_class]) {
+	entry = DRMLISTENTRY(struct _DriFenceObject, list, head);
+
+	/*
+	 * Up refcount so that entry doesn't disappear from under us
+	 * when we unlock-relock mgr to get the correct locking order.
+	 */
+
+	++entry->refCount;
+	pipe_mutex_unlock(mgr->mutex);
+	pipe_mutex_lock(entry->mutex);
+	pipe_mutex_lock(mgr->mutex);
+
+	prev = list->prev;
+
+
+
+	if (list->prev == list) {
+
+		/*
+		 * Somebody else removed the entry from the list.
+		 */
+
+		pipe_mutex_unlock(entry->mutex);
+		driFenceUnReferenceLocked(&entry);
+		return;
+	}
+
+	entry->signaled_type |= (fence_type & entry->fence_type);
+	if (entry->signaled_type == entry->fence_type) {
+	    DRMLISTDELINIT(list);
+	    mgr->info.unreference(mgr, &entry->private);
+	}
+	pipe_mutex_unlock(entry->mutex);
+	driFenceUnReferenceLocked(&entry);
+	list = prev;
+    }
+}
+
+
+int
+driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type,
+	       int lazy_hint)
+{
+    struct _DriFenceMgr *mgr = fence->mgr;
+    int ret = 0;
+
+    pipe_mutex_lock(fence->mutex);
+
+    if ((fence->signaled_type & fence_type) == fence_type)
+	goto out0;
+
+    ret = mgr->info.finish(mgr, fence->private, fence_type, lazy_hint);
+    if (ret)
+	goto out0;
+
+    pipe_mutex_lock(mgr->mutex);
+    pipe_mutex_unlock(fence->mutex);
+
+    driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class,
+				  fence_type);
+    pipe_mutex_unlock(mgr->mutex);
+    return 0;
+
+  out0:
+    pipe_mutex_unlock(fence->mutex);
+    return ret;
+}
+
+uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence)
+{
+    uint32_t ret;
+
+    pipe_mutex_lock(fence->mutex);
+    ret = fence->signaled_type;
+    pipe_mutex_unlock(fence->mutex);
+
+    return ret;
+}
+
+int
+driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type,
+		 uint32_t *signaled)
+{
+    int ret = 0;
+    struct _DriFenceMgr *mgr;
+
+    pipe_mutex_lock(fence->mutex);
+    mgr = fence->mgr;
+    *signaled = fence->signaled_type;
+    if ((fence->signaled_type & flush_type) == flush_type)
+	goto out0;
+
+    ret = mgr->info.signaled(mgr, fence->private, flush_type, signaled);
+    if (ret) {
+	*signaled = fence->signaled_type;
+	goto out0;
+    }
+
+    if ((fence->signaled_type | *signaled) == fence->signaled_type)
+	goto out0;
+
+    pipe_mutex_lock(mgr->mutex);
+    pipe_mutex_unlock(fence->mutex);
+
+    driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class,
+				  *signaled);
+
+    pipe_mutex_unlock(mgr->mutex);
+    return 0;
+  out0:
+    pipe_mutex_unlock(fence->mutex);
+    return ret;
+}
+
+struct _DriFenceObject *
+driFenceReference(struct _DriFenceObject *fence)
+{
+    pipe_mutex_lock(fence->mgr->mutex);
+    ++fence->refCount;
+    pipe_mutex_unlock(fence->mgr->mutex);
+    return fence;
+}
+
+void
+driFenceUnReference(struct _DriFenceObject **pFence)
+{
+    struct _DriFenceMgr *mgr;
+
+    if (*pFence == NULL)
+	return;
+
+    mgr = (*pFence)->mgr;
+    pipe_mutex_lock(mgr->mutex);
+    ++mgr->refCount;
+    driFenceUnReferenceLocked(pFence);
+    driFenceMgrUnrefUnlock(&mgr);
+}
+
+struct _DriFenceObject
+*driFenceCreate(struct _DriFenceMgr *mgr, uint32_t fence_class,
+		uint32_t fence_type, void *private, size_t private_size)
+{
+    struct _DriFenceObject *fence;
+    size_t fence_size = sizeof(*fence);
+
+    if (private_size)
+	fence_size = ((fence_size + 15) & ~15);
+
+    fence = calloc(1, fence_size + private_size);
+
+    if (!fence) {
+	int ret = mgr->info.finish(mgr, private, fence_type, 0);
+
+	if (ret)
+	    usleep(10000000);
+
+	return NULL;
+    }
+
+    pipe_mutex_init(fence->mutex);
+    pipe_mutex_lock(fence->mutex);
+    pipe_mutex_lock(mgr->mutex);
+    fence->refCount = 1;
+    DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]);
+    fence->mgr = mgr;
+    ++mgr->refCount;
+    ++mgr->num_fences;
+    pipe_mutex_unlock(mgr->mutex);
+    fence->fence_class = fence_class;
+    fence->fence_type = fence_type;
+    fence->signaled_type = 0;
+    fence->private = private;
+    if (private_size) {
+        fence->private = (void *)(((uint8_t *) fence) + fence_size);
+	memcpy(fence->private, private, private_size);
+    }
+
+    pipe_mutex_unlock(fence->mutex);
+    return fence;
+}
+
+
+static int
+tSignaled(struct _DriFenceMgr *mgr, void *private, uint32_t flush_type,
+	  uint32_t *signaled_type)
+{
+  long fd = (long) mgr->private;
+  int dummy;
+  drmFence *fence = (drmFence *) private;
+  int ret;
+
+  *signaled_type = 0;
+  ret = drmFenceSignaled((int) fd, fence, flush_type, &dummy);
+  if (ret)
+    return ret;
+
+  *signaled_type = fence->signaled;
+
+  return 0;
+}
+
+static int
+tFinish(struct _DriFenceMgr *mgr, void *private, uint32_t fence_type,
+	int lazy_hint)
+{
+  long fd = (long) mgr->private;
+  unsigned flags = lazy_hint ? DRM_FENCE_FLAG_WAIT_LAZY : 0;
+
+  return drmFenceWait((int)fd, flags, (drmFence *) private, fence_type);
+}
+
+static int
+tUnref(struct _DriFenceMgr *mgr, void **private)
+{
+  long fd = (long) mgr->private;
+  drmFence *fence = (drmFence *) *private;
+  *private = NULL;
+
+  return drmFenceUnreference(fd, fence);
+}
+
+struct _DriFenceMgr *driFenceMgrTTMInit(int fd)
+{
+  struct _DriFenceMgrCreateInfo info;
+  struct _DriFenceMgr *mgr;
+
+  info.flags = DRI_FENCE_CLASS_ORDERED;
+  info.num_classes = 4;
+  info.signaled = tSignaled;
+  info.finish = tFinish;
+  info.unreference = tUnref;
+
+  mgr = driFenceMgrCreate(&info);
+  if (mgr == NULL)
+    return NULL;
+
+  mgr->private = (void *) (long) fd;
+  return mgr;
+}
+
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.h b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.h
new file mode 100644
index 0000000000..4ea58dfe18
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.h
@@ -0,0 +1,115 @@
+#ifndef DRI_FENCEMGR_H
+#define DRI_FENCEMGR_H
+
+#include <stdint.h>
+#include <stdlib.h>
+
+struct _DriFenceObject;
+struct _DriFenceMgr;
+
+/*
+ * Do a quick check to see if the fence manager has registered the fence
+ * object as signaled. Note that this function may return a false negative
+ * answer.
+ */
+extern uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence);
+
+/*
+ * Check if the fence object is signaled. This function can be substantially
+ * more expensive to call than the above function, but will not return a false
+ * negative answer. The argument "flush_type" sets the types that the
+ * underlying mechanism must make sure will eventually signal.
+ */
+extern int driFenceSignaledType(struct _DriFenceObject *fence,
+				uint32_t flush_type, uint32_t *signaled);
+
+/*
+ * Convenience functions.
+ */
+
+static inline int driFenceSignaled(struct _DriFenceObject *fence,
+				   uint32_t flush_type)
+{
+    uint32_t signaled_types;
+    int ret = driFenceSignaledType(fence, flush_type, &signaled_types);
+    if (ret)
+	return 0;
+    return ((signaled_types & flush_type) == flush_type);
+}
+
+static inline int driFenceSignaledCached(struct _DriFenceObject *fence,
+					 uint32_t flush_type)
+{
+    uint32_t signaled_types =
+	driFenceSignaledTypeCached(fence);
+
+    return ((signaled_types & flush_type) == flush_type);
+}
+
+/*
+ * Reference a fence object.
+ */
+extern struct _DriFenceObject *driFenceReference(struct _DriFenceObject *fence);
+
+/*
+ * Unreference a fence object. The fence object pointer will be reset to NULL.
+ */
+
+extern void driFenceUnReference(struct _DriFenceObject **pFence);
+
+
+/*
+ * Wait for a fence to signal the indicated fence_type.
+ * If "lazy_hint" is true, it indicates that the wait may sleep to avoid
+ * busy-wait polling.
+ */
+extern int driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type,
+			  int lazy_hint);
+
+/*
+ * Create a DriFenceObject for manager "mgr".
+ *
+ * "private" is a pointer that should be used for the callbacks in
+ * struct _DriFenceMgrCreateInfo.
+ *
+ * if private_size is nonzero, then the info stored at *private, with size
+ * private size will be copied and the fence manager will instead use a
+ * pointer to the copied data for the callbacks in
+ * struct _DriFenceMgrCreateInfo. In that case, the object pointed to by
+ * "private" may be destroyed after the call to driFenceCreate.
+ */
+extern struct _DriFenceObject *driFenceCreate(struct _DriFenceMgr *mgr,
+					      uint32_t fence_class,
+					      uint32_t fence_type,
+					      void *private,
+					      size_t private_size);
+
+extern uint32_t driFenceType(struct _DriFenceObject *fence);
+
+/*
+ * Fence creations are ordered. If a fence signals a fence_type,
+ * it is safe to assume that all fences of the same class that was
+ * created before that fence has signaled the same type.
+ */
+
+#define DRI_FENCE_CLASS_ORDERED (1 << 0)
+
+struct _DriFenceMgrCreateInfo {
+    uint32_t flags;
+    uint32_t num_classes;
+    int (*signaled) (struct _DriFenceMgr *mgr, void *private, uint32_t flush_type,
+		     uint32_t *signaled_type);
+    int (*finish) (struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, int lazy_hint);
+    int (*unreference) (struct _DriFenceMgr *mgr, void **private);
+};
+
+extern struct _DriFenceMgr *
+driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info);
+
+void
+driFenceMgrUnReference(struct _DriFenceMgr **pMgr);
+
+extern struct _DriFenceMgr *
+driFenceMgrTTMInit(int fd);
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c
new file mode 100644
index 0000000000..60924eac9e
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c
@@ -0,0 +1,161 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, TX., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+/*
+ * Authors: Thomas Hellström <thomas-at-tungstengraphics-dot-com>
+ */
+
+#include <xf86drm.h>
+#include <stdlib.h>
+#include <errno.h>
+#include "pipe/p_debug.h"
+#include "pipe/p_thread.h"
+#include "ws_dri_bufpool.h"
+#include "ws_dri_bufmgr.h"
+
+static void *
+pool_create(struct _DriBufferPool *pool,
+            unsigned long size, uint64_t flags, unsigned hint,
+            unsigned alignment)
+{
+    unsigned long *private = malloc(size + 2*sizeof(unsigned long));
+    if ((flags & DRM_BO_MASK_MEM) != DRM_BO_FLAG_MEM_LOCAL)
+      abort();
+
+    *private = size;
+    return (void *)private;
+}
+
+
+static int
+pool_destroy(struct _DriBufferPool *pool, void *private)
+{
+    free(private);
+    return 0;
+}
+
+static int
+pool_waitIdle(struct _DriBufferPool *pool, void *private,
+	      pipe_mutex *mutex, int lazy)
+{
+    return 0;
+}
+
+static int
+pool_map(struct _DriBufferPool *pool, void *private, unsigned flags,
+         int hint, pipe_mutex *mutex, void **virtual)
+{
+    *virtual = (void *)((unsigned long *)private + 2);
+    return 0;
+}
+
+static int
+pool_unmap(struct _DriBufferPool *pool, void *private)
+{
+    return 0;
+}
+
+static unsigned long
+pool_offset(struct _DriBufferPool *pool, void *private)
+{
+    /*
+     * BUG
+     */
+    abort();
+    return 0UL;
+}
+
+static unsigned long
+pool_poolOffset(struct _DriBufferPool *pool, void *private)
+{
+    /*
+     * BUG
+     */
+    abort();
+}
+
+static uint64_t
+pool_flags(struct _DriBufferPool *pool, void *private)
+{
+    return DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED;
+}
+
+static unsigned long
+pool_size(struct _DriBufferPool *pool, void *private)
+{
+    return *(unsigned long *) private;
+}
+
+
+static int
+pool_fence(struct _DriBufferPool *pool, void *private,
+           struct _DriFenceObject *fence)
+{
+    abort();
+    return 0UL;
+}
+
+static drmBO *
+pool_kernel(struct _DriBufferPool *pool, void *private)
+{
+    abort();
+    return NULL;
+}
+
+static void
+pool_takedown(struct _DriBufferPool *pool)
+{
+    free(pool);
+}
+
+
+struct _DriBufferPool *
+driMallocPoolInit(void)
+{
+   struct _DriBufferPool *pool;
+
+   pool = (struct _DriBufferPool *) malloc(sizeof(*pool));
+   if (!pool)
+       return NULL;
+
+   pool->data = NULL;
+   pool->fd = -1;
+   pool->map = &pool_map;
+   pool->unmap = &pool_unmap;
+   pool->destroy = &pool_destroy;
+   pool->offset = &pool_offset;
+   pool->poolOffset = &pool_poolOffset;
+   pool->flags = &pool_flags;
+   pool->size = &pool_size;
+   pool->create = &pool_create;
+   pool->fence = &pool_fence;
+   pool->kernel = &pool_kernel;
+   pool->validate = NULL;
+   pool->waitIdle = &pool_waitIdle;
+   pool->takeDown = &pool_takedown;
+   return pool;
+}
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c
new file mode 100644
index 0000000000..391cea50a7
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c
@@ -0,0 +1,968 @@
+/**************************************************************************
+ *
+ * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+/*
+ * Authors: Thomas Hellstrom <thomas-at-tungstengraphics-dot-com>
+ */
+
+#include <stdint.h>
+#include <sys/time.h>
+#include <errno.h>
+#include <unistd.h>
+#include <assert.h>
+#include "ws_dri_bufpool.h"
+#include "ws_dri_fencemgr.h"
+#include "ws_dri_bufmgr.h"
+#include "pipe/p_thread.h"
+
+#define DRI_SLABPOOL_ALLOC_RETRIES 100
+
+struct _DriSlab;
+
+struct _DriSlabBuffer {
+    int isSlabBuffer;
+    drmBO *bo;
+    struct _DriFenceObject *fence;
+    struct _DriSlab *parent;
+    drmMMListHead head;
+    uint32_t mapCount;
+    uint32_t start;
+    uint32_t fenceType;
+    int unFenced;
+    pipe_condvar event;
+};
+
+struct _DriKernelBO {
+    int fd;
+    drmBO bo;
+    drmMMListHead timeoutHead;
+    drmMMListHead head;
+    struct timeval timeFreed;
+    uint32_t pageAlignment;
+    void *virtual;
+};
+
+struct _DriSlab{
+    drmMMListHead head;
+    drmMMListHead freeBuffers;
+    uint32_t numBuffers;
+    uint32_t numFree;
+    struct _DriSlabBuffer *buffers;
+    struct _DriSlabSizeHeader *header;
+    struct _DriKernelBO *kbo;
+};
+
+
+struct _DriSlabSizeHeader {
+    drmMMListHead slabs;
+    drmMMListHead freeSlabs;
+    drmMMListHead delayedBuffers;
+    uint32_t numDelayed;
+    struct _DriSlabPool *slabPool;
+    uint32_t bufSize;
+    pipe_mutex mutex;
+};
+
+struct _DriFreeSlabManager {
+    struct timeval slabTimeout;
+    struct timeval checkInterval;
+    struct timeval nextCheck;
+    drmMMListHead timeoutList;
+    drmMMListHead unCached;
+    drmMMListHead cached;
+    pipe_mutex mutex;
+};
+
+
+struct _DriSlabPool {
+
+    /*
+     * The data of this structure remains constant after
+     * initialization and thus needs no mutex protection.
+     */
+
+    struct _DriFreeSlabManager *fMan;
+    uint64_t proposedFlags;
+    uint64_t validMask;
+    uint32_t *bucketSizes;
+    uint32_t numBuckets;
+    uint32_t pageSize;
+    int fd;
+    int pageAlignment;
+    int maxSlabSize;
+    int desiredNumBuffers;
+    struct _DriSlabSizeHeader *headers;
+};
+
+/*
+ * FIXME: Perhaps arrange timeout slabs in size buckets for fast
+ * retreival??
+ */
+
+
+static inline int
+driTimeAfterEq(struct timeval *arg1, struct timeval *arg2)
+{
+    return ((arg1->tv_sec > arg2->tv_sec) ||
+	    ((arg1->tv_sec == arg2->tv_sec) &&
+	     (arg1->tv_usec > arg2->tv_usec)));
+}
+
+static inline void
+driTimeAdd(struct timeval *arg, struct timeval *add)
+{
+    unsigned int sec;
+
+    arg->tv_sec += add->tv_sec;
+    arg->tv_usec += add->tv_usec;
+    sec = arg->tv_usec / 1000000;
+    arg->tv_sec += sec;
+    arg->tv_usec -= sec*1000000;
+}
+
+static void
+driFreeKernelBO(struct _DriKernelBO *kbo)
+{
+    if (!kbo)
+	return;
+
+    (void) drmBOUnreference(kbo->fd, &kbo->bo);
+    free(kbo);
+}
+
+
+static void
+driFreeTimeoutKBOsLocked(struct _DriFreeSlabManager *fMan,
+			 struct timeval *time)
+{
+    drmMMListHead *list, *next;
+    struct _DriKernelBO *kbo;
+
+    if (!driTimeAfterEq(time, &fMan->nextCheck))
+	return;
+
+    for (list = fMan->timeoutList.next, next = list->next;
+	 list != &fMan->timeoutList;
+	 list = next, next = list->next) {
+
+	kbo = DRMLISTENTRY(struct _DriKernelBO, list, timeoutHead);
+
+	if (!driTimeAfterEq(time, &kbo->timeFreed))
+	    break;
+
+	DRMLISTDELINIT(&kbo->timeoutHead);
+	DRMLISTDELINIT(&kbo->head);
+	driFreeKernelBO(kbo);
+    }
+
+    fMan->nextCheck = *time;
+    driTimeAdd(&fMan->nextCheck, &fMan->checkInterval);
+}
+
+
+/*
+ * Add a _DriKernelBO to the free slab manager.
+ * This means that it is available for reuse, but if it's not
+ * reused in a while, it will be freed.
+ */
+
+static void
+driSetKernelBOFree(struct _DriFreeSlabManager *fMan,
+		   struct _DriKernelBO *kbo)
+{
+    struct timeval time;
+
+    pipe_mutex_lock(fMan->mutex);
+    gettimeofday(&time, NULL);
+    driTimeAdd(&time, &fMan->slabTimeout);
+
+    kbo->timeFreed = time;
+
+    if (kbo->bo.flags & DRM_BO_FLAG_CACHED)
+	DRMLISTADD(&kbo->head, &fMan->cached);
+    else
+	DRMLISTADD(&kbo->head, &fMan->unCached);
+
+    DRMLISTADDTAIL(&kbo->timeoutHead, &fMan->timeoutList);
+    driFreeTimeoutKBOsLocked(fMan, &time);
+
+    pipe_mutex_unlock(fMan->mutex);
+}
+
+/*
+ * Get a _DriKernelBO for us to use as storage for a slab.
+ *
+ */
+
+static struct _DriKernelBO *
+driAllocKernelBO(struct _DriSlabSizeHeader *header)
+
+{
+    struct _DriSlabPool *slabPool = header->slabPool;
+    struct _DriFreeSlabManager *fMan = slabPool->fMan;
+    drmMMListHead *list, *next, *head;
+    uint32_t size = header->bufSize * slabPool->desiredNumBuffers;
+    struct _DriKernelBO *kbo;
+    struct _DriKernelBO *kboTmp;
+    int ret;
+
+    /*
+     * FIXME: We should perhaps allow some variation in slabsize in order
+     * to efficiently reuse slabs.
+     */
+
+    size = (size <= slabPool->maxSlabSize) ? size : slabPool->maxSlabSize;
+    size = (size + slabPool->pageSize - 1) & ~(slabPool->pageSize - 1);
+    pipe_mutex_lock(fMan->mutex);
+
+    kbo = NULL;
+
+  retry:
+    head = (slabPool->proposedFlags & DRM_BO_FLAG_CACHED) ?
+	&fMan->cached : &fMan->unCached;
+
+    for (list = head->next, next = list->next;
+	 list != head;
+	 list = next, next = list->next) {
+
+	kboTmp = DRMLISTENTRY(struct _DriKernelBO, list, head);
+
+	if ((kboTmp->bo.size == size) &&
+	    (slabPool->pageAlignment == 0 ||
+	     (kboTmp->pageAlignment % slabPool->pageAlignment) == 0)) {
+
+	    if (!kbo)
+		kbo = kboTmp;
+
+	    if ((kbo->bo.proposedFlags ^ slabPool->proposedFlags) == 0)
+		break;
+
+	}
+    }
+
+    if (kbo) {
+	DRMLISTDELINIT(&kbo->head);
+	DRMLISTDELINIT(&kbo->timeoutHead);
+    }
+
+    pipe_mutex_unlock(fMan->mutex);
+
+    if (kbo) {
+        uint64_t new_mask = kbo->bo.proposedFlags ^ slabPool->proposedFlags;
+
+	ret = 0;
+	if (new_mask) {
+	    ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags,
+				 new_mask, DRM_BO_HINT_DONT_FENCE, 0, 0);
+	}
+	if (ret == 0)
+	    return kbo;
+
+	driFreeKernelBO(kbo);
+	kbo = NULL;
+	goto retry;
+    }
+
+    kbo = calloc(1, sizeof(struct _DriKernelBO));
+    if (!kbo)
+	return NULL;
+
+    kbo->fd = slabPool->fd;
+    DRMINITLISTHEAD(&kbo->head);
+    DRMINITLISTHEAD(&kbo->timeoutHead);
+    ret = drmBOCreate(kbo->fd, size, slabPool->pageAlignment, NULL,
+		      slabPool->proposedFlags,
+		      DRM_BO_HINT_DONT_FENCE, &kbo->bo);
+    if (ret)
+	goto out_err0;
+
+    ret = drmBOMap(kbo->fd, &kbo->bo,
+		   DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE,
+		   0, &kbo->virtual);
+
+    if (ret)
+	goto out_err1;
+
+    ret = drmBOUnmap(kbo->fd, &kbo->bo);
+    if (ret)
+	goto out_err1;
+
+    return kbo;
+
+  out_err1:
+    drmBOUnreference(kbo->fd, &kbo->bo);
+  out_err0:
+    free(kbo);
+    return NULL;
+}
+
+
+static int
+driAllocSlab(struct _DriSlabSizeHeader *header)
+{
+    struct _DriSlab *slab;
+    struct _DriSlabBuffer *buf;
+    uint32_t numBuffers;
+    int ret;
+    int i;
+
+    slab = calloc(1, sizeof(*slab));
+    if (!slab)
+	return -ENOMEM;
+
+    slab->kbo = driAllocKernelBO(header);
+    if (!slab->kbo) {
+	ret = -ENOMEM;
+	goto out_err0;
+    }
+
+    numBuffers = slab->kbo->bo.size / header->bufSize;
+
+    slab->buffers = calloc(numBuffers, sizeof(*slab->buffers));
+    if (!slab->buffers) {
+	ret = -ENOMEM;
+	goto out_err1;
+    }
+
+    DRMINITLISTHEAD(&slab->head);
+    DRMINITLISTHEAD(&slab->freeBuffers);
+    slab->numBuffers = numBuffers;
+    slab->numFree = 0;
+    slab->header = header;
+
+    buf = slab->buffers;
+    for (i=0; i < numBuffers; ++i) {
+	buf->parent = slab;
+	buf->start = i* header->bufSize;
+	buf->mapCount = 0;
+	buf->isSlabBuffer = 1;
+	pipe_condvar_init(buf->event);
+	DRMLISTADDTAIL(&buf->head, &slab->freeBuffers);
+	slab->numFree++;
+	buf++;
+    }
+
+    DRMLISTADDTAIL(&slab->head, &header->slabs);
+
+    return 0;
+
+  out_err1:
+    driSetKernelBOFree(header->slabPool->fMan, slab->kbo);
+    free(slab->buffers);
+  out_err0:
+    free(slab);
+    return ret;
+}
+
+/*
+ * Delete a buffer from the slab header delayed list and put
+ * it on the slab free list.
+ */
+
+static void
+driSlabFreeBufferLocked(struct _DriSlabBuffer *buf)
+{
+    struct _DriSlab *slab = buf->parent;
+    struct _DriSlabSizeHeader *header = slab->header;
+    drmMMListHead *list = &buf->head;
+
+    DRMLISTDEL(list);
+    DRMLISTADDTAIL(list, &slab->freeBuffers);
+    slab->numFree++;
+
+    if (slab->head.next == &slab->head)
+	DRMLISTADDTAIL(&slab->head, &header->slabs);
+
+    if (slab->numFree == slab->numBuffers) {
+	list = &slab->head;
+	DRMLISTDEL(list);
+	DRMLISTADDTAIL(list, &header->freeSlabs);
+    }
+
+    if (header->slabs.next == &header->slabs ||
+	slab->numFree != slab->numBuffers) {
+
+	drmMMListHead *next;
+	struct _DriFreeSlabManager *fMan = header->slabPool->fMan;
+
+	for (list = header->freeSlabs.next, next = list->next;
+	     list != &header->freeSlabs;
+	     list = next, next = list->next) {
+
+	    slab = DRMLISTENTRY(struct _DriSlab, list, head);
+
+	    DRMLISTDELINIT(list);
+	    driSetKernelBOFree(fMan, slab->kbo);
+	    free(slab->buffers);
+	    free(slab);
+	}
+    }
+}
+
+static void
+driSlabCheckFreeLocked(struct _DriSlabSizeHeader *header, int wait)
+{
+  drmMMListHead *list, *prev, *first;
+   struct _DriSlabBuffer *buf;
+   struct _DriSlab *slab;
+   int firstWasSignaled = 1;
+   int signaled;
+   int i;
+   int ret;
+
+   /*
+    * Rerun the freeing test if the youngest tested buffer
+    * was signaled, since there might be more idle buffers
+    * in the delay list.
+    */
+
+   while (firstWasSignaled) {
+       firstWasSignaled = 0;
+       signaled = 0;
+       first = header->delayedBuffers.next;
+
+       /* Only examine the oldest 1/3 of delayed buffers:
+	*/
+       if (header->numDelayed > 3) {
+	   for (i = 0; i < header->numDelayed; i += 3) {
+	       first = first->next;
+	   }
+       }
+
+       for (list = first, prev = list->prev;
+	    list != &header->delayedBuffers;
+	    list = prev, prev = list->prev) {
+	   buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head);
+	   slab = buf->parent;
+
+	   if (!signaled) {
+	       if (wait) {
+		   ret = driFenceFinish(buf->fence, buf->fenceType, 0);
+		   if (ret)
+		       break;
+		   signaled = 1;
+		   wait = 0;
+	       } else {
+		   signaled = driFenceSignaled(buf->fence, buf->fenceType);
+	       }
+	       if (signaled) {
+		   if (list == first)
+		       firstWasSignaled = 1;
+		   driFenceUnReference(&buf->fence);
+		   header->numDelayed--;
+		   driSlabFreeBufferLocked(buf);
+	       }
+	   } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) {
+	       driFenceUnReference(&buf->fence);
+	       header->numDelayed--;
+	       driSlabFreeBufferLocked(buf);
+	   }
+       }
+   }
+}
+
+
+static struct _DriSlabBuffer *
+driSlabAllocBuffer(struct _DriSlabSizeHeader *header)
+{
+    static struct _DriSlabBuffer *buf;
+    struct _DriSlab *slab;
+    drmMMListHead *list;
+    int count = DRI_SLABPOOL_ALLOC_RETRIES;
+
+    pipe_mutex_lock(header->mutex);
+    while(header->slabs.next == &header->slabs && count > 0) {
+        driSlabCheckFreeLocked(header, 0);
+	if (header->slabs.next != &header->slabs)
+	  break;
+
+	pipe_mutex_unlock(header->mutex);
+	if (count != DRI_SLABPOOL_ALLOC_RETRIES)
+	    usleep(1);
+	pipe_mutex_lock(header->mutex);
+	(void) driAllocSlab(header);
+	count--;
+    }
+
+    list = header->slabs.next;
+    if (list == &header->slabs) {
+	pipe_mutex_unlock(header->mutex);
+	return NULL;
+    }
+    slab = DRMLISTENTRY(struct _DriSlab, list, head);
+    if (--slab->numFree == 0)
+	DRMLISTDELINIT(list);
+
+    list = slab->freeBuffers.next;
+    DRMLISTDELINIT(list);
+
+    pipe_mutex_unlock(header->mutex);
+    buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head);
+    return buf;
+}
+
+static void *
+pool_create(struct _DriBufferPool *driPool, unsigned long size,
+	    uint64_t flags, unsigned hint, unsigned alignment)
+{
+    struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data;
+    struct _DriSlabSizeHeader *header;
+    struct _DriSlabBuffer *buf;
+    void *dummy;
+    int i;
+    int ret;
+
+    /*
+     * FIXME: Check for compatibility.
+     */
+
+    header = pool->headers;
+    for (i=0; i<pool->numBuckets; ++i) {
+      if (header->bufSize >= size)
+	break;
+      header++;
+    }
+
+    if (i < pool->numBuckets)
+	return driSlabAllocBuffer(header);
+
+
+    /*
+     * Fall back to allocate a buffer object directly from DRM.
+     * and wrap it in a driBO structure.
+     */
+
+
+    buf = calloc(1, sizeof(*buf));
+
+    if (!buf)
+	return NULL;
+
+    buf->bo = calloc(1, sizeof(*buf->bo));
+    if (!buf->bo)
+	goto out_err0;
+
+    if (alignment) {
+	if ((alignment < pool->pageSize) && (pool->pageSize % alignment))
+	    goto out_err1;
+	if ((alignment > pool->pageSize) && (alignment % pool->pageSize))
+	    goto out_err1;
+    }
+
+    ret = drmBOCreate(pool->fd, size, alignment / pool->pageSize, NULL,
+			flags, hint, buf->bo);
+    if (ret)
+	goto out_err1;
+
+    ret  = drmBOMap(pool->fd, buf->bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE,
+		    0, &dummy);
+    if (ret)
+	goto out_err2;
+
+    ret = drmBOUnmap(pool->fd, buf->bo);
+    if (ret)
+	goto out_err2;
+
+    return buf;
+  out_err2:
+    drmBOUnreference(pool->fd, buf->bo);
+  out_err1:
+    free(buf->bo);
+  out_err0:
+    free(buf);
+    return NULL;
+}
+
+static int
+pool_destroy(struct _DriBufferPool *driPool, void *private)
+{
+    struct _DriSlabBuffer *buf =
+	(struct _DriSlabBuffer *) private;
+    struct _DriSlab *slab;
+    struct _DriSlabSizeHeader *header;
+
+    if (!buf->isSlabBuffer) {
+	struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data;
+	int ret;
+
+	ret = drmBOUnreference(pool->fd, buf->bo);
+	free(buf->bo);
+	free(buf);
+	return ret;
+    }
+
+    slab = buf->parent;
+    header = slab->header;
+
+    pipe_mutex_lock(header->mutex);
+    buf->unFenced = 0;
+    buf->mapCount = 0;
+
+    if (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)) {
+	DRMLISTADDTAIL(&buf->head, &header->delayedBuffers);
+	header->numDelayed++;
+    } else {
+	if (buf->fence)
+	    driFenceUnReference(&buf->fence);
+	driSlabFreeBufferLocked(buf);
+    }
+
+    pipe_mutex_unlock(header->mutex);
+    return 0;
+}
+
+static int
+pool_waitIdle(struct _DriBufferPool *driPool, void *private,
+	      pipe_mutex *mutex, int lazy)
+{
+   struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
+
+   while(buf->unFenced)
+       pipe_condvar_wait(buf->event, *mutex);
+
+   if (!buf->fence)
+     return 0;
+
+   driFenceFinish(buf->fence, buf->fenceType, lazy);
+   driFenceUnReference(&buf->fence);
+
+   return 0;
+}
+
+static int
+pool_map(struct _DriBufferPool *pool, void *private, unsigned flags,
+         int hint, pipe_mutex *mutex, void **virtual)
+{
+   struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
+   int busy;
+
+   if (buf->isSlabBuffer)
+       busy = buf->unFenced || (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType));
+   else
+       busy = buf->fence && !driFenceSignaled(buf->fence, buf->fenceType);
+
+
+   if (busy) {
+       if (hint & DRM_BO_HINT_DONT_BLOCK)
+	   return -EBUSY;
+       else {
+	   (void) pool_waitIdle(pool, private, mutex, 0);
+       }
+   }
+
+   ++buf->mapCount;
+   *virtual = (buf->isSlabBuffer) ?
+       (void *) ((uint8_t *) buf->parent->kbo->virtual + buf->start) :
+       (void *) buf->bo->virtual;
+
+   return 0;
+}
+
+static int
+pool_unmap(struct _DriBufferPool *pool, void *private)
+{
+   struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
+
+   --buf->mapCount;
+   if (buf->mapCount == 0 && buf->isSlabBuffer)
+      pipe_condvar_broadcast(buf->event);
+
+   return 0;
+}
+
+static unsigned long
+pool_offset(struct _DriBufferPool *pool, void *private)
+{
+   struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
+   struct _DriSlab *slab;
+   struct _DriSlabSizeHeader *header;
+
+   if (!buf->isSlabBuffer) {
+       assert(buf->bo->proposedFlags & DRM_BO_FLAG_NO_MOVE);
+       return buf->bo->offset;
+   }
+
+   slab = buf->parent;
+   header = slab->header;
+
+   (void) header;
+   assert(header->slabPool->proposedFlags & DRM_BO_FLAG_NO_MOVE);
+   return slab->kbo->bo.offset + buf->start;
+}
+
+static unsigned long
+pool_poolOffset(struct _DriBufferPool *pool, void *private)
+{
+   struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
+
+   return buf->start;
+}
+
+static uint64_t
+pool_flags(struct _DriBufferPool *pool, void *private)
+{
+   struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
+
+   if (!buf->isSlabBuffer)
+       return buf->bo->flags;
+
+   return buf->parent->kbo->bo.flags;
+}
+
+static unsigned long
+pool_size(struct _DriBufferPool *pool, void *private)
+{
+   struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
+   if (!buf->isSlabBuffer)
+       return buf->bo->size;
+
+   return buf->parent->header->bufSize;
+}
+
+static int
+pool_fence(struct _DriBufferPool *pool, void *private,
+           struct _DriFenceObject *fence)
+{
+   struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
+   drmBO *bo;
+
+   if (buf->fence)
+      driFenceUnReference(&buf->fence);
+
+   buf->fence = driFenceReference(fence);
+   bo = (buf->isSlabBuffer) ?
+     &buf->parent->kbo->bo:
+     buf->bo;
+   buf->fenceType = bo->fenceFlags;
+
+   buf->unFenced = 0;
+   pipe_condvar_broadcast(buf->event);
+
+   return 0;
+}
+
+static drmBO *
+pool_kernel(struct _DriBufferPool *pool, void *private)
+{
+   struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
+
+   return (buf->isSlabBuffer) ? &buf->parent->kbo->bo : buf->bo;
+}
+
+static int
+pool_validate(struct _DriBufferPool *pool, void *private,
+	      pipe_mutex *mutex)
+{
+   struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
+
+   if (!buf->isSlabBuffer)
+       return 0;
+
+   while(buf->mapCount != 0)
+      pipe_condvar_wait(buf->event, *mutex);
+
+   buf->unFenced = 1;
+   return 0;
+}
+
+
+struct _DriFreeSlabManager *
+driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec)
+{
+    struct _DriFreeSlabManager *tmp;
+
+    tmp = calloc(1, sizeof(*tmp));
+    if (!tmp)
+	return NULL;
+
+    pipe_mutex_init(tmp->mutex);
+    pipe_mutex_lock(tmp->mutex);
+    tmp->slabTimeout.tv_usec = slabTimeoutMsec*1000;
+    tmp->slabTimeout.tv_sec = tmp->slabTimeout.tv_usec / 1000000;
+    tmp->slabTimeout.tv_usec -=  tmp->slabTimeout.tv_sec*1000000;
+
+    tmp->checkInterval.tv_usec = checkIntervalMsec*1000;
+    tmp->checkInterval.tv_sec = tmp->checkInterval.tv_usec / 1000000;
+    tmp->checkInterval.tv_usec -=  tmp->checkInterval.tv_sec*1000000;
+
+    gettimeofday(&tmp->nextCheck, NULL);
+    driTimeAdd(&tmp->nextCheck, &tmp->checkInterval);
+    DRMINITLISTHEAD(&tmp->timeoutList);
+    DRMINITLISTHEAD(&tmp->unCached);
+    DRMINITLISTHEAD(&tmp->cached);
+    pipe_mutex_unlock(tmp->mutex);
+
+    return tmp;
+}
+
+void
+driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan)
+{
+    struct timeval time;
+
+    time = fMan->nextCheck;
+    driTimeAdd(&time, &fMan->checkInterval);
+
+    pipe_mutex_lock(fMan->mutex);
+    driFreeTimeoutKBOsLocked(fMan, &time);
+    pipe_mutex_unlock(fMan->mutex);
+
+    assert(fMan->timeoutList.next == &fMan->timeoutList);
+    assert(fMan->unCached.next == &fMan->unCached);
+    assert(fMan->cached.next == &fMan->cached);
+
+    free(fMan);
+}
+
+static void
+driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size,
+		  struct _DriSlabSizeHeader *header)
+{
+    pipe_mutex_init(header->mutex);
+    pipe_mutex_lock(header->mutex);
+
+    DRMINITLISTHEAD(&header->slabs);
+    DRMINITLISTHEAD(&header->freeSlabs);
+    DRMINITLISTHEAD(&header->delayedBuffers);
+
+    header->numDelayed = 0;
+    header->slabPool = pool;
+    header->bufSize = size;
+
+    pipe_mutex_unlock(header->mutex);
+}
+
+static void
+driFinishSizeHeader(struct _DriSlabSizeHeader *header)
+{
+    drmMMListHead *list, *next;
+    struct _DriSlabBuffer *buf;
+
+    pipe_mutex_lock(header->mutex);
+    for (list = header->delayedBuffers.next, next = list->next;
+	 list != &header->delayedBuffers;
+	 list = next, next = list->next) {
+
+	buf = DRMLISTENTRY(struct _DriSlabBuffer, list , head);
+	if (buf->fence) {
+	    (void) driFenceFinish(buf->fence, buf->fenceType, 0);
+	    driFenceUnReference(&buf->fence);
+	}
+	header->numDelayed--;
+	driSlabFreeBufferLocked(buf);
+    }
+    pipe_mutex_unlock(header->mutex);
+}
+
+static void
+pool_takedown(struct _DriBufferPool *driPool)
+{
+   struct _DriSlabPool *pool = driPool->data;
+   int i;
+
+   for (i=0; i<pool->numBuckets; ++i) {
+     driFinishSizeHeader(&pool->headers[i]);
+   }
+
+   free(pool->headers);
+   free(pool->bucketSizes);
+   free(pool);
+   free(driPool);
+}
+
+struct _DriBufferPool *
+driSlabPoolInit(int fd, uint64_t flags,
+		uint64_t validMask,
+		uint32_t smallestSize,
+		uint32_t numSizes,
+		uint32_t desiredNumBuffers,
+		uint32_t maxSlabSize,
+		uint32_t pageAlignment,
+		struct _DriFreeSlabManager *fMan)
+{
+    struct _DriBufferPool *driPool;
+    struct _DriSlabPool *pool;
+    uint32_t i;
+
+    driPool = calloc(1, sizeof(*driPool));
+    if (!driPool)
+	return NULL;
+
+    pool = calloc(1, sizeof(*pool));
+    if (!pool)
+	goto out_err0;
+
+    pool->bucketSizes = calloc(numSizes, sizeof(*pool->bucketSizes));
+    if (!pool->bucketSizes)
+	goto out_err1;
+
+    pool->headers = calloc(numSizes, sizeof(*pool->headers));
+    if (!pool->headers)
+	goto out_err2;
+
+    pool->fMan = fMan;
+    pool->proposedFlags = flags;
+    pool->validMask = validMask;
+    pool->numBuckets = numSizes;
+    pool->pageSize = getpagesize();
+    pool->fd = fd;
+    pool->pageAlignment = pageAlignment;
+    pool->maxSlabSize = maxSlabSize;
+    pool->desiredNumBuffers = desiredNumBuffers;
+
+    for (i=0; i<pool->numBuckets; ++i) {
+	pool->bucketSizes[i] = (smallestSize << i);
+	driInitSizeHeader(pool, pool->bucketSizes[i],
+			  &pool->headers[i]);
+    }
+
+    driPool->data = (void *) pool;
+    driPool->map = &pool_map;
+    driPool->unmap = &pool_unmap;
+    driPool->destroy = &pool_destroy;
+    driPool->offset = &pool_offset;
+    driPool->poolOffset = &pool_poolOffset;
+    driPool->flags = &pool_flags;
+    driPool->size = &pool_size;
+    driPool->create = &pool_create;
+    driPool->fence = &pool_fence;
+    driPool->kernel = &pool_kernel;
+    driPool->validate = &pool_validate;
+    driPool->waitIdle = &pool_waitIdle;
+    driPool->takeDown = &pool_takedown;
+
+    return driPool;
+
+  out_err2:
+    free(pool->bucketSizes);
+  out_err1:
+    free(pool);
+  out_err0:
+    free(driPool);
+
+    return NULL;
+}
diff --git a/src/gallium/winsys/drm/intel/dri/Makefile b/src/gallium/winsys/drm/intel/dri/Makefile
new file mode 100644
index 0000000000..2046441a22
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/Makefile
@@ -0,0 +1,33 @@
+TOP = ../../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = i915_dri.so
+LIBNAME_EGL = egl_i915_dri.so
+
+PIPE_DRIVERS = \
+	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	../common/libinteldrm.a \
+	$(TOP)/src/gallium/drivers/i915simple/libi915simple.a
+
+
+DRIVER_SOURCES = \
+	intel_winsys_softpipe.c \
+	intel_swapbuffers.c \
+	intel_context.c \
+	intel_lock.c \
+	intel_screen.c
+
+C_SOURCES = \
+	$(COMMON_GALLIUM_SOURCES) \
+	$(DRIVER_SOURCES)
+
+ASM_SOURCES = 
+
+DRIVER_DEFINES = -I../common $(shell pkg-config libdrm --atleast-version=2.3.1 \
+				&& echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP")
+
+include ../../Makefile.template
+
+#intel_tex_layout.o: $(TOP)/src/mesa/drivers/dri/intel/intel_tex_layout.c
+
+symlinks:
diff --git a/src/gallium/winsys/drm/intel/dri/SConscript b/src/gallium/winsys/drm/intel/dri/SConscript
new file mode 100644
index 0000000000..6a4f50afcc
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/SConscript
@@ -0,0 +1,41 @@
+Import('*')
+
+if 'mesa' in env['statetrackers']:
+
+	env = drienv.Clone()
+
+	env.Append(CPPPATH = [
+		'../intel',
+		'server'
+	])
+
+	#MINIGLX_SOURCES = server/intel_dri.c
+
+	DRIVER_SOURCES = [
+		'intel_winsys_pipe.c',
+		'intel_winsys_softpipe.c',
+		'intel_winsys_i915.c',
+		'intel_batchbuffer.c',
+		'intel_swapbuffers.c',
+		'intel_context.c',
+		'intel_lock.c',
+		'intel_screen.c',
+		'intel_batchpool.c',
+	]
+
+	sources = \
+		COMMON_GALLIUM_SOURCES + \
+		COMMON_BM_SOURCES + \
+		DRIVER_SOURCES
+
+	drivers = [
+		softpipe,
+		i915simple
+	]
+
+	# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions
+	env.SharedLibrary(
+		target ='i915tex_dri.so',
+		source = sources,
+		LIBS = drivers + mesa + auxiliaries + env['LIBS'],
+	)
diff --git a/src/gallium/winsys/drm/intel/dri/intel_batchbuffer.h b/src/gallium/winsys/drm/intel/dri/intel_batchbuffer.h
new file mode 100644
index 0000000000..3e95326168
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_batchbuffer.h
@@ -0,0 +1,24 @@
+#ifndef INTEL_BATCHBUFFER_H
+#define INTEL_BATCHBUFFER_H
+
+#include "intel_be_batchbuffer.h"
+
+/*
+ * Need to redefine the BATCH defines
+ */
+
+#undef BEGIN_BATCH
+#define BEGIN_BATCH(dwords, relocs) \
+   (i915_batchbuffer_check(&intel->base.batch->base, dwords, relocs))
+
+#undef OUT_BATCH
+#define OUT_BATCH(d) \
+   i915_batchbuffer_dword(&intel->base.batch->base, d)
+
+#undef OUT_RELOC
+#define OUT_RELOC(buf,flags,mask,delta) do { 					\
+   assert((delta) >= 0);							\
+   intel_be_offset_relocation(intel->base.batch, delta, buf, flags, mask); 	\
+} while (0)
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/dri/intel_context.c b/src/gallium/winsys/drm/intel/dri/intel_context.c
new file mode 100644
index 0000000000..97ef731aaa
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_context.c
@@ -0,0 +1,337 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "i830_dri.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_swapbuffers.h"
+#include "intel_batchbuffer.h"
+#include "intel_winsys_softpipe.h"
+
+#include "i915simple/i915_screen.h"
+
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+
+#include "utils.h"
+
+
+#ifdef DEBUG
+int __intel_debug = 0;
+#endif
+
+
+#define need_GL_ARB_multisample
+#define need_GL_ARB_point_parameters
+#define need_GL_ARB_texture_compression
+#define need_GL_ARB_vertex_buffer_object
+#define need_GL_ARB_vertex_program
+#define need_GL_ARB_window_pos
+#define need_GL_EXT_blend_color
+#define need_GL_EXT_blend_equation_separate
+#define need_GL_EXT_blend_func_separate
+#define need_GL_EXT_blend_minmax
+#define need_GL_EXT_cull_vertex
+#define need_GL_EXT_fog_coord
+#define need_GL_EXT_framebuffer_object
+#define need_GL_EXT_multi_draw_arrays
+#define need_GL_EXT_secondary_color
+#define need_GL_NV_vertex_program
+#include "extension_helper.h"
+
+
+/**
+ * Extension strings exported by the intel driver.
+ *
+ * \note
+ * It appears that ARB_texture_env_crossbar has "disappeared" compared to the
+ * old i830-specific driver.
+ */
+const struct dri_extension card_extensions[] = {
+   {"GL_ARB_multisample", GL_ARB_multisample_functions},
+   {"GL_ARB_multitexture", NULL},
+   {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions},
+   {"GL_ARB_texture_border_clamp", NULL},
+   {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions},
+   {"GL_ARB_texture_cube_map", NULL},
+   {"GL_ARB_texture_env_add", NULL},
+   {"GL_ARB_texture_env_combine", NULL},
+   {"GL_ARB_texture_env_dot3", NULL},
+   {"GL_ARB_texture_mirrored_repeat", NULL},
+   {"GL_ARB_texture_rectangle", NULL},
+   {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions},
+   {"GL_ARB_pixel_buffer_object", NULL},
+   {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
+   {"GL_ARB_window_pos", GL_ARB_window_pos_functions},
+   {"GL_EXT_blend_color", GL_EXT_blend_color_functions},
+   {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions},
+   {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
+   {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
+   {"GL_EXT_blend_subtract", NULL},
+   {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions},
+   {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions},
+   {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions},
+   {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions},
+   {"GL_EXT_packed_depth_stencil", NULL},
+   {"GL_EXT_pixel_buffer_object", NULL},
+   {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions},
+   {"GL_EXT_stencil_wrap", NULL},
+   {"GL_EXT_texture_edge_clamp", NULL},
+   {"GL_EXT_texture_env_combine", NULL},
+   {"GL_EXT_texture_env_dot3", NULL},
+   {"GL_EXT_texture_filter_anisotropic", NULL},
+   {"GL_EXT_texture_lod_bias", NULL},
+   {"GL_3DFX_texture_compression_FXT1", NULL},
+   {"GL_APPLE_client_storage", NULL},
+   {"GL_MESA_pack_invert", NULL},
+   {"GL_MESA_ycbcr_texture", NULL},
+   {"GL_NV_blend_square", NULL},
+   {"GL_NV_vertex_program", GL_NV_vertex_program_functions},
+   {"GL_NV_vertex_program1_1", NULL},
+   {"GL_SGIS_generate_mipmap", NULL },
+   {NULL, NULL}
+};
+
+
+
+#ifdef DEBUG
+static const struct dri_debug_control debug_control[] = {
+   {"ioctl", DEBUG_IOCTL},
+   {"bat", DEBUG_BATCH},
+   {"lock", DEBUG_LOCK},
+   {"swap", DEBUG_SWAP},
+   {NULL, 0}
+};
+#endif
+
+
+
+static void
+intel_lock_hardware(struct intel_be_context *context)
+{
+   struct intel_context *intel = (struct intel_context *)context;
+   LOCK_HARDWARE(intel);
+}
+
+static void
+intel_unlock_hardware(struct intel_be_context *context)
+{
+   struct intel_context *intel = (struct intel_context *)context;
+   UNLOCK_HARDWARE(intel);
+}
+
+static boolean
+intel_locked_hardware(struct intel_be_context *context)
+{
+   struct intel_context *intel = (struct intel_context *)context;
+   return intel->locked ? TRUE : FALSE;
+}
+
+GLboolean
+intelCreateContext(const __GLcontextModes * visual,
+                   __DRIcontextPrivate * driContextPriv,
+                   void *sharedContextPrivate)
+{
+   struct intel_context *intel = CALLOC_STRUCT(intel_context);
+   __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
+   struct intel_screen *intelScreen = intel_screen(sPriv);
+   drmI830Sarea *saPriv = intelScreen->sarea;
+   int fthrottle_mode;
+   GLboolean havePools;
+   struct pipe_context *pipe;
+   struct st_context *st_share = NULL;
+
+   if (sharedContextPrivate) {
+      st_share = ((struct intel_context *) sharedContextPrivate)->st;
+   }
+
+   driContextPriv->driverPrivate = intel;
+   intel->intelScreen = intelScreen;
+   intel->driScreen = sPriv;
+   intel->sarea = saPriv;
+
+   driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache,
+                       intel->driScreen->myNum, "i915");
+
+
+   /*
+    * memory pools
+    */
+   DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext);
+   // ZZZ JB should be per screen and not be done per context
+   havePools = intelCreatePools(sPriv);
+   DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext);
+   if (!havePools)
+      return GL_FALSE;
+
+
+   /* Dri stuff */
+   intel->hHWContext = driContextPriv->hHWContext;
+   intel->driFd = sPriv->fd;
+   intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock;
+
+   fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode");
+   intel->iw.irq_seq = -1;
+   intel->irqsEmitted = 0;
+
+   intel->last_swap_fence = NULL;
+   intel->first_swap_fence = NULL;
+
+#ifdef DEBUG
+   __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control);
+#endif
+   intel->base.hardware_lock = intel_lock_hardware;
+   intel->base.hardware_unlock = intel_unlock_hardware;
+   intel->base.hardware_locked = intel_locked_hardware;
+
+   intel_be_init_context(&intel->base, &intelScreen->base);
+
+   /*
+    * Pipe-related setup
+    */
+   if (getenv("INTEL_SP")) {
+      /* use softpipe driver instead of hw */
+      pipe = intel_create_softpipe( intel, &intelScreen->base.base );
+   }
+   else {
+      switch (intel->intelScreen->deviceID) {
+      case PCI_CHIP_I945_G:
+      case PCI_CHIP_I945_GM:
+      case PCI_CHIP_I945_GME:
+      case PCI_CHIP_G33_G:
+      case PCI_CHIP_Q33_G:
+      case PCI_CHIP_Q35_G:
+      case PCI_CHIP_I915_G:
+      case PCI_CHIP_I915_GM:
+	 pipe = i915_create_context(intelScreen->base.screen,
+				    &intelScreen->base.base,
+				    &intel->base.base);
+	 break;
+      default:
+	 fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n",
+                 intel->intelScreen->deviceID, __FUNCTION__);
+
+	 pipe = intel_create_softpipe( intel, &intelScreen->base.base );
+	 break;
+      }
+   }
+
+   pipe->priv = intel;
+
+   intel->st = st_create_context(pipe, visual, st_share);
+
+   driInitExtensions( intel->st->ctx, card_extensions, GL_TRUE );
+
+   return GL_TRUE;
+}
+
+
+void
+intelDestroyContext(__DRIcontextPrivate * driContextPriv)
+{
+   struct intel_context *intel = intel_context(driContextPriv);
+
+   assert(intel);               /* should never be null */
+   if (intel) {
+      st_finish(intel->st);
+
+      if (intel->last_swap_fence) {
+	 driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE);
+	 driFenceUnReference(&intel->last_swap_fence);
+	 intel->last_swap_fence = NULL;
+      }
+      if (intel->first_swap_fence) {
+	 driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE);
+	 driFenceUnReference(&intel->first_swap_fence);
+	 intel->first_swap_fence = NULL;
+      }
+
+      if (intel->intelScreen->dummyContext == intel)
+         intel->intelScreen->dummyContext = NULL;
+
+      st_destroy_context(intel->st);
+      intel_be_destroy_context(&intel->base);
+      free(intel);
+   }
+}
+
+
+GLboolean
+intelUnbindContext(__DRIcontextPrivate * driContextPriv)
+{
+   struct intel_context *intel = intel_context(driContextPriv);
+   st_flush(intel->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+   /* XXX make_current(NULL)? */
+   return GL_TRUE;
+}
+
+
+GLboolean
+intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
+                 __DRIdrawablePrivate * driDrawPriv,
+                 __DRIdrawablePrivate * driReadPriv)
+{
+   if (driContextPriv) {
+      struct intel_context *intel = intel_context(driContextPriv);
+      struct intel_framebuffer *draw_fb = intel_framebuffer(driDrawPriv);
+      struct intel_framebuffer *read_fb = intel_framebuffer(driReadPriv);
+
+      assert(draw_fb->stfb);
+      assert(read_fb->stfb);
+
+      /* This is for situations in which we need a rendering context but
+       * there may not be any currently bound.
+       */
+      intel->intelScreen->dummyContext = intel;
+
+      st_make_current(intel->st, draw_fb->stfb, read_fb->stfb);
+
+      if ((intel->driDrawable != driDrawPriv) ||
+	  (intel->lastStamp != driDrawPriv->lastStamp)) {
+         intel->driDrawable = driDrawPriv;
+         intelUpdateWindowSize(driDrawPriv);
+         intel->lastStamp = driDrawPriv->lastStamp;
+      }
+
+      /* The size of the draw buffer will have been updated above.
+       * If the readbuffer is a different window, check/update its size now.
+       */
+      if (driReadPriv != driDrawPriv) {
+         intelUpdateWindowSize(driReadPriv);
+      }
+
+   }
+   else {
+      st_make_current(NULL, NULL, NULL);
+   }
+
+   return GL_TRUE;
+}
diff --git a/src/gallium/winsys/drm/intel/dri/intel_context.h b/src/gallium/winsys/drm/intel/dri/intel_context.h
new file mode 100644
index 0000000000..5d22a422af
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_context.h
@@ -0,0 +1,164 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef INTEL_CONTEXT_H
+#define INTEL_CONTEXT_H
+
+#include <stdint.h>
+#include "drm.h"
+
+#include "pipe/p_debug.h"
+
+#include "intel_screen.h"
+#include "i915_drm.h"
+
+#include "intel_be_context.h"
+
+
+struct pipe_context;
+struct intel_context;
+struct _DriBufferObject;
+struct st_context;
+
+
+#define INTEL_MAX_FIXUP 64
+
+/**
+ * Intel rendering context, contains a state tracker and intel-specific info.
+ */
+struct intel_context
+{
+   struct intel_be_context base;
+   struct st_context *st;
+
+   struct _DriFenceObject *last_swap_fence;
+   struct _DriFenceObject *first_swap_fence;
+
+//   struct intel_batchbuffer *batch;
+
+   boolean locked;
+   char *prevLockFile;
+   int prevLockLine;
+
+   uint irqsEmitted;
+   drm_i915_irq_wait_t iw;
+
+   drm_context_t hHWContext;
+   drmLock *driHwLock;
+   int driFd;
+
+   __DRIdrawablePrivate *driDrawable;
+   __DRIscreenPrivate *driScreen;
+   struct intel_screen *intelScreen;
+   drmI830Sarea *sarea;
+
+   uint lastStamp;
+
+   /**
+    * Configuration cache
+    */
+   driOptionCache optionCache;
+};
+
+
+
+/**
+ * Intel framebuffer.
+ */
+struct intel_framebuffer
+{
+   struct st_framebuffer *stfb;
+
+   /* other fields TBD */
+   int other;
+};
+
+
+
+
+/* These are functions now:
+ */
+void LOCK_HARDWARE( struct intel_context *intel );
+void UNLOCK_HARDWARE( struct intel_context *intel );
+
+extern char *__progname;
+
+
+
+/* ================================================================
+ * Debugging:
+ */
+#ifdef DEBUG
+extern int __intel_debug;
+
+#define DEBUG_SWAP	0x1
+#define DEBUG_LOCK      0x2
+#define DEBUG_IOCTL	0x4
+#define DEBUG_BATCH     0x8
+
+#define DBG(flag, ...)  do { 			\
+   if (__intel_debug & (DEBUG_##flag)) 		\
+      printf(__VA_ARGS__); 		\
+} while(0)
+
+#else
+#define DBG(flag, ...)
+#endif
+
+
+
+#define PCI_CHIP_845_G			0x2562
+#define PCI_CHIP_I830_M			0x3577
+#define PCI_CHIP_I855_GM		0x3582
+#define PCI_CHIP_I865_G			0x2572
+#define PCI_CHIP_I915_G			0x2582
+#define PCI_CHIP_I915_GM		0x2592
+#define PCI_CHIP_I945_G			0x2772
+#define PCI_CHIP_I945_GM		0x27A2
+#define PCI_CHIP_I945_GME		0x27AE
+#define PCI_CHIP_G33_G			0x29C2
+#define PCI_CHIP_Q35_G			0x29B2
+#define PCI_CHIP_Q33_G			0x29D2
+
+
+/** Cast wrapper */
+static INLINE struct intel_context *
+intel_context(__DRIcontextPrivate *driContextPriv)
+{
+   return (struct intel_context *) driContextPriv->driverPrivate;
+}
+
+
+/** Cast wrapper */
+static INLINE struct intel_framebuffer *
+intel_framebuffer(__DRIdrawablePrivate * driDrawPriv)
+{
+   return (struct intel_framebuffer *) driDrawPriv->driverPrivate;
+}
+
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/dri/intel_lock.c b/src/gallium/winsys/drm/intel/dri/intel_lock.c
new file mode 100644
index 0000000000..ad1c202429
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_lock.c
@@ -0,0 +1,102 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "main/glheader.h"
+#include "pipe/p_thread.h"
+#include <GL/internal/glcore.h>
+#include "state_tracker/st_public.h"
+#include "intel_context.h"
+#include "i830_dri.h"
+
+
+
+pipe_static_mutex( lockMutex );
+
+
+static void
+intelContendedLock(struct intel_context *intel, uint flags)
+{
+   __DRIdrawablePrivate *dPriv = intel->driDrawable;
+   __DRIscreenPrivate *sPriv = intel->driScreen;
+   struct intel_screen *intelScreen = intel_screen(sPriv);
+   drmI830Sarea *sarea = intel->sarea;
+
+   drmGetLock(intel->driFd, intel->hHWContext, flags);
+
+   DBG(LOCK, "%s - got contended lock\n", __progname);
+
+   /* If the window moved, may need to set a new cliprect now.
+    *
+    * NOTE: This releases and regains the hw lock, so all state
+    * checking must be done *after* this call:
+    */
+   if (dPriv)
+      DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv);
+
+   if (sarea->width != intelScreen->front.width ||
+       sarea->height != intelScreen->front.height) {
+
+      intelUpdateScreenRotation(sPriv, sarea);
+   }
+}
+
+
+/* Lock the hardware and validate our state.
+ */
+void LOCK_HARDWARE( struct intel_context *intel )
+{
+    char __ret = 0;
+
+    pipe_mutex_lock(lockMutex);
+    assert(!intel->locked);
+
+    DRM_CAS(intel->driHwLock, intel->hHWContext,
+            (DRM_LOCK_HELD|intel->hHWContext), __ret);
+
+    if (__ret)
+       intelContendedLock( intel, 0 );
+
+    DBG(LOCK, "%s - locked\n", __progname);
+
+    intel->locked = 1;
+}
+
+
+/* Unlock the hardware using the global current context
+ */
+void UNLOCK_HARDWARE( struct intel_context *intel )
+{
+   assert(intel->locked);
+   intel->locked = 0;
+
+   DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext);
+
+   pipe_mutex_unlock(lockMutex);
+
+   DBG(LOCK, "%s - unlocked\n", __progname);
+}
diff --git a/src/gallium/winsys/drm/intel/dri/intel_reg.h b/src/gallium/winsys/drm/intel/dri/intel_reg.h
new file mode 100644
index 0000000000..4f33bee438
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_reg.h
@@ -0,0 +1,53 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#ifndef _INTEL_REG_H_
+#define _INTEL_REG_H_
+
+
+#define BR00_BITBLT_CLIENT   0x40000000
+#define BR00_OP_COLOR_BLT    0x10000000
+#define BR00_OP_SRC_COPY_BLT 0x10C00000
+#define BR13_SOLID_PATTERN   0x80000000
+
+#define XY_COLOR_BLT_CMD		((2<<29)|(0x50<<22)|0x4)
+#define XY_COLOR_BLT_WRITE_ALPHA	(1<<21)
+#define XY_COLOR_BLT_WRITE_RGB		(1<<20)
+
+#define XY_SRC_COPY_BLT_CMD             ((2<<29)|(0x53<<22)|6)
+#define XY_SRC_COPY_BLT_WRITE_ALPHA     (1<<21)
+#define XY_SRC_COPY_BLT_WRITE_RGB       (1<<20)
+
+#define MI_WAIT_FOR_EVENT               ((0x3<<23))
+#define MI_WAIT_FOR_PLANE_B_FLIP        (1<<6)
+#define MI_WAIT_FOR_PLANE_A_FLIP        (1<<2)
+
+#define MI_BATCH_BUFFER_END 	        (0xA<<23)
+
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/dri/intel_screen.c b/src/gallium/winsys/drm/intel/dri/intel_screen.c
new file mode 100644
index 0000000000..ed75368982
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_screen.c
@@ -0,0 +1,703 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "utils.h"
+#include "vblank.h"
+#include "xmlpool.h"
+
+#include "intel_context.h"
+#include "intel_screen.h"
+#include "intel_batchbuffer.h"
+#include "intel_swapbuffers.h"
+
+#include "i830_dri.h"
+#include "ws_dri_bufpool.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_cb_fbo.h"
+
+static void
+intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle);
+
+static void
+intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle)
+{
+   struct pipe_screen *screen = intelScreen->base.screen;
+   struct pipe_texture *texture;
+   struct pipe_texture templat;
+   struct pipe_surface *surface;
+   struct pipe_buffer *buffer;
+   unsigned pitch;
+
+   assert(intelScreen->front.cpp == 4);
+
+   buffer = intel_be_buffer_from_handle(&intelScreen->base,
+                                        "front", handle);
+
+   if (!buffer)
+      return;
+
+   intelScreen->front.buffer = dri_bo(buffer);
+
+   memset(&templat, 0, sizeof(templat));
+   templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+   templat.target = PIPE_TEXTURE_2D;
+   templat.last_level = 0;
+   templat.depth[0] = 1;
+   templat.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+   templat.width[0] = intelScreen->front.width;
+   templat.height[0] = intelScreen->front.height;
+   pf_get_block(templat.format, &templat.block);
+   pitch = intelScreen->front.pitch;
+
+   texture = screen->texture_blanket(screen,
+                                     &templat,
+                                     &pitch,
+                                     buffer);
+
+   /* Unref the buffer we don't need it anyways */
+   pipe_buffer_reference(screen, &buffer, NULL);
+
+   surface = screen->get_tex_surface(screen,
+                                     texture,
+                                     0,
+                                     0,
+                                     0,
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+   intelScreen->front.texture = texture;
+   intelScreen->front.surface = surface;
+}
+
+PUBLIC const char __driConfigOptions[] =
+   DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE
+   DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS)
+   DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0)
+   DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY
+//   DRI_CONF_FORCE_S3TC_ENABLE(false)
+   DRI_CONF_ALLOW_LARGE_TEXTURES(1)
+   DRI_CONF_SECTION_END DRI_CONF_END;
+
+const uint __driNConfigOptions = 3;
+
+#ifdef USE_NEW_INTERFACE
+static PFNGLXCREATECONTEXTMODES create_context_modes = NULL;
+#endif /*USE_NEW_INTERFACE */
+
+extern const struct dri_extension card_extensions[];
+
+static GLboolean
+intel_get_param(__DRIscreenPrivate *psp, int param, int *value)
+{
+   int ret;
+   struct drm_i915_getparam gp;
+
+   gp.param = param;
+   gp.value = value;
+
+   ret = drmCommandWriteRead(psp->fd, DRM_I915_GETPARAM, &gp, sizeof(gp));
+   if (ret) {
+      fprintf(stderr, "drm_i915_getparam: %d\n", ret);
+      return GL_FALSE;
+   }
+
+   return GL_TRUE;
+}
+
+static void
+intelSetTexOffset(__DRIcontext *pDRICtx, int texname,
+		  unsigned long long offset, int depth, uint pitch)
+{
+   abort();
+#if 0
+   struct intel_context *intel = (struct intel_context*)
+      ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate;
+   struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname);
+   struct st_texture_object *stObj = st_texture_object(tObj);
+
+   if (!stObj)
+      return;
+
+   if (stObj->pt)
+      st->pipe->texture_release(intel->st->pipe, &stObj->pt);
+
+   stObj->imageOverride = GL_TRUE;
+   stObj->depthOverride = depth;
+   stObj->pitchOverride = pitch;
+
+   if (offset)
+      stObj->textureOffset = offset;
+#endif
+}
+
+
+#if 0
+static void
+intelHandleDrawableConfig(__DRIdrawablePrivate *dPriv,
+			  __DRIcontextPrivate *pcp,
+			  __DRIDrawableConfigEvent *event)
+{
+   (void) dPriv;
+   (void) pcp;
+   (void) event;
+}
+#endif
+
+#if 0
+static void
+intelHandleBufferAttach(__DRIdrawablePrivate *dPriv,
+			__DRIcontextPrivate *pcp,
+			__DRIBufferAttachEvent *ba)
+{
+   struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv);
+
+   switch (ba->buffer.attachment) {
+   case DRI_DRAWABLE_BUFFER_FRONT_LEFT:
+      intelScreen->front.width = dPriv->w;
+      intelScreen->front.height = dPriv->h;
+      intelScreen->front.cpp = ba->buffer.cpp;
+      intelScreen->front.pitch = ba->buffer.pitch;
+      driGenBuffers(intelScreen->base.staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0);
+      driBOSetReferenced(intelScreen->front.buffer, ba->buffer.handle);
+      break;
+
+   case DRI_DRAWABLE_BUFFER_BACK_LEFT:
+   case DRI_DRAWABLE_BUFFER_DEPTH:
+   case DRI_DRAWABLE_BUFFER_STENCIL:
+   case DRI_DRAWABLE_BUFFER_ACCUM:
+      /* anything ?? */
+      break;
+
+   default:
+      fprintf(stderr, "unhandled buffer attach event, attachment type %d\n",
+	      ba->buffer.attachment);
+      return;
+   }
+}
+#endif
+
+static const __DRItexOffsetExtension intelTexOffsetExtension = {
+   { __DRI_TEX_OFFSET },
+   intelSetTexOffset,
+};
+
+#if 0
+static const __DRItexBufferExtension intelTexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   intelSetTexBuffer,
+};
+#endif
+
+static const __DRIextension *intelScreenExtensions[] = {
+    &driReadDrawableExtension,
+    &driCopySubBufferExtension.base,
+    &driSwapControlExtension.base,
+    &driFrameTrackingExtension.base,
+    &driMediaStreamCounterExtension.base,
+    &intelTexOffsetExtension.base,
+//    &intelTexBufferExtension.base,
+    NULL
+};
+
+
+static void
+intelPrintDRIInfo(struct intel_screen * intelScreen,
+                  __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv)
+{
+   fprintf(stderr, "*** Front size:   0x%x  offset: 0x%x  pitch: %d\n",
+           intelScreen->front.size, intelScreen->front.offset,
+           intelScreen->front.pitch);
+   fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem);
+}
+
+
+#if 0
+static void
+intelPrintSAREA(const drmI830Sarea * sarea)
+{
+   fprintf(stderr, "SAREA: sarea width %d  height %d\n", sarea->width,
+           sarea->height);
+   fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch);
+   fprintf(stderr,
+           "SAREA: front offset: 0x%08x  size: 0x%x  handle: 0x%x\n",
+           sarea->front_offset, sarea->front_size,
+           (unsigned) sarea->front_handle);
+   fprintf(stderr,
+           "SAREA: back  offset: 0x%08x  size: 0x%x  handle: 0x%x\n",
+           sarea->back_offset, sarea->back_size,
+           (unsigned) sarea->back_handle);
+   fprintf(stderr, "SAREA: depth offset: 0x%08x  size: 0x%x  handle: 0x%x\n",
+           sarea->depth_offset, sarea->depth_size,
+           (unsigned) sarea->depth_handle);
+   fprintf(stderr, "SAREA: tex   offset: 0x%08x  size: 0x%x  handle: 0x%x\n",
+           sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle);
+   fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation);
+   fprintf(stderr,
+           "SAREA: rotated offset: 0x%08x  size: 0x%x\n",
+           sarea->rotated_offset, sarea->rotated_size);
+   fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch);
+}
+#endif
+
+
+/**
+ * Use the information in the sarea to update the screen parameters
+ * related to screen rotation. Needs to be called locked.
+ */
+void
+intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea)
+{
+   struct intel_screen *intelScreen = intel_screen(sPriv);
+
+   if (intelScreen->front.map) {
+      drmUnmap(intelScreen->front.map, intelScreen->front.size);
+      intelScreen->front.map = NULL;
+   }
+
+   if (intelScreen->front.buffer)
+      driDeleteBuffers(1, &intelScreen->front.buffer);
+
+   intelScreen->front.width = sarea->width;
+   intelScreen->front.height = sarea->height;
+   intelScreen->front.offset = sarea->front_offset;
+   intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp;
+   intelScreen->front.size = sarea->front_size;
+   intelScreen->front.handle = sarea->front_handle;
+
+   assert( sarea->front_size >=
+	   intelScreen->front.pitch * intelScreen->front.height );
+
+#if 0 /* JB not important */
+   if (!sarea->front_handle)
+      return;
+
+   if (drmMap(sPriv->fd,
+	      sarea->front_handle,
+	      intelScreen->front.size,
+	      (drmAddress *) & intelScreen->front.map) != 0) {
+      fprintf(stderr, "drmMap(frontbuffer) failed!\n");
+      return;
+   }
+#endif
+
+#if 0 /* JB */
+   if (intelScreen->staticPool) {
+      driGenBuffers(intelScreen->staticPool, "static region", 1,
+		    &intelScreen->front.buffer, 64,
+		    DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE |
+		    DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0);
+
+      driBOSetStatic(intelScreen->front.buffer,
+		     intelScreen->front.offset,
+		     intelScreen->front.pitch * intelScreen->front.height,
+		     intelScreen->front.map, 0);
+   }
+#else
+   if (intelScreen->base.staticPool) {
+      if (intelScreen->front.buffer) {
+	 driBOUnReference(intelScreen->front.buffer);
+	 pipe_surface_reference(&intelScreen->front.surface, NULL);
+	 pipe_texture_reference(&intelScreen->front.texture, NULL);
+      }
+      intelCreateSurface(intelScreen, &intelScreen->base.base, sarea->front_bo_handle);
+   }
+#endif
+}
+
+
+boolean
+intelCreatePools(__DRIscreenPrivate * sPriv)
+{
+   //unsigned batchPoolSize = 1024*1024;
+   struct intel_screen *intelScreen = intel_screen(sPriv);
+
+   if (intelScreen->havePools)
+      return GL_TRUE;
+
+   intelScreen->havePools = GL_TRUE;
+
+   if (intelScreen->sarea)
+	intelUpdateScreenRotation(sPriv, intelScreen->sarea);
+
+   return GL_TRUE;
+}
+
+static const char *
+intel_get_name( struct pipe_winsys *winsys )
+{
+   return "Intel/DRI/ttm";
+}
+
+/*
+ * The state tracker (should!) keep track of whether the fake
+ * frontbuffer has been touched by any rendering since the last time
+ * we copied its contents to the real frontbuffer.  Our task is easy:
+ */
+static void
+intel_flush_frontbuffer( struct pipe_winsys *winsys,
+                         struct pipe_surface *surf,
+                         void *context_private)
+{
+   struct intel_context *intel = (struct intel_context *) context_private;
+   __DRIdrawablePrivate *dPriv = intel->driDrawable;
+
+   intelDisplaySurface(dPriv, surf, NULL);
+}
+
+static boolean
+intelInitDriver(__DRIscreenPrivate * sPriv)
+{
+   struct intel_screen *intelScreen;
+   I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv;
+
+   if (sPriv->devPrivSize != sizeof(I830DRIRec)) {
+      fprintf(stderr,
+              "\nERROR!  sizeof(I830DRIRec) does not match passed size from device driver\n");
+      return GL_FALSE;
+   }
+
+   /* Allocate the private area */
+   intelScreen = CALLOC_STRUCT(intel_screen);
+   if (!intelScreen)
+      return GL_FALSE;
+
+   /* parse information in __driConfigOptions */
+   driParseOptionInfo(&intelScreen->optionCache,
+                      __driConfigOptions, __driNConfigOptions);
+
+   sPriv->private = (void *) intelScreen;
+   intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) +
+                                            gDRIPriv->sarea_priv_offset);
+
+   intelScreen->deviceID = gDRIPriv->deviceID;
+
+   intelScreen->front.cpp = gDRIPriv->cpp;
+   intelScreen->drmMinor = sPriv->drm_version.minor;
+   intelUpdateScreenRotation(sPriv, intelScreen->sarea);
+
+   if (0)
+      intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv);
+
+   sPriv->extensions = intelScreenExtensions;
+
+   intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer;
+   intelScreen->base.base.get_name = intel_get_name;
+   intel_be_init_device(&intelScreen->base, sPriv->fd, intelScreen->deviceID);
+
+   return GL_TRUE;
+}
+
+
+static void
+intelDestroyScreen(__DRIscreenPrivate * sPriv)
+{
+   struct intel_screen *intelScreen = intel_screen(sPriv);
+
+   intel_be_destroy_device(&intelScreen->base);
+   /*  intelUnmapScreenRegions(intelScreen); */
+
+   FREE(intelScreen);
+   sPriv->private = NULL;
+}
+
+
+/**
+ * This is called when we need to set up GL rendering to a new X window.
+ */
+static boolean
+intelCreateBuffer(__DRIscreenPrivate * driScrnPriv,
+                  __DRIdrawablePrivate * driDrawPriv,
+                  const __GLcontextModes * visual, boolean isPixmap)
+{
+   if (isPixmap) {
+      return GL_FALSE;          /* not implemented */
+   }
+   else {
+      enum pipe_format colorFormat, depthFormat, stencilFormat;
+      struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer);
+
+      if (!intelfb)
+         return GL_FALSE;
+
+      if (visual->redBits == 5)
+         colorFormat = PIPE_FORMAT_R5G6B5_UNORM;
+      else
+         colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+      if (visual->depthBits == 16)
+         depthFormat = PIPE_FORMAT_Z16_UNORM;
+      else if (visual->depthBits == 24)
+         depthFormat = PIPE_FORMAT_S8Z24_UNORM;
+      else
+         depthFormat = PIPE_FORMAT_NONE;
+
+      if (visual->stencilBits == 8)
+         stencilFormat = PIPE_FORMAT_S8Z24_UNORM;
+      else
+         stencilFormat = PIPE_FORMAT_NONE;
+
+      intelfb->stfb = st_create_framebuffer(visual,
+                                            colorFormat,
+                                            depthFormat,
+                                            stencilFormat,
+                                            driDrawPriv->w,
+                                            driDrawPriv->h,
+                                            (void*) intelfb);
+      if (!intelfb->stfb) {
+         free(intelfb);
+         return GL_FALSE;
+      }
+
+      driDrawPriv->driverPrivate = (void *) intelfb;
+      return GL_TRUE;
+   }
+}
+
+static void
+intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv)
+{
+   struct intel_framebuffer *intelfb = intel_framebuffer(driDrawPriv);
+   assert(intelfb->stfb);
+   st_unreference_framebuffer(intelfb->stfb);
+   free(intelfb);
+}
+
+
+/**
+ * Get information about previous buffer swaps.
+ */
+static int
+intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo)
+{
+   if ((dPriv == NULL) || (dPriv->driverPrivate == NULL)
+       || (sInfo == NULL)) {
+      return -1;
+   }
+
+   return 0;
+}
+
+static __DRIconfig **
+intelFillInModes(__DRIscreenPrivate *psp,
+		 unsigned pixel_bits, unsigned depth_bits,
+                 unsigned stencil_bits, GLboolean have_back_buffer)
+{
+   __DRIconfig **configs;
+   __GLcontextModes *m;
+   unsigned num_modes;
+   unsigned depth_buffer_factor;
+   unsigned back_buffer_factor;
+   GLenum fb_format;
+   GLenum fb_type;
+   int i;
+
+   /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't
+    * support pageflipping at all.
+    */
+   static const GLenum back_buffer_modes[] = {
+      GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML
+   };
+
+   uint8_t depth_bits_array[3];
+   uint8_t stencil_bits_array[3];
+   uint8_t msaa_samples_array[1];
+
+
+   depth_bits_array[0] = 0;
+   depth_bits_array[1] = depth_bits;
+   depth_bits_array[2] = depth_bits;
+   msaa_samples_array[0] = 0;
+
+   /* Just like with the accumulation buffer, always provide some modes
+    * with a stencil buffer.  It will be a sw fallback, but some apps won't
+    * care about that.
+    */
+   stencil_bits_array[0] = 0;
+   stencil_bits_array[1] = 0;
+   if (depth_bits == 24)
+      stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
+
+   stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
+
+   depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
+   back_buffer_factor = (have_back_buffer) ? 3 : 1;
+
+   num_modes = depth_buffer_factor * back_buffer_factor * 4;
+
+   if (pixel_bits == 16) {
+      fb_format = GL_RGB;
+      fb_type = GL_UNSIGNED_SHORT_5_6_5;
+   }
+   else {
+      fb_format = GL_BGRA;
+      fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+   }
+
+   configs = driCreateConfigs(fb_format, fb_type,
+			      depth_bits_array, stencil_bits_array,
+			      depth_buffer_factor, back_buffer_modes,
+			      back_buffer_factor, msaa_samples_array, 1);
+   if (configs == NULL) {
+    fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
+              __LINE__);
+      return NULL;
+   }
+
+   /* Mark the visual as slow if there are "fake" stencil bits.
+    */
+   for (i = 0; configs[i]; i++) {
+      m = &configs[i]->modes;
+      if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) {
+         m->visualRating = GLX_SLOW_CONFIG;
+      }
+   }
+
+   return configs;
+}
+
+/**
+ * This is the driver specific part of the createNewScreen entry point.
+ * 
+ * \todo maybe fold this into intelInitDriver
+ *
+ * \return the __GLcontextModes supported by this driver
+ */
+static const __DRIconfig **intelInitScreen(__DRIscreenPrivate *psp)
+{
+#ifdef I915
+   static const __DRIversion ddx_expected = { 1, 5, 0 };
+#else
+   static const __DRIversion ddx_expected = { 1, 6, 0 };
+#endif
+   static const __DRIversion dri_expected = { 4, 0, 0 };
+   static const __DRIversion drm_expected = { 1, 5, 0 };
+   I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv;
+
+   if (!driCheckDriDdxDrmVersions2("i915",
+                                   &psp->dri_version, &dri_expected,
+                                   &psp->ddx_version, &ddx_expected,
+                                   &psp->drm_version, &drm_expected)) {
+      return NULL;
+   }
+
+   /* Calling driInitExtensions here, with a NULL context pointer,
+    * does not actually enable the extensions.  It just makes sure
+    * that all the dispatch offsets for all the extensions that
+    * *might* be enables are known.  This is needed because the
+    * dispatch offsets need to be known when _mesa_context_create is
+    * called, but we can't enable the extensions until we have a
+    * context pointer.
+    *
+    * Hello chicken.  Hello egg.  How are you two today?
+    */
+   driInitExtensions( NULL, card_extensions, GL_FALSE );
+   //intelInitExtensions(NULL, GL_TRUE);
+	   
+   if (!intelInitDriver(psp))
+       return NULL;
+
+   psp->extensions = intelScreenExtensions;
+
+   return (const __DRIconfig **)
+       intelFillInModes(psp, dri_priv->cpp * 8,
+			(dri_priv->cpp == 2) ? 16 : 24,
+			(dri_priv->cpp == 2) ? 0  : 8, 1);
+}
+
+/**
+ * This is the driver specific part of the createNewScreen entry point.
+ * 
+ * \return the __GLcontextModes supported by this driver
+ */
+static const
+__DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp)
+{
+   struct intel_screen *intelScreen;
+
+   /* Calling driInitExtensions here, with a NULL context pointer,
+    * does not actually enable the extensions.  It just makes sure
+    * that all the dispatch offsets for all the extensions that
+    * *might* be enables are known.  This is needed because the
+    * dispatch offsets need to be known when _mesa_context_create is
+    * called, but we can't enable the extensions until we have a
+    * context pointer.
+    *
+    * Hello chicken.  Hello egg.  How are you two today?
+    */
+   //intelInitExtensions(NULL, GL_TRUE);
+
+   /* Allocate the private area */
+   intelScreen = CALLOC_STRUCT(intel_screen);
+   if (!intelScreen) {
+      fprintf(stderr, "\nERROR!  Allocating private area failed\n");
+      return GL_FALSE;
+   }
+   /* parse information in __driConfigOptions */
+   driParseOptionInfo(&intelScreen->optionCache,
+                      __driConfigOptions, __driNConfigOptions);
+
+   psp->private = (void *) intelScreen;
+
+   intelScreen->drmMinor = psp->drm_version.minor;
+
+   /* Determine chipset ID? */
+   if (!intel_get_param(psp, I915_PARAM_CHIPSET_ID,
+			&intelScreen->deviceID))
+      return GL_FALSE;
+
+   psp->extensions = intelScreenExtensions;
+
+   intel_be_init_device(&intelScreen->base, psp->fd, intelScreen->deviceID);
+   intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer;
+   intelScreen->base.base.get_name = intel_get_name;
+
+   return driConcatConfigs(intelFillInModes(psp, 16, 16, 0, 1),
+			   intelFillInModes(psp, 32, 24, 8, 1));
+}
+
+const struct __DriverAPIRec driDriverAPI = {
+   .InitScreen		 = intelInitScreen,
+   .DestroyScreen	 = intelDestroyScreen,
+   .CreateContext	 = intelCreateContext,
+   .DestroyContext	 = intelDestroyContext,
+   .CreateBuffer	 = intelCreateBuffer,
+   .DestroyBuffer	 = intelDestroyBuffer,
+   .SwapBuffers		 = intelSwapBuffers,
+   .MakeCurrent		 = intelMakeCurrent,
+   .UnbindContext	 = intelUnbindContext,
+   .GetSwapInfo		 = intelGetSwapInfo,
+   .GetDrawableMSC	 = driDrawableGetMSC32,
+   .WaitForMSC		 = driWaitForMSC32,
+   .CopySubBuffer	 = intelCopySubBuffer,
+
+   //.InitScreen2		 = intelInitScreen2,
+   //.HandleDrawableConfig = intelHandleDrawableConfig,
+   //.HandleBufferAttach	 = intelHandleBufferAttach,
+};
diff --git a/src/gallium/winsys/drm/intel/dri/intel_screen.h b/src/gallium/winsys/drm/intel/dri/intel_screen.h
new file mode 100644
index 0000000000..0bb43a915c
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_screen.h
@@ -0,0 +1,122 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef _INTEL_SCREEN_H_
+#define _INTEL_SCREEN_H_
+
+#include "dri_util.h"
+#include "i830_common.h"
+#include "xmlconfig.h"
+#include "ws_dri_bufpool.h"
+
+#include "pipe/p_compiler.h"
+
+#include "intel_be_device.h"
+
+struct intel_screen
+{
+   struct intel_be_device base;
+
+   struct {
+      drm_handle_t handle;
+
+      /* We create a static dri buffer for the frontbuffer.
+       */
+      struct _DriBufferObject *buffer;
+      struct pipe_surface *surface;
+      struct pipe_texture *texture;
+
+      char *map;                   /* memory map */
+      int offset;                  /* from start of video mem, in bytes */
+      int pitch;                   /* row stride, in bytes */
+      int width;
+      int height;
+      int size;
+      int cpp;                     /* for front and back buffers */
+   } front;
+
+   int deviceID;
+   int drmMinor;
+
+   drmI830Sarea *sarea;
+
+   /**
+   * Configuration cache with default values for all contexts
+   */
+   driOptionCache optionCache;
+
+   boolean havePools;
+
+   /**
+    * Temporary(?) context to use for SwapBuffers or other situations in
+    * which we need a rendering context, but none is currently bound.
+    */
+   struct intel_context *dummyContext;
+
+   /*
+    * New stuff form the i915tex integration
+    */
+   unsigned batch_id;
+
+
+   struct pipe_winsys *winsys;
+};
+
+
+
+/** cast wrapper */
+static INLINE struct intel_screen *
+intel_screen(__DRIscreenPrivate *sPriv)
+{
+   return (struct intel_screen *) sPriv->private;
+}
+
+
+extern void
+intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea);
+
+
+extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv);
+
+extern boolean intelUnbindContext(__DRIcontextPrivate * driContextPriv);
+
+extern boolean
+intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
+                 __DRIdrawablePrivate * driDrawPriv,
+                 __DRIdrawablePrivate * driReadPriv);
+
+
+extern boolean
+intelCreatePools(__DRIscreenPrivate *sPriv);
+
+extern boolean
+intelCreateContext(const __GLcontextModes * visual,
+                   __DRIcontextPrivate * driContextPriv,
+                   void *sharedContextPrivate);
+
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.c b/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.c
new file mode 100644
index 0000000000..34ad7eebe1
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.c
@@ -0,0 +1,260 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_swapbuffers.h"
+
+#include "intel_reg.h"
+
+#include "pipe/p_context.h"
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_context.h"
+#include "state_tracker/st_cb_fbo.h"
+
+#include "ws_dri_bufmgr.h"
+#include "intel_batchbuffer.h"
+
+/**
+ * Display a colorbuffer surface in an X window.
+ * Used for SwapBuffers and flushing front buffer rendering.
+ *
+ * \param dPriv  the window/drawable to display into
+ * \param surf  the surface to display
+ * \param rect  optional subrect of surface to display (may be NULL).
+ */
+void
+intelDisplaySurface(__DRIdrawablePrivate *dPriv,
+                    struct pipe_surface *surf,
+                    const drm_clip_rect_t *rect)
+{
+   struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv);
+   struct intel_context *intel = intelScreen->dummyContext;
+
+   DBG(SWAP, "%s\n", __FUNCTION__);
+
+   if (!intel) {
+      /* XXX this is where some kind of extra/meta context could be useful */
+      return;
+   }
+
+   if (intel->last_swap_fence) {
+      driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE);
+      driFenceUnReference(&intel->last_swap_fence);
+      intel->last_swap_fence = NULL;
+   }
+   intel->last_swap_fence = intel->first_swap_fence;
+   intel->first_swap_fence = NULL;
+
+   /* The LOCK_HARDWARE is required for the cliprects.  Buffer offsets
+    * should work regardless.
+    */
+   LOCK_HARDWARE(intel);
+   /* if this drawable isn't currently bound the LOCK_HARDWARE done on the
+    * current context (which is what intelScreenContext should return) might
+    * not get a contended lock and thus cliprects not updated (tests/manywin)
+    */
+   if (intel_context(dPriv->driContextPriv) != intel)
+      DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv);
+
+
+   if (dPriv && dPriv->numClipRects) {
+      const int srcWidth = surf->width;
+      const int srcHeight = surf->height;
+      const int nbox = dPriv->numClipRects;
+      const drm_clip_rect_t *pbox = dPriv->pClipRects;
+      const int pitch = intelScreen->front.pitch / intelScreen->front.cpp;
+      const int cpp = intelScreen->front.cpp;
+      const int srcpitch = surf->stride / cpp;
+      int BR13, CMD;
+      int i;
+
+      ASSERT(surf->buffer);
+
+      DBG(SWAP, "screen pitch %d  src surface pitch %d\n",
+	  pitch, surf->stride);
+
+      if (cpp == 2) {
+	 BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24);
+	 CMD = XY_SRC_COPY_BLT_CMD;
+      }
+      else {
+	 BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25);
+	 CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA |
+		XY_SRC_COPY_BLT_WRITE_RGB);
+      }
+
+      for (i = 0; i < nbox; i++, pbox++) {
+	 drm_clip_rect_t box;
+	 drm_clip_rect_t sbox;
+
+	 if (pbox->x1 > pbox->x2 ||
+	     pbox->y1 > pbox->y2 ||
+	     pbox->x2 > intelScreen->front.width ||
+	     pbox->y2 > intelScreen->front.height) {
+            /* invalid cliprect, skip it */
+	    continue;
+         }
+
+	 box = *pbox;
+
+	 if (rect) {
+            /* intersect cliprect with user-provided src rect */
+	    drm_clip_rect_t rrect;
+
+	    rrect.x1 = dPriv->x + rect->x1;
+	    rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y;
+	    rrect.x2 = rect->x2 + rrect.x1;
+	    rrect.y2 = rect->y2 + rrect.y1;
+	    if (rrect.x1 > box.x1)
+	       box.x1 = rrect.x1;
+	    if (rrect.y1 > box.y1)
+	       box.y1 = rrect.y1;
+	    if (rrect.x2 < box.x2)
+	       box.x2 = rrect.x2;
+	    if (rrect.y2 < box.y2)
+	       box.y2 = rrect.y2;
+
+	    if (box.x1 > box.x2 || box.y1 > box.y2)
+	       continue;
+	 }
+
+	 /* restrict blit to size of actually rendered area */
+	 if (box.x2 - box.x1 > srcWidth)
+	    box.x2 = srcWidth + box.x1;
+	 if (box.y2 - box.y1 > srcHeight)
+	    box.y2 = srcHeight + box.y1;
+
+	 DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n",
+	     box.x1, box.x2, box.y1, box.y2);
+
+	 sbox.x1 = box.x1 - dPriv->x;
+	 sbox.y1 = box.y1 - dPriv->y;
+
+         assert(box.x1 < box.x2);
+         assert(box.y1 < box.y2);
+
+         /* XXX this could be done with pipe->surface_copy() */
+	 /* XXX should have its own batch buffer */
+	 if (!BEGIN_BATCH(8, 2)) {
+	    /*
+	     * Since we share this batch buffer with a context
+	     * we can't flush it since that risks a GPU lockup
+	     */
+	    assert(0);
+	    continue;
+	 }
+
+	 OUT_BATCH(CMD);
+	 OUT_BATCH(BR13);
+	 OUT_BATCH((box.y1 << 16) | box.x1);
+	 OUT_BATCH((box.y2 << 16) | box.x2);
+
+	 OUT_RELOC(intelScreen->front.buffer,
+		   DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE,
+		   DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0);
+	 OUT_BATCH((sbox.y1 << 16) | sbox.x1);
+	 OUT_BATCH((srcpitch * cpp) & 0xffff);
+	 OUT_RELOC(dri_bo(surf->buffer),
+                   DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ,
+		   DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0);
+
+      }
+
+      if (intel->first_swap_fence)
+	 driFenceUnReference(&intel->first_swap_fence);
+      intel->first_swap_fence = intel_be_batchbuffer_flush(intel->base.batch);
+   }
+
+   UNLOCK_HARDWARE(intel);
+
+   if (intel->lastStamp != dPriv->lastStamp) {
+      intelUpdateWindowSize(dPriv);
+      intel->lastStamp = dPriv->lastStamp;
+   }
+}
+
+
+
+/**
+ * This will be called whenever the currently bound window is moved/resized.
+ */
+void
+intelUpdateWindowSize(__DRIdrawablePrivate *dPriv)
+{
+   struct intel_framebuffer *intelfb = intel_framebuffer(dPriv);
+   assert(intelfb->stfb);
+   st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h);
+}
+
+
+
+void
+intelSwapBuffers(__DRIdrawablePrivate * dPriv)
+{
+   struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv);
+   struct pipe_surface *back_surf;
+
+   assert(intel_fb);
+   assert(intel_fb->stfb);
+
+   back_surf = st_get_framebuffer_surface(intel_fb->stfb,
+                                          ST_SURFACE_BACK_LEFT);
+   if (back_surf) {
+      st_notify_swapbuffers(intel_fb->stfb);
+      intelDisplaySurface(dPriv, back_surf, NULL);
+      st_notify_swapbuffers_complete(intel_fb->stfb);
+   }
+}
+
+
+/**
+ * Called via glXCopySubBufferMESA() to copy a subrect of the back
+ * buffer to the front buffer/screen.
+ */
+void
+intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h)
+{
+   struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv);
+   struct pipe_surface *back_surf;
+
+   assert(intel_fb);
+   assert(intel_fb->stfb);
+
+   back_surf = st_get_framebuffer_surface(intel_fb->stfb,
+                                          ST_SURFACE_BACK_LEFT);
+   if (back_surf) {
+      drm_clip_rect_t rect;
+      rect.x1 = x;
+      rect.y1 = y;
+      rect.x2 = w;
+      rect.y2 = h;
+
+      st_notify_swapbuffers(intel_fb->stfb);
+      intelDisplaySurface(dPriv, back_surf, &rect);
+   }
+}
diff --git a/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.h b/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.h
new file mode 100644
index 0000000000..46c9bab3af
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.h
@@ -0,0 +1,47 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef INTEL_SWAPBUFFERS_H
+#define INTEL_SWAPBUFFERS_H
+
+
+struct pipe_surface;
+
+
+extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv,
+                                struct pipe_surface *surf,
+                                const drm_clip_rect_t * rect);
+
+extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv);
+
+extern void intelCopySubBuffer(__DRIdrawablePrivate * dPriv,
+                               int x, int y, int w, int h);
+
+extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv);
+
+
+#endif /* INTEL_SWAPBUFFERS_H */
diff --git a/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c
new file mode 100644
index 0000000000..20920a2052
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c
@@ -0,0 +1,82 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+/*
+ * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com>
+ */
+
+#include "intel_context.h"
+#include "intel_winsys_softpipe.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_format.h"
+#include "util/u_memory.h"
+#include "softpipe/sp_winsys.h"
+
+
+struct intel_softpipe_winsys {
+   struct softpipe_winsys sws;
+   struct intel_context *intel;
+};
+
+/**
+ * Return list of surface formats supported by this driver.
+ */
+static boolean
+intel_is_format_supported(struct softpipe_winsys *sws,
+                          enum pipe_format format)
+{
+   switch(format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+   case PIPE_FORMAT_R5G6B5_UNORM:
+   case PIPE_FORMAT_S8Z24_UNORM:
+      return TRUE;
+   default:
+      return FALSE;
+   }
+}
+
+
+/**
+ * Create rendering context which uses software rendering.
+ */
+struct pipe_context *
+intel_create_softpipe( struct intel_context *intel,
+                       struct pipe_winsys *winsys )
+{
+   struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys );
+   struct pipe_screen *screen = softpipe_create_screen(winsys);
+
+   /* Fill in this struct with callbacks that softpipe will need to
+    * communicate with the window system, buffer manager, etc.
+    */
+   isws->sws.is_format_supported = intel_is_format_supported;
+   isws->intel = intel;
+
+   /* Create the softpipe context:
+    */
+   return softpipe_create( screen, winsys, &isws->sws );
+}
diff --git a/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.h b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.h
new file mode 100644
index 0000000000..5fa14cb749
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ *
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef INTEL_SOFTPIPE_H
+#define INTEL_SOFTPIPE_H
+
+struct pipe_winsys;
+struct pipe_context;
+struct intel_context;
+
+struct pipe_context *
+intel_create_softpipe( struct intel_context *intel,
+                       struct pipe_winsys *winsys );
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/dri/server/i830_common.h b/src/gallium/winsys/drm/intel/dri/server/i830_common.h
new file mode 100644
index 0000000000..3452ddb3c9
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/server/i830_common.h
@@ -0,0 +1,255 @@
+/**************************************************************************
+
+Copyright 2001 VA Linux Systems Inc., Fremont, California.
+Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
+
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, and/or sell copies of the Software, and to permit persons to whom
+the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice (including the next
+paragraph) shall be included in all copies or substantial portions of the
+Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+
+#ifndef _I830_COMMON_H_
+#define _I830_COMMON_H_
+
+
+#define I830_NR_TEX_REGIONS 255	/* maximum due to use of chars for next/prev */
+#define I830_LOG_MIN_TEX_REGION_SIZE 14
+
+
+/* Driver specific DRM command indices
+ * NOTE: these are not OS specific, but they are driver specific
+ */
+#define DRM_I830_INIT                     0x00
+#define DRM_I830_FLUSH                    0x01
+#define DRM_I830_FLIP                     0x02
+#define DRM_I830_BATCHBUFFER              0x03
+#define DRM_I830_IRQ_EMIT                 0x04
+#define DRM_I830_IRQ_WAIT                 0x05
+#define DRM_I830_GETPARAM                 0x06
+#define DRM_I830_SETPARAM                 0x07
+#define DRM_I830_ALLOC                    0x08
+#define DRM_I830_FREE                     0x09
+#define DRM_I830_INIT_HEAP                0x0a
+#define DRM_I830_CMDBUFFER                0x0b
+#define DRM_I830_DESTROY_HEAP             0x0c
+#define DRM_I830_SET_VBLANK_PIPE          0x0d
+#define DRM_I830_GET_VBLANK_PIPE          0x0e
+#define DRM_I830_MMIO		       	  0x10
+
+typedef struct {
+   enum {
+      I830_INIT_DMA = 0x01,
+      I830_CLEANUP_DMA = 0x02,
+      I830_RESUME_DMA = 0x03
+   } func;
+   unsigned int mmio_offset;
+   int sarea_priv_offset;
+   unsigned int ring_start;
+   unsigned int ring_end;
+   unsigned int ring_size;
+   unsigned int front_offset;
+   unsigned int back_offset;
+   unsigned int depth_offset;
+   unsigned int w;
+   unsigned int h;
+   unsigned int pitch;
+   unsigned int pitch_bits;
+   unsigned int back_pitch;
+   unsigned int depth_pitch;
+   unsigned int cpp;
+   unsigned int chipset;
+} drmI830Init;
+
+typedef struct {
+	drmTextureRegion texList[I830_NR_TEX_REGIONS+1];
+        int last_upload;	/* last time texture was uploaded */
+        int last_enqueue;	/* last time a buffer was enqueued */
+	int last_dispatch;	/* age of the most recently dispatched buffer */
+	int ctxOwner;		/* last context to upload state */
+	/** Last context that used the buffer manager. */
+	int texAge;
+        int pf_enabled;		/* is pageflipping allowed? */
+        int pf_active;
+        int pf_current_page;	/* which buffer is being displayed? */
+        int perf_boxes;	        /* performance boxes to be displayed */
+	int width, height;      /* screen size in pixels */
+
+	drm_handle_t front_handle;
+	int front_offset;
+	int front_size;
+
+	drm_handle_t back_handle;
+	int back_offset;
+	int back_size;
+
+	drm_handle_t depth_handle;
+	int depth_offset;
+	int depth_size;
+
+	drm_handle_t tex_handle;
+	int tex_offset;
+	int tex_size;
+	int log_tex_granularity;
+	int pitch;
+	int rotation;           /* 0, 90, 180 or 270 */
+	int rotated_offset;
+	int rotated_size;
+	int rotated_pitch;
+	int virtualX, virtualY;
+
+	unsigned int front_tiled;
+	unsigned int back_tiled;
+	unsigned int depth_tiled;
+	unsigned int rotated_tiled;
+	unsigned int rotated2_tiled;
+
+	int planeA_x;
+	int planeA_y;
+	int planeA_w;
+	int planeA_h;
+	int planeB_x;
+	int planeB_y;
+	int planeB_w;
+	int planeB_h;
+
+	/* Triple buffering */
+	drm_handle_t third_handle;
+	int third_offset;
+	int third_size;
+	unsigned int third_tiled;
+
+	/* buffer object handles for the static buffers.  May change
+	 * over the lifetime of the client, though it doesn't in our current
+	 * implementation.
+	 */
+	unsigned int front_bo_handle;
+	unsigned int back_bo_handle;
+	unsigned int third_bo_handle;
+	unsigned int depth_bo_handle;
+} drmI830Sarea;
+
+/* Flags for perf_boxes
+ */
+#define I830_BOX_RING_EMPTY    0x1 /* populated by kernel */
+#define I830_BOX_FLIP          0x2 /* populated by kernel */
+#define I830_BOX_WAIT          0x4 /* populated by kernel & client */
+#define I830_BOX_TEXTURE_LOAD  0x8 /* populated by kernel */
+#define I830_BOX_LOST_CONTEXT  0x10 /* populated by client */
+
+
+typedef struct {
+   	int start;		/* agp offset */
+	int used;		/* nr bytes in use */
+	int DR1;		/* hw flags for GFX_OP_DRAWRECT_INFO */
+        int DR4;		/* window origin for GFX_OP_DRAWRECT_INFO*/
+	int num_cliprects;	/* mulitpass with multiple cliprects? */
+        drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */
+} drmI830BatchBuffer;
+
+typedef struct {
+   	char *buf;		/* agp offset */
+	int sz; 		/* nr bytes in use */
+	int DR1;		/* hw flags for GFX_OP_DRAWRECT_INFO */
+        int DR4;		/* window origin for GFX_OP_DRAWRECT_INFO*/
+	int num_cliprects;	/* mulitpass with multiple cliprects? */
+        drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */
+} drmI830CmdBuffer;
+
+typedef struct {
+	int *irq_seq;
+} drmI830IrqEmit;
+
+typedef struct {
+	int irq_seq;
+} drmI830IrqWait;
+
+typedef struct {
+	int param;
+	int *value;
+} drmI830GetParam;
+
+#define I830_PARAM_IRQ_ACTIVE     1
+#define I830_PARAM_ALLOW_BATCHBUFFER   2
+
+typedef struct {
+	int param;
+	int value;
+} drmI830SetParam;
+
+#define I830_SETPARAM_USE_MI_BATCHBUFFER_START  1
+#define I830_SETPARAM_TEX_LRU_LOG_GRANULARITY   2
+#define I830_SETPARAM_ALLOW_BATCHBUFFER         3
+
+
+/* A memory manager for regions of shared memory:
+ */
+#define I830_MEM_REGION_AGP 1
+
+typedef struct {
+	int region;
+	int alignment;
+	int size;
+	int *region_offset;	/* offset from start of fb or agp */
+} drmI830MemAlloc;
+
+typedef struct {
+	int region;
+	int region_offset;
+} drmI830MemFree;
+
+typedef struct {
+	int region;
+	int size;
+	int start;	
+} drmI830MemInitHeap;
+
+typedef struct {
+	int region;
+} drmI830MemDestroyHeap;
+
+#define DRM_I830_VBLANK_PIPE_A  1
+#define DRM_I830_VBLANK_PIPE_B  2
+
+typedef struct {
+        int pipe;
+} drmI830VBlankPipe;
+
+#define MMIO_READ  0
+#define MMIO_WRITE 1
+
+#define MMIO_REGS_IA_PRIMATIVES_COUNT           0
+#define MMIO_REGS_IA_VERTICES_COUNT             1
+#define MMIO_REGS_VS_INVOCATION_COUNT           2
+#define MMIO_REGS_GS_PRIMITIVES_COUNT           3
+#define MMIO_REGS_GS_INVOCATION_COUNT           4
+#define MMIO_REGS_CL_PRIMITIVES_COUNT           5
+#define MMIO_REGS_CL_INVOCATION_COUNT           6
+#define MMIO_REGS_PS_INVOCATION_COUNT           7
+#define MMIO_REGS_PS_DEPTH_COUNT                8
+
+typedef struct {
+        unsigned int read_write:1;
+        unsigned int reg:31;
+        void __user *data;
+} drmI830MMIO;
+
+#endif /* _I830_DRM_H_ */
diff --git a/src/gallium/winsys/drm/intel/dri/server/i830_dri.h b/src/gallium/winsys/drm/intel/dri/server/i830_dri.h
new file mode 100644
index 0000000000..0d514b6c38
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/dri/server/i830_dri.h
@@ -0,0 +1,62 @@
+
+#ifndef _I830_DRI_H
+#define _I830_DRI_H
+
+#include "xf86drm.h"
+#include "i830_common.h"
+
+#define I830_MAX_DRAWABLES 256
+
+#define I830_MAJOR_VERSION 1
+#define I830_MINOR_VERSION 7
+#define I830_PATCHLEVEL 2
+
+#define I830_REG_SIZE 0x80000
+
+typedef struct _I830DRIRec {
+   drm_handle_t regs;
+   drmSize regsSize;
+
+   drmSize unused1; /* backbufferSize */
+   drm_handle_t unused2; /* backbuffer */
+
+   drmSize unused3; /* depthbufferSize */
+   drm_handle_t unused4; /* depthbuffer */
+
+   drmSize unused5; /* rotatedSize */
+   drm_handle_t unused6; /* rotatedbuffer */
+
+   drm_handle_t unused7; /* textures */
+   int unused8; /* textureSize */
+
+   drm_handle_t unused9; /* agp_buffers */
+   drmSize unused10; /* agp_buf_size */
+
+   int deviceID;
+   int width;
+   int height;
+   int mem;
+   int cpp;
+   int bitsPerPixel;
+
+   int unused11[8]; /* was front/back/depth/rotated offset/pitch */
+
+   int unused12; /* logTextureGranularity */
+   int unused13; /* textureOffset */
+
+   int irq;
+   int sarea_priv_offset;
+} I830DRIRec, *I830DRIPtr;
+
+typedef struct {
+   /* Nothing here yet */
+   int dummy;
+} I830ConfigPrivRec, *I830ConfigPrivPtr;
+
+typedef struct {
+   /* Nothing here yet */
+   int dummy;
+} I830DRIContextRec, *I830DRIContextPtr;
+
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/egl/Makefile b/src/gallium/winsys/drm/intel/egl/Makefile
new file mode 100644
index 0000000000..7147d89e0d
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/egl/Makefile
@@ -0,0 +1,31 @@
+TOP = ../../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = EGL_i915.so
+
+PIPE_DRIVERS = \
+	../gem/libinteldrm.a \
+	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/i915simple/libi915simple.a \
+	$(TOP)/src/gallium/state_trackers/egl/libegldrm.a \
+
+
+DRIVER_SOURCES = \
+	intel_context.c \
+	intel_device.c \
+	intel_api.c
+
+C_SOURCES = \
+	$(COMMON_GALLIUM_SOURCES) \
+	$(DRIVER_SOURCES)
+
+DRIVER_EXTRAS = -ldrm_intel
+
+ASM_SOURCES = 
+
+DRIVER_DEFINES = -I../gem $(shell pkg-config libdrm --atleast-version=2.3.1 \
+				&& echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP")
+
+include ../../Makefile.template
+
+symlinks:
diff --git a/src/gallium/winsys/drm/intel/egl/intel_api.c b/src/gallium/winsys/drm/intel/egl/intel_api.c
new file mode 100644
index 0000000000..5dc4a7b052
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/egl/intel_api.c
@@ -0,0 +1,10 @@
+
+#include "intel_api.h"
+
+struct drm_api drm_api_hocks =
+{
+	.create_screen = intel_create_screen,
+	.create_context = intel_create_context,
+	.buffer_from_handle = intel_be_buffer_from_handle,
+	.handle_from_buffer = intel_be_handle_from_buffer,
+};
diff --git a/src/gallium/winsys/drm/intel/egl/intel_api.h b/src/gallium/winsys/drm/intel/egl/intel_api.h
new file mode 100644
index 0000000000..8ec165ab01
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/egl/intel_api.h
@@ -0,0 +1,14 @@
+
+#ifndef _INTEL_API_H_
+#define _INTEL_API_H_
+
+#include "pipe/p_compiler.h"
+
+#include "state_tracker/drm_api.h"
+
+#include "intel_be_device.h"
+
+struct pipe_screen *intel_create_screen(int drmFD, int pciID);
+struct pipe_context *intel_create_context(struct pipe_screen *screen);
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/egl/intel_context.c b/src/gallium/winsys/drm/intel/egl/intel_context.c
new file mode 100644
index 0000000000..57e5ff7bc1
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/egl/intel_context.c
@@ -0,0 +1,83 @@
+
+#include "i915simple/i915_screen.h"
+
+#include "intel_be_device.h"
+#include "intel_be_context.h"
+
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+
+#include "intel_api.h"
+
+struct intel_context
+{
+	struct intel_be_context base;
+
+	/* stuff */
+};
+
+/*
+ * Hardware lock functions.
+ * Doesn't do anything in EGL
+ */
+
+static void
+intel_lock_hardware(struct intel_be_context *context)
+{
+	(void)context;
+}
+
+static void
+intel_unlock_hardware(struct intel_be_context *context)
+{
+	(void)context;
+}
+
+static boolean
+intel_locked_hardware(struct intel_be_context *context)
+{
+	(void)context;
+	return FALSE;
+}
+
+
+/*
+ * Misc functions.
+ */
+static void
+intel_destroy_be_context(struct i915_winsys *winsys)
+{
+	struct intel_context *intel = (struct intel_context *)winsys;
+
+	intel_be_destroy_context(&intel->base);
+	free(intel);
+}
+
+struct pipe_context *
+intel_create_context(struct pipe_screen *screen)
+{
+	struct intel_context *intel;
+	struct pipe_context *pipe;
+	struct intel_be_device *device = (struct intel_be_device *)screen->winsys;
+
+	intel = (struct intel_context *)malloc(sizeof(*intel));
+	memset(intel, 0, sizeof(*intel));
+
+	intel->base.hardware_lock = intel_lock_hardware;
+	intel->base.hardware_unlock = intel_unlock_hardware;
+	intel->base.hardware_locked = intel_locked_hardware;
+
+	intel_be_init_context(&intel->base, device);
+
+	intel->base.base.destroy = intel_destroy_be_context;
+
+#if 0
+	pipe = intel_create_softpipe(intel, screen->winsys);
+#else
+	pipe = i915_create_context(screen, &device->base, &intel->base.base);
+#endif
+
+	pipe->priv = intel;
+
+	return pipe;
+}
diff --git a/src/gallium/winsys/drm/intel/egl/intel_device.c b/src/gallium/winsys/drm/intel/egl/intel_device.c
new file mode 100644
index 0000000000..6b281402d5
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/egl/intel_device.c
@@ -0,0 +1,48 @@
+
+#include <stdio.h>
+#include "pipe/p_defines.h"
+#include "intel_be_device.h"
+#include "i915simple/i915_screen.h"
+
+#include "intel_api.h"
+
+struct intel_device
+{
+	struct intel_be_device base;
+
+	int deviceID;
+};
+
+static void
+intel_destroy_winsys(struct pipe_winsys *winsys)
+{
+	struct intel_device *dev = (struct intel_device *)winsys;
+
+	intel_be_destroy_device(&dev->base);
+
+	free(dev);
+}
+
+struct pipe_screen *
+intel_create_screen(int drmFD, int deviceID)
+{
+	struct intel_device *dev;
+	struct pipe_screen *screen;
+
+	/* Allocate the private area */
+	dev = malloc(sizeof(*dev));
+	if (!dev)
+		return NULL;
+	memset(dev, 0, sizeof(*dev));
+
+	dev->deviceID = deviceID;
+
+	intel_be_init_device(&dev->base, drmFD, deviceID);
+
+	/* we need to hock our own destroy function in here */
+	dev->base.base.destroy = intel_destroy_winsys;
+
+	screen = i915_create_screen(&dev->base.base, deviceID);
+
+	return screen;
+}
diff --git a/src/gallium/winsys/drm/intel/gem/Makefile b/src/gallium/winsys/drm/intel/gem/Makefile
new file mode 100644
index 0000000000..b25fc258f4
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/Makefile
@@ -0,0 +1,18 @@
+TOP = ../../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = inteldrm
+
+C_SOURCES = \
+	intel_be_batchbuffer.c \
+	intel_be_context.c \
+	intel_be_device.c
+
+
+include ./Makefile.template
+
+DRIVER_DEFINES = $(shell pkg-config libdrm --cflags \
+                && pkg-config libdrm --atleast-version=2.3.1 \
+                && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP")
+symlinks:
+
diff --git a/src/gallium/winsys/drm/intel/gem/Makefile.template b/src/gallium/winsys/drm/intel/gem/Makefile.template
new file mode 100644
index 0000000000..b60e978894
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/Makefile.template
@@ -0,0 +1,64 @@
+# -*-makefile-*-
+
+
+# We still have a dependency on the "dri" buffer manager.  Most likely
+# the interface can be reused in non-dri environments, and also as a
+# frontend to simpler memory managers.
+#
+COMMON_SOURCES =
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+	$(CPP_SOURCES:.cpp=.o) \
+	$(ASM_SOURCES:.S=.o)
+
+
+### Include directories
+INCLUDES = \
+	-I. \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/include \
+	$(DRIVER_INCLUDES)
+
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.cpp.o:
+	$(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES)  $< -o $@
+
+
+##### TARGETS #####
+
+default: depend symlinks lib$(LIBNAME).a
+
+
+lib$(LIBNAME).a: $(OBJECTS) Makefile Makefile.template
+	$(TOP)/bin/mklib -o $(LIBNAME) -static $(OBJECTS) $(DRIVER_LIBS)
+
+
+depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \
+		$(ASM_SOURCES) 2> /dev/null
+
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` `find ../include`
+
+
+# Remove .o and backup files
+clean::
+	-rm -f *.o */*.o *~ *.so *.a *~ server/*.o $(SYMLINKS)
+	-rm -f depend depend.bak
+
+
+include depend
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c
new file mode 100644
index 0000000000..d9556e1f38
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.c
@@ -0,0 +1,140 @@
+
+#include "i915simple/i915_debug.h"
+#include "intel_be_batchbuffer.h"
+#include "intel_be_context.h"
+#include "intel_be_device.h"
+#include "intel_be_fence.h"
+#include <errno.h>
+
+#include "util/u_memory.h"
+
+struct intel_be_batchbuffer *
+intel_be_batchbuffer_alloc(struct intel_be_context *intel)
+{
+	struct intel_be_batchbuffer *batch = CALLOC_STRUCT(intel_be_batchbuffer);
+
+
+	batch->base.buffer = NULL;
+	batch->base.winsys = &intel->base;
+	batch->base.map = NULL;
+	batch->base.ptr = NULL;
+	batch->base.size = 0;
+	batch->base.actual_size = intel->device->max_batch_size;
+	batch->base.relocs = 0;
+	batch->base.max_relocs = INTEL_DEFAULT_RELOCS;
+
+	batch->base.map = malloc(batch->base.actual_size);
+	memset(batch->base.map, 0, batch->base.actual_size);
+
+	batch->base.ptr = batch->base.map;
+
+	intel_be_batchbuffer_reset(batch);
+
+	return batch;
+}
+
+void
+intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch)
+{
+	struct intel_be_context *intel = intel_be_context(batch->base.winsys);
+	struct intel_be_device *dev = intel->device;
+
+	if (batch->bo)
+		drm_intel_bo_unreference(batch->bo);
+
+	memset(batch->base.map, 0, batch->base.actual_size);
+	batch->base.ptr = batch->base.map;
+	batch->base.size = batch->base.actual_size - BATCH_RESERVED;
+
+	batch->base.relocs = 0;
+	batch->base.max_relocs = INTEL_DEFAULT_RELOCS;
+
+	batch->bo = drm_intel_bo_alloc(dev->pools.gem,
+	                               "gallium3d_batch_buffer",
+	                               batch->base.actual_size, 0);
+}
+
+int
+intel_be_offset_relocation(struct intel_be_batchbuffer *batch,
+			   unsigned pre_add,
+			   drm_intel_bo *bo,
+			   uint32_t read_domains,
+			   uint32_t write_domain)
+{
+	unsigned offset;
+	int ret = 0;
+
+	assert(batch->base.relocs < batch->base.max_relocs);
+
+	offset = (unsigned)(batch->base.ptr - batch->base.map);
+
+	ret = drm_intel_bo_emit_reloc(batch->bo, offset,
+	                              bo, pre_add,
+	                              read_domains,
+	                              write_domain);
+
+	((uint32_t*)batch->base.ptr)[0] = bo->offset + pre_add;
+	batch->base.ptr += 4;
+
+	if (!ret)
+		batch->base.relocs++;
+
+	return ret;
+}
+
+void
+intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch,
+			   struct intel_be_fence **fence)
+{
+	struct i915_batchbuffer *i915 = &batch->base;
+	unsigned used = 0;
+	int ret = 0;
+
+	assert(i915_batchbuffer_space(i915) >= 0);
+
+	used = batch->base.ptr - batch->base.map;
+	assert((used & 3) == 0);
+
+	if (used & 4) {
+		i915_batchbuffer_dword(i915, (0x0<<29)|(0x4<<23)|(1<<0)); // MI_FLUSH | FLUSH_MAP_CACHE;
+		i915_batchbuffer_dword(i915, (0x0<<29)|(0x0<<23)); // MI_NOOP
+		i915_batchbuffer_dword(i915, (0x0<<29)|(0xA<<23)); // MI_BATCH_BUFFER_END;
+	} else {
+		i915_batchbuffer_dword(i915, (0x0<<29)|(0x4<<23)|(1<<0)); //MI_FLUSH | FLUSH_MAP_CACHE;
+		i915_batchbuffer_dword(i915, (0x0<<29)|(0xA<<23)); // MI_BATCH_BUFFER_END;
+	}
+
+	used = batch->base.ptr - batch->base.map;
+
+	drm_intel_bo_subdata(batch->bo, 0, used, batch->base.map);
+	ret = drm_intel_bo_exec(batch->bo, used, NULL, 0, 0);
+
+	assert(ret == 0);
+
+	intel_be_batchbuffer_reset(batch);
+
+	if (fence) {
+		if (*fence)
+			intel_be_fence_unreference(*fence);
+
+		(*fence) = CALLOC_STRUCT(intel_be_fence);
+		(*fence)->refcount = 1;
+		(*fence)->bo = NULL;
+	}
+}
+
+void
+intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch)
+{
+
+}
+
+void
+intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch)
+{
+	if (batch->bo)
+		drm_intel_bo_unreference(batch->bo);
+
+	free(batch->base.map);
+	free(batch);
+}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.h b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.h
new file mode 100644
index 0000000000..195bf8dee7
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_batchbuffer.h
@@ -0,0 +1,55 @@
+
+#ifndef INTEL_BE_BATCHBUFFER_H
+#define INTEL_BE_BATCHBUFFER_H
+
+#include "i915simple/i915_batch.h"
+
+#include "drm.h"
+#include "intel_bufmgr.h"
+
+#define BATCH_RESERVED 16
+
+#define INTEL_DEFAULT_RELOCS 100
+#define INTEL_MAX_RELOCS 400
+
+#define INTEL_BATCH_NO_CLIPRECTS 0x1
+#define INTEL_BATCH_CLIPRECTS    0x2
+
+struct intel_be_context;
+struct intel_be_device;
+struct intel_be_fence;
+
+struct intel_be_batchbuffer
+{
+	struct i915_batchbuffer base;
+
+	struct intel_be_context *intel;
+	struct intel_be_device *device;
+
+	drm_intel_bo *bo;
+};
+
+struct intel_be_batchbuffer *
+intel_be_batchbuffer_alloc(struct intel_be_context *intel);
+
+void
+intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch);
+
+void
+intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch);
+
+void
+intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch,
+			   struct intel_be_fence **fence);
+
+void
+intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch);
+
+int
+intel_be_offset_relocation(struct intel_be_batchbuffer *batch,
+			   unsigned pre_add,
+			   drm_intel_bo *bo,
+			   uint32_t read_domains,
+			   uint32_t write_doman);
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.c b/src/gallium/winsys/drm/intel/gem/intel_be_context.c
new file mode 100644
index 0000000000..95e761d78d
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_context.c
@@ -0,0 +1,80 @@
+
+#include "intel_be_device.h"
+#include "intel_be_context.h"
+#include "intel_be_batchbuffer.h"
+
+#include "i915_drm.h"
+
+static struct i915_batchbuffer *
+intel_be_batch_get(struct i915_winsys *sws)
+{
+	struct intel_be_context *intel = intel_be_context(sws);
+	return &intel->batch->base;
+}
+
+static void
+intel_be_batch_reloc(struct i915_winsys *sws,
+		     struct pipe_buffer *buf,
+		     unsigned access_flags,
+		     unsigned delta)
+{
+	struct intel_be_context *intel = intel_be_context(sws);
+	drm_intel_bo *bo = intel_bo(buf);
+	int ret;
+	uint32_t read = 0;
+	uint32_t write = 0;
+
+	if (access_flags & I915_BUFFER_ACCESS_WRITE) {
+		write = I915_GEM_DOMAIN_RENDER;
+		read = I915_GEM_DOMAIN_RENDER;
+	}
+
+	if (access_flags & I915_BUFFER_ACCESS_READ) {
+		read |= I915_GEM_DOMAIN_VERTEX;
+	}
+
+	ret = intel_be_offset_relocation(intel->batch,
+	                                 delta,
+	                                 bo,
+	                                 read,
+	                                 write);
+    assert(ret == 0);
+
+	/* TODO change return type */
+	/* return ret; */
+}
+
+static void
+intel_be_batch_flush(struct i915_winsys *sws,
+		     struct pipe_fence_handle **fence)
+{
+	struct intel_be_context *intel = intel_be_context(sws);
+	struct intel_be_fence **f = (struct intel_be_fence **)fence;
+
+	if (fence && *fence)
+		assert(0);
+
+	intel_be_batchbuffer_flush(intel->batch, f);
+}
+
+boolean
+intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *device)
+{
+	assert(intel);
+	assert(device);
+	intel->device = device;
+
+	intel->base.batch_get = intel_be_batch_get;
+	intel->base.batch_reloc = intel_be_batch_reloc;
+	intel->base.batch_flush = intel_be_batch_flush;
+
+	intel->batch = intel_be_batchbuffer_alloc(intel);
+
+	return true;
+}
+
+void
+intel_be_destroy_context(struct intel_be_context *intel)
+{
+	intel_be_batchbuffer_free(intel->batch);
+}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_context.h b/src/gallium/winsys/drm/intel/gem/intel_be_context.h
new file mode 100644
index 0000000000..9cee1a4e52
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_context.h
@@ -0,0 +1,48 @@
+
+#ifndef INTEL_BE_CONTEXT_H
+#define INTEL_BE_CONTEXT_H
+
+#include "i915simple/i915_winsys.h"
+
+struct intel_be_context
+{
+	/** Interface to i915simple driver */
+	struct i915_winsys base;
+
+	struct intel_be_device *device;
+	struct intel_be_batchbuffer *batch;
+
+	/*
+	 * Hardware lock functions.
+	 *
+	 * Needs to be filled in by the winsys.
+	 */
+	void (*hardware_lock)(struct intel_be_context *context);
+	void (*hardware_unlock)(struct intel_be_context *context);
+	boolean (*hardware_locked)(struct intel_be_context *context);
+};
+
+static INLINE struct intel_be_context *
+intel_be_context(struct i915_winsys *sws)
+{
+	return (struct intel_be_context *)sws;
+}
+
+/**
+ * Intialize a allocated intel_be_context struct.
+ *
+ * Remember to set the hardware_* functions.
+ */
+boolean
+intel_be_init_context(struct intel_be_context *intel,
+		      struct intel_be_device *device);
+
+/**
+ * Destroy a intel_be_context.
+ *
+ * Does not free the struct that is up to the winsys.
+ */
+void
+intel_be_destroy_context(struct intel_be_context *intel);
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.c b/src/gallium/winsys/drm/intel/gem/intel_be_device.c
new file mode 100644
index 0000000000..82c1cb2f32
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.c
@@ -0,0 +1,265 @@
+
+#include "intel_be_device.h"
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "util/u_memory.h"
+
+#include "intel_be_fence.h"
+
+#include "i915simple/i915_screen.h"
+
+
+/**
+ *  Turn a pipe winsys into an intel/pipe winsys:
+ */
+static INLINE struct intel_be_device *
+intel_be_device(struct pipe_winsys *winsys)
+{
+	return (struct intel_be_device *)winsys;
+}
+
+/*
+ * Buffer
+ */
+
+static void *
+intel_be_buffer_map(struct pipe_winsys *winsys,
+		    struct pipe_buffer *buf,
+		    unsigned flags)
+{
+	drm_intel_bo *bo = intel_bo(buf);
+	int write = 0;
+	int ret;
+
+	if (flags & PIPE_BUFFER_USAGE_CPU_WRITE)
+		write = 1;
+
+	ret = drm_intel_bo_map(bo, write);
+
+	if (ret)
+		return NULL;
+
+	return bo->virtual;
+}
+
+static void
+intel_be_buffer_unmap(struct pipe_winsys *winsys,
+		      struct pipe_buffer *buf)
+{
+	drm_intel_bo_unmap(intel_bo(buf));
+}
+
+static void
+intel_be_buffer_destroy(struct pipe_winsys *winsys,
+			struct pipe_buffer *buf)
+{
+	drm_intel_bo_unreference(intel_bo(buf));
+	free(buf);
+}
+
+static struct pipe_buffer *
+intel_be_buffer_create(struct pipe_winsys *winsys,
+		       unsigned alignment,
+		       unsigned usage,
+		       unsigned size)
+{
+	struct intel_be_buffer *buffer = CALLOC_STRUCT(intel_be_buffer);
+	struct intel_be_device *dev = intel_be_device(winsys);
+	drm_intel_bufmgr *pool;
+	char *name;
+
+	if (!buffer)
+		return NULL;
+
+	buffer->base.refcount = 1;
+	buffer->base.alignment = alignment;
+	buffer->base.usage = usage;
+	buffer->base.size = size;
+
+	if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) {
+		/* Local buffer */
+		name = "gallium3d_local";
+		pool = dev->pools.gem;
+	} else if (usage & PIPE_BUFFER_USAGE_CUSTOM) {
+		/* For vertex buffers */
+		name = "gallium3d_internal_vertex";
+		pool = dev->pools.gem;
+	} else {
+		/* Regular buffers */
+		name = "gallium3d_regular";
+		pool = dev->pools.gem;
+	}
+
+	buffer->bo = drm_intel_bo_alloc(pool, name, size, alignment);
+
+	if (!buffer->bo)
+		goto err;
+
+	return &buffer->base;
+
+err:
+	free(buffer);
+	return NULL;
+}
+
+static struct pipe_buffer *
+intel_be_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes)
+{
+	struct intel_be_buffer *buffer = CALLOC_STRUCT(intel_be_buffer);
+	struct intel_be_device *dev = intel_be_device(winsys);
+	int ret;
+
+	if (!buffer)
+		return NULL;
+
+	buffer->base.refcount = 1;
+	buffer->base.alignment = 0;
+	buffer->base.usage = 0;
+	buffer->base.size = bytes;
+
+	buffer->bo = drm_intel_bo_alloc(dev->pools.gem,
+	                                "gallium3d_user_buffer",
+	                                bytes, 0);
+
+	if (!buffer->bo)
+		goto err;
+
+	ret = drm_intel_bo_subdata(buffer->bo,
+	                           0, bytes, ptr);
+
+	if (ret)
+		goto err;
+
+	return &buffer->base;
+
+err:
+	free(buffer);
+	return NULL;
+}
+
+struct pipe_buffer *
+intel_be_buffer_from_handle(struct pipe_winsys *winsys,
+                            const char* name, unsigned handle)
+{
+	struct intel_be_device *dev = intel_be_device(winsys);
+	struct intel_be_buffer *buffer = CALLOC_STRUCT(intel_be_buffer);
+
+	if (!buffer)
+		return NULL;
+
+	buffer->bo = drm_intel_bo_gem_create_from_name(dev->pools.gem, name, handle);
+
+	if (!buffer->bo)
+		goto err;
+
+	buffer->base.refcount = 1;
+	buffer->base.alignment = buffer->bo->align;
+	buffer->base.usage = PIPE_BUFFER_USAGE_GPU_READ |
+	                     PIPE_BUFFER_USAGE_GPU_WRITE |
+	                     PIPE_BUFFER_USAGE_CPU_READ |
+	                     PIPE_BUFFER_USAGE_CPU_WRITE;
+	buffer->base.size = buffer->bo->size;
+
+	return &buffer->base;
+
+err:
+	free(buffer);
+	return NULL;
+}
+
+unsigned
+intel_be_handle_from_buffer(struct pipe_winsys *winsys,
+                            struct pipe_buffer *buf)
+{
+	drm_intel_bo *bo = intel_bo(buf);
+	return bo->handle;
+}
+
+/*
+ * Fence
+ */
+
+static void
+intel_be_fence_refunref(struct pipe_winsys *sws,
+			 struct pipe_fence_handle **ptr,
+			 struct pipe_fence_handle *fence)
+{
+	struct intel_be_fence **p = (struct intel_be_fence **)ptr;
+	struct intel_be_fence *f = (struct intel_be_fence *)fence;
+
+	assert(p);
+
+	if (f)
+		intel_be_fence_reference(f);
+
+	if (*p)
+		intel_be_fence_unreference(*p);
+
+	*p = f;
+}
+
+static int
+intel_be_fence_signalled(struct pipe_winsys *sws,
+			 struct pipe_fence_handle *fence,
+			 unsigned flag)
+{
+	assert(0);
+
+	return 0;
+}
+
+static int
+intel_be_fence_finish(struct pipe_winsys *sws,
+		      struct pipe_fence_handle *fence,
+		      unsigned flag)
+{
+	struct intel_be_fence *f = (struct intel_be_fence *)fence;
+
+	/* fence already expired */
+	if (!f->bo)
+		return 0;
+
+	drm_intel_bo_wait_rendering(f->bo);
+	drm_intel_bo_unreference(f->bo);
+	f->bo = NULL;
+
+	return 0;
+}
+
+/*
+ * Misc functions
+ */
+
+boolean
+intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id)
+{
+	dev->fd = fd;
+	dev->max_batch_size = 16 * 4096;
+	dev->max_vertex_size = 128 * 4096;
+
+	dev->base.buffer_create = intel_be_buffer_create;
+	dev->base.user_buffer_create = intel_be_user_buffer_create;
+	dev->base.buffer_map = intel_be_buffer_map;
+	dev->base.buffer_unmap = intel_be_buffer_unmap;
+	dev->base.buffer_destroy = intel_be_buffer_destroy;
+
+	/* Not used anymore */
+	dev->base.surface_buffer_create = NULL;
+
+	dev->base.fence_reference = intel_be_fence_refunref;
+	dev->base.fence_signalled = intel_be_fence_signalled;
+	dev->base.fence_finish = intel_be_fence_finish;
+
+	dev->pools.gem = drm_intel_bufmgr_gem_init(dev->fd, dev->max_batch_size);
+
+	return true;
+}
+
+void
+intel_be_destroy_device(struct intel_be_device *dev)
+{
+	drm_intel_bufmgr_destroy(dev->pools.gem);
+}
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_device.h b/src/gallium/winsys/drm/intel/gem/intel_be_device.h
new file mode 100644
index 0000000000..f06890163c
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_device.h
@@ -0,0 +1,74 @@
+
+#ifndef INTEL_DRM_DEVICE_H
+#define INTEL_DRM_DEVICE_H
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_context.h"
+
+#include "drm.h"
+#include "intel_bufmgr.h"
+
+/*
+ * Device
+ */
+
+struct intel_be_device
+{
+	struct pipe_winsys base;
+
+	int fd; /**< Drm file discriptor */
+
+	size_t max_batch_size;
+	size_t max_vertex_size;
+
+	struct {
+		drm_intel_bufmgr *gem;
+	} pools;
+};
+
+boolean
+intel_be_init_device(struct intel_be_device *device, int fd, unsigned id);
+
+void
+intel_be_destroy_device(struct intel_be_device *dev);
+
+/*
+ * Buffer
+ */
+
+struct intel_be_buffer {
+	struct pipe_buffer base;
+	drm_intel_bo *bo;
+};
+
+/**
+ * Create a be buffer from a drm bo handle.
+ *
+ * Takes a reference.
+ */
+struct pipe_buffer *
+intel_be_buffer_from_handle(struct pipe_winsys *winsys,
+                            const char* name, unsigned handle);
+
+/**
+ * Gets a handle from a buffer.
+ *
+ * If buffer is destroyed handle may become invalid.
+ */
+unsigned
+intel_be_handle_from_buffer(struct pipe_winsys *winsys,
+                            struct pipe_buffer *buffer);
+
+static INLINE struct intel_be_buffer *
+intel_be_buffer(struct pipe_buffer *buf)
+{
+	return (struct intel_be_buffer *)buf;
+}
+
+static INLINE drm_intel_bo *
+intel_bo(struct pipe_buffer *buf)
+{
+	return intel_be_buffer(buf)->bo;
+}
+
+#endif
diff --git a/src/gallium/winsys/drm/intel/gem/intel_be_fence.h b/src/gallium/winsys/drm/intel/gem/intel_be_fence.h
new file mode 100644
index 0000000000..0fe18f66f8
--- /dev/null
+++ b/src/gallium/winsys/drm/intel/gem/intel_be_fence.h
@@ -0,0 +1,38 @@
+
+#ifndef INTEL_BE_FENCE_H
+#define INTEL_BE_FENCE_H
+
+#include "pipe/p_defines.h"
+
+#include "drm.h"
+#include "intel_bufmgr.h"
+
+/**
+ * Because gem does not have fence's we have to create our own fences.
+ *
+ * They work by keeping the batchbuffer around and checking if that has
+ * been idled. If bo is NULL fence has expired.
+ */
+struct intel_be_fence
+{
+	uint32_t refcount;
+	drm_intel_bo *bo;
+};
+
+static INLINE void
+intel_be_fence_reference(struct intel_be_fence *f)
+{
+	f->refcount++;
+}
+
+static INLINE void
+intel_be_fence_unreference(struct intel_be_fence *f)
+{
+	if (!--f->refcount) {
+		if (f->bo)
+			drm_intel_bo_unreference(f->bo);
+		free(f);
+	}
+}
+
+#endif
diff --git a/src/gallium/winsys/drm/nouveau/Makefile b/src/gallium/winsys/drm/nouveau/Makefile
new file mode 100644
index 0000000000..b5735329ec
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/Makefile
@@ -0,0 +1,25 @@
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+
+SUBDIRS = common dri
+
+
+default: subdirs
+
+
+subdirs:
+	@for dir in $(SUBDIRS) ; do \
+		if [ -d $$dir ] ; then \
+			(cd $$dir && $(MAKE)) || exit 1 ; \
+		fi \
+	done
+
+
+clean:
+	rm -f `find . -name \*.[oa]`
+	rm -f `find . -name depend`
+
+
+# Dummy install target
+install:
diff --git a/src/gallium/winsys/drm/nouveau/common/Makefile b/src/gallium/winsys/drm/nouveau/common/Makefile
new file mode 100644
index 0000000000..c6dd6dd7f9
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/Makefile
@@ -0,0 +1,22 @@
+TOP = ../../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nouveaudrm
+
+C_SOURCES = \
+        nouveau_context.c \
+	nouveau_lock.c \
+	nouveau_screen.c \
+	nouveau_winsys.c \
+	nouveau_winsys_pipe.c \
+	nouveau_winsys_softpipe.c
+
+include ./Makefile.template
+
+DRIVER_DEFINES = $(shell pkg-config libdrm --cflags \
+                && pkg-config libdrm --atleast-version=2.3.1 \
+                && pkg-config libdrm_nouveau --exact-version=0.5 \
+                && pkg-config libdrm_nouveau --cflags \
+                && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP")
+symlinks:
+
diff --git a/src/gallium/winsys/drm/nouveau/common/Makefile.template b/src/gallium/winsys/drm/nouveau/common/Makefile.template
new file mode 100644
index 0000000000..f0d098bb1c
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/Makefile.template
@@ -0,0 +1,59 @@
+# -*-makefile-*-
+
+COMMON_SOURCES = 
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+          $(CPP_SOURCES:.cpp=.o) \
+	  $(ASM_SOURCES:.S=.o) 
+
+
+### Include directories
+INCLUDES = \
+	-I. \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/include \
+        $(DRIVER_INCLUDES)
+
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.cpp.o:
+	$(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES)  $< -o $@
+
+
+##### TARGETS #####
+
+default: depend symlinks lib$(LIBNAME).a
+
+
+lib$(LIBNAME).a: $(OBJECTS) Makefile Makefile.template
+	$(TOP)/bin/mklib -o $(LIBNAME) -static $(OBJECTS) $(DRIVER_LIBS)
+
+
+depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \
+		$(ASM_SOURCES) 2> /dev/null
+
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` `find ../include`
+
+
+# Remove .o and backup files
+clean::
+	-rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS)
+	-rm -f depend depend.bak
+
+
+include depend
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_context.c b/src/gallium/winsys/drm/nouveau/common/nouveau_context.c
new file mode 100644
index 0000000000..d6ae0827cd
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_context.c
@@ -0,0 +1,206 @@
+#include <pipe/p_defines.h>
+#include <pipe/p_context.h>
+#include <pipe/p_screen.h>
+#include <util/u_memory.h>
+#include "nouveau_context.h"
+#include "nouveau_dri.h"
+#include "nouveau_local.h"
+#include "nouveau_screen.h"
+#include "nouveau_winsys_pipe.h"
+
+static void
+nouveau_channel_context_destroy(struct nouveau_channel_context *nvc)
+{
+	nouveau_channel_free(&nvc->channel);
+
+	FREE(nvc);
+}
+
+static struct nouveau_channel_context *
+nouveau_channel_context_create(struct nouveau_device *dev)
+{
+	struct nouveau_channel_context *nvc;
+	int ret;
+
+	nvc = CALLOC_STRUCT(nouveau_channel_context);
+	if (!nvc)
+		return NULL;
+
+	if ((ret = nouveau_channel_alloc(dev, 0x8003d001, 0x8003d002,
+					 &nvc->channel))) {
+		NOUVEAU_ERR("Error creating GPU channel: %d\n", ret);
+		nouveau_channel_context_destroy(nvc);
+		return NULL;
+	}
+
+	nvc->next_handle = 0x77000000;
+	return nvc;
+}
+
+int
+nouveau_context_init(struct nouveau_screen *nv_screen,
+                     drm_context_t hHWContext, drmLock *sarea_lock,
+                     struct nouveau_context *nv_share,
+                     struct nouveau_context *nv)
+{
+	struct pipe_context *pipe = NULL;
+	struct nouveau_channel_context *nvc = NULL;
+	struct nouveau_device *dev = nv_screen->device;
+	int i;
+
+	switch (dev->chipset & 0xf0) {
+	case 0x00:
+		/* NV04 */
+	case 0x10:
+	case 0x20:
+		/* NV10 */
+	case 0x30:
+		/* NV30 */
+	case 0x40:
+	case 0x60:
+		/* NV40 */
+	case 0x50:
+	case 0x80:
+	case 0x90:
+		/* G80 */
+		break;
+	default:
+		NOUVEAU_ERR("Unsupported chipset: NV%02x\n", dev->chipset);
+		return 1;
+	}
+
+	nv->nv_screen  = nv_screen;
+
+	{
+		struct nouveau_device_priv *nvdev = nouveau_device(dev);
+
+		nvdev->ctx  = hHWContext;
+		nvdev->lock = sarea_lock;
+	}
+
+	/* Attempt to share a single channel between multiple contexts from
+	 * a single process.
+	 */
+	nvc = nv_screen->nvc;
+	if (!nvc && nv_share)
+		nvc = nv_share->nvc;
+
+	/*XXX: temporary - disable multi-context/single-channel on pre-NV4x */
+	switch (dev->chipset & 0xf0) {
+	case 0x40:
+	case 0x60:
+		/* NV40 class */
+	case 0x50:
+	case 0x80:
+	case 0x90:
+		/* G80 class */
+		break;
+	default:
+		nvc = NULL;
+		break;
+	}
+
+	if (!nvc) {
+		nvc = nouveau_channel_context_create(dev);
+		if (!nvc) {
+			NOUVEAU_ERR("Failed initialising GPU context\n");
+			return 1;
+		}
+		nv_screen->nvc = nvc;
+	}
+
+	nvc->refcount++;
+	nv->nvc = nvc;
+
+	/* Find a free slot for a pipe context, allocate a new one if needed */
+	nv->pctx_id = -1;
+	for (i = 0; i < nvc->nr_pctx; i++) {
+		if (nvc->pctx[i] == NULL) {
+			nv->pctx_id = i;
+			break;
+		}
+	}
+
+	if (nv->pctx_id < 0) {
+		nv->pctx_id = nvc->nr_pctx++;
+		nvc->pctx =
+			realloc(nvc->pctx,
+				sizeof(struct pipe_context *) * nvc->nr_pctx);
+	}
+
+	/* Create pipe */
+	if (!getenv("NOUVEAU_FORCE_SOFTPIPE")) {
+		struct pipe_screen *pscreen;
+
+		pipe = nouveau_pipe_create(nv);
+		if (!pipe)
+			NOUVEAU_ERR("Couldn't create hw pipe\n");
+		pscreen = nvc->pscreen;
+
+		nv->cap.hw_vertex_buffer =
+			pscreen->get_param(pscreen, NOUVEAU_CAP_HW_VTXBUF);
+		nv->cap.hw_index_buffer =
+			pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF);
+	}
+
+	if (!pipe) {
+		NOUVEAU_MSG("Using softpipe\n");
+		pipe = nouveau_create_softpipe(nv);
+		if (!pipe) {
+			NOUVEAU_ERR("Error creating pipe, bailing\n");
+			return 1;
+		}
+	}
+
+	{
+		struct pipe_texture *fb_tex;
+		struct pipe_surface *fb_surf;
+		struct nouveau_pipe_buffer *fb_buf;
+		enum pipe_format format;
+
+		fb_buf = calloc(1, sizeof(struct nouveau_pipe_buffer));
+		fb_buf->base.refcount = 1;
+		fb_buf->base.usage = PIPE_BUFFER_USAGE_PIXEL;
+
+		nouveau_bo_fake(dev, nv_screen->front_offset, NOUVEAU_BO_VRAM,
+				nv_screen->front_pitch*nv_screen->front_height,
+				NULL, &fb_buf->bo);
+
+		if (nv_screen->front_cpp == 4)
+			format = PIPE_FORMAT_A8R8G8B8_UNORM;
+		else
+			format = PIPE_FORMAT_R5G6B5_UNORM;
+
+		fb_surf = nouveau_surface_buffer_ref(nv, &fb_buf->base, format,
+						     nv_screen->front_pitch /
+						     nv_screen->front_cpp,
+						     nv_screen->front_height,
+						     nv_screen->front_pitch,
+						     &fb_tex);
+
+		nv->frontbuffer = fb_surf;
+		nv->frontbuffer_texture = fb_tex;
+	}
+
+	pipe->priv = nv;
+	return 0;
+}
+
+void
+nouveau_context_cleanup(struct nouveau_context *nv)
+{
+	struct nouveau_channel_context *nvc = nv->nvc;
+
+	assert(nv);
+
+	if (nv->pctx_id >= 0) {
+		nvc->pctx[nv->pctx_id] = NULL;
+		if (--nvc->refcount <= 0) {
+			nouveau_channel_context_destroy(nvc);
+			nv->nv_screen->nvc = NULL;
+		}
+	}
+	
+	/* XXX: Who cleans up the pipe? */
+}
+
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_context.h b/src/gallium/winsys/drm/nouveau/common/nouveau_context.h
new file mode 100644
index 0000000000..02d2745680
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_context.h
@@ -0,0 +1,59 @@
+#ifndef __NOUVEAU_CONTEXT_H__
+#define __NOUVEAU_CONTEXT_H__
+
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau_drmif.h"
+#include "nouveau_device.h"
+#include "nouveau_channel.h"
+#include "nouveau_pushbuf.h"
+#include "nouveau_bo.h"
+#include "nouveau_grobj.h"
+#include "nouveau_notifier.h"
+#include "nouveau_class.h"
+#include "nouveau_local.h"
+
+struct nouveau_channel_context {
+	struct pipe_screen *pscreen;
+	int refcount;
+
+	unsigned cur_pctx;
+	unsigned nr_pctx;
+	struct pipe_context **pctx;
+
+	struct nouveau_channel  *channel;
+	unsigned next_handle;
+};
+
+struct nouveau_context {
+	int locked;
+	struct nouveau_screen *nv_screen;
+	struct pipe_surface *frontbuffer;
+	struct pipe_texture *frontbuffer_texture;
+
+	struct {
+		int hw_vertex_buffer;
+		int hw_index_buffer;
+	} cap;
+
+	/* Hardware context */
+	struct nouveau_channel_context *nvc;
+	int pctx_id;
+};
+
+extern int nouveau_context_init(struct nouveau_screen *nv_screen,
+                                drm_context_t hHWContext, drmLock *sarea_lock,
+                                struct nouveau_context *nv_share,
+                                struct nouveau_context *nv);
+extern void nouveau_context_cleanup(struct nouveau_context *nv);
+
+extern void LOCK_HARDWARE(struct nouveau_context *);
+extern void UNLOCK_HARDWARE(struct nouveau_context *);
+
+extern uint32_t *nouveau_pipe_dma_beginp(struct nouveau_grobj *, int, int);
+extern void nouveau_pipe_dma_kickoff(struct nouveau_channel *);
+
+/* Must be provided by clients of common code */
+extern void
+nouveau_contended_lock(struct nouveau_context *nv);
+
+#endif
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_dri.h b/src/gallium/winsys/drm/nouveau/common/nouveau_dri.h
new file mode 100644
index 0000000000..1207c2d609
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_dri.h
@@ -0,0 +1,28 @@
+#ifndef _NOUVEAU_DRI_
+#define _NOUVEAU_DRI_
+
+#include "xf86drm.h"
+#include "drm.h"
+#include "nouveau_drm.h"
+
+struct nouveau_dri {
+	uint32_t device_id;	/**< \brief PCI device ID */
+	uint32_t width;		/**< \brief width in pixels of display */
+	uint32_t height;	/**< \brief height in scanlines of display */
+	uint32_t depth;		/**< \brief depth of display (8, 15, 16, 24) */
+	uint32_t bpp;		/**< \brief bit depth of display (8, 16, 24, 32) */
+
+	uint32_t bus_type;	/**< \brief ths bus type */
+	uint32_t bus_mode;	/**< \brief bus mode (used for AGP, maybe also for PCI-E ?) */
+
+	uint32_t front_offset;	/**< \brief front buffer offset */
+	uint32_t front_pitch;	/**< \brief front buffer pitch */
+	uint32_t back_offset;	/**< \brief private back buffer offset */
+	uint32_t back_pitch;	/**< \brief private back buffer pitch */
+	uint32_t depth_offset;	/**< \brief private depth buffer offset */
+	uint32_t depth_pitch;	/**< \brief private depth buffer pitch */
+
+};
+
+#endif
+
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_local.h b/src/gallium/winsys/drm/nouveau/common/nouveau_local.h
new file mode 100644
index 0000000000..11175bce7a
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_local.h
@@ -0,0 +1,19 @@
+#ifndef __NOUVEAU_LOCAL_H__
+#define __NOUVEAU_LOCAL_H__
+
+#include "pipe/p_compiler.h"
+#include "nouveau_winsys_pipe.h"
+#include <stdio.h>
+
+/* Debug output */
+#define NOUVEAU_MSG(fmt, args...) do {                                         \
+	fprintf(stdout, "nouveau: "fmt, ##args);                               \
+	fflush(stdout);                                                        \
+} while(0)
+
+#define NOUVEAU_ERR(fmt, args...) do {                                         \
+	fprintf(stderr, "%s:%d -  "fmt, __func__, __LINE__, ##args);           \
+	fflush(stderr);                                                        \
+} while(0)
+
+#endif
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_lock.c b/src/gallium/winsys/drm/nouveau/common/nouveau_lock.c
new file mode 100644
index 0000000000..e8cf051ed9
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_lock.c
@@ -0,0 +1,72 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include <pipe/p_thread.h>
+#include "nouveau_context.h"
+#include "nouveau_screen.h"
+
+pipe_static_mutex(lockMutex);
+
+/* Lock the hardware and validate our state.
+ */
+void
+LOCK_HARDWARE(struct nouveau_context *nv)
+{
+	struct nouveau_screen *nv_screen = nv->nv_screen;
+	struct nouveau_device *dev = nv_screen->device;
+	struct nouveau_device_priv *nvdev = nouveau_device(dev);
+	char __ret=0;
+
+	assert(!nv->locked);
+	pipe_mutex_lock(lockMutex);
+
+	DRM_CAS(nvdev->lock, nvdev->ctx,
+		(DRM_LOCK_HELD | nvdev->ctx), __ret);
+
+	if (__ret) {
+		drmGetLock(nvdev->fd, nvdev->ctx, 0);
+		nouveau_contended_lock(nv);
+	}
+	nv->locked = 1;
+}
+
+/* Unlock the hardware using the global current context 
+ */
+void
+UNLOCK_HARDWARE(struct nouveau_context *nv)
+{
+	struct nouveau_screen *nv_screen = nv->nv_screen;
+	struct nouveau_device *dev = nv_screen->device;
+	struct nouveau_device_priv *nvdev = nouveau_device(dev);
+
+	assert(nv->locked);
+	nv->locked = 0;
+
+	DRM_UNLOCK(nvdev->fd, nvdev->lock, nvdev->ctx);
+
+	pipe_mutex_unlock(lockMutex);
+} 
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_screen.c b/src/gallium/winsys/drm/nouveau/common/nouveau_screen.c
new file mode 100644
index 0000000000..422fbf0207
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_screen.c
@@ -0,0 +1,31 @@
+#include <util/u_memory.h>
+#include "nouveau_dri.h"
+#include "nouveau_local.h"
+#include "nouveau_screen.h"
+
+int
+nouveau_screen_init(struct nouveau_dri *nv_dri, int dev_fd,
+                    struct nouveau_screen *nv_screen)
+{
+	int ret;
+
+	ret = nouveau_device_open_existing(&nv_screen->device, 0,
+	                                   dev_fd, 0);
+	if (ret) {
+		NOUVEAU_ERR("Failed opening nouveau device: %d\n", ret);
+		return 1;
+	}
+
+	nv_screen->front_offset = nv_dri->front_offset;
+	nv_screen->front_pitch  = nv_dri->front_pitch * (nv_dri->bpp / 8);
+	nv_screen->front_cpp = nv_dri->bpp / 8;
+	nv_screen->front_height = nv_dri->height;
+
+	return 0;
+}
+
+void
+nouveau_screen_cleanup(struct nouveau_screen *nv_screen)
+{
+	nouveau_device_close(&nv_screen->device);
+}
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_screen.h b/src/gallium/winsys/drm/nouveau/common/nouveau_screen.h
new file mode 100644
index 0000000000..3e68e219d8
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_screen.h
@@ -0,0 +1,27 @@
+#ifndef __NOUVEAU_SCREEN_H__
+#define __NOUVEAU_SCREEN_H__
+
+#include <stdint.h>
+
+struct nouveau_device;
+struct nouveau_dri;
+
+struct nouveau_screen {
+	struct nouveau_device *device;
+
+	uint32_t front_offset;
+	uint32_t front_pitch;
+	uint32_t front_cpp;
+	uint32_t front_height;
+
+	void *nvc;
+};
+
+int
+nouveau_screen_init(struct nouveau_dri *nv_dri, int dev_fd,
+                    struct nouveau_screen *nv_screen);
+
+void
+nouveau_screen_cleanup(struct nouveau_screen *nv_screen);
+
+#endif
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys.c b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys.c
new file mode 100644
index 0000000000..b6199f8e6d
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys.c
@@ -0,0 +1,141 @@
+#include "util/u_memory.h"
+
+#include "nouveau_context.h"
+#include "nouveau_screen.h"
+#include "nouveau_winsys_pipe.h"
+
+#include "nouveau/nouveau_winsys.h"
+
+static int
+nouveau_pipe_notifier_alloc(struct nouveau_winsys *nvws, int count,
+			    struct nouveau_notifier **notify)
+{
+	struct nouveau_context *nv = nvws->nv;
+
+	return nouveau_notifier_alloc(nv->nvc->channel, nv->nvc->next_handle++,
+				      count, notify);
+}
+
+static int
+nouveau_pipe_grobj_alloc(struct nouveau_winsys *nvws, int grclass,
+			 struct nouveau_grobj **grobj)
+{
+	struct nouveau_context *nv = nvws->nv;
+	struct nouveau_channel *chan = nv->nvc->channel;
+	int ret;
+
+	ret = nouveau_grobj_alloc(chan, nv->nvc->next_handle++,
+				  grclass, grobj);
+	if (ret)
+		return ret;
+
+	BEGIN_RING(chan, *grobj, 0x0000, 1);
+	OUT_RING  (chan, (*grobj)->handle);
+	(*grobj)->bound = NOUVEAU_GROBJ_BOUND_EXPLICIT;
+	return 0;
+}
+
+static int
+nouveau_pipe_push_reloc(struct nouveau_winsys *nvws, void *ptr,
+			struct pipe_buffer *buf, uint32_t data,
+			uint32_t flags, uint32_t vor, uint32_t tor)
+{
+	struct nouveau_bo *bo = nouveau_pipe_buffer(buf)->bo;
+
+	return nouveau_pushbuf_emit_reloc(nvws->channel, ptr, bo,
+					  data, flags, vor, tor);
+}
+
+static int
+nouveau_pipe_push_flush(struct nouveau_winsys *nvws, unsigned size,
+			struct pipe_fence_handle **fence)
+{
+	if (fence)
+		*fence = NULL;
+
+	return nouveau_pushbuf_flush(nvws->channel, size);
+}
+
+static struct nouveau_bo *
+nouveau_pipe_get_bo(struct pipe_buffer *pb)
+{
+	return nouveau_pipe_buffer(pb)->bo;
+}
+
+struct pipe_context *
+nouveau_pipe_create(struct nouveau_context *nv)
+{
+	struct nouveau_channel_context *nvc = nv->nvc;
+	struct nouveau_winsys *nvws = CALLOC_STRUCT(nouveau_winsys);
+	struct pipe_screen *(*hws_create)(struct pipe_winsys *,
+					  struct nouveau_winsys *);
+	struct pipe_context *(*hw_create)(struct pipe_screen *, unsigned);
+	struct pipe_winsys *ws;
+	unsigned chipset = nv->nv_screen->device->chipset;
+
+	if (!nvws)
+		return NULL;
+
+	switch (chipset & 0xf0) {
+	case 0x00:
+		hws_create = nv04_screen_create;
+		hw_create = nv04_create;
+		break;
+	case 0x10:
+		hws_create = nv10_screen_create;
+		hw_create = nv10_create;
+		break;
+	case 0x20:
+		hws_create = nv20_screen_create;
+		hw_create = nv20_create;
+		break;
+	case 0x30:
+		hws_create = nv30_screen_create;
+		hw_create = nv30_create;
+		break;
+	case 0x40:
+	case 0x60:
+		hws_create = nv40_screen_create;
+		hw_create = nv40_create;
+		break;
+	case 0x50:
+	case 0x80:
+	case 0x90:
+		hws_create = nv50_screen_create;
+		hw_create = nv50_create;
+		break;
+	default:
+		NOUVEAU_ERR("Unknown chipset NV%02x\n", chipset);
+		return NULL;
+	}
+
+	nvws->nv		= nv;
+	nvws->channel		= nv->nvc->channel;
+
+	nvws->res_init		= nouveau_resource_init;
+	nvws->res_alloc		= nouveau_resource_alloc;
+	nvws->res_free		= nouveau_resource_free;
+
+	nvws->push_reloc        = nouveau_pipe_push_reloc;
+	nvws->push_flush	= nouveau_pipe_push_flush;
+
+	nvws->grobj_alloc	= nouveau_pipe_grobj_alloc;
+	nvws->grobj_free	= nouveau_grobj_free;
+
+	nvws->notifier_alloc	= nouveau_pipe_notifier_alloc;
+	nvws->notifier_free	= nouveau_notifier_free;
+	nvws->notifier_reset	= nouveau_notifier_reset;
+	nvws->notifier_status	= nouveau_notifier_status;
+	nvws->notifier_retval	= nouveau_notifier_return_val;
+	nvws->notifier_wait	= nouveau_notifier_wait_status;
+
+	nvws->get_bo		= nouveau_pipe_get_bo;
+
+	ws = nouveau_create_pipe_winsys(nv);
+
+	if (!nvc->pscreen)
+		nvc->pscreen = hws_create(ws, nvws);
+	nvc->pctx[nv->pctx_id] = hw_create(nvc->pscreen, nv->pctx_id);
+	return nvc->pctx[nv->pctx_id];
+}
+
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.c b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.c
new file mode 100644
index 0000000000..881df98556
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.c
@@ -0,0 +1,242 @@
+#include "pipe/internal/p_winsys_screen.h"
+#include <pipe/p_defines.h>
+#include <pipe/p_inlines.h>
+#include <util/u_memory.h>
+#include "nouveau_context.h"
+#include "nouveau_local.h"
+#include "nouveau_screen.h"
+#include "nouveau_winsys_pipe.h"
+
+static const char *
+nouveau_get_name(struct pipe_winsys *pws)
+{
+	return "Nouveau/DRI";
+}
+
+static uint32_t
+nouveau_flags_from_usage(struct nouveau_context *nv, unsigned usage)
+{
+	struct nouveau_device *dev = nv->nv_screen->device;
+	uint32_t flags = NOUVEAU_BO_LOCAL;
+
+	if (usage & PIPE_BUFFER_USAGE_PIXEL) {
+		if (usage & NOUVEAU_BUFFER_USAGE_TEXTURE)
+			flags |= NOUVEAU_BO_GART;
+		if (!(usage & PIPE_BUFFER_USAGE_CPU_READ_WRITE))
+			flags |= NOUVEAU_BO_VRAM;
+
+		switch (dev->chipset & 0xf0) {
+		case 0x50:
+		case 0x80:
+		case 0x90:
+			flags |= NOUVEAU_BO_TILED;
+			if (usage & NOUVEAU_BUFFER_USAGE_ZETA)
+				flags |= NOUVEAU_BO_ZTILE;
+			break;
+		default:
+			break;
+		}
+	}
+
+	if (usage & PIPE_BUFFER_USAGE_VERTEX) {
+		if (nv->cap.hw_vertex_buffer)
+			flags |= NOUVEAU_BO_GART;
+	}
+
+	if (usage & PIPE_BUFFER_USAGE_INDEX) {
+		if (nv->cap.hw_index_buffer)
+			flags |= NOUVEAU_BO_GART;
+	}
+
+	return flags;
+}
+
+static struct pipe_buffer *
+nouveau_pipe_bo_create(struct pipe_winsys *pws, unsigned alignment,
+		       unsigned usage, unsigned size)
+{
+	struct nouveau_pipe_winsys *nvpws = (struct nouveau_pipe_winsys *)pws;
+	struct nouveau_context *nv = nvpws->nv;
+	struct nouveau_device *dev = nv->nv_screen->device;
+	struct nouveau_pipe_buffer *nvbuf;
+	uint32_t flags;
+
+	nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer);
+	if (!nvbuf)
+		return NULL;
+	nvbuf->base.refcount = 1;
+	nvbuf->base.alignment = alignment;
+	nvbuf->base.usage = usage;
+	nvbuf->base.size = size;
+
+	flags = nouveau_flags_from_usage(nv, usage);
+
+	if (nouveau_bo_new(dev, flags, alignment, size, &nvbuf->bo)) {
+		FREE(nvbuf);
+		return NULL;
+	}
+
+	return &nvbuf->base;
+}
+
+static struct pipe_buffer *
+nouveau_pipe_bo_user_create(struct pipe_winsys *pws, void *ptr, unsigned bytes)
+{
+	struct nouveau_pipe_winsys *nvpws = (struct nouveau_pipe_winsys *)pws;
+	struct nouveau_device *dev = nvpws->nv->nv_screen->device;
+	struct nouveau_pipe_buffer *nvbuf;
+
+	nvbuf = CALLOC_STRUCT(nouveau_pipe_buffer);
+	if (!nvbuf)
+		return NULL;
+	nvbuf->base.refcount = 1;
+	nvbuf->base.size = bytes;
+
+	if (nouveau_bo_user(dev, ptr, bytes, &nvbuf->bo)) {
+		FREE(nvbuf);
+		return NULL;
+	}
+
+	return &nvbuf->base;
+}
+
+static void
+nouveau_pipe_bo_del(struct pipe_winsys *ws, struct pipe_buffer *buf)
+{
+	struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf);
+
+	nouveau_bo_ref(NULL, &nvbuf->bo);
+	FREE(nvbuf);
+}
+
+static void *
+nouveau_pipe_bo_map(struct pipe_winsys *pws, struct pipe_buffer *buf,
+		    unsigned flags)
+{
+	struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf);
+	uint32_t map_flags = 0;
+
+	if (flags & PIPE_BUFFER_USAGE_CPU_READ)
+		map_flags |= NOUVEAU_BO_RD;
+	if (flags & PIPE_BUFFER_USAGE_CPU_WRITE)
+		map_flags |= NOUVEAU_BO_WR;
+
+#if 0
+	if (flags & PIPE_BUFFER_USAGE_DISCARD &&
+	    !(flags & PIPE_BUFFER_USAGE_CPU_READ) &&
+	    nouveau_bo_busy(nvbuf->bo, map_flags)) {
+		struct nouveau_pipe_winsys *nvpws = (struct nouveau_pipe_winsys *)pws;
+		struct nouveau_context *nv = nvpws->nv;
+		struct nouveau_device *dev = nv->nv_screen->device;
+		struct nouveau_bo *rename;
+		uint32_t flags = nouveau_flags_from_usage(nv, buf->usage);
+
+		if (!nouveau_bo_new(dev, flags, buf->alignment, buf->size, &rename)) {
+			nouveau_bo_ref(NULL, &nvbuf->bo);
+			nvbuf->bo = rename;
+		}
+	}
+#endif
+
+	if (nouveau_bo_map(nvbuf->bo, map_flags))
+		return NULL;
+	return nvbuf->bo->map;
+}
+
+static void
+nouveau_pipe_bo_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf)
+{
+	struct nouveau_pipe_buffer *nvbuf = nouveau_pipe_buffer(buf);
+
+	nouveau_bo_unmap(nvbuf->bo);
+}
+
+static void
+nouveau_pipe_fence_reference(struct pipe_winsys *ws,
+			     struct pipe_fence_handle **ptr,
+			     struct pipe_fence_handle *pfence)
+{
+	*ptr = pfence;
+}
+
+static int
+nouveau_pipe_fence_signalled(struct pipe_winsys *ws,
+			     struct pipe_fence_handle *pfence, unsigned flag)
+{
+	return 0;
+}
+
+static int
+nouveau_pipe_fence_finish(struct pipe_winsys *ws,
+			  struct pipe_fence_handle *pfence, unsigned flag)
+{
+	return 0;
+}
+
+struct pipe_surface *
+nouveau_surface_buffer_ref(struct nouveau_context *nv, struct pipe_buffer *pb,
+			   enum pipe_format format, int w, int h,
+			   unsigned pitch, struct pipe_texture **ppt)
+{
+	struct pipe_screen *pscreen = nv->nvc->pscreen;
+	struct pipe_texture tmpl, *pt;
+	struct pipe_surface *ps;
+
+	memset(&tmpl, 0, sizeof(tmpl));
+	tmpl.tex_usage = PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+			 NOUVEAU_TEXTURE_USAGE_LINEAR;
+	tmpl.target = PIPE_TEXTURE_2D;
+	tmpl.width[0] = w;
+	tmpl.height[0] = h;
+	tmpl.depth[0] = 1;
+	tmpl.format = format;
+	pf_get_block(tmpl.format, &tmpl.block);
+	tmpl.nblocksx[0] = pf_get_nblocksx(&tmpl.block, w);
+	tmpl.nblocksy[0] = pf_get_nblocksy(&tmpl.block, h);
+
+	pt = pscreen->texture_blanket(pscreen, &tmpl, &pitch, pb);
+	if (!pt)
+		return NULL;
+
+	ps = pscreen->get_tex_surface(pscreen, pt, 0, 0, 0,
+				      PIPE_BUFFER_USAGE_GPU_WRITE);
+
+	*ppt = pt;
+	return ps;
+}
+
+static void
+nouveau_destroy(struct pipe_winsys *pws)
+{
+	FREE(pws);
+}
+
+struct pipe_winsys *
+nouveau_create_pipe_winsys(struct nouveau_context *nv)
+{
+	struct nouveau_pipe_winsys *nvpws;
+	struct pipe_winsys *pws;
+
+	nvpws = CALLOC_STRUCT(nouveau_pipe_winsys);
+	if (!nvpws)
+		return NULL;
+	nvpws->nv = nv;
+	pws = &nvpws->pws;
+
+	pws->flush_frontbuffer = nouveau_flush_frontbuffer;
+
+	pws->buffer_create = nouveau_pipe_bo_create;
+	pws->buffer_destroy = nouveau_pipe_bo_del;
+	pws->user_buffer_create = nouveau_pipe_bo_user_create;
+	pws->buffer_map = nouveau_pipe_bo_map;
+	pws->buffer_unmap = nouveau_pipe_bo_unmap;
+
+	pws->fence_reference = nouveau_pipe_fence_reference;
+	pws->fence_signalled = nouveau_pipe_fence_signalled;
+	pws->fence_finish = nouveau_pipe_fence_finish;
+
+	pws->get_name = nouveau_get_name;
+	pws->destroy = nouveau_destroy;
+
+	return &nvpws->pws;
+}
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.h b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.h
new file mode 100644
index 0000000000..1eb8043478
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_pipe.h
@@ -0,0 +1,44 @@
+#ifndef NOUVEAU_PIPE_WINSYS_H
+#define NOUVEAU_PIPE_WINSYS_H
+
+#include "pipe/p_context.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "nouveau_context.h"
+
+struct nouveau_pipe_buffer {
+	struct pipe_buffer base;
+	struct nouveau_bo *bo;
+};
+
+static INLINE struct nouveau_pipe_buffer *
+nouveau_pipe_buffer(struct pipe_buffer *buf)
+{
+	return (struct nouveau_pipe_buffer *)buf;
+}
+
+struct nouveau_pipe_winsys {
+	struct pipe_winsys pws;
+
+	struct nouveau_context *nv;
+};
+
+extern struct pipe_winsys *
+nouveau_create_pipe_winsys(struct nouveau_context *nv);
+
+struct pipe_context *
+nouveau_create_softpipe(struct nouveau_context *nv);
+
+struct pipe_context *
+nouveau_pipe_create(struct nouveau_context *nv);
+
+/* Must be provided by clients of common code */
+extern void
+nouveau_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *surf,
+			  void *context_private);
+
+struct pipe_surface *
+nouveau_surface_buffer_ref(struct nouveau_context *nv, struct pipe_buffer *pb,
+			   enum pipe_format format, int w, int h,
+			   unsigned pitch, struct pipe_texture **ppt);
+
+#endif
diff --git a/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_softpipe.c b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_softpipe.c
new file mode 100644
index 0000000000..396e4f2a2e
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/common/nouveau_winsys_softpipe.c
@@ -0,0 +1,101 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+/*
+ * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com>
+ */
+
+#include "pipe/internal/p_winsys_screen.h"
+#include <pipe/p_screen.h>
+#include <pipe/p_defines.h>
+#include <pipe/p_format.h>
+#include <softpipe/sp_winsys.h>
+#include <util/u_memory.h>
+#include "nouveau_context.h"
+#include "nouveau_winsys_pipe.h"
+
+struct nouveau_softpipe_winsys {
+   struct softpipe_winsys sws;
+   struct nouveau_context *nv;
+};
+
+/**
+ * Return list of surface formats supported by this driver.
+ */
+static boolean
+nouveau_is_format_supported(struct softpipe_winsys *sws,
+                            enum pipe_format format)
+{
+	switch (format) {
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case PIPE_FORMAT_R5G6B5_UNORM:
+	case PIPE_FORMAT_Z24S8_UNORM:
+		return TRUE;
+	default:
+		break;
+	};
+
+	return FALSE;
+}
+
+struct pipe_context *
+nouveau_create_softpipe(struct nouveau_context *nv)
+{
+	struct nouveau_softpipe_winsys *nvsws;
+	struct pipe_screen *pscreen;
+	struct pipe_winsys *ws;
+	struct pipe_context *pipe;
+
+	ws = nouveau_create_pipe_winsys(nv);
+	if (!ws)
+		return NULL;
+	pscreen = softpipe_create_screen(ws);
+	if (!pscreen) {
+		ws->destroy(ws);
+		return NULL;
+	}
+	nvsws = CALLOC_STRUCT(nouveau_softpipe_winsys);
+	if (!nvsws) {
+		ws->destroy(ws);
+		pscreen->destroy(pscreen);
+		return NULL;
+	}
+
+	nvsws->sws.is_format_supported = nouveau_is_format_supported;
+	nvsws->nv = nv;
+
+	pipe = softpipe_create(pscreen, ws, &nvsws->sws);
+	if (!pipe) {
+		ws->destroy(ws);
+		pscreen->destroy(pscreen);
+		FREE(nvsws);
+		return NULL;
+	}
+
+	return pipe;
+}
+
diff --git a/src/gallium/winsys/drm/nouveau/dri/Makefile b/src/gallium/winsys/drm/nouveau/dri/Makefile
new file mode 100644
index 0000000000..3f3553b61d
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/dri/Makefile
@@ -0,0 +1,34 @@
+TOP = ../../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nouveau_dri.so
+
+MINIGLX_SOURCES =
+
+PIPE_DRIVERS = \
+	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/nv04/libnv04.a \
+	$(TOP)/src/gallium/drivers/nv10/libnv10.a \
+	$(TOP)/src/gallium/drivers/nv20/libnv20.a \
+	$(TOP)/src/gallium/drivers/nv30/libnv30.a \
+	$(TOP)/src/gallium/drivers/nv40/libnv40.a \
+	$(TOP)/src/gallium/drivers/nv50/libnv50.a
+
+DRIVER_SOURCES = \
+	nouveau_context_dri.c \
+	nouveau_screen_dri.c \
+	nouveau_swapbuffers.c \
+	../common/libnouveaudrm.a
+
+C_SOURCES = \
+	$(COMMON_GALLIUM_SOURCES) \
+	$(DRIVER_SOURCES)
+
+ASM_SOURCES = 
+
+DRIVER_DEFINES = $(shell pkg-config libdrm_nouveau --cflags)
+DRI_LIB_DEPS += $(shell pkg-config libdrm_nouveau --libs)
+
+include ../../Makefile.template
+
+symlinks:
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context_dri.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_context_dri.c
new file mode 100644
index 0000000000..aacfe984d1
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_context_dri.c
@@ -0,0 +1,124 @@
+#include <main/glheader.h>
+#include <glapi/glthread.h>
+#include <GL/internal/glcore.h>
+#include <utils.h>
+
+#include <state_tracker/st_public.h>
+#include <state_tracker/st_context.h>
+#include <pipe/p_defines.h>
+#include <pipe/p_context.h>
+#include <pipe/p_screen.h>
+
+#include "../common/nouveau_winsys_pipe.h"
+#include "../common/nouveau_dri.h"
+#include "../common/nouveau_local.h"
+#include "nouveau_context_dri.h"
+#include "nouveau_screen_dri.h"
+
+#ifdef DEBUG
+static const struct dri_debug_control debug_control[] = {
+	{ "bo", DEBUG_BO },
+	{ NULL, 0 }
+};
+int __nouveau_debug = 0;
+#endif
+
+GLboolean
+nouveau_context_create(const __GLcontextModes *glVis,
+		       __DRIcontextPrivate *driContextPriv,
+		       void *sharedContextPrivate)
+{
+	__DRIscreenPrivate *driScrnPriv = driContextPriv->driScreenPriv;
+	struct nouveau_screen_dri  *nv_screen = driScrnPriv->private;
+	struct nouveau_context_dri *nv = CALLOC_STRUCT(nouveau_context_dri);
+	struct st_context *st_share = NULL;
+	struct nouveau_context_dri *nv_share = NULL;
+	struct pipe_context *pipe;
+
+	if (sharedContextPrivate) {
+		st_share = ((struct nouveau_context_dri *)sharedContextPrivate)->st;
+		nv_share = st_share->pipe->priv;
+	}
+
+	if (nouveau_context_init(&nv_screen->base, driContextPriv->hHWContext,
+	                         (drmLock *)&driScrnPriv->pSAREA->lock,
+	                         &nv_share->base, &nv->base)) {
+		return GL_FALSE;
+	}
+
+	pipe = nv->base.nvc->pctx[nv->base.pctx_id];
+	driContextPriv->driverPrivate = (void *)nv;
+	//nv->nv_screen  = nv_screen;
+	nv->dri_screen = driScrnPriv;
+
+	driParseConfigFiles(&nv->dri_option_cache, &nv_screen->option_cache,
+			    nv->dri_screen->myNum, "nouveau");
+#ifdef DEBUG
+	__nouveau_debug = driParseDebugString(getenv("NOUVEAU_DEBUG"),
+					      debug_control);
+#endif
+
+	nv->st = st_create_context(pipe, glVis, st_share);
+	return GL_TRUE;
+}
+
+void
+nouveau_context_destroy(__DRIcontextPrivate *driContextPriv)
+{
+	struct nouveau_context_dri *nv = driContextPriv->driverPrivate;
+
+	assert(nv);
+
+	st_finish(nv->st);
+	st_destroy_context(nv->st);
+
+	nouveau_context_cleanup(&nv->base);
+
+	FREE(nv);
+}
+
+GLboolean
+nouveau_context_bind(__DRIcontextPrivate *driContextPriv,
+		     __DRIdrawablePrivate *driDrawPriv,
+		     __DRIdrawablePrivate *driReadPriv)
+{
+	struct nouveau_context_dri *nv;
+	struct nouveau_framebuffer *draw, *read;
+
+	if (!driContextPriv) {
+		st_make_current(NULL, NULL, NULL);
+		return GL_TRUE;
+	}
+
+	nv = driContextPriv->driverPrivate;
+	draw = driDrawPriv->driverPrivate;
+	read = driReadPriv->driverPrivate;
+
+	st_make_current(nv->st, draw->stfb, read->stfb);
+
+	if ((nv->dri_drawable != driDrawPriv) ||
+	    (nv->last_stamp != driDrawPriv->lastStamp)) {
+		nv->dri_drawable = driDrawPriv;
+		st_resize_framebuffer(draw->stfb, driDrawPriv->w,
+				      driDrawPriv->h);
+		nv->last_stamp = driDrawPriv->lastStamp;
+	}
+
+	if (driDrawPriv != driReadPriv) {
+		st_resize_framebuffer(read->stfb, driReadPriv->w,
+				      driReadPriv->h);
+	}
+
+	return GL_TRUE;
+}
+
+GLboolean
+nouveau_context_unbind(__DRIcontextPrivate *driContextPriv)
+{
+	struct nouveau_context_dri *nv = driContextPriv->driverPrivate;
+	(void)nv;
+
+	st_flush(nv->st, 0, NULL);
+	return GL_TRUE;
+}
+
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_context_dri.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_context_dri.h
new file mode 100644
index 0000000000..64cf326411
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_context_dri.h
@@ -0,0 +1,47 @@
+#ifndef __NOUVEAU_CONTEXT_DRI_H__
+#define __NOUVEAU_CONTEXT_DRI_H__
+
+#include <dri_util.h>
+#include <xmlconfig.h>
+#include <nouveau/nouveau_winsys.h>
+#include "../common/nouveau_context.h"
+
+struct nouveau_framebuffer {
+	struct st_framebuffer *stfb;
+};
+
+struct nouveau_context_dri {
+	struct nouveau_context base;
+	struct st_context *st;
+
+	/* DRI stuff */
+	__DRIscreenPrivate    *dri_screen;
+	__DRIdrawablePrivate  *dri_drawable;
+	unsigned int           last_stamp;
+	driOptionCache         dri_option_cache;
+	drm_context_t          drm_context;
+	drmLock                drm_lock;
+};
+
+extern GLboolean nouveau_context_create(const __GLcontextModes *,
+					__DRIcontextPrivate *, void *);
+extern void nouveau_context_destroy(__DRIcontextPrivate *);
+extern GLboolean nouveau_context_bind(__DRIcontextPrivate *,
+				      __DRIdrawablePrivate *draw,
+				      __DRIdrawablePrivate *read);
+extern GLboolean nouveau_context_unbind(__DRIcontextPrivate *);
+
+#ifdef DEBUG
+extern int __nouveau_debug;
+
+#define DEBUG_BO (1 << 0)
+
+#define DBG(flag, ...) do {                   \
+	if (__nouveau_debug & (DEBUG_##flag)) \
+		NOUVEAU_ERR(__VA_ARGS__);     \
+} while(0)
+#else
+#define DBG(flag, ...)
+#endif
+
+#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen_dri.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen_dri.c
new file mode 100644
index 0000000000..964a9028aa
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen_dri.c
@@ -0,0 +1,259 @@
+#include <utils.h>
+#include <vblank.h>
+#include <xmlpool.h>
+
+#include <pipe/p_context.h>
+#include <state_tracker/st_public.h>
+#include <state_tracker/st_cb_fbo.h>
+#include <nouveau_drm.h>
+#include "../common/nouveau_dri.h"
+#include "../common/nouveau_local.h"
+#include "nouveau_context_dri.h"
+#include "nouveau_screen_dri.h"
+#include "nouveau_swapbuffers.h"
+
+#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 12
+#error nouveau_drm.h version does not match expected version
+#endif
+
+/* Extension stuff, enabling of extensions handled by Gallium's GL state
+ * tracker.  But, we still need to define the entry points we want.
+ */
+#define need_GL_ARB_fragment_program
+#define need_GL_ARB_multisample
+#define need_GL_ARB_occlusion_query
+#define need_GL_ARB_point_parameters
+#define need_GL_ARB_shader_objects
+#define need_GL_ARB_texture_compression
+#define need_GL_ARB_vertex_program
+#define need_GL_ARB_vertex_shader
+#define need_GL_ARB_vertex_buffer_object
+#define need_GL_EXT_compiled_vertex_array
+#define need_GL_EXT_fog_coord
+#define need_GL_EXT_secondary_color
+#define need_GL_EXT_framebuffer_object
+#define need_GL_VERSION_2_0
+#define need_GL_VERSION_2_1
+#include "extension_helper.h"
+
+const struct dri_extension card_extensions[] =
+{
+	{ "GL_ARB_multisample", GL_ARB_multisample_functions },
+	{ "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions },
+	{ "GL_ARB_point_parameters", GL_ARB_point_parameters_functions },
+	{ "GL_ARB_shader_objects", GL_ARB_shader_objects_functions },
+	{ "GL_ARB_shading_language_100", GL_VERSION_2_0_functions },
+	{ "GL_ARB_shading_language_120", GL_VERSION_2_1_functions },
+	{ "GL_ARB_texture_compression", GL_ARB_texture_compression_functions },
+	{ "GL_ARB_vertex_program", GL_ARB_vertex_program_functions },
+	{ "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions },
+	{ "GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions },
+	{ "GL_EXT_compiled_vertex_array", GL_EXT_compiled_vertex_array_functions },
+	{ "GL_EXT_fog_coord", GL_EXT_fog_coord_functions },
+	{ "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+	{ "GL_EXT_secondary_color", GL_EXT_secondary_color_functions },
+	{ NULL, 0 }
+};
+
+PUBLIC const char __driConfigOptions[] =
+DRI_CONF_BEGIN
+DRI_CONF_END;
+static const GLuint __driNConfigOptions = 0;
+
+extern const struct dri_extension common_extensions[];
+extern const struct dri_extension nv40_extensions[];
+
+static GLboolean
+nouveau_create_buffer(__DRIscreenPrivate * driScrnPriv,
+		      __DRIdrawablePrivate * driDrawPriv,
+		      const __GLcontextModes *glVis, GLboolean pixmapBuffer)
+{
+	struct nouveau_framebuffer *nvfb;
+	enum pipe_format colour, depth, stencil;
+
+	if (pixmapBuffer)
+		return GL_FALSE;
+
+	nvfb = CALLOC_STRUCT(nouveau_framebuffer);
+	if (!nvfb)
+		return GL_FALSE;
+
+	if (glVis->redBits == 5)
+		colour = PIPE_FORMAT_R5G6B5_UNORM;
+	else
+		colour = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+	if (glVis->depthBits == 16)
+		depth = PIPE_FORMAT_Z16_UNORM;
+	else if (glVis->depthBits == 24)
+		depth = PIPE_FORMAT_Z24S8_UNORM;
+	else
+		depth = PIPE_FORMAT_NONE;
+
+	if (glVis->stencilBits == 8)
+		stencil = PIPE_FORMAT_Z24S8_UNORM;
+	else
+		stencil = PIPE_FORMAT_NONE;
+
+	nvfb->stfb = st_create_framebuffer(glVis, colour, depth, stencil,
+					   driDrawPriv->w, driDrawPriv->h,
+					   (void*)nvfb);
+	if (!nvfb->stfb) {
+		free(nvfb);
+		return  GL_FALSE;
+	}
+
+	driDrawPriv->driverPrivate = (void *)nvfb;
+	return GL_TRUE;
+}
+
+static void
+nouveau_destroy_buffer(__DRIdrawablePrivate * driDrawPriv)
+{
+	struct nouveau_framebuffer *nvfb;
+	
+	nvfb = (struct nouveau_framebuffer *)driDrawPriv->driverPrivate;
+	st_unreference_framebuffer(nvfb->stfb);
+	free(nvfb);
+}
+
+static __DRIconfig **
+nouveau_fill_in_modes(__DRIscreenPrivate *psp,
+		      unsigned pixel_bits, unsigned depth_bits,
+		      unsigned stencil_bits, GLboolean have_back_buffer)
+{
+	__DRIconfig **configs;
+	unsigned depth_buffer_factor;
+	unsigned back_buffer_factor;
+	GLenum fb_format;
+	GLenum fb_type;
+
+	static const GLenum back_buffer_modes[] = {
+		GLX_NONE, GLX_SWAP_UNDEFINED_OML,
+	};
+
+	uint8_t depth_bits_array[3];
+	uint8_t stencil_bits_array[3];
+	uint8_t msaa_samples_array[1];
+
+	depth_bits_array[0] = 0;
+	depth_bits_array[1] = depth_bits;
+	depth_bits_array[2] = depth_bits;
+
+	/* Just like with the accumulation buffer, always provide some modes
+	 * with a stencil buffer.  It will be a sw fallback, but some apps won't
+	 * care about that.
+	 */
+	stencil_bits_array[0] = 0;
+	stencil_bits_array[1] = 0;
+	if (depth_bits == 24)
+		stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
+	stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
+
+	msaa_samples_array[0] = 0;
+
+	depth_buffer_factor =
+		((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
+	back_buffer_factor = (have_back_buffer) ? 3 : 1;
+
+	if (pixel_bits == 16) {
+		fb_format = GL_RGB;
+		fb_type = GL_UNSIGNED_SHORT_5_6_5;
+	}
+	else {
+		fb_format = GL_BGRA;
+		fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+	}
+
+	configs = driCreateConfigs(fb_format, fb_type,
+				   depth_bits_array, stencil_bits_array,
+				   depth_buffer_factor, back_buffer_modes,
+				   back_buffer_factor, msaa_samples_array, 1);
+	if (configs == NULL) {
+	 fprintf(stderr, "[%s:%u] Error creating FBConfig!\n",
+			 __func__, __LINE__);
+		return NULL;
+	}
+
+	return configs;
+}
+
+static const __DRIconfig **
+nouveau_screen_create(__DRIscreenPrivate *psp)
+{
+	struct nouveau_dri *nv_dri = psp->pDevPriv;
+	struct nouveau_screen_dri *nv_screen;
+	static const __DRIversion ddx_expected =
+		{ 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL };
+	static const __DRIversion dri_expected = { 4, 0, 0 };
+	static const __DRIversion drm_expected =
+		{ 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL };
+
+	if (!driCheckDriDdxDrmVersions2("nouveau",
+					&psp->dri_version, &dri_expected,
+					&psp->ddx_version, &ddx_expected,
+					&psp->drm_version, &drm_expected)) {
+		return NULL;
+	}
+
+	if (drm_expected.patch != psp->drm_version.patch) {
+		fprintf(stderr, "Incompatible DRM patch level.\n"
+				"Expected: %d\n" "Current : %d\n",
+			drm_expected.patch, psp->drm_version.patch);
+		return NULL;
+	}
+
+	driInitExtensions(NULL, card_extensions, GL_FALSE);
+
+	if (psp->devPrivSize != sizeof(struct nouveau_dri)) {
+		NOUVEAU_ERR("DRI struct mismatch between DDX/DRI\n");
+		return NULL;
+	}
+
+	nv_screen = CALLOC_STRUCT(nouveau_screen_dri);
+	if (!nv_screen)
+		return NULL;
+
+	driParseOptionInfo(&nv_screen->option_cache,
+			   __driConfigOptions, __driNConfigOptions);
+
+	if (nouveau_screen_init(nv_dri, psp->fd, &nv_screen->base)) {
+		FREE(nv_screen);
+		return NULL;
+	}
+
+	nv_screen->driScrnPriv = psp;
+	psp->private = (void *)nv_screen;
+
+	return (const __DRIconfig **)
+		nouveau_fill_in_modes(psp, nv_dri->bpp,
+				      (nv_dri->bpp == 16) ? 16 : 24,
+				      (nv_dri->bpp == 16) ? 0 : 8, 1);
+}
+
+static void
+nouveau_screen_destroy(__DRIscreenPrivate *driScrnPriv)
+{
+	struct nouveau_screen_dri *nv_screen = driScrnPriv->private;
+
+	driScrnPriv->private = NULL;
+	nouveau_screen_cleanup(&nv_screen->base);
+	FREE(nv_screen);
+}
+
+const struct __DriverAPIRec
+driDriverAPI = {
+	.InitScreen	= nouveau_screen_create,
+	.DestroyScreen	= nouveau_screen_destroy,
+	.CreateContext	= nouveau_context_create,
+	.DestroyContext	= nouveau_context_destroy,
+	.CreateBuffer	= nouveau_create_buffer,
+	.DestroyBuffer	= nouveau_destroy_buffer,
+	.SwapBuffers	= nouveau_swap_buffers,
+	.MakeCurrent	= nouveau_context_bind,
+	.UnbindContext	= nouveau_context_unbind,
+	.CopySubBuffer	= nouveau_copy_sub_buffer,
+
+	.InitScreen2	= NULL, /* one day, I promise! */
+};
+
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_screen_dri.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen_dri.h
new file mode 100644
index 0000000000..1498087819
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_screen_dri.h
@@ -0,0 +1,13 @@
+#ifndef __NOUVEAU_SCREEN_DRI_H__
+#define __NOUVEAU_SCREEN_DRI_H__
+
+#include "../common/nouveau_screen.h"
+#include "xmlconfig.h"
+
+struct nouveau_screen_dri {
+	struct nouveau_screen base;
+	__DRIscreenPrivate *driScrnPriv;
+	driOptionCache      option_cache;
+};
+
+#endif
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c
new file mode 100644
index 0000000000..58cb6f7265
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.c
@@ -0,0 +1,113 @@
+#include <main/glheader.h>
+#include <glapi/glthread.h>
+#include <GL/internal/glcore.h>
+
+#include <pipe/p_context.h>
+#include <state_tracker/st_public.h>
+#include <state_tracker/st_context.h>
+#include <state_tracker/st_cb_fbo.h>
+
+#include "../common/nouveau_local.h"
+#include "nouveau_context_dri.h"
+#include "nouveau_screen_dri.h"
+#include "nouveau_swapbuffers.h"
+
+void
+nouveau_copy_buffer(__DRIdrawablePrivate *dPriv, struct pipe_surface *surf,
+		    const drm_clip_rect_t *rect)
+{
+	struct nouveau_context_dri *nv = dPriv->driContextPriv->driverPrivate;
+	struct pipe_context *pipe = nv->base.nvc->pctx[nv->base.pctx_id];
+	drm_clip_rect_t *pbox;
+	int nbox, i;
+
+	LOCK_HARDWARE(&nv->base);
+	if (!dPriv->numClipRects) {
+		UNLOCK_HARDWARE(&nv->base);
+		return;
+	}
+	pbox = dPriv->pClipRects;
+	nbox = dPriv->numClipRects;
+
+	for (i = 0; i < nbox; i++, pbox++) {
+		int sx, sy, dx, dy, w, h;
+
+		sx = pbox->x1 - dPriv->x;
+		sy = pbox->y1 - dPriv->y;
+		dx = pbox->x1;
+		dy = pbox->y1;
+		w  = pbox->x2 - pbox->x1;
+		h  = pbox->y2 - pbox->y1;
+
+		pipe->surface_copy(pipe, FALSE, nv->base.frontbuffer,
+				   dx, dy, surf, sx, sy, w, h);
+	}
+
+	FIRE_RING(nv->base.nvc->channel);
+	UNLOCK_HARDWARE(&nv->base);
+
+	if (nv->last_stamp != dPriv->lastStamp) {
+		struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
+		st_resize_framebuffer(nvfb->stfb, dPriv->w, dPriv->h);
+		nv->last_stamp = dPriv->lastStamp;
+	}
+}
+
+void
+nouveau_copy_sub_buffer(__DRIdrawablePrivate *dPriv, int x, int y, int w, int h)
+{
+	struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
+	struct pipe_surface *surf;
+
+	st_get_framebuffer_surface(nvfb->stfb, ST_SURFACE_BACK_LEFT, &surf);
+	if (surf) {
+		drm_clip_rect_t rect;
+		rect.x1 = x;
+		rect.y1 = y;
+		rect.x2 = x + w;
+		rect.y2 = y + h;
+
+		st_notify_swapbuffers(nvfb->stfb);
+		nouveau_copy_buffer(dPriv, surf, &rect);
+	}
+}
+
+void
+nouveau_swap_buffers(__DRIdrawablePrivate *dPriv)
+{
+	struct nouveau_framebuffer *nvfb = dPriv->driverPrivate;
+	struct pipe_surface *surf;
+
+	st_get_framebuffer_surface(nvfb->stfb, ST_SURFACE_BACK_LEFT, &surf);
+	if (surf) {
+		st_notify_swapbuffers(nvfb->stfb);
+		nouveau_copy_buffer(dPriv, surf, NULL);
+	}
+}
+
+void
+nouveau_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *surf,
+			  void *context_private)
+{
+	struct nouveau_context_dri *nv = context_private;
+	__DRIdrawablePrivate *dPriv = nv->dri_drawable;
+
+	nouveau_copy_buffer(dPriv, surf, NULL);
+}
+
+void
+nouveau_contended_lock(struct nouveau_context *nv)
+{
+	struct nouveau_context_dri *nv_sub = (struct nouveau_context_dri*)nv;
+	__DRIdrawablePrivate *dPriv = nv_sub->dri_drawable;
+	__DRIscreenPrivate *sPriv = nv_sub->dri_screen;
+
+	/* If the window moved, may need to set a new cliprect now.
+	 *
+	 * NOTE: This releases and regains the hw lock, so all state
+	 * checking must be done *after* this call:
+	 */
+	if (dPriv)
+		DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv);
+}
+
diff --git a/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h
new file mode 100644
index 0000000000..825d3da6da
--- /dev/null
+++ b/src/gallium/winsys/drm/nouveau/dri/nouveau_swapbuffers.h
@@ -0,0 +1,10 @@
+#ifndef __NOUVEAU_SWAPBUFFERS_H__
+#define __NOUVEAU_SWAPBUFFERS_H__
+
+extern void nouveau_copy_buffer(__DRIdrawablePrivate *, struct pipe_surface *,
+				const drm_clip_rect_t *);
+extern void nouveau_copy_sub_buffer(__DRIdrawablePrivate *,
+				    int x, int y, int w, int h);
+extern void nouveau_swap_buffers(__DRIdrawablePrivate *);
+
+#endif
diff --git a/src/gallium/winsys/drm/radeon/Makefile b/src/gallium/winsys/drm/radeon/Makefile
new file mode 100644
index 0000000000..dca1e3233a
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/Makefile
@@ -0,0 +1,32 @@
+
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = radeon_dri.so
+
+MINIGLX_SOURCES =
+
+PIPE_DRIVERS = \
+	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/r300/libr300.a
+
+DRIVER_SOURCES = \
+	radeon_buffer.c \
+	radeon_context.c \
+	radeon_r300.c \
+	radeon_screen.c \
+	radeon_winsys_softpipe.c
+
+C_SOURCES = \
+	$(COMMON_GALLIUM_SOURCES) \
+	$(DRIVER_SOURCES)
+
+ASM_SOURCES = 
+
+DRIVER_DEFINES = -I../../../drivers/r300
+
+include ../Makefile.template
+
+DRI_LIB_DEPS += -ldrm_radeon
+
+symlinks:
diff --git a/src/gallium/winsys/drm/radeon/SConscript b/src/gallium/winsys/drm/radeon/SConscript
new file mode 100644
index 0000000000..2435211a32
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/SConscript
@@ -0,0 +1,29 @@
+Import('*')
+
+if 'mesa' in env['statetrackers']:
+
+	env = drienv.Clone()
+
+	DRIVER_SOURCES = [
+		'radeon_buffer.c',
+		'radeon_context.c',
+		'radeon_screen.c',
+		'radeon_winsys_softpipe.c',
+	]
+
+	sources = \
+		COMMON_GALLIUM_SOURCES + \
+		DRIVER_SOURCES
+
+	drivers = [
+		softpipe,
+		r300
+	]
+
+	# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions
+	env.SharedLibrary(
+		target ='radeon_dri.so',
+		source = sources,
+		LIBS = drivers + mesa + auxiliaries + env['LIBS'],
+	)
+
diff --git a/src/gallium/winsys/drm/radeon/radeon_buffer.c b/src/gallium/winsys/drm/radeon/radeon_buffer.c
new file mode 100644
index 0000000000..259a505c0a
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/radeon_buffer.c
@@ -0,0 +1,239 @@
+/* 
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#include <stdio.h>
+#include "dri_util.h"
+#include "state_tracker/st_public.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "radeon_buffer.h"
+#include "radeon_screen.h"
+#include "radeon_context.h"
+#include "radeon_bo.h"
+#include "radeon_drm.h"
+
+static const char *radeon_get_name(struct pipe_winsys *ws)
+{
+    return "RADEON/DRI2";
+}
+
+static struct pipe_buffer *radeon_buffer_create(struct pipe_winsys *ws,
+                                             unsigned alignment,
+                                             unsigned usage,
+                                             unsigned size)
+{
+    struct radeon_pipe_winsys *radeon_ws = (struct radeon_pipe_winsys *)ws;
+    struct radeon_pipe_buffer *radeon_buffer;
+    uint32_t domain;
+
+    radeon_buffer = calloc(1, sizeof(*radeon_buffer));
+    if (radeon_buffer == NULL) {
+        return NULL;
+    }
+    radeon_buffer->base.refcount = 1;
+    radeon_buffer->base.alignment = alignment;
+    radeon_buffer->base.usage = usage;
+    radeon_buffer->base.size = size;
+
+    domain = 0;
+
+    if (usage & PIPE_BUFFER_USAGE_PIXEL) {
+        domain |= RADEON_GEM_DOMAIN_VRAM;
+    }
+    if (usage & PIPE_BUFFER_USAGE_VERTEX) {
+        domain |= RADEON_GEM_DOMAIN_GTT;
+    }
+
+    if (usage & PIPE_BUFFER_USAGE_INDEX) {
+        domain |= RADEON_GEM_DOMAIN_GTT;
+    }
+    radeon_buffer->bo = radeon_bo_open(radeon_ws->radeon_screen->bom, 0,
+                                    size, alignment, domain, 0);
+    if (radeon_buffer->bo == NULL) {
+        free(radeon_buffer);
+    }
+    return &radeon_buffer->base;
+}
+
+static struct pipe_buffer *radeon_buffer_user_create(struct pipe_winsys *ws,
+                                                  void *ptr,
+                                                  unsigned bytes)
+{
+    struct radeon_pipe_buffer *radeon_buffer;
+
+    radeon_buffer = (struct radeon_pipe_buffer*)radeon_buffer_create(ws, 0, 0, bytes);
+    if (radeon_buffer == NULL) {
+        return NULL;
+    }
+    radeon_bo_map(radeon_buffer->bo, 1);
+    memcpy(radeon_buffer->bo->ptr, ptr, bytes);
+    radeon_bo_unmap(radeon_buffer->bo);
+    return &radeon_buffer->base;
+}
+
+static void radeon_buffer_del(struct pipe_winsys *ws, struct pipe_buffer *buffer)
+{
+    struct radeon_pipe_buffer *radeon_buffer = (struct radeon_pipe_buffer*)buffer;
+
+    radeon_bo_unref(radeon_buffer->bo);
+    free(radeon_buffer);
+}
+
+static void *radeon_buffer_map(struct pipe_winsys *ws,
+                            struct pipe_buffer *buffer,
+                            unsigned flags)
+{
+    struct radeon_pipe_buffer *radeon_buffer = (struct radeon_pipe_buffer*)buffer;
+    int write = 0;
+
+    if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) {
+        write = 1;
+    }
+    if (radeon_bo_map(radeon_buffer->bo, write))
+        return NULL;
+    return radeon_buffer->bo->ptr;
+}
+
+static void radeon_buffer_unmap(struct pipe_winsys *ws, struct pipe_buffer *buffer)
+{
+    struct radeon_pipe_buffer *radeon_buffer = (struct radeon_pipe_buffer*)buffer;
+
+    radeon_bo_unmap(radeon_buffer->bo);
+}
+
+static void radeon_fence_reference(struct pipe_winsys *ws,
+                                struct pipe_fence_handle **ptr,
+                                struct pipe_fence_handle *pfence)
+{
+}
+
+static int radeon_fence_signalled(struct pipe_winsys *ws,
+                               struct pipe_fence_handle *pfence,
+                               unsigned flag)
+{
+    return 1;
+}
+
+static int radeon_fence_finish(struct pipe_winsys *ws,
+                            struct pipe_fence_handle *pfence,
+                            unsigned flag)
+{
+    return 0;
+}
+
+static void radeon_flush_frontbuffer(struct pipe_winsys *pipe_winsys,
+                                  struct pipe_surface *pipe_surface,
+                                  void *context_private)
+{
+    /* TODO: call dri2CopyRegion */
+}
+
+struct pipe_winsys *radeon_pipe_winsys(struct radeon_screen *radeon_screen)
+{
+    struct radeon_pipe_winsys *radeon_ws;
+
+    radeon_ws = calloc(1, sizeof(struct radeon_pipe_winsys));
+    if (radeon_ws == NULL) {
+        return NULL;
+    }
+    radeon_ws->radeon_screen = radeon_screen;
+
+    radeon_ws->winsys.flush_frontbuffer = radeon_flush_frontbuffer;
+
+    radeon_ws->winsys.buffer_create = radeon_buffer_create;
+    radeon_ws->winsys.buffer_destroy = radeon_buffer_del;
+    radeon_ws->winsys.user_buffer_create = radeon_buffer_user_create;
+    radeon_ws->winsys.buffer_map = radeon_buffer_map;
+    radeon_ws->winsys.buffer_unmap = radeon_buffer_unmap;
+
+    radeon_ws->winsys.fence_reference = radeon_fence_reference;
+    radeon_ws->winsys.fence_signalled = radeon_fence_signalled;
+    radeon_ws->winsys.fence_finish = radeon_fence_finish;
+
+    radeon_ws->winsys.get_name = radeon_get_name;
+
+    return &radeon_ws->winsys;
+}
+
+static struct pipe_buffer *radeon_buffer_from_handle(struct radeon_screen *radeon_screen,
+                                                  uint32_t handle)
+{
+    struct radeon_pipe_buffer *radeon_buffer;
+    struct radeon_bo *bo = NULL;
+
+    bo = radeon_bo_open(radeon_screen->bom, handle, 0, 0, 0, 0);
+    if (bo == NULL) {
+        return NULL;
+    }
+    radeon_buffer = calloc(1, sizeof(struct radeon_pipe_buffer));
+    if (radeon_buffer == NULL) {
+        radeon_bo_unref(bo);
+        return NULL;
+    }
+    radeon_buffer->base.refcount = 1;
+    radeon_buffer->base.usage = PIPE_BUFFER_USAGE_PIXEL;
+    radeon_buffer->bo = bo;
+    return &radeon_buffer->base;
+}
+
+struct pipe_surface *radeon_surface_from_handle(struct radeon_context *radeon_context,
+                                             uint32_t handle,
+                                             enum pipe_format format,
+                                             int w, int h, int pitch)
+{
+    struct pipe_screen *pipe_screen = radeon_context->pipe_screen;
+    struct pipe_winsys *pipe_winsys = radeon_context->pipe_winsys;
+    struct pipe_texture tmpl;
+    struct pipe_surface *ps;
+    struct pipe_texture *pt;
+    struct pipe_buffer *pb;
+
+    pb = radeon_buffer_from_handle(radeon_context->radeon_screen, handle);
+    if (pb == NULL) {
+        return NULL;
+    }
+    memset(&tmpl, 0, sizeof(tmpl));
+    tmpl.tex_usage = PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+    tmpl.target = PIPE_TEXTURE_2D;
+    tmpl.width[0] = w;
+    tmpl.height[0] = h;
+    tmpl.depth[0] = 1;
+    tmpl.format = format;
+    pf_get_block(tmpl.format, &tmpl.block);
+    tmpl.nblocksx[0] = pf_get_nblocksx(&tmpl.block, w);
+    tmpl.nblocksy[0] = pf_get_nblocksy(&tmpl.block, h);
+
+    pt = pipe_screen->texture_blanket(pipe_screen, &tmpl, &pitch, pb);
+    if (pt == NULL) {
+        pipe_buffer_reference(pipe_screen, &pb, NULL);
+    }
+    ps = pipe_screen->get_tex_surface(pipe_screen, pt, 0, 0, 0,
+                                      PIPE_BUFFER_USAGE_GPU_WRITE);
+    return ps;
+}
diff --git a/src/gallium/winsys/drm/radeon/radeon_buffer.h b/src/gallium/winsys/drm/radeon/radeon_buffer.h
new file mode 100644
index 0000000000..c626c20229
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/radeon_buffer.h
@@ -0,0 +1,54 @@
+/* 
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_BUFFER_H
+#define RADEON_BUFFER_H
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "radeon_screen.h"
+#include "radeon_context.h"
+#include "radeon_bo.h"
+
+struct radeon_pipe_buffer {
+    struct pipe_buffer  base;
+    struct radeon_bo    *bo;
+};
+
+struct radeon_pipe_winsys {
+    struct pipe_winsys      winsys;
+    struct radeon_screen       *radeon_screen;
+};
+
+struct pipe_winsys *radeon_pipe_winsys(struct radeon_screen *radeon_screen);
+struct pipe_surface *radeon_surface_from_handle(struct radeon_context *radeon_context,
+                                             uint32_t handle,
+                                             enum pipe_format format,
+                                             int w, int h, int pitch);
+
+#endif
diff --git a/src/gallium/winsys/drm/radeon/radeon_context.c b/src/gallium/winsys/drm/radeon/radeon_context.c
new file mode 100644
index 0000000000..a9d1577634
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/radeon_context.c
@@ -0,0 +1,306 @@
+/* 
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#include <stdio.h>
+#include "dri_util.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_context.h"
+#include "radeon_screen.h"
+#include "radeon_context.h"
+#include "radeon_buffer.h"
+#include "radeon_winsys_softpipe.h"
+
+#define need_GL_ARB_fragment_program
+#define need_GL_ARB_multisample
+#define need_GL_ARB_point_parameters
+#define need_GL_ARB_shader_objects
+#define need_GL_ARB_texture_compression
+#define need_GL_ARB_vertex_buffer_object
+#define need_GL_ARB_vertex_program
+#define need_GL_ARB_vertex_shader
+#define need_GL_EXT_blend_color
+#define need_GL_EXT_blend_equation_separate
+#define need_GL_EXT_blend_func_separate
+#define need_GL_EXT_blend_minmax
+#define need_GL_EXT_cull_vertex
+#define need_GL_EXT_compiled_vertex_array
+#define need_GL_EXT_fog_coord
+#define need_GL_EXT_framebuffer_object
+#define need_GL_EXT_multi_draw_arrays
+#define need_GL_EXT_secondary_color
+#define need_GL_VERSION_2_0
+#define need_GL_VERSION_2_1
+#include "extension_helper.h"
+
+/**
+ * Extension strings exported by the radeon driver.
+ */
+const struct dri_extension radeon_card_extensions[] = {
+   {"GL_ARB_multitexture", NULL},
+   {"GL_ARB_texture_border_clamp", NULL},
+   {"GL_ARB_texture_rectangle", NULL},
+   {"GL_ARB_pixel_buffer_object", NULL},
+   {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions},
+   {"GL_ARB_shader_objects", GL_ARB_shader_objects_functions},
+   {"GL_ARB_shading_language_100", GL_VERSION_2_0_functions},
+   {"GL_ARB_shading_language_120", GL_VERSION_2_1_functions},
+   {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions},
+   {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
+   {"GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions},
+   {"GL_EXT_blend_color", GL_EXT_blend_color_functions},
+   {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions},
+   {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
+   {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
+   {"GL_EXT_blend_subtract", NULL},
+   {"GL_EXT_compiled_vertex_array", GL_EXT_compiled_vertex_array_functions},
+   {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions},
+   {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions},
+   {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions},
+   {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions},
+   {"GL_EXT_packed_depth_stencil", NULL},
+   {"GL_EXT_pixel_buffer_object", NULL},
+   {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions},
+   {"GL_EXT_stencil_wrap", NULL},
+   {NULL, NULL}
+};
+
+static void radeon_update_renderbuffers(__DRIcontext *dri_context,
+                                     __DRIdrawable *dri_drawable)
+{
+    struct radeon_framebuffer *radeon_fb;
+    struct radeon_context *radeon_context;
+    unsigned attachments[10];
+    __DRIbuffer *buffers;
+    __DRIscreen *screen;
+    int i, count;
+
+    radeon_context = dri_context->driverPrivate;
+    screen = dri_drawable->driScreenPriv;
+    radeon_fb = dri_drawable->driverPrivate;
+    for (count = 0, i = 0; count < 6; count++) {
+        if (radeon_fb->attachments & (1 << count)) {
+            attachments[i++] = count;
+        }
+    }
+
+    buffers = (*screen->dri2.loader->getBuffers)(dri_drawable,
+                                                 &dri_drawable->w,
+                                                 &dri_drawable->h,
+                                                 attachments,
+                                                 i,
+                                                 &count,
+                                                 dri_drawable->loaderPrivate);
+    if (buffers == NULL) {
+        return;
+    }
+
+    /* set one cliprect to cover the whole dri_drawable */
+    dri_drawable->x = 0;
+    dri_drawable->y = 0;
+    dri_drawable->backX = 0;
+    dri_drawable->backY = 0;
+    dri_drawable->numClipRects = 1;
+    dri_drawable->pClipRects[0].x1 = 0;
+    dri_drawable->pClipRects[0].y1 = 0;
+    dri_drawable->pClipRects[0].x2 = dri_drawable->w;
+    dri_drawable->pClipRects[0].y2 = dri_drawable->h;
+    dri_drawable->numBackClipRects = 1;
+    dri_drawable->pBackClipRects[0].x1 = 0;
+    dri_drawable->pBackClipRects[0].y1 = 0;
+    dri_drawable->pBackClipRects[0].x2 = dri_drawable->w;
+    dri_drawable->pBackClipRects[0].y2 = dri_drawable->h;
+
+    for (i = 0; i < count; i++) {
+        struct pipe_surface *ps;
+        enum pipe_format format = 0;
+        int index = 0;
+
+        switch (buffers[i].attachment) {
+        case __DRI_BUFFER_FRONT_LEFT:
+            index = ST_SURFACE_FRONT_LEFT;
+            switch (buffers[i].cpp) {
+            case 4:
+                format = PIPE_FORMAT_A8R8G8B8_UNORM;
+                break;
+            case 2:
+                format = PIPE_FORMAT_R5G6B5_UNORM;
+                break;
+            default:
+                /* FIXME: error */
+                return;
+            }
+            break;
+        case __DRI_BUFFER_BACK_LEFT:
+            index = ST_SURFACE_BACK_LEFT;
+            switch (buffers[i].cpp) {
+            case 4:
+                format = PIPE_FORMAT_A8R8G8B8_UNORM;
+                break;
+            case 2:
+                format = PIPE_FORMAT_R5G6B5_UNORM;
+                break;
+            default:
+                /* FIXME: error */
+                return;
+            }
+            break;
+        case __DRI_BUFFER_STENCIL:
+        case __DRI_BUFFER_DEPTH:
+            index = ST_SURFACE_DEPTH;
+            switch (buffers[i].cpp) {
+            case 4:
+                format = PIPE_FORMAT_Z24S8_UNORM;
+                break;
+            case 2:
+                format = PIPE_FORMAT_Z16_UNORM;
+                break;
+            default:
+                /* FIXME: error */
+                return;
+            }
+            break;
+        case __DRI_BUFFER_ACCUM:
+        default:
+            fprintf(stderr,
+                    "unhandled buffer attach event, attacment type %d\n",
+                    buffers[i].attachment);
+            return;
+        }
+
+        ps = radeon_surface_from_handle(radeon_context,
+                                     buffers[i].name,
+                                     format,
+                                     dri_drawable->w,
+                                     dri_drawable->h,
+                                     buffers[i].pitch);
+        assert(ps);
+        st_set_framebuffer_surface(radeon_fb->st_framebuffer, index, ps);
+    }
+    st_resize_framebuffer(radeon_fb->st_framebuffer,
+                          dri_drawable->w,
+                          dri_drawable->h);
+}
+
+GLboolean radeon_context_create(const __GLcontextModes *visual,
+                             __DRIcontextPrivate *dri_context,
+                             void *shared_context)
+{
+    __DRIscreenPrivate *dri_screen;
+    struct radeon_context *radeon_context;
+    struct radeon_screen *radeon_screen;
+    struct pipe_context *pipe;
+    struct st_context *shared_st_context = NULL;
+
+    dri_context->driverPrivate = NULL;
+    radeon_context = calloc(1, sizeof(struct radeon_context));
+    if (radeon_context == NULL) {
+        return GL_FALSE;
+    }
+
+    if (shared_context) {
+        shared_st_context = ((struct radeon_context*)shared_context)->st_context;
+    }
+
+    dri_screen = dri_context->driScreenPriv;
+    radeon_screen = dri_screen->private;
+    radeon_context->dri_screen = dri_screen;
+    radeon_context->radeon_screen = radeon_screen;
+    radeon_context->drm_fd = dri_screen->fd;
+
+    radeon_context->pipe_winsys = radeon_pipe_winsys(radeon_screen);
+    if (radeon_context->pipe_winsys == NULL) {
+        free(radeon_context);
+        return GL_FALSE;
+    }
+
+    if (!getenv("RADEON_SOFTPIPE")) {
+        fprintf(stderr, "Creating r300 context...\n");
+        pipe =
+            r300_create_context(NULL,
+                                radeon_context->pipe_winsys,
+                                radeon_create_r300_winsys(radeon_context->drm_fd));
+        radeon_context->pipe_screen = pipe->screen;
+    } else {
+        pipe = radeon_create_softpipe(radeon_context);
+    }
+    radeon_context->st_context = st_create_context(pipe, visual,
+                                                shared_st_context);
+    driInitExtensions(radeon_context->st_context->ctx,
+                      radeon_card_extensions, GL_TRUE);
+    dri_context->driverPrivate = radeon_context;
+    return GL_TRUE;
+}
+
+void radeon_context_destroy(__DRIcontextPrivate *dri_context)
+{
+    struct radeon_context *radeon_context;
+
+    radeon_context = dri_context->driverPrivate;
+    st_finish(radeon_context->st_context);
+    st_destroy_context(radeon_context->st_context);
+    free(radeon_context);
+}
+
+GLboolean radeon_context_bind(__DRIcontextPrivate *dri_context,
+                           __DRIdrawablePrivate *dri_drawable,
+                           __DRIdrawablePrivate *dri_readable)
+{
+    struct radeon_framebuffer *drawable;
+    struct radeon_framebuffer *readable;
+    struct radeon_context *radeon_context;
+
+    if (dri_context == NULL) {
+        st_make_current(NULL, NULL, NULL);
+        return GL_TRUE;
+    }
+
+    radeon_context = dri_context->driverPrivate;
+    drawable = dri_drawable->driverPrivate;
+    readable = dri_readable->driverPrivate;
+    st_make_current(radeon_context->st_context,
+                    drawable->st_framebuffer,
+                    readable->st_framebuffer);
+
+    radeon_update_renderbuffers(dri_context, dri_drawable);
+    if (dri_drawable != dri_readable) {
+        radeon_update_renderbuffers(dri_context, dri_readable);
+    }
+    return GL_TRUE;
+}
+
+GLboolean radeon_context_unbind(__DRIcontextPrivate *dri_context)
+{
+    struct radeon_context *radeon_context;
+
+    radeon_context = dri_context->driverPrivate;
+    st_flush(radeon_context->st_context, PIPE_FLUSH_RENDER_CACHE, NULL);
+    return GL_TRUE;
+}
diff --git a/src/gallium/winsys/drm/radeon/radeon_context.h b/src/gallium/winsys/drm/radeon/radeon_context.h
new file mode 100644
index 0000000000..d7222b4469
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/radeon_context.h
@@ -0,0 +1,70 @@
+/* 
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_CONTEXT_H
+#define RADEON_CONTEXT_H
+
+#include "dri_util.h"
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_context.h"
+#include "radeon_screen.h"
+
+#include "radeon_r300.h"
+
+struct radeon_framebuffer {
+    struct st_framebuffer   *st_framebuffer;
+    unsigned                attachments;
+};
+
+struct radeon_context {
+    /* st */
+    struct st_context       *st_context;
+    /* pipe */
+    struct pipe_screen      *pipe_screen;
+    struct pipe_winsys      *pipe_winsys;
+    /* DRI */
+    __DRIscreenPrivate      *dri_screen;
+    __DRIdrawablePrivate    *dri_drawable;
+    __DRIdrawablePrivate    *dri_readable;
+    /* DRM */
+    int                     drm_fd;
+   /* RADEON */
+    struct radeon_screen       *radeon_screen;
+};
+
+GLboolean radeon_context_create(const __GLcontextModes*,
+                             __DRIcontextPrivate*,
+                             void*);
+void radeon_context_destroy(__DRIcontextPrivate*);
+GLboolean radeon_context_bind(__DRIcontextPrivate*,
+                           __DRIdrawablePrivate*,
+                           __DRIdrawablePrivate*);
+GLboolean radeon_context_unbind(__DRIcontextPrivate*);
+
+#endif
diff --git a/src/gallium/winsys/drm/radeon/radeon_r300.c b/src/gallium/winsys/drm/radeon/radeon_r300.c
new file mode 100644
index 0000000000..8fe2375e34
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/radeon_r300.c
@@ -0,0 +1,96 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "radeon_r300.h"
+
+static boolean radeon_r300_check_cs(struct radeon_cs* cs, int size)
+{
+    /* XXX check size here, lazy ass! */
+    return TRUE;
+}
+
+static void radeon_r300_write_cs_reloc(struct radeon_cs* cs,
+                                    struct pipe_buffer* pbuffer,
+                                    uint32_t rd,
+                                    uint32_t wd,
+                                    uint32_t flags)
+{
+    radeon_cs_write_reloc(cs, ((struct radeon_pipe_buffer*)pbuffer)->bo, rd, wd, flags);
+}
+
+static void radeon_r300_flush_cs(struct radeon_cs* cs)
+{
+    radeon_cs_emit(cs);
+    radeon_cs_erase(cs);
+}
+
+/* Helper function to do the ioctls needed for setup and init. */
+static void do_ioctls(struct r300_winsys* winsys, int fd)
+{
+    drm_radeon_getparam_t gp;
+    uint32_t target;
+    int retval;
+
+    /* XXX is this cast safe? */
+    gp.value = (int*)&target;
+
+    /* First, get PCI ID */
+    gp.param = RADEON_PARAM_DEVICE_ID;
+    retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
+    if (retval) {
+        fprintf(stderr, "%s: Failed to get PCI ID, error number %d",
+                __FUNCTION__, retval);
+        exit(1);
+    }
+    winsys->pci_id = target;
+
+    /* Then, get the number of pixel pipes */
+    gp.param = RADEON_PARAM_NUM_GB_PIPES;
+    retval = drmCommandWriteRead(fd, DRM_RADEON_GETPARAM, &gp, sizeof(gp));
+    if (retval) {
+        fprintf(stderr, "%s: Failed to get GB pipe count, error number %d",
+                __FUNCTION__, retval);
+        exit(1);
+    }
+    winsys->gb_pipes = target;
+
+}
+
+struct r300_winsys* radeon_create_r300_winsys(int fd)
+{
+    struct r300_winsys* winsys = calloc(1, sizeof(struct r300_winsys));
+
+    do_ioctls(winsys, fd);
+
+    struct radeon_cs_manager* csm = radeon_cs_manager_gem_ctor(fd);
+
+    winsys->cs = radeon_cs_create(csm, 1024 * 64 / 4);
+
+    winsys->check_cs = radeon_r300_check_cs;
+    winsys->begin_cs = radeon_cs_begin;
+    winsys->write_cs_dword = radeon_cs_write_dword;
+    winsys->write_cs_reloc = radeon_r300_write_cs_reloc;
+    winsys->end_cs = radeon_cs_end;
+    winsys->flush_cs = radeon_r300_flush_cs;
+
+    return winsys;
+}
diff --git a/src/gallium/winsys/drm/radeon/radeon_r300.h b/src/gallium/winsys/drm/radeon/radeon_r300.h
new file mode 100644
index 0000000000..8ed95a3a9b
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/radeon_r300.h
@@ -0,0 +1,34 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+/* XXX WTF is this! I shouldn't have to include those first three! FUCK! */
+#include <stdint.h>
+#include <stdlib.h>
+#include "drm.h"
+#include "radeon_drm.h"
+#include "radeon_cs.h"
+
+#include "r300_winsys.h"
+
+#include "radeon_buffer.h"
+
+struct r300_winsys* radeon_create_r300_winsys(int fd);
diff --git a/src/gallium/winsys/drm/radeon/radeon_screen.c b/src/gallium/winsys/drm/radeon/radeon_screen.c
new file mode 100644
index 0000000000..e31caff0bf
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/radeon_screen.c
@@ -0,0 +1,288 @@
+/* 
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#include <stdio.h>
+#include "pipe/p_screen.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_context.h"
+#include "utils.h"
+#include "xf86drm.h"
+#include "drm.h"
+#include "dri_util.h"
+#include "radeon_screen.h"
+#include "radeon_context.h"
+#include "radeon_buffer.h"
+#include "radeon_bo.h"
+#include "radeon_bo_gem.h"
+#include "radeon_drm.h"
+
+extern const struct dri_extension radeon_card_extensions[];
+
+static const __DRIextension *radeon_screen_extensions[] = {
+    &driReadDrawableExtension,
+    &driCopySubBufferExtension.base,
+    &driSwapControlExtension.base,
+    &driFrameTrackingExtension.base,
+    &driMediaStreamCounterExtension.base,
+    NULL
+};
+
+static __DRIconfig **radeon_fill_in_modes(unsigned pixel_bits,
+                                       unsigned depth_bits,
+                                       GLboolean have_back_buffer)
+{
+    __DRIconfig **configs;
+    unsigned depth_buffer_factor;
+    unsigned back_buffer_factor;
+    unsigned num_modes;
+    GLenum fb_format;
+    GLenum fb_type;
+    uint8_t depth_bits_array[3];
+    uint8_t stencil_bits_array[3];
+    uint8_t msaa_samples_array[1];
+    /* TODO: pageflipping ? */
+    static const GLenum back_buffer_modes[] = {
+        GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML
+    };
+
+    stencil_bits_array[0] = 0;
+    stencil_bits_array[1] = 0;
+    if (depth_bits == 24) {
+        stencil_bits_array[2] = 8;
+        num_modes = 3;
+    }
+
+    depth_bits_array[0] = 0;
+    depth_bits_array[1] = depth_bits;
+    depth_bits_array[2] = depth_bits;
+    depth_buffer_factor = (depth_bits == 24) ? 3 : 2;
+
+    back_buffer_factor = (have_back_buffer) ? 3 : 1;
+
+    msaa_samples_array[0] = 0;
+
+    if (pixel_bits == 16) {
+        fb_format = GL_RGB;
+        fb_type = GL_UNSIGNED_SHORT_5_6_5;
+    } else {
+        fb_format = GL_BGRA;
+        fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+    }
+
+    configs = (__DRIconfig **)driCreateConfigs(fb_format,
+                                               fb_type,
+                                               depth_bits_array,
+                                               stencil_bits_array,
+                                               depth_buffer_factor,
+                                               back_buffer_modes,
+                                               back_buffer_factor,
+                                               msaa_samples_array,
+                                               1);
+    if (configs == NULL) {
+        fprintf(stderr, "[%s:%u] Error creating FBConfig!\n",
+                __FILE__, __LINE__);
+        return NULL;
+    }
+    return configs;
+}
+
+static void radeon_screen_destroy(__DRIscreenPrivate *dri_screen)
+{
+     struct radeon_screen *radeon_screen = (struct radeon_screen*)dri_screen->private;
+
+     radeon_bo_manager_gem_dtor(radeon_screen->bom); 
+     dri_screen = NULL;
+     free(radeon_screen);
+}
+
+static const __DRIconfig **radeon_screen_init(__DRIscreenPrivate *dri_screen)
+{
+    struct radeon_screen *radeon_screen;
+
+    /* Calling driInitExtensions here, with a NULL context pointer,
+     * does not actually enable the extensions.  It just makes sure
+     * that all the dispatch offsets for all the extensions that
+     * *might* be enables are known.  This is needed because the
+     * dispatch offsets need to be known when _mesa_context_create is
+     * called, but we can't enable the extensions until we have a
+     * context pointer.
+     *
+     * Hello chicken.  Hello egg.  How are you two today?
+     */
+    driInitExtensions(NULL, radeon_card_extensions, GL_FALSE);
+
+    radeon_screen = calloc(1, sizeof(struct radeon_screen));
+    if (radeon_screen == NULL) {
+        fprintf(stderr, "\nERROR!  Allocating private area failed\n");
+        return NULL;
+    }
+    dri_screen->private = (void*)radeon_screen;
+    dri_screen->extensions = radeon_screen_extensions;
+    radeon_screen->dri_screen = dri_screen;
+
+    radeon_screen->bom = radeon_bo_manager_gem_ctor(dri_screen->fd);
+    if (radeon_screen->bom == NULL) {
+        radeon_screen_destroy(dri_screen);
+        return NULL;
+    }
+
+    return driConcatConfigs(radeon_fill_in_modes(16, 16, 1),
+                            radeon_fill_in_modes(32, 24, 1));
+}
+
+static boolean radeon_buffer_create(__DRIscreenPrivate *dri_screen,
+                                 __DRIdrawablePrivate *dri_drawable,
+                                 const __GLcontextModes *visual,
+                                 boolean is_pixmap)
+{
+    if (is_pixmap) {
+        /* TODO: implement ? */
+        return GL_FALSE;
+    } else {
+        enum pipe_format color_format, depth_format, stencil_format;
+        struct radeon_framebuffer *radeon_fb;
+
+        radeon_fb = calloc(1, sizeof(struct radeon_framebuffer));
+        if (radeon_fb == NULL) {
+            return GL_FALSE;
+        }
+
+        switch (visual->redBits) {
+        case 5:
+            color_format = PIPE_FORMAT_R5G6B5_UNORM;
+            break;
+        default:
+            color_format = PIPE_FORMAT_A8R8G8B8_UNORM;
+            break;
+        }
+
+        switch (visual->depthBits) {
+        case 24:
+            depth_format = PIPE_FORMAT_S8Z24_UNORM;
+            break;
+        case 16:
+            depth_format = PIPE_FORMAT_Z16_UNORM;
+            break;
+        default:
+            depth_format = PIPE_FORMAT_NONE;
+            break;
+        }
+
+        switch (visual->stencilBits) {
+        case 8:
+            /* force depth format */
+            depth_format = PIPE_FORMAT_S8Z24_UNORM;
+            stencil_format = PIPE_FORMAT_S8Z24_UNORM;
+            break;
+        default:
+            stencil_format = PIPE_FORMAT_NONE;
+            break;
+        }
+
+        radeon_fb->st_framebuffer = st_create_framebuffer(visual,
+                                                       color_format,
+                                                       depth_format,
+                                                       stencil_format,
+                                                       dri_drawable->w,
+                                                       dri_drawable->h,
+                                                       (void*)radeon_fb);
+        if (radeon_fb->st_framebuffer == NULL) {
+            free(radeon_fb);
+            return GL_FALSE;
+        }
+        dri_drawable->driverPrivate = (void *) radeon_fb;
+
+        radeon_fb->attachments = (1 << __DRI_BUFFER_FRONT_LEFT);
+        if (visual->doubleBufferMode) {
+            radeon_fb->attachments |= (1 << __DRI_BUFFER_BACK_LEFT);
+        }
+        if (visual->depthBits || visual->stencilBits) {
+            radeon_fb->attachments |= (1 << __DRI_BUFFER_DEPTH);
+        }
+
+        return GL_TRUE;
+    }
+}
+
+static void radeon_buffer_destroy(__DRIdrawablePrivate * dri_drawable)
+{
+   struct radeon_framebuffer *radeon_fb;
+   
+   radeon_fb = dri_drawable->driverPrivate;
+   assert(radeon_fb->st_framebuffer);
+   st_unreference_framebuffer(radeon_fb->st_framebuffer);
+   free(radeon_fb);
+}
+
+static void radeon_swap_buffers(__DRIdrawablePrivate *dri_drawable)
+{
+    struct radeon_framebuffer *radeon_fb;
+    struct pipe_surface *back_surf = NULL;
+
+    radeon_fb = dri_drawable->driverPrivate;
+    assert(radeon_fb);
+    assert(radeon_fb->st_framebuffer);
+
+    st_get_framebuffer_surface(radeon_fb->st_framebuffer,
+                               ST_SURFACE_BACK_LEFT,
+                               &back_surf);
+    if (back_surf) {
+        st_notify_swapbuffers(radeon_fb->st_framebuffer);
+        /* TODO: do we want to do anythings ? */
+        st_notify_swapbuffers_complete(radeon_fb->st_framebuffer);
+    }
+}
+
+/**
+ * Called via glXCopySubBufferMESA() to copy a subrect of the back
+ * buffer to the front buffer/screen.
+ */
+static void radeon_copy_sub_buffer(__DRIdrawablePrivate *dri_drawable,
+                         int x, int y, int w, int h)
+{
+    /* TODO: ... */
+}
+
+const struct __DriverAPIRec driDriverAPI = {
+    .InitScreen           = NULL,
+    .DestroyScreen        = radeon_screen_destroy,
+    .CreateContext        = radeon_context_create,
+    .DestroyContext       = radeon_context_destroy,
+    .CreateBuffer         = radeon_buffer_create,
+    .DestroyBuffer        = radeon_buffer_destroy,
+    .SwapBuffers          = radeon_swap_buffers,
+    .MakeCurrent          = radeon_context_bind,
+    .UnbindContext        = radeon_context_unbind,
+    .CopySubBuffer        = radeon_copy_sub_buffer,
+    .InitScreen2          = radeon_screen_init,
+};
diff --git a/src/gallium/winsys/drm/radeon/radeon_screen.h b/src/gallium/winsys/drm/radeon/radeon_screen.h
new file mode 100644
index 0000000000..01b7fa6531
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/radeon_screen.h
@@ -0,0 +1,41 @@
+/* 
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_SCREEN_H
+#define RADEON_SCREEN_H
+
+#include "dri_util.h"
+#include "radeon_bo.h"
+
+struct radeon_screen {
+    __DRIscreenPrivate          *dri_screen;
+    struct radeon_bo_manager    *bom;
+};
+
+#endif
diff --git a/src/gallium/winsys/drm/radeon/radeon_winsys_softpipe.c b/src/gallium/winsys/drm/radeon/radeon_winsys_softpipe.c
new file mode 100644
index 0000000000..8402e1fa5a
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/radeon_winsys_softpipe.c
@@ -0,0 +1,77 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+/*
+ * Authors: Keith Whitwell <keithw-at-tungstengraphics-dot-com>
+ */
+#include <stdio.h>
+#include "imports.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_format.h"
+#include "softpipe/sp_winsys.h"
+#include "radeon_context.h"
+#include "radeon_winsys_softpipe.h"
+
+struct radeon_softpipe_winsys {
+    struct softpipe_winsys  sp_winsys;
+    struct radeon_context      *radeon_context;
+};
+
+/**
+ * Return list of surface formats supported by this driver.
+ */
+static boolean radeon_is_format_supported(struct softpipe_winsys *sws, uint format)
+{
+    switch (format) {
+    case PIPE_FORMAT_A8R8G8B8_UNORM:
+    case PIPE_FORMAT_R5G6B5_UNORM:
+    case PIPE_FORMAT_Z24S8_UNORM:
+        return TRUE;
+    default:
+        break;
+    };
+    return FALSE;
+}
+
+struct pipe_context *radeon_create_softpipe(struct radeon_context *radeon_context)
+{
+    struct radeon_softpipe_winsys *radeon_sp_ws;
+    struct pipe_screen *pipe_screen;
+
+    pipe_screen = softpipe_create_screen(radeon_context->pipe_winsys);
+
+    radeon_sp_ws = CALLOC_STRUCT(radeon_softpipe_winsys);
+    if (radeon_sp_ws == NULL) {
+        return NULL;
+    }
+    radeon_context->pipe_screen = pipe_screen;
+    radeon_sp_ws->radeon_context = radeon_context;
+    radeon_sp_ws->sp_winsys.is_format_supported = radeon_is_format_supported;
+    return softpipe_create(pipe_screen,
+                           radeon_context->pipe_winsys,
+                           &radeon_sp_ws->sp_winsys);
+}
diff --git a/src/gallium/winsys/drm/radeon/radeon_winsys_softpipe.h b/src/gallium/winsys/drm/radeon/radeon_winsys_softpipe.h
new file mode 100644
index 0000000000..519eab769c
--- /dev/null
+++ b/src/gallium/winsys/drm/radeon/radeon_winsys_softpipe.h
@@ -0,0 +1,37 @@
+/* 
+ * Copyright © 2008 Jérôme Glisse
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
+ * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
+ * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ */
+/*
+ * Authors:
+ *      Jérôme Glisse <glisse@freedesktop.org>
+ */
+#ifndef RADEON_WINSYS_SOFTPIPE_H
+#define RADEON_WINSYS_SOFTPIPE_H
+
+#include "radeon_context.h"
+
+struct pipe_context *radeon_create_softpipe(struct radeon_context *radeon_context);
+
+#endif
diff --git a/src/gallium/winsys/egl_xlib/Makefile b/src/gallium/winsys/egl_xlib/Makefile
new file mode 100644
index 0000000000..02ac47caa4
--- /dev/null
+++ b/src/gallium/winsys/egl_xlib/Makefile
@@ -0,0 +1,89 @@
+# src/gallium/winsys/egl_xlib/Makefile
+
+# Build softpipe/xlib/EGL driver library/object: "egl_softpipe.so"
+
+
+TOP = ../../../..
+include $(TOP)/configs/current
+
+
+DRIVER_NAME = egl_softpipe.so
+
+
+INCLUDE_DIRS = \
+	-I$(TOP)/include \
+	-I$(TOP)/src/egl/main \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/mesa/main \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/gallium/auxiliary
+
+WINSYS_SOURCES = \
+	egl_xlib.c \
+	sw_winsys.c
+
+WINSYS_OBJECTS = $(WINSYS_SOURCES:.c=.o)
+
+
+LIBS = \
+	$(GALLIUM_DRIVERS) \
+	$(GALLIUM_AUXILIARIES)
+
+# XXX temporary (should create a separate lib with the GL API funcs and
+# mesa code, as done for ES 1.x, 2.x, OpenVG, etc)
+UNUSED_LIBS = \
+	$(TOP)/src/mesa/libglapi.a \
+	$(TOP)/src/mesa/libmesagallium.a \
+
+
+LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1
+
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(LOCAL_CFLAGS) $< -o $@
+
+
+.PHONY: library
+
+
+default: depend library Makefile
+
+
+library: $(TOP)/$(LIB_DIR)/$(DRIVER_NAME)
+
+
+# Make the egl_softpipe.so library
+$(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(WINSYS_OBJECTS) $(LIBS)
+	$(TOP)/bin/mklib -o $(DRIVER_NAME) \
+		-linker "$(CC)" \
+		-noprefix \
+		-install $(TOP)/$(LIB_DIR) \
+		$(MKLIB_OPTIONS) $(WINSYS_OBJECTS) \
+		--whole-archive $(LIBS) --no-whole-archive
+
+
+depend: $(ALL_SOURCES)
+	@ echo "running $(MKDEP)"
+	@ rm -f depend  # workaround oops on gutsy?!?
+	@ touch depend
+	@ $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \
+		> /dev/null 2>/dev/null
+
+
+install: default
+	$(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR)
+	@if [ -e $(TOP)/$(LIB_DIR) ]; then \
+		$(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(INSTALL_DIR)/$(LIB_DIR); \
+	fi
+
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` $(TOP)/include/GL/*.h
+
+clean:
+	-rm -f *.o *~ *.bak
+
+
+include depend
diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c
new file mode 100644
index 0000000000..c6b0e3d8c5
--- /dev/null
+++ b/src/gallium/winsys/egl_xlib/egl_xlib.c
@@ -0,0 +1,658 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * EGL / softpipe / xlib winsys module
+ *
+ * Authors: Brian Paul
+ */
+
+
+#include <dlfcn.h>
+#include <X11/Xutil.h>
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+#include "pipe/p_state.h"
+#include "pipe/internal/p_winsys_screen.h"
+#include "util/u_memory.h"
+#include "softpipe/sp_winsys.h"
+#include "softpipe/sp_texture.h"
+
+#include "eglconfig.h"
+#include "eglconfigutil.h"
+#include "eglcontext.h"
+#include "egldisplay.h"
+#include "egldriver.h"
+#include "eglglobals.h"
+#include "egllog.h"
+#include "eglsurface.h"
+
+#include "state_tracker/st_public.h"
+
+#include "sw_winsys.h"
+
+
+/** subclass of _EGLDriver */
+struct xlib_egl_driver
+{
+   _EGLDriver Base;   /**< base class */
+
+   struct pipe_winsys *winsys;
+   struct pipe_screen *screen;
+};
+
+
+/** subclass of _EGLContext */
+struct xlib_egl_context
+{
+   _EGLContext Base;   /**< base class */
+
+   struct pipe_context *pipe;   /**< Gallium driver context */
+   struct st_context *Context;  /**< Mesa/gallium state tracker context */
+};
+
+
+/** subclass of _EGLSurface */
+struct xlib_egl_surface
+{
+   _EGLSurface Base;   /**< base class */
+
+   Display *Dpy;  /**< The X Display of the window */
+   Window Win;    /**< The user-created window ID */
+   GC Gc;
+   XVisualInfo VisInfo;
+
+   struct pipe_winsys *winsys;
+
+   struct st_framebuffer *Framebuffer;
+};
+
+
+/** cast wrapper */
+static INLINE struct xlib_egl_driver *
+xlib_egl_driver(_EGLDriver *drv)
+{
+   return (struct xlib_egl_driver *) drv;
+}
+
+
+static struct xlib_egl_surface *
+lookup_surface(EGLSurface surf)
+{
+   _EGLSurface *surface = _eglLookupSurface(surf);
+   return (struct xlib_egl_surface *) surface;
+}
+
+
+static struct xlib_egl_context *
+lookup_context(EGLContext surf)
+{
+   _EGLContext *context = _eglLookupContext(surf);
+   return (struct xlib_egl_context *) context;
+}
+
+
+static unsigned int
+bitcount(unsigned int n)
+{
+   unsigned int bits;
+   for (bits = 0; n > 0; n = n >> 1) {
+      bits += (n & 1);
+   }
+   return bits;
+}
+
+
+/**
+ * Create the EGLConfigs.  (one per X visual)
+ */
+static void
+create_configs(_EGLDriver *drv, EGLDisplay dpy)
+{
+   static const EGLint all_apis = (EGL_OPENGL_ES_BIT |
+                                   EGL_OPENGL_ES2_BIT |
+                                   EGL_OPENVG_BIT |
+                                   EGL_OPENGL_BIT);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   XVisualInfo *visInfo, visTemplate;
+   int num_visuals, i;
+
+   /* get list of all X visuals, create an EGL config for each */
+   visTemplate.screen = DefaultScreen(disp->Xdpy);
+   visInfo = XGetVisualInfo(disp->Xdpy, VisualScreenMask,
+                            &visTemplate, &num_visuals);
+   if (!visInfo) {
+      printf("egl_xlib.c: couldn't get any X visuals\n");
+      abort();
+   }
+
+   for (i = 0; i < num_visuals; i++) {
+      _EGLConfig *config = calloc(1, sizeof(_EGLConfig));
+      int id = i + 1;
+      int rbits = bitcount(visInfo[i].red_mask);
+      int gbits = bitcount(visInfo[i].green_mask);
+      int bbits = bitcount(visInfo[i].blue_mask);
+      int abits = bbits == 8 ? 8 : 0;
+      int zbits = 24;
+      int sbits = 8;
+      int visid = visInfo[i].visualid;
+#if defined(__cplusplus) || defined(c_plusplus)
+      int vistype = visInfo[i].c_class;
+#else
+      int vistype = visInfo[i].class;
+#endif
+
+      _eglInitConfig(config, id);
+      SET_CONFIG_ATTRIB(config, EGL_BUFFER_SIZE, rbits + gbits + bbits + abits);
+      SET_CONFIG_ATTRIB(config, EGL_RED_SIZE, rbits);
+      SET_CONFIG_ATTRIB(config, EGL_GREEN_SIZE, gbits);
+      SET_CONFIG_ATTRIB(config, EGL_BLUE_SIZE, bbits);
+      SET_CONFIG_ATTRIB(config, EGL_ALPHA_SIZE, abits);
+      SET_CONFIG_ATTRIB(config, EGL_DEPTH_SIZE, zbits);
+      SET_CONFIG_ATTRIB(config, EGL_STENCIL_SIZE, sbits);
+      SET_CONFIG_ATTRIB(config, EGL_NATIVE_VISUAL_ID, visid);
+      SET_CONFIG_ATTRIB(config, EGL_NATIVE_VISUAL_TYPE, vistype);
+      SET_CONFIG_ATTRIB(config, EGL_NATIVE_RENDERABLE, EGL_FALSE);
+      SET_CONFIG_ATTRIB(config, EGL_CONFORMANT, all_apis);
+      SET_CONFIG_ATTRIB(config, EGL_RENDERABLE_TYPE, all_apis);
+      SET_CONFIG_ATTRIB(config, EGL_SURFACE_TYPE, EGL_WINDOW_BIT);
+
+      _eglAddConfig(disp, config);
+   }
+}
+
+
+/**
+ * Called via eglInitialize(), drv->API.Initialize().
+ */
+static EGLBoolean
+xlib_eglInitialize(_EGLDriver *drv, EGLDisplay dpy,
+                   EGLint *minor, EGLint *major)
+{
+   create_configs(drv, dpy);
+
+   drv->Initialized = EGL_TRUE;
+
+   /* we're supporting EGL 1.4 */
+   *minor = 1;
+   *major = 4;
+
+   return EGL_TRUE;
+}
+
+
+/**
+ * Called via eglTerminate(), drv->API.Terminate().
+ */
+static EGLBoolean
+xlib_eglTerminate(_EGLDriver *drv, EGLDisplay dpy)
+{
+   return EGL_TRUE;
+}
+
+
+static _EGLProc
+xlib_eglGetProcAddress(const char *procname)
+{
+   return (_EGLProc) st_get_proc_address(procname);
+}
+
+
+static void
+get_drawable_visual_info(Display *dpy, Drawable d, XVisualInfo *visInfo)
+{
+   XWindowAttributes attr;
+   XVisualInfo visTemp, *vis;
+   int num_visuals;
+
+   XGetWindowAttributes(dpy, d, &attr);
+
+   visTemp.screen = DefaultScreen(dpy);
+   visTemp.visualid = attr.visual->visualid;
+   vis = XGetVisualInfo(dpy,
+                        (VisualScreenMask | VisualIDMask),
+                        &visTemp, &num_visuals);
+   if (vis)
+      *visInfo = *vis;
+
+   XFree(vis);
+}
+
+
+
+/** Get size of given window */
+static Status
+get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height)
+{
+   Window root;
+   Status stat;
+   int xpos, ypos;
+   unsigned int w, h, bw, depth;
+   stat = XGetGeometry(dpy, d, &root, &xpos, &ypos, &w, &h, &bw, &depth);
+   *width = w;
+   *height = h;
+   return stat;
+}
+
+
+static void
+check_and_update_buffer_size(struct xlib_egl_surface *surface)
+{
+   uint width, height;
+   get_drawable_size(surface->Dpy, surface->Win, &width, &height);
+   st_resize_framebuffer(surface->Framebuffer, width, height);
+   surface->Base.Width = width;
+   surface->Base.Height = height;
+}
+
+
+
+static void
+display_surface(struct pipe_winsys *pws,
+                struct pipe_surface *psurf,
+                struct xlib_egl_surface *xsurf)
+{
+   XImage *ximage;
+   void *data;
+
+   ximage = XCreateImage(xsurf->Dpy,
+                         xsurf->VisInfo.visual,
+                         xsurf->VisInfo.depth,
+                         ZPixmap, 0,   /* format, offset */
+                         NULL,         /* data */
+                         0, 0,         /* size */
+                         32,           /* bitmap_pad */
+                         0);           /* bytes_per_line */
+
+
+   assert(ximage->format);
+   assert(ximage->bitmap_unit);
+
+   data = pws->buffer_map(pws, softpipe_texture(psurf->texture)->buffer, 0);
+
+   /* update XImage's fields */
+   ximage->data = data;
+   ximage->width = psurf->width;
+   ximage->height = psurf->height;
+   ximage->bytes_per_line = psurf->stride;
+   
+   XPutImage(xsurf->Dpy, xsurf->Win, xsurf->Gc,
+             ximage, 0, 0, 0, 0, psurf->width, psurf->height);
+
+   XSync(xsurf->Dpy, 0);
+
+   ximage->data = NULL;
+   XDestroyImage(ximage);
+
+   pws->buffer_unmap(pws, softpipe_texture(psurf->texture)->buffer);
+}
+
+
+
+/** Display gallium surface in X window */
+static void
+flush_frontbuffer(struct pipe_winsys *pws,
+                  struct pipe_surface *psurf,
+                  void *context_private)
+{
+   struct xlib_egl_surface *xsurf = (struct xlib_egl_surface *) context_private;
+   display_surface(pws, psurf, xsurf);
+}
+
+
+
+/**
+ * Called via eglCreateContext(), drv->API.CreateContext().
+ */
+static EGLContext
+xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+                      EGLContext share_list, const EGLint *attrib_list)
+{
+   struct xlib_egl_driver *xdrv = xlib_egl_driver(drv);
+   _EGLConfig *conf = _eglLookupConfig(drv, dpy, config);
+   struct xlib_egl_context *ctx;
+   struct st_context *share_ctx = NULL; /* XXX fix */
+   __GLcontextModes visual;
+
+   ctx = CALLOC_STRUCT(xlib_egl_context);
+   if (!ctx)
+      return EGL_NO_CONTEXT;
+
+   /* let EGL lib init the common stuff */
+   if (!_eglInitContext(drv, dpy, &ctx->Base, config, attrib_list)) {
+      free(ctx);
+      return EGL_NO_CONTEXT;
+   }
+
+   /* API-dependent context creation */
+   switch (ctx->Base.ClientAPI) {
+   case EGL_OPENVG_API:
+   case EGL_OPENGL_ES_API:
+      _eglLog(_EGL_DEBUG, "Create Context for ES version %d\n",
+              ctx->Base.ClientVersion);
+      /* fall-through */
+   case EGL_OPENGL_API:
+      /* create a softpipe context */
+      ctx->pipe = softpipe_create(xdrv->screen, xdrv->winsys, NULL);
+      /* Now do xlib / state tracker inits here */
+      _eglConfigToContextModesRec(conf, &visual);
+      ctx->Context = st_create_context(ctx->pipe, &visual, share_ctx);
+      break;
+   default:
+      _eglError(EGL_BAD_MATCH, "eglCreateContext(unsupported API)");
+      free(ctx);
+      return EGL_NO_CONTEXT;
+   }
+
+   _eglSaveContext(&ctx->Base);
+
+   return _eglGetContextHandle(&ctx->Base);
+}
+
+
+static EGLBoolean
+xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx)
+{
+   struct xlib_egl_context *context = lookup_context(ctx);
+   if (context) {
+      if (context->Base.IsBound) {
+         context->Base.DeletePending = EGL_TRUE;
+      }
+      else {
+         /* API-dependent clean-up */
+         switch (context->Base.ClientAPI) {
+         case EGL_OPENGL_ES_API:
+            /* fall-through */
+         case EGL_OPENGL_API:
+            st_destroy_context(context->Context);
+            break;
+         default:
+            assert(0);
+         }
+         free(context);
+      }
+      return EGL_TRUE;
+   }
+   else {
+      _eglError(EGL_BAD_CONTEXT, "eglDestroyContext");
+      return EGL_TRUE;
+   }
+}
+
+
+/**
+ * Called via eglMakeCurrent(), drv->API.MakeCurrent().
+ */
+static EGLBoolean
+xlib_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy,
+                    EGLSurface draw, EGLSurface read, EGLContext ctx)
+{
+   struct xlib_egl_context *context = lookup_context(ctx);
+   struct xlib_egl_surface *draw_surf = lookup_surface(draw);
+   struct xlib_egl_surface *read_surf = lookup_surface(read);
+
+   if (!_eglMakeCurrent(drv, dpy, draw, read, context))
+      return EGL_FALSE;
+
+   st_make_current((context ? context->Context : NULL),
+                   (draw_surf ? draw_surf->Framebuffer : NULL),
+                   (read_surf ? read_surf->Framebuffer : NULL));
+
+   if (draw_surf)
+      check_and_update_buffer_size(draw_surf);
+   if (read_surf && read_surf != draw_surf)
+      check_and_update_buffer_size(draw_surf);
+
+   return EGL_TRUE;
+}
+
+
+static enum pipe_format
+choose_color_format(const __GLcontextModes *visual)
+{
+   if (visual->redBits == 8 &&
+       visual->greenBits == 8 &&
+       visual->blueBits == 8 &&
+       visual->alphaBits == 8) {
+      /* XXX this really also depends on the ordering of R,G,B,A */
+      return PIPE_FORMAT_A8R8G8B8_UNORM;
+   }
+   else {
+      assert(0);
+      return PIPE_FORMAT_NONE;
+   }
+}
+
+
+static enum pipe_format
+choose_depth_format(const __GLcontextModes *visual)
+{
+   if (visual->depthBits > 0)
+      return PIPE_FORMAT_S8Z24_UNORM;
+   else
+      return PIPE_FORMAT_NONE;
+}
+
+
+static enum pipe_format
+choose_stencil_format(const __GLcontextModes *visual)
+{
+   if (visual->stencilBits > 0)
+      return PIPE_FORMAT_S8Z24_UNORM;
+   else
+      return PIPE_FORMAT_NONE;
+}
+
+
+/**
+ * Called via eglCreateWindowSurface(), drv->API.CreateWindowSurface().
+ */
+static EGLSurface
+xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config,
+                            NativeWindowType window, const EGLint *attrib_list)
+{
+   struct xlib_egl_driver *xdrv = xlib_egl_driver(drv);
+   _EGLDisplay *disp = _eglLookupDisplay(dpy);
+   _EGLConfig *conf = _eglLookupConfig(drv, dpy, config);
+
+   struct xlib_egl_surface *surf;
+   __GLcontextModes visual;
+   uint width, height;
+
+   surf = CALLOC_STRUCT(xlib_egl_surface);
+   if (!surf)
+      return EGL_NO_SURFACE;
+
+   /* Let EGL lib init the common stuff */
+   if (!_eglInitSurface(drv, dpy, &surf->Base, EGL_WINDOW_BIT,
+                        config, attrib_list)) {
+      free(surf);
+      return EGL_NO_SURFACE;
+   }
+
+   _eglSaveSurface(&surf->Base);
+
+   /*
+    * Now init the Xlib and gallium stuff
+    */
+   surf->Win = (Window) window;  /* The X window ID */
+   surf->Dpy = disp->Xdpy;  /* The X display */
+   surf->Gc = XCreateGC(surf->Dpy, surf->Win, 0, NULL);
+
+   surf->winsys = xdrv->winsys;
+
+   _eglConfigToContextModesRec(conf, &visual);
+   get_drawable_size(surf->Dpy, surf->Win, &width, &height);
+   get_drawable_visual_info(surf->Dpy, surf->Win, &surf->VisInfo);
+
+   surf->Base.Width = width;
+   surf->Base.Height = height;
+
+   /* Create GL statetracker framebuffer */
+   surf->Framebuffer = st_create_framebuffer(&visual,
+                                             choose_color_format(&visual),
+                                             choose_depth_format(&visual),
+                                             choose_stencil_format(&visual),
+                                             width, height,
+                                             (void *) surf);
+
+   st_resize_framebuffer(surf->Framebuffer, width, height);
+
+   return _eglGetSurfaceHandle(&surf->Base);
+}
+
+
+static EGLBoolean
+xlib_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface)
+{
+   struct xlib_egl_surface *surf = lookup_surface(surface);
+   if (surf) {
+      _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface);
+      if (surf->Base.IsBound) {
+         surf->Base.DeletePending = EGL_TRUE;
+      }
+      else {
+         XFreeGC(surf->Dpy, surf->Gc);
+         st_unreference_framebuffer(surf->Framebuffer);
+         free(surf);
+      }
+      return EGL_TRUE;
+   }
+   else {
+      _eglError(EGL_BAD_SURFACE, "eglDestroySurface");
+      return EGL_FALSE;
+   }
+}
+
+
+static EGLBoolean
+xlib_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw)
+{
+   /* error checking step: */
+   if (!_eglSwapBuffers(drv, dpy, draw))
+      return EGL_FALSE;
+
+   {
+      struct xlib_egl_surface *xsurf = lookup_surface(draw);
+      struct pipe_winsys *pws = xsurf->winsys;
+      struct pipe_surface *psurf;
+
+      st_get_framebuffer_surface(xsurf->Framebuffer, ST_SURFACE_BACK_LEFT,
+                                 &psurf);
+
+      st_notify_swapbuffers(xsurf->Framebuffer);
+
+      display_surface(pws, psurf, xsurf);
+
+      check_and_update_buffer_size(xsurf);
+   }
+
+   return EGL_TRUE;
+}
+
+
+/**
+ * Determine which API(s) is(are) present by looking for some specific
+ * global symbols.
+ */
+static EGLint
+find_supported_apis(void)
+{
+   EGLint mask = 0;
+   void *handle;
+
+   handle = dlopen(NULL, 0);
+
+   if (dlsym(handle, "st_api_OpenGL_ES1"))
+      mask |= EGL_OPENGL_ES_BIT;
+
+   if (dlsym(handle, "st_api_OpenGL_ES2"))
+      mask |= EGL_OPENGL_ES2_BIT;
+
+   if (dlsym(handle, "st_api_OpenGL"))
+      mask |= EGL_OPENGL_BIT;
+
+   if (dlsym(handle, "st_api_OpenVG"))
+      mask |= EGL_OPENVG_BIT;
+
+   dlclose(handle);
+
+   return mask;
+}
+
+
+/**
+ * This is the main entrypoint into the driver.
+ * Called by libEGL to instantiate an _EGLDriver object.
+ */
+_EGLDriver *
+_eglMain(_EGLDisplay *dpy, const char *args)
+{
+   struct xlib_egl_driver *xdrv;
+
+   _eglLog(_EGL_INFO, "Entering EGL/Xlib _eglMain(%s)", args);
+
+   xdrv = CALLOC_STRUCT(xlib_egl_driver);
+   if (!xdrv)
+      return NULL;
+
+   if (!dpy->Xdpy) {
+      dpy->Xdpy = XOpenDisplay(NULL);
+   }
+
+   _eglInitDriverFallbacks(&xdrv->Base);
+   xdrv->Base.API.Initialize = xlib_eglInitialize;
+   xdrv->Base.API.Terminate = xlib_eglTerminate;
+   xdrv->Base.API.GetProcAddress = xlib_eglGetProcAddress;
+   xdrv->Base.API.CreateContext = xlib_eglCreateContext;
+   xdrv->Base.API.DestroyContext = xlib_eglDestroyContext;
+   xdrv->Base.API.CreateWindowSurface = xlib_eglCreateWindowSurface;
+   xdrv->Base.API.DestroySurface = xlib_eglDestroySurface;
+   xdrv->Base.API.MakeCurrent = xlib_eglMakeCurrent;
+   xdrv->Base.API.SwapBuffers = xlib_eglSwapBuffers;
+
+   xdrv->Base.ClientAPIsMask = find_supported_apis();
+   if (xdrv->Base.ClientAPIsMask == 0x0) {
+      /* the app isn't directly linked with any EGL-supprted APIs
+       * (such as libGLESv2.so) so use an EGL utility to see what
+       * APIs might be loaded dynamically on this system.
+       */
+      xdrv->Base.ClientAPIsMask = _eglFindAPIs();
+   }      
+
+   xdrv->Base.Name = "Xlib/softpipe";
+
+   /* create one winsys and use it for all contexts/surfaces */
+   xdrv->winsys = create_sw_winsys();
+   xdrv->winsys->flush_frontbuffer = flush_frontbuffer;
+
+   xdrv->screen = softpipe_create_screen(xdrv->winsys);
+
+   return &xdrv->Base;
+}
+
diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c
new file mode 100644
index 0000000000..739bfa1c1a
--- /dev/null
+++ b/src/gallium/winsys/egl_xlib/sw_winsys.c
@@ -0,0 +1,242 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Totally software-based winsys layer.
+ * Note that the one winsys function that we can't implement here
+ * is flush_frontbuffer().
+ * Whoever uses this code will have to provide that.
+ *
+ * Authors: Brian Paul
+ */
+
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_state.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "sw_winsys.h"
+
+
+
+/** Subclass of pipe_winsys */
+struct sw_pipe_winsys
+{
+   struct pipe_winsys Base;
+   /* no extra fields for now */
+};
+
+
+/** subclass of pipe_buffer */
+struct sw_pipe_buffer
+{
+   struct pipe_buffer Base;
+   boolean UserBuffer;  /** Is this a user-space buffer? */
+   void *Data;
+   void *Mapped;
+};
+
+
+/** cast wrapper */
+static INLINE struct sw_pipe_buffer *
+sw_pipe_buffer(struct pipe_buffer *b)
+{
+   return (struct sw_pipe_buffer *) b;
+}
+
+
+/**
+ * Round n up to next multiple.
+ */
+static INLINE unsigned
+round_up(unsigned n, unsigned multiple)
+{
+   return (n + multiple - 1) & ~(multiple - 1);
+}
+
+
+static const char *
+get_name(struct pipe_winsys *pws)
+{
+   return "software";
+}
+
+
+/** Create new pipe_buffer and allocate storage of given size */
+static struct pipe_buffer *
+buffer_create(struct pipe_winsys *pws, 
+              unsigned alignment, 
+              unsigned usage,
+              unsigned size)
+{
+   struct sw_pipe_buffer *buffer = CALLOC_STRUCT(sw_pipe_buffer);
+   if (!buffer)
+      return NULL;
+
+   buffer->Base.refcount = 1;
+   buffer->Base.alignment = alignment;
+   buffer->Base.usage = usage;
+   buffer->Base.size = size;
+
+   /* align to 16-byte multiple for Cell */
+   buffer->Data = align_malloc(size, MAX2(alignment, 16));
+
+   return &buffer->Base;
+}
+
+
+/**
+ * Create buffer which wraps user-space data.
+ */
+static struct pipe_buffer *
+user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes)
+{
+   struct sw_pipe_buffer *buffer = CALLOC_STRUCT(sw_pipe_buffer);
+   if (!buffer)
+      return NULL;
+
+   buffer->Base.refcount = 1;
+   buffer->Base.size = bytes;
+   buffer->UserBuffer = TRUE;
+   buffer->Data = ptr;
+
+   return &buffer->Base;
+}
+
+
+static void *
+buffer_map(struct pipe_winsys *pws, struct pipe_buffer *buf, unsigned flags)
+{
+   struct sw_pipe_buffer *buffer = sw_pipe_buffer(buf);
+   buffer->Mapped = buffer->Data;
+   return buffer->Mapped;
+}
+
+
+static void
+buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf)
+{
+   struct sw_pipe_buffer *buffer = sw_pipe_buffer(buf);
+   buffer->Mapped = NULL;
+}
+
+
+static void
+buffer_destroy(struct pipe_winsys *pws, struct pipe_buffer *buf)
+{
+   struct sw_pipe_buffer *buffer = sw_pipe_buffer(buf);
+
+   if (buffer->Data && !buffer->UserBuffer) {
+      align_free(buffer->Data);
+      buffer->Data = NULL;
+   }
+
+   free(buffer);
+}
+
+
+static struct pipe_buffer *
+surface_buffer_create(struct pipe_winsys *winsys,
+                      unsigned width, unsigned height,
+                      enum pipe_format format, 
+                      unsigned usage,
+                      unsigned *stride)
+{
+   const unsigned alignment = 64;
+   struct pipe_format_block block;
+   unsigned nblocksx, nblocksy;
+
+   pf_get_block(format, &block);
+   nblocksx = pf_get_nblocksx(&block, width);
+   nblocksy = pf_get_nblocksy(&block, height);
+   *stride = round_up(nblocksx * block.size, alignment);
+
+   return winsys->buffer_create(winsys, alignment,
+                                usage,
+                                *stride * nblocksy);
+}
+
+
+static void
+fence_reference(struct pipe_winsys *sws, struct pipe_fence_handle **ptr,
+                struct pipe_fence_handle *fence)
+{
+   /* no-op */
+}
+
+
+static int
+fence_signalled(struct pipe_winsys *sws, struct pipe_fence_handle *fence,
+                unsigned flag)
+{
+   /* no-op */
+   return 0;
+}
+
+
+static int
+fence_finish(struct pipe_winsys *sws, struct pipe_fence_handle *fence,
+             unsigned flag)
+{
+   /* no-op */
+   return 0;
+}
+
+
+/**
+ * Create/return a new pipe_winsys object.
+ */
+struct pipe_winsys *
+create_sw_winsys(void)
+{
+   struct sw_pipe_winsys *ws = CALLOC_STRUCT(sw_pipe_winsys);
+   if (!ws)
+      return NULL;
+
+   /* Fill in this struct with callbacks that pipe will need to
+    * communicate with the window system, buffer manager, etc. 
+    */
+   ws->Base.buffer_create = buffer_create;
+   ws->Base.user_buffer_create = user_buffer_create;
+   ws->Base.buffer_map = buffer_map;
+   ws->Base.buffer_unmap = buffer_unmap;
+   ws->Base.buffer_destroy = buffer_destroy;
+
+   ws->Base.surface_buffer_create = surface_buffer_create;
+
+   ws->Base.fence_reference = fence_reference;
+   ws->Base.fence_signalled = fence_signalled;
+   ws->Base.fence_finish = fence_finish;
+
+   ws->Base.flush_frontbuffer = NULL; /* not implemented here! */
+
+   ws->Base.get_name = get_name;
+
+   return &ws->Base;
+}
diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.h b/src/gallium/winsys/egl_xlib/sw_winsys.h
new file mode 100644
index 0000000000..f96c5a14b0
--- /dev/null
+++ b/src/gallium/winsys/egl_xlib/sw_winsys.h
@@ -0,0 +1,40 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef SW_WINSYS_H
+#define SW_WINSYS_H
+
+
+struct pipe_winsys;
+
+
+extern struct pipe_winsys *
+create_sw_winsys(void);
+
+
+#endif /* SW_WINSYS_H */
diff --git a/src/gallium/winsys/g3dvl/nouveau/Makefile b/src/gallium/winsys/g3dvl/nouveau/Makefile
new file mode 100644
index 0000000000..2997f6b79c
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/Makefile
@@ -0,0 +1,50 @@
+TARGET		= libnouveau_dri.so
+GALLIUMDIR	= ../../..
+DRMDIR		?= /usr
+DRIDIR		= ../../../../driclient
+
+OBJECTS		= nouveau_screen_vl.o nouveau_context_vl.o nouveau_swapbuffers.o
+
+CFLAGS		+= -g -Wall -Werror=implicit-function-declaration -fPIC	\
+		   -I${GALLIUMDIR}/include				\
+		   -I${GALLIUMDIR}/winsys/g3dvl				\
+		   -I${GALLIUMDIR}/winsys/drm/nouveau			\
+		   -I${DRMDIR}/include					\
+		   -I${DRMDIR}/include/drm				\
+		   -I${DRMDIR}/include/nouveau				\
+		   -I${GALLIUMDIR}/drivers				\
+		   -I${GALLIUMDIR}/auxiliary				\
+		   -I${DRIDIR}/include
+
+LDFLAGS		+= -L${DRMDIR}/lib				\
+		   -L${DRIDIR}/lib				\
+		   -L${GALLIUMDIR}/winsys/drm/nouveau/common	\
+		   -L${GALLIUMDIR}/auxiliary/draw		\
+		   -L${GALLIUMDIR}/auxiliary/tgsi		\
+		   -L${GALLIUMDIR}/auxiliary/translate		\
+		   -L${GALLIUMDIR}/auxiliary/rtasm		\
+		   -L${GALLIUMDIR}/auxiliary/cso_cache		\
+		   -L${GALLIUMDIR}/drivers/nv04			\
+		   -L${GALLIUMDIR}/drivers/nv10			\
+		   -L${GALLIUMDIR}/drivers/nv20			\
+		   -L${GALLIUMDIR}/drivers/nv30			\
+		   -L${GALLIUMDIR}/drivers/nv40			\
+		   -L${GALLIUMDIR}/drivers/nv50
+
+LIBS		+= -lnouveaudrm -ldriclient -ldrm_nouveau -ldrm -lnv04 -lnv10 -lnv20 -lnv30 -lnv40 -lnv50 -ldraw -ltgsi -ltranslate -lrtasm -lcso_cache -lm
+
+#############################################
+
+.PHONY	= all clean libdriclient
+
+all: ${TARGET}
+
+${TARGET}: ${OBJECTS} libdriclient
+	$(CC) ${LDFLAGS} -shared -o $@ ${OBJECTS} ${LIBS}
+
+libdriclient:
+	cd ${DRIDIR}/src; ${MAKE}
+
+clean:
+	cd ${DRIDIR}/src; ${MAKE} clean
+	rm -rf ${OBJECTS} ${TARGET}
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_context_vl.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_context_vl.c
new file mode 100644
index 0000000000..dfc4905bc0
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_context_vl.c
@@ -0,0 +1,172 @@
+#include "nouveau_context_vl.h"
+#include <pipe/p_defines.h>
+#include <pipe/p_context.h>
+#include <pipe/p_screen.h>
+#include <util/u_memory.h>
+#include <common/nouveau_dri.h>
+#include <common/nouveau_local.h>
+#include <common/nouveau_winsys_pipe.h>
+#include "nouveau_screen_vl.h"
+
+/*
+#ifdef DEBUG
+static const struct dri_debug_control debug_control[] = {
+	{ "bo", DEBUG_BO },
+	{ NULL, 0 }
+};
+int __nouveau_debug = 0;
+#endif
+*/
+
+int
+nouveau_context_create(dri_context_t *dri_context)
+{
+	dri_screen_t			*dri_screen;
+	struct nouveau_screen_vl	*nv_screen;
+	struct nouveau_context_vl	*nv;
+
+	assert (dri_context);
+
+	dri_screen = dri_context->dri_screen;
+	nv_screen = dri_screen->private;
+	nv = CALLOC_STRUCT(nouveau_context_vl);
+
+	if (!nv)
+		return 1;
+
+	if (nouveau_context_init(&nv_screen->base, dri_context->drm_context,
+	                        (drmLock*)&dri_screen->sarea->lock, NULL, &nv->base))
+	{
+		FREE(nv);
+		return 1;
+	}
+
+	dri_context->private = (void*)nv;
+	nv->dri_context = dri_context;
+	nv->nv_screen  = nv_screen;
+
+	/*
+	driParseConfigFiles(&nv->dri_option_cache, &nv_screen->option_cache,
+			    nv->dri_screen->myNum, "nouveau");
+#ifdef DEBUG
+	__nouveau_debug = driParseDebugString(getenv("NOUVEAU_DEBUG"),
+					      debug_control);
+#endif
+	*/
+
+	nv->base.nvc->pctx[nv->base.pctx_id]->priv = nv;
+
+	return 0;
+}
+
+void
+nouveau_context_destroy(dri_context_t *dri_context)
+{
+	struct nouveau_context_vl *nv = dri_context->private;
+
+	assert(dri_context);
+
+	nouveau_context_cleanup(&nv->base);
+
+	FREE(nv);
+}
+
+int
+nouveau_context_bind(struct nouveau_context_vl *nv, dri_drawable_t *dri_drawable)
+{
+	assert(nv);
+	assert(dri_drawable);
+
+	if (nv->dri_drawable != dri_drawable)
+	{
+		nv->dri_drawable = dri_drawable;
+		dri_drawable->private = nv;
+	}
+
+	return 0;
+}
+
+int
+nouveau_context_unbind(struct nouveau_context_vl *nv)
+{
+	assert(nv);
+
+	nv->dri_drawable = NULL;
+
+	return 0;
+}
+
+/* Show starts here */
+
+int bind_pipe_drawable(struct pipe_context *pipe, Drawable drawable)
+{
+	struct nouveau_context_vl	*nv;
+	dri_drawable_t			*dri_drawable;
+
+	assert(pipe);
+
+	nv = pipe->priv;
+
+	driCreateDrawable(nv->nv_screen->dri_screen, drawable, &dri_drawable);
+
+	nouveau_context_bind(nv, dri_drawable);
+
+	return 0;
+}
+
+int unbind_pipe_drawable(struct pipe_context *pipe)
+{
+	assert (pipe);
+
+	nouveau_context_unbind(pipe->priv);
+
+	return 0;
+}
+
+struct pipe_context* create_pipe_context(Display *display, int screen)
+{
+	dri_screen_t			*dri_screen;
+	dri_framebuffer_t		dri_framebuf;
+	dri_context_t			*dri_context;
+	struct nouveau_context_vl	*nv;
+
+	assert(display);
+
+	driCreateScreen(display, screen, &dri_screen, &dri_framebuf);
+	driCreateContext(dri_screen, XDefaultVisual(display, screen), &dri_context);
+
+	nouveau_screen_create(dri_screen, &dri_framebuf);
+	nouveau_context_create(dri_context);
+
+	nv = dri_context->private;
+
+	return nv->base.nvc->pctx[nv->base.pctx_id];
+}
+
+int destroy_pipe_context(struct pipe_context *pipe)
+{
+	struct pipe_screen		*screen;
+	struct pipe_winsys		*winsys;
+	struct nouveau_context_vl	*nv;
+	dri_screen_t			*dri_screen;
+	dri_context_t			*dri_context;
+
+	assert(pipe);
+
+	screen = pipe->screen;
+	winsys = pipe->winsys;
+	nv = pipe->priv;
+	dri_context = nv->dri_context;
+	dri_screen = dri_context->dri_screen;
+
+	pipe->destroy(pipe);
+	screen->destroy(screen);
+	FREE(winsys);
+
+	nouveau_context_destroy(dri_context);
+	nouveau_screen_destroy(dri_screen);
+	driDestroyContext(dri_context);
+	driDestroyScreen(dri_screen);
+
+	return 0;
+}
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_context_vl.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_context_vl.h
new file mode 100644
index 0000000000..1115c3130c
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_context_vl.h
@@ -0,0 +1,39 @@
+#ifndef __NOUVEAU_CONTEXT_VL_H__
+#define __NOUVEAU_CONTEXT_VL_H__
+
+#include <driclient.h>
+#include <nouveau/nouveau_winsys.h>
+#include <common/nouveau_context.h>
+
+/*#include "xmlconfig.h"*/
+
+struct nouveau_context_vl {
+	struct nouveau_context		base;
+	struct nouveau_screen_vl	*nv_screen;
+	dri_context_t			*dri_context;
+	dri_drawable_t			*dri_drawable;
+	unsigned int			last_stamp;
+	/*driOptionCache		dri_option_cache;*/
+	drm_context_t			drm_context;
+	drmLock				drm_lock;
+};
+
+extern int nouveau_context_create(dri_context_t *);
+extern void nouveau_context_destroy(dri_context_t *);
+extern int nouveau_context_bind(struct nouveau_context_vl *, dri_drawable_t *);
+extern int nouveau_context_unbind(struct nouveau_context_vl *);
+
+#ifdef DEBUG
+extern int __nouveau_debug;
+
+#define DEBUG_BO (1 << 0)
+
+#define DBG(flag, ...) do {                   \
+	if (__nouveau_debug & (DEBUG_##flag)) \
+		NOUVEAU_ERR(__VA_ARGS__);     \
+} while(0)
+#else
+#define DBG(flag, ...)
+#endif
+
+#endif
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_screen_vl.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_screen_vl.c
new file mode 100644
index 0000000000..b7c74f8299
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_screen_vl.c
@@ -0,0 +1,88 @@
+#include "nouveau_screen_vl.h"
+#include <util/u_memory.h>
+#include <nouveau_drm.h>
+#include <common/nouveau_dri.h>
+#include <common/nouveau_local.h>
+
+#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 12
+#error nouveau_drm.h version does not match expected version
+#endif
+
+/*
+PUBLIC const char __driConfigOptions[] =
+DRI_CONF_BEGIN
+DRI_CONF_END;
+static const GLuint __driNConfigOptions = 0;
+*/
+
+int nouveau_check_dri_drm_ddx(dri_version_t *dri, dri_version_t *drm, dri_version_t *ddx)
+{
+	static const dri_version_t ddx_expected = {0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL};
+	static const dri_version_t dri_expected = {4, 0, 0};
+	static const dri_version_t drm_expected = {0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL};
+
+	assert(dri);
+	assert(drm);
+	assert(ddx);
+
+	if (dri->major != dri_expected.major || dri->minor < dri_expected.minor)
+	{
+		NOUVEAU_ERR("Unexpected DRI version.\n");
+		return 1;
+	}
+	if (drm->major != drm_expected.major || drm->minor < drm_expected.minor)
+	{
+		NOUVEAU_ERR("Unexpected DRM version.\n");
+		return 1;
+	}
+	if (ddx->major != ddx_expected.major || ddx->minor < ddx_expected.minor)
+	{
+		NOUVEAU_ERR("Unexpected DDX version.\n");
+		return 1;
+	}
+
+	return 0;
+}
+
+int
+nouveau_screen_create(dri_screen_t *dri_screen, dri_framebuffer_t *dri_framebuf)
+{
+	struct nouveau_dri		*nv_dri = dri_framebuf->private;
+	struct nouveau_screen_vl	*nv_screen;
+
+	assert(dri_screen);
+	assert(dri_framebuf);
+
+	if (nouveau_check_dri_drm_ddx(&dri_screen->dri, &dri_screen->drm, &dri_screen->ddx))
+		return 1;
+
+	nv_screen = CALLOC_STRUCT(nouveau_screen_vl);
+
+	if (!nv_screen)
+		return 1;
+
+	if (nouveau_screen_init(nv_dri, dri_screen->fd, &nv_screen->base))
+	{
+		FREE(nv_screen);
+		return 1;
+	}
+
+	/*
+	driParseOptionInfo(&nv_screen->option_cache,
+			   __driConfigOptions, __driNConfigOptions);
+	*/
+
+	nv_screen->dri_screen = dri_screen;
+	dri_screen->private = (void*)nv_screen;
+
+	return 0;
+}
+
+void
+nouveau_screen_destroy(dri_screen_t *dri_screen)
+{
+	struct nouveau_screen_vl *nv_screen = dri_screen->private;
+
+	nouveau_screen_cleanup(&nv_screen->base);
+	FREE(nv_screen);
+}
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_screen_vl.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_screen_vl.h
new file mode 100644
index 0000000000..0c1ceca6de
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_screen_vl.h
@@ -0,0 +1,20 @@
+#ifndef __NOUVEAU_SCREEN_VL_H__
+#define __NOUVEAU_SCREEN_VL_H__
+
+#include <driclient.h>
+#include <common/nouveau_screen.h>
+
+/* TODO: Investigate using DRI options for interesting things */
+/*#include "xmlconfig.h"*/
+
+struct nouveau_screen_vl
+{
+	struct nouveau_screen		base;
+	dri_screen_t			*dri_screen;
+	/*driOptionCache		option_cache;*/
+};
+
+int nouveau_screen_create(dri_screen_t *dri_screen, dri_framebuffer_t *dri_framebuf);
+void nouveau_screen_destroy(dri_screen_t *dri_screen);
+
+#endif
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c
new file mode 100644
index 0000000000..864be37871
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c
@@ -0,0 +1,94 @@
+#include <driclient.h>
+#include <common/nouveau_local.h>
+#include <common/nouveau_screen.h>
+#include "nouveau_context_vl.h"
+#include "nouveau_swapbuffers.h"
+
+void
+nouveau_copy_buffer(dri_drawable_t *dri_drawable, struct pipe_surface *surf,
+		    const drm_clip_rect_t *rect)
+{
+	struct nouveau_context_vl	*nv = dri_drawable->private;
+	struct pipe_context		*pipe = nv->base.nvc->pctx[nv->base.pctx_id];
+	drm_clip_rect_t			*pbox;
+	int				nbox, i;
+
+	LOCK_HARDWARE(&nv->base);
+	if (!dri_drawable->num_cliprects) {
+		UNLOCK_HARDWARE(&nv->base);
+		return;
+	}
+	pbox = dri_drawable->cliprects;
+	nbox = dri_drawable->num_cliprects;
+
+	for (i = 0; i < nbox; i++, pbox++) {
+		int sx, sy, dx, dy, w, h;
+
+		sx = pbox->x1 - dri_drawable->x;
+		sy = pbox->y1 - dri_drawable->y;
+		dx = pbox->x1;
+		dy = pbox->y1;
+		w  = pbox->x2 - pbox->x1;
+		h  = pbox->y2 - pbox->y1;
+
+		pipe->surface_copy(pipe, FALSE, nv->base.frontbuffer,
+				   dx, dy, surf, sx, sy, w, h);
+	}
+
+	FIRE_RING(nv->base.nvc->channel);
+	UNLOCK_HARDWARE(&nv->base);
+}
+
+void
+nouveau_copy_sub_buffer(dri_drawable_t *dri_drawable, struct pipe_surface *surf, int x, int y, int w, int h)
+{
+	if (surf) {
+		drm_clip_rect_t rect;
+		rect.x1 = x;
+		rect.y1 = y;
+		rect.x2 = x + w;
+		rect.y2 = y + h;
+
+		nouveau_copy_buffer(dri_drawable, surf, &rect);
+	}
+}
+
+void
+nouveau_swap_buffers(dri_drawable_t *dri_drawable, struct pipe_surface *surf)
+{
+	if (surf)
+		nouveau_copy_buffer(dri_drawable, surf, NULL);
+}
+
+void
+nouveau_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *surf,
+			  void *context_private)
+{
+	struct nouveau_context_vl	*nv;
+	dri_drawable_t			*dri_drawable;
+
+	assert(pws);
+	assert(surf);
+	assert(context_private);
+
+	nv = context_private;
+	dri_drawable = nv->dri_drawable;
+
+	nouveau_copy_buffer(dri_drawable, surf, NULL);
+}
+
+void
+nouveau_contended_lock(struct nouveau_context *nv)
+{
+	struct nouveau_context_vl	*nv_vl = (struct nouveau_context_vl*)nv;
+	dri_drawable_t			*dri_drawable = nv_vl->dri_drawable;
+	dri_screen_t			*dri_screen = nv_vl->dri_context->dri_screen;
+
+	/* If the window moved, may need to set a new cliprect now.
+	 *
+	 * NOTE: This releases and regains the hw lock, so all state
+	 * checking must be done *after* this call:
+	 */
+	if (dri_drawable)
+		DRI_VALIDATE_DRAWABLE_INFO(dri_screen, dri_drawable);
+}
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.h
new file mode 100644
index 0000000000..35e934adba
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.h
@@ -0,0 +1,10 @@
+#ifndef __NOUVEAU_SWAPBUFFERS_H__
+#define __NOUVEAU_SWAPBUFFERS_H__
+
+extern void nouveau_copy_buffer(dri_drawable_t *, struct pipe_surface *,
+				const drm_clip_rect_t *);
+extern void nouveau_copy_sub_buffer(dri_drawable_t *, struct pipe_surface *,
+				    int x, int y, int w, int h);
+extern void nouveau_swap_buffers(dri_drawable_t *, struct pipe_surface *);
+
+#endif
diff --git a/src/gallium/winsys/g3dvl/vl_winsys.h b/src/gallium/winsys/g3dvl/vl_winsys.h
new file mode 100644
index 0000000000..c83db28dd9
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/vl_winsys.h
@@ -0,0 +1,14 @@
+#ifndef vl_winsys_h
+#define vl_winsys_h
+
+#include <X11/Xlib.h>
+
+struct pipe_context;
+
+struct pipe_context* create_pipe_context(Display *display, int screen);
+int destroy_pipe_context(struct pipe_context *pipe);
+int bind_pipe_drawable(struct pipe_context *pipe, Drawable drawable);
+int unbind_pipe_drawable(struct pipe_context *pipe);
+
+#endif
+
diff --git a/src/gallium/winsys/g3dvl/xsp_winsys.c b/src/gallium/winsys/g3dvl/xsp_winsys.c
new file mode 100644
index 0000000000..40d683234f
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/xsp_winsys.c
@@ -0,0 +1,295 @@
+#include "vl_winsys.h"
+#include <X11/Xutil.h>
+#include <pipe/p_winsys.h>
+#include <pipe/p_state.h>
+#include <pipe/p_inlines.h>
+#include <util/u_memory.h>
+#include <softpipe/sp_winsys.h>
+
+/* pipe_winsys implementation */
+
+struct xsp_pipe_winsys
+{
+	struct pipe_winsys	base;
+	XImage			fbimage;
+};
+
+struct xsp_context
+{
+	Display			*display;
+	int			screen;
+	Drawable		drawable;
+	int			drawable_bound;
+};
+
+struct xsp_buffer
+{
+	struct pipe_buffer	base;
+	boolean			is_user_buffer;
+	void			*data;
+	void			*mapped_data;
+};
+
+static struct pipe_buffer* xsp_buffer_create(struct pipe_winsys *pws, unsigned alignment, unsigned usage, unsigned size)
+{
+	struct xsp_buffer *buffer;
+
+	assert(pws);
+
+	buffer = calloc(1, sizeof(struct xsp_buffer));
+	buffer->base.refcount = 1;
+	buffer->base.alignment = alignment;
+	buffer->base.usage = usage;
+	buffer->base.size = size;
+	buffer->data = align_malloc(size, alignment);
+
+	return (struct pipe_buffer*)buffer;
+}
+
+static struct pipe_buffer* xsp_user_buffer_create(struct pipe_winsys *pws, void *data, unsigned size)
+{
+	struct xsp_buffer *buffer;
+
+	assert(pws);
+
+	buffer = calloc(1, sizeof(struct xsp_buffer));
+	buffer->base.refcount = 1;
+	buffer->base.size = size;
+	buffer->is_user_buffer = TRUE;
+	buffer->data = data;
+
+	return (struct pipe_buffer*)buffer;
+}
+
+static void* xsp_buffer_map(struct pipe_winsys *pws, struct pipe_buffer *buffer, unsigned flags)
+{
+	struct xsp_buffer *xsp_buf = (struct xsp_buffer*)buffer;
+
+	assert(pws);
+	assert(buffer);
+
+	xsp_buf->mapped_data = xsp_buf->data;
+
+	return xsp_buf->mapped_data;
+}
+
+static void xsp_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buffer)
+{
+	struct xsp_buffer *xsp_buf = (struct xsp_buffer*)buffer;
+
+	assert(pws);
+	assert(buffer);
+
+	xsp_buf->mapped_data = NULL;
+}
+
+static void xsp_buffer_destroy(struct pipe_winsys *pws, struct pipe_buffer *buffer)
+{
+	struct xsp_buffer *xsp_buf = (struct xsp_buffer*)buffer;
+
+	assert(pws);
+	assert(buffer);
+
+	if (!xsp_buf->is_user_buffer)
+		align_free(xsp_buf->data);
+
+	free(xsp_buf);
+}
+
+/* Borrowed from Mesa's xm_winsys */
+static unsigned int round_up(unsigned n, unsigned multiple)
+{
+   return (n + multiple - 1) & ~(multiple - 1);
+}
+
+static struct pipe_buffer* xsp_surface_buffer_create
+(
+	struct pipe_winsys *pws,
+	unsigned width,
+	unsigned height,
+	enum pipe_format format,
+	unsigned usage,
+	unsigned *stride
+)
+{
+	const unsigned int ALIGNMENT = 1;
+	struct pipe_format_block block;
+	unsigned nblocksx, nblocksy;
+
+	pf_get_block(format, &block);
+	nblocksx = pf_get_nblocksx(&block, width);
+	nblocksy = pf_get_nblocksy(&block, height);
+	*stride = round_up(nblocksx * block.size, ALIGNMENT);
+
+	return winsys->buffer_create(winsys, ALIGNMENT,
+				     usage,
+				     *stride * nblocksy);
+}
+
+static void xsp_fence_reference(struct pipe_winsys *pws, struct pipe_fence_handle **ptr, struct pipe_fence_handle *fence)
+{
+	assert(pws);
+	assert(ptr);
+	assert(fence);
+}
+
+static int xsp_fence_signalled(struct pipe_winsys *pws, struct pipe_fence_handle *fence, unsigned flag)
+{
+	assert(pws);
+	assert(fence);
+
+	return 0;
+}
+
+static int xsp_fence_finish(struct pipe_winsys *pws, struct pipe_fence_handle *fence, unsigned flag)
+{
+	assert(pws);
+	assert(fence);
+
+	return 0;
+}
+
+static void xsp_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *surface, void *context_private)
+{
+	struct xsp_pipe_winsys	*xsp_winsys;
+	struct xsp_context	*xsp_context;
+
+	assert(pws);
+	assert(surface);
+	assert(context_private);
+
+	xsp_winsys = (struct xsp_pipe_winsys*)pws;
+	xsp_context = (struct xsp_context*)context_private;
+
+	if (!xsp_context->drawable_bound)
+		return;
+
+	xsp_winsys->fbimage.width = surface->width;
+	xsp_winsys->fbimage.height = surface->height;
+	xsp_winsys->fbimage.bytes_per_line = surface->width * (xsp_winsys->fbimage.bits_per_pixel >> 3);
+	xsp_winsys->fbimage.data = pipe_surface_map(surface, 0);
+
+	XPutImage
+	(
+		xsp_context->display,
+		xsp_context->drawable,
+		XDefaultGC(xsp_context->display, xsp_context->screen),
+		&xsp_winsys->fbimage,
+		0,
+		0,
+		0,
+		0,
+		surface->width,
+		surface->height
+	);
+	XFlush(xsp_context->display);
+	pipe_surface_unmap(surface);
+}
+
+static const char* xsp_get_name(struct pipe_winsys *pws)
+{
+	assert(pws);
+	return "X11 SoftPipe";
+}
+
+/* Show starts here */
+
+int bind_pipe_drawable(struct pipe_context *pipe, Drawable drawable)
+{
+	struct xsp_context *xsp_context;
+
+	assert(pipe);
+
+	xsp_context = pipe->priv;
+	xsp_context->drawable = drawable;
+	xsp_context->drawable_bound = 1;
+
+	return 0;
+}
+
+int unbind_pipe_drawable(struct pipe_context *pipe)
+{
+	struct xsp_context *xsp_context;
+
+	assert(pipe);
+
+	xsp_context = pipe->priv;
+	xsp_context->drawable_bound = 0;
+
+	return 0;
+}
+
+struct pipe_context* create_pipe_context(Display *display, int screen)
+{
+	struct xsp_pipe_winsys	*xsp_winsys;
+	struct xsp_context	*xsp_context;
+	struct pipe_screen	*sp_screen;
+	struct pipe_context	*sp_pipe;
+
+	assert(display);
+
+	xsp_winsys = calloc(1, sizeof(struct xsp_pipe_winsys));
+	xsp_winsys->base.buffer_create = xsp_buffer_create;
+	xsp_winsys->base.user_buffer_create = xsp_user_buffer_create;
+	xsp_winsys->base.buffer_map = xsp_buffer_map;
+	xsp_winsys->base.buffer_unmap = xsp_buffer_unmap;
+	xsp_winsys->base.buffer_destroy = xsp_buffer_destroy;
+	xsp_winsys->base.surface_buffer_create = xsp_surface_buffer_create;
+	xsp_winsys->base.fence_reference = xsp_fence_reference;
+	xsp_winsys->base.fence_signalled = xsp_fence_signalled;
+	xsp_winsys->base.fence_finish = xsp_fence_finish;
+	xsp_winsys->base.flush_frontbuffer = xsp_flush_frontbuffer;
+	xsp_winsys->base.get_name = xsp_get_name;
+
+	{
+		/* XXX: Can't use the returned XImage* directly,
+		since we don't have control over winsys destruction
+		and we wouldn't be able to free it */
+		XImage *template = XCreateImage
+		(
+			display,
+			XDefaultVisual(display, XDefaultScreen(display)),
+			XDefaultDepth(display, XDefaultScreen(display)),
+			ZPixmap,
+			0,
+			NULL,
+			0,	/* Don't know the width and height until flush_frontbuffer */
+			0,
+			32,
+			0
+		);
+
+		memcpy(&xsp_winsys->fbimage, template, sizeof(XImage));
+		XInitImage(&xsp_winsys->fbimage);
+
+		XDestroyImage(template);
+	}
+
+	sp_screen = softpipe_create_screen((struct pipe_winsys*)xsp_winsys);
+	sp_pipe = softpipe_create(sp_screen, (struct pipe_winsys*)xsp_winsys, NULL);
+
+	xsp_context = calloc(1, sizeof(struct xsp_context));
+	xsp_context->display = display;
+	xsp_context->screen = screen;
+
+	sp_pipe->priv = xsp_context;
+
+	return sp_pipe;
+}
+
+int destroy_pipe_context(struct pipe_context *pipe)
+{
+	struct pipe_screen *screen;
+	struct pipe_winsys *winsys;
+
+	assert(pipe);
+
+	screen = pipe->screen;
+	winsys = pipe->winsys;
+	free(pipe->priv);
+	pipe->destroy(pipe);
+	screen->destroy(screen);
+	free(winsys);
+
+	return 0;
+}
diff --git a/src/gallium/winsys/gdi/SConscript b/src/gallium/winsys/gdi/SConscript
new file mode 100644
index 0000000000..f9c1a34668
--- /dev/null
+++ b/src/gallium/winsys/gdi/SConscript
@@ -0,0 +1,39 @@
+#######################################################################
+# SConscript for gdi winsys
+
+Import('*')
+
+if env['platform'] == 'windows':
+
+	env = env.Clone()
+
+	env.Append(CPPPATH = [
+		'#src/gallium/state_trackers/wgl',
+	])
+
+	env.Append(CPPDEFINES = [
+		'__GL_EXPORTS',
+		'BUILD_GL32',
+		'_GNU_H_WINDOWS32_DEFINES',
+	])
+
+	env.Append(LIBS = [
+		'gdi32',
+		'user32',
+		'kernel32',
+	])
+
+	sources = [
+		'#src/gallium/state_trackers/wgl/opengl32.def',
+		'gdi_softpipe_winsys.c',
+	]
+		
+	drivers = [
+		softpipe,
+	]
+
+	env.SharedLibrary(
+		target ='opengl32',
+		source = sources,
+		LIBS = wgl + glapi + mesa + drivers + auxiliaries + env['LIBS'],
+	)
diff --git a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c
new file mode 100644
index 0000000000..4924dbf26a
--- /dev/null
+++ b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c
@@ -0,0 +1,320 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ *
+ **************************************************************************/
+
+/**
+ * @file
+ * Softpipe support.
+ *
+ * @author Keith Whitwell
+ * @author Brian Paul
+ * @author Jose Fonseca
+ */
+
+
+#include <windows.h>
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "softpipe/sp_winsys.h"
+#include "softpipe/sp_texture.h"
+#include "shared/stw_winsys.h"
+
+
+struct gdi_softpipe_buffer
+{
+   struct pipe_buffer base;
+   boolean userBuffer;  /** Is this a user-space buffer? */
+   void *data;
+   void *mapped;
+};
+
+
+/** Cast wrapper */
+static INLINE struct gdi_softpipe_buffer *
+gdi_softpipe_buffer( struct pipe_buffer *buf )
+{
+   return (struct gdi_softpipe_buffer *)buf;
+}
+
+
+static void *
+gdi_softpipe_buffer_map(struct pipe_winsys *winsys,
+                        struct pipe_buffer *buf,
+                        unsigned flags)
+{
+   struct gdi_softpipe_buffer *gdi_softpipe_buf = gdi_softpipe_buffer(buf);
+   gdi_softpipe_buf->mapped = gdi_softpipe_buf->data;
+   return gdi_softpipe_buf->mapped;
+}
+
+
+static void
+gdi_softpipe_buffer_unmap(struct pipe_winsys *winsys,
+                          struct pipe_buffer *buf)
+{
+   struct gdi_softpipe_buffer *gdi_softpipe_buf = gdi_softpipe_buffer(buf);
+   gdi_softpipe_buf->mapped = NULL;
+}
+
+
+static void
+gdi_softpipe_buffer_destroy(struct pipe_winsys *winsys,
+                            struct pipe_buffer *buf)
+{
+   struct gdi_softpipe_buffer *oldBuf = gdi_softpipe_buffer(buf);
+
+   if (oldBuf->data) {
+      if (!oldBuf->userBuffer)
+         align_free(oldBuf->data);
+
+      oldBuf->data = NULL;
+   }
+
+   FREE(oldBuf);
+}
+
+
+static const char *
+gdi_softpipe_get_name(struct pipe_winsys *winsys)
+{
+   return "softpipe";
+}
+
+
+static struct pipe_buffer *
+gdi_softpipe_buffer_create(struct pipe_winsys *winsys,
+                           unsigned alignment,
+                           unsigned usage,
+                           unsigned size)
+{
+   struct gdi_softpipe_buffer *buffer = CALLOC_STRUCT(gdi_softpipe_buffer);
+
+   buffer->base.refcount = 1;
+   buffer->base.alignment = alignment;
+   buffer->base.usage = usage;
+   buffer->base.size = size;
+
+   buffer->data = align_malloc(size, alignment);
+
+   return &buffer->base;
+}
+
+
+/**
+ * Create buffer which wraps user-space data.
+ */
+static struct pipe_buffer *
+gdi_softpipe_user_buffer_create(struct pipe_winsys *winsys,
+                               void *ptr,
+                               unsigned bytes)
+{
+   struct gdi_softpipe_buffer *buffer;
+
+   buffer = CALLOC_STRUCT(gdi_softpipe_buffer);
+   if(!buffer)
+      return NULL;
+
+   buffer->base.refcount = 1;
+   buffer->base.size = bytes;
+   buffer->userBuffer = TRUE;
+   buffer->data = ptr;
+
+   return &buffer->base;
+}
+
+
+/**
+ * Round n up to next multiple.
+ */
+static INLINE unsigned
+round_up(unsigned n, unsigned multiple)
+{
+   return (n + multiple - 1) & ~(multiple - 1);
+}
+
+
+static struct pipe_buffer *
+gdi_softpipe_surface_buffer_create(struct pipe_winsys *winsys,
+                                   unsigned width, unsigned height,
+                                   enum pipe_format format,
+                                   unsigned usage,
+                                   unsigned *stride)
+{
+   const unsigned alignment = 64;
+   struct pipe_format_block block;
+   unsigned nblocksx, nblocksy;
+
+   pf_get_block(format, &block);
+   nblocksx = pf_get_nblocksx(&block, width);
+   nblocksy = pf_get_nblocksy(&block, height);
+   *stride = round_up(nblocksx * block.size, alignment);
+
+   return winsys->buffer_create(winsys, alignment,
+                                usage,
+                                *stride * nblocksy);
+}
+
+
+static void
+gdi_softpipe_dummy_flush_frontbuffer(struct pipe_winsys *winsys,
+                                     struct pipe_surface *surface,
+                                     void *context_private)
+{
+   assert(0);
+}
+
+
+static void
+gdi_softpipe_fence_reference(struct pipe_winsys *winsys,
+                             struct pipe_fence_handle **ptr,
+                             struct pipe_fence_handle *fence)
+{
+}
+
+
+static int
+gdi_softpipe_fence_signalled(struct pipe_winsys *winsys,
+                             struct pipe_fence_handle *fence,
+                             unsigned flag)
+{
+   return 0;
+}
+
+
+static int
+gdi_softpipe_fence_finish(struct pipe_winsys *winsys,
+                          struct pipe_fence_handle *fence,
+                          unsigned flag)
+{
+   return 0;
+}
+
+
+static void
+gdi_softpipe_destroy(struct pipe_winsys *winsys)
+{
+   FREE(winsys);
+}
+
+
+static struct pipe_screen *
+gdi_softpipe_screen_create(void)
+{
+   static struct pipe_winsys *winsys;
+   struct pipe_screen *screen;
+
+   winsys = CALLOC_STRUCT(pipe_winsys);
+   if(!winsys)
+      return NULL;
+
+   winsys->destroy = gdi_softpipe_destroy;
+
+   winsys->buffer_create = gdi_softpipe_buffer_create;
+   winsys->user_buffer_create = gdi_softpipe_user_buffer_create;
+   winsys->buffer_map = gdi_softpipe_buffer_map;
+   winsys->buffer_unmap = gdi_softpipe_buffer_unmap;
+   winsys->buffer_destroy = gdi_softpipe_buffer_destroy;
+
+   winsys->surface_buffer_create = gdi_softpipe_surface_buffer_create;
+
+   winsys->fence_reference = gdi_softpipe_fence_reference;
+   winsys->fence_signalled = gdi_softpipe_fence_signalled;
+   winsys->fence_finish = gdi_softpipe_fence_finish;
+
+   winsys->flush_frontbuffer = gdi_softpipe_dummy_flush_frontbuffer;
+   winsys->get_name = gdi_softpipe_get_name;
+
+   screen = softpipe_create_screen(winsys);
+   if(!screen)
+      gdi_softpipe_destroy(winsys);
+
+   return screen;
+}
+
+
+static struct pipe_context *
+gdi_softpipe_context_create(struct pipe_screen *screen)
+{
+   return softpipe_create(screen, screen->winsys, NULL);
+}
+
+
+static void
+gdi_softpipe_flush_frontbuffer(struct pipe_screen *screen,
+                               struct pipe_surface *surface,
+                               HDC hDC)
+{
+    struct gdi_softpipe_buffer *buffer;
+    BITMAPINFO bmi;
+
+    buffer = gdi_softpipe_buffer(softpipe_texture(surface->texture)->buffer);
+
+    memset(&bmi, 0, sizeof(BITMAPINFO));
+    bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER);
+    bmi.bmiHeader.biWidth = surface->stride / pf_get_size(surface->format);
+    bmi.bmiHeader.biHeight= -(long)surface->height;
+    bmi.bmiHeader.biPlanes = 1;
+    bmi.bmiHeader.biBitCount = pf_get_bits(surface->format);
+    bmi.bmiHeader.biCompression = BI_RGB;
+    bmi.bmiHeader.biSizeImage = 0;
+    bmi.bmiHeader.biXPelsPerMeter = 0;
+    bmi.bmiHeader.biYPelsPerMeter = 0;
+    bmi.bmiHeader.biClrUsed = 0;
+    bmi.bmiHeader.biClrImportant = 0;
+
+    StretchDIBits(hDC,
+                  0, 0, surface->width, surface->height,
+                  0, 0, surface->width, surface->height,
+                  buffer->data, &bmi, 0, SRCCOPY);
+}
+
+
+static const struct stw_winsys stw_winsys = {
+   &gdi_softpipe_screen_create,
+   &gdi_softpipe_context_create,
+   &gdi_softpipe_flush_frontbuffer
+};
+
+
+BOOL WINAPI
+DllMain(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpReserved)
+{
+   switch (fdwReason) {
+   case DLL_PROCESS_ATTACH:
+      return st_init(&stw_winsys);
+
+   case DLL_PROCESS_DETACH:
+      st_cleanup();
+      break;
+   }
+   return TRUE;
+}
diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile
new file mode 100644
index 0000000000..bb187cc14a
--- /dev/null
+++ b/src/gallium/winsys/xlib/Makefile
@@ -0,0 +1,105 @@
+# src/gallium/winsys/xlib/Makefile
+
+# This makefile produces a "stand-alone" libGL.so which is based on
+# Xlib (no DRI HW acceleration)
+
+
+TOP = ../../../..
+include $(TOP)/configs/current
+
+
+GL_MAJOR = 1
+GL_MINOR = 5
+GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY)
+
+
+INCLUDE_DIRS = \
+	-I$(TOP)/include \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/mesa/main \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/gallium/state_trackers/glx/xlib \
+	-I$(TOP)/src/gallium/auxiliary
+
+DEFINES += \
+	-DGALLIUM_SOFTPIPE \
+	-DGALLIUM_TRACE \
+	-DGALLIUM_BRW
+#-DGALLIUM_CELL will be defined by the config */
+
+XLIB_WINSYS_SOURCES = \
+	xlib.c \
+	xlib_cell.c \
+	xlib_brw_aub.c \
+	xlib_brw_context.c \
+	xlib_brw_screen.c \
+	xlib_softpipe.c \
+	xlib_trace.c 
+
+
+XLIB_WINSYS_OBJECTS = $(XLIB_WINSYS_SOURCES:.c=.o)
+
+
+# Note: CELL_SPU_LIB is only defined for cell configs
+
+LIBS = \
+	$(GALLIUM_DRIVERS) \
+	$(TOP)/src/gallium/state_trackers/glx/xlib/libxlib.a \
+	$(TOP)/src/mesa/libglapi.a \
+	$(TOP)/src/mesa/libmesagallium.a \
+	$(GALLIUM_AUXILIARIES) \
+	$(CELL_SPU_LIB) \
+
+
+.SUFFIXES : .cpp
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(DEFINES) $(CFLAGS) $< -o $@
+
+.cpp.o:
+	$(CXX) -c $(INCLUDE_DIRS) $(DEFINES) $(CXXFLAGS) $< -o $@
+
+
+
+default: $(TOP)/$(LIB_DIR)/gallium $(TOP)/$(LIB_DIR)/gallium/$(GL_LIB_NAME)
+
+$(TOP)/$(LIB_DIR)/gallium:
+	@ mkdir -p $(TOP)/$(LIB_DIR)/gallium
+
+# Make the libGL.so library
+$(TOP)/$(LIB_DIR)/gallium/$(GL_LIB_NAME): $(XLIB_WINSYS_OBJECTS) $(LIBS) Makefile
+	$(TOP)/bin/mklib -o $(GL_LIB) \
+		-linker "$(CC)" \
+		-major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \
+		-install $(TOP)/$(LIB_DIR)/gallium \
+		$(MKLIB_OPTIONS) $(XLIB_WINSYS_OBJECTS) \
+		--start-group $(LIBS) --end-group $(GL_LIB_DEPS)
+
+
+depend: $(XLIB_WINSYS_SOURCES)
+	@ echo "running $(MKDEP)"
+	@ rm -f depend  # workaround oops on gutsy?!?
+	@ touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(XLIB_WINSYS_SOURCES) \
+		> /dev/null 2>/dev/null
+
+
+install: default
+	$(INSTALL) -d $(INSTALL_DIR)/include/GL
+	$(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR)
+	$(INSTALL) -m 644 $(TOP)/include/GL/*.h $(INSTALL_DIR)/include/GL
+	@if [ -e $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) ]; then \
+		$(INSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \
+	fi
+
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` $(TOP)/include/GL/*.h
+
+clean:
+	-rm -f *.o
+
+
+include depend
diff --git a/src/gallium/winsys/xlib/SConscript b/src/gallium/winsys/xlib/SConscript
new file mode 100644
index 0000000000..0fb4b50f63
--- /dev/null
+++ b/src/gallium/winsys/xlib/SConscript
@@ -0,0 +1,58 @@
+#######################################################################
+# SConscript for xlib winsys
+
+Import('*')
+
+if env['platform'] == 'linux' \
+        and 'mesa' in env['statetrackers'] \
+        and ('softpipe' or 'i915simple' or 'trace') in env['drivers'] \
+        and not env['dri']:
+
+    env = env.Clone()
+
+    env.Append(CPPPATH = [
+        '#/src/mesa',
+        '#/src/mesa/main',
+        '#src/gallium/state_trackers/glx/xlib',
+    ])
+
+    env.Append(CPPDEFINES = ['USE_XSHM'])
+
+    sources = [
+        'xlib.c',
+    ]
+
+    drivers = []
+        
+    if 'softpipe' in env['drivers']:
+        env.Append(CPPDEFINES = 'GALLIUM_SOFTPIPE')
+        sources += ['xlib_softpipe.c']
+        drivers += [softpipe]
+
+    if 'i965simple' in env['drivers']:
+        env.Append(CPPDEFINES = 'GALLIUM_I965SIMPLE')
+        sources += [
+            'xlib_brw_aub.c',
+            'xlib_brw_context.c',
+            'xlib_brw_screen.c',
+        ]
+        drivers += [i965simple]
+        
+    if 'cell' in env['drivers']:
+        env.Append(CPPDEFINES = 'GALLIUM_CELL')
+        sources += ['xlib_cell.c']
+        drivers += [cell]
+
+    if 'trace' in env['drivers']:
+        env.Append(CPPDEFINES = 'GALLIUM_TRACE')
+        sources += ['xlib_trace.c']
+        drivers += [trace]
+
+    # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions
+    libgl = env.SharedLibrary(
+        target ='GL',
+        source = sources,
+        LIBS = st_xlib + glapi + mesa + drivers + auxiliaries + env['LIBS'],
+    )
+
+    env.InstallSharedLibrary(libgl, version=(1, 5))
diff --git a/src/gallium/winsys/xlib/xlib.c b/src/gallium/winsys/xlib/xlib.c
new file mode 100644
index 0000000000..da72228215
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib.c
@@ -0,0 +1,113 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/*
+ * Authors:
+ *   Keith Whitwell
+ */
+
+#include "xlib.h"
+#include "xm_winsys.h"
+
+#include <stdlib.h>
+#include <assert.h>
+
+/* Todo, replace all this with callback-structs provided by the
+ * individual implementations.
+ */
+
+enum mode {
+   MODE_TRACE,
+   MODE_BRW,
+   MODE_CELL,
+   MODE_SOFTPIPE
+};
+
+
+static enum mode get_mode()
+{
+   if (getenv("XMESA_TRACE"))
+      return MODE_TRACE;
+
+   if (getenv("XMESA_BRW"))
+      return MODE_BRW;
+
+#ifdef GALLIUM_CELL
+   if (!getenv("GALLIUM_NOCELL")) 
+      return MODE_CELL;
+#endif
+
+   return MODE_SOFTPIPE;
+}
+
+static void _init( void ) __attribute__((constructor));
+
+static void _init( void )
+{
+   enum mode xlib_mode = get_mode();
+
+   switch (xlib_mode) {
+   case MODE_TRACE:
+#if defined(GALLIUM_TRACE) && defined(GALLIUM_SOFTPIPE)
+      xmesa_set_driver( &xlib_trace_driver );
+#endif
+      break;
+   case MODE_BRW:
+#if defined(GALLIUM_BRW)
+      xmesa_set_driver( &xlib_brw_driver );
+#endif
+      break;
+   case MODE_CELL:
+#if defined(GALLIUM_CELL)
+      xmesa_set_driver( &xlib_cell_driver );
+#endif
+      break;
+   case MODE_SOFTPIPE:
+#if defined(GALLIUM_SOFTPIPE)
+      xmesa_set_driver( &xlib_softpipe_driver );
+#endif
+      break;
+   default:
+      assert(0);
+      break;
+   }
+}
+
+
+/***********************************************************************
+ *
+ * Butt-ugly hack to convince the linker not to throw away public GL
+ * symbols (they are all referenced from getprocaddress, I guess).
+ */
+extern void (*linker_foo(const unsigned char *procName))();
+extern void (*glXGetProcAddress(const unsigned char *procName))();
+
+extern void (*linker_foo(const unsigned char *procName))()
+{
+   return glXGetProcAddress(procName);
+}
diff --git a/src/gallium/winsys/xlib/xlib.h b/src/gallium/winsys/xlib/xlib.h
new file mode 100644
index 0000000000..d602ab0b13
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib.h
@@ -0,0 +1,14 @@
+
+#ifndef XLIB_H
+#define XLIB_H
+
+#include "pipe/p_compiler.h"
+#include "xm_winsys.h"
+
+extern struct xm_driver xlib_trace_driver;
+extern struct xm_driver xlib_softpipe_driver;
+extern struct xm_driver xlib_cell_driver;
+extern struct xm_driver xlib_brw_driver;
+
+
+#endif
diff --git a/src/gallium/winsys/xlib/xlib_brw.h b/src/gallium/winsys/xlib/xlib_brw.h
new file mode 100644
index 0000000000..be2dd147db
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib_brw.h
@@ -0,0 +1,30 @@
+#ifndef XLIB_BRW_H
+#define XLIB_BRW_H
+
+struct pipe_winsys;
+struct pipe_buffer;
+struct pipe_surface;
+struct xmesa_buffer;
+
+unsigned xlib_brw_get_buffer_offset( struct pipe_winsys *pws,
+                                     struct pipe_buffer *buf,
+                                     unsigned access_flags );
+
+void xlib_brw_buffer_subdata_typed( struct pipe_winsys *pws,
+                                    struct pipe_buffer *buf,
+                                    unsigned long offset, 
+                                    unsigned long size, 
+                                    const void *data,
+                                    unsigned data_type );
+
+
+
+void xlib_brw_commands_aub(struct pipe_winsys *winsys,
+                           unsigned *cmds,
+                           unsigned nr_dwords);
+
+struct pipe_context *
+xlib_create_brw_context( struct pipe_screen *screen,
+                         void *unused );
+
+#endif
diff --git a/src/gallium/winsys/xlib/xlib_brw_aub.c b/src/gallium/winsys/xlib/xlib_brw_aub.c
new file mode 100644
index 0000000000..2956e1b960
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib_brw_aub.c
@@ -0,0 +1,397 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include "xlib_brw_aub.h"
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_debug.h"
+#include "util/u_memory.h"
+
+
+struct brw_aubfile {
+   FILE *file;
+   unsigned next_free_page;
+};
+
+
+extern char *__progname;
+
+
+struct aub_file_header {
+   unsigned int instruction_type;
+   unsigned int pad0:16;
+   unsigned int minor:8;
+   unsigned int major:8;
+   unsigned char application[8*4];
+   unsigned int day:8;
+   unsigned int month:8;
+   unsigned int year:16;
+   unsigned int timezone:8;
+   unsigned int second:8;
+   unsigned int minute:8;
+   unsigned int hour:8;
+   unsigned int comment_length:16;   
+   unsigned int pad1:16;
+};
+
+struct aub_block_header {
+   unsigned int instruction_type;
+   unsigned int operation:8;
+   unsigned int type:8;
+   unsigned int address_space:8;
+   unsigned int pad0:8;
+   unsigned int general_state_type:8;
+   unsigned int surface_state_type:8;
+   unsigned int pad1:16;
+   unsigned int address;
+   unsigned int length;
+};
+
+struct aub_dump_bmp {
+   unsigned int instruction_type;
+   unsigned int xmin:16;
+   unsigned int ymin:16;
+   unsigned int pitch:16;
+   unsigned int bpp:8;
+   unsigned int format:8;
+   unsigned int xsize:16;
+   unsigned int ysize:16;
+   unsigned int addr;
+   unsigned int unknown;
+};
+
+enum bh_operation {
+   BH_COMMENT,
+   BH_DATA_WRITE,
+   BH_COMMAND_WRITE,
+   BH_MMI0_WRITE32,
+   BH_END_SCENE,
+   BH_CONFIG_MEMORY_MAP,
+   BH_MAX_OPERATION
+};
+
+enum command_write_type {
+   CW_HWB_RING = 1,
+   CW_PRIMARY_RING_A,
+   CW_PRIMARY_RING_B,		/* XXX - disagreement with listaub! */
+   CW_PRIMARY_RING_C,
+   CW_MAX_TYPE
+};
+
+enum memory_map_type {
+   MM_DEFAULT,
+   MM_DYNAMIC,
+   MM_MAX_TYPE
+};
+
+enum address_space {
+   ADDR_GTT,
+   ADDR_LOCAL,
+   ADDR_MAIN,
+   ADDR_MAX
+};
+
+
+#define AUB_FILE_HEADER 0xe085000b
+#define AUB_BLOCK_HEADER 0xe0c10003
+#define AUB_DUMP_BMP 0xe09e0004
+
+/* Registers to control page table
+ */
+#define PGETBL_CTL       0x2020
+#define PGETBL_ENABLED   0x1
+
+#define NR_GTT_ENTRIES  65536	/* 256 mb */
+
+#define FAIL										\
+do {											\
+   fprintf(stderr, "failed to write aub data at %s/%d\n", __FUNCTION__, __LINE__);	\
+   exit(1);										\
+} while (0)
+
+
+/* Emit the headers at the top of each aubfile.  Initialize the GTT.
+ */
+static void init_aubfile( FILE *aub_file )
+{   
+   struct aub_file_header fh;
+   struct aub_block_header bh;
+   unsigned int data;
+
+   static int nr;
+   
+   nr++;
+
+   /* Emit the aub header:
+    */
+   memset(&fh, 0, sizeof(fh));
+
+   fh.instruction_type = AUB_FILE_HEADER;
+   fh.minor = 0x0;
+   fh.major = 0x7;
+   memcpy(fh.application, __progname, sizeof(fh.application));
+   fh.day = (nr>>24) & 0xff;
+   fh.month = 0x0;
+   fh.year = 0x0;
+   fh.timezone = 0x0;
+   fh.second = nr & 0xff;
+   fh.minute = (nr>>8) & 0xff;
+   fh.hour = (nr>>16) & 0xff;
+   fh.comment_length = 0x0;   
+
+   if (fwrite(&fh, sizeof(fh), 1, aub_file) < 0) 
+      FAIL;
+         
+   /* Setup the GTT starting at main memory address zero (!):
+    */
+   memset(&bh, 0, sizeof(bh));
+   
+   bh.instruction_type = AUB_BLOCK_HEADER;
+   bh.operation = BH_MMI0_WRITE32;
+   bh.type = 0x0;
+   bh.address_space = ADDR_GTT;	/* ??? */
+   bh.general_state_type = 0x0;
+   bh.surface_state_type = 0x0;
+   bh.address = PGETBL_CTL;
+   bh.length = 0x4;
+
+   if (fwrite(&bh, sizeof(bh), 1, aub_file) < 0) 
+      FAIL;
+
+   data = 0x0 | PGETBL_ENABLED;
+
+   if (fwrite(&data, sizeof(data), 1, aub_file) < 0) 
+      FAIL;
+}
+
+
+static void init_aub_gtt( struct brw_aubfile *aubfile,
+			  unsigned start_offset, 
+			  unsigned size )
+{
+   FILE *aub_file = aubfile->file;
+   struct aub_block_header bh;
+   unsigned int i;
+
+   assert(start_offset + size < NR_GTT_ENTRIES * 4096);
+
+
+   memset(&bh, 0, sizeof(bh));
+   
+   bh.instruction_type = AUB_BLOCK_HEADER;
+   bh.operation = BH_DATA_WRITE;
+   bh.type = 0x0;
+   bh.address_space = ADDR_MAIN;
+   bh.general_state_type = 0x0;
+   bh.surface_state_type = 0x0;
+   bh.address =  start_offset / 4096 * 4;
+   bh.length = size / 4096 * 4;
+
+   if (fwrite(&bh, sizeof(bh), 1, aub_file) < 0) 
+      FAIL;
+
+   for (i = 0; i < size / 4096; i++) {
+      unsigned data = aubfile->next_free_page | 1;
+
+      aubfile->next_free_page += 4096;
+
+      if (fwrite(&data, sizeof(data), 1, aub_file) < 0) 
+	 FAIL;
+   }
+
+}
+
+static void write_block_header( FILE *aub_file,
+				struct aub_block_header *bh,
+				const unsigned *data,
+				unsigned sz )
+{
+   sz = (sz + 3) & ~3;
+
+   if (fwrite(bh, sizeof(*bh), 1, aub_file) < 0) 
+      FAIL;
+
+   if (fwrite(data, sz, 1, aub_file) < 0) 
+      FAIL;
+
+   fflush(aub_file);
+}
+
+
+static void write_dump_bmp( FILE *aub_file,
+			    struct aub_dump_bmp *db )
+{
+   if (fwrite(db, sizeof(*db), 1, aub_file) < 0) 
+      FAIL;
+
+   fflush(aub_file);
+}
+
+
+
+void brw_aub_gtt_data( struct brw_aubfile *aubfile,
+		       unsigned offset,
+		       const void *data,
+		       unsigned sz,
+		       unsigned type,
+		       unsigned state_type )
+{
+   struct aub_block_header bh;
+
+   bh.instruction_type = AUB_BLOCK_HEADER;
+   bh.operation = BH_DATA_WRITE;
+   bh.type = type;
+   bh.address_space = ADDR_GTT;
+   bh.pad0 = 0;
+
+   if (type == DW_GENERAL_STATE) {
+      bh.general_state_type = state_type;
+      bh.surface_state_type = 0;
+   }
+   else {
+      bh.general_state_type = 0;
+      bh.surface_state_type = state_type;
+   }
+
+   bh.pad1 = 0;
+   bh.address = offset;
+   bh.length = sz;
+
+   write_block_header(aubfile->file, &bh, data, sz);
+}
+
+
+
+void brw_aub_gtt_cmds( struct brw_aubfile *aubfile,
+		       unsigned offset,
+		       const void *data,
+		       unsigned sz )
+{
+   struct aub_block_header bh;   
+   unsigned type = CW_PRIMARY_RING_A;
+   
+
+   bh.instruction_type = AUB_BLOCK_HEADER;
+   bh.operation = BH_COMMAND_WRITE;
+   bh.type = type;
+   bh.address_space = ADDR_GTT;
+   bh.pad0 = 0;
+   bh.general_state_type = 0;
+   bh.surface_state_type = 0;
+   bh.pad1 = 0;
+   bh.address = offset;
+   bh.length = sz;
+
+   write_block_header(aubfile->file, &bh, data, sz);
+}
+
+void brw_aub_dump_bmp( struct brw_aubfile *aubfile,
+		       struct pipe_surface *surface,
+		       unsigned gtt_offset )
+{
+   struct aub_dump_bmp db;
+   unsigned format;
+
+   assert(surface->block.width == 1);
+   assert(surface->block.height == 1);
+   
+   if (surface->block.size == 4)
+      format = 0x7;
+   else
+      format = 0x3;
+
+   db.instruction_type = AUB_DUMP_BMP;
+   db.xmin = 0;
+   db.ymin = 0;
+   db.format = format;
+   db.bpp = surface->block.size * 8;
+   db.pitch = surface->stride/surface->block.size;
+   db.xsize = surface->width;
+   db.ysize = surface->height;
+   db.addr = gtt_offset;
+   db.unknown = /* surface->tiled ? 0x4 : */ 0x0;
+
+   write_dump_bmp(aubfile->file, &db);
+}
+
+
+
+struct brw_aubfile *brw_aubfile_create( void )
+{
+   struct brw_aubfile *aubfile = CALLOC_STRUCT(brw_aubfile);
+   char filename[80];
+   int val;
+   static int i = 0;
+
+   i++;
+
+   if (getenv("INTEL_AUBFILE")) {
+      val = snprintf(filename, sizeof(filename), "%s%d.aub", getenv("INTEL_AUBFILE"), i%4);
+      debug_printf("--> Aub file: %s\n", filename);
+      aubfile->file = fopen(filename, "w");
+   }
+   else {
+      val = snprintf(filename, sizeof(filename), "%s.aub", __progname);
+      if (val < 0 || val > sizeof(filename)) 
+	 strcpy(filename, "default.aub");   
+   
+      debug_printf("--> Aub file: %s\n", filename);
+      aubfile->file = fopen(filename, "w");
+   }
+
+   if (!aubfile->file) {
+      debug_printf("couldn't open aubfile\n");
+      exit(1);
+   }
+
+   init_aubfile(aubfile->file);
+
+   /* The GTT is located starting address zero in main memory.  Pages
+    * to populate the gtt start after this point.
+    */
+   aubfile->next_free_page = (NR_GTT_ENTRIES * 4 + 4095) & ~4095;
+
+   /* More or less correspond with all the agp regions mapped by the
+    * driver:
+    */
+   init_aub_gtt(aubfile, 0, 4096*4);
+   init_aub_gtt(aubfile, AUB_BUF_START, AUB_BUF_SIZE);
+
+   return aubfile;
+}
+
+void brw_aub_destroy( struct brw_aubfile *aubfile )
+{
+   fclose(aubfile->file);
+   FREE(aubfile);
+}
diff --git a/src/gallium/winsys/xlib/xlib_brw_aub.h b/src/gallium/winsys/xlib/xlib_brw_aub.h
new file mode 100644
index 0000000000..f5c60c7be2
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib_brw_aub.h
@@ -0,0 +1,114 @@
+/*
+ Copyright (C) Intel Corp.  2006.  All Rights Reserved.
+ Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
+ develop this 3D driver.
+ 
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+ 
+ The above copyright notice and this permission notice (including the
+ next paragraph) shall be included in all copies or substantial
+ portions of the Software.
+ 
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
+ LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+ OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ 
+ **********************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#ifndef BRW_AUB_H
+#define BRW_AUB_H
+
+/* We set up this region, buffers may be allocated here:
+ */
+#define AUB_BUF_START (4096*4)
+#define AUB_BUF_SIZE  (8*1024*1024)
+
+struct intel_context;
+struct pipe_surface;
+
+struct brw_aubfile *brw_aubfile_create( void );
+
+void brw_aub_destroy( struct brw_aubfile *aubfile );
+
+void brw_aub_gtt_data( struct brw_aubfile *aubfile,
+		       unsigned offset,
+		       const void *data,
+		       unsigned sz,
+		       unsigned type,
+		       unsigned state_type );
+
+void brw_aub_gtt_cmds( struct brw_aubfile *aubfile,
+		       unsigned offset,
+		       const void *data,
+		       unsigned sz );
+
+void brw_aub_dump_bmp( struct brw_aubfile *aubfile,
+		       struct pipe_surface *surface,
+		       unsigned gtt_offset );
+
+
+enum data_write_type {
+   DW_NOTYPE,
+   DW_BATCH_BUFFER,
+   DW_BIN_BUFFER,
+   DW_BIN_POINTER_LIST,
+   DW_SLOW_STATE_BUFFER,
+   DW_VERTEX_BUFFER,
+   DW_2D_MAP,
+   DW_CUBE_MAP,
+   DW_INDIRECT_STATE_BUFFER,
+   DW_VOLUME_MAP,
+   DW_1D_MAP,
+   DW_CONSTANT_BUFFER,
+   DW_CONSTANT_URB_ENTRY,
+   DW_INDEX_BUFFER,
+   DW_GENERAL_STATE,
+   DW_SURFACE_STATE,
+   DW_MEDIA_OBJECT_INDIRECT_DATA,
+   DW_MAX_TYPE
+};
+
+enum data_write_general_state_type {
+   DWGS_NOTYPE,
+   DWGS_VERTEX_SHADER_STATE,
+   DWGS_GEOMETRY_SHADER_STATE ,
+   DWGS_CLIPPER_STATE,
+   DWGS_STRIPS_FANS_STATE,
+   DWGS_WINDOWER_IZ_STATE,
+   DWGS_COLOR_CALC_STATE,
+   DWGS_CLIPPER_VIEWPORT_STATE,	/* was 0x7 */
+   DWGS_STRIPS_FANS_VIEWPORT_STATE,
+   DWGS_COLOR_CALC_VIEWPORT_STATE, /* was 0x9 */
+   DWGS_SAMPLER_STATE,
+   DWGS_KERNEL_INSTRUCTIONS,
+   DWGS_SCRATCH_SPACE,
+   DWGS_SAMPLER_DEFAULT_COLOR,
+   DWGS_INTERFACE_DESCRIPTOR,
+   DWGS_VLD_STATE,
+   DWGS_VFE_STATE,
+   DWGS_MAX_TYPE
+};
+
+enum data_write_surface_state_type {
+   DWSS_NOTYPE,
+   DWSS_BINDING_TABLE_STATE,
+   DWSS_SURFACE_STATE,
+   DWSS_MAX_TYPE
+};
+
+
+#endif
diff --git a/src/gallium/winsys/xlib/xlib_brw_context.c b/src/gallium/winsys/xlib/xlib_brw_context.c
new file mode 100644
index 0000000000..09599507f4
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib_brw_context.c
@@ -0,0 +1,209 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/*
+ * Authors:
+ *   Keith Whitwell
+ *   Brian Paul
+ */
+
+
+//#include "glxheader.h"
+//#include "xmesaP.h"
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "i965simple/brw_winsys.h"
+#include "xlib_brw_aub.h"
+#include "xlib_brw.h"
+
+
+
+
+#define XBCWS_BATCHBUFFER_SIZE 1024
+
+
+/* The backend to the brw driver (ie struct brw_winsys) is actually a
+ * per-context entity.
+ */
+struct xlib_brw_context_winsys {
+   struct brw_winsys brw_context_winsys;   /**< batch buffer funcs */
+   struct aub_context *aub;
+                         
+   struct pipe_winsys *pipe_winsys;
+
+   unsigned batch_data[XBCWS_BATCHBUFFER_SIZE];
+   unsigned batch_nr;
+   unsigned batch_size;
+   unsigned batch_alloc;
+};
+
+
+/* Turn a brw_winsys into an xlib_brw_context_winsys:
+ */
+static inline struct xlib_brw_context_winsys *
+xlib_brw_context_winsys( struct brw_winsys *sws )
+{
+   return (struct xlib_brw_context_winsys *)sws;
+}
+
+
+/* Simple batchbuffer interface:
+ */
+
+static unsigned *xbcws_batch_start( struct brw_winsys *sws,
+					 unsigned dwords,
+					 unsigned relocs )
+{
+   struct xlib_brw_context_winsys *xbcws = xlib_brw_context_winsys(sws);
+
+   if (xbcws->batch_size < xbcws->batch_nr + dwords)
+      return NULL;
+
+   xbcws->batch_alloc = xbcws->batch_nr + dwords;
+   return (void *)1;			/* not a valid pointer! */
+}
+
+static void xbcws_batch_dword( struct brw_winsys *sws,
+				    unsigned dword )
+{
+   struct xlib_brw_context_winsys *xbcws = xlib_brw_context_winsys(sws);
+
+   assert(xbcws->batch_nr < xbcws->batch_alloc);
+   xbcws->batch_data[xbcws->batch_nr++] = dword;
+}
+
+static void xbcws_batch_reloc( struct brw_winsys *sws,
+			     struct pipe_buffer *buf,
+			     unsigned access_flags,
+			     unsigned delta )
+{
+   struct xlib_brw_context_winsys *xbcws = xlib_brw_context_winsys(sws);
+
+   assert(xbcws->batch_nr < xbcws->batch_alloc);
+   xbcws->batch_data[xbcws->batch_nr++] = 
+      ( xlib_brw_get_buffer_offset( NULL, buf, access_flags ) +
+        delta );
+}
+
+static void xbcws_batch_end( struct brw_winsys *sws )
+{
+   struct xlib_brw_context_winsys *xbcws = xlib_brw_context_winsys(sws);
+
+   assert(xbcws->batch_nr <= xbcws->batch_alloc);
+   xbcws->batch_alloc = 0;
+}
+
+static void xbcws_batch_flush( struct brw_winsys *sws,
+				    struct pipe_fence_handle **fence )
+{
+   struct xlib_brw_context_winsys *xbcws = xlib_brw_context_winsys(sws);
+   assert(xbcws->batch_nr <= xbcws->batch_size);
+
+   if (xbcws->batch_nr) {
+      xlib_brw_commands_aub( xbcws->pipe_winsys,
+                             xbcws->batch_data,
+                             xbcws->batch_nr );
+   }
+
+   xbcws->batch_nr = 0;
+}
+
+  
+
+/* Really a per-device function, just pass through:
+ */
+static unsigned xbcws_get_buffer_offset( struct brw_winsys *sws,
+                                         struct pipe_buffer *buf,
+                                         unsigned access_flags )
+{
+   struct xlib_brw_context_winsys *xbcws = xlib_brw_context_winsys(sws);
+
+   return xlib_brw_get_buffer_offset( xbcws->pipe_winsys,
+                                      buf,
+                                      access_flags );
+}
+
+
+/* Really a per-device function, just pass through:
+ */
+static void xbcws_buffer_subdata_typed( struct brw_winsys *sws,
+                                       struct pipe_buffer *buf,
+                                       unsigned long offset, 
+                                       unsigned long size, 
+                                       const void *data,
+                                       unsigned data_type )
+{
+   struct xlib_brw_context_winsys *xbcws = xlib_brw_context_winsys(sws);
+
+   xlib_brw_buffer_subdata_typed( xbcws->pipe_winsys,
+                                  buf,
+                                  offset,
+                                  size,
+                                  data,
+                                  data_type );
+}
+
+
+/**
+ * Create i965 hardware rendering context, but plugged into a
+ * dump-to-aubfile backend.
+ */
+struct pipe_context *
+xlib_create_brw_context( struct pipe_screen *screen,
+                         void *unused )
+{
+   struct xlib_brw_context_winsys *xbcws = CALLOC_STRUCT( xlib_brw_context_winsys );
+   
+   /* Fill in this struct with callbacks that i965simple will need to
+    * communicate with the window system, buffer manager, etc. 
+    */
+   xbcws->brw_context_winsys.batch_start = xbcws_batch_start;
+   xbcws->brw_context_winsys.batch_dword = xbcws_batch_dword;
+   xbcws->brw_context_winsys.batch_reloc = xbcws_batch_reloc;
+   xbcws->brw_context_winsys.batch_end = xbcws_batch_end;
+   xbcws->brw_context_winsys.batch_flush = xbcws_batch_flush;
+   xbcws->brw_context_winsys.buffer_subdata_typed = xbcws_buffer_subdata_typed;
+   xbcws->brw_context_winsys.get_buffer_offset = xbcws_get_buffer_offset;
+
+   xbcws->pipe_winsys = screen->winsys; /* redundant */
+
+   xbcws->batch_size = XBCWS_BATCHBUFFER_SIZE;
+
+   /* Create the i965simple context:
+    */
+#ifdef GALLIUM_CELL
+   return NULL;
+#else
+   return brw_create( screen,
+		      &xbcws->brw_context_winsys,
+		      0 );
+#endif
+}
diff --git a/src/gallium/winsys/xlib/xlib_brw_screen.c b/src/gallium/winsys/xlib/xlib_brw_screen.c
new file mode 100644
index 0000000000..8e1bfab2f5
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib_brw_screen.c
@@ -0,0 +1,470 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/*
+ * Authors:
+ *   Keith Whitwell
+ *   Brian Paul
+ */
+
+
+//#include "state_trackers/xlib/glxheader.h"
+//#include "state_trackers/xlib/xmesaP.h"
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "i965simple/brw_winsys.h"
+#include "i965simple/brw_screen.h"
+#include "i965simple/brw_context.h"
+
+
+#include "xlib_brw_aub.h"
+#include "xlib_brw.h"
+#include "xlib.h"
+
+static struct pipe_buffer *
+buffer_from_surface(struct pipe_surface *surface)
+{
+   struct brw_texture *texture = (struct brw_texture *)surface;
+   return texture->buffer;
+}
+
+struct aub_buffer {
+   char *data;
+   unsigned offset;
+   unsigned size;
+   unsigned refcount;
+   unsigned map_count;
+   boolean dump_on_unmap;
+};
+
+
+
+struct aub_pipe_winsys {
+   struct pipe_winsys winsys;
+
+   struct brw_aubfile *aubfile;
+
+   /* This is simple, isn't it:
+    */
+   char *pool;
+   unsigned size;
+   unsigned used;
+};
+
+
+/* Turn a pipe winsys into an aub/pipe winsys:
+ */
+static inline struct aub_pipe_winsys *
+aub_pipe_winsys( struct pipe_winsys *winsys )
+{
+   return (struct aub_pipe_winsys *)winsys;
+}
+
+
+
+static INLINE struct aub_buffer *
+aub_bo( struct pipe_buffer *bo )
+{
+   return (struct aub_buffer *)bo;
+}
+
+static INLINE struct pipe_buffer *
+pipe_bo( struct aub_buffer *bo )
+{
+   return (struct pipe_buffer *)bo;
+}
+
+
+
+
+static void *aub_buffer_map(struct pipe_winsys *winsys, 
+			      struct pipe_buffer *buf,
+			      unsigned flags )
+{
+   struct aub_buffer *sbo = aub_bo(buf);
+
+   assert(sbo->data);
+
+   if (flags & PIPE_BUFFER_USAGE_CPU_WRITE)
+      sbo->dump_on_unmap = 1;
+
+   sbo->map_count++;
+   return sbo->data;
+}
+
+static void aub_buffer_unmap(struct pipe_winsys *winsys, 
+			       struct pipe_buffer *buf)
+{
+   struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys);
+   struct aub_buffer *sbo = aub_bo(buf);
+
+   sbo->map_count--;
+
+   if (sbo->map_count == 0 &&
+       sbo->dump_on_unmap) {
+
+      sbo->dump_on_unmap = 0;
+
+      brw_aub_gtt_data( iws->aubfile, 
+			sbo->offset,
+			sbo->data,
+			sbo->size,
+			0,
+			0);
+   }
+}
+
+
+static void
+aub_buffer_destroy(struct pipe_winsys *winsys,
+		   struct pipe_buffer *buf)
+{
+   free(buf);
+}
+
+
+
+void xlib_brw_commands_aub(struct pipe_winsys *winsys,
+			unsigned *cmds,
+			unsigned nr_dwords)
+{
+   struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys);
+   unsigned size = nr_dwords * 4;
+
+   assert(iws->used + size < iws->size);
+
+   brw_aub_gtt_cmds( iws->aubfile, 
+		     AUB_BUF_START + iws->used,
+		     cmds,
+		     nr_dwords * sizeof(int) );
+
+   iws->used += align(size, 4096);
+}
+
+
+/* XXX: fix me:
+ */
+static struct aub_pipe_winsys *global_winsys = NULL;
+
+
+
+
+/* Pipe has no concept of pools.  We choose the tex/region pool
+ * for all buffers.
+ */
+static struct pipe_buffer *
+aub_buffer_create(struct pipe_winsys *winsys,
+                  unsigned alignment,
+                  unsigned usage,
+                  unsigned size)
+{
+   struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys);
+   struct aub_buffer *sbo = CALLOC_STRUCT(aub_buffer);
+
+   sbo->refcount = 1;
+
+   /* Could reuse buffers that are not referenced in current
+    * batchbuffer.  Can't do that atm, so always reallocate:
+    */
+   assert(iws->used + size < iws->size);
+   sbo->data = iws->pool + iws->used;
+   sbo->offset = AUB_BUF_START + iws->used;
+   iws->used += align(size, 4096);
+
+   sbo->size = size;
+
+   return pipe_bo(sbo);
+}
+
+
+static struct pipe_buffer *
+aub_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes)
+{
+   struct aub_buffer *sbo;
+
+   /* Lets hope this is meant for upload, not as a result!  
+    */
+   sbo = aub_bo(aub_buffer_create( winsys, 0, 0, 0 ));
+
+   sbo->data = ptr;
+   sbo->size = bytes;
+
+   return pipe_bo(sbo);
+}
+
+
+/* The state tracker (should!) keep track of whether the fake
+ * frontbuffer has been touched by any rendering since the last time
+ * we copied its contents to the real frontbuffer.  Our task is easy:
+ */
+static void
+aub_flush_frontbuffer( struct pipe_winsys *winsys,
+                       struct pipe_surface *surface,
+                       void *context_private)
+{
+//   struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys);
+   brw_aub_dump_bmp( global_winsys->aubfile, 
+		     surface,
+		     aub_bo(buffer_from_surface(surface))->offset );
+}
+
+
+/**
+ * Round n up to next multiple.
+ */
+static INLINE unsigned
+round_up(unsigned n, unsigned multiple)
+{
+   return (n + multiple - 1) & ~(multiple - 1);
+}
+
+static struct pipe_buffer *
+aub_i915_surface_buffer_create(struct pipe_winsys *winsys,
+                               unsigned width, unsigned height,
+                               enum pipe_format format,
+                               unsigned usage,
+                               unsigned *stride)
+{
+   const unsigned alignment = 64;
+   struct pipe_format_block block;
+   unsigned nblocksx, nblocksy;
+
+   pf_get_block(format, &block);
+   nblocksx = pf_get_nblocksx(&block, width);
+   nblocksy = pf_get_nblocksy(&block, height);
+   *stride = round_up(nblocksx * block.size, alignment);
+
+   return winsys->buffer_create(winsys, alignment,
+                                usage,
+                                *stride * nblocksy);
+}
+
+
+static const char *
+aub_get_name( struct pipe_winsys *winsys )
+{
+   return "Aub/xlib";
+}
+
+static void
+xlib_brw_destroy_pipe_winsys_aub( struct pipe_winsys *winsys )
+
+{
+   struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys);
+   brw_aub_destroy(iws->aubfile);
+   free(iws->pool);
+   free(iws);
+}
+
+
+
+static struct pipe_winsys *
+xlib_create_brw_winsys( void )
+{
+   struct aub_pipe_winsys *iws = CALLOC_STRUCT( aub_pipe_winsys );
+   
+   /* Fill in this struct with callbacks that pipe will need to
+    * communicate with the window system, buffer manager, etc. 
+    *
+    * Pipe would be happy with a malloc based memory manager, but
+    * the SwapBuffers implementation in this winsys driver requires
+    * that rendering be done to an appropriate _DriBufferObject.  
+    */
+   iws->winsys.buffer_create = aub_buffer_create;
+   iws->winsys.user_buffer_create = aub_user_buffer_create;
+   iws->winsys.buffer_map = aub_buffer_map;
+   iws->winsys.buffer_unmap = aub_buffer_unmap;
+   iws->winsys.buffer_destroy = aub_buffer_destroy;
+   iws->winsys.flush_frontbuffer = aub_flush_frontbuffer;
+   iws->winsys.get_name = aub_get_name;
+   iws->winsys.destroy = xlib_brw_destroy_pipe_winsys_aub;
+
+   iws->winsys.surface_buffer_create = aub_i915_surface_buffer_create;
+
+   iws->aubfile = brw_aubfile_create();
+   iws->size = AUB_BUF_SIZE;
+   iws->pool = malloc(AUB_BUF_SIZE);
+
+   /* HACK: static copy of this pointer:
+    */
+   assert(global_winsys == NULL);
+   global_winsys = iws;
+
+   return &iws->winsys;
+}
+
+
+static struct pipe_screen *
+xlib_create_brw_screen( void )
+{
+#ifndef GALLIUM_CELL
+   struct pipe_winsys *winsys;
+   struct pipe_screen *screen;
+
+   winsys = xlib_create_brw_winsys();
+   if (winsys == NULL)
+      return NULL;
+
+   screen = brw_create_screen(winsys, 0/* XXX pci_id */);
+   if (screen == NULL)
+      goto fail;
+
+   return screen;
+
+fail:
+   if (winsys)
+      winsys->destroy( winsys );
+
+#endif
+   return NULL;
+}
+
+
+/* These per-screen functions are acually made available to the driver
+ * through the brw_winsys (per-context) entity.
+ */
+unsigned xlib_brw_get_buffer_offset( struct pipe_winsys *pws,
+                                     struct pipe_buffer *buf,
+                                     unsigned access_flags )
+{
+   return aub_bo(buf)->offset;
+}
+
+void xlib_brw_buffer_subdata_typed( struct pipe_winsys *pws,
+                                    struct pipe_buffer *buf,
+                                    unsigned long offset, 
+                                    unsigned long size, 
+                                    const void *data,
+                                    unsigned data_type )
+{
+   unsigned aub_type = DW_GENERAL_STATE;
+   unsigned aub_sub_type;
+
+   switch (data_type) {
+   case BRW_CC_VP:
+      aub_sub_type = DWGS_COLOR_CALC_VIEWPORT_STATE;
+      break;
+   case BRW_CC_UNIT:
+      aub_sub_type = DWGS_COLOR_CALC_STATE;
+      break;
+   case BRW_WM_PROG:
+      aub_sub_type = DWGS_KERNEL_INSTRUCTIONS;
+      break;
+   case BRW_SAMPLER_DEFAULT_COLOR:
+      aub_sub_type = DWGS_SAMPLER_DEFAULT_COLOR;
+      break;
+   case BRW_SAMPLER:
+      aub_sub_type = DWGS_SAMPLER_STATE;
+      break;
+   case BRW_WM_UNIT:
+      aub_sub_type = DWGS_WINDOWER_IZ_STATE;
+      break;
+   case BRW_SF_PROG:
+      aub_sub_type = DWGS_KERNEL_INSTRUCTIONS;
+      break;
+   case BRW_SF_VP:
+      aub_sub_type = DWGS_STRIPS_FANS_VIEWPORT_STATE;
+      break;
+   case BRW_SF_UNIT:
+      aub_sub_type = DWGS_STRIPS_FANS_STATE;
+      break;
+   case BRW_VS_UNIT:
+      aub_sub_type = DWGS_VERTEX_SHADER_STATE;
+      break;
+   case BRW_VS_PROG:
+      aub_sub_type = DWGS_KERNEL_INSTRUCTIONS;
+      break;
+   case BRW_GS_UNIT:
+      aub_sub_type = DWGS_GEOMETRY_SHADER_STATE;
+      break;
+   case BRW_GS_PROG:
+      aub_sub_type = DWGS_KERNEL_INSTRUCTIONS;
+      break;
+   case BRW_CLIP_VP:
+      aub_sub_type = DWGS_CLIPPER_VIEWPORT_STATE;
+      break;
+   case BRW_CLIP_UNIT:
+      aub_sub_type = DWGS_CLIPPER_STATE;
+      break;
+   case BRW_CLIP_PROG:
+      aub_sub_type = DWGS_KERNEL_INSTRUCTIONS;
+      break;
+   case BRW_SS_SURFACE:
+      aub_type = DW_SURFACE_STATE;
+      aub_sub_type = DWSS_SURFACE_STATE; 
+      break;
+   case BRW_SS_SURF_BIND:
+      aub_type = DW_SURFACE_STATE;
+      aub_sub_type = DWSS_BINDING_TABLE_STATE; 
+      break;
+   case BRW_CONSTANT_BUFFER:
+      aub_type = DW_CONSTANT_URB_ENTRY;
+      aub_sub_type = 0; 
+      break;
+
+   default:
+      assert(0);
+      break;
+   }
+
+   {
+      struct aub_pipe_winsys *iws = aub_pipe_winsys(pws);
+      struct aub_buffer *sbo = aub_bo(buf);
+
+      assert(sbo->size > offset + size);
+      memcpy(sbo->data + offset, data, size);
+
+      brw_aub_gtt_data( iws->aubfile, 
+                        sbo->offset + offset,
+                        sbo->data + offset,
+                        size,
+                        aub_type,
+                        aub_sub_type );
+   }
+}
+ 
+
+static void
+xlib_brw_display_surface(struct xmesa_buffer *b, 
+                         struct pipe_surface *surf)
+{
+   brw_aub_dump_bmp( global_winsys->aubfile, 
+		     surf,
+		     aub_bo(buffer_from_surface(surf))->offset );
+}
+
+
+struct xm_driver xlib_brw_driver = 
+{
+   .create_pipe_screen = xlib_create_brw_screen,
+   .create_pipe_context = xlib_create_brw_context,
+   .display_surface = xlib_brw_display_surface,
+};
diff --git a/src/gallium/winsys/xlib/xlib_cell.c b/src/gallium/winsys/xlib/xlib_cell.c
new file mode 100644
index 0000000000..c87564f4dc
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib_cell.c
@@ -0,0 +1,437 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/*
+ * Authors:
+ *   Keith Whitwell
+ *   Brian Paul
+ */
+
+#include "xlib.h"
+
+#ifdef GALLIUM_CELL
+
+#include "xm_api.h"
+
+#undef ASSERT
+#undef Elements
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "cell/ppu/cell_context.h"
+#include "cell/ppu/cell_screen.h"
+#include "cell/ppu/cell_winsys.h"
+#include "cell/ppu/cell_texture.h"
+
+
+/**
+ * Subclass of pipe_buffer for Xlib winsys.
+ * Low-level OS/window system memory buffer
+ */
+struct xm_buffer
+{
+   struct pipe_buffer base;
+   boolean userBuffer;  /** Is this a user-space buffer? */
+   void *data;
+   void *mapped;
+   
+   XImage *tempImage;
+   int shm;
+};
+
+
+/**
+ * Subclass of pipe_winsys for Xlib winsys
+ */
+struct xmesa_pipe_winsys
+{
+   struct pipe_winsys base;
+};
+
+
+
+/** Cast wrapper */
+static INLINE struct xm_buffer *
+xm_buffer( struct pipe_buffer *buf )
+{
+   return (struct xm_buffer *)buf;
+}
+
+
+/* Most callbacks map direcly onto dri_bufmgr operations:
+ */
+static void *
+xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer *buf,
+              unsigned flags)
+{
+   struct xm_buffer *xm_buf = xm_buffer(buf);
+   xm_buf->mapped = xm_buf->data;
+   return xm_buf->mapped;
+}
+
+static void
+xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf)
+{
+   struct xm_buffer *xm_buf = xm_buffer(buf);
+   xm_buf->mapped = NULL;
+}
+
+static void
+xm_buffer_destroy(struct pipe_winsys *pws,
+                  struct pipe_buffer *buf)
+{
+   struct xm_buffer *oldBuf = xm_buffer(buf);
+
+   if (oldBuf->data) {
+      {
+         if (!oldBuf->userBuffer) {
+            align_free(oldBuf->data);
+         }
+      }
+
+      oldBuf->data = NULL;
+   }
+
+   free(oldBuf);
+}
+
+
+/**
+ * For Cell.  Basically, rearrange the pixels/quads from this layout:
+ *  +--+--+--+--+
+ *  |p0|p1|p2|p3|....
+ *  +--+--+--+--+
+ *
+ * to this layout:
+ *  +--+--+
+ *  |p0|p1|....
+ *  +--+--+
+ *  |p2|p3|
+ *  +--+--+
+ */
+static void
+twiddle_tile(const uint *tileIn, uint *tileOut)
+{
+   int y, x;
+
+   for (y = 0; y < TILE_SIZE; y+=2) {
+      for (x = 0; x < TILE_SIZE; x+=2) {
+         int k = 4 * (y/2 * TILE_SIZE/2 + x/2);
+         tileOut[y * TILE_SIZE + (x + 0)] = tileIn[k];
+         tileOut[y * TILE_SIZE + (x + 1)] = tileIn[k+1];
+         tileOut[(y + 1) * TILE_SIZE + (x + 0)] = tileIn[k+2];
+         tileOut[(y + 1) * TILE_SIZE + (x + 1)] = tileIn[k+3];
+      }
+   }
+}
+
+
+
+/**
+ * Display a surface that's in a tiled configuration.  That is, all the
+ * pixels for a TILE_SIZExTILE_SIZE block are contiguous in memory.
+ */
+static void
+xlib_cell_display_surface(struct xmesa_buffer *b, struct pipe_surface *surf)
+{
+   XImage *ximage;
+   struct xm_buffer *xm_buf = xm_buffer(
+      cell_texture(surf->texture)->buffer);
+   const uint tilesPerRow = (surf->width + TILE_SIZE - 1) / TILE_SIZE;
+   uint x, y;
+
+   ximage = b->tempImage;
+
+   /* check that the XImage has been previously initialized */
+   assert(ximage->format);
+   assert(ximage->bitmap_unit);
+
+   /* update XImage's fields */
+   ximage->width = TILE_SIZE;
+   ximage->height = TILE_SIZE;
+   ximage->bytes_per_line = TILE_SIZE * 4;
+
+   for (y = 0; y < surf->height; y += TILE_SIZE) {
+      for (x = 0; x < surf->width; x += TILE_SIZE) {
+         uint tmpTile[TILE_SIZE * TILE_SIZE];
+         int tx = x / TILE_SIZE;
+         int ty = y / TILE_SIZE;
+         int offset = ty * tilesPerRow + tx;
+         int w = TILE_SIZE;
+         int h = TILE_SIZE;
+
+         if (y + h > surf->height)
+            h = surf->height - y;
+         if (x + w > surf->width)
+            w = surf->width - x;
+
+         /* offset in pixels */
+         offset *= TILE_SIZE * TILE_SIZE;
+
+         /* twiddle from ximage buffer to temp tile */
+         twiddle_tile((uint *) xm_buf->data + offset, tmpTile);
+         /* display temp tile data */
+         ximage->data = (char *) tmpTile;
+         XPutImage(b->xm_visual->display, b->drawable, b->gc,
+                   ximage, 0, 0, x, y, w, h);
+      }
+   }
+}
+
+
+
+
+
+static void
+xm_flush_frontbuffer(struct pipe_winsys *pws,
+                     struct pipe_surface *surf,
+                     void *context_private)
+{
+   /*
+    * The front color buffer is actually just another XImage buffer.
+    * This function copies that XImage to the actual X Window.
+    */
+   XMesaContext xmctx = (XMesaContext) context_private;
+   xlib_cell_display_surface(xmctx->xm_buffer, surf);
+}
+
+
+
+static const char *
+xm_get_name(struct pipe_winsys *pws)
+{
+   return "Xlib/Cell";
+}
+
+
+static struct pipe_buffer *
+xm_buffer_create(struct pipe_winsys *pws, 
+                 unsigned alignment, 
+                 unsigned usage,
+                 unsigned size)
+{
+   struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer);
+
+   buffer->base.refcount = 1;
+   buffer->base.alignment = alignment;
+   buffer->base.usage = usage;
+   buffer->base.size = size;
+
+
+   if (buffer->data == NULL) {
+      buffer->shm = 0;
+
+      /* align to 16-byte multiple for Cell */
+      buffer->data = align_malloc(size, max(alignment, 16));
+   }
+
+   return &buffer->base;
+}
+
+
+/**
+ * Create buffer which wraps user-space data.
+ */
+static struct pipe_buffer *
+xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes)
+{
+   struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer);
+   buffer->base.refcount = 1;
+   buffer->base.size = bytes;
+   buffer->userBuffer = TRUE;
+   buffer->data = ptr;
+   buffer->shm = 0;
+
+   return &buffer->base;
+}
+
+
+
+/**
+ * Round n up to next multiple.
+ */
+static INLINE unsigned
+round_up(unsigned n, unsigned multiple)
+{
+   return (n + multiple - 1) & ~(multiple - 1);
+}
+
+static struct pipe_buffer *
+xm_surface_buffer_create(struct pipe_winsys *winsys,
+                         unsigned width, unsigned height,
+                         enum pipe_format format,
+                         unsigned usage,
+                         unsigned *stride)
+{
+   const unsigned alignment = 64;
+   struct pipe_format_block block;
+   unsigned nblocksx, nblocksy;
+
+   pf_get_block(format, &block);
+   nblocksx = pf_get_nblocksx(&block, width);
+   nblocksy = pf_get_nblocksy(&block, height);
+   *stride = round_up(nblocksx * block.size, alignment);
+
+   return winsys->buffer_create(winsys, alignment,
+                                usage,
+                                /* XXX a bit of a hack */
+                                *stride * round_up(nblocksy, TILE_SIZE));
+}
+
+
+/*
+ * Fence functions - basically nothing to do, as we don't create any actual
+ * fence objects.
+ */
+
+static void
+xm_fence_reference(struct pipe_winsys *sws, struct pipe_fence_handle **ptr,
+                   struct pipe_fence_handle *fence)
+{
+}
+
+
+static int
+xm_fence_signalled(struct pipe_winsys *sws, struct pipe_fence_handle *fence,
+                   unsigned flag)
+{
+   return 0;
+}
+
+
+static int
+xm_fence_finish(struct pipe_winsys *sws, struct pipe_fence_handle *fence,
+                unsigned flag)
+{
+   return 0;
+}
+
+
+
+static struct pipe_winsys *
+xlib_create_cell_winsys( void )
+{
+   static struct xmesa_pipe_winsys *ws = NULL;
+
+   if (!ws) {
+      ws = CALLOC_STRUCT(xmesa_pipe_winsys);
+
+      /* Fill in this struct with callbacks that pipe will need to
+       * communicate with the window system, buffer manager, etc. 
+       */
+      ws->base.buffer_create = xm_buffer_create;
+      ws->base.user_buffer_create = xm_user_buffer_create;
+      ws->base.buffer_map = xm_buffer_map;
+      ws->base.buffer_unmap = xm_buffer_unmap;
+      ws->base.buffer_destroy = xm_buffer_destroy;
+
+      ws->base.surface_buffer_create = xm_surface_buffer_create;
+
+      ws->base.fence_reference = xm_fence_reference;
+      ws->base.fence_signalled = xm_fence_signalled;
+      ws->base.fence_finish = xm_fence_finish;
+
+      ws->base.flush_frontbuffer = xm_flush_frontbuffer;
+      ws->base.get_name = xm_get_name;
+   }
+
+   return &ws->base;
+}
+
+
+static struct pipe_screen *
+xlib_create_cell_screen( struct pipe_winsys *pws )
+{
+   struct pipe_winsys *winsys;
+   struct pipe_screen *screen;
+
+   winsys = xlib_create_cell_winsys();
+   if (winsys == NULL)
+      return NULL;
+
+   screen = cell_create_screen(winsys);
+   if (screen == NULL)
+      goto fail;
+
+   return screen;
+
+fail:
+   if (winsys)
+      winsys->destroy( winsys );
+
+   return NULL;
+}
+
+
+static struct pipe_context *
+xlib_create_cell_context( struct pipe_screen *screen,
+                          void *priv )
+{
+   struct pipe_context *pipe;
+
+   
+   /* This takes a cell_winsys pointer, but probably that should be
+    * created and stored at screen creation, not context creation.
+    *
+    * The actual cell_winsys value isn't used for anything, so just
+    * passing NULL for now.
+    */
+   pipe = cell_create_context( screen, NULL);
+   if (pipe == NULL)
+      goto fail;
+
+   pipe->priv = priv;
+
+   return pipe;
+
+fail:
+   return NULL;
+}
+
+struct xm_driver xlib_cell_driver = 
+{
+   .create_pipe_screen = xlib_create_cell_screen,
+   .create_pipe_context = xlib_create_cell_context,
+   .display_surface = xlib_cell_display_surface,
+};
+
+#else
+
+struct xm_driver xlib_cell_driver = 
+{
+   .create_pipe_screen = NULL,
+   .create_pipe_context = NULL,
+   .display_surface = NULL,
+};
+
+#endif
diff --git a/src/gallium/winsys/xlib/xlib_softpipe.c b/src/gallium/winsys/xlib/xlib_softpipe.c
new file mode 100644
index 0000000000..586e1dfca5
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib_softpipe.c
@@ -0,0 +1,481 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/*
+ * Authors:
+ *   Keith Whitwell
+ *   Brian Paul
+ */
+
+
+#include "xm_api.h"
+
+#undef ASSERT
+#undef Elements
+
+#include "pipe/internal/p_winsys_screen.h"
+#include "pipe/p_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+#include "softpipe/sp_winsys.h"
+#include "softpipe/sp_texture.h"
+
+#include "xlib.h"
+
+/**
+ * Subclass of pipe_buffer for Xlib winsys.
+ * Low-level OS/window system memory buffer
+ */
+struct xm_buffer
+{
+   struct pipe_buffer base;
+   boolean userBuffer;  /** Is this a user-space buffer? */
+   void *data;
+   void *mapped;
+
+   XImage *tempImage;
+   int shm;
+   XShmSegmentInfo shminfo;
+};
+
+
+/**
+ * Subclass of pipe_winsys for Xlib winsys
+ */
+struct xmesa_pipe_winsys
+{
+   struct pipe_winsys base;
+/*   struct xmesa_visual *xm_visual; */
+   int shm;
+};
+
+
+
+/** Cast wrapper */
+static INLINE struct xm_buffer *
+xm_buffer( struct pipe_buffer *buf )
+{
+   return (struct xm_buffer *)buf;
+}
+
+
+/**
+ * X Shared Memory Image extension code
+ */
+#define XSHM_ENABLED(b) ((b)->shm)
+
+static volatile int mesaXErrorFlag = 0;
+
+/**
+ * Catches potential Xlib errors.
+ */
+static int
+mesaHandleXError(Display *dpy, XErrorEvent *event)
+{
+   (void) dpy;
+   (void) event;
+   mesaXErrorFlag = 1;
+   return 0;
+}
+
+
+static GLboolean alloc_shm(struct xm_buffer *buf, unsigned size)
+{
+   XShmSegmentInfo *const shminfo = & buf->shminfo;
+
+   shminfo->shmid = shmget(IPC_PRIVATE, size, IPC_CREAT|0777);
+   if (shminfo->shmid < 0) {
+      return GL_FALSE;
+   }
+
+   shminfo->shmaddr = (char *) shmat(shminfo->shmid, 0, 0);
+   if (shminfo->shmaddr == (char *) -1) {
+      shmctl(shminfo->shmid, IPC_RMID, 0);
+      return GL_FALSE;
+   }
+
+   shminfo->readOnly = False;
+   return GL_TRUE;
+}
+
+
+/**
+ * Allocate a shared memory XImage back buffer for the given XMesaBuffer.
+ */
+static void
+alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb,
+                 unsigned width, unsigned height)
+{
+   /*
+    * We have to do a _lot_ of error checking here to be sure we can
+    * really use the XSHM extension.  It seems different servers trigger
+    * errors at different points if the extension won't work.  Therefore
+    * we have to be very careful...
+    */
+   int (*old_handler)(Display *, XErrorEvent *);
+
+   b->tempImage = XShmCreateImage(xmb->xm_visual->display,
+                                  xmb->xm_visual->visinfo->visual,
+                                  xmb->xm_visual->visinfo->depth,
+                                  ZPixmap,
+                                  NULL,
+                                  &b->shminfo,
+                                  width, height);
+   if (b->tempImage == NULL) {
+      b->shm = 0;
+      return;
+   }
+
+
+   mesaXErrorFlag = 0;
+   old_handler = XSetErrorHandler(mesaHandleXError);
+   /* This may trigger the X protocol error we're ready to catch: */
+   XShmAttach(xmb->xm_visual->display, &b->shminfo);
+   XSync(xmb->xm_visual->display, False);
+
+   if (mesaXErrorFlag) {
+      /* we are on a remote display, this error is normal, don't print it */
+      XFlush(xmb->xm_visual->display);
+      mesaXErrorFlag = 0;
+      XDestroyImage(b->tempImage);
+      b->tempImage = NULL;
+      b->shm = 0;
+      (void) XSetErrorHandler(old_handler);
+      return;
+   }
+}
+
+
+
+/* Most callbacks map direcly onto dri_bufmgr operations:
+ */
+static void *
+xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer *buf,
+              unsigned flags)
+{
+   struct xm_buffer *xm_buf = xm_buffer(buf);
+   xm_buf->mapped = xm_buf->data;
+   return xm_buf->mapped;
+}
+
+static void
+xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf)
+{
+   struct xm_buffer *xm_buf = xm_buffer(buf);
+   xm_buf->mapped = NULL;
+}
+
+static void
+xm_buffer_destroy(struct pipe_winsys *pws,
+                  struct pipe_buffer *buf)
+{
+   struct xm_buffer *oldBuf = xm_buffer(buf);
+
+   if (oldBuf->data) {
+      if (oldBuf->shminfo.shmid >= 0) {
+         shmdt(oldBuf->shminfo.shmaddr);
+         shmctl(oldBuf->shminfo.shmid, IPC_RMID, 0);
+         
+         oldBuf->shminfo.shmid = -1;
+         oldBuf->shminfo.shmaddr = (char *) -1;
+      }
+      else
+      {
+         if (!oldBuf->userBuffer) {
+            align_free(oldBuf->data);
+         }
+      }
+
+      oldBuf->data = NULL;
+   }
+
+   free(oldBuf);
+}
+
+
+
+/**
+ * Display/copy the image in the surface into the X window specified
+ * by the XMesaBuffer.
+ */
+static void
+xlib_softpipe_display_surface(struct xmesa_buffer *b,
+                              struct pipe_surface *surf)
+{
+   XImage *ximage;
+   struct xm_buffer *xm_buf = xm_buffer(
+      softpipe_texture(surf->texture)->buffer);
+   static boolean no_swap = 0;
+   static boolean firsttime = 1;
+
+   if (firsttime) {
+      no_swap = getenv("SP_NO_RAST") != NULL;
+      firsttime = 0;
+   }
+
+   if (no_swap)
+      return;
+
+   if (XSHM_ENABLED(xm_buf) && (xm_buf->tempImage == NULL)) {
+      assert(surf->block.width == 1);
+      assert(surf->block.height == 1);
+      alloc_shm_ximage(xm_buf, b, surf->stride/surf->block.size, surf->height);
+   }
+
+   ximage = (XSHM_ENABLED(xm_buf)) ? xm_buf->tempImage : b->tempImage;
+   ximage->data = xm_buf->data;
+
+   /* display image in Window */
+   if (XSHM_ENABLED(xm_buf)) {
+      XShmPutImage(b->xm_visual->display, b->drawable, b->gc,
+                   ximage, 0, 0, 0, 0, surf->width, surf->height, False);
+   } else {
+      /* check that the XImage has been previously initialized */
+      assert(ximage->format);
+      assert(ximage->bitmap_unit);
+
+      /* update XImage's fields */
+      ximage->width = surf->width;
+      ximage->height = surf->height;
+      ximage->bytes_per_line = surf->stride;
+
+      XPutImage(b->xm_visual->display, b->drawable, b->gc,
+                ximage, 0, 0, 0, 0, surf->width, surf->height);
+   }
+}
+
+
+static void
+xm_flush_frontbuffer(struct pipe_winsys *pws,
+                     struct pipe_surface *surf,
+                     void *context_private)
+{
+   /*
+    * The front color buffer is actually just another XImage buffer.
+    * This function copies that XImage to the actual X Window.
+    */
+   XMesaContext xmctx = (XMesaContext) context_private;
+   xlib_softpipe_display_surface(xmctx->xm_buffer, surf);
+}
+
+
+
+static const char *
+xm_get_name(struct pipe_winsys *pws)
+{
+   return "Xlib";
+}
+
+
+static struct pipe_buffer *
+xm_buffer_create(struct pipe_winsys *pws, 
+                 unsigned alignment, 
+                 unsigned usage,
+                 unsigned size)
+{
+   struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer);
+   struct xmesa_pipe_winsys *xpws = (struct xmesa_pipe_winsys *) pws;
+
+   buffer->base.refcount = 1;
+   buffer->base.alignment = alignment;
+   buffer->base.usage = usage;
+   buffer->base.size = size;
+   buffer->shminfo.shmid = -1;
+   buffer->shminfo.shmaddr = (char *) -1;
+
+   if (xpws->shm && (usage & PIPE_BUFFER_USAGE_PIXEL) != 0) {
+      buffer->shm = xpws->shm;
+
+      if (alloc_shm(buffer, size)) {
+         buffer->data = buffer->shminfo.shmaddr;
+      }
+   }
+
+   if (buffer->data == NULL) {
+      buffer->shm = 0;
+
+      /* align to 16-byte multiple for Cell */
+      buffer->data = align_malloc(size, max(alignment, 16));
+   }
+
+   return &buffer->base;
+}
+
+
+/**
+ * Create buffer which wraps user-space data.
+ */
+static struct pipe_buffer *
+xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes)
+{
+   struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer);
+   buffer->base.refcount = 1;
+   buffer->base.size = bytes;
+   buffer->userBuffer = TRUE;
+   buffer->data = ptr;
+   buffer->shm = 0;
+
+   return &buffer->base;
+}
+
+
+static struct pipe_buffer *
+xm_surface_buffer_create(struct pipe_winsys *winsys,
+                         unsigned width, unsigned height,
+                         enum pipe_format format,
+                         unsigned usage,
+                         unsigned *stride)
+{
+   const unsigned alignment = 64;
+   struct pipe_format_block block;
+   unsigned nblocksx, nblocksy;
+
+   pf_get_block(format, &block);
+   nblocksx = pf_get_nblocksx(&block, width);
+   nblocksy = pf_get_nblocksy(&block, height);
+   *stride = align(nblocksx * block.size, alignment);
+
+   return winsys->buffer_create(winsys, alignment,
+                                usage,
+                                *stride * nblocksy);
+}
+
+
+/*
+ * Fence functions - basically nothing to do, as we don't create any actual
+ * fence objects.
+ */
+
+static void
+xm_fence_reference(struct pipe_winsys *sws, struct pipe_fence_handle **ptr,
+                   struct pipe_fence_handle *fence)
+{
+}
+
+
+static int
+xm_fence_signalled(struct pipe_winsys *sws, struct pipe_fence_handle *fence,
+                   unsigned flag)
+{
+   return 0;
+}
+
+
+static int
+xm_fence_finish(struct pipe_winsys *sws, struct pipe_fence_handle *fence,
+                unsigned flag)
+{
+   return 0;
+}
+
+
+
+static struct pipe_winsys *
+xlib_create_softpipe_winsys( void )
+{
+   static struct xmesa_pipe_winsys *ws = NULL;
+
+   if (!ws) {
+      ws = CALLOC_STRUCT(xmesa_pipe_winsys);
+
+      /* Fill in this struct with callbacks that pipe will need to
+       * communicate with the window system, buffer manager, etc. 
+       */
+      ws->base.buffer_create = xm_buffer_create;
+      ws->base.user_buffer_create = xm_user_buffer_create;
+      ws->base.buffer_map = xm_buffer_map;
+      ws->base.buffer_unmap = xm_buffer_unmap;
+      ws->base.buffer_destroy = xm_buffer_destroy;
+
+      ws->base.surface_buffer_create = xm_surface_buffer_create;
+
+      ws->base.fence_reference = xm_fence_reference;
+      ws->base.fence_signalled = xm_fence_signalled;
+      ws->base.fence_finish = xm_fence_finish;
+
+      ws->base.flush_frontbuffer = xm_flush_frontbuffer;
+      ws->base.get_name = xm_get_name;
+   }
+
+   return &ws->base;
+}
+
+
+static struct pipe_screen *
+xlib_create_softpipe_screen( void )
+{
+   struct pipe_winsys *winsys;
+   struct pipe_screen *screen;
+
+   winsys = xlib_create_softpipe_winsys();
+   if (winsys == NULL)
+      return NULL;
+
+   screen = softpipe_create_screen(winsys);
+   if (screen == NULL)
+      goto fail;
+
+   return screen;
+
+fail:
+   if (winsys)
+      winsys->destroy( winsys );
+
+   return NULL;
+}
+
+
+static struct pipe_context *
+xlib_create_softpipe_context( struct pipe_screen *screen,
+                              void *context_private )
+{
+   struct pipe_context *pipe;
+   
+   pipe = softpipe_create(screen, screen->winsys, NULL);
+   if (pipe == NULL)
+      goto fail;
+
+   pipe->priv = context_private;
+   return pipe;
+
+fail:
+   /* Free stuff here */
+   return NULL;
+}
+
+struct xm_driver xlib_softpipe_driver = 
+{
+   .create_pipe_screen = xlib_create_softpipe_screen,
+   .create_pipe_context = xlib_create_softpipe_context,
+   .display_surface = xlib_softpipe_display_surface
+};
+
+
+
diff --git a/src/gallium/winsys/xlib/xlib_trace.c b/src/gallium/winsys/xlib/xlib_trace.c
new file mode 100644
index 0000000000..37095c5d8e
--- /dev/null
+++ b/src/gallium/winsys/xlib/xlib_trace.c
@@ -0,0 +1,107 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR 
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE 
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * 
+ **************************************************************************/
+
+/*
+ * Authors:
+ *   Keith Whitwell
+ *   Brian Paul
+ */
+
+
+#include "xlib.h"
+
+#include "trace/tr_screen.h"
+#include "trace/tr_context.h"
+
+#include "pipe/p_screen.h"
+
+
+
+static struct pipe_screen *
+xlib_create_trace_screen( void )
+{
+   struct pipe_screen *screen, *trace_screen;
+
+   screen = xlib_softpipe_driver.create_pipe_screen();
+   if (screen == NULL)
+      goto fail;
+
+   /* Wrap it:
+    */
+   trace_screen = trace_screen_create(screen);
+   if (trace_screen == NULL)
+      goto fail;
+
+   return trace_screen;
+
+fail:
+   if (screen)
+      screen->destroy( screen );
+   return NULL;
+}
+
+static struct pipe_context *
+xlib_create_trace_context( struct pipe_screen *screen,
+                           void *priv )
+{
+   struct pipe_context *pipe, *trace_pipe;
+   
+   pipe = xlib_softpipe_driver.create_pipe_context( screen, priv );
+   if (pipe == NULL)
+      goto fail;
+
+   /* Wrap it:
+    */
+   trace_pipe = trace_context_create(screen, pipe);
+   if (trace_pipe == NULL)
+      goto fail;
+
+   trace_pipe->priv = priv;
+
+   return trace_pipe;
+
+fail:
+   return NULL;
+}
+
+static void
+xlib_trace_display_surface( struct xmesa_buffer *buffer,
+                            struct pipe_surface *surf )
+{
+   /* ??
+    */
+   xlib_softpipe_driver.display_surface( buffer, surf );
+}
+
+
+struct xm_driver xlib_trace_driver = 
+{
+   .create_pipe_screen = xlib_create_trace_screen,
+   .create_pipe_context = xlib_create_trace_context,
+   .display_surface = xlib_trace_display_surface,
+};
diff --git a/src/gallium/winsys/xlib/xmesa.h b/src/gallium/winsys/xlib/xmesa.h
new file mode 100644
index 0000000000..98139af833
--- /dev/null
+++ b/src/gallium/winsys/xlib/xmesa.h
@@ -0,0 +1,424 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ * 
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ * Mesa/X11 interface.  This header file serves as the documentation for
+ * the Mesa/X11 interface functions.
+ *
+ * Note: this interface isn't intended for user programs.  It's primarily
+ * just for implementing the pseudo-GLX interface.
+ */
+
+
+/* Sample Usage:
+
+In addition to the usual X calls to select a visual, create a colormap
+and create a window, you must do the following to use the X/Mesa interface:
+
+1. Call XMesaCreateVisual() to make an XMesaVisual from an XVisualInfo.
+
+2. Call XMesaCreateContext() to create an X/Mesa rendering context, given
+   the XMesaVisual.
+
+3. Call XMesaCreateWindowBuffer() to create an XMesaBuffer from an X window
+   and XMesaVisual.
+
+4. Call XMesaMakeCurrent() to bind the XMesaBuffer to an XMesaContext and
+   to make the context the current one.
+
+5. Make gl* calls to render your graphics.
+
+6. Use XMesaSwapBuffers() when double buffering to swap front/back buffers.
+
+7. Before the X window is destroyed, call XMesaDestroyBuffer().
+
+8. Before exiting, call XMesaDestroyVisual and XMesaDestroyContext.
+
+*/
+
+
+
+
+#ifndef XMESA_H
+#define XMESA_H
+
+#ifdef __VMS
+#include <GL/vms_x_fix.h>
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef XFree86Server
+#include "xmesa_xf86.h"
+#else
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+#include "xmesa_x.h"
+#endif
+#include "GL/gl.h"
+
+#ifdef AMIWIN
+#include <pragmas/xlib_pragmas.h>
+extern struct Library *XLibBase;
+#endif
+
+
+#define XMESA_MAJOR_VERSION 6
+#define XMESA_MINOR_VERSION 3
+
+
+
+/*
+ * Values passed to XMesaGetString:
+ */
+#define XMESA_VERSION 1
+#define XMESA_EXTENSIONS 2
+
+
+/*
+ * Values passed to XMesaSetFXmode:
+ */
+#define XMESA_FX_WINDOW       1
+#define XMESA_FX_FULLSCREEN   2
+
+
+
+typedef struct xmesa_context *XMesaContext;
+
+typedef struct xmesa_visual *XMesaVisual;
+
+typedef struct xmesa_buffer *XMesaBuffer;
+
+
+
+/*
+ * Create a new X/Mesa visual.
+ * Input:  display - X11 display
+ *         visinfo - an XVisualInfo pointer
+ *         rgb_flag - GL_TRUE = RGB mode,
+ *                    GL_FALSE = color index mode
+ *         alpha_flag - alpha buffer requested?
+ *         db_flag - GL_TRUE = double-buffered,
+ *                   GL_FALSE = single buffered
+ *         stereo_flag - stereo visual?
+ *         ximage_flag - GL_TRUE = use an XImage for back buffer,
+ *                       GL_FALSE = use an off-screen pixmap for back buffer
+ *         depth_size - requested bits/depth values, or zero
+ *         stencil_size - requested bits/stencil values, or zero
+ *         accum_red_size - requested bits/red accum values, or zero
+ *         accum_green_size - requested bits/green accum values, or zero
+ *         accum_blue_size - requested bits/blue accum values, or zero
+ *         accum_alpha_size - requested bits/alpha accum values, or zero
+ *         num_samples - number of samples/pixel if multisampling, or zero
+ *         level - visual level, usually 0
+ *         visualCaveat - ala the GLX extension, usually GLX_NONE_EXT
+ * Return;  a new XMesaVisual or 0 if error.
+ */
+extern XMesaVisual XMesaCreateVisual( XMesaDisplay *display,
+                                      XMesaVisualInfo visinfo,
+                                      GLboolean rgb_flag,
+                                      GLboolean alpha_flag,
+                                      GLboolean db_flag,
+                                      GLboolean stereo_flag,
+                                      GLboolean ximage_flag,
+                                      GLint depth_size,
+                                      GLint stencil_size,
+                                      GLint accum_red_size,
+                                      GLint accum_green_size,
+                                      GLint accum_blue_size,
+                                      GLint accum_alpha_size,
+                                      GLint num_samples,
+                                      GLint level,
+                                      GLint visualCaveat );
+
+/*
+ * Destroy an XMesaVisual, but not the associated XVisualInfo.
+ */
+extern void XMesaDestroyVisual( XMesaVisual v );
+
+
+
+/*
+ * Create a new XMesaContext for rendering into an X11 window.
+ *
+ * Input:  visual - an XMesaVisual
+ *         share_list - another XMesaContext with which to share display
+ *                      lists or NULL if no sharing is wanted.
+ * Return:  an XMesaContext or NULL if error.
+ */
+extern XMesaContext XMesaCreateContext( XMesaVisual v,
+					XMesaContext share_list );
+
+
+/*
+ * Destroy a rendering context as returned by XMesaCreateContext()
+ */
+extern void XMesaDestroyContext( XMesaContext c );
+
+
+#ifdef XFree86Server
+/*
+ * These are the extra routines required for integration with XFree86.
+ * None of these routines should be user visible. -KEM
+ */
+extern GLboolean XMesaForceCurrent( XMesaContext c );
+
+extern GLboolean XMesaLoseCurrent( XMesaContext c );
+
+extern GLboolean XMesaCopyContext( XMesaContext src,
+				   XMesaContext dst,
+				   GLuint mask );
+#endif /* XFree86Server */
+
+
+/*
+ * Create an XMesaBuffer from an X window.
+ */
+extern XMesaBuffer XMesaCreateWindowBuffer( XMesaVisual v, XMesaWindow w );
+
+
+/*
+ * Create an XMesaBuffer from an X pixmap.
+ */
+extern XMesaBuffer XMesaCreatePixmapBuffer( XMesaVisual v,
+					    XMesaPixmap p,
+					    XMesaColormap cmap );
+
+
+/*
+ * Destroy an XMesaBuffer, but not the corresponding window or pixmap.
+ */
+extern void XMesaDestroyBuffer( XMesaBuffer b );
+
+
+/*
+ * Return the XMesaBuffer handle which corresponds to an X drawable, if any.
+ *
+ * New in Mesa 2.3.
+ */
+extern XMesaBuffer XMesaFindBuffer( XMesaDisplay *dpy,
+				    XMesaDrawable d );
+
+
+
+/*
+ * Bind a buffer to a context and make the context the current one.
+ */
+extern GLboolean XMesaMakeCurrent( XMesaContext c,
+				   XMesaBuffer b );
+
+
+/*
+ * Bind two buffers (read and draw) to a context and make the
+ * context the current one.
+ * New in Mesa 3.3
+ */
+extern GLboolean XMesaMakeCurrent2( XMesaContext c,
+                                    XMesaBuffer drawBuffer,
+                                    XMesaBuffer readBuffer );
+
+
+/*
+ * Unbind the current context from its buffer.
+ */
+extern GLboolean XMesaUnbindContext( XMesaContext c );
+
+
+/*
+ * Return a handle to the current context.
+ */
+extern XMesaContext XMesaGetCurrentContext( void );
+
+
+/*
+ * Return handle to the current (draw) buffer.
+ */
+extern XMesaBuffer XMesaGetCurrentBuffer( void );
+
+
+/*
+ * Return handle to the current read buffer.
+ * New in Mesa 3.3
+ */
+extern XMesaBuffer XMesaGetCurrentReadBuffer( void );
+
+
+/*
+ * Swap the front and back buffers for the given buffer.  No action is
+ * taken if the buffer is not double buffered.
+ */
+extern void XMesaSwapBuffers( XMesaBuffer b );
+
+
+/*
+ * Copy a sub-region of the back buffer to the front buffer.
+ *
+ * New in Mesa 2.6
+ */
+extern void XMesaCopySubBuffer( XMesaBuffer b,
+				int x,
+				int y,
+				int width,
+				int height );
+
+
+/*
+ * Return a pointer to the the Pixmap or XImage being used as the back
+ * color buffer of an XMesaBuffer.  This function is a way to get "under
+ * the hood" of X/Mesa so one can manipulate the back buffer directly.
+ * Input:  b - the XMesaBuffer
+ * Output:  pixmap - pointer to back buffer's Pixmap, or 0
+ *          ximage - pointer to back buffer's XImage, or NULL
+ * Return:  GL_TRUE = context is double buffered
+ *          GL_FALSE = context is single buffered
+ */
+extern GLboolean XMesaGetBackBuffer( XMesaBuffer b,
+				     XMesaPixmap *pixmap,
+				     XMesaImage **ximage );
+
+
+
+/*
+ * Return the depth buffer associated with an XMesaBuffer.
+ * Input:  b - the XMesa buffer handle
+ * Output:  width, height - size of buffer in pixels
+ *          bytesPerValue - bytes per depth value (2 or 4)
+ *          buffer - pointer to depth buffer values
+ * Return:  GL_TRUE or GL_FALSE to indicate success or failure.
+ *
+ * New in Mesa 2.4.
+ */
+extern GLboolean XMesaGetDepthBuffer( XMesaBuffer b,
+				      GLint *width,
+				      GLint *height,
+				      GLint *bytesPerValue,
+				      void **buffer );
+
+
+
+/*
+ * Flush/sync a context
+ */
+extern void XMesaFlush( XMesaContext c );
+
+
+
+/*
+ * Get an X/Mesa-specific string.
+ * Input:  name - either XMESA_VERSION or XMESA_EXTENSIONS
+ */
+extern const char *XMesaGetString( XMesaContext c, int name );
+
+
+
+/*
+ * Scan for XMesaBuffers whose window/pixmap has been destroyed, then free
+ * any memory used by that buffer.
+ *
+ * New in Mesa 2.3.
+ */
+extern void XMesaGarbageCollect( void );
+
+
+
+/*
+ * Return a dithered pixel value.
+ * Input:  c - XMesaContext
+ *         x, y - window coordinate
+ *         red, green, blue, alpha - color components in [0,1]
+ * Return:  pixel value
+ *
+ * New in Mesa 2.3.
+ */
+extern unsigned long XMesaDitherColor( XMesaContext xmesa,
+				       GLint x,
+				       GLint y,
+				       GLfloat red,
+				       GLfloat green,
+				       GLfloat blue,
+				       GLfloat alpha );
+
+
+
+/*
+ * 3Dfx Glide driver only!
+ * Set 3Dfx/Glide full-screen or window rendering mode.
+ * Input:  mode - either XMESA_FX_WINDOW (window rendering mode) or
+ *                XMESA_FX_FULLSCREEN (full-screen rendering mode)
+ * Return:  GL_TRUE if success
+ *          GL_FALSE if invalid mode or if not using 3Dfx driver
+ *
+ * New in Mesa 2.6.
+ */
+extern GLboolean XMesaSetFXmode( GLint mode );
+
+
+
+/*
+ * Reallocate the back/depth/stencil/accum/etc/ buffers associated with
+ * buffer <b> if its size has changed.
+ *
+ * New in Mesa 4.0.2
+ */
+extern void XMesaResizeBuffers( XMesaBuffer b );
+
+
+
+/*
+ * Create a pbuffer.
+ * New in Mesa 4.1
+ */
+extern XMesaBuffer XMesaCreatePBuffer(XMesaVisual v, XMesaColormap cmap,
+                                      unsigned int width, unsigned int height);
+
+
+
+/*
+ * Texture from Pixmap
+ * New in Mesa 7.1
+ */
+extern void
+XMesaBindTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer,
+                  const int *attrib_list);
+
+extern void
+XMesaReleaseTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer);
+
+
+extern XMesaBuffer
+XMesaCreatePixmapTextureBuffer(XMesaVisual v, XMesaPixmap p,
+                               XMesaColormap cmap,
+                               int format, int target, int mipmap);
+
+
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif
diff --git a/src/gallium/winsys/xlib/xmesa_x.h b/src/gallium/winsys/xlib/xmesa_x.h
new file mode 100644
index 0000000000..865bab4313
--- /dev/null
+++ b/src/gallium/winsys/xlib/xmesa_x.h
@@ -0,0 +1,86 @@
+
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ * When we're building the XMesa driver for stand-alone Mesa we
+ * include this file when building the xm_*.c files.
+ * We need to define some types and macros differently when building
+ * in the Xserver vs. stand-alone Mesa.
+ */
+
+#ifndef _XMESA_X_H_
+#define _XMESA_X_H_
+
+typedef Display      XMesaDisplay;
+typedef Pixmap       XMesaPixmap;
+typedef Colormap     XMesaColormap;
+typedef Drawable     XMesaDrawable;
+typedef Window       XMesaWindow;
+typedef GC           XMesaGC;
+typedef XVisualInfo *XMesaVisualInfo;
+typedef XImage       XMesaImage;
+typedef XPoint       XMesaPoint;
+typedef XColor       XMesaColor;
+
+#define XMesaDestroyImage      XDestroyImage
+
+#define XMesaPutPixel          XPutPixel
+#define XMesaGetPixel          XGetPixel
+
+#define XMesaSetForeground     XSetForeground
+#define XMesaSetBackground     XSetBackground
+#define XMesaSetPlaneMask      XSetPlaneMask
+#define XMesaSetFunction       XSetFunction
+#define XMesaSetFillStyle      XSetFillStyle
+#define XMesaSetTile           XSetTile
+
+#define XMesaDrawPoint         XDrawPoint
+#define XMesaDrawPoints        XDrawPoints
+#define XMesaDrawLine          XDrawLine
+#define XMesaFillRectangle     XFillRectangle
+#define XMesaGetImage          XGetImage
+#define XMesaPutImage          XPutImage
+#define XMesaCopyArea          XCopyArea
+
+#define XMesaCreatePixmap      XCreatePixmap
+#define XMesaFreePixmap        XFreePixmap
+#define XMesaFreeGC            XFreeGC
+
+#define GET_COLORMAP_SIZE(__v)  __v->visinfo->colormap_size
+#define GET_REDMASK(__v)        __v->mesa_visual.redMask
+#define GET_GREENMASK(__v)      __v->mesa_visual.greenMask
+#define GET_BLUEMASK(__v)       __v->mesa_visual.blueMask
+#define GET_VISUAL_DEPTH(__v)   __v->visinfo->depth
+#define GET_BLACK_PIXEL(__v)    BlackPixel(__v->display, __v->mesa_visual.screen)
+#define CHECK_BYTE_ORDER(__v)   host_byte_order()==ImageByteOrder(__v->display)
+#define CHECK_FOR_HPCR(__v)     XInternAtom(__v->display, "_HP_RGB_SMOOTH_MAP_LIST", True)
+
+#endif
diff --git a/src/glew/LICENSE.txt b/src/glew/LICENSE.txt
new file mode 100644
index 0000000000..f7078042e9
--- /dev/null
+++ b/src/glew/LICENSE.txt
@@ -0,0 +1,73 @@
+The OpenGL Extension Wrangler Library
+Copyright (C) 2002-2007, Milan Ikits <milan ikits[]ieee org>
+Copyright (C) 2002-2007, Marcelo E. Magallon <mmagallo[]debian org>
+Copyright (C) 2002, Lev Povalahev
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without 
+modification, are permitted provided that the following conditions are met:
+
+* Redistributions of source code must retain the above copyright notice, 
+  this list of conditions and the following disclaimer.
+* Redistributions in binary form must reproduce the above copyright notice, 
+  this list of conditions and the following disclaimer in the documentation 
+  and/or other materials provided with the distribution.
+* The name of the author may be used to endorse or promote products 
+  derived from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 
+CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 
+SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+THE POSSIBILITY OF SUCH DAMAGE.
+
+
+Mesa 3-D graphics library
+Version:  7.0
+
+Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+the rights to use, copy, modify, merge, publish, distribute, sublicense,
+and/or sell copies of the Software, and to permit persons to whom the
+Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included
+in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+
+Copyright (c) 2007 The Khronos Group Inc.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and/or associated documentation files (the
+"Materials"), to deal in the Materials without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Materials, and to
+permit persons to whom the Materials are furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice shall be included
+in all copies or substantial portions of the Materials.
+
+THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
diff --git a/src/glew/Makefile b/src/glew/Makefile
new file mode 100644
index 0000000000..49ee8b6594
--- /dev/null
+++ b/src/glew/Makefile
@@ -0,0 +1,51 @@
+# src/glew/Makefile
+
+TOP = ../..
+include $(TOP)/configs/current
+
+
+##### MACROS #####
+
+GLEW_LIB_NAME = libglew.a
+
+MAJOR = 1
+MINOR = 5
+TINY = 0
+
+SOURCES = \
+	glew.c
+
+
+OBJECTS = $(SOURCES:.c=.o)
+
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c -I$(TOP)/include $(X11_INCLUDES) $(CFLAGS) $<
+
+
+
+##### TARGETS #####
+
+default: $(TOP)/$(LIB_DIR)/$(GLEW_LIB_NAME)
+
+
+
+# Make the library
+$(TOP)/$(LIB_DIR)/$(GLEW_LIB_NAME): depend $(OBJECTS)
+	$(MKLIB) -o $(GLEW_LIB) -static -install $(TOP)/$(LIB_DIR) $(OBJECTS)
+
+
+clean:
+	-rm -f *.o *~
+	-rm -f depend depend.bak
+
+
+depend: $(SOURCES)
+	@ rm -f depend
+	@ touch depend
+	@ $(MKDEP) $(MKDEP_OPTIONS) -I$(TOP)/include $(SOURCES) \
+		$(X11_INCLUDES) > /dev/null
+
+include depend
diff --git a/src/glew/SConscript b/src/glew/SConscript
new file mode 100644
index 0000000000..1161be6e63
--- /dev/null
+++ b/src/glew/SConscript
@@ -0,0 +1,50 @@
+Import('*')
+
+if env['platform'] not in ['windows', 'linux']:
+    Return()
+
+env = env.Clone()
+
+env.Append(CPPDEFINES = [
+    'GLEW_BUILD',
+    'GLEW_STATIC',
+    #'GLEW_MX', # Multiple Rendering Contexts support
+])
+
+env.PrependUnique(CPPPATH = [
+    '#/include',
+])
+
+glew = env.StaticLibrary(
+    target = 'glew',
+    source = [
+        'glew.c',
+    ],
+)
+
+env = env.Clone()
+
+if env['platform'] == 'windows':
+    env.PrependUnique(LIBS = [
+        'glu32', 
+        'opengl32', 
+        'gdi32', 
+        'user32', 
+    ])
+else:
+    env.PrependUnique(LIBS = [
+        'GLU',
+        'GL',
+        'X11',
+    ])
+env.Prepend(LIBS = [glew])
+
+env.Program(
+    target = 'glewinfo',
+    source = ['glewinfo.c'],
+)
+
+env.Program(
+    target = 'visualinfo',
+    source = ['visualinfo.c'],
+)
diff --git a/src/glew/glew.c b/src/glew/glew.c
new file mode 100644
index 0000000000..24c6a726c3
--- /dev/null
+++ b/src/glew/glew.c
@@ -0,0 +1,12180 @@
+/*
+** The OpenGL Extension Wrangler Library
+** Copyright (C) 2002-2008, Milan Ikits <milan ikits[]ieee org>
+** Copyright (C) 2002-2008, Marcelo E. Magallon <mmagallo[]debian org>
+** Copyright (C) 2002, Lev Povalahev
+** All rights reserved.
+** 
+** Redistribution and use in source and binary forms, with or without 
+** modification, are permitted provided that the following conditions are met:
+** 
+** * Redistributions of source code must retain the above copyright notice, 
+**   this list of conditions and the following disclaimer.
+** * Redistributions in binary form must reproduce the above copyright notice, 
+**   this list of conditions and the following disclaimer in the documentation 
+**   and/or other materials provided with the distribution.
+** * The name of the author may be used to endorse or promote products 
+**   derived from this software without specific prior written permission.
+**
+** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
+** AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
+** IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+** ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 
+** LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 
+** CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 
+** SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+** INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+** CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+** ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+** THE POSSIBILITY OF SUCH DAMAGE.
+*/
+
+#include <GL/glew.h>
+#if defined(_WIN32)
+#  include <GL/wglew.h>
+#elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+#  include <GL/glxew.h>
+#endif
+
+/*
+ * Define glewGetContext and related helper macros.
+ */
+#ifdef GLEW_MX
+#  define glewGetContext() ctx
+#  ifdef _WIN32
+#    define GLEW_CONTEXT_ARG_DEF_INIT GLEWContext* ctx
+#    define GLEW_CONTEXT_ARG_VAR_INIT ctx
+#    define wglewGetContext() ctx
+#    define WGLEW_CONTEXT_ARG_DEF_INIT WGLEWContext* ctx
+#    define WGLEW_CONTEXT_ARG_DEF_LIST WGLEWContext* ctx
+#  else /* _WIN32 */
+#    define GLEW_CONTEXT_ARG_DEF_INIT void
+#    define GLEW_CONTEXT_ARG_VAR_INIT
+#    define glxewGetContext() ctx
+#    define GLXEW_CONTEXT_ARG_DEF_INIT void
+#    define GLXEW_CONTEXT_ARG_DEF_LIST GLXEWContext* ctx
+#  endif /* _WIN32 */
+#  define GLEW_CONTEXT_ARG_DEF_LIST GLEWContext* ctx
+#else /* GLEW_MX */
+#  define GLEW_CONTEXT_ARG_DEF_INIT void
+#  define GLEW_CONTEXT_ARG_VAR_INIT
+#  define GLEW_CONTEXT_ARG_DEF_LIST void
+#  define WGLEW_CONTEXT_ARG_DEF_INIT void
+#  define WGLEW_CONTEXT_ARG_DEF_LIST void
+#  define GLXEW_CONTEXT_ARG_DEF_INIT void
+#  define GLXEW_CONTEXT_ARG_DEF_LIST void
+#endif /* GLEW_MX */
+
+#if defined(__APPLE__)
+#include <mach-o/dyld.h>
+#include <stdlib.h>
+#include <string.h>
+
+void* NSGLGetProcAddress (const GLubyte *name)
+{
+  static const struct mach_header* image = NULL;
+  NSSymbol symbol;
+  char* symbolName;
+  if (NULL == image)
+  {
+    image = NSAddImage("/System/Library/Frameworks/OpenGL.framework/Versions/Current/OpenGL", NSADDIMAGE_OPTION_RETURN_ON_ERROR);
+  }
+  /* prepend a '_' for the Unix C symbol mangling convention */
+  symbolName = malloc(strlen((const char*)name) + 2);
+  strcpy(symbolName+1, (const char*)name);
+  symbolName[0] = '_';
+  symbol = NULL;
+  /* if (NSIsSymbolNameDefined(symbolName))
+	 symbol = NSLookupAndBindSymbol(symbolName); */
+  symbol = image ? NSLookupSymbolInImage(image, symbolName, NSLOOKUPSYMBOLINIMAGE_OPTION_BIND | NSLOOKUPSYMBOLINIMAGE_OPTION_RETURN_ON_ERROR) : NULL;
+  free(symbolName);
+  return symbol ? NSAddressOfSymbol(symbol) : NULL;
+}
+#endif /* __APPLE__ */
+
+#if defined(__sgi) || defined (__sun)
+#include <dlfcn.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+void* dlGetProcAddress (const GLubyte* name)
+{
+  static void* h = NULL;
+  static void* gpa;
+
+  if (h == NULL)
+  {
+    if ((h = dlopen(NULL, RTLD_LAZY | RTLD_LOCAL)) == NULL) return NULL;
+    gpa = dlsym(h, "glXGetProcAddress");
+  }
+
+  if (gpa != NULL)
+    return ((void*(*)(const GLubyte*))gpa)(name);
+  else
+    return dlsym(h, (const char*)name);
+}
+#endif /* __sgi || __sun */
+
+/*
+ * Define glewGetProcAddress.
+ */
+#if defined(_WIN32)
+#  define glewGetProcAddress(name) wglGetProcAddress((LPCSTR)name)
+#else
+#  if defined(__APPLE__)
+#    define glewGetProcAddress(name) NSGLGetProcAddress(name)
+#  else
+#    if defined(__sgi) || defined(__sun)
+#      define glewGetProcAddress(name) dlGetProcAddress(name)
+#    else /* __linux */
+#      define glewGetProcAddress(name) (*glXGetProcAddressARB)(name)
+#    endif
+#  endif
+#endif
+
+/*
+ * Define GLboolean const cast.
+ */
+#define CONST_CAST(x) (*(GLboolean*)&x)
+
+/*
+ * GLEW, just like OpenGL or GLU, does not rely on the standard C library.
+ * These functions implement the functionality required in this file.
+ */
+static GLuint _glewStrLen (const GLubyte* s)
+{
+  GLuint i=0;
+  if (s == NULL) return 0;
+  while (s[i] != '\0') i++;
+  return i;
+}
+
+static GLuint _glewStrCLen (const GLubyte* s, GLubyte c)
+{
+  GLuint i=0;
+  if (s == NULL) return 0;
+  while (s[i] != '\0' && s[i] != c) i++;
+  return (s[i] == '\0' || s[i] == c) ? i : 0;
+}
+
+static GLboolean _glewStrSame (const GLubyte* a, const GLubyte* b, GLuint n)
+{
+  GLuint i=0;
+  if(a == NULL || b == NULL)
+    return (a == NULL && b == NULL && n == 0) ? GL_TRUE : GL_FALSE;
+  while (i < n && a[i] != '\0' && b[i] != '\0' && a[i] == b[i]) i++;
+  return i == n ? GL_TRUE : GL_FALSE;
+}
+
+static GLboolean _glewStrSame1 (GLubyte** a, GLuint* na, const GLubyte* b, GLuint nb)
+{
+  while (*na > 0 && (**a == ' ' || **a == '\n' || **a == '\r' || **a == '\t'))
+  {
+    (*a)++;
+    (*na)--;
+  }
+  if(*na >= nb)
+  {
+    GLuint i=0;
+    while (i < nb && (*a)+i != NULL && b+i != NULL && (*a)[i] == b[i]) i++;
+	if(i == nb)
+	{
+		*a = *a + nb;
+		*na = *na - nb;
+		return GL_TRUE;
+	}
+  }
+  return GL_FALSE;
+}
+
+static GLboolean _glewStrSame2 (GLubyte** a, GLuint* na, const GLubyte* b, GLuint nb)
+{
+  if(*na >= nb)
+  {
+    GLuint i=0;
+    while (i < nb && (*a)+i != NULL && b+i != NULL && (*a)[i] == b[i]) i++;
+	if(i == nb)
+	{
+		*a = *a + nb;
+		*na = *na - nb;
+		return GL_TRUE;
+	}
+  }
+  return GL_FALSE;
+}
+
+static GLboolean _glewStrSame3 (GLubyte** a, GLuint* na, const GLubyte* b, GLuint nb)
+{
+  if(*na >= nb)
+  {
+    GLuint i=0;
+    while (i < nb && (*a)+i != NULL && b+i != NULL && (*a)[i] == b[i]) i++;
+    if (i == nb && (*na == nb || (*a)[i] == ' ' || (*a)[i] == '\n' || (*a)[i] == '\r' || (*a)[i] == '\t'))
+    {
+      *a = *a + nb;
+      *na = *na - nb;
+      return GL_TRUE;
+    }
+  }
+  return GL_FALSE;
+}
+
+#if !defined(_WIN32) || !defined(GLEW_MX)
+
+PFNGLCOPYTEXSUBIMAGE3DPROC __glewCopyTexSubImage3D = NULL;
+PFNGLDRAWRANGEELEMENTSPROC __glewDrawRangeElements = NULL;
+PFNGLTEXIMAGE3DPROC __glewTexImage3D = NULL;
+PFNGLTEXSUBIMAGE3DPROC __glewTexSubImage3D = NULL;
+
+PFNGLACTIVETEXTUREPROC __glewActiveTexture = NULL;
+PFNGLCLIENTACTIVETEXTUREPROC __glewClientActiveTexture = NULL;
+PFNGLCOMPRESSEDTEXIMAGE1DPROC __glewCompressedTexImage1D = NULL;
+PFNGLCOMPRESSEDTEXIMAGE2DPROC __glewCompressedTexImage2D = NULL;
+PFNGLCOMPRESSEDTEXIMAGE3DPROC __glewCompressedTexImage3D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC __glewCompressedTexSubImage1D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC __glewCompressedTexSubImage2D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC __glewCompressedTexSubImage3D = NULL;
+PFNGLGETCOMPRESSEDTEXIMAGEPROC __glewGetCompressedTexImage = NULL;
+PFNGLLOADTRANSPOSEMATRIXDPROC __glewLoadTransposeMatrixd = NULL;
+PFNGLLOADTRANSPOSEMATRIXFPROC __glewLoadTransposeMatrixf = NULL;
+PFNGLMULTTRANSPOSEMATRIXDPROC __glewMultTransposeMatrixd = NULL;
+PFNGLMULTTRANSPOSEMATRIXFPROC __glewMultTransposeMatrixf = NULL;
+PFNGLMULTITEXCOORD1DPROC __glewMultiTexCoord1d = NULL;
+PFNGLMULTITEXCOORD1DVPROC __glewMultiTexCoord1dv = NULL;
+PFNGLMULTITEXCOORD1FPROC __glewMultiTexCoord1f = NULL;
+PFNGLMULTITEXCOORD1FVPROC __glewMultiTexCoord1fv = NULL;
+PFNGLMULTITEXCOORD1IPROC __glewMultiTexCoord1i = NULL;
+PFNGLMULTITEXCOORD1IVPROC __glewMultiTexCoord1iv = NULL;
+PFNGLMULTITEXCOORD1SPROC __glewMultiTexCoord1s = NULL;
+PFNGLMULTITEXCOORD1SVPROC __glewMultiTexCoord1sv = NULL;
+PFNGLMULTITEXCOORD2DPROC __glewMultiTexCoord2d = NULL;
+PFNGLMULTITEXCOORD2DVPROC __glewMultiTexCoord2dv = NULL;
+PFNGLMULTITEXCOORD2FPROC __glewMultiTexCoord2f = NULL;
+PFNGLMULTITEXCOORD2FVPROC __glewMultiTexCoord2fv = NULL;
+PFNGLMULTITEXCOORD2IPROC __glewMultiTexCoord2i = NULL;
+PFNGLMULTITEXCOORD2IVPROC __glewMultiTexCoord2iv = NULL;
+PFNGLMULTITEXCOORD2SPROC __glewMultiTexCoord2s = NULL;
+PFNGLMULTITEXCOORD2SVPROC __glewMultiTexCoord2sv = NULL;
+PFNGLMULTITEXCOORD3DPROC __glewMultiTexCoord3d = NULL;
+PFNGLMULTITEXCOORD3DVPROC __glewMultiTexCoord3dv = NULL;
+PFNGLMULTITEXCOORD3FPROC __glewMultiTexCoord3f = NULL;
+PFNGLMULTITEXCOORD3FVPROC __glewMultiTexCoord3fv = NULL;
+PFNGLMULTITEXCOORD3IPROC __glewMultiTexCoord3i = NULL;
+PFNGLMULTITEXCOORD3IVPROC __glewMultiTexCoord3iv = NULL;
+PFNGLMULTITEXCOORD3SPROC __glewMultiTexCoord3s = NULL;
+PFNGLMULTITEXCOORD3SVPROC __glewMultiTexCoord3sv = NULL;
+PFNGLMULTITEXCOORD4DPROC __glewMultiTexCoord4d = NULL;
+PFNGLMULTITEXCOORD4DVPROC __glewMultiTexCoord4dv = NULL;
+PFNGLMULTITEXCOORD4FPROC __glewMultiTexCoord4f = NULL;
+PFNGLMULTITEXCOORD4FVPROC __glewMultiTexCoord4fv = NULL;
+PFNGLMULTITEXCOORD4IPROC __glewMultiTexCoord4i = NULL;
+PFNGLMULTITEXCOORD4IVPROC __glewMultiTexCoord4iv = NULL;
+PFNGLMULTITEXCOORD4SPROC __glewMultiTexCoord4s = NULL;
+PFNGLMULTITEXCOORD4SVPROC __glewMultiTexCoord4sv = NULL;
+PFNGLSAMPLECOVERAGEPROC __glewSampleCoverage = NULL;
+
+PFNGLBLENDCOLORPROC __glewBlendColor = NULL;
+PFNGLBLENDEQUATIONPROC __glewBlendEquation = NULL;
+PFNGLBLENDFUNCSEPARATEPROC __glewBlendFuncSeparate = NULL;
+PFNGLFOGCOORDPOINTERPROC __glewFogCoordPointer = NULL;
+PFNGLFOGCOORDDPROC __glewFogCoordd = NULL;
+PFNGLFOGCOORDDVPROC __glewFogCoorddv = NULL;
+PFNGLFOGCOORDFPROC __glewFogCoordf = NULL;
+PFNGLFOGCOORDFVPROC __glewFogCoordfv = NULL;
+PFNGLMULTIDRAWARRAYSPROC __glewMultiDrawArrays = NULL;
+PFNGLMULTIDRAWELEMENTSPROC __glewMultiDrawElements = NULL;
+PFNGLPOINTPARAMETERFPROC __glewPointParameterf = NULL;
+PFNGLPOINTPARAMETERFVPROC __glewPointParameterfv = NULL;
+PFNGLPOINTPARAMETERIPROC __glewPointParameteri = NULL;
+PFNGLPOINTPARAMETERIVPROC __glewPointParameteriv = NULL;
+PFNGLSECONDARYCOLOR3BPROC __glewSecondaryColor3b = NULL;
+PFNGLSECONDARYCOLOR3BVPROC __glewSecondaryColor3bv = NULL;
+PFNGLSECONDARYCOLOR3DPROC __glewSecondaryColor3d = NULL;
+PFNGLSECONDARYCOLOR3DVPROC __glewSecondaryColor3dv = NULL;
+PFNGLSECONDARYCOLOR3FPROC __glewSecondaryColor3f = NULL;
+PFNGLSECONDARYCOLOR3FVPROC __glewSecondaryColor3fv = NULL;
+PFNGLSECONDARYCOLOR3IPROC __glewSecondaryColor3i = NULL;
+PFNGLSECONDARYCOLOR3IVPROC __glewSecondaryColor3iv = NULL;
+PFNGLSECONDARYCOLOR3SPROC __glewSecondaryColor3s = NULL;
+PFNGLSECONDARYCOLOR3SVPROC __glewSecondaryColor3sv = NULL;
+PFNGLSECONDARYCOLOR3UBPROC __glewSecondaryColor3ub = NULL;
+PFNGLSECONDARYCOLOR3UBVPROC __glewSecondaryColor3ubv = NULL;
+PFNGLSECONDARYCOLOR3UIPROC __glewSecondaryColor3ui = NULL;
+PFNGLSECONDARYCOLOR3UIVPROC __glewSecondaryColor3uiv = NULL;
+PFNGLSECONDARYCOLOR3USPROC __glewSecondaryColor3us = NULL;
+PFNGLSECONDARYCOLOR3USVPROC __glewSecondaryColor3usv = NULL;
+PFNGLSECONDARYCOLORPOINTERPROC __glewSecondaryColorPointer = NULL;
+PFNGLWINDOWPOS2DPROC __glewWindowPos2d = NULL;
+PFNGLWINDOWPOS2DVPROC __glewWindowPos2dv = NULL;
+PFNGLWINDOWPOS2FPROC __glewWindowPos2f = NULL;
+PFNGLWINDOWPOS2FVPROC __glewWindowPos2fv = NULL;
+PFNGLWINDOWPOS2IPROC __glewWindowPos2i = NULL;
+PFNGLWINDOWPOS2IVPROC __glewWindowPos2iv = NULL;
+PFNGLWINDOWPOS2SPROC __glewWindowPos2s = NULL;
+PFNGLWINDOWPOS2SVPROC __glewWindowPos2sv = NULL;
+PFNGLWINDOWPOS3DPROC __glewWindowPos3d = NULL;
+PFNGLWINDOWPOS3DVPROC __glewWindowPos3dv = NULL;
+PFNGLWINDOWPOS3FPROC __glewWindowPos3f = NULL;
+PFNGLWINDOWPOS3FVPROC __glewWindowPos3fv = NULL;
+PFNGLWINDOWPOS3IPROC __glewWindowPos3i = NULL;
+PFNGLWINDOWPOS3IVPROC __glewWindowPos3iv = NULL;
+PFNGLWINDOWPOS3SPROC __glewWindowPos3s = NULL;
+PFNGLWINDOWPOS3SVPROC __glewWindowPos3sv = NULL;
+
+PFNGLBEGINQUERYPROC __glewBeginQuery = NULL;
+PFNGLBINDBUFFERPROC __glewBindBuffer = NULL;
+PFNGLBUFFERDATAPROC __glewBufferData = NULL;
+PFNGLBUFFERSUBDATAPROC __glewBufferSubData = NULL;
+PFNGLDELETEBUFFERSPROC __glewDeleteBuffers = NULL;
+PFNGLDELETEQUERIESPROC __glewDeleteQueries = NULL;
+PFNGLENDQUERYPROC __glewEndQuery = NULL;
+PFNGLGENBUFFERSPROC __glewGenBuffers = NULL;
+PFNGLGENQUERIESPROC __glewGenQueries = NULL;
+PFNGLGETBUFFERPARAMETERIVPROC __glewGetBufferParameteriv = NULL;
+PFNGLGETBUFFERPOINTERVPROC __glewGetBufferPointerv = NULL;
+PFNGLGETBUFFERSUBDATAPROC __glewGetBufferSubData = NULL;
+PFNGLGETQUERYOBJECTIVPROC __glewGetQueryObjectiv = NULL;
+PFNGLGETQUERYOBJECTUIVPROC __glewGetQueryObjectuiv = NULL;
+PFNGLGETQUERYIVPROC __glewGetQueryiv = NULL;
+PFNGLISBUFFERPROC __glewIsBuffer = NULL;
+PFNGLISQUERYPROC __glewIsQuery = NULL;
+PFNGLMAPBUFFERPROC __glewMapBuffer = NULL;
+PFNGLUNMAPBUFFERPROC __glewUnmapBuffer = NULL;
+
+PFNGLATTACHSHADERPROC __glewAttachShader = NULL;
+PFNGLBINDATTRIBLOCATIONPROC __glewBindAttribLocation = NULL;
+PFNGLBLENDEQUATIONSEPARATEPROC __glewBlendEquationSeparate = NULL;
+PFNGLCOMPILESHADERPROC __glewCompileShader = NULL;
+PFNGLCREATEPROGRAMPROC __glewCreateProgram = NULL;
+PFNGLCREATESHADERPROC __glewCreateShader = NULL;
+PFNGLDELETEPROGRAMPROC __glewDeleteProgram = NULL;
+PFNGLDELETESHADERPROC __glewDeleteShader = NULL;
+PFNGLDETACHSHADERPROC __glewDetachShader = NULL;
+PFNGLDISABLEVERTEXATTRIBARRAYPROC __glewDisableVertexAttribArray = NULL;
+PFNGLDRAWBUFFERSPROC __glewDrawBuffers = NULL;
+PFNGLENABLEVERTEXATTRIBARRAYPROC __glewEnableVertexAttribArray = NULL;
+PFNGLGETACTIVEATTRIBPROC __glewGetActiveAttrib = NULL;
+PFNGLGETACTIVEUNIFORMPROC __glewGetActiveUniform = NULL;
+PFNGLGETATTACHEDSHADERSPROC __glewGetAttachedShaders = NULL;
+PFNGLGETATTRIBLOCATIONPROC __glewGetAttribLocation = NULL;
+PFNGLGETPROGRAMINFOLOGPROC __glewGetProgramInfoLog = NULL;
+PFNGLGETPROGRAMIVPROC __glewGetProgramiv = NULL;
+PFNGLGETSHADERINFOLOGPROC __glewGetShaderInfoLog = NULL;
+PFNGLGETSHADERSOURCEPROC __glewGetShaderSource = NULL;
+PFNGLGETSHADERIVPROC __glewGetShaderiv = NULL;
+PFNGLGETUNIFORMLOCATIONPROC __glewGetUniformLocation = NULL;
+PFNGLGETUNIFORMFVPROC __glewGetUniformfv = NULL;
+PFNGLGETUNIFORMIVPROC __glewGetUniformiv = NULL;
+PFNGLGETVERTEXATTRIBPOINTERVPROC __glewGetVertexAttribPointerv = NULL;
+PFNGLGETVERTEXATTRIBDVPROC __glewGetVertexAttribdv = NULL;
+PFNGLGETVERTEXATTRIBFVPROC __glewGetVertexAttribfv = NULL;
+PFNGLGETVERTEXATTRIBIVPROC __glewGetVertexAttribiv = NULL;
+PFNGLISPROGRAMPROC __glewIsProgram = NULL;
+PFNGLISSHADERPROC __glewIsShader = NULL;
+PFNGLLINKPROGRAMPROC __glewLinkProgram = NULL;
+PFNGLSHADERSOURCEPROC __glewShaderSource = NULL;
+PFNGLSTENCILFUNCSEPARATEPROC __glewStencilFuncSeparate = NULL;
+PFNGLSTENCILMASKSEPARATEPROC __glewStencilMaskSeparate = NULL;
+PFNGLSTENCILOPSEPARATEPROC __glewStencilOpSeparate = NULL;
+PFNGLUNIFORM1FPROC __glewUniform1f = NULL;
+PFNGLUNIFORM1FVPROC __glewUniform1fv = NULL;
+PFNGLUNIFORM1IPROC __glewUniform1i = NULL;
+PFNGLUNIFORM1IVPROC __glewUniform1iv = NULL;
+PFNGLUNIFORM2FPROC __glewUniform2f = NULL;
+PFNGLUNIFORM2FVPROC __glewUniform2fv = NULL;
+PFNGLUNIFORM2IPROC __glewUniform2i = NULL;
+PFNGLUNIFORM2IVPROC __glewUniform2iv = NULL;
+PFNGLUNIFORM3FPROC __glewUniform3f = NULL;
+PFNGLUNIFORM3FVPROC __glewUniform3fv = NULL;
+PFNGLUNIFORM3IPROC __glewUniform3i = NULL;
+PFNGLUNIFORM3IVPROC __glewUniform3iv = NULL;
+PFNGLUNIFORM4FPROC __glewUniform4f = NULL;
+PFNGLUNIFORM4FVPROC __glewUniform4fv = NULL;
+PFNGLUNIFORM4IPROC __glewUniform4i = NULL;
+PFNGLUNIFORM4IVPROC __glewUniform4iv = NULL;
+PFNGLUNIFORMMATRIX2FVPROC __glewUniformMatrix2fv = NULL;
+PFNGLUNIFORMMATRIX3FVPROC __glewUniformMatrix3fv = NULL;
+PFNGLUNIFORMMATRIX4FVPROC __glewUniformMatrix4fv = NULL;
+PFNGLUSEPROGRAMPROC __glewUseProgram = NULL;
+PFNGLVALIDATEPROGRAMPROC __glewValidateProgram = NULL;
+PFNGLVERTEXATTRIB1DPROC __glewVertexAttrib1d = NULL;
+PFNGLVERTEXATTRIB1DVPROC __glewVertexAttrib1dv = NULL;
+PFNGLVERTEXATTRIB1FPROC __glewVertexAttrib1f = NULL;
+PFNGLVERTEXATTRIB1FVPROC __glewVertexAttrib1fv = NULL;
+PFNGLVERTEXATTRIB1SPROC __glewVertexAttrib1s = NULL;
+PFNGLVERTEXATTRIB1SVPROC __glewVertexAttrib1sv = NULL;
+PFNGLVERTEXATTRIB2DPROC __glewVertexAttrib2d = NULL;
+PFNGLVERTEXATTRIB2DVPROC __glewVertexAttrib2dv = NULL;
+PFNGLVERTEXATTRIB2FPROC __glewVertexAttrib2f = NULL;
+PFNGLVERTEXATTRIB2FVPROC __glewVertexAttrib2fv = NULL;
+PFNGLVERTEXATTRIB2SPROC __glewVertexAttrib2s = NULL;
+PFNGLVERTEXATTRIB2SVPROC __glewVertexAttrib2sv = NULL;
+PFNGLVERTEXATTRIB3DPROC __glewVertexAttrib3d = NULL;
+PFNGLVERTEXATTRIB3DVPROC __glewVertexAttrib3dv = NULL;
+PFNGLVERTEXATTRIB3FPROC __glewVertexAttrib3f = NULL;
+PFNGLVERTEXATTRIB3FVPROC __glewVertexAttrib3fv = NULL;
+PFNGLVERTEXATTRIB3SPROC __glewVertexAttrib3s = NULL;
+PFNGLVERTEXATTRIB3SVPROC __glewVertexAttrib3sv = NULL;
+PFNGLVERTEXATTRIB4NBVPROC __glewVertexAttrib4Nbv = NULL;
+PFNGLVERTEXATTRIB4NIVPROC __glewVertexAttrib4Niv = NULL;
+PFNGLVERTEXATTRIB4NSVPROC __glewVertexAttrib4Nsv = NULL;
+PFNGLVERTEXATTRIB4NUBPROC __glewVertexAttrib4Nub = NULL;
+PFNGLVERTEXATTRIB4NUBVPROC __glewVertexAttrib4Nubv = NULL;
+PFNGLVERTEXATTRIB4NUIVPROC __glewVertexAttrib4Nuiv = NULL;
+PFNGLVERTEXATTRIB4NUSVPROC __glewVertexAttrib4Nusv = NULL;
+PFNGLVERTEXATTRIB4BVPROC __glewVertexAttrib4bv = NULL;
+PFNGLVERTEXATTRIB4DPROC __glewVertexAttrib4d = NULL;
+PFNGLVERTEXATTRIB4DVPROC __glewVertexAttrib4dv = NULL;
+PFNGLVERTEXATTRIB4FPROC __glewVertexAttrib4f = NULL;
+PFNGLVERTEXATTRIB4FVPROC __glewVertexAttrib4fv = NULL;
+PFNGLVERTEXATTRIB4IVPROC __glewVertexAttrib4iv = NULL;
+PFNGLVERTEXATTRIB4SPROC __glewVertexAttrib4s = NULL;
+PFNGLVERTEXATTRIB4SVPROC __glewVertexAttrib4sv = NULL;
+PFNGLVERTEXATTRIB4UBVPROC __glewVertexAttrib4ubv = NULL;
+PFNGLVERTEXATTRIB4UIVPROC __glewVertexAttrib4uiv = NULL;
+PFNGLVERTEXATTRIB4USVPROC __glewVertexAttrib4usv = NULL;
+PFNGLVERTEXATTRIBPOINTERPROC __glewVertexAttribPointer = NULL;
+
+PFNGLUNIFORMMATRIX2X3FVPROC __glewUniformMatrix2x3fv = NULL;
+PFNGLUNIFORMMATRIX2X4FVPROC __glewUniformMatrix2x4fv = NULL;
+PFNGLUNIFORMMATRIX3X2FVPROC __glewUniformMatrix3x2fv = NULL;
+PFNGLUNIFORMMATRIX3X4FVPROC __glewUniformMatrix3x4fv = NULL;
+PFNGLUNIFORMMATRIX4X2FVPROC __glewUniformMatrix4x2fv = NULL;
+PFNGLUNIFORMMATRIX4X3FVPROC __glewUniformMatrix4x3fv = NULL;
+
+PFNGLBEGINCONDITIONALRENDERPROC __glewBeginConditionalRender = NULL;
+PFNGLBEGINTRANSFORMFEEDBACKPROC __glewBeginTransformFeedback = NULL;
+PFNGLBINDBUFFERBASEPROC __glewBindBufferBase = NULL;
+PFNGLBINDBUFFERRANGEPROC __glewBindBufferRange = NULL;
+PFNGLBINDFRAGDATALOCATIONPROC __glewBindFragDataLocation = NULL;
+PFNGLCLAMPCOLORPROC __glewClampColor = NULL;
+PFNGLCLEARBUFFERFIPROC __glewClearBufferfi = NULL;
+PFNGLCLEARBUFFERFVPROC __glewClearBufferfv = NULL;
+PFNGLCLEARBUFFERIVPROC __glewClearBufferiv = NULL;
+PFNGLCLEARBUFFERUIVPROC __glewClearBufferuiv = NULL;
+PFNGLCOLORMASKIPROC __glewColorMaski = NULL;
+PFNGLDISABLEIPROC __glewDisablei = NULL;
+PFNGLENABLEIPROC __glewEnablei = NULL;
+PFNGLENDCONDITIONALRENDERPROC __glewEndConditionalRender = NULL;
+PFNGLENDTRANSFORMFEEDBACKPROC __glewEndTransformFeedback = NULL;
+PFNGLGETBOOLEANI_VPROC __glewGetBooleani_v = NULL;
+PFNGLGETFRAGDATALOCATIONPROC __glewGetFragDataLocation = NULL;
+PFNGLGETINTEGERI_VPROC __glewGetIntegeri_v = NULL;
+PFNGLGETSTRINGIPROC __glewGetStringi = NULL;
+PFNGLGETTEXPARAMETERIIVPROC __glewGetTexParameterIiv = NULL;
+PFNGLGETTEXPARAMETERIUIVPROC __glewGetTexParameterIuiv = NULL;
+PFNGLGETTRANSFORMFEEDBACKVARYINGPROC __glewGetTransformFeedbackVarying = NULL;
+PFNGLGETUNIFORMUIVPROC __glewGetUniformuiv = NULL;
+PFNGLGETVERTEXATTRIBIIVPROC __glewGetVertexAttribIiv = NULL;
+PFNGLGETVERTEXATTRIBIUIVPROC __glewGetVertexAttribIuiv = NULL;
+PFNGLISENABLEDIPROC __glewIsEnabledi = NULL;
+PFNGLTEXPARAMETERIIVPROC __glewTexParameterIiv = NULL;
+PFNGLTEXPARAMETERIUIVPROC __glewTexParameterIuiv = NULL;
+PFNGLTRANSFORMFEEDBACKVARYINGSPROC __glewTransformFeedbackVaryings = NULL;
+PFNGLUNIFORM1UIPROC __glewUniform1ui = NULL;
+PFNGLUNIFORM1UIVPROC __glewUniform1uiv = NULL;
+PFNGLUNIFORM2UIPROC __glewUniform2ui = NULL;
+PFNGLUNIFORM2UIVPROC __glewUniform2uiv = NULL;
+PFNGLUNIFORM3UIPROC __glewUniform3ui = NULL;
+PFNGLUNIFORM3UIVPROC __glewUniform3uiv = NULL;
+PFNGLUNIFORM4UIPROC __glewUniform4ui = NULL;
+PFNGLUNIFORM4UIVPROC __glewUniform4uiv = NULL;
+PFNGLVERTEXATTRIBI1IPROC __glewVertexAttribI1i = NULL;
+PFNGLVERTEXATTRIBI1IVPROC __glewVertexAttribI1iv = NULL;
+PFNGLVERTEXATTRIBI1UIPROC __glewVertexAttribI1ui = NULL;
+PFNGLVERTEXATTRIBI1UIVPROC __glewVertexAttribI1uiv = NULL;
+PFNGLVERTEXATTRIBI2IPROC __glewVertexAttribI2i = NULL;
+PFNGLVERTEXATTRIBI2IVPROC __glewVertexAttribI2iv = NULL;
+PFNGLVERTEXATTRIBI2UIPROC __glewVertexAttribI2ui = NULL;
+PFNGLVERTEXATTRIBI2UIVPROC __glewVertexAttribI2uiv = NULL;
+PFNGLVERTEXATTRIBI3IPROC __glewVertexAttribI3i = NULL;
+PFNGLVERTEXATTRIBI3IVPROC __glewVertexAttribI3iv = NULL;
+PFNGLVERTEXATTRIBI3UIPROC __glewVertexAttribI3ui = NULL;
+PFNGLVERTEXATTRIBI3UIVPROC __glewVertexAttribI3uiv = NULL;
+PFNGLVERTEXATTRIBI4BVPROC __glewVertexAttribI4bv = NULL;
+PFNGLVERTEXATTRIBI4IPROC __glewVertexAttribI4i = NULL;
+PFNGLVERTEXATTRIBI4IVPROC __glewVertexAttribI4iv = NULL;
+PFNGLVERTEXATTRIBI4SVPROC __glewVertexAttribI4sv = NULL;
+PFNGLVERTEXATTRIBI4UBVPROC __glewVertexAttribI4ubv = NULL;
+PFNGLVERTEXATTRIBI4UIPROC __glewVertexAttribI4ui = NULL;
+PFNGLVERTEXATTRIBI4UIVPROC __glewVertexAttribI4uiv = NULL;
+PFNGLVERTEXATTRIBI4USVPROC __glewVertexAttribI4usv = NULL;
+PFNGLVERTEXATTRIBIPOINTERPROC __glewVertexAttribIPointer = NULL;
+
+PFNGLTBUFFERMASK3DFXPROC __glewTbufferMask3DFX = NULL;
+
+PFNGLDRAWELEMENTARRAYAPPLEPROC __glewDrawElementArrayAPPLE = NULL;
+PFNGLDRAWRANGEELEMENTARRAYAPPLEPROC __glewDrawRangeElementArrayAPPLE = NULL;
+PFNGLELEMENTPOINTERAPPLEPROC __glewElementPointerAPPLE = NULL;
+PFNGLMULTIDRAWELEMENTARRAYAPPLEPROC __glewMultiDrawElementArrayAPPLE = NULL;
+PFNGLMULTIDRAWRANGEELEMENTARRAYAPPLEPROC __glewMultiDrawRangeElementArrayAPPLE = NULL;
+
+PFNGLDELETEFENCESAPPLEPROC __glewDeleteFencesAPPLE = NULL;
+PFNGLFINISHFENCEAPPLEPROC __glewFinishFenceAPPLE = NULL;
+PFNGLFINISHOBJECTAPPLEPROC __glewFinishObjectAPPLE = NULL;
+PFNGLGENFENCESAPPLEPROC __glewGenFencesAPPLE = NULL;
+PFNGLISFENCEAPPLEPROC __glewIsFenceAPPLE = NULL;
+PFNGLSETFENCEAPPLEPROC __glewSetFenceAPPLE = NULL;
+PFNGLTESTFENCEAPPLEPROC __glewTestFenceAPPLE = NULL;
+PFNGLTESTOBJECTAPPLEPROC __glewTestObjectAPPLE = NULL;
+
+PFNGLBUFFERPARAMETERIAPPLEPROC __glewBufferParameteriAPPLE = NULL;
+PFNGLFLUSHMAPPEDBUFFERRANGEAPPLEPROC __glewFlushMappedBufferRangeAPPLE = NULL;
+
+PFNGLGETTEXPARAMETERPOINTERVAPPLEPROC __glewGetTexParameterPointervAPPLE = NULL;
+PFNGLTEXTURERANGEAPPLEPROC __glewTextureRangeAPPLE = NULL;
+
+PFNGLBINDVERTEXARRAYAPPLEPROC __glewBindVertexArrayAPPLE = NULL;
+PFNGLDELETEVERTEXARRAYSAPPLEPROC __glewDeleteVertexArraysAPPLE = NULL;
+PFNGLGENVERTEXARRAYSAPPLEPROC __glewGenVertexArraysAPPLE = NULL;
+PFNGLISVERTEXARRAYAPPLEPROC __glewIsVertexArrayAPPLE = NULL;
+
+PFNGLFLUSHVERTEXARRAYRANGEAPPLEPROC __glewFlushVertexArrayRangeAPPLE = NULL;
+PFNGLVERTEXARRAYPARAMETERIAPPLEPROC __glewVertexArrayParameteriAPPLE = NULL;
+PFNGLVERTEXARRAYRANGEAPPLEPROC __glewVertexArrayRangeAPPLE = NULL;
+
+PFNGLCLAMPCOLORARBPROC __glewClampColorARB = NULL;
+
+PFNGLDRAWBUFFERSARBPROC __glewDrawBuffersARB = NULL;
+
+PFNGLDRAWARRAYSINSTANCEDARBPROC __glewDrawArraysInstancedARB = NULL;
+PFNGLDRAWELEMENTSINSTANCEDARBPROC __glewDrawElementsInstancedARB = NULL;
+
+PFNGLBINDFRAMEBUFFERPROC __glewBindFramebuffer = NULL;
+PFNGLBINDRENDERBUFFERPROC __glewBindRenderbuffer = NULL;
+PFNGLBLITFRAMEBUFFERPROC __glewBlitFramebuffer = NULL;
+PFNGLCHECKFRAMEBUFFERSTATUSPROC __glewCheckFramebufferStatus = NULL;
+PFNGLDELETEFRAMEBUFFERSPROC __glewDeleteFramebuffers = NULL;
+PFNGLDELETERENDERBUFFERSPROC __glewDeleteRenderbuffers = NULL;
+PFNGLFRAMEBUFFERRENDERBUFFERPROC __glewFramebufferRenderbuffer = NULL;
+PFNGLFRAMEBUFFERTEXTURLAYERPROC __glewFramebufferTexturLayer = NULL;
+PFNGLFRAMEBUFFERTEXTURE1DPROC __glewFramebufferTexture1D = NULL;
+PFNGLFRAMEBUFFERTEXTURE2DPROC __glewFramebufferTexture2D = NULL;
+PFNGLFRAMEBUFFERTEXTURE3DPROC __glewFramebufferTexture3D = NULL;
+PFNGLGENFRAMEBUFFERSPROC __glewGenFramebuffers = NULL;
+PFNGLGENRENDERBUFFERSPROC __glewGenRenderbuffers = NULL;
+PFNGLGENERATEMIPMAPPROC __glewGenerateMipmap = NULL;
+PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC __glewGetFramebufferAttachmentParameteriv = NULL;
+PFNGLGETRENDERBUFFERPARAMETERIVPROC __glewGetRenderbufferParameteriv = NULL;
+PFNGLISFRAMEBUFFERPROC __glewIsFramebuffer = NULL;
+PFNGLISRENDERBUFFERPROC __glewIsRenderbuffer = NULL;
+PFNGLRENDERBUFFERSTORAGEPROC __glewRenderbufferStorage = NULL;
+PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC __glewRenderbufferStorageMultisample = NULL;
+
+PFNGLFRAMEBUFFERTEXTUREARBPROC __glewFramebufferTextureARB = NULL;
+PFNGLFRAMEBUFFERTEXTUREFACEARBPROC __glewFramebufferTextureFaceARB = NULL;
+PFNGLFRAMEBUFFERTEXTURELAYERARBPROC __glewFramebufferTextureLayerARB = NULL;
+PFNGLPROGRAMPARAMETERIARBPROC __glewProgramParameteriARB = NULL;
+
+PFNGLCOLORSUBTABLEPROC __glewColorSubTable = NULL;
+PFNGLCOLORTABLEPROC __glewColorTable = NULL;
+PFNGLCOLORTABLEPARAMETERFVPROC __glewColorTableParameterfv = NULL;
+PFNGLCOLORTABLEPARAMETERIVPROC __glewColorTableParameteriv = NULL;
+PFNGLCONVOLUTIONFILTER1DPROC __glewConvolutionFilter1D = NULL;
+PFNGLCONVOLUTIONFILTER2DPROC __glewConvolutionFilter2D = NULL;
+PFNGLCONVOLUTIONPARAMETERFPROC __glewConvolutionParameterf = NULL;
+PFNGLCONVOLUTIONPARAMETERFVPROC __glewConvolutionParameterfv = NULL;
+PFNGLCONVOLUTIONPARAMETERIPROC __glewConvolutionParameteri = NULL;
+PFNGLCONVOLUTIONPARAMETERIVPROC __glewConvolutionParameteriv = NULL;
+PFNGLCOPYCOLORSUBTABLEPROC __glewCopyColorSubTable = NULL;
+PFNGLCOPYCOLORTABLEPROC __glewCopyColorTable = NULL;
+PFNGLCOPYCONVOLUTIONFILTER1DPROC __glewCopyConvolutionFilter1D = NULL;
+PFNGLCOPYCONVOLUTIONFILTER2DPROC __glewCopyConvolutionFilter2D = NULL;
+PFNGLGETCOLORTABLEPROC __glewGetColorTable = NULL;
+PFNGLGETCOLORTABLEPARAMETERFVPROC __glewGetColorTableParameterfv = NULL;
+PFNGLGETCOLORTABLEPARAMETERIVPROC __glewGetColorTableParameteriv = NULL;
+PFNGLGETCONVOLUTIONFILTERPROC __glewGetConvolutionFilter = NULL;
+PFNGLGETCONVOLUTIONPARAMETERFVPROC __glewGetConvolutionParameterfv = NULL;
+PFNGLGETCONVOLUTIONPARAMETERIVPROC __glewGetConvolutionParameteriv = NULL;
+PFNGLGETHISTOGRAMPROC __glewGetHistogram = NULL;
+PFNGLGETHISTOGRAMPARAMETERFVPROC __glewGetHistogramParameterfv = NULL;
+PFNGLGETHISTOGRAMPARAMETERIVPROC __glewGetHistogramParameteriv = NULL;
+PFNGLGETMINMAXPROC __glewGetMinmax = NULL;
+PFNGLGETMINMAXPARAMETERFVPROC __glewGetMinmaxParameterfv = NULL;
+PFNGLGETMINMAXPARAMETERIVPROC __glewGetMinmaxParameteriv = NULL;
+PFNGLGETSEPARABLEFILTERPROC __glewGetSeparableFilter = NULL;
+PFNGLHISTOGRAMPROC __glewHistogram = NULL;
+PFNGLMINMAXPROC __glewMinmax = NULL;
+PFNGLRESETHISTOGRAMPROC __glewResetHistogram = NULL;
+PFNGLRESETMINMAXPROC __glewResetMinmax = NULL;
+PFNGLSEPARABLEFILTER2DPROC __glewSeparableFilter2D = NULL;
+
+PFNGLVERTEXATTRIBDIVISORARBPROC __glewVertexAttribDivisorARB = NULL;
+
+PFNGLFLUSHMAPPEDBUFFERRANGEPROC __glewFlushMappedBufferRange = NULL;
+PFNGLMAPBUFFERRANGEPROC __glewMapBufferRange = NULL;
+
+PFNGLCURRENTPALETTEMATRIXARBPROC __glewCurrentPaletteMatrixARB = NULL;
+PFNGLMATRIXINDEXPOINTERARBPROC __glewMatrixIndexPointerARB = NULL;
+PFNGLMATRIXINDEXUBVARBPROC __glewMatrixIndexubvARB = NULL;
+PFNGLMATRIXINDEXUIVARBPROC __glewMatrixIndexuivARB = NULL;
+PFNGLMATRIXINDEXUSVARBPROC __glewMatrixIndexusvARB = NULL;
+
+PFNGLSAMPLECOVERAGEARBPROC __glewSampleCoverageARB = NULL;
+
+PFNGLACTIVETEXTUREARBPROC __glewActiveTextureARB = NULL;
+PFNGLCLIENTACTIVETEXTUREARBPROC __glewClientActiveTextureARB = NULL;
+PFNGLMULTITEXCOORD1DARBPROC __glewMultiTexCoord1dARB = NULL;
+PFNGLMULTITEXCOORD1DVARBPROC __glewMultiTexCoord1dvARB = NULL;
+PFNGLMULTITEXCOORD1FARBPROC __glewMultiTexCoord1fARB = NULL;
+PFNGLMULTITEXCOORD1FVARBPROC __glewMultiTexCoord1fvARB = NULL;
+PFNGLMULTITEXCOORD1IARBPROC __glewMultiTexCoord1iARB = NULL;
+PFNGLMULTITEXCOORD1IVARBPROC __glewMultiTexCoord1ivARB = NULL;
+PFNGLMULTITEXCOORD1SARBPROC __glewMultiTexCoord1sARB = NULL;
+PFNGLMULTITEXCOORD1SVARBPROC __glewMultiTexCoord1svARB = NULL;
+PFNGLMULTITEXCOORD2DARBPROC __glewMultiTexCoord2dARB = NULL;
+PFNGLMULTITEXCOORD2DVARBPROC __glewMultiTexCoord2dvARB = NULL;
+PFNGLMULTITEXCOORD2FARBPROC __glewMultiTexCoord2fARB = NULL;
+PFNGLMULTITEXCOORD2FVARBPROC __glewMultiTexCoord2fvARB = NULL;
+PFNGLMULTITEXCOORD2IARBPROC __glewMultiTexCoord2iARB = NULL;
+PFNGLMULTITEXCOORD2IVARBPROC __glewMultiTexCoord2ivARB = NULL;
+PFNGLMULTITEXCOORD2SARBPROC __glewMultiTexCoord2sARB = NULL;
+PFNGLMULTITEXCOORD2SVARBPROC __glewMultiTexCoord2svARB = NULL;
+PFNGLMULTITEXCOORD3DARBPROC __glewMultiTexCoord3dARB = NULL;
+PFNGLMULTITEXCOORD3DVARBPROC __glewMultiTexCoord3dvARB = NULL;
+PFNGLMULTITEXCOORD3FARBPROC __glewMultiTexCoord3fARB = NULL;
+PFNGLMULTITEXCOORD3FVARBPROC __glewMultiTexCoord3fvARB = NULL;
+PFNGLMULTITEXCOORD3IARBPROC __glewMultiTexCoord3iARB = NULL;
+PFNGLMULTITEXCOORD3IVARBPROC __glewMultiTexCoord3ivARB = NULL;
+PFNGLMULTITEXCOORD3SARBPROC __glewMultiTexCoord3sARB = NULL;
+PFNGLMULTITEXCOORD3SVARBPROC __glewMultiTexCoord3svARB = NULL;
+PFNGLMULTITEXCOORD4DARBPROC __glewMultiTexCoord4dARB = NULL;
+PFNGLMULTITEXCOORD4DVARBPROC __glewMultiTexCoord4dvARB = NULL;
+PFNGLMULTITEXCOORD4FARBPROC __glewMultiTexCoord4fARB = NULL;
+PFNGLMULTITEXCOORD4FVARBPROC __glewMultiTexCoord4fvARB = NULL;
+PFNGLMULTITEXCOORD4IARBPROC __glewMultiTexCoord4iARB = NULL;
+PFNGLMULTITEXCOORD4IVARBPROC __glewMultiTexCoord4ivARB = NULL;
+PFNGLMULTITEXCOORD4SARBPROC __glewMultiTexCoord4sARB = NULL;
+PFNGLMULTITEXCOORD4SVARBPROC __glewMultiTexCoord4svARB = NULL;
+
+PFNGLBEGINQUERYARBPROC __glewBeginQueryARB = NULL;
+PFNGLDELETEQUERIESARBPROC __glewDeleteQueriesARB = NULL;
+PFNGLENDQUERYARBPROC __glewEndQueryARB = NULL;
+PFNGLGENQUERIESARBPROC __glewGenQueriesARB = NULL;
+PFNGLGETQUERYOBJECTIVARBPROC __glewGetQueryObjectivARB = NULL;
+PFNGLGETQUERYOBJECTUIVARBPROC __glewGetQueryObjectuivARB = NULL;
+PFNGLGETQUERYIVARBPROC __glewGetQueryivARB = NULL;
+PFNGLISQUERYARBPROC __glewIsQueryARB = NULL;
+
+PFNGLPOINTPARAMETERFARBPROC __glewPointParameterfARB = NULL;
+PFNGLPOINTPARAMETERFVARBPROC __glewPointParameterfvARB = NULL;
+
+PFNGLATTACHOBJECTARBPROC __glewAttachObjectARB = NULL;
+PFNGLCOMPILESHADERARBPROC __glewCompileShaderARB = NULL;
+PFNGLCREATEPROGRAMOBJECTARBPROC __glewCreateProgramObjectARB = NULL;
+PFNGLCREATESHADEROBJECTARBPROC __glewCreateShaderObjectARB = NULL;
+PFNGLDELETEOBJECTARBPROC __glewDeleteObjectARB = NULL;
+PFNGLDETACHOBJECTARBPROC __glewDetachObjectARB = NULL;
+PFNGLGETACTIVEUNIFORMARBPROC __glewGetActiveUniformARB = NULL;
+PFNGLGETATTACHEDOBJECTSARBPROC __glewGetAttachedObjectsARB = NULL;
+PFNGLGETHANDLEARBPROC __glewGetHandleARB = NULL;
+PFNGLGETINFOLOGARBPROC __glewGetInfoLogARB = NULL;
+PFNGLGETOBJECTPARAMETERFVARBPROC __glewGetObjectParameterfvARB = NULL;
+PFNGLGETOBJECTPARAMETERIVARBPROC __glewGetObjectParameterivARB = NULL;
+PFNGLGETSHADERSOURCEARBPROC __glewGetShaderSourceARB = NULL;
+PFNGLGETUNIFORMLOCATIONARBPROC __glewGetUniformLocationARB = NULL;
+PFNGLGETUNIFORMFVARBPROC __glewGetUniformfvARB = NULL;
+PFNGLGETUNIFORMIVARBPROC __glewGetUniformivARB = NULL;
+PFNGLLINKPROGRAMARBPROC __glewLinkProgramARB = NULL;
+PFNGLSHADERSOURCEARBPROC __glewShaderSourceARB = NULL;
+PFNGLUNIFORM1FARBPROC __glewUniform1fARB = NULL;
+PFNGLUNIFORM1FVARBPROC __glewUniform1fvARB = NULL;
+PFNGLUNIFORM1IARBPROC __glewUniform1iARB = NULL;
+PFNGLUNIFORM1IVARBPROC __glewUniform1ivARB = NULL;
+PFNGLUNIFORM2FARBPROC __glewUniform2fARB = NULL;
+PFNGLUNIFORM2FVARBPROC __glewUniform2fvARB = NULL;
+PFNGLUNIFORM2IARBPROC __glewUniform2iARB = NULL;
+PFNGLUNIFORM2IVARBPROC __glewUniform2ivARB = NULL;
+PFNGLUNIFORM3FARBPROC __glewUniform3fARB = NULL;
+PFNGLUNIFORM3FVARBPROC __glewUniform3fvARB = NULL;
+PFNGLUNIFORM3IARBPROC __glewUniform3iARB = NULL;
+PFNGLUNIFORM3IVARBPROC __glewUniform3ivARB = NULL;
+PFNGLUNIFORM4FARBPROC __glewUniform4fARB = NULL;
+PFNGLUNIFORM4FVARBPROC __glewUniform4fvARB = NULL;
+PFNGLUNIFORM4IARBPROC __glewUniform4iARB = NULL;
+PFNGLUNIFORM4IVARBPROC __glewUniform4ivARB = NULL;
+PFNGLUNIFORMMATRIX2FVARBPROC __glewUniformMatrix2fvARB = NULL;
+PFNGLUNIFORMMATRIX3FVARBPROC __glewUniformMatrix3fvARB = NULL;
+PFNGLUNIFORMMATRIX4FVARBPROC __glewUniformMatrix4fvARB = NULL;
+PFNGLUSEPROGRAMOBJECTARBPROC __glewUseProgramObjectARB = NULL;
+PFNGLVALIDATEPROGRAMARBPROC __glewValidateProgramARB = NULL;
+
+PFNGLTEXBUFFERARBPROC __glewTexBufferARB = NULL;
+
+PFNGLCOMPRESSEDTEXIMAGE1DARBPROC __glewCompressedTexImage1DARB = NULL;
+PFNGLCOMPRESSEDTEXIMAGE2DARBPROC __glewCompressedTexImage2DARB = NULL;
+PFNGLCOMPRESSEDTEXIMAGE3DARBPROC __glewCompressedTexImage3DARB = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE1DARBPROC __glewCompressedTexSubImage1DARB = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE2DARBPROC __glewCompressedTexSubImage2DARB = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE3DARBPROC __glewCompressedTexSubImage3DARB = NULL;
+PFNGLGETCOMPRESSEDTEXIMAGEARBPROC __glewGetCompressedTexImageARB = NULL;
+
+PFNGLLOADTRANSPOSEMATRIXDARBPROC __glewLoadTransposeMatrixdARB = NULL;
+PFNGLLOADTRANSPOSEMATRIXFARBPROC __glewLoadTransposeMatrixfARB = NULL;
+PFNGLMULTTRANSPOSEMATRIXDARBPROC __glewMultTransposeMatrixdARB = NULL;
+PFNGLMULTTRANSPOSEMATRIXFARBPROC __glewMultTransposeMatrixfARB = NULL;
+
+PFNGLBINDVERTEXARRAYPROC __glewBindVertexArray = NULL;
+PFNGLDELETEVERTEXARRAYSPROC __glewDeleteVertexArrays = NULL;
+PFNGLGENVERTEXARRAYSPROC __glewGenVertexArrays = NULL;
+PFNGLISVERTEXARRAYPROC __glewIsVertexArray = NULL;
+
+PFNGLVERTEXBLENDARBPROC __glewVertexBlendARB = NULL;
+PFNGLWEIGHTPOINTERARBPROC __glewWeightPointerARB = NULL;
+PFNGLWEIGHTBVARBPROC __glewWeightbvARB = NULL;
+PFNGLWEIGHTDVARBPROC __glewWeightdvARB = NULL;
+PFNGLWEIGHTFVARBPROC __glewWeightfvARB = NULL;
+PFNGLWEIGHTIVARBPROC __glewWeightivARB = NULL;
+PFNGLWEIGHTSVARBPROC __glewWeightsvARB = NULL;
+PFNGLWEIGHTUBVARBPROC __glewWeightubvARB = NULL;
+PFNGLWEIGHTUIVARBPROC __glewWeightuivARB = NULL;
+PFNGLWEIGHTUSVARBPROC __glewWeightusvARB = NULL;
+
+PFNGLBINDBUFFERARBPROC __glewBindBufferARB = NULL;
+PFNGLBUFFERDATAARBPROC __glewBufferDataARB = NULL;
+PFNGLBUFFERSUBDATAARBPROC __glewBufferSubDataARB = NULL;
+PFNGLDELETEBUFFERSARBPROC __glewDeleteBuffersARB = NULL;
+PFNGLGENBUFFERSARBPROC __glewGenBuffersARB = NULL;
+PFNGLGETBUFFERPARAMETERIVARBPROC __glewGetBufferParameterivARB = NULL;
+PFNGLGETBUFFERPOINTERVARBPROC __glewGetBufferPointervARB = NULL;
+PFNGLGETBUFFERSUBDATAARBPROC __glewGetBufferSubDataARB = NULL;
+PFNGLISBUFFERARBPROC __glewIsBufferARB = NULL;
+PFNGLMAPBUFFERARBPROC __glewMapBufferARB = NULL;
+PFNGLUNMAPBUFFERARBPROC __glewUnmapBufferARB = NULL;
+
+PFNGLBINDPROGRAMARBPROC __glewBindProgramARB = NULL;
+PFNGLDELETEPROGRAMSARBPROC __glewDeleteProgramsARB = NULL;
+PFNGLDISABLEVERTEXATTRIBARRAYARBPROC __glewDisableVertexAttribArrayARB = NULL;
+PFNGLENABLEVERTEXATTRIBARRAYARBPROC __glewEnableVertexAttribArrayARB = NULL;
+PFNGLGENPROGRAMSARBPROC __glewGenProgramsARB = NULL;
+PFNGLGETPROGRAMENVPARAMETERDVARBPROC __glewGetProgramEnvParameterdvARB = NULL;
+PFNGLGETPROGRAMENVPARAMETERFVARBPROC __glewGetProgramEnvParameterfvARB = NULL;
+PFNGLGETPROGRAMLOCALPARAMETERDVARBPROC __glewGetProgramLocalParameterdvARB = NULL;
+PFNGLGETPROGRAMLOCALPARAMETERFVARBPROC __glewGetProgramLocalParameterfvARB = NULL;
+PFNGLGETPROGRAMSTRINGARBPROC __glewGetProgramStringARB = NULL;
+PFNGLGETPROGRAMIVARBPROC __glewGetProgramivARB = NULL;
+PFNGLGETVERTEXATTRIBPOINTERVARBPROC __glewGetVertexAttribPointervARB = NULL;
+PFNGLGETVERTEXATTRIBDVARBPROC __glewGetVertexAttribdvARB = NULL;
+PFNGLGETVERTEXATTRIBFVARBPROC __glewGetVertexAttribfvARB = NULL;
+PFNGLGETVERTEXATTRIBIVARBPROC __glewGetVertexAttribivARB = NULL;
+PFNGLISPROGRAMARBPROC __glewIsProgramARB = NULL;
+PFNGLPROGRAMENVPARAMETER4DARBPROC __glewProgramEnvParameter4dARB = NULL;
+PFNGLPROGRAMENVPARAMETER4DVARBPROC __glewProgramEnvParameter4dvARB = NULL;
+PFNGLPROGRAMENVPARAMETER4FARBPROC __glewProgramEnvParameter4fARB = NULL;
+PFNGLPROGRAMENVPARAMETER4FVARBPROC __glewProgramEnvParameter4fvARB = NULL;
+PFNGLPROGRAMLOCALPARAMETER4DARBPROC __glewProgramLocalParameter4dARB = NULL;
+PFNGLPROGRAMLOCALPARAMETER4DVARBPROC __glewProgramLocalParameter4dvARB = NULL;
+PFNGLPROGRAMLOCALPARAMETER4FARBPROC __glewProgramLocalParameter4fARB = NULL;
+PFNGLPROGRAMLOCALPARAMETER4FVARBPROC __glewProgramLocalParameter4fvARB = NULL;
+PFNGLPROGRAMSTRINGARBPROC __glewProgramStringARB = NULL;
+PFNGLVERTEXATTRIB1DARBPROC __glewVertexAttrib1dARB = NULL;
+PFNGLVERTEXATTRIB1DVARBPROC __glewVertexAttrib1dvARB = NULL;
+PFNGLVERTEXATTRIB1FARBPROC __glewVertexAttrib1fARB = NULL;
+PFNGLVERTEXATTRIB1FVARBPROC __glewVertexAttrib1fvARB = NULL;
+PFNGLVERTEXATTRIB1SARBPROC __glewVertexAttrib1sARB = NULL;
+PFNGLVERTEXATTRIB1SVARBPROC __glewVertexAttrib1svARB = NULL;
+PFNGLVERTEXATTRIB2DARBPROC __glewVertexAttrib2dARB = NULL;
+PFNGLVERTEXATTRIB2DVARBPROC __glewVertexAttrib2dvARB = NULL;
+PFNGLVERTEXATTRIB2FARBPROC __glewVertexAttrib2fARB = NULL;
+PFNGLVERTEXATTRIB2FVARBPROC __glewVertexAttrib2fvARB = NULL;
+PFNGLVERTEXATTRIB2SARBPROC __glewVertexAttrib2sARB = NULL;
+PFNGLVERTEXATTRIB2SVARBPROC __glewVertexAttrib2svARB = NULL;
+PFNGLVERTEXATTRIB3DARBPROC __glewVertexAttrib3dARB = NULL;
+PFNGLVERTEXATTRIB3DVARBPROC __glewVertexAttrib3dvARB = NULL;
+PFNGLVERTEXATTRIB3FARBPROC __glewVertexAttrib3fARB = NULL;
+PFNGLVERTEXATTRIB3FVARBPROC __glewVertexAttrib3fvARB = NULL;
+PFNGLVERTEXATTRIB3SARBPROC __glewVertexAttrib3sARB = NULL;
+PFNGLVERTEXATTRIB3SVARBPROC __glewVertexAttrib3svARB = NULL;
+PFNGLVERTEXATTRIB4NBVARBPROC __glewVertexAttrib4NbvARB = NULL;
+PFNGLVERTEXATTRIB4NIVARBPROC __glewVertexAttrib4NivARB = NULL;
+PFNGLVERTEXATTRIB4NSVARBPROC __glewVertexAttrib4NsvARB = NULL;
+PFNGLVERTEXATTRIB4NUBARBPROC __glewVertexAttrib4NubARB = NULL;
+PFNGLVERTEXATTRIB4NUBVARBPROC __glewVertexAttrib4NubvARB = NULL;
+PFNGLVERTEXATTRIB4NUIVARBPROC __glewVertexAttrib4NuivARB = NULL;
+PFNGLVERTEXATTRIB4NUSVARBPROC __glewVertexAttrib4NusvARB = NULL;
+PFNGLVERTEXATTRIB4BVARBPROC __glewVertexAttrib4bvARB = NULL;
+PFNGLVERTEXATTRIB4DARBPROC __glewVertexAttrib4dARB = NULL;
+PFNGLVERTEXATTRIB4DVARBPROC __glewVertexAttrib4dvARB = NULL;
+PFNGLVERTEXATTRIB4FARBPROC __glewVertexAttrib4fARB = NULL;
+PFNGLVERTEXATTRIB4FVARBPROC __glewVertexAttrib4fvARB = NULL;
+PFNGLVERTEXATTRIB4IVARBPROC __glewVertexAttrib4ivARB = NULL;
+PFNGLVERTEXATTRIB4SARBPROC __glewVertexAttrib4sARB = NULL;
+PFNGLVERTEXATTRIB4SVARBPROC __glewVertexAttrib4svARB = NULL;
+PFNGLVERTEXATTRIB4UBVARBPROC __glewVertexAttrib4ubvARB = NULL;
+PFNGLVERTEXATTRIB4UIVARBPROC __glewVertexAttrib4uivARB = NULL;
+PFNGLVERTEXATTRIB4USVARBPROC __glewVertexAttrib4usvARB = NULL;
+PFNGLVERTEXATTRIBPOINTERARBPROC __glewVertexAttribPointerARB = NULL;
+
+PFNGLBINDATTRIBLOCATIONARBPROC __glewBindAttribLocationARB = NULL;
+PFNGLGETACTIVEATTRIBARBPROC __glewGetActiveAttribARB = NULL;
+PFNGLGETATTRIBLOCATIONARBPROC __glewGetAttribLocationARB = NULL;
+
+PFNGLWINDOWPOS2DARBPROC __glewWindowPos2dARB = NULL;
+PFNGLWINDOWPOS2DVARBPROC __glewWindowPos2dvARB = NULL;
+PFNGLWINDOWPOS2FARBPROC __glewWindowPos2fARB = NULL;
+PFNGLWINDOWPOS2FVARBPROC __glewWindowPos2fvARB = NULL;
+PFNGLWINDOWPOS2IARBPROC __glewWindowPos2iARB = NULL;
+PFNGLWINDOWPOS2IVARBPROC __glewWindowPos2ivARB = NULL;
+PFNGLWINDOWPOS2SARBPROC __glewWindowPos2sARB = NULL;
+PFNGLWINDOWPOS2SVARBPROC __glewWindowPos2svARB = NULL;
+PFNGLWINDOWPOS3DARBPROC __glewWindowPos3dARB = NULL;
+PFNGLWINDOWPOS3DVARBPROC __glewWindowPos3dvARB = NULL;
+PFNGLWINDOWPOS3FARBPROC __glewWindowPos3fARB = NULL;
+PFNGLWINDOWPOS3FVARBPROC __glewWindowPos3fvARB = NULL;
+PFNGLWINDOWPOS3IARBPROC __glewWindowPos3iARB = NULL;
+PFNGLWINDOWPOS3IVARBPROC __glewWindowPos3ivARB = NULL;
+PFNGLWINDOWPOS3SARBPROC __glewWindowPos3sARB = NULL;
+PFNGLWINDOWPOS3SVARBPROC __glewWindowPos3svARB = NULL;
+
+PFNGLDRAWBUFFERSATIPROC __glewDrawBuffersATI = NULL;
+
+PFNGLDRAWELEMENTARRAYATIPROC __glewDrawElementArrayATI = NULL;
+PFNGLDRAWRANGEELEMENTARRAYATIPROC __glewDrawRangeElementArrayATI = NULL;
+PFNGLELEMENTPOINTERATIPROC __glewElementPointerATI = NULL;
+
+PFNGLGETTEXBUMPPARAMETERFVATIPROC __glewGetTexBumpParameterfvATI = NULL;
+PFNGLGETTEXBUMPPARAMETERIVATIPROC __glewGetTexBumpParameterivATI = NULL;
+PFNGLTEXBUMPPARAMETERFVATIPROC __glewTexBumpParameterfvATI = NULL;
+PFNGLTEXBUMPPARAMETERIVATIPROC __glewTexBumpParameterivATI = NULL;
+
+PFNGLALPHAFRAGMENTOP1ATIPROC __glewAlphaFragmentOp1ATI = NULL;
+PFNGLALPHAFRAGMENTOP2ATIPROC __glewAlphaFragmentOp2ATI = NULL;
+PFNGLALPHAFRAGMENTOP3ATIPROC __glewAlphaFragmentOp3ATI = NULL;
+PFNGLBEGINFRAGMENTSHADERATIPROC __glewBeginFragmentShaderATI = NULL;
+PFNGLBINDFRAGMENTSHADERATIPROC __glewBindFragmentShaderATI = NULL;
+PFNGLCOLORFRAGMENTOP1ATIPROC __glewColorFragmentOp1ATI = NULL;
+PFNGLCOLORFRAGMENTOP2ATIPROC __glewColorFragmentOp2ATI = NULL;
+PFNGLCOLORFRAGMENTOP3ATIPROC __glewColorFragmentOp3ATI = NULL;
+PFNGLDELETEFRAGMENTSHADERATIPROC __glewDeleteFragmentShaderATI = NULL;
+PFNGLENDFRAGMENTSHADERATIPROC __glewEndFragmentShaderATI = NULL;
+PFNGLGENFRAGMENTSHADERSATIPROC __glewGenFragmentShadersATI = NULL;
+PFNGLPASSTEXCOORDATIPROC __glewPassTexCoordATI = NULL;
+PFNGLSAMPLEMAPATIPROC __glewSampleMapATI = NULL;
+PFNGLSETFRAGMENTSHADERCONSTANTATIPROC __glewSetFragmentShaderConstantATI = NULL;
+
+PFNGLMAPOBJECTBUFFERATIPROC __glewMapObjectBufferATI = NULL;
+PFNGLUNMAPOBJECTBUFFERATIPROC __glewUnmapObjectBufferATI = NULL;
+
+PFNGLPNTRIANGLESFATIPROC __glPNTrianglewesfATI = NULL;
+PFNGLPNTRIANGLESIATIPROC __glPNTrianglewesiATI = NULL;
+
+PFNGLSTENCILFUNCSEPARATEATIPROC __glewStencilFuncSeparateATI = NULL;
+PFNGLSTENCILOPSEPARATEATIPROC __glewStencilOpSeparateATI = NULL;
+
+PFNGLARRAYOBJECTATIPROC __glewArrayObjectATI = NULL;
+PFNGLFREEOBJECTBUFFERATIPROC __glewFreeObjectBufferATI = NULL;
+PFNGLGETARRAYOBJECTFVATIPROC __glewGetArrayObjectfvATI = NULL;
+PFNGLGETARRAYOBJECTIVATIPROC __glewGetArrayObjectivATI = NULL;
+PFNGLGETOBJECTBUFFERFVATIPROC __glewGetObjectBufferfvATI = NULL;
+PFNGLGETOBJECTBUFFERIVATIPROC __glewGetObjectBufferivATI = NULL;
+PFNGLGETVARIANTARRAYOBJECTFVATIPROC __glewGetVariantArrayObjectfvATI = NULL;
+PFNGLGETVARIANTARRAYOBJECTIVATIPROC __glewGetVariantArrayObjectivATI = NULL;
+PFNGLISOBJECTBUFFERATIPROC __glewIsObjectBufferATI = NULL;
+PFNGLNEWOBJECTBUFFERATIPROC __glewNewObjectBufferATI = NULL;
+PFNGLUPDATEOBJECTBUFFERATIPROC __glewUpdateObjectBufferATI = NULL;
+PFNGLVARIANTARRAYOBJECTATIPROC __glewVariantArrayObjectATI = NULL;
+
+PFNGLGETVERTEXATTRIBARRAYOBJECTFVATIPROC __glewGetVertexAttribArrayObjectfvATI = NULL;
+PFNGLGETVERTEXATTRIBARRAYOBJECTIVATIPROC __glewGetVertexAttribArrayObjectivATI = NULL;
+PFNGLVERTEXATTRIBARRAYOBJECTATIPROC __glewVertexAttribArrayObjectATI = NULL;
+
+PFNGLCLIENTACTIVEVERTEXSTREAMATIPROC __glewClientActiveVertexStreamATI = NULL;
+PFNGLNORMALSTREAM3BATIPROC __glewNormalStream3bATI = NULL;
+PFNGLNORMALSTREAM3BVATIPROC __glewNormalStream3bvATI = NULL;
+PFNGLNORMALSTREAM3DATIPROC __glewNormalStream3dATI = NULL;
+PFNGLNORMALSTREAM3DVATIPROC __glewNormalStream3dvATI = NULL;
+PFNGLNORMALSTREAM3FATIPROC __glewNormalStream3fATI = NULL;
+PFNGLNORMALSTREAM3FVATIPROC __glewNormalStream3fvATI = NULL;
+PFNGLNORMALSTREAM3IATIPROC __glewNormalStream3iATI = NULL;
+PFNGLNORMALSTREAM3IVATIPROC __glewNormalStream3ivATI = NULL;
+PFNGLNORMALSTREAM3SATIPROC __glewNormalStream3sATI = NULL;
+PFNGLNORMALSTREAM3SVATIPROC __glewNormalStream3svATI = NULL;
+PFNGLVERTEXBLENDENVFATIPROC __glewVertexBlendEnvfATI = NULL;
+PFNGLVERTEXBLENDENVIATIPROC __glewVertexBlendEnviATI = NULL;
+PFNGLVERTEXSTREAM2DATIPROC __glewVertexStream2dATI = NULL;
+PFNGLVERTEXSTREAM2DVATIPROC __glewVertexStream2dvATI = NULL;
+PFNGLVERTEXSTREAM2FATIPROC __glewVertexStream2fATI = NULL;
+PFNGLVERTEXSTREAM2FVATIPROC __glewVertexStream2fvATI = NULL;
+PFNGLVERTEXSTREAM2IATIPROC __glewVertexStream2iATI = NULL;
+PFNGLVERTEXSTREAM2IVATIPROC __glewVertexStream2ivATI = NULL;
+PFNGLVERTEXSTREAM2SATIPROC __glewVertexStream2sATI = NULL;
+PFNGLVERTEXSTREAM2SVATIPROC __glewVertexStream2svATI = NULL;
+PFNGLVERTEXSTREAM3DATIPROC __glewVertexStream3dATI = NULL;
+PFNGLVERTEXSTREAM3DVATIPROC __glewVertexStream3dvATI = NULL;
+PFNGLVERTEXSTREAM3FATIPROC __glewVertexStream3fATI = NULL;
+PFNGLVERTEXSTREAM3FVATIPROC __glewVertexStream3fvATI = NULL;
+PFNGLVERTEXSTREAM3IATIPROC __glewVertexStream3iATI = NULL;
+PFNGLVERTEXSTREAM3IVATIPROC __glewVertexStream3ivATI = NULL;
+PFNGLVERTEXSTREAM3SATIPROC __glewVertexStream3sATI = NULL;
+PFNGLVERTEXSTREAM3SVATIPROC __glewVertexStream3svATI = NULL;
+PFNGLVERTEXSTREAM4DATIPROC __glewVertexStream4dATI = NULL;
+PFNGLVERTEXSTREAM4DVATIPROC __glewVertexStream4dvATI = NULL;
+PFNGLVERTEXSTREAM4FATIPROC __glewVertexStream4fATI = NULL;
+PFNGLVERTEXSTREAM4FVATIPROC __glewVertexStream4fvATI = NULL;
+PFNGLVERTEXSTREAM4IATIPROC __glewVertexStream4iATI = NULL;
+PFNGLVERTEXSTREAM4IVATIPROC __glewVertexStream4ivATI = NULL;
+PFNGLVERTEXSTREAM4SATIPROC __glewVertexStream4sATI = NULL;
+PFNGLVERTEXSTREAM4SVATIPROC __glewVertexStream4svATI = NULL;
+
+PFNGLGETUNIFORMBUFFERSIZEEXTPROC __glewGetUniformBufferSizeEXT = NULL;
+PFNGLGETUNIFORMOFFSETEXTPROC __glewGetUniformOffsetEXT = NULL;
+PFNGLUNIFORMBUFFEREXTPROC __glewUniformBufferEXT = NULL;
+
+PFNGLBLENDCOLOREXTPROC __glewBlendColorEXT = NULL;
+
+PFNGLBLENDEQUATIONSEPARATEEXTPROC __glewBlendEquationSeparateEXT = NULL;
+
+PFNGLBLENDFUNCSEPARATEEXTPROC __glewBlendFuncSeparateEXT = NULL;
+
+PFNGLBLENDEQUATIONEXTPROC __glewBlendEquationEXT = NULL;
+
+PFNGLCOLORSUBTABLEEXTPROC __glewColorSubTableEXT = NULL;
+PFNGLCOPYCOLORSUBTABLEEXTPROC __glewCopyColorSubTableEXT = NULL;
+
+PFNGLLOCKARRAYSEXTPROC __glewLockArraysEXT = NULL;
+PFNGLUNLOCKARRAYSEXTPROC __glewUnlockArraysEXT = NULL;
+
+PFNGLCONVOLUTIONFILTER1DEXTPROC __glewConvolutionFilter1DEXT = NULL;
+PFNGLCONVOLUTIONFILTER2DEXTPROC __glewConvolutionFilter2DEXT = NULL;
+PFNGLCONVOLUTIONPARAMETERFEXTPROC __glewConvolutionParameterfEXT = NULL;
+PFNGLCONVOLUTIONPARAMETERFVEXTPROC __glewConvolutionParameterfvEXT = NULL;
+PFNGLCONVOLUTIONPARAMETERIEXTPROC __glewConvolutionParameteriEXT = NULL;
+PFNGLCONVOLUTIONPARAMETERIVEXTPROC __glewConvolutionParameterivEXT = NULL;
+PFNGLCOPYCONVOLUTIONFILTER1DEXTPROC __glewCopyConvolutionFilter1DEXT = NULL;
+PFNGLCOPYCONVOLUTIONFILTER2DEXTPROC __glewCopyConvolutionFilter2DEXT = NULL;
+PFNGLGETCONVOLUTIONFILTEREXTPROC __glewGetConvolutionFilterEXT = NULL;
+PFNGLGETCONVOLUTIONPARAMETERFVEXTPROC __glewGetConvolutionParameterfvEXT = NULL;
+PFNGLGETCONVOLUTIONPARAMETERIVEXTPROC __glewGetConvolutionParameterivEXT = NULL;
+PFNGLGETSEPARABLEFILTEREXTPROC __glewGetSeparableFilterEXT = NULL;
+PFNGLSEPARABLEFILTER2DEXTPROC __glewSeparableFilter2DEXT = NULL;
+
+PFNGLBINORMALPOINTEREXTPROC __glewBinormalPointerEXT = NULL;
+PFNGLTANGENTPOINTEREXTPROC __glewTangentPointerEXT = NULL;
+
+PFNGLCOPYTEXIMAGE1DEXTPROC __glewCopyTexImage1DEXT = NULL;
+PFNGLCOPYTEXIMAGE2DEXTPROC __glewCopyTexImage2DEXT = NULL;
+PFNGLCOPYTEXSUBIMAGE1DEXTPROC __glewCopyTexSubImage1DEXT = NULL;
+PFNGLCOPYTEXSUBIMAGE2DEXTPROC __glewCopyTexSubImage2DEXT = NULL;
+PFNGLCOPYTEXSUBIMAGE3DEXTPROC __glewCopyTexSubImage3DEXT = NULL;
+
+PFNGLCULLPARAMETERDVEXTPROC __glewCullParameterdvEXT = NULL;
+PFNGLCULLPARAMETERFVEXTPROC __glewCullParameterfvEXT = NULL;
+
+PFNGLDEPTHBOUNDSEXTPROC __glewDepthBoundsEXT = NULL;
+
+PFNGLBINDMULTITEXTUREEXTPROC __glewBindMultiTextureEXT = NULL;
+PFNGLCHECKNAMEDFRAMEBUFFERSTATUSEXTPROC __glewCheckNamedFramebufferStatusEXT = NULL;
+PFNGLCLIENTATTRIBDEFAULTEXTPROC __glewClientAttribDefaultEXT = NULL;
+PFNGLCOMPRESSEDMULTITEXIMAGE1DEXTPROC __glewCompressedMultiTexImage1DEXT = NULL;
+PFNGLCOMPRESSEDMULTITEXIMAGE2DEXTPROC __glewCompressedMultiTexImage2DEXT = NULL;
+PFNGLCOMPRESSEDMULTITEXIMAGE3DEXTPROC __glewCompressedMultiTexImage3DEXT = NULL;
+PFNGLCOMPRESSEDMULTITEXSUBIMAGE1DEXTPROC __glewCompressedMultiTexSubImage1DEXT = NULL;
+PFNGLCOMPRESSEDMULTITEXSUBIMAGE2DEXTPROC __glewCompressedMultiTexSubImage2DEXT = NULL;
+PFNGLCOMPRESSEDMULTITEXSUBIMAGE3DEXTPROC __glewCompressedMultiTexSubImage3DEXT = NULL;
+PFNGLCOMPRESSEDTEXTUREIMAGE1DEXTPROC __glewCompressedTextureImage1DEXT = NULL;
+PFNGLCOMPRESSEDTEXTUREIMAGE2DEXTPROC __glewCompressedTextureImage2DEXT = NULL;
+PFNGLCOMPRESSEDTEXTUREIMAGE3DEXTPROC __glewCompressedTextureImage3DEXT = NULL;
+PFNGLCOMPRESSEDTEXTURESUBIMAGE1DEXTPROC __glewCompressedTextureSubImage1DEXT = NULL;
+PFNGLCOMPRESSEDTEXTURESUBIMAGE2DEXTPROC __glewCompressedTextureSubImage2DEXT = NULL;
+PFNGLCOMPRESSEDTEXTURESUBIMAGE3DEXTPROC __glewCompressedTextureSubImage3DEXT = NULL;
+PFNGLCOPYMULTITEXIMAGE1DEXTPROC __glewCopyMultiTexImage1DEXT = NULL;
+PFNGLCOPYMULTITEXIMAGE2DEXTPROC __glewCopyMultiTexImage2DEXT = NULL;
+PFNGLCOPYMULTITEXSUBIMAGE1DEXTPROC __glewCopyMultiTexSubImage1DEXT = NULL;
+PFNGLCOPYMULTITEXSUBIMAGE2DEXTPROC __glewCopyMultiTexSubImage2DEXT = NULL;
+PFNGLCOPYMULTITEXSUBIMAGE3DEXTPROC __glewCopyMultiTexSubImage3DEXT = NULL;
+PFNGLCOPYTEXTUREIMAGE1DEXTPROC __glewCopyTextureImage1DEXT = NULL;
+PFNGLCOPYTEXTUREIMAGE2DEXTPROC __glewCopyTextureImage2DEXT = NULL;
+PFNGLCOPYTEXTURESUBIMAGE1DEXTPROC __glewCopyTextureSubImage1DEXT = NULL;
+PFNGLCOPYTEXTURESUBIMAGE2DEXTPROC __glewCopyTextureSubImage2DEXT = NULL;
+PFNGLCOPYTEXTURESUBIMAGE3DEXTPROC __glewCopyTextureSubImage3DEXT = NULL;
+PFNGLDISABLECLIENTSTATEINDEXEDEXTPROC __glewDisableClientStateIndexedEXT = NULL;
+PFNGLENABLECLIENTSTATEINDEXEDEXTPROC __glewEnableClientStateIndexedEXT = NULL;
+PFNGLFRAMEBUFFERDRAWBUFFEREXTPROC __glewFramebufferDrawBufferEXT = NULL;
+PFNGLFRAMEBUFFERDRAWBUFFERSEXTPROC __glewFramebufferDrawBuffersEXT = NULL;
+PFNGLFRAMEBUFFERREADBUFFEREXTPROC __glewFramebufferReadBufferEXT = NULL;
+PFNGLGENERATEMULTITEXMIPMAPEXTPROC __glewGenerateMultiTexMipmapEXT = NULL;
+PFNGLGENERATETEXTUREMIPMAPEXTPROC __glewGenerateTextureMipmapEXT = NULL;
+PFNGLGETCOMPRESSEDMULTITEXIMAGEEXTPROC __glewGetCompressedMultiTexImageEXT = NULL;
+PFNGLGETCOMPRESSEDTEXTUREIMAGEEXTPROC __glewGetCompressedTextureImageEXT = NULL;
+PFNGLGETDOUBLEINDEXEDVEXTPROC __glewGetDoubleIndexedvEXT = NULL;
+PFNGLGETFLOATINDEXEDVEXTPROC __glewGetFloatIndexedvEXT = NULL;
+PFNGLGETFRAMEBUFFERPARAMETERIVEXTPROC __glewGetFramebufferParameterivEXT = NULL;
+PFNGLGETMULTITEXENVFVEXTPROC __glewGetMultiTexEnvfvEXT = NULL;
+PFNGLGETMULTITEXENVIVEXTPROC __glewGetMultiTexEnvivEXT = NULL;
+PFNGLGETMULTITEXGENDVEXTPROC __glewGetMultiTexGendvEXT = NULL;
+PFNGLGETMULTITEXGENFVEXTPROC __glewGetMultiTexGenfvEXT = NULL;
+PFNGLGETMULTITEXGENIVEXTPROC __glewGetMultiTexGenivEXT = NULL;
+PFNGLGETMULTITEXIMAGEEXTPROC __glewGetMultiTexImageEXT = NULL;
+PFNGLGETMULTITEXLEVELPARAMETERFVEXTPROC __glewGetMultiTexLevelParameterfvEXT = NULL;
+PFNGLGETMULTITEXLEVELPARAMETERIVEXTPROC __glewGetMultiTexLevelParameterivEXT = NULL;
+PFNGLGETMULTITEXPARAMETERIIVEXTPROC __glewGetMultiTexParameterIivEXT = NULL;
+PFNGLGETMULTITEXPARAMETERIUIVEXTPROC __glewGetMultiTexParameterIuivEXT = NULL;
+PFNGLGETMULTITEXPARAMETERFVEXTPROC __glewGetMultiTexParameterfvEXT = NULL;
+PFNGLGETMULTITEXPARAMETERIVEXTPROC __glewGetMultiTexParameterivEXT = NULL;
+PFNGLGETNAMEDBUFFERPARAMETERIVEXTPROC __glewGetNamedBufferParameterivEXT = NULL;
+PFNGLGETNAMEDBUFFERPOINTERVEXTPROC __glewGetNamedBufferPointervEXT = NULL;
+PFNGLGETNAMEDBUFFERSUBDATAEXTPROC __glewGetNamedBufferSubDataEXT = NULL;
+PFNGLGETNAMEDFRAMEBUFFERATTACHMENTPARAMETERIVEXTPROC __glewGetNamedFramebufferAttachmentParameterivEXT = NULL;
+PFNGLGETNAMEDPROGRAMLOCALPARAMETERIIVEXTPROC __glewGetNamedProgramLocalParameterIivEXT = NULL;
+PFNGLGETNAMEDPROGRAMLOCALPARAMETERIUIVEXTPROC __glewGetNamedProgramLocalParameterIuivEXT = NULL;
+PFNGLGETNAMEDPROGRAMLOCALPARAMETERDVEXTPROC __glewGetNamedProgramLocalParameterdvEXT = NULL;
+PFNGLGETNAMEDPROGRAMLOCALPARAMETERFVEXTPROC __glewGetNamedProgramLocalParameterfvEXT = NULL;
+PFNGLGETNAMEDPROGRAMSTRINGEXTPROC __glewGetNamedProgramStringEXT = NULL;
+PFNGLGETNAMEDPROGRAMIVEXTPROC __glewGetNamedProgramivEXT = NULL;
+PFNGLGETNAMEDRENDERBUFFERPARAMETERIVEXTPROC __glewGetNamedRenderbufferParameterivEXT = NULL;
+PFNGLGETPOINTERINDEXEDVEXTPROC __glewGetPointerIndexedvEXT = NULL;
+PFNGLGETTEXTUREIMAGEEXTPROC __glewGetTextureImageEXT = NULL;
+PFNGLGETTEXTURELEVELPARAMETERFVEXTPROC __glewGetTextureLevelParameterfvEXT = NULL;
+PFNGLGETTEXTURELEVELPARAMETERIVEXTPROC __glewGetTextureLevelParameterivEXT = NULL;
+PFNGLGETTEXTUREPARAMETERIIVEXTPROC __glewGetTextureParameterIivEXT = NULL;
+PFNGLGETTEXTUREPARAMETERIUIVEXTPROC __glewGetTextureParameterIuivEXT = NULL;
+PFNGLGETTEXTUREPARAMETERFVEXTPROC __glewGetTextureParameterfvEXT = NULL;
+PFNGLGETTEXTUREPARAMETERIVEXTPROC __glewGetTextureParameterivEXT = NULL;
+PFNGLMAPNAMEDBUFFEREXTPROC __glewMapNamedBufferEXT = NULL;
+PFNGLMATRIXFRUSTUMEXTPROC __glewMatrixFrustumEXT = NULL;
+PFNGLMATRIXLOADIDENTITYEXTPROC __glewMatrixLoadIdentityEXT = NULL;
+PFNGLMATRIXLOADTRANSPOSEDEXTPROC __glewMatrixLoadTransposedEXT = NULL;
+PFNGLMATRIXLOADTRANSPOSEFEXTPROC __glewMatrixLoadTransposefEXT = NULL;
+PFNGLMATRIXLOADDEXTPROC __glewMatrixLoaddEXT = NULL;
+PFNGLMATRIXLOADFEXTPROC __glewMatrixLoadfEXT = NULL;
+PFNGLMATRIXMULTTRANSPOSEDEXTPROC __glewMatrixMultTransposedEXT = NULL;
+PFNGLMATRIXMULTTRANSPOSEFEXTPROC __glewMatrixMultTransposefEXT = NULL;
+PFNGLMATRIXMULTDEXTPROC __glewMatrixMultdEXT = NULL;
+PFNGLMATRIXMULTFEXTPROC __glewMatrixMultfEXT = NULL;
+PFNGLMATRIXORTHOEXTPROC __glewMatrixOrthoEXT = NULL;
+PFNGLMATRIXPOPEXTPROC __glewMatrixPopEXT = NULL;
+PFNGLMATRIXPUSHEXTPROC __glewMatrixPushEXT = NULL;
+PFNGLMATRIXROTATEDEXTPROC __glewMatrixRotatedEXT = NULL;
+PFNGLMATRIXROTATEFEXTPROC __glewMatrixRotatefEXT = NULL;
+PFNGLMATRIXSCALEDEXTPROC __glewMatrixScaledEXT = NULL;
+PFNGLMATRIXSCALEFEXTPROC __glewMatrixScalefEXT = NULL;
+PFNGLMATRIXTRANSLATEDEXTPROC __glewMatrixTranslatedEXT = NULL;
+PFNGLMATRIXTRANSLATEFEXTPROC __glewMatrixTranslatefEXT = NULL;
+PFNGLMULTITEXBUFFEREXTPROC __glewMultiTexBufferEXT = NULL;
+PFNGLMULTITEXCOORDPOINTEREXTPROC __glewMultiTexCoordPointerEXT = NULL;
+PFNGLMULTITEXENVFEXTPROC __glewMultiTexEnvfEXT = NULL;
+PFNGLMULTITEXENVFVEXTPROC __glewMultiTexEnvfvEXT = NULL;
+PFNGLMULTITEXENVIEXTPROC __glewMultiTexEnviEXT = NULL;
+PFNGLMULTITEXENVIVEXTPROC __glewMultiTexEnvivEXT = NULL;
+PFNGLMULTITEXGENDEXTPROC __glewMultiTexGendEXT = NULL;
+PFNGLMULTITEXGENDVEXTPROC __glewMultiTexGendvEXT = NULL;
+PFNGLMULTITEXGENFEXTPROC __glewMultiTexGenfEXT = NULL;
+PFNGLMULTITEXGENFVEXTPROC __glewMultiTexGenfvEXT = NULL;
+PFNGLMULTITEXGENIEXTPROC __glewMultiTexGeniEXT = NULL;
+PFNGLMULTITEXGENIVEXTPROC __glewMultiTexGenivEXT = NULL;
+PFNGLMULTITEXIMAGE1DEXTPROC __glewMultiTexImage1DEXT = NULL;
+PFNGLMULTITEXIMAGE2DEXTPROC __glewMultiTexImage2DEXT = NULL;
+PFNGLMULTITEXIMAGE3DEXTPROC __glewMultiTexImage3DEXT = NULL;
+PFNGLMULTITEXPARAMETERIIVEXTPROC __glewMultiTexParameterIivEXT = NULL;
+PFNGLMULTITEXPARAMETERIUIVEXTPROC __glewMultiTexParameterIuivEXT = NULL;
+PFNGLMULTITEXPARAMETERFEXTPROC __glewMultiTexParameterfEXT = NULL;
+PFNGLMULTITEXPARAMETERFVEXTPROC __glewMultiTexParameterfvEXT = NULL;
+PFNGLMULTITEXPARAMETERIEXTPROC __glewMultiTexParameteriEXT = NULL;
+PFNGLMULTITEXPARAMETERIVEXTPROC __glewMultiTexParameterivEXT = NULL;
+PFNGLMULTITEXRENDERBUFFEREXTPROC __glewMultiTexRenderbufferEXT = NULL;
+PFNGLMULTITEXSUBIMAGE1DEXTPROC __glewMultiTexSubImage1DEXT = NULL;
+PFNGLMULTITEXSUBIMAGE2DEXTPROC __glewMultiTexSubImage2DEXT = NULL;
+PFNGLMULTITEXSUBIMAGE3DEXTPROC __glewMultiTexSubImage3DEXT = NULL;
+PFNGLNAMEDBUFFERDATAEXTPROC __glewNamedBufferDataEXT = NULL;
+PFNGLNAMEDBUFFERSUBDATAEXTPROC __glewNamedBufferSubDataEXT = NULL;
+PFNGLNAMEDFRAMEBUFFERRENDERBUFFEREXTPROC __glewNamedFramebufferRenderbufferEXT = NULL;
+PFNGLNAMEDFRAMEBUFFERTEXTURE1DEXTPROC __glewNamedFramebufferTexture1DEXT = NULL;
+PFNGLNAMEDFRAMEBUFFERTEXTURE2DEXTPROC __glewNamedFramebufferTexture2DEXT = NULL;
+PFNGLNAMEDFRAMEBUFFERTEXTURE3DEXTPROC __glewNamedFramebufferTexture3DEXT = NULL;
+PFNGLNAMEDFRAMEBUFFERTEXTUREEXTPROC __glewNamedFramebufferTextureEXT = NULL;
+PFNGLNAMEDFRAMEBUFFERTEXTUREFACEEXTPROC __glewNamedFramebufferTextureFaceEXT = NULL;
+PFNGLNAMEDFRAMEBUFFERTEXTURELAYEREXTPROC __glewNamedFramebufferTextureLayerEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETER4DEXTPROC __glewNamedProgramLocalParameter4dEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETER4DVEXTPROC __glewNamedProgramLocalParameter4dvEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETER4FEXTPROC __glewNamedProgramLocalParameter4fEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETER4FVEXTPROC __glewNamedProgramLocalParameter4fvEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETERI4IEXTPROC __glewNamedProgramLocalParameterI4iEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETERI4IVEXTPROC __glewNamedProgramLocalParameterI4ivEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETERI4UIEXTPROC __glewNamedProgramLocalParameterI4uiEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETERI4UIVEXTPROC __glewNamedProgramLocalParameterI4uivEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETERS4FVEXTPROC __glewNamedProgramLocalParameters4fvEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETERSI4IVEXTPROC __glewNamedProgramLocalParametersI4ivEXT = NULL;
+PFNGLNAMEDPROGRAMLOCALPARAMETERSI4UIVEXTPROC __glewNamedProgramLocalParametersI4uivEXT = NULL;
+PFNGLNAMEDPROGRAMSTRINGEXTPROC __glewNamedProgramStringEXT = NULL;
+PFNGLNAMEDRENDERBUFFERSTORAGEEXTPROC __glewNamedRenderbufferStorageEXT = NULL;
+PFNGLNAMEDRENDERBUFFERSTORAGEMULTISAMPLECOVERAGEEXTPROC __glewNamedRenderbufferStorageMultisampleCoverageEXT = NULL;
+PFNGLNAMEDRENDERBUFFERSTORAGEMULTISAMPLEEXTPROC __glewNamedRenderbufferStorageMultisampleEXT = NULL;
+PFNGLPROGRAMUNIFORM1FEXTPROC __glewProgramUniform1fEXT = NULL;
+PFNGLPROGRAMUNIFORM1FVEXTPROC __glewProgramUniform1fvEXT = NULL;
+PFNGLPROGRAMUNIFORM1IEXTPROC __glewProgramUniform1iEXT = NULL;
+PFNGLPROGRAMUNIFORM1IVEXTPROC __glewProgramUniform1ivEXT = NULL;
+PFNGLPROGRAMUNIFORM1UIEXTPROC __glewProgramUniform1uiEXT = NULL;
+PFNGLPROGRAMUNIFORM1UIVEXTPROC __glewProgramUniform1uivEXT = NULL;
+PFNGLPROGRAMUNIFORM2FEXTPROC __glewProgramUniform2fEXT = NULL;
+PFNGLPROGRAMUNIFORM2FVEXTPROC __glewProgramUniform2fvEXT = NULL;
+PFNGLPROGRAMUNIFORM2IEXTPROC __glewProgramUniform2iEXT = NULL;
+PFNGLPROGRAMUNIFORM2IVEXTPROC __glewProgramUniform2ivEXT = NULL;
+PFNGLPROGRAMUNIFORM2UIEXTPROC __glewProgramUniform2uiEXT = NULL;
+PFNGLPROGRAMUNIFORM2UIVEXTPROC __glewProgramUniform2uivEXT = NULL;
+PFNGLPROGRAMUNIFORM3FEXTPROC __glewProgramUniform3fEXT = NULL;
+PFNGLPROGRAMUNIFORM3FVEXTPROC __glewProgramUniform3fvEXT = NULL;
+PFNGLPROGRAMUNIFORM3IEXTPROC __glewProgramUniform3iEXT = NULL;
+PFNGLPROGRAMUNIFORM3IVEXTPROC __glewProgramUniform3ivEXT = NULL;
+PFNGLPROGRAMUNIFORM3UIEXTPROC __glewProgramUniform3uiEXT = NULL;
+PFNGLPROGRAMUNIFORM3UIVEXTPROC __glewProgramUniform3uivEXT = NULL;
+PFNGLPROGRAMUNIFORM4FEXTPROC __glewProgramUniform4fEXT = NULL;
+PFNGLPROGRAMUNIFORM4FVEXTPROC __glewProgramUniform4fvEXT = NULL;
+PFNGLPROGRAMUNIFORM4IEXTPROC __glewProgramUniform4iEXT = NULL;
+PFNGLPROGRAMUNIFORM4IVEXTPROC __glewProgramUniform4ivEXT = NULL;
+PFNGLPROGRAMUNIFORM4UIEXTPROC __glewProgramUniform4uiEXT = NULL;
+PFNGLPROGRAMUNIFORM4UIVEXTPROC __glewProgramUniform4uivEXT = NULL;
+PFNGLPROGRAMUNIFORMMATRIX2FVEXTPROC __glewProgramUniformMatrix2fvEXT = NULL;
+PFNGLPROGRAMUNIFORMMATRIX2X3FVEXTPROC __glewProgramUniformMatrix2x3fvEXT = NULL;
+PFNGLPROGRAMUNIFORMMATRIX2X4FVEXTPROC __glewProgramUniformMatrix2x4fvEXT = NULL;
+PFNGLPROGRAMUNIFORMMATRIX3FVEXTPROC __glewProgramUniformMatrix3fvEXT = NULL;
+PFNGLPROGRAMUNIFORMMATRIX3X2FVEXTPROC __glewProgramUniformMatrix3x2fvEXT = NULL;
+PFNGLPROGRAMUNIFORMMATRIX3X4FVEXTPROC __glewProgramUniformMatrix3x4fvEXT = NULL;
+PFNGLPROGRAMUNIFORMMATRIX4FVEXTPROC __glewProgramUniformMatrix4fvEXT = NULL;
+PFNGLPROGRAMUNIFORMMATRIX4X2FVEXTPROC __glewProgramUniformMatrix4x2fvEXT = NULL;
+PFNGLPROGRAMUNIFORMMATRIX4X3FVEXTPROC __glewProgramUniformMatrix4x3fvEXT = NULL;
+PFNGLPUSHCLIENTATTRIBDEFAULTEXTPROC __glewPushClientAttribDefaultEXT = NULL;
+PFNGLTEXTUREBUFFEREXTPROC __glewTextureBufferEXT = NULL;
+PFNGLTEXTUREIMAGE1DEXTPROC __glewTextureImage1DEXT = NULL;
+PFNGLTEXTUREIMAGE2DEXTPROC __glewTextureImage2DEXT = NULL;
+PFNGLTEXTUREIMAGE3DEXTPROC __glewTextureImage3DEXT = NULL;
+PFNGLTEXTUREPARAMETERIIVEXTPROC __glewTextureParameterIivEXT = NULL;
+PFNGLTEXTUREPARAMETERIUIVEXTPROC __glewTextureParameterIuivEXT = NULL;
+PFNGLTEXTUREPARAMETERFEXTPROC __glewTextureParameterfEXT = NULL;
+PFNGLTEXTUREPARAMETERFVEXTPROC __glewTextureParameterfvEXT = NULL;
+PFNGLTEXTUREPARAMETERIEXTPROC __glewTextureParameteriEXT = NULL;
+PFNGLTEXTUREPARAMETERIVEXTPROC __glewTextureParameterivEXT = NULL;
+PFNGLTEXTURERENDERBUFFEREXTPROC __glewTextureRenderbufferEXT = NULL;
+PFNGLTEXTURESUBIMAGE1DEXTPROC __glewTextureSubImage1DEXT = NULL;
+PFNGLTEXTURESUBIMAGE2DEXTPROC __glewTextureSubImage2DEXT = NULL;
+PFNGLTEXTURESUBIMAGE3DEXTPROC __glewTextureSubImage3DEXT = NULL;
+PFNGLUNMAPNAMEDBUFFEREXTPROC __glewUnmapNamedBufferEXT = NULL;
+
+PFNGLCOLORMASKINDEXEDEXTPROC __glewColorMaskIndexedEXT = NULL;
+PFNGLDISABLEINDEXEDEXTPROC __glewDisableIndexedEXT = NULL;
+PFNGLENABLEINDEXEDEXTPROC __glewEnableIndexedEXT = NULL;
+PFNGLGETBOOLEANINDEXEDVEXTPROC __glewGetBooleanIndexedvEXT = NULL;
+PFNGLGETINTEGERINDEXEDVEXTPROC __glewGetIntegerIndexedvEXT = NULL;
+PFNGLISENABLEDINDEXEDEXTPROC __glewIsEnabledIndexedEXT = NULL;
+
+PFNGLDRAWARRAYSINSTANCEDEXTPROC __glewDrawArraysInstancedEXT = NULL;
+PFNGLDRAWELEMENTSINSTANCEDEXTPROC __glewDrawElementsInstancedEXT = NULL;
+
+PFNGLDRAWRANGEELEMENTSEXTPROC __glewDrawRangeElementsEXT = NULL;
+
+PFNGLFOGCOORDPOINTEREXTPROC __glewFogCoordPointerEXT = NULL;
+PFNGLFOGCOORDDEXTPROC __glewFogCoorddEXT = NULL;
+PFNGLFOGCOORDDVEXTPROC __glewFogCoorddvEXT = NULL;
+PFNGLFOGCOORDFEXTPROC __glewFogCoordfEXT = NULL;
+PFNGLFOGCOORDFVEXTPROC __glewFogCoordfvEXT = NULL;
+
+PFNGLFRAGMENTCOLORMATERIALEXTPROC __glewFragmentColorMaterialEXT = NULL;
+PFNGLFRAGMENTLIGHTMODELFEXTPROC __glewFragmentLightModelfEXT = NULL;
+PFNGLFRAGMENTLIGHTMODELFVEXTPROC __glewFragmentLightModelfvEXT = NULL;
+PFNGLFRAGMENTLIGHTMODELIEXTPROC __glewFragmentLightModeliEXT = NULL;
+PFNGLFRAGMENTLIGHTMODELIVEXTPROC __glewFragmentLightModelivEXT = NULL;
+PFNGLFRAGMENTLIGHTFEXTPROC __glewFragmentLightfEXT = NULL;
+PFNGLFRAGMENTLIGHTFVEXTPROC __glewFragmentLightfvEXT = NULL;
+PFNGLFRAGMENTLIGHTIEXTPROC __glewFragmentLightiEXT = NULL;
+PFNGLFRAGMENTLIGHTIVEXTPROC __glewFragmentLightivEXT = NULL;
+PFNGLFRAGMENTMATERIALFEXTPROC __glewFragmentMaterialfEXT = NULL;
+PFNGLFRAGMENTMATERIALFVEXTPROC __glewFragmentMaterialfvEXT = NULL;
+PFNGLFRAGMENTMATERIALIEXTPROC __glewFragmentMaterialiEXT = NULL;
+PFNGLFRAGMENTMATERIALIVEXTPROC __glewFragmentMaterialivEXT = NULL;
+PFNGLGETFRAGMENTLIGHTFVEXTPROC __glewGetFragmentLightfvEXT = NULL;
+PFNGLGETFRAGMENTLIGHTIVEXTPROC __glewGetFragmentLightivEXT = NULL;
+PFNGLGETFRAGMENTMATERIALFVEXTPROC __glewGetFragmentMaterialfvEXT = NULL;
+PFNGLGETFRAGMENTMATERIALIVEXTPROC __glewGetFragmentMaterialivEXT = NULL;
+PFNGLLIGHTENVIEXTPROC __glewLightEnviEXT = NULL;
+
+PFNGLBLITFRAMEBUFFEREXTPROC __glewBlitFramebufferEXT = NULL;
+
+PFNGLRENDERBUFFERSTORAGEMULTISAMPLEEXTPROC __glewRenderbufferStorageMultisampleEXT = NULL;
+
+PFNGLBINDFRAMEBUFFEREXTPROC __glewBindFramebufferEXT = NULL;
+PFNGLBINDRENDERBUFFEREXTPROC __glewBindRenderbufferEXT = NULL;
+PFNGLCHECKFRAMEBUFFERSTATUSEXTPROC __glewCheckFramebufferStatusEXT = NULL;
+PFNGLDELETEFRAMEBUFFERSEXTPROC __glewDeleteFramebuffersEXT = NULL;
+PFNGLDELETERENDERBUFFERSEXTPROC __glewDeleteRenderbuffersEXT = NULL;
+PFNGLFRAMEBUFFERRENDERBUFFEREXTPROC __glewFramebufferRenderbufferEXT = NULL;
+PFNGLFRAMEBUFFERTEXTURE1DEXTPROC __glewFramebufferTexture1DEXT = NULL;
+PFNGLFRAMEBUFFERTEXTURE2DEXTPROC __glewFramebufferTexture2DEXT = NULL;
+PFNGLFRAMEBUFFERTEXTURE3DEXTPROC __glewFramebufferTexture3DEXT = NULL;
+PFNGLGENFRAMEBUFFERSEXTPROC __glewGenFramebuffersEXT = NULL;
+PFNGLGENRENDERBUFFERSEXTPROC __glewGenRenderbuffersEXT = NULL;
+PFNGLGENERATEMIPMAPEXTPROC __glewGenerateMipmapEXT = NULL;
+PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVEXTPROC __glewGetFramebufferAttachmentParameterivEXT = NULL;
+PFNGLGETRENDERBUFFERPARAMETERIVEXTPROC __glewGetRenderbufferParameterivEXT = NULL;
+PFNGLISFRAMEBUFFEREXTPROC __glewIsFramebufferEXT = NULL;
+PFNGLISRENDERBUFFEREXTPROC __glewIsRenderbufferEXT = NULL;
+PFNGLRENDERBUFFERSTORAGEEXTPROC __glewRenderbufferStorageEXT = NULL;
+
+PFNGLFRAMEBUFFERTEXTUREEXTPROC __glewFramebufferTextureEXT = NULL;
+PFNGLFRAMEBUFFERTEXTUREFACEEXTPROC __glewFramebufferTextureFaceEXT = NULL;
+PFNGLFRAMEBUFFERTEXTURELAYEREXTPROC __glewFramebufferTextureLayerEXT = NULL;
+PFNGLPROGRAMPARAMETERIEXTPROC __glewProgramParameteriEXT = NULL;
+
+PFNGLPROGRAMENVPARAMETERS4FVEXTPROC __glewProgramEnvParameters4fvEXT = NULL;
+PFNGLPROGRAMLOCALPARAMETERS4FVEXTPROC __glewProgramLocalParameters4fvEXT = NULL;
+
+PFNGLBINDFRAGDATALOCATIONEXTPROC __glewBindFragDataLocationEXT = NULL;
+PFNGLGETFRAGDATALOCATIONEXTPROC __glewGetFragDataLocationEXT = NULL;
+PFNGLGETUNIFORMUIVEXTPROC __glewGetUniformuivEXT = NULL;
+PFNGLGETVERTEXATTRIBIIVEXTPROC __glewGetVertexAttribIivEXT = NULL;
+PFNGLGETVERTEXATTRIBIUIVEXTPROC __glewGetVertexAttribIuivEXT = NULL;
+PFNGLUNIFORM1UIEXTPROC __glewUniform1uiEXT = NULL;
+PFNGLUNIFORM1UIVEXTPROC __glewUniform1uivEXT = NULL;
+PFNGLUNIFORM2UIEXTPROC __glewUniform2uiEXT = NULL;
+PFNGLUNIFORM2UIVEXTPROC __glewUniform2uivEXT = NULL;
+PFNGLUNIFORM3UIEXTPROC __glewUniform3uiEXT = NULL;
+PFNGLUNIFORM3UIVEXTPROC __glewUniform3uivEXT = NULL;
+PFNGLUNIFORM4UIEXTPROC __glewUniform4uiEXT = NULL;
+PFNGLUNIFORM4UIVEXTPROC __glewUniform4uivEXT = NULL;
+PFNGLVERTEXATTRIBI1IEXTPROC __glewVertexAttribI1iEXT = NULL;
+PFNGLVERTEXATTRIBI1IVEXTPROC __glewVertexAttribI1ivEXT = NULL;
+PFNGLVERTEXATTRIBI1UIEXTPROC __glewVertexAttribI1uiEXT = NULL;
+PFNGLVERTEXATTRIBI1UIVEXTPROC __glewVertexAttribI1uivEXT = NULL;
+PFNGLVERTEXATTRIBI2IEXTPROC __glewVertexAttribI2iEXT = NULL;
+PFNGLVERTEXATTRIBI2IVEXTPROC __glewVertexAttribI2ivEXT = NULL;
+PFNGLVERTEXATTRIBI2UIEXTPROC __glewVertexAttribI2uiEXT = NULL;
+PFNGLVERTEXATTRIBI2UIVEXTPROC __glewVertexAttribI2uivEXT = NULL;
+PFNGLVERTEXATTRIBI3IEXTPROC __glewVertexAttribI3iEXT = NULL;
+PFNGLVERTEXATTRIBI3IVEXTPROC __glewVertexAttribI3ivEXT = NULL;
+PFNGLVERTEXATTRIBI3UIEXTPROC __glewVertexAttribI3uiEXT = NULL;
+PFNGLVERTEXATTRIBI3UIVEXTPROC __glewVertexAttribI3uivEXT = NULL;
+PFNGLVERTEXATTRIBI4BVEXTPROC __glewVertexAttribI4bvEXT = NULL;
+PFNGLVERTEXATTRIBI4IEXTPROC __glewVertexAttribI4iEXT = NULL;
+PFNGLVERTEXATTRIBI4IVEXTPROC __glewVertexAttribI4ivEXT = NULL;
+PFNGLVERTEXATTRIBI4SVEXTPROC __glewVertexAttribI4svEXT = NULL;
+PFNGLVERTEXATTRIBI4UBVEXTPROC __glewVertexAttribI4ubvEXT = NULL;
+PFNGLVERTEXATTRIBI4UIEXTPROC __glewVertexAttribI4uiEXT = NULL;
+PFNGLVERTEXATTRIBI4UIVEXTPROC __glewVertexAttribI4uivEXT = NULL;
+PFNGLVERTEXATTRIBI4USVEXTPROC __glewVertexAttribI4usvEXT = NULL;
+PFNGLVERTEXATTRIBIPOINTEREXTPROC __glewVertexAttribIPointerEXT = NULL;
+
+PFNGLGETHISTOGRAMEXTPROC __glewGetHistogramEXT = NULL;
+PFNGLGETHISTOGRAMPARAMETERFVEXTPROC __glewGetHistogramParameterfvEXT = NULL;
+PFNGLGETHISTOGRAMPARAMETERIVEXTPROC __glewGetHistogramParameterivEXT = NULL;
+PFNGLGETMINMAXEXTPROC __glewGetMinmaxEXT = NULL;
+PFNGLGETMINMAXPARAMETERFVEXTPROC __glewGetMinmaxParameterfvEXT = NULL;
+PFNGLGETMINMAXPARAMETERIVEXTPROC __glewGetMinmaxParameterivEXT = NULL;
+PFNGLHISTOGRAMEXTPROC __glewHistogramEXT = NULL;
+PFNGLMINMAXEXTPROC __glewMinmaxEXT = NULL;
+PFNGLRESETHISTOGRAMEXTPROC __glewResetHistogramEXT = NULL;
+PFNGLRESETMINMAXEXTPROC __glewResetMinmaxEXT = NULL;
+
+PFNGLINDEXFUNCEXTPROC __glewIndexFuncEXT = NULL;
+
+PFNGLINDEXMATERIALEXTPROC __glewIndexMaterialEXT = NULL;
+
+PFNGLAPPLYTEXTUREEXTPROC __glewApplyTextureEXT = NULL;
+PFNGLTEXTURELIGHTEXTPROC __glewTextureLightEXT = NULL;
+PFNGLTEXTUREMATERIALEXTPROC __glewTextureMaterialEXT = NULL;
+
+PFNGLMULTIDRAWARRAYSEXTPROC __glewMultiDrawArraysEXT = NULL;
+PFNGLMULTIDRAWELEMENTSEXTPROC __glewMultiDrawElementsEXT = NULL;
+
+PFNGLSAMPLEMASKEXTPROC __glewSampleMaskEXT = NULL;
+PFNGLSAMPLEPATTERNEXTPROC __glewSamplePatternEXT = NULL;
+
+PFNGLCOLORTABLEEXTPROC __glewColorTableEXT = NULL;
+PFNGLGETCOLORTABLEEXTPROC __glewGetColorTableEXT = NULL;
+PFNGLGETCOLORTABLEPARAMETERFVEXTPROC __glewGetColorTableParameterfvEXT = NULL;
+PFNGLGETCOLORTABLEPARAMETERIVEXTPROC __glewGetColorTableParameterivEXT = NULL;
+
+PFNGLGETPIXELTRANSFORMPARAMETERFVEXTPROC __glewGetPixelTransformParameterfvEXT = NULL;
+PFNGLGETPIXELTRANSFORMPARAMETERIVEXTPROC __glewGetPixelTransformParameterivEXT = NULL;
+PFNGLPIXELTRANSFORMPARAMETERFEXTPROC __glewPixelTransformParameterfEXT = NULL;
+PFNGLPIXELTRANSFORMPARAMETERFVEXTPROC __glewPixelTransformParameterfvEXT = NULL;
+PFNGLPIXELTRANSFORMPARAMETERIEXTPROC __glewPixelTransformParameteriEXT = NULL;
+PFNGLPIXELTRANSFORMPARAMETERIVEXTPROC __glewPixelTransformParameterivEXT = NULL;
+
+PFNGLPOINTPARAMETERFEXTPROC __glewPointParameterfEXT = NULL;
+PFNGLPOINTPARAMETERFVEXTPROC __glewPointParameterfvEXT = NULL;
+
+PFNGLPOLYGONOFFSETEXTPROC __glewPolygonOffsetEXT = NULL;
+
+PFNGLBEGINSCENEEXTPROC __glewBeginSceneEXT = NULL;
+PFNGLENDSCENEEXTPROC __glewEndSceneEXT = NULL;
+
+PFNGLSECONDARYCOLOR3BEXTPROC __glewSecondaryColor3bEXT = NULL;
+PFNGLSECONDARYCOLOR3BVEXTPROC __glewSecondaryColor3bvEXT = NULL;
+PFNGLSECONDARYCOLOR3DEXTPROC __glewSecondaryColor3dEXT = NULL;
+PFNGLSECONDARYCOLOR3DVEXTPROC __glewSecondaryColor3dvEXT = NULL;
+PFNGLSECONDARYCOLOR3FEXTPROC __glewSecondaryColor3fEXT = NULL;
+PFNGLSECONDARYCOLOR3FVEXTPROC __glewSecondaryColor3fvEXT = NULL;
+PFNGLSECONDARYCOLOR3IEXTPROC __glewSecondaryColor3iEXT = NULL;
+PFNGLSECONDARYCOLOR3IVEXTPROC __glewSecondaryColor3ivEXT = NULL;
+PFNGLSECONDARYCOLOR3SEXTPROC __glewSecondaryColor3sEXT = NULL;
+PFNGLSECONDARYCOLOR3SVEXTPROC __glewSecondaryColor3svEXT = NULL;
+PFNGLSECONDARYCOLOR3UBEXTPROC __glewSecondaryColor3ubEXT = NULL;
+PFNGLSECONDARYCOLOR3UBVEXTPROC __glewSecondaryColor3ubvEXT = NULL;
+PFNGLSECONDARYCOLOR3UIEXTPROC __glewSecondaryColor3uiEXT = NULL;
+PFNGLSECONDARYCOLOR3UIVEXTPROC __glewSecondaryColor3uivEXT = NULL;
+PFNGLSECONDARYCOLOR3USEXTPROC __glewSecondaryColor3usEXT = NULL;
+PFNGLSECONDARYCOLOR3USVEXTPROC __glewSecondaryColor3usvEXT = NULL;
+PFNGLSECONDARYCOLORPOINTEREXTPROC __glewSecondaryColorPointerEXT = NULL;
+
+PFNGLACTIVESTENCILFACEEXTPROC __glewActiveStencilFaceEXT = NULL;
+
+PFNGLTEXSUBIMAGE1DEXTPROC __glewTexSubImage1DEXT = NULL;
+PFNGLTEXSUBIMAGE2DEXTPROC __glewTexSubImage2DEXT = NULL;
+PFNGLTEXSUBIMAGE3DEXTPROC __glewTexSubImage3DEXT = NULL;
+
+PFNGLTEXIMAGE3DEXTPROC __glewTexImage3DEXT = NULL;
+
+PFNGLTEXBUFFEREXTPROC __glewTexBufferEXT = NULL;
+
+PFNGLCLEARCOLORIIEXTPROC __glewClearColorIiEXT = NULL;
+PFNGLCLEARCOLORIUIEXTPROC __glewClearColorIuiEXT = NULL;
+PFNGLGETTEXPARAMETERIIVEXTPROC __glewGetTexParameterIivEXT = NULL;
+PFNGLGETTEXPARAMETERIUIVEXTPROC __glewGetTexParameterIuivEXT = NULL;
+PFNGLTEXPARAMETERIIVEXTPROC __glewTexParameterIivEXT = NULL;
+PFNGLTEXPARAMETERIUIVEXTPROC __glewTexParameterIuivEXT = NULL;
+
+PFNGLARETEXTURESRESIDENTEXTPROC __glewAreTexturesResidentEXT = NULL;
+PFNGLBINDTEXTUREEXTPROC __glewBindTextureEXT = NULL;
+PFNGLDELETETEXTURESEXTPROC __glewDeleteTexturesEXT = NULL;
+PFNGLGENTEXTURESEXTPROC __glewGenTexturesEXT = NULL;
+PFNGLISTEXTUREEXTPROC __glewIsTextureEXT = NULL;
+PFNGLPRIORITIZETEXTURESEXTPROC __glewPrioritizeTexturesEXT = NULL;
+
+PFNGLTEXTURENORMALEXTPROC __glewTextureNormalEXT = NULL;
+
+PFNGLGETQUERYOBJECTI64VEXTPROC __glewGetQueryObjecti64vEXT = NULL;
+PFNGLGETQUERYOBJECTUI64VEXTPROC __glewGetQueryObjectui64vEXT = NULL;
+
+PFNGLBEGINTRANSFORMFEEDBACKEXTPROC __glewBeginTransformFeedbackEXT = NULL;
+PFNGLBINDBUFFERBASEEXTPROC __glewBindBufferBaseEXT = NULL;
+PFNGLBINDBUFFEROFFSETEXTPROC __glewBindBufferOffsetEXT = NULL;
+PFNGLBINDBUFFERRANGEEXTPROC __glewBindBufferRangeEXT = NULL;
+PFNGLENDTRANSFORMFEEDBACKEXTPROC __glewEndTransformFeedbackEXT = NULL;
+PFNGLGETTRANSFORMFEEDBACKVARYINGEXTPROC __glewGetTransformFeedbackVaryingEXT = NULL;
+PFNGLTRANSFORMFEEDBACKVARYINGSEXTPROC __glewTransformFeedbackVaryingsEXT = NULL;
+
+PFNGLARRAYELEMENTEXTPROC __glewArrayElementEXT = NULL;
+PFNGLCOLORPOINTEREXTPROC __glewColorPointerEXT = NULL;
+PFNGLDRAWARRAYSEXTPROC __glewDrawArraysEXT = NULL;
+PFNGLEDGEFLAGPOINTEREXTPROC __glewEdgeFlagPointerEXT = NULL;
+PFNGLGETPOINTERVEXTPROC __glewGetPointervEXT = NULL;
+PFNGLINDEXPOINTEREXTPROC __glewIndexPointerEXT = NULL;
+PFNGLNORMALPOINTEREXTPROC __glewNormalPointerEXT = NULL;
+PFNGLTEXCOORDPOINTEREXTPROC __glewTexCoordPointerEXT = NULL;
+PFNGLVERTEXPOINTEREXTPROC __glewVertexPointerEXT = NULL;
+
+PFNGLBEGINVERTEXSHADEREXTPROC __glewBeginVertexShaderEXT = NULL;
+PFNGLBINDLIGHTPARAMETEREXTPROC __glewBindLightParameterEXT = NULL;
+PFNGLBINDMATERIALPARAMETEREXTPROC __glewBindMaterialParameterEXT = NULL;
+PFNGLBINDPARAMETEREXTPROC __glewBindParameterEXT = NULL;
+PFNGLBINDTEXGENPARAMETEREXTPROC __glewBindTexGenParameterEXT = NULL;
+PFNGLBINDTEXTUREUNITPARAMETEREXTPROC __glewBindTextureUnitParameterEXT = NULL;
+PFNGLBINDVERTEXSHADEREXTPROC __glewBindVertexShaderEXT = NULL;
+PFNGLDELETEVERTEXSHADEREXTPROC __glewDeleteVertexShaderEXT = NULL;
+PFNGLDISABLEVARIANTCLIENTSTATEEXTPROC __glewDisableVariantClientStateEXT = NULL;
+PFNGLENABLEVARIANTCLIENTSTATEEXTPROC __glewEnableVariantClientStateEXT = NULL;
+PFNGLENDVERTEXSHADEREXTPROC __glewEndVertexShaderEXT = NULL;
+PFNGLEXTRACTCOMPONENTEXTPROC __glewExtractComponentEXT = NULL;
+PFNGLGENSYMBOLSEXTPROC __glewGenSymbolsEXT = NULL;
+PFNGLGENVERTEXSHADERSEXTPROC __glewGenVertexShadersEXT = NULL;
+PFNGLGETINVARIANTBOOLEANVEXTPROC __glewGetInvariantBooleanvEXT = NULL;
+PFNGLGETINVARIANTFLOATVEXTPROC __glewGetInvariantFloatvEXT = NULL;
+PFNGLGETINVARIANTINTEGERVEXTPROC __glewGetInvariantIntegervEXT = NULL;
+PFNGLGETLOCALCONSTANTBOOLEANVEXTPROC __glewGetLocalConstantBooleanvEXT = NULL;
+PFNGLGETLOCALCONSTANTFLOATVEXTPROC __glewGetLocalConstantFloatvEXT = NULL;
+PFNGLGETLOCALCONSTANTINTEGERVEXTPROC __glewGetLocalConstantIntegervEXT = NULL;
+PFNGLGETVARIANTBOOLEANVEXTPROC __glewGetVariantBooleanvEXT = NULL;
+PFNGLGETVARIANTFLOATVEXTPROC __glewGetVariantFloatvEXT = NULL;
+PFNGLGETVARIANTINTEGERVEXTPROC __glewGetVariantIntegervEXT = NULL;
+PFNGLGETVARIANTPOINTERVEXTPROC __glewGetVariantPointervEXT = NULL;
+PFNGLINSERTCOMPONENTEXTPROC __glewInsertComponentEXT = NULL;
+PFNGLISVARIANTENABLEDEXTPROC __glewIsVariantEnabledEXT = NULL;
+PFNGLSETINVARIANTEXTPROC __glewSetInvariantEXT = NULL;
+PFNGLSETLOCALCONSTANTEXTPROC __glewSetLocalConstantEXT = NULL;
+PFNGLSHADEROP1EXTPROC __glewShaderOp1EXT = NULL;
+PFNGLSHADEROP2EXTPROC __glewShaderOp2EXT = NULL;
+PFNGLSHADEROP3EXTPROC __glewShaderOp3EXT = NULL;
+PFNGLSWIZZLEEXTPROC __glewSwizzleEXT = NULL;
+PFNGLVARIANTPOINTEREXTPROC __glewVariantPointerEXT = NULL;
+PFNGLVARIANTBVEXTPROC __glewVariantbvEXT = NULL;
+PFNGLVARIANTDVEXTPROC __glewVariantdvEXT = NULL;
+PFNGLVARIANTFVEXTPROC __glewVariantfvEXT = NULL;
+PFNGLVARIANTIVEXTPROC __glewVariantivEXT = NULL;
+PFNGLVARIANTSVEXTPROC __glewVariantsvEXT = NULL;
+PFNGLVARIANTUBVEXTPROC __glewVariantubvEXT = NULL;
+PFNGLVARIANTUIVEXTPROC __glewVariantuivEXT = NULL;
+PFNGLVARIANTUSVEXTPROC __glewVariantusvEXT = NULL;
+PFNGLWRITEMASKEXTPROC __glewWriteMaskEXT = NULL;
+
+PFNGLVERTEXWEIGHTPOINTEREXTPROC __glewVertexWeightPointerEXT = NULL;
+PFNGLVERTEXWEIGHTFEXTPROC __glewVertexWeightfEXT = NULL;
+PFNGLVERTEXWEIGHTFVEXTPROC __glewVertexWeightfvEXT = NULL;
+
+PFNGLFRAMETERMINATORGREMEDYPROC __glewFrameTerminatorGREMEDY = NULL;
+
+PFNGLSTRINGMARKERGREMEDYPROC __glewStringMarkerGREMEDY = NULL;
+
+PFNGLGETIMAGETRANSFORMPARAMETERFVHPPROC __glewGetImageTransformParameterfvHP = NULL;
+PFNGLGETIMAGETRANSFORMPARAMETERIVHPPROC __glewGetImageTransformParameterivHP = NULL;
+PFNGLIMAGETRANSFORMPARAMETERFHPPROC __glewImageTransformParameterfHP = NULL;
+PFNGLIMAGETRANSFORMPARAMETERFVHPPROC __glewImageTransformParameterfvHP = NULL;
+PFNGLIMAGETRANSFORMPARAMETERIHPPROC __glewImageTransformParameteriHP = NULL;
+PFNGLIMAGETRANSFORMPARAMETERIVHPPROC __glewImageTransformParameterivHP = NULL;
+
+PFNGLMULTIMODEDRAWARRAYSIBMPROC __glewMultiModeDrawArraysIBM = NULL;
+PFNGLMULTIMODEDRAWELEMENTSIBMPROC __glewMultiModeDrawElementsIBM = NULL;
+
+PFNGLCOLORPOINTERLISTIBMPROC __glewColorPointerListIBM = NULL;
+PFNGLEDGEFLAGPOINTERLISTIBMPROC __glewEdgeFlagPointerListIBM = NULL;
+PFNGLFOGCOORDPOINTERLISTIBMPROC __glewFogCoordPointerListIBM = NULL;
+PFNGLINDEXPOINTERLISTIBMPROC __glewIndexPointerListIBM = NULL;
+PFNGLNORMALPOINTERLISTIBMPROC __glewNormalPointerListIBM = NULL;
+PFNGLSECONDARYCOLORPOINTERLISTIBMPROC __glewSecondaryColorPointerListIBM = NULL;
+PFNGLTEXCOORDPOINTERLISTIBMPROC __glewTexCoordPointerListIBM = NULL;
+PFNGLVERTEXPOINTERLISTIBMPROC __glewVertexPointerListIBM = NULL;
+
+PFNGLCOLORPOINTERVINTELPROC __glewColorPointervINTEL = NULL;
+PFNGLNORMALPOINTERVINTELPROC __glewNormalPointervINTEL = NULL;
+PFNGLTEXCOORDPOINTERVINTELPROC __glewTexCoordPointervINTEL = NULL;
+PFNGLVERTEXPOINTERVINTELPROC __glewVertexPointervINTEL = NULL;
+
+PFNGLTEXSCISSORFUNCINTELPROC __glewTexScissorFuncINTEL = NULL;
+PFNGLTEXSCISSORINTELPROC __glewTexScissorINTEL = NULL;
+
+PFNGLBUFFERREGIONENABLEDEXTPROC __glewBufferRegionEnabledEXT = NULL;
+PFNGLDELETEBUFFERREGIONEXTPROC __glewDeleteBufferRegionEXT = NULL;
+PFNGLDRAWBUFFERREGIONEXTPROC __glewDrawBufferRegionEXT = NULL;
+PFNGLNEWBUFFERREGIONEXTPROC __glewNewBufferRegionEXT = NULL;
+PFNGLREADBUFFERREGIONEXTPROC __glewReadBufferRegionEXT = NULL;
+
+PFNGLRESIZEBUFFERSMESAPROC __glewResizeBuffersMESA = NULL;
+
+PFNGLWINDOWPOS2DMESAPROC __glewWindowPos2dMESA = NULL;
+PFNGLWINDOWPOS2DVMESAPROC __glewWindowPos2dvMESA = NULL;
+PFNGLWINDOWPOS2FMESAPROC __glewWindowPos2fMESA = NULL;
+PFNGLWINDOWPOS2FVMESAPROC __glewWindowPos2fvMESA = NULL;
+PFNGLWINDOWPOS2IMESAPROC __glewWindowPos2iMESA = NULL;
+PFNGLWINDOWPOS2IVMESAPROC __glewWindowPos2ivMESA = NULL;
+PFNGLWINDOWPOS2SMESAPROC __glewWindowPos2sMESA = NULL;
+PFNGLWINDOWPOS2SVMESAPROC __glewWindowPos2svMESA = NULL;
+PFNGLWINDOWPOS3DMESAPROC __glewWindowPos3dMESA = NULL;
+PFNGLWINDOWPOS3DVMESAPROC __glewWindowPos3dvMESA = NULL;
+PFNGLWINDOWPOS3FMESAPROC __glewWindowPos3fMESA = NULL;
+PFNGLWINDOWPOS3FVMESAPROC __glewWindowPos3fvMESA = NULL;
+PFNGLWINDOWPOS3IMESAPROC __glewWindowPos3iMESA = NULL;
+PFNGLWINDOWPOS3IVMESAPROC __glewWindowPos3ivMESA = NULL;
+PFNGLWINDOWPOS3SMESAPROC __glewWindowPos3sMESA = NULL;
+PFNGLWINDOWPOS3SVMESAPROC __glewWindowPos3svMESA = NULL;
+PFNGLWINDOWPOS4DMESAPROC __glewWindowPos4dMESA = NULL;
+PFNGLWINDOWPOS4DVMESAPROC __glewWindowPos4dvMESA = NULL;
+PFNGLWINDOWPOS4FMESAPROC __glewWindowPos4fMESA = NULL;
+PFNGLWINDOWPOS4FVMESAPROC __glewWindowPos4fvMESA = NULL;
+PFNGLWINDOWPOS4IMESAPROC __glewWindowPos4iMESA = NULL;
+PFNGLWINDOWPOS4IVMESAPROC __glewWindowPos4ivMESA = NULL;
+PFNGLWINDOWPOS4SMESAPROC __glewWindowPos4sMESA = NULL;
+PFNGLWINDOWPOS4SVMESAPROC __glewWindowPos4svMESA = NULL;
+
+PFNGLBEGINCONDITIONALRENDERNVPROC __glewBeginConditionalRenderNV = NULL;
+PFNGLENDCONDITIONALRENDERNVPROC __glewEndConditionalRenderNV = NULL;
+
+PFNGLCLEARDEPTHDNVPROC __glewClearDepthdNV = NULL;
+PFNGLDEPTHBOUNDSDNVPROC __glewDepthBoundsdNV = NULL;
+PFNGLDEPTHRANGEDNVPROC __glewDepthRangedNV = NULL;
+
+PFNGLEVALMAPSNVPROC __glewEvalMapsNV = NULL;
+PFNGLGETMAPATTRIBPARAMETERFVNVPROC __glewGetMapAttribParameterfvNV = NULL;
+PFNGLGETMAPATTRIBPARAMETERIVNVPROC __glewGetMapAttribParameterivNV = NULL;
+PFNGLGETMAPCONTROLPOINTSNVPROC __glewGetMapControlPointsNV = NULL;
+PFNGLGETMAPPARAMETERFVNVPROC __glewGetMapParameterfvNV = NULL;
+PFNGLGETMAPPARAMETERIVNVPROC __glewGetMapParameterivNV = NULL;
+PFNGLMAPCONTROLPOINTSNVPROC __glewMapControlPointsNV = NULL;
+PFNGLMAPPARAMETERFVNVPROC __glewMapParameterfvNV = NULL;
+PFNGLMAPPARAMETERIVNVPROC __glewMapParameterivNV = NULL;
+
+PFNGLGETMULTISAMPLEFVNVPROC __glewGetMultisamplefvNV = NULL;
+PFNGLSAMPLEMASKINDEXEDNVPROC __glewSampleMaskIndexedNV = NULL;
+PFNGLTEXRENDERBUFFERNVPROC __glewTexRenderbufferNV = NULL;
+
+PFNGLDELETEFENCESNVPROC __glewDeleteFencesNV = NULL;
+PFNGLFINISHFENCENVPROC __glewFinishFenceNV = NULL;
+PFNGLGENFENCESNVPROC __glewGenFencesNV = NULL;
+PFNGLGETFENCEIVNVPROC __glewGetFenceivNV = NULL;
+PFNGLISFENCENVPROC __glewIsFenceNV = NULL;
+PFNGLSETFENCENVPROC __glewSetFenceNV = NULL;
+PFNGLTESTFENCENVPROC __glewTestFenceNV = NULL;
+
+PFNGLGETPROGRAMNAMEDPARAMETERDVNVPROC __glewGetProgramNamedParameterdvNV = NULL;
+PFNGLGETPROGRAMNAMEDPARAMETERFVNVPROC __glewGetProgramNamedParameterfvNV = NULL;
+PFNGLPROGRAMNAMEDPARAMETER4DNVPROC __glewProgramNamedParameter4dNV = NULL;
+PFNGLPROGRAMNAMEDPARAMETER4DVNVPROC __glewProgramNamedParameter4dvNV = NULL;
+PFNGLPROGRAMNAMEDPARAMETER4FNVPROC __glewProgramNamedParameter4fNV = NULL;
+PFNGLPROGRAMNAMEDPARAMETER4FVNVPROC __glewProgramNamedParameter4fvNV = NULL;
+
+PFNGLRENDERBUFFERSTORAGEMULTISAMPLECOVERAGENVPROC __glewRenderbufferStorageMultisampleCoverageNV = NULL;
+
+PFNGLPROGRAMVERTEXLIMITNVPROC __glewProgramVertexLimitNV = NULL;
+
+PFNGLPROGRAMENVPARAMETERI4INVPROC __glewProgramEnvParameterI4iNV = NULL;
+PFNGLPROGRAMENVPARAMETERI4IVNVPROC __glewProgramEnvParameterI4ivNV = NULL;
+PFNGLPROGRAMENVPARAMETERI4UINVPROC __glewProgramEnvParameterI4uiNV = NULL;
+PFNGLPROGRAMENVPARAMETERI4UIVNVPROC __glewProgramEnvParameterI4uivNV = NULL;
+PFNGLPROGRAMENVPARAMETERSI4IVNVPROC __glewProgramEnvParametersI4ivNV = NULL;
+PFNGLPROGRAMENVPARAMETERSI4UIVNVPROC __glewProgramEnvParametersI4uivNV = NULL;
+PFNGLPROGRAMLOCALPARAMETERI4INVPROC __glewProgramLocalParameterI4iNV = NULL;
+PFNGLPROGRAMLOCALPARAMETERI4IVNVPROC __glewProgramLocalParameterI4ivNV = NULL;
+PFNGLPROGRAMLOCALPARAMETERI4UINVPROC __glewProgramLocalParameterI4uiNV = NULL;
+PFNGLPROGRAMLOCALPARAMETERI4UIVNVPROC __glewProgramLocalParameterI4uivNV = NULL;
+PFNGLPROGRAMLOCALPARAMETERSI4IVNVPROC __glewProgramLocalParametersI4ivNV = NULL;
+PFNGLPROGRAMLOCALPARAMETERSI4UIVNVPROC __glewProgramLocalParametersI4uivNV = NULL;
+
+PFNGLCOLOR3HNVPROC __glewColor3hNV = NULL;
+PFNGLCOLOR3HVNVPROC __glewColor3hvNV = NULL;
+PFNGLCOLOR4HNVPROC __glewColor4hNV = NULL;
+PFNGLCOLOR4HVNVPROC __glewColor4hvNV = NULL;
+PFNGLFOGCOORDHNVPROC __glewFogCoordhNV = NULL;
+PFNGLFOGCOORDHVNVPROC __glewFogCoordhvNV = NULL;
+PFNGLMULTITEXCOORD1HNVPROC __glewMultiTexCoord1hNV = NULL;
+PFNGLMULTITEXCOORD1HVNVPROC __glewMultiTexCoord1hvNV = NULL;
+PFNGLMULTITEXCOORD2HNVPROC __glewMultiTexCoord2hNV = NULL;
+PFNGLMULTITEXCOORD2HVNVPROC __glewMultiTexCoord2hvNV = NULL;
+PFNGLMULTITEXCOORD3HNVPROC __glewMultiTexCoord3hNV = NULL;
+PFNGLMULTITEXCOORD3HVNVPROC __glewMultiTexCoord3hvNV = NULL;
+PFNGLMULTITEXCOORD4HNVPROC __glewMultiTexCoord4hNV = NULL;
+PFNGLMULTITEXCOORD4HVNVPROC __glewMultiTexCoord4hvNV = NULL;
+PFNGLNORMAL3HNVPROC __glewNormal3hNV = NULL;
+PFNGLNORMAL3HVNVPROC __glewNormal3hvNV = NULL;
+PFNGLSECONDARYCOLOR3HNVPROC __glewSecondaryColor3hNV = NULL;
+PFNGLSECONDARYCOLOR3HVNVPROC __glewSecondaryColor3hvNV = NULL;
+PFNGLTEXCOORD1HNVPROC __glewTexCoord1hNV = NULL;
+PFNGLTEXCOORD1HVNVPROC __glewTexCoord1hvNV = NULL;
+PFNGLTEXCOORD2HNVPROC __glewTexCoord2hNV = NULL;
+PFNGLTEXCOORD2HVNVPROC __glewTexCoord2hvNV = NULL;
+PFNGLTEXCOORD3HNVPROC __glewTexCoord3hNV = NULL;
+PFNGLTEXCOORD3HVNVPROC __glewTexCoord3hvNV = NULL;
+PFNGLTEXCOORD4HNVPROC __glewTexCoord4hNV = NULL;
+PFNGLTEXCOORD4HVNVPROC __glewTexCoord4hvNV = NULL;
+PFNGLVERTEX2HNVPROC __glewVertex2hNV = NULL;
+PFNGLVERTEX2HVNVPROC __glewVertex2hvNV = NULL;
+PFNGLVERTEX3HNVPROC __glewVertex3hNV = NULL;
+PFNGLVERTEX3HVNVPROC __glewVertex3hvNV = NULL;
+PFNGLVERTEX4HNVPROC __glewVertex4hNV = NULL;
+PFNGLVERTEX4HVNVPROC __glewVertex4hvNV = NULL;
+PFNGLVERTEXATTRIB1HNVPROC __glewVertexAttrib1hNV = NULL;
+PFNGLVERTEXATTRIB1HVNVPROC __glewVertexAttrib1hvNV = NULL;
+PFNGLVERTEXATTRIB2HNVPROC __glewVertexAttrib2hNV = NULL;
+PFNGLVERTEXATTRIB2HVNVPROC __glewVertexAttrib2hvNV = NULL;
+PFNGLVERTEXATTRIB3HNVPROC __glewVertexAttrib3hNV = NULL;
+PFNGLVERTEXATTRIB3HVNVPROC __glewVertexAttrib3hvNV = NULL;
+PFNGLVERTEXATTRIB4HNVPROC __glewVertexAttrib4hNV = NULL;
+PFNGLVERTEXATTRIB4HVNVPROC __glewVertexAttrib4hvNV = NULL;
+PFNGLVERTEXATTRIBS1HVNVPROC __glewVertexAttribs1hvNV = NULL;
+PFNGLVERTEXATTRIBS2HVNVPROC __glewVertexAttribs2hvNV = NULL;
+PFNGLVERTEXATTRIBS3HVNVPROC __glewVertexAttribs3hvNV = NULL;
+PFNGLVERTEXATTRIBS4HVNVPROC __glewVertexAttribs4hvNV = NULL;
+PFNGLVERTEXWEIGHTHNVPROC __glewVertexWeighthNV = NULL;
+PFNGLVERTEXWEIGHTHVNVPROC __glewVertexWeighthvNV = NULL;
+
+PFNGLBEGINOCCLUSIONQUERYNVPROC __glewBeginOcclusionQueryNV = NULL;
+PFNGLDELETEOCCLUSIONQUERIESNVPROC __glewDeleteOcclusionQueriesNV = NULL;
+PFNGLENDOCCLUSIONQUERYNVPROC __glewEndOcclusionQueryNV = NULL;
+PFNGLGENOCCLUSIONQUERIESNVPROC __glewGenOcclusionQueriesNV = NULL;
+PFNGLGETOCCLUSIONQUERYIVNVPROC __glewGetOcclusionQueryivNV = NULL;
+PFNGLGETOCCLUSIONQUERYUIVNVPROC __glewGetOcclusionQueryuivNV = NULL;
+PFNGLISOCCLUSIONQUERYNVPROC __glewIsOcclusionQueryNV = NULL;
+
+PFNGLPROGRAMBUFFERPARAMETERSIIVNVPROC __glewProgramBufferParametersIivNV = NULL;
+PFNGLPROGRAMBUFFERPARAMETERSIUIVNVPROC __glewProgramBufferParametersIuivNV = NULL;
+PFNGLPROGRAMBUFFERPARAMETERSFVNVPROC __glewProgramBufferParametersfvNV = NULL;
+
+PFNGLFLUSHPIXELDATARANGENVPROC __glewFlushPixelDataRangeNV = NULL;
+PFNGLPIXELDATARANGENVPROC __glewPixelDataRangeNV = NULL;
+
+PFNGLPOINTPARAMETERINVPROC __glewPointParameteriNV = NULL;
+PFNGLPOINTPARAMETERIVNVPROC __glewPointParameterivNV = NULL;
+
+PFNGLGETVIDEOI64VNVPROC __glewGetVideoi64vNV = NULL;
+PFNGLGETVIDEOIVNVPROC __glewGetVideoivNV = NULL;
+PFNGLGETVIDEOUI64VNVPROC __glewGetVideoui64vNV = NULL;
+PFNGLGETVIDEOUIVNVPROC __glewGetVideouivNV = NULL;
+PFNGLPRESENTFRAMEDUALFILLNVPROC __glewPresentFrameDualFillNV = NULL;
+PFNGLPRESENTFRAMEKEYEDNVPROC __glewPresentFrameKeyedNV = NULL;
+PFNGLVIDEOPARAMETERIVNVPROC __glewVideoParameterivNV = NULL;
+
+PFNGLPRIMITIVERESTARTINDEXNVPROC __glewPrimitiveRestartIndexNV = NULL;
+PFNGLPRIMITIVERESTARTNVPROC __glewPrimitiveRestartNV = NULL;
+
+PFNGLCOMBINERINPUTNVPROC __glewCombinerInputNV = NULL;
+PFNGLCOMBINEROUTPUTNVPROC __glewCombinerOutputNV = NULL;
+PFNGLCOMBINERPARAMETERFNVPROC __glewCombinerParameterfNV = NULL;
+PFNGLCOMBINERPARAMETERFVNVPROC __glewCombinerParameterfvNV = NULL;
+PFNGLCOMBINERPARAMETERINVPROC __glewCombinerParameteriNV = NULL;
+PFNGLCOMBINERPARAMETERIVNVPROC __glewCombinerParameterivNV = NULL;
+PFNGLFINALCOMBINERINPUTNVPROC __glewFinalCombinerInputNV = NULL;
+PFNGLGETCOMBINERINPUTPARAMETERFVNVPROC __glewGetCombinerInputParameterfvNV = NULL;
+PFNGLGETCOMBINERINPUTPARAMETERIVNVPROC __glewGetCombinerInputParameterivNV = NULL;
+PFNGLGETCOMBINEROUTPUTPARAMETERFVNVPROC __glewGetCombinerOutputParameterfvNV = NULL;
+PFNGLGETCOMBINEROUTPUTPARAMETERIVNVPROC __glewGetCombinerOutputParameterivNV = NULL;
+PFNGLGETFINALCOMBINERINPUTPARAMETERFVNVPROC __glewGetFinalCombinerInputParameterfvNV = NULL;
+PFNGLGETFINALCOMBINERINPUTPARAMETERIVNVPROC __glewGetFinalCombinerInputParameterivNV = NULL;
+
+PFNGLCOMBINERSTAGEPARAMETERFVNVPROC __glewCombinerStageParameterfvNV = NULL;
+PFNGLGETCOMBINERSTAGEPARAMETERFVNVPROC __glewGetCombinerStageParameterfvNV = NULL;
+
+PFNGLACTIVEVARYINGNVPROC __glewActiveVaryingNV = NULL;
+PFNGLBEGINTRANSFORMFEEDBACKNVPROC __glewBeginTransformFeedbackNV = NULL;
+PFNGLBINDBUFFERBASENVPROC __glewBindBufferBaseNV = NULL;
+PFNGLBINDBUFFEROFFSETNVPROC __glewBindBufferOffsetNV = NULL;
+PFNGLBINDBUFFERRANGENVPROC __glewBindBufferRangeNV = NULL;
+PFNGLENDTRANSFORMFEEDBACKNVPROC __glewEndTransformFeedbackNV = NULL;
+PFNGLGETACTIVEVARYINGNVPROC __glewGetActiveVaryingNV = NULL;
+PFNGLGETTRANSFORMFEEDBACKVARYINGNVPROC __glewGetTransformFeedbackVaryingNV = NULL;
+PFNGLGETVARYINGLOCATIONNVPROC __glewGetVaryingLocationNV = NULL;
+PFNGLTRANSFORMFEEDBACKATTRIBSNVPROC __glewTransformFeedbackAttribsNV = NULL;
+PFNGLTRANSFORMFEEDBACKVARYINGSNVPROC __glewTransformFeedbackVaryingsNV = NULL;
+
+PFNGLFLUSHVERTEXARRAYRANGENVPROC __glewFlushVertexArrayRangeNV = NULL;
+PFNGLVERTEXARRAYRANGENVPROC __glewVertexArrayRangeNV = NULL;
+
+PFNGLAREPROGRAMSRESIDENTNVPROC __glewAreProgramsResidentNV = NULL;
+PFNGLBINDPROGRAMNVPROC __glewBindProgramNV = NULL;
+PFNGLDELETEPROGRAMSNVPROC __glewDeleteProgramsNV = NULL;
+PFNGLEXECUTEPROGRAMNVPROC __glewExecuteProgramNV = NULL;
+PFNGLGENPROGRAMSNVPROC __glewGenProgramsNV = NULL;
+PFNGLGETPROGRAMPARAMETERDVNVPROC __glewGetProgramParameterdvNV = NULL;
+PFNGLGETPROGRAMPARAMETERFVNVPROC __glewGetProgramParameterfvNV = NULL;
+PFNGLGETPROGRAMSTRINGNVPROC __glewGetProgramStringNV = NULL;
+PFNGLGETPROGRAMIVNVPROC __glewGetProgramivNV = NULL;
+PFNGLGETTRACKMATRIXIVNVPROC __glewGetTrackMatrixivNV = NULL;
+PFNGLGETVERTEXATTRIBPOINTERVNVPROC __glewGetVertexAttribPointervNV = NULL;
+PFNGLGETVERTEXATTRIBDVNVPROC __glewGetVertexAttribdvNV = NULL;
+PFNGLGETVERTEXATTRIBFVNVPROC __glewGetVertexAttribfvNV = NULL;
+PFNGLGETVERTEXATTRIBIVNVPROC __glewGetVertexAttribivNV = NULL;
+PFNGLISPROGRAMNVPROC __glewIsProgramNV = NULL;
+PFNGLLOADPROGRAMNVPROC __glewLoadProgramNV = NULL;
+PFNGLPROGRAMPARAMETER4DNVPROC __glewProgramParameter4dNV = NULL;
+PFNGLPROGRAMPARAMETER4DVNVPROC __glewProgramParameter4dvNV = NULL;
+PFNGLPROGRAMPARAMETER4FNVPROC __glewProgramParameter4fNV = NULL;
+PFNGLPROGRAMPARAMETER4FVNVPROC __glewProgramParameter4fvNV = NULL;
+PFNGLPROGRAMPARAMETERS4DVNVPROC __glewProgramParameters4dvNV = NULL;
+PFNGLPROGRAMPARAMETERS4FVNVPROC __glewProgramParameters4fvNV = NULL;
+PFNGLREQUESTRESIDENTPROGRAMSNVPROC __glewRequestResidentProgramsNV = NULL;
+PFNGLTRACKMATRIXNVPROC __glewTrackMatrixNV = NULL;
+PFNGLVERTEXATTRIB1DNVPROC __glewVertexAttrib1dNV = NULL;
+PFNGLVERTEXATTRIB1DVNVPROC __glewVertexAttrib1dvNV = NULL;
+PFNGLVERTEXATTRIB1FNVPROC __glewVertexAttrib1fNV = NULL;
+PFNGLVERTEXATTRIB1FVNVPROC __glewVertexAttrib1fvNV = NULL;
+PFNGLVERTEXATTRIB1SNVPROC __glewVertexAttrib1sNV = NULL;
+PFNGLVERTEXATTRIB1SVNVPROC __glewVertexAttrib1svNV = NULL;
+PFNGLVERTEXATTRIB2DNVPROC __glewVertexAttrib2dNV = NULL;
+PFNGLVERTEXATTRIB2DVNVPROC __glewVertexAttrib2dvNV = NULL;
+PFNGLVERTEXATTRIB2FNVPROC __glewVertexAttrib2fNV = NULL;
+PFNGLVERTEXATTRIB2FVNVPROC __glewVertexAttrib2fvNV = NULL;
+PFNGLVERTEXATTRIB2SNVPROC __glewVertexAttrib2sNV = NULL;
+PFNGLVERTEXATTRIB2SVNVPROC __glewVertexAttrib2svNV = NULL;
+PFNGLVERTEXATTRIB3DNVPROC __glewVertexAttrib3dNV = NULL;
+PFNGLVERTEXATTRIB3DVNVPROC __glewVertexAttrib3dvNV = NULL;
+PFNGLVERTEXATTRIB3FNVPROC __glewVertexAttrib3fNV = NULL;
+PFNGLVERTEXATTRIB3FVNVPROC __glewVertexAttrib3fvNV = NULL;
+PFNGLVERTEXATTRIB3SNVPROC __glewVertexAttrib3sNV = NULL;
+PFNGLVERTEXATTRIB3SVNVPROC __glewVertexAttrib3svNV = NULL;
+PFNGLVERTEXATTRIB4DNVPROC __glewVertexAttrib4dNV = NULL;
+PFNGLVERTEXATTRIB4DVNVPROC __glewVertexAttrib4dvNV = NULL;
+PFNGLVERTEXATTRIB4FNVPROC __glewVertexAttrib4fNV = NULL;
+PFNGLVERTEXATTRIB4FVNVPROC __glewVertexAttrib4fvNV = NULL;
+PFNGLVERTEXATTRIB4SNVPROC __glewVertexAttrib4sNV = NULL;
+PFNGLVERTEXATTRIB4SVNVPROC __glewVertexAttrib4svNV = NULL;
+PFNGLVERTEXATTRIB4UBNVPROC __glewVertexAttrib4ubNV = NULL;
+PFNGLVERTEXATTRIB4UBVNVPROC __glewVertexAttrib4ubvNV = NULL;
+PFNGLVERTEXATTRIBPOINTERNVPROC __glewVertexAttribPointerNV = NULL;
+PFNGLVERTEXATTRIBS1DVNVPROC __glewVertexAttribs1dvNV = NULL;
+PFNGLVERTEXATTRIBS1FVNVPROC __glewVertexAttribs1fvNV = NULL;
+PFNGLVERTEXATTRIBS1SVNVPROC __glewVertexAttribs1svNV = NULL;
+PFNGLVERTEXATTRIBS2DVNVPROC __glewVertexAttribs2dvNV = NULL;
+PFNGLVERTEXATTRIBS2FVNVPROC __glewVertexAttribs2fvNV = NULL;
+PFNGLVERTEXATTRIBS2SVNVPROC __glewVertexAttribs2svNV = NULL;
+PFNGLVERTEXATTRIBS3DVNVPROC __glewVertexAttribs3dvNV = NULL;
+PFNGLVERTEXATTRIBS3FVNVPROC __glewVertexAttribs3fvNV = NULL;
+PFNGLVERTEXATTRIBS3SVNVPROC __glewVertexAttribs3svNV = NULL;
+PFNGLVERTEXATTRIBS4DVNVPROC __glewVertexAttribs4dvNV = NULL;
+PFNGLVERTEXATTRIBS4FVNVPROC __glewVertexAttribs4fvNV = NULL;
+PFNGLVERTEXATTRIBS4SVNVPROC __glewVertexAttribs4svNV = NULL;
+PFNGLVERTEXATTRIBS4UBVNVPROC __glewVertexAttribs4ubvNV = NULL;
+
+PFNGLCLEARDEPTHFOESPROC __glewClearDepthfOES = NULL;
+PFNGLCLIPPLANEFOESPROC __glewClipPlanefOES = NULL;
+PFNGLDEPTHRANGEFOESPROC __glewDepthRangefOES = NULL;
+PFNGLFRUSTUMFOESPROC __glewFrustumfOES = NULL;
+PFNGLGETCLIPPLANEFOESPROC __glewGetClipPlanefOES = NULL;
+PFNGLORTHOFOESPROC __glewOrthofOES = NULL;
+
+PFNGLDETAILTEXFUNCSGISPROC __glewDetailTexFuncSGIS = NULL;
+PFNGLGETDETAILTEXFUNCSGISPROC __glewGetDetailTexFuncSGIS = NULL;
+
+PFNGLFOGFUNCSGISPROC __glewFogFuncSGIS = NULL;
+PFNGLGETFOGFUNCSGISPROC __glewGetFogFuncSGIS = NULL;
+
+PFNGLSAMPLEMASKSGISPROC __glewSampleMaskSGIS = NULL;
+PFNGLSAMPLEPATTERNSGISPROC __glewSamplePatternSGIS = NULL;
+
+PFNGLGETSHARPENTEXFUNCSGISPROC __glewGetSharpenTexFuncSGIS = NULL;
+PFNGLSHARPENTEXFUNCSGISPROC __glewSharpenTexFuncSGIS = NULL;
+
+PFNGLTEXIMAGE4DSGISPROC __glewTexImage4DSGIS = NULL;
+PFNGLTEXSUBIMAGE4DSGISPROC __glewTexSubImage4DSGIS = NULL;
+
+PFNGLGETTEXFILTERFUNCSGISPROC __glewGetTexFilterFuncSGIS = NULL;
+PFNGLTEXFILTERFUNCSGISPROC __glewTexFilterFuncSGIS = NULL;
+
+PFNGLASYNCMARKERSGIXPROC __glewAsyncMarkerSGIX = NULL;
+PFNGLDELETEASYNCMARKERSSGIXPROC __glewDeleteAsyncMarkersSGIX = NULL;
+PFNGLFINISHASYNCSGIXPROC __glewFinishAsyncSGIX = NULL;
+PFNGLGENASYNCMARKERSSGIXPROC __glewGenAsyncMarkersSGIX = NULL;
+PFNGLISASYNCMARKERSGIXPROC __glewIsAsyncMarkerSGIX = NULL;
+PFNGLPOLLASYNCSGIXPROC __glewPollAsyncSGIX = NULL;
+
+PFNGLFLUSHRASTERSGIXPROC __glewFlushRasterSGIX = NULL;
+
+PFNGLTEXTUREFOGSGIXPROC __glewTextureFogSGIX = NULL;
+
+PFNGLFRAGMENTCOLORMATERIALSGIXPROC __glewFragmentColorMaterialSGIX = NULL;
+PFNGLFRAGMENTLIGHTMODELFSGIXPROC __glewFragmentLightModelfSGIX = NULL;
+PFNGLFRAGMENTLIGHTMODELFVSGIXPROC __glewFragmentLightModelfvSGIX = NULL;
+PFNGLFRAGMENTLIGHTMODELISGIXPROC __glewFragmentLightModeliSGIX = NULL;
+PFNGLFRAGMENTLIGHTMODELIVSGIXPROC __glewFragmentLightModelivSGIX = NULL;
+PFNGLFRAGMENTLIGHTFSGIXPROC __glewFragmentLightfSGIX = NULL;
+PFNGLFRAGMENTLIGHTFVSGIXPROC __glewFragmentLightfvSGIX = NULL;
+PFNGLFRAGMENTLIGHTISGIXPROC __glewFragmentLightiSGIX = NULL;
+PFNGLFRAGMENTLIGHTIVSGIXPROC __glewFragmentLightivSGIX = NULL;
+PFNGLFRAGMENTMATERIALFSGIXPROC __glewFragmentMaterialfSGIX = NULL;
+PFNGLFRAGMENTMATERIALFVSGIXPROC __glewFragmentMaterialfvSGIX = NULL;
+PFNGLFRAGMENTMATERIALISGIXPROC __glewFragmentMaterialiSGIX = NULL;
+PFNGLFRAGMENTMATERIALIVSGIXPROC __glewFragmentMaterialivSGIX = NULL;
+PFNGLGETFRAGMENTLIGHTFVSGIXPROC __glewGetFragmentLightfvSGIX = NULL;
+PFNGLGETFRAGMENTLIGHTIVSGIXPROC __glewGetFragmentLightivSGIX = NULL;
+PFNGLGETFRAGMENTMATERIALFVSGIXPROC __glewGetFragmentMaterialfvSGIX = NULL;
+PFNGLGETFRAGMENTMATERIALIVSGIXPROC __glewGetFragmentMaterialivSGIX = NULL;
+
+PFNGLFRAMEZOOMSGIXPROC __glewFrameZoomSGIX = NULL;
+
+PFNGLPIXELTEXGENSGIXPROC __glewPixelTexGenSGIX = NULL;
+
+PFNGLREFERENCEPLANESGIXPROC __glewReferencePlaneSGIX = NULL;
+
+PFNGLSPRITEPARAMETERFSGIXPROC __glewSpriteParameterfSGIX = NULL;
+PFNGLSPRITEPARAMETERFVSGIXPROC __glewSpriteParameterfvSGIX = NULL;
+PFNGLSPRITEPARAMETERISGIXPROC __glewSpriteParameteriSGIX = NULL;
+PFNGLSPRITEPARAMETERIVSGIXPROC __glewSpriteParameterivSGIX = NULL;
+
+PFNGLTAGSAMPLEBUFFERSGIXPROC __glewTagSampleBufferSGIX = NULL;
+
+PFNGLCOLORTABLEPARAMETERFVSGIPROC __glewColorTableParameterfvSGI = NULL;
+PFNGLCOLORTABLEPARAMETERIVSGIPROC __glewColorTableParameterivSGI = NULL;
+PFNGLCOLORTABLESGIPROC __glewColorTableSGI = NULL;
+PFNGLCOPYCOLORTABLESGIPROC __glewCopyColorTableSGI = NULL;
+PFNGLGETCOLORTABLEPARAMETERFVSGIPROC __glewGetColorTableParameterfvSGI = NULL;
+PFNGLGETCOLORTABLEPARAMETERIVSGIPROC __glewGetColorTableParameterivSGI = NULL;
+PFNGLGETCOLORTABLESGIPROC __glewGetColorTableSGI = NULL;
+
+PFNGLFINISHTEXTURESUNXPROC __glewFinishTextureSUNX = NULL;
+
+PFNGLGLOBALALPHAFACTORBSUNPROC __glewGlobalAlphaFactorbSUN = NULL;
+PFNGLGLOBALALPHAFACTORDSUNPROC __glewGlobalAlphaFactordSUN = NULL;
+PFNGLGLOBALALPHAFACTORFSUNPROC __glewGlobalAlphaFactorfSUN = NULL;
+PFNGLGLOBALALPHAFACTORISUNPROC __glewGlobalAlphaFactoriSUN = NULL;
+PFNGLGLOBALALPHAFACTORSSUNPROC __glewGlobalAlphaFactorsSUN = NULL;
+PFNGLGLOBALALPHAFACTORUBSUNPROC __glewGlobalAlphaFactorubSUN = NULL;
+PFNGLGLOBALALPHAFACTORUISUNPROC __glewGlobalAlphaFactoruiSUN = NULL;
+PFNGLGLOBALALPHAFACTORUSSUNPROC __glewGlobalAlphaFactorusSUN = NULL;
+
+PFNGLREADVIDEOPIXELSSUNPROC __glewReadVideoPixelsSUN = NULL;
+
+PFNGLREPLACEMENTCODEPOINTERSUNPROC __glewReplacementCodePointerSUN = NULL;
+PFNGLREPLACEMENTCODEUBSUNPROC __glewReplacementCodeubSUN = NULL;
+PFNGLREPLACEMENTCODEUBVSUNPROC __glewReplacementCodeubvSUN = NULL;
+PFNGLREPLACEMENTCODEUISUNPROC __glewReplacementCodeuiSUN = NULL;
+PFNGLREPLACEMENTCODEUIVSUNPROC __glewReplacementCodeuivSUN = NULL;
+PFNGLREPLACEMENTCODEUSSUNPROC __glewReplacementCodeusSUN = NULL;
+PFNGLREPLACEMENTCODEUSVSUNPROC __glewReplacementCodeusvSUN = NULL;
+
+PFNGLCOLOR3FVERTEX3FSUNPROC __glewColor3fVertex3fSUN = NULL;
+PFNGLCOLOR3FVERTEX3FVSUNPROC __glewColor3fVertex3fvSUN = NULL;
+PFNGLCOLOR4FNORMAL3FVERTEX3FSUNPROC __glewColor4fNormal3fVertex3fSUN = NULL;
+PFNGLCOLOR4FNORMAL3FVERTEX3FVSUNPROC __glewColor4fNormal3fVertex3fvSUN = NULL;
+PFNGLCOLOR4UBVERTEX2FSUNPROC __glewColor4ubVertex2fSUN = NULL;
+PFNGLCOLOR4UBVERTEX2FVSUNPROC __glewColor4ubVertex2fvSUN = NULL;
+PFNGLCOLOR4UBVERTEX3FSUNPROC __glewColor4ubVertex3fSUN = NULL;
+PFNGLCOLOR4UBVERTEX3FVSUNPROC __glewColor4ubVertex3fvSUN = NULL;
+PFNGLNORMAL3FVERTEX3FSUNPROC __glewNormal3fVertex3fSUN = NULL;
+PFNGLNORMAL3FVERTEX3FVSUNPROC __glewNormal3fVertex3fvSUN = NULL;
+PFNGLREPLACEMENTCODEUICOLOR3FVERTEX3FSUNPROC __glewReplacementCodeuiColor3fVertex3fSUN = NULL;
+PFNGLREPLACEMENTCODEUICOLOR3FVERTEX3FVSUNPROC __glewReplacementCodeuiColor3fVertex3fvSUN = NULL;
+PFNGLREPLACEMENTCODEUICOLOR4FNORMAL3FVERTEX3FSUNPROC __glewReplacementCodeuiColor4fNormal3fVertex3fSUN = NULL;
+PFNGLREPLACEMENTCODEUICOLOR4FNORMAL3FVERTEX3FVSUNPROC __glewReplacementCodeuiColor4fNormal3fVertex3fvSUN = NULL;
+PFNGLREPLACEMENTCODEUICOLOR4UBVERTEX3FSUNPROC __glewReplacementCodeuiColor4ubVertex3fSUN = NULL;
+PFNGLREPLACEMENTCODEUICOLOR4UBVERTEX3FVSUNPROC __glewReplacementCodeuiColor4ubVertex3fvSUN = NULL;
+PFNGLREPLACEMENTCODEUINORMAL3FVERTEX3FSUNPROC __glewReplacementCodeuiNormal3fVertex3fSUN = NULL;
+PFNGLREPLACEMENTCODEUINORMAL3FVERTEX3FVSUNPROC __glewReplacementCodeuiNormal3fVertex3fvSUN = NULL;
+PFNGLREPLACEMENTCODEUITEXCOORD2FCOLOR4FNORMAL3FVERTEX3FSUNPROC __glewReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN = NULL;
+PFNGLREPLACEMENTCODEUITEXCOORD2FCOLOR4FNORMAL3FVERTEX3FVSUNPROC __glewReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN = NULL;
+PFNGLREPLACEMENTCODEUITEXCOORD2FNORMAL3FVERTEX3FSUNPROC __glewReplacementCodeuiTexCoord2fNormal3fVertex3fSUN = NULL;
+PFNGLREPLACEMENTCODEUITEXCOORD2FNORMAL3FVERTEX3FVSUNPROC __glewReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN = NULL;
+PFNGLREPLACEMENTCODEUITEXCOORD2FVERTEX3FSUNPROC __glewReplacementCodeuiTexCoord2fVertex3fSUN = NULL;
+PFNGLREPLACEMENTCODEUITEXCOORD2FVERTEX3FVSUNPROC __glewReplacementCodeuiTexCoord2fVertex3fvSUN = NULL;
+PFNGLREPLACEMENTCODEUIVERTEX3FSUNPROC __glewReplacementCodeuiVertex3fSUN = NULL;
+PFNGLREPLACEMENTCODEUIVERTEX3FVSUNPROC __glewReplacementCodeuiVertex3fvSUN = NULL;
+PFNGLTEXCOORD2FCOLOR3FVERTEX3FSUNPROC __glewTexCoord2fColor3fVertex3fSUN = NULL;
+PFNGLTEXCOORD2FCOLOR3FVERTEX3FVSUNPROC __glewTexCoord2fColor3fVertex3fvSUN = NULL;
+PFNGLTEXCOORD2FCOLOR4FNORMAL3FVERTEX3FSUNPROC __glewTexCoord2fColor4fNormal3fVertex3fSUN = NULL;
+PFNGLTEXCOORD2FCOLOR4FNORMAL3FVERTEX3FVSUNPROC __glewTexCoord2fColor4fNormal3fVertex3fvSUN = NULL;
+PFNGLTEXCOORD2FCOLOR4UBVERTEX3FSUNPROC __glewTexCoord2fColor4ubVertex3fSUN = NULL;
+PFNGLTEXCOORD2FCOLOR4UBVERTEX3FVSUNPROC __glewTexCoord2fColor4ubVertex3fvSUN = NULL;
+PFNGLTEXCOORD2FNORMAL3FVERTEX3FSUNPROC __glewTexCoord2fNormal3fVertex3fSUN = NULL;
+PFNGLTEXCOORD2FNORMAL3FVERTEX3FVSUNPROC __glewTexCoord2fNormal3fVertex3fvSUN = NULL;
+PFNGLTEXCOORD2FVERTEX3FSUNPROC __glewTexCoord2fVertex3fSUN = NULL;
+PFNGLTEXCOORD2FVERTEX3FVSUNPROC __glewTexCoord2fVertex3fvSUN = NULL;
+PFNGLTEXCOORD4FCOLOR4FNORMAL3FVERTEX4FSUNPROC __glewTexCoord4fColor4fNormal3fVertex4fSUN = NULL;
+PFNGLTEXCOORD4FCOLOR4FNORMAL3FVERTEX4FVSUNPROC __glewTexCoord4fColor4fNormal3fVertex4fvSUN = NULL;
+PFNGLTEXCOORD4FVERTEX4FSUNPROC __glewTexCoord4fVertex4fSUN = NULL;
+PFNGLTEXCOORD4FVERTEX4FVSUNPROC __glewTexCoord4fVertex4fvSUN = NULL;
+
+PFNGLADDSWAPHINTRECTWINPROC __glewAddSwapHintRectWIN = NULL;
+
+#endif /* !WIN32 || !GLEW_MX */
+
+#if !defined(GLEW_MX)
+
+GLboolean __GLEW_VERSION_1_1 = GL_FALSE;
+GLboolean __GLEW_VERSION_1_2 = GL_FALSE;
+GLboolean __GLEW_VERSION_1_3 = GL_FALSE;
+GLboolean __GLEW_VERSION_1_4 = GL_FALSE;
+GLboolean __GLEW_VERSION_1_5 = GL_FALSE;
+GLboolean __GLEW_VERSION_2_0 = GL_FALSE;
+GLboolean __GLEW_VERSION_2_1 = GL_FALSE;
+GLboolean __GLEW_VERSION_3_0 = GL_FALSE;
+GLboolean __GLEW_3DFX_multisample = GL_FALSE;
+GLboolean __GLEW_3DFX_tbuffer = GL_FALSE;
+GLboolean __GLEW_3DFX_texture_compression_FXT1 = GL_FALSE;
+GLboolean __GLEW_APPLE_client_storage = GL_FALSE;
+GLboolean __GLEW_APPLE_element_array = GL_FALSE;
+GLboolean __GLEW_APPLE_fence = GL_FALSE;
+GLboolean __GLEW_APPLE_float_pixels = GL_FALSE;
+GLboolean __GLEW_APPLE_flush_buffer_range = GL_FALSE;
+GLboolean __GLEW_APPLE_pixel_buffer = GL_FALSE;
+GLboolean __GLEW_APPLE_specular_vector = GL_FALSE;
+GLboolean __GLEW_APPLE_texture_range = GL_FALSE;
+GLboolean __GLEW_APPLE_transform_hint = GL_FALSE;
+GLboolean __GLEW_APPLE_vertex_array_object = GL_FALSE;
+GLboolean __GLEW_APPLE_vertex_array_range = GL_FALSE;
+GLboolean __GLEW_APPLE_ycbcr_422 = GL_FALSE;
+GLboolean __GLEW_ARB_color_buffer_float = GL_FALSE;
+GLboolean __GLEW_ARB_depth_buffer_float = GL_FALSE;
+GLboolean __GLEW_ARB_depth_texture = GL_FALSE;
+GLboolean __GLEW_ARB_draw_buffers = GL_FALSE;
+GLboolean __GLEW_ARB_draw_instanced = GL_FALSE;
+GLboolean __GLEW_ARB_fragment_program = GL_FALSE;
+GLboolean __GLEW_ARB_fragment_program_shadow = GL_FALSE;
+GLboolean __GLEW_ARB_fragment_shader = GL_FALSE;
+GLboolean __GLEW_ARB_framebuffer_object = GL_FALSE;
+GLboolean __GLEW_ARB_framebuffer_sRGB = GL_FALSE;
+GLboolean __GLEW_ARB_geometry_shader4 = GL_FALSE;
+GLboolean __GLEW_ARB_half_float_pixel = GL_FALSE;
+GLboolean __GLEW_ARB_half_float_vertex = GL_FALSE;
+GLboolean __GLEW_ARB_imaging = GL_FALSE;
+GLboolean __GLEW_ARB_instanced_arrays = GL_FALSE;
+GLboolean __GLEW_ARB_map_buffer_range = GL_FALSE;
+GLboolean __GLEW_ARB_matrix_palette = GL_FALSE;
+GLboolean __GLEW_ARB_multisample = GL_FALSE;
+GLboolean __GLEW_ARB_multitexture = GL_FALSE;
+GLboolean __GLEW_ARB_occlusion_query = GL_FALSE;
+GLboolean __GLEW_ARB_pixel_buffer_object = GL_FALSE;
+GLboolean __GLEW_ARB_point_parameters = GL_FALSE;
+GLboolean __GLEW_ARB_point_sprite = GL_FALSE;
+GLboolean __GLEW_ARB_shader_objects = GL_FALSE;
+GLboolean __GLEW_ARB_shading_language_100 = GL_FALSE;
+GLboolean __GLEW_ARB_shadow = GL_FALSE;
+GLboolean __GLEW_ARB_shadow_ambient = GL_FALSE;
+GLboolean __GLEW_ARB_texture_border_clamp = GL_FALSE;
+GLboolean __GLEW_ARB_texture_buffer_object = GL_FALSE;
+GLboolean __GLEW_ARB_texture_compression = GL_FALSE;
+GLboolean __GLEW_ARB_texture_compression_rgtc = GL_FALSE;
+GLboolean __GLEW_ARB_texture_cube_map = GL_FALSE;
+GLboolean __GLEW_ARB_texture_env_add = GL_FALSE;
+GLboolean __GLEW_ARB_texture_env_combine = GL_FALSE;
+GLboolean __GLEW_ARB_texture_env_crossbar = GL_FALSE;
+GLboolean __GLEW_ARB_texture_env_dot3 = GL_FALSE;
+GLboolean __GLEW_ARB_texture_float = GL_FALSE;
+GLboolean __GLEW_ARB_texture_mirrored_repeat = GL_FALSE;
+GLboolean __GLEW_ARB_texture_non_power_of_two = GL_FALSE;
+GLboolean __GLEW_ARB_texture_rectangle = GL_FALSE;
+GLboolean __GLEW_ARB_texture_rg = GL_FALSE;
+GLboolean __GLEW_ARB_transpose_matrix = GL_FALSE;
+GLboolean __GLEW_ARB_vertex_array_object = GL_FALSE;
+GLboolean __GLEW_ARB_vertex_blend = GL_FALSE;
+GLboolean __GLEW_ARB_vertex_buffer_object = GL_FALSE;
+GLboolean __GLEW_ARB_vertex_program = GL_FALSE;
+GLboolean __GLEW_ARB_vertex_shader = GL_FALSE;
+GLboolean __GLEW_ARB_window_pos = GL_FALSE;
+GLboolean __GLEW_ATIX_point_sprites = GL_FALSE;
+GLboolean __GLEW_ATIX_texture_env_combine3 = GL_FALSE;
+GLboolean __GLEW_ATIX_texture_env_route = GL_FALSE;
+GLboolean __GLEW_ATIX_vertex_shader_output_point_size = GL_FALSE;
+GLboolean __GLEW_ATI_draw_buffers = GL_FALSE;
+GLboolean __GLEW_ATI_element_array = GL_FALSE;
+GLboolean __GLEW_ATI_envmap_bumpmap = GL_FALSE;
+GLboolean __GLEW_ATI_fragment_shader = GL_FALSE;
+GLboolean __GLEW_ATI_map_object_buffer = GL_FALSE;
+GLboolean __GLEW_ATI_pn_triangles = GL_FALSE;
+GLboolean __GLEW_ATI_separate_stencil = GL_FALSE;
+GLboolean __GLEW_ATI_shader_texture_lod = GL_FALSE;
+GLboolean __GLEW_ATI_text_fragment_shader = GL_FALSE;
+GLboolean __GLEW_ATI_texture_compression_3dc = GL_FALSE;
+GLboolean __GLEW_ATI_texture_env_combine3 = GL_FALSE;
+GLboolean __GLEW_ATI_texture_float = GL_FALSE;
+GLboolean __GLEW_ATI_texture_mirror_once = GL_FALSE;
+GLboolean __GLEW_ATI_vertex_array_object = GL_FALSE;
+GLboolean __GLEW_ATI_vertex_attrib_array_object = GL_FALSE;
+GLboolean __GLEW_ATI_vertex_streams = GL_FALSE;
+GLboolean __GLEW_EXT_422_pixels = GL_FALSE;
+GLboolean __GLEW_EXT_Cg_shader = GL_FALSE;
+GLboolean __GLEW_EXT_abgr = GL_FALSE;
+GLboolean __GLEW_EXT_bgra = GL_FALSE;
+GLboolean __GLEW_EXT_bindable_uniform = GL_FALSE;
+GLboolean __GLEW_EXT_blend_color = GL_FALSE;
+GLboolean __GLEW_EXT_blend_equation_separate = GL_FALSE;
+GLboolean __GLEW_EXT_blend_func_separate = GL_FALSE;
+GLboolean __GLEW_EXT_blend_logic_op = GL_FALSE;
+GLboolean __GLEW_EXT_blend_minmax = GL_FALSE;
+GLboolean __GLEW_EXT_blend_subtract = GL_FALSE;
+GLboolean __GLEW_EXT_clip_volume_hint = GL_FALSE;
+GLboolean __GLEW_EXT_cmyka = GL_FALSE;
+GLboolean __GLEW_EXT_color_subtable = GL_FALSE;
+GLboolean __GLEW_EXT_compiled_vertex_array = GL_FALSE;
+GLboolean __GLEW_EXT_convolution = GL_FALSE;
+GLboolean __GLEW_EXT_coordinate_frame = GL_FALSE;
+GLboolean __GLEW_EXT_copy_texture = GL_FALSE;
+GLboolean __GLEW_EXT_cull_vertex = GL_FALSE;
+GLboolean __GLEW_EXT_depth_bounds_test = GL_FALSE;
+GLboolean __GLEW_EXT_direct_state_access = GL_FALSE;
+GLboolean __GLEW_EXT_draw_buffers2 = GL_FALSE;
+GLboolean __GLEW_EXT_draw_instanced = GL_FALSE;
+GLboolean __GLEW_EXT_draw_range_elements = GL_FALSE;
+GLboolean __GLEW_EXT_fog_coord = GL_FALSE;
+GLboolean __GLEW_EXT_fragment_lighting = GL_FALSE;
+GLboolean __GLEW_EXT_framebuffer_blit = GL_FALSE;
+GLboolean __GLEW_EXT_framebuffer_multisample = GL_FALSE;
+GLboolean __GLEW_EXT_framebuffer_object = GL_FALSE;
+GLboolean __GLEW_EXT_framebuffer_sRGB = GL_FALSE;
+GLboolean __GLEW_EXT_geometry_shader4 = GL_FALSE;
+GLboolean __GLEW_EXT_gpu_program_parameters = GL_FALSE;
+GLboolean __GLEW_EXT_gpu_shader4 = GL_FALSE;
+GLboolean __GLEW_EXT_histogram = GL_FALSE;
+GLboolean __GLEW_EXT_index_array_formats = GL_FALSE;
+GLboolean __GLEW_EXT_index_func = GL_FALSE;
+GLboolean __GLEW_EXT_index_material = GL_FALSE;
+GLboolean __GLEW_EXT_index_texture = GL_FALSE;
+GLboolean __GLEW_EXT_light_texture = GL_FALSE;
+GLboolean __GLEW_EXT_misc_attribute = GL_FALSE;
+GLboolean __GLEW_EXT_multi_draw_arrays = GL_FALSE;
+GLboolean __GLEW_EXT_multisample = GL_FALSE;
+GLboolean __GLEW_EXT_packed_depth_stencil = GL_FALSE;
+GLboolean __GLEW_EXT_packed_float = GL_FALSE;
+GLboolean __GLEW_EXT_packed_pixels = GL_FALSE;
+GLboolean __GLEW_EXT_paletted_texture = GL_FALSE;
+GLboolean __GLEW_EXT_pixel_buffer_object = GL_FALSE;
+GLboolean __GLEW_EXT_pixel_transform = GL_FALSE;
+GLboolean __GLEW_EXT_pixel_transform_color_table = GL_FALSE;
+GLboolean __GLEW_EXT_point_parameters = GL_FALSE;
+GLboolean __GLEW_EXT_polygon_offset = GL_FALSE;
+GLboolean __GLEW_EXT_rescale_normal = GL_FALSE;
+GLboolean __GLEW_EXT_scene_marker = GL_FALSE;
+GLboolean __GLEW_EXT_secondary_color = GL_FALSE;
+GLboolean __GLEW_EXT_separate_specular_color = GL_FALSE;
+GLboolean __GLEW_EXT_shadow_funcs = GL_FALSE;
+GLboolean __GLEW_EXT_shared_texture_palette = GL_FALSE;
+GLboolean __GLEW_EXT_stencil_clear_tag = GL_FALSE;
+GLboolean __GLEW_EXT_stencil_two_side = GL_FALSE;
+GLboolean __GLEW_EXT_stencil_wrap = GL_FALSE;
+GLboolean __GLEW_EXT_subtexture = GL_FALSE;
+GLboolean __GLEW_EXT_texture = GL_FALSE;
+GLboolean __GLEW_EXT_texture3D = GL_FALSE;
+GLboolean __GLEW_EXT_texture_array = GL_FALSE;
+GLboolean __GLEW_EXT_texture_buffer_object = GL_FALSE;
+GLboolean __GLEW_EXT_texture_compression_dxt1 = GL_FALSE;
+GLboolean __GLEW_EXT_texture_compression_latc = GL_FALSE;
+GLboolean __GLEW_EXT_texture_compression_rgtc = GL_FALSE;
+GLboolean __GLEW_EXT_texture_compression_s3tc = GL_FALSE;
+GLboolean __GLEW_EXT_texture_cube_map = GL_FALSE;
+GLboolean __GLEW_EXT_texture_edge_clamp = GL_FALSE;
+GLboolean __GLEW_EXT_texture_env = GL_FALSE;
+GLboolean __GLEW_EXT_texture_env_add = GL_FALSE;
+GLboolean __GLEW_EXT_texture_env_combine = GL_FALSE;
+GLboolean __GLEW_EXT_texture_env_dot3 = GL_FALSE;
+GLboolean __GLEW_EXT_texture_filter_anisotropic = GL_FALSE;
+GLboolean __GLEW_EXT_texture_integer = GL_FALSE;
+GLboolean __GLEW_EXT_texture_lod_bias = GL_FALSE;
+GLboolean __GLEW_EXT_texture_mirror_clamp = GL_FALSE;
+GLboolean __GLEW_EXT_texture_object = GL_FALSE;
+GLboolean __GLEW_EXT_texture_perturb_normal = GL_FALSE;
+GLboolean __GLEW_EXT_texture_rectangle = GL_FALSE;
+GLboolean __GLEW_EXT_texture_sRGB = GL_FALSE;
+GLboolean __GLEW_EXT_texture_shared_exponent = GL_FALSE;
+GLboolean __GLEW_EXT_texture_swizzle = GL_FALSE;
+GLboolean __GLEW_EXT_timer_query = GL_FALSE;
+GLboolean __GLEW_EXT_transform_feedback = GL_FALSE;
+GLboolean __GLEW_EXT_vertex_array = GL_FALSE;
+GLboolean __GLEW_EXT_vertex_array_bgra = GL_FALSE;
+GLboolean __GLEW_EXT_vertex_shader = GL_FALSE;
+GLboolean __GLEW_EXT_vertex_weighting = GL_FALSE;
+GLboolean __GLEW_GREMEDY_frame_terminator = GL_FALSE;
+GLboolean __GLEW_GREMEDY_string_marker = GL_FALSE;
+GLboolean __GLEW_HP_convolution_border_modes = GL_FALSE;
+GLboolean __GLEW_HP_image_transform = GL_FALSE;
+GLboolean __GLEW_HP_occlusion_test = GL_FALSE;
+GLboolean __GLEW_HP_texture_lighting = GL_FALSE;
+GLboolean __GLEW_IBM_cull_vertex = GL_FALSE;
+GLboolean __GLEW_IBM_multimode_draw_arrays = GL_FALSE;
+GLboolean __GLEW_IBM_rasterpos_clip = GL_FALSE;
+GLboolean __GLEW_IBM_static_data = GL_FALSE;
+GLboolean __GLEW_IBM_texture_mirrored_repeat = GL_FALSE;
+GLboolean __GLEW_IBM_vertex_array_lists = GL_FALSE;
+GLboolean __GLEW_INGR_color_clamp = GL_FALSE;
+GLboolean __GLEW_INGR_interlace_read = GL_FALSE;
+GLboolean __GLEW_INTEL_parallel_arrays = GL_FALSE;
+GLboolean __GLEW_INTEL_texture_scissor = GL_FALSE;
+GLboolean __GLEW_KTX_buffer_region = GL_FALSE;
+GLboolean __GLEW_MESAX_texture_stack = GL_FALSE;
+GLboolean __GLEW_MESA_pack_invert = GL_FALSE;
+GLboolean __GLEW_MESA_resize_buffers = GL_FALSE;
+GLboolean __GLEW_MESA_window_pos = GL_FALSE;
+GLboolean __GLEW_MESA_ycbcr_texture = GL_FALSE;
+GLboolean __GLEW_NV_blend_square = GL_FALSE;
+GLboolean __GLEW_NV_conditional_render = GL_FALSE;
+GLboolean __GLEW_NV_copy_depth_to_color = GL_FALSE;
+GLboolean __GLEW_NV_depth_buffer_float = GL_FALSE;
+GLboolean __GLEW_NV_depth_clamp = GL_FALSE;
+GLboolean __GLEW_NV_depth_range_unclamped = GL_FALSE;
+GLboolean __GLEW_NV_evaluators = GL_FALSE;
+GLboolean __GLEW_NV_explicit_multisample = GL_FALSE;
+GLboolean __GLEW_NV_fence = GL_FALSE;
+GLboolean __GLEW_NV_float_buffer = GL_FALSE;
+GLboolean __GLEW_NV_fog_distance = GL_FALSE;
+GLboolean __GLEW_NV_fragment_program = GL_FALSE;
+GLboolean __GLEW_NV_fragment_program2 = GL_FALSE;
+GLboolean __GLEW_NV_fragment_program4 = GL_FALSE;
+GLboolean __GLEW_NV_fragment_program_option = GL_FALSE;
+GLboolean __GLEW_NV_framebuffer_multisample_coverage = GL_FALSE;
+GLboolean __GLEW_NV_geometry_program4 = GL_FALSE;
+GLboolean __GLEW_NV_geometry_shader4 = GL_FALSE;
+GLboolean __GLEW_NV_gpu_program4 = GL_FALSE;
+GLboolean __GLEW_NV_half_float = GL_FALSE;
+GLboolean __GLEW_NV_light_max_exponent = GL_FALSE;
+GLboolean __GLEW_NV_multisample_filter_hint = GL_FALSE;
+GLboolean __GLEW_NV_occlusion_query = GL_FALSE;
+GLboolean __GLEW_NV_packed_depth_stencil = GL_FALSE;
+GLboolean __GLEW_NV_parameter_buffer_object = GL_FALSE;
+GLboolean __GLEW_NV_pixel_data_range = GL_FALSE;
+GLboolean __GLEW_NV_point_sprite = GL_FALSE;
+GLboolean __GLEW_NV_present_video = GL_FALSE;
+GLboolean __GLEW_NV_primitive_restart = GL_FALSE;
+GLboolean __GLEW_NV_register_combiners = GL_FALSE;
+GLboolean __GLEW_NV_register_combiners2 = GL_FALSE;
+GLboolean __GLEW_NV_texgen_emboss = GL_FALSE;
+GLboolean __GLEW_NV_texgen_reflection = GL_FALSE;
+GLboolean __GLEW_NV_texture_compression_vtc = GL_FALSE;
+GLboolean __GLEW_NV_texture_env_combine4 = GL_FALSE;
+GLboolean __GLEW_NV_texture_expand_normal = GL_FALSE;
+GLboolean __GLEW_NV_texture_rectangle = GL_FALSE;
+GLboolean __GLEW_NV_texture_shader = GL_FALSE;
+GLboolean __GLEW_NV_texture_shader2 = GL_FALSE;
+GLboolean __GLEW_NV_texture_shader3 = GL_FALSE;
+GLboolean __GLEW_NV_transform_feedback = GL_FALSE;
+GLboolean __GLEW_NV_vertex_array_range = GL_FALSE;
+GLboolean __GLEW_NV_vertex_array_range2 = GL_FALSE;
+GLboolean __GLEW_NV_vertex_program = GL_FALSE;
+GLboolean __GLEW_NV_vertex_program1_1 = GL_FALSE;
+GLboolean __GLEW_NV_vertex_program2 = GL_FALSE;
+GLboolean __GLEW_NV_vertex_program2_option = GL_FALSE;
+GLboolean __GLEW_NV_vertex_program3 = GL_FALSE;
+GLboolean __GLEW_NV_vertex_program4 = GL_FALSE;
+GLboolean __GLEW_OES_byte_coordinates = GL_FALSE;
+GLboolean __GLEW_OES_compressed_paletted_texture = GL_FALSE;
+GLboolean __GLEW_OES_read_format = GL_FALSE;
+GLboolean __GLEW_OES_single_precision = GL_FALSE;
+GLboolean __GLEW_OML_interlace = GL_FALSE;
+GLboolean __GLEW_OML_resample = GL_FALSE;
+GLboolean __GLEW_OML_subsample = GL_FALSE;
+GLboolean __GLEW_PGI_misc_hints = GL_FALSE;
+GLboolean __GLEW_PGI_vertex_hints = GL_FALSE;
+GLboolean __GLEW_REND_screen_coordinates = GL_FALSE;
+GLboolean __GLEW_S3_s3tc = GL_FALSE;
+GLboolean __GLEW_SGIS_color_range = GL_FALSE;
+GLboolean __GLEW_SGIS_detail_texture = GL_FALSE;
+GLboolean __GLEW_SGIS_fog_function = GL_FALSE;
+GLboolean __GLEW_SGIS_generate_mipmap = GL_FALSE;
+GLboolean __GLEW_SGIS_multisample = GL_FALSE;
+GLboolean __GLEW_SGIS_pixel_texture = GL_FALSE;
+GLboolean __GLEW_SGIS_point_line_texgen = GL_FALSE;
+GLboolean __GLEW_SGIS_sharpen_texture = GL_FALSE;
+GLboolean __GLEW_SGIS_texture4D = GL_FALSE;
+GLboolean __GLEW_SGIS_texture_border_clamp = GL_FALSE;
+GLboolean __GLEW_SGIS_texture_edge_clamp = GL_FALSE;
+GLboolean __GLEW_SGIS_texture_filter4 = GL_FALSE;
+GLboolean __GLEW_SGIS_texture_lod = GL_FALSE;
+GLboolean __GLEW_SGIS_texture_select = GL_FALSE;
+GLboolean __GLEW_SGIX_async = GL_FALSE;
+GLboolean __GLEW_SGIX_async_histogram = GL_FALSE;
+GLboolean __GLEW_SGIX_async_pixel = GL_FALSE;
+GLboolean __GLEW_SGIX_blend_alpha_minmax = GL_FALSE;
+GLboolean __GLEW_SGIX_clipmap = GL_FALSE;
+GLboolean __GLEW_SGIX_convolution_accuracy = GL_FALSE;
+GLboolean __GLEW_SGIX_depth_texture = GL_FALSE;
+GLboolean __GLEW_SGIX_flush_raster = GL_FALSE;
+GLboolean __GLEW_SGIX_fog_offset = GL_FALSE;
+GLboolean __GLEW_SGIX_fog_texture = GL_FALSE;
+GLboolean __GLEW_SGIX_fragment_specular_lighting = GL_FALSE;
+GLboolean __GLEW_SGIX_framezoom = GL_FALSE;
+GLboolean __GLEW_SGIX_interlace = GL_FALSE;
+GLboolean __GLEW_SGIX_ir_instrument1 = GL_FALSE;
+GLboolean __GLEW_SGIX_list_priority = GL_FALSE;
+GLboolean __GLEW_SGIX_pixel_texture = GL_FALSE;
+GLboolean __GLEW_SGIX_pixel_texture_bits = GL_FALSE;
+GLboolean __GLEW_SGIX_reference_plane = GL_FALSE;
+GLboolean __GLEW_SGIX_resample = GL_FALSE;
+GLboolean __GLEW_SGIX_shadow = GL_FALSE;
+GLboolean __GLEW_SGIX_shadow_ambient = GL_FALSE;
+GLboolean __GLEW_SGIX_sprite = GL_FALSE;
+GLboolean __GLEW_SGIX_tag_sample_buffer = GL_FALSE;
+GLboolean __GLEW_SGIX_texture_add_env = GL_FALSE;
+GLboolean __GLEW_SGIX_texture_coordinate_clamp = GL_FALSE;
+GLboolean __GLEW_SGIX_texture_lod_bias = GL_FALSE;
+GLboolean __GLEW_SGIX_texture_multi_buffer = GL_FALSE;
+GLboolean __GLEW_SGIX_texture_range = GL_FALSE;
+GLboolean __GLEW_SGIX_texture_scale_bias = GL_FALSE;
+GLboolean __GLEW_SGIX_vertex_preclip = GL_FALSE;
+GLboolean __GLEW_SGIX_vertex_preclip_hint = GL_FALSE;
+GLboolean __GLEW_SGIX_ycrcb = GL_FALSE;
+GLboolean __GLEW_SGI_color_matrix = GL_FALSE;
+GLboolean __GLEW_SGI_color_table = GL_FALSE;
+GLboolean __GLEW_SGI_texture_color_table = GL_FALSE;
+GLboolean __GLEW_SUNX_constant_data = GL_FALSE;
+GLboolean __GLEW_SUN_convolution_border_modes = GL_FALSE;
+GLboolean __GLEW_SUN_global_alpha = GL_FALSE;
+GLboolean __GLEW_SUN_mesh_array = GL_FALSE;
+GLboolean __GLEW_SUN_read_video_pixels = GL_FALSE;
+GLboolean __GLEW_SUN_slice_accum = GL_FALSE;
+GLboolean __GLEW_SUN_triangle_list = GL_FALSE;
+GLboolean __GLEW_SUN_vertex = GL_FALSE;
+GLboolean __GLEW_WIN_phong_shading = GL_FALSE;
+GLboolean __GLEW_WIN_specular_fog = GL_FALSE;
+GLboolean __GLEW_WIN_swap_hint = GL_FALSE;
+
+#endif /* !GLEW_MX */
+
+#ifdef GL_VERSION_1_2
+
+static GLboolean _glewInit_GL_VERSION_1_2 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glCopyTexSubImage3D = (PFNGLCOPYTEXSUBIMAGE3DPROC)glewGetProcAddress((const GLubyte*)"glCopyTexSubImage3D")) == NULL) || r;
+  r = ((glDrawRangeElements = (PFNGLDRAWRANGEELEMENTSPROC)glewGetProcAddress((const GLubyte*)"glDrawRangeElements")) == NULL) || r;
+  r = ((glTexImage3D = (PFNGLTEXIMAGE3DPROC)glewGetProcAddress((const GLubyte*)"glTexImage3D")) == NULL) || r;
+  r = ((glTexSubImage3D = (PFNGLTEXSUBIMAGE3DPROC)glewGetProcAddress((const GLubyte*)"glTexSubImage3D")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_VERSION_1_2 */
+
+#ifdef GL_VERSION_1_3
+
+static GLboolean _glewInit_GL_VERSION_1_3 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glActiveTexture = (PFNGLACTIVETEXTUREPROC)glewGetProcAddress((const GLubyte*)"glActiveTexture")) == NULL) || r;
+  r = ((glClientActiveTexture = (PFNGLCLIENTACTIVETEXTUREPROC)glewGetProcAddress((const GLubyte*)"glClientActiveTexture")) == NULL) || r;
+  r = ((glCompressedTexImage1D = (PFNGLCOMPRESSEDTEXIMAGE1DPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexImage1D")) == NULL) || r;
+  r = ((glCompressedTexImage2D = (PFNGLCOMPRESSEDTEXIMAGE2DPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexImage2D")) == NULL) || r;
+  r = ((glCompressedTexImage3D = (PFNGLCOMPRESSEDTEXIMAGE3DPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexImage3D")) == NULL) || r;
+  r = ((glCompressedTexSubImage1D = (PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexSubImage1D")) == NULL) || r;
+  r = ((glCompressedTexSubImage2D = (PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexSubImage2D")) == NULL) || r;
+  r = ((glCompressedTexSubImage3D = (PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexSubImage3D")) == NULL) || r;
+  r = ((glGetCompressedTexImage = (PFNGLGETCOMPRESSEDTEXIMAGEPROC)glewGetProcAddress((const GLubyte*)"glGetCompressedTexImage")) == NULL) || r;
+  r = ((glLoadTransposeMatrixd = (PFNGLLOADTRANSPOSEMATRIXDPROC)glewGetProcAddress((const GLubyte*)"glLoadTransposeMatrixd")) == NULL) || r;
+  r = ((glLoadTransposeMatrixf = (PFNGLLOADTRANSPOSEMATRIXFPROC)glewGetProcAddress((const GLubyte*)"glLoadTransposeMatrixf")) == NULL) || r;
+  r = ((glMultTransposeMatrixd = (PFNGLMULTTRANSPOSEMATRIXDPROC)glewGetProcAddress((const GLubyte*)"glMultTransposeMatrixd")) == NULL) || r;
+  r = ((glMultTransposeMatrixf = (PFNGLMULTTRANSPOSEMATRIXFPROC)glewGetProcAddress((const GLubyte*)"glMultTransposeMatrixf")) == NULL) || r;
+  r = ((glMultiTexCoord1d = (PFNGLMULTITEXCOORD1DPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1d")) == NULL) || r;
+  r = ((glMultiTexCoord1dv = (PFNGLMULTITEXCOORD1DVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1dv")) == NULL) || r;
+  r = ((glMultiTexCoord1f = (PFNGLMULTITEXCOORD1FPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1f")) == NULL) || r;
+  r = ((glMultiTexCoord1fv = (PFNGLMULTITEXCOORD1FVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1fv")) == NULL) || r;
+  r = ((glMultiTexCoord1i = (PFNGLMULTITEXCOORD1IPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1i")) == NULL) || r;
+  r = ((glMultiTexCoord1iv = (PFNGLMULTITEXCOORD1IVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1iv")) == NULL) || r;
+  r = ((glMultiTexCoord1s = (PFNGLMULTITEXCOORD1SPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1s")) == NULL) || r;
+  r = ((glMultiTexCoord1sv = (PFNGLMULTITEXCOORD1SVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1sv")) == NULL) || r;
+  r = ((glMultiTexCoord2d = (PFNGLMULTITEXCOORD2DPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2d")) == NULL) || r;
+  r = ((glMultiTexCoord2dv = (PFNGLMULTITEXCOORD2DVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2dv")) == NULL) || r;
+  r = ((glMultiTexCoord2f = (PFNGLMULTITEXCOORD2FPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2f")) == NULL) || r;
+  r = ((glMultiTexCoord2fv = (PFNGLMULTITEXCOORD2FVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2fv")) == NULL) || r;
+  r = ((glMultiTexCoord2i = (PFNGLMULTITEXCOORD2IPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2i")) == NULL) || r;
+  r = ((glMultiTexCoord2iv = (PFNGLMULTITEXCOORD2IVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2iv")) == NULL) || r;
+  r = ((glMultiTexCoord2s = (PFNGLMULTITEXCOORD2SPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2s")) == NULL) || r;
+  r = ((glMultiTexCoord2sv = (PFNGLMULTITEXCOORD2SVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2sv")) == NULL) || r;
+  r = ((glMultiTexCoord3d = (PFNGLMULTITEXCOORD3DPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3d")) == NULL) || r;
+  r = ((glMultiTexCoord3dv = (PFNGLMULTITEXCOORD3DVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3dv")) == NULL) || r;
+  r = ((glMultiTexCoord3f = (PFNGLMULTITEXCOORD3FPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3f")) == NULL) || r;
+  r = ((glMultiTexCoord3fv = (PFNGLMULTITEXCOORD3FVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3fv")) == NULL) || r;
+  r = ((glMultiTexCoord3i = (PFNGLMULTITEXCOORD3IPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3i")) == NULL) || r;
+  r = ((glMultiTexCoord3iv = (PFNGLMULTITEXCOORD3IVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3iv")) == NULL) || r;
+  r = ((glMultiTexCoord3s = (PFNGLMULTITEXCOORD3SPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3s")) == NULL) || r;
+  r = ((glMultiTexCoord3sv = (PFNGLMULTITEXCOORD3SVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3sv")) == NULL) || r;
+  r = ((glMultiTexCoord4d = (PFNGLMULTITEXCOORD4DPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4d")) == NULL) || r;
+  r = ((glMultiTexCoord4dv = (PFNGLMULTITEXCOORD4DVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4dv")) == NULL) || r;
+  r = ((glMultiTexCoord4f = (PFNGLMULTITEXCOORD4FPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4f")) == NULL) || r;
+  r = ((glMultiTexCoord4fv = (PFNGLMULTITEXCOORD4FVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4fv")) == NULL) || r;
+  r = ((glMultiTexCoord4i = (PFNGLMULTITEXCOORD4IPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4i")) == NULL) || r;
+  r = ((glMultiTexCoord4iv = (PFNGLMULTITEXCOORD4IVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4iv")) == NULL) || r;
+  r = ((glMultiTexCoord4s = (PFNGLMULTITEXCOORD4SPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4s")) == NULL) || r;
+  r = ((glMultiTexCoord4sv = (PFNGLMULTITEXCOORD4SVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4sv")) == NULL) || r;
+  r = ((glSampleCoverage = (PFNGLSAMPLECOVERAGEPROC)glewGetProcAddress((const GLubyte*)"glSampleCoverage")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_VERSION_1_3 */
+
+#ifdef GL_VERSION_1_4
+
+static GLboolean _glewInit_GL_VERSION_1_4 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBlendColor = (PFNGLBLENDCOLORPROC)glewGetProcAddress((const GLubyte*)"glBlendColor")) == NULL) || r;
+  r = ((glBlendEquation = (PFNGLBLENDEQUATIONPROC)glewGetProcAddress((const GLubyte*)"glBlendEquation")) == NULL) || r;
+  r = ((glBlendFuncSeparate = (PFNGLBLENDFUNCSEPARATEPROC)glewGetProcAddress((const GLubyte*)"glBlendFuncSeparate")) == NULL) || r;
+  r = ((glFogCoordPointer = (PFNGLFOGCOORDPOINTERPROC)glewGetProcAddress((const GLubyte*)"glFogCoordPointer")) == NULL) || r;
+  r = ((glFogCoordd = (PFNGLFOGCOORDDPROC)glewGetProcAddress((const GLubyte*)"glFogCoordd")) == NULL) || r;
+  r = ((glFogCoorddv = (PFNGLFOGCOORDDVPROC)glewGetProcAddress((const GLubyte*)"glFogCoorddv")) == NULL) || r;
+  r = ((glFogCoordf = (PFNGLFOGCOORDFPROC)glewGetProcAddress((const GLubyte*)"glFogCoordf")) == NULL) || r;
+  r = ((glFogCoordfv = (PFNGLFOGCOORDFVPROC)glewGetProcAddress((const GLubyte*)"glFogCoordfv")) == NULL) || r;
+  r = ((glMultiDrawArrays = (PFNGLMULTIDRAWARRAYSPROC)glewGetProcAddress((const GLubyte*)"glMultiDrawArrays")) == NULL) || r;
+  r = ((glMultiDrawElements = (PFNGLMULTIDRAWELEMENTSPROC)glewGetProcAddress((const GLubyte*)"glMultiDrawElements")) == NULL) || r;
+  r = ((glPointParameterf = (PFNGLPOINTPARAMETERFPROC)glewGetProcAddress((const GLubyte*)"glPointParameterf")) == NULL) || r;
+  r = ((glPointParameterfv = (PFNGLPOINTPARAMETERFVPROC)glewGetProcAddress((const GLubyte*)"glPointParameterfv")) == NULL) || r;
+  r = ((glPointParameteri = (PFNGLPOINTPARAMETERIPROC)glewGetProcAddress((const GLubyte*)"glPointParameteri")) == NULL) || r;
+  r = ((glPointParameteriv = (PFNGLPOINTPARAMETERIVPROC)glewGetProcAddress((const GLubyte*)"glPointParameteriv")) == NULL) || r;
+  r = ((glSecondaryColor3b = (PFNGLSECONDARYCOLOR3BPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3b")) == NULL) || r;
+  r = ((glSecondaryColor3bv = (PFNGLSECONDARYCOLOR3BVPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3bv")) == NULL) || r;
+  r = ((glSecondaryColor3d = (PFNGLSECONDARYCOLOR3DPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3d")) == NULL) || r;
+  r = ((glSecondaryColor3dv = (PFNGLSECONDARYCOLOR3DVPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3dv")) == NULL) || r;
+  r = ((glSecondaryColor3f = (PFNGLSECONDARYCOLOR3FPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3f")) == NULL) || r;
+  r = ((glSecondaryColor3fv = (PFNGLSECONDARYCOLOR3FVPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3fv")) == NULL) || r;
+  r = ((glSecondaryColor3i = (PFNGLSECONDARYCOLOR3IPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3i")) == NULL) || r;
+  r = ((glSecondaryColor3iv = (PFNGLSECONDARYCOLOR3IVPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3iv")) == NULL) || r;
+  r = ((glSecondaryColor3s = (PFNGLSECONDARYCOLOR3SPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3s")) == NULL) || r;
+  r = ((glSecondaryColor3sv = (PFNGLSECONDARYCOLOR3SVPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3sv")) == NULL) || r;
+  r = ((glSecondaryColor3ub = (PFNGLSECONDARYCOLOR3UBPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3ub")) == NULL) || r;
+  r = ((glSecondaryColor3ubv = (PFNGLSECONDARYCOLOR3UBVPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3ubv")) == NULL) || r;
+  r = ((glSecondaryColor3ui = (PFNGLSECONDARYCOLOR3UIPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3ui")) == NULL) || r;
+  r = ((glSecondaryColor3uiv = (PFNGLSECONDARYCOLOR3UIVPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3uiv")) == NULL) || r;
+  r = ((glSecondaryColor3us = (PFNGLSECONDARYCOLOR3USPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3us")) == NULL) || r;
+  r = ((glSecondaryColor3usv = (PFNGLSECONDARYCOLOR3USVPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3usv")) == NULL) || r;
+  r = ((glSecondaryColorPointer = (PFNGLSECONDARYCOLORPOINTERPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColorPointer")) == NULL) || r;
+  r = ((glWindowPos2d = (PFNGLWINDOWPOS2DPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2d")) == NULL) || r;
+  r = ((glWindowPos2dv = (PFNGLWINDOWPOS2DVPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2dv")) == NULL) || r;
+  r = ((glWindowPos2f = (PFNGLWINDOWPOS2FPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2f")) == NULL) || r;
+  r = ((glWindowPos2fv = (PFNGLWINDOWPOS2FVPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2fv")) == NULL) || r;
+  r = ((glWindowPos2i = (PFNGLWINDOWPOS2IPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2i")) == NULL) || r;
+  r = ((glWindowPos2iv = (PFNGLWINDOWPOS2IVPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2iv")) == NULL) || r;
+  r = ((glWindowPos2s = (PFNGLWINDOWPOS2SPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2s")) == NULL) || r;
+  r = ((glWindowPos2sv = (PFNGLWINDOWPOS2SVPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2sv")) == NULL) || r;
+  r = ((glWindowPos3d = (PFNGLWINDOWPOS3DPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3d")) == NULL) || r;
+  r = ((glWindowPos3dv = (PFNGLWINDOWPOS3DVPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3dv")) == NULL) || r;
+  r = ((glWindowPos3f = (PFNGLWINDOWPOS3FPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3f")) == NULL) || r;
+  r = ((glWindowPos3fv = (PFNGLWINDOWPOS3FVPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3fv")) == NULL) || r;
+  r = ((glWindowPos3i = (PFNGLWINDOWPOS3IPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3i")) == NULL) || r;
+  r = ((glWindowPos3iv = (PFNGLWINDOWPOS3IVPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3iv")) == NULL) || r;
+  r = ((glWindowPos3s = (PFNGLWINDOWPOS3SPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3s")) == NULL) || r;
+  r = ((glWindowPos3sv = (PFNGLWINDOWPOS3SVPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3sv")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_VERSION_1_4 */
+
+#ifdef GL_VERSION_1_5
+
+static GLboolean _glewInit_GL_VERSION_1_5 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBeginQuery = (PFNGLBEGINQUERYPROC)glewGetProcAddress((const GLubyte*)"glBeginQuery")) == NULL) || r;
+  r = ((glBindBuffer = (PFNGLBINDBUFFERPROC)glewGetProcAddress((const GLubyte*)"glBindBuffer")) == NULL) || r;
+  r = ((glBufferData = (PFNGLBUFFERDATAPROC)glewGetProcAddress((const GLubyte*)"glBufferData")) == NULL) || r;
+  r = ((glBufferSubData = (PFNGLBUFFERSUBDATAPROC)glewGetProcAddress((const GLubyte*)"glBufferSubData")) == NULL) || r;
+  r = ((glDeleteBuffers = (PFNGLDELETEBUFFERSPROC)glewGetProcAddress((const GLubyte*)"glDeleteBuffers")) == NULL) || r;
+  r = ((glDeleteQueries = (PFNGLDELETEQUERIESPROC)glewGetProcAddress((const GLubyte*)"glDeleteQueries")) == NULL) || r;
+  r = ((glEndQuery = (PFNGLENDQUERYPROC)glewGetProcAddress((const GLubyte*)"glEndQuery")) == NULL) || r;
+  r = ((glGenBuffers = (PFNGLGENBUFFERSPROC)glewGetProcAddress((const GLubyte*)"glGenBuffers")) == NULL) || r;
+  r = ((glGenQueries = (PFNGLGENQUERIESPROC)glewGetProcAddress((const GLubyte*)"glGenQueries")) == NULL) || r;
+  r = ((glGetBufferParameteriv = (PFNGLGETBUFFERPARAMETERIVPROC)glewGetProcAddress((const GLubyte*)"glGetBufferParameteriv")) == NULL) || r;
+  r = ((glGetBufferPointerv = (PFNGLGETBUFFERPOINTERVPROC)glewGetProcAddress((const GLubyte*)"glGetBufferPointerv")) == NULL) || r;
+  r = ((glGetBufferSubData = (PFNGLGETBUFFERSUBDATAPROC)glewGetProcAddress((const GLubyte*)"glGetBufferSubData")) == NULL) || r;
+  r = ((glGetQueryObjectiv = (PFNGLGETQUERYOBJECTIVPROC)glewGetProcAddress((const GLubyte*)"glGetQueryObjectiv")) == NULL) || r;
+  r = ((glGetQueryObjectuiv = (PFNGLGETQUERYOBJECTUIVPROC)glewGetProcAddress((const GLubyte*)"glGetQueryObjectuiv")) == NULL) || r;
+  r = ((glGetQueryiv = (PFNGLGETQUERYIVPROC)glewGetProcAddress((const GLubyte*)"glGetQueryiv")) == NULL) || r;
+  r = ((glIsBuffer = (PFNGLISBUFFERPROC)glewGetProcAddress((const GLubyte*)"glIsBuffer")) == NULL) || r;
+  r = ((glIsQuery = (PFNGLISQUERYPROC)glewGetProcAddress((const GLubyte*)"glIsQuery")) == NULL) || r;
+  r = ((glMapBuffer = (PFNGLMAPBUFFERPROC)glewGetProcAddress((const GLubyte*)"glMapBuffer")) == NULL) || r;
+  r = ((glUnmapBuffer = (PFNGLUNMAPBUFFERPROC)glewGetProcAddress((const GLubyte*)"glUnmapBuffer")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_VERSION_1_5 */
+
+#ifdef GL_VERSION_2_0
+
+static GLboolean _glewInit_GL_VERSION_2_0 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glAttachShader = (PFNGLATTACHSHADERPROC)glewGetProcAddress((const GLubyte*)"glAttachShader")) == NULL) || r;
+  r = ((glBindAttribLocation = (PFNGLBINDATTRIBLOCATIONPROC)glewGetProcAddress((const GLubyte*)"glBindAttribLocation")) == NULL) || r;
+  r = ((glBlendEquationSeparate = (PFNGLBLENDEQUATIONSEPARATEPROC)glewGetProcAddress((const GLubyte*)"glBlendEquationSeparate")) == NULL) || r;
+  r = ((glCompileShader = (PFNGLCOMPILESHADERPROC)glewGetProcAddress((const GLubyte*)"glCompileShader")) == NULL) || r;
+  r = ((glCreateProgram = (PFNGLCREATEPROGRAMPROC)glewGetProcAddress((const GLubyte*)"glCreateProgram")) == NULL) || r;
+  r = ((glCreateShader = (PFNGLCREATESHADERPROC)glewGetProcAddress((const GLubyte*)"glCreateShader")) == NULL) || r;
+  r = ((glDeleteProgram = (PFNGLDELETEPROGRAMPROC)glewGetProcAddress((const GLubyte*)"glDeleteProgram")) == NULL) || r;
+  r = ((glDeleteShader = (PFNGLDELETESHADERPROC)glewGetProcAddress((const GLubyte*)"glDeleteShader")) == NULL) || r;
+  r = ((glDetachShader = (PFNGLDETACHSHADERPROC)glewGetProcAddress((const GLubyte*)"glDetachShader")) == NULL) || r;
+  r = ((glDisableVertexAttribArray = (PFNGLDISABLEVERTEXATTRIBARRAYPROC)glewGetProcAddress((const GLubyte*)"glDisableVertexAttribArray")) == NULL) || r;
+  r = ((glDrawBuffers = (PFNGLDRAWBUFFERSPROC)glewGetProcAddress((const GLubyte*)"glDrawBuffers")) == NULL) || r;
+  r = ((glEnableVertexAttribArray = (PFNGLENABLEVERTEXATTRIBARRAYPROC)glewGetProcAddress((const GLubyte*)"glEnableVertexAttribArray")) == NULL) || r;
+  r = ((glGetActiveAttrib = (PFNGLGETACTIVEATTRIBPROC)glewGetProcAddress((const GLubyte*)"glGetActiveAttrib")) == NULL) || r;
+  r = ((glGetActiveUniform = (PFNGLGETACTIVEUNIFORMPROC)glewGetProcAddress((const GLubyte*)"glGetActiveUniform")) == NULL) || r;
+  r = ((glGetAttachedShaders = (PFNGLGETATTACHEDSHADERSPROC)glewGetProcAddress((const GLubyte*)"glGetAttachedShaders")) == NULL) || r;
+  r = ((glGetAttribLocation = (PFNGLGETATTRIBLOCATIONPROC)glewGetProcAddress((const GLubyte*)"glGetAttribLocation")) == NULL) || r;
+  r = ((glGetProgramInfoLog = (PFNGLGETPROGRAMINFOLOGPROC)glewGetProcAddress((const GLubyte*)"glGetProgramInfoLog")) == NULL) || r;
+  r = ((glGetProgramiv = (PFNGLGETPROGRAMIVPROC)glewGetProcAddress((const GLubyte*)"glGetProgramiv")) == NULL) || r;
+  r = ((glGetShaderInfoLog = (PFNGLGETSHADERINFOLOGPROC)glewGetProcAddress((const GLubyte*)"glGetShaderInfoLog")) == NULL) || r;
+  r = ((glGetShaderSource = (PFNGLGETSHADERSOURCEPROC)glewGetProcAddress((const GLubyte*)"glGetShaderSource")) == NULL) || r;
+  r = ((glGetShaderiv = (PFNGLGETSHADERIVPROC)glewGetProcAddress((const GLubyte*)"glGetShaderiv")) == NULL) || r;
+  r = ((glGetUniformLocation = (PFNGLGETUNIFORMLOCATIONPROC)glewGetProcAddress((const GLubyte*)"glGetUniformLocation")) == NULL) || r;
+  r = ((glGetUniformfv = (PFNGLGETUNIFORMFVPROC)glewGetProcAddress((const GLubyte*)"glGetUniformfv")) == NULL) || r;
+  r = ((glGetUniformiv = (PFNGLGETUNIFORMIVPROC)glewGetProcAddress((const GLubyte*)"glGetUniformiv")) == NULL) || r;
+  r = ((glGetVertexAttribPointerv = (PFNGLGETVERTEXATTRIBPOINTERVPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribPointerv")) == NULL) || r;
+  r = ((glGetVertexAttribdv = (PFNGLGETVERTEXATTRIBDVPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribdv")) == NULL) || r;
+  r = ((glGetVertexAttribfv = (PFNGLGETVERTEXATTRIBFVPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribfv")) == NULL) || r;
+  r = ((glGetVertexAttribiv = (PFNGLGETVERTEXATTRIBIVPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribiv")) == NULL) || r;
+  r = ((glIsProgram = (PFNGLISPROGRAMPROC)glewGetProcAddress((const GLubyte*)"glIsProgram")) == NULL) || r;
+  r = ((glIsShader = (PFNGLISSHADERPROC)glewGetProcAddress((const GLubyte*)"glIsShader")) == NULL) || r;
+  r = ((glLinkProgram = (PFNGLLINKPROGRAMPROC)glewGetProcAddress((const GLubyte*)"glLinkProgram")) == NULL) || r;
+  r = ((glShaderSource = (PFNGLSHADERSOURCEPROC)glewGetProcAddress((const GLubyte*)"glShaderSource")) == NULL) || r;
+  r = ((glStencilFuncSeparate = (PFNGLSTENCILFUNCSEPARATEPROC)glewGetProcAddress((const GLubyte*)"glStencilFuncSeparate")) == NULL) || r;
+  r = ((glStencilMaskSeparate = (PFNGLSTENCILMASKSEPARATEPROC)glewGetProcAddress((const GLubyte*)"glStencilMaskSeparate")) == NULL) || r;
+  r = ((glStencilOpSeparate = (PFNGLSTENCILOPSEPARATEPROC)glewGetProcAddress((const GLubyte*)"glStencilOpSeparate")) == NULL) || r;
+  r = ((glUniform1f = (PFNGLUNIFORM1FPROC)glewGetProcAddress((const GLubyte*)"glUniform1f")) == NULL) || r;
+  r = ((glUniform1fv = (PFNGLUNIFORM1FVPROC)glewGetProcAddress((const GLubyte*)"glUniform1fv")) == NULL) || r;
+  r = ((glUniform1i = (PFNGLUNIFORM1IPROC)glewGetProcAddress((const GLubyte*)"glUniform1i")) == NULL) || r;
+  r = ((glUniform1iv = (PFNGLUNIFORM1IVPROC)glewGetProcAddress((const GLubyte*)"glUniform1iv")) == NULL) || r;
+  r = ((glUniform2f = (PFNGLUNIFORM2FPROC)glewGetProcAddress((const GLubyte*)"glUniform2f")) == NULL) || r;
+  r = ((glUniform2fv = (PFNGLUNIFORM2FVPROC)glewGetProcAddress((const GLubyte*)"glUniform2fv")) == NULL) || r;
+  r = ((glUniform2i = (PFNGLUNIFORM2IPROC)glewGetProcAddress((const GLubyte*)"glUniform2i")) == NULL) || r;
+  r = ((glUniform2iv = (PFNGLUNIFORM2IVPROC)glewGetProcAddress((const GLubyte*)"glUniform2iv")) == NULL) || r;
+  r = ((glUniform3f = (PFNGLUNIFORM3FPROC)glewGetProcAddress((const GLubyte*)"glUniform3f")) == NULL) || r;
+  r = ((glUniform3fv = (PFNGLUNIFORM3FVPROC)glewGetProcAddress((const GLubyte*)"glUniform3fv")) == NULL) || r;
+  r = ((glUniform3i = (PFNGLUNIFORM3IPROC)glewGetProcAddress((const GLubyte*)"glUniform3i")) == NULL) || r;
+  r = ((glUniform3iv = (PFNGLUNIFORM3IVPROC)glewGetProcAddress((const GLubyte*)"glUniform3iv")) == NULL) || r;
+  r = ((glUniform4f = (PFNGLUNIFORM4FPROC)glewGetProcAddress((const GLubyte*)"glUniform4f")) == NULL) || r;
+  r = ((glUniform4fv = (PFNGLUNIFORM4FVPROC)glewGetProcAddress((const GLubyte*)"glUniform4fv")) == NULL) || r;
+  r = ((glUniform4i = (PFNGLUNIFORM4IPROC)glewGetProcAddress((const GLubyte*)"glUniform4i")) == NULL) || r;
+  r = ((glUniform4iv = (PFNGLUNIFORM4IVPROC)glewGetProcAddress((const GLubyte*)"glUniform4iv")) == NULL) || r;
+  r = ((glUniformMatrix2fv = (PFNGLUNIFORMMATRIX2FVPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix2fv")) == NULL) || r;
+  r = ((glUniformMatrix3fv = (PFNGLUNIFORMMATRIX3FVPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix3fv")) == NULL) || r;
+  r = ((glUniformMatrix4fv = (PFNGLUNIFORMMATRIX4FVPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix4fv")) == NULL) || r;
+  r = ((glUseProgram = (PFNGLUSEPROGRAMPROC)glewGetProcAddress((const GLubyte*)"glUseProgram")) == NULL) || r;
+  r = ((glValidateProgram = (PFNGLVALIDATEPROGRAMPROC)glewGetProcAddress((const GLubyte*)"glValidateProgram")) == NULL) || r;
+  r = ((glVertexAttrib1d = (PFNGLVERTEXATTRIB1DPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1d")) == NULL) || r;
+  r = ((glVertexAttrib1dv = (PFNGLVERTEXATTRIB1DVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1dv")) == NULL) || r;
+  r = ((glVertexAttrib1f = (PFNGLVERTEXATTRIB1FPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1f")) == NULL) || r;
+  r = ((glVertexAttrib1fv = (PFNGLVERTEXATTRIB1FVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1fv")) == NULL) || r;
+  r = ((glVertexAttrib1s = (PFNGLVERTEXATTRIB1SPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1s")) == NULL) || r;
+  r = ((glVertexAttrib1sv = (PFNGLVERTEXATTRIB1SVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1sv")) == NULL) || r;
+  r = ((glVertexAttrib2d = (PFNGLVERTEXATTRIB2DPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2d")) == NULL) || r;
+  r = ((glVertexAttrib2dv = (PFNGLVERTEXATTRIB2DVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2dv")) == NULL) || r;
+  r = ((glVertexAttrib2f = (PFNGLVERTEXATTRIB2FPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2f")) == NULL) || r;
+  r = ((glVertexAttrib2fv = (PFNGLVERTEXATTRIB2FVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2fv")) == NULL) || r;
+  r = ((glVertexAttrib2s = (PFNGLVERTEXATTRIB2SPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2s")) == NULL) || r;
+  r = ((glVertexAttrib2sv = (PFNGLVERTEXATTRIB2SVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2sv")) == NULL) || r;
+  r = ((glVertexAttrib3d = (PFNGLVERTEXATTRIB3DPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3d")) == NULL) || r;
+  r = ((glVertexAttrib3dv = (PFNGLVERTEXATTRIB3DVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3dv")) == NULL) || r;
+  r = ((glVertexAttrib3f = (PFNGLVERTEXATTRIB3FPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3f")) == NULL) || r;
+  r = ((glVertexAttrib3fv = (PFNGLVERTEXATTRIB3FVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3fv")) == NULL) || r;
+  r = ((glVertexAttrib3s = (PFNGLVERTEXATTRIB3SPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3s")) == NULL) || r;
+  r = ((glVertexAttrib3sv = (PFNGLVERTEXATTRIB3SVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3sv")) == NULL) || r;
+  r = ((glVertexAttrib4Nbv = (PFNGLVERTEXATTRIB4NBVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4Nbv")) == NULL) || r;
+  r = ((glVertexAttrib4Niv = (PFNGLVERTEXATTRIB4NIVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4Niv")) == NULL) || r;
+  r = ((glVertexAttrib4Nsv = (PFNGLVERTEXATTRIB4NSVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4Nsv")) == NULL) || r;
+  r = ((glVertexAttrib4Nub = (PFNGLVERTEXATTRIB4NUBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4Nub")) == NULL) || r;
+  r = ((glVertexAttrib4Nubv = (PFNGLVERTEXATTRIB4NUBVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4Nubv")) == NULL) || r;
+  r = ((glVertexAttrib4Nuiv = (PFNGLVERTEXATTRIB4NUIVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4Nuiv")) == NULL) || r;
+  r = ((glVertexAttrib4Nusv = (PFNGLVERTEXATTRIB4NUSVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4Nusv")) == NULL) || r;
+  r = ((glVertexAttrib4bv = (PFNGLVERTEXATTRIB4BVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4bv")) == NULL) || r;
+  r = ((glVertexAttrib4d = (PFNGLVERTEXATTRIB4DPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4d")) == NULL) || r;
+  r = ((glVertexAttrib4dv = (PFNGLVERTEXATTRIB4DVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4dv")) == NULL) || r;
+  r = ((glVertexAttrib4f = (PFNGLVERTEXATTRIB4FPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4f")) == NULL) || r;
+  r = ((glVertexAttrib4fv = (PFNGLVERTEXATTRIB4FVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4fv")) == NULL) || r;
+  r = ((glVertexAttrib4iv = (PFNGLVERTEXATTRIB4IVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4iv")) == NULL) || r;
+  r = ((glVertexAttrib4s = (PFNGLVERTEXATTRIB4SPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4s")) == NULL) || r;
+  r = ((glVertexAttrib4sv = (PFNGLVERTEXATTRIB4SVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4sv")) == NULL) || r;
+  r = ((glVertexAttrib4ubv = (PFNGLVERTEXATTRIB4UBVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4ubv")) == NULL) || r;
+  r = ((glVertexAttrib4uiv = (PFNGLVERTEXATTRIB4UIVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4uiv")) == NULL) || r;
+  r = ((glVertexAttrib4usv = (PFNGLVERTEXATTRIB4USVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4usv")) == NULL) || r;
+  r = ((glVertexAttribPointer = (PFNGLVERTEXATTRIBPOINTERPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribPointer")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_VERSION_2_0 */
+
+#ifdef GL_VERSION_2_1
+
+static GLboolean _glewInit_GL_VERSION_2_1 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glUniformMatrix2x3fv = (PFNGLUNIFORMMATRIX2X3FVPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix2x3fv")) == NULL) || r;
+  r = ((glUniformMatrix2x4fv = (PFNGLUNIFORMMATRIX2X4FVPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix2x4fv")) == NULL) || r;
+  r = ((glUniformMatrix3x2fv = (PFNGLUNIFORMMATRIX3X2FVPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix3x2fv")) == NULL) || r;
+  r = ((glUniformMatrix3x4fv = (PFNGLUNIFORMMATRIX3X4FVPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix3x4fv")) == NULL) || r;
+  r = ((glUniformMatrix4x2fv = (PFNGLUNIFORMMATRIX4X2FVPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix4x2fv")) == NULL) || r;
+  r = ((glUniformMatrix4x3fv = (PFNGLUNIFORMMATRIX4X3FVPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix4x3fv")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_VERSION_2_1 */
+
+#ifdef GL_VERSION_3_0
+
+static GLboolean _glewInit_GL_VERSION_3_0 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBeginConditionalRender = (PFNGLBEGINCONDITIONALRENDERPROC)glewGetProcAddress((const GLubyte*)"glBeginConditionalRender")) == NULL) || r;
+  r = ((glBeginTransformFeedback = (PFNGLBEGINTRANSFORMFEEDBACKPROC)glewGetProcAddress((const GLubyte*)"glBeginTransformFeedback")) == NULL) || r;
+  r = ((glBindBufferBase = (PFNGLBINDBUFFERBASEPROC)glewGetProcAddress((const GLubyte*)"glBindBufferBase")) == NULL) || r;
+  r = ((glBindBufferRange = (PFNGLBINDBUFFERRANGEPROC)glewGetProcAddress((const GLubyte*)"glBindBufferRange")) == NULL) || r;
+  r = ((glBindFragDataLocation = (PFNGLBINDFRAGDATALOCATIONPROC)glewGetProcAddress((const GLubyte*)"glBindFragDataLocation")) == NULL) || r;
+  r = ((glClampColor = (PFNGLCLAMPCOLORPROC)glewGetProcAddress((const GLubyte*)"glClampColor")) == NULL) || r;
+  r = ((glClearBufferfi = (PFNGLCLEARBUFFERFIPROC)glewGetProcAddress((const GLubyte*)"glClearBufferfi")) == NULL) || r;
+  r = ((glClearBufferfv = (PFNGLCLEARBUFFERFVPROC)glewGetProcAddress((const GLubyte*)"glClearBufferfv")) == NULL) || r;
+  r = ((glClearBufferiv = (PFNGLCLEARBUFFERIVPROC)glewGetProcAddress((const GLubyte*)"glClearBufferiv")) == NULL) || r;
+  r = ((glClearBufferuiv = (PFNGLCLEARBUFFERUIVPROC)glewGetProcAddress((const GLubyte*)"glClearBufferuiv")) == NULL) || r;
+  r = ((glColorMaski = (PFNGLCOLORMASKIPROC)glewGetProcAddress((const GLubyte*)"glColorMaski")) == NULL) || r;
+  r = ((glDisablei = (PFNGLDISABLEIPROC)glewGetProcAddress((const GLubyte*)"glDisablei")) == NULL) || r;
+  r = ((glEnablei = (PFNGLENABLEIPROC)glewGetProcAddress((const GLubyte*)"glEnablei")) == NULL) || r;
+  r = ((glEndConditionalRender = (PFNGLENDCONDITIONALRENDERPROC)glewGetProcAddress((const GLubyte*)"glEndConditionalRender")) == NULL) || r;
+  r = ((glEndTransformFeedback = (PFNGLENDTRANSFORMFEEDBACKPROC)glewGetProcAddress((const GLubyte*)"glEndTransformFeedback")) == NULL) || r;
+  r = ((glGetBooleani_v = (PFNGLGETBOOLEANI_VPROC)glewGetProcAddress((const GLubyte*)"glGetBooleani_v")) == NULL) || r;
+  r = ((glGetFragDataLocation = (PFNGLGETFRAGDATALOCATIONPROC)glewGetProcAddress((const GLubyte*)"glGetFragDataLocation")) == NULL) || r;
+  r = ((glGetIntegeri_v = (PFNGLGETINTEGERI_VPROC)glewGetProcAddress((const GLubyte*)"glGetIntegeri_v")) == NULL) || r;
+  r = ((glGetStringi = (PFNGLGETSTRINGIPROC)glewGetProcAddress((const GLubyte*)"glGetStringi")) == NULL) || r;
+  r = ((glGetTexParameterIiv = (PFNGLGETTEXPARAMETERIIVPROC)glewGetProcAddress((const GLubyte*)"glGetTexParameterIiv")) == NULL) || r;
+  r = ((glGetTexParameterIuiv = (PFNGLGETTEXPARAMETERIUIVPROC)glewGetProcAddress((const GLubyte*)"glGetTexParameterIuiv")) == NULL) || r;
+  r = ((glGetTransformFeedbackVarying = (PFNGLGETTRANSFORMFEEDBACKVARYINGPROC)glewGetProcAddress((const GLubyte*)"glGetTransformFeedbackVarying")) == NULL) || r;
+  r = ((glGetUniformuiv = (PFNGLGETUNIFORMUIVPROC)glewGetProcAddress((const GLubyte*)"glGetUniformuiv")) == NULL) || r;
+  r = ((glGetVertexAttribIiv = (PFNGLGETVERTEXATTRIBIIVPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribIiv")) == NULL) || r;
+  r = ((glGetVertexAttribIuiv = (PFNGLGETVERTEXATTRIBIUIVPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribIuiv")) == NULL) || r;
+  r = ((glIsEnabledi = (PFNGLISENABLEDIPROC)glewGetProcAddress((const GLubyte*)"glIsEnabledi")) == NULL) || r;
+  r = ((glTexParameterIiv = (PFNGLTEXPARAMETERIIVPROC)glewGetProcAddress((const GLubyte*)"glTexParameterIiv")) == NULL) || r;
+  r = ((glTexParameterIuiv = (PFNGLTEXPARAMETERIUIVPROC)glewGetProcAddress((const GLubyte*)"glTexParameterIuiv")) == NULL) || r;
+  r = ((glTransformFeedbackVaryings = (PFNGLTRANSFORMFEEDBACKVARYINGSPROC)glewGetProcAddress((const GLubyte*)"glTransformFeedbackVaryings")) == NULL) || r;
+  r = ((glUniform1ui = (PFNGLUNIFORM1UIPROC)glewGetProcAddress((const GLubyte*)"glUniform1ui")) == NULL) || r;
+  r = ((glUniform1uiv = (PFNGLUNIFORM1UIVPROC)glewGetProcAddress((const GLubyte*)"glUniform1uiv")) == NULL) || r;
+  r = ((glUniform2ui = (PFNGLUNIFORM2UIPROC)glewGetProcAddress((const GLubyte*)"glUniform2ui")) == NULL) || r;
+  r = ((glUniform2uiv = (PFNGLUNIFORM2UIVPROC)glewGetProcAddress((const GLubyte*)"glUniform2uiv")) == NULL) || r;
+  r = ((glUniform3ui = (PFNGLUNIFORM3UIPROC)glewGetProcAddress((const GLubyte*)"glUniform3ui")) == NULL) || r;
+  r = ((glUniform3uiv = (PFNGLUNIFORM3UIVPROC)glewGetProcAddress((const GLubyte*)"glUniform3uiv")) == NULL) || r;
+  r = ((glUniform4ui = (PFNGLUNIFORM4UIPROC)glewGetProcAddress((const GLubyte*)"glUniform4ui")) == NULL) || r;
+  r = ((glUniform4uiv = (PFNGLUNIFORM4UIVPROC)glewGetProcAddress((const GLubyte*)"glUniform4uiv")) == NULL) || r;
+  r = ((glVertexAttribI1i = (PFNGLVERTEXATTRIBI1IPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI1i")) == NULL) || r;
+  r = ((glVertexAttribI1iv = (PFNGLVERTEXATTRIBI1IVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI1iv")) == NULL) || r;
+  r = ((glVertexAttribI1ui = (PFNGLVERTEXATTRIBI1UIPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI1ui")) == NULL) || r;
+  r = ((glVertexAttribI1uiv = (PFNGLVERTEXATTRIBI1UIVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI1uiv")) == NULL) || r;
+  r = ((glVertexAttribI2i = (PFNGLVERTEXATTRIBI2IPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI2i")) == NULL) || r;
+  r = ((glVertexAttribI2iv = (PFNGLVERTEXATTRIBI2IVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI2iv")) == NULL) || r;
+  r = ((glVertexAttribI2ui = (PFNGLVERTEXATTRIBI2UIPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI2ui")) == NULL) || r;
+  r = ((glVertexAttribI2uiv = (PFNGLVERTEXATTRIBI2UIVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI2uiv")) == NULL) || r;
+  r = ((glVertexAttribI3i = (PFNGLVERTEXATTRIBI3IPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI3i")) == NULL) || r;
+  r = ((glVertexAttribI3iv = (PFNGLVERTEXATTRIBI3IVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI3iv")) == NULL) || r;
+  r = ((glVertexAttribI3ui = (PFNGLVERTEXATTRIBI3UIPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI3ui")) == NULL) || r;
+  r = ((glVertexAttribI3uiv = (PFNGLVERTEXATTRIBI3UIVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI3uiv")) == NULL) || r;
+  r = ((glVertexAttribI4bv = (PFNGLVERTEXATTRIBI4BVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4bv")) == NULL) || r;
+  r = ((glVertexAttribI4i = (PFNGLVERTEXATTRIBI4IPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4i")) == NULL) || r;
+  r = ((glVertexAttribI4iv = (PFNGLVERTEXATTRIBI4IVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4iv")) == NULL) || r;
+  r = ((glVertexAttribI4sv = (PFNGLVERTEXATTRIBI4SVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4sv")) == NULL) || r;
+  r = ((glVertexAttribI4ubv = (PFNGLVERTEXATTRIBI4UBVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4ubv")) == NULL) || r;
+  r = ((glVertexAttribI4ui = (PFNGLVERTEXATTRIBI4UIPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4ui")) == NULL) || r;
+  r = ((glVertexAttribI4uiv = (PFNGLVERTEXATTRIBI4UIVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4uiv")) == NULL) || r;
+  r = ((glVertexAttribI4usv = (PFNGLVERTEXATTRIBI4USVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4usv")) == NULL) || r;
+  r = ((glVertexAttribIPointer = (PFNGLVERTEXATTRIBIPOINTERPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribIPointer")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_VERSION_3_0 */
+
+#ifdef GL_3DFX_multisample
+
+#endif /* GL_3DFX_multisample */
+
+#ifdef GL_3DFX_tbuffer
+
+static GLboolean _glewInit_GL_3DFX_tbuffer (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glTbufferMask3DFX = (PFNGLTBUFFERMASK3DFXPROC)glewGetProcAddress((const GLubyte*)"glTbufferMask3DFX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_3DFX_tbuffer */
+
+#ifdef GL_3DFX_texture_compression_FXT1
+
+#endif /* GL_3DFX_texture_compression_FXT1 */
+
+#ifdef GL_APPLE_client_storage
+
+#endif /* GL_APPLE_client_storage */
+
+#ifdef GL_APPLE_element_array
+
+static GLboolean _glewInit_GL_APPLE_element_array (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDrawElementArrayAPPLE = (PFNGLDRAWELEMENTARRAYAPPLEPROC)glewGetProcAddress((const GLubyte*)"glDrawElementArrayAPPLE")) == NULL) || r;
+  r = ((glDrawRangeElementArrayAPPLE = (PFNGLDRAWRANGEELEMENTARRAYAPPLEPROC)glewGetProcAddress((const GLubyte*)"glDrawRangeElementArrayAPPLE")) == NULL) || r;
+  r = ((glElementPointerAPPLE = (PFNGLELEMENTPOINTERAPPLEPROC)glewGetProcAddress((const GLubyte*)"glElementPointerAPPLE")) == NULL) || r;
+  r = ((glMultiDrawElementArrayAPPLE = (PFNGLMULTIDRAWELEMENTARRAYAPPLEPROC)glewGetProcAddress((const GLubyte*)"glMultiDrawElementArrayAPPLE")) == NULL) || r;
+  r = ((glMultiDrawRangeElementArrayAPPLE = (PFNGLMULTIDRAWRANGEELEMENTARRAYAPPLEPROC)glewGetProcAddress((const GLubyte*)"glMultiDrawRangeElementArrayAPPLE")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_APPLE_element_array */
+
+#ifdef GL_APPLE_fence
+
+static GLboolean _glewInit_GL_APPLE_fence (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDeleteFencesAPPLE = (PFNGLDELETEFENCESAPPLEPROC)glewGetProcAddress((const GLubyte*)"glDeleteFencesAPPLE")) == NULL) || r;
+  r = ((glFinishFenceAPPLE = (PFNGLFINISHFENCEAPPLEPROC)glewGetProcAddress((const GLubyte*)"glFinishFenceAPPLE")) == NULL) || r;
+  r = ((glFinishObjectAPPLE = (PFNGLFINISHOBJECTAPPLEPROC)glewGetProcAddress((const GLubyte*)"glFinishObjectAPPLE")) == NULL) || r;
+  r = ((glGenFencesAPPLE = (PFNGLGENFENCESAPPLEPROC)glewGetProcAddress((const GLubyte*)"glGenFencesAPPLE")) == NULL) || r;
+  r = ((glIsFenceAPPLE = (PFNGLISFENCEAPPLEPROC)glewGetProcAddress((const GLubyte*)"glIsFenceAPPLE")) == NULL) || r;
+  r = ((glSetFenceAPPLE = (PFNGLSETFENCEAPPLEPROC)glewGetProcAddress((const GLubyte*)"glSetFenceAPPLE")) == NULL) || r;
+  r = ((glTestFenceAPPLE = (PFNGLTESTFENCEAPPLEPROC)glewGetProcAddress((const GLubyte*)"glTestFenceAPPLE")) == NULL) || r;
+  r = ((glTestObjectAPPLE = (PFNGLTESTOBJECTAPPLEPROC)glewGetProcAddress((const GLubyte*)"glTestObjectAPPLE")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_APPLE_fence */
+
+#ifdef GL_APPLE_float_pixels
+
+#endif /* GL_APPLE_float_pixels */
+
+#ifdef GL_APPLE_flush_buffer_range
+
+static GLboolean _glewInit_GL_APPLE_flush_buffer_range (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBufferParameteriAPPLE = (PFNGLBUFFERPARAMETERIAPPLEPROC)glewGetProcAddress((const GLubyte*)"glBufferParameteriAPPLE")) == NULL) || r;
+  r = ((glFlushMappedBufferRangeAPPLE = (PFNGLFLUSHMAPPEDBUFFERRANGEAPPLEPROC)glewGetProcAddress((const GLubyte*)"glFlushMappedBufferRangeAPPLE")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_APPLE_flush_buffer_range */
+
+#ifdef GL_APPLE_pixel_buffer
+
+#endif /* GL_APPLE_pixel_buffer */
+
+#ifdef GL_APPLE_specular_vector
+
+#endif /* GL_APPLE_specular_vector */
+
+#ifdef GL_APPLE_texture_range
+
+static GLboolean _glewInit_GL_APPLE_texture_range (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetTexParameterPointervAPPLE = (PFNGLGETTEXPARAMETERPOINTERVAPPLEPROC)glewGetProcAddress((const GLubyte*)"glGetTexParameterPointervAPPLE")) == NULL) || r;
+  r = ((glTextureRangeAPPLE = (PFNGLTEXTURERANGEAPPLEPROC)glewGetProcAddress((const GLubyte*)"glTextureRangeAPPLE")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_APPLE_texture_range */
+
+#ifdef GL_APPLE_transform_hint
+
+#endif /* GL_APPLE_transform_hint */
+
+#ifdef GL_APPLE_vertex_array_object
+
+static GLboolean _glewInit_GL_APPLE_vertex_array_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBindVertexArrayAPPLE = (PFNGLBINDVERTEXARRAYAPPLEPROC)glewGetProcAddress((const GLubyte*)"glBindVertexArrayAPPLE")) == NULL) || r;
+  r = ((glDeleteVertexArraysAPPLE = (PFNGLDELETEVERTEXARRAYSAPPLEPROC)glewGetProcAddress((const GLubyte*)"glDeleteVertexArraysAPPLE")) == NULL) || r;
+  r = ((glGenVertexArraysAPPLE = (PFNGLGENVERTEXARRAYSAPPLEPROC)glewGetProcAddress((const GLubyte*)"glGenVertexArraysAPPLE")) == NULL) || r;
+  r = ((glIsVertexArrayAPPLE = (PFNGLISVERTEXARRAYAPPLEPROC)glewGetProcAddress((const GLubyte*)"glIsVertexArrayAPPLE")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_APPLE_vertex_array_object */
+
+#ifdef GL_APPLE_vertex_array_range
+
+static GLboolean _glewInit_GL_APPLE_vertex_array_range (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFlushVertexArrayRangeAPPLE = (PFNGLFLUSHVERTEXARRAYRANGEAPPLEPROC)glewGetProcAddress((const GLubyte*)"glFlushVertexArrayRangeAPPLE")) == NULL) || r;
+  r = ((glVertexArrayParameteriAPPLE = (PFNGLVERTEXARRAYPARAMETERIAPPLEPROC)glewGetProcAddress((const GLubyte*)"glVertexArrayParameteriAPPLE")) == NULL) || r;
+  r = ((glVertexArrayRangeAPPLE = (PFNGLVERTEXARRAYRANGEAPPLEPROC)glewGetProcAddress((const GLubyte*)"glVertexArrayRangeAPPLE")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_APPLE_vertex_array_range */
+
+#ifdef GL_APPLE_ycbcr_422
+
+#endif /* GL_APPLE_ycbcr_422 */
+
+#ifdef GL_ARB_color_buffer_float
+
+static GLboolean _glewInit_GL_ARB_color_buffer_float (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glClampColorARB = (PFNGLCLAMPCOLORARBPROC)glewGetProcAddress((const GLubyte*)"glClampColorARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_color_buffer_float */
+
+#ifdef GL_ARB_depth_buffer_float
+
+#endif /* GL_ARB_depth_buffer_float */
+
+#ifdef GL_ARB_depth_texture
+
+#endif /* GL_ARB_depth_texture */
+
+#ifdef GL_ARB_draw_buffers
+
+static GLboolean _glewInit_GL_ARB_draw_buffers (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDrawBuffersARB = (PFNGLDRAWBUFFERSARBPROC)glewGetProcAddress((const GLubyte*)"glDrawBuffersARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_draw_buffers */
+
+#ifdef GL_ARB_draw_instanced
+
+static GLboolean _glewInit_GL_ARB_draw_instanced (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDrawArraysInstancedARB = (PFNGLDRAWARRAYSINSTANCEDARBPROC)glewGetProcAddress((const GLubyte*)"glDrawArraysInstancedARB")) == NULL) || r;
+  r = ((glDrawElementsInstancedARB = (PFNGLDRAWELEMENTSINSTANCEDARBPROC)glewGetProcAddress((const GLubyte*)"glDrawElementsInstancedARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_draw_instanced */
+
+#ifdef GL_ARB_fragment_program
+
+#endif /* GL_ARB_fragment_program */
+
+#ifdef GL_ARB_fragment_program_shadow
+
+#endif /* GL_ARB_fragment_program_shadow */
+
+#ifdef GL_ARB_fragment_shader
+
+#endif /* GL_ARB_fragment_shader */
+
+#ifdef GL_ARB_framebuffer_object
+
+static GLboolean _glewInit_GL_ARB_framebuffer_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBindFramebuffer = (PFNGLBINDFRAMEBUFFERPROC)glewGetProcAddress((const GLubyte*)"glBindFramebuffer")) == NULL) || r;
+  r = ((glBindRenderbuffer = (PFNGLBINDRENDERBUFFERPROC)glewGetProcAddress((const GLubyte*)"glBindRenderbuffer")) == NULL) || r;
+  r = ((glBlitFramebuffer = (PFNGLBLITFRAMEBUFFERPROC)glewGetProcAddress((const GLubyte*)"glBlitFramebuffer")) == NULL) || r;
+  r = ((glCheckFramebufferStatus = (PFNGLCHECKFRAMEBUFFERSTATUSPROC)glewGetProcAddress((const GLubyte*)"glCheckFramebufferStatus")) == NULL) || r;
+  r = ((glDeleteFramebuffers = (PFNGLDELETEFRAMEBUFFERSPROC)glewGetProcAddress((const GLubyte*)"glDeleteFramebuffers")) == NULL) || r;
+  r = ((glDeleteRenderbuffers = (PFNGLDELETERENDERBUFFERSPROC)glewGetProcAddress((const GLubyte*)"glDeleteRenderbuffers")) == NULL) || r;
+  r = ((glFramebufferRenderbuffer = (PFNGLFRAMEBUFFERRENDERBUFFERPROC)glewGetProcAddress((const GLubyte*)"glFramebufferRenderbuffer")) == NULL) || r;
+  r = ((glFramebufferTexturLayer = (PFNGLFRAMEBUFFERTEXTURLAYERPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTexturLayer")) == NULL) || r;
+  r = ((glFramebufferTexture1D = (PFNGLFRAMEBUFFERTEXTURE1DPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTexture1D")) == NULL) || r;
+  r = ((glFramebufferTexture2D = (PFNGLFRAMEBUFFERTEXTURE2DPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTexture2D")) == NULL) || r;
+  r = ((glFramebufferTexture3D = (PFNGLFRAMEBUFFERTEXTURE3DPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTexture3D")) == NULL) || r;
+  r = ((glGenFramebuffers = (PFNGLGENFRAMEBUFFERSPROC)glewGetProcAddress((const GLubyte*)"glGenFramebuffers")) == NULL) || r;
+  r = ((glGenRenderbuffers = (PFNGLGENRENDERBUFFERSPROC)glewGetProcAddress((const GLubyte*)"glGenRenderbuffers")) == NULL) || r;
+  r = ((glGenerateMipmap = (PFNGLGENERATEMIPMAPPROC)glewGetProcAddress((const GLubyte*)"glGenerateMipmap")) == NULL) || r;
+  r = ((glGetFramebufferAttachmentParameteriv = (PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC)glewGetProcAddress((const GLubyte*)"glGetFramebufferAttachmentParameteriv")) == NULL) || r;
+  r = ((glGetRenderbufferParameteriv = (PFNGLGETRENDERBUFFERPARAMETERIVPROC)glewGetProcAddress((const GLubyte*)"glGetRenderbufferParameteriv")) == NULL) || r;
+  r = ((glIsFramebuffer = (PFNGLISFRAMEBUFFERPROC)glewGetProcAddress((const GLubyte*)"glIsFramebuffer")) == NULL) || r;
+  r = ((glIsRenderbuffer = (PFNGLISRENDERBUFFERPROC)glewGetProcAddress((const GLubyte*)"glIsRenderbuffer")) == NULL) || r;
+  r = ((glRenderbufferStorage = (PFNGLRENDERBUFFERSTORAGEPROC)glewGetProcAddress((const GLubyte*)"glRenderbufferStorage")) == NULL) || r;
+  r = ((glRenderbufferStorageMultisample = (PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC)glewGetProcAddress((const GLubyte*)"glRenderbufferStorageMultisample")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_framebuffer_object */
+
+#ifdef GL_ARB_framebuffer_sRGB
+
+#endif /* GL_ARB_framebuffer_sRGB */
+
+#ifdef GL_ARB_geometry_shader4
+
+static GLboolean _glewInit_GL_ARB_geometry_shader4 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFramebufferTextureARB = (PFNGLFRAMEBUFFERTEXTUREARBPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTextureARB")) == NULL) || r;
+  r = ((glFramebufferTextureFaceARB = (PFNGLFRAMEBUFFERTEXTUREFACEARBPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTextureFaceARB")) == NULL) || r;
+  r = ((glFramebufferTextureLayerARB = (PFNGLFRAMEBUFFERTEXTURELAYERARBPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTextureLayerARB")) == NULL) || r;
+  r = ((glProgramParameteriARB = (PFNGLPROGRAMPARAMETERIARBPROC)glewGetProcAddress((const GLubyte*)"glProgramParameteriARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_geometry_shader4 */
+
+#ifdef GL_ARB_half_float_pixel
+
+#endif /* GL_ARB_half_float_pixel */
+
+#ifdef GL_ARB_half_float_vertex
+
+#endif /* GL_ARB_half_float_vertex */
+
+#ifdef GL_ARB_imaging
+
+static GLboolean _glewInit_GL_ARB_imaging (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBlendEquation = (PFNGLBLENDEQUATIONPROC)glewGetProcAddress((const GLubyte*)"glBlendEquation")) == NULL) || r;
+  r = ((glColorSubTable = (PFNGLCOLORSUBTABLEPROC)glewGetProcAddress((const GLubyte*)"glColorSubTable")) == NULL) || r;
+  r = ((glColorTable = (PFNGLCOLORTABLEPROC)glewGetProcAddress((const GLubyte*)"glColorTable")) == NULL) || r;
+  r = ((glColorTableParameterfv = (PFNGLCOLORTABLEPARAMETERFVPROC)glewGetProcAddress((const GLubyte*)"glColorTableParameterfv")) == NULL) || r;
+  r = ((glColorTableParameteriv = (PFNGLCOLORTABLEPARAMETERIVPROC)glewGetProcAddress((const GLubyte*)"glColorTableParameteriv")) == NULL) || r;
+  r = ((glConvolutionFilter1D = (PFNGLCONVOLUTIONFILTER1DPROC)glewGetProcAddress((const GLubyte*)"glConvolutionFilter1D")) == NULL) || r;
+  r = ((glConvolutionFilter2D = (PFNGLCONVOLUTIONFILTER2DPROC)glewGetProcAddress((const GLubyte*)"glConvolutionFilter2D")) == NULL) || r;
+  r = ((glConvolutionParameterf = (PFNGLCONVOLUTIONPARAMETERFPROC)glewGetProcAddress((const GLubyte*)"glConvolutionParameterf")) == NULL) || r;
+  r = ((glConvolutionParameterfv = (PFNGLCONVOLUTIONPARAMETERFVPROC)glewGetProcAddress((const GLubyte*)"glConvolutionParameterfv")) == NULL) || r;
+  r = ((glConvolutionParameteri = (PFNGLCONVOLUTIONPARAMETERIPROC)glewGetProcAddress((const GLubyte*)"glConvolutionParameteri")) == NULL) || r;
+  r = ((glConvolutionParameteriv = (PFNGLCONVOLUTIONPARAMETERIVPROC)glewGetProcAddress((const GLubyte*)"glConvolutionParameteriv")) == NULL) || r;
+  r = ((glCopyColorSubTable = (PFNGLCOPYCOLORSUBTABLEPROC)glewGetProcAddress((const GLubyte*)"glCopyColorSubTable")) == NULL) || r;
+  r = ((glCopyColorTable = (PFNGLCOPYCOLORTABLEPROC)glewGetProcAddress((const GLubyte*)"glCopyColorTable")) == NULL) || r;
+  r = ((glCopyConvolutionFilter1D = (PFNGLCOPYCONVOLUTIONFILTER1DPROC)glewGetProcAddress((const GLubyte*)"glCopyConvolutionFilter1D")) == NULL) || r;
+  r = ((glCopyConvolutionFilter2D = (PFNGLCOPYCONVOLUTIONFILTER2DPROC)glewGetProcAddress((const GLubyte*)"glCopyConvolutionFilter2D")) == NULL) || r;
+  r = ((glGetColorTable = (PFNGLGETCOLORTABLEPROC)glewGetProcAddress((const GLubyte*)"glGetColorTable")) == NULL) || r;
+  r = ((glGetColorTableParameterfv = (PFNGLGETCOLORTABLEPARAMETERFVPROC)glewGetProcAddress((const GLubyte*)"glGetColorTableParameterfv")) == NULL) || r;
+  r = ((glGetColorTableParameteriv = (PFNGLGETCOLORTABLEPARAMETERIVPROC)glewGetProcAddress((const GLubyte*)"glGetColorTableParameteriv")) == NULL) || r;
+  r = ((glGetConvolutionFilter = (PFNGLGETCONVOLUTIONFILTERPROC)glewGetProcAddress((const GLubyte*)"glGetConvolutionFilter")) == NULL) || r;
+  r = ((glGetConvolutionParameterfv = (PFNGLGETCONVOLUTIONPARAMETERFVPROC)glewGetProcAddress((const GLubyte*)"glGetConvolutionParameterfv")) == NULL) || r;
+  r = ((glGetConvolutionParameteriv = (PFNGLGETCONVOLUTIONPARAMETERIVPROC)glewGetProcAddress((const GLubyte*)"glGetConvolutionParameteriv")) == NULL) || r;
+  r = ((glGetHistogram = (PFNGLGETHISTOGRAMPROC)glewGetProcAddress((const GLubyte*)"glGetHistogram")) == NULL) || r;
+  r = ((glGetHistogramParameterfv = (PFNGLGETHISTOGRAMPARAMETERFVPROC)glewGetProcAddress((const GLubyte*)"glGetHistogramParameterfv")) == NULL) || r;
+  r = ((glGetHistogramParameteriv = (PFNGLGETHISTOGRAMPARAMETERIVPROC)glewGetProcAddress((const GLubyte*)"glGetHistogramParameteriv")) == NULL) || r;
+  r = ((glGetMinmax = (PFNGLGETMINMAXPROC)glewGetProcAddress((const GLubyte*)"glGetMinmax")) == NULL) || r;
+  r = ((glGetMinmaxParameterfv = (PFNGLGETMINMAXPARAMETERFVPROC)glewGetProcAddress((const GLubyte*)"glGetMinmaxParameterfv")) == NULL) || r;
+  r = ((glGetMinmaxParameteriv = (PFNGLGETMINMAXPARAMETERIVPROC)glewGetProcAddress((const GLubyte*)"glGetMinmaxParameteriv")) == NULL) || r;
+  r = ((glGetSeparableFilter = (PFNGLGETSEPARABLEFILTERPROC)glewGetProcAddress((const GLubyte*)"glGetSeparableFilter")) == NULL) || r;
+  r = ((glHistogram = (PFNGLHISTOGRAMPROC)glewGetProcAddress((const GLubyte*)"glHistogram")) == NULL) || r;
+  r = ((glMinmax = (PFNGLMINMAXPROC)glewGetProcAddress((const GLubyte*)"glMinmax")) == NULL) || r;
+  r = ((glResetHistogram = (PFNGLRESETHISTOGRAMPROC)glewGetProcAddress((const GLubyte*)"glResetHistogram")) == NULL) || r;
+  r = ((glResetMinmax = (PFNGLRESETMINMAXPROC)glewGetProcAddress((const GLubyte*)"glResetMinmax")) == NULL) || r;
+  r = ((glSeparableFilter2D = (PFNGLSEPARABLEFILTER2DPROC)glewGetProcAddress((const GLubyte*)"glSeparableFilter2D")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_imaging */
+
+#ifdef GL_ARB_instanced_arrays
+
+static GLboolean _glewInit_GL_ARB_instanced_arrays (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glVertexAttribDivisorARB = (PFNGLVERTEXATTRIBDIVISORARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribDivisorARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_instanced_arrays */
+
+#ifdef GL_ARB_map_buffer_range
+
+static GLboolean _glewInit_GL_ARB_map_buffer_range (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFlushMappedBufferRange = (PFNGLFLUSHMAPPEDBUFFERRANGEPROC)glewGetProcAddress((const GLubyte*)"glFlushMappedBufferRange")) == NULL) || r;
+  r = ((glMapBufferRange = (PFNGLMAPBUFFERRANGEPROC)glewGetProcAddress((const GLubyte*)"glMapBufferRange")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_map_buffer_range */
+
+#ifdef GL_ARB_matrix_palette
+
+static GLboolean _glewInit_GL_ARB_matrix_palette (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glCurrentPaletteMatrixARB = (PFNGLCURRENTPALETTEMATRIXARBPROC)glewGetProcAddress((const GLubyte*)"glCurrentPaletteMatrixARB")) == NULL) || r;
+  r = ((glMatrixIndexPointerARB = (PFNGLMATRIXINDEXPOINTERARBPROC)glewGetProcAddress((const GLubyte*)"glMatrixIndexPointerARB")) == NULL) || r;
+  r = ((glMatrixIndexubvARB = (PFNGLMATRIXINDEXUBVARBPROC)glewGetProcAddress((const GLubyte*)"glMatrixIndexubvARB")) == NULL) || r;
+  r = ((glMatrixIndexuivARB = (PFNGLMATRIXINDEXUIVARBPROC)glewGetProcAddress((const GLubyte*)"glMatrixIndexuivARB")) == NULL) || r;
+  r = ((glMatrixIndexusvARB = (PFNGLMATRIXINDEXUSVARBPROC)glewGetProcAddress((const GLubyte*)"glMatrixIndexusvARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_matrix_palette */
+
+#ifdef GL_ARB_multisample
+
+static GLboolean _glewInit_GL_ARB_multisample (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glSampleCoverageARB = (PFNGLSAMPLECOVERAGEARBPROC)glewGetProcAddress((const GLubyte*)"glSampleCoverageARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_multisample */
+
+#ifdef GL_ARB_multitexture
+
+static GLboolean _glewInit_GL_ARB_multitexture (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glActiveTextureARB = (PFNGLACTIVETEXTUREARBPROC)glewGetProcAddress((const GLubyte*)"glActiveTextureARB")) == NULL) || r;
+  r = ((glClientActiveTextureARB = (PFNGLCLIENTACTIVETEXTUREARBPROC)glewGetProcAddress((const GLubyte*)"glClientActiveTextureARB")) == NULL) || r;
+  r = ((glMultiTexCoord1dARB = (PFNGLMULTITEXCOORD1DARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1dARB")) == NULL) || r;
+  r = ((glMultiTexCoord1dvARB = (PFNGLMULTITEXCOORD1DVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1dvARB")) == NULL) || r;
+  r = ((glMultiTexCoord1fARB = (PFNGLMULTITEXCOORD1FARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1fARB")) == NULL) || r;
+  r = ((glMultiTexCoord1fvARB = (PFNGLMULTITEXCOORD1FVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1fvARB")) == NULL) || r;
+  r = ((glMultiTexCoord1iARB = (PFNGLMULTITEXCOORD1IARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1iARB")) == NULL) || r;
+  r = ((glMultiTexCoord1ivARB = (PFNGLMULTITEXCOORD1IVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1ivARB")) == NULL) || r;
+  r = ((glMultiTexCoord1sARB = (PFNGLMULTITEXCOORD1SARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1sARB")) == NULL) || r;
+  r = ((glMultiTexCoord1svARB = (PFNGLMULTITEXCOORD1SVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1svARB")) == NULL) || r;
+  r = ((glMultiTexCoord2dARB = (PFNGLMULTITEXCOORD2DARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2dARB")) == NULL) || r;
+  r = ((glMultiTexCoord2dvARB = (PFNGLMULTITEXCOORD2DVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2dvARB")) == NULL) || r;
+  r = ((glMultiTexCoord2fARB = (PFNGLMULTITEXCOORD2FARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2fARB")) == NULL) || r;
+  r = ((glMultiTexCoord2fvARB = (PFNGLMULTITEXCOORD2FVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2fvARB")) == NULL) || r;
+  r = ((glMultiTexCoord2iARB = (PFNGLMULTITEXCOORD2IARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2iARB")) == NULL) || r;
+  r = ((glMultiTexCoord2ivARB = (PFNGLMULTITEXCOORD2IVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2ivARB")) == NULL) || r;
+  r = ((glMultiTexCoord2sARB = (PFNGLMULTITEXCOORD2SARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2sARB")) == NULL) || r;
+  r = ((glMultiTexCoord2svARB = (PFNGLMULTITEXCOORD2SVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2svARB")) == NULL) || r;
+  r = ((glMultiTexCoord3dARB = (PFNGLMULTITEXCOORD3DARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3dARB")) == NULL) || r;
+  r = ((glMultiTexCoord3dvARB = (PFNGLMULTITEXCOORD3DVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3dvARB")) == NULL) || r;
+  r = ((glMultiTexCoord3fARB = (PFNGLMULTITEXCOORD3FARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3fARB")) == NULL) || r;
+  r = ((glMultiTexCoord3fvARB = (PFNGLMULTITEXCOORD3FVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3fvARB")) == NULL) || r;
+  r = ((glMultiTexCoord3iARB = (PFNGLMULTITEXCOORD3IARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3iARB")) == NULL) || r;
+  r = ((glMultiTexCoord3ivARB = (PFNGLMULTITEXCOORD3IVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3ivARB")) == NULL) || r;
+  r = ((glMultiTexCoord3sARB = (PFNGLMULTITEXCOORD3SARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3sARB")) == NULL) || r;
+  r = ((glMultiTexCoord3svARB = (PFNGLMULTITEXCOORD3SVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3svARB")) == NULL) || r;
+  r = ((glMultiTexCoord4dARB = (PFNGLMULTITEXCOORD4DARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4dARB")) == NULL) || r;
+  r = ((glMultiTexCoord4dvARB = (PFNGLMULTITEXCOORD4DVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4dvARB")) == NULL) || r;
+  r = ((glMultiTexCoord4fARB = (PFNGLMULTITEXCOORD4FARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4fARB")) == NULL) || r;
+  r = ((glMultiTexCoord4fvARB = (PFNGLMULTITEXCOORD4FVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4fvARB")) == NULL) || r;
+  r = ((glMultiTexCoord4iARB = (PFNGLMULTITEXCOORD4IARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4iARB")) == NULL) || r;
+  r = ((glMultiTexCoord4ivARB = (PFNGLMULTITEXCOORD4IVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4ivARB")) == NULL) || r;
+  r = ((glMultiTexCoord4sARB = (PFNGLMULTITEXCOORD4SARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4sARB")) == NULL) || r;
+  r = ((glMultiTexCoord4svARB = (PFNGLMULTITEXCOORD4SVARBPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4svARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_multitexture */
+
+#ifdef GL_ARB_occlusion_query
+
+static GLboolean _glewInit_GL_ARB_occlusion_query (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBeginQueryARB = (PFNGLBEGINQUERYARBPROC)glewGetProcAddress((const GLubyte*)"glBeginQueryARB")) == NULL) || r;
+  r = ((glDeleteQueriesARB = (PFNGLDELETEQUERIESARBPROC)glewGetProcAddress((const GLubyte*)"glDeleteQueriesARB")) == NULL) || r;
+  r = ((glEndQueryARB = (PFNGLENDQUERYARBPROC)glewGetProcAddress((const GLubyte*)"glEndQueryARB")) == NULL) || r;
+  r = ((glGenQueriesARB = (PFNGLGENQUERIESARBPROC)glewGetProcAddress((const GLubyte*)"glGenQueriesARB")) == NULL) || r;
+  r = ((glGetQueryObjectivARB = (PFNGLGETQUERYOBJECTIVARBPROC)glewGetProcAddress((const GLubyte*)"glGetQueryObjectivARB")) == NULL) || r;
+  r = ((glGetQueryObjectuivARB = (PFNGLGETQUERYOBJECTUIVARBPROC)glewGetProcAddress((const GLubyte*)"glGetQueryObjectuivARB")) == NULL) || r;
+  r = ((glGetQueryivARB = (PFNGLGETQUERYIVARBPROC)glewGetProcAddress((const GLubyte*)"glGetQueryivARB")) == NULL) || r;
+  r = ((glIsQueryARB = (PFNGLISQUERYARBPROC)glewGetProcAddress((const GLubyte*)"glIsQueryARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_occlusion_query */
+
+#ifdef GL_ARB_pixel_buffer_object
+
+#endif /* GL_ARB_pixel_buffer_object */
+
+#ifdef GL_ARB_point_parameters
+
+static GLboolean _glewInit_GL_ARB_point_parameters (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glPointParameterfARB = (PFNGLPOINTPARAMETERFARBPROC)glewGetProcAddress((const GLubyte*)"glPointParameterfARB")) == NULL) || r;
+  r = ((glPointParameterfvARB = (PFNGLPOINTPARAMETERFVARBPROC)glewGetProcAddress((const GLubyte*)"glPointParameterfvARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_point_parameters */
+
+#ifdef GL_ARB_point_sprite
+
+#endif /* GL_ARB_point_sprite */
+
+#ifdef GL_ARB_shader_objects
+
+static GLboolean _glewInit_GL_ARB_shader_objects (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glAttachObjectARB = (PFNGLATTACHOBJECTARBPROC)glewGetProcAddress((const GLubyte*)"glAttachObjectARB")) == NULL) || r;
+  r = ((glCompileShaderARB = (PFNGLCOMPILESHADERARBPROC)glewGetProcAddress((const GLubyte*)"glCompileShaderARB")) == NULL) || r;
+  r = ((glCreateProgramObjectARB = (PFNGLCREATEPROGRAMOBJECTARBPROC)glewGetProcAddress((const GLubyte*)"glCreateProgramObjectARB")) == NULL) || r;
+  r = ((glCreateShaderObjectARB = (PFNGLCREATESHADEROBJECTARBPROC)glewGetProcAddress((const GLubyte*)"glCreateShaderObjectARB")) == NULL) || r;
+  r = ((glDeleteObjectARB = (PFNGLDELETEOBJECTARBPROC)glewGetProcAddress((const GLubyte*)"glDeleteObjectARB")) == NULL) || r;
+  r = ((glDetachObjectARB = (PFNGLDETACHOBJECTARBPROC)glewGetProcAddress((const GLubyte*)"glDetachObjectARB")) == NULL) || r;
+  r = ((glGetActiveUniformARB = (PFNGLGETACTIVEUNIFORMARBPROC)glewGetProcAddress((const GLubyte*)"glGetActiveUniformARB")) == NULL) || r;
+  r = ((glGetAttachedObjectsARB = (PFNGLGETATTACHEDOBJECTSARBPROC)glewGetProcAddress((const GLubyte*)"glGetAttachedObjectsARB")) == NULL) || r;
+  r = ((glGetHandleARB = (PFNGLGETHANDLEARBPROC)glewGetProcAddress((const GLubyte*)"glGetHandleARB")) == NULL) || r;
+  r = ((glGetInfoLogARB = (PFNGLGETINFOLOGARBPROC)glewGetProcAddress((const GLubyte*)"glGetInfoLogARB")) == NULL) || r;
+  r = ((glGetObjectParameterfvARB = (PFNGLGETOBJECTPARAMETERFVARBPROC)glewGetProcAddress((const GLubyte*)"glGetObjectParameterfvARB")) == NULL) || r;
+  r = ((glGetObjectParameterivARB = (PFNGLGETOBJECTPARAMETERIVARBPROC)glewGetProcAddress((const GLubyte*)"glGetObjectParameterivARB")) == NULL) || r;
+  r = ((glGetShaderSourceARB = (PFNGLGETSHADERSOURCEARBPROC)glewGetProcAddress((const GLubyte*)"glGetShaderSourceARB")) == NULL) || r;
+  r = ((glGetUniformLocationARB = (PFNGLGETUNIFORMLOCATIONARBPROC)glewGetProcAddress((const GLubyte*)"glGetUniformLocationARB")) == NULL) || r;
+  r = ((glGetUniformfvARB = (PFNGLGETUNIFORMFVARBPROC)glewGetProcAddress((const GLubyte*)"glGetUniformfvARB")) == NULL) || r;
+  r = ((glGetUniformivARB = (PFNGLGETUNIFORMIVARBPROC)glewGetProcAddress((const GLubyte*)"glGetUniformivARB")) == NULL) || r;
+  r = ((glLinkProgramARB = (PFNGLLINKPROGRAMARBPROC)glewGetProcAddress((const GLubyte*)"glLinkProgramARB")) == NULL) || r;
+  r = ((glShaderSourceARB = (PFNGLSHADERSOURCEARBPROC)glewGetProcAddress((const GLubyte*)"glShaderSourceARB")) == NULL) || r;
+  r = ((glUniform1fARB = (PFNGLUNIFORM1FARBPROC)glewGetProcAddress((const GLubyte*)"glUniform1fARB")) == NULL) || r;
+  r = ((glUniform1fvARB = (PFNGLUNIFORM1FVARBPROC)glewGetProcAddress((const GLubyte*)"glUniform1fvARB")) == NULL) || r;
+  r = ((glUniform1iARB = (PFNGLUNIFORM1IARBPROC)glewGetProcAddress((const GLubyte*)"glUniform1iARB")) == NULL) || r;
+  r = ((glUniform1ivARB = (PFNGLUNIFORM1IVARBPROC)glewGetProcAddress((const GLubyte*)"glUniform1ivARB")) == NULL) || r;
+  r = ((glUniform2fARB = (PFNGLUNIFORM2FARBPROC)glewGetProcAddress((const GLubyte*)"glUniform2fARB")) == NULL) || r;
+  r = ((glUniform2fvARB = (PFNGLUNIFORM2FVARBPROC)glewGetProcAddress((const GLubyte*)"glUniform2fvARB")) == NULL) || r;
+  r = ((glUniform2iARB = (PFNGLUNIFORM2IARBPROC)glewGetProcAddress((const GLubyte*)"glUniform2iARB")) == NULL) || r;
+  r = ((glUniform2ivARB = (PFNGLUNIFORM2IVARBPROC)glewGetProcAddress((const GLubyte*)"glUniform2ivARB")) == NULL) || r;
+  r = ((glUniform3fARB = (PFNGLUNIFORM3FARBPROC)glewGetProcAddress((const GLubyte*)"glUniform3fARB")) == NULL) || r;
+  r = ((glUniform3fvARB = (PFNGLUNIFORM3FVARBPROC)glewGetProcAddress((const GLubyte*)"glUniform3fvARB")) == NULL) || r;
+  r = ((glUniform3iARB = (PFNGLUNIFORM3IARBPROC)glewGetProcAddress((const GLubyte*)"glUniform3iARB")) == NULL) || r;
+  r = ((glUniform3ivARB = (PFNGLUNIFORM3IVARBPROC)glewGetProcAddress((const GLubyte*)"glUniform3ivARB")) == NULL) || r;
+  r = ((glUniform4fARB = (PFNGLUNIFORM4FARBPROC)glewGetProcAddress((const GLubyte*)"glUniform4fARB")) == NULL) || r;
+  r = ((glUniform4fvARB = (PFNGLUNIFORM4FVARBPROC)glewGetProcAddress((const GLubyte*)"glUniform4fvARB")) == NULL) || r;
+  r = ((glUniform4iARB = (PFNGLUNIFORM4IARBPROC)glewGetProcAddress((const GLubyte*)"glUniform4iARB")) == NULL) || r;
+  r = ((glUniform4ivARB = (PFNGLUNIFORM4IVARBPROC)glewGetProcAddress((const GLubyte*)"glUniform4ivARB")) == NULL) || r;
+  r = ((glUniformMatrix2fvARB = (PFNGLUNIFORMMATRIX2FVARBPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix2fvARB")) == NULL) || r;
+  r = ((glUniformMatrix3fvARB = (PFNGLUNIFORMMATRIX3FVARBPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix3fvARB")) == NULL) || r;
+  r = ((glUniformMatrix4fvARB = (PFNGLUNIFORMMATRIX4FVARBPROC)glewGetProcAddress((const GLubyte*)"glUniformMatrix4fvARB")) == NULL) || r;
+  r = ((glUseProgramObjectARB = (PFNGLUSEPROGRAMOBJECTARBPROC)glewGetProcAddress((const GLubyte*)"glUseProgramObjectARB")) == NULL) || r;
+  r = ((glValidateProgramARB = (PFNGLVALIDATEPROGRAMARBPROC)glewGetProcAddress((const GLubyte*)"glValidateProgramARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_shader_objects */
+
+#ifdef GL_ARB_shading_language_100
+
+#endif /* GL_ARB_shading_language_100 */
+
+#ifdef GL_ARB_shadow
+
+#endif /* GL_ARB_shadow */
+
+#ifdef GL_ARB_shadow_ambient
+
+#endif /* GL_ARB_shadow_ambient */
+
+#ifdef GL_ARB_texture_border_clamp
+
+#endif /* GL_ARB_texture_border_clamp */
+
+#ifdef GL_ARB_texture_buffer_object
+
+static GLboolean _glewInit_GL_ARB_texture_buffer_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glTexBufferARB = (PFNGLTEXBUFFERARBPROC)glewGetProcAddress((const GLubyte*)"glTexBufferARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_texture_buffer_object */
+
+#ifdef GL_ARB_texture_compression
+
+static GLboolean _glewInit_GL_ARB_texture_compression (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glCompressedTexImage1DARB = (PFNGLCOMPRESSEDTEXIMAGE1DARBPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexImage1DARB")) == NULL) || r;
+  r = ((glCompressedTexImage2DARB = (PFNGLCOMPRESSEDTEXIMAGE2DARBPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexImage2DARB")) == NULL) || r;
+  r = ((glCompressedTexImage3DARB = (PFNGLCOMPRESSEDTEXIMAGE3DARBPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexImage3DARB")) == NULL) || r;
+  r = ((glCompressedTexSubImage1DARB = (PFNGLCOMPRESSEDTEXSUBIMAGE1DARBPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexSubImage1DARB")) == NULL) || r;
+  r = ((glCompressedTexSubImage2DARB = (PFNGLCOMPRESSEDTEXSUBIMAGE2DARBPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexSubImage2DARB")) == NULL) || r;
+  r = ((glCompressedTexSubImage3DARB = (PFNGLCOMPRESSEDTEXSUBIMAGE3DARBPROC)glewGetProcAddress((const GLubyte*)"glCompressedTexSubImage3DARB")) == NULL) || r;
+  r = ((glGetCompressedTexImageARB = (PFNGLGETCOMPRESSEDTEXIMAGEARBPROC)glewGetProcAddress((const GLubyte*)"glGetCompressedTexImageARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_texture_compression */
+
+#ifdef GL_ARB_texture_compression_rgtc
+
+#endif /* GL_ARB_texture_compression_rgtc */
+
+#ifdef GL_ARB_texture_cube_map
+
+#endif /* GL_ARB_texture_cube_map */
+
+#ifdef GL_ARB_texture_env_add
+
+#endif /* GL_ARB_texture_env_add */
+
+#ifdef GL_ARB_texture_env_combine
+
+#endif /* GL_ARB_texture_env_combine */
+
+#ifdef GL_ARB_texture_env_crossbar
+
+#endif /* GL_ARB_texture_env_crossbar */
+
+#ifdef GL_ARB_texture_env_dot3
+
+#endif /* GL_ARB_texture_env_dot3 */
+
+#ifdef GL_ARB_texture_float
+
+#endif /* GL_ARB_texture_float */
+
+#ifdef GL_ARB_texture_mirrored_repeat
+
+#endif /* GL_ARB_texture_mirrored_repeat */
+
+#ifdef GL_ARB_texture_non_power_of_two
+
+#endif /* GL_ARB_texture_non_power_of_two */
+
+#ifdef GL_ARB_texture_rectangle
+
+#endif /* GL_ARB_texture_rectangle */
+
+#ifdef GL_ARB_texture_rg
+
+#endif /* GL_ARB_texture_rg */
+
+#ifdef GL_ARB_transpose_matrix
+
+static GLboolean _glewInit_GL_ARB_transpose_matrix (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glLoadTransposeMatrixdARB = (PFNGLLOADTRANSPOSEMATRIXDARBPROC)glewGetProcAddress((const GLubyte*)"glLoadTransposeMatrixdARB")) == NULL) || r;
+  r = ((glLoadTransposeMatrixfARB = (PFNGLLOADTRANSPOSEMATRIXFARBPROC)glewGetProcAddress((const GLubyte*)"glLoadTransposeMatrixfARB")) == NULL) || r;
+  r = ((glMultTransposeMatrixdARB = (PFNGLMULTTRANSPOSEMATRIXDARBPROC)glewGetProcAddress((const GLubyte*)"glMultTransposeMatrixdARB")) == NULL) || r;
+  r = ((glMultTransposeMatrixfARB = (PFNGLMULTTRANSPOSEMATRIXFARBPROC)glewGetProcAddress((const GLubyte*)"glMultTransposeMatrixfARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_transpose_matrix */
+
+#ifdef GL_ARB_vertex_array_object
+
+static GLboolean _glewInit_GL_ARB_vertex_array_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBindVertexArray = (PFNGLBINDVERTEXARRAYPROC)glewGetProcAddress((const GLubyte*)"glBindVertexArray")) == NULL) || r;
+  r = ((glDeleteVertexArrays = (PFNGLDELETEVERTEXARRAYSPROC)glewGetProcAddress((const GLubyte*)"glDeleteVertexArrays")) == NULL) || r;
+  r = ((glGenVertexArrays = (PFNGLGENVERTEXARRAYSPROC)glewGetProcAddress((const GLubyte*)"glGenVertexArrays")) == NULL) || r;
+  r = ((glIsVertexArray = (PFNGLISVERTEXARRAYPROC)glewGetProcAddress((const GLubyte*)"glIsVertexArray")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_vertex_array_object */
+
+#ifdef GL_ARB_vertex_blend
+
+static GLboolean _glewInit_GL_ARB_vertex_blend (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glVertexBlendARB = (PFNGLVERTEXBLENDARBPROC)glewGetProcAddress((const GLubyte*)"glVertexBlendARB")) == NULL) || r;
+  r = ((glWeightPointerARB = (PFNGLWEIGHTPOINTERARBPROC)glewGetProcAddress((const GLubyte*)"glWeightPointerARB")) == NULL) || r;
+  r = ((glWeightbvARB = (PFNGLWEIGHTBVARBPROC)glewGetProcAddress((const GLubyte*)"glWeightbvARB")) == NULL) || r;
+  r = ((glWeightdvARB = (PFNGLWEIGHTDVARBPROC)glewGetProcAddress((const GLubyte*)"glWeightdvARB")) == NULL) || r;
+  r = ((glWeightfvARB = (PFNGLWEIGHTFVARBPROC)glewGetProcAddress((const GLubyte*)"glWeightfvARB")) == NULL) || r;
+  r = ((glWeightivARB = (PFNGLWEIGHTIVARBPROC)glewGetProcAddress((const GLubyte*)"glWeightivARB")) == NULL) || r;
+  r = ((glWeightsvARB = (PFNGLWEIGHTSVARBPROC)glewGetProcAddress((const GLubyte*)"glWeightsvARB")) == NULL) || r;
+  r = ((glWeightubvARB = (PFNGLWEIGHTUBVARBPROC)glewGetProcAddress((const GLubyte*)"glWeightubvARB")) == NULL) || r;
+  r = ((glWeightuivARB = (PFNGLWEIGHTUIVARBPROC)glewGetProcAddress((const GLubyte*)"glWeightuivARB")) == NULL) || r;
+  r = ((glWeightusvARB = (PFNGLWEIGHTUSVARBPROC)glewGetProcAddress((const GLubyte*)"glWeightusvARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_vertex_blend */
+
+#ifdef GL_ARB_vertex_buffer_object
+
+static GLboolean _glewInit_GL_ARB_vertex_buffer_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBindBufferARB = (PFNGLBINDBUFFERARBPROC)glewGetProcAddress((const GLubyte*)"glBindBufferARB")) == NULL) || r;
+  r = ((glBufferDataARB = (PFNGLBUFFERDATAARBPROC)glewGetProcAddress((const GLubyte*)"glBufferDataARB")) == NULL) || r;
+  r = ((glBufferSubDataARB = (PFNGLBUFFERSUBDATAARBPROC)glewGetProcAddress((const GLubyte*)"glBufferSubDataARB")) == NULL) || r;
+  r = ((glDeleteBuffersARB = (PFNGLDELETEBUFFERSARBPROC)glewGetProcAddress((const GLubyte*)"glDeleteBuffersARB")) == NULL) || r;
+  r = ((glGenBuffersARB = (PFNGLGENBUFFERSARBPROC)glewGetProcAddress((const GLubyte*)"glGenBuffersARB")) == NULL) || r;
+  r = ((glGetBufferParameterivARB = (PFNGLGETBUFFERPARAMETERIVARBPROC)glewGetProcAddress((const GLubyte*)"glGetBufferParameterivARB")) == NULL) || r;
+  r = ((glGetBufferPointervARB = (PFNGLGETBUFFERPOINTERVARBPROC)glewGetProcAddress((const GLubyte*)"glGetBufferPointervARB")) == NULL) || r;
+  r = ((glGetBufferSubDataARB = (PFNGLGETBUFFERSUBDATAARBPROC)glewGetProcAddress((const GLubyte*)"glGetBufferSubDataARB")) == NULL) || r;
+  r = ((glIsBufferARB = (PFNGLISBUFFERARBPROC)glewGetProcAddress((const GLubyte*)"glIsBufferARB")) == NULL) || r;
+  r = ((glMapBufferARB = (PFNGLMAPBUFFERARBPROC)glewGetProcAddress((const GLubyte*)"glMapBufferARB")) == NULL) || r;
+  r = ((glUnmapBufferARB = (PFNGLUNMAPBUFFERARBPROC)glewGetProcAddress((const GLubyte*)"glUnmapBufferARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_vertex_buffer_object */
+
+#ifdef GL_ARB_vertex_program
+
+static GLboolean _glewInit_GL_ARB_vertex_program (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBindProgramARB = (PFNGLBINDPROGRAMARBPROC)glewGetProcAddress((const GLubyte*)"glBindProgramARB")) == NULL) || r;
+  r = ((glDeleteProgramsARB = (PFNGLDELETEPROGRAMSARBPROC)glewGetProcAddress((const GLubyte*)"glDeleteProgramsARB")) == NULL) || r;
+  r = ((glDisableVertexAttribArrayARB = (PFNGLDISABLEVERTEXATTRIBARRAYARBPROC)glewGetProcAddress((const GLubyte*)"glDisableVertexAttribArrayARB")) == NULL) || r;
+  r = ((glEnableVertexAttribArrayARB = (PFNGLENABLEVERTEXATTRIBARRAYARBPROC)glewGetProcAddress((const GLubyte*)"glEnableVertexAttribArrayARB")) == NULL) || r;
+  r = ((glGenProgramsARB = (PFNGLGENPROGRAMSARBPROC)glewGetProcAddress((const GLubyte*)"glGenProgramsARB")) == NULL) || r;
+  r = ((glGetProgramEnvParameterdvARB = (PFNGLGETPROGRAMENVPARAMETERDVARBPROC)glewGetProcAddress((const GLubyte*)"glGetProgramEnvParameterdvARB")) == NULL) || r;
+  r = ((glGetProgramEnvParameterfvARB = (PFNGLGETPROGRAMENVPARAMETERFVARBPROC)glewGetProcAddress((const GLubyte*)"glGetProgramEnvParameterfvARB")) == NULL) || r;
+  r = ((glGetProgramLocalParameterdvARB = (PFNGLGETPROGRAMLOCALPARAMETERDVARBPROC)glewGetProcAddress((const GLubyte*)"glGetProgramLocalParameterdvARB")) == NULL) || r;
+  r = ((glGetProgramLocalParameterfvARB = (PFNGLGETPROGRAMLOCALPARAMETERFVARBPROC)glewGetProcAddress((const GLubyte*)"glGetProgramLocalParameterfvARB")) == NULL) || r;
+  r = ((glGetProgramStringARB = (PFNGLGETPROGRAMSTRINGARBPROC)glewGetProcAddress((const GLubyte*)"glGetProgramStringARB")) == NULL) || r;
+  r = ((glGetProgramivARB = (PFNGLGETPROGRAMIVARBPROC)glewGetProcAddress((const GLubyte*)"glGetProgramivARB")) == NULL) || r;
+  r = ((glGetVertexAttribPointervARB = (PFNGLGETVERTEXATTRIBPOINTERVARBPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribPointervARB")) == NULL) || r;
+  r = ((glGetVertexAttribdvARB = (PFNGLGETVERTEXATTRIBDVARBPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribdvARB")) == NULL) || r;
+  r = ((glGetVertexAttribfvARB = (PFNGLGETVERTEXATTRIBFVARBPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribfvARB")) == NULL) || r;
+  r = ((glGetVertexAttribivARB = (PFNGLGETVERTEXATTRIBIVARBPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribivARB")) == NULL) || r;
+  r = ((glIsProgramARB = (PFNGLISPROGRAMARBPROC)glewGetProcAddress((const GLubyte*)"glIsProgramARB")) == NULL) || r;
+  r = ((glProgramEnvParameter4dARB = (PFNGLPROGRAMENVPARAMETER4DARBPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParameter4dARB")) == NULL) || r;
+  r = ((glProgramEnvParameter4dvARB = (PFNGLPROGRAMENVPARAMETER4DVARBPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParameter4dvARB")) == NULL) || r;
+  r = ((glProgramEnvParameter4fARB = (PFNGLPROGRAMENVPARAMETER4FARBPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParameter4fARB")) == NULL) || r;
+  r = ((glProgramEnvParameter4fvARB = (PFNGLPROGRAMENVPARAMETER4FVARBPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParameter4fvARB")) == NULL) || r;
+  r = ((glProgramLocalParameter4dARB = (PFNGLPROGRAMLOCALPARAMETER4DARBPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParameter4dARB")) == NULL) || r;
+  r = ((glProgramLocalParameter4dvARB = (PFNGLPROGRAMLOCALPARAMETER4DVARBPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParameter4dvARB")) == NULL) || r;
+  r = ((glProgramLocalParameter4fARB = (PFNGLPROGRAMLOCALPARAMETER4FARBPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParameter4fARB")) == NULL) || r;
+  r = ((glProgramLocalParameter4fvARB = (PFNGLPROGRAMLOCALPARAMETER4FVARBPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParameter4fvARB")) == NULL) || r;
+  r = ((glProgramStringARB = (PFNGLPROGRAMSTRINGARBPROC)glewGetProcAddress((const GLubyte*)"glProgramStringARB")) == NULL) || r;
+  r = ((glVertexAttrib1dARB = (PFNGLVERTEXATTRIB1DARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1dARB")) == NULL) || r;
+  r = ((glVertexAttrib1dvARB = (PFNGLVERTEXATTRIB1DVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1dvARB")) == NULL) || r;
+  r = ((glVertexAttrib1fARB = (PFNGLVERTEXATTRIB1FARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1fARB")) == NULL) || r;
+  r = ((glVertexAttrib1fvARB = (PFNGLVERTEXATTRIB1FVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1fvARB")) == NULL) || r;
+  r = ((glVertexAttrib1sARB = (PFNGLVERTEXATTRIB1SARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1sARB")) == NULL) || r;
+  r = ((glVertexAttrib1svARB = (PFNGLVERTEXATTRIB1SVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1svARB")) == NULL) || r;
+  r = ((glVertexAttrib2dARB = (PFNGLVERTEXATTRIB2DARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2dARB")) == NULL) || r;
+  r = ((glVertexAttrib2dvARB = (PFNGLVERTEXATTRIB2DVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2dvARB")) == NULL) || r;
+  r = ((glVertexAttrib2fARB = (PFNGLVERTEXATTRIB2FARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2fARB")) == NULL) || r;
+  r = ((glVertexAttrib2fvARB = (PFNGLVERTEXATTRIB2FVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2fvARB")) == NULL) || r;
+  r = ((glVertexAttrib2sARB = (PFNGLVERTEXATTRIB2SARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2sARB")) == NULL) || r;
+  r = ((glVertexAttrib2svARB = (PFNGLVERTEXATTRIB2SVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2svARB")) == NULL) || r;
+  r = ((glVertexAttrib3dARB = (PFNGLVERTEXATTRIB3DARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3dARB")) == NULL) || r;
+  r = ((glVertexAttrib3dvARB = (PFNGLVERTEXATTRIB3DVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3dvARB")) == NULL) || r;
+  r = ((glVertexAttrib3fARB = (PFNGLVERTEXATTRIB3FARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3fARB")) == NULL) || r;
+  r = ((glVertexAttrib3fvARB = (PFNGLVERTEXATTRIB3FVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3fvARB")) == NULL) || r;
+  r = ((glVertexAttrib3sARB = (PFNGLVERTEXATTRIB3SARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3sARB")) == NULL) || r;
+  r = ((glVertexAttrib3svARB = (PFNGLVERTEXATTRIB3SVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3svARB")) == NULL) || r;
+  r = ((glVertexAttrib4NbvARB = (PFNGLVERTEXATTRIB4NBVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4NbvARB")) == NULL) || r;
+  r = ((glVertexAttrib4NivARB = (PFNGLVERTEXATTRIB4NIVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4NivARB")) == NULL) || r;
+  r = ((glVertexAttrib4NsvARB = (PFNGLVERTEXATTRIB4NSVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4NsvARB")) == NULL) || r;
+  r = ((glVertexAttrib4NubARB = (PFNGLVERTEXATTRIB4NUBARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4NubARB")) == NULL) || r;
+  r = ((glVertexAttrib4NubvARB = (PFNGLVERTEXATTRIB4NUBVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4NubvARB")) == NULL) || r;
+  r = ((glVertexAttrib4NuivARB = (PFNGLVERTEXATTRIB4NUIVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4NuivARB")) == NULL) || r;
+  r = ((glVertexAttrib4NusvARB = (PFNGLVERTEXATTRIB4NUSVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4NusvARB")) == NULL) || r;
+  r = ((glVertexAttrib4bvARB = (PFNGLVERTEXATTRIB4BVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4bvARB")) == NULL) || r;
+  r = ((glVertexAttrib4dARB = (PFNGLVERTEXATTRIB4DARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4dARB")) == NULL) || r;
+  r = ((glVertexAttrib4dvARB = (PFNGLVERTEXATTRIB4DVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4dvARB")) == NULL) || r;
+  r = ((glVertexAttrib4fARB = (PFNGLVERTEXATTRIB4FARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4fARB")) == NULL) || r;
+  r = ((glVertexAttrib4fvARB = (PFNGLVERTEXATTRIB4FVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4fvARB")) == NULL) || r;
+  r = ((glVertexAttrib4ivARB = (PFNGLVERTEXATTRIB4IVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4ivARB")) == NULL) || r;
+  r = ((glVertexAttrib4sARB = (PFNGLVERTEXATTRIB4SARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4sARB")) == NULL) || r;
+  r = ((glVertexAttrib4svARB = (PFNGLVERTEXATTRIB4SVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4svARB")) == NULL) || r;
+  r = ((glVertexAttrib4ubvARB = (PFNGLVERTEXATTRIB4UBVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4ubvARB")) == NULL) || r;
+  r = ((glVertexAttrib4uivARB = (PFNGLVERTEXATTRIB4UIVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4uivARB")) == NULL) || r;
+  r = ((glVertexAttrib4usvARB = (PFNGLVERTEXATTRIB4USVARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4usvARB")) == NULL) || r;
+  r = ((glVertexAttribPointerARB = (PFNGLVERTEXATTRIBPOINTERARBPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribPointerARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_vertex_program */
+
+#ifdef GL_ARB_vertex_shader
+
+static GLboolean _glewInit_GL_ARB_vertex_shader (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBindAttribLocationARB = (PFNGLBINDATTRIBLOCATIONARBPROC)glewGetProcAddress((const GLubyte*)"glBindAttribLocationARB")) == NULL) || r;
+  r = ((glGetActiveAttribARB = (PFNGLGETACTIVEATTRIBARBPROC)glewGetProcAddress((const GLubyte*)"glGetActiveAttribARB")) == NULL) || r;
+  r = ((glGetAttribLocationARB = (PFNGLGETATTRIBLOCATIONARBPROC)glewGetProcAddress((const GLubyte*)"glGetAttribLocationARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_vertex_shader */
+
+#ifdef GL_ARB_window_pos
+
+static GLboolean _glewInit_GL_ARB_window_pos (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glWindowPos2dARB = (PFNGLWINDOWPOS2DARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2dARB")) == NULL) || r;
+  r = ((glWindowPos2dvARB = (PFNGLWINDOWPOS2DVARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2dvARB")) == NULL) || r;
+  r = ((glWindowPos2fARB = (PFNGLWINDOWPOS2FARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2fARB")) == NULL) || r;
+  r = ((glWindowPos2fvARB = (PFNGLWINDOWPOS2FVARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2fvARB")) == NULL) || r;
+  r = ((glWindowPos2iARB = (PFNGLWINDOWPOS2IARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2iARB")) == NULL) || r;
+  r = ((glWindowPos2ivARB = (PFNGLWINDOWPOS2IVARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2ivARB")) == NULL) || r;
+  r = ((glWindowPos2sARB = (PFNGLWINDOWPOS2SARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2sARB")) == NULL) || r;
+  r = ((glWindowPos2svARB = (PFNGLWINDOWPOS2SVARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2svARB")) == NULL) || r;
+  r = ((glWindowPos3dARB = (PFNGLWINDOWPOS3DARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3dARB")) == NULL) || r;
+  r = ((glWindowPos3dvARB = (PFNGLWINDOWPOS3DVARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3dvARB")) == NULL) || r;
+  r = ((glWindowPos3fARB = (PFNGLWINDOWPOS3FARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3fARB")) == NULL) || r;
+  r = ((glWindowPos3fvARB = (PFNGLWINDOWPOS3FVARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3fvARB")) == NULL) || r;
+  r = ((glWindowPos3iARB = (PFNGLWINDOWPOS3IARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3iARB")) == NULL) || r;
+  r = ((glWindowPos3ivARB = (PFNGLWINDOWPOS3IVARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3ivARB")) == NULL) || r;
+  r = ((glWindowPos3sARB = (PFNGLWINDOWPOS3SARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3sARB")) == NULL) || r;
+  r = ((glWindowPos3svARB = (PFNGLWINDOWPOS3SVARBPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3svARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ARB_window_pos */
+
+#ifdef GL_ATIX_point_sprites
+
+#endif /* GL_ATIX_point_sprites */
+
+#ifdef GL_ATIX_texture_env_combine3
+
+#endif /* GL_ATIX_texture_env_combine3 */
+
+#ifdef GL_ATIX_texture_env_route
+
+#endif /* GL_ATIX_texture_env_route */
+
+#ifdef GL_ATIX_vertex_shader_output_point_size
+
+#endif /* GL_ATIX_vertex_shader_output_point_size */
+
+#ifdef GL_ATI_draw_buffers
+
+static GLboolean _glewInit_GL_ATI_draw_buffers (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDrawBuffersATI = (PFNGLDRAWBUFFERSATIPROC)glewGetProcAddress((const GLubyte*)"glDrawBuffersATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ATI_draw_buffers */
+
+#ifdef GL_ATI_element_array
+
+static GLboolean _glewInit_GL_ATI_element_array (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDrawElementArrayATI = (PFNGLDRAWELEMENTARRAYATIPROC)glewGetProcAddress((const GLubyte*)"glDrawElementArrayATI")) == NULL) || r;
+  r = ((glDrawRangeElementArrayATI = (PFNGLDRAWRANGEELEMENTARRAYATIPROC)glewGetProcAddress((const GLubyte*)"glDrawRangeElementArrayATI")) == NULL) || r;
+  r = ((glElementPointerATI = (PFNGLELEMENTPOINTERATIPROC)glewGetProcAddress((const GLubyte*)"glElementPointerATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ATI_element_array */
+
+#ifdef GL_ATI_envmap_bumpmap
+
+static GLboolean _glewInit_GL_ATI_envmap_bumpmap (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetTexBumpParameterfvATI = (PFNGLGETTEXBUMPPARAMETERFVATIPROC)glewGetProcAddress((const GLubyte*)"glGetTexBumpParameterfvATI")) == NULL) || r;
+  r = ((glGetTexBumpParameterivATI = (PFNGLGETTEXBUMPPARAMETERIVATIPROC)glewGetProcAddress((const GLubyte*)"glGetTexBumpParameterivATI")) == NULL) || r;
+  r = ((glTexBumpParameterfvATI = (PFNGLTEXBUMPPARAMETERFVATIPROC)glewGetProcAddress((const GLubyte*)"glTexBumpParameterfvATI")) == NULL) || r;
+  r = ((glTexBumpParameterivATI = (PFNGLTEXBUMPPARAMETERIVATIPROC)glewGetProcAddress((const GLubyte*)"glTexBumpParameterivATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ATI_envmap_bumpmap */
+
+#ifdef GL_ATI_fragment_shader
+
+static GLboolean _glewInit_GL_ATI_fragment_shader (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glAlphaFragmentOp1ATI = (PFNGLALPHAFRAGMENTOP1ATIPROC)glewGetProcAddress((const GLubyte*)"glAlphaFragmentOp1ATI")) == NULL) || r;
+  r = ((glAlphaFragmentOp2ATI = (PFNGLALPHAFRAGMENTOP2ATIPROC)glewGetProcAddress((const GLubyte*)"glAlphaFragmentOp2ATI")) == NULL) || r;
+  r = ((glAlphaFragmentOp3ATI = (PFNGLALPHAFRAGMENTOP3ATIPROC)glewGetProcAddress((const GLubyte*)"glAlphaFragmentOp3ATI")) == NULL) || r;
+  r = ((glBeginFragmentShaderATI = (PFNGLBEGINFRAGMENTSHADERATIPROC)glewGetProcAddress((const GLubyte*)"glBeginFragmentShaderATI")) == NULL) || r;
+  r = ((glBindFragmentShaderATI = (PFNGLBINDFRAGMENTSHADERATIPROC)glewGetProcAddress((const GLubyte*)"glBindFragmentShaderATI")) == NULL) || r;
+  r = ((glColorFragmentOp1ATI = (PFNGLCOLORFRAGMENTOP1ATIPROC)glewGetProcAddress((const GLubyte*)"glColorFragmentOp1ATI")) == NULL) || r;
+  r = ((glColorFragmentOp2ATI = (PFNGLCOLORFRAGMENTOP2ATIPROC)glewGetProcAddress((const GLubyte*)"glColorFragmentOp2ATI")) == NULL) || r;
+  r = ((glColorFragmentOp3ATI = (PFNGLCOLORFRAGMENTOP3ATIPROC)glewGetProcAddress((const GLubyte*)"glColorFragmentOp3ATI")) == NULL) || r;
+  r = ((glDeleteFragmentShaderATI = (PFNGLDELETEFRAGMENTSHADERATIPROC)glewGetProcAddress((const GLubyte*)"glDeleteFragmentShaderATI")) == NULL) || r;
+  r = ((glEndFragmentShaderATI = (PFNGLENDFRAGMENTSHADERATIPROC)glewGetProcAddress((const GLubyte*)"glEndFragmentShaderATI")) == NULL) || r;
+  r = ((glGenFragmentShadersATI = (PFNGLGENFRAGMENTSHADERSATIPROC)glewGetProcAddress((const GLubyte*)"glGenFragmentShadersATI")) == NULL) || r;
+  r = ((glPassTexCoordATI = (PFNGLPASSTEXCOORDATIPROC)glewGetProcAddress((const GLubyte*)"glPassTexCoordATI")) == NULL) || r;
+  r = ((glSampleMapATI = (PFNGLSAMPLEMAPATIPROC)glewGetProcAddress((const GLubyte*)"glSampleMapATI")) == NULL) || r;
+  r = ((glSetFragmentShaderConstantATI = (PFNGLSETFRAGMENTSHADERCONSTANTATIPROC)glewGetProcAddress((const GLubyte*)"glSetFragmentShaderConstantATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ATI_fragment_shader */
+
+#ifdef GL_ATI_map_object_buffer
+
+static GLboolean _glewInit_GL_ATI_map_object_buffer (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glMapObjectBufferATI = (PFNGLMAPOBJECTBUFFERATIPROC)glewGetProcAddress((const GLubyte*)"glMapObjectBufferATI")) == NULL) || r;
+  r = ((glUnmapObjectBufferATI = (PFNGLUNMAPOBJECTBUFFERATIPROC)glewGetProcAddress((const GLubyte*)"glUnmapObjectBufferATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ATI_map_object_buffer */
+
+#ifdef GL_ATI_pn_triangles
+
+static GLboolean _glewInit_GL_ATI_pn_triangles (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glPNTrianglesfATI = (PFNGLPNTRIANGLESFATIPROC)glewGetProcAddress((const GLubyte*)"glPNTrianglesfATI")) == NULL) || r;
+  r = ((glPNTrianglesiATI = (PFNGLPNTRIANGLESIATIPROC)glewGetProcAddress((const GLubyte*)"glPNTrianglesiATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ATI_pn_triangles */
+
+#ifdef GL_ATI_separate_stencil
+
+static GLboolean _glewInit_GL_ATI_separate_stencil (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glStencilFuncSeparateATI = (PFNGLSTENCILFUNCSEPARATEATIPROC)glewGetProcAddress((const GLubyte*)"glStencilFuncSeparateATI")) == NULL) || r;
+  r = ((glStencilOpSeparateATI = (PFNGLSTENCILOPSEPARATEATIPROC)glewGetProcAddress((const GLubyte*)"glStencilOpSeparateATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ATI_separate_stencil */
+
+#ifdef GL_ATI_shader_texture_lod
+
+#endif /* GL_ATI_shader_texture_lod */
+
+#ifdef GL_ATI_text_fragment_shader
+
+#endif /* GL_ATI_text_fragment_shader */
+
+#ifdef GL_ATI_texture_compression_3dc
+
+#endif /* GL_ATI_texture_compression_3dc */
+
+#ifdef GL_ATI_texture_env_combine3
+
+#endif /* GL_ATI_texture_env_combine3 */
+
+#ifdef GL_ATI_texture_float
+
+#endif /* GL_ATI_texture_float */
+
+#ifdef GL_ATI_texture_mirror_once
+
+#endif /* GL_ATI_texture_mirror_once */
+
+#ifdef GL_ATI_vertex_array_object
+
+static GLboolean _glewInit_GL_ATI_vertex_array_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glArrayObjectATI = (PFNGLARRAYOBJECTATIPROC)glewGetProcAddress((const GLubyte*)"glArrayObjectATI")) == NULL) || r;
+  r = ((glFreeObjectBufferATI = (PFNGLFREEOBJECTBUFFERATIPROC)glewGetProcAddress((const GLubyte*)"glFreeObjectBufferATI")) == NULL) || r;
+  r = ((glGetArrayObjectfvATI = (PFNGLGETARRAYOBJECTFVATIPROC)glewGetProcAddress((const GLubyte*)"glGetArrayObjectfvATI")) == NULL) || r;
+  r = ((glGetArrayObjectivATI = (PFNGLGETARRAYOBJECTIVATIPROC)glewGetProcAddress((const GLubyte*)"glGetArrayObjectivATI")) == NULL) || r;
+  r = ((glGetObjectBufferfvATI = (PFNGLGETOBJECTBUFFERFVATIPROC)glewGetProcAddress((const GLubyte*)"glGetObjectBufferfvATI")) == NULL) || r;
+  r = ((glGetObjectBufferivATI = (PFNGLGETOBJECTBUFFERIVATIPROC)glewGetProcAddress((const GLubyte*)"glGetObjectBufferivATI")) == NULL) || r;
+  r = ((glGetVariantArrayObjectfvATI = (PFNGLGETVARIANTARRAYOBJECTFVATIPROC)glewGetProcAddress((const GLubyte*)"glGetVariantArrayObjectfvATI")) == NULL) || r;
+  r = ((glGetVariantArrayObjectivATI = (PFNGLGETVARIANTARRAYOBJECTIVATIPROC)glewGetProcAddress((const GLubyte*)"glGetVariantArrayObjectivATI")) == NULL) || r;
+  r = ((glIsObjectBufferATI = (PFNGLISOBJECTBUFFERATIPROC)glewGetProcAddress((const GLubyte*)"glIsObjectBufferATI")) == NULL) || r;
+  r = ((glNewObjectBufferATI = (PFNGLNEWOBJECTBUFFERATIPROC)glewGetProcAddress((const GLubyte*)"glNewObjectBufferATI")) == NULL) || r;
+  r = ((glUpdateObjectBufferATI = (PFNGLUPDATEOBJECTBUFFERATIPROC)glewGetProcAddress((const GLubyte*)"glUpdateObjectBufferATI")) == NULL) || r;
+  r = ((glVariantArrayObjectATI = (PFNGLVARIANTARRAYOBJECTATIPROC)glewGetProcAddress((const GLubyte*)"glVariantArrayObjectATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ATI_vertex_array_object */
+
+#ifdef GL_ATI_vertex_attrib_array_object
+
+static GLboolean _glewInit_GL_ATI_vertex_attrib_array_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetVertexAttribArrayObjectfvATI = (PFNGLGETVERTEXATTRIBARRAYOBJECTFVATIPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribArrayObjectfvATI")) == NULL) || r;
+  r = ((glGetVertexAttribArrayObjectivATI = (PFNGLGETVERTEXATTRIBARRAYOBJECTIVATIPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribArrayObjectivATI")) == NULL) || r;
+  r = ((glVertexAttribArrayObjectATI = (PFNGLVERTEXATTRIBARRAYOBJECTATIPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribArrayObjectATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ATI_vertex_attrib_array_object */
+
+#ifdef GL_ATI_vertex_streams
+
+static GLboolean _glewInit_GL_ATI_vertex_streams (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glClientActiveVertexStreamATI = (PFNGLCLIENTACTIVEVERTEXSTREAMATIPROC)glewGetProcAddress((const GLubyte*)"glClientActiveVertexStreamATI")) == NULL) || r;
+  r = ((glNormalStream3bATI = (PFNGLNORMALSTREAM3BATIPROC)glewGetProcAddress((const GLubyte*)"glNormalStream3bATI")) == NULL) || r;
+  r = ((glNormalStream3bvATI = (PFNGLNORMALSTREAM3BVATIPROC)glewGetProcAddress((const GLubyte*)"glNormalStream3bvATI")) == NULL) || r;
+  r = ((glNormalStream3dATI = (PFNGLNORMALSTREAM3DATIPROC)glewGetProcAddress((const GLubyte*)"glNormalStream3dATI")) == NULL) || r;
+  r = ((glNormalStream3dvATI = (PFNGLNORMALSTREAM3DVATIPROC)glewGetProcAddress((const GLubyte*)"glNormalStream3dvATI")) == NULL) || r;
+  r = ((glNormalStream3fATI = (PFNGLNORMALSTREAM3FATIPROC)glewGetProcAddress((const GLubyte*)"glNormalStream3fATI")) == NULL) || r;
+  r = ((glNormalStream3fvATI = (PFNGLNORMALSTREAM3FVATIPROC)glewGetProcAddress((const GLubyte*)"glNormalStream3fvATI")) == NULL) || r;
+  r = ((glNormalStream3iATI = (PFNGLNORMALSTREAM3IATIPROC)glewGetProcAddress((const GLubyte*)"glNormalStream3iATI")) == NULL) || r;
+  r = ((glNormalStream3ivATI = (PFNGLNORMALSTREAM3IVATIPROC)glewGetProcAddress((const GLubyte*)"glNormalStream3ivATI")) == NULL) || r;
+  r = ((glNormalStream3sATI = (PFNGLNORMALSTREAM3SATIPROC)glewGetProcAddress((const GLubyte*)"glNormalStream3sATI")) == NULL) || r;
+  r = ((glNormalStream3svATI = (PFNGLNORMALSTREAM3SVATIPROC)glewGetProcAddress((const GLubyte*)"glNormalStream3svATI")) == NULL) || r;
+  r = ((glVertexBlendEnvfATI = (PFNGLVERTEXBLENDENVFATIPROC)glewGetProcAddress((const GLubyte*)"glVertexBlendEnvfATI")) == NULL) || r;
+  r = ((glVertexBlendEnviATI = (PFNGLVERTEXBLENDENVIATIPROC)glewGetProcAddress((const GLubyte*)"glVertexBlendEnviATI")) == NULL) || r;
+  r = ((glVertexStream2dATI = (PFNGLVERTEXSTREAM2DATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream2dATI")) == NULL) || r;
+  r = ((glVertexStream2dvATI = (PFNGLVERTEXSTREAM2DVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream2dvATI")) == NULL) || r;
+  r = ((glVertexStream2fATI = (PFNGLVERTEXSTREAM2FATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream2fATI")) == NULL) || r;
+  r = ((glVertexStream2fvATI = (PFNGLVERTEXSTREAM2FVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream2fvATI")) == NULL) || r;
+  r = ((glVertexStream2iATI = (PFNGLVERTEXSTREAM2IATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream2iATI")) == NULL) || r;
+  r = ((glVertexStream2ivATI = (PFNGLVERTEXSTREAM2IVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream2ivATI")) == NULL) || r;
+  r = ((glVertexStream2sATI = (PFNGLVERTEXSTREAM2SATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream2sATI")) == NULL) || r;
+  r = ((glVertexStream2svATI = (PFNGLVERTEXSTREAM2SVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream2svATI")) == NULL) || r;
+  r = ((glVertexStream3dATI = (PFNGLVERTEXSTREAM3DATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream3dATI")) == NULL) || r;
+  r = ((glVertexStream3dvATI = (PFNGLVERTEXSTREAM3DVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream3dvATI")) == NULL) || r;
+  r = ((glVertexStream3fATI = (PFNGLVERTEXSTREAM3FATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream3fATI")) == NULL) || r;
+  r = ((glVertexStream3fvATI = (PFNGLVERTEXSTREAM3FVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream3fvATI")) == NULL) || r;
+  r = ((glVertexStream3iATI = (PFNGLVERTEXSTREAM3IATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream3iATI")) == NULL) || r;
+  r = ((glVertexStream3ivATI = (PFNGLVERTEXSTREAM3IVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream3ivATI")) == NULL) || r;
+  r = ((glVertexStream3sATI = (PFNGLVERTEXSTREAM3SATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream3sATI")) == NULL) || r;
+  r = ((glVertexStream3svATI = (PFNGLVERTEXSTREAM3SVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream3svATI")) == NULL) || r;
+  r = ((glVertexStream4dATI = (PFNGLVERTEXSTREAM4DATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream4dATI")) == NULL) || r;
+  r = ((glVertexStream4dvATI = (PFNGLVERTEXSTREAM4DVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream4dvATI")) == NULL) || r;
+  r = ((glVertexStream4fATI = (PFNGLVERTEXSTREAM4FATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream4fATI")) == NULL) || r;
+  r = ((glVertexStream4fvATI = (PFNGLVERTEXSTREAM4FVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream4fvATI")) == NULL) || r;
+  r = ((glVertexStream4iATI = (PFNGLVERTEXSTREAM4IATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream4iATI")) == NULL) || r;
+  r = ((glVertexStream4ivATI = (PFNGLVERTEXSTREAM4IVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream4ivATI")) == NULL) || r;
+  r = ((glVertexStream4sATI = (PFNGLVERTEXSTREAM4SATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream4sATI")) == NULL) || r;
+  r = ((glVertexStream4svATI = (PFNGLVERTEXSTREAM4SVATIPROC)glewGetProcAddress((const GLubyte*)"glVertexStream4svATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_ATI_vertex_streams */
+
+#ifdef GL_EXT_422_pixels
+
+#endif /* GL_EXT_422_pixels */
+
+#ifdef GL_EXT_Cg_shader
+
+#endif /* GL_EXT_Cg_shader */
+
+#ifdef GL_EXT_abgr
+
+#endif /* GL_EXT_abgr */
+
+#ifdef GL_EXT_bgra
+
+#endif /* GL_EXT_bgra */
+
+#ifdef GL_EXT_bindable_uniform
+
+static GLboolean _glewInit_GL_EXT_bindable_uniform (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetUniformBufferSizeEXT = (PFNGLGETUNIFORMBUFFERSIZEEXTPROC)glewGetProcAddress((const GLubyte*)"glGetUniformBufferSizeEXT")) == NULL) || r;
+  r = ((glGetUniformOffsetEXT = (PFNGLGETUNIFORMOFFSETEXTPROC)glewGetProcAddress((const GLubyte*)"glGetUniformOffsetEXT")) == NULL) || r;
+  r = ((glUniformBufferEXT = (PFNGLUNIFORMBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glUniformBufferEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_bindable_uniform */
+
+#ifdef GL_EXT_blend_color
+
+static GLboolean _glewInit_GL_EXT_blend_color (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBlendColorEXT = (PFNGLBLENDCOLOREXTPROC)glewGetProcAddress((const GLubyte*)"glBlendColorEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_blend_color */
+
+#ifdef GL_EXT_blend_equation_separate
+
+static GLboolean _glewInit_GL_EXT_blend_equation_separate (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBlendEquationSeparateEXT = (PFNGLBLENDEQUATIONSEPARATEEXTPROC)glewGetProcAddress((const GLubyte*)"glBlendEquationSeparateEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_blend_equation_separate */
+
+#ifdef GL_EXT_blend_func_separate
+
+static GLboolean _glewInit_GL_EXT_blend_func_separate (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBlendFuncSeparateEXT = (PFNGLBLENDFUNCSEPARATEEXTPROC)glewGetProcAddress((const GLubyte*)"glBlendFuncSeparateEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_blend_func_separate */
+
+#ifdef GL_EXT_blend_logic_op
+
+#endif /* GL_EXT_blend_logic_op */
+
+#ifdef GL_EXT_blend_minmax
+
+static GLboolean _glewInit_GL_EXT_blend_minmax (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBlendEquationEXT = (PFNGLBLENDEQUATIONEXTPROC)glewGetProcAddress((const GLubyte*)"glBlendEquationEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_blend_minmax */
+
+#ifdef GL_EXT_blend_subtract
+
+#endif /* GL_EXT_blend_subtract */
+
+#ifdef GL_EXT_clip_volume_hint
+
+#endif /* GL_EXT_clip_volume_hint */
+
+#ifdef GL_EXT_cmyka
+
+#endif /* GL_EXT_cmyka */
+
+#ifdef GL_EXT_color_subtable
+
+static GLboolean _glewInit_GL_EXT_color_subtable (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glColorSubTableEXT = (PFNGLCOLORSUBTABLEEXTPROC)glewGetProcAddress((const GLubyte*)"glColorSubTableEXT")) == NULL) || r;
+  r = ((glCopyColorSubTableEXT = (PFNGLCOPYCOLORSUBTABLEEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyColorSubTableEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_color_subtable */
+
+#ifdef GL_EXT_compiled_vertex_array
+
+static GLboolean _glewInit_GL_EXT_compiled_vertex_array (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glLockArraysEXT = (PFNGLLOCKARRAYSEXTPROC)glewGetProcAddress((const GLubyte*)"glLockArraysEXT")) == NULL) || r;
+  r = ((glUnlockArraysEXT = (PFNGLUNLOCKARRAYSEXTPROC)glewGetProcAddress((const GLubyte*)"glUnlockArraysEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_compiled_vertex_array */
+
+#ifdef GL_EXT_convolution
+
+static GLboolean _glewInit_GL_EXT_convolution (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glConvolutionFilter1DEXT = (PFNGLCONVOLUTIONFILTER1DEXTPROC)glewGetProcAddress((const GLubyte*)"glConvolutionFilter1DEXT")) == NULL) || r;
+  r = ((glConvolutionFilter2DEXT = (PFNGLCONVOLUTIONFILTER2DEXTPROC)glewGetProcAddress((const GLubyte*)"glConvolutionFilter2DEXT")) == NULL) || r;
+  r = ((glConvolutionParameterfEXT = (PFNGLCONVOLUTIONPARAMETERFEXTPROC)glewGetProcAddress((const GLubyte*)"glConvolutionParameterfEXT")) == NULL) || r;
+  r = ((glConvolutionParameterfvEXT = (PFNGLCONVOLUTIONPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glConvolutionParameterfvEXT")) == NULL) || r;
+  r = ((glConvolutionParameteriEXT = (PFNGLCONVOLUTIONPARAMETERIEXTPROC)glewGetProcAddress((const GLubyte*)"glConvolutionParameteriEXT")) == NULL) || r;
+  r = ((glConvolutionParameterivEXT = (PFNGLCONVOLUTIONPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glConvolutionParameterivEXT")) == NULL) || r;
+  r = ((glCopyConvolutionFilter1DEXT = (PFNGLCOPYCONVOLUTIONFILTER1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyConvolutionFilter1DEXT")) == NULL) || r;
+  r = ((glCopyConvolutionFilter2DEXT = (PFNGLCOPYCONVOLUTIONFILTER2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyConvolutionFilter2DEXT")) == NULL) || r;
+  r = ((glGetConvolutionFilterEXT = (PFNGLGETCONVOLUTIONFILTEREXTPROC)glewGetProcAddress((const GLubyte*)"glGetConvolutionFilterEXT")) == NULL) || r;
+  r = ((glGetConvolutionParameterfvEXT = (PFNGLGETCONVOLUTIONPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetConvolutionParameterfvEXT")) == NULL) || r;
+  r = ((glGetConvolutionParameterivEXT = (PFNGLGETCONVOLUTIONPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetConvolutionParameterivEXT")) == NULL) || r;
+  r = ((glGetSeparableFilterEXT = (PFNGLGETSEPARABLEFILTEREXTPROC)glewGetProcAddress((const GLubyte*)"glGetSeparableFilterEXT")) == NULL) || r;
+  r = ((glSeparableFilter2DEXT = (PFNGLSEPARABLEFILTER2DEXTPROC)glewGetProcAddress((const GLubyte*)"glSeparableFilter2DEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_convolution */
+
+#ifdef GL_EXT_coordinate_frame
+
+static GLboolean _glewInit_GL_EXT_coordinate_frame (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBinormalPointerEXT = (PFNGLBINORMALPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glBinormalPointerEXT")) == NULL) || r;
+  r = ((glTangentPointerEXT = (PFNGLTANGENTPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glTangentPointerEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_coordinate_frame */
+
+#ifdef GL_EXT_copy_texture
+
+static GLboolean _glewInit_GL_EXT_copy_texture (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glCopyTexImage1DEXT = (PFNGLCOPYTEXIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyTexImage1DEXT")) == NULL) || r;
+  r = ((glCopyTexImage2DEXT = (PFNGLCOPYTEXIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyTexImage2DEXT")) == NULL) || r;
+  r = ((glCopyTexSubImage1DEXT = (PFNGLCOPYTEXSUBIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyTexSubImage1DEXT")) == NULL) || r;
+  r = ((glCopyTexSubImage2DEXT = (PFNGLCOPYTEXSUBIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyTexSubImage2DEXT")) == NULL) || r;
+  r = ((glCopyTexSubImage3DEXT = (PFNGLCOPYTEXSUBIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyTexSubImage3DEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_copy_texture */
+
+#ifdef GL_EXT_cull_vertex
+
+static GLboolean _glewInit_GL_EXT_cull_vertex (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glCullParameterdvEXT = (PFNGLCULLPARAMETERDVEXTPROC)glewGetProcAddress((const GLubyte*)"glCullParameterdvEXT")) == NULL) || r;
+  r = ((glCullParameterfvEXT = (PFNGLCULLPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glCullParameterfvEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_cull_vertex */
+
+#ifdef GL_EXT_depth_bounds_test
+
+static GLboolean _glewInit_GL_EXT_depth_bounds_test (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDepthBoundsEXT = (PFNGLDEPTHBOUNDSEXTPROC)glewGetProcAddress((const GLubyte*)"glDepthBoundsEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_depth_bounds_test */
+
+#ifdef GL_EXT_direct_state_access
+
+static GLboolean _glewInit_GL_EXT_direct_state_access (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBindMultiTextureEXT = (PFNGLBINDMULTITEXTUREEXTPROC)glewGetProcAddress((const GLubyte*)"glBindMultiTextureEXT")) == NULL) || r;
+  r = ((glCheckNamedFramebufferStatusEXT = (PFNGLCHECKNAMEDFRAMEBUFFERSTATUSEXTPROC)glewGetProcAddress((const GLubyte*)"glCheckNamedFramebufferStatusEXT")) == NULL) || r;
+  r = ((glClientAttribDefaultEXT = (PFNGLCLIENTATTRIBDEFAULTEXTPROC)glewGetProcAddress((const GLubyte*)"glClientAttribDefaultEXT")) == NULL) || r;
+  r = ((glCompressedMultiTexImage1DEXT = (PFNGLCOMPRESSEDMULTITEXIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedMultiTexImage1DEXT")) == NULL) || r;
+  r = ((glCompressedMultiTexImage2DEXT = (PFNGLCOMPRESSEDMULTITEXIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedMultiTexImage2DEXT")) == NULL) || r;
+  r = ((glCompressedMultiTexImage3DEXT = (PFNGLCOMPRESSEDMULTITEXIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedMultiTexImage3DEXT")) == NULL) || r;
+  r = ((glCompressedMultiTexSubImage1DEXT = (PFNGLCOMPRESSEDMULTITEXSUBIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedMultiTexSubImage1DEXT")) == NULL) || r;
+  r = ((glCompressedMultiTexSubImage2DEXT = (PFNGLCOMPRESSEDMULTITEXSUBIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedMultiTexSubImage2DEXT")) == NULL) || r;
+  r = ((glCompressedMultiTexSubImage3DEXT = (PFNGLCOMPRESSEDMULTITEXSUBIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedMultiTexSubImage3DEXT")) == NULL) || r;
+  r = ((glCompressedTextureImage1DEXT = (PFNGLCOMPRESSEDTEXTUREIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedTextureImage1DEXT")) == NULL) || r;
+  r = ((glCompressedTextureImage2DEXT = (PFNGLCOMPRESSEDTEXTUREIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedTextureImage2DEXT")) == NULL) || r;
+  r = ((glCompressedTextureImage3DEXT = (PFNGLCOMPRESSEDTEXTUREIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedTextureImage3DEXT")) == NULL) || r;
+  r = ((glCompressedTextureSubImage1DEXT = (PFNGLCOMPRESSEDTEXTURESUBIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedTextureSubImage1DEXT")) == NULL) || r;
+  r = ((glCompressedTextureSubImage2DEXT = (PFNGLCOMPRESSEDTEXTURESUBIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedTextureSubImage2DEXT")) == NULL) || r;
+  r = ((glCompressedTextureSubImage3DEXT = (PFNGLCOMPRESSEDTEXTURESUBIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glCompressedTextureSubImage3DEXT")) == NULL) || r;
+  r = ((glCopyMultiTexImage1DEXT = (PFNGLCOPYMULTITEXIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyMultiTexImage1DEXT")) == NULL) || r;
+  r = ((glCopyMultiTexImage2DEXT = (PFNGLCOPYMULTITEXIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyMultiTexImage2DEXT")) == NULL) || r;
+  r = ((glCopyMultiTexSubImage1DEXT = (PFNGLCOPYMULTITEXSUBIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyMultiTexSubImage1DEXT")) == NULL) || r;
+  r = ((glCopyMultiTexSubImage2DEXT = (PFNGLCOPYMULTITEXSUBIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyMultiTexSubImage2DEXT")) == NULL) || r;
+  r = ((glCopyMultiTexSubImage3DEXT = (PFNGLCOPYMULTITEXSUBIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyMultiTexSubImage3DEXT")) == NULL) || r;
+  r = ((glCopyTextureImage1DEXT = (PFNGLCOPYTEXTUREIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyTextureImage1DEXT")) == NULL) || r;
+  r = ((glCopyTextureImage2DEXT = (PFNGLCOPYTEXTUREIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyTextureImage2DEXT")) == NULL) || r;
+  r = ((glCopyTextureSubImage1DEXT = (PFNGLCOPYTEXTURESUBIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyTextureSubImage1DEXT")) == NULL) || r;
+  r = ((glCopyTextureSubImage2DEXT = (PFNGLCOPYTEXTURESUBIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyTextureSubImage2DEXT")) == NULL) || r;
+  r = ((glCopyTextureSubImage3DEXT = (PFNGLCOPYTEXTURESUBIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glCopyTextureSubImage3DEXT")) == NULL) || r;
+  r = ((glDisableClientStateIndexedEXT = (PFNGLDISABLECLIENTSTATEINDEXEDEXTPROC)glewGetProcAddress((const GLubyte*)"glDisableClientStateIndexedEXT")) == NULL) || r;
+  r = ((glEnableClientStateIndexedEXT = (PFNGLENABLECLIENTSTATEINDEXEDEXTPROC)glewGetProcAddress((const GLubyte*)"glEnableClientStateIndexedEXT")) == NULL) || r;
+  r = ((glFramebufferDrawBufferEXT = (PFNGLFRAMEBUFFERDRAWBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glFramebufferDrawBufferEXT")) == NULL) || r;
+  r = ((glFramebufferDrawBuffersEXT = (PFNGLFRAMEBUFFERDRAWBUFFERSEXTPROC)glewGetProcAddress((const GLubyte*)"glFramebufferDrawBuffersEXT")) == NULL) || r;
+  r = ((glFramebufferReadBufferEXT = (PFNGLFRAMEBUFFERREADBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glFramebufferReadBufferEXT")) == NULL) || r;
+  r = ((glGenerateMultiTexMipmapEXT = (PFNGLGENERATEMULTITEXMIPMAPEXTPROC)glewGetProcAddress((const GLubyte*)"glGenerateMultiTexMipmapEXT")) == NULL) || r;
+  r = ((glGenerateTextureMipmapEXT = (PFNGLGENERATETEXTUREMIPMAPEXTPROC)glewGetProcAddress((const GLubyte*)"glGenerateTextureMipmapEXT")) == NULL) || r;
+  r = ((glGetCompressedMultiTexImageEXT = (PFNGLGETCOMPRESSEDMULTITEXIMAGEEXTPROC)glewGetProcAddress((const GLubyte*)"glGetCompressedMultiTexImageEXT")) == NULL) || r;
+  r = ((glGetCompressedTextureImageEXT = (PFNGLGETCOMPRESSEDTEXTUREIMAGEEXTPROC)glewGetProcAddress((const GLubyte*)"glGetCompressedTextureImageEXT")) == NULL) || r;
+  r = ((glGetDoubleIndexedvEXT = (PFNGLGETDOUBLEINDEXEDVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetDoubleIndexedvEXT")) == NULL) || r;
+  r = ((glGetFloatIndexedvEXT = (PFNGLGETFLOATINDEXEDVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetFloatIndexedvEXT")) == NULL) || r;
+  r = ((glGetFramebufferParameterivEXT = (PFNGLGETFRAMEBUFFERPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetFramebufferParameterivEXT")) == NULL) || r;
+  r = ((glGetMultiTexEnvfvEXT = (PFNGLGETMULTITEXENVFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexEnvfvEXT")) == NULL) || r;
+  r = ((glGetMultiTexEnvivEXT = (PFNGLGETMULTITEXENVIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexEnvivEXT")) == NULL) || r;
+  r = ((glGetMultiTexGendvEXT = (PFNGLGETMULTITEXGENDVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexGendvEXT")) == NULL) || r;
+  r = ((glGetMultiTexGenfvEXT = (PFNGLGETMULTITEXGENFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexGenfvEXT")) == NULL) || r;
+  r = ((glGetMultiTexGenivEXT = (PFNGLGETMULTITEXGENIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexGenivEXT")) == NULL) || r;
+  r = ((glGetMultiTexImageEXT = (PFNGLGETMULTITEXIMAGEEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexImageEXT")) == NULL) || r;
+  r = ((glGetMultiTexLevelParameterfvEXT = (PFNGLGETMULTITEXLEVELPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexLevelParameterfvEXT")) == NULL) || r;
+  r = ((glGetMultiTexLevelParameterivEXT = (PFNGLGETMULTITEXLEVELPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexLevelParameterivEXT")) == NULL) || r;
+  r = ((glGetMultiTexParameterIivEXT = (PFNGLGETMULTITEXPARAMETERIIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexParameterIivEXT")) == NULL) || r;
+  r = ((glGetMultiTexParameterIuivEXT = (PFNGLGETMULTITEXPARAMETERIUIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexParameterIuivEXT")) == NULL) || r;
+  r = ((glGetMultiTexParameterfvEXT = (PFNGLGETMULTITEXPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexParameterfvEXT")) == NULL) || r;
+  r = ((glGetMultiTexParameterivEXT = (PFNGLGETMULTITEXPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMultiTexParameterivEXT")) == NULL) || r;
+  r = ((glGetNamedBufferParameterivEXT = (PFNGLGETNAMEDBUFFERPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedBufferParameterivEXT")) == NULL) || r;
+  r = ((glGetNamedBufferPointervEXT = (PFNGLGETNAMEDBUFFERPOINTERVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedBufferPointervEXT")) == NULL) || r;
+  r = ((glGetNamedBufferSubDataEXT = (PFNGLGETNAMEDBUFFERSUBDATAEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedBufferSubDataEXT")) == NULL) || r;
+  r = ((glGetNamedFramebufferAttachmentParameterivEXT = (PFNGLGETNAMEDFRAMEBUFFERATTACHMENTPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedFramebufferAttachmentParameterivEXT")) == NULL) || r;
+  r = ((glGetNamedProgramLocalParameterIivEXT = (PFNGLGETNAMEDPROGRAMLOCALPARAMETERIIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedProgramLocalParameterIivEXT")) == NULL) || r;
+  r = ((glGetNamedProgramLocalParameterIuivEXT = (PFNGLGETNAMEDPROGRAMLOCALPARAMETERIUIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedProgramLocalParameterIuivEXT")) == NULL) || r;
+  r = ((glGetNamedProgramLocalParameterdvEXT = (PFNGLGETNAMEDPROGRAMLOCALPARAMETERDVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedProgramLocalParameterdvEXT")) == NULL) || r;
+  r = ((glGetNamedProgramLocalParameterfvEXT = (PFNGLGETNAMEDPROGRAMLOCALPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedProgramLocalParameterfvEXT")) == NULL) || r;
+  r = ((glGetNamedProgramStringEXT = (PFNGLGETNAMEDPROGRAMSTRINGEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedProgramStringEXT")) == NULL) || r;
+  r = ((glGetNamedProgramivEXT = (PFNGLGETNAMEDPROGRAMIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedProgramivEXT")) == NULL) || r;
+  r = ((glGetNamedRenderbufferParameterivEXT = (PFNGLGETNAMEDRENDERBUFFERPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetNamedRenderbufferParameterivEXT")) == NULL) || r;
+  r = ((glGetPointerIndexedvEXT = (PFNGLGETPOINTERINDEXEDVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetPointerIndexedvEXT")) == NULL) || r;
+  r = ((glGetTextureImageEXT = (PFNGLGETTEXTUREIMAGEEXTPROC)glewGetProcAddress((const GLubyte*)"glGetTextureImageEXT")) == NULL) || r;
+  r = ((glGetTextureLevelParameterfvEXT = (PFNGLGETTEXTURELEVELPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetTextureLevelParameterfvEXT")) == NULL) || r;
+  r = ((glGetTextureLevelParameterivEXT = (PFNGLGETTEXTURELEVELPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetTextureLevelParameterivEXT")) == NULL) || r;
+  r = ((glGetTextureParameterIivEXT = (PFNGLGETTEXTUREPARAMETERIIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetTextureParameterIivEXT")) == NULL) || r;
+  r = ((glGetTextureParameterIuivEXT = (PFNGLGETTEXTUREPARAMETERIUIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetTextureParameterIuivEXT")) == NULL) || r;
+  r = ((glGetTextureParameterfvEXT = (PFNGLGETTEXTUREPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetTextureParameterfvEXT")) == NULL) || r;
+  r = ((glGetTextureParameterivEXT = (PFNGLGETTEXTUREPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetTextureParameterivEXT")) == NULL) || r;
+  r = ((glMapNamedBufferEXT = (PFNGLMAPNAMEDBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glMapNamedBufferEXT")) == NULL) || r;
+  r = ((glMatrixFrustumEXT = (PFNGLMATRIXFRUSTUMEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixFrustumEXT")) == NULL) || r;
+  r = ((glMatrixLoadIdentityEXT = (PFNGLMATRIXLOADIDENTITYEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixLoadIdentityEXT")) == NULL) || r;
+  r = ((glMatrixLoadTransposedEXT = (PFNGLMATRIXLOADTRANSPOSEDEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixLoadTransposedEXT")) == NULL) || r;
+  r = ((glMatrixLoadTransposefEXT = (PFNGLMATRIXLOADTRANSPOSEFEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixLoadTransposefEXT")) == NULL) || r;
+  r = ((glMatrixLoaddEXT = (PFNGLMATRIXLOADDEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixLoaddEXT")) == NULL) || r;
+  r = ((glMatrixLoadfEXT = (PFNGLMATRIXLOADFEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixLoadfEXT")) == NULL) || r;
+  r = ((glMatrixMultTransposedEXT = (PFNGLMATRIXMULTTRANSPOSEDEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixMultTransposedEXT")) == NULL) || r;
+  r = ((glMatrixMultTransposefEXT = (PFNGLMATRIXMULTTRANSPOSEFEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixMultTransposefEXT")) == NULL) || r;
+  r = ((glMatrixMultdEXT = (PFNGLMATRIXMULTDEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixMultdEXT")) == NULL) || r;
+  r = ((glMatrixMultfEXT = (PFNGLMATRIXMULTFEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixMultfEXT")) == NULL) || r;
+  r = ((glMatrixOrthoEXT = (PFNGLMATRIXORTHOEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixOrthoEXT")) == NULL) || r;
+  r = ((glMatrixPopEXT = (PFNGLMATRIXPOPEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixPopEXT")) == NULL) || r;
+  r = ((glMatrixPushEXT = (PFNGLMATRIXPUSHEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixPushEXT")) == NULL) || r;
+  r = ((glMatrixRotatedEXT = (PFNGLMATRIXROTATEDEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixRotatedEXT")) == NULL) || r;
+  r = ((glMatrixRotatefEXT = (PFNGLMATRIXROTATEFEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixRotatefEXT")) == NULL) || r;
+  r = ((glMatrixScaledEXT = (PFNGLMATRIXSCALEDEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixScaledEXT")) == NULL) || r;
+  r = ((glMatrixScalefEXT = (PFNGLMATRIXSCALEFEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixScalefEXT")) == NULL) || r;
+  r = ((glMatrixTranslatedEXT = (PFNGLMATRIXTRANSLATEDEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixTranslatedEXT")) == NULL) || r;
+  r = ((glMatrixTranslatefEXT = (PFNGLMATRIXTRANSLATEFEXTPROC)glewGetProcAddress((const GLubyte*)"glMatrixTranslatefEXT")) == NULL) || r;
+  r = ((glMultiTexBufferEXT = (PFNGLMULTITEXBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexBufferEXT")) == NULL) || r;
+  r = ((glMultiTexCoordPointerEXT = (PFNGLMULTITEXCOORDPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoordPointerEXT")) == NULL) || r;
+  r = ((glMultiTexEnvfEXT = (PFNGLMULTITEXENVFEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexEnvfEXT")) == NULL) || r;
+  r = ((glMultiTexEnvfvEXT = (PFNGLMULTITEXENVFVEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexEnvfvEXT")) == NULL) || r;
+  r = ((glMultiTexEnviEXT = (PFNGLMULTITEXENVIEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexEnviEXT")) == NULL) || r;
+  r = ((glMultiTexEnvivEXT = (PFNGLMULTITEXENVIVEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexEnvivEXT")) == NULL) || r;
+  r = ((glMultiTexGendEXT = (PFNGLMULTITEXGENDEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexGendEXT")) == NULL) || r;
+  r = ((glMultiTexGendvEXT = (PFNGLMULTITEXGENDVEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexGendvEXT")) == NULL) || r;
+  r = ((glMultiTexGenfEXT = (PFNGLMULTITEXGENFEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexGenfEXT")) == NULL) || r;
+  r = ((glMultiTexGenfvEXT = (PFNGLMULTITEXGENFVEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexGenfvEXT")) == NULL) || r;
+  r = ((glMultiTexGeniEXT = (PFNGLMULTITEXGENIEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexGeniEXT")) == NULL) || r;
+  r = ((glMultiTexGenivEXT = (PFNGLMULTITEXGENIVEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexGenivEXT")) == NULL) || r;
+  r = ((glMultiTexImage1DEXT = (PFNGLMULTITEXIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexImage1DEXT")) == NULL) || r;
+  r = ((glMultiTexImage2DEXT = (PFNGLMULTITEXIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexImage2DEXT")) == NULL) || r;
+  r = ((glMultiTexImage3DEXT = (PFNGLMULTITEXIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexImage3DEXT")) == NULL) || r;
+  r = ((glMultiTexParameterIivEXT = (PFNGLMULTITEXPARAMETERIIVEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexParameterIivEXT")) == NULL) || r;
+  r = ((glMultiTexParameterIuivEXT = (PFNGLMULTITEXPARAMETERIUIVEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexParameterIuivEXT")) == NULL) || r;
+  r = ((glMultiTexParameterfEXT = (PFNGLMULTITEXPARAMETERFEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexParameterfEXT")) == NULL) || r;
+  r = ((glMultiTexParameterfvEXT = (PFNGLMULTITEXPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexParameterfvEXT")) == NULL) || r;
+  r = ((glMultiTexParameteriEXT = (PFNGLMULTITEXPARAMETERIEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexParameteriEXT")) == NULL) || r;
+  r = ((glMultiTexParameterivEXT = (PFNGLMULTITEXPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexParameterivEXT")) == NULL) || r;
+  r = ((glMultiTexRenderbufferEXT = (PFNGLMULTITEXRENDERBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexRenderbufferEXT")) == NULL) || r;
+  r = ((glMultiTexSubImage1DEXT = (PFNGLMULTITEXSUBIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexSubImage1DEXT")) == NULL) || r;
+  r = ((glMultiTexSubImage2DEXT = (PFNGLMULTITEXSUBIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexSubImage2DEXT")) == NULL) || r;
+  r = ((glMultiTexSubImage3DEXT = (PFNGLMULTITEXSUBIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiTexSubImage3DEXT")) == NULL) || r;
+  r = ((glNamedBufferDataEXT = (PFNGLNAMEDBUFFERDATAEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedBufferDataEXT")) == NULL) || r;
+  r = ((glNamedBufferSubDataEXT = (PFNGLNAMEDBUFFERSUBDATAEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedBufferSubDataEXT")) == NULL) || r;
+  r = ((glNamedFramebufferRenderbufferEXT = (PFNGLNAMEDFRAMEBUFFERRENDERBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glNamedFramebufferRenderbufferEXT")) == NULL) || r;
+  r = ((glNamedFramebufferTexture1DEXT = (PFNGLNAMEDFRAMEBUFFERTEXTURE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedFramebufferTexture1DEXT")) == NULL) || r;
+  r = ((glNamedFramebufferTexture2DEXT = (PFNGLNAMEDFRAMEBUFFERTEXTURE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedFramebufferTexture2DEXT")) == NULL) || r;
+  r = ((glNamedFramebufferTexture3DEXT = (PFNGLNAMEDFRAMEBUFFERTEXTURE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedFramebufferTexture3DEXT")) == NULL) || r;
+  r = ((glNamedFramebufferTextureEXT = (PFNGLNAMEDFRAMEBUFFERTEXTUREEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedFramebufferTextureEXT")) == NULL) || r;
+  r = ((glNamedFramebufferTextureFaceEXT = (PFNGLNAMEDFRAMEBUFFERTEXTUREFACEEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedFramebufferTextureFaceEXT")) == NULL) || r;
+  r = ((glNamedFramebufferTextureLayerEXT = (PFNGLNAMEDFRAMEBUFFERTEXTURELAYEREXTPROC)glewGetProcAddress((const GLubyte*)"glNamedFramebufferTextureLayerEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParameter4dEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETER4DEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParameter4dEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParameter4dvEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETER4DVEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParameter4dvEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParameter4fEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETER4FEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParameter4fEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParameter4fvEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETER4FVEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParameter4fvEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParameterI4iEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETERI4IEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParameterI4iEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParameterI4ivEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETERI4IVEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParameterI4ivEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParameterI4uiEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETERI4UIEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParameterI4uiEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParameterI4uivEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETERI4UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParameterI4uivEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParameters4fvEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETERS4FVEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParameters4fvEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParametersI4ivEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETERSI4IVEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParametersI4ivEXT")) == NULL) || r;
+  r = ((glNamedProgramLocalParametersI4uivEXT = (PFNGLNAMEDPROGRAMLOCALPARAMETERSI4UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramLocalParametersI4uivEXT")) == NULL) || r;
+  r = ((glNamedProgramStringEXT = (PFNGLNAMEDPROGRAMSTRINGEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedProgramStringEXT")) == NULL) || r;
+  r = ((glNamedRenderbufferStorageEXT = (PFNGLNAMEDRENDERBUFFERSTORAGEEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedRenderbufferStorageEXT")) == NULL) || r;
+  r = ((glNamedRenderbufferStorageMultisampleCoverageEXT = (PFNGLNAMEDRENDERBUFFERSTORAGEMULTISAMPLECOVERAGEEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedRenderbufferStorageMultisampleCoverageEXT")) == NULL) || r;
+  r = ((glNamedRenderbufferStorageMultisampleEXT = (PFNGLNAMEDRENDERBUFFERSTORAGEMULTISAMPLEEXTPROC)glewGetProcAddress((const GLubyte*)"glNamedRenderbufferStorageMultisampleEXT")) == NULL) || r;
+  r = ((glProgramUniform1fEXT = (PFNGLPROGRAMUNIFORM1FEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform1fEXT")) == NULL) || r;
+  r = ((glProgramUniform1fvEXT = (PFNGLPROGRAMUNIFORM1FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform1fvEXT")) == NULL) || r;
+  r = ((glProgramUniform1iEXT = (PFNGLPROGRAMUNIFORM1IEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform1iEXT")) == NULL) || r;
+  r = ((glProgramUniform1ivEXT = (PFNGLPROGRAMUNIFORM1IVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform1ivEXT")) == NULL) || r;
+  r = ((glProgramUniform1uiEXT = (PFNGLPROGRAMUNIFORM1UIEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform1uiEXT")) == NULL) || r;
+  r = ((glProgramUniform1uivEXT = (PFNGLPROGRAMUNIFORM1UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform1uivEXT")) == NULL) || r;
+  r = ((glProgramUniform2fEXT = (PFNGLPROGRAMUNIFORM2FEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform2fEXT")) == NULL) || r;
+  r = ((glProgramUniform2fvEXT = (PFNGLPROGRAMUNIFORM2FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform2fvEXT")) == NULL) || r;
+  r = ((glProgramUniform2iEXT = (PFNGLPROGRAMUNIFORM2IEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform2iEXT")) == NULL) || r;
+  r = ((glProgramUniform2ivEXT = (PFNGLPROGRAMUNIFORM2IVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform2ivEXT")) == NULL) || r;
+  r = ((glProgramUniform2uiEXT = (PFNGLPROGRAMUNIFORM2UIEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform2uiEXT")) == NULL) || r;
+  r = ((glProgramUniform2uivEXT = (PFNGLPROGRAMUNIFORM2UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform2uivEXT")) == NULL) || r;
+  r = ((glProgramUniform3fEXT = (PFNGLPROGRAMUNIFORM3FEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform3fEXT")) == NULL) || r;
+  r = ((glProgramUniform3fvEXT = (PFNGLPROGRAMUNIFORM3FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform3fvEXT")) == NULL) || r;
+  r = ((glProgramUniform3iEXT = (PFNGLPROGRAMUNIFORM3IEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform3iEXT")) == NULL) || r;
+  r = ((glProgramUniform3ivEXT = (PFNGLPROGRAMUNIFORM3IVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform3ivEXT")) == NULL) || r;
+  r = ((glProgramUniform3uiEXT = (PFNGLPROGRAMUNIFORM3UIEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform3uiEXT")) == NULL) || r;
+  r = ((glProgramUniform3uivEXT = (PFNGLPROGRAMUNIFORM3UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform3uivEXT")) == NULL) || r;
+  r = ((glProgramUniform4fEXT = (PFNGLPROGRAMUNIFORM4FEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform4fEXT")) == NULL) || r;
+  r = ((glProgramUniform4fvEXT = (PFNGLPROGRAMUNIFORM4FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform4fvEXT")) == NULL) || r;
+  r = ((glProgramUniform4iEXT = (PFNGLPROGRAMUNIFORM4IEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform4iEXT")) == NULL) || r;
+  r = ((glProgramUniform4ivEXT = (PFNGLPROGRAMUNIFORM4IVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform4ivEXT")) == NULL) || r;
+  r = ((glProgramUniform4uiEXT = (PFNGLPROGRAMUNIFORM4UIEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform4uiEXT")) == NULL) || r;
+  r = ((glProgramUniform4uivEXT = (PFNGLPROGRAMUNIFORM4UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniform4uivEXT")) == NULL) || r;
+  r = ((glProgramUniformMatrix2fvEXT = (PFNGLPROGRAMUNIFORMMATRIX2FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniformMatrix2fvEXT")) == NULL) || r;
+  r = ((glProgramUniformMatrix2x3fvEXT = (PFNGLPROGRAMUNIFORMMATRIX2X3FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniformMatrix2x3fvEXT")) == NULL) || r;
+  r = ((glProgramUniformMatrix2x4fvEXT = (PFNGLPROGRAMUNIFORMMATRIX2X4FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniformMatrix2x4fvEXT")) == NULL) || r;
+  r = ((glProgramUniformMatrix3fvEXT = (PFNGLPROGRAMUNIFORMMATRIX3FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniformMatrix3fvEXT")) == NULL) || r;
+  r = ((glProgramUniformMatrix3x2fvEXT = (PFNGLPROGRAMUNIFORMMATRIX3X2FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniformMatrix3x2fvEXT")) == NULL) || r;
+  r = ((glProgramUniformMatrix3x4fvEXT = (PFNGLPROGRAMUNIFORMMATRIX3X4FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniformMatrix3x4fvEXT")) == NULL) || r;
+  r = ((glProgramUniformMatrix4fvEXT = (PFNGLPROGRAMUNIFORMMATRIX4FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniformMatrix4fvEXT")) == NULL) || r;
+  r = ((glProgramUniformMatrix4x2fvEXT = (PFNGLPROGRAMUNIFORMMATRIX4X2FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniformMatrix4x2fvEXT")) == NULL) || r;
+  r = ((glProgramUniformMatrix4x3fvEXT = (PFNGLPROGRAMUNIFORMMATRIX4X3FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramUniformMatrix4x3fvEXT")) == NULL) || r;
+  r = ((glPushClientAttribDefaultEXT = (PFNGLPUSHCLIENTATTRIBDEFAULTEXTPROC)glewGetProcAddress((const GLubyte*)"glPushClientAttribDefaultEXT")) == NULL) || r;
+  r = ((glTextureBufferEXT = (PFNGLTEXTUREBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glTextureBufferEXT")) == NULL) || r;
+  r = ((glTextureImage1DEXT = (PFNGLTEXTUREIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureImage1DEXT")) == NULL) || r;
+  r = ((glTextureImage2DEXT = (PFNGLTEXTUREIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureImage2DEXT")) == NULL) || r;
+  r = ((glTextureImage3DEXT = (PFNGLTEXTUREIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureImage3DEXT")) == NULL) || r;
+  r = ((glTextureParameterIivEXT = (PFNGLTEXTUREPARAMETERIIVEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureParameterIivEXT")) == NULL) || r;
+  r = ((glTextureParameterIuivEXT = (PFNGLTEXTUREPARAMETERIUIVEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureParameterIuivEXT")) == NULL) || r;
+  r = ((glTextureParameterfEXT = (PFNGLTEXTUREPARAMETERFEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureParameterfEXT")) == NULL) || r;
+  r = ((glTextureParameterfvEXT = (PFNGLTEXTUREPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureParameterfvEXT")) == NULL) || r;
+  r = ((glTextureParameteriEXT = (PFNGLTEXTUREPARAMETERIEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureParameteriEXT")) == NULL) || r;
+  r = ((glTextureParameterivEXT = (PFNGLTEXTUREPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureParameterivEXT")) == NULL) || r;
+  r = ((glTextureRenderbufferEXT = (PFNGLTEXTURERENDERBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glTextureRenderbufferEXT")) == NULL) || r;
+  r = ((glTextureSubImage1DEXT = (PFNGLTEXTURESUBIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureSubImage1DEXT")) == NULL) || r;
+  r = ((glTextureSubImage2DEXT = (PFNGLTEXTURESUBIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureSubImage2DEXT")) == NULL) || r;
+  r = ((glTextureSubImage3DEXT = (PFNGLTEXTURESUBIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureSubImage3DEXT")) == NULL) || r;
+  r = ((glUnmapNamedBufferEXT = (PFNGLUNMAPNAMEDBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glUnmapNamedBufferEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_direct_state_access */
+
+#ifdef GL_EXT_draw_buffers2
+
+static GLboolean _glewInit_GL_EXT_draw_buffers2 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glColorMaskIndexedEXT = (PFNGLCOLORMASKINDEXEDEXTPROC)glewGetProcAddress((const GLubyte*)"glColorMaskIndexedEXT")) == NULL) || r;
+  r = ((glDisableIndexedEXT = (PFNGLDISABLEINDEXEDEXTPROC)glewGetProcAddress((const GLubyte*)"glDisableIndexedEXT")) == NULL) || r;
+  r = ((glEnableIndexedEXT = (PFNGLENABLEINDEXEDEXTPROC)glewGetProcAddress((const GLubyte*)"glEnableIndexedEXT")) == NULL) || r;
+  r = ((glGetBooleanIndexedvEXT = (PFNGLGETBOOLEANINDEXEDVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetBooleanIndexedvEXT")) == NULL) || r;
+  r = ((glGetIntegerIndexedvEXT = (PFNGLGETINTEGERINDEXEDVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetIntegerIndexedvEXT")) == NULL) || r;
+  r = ((glIsEnabledIndexedEXT = (PFNGLISENABLEDINDEXEDEXTPROC)glewGetProcAddress((const GLubyte*)"glIsEnabledIndexedEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_draw_buffers2 */
+
+#ifdef GL_EXT_draw_instanced
+
+static GLboolean _glewInit_GL_EXT_draw_instanced (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDrawArraysInstancedEXT = (PFNGLDRAWARRAYSINSTANCEDEXTPROC)glewGetProcAddress((const GLubyte*)"glDrawArraysInstancedEXT")) == NULL) || r;
+  r = ((glDrawElementsInstancedEXT = (PFNGLDRAWELEMENTSINSTANCEDEXTPROC)glewGetProcAddress((const GLubyte*)"glDrawElementsInstancedEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_draw_instanced */
+
+#ifdef GL_EXT_draw_range_elements
+
+static GLboolean _glewInit_GL_EXT_draw_range_elements (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDrawRangeElementsEXT = (PFNGLDRAWRANGEELEMENTSEXTPROC)glewGetProcAddress((const GLubyte*)"glDrawRangeElementsEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_draw_range_elements */
+
+#ifdef GL_EXT_fog_coord
+
+static GLboolean _glewInit_GL_EXT_fog_coord (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFogCoordPointerEXT = (PFNGLFOGCOORDPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glFogCoordPointerEXT")) == NULL) || r;
+  r = ((glFogCoorddEXT = (PFNGLFOGCOORDDEXTPROC)glewGetProcAddress((const GLubyte*)"glFogCoorddEXT")) == NULL) || r;
+  r = ((glFogCoorddvEXT = (PFNGLFOGCOORDDVEXTPROC)glewGetProcAddress((const GLubyte*)"glFogCoorddvEXT")) == NULL) || r;
+  r = ((glFogCoordfEXT = (PFNGLFOGCOORDFEXTPROC)glewGetProcAddress((const GLubyte*)"glFogCoordfEXT")) == NULL) || r;
+  r = ((glFogCoordfvEXT = (PFNGLFOGCOORDFVEXTPROC)glewGetProcAddress((const GLubyte*)"glFogCoordfvEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_fog_coord */
+
+#ifdef GL_EXT_fragment_lighting
+
+static GLboolean _glewInit_GL_EXT_fragment_lighting (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFragmentColorMaterialEXT = (PFNGLFRAGMENTCOLORMATERIALEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentColorMaterialEXT")) == NULL) || r;
+  r = ((glFragmentLightModelfEXT = (PFNGLFRAGMENTLIGHTMODELFEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightModelfEXT")) == NULL) || r;
+  r = ((glFragmentLightModelfvEXT = (PFNGLFRAGMENTLIGHTMODELFVEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightModelfvEXT")) == NULL) || r;
+  r = ((glFragmentLightModeliEXT = (PFNGLFRAGMENTLIGHTMODELIEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightModeliEXT")) == NULL) || r;
+  r = ((glFragmentLightModelivEXT = (PFNGLFRAGMENTLIGHTMODELIVEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightModelivEXT")) == NULL) || r;
+  r = ((glFragmentLightfEXT = (PFNGLFRAGMENTLIGHTFEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightfEXT")) == NULL) || r;
+  r = ((glFragmentLightfvEXT = (PFNGLFRAGMENTLIGHTFVEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightfvEXT")) == NULL) || r;
+  r = ((glFragmentLightiEXT = (PFNGLFRAGMENTLIGHTIEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightiEXT")) == NULL) || r;
+  r = ((glFragmentLightivEXT = (PFNGLFRAGMENTLIGHTIVEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightivEXT")) == NULL) || r;
+  r = ((glFragmentMaterialfEXT = (PFNGLFRAGMENTMATERIALFEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentMaterialfEXT")) == NULL) || r;
+  r = ((glFragmentMaterialfvEXT = (PFNGLFRAGMENTMATERIALFVEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentMaterialfvEXT")) == NULL) || r;
+  r = ((glFragmentMaterialiEXT = (PFNGLFRAGMENTMATERIALIEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentMaterialiEXT")) == NULL) || r;
+  r = ((glFragmentMaterialivEXT = (PFNGLFRAGMENTMATERIALIVEXTPROC)glewGetProcAddress((const GLubyte*)"glFragmentMaterialivEXT")) == NULL) || r;
+  r = ((glGetFragmentLightfvEXT = (PFNGLGETFRAGMENTLIGHTFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetFragmentLightfvEXT")) == NULL) || r;
+  r = ((glGetFragmentLightivEXT = (PFNGLGETFRAGMENTLIGHTIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetFragmentLightivEXT")) == NULL) || r;
+  r = ((glGetFragmentMaterialfvEXT = (PFNGLGETFRAGMENTMATERIALFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetFragmentMaterialfvEXT")) == NULL) || r;
+  r = ((glGetFragmentMaterialivEXT = (PFNGLGETFRAGMENTMATERIALIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetFragmentMaterialivEXT")) == NULL) || r;
+  r = ((glLightEnviEXT = (PFNGLLIGHTENVIEXTPROC)glewGetProcAddress((const GLubyte*)"glLightEnviEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_fragment_lighting */
+
+#ifdef GL_EXT_framebuffer_blit
+
+static GLboolean _glewInit_GL_EXT_framebuffer_blit (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBlitFramebufferEXT = (PFNGLBLITFRAMEBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glBlitFramebufferEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_framebuffer_blit */
+
+#ifdef GL_EXT_framebuffer_multisample
+
+static GLboolean _glewInit_GL_EXT_framebuffer_multisample (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glRenderbufferStorageMultisampleEXT = (PFNGLRENDERBUFFERSTORAGEMULTISAMPLEEXTPROC)glewGetProcAddress((const GLubyte*)"glRenderbufferStorageMultisampleEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_framebuffer_multisample */
+
+#ifdef GL_EXT_framebuffer_object
+
+static GLboolean _glewInit_GL_EXT_framebuffer_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBindFramebufferEXT = (PFNGLBINDFRAMEBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glBindFramebufferEXT")) == NULL) || r;
+  r = ((glBindRenderbufferEXT = (PFNGLBINDRENDERBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glBindRenderbufferEXT")) == NULL) || r;
+  r = ((glCheckFramebufferStatusEXT = (PFNGLCHECKFRAMEBUFFERSTATUSEXTPROC)glewGetProcAddress((const GLubyte*)"glCheckFramebufferStatusEXT")) == NULL) || r;
+  r = ((glDeleteFramebuffersEXT = (PFNGLDELETEFRAMEBUFFERSEXTPROC)glewGetProcAddress((const GLubyte*)"glDeleteFramebuffersEXT")) == NULL) || r;
+  r = ((glDeleteRenderbuffersEXT = (PFNGLDELETERENDERBUFFERSEXTPROC)glewGetProcAddress((const GLubyte*)"glDeleteRenderbuffersEXT")) == NULL) || r;
+  r = ((glFramebufferRenderbufferEXT = (PFNGLFRAMEBUFFERRENDERBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glFramebufferRenderbufferEXT")) == NULL) || r;
+  r = ((glFramebufferTexture1DEXT = (PFNGLFRAMEBUFFERTEXTURE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTexture1DEXT")) == NULL) || r;
+  r = ((glFramebufferTexture2DEXT = (PFNGLFRAMEBUFFERTEXTURE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTexture2DEXT")) == NULL) || r;
+  r = ((glFramebufferTexture3DEXT = (PFNGLFRAMEBUFFERTEXTURE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTexture3DEXT")) == NULL) || r;
+  r = ((glGenFramebuffersEXT = (PFNGLGENFRAMEBUFFERSEXTPROC)glewGetProcAddress((const GLubyte*)"glGenFramebuffersEXT")) == NULL) || r;
+  r = ((glGenRenderbuffersEXT = (PFNGLGENRENDERBUFFERSEXTPROC)glewGetProcAddress((const GLubyte*)"glGenRenderbuffersEXT")) == NULL) || r;
+  r = ((glGenerateMipmapEXT = (PFNGLGENERATEMIPMAPEXTPROC)glewGetProcAddress((const GLubyte*)"glGenerateMipmapEXT")) == NULL) || r;
+  r = ((glGetFramebufferAttachmentParameterivEXT = (PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetFramebufferAttachmentParameterivEXT")) == NULL) || r;
+  r = ((glGetRenderbufferParameterivEXT = (PFNGLGETRENDERBUFFERPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetRenderbufferParameterivEXT")) == NULL) || r;
+  r = ((glIsFramebufferEXT = (PFNGLISFRAMEBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glIsFramebufferEXT")) == NULL) || r;
+  r = ((glIsRenderbufferEXT = (PFNGLISRENDERBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glIsRenderbufferEXT")) == NULL) || r;
+  r = ((glRenderbufferStorageEXT = (PFNGLRENDERBUFFERSTORAGEEXTPROC)glewGetProcAddress((const GLubyte*)"glRenderbufferStorageEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_framebuffer_object */
+
+#ifdef GL_EXT_framebuffer_sRGB
+
+#endif /* GL_EXT_framebuffer_sRGB */
+
+#ifdef GL_EXT_geometry_shader4
+
+static GLboolean _glewInit_GL_EXT_geometry_shader4 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFramebufferTextureEXT = (PFNGLFRAMEBUFFERTEXTUREEXTPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTextureEXT")) == NULL) || r;
+  r = ((glFramebufferTextureFaceEXT = (PFNGLFRAMEBUFFERTEXTUREFACEEXTPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTextureFaceEXT")) == NULL) || r;
+  r = ((glFramebufferTextureLayerEXT = (PFNGLFRAMEBUFFERTEXTURELAYEREXTPROC)glewGetProcAddress((const GLubyte*)"glFramebufferTextureLayerEXT")) == NULL) || r;
+  r = ((glProgramParameteriEXT = (PFNGLPROGRAMPARAMETERIEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramParameteriEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_geometry_shader4 */
+
+#ifdef GL_EXT_gpu_program_parameters
+
+static GLboolean _glewInit_GL_EXT_gpu_program_parameters (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glProgramEnvParameters4fvEXT = (PFNGLPROGRAMENVPARAMETERS4FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParameters4fvEXT")) == NULL) || r;
+  r = ((glProgramLocalParameters4fvEXT = (PFNGLPROGRAMLOCALPARAMETERS4FVEXTPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParameters4fvEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_gpu_program_parameters */
+
+#ifdef GL_EXT_gpu_shader4
+
+static GLboolean _glewInit_GL_EXT_gpu_shader4 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBindFragDataLocationEXT = (PFNGLBINDFRAGDATALOCATIONEXTPROC)glewGetProcAddress((const GLubyte*)"glBindFragDataLocationEXT")) == NULL) || r;
+  r = ((glGetFragDataLocationEXT = (PFNGLGETFRAGDATALOCATIONEXTPROC)glewGetProcAddress((const GLubyte*)"glGetFragDataLocationEXT")) == NULL) || r;
+  r = ((glGetUniformuivEXT = (PFNGLGETUNIFORMUIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetUniformuivEXT")) == NULL) || r;
+  r = ((glGetVertexAttribIivEXT = (PFNGLGETVERTEXATTRIBIIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribIivEXT")) == NULL) || r;
+  r = ((glGetVertexAttribIuivEXT = (PFNGLGETVERTEXATTRIBIUIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribIuivEXT")) == NULL) || r;
+  r = ((glUniform1uiEXT = (PFNGLUNIFORM1UIEXTPROC)glewGetProcAddress((const GLubyte*)"glUniform1uiEXT")) == NULL) || r;
+  r = ((glUniform1uivEXT = (PFNGLUNIFORM1UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glUniform1uivEXT")) == NULL) || r;
+  r = ((glUniform2uiEXT = (PFNGLUNIFORM2UIEXTPROC)glewGetProcAddress((const GLubyte*)"glUniform2uiEXT")) == NULL) || r;
+  r = ((glUniform2uivEXT = (PFNGLUNIFORM2UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glUniform2uivEXT")) == NULL) || r;
+  r = ((glUniform3uiEXT = (PFNGLUNIFORM3UIEXTPROC)glewGetProcAddress((const GLubyte*)"glUniform3uiEXT")) == NULL) || r;
+  r = ((glUniform3uivEXT = (PFNGLUNIFORM3UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glUniform3uivEXT")) == NULL) || r;
+  r = ((glUniform4uiEXT = (PFNGLUNIFORM4UIEXTPROC)glewGetProcAddress((const GLubyte*)"glUniform4uiEXT")) == NULL) || r;
+  r = ((glUniform4uivEXT = (PFNGLUNIFORM4UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glUniform4uivEXT")) == NULL) || r;
+  r = ((glVertexAttribI1iEXT = (PFNGLVERTEXATTRIBI1IEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI1iEXT")) == NULL) || r;
+  r = ((glVertexAttribI1ivEXT = (PFNGLVERTEXATTRIBI1IVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI1ivEXT")) == NULL) || r;
+  r = ((glVertexAttribI1uiEXT = (PFNGLVERTEXATTRIBI1UIEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI1uiEXT")) == NULL) || r;
+  r = ((glVertexAttribI1uivEXT = (PFNGLVERTEXATTRIBI1UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI1uivEXT")) == NULL) || r;
+  r = ((glVertexAttribI2iEXT = (PFNGLVERTEXATTRIBI2IEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI2iEXT")) == NULL) || r;
+  r = ((glVertexAttribI2ivEXT = (PFNGLVERTEXATTRIBI2IVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI2ivEXT")) == NULL) || r;
+  r = ((glVertexAttribI2uiEXT = (PFNGLVERTEXATTRIBI2UIEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI2uiEXT")) == NULL) || r;
+  r = ((glVertexAttribI2uivEXT = (PFNGLVERTEXATTRIBI2UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI2uivEXT")) == NULL) || r;
+  r = ((glVertexAttribI3iEXT = (PFNGLVERTEXATTRIBI3IEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI3iEXT")) == NULL) || r;
+  r = ((glVertexAttribI3ivEXT = (PFNGLVERTEXATTRIBI3IVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI3ivEXT")) == NULL) || r;
+  r = ((glVertexAttribI3uiEXT = (PFNGLVERTEXATTRIBI3UIEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI3uiEXT")) == NULL) || r;
+  r = ((glVertexAttribI3uivEXT = (PFNGLVERTEXATTRIBI3UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI3uivEXT")) == NULL) || r;
+  r = ((glVertexAttribI4bvEXT = (PFNGLVERTEXATTRIBI4BVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4bvEXT")) == NULL) || r;
+  r = ((glVertexAttribI4iEXT = (PFNGLVERTEXATTRIBI4IEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4iEXT")) == NULL) || r;
+  r = ((glVertexAttribI4ivEXT = (PFNGLVERTEXATTRIBI4IVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4ivEXT")) == NULL) || r;
+  r = ((glVertexAttribI4svEXT = (PFNGLVERTEXATTRIBI4SVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4svEXT")) == NULL) || r;
+  r = ((glVertexAttribI4ubvEXT = (PFNGLVERTEXATTRIBI4UBVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4ubvEXT")) == NULL) || r;
+  r = ((glVertexAttribI4uiEXT = (PFNGLVERTEXATTRIBI4UIEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4uiEXT")) == NULL) || r;
+  r = ((glVertexAttribI4uivEXT = (PFNGLVERTEXATTRIBI4UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4uivEXT")) == NULL) || r;
+  r = ((glVertexAttribI4usvEXT = (PFNGLVERTEXATTRIBI4USVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribI4usvEXT")) == NULL) || r;
+  r = ((glVertexAttribIPointerEXT = (PFNGLVERTEXATTRIBIPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribIPointerEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_gpu_shader4 */
+
+#ifdef GL_EXT_histogram
+
+static GLboolean _glewInit_GL_EXT_histogram (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetHistogramEXT = (PFNGLGETHISTOGRAMEXTPROC)glewGetProcAddress((const GLubyte*)"glGetHistogramEXT")) == NULL) || r;
+  r = ((glGetHistogramParameterfvEXT = (PFNGLGETHISTOGRAMPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetHistogramParameterfvEXT")) == NULL) || r;
+  r = ((glGetHistogramParameterivEXT = (PFNGLGETHISTOGRAMPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetHistogramParameterivEXT")) == NULL) || r;
+  r = ((glGetMinmaxEXT = (PFNGLGETMINMAXEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMinmaxEXT")) == NULL) || r;
+  r = ((glGetMinmaxParameterfvEXT = (PFNGLGETMINMAXPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMinmaxParameterfvEXT")) == NULL) || r;
+  r = ((glGetMinmaxParameterivEXT = (PFNGLGETMINMAXPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetMinmaxParameterivEXT")) == NULL) || r;
+  r = ((glHistogramEXT = (PFNGLHISTOGRAMEXTPROC)glewGetProcAddress((const GLubyte*)"glHistogramEXT")) == NULL) || r;
+  r = ((glMinmaxEXT = (PFNGLMINMAXEXTPROC)glewGetProcAddress((const GLubyte*)"glMinmaxEXT")) == NULL) || r;
+  r = ((glResetHistogramEXT = (PFNGLRESETHISTOGRAMEXTPROC)glewGetProcAddress((const GLubyte*)"glResetHistogramEXT")) == NULL) || r;
+  r = ((glResetMinmaxEXT = (PFNGLRESETMINMAXEXTPROC)glewGetProcAddress((const GLubyte*)"glResetMinmaxEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_histogram */
+
+#ifdef GL_EXT_index_array_formats
+
+#endif /* GL_EXT_index_array_formats */
+
+#ifdef GL_EXT_index_func
+
+static GLboolean _glewInit_GL_EXT_index_func (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glIndexFuncEXT = (PFNGLINDEXFUNCEXTPROC)glewGetProcAddress((const GLubyte*)"glIndexFuncEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_index_func */
+
+#ifdef GL_EXT_index_material
+
+static GLboolean _glewInit_GL_EXT_index_material (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glIndexMaterialEXT = (PFNGLINDEXMATERIALEXTPROC)glewGetProcAddress((const GLubyte*)"glIndexMaterialEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_index_material */
+
+#ifdef GL_EXT_index_texture
+
+#endif /* GL_EXT_index_texture */
+
+#ifdef GL_EXT_light_texture
+
+static GLboolean _glewInit_GL_EXT_light_texture (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glApplyTextureEXT = (PFNGLAPPLYTEXTUREEXTPROC)glewGetProcAddress((const GLubyte*)"glApplyTextureEXT")) == NULL) || r;
+  r = ((glTextureLightEXT = (PFNGLTEXTURELIGHTEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureLightEXT")) == NULL) || r;
+  r = ((glTextureMaterialEXT = (PFNGLTEXTUREMATERIALEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureMaterialEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_light_texture */
+
+#ifdef GL_EXT_misc_attribute
+
+#endif /* GL_EXT_misc_attribute */
+
+#ifdef GL_EXT_multi_draw_arrays
+
+static GLboolean _glewInit_GL_EXT_multi_draw_arrays (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glMultiDrawArraysEXT = (PFNGLMULTIDRAWARRAYSEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiDrawArraysEXT")) == NULL) || r;
+  r = ((glMultiDrawElementsEXT = (PFNGLMULTIDRAWELEMENTSEXTPROC)glewGetProcAddress((const GLubyte*)"glMultiDrawElementsEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_multi_draw_arrays */
+
+#ifdef GL_EXT_multisample
+
+static GLboolean _glewInit_GL_EXT_multisample (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glSampleMaskEXT = (PFNGLSAMPLEMASKEXTPROC)glewGetProcAddress((const GLubyte*)"glSampleMaskEXT")) == NULL) || r;
+  r = ((glSamplePatternEXT = (PFNGLSAMPLEPATTERNEXTPROC)glewGetProcAddress((const GLubyte*)"glSamplePatternEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_multisample */
+
+#ifdef GL_EXT_packed_depth_stencil
+
+#endif /* GL_EXT_packed_depth_stencil */
+
+#ifdef GL_EXT_packed_float
+
+#endif /* GL_EXT_packed_float */
+
+#ifdef GL_EXT_packed_pixels
+
+#endif /* GL_EXT_packed_pixels */
+
+#ifdef GL_EXT_paletted_texture
+
+static GLboolean _glewInit_GL_EXT_paletted_texture (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glColorTableEXT = (PFNGLCOLORTABLEEXTPROC)glewGetProcAddress((const GLubyte*)"glColorTableEXT")) == NULL) || r;
+  r = ((glGetColorTableEXT = (PFNGLGETCOLORTABLEEXTPROC)glewGetProcAddress((const GLubyte*)"glGetColorTableEXT")) == NULL) || r;
+  r = ((glGetColorTableParameterfvEXT = (PFNGLGETCOLORTABLEPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetColorTableParameterfvEXT")) == NULL) || r;
+  r = ((glGetColorTableParameterivEXT = (PFNGLGETCOLORTABLEPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetColorTableParameterivEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_paletted_texture */
+
+#ifdef GL_EXT_pixel_buffer_object
+
+#endif /* GL_EXT_pixel_buffer_object */
+
+#ifdef GL_EXT_pixel_transform
+
+static GLboolean _glewInit_GL_EXT_pixel_transform (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetPixelTransformParameterfvEXT = (PFNGLGETPIXELTRANSFORMPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetPixelTransformParameterfvEXT")) == NULL) || r;
+  r = ((glGetPixelTransformParameterivEXT = (PFNGLGETPIXELTRANSFORMPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetPixelTransformParameterivEXT")) == NULL) || r;
+  r = ((glPixelTransformParameterfEXT = (PFNGLPIXELTRANSFORMPARAMETERFEXTPROC)glewGetProcAddress((const GLubyte*)"glPixelTransformParameterfEXT")) == NULL) || r;
+  r = ((glPixelTransformParameterfvEXT = (PFNGLPIXELTRANSFORMPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glPixelTransformParameterfvEXT")) == NULL) || r;
+  r = ((glPixelTransformParameteriEXT = (PFNGLPIXELTRANSFORMPARAMETERIEXTPROC)glewGetProcAddress((const GLubyte*)"glPixelTransformParameteriEXT")) == NULL) || r;
+  r = ((glPixelTransformParameterivEXT = (PFNGLPIXELTRANSFORMPARAMETERIVEXTPROC)glewGetProcAddress((const GLubyte*)"glPixelTransformParameterivEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_pixel_transform */
+
+#ifdef GL_EXT_pixel_transform_color_table
+
+#endif /* GL_EXT_pixel_transform_color_table */
+
+#ifdef GL_EXT_point_parameters
+
+static GLboolean _glewInit_GL_EXT_point_parameters (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glPointParameterfEXT = (PFNGLPOINTPARAMETERFEXTPROC)glewGetProcAddress((const GLubyte*)"glPointParameterfEXT")) == NULL) || r;
+  r = ((glPointParameterfvEXT = (PFNGLPOINTPARAMETERFVEXTPROC)glewGetProcAddress((const GLubyte*)"glPointParameterfvEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_point_parameters */
+
+#ifdef GL_EXT_polygon_offset
+
+static GLboolean _glewInit_GL_EXT_polygon_offset (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glPolygonOffsetEXT = (PFNGLPOLYGONOFFSETEXTPROC)glewGetProcAddress((const GLubyte*)"glPolygonOffsetEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_polygon_offset */
+
+#ifdef GL_EXT_rescale_normal
+
+#endif /* GL_EXT_rescale_normal */
+
+#ifdef GL_EXT_scene_marker
+
+static GLboolean _glewInit_GL_EXT_scene_marker (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBeginSceneEXT = (PFNGLBEGINSCENEEXTPROC)glewGetProcAddress((const GLubyte*)"glBeginSceneEXT")) == NULL) || r;
+  r = ((glEndSceneEXT = (PFNGLENDSCENEEXTPROC)glewGetProcAddress((const GLubyte*)"glEndSceneEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_scene_marker */
+
+#ifdef GL_EXT_secondary_color
+
+static GLboolean _glewInit_GL_EXT_secondary_color (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glSecondaryColor3bEXT = (PFNGLSECONDARYCOLOR3BEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3bEXT")) == NULL) || r;
+  r = ((glSecondaryColor3bvEXT = (PFNGLSECONDARYCOLOR3BVEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3bvEXT")) == NULL) || r;
+  r = ((glSecondaryColor3dEXT = (PFNGLSECONDARYCOLOR3DEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3dEXT")) == NULL) || r;
+  r = ((glSecondaryColor3dvEXT = (PFNGLSECONDARYCOLOR3DVEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3dvEXT")) == NULL) || r;
+  r = ((glSecondaryColor3fEXT = (PFNGLSECONDARYCOLOR3FEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3fEXT")) == NULL) || r;
+  r = ((glSecondaryColor3fvEXT = (PFNGLSECONDARYCOLOR3FVEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3fvEXT")) == NULL) || r;
+  r = ((glSecondaryColor3iEXT = (PFNGLSECONDARYCOLOR3IEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3iEXT")) == NULL) || r;
+  r = ((glSecondaryColor3ivEXT = (PFNGLSECONDARYCOLOR3IVEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3ivEXT")) == NULL) || r;
+  r = ((glSecondaryColor3sEXT = (PFNGLSECONDARYCOLOR3SEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3sEXT")) == NULL) || r;
+  r = ((glSecondaryColor3svEXT = (PFNGLSECONDARYCOLOR3SVEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3svEXT")) == NULL) || r;
+  r = ((glSecondaryColor3ubEXT = (PFNGLSECONDARYCOLOR3UBEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3ubEXT")) == NULL) || r;
+  r = ((glSecondaryColor3ubvEXT = (PFNGLSECONDARYCOLOR3UBVEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3ubvEXT")) == NULL) || r;
+  r = ((glSecondaryColor3uiEXT = (PFNGLSECONDARYCOLOR3UIEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3uiEXT")) == NULL) || r;
+  r = ((glSecondaryColor3uivEXT = (PFNGLSECONDARYCOLOR3UIVEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3uivEXT")) == NULL) || r;
+  r = ((glSecondaryColor3usEXT = (PFNGLSECONDARYCOLOR3USEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3usEXT")) == NULL) || r;
+  r = ((glSecondaryColor3usvEXT = (PFNGLSECONDARYCOLOR3USVEXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3usvEXT")) == NULL) || r;
+  r = ((glSecondaryColorPointerEXT = (PFNGLSECONDARYCOLORPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColorPointerEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_secondary_color */
+
+#ifdef GL_EXT_separate_specular_color
+
+#endif /* GL_EXT_separate_specular_color */
+
+#ifdef GL_EXT_shadow_funcs
+
+#endif /* GL_EXT_shadow_funcs */
+
+#ifdef GL_EXT_shared_texture_palette
+
+#endif /* GL_EXT_shared_texture_palette */
+
+#ifdef GL_EXT_stencil_clear_tag
+
+#endif /* GL_EXT_stencil_clear_tag */
+
+#ifdef GL_EXT_stencil_two_side
+
+static GLboolean _glewInit_GL_EXT_stencil_two_side (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glActiveStencilFaceEXT = (PFNGLACTIVESTENCILFACEEXTPROC)glewGetProcAddress((const GLubyte*)"glActiveStencilFaceEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_stencil_two_side */
+
+#ifdef GL_EXT_stencil_wrap
+
+#endif /* GL_EXT_stencil_wrap */
+
+#ifdef GL_EXT_subtexture
+
+static GLboolean _glewInit_GL_EXT_subtexture (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glTexSubImage1DEXT = (PFNGLTEXSUBIMAGE1DEXTPROC)glewGetProcAddress((const GLubyte*)"glTexSubImage1DEXT")) == NULL) || r;
+  r = ((glTexSubImage2DEXT = (PFNGLTEXSUBIMAGE2DEXTPROC)glewGetProcAddress((const GLubyte*)"glTexSubImage2DEXT")) == NULL) || r;
+  r = ((glTexSubImage3DEXT = (PFNGLTEXSUBIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glTexSubImage3DEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_subtexture */
+
+#ifdef GL_EXT_texture
+
+#endif /* GL_EXT_texture */
+
+#ifdef GL_EXT_texture3D
+
+static GLboolean _glewInit_GL_EXT_texture3D (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glTexImage3DEXT = (PFNGLTEXIMAGE3DEXTPROC)glewGetProcAddress((const GLubyte*)"glTexImage3DEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_texture3D */
+
+#ifdef GL_EXT_texture_array
+
+#endif /* GL_EXT_texture_array */
+
+#ifdef GL_EXT_texture_buffer_object
+
+static GLboolean _glewInit_GL_EXT_texture_buffer_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glTexBufferEXT = (PFNGLTEXBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"glTexBufferEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_texture_buffer_object */
+
+#ifdef GL_EXT_texture_compression_dxt1
+
+#endif /* GL_EXT_texture_compression_dxt1 */
+
+#ifdef GL_EXT_texture_compression_latc
+
+#endif /* GL_EXT_texture_compression_latc */
+
+#ifdef GL_EXT_texture_compression_rgtc
+
+#endif /* GL_EXT_texture_compression_rgtc */
+
+#ifdef GL_EXT_texture_compression_s3tc
+
+#endif /* GL_EXT_texture_compression_s3tc */
+
+#ifdef GL_EXT_texture_cube_map
+
+#endif /* GL_EXT_texture_cube_map */
+
+#ifdef GL_EXT_texture_edge_clamp
+
+#endif /* GL_EXT_texture_edge_clamp */
+
+#ifdef GL_EXT_texture_env
+
+#endif /* GL_EXT_texture_env */
+
+#ifdef GL_EXT_texture_env_add
+
+#endif /* GL_EXT_texture_env_add */
+
+#ifdef GL_EXT_texture_env_combine
+
+#endif /* GL_EXT_texture_env_combine */
+
+#ifdef GL_EXT_texture_env_dot3
+
+#endif /* GL_EXT_texture_env_dot3 */
+
+#ifdef GL_EXT_texture_filter_anisotropic
+
+#endif /* GL_EXT_texture_filter_anisotropic */
+
+#ifdef GL_EXT_texture_integer
+
+static GLboolean _glewInit_GL_EXT_texture_integer (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glClearColorIiEXT = (PFNGLCLEARCOLORIIEXTPROC)glewGetProcAddress((const GLubyte*)"glClearColorIiEXT")) == NULL) || r;
+  r = ((glClearColorIuiEXT = (PFNGLCLEARCOLORIUIEXTPROC)glewGetProcAddress((const GLubyte*)"glClearColorIuiEXT")) == NULL) || r;
+  r = ((glGetTexParameterIivEXT = (PFNGLGETTEXPARAMETERIIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetTexParameterIivEXT")) == NULL) || r;
+  r = ((glGetTexParameterIuivEXT = (PFNGLGETTEXPARAMETERIUIVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetTexParameterIuivEXT")) == NULL) || r;
+  r = ((glTexParameterIivEXT = (PFNGLTEXPARAMETERIIVEXTPROC)glewGetProcAddress((const GLubyte*)"glTexParameterIivEXT")) == NULL) || r;
+  r = ((glTexParameterIuivEXT = (PFNGLTEXPARAMETERIUIVEXTPROC)glewGetProcAddress((const GLubyte*)"glTexParameterIuivEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_texture_integer */
+
+#ifdef GL_EXT_texture_lod_bias
+
+#endif /* GL_EXT_texture_lod_bias */
+
+#ifdef GL_EXT_texture_mirror_clamp
+
+#endif /* GL_EXT_texture_mirror_clamp */
+
+#ifdef GL_EXT_texture_object
+
+static GLboolean _glewInit_GL_EXT_texture_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glAreTexturesResidentEXT = (PFNGLARETEXTURESRESIDENTEXTPROC)glewGetProcAddress((const GLubyte*)"glAreTexturesResidentEXT")) == NULL) || r;
+  r = ((glBindTextureEXT = (PFNGLBINDTEXTUREEXTPROC)glewGetProcAddress((const GLubyte*)"glBindTextureEXT")) == NULL) || r;
+  r = ((glDeleteTexturesEXT = (PFNGLDELETETEXTURESEXTPROC)glewGetProcAddress((const GLubyte*)"glDeleteTexturesEXT")) == NULL) || r;
+  r = ((glGenTexturesEXT = (PFNGLGENTEXTURESEXTPROC)glewGetProcAddress((const GLubyte*)"glGenTexturesEXT")) == NULL) || r;
+  r = ((glIsTextureEXT = (PFNGLISTEXTUREEXTPROC)glewGetProcAddress((const GLubyte*)"glIsTextureEXT")) == NULL) || r;
+  r = ((glPrioritizeTexturesEXT = (PFNGLPRIORITIZETEXTURESEXTPROC)glewGetProcAddress((const GLubyte*)"glPrioritizeTexturesEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_texture_object */
+
+#ifdef GL_EXT_texture_perturb_normal
+
+static GLboolean _glewInit_GL_EXT_texture_perturb_normal (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glTextureNormalEXT = (PFNGLTEXTURENORMALEXTPROC)glewGetProcAddress((const GLubyte*)"glTextureNormalEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_texture_perturb_normal */
+
+#ifdef GL_EXT_texture_rectangle
+
+#endif /* GL_EXT_texture_rectangle */
+
+#ifdef GL_EXT_texture_sRGB
+
+#endif /* GL_EXT_texture_sRGB */
+
+#ifdef GL_EXT_texture_shared_exponent
+
+#endif /* GL_EXT_texture_shared_exponent */
+
+#ifdef GL_EXT_texture_swizzle
+
+#endif /* GL_EXT_texture_swizzle */
+
+#ifdef GL_EXT_timer_query
+
+static GLboolean _glewInit_GL_EXT_timer_query (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetQueryObjecti64vEXT = (PFNGLGETQUERYOBJECTI64VEXTPROC)glewGetProcAddress((const GLubyte*)"glGetQueryObjecti64vEXT")) == NULL) || r;
+  r = ((glGetQueryObjectui64vEXT = (PFNGLGETQUERYOBJECTUI64VEXTPROC)glewGetProcAddress((const GLubyte*)"glGetQueryObjectui64vEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_timer_query */
+
+#ifdef GL_EXT_transform_feedback
+
+static GLboolean _glewInit_GL_EXT_transform_feedback (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBeginTransformFeedbackEXT = (PFNGLBEGINTRANSFORMFEEDBACKEXTPROC)glewGetProcAddress((const GLubyte*)"glBeginTransformFeedbackEXT")) == NULL) || r;
+  r = ((glBindBufferBaseEXT = (PFNGLBINDBUFFERBASEEXTPROC)glewGetProcAddress((const GLubyte*)"glBindBufferBaseEXT")) == NULL) || r;
+  r = ((glBindBufferOffsetEXT = (PFNGLBINDBUFFEROFFSETEXTPROC)glewGetProcAddress((const GLubyte*)"glBindBufferOffsetEXT")) == NULL) || r;
+  r = ((glBindBufferRangeEXT = (PFNGLBINDBUFFERRANGEEXTPROC)glewGetProcAddress((const GLubyte*)"glBindBufferRangeEXT")) == NULL) || r;
+  r = ((glEndTransformFeedbackEXT = (PFNGLENDTRANSFORMFEEDBACKEXTPROC)glewGetProcAddress((const GLubyte*)"glEndTransformFeedbackEXT")) == NULL) || r;
+  r = ((glGetTransformFeedbackVaryingEXT = (PFNGLGETTRANSFORMFEEDBACKVARYINGEXTPROC)glewGetProcAddress((const GLubyte*)"glGetTransformFeedbackVaryingEXT")) == NULL) || r;
+  r = ((glTransformFeedbackVaryingsEXT = (PFNGLTRANSFORMFEEDBACKVARYINGSEXTPROC)glewGetProcAddress((const GLubyte*)"glTransformFeedbackVaryingsEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_transform_feedback */
+
+#ifdef GL_EXT_vertex_array
+
+static GLboolean _glewInit_GL_EXT_vertex_array (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glArrayElementEXT = (PFNGLARRAYELEMENTEXTPROC)glewGetProcAddress((const GLubyte*)"glArrayElementEXT")) == NULL) || r;
+  r = ((glColorPointerEXT = (PFNGLCOLORPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glColorPointerEXT")) == NULL) || r;
+  r = ((glDrawArraysEXT = (PFNGLDRAWARRAYSEXTPROC)glewGetProcAddress((const GLubyte*)"glDrawArraysEXT")) == NULL) || r;
+  r = ((glEdgeFlagPointerEXT = (PFNGLEDGEFLAGPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glEdgeFlagPointerEXT")) == NULL) || r;
+  r = ((glGetPointervEXT = (PFNGLGETPOINTERVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetPointervEXT")) == NULL) || r;
+  r = ((glIndexPointerEXT = (PFNGLINDEXPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glIndexPointerEXT")) == NULL) || r;
+  r = ((glNormalPointerEXT = (PFNGLNORMALPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glNormalPointerEXT")) == NULL) || r;
+  r = ((glTexCoordPointerEXT = (PFNGLTEXCOORDPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glTexCoordPointerEXT")) == NULL) || r;
+  r = ((glVertexPointerEXT = (PFNGLVERTEXPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glVertexPointerEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_vertex_array */
+
+#ifdef GL_EXT_vertex_array_bgra
+
+#endif /* GL_EXT_vertex_array_bgra */
+
+#ifdef GL_EXT_vertex_shader
+
+static GLboolean _glewInit_GL_EXT_vertex_shader (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBeginVertexShaderEXT = (PFNGLBEGINVERTEXSHADEREXTPROC)glewGetProcAddress((const GLubyte*)"glBeginVertexShaderEXT")) == NULL) || r;
+  r = ((glBindLightParameterEXT = (PFNGLBINDLIGHTPARAMETEREXTPROC)glewGetProcAddress((const GLubyte*)"glBindLightParameterEXT")) == NULL) || r;
+  r = ((glBindMaterialParameterEXT = (PFNGLBINDMATERIALPARAMETEREXTPROC)glewGetProcAddress((const GLubyte*)"glBindMaterialParameterEXT")) == NULL) || r;
+  r = ((glBindParameterEXT = (PFNGLBINDPARAMETEREXTPROC)glewGetProcAddress((const GLubyte*)"glBindParameterEXT")) == NULL) || r;
+  r = ((glBindTexGenParameterEXT = (PFNGLBINDTEXGENPARAMETEREXTPROC)glewGetProcAddress((const GLubyte*)"glBindTexGenParameterEXT")) == NULL) || r;
+  r = ((glBindTextureUnitParameterEXT = (PFNGLBINDTEXTUREUNITPARAMETEREXTPROC)glewGetProcAddress((const GLubyte*)"glBindTextureUnitParameterEXT")) == NULL) || r;
+  r = ((glBindVertexShaderEXT = (PFNGLBINDVERTEXSHADEREXTPROC)glewGetProcAddress((const GLubyte*)"glBindVertexShaderEXT")) == NULL) || r;
+  r = ((glDeleteVertexShaderEXT = (PFNGLDELETEVERTEXSHADEREXTPROC)glewGetProcAddress((const GLubyte*)"glDeleteVertexShaderEXT")) == NULL) || r;
+  r = ((glDisableVariantClientStateEXT = (PFNGLDISABLEVARIANTCLIENTSTATEEXTPROC)glewGetProcAddress((const GLubyte*)"glDisableVariantClientStateEXT")) == NULL) || r;
+  r = ((glEnableVariantClientStateEXT = (PFNGLENABLEVARIANTCLIENTSTATEEXTPROC)glewGetProcAddress((const GLubyte*)"glEnableVariantClientStateEXT")) == NULL) || r;
+  r = ((glEndVertexShaderEXT = (PFNGLENDVERTEXSHADEREXTPROC)glewGetProcAddress((const GLubyte*)"glEndVertexShaderEXT")) == NULL) || r;
+  r = ((glExtractComponentEXT = (PFNGLEXTRACTCOMPONENTEXTPROC)glewGetProcAddress((const GLubyte*)"glExtractComponentEXT")) == NULL) || r;
+  r = ((glGenSymbolsEXT = (PFNGLGENSYMBOLSEXTPROC)glewGetProcAddress((const GLubyte*)"glGenSymbolsEXT")) == NULL) || r;
+  r = ((glGenVertexShadersEXT = (PFNGLGENVERTEXSHADERSEXTPROC)glewGetProcAddress((const GLubyte*)"glGenVertexShadersEXT")) == NULL) || r;
+  r = ((glGetInvariantBooleanvEXT = (PFNGLGETINVARIANTBOOLEANVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetInvariantBooleanvEXT")) == NULL) || r;
+  r = ((glGetInvariantFloatvEXT = (PFNGLGETINVARIANTFLOATVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetInvariantFloatvEXT")) == NULL) || r;
+  r = ((glGetInvariantIntegervEXT = (PFNGLGETINVARIANTINTEGERVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetInvariantIntegervEXT")) == NULL) || r;
+  r = ((glGetLocalConstantBooleanvEXT = (PFNGLGETLOCALCONSTANTBOOLEANVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetLocalConstantBooleanvEXT")) == NULL) || r;
+  r = ((glGetLocalConstantFloatvEXT = (PFNGLGETLOCALCONSTANTFLOATVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetLocalConstantFloatvEXT")) == NULL) || r;
+  r = ((glGetLocalConstantIntegervEXT = (PFNGLGETLOCALCONSTANTINTEGERVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetLocalConstantIntegervEXT")) == NULL) || r;
+  r = ((glGetVariantBooleanvEXT = (PFNGLGETVARIANTBOOLEANVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetVariantBooleanvEXT")) == NULL) || r;
+  r = ((glGetVariantFloatvEXT = (PFNGLGETVARIANTFLOATVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetVariantFloatvEXT")) == NULL) || r;
+  r = ((glGetVariantIntegervEXT = (PFNGLGETVARIANTINTEGERVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetVariantIntegervEXT")) == NULL) || r;
+  r = ((glGetVariantPointervEXT = (PFNGLGETVARIANTPOINTERVEXTPROC)glewGetProcAddress((const GLubyte*)"glGetVariantPointervEXT")) == NULL) || r;
+  r = ((glInsertComponentEXT = (PFNGLINSERTCOMPONENTEXTPROC)glewGetProcAddress((const GLubyte*)"glInsertComponentEXT")) == NULL) || r;
+  r = ((glIsVariantEnabledEXT = (PFNGLISVARIANTENABLEDEXTPROC)glewGetProcAddress((const GLubyte*)"glIsVariantEnabledEXT")) == NULL) || r;
+  r = ((glSetInvariantEXT = (PFNGLSETINVARIANTEXTPROC)glewGetProcAddress((const GLubyte*)"glSetInvariantEXT")) == NULL) || r;
+  r = ((glSetLocalConstantEXT = (PFNGLSETLOCALCONSTANTEXTPROC)glewGetProcAddress((const GLubyte*)"glSetLocalConstantEXT")) == NULL) || r;
+  r = ((glShaderOp1EXT = (PFNGLSHADEROP1EXTPROC)glewGetProcAddress((const GLubyte*)"glShaderOp1EXT")) == NULL) || r;
+  r = ((glShaderOp2EXT = (PFNGLSHADEROP2EXTPROC)glewGetProcAddress((const GLubyte*)"glShaderOp2EXT")) == NULL) || r;
+  r = ((glShaderOp3EXT = (PFNGLSHADEROP3EXTPROC)glewGetProcAddress((const GLubyte*)"glShaderOp3EXT")) == NULL) || r;
+  r = ((glSwizzleEXT = (PFNGLSWIZZLEEXTPROC)glewGetProcAddress((const GLubyte*)"glSwizzleEXT")) == NULL) || r;
+  r = ((glVariantPointerEXT = (PFNGLVARIANTPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glVariantPointerEXT")) == NULL) || r;
+  r = ((glVariantbvEXT = (PFNGLVARIANTBVEXTPROC)glewGetProcAddress((const GLubyte*)"glVariantbvEXT")) == NULL) || r;
+  r = ((glVariantdvEXT = (PFNGLVARIANTDVEXTPROC)glewGetProcAddress((const GLubyte*)"glVariantdvEXT")) == NULL) || r;
+  r = ((glVariantfvEXT = (PFNGLVARIANTFVEXTPROC)glewGetProcAddress((const GLubyte*)"glVariantfvEXT")) == NULL) || r;
+  r = ((glVariantivEXT = (PFNGLVARIANTIVEXTPROC)glewGetProcAddress((const GLubyte*)"glVariantivEXT")) == NULL) || r;
+  r = ((glVariantsvEXT = (PFNGLVARIANTSVEXTPROC)glewGetProcAddress((const GLubyte*)"glVariantsvEXT")) == NULL) || r;
+  r = ((glVariantubvEXT = (PFNGLVARIANTUBVEXTPROC)glewGetProcAddress((const GLubyte*)"glVariantubvEXT")) == NULL) || r;
+  r = ((glVariantuivEXT = (PFNGLVARIANTUIVEXTPROC)glewGetProcAddress((const GLubyte*)"glVariantuivEXT")) == NULL) || r;
+  r = ((glVariantusvEXT = (PFNGLVARIANTUSVEXTPROC)glewGetProcAddress((const GLubyte*)"glVariantusvEXT")) == NULL) || r;
+  r = ((glWriteMaskEXT = (PFNGLWRITEMASKEXTPROC)glewGetProcAddress((const GLubyte*)"glWriteMaskEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_vertex_shader */
+
+#ifdef GL_EXT_vertex_weighting
+
+static GLboolean _glewInit_GL_EXT_vertex_weighting (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glVertexWeightPointerEXT = (PFNGLVERTEXWEIGHTPOINTEREXTPROC)glewGetProcAddress((const GLubyte*)"glVertexWeightPointerEXT")) == NULL) || r;
+  r = ((glVertexWeightfEXT = (PFNGLVERTEXWEIGHTFEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexWeightfEXT")) == NULL) || r;
+  r = ((glVertexWeightfvEXT = (PFNGLVERTEXWEIGHTFVEXTPROC)glewGetProcAddress((const GLubyte*)"glVertexWeightfvEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_EXT_vertex_weighting */
+
+#ifdef GL_GREMEDY_frame_terminator
+
+static GLboolean _glewInit_GL_GREMEDY_frame_terminator (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFrameTerminatorGREMEDY = (PFNGLFRAMETERMINATORGREMEDYPROC)glewGetProcAddress((const GLubyte*)"glFrameTerminatorGREMEDY")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_GREMEDY_frame_terminator */
+
+#ifdef GL_GREMEDY_string_marker
+
+static GLboolean _glewInit_GL_GREMEDY_string_marker (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glStringMarkerGREMEDY = (PFNGLSTRINGMARKERGREMEDYPROC)glewGetProcAddress((const GLubyte*)"glStringMarkerGREMEDY")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_GREMEDY_string_marker */
+
+#ifdef GL_HP_convolution_border_modes
+
+#endif /* GL_HP_convolution_border_modes */
+
+#ifdef GL_HP_image_transform
+
+static GLboolean _glewInit_GL_HP_image_transform (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetImageTransformParameterfvHP = (PFNGLGETIMAGETRANSFORMPARAMETERFVHPPROC)glewGetProcAddress((const GLubyte*)"glGetImageTransformParameterfvHP")) == NULL) || r;
+  r = ((glGetImageTransformParameterivHP = (PFNGLGETIMAGETRANSFORMPARAMETERIVHPPROC)glewGetProcAddress((const GLubyte*)"glGetImageTransformParameterivHP")) == NULL) || r;
+  r = ((glImageTransformParameterfHP = (PFNGLIMAGETRANSFORMPARAMETERFHPPROC)glewGetProcAddress((const GLubyte*)"glImageTransformParameterfHP")) == NULL) || r;
+  r = ((glImageTransformParameterfvHP = (PFNGLIMAGETRANSFORMPARAMETERFVHPPROC)glewGetProcAddress((const GLubyte*)"glImageTransformParameterfvHP")) == NULL) || r;
+  r = ((glImageTransformParameteriHP = (PFNGLIMAGETRANSFORMPARAMETERIHPPROC)glewGetProcAddress((const GLubyte*)"glImageTransformParameteriHP")) == NULL) || r;
+  r = ((glImageTransformParameterivHP = (PFNGLIMAGETRANSFORMPARAMETERIVHPPROC)glewGetProcAddress((const GLubyte*)"glImageTransformParameterivHP")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_HP_image_transform */
+
+#ifdef GL_HP_occlusion_test
+
+#endif /* GL_HP_occlusion_test */
+
+#ifdef GL_HP_texture_lighting
+
+#endif /* GL_HP_texture_lighting */
+
+#ifdef GL_IBM_cull_vertex
+
+#endif /* GL_IBM_cull_vertex */
+
+#ifdef GL_IBM_multimode_draw_arrays
+
+static GLboolean _glewInit_GL_IBM_multimode_draw_arrays (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glMultiModeDrawArraysIBM = (PFNGLMULTIMODEDRAWARRAYSIBMPROC)glewGetProcAddress((const GLubyte*)"glMultiModeDrawArraysIBM")) == NULL) || r;
+  r = ((glMultiModeDrawElementsIBM = (PFNGLMULTIMODEDRAWELEMENTSIBMPROC)glewGetProcAddress((const GLubyte*)"glMultiModeDrawElementsIBM")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_IBM_multimode_draw_arrays */
+
+#ifdef GL_IBM_rasterpos_clip
+
+#endif /* GL_IBM_rasterpos_clip */
+
+#ifdef GL_IBM_static_data
+
+#endif /* GL_IBM_static_data */
+
+#ifdef GL_IBM_texture_mirrored_repeat
+
+#endif /* GL_IBM_texture_mirrored_repeat */
+
+#ifdef GL_IBM_vertex_array_lists
+
+static GLboolean _glewInit_GL_IBM_vertex_array_lists (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glColorPointerListIBM = (PFNGLCOLORPOINTERLISTIBMPROC)glewGetProcAddress((const GLubyte*)"glColorPointerListIBM")) == NULL) || r;
+  r = ((glEdgeFlagPointerListIBM = (PFNGLEDGEFLAGPOINTERLISTIBMPROC)glewGetProcAddress((const GLubyte*)"glEdgeFlagPointerListIBM")) == NULL) || r;
+  r = ((glFogCoordPointerListIBM = (PFNGLFOGCOORDPOINTERLISTIBMPROC)glewGetProcAddress((const GLubyte*)"glFogCoordPointerListIBM")) == NULL) || r;
+  r = ((glIndexPointerListIBM = (PFNGLINDEXPOINTERLISTIBMPROC)glewGetProcAddress((const GLubyte*)"glIndexPointerListIBM")) == NULL) || r;
+  r = ((glNormalPointerListIBM = (PFNGLNORMALPOINTERLISTIBMPROC)glewGetProcAddress((const GLubyte*)"glNormalPointerListIBM")) == NULL) || r;
+  r = ((glSecondaryColorPointerListIBM = (PFNGLSECONDARYCOLORPOINTERLISTIBMPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColorPointerListIBM")) == NULL) || r;
+  r = ((glTexCoordPointerListIBM = (PFNGLTEXCOORDPOINTERLISTIBMPROC)glewGetProcAddress((const GLubyte*)"glTexCoordPointerListIBM")) == NULL) || r;
+  r = ((glVertexPointerListIBM = (PFNGLVERTEXPOINTERLISTIBMPROC)glewGetProcAddress((const GLubyte*)"glVertexPointerListIBM")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_IBM_vertex_array_lists */
+
+#ifdef GL_INGR_color_clamp
+
+#endif /* GL_INGR_color_clamp */
+
+#ifdef GL_INGR_interlace_read
+
+#endif /* GL_INGR_interlace_read */
+
+#ifdef GL_INTEL_parallel_arrays
+
+static GLboolean _glewInit_GL_INTEL_parallel_arrays (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glColorPointervINTEL = (PFNGLCOLORPOINTERVINTELPROC)glewGetProcAddress((const GLubyte*)"glColorPointervINTEL")) == NULL) || r;
+  r = ((glNormalPointervINTEL = (PFNGLNORMALPOINTERVINTELPROC)glewGetProcAddress((const GLubyte*)"glNormalPointervINTEL")) == NULL) || r;
+  r = ((glTexCoordPointervINTEL = (PFNGLTEXCOORDPOINTERVINTELPROC)glewGetProcAddress((const GLubyte*)"glTexCoordPointervINTEL")) == NULL) || r;
+  r = ((glVertexPointervINTEL = (PFNGLVERTEXPOINTERVINTELPROC)glewGetProcAddress((const GLubyte*)"glVertexPointervINTEL")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_INTEL_parallel_arrays */
+
+#ifdef GL_INTEL_texture_scissor
+
+static GLboolean _glewInit_GL_INTEL_texture_scissor (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glTexScissorFuncINTEL = (PFNGLTEXSCISSORFUNCINTELPROC)glewGetProcAddress((const GLubyte*)"glTexScissorFuncINTEL")) == NULL) || r;
+  r = ((glTexScissorINTEL = (PFNGLTEXSCISSORINTELPROC)glewGetProcAddress((const GLubyte*)"glTexScissorINTEL")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_INTEL_texture_scissor */
+
+#ifdef GL_KTX_buffer_region
+
+static GLboolean _glewInit_GL_KTX_buffer_region (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBufferRegionEnabledEXT = (PFNGLBUFFERREGIONENABLEDEXTPROC)glewGetProcAddress((const GLubyte*)"glBufferRegionEnabledEXT")) == NULL) || r;
+  r = ((glDeleteBufferRegionEXT = (PFNGLDELETEBUFFERREGIONEXTPROC)glewGetProcAddress((const GLubyte*)"glDeleteBufferRegionEXT")) == NULL) || r;
+  r = ((glDrawBufferRegionEXT = (PFNGLDRAWBUFFERREGIONEXTPROC)glewGetProcAddress((const GLubyte*)"glDrawBufferRegionEXT")) == NULL) || r;
+  r = ((glNewBufferRegionEXT = (PFNGLNEWBUFFERREGIONEXTPROC)glewGetProcAddress((const GLubyte*)"glNewBufferRegionEXT")) == NULL) || r;
+  r = ((glReadBufferRegionEXT = (PFNGLREADBUFFERREGIONEXTPROC)glewGetProcAddress((const GLubyte*)"glReadBufferRegionEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_KTX_buffer_region */
+
+#ifdef GL_MESAX_texture_stack
+
+#endif /* GL_MESAX_texture_stack */
+
+#ifdef GL_MESA_pack_invert
+
+#endif /* GL_MESA_pack_invert */
+
+#ifdef GL_MESA_resize_buffers
+
+static GLboolean _glewInit_GL_MESA_resize_buffers (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glResizeBuffersMESA = (PFNGLRESIZEBUFFERSMESAPROC)glewGetProcAddress((const GLubyte*)"glResizeBuffersMESA")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_MESA_resize_buffers */
+
+#ifdef GL_MESA_window_pos
+
+static GLboolean _glewInit_GL_MESA_window_pos (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glWindowPos2dMESA = (PFNGLWINDOWPOS2DMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2dMESA")) == NULL) || r;
+  r = ((glWindowPos2dvMESA = (PFNGLWINDOWPOS2DVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2dvMESA")) == NULL) || r;
+  r = ((glWindowPos2fMESA = (PFNGLWINDOWPOS2FMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2fMESA")) == NULL) || r;
+  r = ((glWindowPos2fvMESA = (PFNGLWINDOWPOS2FVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2fvMESA")) == NULL) || r;
+  r = ((glWindowPos2iMESA = (PFNGLWINDOWPOS2IMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2iMESA")) == NULL) || r;
+  r = ((glWindowPos2ivMESA = (PFNGLWINDOWPOS2IVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2ivMESA")) == NULL) || r;
+  r = ((glWindowPos2sMESA = (PFNGLWINDOWPOS2SMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2sMESA")) == NULL) || r;
+  r = ((glWindowPos2svMESA = (PFNGLWINDOWPOS2SVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos2svMESA")) == NULL) || r;
+  r = ((glWindowPos3dMESA = (PFNGLWINDOWPOS3DMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3dMESA")) == NULL) || r;
+  r = ((glWindowPos3dvMESA = (PFNGLWINDOWPOS3DVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3dvMESA")) == NULL) || r;
+  r = ((glWindowPos3fMESA = (PFNGLWINDOWPOS3FMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3fMESA")) == NULL) || r;
+  r = ((glWindowPos3fvMESA = (PFNGLWINDOWPOS3FVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3fvMESA")) == NULL) || r;
+  r = ((glWindowPos3iMESA = (PFNGLWINDOWPOS3IMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3iMESA")) == NULL) || r;
+  r = ((glWindowPos3ivMESA = (PFNGLWINDOWPOS3IVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3ivMESA")) == NULL) || r;
+  r = ((glWindowPos3sMESA = (PFNGLWINDOWPOS3SMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3sMESA")) == NULL) || r;
+  r = ((glWindowPos3svMESA = (PFNGLWINDOWPOS3SVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos3svMESA")) == NULL) || r;
+  r = ((glWindowPos4dMESA = (PFNGLWINDOWPOS4DMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos4dMESA")) == NULL) || r;
+  r = ((glWindowPos4dvMESA = (PFNGLWINDOWPOS4DVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos4dvMESA")) == NULL) || r;
+  r = ((glWindowPos4fMESA = (PFNGLWINDOWPOS4FMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos4fMESA")) == NULL) || r;
+  r = ((glWindowPos4fvMESA = (PFNGLWINDOWPOS4FVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos4fvMESA")) == NULL) || r;
+  r = ((glWindowPos4iMESA = (PFNGLWINDOWPOS4IMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos4iMESA")) == NULL) || r;
+  r = ((glWindowPos4ivMESA = (PFNGLWINDOWPOS4IVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos4ivMESA")) == NULL) || r;
+  r = ((glWindowPos4sMESA = (PFNGLWINDOWPOS4SMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos4sMESA")) == NULL) || r;
+  r = ((glWindowPos4svMESA = (PFNGLWINDOWPOS4SVMESAPROC)glewGetProcAddress((const GLubyte*)"glWindowPos4svMESA")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_MESA_window_pos */
+
+#ifdef GL_MESA_ycbcr_texture
+
+#endif /* GL_MESA_ycbcr_texture */
+
+#ifdef GL_NV_blend_square
+
+#endif /* GL_NV_blend_square */
+
+#ifdef GL_NV_conditional_render
+
+static GLboolean _glewInit_GL_NV_conditional_render (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBeginConditionalRenderNV = (PFNGLBEGINCONDITIONALRENDERNVPROC)glewGetProcAddress((const GLubyte*)"glBeginConditionalRenderNV")) == NULL) || r;
+  r = ((glEndConditionalRenderNV = (PFNGLENDCONDITIONALRENDERNVPROC)glewGetProcAddress((const GLubyte*)"glEndConditionalRenderNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_conditional_render */
+
+#ifdef GL_NV_copy_depth_to_color
+
+#endif /* GL_NV_copy_depth_to_color */
+
+#ifdef GL_NV_depth_buffer_float
+
+static GLboolean _glewInit_GL_NV_depth_buffer_float (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glClearDepthdNV = (PFNGLCLEARDEPTHDNVPROC)glewGetProcAddress((const GLubyte*)"glClearDepthdNV")) == NULL) || r;
+  r = ((glDepthBoundsdNV = (PFNGLDEPTHBOUNDSDNVPROC)glewGetProcAddress((const GLubyte*)"glDepthBoundsdNV")) == NULL) || r;
+  r = ((glDepthRangedNV = (PFNGLDEPTHRANGEDNVPROC)glewGetProcAddress((const GLubyte*)"glDepthRangedNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_depth_buffer_float */
+
+#ifdef GL_NV_depth_clamp
+
+#endif /* GL_NV_depth_clamp */
+
+#ifdef GL_NV_depth_range_unclamped
+
+#endif /* GL_NV_depth_range_unclamped */
+
+#ifdef GL_NV_evaluators
+
+static GLboolean _glewInit_GL_NV_evaluators (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glEvalMapsNV = (PFNGLEVALMAPSNVPROC)glewGetProcAddress((const GLubyte*)"glEvalMapsNV")) == NULL) || r;
+  r = ((glGetMapAttribParameterfvNV = (PFNGLGETMAPATTRIBPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glGetMapAttribParameterfvNV")) == NULL) || r;
+  r = ((glGetMapAttribParameterivNV = (PFNGLGETMAPATTRIBPARAMETERIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetMapAttribParameterivNV")) == NULL) || r;
+  r = ((glGetMapControlPointsNV = (PFNGLGETMAPCONTROLPOINTSNVPROC)glewGetProcAddress((const GLubyte*)"glGetMapControlPointsNV")) == NULL) || r;
+  r = ((glGetMapParameterfvNV = (PFNGLGETMAPPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glGetMapParameterfvNV")) == NULL) || r;
+  r = ((glGetMapParameterivNV = (PFNGLGETMAPPARAMETERIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetMapParameterivNV")) == NULL) || r;
+  r = ((glMapControlPointsNV = (PFNGLMAPCONTROLPOINTSNVPROC)glewGetProcAddress((const GLubyte*)"glMapControlPointsNV")) == NULL) || r;
+  r = ((glMapParameterfvNV = (PFNGLMAPPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glMapParameterfvNV")) == NULL) || r;
+  r = ((glMapParameterivNV = (PFNGLMAPPARAMETERIVNVPROC)glewGetProcAddress((const GLubyte*)"glMapParameterivNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_evaluators */
+
+#ifdef GL_NV_explicit_multisample
+
+static GLboolean _glewInit_GL_NV_explicit_multisample (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetMultisamplefvNV = (PFNGLGETMULTISAMPLEFVNVPROC)glewGetProcAddress((const GLubyte*)"glGetMultisamplefvNV")) == NULL) || r;
+  r = ((glSampleMaskIndexedNV = (PFNGLSAMPLEMASKINDEXEDNVPROC)glewGetProcAddress((const GLubyte*)"glSampleMaskIndexedNV")) == NULL) || r;
+  r = ((glTexRenderbufferNV = (PFNGLTEXRENDERBUFFERNVPROC)glewGetProcAddress((const GLubyte*)"glTexRenderbufferNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_explicit_multisample */
+
+#ifdef GL_NV_fence
+
+static GLboolean _glewInit_GL_NV_fence (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDeleteFencesNV = (PFNGLDELETEFENCESNVPROC)glewGetProcAddress((const GLubyte*)"glDeleteFencesNV")) == NULL) || r;
+  r = ((glFinishFenceNV = (PFNGLFINISHFENCENVPROC)glewGetProcAddress((const GLubyte*)"glFinishFenceNV")) == NULL) || r;
+  r = ((glGenFencesNV = (PFNGLGENFENCESNVPROC)glewGetProcAddress((const GLubyte*)"glGenFencesNV")) == NULL) || r;
+  r = ((glGetFenceivNV = (PFNGLGETFENCEIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetFenceivNV")) == NULL) || r;
+  r = ((glIsFenceNV = (PFNGLISFENCENVPROC)glewGetProcAddress((const GLubyte*)"glIsFenceNV")) == NULL) || r;
+  r = ((glSetFenceNV = (PFNGLSETFENCENVPROC)glewGetProcAddress((const GLubyte*)"glSetFenceNV")) == NULL) || r;
+  r = ((glTestFenceNV = (PFNGLTESTFENCENVPROC)glewGetProcAddress((const GLubyte*)"glTestFenceNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_fence */
+
+#ifdef GL_NV_float_buffer
+
+#endif /* GL_NV_float_buffer */
+
+#ifdef GL_NV_fog_distance
+
+#endif /* GL_NV_fog_distance */
+
+#ifdef GL_NV_fragment_program
+
+static GLboolean _glewInit_GL_NV_fragment_program (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetProgramNamedParameterdvNV = (PFNGLGETPROGRAMNAMEDPARAMETERDVNVPROC)glewGetProcAddress((const GLubyte*)"glGetProgramNamedParameterdvNV")) == NULL) || r;
+  r = ((glGetProgramNamedParameterfvNV = (PFNGLGETPROGRAMNAMEDPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glGetProgramNamedParameterfvNV")) == NULL) || r;
+  r = ((glProgramNamedParameter4dNV = (PFNGLPROGRAMNAMEDPARAMETER4DNVPROC)glewGetProcAddress((const GLubyte*)"glProgramNamedParameter4dNV")) == NULL) || r;
+  r = ((glProgramNamedParameter4dvNV = (PFNGLPROGRAMNAMEDPARAMETER4DVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramNamedParameter4dvNV")) == NULL) || r;
+  r = ((glProgramNamedParameter4fNV = (PFNGLPROGRAMNAMEDPARAMETER4FNVPROC)glewGetProcAddress((const GLubyte*)"glProgramNamedParameter4fNV")) == NULL) || r;
+  r = ((glProgramNamedParameter4fvNV = (PFNGLPROGRAMNAMEDPARAMETER4FVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramNamedParameter4fvNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_fragment_program */
+
+#ifdef GL_NV_fragment_program2
+
+#endif /* GL_NV_fragment_program2 */
+
+#ifdef GL_NV_fragment_program4
+
+#endif /* GL_NV_fragment_program4 */
+
+#ifdef GL_NV_fragment_program_option
+
+#endif /* GL_NV_fragment_program_option */
+
+#ifdef GL_NV_framebuffer_multisample_coverage
+
+static GLboolean _glewInit_GL_NV_framebuffer_multisample_coverage (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glRenderbufferStorageMultisampleCoverageNV = (PFNGLRENDERBUFFERSTORAGEMULTISAMPLECOVERAGENVPROC)glewGetProcAddress((const GLubyte*)"glRenderbufferStorageMultisampleCoverageNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_framebuffer_multisample_coverage */
+
+#ifdef GL_NV_geometry_program4
+
+static GLboolean _glewInit_GL_NV_geometry_program4 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glProgramVertexLimitNV = (PFNGLPROGRAMVERTEXLIMITNVPROC)glewGetProcAddress((const GLubyte*)"glProgramVertexLimitNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_geometry_program4 */
+
+#ifdef GL_NV_geometry_shader4
+
+#endif /* GL_NV_geometry_shader4 */
+
+#ifdef GL_NV_gpu_program4
+
+static GLboolean _glewInit_GL_NV_gpu_program4 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glProgramEnvParameterI4iNV = (PFNGLPROGRAMENVPARAMETERI4INVPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParameterI4iNV")) == NULL) || r;
+  r = ((glProgramEnvParameterI4ivNV = (PFNGLPROGRAMENVPARAMETERI4IVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParameterI4ivNV")) == NULL) || r;
+  r = ((glProgramEnvParameterI4uiNV = (PFNGLPROGRAMENVPARAMETERI4UINVPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParameterI4uiNV")) == NULL) || r;
+  r = ((glProgramEnvParameterI4uivNV = (PFNGLPROGRAMENVPARAMETERI4UIVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParameterI4uivNV")) == NULL) || r;
+  r = ((glProgramEnvParametersI4ivNV = (PFNGLPROGRAMENVPARAMETERSI4IVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParametersI4ivNV")) == NULL) || r;
+  r = ((glProgramEnvParametersI4uivNV = (PFNGLPROGRAMENVPARAMETERSI4UIVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramEnvParametersI4uivNV")) == NULL) || r;
+  r = ((glProgramLocalParameterI4iNV = (PFNGLPROGRAMLOCALPARAMETERI4INVPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParameterI4iNV")) == NULL) || r;
+  r = ((glProgramLocalParameterI4ivNV = (PFNGLPROGRAMLOCALPARAMETERI4IVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParameterI4ivNV")) == NULL) || r;
+  r = ((glProgramLocalParameterI4uiNV = (PFNGLPROGRAMLOCALPARAMETERI4UINVPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParameterI4uiNV")) == NULL) || r;
+  r = ((glProgramLocalParameterI4uivNV = (PFNGLPROGRAMLOCALPARAMETERI4UIVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParameterI4uivNV")) == NULL) || r;
+  r = ((glProgramLocalParametersI4ivNV = (PFNGLPROGRAMLOCALPARAMETERSI4IVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParametersI4ivNV")) == NULL) || r;
+  r = ((glProgramLocalParametersI4uivNV = (PFNGLPROGRAMLOCALPARAMETERSI4UIVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramLocalParametersI4uivNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_gpu_program4 */
+
+#ifdef GL_NV_half_float
+
+static GLboolean _glewInit_GL_NV_half_float (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glColor3hNV = (PFNGLCOLOR3HNVPROC)glewGetProcAddress((const GLubyte*)"glColor3hNV")) == NULL) || r;
+  r = ((glColor3hvNV = (PFNGLCOLOR3HVNVPROC)glewGetProcAddress((const GLubyte*)"glColor3hvNV")) == NULL) || r;
+  r = ((glColor4hNV = (PFNGLCOLOR4HNVPROC)glewGetProcAddress((const GLubyte*)"glColor4hNV")) == NULL) || r;
+  r = ((glColor4hvNV = (PFNGLCOLOR4HVNVPROC)glewGetProcAddress((const GLubyte*)"glColor4hvNV")) == NULL) || r;
+  r = ((glFogCoordhNV = (PFNGLFOGCOORDHNVPROC)glewGetProcAddress((const GLubyte*)"glFogCoordhNV")) == NULL) || r;
+  r = ((glFogCoordhvNV = (PFNGLFOGCOORDHVNVPROC)glewGetProcAddress((const GLubyte*)"glFogCoordhvNV")) == NULL) || r;
+  r = ((glMultiTexCoord1hNV = (PFNGLMULTITEXCOORD1HNVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1hNV")) == NULL) || r;
+  r = ((glMultiTexCoord1hvNV = (PFNGLMULTITEXCOORD1HVNVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord1hvNV")) == NULL) || r;
+  r = ((glMultiTexCoord2hNV = (PFNGLMULTITEXCOORD2HNVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2hNV")) == NULL) || r;
+  r = ((glMultiTexCoord2hvNV = (PFNGLMULTITEXCOORD2HVNVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord2hvNV")) == NULL) || r;
+  r = ((glMultiTexCoord3hNV = (PFNGLMULTITEXCOORD3HNVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3hNV")) == NULL) || r;
+  r = ((glMultiTexCoord3hvNV = (PFNGLMULTITEXCOORD3HVNVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord3hvNV")) == NULL) || r;
+  r = ((glMultiTexCoord4hNV = (PFNGLMULTITEXCOORD4HNVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4hNV")) == NULL) || r;
+  r = ((glMultiTexCoord4hvNV = (PFNGLMULTITEXCOORD4HVNVPROC)glewGetProcAddress((const GLubyte*)"glMultiTexCoord4hvNV")) == NULL) || r;
+  r = ((glNormal3hNV = (PFNGLNORMAL3HNVPROC)glewGetProcAddress((const GLubyte*)"glNormal3hNV")) == NULL) || r;
+  r = ((glNormal3hvNV = (PFNGLNORMAL3HVNVPROC)glewGetProcAddress((const GLubyte*)"glNormal3hvNV")) == NULL) || r;
+  r = ((glSecondaryColor3hNV = (PFNGLSECONDARYCOLOR3HNVPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3hNV")) == NULL) || r;
+  r = ((glSecondaryColor3hvNV = (PFNGLSECONDARYCOLOR3HVNVPROC)glewGetProcAddress((const GLubyte*)"glSecondaryColor3hvNV")) == NULL) || r;
+  r = ((glTexCoord1hNV = (PFNGLTEXCOORD1HNVPROC)glewGetProcAddress((const GLubyte*)"glTexCoord1hNV")) == NULL) || r;
+  r = ((glTexCoord1hvNV = (PFNGLTEXCOORD1HVNVPROC)glewGetProcAddress((const GLubyte*)"glTexCoord1hvNV")) == NULL) || r;
+  r = ((glTexCoord2hNV = (PFNGLTEXCOORD2HNVPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2hNV")) == NULL) || r;
+  r = ((glTexCoord2hvNV = (PFNGLTEXCOORD2HVNVPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2hvNV")) == NULL) || r;
+  r = ((glTexCoord3hNV = (PFNGLTEXCOORD3HNVPROC)glewGetProcAddress((const GLubyte*)"glTexCoord3hNV")) == NULL) || r;
+  r = ((glTexCoord3hvNV = (PFNGLTEXCOORD3HVNVPROC)glewGetProcAddress((const GLubyte*)"glTexCoord3hvNV")) == NULL) || r;
+  r = ((glTexCoord4hNV = (PFNGLTEXCOORD4HNVPROC)glewGetProcAddress((const GLubyte*)"glTexCoord4hNV")) == NULL) || r;
+  r = ((glTexCoord4hvNV = (PFNGLTEXCOORD4HVNVPROC)glewGetProcAddress((const GLubyte*)"glTexCoord4hvNV")) == NULL) || r;
+  r = ((glVertex2hNV = (PFNGLVERTEX2HNVPROC)glewGetProcAddress((const GLubyte*)"glVertex2hNV")) == NULL) || r;
+  r = ((glVertex2hvNV = (PFNGLVERTEX2HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertex2hvNV")) == NULL) || r;
+  r = ((glVertex3hNV = (PFNGLVERTEX3HNVPROC)glewGetProcAddress((const GLubyte*)"glVertex3hNV")) == NULL) || r;
+  r = ((glVertex3hvNV = (PFNGLVERTEX3HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertex3hvNV")) == NULL) || r;
+  r = ((glVertex4hNV = (PFNGLVERTEX4HNVPROC)glewGetProcAddress((const GLubyte*)"glVertex4hNV")) == NULL) || r;
+  r = ((glVertex4hvNV = (PFNGLVERTEX4HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertex4hvNV")) == NULL) || r;
+  r = ((glVertexAttrib1hNV = (PFNGLVERTEXATTRIB1HNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1hNV")) == NULL) || r;
+  r = ((glVertexAttrib1hvNV = (PFNGLVERTEXATTRIB1HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1hvNV")) == NULL) || r;
+  r = ((glVertexAttrib2hNV = (PFNGLVERTEXATTRIB2HNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2hNV")) == NULL) || r;
+  r = ((glVertexAttrib2hvNV = (PFNGLVERTEXATTRIB2HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2hvNV")) == NULL) || r;
+  r = ((glVertexAttrib3hNV = (PFNGLVERTEXATTRIB3HNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3hNV")) == NULL) || r;
+  r = ((glVertexAttrib3hvNV = (PFNGLVERTEXATTRIB3HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3hvNV")) == NULL) || r;
+  r = ((glVertexAttrib4hNV = (PFNGLVERTEXATTRIB4HNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4hNV")) == NULL) || r;
+  r = ((glVertexAttrib4hvNV = (PFNGLVERTEXATTRIB4HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4hvNV")) == NULL) || r;
+  r = ((glVertexAttribs1hvNV = (PFNGLVERTEXATTRIBS1HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs1hvNV")) == NULL) || r;
+  r = ((glVertexAttribs2hvNV = (PFNGLVERTEXATTRIBS2HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs2hvNV")) == NULL) || r;
+  r = ((glVertexAttribs3hvNV = (PFNGLVERTEXATTRIBS3HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs3hvNV")) == NULL) || r;
+  r = ((glVertexAttribs4hvNV = (PFNGLVERTEXATTRIBS4HVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs4hvNV")) == NULL) || r;
+  r = ((glVertexWeighthNV = (PFNGLVERTEXWEIGHTHNVPROC)glewGetProcAddress((const GLubyte*)"glVertexWeighthNV")) == NULL) || r;
+  r = ((glVertexWeighthvNV = (PFNGLVERTEXWEIGHTHVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexWeighthvNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_half_float */
+
+#ifdef GL_NV_light_max_exponent
+
+#endif /* GL_NV_light_max_exponent */
+
+#ifdef GL_NV_multisample_filter_hint
+
+#endif /* GL_NV_multisample_filter_hint */
+
+#ifdef GL_NV_occlusion_query
+
+static GLboolean _glewInit_GL_NV_occlusion_query (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glBeginOcclusionQueryNV = (PFNGLBEGINOCCLUSIONQUERYNVPROC)glewGetProcAddress((const GLubyte*)"glBeginOcclusionQueryNV")) == NULL) || r;
+  r = ((glDeleteOcclusionQueriesNV = (PFNGLDELETEOCCLUSIONQUERIESNVPROC)glewGetProcAddress((const GLubyte*)"glDeleteOcclusionQueriesNV")) == NULL) || r;
+  r = ((glEndOcclusionQueryNV = (PFNGLENDOCCLUSIONQUERYNVPROC)glewGetProcAddress((const GLubyte*)"glEndOcclusionQueryNV")) == NULL) || r;
+  r = ((glGenOcclusionQueriesNV = (PFNGLGENOCCLUSIONQUERIESNVPROC)glewGetProcAddress((const GLubyte*)"glGenOcclusionQueriesNV")) == NULL) || r;
+  r = ((glGetOcclusionQueryivNV = (PFNGLGETOCCLUSIONQUERYIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetOcclusionQueryivNV")) == NULL) || r;
+  r = ((glGetOcclusionQueryuivNV = (PFNGLGETOCCLUSIONQUERYUIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetOcclusionQueryuivNV")) == NULL) || r;
+  r = ((glIsOcclusionQueryNV = (PFNGLISOCCLUSIONQUERYNVPROC)glewGetProcAddress((const GLubyte*)"glIsOcclusionQueryNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_occlusion_query */
+
+#ifdef GL_NV_packed_depth_stencil
+
+#endif /* GL_NV_packed_depth_stencil */
+
+#ifdef GL_NV_parameter_buffer_object
+
+static GLboolean _glewInit_GL_NV_parameter_buffer_object (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glProgramBufferParametersIivNV = (PFNGLPROGRAMBUFFERPARAMETERSIIVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramBufferParametersIivNV")) == NULL) || r;
+  r = ((glProgramBufferParametersIuivNV = (PFNGLPROGRAMBUFFERPARAMETERSIUIVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramBufferParametersIuivNV")) == NULL) || r;
+  r = ((glProgramBufferParametersfvNV = (PFNGLPROGRAMBUFFERPARAMETERSFVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramBufferParametersfvNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_parameter_buffer_object */
+
+#ifdef GL_NV_pixel_data_range
+
+static GLboolean _glewInit_GL_NV_pixel_data_range (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFlushPixelDataRangeNV = (PFNGLFLUSHPIXELDATARANGENVPROC)glewGetProcAddress((const GLubyte*)"glFlushPixelDataRangeNV")) == NULL) || r;
+  r = ((glPixelDataRangeNV = (PFNGLPIXELDATARANGENVPROC)glewGetProcAddress((const GLubyte*)"glPixelDataRangeNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_pixel_data_range */
+
+#ifdef GL_NV_point_sprite
+
+static GLboolean _glewInit_GL_NV_point_sprite (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glPointParameteriNV = (PFNGLPOINTPARAMETERINVPROC)glewGetProcAddress((const GLubyte*)"glPointParameteriNV")) == NULL) || r;
+  r = ((glPointParameterivNV = (PFNGLPOINTPARAMETERIVNVPROC)glewGetProcAddress((const GLubyte*)"glPointParameterivNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_point_sprite */
+
+#ifdef GL_NV_present_video
+
+static GLboolean _glewInit_GL_NV_present_video (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetVideoi64vNV = (PFNGLGETVIDEOI64VNVPROC)glewGetProcAddress((const GLubyte*)"glGetVideoi64vNV")) == NULL) || r;
+  r = ((glGetVideoivNV = (PFNGLGETVIDEOIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetVideoivNV")) == NULL) || r;
+  r = ((glGetVideoui64vNV = (PFNGLGETVIDEOUI64VNVPROC)glewGetProcAddress((const GLubyte*)"glGetVideoui64vNV")) == NULL) || r;
+  r = ((glGetVideouivNV = (PFNGLGETVIDEOUIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetVideouivNV")) == NULL) || r;
+  r = ((glPresentFrameDualFillNV = (PFNGLPRESENTFRAMEDUALFILLNVPROC)glewGetProcAddress((const GLubyte*)"glPresentFrameDualFillNV")) == NULL) || r;
+  r = ((glPresentFrameKeyedNV = (PFNGLPRESENTFRAMEKEYEDNVPROC)glewGetProcAddress((const GLubyte*)"glPresentFrameKeyedNV")) == NULL) || r;
+  r = ((glVideoParameterivNV = (PFNGLVIDEOPARAMETERIVNVPROC)glewGetProcAddress((const GLubyte*)"glVideoParameterivNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_present_video */
+
+#ifdef GL_NV_primitive_restart
+
+static GLboolean _glewInit_GL_NV_primitive_restart (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glPrimitiveRestartIndexNV = (PFNGLPRIMITIVERESTARTINDEXNVPROC)glewGetProcAddress((const GLubyte*)"glPrimitiveRestartIndexNV")) == NULL) || r;
+  r = ((glPrimitiveRestartNV = (PFNGLPRIMITIVERESTARTNVPROC)glewGetProcAddress((const GLubyte*)"glPrimitiveRestartNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_primitive_restart */
+
+#ifdef GL_NV_register_combiners
+
+static GLboolean _glewInit_GL_NV_register_combiners (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glCombinerInputNV = (PFNGLCOMBINERINPUTNVPROC)glewGetProcAddress((const GLubyte*)"glCombinerInputNV")) == NULL) || r;
+  r = ((glCombinerOutputNV = (PFNGLCOMBINEROUTPUTNVPROC)glewGetProcAddress((const GLubyte*)"glCombinerOutputNV")) == NULL) || r;
+  r = ((glCombinerParameterfNV = (PFNGLCOMBINERPARAMETERFNVPROC)glewGetProcAddress((const GLubyte*)"glCombinerParameterfNV")) == NULL) || r;
+  r = ((glCombinerParameterfvNV = (PFNGLCOMBINERPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glCombinerParameterfvNV")) == NULL) || r;
+  r = ((glCombinerParameteriNV = (PFNGLCOMBINERPARAMETERINVPROC)glewGetProcAddress((const GLubyte*)"glCombinerParameteriNV")) == NULL) || r;
+  r = ((glCombinerParameterivNV = (PFNGLCOMBINERPARAMETERIVNVPROC)glewGetProcAddress((const GLubyte*)"glCombinerParameterivNV")) == NULL) || r;
+  r = ((glFinalCombinerInputNV = (PFNGLFINALCOMBINERINPUTNVPROC)glewGetProcAddress((const GLubyte*)"glFinalCombinerInputNV")) == NULL) || r;
+  r = ((glGetCombinerInputParameterfvNV = (PFNGLGETCOMBINERINPUTPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glGetCombinerInputParameterfvNV")) == NULL) || r;
+  r = ((glGetCombinerInputParameterivNV = (PFNGLGETCOMBINERINPUTPARAMETERIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetCombinerInputParameterivNV")) == NULL) || r;
+  r = ((glGetCombinerOutputParameterfvNV = (PFNGLGETCOMBINEROUTPUTPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glGetCombinerOutputParameterfvNV")) == NULL) || r;
+  r = ((glGetCombinerOutputParameterivNV = (PFNGLGETCOMBINEROUTPUTPARAMETERIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetCombinerOutputParameterivNV")) == NULL) || r;
+  r = ((glGetFinalCombinerInputParameterfvNV = (PFNGLGETFINALCOMBINERINPUTPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glGetFinalCombinerInputParameterfvNV")) == NULL) || r;
+  r = ((glGetFinalCombinerInputParameterivNV = (PFNGLGETFINALCOMBINERINPUTPARAMETERIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetFinalCombinerInputParameterivNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_register_combiners */
+
+#ifdef GL_NV_register_combiners2
+
+static GLboolean _glewInit_GL_NV_register_combiners2 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glCombinerStageParameterfvNV = (PFNGLCOMBINERSTAGEPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glCombinerStageParameterfvNV")) == NULL) || r;
+  r = ((glGetCombinerStageParameterfvNV = (PFNGLGETCOMBINERSTAGEPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glGetCombinerStageParameterfvNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_register_combiners2 */
+
+#ifdef GL_NV_texgen_emboss
+
+#endif /* GL_NV_texgen_emboss */
+
+#ifdef GL_NV_texgen_reflection
+
+#endif /* GL_NV_texgen_reflection */
+
+#ifdef GL_NV_texture_compression_vtc
+
+#endif /* GL_NV_texture_compression_vtc */
+
+#ifdef GL_NV_texture_env_combine4
+
+#endif /* GL_NV_texture_env_combine4 */
+
+#ifdef GL_NV_texture_expand_normal
+
+#endif /* GL_NV_texture_expand_normal */
+
+#ifdef GL_NV_texture_rectangle
+
+#endif /* GL_NV_texture_rectangle */
+
+#ifdef GL_NV_texture_shader
+
+#endif /* GL_NV_texture_shader */
+
+#ifdef GL_NV_texture_shader2
+
+#endif /* GL_NV_texture_shader2 */
+
+#ifdef GL_NV_texture_shader3
+
+#endif /* GL_NV_texture_shader3 */
+
+#ifdef GL_NV_transform_feedback
+
+static GLboolean _glewInit_GL_NV_transform_feedback (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glActiveVaryingNV = (PFNGLACTIVEVARYINGNVPROC)glewGetProcAddress((const GLubyte*)"glActiveVaryingNV")) == NULL) || r;
+  r = ((glBeginTransformFeedbackNV = (PFNGLBEGINTRANSFORMFEEDBACKNVPROC)glewGetProcAddress((const GLubyte*)"glBeginTransformFeedbackNV")) == NULL) || r;
+  r = ((glBindBufferBaseNV = (PFNGLBINDBUFFERBASENVPROC)glewGetProcAddress((const GLubyte*)"glBindBufferBaseNV")) == NULL) || r;
+  r = ((glBindBufferOffsetNV = (PFNGLBINDBUFFEROFFSETNVPROC)glewGetProcAddress((const GLubyte*)"glBindBufferOffsetNV")) == NULL) || r;
+  r = ((glBindBufferRangeNV = (PFNGLBINDBUFFERRANGENVPROC)glewGetProcAddress((const GLubyte*)"glBindBufferRangeNV")) == NULL) || r;
+  r = ((glEndTransformFeedbackNV = (PFNGLENDTRANSFORMFEEDBACKNVPROC)glewGetProcAddress((const GLubyte*)"glEndTransformFeedbackNV")) == NULL) || r;
+  r = ((glGetActiveVaryingNV = (PFNGLGETACTIVEVARYINGNVPROC)glewGetProcAddress((const GLubyte*)"glGetActiveVaryingNV")) == NULL) || r;
+  r = ((glGetTransformFeedbackVaryingNV = (PFNGLGETTRANSFORMFEEDBACKVARYINGNVPROC)glewGetProcAddress((const GLubyte*)"glGetTransformFeedbackVaryingNV")) == NULL) || r;
+  r = ((glGetVaryingLocationNV = (PFNGLGETVARYINGLOCATIONNVPROC)glewGetProcAddress((const GLubyte*)"glGetVaryingLocationNV")) == NULL) || r;
+  r = ((glTransformFeedbackAttribsNV = (PFNGLTRANSFORMFEEDBACKATTRIBSNVPROC)glewGetProcAddress((const GLubyte*)"glTransformFeedbackAttribsNV")) == NULL) || r;
+  r = ((glTransformFeedbackVaryingsNV = (PFNGLTRANSFORMFEEDBACKVARYINGSNVPROC)glewGetProcAddress((const GLubyte*)"glTransformFeedbackVaryingsNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_transform_feedback */
+
+#ifdef GL_NV_vertex_array_range
+
+static GLboolean _glewInit_GL_NV_vertex_array_range (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFlushVertexArrayRangeNV = (PFNGLFLUSHVERTEXARRAYRANGENVPROC)glewGetProcAddress((const GLubyte*)"glFlushVertexArrayRangeNV")) == NULL) || r;
+  r = ((glVertexArrayRangeNV = (PFNGLVERTEXARRAYRANGENVPROC)glewGetProcAddress((const GLubyte*)"glVertexArrayRangeNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_vertex_array_range */
+
+#ifdef GL_NV_vertex_array_range2
+
+#endif /* GL_NV_vertex_array_range2 */
+
+#ifdef GL_NV_vertex_program
+
+static GLboolean _glewInit_GL_NV_vertex_program (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glAreProgramsResidentNV = (PFNGLAREPROGRAMSRESIDENTNVPROC)glewGetProcAddress((const GLubyte*)"glAreProgramsResidentNV")) == NULL) || r;
+  r = ((glBindProgramNV = (PFNGLBINDPROGRAMNVPROC)glewGetProcAddress((const GLubyte*)"glBindProgramNV")) == NULL) || r;
+  r = ((glDeleteProgramsNV = (PFNGLDELETEPROGRAMSNVPROC)glewGetProcAddress((const GLubyte*)"glDeleteProgramsNV")) == NULL) || r;
+  r = ((glExecuteProgramNV = (PFNGLEXECUTEPROGRAMNVPROC)glewGetProcAddress((const GLubyte*)"glExecuteProgramNV")) == NULL) || r;
+  r = ((glGenProgramsNV = (PFNGLGENPROGRAMSNVPROC)glewGetProcAddress((const GLubyte*)"glGenProgramsNV")) == NULL) || r;
+  r = ((glGetProgramParameterdvNV = (PFNGLGETPROGRAMPARAMETERDVNVPROC)glewGetProcAddress((const GLubyte*)"glGetProgramParameterdvNV")) == NULL) || r;
+  r = ((glGetProgramParameterfvNV = (PFNGLGETPROGRAMPARAMETERFVNVPROC)glewGetProcAddress((const GLubyte*)"glGetProgramParameterfvNV")) == NULL) || r;
+  r = ((glGetProgramStringNV = (PFNGLGETPROGRAMSTRINGNVPROC)glewGetProcAddress((const GLubyte*)"glGetProgramStringNV")) == NULL) || r;
+  r = ((glGetProgramivNV = (PFNGLGETPROGRAMIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetProgramivNV")) == NULL) || r;
+  r = ((glGetTrackMatrixivNV = (PFNGLGETTRACKMATRIXIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetTrackMatrixivNV")) == NULL) || r;
+  r = ((glGetVertexAttribPointervNV = (PFNGLGETVERTEXATTRIBPOINTERVNVPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribPointervNV")) == NULL) || r;
+  r = ((glGetVertexAttribdvNV = (PFNGLGETVERTEXATTRIBDVNVPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribdvNV")) == NULL) || r;
+  r = ((glGetVertexAttribfvNV = (PFNGLGETVERTEXATTRIBFVNVPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribfvNV")) == NULL) || r;
+  r = ((glGetVertexAttribivNV = (PFNGLGETVERTEXATTRIBIVNVPROC)glewGetProcAddress((const GLubyte*)"glGetVertexAttribivNV")) == NULL) || r;
+  r = ((glIsProgramNV = (PFNGLISPROGRAMNVPROC)glewGetProcAddress((const GLubyte*)"glIsProgramNV")) == NULL) || r;
+  r = ((glLoadProgramNV = (PFNGLLOADPROGRAMNVPROC)glewGetProcAddress((const GLubyte*)"glLoadProgramNV")) == NULL) || r;
+  r = ((glProgramParameter4dNV = (PFNGLPROGRAMPARAMETER4DNVPROC)glewGetProcAddress((const GLubyte*)"glProgramParameter4dNV")) == NULL) || r;
+  r = ((glProgramParameter4dvNV = (PFNGLPROGRAMPARAMETER4DVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramParameter4dvNV")) == NULL) || r;
+  r = ((glProgramParameter4fNV = (PFNGLPROGRAMPARAMETER4FNVPROC)glewGetProcAddress((const GLubyte*)"glProgramParameter4fNV")) == NULL) || r;
+  r = ((glProgramParameter4fvNV = (PFNGLPROGRAMPARAMETER4FVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramParameter4fvNV")) == NULL) || r;
+  r = ((glProgramParameters4dvNV = (PFNGLPROGRAMPARAMETERS4DVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramParameters4dvNV")) == NULL) || r;
+  r = ((glProgramParameters4fvNV = (PFNGLPROGRAMPARAMETERS4FVNVPROC)glewGetProcAddress((const GLubyte*)"glProgramParameters4fvNV")) == NULL) || r;
+  r = ((glRequestResidentProgramsNV = (PFNGLREQUESTRESIDENTPROGRAMSNVPROC)glewGetProcAddress((const GLubyte*)"glRequestResidentProgramsNV")) == NULL) || r;
+  r = ((glTrackMatrixNV = (PFNGLTRACKMATRIXNVPROC)glewGetProcAddress((const GLubyte*)"glTrackMatrixNV")) == NULL) || r;
+  r = ((glVertexAttrib1dNV = (PFNGLVERTEXATTRIB1DNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1dNV")) == NULL) || r;
+  r = ((glVertexAttrib1dvNV = (PFNGLVERTEXATTRIB1DVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1dvNV")) == NULL) || r;
+  r = ((glVertexAttrib1fNV = (PFNGLVERTEXATTRIB1FNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1fNV")) == NULL) || r;
+  r = ((glVertexAttrib1fvNV = (PFNGLVERTEXATTRIB1FVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1fvNV")) == NULL) || r;
+  r = ((glVertexAttrib1sNV = (PFNGLVERTEXATTRIB1SNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1sNV")) == NULL) || r;
+  r = ((glVertexAttrib1svNV = (PFNGLVERTEXATTRIB1SVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib1svNV")) == NULL) || r;
+  r = ((glVertexAttrib2dNV = (PFNGLVERTEXATTRIB2DNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2dNV")) == NULL) || r;
+  r = ((glVertexAttrib2dvNV = (PFNGLVERTEXATTRIB2DVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2dvNV")) == NULL) || r;
+  r = ((glVertexAttrib2fNV = (PFNGLVERTEXATTRIB2FNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2fNV")) == NULL) || r;
+  r = ((glVertexAttrib2fvNV = (PFNGLVERTEXATTRIB2FVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2fvNV")) == NULL) || r;
+  r = ((glVertexAttrib2sNV = (PFNGLVERTEXATTRIB2SNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2sNV")) == NULL) || r;
+  r = ((glVertexAttrib2svNV = (PFNGLVERTEXATTRIB2SVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib2svNV")) == NULL) || r;
+  r = ((glVertexAttrib3dNV = (PFNGLVERTEXATTRIB3DNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3dNV")) == NULL) || r;
+  r = ((glVertexAttrib3dvNV = (PFNGLVERTEXATTRIB3DVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3dvNV")) == NULL) || r;
+  r = ((glVertexAttrib3fNV = (PFNGLVERTEXATTRIB3FNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3fNV")) == NULL) || r;
+  r = ((glVertexAttrib3fvNV = (PFNGLVERTEXATTRIB3FVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3fvNV")) == NULL) || r;
+  r = ((glVertexAttrib3sNV = (PFNGLVERTEXATTRIB3SNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3sNV")) == NULL) || r;
+  r = ((glVertexAttrib3svNV = (PFNGLVERTEXATTRIB3SVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib3svNV")) == NULL) || r;
+  r = ((glVertexAttrib4dNV = (PFNGLVERTEXATTRIB4DNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4dNV")) == NULL) || r;
+  r = ((glVertexAttrib4dvNV = (PFNGLVERTEXATTRIB4DVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4dvNV")) == NULL) || r;
+  r = ((glVertexAttrib4fNV = (PFNGLVERTEXATTRIB4FNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4fNV")) == NULL) || r;
+  r = ((glVertexAttrib4fvNV = (PFNGLVERTEXATTRIB4FVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4fvNV")) == NULL) || r;
+  r = ((glVertexAttrib4sNV = (PFNGLVERTEXATTRIB4SNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4sNV")) == NULL) || r;
+  r = ((glVertexAttrib4svNV = (PFNGLVERTEXATTRIB4SVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4svNV")) == NULL) || r;
+  r = ((glVertexAttrib4ubNV = (PFNGLVERTEXATTRIB4UBNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4ubNV")) == NULL) || r;
+  r = ((glVertexAttrib4ubvNV = (PFNGLVERTEXATTRIB4UBVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttrib4ubvNV")) == NULL) || r;
+  r = ((glVertexAttribPointerNV = (PFNGLVERTEXATTRIBPOINTERNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribPointerNV")) == NULL) || r;
+  r = ((glVertexAttribs1dvNV = (PFNGLVERTEXATTRIBS1DVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs1dvNV")) == NULL) || r;
+  r = ((glVertexAttribs1fvNV = (PFNGLVERTEXATTRIBS1FVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs1fvNV")) == NULL) || r;
+  r = ((glVertexAttribs1svNV = (PFNGLVERTEXATTRIBS1SVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs1svNV")) == NULL) || r;
+  r = ((glVertexAttribs2dvNV = (PFNGLVERTEXATTRIBS2DVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs2dvNV")) == NULL) || r;
+  r = ((glVertexAttribs2fvNV = (PFNGLVERTEXATTRIBS2FVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs2fvNV")) == NULL) || r;
+  r = ((glVertexAttribs2svNV = (PFNGLVERTEXATTRIBS2SVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs2svNV")) == NULL) || r;
+  r = ((glVertexAttribs3dvNV = (PFNGLVERTEXATTRIBS3DVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs3dvNV")) == NULL) || r;
+  r = ((glVertexAttribs3fvNV = (PFNGLVERTEXATTRIBS3FVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs3fvNV")) == NULL) || r;
+  r = ((glVertexAttribs3svNV = (PFNGLVERTEXATTRIBS3SVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs3svNV")) == NULL) || r;
+  r = ((glVertexAttribs4dvNV = (PFNGLVERTEXATTRIBS4DVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs4dvNV")) == NULL) || r;
+  r = ((glVertexAttribs4fvNV = (PFNGLVERTEXATTRIBS4FVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs4fvNV")) == NULL) || r;
+  r = ((glVertexAttribs4svNV = (PFNGLVERTEXATTRIBS4SVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs4svNV")) == NULL) || r;
+  r = ((glVertexAttribs4ubvNV = (PFNGLVERTEXATTRIBS4UBVNVPROC)glewGetProcAddress((const GLubyte*)"glVertexAttribs4ubvNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_NV_vertex_program */
+
+#ifdef GL_NV_vertex_program1_1
+
+#endif /* GL_NV_vertex_program1_1 */
+
+#ifdef GL_NV_vertex_program2
+
+#endif /* GL_NV_vertex_program2 */
+
+#ifdef GL_NV_vertex_program2_option
+
+#endif /* GL_NV_vertex_program2_option */
+
+#ifdef GL_NV_vertex_program3
+
+#endif /* GL_NV_vertex_program3 */
+
+#ifdef GL_NV_vertex_program4
+
+#endif /* GL_NV_vertex_program4 */
+
+#ifdef GL_OES_byte_coordinates
+
+#endif /* GL_OES_byte_coordinates */
+
+#ifdef GL_OES_compressed_paletted_texture
+
+#endif /* GL_OES_compressed_paletted_texture */
+
+#ifdef GL_OES_read_format
+
+#endif /* GL_OES_read_format */
+
+#ifdef GL_OES_single_precision
+
+static GLboolean _glewInit_GL_OES_single_precision (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glClearDepthfOES = (PFNGLCLEARDEPTHFOESPROC)glewGetProcAddress((const GLubyte*)"glClearDepthfOES")) == NULL) || r;
+  r = ((glClipPlanefOES = (PFNGLCLIPPLANEFOESPROC)glewGetProcAddress((const GLubyte*)"glClipPlanefOES")) == NULL) || r;
+  r = ((glDepthRangefOES = (PFNGLDEPTHRANGEFOESPROC)glewGetProcAddress((const GLubyte*)"glDepthRangefOES")) == NULL) || r;
+  r = ((glFrustumfOES = (PFNGLFRUSTUMFOESPROC)glewGetProcAddress((const GLubyte*)"glFrustumfOES")) == NULL) || r;
+  r = ((glGetClipPlanefOES = (PFNGLGETCLIPPLANEFOESPROC)glewGetProcAddress((const GLubyte*)"glGetClipPlanefOES")) == NULL) || r;
+  r = ((glOrthofOES = (PFNGLORTHOFOESPROC)glewGetProcAddress((const GLubyte*)"glOrthofOES")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_OES_single_precision */
+
+#ifdef GL_OML_interlace
+
+#endif /* GL_OML_interlace */
+
+#ifdef GL_OML_resample
+
+#endif /* GL_OML_resample */
+
+#ifdef GL_OML_subsample
+
+#endif /* GL_OML_subsample */
+
+#ifdef GL_PGI_misc_hints
+
+#endif /* GL_PGI_misc_hints */
+
+#ifdef GL_PGI_vertex_hints
+
+#endif /* GL_PGI_vertex_hints */
+
+#ifdef GL_REND_screen_coordinates
+
+#endif /* GL_REND_screen_coordinates */
+
+#ifdef GL_S3_s3tc
+
+#endif /* GL_S3_s3tc */
+
+#ifdef GL_SGIS_color_range
+
+#endif /* GL_SGIS_color_range */
+
+#ifdef GL_SGIS_detail_texture
+
+static GLboolean _glewInit_GL_SGIS_detail_texture (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glDetailTexFuncSGIS = (PFNGLDETAILTEXFUNCSGISPROC)glewGetProcAddress((const GLubyte*)"glDetailTexFuncSGIS")) == NULL) || r;
+  r = ((glGetDetailTexFuncSGIS = (PFNGLGETDETAILTEXFUNCSGISPROC)glewGetProcAddress((const GLubyte*)"glGetDetailTexFuncSGIS")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIS_detail_texture */
+
+#ifdef GL_SGIS_fog_function
+
+static GLboolean _glewInit_GL_SGIS_fog_function (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFogFuncSGIS = (PFNGLFOGFUNCSGISPROC)glewGetProcAddress((const GLubyte*)"glFogFuncSGIS")) == NULL) || r;
+  r = ((glGetFogFuncSGIS = (PFNGLGETFOGFUNCSGISPROC)glewGetProcAddress((const GLubyte*)"glGetFogFuncSGIS")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIS_fog_function */
+
+#ifdef GL_SGIS_generate_mipmap
+
+#endif /* GL_SGIS_generate_mipmap */
+
+#ifdef GL_SGIS_multisample
+
+static GLboolean _glewInit_GL_SGIS_multisample (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glSampleMaskSGIS = (PFNGLSAMPLEMASKSGISPROC)glewGetProcAddress((const GLubyte*)"glSampleMaskSGIS")) == NULL) || r;
+  r = ((glSamplePatternSGIS = (PFNGLSAMPLEPATTERNSGISPROC)glewGetProcAddress((const GLubyte*)"glSamplePatternSGIS")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIS_multisample */
+
+#ifdef GL_SGIS_pixel_texture
+
+#endif /* GL_SGIS_pixel_texture */
+
+#ifdef GL_SGIS_point_line_texgen
+
+#endif /* GL_SGIS_point_line_texgen */
+
+#ifdef GL_SGIS_sharpen_texture
+
+static GLboolean _glewInit_GL_SGIS_sharpen_texture (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetSharpenTexFuncSGIS = (PFNGLGETSHARPENTEXFUNCSGISPROC)glewGetProcAddress((const GLubyte*)"glGetSharpenTexFuncSGIS")) == NULL) || r;
+  r = ((glSharpenTexFuncSGIS = (PFNGLSHARPENTEXFUNCSGISPROC)glewGetProcAddress((const GLubyte*)"glSharpenTexFuncSGIS")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIS_sharpen_texture */
+
+#ifdef GL_SGIS_texture4D
+
+static GLboolean _glewInit_GL_SGIS_texture4D (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glTexImage4DSGIS = (PFNGLTEXIMAGE4DSGISPROC)glewGetProcAddress((const GLubyte*)"glTexImage4DSGIS")) == NULL) || r;
+  r = ((glTexSubImage4DSGIS = (PFNGLTEXSUBIMAGE4DSGISPROC)glewGetProcAddress((const GLubyte*)"glTexSubImage4DSGIS")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIS_texture4D */
+
+#ifdef GL_SGIS_texture_border_clamp
+
+#endif /* GL_SGIS_texture_border_clamp */
+
+#ifdef GL_SGIS_texture_edge_clamp
+
+#endif /* GL_SGIS_texture_edge_clamp */
+
+#ifdef GL_SGIS_texture_filter4
+
+static GLboolean _glewInit_GL_SGIS_texture_filter4 (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGetTexFilterFuncSGIS = (PFNGLGETTEXFILTERFUNCSGISPROC)glewGetProcAddress((const GLubyte*)"glGetTexFilterFuncSGIS")) == NULL) || r;
+  r = ((glTexFilterFuncSGIS = (PFNGLTEXFILTERFUNCSGISPROC)glewGetProcAddress((const GLubyte*)"glTexFilterFuncSGIS")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIS_texture_filter4 */
+
+#ifdef GL_SGIS_texture_lod
+
+#endif /* GL_SGIS_texture_lod */
+
+#ifdef GL_SGIS_texture_select
+
+#endif /* GL_SGIS_texture_select */
+
+#ifdef GL_SGIX_async
+
+static GLboolean _glewInit_GL_SGIX_async (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glAsyncMarkerSGIX = (PFNGLASYNCMARKERSGIXPROC)glewGetProcAddress((const GLubyte*)"glAsyncMarkerSGIX")) == NULL) || r;
+  r = ((glDeleteAsyncMarkersSGIX = (PFNGLDELETEASYNCMARKERSSGIXPROC)glewGetProcAddress((const GLubyte*)"glDeleteAsyncMarkersSGIX")) == NULL) || r;
+  r = ((glFinishAsyncSGIX = (PFNGLFINISHASYNCSGIXPROC)glewGetProcAddress((const GLubyte*)"glFinishAsyncSGIX")) == NULL) || r;
+  r = ((glGenAsyncMarkersSGIX = (PFNGLGENASYNCMARKERSSGIXPROC)glewGetProcAddress((const GLubyte*)"glGenAsyncMarkersSGIX")) == NULL) || r;
+  r = ((glIsAsyncMarkerSGIX = (PFNGLISASYNCMARKERSGIXPROC)glewGetProcAddress((const GLubyte*)"glIsAsyncMarkerSGIX")) == NULL) || r;
+  r = ((glPollAsyncSGIX = (PFNGLPOLLASYNCSGIXPROC)glewGetProcAddress((const GLubyte*)"glPollAsyncSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIX_async */
+
+#ifdef GL_SGIX_async_histogram
+
+#endif /* GL_SGIX_async_histogram */
+
+#ifdef GL_SGIX_async_pixel
+
+#endif /* GL_SGIX_async_pixel */
+
+#ifdef GL_SGIX_blend_alpha_minmax
+
+#endif /* GL_SGIX_blend_alpha_minmax */
+
+#ifdef GL_SGIX_clipmap
+
+#endif /* GL_SGIX_clipmap */
+
+#ifdef GL_SGIX_convolution_accuracy
+
+#endif /* GL_SGIX_convolution_accuracy */
+
+#ifdef GL_SGIX_depth_texture
+
+#endif /* GL_SGIX_depth_texture */
+
+#ifdef GL_SGIX_flush_raster
+
+static GLboolean _glewInit_GL_SGIX_flush_raster (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFlushRasterSGIX = (PFNGLFLUSHRASTERSGIXPROC)glewGetProcAddress((const GLubyte*)"glFlushRasterSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIX_flush_raster */
+
+#ifdef GL_SGIX_fog_offset
+
+#endif /* GL_SGIX_fog_offset */
+
+#ifdef GL_SGIX_fog_texture
+
+static GLboolean _glewInit_GL_SGIX_fog_texture (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glTextureFogSGIX = (PFNGLTEXTUREFOGSGIXPROC)glewGetProcAddress((const GLubyte*)"glTextureFogSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIX_fog_texture */
+
+#ifdef GL_SGIX_fragment_specular_lighting
+
+static GLboolean _glewInit_GL_SGIX_fragment_specular_lighting (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFragmentColorMaterialSGIX = (PFNGLFRAGMENTCOLORMATERIALSGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentColorMaterialSGIX")) == NULL) || r;
+  r = ((glFragmentLightModelfSGIX = (PFNGLFRAGMENTLIGHTMODELFSGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightModelfSGIX")) == NULL) || r;
+  r = ((glFragmentLightModelfvSGIX = (PFNGLFRAGMENTLIGHTMODELFVSGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightModelfvSGIX")) == NULL) || r;
+  r = ((glFragmentLightModeliSGIX = (PFNGLFRAGMENTLIGHTMODELISGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightModeliSGIX")) == NULL) || r;
+  r = ((glFragmentLightModelivSGIX = (PFNGLFRAGMENTLIGHTMODELIVSGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightModelivSGIX")) == NULL) || r;
+  r = ((glFragmentLightfSGIX = (PFNGLFRAGMENTLIGHTFSGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightfSGIX")) == NULL) || r;
+  r = ((glFragmentLightfvSGIX = (PFNGLFRAGMENTLIGHTFVSGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightfvSGIX")) == NULL) || r;
+  r = ((glFragmentLightiSGIX = (PFNGLFRAGMENTLIGHTISGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightiSGIX")) == NULL) || r;
+  r = ((glFragmentLightivSGIX = (PFNGLFRAGMENTLIGHTIVSGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentLightivSGIX")) == NULL) || r;
+  r = ((glFragmentMaterialfSGIX = (PFNGLFRAGMENTMATERIALFSGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentMaterialfSGIX")) == NULL) || r;
+  r = ((glFragmentMaterialfvSGIX = (PFNGLFRAGMENTMATERIALFVSGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentMaterialfvSGIX")) == NULL) || r;
+  r = ((glFragmentMaterialiSGIX = (PFNGLFRAGMENTMATERIALISGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentMaterialiSGIX")) == NULL) || r;
+  r = ((glFragmentMaterialivSGIX = (PFNGLFRAGMENTMATERIALIVSGIXPROC)glewGetProcAddress((const GLubyte*)"glFragmentMaterialivSGIX")) == NULL) || r;
+  r = ((glGetFragmentLightfvSGIX = (PFNGLGETFRAGMENTLIGHTFVSGIXPROC)glewGetProcAddress((const GLubyte*)"glGetFragmentLightfvSGIX")) == NULL) || r;
+  r = ((glGetFragmentLightivSGIX = (PFNGLGETFRAGMENTLIGHTIVSGIXPROC)glewGetProcAddress((const GLubyte*)"glGetFragmentLightivSGIX")) == NULL) || r;
+  r = ((glGetFragmentMaterialfvSGIX = (PFNGLGETFRAGMENTMATERIALFVSGIXPROC)glewGetProcAddress((const GLubyte*)"glGetFragmentMaterialfvSGIX")) == NULL) || r;
+  r = ((glGetFragmentMaterialivSGIX = (PFNGLGETFRAGMENTMATERIALIVSGIXPROC)glewGetProcAddress((const GLubyte*)"glGetFragmentMaterialivSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIX_fragment_specular_lighting */
+
+#ifdef GL_SGIX_framezoom
+
+static GLboolean _glewInit_GL_SGIX_framezoom (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFrameZoomSGIX = (PFNGLFRAMEZOOMSGIXPROC)glewGetProcAddress((const GLubyte*)"glFrameZoomSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIX_framezoom */
+
+#ifdef GL_SGIX_interlace
+
+#endif /* GL_SGIX_interlace */
+
+#ifdef GL_SGIX_ir_instrument1
+
+#endif /* GL_SGIX_ir_instrument1 */
+
+#ifdef GL_SGIX_list_priority
+
+#endif /* GL_SGIX_list_priority */
+
+#ifdef GL_SGIX_pixel_texture
+
+static GLboolean _glewInit_GL_SGIX_pixel_texture (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glPixelTexGenSGIX = (PFNGLPIXELTEXGENSGIXPROC)glewGetProcAddress((const GLubyte*)"glPixelTexGenSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIX_pixel_texture */
+
+#ifdef GL_SGIX_pixel_texture_bits
+
+#endif /* GL_SGIX_pixel_texture_bits */
+
+#ifdef GL_SGIX_reference_plane
+
+static GLboolean _glewInit_GL_SGIX_reference_plane (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glReferencePlaneSGIX = (PFNGLREFERENCEPLANESGIXPROC)glewGetProcAddress((const GLubyte*)"glReferencePlaneSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIX_reference_plane */
+
+#ifdef GL_SGIX_resample
+
+#endif /* GL_SGIX_resample */
+
+#ifdef GL_SGIX_shadow
+
+#endif /* GL_SGIX_shadow */
+
+#ifdef GL_SGIX_shadow_ambient
+
+#endif /* GL_SGIX_shadow_ambient */
+
+#ifdef GL_SGIX_sprite
+
+static GLboolean _glewInit_GL_SGIX_sprite (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glSpriteParameterfSGIX = (PFNGLSPRITEPARAMETERFSGIXPROC)glewGetProcAddress((const GLubyte*)"glSpriteParameterfSGIX")) == NULL) || r;
+  r = ((glSpriteParameterfvSGIX = (PFNGLSPRITEPARAMETERFVSGIXPROC)glewGetProcAddress((const GLubyte*)"glSpriteParameterfvSGIX")) == NULL) || r;
+  r = ((glSpriteParameteriSGIX = (PFNGLSPRITEPARAMETERISGIXPROC)glewGetProcAddress((const GLubyte*)"glSpriteParameteriSGIX")) == NULL) || r;
+  r = ((glSpriteParameterivSGIX = (PFNGLSPRITEPARAMETERIVSGIXPROC)glewGetProcAddress((const GLubyte*)"glSpriteParameterivSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIX_sprite */
+
+#ifdef GL_SGIX_tag_sample_buffer
+
+static GLboolean _glewInit_GL_SGIX_tag_sample_buffer (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glTagSampleBufferSGIX = (PFNGLTAGSAMPLEBUFFERSGIXPROC)glewGetProcAddress((const GLubyte*)"glTagSampleBufferSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGIX_tag_sample_buffer */
+
+#ifdef GL_SGIX_texture_add_env
+
+#endif /* GL_SGIX_texture_add_env */
+
+#ifdef GL_SGIX_texture_coordinate_clamp
+
+#endif /* GL_SGIX_texture_coordinate_clamp */
+
+#ifdef GL_SGIX_texture_lod_bias
+
+#endif /* GL_SGIX_texture_lod_bias */
+
+#ifdef GL_SGIX_texture_multi_buffer
+
+#endif /* GL_SGIX_texture_multi_buffer */
+
+#ifdef GL_SGIX_texture_range
+
+#endif /* GL_SGIX_texture_range */
+
+#ifdef GL_SGIX_texture_scale_bias
+
+#endif /* GL_SGIX_texture_scale_bias */
+
+#ifdef GL_SGIX_vertex_preclip
+
+#endif /* GL_SGIX_vertex_preclip */
+
+#ifdef GL_SGIX_vertex_preclip_hint
+
+#endif /* GL_SGIX_vertex_preclip_hint */
+
+#ifdef GL_SGIX_ycrcb
+
+#endif /* GL_SGIX_ycrcb */
+
+#ifdef GL_SGI_color_matrix
+
+#endif /* GL_SGI_color_matrix */
+
+#ifdef GL_SGI_color_table
+
+static GLboolean _glewInit_GL_SGI_color_table (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glColorTableParameterfvSGI = (PFNGLCOLORTABLEPARAMETERFVSGIPROC)glewGetProcAddress((const GLubyte*)"glColorTableParameterfvSGI")) == NULL) || r;
+  r = ((glColorTableParameterivSGI = (PFNGLCOLORTABLEPARAMETERIVSGIPROC)glewGetProcAddress((const GLubyte*)"glColorTableParameterivSGI")) == NULL) || r;
+  r = ((glColorTableSGI = (PFNGLCOLORTABLESGIPROC)glewGetProcAddress((const GLubyte*)"glColorTableSGI")) == NULL) || r;
+  r = ((glCopyColorTableSGI = (PFNGLCOPYCOLORTABLESGIPROC)glewGetProcAddress((const GLubyte*)"glCopyColorTableSGI")) == NULL) || r;
+  r = ((glGetColorTableParameterfvSGI = (PFNGLGETCOLORTABLEPARAMETERFVSGIPROC)glewGetProcAddress((const GLubyte*)"glGetColorTableParameterfvSGI")) == NULL) || r;
+  r = ((glGetColorTableParameterivSGI = (PFNGLGETCOLORTABLEPARAMETERIVSGIPROC)glewGetProcAddress((const GLubyte*)"glGetColorTableParameterivSGI")) == NULL) || r;
+  r = ((glGetColorTableSGI = (PFNGLGETCOLORTABLESGIPROC)glewGetProcAddress((const GLubyte*)"glGetColorTableSGI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SGI_color_table */
+
+#ifdef GL_SGI_texture_color_table
+
+#endif /* GL_SGI_texture_color_table */
+
+#ifdef GL_SUNX_constant_data
+
+static GLboolean _glewInit_GL_SUNX_constant_data (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glFinishTextureSUNX = (PFNGLFINISHTEXTURESUNXPROC)glewGetProcAddress((const GLubyte*)"glFinishTextureSUNX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SUNX_constant_data */
+
+#ifdef GL_SUN_convolution_border_modes
+
+#endif /* GL_SUN_convolution_border_modes */
+
+#ifdef GL_SUN_global_alpha
+
+static GLboolean _glewInit_GL_SUN_global_alpha (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glGlobalAlphaFactorbSUN = (PFNGLGLOBALALPHAFACTORBSUNPROC)glewGetProcAddress((const GLubyte*)"glGlobalAlphaFactorbSUN")) == NULL) || r;
+  r = ((glGlobalAlphaFactordSUN = (PFNGLGLOBALALPHAFACTORDSUNPROC)glewGetProcAddress((const GLubyte*)"glGlobalAlphaFactordSUN")) == NULL) || r;
+  r = ((glGlobalAlphaFactorfSUN = (PFNGLGLOBALALPHAFACTORFSUNPROC)glewGetProcAddress((const GLubyte*)"glGlobalAlphaFactorfSUN")) == NULL) || r;
+  r = ((glGlobalAlphaFactoriSUN = (PFNGLGLOBALALPHAFACTORISUNPROC)glewGetProcAddress((const GLubyte*)"glGlobalAlphaFactoriSUN")) == NULL) || r;
+  r = ((glGlobalAlphaFactorsSUN = (PFNGLGLOBALALPHAFACTORSSUNPROC)glewGetProcAddress((const GLubyte*)"glGlobalAlphaFactorsSUN")) == NULL) || r;
+  r = ((glGlobalAlphaFactorubSUN = (PFNGLGLOBALALPHAFACTORUBSUNPROC)glewGetProcAddress((const GLubyte*)"glGlobalAlphaFactorubSUN")) == NULL) || r;
+  r = ((glGlobalAlphaFactoruiSUN = (PFNGLGLOBALALPHAFACTORUISUNPROC)glewGetProcAddress((const GLubyte*)"glGlobalAlphaFactoruiSUN")) == NULL) || r;
+  r = ((glGlobalAlphaFactorusSUN = (PFNGLGLOBALALPHAFACTORUSSUNPROC)glewGetProcAddress((const GLubyte*)"glGlobalAlphaFactorusSUN")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SUN_global_alpha */
+
+#ifdef GL_SUN_mesh_array
+
+#endif /* GL_SUN_mesh_array */
+
+#ifdef GL_SUN_read_video_pixels
+
+static GLboolean _glewInit_GL_SUN_read_video_pixels (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glReadVideoPixelsSUN = (PFNGLREADVIDEOPIXELSSUNPROC)glewGetProcAddress((const GLubyte*)"glReadVideoPixelsSUN")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SUN_read_video_pixels */
+
+#ifdef GL_SUN_slice_accum
+
+#endif /* GL_SUN_slice_accum */
+
+#ifdef GL_SUN_triangle_list
+
+static GLboolean _glewInit_GL_SUN_triangle_list (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glReplacementCodePointerSUN = (PFNGLREPLACEMENTCODEPOINTERSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodePointerSUN")) == NULL) || r;
+  r = ((glReplacementCodeubSUN = (PFNGLREPLACEMENTCODEUBSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeubSUN")) == NULL) || r;
+  r = ((glReplacementCodeubvSUN = (PFNGLREPLACEMENTCODEUBVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeubvSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiSUN = (PFNGLREPLACEMENTCODEUISUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiSUN")) == NULL) || r;
+  r = ((glReplacementCodeuivSUN = (PFNGLREPLACEMENTCODEUIVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuivSUN")) == NULL) || r;
+  r = ((glReplacementCodeusSUN = (PFNGLREPLACEMENTCODEUSSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeusSUN")) == NULL) || r;
+  r = ((glReplacementCodeusvSUN = (PFNGLREPLACEMENTCODEUSVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeusvSUN")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SUN_triangle_list */
+
+#ifdef GL_SUN_vertex
+
+static GLboolean _glewInit_GL_SUN_vertex (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glColor3fVertex3fSUN = (PFNGLCOLOR3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glColor3fVertex3fSUN")) == NULL) || r;
+  r = ((glColor3fVertex3fvSUN = (PFNGLCOLOR3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glColor3fVertex3fvSUN")) == NULL) || r;
+  r = ((glColor4fNormal3fVertex3fSUN = (PFNGLCOLOR4FNORMAL3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glColor4fNormal3fVertex3fSUN")) == NULL) || r;
+  r = ((glColor4fNormal3fVertex3fvSUN = (PFNGLCOLOR4FNORMAL3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glColor4fNormal3fVertex3fvSUN")) == NULL) || r;
+  r = ((glColor4ubVertex2fSUN = (PFNGLCOLOR4UBVERTEX2FSUNPROC)glewGetProcAddress((const GLubyte*)"glColor4ubVertex2fSUN")) == NULL) || r;
+  r = ((glColor4ubVertex2fvSUN = (PFNGLCOLOR4UBVERTEX2FVSUNPROC)glewGetProcAddress((const GLubyte*)"glColor4ubVertex2fvSUN")) == NULL) || r;
+  r = ((glColor4ubVertex3fSUN = (PFNGLCOLOR4UBVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glColor4ubVertex3fSUN")) == NULL) || r;
+  r = ((glColor4ubVertex3fvSUN = (PFNGLCOLOR4UBVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glColor4ubVertex3fvSUN")) == NULL) || r;
+  r = ((glNormal3fVertex3fSUN = (PFNGLNORMAL3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glNormal3fVertex3fSUN")) == NULL) || r;
+  r = ((glNormal3fVertex3fvSUN = (PFNGLNORMAL3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glNormal3fVertex3fvSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiColor3fVertex3fSUN = (PFNGLREPLACEMENTCODEUICOLOR3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiColor3fVertex3fSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiColor3fVertex3fvSUN = (PFNGLREPLACEMENTCODEUICOLOR3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiColor3fVertex3fvSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiColor4fNormal3fVertex3fSUN = (PFNGLREPLACEMENTCODEUICOLOR4FNORMAL3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiColor4fNormal3fVertex3fSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiColor4fNormal3fVertex3fvSUN = (PFNGLREPLACEMENTCODEUICOLOR4FNORMAL3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiColor4fNormal3fVertex3fvSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiColor4ubVertex3fSUN = (PFNGLREPLACEMENTCODEUICOLOR4UBVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiColor4ubVertex3fSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiColor4ubVertex3fvSUN = (PFNGLREPLACEMENTCODEUICOLOR4UBVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiColor4ubVertex3fvSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiNormal3fVertex3fSUN = (PFNGLREPLACEMENTCODEUINORMAL3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiNormal3fVertex3fSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiNormal3fVertex3fvSUN = (PFNGLREPLACEMENTCODEUINORMAL3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiNormal3fVertex3fvSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN = (PFNGLREPLACEMENTCODEUITEXCOORD2FCOLOR4FNORMAL3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN = (PFNGLREPLACEMENTCODEUITEXCOORD2FCOLOR4FNORMAL3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiTexCoord2fNormal3fVertex3fSUN = (PFNGLREPLACEMENTCODEUITEXCOORD2FNORMAL3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiTexCoord2fNormal3fVertex3fSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN = (PFNGLREPLACEMENTCODEUITEXCOORD2FNORMAL3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiTexCoord2fVertex3fSUN = (PFNGLREPLACEMENTCODEUITEXCOORD2FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiTexCoord2fVertex3fSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiTexCoord2fVertex3fvSUN = (PFNGLREPLACEMENTCODEUITEXCOORD2FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiTexCoord2fVertex3fvSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiVertex3fSUN = (PFNGLREPLACEMENTCODEUIVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiVertex3fSUN")) == NULL) || r;
+  r = ((glReplacementCodeuiVertex3fvSUN = (PFNGLREPLACEMENTCODEUIVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glReplacementCodeuiVertex3fvSUN")) == NULL) || r;
+  r = ((glTexCoord2fColor3fVertex3fSUN = (PFNGLTEXCOORD2FCOLOR3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2fColor3fVertex3fSUN")) == NULL) || r;
+  r = ((glTexCoord2fColor3fVertex3fvSUN = (PFNGLTEXCOORD2FCOLOR3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2fColor3fVertex3fvSUN")) == NULL) || r;
+  r = ((glTexCoord2fColor4fNormal3fVertex3fSUN = (PFNGLTEXCOORD2FCOLOR4FNORMAL3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2fColor4fNormal3fVertex3fSUN")) == NULL) || r;
+  r = ((glTexCoord2fColor4fNormal3fVertex3fvSUN = (PFNGLTEXCOORD2FCOLOR4FNORMAL3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2fColor4fNormal3fVertex3fvSUN")) == NULL) || r;
+  r = ((glTexCoord2fColor4ubVertex3fSUN = (PFNGLTEXCOORD2FCOLOR4UBVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2fColor4ubVertex3fSUN")) == NULL) || r;
+  r = ((glTexCoord2fColor4ubVertex3fvSUN = (PFNGLTEXCOORD2FCOLOR4UBVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2fColor4ubVertex3fvSUN")) == NULL) || r;
+  r = ((glTexCoord2fNormal3fVertex3fSUN = (PFNGLTEXCOORD2FNORMAL3FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2fNormal3fVertex3fSUN")) == NULL) || r;
+  r = ((glTexCoord2fNormal3fVertex3fvSUN = (PFNGLTEXCOORD2FNORMAL3FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2fNormal3fVertex3fvSUN")) == NULL) || r;
+  r = ((glTexCoord2fVertex3fSUN = (PFNGLTEXCOORD2FVERTEX3FSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2fVertex3fSUN")) == NULL) || r;
+  r = ((glTexCoord2fVertex3fvSUN = (PFNGLTEXCOORD2FVERTEX3FVSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord2fVertex3fvSUN")) == NULL) || r;
+  r = ((glTexCoord4fColor4fNormal3fVertex4fSUN = (PFNGLTEXCOORD4FCOLOR4FNORMAL3FVERTEX4FSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord4fColor4fNormal3fVertex4fSUN")) == NULL) || r;
+  r = ((glTexCoord4fColor4fNormal3fVertex4fvSUN = (PFNGLTEXCOORD4FCOLOR4FNORMAL3FVERTEX4FVSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord4fColor4fNormal3fVertex4fvSUN")) == NULL) || r;
+  r = ((glTexCoord4fVertex4fSUN = (PFNGLTEXCOORD4FVERTEX4FSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord4fVertex4fSUN")) == NULL) || r;
+  r = ((glTexCoord4fVertex4fvSUN = (PFNGLTEXCOORD4FVERTEX4FVSUNPROC)glewGetProcAddress((const GLubyte*)"glTexCoord4fVertex4fvSUN")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_SUN_vertex */
+
+#ifdef GL_WIN_phong_shading
+
+#endif /* GL_WIN_phong_shading */
+
+#ifdef GL_WIN_specular_fog
+
+#endif /* GL_WIN_specular_fog */
+
+#ifdef GL_WIN_swap_hint
+
+static GLboolean _glewInit_GL_WIN_swap_hint (GLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glAddSwapHintRectWIN = (PFNGLADDSWAPHINTRECTWINPROC)glewGetProcAddress((const GLubyte*)"glAddSwapHintRectWIN")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GL_WIN_swap_hint */
+
+/* ------------------------------------------------------------------------- */
+
+/* 
+ * Search for name in the extensions string. Use of strstr()
+ * is not sufficient because extension names can be prefixes of
+ * other extension names. Could use strtok() but the constant
+ * string returned by glGetString might be in read-only memory.
+ */
+GLboolean glewGetExtension (const char* name)
+{    
+  GLubyte* p;
+  GLubyte* end;
+  GLuint len = _glewStrLen((const GLubyte*)name);
+  p = (GLubyte*)glGetString(GL_EXTENSIONS);
+  if (0 == p) return GL_FALSE;
+  end = p + _glewStrLen(p);
+  while (p < end)
+  {
+    GLuint n = _glewStrCLen(p, ' ');
+    if (len == n && _glewStrSame((const GLubyte*)name, p, n)) return GL_TRUE;
+    p += n+1;
+  }
+  return GL_FALSE;
+}
+
+/* ------------------------------------------------------------------------- */
+
+#ifndef GLEW_MX
+static
+#endif
+GLenum glewContextInit (GLEW_CONTEXT_ARG_DEF_LIST)
+{
+  const GLubyte* s;
+  GLuint dot, major, minor;
+  /* query opengl version */
+  s = glGetString(GL_VERSION);
+  dot = _glewStrCLen(s, '.');
+  major = dot-1;
+  minor = dot+1;
+  if (dot == 0 || s[minor] == '\0')
+    return GLEW_ERROR_NO_GL_VERSION;
+  if (s[major] == '1' && s[minor] == '0')
+  {
+	return GLEW_ERROR_GL_VERSION_10_ONLY;
+  }
+  else
+  {
+    CONST_CAST(GLEW_VERSION_1_1) = GL_TRUE;
+	if (s[major] >= '2')
+	{
+      CONST_CAST(GLEW_VERSION_1_2) = GL_TRUE;
+      CONST_CAST(GLEW_VERSION_1_3) = GL_TRUE;
+      CONST_CAST(GLEW_VERSION_1_4) = GL_TRUE;
+	  CONST_CAST(GLEW_VERSION_1_5) = GL_TRUE;
+	  CONST_CAST(GLEW_VERSION_2_0) = GL_TRUE;
+	  if (s[minor] >= '1')
+	  {
+	    CONST_CAST(GLEW_VERSION_2_1) = GL_TRUE;
+      }
+	}
+	else
+	{
+	  if (s[minor] >= '5')
+	  {
+		CONST_CAST(GLEW_VERSION_1_2) = GL_TRUE;
+		CONST_CAST(GLEW_VERSION_1_3) = GL_TRUE;
+		CONST_CAST(GLEW_VERSION_1_4) = GL_TRUE;
+		CONST_CAST(GLEW_VERSION_1_5) = GL_TRUE;
+		CONST_CAST(GLEW_VERSION_2_0) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_2_1) = GL_FALSE;
+	  }
+	  if (s[minor] == '4')
+	  {
+		CONST_CAST(GLEW_VERSION_1_2) = GL_TRUE;
+		CONST_CAST(GLEW_VERSION_1_3) = GL_TRUE;
+		CONST_CAST(GLEW_VERSION_1_4) = GL_TRUE;
+		CONST_CAST(GLEW_VERSION_1_5) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_2_0) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_2_1) = GL_FALSE;
+	  }
+	  if (s[minor] == '3')
+	  {
+		CONST_CAST(GLEW_VERSION_1_2) = GL_TRUE;
+		CONST_CAST(GLEW_VERSION_1_3) = GL_TRUE;
+		CONST_CAST(GLEW_VERSION_1_4) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_1_5) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_2_0) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_2_1) = GL_FALSE;
+	  }
+	  if (s[minor] == '2')
+	  {
+		CONST_CAST(GLEW_VERSION_1_2) = GL_TRUE;
+		CONST_CAST(GLEW_VERSION_1_3) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_1_4) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_1_5) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_2_0) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_2_1) = GL_FALSE;
+	  }
+	  if (s[minor] < '2')
+	  {
+		CONST_CAST(GLEW_VERSION_1_2) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_1_3) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_1_4) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_1_5) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_2_0) = GL_FALSE;
+		CONST_CAST(GLEW_VERSION_2_1) = GL_FALSE;
+	  }
+	}
+  }
+  /* initialize extensions */
+#ifdef GL_VERSION_1_2
+  if (glewExperimental || GLEW_VERSION_1_2) CONST_CAST(GLEW_VERSION_1_2) = !_glewInit_GL_VERSION_1_2(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_VERSION_1_2 */
+#ifdef GL_VERSION_1_3
+  if (glewExperimental || GLEW_VERSION_1_3) CONST_CAST(GLEW_VERSION_1_3) = !_glewInit_GL_VERSION_1_3(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_VERSION_1_3 */
+#ifdef GL_VERSION_1_4
+  if (glewExperimental || GLEW_VERSION_1_4) CONST_CAST(GLEW_VERSION_1_4) = !_glewInit_GL_VERSION_1_4(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_VERSION_1_4 */
+#ifdef GL_VERSION_1_5
+  if (glewExperimental || GLEW_VERSION_1_5) CONST_CAST(GLEW_VERSION_1_5) = !_glewInit_GL_VERSION_1_5(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_VERSION_1_5 */
+#ifdef GL_VERSION_2_0
+  if (glewExperimental || GLEW_VERSION_2_0) CONST_CAST(GLEW_VERSION_2_0) = !_glewInit_GL_VERSION_2_0(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_VERSION_2_0 */
+#ifdef GL_VERSION_2_1
+  if (glewExperimental || GLEW_VERSION_2_1) CONST_CAST(GLEW_VERSION_2_1) = !_glewInit_GL_VERSION_2_1(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_VERSION_2_1 */
+#ifdef GL_VERSION_3_0
+  if (glewExperimental || GLEW_VERSION_3_0) CONST_CAST(GLEW_VERSION_3_0) = !_glewInit_GL_VERSION_3_0(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_VERSION_3_0 */
+#ifdef GL_3DFX_multisample
+  CONST_CAST(GLEW_3DFX_multisample) = glewGetExtension("GL_3DFX_multisample");
+#endif /* GL_3DFX_multisample */
+#ifdef GL_3DFX_tbuffer
+  CONST_CAST(GLEW_3DFX_tbuffer) = glewGetExtension("GL_3DFX_tbuffer");
+  if (glewExperimental || GLEW_3DFX_tbuffer) CONST_CAST(GLEW_3DFX_tbuffer) = !_glewInit_GL_3DFX_tbuffer(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_3DFX_tbuffer */
+#ifdef GL_3DFX_texture_compression_FXT1
+  CONST_CAST(GLEW_3DFX_texture_compression_FXT1) = glewGetExtension("GL_3DFX_texture_compression_FXT1");
+#endif /* GL_3DFX_texture_compression_FXT1 */
+#ifdef GL_APPLE_client_storage
+  CONST_CAST(GLEW_APPLE_client_storage) = glewGetExtension("GL_APPLE_client_storage");
+#endif /* GL_APPLE_client_storage */
+#ifdef GL_APPLE_element_array
+  CONST_CAST(GLEW_APPLE_element_array) = glewGetExtension("GL_APPLE_element_array");
+  if (glewExperimental || GLEW_APPLE_element_array) CONST_CAST(GLEW_APPLE_element_array) = !_glewInit_GL_APPLE_element_array(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_APPLE_element_array */
+#ifdef GL_APPLE_fence
+  CONST_CAST(GLEW_APPLE_fence) = glewGetExtension("GL_APPLE_fence");
+  if (glewExperimental || GLEW_APPLE_fence) CONST_CAST(GLEW_APPLE_fence) = !_glewInit_GL_APPLE_fence(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_APPLE_fence */
+#ifdef GL_APPLE_float_pixels
+  CONST_CAST(GLEW_APPLE_float_pixels) = glewGetExtension("GL_APPLE_float_pixels");
+#endif /* GL_APPLE_float_pixels */
+#ifdef GL_APPLE_flush_buffer_range
+  CONST_CAST(GLEW_APPLE_flush_buffer_range) = glewGetExtension("GL_APPLE_flush_buffer_range");
+  if (glewExperimental || GLEW_APPLE_flush_buffer_range) CONST_CAST(GLEW_APPLE_flush_buffer_range) = !_glewInit_GL_APPLE_flush_buffer_range(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_APPLE_flush_buffer_range */
+#ifdef GL_APPLE_pixel_buffer
+  CONST_CAST(GLEW_APPLE_pixel_buffer) = glewGetExtension("GL_APPLE_pixel_buffer");
+#endif /* GL_APPLE_pixel_buffer */
+#ifdef GL_APPLE_specular_vector
+  CONST_CAST(GLEW_APPLE_specular_vector) = glewGetExtension("GL_APPLE_specular_vector");
+#endif /* GL_APPLE_specular_vector */
+#ifdef GL_APPLE_texture_range
+  CONST_CAST(GLEW_APPLE_texture_range) = glewGetExtension("GL_APPLE_texture_range");
+  if (glewExperimental || GLEW_APPLE_texture_range) CONST_CAST(GLEW_APPLE_texture_range) = !_glewInit_GL_APPLE_texture_range(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_APPLE_texture_range */
+#ifdef GL_APPLE_transform_hint
+  CONST_CAST(GLEW_APPLE_transform_hint) = glewGetExtension("GL_APPLE_transform_hint");
+#endif /* GL_APPLE_transform_hint */
+#ifdef GL_APPLE_vertex_array_object
+  CONST_CAST(GLEW_APPLE_vertex_array_object) = glewGetExtension("GL_APPLE_vertex_array_object");
+  if (glewExperimental || GLEW_APPLE_vertex_array_object) CONST_CAST(GLEW_APPLE_vertex_array_object) = !_glewInit_GL_APPLE_vertex_array_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_APPLE_vertex_array_object */
+#ifdef GL_APPLE_vertex_array_range
+  CONST_CAST(GLEW_APPLE_vertex_array_range) = glewGetExtension("GL_APPLE_vertex_array_range");
+  if (glewExperimental || GLEW_APPLE_vertex_array_range) CONST_CAST(GLEW_APPLE_vertex_array_range) = !_glewInit_GL_APPLE_vertex_array_range(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_APPLE_vertex_array_range */
+#ifdef GL_APPLE_ycbcr_422
+  CONST_CAST(GLEW_APPLE_ycbcr_422) = glewGetExtension("GL_APPLE_ycbcr_422");
+#endif /* GL_APPLE_ycbcr_422 */
+#ifdef GL_ARB_color_buffer_float
+  CONST_CAST(GLEW_ARB_color_buffer_float) = glewGetExtension("GL_ARB_color_buffer_float");
+  if (glewExperimental || GLEW_ARB_color_buffer_float) CONST_CAST(GLEW_ARB_color_buffer_float) = !_glewInit_GL_ARB_color_buffer_float(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_color_buffer_float */
+#ifdef GL_ARB_depth_buffer_float
+  CONST_CAST(GLEW_ARB_depth_buffer_float) = glewGetExtension("GL_ARB_depth_buffer_float");
+#endif /* GL_ARB_depth_buffer_float */
+#ifdef GL_ARB_depth_texture
+  CONST_CAST(GLEW_ARB_depth_texture) = glewGetExtension("GL_ARB_depth_texture");
+#endif /* GL_ARB_depth_texture */
+#ifdef GL_ARB_draw_buffers
+  CONST_CAST(GLEW_ARB_draw_buffers) = glewGetExtension("GL_ARB_draw_buffers");
+  if (glewExperimental || GLEW_ARB_draw_buffers) CONST_CAST(GLEW_ARB_draw_buffers) = !_glewInit_GL_ARB_draw_buffers(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_draw_buffers */
+#ifdef GL_ARB_draw_instanced
+  CONST_CAST(GLEW_ARB_draw_instanced) = glewGetExtension("GL_ARB_draw_instanced");
+  if (glewExperimental || GLEW_ARB_draw_instanced) CONST_CAST(GLEW_ARB_draw_instanced) = !_glewInit_GL_ARB_draw_instanced(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_draw_instanced */
+#ifdef GL_ARB_fragment_program
+  CONST_CAST(GLEW_ARB_fragment_program) = glewGetExtension("GL_ARB_fragment_program");
+#endif /* GL_ARB_fragment_program */
+#ifdef GL_ARB_fragment_program_shadow
+  CONST_CAST(GLEW_ARB_fragment_program_shadow) = glewGetExtension("GL_ARB_fragment_program_shadow");
+#endif /* GL_ARB_fragment_program_shadow */
+#ifdef GL_ARB_fragment_shader
+  CONST_CAST(GLEW_ARB_fragment_shader) = glewGetExtension("GL_ARB_fragment_shader");
+#endif /* GL_ARB_fragment_shader */
+#ifdef GL_ARB_framebuffer_object
+  CONST_CAST(GLEW_ARB_framebuffer_object) = glewGetExtension("GL_ARB_framebuffer_object");
+  if (glewExperimental || GLEW_ARB_framebuffer_object) CONST_CAST(GLEW_ARB_framebuffer_object) = !_glewInit_GL_ARB_framebuffer_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_framebuffer_object */
+#ifdef GL_ARB_framebuffer_sRGB
+  CONST_CAST(GLEW_ARB_framebuffer_sRGB) = glewGetExtension("GL_ARB_framebuffer_sRGB");
+#endif /* GL_ARB_framebuffer_sRGB */
+#ifdef GL_ARB_geometry_shader4
+  CONST_CAST(GLEW_ARB_geometry_shader4) = glewGetExtension("GL_ARB_geometry_shader4");
+  if (glewExperimental || GLEW_ARB_geometry_shader4) CONST_CAST(GLEW_ARB_geometry_shader4) = !_glewInit_GL_ARB_geometry_shader4(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_geometry_shader4 */
+#ifdef GL_ARB_half_float_pixel
+  CONST_CAST(GLEW_ARB_half_float_pixel) = glewGetExtension("GL_ARB_half_float_pixel");
+#endif /* GL_ARB_half_float_pixel */
+#ifdef GL_ARB_half_float_vertex
+  CONST_CAST(GLEW_ARB_half_float_vertex) = glewGetExtension("GL_ARB_half_float_vertex");
+#endif /* GL_ARB_half_float_vertex */
+#ifdef GL_ARB_imaging
+  CONST_CAST(GLEW_ARB_imaging) = glewGetExtension("GL_ARB_imaging");
+  if (glewExperimental || GLEW_ARB_imaging) CONST_CAST(GLEW_ARB_imaging) = !_glewInit_GL_ARB_imaging(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_imaging */
+#ifdef GL_ARB_instanced_arrays
+  CONST_CAST(GLEW_ARB_instanced_arrays) = glewGetExtension("GL_ARB_instanced_arrays");
+  if (glewExperimental || GLEW_ARB_instanced_arrays) CONST_CAST(GLEW_ARB_instanced_arrays) = !_glewInit_GL_ARB_instanced_arrays(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_instanced_arrays */
+#ifdef GL_ARB_map_buffer_range
+  CONST_CAST(GLEW_ARB_map_buffer_range) = glewGetExtension("GL_ARB_map_buffer_range");
+  if (glewExperimental || GLEW_ARB_map_buffer_range) CONST_CAST(GLEW_ARB_map_buffer_range) = !_glewInit_GL_ARB_map_buffer_range(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_map_buffer_range */
+#ifdef GL_ARB_matrix_palette
+  CONST_CAST(GLEW_ARB_matrix_palette) = glewGetExtension("GL_ARB_matrix_palette");
+  if (glewExperimental || GLEW_ARB_matrix_palette) CONST_CAST(GLEW_ARB_matrix_palette) = !_glewInit_GL_ARB_matrix_palette(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_matrix_palette */
+#ifdef GL_ARB_multisample
+  CONST_CAST(GLEW_ARB_multisample) = glewGetExtension("GL_ARB_multisample");
+  if (glewExperimental || GLEW_ARB_multisample) CONST_CAST(GLEW_ARB_multisample) = !_glewInit_GL_ARB_multisample(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_multisample */
+#ifdef GL_ARB_multitexture
+  CONST_CAST(GLEW_ARB_multitexture) = glewGetExtension("GL_ARB_multitexture");
+  if (glewExperimental || GLEW_ARB_multitexture) CONST_CAST(GLEW_ARB_multitexture) = !_glewInit_GL_ARB_multitexture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_multitexture */
+#ifdef GL_ARB_occlusion_query
+  CONST_CAST(GLEW_ARB_occlusion_query) = glewGetExtension("GL_ARB_occlusion_query");
+  if (glewExperimental || GLEW_ARB_occlusion_query) CONST_CAST(GLEW_ARB_occlusion_query) = !_glewInit_GL_ARB_occlusion_query(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_occlusion_query */
+#ifdef GL_ARB_pixel_buffer_object
+  CONST_CAST(GLEW_ARB_pixel_buffer_object) = glewGetExtension("GL_ARB_pixel_buffer_object");
+#endif /* GL_ARB_pixel_buffer_object */
+#ifdef GL_ARB_point_parameters
+  CONST_CAST(GLEW_ARB_point_parameters) = glewGetExtension("GL_ARB_point_parameters");
+  if (glewExperimental || GLEW_ARB_point_parameters) CONST_CAST(GLEW_ARB_point_parameters) = !_glewInit_GL_ARB_point_parameters(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_point_parameters */
+#ifdef GL_ARB_point_sprite
+  CONST_CAST(GLEW_ARB_point_sprite) = glewGetExtension("GL_ARB_point_sprite");
+#endif /* GL_ARB_point_sprite */
+#ifdef GL_ARB_shader_objects
+  CONST_CAST(GLEW_ARB_shader_objects) = glewGetExtension("GL_ARB_shader_objects");
+  if (glewExperimental || GLEW_ARB_shader_objects) CONST_CAST(GLEW_ARB_shader_objects) = !_glewInit_GL_ARB_shader_objects(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_shader_objects */
+#ifdef GL_ARB_shading_language_100
+  CONST_CAST(GLEW_ARB_shading_language_100) = glewGetExtension("GL_ARB_shading_language_100");
+#endif /* GL_ARB_shading_language_100 */
+#ifdef GL_ARB_shadow
+  CONST_CAST(GLEW_ARB_shadow) = glewGetExtension("GL_ARB_shadow");
+#endif /* GL_ARB_shadow */
+#ifdef GL_ARB_shadow_ambient
+  CONST_CAST(GLEW_ARB_shadow_ambient) = glewGetExtension("GL_ARB_shadow_ambient");
+#endif /* GL_ARB_shadow_ambient */
+#ifdef GL_ARB_texture_border_clamp
+  CONST_CAST(GLEW_ARB_texture_border_clamp) = glewGetExtension("GL_ARB_texture_border_clamp");
+#endif /* GL_ARB_texture_border_clamp */
+#ifdef GL_ARB_texture_buffer_object
+  CONST_CAST(GLEW_ARB_texture_buffer_object) = glewGetExtension("GL_ARB_texture_buffer_object");
+  if (glewExperimental || GLEW_ARB_texture_buffer_object) CONST_CAST(GLEW_ARB_texture_buffer_object) = !_glewInit_GL_ARB_texture_buffer_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_texture_buffer_object */
+#ifdef GL_ARB_texture_compression
+  CONST_CAST(GLEW_ARB_texture_compression) = glewGetExtension("GL_ARB_texture_compression");
+  if (glewExperimental || GLEW_ARB_texture_compression) CONST_CAST(GLEW_ARB_texture_compression) = !_glewInit_GL_ARB_texture_compression(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_texture_compression */
+#ifdef GL_ARB_texture_compression_rgtc
+  CONST_CAST(GLEW_ARB_texture_compression_rgtc) = glewGetExtension("GL_ARB_texture_compression_rgtc");
+#endif /* GL_ARB_texture_compression_rgtc */
+#ifdef GL_ARB_texture_cube_map
+  CONST_CAST(GLEW_ARB_texture_cube_map) = glewGetExtension("GL_ARB_texture_cube_map");
+#endif /* GL_ARB_texture_cube_map */
+#ifdef GL_ARB_texture_env_add
+  CONST_CAST(GLEW_ARB_texture_env_add) = glewGetExtension("GL_ARB_texture_env_add");
+#endif /* GL_ARB_texture_env_add */
+#ifdef GL_ARB_texture_env_combine
+  CONST_CAST(GLEW_ARB_texture_env_combine) = glewGetExtension("GL_ARB_texture_env_combine");
+#endif /* GL_ARB_texture_env_combine */
+#ifdef GL_ARB_texture_env_crossbar
+  CONST_CAST(GLEW_ARB_texture_env_crossbar) = glewGetExtension("GL_ARB_texture_env_crossbar");
+#endif /* GL_ARB_texture_env_crossbar */
+#ifdef GL_ARB_texture_env_dot3
+  CONST_CAST(GLEW_ARB_texture_env_dot3) = glewGetExtension("GL_ARB_texture_env_dot3");
+#endif /* GL_ARB_texture_env_dot3 */
+#ifdef GL_ARB_texture_float
+  CONST_CAST(GLEW_ARB_texture_float) = glewGetExtension("GL_ARB_texture_float");
+#endif /* GL_ARB_texture_float */
+#ifdef GL_ARB_texture_mirrored_repeat
+  CONST_CAST(GLEW_ARB_texture_mirrored_repeat) = glewGetExtension("GL_ARB_texture_mirrored_repeat");
+#endif /* GL_ARB_texture_mirrored_repeat */
+#ifdef GL_ARB_texture_non_power_of_two
+  CONST_CAST(GLEW_ARB_texture_non_power_of_two) = glewGetExtension("GL_ARB_texture_non_power_of_two");
+#endif /* GL_ARB_texture_non_power_of_two */
+#ifdef GL_ARB_texture_rectangle
+  CONST_CAST(GLEW_ARB_texture_rectangle) = glewGetExtension("GL_ARB_texture_rectangle");
+#endif /* GL_ARB_texture_rectangle */
+#ifdef GL_ARB_texture_rg
+  CONST_CAST(GLEW_ARB_texture_rg) = glewGetExtension("GL_ARB_texture_rg");
+#endif /* GL_ARB_texture_rg */
+#ifdef GL_ARB_transpose_matrix
+  CONST_CAST(GLEW_ARB_transpose_matrix) = glewGetExtension("GL_ARB_transpose_matrix");
+  if (glewExperimental || GLEW_ARB_transpose_matrix) CONST_CAST(GLEW_ARB_transpose_matrix) = !_glewInit_GL_ARB_transpose_matrix(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_transpose_matrix */
+#ifdef GL_ARB_vertex_array_object
+  CONST_CAST(GLEW_ARB_vertex_array_object) = glewGetExtension("GL_ARB_vertex_array_object");
+  if (glewExperimental || GLEW_ARB_vertex_array_object) CONST_CAST(GLEW_ARB_vertex_array_object) = !_glewInit_GL_ARB_vertex_array_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_vertex_array_object */
+#ifdef GL_ARB_vertex_blend
+  CONST_CAST(GLEW_ARB_vertex_blend) = glewGetExtension("GL_ARB_vertex_blend");
+  if (glewExperimental || GLEW_ARB_vertex_blend) CONST_CAST(GLEW_ARB_vertex_blend) = !_glewInit_GL_ARB_vertex_blend(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_vertex_blend */
+#ifdef GL_ARB_vertex_buffer_object
+  CONST_CAST(GLEW_ARB_vertex_buffer_object) = glewGetExtension("GL_ARB_vertex_buffer_object");
+  if (glewExperimental || GLEW_ARB_vertex_buffer_object) CONST_CAST(GLEW_ARB_vertex_buffer_object) = !_glewInit_GL_ARB_vertex_buffer_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_vertex_buffer_object */
+#ifdef GL_ARB_vertex_program
+  CONST_CAST(GLEW_ARB_vertex_program) = glewGetExtension("GL_ARB_vertex_program");
+  if (glewExperimental || GLEW_ARB_vertex_program) CONST_CAST(GLEW_ARB_vertex_program) = !_glewInit_GL_ARB_vertex_program(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_vertex_program */
+#ifdef GL_ARB_vertex_shader
+  CONST_CAST(GLEW_ARB_vertex_shader) = glewGetExtension("GL_ARB_vertex_shader");
+  if (glewExperimental || GLEW_ARB_vertex_shader) CONST_CAST(GLEW_ARB_vertex_shader) = !_glewInit_GL_ARB_vertex_shader(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_vertex_shader */
+#ifdef GL_ARB_window_pos
+  CONST_CAST(GLEW_ARB_window_pos) = glewGetExtension("GL_ARB_window_pos");
+  if (glewExperimental || GLEW_ARB_window_pos) CONST_CAST(GLEW_ARB_window_pos) = !_glewInit_GL_ARB_window_pos(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ARB_window_pos */
+#ifdef GL_ATIX_point_sprites
+  CONST_CAST(GLEW_ATIX_point_sprites) = glewGetExtension("GL_ATIX_point_sprites");
+#endif /* GL_ATIX_point_sprites */
+#ifdef GL_ATIX_texture_env_combine3
+  CONST_CAST(GLEW_ATIX_texture_env_combine3) = glewGetExtension("GL_ATIX_texture_env_combine3");
+#endif /* GL_ATIX_texture_env_combine3 */
+#ifdef GL_ATIX_texture_env_route
+  CONST_CAST(GLEW_ATIX_texture_env_route) = glewGetExtension("GL_ATIX_texture_env_route");
+#endif /* GL_ATIX_texture_env_route */
+#ifdef GL_ATIX_vertex_shader_output_point_size
+  CONST_CAST(GLEW_ATIX_vertex_shader_output_point_size) = glewGetExtension("GL_ATIX_vertex_shader_output_point_size");
+#endif /* GL_ATIX_vertex_shader_output_point_size */
+#ifdef GL_ATI_draw_buffers
+  CONST_CAST(GLEW_ATI_draw_buffers) = glewGetExtension("GL_ATI_draw_buffers");
+  if (glewExperimental || GLEW_ATI_draw_buffers) CONST_CAST(GLEW_ATI_draw_buffers) = !_glewInit_GL_ATI_draw_buffers(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ATI_draw_buffers */
+#ifdef GL_ATI_element_array
+  CONST_CAST(GLEW_ATI_element_array) = glewGetExtension("GL_ATI_element_array");
+  if (glewExperimental || GLEW_ATI_element_array) CONST_CAST(GLEW_ATI_element_array) = !_glewInit_GL_ATI_element_array(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ATI_element_array */
+#ifdef GL_ATI_envmap_bumpmap
+  CONST_CAST(GLEW_ATI_envmap_bumpmap) = glewGetExtension("GL_ATI_envmap_bumpmap");
+  if (glewExperimental || GLEW_ATI_envmap_bumpmap) CONST_CAST(GLEW_ATI_envmap_bumpmap) = !_glewInit_GL_ATI_envmap_bumpmap(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ATI_envmap_bumpmap */
+#ifdef GL_ATI_fragment_shader
+  CONST_CAST(GLEW_ATI_fragment_shader) = glewGetExtension("GL_ATI_fragment_shader");
+  if (glewExperimental || GLEW_ATI_fragment_shader) CONST_CAST(GLEW_ATI_fragment_shader) = !_glewInit_GL_ATI_fragment_shader(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ATI_fragment_shader */
+#ifdef GL_ATI_map_object_buffer
+  CONST_CAST(GLEW_ATI_map_object_buffer) = glewGetExtension("GL_ATI_map_object_buffer");
+  if (glewExperimental || GLEW_ATI_map_object_buffer) CONST_CAST(GLEW_ATI_map_object_buffer) = !_glewInit_GL_ATI_map_object_buffer(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ATI_map_object_buffer */
+#ifdef GL_ATI_pn_triangles
+  CONST_CAST(GLEW_ATI_pn_triangles) = glewGetExtension("GL_ATI_pn_triangles");
+  if (glewExperimental || GLEW_ATI_pn_triangles) CONST_CAST(GLEW_ATI_pn_triangles) = !_glewInit_GL_ATI_pn_triangles(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ATI_pn_triangles */
+#ifdef GL_ATI_separate_stencil
+  CONST_CAST(GLEW_ATI_separate_stencil) = glewGetExtension("GL_ATI_separate_stencil");
+  if (glewExperimental || GLEW_ATI_separate_stencil) CONST_CAST(GLEW_ATI_separate_stencil) = !_glewInit_GL_ATI_separate_stencil(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ATI_separate_stencil */
+#ifdef GL_ATI_shader_texture_lod
+  CONST_CAST(GLEW_ATI_shader_texture_lod) = glewGetExtension("GL_ATI_shader_texture_lod");
+#endif /* GL_ATI_shader_texture_lod */
+#ifdef GL_ATI_text_fragment_shader
+  CONST_CAST(GLEW_ATI_text_fragment_shader) = glewGetExtension("GL_ATI_text_fragment_shader");
+#endif /* GL_ATI_text_fragment_shader */
+#ifdef GL_ATI_texture_compression_3dc
+  CONST_CAST(GLEW_ATI_texture_compression_3dc) = glewGetExtension("GL_ATI_texture_compression_3dc");
+#endif /* GL_ATI_texture_compression_3dc */
+#ifdef GL_ATI_texture_env_combine3
+  CONST_CAST(GLEW_ATI_texture_env_combine3) = glewGetExtension("GL_ATI_texture_env_combine3");
+#endif /* GL_ATI_texture_env_combine3 */
+#ifdef GL_ATI_texture_float
+  CONST_CAST(GLEW_ATI_texture_float) = glewGetExtension("GL_ATI_texture_float");
+#endif /* GL_ATI_texture_float */
+#ifdef GL_ATI_texture_mirror_once
+  CONST_CAST(GLEW_ATI_texture_mirror_once) = glewGetExtension("GL_ATI_texture_mirror_once");
+#endif /* GL_ATI_texture_mirror_once */
+#ifdef GL_ATI_vertex_array_object
+  CONST_CAST(GLEW_ATI_vertex_array_object) = glewGetExtension("GL_ATI_vertex_array_object");
+  if (glewExperimental || GLEW_ATI_vertex_array_object) CONST_CAST(GLEW_ATI_vertex_array_object) = !_glewInit_GL_ATI_vertex_array_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ATI_vertex_array_object */
+#ifdef GL_ATI_vertex_attrib_array_object
+  CONST_CAST(GLEW_ATI_vertex_attrib_array_object) = glewGetExtension("GL_ATI_vertex_attrib_array_object");
+  if (glewExperimental || GLEW_ATI_vertex_attrib_array_object) CONST_CAST(GLEW_ATI_vertex_attrib_array_object) = !_glewInit_GL_ATI_vertex_attrib_array_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ATI_vertex_attrib_array_object */
+#ifdef GL_ATI_vertex_streams
+  CONST_CAST(GLEW_ATI_vertex_streams) = glewGetExtension("GL_ATI_vertex_streams");
+  if (glewExperimental || GLEW_ATI_vertex_streams) CONST_CAST(GLEW_ATI_vertex_streams) = !_glewInit_GL_ATI_vertex_streams(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_ATI_vertex_streams */
+#ifdef GL_EXT_422_pixels
+  CONST_CAST(GLEW_EXT_422_pixels) = glewGetExtension("GL_EXT_422_pixels");
+#endif /* GL_EXT_422_pixels */
+#ifdef GL_EXT_Cg_shader
+  CONST_CAST(GLEW_EXT_Cg_shader) = glewGetExtension("GL_EXT_Cg_shader");
+#endif /* GL_EXT_Cg_shader */
+#ifdef GL_EXT_abgr
+  CONST_CAST(GLEW_EXT_abgr) = glewGetExtension("GL_EXT_abgr");
+#endif /* GL_EXT_abgr */
+#ifdef GL_EXT_bgra
+  CONST_CAST(GLEW_EXT_bgra) = glewGetExtension("GL_EXT_bgra");
+#endif /* GL_EXT_bgra */
+#ifdef GL_EXT_bindable_uniform
+  CONST_CAST(GLEW_EXT_bindable_uniform) = glewGetExtension("GL_EXT_bindable_uniform");
+  if (glewExperimental || GLEW_EXT_bindable_uniform) CONST_CAST(GLEW_EXT_bindable_uniform) = !_glewInit_GL_EXT_bindable_uniform(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_bindable_uniform */
+#ifdef GL_EXT_blend_color
+  CONST_CAST(GLEW_EXT_blend_color) = glewGetExtension("GL_EXT_blend_color");
+  if (glewExperimental || GLEW_EXT_blend_color) CONST_CAST(GLEW_EXT_blend_color) = !_glewInit_GL_EXT_blend_color(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_blend_color */
+#ifdef GL_EXT_blend_equation_separate
+  CONST_CAST(GLEW_EXT_blend_equation_separate) = glewGetExtension("GL_EXT_blend_equation_separate");
+  if (glewExperimental || GLEW_EXT_blend_equation_separate) CONST_CAST(GLEW_EXT_blend_equation_separate) = !_glewInit_GL_EXT_blend_equation_separate(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_blend_equation_separate */
+#ifdef GL_EXT_blend_func_separate
+  CONST_CAST(GLEW_EXT_blend_func_separate) = glewGetExtension("GL_EXT_blend_func_separate");
+  if (glewExperimental || GLEW_EXT_blend_func_separate) CONST_CAST(GLEW_EXT_blend_func_separate) = !_glewInit_GL_EXT_blend_func_separate(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_blend_func_separate */
+#ifdef GL_EXT_blend_logic_op
+  CONST_CAST(GLEW_EXT_blend_logic_op) = glewGetExtension("GL_EXT_blend_logic_op");
+#endif /* GL_EXT_blend_logic_op */
+#ifdef GL_EXT_blend_minmax
+  CONST_CAST(GLEW_EXT_blend_minmax) = glewGetExtension("GL_EXT_blend_minmax");
+  if (glewExperimental || GLEW_EXT_blend_minmax) CONST_CAST(GLEW_EXT_blend_minmax) = !_glewInit_GL_EXT_blend_minmax(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_blend_minmax */
+#ifdef GL_EXT_blend_subtract
+  CONST_CAST(GLEW_EXT_blend_subtract) = glewGetExtension("GL_EXT_blend_subtract");
+#endif /* GL_EXT_blend_subtract */
+#ifdef GL_EXT_clip_volume_hint
+  CONST_CAST(GLEW_EXT_clip_volume_hint) = glewGetExtension("GL_EXT_clip_volume_hint");
+#endif /* GL_EXT_clip_volume_hint */
+#ifdef GL_EXT_cmyka
+  CONST_CAST(GLEW_EXT_cmyka) = glewGetExtension("GL_EXT_cmyka");
+#endif /* GL_EXT_cmyka */
+#ifdef GL_EXT_color_subtable
+  CONST_CAST(GLEW_EXT_color_subtable) = glewGetExtension("GL_EXT_color_subtable");
+  if (glewExperimental || GLEW_EXT_color_subtable) CONST_CAST(GLEW_EXT_color_subtable) = !_glewInit_GL_EXT_color_subtable(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_color_subtable */
+#ifdef GL_EXT_compiled_vertex_array
+  CONST_CAST(GLEW_EXT_compiled_vertex_array) = glewGetExtension("GL_EXT_compiled_vertex_array");
+  if (glewExperimental || GLEW_EXT_compiled_vertex_array) CONST_CAST(GLEW_EXT_compiled_vertex_array) = !_glewInit_GL_EXT_compiled_vertex_array(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_compiled_vertex_array */
+#ifdef GL_EXT_convolution
+  CONST_CAST(GLEW_EXT_convolution) = glewGetExtension("GL_EXT_convolution");
+  if (glewExperimental || GLEW_EXT_convolution) CONST_CAST(GLEW_EXT_convolution) = !_glewInit_GL_EXT_convolution(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_convolution */
+#ifdef GL_EXT_coordinate_frame
+  CONST_CAST(GLEW_EXT_coordinate_frame) = glewGetExtension("GL_EXT_coordinate_frame");
+  if (glewExperimental || GLEW_EXT_coordinate_frame) CONST_CAST(GLEW_EXT_coordinate_frame) = !_glewInit_GL_EXT_coordinate_frame(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_coordinate_frame */
+#ifdef GL_EXT_copy_texture
+  CONST_CAST(GLEW_EXT_copy_texture) = glewGetExtension("GL_EXT_copy_texture");
+  if (glewExperimental || GLEW_EXT_copy_texture) CONST_CAST(GLEW_EXT_copy_texture) = !_glewInit_GL_EXT_copy_texture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_copy_texture */
+#ifdef GL_EXT_cull_vertex
+  CONST_CAST(GLEW_EXT_cull_vertex) = glewGetExtension("GL_EXT_cull_vertex");
+  if (glewExperimental || GLEW_EXT_cull_vertex) CONST_CAST(GLEW_EXT_cull_vertex) = !_glewInit_GL_EXT_cull_vertex(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_cull_vertex */
+#ifdef GL_EXT_depth_bounds_test
+  CONST_CAST(GLEW_EXT_depth_bounds_test) = glewGetExtension("GL_EXT_depth_bounds_test");
+  if (glewExperimental || GLEW_EXT_depth_bounds_test) CONST_CAST(GLEW_EXT_depth_bounds_test) = !_glewInit_GL_EXT_depth_bounds_test(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_depth_bounds_test */
+#ifdef GL_EXT_direct_state_access
+  CONST_CAST(GLEW_EXT_direct_state_access) = glewGetExtension("GL_EXT_direct_state_access");
+  if (glewExperimental || GLEW_EXT_direct_state_access) CONST_CAST(GLEW_EXT_direct_state_access) = !_glewInit_GL_EXT_direct_state_access(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_direct_state_access */
+#ifdef GL_EXT_draw_buffers2
+  CONST_CAST(GLEW_EXT_draw_buffers2) = glewGetExtension("GL_EXT_draw_buffers2");
+  if (glewExperimental || GLEW_EXT_draw_buffers2) CONST_CAST(GLEW_EXT_draw_buffers2) = !_glewInit_GL_EXT_draw_buffers2(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_draw_buffers2 */
+#ifdef GL_EXT_draw_instanced
+  CONST_CAST(GLEW_EXT_draw_instanced) = glewGetExtension("GL_EXT_draw_instanced");
+  if (glewExperimental || GLEW_EXT_draw_instanced) CONST_CAST(GLEW_EXT_draw_instanced) = !_glewInit_GL_EXT_draw_instanced(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_draw_instanced */
+#ifdef GL_EXT_draw_range_elements
+  CONST_CAST(GLEW_EXT_draw_range_elements) = glewGetExtension("GL_EXT_draw_range_elements");
+  if (glewExperimental || GLEW_EXT_draw_range_elements) CONST_CAST(GLEW_EXT_draw_range_elements) = !_glewInit_GL_EXT_draw_range_elements(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_draw_range_elements */
+#ifdef GL_EXT_fog_coord
+  CONST_CAST(GLEW_EXT_fog_coord) = glewGetExtension("GL_EXT_fog_coord");
+  if (glewExperimental || GLEW_EXT_fog_coord) CONST_CAST(GLEW_EXT_fog_coord) = !_glewInit_GL_EXT_fog_coord(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_fog_coord */
+#ifdef GL_EXT_fragment_lighting
+  CONST_CAST(GLEW_EXT_fragment_lighting) = glewGetExtension("GL_EXT_fragment_lighting");
+  if (glewExperimental || GLEW_EXT_fragment_lighting) CONST_CAST(GLEW_EXT_fragment_lighting) = !_glewInit_GL_EXT_fragment_lighting(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_fragment_lighting */
+#ifdef GL_EXT_framebuffer_blit
+  CONST_CAST(GLEW_EXT_framebuffer_blit) = glewGetExtension("GL_EXT_framebuffer_blit");
+  if (glewExperimental || GLEW_EXT_framebuffer_blit) CONST_CAST(GLEW_EXT_framebuffer_blit) = !_glewInit_GL_EXT_framebuffer_blit(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_framebuffer_blit */
+#ifdef GL_EXT_framebuffer_multisample
+  CONST_CAST(GLEW_EXT_framebuffer_multisample) = glewGetExtension("GL_EXT_framebuffer_multisample");
+  if (glewExperimental || GLEW_EXT_framebuffer_multisample) CONST_CAST(GLEW_EXT_framebuffer_multisample) = !_glewInit_GL_EXT_framebuffer_multisample(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_framebuffer_multisample */
+#ifdef GL_EXT_framebuffer_object
+  CONST_CAST(GLEW_EXT_framebuffer_object) = glewGetExtension("GL_EXT_framebuffer_object");
+  if (glewExperimental || GLEW_EXT_framebuffer_object) CONST_CAST(GLEW_EXT_framebuffer_object) = !_glewInit_GL_EXT_framebuffer_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_framebuffer_object */
+#ifdef GL_EXT_framebuffer_sRGB
+  CONST_CAST(GLEW_EXT_framebuffer_sRGB) = glewGetExtension("GL_EXT_framebuffer_sRGB");
+#endif /* GL_EXT_framebuffer_sRGB */
+#ifdef GL_EXT_geometry_shader4
+  CONST_CAST(GLEW_EXT_geometry_shader4) = glewGetExtension("GL_EXT_geometry_shader4");
+  if (glewExperimental || GLEW_EXT_geometry_shader4) CONST_CAST(GLEW_EXT_geometry_shader4) = !_glewInit_GL_EXT_geometry_shader4(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_geometry_shader4 */
+#ifdef GL_EXT_gpu_program_parameters
+  CONST_CAST(GLEW_EXT_gpu_program_parameters) = glewGetExtension("GL_EXT_gpu_program_parameters");
+  if (glewExperimental || GLEW_EXT_gpu_program_parameters) CONST_CAST(GLEW_EXT_gpu_program_parameters) = !_glewInit_GL_EXT_gpu_program_parameters(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_gpu_program_parameters */
+#ifdef GL_EXT_gpu_shader4
+  CONST_CAST(GLEW_EXT_gpu_shader4) = glewGetExtension("GL_EXT_gpu_shader4");
+  if (glewExperimental || GLEW_EXT_gpu_shader4) CONST_CAST(GLEW_EXT_gpu_shader4) = !_glewInit_GL_EXT_gpu_shader4(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_gpu_shader4 */
+#ifdef GL_EXT_histogram
+  CONST_CAST(GLEW_EXT_histogram) = glewGetExtension("GL_EXT_histogram");
+  if (glewExperimental || GLEW_EXT_histogram) CONST_CAST(GLEW_EXT_histogram) = !_glewInit_GL_EXT_histogram(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_histogram */
+#ifdef GL_EXT_index_array_formats
+  CONST_CAST(GLEW_EXT_index_array_formats) = glewGetExtension("GL_EXT_index_array_formats");
+#endif /* GL_EXT_index_array_formats */
+#ifdef GL_EXT_index_func
+  CONST_CAST(GLEW_EXT_index_func) = glewGetExtension("GL_EXT_index_func");
+  if (glewExperimental || GLEW_EXT_index_func) CONST_CAST(GLEW_EXT_index_func) = !_glewInit_GL_EXT_index_func(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_index_func */
+#ifdef GL_EXT_index_material
+  CONST_CAST(GLEW_EXT_index_material) = glewGetExtension("GL_EXT_index_material");
+  if (glewExperimental || GLEW_EXT_index_material) CONST_CAST(GLEW_EXT_index_material) = !_glewInit_GL_EXT_index_material(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_index_material */
+#ifdef GL_EXT_index_texture
+  CONST_CAST(GLEW_EXT_index_texture) = glewGetExtension("GL_EXT_index_texture");
+#endif /* GL_EXT_index_texture */
+#ifdef GL_EXT_light_texture
+  CONST_CAST(GLEW_EXT_light_texture) = glewGetExtension("GL_EXT_light_texture");
+  if (glewExperimental || GLEW_EXT_light_texture) CONST_CAST(GLEW_EXT_light_texture) = !_glewInit_GL_EXT_light_texture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_light_texture */
+#ifdef GL_EXT_misc_attribute
+  CONST_CAST(GLEW_EXT_misc_attribute) = glewGetExtension("GL_EXT_misc_attribute");
+#endif /* GL_EXT_misc_attribute */
+#ifdef GL_EXT_multi_draw_arrays
+  CONST_CAST(GLEW_EXT_multi_draw_arrays) = glewGetExtension("GL_EXT_multi_draw_arrays");
+  if (glewExperimental || GLEW_EXT_multi_draw_arrays) CONST_CAST(GLEW_EXT_multi_draw_arrays) = !_glewInit_GL_EXT_multi_draw_arrays(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_multi_draw_arrays */
+#ifdef GL_EXT_multisample
+  CONST_CAST(GLEW_EXT_multisample) = glewGetExtension("GL_EXT_multisample");
+  if (glewExperimental || GLEW_EXT_multisample) CONST_CAST(GLEW_EXT_multisample) = !_glewInit_GL_EXT_multisample(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_multisample */
+#ifdef GL_EXT_packed_depth_stencil
+  CONST_CAST(GLEW_EXT_packed_depth_stencil) = glewGetExtension("GL_EXT_packed_depth_stencil");
+#endif /* GL_EXT_packed_depth_stencil */
+#ifdef GL_EXT_packed_float
+  CONST_CAST(GLEW_EXT_packed_float) = glewGetExtension("GL_EXT_packed_float");
+#endif /* GL_EXT_packed_float */
+#ifdef GL_EXT_packed_pixels
+  CONST_CAST(GLEW_EXT_packed_pixels) = glewGetExtension("GL_EXT_packed_pixels");
+#endif /* GL_EXT_packed_pixels */
+#ifdef GL_EXT_paletted_texture
+  CONST_CAST(GLEW_EXT_paletted_texture) = glewGetExtension("GL_EXT_paletted_texture");
+  if (glewExperimental || GLEW_EXT_paletted_texture) CONST_CAST(GLEW_EXT_paletted_texture) = !_glewInit_GL_EXT_paletted_texture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_paletted_texture */
+#ifdef GL_EXT_pixel_buffer_object
+  CONST_CAST(GLEW_EXT_pixel_buffer_object) = glewGetExtension("GL_EXT_pixel_buffer_object");
+#endif /* GL_EXT_pixel_buffer_object */
+#ifdef GL_EXT_pixel_transform
+  CONST_CAST(GLEW_EXT_pixel_transform) = glewGetExtension("GL_EXT_pixel_transform");
+  if (glewExperimental || GLEW_EXT_pixel_transform) CONST_CAST(GLEW_EXT_pixel_transform) = !_glewInit_GL_EXT_pixel_transform(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_pixel_transform */
+#ifdef GL_EXT_pixel_transform_color_table
+  CONST_CAST(GLEW_EXT_pixel_transform_color_table) = glewGetExtension("GL_EXT_pixel_transform_color_table");
+#endif /* GL_EXT_pixel_transform_color_table */
+#ifdef GL_EXT_point_parameters
+  CONST_CAST(GLEW_EXT_point_parameters) = glewGetExtension("GL_EXT_point_parameters");
+  if (glewExperimental || GLEW_EXT_point_parameters) CONST_CAST(GLEW_EXT_point_parameters) = !_glewInit_GL_EXT_point_parameters(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_point_parameters */
+#ifdef GL_EXT_polygon_offset
+  CONST_CAST(GLEW_EXT_polygon_offset) = glewGetExtension("GL_EXT_polygon_offset");
+  if (glewExperimental || GLEW_EXT_polygon_offset) CONST_CAST(GLEW_EXT_polygon_offset) = !_glewInit_GL_EXT_polygon_offset(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_polygon_offset */
+#ifdef GL_EXT_rescale_normal
+  CONST_CAST(GLEW_EXT_rescale_normal) = glewGetExtension("GL_EXT_rescale_normal");
+#endif /* GL_EXT_rescale_normal */
+#ifdef GL_EXT_scene_marker
+  CONST_CAST(GLEW_EXT_scene_marker) = glewGetExtension("GL_EXT_scene_marker");
+  if (glewExperimental || GLEW_EXT_scene_marker) CONST_CAST(GLEW_EXT_scene_marker) = !_glewInit_GL_EXT_scene_marker(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_scene_marker */
+#ifdef GL_EXT_secondary_color
+  CONST_CAST(GLEW_EXT_secondary_color) = glewGetExtension("GL_EXT_secondary_color");
+  if (glewExperimental || GLEW_EXT_secondary_color) CONST_CAST(GLEW_EXT_secondary_color) = !_glewInit_GL_EXT_secondary_color(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_secondary_color */
+#ifdef GL_EXT_separate_specular_color
+  CONST_CAST(GLEW_EXT_separate_specular_color) = glewGetExtension("GL_EXT_separate_specular_color");
+#endif /* GL_EXT_separate_specular_color */
+#ifdef GL_EXT_shadow_funcs
+  CONST_CAST(GLEW_EXT_shadow_funcs) = glewGetExtension("GL_EXT_shadow_funcs");
+#endif /* GL_EXT_shadow_funcs */
+#ifdef GL_EXT_shared_texture_palette
+  CONST_CAST(GLEW_EXT_shared_texture_palette) = glewGetExtension("GL_EXT_shared_texture_palette");
+#endif /* GL_EXT_shared_texture_palette */
+#ifdef GL_EXT_stencil_clear_tag
+  CONST_CAST(GLEW_EXT_stencil_clear_tag) = glewGetExtension("GL_EXT_stencil_clear_tag");
+#endif /* GL_EXT_stencil_clear_tag */
+#ifdef GL_EXT_stencil_two_side
+  CONST_CAST(GLEW_EXT_stencil_two_side) = glewGetExtension("GL_EXT_stencil_two_side");
+  if (glewExperimental || GLEW_EXT_stencil_two_side) CONST_CAST(GLEW_EXT_stencil_two_side) = !_glewInit_GL_EXT_stencil_two_side(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_stencil_two_side */
+#ifdef GL_EXT_stencil_wrap
+  CONST_CAST(GLEW_EXT_stencil_wrap) = glewGetExtension("GL_EXT_stencil_wrap");
+#endif /* GL_EXT_stencil_wrap */
+#ifdef GL_EXT_subtexture
+  CONST_CAST(GLEW_EXT_subtexture) = glewGetExtension("GL_EXT_subtexture");
+  if (glewExperimental || GLEW_EXT_subtexture) CONST_CAST(GLEW_EXT_subtexture) = !_glewInit_GL_EXT_subtexture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_subtexture */
+#ifdef GL_EXT_texture
+  CONST_CAST(GLEW_EXT_texture) = glewGetExtension("GL_EXT_texture");
+#endif /* GL_EXT_texture */
+#ifdef GL_EXT_texture3D
+  CONST_CAST(GLEW_EXT_texture3D) = glewGetExtension("GL_EXT_texture3D");
+  if (glewExperimental || GLEW_EXT_texture3D) CONST_CAST(GLEW_EXT_texture3D) = !_glewInit_GL_EXT_texture3D(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_texture3D */
+#ifdef GL_EXT_texture_array
+  CONST_CAST(GLEW_EXT_texture_array) = glewGetExtension("GL_EXT_texture_array");
+#endif /* GL_EXT_texture_array */
+#ifdef GL_EXT_texture_buffer_object
+  CONST_CAST(GLEW_EXT_texture_buffer_object) = glewGetExtension("GL_EXT_texture_buffer_object");
+  if (glewExperimental || GLEW_EXT_texture_buffer_object) CONST_CAST(GLEW_EXT_texture_buffer_object) = !_glewInit_GL_EXT_texture_buffer_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_texture_buffer_object */
+#ifdef GL_EXT_texture_compression_dxt1
+  CONST_CAST(GLEW_EXT_texture_compression_dxt1) = glewGetExtension("GL_EXT_texture_compression_dxt1");
+#endif /* GL_EXT_texture_compression_dxt1 */
+#ifdef GL_EXT_texture_compression_latc
+  CONST_CAST(GLEW_EXT_texture_compression_latc) = glewGetExtension("GL_EXT_texture_compression_latc");
+#endif /* GL_EXT_texture_compression_latc */
+#ifdef GL_EXT_texture_compression_rgtc
+  CONST_CAST(GLEW_EXT_texture_compression_rgtc) = glewGetExtension("GL_EXT_texture_compression_rgtc");
+#endif /* GL_EXT_texture_compression_rgtc */
+#ifdef GL_EXT_texture_compression_s3tc
+  CONST_CAST(GLEW_EXT_texture_compression_s3tc) = glewGetExtension("GL_EXT_texture_compression_s3tc");
+#endif /* GL_EXT_texture_compression_s3tc */
+#ifdef GL_EXT_texture_cube_map
+  CONST_CAST(GLEW_EXT_texture_cube_map) = glewGetExtension("GL_EXT_texture_cube_map");
+#endif /* GL_EXT_texture_cube_map */
+#ifdef GL_EXT_texture_edge_clamp
+  CONST_CAST(GLEW_EXT_texture_edge_clamp) = glewGetExtension("GL_EXT_texture_edge_clamp");
+#endif /* GL_EXT_texture_edge_clamp */
+#ifdef GL_EXT_texture_env
+  CONST_CAST(GLEW_EXT_texture_env) = glewGetExtension("GL_EXT_texture_env");
+#endif /* GL_EXT_texture_env */
+#ifdef GL_EXT_texture_env_add
+  CONST_CAST(GLEW_EXT_texture_env_add) = glewGetExtension("GL_EXT_texture_env_add");
+#endif /* GL_EXT_texture_env_add */
+#ifdef GL_EXT_texture_env_combine
+  CONST_CAST(GLEW_EXT_texture_env_combine) = glewGetExtension("GL_EXT_texture_env_combine");
+#endif /* GL_EXT_texture_env_combine */
+#ifdef GL_EXT_texture_env_dot3
+  CONST_CAST(GLEW_EXT_texture_env_dot3) = glewGetExtension("GL_EXT_texture_env_dot3");
+#endif /* GL_EXT_texture_env_dot3 */
+#ifdef GL_EXT_texture_filter_anisotropic
+  CONST_CAST(GLEW_EXT_texture_filter_anisotropic) = glewGetExtension("GL_EXT_texture_filter_anisotropic");
+#endif /* GL_EXT_texture_filter_anisotropic */
+#ifdef GL_EXT_texture_integer
+  CONST_CAST(GLEW_EXT_texture_integer) = glewGetExtension("GL_EXT_texture_integer");
+  if (glewExperimental || GLEW_EXT_texture_integer) CONST_CAST(GLEW_EXT_texture_integer) = !_glewInit_GL_EXT_texture_integer(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_texture_integer */
+#ifdef GL_EXT_texture_lod_bias
+  CONST_CAST(GLEW_EXT_texture_lod_bias) = glewGetExtension("GL_EXT_texture_lod_bias");
+#endif /* GL_EXT_texture_lod_bias */
+#ifdef GL_EXT_texture_mirror_clamp
+  CONST_CAST(GLEW_EXT_texture_mirror_clamp) = glewGetExtension("GL_EXT_texture_mirror_clamp");
+#endif /* GL_EXT_texture_mirror_clamp */
+#ifdef GL_EXT_texture_object
+  CONST_CAST(GLEW_EXT_texture_object) = glewGetExtension("GL_EXT_texture_object");
+  if (glewExperimental || GLEW_EXT_texture_object) CONST_CAST(GLEW_EXT_texture_object) = !_glewInit_GL_EXT_texture_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_texture_object */
+#ifdef GL_EXT_texture_perturb_normal
+  CONST_CAST(GLEW_EXT_texture_perturb_normal) = glewGetExtension("GL_EXT_texture_perturb_normal");
+  if (glewExperimental || GLEW_EXT_texture_perturb_normal) CONST_CAST(GLEW_EXT_texture_perturb_normal) = !_glewInit_GL_EXT_texture_perturb_normal(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_texture_perturb_normal */
+#ifdef GL_EXT_texture_rectangle
+  CONST_CAST(GLEW_EXT_texture_rectangle) = glewGetExtension("GL_EXT_texture_rectangle");
+#endif /* GL_EXT_texture_rectangle */
+#ifdef GL_EXT_texture_sRGB
+  CONST_CAST(GLEW_EXT_texture_sRGB) = glewGetExtension("GL_EXT_texture_sRGB");
+#endif /* GL_EXT_texture_sRGB */
+#ifdef GL_EXT_texture_shared_exponent
+  CONST_CAST(GLEW_EXT_texture_shared_exponent) = glewGetExtension("GL_EXT_texture_shared_exponent");
+#endif /* GL_EXT_texture_shared_exponent */
+#ifdef GL_EXT_texture_swizzle
+  CONST_CAST(GLEW_EXT_texture_swizzle) = glewGetExtension("GL_EXT_texture_swizzle");
+#endif /* GL_EXT_texture_swizzle */
+#ifdef GL_EXT_timer_query
+  CONST_CAST(GLEW_EXT_timer_query) = glewGetExtension("GL_EXT_timer_query");
+  if (glewExperimental || GLEW_EXT_timer_query) CONST_CAST(GLEW_EXT_timer_query) = !_glewInit_GL_EXT_timer_query(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_timer_query */
+#ifdef GL_EXT_transform_feedback
+  CONST_CAST(GLEW_EXT_transform_feedback) = glewGetExtension("GL_EXT_transform_feedback");
+  if (glewExperimental || GLEW_EXT_transform_feedback) CONST_CAST(GLEW_EXT_transform_feedback) = !_glewInit_GL_EXT_transform_feedback(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_transform_feedback */
+#ifdef GL_EXT_vertex_array
+  CONST_CAST(GLEW_EXT_vertex_array) = glewGetExtension("GL_EXT_vertex_array");
+  if (glewExperimental || GLEW_EXT_vertex_array) CONST_CAST(GLEW_EXT_vertex_array) = !_glewInit_GL_EXT_vertex_array(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_vertex_array */
+#ifdef GL_EXT_vertex_array_bgra
+  CONST_CAST(GLEW_EXT_vertex_array_bgra) = glewGetExtension("GL_EXT_vertex_array_bgra");
+#endif /* GL_EXT_vertex_array_bgra */
+#ifdef GL_EXT_vertex_shader
+  CONST_CAST(GLEW_EXT_vertex_shader) = glewGetExtension("GL_EXT_vertex_shader");
+  if (glewExperimental || GLEW_EXT_vertex_shader) CONST_CAST(GLEW_EXT_vertex_shader) = !_glewInit_GL_EXT_vertex_shader(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_vertex_shader */
+#ifdef GL_EXT_vertex_weighting
+  CONST_CAST(GLEW_EXT_vertex_weighting) = glewGetExtension("GL_EXT_vertex_weighting");
+  if (glewExperimental || GLEW_EXT_vertex_weighting) CONST_CAST(GLEW_EXT_vertex_weighting) = !_glewInit_GL_EXT_vertex_weighting(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_EXT_vertex_weighting */
+#ifdef GL_GREMEDY_frame_terminator
+  CONST_CAST(GLEW_GREMEDY_frame_terminator) = glewGetExtension("GL_GREMEDY_frame_terminator");
+  if (glewExperimental || GLEW_GREMEDY_frame_terminator) CONST_CAST(GLEW_GREMEDY_frame_terminator) = !_glewInit_GL_GREMEDY_frame_terminator(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_GREMEDY_frame_terminator */
+#ifdef GL_GREMEDY_string_marker
+  CONST_CAST(GLEW_GREMEDY_string_marker) = glewGetExtension("GL_GREMEDY_string_marker");
+  if (glewExperimental || GLEW_GREMEDY_string_marker) CONST_CAST(GLEW_GREMEDY_string_marker) = !_glewInit_GL_GREMEDY_string_marker(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_GREMEDY_string_marker */
+#ifdef GL_HP_convolution_border_modes
+  CONST_CAST(GLEW_HP_convolution_border_modes) = glewGetExtension("GL_HP_convolution_border_modes");
+#endif /* GL_HP_convolution_border_modes */
+#ifdef GL_HP_image_transform
+  CONST_CAST(GLEW_HP_image_transform) = glewGetExtension("GL_HP_image_transform");
+  if (glewExperimental || GLEW_HP_image_transform) CONST_CAST(GLEW_HP_image_transform) = !_glewInit_GL_HP_image_transform(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_HP_image_transform */
+#ifdef GL_HP_occlusion_test
+  CONST_CAST(GLEW_HP_occlusion_test) = glewGetExtension("GL_HP_occlusion_test");
+#endif /* GL_HP_occlusion_test */
+#ifdef GL_HP_texture_lighting
+  CONST_CAST(GLEW_HP_texture_lighting) = glewGetExtension("GL_HP_texture_lighting");
+#endif /* GL_HP_texture_lighting */
+#ifdef GL_IBM_cull_vertex
+  CONST_CAST(GLEW_IBM_cull_vertex) = glewGetExtension("GL_IBM_cull_vertex");
+#endif /* GL_IBM_cull_vertex */
+#ifdef GL_IBM_multimode_draw_arrays
+  CONST_CAST(GLEW_IBM_multimode_draw_arrays) = glewGetExtension("GL_IBM_multimode_draw_arrays");
+  if (glewExperimental || GLEW_IBM_multimode_draw_arrays) CONST_CAST(GLEW_IBM_multimode_draw_arrays) = !_glewInit_GL_IBM_multimode_draw_arrays(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_IBM_multimode_draw_arrays */
+#ifdef GL_IBM_rasterpos_clip
+  CONST_CAST(GLEW_IBM_rasterpos_clip) = glewGetExtension("GL_IBM_rasterpos_clip");
+#endif /* GL_IBM_rasterpos_clip */
+#ifdef GL_IBM_static_data
+  CONST_CAST(GLEW_IBM_static_data) = glewGetExtension("GL_IBM_static_data");
+#endif /* GL_IBM_static_data */
+#ifdef GL_IBM_texture_mirrored_repeat
+  CONST_CAST(GLEW_IBM_texture_mirrored_repeat) = glewGetExtension("GL_IBM_texture_mirrored_repeat");
+#endif /* GL_IBM_texture_mirrored_repeat */
+#ifdef GL_IBM_vertex_array_lists
+  CONST_CAST(GLEW_IBM_vertex_array_lists) = glewGetExtension("GL_IBM_vertex_array_lists");
+  if (glewExperimental || GLEW_IBM_vertex_array_lists) CONST_CAST(GLEW_IBM_vertex_array_lists) = !_glewInit_GL_IBM_vertex_array_lists(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_IBM_vertex_array_lists */
+#ifdef GL_INGR_color_clamp
+  CONST_CAST(GLEW_INGR_color_clamp) = glewGetExtension("GL_INGR_color_clamp");
+#endif /* GL_INGR_color_clamp */
+#ifdef GL_INGR_interlace_read
+  CONST_CAST(GLEW_INGR_interlace_read) = glewGetExtension("GL_INGR_interlace_read");
+#endif /* GL_INGR_interlace_read */
+#ifdef GL_INTEL_parallel_arrays
+  CONST_CAST(GLEW_INTEL_parallel_arrays) = glewGetExtension("GL_INTEL_parallel_arrays");
+  if (glewExperimental || GLEW_INTEL_parallel_arrays) CONST_CAST(GLEW_INTEL_parallel_arrays) = !_glewInit_GL_INTEL_parallel_arrays(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_INTEL_parallel_arrays */
+#ifdef GL_INTEL_texture_scissor
+  CONST_CAST(GLEW_INTEL_texture_scissor) = glewGetExtension("GL_INTEL_texture_scissor");
+  if (glewExperimental || GLEW_INTEL_texture_scissor) CONST_CAST(GLEW_INTEL_texture_scissor) = !_glewInit_GL_INTEL_texture_scissor(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_INTEL_texture_scissor */
+#ifdef GL_KTX_buffer_region
+  CONST_CAST(GLEW_KTX_buffer_region) = glewGetExtension("GL_KTX_buffer_region");
+  if (glewExperimental || GLEW_KTX_buffer_region) CONST_CAST(GLEW_KTX_buffer_region) = !_glewInit_GL_KTX_buffer_region(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_KTX_buffer_region */
+#ifdef GL_MESAX_texture_stack
+  CONST_CAST(GLEW_MESAX_texture_stack) = glewGetExtension("GL_MESAX_texture_stack");
+#endif /* GL_MESAX_texture_stack */
+#ifdef GL_MESA_pack_invert
+  CONST_CAST(GLEW_MESA_pack_invert) = glewGetExtension("GL_MESA_pack_invert");
+#endif /* GL_MESA_pack_invert */
+#ifdef GL_MESA_resize_buffers
+  CONST_CAST(GLEW_MESA_resize_buffers) = glewGetExtension("GL_MESA_resize_buffers");
+  if (glewExperimental || GLEW_MESA_resize_buffers) CONST_CAST(GLEW_MESA_resize_buffers) = !_glewInit_GL_MESA_resize_buffers(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_MESA_resize_buffers */
+#ifdef GL_MESA_window_pos
+  CONST_CAST(GLEW_MESA_window_pos) = glewGetExtension("GL_MESA_window_pos");
+  if (glewExperimental || GLEW_MESA_window_pos) CONST_CAST(GLEW_MESA_window_pos) = !_glewInit_GL_MESA_window_pos(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_MESA_window_pos */
+#ifdef GL_MESA_ycbcr_texture
+  CONST_CAST(GLEW_MESA_ycbcr_texture) = glewGetExtension("GL_MESA_ycbcr_texture");
+#endif /* GL_MESA_ycbcr_texture */
+#ifdef GL_NV_blend_square
+  CONST_CAST(GLEW_NV_blend_square) = glewGetExtension("GL_NV_blend_square");
+#endif /* GL_NV_blend_square */
+#ifdef GL_NV_conditional_render
+  CONST_CAST(GLEW_NV_conditional_render) = glewGetExtension("GL_NV_conditional_render");
+  if (glewExperimental || GLEW_NV_conditional_render) CONST_CAST(GLEW_NV_conditional_render) = !_glewInit_GL_NV_conditional_render(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_conditional_render */
+#ifdef GL_NV_copy_depth_to_color
+  CONST_CAST(GLEW_NV_copy_depth_to_color) = glewGetExtension("GL_NV_copy_depth_to_color");
+#endif /* GL_NV_copy_depth_to_color */
+#ifdef GL_NV_depth_buffer_float
+  CONST_CAST(GLEW_NV_depth_buffer_float) = glewGetExtension("GL_NV_depth_buffer_float");
+  if (glewExperimental || GLEW_NV_depth_buffer_float) CONST_CAST(GLEW_NV_depth_buffer_float) = !_glewInit_GL_NV_depth_buffer_float(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_depth_buffer_float */
+#ifdef GL_NV_depth_clamp
+  CONST_CAST(GLEW_NV_depth_clamp) = glewGetExtension("GL_NV_depth_clamp");
+#endif /* GL_NV_depth_clamp */
+#ifdef GL_NV_depth_range_unclamped
+  CONST_CAST(GLEW_NV_depth_range_unclamped) = glewGetExtension("GL_NV_depth_range_unclamped");
+#endif /* GL_NV_depth_range_unclamped */
+#ifdef GL_NV_evaluators
+  CONST_CAST(GLEW_NV_evaluators) = glewGetExtension("GL_NV_evaluators");
+  if (glewExperimental || GLEW_NV_evaluators) CONST_CAST(GLEW_NV_evaluators) = !_glewInit_GL_NV_evaluators(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_evaluators */
+#ifdef GL_NV_explicit_multisample
+  CONST_CAST(GLEW_NV_explicit_multisample) = glewGetExtension("GL_NV_explicit_multisample");
+  if (glewExperimental || GLEW_NV_explicit_multisample) CONST_CAST(GLEW_NV_explicit_multisample) = !_glewInit_GL_NV_explicit_multisample(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_explicit_multisample */
+#ifdef GL_NV_fence
+  CONST_CAST(GLEW_NV_fence) = glewGetExtension("GL_NV_fence");
+  if (glewExperimental || GLEW_NV_fence) CONST_CAST(GLEW_NV_fence) = !_glewInit_GL_NV_fence(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_fence */
+#ifdef GL_NV_float_buffer
+  CONST_CAST(GLEW_NV_float_buffer) = glewGetExtension("GL_NV_float_buffer");
+#endif /* GL_NV_float_buffer */
+#ifdef GL_NV_fog_distance
+  CONST_CAST(GLEW_NV_fog_distance) = glewGetExtension("GL_NV_fog_distance");
+#endif /* GL_NV_fog_distance */
+#ifdef GL_NV_fragment_program
+  CONST_CAST(GLEW_NV_fragment_program) = glewGetExtension("GL_NV_fragment_program");
+  if (glewExperimental || GLEW_NV_fragment_program) CONST_CAST(GLEW_NV_fragment_program) = !_glewInit_GL_NV_fragment_program(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_fragment_program */
+#ifdef GL_NV_fragment_program2
+  CONST_CAST(GLEW_NV_fragment_program2) = glewGetExtension("GL_NV_fragment_program2");
+#endif /* GL_NV_fragment_program2 */
+#ifdef GL_NV_fragment_program4
+  CONST_CAST(GLEW_NV_fragment_program4) = glewGetExtension("GL_NV_fragment_program4");
+#endif /* GL_NV_fragment_program4 */
+#ifdef GL_NV_fragment_program_option
+  CONST_CAST(GLEW_NV_fragment_program_option) = glewGetExtension("GL_NV_fragment_program_option");
+#endif /* GL_NV_fragment_program_option */
+#ifdef GL_NV_framebuffer_multisample_coverage
+  CONST_CAST(GLEW_NV_framebuffer_multisample_coverage) = glewGetExtension("GL_NV_framebuffer_multisample_coverage");
+  if (glewExperimental || GLEW_NV_framebuffer_multisample_coverage) CONST_CAST(GLEW_NV_framebuffer_multisample_coverage) = !_glewInit_GL_NV_framebuffer_multisample_coverage(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_framebuffer_multisample_coverage */
+#ifdef GL_NV_geometry_program4
+  CONST_CAST(GLEW_NV_geometry_program4) = glewGetExtension("GL_NV_geometry_program4");
+  if (glewExperimental || GLEW_NV_geometry_program4) CONST_CAST(GLEW_NV_geometry_program4) = !_glewInit_GL_NV_geometry_program4(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_geometry_program4 */
+#ifdef GL_NV_geometry_shader4
+  CONST_CAST(GLEW_NV_geometry_shader4) = glewGetExtension("GL_NV_geometry_shader4");
+#endif /* GL_NV_geometry_shader4 */
+#ifdef GL_NV_gpu_program4
+  CONST_CAST(GLEW_NV_gpu_program4) = glewGetExtension("GL_NV_gpu_program4");
+  if (glewExperimental || GLEW_NV_gpu_program4) CONST_CAST(GLEW_NV_gpu_program4) = !_glewInit_GL_NV_gpu_program4(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_gpu_program4 */
+#ifdef GL_NV_half_float
+  CONST_CAST(GLEW_NV_half_float) = glewGetExtension("GL_NV_half_float");
+  if (glewExperimental || GLEW_NV_half_float) CONST_CAST(GLEW_NV_half_float) = !_glewInit_GL_NV_half_float(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_half_float */
+#ifdef GL_NV_light_max_exponent
+  CONST_CAST(GLEW_NV_light_max_exponent) = glewGetExtension("GL_NV_light_max_exponent");
+#endif /* GL_NV_light_max_exponent */
+#ifdef GL_NV_multisample_filter_hint
+  CONST_CAST(GLEW_NV_multisample_filter_hint) = glewGetExtension("GL_NV_multisample_filter_hint");
+#endif /* GL_NV_multisample_filter_hint */
+#ifdef GL_NV_occlusion_query
+  CONST_CAST(GLEW_NV_occlusion_query) = glewGetExtension("GL_NV_occlusion_query");
+  if (glewExperimental || GLEW_NV_occlusion_query) CONST_CAST(GLEW_NV_occlusion_query) = !_glewInit_GL_NV_occlusion_query(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_occlusion_query */
+#ifdef GL_NV_packed_depth_stencil
+  CONST_CAST(GLEW_NV_packed_depth_stencil) = glewGetExtension("GL_NV_packed_depth_stencil");
+#endif /* GL_NV_packed_depth_stencil */
+#ifdef GL_NV_parameter_buffer_object
+  CONST_CAST(GLEW_NV_parameter_buffer_object) = glewGetExtension("GL_NV_parameter_buffer_object");
+  if (glewExperimental || GLEW_NV_parameter_buffer_object) CONST_CAST(GLEW_NV_parameter_buffer_object) = !_glewInit_GL_NV_parameter_buffer_object(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_parameter_buffer_object */
+#ifdef GL_NV_pixel_data_range
+  CONST_CAST(GLEW_NV_pixel_data_range) = glewGetExtension("GL_NV_pixel_data_range");
+  if (glewExperimental || GLEW_NV_pixel_data_range) CONST_CAST(GLEW_NV_pixel_data_range) = !_glewInit_GL_NV_pixel_data_range(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_pixel_data_range */
+#ifdef GL_NV_point_sprite
+  CONST_CAST(GLEW_NV_point_sprite) = glewGetExtension("GL_NV_point_sprite");
+  if (glewExperimental || GLEW_NV_point_sprite) CONST_CAST(GLEW_NV_point_sprite) = !_glewInit_GL_NV_point_sprite(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_point_sprite */
+#ifdef GL_NV_present_video
+  CONST_CAST(GLEW_NV_present_video) = glewGetExtension("GL_NV_present_video");
+  if (glewExperimental || GLEW_NV_present_video) CONST_CAST(GLEW_NV_present_video) = !_glewInit_GL_NV_present_video(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_present_video */
+#ifdef GL_NV_primitive_restart
+  CONST_CAST(GLEW_NV_primitive_restart) = glewGetExtension("GL_NV_primitive_restart");
+  if (glewExperimental || GLEW_NV_primitive_restart) CONST_CAST(GLEW_NV_primitive_restart) = !_glewInit_GL_NV_primitive_restart(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_primitive_restart */
+#ifdef GL_NV_register_combiners
+  CONST_CAST(GLEW_NV_register_combiners) = glewGetExtension("GL_NV_register_combiners");
+  if (glewExperimental || GLEW_NV_register_combiners) CONST_CAST(GLEW_NV_register_combiners) = !_glewInit_GL_NV_register_combiners(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_register_combiners */
+#ifdef GL_NV_register_combiners2
+  CONST_CAST(GLEW_NV_register_combiners2) = glewGetExtension("GL_NV_register_combiners2");
+  if (glewExperimental || GLEW_NV_register_combiners2) CONST_CAST(GLEW_NV_register_combiners2) = !_glewInit_GL_NV_register_combiners2(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_register_combiners2 */
+#ifdef GL_NV_texgen_emboss
+  CONST_CAST(GLEW_NV_texgen_emboss) = glewGetExtension("GL_NV_texgen_emboss");
+#endif /* GL_NV_texgen_emboss */
+#ifdef GL_NV_texgen_reflection
+  CONST_CAST(GLEW_NV_texgen_reflection) = glewGetExtension("GL_NV_texgen_reflection");
+#endif /* GL_NV_texgen_reflection */
+#ifdef GL_NV_texture_compression_vtc
+  CONST_CAST(GLEW_NV_texture_compression_vtc) = glewGetExtension("GL_NV_texture_compression_vtc");
+#endif /* GL_NV_texture_compression_vtc */
+#ifdef GL_NV_texture_env_combine4
+  CONST_CAST(GLEW_NV_texture_env_combine4) = glewGetExtension("GL_NV_texture_env_combine4");
+#endif /* GL_NV_texture_env_combine4 */
+#ifdef GL_NV_texture_expand_normal
+  CONST_CAST(GLEW_NV_texture_expand_normal) = glewGetExtension("GL_NV_texture_expand_normal");
+#endif /* GL_NV_texture_expand_normal */
+#ifdef GL_NV_texture_rectangle
+  CONST_CAST(GLEW_NV_texture_rectangle) = glewGetExtension("GL_NV_texture_rectangle");
+#endif /* GL_NV_texture_rectangle */
+#ifdef GL_NV_texture_shader
+  CONST_CAST(GLEW_NV_texture_shader) = glewGetExtension("GL_NV_texture_shader");
+#endif /* GL_NV_texture_shader */
+#ifdef GL_NV_texture_shader2
+  CONST_CAST(GLEW_NV_texture_shader2) = glewGetExtension("GL_NV_texture_shader2");
+#endif /* GL_NV_texture_shader2 */
+#ifdef GL_NV_texture_shader3
+  CONST_CAST(GLEW_NV_texture_shader3) = glewGetExtension("GL_NV_texture_shader3");
+#endif /* GL_NV_texture_shader3 */
+#ifdef GL_NV_transform_feedback
+  CONST_CAST(GLEW_NV_transform_feedback) = glewGetExtension("GL_NV_transform_feedback");
+  if (glewExperimental || GLEW_NV_transform_feedback) CONST_CAST(GLEW_NV_transform_feedback) = !_glewInit_GL_NV_transform_feedback(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_transform_feedback */
+#ifdef GL_NV_vertex_array_range
+  CONST_CAST(GLEW_NV_vertex_array_range) = glewGetExtension("GL_NV_vertex_array_range");
+  if (glewExperimental || GLEW_NV_vertex_array_range) CONST_CAST(GLEW_NV_vertex_array_range) = !_glewInit_GL_NV_vertex_array_range(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_vertex_array_range */
+#ifdef GL_NV_vertex_array_range2
+  CONST_CAST(GLEW_NV_vertex_array_range2) = glewGetExtension("GL_NV_vertex_array_range2");
+#endif /* GL_NV_vertex_array_range2 */
+#ifdef GL_NV_vertex_program
+  CONST_CAST(GLEW_NV_vertex_program) = glewGetExtension("GL_NV_vertex_program");
+  if (glewExperimental || GLEW_NV_vertex_program) CONST_CAST(GLEW_NV_vertex_program) = !_glewInit_GL_NV_vertex_program(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_NV_vertex_program */
+#ifdef GL_NV_vertex_program1_1
+  CONST_CAST(GLEW_NV_vertex_program1_1) = glewGetExtension("GL_NV_vertex_program1_1");
+#endif /* GL_NV_vertex_program1_1 */
+#ifdef GL_NV_vertex_program2
+  CONST_CAST(GLEW_NV_vertex_program2) = glewGetExtension("GL_NV_vertex_program2");
+#endif /* GL_NV_vertex_program2 */
+#ifdef GL_NV_vertex_program2_option
+  CONST_CAST(GLEW_NV_vertex_program2_option) = glewGetExtension("GL_NV_vertex_program2_option");
+#endif /* GL_NV_vertex_program2_option */
+#ifdef GL_NV_vertex_program3
+  CONST_CAST(GLEW_NV_vertex_program3) = glewGetExtension("GL_NV_vertex_program3");
+#endif /* GL_NV_vertex_program3 */
+#ifdef GL_NV_vertex_program4
+  CONST_CAST(GLEW_NV_vertex_program4) = glewGetExtension("GL_NV_vertex_program4");
+#endif /* GL_NV_vertex_program4 */
+#ifdef GL_OES_byte_coordinates
+  CONST_CAST(GLEW_OES_byte_coordinates) = glewGetExtension("GL_OES_byte_coordinates");
+#endif /* GL_OES_byte_coordinates */
+#ifdef GL_OES_compressed_paletted_texture
+  CONST_CAST(GLEW_OES_compressed_paletted_texture) = glewGetExtension("GL_OES_compressed_paletted_texture");
+#endif /* GL_OES_compressed_paletted_texture */
+#ifdef GL_OES_read_format
+  CONST_CAST(GLEW_OES_read_format) = glewGetExtension("GL_OES_read_format");
+#endif /* GL_OES_read_format */
+#ifdef GL_OES_single_precision
+  CONST_CAST(GLEW_OES_single_precision) = glewGetExtension("GL_OES_single_precision");
+  if (glewExperimental || GLEW_OES_single_precision) CONST_CAST(GLEW_OES_single_precision) = !_glewInit_GL_OES_single_precision(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_OES_single_precision */
+#ifdef GL_OML_interlace
+  CONST_CAST(GLEW_OML_interlace) = glewGetExtension("GL_OML_interlace");
+#endif /* GL_OML_interlace */
+#ifdef GL_OML_resample
+  CONST_CAST(GLEW_OML_resample) = glewGetExtension("GL_OML_resample");
+#endif /* GL_OML_resample */
+#ifdef GL_OML_subsample
+  CONST_CAST(GLEW_OML_subsample) = glewGetExtension("GL_OML_subsample");
+#endif /* GL_OML_subsample */
+#ifdef GL_PGI_misc_hints
+  CONST_CAST(GLEW_PGI_misc_hints) = glewGetExtension("GL_PGI_misc_hints");
+#endif /* GL_PGI_misc_hints */
+#ifdef GL_PGI_vertex_hints
+  CONST_CAST(GLEW_PGI_vertex_hints) = glewGetExtension("GL_PGI_vertex_hints");
+#endif /* GL_PGI_vertex_hints */
+#ifdef GL_REND_screen_coordinates
+  CONST_CAST(GLEW_REND_screen_coordinates) = glewGetExtension("GL_REND_screen_coordinates");
+#endif /* GL_REND_screen_coordinates */
+#ifdef GL_S3_s3tc
+  CONST_CAST(GLEW_S3_s3tc) = glewGetExtension("GL_S3_s3tc");
+#endif /* GL_S3_s3tc */
+#ifdef GL_SGIS_color_range
+  CONST_CAST(GLEW_SGIS_color_range) = glewGetExtension("GL_SGIS_color_range");
+#endif /* GL_SGIS_color_range */
+#ifdef GL_SGIS_detail_texture
+  CONST_CAST(GLEW_SGIS_detail_texture) = glewGetExtension("GL_SGIS_detail_texture");
+  if (glewExperimental || GLEW_SGIS_detail_texture) CONST_CAST(GLEW_SGIS_detail_texture) = !_glewInit_GL_SGIS_detail_texture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIS_detail_texture */
+#ifdef GL_SGIS_fog_function
+  CONST_CAST(GLEW_SGIS_fog_function) = glewGetExtension("GL_SGIS_fog_function");
+  if (glewExperimental || GLEW_SGIS_fog_function) CONST_CAST(GLEW_SGIS_fog_function) = !_glewInit_GL_SGIS_fog_function(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIS_fog_function */
+#ifdef GL_SGIS_generate_mipmap
+  CONST_CAST(GLEW_SGIS_generate_mipmap) = glewGetExtension("GL_SGIS_generate_mipmap");
+#endif /* GL_SGIS_generate_mipmap */
+#ifdef GL_SGIS_multisample
+  CONST_CAST(GLEW_SGIS_multisample) = glewGetExtension("GL_SGIS_multisample");
+  if (glewExperimental || GLEW_SGIS_multisample) CONST_CAST(GLEW_SGIS_multisample) = !_glewInit_GL_SGIS_multisample(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIS_multisample */
+#ifdef GL_SGIS_pixel_texture
+  CONST_CAST(GLEW_SGIS_pixel_texture) = glewGetExtension("GL_SGIS_pixel_texture");
+#endif /* GL_SGIS_pixel_texture */
+#ifdef GL_SGIS_point_line_texgen
+  CONST_CAST(GLEW_SGIS_point_line_texgen) = glewGetExtension("GL_SGIS_point_line_texgen");
+#endif /* GL_SGIS_point_line_texgen */
+#ifdef GL_SGIS_sharpen_texture
+  CONST_CAST(GLEW_SGIS_sharpen_texture) = glewGetExtension("GL_SGIS_sharpen_texture");
+  if (glewExperimental || GLEW_SGIS_sharpen_texture) CONST_CAST(GLEW_SGIS_sharpen_texture) = !_glewInit_GL_SGIS_sharpen_texture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIS_sharpen_texture */
+#ifdef GL_SGIS_texture4D
+  CONST_CAST(GLEW_SGIS_texture4D) = glewGetExtension("GL_SGIS_texture4D");
+  if (glewExperimental || GLEW_SGIS_texture4D) CONST_CAST(GLEW_SGIS_texture4D) = !_glewInit_GL_SGIS_texture4D(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIS_texture4D */
+#ifdef GL_SGIS_texture_border_clamp
+  CONST_CAST(GLEW_SGIS_texture_border_clamp) = glewGetExtension("GL_SGIS_texture_border_clamp");
+#endif /* GL_SGIS_texture_border_clamp */
+#ifdef GL_SGIS_texture_edge_clamp
+  CONST_CAST(GLEW_SGIS_texture_edge_clamp) = glewGetExtension("GL_SGIS_texture_edge_clamp");
+#endif /* GL_SGIS_texture_edge_clamp */
+#ifdef GL_SGIS_texture_filter4
+  CONST_CAST(GLEW_SGIS_texture_filter4) = glewGetExtension("GL_SGIS_texture_filter4");
+  if (glewExperimental || GLEW_SGIS_texture_filter4) CONST_CAST(GLEW_SGIS_texture_filter4) = !_glewInit_GL_SGIS_texture_filter4(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIS_texture_filter4 */
+#ifdef GL_SGIS_texture_lod
+  CONST_CAST(GLEW_SGIS_texture_lod) = glewGetExtension("GL_SGIS_texture_lod");
+#endif /* GL_SGIS_texture_lod */
+#ifdef GL_SGIS_texture_select
+  CONST_CAST(GLEW_SGIS_texture_select) = glewGetExtension("GL_SGIS_texture_select");
+#endif /* GL_SGIS_texture_select */
+#ifdef GL_SGIX_async
+  CONST_CAST(GLEW_SGIX_async) = glewGetExtension("GL_SGIX_async");
+  if (glewExperimental || GLEW_SGIX_async) CONST_CAST(GLEW_SGIX_async) = !_glewInit_GL_SGIX_async(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIX_async */
+#ifdef GL_SGIX_async_histogram
+  CONST_CAST(GLEW_SGIX_async_histogram) = glewGetExtension("GL_SGIX_async_histogram");
+#endif /* GL_SGIX_async_histogram */
+#ifdef GL_SGIX_async_pixel
+  CONST_CAST(GLEW_SGIX_async_pixel) = glewGetExtension("GL_SGIX_async_pixel");
+#endif /* GL_SGIX_async_pixel */
+#ifdef GL_SGIX_blend_alpha_minmax
+  CONST_CAST(GLEW_SGIX_blend_alpha_minmax) = glewGetExtension("GL_SGIX_blend_alpha_minmax");
+#endif /* GL_SGIX_blend_alpha_minmax */
+#ifdef GL_SGIX_clipmap
+  CONST_CAST(GLEW_SGIX_clipmap) = glewGetExtension("GL_SGIX_clipmap");
+#endif /* GL_SGIX_clipmap */
+#ifdef GL_SGIX_convolution_accuracy
+  CONST_CAST(GLEW_SGIX_convolution_accuracy) = glewGetExtension("GL_SGIX_convolution_accuracy");
+#endif /* GL_SGIX_convolution_accuracy */
+#ifdef GL_SGIX_depth_texture
+  CONST_CAST(GLEW_SGIX_depth_texture) = glewGetExtension("GL_SGIX_depth_texture");
+#endif /* GL_SGIX_depth_texture */
+#ifdef GL_SGIX_flush_raster
+  CONST_CAST(GLEW_SGIX_flush_raster) = glewGetExtension("GL_SGIX_flush_raster");
+  if (glewExperimental || GLEW_SGIX_flush_raster) CONST_CAST(GLEW_SGIX_flush_raster) = !_glewInit_GL_SGIX_flush_raster(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIX_flush_raster */
+#ifdef GL_SGIX_fog_offset
+  CONST_CAST(GLEW_SGIX_fog_offset) = glewGetExtension("GL_SGIX_fog_offset");
+#endif /* GL_SGIX_fog_offset */
+#ifdef GL_SGIX_fog_texture
+  CONST_CAST(GLEW_SGIX_fog_texture) = glewGetExtension("GL_SGIX_fog_texture");
+  if (glewExperimental || GLEW_SGIX_fog_texture) CONST_CAST(GLEW_SGIX_fog_texture) = !_glewInit_GL_SGIX_fog_texture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIX_fog_texture */
+#ifdef GL_SGIX_fragment_specular_lighting
+  CONST_CAST(GLEW_SGIX_fragment_specular_lighting) = glewGetExtension("GL_SGIX_fragment_specular_lighting");
+  if (glewExperimental || GLEW_SGIX_fragment_specular_lighting) CONST_CAST(GLEW_SGIX_fragment_specular_lighting) = !_glewInit_GL_SGIX_fragment_specular_lighting(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIX_fragment_specular_lighting */
+#ifdef GL_SGIX_framezoom
+  CONST_CAST(GLEW_SGIX_framezoom) = glewGetExtension("GL_SGIX_framezoom");
+  if (glewExperimental || GLEW_SGIX_framezoom) CONST_CAST(GLEW_SGIX_framezoom) = !_glewInit_GL_SGIX_framezoom(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIX_framezoom */
+#ifdef GL_SGIX_interlace
+  CONST_CAST(GLEW_SGIX_interlace) = glewGetExtension("GL_SGIX_interlace");
+#endif /* GL_SGIX_interlace */
+#ifdef GL_SGIX_ir_instrument1
+  CONST_CAST(GLEW_SGIX_ir_instrument1) = glewGetExtension("GL_SGIX_ir_instrument1");
+#endif /* GL_SGIX_ir_instrument1 */
+#ifdef GL_SGIX_list_priority
+  CONST_CAST(GLEW_SGIX_list_priority) = glewGetExtension("GL_SGIX_list_priority");
+#endif /* GL_SGIX_list_priority */
+#ifdef GL_SGIX_pixel_texture
+  CONST_CAST(GLEW_SGIX_pixel_texture) = glewGetExtension("GL_SGIX_pixel_texture");
+  if (glewExperimental || GLEW_SGIX_pixel_texture) CONST_CAST(GLEW_SGIX_pixel_texture) = !_glewInit_GL_SGIX_pixel_texture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIX_pixel_texture */
+#ifdef GL_SGIX_pixel_texture_bits
+  CONST_CAST(GLEW_SGIX_pixel_texture_bits) = glewGetExtension("GL_SGIX_pixel_texture_bits");
+#endif /* GL_SGIX_pixel_texture_bits */
+#ifdef GL_SGIX_reference_plane
+  CONST_CAST(GLEW_SGIX_reference_plane) = glewGetExtension("GL_SGIX_reference_plane");
+  if (glewExperimental || GLEW_SGIX_reference_plane) CONST_CAST(GLEW_SGIX_reference_plane) = !_glewInit_GL_SGIX_reference_plane(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIX_reference_plane */
+#ifdef GL_SGIX_resample
+  CONST_CAST(GLEW_SGIX_resample) = glewGetExtension("GL_SGIX_resample");
+#endif /* GL_SGIX_resample */
+#ifdef GL_SGIX_shadow
+  CONST_CAST(GLEW_SGIX_shadow) = glewGetExtension("GL_SGIX_shadow");
+#endif /* GL_SGIX_shadow */
+#ifdef GL_SGIX_shadow_ambient
+  CONST_CAST(GLEW_SGIX_shadow_ambient) = glewGetExtension("GL_SGIX_shadow_ambient");
+#endif /* GL_SGIX_shadow_ambient */
+#ifdef GL_SGIX_sprite
+  CONST_CAST(GLEW_SGIX_sprite) = glewGetExtension("GL_SGIX_sprite");
+  if (glewExperimental || GLEW_SGIX_sprite) CONST_CAST(GLEW_SGIX_sprite) = !_glewInit_GL_SGIX_sprite(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIX_sprite */
+#ifdef GL_SGIX_tag_sample_buffer
+  CONST_CAST(GLEW_SGIX_tag_sample_buffer) = glewGetExtension("GL_SGIX_tag_sample_buffer");
+  if (glewExperimental || GLEW_SGIX_tag_sample_buffer) CONST_CAST(GLEW_SGIX_tag_sample_buffer) = !_glewInit_GL_SGIX_tag_sample_buffer(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGIX_tag_sample_buffer */
+#ifdef GL_SGIX_texture_add_env
+  CONST_CAST(GLEW_SGIX_texture_add_env) = glewGetExtension("GL_SGIX_texture_add_env");
+#endif /* GL_SGIX_texture_add_env */
+#ifdef GL_SGIX_texture_coordinate_clamp
+  CONST_CAST(GLEW_SGIX_texture_coordinate_clamp) = glewGetExtension("GL_SGIX_texture_coordinate_clamp");
+#endif /* GL_SGIX_texture_coordinate_clamp */
+#ifdef GL_SGIX_texture_lod_bias
+  CONST_CAST(GLEW_SGIX_texture_lod_bias) = glewGetExtension("GL_SGIX_texture_lod_bias");
+#endif /* GL_SGIX_texture_lod_bias */
+#ifdef GL_SGIX_texture_multi_buffer
+  CONST_CAST(GLEW_SGIX_texture_multi_buffer) = glewGetExtension("GL_SGIX_texture_multi_buffer");
+#endif /* GL_SGIX_texture_multi_buffer */
+#ifdef GL_SGIX_texture_range
+  CONST_CAST(GLEW_SGIX_texture_range) = glewGetExtension("GL_SGIX_texture_range");
+#endif /* GL_SGIX_texture_range */
+#ifdef GL_SGIX_texture_scale_bias
+  CONST_CAST(GLEW_SGIX_texture_scale_bias) = glewGetExtension("GL_SGIX_texture_scale_bias");
+#endif /* GL_SGIX_texture_scale_bias */
+#ifdef GL_SGIX_vertex_preclip
+  CONST_CAST(GLEW_SGIX_vertex_preclip) = glewGetExtension("GL_SGIX_vertex_preclip");
+#endif /* GL_SGIX_vertex_preclip */
+#ifdef GL_SGIX_vertex_preclip_hint
+  CONST_CAST(GLEW_SGIX_vertex_preclip_hint) = glewGetExtension("GL_SGIX_vertex_preclip_hint");
+#endif /* GL_SGIX_vertex_preclip_hint */
+#ifdef GL_SGIX_ycrcb
+  CONST_CAST(GLEW_SGIX_ycrcb) = glewGetExtension("GL_SGIX_ycrcb");
+#endif /* GL_SGIX_ycrcb */
+#ifdef GL_SGI_color_matrix
+  CONST_CAST(GLEW_SGI_color_matrix) = glewGetExtension("GL_SGI_color_matrix");
+#endif /* GL_SGI_color_matrix */
+#ifdef GL_SGI_color_table
+  CONST_CAST(GLEW_SGI_color_table) = glewGetExtension("GL_SGI_color_table");
+  if (glewExperimental || GLEW_SGI_color_table) CONST_CAST(GLEW_SGI_color_table) = !_glewInit_GL_SGI_color_table(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SGI_color_table */
+#ifdef GL_SGI_texture_color_table
+  CONST_CAST(GLEW_SGI_texture_color_table) = glewGetExtension("GL_SGI_texture_color_table");
+#endif /* GL_SGI_texture_color_table */
+#ifdef GL_SUNX_constant_data
+  CONST_CAST(GLEW_SUNX_constant_data) = glewGetExtension("GL_SUNX_constant_data");
+  if (glewExperimental || GLEW_SUNX_constant_data) CONST_CAST(GLEW_SUNX_constant_data) = !_glewInit_GL_SUNX_constant_data(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SUNX_constant_data */
+#ifdef GL_SUN_convolution_border_modes
+  CONST_CAST(GLEW_SUN_convolution_border_modes) = glewGetExtension("GL_SUN_convolution_border_modes");
+#endif /* GL_SUN_convolution_border_modes */
+#ifdef GL_SUN_global_alpha
+  CONST_CAST(GLEW_SUN_global_alpha) = glewGetExtension("GL_SUN_global_alpha");
+  if (glewExperimental || GLEW_SUN_global_alpha) CONST_CAST(GLEW_SUN_global_alpha) = !_glewInit_GL_SUN_global_alpha(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SUN_global_alpha */
+#ifdef GL_SUN_mesh_array
+  CONST_CAST(GLEW_SUN_mesh_array) = glewGetExtension("GL_SUN_mesh_array");
+#endif /* GL_SUN_mesh_array */
+#ifdef GL_SUN_read_video_pixels
+  CONST_CAST(GLEW_SUN_read_video_pixels) = glewGetExtension("GL_SUN_read_video_pixels");
+  if (glewExperimental || GLEW_SUN_read_video_pixels) CONST_CAST(GLEW_SUN_read_video_pixels) = !_glewInit_GL_SUN_read_video_pixels(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SUN_read_video_pixels */
+#ifdef GL_SUN_slice_accum
+  CONST_CAST(GLEW_SUN_slice_accum) = glewGetExtension("GL_SUN_slice_accum");
+#endif /* GL_SUN_slice_accum */
+#ifdef GL_SUN_triangle_list
+  CONST_CAST(GLEW_SUN_triangle_list) = glewGetExtension("GL_SUN_triangle_list");
+  if (glewExperimental || GLEW_SUN_triangle_list) CONST_CAST(GLEW_SUN_triangle_list) = !_glewInit_GL_SUN_triangle_list(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SUN_triangle_list */
+#ifdef GL_SUN_vertex
+  CONST_CAST(GLEW_SUN_vertex) = glewGetExtension("GL_SUN_vertex");
+  if (glewExperimental || GLEW_SUN_vertex) CONST_CAST(GLEW_SUN_vertex) = !_glewInit_GL_SUN_vertex(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_SUN_vertex */
+#ifdef GL_WIN_phong_shading
+  CONST_CAST(GLEW_WIN_phong_shading) = glewGetExtension("GL_WIN_phong_shading");
+#endif /* GL_WIN_phong_shading */
+#ifdef GL_WIN_specular_fog
+  CONST_CAST(GLEW_WIN_specular_fog) = glewGetExtension("GL_WIN_specular_fog");
+#endif /* GL_WIN_specular_fog */
+#ifdef GL_WIN_swap_hint
+  CONST_CAST(GLEW_WIN_swap_hint) = glewGetExtension("GL_WIN_swap_hint");
+  if (glewExperimental || GLEW_WIN_swap_hint) CONST_CAST(GLEW_WIN_swap_hint) = !_glewInit_GL_WIN_swap_hint(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GL_WIN_swap_hint */
+
+  return GLEW_OK;
+}
+
+
+#if defined(_WIN32)
+
+#if !defined(GLEW_MX)
+
+PFNWGLSETSTEREOEMITTERSTATE3DLPROC __wglewSetStereoEmitterState3DL = NULL;
+
+PFNWGLCREATEBUFFERREGIONARBPROC __wglewCreateBufferRegionARB = NULL;
+PFNWGLDELETEBUFFERREGIONARBPROC __wglewDeleteBufferRegionARB = NULL;
+PFNWGLRESTOREBUFFERREGIONARBPROC __wglewRestoreBufferRegionARB = NULL;
+PFNWGLSAVEBUFFERREGIONARBPROC __wglewSaveBufferRegionARB = NULL;
+
+PFNWGLCREATECONTEXTATTRIBSARBPROC __wglewCreateContextAttribsARB = NULL;
+
+PFNWGLGETEXTENSIONSSTRINGARBPROC __wglewGetExtensionsStringARB = NULL;
+
+PFNWGLGETCURRENTREADDCARBPROC __wglewGetCurrentReadDCARB = NULL;
+PFNWGLMAKECONTEXTCURRENTARBPROC __wglewMakeContextCurrentARB = NULL;
+
+PFNWGLCREATEPBUFFERARBPROC __wglewCreatePbufferARB = NULL;
+PFNWGLDESTROYPBUFFERARBPROC __wglewDestroyPbufferARB = NULL;
+PFNWGLGETPBUFFERDCARBPROC __wglewGetPbufferDCARB = NULL;
+PFNWGLQUERYPBUFFERARBPROC __wglewQueryPbufferARB = NULL;
+PFNWGLRELEASEPBUFFERDCARBPROC __wglewReleasePbufferDCARB = NULL;
+
+PFNWGLCHOOSEPIXELFORMATARBPROC __wglewChoosePixelFormatARB = NULL;
+PFNWGLGETPIXELFORMATATTRIBFVARBPROC __wglewGetPixelFormatAttribfvARB = NULL;
+PFNWGLGETPIXELFORMATATTRIBIVARBPROC __wglewGetPixelFormatAttribivARB = NULL;
+
+PFNWGLBINDTEXIMAGEARBPROC __wglewBindTexImageARB = NULL;
+PFNWGLRELEASETEXIMAGEARBPROC __wglewReleaseTexImageARB = NULL;
+PFNWGLSETPBUFFERATTRIBARBPROC __wglewSetPbufferAttribARB = NULL;
+
+PFNWGLBINDDISPLAYCOLORTABLEEXTPROC __wglewBindDisplayColorTableEXT = NULL;
+PFNWGLCREATEDISPLAYCOLORTABLEEXTPROC __wglewCreateDisplayColorTableEXT = NULL;
+PFNWGLDESTROYDISPLAYCOLORTABLEEXTPROC __wglewDestroyDisplayColorTableEXT = NULL;
+PFNWGLLOADDISPLAYCOLORTABLEEXTPROC __wglewLoadDisplayColorTableEXT = NULL;
+
+PFNWGLGETEXTENSIONSSTRINGEXTPROC __wglewGetExtensionsStringEXT = NULL;
+
+PFNWGLGETCURRENTREADDCEXTPROC __wglewGetCurrentReadDCEXT = NULL;
+PFNWGLMAKECONTEXTCURRENTEXTPROC __wglewMakeContextCurrentEXT = NULL;
+
+PFNWGLCREATEPBUFFEREXTPROC __wglewCreatePbufferEXT = NULL;
+PFNWGLDESTROYPBUFFEREXTPROC __wglewDestroyPbufferEXT = NULL;
+PFNWGLGETPBUFFERDCEXTPROC __wglewGetPbufferDCEXT = NULL;
+PFNWGLQUERYPBUFFEREXTPROC __wglewQueryPbufferEXT = NULL;
+PFNWGLRELEASEPBUFFERDCEXTPROC __wglewReleasePbufferDCEXT = NULL;
+
+PFNWGLCHOOSEPIXELFORMATEXTPROC __wglewChoosePixelFormatEXT = NULL;
+PFNWGLGETPIXELFORMATATTRIBFVEXTPROC __wglewGetPixelFormatAttribfvEXT = NULL;
+PFNWGLGETPIXELFORMATATTRIBIVEXTPROC __wglewGetPixelFormatAttribivEXT = NULL;
+
+PFNWGLGETSWAPINTERVALEXTPROC __wglewGetSwapIntervalEXT = NULL;
+PFNWGLSWAPINTERVALEXTPROC __wglewSwapIntervalEXT = NULL;
+
+PFNWGLGETDIGITALVIDEOPARAMETERSI3DPROC __wglewGetDigitalVideoParametersI3D = NULL;
+PFNWGLSETDIGITALVIDEOPARAMETERSI3DPROC __wglewSetDigitalVideoParametersI3D = NULL;
+
+PFNWGLGETGAMMATABLEI3DPROC __wglewGetGammaTableI3D = NULL;
+PFNWGLGETGAMMATABLEPARAMETERSI3DPROC __wglewGetGammaTableParametersI3D = NULL;
+PFNWGLSETGAMMATABLEI3DPROC __wglewSetGammaTableI3D = NULL;
+PFNWGLSETGAMMATABLEPARAMETERSI3DPROC __wglewSetGammaTableParametersI3D = NULL;
+
+PFNWGLDISABLEGENLOCKI3DPROC __wglewDisableGenlockI3D = NULL;
+PFNWGLENABLEGENLOCKI3DPROC __wglewEnableGenlockI3D = NULL;
+PFNWGLGENLOCKSAMPLERATEI3DPROC __wglewGenlockSampleRateI3D = NULL;
+PFNWGLGENLOCKSOURCEDELAYI3DPROC __wglewGenlockSourceDelayI3D = NULL;
+PFNWGLGENLOCKSOURCEEDGEI3DPROC __wglewGenlockSourceEdgeI3D = NULL;
+PFNWGLGENLOCKSOURCEI3DPROC __wglewGenlockSourceI3D = NULL;
+PFNWGLGETGENLOCKSAMPLERATEI3DPROC __wglewGetGenlockSampleRateI3D = NULL;
+PFNWGLGETGENLOCKSOURCEDELAYI3DPROC __wglewGetGenlockSourceDelayI3D = NULL;
+PFNWGLGETGENLOCKSOURCEEDGEI3DPROC __wglewGetGenlockSourceEdgeI3D = NULL;
+PFNWGLGETGENLOCKSOURCEI3DPROC __wglewGetGenlockSourceI3D = NULL;
+PFNWGLISENABLEDGENLOCKI3DPROC __wglewIsEnabledGenlockI3D = NULL;
+PFNWGLQUERYGENLOCKMAXSOURCEDELAYI3DPROC __wglewQueryGenlockMaxSourceDelayI3D = NULL;
+
+PFNWGLASSOCIATEIMAGEBUFFEREVENTSI3DPROC __wglewAssociateImageBufferEventsI3D = NULL;
+PFNWGLCREATEIMAGEBUFFERI3DPROC __wglewCreateImageBufferI3D = NULL;
+PFNWGLDESTROYIMAGEBUFFERI3DPROC __wglewDestroyImageBufferI3D = NULL;
+PFNWGLRELEASEIMAGEBUFFEREVENTSI3DPROC __wglewReleaseImageBufferEventsI3D = NULL;
+
+PFNWGLDISABLEFRAMELOCKI3DPROC __wglewDisableFrameLockI3D = NULL;
+PFNWGLENABLEFRAMELOCKI3DPROC __wglewEnableFrameLockI3D = NULL;
+PFNWGLISENABLEDFRAMELOCKI3DPROC __wglewIsEnabledFrameLockI3D = NULL;
+PFNWGLQUERYFRAMELOCKMASTERI3DPROC __wglewQueryFrameLockMasterI3D = NULL;
+
+PFNWGLBEGINFRAMETRACKINGI3DPROC __wglewBeginFrameTrackingI3D = NULL;
+PFNWGLENDFRAMETRACKINGI3DPROC __wglewEndFrameTrackingI3D = NULL;
+PFNWGLGETFRAMEUSAGEI3DPROC __wglewGetFrameUsageI3D = NULL;
+PFNWGLQUERYFRAMETRACKINGI3DPROC __wglewQueryFrameTrackingI3D = NULL;
+
+PFNWGLCREATEAFFINITYDCNVPROC __wglewCreateAffinityDCNV = NULL;
+PFNWGLDELETEDCNVPROC __wglewDeleteDCNV = NULL;
+PFNWGLENUMGPUDEVICESNVPROC __wglewEnumGpuDevicesNV = NULL;
+PFNWGLENUMGPUSFROMAFFINITYDCNVPROC __wglewEnumGpusFromAffinityDCNV = NULL;
+PFNWGLENUMGPUSNVPROC __wglewEnumGpusNV = NULL;
+
+PFNWGLBINDVIDEODEVICENVPROC __wglewBindVideoDeviceNV = NULL;
+PFNWGLENUMERATEVIDEODEVICESNVPROC __wglewEnumerateVideoDevicesNV = NULL;
+PFNWGLQUERYCURRENTCONTEXTNVPROC __wglewQueryCurrentContextNV = NULL;
+
+PFNWGLBINDSWAPBARRIERNVPROC __wglewBindSwapBarrierNV = NULL;
+PFNWGLJOINSWAPGROUPNVPROC __wglewJoinSwapGroupNV = NULL;
+PFNWGLQUERYFRAMECOUNTNVPROC __wglewQueryFrameCountNV = NULL;
+PFNWGLQUERYMAXSWAPGROUPSNVPROC __wglewQueryMaxSwapGroupsNV = NULL;
+PFNWGLQUERYSWAPGROUPNVPROC __wglewQuerySwapGroupNV = NULL;
+PFNWGLRESETFRAMECOUNTNVPROC __wglewResetFrameCountNV = NULL;
+
+PFNWGLALLOCATEMEMORYNVPROC __wglewAllocateMemoryNV = NULL;
+PFNWGLFREEMEMORYNVPROC __wglewFreeMemoryNV = NULL;
+
+PFNWGLBINDVIDEOIMAGENVPROC __wglewBindVideoImageNV = NULL;
+PFNWGLGETVIDEODEVICENVPROC __wglewGetVideoDeviceNV = NULL;
+PFNWGLGETVIDEOINFONVPROC __wglewGetVideoInfoNV = NULL;
+PFNWGLRELEASEVIDEODEVICENVPROC __wglewReleaseVideoDeviceNV = NULL;
+PFNWGLRELEASEVIDEOIMAGENVPROC __wglewReleaseVideoImageNV = NULL;
+PFNWGLSENDPBUFFERTOVIDEONVPROC __wglewSendPbufferToVideoNV = NULL;
+
+PFNWGLGETMSCRATEOMLPROC __wglewGetMscRateOML = NULL;
+PFNWGLGETSYNCVALUESOMLPROC __wglewGetSyncValuesOML = NULL;
+PFNWGLSWAPBUFFERSMSCOMLPROC __wglewSwapBuffersMscOML = NULL;
+PFNWGLSWAPLAYERBUFFERSMSCOMLPROC __wglewSwapLayerBuffersMscOML = NULL;
+PFNWGLWAITFORMSCOMLPROC __wglewWaitForMscOML = NULL;
+PFNWGLWAITFORSBCOMLPROC __wglewWaitForSbcOML = NULL;
+GLboolean __WGLEW_3DFX_multisample = GL_FALSE;
+GLboolean __WGLEW_3DL_stereo_control = GL_FALSE;
+GLboolean __WGLEW_ARB_buffer_region = GL_FALSE;
+GLboolean __WGLEW_ARB_create_context = GL_FALSE;
+GLboolean __WGLEW_ARB_extensions_string = GL_FALSE;
+GLboolean __WGLEW_ARB_framebuffer_sRGB = GL_FALSE;
+GLboolean __WGLEW_ARB_make_current_read = GL_FALSE;
+GLboolean __WGLEW_ARB_multisample = GL_FALSE;
+GLboolean __WGLEW_ARB_pbuffer = GL_FALSE;
+GLboolean __WGLEW_ARB_pixel_format = GL_FALSE;
+GLboolean __WGLEW_ARB_pixel_format_float = GL_FALSE;
+GLboolean __WGLEW_ARB_render_texture = GL_FALSE;
+GLboolean __WGLEW_ATI_pixel_format_float = GL_FALSE;
+GLboolean __WGLEW_ATI_render_texture_rectangle = GL_FALSE;
+GLboolean __WGLEW_EXT_depth_float = GL_FALSE;
+GLboolean __WGLEW_EXT_display_color_table = GL_FALSE;
+GLboolean __WGLEW_EXT_extensions_string = GL_FALSE;
+GLboolean __WGLEW_EXT_framebuffer_sRGB = GL_FALSE;
+GLboolean __WGLEW_EXT_make_current_read = GL_FALSE;
+GLboolean __WGLEW_EXT_multisample = GL_FALSE;
+GLboolean __WGLEW_EXT_pbuffer = GL_FALSE;
+GLboolean __WGLEW_EXT_pixel_format = GL_FALSE;
+GLboolean __WGLEW_EXT_pixel_format_packed_float = GL_FALSE;
+GLboolean __WGLEW_EXT_swap_control = GL_FALSE;
+GLboolean __WGLEW_I3D_digital_video_control = GL_FALSE;
+GLboolean __WGLEW_I3D_gamma = GL_FALSE;
+GLboolean __WGLEW_I3D_genlock = GL_FALSE;
+GLboolean __WGLEW_I3D_image_buffer = GL_FALSE;
+GLboolean __WGLEW_I3D_swap_frame_lock = GL_FALSE;
+GLboolean __WGLEW_I3D_swap_frame_usage = GL_FALSE;
+GLboolean __WGLEW_NV_float_buffer = GL_FALSE;
+GLboolean __WGLEW_NV_gpu_affinity = GL_FALSE;
+GLboolean __WGLEW_NV_present_video = GL_FALSE;
+GLboolean __WGLEW_NV_render_depth_texture = GL_FALSE;
+GLboolean __WGLEW_NV_render_texture_rectangle = GL_FALSE;
+GLboolean __WGLEW_NV_swap_group = GL_FALSE;
+GLboolean __WGLEW_NV_vertex_array_range = GL_FALSE;
+GLboolean __WGLEW_NV_video_output = GL_FALSE;
+GLboolean __WGLEW_OML_sync_control = GL_FALSE;
+
+#endif /* !GLEW_MX */
+
+#ifdef WGL_3DFX_multisample
+
+#endif /* WGL_3DFX_multisample */
+
+#ifdef WGL_3DL_stereo_control
+
+static GLboolean _glewInit_WGL_3DL_stereo_control (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglSetStereoEmitterState3DL = (PFNWGLSETSTEREOEMITTERSTATE3DLPROC)glewGetProcAddress((const GLubyte*)"wglSetStereoEmitterState3DL")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_3DL_stereo_control */
+
+#ifdef WGL_ARB_buffer_region
+
+static GLboolean _glewInit_WGL_ARB_buffer_region (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglCreateBufferRegionARB = (PFNWGLCREATEBUFFERREGIONARBPROC)glewGetProcAddress((const GLubyte*)"wglCreateBufferRegionARB")) == NULL) || r;
+  r = ((wglDeleteBufferRegionARB = (PFNWGLDELETEBUFFERREGIONARBPROC)glewGetProcAddress((const GLubyte*)"wglDeleteBufferRegionARB")) == NULL) || r;
+  r = ((wglRestoreBufferRegionARB = (PFNWGLRESTOREBUFFERREGIONARBPROC)glewGetProcAddress((const GLubyte*)"wglRestoreBufferRegionARB")) == NULL) || r;
+  r = ((wglSaveBufferRegionARB = (PFNWGLSAVEBUFFERREGIONARBPROC)glewGetProcAddress((const GLubyte*)"wglSaveBufferRegionARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_ARB_buffer_region */
+
+#ifdef WGL_ARB_create_context
+
+static GLboolean _glewInit_WGL_ARB_create_context (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglCreateContextAttribsARB = (PFNWGLCREATECONTEXTATTRIBSARBPROC)glewGetProcAddress((const GLubyte*)"wglCreateContextAttribsARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_ARB_create_context */
+
+#ifdef WGL_ARB_extensions_string
+
+static GLboolean _glewInit_WGL_ARB_extensions_string (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglGetExtensionsStringARB = (PFNWGLGETEXTENSIONSSTRINGARBPROC)glewGetProcAddress((const GLubyte*)"wglGetExtensionsStringARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_ARB_extensions_string */
+
+#ifdef WGL_ARB_framebuffer_sRGB
+
+#endif /* WGL_ARB_framebuffer_sRGB */
+
+#ifdef WGL_ARB_make_current_read
+
+static GLboolean _glewInit_WGL_ARB_make_current_read (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglGetCurrentReadDCARB = (PFNWGLGETCURRENTREADDCARBPROC)glewGetProcAddress((const GLubyte*)"wglGetCurrentReadDCARB")) == NULL) || r;
+  r = ((wglMakeContextCurrentARB = (PFNWGLMAKECONTEXTCURRENTARBPROC)glewGetProcAddress((const GLubyte*)"wglMakeContextCurrentARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_ARB_make_current_read */
+
+#ifdef WGL_ARB_multisample
+
+#endif /* WGL_ARB_multisample */
+
+#ifdef WGL_ARB_pbuffer
+
+static GLboolean _glewInit_WGL_ARB_pbuffer (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglCreatePbufferARB = (PFNWGLCREATEPBUFFERARBPROC)glewGetProcAddress((const GLubyte*)"wglCreatePbufferARB")) == NULL) || r;
+  r = ((wglDestroyPbufferARB = (PFNWGLDESTROYPBUFFERARBPROC)glewGetProcAddress((const GLubyte*)"wglDestroyPbufferARB")) == NULL) || r;
+  r = ((wglGetPbufferDCARB = (PFNWGLGETPBUFFERDCARBPROC)glewGetProcAddress((const GLubyte*)"wglGetPbufferDCARB")) == NULL) || r;
+  r = ((wglQueryPbufferARB = (PFNWGLQUERYPBUFFERARBPROC)glewGetProcAddress((const GLubyte*)"wglQueryPbufferARB")) == NULL) || r;
+  r = ((wglReleasePbufferDCARB = (PFNWGLRELEASEPBUFFERDCARBPROC)glewGetProcAddress((const GLubyte*)"wglReleasePbufferDCARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_ARB_pbuffer */
+
+#ifdef WGL_ARB_pixel_format
+
+static GLboolean _glewInit_WGL_ARB_pixel_format (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglChoosePixelFormatARB = (PFNWGLCHOOSEPIXELFORMATARBPROC)glewGetProcAddress((const GLubyte*)"wglChoosePixelFormatARB")) == NULL) || r;
+  r = ((wglGetPixelFormatAttribfvARB = (PFNWGLGETPIXELFORMATATTRIBFVARBPROC)glewGetProcAddress((const GLubyte*)"wglGetPixelFormatAttribfvARB")) == NULL) || r;
+  r = ((wglGetPixelFormatAttribivARB = (PFNWGLGETPIXELFORMATATTRIBIVARBPROC)glewGetProcAddress((const GLubyte*)"wglGetPixelFormatAttribivARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_ARB_pixel_format */
+
+#ifdef WGL_ARB_pixel_format_float
+
+#endif /* WGL_ARB_pixel_format_float */
+
+#ifdef WGL_ARB_render_texture
+
+static GLboolean _glewInit_WGL_ARB_render_texture (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglBindTexImageARB = (PFNWGLBINDTEXIMAGEARBPROC)glewGetProcAddress((const GLubyte*)"wglBindTexImageARB")) == NULL) || r;
+  r = ((wglReleaseTexImageARB = (PFNWGLRELEASETEXIMAGEARBPROC)glewGetProcAddress((const GLubyte*)"wglReleaseTexImageARB")) == NULL) || r;
+  r = ((wglSetPbufferAttribARB = (PFNWGLSETPBUFFERATTRIBARBPROC)glewGetProcAddress((const GLubyte*)"wglSetPbufferAttribARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_ARB_render_texture */
+
+#ifdef WGL_ATI_pixel_format_float
+
+#endif /* WGL_ATI_pixel_format_float */
+
+#ifdef WGL_ATI_render_texture_rectangle
+
+#endif /* WGL_ATI_render_texture_rectangle */
+
+#ifdef WGL_EXT_depth_float
+
+#endif /* WGL_EXT_depth_float */
+
+#ifdef WGL_EXT_display_color_table
+
+static GLboolean _glewInit_WGL_EXT_display_color_table (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglBindDisplayColorTableEXT = (PFNWGLBINDDISPLAYCOLORTABLEEXTPROC)glewGetProcAddress((const GLubyte*)"wglBindDisplayColorTableEXT")) == NULL) || r;
+  r = ((wglCreateDisplayColorTableEXT = (PFNWGLCREATEDISPLAYCOLORTABLEEXTPROC)glewGetProcAddress((const GLubyte*)"wglCreateDisplayColorTableEXT")) == NULL) || r;
+  r = ((wglDestroyDisplayColorTableEXT = (PFNWGLDESTROYDISPLAYCOLORTABLEEXTPROC)glewGetProcAddress((const GLubyte*)"wglDestroyDisplayColorTableEXT")) == NULL) || r;
+  r = ((wglLoadDisplayColorTableEXT = (PFNWGLLOADDISPLAYCOLORTABLEEXTPROC)glewGetProcAddress((const GLubyte*)"wglLoadDisplayColorTableEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_EXT_display_color_table */
+
+#ifdef WGL_EXT_extensions_string
+
+static GLboolean _glewInit_WGL_EXT_extensions_string (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglGetExtensionsStringEXT = (PFNWGLGETEXTENSIONSSTRINGEXTPROC)glewGetProcAddress((const GLubyte*)"wglGetExtensionsStringEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_EXT_extensions_string */
+
+#ifdef WGL_EXT_framebuffer_sRGB
+
+#endif /* WGL_EXT_framebuffer_sRGB */
+
+#ifdef WGL_EXT_make_current_read
+
+static GLboolean _glewInit_WGL_EXT_make_current_read (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglGetCurrentReadDCEXT = (PFNWGLGETCURRENTREADDCEXTPROC)glewGetProcAddress((const GLubyte*)"wglGetCurrentReadDCEXT")) == NULL) || r;
+  r = ((wglMakeContextCurrentEXT = (PFNWGLMAKECONTEXTCURRENTEXTPROC)glewGetProcAddress((const GLubyte*)"wglMakeContextCurrentEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_EXT_make_current_read */
+
+#ifdef WGL_EXT_multisample
+
+#endif /* WGL_EXT_multisample */
+
+#ifdef WGL_EXT_pbuffer
+
+static GLboolean _glewInit_WGL_EXT_pbuffer (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglCreatePbufferEXT = (PFNWGLCREATEPBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"wglCreatePbufferEXT")) == NULL) || r;
+  r = ((wglDestroyPbufferEXT = (PFNWGLDESTROYPBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"wglDestroyPbufferEXT")) == NULL) || r;
+  r = ((wglGetPbufferDCEXT = (PFNWGLGETPBUFFERDCEXTPROC)glewGetProcAddress((const GLubyte*)"wglGetPbufferDCEXT")) == NULL) || r;
+  r = ((wglQueryPbufferEXT = (PFNWGLQUERYPBUFFEREXTPROC)glewGetProcAddress((const GLubyte*)"wglQueryPbufferEXT")) == NULL) || r;
+  r = ((wglReleasePbufferDCEXT = (PFNWGLRELEASEPBUFFERDCEXTPROC)glewGetProcAddress((const GLubyte*)"wglReleasePbufferDCEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_EXT_pbuffer */
+
+#ifdef WGL_EXT_pixel_format
+
+static GLboolean _glewInit_WGL_EXT_pixel_format (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglChoosePixelFormatEXT = (PFNWGLCHOOSEPIXELFORMATEXTPROC)glewGetProcAddress((const GLubyte*)"wglChoosePixelFormatEXT")) == NULL) || r;
+  r = ((wglGetPixelFormatAttribfvEXT = (PFNWGLGETPIXELFORMATATTRIBFVEXTPROC)glewGetProcAddress((const GLubyte*)"wglGetPixelFormatAttribfvEXT")) == NULL) || r;
+  r = ((wglGetPixelFormatAttribivEXT = (PFNWGLGETPIXELFORMATATTRIBIVEXTPROC)glewGetProcAddress((const GLubyte*)"wglGetPixelFormatAttribivEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_EXT_pixel_format */
+
+#ifdef WGL_EXT_pixel_format_packed_float
+
+#endif /* WGL_EXT_pixel_format_packed_float */
+
+#ifdef WGL_EXT_swap_control
+
+static GLboolean _glewInit_WGL_EXT_swap_control (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglGetSwapIntervalEXT = (PFNWGLGETSWAPINTERVALEXTPROC)glewGetProcAddress((const GLubyte*)"wglGetSwapIntervalEXT")) == NULL) || r;
+  r = ((wglSwapIntervalEXT = (PFNWGLSWAPINTERVALEXTPROC)glewGetProcAddress((const GLubyte*)"wglSwapIntervalEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_EXT_swap_control */
+
+#ifdef WGL_I3D_digital_video_control
+
+static GLboolean _glewInit_WGL_I3D_digital_video_control (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglGetDigitalVideoParametersI3D = (PFNWGLGETDIGITALVIDEOPARAMETERSI3DPROC)glewGetProcAddress((const GLubyte*)"wglGetDigitalVideoParametersI3D")) == NULL) || r;
+  r = ((wglSetDigitalVideoParametersI3D = (PFNWGLSETDIGITALVIDEOPARAMETERSI3DPROC)glewGetProcAddress((const GLubyte*)"wglSetDigitalVideoParametersI3D")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_I3D_digital_video_control */
+
+#ifdef WGL_I3D_gamma
+
+static GLboolean _glewInit_WGL_I3D_gamma (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglGetGammaTableI3D = (PFNWGLGETGAMMATABLEI3DPROC)glewGetProcAddress((const GLubyte*)"wglGetGammaTableI3D")) == NULL) || r;
+  r = ((wglGetGammaTableParametersI3D = (PFNWGLGETGAMMATABLEPARAMETERSI3DPROC)glewGetProcAddress((const GLubyte*)"wglGetGammaTableParametersI3D")) == NULL) || r;
+  r = ((wglSetGammaTableI3D = (PFNWGLSETGAMMATABLEI3DPROC)glewGetProcAddress((const GLubyte*)"wglSetGammaTableI3D")) == NULL) || r;
+  r = ((wglSetGammaTableParametersI3D = (PFNWGLSETGAMMATABLEPARAMETERSI3DPROC)glewGetProcAddress((const GLubyte*)"wglSetGammaTableParametersI3D")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_I3D_gamma */
+
+#ifdef WGL_I3D_genlock
+
+static GLboolean _glewInit_WGL_I3D_genlock (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglDisableGenlockI3D = (PFNWGLDISABLEGENLOCKI3DPROC)glewGetProcAddress((const GLubyte*)"wglDisableGenlockI3D")) == NULL) || r;
+  r = ((wglEnableGenlockI3D = (PFNWGLENABLEGENLOCKI3DPROC)glewGetProcAddress((const GLubyte*)"wglEnableGenlockI3D")) == NULL) || r;
+  r = ((wglGenlockSampleRateI3D = (PFNWGLGENLOCKSAMPLERATEI3DPROC)glewGetProcAddress((const GLubyte*)"wglGenlockSampleRateI3D")) == NULL) || r;
+  r = ((wglGenlockSourceDelayI3D = (PFNWGLGENLOCKSOURCEDELAYI3DPROC)glewGetProcAddress((const GLubyte*)"wglGenlockSourceDelayI3D")) == NULL) || r;
+  r = ((wglGenlockSourceEdgeI3D = (PFNWGLGENLOCKSOURCEEDGEI3DPROC)glewGetProcAddress((const GLubyte*)"wglGenlockSourceEdgeI3D")) == NULL) || r;
+  r = ((wglGenlockSourceI3D = (PFNWGLGENLOCKSOURCEI3DPROC)glewGetProcAddress((const GLubyte*)"wglGenlockSourceI3D")) == NULL) || r;
+  r = ((wglGetGenlockSampleRateI3D = (PFNWGLGETGENLOCKSAMPLERATEI3DPROC)glewGetProcAddress((const GLubyte*)"wglGetGenlockSampleRateI3D")) == NULL) || r;
+  r = ((wglGetGenlockSourceDelayI3D = (PFNWGLGETGENLOCKSOURCEDELAYI3DPROC)glewGetProcAddress((const GLubyte*)"wglGetGenlockSourceDelayI3D")) == NULL) || r;
+  r = ((wglGetGenlockSourceEdgeI3D = (PFNWGLGETGENLOCKSOURCEEDGEI3DPROC)glewGetProcAddress((const GLubyte*)"wglGetGenlockSourceEdgeI3D")) == NULL) || r;
+  r = ((wglGetGenlockSourceI3D = (PFNWGLGETGENLOCKSOURCEI3DPROC)glewGetProcAddress((const GLubyte*)"wglGetGenlockSourceI3D")) == NULL) || r;
+  r = ((wglIsEnabledGenlockI3D = (PFNWGLISENABLEDGENLOCKI3DPROC)glewGetProcAddress((const GLubyte*)"wglIsEnabledGenlockI3D")) == NULL) || r;
+  r = ((wglQueryGenlockMaxSourceDelayI3D = (PFNWGLQUERYGENLOCKMAXSOURCEDELAYI3DPROC)glewGetProcAddress((const GLubyte*)"wglQueryGenlockMaxSourceDelayI3D")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_I3D_genlock */
+
+#ifdef WGL_I3D_image_buffer
+
+static GLboolean _glewInit_WGL_I3D_image_buffer (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglAssociateImageBufferEventsI3D = (PFNWGLASSOCIATEIMAGEBUFFEREVENTSI3DPROC)glewGetProcAddress((const GLubyte*)"wglAssociateImageBufferEventsI3D")) == NULL) || r;
+  r = ((wglCreateImageBufferI3D = (PFNWGLCREATEIMAGEBUFFERI3DPROC)glewGetProcAddress((const GLubyte*)"wglCreateImageBufferI3D")) == NULL) || r;
+  r = ((wglDestroyImageBufferI3D = (PFNWGLDESTROYIMAGEBUFFERI3DPROC)glewGetProcAddress((const GLubyte*)"wglDestroyImageBufferI3D")) == NULL) || r;
+  r = ((wglReleaseImageBufferEventsI3D = (PFNWGLRELEASEIMAGEBUFFEREVENTSI3DPROC)glewGetProcAddress((const GLubyte*)"wglReleaseImageBufferEventsI3D")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_I3D_image_buffer */
+
+#ifdef WGL_I3D_swap_frame_lock
+
+static GLboolean _glewInit_WGL_I3D_swap_frame_lock (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglDisableFrameLockI3D = (PFNWGLDISABLEFRAMELOCKI3DPROC)glewGetProcAddress((const GLubyte*)"wglDisableFrameLockI3D")) == NULL) || r;
+  r = ((wglEnableFrameLockI3D = (PFNWGLENABLEFRAMELOCKI3DPROC)glewGetProcAddress((const GLubyte*)"wglEnableFrameLockI3D")) == NULL) || r;
+  r = ((wglIsEnabledFrameLockI3D = (PFNWGLISENABLEDFRAMELOCKI3DPROC)glewGetProcAddress((const GLubyte*)"wglIsEnabledFrameLockI3D")) == NULL) || r;
+  r = ((wglQueryFrameLockMasterI3D = (PFNWGLQUERYFRAMELOCKMASTERI3DPROC)glewGetProcAddress((const GLubyte*)"wglQueryFrameLockMasterI3D")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_I3D_swap_frame_lock */
+
+#ifdef WGL_I3D_swap_frame_usage
+
+static GLboolean _glewInit_WGL_I3D_swap_frame_usage (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglBeginFrameTrackingI3D = (PFNWGLBEGINFRAMETRACKINGI3DPROC)glewGetProcAddress((const GLubyte*)"wglBeginFrameTrackingI3D")) == NULL) || r;
+  r = ((wglEndFrameTrackingI3D = (PFNWGLENDFRAMETRACKINGI3DPROC)glewGetProcAddress((const GLubyte*)"wglEndFrameTrackingI3D")) == NULL) || r;
+  r = ((wglGetFrameUsageI3D = (PFNWGLGETFRAMEUSAGEI3DPROC)glewGetProcAddress((const GLubyte*)"wglGetFrameUsageI3D")) == NULL) || r;
+  r = ((wglQueryFrameTrackingI3D = (PFNWGLQUERYFRAMETRACKINGI3DPROC)glewGetProcAddress((const GLubyte*)"wglQueryFrameTrackingI3D")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_I3D_swap_frame_usage */
+
+#ifdef WGL_NV_float_buffer
+
+#endif /* WGL_NV_float_buffer */
+
+#ifdef WGL_NV_gpu_affinity
+
+static GLboolean _glewInit_WGL_NV_gpu_affinity (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglCreateAffinityDCNV = (PFNWGLCREATEAFFINITYDCNVPROC)glewGetProcAddress((const GLubyte*)"wglCreateAffinityDCNV")) == NULL) || r;
+  r = ((wglDeleteDCNV = (PFNWGLDELETEDCNVPROC)glewGetProcAddress((const GLubyte*)"wglDeleteDCNV")) == NULL) || r;
+  r = ((wglEnumGpuDevicesNV = (PFNWGLENUMGPUDEVICESNVPROC)glewGetProcAddress((const GLubyte*)"wglEnumGpuDevicesNV")) == NULL) || r;
+  r = ((wglEnumGpusFromAffinityDCNV = (PFNWGLENUMGPUSFROMAFFINITYDCNVPROC)glewGetProcAddress((const GLubyte*)"wglEnumGpusFromAffinityDCNV")) == NULL) || r;
+  r = ((wglEnumGpusNV = (PFNWGLENUMGPUSNVPROC)glewGetProcAddress((const GLubyte*)"wglEnumGpusNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_NV_gpu_affinity */
+
+#ifdef WGL_NV_present_video
+
+static GLboolean _glewInit_WGL_NV_present_video (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglBindVideoDeviceNV = (PFNWGLBINDVIDEODEVICENVPROC)glewGetProcAddress((const GLubyte*)"wglBindVideoDeviceNV")) == NULL) || r;
+  r = ((wglEnumerateVideoDevicesNV = (PFNWGLENUMERATEVIDEODEVICESNVPROC)glewGetProcAddress((const GLubyte*)"wglEnumerateVideoDevicesNV")) == NULL) || r;
+  r = ((wglQueryCurrentContextNV = (PFNWGLQUERYCURRENTCONTEXTNVPROC)glewGetProcAddress((const GLubyte*)"wglQueryCurrentContextNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_NV_present_video */
+
+#ifdef WGL_NV_render_depth_texture
+
+#endif /* WGL_NV_render_depth_texture */
+
+#ifdef WGL_NV_render_texture_rectangle
+
+#endif /* WGL_NV_render_texture_rectangle */
+
+#ifdef WGL_NV_swap_group
+
+static GLboolean _glewInit_WGL_NV_swap_group (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglBindSwapBarrierNV = (PFNWGLBINDSWAPBARRIERNVPROC)glewGetProcAddress((const GLubyte*)"wglBindSwapBarrierNV")) == NULL) || r;
+  r = ((wglJoinSwapGroupNV = (PFNWGLJOINSWAPGROUPNVPROC)glewGetProcAddress((const GLubyte*)"wglJoinSwapGroupNV")) == NULL) || r;
+  r = ((wglQueryFrameCountNV = (PFNWGLQUERYFRAMECOUNTNVPROC)glewGetProcAddress((const GLubyte*)"wglQueryFrameCountNV")) == NULL) || r;
+  r = ((wglQueryMaxSwapGroupsNV = (PFNWGLQUERYMAXSWAPGROUPSNVPROC)glewGetProcAddress((const GLubyte*)"wglQueryMaxSwapGroupsNV")) == NULL) || r;
+  r = ((wglQuerySwapGroupNV = (PFNWGLQUERYSWAPGROUPNVPROC)glewGetProcAddress((const GLubyte*)"wglQuerySwapGroupNV")) == NULL) || r;
+  r = ((wglResetFrameCountNV = (PFNWGLRESETFRAMECOUNTNVPROC)glewGetProcAddress((const GLubyte*)"wglResetFrameCountNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_NV_swap_group */
+
+#ifdef WGL_NV_vertex_array_range
+
+static GLboolean _glewInit_WGL_NV_vertex_array_range (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglAllocateMemoryNV = (PFNWGLALLOCATEMEMORYNVPROC)glewGetProcAddress((const GLubyte*)"wglAllocateMemoryNV")) == NULL) || r;
+  r = ((wglFreeMemoryNV = (PFNWGLFREEMEMORYNVPROC)glewGetProcAddress((const GLubyte*)"wglFreeMemoryNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_NV_vertex_array_range */
+
+#ifdef WGL_NV_video_output
+
+static GLboolean _glewInit_WGL_NV_video_output (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglBindVideoImageNV = (PFNWGLBINDVIDEOIMAGENVPROC)glewGetProcAddress((const GLubyte*)"wglBindVideoImageNV")) == NULL) || r;
+  r = ((wglGetVideoDeviceNV = (PFNWGLGETVIDEODEVICENVPROC)glewGetProcAddress((const GLubyte*)"wglGetVideoDeviceNV")) == NULL) || r;
+  r = ((wglGetVideoInfoNV = (PFNWGLGETVIDEOINFONVPROC)glewGetProcAddress((const GLubyte*)"wglGetVideoInfoNV")) == NULL) || r;
+  r = ((wglReleaseVideoDeviceNV = (PFNWGLRELEASEVIDEODEVICENVPROC)glewGetProcAddress((const GLubyte*)"wglReleaseVideoDeviceNV")) == NULL) || r;
+  r = ((wglReleaseVideoImageNV = (PFNWGLRELEASEVIDEOIMAGENVPROC)glewGetProcAddress((const GLubyte*)"wglReleaseVideoImageNV")) == NULL) || r;
+  r = ((wglSendPbufferToVideoNV = (PFNWGLSENDPBUFFERTOVIDEONVPROC)glewGetProcAddress((const GLubyte*)"wglSendPbufferToVideoNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_NV_video_output */
+
+#ifdef WGL_OML_sync_control
+
+static GLboolean _glewInit_WGL_OML_sync_control (WGLEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((wglGetMscRateOML = (PFNWGLGETMSCRATEOMLPROC)glewGetProcAddress((const GLubyte*)"wglGetMscRateOML")) == NULL) || r;
+  r = ((wglGetSyncValuesOML = (PFNWGLGETSYNCVALUESOMLPROC)glewGetProcAddress((const GLubyte*)"wglGetSyncValuesOML")) == NULL) || r;
+  r = ((wglSwapBuffersMscOML = (PFNWGLSWAPBUFFERSMSCOMLPROC)glewGetProcAddress((const GLubyte*)"wglSwapBuffersMscOML")) == NULL) || r;
+  r = ((wglSwapLayerBuffersMscOML = (PFNWGLSWAPLAYERBUFFERSMSCOMLPROC)glewGetProcAddress((const GLubyte*)"wglSwapLayerBuffersMscOML")) == NULL) || r;
+  r = ((wglWaitForMscOML = (PFNWGLWAITFORMSCOMLPROC)glewGetProcAddress((const GLubyte*)"wglWaitForMscOML")) == NULL) || r;
+  r = ((wglWaitForSbcOML = (PFNWGLWAITFORSBCOMLPROC)glewGetProcAddress((const GLubyte*)"wglWaitForSbcOML")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* WGL_OML_sync_control */
+
+/* ------------------------------------------------------------------------- */
+
+static PFNWGLGETEXTENSIONSSTRINGARBPROC _wglewGetExtensionsStringARB = NULL;
+static PFNWGLGETEXTENSIONSSTRINGEXTPROC _wglewGetExtensionsStringEXT = NULL;
+
+GLboolean wglewGetExtension (const char* name)
+{    
+  GLubyte* p;
+  GLubyte* end;
+  GLuint len = _glewStrLen((const GLubyte*)name);
+  if (_wglewGetExtensionsStringARB == NULL)
+    if (_wglewGetExtensionsStringEXT == NULL)
+      return GL_FALSE;
+    else
+      p = (GLubyte*)_wglewGetExtensionsStringEXT();
+  else
+    p = (GLubyte*)_wglewGetExtensionsStringARB(wglGetCurrentDC());
+  if (0 == p) return GL_FALSE;
+  end = p + _glewStrLen(p);
+  while (p < end)
+  {
+    GLuint n = _glewStrCLen(p, ' ');
+    if (len == n && _glewStrSame((const GLubyte*)name, p, n)) return GL_TRUE;
+    p += n+1;
+  }
+  return GL_FALSE;
+}
+
+GLenum wglewContextInit (WGLEW_CONTEXT_ARG_DEF_LIST)
+{
+  GLboolean crippled;
+  /* find wgl extension string query functions */
+  _wglewGetExtensionsStringARB = (PFNWGLGETEXTENSIONSSTRINGARBPROC)glewGetProcAddress((const GLubyte*)"wglGetExtensionsStringARB");
+  _wglewGetExtensionsStringEXT = (PFNWGLGETEXTENSIONSSTRINGEXTPROC)glewGetProcAddress((const GLubyte*)"wglGetExtensionsStringEXT");
+  /* initialize extensions */
+  crippled = _wglewGetExtensionsStringARB == NULL && _wglewGetExtensionsStringEXT == NULL;
+#ifdef WGL_3DFX_multisample
+  CONST_CAST(WGLEW_3DFX_multisample) = wglewGetExtension("WGL_3DFX_multisample");
+#endif /* WGL_3DFX_multisample */
+#ifdef WGL_3DL_stereo_control
+  CONST_CAST(WGLEW_3DL_stereo_control) = wglewGetExtension("WGL_3DL_stereo_control");
+  if (glewExperimental || WGLEW_3DL_stereo_control|| crippled) CONST_CAST(WGLEW_3DL_stereo_control)= !_glewInit_WGL_3DL_stereo_control(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_3DL_stereo_control */
+#ifdef WGL_ARB_buffer_region
+  CONST_CAST(WGLEW_ARB_buffer_region) = wglewGetExtension("WGL_ARB_buffer_region");
+  if (glewExperimental || WGLEW_ARB_buffer_region|| crippled) CONST_CAST(WGLEW_ARB_buffer_region)= !_glewInit_WGL_ARB_buffer_region(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_ARB_buffer_region */
+#ifdef WGL_ARB_create_context
+  CONST_CAST(WGLEW_ARB_create_context) = wglewGetExtension("WGL_ARB_create_context");
+  if (glewExperimental || WGLEW_ARB_create_context|| crippled) CONST_CAST(WGLEW_ARB_create_context)= !_glewInit_WGL_ARB_create_context(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_ARB_create_context */
+#ifdef WGL_ARB_extensions_string
+  CONST_CAST(WGLEW_ARB_extensions_string) = wglewGetExtension("WGL_ARB_extensions_string");
+  if (glewExperimental || WGLEW_ARB_extensions_string|| crippled) CONST_CAST(WGLEW_ARB_extensions_string)= !_glewInit_WGL_ARB_extensions_string(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_ARB_extensions_string */
+#ifdef WGL_ARB_framebuffer_sRGB
+  CONST_CAST(WGLEW_ARB_framebuffer_sRGB) = wglewGetExtension("WGL_ARB_framebuffer_sRGB");
+#endif /* WGL_ARB_framebuffer_sRGB */
+#ifdef WGL_ARB_make_current_read
+  CONST_CAST(WGLEW_ARB_make_current_read) = wglewGetExtension("WGL_ARB_make_current_read");
+  if (glewExperimental || WGLEW_ARB_make_current_read|| crippled) CONST_CAST(WGLEW_ARB_make_current_read)= !_glewInit_WGL_ARB_make_current_read(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_ARB_make_current_read */
+#ifdef WGL_ARB_multisample
+  CONST_CAST(WGLEW_ARB_multisample) = wglewGetExtension("WGL_ARB_multisample");
+#endif /* WGL_ARB_multisample */
+#ifdef WGL_ARB_pbuffer
+  CONST_CAST(WGLEW_ARB_pbuffer) = wglewGetExtension("WGL_ARB_pbuffer");
+  if (glewExperimental || WGLEW_ARB_pbuffer|| crippled) CONST_CAST(WGLEW_ARB_pbuffer)= !_glewInit_WGL_ARB_pbuffer(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_ARB_pbuffer */
+#ifdef WGL_ARB_pixel_format
+  CONST_CAST(WGLEW_ARB_pixel_format) = wglewGetExtension("WGL_ARB_pixel_format");
+  if (glewExperimental || WGLEW_ARB_pixel_format|| crippled) CONST_CAST(WGLEW_ARB_pixel_format)= !_glewInit_WGL_ARB_pixel_format(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_ARB_pixel_format */
+#ifdef WGL_ARB_pixel_format_float
+  CONST_CAST(WGLEW_ARB_pixel_format_float) = wglewGetExtension("WGL_ARB_pixel_format_float");
+#endif /* WGL_ARB_pixel_format_float */
+#ifdef WGL_ARB_render_texture
+  CONST_CAST(WGLEW_ARB_render_texture) = wglewGetExtension("WGL_ARB_render_texture");
+  if (glewExperimental || WGLEW_ARB_render_texture|| crippled) CONST_CAST(WGLEW_ARB_render_texture)= !_glewInit_WGL_ARB_render_texture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_ARB_render_texture */
+#ifdef WGL_ATI_pixel_format_float
+  CONST_CAST(WGLEW_ATI_pixel_format_float) = wglewGetExtension("WGL_ATI_pixel_format_float");
+#endif /* WGL_ATI_pixel_format_float */
+#ifdef WGL_ATI_render_texture_rectangle
+  CONST_CAST(WGLEW_ATI_render_texture_rectangle) = wglewGetExtension("WGL_ATI_render_texture_rectangle");
+#endif /* WGL_ATI_render_texture_rectangle */
+#ifdef WGL_EXT_depth_float
+  CONST_CAST(WGLEW_EXT_depth_float) = wglewGetExtension("WGL_EXT_depth_float");
+#endif /* WGL_EXT_depth_float */
+#ifdef WGL_EXT_display_color_table
+  CONST_CAST(WGLEW_EXT_display_color_table) = wglewGetExtension("WGL_EXT_display_color_table");
+  if (glewExperimental || WGLEW_EXT_display_color_table|| crippled) CONST_CAST(WGLEW_EXT_display_color_table)= !_glewInit_WGL_EXT_display_color_table(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_EXT_display_color_table */
+#ifdef WGL_EXT_extensions_string
+  CONST_CAST(WGLEW_EXT_extensions_string) = wglewGetExtension("WGL_EXT_extensions_string");
+  if (glewExperimental || WGLEW_EXT_extensions_string|| crippled) CONST_CAST(WGLEW_EXT_extensions_string)= !_glewInit_WGL_EXT_extensions_string(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_EXT_extensions_string */
+#ifdef WGL_EXT_framebuffer_sRGB
+  CONST_CAST(WGLEW_EXT_framebuffer_sRGB) = wglewGetExtension("WGL_EXT_framebuffer_sRGB");
+#endif /* WGL_EXT_framebuffer_sRGB */
+#ifdef WGL_EXT_make_current_read
+  CONST_CAST(WGLEW_EXT_make_current_read) = wglewGetExtension("WGL_EXT_make_current_read");
+  if (glewExperimental || WGLEW_EXT_make_current_read|| crippled) CONST_CAST(WGLEW_EXT_make_current_read)= !_glewInit_WGL_EXT_make_current_read(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_EXT_make_current_read */
+#ifdef WGL_EXT_multisample
+  CONST_CAST(WGLEW_EXT_multisample) = wglewGetExtension("WGL_EXT_multisample");
+#endif /* WGL_EXT_multisample */
+#ifdef WGL_EXT_pbuffer
+  CONST_CAST(WGLEW_EXT_pbuffer) = wglewGetExtension("WGL_EXT_pbuffer");
+  if (glewExperimental || WGLEW_EXT_pbuffer|| crippled) CONST_CAST(WGLEW_EXT_pbuffer)= !_glewInit_WGL_EXT_pbuffer(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_EXT_pbuffer */
+#ifdef WGL_EXT_pixel_format
+  CONST_CAST(WGLEW_EXT_pixel_format) = wglewGetExtension("WGL_EXT_pixel_format");
+  if (glewExperimental || WGLEW_EXT_pixel_format|| crippled) CONST_CAST(WGLEW_EXT_pixel_format)= !_glewInit_WGL_EXT_pixel_format(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_EXT_pixel_format */
+#ifdef WGL_EXT_pixel_format_packed_float
+  CONST_CAST(WGLEW_EXT_pixel_format_packed_float) = wglewGetExtension("WGL_EXT_pixel_format_packed_float");
+#endif /* WGL_EXT_pixel_format_packed_float */
+#ifdef WGL_EXT_swap_control
+  CONST_CAST(WGLEW_EXT_swap_control) = wglewGetExtension("WGL_EXT_swap_control");
+  if (glewExperimental || WGLEW_EXT_swap_control|| crippled) CONST_CAST(WGLEW_EXT_swap_control)= !_glewInit_WGL_EXT_swap_control(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_EXT_swap_control */
+#ifdef WGL_I3D_digital_video_control
+  CONST_CAST(WGLEW_I3D_digital_video_control) = wglewGetExtension("WGL_I3D_digital_video_control");
+  if (glewExperimental || WGLEW_I3D_digital_video_control|| crippled) CONST_CAST(WGLEW_I3D_digital_video_control)= !_glewInit_WGL_I3D_digital_video_control(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_I3D_digital_video_control */
+#ifdef WGL_I3D_gamma
+  CONST_CAST(WGLEW_I3D_gamma) = wglewGetExtension("WGL_I3D_gamma");
+  if (glewExperimental || WGLEW_I3D_gamma|| crippled) CONST_CAST(WGLEW_I3D_gamma)= !_glewInit_WGL_I3D_gamma(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_I3D_gamma */
+#ifdef WGL_I3D_genlock
+  CONST_CAST(WGLEW_I3D_genlock) = wglewGetExtension("WGL_I3D_genlock");
+  if (glewExperimental || WGLEW_I3D_genlock|| crippled) CONST_CAST(WGLEW_I3D_genlock)= !_glewInit_WGL_I3D_genlock(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_I3D_genlock */
+#ifdef WGL_I3D_image_buffer
+  CONST_CAST(WGLEW_I3D_image_buffer) = wglewGetExtension("WGL_I3D_image_buffer");
+  if (glewExperimental || WGLEW_I3D_image_buffer|| crippled) CONST_CAST(WGLEW_I3D_image_buffer)= !_glewInit_WGL_I3D_image_buffer(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_I3D_image_buffer */
+#ifdef WGL_I3D_swap_frame_lock
+  CONST_CAST(WGLEW_I3D_swap_frame_lock) = wglewGetExtension("WGL_I3D_swap_frame_lock");
+  if (glewExperimental || WGLEW_I3D_swap_frame_lock|| crippled) CONST_CAST(WGLEW_I3D_swap_frame_lock)= !_glewInit_WGL_I3D_swap_frame_lock(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_I3D_swap_frame_lock */
+#ifdef WGL_I3D_swap_frame_usage
+  CONST_CAST(WGLEW_I3D_swap_frame_usage) = wglewGetExtension("WGL_I3D_swap_frame_usage");
+  if (glewExperimental || WGLEW_I3D_swap_frame_usage|| crippled) CONST_CAST(WGLEW_I3D_swap_frame_usage)= !_glewInit_WGL_I3D_swap_frame_usage(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_I3D_swap_frame_usage */
+#ifdef WGL_NV_float_buffer
+  CONST_CAST(WGLEW_NV_float_buffer) = wglewGetExtension("WGL_NV_float_buffer");
+#endif /* WGL_NV_float_buffer */
+#ifdef WGL_NV_gpu_affinity
+  CONST_CAST(WGLEW_NV_gpu_affinity) = wglewGetExtension("WGL_NV_gpu_affinity");
+  if (glewExperimental || WGLEW_NV_gpu_affinity|| crippled) CONST_CAST(WGLEW_NV_gpu_affinity)= !_glewInit_WGL_NV_gpu_affinity(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_NV_gpu_affinity */
+#ifdef WGL_NV_present_video
+  CONST_CAST(WGLEW_NV_present_video) = wglewGetExtension("WGL_NV_present_video");
+  if (glewExperimental || WGLEW_NV_present_video|| crippled) CONST_CAST(WGLEW_NV_present_video)= !_glewInit_WGL_NV_present_video(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_NV_present_video */
+#ifdef WGL_NV_render_depth_texture
+  CONST_CAST(WGLEW_NV_render_depth_texture) = wglewGetExtension("WGL_NV_render_depth_texture");
+#endif /* WGL_NV_render_depth_texture */
+#ifdef WGL_NV_render_texture_rectangle
+  CONST_CAST(WGLEW_NV_render_texture_rectangle) = wglewGetExtension("WGL_NV_render_texture_rectangle");
+#endif /* WGL_NV_render_texture_rectangle */
+#ifdef WGL_NV_swap_group
+  CONST_CAST(WGLEW_NV_swap_group) = wglewGetExtension("WGL_NV_swap_group");
+  if (glewExperimental || WGLEW_NV_swap_group|| crippled) CONST_CAST(WGLEW_NV_swap_group)= !_glewInit_WGL_NV_swap_group(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_NV_swap_group */
+#ifdef WGL_NV_vertex_array_range
+  CONST_CAST(WGLEW_NV_vertex_array_range) = wglewGetExtension("WGL_NV_vertex_array_range");
+  if (glewExperimental || WGLEW_NV_vertex_array_range|| crippled) CONST_CAST(WGLEW_NV_vertex_array_range)= !_glewInit_WGL_NV_vertex_array_range(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_NV_vertex_array_range */
+#ifdef WGL_NV_video_output
+  CONST_CAST(WGLEW_NV_video_output) = wglewGetExtension("WGL_NV_video_output");
+  if (glewExperimental || WGLEW_NV_video_output|| crippled) CONST_CAST(WGLEW_NV_video_output)= !_glewInit_WGL_NV_video_output(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_NV_video_output */
+#ifdef WGL_OML_sync_control
+  CONST_CAST(WGLEW_OML_sync_control) = wglewGetExtension("WGL_OML_sync_control");
+  if (glewExperimental || WGLEW_OML_sync_control|| crippled) CONST_CAST(WGLEW_OML_sync_control)= !_glewInit_WGL_OML_sync_control(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* WGL_OML_sync_control */
+
+  return GLEW_OK;
+}
+
+#elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+
+PFNGLXGETCURRENTDISPLAYPROC __glewXGetCurrentDisplay = NULL;
+
+PFNGLXCHOOSEFBCONFIGPROC __glewXChooseFBConfig = NULL;
+PFNGLXCREATENEWCONTEXTPROC __glewXCreateNewContext = NULL;
+PFNGLXCREATEPBUFFERPROC __glewXCreatePbuffer = NULL;
+PFNGLXCREATEPIXMAPPROC __glewXCreatePixmap = NULL;
+PFNGLXCREATEWINDOWPROC __glewXCreateWindow = NULL;
+PFNGLXDESTROYPBUFFERPROC __glewXDestroyPbuffer = NULL;
+PFNGLXDESTROYPIXMAPPROC __glewXDestroyPixmap = NULL;
+PFNGLXDESTROYWINDOWPROC __glewXDestroyWindow = NULL;
+PFNGLXGETCURRENTREADDRAWABLEPROC __glewXGetCurrentReadDrawable = NULL;
+PFNGLXGETFBCONFIGATTRIBPROC __glewXGetFBConfigAttrib = NULL;
+PFNGLXGETFBCONFIGSPROC __glewXGetFBConfigs = NULL;
+PFNGLXGETSELECTEDEVENTPROC __glewXGetSelectedEvent = NULL;
+PFNGLXGETVISUALFROMFBCONFIGPROC __glewXGetVisualFromFBConfig = NULL;
+PFNGLXMAKECONTEXTCURRENTPROC __glewXMakeContextCurrent = NULL;
+PFNGLXQUERYCONTEXTPROC __glewXQueryContext = NULL;
+PFNGLXQUERYDRAWABLEPROC __glewXQueryDrawable = NULL;
+PFNGLXSELECTEVENTPROC __glewXSelectEvent = NULL;
+
+PFNGLXCREATECONTEXTATTRIBSARBPROC __glewXCreateContextAttribsARB = NULL;
+
+PFNGLXBINDTEXIMAGEATIPROC __glewXBindTexImageATI = NULL;
+PFNGLXDRAWABLEATTRIBATIPROC __glewXDrawableAttribATI = NULL;
+PFNGLXRELEASETEXIMAGEATIPROC __glewXReleaseTexImageATI = NULL;
+
+PFNGLXFREECONTEXTEXTPROC __glewXFreeContextEXT = NULL;
+PFNGLXGETCONTEXTIDEXTPROC __glewXGetContextIDEXT = NULL;
+PFNGLXIMPORTCONTEXTEXTPROC __glewXImportContextEXT = NULL;
+PFNGLXQUERYCONTEXTINFOEXTPROC __glewXQueryContextInfoEXT = NULL;
+
+PFNGLXBINDTEXIMAGEEXTPROC __glewXBindTexImageEXT = NULL;
+PFNGLXRELEASETEXIMAGEEXTPROC __glewXReleaseTexImageEXT = NULL;
+
+PFNGLXGETAGPOFFSETMESAPROC __glewXGetAGPOffsetMESA = NULL;
+
+PFNGLXCOPYSUBBUFFERMESAPROC __glewXCopySubBufferMESA = NULL;
+
+PFNGLXCREATEGLXPIXMAPMESAPROC __glewXCreateGLXPixmapMESA = NULL;
+
+PFNGLXRELEASEBUFFERSMESAPROC __glewXReleaseBuffersMESA = NULL;
+
+PFNGLXSET3DFXMODEMESAPROC __glewXSet3DfxModeMESA = NULL;
+
+PFNGLXBINDVIDEODEVICENVPROC __glewXBindVideoDeviceNV = NULL;
+PFNGLXENUMERATEVIDEODEVICESNVPROC __glewXEnumerateVideoDevicesNV = NULL;
+
+PFNGLXBINDSWAPBARRIERNVPROC __glewXBindSwapBarrierNV = NULL;
+PFNGLXJOINSWAPGROUPNVPROC __glewXJoinSwapGroupNV = NULL;
+PFNGLXQUERYFRAMECOUNTNVPROC __glewXQueryFrameCountNV = NULL;
+PFNGLXQUERYMAXSWAPGROUPSNVPROC __glewXQueryMaxSwapGroupsNV = NULL;
+PFNGLXQUERYSWAPGROUPNVPROC __glewXQuerySwapGroupNV = NULL;
+PFNGLXRESETFRAMECOUNTNVPROC __glewXResetFrameCountNV = NULL;
+
+PFNGLXALLOCATEMEMORYNVPROC __glewXAllocateMemoryNV = NULL;
+PFNGLXFREEMEMORYNVPROC __glewXFreeMemoryNV = NULL;
+
+PFNGLXBINDVIDEOIMAGENVPROC __glewXBindVideoImageNV = NULL;
+PFNGLXGETVIDEODEVICENVPROC __glewXGetVideoDeviceNV = NULL;
+PFNGLXGETVIDEOINFONVPROC __glewXGetVideoInfoNV = NULL;
+PFNGLXRELEASEVIDEODEVICENVPROC __glewXReleaseVideoDeviceNV = NULL;
+PFNGLXRELEASEVIDEOIMAGENVPROC __glewXReleaseVideoImageNV = NULL;
+PFNGLXSENDPBUFFERTOVIDEONVPROC __glewXSendPbufferToVideoNV = NULL;
+
+#ifdef GLX_OML_sync_control
+PFNGLXGETMSCRATEOMLPROC __glewXGetMscRateOML = NULL;
+PFNGLXGETSYNCVALUESOMLPROC __glewXGetSyncValuesOML = NULL;
+PFNGLXSWAPBUFFERSMSCOMLPROC __glewXSwapBuffersMscOML = NULL;
+PFNGLXWAITFORMSCOMLPROC __glewXWaitForMscOML = NULL;
+PFNGLXWAITFORSBCOMLPROC __glewXWaitForSbcOML = NULL;
+#endif
+
+PFNGLXCHOOSEFBCONFIGSGIXPROC __glewXChooseFBConfigSGIX = NULL;
+PFNGLXCREATECONTEXTWITHCONFIGSGIXPROC __glewXCreateContextWithConfigSGIX = NULL;
+PFNGLXCREATEGLXPIXMAPWITHCONFIGSGIXPROC __glewXCreateGLXPixmapWithConfigSGIX = NULL;
+PFNGLXGETFBCONFIGATTRIBSGIXPROC __glewXGetFBConfigAttribSGIX = NULL;
+PFNGLXGETFBCONFIGFROMVISUALSGIXPROC __glewXGetFBConfigFromVisualSGIX = NULL;
+PFNGLXGETVISUALFROMFBCONFIGSGIXPROC __glewXGetVisualFromFBConfigSGIX = NULL;
+
+PFNGLXBINDHYPERPIPESGIXPROC __glewXBindHyperpipeSGIX = NULL;
+PFNGLXDESTROYHYPERPIPECONFIGSGIXPROC __glewXDestroyHyperpipeConfigSGIX = NULL;
+PFNGLXHYPERPIPEATTRIBSGIXPROC __glewXHyperpipeAttribSGIX = NULL;
+PFNGLXHYPERPIPECONFIGSGIXPROC __glewXHyperpipeConfigSGIX = NULL;
+PFNGLXQUERYHYPERPIPEATTRIBSGIXPROC __glewXQueryHyperpipeAttribSGIX = NULL;
+PFNGLXQUERYHYPERPIPEBESTATTRIBSGIXPROC __glewXQueryHyperpipeBestAttribSGIX = NULL;
+PFNGLXQUERYHYPERPIPECONFIGSGIXPROC __glewXQueryHyperpipeConfigSGIX = NULL;
+PFNGLXQUERYHYPERPIPENETWORKSGIXPROC __glewXQueryHyperpipeNetworkSGIX = NULL;
+
+PFNGLXCREATEGLXPBUFFERSGIXPROC __glewXCreateGLXPbufferSGIX = NULL;
+PFNGLXDESTROYGLXPBUFFERSGIXPROC __glewXDestroyGLXPbufferSGIX = NULL;
+PFNGLXGETSELECTEDEVENTSGIXPROC __glewXGetSelectedEventSGIX = NULL;
+PFNGLXQUERYGLXPBUFFERSGIXPROC __glewXQueryGLXPbufferSGIX = NULL;
+PFNGLXSELECTEVENTSGIXPROC __glewXSelectEventSGIX = NULL;
+
+PFNGLXBINDSWAPBARRIERSGIXPROC __glewXBindSwapBarrierSGIX = NULL;
+PFNGLXQUERYMAXSWAPBARRIERSSGIXPROC __glewXQueryMaxSwapBarriersSGIX = NULL;
+
+PFNGLXJOINSWAPGROUPSGIXPROC __glewXJoinSwapGroupSGIX = NULL;
+
+PFNGLXBINDCHANNELTOWINDOWSGIXPROC __glewXBindChannelToWindowSGIX = NULL;
+PFNGLXCHANNELRECTSGIXPROC __glewXChannelRectSGIX = NULL;
+PFNGLXCHANNELRECTSYNCSGIXPROC __glewXChannelRectSyncSGIX = NULL;
+PFNGLXQUERYCHANNELDELTASSGIXPROC __glewXQueryChannelDeltasSGIX = NULL;
+PFNGLXQUERYCHANNELRECTSGIXPROC __glewXQueryChannelRectSGIX = NULL;
+
+PFNGLXCUSHIONSGIPROC __glewXCushionSGI = NULL;
+
+PFNGLXGETCURRENTREADDRAWABLESGIPROC __glewXGetCurrentReadDrawableSGI = NULL;
+PFNGLXMAKECURRENTREADSGIPROC __glewXMakeCurrentReadSGI = NULL;
+
+PFNGLXSWAPINTERVALSGIPROC __glewXSwapIntervalSGI = NULL;
+
+PFNGLXGETVIDEOSYNCSGIPROC __glewXGetVideoSyncSGI = NULL;
+PFNGLXWAITVIDEOSYNCSGIPROC __glewXWaitVideoSyncSGI = NULL;
+
+PFNGLXGETTRANSPARENTINDEXSUNPROC __glewXGetTransparentIndexSUN = NULL;
+
+PFNGLXGETVIDEORESIZESUNPROC __glewXGetVideoResizeSUN = NULL;
+PFNGLXVIDEORESIZESUNPROC __glewXVideoResizeSUN = NULL;
+
+#if !defined(GLEW_MX)
+
+GLboolean __GLXEW_VERSION_1_0 = GL_FALSE;
+GLboolean __GLXEW_VERSION_1_1 = GL_FALSE;
+GLboolean __GLXEW_VERSION_1_2 = GL_FALSE;
+GLboolean __GLXEW_VERSION_1_3 = GL_FALSE;
+GLboolean __GLXEW_VERSION_1_4 = GL_FALSE;
+GLboolean __GLXEW_3DFX_multisample = GL_FALSE;
+GLboolean __GLXEW_ARB_create_context = GL_FALSE;
+GLboolean __GLXEW_ARB_fbconfig_float = GL_FALSE;
+GLboolean __GLXEW_ARB_framebuffer_sRGB = GL_FALSE;
+GLboolean __GLXEW_ARB_get_proc_address = GL_FALSE;
+GLboolean __GLXEW_ARB_multisample = GL_FALSE;
+GLboolean __GLXEW_ATI_pixel_format_float = GL_FALSE;
+GLboolean __GLXEW_ATI_render_texture = GL_FALSE;
+GLboolean __GLXEW_EXT_fbconfig_packed_float = GL_FALSE;
+GLboolean __GLXEW_EXT_framebuffer_sRGB = GL_FALSE;
+GLboolean __GLXEW_EXT_import_context = GL_FALSE;
+GLboolean __GLXEW_EXT_scene_marker = GL_FALSE;
+GLboolean __GLXEW_EXT_texture_from_pixmap = GL_FALSE;
+GLboolean __GLXEW_EXT_visual_info = GL_FALSE;
+GLboolean __GLXEW_EXT_visual_rating = GL_FALSE;
+GLboolean __GLXEW_MESA_agp_offset = GL_FALSE;
+GLboolean __GLXEW_MESA_copy_sub_buffer = GL_FALSE;
+GLboolean __GLXEW_MESA_pixmap_colormap = GL_FALSE;
+GLboolean __GLXEW_MESA_release_buffers = GL_FALSE;
+GLboolean __GLXEW_MESA_set_3dfx_mode = GL_FALSE;
+GLboolean __GLXEW_NV_float_buffer = GL_FALSE;
+GLboolean __GLXEW_NV_present_video = GL_FALSE;
+GLboolean __GLXEW_NV_swap_group = GL_FALSE;
+GLboolean __GLXEW_NV_vertex_array_range = GL_FALSE;
+GLboolean __GLXEW_NV_video_output = GL_FALSE;
+GLboolean __GLXEW_OML_swap_method = GL_FALSE;
+#ifdef GLX_OML_sync_control
+GLboolean __GLXEW_OML_sync_control = GL_FALSE;
+#endif
+GLboolean __GLXEW_SGIS_blended_overlay = GL_FALSE;
+GLboolean __GLXEW_SGIS_color_range = GL_FALSE;
+GLboolean __GLXEW_SGIS_multisample = GL_FALSE;
+GLboolean __GLXEW_SGIS_shared_multisample = GL_FALSE;
+GLboolean __GLXEW_SGIX_fbconfig = GL_FALSE;
+GLboolean __GLXEW_SGIX_hyperpipe = GL_FALSE;
+GLboolean __GLXEW_SGIX_pbuffer = GL_FALSE;
+GLboolean __GLXEW_SGIX_swap_barrier = GL_FALSE;
+GLboolean __GLXEW_SGIX_swap_group = GL_FALSE;
+GLboolean __GLXEW_SGIX_video_resize = GL_FALSE;
+GLboolean __GLXEW_SGIX_visual_select_group = GL_FALSE;
+GLboolean __GLXEW_SGI_cushion = GL_FALSE;
+GLboolean __GLXEW_SGI_make_current_read = GL_FALSE;
+GLboolean __GLXEW_SGI_swap_control = GL_FALSE;
+GLboolean __GLXEW_SGI_video_sync = GL_FALSE;
+GLboolean __GLXEW_SUN_get_transparent_index = GL_FALSE;
+GLboolean __GLXEW_SUN_video_resize = GL_FALSE;
+
+#endif /* !GLEW_MX */
+
+#ifdef GLX_VERSION_1_2
+
+static GLboolean _glewInit_GLX_VERSION_1_2 (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXGetCurrentDisplay = (PFNGLXGETCURRENTDISPLAYPROC)glewGetProcAddress((const GLubyte*)"glXGetCurrentDisplay")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_VERSION_1_2 */
+
+#ifdef GLX_VERSION_1_3
+
+static GLboolean _glewInit_GLX_VERSION_1_3 (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXChooseFBConfig = (PFNGLXCHOOSEFBCONFIGPROC)glewGetProcAddress((const GLubyte*)"glXChooseFBConfig")) == NULL) || r;
+  r = ((glXCreateNewContext = (PFNGLXCREATENEWCONTEXTPROC)glewGetProcAddress((const GLubyte*)"glXCreateNewContext")) == NULL) || r;
+  r = ((glXCreatePbuffer = (PFNGLXCREATEPBUFFERPROC)glewGetProcAddress((const GLubyte*)"glXCreatePbuffer")) == NULL) || r;
+  r = ((glXCreatePixmap = (PFNGLXCREATEPIXMAPPROC)glewGetProcAddress((const GLubyte*)"glXCreatePixmap")) == NULL) || r;
+  r = ((glXCreateWindow = (PFNGLXCREATEWINDOWPROC)glewGetProcAddress((const GLubyte*)"glXCreateWindow")) == NULL) || r;
+  r = ((glXDestroyPbuffer = (PFNGLXDESTROYPBUFFERPROC)glewGetProcAddress((const GLubyte*)"glXDestroyPbuffer")) == NULL) || r;
+  r = ((glXDestroyPixmap = (PFNGLXDESTROYPIXMAPPROC)glewGetProcAddress((const GLubyte*)"glXDestroyPixmap")) == NULL) || r;
+  r = ((glXDestroyWindow = (PFNGLXDESTROYWINDOWPROC)glewGetProcAddress((const GLubyte*)"glXDestroyWindow")) == NULL) || r;
+  r = ((glXGetCurrentReadDrawable = (PFNGLXGETCURRENTREADDRAWABLEPROC)glewGetProcAddress((const GLubyte*)"glXGetCurrentReadDrawable")) == NULL) || r;
+  r = ((glXGetFBConfigAttrib = (PFNGLXGETFBCONFIGATTRIBPROC)glewGetProcAddress((const GLubyte*)"glXGetFBConfigAttrib")) == NULL) || r;
+  r = ((glXGetFBConfigs = (PFNGLXGETFBCONFIGSPROC)glewGetProcAddress((const GLubyte*)"glXGetFBConfigs")) == NULL) || r;
+  r = ((glXGetSelectedEvent = (PFNGLXGETSELECTEDEVENTPROC)glewGetProcAddress((const GLubyte*)"glXGetSelectedEvent")) == NULL) || r;
+  r = ((glXGetVisualFromFBConfig = (PFNGLXGETVISUALFROMFBCONFIGPROC)glewGetProcAddress((const GLubyte*)"glXGetVisualFromFBConfig")) == NULL) || r;
+  r = ((glXMakeContextCurrent = (PFNGLXMAKECONTEXTCURRENTPROC)glewGetProcAddress((const GLubyte*)"glXMakeContextCurrent")) == NULL) || r;
+  r = ((glXQueryContext = (PFNGLXQUERYCONTEXTPROC)glewGetProcAddress((const GLubyte*)"glXQueryContext")) == NULL) || r;
+  r = ((glXQueryDrawable = (PFNGLXQUERYDRAWABLEPROC)glewGetProcAddress((const GLubyte*)"glXQueryDrawable")) == NULL) || r;
+  r = ((glXSelectEvent = (PFNGLXSELECTEVENTPROC)glewGetProcAddress((const GLubyte*)"glXSelectEvent")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_VERSION_1_3 */
+
+#ifdef GLX_VERSION_1_4
+
+#endif /* GLX_VERSION_1_4 */
+
+#ifdef GLX_3DFX_multisample
+
+#endif /* GLX_3DFX_multisample */
+
+#ifdef GLX_ARB_create_context
+
+static GLboolean _glewInit_GLX_ARB_create_context (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXCreateContextAttribsARB = (PFNGLXCREATECONTEXTATTRIBSARBPROC)glewGetProcAddress((const GLubyte*)"glXCreateContextAttribsARB")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_ARB_create_context */
+
+#ifdef GLX_ARB_fbconfig_float
+
+#endif /* GLX_ARB_fbconfig_float */
+
+#ifdef GLX_ARB_framebuffer_sRGB
+
+#endif /* GLX_ARB_framebuffer_sRGB */
+
+#ifdef GLX_ARB_get_proc_address
+
+#endif /* GLX_ARB_get_proc_address */
+
+#ifdef GLX_ARB_multisample
+
+#endif /* GLX_ARB_multisample */
+
+#ifdef GLX_ATI_pixel_format_float
+
+#endif /* GLX_ATI_pixel_format_float */
+
+#ifdef GLX_ATI_render_texture
+
+static GLboolean _glewInit_GLX_ATI_render_texture (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXBindTexImageATI = (PFNGLXBINDTEXIMAGEATIPROC)glewGetProcAddress((const GLubyte*)"glXBindTexImageATI")) == NULL) || r;
+  r = ((glXDrawableAttribATI = (PFNGLXDRAWABLEATTRIBATIPROC)glewGetProcAddress((const GLubyte*)"glXDrawableAttribATI")) == NULL) || r;
+  r = ((glXReleaseTexImageATI = (PFNGLXRELEASETEXIMAGEATIPROC)glewGetProcAddress((const GLubyte*)"glXReleaseTexImageATI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_ATI_render_texture */
+
+#ifdef GLX_EXT_fbconfig_packed_float
+
+#endif /* GLX_EXT_fbconfig_packed_float */
+
+#ifdef GLX_EXT_framebuffer_sRGB
+
+#endif /* GLX_EXT_framebuffer_sRGB */
+
+#ifdef GLX_EXT_import_context
+
+static GLboolean _glewInit_GLX_EXT_import_context (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXFreeContextEXT = (PFNGLXFREECONTEXTEXTPROC)glewGetProcAddress((const GLubyte*)"glXFreeContextEXT")) == NULL) || r;
+  r = ((glXGetContextIDEXT = (PFNGLXGETCONTEXTIDEXTPROC)glewGetProcAddress((const GLubyte*)"glXGetContextIDEXT")) == NULL) || r;
+  r = ((glXImportContextEXT = (PFNGLXIMPORTCONTEXTEXTPROC)glewGetProcAddress((const GLubyte*)"glXImportContextEXT")) == NULL) || r;
+  r = ((glXQueryContextInfoEXT = (PFNGLXQUERYCONTEXTINFOEXTPROC)glewGetProcAddress((const GLubyte*)"glXQueryContextInfoEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_EXT_import_context */
+
+#ifdef GLX_EXT_scene_marker
+
+#endif /* GLX_EXT_scene_marker */
+
+#ifdef GLX_EXT_texture_from_pixmap
+
+static GLboolean _glewInit_GLX_EXT_texture_from_pixmap (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXBindTexImageEXT = (PFNGLXBINDTEXIMAGEEXTPROC)glewGetProcAddress((const GLubyte*)"glXBindTexImageEXT")) == NULL) || r;
+  r = ((glXReleaseTexImageEXT = (PFNGLXRELEASETEXIMAGEEXTPROC)glewGetProcAddress((const GLubyte*)"glXReleaseTexImageEXT")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_EXT_texture_from_pixmap */
+
+#ifdef GLX_EXT_visual_info
+
+#endif /* GLX_EXT_visual_info */
+
+#ifdef GLX_EXT_visual_rating
+
+#endif /* GLX_EXT_visual_rating */
+
+#ifdef GLX_MESA_agp_offset
+
+static GLboolean _glewInit_GLX_MESA_agp_offset (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXGetAGPOffsetMESA = (PFNGLXGETAGPOFFSETMESAPROC)glewGetProcAddress((const GLubyte*)"glXGetAGPOffsetMESA")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_MESA_agp_offset */
+
+#ifdef GLX_MESA_copy_sub_buffer
+
+static GLboolean _glewInit_GLX_MESA_copy_sub_buffer (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXCopySubBufferMESA = (PFNGLXCOPYSUBBUFFERMESAPROC)glewGetProcAddress((const GLubyte*)"glXCopySubBufferMESA")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_MESA_copy_sub_buffer */
+
+#ifdef GLX_MESA_pixmap_colormap
+
+static GLboolean _glewInit_GLX_MESA_pixmap_colormap (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXCreateGLXPixmapMESA = (PFNGLXCREATEGLXPIXMAPMESAPROC)glewGetProcAddress((const GLubyte*)"glXCreateGLXPixmapMESA")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_MESA_pixmap_colormap */
+
+#ifdef GLX_MESA_release_buffers
+
+static GLboolean _glewInit_GLX_MESA_release_buffers (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXReleaseBuffersMESA = (PFNGLXRELEASEBUFFERSMESAPROC)glewGetProcAddress((const GLubyte*)"glXReleaseBuffersMESA")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_MESA_release_buffers */
+
+#ifdef GLX_MESA_set_3dfx_mode
+
+static GLboolean _glewInit_GLX_MESA_set_3dfx_mode (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXSet3DfxModeMESA = (PFNGLXSET3DFXMODEMESAPROC)glewGetProcAddress((const GLubyte*)"glXSet3DfxModeMESA")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_MESA_set_3dfx_mode */
+
+#ifdef GLX_NV_float_buffer
+
+#endif /* GLX_NV_float_buffer */
+
+#ifdef GLX_NV_present_video
+
+static GLboolean _glewInit_GLX_NV_present_video (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXBindVideoDeviceNV = (PFNGLXBINDVIDEODEVICENVPROC)glewGetProcAddress((const GLubyte*)"glXBindVideoDeviceNV")) == NULL) || r;
+  r = ((glXEnumerateVideoDevicesNV = (PFNGLXENUMERATEVIDEODEVICESNVPROC)glewGetProcAddress((const GLubyte*)"glXEnumerateVideoDevicesNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_NV_present_video */
+
+#ifdef GLX_NV_swap_group
+
+static GLboolean _glewInit_GLX_NV_swap_group (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXBindSwapBarrierNV = (PFNGLXBINDSWAPBARRIERNVPROC)glewGetProcAddress((const GLubyte*)"glXBindSwapBarrierNV")) == NULL) || r;
+  r = ((glXJoinSwapGroupNV = (PFNGLXJOINSWAPGROUPNVPROC)glewGetProcAddress((const GLubyte*)"glXJoinSwapGroupNV")) == NULL) || r;
+  r = ((glXQueryFrameCountNV = (PFNGLXQUERYFRAMECOUNTNVPROC)glewGetProcAddress((const GLubyte*)"glXQueryFrameCountNV")) == NULL) || r;
+  r = ((glXQueryMaxSwapGroupsNV = (PFNGLXQUERYMAXSWAPGROUPSNVPROC)glewGetProcAddress((const GLubyte*)"glXQueryMaxSwapGroupsNV")) == NULL) || r;
+  r = ((glXQuerySwapGroupNV = (PFNGLXQUERYSWAPGROUPNVPROC)glewGetProcAddress((const GLubyte*)"glXQuerySwapGroupNV")) == NULL) || r;
+  r = ((glXResetFrameCountNV = (PFNGLXRESETFRAMECOUNTNVPROC)glewGetProcAddress((const GLubyte*)"glXResetFrameCountNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_NV_swap_group */
+
+#ifdef GLX_NV_vertex_array_range
+
+static GLboolean _glewInit_GLX_NV_vertex_array_range (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXAllocateMemoryNV = (PFNGLXALLOCATEMEMORYNVPROC)glewGetProcAddress((const GLubyte*)"glXAllocateMemoryNV")) == NULL) || r;
+  r = ((glXFreeMemoryNV = (PFNGLXFREEMEMORYNVPROC)glewGetProcAddress((const GLubyte*)"glXFreeMemoryNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_NV_vertex_array_range */
+
+#ifdef GLX_NV_video_output
+
+static GLboolean _glewInit_GLX_NV_video_output (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXBindVideoImageNV = (PFNGLXBINDVIDEOIMAGENVPROC)glewGetProcAddress((const GLubyte*)"glXBindVideoImageNV")) == NULL) || r;
+  r = ((glXGetVideoDeviceNV = (PFNGLXGETVIDEODEVICENVPROC)glewGetProcAddress((const GLubyte*)"glXGetVideoDeviceNV")) == NULL) || r;
+  r = ((glXGetVideoInfoNV = (PFNGLXGETVIDEOINFONVPROC)glewGetProcAddress((const GLubyte*)"glXGetVideoInfoNV")) == NULL) || r;
+  r = ((glXReleaseVideoDeviceNV = (PFNGLXRELEASEVIDEODEVICENVPROC)glewGetProcAddress((const GLubyte*)"glXReleaseVideoDeviceNV")) == NULL) || r;
+  r = ((glXReleaseVideoImageNV = (PFNGLXRELEASEVIDEOIMAGENVPROC)glewGetProcAddress((const GLubyte*)"glXReleaseVideoImageNV")) == NULL) || r;
+  r = ((glXSendPbufferToVideoNV = (PFNGLXSENDPBUFFERTOVIDEONVPROC)glewGetProcAddress((const GLubyte*)"glXSendPbufferToVideoNV")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_NV_video_output */
+
+#ifdef GLX_OML_swap_method
+
+#endif /* GLX_OML_swap_method */
+
+#if defined(GLX_OML_sync_control) && defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)
+#include <inttypes.h>
+
+static GLboolean _glewInit_GLX_OML_sync_control (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXGetMscRateOML = (PFNGLXGETMSCRATEOMLPROC)glewGetProcAddress((const GLubyte*)"glXGetMscRateOML")) == NULL) || r;
+  r = ((glXGetSyncValuesOML = (PFNGLXGETSYNCVALUESOMLPROC)glewGetProcAddress((const GLubyte*)"glXGetSyncValuesOML")) == NULL) || r;
+  r = ((glXSwapBuffersMscOML = (PFNGLXSWAPBUFFERSMSCOMLPROC)glewGetProcAddress((const GLubyte*)"glXSwapBuffersMscOML")) == NULL) || r;
+  r = ((glXWaitForMscOML = (PFNGLXWAITFORMSCOMLPROC)glewGetProcAddress((const GLubyte*)"glXWaitForMscOML")) == NULL) || r;
+  r = ((glXWaitForSbcOML = (PFNGLXWAITFORSBCOMLPROC)glewGetProcAddress((const GLubyte*)"glXWaitForSbcOML")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_OML_sync_control */
+
+#ifdef GLX_SGIS_blended_overlay
+
+#endif /* GLX_SGIS_blended_overlay */
+
+#ifdef GLX_SGIS_color_range
+
+#endif /* GLX_SGIS_color_range */
+
+#ifdef GLX_SGIS_multisample
+
+#endif /* GLX_SGIS_multisample */
+
+#ifdef GLX_SGIS_shared_multisample
+
+#endif /* GLX_SGIS_shared_multisample */
+
+#ifdef GLX_SGIX_fbconfig
+
+static GLboolean _glewInit_GLX_SGIX_fbconfig (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXChooseFBConfigSGIX = (PFNGLXCHOOSEFBCONFIGSGIXPROC)glewGetProcAddress((const GLubyte*)"glXChooseFBConfigSGIX")) == NULL) || r;
+  r = ((glXCreateContextWithConfigSGIX = (PFNGLXCREATECONTEXTWITHCONFIGSGIXPROC)glewGetProcAddress((const GLubyte*)"glXCreateContextWithConfigSGIX")) == NULL) || r;
+  r = ((glXCreateGLXPixmapWithConfigSGIX = (PFNGLXCREATEGLXPIXMAPWITHCONFIGSGIXPROC)glewGetProcAddress((const GLubyte*)"glXCreateGLXPixmapWithConfigSGIX")) == NULL) || r;
+  r = ((glXGetFBConfigAttribSGIX = (PFNGLXGETFBCONFIGATTRIBSGIXPROC)glewGetProcAddress((const GLubyte*)"glXGetFBConfigAttribSGIX")) == NULL) || r;
+  r = ((glXGetFBConfigFromVisualSGIX = (PFNGLXGETFBCONFIGFROMVISUALSGIXPROC)glewGetProcAddress((const GLubyte*)"glXGetFBConfigFromVisualSGIX")) == NULL) || r;
+  r = ((glXGetVisualFromFBConfigSGIX = (PFNGLXGETVISUALFROMFBCONFIGSGIXPROC)glewGetProcAddress((const GLubyte*)"glXGetVisualFromFBConfigSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SGIX_fbconfig */
+
+#ifdef GLX_SGIX_hyperpipe
+
+static GLboolean _glewInit_GLX_SGIX_hyperpipe (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXBindHyperpipeSGIX = (PFNGLXBINDHYPERPIPESGIXPROC)glewGetProcAddress((const GLubyte*)"glXBindHyperpipeSGIX")) == NULL) || r;
+  r = ((glXDestroyHyperpipeConfigSGIX = (PFNGLXDESTROYHYPERPIPECONFIGSGIXPROC)glewGetProcAddress((const GLubyte*)"glXDestroyHyperpipeConfigSGIX")) == NULL) || r;
+  r = ((glXHyperpipeAttribSGIX = (PFNGLXHYPERPIPEATTRIBSGIXPROC)glewGetProcAddress((const GLubyte*)"glXHyperpipeAttribSGIX")) == NULL) || r;
+  r = ((glXHyperpipeConfigSGIX = (PFNGLXHYPERPIPECONFIGSGIXPROC)glewGetProcAddress((const GLubyte*)"glXHyperpipeConfigSGIX")) == NULL) || r;
+  r = ((glXQueryHyperpipeAttribSGIX = (PFNGLXQUERYHYPERPIPEATTRIBSGIXPROC)glewGetProcAddress((const GLubyte*)"glXQueryHyperpipeAttribSGIX")) == NULL) || r;
+  r = ((glXQueryHyperpipeBestAttribSGIX = (PFNGLXQUERYHYPERPIPEBESTATTRIBSGIXPROC)glewGetProcAddress((const GLubyte*)"glXQueryHyperpipeBestAttribSGIX")) == NULL) || r;
+  r = ((glXQueryHyperpipeConfigSGIX = (PFNGLXQUERYHYPERPIPECONFIGSGIXPROC)glewGetProcAddress((const GLubyte*)"glXQueryHyperpipeConfigSGIX")) == NULL) || r;
+  r = ((glXQueryHyperpipeNetworkSGIX = (PFNGLXQUERYHYPERPIPENETWORKSGIXPROC)glewGetProcAddress((const GLubyte*)"glXQueryHyperpipeNetworkSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SGIX_hyperpipe */
+
+#ifdef GLX_SGIX_pbuffer
+
+static GLboolean _glewInit_GLX_SGIX_pbuffer (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXCreateGLXPbufferSGIX = (PFNGLXCREATEGLXPBUFFERSGIXPROC)glewGetProcAddress((const GLubyte*)"glXCreateGLXPbufferSGIX")) == NULL) || r;
+  r = ((glXDestroyGLXPbufferSGIX = (PFNGLXDESTROYGLXPBUFFERSGIXPROC)glewGetProcAddress((const GLubyte*)"glXDestroyGLXPbufferSGIX")) == NULL) || r;
+  r = ((glXGetSelectedEventSGIX = (PFNGLXGETSELECTEDEVENTSGIXPROC)glewGetProcAddress((const GLubyte*)"glXGetSelectedEventSGIX")) == NULL) || r;
+  r = ((glXQueryGLXPbufferSGIX = (PFNGLXQUERYGLXPBUFFERSGIXPROC)glewGetProcAddress((const GLubyte*)"glXQueryGLXPbufferSGIX")) == NULL) || r;
+  r = ((glXSelectEventSGIX = (PFNGLXSELECTEVENTSGIXPROC)glewGetProcAddress((const GLubyte*)"glXSelectEventSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SGIX_pbuffer */
+
+#ifdef GLX_SGIX_swap_barrier
+
+static GLboolean _glewInit_GLX_SGIX_swap_barrier (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXBindSwapBarrierSGIX = (PFNGLXBINDSWAPBARRIERSGIXPROC)glewGetProcAddress((const GLubyte*)"glXBindSwapBarrierSGIX")) == NULL) || r;
+  r = ((glXQueryMaxSwapBarriersSGIX = (PFNGLXQUERYMAXSWAPBARRIERSSGIXPROC)glewGetProcAddress((const GLubyte*)"glXQueryMaxSwapBarriersSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SGIX_swap_barrier */
+
+#ifdef GLX_SGIX_swap_group
+
+static GLboolean _glewInit_GLX_SGIX_swap_group (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXJoinSwapGroupSGIX = (PFNGLXJOINSWAPGROUPSGIXPROC)glewGetProcAddress((const GLubyte*)"glXJoinSwapGroupSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SGIX_swap_group */
+
+#ifdef GLX_SGIX_video_resize
+
+static GLboolean _glewInit_GLX_SGIX_video_resize (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXBindChannelToWindowSGIX = (PFNGLXBINDCHANNELTOWINDOWSGIXPROC)glewGetProcAddress((const GLubyte*)"glXBindChannelToWindowSGIX")) == NULL) || r;
+  r = ((glXChannelRectSGIX = (PFNGLXCHANNELRECTSGIXPROC)glewGetProcAddress((const GLubyte*)"glXChannelRectSGIX")) == NULL) || r;
+  r = ((glXChannelRectSyncSGIX = (PFNGLXCHANNELRECTSYNCSGIXPROC)glewGetProcAddress((const GLubyte*)"glXChannelRectSyncSGIX")) == NULL) || r;
+  r = ((glXQueryChannelDeltasSGIX = (PFNGLXQUERYCHANNELDELTASSGIXPROC)glewGetProcAddress((const GLubyte*)"glXQueryChannelDeltasSGIX")) == NULL) || r;
+  r = ((glXQueryChannelRectSGIX = (PFNGLXQUERYCHANNELRECTSGIXPROC)glewGetProcAddress((const GLubyte*)"glXQueryChannelRectSGIX")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SGIX_video_resize */
+
+#ifdef GLX_SGIX_visual_select_group
+
+#endif /* GLX_SGIX_visual_select_group */
+
+#ifdef GLX_SGI_cushion
+
+static GLboolean _glewInit_GLX_SGI_cushion (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXCushionSGI = (PFNGLXCUSHIONSGIPROC)glewGetProcAddress((const GLubyte*)"glXCushionSGI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SGI_cushion */
+
+#ifdef GLX_SGI_make_current_read
+
+static GLboolean _glewInit_GLX_SGI_make_current_read (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXGetCurrentReadDrawableSGI = (PFNGLXGETCURRENTREADDRAWABLESGIPROC)glewGetProcAddress((const GLubyte*)"glXGetCurrentReadDrawableSGI")) == NULL) || r;
+  r = ((glXMakeCurrentReadSGI = (PFNGLXMAKECURRENTREADSGIPROC)glewGetProcAddress((const GLubyte*)"glXMakeCurrentReadSGI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SGI_make_current_read */
+
+#ifdef GLX_SGI_swap_control
+
+static GLboolean _glewInit_GLX_SGI_swap_control (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXSwapIntervalSGI = (PFNGLXSWAPINTERVALSGIPROC)glewGetProcAddress((const GLubyte*)"glXSwapIntervalSGI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SGI_swap_control */
+
+#ifdef GLX_SGI_video_sync
+
+static GLboolean _glewInit_GLX_SGI_video_sync (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXGetVideoSyncSGI = (PFNGLXGETVIDEOSYNCSGIPROC)glewGetProcAddress((const GLubyte*)"glXGetVideoSyncSGI")) == NULL) || r;
+  r = ((glXWaitVideoSyncSGI = (PFNGLXWAITVIDEOSYNCSGIPROC)glewGetProcAddress((const GLubyte*)"glXWaitVideoSyncSGI")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SGI_video_sync */
+
+#ifdef GLX_SUN_get_transparent_index
+
+static GLboolean _glewInit_GLX_SUN_get_transparent_index (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXGetTransparentIndexSUN = (PFNGLXGETTRANSPARENTINDEXSUNPROC)glewGetProcAddress((const GLubyte*)"glXGetTransparentIndexSUN")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SUN_get_transparent_index */
+
+#ifdef GLX_SUN_video_resize
+
+static GLboolean _glewInit_GLX_SUN_video_resize (GLXEW_CONTEXT_ARG_DEF_INIT)
+{
+  GLboolean r = GL_FALSE;
+
+  r = ((glXGetVideoResizeSUN = (PFNGLXGETVIDEORESIZESUNPROC)glewGetProcAddress((const GLubyte*)"glXGetVideoResizeSUN")) == NULL) || r;
+  r = ((glXVideoResizeSUN = (PFNGLXVIDEORESIZESUNPROC)glewGetProcAddress((const GLubyte*)"glXVideoResizeSUN")) == NULL) || r;
+
+  return r;
+}
+
+#endif /* GLX_SUN_video_resize */
+
+/* ------------------------------------------------------------------------ */
+
+GLboolean glxewGetExtension (const char* name)
+{    
+  GLubyte* p;
+  GLubyte* end;
+  GLuint len = _glewStrLen((const GLubyte*)name);
+/*   if (glXQueryExtensionsString == NULL || glXGetCurrentDisplay == NULL) return GL_FALSE; */
+/*   p = (GLubyte*)glXQueryExtensionsString(glXGetCurrentDisplay(), DefaultScreen(glXGetCurrentDisplay())); */
+  if (glXGetClientString == NULL || glXGetCurrentDisplay == NULL) return GL_FALSE;
+  p = (GLubyte*)glXGetClientString(glXGetCurrentDisplay(), GLX_EXTENSIONS);
+  if (0 == p) return GL_FALSE;
+  end = p + _glewStrLen(p);
+  while (p < end)
+  {
+    GLuint n = _glewStrCLen(p, ' ');
+    if (len == n && _glewStrSame((const GLubyte*)name, p, n)) return GL_TRUE;
+    p += n+1;
+  }
+  return GL_FALSE;
+}
+
+GLenum glxewContextInit (GLXEW_CONTEXT_ARG_DEF_LIST)
+{
+  int major, minor;
+  /* initialize core GLX 1.2 */
+  if (_glewInit_GLX_VERSION_1_2(GLEW_CONTEXT_ARG_VAR_INIT)) return GLEW_ERROR_GLX_VERSION_11_ONLY;
+  /* initialize flags */
+  CONST_CAST(GLXEW_VERSION_1_0) = GL_TRUE;
+  CONST_CAST(GLXEW_VERSION_1_1) = GL_TRUE;
+  CONST_CAST(GLXEW_VERSION_1_2) = GL_TRUE;
+  CONST_CAST(GLXEW_VERSION_1_3) = GL_TRUE;
+  CONST_CAST(GLXEW_VERSION_1_4) = GL_TRUE;
+  /* query GLX version */
+  glXQueryVersion(glXGetCurrentDisplay(), &major, &minor);
+  if (major == 1 && minor <= 3)
+  {
+    switch (minor)
+    {
+      case 3:
+      CONST_CAST(GLXEW_VERSION_1_4) = GL_FALSE;
+      break;
+      case 2:
+      CONST_CAST(GLXEW_VERSION_1_4) = GL_FALSE;
+      CONST_CAST(GLXEW_VERSION_1_3) = GL_FALSE;
+      break;
+      default:
+      return GLEW_ERROR_GLX_VERSION_11_ONLY;
+      break;
+    }
+  }
+  /* initialize extensions */
+#ifdef GLX_VERSION_1_3
+  if (glewExperimental || GLXEW_VERSION_1_3) CONST_CAST(GLXEW_VERSION_1_3) = !_glewInit_GLX_VERSION_1_3(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_VERSION_1_3 */
+#ifdef GLX_3DFX_multisample
+  CONST_CAST(GLXEW_3DFX_multisample) = glxewGetExtension("GLX_3DFX_multisample");
+#endif /* GLX_3DFX_multisample */
+#ifdef GLX_ARB_create_context
+  CONST_CAST(GLXEW_ARB_create_context) = glxewGetExtension("GLX_ARB_create_context");
+  if (glewExperimental || GLXEW_ARB_create_context) CONST_CAST(GLXEW_ARB_create_context) = !_glewInit_GLX_ARB_create_context(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_ARB_create_context */
+#ifdef GLX_ARB_fbconfig_float
+  CONST_CAST(GLXEW_ARB_fbconfig_float) = glxewGetExtension("GLX_ARB_fbconfig_float");
+#endif /* GLX_ARB_fbconfig_float */
+#ifdef GLX_ARB_framebuffer_sRGB
+  CONST_CAST(GLXEW_ARB_framebuffer_sRGB) = glxewGetExtension("GLX_ARB_framebuffer_sRGB");
+#endif /* GLX_ARB_framebuffer_sRGB */
+#ifdef GLX_ARB_get_proc_address
+  CONST_CAST(GLXEW_ARB_get_proc_address) = glxewGetExtension("GLX_ARB_get_proc_address");
+#endif /* GLX_ARB_get_proc_address */
+#ifdef GLX_ARB_multisample
+  CONST_CAST(GLXEW_ARB_multisample) = glxewGetExtension("GLX_ARB_multisample");
+#endif /* GLX_ARB_multisample */
+#ifdef GLX_ATI_pixel_format_float
+  CONST_CAST(GLXEW_ATI_pixel_format_float) = glxewGetExtension("GLX_ATI_pixel_format_float");
+#endif /* GLX_ATI_pixel_format_float */
+#ifdef GLX_ATI_render_texture
+  CONST_CAST(GLXEW_ATI_render_texture) = glxewGetExtension("GLX_ATI_render_texture");
+  if (glewExperimental || GLXEW_ATI_render_texture) CONST_CAST(GLXEW_ATI_render_texture) = !_glewInit_GLX_ATI_render_texture(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_ATI_render_texture */
+#ifdef GLX_EXT_fbconfig_packed_float
+  CONST_CAST(GLXEW_EXT_fbconfig_packed_float) = glxewGetExtension("GLX_EXT_fbconfig_packed_float");
+#endif /* GLX_EXT_fbconfig_packed_float */
+#ifdef GLX_EXT_framebuffer_sRGB
+  CONST_CAST(GLXEW_EXT_framebuffer_sRGB) = glxewGetExtension("GLX_EXT_framebuffer_sRGB");
+#endif /* GLX_EXT_framebuffer_sRGB */
+#ifdef GLX_EXT_import_context
+  CONST_CAST(GLXEW_EXT_import_context) = glxewGetExtension("GLX_EXT_import_context");
+  if (glewExperimental || GLXEW_EXT_import_context) CONST_CAST(GLXEW_EXT_import_context) = !_glewInit_GLX_EXT_import_context(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_EXT_import_context */
+#ifdef GLX_EXT_scene_marker
+  CONST_CAST(GLXEW_EXT_scene_marker) = glxewGetExtension("GLX_EXT_scene_marker");
+#endif /* GLX_EXT_scene_marker */
+#ifdef GLX_EXT_texture_from_pixmap
+  CONST_CAST(GLXEW_EXT_texture_from_pixmap) = glxewGetExtension("GLX_EXT_texture_from_pixmap");
+  if (glewExperimental || GLXEW_EXT_texture_from_pixmap) CONST_CAST(GLXEW_EXT_texture_from_pixmap) = !_glewInit_GLX_EXT_texture_from_pixmap(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_EXT_texture_from_pixmap */
+#ifdef GLX_EXT_visual_info
+  CONST_CAST(GLXEW_EXT_visual_info) = glxewGetExtension("GLX_EXT_visual_info");
+#endif /* GLX_EXT_visual_info */
+#ifdef GLX_EXT_visual_rating
+  CONST_CAST(GLXEW_EXT_visual_rating) = glxewGetExtension("GLX_EXT_visual_rating");
+#endif /* GLX_EXT_visual_rating */
+#ifdef GLX_MESA_agp_offset
+  CONST_CAST(GLXEW_MESA_agp_offset) = glxewGetExtension("GLX_MESA_agp_offset");
+  if (glewExperimental || GLXEW_MESA_agp_offset) CONST_CAST(GLXEW_MESA_agp_offset) = !_glewInit_GLX_MESA_agp_offset(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_MESA_agp_offset */
+#ifdef GLX_MESA_copy_sub_buffer
+  CONST_CAST(GLXEW_MESA_copy_sub_buffer) = glxewGetExtension("GLX_MESA_copy_sub_buffer");
+  if (glewExperimental || GLXEW_MESA_copy_sub_buffer) CONST_CAST(GLXEW_MESA_copy_sub_buffer) = !_glewInit_GLX_MESA_copy_sub_buffer(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_MESA_copy_sub_buffer */
+#ifdef GLX_MESA_pixmap_colormap
+  CONST_CAST(GLXEW_MESA_pixmap_colormap) = glxewGetExtension("GLX_MESA_pixmap_colormap");
+  if (glewExperimental || GLXEW_MESA_pixmap_colormap) CONST_CAST(GLXEW_MESA_pixmap_colormap) = !_glewInit_GLX_MESA_pixmap_colormap(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_MESA_pixmap_colormap */
+#ifdef GLX_MESA_release_buffers
+  CONST_CAST(GLXEW_MESA_release_buffers) = glxewGetExtension("GLX_MESA_release_buffers");
+  if (glewExperimental || GLXEW_MESA_release_buffers) CONST_CAST(GLXEW_MESA_release_buffers) = !_glewInit_GLX_MESA_release_buffers(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_MESA_release_buffers */
+#ifdef GLX_MESA_set_3dfx_mode
+  CONST_CAST(GLXEW_MESA_set_3dfx_mode) = glxewGetExtension("GLX_MESA_set_3dfx_mode");
+  if (glewExperimental || GLXEW_MESA_set_3dfx_mode) CONST_CAST(GLXEW_MESA_set_3dfx_mode) = !_glewInit_GLX_MESA_set_3dfx_mode(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_MESA_set_3dfx_mode */
+#ifdef GLX_NV_float_buffer
+  CONST_CAST(GLXEW_NV_float_buffer) = glxewGetExtension("GLX_NV_float_buffer");
+#endif /* GLX_NV_float_buffer */
+#ifdef GLX_NV_present_video
+  CONST_CAST(GLXEW_NV_present_video) = glxewGetExtension("GLX_NV_present_video");
+  if (glewExperimental || GLXEW_NV_present_video) CONST_CAST(GLXEW_NV_present_video) = !_glewInit_GLX_NV_present_video(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_NV_present_video */
+#ifdef GLX_NV_swap_group
+  CONST_CAST(GLXEW_NV_swap_group) = glxewGetExtension("GLX_NV_swap_group");
+  if (glewExperimental || GLXEW_NV_swap_group) CONST_CAST(GLXEW_NV_swap_group) = !_glewInit_GLX_NV_swap_group(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_NV_swap_group */
+#ifdef GLX_NV_vertex_array_range
+  CONST_CAST(GLXEW_NV_vertex_array_range) = glxewGetExtension("GLX_NV_vertex_array_range");
+  if (glewExperimental || GLXEW_NV_vertex_array_range) CONST_CAST(GLXEW_NV_vertex_array_range) = !_glewInit_GLX_NV_vertex_array_range(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_NV_vertex_array_range */
+#ifdef GLX_NV_video_output
+  CONST_CAST(GLXEW_NV_video_output) = glxewGetExtension("GLX_NV_video_output");
+  if (glewExperimental || GLXEW_NV_video_output) CONST_CAST(GLXEW_NV_video_output) = !_glewInit_GLX_NV_video_output(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_NV_video_output */
+#ifdef GLX_OML_swap_method
+  CONST_CAST(GLXEW_OML_swap_method) = glxewGetExtension("GLX_OML_swap_method");
+#endif /* GLX_OML_swap_method */
+#if defined(GLX_OML_sync_control) && defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)
+#include <inttypes.h>
+  CONST_CAST(GLXEW_OML_sync_control) = glxewGetExtension("GLX_OML_sync_control");
+  if (glewExperimental || GLXEW_OML_sync_control) CONST_CAST(GLXEW_OML_sync_control) = !_glewInit_GLX_OML_sync_control(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_OML_sync_control */
+#ifdef GLX_SGIS_blended_overlay
+  CONST_CAST(GLXEW_SGIS_blended_overlay) = glxewGetExtension("GLX_SGIS_blended_overlay");
+#endif /* GLX_SGIS_blended_overlay */
+#ifdef GLX_SGIS_color_range
+  CONST_CAST(GLXEW_SGIS_color_range) = glxewGetExtension("GLX_SGIS_color_range");
+#endif /* GLX_SGIS_color_range */
+#ifdef GLX_SGIS_multisample
+  CONST_CAST(GLXEW_SGIS_multisample) = glxewGetExtension("GLX_SGIS_multisample");
+#endif /* GLX_SGIS_multisample */
+#ifdef GLX_SGIS_shared_multisample
+  CONST_CAST(GLXEW_SGIS_shared_multisample) = glxewGetExtension("GLX_SGIS_shared_multisample");
+#endif /* GLX_SGIS_shared_multisample */
+#ifdef GLX_SGIX_fbconfig
+  CONST_CAST(GLXEW_SGIX_fbconfig) = glxewGetExtension("GLX_SGIX_fbconfig");
+  if (glewExperimental || GLXEW_SGIX_fbconfig) CONST_CAST(GLXEW_SGIX_fbconfig) = !_glewInit_GLX_SGIX_fbconfig(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SGIX_fbconfig */
+#ifdef GLX_SGIX_hyperpipe
+  CONST_CAST(GLXEW_SGIX_hyperpipe) = glxewGetExtension("GLX_SGIX_hyperpipe");
+  if (glewExperimental || GLXEW_SGIX_hyperpipe) CONST_CAST(GLXEW_SGIX_hyperpipe) = !_glewInit_GLX_SGIX_hyperpipe(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SGIX_hyperpipe */
+#ifdef GLX_SGIX_pbuffer
+  CONST_CAST(GLXEW_SGIX_pbuffer) = glxewGetExtension("GLX_SGIX_pbuffer");
+  if (glewExperimental || GLXEW_SGIX_pbuffer) CONST_CAST(GLXEW_SGIX_pbuffer) = !_glewInit_GLX_SGIX_pbuffer(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SGIX_pbuffer */
+#ifdef GLX_SGIX_swap_barrier
+  CONST_CAST(GLXEW_SGIX_swap_barrier) = glxewGetExtension("GLX_SGIX_swap_barrier");
+  if (glewExperimental || GLXEW_SGIX_swap_barrier) CONST_CAST(GLXEW_SGIX_swap_barrier) = !_glewInit_GLX_SGIX_swap_barrier(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SGIX_swap_barrier */
+#ifdef GLX_SGIX_swap_group
+  CONST_CAST(GLXEW_SGIX_swap_group) = glxewGetExtension("GLX_SGIX_swap_group");
+  if (glewExperimental || GLXEW_SGIX_swap_group) CONST_CAST(GLXEW_SGIX_swap_group) = !_glewInit_GLX_SGIX_swap_group(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SGIX_swap_group */
+#ifdef GLX_SGIX_video_resize
+  CONST_CAST(GLXEW_SGIX_video_resize) = glxewGetExtension("GLX_SGIX_video_resize");
+  if (glewExperimental || GLXEW_SGIX_video_resize) CONST_CAST(GLXEW_SGIX_video_resize) = !_glewInit_GLX_SGIX_video_resize(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SGIX_video_resize */
+#ifdef GLX_SGIX_visual_select_group
+  CONST_CAST(GLXEW_SGIX_visual_select_group) = glxewGetExtension("GLX_SGIX_visual_select_group");
+#endif /* GLX_SGIX_visual_select_group */
+#ifdef GLX_SGI_cushion
+  CONST_CAST(GLXEW_SGI_cushion) = glxewGetExtension("GLX_SGI_cushion");
+  if (glewExperimental || GLXEW_SGI_cushion) CONST_CAST(GLXEW_SGI_cushion) = !_glewInit_GLX_SGI_cushion(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SGI_cushion */
+#ifdef GLX_SGI_make_current_read
+  CONST_CAST(GLXEW_SGI_make_current_read) = glxewGetExtension("GLX_SGI_make_current_read");
+  if (glewExperimental || GLXEW_SGI_make_current_read) CONST_CAST(GLXEW_SGI_make_current_read) = !_glewInit_GLX_SGI_make_current_read(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SGI_make_current_read */
+#ifdef GLX_SGI_swap_control
+  CONST_CAST(GLXEW_SGI_swap_control) = glxewGetExtension("GLX_SGI_swap_control");
+  if (glewExperimental || GLXEW_SGI_swap_control) CONST_CAST(GLXEW_SGI_swap_control) = !_glewInit_GLX_SGI_swap_control(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SGI_swap_control */
+#ifdef GLX_SGI_video_sync
+  CONST_CAST(GLXEW_SGI_video_sync) = glxewGetExtension("GLX_SGI_video_sync");
+  if (glewExperimental || GLXEW_SGI_video_sync) CONST_CAST(GLXEW_SGI_video_sync) = !_glewInit_GLX_SGI_video_sync(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SGI_video_sync */
+#ifdef GLX_SUN_get_transparent_index
+  CONST_CAST(GLXEW_SUN_get_transparent_index) = glxewGetExtension("GLX_SUN_get_transparent_index");
+  if (glewExperimental || GLXEW_SUN_get_transparent_index) CONST_CAST(GLXEW_SUN_get_transparent_index) = !_glewInit_GLX_SUN_get_transparent_index(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SUN_get_transparent_index */
+#ifdef GLX_SUN_video_resize
+  CONST_CAST(GLXEW_SUN_video_resize) = glxewGetExtension("GLX_SUN_video_resize");
+  if (glewExperimental || GLXEW_SUN_video_resize) CONST_CAST(GLXEW_SUN_video_resize) = !_glewInit_GLX_SUN_video_resize(GLEW_CONTEXT_ARG_VAR_INIT);
+#endif /* GLX_SUN_video_resize */
+
+  return GLEW_OK;
+}
+
+#endif /* !__APPLE__ || GLEW_APPLE_GLX */
+
+/* ------------------------------------------------------------------------ */
+
+const GLubyte* glewGetErrorString (GLenum error)
+{
+  static const GLubyte* _glewErrorString[] =
+  {
+    (const GLubyte*)"No error",
+    (const GLubyte*)"Missing GL version",
+    (const GLubyte*)"GL 1.1 and up are not supported",
+    (const GLubyte*)"GLX 1.2 and up are not supported",
+    (const GLubyte*)"Unknown error"
+  };
+  const int max_error = sizeof(_glewErrorString)/sizeof(*_glewErrorString) - 1;
+  return _glewErrorString[(int)error > max_error ? max_error : (int)error];
+}
+
+const GLubyte* glewGetString (GLenum name)
+{
+  static const GLubyte* _glewString[] =
+  {
+    (const GLubyte*)NULL,
+    (const GLubyte*)"1.5.1",
+    (const GLubyte*)"1",
+    (const GLubyte*)"5",
+    (const GLubyte*)"1"
+  };
+  const int max_string = sizeof(_glewString)/sizeof(*_glewString) - 1;
+  return _glewString[(int)name > max_string ? 0 : (int)name];
+}
+
+/* ------------------------------------------------------------------------ */
+
+GLboolean glewExperimental = GL_FALSE;
+
+#if !defined(GLEW_MX)
+
+#if defined(_WIN32)
+extern GLenum wglewContextInit (void);
+#elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX) /* _UNIX */
+extern GLenum glxewContextInit (void);
+#endif /* _WIN32 */
+
+GLenum glewInit ()
+{
+  GLenum r;
+  if ( (r = glewContextInit()) ) return r;
+#if defined(_WIN32)
+  return wglewContextInit();
+#elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX) /* _UNIX */
+  return glxewContextInit();
+#else
+  return r;
+#endif /* _WIN32 */
+}
+
+#endif /* !GLEW_MX */
+#ifdef GLEW_MX
+GLboolean glewContextIsSupported (GLEWContext* ctx, const char* name)
+#else
+GLboolean glewIsSupported (const char* name)
+#endif
+{
+  GLubyte* pos = (GLubyte*)name;
+  GLuint len = _glewStrLen(pos);
+  GLboolean ret = GL_TRUE;
+  while (ret && len > 0)
+  {
+    if (_glewStrSame1(&pos, &len, (const GLubyte*)"GL_", 3))
+    {
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"VERSION_", 8))
+      {
+#ifdef GL_VERSION_1_2
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"1_2", 3))
+        {
+          ret = GLEW_VERSION_1_2;
+          continue;
+        }
+#endif
+#ifdef GL_VERSION_1_3
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"1_3", 3))
+        {
+          ret = GLEW_VERSION_1_3;
+          continue;
+        }
+#endif
+#ifdef GL_VERSION_1_4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"1_4", 3))
+        {
+          ret = GLEW_VERSION_1_4;
+          continue;
+        }
+#endif
+#ifdef GL_VERSION_1_5
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"1_5", 3))
+        {
+          ret = GLEW_VERSION_1_5;
+          continue;
+        }
+#endif
+#ifdef GL_VERSION_2_0
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"2_0", 3))
+        {
+          ret = GLEW_VERSION_2_0;
+          continue;
+        }
+#endif
+#ifdef GL_VERSION_2_1
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"2_1", 3))
+        {
+          ret = GLEW_VERSION_2_1;
+          continue;
+        }
+#endif
+#ifdef GL_VERSION_3_0
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"3_0", 3))
+        {
+          ret = GLEW_VERSION_3_0;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"3DFX_", 5))
+      {
+#ifdef GL_3DFX_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample", 11))
+        {
+          ret = GLEW_3DFX_multisample;
+          continue;
+        }
+#endif
+#ifdef GL_3DFX_tbuffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"tbuffer", 7))
+        {
+          ret = GLEW_3DFX_tbuffer;
+          continue;
+        }
+#endif
+#ifdef GL_3DFX_texture_compression_FXT1
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_compression_FXT1", 24))
+        {
+          ret = GLEW_3DFX_texture_compression_FXT1;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"APPLE_", 6))
+      {
+#ifdef GL_APPLE_client_storage
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"client_storage", 14))
+        {
+          ret = GLEW_APPLE_client_storage;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_element_array
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"element_array", 13))
+        {
+          ret = GLEW_APPLE_element_array;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_fence
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fence", 5))
+        {
+          ret = GLEW_APPLE_fence;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_float_pixels
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"float_pixels", 12))
+        {
+          ret = GLEW_APPLE_float_pixels;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_flush_buffer_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"flush_buffer_range", 18))
+        {
+          ret = GLEW_APPLE_flush_buffer_range;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_pixel_buffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_buffer", 12))
+        {
+          ret = GLEW_APPLE_pixel_buffer;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_specular_vector
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"specular_vector", 15))
+        {
+          ret = GLEW_APPLE_specular_vector;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_texture_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_range", 13))
+        {
+          ret = GLEW_APPLE_texture_range;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_transform_hint
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"transform_hint", 14))
+        {
+          ret = GLEW_APPLE_transform_hint;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_vertex_array_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array_object", 19))
+        {
+          ret = GLEW_APPLE_vertex_array_object;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_vertex_array_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array_range", 18))
+        {
+          ret = GLEW_APPLE_vertex_array_range;
+          continue;
+        }
+#endif
+#ifdef GL_APPLE_ycbcr_422
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"ycbcr_422", 9))
+        {
+          ret = GLEW_APPLE_ycbcr_422;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"ARB_", 4))
+      {
+#ifdef GL_ARB_color_buffer_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"color_buffer_float", 18))
+        {
+          ret = GLEW_ARB_color_buffer_float;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_depth_buffer_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"depth_buffer_float", 18))
+        {
+          ret = GLEW_ARB_depth_buffer_float;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_depth_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"depth_texture", 13))
+        {
+          ret = GLEW_ARB_depth_texture;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_draw_buffers
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"draw_buffers", 12))
+        {
+          ret = GLEW_ARB_draw_buffers;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_draw_instanced
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"draw_instanced", 14))
+        {
+          ret = GLEW_ARB_draw_instanced;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_fragment_program
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fragment_program", 16))
+        {
+          ret = GLEW_ARB_fragment_program;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_fragment_program_shadow
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fragment_program_shadow", 23))
+        {
+          ret = GLEW_ARB_fragment_program_shadow;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_fragment_shader
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fragment_shader", 15))
+        {
+          ret = GLEW_ARB_fragment_shader;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_framebuffer_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_object", 18))
+        {
+          ret = GLEW_ARB_framebuffer_object;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_framebuffer_sRGB
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_sRGB", 16))
+        {
+          ret = GLEW_ARB_framebuffer_sRGB;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_geometry_shader4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"geometry_shader4", 16))
+        {
+          ret = GLEW_ARB_geometry_shader4;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_half_float_pixel
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"half_float_pixel", 16))
+        {
+          ret = GLEW_ARB_half_float_pixel;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_half_float_vertex
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"half_float_vertex", 17))
+        {
+          ret = GLEW_ARB_half_float_vertex;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_imaging
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"imaging", 7))
+        {
+          ret = GLEW_ARB_imaging;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_instanced_arrays
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"instanced_arrays", 16))
+        {
+          ret = GLEW_ARB_instanced_arrays;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_map_buffer_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"map_buffer_range", 16))
+        {
+          ret = GLEW_ARB_map_buffer_range;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_matrix_palette
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"matrix_palette", 14))
+        {
+          ret = GLEW_ARB_matrix_palette;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample", 11))
+        {
+          ret = GLEW_ARB_multisample;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_multitexture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multitexture", 12))
+        {
+          ret = GLEW_ARB_multitexture;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_occlusion_query
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"occlusion_query", 15))
+        {
+          ret = GLEW_ARB_occlusion_query;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_pixel_buffer_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_buffer_object", 19))
+        {
+          ret = GLEW_ARB_pixel_buffer_object;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_point_parameters
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"point_parameters", 16))
+        {
+          ret = GLEW_ARB_point_parameters;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_point_sprite
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"point_sprite", 12))
+        {
+          ret = GLEW_ARB_point_sprite;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_shader_objects
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"shader_objects", 14))
+        {
+          ret = GLEW_ARB_shader_objects;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_shading_language_100
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"shading_language_100", 20))
+        {
+          ret = GLEW_ARB_shading_language_100;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_shadow
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"shadow", 6))
+        {
+          ret = GLEW_ARB_shadow;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_shadow_ambient
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"shadow_ambient", 14))
+        {
+          ret = GLEW_ARB_shadow_ambient;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_border_clamp
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_border_clamp", 20))
+        {
+          ret = GLEW_ARB_texture_border_clamp;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_buffer_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_buffer_object", 21))
+        {
+          ret = GLEW_ARB_texture_buffer_object;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_compression
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_compression", 19))
+        {
+          ret = GLEW_ARB_texture_compression;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_compression_rgtc
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_compression_rgtc", 24))
+        {
+          ret = GLEW_ARB_texture_compression_rgtc;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_cube_map
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_cube_map", 16))
+        {
+          ret = GLEW_ARB_texture_cube_map;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_env_add
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_add", 15))
+        {
+          ret = GLEW_ARB_texture_env_add;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_env_combine
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_combine", 19))
+        {
+          ret = GLEW_ARB_texture_env_combine;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_env_crossbar
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_crossbar", 20))
+        {
+          ret = GLEW_ARB_texture_env_crossbar;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_env_dot3
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_dot3", 16))
+        {
+          ret = GLEW_ARB_texture_env_dot3;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_float", 13))
+        {
+          ret = GLEW_ARB_texture_float;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_mirrored_repeat
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_mirrored_repeat", 23))
+        {
+          ret = GLEW_ARB_texture_mirrored_repeat;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_non_power_of_two
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_non_power_of_two", 24))
+        {
+          ret = GLEW_ARB_texture_non_power_of_two;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_rectangle
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_rectangle", 17))
+        {
+          ret = GLEW_ARB_texture_rectangle;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_texture_rg
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_rg", 10))
+        {
+          ret = GLEW_ARB_texture_rg;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_transpose_matrix
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"transpose_matrix", 16))
+        {
+          ret = GLEW_ARB_transpose_matrix;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_vertex_array_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array_object", 19))
+        {
+          ret = GLEW_ARB_vertex_array_object;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_vertex_blend
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_blend", 12))
+        {
+          ret = GLEW_ARB_vertex_blend;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_vertex_buffer_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_buffer_object", 20))
+        {
+          ret = GLEW_ARB_vertex_buffer_object;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_vertex_program
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_program", 14))
+        {
+          ret = GLEW_ARB_vertex_program;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_vertex_shader
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_shader", 13))
+        {
+          ret = GLEW_ARB_vertex_shader;
+          continue;
+        }
+#endif
+#ifdef GL_ARB_window_pos
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"window_pos", 10))
+        {
+          ret = GLEW_ARB_window_pos;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"ATIX_", 5))
+      {
+#ifdef GL_ATIX_point_sprites
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"point_sprites", 13))
+        {
+          ret = GLEW_ATIX_point_sprites;
+          continue;
+        }
+#endif
+#ifdef GL_ATIX_texture_env_combine3
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_combine3", 20))
+        {
+          ret = GLEW_ATIX_texture_env_combine3;
+          continue;
+        }
+#endif
+#ifdef GL_ATIX_texture_env_route
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_route", 17))
+        {
+          ret = GLEW_ATIX_texture_env_route;
+          continue;
+        }
+#endif
+#ifdef GL_ATIX_vertex_shader_output_point_size
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_shader_output_point_size", 31))
+        {
+          ret = GLEW_ATIX_vertex_shader_output_point_size;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"ATI_", 4))
+      {
+#ifdef GL_ATI_draw_buffers
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"draw_buffers", 12))
+        {
+          ret = GLEW_ATI_draw_buffers;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_element_array
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"element_array", 13))
+        {
+          ret = GLEW_ATI_element_array;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_envmap_bumpmap
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"envmap_bumpmap", 14))
+        {
+          ret = GLEW_ATI_envmap_bumpmap;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_fragment_shader
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fragment_shader", 15))
+        {
+          ret = GLEW_ATI_fragment_shader;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_map_object_buffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"map_object_buffer", 17))
+        {
+          ret = GLEW_ATI_map_object_buffer;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_pn_triangles
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pn_triangles", 12))
+        {
+          ret = GLEW_ATI_pn_triangles;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_separate_stencil
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"separate_stencil", 16))
+        {
+          ret = GLEW_ATI_separate_stencil;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_shader_texture_lod
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"shader_texture_lod", 18))
+        {
+          ret = GLEW_ATI_shader_texture_lod;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_text_fragment_shader
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"text_fragment_shader", 20))
+        {
+          ret = GLEW_ATI_text_fragment_shader;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_texture_compression_3dc
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_compression_3dc", 23))
+        {
+          ret = GLEW_ATI_texture_compression_3dc;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_texture_env_combine3
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_combine3", 20))
+        {
+          ret = GLEW_ATI_texture_env_combine3;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_texture_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_float", 13))
+        {
+          ret = GLEW_ATI_texture_float;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_texture_mirror_once
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_mirror_once", 19))
+        {
+          ret = GLEW_ATI_texture_mirror_once;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_vertex_array_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array_object", 19))
+        {
+          ret = GLEW_ATI_vertex_array_object;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_vertex_attrib_array_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_attrib_array_object", 26))
+        {
+          ret = GLEW_ATI_vertex_attrib_array_object;
+          continue;
+        }
+#endif
+#ifdef GL_ATI_vertex_streams
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_streams", 14))
+        {
+          ret = GLEW_ATI_vertex_streams;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"EXT_", 4))
+      {
+#ifdef GL_EXT_422_pixels
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"422_pixels", 10))
+        {
+          ret = GLEW_EXT_422_pixels;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_Cg_shader
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"Cg_shader", 9))
+        {
+          ret = GLEW_EXT_Cg_shader;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_abgr
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"abgr", 4))
+        {
+          ret = GLEW_EXT_abgr;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_bgra
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"bgra", 4))
+        {
+          ret = GLEW_EXT_bgra;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_bindable_uniform
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"bindable_uniform", 16))
+        {
+          ret = GLEW_EXT_bindable_uniform;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_blend_color
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"blend_color", 11))
+        {
+          ret = GLEW_EXT_blend_color;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_blend_equation_separate
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"blend_equation_separate", 23))
+        {
+          ret = GLEW_EXT_blend_equation_separate;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_blend_func_separate
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"blend_func_separate", 19))
+        {
+          ret = GLEW_EXT_blend_func_separate;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_blend_logic_op
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"blend_logic_op", 14))
+        {
+          ret = GLEW_EXT_blend_logic_op;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_blend_minmax
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"blend_minmax", 12))
+        {
+          ret = GLEW_EXT_blend_minmax;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_blend_subtract
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"blend_subtract", 14))
+        {
+          ret = GLEW_EXT_blend_subtract;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_clip_volume_hint
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"clip_volume_hint", 16))
+        {
+          ret = GLEW_EXT_clip_volume_hint;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_cmyka
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"cmyka", 5))
+        {
+          ret = GLEW_EXT_cmyka;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_color_subtable
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"color_subtable", 14))
+        {
+          ret = GLEW_EXT_color_subtable;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_compiled_vertex_array
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"compiled_vertex_array", 21))
+        {
+          ret = GLEW_EXT_compiled_vertex_array;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_convolution
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"convolution", 11))
+        {
+          ret = GLEW_EXT_convolution;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_coordinate_frame
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"coordinate_frame", 16))
+        {
+          ret = GLEW_EXT_coordinate_frame;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_copy_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"copy_texture", 12))
+        {
+          ret = GLEW_EXT_copy_texture;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_cull_vertex
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"cull_vertex", 11))
+        {
+          ret = GLEW_EXT_cull_vertex;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_depth_bounds_test
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"depth_bounds_test", 17))
+        {
+          ret = GLEW_EXT_depth_bounds_test;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_direct_state_access
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"direct_state_access", 19))
+        {
+          ret = GLEW_EXT_direct_state_access;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_draw_buffers2
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"draw_buffers2", 13))
+        {
+          ret = GLEW_EXT_draw_buffers2;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_draw_instanced
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"draw_instanced", 14))
+        {
+          ret = GLEW_EXT_draw_instanced;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_draw_range_elements
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"draw_range_elements", 19))
+        {
+          ret = GLEW_EXT_draw_range_elements;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_fog_coord
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fog_coord", 9))
+        {
+          ret = GLEW_EXT_fog_coord;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_fragment_lighting
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fragment_lighting", 17))
+        {
+          ret = GLEW_EXT_fragment_lighting;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_framebuffer_blit
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_blit", 16))
+        {
+          ret = GLEW_EXT_framebuffer_blit;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_framebuffer_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_multisample", 23))
+        {
+          ret = GLEW_EXT_framebuffer_multisample;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_framebuffer_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_object", 18))
+        {
+          ret = GLEW_EXT_framebuffer_object;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_framebuffer_sRGB
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_sRGB", 16))
+        {
+          ret = GLEW_EXT_framebuffer_sRGB;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_geometry_shader4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"geometry_shader4", 16))
+        {
+          ret = GLEW_EXT_geometry_shader4;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_gpu_program_parameters
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"gpu_program_parameters", 22))
+        {
+          ret = GLEW_EXT_gpu_program_parameters;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_gpu_shader4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"gpu_shader4", 11))
+        {
+          ret = GLEW_EXT_gpu_shader4;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_histogram
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"histogram", 9))
+        {
+          ret = GLEW_EXT_histogram;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_index_array_formats
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"index_array_formats", 19))
+        {
+          ret = GLEW_EXT_index_array_formats;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_index_func
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"index_func", 10))
+        {
+          ret = GLEW_EXT_index_func;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_index_material
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"index_material", 14))
+        {
+          ret = GLEW_EXT_index_material;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_index_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"index_texture", 13))
+        {
+          ret = GLEW_EXT_index_texture;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_light_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"light_texture", 13))
+        {
+          ret = GLEW_EXT_light_texture;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_misc_attribute
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"misc_attribute", 14))
+        {
+          ret = GLEW_EXT_misc_attribute;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_multi_draw_arrays
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multi_draw_arrays", 17))
+        {
+          ret = GLEW_EXT_multi_draw_arrays;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample", 11))
+        {
+          ret = GLEW_EXT_multisample;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_packed_depth_stencil
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"packed_depth_stencil", 20))
+        {
+          ret = GLEW_EXT_packed_depth_stencil;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_packed_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"packed_float", 12))
+        {
+          ret = GLEW_EXT_packed_float;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_packed_pixels
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"packed_pixels", 13))
+        {
+          ret = GLEW_EXT_packed_pixels;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_paletted_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"paletted_texture", 16))
+        {
+          ret = GLEW_EXT_paletted_texture;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_pixel_buffer_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_buffer_object", 19))
+        {
+          ret = GLEW_EXT_pixel_buffer_object;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_pixel_transform
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_transform", 15))
+        {
+          ret = GLEW_EXT_pixel_transform;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_pixel_transform_color_table
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_transform_color_table", 27))
+        {
+          ret = GLEW_EXT_pixel_transform_color_table;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_point_parameters
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"point_parameters", 16))
+        {
+          ret = GLEW_EXT_point_parameters;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_polygon_offset
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"polygon_offset", 14))
+        {
+          ret = GLEW_EXT_polygon_offset;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_rescale_normal
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"rescale_normal", 14))
+        {
+          ret = GLEW_EXT_rescale_normal;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_scene_marker
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"scene_marker", 12))
+        {
+          ret = GLEW_EXT_scene_marker;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_secondary_color
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"secondary_color", 15))
+        {
+          ret = GLEW_EXT_secondary_color;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_separate_specular_color
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"separate_specular_color", 23))
+        {
+          ret = GLEW_EXT_separate_specular_color;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_shadow_funcs
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"shadow_funcs", 12))
+        {
+          ret = GLEW_EXT_shadow_funcs;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_shared_texture_palette
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"shared_texture_palette", 22))
+        {
+          ret = GLEW_EXT_shared_texture_palette;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_stencil_clear_tag
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"stencil_clear_tag", 17))
+        {
+          ret = GLEW_EXT_stencil_clear_tag;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_stencil_two_side
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"stencil_two_side", 16))
+        {
+          ret = GLEW_EXT_stencil_two_side;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_stencil_wrap
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"stencil_wrap", 12))
+        {
+          ret = GLEW_EXT_stencil_wrap;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_subtexture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"subtexture", 10))
+        {
+          ret = GLEW_EXT_subtexture;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture", 7))
+        {
+          ret = GLEW_EXT_texture;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture3D
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture3D", 9))
+        {
+          ret = GLEW_EXT_texture3D;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_array
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_array", 13))
+        {
+          ret = GLEW_EXT_texture_array;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_buffer_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_buffer_object", 21))
+        {
+          ret = GLEW_EXT_texture_buffer_object;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_compression_dxt1
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_compression_dxt1", 24))
+        {
+          ret = GLEW_EXT_texture_compression_dxt1;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_compression_latc
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_compression_latc", 24))
+        {
+          ret = GLEW_EXT_texture_compression_latc;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_compression_rgtc
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_compression_rgtc", 24))
+        {
+          ret = GLEW_EXT_texture_compression_rgtc;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_compression_s3tc
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_compression_s3tc", 24))
+        {
+          ret = GLEW_EXT_texture_compression_s3tc;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_cube_map
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_cube_map", 16))
+        {
+          ret = GLEW_EXT_texture_cube_map;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_edge_clamp
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_edge_clamp", 18))
+        {
+          ret = GLEW_EXT_texture_edge_clamp;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_env
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env", 11))
+        {
+          ret = GLEW_EXT_texture_env;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_env_add
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_add", 15))
+        {
+          ret = GLEW_EXT_texture_env_add;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_env_combine
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_combine", 19))
+        {
+          ret = GLEW_EXT_texture_env_combine;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_env_dot3
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_dot3", 16))
+        {
+          ret = GLEW_EXT_texture_env_dot3;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_filter_anisotropic
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_filter_anisotropic", 26))
+        {
+          ret = GLEW_EXT_texture_filter_anisotropic;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_integer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_integer", 15))
+        {
+          ret = GLEW_EXT_texture_integer;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_lod_bias
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_lod_bias", 16))
+        {
+          ret = GLEW_EXT_texture_lod_bias;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_mirror_clamp
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_mirror_clamp", 20))
+        {
+          ret = GLEW_EXT_texture_mirror_clamp;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_object", 14))
+        {
+          ret = GLEW_EXT_texture_object;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_perturb_normal
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_perturb_normal", 22))
+        {
+          ret = GLEW_EXT_texture_perturb_normal;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_rectangle
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_rectangle", 17))
+        {
+          ret = GLEW_EXT_texture_rectangle;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_sRGB
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_sRGB", 12))
+        {
+          ret = GLEW_EXT_texture_sRGB;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_shared_exponent
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_shared_exponent", 23))
+        {
+          ret = GLEW_EXT_texture_shared_exponent;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_texture_swizzle
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_swizzle", 15))
+        {
+          ret = GLEW_EXT_texture_swizzle;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_timer_query
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"timer_query", 11))
+        {
+          ret = GLEW_EXT_timer_query;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_transform_feedback
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"transform_feedback", 18))
+        {
+          ret = GLEW_EXT_transform_feedback;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_vertex_array
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array", 12))
+        {
+          ret = GLEW_EXT_vertex_array;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_vertex_array_bgra
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array_bgra", 17))
+        {
+          ret = GLEW_EXT_vertex_array_bgra;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_vertex_shader
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_shader", 13))
+        {
+          ret = GLEW_EXT_vertex_shader;
+          continue;
+        }
+#endif
+#ifdef GL_EXT_vertex_weighting
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_weighting", 16))
+        {
+          ret = GLEW_EXT_vertex_weighting;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"GREMEDY_", 8))
+      {
+#ifdef GL_GREMEDY_frame_terminator
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"frame_terminator", 16))
+        {
+          ret = GLEW_GREMEDY_frame_terminator;
+          continue;
+        }
+#endif
+#ifdef GL_GREMEDY_string_marker
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"string_marker", 13))
+        {
+          ret = GLEW_GREMEDY_string_marker;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"HP_", 3))
+      {
+#ifdef GL_HP_convolution_border_modes
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"convolution_border_modes", 24))
+        {
+          ret = GLEW_HP_convolution_border_modes;
+          continue;
+        }
+#endif
+#ifdef GL_HP_image_transform
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"image_transform", 15))
+        {
+          ret = GLEW_HP_image_transform;
+          continue;
+        }
+#endif
+#ifdef GL_HP_occlusion_test
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"occlusion_test", 14))
+        {
+          ret = GLEW_HP_occlusion_test;
+          continue;
+        }
+#endif
+#ifdef GL_HP_texture_lighting
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_lighting", 16))
+        {
+          ret = GLEW_HP_texture_lighting;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"IBM_", 4))
+      {
+#ifdef GL_IBM_cull_vertex
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"cull_vertex", 11))
+        {
+          ret = GLEW_IBM_cull_vertex;
+          continue;
+        }
+#endif
+#ifdef GL_IBM_multimode_draw_arrays
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multimode_draw_arrays", 21))
+        {
+          ret = GLEW_IBM_multimode_draw_arrays;
+          continue;
+        }
+#endif
+#ifdef GL_IBM_rasterpos_clip
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"rasterpos_clip", 14))
+        {
+          ret = GLEW_IBM_rasterpos_clip;
+          continue;
+        }
+#endif
+#ifdef GL_IBM_static_data
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"static_data", 11))
+        {
+          ret = GLEW_IBM_static_data;
+          continue;
+        }
+#endif
+#ifdef GL_IBM_texture_mirrored_repeat
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_mirrored_repeat", 23))
+        {
+          ret = GLEW_IBM_texture_mirrored_repeat;
+          continue;
+        }
+#endif
+#ifdef GL_IBM_vertex_array_lists
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array_lists", 18))
+        {
+          ret = GLEW_IBM_vertex_array_lists;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"INGR_", 5))
+      {
+#ifdef GL_INGR_color_clamp
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"color_clamp", 11))
+        {
+          ret = GLEW_INGR_color_clamp;
+          continue;
+        }
+#endif
+#ifdef GL_INGR_interlace_read
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"interlace_read", 14))
+        {
+          ret = GLEW_INGR_interlace_read;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"INTEL_", 6))
+      {
+#ifdef GL_INTEL_parallel_arrays
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"parallel_arrays", 15))
+        {
+          ret = GLEW_INTEL_parallel_arrays;
+          continue;
+        }
+#endif
+#ifdef GL_INTEL_texture_scissor
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_scissor", 15))
+        {
+          ret = GLEW_INTEL_texture_scissor;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"KTX_", 4))
+      {
+#ifdef GL_KTX_buffer_region
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"buffer_region", 13))
+        {
+          ret = GLEW_KTX_buffer_region;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"MESAX_", 6))
+      {
+#ifdef GL_MESAX_texture_stack
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_stack", 13))
+        {
+          ret = GLEW_MESAX_texture_stack;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"MESA_", 5))
+      {
+#ifdef GL_MESA_pack_invert
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pack_invert", 11))
+        {
+          ret = GLEW_MESA_pack_invert;
+          continue;
+        }
+#endif
+#ifdef GL_MESA_resize_buffers
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"resize_buffers", 14))
+        {
+          ret = GLEW_MESA_resize_buffers;
+          continue;
+        }
+#endif
+#ifdef GL_MESA_window_pos
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"window_pos", 10))
+        {
+          ret = GLEW_MESA_window_pos;
+          continue;
+        }
+#endif
+#ifdef GL_MESA_ycbcr_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"ycbcr_texture", 13))
+        {
+          ret = GLEW_MESA_ycbcr_texture;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"NV_", 3))
+      {
+#ifdef GL_NV_blend_square
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"blend_square", 12))
+        {
+          ret = GLEW_NV_blend_square;
+          continue;
+        }
+#endif
+#ifdef GL_NV_conditional_render
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"conditional_render", 18))
+        {
+          ret = GLEW_NV_conditional_render;
+          continue;
+        }
+#endif
+#ifdef GL_NV_copy_depth_to_color
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"copy_depth_to_color", 19))
+        {
+          ret = GLEW_NV_copy_depth_to_color;
+          continue;
+        }
+#endif
+#ifdef GL_NV_depth_buffer_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"depth_buffer_float", 18))
+        {
+          ret = GLEW_NV_depth_buffer_float;
+          continue;
+        }
+#endif
+#ifdef GL_NV_depth_clamp
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"depth_clamp", 11))
+        {
+          ret = GLEW_NV_depth_clamp;
+          continue;
+        }
+#endif
+#ifdef GL_NV_depth_range_unclamped
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"depth_range_unclamped", 21))
+        {
+          ret = GLEW_NV_depth_range_unclamped;
+          continue;
+        }
+#endif
+#ifdef GL_NV_evaluators
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"evaluators", 10))
+        {
+          ret = GLEW_NV_evaluators;
+          continue;
+        }
+#endif
+#ifdef GL_NV_explicit_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"explicit_multisample", 20))
+        {
+          ret = GLEW_NV_explicit_multisample;
+          continue;
+        }
+#endif
+#ifdef GL_NV_fence
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fence", 5))
+        {
+          ret = GLEW_NV_fence;
+          continue;
+        }
+#endif
+#ifdef GL_NV_float_buffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"float_buffer", 12))
+        {
+          ret = GLEW_NV_float_buffer;
+          continue;
+        }
+#endif
+#ifdef GL_NV_fog_distance
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fog_distance", 12))
+        {
+          ret = GLEW_NV_fog_distance;
+          continue;
+        }
+#endif
+#ifdef GL_NV_fragment_program
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fragment_program", 16))
+        {
+          ret = GLEW_NV_fragment_program;
+          continue;
+        }
+#endif
+#ifdef GL_NV_fragment_program2
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fragment_program2", 17))
+        {
+          ret = GLEW_NV_fragment_program2;
+          continue;
+        }
+#endif
+#ifdef GL_NV_fragment_program4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fragment_program4", 17))
+        {
+          ret = GLEW_NV_fragment_program4;
+          continue;
+        }
+#endif
+#ifdef GL_NV_fragment_program_option
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fragment_program_option", 23))
+        {
+          ret = GLEW_NV_fragment_program_option;
+          continue;
+        }
+#endif
+#ifdef GL_NV_framebuffer_multisample_coverage
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_multisample_coverage", 32))
+        {
+          ret = GLEW_NV_framebuffer_multisample_coverage;
+          continue;
+        }
+#endif
+#ifdef GL_NV_geometry_program4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"geometry_program4", 17))
+        {
+          ret = GLEW_NV_geometry_program4;
+          continue;
+        }
+#endif
+#ifdef GL_NV_geometry_shader4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"geometry_shader4", 16))
+        {
+          ret = GLEW_NV_geometry_shader4;
+          continue;
+        }
+#endif
+#ifdef GL_NV_gpu_program4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"gpu_program4", 12))
+        {
+          ret = GLEW_NV_gpu_program4;
+          continue;
+        }
+#endif
+#ifdef GL_NV_half_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"half_float", 10))
+        {
+          ret = GLEW_NV_half_float;
+          continue;
+        }
+#endif
+#ifdef GL_NV_light_max_exponent
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"light_max_exponent", 18))
+        {
+          ret = GLEW_NV_light_max_exponent;
+          continue;
+        }
+#endif
+#ifdef GL_NV_multisample_filter_hint
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample_filter_hint", 23))
+        {
+          ret = GLEW_NV_multisample_filter_hint;
+          continue;
+        }
+#endif
+#ifdef GL_NV_occlusion_query
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"occlusion_query", 15))
+        {
+          ret = GLEW_NV_occlusion_query;
+          continue;
+        }
+#endif
+#ifdef GL_NV_packed_depth_stencil
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"packed_depth_stencil", 20))
+        {
+          ret = GLEW_NV_packed_depth_stencil;
+          continue;
+        }
+#endif
+#ifdef GL_NV_parameter_buffer_object
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"parameter_buffer_object", 23))
+        {
+          ret = GLEW_NV_parameter_buffer_object;
+          continue;
+        }
+#endif
+#ifdef GL_NV_pixel_data_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_data_range", 16))
+        {
+          ret = GLEW_NV_pixel_data_range;
+          continue;
+        }
+#endif
+#ifdef GL_NV_point_sprite
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"point_sprite", 12))
+        {
+          ret = GLEW_NV_point_sprite;
+          continue;
+        }
+#endif
+#ifdef GL_NV_present_video
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"present_video", 13))
+        {
+          ret = GLEW_NV_present_video;
+          continue;
+        }
+#endif
+#ifdef GL_NV_primitive_restart
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"primitive_restart", 17))
+        {
+          ret = GLEW_NV_primitive_restart;
+          continue;
+        }
+#endif
+#ifdef GL_NV_register_combiners
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"register_combiners", 18))
+        {
+          ret = GLEW_NV_register_combiners;
+          continue;
+        }
+#endif
+#ifdef GL_NV_register_combiners2
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"register_combiners2", 19))
+        {
+          ret = GLEW_NV_register_combiners2;
+          continue;
+        }
+#endif
+#ifdef GL_NV_texgen_emboss
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texgen_emboss", 13))
+        {
+          ret = GLEW_NV_texgen_emboss;
+          continue;
+        }
+#endif
+#ifdef GL_NV_texgen_reflection
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texgen_reflection", 17))
+        {
+          ret = GLEW_NV_texgen_reflection;
+          continue;
+        }
+#endif
+#ifdef GL_NV_texture_compression_vtc
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_compression_vtc", 23))
+        {
+          ret = GLEW_NV_texture_compression_vtc;
+          continue;
+        }
+#endif
+#ifdef GL_NV_texture_env_combine4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_env_combine4", 20))
+        {
+          ret = GLEW_NV_texture_env_combine4;
+          continue;
+        }
+#endif
+#ifdef GL_NV_texture_expand_normal
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_expand_normal", 21))
+        {
+          ret = GLEW_NV_texture_expand_normal;
+          continue;
+        }
+#endif
+#ifdef GL_NV_texture_rectangle
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_rectangle", 17))
+        {
+          ret = GLEW_NV_texture_rectangle;
+          continue;
+        }
+#endif
+#ifdef GL_NV_texture_shader
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_shader", 14))
+        {
+          ret = GLEW_NV_texture_shader;
+          continue;
+        }
+#endif
+#ifdef GL_NV_texture_shader2
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_shader2", 15))
+        {
+          ret = GLEW_NV_texture_shader2;
+          continue;
+        }
+#endif
+#ifdef GL_NV_texture_shader3
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_shader3", 15))
+        {
+          ret = GLEW_NV_texture_shader3;
+          continue;
+        }
+#endif
+#ifdef GL_NV_transform_feedback
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"transform_feedback", 18))
+        {
+          ret = GLEW_NV_transform_feedback;
+          continue;
+        }
+#endif
+#ifdef GL_NV_vertex_array_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array_range", 18))
+        {
+          ret = GLEW_NV_vertex_array_range;
+          continue;
+        }
+#endif
+#ifdef GL_NV_vertex_array_range2
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array_range2", 19))
+        {
+          ret = GLEW_NV_vertex_array_range2;
+          continue;
+        }
+#endif
+#ifdef GL_NV_vertex_program
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_program", 14))
+        {
+          ret = GLEW_NV_vertex_program;
+          continue;
+        }
+#endif
+#ifdef GL_NV_vertex_program1_1
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_program1_1", 17))
+        {
+          ret = GLEW_NV_vertex_program1_1;
+          continue;
+        }
+#endif
+#ifdef GL_NV_vertex_program2
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_program2", 15))
+        {
+          ret = GLEW_NV_vertex_program2;
+          continue;
+        }
+#endif
+#ifdef GL_NV_vertex_program2_option
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_program2_option", 22))
+        {
+          ret = GLEW_NV_vertex_program2_option;
+          continue;
+        }
+#endif
+#ifdef GL_NV_vertex_program3
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_program3", 15))
+        {
+          ret = GLEW_NV_vertex_program3;
+          continue;
+        }
+#endif
+#ifdef GL_NV_vertex_program4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_program4", 15))
+        {
+          ret = GLEW_NV_vertex_program4;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"OES_", 4))
+      {
+#ifdef GL_OES_byte_coordinates
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"byte_coordinates", 16))
+        {
+          ret = GLEW_OES_byte_coordinates;
+          continue;
+        }
+#endif
+#ifdef GL_OES_compressed_paletted_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"compressed_paletted_texture", 27))
+        {
+          ret = GLEW_OES_compressed_paletted_texture;
+          continue;
+        }
+#endif
+#ifdef GL_OES_read_format
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"read_format", 11))
+        {
+          ret = GLEW_OES_read_format;
+          continue;
+        }
+#endif
+#ifdef GL_OES_single_precision
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"single_precision", 16))
+        {
+          ret = GLEW_OES_single_precision;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"OML_", 4))
+      {
+#ifdef GL_OML_interlace
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"interlace", 9))
+        {
+          ret = GLEW_OML_interlace;
+          continue;
+        }
+#endif
+#ifdef GL_OML_resample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"resample", 8))
+        {
+          ret = GLEW_OML_resample;
+          continue;
+        }
+#endif
+#ifdef GL_OML_subsample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"subsample", 9))
+        {
+          ret = GLEW_OML_subsample;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"PGI_", 4))
+      {
+#ifdef GL_PGI_misc_hints
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"misc_hints", 10))
+        {
+          ret = GLEW_PGI_misc_hints;
+          continue;
+        }
+#endif
+#ifdef GL_PGI_vertex_hints
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_hints", 12))
+        {
+          ret = GLEW_PGI_vertex_hints;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"REND_", 5))
+      {
+#ifdef GL_REND_screen_coordinates
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"screen_coordinates", 18))
+        {
+          ret = GLEW_REND_screen_coordinates;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"S3_", 3))
+      {
+#ifdef GL_S3_s3tc
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"s3tc", 4))
+        {
+          ret = GLEW_S3_s3tc;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"SGIS_", 5))
+      {
+#ifdef GL_SGIS_color_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"color_range", 11))
+        {
+          ret = GLEW_SGIS_color_range;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_detail_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"detail_texture", 14))
+        {
+          ret = GLEW_SGIS_detail_texture;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_fog_function
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fog_function", 12))
+        {
+          ret = GLEW_SGIS_fog_function;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_generate_mipmap
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"generate_mipmap", 15))
+        {
+          ret = GLEW_SGIS_generate_mipmap;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample", 11))
+        {
+          ret = GLEW_SGIS_multisample;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_pixel_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_texture", 13))
+        {
+          ret = GLEW_SGIS_pixel_texture;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_point_line_texgen
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"point_line_texgen", 17))
+        {
+          ret = GLEW_SGIS_point_line_texgen;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_sharpen_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"sharpen_texture", 15))
+        {
+          ret = GLEW_SGIS_sharpen_texture;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_texture4D
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture4D", 9))
+        {
+          ret = GLEW_SGIS_texture4D;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_texture_border_clamp
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_border_clamp", 20))
+        {
+          ret = GLEW_SGIS_texture_border_clamp;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_texture_edge_clamp
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_edge_clamp", 18))
+        {
+          ret = GLEW_SGIS_texture_edge_clamp;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_texture_filter4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_filter4", 15))
+        {
+          ret = GLEW_SGIS_texture_filter4;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_texture_lod
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_lod", 11))
+        {
+          ret = GLEW_SGIS_texture_lod;
+          continue;
+        }
+#endif
+#ifdef GL_SGIS_texture_select
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_select", 14))
+        {
+          ret = GLEW_SGIS_texture_select;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"SGIX_", 5))
+      {
+#ifdef GL_SGIX_async
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"async", 5))
+        {
+          ret = GLEW_SGIX_async;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_async_histogram
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"async_histogram", 15))
+        {
+          ret = GLEW_SGIX_async_histogram;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_async_pixel
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"async_pixel", 11))
+        {
+          ret = GLEW_SGIX_async_pixel;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_blend_alpha_minmax
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"blend_alpha_minmax", 18))
+        {
+          ret = GLEW_SGIX_blend_alpha_minmax;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_clipmap
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"clipmap", 7))
+        {
+          ret = GLEW_SGIX_clipmap;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_convolution_accuracy
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"convolution_accuracy", 20))
+        {
+          ret = GLEW_SGIX_convolution_accuracy;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_depth_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"depth_texture", 13))
+        {
+          ret = GLEW_SGIX_depth_texture;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_flush_raster
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"flush_raster", 12))
+        {
+          ret = GLEW_SGIX_flush_raster;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_fog_offset
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fog_offset", 10))
+        {
+          ret = GLEW_SGIX_fog_offset;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_fog_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fog_texture", 11))
+        {
+          ret = GLEW_SGIX_fog_texture;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_fragment_specular_lighting
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fragment_specular_lighting", 26))
+        {
+          ret = GLEW_SGIX_fragment_specular_lighting;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_framezoom
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framezoom", 9))
+        {
+          ret = GLEW_SGIX_framezoom;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_interlace
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"interlace", 9))
+        {
+          ret = GLEW_SGIX_interlace;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_ir_instrument1
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"ir_instrument1", 14))
+        {
+          ret = GLEW_SGIX_ir_instrument1;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_list_priority
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"list_priority", 13))
+        {
+          ret = GLEW_SGIX_list_priority;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_pixel_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_texture", 13))
+        {
+          ret = GLEW_SGIX_pixel_texture;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_pixel_texture_bits
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_texture_bits", 18))
+        {
+          ret = GLEW_SGIX_pixel_texture_bits;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_reference_plane
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"reference_plane", 15))
+        {
+          ret = GLEW_SGIX_reference_plane;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_resample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"resample", 8))
+        {
+          ret = GLEW_SGIX_resample;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_shadow
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"shadow", 6))
+        {
+          ret = GLEW_SGIX_shadow;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_shadow_ambient
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"shadow_ambient", 14))
+        {
+          ret = GLEW_SGIX_shadow_ambient;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_sprite
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"sprite", 6))
+        {
+          ret = GLEW_SGIX_sprite;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_tag_sample_buffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"tag_sample_buffer", 17))
+        {
+          ret = GLEW_SGIX_tag_sample_buffer;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_texture_add_env
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_add_env", 15))
+        {
+          ret = GLEW_SGIX_texture_add_env;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_texture_coordinate_clamp
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_coordinate_clamp", 24))
+        {
+          ret = GLEW_SGIX_texture_coordinate_clamp;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_texture_lod_bias
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_lod_bias", 16))
+        {
+          ret = GLEW_SGIX_texture_lod_bias;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_texture_multi_buffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_multi_buffer", 20))
+        {
+          ret = GLEW_SGIX_texture_multi_buffer;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_texture_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_range", 13))
+        {
+          ret = GLEW_SGIX_texture_range;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_texture_scale_bias
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_scale_bias", 18))
+        {
+          ret = GLEW_SGIX_texture_scale_bias;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_vertex_preclip
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_preclip", 14))
+        {
+          ret = GLEW_SGIX_vertex_preclip;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_vertex_preclip_hint
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_preclip_hint", 19))
+        {
+          ret = GLEW_SGIX_vertex_preclip_hint;
+          continue;
+        }
+#endif
+#ifdef GL_SGIX_ycrcb
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"ycrcb", 5))
+        {
+          ret = GLEW_SGIX_ycrcb;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"SGI_", 4))
+      {
+#ifdef GL_SGI_color_matrix
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"color_matrix", 12))
+        {
+          ret = GLEW_SGI_color_matrix;
+          continue;
+        }
+#endif
+#ifdef GL_SGI_color_table
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"color_table", 11))
+        {
+          ret = GLEW_SGI_color_table;
+          continue;
+        }
+#endif
+#ifdef GL_SGI_texture_color_table
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_color_table", 19))
+        {
+          ret = GLEW_SGI_texture_color_table;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"SUNX_", 5))
+      {
+#ifdef GL_SUNX_constant_data
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"constant_data", 13))
+        {
+          ret = GLEW_SUNX_constant_data;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"SUN_", 4))
+      {
+#ifdef GL_SUN_convolution_border_modes
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"convolution_border_modes", 24))
+        {
+          ret = GLEW_SUN_convolution_border_modes;
+          continue;
+        }
+#endif
+#ifdef GL_SUN_global_alpha
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"global_alpha", 12))
+        {
+          ret = GLEW_SUN_global_alpha;
+          continue;
+        }
+#endif
+#ifdef GL_SUN_mesh_array
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"mesh_array", 10))
+        {
+          ret = GLEW_SUN_mesh_array;
+          continue;
+        }
+#endif
+#ifdef GL_SUN_read_video_pixels
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"read_video_pixels", 17))
+        {
+          ret = GLEW_SUN_read_video_pixels;
+          continue;
+        }
+#endif
+#ifdef GL_SUN_slice_accum
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"slice_accum", 11))
+        {
+          ret = GLEW_SUN_slice_accum;
+          continue;
+        }
+#endif
+#ifdef GL_SUN_triangle_list
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"triangle_list", 13))
+        {
+          ret = GLEW_SUN_triangle_list;
+          continue;
+        }
+#endif
+#ifdef GL_SUN_vertex
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex", 6))
+        {
+          ret = GLEW_SUN_vertex;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"WIN_", 4))
+      {
+#ifdef GL_WIN_phong_shading
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"phong_shading", 13))
+        {
+          ret = GLEW_WIN_phong_shading;
+          continue;
+        }
+#endif
+#ifdef GL_WIN_specular_fog
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"specular_fog", 12))
+        {
+          ret = GLEW_WIN_specular_fog;
+          continue;
+        }
+#endif
+#ifdef GL_WIN_swap_hint
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"swap_hint", 9))
+        {
+          ret = GLEW_WIN_swap_hint;
+          continue;
+        }
+#endif
+      }
+    }
+    ret = (len == 0);
+  }
+  return ret;
+}
+
+#if defined(_WIN32)
+
+#if defined(GLEW_MX)
+GLboolean wglewContextIsSupported (WGLEWContext* ctx, const char* name)
+#else
+GLboolean wglewIsSupported (const char* name)
+#endif
+{
+  GLubyte* pos = (GLubyte*)name;
+  GLuint len = _glewStrLen(pos);
+  GLboolean ret = GL_TRUE;
+  while (ret && len > 0)
+  {
+    if (_glewStrSame1(&pos, &len, (const GLubyte*)"WGL_", 4))
+    {
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"3DFX_", 5))
+      {
+#ifdef WGL_3DFX_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample", 11))
+        {
+          ret = WGLEW_3DFX_multisample;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"3DL_", 4))
+      {
+#ifdef WGL_3DL_stereo_control
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"stereo_control", 14))
+        {
+          ret = WGLEW_3DL_stereo_control;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"ARB_", 4))
+      {
+#ifdef WGL_ARB_buffer_region
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"buffer_region", 13))
+        {
+          ret = WGLEW_ARB_buffer_region;
+          continue;
+        }
+#endif
+#ifdef WGL_ARB_create_context
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"create_context", 14))
+        {
+          ret = WGLEW_ARB_create_context;
+          continue;
+        }
+#endif
+#ifdef WGL_ARB_extensions_string
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"extensions_string", 17))
+        {
+          ret = WGLEW_ARB_extensions_string;
+          continue;
+        }
+#endif
+#ifdef WGL_ARB_framebuffer_sRGB
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_sRGB", 16))
+        {
+          ret = WGLEW_ARB_framebuffer_sRGB;
+          continue;
+        }
+#endif
+#ifdef WGL_ARB_make_current_read
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"make_current_read", 17))
+        {
+          ret = WGLEW_ARB_make_current_read;
+          continue;
+        }
+#endif
+#ifdef WGL_ARB_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample", 11))
+        {
+          ret = WGLEW_ARB_multisample;
+          continue;
+        }
+#endif
+#ifdef WGL_ARB_pbuffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pbuffer", 7))
+        {
+          ret = WGLEW_ARB_pbuffer;
+          continue;
+        }
+#endif
+#ifdef WGL_ARB_pixel_format
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_format", 12))
+        {
+          ret = WGLEW_ARB_pixel_format;
+          continue;
+        }
+#endif
+#ifdef WGL_ARB_pixel_format_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_format_float", 18))
+        {
+          ret = WGLEW_ARB_pixel_format_float;
+          continue;
+        }
+#endif
+#ifdef WGL_ARB_render_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"render_texture", 14))
+        {
+          ret = WGLEW_ARB_render_texture;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"ATI_", 4))
+      {
+#ifdef WGL_ATI_pixel_format_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_format_float", 18))
+        {
+          ret = WGLEW_ATI_pixel_format_float;
+          continue;
+        }
+#endif
+#ifdef WGL_ATI_render_texture_rectangle
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"render_texture_rectangle", 24))
+        {
+          ret = WGLEW_ATI_render_texture_rectangle;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"EXT_", 4))
+      {
+#ifdef WGL_EXT_depth_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"depth_float", 11))
+        {
+          ret = WGLEW_EXT_depth_float;
+          continue;
+        }
+#endif
+#ifdef WGL_EXT_display_color_table
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"display_color_table", 19))
+        {
+          ret = WGLEW_EXT_display_color_table;
+          continue;
+        }
+#endif
+#ifdef WGL_EXT_extensions_string
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"extensions_string", 17))
+        {
+          ret = WGLEW_EXT_extensions_string;
+          continue;
+        }
+#endif
+#ifdef WGL_EXT_framebuffer_sRGB
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_sRGB", 16))
+        {
+          ret = WGLEW_EXT_framebuffer_sRGB;
+          continue;
+        }
+#endif
+#ifdef WGL_EXT_make_current_read
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"make_current_read", 17))
+        {
+          ret = WGLEW_EXT_make_current_read;
+          continue;
+        }
+#endif
+#ifdef WGL_EXT_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample", 11))
+        {
+          ret = WGLEW_EXT_multisample;
+          continue;
+        }
+#endif
+#ifdef WGL_EXT_pbuffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pbuffer", 7))
+        {
+          ret = WGLEW_EXT_pbuffer;
+          continue;
+        }
+#endif
+#ifdef WGL_EXT_pixel_format
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_format", 12))
+        {
+          ret = WGLEW_EXT_pixel_format;
+          continue;
+        }
+#endif
+#ifdef WGL_EXT_pixel_format_packed_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_format_packed_float", 25))
+        {
+          ret = WGLEW_EXT_pixel_format_packed_float;
+          continue;
+        }
+#endif
+#ifdef WGL_EXT_swap_control
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"swap_control", 12))
+        {
+          ret = WGLEW_EXT_swap_control;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"I3D_", 4))
+      {
+#ifdef WGL_I3D_digital_video_control
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"digital_video_control", 21))
+        {
+          ret = WGLEW_I3D_digital_video_control;
+          continue;
+        }
+#endif
+#ifdef WGL_I3D_gamma
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"gamma", 5))
+        {
+          ret = WGLEW_I3D_gamma;
+          continue;
+        }
+#endif
+#ifdef WGL_I3D_genlock
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"genlock", 7))
+        {
+          ret = WGLEW_I3D_genlock;
+          continue;
+        }
+#endif
+#ifdef WGL_I3D_image_buffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"image_buffer", 12))
+        {
+          ret = WGLEW_I3D_image_buffer;
+          continue;
+        }
+#endif
+#ifdef WGL_I3D_swap_frame_lock
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"swap_frame_lock", 15))
+        {
+          ret = WGLEW_I3D_swap_frame_lock;
+          continue;
+        }
+#endif
+#ifdef WGL_I3D_swap_frame_usage
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"swap_frame_usage", 16))
+        {
+          ret = WGLEW_I3D_swap_frame_usage;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"NV_", 3))
+      {
+#ifdef WGL_NV_float_buffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"float_buffer", 12))
+        {
+          ret = WGLEW_NV_float_buffer;
+          continue;
+        }
+#endif
+#ifdef WGL_NV_gpu_affinity
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"gpu_affinity", 12))
+        {
+          ret = WGLEW_NV_gpu_affinity;
+          continue;
+        }
+#endif
+#ifdef WGL_NV_present_video
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"present_video", 13))
+        {
+          ret = WGLEW_NV_present_video;
+          continue;
+        }
+#endif
+#ifdef WGL_NV_render_depth_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"render_depth_texture", 20))
+        {
+          ret = WGLEW_NV_render_depth_texture;
+          continue;
+        }
+#endif
+#ifdef WGL_NV_render_texture_rectangle
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"render_texture_rectangle", 24))
+        {
+          ret = WGLEW_NV_render_texture_rectangle;
+          continue;
+        }
+#endif
+#ifdef WGL_NV_swap_group
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"swap_group", 10))
+        {
+          ret = WGLEW_NV_swap_group;
+          continue;
+        }
+#endif
+#ifdef WGL_NV_vertex_array_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array_range", 18))
+        {
+          ret = WGLEW_NV_vertex_array_range;
+          continue;
+        }
+#endif
+#ifdef WGL_NV_video_output
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"video_output", 12))
+        {
+          ret = WGLEW_NV_video_output;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"OML_", 4))
+      {
+#ifdef WGL_OML_sync_control
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"sync_control", 12))
+        {
+          ret = WGLEW_OML_sync_control;
+          continue;
+        }
+#endif
+      }
+    }
+    ret = (len == 0);
+  }
+  return ret;
+}
+
+#elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+
+#if defined(GLEW_MX)
+GLboolean glxewContextIsSupported (GLXEWContext* ctx, const char* name)
+#else
+GLboolean glxewIsSupported (const char* name)
+#endif
+{
+  GLubyte* pos = (GLubyte*)name;
+  GLuint len = _glewStrLen(pos);
+  GLboolean ret = GL_TRUE;
+  while (ret && len > 0)
+  {
+    if(_glewStrSame1(&pos, &len, (const GLubyte*)"GLX_", 4))
+    {
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"VERSION_", 8))
+      {
+#ifdef GLX_VERSION_1_2
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"1_2", 3))
+        {
+          ret = GLXEW_VERSION_1_2;
+          continue;
+        }
+#endif
+#ifdef GLX_VERSION_1_3
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"1_3", 3))
+        {
+          ret = GLXEW_VERSION_1_3;
+          continue;
+        }
+#endif
+#ifdef GLX_VERSION_1_4
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"1_4", 3))
+        {
+          ret = GLXEW_VERSION_1_4;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"3DFX_", 5))
+      {
+#ifdef GLX_3DFX_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample", 11))
+        {
+          ret = GLXEW_3DFX_multisample;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"ARB_", 4))
+      {
+#ifdef GLX_ARB_create_context
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"create_context", 14))
+        {
+          ret = GLXEW_ARB_create_context;
+          continue;
+        }
+#endif
+#ifdef GLX_ARB_fbconfig_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fbconfig_float", 14))
+        {
+          ret = GLXEW_ARB_fbconfig_float;
+          continue;
+        }
+#endif
+#ifdef GLX_ARB_framebuffer_sRGB
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_sRGB", 16))
+        {
+          ret = GLXEW_ARB_framebuffer_sRGB;
+          continue;
+        }
+#endif
+#ifdef GLX_ARB_get_proc_address
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"get_proc_address", 16))
+        {
+          ret = GLXEW_ARB_get_proc_address;
+          continue;
+        }
+#endif
+#ifdef GLX_ARB_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample", 11))
+        {
+          ret = GLXEW_ARB_multisample;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"ATI_", 4))
+      {
+#ifdef GLX_ATI_pixel_format_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixel_format_float", 18))
+        {
+          ret = GLXEW_ATI_pixel_format_float;
+          continue;
+        }
+#endif
+#ifdef GLX_ATI_render_texture
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"render_texture", 14))
+        {
+          ret = GLXEW_ATI_render_texture;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"EXT_", 4))
+      {
+#ifdef GLX_EXT_fbconfig_packed_float
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fbconfig_packed_float", 21))
+        {
+          ret = GLXEW_EXT_fbconfig_packed_float;
+          continue;
+        }
+#endif
+#ifdef GLX_EXT_framebuffer_sRGB
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"framebuffer_sRGB", 16))
+        {
+          ret = GLXEW_EXT_framebuffer_sRGB;
+          continue;
+        }
+#endif
+#ifdef GLX_EXT_import_context
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"import_context", 14))
+        {
+          ret = GLXEW_EXT_import_context;
+          continue;
+        }
+#endif
+#ifdef GLX_EXT_scene_marker
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"scene_marker", 12))
+        {
+          ret = GLXEW_EXT_scene_marker;
+          continue;
+        }
+#endif
+#ifdef GLX_EXT_texture_from_pixmap
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"texture_from_pixmap", 19))
+        {
+          ret = GLXEW_EXT_texture_from_pixmap;
+          continue;
+        }
+#endif
+#ifdef GLX_EXT_visual_info
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"visual_info", 11))
+        {
+          ret = GLXEW_EXT_visual_info;
+          continue;
+        }
+#endif
+#ifdef GLX_EXT_visual_rating
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"visual_rating", 13))
+        {
+          ret = GLXEW_EXT_visual_rating;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"MESA_", 5))
+      {
+#ifdef GLX_MESA_agp_offset
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"agp_offset", 10))
+        {
+          ret = GLXEW_MESA_agp_offset;
+          continue;
+        }
+#endif
+#ifdef GLX_MESA_copy_sub_buffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"copy_sub_buffer", 15))
+        {
+          ret = GLXEW_MESA_copy_sub_buffer;
+          continue;
+        }
+#endif
+#ifdef GLX_MESA_pixmap_colormap
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pixmap_colormap", 15))
+        {
+          ret = GLXEW_MESA_pixmap_colormap;
+          continue;
+        }
+#endif
+#ifdef GLX_MESA_release_buffers
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"release_buffers", 15))
+        {
+          ret = GLXEW_MESA_release_buffers;
+          continue;
+        }
+#endif
+#ifdef GLX_MESA_set_3dfx_mode
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"set_3dfx_mode", 13))
+        {
+          ret = GLXEW_MESA_set_3dfx_mode;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"NV_", 3))
+      {
+#ifdef GLX_NV_float_buffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"float_buffer", 12))
+        {
+          ret = GLXEW_NV_float_buffer;
+          continue;
+        }
+#endif
+#ifdef GLX_NV_present_video
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"present_video", 13))
+        {
+          ret = GLXEW_NV_present_video;
+          continue;
+        }
+#endif
+#ifdef GLX_NV_swap_group
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"swap_group", 10))
+        {
+          ret = GLXEW_NV_swap_group;
+          continue;
+        }
+#endif
+#ifdef GLX_NV_vertex_array_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"vertex_array_range", 18))
+        {
+          ret = GLXEW_NV_vertex_array_range;
+          continue;
+        }
+#endif
+#ifdef GLX_NV_video_output
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"video_output", 12))
+        {
+          ret = GLXEW_NV_video_output;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"OML_", 4))
+      {
+#ifdef GLX_OML_swap_method
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"swap_method", 11))
+        {
+          ret = GLXEW_OML_swap_method;
+          continue;
+        }
+#endif
+#if defined(GLX_OML_sync_control) && defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)
+#include <inttypes.h>
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"sync_control", 12))
+        {
+          ret = GLXEW_OML_sync_control;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"SGIS_", 5))
+      {
+#ifdef GLX_SGIS_blended_overlay
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"blended_overlay", 15))
+        {
+          ret = GLXEW_SGIS_blended_overlay;
+          continue;
+        }
+#endif
+#ifdef GLX_SGIS_color_range
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"color_range", 11))
+        {
+          ret = GLXEW_SGIS_color_range;
+          continue;
+        }
+#endif
+#ifdef GLX_SGIS_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"multisample", 11))
+        {
+          ret = GLXEW_SGIS_multisample;
+          continue;
+        }
+#endif
+#ifdef GLX_SGIS_shared_multisample
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"shared_multisample", 18))
+        {
+          ret = GLXEW_SGIS_shared_multisample;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"SGIX_", 5))
+      {
+#ifdef GLX_SGIX_fbconfig
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"fbconfig", 8))
+        {
+          ret = GLXEW_SGIX_fbconfig;
+          continue;
+        }
+#endif
+#ifdef GLX_SGIX_hyperpipe
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"hyperpipe", 9))
+        {
+          ret = GLXEW_SGIX_hyperpipe;
+          continue;
+        }
+#endif
+#ifdef GLX_SGIX_pbuffer
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"pbuffer", 7))
+        {
+          ret = GLXEW_SGIX_pbuffer;
+          continue;
+        }
+#endif
+#ifdef GLX_SGIX_swap_barrier
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"swap_barrier", 12))
+        {
+          ret = GLXEW_SGIX_swap_barrier;
+          continue;
+        }
+#endif
+#ifdef GLX_SGIX_swap_group
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"swap_group", 10))
+        {
+          ret = GLXEW_SGIX_swap_group;
+          continue;
+        }
+#endif
+#ifdef GLX_SGIX_video_resize
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"video_resize", 12))
+        {
+          ret = GLXEW_SGIX_video_resize;
+          continue;
+        }
+#endif
+#ifdef GLX_SGIX_visual_select_group
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"visual_select_group", 19))
+        {
+          ret = GLXEW_SGIX_visual_select_group;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"SGI_", 4))
+      {
+#ifdef GLX_SGI_cushion
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"cushion", 7))
+        {
+          ret = GLXEW_SGI_cushion;
+          continue;
+        }
+#endif
+#ifdef GLX_SGI_make_current_read
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"make_current_read", 17))
+        {
+          ret = GLXEW_SGI_make_current_read;
+          continue;
+        }
+#endif
+#ifdef GLX_SGI_swap_control
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"swap_control", 12))
+        {
+          ret = GLXEW_SGI_swap_control;
+          continue;
+        }
+#endif
+#ifdef GLX_SGI_video_sync
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"video_sync", 10))
+        {
+          ret = GLXEW_SGI_video_sync;
+          continue;
+        }
+#endif
+      }
+      if (_glewStrSame2(&pos, &len, (const GLubyte*)"SUN_", 4))
+      {
+#ifdef GLX_SUN_get_transparent_index
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"get_transparent_index", 21))
+        {
+          ret = GLXEW_SUN_get_transparent_index;
+          continue;
+        }
+#endif
+#ifdef GLX_SUN_video_resize
+        if (_glewStrSame3(&pos, &len, (const GLubyte*)"video_resize", 12))
+        {
+          ret = GLXEW_SUN_video_resize;
+          continue;
+        }
+#endif
+      }
+    }
+    ret = (len == 0);
+  }
+  return ret;
+}
+
+#endif /* _WIN32 */
diff --git a/src/glew/glewinfo.c b/src/glew/glewinfo.c
new file mode 100644
index 0000000000..13c6d05bd1
--- /dev/null
+++ b/src/glew/glewinfo.c
@@ -0,0 +1,7180 @@
+/*
+** The OpenGL Extension Wrangler Library
+** Copyright (C) 2002-2008, Milan Ikits <milan ikits[]ieee org>
+** Copyright (C) 2002-2008, Marcelo E. Magallon <mmagallo[]debian org>
+** Copyright (C) 2002, Lev Povalahev
+** All rights reserved.
+** 
+** Redistribution and use in source and binary forms, with or without 
+** modification, are permitted provided that the following conditions are met:
+** 
+** * Redistributions of source code must retain the above copyright notice, 
+**   this list of conditions and the following disclaimer.
+** * Redistributions in binary form must reproduce the above copyright notice, 
+**   this list of conditions and the following disclaimer in the documentation 
+**   and/or other materials provided with the distribution.
+** * The name of the author may be used to endorse or promote products 
+**   derived from this software without specific prior written permission.
+**
+** THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
+** AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 
+** IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+** ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 
+** LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 
+** CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 
+** SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+** INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+** CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+** ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+** THE POSSIBILITY OF SUCH DAMAGE.
+*/
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <GL/glew.h>
+#if defined(_WIN32)
+#include <GL/wglew.h>
+#elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+#include <GL/glxew.h>
+#endif
+
+static FILE* f;
+
+#ifdef GLEW_MX
+GLEWContext _glewctx;
+#define glewGetContext() (&_glewctx)
+#ifdef _WIN32
+WGLEWContext _wglewctx;
+#define wglewGetContext() (&_wglewctx)
+#elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+GLXEWContext _glxewctx;
+#define glxewGetContext() (&_glxewctx)
+#endif
+#endif
+
+#if defined(_WIN32)
+GLboolean glewCreateContext (int* pixelformat);
+#elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+GLboolean glewCreateContext (const char* display, int* visual);
+#else
+GLboolean glewCreateContext ();
+#endif
+
+#if defined(_WIN32) || !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+GLboolean glewParseArgs (int argc, char** argv, char** display, int* visual);
+#endif
+
+void glewDestroyContext ();
+
+/* ------------------------------------------------------------------------- */
+
+static void glewPrintExt (const char* name, GLboolean def1, GLboolean def2, GLboolean def3)
+{
+  unsigned int i;
+  fprintf(f, "\n%s:", name);
+  for (i=0; i<62-strlen(name); i++) fprintf(f, " ");
+  fprintf(f, "%s ", def1 ? "OK" : "MISSING");
+  if (def1 != def2)
+    fprintf(f, "[%s] ", def2 ? "OK" : "MISSING");
+  if (def1 != def3)
+    fprintf(f, "[%s]\n", def3 ? "OK" : "MISSING");
+  else
+    fprintf(f, "\n");
+  for (i=0; i<strlen(name)+1; i++) fprintf(f, "-");
+  fprintf(f, "\n");
+  fflush(f);
+}
+
+static void glewInfoFunc (const char* name, GLint undefined)
+{
+  unsigned int i;
+  fprintf(f, "  %s:", name);
+  for (i=0; i<60-strlen(name); i++) fprintf(f, " ");
+  fprintf(f, "%s\n", undefined ? "MISSING" : "OK");
+  fflush(f);
+}
+
+/* ----------------------------- GL_VERSION_1_1 ---------------------------- */
+
+#ifdef GL_VERSION_1_1
+
+static void _glewInfo_GL_VERSION_1_1 (void)
+{
+  glewPrintExt("GL_VERSION_1_1", GLEW_VERSION_1_1, GLEW_VERSION_1_1, GLEW_VERSION_1_1);
+}
+
+#endif /* GL_VERSION_1_1 */
+
+#ifdef GL_VERSION_1_2
+
+static void _glewInfo_GL_VERSION_1_2 (void)
+{
+  glewPrintExt("GL_VERSION_1_2", GLEW_VERSION_1_2, GLEW_VERSION_1_2, GLEW_VERSION_1_2);
+
+  glewInfoFunc("glCopyTexSubImage3D", glCopyTexSubImage3D == NULL);
+  glewInfoFunc("glDrawRangeElements", glDrawRangeElements == NULL);
+  glewInfoFunc("glTexImage3D", glTexImage3D == NULL);
+  glewInfoFunc("glTexSubImage3D", glTexSubImage3D == NULL);
+}
+
+#endif /* GL_VERSION_1_2 */
+
+#ifdef GL_VERSION_1_3
+
+static void _glewInfo_GL_VERSION_1_3 (void)
+{
+  glewPrintExt("GL_VERSION_1_3", GLEW_VERSION_1_3, GLEW_VERSION_1_3, GLEW_VERSION_1_3);
+
+  glewInfoFunc("glActiveTexture", glActiveTexture == NULL);
+  glewInfoFunc("glClientActiveTexture", glClientActiveTexture == NULL);
+  glewInfoFunc("glCompressedTexImage1D", glCompressedTexImage1D == NULL);
+  glewInfoFunc("glCompressedTexImage2D", glCompressedTexImage2D == NULL);
+  glewInfoFunc("glCompressedTexImage3D", glCompressedTexImage3D == NULL);
+  glewInfoFunc("glCompressedTexSubImage1D", glCompressedTexSubImage1D == NULL);
+  glewInfoFunc("glCompressedTexSubImage2D", glCompressedTexSubImage2D == NULL);
+  glewInfoFunc("glCompressedTexSubImage3D", glCompressedTexSubImage3D == NULL);
+  glewInfoFunc("glGetCompressedTexImage", glGetCompressedTexImage == NULL);
+  glewInfoFunc("glLoadTransposeMatrixd", glLoadTransposeMatrixd == NULL);
+  glewInfoFunc("glLoadTransposeMatrixf", glLoadTransposeMatrixf == NULL);
+  glewInfoFunc("glMultTransposeMatrixd", glMultTransposeMatrixd == NULL);
+  glewInfoFunc("glMultTransposeMatrixf", glMultTransposeMatrixf == NULL);
+  glewInfoFunc("glMultiTexCoord1d", glMultiTexCoord1d == NULL);
+  glewInfoFunc("glMultiTexCoord1dv", glMultiTexCoord1dv == NULL);
+  glewInfoFunc("glMultiTexCoord1f", glMultiTexCoord1f == NULL);
+  glewInfoFunc("glMultiTexCoord1fv", glMultiTexCoord1fv == NULL);
+  glewInfoFunc("glMultiTexCoord1i", glMultiTexCoord1i == NULL);
+  glewInfoFunc("glMultiTexCoord1iv", glMultiTexCoord1iv == NULL);
+  glewInfoFunc("glMultiTexCoord1s", glMultiTexCoord1s == NULL);
+  glewInfoFunc("glMultiTexCoord1sv", glMultiTexCoord1sv == NULL);
+  glewInfoFunc("glMultiTexCoord2d", glMultiTexCoord2d == NULL);
+  glewInfoFunc("glMultiTexCoord2dv", glMultiTexCoord2dv == NULL);
+  glewInfoFunc("glMultiTexCoord2f", glMultiTexCoord2f == NULL);
+  glewInfoFunc("glMultiTexCoord2fv", glMultiTexCoord2fv == NULL);
+  glewInfoFunc("glMultiTexCoord2i", glMultiTexCoord2i == NULL);
+  glewInfoFunc("glMultiTexCoord2iv", glMultiTexCoord2iv == NULL);
+  glewInfoFunc("glMultiTexCoord2s", glMultiTexCoord2s == NULL);
+  glewInfoFunc("glMultiTexCoord2sv", glMultiTexCoord2sv == NULL);
+  glewInfoFunc("glMultiTexCoord3d", glMultiTexCoord3d == NULL);
+  glewInfoFunc("glMultiTexCoord3dv", glMultiTexCoord3dv == NULL);
+  glewInfoFunc("glMultiTexCoord3f", glMultiTexCoord3f == NULL);
+  glewInfoFunc("glMultiTexCoord3fv", glMultiTexCoord3fv == NULL);
+  glewInfoFunc("glMultiTexCoord3i", glMultiTexCoord3i == NULL);
+  glewInfoFunc("glMultiTexCoord3iv", glMultiTexCoord3iv == NULL);
+  glewInfoFunc("glMultiTexCoord3s", glMultiTexCoord3s == NULL);
+  glewInfoFunc("glMultiTexCoord3sv", glMultiTexCoord3sv == NULL);
+  glewInfoFunc("glMultiTexCoord4d", glMultiTexCoord4d == NULL);
+  glewInfoFunc("glMultiTexCoord4dv", glMultiTexCoord4dv == NULL);
+  glewInfoFunc("glMultiTexCoord4f", glMultiTexCoord4f == NULL);
+  glewInfoFunc("glMultiTexCoord4fv", glMultiTexCoord4fv == NULL);
+  glewInfoFunc("glMultiTexCoord4i", glMultiTexCoord4i == NULL);
+  glewInfoFunc("glMultiTexCoord4iv", glMultiTexCoord4iv == NULL);
+  glewInfoFunc("glMultiTexCoord4s", glMultiTexCoord4s == NULL);
+  glewInfoFunc("glMultiTexCoord4sv", glMultiTexCoord4sv == NULL);
+  glewInfoFunc("glSampleCoverage", glSampleCoverage == NULL);
+}
+
+#endif /* GL_VERSION_1_3 */
+
+#ifdef GL_VERSION_1_4
+
+static void _glewInfo_GL_VERSION_1_4 (void)
+{
+  glewPrintExt("GL_VERSION_1_4", GLEW_VERSION_1_4, GLEW_VERSION_1_4, GLEW_VERSION_1_4);
+
+  glewInfoFunc("glBlendColor", glBlendColor == NULL);
+  glewInfoFunc("glBlendEquation", glBlendEquation == NULL);
+  glewInfoFunc("glBlendFuncSeparate", glBlendFuncSeparate == NULL);
+  glewInfoFunc("glFogCoordPointer", glFogCoordPointer == NULL);
+  glewInfoFunc("glFogCoordd", glFogCoordd == NULL);
+  glewInfoFunc("glFogCoorddv", glFogCoorddv == NULL);
+  glewInfoFunc("glFogCoordf", glFogCoordf == NULL);
+  glewInfoFunc("glFogCoordfv", glFogCoordfv == NULL);
+  glewInfoFunc("glMultiDrawArrays", glMultiDrawArrays == NULL);
+  glewInfoFunc("glMultiDrawElements", glMultiDrawElements == NULL);
+  glewInfoFunc("glPointParameterf", glPointParameterf == NULL);
+  glewInfoFunc("glPointParameterfv", glPointParameterfv == NULL);
+  glewInfoFunc("glPointParameteri", glPointParameteri == NULL);
+  glewInfoFunc("glPointParameteriv", glPointParameteriv == NULL);
+  glewInfoFunc("glSecondaryColor3b", glSecondaryColor3b == NULL);
+  glewInfoFunc("glSecondaryColor3bv", glSecondaryColor3bv == NULL);
+  glewInfoFunc("glSecondaryColor3d", glSecondaryColor3d == NULL);
+  glewInfoFunc("glSecondaryColor3dv", glSecondaryColor3dv == NULL);
+  glewInfoFunc("glSecondaryColor3f", glSecondaryColor3f == NULL);
+  glewInfoFunc("glSecondaryColor3fv", glSecondaryColor3fv == NULL);
+  glewInfoFunc("glSecondaryColor3i", glSecondaryColor3i == NULL);
+  glewInfoFunc("glSecondaryColor3iv", glSecondaryColor3iv == NULL);
+  glewInfoFunc("glSecondaryColor3s", glSecondaryColor3s == NULL);
+  glewInfoFunc("glSecondaryColor3sv", glSecondaryColor3sv == NULL);
+  glewInfoFunc("glSecondaryColor3ub", glSecondaryColor3ub == NULL);
+  glewInfoFunc("glSecondaryColor3ubv", glSecondaryColor3ubv == NULL);
+  glewInfoFunc("glSecondaryColor3ui", glSecondaryColor3ui == NULL);
+  glewInfoFunc("glSecondaryColor3uiv", glSecondaryColor3uiv == NULL);
+  glewInfoFunc("glSecondaryColor3us", glSecondaryColor3us == NULL);
+  glewInfoFunc("glSecondaryColor3usv", glSecondaryColor3usv == NULL);
+  glewInfoFunc("glSecondaryColorPointer", glSecondaryColorPointer == NULL);
+  glewInfoFunc("glWindowPos2d", glWindowPos2d == NULL);
+  glewInfoFunc("glWindowPos2dv", glWindowPos2dv == NULL);
+  glewInfoFunc("glWindowPos2f", glWindowPos2f == NULL);
+  glewInfoFunc("glWindowPos2fv", glWindowPos2fv == NULL);
+  glewInfoFunc("glWindowPos2i", glWindowPos2i == NULL);
+  glewInfoFunc("glWindowPos2iv", glWindowPos2iv == NULL);
+  glewInfoFunc("glWindowPos2s", glWindowPos2s == NULL);
+  glewInfoFunc("glWindowPos2sv", glWindowPos2sv == NULL);
+  glewInfoFunc("glWindowPos3d", glWindowPos3d == NULL);
+  glewInfoFunc("glWindowPos3dv", glWindowPos3dv == NULL);
+  glewInfoFunc("glWindowPos3f", glWindowPos3f == NULL);
+  glewInfoFunc("glWindowPos3fv", glWindowPos3fv == NULL);
+  glewInfoFunc("glWindowPos3i", glWindowPos3i == NULL);
+  glewInfoFunc("glWindowPos3iv", glWindowPos3iv == NULL);
+  glewInfoFunc("glWindowPos3s", glWindowPos3s == NULL);
+  glewInfoFunc("glWindowPos3sv", glWindowPos3sv == NULL);
+}
+
+#endif /* GL_VERSION_1_4 */
+
+#ifdef GL_VERSION_1_5
+
+static void _glewInfo_GL_VERSION_1_5 (void)
+{
+  glewPrintExt("GL_VERSION_1_5", GLEW_VERSION_1_5, GLEW_VERSION_1_5, GLEW_VERSION_1_5);
+
+  glewInfoFunc("glBeginQuery", glBeginQuery == NULL);
+  glewInfoFunc("glBindBuffer", glBindBuffer == NULL);
+  glewInfoFunc("glBufferData", glBufferData == NULL);
+  glewInfoFunc("glBufferSubData", glBufferSubData == NULL);
+  glewInfoFunc("glDeleteBuffers", glDeleteBuffers == NULL);
+  glewInfoFunc("glDeleteQueries", glDeleteQueries == NULL);
+  glewInfoFunc("glEndQuery", glEndQuery == NULL);
+  glewInfoFunc("glGenBuffers", glGenBuffers == NULL);
+  glewInfoFunc("glGenQueries", glGenQueries == NULL);
+  glewInfoFunc("glGetBufferParameteriv", glGetBufferParameteriv == NULL);
+  glewInfoFunc("glGetBufferPointerv", glGetBufferPointerv == NULL);
+  glewInfoFunc("glGetBufferSubData", glGetBufferSubData == NULL);
+  glewInfoFunc("glGetQueryObjectiv", glGetQueryObjectiv == NULL);
+  glewInfoFunc("glGetQueryObjectuiv", glGetQueryObjectuiv == NULL);
+  glewInfoFunc("glGetQueryiv", glGetQueryiv == NULL);
+  glewInfoFunc("glIsBuffer", glIsBuffer == NULL);
+  glewInfoFunc("glIsQuery", glIsQuery == NULL);
+  glewInfoFunc("glMapBuffer", glMapBuffer == NULL);
+  glewInfoFunc("glUnmapBuffer", glUnmapBuffer == NULL);
+}
+
+#endif /* GL_VERSION_1_5 */
+
+#ifdef GL_VERSION_2_0
+
+static void _glewInfo_GL_VERSION_2_0 (void)
+{
+  glewPrintExt("GL_VERSION_2_0", GLEW_VERSION_2_0, GLEW_VERSION_2_0, GLEW_VERSION_2_0);
+
+  glewInfoFunc("glAttachShader", glAttachShader == NULL);
+  glewInfoFunc("glBindAttribLocation", glBindAttribLocation == NULL);
+  glewInfoFunc("glBlendEquationSeparate", glBlendEquationSeparate == NULL);
+  glewInfoFunc("glCompileShader", glCompileShader == NULL);
+  glewInfoFunc("glCreateProgram", glCreateProgram == NULL);
+  glewInfoFunc("glCreateShader", glCreateShader == NULL);
+  glewInfoFunc("glDeleteProgram", glDeleteProgram == NULL);
+  glewInfoFunc("glDeleteShader", glDeleteShader == NULL);
+  glewInfoFunc("glDetachShader", glDetachShader == NULL);
+  glewInfoFunc("glDisableVertexAttribArray", glDisableVertexAttribArray == NULL);
+  glewInfoFunc("glDrawBuffers", glDrawBuffers == NULL);
+  glewInfoFunc("glEnableVertexAttribArray", glEnableVertexAttribArray == NULL);
+  glewInfoFunc("glGetActiveAttrib", glGetActiveAttrib == NULL);
+  glewInfoFunc("glGetActiveUniform", glGetActiveUniform == NULL);
+  glewInfoFunc("glGetAttachedShaders", glGetAttachedShaders == NULL);
+  glewInfoFunc("glGetAttribLocation", glGetAttribLocation == NULL);
+  glewInfoFunc("glGetProgramInfoLog", glGetProgramInfoLog == NULL);
+  glewInfoFunc("glGetProgramiv", glGetProgramiv == NULL);
+  glewInfoFunc("glGetShaderInfoLog", glGetShaderInfoLog == NULL);
+  glewInfoFunc("glGetShaderSource", glGetShaderSource == NULL);
+  glewInfoFunc("glGetShaderiv", glGetShaderiv == NULL);
+  glewInfoFunc("glGetUniformLocation", glGetUniformLocation == NULL);
+  glewInfoFunc("glGetUniformfv", glGetUniformfv == NULL);
+  glewInfoFunc("glGetUniformiv", glGetUniformiv == NULL);
+  glewInfoFunc("glGetVertexAttribPointerv", glGetVertexAttribPointerv == NULL);
+  glewInfoFunc("glGetVertexAttribdv", glGetVertexAttribdv == NULL);
+  glewInfoFunc("glGetVertexAttribfv", glGetVertexAttribfv == NULL);
+  glewInfoFunc("glGetVertexAttribiv", glGetVertexAttribiv == NULL);
+  glewInfoFunc("glIsProgram", glIsProgram == NULL);
+  glewInfoFunc("glIsShader", glIsShader == NULL);
+  glewInfoFunc("glLinkProgram", glLinkProgram == NULL);
+  glewInfoFunc("glShaderSource", glShaderSource == NULL);
+  glewInfoFunc("glStencilFuncSeparate", glStencilFuncSeparate == NULL);
+  glewInfoFunc("glStencilMaskSeparate", glStencilMaskSeparate == NULL);
+  glewInfoFunc("glStencilOpSeparate", glStencilOpSeparate == NULL);
+  glewInfoFunc("glUniform1f", glUniform1f == NULL);
+  glewInfoFunc("glUniform1fv", glUniform1fv == NULL);
+  glewInfoFunc("glUniform1i", glUniform1i == NULL);
+  glewInfoFunc("glUniform1iv", glUniform1iv == NULL);
+  glewInfoFunc("glUniform2f", glUniform2f == NULL);
+  glewInfoFunc("glUniform2fv", glUniform2fv == NULL);
+  glewInfoFunc("glUniform2i", glUniform2i == NULL);
+  glewInfoFunc("glUniform2iv", glUniform2iv == NULL);
+  glewInfoFunc("glUniform3f", glUniform3f == NULL);
+  glewInfoFunc("glUniform3fv", glUniform3fv == NULL);
+  glewInfoFunc("glUniform3i", glUniform3i == NULL);
+  glewInfoFunc("glUniform3iv", glUniform3iv == NULL);
+  glewInfoFunc("glUniform4f", glUniform4f == NULL);
+  glewInfoFunc("glUniform4fv", glUniform4fv == NULL);
+  glewInfoFunc("glUniform4i", glUniform4i == NULL);
+  glewInfoFunc("glUniform4iv", glUniform4iv == NULL);
+  glewInfoFunc("glUniformMatrix2fv", glUniformMatrix2fv == NULL);
+  glewInfoFunc("glUniformMatrix3fv", glUniformMatrix3fv == NULL);
+  glewInfoFunc("glUniformMatrix4fv", glUniformMatrix4fv == NULL);
+  glewInfoFunc("glUseProgram", glUseProgram == NULL);
+  glewInfoFunc("glValidateProgram", glValidateProgram == NULL);
+  glewInfoFunc("glVertexAttrib1d", glVertexAttrib1d == NULL);
+  glewInfoFunc("glVertexAttrib1dv", glVertexAttrib1dv == NULL);
+  glewInfoFunc("glVertexAttrib1f", glVertexAttrib1f == NULL);
+  glewInfoFunc("glVertexAttrib1fv", glVertexAttrib1fv == NULL);
+  glewInfoFunc("glVertexAttrib1s", glVertexAttrib1s == NULL);
+  glewInfoFunc("glVertexAttrib1sv", glVertexAttrib1sv == NULL);
+  glewInfoFunc("glVertexAttrib2d", glVertexAttrib2d == NULL);
+  glewInfoFunc("glVertexAttrib2dv", glVertexAttrib2dv == NULL);
+  glewInfoFunc("glVertexAttrib2f", glVertexAttrib2f == NULL);
+  glewInfoFunc("glVertexAttrib2fv", glVertexAttrib2fv == NULL);
+  glewInfoFunc("glVertexAttrib2s", glVertexAttrib2s == NULL);
+  glewInfoFunc("glVertexAttrib2sv", glVertexAttrib2sv == NULL);
+  glewInfoFunc("glVertexAttrib3d", glVertexAttrib3d == NULL);
+  glewInfoFunc("glVertexAttrib3dv", glVertexAttrib3dv == NULL);
+  glewInfoFunc("glVertexAttrib3f", glVertexAttrib3f == NULL);
+  glewInfoFunc("glVertexAttrib3fv", glVertexAttrib3fv == NULL);
+  glewInfoFunc("glVertexAttrib3s", glVertexAttrib3s == NULL);
+  glewInfoFunc("glVertexAttrib3sv", glVertexAttrib3sv == NULL);
+  glewInfoFunc("glVertexAttrib4Nbv", glVertexAttrib4Nbv == NULL);
+  glewInfoFunc("glVertexAttrib4Niv", glVertexAttrib4Niv == NULL);
+  glewInfoFunc("glVertexAttrib4Nsv", glVertexAttrib4Nsv == NULL);
+  glewInfoFunc("glVertexAttrib4Nub", glVertexAttrib4Nub == NULL);
+  glewInfoFunc("glVertexAttrib4Nubv", glVertexAttrib4Nubv == NULL);
+  glewInfoFunc("glVertexAttrib4Nuiv", glVertexAttrib4Nuiv == NULL);
+  glewInfoFunc("glVertexAttrib4Nusv", glVertexAttrib4Nusv == NULL);
+  glewInfoFunc("glVertexAttrib4bv", glVertexAttrib4bv == NULL);
+  glewInfoFunc("glVertexAttrib4d", glVertexAttrib4d == NULL);
+  glewInfoFunc("glVertexAttrib4dv", glVertexAttrib4dv == NULL);
+  glewInfoFunc("glVertexAttrib4f", glVertexAttrib4f == NULL);
+  glewInfoFunc("glVertexAttrib4fv", glVertexAttrib4fv == NULL);
+  glewInfoFunc("glVertexAttrib4iv", glVertexAttrib4iv == NULL);
+  glewInfoFunc("glVertexAttrib4s", glVertexAttrib4s == NULL);
+  glewInfoFunc("glVertexAttrib4sv", glVertexAttrib4sv == NULL);
+  glewInfoFunc("glVertexAttrib4ubv", glVertexAttrib4ubv == NULL);
+  glewInfoFunc("glVertexAttrib4uiv", glVertexAttrib4uiv == NULL);
+  glewInfoFunc("glVertexAttrib4usv", glVertexAttrib4usv == NULL);
+  glewInfoFunc("glVertexAttribPointer", glVertexAttribPointer == NULL);
+}
+
+#endif /* GL_VERSION_2_0 */
+
+#ifdef GL_VERSION_2_1
+
+static void _glewInfo_GL_VERSION_2_1 (void)
+{
+  glewPrintExt("GL_VERSION_2_1", GLEW_VERSION_2_1, GLEW_VERSION_2_1, GLEW_VERSION_2_1);
+
+  glewInfoFunc("glUniformMatrix2x3fv", glUniformMatrix2x3fv == NULL);
+  glewInfoFunc("glUniformMatrix2x4fv", glUniformMatrix2x4fv == NULL);
+  glewInfoFunc("glUniformMatrix3x2fv", glUniformMatrix3x2fv == NULL);
+  glewInfoFunc("glUniformMatrix3x4fv", glUniformMatrix3x4fv == NULL);
+  glewInfoFunc("glUniformMatrix4x2fv", glUniformMatrix4x2fv == NULL);
+  glewInfoFunc("glUniformMatrix4x3fv", glUniformMatrix4x3fv == NULL);
+}
+
+#endif /* GL_VERSION_2_1 */
+
+#ifdef GL_VERSION_3_0
+
+static void _glewInfo_GL_VERSION_3_0 (void)
+{
+  glewPrintExt("GL_VERSION_3_0", GLEW_VERSION_3_0, GLEW_VERSION_3_0, GLEW_VERSION_3_0);
+
+  glewInfoFunc("glBeginConditionalRender", glBeginConditionalRender == NULL);
+  glewInfoFunc("glBeginTransformFeedback", glBeginTransformFeedback == NULL);
+  glewInfoFunc("glBindBufferBase", glBindBufferBase == NULL);
+  glewInfoFunc("glBindBufferRange", glBindBufferRange == NULL);
+  glewInfoFunc("glBindFragDataLocation", glBindFragDataLocation == NULL);
+  glewInfoFunc("glClampColor", glClampColor == NULL);
+  glewInfoFunc("glClearBufferfi", glClearBufferfi == NULL);
+  glewInfoFunc("glClearBufferfv", glClearBufferfv == NULL);
+  glewInfoFunc("glClearBufferiv", glClearBufferiv == NULL);
+  glewInfoFunc("glClearBufferuiv", glClearBufferuiv == NULL);
+  glewInfoFunc("glColorMaski", glColorMaski == NULL);
+  glewInfoFunc("glDisablei", glDisablei == NULL);
+  glewInfoFunc("glEnablei", glEnablei == NULL);
+  glewInfoFunc("glEndConditionalRender", glEndConditionalRender == NULL);
+  glewInfoFunc("glEndTransformFeedback", glEndTransformFeedback == NULL);
+  glewInfoFunc("glGetBooleani_v", glGetBooleani_v == NULL);
+  glewInfoFunc("glGetFragDataLocation", glGetFragDataLocation == NULL);
+  glewInfoFunc("glGetIntegeri_v", glGetIntegeri_v == NULL);
+  glewInfoFunc("glGetStringi", glGetStringi == NULL);
+  glewInfoFunc("glGetTexParameterIiv", glGetTexParameterIiv == NULL);
+  glewInfoFunc("glGetTexParameterIuiv", glGetTexParameterIuiv == NULL);
+  glewInfoFunc("glGetTransformFeedbackVarying", glGetTransformFeedbackVarying == NULL);
+  glewInfoFunc("glGetUniformuiv", glGetUniformuiv == NULL);
+  glewInfoFunc("glGetVertexAttribIiv", glGetVertexAttribIiv == NULL);
+  glewInfoFunc("glGetVertexAttribIuiv", glGetVertexAttribIuiv == NULL);
+  glewInfoFunc("glIsEnabledi", glIsEnabledi == NULL);
+  glewInfoFunc("glTexParameterIiv", glTexParameterIiv == NULL);
+  glewInfoFunc("glTexParameterIuiv", glTexParameterIuiv == NULL);
+  glewInfoFunc("glTransformFeedbackVaryings", glTransformFeedbackVaryings == NULL);
+  glewInfoFunc("glUniform1ui", glUniform1ui == NULL);
+  glewInfoFunc("glUniform1uiv", glUniform1uiv == NULL);
+  glewInfoFunc("glUniform2ui", glUniform2ui == NULL);
+  glewInfoFunc("glUniform2uiv", glUniform2uiv == NULL);
+  glewInfoFunc("glUniform3ui", glUniform3ui == NULL);
+  glewInfoFunc("glUniform3uiv", glUniform3uiv == NULL);
+  glewInfoFunc("glUniform4ui", glUniform4ui == NULL);
+  glewInfoFunc("glUniform4uiv", glUniform4uiv == NULL);
+  glewInfoFunc("glVertexAttribI1i", glVertexAttribI1i == NULL);
+  glewInfoFunc("glVertexAttribI1iv", glVertexAttribI1iv == NULL);
+  glewInfoFunc("glVertexAttribI1ui", glVertexAttribI1ui == NULL);
+  glewInfoFunc("glVertexAttribI1uiv", glVertexAttribI1uiv == NULL);
+  glewInfoFunc("glVertexAttribI2i", glVertexAttribI2i == NULL);
+  glewInfoFunc("glVertexAttribI2iv", glVertexAttribI2iv == NULL);
+  glewInfoFunc("glVertexAttribI2ui", glVertexAttribI2ui == NULL);
+  glewInfoFunc("glVertexAttribI2uiv", glVertexAttribI2uiv == NULL);
+  glewInfoFunc("glVertexAttribI3i", glVertexAttribI3i == NULL);
+  glewInfoFunc("glVertexAttribI3iv", glVertexAttribI3iv == NULL);
+  glewInfoFunc("glVertexAttribI3ui", glVertexAttribI3ui == NULL);
+  glewInfoFunc("glVertexAttribI3uiv", glVertexAttribI3uiv == NULL);
+  glewInfoFunc("glVertexAttribI4bv", glVertexAttribI4bv == NULL);
+  glewInfoFunc("glVertexAttribI4i", glVertexAttribI4i == NULL);
+  glewInfoFunc("glVertexAttribI4iv", glVertexAttribI4iv == NULL);
+  glewInfoFunc("glVertexAttribI4sv", glVertexAttribI4sv == NULL);
+  glewInfoFunc("glVertexAttribI4ubv", glVertexAttribI4ubv == NULL);
+  glewInfoFunc("glVertexAttribI4ui", glVertexAttribI4ui == NULL);
+  glewInfoFunc("glVertexAttribI4uiv", glVertexAttribI4uiv == NULL);
+  glewInfoFunc("glVertexAttribI4usv", glVertexAttribI4usv == NULL);
+  glewInfoFunc("glVertexAttribIPointer", glVertexAttribIPointer == NULL);
+}
+
+#endif /* GL_VERSION_3_0 */
+
+#ifdef GL_3DFX_multisample
+
+static void _glewInfo_GL_3DFX_multisample (void)
+{
+  glewPrintExt("GL_3DFX_multisample", GLEW_3DFX_multisample, glewIsSupported("GL_3DFX_multisample"), glewGetExtension("GL_3DFX_multisample"));
+}
+
+#endif /* GL_3DFX_multisample */
+
+#ifdef GL_3DFX_tbuffer
+
+static void _glewInfo_GL_3DFX_tbuffer (void)
+{
+  glewPrintExt("GL_3DFX_tbuffer", GLEW_3DFX_tbuffer, glewIsSupported("GL_3DFX_tbuffer"), glewGetExtension("GL_3DFX_tbuffer"));
+
+  glewInfoFunc("glTbufferMask3DFX", glTbufferMask3DFX == NULL);
+}
+
+#endif /* GL_3DFX_tbuffer */
+
+#ifdef GL_3DFX_texture_compression_FXT1
+
+static void _glewInfo_GL_3DFX_texture_compression_FXT1 (void)
+{
+  glewPrintExt("GL_3DFX_texture_compression_FXT1", GLEW_3DFX_texture_compression_FXT1, glewIsSupported("GL_3DFX_texture_compression_FXT1"), glewGetExtension("GL_3DFX_texture_compression_FXT1"));
+}
+
+#endif /* GL_3DFX_texture_compression_FXT1 */
+
+#ifdef GL_APPLE_client_storage
+
+static void _glewInfo_GL_APPLE_client_storage (void)
+{
+  glewPrintExt("GL_APPLE_client_storage", GLEW_APPLE_client_storage, glewIsSupported("GL_APPLE_client_storage"), glewGetExtension("GL_APPLE_client_storage"));
+}
+
+#endif /* GL_APPLE_client_storage */
+
+#ifdef GL_APPLE_element_array
+
+static void _glewInfo_GL_APPLE_element_array (void)
+{
+  glewPrintExt("GL_APPLE_element_array", GLEW_APPLE_element_array, glewIsSupported("GL_APPLE_element_array"), glewGetExtension("GL_APPLE_element_array"));
+
+  glewInfoFunc("glDrawElementArrayAPPLE", glDrawElementArrayAPPLE == NULL);
+  glewInfoFunc("glDrawRangeElementArrayAPPLE", glDrawRangeElementArrayAPPLE == NULL);
+  glewInfoFunc("glElementPointerAPPLE", glElementPointerAPPLE == NULL);
+  glewInfoFunc("glMultiDrawElementArrayAPPLE", glMultiDrawElementArrayAPPLE == NULL);
+  glewInfoFunc("glMultiDrawRangeElementArrayAPPLE", glMultiDrawRangeElementArrayAPPLE == NULL);
+}
+
+#endif /* GL_APPLE_element_array */
+
+#ifdef GL_APPLE_fence
+
+static void _glewInfo_GL_APPLE_fence (void)
+{
+  glewPrintExt("GL_APPLE_fence", GLEW_APPLE_fence, glewIsSupported("GL_APPLE_fence"), glewGetExtension("GL_APPLE_fence"));
+
+  glewInfoFunc("glDeleteFencesAPPLE", glDeleteFencesAPPLE == NULL);
+  glewInfoFunc("glFinishFenceAPPLE", glFinishFenceAPPLE == NULL);
+  glewInfoFunc("glFinishObjectAPPLE", glFinishObjectAPPLE == NULL);
+  glewInfoFunc("glGenFencesAPPLE", glGenFencesAPPLE == NULL);
+  glewInfoFunc("glIsFenceAPPLE", glIsFenceAPPLE == NULL);
+  glewInfoFunc("glSetFenceAPPLE", glSetFenceAPPLE == NULL);
+  glewInfoFunc("glTestFenceAPPLE", glTestFenceAPPLE == NULL);
+  glewInfoFunc("glTestObjectAPPLE", glTestObjectAPPLE == NULL);
+}
+
+#endif /* GL_APPLE_fence */
+
+#ifdef GL_APPLE_float_pixels
+
+static void _glewInfo_GL_APPLE_float_pixels (void)
+{
+  glewPrintExt("GL_APPLE_float_pixels", GLEW_APPLE_float_pixels, glewIsSupported("GL_APPLE_float_pixels"), glewGetExtension("GL_APPLE_float_pixels"));
+}
+
+#endif /* GL_APPLE_float_pixels */
+
+#ifdef GL_APPLE_flush_buffer_range
+
+static void _glewInfo_GL_APPLE_flush_buffer_range (void)
+{
+  glewPrintExt("GL_APPLE_flush_buffer_range", GLEW_APPLE_flush_buffer_range, glewIsSupported("GL_APPLE_flush_buffer_range"), glewGetExtension("GL_APPLE_flush_buffer_range"));
+
+  glewInfoFunc("glBufferParameteriAPPLE", glBufferParameteriAPPLE == NULL);
+  glewInfoFunc("glFlushMappedBufferRangeAPPLE", glFlushMappedBufferRangeAPPLE == NULL);
+}
+
+#endif /* GL_APPLE_flush_buffer_range */
+
+#ifdef GL_APPLE_pixel_buffer
+
+static void _glewInfo_GL_APPLE_pixel_buffer (void)
+{
+  glewPrintExt("GL_APPLE_pixel_buffer", GLEW_APPLE_pixel_buffer, glewIsSupported("GL_APPLE_pixel_buffer"), glewGetExtension("GL_APPLE_pixel_buffer"));
+}
+
+#endif /* GL_APPLE_pixel_buffer */
+
+#ifdef GL_APPLE_specular_vector
+
+static void _glewInfo_GL_APPLE_specular_vector (void)
+{
+  glewPrintExt("GL_APPLE_specular_vector", GLEW_APPLE_specular_vector, glewIsSupported("GL_APPLE_specular_vector"), glewGetExtension("GL_APPLE_specular_vector"));
+}
+
+#endif /* GL_APPLE_specular_vector */
+
+#ifdef GL_APPLE_texture_range
+
+static void _glewInfo_GL_APPLE_texture_range (void)
+{
+  glewPrintExt("GL_APPLE_texture_range", GLEW_APPLE_texture_range, glewIsSupported("GL_APPLE_texture_range"), glewGetExtension("GL_APPLE_texture_range"));
+
+  glewInfoFunc("glGetTexParameterPointervAPPLE", glGetTexParameterPointervAPPLE == NULL);
+  glewInfoFunc("glTextureRangeAPPLE", glTextureRangeAPPLE == NULL);
+}
+
+#endif /* GL_APPLE_texture_range */
+
+#ifdef GL_APPLE_transform_hint
+
+static void _glewInfo_GL_APPLE_transform_hint (void)
+{
+  glewPrintExt("GL_APPLE_transform_hint", GLEW_APPLE_transform_hint, glewIsSupported("GL_APPLE_transform_hint"), glewGetExtension("GL_APPLE_transform_hint"));
+}
+
+#endif /* GL_APPLE_transform_hint */
+
+#ifdef GL_APPLE_vertex_array_object
+
+static void _glewInfo_GL_APPLE_vertex_array_object (void)
+{
+  glewPrintExt("GL_APPLE_vertex_array_object", GLEW_APPLE_vertex_array_object, glewIsSupported("GL_APPLE_vertex_array_object"), glewGetExtension("GL_APPLE_vertex_array_object"));
+
+  glewInfoFunc("glBindVertexArrayAPPLE", glBindVertexArrayAPPLE == NULL);
+  glewInfoFunc("glDeleteVertexArraysAPPLE", glDeleteVertexArraysAPPLE == NULL);
+  glewInfoFunc("glGenVertexArraysAPPLE", glGenVertexArraysAPPLE == NULL);
+  glewInfoFunc("glIsVertexArrayAPPLE", glIsVertexArrayAPPLE == NULL);
+}
+
+#endif /* GL_APPLE_vertex_array_object */
+
+#ifdef GL_APPLE_vertex_array_range
+
+static void _glewInfo_GL_APPLE_vertex_array_range (void)
+{
+  glewPrintExt("GL_APPLE_vertex_array_range", GLEW_APPLE_vertex_array_range, glewIsSupported("GL_APPLE_vertex_array_range"), glewGetExtension("GL_APPLE_vertex_array_range"));
+
+  glewInfoFunc("glFlushVertexArrayRangeAPPLE", glFlushVertexArrayRangeAPPLE == NULL);
+  glewInfoFunc("glVertexArrayParameteriAPPLE", glVertexArrayParameteriAPPLE == NULL);
+  glewInfoFunc("glVertexArrayRangeAPPLE", glVertexArrayRangeAPPLE == NULL);
+}
+
+#endif /* GL_APPLE_vertex_array_range */
+
+#ifdef GL_APPLE_ycbcr_422
+
+static void _glewInfo_GL_APPLE_ycbcr_422 (void)
+{
+  glewPrintExt("GL_APPLE_ycbcr_422", GLEW_APPLE_ycbcr_422, glewIsSupported("GL_APPLE_ycbcr_422"), glewGetExtension("GL_APPLE_ycbcr_422"));
+}
+
+#endif /* GL_APPLE_ycbcr_422 */
+
+#ifdef GL_ARB_color_buffer_float
+
+static void _glewInfo_GL_ARB_color_buffer_float (void)
+{
+  glewPrintExt("GL_ARB_color_buffer_float", GLEW_ARB_color_buffer_float, glewIsSupported("GL_ARB_color_buffer_float"), glewGetExtension("GL_ARB_color_buffer_float"));
+
+  glewInfoFunc("glClampColorARB", glClampColorARB == NULL);
+}
+
+#endif /* GL_ARB_color_buffer_float */
+
+#ifdef GL_ARB_depth_buffer_float
+
+static void _glewInfo_GL_ARB_depth_buffer_float (void)
+{
+  glewPrintExt("GL_ARB_depth_buffer_float", GLEW_ARB_depth_buffer_float, glewIsSupported("GL_ARB_depth_buffer_float"), glewGetExtension("GL_ARB_depth_buffer_float"));
+}
+
+#endif /* GL_ARB_depth_buffer_float */
+
+#ifdef GL_ARB_depth_texture
+
+static void _glewInfo_GL_ARB_depth_texture (void)
+{
+  glewPrintExt("GL_ARB_depth_texture", GLEW_ARB_depth_texture, glewIsSupported("GL_ARB_depth_texture"), glewGetExtension("GL_ARB_depth_texture"));
+}
+
+#endif /* GL_ARB_depth_texture */
+
+#ifdef GL_ARB_draw_buffers
+
+static void _glewInfo_GL_ARB_draw_buffers (void)
+{
+  glewPrintExt("GL_ARB_draw_buffers", GLEW_ARB_draw_buffers, glewIsSupported("GL_ARB_draw_buffers"), glewGetExtension("GL_ARB_draw_buffers"));
+
+  glewInfoFunc("glDrawBuffersARB", glDrawBuffersARB == NULL);
+}
+
+#endif /* GL_ARB_draw_buffers */
+
+#ifdef GL_ARB_draw_instanced
+
+static void _glewInfo_GL_ARB_draw_instanced (void)
+{
+  glewPrintExt("GL_ARB_draw_instanced", GLEW_ARB_draw_instanced, glewIsSupported("GL_ARB_draw_instanced"), glewGetExtension("GL_ARB_draw_instanced"));
+
+  glewInfoFunc("glDrawArraysInstancedARB", glDrawArraysInstancedARB == NULL);
+  glewInfoFunc("glDrawElementsInstancedARB", glDrawElementsInstancedARB == NULL);
+}
+
+#endif /* GL_ARB_draw_instanced */
+
+#ifdef GL_ARB_fragment_program
+
+static void _glewInfo_GL_ARB_fragment_program (void)
+{
+  glewPrintExt("GL_ARB_fragment_program", GLEW_ARB_fragment_program, glewIsSupported("GL_ARB_fragment_program"), glewGetExtension("GL_ARB_fragment_program"));
+}
+
+#endif /* GL_ARB_fragment_program */
+
+#ifdef GL_ARB_fragment_program_shadow
+
+static void _glewInfo_GL_ARB_fragment_program_shadow (void)
+{
+  glewPrintExt("GL_ARB_fragment_program_shadow", GLEW_ARB_fragment_program_shadow, glewIsSupported("GL_ARB_fragment_program_shadow"), glewGetExtension("GL_ARB_fragment_program_shadow"));
+}
+
+#endif /* GL_ARB_fragment_program_shadow */
+
+#ifdef GL_ARB_fragment_shader
+
+static void _glewInfo_GL_ARB_fragment_shader (void)
+{
+  glewPrintExt("GL_ARB_fragment_shader", GLEW_ARB_fragment_shader, glewIsSupported("GL_ARB_fragment_shader"), glewGetExtension("GL_ARB_fragment_shader"));
+}
+
+#endif /* GL_ARB_fragment_shader */
+
+#ifdef GL_ARB_framebuffer_object
+
+static void _glewInfo_GL_ARB_framebuffer_object (void)
+{
+  glewPrintExt("GL_ARB_framebuffer_object", GLEW_ARB_framebuffer_object, glewIsSupported("GL_ARB_framebuffer_object"), glewGetExtension("GL_ARB_framebuffer_object"));
+
+  glewInfoFunc("glBindFramebuffer", glBindFramebuffer == NULL);
+  glewInfoFunc("glBindRenderbuffer", glBindRenderbuffer == NULL);
+  glewInfoFunc("glBlitFramebuffer", glBlitFramebuffer == NULL);
+  glewInfoFunc("glCheckFramebufferStatus", glCheckFramebufferStatus == NULL);
+  glewInfoFunc("glDeleteFramebuffers", glDeleteFramebuffers == NULL);
+  glewInfoFunc("glDeleteRenderbuffers", glDeleteRenderbuffers == NULL);
+  glewInfoFunc("glFramebufferRenderbuffer", glFramebufferRenderbuffer == NULL);
+  glewInfoFunc("glFramebufferTexturLayer", glFramebufferTexturLayer == NULL);
+  glewInfoFunc("glFramebufferTexture1D", glFramebufferTexture1D == NULL);
+  glewInfoFunc("glFramebufferTexture2D", glFramebufferTexture2D == NULL);
+  glewInfoFunc("glFramebufferTexture3D", glFramebufferTexture3D == NULL);
+  glewInfoFunc("glGenFramebuffers", glGenFramebuffers == NULL);
+  glewInfoFunc("glGenRenderbuffers", glGenRenderbuffers == NULL);
+  glewInfoFunc("glGenerateMipmap", glGenerateMipmap == NULL);
+  glewInfoFunc("glGetFramebufferAttachmentParameteriv", glGetFramebufferAttachmentParameteriv == NULL);
+  glewInfoFunc("glGetRenderbufferParameteriv", glGetRenderbufferParameteriv == NULL);
+  glewInfoFunc("glIsFramebuffer", glIsFramebuffer == NULL);
+  glewInfoFunc("glIsRenderbuffer", glIsRenderbuffer == NULL);
+  glewInfoFunc("glRenderbufferStorage", glRenderbufferStorage == NULL);
+  glewInfoFunc("glRenderbufferStorageMultisample", glRenderbufferStorageMultisample == NULL);
+}
+
+#endif /* GL_ARB_framebuffer_object */
+
+#ifdef GL_ARB_framebuffer_sRGB
+
+static void _glewInfo_GL_ARB_framebuffer_sRGB (void)
+{
+  glewPrintExt("GL_ARB_framebuffer_sRGB", GLEW_ARB_framebuffer_sRGB, glewIsSupported("GL_ARB_framebuffer_sRGB"), glewGetExtension("GL_ARB_framebuffer_sRGB"));
+}
+
+#endif /* GL_ARB_framebuffer_sRGB */
+
+#ifdef GL_ARB_geometry_shader4
+
+static void _glewInfo_GL_ARB_geometry_shader4 (void)
+{
+  glewPrintExt("GL_ARB_geometry_shader4", GLEW_ARB_geometry_shader4, glewIsSupported("GL_ARB_geometry_shader4"), glewGetExtension("GL_ARB_geometry_shader4"));
+
+  glewInfoFunc("glFramebufferTextureARB", glFramebufferTextureARB == NULL);
+  glewInfoFunc("glFramebufferTextureFaceARB", glFramebufferTextureFaceARB == NULL);
+  glewInfoFunc("glFramebufferTextureLayerARB", glFramebufferTextureLayerARB == NULL);
+  glewInfoFunc("glProgramParameteriARB", glProgramParameteriARB == NULL);
+}
+
+#endif /* GL_ARB_geometry_shader4 */
+
+#ifdef GL_ARB_half_float_pixel
+
+static void _glewInfo_GL_ARB_half_float_pixel (void)
+{
+  glewPrintExt("GL_ARB_half_float_pixel", GLEW_ARB_half_float_pixel, glewIsSupported("GL_ARB_half_float_pixel"), glewGetExtension("GL_ARB_half_float_pixel"));
+}
+
+#endif /* GL_ARB_half_float_pixel */
+
+#ifdef GL_ARB_half_float_vertex
+
+static void _glewInfo_GL_ARB_half_float_vertex (void)
+{
+  glewPrintExt("GL_ARB_half_float_vertex", GLEW_ARB_half_float_vertex, glewIsSupported("GL_ARB_half_float_vertex"), glewGetExtension("GL_ARB_half_float_vertex"));
+}
+
+#endif /* GL_ARB_half_float_vertex */
+
+#ifdef GL_ARB_imaging
+
+static void _glewInfo_GL_ARB_imaging (void)
+{
+  glewPrintExt("GL_ARB_imaging", GLEW_ARB_imaging, glewIsSupported("GL_ARB_imaging"), glewGetExtension("GL_ARB_imaging"));
+
+  glewInfoFunc("glBlendEquation", glBlendEquation == NULL);
+  glewInfoFunc("glColorSubTable", glColorSubTable == NULL);
+  glewInfoFunc("glColorTable", glColorTable == NULL);
+  glewInfoFunc("glColorTableParameterfv", glColorTableParameterfv == NULL);
+  glewInfoFunc("glColorTableParameteriv", glColorTableParameteriv == NULL);
+  glewInfoFunc("glConvolutionFilter1D", glConvolutionFilter1D == NULL);
+  glewInfoFunc("glConvolutionFilter2D", glConvolutionFilter2D == NULL);
+  glewInfoFunc("glConvolutionParameterf", glConvolutionParameterf == NULL);
+  glewInfoFunc("glConvolutionParameterfv", glConvolutionParameterfv == NULL);
+  glewInfoFunc("glConvolutionParameteri", glConvolutionParameteri == NULL);
+  glewInfoFunc("glConvolutionParameteriv", glConvolutionParameteriv == NULL);
+  glewInfoFunc("glCopyColorSubTable", glCopyColorSubTable == NULL);
+  glewInfoFunc("glCopyColorTable", glCopyColorTable == NULL);
+  glewInfoFunc("glCopyConvolutionFilter1D", glCopyConvolutionFilter1D == NULL);
+  glewInfoFunc("glCopyConvolutionFilter2D", glCopyConvolutionFilter2D == NULL);
+  glewInfoFunc("glGetColorTable", glGetColorTable == NULL);
+  glewInfoFunc("glGetColorTableParameterfv", glGetColorTableParameterfv == NULL);
+  glewInfoFunc("glGetColorTableParameteriv", glGetColorTableParameteriv == NULL);
+  glewInfoFunc("glGetConvolutionFilter", glGetConvolutionFilter == NULL);
+  glewInfoFunc("glGetConvolutionParameterfv", glGetConvolutionParameterfv == NULL);
+  glewInfoFunc("glGetConvolutionParameteriv", glGetConvolutionParameteriv == NULL);
+  glewInfoFunc("glGetHistogram", glGetHistogram == NULL);
+  glewInfoFunc("glGetHistogramParameterfv", glGetHistogramParameterfv == NULL);
+  glewInfoFunc("glGetHistogramParameteriv", glGetHistogramParameteriv == NULL);
+  glewInfoFunc("glGetMinmax", glGetMinmax == NULL);
+  glewInfoFunc("glGetMinmaxParameterfv", glGetMinmaxParameterfv == NULL);
+  glewInfoFunc("glGetMinmaxParameteriv", glGetMinmaxParameteriv == NULL);
+  glewInfoFunc("glGetSeparableFilter", glGetSeparableFilter == NULL);
+  glewInfoFunc("glHistogram", glHistogram == NULL);
+  glewInfoFunc("glMinmax", glMinmax == NULL);
+  glewInfoFunc("glResetHistogram", glResetHistogram == NULL);
+  glewInfoFunc("glResetMinmax", glResetMinmax == NULL);
+  glewInfoFunc("glSeparableFilter2D", glSeparableFilter2D == NULL);
+}
+
+#endif /* GL_ARB_imaging */
+
+#ifdef GL_ARB_instanced_arrays
+
+static void _glewInfo_GL_ARB_instanced_arrays (void)
+{
+  glewPrintExt("GL_ARB_instanced_arrays", GLEW_ARB_instanced_arrays, glewIsSupported("GL_ARB_instanced_arrays"), glewGetExtension("GL_ARB_instanced_arrays"));
+
+  glewInfoFunc("glVertexAttribDivisorARB", glVertexAttribDivisorARB == NULL);
+}
+
+#endif /* GL_ARB_instanced_arrays */
+
+#ifdef GL_ARB_map_buffer_range
+
+static void _glewInfo_GL_ARB_map_buffer_range (void)
+{
+  glewPrintExt("GL_ARB_map_buffer_range", GLEW_ARB_map_buffer_range, glewIsSupported("GL_ARB_map_buffer_range"), glewGetExtension("GL_ARB_map_buffer_range"));
+
+  glewInfoFunc("glFlushMappedBufferRange", glFlushMappedBufferRange == NULL);
+  glewInfoFunc("glMapBufferRange", glMapBufferRange == NULL);
+}
+
+#endif /* GL_ARB_map_buffer_range */
+
+#ifdef GL_ARB_matrix_palette
+
+static void _glewInfo_GL_ARB_matrix_palette (void)
+{
+  glewPrintExt("GL_ARB_matrix_palette", GLEW_ARB_matrix_palette, glewIsSupported("GL_ARB_matrix_palette"), glewGetExtension("GL_ARB_matrix_palette"));
+
+  glewInfoFunc("glCurrentPaletteMatrixARB", glCurrentPaletteMatrixARB == NULL);
+  glewInfoFunc("glMatrixIndexPointerARB", glMatrixIndexPointerARB == NULL);
+  glewInfoFunc("glMatrixIndexubvARB", glMatrixIndexubvARB == NULL);
+  glewInfoFunc("glMatrixIndexuivARB", glMatrixIndexuivARB == NULL);
+  glewInfoFunc("glMatrixIndexusvARB", glMatrixIndexusvARB == NULL);
+}
+
+#endif /* GL_ARB_matrix_palette */
+
+#ifdef GL_ARB_multisample
+
+static void _glewInfo_GL_ARB_multisample (void)
+{
+  glewPrintExt("GL_ARB_multisample", GLEW_ARB_multisample, glewIsSupported("GL_ARB_multisample"), glewGetExtension("GL_ARB_multisample"));
+
+  glewInfoFunc("glSampleCoverageARB", glSampleCoverageARB == NULL);
+}
+
+#endif /* GL_ARB_multisample */
+
+#ifdef GL_ARB_multitexture
+
+static void _glewInfo_GL_ARB_multitexture (void)
+{
+  glewPrintExt("GL_ARB_multitexture", GLEW_ARB_multitexture, glewIsSupported("GL_ARB_multitexture"), glewGetExtension("GL_ARB_multitexture"));
+
+  glewInfoFunc("glActiveTextureARB", glActiveTextureARB == NULL);
+  glewInfoFunc("glClientActiveTextureARB", glClientActiveTextureARB == NULL);
+  glewInfoFunc("glMultiTexCoord1dARB", glMultiTexCoord1dARB == NULL);
+  glewInfoFunc("glMultiTexCoord1dvARB", glMultiTexCoord1dvARB == NULL);
+  glewInfoFunc("glMultiTexCoord1fARB", glMultiTexCoord1fARB == NULL);
+  glewInfoFunc("glMultiTexCoord1fvARB", glMultiTexCoord1fvARB == NULL);
+  glewInfoFunc("glMultiTexCoord1iARB", glMultiTexCoord1iARB == NULL);
+  glewInfoFunc("glMultiTexCoord1ivARB", glMultiTexCoord1ivARB == NULL);
+  glewInfoFunc("glMultiTexCoord1sARB", glMultiTexCoord1sARB == NULL);
+  glewInfoFunc("glMultiTexCoord1svARB", glMultiTexCoord1svARB == NULL);
+  glewInfoFunc("glMultiTexCoord2dARB", glMultiTexCoord2dARB == NULL);
+  glewInfoFunc("glMultiTexCoord2dvARB", glMultiTexCoord2dvARB == NULL);
+  glewInfoFunc("glMultiTexCoord2fARB", glMultiTexCoord2fARB == NULL);
+  glewInfoFunc("glMultiTexCoord2fvARB", glMultiTexCoord2fvARB == NULL);
+  glewInfoFunc("glMultiTexCoord2iARB", glMultiTexCoord2iARB == NULL);
+  glewInfoFunc("glMultiTexCoord2ivARB", glMultiTexCoord2ivARB == NULL);
+  glewInfoFunc("glMultiTexCoord2sARB", glMultiTexCoord2sARB == NULL);
+  glewInfoFunc("glMultiTexCoord2svARB", glMultiTexCoord2svARB == NULL);
+  glewInfoFunc("glMultiTexCoord3dARB", glMultiTexCoord3dARB == NULL);
+  glewInfoFunc("glMultiTexCoord3dvARB", glMultiTexCoord3dvARB == NULL);
+  glewInfoFunc("glMultiTexCoord3fARB", glMultiTexCoord3fARB == NULL);
+  glewInfoFunc("glMultiTexCoord3fvARB", glMultiTexCoord3fvARB == NULL);
+  glewInfoFunc("glMultiTexCoord3iARB", glMultiTexCoord3iARB == NULL);
+  glewInfoFunc("glMultiTexCoord3ivARB", glMultiTexCoord3ivARB == NULL);
+  glewInfoFunc("glMultiTexCoord3sARB", glMultiTexCoord3sARB == NULL);
+  glewInfoFunc("glMultiTexCoord3svARB", glMultiTexCoord3svARB == NULL);
+  glewInfoFunc("glMultiTexCoord4dARB", glMultiTexCoord4dARB == NULL);
+  glewInfoFunc("glMultiTexCoord4dvARB", glMultiTexCoord4dvARB == NULL);
+  glewInfoFunc("glMultiTexCoord4fARB", glMultiTexCoord4fARB == NULL);
+  glewInfoFunc("glMultiTexCoord4fvARB", glMultiTexCoord4fvARB == NULL);
+  glewInfoFunc("glMultiTexCoord4iARB", glMultiTexCoord4iARB == NULL);
+  glewInfoFunc("glMultiTexCoord4ivARB", glMultiTexCoord4ivARB == NULL);
+  glewInfoFunc("glMultiTexCoord4sARB", glMultiTexCoord4sARB == NULL);
+  glewInfoFunc("glMultiTexCoord4svARB", glMultiTexCoord4svARB == NULL);
+}
+
+#endif /* GL_ARB_multitexture */
+
+#ifdef GL_ARB_occlusion_query
+
+static void _glewInfo_GL_ARB_occlusion_query (void)
+{
+  glewPrintExt("GL_ARB_occlusion_query", GLEW_ARB_occlusion_query, glewIsSupported("GL_ARB_occlusion_query"), glewGetExtension("GL_ARB_occlusion_query"));
+
+  glewInfoFunc("glBeginQueryARB", glBeginQueryARB == NULL);
+  glewInfoFunc("glDeleteQueriesARB", glDeleteQueriesARB == NULL);
+  glewInfoFunc("glEndQueryARB", glEndQueryARB == NULL);
+  glewInfoFunc("glGenQueriesARB", glGenQueriesARB == NULL);
+  glewInfoFunc("glGetQueryObjectivARB", glGetQueryObjectivARB == NULL);
+  glewInfoFunc("glGetQueryObjectuivARB", glGetQueryObjectuivARB == NULL);
+  glewInfoFunc("glGetQueryivARB", glGetQueryivARB == NULL);
+  glewInfoFunc("glIsQueryARB", glIsQueryARB == NULL);
+}
+
+#endif /* GL_ARB_occlusion_query */
+
+#ifdef GL_ARB_pixel_buffer_object
+
+static void _glewInfo_GL_ARB_pixel_buffer_object (void)
+{
+  glewPrintExt("GL_ARB_pixel_buffer_object", GLEW_ARB_pixel_buffer_object, glewIsSupported("GL_ARB_pixel_buffer_object"), glewGetExtension("GL_ARB_pixel_buffer_object"));
+}
+
+#endif /* GL_ARB_pixel_buffer_object */
+
+#ifdef GL_ARB_point_parameters
+
+static void _glewInfo_GL_ARB_point_parameters (void)
+{
+  glewPrintExt("GL_ARB_point_parameters", GLEW_ARB_point_parameters, glewIsSupported("GL_ARB_point_parameters"), glewGetExtension("GL_ARB_point_parameters"));
+
+  glewInfoFunc("glPointParameterfARB", glPointParameterfARB == NULL);
+  glewInfoFunc("glPointParameterfvARB", glPointParameterfvARB == NULL);
+}
+
+#endif /* GL_ARB_point_parameters */
+
+#ifdef GL_ARB_point_sprite
+
+static void _glewInfo_GL_ARB_point_sprite (void)
+{
+  glewPrintExt("GL_ARB_point_sprite", GLEW_ARB_point_sprite, glewIsSupported("GL_ARB_point_sprite"), glewGetExtension("GL_ARB_point_sprite"));
+}
+
+#endif /* GL_ARB_point_sprite */
+
+#ifdef GL_ARB_shader_objects
+
+static void _glewInfo_GL_ARB_shader_objects (void)
+{
+  glewPrintExt("GL_ARB_shader_objects", GLEW_ARB_shader_objects, glewIsSupported("GL_ARB_shader_objects"), glewGetExtension("GL_ARB_shader_objects"));
+
+  glewInfoFunc("glAttachObjectARB", glAttachObjectARB == NULL);
+  glewInfoFunc("glCompileShaderARB", glCompileShaderARB == NULL);
+  glewInfoFunc("glCreateProgramObjectARB", glCreateProgramObjectARB == NULL);
+  glewInfoFunc("glCreateShaderObjectARB", glCreateShaderObjectARB == NULL);
+  glewInfoFunc("glDeleteObjectARB", glDeleteObjectARB == NULL);
+  glewInfoFunc("glDetachObjectARB", glDetachObjectARB == NULL);
+  glewInfoFunc("glGetActiveUniformARB", glGetActiveUniformARB == NULL);
+  glewInfoFunc("glGetAttachedObjectsARB", glGetAttachedObjectsARB == NULL);
+  glewInfoFunc("glGetHandleARB", glGetHandleARB == NULL);
+  glewInfoFunc("glGetInfoLogARB", glGetInfoLogARB == NULL);
+  glewInfoFunc("glGetObjectParameterfvARB", glGetObjectParameterfvARB == NULL);
+  glewInfoFunc("glGetObjectParameterivARB", glGetObjectParameterivARB == NULL);
+  glewInfoFunc("glGetShaderSourceARB", glGetShaderSourceARB == NULL);
+  glewInfoFunc("glGetUniformLocationARB", glGetUniformLocationARB == NULL);
+  glewInfoFunc("glGetUniformfvARB", glGetUniformfvARB == NULL);
+  glewInfoFunc("glGetUniformivARB", glGetUniformivARB == NULL);
+  glewInfoFunc("glLinkProgramARB", glLinkProgramARB == NULL);
+  glewInfoFunc("glShaderSourceARB", glShaderSourceARB == NULL);
+  glewInfoFunc("glUniform1fARB", glUniform1fARB == NULL);
+  glewInfoFunc("glUniform1fvARB", glUniform1fvARB == NULL);
+  glewInfoFunc("glUniform1iARB", glUniform1iARB == NULL);
+  glewInfoFunc("glUniform1ivARB", glUniform1ivARB == NULL);
+  glewInfoFunc("glUniform2fARB", glUniform2fARB == NULL);
+  glewInfoFunc("glUniform2fvARB", glUniform2fvARB == NULL);
+  glewInfoFunc("glUniform2iARB", glUniform2iARB == NULL);
+  glewInfoFunc("glUniform2ivARB", glUniform2ivARB == NULL);
+  glewInfoFunc("glUniform3fARB", glUniform3fARB == NULL);
+  glewInfoFunc("glUniform3fvARB", glUniform3fvARB == NULL);
+  glewInfoFunc("glUniform3iARB", glUniform3iARB == NULL);
+  glewInfoFunc("glUniform3ivARB", glUniform3ivARB == NULL);
+  glewInfoFunc("glUniform4fARB", glUniform4fARB == NULL);
+  glewInfoFunc("glUniform4fvARB", glUniform4fvARB == NULL);
+  glewInfoFunc("glUniform4iARB", glUniform4iARB == NULL);
+  glewInfoFunc("glUniform4ivARB", glUniform4ivARB == NULL);
+  glewInfoFunc("glUniformMatrix2fvARB", glUniformMatrix2fvARB == NULL);
+  glewInfoFunc("glUniformMatrix3fvARB", glUniformMatrix3fvARB == NULL);
+  glewInfoFunc("glUniformMatrix4fvARB", glUniformMatrix4fvARB == NULL);
+  glewInfoFunc("glUseProgramObjectARB", glUseProgramObjectARB == NULL);
+  glewInfoFunc("glValidateProgramARB", glValidateProgramARB == NULL);
+}
+
+#endif /* GL_ARB_shader_objects */
+
+#ifdef GL_ARB_shading_language_100
+
+static void _glewInfo_GL_ARB_shading_language_100 (void)
+{
+  glewPrintExt("GL_ARB_shading_language_100", GLEW_ARB_shading_language_100, glewIsSupported("GL_ARB_shading_language_100"), glewGetExtension("GL_ARB_shading_language_100"));
+}
+
+#endif /* GL_ARB_shading_language_100 */
+
+#ifdef GL_ARB_shadow
+
+static void _glewInfo_GL_ARB_shadow (void)
+{
+  glewPrintExt("GL_ARB_shadow", GLEW_ARB_shadow, glewIsSupported("GL_ARB_shadow"), glewGetExtension("GL_ARB_shadow"));
+}
+
+#endif /* GL_ARB_shadow */
+
+#ifdef GL_ARB_shadow_ambient
+
+static void _glewInfo_GL_ARB_shadow_ambient (void)
+{
+  glewPrintExt("GL_ARB_shadow_ambient", GLEW_ARB_shadow_ambient, glewIsSupported("GL_ARB_shadow_ambient"), glewGetExtension("GL_ARB_shadow_ambient"));
+}
+
+#endif /* GL_ARB_shadow_ambient */
+
+#ifdef GL_ARB_texture_border_clamp
+
+static void _glewInfo_GL_ARB_texture_border_clamp (void)
+{
+  glewPrintExt("GL_ARB_texture_border_clamp", GLEW_ARB_texture_border_clamp, glewIsSupported("GL_ARB_texture_border_clamp"), glewGetExtension("GL_ARB_texture_border_clamp"));
+}
+
+#endif /* GL_ARB_texture_border_clamp */
+
+#ifdef GL_ARB_texture_buffer_object
+
+static void _glewInfo_GL_ARB_texture_buffer_object (void)
+{
+  glewPrintExt("GL_ARB_texture_buffer_object", GLEW_ARB_texture_buffer_object, glewIsSupported("GL_ARB_texture_buffer_object"), glewGetExtension("GL_ARB_texture_buffer_object"));
+
+  glewInfoFunc("glTexBufferARB", glTexBufferARB == NULL);
+}
+
+#endif /* GL_ARB_texture_buffer_object */
+
+#ifdef GL_ARB_texture_compression
+
+static void _glewInfo_GL_ARB_texture_compression (void)
+{
+  glewPrintExt("GL_ARB_texture_compression", GLEW_ARB_texture_compression, glewIsSupported("GL_ARB_texture_compression"), glewGetExtension("GL_ARB_texture_compression"));
+
+  glewInfoFunc("glCompressedTexImage1DARB", glCompressedTexImage1DARB == NULL);
+  glewInfoFunc("glCompressedTexImage2DARB", glCompressedTexImage2DARB == NULL);
+  glewInfoFunc("glCompressedTexImage3DARB", glCompressedTexImage3DARB == NULL);
+  glewInfoFunc("glCompressedTexSubImage1DARB", glCompressedTexSubImage1DARB == NULL);
+  glewInfoFunc("glCompressedTexSubImage2DARB", glCompressedTexSubImage2DARB == NULL);
+  glewInfoFunc("glCompressedTexSubImage3DARB", glCompressedTexSubImage3DARB == NULL);
+  glewInfoFunc("glGetCompressedTexImageARB", glGetCompressedTexImageARB == NULL);
+}
+
+#endif /* GL_ARB_texture_compression */
+
+#ifdef GL_ARB_texture_compression_rgtc
+
+static void _glewInfo_GL_ARB_texture_compression_rgtc (void)
+{
+  glewPrintExt("GL_ARB_texture_compression_rgtc", GLEW_ARB_texture_compression_rgtc, glewIsSupported("GL_ARB_texture_compression_rgtc"), glewGetExtension("GL_ARB_texture_compression_rgtc"));
+}
+
+#endif /* GL_ARB_texture_compression_rgtc */
+
+#ifdef GL_ARB_texture_cube_map
+
+static void _glewInfo_GL_ARB_texture_cube_map (void)
+{
+  glewPrintExt("GL_ARB_texture_cube_map", GLEW_ARB_texture_cube_map, glewIsSupported("GL_ARB_texture_cube_map"), glewGetExtension("GL_ARB_texture_cube_map"));
+}
+
+#endif /* GL_ARB_texture_cube_map */
+
+#ifdef GL_ARB_texture_env_add
+
+static void _glewInfo_GL_ARB_texture_env_add (void)
+{
+  glewPrintExt("GL_ARB_texture_env_add", GLEW_ARB_texture_env_add, glewIsSupported("GL_ARB_texture_env_add"), glewGetExtension("GL_ARB_texture_env_add"));
+}
+
+#endif /* GL_ARB_texture_env_add */
+
+#ifdef GL_ARB_texture_env_combine
+
+static void _glewInfo_GL_ARB_texture_env_combine (void)
+{
+  glewPrintExt("GL_ARB_texture_env_combine", GLEW_ARB_texture_env_combine, glewIsSupported("GL_ARB_texture_env_combine"), glewGetExtension("GL_ARB_texture_env_combine"));
+}
+
+#endif /* GL_ARB_texture_env_combine */
+
+#ifdef GL_ARB_texture_env_crossbar
+
+static void _glewInfo_GL_ARB_texture_env_crossbar (void)
+{
+  glewPrintExt("GL_ARB_texture_env_crossbar", GLEW_ARB_texture_env_crossbar, glewIsSupported("GL_ARB_texture_env_crossbar"), glewGetExtension("GL_ARB_texture_env_crossbar"));
+}
+
+#endif /* GL_ARB_texture_env_crossbar */
+
+#ifdef GL_ARB_texture_env_dot3
+
+static void _glewInfo_GL_ARB_texture_env_dot3 (void)
+{
+  glewPrintExt("GL_ARB_texture_env_dot3", GLEW_ARB_texture_env_dot3, glewIsSupported("GL_ARB_texture_env_dot3"), glewGetExtension("GL_ARB_texture_env_dot3"));
+}
+
+#endif /* GL_ARB_texture_env_dot3 */
+
+#ifdef GL_ARB_texture_float
+
+static void _glewInfo_GL_ARB_texture_float (void)
+{
+  glewPrintExt("GL_ARB_texture_float", GLEW_ARB_texture_float, glewIsSupported("GL_ARB_texture_float"), glewGetExtension("GL_ARB_texture_float"));
+}
+
+#endif /* GL_ARB_texture_float */
+
+#ifdef GL_ARB_texture_mirrored_repeat
+
+static void _glewInfo_GL_ARB_texture_mirrored_repeat (void)
+{
+  glewPrintExt("GL_ARB_texture_mirrored_repeat", GLEW_ARB_texture_mirrored_repeat, glewIsSupported("GL_ARB_texture_mirrored_repeat"), glewGetExtension("GL_ARB_texture_mirrored_repeat"));
+}
+
+#endif /* GL_ARB_texture_mirrored_repeat */
+
+#ifdef GL_ARB_texture_non_power_of_two
+
+static void _glewInfo_GL_ARB_texture_non_power_of_two (void)
+{
+  glewPrintExt("GL_ARB_texture_non_power_of_two", GLEW_ARB_texture_non_power_of_two, glewIsSupported("GL_ARB_texture_non_power_of_two"), glewGetExtension("GL_ARB_texture_non_power_of_two"));
+}
+
+#endif /* GL_ARB_texture_non_power_of_two */
+
+#ifdef GL_ARB_texture_rectangle
+
+static void _glewInfo_GL_ARB_texture_rectangle (void)
+{
+  glewPrintExt("GL_ARB_texture_rectangle", GLEW_ARB_texture_rectangle, glewIsSupported("GL_ARB_texture_rectangle"), glewGetExtension("GL_ARB_texture_rectangle"));
+}
+
+#endif /* GL_ARB_texture_rectangle */
+
+#ifdef GL_ARB_texture_rg
+
+static void _glewInfo_GL_ARB_texture_rg (void)
+{
+  glewPrintExt("GL_ARB_texture_rg", GLEW_ARB_texture_rg, glewIsSupported("GL_ARB_texture_rg"), glewGetExtension("GL_ARB_texture_rg"));
+}
+
+#endif /* GL_ARB_texture_rg */
+
+#ifdef GL_ARB_transpose_matrix
+
+static void _glewInfo_GL_ARB_transpose_matrix (void)
+{
+  glewPrintExt("GL_ARB_transpose_matrix", GLEW_ARB_transpose_matrix, glewIsSupported("GL_ARB_transpose_matrix"), glewGetExtension("GL_ARB_transpose_matrix"));
+
+  glewInfoFunc("glLoadTransposeMatrixdARB", glLoadTransposeMatrixdARB == NULL);
+  glewInfoFunc("glLoadTransposeMatrixfARB", glLoadTransposeMatrixfARB == NULL);
+  glewInfoFunc("glMultTransposeMatrixdARB", glMultTransposeMatrixdARB == NULL);
+  glewInfoFunc("glMultTransposeMatrixfARB", glMultTransposeMatrixfARB == NULL);
+}
+
+#endif /* GL_ARB_transpose_matrix */
+
+#ifdef GL_ARB_vertex_array_object
+
+static void _glewInfo_GL_ARB_vertex_array_object (void)
+{
+  glewPrintExt("GL_ARB_vertex_array_object", GLEW_ARB_vertex_array_object, glewIsSupported("GL_ARB_vertex_array_object"), glewGetExtension("GL_ARB_vertex_array_object"));
+
+  glewInfoFunc("glBindVertexArray", glBindVertexArray == NULL);
+  glewInfoFunc("glDeleteVertexArrays", glDeleteVertexArrays == NULL);
+  glewInfoFunc("glGenVertexArrays", glGenVertexArrays == NULL);
+  glewInfoFunc("glIsVertexArray", glIsVertexArray == NULL);
+}
+
+#endif /* GL_ARB_vertex_array_object */
+
+#ifdef GL_ARB_vertex_blend
+
+static void _glewInfo_GL_ARB_vertex_blend (void)
+{
+  glewPrintExt("GL_ARB_vertex_blend", GLEW_ARB_vertex_blend, glewIsSupported("GL_ARB_vertex_blend"), glewGetExtension("GL_ARB_vertex_blend"));
+
+  glewInfoFunc("glVertexBlendARB", glVertexBlendARB == NULL);
+  glewInfoFunc("glWeightPointerARB", glWeightPointerARB == NULL);
+  glewInfoFunc("glWeightbvARB", glWeightbvARB == NULL);
+  glewInfoFunc("glWeightdvARB", glWeightdvARB == NULL);
+  glewInfoFunc("glWeightfvARB", glWeightfvARB == NULL);
+  glewInfoFunc("glWeightivARB", glWeightivARB == NULL);
+  glewInfoFunc("glWeightsvARB", glWeightsvARB == NULL);
+  glewInfoFunc("glWeightubvARB", glWeightubvARB == NULL);
+  glewInfoFunc("glWeightuivARB", glWeightuivARB == NULL);
+  glewInfoFunc("glWeightusvARB", glWeightusvARB == NULL);
+}
+
+#endif /* GL_ARB_vertex_blend */
+
+#ifdef GL_ARB_vertex_buffer_object
+
+static void _glewInfo_GL_ARB_vertex_buffer_object (void)
+{
+  glewPrintExt("GL_ARB_vertex_buffer_object", GLEW_ARB_vertex_buffer_object, glewIsSupported("GL_ARB_vertex_buffer_object"), glewGetExtension("GL_ARB_vertex_buffer_object"));
+
+  glewInfoFunc("glBindBufferARB", glBindBufferARB == NULL);
+  glewInfoFunc("glBufferDataARB", glBufferDataARB == NULL);
+  glewInfoFunc("glBufferSubDataARB", glBufferSubDataARB == NULL);
+  glewInfoFunc("glDeleteBuffersARB", glDeleteBuffersARB == NULL);
+  glewInfoFunc("glGenBuffersARB", glGenBuffersARB == NULL);
+  glewInfoFunc("glGetBufferParameterivARB", glGetBufferParameterivARB == NULL);
+  glewInfoFunc("glGetBufferPointervARB", glGetBufferPointervARB == NULL);
+  glewInfoFunc("glGetBufferSubDataARB", glGetBufferSubDataARB == NULL);
+  glewInfoFunc("glIsBufferARB", glIsBufferARB == NULL);
+  glewInfoFunc("glMapBufferARB", glMapBufferARB == NULL);
+  glewInfoFunc("glUnmapBufferARB", glUnmapBufferARB == NULL);
+}
+
+#endif /* GL_ARB_vertex_buffer_object */
+
+#ifdef GL_ARB_vertex_program
+
+static void _glewInfo_GL_ARB_vertex_program (void)
+{
+  glewPrintExt("GL_ARB_vertex_program", GLEW_ARB_vertex_program, glewIsSupported("GL_ARB_vertex_program"), glewGetExtension("GL_ARB_vertex_program"));
+
+  glewInfoFunc("glBindProgramARB", glBindProgramARB == NULL);
+  glewInfoFunc("glDeleteProgramsARB", glDeleteProgramsARB == NULL);
+  glewInfoFunc("glDisableVertexAttribArrayARB", glDisableVertexAttribArrayARB == NULL);
+  glewInfoFunc("glEnableVertexAttribArrayARB", glEnableVertexAttribArrayARB == NULL);
+  glewInfoFunc("glGenProgramsARB", glGenProgramsARB == NULL);
+  glewInfoFunc("glGetProgramEnvParameterdvARB", glGetProgramEnvParameterdvARB == NULL);
+  glewInfoFunc("glGetProgramEnvParameterfvARB", glGetProgramEnvParameterfvARB == NULL);
+  glewInfoFunc("glGetProgramLocalParameterdvARB", glGetProgramLocalParameterdvARB == NULL);
+  glewInfoFunc("glGetProgramLocalParameterfvARB", glGetProgramLocalParameterfvARB == NULL);
+  glewInfoFunc("glGetProgramStringARB", glGetProgramStringARB == NULL);
+  glewInfoFunc("glGetProgramivARB", glGetProgramivARB == NULL);
+  glewInfoFunc("glGetVertexAttribPointervARB", glGetVertexAttribPointervARB == NULL);
+  glewInfoFunc("glGetVertexAttribdvARB", glGetVertexAttribdvARB == NULL);
+  glewInfoFunc("glGetVertexAttribfvARB", glGetVertexAttribfvARB == NULL);
+  glewInfoFunc("glGetVertexAttribivARB", glGetVertexAttribivARB == NULL);
+  glewInfoFunc("glIsProgramARB", glIsProgramARB == NULL);
+  glewInfoFunc("glProgramEnvParameter4dARB", glProgramEnvParameter4dARB == NULL);
+  glewInfoFunc("glProgramEnvParameter4dvARB", glProgramEnvParameter4dvARB == NULL);
+  glewInfoFunc("glProgramEnvParameter4fARB", glProgramEnvParameter4fARB == NULL);
+  glewInfoFunc("glProgramEnvParameter4fvARB", glProgramEnvParameter4fvARB == NULL);
+  glewInfoFunc("glProgramLocalParameter4dARB", glProgramLocalParameter4dARB == NULL);
+  glewInfoFunc("glProgramLocalParameter4dvARB", glProgramLocalParameter4dvARB == NULL);
+  glewInfoFunc("glProgramLocalParameter4fARB", glProgramLocalParameter4fARB == NULL);
+  glewInfoFunc("glProgramLocalParameter4fvARB", glProgramLocalParameter4fvARB == NULL);
+  glewInfoFunc("glProgramStringARB", glProgramStringARB == NULL);
+  glewInfoFunc("glVertexAttrib1dARB", glVertexAttrib1dARB == NULL);
+  glewInfoFunc("glVertexAttrib1dvARB", glVertexAttrib1dvARB == NULL);
+  glewInfoFunc("glVertexAttrib1fARB", glVertexAttrib1fARB == NULL);
+  glewInfoFunc("glVertexAttrib1fvARB", glVertexAttrib1fvARB == NULL);
+  glewInfoFunc("glVertexAttrib1sARB", glVertexAttrib1sARB == NULL);
+  glewInfoFunc("glVertexAttrib1svARB", glVertexAttrib1svARB == NULL);
+  glewInfoFunc("glVertexAttrib2dARB", glVertexAttrib2dARB == NULL);
+  glewInfoFunc("glVertexAttrib2dvARB", glVertexAttrib2dvARB == NULL);
+  glewInfoFunc("glVertexAttrib2fARB", glVertexAttrib2fARB == NULL);
+  glewInfoFunc("glVertexAttrib2fvARB", glVertexAttrib2fvARB == NULL);
+  glewInfoFunc("glVertexAttrib2sARB", glVertexAttrib2sARB == NULL);
+  glewInfoFunc("glVertexAttrib2svARB", glVertexAttrib2svARB == NULL);
+  glewInfoFunc("glVertexAttrib3dARB", glVertexAttrib3dARB == NULL);
+  glewInfoFunc("glVertexAttrib3dvARB", glVertexAttrib3dvARB == NULL);
+  glewInfoFunc("glVertexAttrib3fARB", glVertexAttrib3fARB == NULL);
+  glewInfoFunc("glVertexAttrib3fvARB", glVertexAttrib3fvARB == NULL);
+  glewInfoFunc("glVertexAttrib3sARB", glVertexAttrib3sARB == NULL);
+  glewInfoFunc("glVertexAttrib3svARB", glVertexAttrib3svARB == NULL);
+  glewInfoFunc("glVertexAttrib4NbvARB", glVertexAttrib4NbvARB == NULL);
+  glewInfoFunc("glVertexAttrib4NivARB", glVertexAttrib4NivARB == NULL);
+  glewInfoFunc("glVertexAttrib4NsvARB", glVertexAttrib4NsvARB == NULL);
+  glewInfoFunc("glVertexAttrib4NubARB", glVertexAttrib4NubARB == NULL);
+  glewInfoFunc("glVertexAttrib4NubvARB", glVertexAttrib4NubvARB == NULL);
+  glewInfoFunc("glVertexAttrib4NuivARB", glVertexAttrib4NuivARB == NULL);
+  glewInfoFunc("glVertexAttrib4NusvARB", glVertexAttrib4NusvARB == NULL);
+  glewInfoFunc("glVertexAttrib4bvARB", glVertexAttrib4bvARB == NULL);
+  glewInfoFunc("glVertexAttrib4dARB", glVertexAttrib4dARB == NULL);
+  glewInfoFunc("glVertexAttrib4dvARB", glVertexAttrib4dvARB == NULL);
+  glewInfoFunc("glVertexAttrib4fARB", glVertexAttrib4fARB == NULL);
+  glewInfoFunc("glVertexAttrib4fvARB", glVertexAttrib4fvARB == NULL);
+  glewInfoFunc("glVertexAttrib4ivARB", glVertexAttrib4ivARB == NULL);
+  glewInfoFunc("glVertexAttrib4sARB", glVertexAttrib4sARB == NULL);
+  glewInfoFunc("glVertexAttrib4svARB", glVertexAttrib4svARB == NULL);
+  glewInfoFunc("glVertexAttrib4ubvARB", glVertexAttrib4ubvARB == NULL);
+  glewInfoFunc("glVertexAttrib4uivARB", glVertexAttrib4uivARB == NULL);
+  glewInfoFunc("glVertexAttrib4usvARB", glVertexAttrib4usvARB == NULL);
+  glewInfoFunc("glVertexAttribPointerARB", glVertexAttribPointerARB == NULL);
+}
+
+#endif /* GL_ARB_vertex_program */
+
+#ifdef GL_ARB_vertex_shader
+
+static void _glewInfo_GL_ARB_vertex_shader (void)
+{
+  glewPrintExt("GL_ARB_vertex_shader", GLEW_ARB_vertex_shader, glewIsSupported("GL_ARB_vertex_shader"), glewGetExtension("GL_ARB_vertex_shader"));
+
+  glewInfoFunc("glBindAttribLocationARB", glBindAttribLocationARB == NULL);
+  glewInfoFunc("glGetActiveAttribARB", glGetActiveAttribARB == NULL);
+  glewInfoFunc("glGetAttribLocationARB", glGetAttribLocationARB == NULL);
+}
+
+#endif /* GL_ARB_vertex_shader */
+
+#ifdef GL_ARB_window_pos
+
+static void _glewInfo_GL_ARB_window_pos (void)
+{
+  glewPrintExt("GL_ARB_window_pos", GLEW_ARB_window_pos, glewIsSupported("GL_ARB_window_pos"), glewGetExtension("GL_ARB_window_pos"));
+
+  glewInfoFunc("glWindowPos2dARB", glWindowPos2dARB == NULL);
+  glewInfoFunc("glWindowPos2dvARB", glWindowPos2dvARB == NULL);
+  glewInfoFunc("glWindowPos2fARB", glWindowPos2fARB == NULL);
+  glewInfoFunc("glWindowPos2fvARB", glWindowPos2fvARB == NULL);
+  glewInfoFunc("glWindowPos2iARB", glWindowPos2iARB == NULL);
+  glewInfoFunc("glWindowPos2ivARB", glWindowPos2ivARB == NULL);
+  glewInfoFunc("glWindowPos2sARB", glWindowPos2sARB == NULL);
+  glewInfoFunc("glWindowPos2svARB", glWindowPos2svARB == NULL);
+  glewInfoFunc("glWindowPos3dARB", glWindowPos3dARB == NULL);
+  glewInfoFunc("glWindowPos3dvARB", glWindowPos3dvARB == NULL);
+  glewInfoFunc("glWindowPos3fARB", glWindowPos3fARB == NULL);
+  glewInfoFunc("glWindowPos3fvARB", glWindowPos3fvARB == NULL);
+  glewInfoFunc("glWindowPos3iARB", glWindowPos3iARB == NULL);
+  glewInfoFunc("glWindowPos3ivARB", glWindowPos3ivARB == NULL);
+  glewInfoFunc("glWindowPos3sARB", glWindowPos3sARB == NULL);
+  glewInfoFunc("glWindowPos3svARB", glWindowPos3svARB == NULL);
+}
+
+#endif /* GL_ARB_window_pos */
+
+#ifdef GL_ATIX_point_sprites
+
+static void _glewInfo_GL_ATIX_point_sprites (void)
+{
+  glewPrintExt("GL_ATIX_point_sprites", GLEW_ATIX_point_sprites, glewIsSupported("GL_ATIX_point_sprites"), glewGetExtension("GL_ATIX_point_sprites"));
+}
+
+#endif /* GL_ATIX_point_sprites */
+
+#ifdef GL_ATIX_texture_env_combine3
+
+static void _glewInfo_GL_ATIX_texture_env_combine3 (void)
+{
+  glewPrintExt("GL_ATIX_texture_env_combine3", GLEW_ATIX_texture_env_combine3, glewIsSupported("GL_ATIX_texture_env_combine3"), glewGetExtension("GL_ATIX_texture_env_combine3"));
+}
+
+#endif /* GL_ATIX_texture_env_combine3 */
+
+#ifdef GL_ATIX_texture_env_route
+
+static void _glewInfo_GL_ATIX_texture_env_route (void)
+{
+  glewPrintExt("GL_ATIX_texture_env_route", GLEW_ATIX_texture_env_route, glewIsSupported("GL_ATIX_texture_env_route"), glewGetExtension("GL_ATIX_texture_env_route"));
+}
+
+#endif /* GL_ATIX_texture_env_route */
+
+#ifdef GL_ATIX_vertex_shader_output_point_size
+
+static void _glewInfo_GL_ATIX_vertex_shader_output_point_size (void)
+{
+  glewPrintExt("GL_ATIX_vertex_shader_output_point_size", GLEW_ATIX_vertex_shader_output_point_size, glewIsSupported("GL_ATIX_vertex_shader_output_point_size"), glewGetExtension("GL_ATIX_vertex_shader_output_point_size"));
+}
+
+#endif /* GL_ATIX_vertex_shader_output_point_size */
+
+#ifdef GL_ATI_draw_buffers
+
+static void _glewInfo_GL_ATI_draw_buffers (void)
+{
+  glewPrintExt("GL_ATI_draw_buffers", GLEW_ATI_draw_buffers, glewIsSupported("GL_ATI_draw_buffers"), glewGetExtension("GL_ATI_draw_buffers"));
+
+  glewInfoFunc("glDrawBuffersATI", glDrawBuffersATI == NULL);
+}
+
+#endif /* GL_ATI_draw_buffers */
+
+#ifdef GL_ATI_element_array
+
+static void _glewInfo_GL_ATI_element_array (void)
+{
+  glewPrintExt("GL_ATI_element_array", GLEW_ATI_element_array, glewIsSupported("GL_ATI_element_array"), glewGetExtension("GL_ATI_element_array"));
+
+  glewInfoFunc("glDrawElementArrayATI", glDrawElementArrayATI == NULL);
+  glewInfoFunc("glDrawRangeElementArrayATI", glDrawRangeElementArrayATI == NULL);
+  glewInfoFunc("glElementPointerATI", glElementPointerATI == NULL);
+}
+
+#endif /* GL_ATI_element_array */
+
+#ifdef GL_ATI_envmap_bumpmap
+
+static void _glewInfo_GL_ATI_envmap_bumpmap (void)
+{
+  glewPrintExt("GL_ATI_envmap_bumpmap", GLEW_ATI_envmap_bumpmap, glewIsSupported("GL_ATI_envmap_bumpmap"), glewGetExtension("GL_ATI_envmap_bumpmap"));
+
+  glewInfoFunc("glGetTexBumpParameterfvATI", glGetTexBumpParameterfvATI == NULL);
+  glewInfoFunc("glGetTexBumpParameterivATI", glGetTexBumpParameterivATI == NULL);
+  glewInfoFunc("glTexBumpParameterfvATI", glTexBumpParameterfvATI == NULL);
+  glewInfoFunc("glTexBumpParameterivATI", glTexBumpParameterivATI == NULL);
+}
+
+#endif /* GL_ATI_envmap_bumpmap */
+
+#ifdef GL_ATI_fragment_shader
+
+static void _glewInfo_GL_ATI_fragment_shader (void)
+{
+  glewPrintExt("GL_ATI_fragment_shader", GLEW_ATI_fragment_shader, glewIsSupported("GL_ATI_fragment_shader"), glewGetExtension("GL_ATI_fragment_shader"));
+
+  glewInfoFunc("glAlphaFragmentOp1ATI", glAlphaFragmentOp1ATI == NULL);
+  glewInfoFunc("glAlphaFragmentOp2ATI", glAlphaFragmentOp2ATI == NULL);
+  glewInfoFunc("glAlphaFragmentOp3ATI", glAlphaFragmentOp3ATI == NULL);
+  glewInfoFunc("glBeginFragmentShaderATI", glBeginFragmentShaderATI == NULL);
+  glewInfoFunc("glBindFragmentShaderATI", glBindFragmentShaderATI == NULL);
+  glewInfoFunc("glColorFragmentOp1ATI", glColorFragmentOp1ATI == NULL);
+  glewInfoFunc("glColorFragmentOp2ATI", glColorFragmentOp2ATI == NULL);
+  glewInfoFunc("glColorFragmentOp3ATI", glColorFragmentOp3ATI == NULL);
+  glewInfoFunc("glDeleteFragmentShaderATI", glDeleteFragmentShaderATI == NULL);
+  glewInfoFunc("glEndFragmentShaderATI", glEndFragmentShaderATI == NULL);
+  glewInfoFunc("glGenFragmentShadersATI", glGenFragmentShadersATI == NULL);
+  glewInfoFunc("glPassTexCoordATI", glPassTexCoordATI == NULL);
+  glewInfoFunc("glSampleMapATI", glSampleMapATI == NULL);
+  glewInfoFunc("glSetFragmentShaderConstantATI", glSetFragmentShaderConstantATI == NULL);
+}
+
+#endif /* GL_ATI_fragment_shader */
+
+#ifdef GL_ATI_map_object_buffer
+
+static void _glewInfo_GL_ATI_map_object_buffer (void)
+{
+  glewPrintExt("GL_ATI_map_object_buffer", GLEW_ATI_map_object_buffer, glewIsSupported("GL_ATI_map_object_buffer"), glewGetExtension("GL_ATI_map_object_buffer"));
+
+  glewInfoFunc("glMapObjectBufferATI", glMapObjectBufferATI == NULL);
+  glewInfoFunc("glUnmapObjectBufferATI", glUnmapObjectBufferATI == NULL);
+}
+
+#endif /* GL_ATI_map_object_buffer */
+
+#ifdef GL_ATI_pn_triangles
+
+static void _glewInfo_GL_ATI_pn_triangles (void)
+{
+  glewPrintExt("GL_ATI_pn_triangles", GLEW_ATI_pn_triangles, glewIsSupported("GL_ATI_pn_triangles"), glewGetExtension("GL_ATI_pn_triangles"));
+
+  glewInfoFunc("glPNTrianglesfATI", glPNTrianglesfATI == NULL);
+  glewInfoFunc("glPNTrianglesiATI", glPNTrianglesiATI == NULL);
+}
+
+#endif /* GL_ATI_pn_triangles */
+
+#ifdef GL_ATI_separate_stencil
+
+static void _glewInfo_GL_ATI_separate_stencil (void)
+{
+  glewPrintExt("GL_ATI_separate_stencil", GLEW_ATI_separate_stencil, glewIsSupported("GL_ATI_separate_stencil"), glewGetExtension("GL_ATI_separate_stencil"));
+
+  glewInfoFunc("glStencilFuncSeparateATI", glStencilFuncSeparateATI == NULL);
+  glewInfoFunc("glStencilOpSeparateATI", glStencilOpSeparateATI == NULL);
+}
+
+#endif /* GL_ATI_separate_stencil */
+
+#ifdef GL_ATI_shader_texture_lod
+
+static void _glewInfo_GL_ATI_shader_texture_lod (void)
+{
+  glewPrintExt("GL_ATI_shader_texture_lod", GLEW_ATI_shader_texture_lod, glewIsSupported("GL_ATI_shader_texture_lod"), glewGetExtension("GL_ATI_shader_texture_lod"));
+}
+
+#endif /* GL_ATI_shader_texture_lod */
+
+#ifdef GL_ATI_text_fragment_shader
+
+static void _glewInfo_GL_ATI_text_fragment_shader (void)
+{
+  glewPrintExt("GL_ATI_text_fragment_shader", GLEW_ATI_text_fragment_shader, glewIsSupported("GL_ATI_text_fragment_shader"), glewGetExtension("GL_ATI_text_fragment_shader"));
+}
+
+#endif /* GL_ATI_text_fragment_shader */
+
+#ifdef GL_ATI_texture_compression_3dc
+
+static void _glewInfo_GL_ATI_texture_compression_3dc (void)
+{
+  glewPrintExt("GL_ATI_texture_compression_3dc", GLEW_ATI_texture_compression_3dc, glewIsSupported("GL_ATI_texture_compression_3dc"), glewGetExtension("GL_ATI_texture_compression_3dc"));
+}
+
+#endif /* GL_ATI_texture_compression_3dc */
+
+#ifdef GL_ATI_texture_env_combine3
+
+static void _glewInfo_GL_ATI_texture_env_combine3 (void)
+{
+  glewPrintExt("GL_ATI_texture_env_combine3", GLEW_ATI_texture_env_combine3, glewIsSupported("GL_ATI_texture_env_combine3"), glewGetExtension("GL_ATI_texture_env_combine3"));
+}
+
+#endif /* GL_ATI_texture_env_combine3 */
+
+#ifdef GL_ATI_texture_float
+
+static void _glewInfo_GL_ATI_texture_float (void)
+{
+  glewPrintExt("GL_ATI_texture_float", GLEW_ATI_texture_float, glewIsSupported("GL_ATI_texture_float"), glewGetExtension("GL_ATI_texture_float"));
+}
+
+#endif /* GL_ATI_texture_float */
+
+#ifdef GL_ATI_texture_mirror_once
+
+static void _glewInfo_GL_ATI_texture_mirror_once (void)
+{
+  glewPrintExt("GL_ATI_texture_mirror_once", GLEW_ATI_texture_mirror_once, glewIsSupported("GL_ATI_texture_mirror_once"), glewGetExtension("GL_ATI_texture_mirror_once"));
+}
+
+#endif /* GL_ATI_texture_mirror_once */
+
+#ifdef GL_ATI_vertex_array_object
+
+static void _glewInfo_GL_ATI_vertex_array_object (void)
+{
+  glewPrintExt("GL_ATI_vertex_array_object", GLEW_ATI_vertex_array_object, glewIsSupported("GL_ATI_vertex_array_object"), glewGetExtension("GL_ATI_vertex_array_object"));
+
+  glewInfoFunc("glArrayObjectATI", glArrayObjectATI == NULL);
+  glewInfoFunc("glFreeObjectBufferATI", glFreeObjectBufferATI == NULL);
+  glewInfoFunc("glGetArrayObjectfvATI", glGetArrayObjectfvATI == NULL);
+  glewInfoFunc("glGetArrayObjectivATI", glGetArrayObjectivATI == NULL);
+  glewInfoFunc("glGetObjectBufferfvATI", glGetObjectBufferfvATI == NULL);
+  glewInfoFunc("glGetObjectBufferivATI", glGetObjectBufferivATI == NULL);
+  glewInfoFunc("glGetVariantArrayObjectfvATI", glGetVariantArrayObjectfvATI == NULL);
+  glewInfoFunc("glGetVariantArrayObjectivATI", glGetVariantArrayObjectivATI == NULL);
+  glewInfoFunc("glIsObjectBufferATI", glIsObjectBufferATI == NULL);
+  glewInfoFunc("glNewObjectBufferATI", glNewObjectBufferATI == NULL);
+  glewInfoFunc("glUpdateObjectBufferATI", glUpdateObjectBufferATI == NULL);
+  glewInfoFunc("glVariantArrayObjectATI", glVariantArrayObjectATI == NULL);
+}
+
+#endif /* GL_ATI_vertex_array_object */
+
+#ifdef GL_ATI_vertex_attrib_array_object
+
+static void _glewInfo_GL_ATI_vertex_attrib_array_object (void)
+{
+  glewPrintExt("GL_ATI_vertex_attrib_array_object", GLEW_ATI_vertex_attrib_array_object, glewIsSupported("GL_ATI_vertex_attrib_array_object"), glewGetExtension("GL_ATI_vertex_attrib_array_object"));
+
+  glewInfoFunc("glGetVertexAttribArrayObjectfvATI", glGetVertexAttribArrayObjectfvATI == NULL);
+  glewInfoFunc("glGetVertexAttribArrayObjectivATI", glGetVertexAttribArrayObjectivATI == NULL);
+  glewInfoFunc("glVertexAttribArrayObjectATI", glVertexAttribArrayObjectATI == NULL);
+}
+
+#endif /* GL_ATI_vertex_attrib_array_object */
+
+#ifdef GL_ATI_vertex_streams
+
+static void _glewInfo_GL_ATI_vertex_streams (void)
+{
+  glewPrintExt("GL_ATI_vertex_streams", GLEW_ATI_vertex_streams, glewIsSupported("GL_ATI_vertex_streams"), glewGetExtension("GL_ATI_vertex_streams"));
+
+  glewInfoFunc("glClientActiveVertexStreamATI", glClientActiveVertexStreamATI == NULL);
+  glewInfoFunc("glNormalStream3bATI", glNormalStream3bATI == NULL);
+  glewInfoFunc("glNormalStream3bvATI", glNormalStream3bvATI == NULL);
+  glewInfoFunc("glNormalStream3dATI", glNormalStream3dATI == NULL);
+  glewInfoFunc("glNormalStream3dvATI", glNormalStream3dvATI == NULL);
+  glewInfoFunc("glNormalStream3fATI", glNormalStream3fATI == NULL);
+  glewInfoFunc("glNormalStream3fvATI", glNormalStream3fvATI == NULL);
+  glewInfoFunc("glNormalStream3iATI", glNormalStream3iATI == NULL);
+  glewInfoFunc("glNormalStream3ivATI", glNormalStream3ivATI == NULL);
+  glewInfoFunc("glNormalStream3sATI", glNormalStream3sATI == NULL);
+  glewInfoFunc("glNormalStream3svATI", glNormalStream3svATI == NULL);
+  glewInfoFunc("glVertexBlendEnvfATI", glVertexBlendEnvfATI == NULL);
+  glewInfoFunc("glVertexBlendEnviATI", glVertexBlendEnviATI == NULL);
+  glewInfoFunc("glVertexStream2dATI", glVertexStream2dATI == NULL);
+  glewInfoFunc("glVertexStream2dvATI", glVertexStream2dvATI == NULL);
+  glewInfoFunc("glVertexStream2fATI", glVertexStream2fATI == NULL);
+  glewInfoFunc("glVertexStream2fvATI", glVertexStream2fvATI == NULL);
+  glewInfoFunc("glVertexStream2iATI", glVertexStream2iATI == NULL);
+  glewInfoFunc("glVertexStream2ivATI", glVertexStream2ivATI == NULL);
+  glewInfoFunc("glVertexStream2sATI", glVertexStream2sATI == NULL);
+  glewInfoFunc("glVertexStream2svATI", glVertexStream2svATI == NULL);
+  glewInfoFunc("glVertexStream3dATI", glVertexStream3dATI == NULL);
+  glewInfoFunc("glVertexStream3dvATI", glVertexStream3dvATI == NULL);
+  glewInfoFunc("glVertexStream3fATI", glVertexStream3fATI == NULL);
+  glewInfoFunc("glVertexStream3fvATI", glVertexStream3fvATI == NULL);
+  glewInfoFunc("glVertexStream3iATI", glVertexStream3iATI == NULL);
+  glewInfoFunc("glVertexStream3ivATI", glVertexStream3ivATI == NULL);
+  glewInfoFunc("glVertexStream3sATI", glVertexStream3sATI == NULL);
+  glewInfoFunc("glVertexStream3svATI", glVertexStream3svATI == NULL);
+  glewInfoFunc("glVertexStream4dATI", glVertexStream4dATI == NULL);
+  glewInfoFunc("glVertexStream4dvATI", glVertexStream4dvATI == NULL);
+  glewInfoFunc("glVertexStream4fATI", glVertexStream4fATI == NULL);
+  glewInfoFunc("glVertexStream4fvATI", glVertexStream4fvATI == NULL);
+  glewInfoFunc("glVertexStream4iATI", glVertexStream4iATI == NULL);
+  glewInfoFunc("glVertexStream4ivATI", glVertexStream4ivATI == NULL);
+  glewInfoFunc("glVertexStream4sATI", glVertexStream4sATI == NULL);
+  glewInfoFunc("glVertexStream4svATI", glVertexStream4svATI == NULL);
+}
+
+#endif /* GL_ATI_vertex_streams */
+
+#ifdef GL_EXT_422_pixels
+
+static void _glewInfo_GL_EXT_422_pixels (void)
+{
+  glewPrintExt("GL_EXT_422_pixels", GLEW_EXT_422_pixels, glewIsSupported("GL_EXT_422_pixels"), glewGetExtension("GL_EXT_422_pixels"));
+}
+
+#endif /* GL_EXT_422_pixels */
+
+#ifdef GL_EXT_Cg_shader
+
+static void _glewInfo_GL_EXT_Cg_shader (void)
+{
+  glewPrintExt("GL_EXT_Cg_shader", GLEW_EXT_Cg_shader, glewIsSupported("GL_EXT_Cg_shader"), glewGetExtension("GL_EXT_Cg_shader"));
+}
+
+#endif /* GL_EXT_Cg_shader */
+
+#ifdef GL_EXT_abgr
+
+static void _glewInfo_GL_EXT_abgr (void)
+{
+  glewPrintExt("GL_EXT_abgr", GLEW_EXT_abgr, glewIsSupported("GL_EXT_abgr"), glewGetExtension("GL_EXT_abgr"));
+}
+
+#endif /* GL_EXT_abgr */
+
+#ifdef GL_EXT_bgra
+
+static void _glewInfo_GL_EXT_bgra (void)
+{
+  glewPrintExt("GL_EXT_bgra", GLEW_EXT_bgra, glewIsSupported("GL_EXT_bgra"), glewGetExtension("GL_EXT_bgra"));
+}
+
+#endif /* GL_EXT_bgra */
+
+#ifdef GL_EXT_bindable_uniform
+
+static void _glewInfo_GL_EXT_bindable_uniform (void)
+{
+  glewPrintExt("GL_EXT_bindable_uniform", GLEW_EXT_bindable_uniform, glewIsSupported("GL_EXT_bindable_uniform"), glewGetExtension("GL_EXT_bindable_uniform"));
+
+  glewInfoFunc("glGetUniformBufferSizeEXT", glGetUniformBufferSizeEXT == NULL);
+  glewInfoFunc("glGetUniformOffsetEXT", glGetUniformOffsetEXT == NULL);
+  glewInfoFunc("glUniformBufferEXT", glUniformBufferEXT == NULL);
+}
+
+#endif /* GL_EXT_bindable_uniform */
+
+#ifdef GL_EXT_blend_color
+
+static void _glewInfo_GL_EXT_blend_color (void)
+{
+  glewPrintExt("GL_EXT_blend_color", GLEW_EXT_blend_color, glewIsSupported("GL_EXT_blend_color"), glewGetExtension("GL_EXT_blend_color"));
+
+  glewInfoFunc("glBlendColorEXT", glBlendColorEXT == NULL);
+}
+
+#endif /* GL_EXT_blend_color */
+
+#ifdef GL_EXT_blend_equation_separate
+
+static void _glewInfo_GL_EXT_blend_equation_separate (void)
+{
+  glewPrintExt("GL_EXT_blend_equation_separate", GLEW_EXT_blend_equation_separate, glewIsSupported("GL_EXT_blend_equation_separate"), glewGetExtension("GL_EXT_blend_equation_separate"));
+
+  glewInfoFunc("glBlendEquationSeparateEXT", glBlendEquationSeparateEXT == NULL);
+}
+
+#endif /* GL_EXT_blend_equation_separate */
+
+#ifdef GL_EXT_blend_func_separate
+
+static void _glewInfo_GL_EXT_blend_func_separate (void)
+{
+  glewPrintExt("GL_EXT_blend_func_separate", GLEW_EXT_blend_func_separate, glewIsSupported("GL_EXT_blend_func_separate"), glewGetExtension("GL_EXT_blend_func_separate"));
+
+  glewInfoFunc("glBlendFuncSeparateEXT", glBlendFuncSeparateEXT == NULL);
+}
+
+#endif /* GL_EXT_blend_func_separate */
+
+#ifdef GL_EXT_blend_logic_op
+
+static void _glewInfo_GL_EXT_blend_logic_op (void)
+{
+  glewPrintExt("GL_EXT_blend_logic_op", GLEW_EXT_blend_logic_op, glewIsSupported("GL_EXT_blend_logic_op"), glewGetExtension("GL_EXT_blend_logic_op"));
+}
+
+#endif /* GL_EXT_blend_logic_op */
+
+#ifdef GL_EXT_blend_minmax
+
+static void _glewInfo_GL_EXT_blend_minmax (void)
+{
+  glewPrintExt("GL_EXT_blend_minmax", GLEW_EXT_blend_minmax, glewIsSupported("GL_EXT_blend_minmax"), glewGetExtension("GL_EXT_blend_minmax"));
+
+  glewInfoFunc("glBlendEquationEXT", glBlendEquationEXT == NULL);
+}
+
+#endif /* GL_EXT_blend_minmax */
+
+#ifdef GL_EXT_blend_subtract
+
+static void _glewInfo_GL_EXT_blend_subtract (void)
+{
+  glewPrintExt("GL_EXT_blend_subtract", GLEW_EXT_blend_subtract, glewIsSupported("GL_EXT_blend_subtract"), glewGetExtension("GL_EXT_blend_subtract"));
+}
+
+#endif /* GL_EXT_blend_subtract */
+
+#ifdef GL_EXT_clip_volume_hint
+
+static void _glewInfo_GL_EXT_clip_volume_hint (void)
+{
+  glewPrintExt("GL_EXT_clip_volume_hint", GLEW_EXT_clip_volume_hint, glewIsSupported("GL_EXT_clip_volume_hint"), glewGetExtension("GL_EXT_clip_volume_hint"));
+}
+
+#endif /* GL_EXT_clip_volume_hint */
+
+#ifdef GL_EXT_cmyka
+
+static void _glewInfo_GL_EXT_cmyka (void)
+{
+  glewPrintExt("GL_EXT_cmyka", GLEW_EXT_cmyka, glewIsSupported("GL_EXT_cmyka"), glewGetExtension("GL_EXT_cmyka"));
+}
+
+#endif /* GL_EXT_cmyka */
+
+#ifdef GL_EXT_color_subtable
+
+static void _glewInfo_GL_EXT_color_subtable (void)
+{
+  glewPrintExt("GL_EXT_color_subtable", GLEW_EXT_color_subtable, glewIsSupported("GL_EXT_color_subtable"), glewGetExtension("GL_EXT_color_subtable"));
+
+  glewInfoFunc("glColorSubTableEXT", glColorSubTableEXT == NULL);
+  glewInfoFunc("glCopyColorSubTableEXT", glCopyColorSubTableEXT == NULL);
+}
+
+#endif /* GL_EXT_color_subtable */
+
+#ifdef GL_EXT_compiled_vertex_array
+
+static void _glewInfo_GL_EXT_compiled_vertex_array (void)
+{
+  glewPrintExt("GL_EXT_compiled_vertex_array", GLEW_EXT_compiled_vertex_array, glewIsSupported("GL_EXT_compiled_vertex_array"), glewGetExtension("GL_EXT_compiled_vertex_array"));
+
+  glewInfoFunc("glLockArraysEXT", glLockArraysEXT == NULL);
+  glewInfoFunc("glUnlockArraysEXT", glUnlockArraysEXT == NULL);
+}
+
+#endif /* GL_EXT_compiled_vertex_array */
+
+#ifdef GL_EXT_convolution
+
+static void _glewInfo_GL_EXT_convolution (void)
+{
+  glewPrintExt("GL_EXT_convolution", GLEW_EXT_convolution, glewIsSupported("GL_EXT_convolution"), glewGetExtension("GL_EXT_convolution"));
+
+  glewInfoFunc("glConvolutionFilter1DEXT", glConvolutionFilter1DEXT == NULL);
+  glewInfoFunc("glConvolutionFilter2DEXT", glConvolutionFilter2DEXT == NULL);
+  glewInfoFunc("glConvolutionParameterfEXT", glConvolutionParameterfEXT == NULL);
+  glewInfoFunc("glConvolutionParameterfvEXT", glConvolutionParameterfvEXT == NULL);
+  glewInfoFunc("glConvolutionParameteriEXT", glConvolutionParameteriEXT == NULL);
+  glewInfoFunc("glConvolutionParameterivEXT", glConvolutionParameterivEXT == NULL);
+  glewInfoFunc("glCopyConvolutionFilter1DEXT", glCopyConvolutionFilter1DEXT == NULL);
+  glewInfoFunc("glCopyConvolutionFilter2DEXT", glCopyConvolutionFilter2DEXT == NULL);
+  glewInfoFunc("glGetConvolutionFilterEXT", glGetConvolutionFilterEXT == NULL);
+  glewInfoFunc("glGetConvolutionParameterfvEXT", glGetConvolutionParameterfvEXT == NULL);
+  glewInfoFunc("glGetConvolutionParameterivEXT", glGetConvolutionParameterivEXT == NULL);
+  glewInfoFunc("glGetSeparableFilterEXT", glGetSeparableFilterEXT == NULL);
+  glewInfoFunc("glSeparableFilter2DEXT", glSeparableFilter2DEXT == NULL);
+}
+
+#endif /* GL_EXT_convolution */
+
+#ifdef GL_EXT_coordinate_frame
+
+static void _glewInfo_GL_EXT_coordinate_frame (void)
+{
+  glewPrintExt("GL_EXT_coordinate_frame", GLEW_EXT_coordinate_frame, glewIsSupported("GL_EXT_coordinate_frame"), glewGetExtension("GL_EXT_coordinate_frame"));
+
+  glewInfoFunc("glBinormalPointerEXT", glBinormalPointerEXT == NULL);
+  glewInfoFunc("glTangentPointerEXT", glTangentPointerEXT == NULL);
+}
+
+#endif /* GL_EXT_coordinate_frame */
+
+#ifdef GL_EXT_copy_texture
+
+static void _glewInfo_GL_EXT_copy_texture (void)
+{
+  glewPrintExt("GL_EXT_copy_texture", GLEW_EXT_copy_texture, glewIsSupported("GL_EXT_copy_texture"), glewGetExtension("GL_EXT_copy_texture"));
+
+  glewInfoFunc("glCopyTexImage1DEXT", glCopyTexImage1DEXT == NULL);
+  glewInfoFunc("glCopyTexImage2DEXT", glCopyTexImage2DEXT == NULL);
+  glewInfoFunc("glCopyTexSubImage1DEXT", glCopyTexSubImage1DEXT == NULL);
+  glewInfoFunc("glCopyTexSubImage2DEXT", glCopyTexSubImage2DEXT == NULL);
+  glewInfoFunc("glCopyTexSubImage3DEXT", glCopyTexSubImage3DEXT == NULL);
+}
+
+#endif /* GL_EXT_copy_texture */
+
+#ifdef GL_EXT_cull_vertex
+
+static void _glewInfo_GL_EXT_cull_vertex (void)
+{
+  glewPrintExt("GL_EXT_cull_vertex", GLEW_EXT_cull_vertex, glewIsSupported("GL_EXT_cull_vertex"), glewGetExtension("GL_EXT_cull_vertex"));
+
+  glewInfoFunc("glCullParameterdvEXT", glCullParameterdvEXT == NULL);
+  glewInfoFunc("glCullParameterfvEXT", glCullParameterfvEXT == NULL);
+}
+
+#endif /* GL_EXT_cull_vertex */
+
+#ifdef GL_EXT_depth_bounds_test
+
+static void _glewInfo_GL_EXT_depth_bounds_test (void)
+{
+  glewPrintExt("GL_EXT_depth_bounds_test", GLEW_EXT_depth_bounds_test, glewIsSupported("GL_EXT_depth_bounds_test"), glewGetExtension("GL_EXT_depth_bounds_test"));
+
+  glewInfoFunc("glDepthBoundsEXT", glDepthBoundsEXT == NULL);
+}
+
+#endif /* GL_EXT_depth_bounds_test */
+
+#ifdef GL_EXT_direct_state_access
+
+static void _glewInfo_GL_EXT_direct_state_access (void)
+{
+  glewPrintExt("GL_EXT_direct_state_access", GLEW_EXT_direct_state_access, glewIsSupported("GL_EXT_direct_state_access"), glewGetExtension("GL_EXT_direct_state_access"));
+
+  glewInfoFunc("glBindMultiTextureEXT", glBindMultiTextureEXT == NULL);
+  glewInfoFunc("glCheckNamedFramebufferStatusEXT", glCheckNamedFramebufferStatusEXT == NULL);
+  glewInfoFunc("glClientAttribDefaultEXT", glClientAttribDefaultEXT == NULL);
+  glewInfoFunc("glCompressedMultiTexImage1DEXT", glCompressedMultiTexImage1DEXT == NULL);
+  glewInfoFunc("glCompressedMultiTexImage2DEXT", glCompressedMultiTexImage2DEXT == NULL);
+  glewInfoFunc("glCompressedMultiTexImage3DEXT", glCompressedMultiTexImage3DEXT == NULL);
+  glewInfoFunc("glCompressedMultiTexSubImage1DEXT", glCompressedMultiTexSubImage1DEXT == NULL);
+  glewInfoFunc("glCompressedMultiTexSubImage2DEXT", glCompressedMultiTexSubImage2DEXT == NULL);
+  glewInfoFunc("glCompressedMultiTexSubImage3DEXT", glCompressedMultiTexSubImage3DEXT == NULL);
+  glewInfoFunc("glCompressedTextureImage1DEXT", glCompressedTextureImage1DEXT == NULL);
+  glewInfoFunc("glCompressedTextureImage2DEXT", glCompressedTextureImage2DEXT == NULL);
+  glewInfoFunc("glCompressedTextureImage3DEXT", glCompressedTextureImage3DEXT == NULL);
+  glewInfoFunc("glCompressedTextureSubImage1DEXT", glCompressedTextureSubImage1DEXT == NULL);
+  glewInfoFunc("glCompressedTextureSubImage2DEXT", glCompressedTextureSubImage2DEXT == NULL);
+  glewInfoFunc("glCompressedTextureSubImage3DEXT", glCompressedTextureSubImage3DEXT == NULL);
+  glewInfoFunc("glCopyMultiTexImage1DEXT", glCopyMultiTexImage1DEXT == NULL);
+  glewInfoFunc("glCopyMultiTexImage2DEXT", glCopyMultiTexImage2DEXT == NULL);
+  glewInfoFunc("glCopyMultiTexSubImage1DEXT", glCopyMultiTexSubImage1DEXT == NULL);
+  glewInfoFunc("glCopyMultiTexSubImage2DEXT", glCopyMultiTexSubImage2DEXT == NULL);
+  glewInfoFunc("glCopyMultiTexSubImage3DEXT", glCopyMultiTexSubImage3DEXT == NULL);
+  glewInfoFunc("glCopyTextureImage1DEXT", glCopyTextureImage1DEXT == NULL);
+  glewInfoFunc("glCopyTextureImage2DEXT", glCopyTextureImage2DEXT == NULL);
+  glewInfoFunc("glCopyTextureSubImage1DEXT", glCopyTextureSubImage1DEXT == NULL);
+  glewInfoFunc("glCopyTextureSubImage2DEXT", glCopyTextureSubImage2DEXT == NULL);
+  glewInfoFunc("glCopyTextureSubImage3DEXT", glCopyTextureSubImage3DEXT == NULL);
+  glewInfoFunc("glDisableClientStateIndexedEXT", glDisableClientStateIndexedEXT == NULL);
+  glewInfoFunc("glEnableClientStateIndexedEXT", glEnableClientStateIndexedEXT == NULL);
+  glewInfoFunc("glFramebufferDrawBufferEXT", glFramebufferDrawBufferEXT == NULL);
+  glewInfoFunc("glFramebufferDrawBuffersEXT", glFramebufferDrawBuffersEXT == NULL);
+  glewInfoFunc("glFramebufferReadBufferEXT", glFramebufferReadBufferEXT == NULL);
+  glewInfoFunc("glGenerateMultiTexMipmapEXT", glGenerateMultiTexMipmapEXT == NULL);
+  glewInfoFunc("glGenerateTextureMipmapEXT", glGenerateTextureMipmapEXT == NULL);
+  glewInfoFunc("glGetCompressedMultiTexImageEXT", glGetCompressedMultiTexImageEXT == NULL);
+  glewInfoFunc("glGetCompressedTextureImageEXT", glGetCompressedTextureImageEXT == NULL);
+  glewInfoFunc("glGetDoubleIndexedvEXT", glGetDoubleIndexedvEXT == NULL);
+  glewInfoFunc("glGetFloatIndexedvEXT", glGetFloatIndexedvEXT == NULL);
+  glewInfoFunc("glGetFramebufferParameterivEXT", glGetFramebufferParameterivEXT == NULL);
+  glewInfoFunc("glGetMultiTexEnvfvEXT", glGetMultiTexEnvfvEXT == NULL);
+  glewInfoFunc("glGetMultiTexEnvivEXT", glGetMultiTexEnvivEXT == NULL);
+  glewInfoFunc("glGetMultiTexGendvEXT", glGetMultiTexGendvEXT == NULL);
+  glewInfoFunc("glGetMultiTexGenfvEXT", glGetMultiTexGenfvEXT == NULL);
+  glewInfoFunc("glGetMultiTexGenivEXT", glGetMultiTexGenivEXT == NULL);
+  glewInfoFunc("glGetMultiTexImageEXT", glGetMultiTexImageEXT == NULL);
+  glewInfoFunc("glGetMultiTexLevelParameterfvEXT", glGetMultiTexLevelParameterfvEXT == NULL);
+  glewInfoFunc("glGetMultiTexLevelParameterivEXT", glGetMultiTexLevelParameterivEXT == NULL);
+  glewInfoFunc("glGetMultiTexParameterIivEXT", glGetMultiTexParameterIivEXT == NULL);
+  glewInfoFunc("glGetMultiTexParameterIuivEXT", glGetMultiTexParameterIuivEXT == NULL);
+  glewInfoFunc("glGetMultiTexParameterfvEXT", glGetMultiTexParameterfvEXT == NULL);
+  glewInfoFunc("glGetMultiTexParameterivEXT", glGetMultiTexParameterivEXT == NULL);
+  glewInfoFunc("glGetNamedBufferParameterivEXT", glGetNamedBufferParameterivEXT == NULL);
+  glewInfoFunc("glGetNamedBufferPointervEXT", glGetNamedBufferPointervEXT == NULL);
+  glewInfoFunc("glGetNamedBufferSubDataEXT", glGetNamedBufferSubDataEXT == NULL);
+  glewInfoFunc("glGetNamedFramebufferAttachmentParameterivEXT", glGetNamedFramebufferAttachmentParameterivEXT == NULL);
+  glewInfoFunc("glGetNamedProgramLocalParameterIivEXT", glGetNamedProgramLocalParameterIivEXT == NULL);
+  glewInfoFunc("glGetNamedProgramLocalParameterIuivEXT", glGetNamedProgramLocalParameterIuivEXT == NULL);
+  glewInfoFunc("glGetNamedProgramLocalParameterdvEXT", glGetNamedProgramLocalParameterdvEXT == NULL);
+  glewInfoFunc("glGetNamedProgramLocalParameterfvEXT", glGetNamedProgramLocalParameterfvEXT == NULL);
+  glewInfoFunc("glGetNamedProgramStringEXT", glGetNamedProgramStringEXT == NULL);
+  glewInfoFunc("glGetNamedProgramivEXT", glGetNamedProgramivEXT == NULL);
+  glewInfoFunc("glGetNamedRenderbufferParameterivEXT", glGetNamedRenderbufferParameterivEXT == NULL);
+  glewInfoFunc("glGetPointerIndexedvEXT", glGetPointerIndexedvEXT == NULL);
+  glewInfoFunc("glGetTextureImageEXT", glGetTextureImageEXT == NULL);
+  glewInfoFunc("glGetTextureLevelParameterfvEXT", glGetTextureLevelParameterfvEXT == NULL);
+  glewInfoFunc("glGetTextureLevelParameterivEXT", glGetTextureLevelParameterivEXT == NULL);
+  glewInfoFunc("glGetTextureParameterIivEXT", glGetTextureParameterIivEXT == NULL);
+  glewInfoFunc("glGetTextureParameterIuivEXT", glGetTextureParameterIuivEXT == NULL);
+  glewInfoFunc("glGetTextureParameterfvEXT", glGetTextureParameterfvEXT == NULL);
+  glewInfoFunc("glGetTextureParameterivEXT", glGetTextureParameterivEXT == NULL);
+  glewInfoFunc("glMapNamedBufferEXT", glMapNamedBufferEXT == NULL);
+  glewInfoFunc("glMatrixFrustumEXT", glMatrixFrustumEXT == NULL);
+  glewInfoFunc("glMatrixLoadIdentityEXT", glMatrixLoadIdentityEXT == NULL);
+  glewInfoFunc("glMatrixLoadTransposedEXT", glMatrixLoadTransposedEXT == NULL);
+  glewInfoFunc("glMatrixLoadTransposefEXT", glMatrixLoadTransposefEXT == NULL);
+  glewInfoFunc("glMatrixLoaddEXT", glMatrixLoaddEXT == NULL);
+  glewInfoFunc("glMatrixLoadfEXT", glMatrixLoadfEXT == NULL);
+  glewInfoFunc("glMatrixMultTransposedEXT", glMatrixMultTransposedEXT == NULL);
+  glewInfoFunc("glMatrixMultTransposefEXT", glMatrixMultTransposefEXT == NULL);
+  glewInfoFunc("glMatrixMultdEXT", glMatrixMultdEXT == NULL);
+  glewInfoFunc("glMatrixMultfEXT", glMatrixMultfEXT == NULL);
+  glewInfoFunc("glMatrixOrthoEXT", glMatrixOrthoEXT == NULL);
+  glewInfoFunc("glMatrixPopEXT", glMatrixPopEXT == NULL);
+  glewInfoFunc("glMatrixPushEXT", glMatrixPushEXT == NULL);
+  glewInfoFunc("glMatrixRotatedEXT", glMatrixRotatedEXT == NULL);
+  glewInfoFunc("glMatrixRotatefEXT", glMatrixRotatefEXT == NULL);
+  glewInfoFunc("glMatrixScaledEXT", glMatrixScaledEXT == NULL);
+  glewInfoFunc("glMatrixScalefEXT", glMatrixScalefEXT == NULL);
+  glewInfoFunc("glMatrixTranslatedEXT", glMatrixTranslatedEXT == NULL);
+  glewInfoFunc("glMatrixTranslatefEXT", glMatrixTranslatefEXT == NULL);
+  glewInfoFunc("glMultiTexBufferEXT", glMultiTexBufferEXT == NULL);
+  glewInfoFunc("glMultiTexCoordPointerEXT", glMultiTexCoordPointerEXT == NULL);
+  glewInfoFunc("glMultiTexEnvfEXT", glMultiTexEnvfEXT == NULL);
+  glewInfoFunc("glMultiTexEnvfvEXT", glMultiTexEnvfvEXT == NULL);
+  glewInfoFunc("glMultiTexEnviEXT", glMultiTexEnviEXT == NULL);
+  glewInfoFunc("glMultiTexEnvivEXT", glMultiTexEnvivEXT == NULL);
+  glewInfoFunc("glMultiTexGendEXT", glMultiTexGendEXT == NULL);
+  glewInfoFunc("glMultiTexGendvEXT", glMultiTexGendvEXT == NULL);
+  glewInfoFunc("glMultiTexGenfEXT", glMultiTexGenfEXT == NULL);
+  glewInfoFunc("glMultiTexGenfvEXT", glMultiTexGenfvEXT == NULL);
+  glewInfoFunc("glMultiTexGeniEXT", glMultiTexGeniEXT == NULL);
+  glewInfoFunc("glMultiTexGenivEXT", glMultiTexGenivEXT == NULL);
+  glewInfoFunc("glMultiTexImage1DEXT", glMultiTexImage1DEXT == NULL);
+  glewInfoFunc("glMultiTexImage2DEXT", glMultiTexImage2DEXT == NULL);
+  glewInfoFunc("glMultiTexImage3DEXT", glMultiTexImage3DEXT == NULL);
+  glewInfoFunc("glMultiTexParameterIivEXT", glMultiTexParameterIivEXT == NULL);
+  glewInfoFunc("glMultiTexParameterIuivEXT", glMultiTexParameterIuivEXT == NULL);
+  glewInfoFunc("glMultiTexParameterfEXT", glMultiTexParameterfEXT == NULL);
+  glewInfoFunc("glMultiTexParameterfvEXT", glMultiTexParameterfvEXT == NULL);
+  glewInfoFunc("glMultiTexParameteriEXT", glMultiTexParameteriEXT == NULL);
+  glewInfoFunc("glMultiTexParameterivEXT", glMultiTexParameterivEXT == NULL);
+  glewInfoFunc("glMultiTexRenderbufferEXT", glMultiTexRenderbufferEXT == NULL);
+  glewInfoFunc("glMultiTexSubImage1DEXT", glMultiTexSubImage1DEXT == NULL);
+  glewInfoFunc("glMultiTexSubImage2DEXT", glMultiTexSubImage2DEXT == NULL);
+  glewInfoFunc("glMultiTexSubImage3DEXT", glMultiTexSubImage3DEXT == NULL);
+  glewInfoFunc("glNamedBufferDataEXT", glNamedBufferDataEXT == NULL);
+  glewInfoFunc("glNamedBufferSubDataEXT", glNamedBufferSubDataEXT == NULL);
+  glewInfoFunc("glNamedFramebufferRenderbufferEXT", glNamedFramebufferRenderbufferEXT == NULL);
+  glewInfoFunc("glNamedFramebufferTexture1DEXT", glNamedFramebufferTexture1DEXT == NULL);
+  glewInfoFunc("glNamedFramebufferTexture2DEXT", glNamedFramebufferTexture2DEXT == NULL);
+  glewInfoFunc("glNamedFramebufferTexture3DEXT", glNamedFramebufferTexture3DEXT == NULL);
+  glewInfoFunc("glNamedFramebufferTextureEXT", glNamedFramebufferTextureEXT == NULL);
+  glewInfoFunc("glNamedFramebufferTextureFaceEXT", glNamedFramebufferTextureFaceEXT == NULL);
+  glewInfoFunc("glNamedFramebufferTextureLayerEXT", glNamedFramebufferTextureLayerEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParameter4dEXT", glNamedProgramLocalParameter4dEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParameter4dvEXT", glNamedProgramLocalParameter4dvEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParameter4fEXT", glNamedProgramLocalParameter4fEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParameter4fvEXT", glNamedProgramLocalParameter4fvEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParameterI4iEXT", glNamedProgramLocalParameterI4iEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParameterI4ivEXT", glNamedProgramLocalParameterI4ivEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParameterI4uiEXT", glNamedProgramLocalParameterI4uiEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParameterI4uivEXT", glNamedProgramLocalParameterI4uivEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParameters4fvEXT", glNamedProgramLocalParameters4fvEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParametersI4ivEXT", glNamedProgramLocalParametersI4ivEXT == NULL);
+  glewInfoFunc("glNamedProgramLocalParametersI4uivEXT", glNamedProgramLocalParametersI4uivEXT == NULL);
+  glewInfoFunc("glNamedProgramStringEXT", glNamedProgramStringEXT == NULL);
+  glewInfoFunc("glNamedRenderbufferStorageEXT", glNamedRenderbufferStorageEXT == NULL);
+  glewInfoFunc("glNamedRenderbufferStorageMultisampleCoverageEXT", glNamedRenderbufferStorageMultisampleCoverageEXT == NULL);
+  glewInfoFunc("glNamedRenderbufferStorageMultisampleEXT", glNamedRenderbufferStorageMultisampleEXT == NULL);
+  glewInfoFunc("glProgramUniform1fEXT", glProgramUniform1fEXT == NULL);
+  glewInfoFunc("glProgramUniform1fvEXT", glProgramUniform1fvEXT == NULL);
+  glewInfoFunc("glProgramUniform1iEXT", glProgramUniform1iEXT == NULL);
+  glewInfoFunc("glProgramUniform1ivEXT", glProgramUniform1ivEXT == NULL);
+  glewInfoFunc("glProgramUniform1uiEXT", glProgramUniform1uiEXT == NULL);
+  glewInfoFunc("glProgramUniform1uivEXT", glProgramUniform1uivEXT == NULL);
+  glewInfoFunc("glProgramUniform2fEXT", glProgramUniform2fEXT == NULL);
+  glewInfoFunc("glProgramUniform2fvEXT", glProgramUniform2fvEXT == NULL);
+  glewInfoFunc("glProgramUniform2iEXT", glProgramUniform2iEXT == NULL);
+  glewInfoFunc("glProgramUniform2ivEXT", glProgramUniform2ivEXT == NULL);
+  glewInfoFunc("glProgramUniform2uiEXT", glProgramUniform2uiEXT == NULL);
+  glewInfoFunc("glProgramUniform2uivEXT", glProgramUniform2uivEXT == NULL);
+  glewInfoFunc("glProgramUniform3fEXT", glProgramUniform3fEXT == NULL);
+  glewInfoFunc("glProgramUniform3fvEXT", glProgramUniform3fvEXT == NULL);
+  glewInfoFunc("glProgramUniform3iEXT", glProgramUniform3iEXT == NULL);
+  glewInfoFunc("glProgramUniform3ivEXT", glProgramUniform3ivEXT == NULL);
+  glewInfoFunc("glProgramUniform3uiEXT", glProgramUniform3uiEXT == NULL);
+  glewInfoFunc("glProgramUniform3uivEXT", glProgramUniform3uivEXT == NULL);
+  glewInfoFunc("glProgramUniform4fEXT", glProgramUniform4fEXT == NULL);
+  glewInfoFunc("glProgramUniform4fvEXT", glProgramUniform4fvEXT == NULL);
+  glewInfoFunc("glProgramUniform4iEXT", glProgramUniform4iEXT == NULL);
+  glewInfoFunc("glProgramUniform4ivEXT", glProgramUniform4ivEXT == NULL);
+  glewInfoFunc("glProgramUniform4uiEXT", glProgramUniform4uiEXT == NULL);
+  glewInfoFunc("glProgramUniform4uivEXT", glProgramUniform4uivEXT == NULL);
+  glewInfoFunc("glProgramUniformMatrix2fvEXT", glProgramUniformMatrix2fvEXT == NULL);
+  glewInfoFunc("glProgramUniformMatrix2x3fvEXT", glProgramUniformMatrix2x3fvEXT == NULL);
+  glewInfoFunc("glProgramUniformMatrix2x4fvEXT", glProgramUniformMatrix2x4fvEXT == NULL);
+  glewInfoFunc("glProgramUniformMatrix3fvEXT", glProgramUniformMatrix3fvEXT == NULL);
+  glewInfoFunc("glProgramUniformMatrix3x2fvEXT", glProgramUniformMatrix3x2fvEXT == NULL);
+  glewInfoFunc("glProgramUniformMatrix3x4fvEXT", glProgramUniformMatrix3x4fvEXT == NULL);
+  glewInfoFunc("glProgramUniformMatrix4fvEXT", glProgramUniformMatrix4fvEXT == NULL);
+  glewInfoFunc("glProgramUniformMatrix4x2fvEXT", glProgramUniformMatrix4x2fvEXT == NULL);
+  glewInfoFunc("glProgramUniformMatrix4x3fvEXT", glProgramUniformMatrix4x3fvEXT == NULL);
+  glewInfoFunc("glPushClientAttribDefaultEXT", glPushClientAttribDefaultEXT == NULL);
+  glewInfoFunc("glTextureBufferEXT", glTextureBufferEXT == NULL);
+  glewInfoFunc("glTextureImage1DEXT", glTextureImage1DEXT == NULL);
+  glewInfoFunc("glTextureImage2DEXT", glTextureImage2DEXT == NULL);
+  glewInfoFunc("glTextureImage3DEXT", glTextureImage3DEXT == NULL);
+  glewInfoFunc("glTextureParameterIivEXT", glTextureParameterIivEXT == NULL);
+  glewInfoFunc("glTextureParameterIuivEXT", glTextureParameterIuivEXT == NULL);
+  glewInfoFunc("glTextureParameterfEXT", glTextureParameterfEXT == NULL);
+  glewInfoFunc("glTextureParameterfvEXT", glTextureParameterfvEXT == NULL);
+  glewInfoFunc("glTextureParameteriEXT", glTextureParameteriEXT == NULL);
+  glewInfoFunc("glTextureParameterivEXT", glTextureParameterivEXT == NULL);
+  glewInfoFunc("glTextureRenderbufferEXT", glTextureRenderbufferEXT == NULL);
+  glewInfoFunc("glTextureSubImage1DEXT", glTextureSubImage1DEXT == NULL);
+  glewInfoFunc("glTextureSubImage2DEXT", glTextureSubImage2DEXT == NULL);
+  glewInfoFunc("glTextureSubImage3DEXT", glTextureSubImage3DEXT == NULL);
+  glewInfoFunc("glUnmapNamedBufferEXT", glUnmapNamedBufferEXT == NULL);
+}
+
+#endif /* GL_EXT_direct_state_access */
+
+#ifdef GL_EXT_draw_buffers2
+
+static void _glewInfo_GL_EXT_draw_buffers2 (void)
+{
+  glewPrintExt("GL_EXT_draw_buffers2", GLEW_EXT_draw_buffers2, glewIsSupported("GL_EXT_draw_buffers2"), glewGetExtension("GL_EXT_draw_buffers2"));
+
+  glewInfoFunc("glColorMaskIndexedEXT", glColorMaskIndexedEXT == NULL);
+  glewInfoFunc("glDisableIndexedEXT", glDisableIndexedEXT == NULL);
+  glewInfoFunc("glEnableIndexedEXT", glEnableIndexedEXT == NULL);
+  glewInfoFunc("glGetBooleanIndexedvEXT", glGetBooleanIndexedvEXT == NULL);
+  glewInfoFunc("glGetIntegerIndexedvEXT", glGetIntegerIndexedvEXT == NULL);
+  glewInfoFunc("glIsEnabledIndexedEXT", glIsEnabledIndexedEXT == NULL);
+}
+
+#endif /* GL_EXT_draw_buffers2 */
+
+#ifdef GL_EXT_draw_instanced
+
+static void _glewInfo_GL_EXT_draw_instanced (void)
+{
+  glewPrintExt("GL_EXT_draw_instanced", GLEW_EXT_draw_instanced, glewIsSupported("GL_EXT_draw_instanced"), glewGetExtension("GL_EXT_draw_instanced"));
+
+  glewInfoFunc("glDrawArraysInstancedEXT", glDrawArraysInstancedEXT == NULL);
+  glewInfoFunc("glDrawElementsInstancedEXT", glDrawElementsInstancedEXT == NULL);
+}
+
+#endif /* GL_EXT_draw_instanced */
+
+#ifdef GL_EXT_draw_range_elements
+
+static void _glewInfo_GL_EXT_draw_range_elements (void)
+{
+  glewPrintExt("GL_EXT_draw_range_elements", GLEW_EXT_draw_range_elements, glewIsSupported("GL_EXT_draw_range_elements"), glewGetExtension("GL_EXT_draw_range_elements"));
+
+  glewInfoFunc("glDrawRangeElementsEXT", glDrawRangeElementsEXT == NULL);
+}
+
+#endif /* GL_EXT_draw_range_elements */
+
+#ifdef GL_EXT_fog_coord
+
+static void _glewInfo_GL_EXT_fog_coord (void)
+{
+  glewPrintExt("GL_EXT_fog_coord", GLEW_EXT_fog_coord, glewIsSupported("GL_EXT_fog_coord"), glewGetExtension("GL_EXT_fog_coord"));
+
+  glewInfoFunc("glFogCoordPointerEXT", glFogCoordPointerEXT == NULL);
+  glewInfoFunc("glFogCoorddEXT", glFogCoorddEXT == NULL);
+  glewInfoFunc("glFogCoorddvEXT", glFogCoorddvEXT == NULL);
+  glewInfoFunc("glFogCoordfEXT", glFogCoordfEXT == NULL);
+  glewInfoFunc("glFogCoordfvEXT", glFogCoordfvEXT == NULL);
+}
+
+#endif /* GL_EXT_fog_coord */
+
+#ifdef GL_EXT_fragment_lighting
+
+static void _glewInfo_GL_EXT_fragment_lighting (void)
+{
+  glewPrintExt("GL_EXT_fragment_lighting", GLEW_EXT_fragment_lighting, glewIsSupported("GL_EXT_fragment_lighting"), glewGetExtension("GL_EXT_fragment_lighting"));
+
+  glewInfoFunc("glFragmentColorMaterialEXT", glFragmentColorMaterialEXT == NULL);
+  glewInfoFunc("glFragmentLightModelfEXT", glFragmentLightModelfEXT == NULL);
+  glewInfoFunc("glFragmentLightModelfvEXT", glFragmentLightModelfvEXT == NULL);
+  glewInfoFunc("glFragmentLightModeliEXT", glFragmentLightModeliEXT == NULL);
+  glewInfoFunc("glFragmentLightModelivEXT", glFragmentLightModelivEXT == NULL);
+  glewInfoFunc("glFragmentLightfEXT", glFragmentLightfEXT == NULL);
+  glewInfoFunc("glFragmentLightfvEXT", glFragmentLightfvEXT == NULL);
+  glewInfoFunc("glFragmentLightiEXT", glFragmentLightiEXT == NULL);
+  glewInfoFunc("glFragmentLightivEXT", glFragmentLightivEXT == NULL);
+  glewInfoFunc("glFragmentMaterialfEXT", glFragmentMaterialfEXT == NULL);
+  glewInfoFunc("glFragmentMaterialfvEXT", glFragmentMaterialfvEXT == NULL);
+  glewInfoFunc("glFragmentMaterialiEXT", glFragmentMaterialiEXT == NULL);
+  glewInfoFunc("glFragmentMaterialivEXT", glFragmentMaterialivEXT == NULL);
+  glewInfoFunc("glGetFragmentLightfvEXT", glGetFragmentLightfvEXT == NULL);
+  glewInfoFunc("glGetFragmentLightivEXT", glGetFragmentLightivEXT == NULL);
+  glewInfoFunc("glGetFragmentMaterialfvEXT", glGetFragmentMaterialfvEXT == NULL);
+  glewInfoFunc("glGetFragmentMaterialivEXT", glGetFragmentMaterialivEXT == NULL);
+  glewInfoFunc("glLightEnviEXT", glLightEnviEXT == NULL);
+}
+
+#endif /* GL_EXT_fragment_lighting */
+
+#ifdef GL_EXT_framebuffer_blit
+
+static void _glewInfo_GL_EXT_framebuffer_blit (void)
+{
+  glewPrintExt("GL_EXT_framebuffer_blit", GLEW_EXT_framebuffer_blit, glewIsSupported("GL_EXT_framebuffer_blit"), glewGetExtension("GL_EXT_framebuffer_blit"));
+
+  glewInfoFunc("glBlitFramebufferEXT", glBlitFramebufferEXT == NULL);
+}
+
+#endif /* GL_EXT_framebuffer_blit */
+
+#ifdef GL_EXT_framebuffer_multisample
+
+static void _glewInfo_GL_EXT_framebuffer_multisample (void)
+{
+  glewPrintExt("GL_EXT_framebuffer_multisample", GLEW_EXT_framebuffer_multisample, glewIsSupported("GL_EXT_framebuffer_multisample"), glewGetExtension("GL_EXT_framebuffer_multisample"));
+
+  glewInfoFunc("glRenderbufferStorageMultisampleEXT", glRenderbufferStorageMultisampleEXT == NULL);
+}
+
+#endif /* GL_EXT_framebuffer_multisample */
+
+#ifdef GL_EXT_framebuffer_object
+
+static void _glewInfo_GL_EXT_framebuffer_object (void)
+{
+  glewPrintExt("GL_EXT_framebuffer_object", GLEW_EXT_framebuffer_object, glewIsSupported("GL_EXT_framebuffer_object"), glewGetExtension("GL_EXT_framebuffer_object"));
+
+  glewInfoFunc("glBindFramebufferEXT", glBindFramebufferEXT == NULL);
+  glewInfoFunc("glBindRenderbufferEXT", glBindRenderbufferEXT == NULL);
+  glewInfoFunc("glCheckFramebufferStatusEXT", glCheckFramebufferStatusEXT == NULL);
+  glewInfoFunc("glDeleteFramebuffersEXT", glDeleteFramebuffersEXT == NULL);
+  glewInfoFunc("glDeleteRenderbuffersEXT", glDeleteRenderbuffersEXT == NULL);
+  glewInfoFunc("glFramebufferRenderbufferEXT", glFramebufferRenderbufferEXT == NULL);
+  glewInfoFunc("glFramebufferTexture1DEXT", glFramebufferTexture1DEXT == NULL);
+  glewInfoFunc("glFramebufferTexture2DEXT", glFramebufferTexture2DEXT == NULL);
+  glewInfoFunc("glFramebufferTexture3DEXT", glFramebufferTexture3DEXT == NULL);
+  glewInfoFunc("glGenFramebuffersEXT", glGenFramebuffersEXT == NULL);
+  glewInfoFunc("glGenRenderbuffersEXT", glGenRenderbuffersEXT == NULL);
+  glewInfoFunc("glGenerateMipmapEXT", glGenerateMipmapEXT == NULL);
+  glewInfoFunc("glGetFramebufferAttachmentParameterivEXT", glGetFramebufferAttachmentParameterivEXT == NULL);
+  glewInfoFunc("glGetRenderbufferParameterivEXT", glGetRenderbufferParameterivEXT == NULL);
+  glewInfoFunc("glIsFramebufferEXT", glIsFramebufferEXT == NULL);
+  glewInfoFunc("glIsRenderbufferEXT", glIsRenderbufferEXT == NULL);
+  glewInfoFunc("glRenderbufferStorageEXT", glRenderbufferStorageEXT == NULL);
+}
+
+#endif /* GL_EXT_framebuffer_object */
+
+#ifdef GL_EXT_framebuffer_sRGB
+
+static void _glewInfo_GL_EXT_framebuffer_sRGB (void)
+{
+  glewPrintExt("GL_EXT_framebuffer_sRGB", GLEW_EXT_framebuffer_sRGB, glewIsSupported("GL_EXT_framebuffer_sRGB"), glewGetExtension("GL_EXT_framebuffer_sRGB"));
+}
+
+#endif /* GL_EXT_framebuffer_sRGB */
+
+#ifdef GL_EXT_geometry_shader4
+
+static void _glewInfo_GL_EXT_geometry_shader4 (void)
+{
+  glewPrintExt("GL_EXT_geometry_shader4", GLEW_EXT_geometry_shader4, glewIsSupported("GL_EXT_geometry_shader4"), glewGetExtension("GL_EXT_geometry_shader4"));
+
+  glewInfoFunc("glFramebufferTextureEXT", glFramebufferTextureEXT == NULL);
+  glewInfoFunc("glFramebufferTextureFaceEXT", glFramebufferTextureFaceEXT == NULL);
+  glewInfoFunc("glFramebufferTextureLayerEXT", glFramebufferTextureLayerEXT == NULL);
+  glewInfoFunc("glProgramParameteriEXT", glProgramParameteriEXT == NULL);
+}
+
+#endif /* GL_EXT_geometry_shader4 */
+
+#ifdef GL_EXT_gpu_program_parameters
+
+static void _glewInfo_GL_EXT_gpu_program_parameters (void)
+{
+  glewPrintExt("GL_EXT_gpu_program_parameters", GLEW_EXT_gpu_program_parameters, glewIsSupported("GL_EXT_gpu_program_parameters"), glewGetExtension("GL_EXT_gpu_program_parameters"));
+
+  glewInfoFunc("glProgramEnvParameters4fvEXT", glProgramEnvParameters4fvEXT == NULL);
+  glewInfoFunc("glProgramLocalParameters4fvEXT", glProgramLocalParameters4fvEXT == NULL);
+}
+
+#endif /* GL_EXT_gpu_program_parameters */
+
+#ifdef GL_EXT_gpu_shader4
+
+static void _glewInfo_GL_EXT_gpu_shader4 (void)
+{
+  glewPrintExt("GL_EXT_gpu_shader4", GLEW_EXT_gpu_shader4, glewIsSupported("GL_EXT_gpu_shader4"), glewGetExtension("GL_EXT_gpu_shader4"));
+
+  glewInfoFunc("glBindFragDataLocationEXT", glBindFragDataLocationEXT == NULL);
+  glewInfoFunc("glGetFragDataLocationEXT", glGetFragDataLocationEXT == NULL);
+  glewInfoFunc("glGetUniformuivEXT", glGetUniformuivEXT == NULL);
+  glewInfoFunc("glGetVertexAttribIivEXT", glGetVertexAttribIivEXT == NULL);
+  glewInfoFunc("glGetVertexAttribIuivEXT", glGetVertexAttribIuivEXT == NULL);
+  glewInfoFunc("glUniform1uiEXT", glUniform1uiEXT == NULL);
+  glewInfoFunc("glUniform1uivEXT", glUniform1uivEXT == NULL);
+  glewInfoFunc("glUniform2uiEXT", glUniform2uiEXT == NULL);
+  glewInfoFunc("glUniform2uivEXT", glUniform2uivEXT == NULL);
+  glewInfoFunc("glUniform3uiEXT", glUniform3uiEXT == NULL);
+  glewInfoFunc("glUniform3uivEXT", glUniform3uivEXT == NULL);
+  glewInfoFunc("glUniform4uiEXT", glUniform4uiEXT == NULL);
+  glewInfoFunc("glUniform4uivEXT", glUniform4uivEXT == NULL);
+  glewInfoFunc("glVertexAttribI1iEXT", glVertexAttribI1iEXT == NULL);
+  glewInfoFunc("glVertexAttribI1ivEXT", glVertexAttribI1ivEXT == NULL);
+  glewInfoFunc("glVertexAttribI1uiEXT", glVertexAttribI1uiEXT == NULL);
+  glewInfoFunc("glVertexAttribI1uivEXT", glVertexAttribI1uivEXT == NULL);
+  glewInfoFunc("glVertexAttribI2iEXT", glVertexAttribI2iEXT == NULL);
+  glewInfoFunc("glVertexAttribI2ivEXT", glVertexAttribI2ivEXT == NULL);
+  glewInfoFunc("glVertexAttribI2uiEXT", glVertexAttribI2uiEXT == NULL);
+  glewInfoFunc("glVertexAttribI2uivEXT", glVertexAttribI2uivEXT == NULL);
+  glewInfoFunc("glVertexAttribI3iEXT", glVertexAttribI3iEXT == NULL);
+  glewInfoFunc("glVertexAttribI3ivEXT", glVertexAttribI3ivEXT == NULL);
+  glewInfoFunc("glVertexAttribI3uiEXT", glVertexAttribI3uiEXT == NULL);
+  glewInfoFunc("glVertexAttribI3uivEXT", glVertexAttribI3uivEXT == NULL);
+  glewInfoFunc("glVertexAttribI4bvEXT", glVertexAttribI4bvEXT == NULL);
+  glewInfoFunc("glVertexAttribI4iEXT", glVertexAttribI4iEXT == NULL);
+  glewInfoFunc("glVertexAttribI4ivEXT", glVertexAttribI4ivEXT == NULL);
+  glewInfoFunc("glVertexAttribI4svEXT", glVertexAttribI4svEXT == NULL);
+  glewInfoFunc("glVertexAttribI4ubvEXT", glVertexAttribI4ubvEXT == NULL);
+  glewInfoFunc("glVertexAttribI4uiEXT", glVertexAttribI4uiEXT == NULL);
+  glewInfoFunc("glVertexAttribI4uivEXT", glVertexAttribI4uivEXT == NULL);
+  glewInfoFunc("glVertexAttribI4usvEXT", glVertexAttribI4usvEXT == NULL);
+  glewInfoFunc("glVertexAttribIPointerEXT", glVertexAttribIPointerEXT == NULL);
+}
+
+#endif /* GL_EXT_gpu_shader4 */
+
+#ifdef GL_EXT_histogram
+
+static void _glewInfo_GL_EXT_histogram (void)
+{
+  glewPrintExt("GL_EXT_histogram", GLEW_EXT_histogram, glewIsSupported("GL_EXT_histogram"), glewGetExtension("GL_EXT_histogram"));
+
+  glewInfoFunc("glGetHistogramEXT", glGetHistogramEXT == NULL);
+  glewInfoFunc("glGetHistogramParameterfvEXT", glGetHistogramParameterfvEXT == NULL);
+  glewInfoFunc("glGetHistogramParameterivEXT", glGetHistogramParameterivEXT == NULL);
+  glewInfoFunc("glGetMinmaxEXT", glGetMinmaxEXT == NULL);
+  glewInfoFunc("glGetMinmaxParameterfvEXT", glGetMinmaxParameterfvEXT == NULL);
+  glewInfoFunc("glGetMinmaxParameterivEXT", glGetMinmaxParameterivEXT == NULL);
+  glewInfoFunc("glHistogramEXT", glHistogramEXT == NULL);
+  glewInfoFunc("glMinmaxEXT", glMinmaxEXT == NULL);
+  glewInfoFunc("glResetHistogramEXT", glResetHistogramEXT == NULL);
+  glewInfoFunc("glResetMinmaxEXT", glResetMinmaxEXT == NULL);
+}
+
+#endif /* GL_EXT_histogram */
+
+#ifdef GL_EXT_index_array_formats
+
+static void _glewInfo_GL_EXT_index_array_formats (void)
+{
+  glewPrintExt("GL_EXT_index_array_formats", GLEW_EXT_index_array_formats, glewIsSupported("GL_EXT_index_array_formats"), glewGetExtension("GL_EXT_index_array_formats"));
+}
+
+#endif /* GL_EXT_index_array_formats */
+
+#ifdef GL_EXT_index_func
+
+static void _glewInfo_GL_EXT_index_func (void)
+{
+  glewPrintExt("GL_EXT_index_func", GLEW_EXT_index_func, glewIsSupported("GL_EXT_index_func"), glewGetExtension("GL_EXT_index_func"));
+
+  glewInfoFunc("glIndexFuncEXT", glIndexFuncEXT == NULL);
+}
+
+#endif /* GL_EXT_index_func */
+
+#ifdef GL_EXT_index_material
+
+static void _glewInfo_GL_EXT_index_material (void)
+{
+  glewPrintExt("GL_EXT_index_material", GLEW_EXT_index_material, glewIsSupported("GL_EXT_index_material"), glewGetExtension("GL_EXT_index_material"));
+
+  glewInfoFunc("glIndexMaterialEXT", glIndexMaterialEXT == NULL);
+}
+
+#endif /* GL_EXT_index_material */
+
+#ifdef GL_EXT_index_texture
+
+static void _glewInfo_GL_EXT_index_texture (void)
+{
+  glewPrintExt("GL_EXT_index_texture", GLEW_EXT_index_texture, glewIsSupported("GL_EXT_index_texture"), glewGetExtension("GL_EXT_index_texture"));
+}
+
+#endif /* GL_EXT_index_texture */
+
+#ifdef GL_EXT_light_texture
+
+static void _glewInfo_GL_EXT_light_texture (void)
+{
+  glewPrintExt("GL_EXT_light_texture", GLEW_EXT_light_texture, glewIsSupported("GL_EXT_light_texture"), glewGetExtension("GL_EXT_light_texture"));
+
+  glewInfoFunc("glApplyTextureEXT", glApplyTextureEXT == NULL);
+  glewInfoFunc("glTextureLightEXT", glTextureLightEXT == NULL);
+  glewInfoFunc("glTextureMaterialEXT", glTextureMaterialEXT == NULL);
+}
+
+#endif /* GL_EXT_light_texture */
+
+#ifdef GL_EXT_misc_attribute
+
+static void _glewInfo_GL_EXT_misc_attribute (void)
+{
+  glewPrintExt("GL_EXT_misc_attribute", GLEW_EXT_misc_attribute, glewIsSupported("GL_EXT_misc_attribute"), glewGetExtension("GL_EXT_misc_attribute"));
+}
+
+#endif /* GL_EXT_misc_attribute */
+
+#ifdef GL_EXT_multi_draw_arrays
+
+static void _glewInfo_GL_EXT_multi_draw_arrays (void)
+{
+  glewPrintExt("GL_EXT_multi_draw_arrays", GLEW_EXT_multi_draw_arrays, glewIsSupported("GL_EXT_multi_draw_arrays"), glewGetExtension("GL_EXT_multi_draw_arrays"));
+
+  glewInfoFunc("glMultiDrawArraysEXT", glMultiDrawArraysEXT == NULL);
+  glewInfoFunc("glMultiDrawElementsEXT", glMultiDrawElementsEXT == NULL);
+}
+
+#endif /* GL_EXT_multi_draw_arrays */
+
+#ifdef GL_EXT_multisample
+
+static void _glewInfo_GL_EXT_multisample (void)
+{
+  glewPrintExt("GL_EXT_multisample", GLEW_EXT_multisample, glewIsSupported("GL_EXT_multisample"), glewGetExtension("GL_EXT_multisample"));
+
+  glewInfoFunc("glSampleMaskEXT", glSampleMaskEXT == NULL);
+  glewInfoFunc("glSamplePatternEXT", glSamplePatternEXT == NULL);
+}
+
+#endif /* GL_EXT_multisample */
+
+#ifdef GL_EXT_packed_depth_stencil
+
+static void _glewInfo_GL_EXT_packed_depth_stencil (void)
+{
+  glewPrintExt("GL_EXT_packed_depth_stencil", GLEW_EXT_packed_depth_stencil, glewIsSupported("GL_EXT_packed_depth_stencil"), glewGetExtension("GL_EXT_packed_depth_stencil"));
+}
+
+#endif /* GL_EXT_packed_depth_stencil */
+
+#ifdef GL_EXT_packed_float
+
+static void _glewInfo_GL_EXT_packed_float (void)
+{
+  glewPrintExt("GL_EXT_packed_float", GLEW_EXT_packed_float, glewIsSupported("GL_EXT_packed_float"), glewGetExtension("GL_EXT_packed_float"));
+}
+
+#endif /* GL_EXT_packed_float */
+
+#ifdef GL_EXT_packed_pixels
+
+static void _glewInfo_GL_EXT_packed_pixels (void)
+{
+  glewPrintExt("GL_EXT_packed_pixels", GLEW_EXT_packed_pixels, glewIsSupported("GL_EXT_packed_pixels"), glewGetExtension("GL_EXT_packed_pixels"));
+}
+
+#endif /* GL_EXT_packed_pixels */
+
+#ifdef GL_EXT_paletted_texture
+
+static void _glewInfo_GL_EXT_paletted_texture (void)
+{
+  glewPrintExt("GL_EXT_paletted_texture", GLEW_EXT_paletted_texture, glewIsSupported("GL_EXT_paletted_texture"), glewGetExtension("GL_EXT_paletted_texture"));
+
+  glewInfoFunc("glColorTableEXT", glColorTableEXT == NULL);
+  glewInfoFunc("glGetColorTableEXT", glGetColorTableEXT == NULL);
+  glewInfoFunc("glGetColorTableParameterfvEXT", glGetColorTableParameterfvEXT == NULL);
+  glewInfoFunc("glGetColorTableParameterivEXT", glGetColorTableParameterivEXT == NULL);
+}
+
+#endif /* GL_EXT_paletted_texture */
+
+#ifdef GL_EXT_pixel_buffer_object
+
+static void _glewInfo_GL_EXT_pixel_buffer_object (void)
+{
+  glewPrintExt("GL_EXT_pixel_buffer_object", GLEW_EXT_pixel_buffer_object, glewIsSupported("GL_EXT_pixel_buffer_object"), glewGetExtension("GL_EXT_pixel_buffer_object"));
+}
+
+#endif /* GL_EXT_pixel_buffer_object */
+
+#ifdef GL_EXT_pixel_transform
+
+static void _glewInfo_GL_EXT_pixel_transform (void)
+{
+  glewPrintExt("GL_EXT_pixel_transform", GLEW_EXT_pixel_transform, glewIsSupported("GL_EXT_pixel_transform"), glewGetExtension("GL_EXT_pixel_transform"));
+
+  glewInfoFunc("glGetPixelTransformParameterfvEXT", glGetPixelTransformParameterfvEXT == NULL);
+  glewInfoFunc("glGetPixelTransformParameterivEXT", glGetPixelTransformParameterivEXT == NULL);
+  glewInfoFunc("glPixelTransformParameterfEXT", glPixelTransformParameterfEXT == NULL);
+  glewInfoFunc("glPixelTransformParameterfvEXT", glPixelTransformParameterfvEXT == NULL);
+  glewInfoFunc("glPixelTransformParameteriEXT", glPixelTransformParameteriEXT == NULL);
+  glewInfoFunc("glPixelTransformParameterivEXT", glPixelTransformParameterivEXT == NULL);
+}
+
+#endif /* GL_EXT_pixel_transform */
+
+#ifdef GL_EXT_pixel_transform_color_table
+
+static void _glewInfo_GL_EXT_pixel_transform_color_table (void)
+{
+  glewPrintExt("GL_EXT_pixel_transform_color_table", GLEW_EXT_pixel_transform_color_table, glewIsSupported("GL_EXT_pixel_transform_color_table"), glewGetExtension("GL_EXT_pixel_transform_color_table"));
+}
+
+#endif /* GL_EXT_pixel_transform_color_table */
+
+#ifdef GL_EXT_point_parameters
+
+static void _glewInfo_GL_EXT_point_parameters (void)
+{
+  glewPrintExt("GL_EXT_point_parameters", GLEW_EXT_point_parameters, glewIsSupported("GL_EXT_point_parameters"), glewGetExtension("GL_EXT_point_parameters"));
+
+  glewInfoFunc("glPointParameterfEXT", glPointParameterfEXT == NULL);
+  glewInfoFunc("glPointParameterfvEXT", glPointParameterfvEXT == NULL);
+}
+
+#endif /* GL_EXT_point_parameters */
+
+#ifdef GL_EXT_polygon_offset
+
+static void _glewInfo_GL_EXT_polygon_offset (void)
+{
+  glewPrintExt("GL_EXT_polygon_offset", GLEW_EXT_polygon_offset, glewIsSupported("GL_EXT_polygon_offset"), glewGetExtension("GL_EXT_polygon_offset"));
+
+  glewInfoFunc("glPolygonOffsetEXT", glPolygonOffsetEXT == NULL);
+}
+
+#endif /* GL_EXT_polygon_offset */
+
+#ifdef GL_EXT_rescale_normal
+
+static void _glewInfo_GL_EXT_rescale_normal (void)
+{
+  glewPrintExt("GL_EXT_rescale_normal", GLEW_EXT_rescale_normal, glewIsSupported("GL_EXT_rescale_normal"), glewGetExtension("GL_EXT_rescale_normal"));
+}
+
+#endif /* GL_EXT_rescale_normal */
+
+#ifdef GL_EXT_scene_marker
+
+static void _glewInfo_GL_EXT_scene_marker (void)
+{
+  glewPrintExt("GL_EXT_scene_marker", GLEW_EXT_scene_marker, glewIsSupported("GL_EXT_scene_marker"), glewGetExtension("GL_EXT_scene_marker"));
+
+  glewInfoFunc("glBeginSceneEXT", glBeginSceneEXT == NULL);
+  glewInfoFunc("glEndSceneEXT", glEndSceneEXT == NULL);
+}
+
+#endif /* GL_EXT_scene_marker */
+
+#ifdef GL_EXT_secondary_color
+
+static void _glewInfo_GL_EXT_secondary_color (void)
+{
+  glewPrintExt("GL_EXT_secondary_color", GLEW_EXT_secondary_color, glewIsSupported("GL_EXT_secondary_color"), glewGetExtension("GL_EXT_secondary_color"));
+
+  glewInfoFunc("glSecondaryColor3bEXT", glSecondaryColor3bEXT == NULL);
+  glewInfoFunc("glSecondaryColor3bvEXT", glSecondaryColor3bvEXT == NULL);
+  glewInfoFunc("glSecondaryColor3dEXT", glSecondaryColor3dEXT == NULL);
+  glewInfoFunc("glSecondaryColor3dvEXT", glSecondaryColor3dvEXT == NULL);
+  glewInfoFunc("glSecondaryColor3fEXT", glSecondaryColor3fEXT == NULL);
+  glewInfoFunc("glSecondaryColor3fvEXT", glSecondaryColor3fvEXT == NULL);
+  glewInfoFunc("glSecondaryColor3iEXT", glSecondaryColor3iEXT == NULL);
+  glewInfoFunc("glSecondaryColor3ivEXT", glSecondaryColor3ivEXT == NULL);
+  glewInfoFunc("glSecondaryColor3sEXT", glSecondaryColor3sEXT == NULL);
+  glewInfoFunc("glSecondaryColor3svEXT", glSecondaryColor3svEXT == NULL);
+  glewInfoFunc("glSecondaryColor3ubEXT", glSecondaryColor3ubEXT == NULL);
+  glewInfoFunc("glSecondaryColor3ubvEXT", glSecondaryColor3ubvEXT == NULL);
+  glewInfoFunc("glSecondaryColor3uiEXT", glSecondaryColor3uiEXT == NULL);
+  glewInfoFunc("glSecondaryColor3uivEXT", glSecondaryColor3uivEXT == NULL);
+  glewInfoFunc("glSecondaryColor3usEXT", glSecondaryColor3usEXT == NULL);
+  glewInfoFunc("glSecondaryColor3usvEXT", glSecondaryColor3usvEXT == NULL);
+  glewInfoFunc("glSecondaryColorPointerEXT", glSecondaryColorPointerEXT == NULL);
+}
+
+#endif /* GL_EXT_secondary_color */
+
+#ifdef GL_EXT_separate_specular_color
+
+static void _glewInfo_GL_EXT_separate_specular_color (void)
+{
+  glewPrintExt("GL_EXT_separate_specular_color", GLEW_EXT_separate_specular_color, glewIsSupported("GL_EXT_separate_specular_color"), glewGetExtension("GL_EXT_separate_specular_color"));
+}
+
+#endif /* GL_EXT_separate_specular_color */
+
+#ifdef GL_EXT_shadow_funcs
+
+static void _glewInfo_GL_EXT_shadow_funcs (void)
+{
+  glewPrintExt("GL_EXT_shadow_funcs", GLEW_EXT_shadow_funcs, glewIsSupported("GL_EXT_shadow_funcs"), glewGetExtension("GL_EXT_shadow_funcs"));
+}
+
+#endif /* GL_EXT_shadow_funcs */
+
+#ifdef GL_EXT_shared_texture_palette
+
+static void _glewInfo_GL_EXT_shared_texture_palette (void)
+{
+  glewPrintExt("GL_EXT_shared_texture_palette", GLEW_EXT_shared_texture_palette, glewIsSupported("GL_EXT_shared_texture_palette"), glewGetExtension("GL_EXT_shared_texture_palette"));
+}
+
+#endif /* GL_EXT_shared_texture_palette */
+
+#ifdef GL_EXT_stencil_clear_tag
+
+static void _glewInfo_GL_EXT_stencil_clear_tag (void)
+{
+  glewPrintExt("GL_EXT_stencil_clear_tag", GLEW_EXT_stencil_clear_tag, glewIsSupported("GL_EXT_stencil_clear_tag"), glewGetExtension("GL_EXT_stencil_clear_tag"));
+}
+
+#endif /* GL_EXT_stencil_clear_tag */
+
+#ifdef GL_EXT_stencil_two_side
+
+static void _glewInfo_GL_EXT_stencil_two_side (void)
+{
+  glewPrintExt("GL_EXT_stencil_two_side", GLEW_EXT_stencil_two_side, glewIsSupported("GL_EXT_stencil_two_side"), glewGetExtension("GL_EXT_stencil_two_side"));
+
+  glewInfoFunc("glActiveStencilFaceEXT", glActiveStencilFaceEXT == NULL);
+}
+
+#endif /* GL_EXT_stencil_two_side */
+
+#ifdef GL_EXT_stencil_wrap
+
+static void _glewInfo_GL_EXT_stencil_wrap (void)
+{
+  glewPrintExt("GL_EXT_stencil_wrap", GLEW_EXT_stencil_wrap, glewIsSupported("GL_EXT_stencil_wrap"), glewGetExtension("GL_EXT_stencil_wrap"));
+}
+
+#endif /* GL_EXT_stencil_wrap */
+
+#ifdef GL_EXT_subtexture
+
+static void _glewInfo_GL_EXT_subtexture (void)
+{
+  glewPrintExt("GL_EXT_subtexture", GLEW_EXT_subtexture, glewIsSupported("GL_EXT_subtexture"), glewGetExtension("GL_EXT_subtexture"));
+
+  glewInfoFunc("glTexSubImage1DEXT", glTexSubImage1DEXT == NULL);
+  glewInfoFunc("glTexSubImage2DEXT", glTexSubImage2DEXT == NULL);
+  glewInfoFunc("glTexSubImage3DEXT", glTexSubImage3DEXT == NULL);
+}
+
+#endif /* GL_EXT_subtexture */
+
+#ifdef GL_EXT_texture
+
+static void _glewInfo_GL_EXT_texture (void)
+{
+  glewPrintExt("GL_EXT_texture", GLEW_EXT_texture, glewIsSupported("GL_EXT_texture"), glewGetExtension("GL_EXT_texture"));
+}
+
+#endif /* GL_EXT_texture */
+
+#ifdef GL_EXT_texture3D
+
+static void _glewInfo_GL_EXT_texture3D (void)
+{
+  glewPrintExt("GL_EXT_texture3D", GLEW_EXT_texture3D, glewIsSupported("GL_EXT_texture3D"), glewGetExtension("GL_EXT_texture3D"));
+
+  glewInfoFunc("glTexImage3DEXT", glTexImage3DEXT == NULL);
+}
+
+#endif /* GL_EXT_texture3D */
+
+#ifdef GL_EXT_texture_array
+
+static void _glewInfo_GL_EXT_texture_array (void)
+{
+  glewPrintExt("GL_EXT_texture_array", GLEW_EXT_texture_array, glewIsSupported("GL_EXT_texture_array"), glewGetExtension("GL_EXT_texture_array"));
+}
+
+#endif /* GL_EXT_texture_array */
+
+#ifdef GL_EXT_texture_buffer_object
+
+static void _glewInfo_GL_EXT_texture_buffer_object (void)
+{
+  glewPrintExt("GL_EXT_texture_buffer_object", GLEW_EXT_texture_buffer_object, glewIsSupported("GL_EXT_texture_buffer_object"), glewGetExtension("GL_EXT_texture_buffer_object"));
+
+  glewInfoFunc("glTexBufferEXT", glTexBufferEXT == NULL);
+}
+
+#endif /* GL_EXT_texture_buffer_object */
+
+#ifdef GL_EXT_texture_compression_dxt1
+
+static void _glewInfo_GL_EXT_texture_compression_dxt1 (void)
+{
+  glewPrintExt("GL_EXT_texture_compression_dxt1", GLEW_EXT_texture_compression_dxt1, glewIsSupported("GL_EXT_texture_compression_dxt1"), glewGetExtension("GL_EXT_texture_compression_dxt1"));
+}
+
+#endif /* GL_EXT_texture_compression_dxt1 */
+
+#ifdef GL_EXT_texture_compression_latc
+
+static void _glewInfo_GL_EXT_texture_compression_latc (void)
+{
+  glewPrintExt("GL_EXT_texture_compression_latc", GLEW_EXT_texture_compression_latc, glewIsSupported("GL_EXT_texture_compression_latc"), glewGetExtension("GL_EXT_texture_compression_latc"));
+}
+
+#endif /* GL_EXT_texture_compression_latc */
+
+#ifdef GL_EXT_texture_compression_rgtc
+
+static void _glewInfo_GL_EXT_texture_compression_rgtc (void)
+{
+  glewPrintExt("GL_EXT_texture_compression_rgtc", GLEW_EXT_texture_compression_rgtc, glewIsSupported("GL_EXT_texture_compression_rgtc"), glewGetExtension("GL_EXT_texture_compression_rgtc"));
+}
+
+#endif /* GL_EXT_texture_compression_rgtc */
+
+#ifdef GL_EXT_texture_compression_s3tc
+
+static void _glewInfo_GL_EXT_texture_compression_s3tc (void)
+{
+  glewPrintExt("GL_EXT_texture_compression_s3tc", GLEW_EXT_texture_compression_s3tc, glewIsSupported("GL_EXT_texture_compression_s3tc"), glewGetExtension("GL_EXT_texture_compression_s3tc"));
+}
+
+#endif /* GL_EXT_texture_compression_s3tc */
+
+#ifdef GL_EXT_texture_cube_map
+
+static void _glewInfo_GL_EXT_texture_cube_map (void)
+{
+  glewPrintExt("GL_EXT_texture_cube_map", GLEW_EXT_texture_cube_map, glewIsSupported("GL_EXT_texture_cube_map"), glewGetExtension("GL_EXT_texture_cube_map"));
+}
+
+#endif /* GL_EXT_texture_cube_map */
+
+#ifdef GL_EXT_texture_edge_clamp
+
+static void _glewInfo_GL_EXT_texture_edge_clamp (void)
+{
+  glewPrintExt("GL_EXT_texture_edge_clamp", GLEW_EXT_texture_edge_clamp, glewIsSupported("GL_EXT_texture_edge_clamp"), glewGetExtension("GL_EXT_texture_edge_clamp"));
+}
+
+#endif /* GL_EXT_texture_edge_clamp */
+
+#ifdef GL_EXT_texture_env
+
+static void _glewInfo_GL_EXT_texture_env (void)
+{
+  glewPrintExt("GL_EXT_texture_env", GLEW_EXT_texture_env, glewIsSupported("GL_EXT_texture_env"), glewGetExtension("GL_EXT_texture_env"));
+}
+
+#endif /* GL_EXT_texture_env */
+
+#ifdef GL_EXT_texture_env_add
+
+static void _glewInfo_GL_EXT_texture_env_add (void)
+{
+  glewPrintExt("GL_EXT_texture_env_add", GLEW_EXT_texture_env_add, glewIsSupported("GL_EXT_texture_env_add"), glewGetExtension("GL_EXT_texture_env_add"));
+}
+
+#endif /* GL_EXT_texture_env_add */
+
+#ifdef GL_EXT_texture_env_combine
+
+static void _glewInfo_GL_EXT_texture_env_combine (void)
+{
+  glewPrintExt("GL_EXT_texture_env_combine", GLEW_EXT_texture_env_combine, glewIsSupported("GL_EXT_texture_env_combine"), glewGetExtension("GL_EXT_texture_env_combine"));
+}
+
+#endif /* GL_EXT_texture_env_combine */
+
+#ifdef GL_EXT_texture_env_dot3
+
+static void _glewInfo_GL_EXT_texture_env_dot3 (void)
+{
+  glewPrintExt("GL_EXT_texture_env_dot3", GLEW_EXT_texture_env_dot3, glewIsSupported("GL_EXT_texture_env_dot3"), glewGetExtension("GL_EXT_texture_env_dot3"));
+}
+
+#endif /* GL_EXT_texture_env_dot3 */
+
+#ifdef GL_EXT_texture_filter_anisotropic
+
+static void _glewInfo_GL_EXT_texture_filter_anisotropic (void)
+{
+  glewPrintExt("GL_EXT_texture_filter_anisotropic", GLEW_EXT_texture_filter_anisotropic, glewIsSupported("GL_EXT_texture_filter_anisotropic"), glewGetExtension("GL_EXT_texture_filter_anisotropic"));
+}
+
+#endif /* GL_EXT_texture_filter_anisotropic */
+
+#ifdef GL_EXT_texture_integer
+
+static void _glewInfo_GL_EXT_texture_integer (void)
+{
+  glewPrintExt("GL_EXT_texture_integer", GLEW_EXT_texture_integer, glewIsSupported("GL_EXT_texture_integer"), glewGetExtension("GL_EXT_texture_integer"));
+
+  glewInfoFunc("glClearColorIiEXT", glClearColorIiEXT == NULL);
+  glewInfoFunc("glClearColorIuiEXT", glClearColorIuiEXT == NULL);
+  glewInfoFunc("glGetTexParameterIivEXT", glGetTexParameterIivEXT == NULL);
+  glewInfoFunc("glGetTexParameterIuivEXT", glGetTexParameterIuivEXT == NULL);
+  glewInfoFunc("glTexParameterIivEXT", glTexParameterIivEXT == NULL);
+  glewInfoFunc("glTexParameterIuivEXT", glTexParameterIuivEXT == NULL);
+}
+
+#endif /* GL_EXT_texture_integer */
+
+#ifdef GL_EXT_texture_lod_bias
+
+static void _glewInfo_GL_EXT_texture_lod_bias (void)
+{
+  glewPrintExt("GL_EXT_texture_lod_bias", GLEW_EXT_texture_lod_bias, glewIsSupported("GL_EXT_texture_lod_bias"), glewGetExtension("GL_EXT_texture_lod_bias"));
+}
+
+#endif /* GL_EXT_texture_lod_bias */
+
+#ifdef GL_EXT_texture_mirror_clamp
+
+static void _glewInfo_GL_EXT_texture_mirror_clamp (void)
+{
+  glewPrintExt("GL_EXT_texture_mirror_clamp", GLEW_EXT_texture_mirror_clamp, glewIsSupported("GL_EXT_texture_mirror_clamp"), glewGetExtension("GL_EXT_texture_mirror_clamp"));
+}
+
+#endif /* GL_EXT_texture_mirror_clamp */
+
+#ifdef GL_EXT_texture_object
+
+static void _glewInfo_GL_EXT_texture_object (void)
+{
+  glewPrintExt("GL_EXT_texture_object", GLEW_EXT_texture_object, glewIsSupported("GL_EXT_texture_object"), glewGetExtension("GL_EXT_texture_object"));
+
+  glewInfoFunc("glAreTexturesResidentEXT", glAreTexturesResidentEXT == NULL);
+  glewInfoFunc("glBindTextureEXT", glBindTextureEXT == NULL);
+  glewInfoFunc("glDeleteTexturesEXT", glDeleteTexturesEXT == NULL);
+  glewInfoFunc("glGenTexturesEXT", glGenTexturesEXT == NULL);
+  glewInfoFunc("glIsTextureEXT", glIsTextureEXT == NULL);
+  glewInfoFunc("glPrioritizeTexturesEXT", glPrioritizeTexturesEXT == NULL);
+}
+
+#endif /* GL_EXT_texture_object */
+
+#ifdef GL_EXT_texture_perturb_normal
+
+static void _glewInfo_GL_EXT_texture_perturb_normal (void)
+{
+  glewPrintExt("GL_EXT_texture_perturb_normal", GLEW_EXT_texture_perturb_normal, glewIsSupported("GL_EXT_texture_perturb_normal"), glewGetExtension("GL_EXT_texture_perturb_normal"));
+
+  glewInfoFunc("glTextureNormalEXT", glTextureNormalEXT == NULL);
+}
+
+#endif /* GL_EXT_texture_perturb_normal */
+
+#ifdef GL_EXT_texture_rectangle
+
+static void _glewInfo_GL_EXT_texture_rectangle (void)
+{
+  glewPrintExt("GL_EXT_texture_rectangle", GLEW_EXT_texture_rectangle, glewIsSupported("GL_EXT_texture_rectangle"), glewGetExtension("GL_EXT_texture_rectangle"));
+}
+
+#endif /* GL_EXT_texture_rectangle */
+
+#ifdef GL_EXT_texture_sRGB
+
+static void _glewInfo_GL_EXT_texture_sRGB (void)
+{
+  glewPrintExt("GL_EXT_texture_sRGB", GLEW_EXT_texture_sRGB, glewIsSupported("GL_EXT_texture_sRGB"), glewGetExtension("GL_EXT_texture_sRGB"));
+}
+
+#endif /* GL_EXT_texture_sRGB */
+
+#ifdef GL_EXT_texture_shared_exponent
+
+static void _glewInfo_GL_EXT_texture_shared_exponent (void)
+{
+  glewPrintExt("GL_EXT_texture_shared_exponent", GLEW_EXT_texture_shared_exponent, glewIsSupported("GL_EXT_texture_shared_exponent"), glewGetExtension("GL_EXT_texture_shared_exponent"));
+}
+
+#endif /* GL_EXT_texture_shared_exponent */
+
+#ifdef GL_EXT_texture_swizzle
+
+static void _glewInfo_GL_EXT_texture_swizzle (void)
+{
+  glewPrintExt("GL_EXT_texture_swizzle", GLEW_EXT_texture_swizzle, glewIsSupported("GL_EXT_texture_swizzle"), glewGetExtension("GL_EXT_texture_swizzle"));
+}
+
+#endif /* GL_EXT_texture_swizzle */
+
+#ifdef GL_EXT_timer_query
+
+static void _glewInfo_GL_EXT_timer_query (void)
+{
+  glewPrintExt("GL_EXT_timer_query", GLEW_EXT_timer_query, glewIsSupported("GL_EXT_timer_query"), glewGetExtension("GL_EXT_timer_query"));
+
+  glewInfoFunc("glGetQueryObjecti64vEXT", glGetQueryObjecti64vEXT == NULL);
+  glewInfoFunc("glGetQueryObjectui64vEXT", glGetQueryObjectui64vEXT == NULL);
+}
+
+#endif /* GL_EXT_timer_query */
+
+#ifdef GL_EXT_transform_feedback
+
+static void _glewInfo_GL_EXT_transform_feedback (void)
+{
+  glewPrintExt("GL_EXT_transform_feedback", GLEW_EXT_transform_feedback, glewIsSupported("GL_EXT_transform_feedback"), glewGetExtension("GL_EXT_transform_feedback"));
+
+  glewInfoFunc("glBeginTransformFeedbackEXT", glBeginTransformFeedbackEXT == NULL);
+  glewInfoFunc("glBindBufferBaseEXT", glBindBufferBaseEXT == NULL);
+  glewInfoFunc("glBindBufferOffsetEXT", glBindBufferOffsetEXT == NULL);
+  glewInfoFunc("glBindBufferRangeEXT", glBindBufferRangeEXT == NULL);
+  glewInfoFunc("glEndTransformFeedbackEXT", glEndTransformFeedbackEXT == NULL);
+  glewInfoFunc("glGetTransformFeedbackVaryingEXT", glGetTransformFeedbackVaryingEXT == NULL);
+  glewInfoFunc("glTransformFeedbackVaryingsEXT", glTransformFeedbackVaryingsEXT == NULL);
+}
+
+#endif /* GL_EXT_transform_feedback */
+
+#ifdef GL_EXT_vertex_array
+
+static void _glewInfo_GL_EXT_vertex_array (void)
+{
+  glewPrintExt("GL_EXT_vertex_array", GLEW_EXT_vertex_array, glewIsSupported("GL_EXT_vertex_array"), glewGetExtension("GL_EXT_vertex_array"));
+
+  glewInfoFunc("glArrayElementEXT", glArrayElementEXT == NULL);
+  glewInfoFunc("glColorPointerEXT", glColorPointerEXT == NULL);
+  glewInfoFunc("glDrawArraysEXT", glDrawArraysEXT == NULL);
+  glewInfoFunc("glEdgeFlagPointerEXT", glEdgeFlagPointerEXT == NULL);
+  glewInfoFunc("glGetPointervEXT", glGetPointervEXT == NULL);
+  glewInfoFunc("glIndexPointerEXT", glIndexPointerEXT == NULL);
+  glewInfoFunc("glNormalPointerEXT", glNormalPointerEXT == NULL);
+  glewInfoFunc("glTexCoordPointerEXT", glTexCoordPointerEXT == NULL);
+  glewInfoFunc("glVertexPointerEXT", glVertexPointerEXT == NULL);
+}
+
+#endif /* GL_EXT_vertex_array */
+
+#ifdef GL_EXT_vertex_array_bgra
+
+static void _glewInfo_GL_EXT_vertex_array_bgra (void)
+{
+  glewPrintExt("GL_EXT_vertex_array_bgra", GLEW_EXT_vertex_array_bgra, glewIsSupported("GL_EXT_vertex_array_bgra"), glewGetExtension("GL_EXT_vertex_array_bgra"));
+}
+
+#endif /* GL_EXT_vertex_array_bgra */
+
+#ifdef GL_EXT_vertex_shader
+
+static void _glewInfo_GL_EXT_vertex_shader (void)
+{
+  glewPrintExt("GL_EXT_vertex_shader", GLEW_EXT_vertex_shader, glewIsSupported("GL_EXT_vertex_shader"), glewGetExtension("GL_EXT_vertex_shader"));
+
+  glewInfoFunc("glBeginVertexShaderEXT", glBeginVertexShaderEXT == NULL);
+  glewInfoFunc("glBindLightParameterEXT", glBindLightParameterEXT == NULL);
+  glewInfoFunc("glBindMaterialParameterEXT", glBindMaterialParameterEXT == NULL);
+  glewInfoFunc("glBindParameterEXT", glBindParameterEXT == NULL);
+  glewInfoFunc("glBindTexGenParameterEXT", glBindTexGenParameterEXT == NULL);
+  glewInfoFunc("glBindTextureUnitParameterEXT", glBindTextureUnitParameterEXT == NULL);
+  glewInfoFunc("glBindVertexShaderEXT", glBindVertexShaderEXT == NULL);
+  glewInfoFunc("glDeleteVertexShaderEXT", glDeleteVertexShaderEXT == NULL);
+  glewInfoFunc("glDisableVariantClientStateEXT", glDisableVariantClientStateEXT == NULL);
+  glewInfoFunc("glEnableVariantClientStateEXT", glEnableVariantClientStateEXT == NULL);
+  glewInfoFunc("glEndVertexShaderEXT", glEndVertexShaderEXT == NULL);
+  glewInfoFunc("glExtractComponentEXT", glExtractComponentEXT == NULL);
+  glewInfoFunc("glGenSymbolsEXT", glGenSymbolsEXT == NULL);
+  glewInfoFunc("glGenVertexShadersEXT", glGenVertexShadersEXT == NULL);
+  glewInfoFunc("glGetInvariantBooleanvEXT", glGetInvariantBooleanvEXT == NULL);
+  glewInfoFunc("glGetInvariantFloatvEXT", glGetInvariantFloatvEXT == NULL);
+  glewInfoFunc("glGetInvariantIntegervEXT", glGetInvariantIntegervEXT == NULL);
+  glewInfoFunc("glGetLocalConstantBooleanvEXT", glGetLocalConstantBooleanvEXT == NULL);
+  glewInfoFunc("glGetLocalConstantFloatvEXT", glGetLocalConstantFloatvEXT == NULL);
+  glewInfoFunc("glGetLocalConstantIntegervEXT", glGetLocalConstantIntegervEXT == NULL);
+  glewInfoFunc("glGetVariantBooleanvEXT", glGetVariantBooleanvEXT == NULL);
+  glewInfoFunc("glGetVariantFloatvEXT", glGetVariantFloatvEXT == NULL);
+  glewInfoFunc("glGetVariantIntegervEXT", glGetVariantIntegervEXT == NULL);
+  glewInfoFunc("glGetVariantPointervEXT", glGetVariantPointervEXT == NULL);
+  glewInfoFunc("glInsertComponentEXT", glInsertComponentEXT == NULL);
+  glewInfoFunc("glIsVariantEnabledEXT", glIsVariantEnabledEXT == NULL);
+  glewInfoFunc("glSetInvariantEXT", glSetInvariantEXT == NULL);
+  glewInfoFunc("glSetLocalConstantEXT", glSetLocalConstantEXT == NULL);
+  glewInfoFunc("glShaderOp1EXT", glShaderOp1EXT == NULL);
+  glewInfoFunc("glShaderOp2EXT", glShaderOp2EXT == NULL);
+  glewInfoFunc("glShaderOp3EXT", glShaderOp3EXT == NULL);
+  glewInfoFunc("glSwizzleEXT", glSwizzleEXT == NULL);
+  glewInfoFunc("glVariantPointerEXT", glVariantPointerEXT == NULL);
+  glewInfoFunc("glVariantbvEXT", glVariantbvEXT == NULL);
+  glewInfoFunc("glVariantdvEXT", glVariantdvEXT == NULL);
+  glewInfoFunc("glVariantfvEXT", glVariantfvEXT == NULL);
+  glewInfoFunc("glVariantivEXT", glVariantivEXT == NULL);
+  glewInfoFunc("glVariantsvEXT", glVariantsvEXT == NULL);
+  glewInfoFunc("glVariantubvEXT", glVariantubvEXT == NULL);
+  glewInfoFunc("glVariantuivEXT", glVariantuivEXT == NULL);
+  glewInfoFunc("glVariantusvEXT", glVariantusvEXT == NULL);
+  glewInfoFunc("glWriteMaskEXT", glWriteMaskEXT == NULL);
+}
+
+#endif /* GL_EXT_vertex_shader */
+
+#ifdef GL_EXT_vertex_weighting
+
+static void _glewInfo_GL_EXT_vertex_weighting (void)
+{
+  glewPrintExt("GL_EXT_vertex_weighting", GLEW_EXT_vertex_weighting, glewIsSupported("GL_EXT_vertex_weighting"), glewGetExtension("GL_EXT_vertex_weighting"));
+
+  glewInfoFunc("glVertexWeightPointerEXT", glVertexWeightPointerEXT == NULL);
+  glewInfoFunc("glVertexWeightfEXT", glVertexWeightfEXT == NULL);
+  glewInfoFunc("glVertexWeightfvEXT", glVertexWeightfvEXT == NULL);
+}
+
+#endif /* GL_EXT_vertex_weighting */
+
+#ifdef GL_GREMEDY_frame_terminator
+
+static void _glewInfo_GL_GREMEDY_frame_terminator (void)
+{
+  glewPrintExt("GL_GREMEDY_frame_terminator", GLEW_GREMEDY_frame_terminator, glewIsSupported("GL_GREMEDY_frame_terminator"), glewGetExtension("GL_GREMEDY_frame_terminator"));
+
+  glewInfoFunc("glFrameTerminatorGREMEDY", glFrameTerminatorGREMEDY == NULL);
+}
+
+#endif /* GL_GREMEDY_frame_terminator */
+
+#ifdef GL_GREMEDY_string_marker
+
+static void _glewInfo_GL_GREMEDY_string_marker (void)
+{
+  glewPrintExt("GL_GREMEDY_string_marker", GLEW_GREMEDY_string_marker, glewIsSupported("GL_GREMEDY_string_marker"), glewGetExtension("GL_GREMEDY_string_marker"));
+
+  glewInfoFunc("glStringMarkerGREMEDY", glStringMarkerGREMEDY == NULL);
+}
+
+#endif /* GL_GREMEDY_string_marker */
+
+#ifdef GL_HP_convolution_border_modes
+
+static void _glewInfo_GL_HP_convolution_border_modes (void)
+{
+  glewPrintExt("GL_HP_convolution_border_modes", GLEW_HP_convolution_border_modes, glewIsSupported("GL_HP_convolution_border_modes"), glewGetExtension("GL_HP_convolution_border_modes"));
+}
+
+#endif /* GL_HP_convolution_border_modes */
+
+#ifdef GL_HP_image_transform
+
+static void _glewInfo_GL_HP_image_transform (void)
+{
+  glewPrintExt("GL_HP_image_transform", GLEW_HP_image_transform, glewIsSupported("GL_HP_image_transform"), glewGetExtension("GL_HP_image_transform"));
+
+  glewInfoFunc("glGetImageTransformParameterfvHP", glGetImageTransformParameterfvHP == NULL);
+  glewInfoFunc("glGetImageTransformParameterivHP", glGetImageTransformParameterivHP == NULL);
+  glewInfoFunc("glImageTransformParameterfHP", glImageTransformParameterfHP == NULL);
+  glewInfoFunc("glImageTransformParameterfvHP", glImageTransformParameterfvHP == NULL);
+  glewInfoFunc("glImageTransformParameteriHP", glImageTransformParameteriHP == NULL);
+  glewInfoFunc("glImageTransformParameterivHP", glImageTransformParameterivHP == NULL);
+}
+
+#endif /* GL_HP_image_transform */
+
+#ifdef GL_HP_occlusion_test
+
+static void _glewInfo_GL_HP_occlusion_test (void)
+{
+  glewPrintExt("GL_HP_occlusion_test", GLEW_HP_occlusion_test, glewIsSupported("GL_HP_occlusion_test"), glewGetExtension("GL_HP_occlusion_test"));
+}
+
+#endif /* GL_HP_occlusion_test */
+
+#ifdef GL_HP_texture_lighting
+
+static void _glewInfo_GL_HP_texture_lighting (void)
+{
+  glewPrintExt("GL_HP_texture_lighting", GLEW_HP_texture_lighting, glewIsSupported("GL_HP_texture_lighting"), glewGetExtension("GL_HP_texture_lighting"));
+}
+
+#endif /* GL_HP_texture_lighting */
+
+#ifdef GL_IBM_cull_vertex
+
+static void _glewInfo_GL_IBM_cull_vertex (void)
+{
+  glewPrintExt("GL_IBM_cull_vertex", GLEW_IBM_cull_vertex, glewIsSupported("GL_IBM_cull_vertex"), glewGetExtension("GL_IBM_cull_vertex"));
+}
+
+#endif /* GL_IBM_cull_vertex */
+
+#ifdef GL_IBM_multimode_draw_arrays
+
+static void _glewInfo_GL_IBM_multimode_draw_arrays (void)
+{
+  glewPrintExt("GL_IBM_multimode_draw_arrays", GLEW_IBM_multimode_draw_arrays, glewIsSupported("GL_IBM_multimode_draw_arrays"), glewGetExtension("GL_IBM_multimode_draw_arrays"));
+
+  glewInfoFunc("glMultiModeDrawArraysIBM", glMultiModeDrawArraysIBM == NULL);
+  glewInfoFunc("glMultiModeDrawElementsIBM", glMultiModeDrawElementsIBM == NULL);
+}
+
+#endif /* GL_IBM_multimode_draw_arrays */
+
+#ifdef GL_IBM_rasterpos_clip
+
+static void _glewInfo_GL_IBM_rasterpos_clip (void)
+{
+  glewPrintExt("GL_IBM_rasterpos_clip", GLEW_IBM_rasterpos_clip, glewIsSupported("GL_IBM_rasterpos_clip"), glewGetExtension("GL_IBM_rasterpos_clip"));
+}
+
+#endif /* GL_IBM_rasterpos_clip */
+
+#ifdef GL_IBM_static_data
+
+static void _glewInfo_GL_IBM_static_data (void)
+{
+  glewPrintExt("GL_IBM_static_data", GLEW_IBM_static_data, glewIsSupported("GL_IBM_static_data"), glewGetExtension("GL_IBM_static_data"));
+}
+
+#endif /* GL_IBM_static_data */
+
+#ifdef GL_IBM_texture_mirrored_repeat
+
+static void _glewInfo_GL_IBM_texture_mirrored_repeat (void)
+{
+  glewPrintExt("GL_IBM_texture_mirrored_repeat", GLEW_IBM_texture_mirrored_repeat, glewIsSupported("GL_IBM_texture_mirrored_repeat"), glewGetExtension("GL_IBM_texture_mirrored_repeat"));
+}
+
+#endif /* GL_IBM_texture_mirrored_repeat */
+
+#ifdef GL_IBM_vertex_array_lists
+
+static void _glewInfo_GL_IBM_vertex_array_lists (void)
+{
+  glewPrintExt("GL_IBM_vertex_array_lists", GLEW_IBM_vertex_array_lists, glewIsSupported("GL_IBM_vertex_array_lists"), glewGetExtension("GL_IBM_vertex_array_lists"));
+
+  glewInfoFunc("glColorPointerListIBM", glColorPointerListIBM == NULL);
+  glewInfoFunc("glEdgeFlagPointerListIBM", glEdgeFlagPointerListIBM == NULL);
+  glewInfoFunc("glFogCoordPointerListIBM", glFogCoordPointerListIBM == NULL);
+  glewInfoFunc("glIndexPointerListIBM", glIndexPointerListIBM == NULL);
+  glewInfoFunc("glNormalPointerListIBM", glNormalPointerListIBM == NULL);
+  glewInfoFunc("glSecondaryColorPointerListIBM", glSecondaryColorPointerListIBM == NULL);
+  glewInfoFunc("glTexCoordPointerListIBM", glTexCoordPointerListIBM == NULL);
+  glewInfoFunc("glVertexPointerListIBM", glVertexPointerListIBM == NULL);
+}
+
+#endif /* GL_IBM_vertex_array_lists */
+
+#ifdef GL_INGR_color_clamp
+
+static void _glewInfo_GL_INGR_color_clamp (void)
+{
+  glewPrintExt("GL_INGR_color_clamp", GLEW_INGR_color_clamp, glewIsSupported("GL_INGR_color_clamp"), glewGetExtension("GL_INGR_color_clamp"));
+}
+
+#endif /* GL_INGR_color_clamp */
+
+#ifdef GL_INGR_interlace_read
+
+static void _glewInfo_GL_INGR_interlace_read (void)
+{
+  glewPrintExt("GL_INGR_interlace_read", GLEW_INGR_interlace_read, glewIsSupported("GL_INGR_interlace_read"), glewGetExtension("GL_INGR_interlace_read"));
+}
+
+#endif /* GL_INGR_interlace_read */
+
+#ifdef GL_INTEL_parallel_arrays
+
+static void _glewInfo_GL_INTEL_parallel_arrays (void)
+{
+  glewPrintExt("GL_INTEL_parallel_arrays", GLEW_INTEL_parallel_arrays, glewIsSupported("GL_INTEL_parallel_arrays"), glewGetExtension("GL_INTEL_parallel_arrays"));
+
+  glewInfoFunc("glColorPointervINTEL", glColorPointervINTEL == NULL);
+  glewInfoFunc("glNormalPointervINTEL", glNormalPointervINTEL == NULL);
+  glewInfoFunc("glTexCoordPointervINTEL", glTexCoordPointervINTEL == NULL);
+  glewInfoFunc("glVertexPointervINTEL", glVertexPointervINTEL == NULL);
+}
+
+#endif /* GL_INTEL_parallel_arrays */
+
+#ifdef GL_INTEL_texture_scissor
+
+static void _glewInfo_GL_INTEL_texture_scissor (void)
+{
+  glewPrintExt("GL_INTEL_texture_scissor", GLEW_INTEL_texture_scissor, glewIsSupported("GL_INTEL_texture_scissor"), glewGetExtension("GL_INTEL_texture_scissor"));
+
+  glewInfoFunc("glTexScissorFuncINTEL", glTexScissorFuncINTEL == NULL);
+  glewInfoFunc("glTexScissorINTEL", glTexScissorINTEL == NULL);
+}
+
+#endif /* GL_INTEL_texture_scissor */
+
+#ifdef GL_KTX_buffer_region
+
+static void _glewInfo_GL_KTX_buffer_region (void)
+{
+  glewPrintExt("GL_KTX_buffer_region", GLEW_KTX_buffer_region, glewIsSupported("GL_KTX_buffer_region"), glewGetExtension("GL_KTX_buffer_region"));
+
+  glewInfoFunc("glBufferRegionEnabledEXT", glBufferRegionEnabledEXT == NULL);
+  glewInfoFunc("glDeleteBufferRegionEXT", glDeleteBufferRegionEXT == NULL);
+  glewInfoFunc("glDrawBufferRegionEXT", glDrawBufferRegionEXT == NULL);
+  glewInfoFunc("glNewBufferRegionEXT", glNewBufferRegionEXT == NULL);
+  glewInfoFunc("glReadBufferRegionEXT", glReadBufferRegionEXT == NULL);
+}
+
+#endif /* GL_KTX_buffer_region */
+
+#ifdef GL_MESAX_texture_stack
+
+static void _glewInfo_GL_MESAX_texture_stack (void)
+{
+  glewPrintExt("GL_MESAX_texture_stack", GLEW_MESAX_texture_stack, glewIsSupported("GL_MESAX_texture_stack"), glewGetExtension("GL_MESAX_texture_stack"));
+}
+
+#endif /* GL_MESAX_texture_stack */
+
+#ifdef GL_MESA_pack_invert
+
+static void _glewInfo_GL_MESA_pack_invert (void)
+{
+  glewPrintExt("GL_MESA_pack_invert", GLEW_MESA_pack_invert, glewIsSupported("GL_MESA_pack_invert"), glewGetExtension("GL_MESA_pack_invert"));
+}
+
+#endif /* GL_MESA_pack_invert */
+
+#ifdef GL_MESA_resize_buffers
+
+static void _glewInfo_GL_MESA_resize_buffers (void)
+{
+  glewPrintExt("GL_MESA_resize_buffers", GLEW_MESA_resize_buffers, glewIsSupported("GL_MESA_resize_buffers"), glewGetExtension("GL_MESA_resize_buffers"));
+
+  glewInfoFunc("glResizeBuffersMESA", glResizeBuffersMESA == NULL);
+}
+
+#endif /* GL_MESA_resize_buffers */
+
+#ifdef GL_MESA_window_pos
+
+static void _glewInfo_GL_MESA_window_pos (void)
+{
+  glewPrintExt("GL_MESA_window_pos", GLEW_MESA_window_pos, glewIsSupported("GL_MESA_window_pos"), glewGetExtension("GL_MESA_window_pos"));
+
+  glewInfoFunc("glWindowPos2dMESA", glWindowPos2dMESA == NULL);
+  glewInfoFunc("glWindowPos2dvMESA", glWindowPos2dvMESA == NULL);
+  glewInfoFunc("glWindowPos2fMESA", glWindowPos2fMESA == NULL);
+  glewInfoFunc("glWindowPos2fvMESA", glWindowPos2fvMESA == NULL);
+  glewInfoFunc("glWindowPos2iMESA", glWindowPos2iMESA == NULL);
+  glewInfoFunc("glWindowPos2ivMESA", glWindowPos2ivMESA == NULL);
+  glewInfoFunc("glWindowPos2sMESA", glWindowPos2sMESA == NULL);
+  glewInfoFunc("glWindowPos2svMESA", glWindowPos2svMESA == NULL);
+  glewInfoFunc("glWindowPos3dMESA", glWindowPos3dMESA == NULL);
+  glewInfoFunc("glWindowPos3dvMESA", glWindowPos3dvMESA == NULL);
+  glewInfoFunc("glWindowPos3fMESA", glWindowPos3fMESA == NULL);
+  glewInfoFunc("glWindowPos3fvMESA", glWindowPos3fvMESA == NULL);
+  glewInfoFunc("glWindowPos3iMESA", glWindowPos3iMESA == NULL);
+  glewInfoFunc("glWindowPos3ivMESA", glWindowPos3ivMESA == NULL);
+  glewInfoFunc("glWindowPos3sMESA", glWindowPos3sMESA == NULL);
+  glewInfoFunc("glWindowPos3svMESA", glWindowPos3svMESA == NULL);
+  glewInfoFunc("glWindowPos4dMESA", glWindowPos4dMESA == NULL);
+  glewInfoFunc("glWindowPos4dvMESA", glWindowPos4dvMESA == NULL);
+  glewInfoFunc("glWindowPos4fMESA", glWindowPos4fMESA == NULL);
+  glewInfoFunc("glWindowPos4fvMESA", glWindowPos4fvMESA == NULL);
+  glewInfoFunc("glWindowPos4iMESA", glWindowPos4iMESA == NULL);
+  glewInfoFunc("glWindowPos4ivMESA", glWindowPos4ivMESA == NULL);
+  glewInfoFunc("glWindowPos4sMESA", glWindowPos4sMESA == NULL);
+  glewInfoFunc("glWindowPos4svMESA", glWindowPos4svMESA == NULL);
+}
+
+#endif /* GL_MESA_window_pos */
+
+#ifdef GL_MESA_ycbcr_texture
+
+static void _glewInfo_GL_MESA_ycbcr_texture (void)
+{
+  glewPrintExt("GL_MESA_ycbcr_texture", GLEW_MESA_ycbcr_texture, glewIsSupported("GL_MESA_ycbcr_texture"), glewGetExtension("GL_MESA_ycbcr_texture"));
+}
+
+#endif /* GL_MESA_ycbcr_texture */
+
+#ifdef GL_NV_blend_square
+
+static void _glewInfo_GL_NV_blend_square (void)
+{
+  glewPrintExt("GL_NV_blend_square", GLEW_NV_blend_square, glewIsSupported("GL_NV_blend_square"), glewGetExtension("GL_NV_blend_square"));
+}
+
+#endif /* GL_NV_blend_square */
+
+#ifdef GL_NV_conditional_render
+
+static void _glewInfo_GL_NV_conditional_render (void)
+{
+  glewPrintExt("GL_NV_conditional_render", GLEW_NV_conditional_render, glewIsSupported("GL_NV_conditional_render"), glewGetExtension("GL_NV_conditional_render"));
+
+  glewInfoFunc("glBeginConditionalRenderNV", glBeginConditionalRenderNV == NULL);
+  glewInfoFunc("glEndConditionalRenderNV", glEndConditionalRenderNV == NULL);
+}
+
+#endif /* GL_NV_conditional_render */
+
+#ifdef GL_NV_copy_depth_to_color
+
+static void _glewInfo_GL_NV_copy_depth_to_color (void)
+{
+  glewPrintExt("GL_NV_copy_depth_to_color", GLEW_NV_copy_depth_to_color, glewIsSupported("GL_NV_copy_depth_to_color"), glewGetExtension("GL_NV_copy_depth_to_color"));
+}
+
+#endif /* GL_NV_copy_depth_to_color */
+
+#ifdef GL_NV_depth_buffer_float
+
+static void _glewInfo_GL_NV_depth_buffer_float (void)
+{
+  glewPrintExt("GL_NV_depth_buffer_float", GLEW_NV_depth_buffer_float, glewIsSupported("GL_NV_depth_buffer_float"), glewGetExtension("GL_NV_depth_buffer_float"));
+
+  glewInfoFunc("glClearDepthdNV", glClearDepthdNV == NULL);
+  glewInfoFunc("glDepthBoundsdNV", glDepthBoundsdNV == NULL);
+  glewInfoFunc("glDepthRangedNV", glDepthRangedNV == NULL);
+}
+
+#endif /* GL_NV_depth_buffer_float */
+
+#ifdef GL_NV_depth_clamp
+
+static void _glewInfo_GL_NV_depth_clamp (void)
+{
+  glewPrintExt("GL_NV_depth_clamp", GLEW_NV_depth_clamp, glewIsSupported("GL_NV_depth_clamp"), glewGetExtension("GL_NV_depth_clamp"));
+}
+
+#endif /* GL_NV_depth_clamp */
+
+#ifdef GL_NV_depth_range_unclamped
+
+static void _glewInfo_GL_NV_depth_range_unclamped (void)
+{
+  glewPrintExt("GL_NV_depth_range_unclamped", GLEW_NV_depth_range_unclamped, glewIsSupported("GL_NV_depth_range_unclamped"), glewGetExtension("GL_NV_depth_range_unclamped"));
+}
+
+#endif /* GL_NV_depth_range_unclamped */
+
+#ifdef GL_NV_evaluators
+
+static void _glewInfo_GL_NV_evaluators (void)
+{
+  glewPrintExt("GL_NV_evaluators", GLEW_NV_evaluators, glewIsSupported("GL_NV_evaluators"), glewGetExtension("GL_NV_evaluators"));
+
+  glewInfoFunc("glEvalMapsNV", glEvalMapsNV == NULL);
+  glewInfoFunc("glGetMapAttribParameterfvNV", glGetMapAttribParameterfvNV == NULL);
+  glewInfoFunc("glGetMapAttribParameterivNV", glGetMapAttribParameterivNV == NULL);
+  glewInfoFunc("glGetMapControlPointsNV", glGetMapControlPointsNV == NULL);
+  glewInfoFunc("glGetMapParameterfvNV", glGetMapParameterfvNV == NULL);
+  glewInfoFunc("glGetMapParameterivNV", glGetMapParameterivNV == NULL);
+  glewInfoFunc("glMapControlPointsNV", glMapControlPointsNV == NULL);
+  glewInfoFunc("glMapParameterfvNV", glMapParameterfvNV == NULL);
+  glewInfoFunc("glMapParameterivNV", glMapParameterivNV == NULL);
+}
+
+#endif /* GL_NV_evaluators */
+
+#ifdef GL_NV_explicit_multisample
+
+static void _glewInfo_GL_NV_explicit_multisample (void)
+{
+  glewPrintExt("GL_NV_explicit_multisample", GLEW_NV_explicit_multisample, glewIsSupported("GL_NV_explicit_multisample"), glewGetExtension("GL_NV_explicit_multisample"));
+
+  glewInfoFunc("glGetMultisamplefvNV", glGetMultisamplefvNV == NULL);
+  glewInfoFunc("glSampleMaskIndexedNV", glSampleMaskIndexedNV == NULL);
+  glewInfoFunc("glTexRenderbufferNV", glTexRenderbufferNV == NULL);
+}
+
+#endif /* GL_NV_explicit_multisample */
+
+#ifdef GL_NV_fence
+
+static void _glewInfo_GL_NV_fence (void)
+{
+  glewPrintExt("GL_NV_fence", GLEW_NV_fence, glewIsSupported("GL_NV_fence"), glewGetExtension("GL_NV_fence"));
+
+  glewInfoFunc("glDeleteFencesNV", glDeleteFencesNV == NULL);
+  glewInfoFunc("glFinishFenceNV", glFinishFenceNV == NULL);
+  glewInfoFunc("glGenFencesNV", glGenFencesNV == NULL);
+  glewInfoFunc("glGetFenceivNV", glGetFenceivNV == NULL);
+  glewInfoFunc("glIsFenceNV", glIsFenceNV == NULL);
+  glewInfoFunc("glSetFenceNV", glSetFenceNV == NULL);
+  glewInfoFunc("glTestFenceNV", glTestFenceNV == NULL);
+}
+
+#endif /* GL_NV_fence */
+
+#ifdef GL_NV_float_buffer
+
+static void _glewInfo_GL_NV_float_buffer (void)
+{
+  glewPrintExt("GL_NV_float_buffer", GLEW_NV_float_buffer, glewIsSupported("GL_NV_float_buffer"), glewGetExtension("GL_NV_float_buffer"));
+}
+
+#endif /* GL_NV_float_buffer */
+
+#ifdef GL_NV_fog_distance
+
+static void _glewInfo_GL_NV_fog_distance (void)
+{
+  glewPrintExt("GL_NV_fog_distance", GLEW_NV_fog_distance, glewIsSupported("GL_NV_fog_distance"), glewGetExtension("GL_NV_fog_distance"));
+}
+
+#endif /* GL_NV_fog_distance */
+
+#ifdef GL_NV_fragment_program
+
+static void _glewInfo_GL_NV_fragment_program (void)
+{
+  glewPrintExt("GL_NV_fragment_program", GLEW_NV_fragment_program, glewIsSupported("GL_NV_fragment_program"), glewGetExtension("GL_NV_fragment_program"));
+
+  glewInfoFunc("glGetProgramNamedParameterdvNV", glGetProgramNamedParameterdvNV == NULL);
+  glewInfoFunc("glGetProgramNamedParameterfvNV", glGetProgramNamedParameterfvNV == NULL);
+  glewInfoFunc("glProgramNamedParameter4dNV", glProgramNamedParameter4dNV == NULL);
+  glewInfoFunc("glProgramNamedParameter4dvNV", glProgramNamedParameter4dvNV == NULL);
+  glewInfoFunc("glProgramNamedParameter4fNV", glProgramNamedParameter4fNV == NULL);
+  glewInfoFunc("glProgramNamedParameter4fvNV", glProgramNamedParameter4fvNV == NULL);
+}
+
+#endif /* GL_NV_fragment_program */
+
+#ifdef GL_NV_fragment_program2
+
+static void _glewInfo_GL_NV_fragment_program2 (void)
+{
+  glewPrintExt("GL_NV_fragment_program2", GLEW_NV_fragment_program2, glewIsSupported("GL_NV_fragment_program2"), glewGetExtension("GL_NV_fragment_program2"));
+}
+
+#endif /* GL_NV_fragment_program2 */
+
+#ifdef GL_NV_fragment_program4
+
+static void _glewInfo_GL_NV_fragment_program4 (void)
+{
+  glewPrintExt("GL_NV_fragment_program4", GLEW_NV_fragment_program4, glewIsSupported("GL_NV_fragment_program4"), glewGetExtension("GL_NV_fragment_program4"));
+}
+
+#endif /* GL_NV_fragment_program4 */
+
+#ifdef GL_NV_fragment_program_option
+
+static void _glewInfo_GL_NV_fragment_program_option (void)
+{
+  glewPrintExt("GL_NV_fragment_program_option", GLEW_NV_fragment_program_option, glewIsSupported("GL_NV_fragment_program_option"), glewGetExtension("GL_NV_fragment_program_option"));
+}
+
+#endif /* GL_NV_fragment_program_option */
+
+#ifdef GL_NV_framebuffer_multisample_coverage
+
+static void _glewInfo_GL_NV_framebuffer_multisample_coverage (void)
+{
+  glewPrintExt("GL_NV_framebuffer_multisample_coverage", GLEW_NV_framebuffer_multisample_coverage, glewIsSupported("GL_NV_framebuffer_multisample_coverage"), glewGetExtension("GL_NV_framebuffer_multisample_coverage"));
+
+  glewInfoFunc("glRenderbufferStorageMultisampleCoverageNV", glRenderbufferStorageMultisampleCoverageNV == NULL);
+}
+
+#endif /* GL_NV_framebuffer_multisample_coverage */
+
+#ifdef GL_NV_geometry_program4
+
+static void _glewInfo_GL_NV_geometry_program4 (void)
+{
+  glewPrintExt("GL_NV_geometry_program4", GLEW_NV_geometry_program4, glewIsSupported("GL_NV_geometry_program4"), glewGetExtension("GL_NV_geometry_program4"));
+
+  glewInfoFunc("glProgramVertexLimitNV", glProgramVertexLimitNV == NULL);
+}
+
+#endif /* GL_NV_geometry_program4 */
+
+#ifdef GL_NV_geometry_shader4
+
+static void _glewInfo_GL_NV_geometry_shader4 (void)
+{
+  glewPrintExt("GL_NV_geometry_shader4", GLEW_NV_geometry_shader4, glewIsSupported("GL_NV_geometry_shader4"), glewGetExtension("GL_NV_geometry_shader4"));
+}
+
+#endif /* GL_NV_geometry_shader4 */
+
+#ifdef GL_NV_gpu_program4
+
+static void _glewInfo_GL_NV_gpu_program4 (void)
+{
+  glewPrintExt("GL_NV_gpu_program4", GLEW_NV_gpu_program4, glewIsSupported("GL_NV_gpu_program4"), glewGetExtension("GL_NV_gpu_program4"));
+
+  glewInfoFunc("glProgramEnvParameterI4iNV", glProgramEnvParameterI4iNV == NULL);
+  glewInfoFunc("glProgramEnvParameterI4ivNV", glProgramEnvParameterI4ivNV == NULL);
+  glewInfoFunc("glProgramEnvParameterI4uiNV", glProgramEnvParameterI4uiNV == NULL);
+  glewInfoFunc("glProgramEnvParameterI4uivNV", glProgramEnvParameterI4uivNV == NULL);
+  glewInfoFunc("glProgramEnvParametersI4ivNV", glProgramEnvParametersI4ivNV == NULL);
+  glewInfoFunc("glProgramEnvParametersI4uivNV", glProgramEnvParametersI4uivNV == NULL);
+  glewInfoFunc("glProgramLocalParameterI4iNV", glProgramLocalParameterI4iNV == NULL);
+  glewInfoFunc("glProgramLocalParameterI4ivNV", glProgramLocalParameterI4ivNV == NULL);
+  glewInfoFunc("glProgramLocalParameterI4uiNV", glProgramLocalParameterI4uiNV == NULL);
+  glewInfoFunc("glProgramLocalParameterI4uivNV", glProgramLocalParameterI4uivNV == NULL);
+  glewInfoFunc("glProgramLocalParametersI4ivNV", glProgramLocalParametersI4ivNV == NULL);
+  glewInfoFunc("glProgramLocalParametersI4uivNV", glProgramLocalParametersI4uivNV == NULL);
+}
+
+#endif /* GL_NV_gpu_program4 */
+
+#ifdef GL_NV_half_float
+
+static void _glewInfo_GL_NV_half_float (void)
+{
+  glewPrintExt("GL_NV_half_float", GLEW_NV_half_float, glewIsSupported("GL_NV_half_float"), glewGetExtension("GL_NV_half_float"));
+
+  glewInfoFunc("glColor3hNV", glColor3hNV == NULL);
+  glewInfoFunc("glColor3hvNV", glColor3hvNV == NULL);
+  glewInfoFunc("glColor4hNV", glColor4hNV == NULL);
+  glewInfoFunc("glColor4hvNV", glColor4hvNV == NULL);
+  glewInfoFunc("glFogCoordhNV", glFogCoordhNV == NULL);
+  glewInfoFunc("glFogCoordhvNV", glFogCoordhvNV == NULL);
+  glewInfoFunc("glMultiTexCoord1hNV", glMultiTexCoord1hNV == NULL);
+  glewInfoFunc("glMultiTexCoord1hvNV", glMultiTexCoord1hvNV == NULL);
+  glewInfoFunc("glMultiTexCoord2hNV", glMultiTexCoord2hNV == NULL);
+  glewInfoFunc("glMultiTexCoord2hvNV", glMultiTexCoord2hvNV == NULL);
+  glewInfoFunc("glMultiTexCoord3hNV", glMultiTexCoord3hNV == NULL);
+  glewInfoFunc("glMultiTexCoord3hvNV", glMultiTexCoord3hvNV == NULL);
+  glewInfoFunc("glMultiTexCoord4hNV", glMultiTexCoord4hNV == NULL);
+  glewInfoFunc("glMultiTexCoord4hvNV", glMultiTexCoord4hvNV == NULL);
+  glewInfoFunc("glNormal3hNV", glNormal3hNV == NULL);
+  glewInfoFunc("glNormal3hvNV", glNormal3hvNV == NULL);
+  glewInfoFunc("glSecondaryColor3hNV", glSecondaryColor3hNV == NULL);
+  glewInfoFunc("glSecondaryColor3hvNV", glSecondaryColor3hvNV == NULL);
+  glewInfoFunc("glTexCoord1hNV", glTexCoord1hNV == NULL);
+  glewInfoFunc("glTexCoord1hvNV", glTexCoord1hvNV == NULL);
+  glewInfoFunc("glTexCoord2hNV", glTexCoord2hNV == NULL);
+  glewInfoFunc("glTexCoord2hvNV", glTexCoord2hvNV == NULL);
+  glewInfoFunc("glTexCoord3hNV", glTexCoord3hNV == NULL);
+  glewInfoFunc("glTexCoord3hvNV", glTexCoord3hvNV == NULL);
+  glewInfoFunc("glTexCoord4hNV", glTexCoord4hNV == NULL);
+  glewInfoFunc("glTexCoord4hvNV", glTexCoord4hvNV == NULL);
+  glewInfoFunc("glVertex2hNV", glVertex2hNV == NULL);
+  glewInfoFunc("glVertex2hvNV", glVertex2hvNV == NULL);
+  glewInfoFunc("glVertex3hNV", glVertex3hNV == NULL);
+  glewInfoFunc("glVertex3hvNV", glVertex3hvNV == NULL);
+  glewInfoFunc("glVertex4hNV", glVertex4hNV == NULL);
+  glewInfoFunc("glVertex4hvNV", glVertex4hvNV == NULL);
+  glewInfoFunc("glVertexAttrib1hNV", glVertexAttrib1hNV == NULL);
+  glewInfoFunc("glVertexAttrib1hvNV", glVertexAttrib1hvNV == NULL);
+  glewInfoFunc("glVertexAttrib2hNV", glVertexAttrib2hNV == NULL);
+  glewInfoFunc("glVertexAttrib2hvNV", glVertexAttrib2hvNV == NULL);
+  glewInfoFunc("glVertexAttrib3hNV", glVertexAttrib3hNV == NULL);
+  glewInfoFunc("glVertexAttrib3hvNV", glVertexAttrib3hvNV == NULL);
+  glewInfoFunc("glVertexAttrib4hNV", glVertexAttrib4hNV == NULL);
+  glewInfoFunc("glVertexAttrib4hvNV", glVertexAttrib4hvNV == NULL);
+  glewInfoFunc("glVertexAttribs1hvNV", glVertexAttribs1hvNV == NULL);
+  glewInfoFunc("glVertexAttribs2hvNV", glVertexAttribs2hvNV == NULL);
+  glewInfoFunc("glVertexAttribs3hvNV", glVertexAttribs3hvNV == NULL);
+  glewInfoFunc("glVertexAttribs4hvNV", glVertexAttribs4hvNV == NULL);
+  glewInfoFunc("glVertexWeighthNV", glVertexWeighthNV == NULL);
+  glewInfoFunc("glVertexWeighthvNV", glVertexWeighthvNV == NULL);
+}
+
+#endif /* GL_NV_half_float */
+
+#ifdef GL_NV_light_max_exponent
+
+static void _glewInfo_GL_NV_light_max_exponent (void)
+{
+  glewPrintExt("GL_NV_light_max_exponent", GLEW_NV_light_max_exponent, glewIsSupported("GL_NV_light_max_exponent"), glewGetExtension("GL_NV_light_max_exponent"));
+}
+
+#endif /* GL_NV_light_max_exponent */
+
+#ifdef GL_NV_multisample_filter_hint
+
+static void _glewInfo_GL_NV_multisample_filter_hint (void)
+{
+  glewPrintExt("GL_NV_multisample_filter_hint", GLEW_NV_multisample_filter_hint, glewIsSupported("GL_NV_multisample_filter_hint"), glewGetExtension("GL_NV_multisample_filter_hint"));
+}
+
+#endif /* GL_NV_multisample_filter_hint */
+
+#ifdef GL_NV_occlusion_query
+
+static void _glewInfo_GL_NV_occlusion_query (void)
+{
+  glewPrintExt("GL_NV_occlusion_query", GLEW_NV_occlusion_query, glewIsSupported("GL_NV_occlusion_query"), glewGetExtension("GL_NV_occlusion_query"));
+
+  glewInfoFunc("glBeginOcclusionQueryNV", glBeginOcclusionQueryNV == NULL);
+  glewInfoFunc("glDeleteOcclusionQueriesNV", glDeleteOcclusionQueriesNV == NULL);
+  glewInfoFunc("glEndOcclusionQueryNV", glEndOcclusionQueryNV == NULL);
+  glewInfoFunc("glGenOcclusionQueriesNV", glGenOcclusionQueriesNV == NULL);
+  glewInfoFunc("glGetOcclusionQueryivNV", glGetOcclusionQueryivNV == NULL);
+  glewInfoFunc("glGetOcclusionQueryuivNV", glGetOcclusionQueryuivNV == NULL);
+  glewInfoFunc("glIsOcclusionQueryNV", glIsOcclusionQueryNV == NULL);
+}
+
+#endif /* GL_NV_occlusion_query */
+
+#ifdef GL_NV_packed_depth_stencil
+
+static void _glewInfo_GL_NV_packed_depth_stencil (void)
+{
+  glewPrintExt("GL_NV_packed_depth_stencil", GLEW_NV_packed_depth_stencil, glewIsSupported("GL_NV_packed_depth_stencil"), glewGetExtension("GL_NV_packed_depth_stencil"));
+}
+
+#endif /* GL_NV_packed_depth_stencil */
+
+#ifdef GL_NV_parameter_buffer_object
+
+static void _glewInfo_GL_NV_parameter_buffer_object (void)
+{
+  glewPrintExt("GL_NV_parameter_buffer_object", GLEW_NV_parameter_buffer_object, glewIsSupported("GL_NV_parameter_buffer_object"), glewGetExtension("GL_NV_parameter_buffer_object"));
+
+  glewInfoFunc("glProgramBufferParametersIivNV", glProgramBufferParametersIivNV == NULL);
+  glewInfoFunc("glProgramBufferParametersIuivNV", glProgramBufferParametersIuivNV == NULL);
+  glewInfoFunc("glProgramBufferParametersfvNV", glProgramBufferParametersfvNV == NULL);
+}
+
+#endif /* GL_NV_parameter_buffer_object */
+
+#ifdef GL_NV_pixel_data_range
+
+static void _glewInfo_GL_NV_pixel_data_range (void)
+{
+  glewPrintExt("GL_NV_pixel_data_range", GLEW_NV_pixel_data_range, glewIsSupported("GL_NV_pixel_data_range"), glewGetExtension("GL_NV_pixel_data_range"));
+
+  glewInfoFunc("glFlushPixelDataRangeNV", glFlushPixelDataRangeNV == NULL);
+  glewInfoFunc("glPixelDataRangeNV", glPixelDataRangeNV == NULL);
+}
+
+#endif /* GL_NV_pixel_data_range */
+
+#ifdef GL_NV_point_sprite
+
+static void _glewInfo_GL_NV_point_sprite (void)
+{
+  glewPrintExt("GL_NV_point_sprite", GLEW_NV_point_sprite, glewIsSupported("GL_NV_point_sprite"), glewGetExtension("GL_NV_point_sprite"));
+
+  glewInfoFunc("glPointParameteriNV", glPointParameteriNV == NULL);
+  glewInfoFunc("glPointParameterivNV", glPointParameterivNV == NULL);
+}
+
+#endif /* GL_NV_point_sprite */
+
+#ifdef GL_NV_present_video
+
+static void _glewInfo_GL_NV_present_video (void)
+{
+  glewPrintExt("GL_NV_present_video", GLEW_NV_present_video, glewIsSupported("GL_NV_present_video"), glewGetExtension("GL_NV_present_video"));
+
+  glewInfoFunc("glGetVideoi64vNV", glGetVideoi64vNV == NULL);
+  glewInfoFunc("glGetVideoivNV", glGetVideoivNV == NULL);
+  glewInfoFunc("glGetVideoui64vNV", glGetVideoui64vNV == NULL);
+  glewInfoFunc("glGetVideouivNV", glGetVideouivNV == NULL);
+  glewInfoFunc("glPresentFrameDualFillNV", glPresentFrameDualFillNV == NULL);
+  glewInfoFunc("glPresentFrameKeyedNV", glPresentFrameKeyedNV == NULL);
+  glewInfoFunc("glVideoParameterivNV", glVideoParameterivNV == NULL);
+}
+
+#endif /* GL_NV_present_video */
+
+#ifdef GL_NV_primitive_restart
+
+static void _glewInfo_GL_NV_primitive_restart (void)
+{
+  glewPrintExt("GL_NV_primitive_restart", GLEW_NV_primitive_restart, glewIsSupported("GL_NV_primitive_restart"), glewGetExtension("GL_NV_primitive_restart"));
+
+  glewInfoFunc("glPrimitiveRestartIndexNV", glPrimitiveRestartIndexNV == NULL);
+  glewInfoFunc("glPrimitiveRestartNV", glPrimitiveRestartNV == NULL);
+}
+
+#endif /* GL_NV_primitive_restart */
+
+#ifdef GL_NV_register_combiners
+
+static void _glewInfo_GL_NV_register_combiners (void)
+{
+  glewPrintExt("GL_NV_register_combiners", GLEW_NV_register_combiners, glewIsSupported("GL_NV_register_combiners"), glewGetExtension("GL_NV_register_combiners"));
+
+  glewInfoFunc("glCombinerInputNV", glCombinerInputNV == NULL);
+  glewInfoFunc("glCombinerOutputNV", glCombinerOutputNV == NULL);
+  glewInfoFunc("glCombinerParameterfNV", glCombinerParameterfNV == NULL);
+  glewInfoFunc("glCombinerParameterfvNV", glCombinerParameterfvNV == NULL);
+  glewInfoFunc("glCombinerParameteriNV", glCombinerParameteriNV == NULL);
+  glewInfoFunc("glCombinerParameterivNV", glCombinerParameterivNV == NULL);
+  glewInfoFunc("glFinalCombinerInputNV", glFinalCombinerInputNV == NULL);
+  glewInfoFunc("glGetCombinerInputParameterfvNV", glGetCombinerInputParameterfvNV == NULL);
+  glewInfoFunc("glGetCombinerInputParameterivNV", glGetCombinerInputParameterivNV == NULL);
+  glewInfoFunc("glGetCombinerOutputParameterfvNV", glGetCombinerOutputParameterfvNV == NULL);
+  glewInfoFunc("glGetCombinerOutputParameterivNV", glGetCombinerOutputParameterivNV == NULL);
+  glewInfoFunc("glGetFinalCombinerInputParameterfvNV", glGetFinalCombinerInputParameterfvNV == NULL);
+  glewInfoFunc("glGetFinalCombinerInputParameterivNV", glGetFinalCombinerInputParameterivNV == NULL);
+}
+
+#endif /* GL_NV_register_combiners */
+
+#ifdef GL_NV_register_combiners2
+
+static void _glewInfo_GL_NV_register_combiners2 (void)
+{
+  glewPrintExt("GL_NV_register_combiners2", GLEW_NV_register_combiners2, glewIsSupported("GL_NV_register_combiners2"), glewGetExtension("GL_NV_register_combiners2"));
+
+  glewInfoFunc("glCombinerStageParameterfvNV", glCombinerStageParameterfvNV == NULL);
+  glewInfoFunc("glGetCombinerStageParameterfvNV", glGetCombinerStageParameterfvNV == NULL);
+}
+
+#endif /* GL_NV_register_combiners2 */
+
+#ifdef GL_NV_texgen_emboss
+
+static void _glewInfo_GL_NV_texgen_emboss (void)
+{
+  glewPrintExt("GL_NV_texgen_emboss", GLEW_NV_texgen_emboss, glewIsSupported("GL_NV_texgen_emboss"), glewGetExtension("GL_NV_texgen_emboss"));
+}
+
+#endif /* GL_NV_texgen_emboss */
+
+#ifdef GL_NV_texgen_reflection
+
+static void _glewInfo_GL_NV_texgen_reflection (void)
+{
+  glewPrintExt("GL_NV_texgen_reflection", GLEW_NV_texgen_reflection, glewIsSupported("GL_NV_texgen_reflection"), glewGetExtension("GL_NV_texgen_reflection"));
+}
+
+#endif /* GL_NV_texgen_reflection */
+
+#ifdef GL_NV_texture_compression_vtc
+
+static void _glewInfo_GL_NV_texture_compression_vtc (void)
+{
+  glewPrintExt("GL_NV_texture_compression_vtc", GLEW_NV_texture_compression_vtc, glewIsSupported("GL_NV_texture_compression_vtc"), glewGetExtension("GL_NV_texture_compression_vtc"));
+}
+
+#endif /* GL_NV_texture_compression_vtc */
+
+#ifdef GL_NV_texture_env_combine4
+
+static void _glewInfo_GL_NV_texture_env_combine4 (void)
+{
+  glewPrintExt("GL_NV_texture_env_combine4", GLEW_NV_texture_env_combine4, glewIsSupported("GL_NV_texture_env_combine4"), glewGetExtension("GL_NV_texture_env_combine4"));
+}
+
+#endif /* GL_NV_texture_env_combine4 */
+
+#ifdef GL_NV_texture_expand_normal
+
+static void _glewInfo_GL_NV_texture_expand_normal (void)
+{
+  glewPrintExt("GL_NV_texture_expand_normal", GLEW_NV_texture_expand_normal, glewIsSupported("GL_NV_texture_expand_normal"), glewGetExtension("GL_NV_texture_expand_normal"));
+}
+
+#endif /* GL_NV_texture_expand_normal */
+
+#ifdef GL_NV_texture_rectangle
+
+static void _glewInfo_GL_NV_texture_rectangle (void)
+{
+  glewPrintExt("GL_NV_texture_rectangle", GLEW_NV_texture_rectangle, glewIsSupported("GL_NV_texture_rectangle"), glewGetExtension("GL_NV_texture_rectangle"));
+}
+
+#endif /* GL_NV_texture_rectangle */
+
+#ifdef GL_NV_texture_shader
+
+static void _glewInfo_GL_NV_texture_shader (void)
+{
+  glewPrintExt("GL_NV_texture_shader", GLEW_NV_texture_shader, glewIsSupported("GL_NV_texture_shader"), glewGetExtension("GL_NV_texture_shader"));
+}
+
+#endif /* GL_NV_texture_shader */
+
+#ifdef GL_NV_texture_shader2
+
+static void _glewInfo_GL_NV_texture_shader2 (void)
+{
+  glewPrintExt("GL_NV_texture_shader2", GLEW_NV_texture_shader2, glewIsSupported("GL_NV_texture_shader2"), glewGetExtension("GL_NV_texture_shader2"));
+}
+
+#endif /* GL_NV_texture_shader2 */
+
+#ifdef GL_NV_texture_shader3
+
+static void _glewInfo_GL_NV_texture_shader3 (void)
+{
+  glewPrintExt("GL_NV_texture_shader3", GLEW_NV_texture_shader3, glewIsSupported("GL_NV_texture_shader3"), glewGetExtension("GL_NV_texture_shader3"));
+}
+
+#endif /* GL_NV_texture_shader3 */
+
+#ifdef GL_NV_transform_feedback
+
+static void _glewInfo_GL_NV_transform_feedback (void)
+{
+  glewPrintExt("GL_NV_transform_feedback", GLEW_NV_transform_feedback, glewIsSupported("GL_NV_transform_feedback"), glewGetExtension("GL_NV_transform_feedback"));
+
+  glewInfoFunc("glActiveVaryingNV", glActiveVaryingNV == NULL);
+  glewInfoFunc("glBeginTransformFeedbackNV", glBeginTransformFeedbackNV == NULL);
+  glewInfoFunc("glBindBufferBaseNV", glBindBufferBaseNV == NULL);
+  glewInfoFunc("glBindBufferOffsetNV", glBindBufferOffsetNV == NULL);
+  glewInfoFunc("glBindBufferRangeNV", glBindBufferRangeNV == NULL);
+  glewInfoFunc("glEndTransformFeedbackNV", glEndTransformFeedbackNV == NULL);
+  glewInfoFunc("glGetActiveVaryingNV", glGetActiveVaryingNV == NULL);
+  glewInfoFunc("glGetTransformFeedbackVaryingNV", glGetTransformFeedbackVaryingNV == NULL);
+  glewInfoFunc("glGetVaryingLocationNV", glGetVaryingLocationNV == NULL);
+  glewInfoFunc("glTransformFeedbackAttribsNV", glTransformFeedbackAttribsNV == NULL);
+  glewInfoFunc("glTransformFeedbackVaryingsNV", glTransformFeedbackVaryingsNV == NULL);
+}
+
+#endif /* GL_NV_transform_feedback */
+
+#ifdef GL_NV_vertex_array_range
+
+static void _glewInfo_GL_NV_vertex_array_range (void)
+{
+  glewPrintExt("GL_NV_vertex_array_range", GLEW_NV_vertex_array_range, glewIsSupported("GL_NV_vertex_array_range"), glewGetExtension("GL_NV_vertex_array_range"));
+
+  glewInfoFunc("glFlushVertexArrayRangeNV", glFlushVertexArrayRangeNV == NULL);
+  glewInfoFunc("glVertexArrayRangeNV", glVertexArrayRangeNV == NULL);
+}
+
+#endif /* GL_NV_vertex_array_range */
+
+#ifdef GL_NV_vertex_array_range2
+
+static void _glewInfo_GL_NV_vertex_array_range2 (void)
+{
+  glewPrintExt("GL_NV_vertex_array_range2", GLEW_NV_vertex_array_range2, glewIsSupported("GL_NV_vertex_array_range2"), glewGetExtension("GL_NV_vertex_array_range2"));
+}
+
+#endif /* GL_NV_vertex_array_range2 */
+
+#ifdef GL_NV_vertex_program
+
+static void _glewInfo_GL_NV_vertex_program (void)
+{
+  glewPrintExt("GL_NV_vertex_program", GLEW_NV_vertex_program, glewIsSupported("GL_NV_vertex_program"), glewGetExtension("GL_NV_vertex_program"));
+
+  glewInfoFunc("glAreProgramsResidentNV", glAreProgramsResidentNV == NULL);
+  glewInfoFunc("glBindProgramNV", glBindProgramNV == NULL);
+  glewInfoFunc("glDeleteProgramsNV", glDeleteProgramsNV == NULL);
+  glewInfoFunc("glExecuteProgramNV", glExecuteProgramNV == NULL);
+  glewInfoFunc("glGenProgramsNV", glGenProgramsNV == NULL);
+  glewInfoFunc("glGetProgramParameterdvNV", glGetProgramParameterdvNV == NULL);
+  glewInfoFunc("glGetProgramParameterfvNV", glGetProgramParameterfvNV == NULL);
+  glewInfoFunc("glGetProgramStringNV", glGetProgramStringNV == NULL);
+  glewInfoFunc("glGetProgramivNV", glGetProgramivNV == NULL);
+  glewInfoFunc("glGetTrackMatrixivNV", glGetTrackMatrixivNV == NULL);
+  glewInfoFunc("glGetVertexAttribPointervNV", glGetVertexAttribPointervNV == NULL);
+  glewInfoFunc("glGetVertexAttribdvNV", glGetVertexAttribdvNV == NULL);
+  glewInfoFunc("glGetVertexAttribfvNV", glGetVertexAttribfvNV == NULL);
+  glewInfoFunc("glGetVertexAttribivNV", glGetVertexAttribivNV == NULL);
+  glewInfoFunc("glIsProgramNV", glIsProgramNV == NULL);
+  glewInfoFunc("glLoadProgramNV", glLoadProgramNV == NULL);
+  glewInfoFunc("glProgramParameter4dNV", glProgramParameter4dNV == NULL);
+  glewInfoFunc("glProgramParameter4dvNV", glProgramParameter4dvNV == NULL);
+  glewInfoFunc("glProgramParameter4fNV", glProgramParameter4fNV == NULL);
+  glewInfoFunc("glProgramParameter4fvNV", glProgramParameter4fvNV == NULL);
+  glewInfoFunc("glProgramParameters4dvNV", glProgramParameters4dvNV == NULL);
+  glewInfoFunc("glProgramParameters4fvNV", glProgramParameters4fvNV == NULL);
+  glewInfoFunc("glRequestResidentProgramsNV", glRequestResidentProgramsNV == NULL);
+  glewInfoFunc("glTrackMatrixNV", glTrackMatrixNV == NULL);
+  glewInfoFunc("glVertexAttrib1dNV", glVertexAttrib1dNV == NULL);
+  glewInfoFunc("glVertexAttrib1dvNV", glVertexAttrib1dvNV == NULL);
+  glewInfoFunc("glVertexAttrib1fNV", glVertexAttrib1fNV == NULL);
+  glewInfoFunc("glVertexAttrib1fvNV", glVertexAttrib1fvNV == NULL);
+  glewInfoFunc("glVertexAttrib1sNV", glVertexAttrib1sNV == NULL);
+  glewInfoFunc("glVertexAttrib1svNV", glVertexAttrib1svNV == NULL);
+  glewInfoFunc("glVertexAttrib2dNV", glVertexAttrib2dNV == NULL);
+  glewInfoFunc("glVertexAttrib2dvNV", glVertexAttrib2dvNV == NULL);
+  glewInfoFunc("glVertexAttrib2fNV", glVertexAttrib2fNV == NULL);
+  glewInfoFunc("glVertexAttrib2fvNV", glVertexAttrib2fvNV == NULL);
+  glewInfoFunc("glVertexAttrib2sNV", glVertexAttrib2sNV == NULL);
+  glewInfoFunc("glVertexAttrib2svNV", glVertexAttrib2svNV == NULL);
+  glewInfoFunc("glVertexAttrib3dNV", glVertexAttrib3dNV == NULL);
+  glewInfoFunc("glVertexAttrib3dvNV", glVertexAttrib3dvNV == NULL);
+  glewInfoFunc("glVertexAttrib3fNV", glVertexAttrib3fNV == NULL);
+  glewInfoFunc("glVertexAttrib3fvNV", glVertexAttrib3fvNV == NULL);
+  glewInfoFunc("glVertexAttrib3sNV", glVertexAttrib3sNV == NULL);
+  glewInfoFunc("glVertexAttrib3svNV", glVertexAttrib3svNV == NULL);
+  glewInfoFunc("glVertexAttrib4dNV", glVertexAttrib4dNV == NULL);
+  glewInfoFunc("glVertexAttrib4dvNV", glVertexAttrib4dvNV == NULL);
+  glewInfoFunc("glVertexAttrib4fNV", glVertexAttrib4fNV == NULL);
+  glewInfoFunc("glVertexAttrib4fvNV", glVertexAttrib4fvNV == NULL);
+  glewInfoFunc("glVertexAttrib4sNV", glVertexAttrib4sNV == NULL);
+  glewInfoFunc("glVertexAttrib4svNV", glVertexAttrib4svNV == NULL);
+  glewInfoFunc("glVertexAttrib4ubNV", glVertexAttrib4ubNV == NULL);
+  glewInfoFunc("glVertexAttrib4ubvNV", glVertexAttrib4ubvNV == NULL);
+  glewInfoFunc("glVertexAttribPointerNV", glVertexAttribPointerNV == NULL);
+  glewInfoFunc("glVertexAttribs1dvNV", glVertexAttribs1dvNV == NULL);
+  glewInfoFunc("glVertexAttribs1fvNV", glVertexAttribs1fvNV == NULL);
+  glewInfoFunc("glVertexAttribs1svNV", glVertexAttribs1svNV == NULL);
+  glewInfoFunc("glVertexAttribs2dvNV", glVertexAttribs2dvNV == NULL);
+  glewInfoFunc("glVertexAttribs2fvNV", glVertexAttribs2fvNV == NULL);
+  glewInfoFunc("glVertexAttribs2svNV", glVertexAttribs2svNV == NULL);
+  glewInfoFunc("glVertexAttribs3dvNV", glVertexAttribs3dvNV == NULL);
+  glewInfoFunc("glVertexAttribs3fvNV", glVertexAttribs3fvNV == NULL);
+  glewInfoFunc("glVertexAttribs3svNV", glVertexAttribs3svNV == NULL);
+  glewInfoFunc("glVertexAttribs4dvNV", glVertexAttribs4dvNV == NULL);
+  glewInfoFunc("glVertexAttribs4fvNV", glVertexAttribs4fvNV == NULL);
+  glewInfoFunc("glVertexAttribs4svNV", glVertexAttribs4svNV == NULL);
+  glewInfoFunc("glVertexAttribs4ubvNV", glVertexAttribs4ubvNV == NULL);
+}
+
+#endif /* GL_NV_vertex_program */
+
+#ifdef GL_NV_vertex_program1_1
+
+static void _glewInfo_GL_NV_vertex_program1_1 (void)
+{
+  glewPrintExt("GL_NV_vertex_program1_1", GLEW_NV_vertex_program1_1, glewIsSupported("GL_NV_vertex_program1_1"), glewGetExtension("GL_NV_vertex_program1_1"));
+}
+
+#endif /* GL_NV_vertex_program1_1 */
+
+#ifdef GL_NV_vertex_program2
+
+static void _glewInfo_GL_NV_vertex_program2 (void)
+{
+  glewPrintExt("GL_NV_vertex_program2", GLEW_NV_vertex_program2, glewIsSupported("GL_NV_vertex_program2"), glewGetExtension("GL_NV_vertex_program2"));
+}
+
+#endif /* GL_NV_vertex_program2 */
+
+#ifdef GL_NV_vertex_program2_option
+
+static void _glewInfo_GL_NV_vertex_program2_option (void)
+{
+  glewPrintExt("GL_NV_vertex_program2_option", GLEW_NV_vertex_program2_option, glewIsSupported("GL_NV_vertex_program2_option"), glewGetExtension("GL_NV_vertex_program2_option"));
+}
+
+#endif /* GL_NV_vertex_program2_option */
+
+#ifdef GL_NV_vertex_program3
+
+static void _glewInfo_GL_NV_vertex_program3 (void)
+{
+  glewPrintExt("GL_NV_vertex_program3", GLEW_NV_vertex_program3, glewIsSupported("GL_NV_vertex_program3"), glewGetExtension("GL_NV_vertex_program3"));
+}
+
+#endif /* GL_NV_vertex_program3 */
+
+#ifdef GL_NV_vertex_program4
+
+static void _glewInfo_GL_NV_vertex_program4 (void)
+{
+  glewPrintExt("GL_NV_vertex_program4", GLEW_NV_vertex_program4, glewIsSupported("GL_NV_vertex_program4"), glewGetExtension("GL_NV_vertex_program4"));
+}
+
+#endif /* GL_NV_vertex_program4 */
+
+#ifdef GL_OES_byte_coordinates
+
+static void _glewInfo_GL_OES_byte_coordinates (void)
+{
+  glewPrintExt("GL_OES_byte_coordinates", GLEW_OES_byte_coordinates, glewIsSupported("GL_OES_byte_coordinates"), glewGetExtension("GL_OES_byte_coordinates"));
+}
+
+#endif /* GL_OES_byte_coordinates */
+
+#ifdef GL_OES_compressed_paletted_texture
+
+static void _glewInfo_GL_OES_compressed_paletted_texture (void)
+{
+  glewPrintExt("GL_OES_compressed_paletted_texture", GLEW_OES_compressed_paletted_texture, glewIsSupported("GL_OES_compressed_paletted_texture"), glewGetExtension("GL_OES_compressed_paletted_texture"));
+}
+
+#endif /* GL_OES_compressed_paletted_texture */
+
+#ifdef GL_OES_read_format
+
+static void _glewInfo_GL_OES_read_format (void)
+{
+  glewPrintExt("GL_OES_read_format", GLEW_OES_read_format, glewIsSupported("GL_OES_read_format"), glewGetExtension("GL_OES_read_format"));
+}
+
+#endif /* GL_OES_read_format */
+
+#ifdef GL_OES_single_precision
+
+static void _glewInfo_GL_OES_single_precision (void)
+{
+  glewPrintExt("GL_OES_single_precision", GLEW_OES_single_precision, glewIsSupported("GL_OES_single_precision"), glewGetExtension("GL_OES_single_precision"));
+
+  glewInfoFunc("glClearDepthfOES", glClearDepthfOES == NULL);
+  glewInfoFunc("glClipPlanefOES", glClipPlanefOES == NULL);
+  glewInfoFunc("glDepthRangefOES", glDepthRangefOES == NULL);
+  glewInfoFunc("glFrustumfOES", glFrustumfOES == NULL);
+  glewInfoFunc("glGetClipPlanefOES", glGetClipPlanefOES == NULL);
+  glewInfoFunc("glOrthofOES", glOrthofOES == NULL);
+}
+
+#endif /* GL_OES_single_precision */
+
+#ifdef GL_OML_interlace
+
+static void _glewInfo_GL_OML_interlace (void)
+{
+  glewPrintExt("GL_OML_interlace", GLEW_OML_interlace, glewIsSupported("GL_OML_interlace"), glewGetExtension("GL_OML_interlace"));
+}
+
+#endif /* GL_OML_interlace */
+
+#ifdef GL_OML_resample
+
+static void _glewInfo_GL_OML_resample (void)
+{
+  glewPrintExt("GL_OML_resample", GLEW_OML_resample, glewIsSupported("GL_OML_resample"), glewGetExtension("GL_OML_resample"));
+}
+
+#endif /* GL_OML_resample */
+
+#ifdef GL_OML_subsample
+
+static void _glewInfo_GL_OML_subsample (void)
+{
+  glewPrintExt("GL_OML_subsample", GLEW_OML_subsample, glewIsSupported("GL_OML_subsample"), glewGetExtension("GL_OML_subsample"));
+}
+
+#endif /* GL_OML_subsample */
+
+#ifdef GL_PGI_misc_hints
+
+static void _glewInfo_GL_PGI_misc_hints (void)
+{
+  glewPrintExt("GL_PGI_misc_hints", GLEW_PGI_misc_hints, glewIsSupported("GL_PGI_misc_hints"), glewGetExtension("GL_PGI_misc_hints"));
+}
+
+#endif /* GL_PGI_misc_hints */
+
+#ifdef GL_PGI_vertex_hints
+
+static void _glewInfo_GL_PGI_vertex_hints (void)
+{
+  glewPrintExt("GL_PGI_vertex_hints", GLEW_PGI_vertex_hints, glewIsSupported("GL_PGI_vertex_hints"), glewGetExtension("GL_PGI_vertex_hints"));
+}
+
+#endif /* GL_PGI_vertex_hints */
+
+#ifdef GL_REND_screen_coordinates
+
+static void _glewInfo_GL_REND_screen_coordinates (void)
+{
+  glewPrintExt("GL_REND_screen_coordinates", GLEW_REND_screen_coordinates, glewIsSupported("GL_REND_screen_coordinates"), glewGetExtension("GL_REND_screen_coordinates"));
+}
+
+#endif /* GL_REND_screen_coordinates */
+
+#ifdef GL_S3_s3tc
+
+static void _glewInfo_GL_S3_s3tc (void)
+{
+  glewPrintExt("GL_S3_s3tc", GLEW_S3_s3tc, glewIsSupported("GL_S3_s3tc"), glewGetExtension("GL_S3_s3tc"));
+}
+
+#endif /* GL_S3_s3tc */
+
+#ifdef GL_SGIS_color_range
+
+static void _glewInfo_GL_SGIS_color_range (void)
+{
+  glewPrintExt("GL_SGIS_color_range", GLEW_SGIS_color_range, glewIsSupported("GL_SGIS_color_range"), glewGetExtension("GL_SGIS_color_range"));
+}
+
+#endif /* GL_SGIS_color_range */
+
+#ifdef GL_SGIS_detail_texture
+
+static void _glewInfo_GL_SGIS_detail_texture (void)
+{
+  glewPrintExt("GL_SGIS_detail_texture", GLEW_SGIS_detail_texture, glewIsSupported("GL_SGIS_detail_texture"), glewGetExtension("GL_SGIS_detail_texture"));
+
+  glewInfoFunc("glDetailTexFuncSGIS", glDetailTexFuncSGIS == NULL);
+  glewInfoFunc("glGetDetailTexFuncSGIS", glGetDetailTexFuncSGIS == NULL);
+}
+
+#endif /* GL_SGIS_detail_texture */
+
+#ifdef GL_SGIS_fog_function
+
+static void _glewInfo_GL_SGIS_fog_function (void)
+{
+  glewPrintExt("GL_SGIS_fog_function", GLEW_SGIS_fog_function, glewIsSupported("GL_SGIS_fog_function"), glewGetExtension("GL_SGIS_fog_function"));
+
+  glewInfoFunc("glFogFuncSGIS", glFogFuncSGIS == NULL);
+  glewInfoFunc("glGetFogFuncSGIS", glGetFogFuncSGIS == NULL);
+}
+
+#endif /* GL_SGIS_fog_function */
+
+#ifdef GL_SGIS_generate_mipmap
+
+static void _glewInfo_GL_SGIS_generate_mipmap (void)
+{
+  glewPrintExt("GL_SGIS_generate_mipmap", GLEW_SGIS_generate_mipmap, glewIsSupported("GL_SGIS_generate_mipmap"), glewGetExtension("GL_SGIS_generate_mipmap"));
+}
+
+#endif /* GL_SGIS_generate_mipmap */
+
+#ifdef GL_SGIS_multisample
+
+static void _glewInfo_GL_SGIS_multisample (void)
+{
+  glewPrintExt("GL_SGIS_multisample", GLEW_SGIS_multisample, glewIsSupported("GL_SGIS_multisample"), glewGetExtension("GL_SGIS_multisample"));
+
+  glewInfoFunc("glSampleMaskSGIS", glSampleMaskSGIS == NULL);
+  glewInfoFunc("glSamplePatternSGIS", glSamplePatternSGIS == NULL);
+}
+
+#endif /* GL_SGIS_multisample */
+
+#ifdef GL_SGIS_pixel_texture
+
+static void _glewInfo_GL_SGIS_pixel_texture (void)
+{
+  glewPrintExt("GL_SGIS_pixel_texture", GLEW_SGIS_pixel_texture, glewIsSupported("GL_SGIS_pixel_texture"), glewGetExtension("GL_SGIS_pixel_texture"));
+}
+
+#endif /* GL_SGIS_pixel_texture */
+
+#ifdef GL_SGIS_point_line_texgen
+
+static void _glewInfo_GL_SGIS_point_line_texgen (void)
+{
+  glewPrintExt("GL_SGIS_point_line_texgen", GLEW_SGIS_point_line_texgen, glewIsSupported("GL_SGIS_point_line_texgen"), glewGetExtension("GL_SGIS_point_line_texgen"));
+}
+
+#endif /* GL_SGIS_point_line_texgen */
+
+#ifdef GL_SGIS_sharpen_texture
+
+static void _glewInfo_GL_SGIS_sharpen_texture (void)
+{
+  glewPrintExt("GL_SGIS_sharpen_texture", GLEW_SGIS_sharpen_texture, glewIsSupported("GL_SGIS_sharpen_texture"), glewGetExtension("GL_SGIS_sharpen_texture"));
+
+  glewInfoFunc("glGetSharpenTexFuncSGIS", glGetSharpenTexFuncSGIS == NULL);
+  glewInfoFunc("glSharpenTexFuncSGIS", glSharpenTexFuncSGIS == NULL);
+}
+
+#endif /* GL_SGIS_sharpen_texture */
+
+#ifdef GL_SGIS_texture4D
+
+static void _glewInfo_GL_SGIS_texture4D (void)
+{
+  glewPrintExt("GL_SGIS_texture4D", GLEW_SGIS_texture4D, glewIsSupported("GL_SGIS_texture4D"), glewGetExtension("GL_SGIS_texture4D"));
+
+  glewInfoFunc("glTexImage4DSGIS", glTexImage4DSGIS == NULL);
+  glewInfoFunc("glTexSubImage4DSGIS", glTexSubImage4DSGIS == NULL);
+}
+
+#endif /* GL_SGIS_texture4D */
+
+#ifdef GL_SGIS_texture_border_clamp
+
+static void _glewInfo_GL_SGIS_texture_border_clamp (void)
+{
+  glewPrintExt("GL_SGIS_texture_border_clamp", GLEW_SGIS_texture_border_clamp, glewIsSupported("GL_SGIS_texture_border_clamp"), glewGetExtension("GL_SGIS_texture_border_clamp"));
+}
+
+#endif /* GL_SGIS_texture_border_clamp */
+
+#ifdef GL_SGIS_texture_edge_clamp
+
+static void _glewInfo_GL_SGIS_texture_edge_clamp (void)
+{
+  glewPrintExt("GL_SGIS_texture_edge_clamp", GLEW_SGIS_texture_edge_clamp, glewIsSupported("GL_SGIS_texture_edge_clamp"), glewGetExtension("GL_SGIS_texture_edge_clamp"));
+}
+
+#endif /* GL_SGIS_texture_edge_clamp */
+
+#ifdef GL_SGIS_texture_filter4
+
+static void _glewInfo_GL_SGIS_texture_filter4 (void)
+{
+  glewPrintExt("GL_SGIS_texture_filter4", GLEW_SGIS_texture_filter4, glewIsSupported("GL_SGIS_texture_filter4"), glewGetExtension("GL_SGIS_texture_filter4"));
+
+  glewInfoFunc("glGetTexFilterFuncSGIS", glGetTexFilterFuncSGIS == NULL);
+  glewInfoFunc("glTexFilterFuncSGIS", glTexFilterFuncSGIS == NULL);
+}
+
+#endif /* GL_SGIS_texture_filter4 */
+
+#ifdef GL_SGIS_texture_lod
+
+static void _glewInfo_GL_SGIS_texture_lod (void)
+{
+  glewPrintExt("GL_SGIS_texture_lod", GLEW_SGIS_texture_lod, glewIsSupported("GL_SGIS_texture_lod"), glewGetExtension("GL_SGIS_texture_lod"));
+}
+
+#endif /* GL_SGIS_texture_lod */
+
+#ifdef GL_SGIS_texture_select
+
+static void _glewInfo_GL_SGIS_texture_select (void)
+{
+  glewPrintExt("GL_SGIS_texture_select", GLEW_SGIS_texture_select, glewIsSupported("GL_SGIS_texture_select"), glewGetExtension("GL_SGIS_texture_select"));
+}
+
+#endif /* GL_SGIS_texture_select */
+
+#ifdef GL_SGIX_async
+
+static void _glewInfo_GL_SGIX_async (void)
+{
+  glewPrintExt("GL_SGIX_async", GLEW_SGIX_async, glewIsSupported("GL_SGIX_async"), glewGetExtension("GL_SGIX_async"));
+
+  glewInfoFunc("glAsyncMarkerSGIX", glAsyncMarkerSGIX == NULL);
+  glewInfoFunc("glDeleteAsyncMarkersSGIX", glDeleteAsyncMarkersSGIX == NULL);
+  glewInfoFunc("glFinishAsyncSGIX", glFinishAsyncSGIX == NULL);
+  glewInfoFunc("glGenAsyncMarkersSGIX", glGenAsyncMarkersSGIX == NULL);
+  glewInfoFunc("glIsAsyncMarkerSGIX", glIsAsyncMarkerSGIX == NULL);
+  glewInfoFunc("glPollAsyncSGIX", glPollAsyncSGIX == NULL);
+}
+
+#endif /* GL_SGIX_async */
+
+#ifdef GL_SGIX_async_histogram
+
+static void _glewInfo_GL_SGIX_async_histogram (void)
+{
+  glewPrintExt("GL_SGIX_async_histogram", GLEW_SGIX_async_histogram, glewIsSupported("GL_SGIX_async_histogram"), glewGetExtension("GL_SGIX_async_histogram"));
+}
+
+#endif /* GL_SGIX_async_histogram */
+
+#ifdef GL_SGIX_async_pixel
+
+static void _glewInfo_GL_SGIX_async_pixel (void)
+{
+  glewPrintExt("GL_SGIX_async_pixel", GLEW_SGIX_async_pixel, glewIsSupported("GL_SGIX_async_pixel"), glewGetExtension("GL_SGIX_async_pixel"));
+}
+
+#endif /* GL_SGIX_async_pixel */
+
+#ifdef GL_SGIX_blend_alpha_minmax
+
+static void _glewInfo_GL_SGIX_blend_alpha_minmax (void)
+{
+  glewPrintExt("GL_SGIX_blend_alpha_minmax", GLEW_SGIX_blend_alpha_minmax, glewIsSupported("GL_SGIX_blend_alpha_minmax"), glewGetExtension("GL_SGIX_blend_alpha_minmax"));
+}
+
+#endif /* GL_SGIX_blend_alpha_minmax */
+
+#ifdef GL_SGIX_clipmap
+
+static void _glewInfo_GL_SGIX_clipmap (void)
+{
+  glewPrintExt("GL_SGIX_clipmap", GLEW_SGIX_clipmap, glewIsSupported("GL_SGIX_clipmap"), glewGetExtension("GL_SGIX_clipmap"));
+}
+
+#endif /* GL_SGIX_clipmap */
+
+#ifdef GL_SGIX_convolution_accuracy
+
+static void _glewInfo_GL_SGIX_convolution_accuracy (void)
+{
+  glewPrintExt("GL_SGIX_convolution_accuracy", GLEW_SGIX_convolution_accuracy, glewIsSupported("GL_SGIX_convolution_accuracy"), glewGetExtension("GL_SGIX_convolution_accuracy"));
+}
+
+#endif /* GL_SGIX_convolution_accuracy */
+
+#ifdef GL_SGIX_depth_texture
+
+static void _glewInfo_GL_SGIX_depth_texture (void)
+{
+  glewPrintExt("GL_SGIX_depth_texture", GLEW_SGIX_depth_texture, glewIsSupported("GL_SGIX_depth_texture"), glewGetExtension("GL_SGIX_depth_texture"));
+}
+
+#endif /* GL_SGIX_depth_texture */
+
+#ifdef GL_SGIX_flush_raster
+
+static void _glewInfo_GL_SGIX_flush_raster (void)
+{
+  glewPrintExt("GL_SGIX_flush_raster", GLEW_SGIX_flush_raster, glewIsSupported("GL_SGIX_flush_raster"), glewGetExtension("GL_SGIX_flush_raster"));
+
+  glewInfoFunc("glFlushRasterSGIX", glFlushRasterSGIX == NULL);
+}
+
+#endif /* GL_SGIX_flush_raster */
+
+#ifdef GL_SGIX_fog_offset
+
+static void _glewInfo_GL_SGIX_fog_offset (void)
+{
+  glewPrintExt("GL_SGIX_fog_offset", GLEW_SGIX_fog_offset, glewIsSupported("GL_SGIX_fog_offset"), glewGetExtension("GL_SGIX_fog_offset"));
+}
+
+#endif /* GL_SGIX_fog_offset */
+
+#ifdef GL_SGIX_fog_texture
+
+static void _glewInfo_GL_SGIX_fog_texture (void)
+{
+  glewPrintExt("GL_SGIX_fog_texture", GLEW_SGIX_fog_texture, glewIsSupported("GL_SGIX_fog_texture"), glewGetExtension("GL_SGIX_fog_texture"));
+
+  glewInfoFunc("glTextureFogSGIX", glTextureFogSGIX == NULL);
+}
+
+#endif /* GL_SGIX_fog_texture */
+
+#ifdef GL_SGIX_fragment_specular_lighting
+
+static void _glewInfo_GL_SGIX_fragment_specular_lighting (void)
+{
+  glewPrintExt("GL_SGIX_fragment_specular_lighting", GLEW_SGIX_fragment_specular_lighting, glewIsSupported("GL_SGIX_fragment_specular_lighting"), glewGetExtension("GL_SGIX_fragment_specular_lighting"));
+
+  glewInfoFunc("glFragmentColorMaterialSGIX", glFragmentColorMaterialSGIX == NULL);
+  glewInfoFunc("glFragmentLightModelfSGIX", glFragmentLightModelfSGIX == NULL);
+  glewInfoFunc("glFragmentLightModelfvSGIX", glFragmentLightModelfvSGIX == NULL);
+  glewInfoFunc("glFragmentLightModeliSGIX", glFragmentLightModeliSGIX == NULL);
+  glewInfoFunc("glFragmentLightModelivSGIX", glFragmentLightModelivSGIX == NULL);
+  glewInfoFunc("glFragmentLightfSGIX", glFragmentLightfSGIX == NULL);
+  glewInfoFunc("glFragmentLightfvSGIX", glFragmentLightfvSGIX == NULL);
+  glewInfoFunc("glFragmentLightiSGIX", glFragmentLightiSGIX == NULL);
+  glewInfoFunc("glFragmentLightivSGIX", glFragmentLightivSGIX == NULL);
+  glewInfoFunc("glFragmentMaterialfSGIX", glFragmentMaterialfSGIX == NULL);
+  glewInfoFunc("glFragmentMaterialfvSGIX", glFragmentMaterialfvSGIX == NULL);
+  glewInfoFunc("glFragmentMaterialiSGIX", glFragmentMaterialiSGIX == NULL);
+  glewInfoFunc("glFragmentMaterialivSGIX", glFragmentMaterialivSGIX == NULL);
+  glewInfoFunc("glGetFragmentLightfvSGIX", glGetFragmentLightfvSGIX == NULL);
+  glewInfoFunc("glGetFragmentLightivSGIX", glGetFragmentLightivSGIX == NULL);
+  glewInfoFunc("glGetFragmentMaterialfvSGIX", glGetFragmentMaterialfvSGIX == NULL);
+  glewInfoFunc("glGetFragmentMaterialivSGIX", glGetFragmentMaterialivSGIX == NULL);
+}
+
+#endif /* GL_SGIX_fragment_specular_lighting */
+
+#ifdef GL_SGIX_framezoom
+
+static void _glewInfo_GL_SGIX_framezoom (void)
+{
+  glewPrintExt("GL_SGIX_framezoom", GLEW_SGIX_framezoom, glewIsSupported("GL_SGIX_framezoom"), glewGetExtension("GL_SGIX_framezoom"));
+
+  glewInfoFunc("glFrameZoomSGIX", glFrameZoomSGIX == NULL);
+}
+
+#endif /* GL_SGIX_framezoom */
+
+#ifdef GL_SGIX_interlace
+
+static void _glewInfo_GL_SGIX_interlace (void)
+{
+  glewPrintExt("GL_SGIX_interlace", GLEW_SGIX_interlace, glewIsSupported("GL_SGIX_interlace"), glewGetExtension("GL_SGIX_interlace"));
+}
+
+#endif /* GL_SGIX_interlace */
+
+#ifdef GL_SGIX_ir_instrument1
+
+static void _glewInfo_GL_SGIX_ir_instrument1 (void)
+{
+  glewPrintExt("GL_SGIX_ir_instrument1", GLEW_SGIX_ir_instrument1, glewIsSupported("GL_SGIX_ir_instrument1"), glewGetExtension("GL_SGIX_ir_instrument1"));
+}
+
+#endif /* GL_SGIX_ir_instrument1 */
+
+#ifdef GL_SGIX_list_priority
+
+static void _glewInfo_GL_SGIX_list_priority (void)
+{
+  glewPrintExt("GL_SGIX_list_priority", GLEW_SGIX_list_priority, glewIsSupported("GL_SGIX_list_priority"), glewGetExtension("GL_SGIX_list_priority"));
+}
+
+#endif /* GL_SGIX_list_priority */
+
+#ifdef GL_SGIX_pixel_texture
+
+static void _glewInfo_GL_SGIX_pixel_texture (void)
+{
+  glewPrintExt("GL_SGIX_pixel_texture", GLEW_SGIX_pixel_texture, glewIsSupported("GL_SGIX_pixel_texture"), glewGetExtension("GL_SGIX_pixel_texture"));
+
+  glewInfoFunc("glPixelTexGenSGIX", glPixelTexGenSGIX == NULL);
+}
+
+#endif /* GL_SGIX_pixel_texture */
+
+#ifdef GL_SGIX_pixel_texture_bits
+
+static void _glewInfo_GL_SGIX_pixel_texture_bits (void)
+{
+  glewPrintExt("GL_SGIX_pixel_texture_bits", GLEW_SGIX_pixel_texture_bits, glewIsSupported("GL_SGIX_pixel_texture_bits"), glewGetExtension("GL_SGIX_pixel_texture_bits"));
+}
+
+#endif /* GL_SGIX_pixel_texture_bits */
+
+#ifdef GL_SGIX_reference_plane
+
+static void _glewInfo_GL_SGIX_reference_plane (void)
+{
+  glewPrintExt("GL_SGIX_reference_plane", GLEW_SGIX_reference_plane, glewIsSupported("GL_SGIX_reference_plane"), glewGetExtension("GL_SGIX_reference_plane"));
+
+  glewInfoFunc("glReferencePlaneSGIX", glReferencePlaneSGIX == NULL);
+}
+
+#endif /* GL_SGIX_reference_plane */
+
+#ifdef GL_SGIX_resample
+
+static void _glewInfo_GL_SGIX_resample (void)
+{
+  glewPrintExt("GL_SGIX_resample", GLEW_SGIX_resample, glewIsSupported("GL_SGIX_resample"), glewGetExtension("GL_SGIX_resample"));
+}
+
+#endif /* GL_SGIX_resample */
+
+#ifdef GL_SGIX_shadow
+
+static void _glewInfo_GL_SGIX_shadow (void)
+{
+  glewPrintExt("GL_SGIX_shadow", GLEW_SGIX_shadow, glewIsSupported("GL_SGIX_shadow"), glewGetExtension("GL_SGIX_shadow"));
+}
+
+#endif /* GL_SGIX_shadow */
+
+#ifdef GL_SGIX_shadow_ambient
+
+static void _glewInfo_GL_SGIX_shadow_ambient (void)
+{
+  glewPrintExt("GL_SGIX_shadow_ambient", GLEW_SGIX_shadow_ambient, glewIsSupported("GL_SGIX_shadow_ambient"), glewGetExtension("GL_SGIX_shadow_ambient"));
+}
+
+#endif /* GL_SGIX_shadow_ambient */
+
+#ifdef GL_SGIX_sprite
+
+static void _glewInfo_GL_SGIX_sprite (void)
+{
+  glewPrintExt("GL_SGIX_sprite", GLEW_SGIX_sprite, glewIsSupported("GL_SGIX_sprite"), glewGetExtension("GL_SGIX_sprite"));
+
+  glewInfoFunc("glSpriteParameterfSGIX", glSpriteParameterfSGIX == NULL);
+  glewInfoFunc("glSpriteParameterfvSGIX", glSpriteParameterfvSGIX == NULL);
+  glewInfoFunc("glSpriteParameteriSGIX", glSpriteParameteriSGIX == NULL);
+  glewInfoFunc("glSpriteParameterivSGIX", glSpriteParameterivSGIX == NULL);
+}
+
+#endif /* GL_SGIX_sprite */
+
+#ifdef GL_SGIX_tag_sample_buffer
+
+static void _glewInfo_GL_SGIX_tag_sample_buffer (void)
+{
+  glewPrintExt("GL_SGIX_tag_sample_buffer", GLEW_SGIX_tag_sample_buffer, glewIsSupported("GL_SGIX_tag_sample_buffer"), glewGetExtension("GL_SGIX_tag_sample_buffer"));
+
+  glewInfoFunc("glTagSampleBufferSGIX", glTagSampleBufferSGIX == NULL);
+}
+
+#endif /* GL_SGIX_tag_sample_buffer */
+
+#ifdef GL_SGIX_texture_add_env
+
+static void _glewInfo_GL_SGIX_texture_add_env (void)
+{
+  glewPrintExt("GL_SGIX_texture_add_env", GLEW_SGIX_texture_add_env, glewIsSupported("GL_SGIX_texture_add_env"), glewGetExtension("GL_SGIX_texture_add_env"));
+}
+
+#endif /* GL_SGIX_texture_add_env */
+
+#ifdef GL_SGIX_texture_coordinate_clamp
+
+static void _glewInfo_GL_SGIX_texture_coordinate_clamp (void)
+{
+  glewPrintExt("GL_SGIX_texture_coordinate_clamp", GLEW_SGIX_texture_coordinate_clamp, glewIsSupported("GL_SGIX_texture_coordinate_clamp"), glewGetExtension("GL_SGIX_texture_coordinate_clamp"));
+}
+
+#endif /* GL_SGIX_texture_coordinate_clamp */
+
+#ifdef GL_SGIX_texture_lod_bias
+
+static void _glewInfo_GL_SGIX_texture_lod_bias (void)
+{
+  glewPrintExt("GL_SGIX_texture_lod_bias", GLEW_SGIX_texture_lod_bias, glewIsSupported("GL_SGIX_texture_lod_bias"), glewGetExtension("GL_SGIX_texture_lod_bias"));
+}
+
+#endif /* GL_SGIX_texture_lod_bias */
+
+#ifdef GL_SGIX_texture_multi_buffer
+
+static void _glewInfo_GL_SGIX_texture_multi_buffer (void)
+{
+  glewPrintExt("GL_SGIX_texture_multi_buffer", GLEW_SGIX_texture_multi_buffer, glewIsSupported("GL_SGIX_texture_multi_buffer"), glewGetExtension("GL_SGIX_texture_multi_buffer"));
+}
+
+#endif /* GL_SGIX_texture_multi_buffer */
+
+#ifdef GL_SGIX_texture_range
+
+static void _glewInfo_GL_SGIX_texture_range (void)
+{
+  glewPrintExt("GL_SGIX_texture_range", GLEW_SGIX_texture_range, glewIsSupported("GL_SGIX_texture_range"), glewGetExtension("GL_SGIX_texture_range"));
+}
+
+#endif /* GL_SGIX_texture_range */
+
+#ifdef GL_SGIX_texture_scale_bias
+
+static void _glewInfo_GL_SGIX_texture_scale_bias (void)
+{
+  glewPrintExt("GL_SGIX_texture_scale_bias", GLEW_SGIX_texture_scale_bias, glewIsSupported("GL_SGIX_texture_scale_bias"), glewGetExtension("GL_SGIX_texture_scale_bias"));
+}
+
+#endif /* GL_SGIX_texture_scale_bias */
+
+#ifdef GL_SGIX_vertex_preclip
+
+static void _glewInfo_GL_SGIX_vertex_preclip (void)
+{
+  glewPrintExt("GL_SGIX_vertex_preclip", GLEW_SGIX_vertex_preclip, glewIsSupported("GL_SGIX_vertex_preclip"), glewGetExtension("GL_SGIX_vertex_preclip"));
+}
+
+#endif /* GL_SGIX_vertex_preclip */
+
+#ifdef GL_SGIX_vertex_preclip_hint
+
+static void _glewInfo_GL_SGIX_vertex_preclip_hint (void)
+{
+  glewPrintExt("GL_SGIX_vertex_preclip_hint", GLEW_SGIX_vertex_preclip_hint, glewIsSupported("GL_SGIX_vertex_preclip_hint"), glewGetExtension("GL_SGIX_vertex_preclip_hint"));
+}
+
+#endif /* GL_SGIX_vertex_preclip_hint */
+
+#ifdef GL_SGIX_ycrcb
+
+static void _glewInfo_GL_SGIX_ycrcb (void)
+{
+  glewPrintExt("GL_SGIX_ycrcb", GLEW_SGIX_ycrcb, glewIsSupported("GL_SGIX_ycrcb"), glewGetExtension("GL_SGIX_ycrcb"));
+}
+
+#endif /* GL_SGIX_ycrcb */
+
+#ifdef GL_SGI_color_matrix
+
+static void _glewInfo_GL_SGI_color_matrix (void)
+{
+  glewPrintExt("GL_SGI_color_matrix", GLEW_SGI_color_matrix, glewIsSupported("GL_SGI_color_matrix"), glewGetExtension("GL_SGI_color_matrix"));
+}
+
+#endif /* GL_SGI_color_matrix */
+
+#ifdef GL_SGI_color_table
+
+static void _glewInfo_GL_SGI_color_table (void)
+{
+  glewPrintExt("GL_SGI_color_table", GLEW_SGI_color_table, glewIsSupported("GL_SGI_color_table"), glewGetExtension("GL_SGI_color_table"));
+
+  glewInfoFunc("glColorTableParameterfvSGI", glColorTableParameterfvSGI == NULL);
+  glewInfoFunc("glColorTableParameterivSGI", glColorTableParameterivSGI == NULL);
+  glewInfoFunc("glColorTableSGI", glColorTableSGI == NULL);
+  glewInfoFunc("glCopyColorTableSGI", glCopyColorTableSGI == NULL);
+  glewInfoFunc("glGetColorTableParameterfvSGI", glGetColorTableParameterfvSGI == NULL);
+  glewInfoFunc("glGetColorTableParameterivSGI", glGetColorTableParameterivSGI == NULL);
+  glewInfoFunc("glGetColorTableSGI", glGetColorTableSGI == NULL);
+}
+
+#endif /* GL_SGI_color_table */
+
+#ifdef GL_SGI_texture_color_table
+
+static void _glewInfo_GL_SGI_texture_color_table (void)
+{
+  glewPrintExt("GL_SGI_texture_color_table", GLEW_SGI_texture_color_table, glewIsSupported("GL_SGI_texture_color_table"), glewGetExtension("GL_SGI_texture_color_table"));
+}
+
+#endif /* GL_SGI_texture_color_table */
+
+#ifdef GL_SUNX_constant_data
+
+static void _glewInfo_GL_SUNX_constant_data (void)
+{
+  glewPrintExt("GL_SUNX_constant_data", GLEW_SUNX_constant_data, glewIsSupported("GL_SUNX_constant_data"), glewGetExtension("GL_SUNX_constant_data"));
+
+  glewInfoFunc("glFinishTextureSUNX", glFinishTextureSUNX == NULL);
+}
+
+#endif /* GL_SUNX_constant_data */
+
+#ifdef GL_SUN_convolution_border_modes
+
+static void _glewInfo_GL_SUN_convolution_border_modes (void)
+{
+  glewPrintExt("GL_SUN_convolution_border_modes", GLEW_SUN_convolution_border_modes, glewIsSupported("GL_SUN_convolution_border_modes"), glewGetExtension("GL_SUN_convolution_border_modes"));
+}
+
+#endif /* GL_SUN_convolution_border_modes */
+
+#ifdef GL_SUN_global_alpha
+
+static void _glewInfo_GL_SUN_global_alpha (void)
+{
+  glewPrintExt("GL_SUN_global_alpha", GLEW_SUN_global_alpha, glewIsSupported("GL_SUN_global_alpha"), glewGetExtension("GL_SUN_global_alpha"));
+
+  glewInfoFunc("glGlobalAlphaFactorbSUN", glGlobalAlphaFactorbSUN == NULL);
+  glewInfoFunc("glGlobalAlphaFactordSUN", glGlobalAlphaFactordSUN == NULL);
+  glewInfoFunc("glGlobalAlphaFactorfSUN", glGlobalAlphaFactorfSUN == NULL);
+  glewInfoFunc("glGlobalAlphaFactoriSUN", glGlobalAlphaFactoriSUN == NULL);
+  glewInfoFunc("glGlobalAlphaFactorsSUN", glGlobalAlphaFactorsSUN == NULL);
+  glewInfoFunc("glGlobalAlphaFactorubSUN", glGlobalAlphaFactorubSUN == NULL);
+  glewInfoFunc("glGlobalAlphaFactoruiSUN", glGlobalAlphaFactoruiSUN == NULL);
+  glewInfoFunc("glGlobalAlphaFactorusSUN", glGlobalAlphaFactorusSUN == NULL);
+}
+
+#endif /* GL_SUN_global_alpha */
+
+#ifdef GL_SUN_mesh_array
+
+static void _glewInfo_GL_SUN_mesh_array (void)
+{
+  glewPrintExt("GL_SUN_mesh_array", GLEW_SUN_mesh_array, glewIsSupported("GL_SUN_mesh_array"), glewGetExtension("GL_SUN_mesh_array"));
+}
+
+#endif /* GL_SUN_mesh_array */
+
+#ifdef GL_SUN_read_video_pixels
+
+static void _glewInfo_GL_SUN_read_video_pixels (void)
+{
+  glewPrintExt("GL_SUN_read_video_pixels", GLEW_SUN_read_video_pixels, glewIsSupported("GL_SUN_read_video_pixels"), glewGetExtension("GL_SUN_read_video_pixels"));
+
+  glewInfoFunc("glReadVideoPixelsSUN", glReadVideoPixelsSUN == NULL);
+}
+
+#endif /* GL_SUN_read_video_pixels */
+
+#ifdef GL_SUN_slice_accum
+
+static void _glewInfo_GL_SUN_slice_accum (void)
+{
+  glewPrintExt("GL_SUN_slice_accum", GLEW_SUN_slice_accum, glewIsSupported("GL_SUN_slice_accum"), glewGetExtension("GL_SUN_slice_accum"));
+}
+
+#endif /* GL_SUN_slice_accum */
+
+#ifdef GL_SUN_triangle_list
+
+static void _glewInfo_GL_SUN_triangle_list (void)
+{
+  glewPrintExt("GL_SUN_triangle_list", GLEW_SUN_triangle_list, glewIsSupported("GL_SUN_triangle_list"), glewGetExtension("GL_SUN_triangle_list"));
+
+  glewInfoFunc("glReplacementCodePointerSUN", glReplacementCodePointerSUN == NULL);
+  glewInfoFunc("glReplacementCodeubSUN", glReplacementCodeubSUN == NULL);
+  glewInfoFunc("glReplacementCodeubvSUN", glReplacementCodeubvSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiSUN", glReplacementCodeuiSUN == NULL);
+  glewInfoFunc("glReplacementCodeuivSUN", glReplacementCodeuivSUN == NULL);
+  glewInfoFunc("glReplacementCodeusSUN", glReplacementCodeusSUN == NULL);
+  glewInfoFunc("glReplacementCodeusvSUN", glReplacementCodeusvSUN == NULL);
+}
+
+#endif /* GL_SUN_triangle_list */
+
+#ifdef GL_SUN_vertex
+
+static void _glewInfo_GL_SUN_vertex (void)
+{
+  glewPrintExt("GL_SUN_vertex", GLEW_SUN_vertex, glewIsSupported("GL_SUN_vertex"), glewGetExtension("GL_SUN_vertex"));
+
+  glewInfoFunc("glColor3fVertex3fSUN", glColor3fVertex3fSUN == NULL);
+  glewInfoFunc("glColor3fVertex3fvSUN", glColor3fVertex3fvSUN == NULL);
+  glewInfoFunc("glColor4fNormal3fVertex3fSUN", glColor4fNormal3fVertex3fSUN == NULL);
+  glewInfoFunc("glColor4fNormal3fVertex3fvSUN", glColor4fNormal3fVertex3fvSUN == NULL);
+  glewInfoFunc("glColor4ubVertex2fSUN", glColor4ubVertex2fSUN == NULL);
+  glewInfoFunc("glColor4ubVertex2fvSUN", glColor4ubVertex2fvSUN == NULL);
+  glewInfoFunc("glColor4ubVertex3fSUN", glColor4ubVertex3fSUN == NULL);
+  glewInfoFunc("glColor4ubVertex3fvSUN", glColor4ubVertex3fvSUN == NULL);
+  glewInfoFunc("glNormal3fVertex3fSUN", glNormal3fVertex3fSUN == NULL);
+  glewInfoFunc("glNormal3fVertex3fvSUN", glNormal3fVertex3fvSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiColor3fVertex3fSUN", glReplacementCodeuiColor3fVertex3fSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiColor3fVertex3fvSUN", glReplacementCodeuiColor3fVertex3fvSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiColor4fNormal3fVertex3fSUN", glReplacementCodeuiColor4fNormal3fVertex3fSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiColor4fNormal3fVertex3fvSUN", glReplacementCodeuiColor4fNormal3fVertex3fvSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiColor4ubVertex3fSUN", glReplacementCodeuiColor4ubVertex3fSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiColor4ubVertex3fvSUN", glReplacementCodeuiColor4ubVertex3fvSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiNormal3fVertex3fSUN", glReplacementCodeuiNormal3fVertex3fSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiNormal3fVertex3fvSUN", glReplacementCodeuiNormal3fVertex3fvSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN", glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN", glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiTexCoord2fNormal3fVertex3fSUN", glReplacementCodeuiTexCoord2fNormal3fVertex3fSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN", glReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiTexCoord2fVertex3fSUN", glReplacementCodeuiTexCoord2fVertex3fSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiTexCoord2fVertex3fvSUN", glReplacementCodeuiTexCoord2fVertex3fvSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiVertex3fSUN", glReplacementCodeuiVertex3fSUN == NULL);
+  glewInfoFunc("glReplacementCodeuiVertex3fvSUN", glReplacementCodeuiVertex3fvSUN == NULL);
+  glewInfoFunc("glTexCoord2fColor3fVertex3fSUN", glTexCoord2fColor3fVertex3fSUN == NULL);
+  glewInfoFunc("glTexCoord2fColor3fVertex3fvSUN", glTexCoord2fColor3fVertex3fvSUN == NULL);
+  glewInfoFunc("glTexCoord2fColor4fNormal3fVertex3fSUN", glTexCoord2fColor4fNormal3fVertex3fSUN == NULL);
+  glewInfoFunc("glTexCoord2fColor4fNormal3fVertex3fvSUN", glTexCoord2fColor4fNormal3fVertex3fvSUN == NULL);
+  glewInfoFunc("glTexCoord2fColor4ubVertex3fSUN", glTexCoord2fColor4ubVertex3fSUN == NULL);
+  glewInfoFunc("glTexCoord2fColor4ubVertex3fvSUN", glTexCoord2fColor4ubVertex3fvSUN == NULL);
+  glewInfoFunc("glTexCoord2fNormal3fVertex3fSUN", glTexCoord2fNormal3fVertex3fSUN == NULL);
+  glewInfoFunc("glTexCoord2fNormal3fVertex3fvSUN", glTexCoord2fNormal3fVertex3fvSUN == NULL);
+  glewInfoFunc("glTexCoord2fVertex3fSUN", glTexCoord2fVertex3fSUN == NULL);
+  glewInfoFunc("glTexCoord2fVertex3fvSUN", glTexCoord2fVertex3fvSUN == NULL);
+  glewInfoFunc("glTexCoord4fColor4fNormal3fVertex4fSUN", glTexCoord4fColor4fNormal3fVertex4fSUN == NULL);
+  glewInfoFunc("glTexCoord4fColor4fNormal3fVertex4fvSUN", glTexCoord4fColor4fNormal3fVertex4fvSUN == NULL);
+  glewInfoFunc("glTexCoord4fVertex4fSUN", glTexCoord4fVertex4fSUN == NULL);
+  glewInfoFunc("glTexCoord4fVertex4fvSUN", glTexCoord4fVertex4fvSUN == NULL);
+}
+
+#endif /* GL_SUN_vertex */
+
+#ifdef GL_WIN_phong_shading
+
+static void _glewInfo_GL_WIN_phong_shading (void)
+{
+  glewPrintExt("GL_WIN_phong_shading", GLEW_WIN_phong_shading, glewIsSupported("GL_WIN_phong_shading"), glewGetExtension("GL_WIN_phong_shading"));
+}
+
+#endif /* GL_WIN_phong_shading */
+
+#ifdef GL_WIN_specular_fog
+
+static void _glewInfo_GL_WIN_specular_fog (void)
+{
+  glewPrintExt("GL_WIN_specular_fog", GLEW_WIN_specular_fog, glewIsSupported("GL_WIN_specular_fog"), glewGetExtension("GL_WIN_specular_fog"));
+}
+
+#endif /* GL_WIN_specular_fog */
+
+#ifdef GL_WIN_swap_hint
+
+static void _glewInfo_GL_WIN_swap_hint (void)
+{
+  glewPrintExt("GL_WIN_swap_hint", GLEW_WIN_swap_hint, glewIsSupported("GL_WIN_swap_hint"), glewGetExtension("GL_WIN_swap_hint"));
+
+  glewInfoFunc("glAddSwapHintRectWIN", glAddSwapHintRectWIN == NULL);
+}
+
+#endif /* GL_WIN_swap_hint */
+
+#ifdef _WIN32
+
+#ifdef WGL_3DFX_multisample
+
+static void _glewInfo_WGL_3DFX_multisample (void)
+{
+  glewPrintExt("WGL_3DFX_multisample", WGLEW_3DFX_multisample, wglewIsSupported("WGL_3DFX_multisample"), wglewGetExtension("WGL_3DFX_multisample"));
+}
+
+#endif /* WGL_3DFX_multisample */
+
+#ifdef WGL_3DL_stereo_control
+
+static void _glewInfo_WGL_3DL_stereo_control (void)
+{
+  glewPrintExt("WGL_3DL_stereo_control", WGLEW_3DL_stereo_control, wglewIsSupported("WGL_3DL_stereo_control"), wglewGetExtension("WGL_3DL_stereo_control"));
+
+  glewInfoFunc("wglSetStereoEmitterState3DL", wglSetStereoEmitterState3DL == NULL);
+}
+
+#endif /* WGL_3DL_stereo_control */
+
+#ifdef WGL_ARB_buffer_region
+
+static void _glewInfo_WGL_ARB_buffer_region (void)
+{
+  glewPrintExt("WGL_ARB_buffer_region", WGLEW_ARB_buffer_region, wglewIsSupported("WGL_ARB_buffer_region"), wglewGetExtension("WGL_ARB_buffer_region"));
+
+  glewInfoFunc("wglCreateBufferRegionARB", wglCreateBufferRegionARB == NULL);
+  glewInfoFunc("wglDeleteBufferRegionARB", wglDeleteBufferRegionARB == NULL);
+  glewInfoFunc("wglRestoreBufferRegionARB", wglRestoreBufferRegionARB == NULL);
+  glewInfoFunc("wglSaveBufferRegionARB", wglSaveBufferRegionARB == NULL);
+}
+
+#endif /* WGL_ARB_buffer_region */
+
+#ifdef WGL_ARB_create_context
+
+static void _glewInfo_WGL_ARB_create_context (void)
+{
+  glewPrintExt("WGL_ARB_create_context", WGLEW_ARB_create_context, wglewIsSupported("WGL_ARB_create_context"), wglewGetExtension("WGL_ARB_create_context"));
+
+  glewInfoFunc("wglCreateContextAttribsARB", wglCreateContextAttribsARB == NULL);
+}
+
+#endif /* WGL_ARB_create_context */
+
+#ifdef WGL_ARB_extensions_string
+
+static void _glewInfo_WGL_ARB_extensions_string (void)
+{
+  glewPrintExt("WGL_ARB_extensions_string", WGLEW_ARB_extensions_string, wglewIsSupported("WGL_ARB_extensions_string"), wglewGetExtension("WGL_ARB_extensions_string"));
+
+  glewInfoFunc("wglGetExtensionsStringARB", wglGetExtensionsStringARB == NULL);
+}
+
+#endif /* WGL_ARB_extensions_string */
+
+#ifdef WGL_ARB_framebuffer_sRGB
+
+static void _glewInfo_WGL_ARB_framebuffer_sRGB (void)
+{
+  glewPrintExt("WGL_ARB_framebuffer_sRGB", WGLEW_ARB_framebuffer_sRGB, wglewIsSupported("WGL_ARB_framebuffer_sRGB"), wglewGetExtension("WGL_ARB_framebuffer_sRGB"));
+}
+
+#endif /* WGL_ARB_framebuffer_sRGB */
+
+#ifdef WGL_ARB_make_current_read
+
+static void _glewInfo_WGL_ARB_make_current_read (void)
+{
+  glewPrintExt("WGL_ARB_make_current_read", WGLEW_ARB_make_current_read, wglewIsSupported("WGL_ARB_make_current_read"), wglewGetExtension("WGL_ARB_make_current_read"));
+
+  glewInfoFunc("wglGetCurrentReadDCARB", wglGetCurrentReadDCARB == NULL);
+  glewInfoFunc("wglMakeContextCurrentARB", wglMakeContextCurrentARB == NULL);
+}
+
+#endif /* WGL_ARB_make_current_read */
+
+#ifdef WGL_ARB_multisample
+
+static void _glewInfo_WGL_ARB_multisample (void)
+{
+  glewPrintExt("WGL_ARB_multisample", WGLEW_ARB_multisample, wglewIsSupported("WGL_ARB_multisample"), wglewGetExtension("WGL_ARB_multisample"));
+}
+
+#endif /* WGL_ARB_multisample */
+
+#ifdef WGL_ARB_pbuffer
+
+static void _glewInfo_WGL_ARB_pbuffer (void)
+{
+  glewPrintExt("WGL_ARB_pbuffer", WGLEW_ARB_pbuffer, wglewIsSupported("WGL_ARB_pbuffer"), wglewGetExtension("WGL_ARB_pbuffer"));
+
+  glewInfoFunc("wglCreatePbufferARB", wglCreatePbufferARB == NULL);
+  glewInfoFunc("wglDestroyPbufferARB", wglDestroyPbufferARB == NULL);
+  glewInfoFunc("wglGetPbufferDCARB", wglGetPbufferDCARB == NULL);
+  glewInfoFunc("wglQueryPbufferARB", wglQueryPbufferARB == NULL);
+  glewInfoFunc("wglReleasePbufferDCARB", wglReleasePbufferDCARB == NULL);
+}
+
+#endif /* WGL_ARB_pbuffer */
+
+#ifdef WGL_ARB_pixel_format
+
+static void _glewInfo_WGL_ARB_pixel_format (void)
+{
+  glewPrintExt("WGL_ARB_pixel_format", WGLEW_ARB_pixel_format, wglewIsSupported("WGL_ARB_pixel_format"), wglewGetExtension("WGL_ARB_pixel_format"));
+
+  glewInfoFunc("wglChoosePixelFormatARB", wglChoosePixelFormatARB == NULL);
+  glewInfoFunc("wglGetPixelFormatAttribfvARB", wglGetPixelFormatAttribfvARB == NULL);
+  glewInfoFunc("wglGetPixelFormatAttribivARB", wglGetPixelFormatAttribivARB == NULL);
+}
+
+#endif /* WGL_ARB_pixel_format */
+
+#ifdef WGL_ARB_pixel_format_float
+
+static void _glewInfo_WGL_ARB_pixel_format_float (void)
+{
+  glewPrintExt("WGL_ARB_pixel_format_float", WGLEW_ARB_pixel_format_float, wglewIsSupported("WGL_ARB_pixel_format_float"), wglewGetExtension("WGL_ARB_pixel_format_float"));
+}
+
+#endif /* WGL_ARB_pixel_format_float */
+
+#ifdef WGL_ARB_render_texture
+
+static void _glewInfo_WGL_ARB_render_texture (void)
+{
+  glewPrintExt("WGL_ARB_render_texture", WGLEW_ARB_render_texture, wglewIsSupported("WGL_ARB_render_texture"), wglewGetExtension("WGL_ARB_render_texture"));
+
+  glewInfoFunc("wglBindTexImageARB", wglBindTexImageARB == NULL);
+  glewInfoFunc("wglReleaseTexImageARB", wglReleaseTexImageARB == NULL);
+  glewInfoFunc("wglSetPbufferAttribARB", wglSetPbufferAttribARB == NULL);
+}
+
+#endif /* WGL_ARB_render_texture */
+
+#ifdef WGL_ATI_pixel_format_float
+
+static void _glewInfo_WGL_ATI_pixel_format_float (void)
+{
+  glewPrintExt("WGL_ATI_pixel_format_float", WGLEW_ATI_pixel_format_float, wglewIsSupported("WGL_ATI_pixel_format_float"), wglewGetExtension("WGL_ATI_pixel_format_float"));
+}
+
+#endif /* WGL_ATI_pixel_format_float */
+
+#ifdef WGL_ATI_render_texture_rectangle
+
+static void _glewInfo_WGL_ATI_render_texture_rectangle (void)
+{
+  glewPrintExt("WGL_ATI_render_texture_rectangle", WGLEW_ATI_render_texture_rectangle, wglewIsSupported("WGL_ATI_render_texture_rectangle"), wglewGetExtension("WGL_ATI_render_texture_rectangle"));
+}
+
+#endif /* WGL_ATI_render_texture_rectangle */
+
+#ifdef WGL_EXT_depth_float
+
+static void _glewInfo_WGL_EXT_depth_float (void)
+{
+  glewPrintExt("WGL_EXT_depth_float", WGLEW_EXT_depth_float, wglewIsSupported("WGL_EXT_depth_float"), wglewGetExtension("WGL_EXT_depth_float"));
+}
+
+#endif /* WGL_EXT_depth_float */
+
+#ifdef WGL_EXT_display_color_table
+
+static void _glewInfo_WGL_EXT_display_color_table (void)
+{
+  glewPrintExt("WGL_EXT_display_color_table", WGLEW_EXT_display_color_table, wglewIsSupported("WGL_EXT_display_color_table"), wglewGetExtension("WGL_EXT_display_color_table"));
+
+  glewInfoFunc("wglBindDisplayColorTableEXT", wglBindDisplayColorTableEXT == NULL);
+  glewInfoFunc("wglCreateDisplayColorTableEXT", wglCreateDisplayColorTableEXT == NULL);
+  glewInfoFunc("wglDestroyDisplayColorTableEXT", wglDestroyDisplayColorTableEXT == NULL);
+  glewInfoFunc("wglLoadDisplayColorTableEXT", wglLoadDisplayColorTableEXT == NULL);
+}
+
+#endif /* WGL_EXT_display_color_table */
+
+#ifdef WGL_EXT_extensions_string
+
+static void _glewInfo_WGL_EXT_extensions_string (void)
+{
+  glewPrintExt("WGL_EXT_extensions_string", WGLEW_EXT_extensions_string, wglewIsSupported("WGL_EXT_extensions_string"), wglewGetExtension("WGL_EXT_extensions_string"));
+
+  glewInfoFunc("wglGetExtensionsStringEXT", wglGetExtensionsStringEXT == NULL);
+}
+
+#endif /* WGL_EXT_extensions_string */
+
+#ifdef WGL_EXT_framebuffer_sRGB
+
+static void _glewInfo_WGL_EXT_framebuffer_sRGB (void)
+{
+  glewPrintExt("WGL_EXT_framebuffer_sRGB", WGLEW_EXT_framebuffer_sRGB, wglewIsSupported("WGL_EXT_framebuffer_sRGB"), wglewGetExtension("WGL_EXT_framebuffer_sRGB"));
+}
+
+#endif /* WGL_EXT_framebuffer_sRGB */
+
+#ifdef WGL_EXT_make_current_read
+
+static void _glewInfo_WGL_EXT_make_current_read (void)
+{
+  glewPrintExt("WGL_EXT_make_current_read", WGLEW_EXT_make_current_read, wglewIsSupported("WGL_EXT_make_current_read"), wglewGetExtension("WGL_EXT_make_current_read"));
+
+  glewInfoFunc("wglGetCurrentReadDCEXT", wglGetCurrentReadDCEXT == NULL);
+  glewInfoFunc("wglMakeContextCurrentEXT", wglMakeContextCurrentEXT == NULL);
+}
+
+#endif /* WGL_EXT_make_current_read */
+
+#ifdef WGL_EXT_multisample
+
+static void _glewInfo_WGL_EXT_multisample (void)
+{
+  glewPrintExt("WGL_EXT_multisample", WGLEW_EXT_multisample, wglewIsSupported("WGL_EXT_multisample"), wglewGetExtension("WGL_EXT_multisample"));
+}
+
+#endif /* WGL_EXT_multisample */
+
+#ifdef WGL_EXT_pbuffer
+
+static void _glewInfo_WGL_EXT_pbuffer (void)
+{
+  glewPrintExt("WGL_EXT_pbuffer", WGLEW_EXT_pbuffer, wglewIsSupported("WGL_EXT_pbuffer"), wglewGetExtension("WGL_EXT_pbuffer"));
+
+  glewInfoFunc("wglCreatePbufferEXT", wglCreatePbufferEXT == NULL);
+  glewInfoFunc("wglDestroyPbufferEXT", wglDestroyPbufferEXT == NULL);
+  glewInfoFunc("wglGetPbufferDCEXT", wglGetPbufferDCEXT == NULL);
+  glewInfoFunc("wglQueryPbufferEXT", wglQueryPbufferEXT == NULL);
+  glewInfoFunc("wglReleasePbufferDCEXT", wglReleasePbufferDCEXT == NULL);
+}
+
+#endif /* WGL_EXT_pbuffer */
+
+#ifdef WGL_EXT_pixel_format
+
+static void _glewInfo_WGL_EXT_pixel_format (void)
+{
+  glewPrintExt("WGL_EXT_pixel_format", WGLEW_EXT_pixel_format, wglewIsSupported("WGL_EXT_pixel_format"), wglewGetExtension("WGL_EXT_pixel_format"));
+
+  glewInfoFunc("wglChoosePixelFormatEXT", wglChoosePixelFormatEXT == NULL);
+  glewInfoFunc("wglGetPixelFormatAttribfvEXT", wglGetPixelFormatAttribfvEXT == NULL);
+  glewInfoFunc("wglGetPixelFormatAttribivEXT", wglGetPixelFormatAttribivEXT == NULL);
+}
+
+#endif /* WGL_EXT_pixel_format */
+
+#ifdef WGL_EXT_pixel_format_packed_float
+
+static void _glewInfo_WGL_EXT_pixel_format_packed_float (void)
+{
+  glewPrintExt("WGL_EXT_pixel_format_packed_float", WGLEW_EXT_pixel_format_packed_float, wglewIsSupported("WGL_EXT_pixel_format_packed_float"), wglewGetExtension("WGL_EXT_pixel_format_packed_float"));
+}
+
+#endif /* WGL_EXT_pixel_format_packed_float */
+
+#ifdef WGL_EXT_swap_control
+
+static void _glewInfo_WGL_EXT_swap_control (void)
+{
+  glewPrintExt("WGL_EXT_swap_control", WGLEW_EXT_swap_control, wglewIsSupported("WGL_EXT_swap_control"), wglewGetExtension("WGL_EXT_swap_control"));
+
+  glewInfoFunc("wglGetSwapIntervalEXT", wglGetSwapIntervalEXT == NULL);
+  glewInfoFunc("wglSwapIntervalEXT", wglSwapIntervalEXT == NULL);
+}
+
+#endif /* WGL_EXT_swap_control */
+
+#ifdef WGL_I3D_digital_video_control
+
+static void _glewInfo_WGL_I3D_digital_video_control (void)
+{
+  glewPrintExt("WGL_I3D_digital_video_control", WGLEW_I3D_digital_video_control, wglewIsSupported("WGL_I3D_digital_video_control"), wglewGetExtension("WGL_I3D_digital_video_control"));
+
+  glewInfoFunc("wglGetDigitalVideoParametersI3D", wglGetDigitalVideoParametersI3D == NULL);
+  glewInfoFunc("wglSetDigitalVideoParametersI3D", wglSetDigitalVideoParametersI3D == NULL);
+}
+
+#endif /* WGL_I3D_digital_video_control */
+
+#ifdef WGL_I3D_gamma
+
+static void _glewInfo_WGL_I3D_gamma (void)
+{
+  glewPrintExt("WGL_I3D_gamma", WGLEW_I3D_gamma, wglewIsSupported("WGL_I3D_gamma"), wglewGetExtension("WGL_I3D_gamma"));
+
+  glewInfoFunc("wglGetGammaTableI3D", wglGetGammaTableI3D == NULL);
+  glewInfoFunc("wglGetGammaTableParametersI3D", wglGetGammaTableParametersI3D == NULL);
+  glewInfoFunc("wglSetGammaTableI3D", wglSetGammaTableI3D == NULL);
+  glewInfoFunc("wglSetGammaTableParametersI3D", wglSetGammaTableParametersI3D == NULL);
+}
+
+#endif /* WGL_I3D_gamma */
+
+#ifdef WGL_I3D_genlock
+
+static void _glewInfo_WGL_I3D_genlock (void)
+{
+  glewPrintExt("WGL_I3D_genlock", WGLEW_I3D_genlock, wglewIsSupported("WGL_I3D_genlock"), wglewGetExtension("WGL_I3D_genlock"));
+
+  glewInfoFunc("wglDisableGenlockI3D", wglDisableGenlockI3D == NULL);
+  glewInfoFunc("wglEnableGenlockI3D", wglEnableGenlockI3D == NULL);
+  glewInfoFunc("wglGenlockSampleRateI3D", wglGenlockSampleRateI3D == NULL);
+  glewInfoFunc("wglGenlockSourceDelayI3D", wglGenlockSourceDelayI3D == NULL);
+  glewInfoFunc("wglGenlockSourceEdgeI3D", wglGenlockSourceEdgeI3D == NULL);
+  glewInfoFunc("wglGenlockSourceI3D", wglGenlockSourceI3D == NULL);
+  glewInfoFunc("wglGetGenlockSampleRateI3D", wglGetGenlockSampleRateI3D == NULL);
+  glewInfoFunc("wglGetGenlockSourceDelayI3D", wglGetGenlockSourceDelayI3D == NULL);
+  glewInfoFunc("wglGetGenlockSourceEdgeI3D", wglGetGenlockSourceEdgeI3D == NULL);
+  glewInfoFunc("wglGetGenlockSourceI3D", wglGetGenlockSourceI3D == NULL);
+  glewInfoFunc("wglIsEnabledGenlockI3D", wglIsEnabledGenlockI3D == NULL);
+  glewInfoFunc("wglQueryGenlockMaxSourceDelayI3D", wglQueryGenlockMaxSourceDelayI3D == NULL);
+}
+
+#endif /* WGL_I3D_genlock */
+
+#ifdef WGL_I3D_image_buffer
+
+static void _glewInfo_WGL_I3D_image_buffer (void)
+{
+  glewPrintExt("WGL_I3D_image_buffer", WGLEW_I3D_image_buffer, wglewIsSupported("WGL_I3D_image_buffer"), wglewGetExtension("WGL_I3D_image_buffer"));
+
+  glewInfoFunc("wglAssociateImageBufferEventsI3D", wglAssociateImageBufferEventsI3D == NULL);
+  glewInfoFunc("wglCreateImageBufferI3D", wglCreateImageBufferI3D == NULL);
+  glewInfoFunc("wglDestroyImageBufferI3D", wglDestroyImageBufferI3D == NULL);
+  glewInfoFunc("wglReleaseImageBufferEventsI3D", wglReleaseImageBufferEventsI3D == NULL);
+}
+
+#endif /* WGL_I3D_image_buffer */
+
+#ifdef WGL_I3D_swap_frame_lock
+
+static void _glewInfo_WGL_I3D_swap_frame_lock (void)
+{
+  glewPrintExt("WGL_I3D_swap_frame_lock", WGLEW_I3D_swap_frame_lock, wglewIsSupported("WGL_I3D_swap_frame_lock"), wglewGetExtension("WGL_I3D_swap_frame_lock"));
+
+  glewInfoFunc("wglDisableFrameLockI3D", wglDisableFrameLockI3D == NULL);
+  glewInfoFunc("wglEnableFrameLockI3D", wglEnableFrameLockI3D == NULL);
+  glewInfoFunc("wglIsEnabledFrameLockI3D", wglIsEnabledFrameLockI3D == NULL);
+  glewInfoFunc("wglQueryFrameLockMasterI3D", wglQueryFrameLockMasterI3D == NULL);
+}
+
+#endif /* WGL_I3D_swap_frame_lock */
+
+#ifdef WGL_I3D_swap_frame_usage
+
+static void _glewInfo_WGL_I3D_swap_frame_usage (void)
+{
+  glewPrintExt("WGL_I3D_swap_frame_usage", WGLEW_I3D_swap_frame_usage, wglewIsSupported("WGL_I3D_swap_frame_usage"), wglewGetExtension("WGL_I3D_swap_frame_usage"));
+
+  glewInfoFunc("wglBeginFrameTrackingI3D", wglBeginFrameTrackingI3D == NULL);
+  glewInfoFunc("wglEndFrameTrackingI3D", wglEndFrameTrackingI3D == NULL);
+  glewInfoFunc("wglGetFrameUsageI3D", wglGetFrameUsageI3D == NULL);
+  glewInfoFunc("wglQueryFrameTrackingI3D", wglQueryFrameTrackingI3D == NULL);
+}
+
+#endif /* WGL_I3D_swap_frame_usage */
+
+#ifdef WGL_NV_float_buffer
+
+static void _glewInfo_WGL_NV_float_buffer (void)
+{
+  glewPrintExt("WGL_NV_float_buffer", WGLEW_NV_float_buffer, wglewIsSupported("WGL_NV_float_buffer"), wglewGetExtension("WGL_NV_float_buffer"));
+}
+
+#endif /* WGL_NV_float_buffer */
+
+#ifdef WGL_NV_gpu_affinity
+
+static void _glewInfo_WGL_NV_gpu_affinity (void)
+{
+  glewPrintExt("WGL_NV_gpu_affinity", WGLEW_NV_gpu_affinity, wglewIsSupported("WGL_NV_gpu_affinity"), wglewGetExtension("WGL_NV_gpu_affinity"));
+
+  glewInfoFunc("wglCreateAffinityDCNV", wglCreateAffinityDCNV == NULL);
+  glewInfoFunc("wglDeleteDCNV", wglDeleteDCNV == NULL);
+  glewInfoFunc("wglEnumGpuDevicesNV", wglEnumGpuDevicesNV == NULL);
+  glewInfoFunc("wglEnumGpusFromAffinityDCNV", wglEnumGpusFromAffinityDCNV == NULL);
+  glewInfoFunc("wglEnumGpusNV", wglEnumGpusNV == NULL);
+}
+
+#endif /* WGL_NV_gpu_affinity */
+
+#ifdef WGL_NV_present_video
+
+static void _glewInfo_WGL_NV_present_video (void)
+{
+  glewPrintExt("WGL_NV_present_video", WGLEW_NV_present_video, wglewIsSupported("WGL_NV_present_video"), wglewGetExtension("WGL_NV_present_video"));
+
+  glewInfoFunc("wglBindVideoDeviceNV", wglBindVideoDeviceNV == NULL);
+  glewInfoFunc("wglEnumerateVideoDevicesNV", wglEnumerateVideoDevicesNV == NULL);
+  glewInfoFunc("wglQueryCurrentContextNV", wglQueryCurrentContextNV == NULL);
+}
+
+#endif /* WGL_NV_present_video */
+
+#ifdef WGL_NV_render_depth_texture
+
+static void _glewInfo_WGL_NV_render_depth_texture (void)
+{
+  glewPrintExt("WGL_NV_render_depth_texture", WGLEW_NV_render_depth_texture, wglewIsSupported("WGL_NV_render_depth_texture"), wglewGetExtension("WGL_NV_render_depth_texture"));
+}
+
+#endif /* WGL_NV_render_depth_texture */
+
+#ifdef WGL_NV_render_texture_rectangle
+
+static void _glewInfo_WGL_NV_render_texture_rectangle (void)
+{
+  glewPrintExt("WGL_NV_render_texture_rectangle", WGLEW_NV_render_texture_rectangle, wglewIsSupported("WGL_NV_render_texture_rectangle"), wglewGetExtension("WGL_NV_render_texture_rectangle"));
+}
+
+#endif /* WGL_NV_render_texture_rectangle */
+
+#ifdef WGL_NV_swap_group
+
+static void _glewInfo_WGL_NV_swap_group (void)
+{
+  glewPrintExt("WGL_NV_swap_group", WGLEW_NV_swap_group, wglewIsSupported("WGL_NV_swap_group"), wglewGetExtension("WGL_NV_swap_group"));
+
+  glewInfoFunc("wglBindSwapBarrierNV", wglBindSwapBarrierNV == NULL);
+  glewInfoFunc("wglJoinSwapGroupNV", wglJoinSwapGroupNV == NULL);
+  glewInfoFunc("wglQueryFrameCountNV", wglQueryFrameCountNV == NULL);
+  glewInfoFunc("wglQueryMaxSwapGroupsNV", wglQueryMaxSwapGroupsNV == NULL);
+  glewInfoFunc("wglQuerySwapGroupNV", wglQuerySwapGroupNV == NULL);
+  glewInfoFunc("wglResetFrameCountNV", wglResetFrameCountNV == NULL);
+}
+
+#endif /* WGL_NV_swap_group */
+
+#ifdef WGL_NV_vertex_array_range
+
+static void _glewInfo_WGL_NV_vertex_array_range (void)
+{
+  glewPrintExt("WGL_NV_vertex_array_range", WGLEW_NV_vertex_array_range, wglewIsSupported("WGL_NV_vertex_array_range"), wglewGetExtension("WGL_NV_vertex_array_range"));
+
+  glewInfoFunc("wglAllocateMemoryNV", wglAllocateMemoryNV == NULL);
+  glewInfoFunc("wglFreeMemoryNV", wglFreeMemoryNV == NULL);
+}
+
+#endif /* WGL_NV_vertex_array_range */
+
+#ifdef WGL_NV_video_output
+
+static void _glewInfo_WGL_NV_video_output (void)
+{
+  glewPrintExt("WGL_NV_video_output", WGLEW_NV_video_output, wglewIsSupported("WGL_NV_video_output"), wglewGetExtension("WGL_NV_video_output"));
+
+  glewInfoFunc("wglBindVideoImageNV", wglBindVideoImageNV == NULL);
+  glewInfoFunc("wglGetVideoDeviceNV", wglGetVideoDeviceNV == NULL);
+  glewInfoFunc("wglGetVideoInfoNV", wglGetVideoInfoNV == NULL);
+  glewInfoFunc("wglReleaseVideoDeviceNV", wglReleaseVideoDeviceNV == NULL);
+  glewInfoFunc("wglReleaseVideoImageNV", wglReleaseVideoImageNV == NULL);
+  glewInfoFunc("wglSendPbufferToVideoNV", wglSendPbufferToVideoNV == NULL);
+}
+
+#endif /* WGL_NV_video_output */
+
+#ifdef WGL_OML_sync_control
+
+static void _glewInfo_WGL_OML_sync_control (void)
+{
+  glewPrintExt("WGL_OML_sync_control", WGLEW_OML_sync_control, wglewIsSupported("WGL_OML_sync_control"), wglewGetExtension("WGL_OML_sync_control"));
+
+  glewInfoFunc("wglGetMscRateOML", wglGetMscRateOML == NULL);
+  glewInfoFunc("wglGetSyncValuesOML", wglGetSyncValuesOML == NULL);
+  glewInfoFunc("wglSwapBuffersMscOML", wglSwapBuffersMscOML == NULL);
+  glewInfoFunc("wglSwapLayerBuffersMscOML", wglSwapLayerBuffersMscOML == NULL);
+  glewInfoFunc("wglWaitForMscOML", wglWaitForMscOML == NULL);
+  glewInfoFunc("wglWaitForSbcOML", wglWaitForSbcOML == NULL);
+}
+
+#endif /* WGL_OML_sync_control */
+
+#else /* _UNIX */
+
+#ifdef GLX_VERSION_1_2
+
+static void _glewInfo_GLX_VERSION_1_2 (void)
+{
+  glewPrintExt("GLX_VERSION_1_2", GLXEW_VERSION_1_2, GLXEW_VERSION_1_2, GLXEW_VERSION_1_2);
+
+  glewInfoFunc("glXGetCurrentDisplay", glXGetCurrentDisplay == NULL);
+}
+
+#endif /* GLX_VERSION_1_2 */
+
+#ifdef GLX_VERSION_1_3
+
+static void _glewInfo_GLX_VERSION_1_3 (void)
+{
+  glewPrintExt("GLX_VERSION_1_3", GLXEW_VERSION_1_3, GLXEW_VERSION_1_3, GLXEW_VERSION_1_3);
+
+  glewInfoFunc("glXChooseFBConfig", glXChooseFBConfig == NULL);
+  glewInfoFunc("glXCreateNewContext", glXCreateNewContext == NULL);
+  glewInfoFunc("glXCreatePbuffer", glXCreatePbuffer == NULL);
+  glewInfoFunc("glXCreatePixmap", glXCreatePixmap == NULL);
+  glewInfoFunc("glXCreateWindow", glXCreateWindow == NULL);
+  glewInfoFunc("glXDestroyPbuffer", glXDestroyPbuffer == NULL);
+  glewInfoFunc("glXDestroyPixmap", glXDestroyPixmap == NULL);
+  glewInfoFunc("glXDestroyWindow", glXDestroyWindow == NULL);
+  glewInfoFunc("glXGetCurrentReadDrawable", glXGetCurrentReadDrawable == NULL);
+  glewInfoFunc("glXGetFBConfigAttrib", glXGetFBConfigAttrib == NULL);
+  glewInfoFunc("glXGetFBConfigs", glXGetFBConfigs == NULL);
+  glewInfoFunc("glXGetSelectedEvent", glXGetSelectedEvent == NULL);
+  glewInfoFunc("glXGetVisualFromFBConfig", glXGetVisualFromFBConfig == NULL);
+  glewInfoFunc("glXMakeContextCurrent", glXMakeContextCurrent == NULL);
+  glewInfoFunc("glXQueryContext", glXQueryContext == NULL);
+  glewInfoFunc("glXQueryDrawable", glXQueryDrawable == NULL);
+  glewInfoFunc("glXSelectEvent", glXSelectEvent == NULL);
+}
+
+#endif /* GLX_VERSION_1_3 */
+
+#ifdef GLX_VERSION_1_4
+
+static void _glewInfo_GLX_VERSION_1_4 (void)
+{
+  glewPrintExt("GLX_VERSION_1_4", GLXEW_VERSION_1_4, GLXEW_VERSION_1_4, GLXEW_VERSION_1_4);
+}
+
+#endif /* GLX_VERSION_1_4 */
+
+#ifdef GLX_3DFX_multisample
+
+static void _glewInfo_GLX_3DFX_multisample (void)
+{
+  glewPrintExt("GLX_3DFX_multisample", GLXEW_3DFX_multisample, glxewIsSupported("GLX_3DFX_multisample"), glxewGetExtension("GLX_3DFX_multisample"));
+}
+
+#endif /* GLX_3DFX_multisample */
+
+#ifdef GLX_ARB_create_context
+
+static void _glewInfo_GLX_ARB_create_context (void)
+{
+  glewPrintExt("GLX_ARB_create_context", GLXEW_ARB_create_context, glxewIsSupported("GLX_ARB_create_context"), glxewGetExtension("GLX_ARB_create_context"));
+
+  glewInfoFunc("glXCreateContextAttribsARB", glXCreateContextAttribsARB == NULL);
+}
+
+#endif /* GLX_ARB_create_context */
+
+#ifdef GLX_ARB_fbconfig_float
+
+static void _glewInfo_GLX_ARB_fbconfig_float (void)
+{
+  glewPrintExt("GLX_ARB_fbconfig_float", GLXEW_ARB_fbconfig_float, glxewIsSupported("GLX_ARB_fbconfig_float"), glxewGetExtension("GLX_ARB_fbconfig_float"));
+}
+
+#endif /* GLX_ARB_fbconfig_float */
+
+#ifdef GLX_ARB_framebuffer_sRGB
+
+static void _glewInfo_GLX_ARB_framebuffer_sRGB (void)
+{
+  glewPrintExt("GLX_ARB_framebuffer_sRGB", GLXEW_ARB_framebuffer_sRGB, glxewIsSupported("GLX_ARB_framebuffer_sRGB"), glxewGetExtension("GLX_ARB_framebuffer_sRGB"));
+}
+
+#endif /* GLX_ARB_framebuffer_sRGB */
+
+#ifdef GLX_ARB_get_proc_address
+
+static void _glewInfo_GLX_ARB_get_proc_address (void)
+{
+  glewPrintExt("GLX_ARB_get_proc_address", GLXEW_ARB_get_proc_address, glxewIsSupported("GLX_ARB_get_proc_address"), glxewGetExtension("GLX_ARB_get_proc_address"));
+}
+
+#endif /* GLX_ARB_get_proc_address */
+
+#ifdef GLX_ARB_multisample
+
+static void _glewInfo_GLX_ARB_multisample (void)
+{
+  glewPrintExt("GLX_ARB_multisample", GLXEW_ARB_multisample, glxewIsSupported("GLX_ARB_multisample"), glxewGetExtension("GLX_ARB_multisample"));
+}
+
+#endif /* GLX_ARB_multisample */
+
+#ifdef GLX_ATI_pixel_format_float
+
+static void _glewInfo_GLX_ATI_pixel_format_float (void)
+{
+  glewPrintExt("GLX_ATI_pixel_format_float", GLXEW_ATI_pixel_format_float, glxewIsSupported("GLX_ATI_pixel_format_float"), glxewGetExtension("GLX_ATI_pixel_format_float"));
+}
+
+#endif /* GLX_ATI_pixel_format_float */
+
+#ifdef GLX_ATI_render_texture
+
+static void _glewInfo_GLX_ATI_render_texture (void)
+{
+  glewPrintExt("GLX_ATI_render_texture", GLXEW_ATI_render_texture, glxewIsSupported("GLX_ATI_render_texture"), glxewGetExtension("GLX_ATI_render_texture"));
+
+  glewInfoFunc("glXBindTexImageATI", glXBindTexImageATI == NULL);
+  glewInfoFunc("glXDrawableAttribATI", glXDrawableAttribATI == NULL);
+  glewInfoFunc("glXReleaseTexImageATI", glXReleaseTexImageATI == NULL);
+}
+
+#endif /* GLX_ATI_render_texture */
+
+#ifdef GLX_EXT_fbconfig_packed_float
+
+static void _glewInfo_GLX_EXT_fbconfig_packed_float (void)
+{
+  glewPrintExt("GLX_EXT_fbconfig_packed_float", GLXEW_EXT_fbconfig_packed_float, glxewIsSupported("GLX_EXT_fbconfig_packed_float"), glxewGetExtension("GLX_EXT_fbconfig_packed_float"));
+}
+
+#endif /* GLX_EXT_fbconfig_packed_float */
+
+#ifdef GLX_EXT_framebuffer_sRGB
+
+static void _glewInfo_GLX_EXT_framebuffer_sRGB (void)
+{
+  glewPrintExt("GLX_EXT_framebuffer_sRGB", GLXEW_EXT_framebuffer_sRGB, glxewIsSupported("GLX_EXT_framebuffer_sRGB"), glxewGetExtension("GLX_EXT_framebuffer_sRGB"));
+}
+
+#endif /* GLX_EXT_framebuffer_sRGB */
+
+#ifdef GLX_EXT_import_context
+
+static void _glewInfo_GLX_EXT_import_context (void)
+{
+  glewPrintExt("GLX_EXT_import_context", GLXEW_EXT_import_context, glxewIsSupported("GLX_EXT_import_context"), glxewGetExtension("GLX_EXT_import_context"));
+
+  glewInfoFunc("glXFreeContextEXT", glXFreeContextEXT == NULL);
+  glewInfoFunc("glXGetContextIDEXT", glXGetContextIDEXT == NULL);
+  glewInfoFunc("glXImportContextEXT", glXImportContextEXT == NULL);
+  glewInfoFunc("glXQueryContextInfoEXT", glXQueryContextInfoEXT == NULL);
+}
+
+#endif /* GLX_EXT_import_context */
+
+#ifdef GLX_EXT_scene_marker
+
+static void _glewInfo_GLX_EXT_scene_marker (void)
+{
+  glewPrintExt("GLX_EXT_scene_marker", GLXEW_EXT_scene_marker, glxewIsSupported("GLX_EXT_scene_marker"), glxewGetExtension("GLX_EXT_scene_marker"));
+}
+
+#endif /* GLX_EXT_scene_marker */
+
+#ifdef GLX_EXT_texture_from_pixmap
+
+static void _glewInfo_GLX_EXT_texture_from_pixmap (void)
+{
+  glewPrintExt("GLX_EXT_texture_from_pixmap", GLXEW_EXT_texture_from_pixmap, glxewIsSupported("GLX_EXT_texture_from_pixmap"), glxewGetExtension("GLX_EXT_texture_from_pixmap"));
+
+  glewInfoFunc("glXBindTexImageEXT", glXBindTexImageEXT == NULL);
+  glewInfoFunc("glXReleaseTexImageEXT", glXReleaseTexImageEXT == NULL);
+}
+
+#endif /* GLX_EXT_texture_from_pixmap */
+
+#ifdef GLX_EXT_visual_info
+
+static void _glewInfo_GLX_EXT_visual_info (void)
+{
+  glewPrintExt("GLX_EXT_visual_info", GLXEW_EXT_visual_info, glxewIsSupported("GLX_EXT_visual_info"), glxewGetExtension("GLX_EXT_visual_info"));
+}
+
+#endif /* GLX_EXT_visual_info */
+
+#ifdef GLX_EXT_visual_rating
+
+static void _glewInfo_GLX_EXT_visual_rating (void)
+{
+  glewPrintExt("GLX_EXT_visual_rating", GLXEW_EXT_visual_rating, glxewIsSupported("GLX_EXT_visual_rating"), glxewGetExtension("GLX_EXT_visual_rating"));
+}
+
+#endif /* GLX_EXT_visual_rating */
+
+#ifdef GLX_MESA_agp_offset
+
+static void _glewInfo_GLX_MESA_agp_offset (void)
+{
+  glewPrintExt("GLX_MESA_agp_offset", GLXEW_MESA_agp_offset, glxewIsSupported("GLX_MESA_agp_offset"), glxewGetExtension("GLX_MESA_agp_offset"));
+
+  glewInfoFunc("glXGetAGPOffsetMESA", glXGetAGPOffsetMESA == NULL);
+}
+
+#endif /* GLX_MESA_agp_offset */
+
+#ifdef GLX_MESA_copy_sub_buffer
+
+static void _glewInfo_GLX_MESA_copy_sub_buffer (void)
+{
+  glewPrintExt("GLX_MESA_copy_sub_buffer", GLXEW_MESA_copy_sub_buffer, glxewIsSupported("GLX_MESA_copy_sub_buffer"), glxewGetExtension("GLX_MESA_copy_sub_buffer"));
+
+  glewInfoFunc("glXCopySubBufferMESA", glXCopySubBufferMESA == NULL);
+}
+
+#endif /* GLX_MESA_copy_sub_buffer */
+
+#ifdef GLX_MESA_pixmap_colormap
+
+static void _glewInfo_GLX_MESA_pixmap_colormap (void)
+{
+  glewPrintExt("GLX_MESA_pixmap_colormap", GLXEW_MESA_pixmap_colormap, glxewIsSupported("GLX_MESA_pixmap_colormap"), glxewGetExtension("GLX_MESA_pixmap_colormap"));
+
+  glewInfoFunc("glXCreateGLXPixmapMESA", glXCreateGLXPixmapMESA == NULL);
+}
+
+#endif /* GLX_MESA_pixmap_colormap */
+
+#ifdef GLX_MESA_release_buffers
+
+static void _glewInfo_GLX_MESA_release_buffers (void)
+{
+  glewPrintExt("GLX_MESA_release_buffers", GLXEW_MESA_release_buffers, glxewIsSupported("GLX_MESA_release_buffers"), glxewGetExtension("GLX_MESA_release_buffers"));
+
+  glewInfoFunc("glXReleaseBuffersMESA", glXReleaseBuffersMESA == NULL);
+}
+
+#endif /* GLX_MESA_release_buffers */
+
+#ifdef GLX_MESA_set_3dfx_mode
+
+static void _glewInfo_GLX_MESA_set_3dfx_mode (void)
+{
+  glewPrintExt("GLX_MESA_set_3dfx_mode", GLXEW_MESA_set_3dfx_mode, glxewIsSupported("GLX_MESA_set_3dfx_mode"), glxewGetExtension("GLX_MESA_set_3dfx_mode"));
+
+  glewInfoFunc("glXSet3DfxModeMESA", glXSet3DfxModeMESA == NULL);
+}
+
+#endif /* GLX_MESA_set_3dfx_mode */
+
+#ifdef GLX_NV_float_buffer
+
+static void _glewInfo_GLX_NV_float_buffer (void)
+{
+  glewPrintExt("GLX_NV_float_buffer", GLXEW_NV_float_buffer, glxewIsSupported("GLX_NV_float_buffer"), glxewGetExtension("GLX_NV_float_buffer"));
+}
+
+#endif /* GLX_NV_float_buffer */
+
+#ifdef GLX_NV_present_video
+
+static void _glewInfo_GLX_NV_present_video (void)
+{
+  glewPrintExt("GLX_NV_present_video", GLXEW_NV_present_video, glxewIsSupported("GLX_NV_present_video"), glxewGetExtension("GLX_NV_present_video"));
+
+  glewInfoFunc("glXBindVideoDeviceNV", glXBindVideoDeviceNV == NULL);
+  glewInfoFunc("glXEnumerateVideoDevicesNV", glXEnumerateVideoDevicesNV == NULL);
+}
+
+#endif /* GLX_NV_present_video */
+
+#ifdef GLX_NV_swap_group
+
+static void _glewInfo_GLX_NV_swap_group (void)
+{
+  glewPrintExt("GLX_NV_swap_group", GLXEW_NV_swap_group, glxewIsSupported("GLX_NV_swap_group"), glxewGetExtension("GLX_NV_swap_group"));
+
+  glewInfoFunc("glXBindSwapBarrierNV", glXBindSwapBarrierNV == NULL);
+  glewInfoFunc("glXJoinSwapGroupNV", glXJoinSwapGroupNV == NULL);
+  glewInfoFunc("glXQueryFrameCountNV", glXQueryFrameCountNV == NULL);
+  glewInfoFunc("glXQueryMaxSwapGroupsNV", glXQueryMaxSwapGroupsNV == NULL);
+  glewInfoFunc("glXQuerySwapGroupNV", glXQuerySwapGroupNV == NULL);
+  glewInfoFunc("glXResetFrameCountNV", glXResetFrameCountNV == NULL);
+}
+
+#endif /* GLX_NV_swap_group */
+
+#ifdef GLX_NV_vertex_array_range
+
+static void _glewInfo_GLX_NV_vertex_array_range (void)
+{
+  glewPrintExt("GLX_NV_vertex_array_range", GLXEW_NV_vertex_array_range, glxewIsSupported("GLX_NV_vertex_array_range"), glxewGetExtension("GLX_NV_vertex_array_range"));
+
+  glewInfoFunc("glXAllocateMemoryNV", glXAllocateMemoryNV == NULL);
+  glewInfoFunc("glXFreeMemoryNV", glXFreeMemoryNV == NULL);
+}
+
+#endif /* GLX_NV_vertex_array_range */
+
+#ifdef GLX_NV_video_output
+
+static void _glewInfo_GLX_NV_video_output (void)
+{
+  glewPrintExt("GLX_NV_video_output", GLXEW_NV_video_output, glxewIsSupported("GLX_NV_video_output"), glxewGetExtension("GLX_NV_video_output"));
+
+  glewInfoFunc("glXBindVideoImageNV", glXBindVideoImageNV == NULL);
+  glewInfoFunc("glXGetVideoDeviceNV", glXGetVideoDeviceNV == NULL);
+  glewInfoFunc("glXGetVideoInfoNV", glXGetVideoInfoNV == NULL);
+  glewInfoFunc("glXReleaseVideoDeviceNV", glXReleaseVideoDeviceNV == NULL);
+  glewInfoFunc("glXReleaseVideoImageNV", glXReleaseVideoImageNV == NULL);
+  glewInfoFunc("glXSendPbufferToVideoNV", glXSendPbufferToVideoNV == NULL);
+}
+
+#endif /* GLX_NV_video_output */
+
+#ifdef GLX_OML_swap_method
+
+static void _glewInfo_GLX_OML_swap_method (void)
+{
+  glewPrintExt("GLX_OML_swap_method", GLXEW_OML_swap_method, glxewIsSupported("GLX_OML_swap_method"), glxewGetExtension("GLX_OML_swap_method"));
+}
+
+#endif /* GLX_OML_swap_method */
+
+#if defined(GLX_OML_sync_control) && defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)
+#include <inttypes.h>
+
+static void _glewInfo_GLX_OML_sync_control (void)
+{
+  glewPrintExt("GLX_OML_sync_control", GLXEW_OML_sync_control, glxewIsSupported("GLX_OML_sync_control"), glxewGetExtension("GLX_OML_sync_control"));
+
+  glewInfoFunc("glXGetMscRateOML", glXGetMscRateOML == NULL);
+  glewInfoFunc("glXGetSyncValuesOML", glXGetSyncValuesOML == NULL);
+  glewInfoFunc("glXSwapBuffersMscOML", glXSwapBuffersMscOML == NULL);
+  glewInfoFunc("glXWaitForMscOML", glXWaitForMscOML == NULL);
+  glewInfoFunc("glXWaitForSbcOML", glXWaitForSbcOML == NULL);
+}
+
+#endif /* GLX_OML_sync_control */
+
+#ifdef GLX_SGIS_blended_overlay
+
+static void _glewInfo_GLX_SGIS_blended_overlay (void)
+{
+  glewPrintExt("GLX_SGIS_blended_overlay", GLXEW_SGIS_blended_overlay, glxewIsSupported("GLX_SGIS_blended_overlay"), glxewGetExtension("GLX_SGIS_blended_overlay"));
+}
+
+#endif /* GLX_SGIS_blended_overlay */
+
+#ifdef GLX_SGIS_color_range
+
+static void _glewInfo_GLX_SGIS_color_range (void)
+{
+  glewPrintExt("GLX_SGIS_color_range", GLXEW_SGIS_color_range, glxewIsSupported("GLX_SGIS_color_range"), glxewGetExtension("GLX_SGIS_color_range"));
+}
+
+#endif /* GLX_SGIS_color_range */
+
+#ifdef GLX_SGIS_multisample
+
+static void _glewInfo_GLX_SGIS_multisample (void)
+{
+  glewPrintExt("GLX_SGIS_multisample", GLXEW_SGIS_multisample, glxewIsSupported("GLX_SGIS_multisample"), glxewGetExtension("GLX_SGIS_multisample"));
+}
+
+#endif /* GLX_SGIS_multisample */
+
+#ifdef GLX_SGIS_shared_multisample
+
+static void _glewInfo_GLX_SGIS_shared_multisample (void)
+{
+  glewPrintExt("GLX_SGIS_shared_multisample", GLXEW_SGIS_shared_multisample, glxewIsSupported("GLX_SGIS_shared_multisample"), glxewGetExtension("GLX_SGIS_shared_multisample"));
+}
+
+#endif /* GLX_SGIS_shared_multisample */
+
+#ifdef GLX_SGIX_fbconfig
+
+static void _glewInfo_GLX_SGIX_fbconfig (void)
+{
+  glewPrintExt("GLX_SGIX_fbconfig", GLXEW_SGIX_fbconfig, glxewIsSupported("GLX_SGIX_fbconfig"), glxewGetExtension("GLX_SGIX_fbconfig"));
+
+  glewInfoFunc("glXChooseFBConfigSGIX", glXChooseFBConfigSGIX == NULL);
+  glewInfoFunc("glXCreateContextWithConfigSGIX", glXCreateContextWithConfigSGIX == NULL);
+  glewInfoFunc("glXCreateGLXPixmapWithConfigSGIX", glXCreateGLXPixmapWithConfigSGIX == NULL);
+  glewInfoFunc("glXGetFBConfigAttribSGIX", glXGetFBConfigAttribSGIX == NULL);
+  glewInfoFunc("glXGetFBConfigFromVisualSGIX", glXGetFBConfigFromVisualSGIX == NULL);
+  glewInfoFunc("glXGetVisualFromFBConfigSGIX", glXGetVisualFromFBConfigSGIX == NULL);
+}
+
+#endif /* GLX_SGIX_fbconfig */
+
+#ifdef GLX_SGIX_hyperpipe
+
+static void _glewInfo_GLX_SGIX_hyperpipe (void)
+{
+  glewPrintExt("GLX_SGIX_hyperpipe", GLXEW_SGIX_hyperpipe, glxewIsSupported("GLX_SGIX_hyperpipe"), glxewGetExtension("GLX_SGIX_hyperpipe"));
+
+  glewInfoFunc("glXBindHyperpipeSGIX", glXBindHyperpipeSGIX == NULL);
+  glewInfoFunc("glXDestroyHyperpipeConfigSGIX", glXDestroyHyperpipeConfigSGIX == NULL);
+  glewInfoFunc("glXHyperpipeAttribSGIX", glXHyperpipeAttribSGIX == NULL);
+  glewInfoFunc("glXHyperpipeConfigSGIX", glXHyperpipeConfigSGIX == NULL);
+  glewInfoFunc("glXQueryHyperpipeAttribSGIX", glXQueryHyperpipeAttribSGIX == NULL);
+  glewInfoFunc("glXQueryHyperpipeBestAttribSGIX", glXQueryHyperpipeBestAttribSGIX == NULL);
+  glewInfoFunc("glXQueryHyperpipeConfigSGIX", glXQueryHyperpipeConfigSGIX == NULL);
+  glewInfoFunc("glXQueryHyperpipeNetworkSGIX", glXQueryHyperpipeNetworkSGIX == NULL);
+}
+
+#endif /* GLX_SGIX_hyperpipe */
+
+#ifdef GLX_SGIX_pbuffer
+
+static void _glewInfo_GLX_SGIX_pbuffer (void)
+{
+  glewPrintExt("GLX_SGIX_pbuffer", GLXEW_SGIX_pbuffer, glxewIsSupported("GLX_SGIX_pbuffer"), glxewGetExtension("GLX_SGIX_pbuffer"));
+
+  glewInfoFunc("glXCreateGLXPbufferSGIX", glXCreateGLXPbufferSGIX == NULL);
+  glewInfoFunc("glXDestroyGLXPbufferSGIX", glXDestroyGLXPbufferSGIX == NULL);
+  glewInfoFunc("glXGetSelectedEventSGIX", glXGetSelectedEventSGIX == NULL);
+  glewInfoFunc("glXQueryGLXPbufferSGIX", glXQueryGLXPbufferSGIX == NULL);
+  glewInfoFunc("glXSelectEventSGIX", glXSelectEventSGIX == NULL);
+}
+
+#endif /* GLX_SGIX_pbuffer */
+
+#ifdef GLX_SGIX_swap_barrier
+
+static void _glewInfo_GLX_SGIX_swap_barrier (void)
+{
+  glewPrintExt("GLX_SGIX_swap_barrier", GLXEW_SGIX_swap_barrier, glxewIsSupported("GLX_SGIX_swap_barrier"), glxewGetExtension("GLX_SGIX_swap_barrier"));
+
+  glewInfoFunc("glXBindSwapBarrierSGIX", glXBindSwapBarrierSGIX == NULL);
+  glewInfoFunc("glXQueryMaxSwapBarriersSGIX", glXQueryMaxSwapBarriersSGIX == NULL);
+}
+
+#endif /* GLX_SGIX_swap_barrier */
+
+#ifdef GLX_SGIX_swap_group
+
+static void _glewInfo_GLX_SGIX_swap_group (void)
+{
+  glewPrintExt("GLX_SGIX_swap_group", GLXEW_SGIX_swap_group, glxewIsSupported("GLX_SGIX_swap_group"), glxewGetExtension("GLX_SGIX_swap_group"));
+
+  glewInfoFunc("glXJoinSwapGroupSGIX", glXJoinSwapGroupSGIX == NULL);
+}
+
+#endif /* GLX_SGIX_swap_group */
+
+#ifdef GLX_SGIX_video_resize
+
+static void _glewInfo_GLX_SGIX_video_resize (void)
+{
+  glewPrintExt("GLX_SGIX_video_resize", GLXEW_SGIX_video_resize, glxewIsSupported("GLX_SGIX_video_resize"), glxewGetExtension("GLX_SGIX_video_resize"));
+
+  glewInfoFunc("glXBindChannelToWindowSGIX", glXBindChannelToWindowSGIX == NULL);
+  glewInfoFunc("glXChannelRectSGIX", glXChannelRectSGIX == NULL);
+  glewInfoFunc("glXChannelRectSyncSGIX", glXChannelRectSyncSGIX == NULL);
+  glewInfoFunc("glXQueryChannelDeltasSGIX", glXQueryChannelDeltasSGIX == NULL);
+  glewInfoFunc("glXQueryChannelRectSGIX", glXQueryChannelRectSGIX == NULL);
+}
+
+#endif /* GLX_SGIX_video_resize */
+
+#ifdef GLX_SGIX_visual_select_group
+
+static void _glewInfo_GLX_SGIX_visual_select_group (void)
+{
+  glewPrintExt("GLX_SGIX_visual_select_group", GLXEW_SGIX_visual_select_group, glxewIsSupported("GLX_SGIX_visual_select_group"), glxewGetExtension("GLX_SGIX_visual_select_group"));
+}
+
+#endif /* GLX_SGIX_visual_select_group */
+
+#ifdef GLX_SGI_cushion
+
+static void _glewInfo_GLX_SGI_cushion (void)
+{
+  glewPrintExt("GLX_SGI_cushion", GLXEW_SGI_cushion, glxewIsSupported("GLX_SGI_cushion"), glxewGetExtension("GLX_SGI_cushion"));
+
+  glewInfoFunc("glXCushionSGI", glXCushionSGI == NULL);
+}
+
+#endif /* GLX_SGI_cushion */
+
+#ifdef GLX_SGI_make_current_read
+
+static void _glewInfo_GLX_SGI_make_current_read (void)
+{
+  glewPrintExt("GLX_SGI_make_current_read", GLXEW_SGI_make_current_read, glxewIsSupported("GLX_SGI_make_current_read"), glxewGetExtension("GLX_SGI_make_current_read"));
+
+  glewInfoFunc("glXGetCurrentReadDrawableSGI", glXGetCurrentReadDrawableSGI == NULL);
+  glewInfoFunc("glXMakeCurrentReadSGI", glXMakeCurrentReadSGI == NULL);
+}
+
+#endif /* GLX_SGI_make_current_read */
+
+#ifdef GLX_SGI_swap_control
+
+static void _glewInfo_GLX_SGI_swap_control (void)
+{
+  glewPrintExt("GLX_SGI_swap_control", GLXEW_SGI_swap_control, glxewIsSupported("GLX_SGI_swap_control"), glxewGetExtension("GLX_SGI_swap_control"));
+
+  glewInfoFunc("glXSwapIntervalSGI", glXSwapIntervalSGI == NULL);
+}
+
+#endif /* GLX_SGI_swap_control */
+
+#ifdef GLX_SGI_video_sync
+
+static void _glewInfo_GLX_SGI_video_sync (void)
+{
+  glewPrintExt("GLX_SGI_video_sync", GLXEW_SGI_video_sync, glxewIsSupported("GLX_SGI_video_sync"), glxewGetExtension("GLX_SGI_video_sync"));
+
+  glewInfoFunc("glXGetVideoSyncSGI", glXGetVideoSyncSGI == NULL);
+  glewInfoFunc("glXWaitVideoSyncSGI", glXWaitVideoSyncSGI == NULL);
+}
+
+#endif /* GLX_SGI_video_sync */
+
+#ifdef GLX_SUN_get_transparent_index
+
+static void _glewInfo_GLX_SUN_get_transparent_index (void)
+{
+  glewPrintExt("GLX_SUN_get_transparent_index", GLXEW_SUN_get_transparent_index, glxewIsSupported("GLX_SUN_get_transparent_index"), glxewGetExtension("GLX_SUN_get_transparent_index"));
+
+  glewInfoFunc("glXGetTransparentIndexSUN", glXGetTransparentIndexSUN == NULL);
+}
+
+#endif /* GLX_SUN_get_transparent_index */
+
+#ifdef GLX_SUN_video_resize
+
+static void _glewInfo_GLX_SUN_video_resize (void)
+{
+  glewPrintExt("GLX_SUN_video_resize", GLXEW_SUN_video_resize, glxewIsSupported("GLX_SUN_video_resize"), glxewGetExtension("GLX_SUN_video_resize"));
+
+  glewInfoFunc("glXGetVideoResizeSUN", glXGetVideoResizeSUN == NULL);
+  glewInfoFunc("glXVideoResizeSUN", glXVideoResizeSUN == NULL);
+}
+
+#endif /* GLX_SUN_video_resize */
+
+#endif /* _WIN32 */
+
+/* ------------------------------------------------------------------------ */
+
+static void glewInfo (void)
+{
+#ifdef GL_VERSION_1_1
+  _glewInfo_GL_VERSION_1_1();
+#endif /* GL_VERSION_1_1 */
+#ifdef GL_VERSION_1_2
+  _glewInfo_GL_VERSION_1_2();
+#endif /* GL_VERSION_1_2 */
+#ifdef GL_VERSION_1_3
+  _glewInfo_GL_VERSION_1_3();
+#endif /* GL_VERSION_1_3 */
+#ifdef GL_VERSION_1_4
+  _glewInfo_GL_VERSION_1_4();
+#endif /* GL_VERSION_1_4 */
+#ifdef GL_VERSION_1_5
+  _glewInfo_GL_VERSION_1_5();
+#endif /* GL_VERSION_1_5 */
+#ifdef GL_VERSION_2_0
+  _glewInfo_GL_VERSION_2_0();
+#endif /* GL_VERSION_2_0 */
+#ifdef GL_VERSION_2_1
+  _glewInfo_GL_VERSION_2_1();
+#endif /* GL_VERSION_2_1 */
+#ifdef GL_VERSION_3_0
+  _glewInfo_GL_VERSION_3_0();
+#endif /* GL_VERSION_3_0 */
+#ifdef GL_3DFX_multisample
+  _glewInfo_GL_3DFX_multisample();
+#endif /* GL_3DFX_multisample */
+#ifdef GL_3DFX_tbuffer
+  _glewInfo_GL_3DFX_tbuffer();
+#endif /* GL_3DFX_tbuffer */
+#ifdef GL_3DFX_texture_compression_FXT1
+  _glewInfo_GL_3DFX_texture_compression_FXT1();
+#endif /* GL_3DFX_texture_compression_FXT1 */
+#ifdef GL_APPLE_client_storage
+  _glewInfo_GL_APPLE_client_storage();
+#endif /* GL_APPLE_client_storage */
+#ifdef GL_APPLE_element_array
+  _glewInfo_GL_APPLE_element_array();
+#endif /* GL_APPLE_element_array */
+#ifdef GL_APPLE_fence
+  _glewInfo_GL_APPLE_fence();
+#endif /* GL_APPLE_fence */
+#ifdef GL_APPLE_float_pixels
+  _glewInfo_GL_APPLE_float_pixels();
+#endif /* GL_APPLE_float_pixels */
+#ifdef GL_APPLE_flush_buffer_range
+  _glewInfo_GL_APPLE_flush_buffer_range();
+#endif /* GL_APPLE_flush_buffer_range */
+#ifdef GL_APPLE_pixel_buffer
+  _glewInfo_GL_APPLE_pixel_buffer();
+#endif /* GL_APPLE_pixel_buffer */
+#ifdef GL_APPLE_specular_vector
+  _glewInfo_GL_APPLE_specular_vector();
+#endif /* GL_APPLE_specular_vector */
+#ifdef GL_APPLE_texture_range
+  _glewInfo_GL_APPLE_texture_range();
+#endif /* GL_APPLE_texture_range */
+#ifdef GL_APPLE_transform_hint
+  _glewInfo_GL_APPLE_transform_hint();
+#endif /* GL_APPLE_transform_hint */
+#ifdef GL_APPLE_vertex_array_object
+  _glewInfo_GL_APPLE_vertex_array_object();
+#endif /* GL_APPLE_vertex_array_object */
+#ifdef GL_APPLE_vertex_array_range
+  _glewInfo_GL_APPLE_vertex_array_range();
+#endif /* GL_APPLE_vertex_array_range */
+#ifdef GL_APPLE_ycbcr_422
+  _glewInfo_GL_APPLE_ycbcr_422();
+#endif /* GL_APPLE_ycbcr_422 */
+#ifdef GL_ARB_color_buffer_float
+  _glewInfo_GL_ARB_color_buffer_float();
+#endif /* GL_ARB_color_buffer_float */
+#ifdef GL_ARB_depth_buffer_float
+  _glewInfo_GL_ARB_depth_buffer_float();
+#endif /* GL_ARB_depth_buffer_float */
+#ifdef GL_ARB_depth_texture
+  _glewInfo_GL_ARB_depth_texture();
+#endif /* GL_ARB_depth_texture */
+#ifdef GL_ARB_draw_buffers
+  _glewInfo_GL_ARB_draw_buffers();
+#endif /* GL_ARB_draw_buffers */
+#ifdef GL_ARB_draw_instanced
+  _glewInfo_GL_ARB_draw_instanced();
+#endif /* GL_ARB_draw_instanced */
+#ifdef GL_ARB_fragment_program
+  _glewInfo_GL_ARB_fragment_program();
+#endif /* GL_ARB_fragment_program */
+#ifdef GL_ARB_fragment_program_shadow
+  _glewInfo_GL_ARB_fragment_program_shadow();
+#endif /* GL_ARB_fragment_program_shadow */
+#ifdef GL_ARB_fragment_shader
+  _glewInfo_GL_ARB_fragment_shader();
+#endif /* GL_ARB_fragment_shader */
+#ifdef GL_ARB_framebuffer_object
+  _glewInfo_GL_ARB_framebuffer_object();
+#endif /* GL_ARB_framebuffer_object */
+#ifdef GL_ARB_framebuffer_sRGB
+  _glewInfo_GL_ARB_framebuffer_sRGB();
+#endif /* GL_ARB_framebuffer_sRGB */
+#ifdef GL_ARB_geometry_shader4
+  _glewInfo_GL_ARB_geometry_shader4();
+#endif /* GL_ARB_geometry_shader4 */
+#ifdef GL_ARB_half_float_pixel
+  _glewInfo_GL_ARB_half_float_pixel();
+#endif /* GL_ARB_half_float_pixel */
+#ifdef GL_ARB_half_float_vertex
+  _glewInfo_GL_ARB_half_float_vertex();
+#endif /* GL_ARB_half_float_vertex */
+#ifdef GL_ARB_imaging
+  _glewInfo_GL_ARB_imaging();
+#endif /* GL_ARB_imaging */
+#ifdef GL_ARB_instanced_arrays
+  _glewInfo_GL_ARB_instanced_arrays();
+#endif /* GL_ARB_instanced_arrays */
+#ifdef GL_ARB_map_buffer_range
+  _glewInfo_GL_ARB_map_buffer_range();
+#endif /* GL_ARB_map_buffer_range */
+#ifdef GL_ARB_matrix_palette
+  _glewInfo_GL_ARB_matrix_palette();
+#endif /* GL_ARB_matrix_palette */
+#ifdef GL_ARB_multisample
+  _glewInfo_GL_ARB_multisample();
+#endif /* GL_ARB_multisample */
+#ifdef GL_ARB_multitexture
+  _glewInfo_GL_ARB_multitexture();
+#endif /* GL_ARB_multitexture */
+#ifdef GL_ARB_occlusion_query
+  _glewInfo_GL_ARB_occlusion_query();
+#endif /* GL_ARB_occlusion_query */
+#ifdef GL_ARB_pixel_buffer_object
+  _glewInfo_GL_ARB_pixel_buffer_object();
+#endif /* GL_ARB_pixel_buffer_object */
+#ifdef GL_ARB_point_parameters
+  _glewInfo_GL_ARB_point_parameters();
+#endif /* GL_ARB_point_parameters */
+#ifdef GL_ARB_point_sprite
+  _glewInfo_GL_ARB_point_sprite();
+#endif /* GL_ARB_point_sprite */
+#ifdef GL_ARB_shader_objects
+  _glewInfo_GL_ARB_shader_objects();
+#endif /* GL_ARB_shader_objects */
+#ifdef GL_ARB_shading_language_100
+  _glewInfo_GL_ARB_shading_language_100();
+#endif /* GL_ARB_shading_language_100 */
+#ifdef GL_ARB_shadow
+  _glewInfo_GL_ARB_shadow();
+#endif /* GL_ARB_shadow */
+#ifdef GL_ARB_shadow_ambient
+  _glewInfo_GL_ARB_shadow_ambient();
+#endif /* GL_ARB_shadow_ambient */
+#ifdef GL_ARB_texture_border_clamp
+  _glewInfo_GL_ARB_texture_border_clamp();
+#endif /* GL_ARB_texture_border_clamp */
+#ifdef GL_ARB_texture_buffer_object
+  _glewInfo_GL_ARB_texture_buffer_object();
+#endif /* GL_ARB_texture_buffer_object */
+#ifdef GL_ARB_texture_compression
+  _glewInfo_GL_ARB_texture_compression();
+#endif /* GL_ARB_texture_compression */
+#ifdef GL_ARB_texture_compression_rgtc
+  _glewInfo_GL_ARB_texture_compression_rgtc();
+#endif /* GL_ARB_texture_compression_rgtc */
+#ifdef GL_ARB_texture_cube_map
+  _glewInfo_GL_ARB_texture_cube_map();
+#endif /* GL_ARB_texture_cube_map */
+#ifdef GL_ARB_texture_env_add
+  _glewInfo_GL_ARB_texture_env_add();
+#endif /* GL_ARB_texture_env_add */
+#ifdef GL_ARB_texture_env_combine
+  _glewInfo_GL_ARB_texture_env_combine();
+#endif /* GL_ARB_texture_env_combine */
+#ifdef GL_ARB_texture_env_crossbar
+  _glewInfo_GL_ARB_texture_env_crossbar();
+#endif /* GL_ARB_texture_env_crossbar */
+#ifdef GL_ARB_texture_env_dot3
+  _glewInfo_GL_ARB_texture_env_dot3();
+#endif /* GL_ARB_texture_env_dot3 */
+#ifdef GL_ARB_texture_float
+  _glewInfo_GL_ARB_texture_float();
+#endif /* GL_ARB_texture_float */
+#ifdef GL_ARB_texture_mirrored_repeat
+  _glewInfo_GL_ARB_texture_mirrored_repeat();
+#endif /* GL_ARB_texture_mirrored_repeat */
+#ifdef GL_ARB_texture_non_power_of_two
+  _glewInfo_GL_ARB_texture_non_power_of_two();
+#endif /* GL_ARB_texture_non_power_of_two */
+#ifdef GL_ARB_texture_rectangle
+  _glewInfo_GL_ARB_texture_rectangle();
+#endif /* GL_ARB_texture_rectangle */
+#ifdef GL_ARB_texture_rg
+  _glewInfo_GL_ARB_texture_rg();
+#endif /* GL_ARB_texture_rg */
+#ifdef GL_ARB_transpose_matrix
+  _glewInfo_GL_ARB_transpose_matrix();
+#endif /* GL_ARB_transpose_matrix */
+#ifdef GL_ARB_vertex_array_object
+  _glewInfo_GL_ARB_vertex_array_object();
+#endif /* GL_ARB_vertex_array_object */
+#ifdef GL_ARB_vertex_blend
+  _glewInfo_GL_ARB_vertex_blend();
+#endif /* GL_ARB_vertex_blend */
+#ifdef GL_ARB_vertex_buffer_object
+  _glewInfo_GL_ARB_vertex_buffer_object();
+#endif /* GL_ARB_vertex_buffer_object */
+#ifdef GL_ARB_vertex_program
+  _glewInfo_GL_ARB_vertex_program();
+#endif /* GL_ARB_vertex_program */
+#ifdef GL_ARB_vertex_shader
+  _glewInfo_GL_ARB_vertex_shader();
+#endif /* GL_ARB_vertex_shader */
+#ifdef GL_ARB_window_pos
+  _glewInfo_GL_ARB_window_pos();
+#endif /* GL_ARB_window_pos */
+#ifdef GL_ATIX_point_sprites
+  _glewInfo_GL_ATIX_point_sprites();
+#endif /* GL_ATIX_point_sprites */
+#ifdef GL_ATIX_texture_env_combine3
+  _glewInfo_GL_ATIX_texture_env_combine3();
+#endif /* GL_ATIX_texture_env_combine3 */
+#ifdef GL_ATIX_texture_env_route
+  _glewInfo_GL_ATIX_texture_env_route();
+#endif /* GL_ATIX_texture_env_route */
+#ifdef GL_ATIX_vertex_shader_output_point_size
+  _glewInfo_GL_ATIX_vertex_shader_output_point_size();
+#endif /* GL_ATIX_vertex_shader_output_point_size */
+#ifdef GL_ATI_draw_buffers
+  _glewInfo_GL_ATI_draw_buffers();
+#endif /* GL_ATI_draw_buffers */
+#ifdef GL_ATI_element_array
+  _glewInfo_GL_ATI_element_array();
+#endif /* GL_ATI_element_array */
+#ifdef GL_ATI_envmap_bumpmap
+  _glewInfo_GL_ATI_envmap_bumpmap();
+#endif /* GL_ATI_envmap_bumpmap */
+#ifdef GL_ATI_fragment_shader
+  _glewInfo_GL_ATI_fragment_shader();
+#endif /* GL_ATI_fragment_shader */
+#ifdef GL_ATI_map_object_buffer
+  _glewInfo_GL_ATI_map_object_buffer();
+#endif /* GL_ATI_map_object_buffer */
+#ifdef GL_ATI_pn_triangles
+  _glewInfo_GL_ATI_pn_triangles();
+#endif /* GL_ATI_pn_triangles */
+#ifdef GL_ATI_separate_stencil
+  _glewInfo_GL_ATI_separate_stencil();
+#endif /* GL_ATI_separate_stencil */
+#ifdef GL_ATI_shader_texture_lod
+  _glewInfo_GL_ATI_shader_texture_lod();
+#endif /* GL_ATI_shader_texture_lod */
+#ifdef GL_ATI_text_fragment_shader
+  _glewInfo_GL_ATI_text_fragment_shader();
+#endif /* GL_ATI_text_fragment_shader */
+#ifdef GL_ATI_texture_compression_3dc
+  _glewInfo_GL_ATI_texture_compression_3dc();
+#endif /* GL_ATI_texture_compression_3dc */
+#ifdef GL_ATI_texture_env_combine3
+  _glewInfo_GL_ATI_texture_env_combine3();
+#endif /* GL_ATI_texture_env_combine3 */
+#ifdef GL_ATI_texture_float
+  _glewInfo_GL_ATI_texture_float();
+#endif /* GL_ATI_texture_float */
+#ifdef GL_ATI_texture_mirror_once
+  _glewInfo_GL_ATI_texture_mirror_once();
+#endif /* GL_ATI_texture_mirror_once */
+#ifdef GL_ATI_vertex_array_object
+  _glewInfo_GL_ATI_vertex_array_object();
+#endif /* GL_ATI_vertex_array_object */
+#ifdef GL_ATI_vertex_attrib_array_object
+  _glewInfo_GL_ATI_vertex_attrib_array_object();
+#endif /* GL_ATI_vertex_attrib_array_object */
+#ifdef GL_ATI_vertex_streams
+  _glewInfo_GL_ATI_vertex_streams();
+#endif /* GL_ATI_vertex_streams */
+#ifdef GL_EXT_422_pixels
+  _glewInfo_GL_EXT_422_pixels();
+#endif /* GL_EXT_422_pixels */
+#ifdef GL_EXT_Cg_shader
+  _glewInfo_GL_EXT_Cg_shader();
+#endif /* GL_EXT_Cg_shader */
+#ifdef GL_EXT_abgr
+  _glewInfo_GL_EXT_abgr();
+#endif /* GL_EXT_abgr */
+#ifdef GL_EXT_bgra
+  _glewInfo_GL_EXT_bgra();
+#endif /* GL_EXT_bgra */
+#ifdef GL_EXT_bindable_uniform
+  _glewInfo_GL_EXT_bindable_uniform();
+#endif /* GL_EXT_bindable_uniform */
+#ifdef GL_EXT_blend_color
+  _glewInfo_GL_EXT_blend_color();
+#endif /* GL_EXT_blend_color */
+#ifdef GL_EXT_blend_equation_separate
+  _glewInfo_GL_EXT_blend_equation_separate();
+#endif /* GL_EXT_blend_equation_separate */
+#ifdef GL_EXT_blend_func_separate
+  _glewInfo_GL_EXT_blend_func_separate();
+#endif /* GL_EXT_blend_func_separate */
+#ifdef GL_EXT_blend_logic_op
+  _glewInfo_GL_EXT_blend_logic_op();
+#endif /* GL_EXT_blend_logic_op */
+#ifdef GL_EXT_blend_minmax
+  _glewInfo_GL_EXT_blend_minmax();
+#endif /* GL_EXT_blend_minmax */
+#ifdef GL_EXT_blend_subtract
+  _glewInfo_GL_EXT_blend_subtract();
+#endif /* GL_EXT_blend_subtract */
+#ifdef GL_EXT_clip_volume_hint
+  _glewInfo_GL_EXT_clip_volume_hint();
+#endif /* GL_EXT_clip_volume_hint */
+#ifdef GL_EXT_cmyka
+  _glewInfo_GL_EXT_cmyka();
+#endif /* GL_EXT_cmyka */
+#ifdef GL_EXT_color_subtable
+  _glewInfo_GL_EXT_color_subtable();
+#endif /* GL_EXT_color_subtable */
+#ifdef GL_EXT_compiled_vertex_array
+  _glewInfo_GL_EXT_compiled_vertex_array();
+#endif /* GL_EXT_compiled_vertex_array */
+#ifdef GL_EXT_convolution
+  _glewInfo_GL_EXT_convolution();
+#endif /* GL_EXT_convolution */
+#ifdef GL_EXT_coordinate_frame
+  _glewInfo_GL_EXT_coordinate_frame();
+#endif /* GL_EXT_coordinate_frame */
+#ifdef GL_EXT_copy_texture
+  _glewInfo_GL_EXT_copy_texture();
+#endif /* GL_EXT_copy_texture */
+#ifdef GL_EXT_cull_vertex
+  _glewInfo_GL_EXT_cull_vertex();
+#endif /* GL_EXT_cull_vertex */
+#ifdef GL_EXT_depth_bounds_test
+  _glewInfo_GL_EXT_depth_bounds_test();
+#endif /* GL_EXT_depth_bounds_test */
+#ifdef GL_EXT_direct_state_access
+  _glewInfo_GL_EXT_direct_state_access();
+#endif /* GL_EXT_direct_state_access */
+#ifdef GL_EXT_draw_buffers2
+  _glewInfo_GL_EXT_draw_buffers2();
+#endif /* GL_EXT_draw_buffers2 */
+#ifdef GL_EXT_draw_instanced
+  _glewInfo_GL_EXT_draw_instanced();
+#endif /* GL_EXT_draw_instanced */
+#ifdef GL_EXT_draw_range_elements
+  _glewInfo_GL_EXT_draw_range_elements();
+#endif /* GL_EXT_draw_range_elements */
+#ifdef GL_EXT_fog_coord
+  _glewInfo_GL_EXT_fog_coord();
+#endif /* GL_EXT_fog_coord */
+#ifdef GL_EXT_fragment_lighting
+  _glewInfo_GL_EXT_fragment_lighting();
+#endif /* GL_EXT_fragment_lighting */
+#ifdef GL_EXT_framebuffer_blit
+  _glewInfo_GL_EXT_framebuffer_blit();
+#endif /* GL_EXT_framebuffer_blit */
+#ifdef GL_EXT_framebuffer_multisample
+  _glewInfo_GL_EXT_framebuffer_multisample();
+#endif /* GL_EXT_framebuffer_multisample */
+#ifdef GL_EXT_framebuffer_object
+  _glewInfo_GL_EXT_framebuffer_object();
+#endif /* GL_EXT_framebuffer_object */
+#ifdef GL_EXT_framebuffer_sRGB
+  _glewInfo_GL_EXT_framebuffer_sRGB();
+#endif /* GL_EXT_framebuffer_sRGB */
+#ifdef GL_EXT_geometry_shader4
+  _glewInfo_GL_EXT_geometry_shader4();
+#endif /* GL_EXT_geometry_shader4 */
+#ifdef GL_EXT_gpu_program_parameters
+  _glewInfo_GL_EXT_gpu_program_parameters();
+#endif /* GL_EXT_gpu_program_parameters */
+#ifdef GL_EXT_gpu_shader4
+  _glewInfo_GL_EXT_gpu_shader4();
+#endif /* GL_EXT_gpu_shader4 */
+#ifdef GL_EXT_histogram
+  _glewInfo_GL_EXT_histogram();
+#endif /* GL_EXT_histogram */
+#ifdef GL_EXT_index_array_formats
+  _glewInfo_GL_EXT_index_array_formats();
+#endif /* GL_EXT_index_array_formats */
+#ifdef GL_EXT_index_func
+  _glewInfo_GL_EXT_index_func();
+#endif /* GL_EXT_index_func */
+#ifdef GL_EXT_index_material
+  _glewInfo_GL_EXT_index_material();
+#endif /* GL_EXT_index_material */
+#ifdef GL_EXT_index_texture
+  _glewInfo_GL_EXT_index_texture();
+#endif /* GL_EXT_index_texture */
+#ifdef GL_EXT_light_texture
+  _glewInfo_GL_EXT_light_texture();
+#endif /* GL_EXT_light_texture */
+#ifdef GL_EXT_misc_attribute
+  _glewInfo_GL_EXT_misc_attribute();
+#endif /* GL_EXT_misc_attribute */
+#ifdef GL_EXT_multi_draw_arrays
+  _glewInfo_GL_EXT_multi_draw_arrays();
+#endif /* GL_EXT_multi_draw_arrays */
+#ifdef GL_EXT_multisample
+  _glewInfo_GL_EXT_multisample();
+#endif /* GL_EXT_multisample */
+#ifdef GL_EXT_packed_depth_stencil
+  _glewInfo_GL_EXT_packed_depth_stencil();
+#endif /* GL_EXT_packed_depth_stencil */
+#ifdef GL_EXT_packed_float
+  _glewInfo_GL_EXT_packed_float();
+#endif /* GL_EXT_packed_float */
+#ifdef GL_EXT_packed_pixels
+  _glewInfo_GL_EXT_packed_pixels();
+#endif /* GL_EXT_packed_pixels */
+#ifdef GL_EXT_paletted_texture
+  _glewInfo_GL_EXT_paletted_texture();
+#endif /* GL_EXT_paletted_texture */
+#ifdef GL_EXT_pixel_buffer_object
+  _glewInfo_GL_EXT_pixel_buffer_object();
+#endif /* GL_EXT_pixel_buffer_object */
+#ifdef GL_EXT_pixel_transform
+  _glewInfo_GL_EXT_pixel_transform();
+#endif /* GL_EXT_pixel_transform */
+#ifdef GL_EXT_pixel_transform_color_table
+  _glewInfo_GL_EXT_pixel_transform_color_table();
+#endif /* GL_EXT_pixel_transform_color_table */
+#ifdef GL_EXT_point_parameters
+  _glewInfo_GL_EXT_point_parameters();
+#endif /* GL_EXT_point_parameters */
+#ifdef GL_EXT_polygon_offset
+  _glewInfo_GL_EXT_polygon_offset();
+#endif /* GL_EXT_polygon_offset */
+#ifdef GL_EXT_rescale_normal
+  _glewInfo_GL_EXT_rescale_normal();
+#endif /* GL_EXT_rescale_normal */
+#ifdef GL_EXT_scene_marker
+  _glewInfo_GL_EXT_scene_marker();
+#endif /* GL_EXT_scene_marker */
+#ifdef GL_EXT_secondary_color
+  _glewInfo_GL_EXT_secondary_color();
+#endif /* GL_EXT_secondary_color */
+#ifdef GL_EXT_separate_specular_color
+  _glewInfo_GL_EXT_separate_specular_color();
+#endif /* GL_EXT_separate_specular_color */
+#ifdef GL_EXT_shadow_funcs
+  _glewInfo_GL_EXT_shadow_funcs();
+#endif /* GL_EXT_shadow_funcs */
+#ifdef GL_EXT_shared_texture_palette
+  _glewInfo_GL_EXT_shared_texture_palette();
+#endif /* GL_EXT_shared_texture_palette */
+#ifdef GL_EXT_stencil_clear_tag
+  _glewInfo_GL_EXT_stencil_clear_tag();
+#endif /* GL_EXT_stencil_clear_tag */
+#ifdef GL_EXT_stencil_two_side
+  _glewInfo_GL_EXT_stencil_two_side();
+#endif /* GL_EXT_stencil_two_side */
+#ifdef GL_EXT_stencil_wrap
+  _glewInfo_GL_EXT_stencil_wrap();
+#endif /* GL_EXT_stencil_wrap */
+#ifdef GL_EXT_subtexture
+  _glewInfo_GL_EXT_subtexture();
+#endif /* GL_EXT_subtexture */
+#ifdef GL_EXT_texture
+  _glewInfo_GL_EXT_texture();
+#endif /* GL_EXT_texture */
+#ifdef GL_EXT_texture3D
+  _glewInfo_GL_EXT_texture3D();
+#endif /* GL_EXT_texture3D */
+#ifdef GL_EXT_texture_array
+  _glewInfo_GL_EXT_texture_array();
+#endif /* GL_EXT_texture_array */
+#ifdef GL_EXT_texture_buffer_object
+  _glewInfo_GL_EXT_texture_buffer_object();
+#endif /* GL_EXT_texture_buffer_object */
+#ifdef GL_EXT_texture_compression_dxt1
+  _glewInfo_GL_EXT_texture_compression_dxt1();
+#endif /* GL_EXT_texture_compression_dxt1 */
+#ifdef GL_EXT_texture_compression_latc
+  _glewInfo_GL_EXT_texture_compression_latc();
+#endif /* GL_EXT_texture_compression_latc */
+#ifdef GL_EXT_texture_compression_rgtc
+  _glewInfo_GL_EXT_texture_compression_rgtc();
+#endif /* GL_EXT_texture_compression_rgtc */
+#ifdef GL_EXT_texture_compression_s3tc
+  _glewInfo_GL_EXT_texture_compression_s3tc();
+#endif /* GL_EXT_texture_compression_s3tc */
+#ifdef GL_EXT_texture_cube_map
+  _glewInfo_GL_EXT_texture_cube_map();
+#endif /* GL_EXT_texture_cube_map */
+#ifdef GL_EXT_texture_edge_clamp
+  _glewInfo_GL_EXT_texture_edge_clamp();
+#endif /* GL_EXT_texture_edge_clamp */
+#ifdef GL_EXT_texture_env
+  _glewInfo_GL_EXT_texture_env();
+#endif /* GL_EXT_texture_env */
+#ifdef GL_EXT_texture_env_add
+  _glewInfo_GL_EXT_texture_env_add();
+#endif /* GL_EXT_texture_env_add */
+#ifdef GL_EXT_texture_env_combine
+  _glewInfo_GL_EXT_texture_env_combine();
+#endif /* GL_EXT_texture_env_combine */
+#ifdef GL_EXT_texture_env_dot3
+  _glewInfo_GL_EXT_texture_env_dot3();
+#endif /* GL_EXT_texture_env_dot3 */
+#ifdef GL_EXT_texture_filter_anisotropic
+  _glewInfo_GL_EXT_texture_filter_anisotropic();
+#endif /* GL_EXT_texture_filter_anisotropic */
+#ifdef GL_EXT_texture_integer
+  _glewInfo_GL_EXT_texture_integer();
+#endif /* GL_EXT_texture_integer */
+#ifdef GL_EXT_texture_lod_bias
+  _glewInfo_GL_EXT_texture_lod_bias();
+#endif /* GL_EXT_texture_lod_bias */
+#ifdef GL_EXT_texture_mirror_clamp
+  _glewInfo_GL_EXT_texture_mirror_clamp();
+#endif /* GL_EXT_texture_mirror_clamp */
+#ifdef GL_EXT_texture_object
+  _glewInfo_GL_EXT_texture_object();
+#endif /* GL_EXT_texture_object */
+#ifdef GL_EXT_texture_perturb_normal
+  _glewInfo_GL_EXT_texture_perturb_normal();
+#endif /* GL_EXT_texture_perturb_normal */
+#ifdef GL_EXT_texture_rectangle
+  _glewInfo_GL_EXT_texture_rectangle();
+#endif /* GL_EXT_texture_rectangle */
+#ifdef GL_EXT_texture_sRGB
+  _glewInfo_GL_EXT_texture_sRGB();
+#endif /* GL_EXT_texture_sRGB */
+#ifdef GL_EXT_texture_shared_exponent
+  _glewInfo_GL_EXT_texture_shared_exponent();
+#endif /* GL_EXT_texture_shared_exponent */
+#ifdef GL_EXT_texture_swizzle
+  _glewInfo_GL_EXT_texture_swizzle();
+#endif /* GL_EXT_texture_swizzle */
+#ifdef GL_EXT_timer_query
+  _glewInfo_GL_EXT_timer_query();
+#endif /* GL_EXT_timer_query */
+#ifdef GL_EXT_transform_feedback
+  _glewInfo_GL_EXT_transform_feedback();
+#endif /* GL_EXT_transform_feedback */
+#ifdef GL_EXT_vertex_array
+  _glewInfo_GL_EXT_vertex_array();
+#endif /* GL_EXT_vertex_array */
+#ifdef GL_EXT_vertex_array_bgra
+  _glewInfo_GL_EXT_vertex_array_bgra();
+#endif /* GL_EXT_vertex_array_bgra */
+#ifdef GL_EXT_vertex_shader
+  _glewInfo_GL_EXT_vertex_shader();
+#endif /* GL_EXT_vertex_shader */
+#ifdef GL_EXT_vertex_weighting
+  _glewInfo_GL_EXT_vertex_weighting();
+#endif /* GL_EXT_vertex_weighting */
+#ifdef GL_GREMEDY_frame_terminator
+  _glewInfo_GL_GREMEDY_frame_terminator();
+#endif /* GL_GREMEDY_frame_terminator */
+#ifdef GL_GREMEDY_string_marker
+  _glewInfo_GL_GREMEDY_string_marker();
+#endif /* GL_GREMEDY_string_marker */
+#ifdef GL_HP_convolution_border_modes
+  _glewInfo_GL_HP_convolution_border_modes();
+#endif /* GL_HP_convolution_border_modes */
+#ifdef GL_HP_image_transform
+  _glewInfo_GL_HP_image_transform();
+#endif /* GL_HP_image_transform */
+#ifdef GL_HP_occlusion_test
+  _glewInfo_GL_HP_occlusion_test();
+#endif /* GL_HP_occlusion_test */
+#ifdef GL_HP_texture_lighting
+  _glewInfo_GL_HP_texture_lighting();
+#endif /* GL_HP_texture_lighting */
+#ifdef GL_IBM_cull_vertex
+  _glewInfo_GL_IBM_cull_vertex();
+#endif /* GL_IBM_cull_vertex */
+#ifdef GL_IBM_multimode_draw_arrays
+  _glewInfo_GL_IBM_multimode_draw_arrays();
+#endif /* GL_IBM_multimode_draw_arrays */
+#ifdef GL_IBM_rasterpos_clip
+  _glewInfo_GL_IBM_rasterpos_clip();
+#endif /* GL_IBM_rasterpos_clip */
+#ifdef GL_IBM_static_data
+  _glewInfo_GL_IBM_static_data();
+#endif /* GL_IBM_static_data */
+#ifdef GL_IBM_texture_mirrored_repeat
+  _glewInfo_GL_IBM_texture_mirrored_repeat();
+#endif /* GL_IBM_texture_mirrored_repeat */
+#ifdef GL_IBM_vertex_array_lists
+  _glewInfo_GL_IBM_vertex_array_lists();
+#endif /* GL_IBM_vertex_array_lists */
+#ifdef GL_INGR_color_clamp
+  _glewInfo_GL_INGR_color_clamp();
+#endif /* GL_INGR_color_clamp */
+#ifdef GL_INGR_interlace_read
+  _glewInfo_GL_INGR_interlace_read();
+#endif /* GL_INGR_interlace_read */
+#ifdef GL_INTEL_parallel_arrays
+  _glewInfo_GL_INTEL_parallel_arrays();
+#endif /* GL_INTEL_parallel_arrays */
+#ifdef GL_INTEL_texture_scissor
+  _glewInfo_GL_INTEL_texture_scissor();
+#endif /* GL_INTEL_texture_scissor */
+#ifdef GL_KTX_buffer_region
+  _glewInfo_GL_KTX_buffer_region();
+#endif /* GL_KTX_buffer_region */
+#ifdef GL_MESAX_texture_stack
+  _glewInfo_GL_MESAX_texture_stack();
+#endif /* GL_MESAX_texture_stack */
+#ifdef GL_MESA_pack_invert
+  _glewInfo_GL_MESA_pack_invert();
+#endif /* GL_MESA_pack_invert */
+#ifdef GL_MESA_resize_buffers
+  _glewInfo_GL_MESA_resize_buffers();
+#endif /* GL_MESA_resize_buffers */
+#ifdef GL_MESA_window_pos
+  _glewInfo_GL_MESA_window_pos();
+#endif /* GL_MESA_window_pos */
+#ifdef GL_MESA_ycbcr_texture
+  _glewInfo_GL_MESA_ycbcr_texture();
+#endif /* GL_MESA_ycbcr_texture */
+#ifdef GL_NV_blend_square
+  _glewInfo_GL_NV_blend_square();
+#endif /* GL_NV_blend_square */
+#ifdef GL_NV_conditional_render
+  _glewInfo_GL_NV_conditional_render();
+#endif /* GL_NV_conditional_render */
+#ifdef GL_NV_copy_depth_to_color
+  _glewInfo_GL_NV_copy_depth_to_color();
+#endif /* GL_NV_copy_depth_to_color */
+#ifdef GL_NV_depth_buffer_float
+  _glewInfo_GL_NV_depth_buffer_float();
+#endif /* GL_NV_depth_buffer_float */
+#ifdef GL_NV_depth_clamp
+  _glewInfo_GL_NV_depth_clamp();
+#endif /* GL_NV_depth_clamp */
+#ifdef GL_NV_depth_range_unclamped
+  _glewInfo_GL_NV_depth_range_unclamped();
+#endif /* GL_NV_depth_range_unclamped */
+#ifdef GL_NV_evaluators
+  _glewInfo_GL_NV_evaluators();
+#endif /* GL_NV_evaluators */
+#ifdef GL_NV_explicit_multisample
+  _glewInfo_GL_NV_explicit_multisample();
+#endif /* GL_NV_explicit_multisample */
+#ifdef GL_NV_fence
+  _glewInfo_GL_NV_fence();
+#endif /* GL_NV_fence */
+#ifdef GL_NV_float_buffer
+  _glewInfo_GL_NV_float_buffer();
+#endif /* GL_NV_float_buffer */
+#ifdef GL_NV_fog_distance
+  _glewInfo_GL_NV_fog_distance();
+#endif /* GL_NV_fog_distance */
+#ifdef GL_NV_fragment_program
+  _glewInfo_GL_NV_fragment_program();
+#endif /* GL_NV_fragment_program */
+#ifdef GL_NV_fragment_program2
+  _glewInfo_GL_NV_fragment_program2();
+#endif /* GL_NV_fragment_program2 */
+#ifdef GL_NV_fragment_program4
+  _glewInfo_GL_NV_fragment_program4();
+#endif /* GL_NV_fragment_program4 */
+#ifdef GL_NV_fragment_program_option
+  _glewInfo_GL_NV_fragment_program_option();
+#endif /* GL_NV_fragment_program_option */
+#ifdef GL_NV_framebuffer_multisample_coverage
+  _glewInfo_GL_NV_framebuffer_multisample_coverage();
+#endif /* GL_NV_framebuffer_multisample_coverage */
+#ifdef GL_NV_geometry_program4
+  _glewInfo_GL_NV_geometry_program4();
+#endif /* GL_NV_geometry_program4 */
+#ifdef GL_NV_geometry_shader4
+  _glewInfo_GL_NV_geometry_shader4();
+#endif /* GL_NV_geometry_shader4 */
+#ifdef GL_NV_gpu_program4
+  _glewInfo_GL_NV_gpu_program4();
+#endif /* GL_NV_gpu_program4 */
+#ifdef GL_NV_half_float
+  _glewInfo_GL_NV_half_float();
+#endif /* GL_NV_half_float */
+#ifdef GL_NV_light_max_exponent
+  _glewInfo_GL_NV_light_max_exponent();
+#endif /* GL_NV_light_max_exponent */
+#ifdef GL_NV_multisample_filter_hint
+  _glewInfo_GL_NV_multisample_filter_hint();
+#endif /* GL_NV_multisample_filter_hint */
+#ifdef GL_NV_occlusion_query
+  _glewInfo_GL_NV_occlusion_query();
+#endif /* GL_NV_occlusion_query */
+#ifdef GL_NV_packed_depth_stencil
+  _glewInfo_GL_NV_packed_depth_stencil();
+#endif /* GL_NV_packed_depth_stencil */
+#ifdef GL_NV_parameter_buffer_object
+  _glewInfo_GL_NV_parameter_buffer_object();
+#endif /* GL_NV_parameter_buffer_object */
+#ifdef GL_NV_pixel_data_range
+  _glewInfo_GL_NV_pixel_data_range();
+#endif /* GL_NV_pixel_data_range */
+#ifdef GL_NV_point_sprite
+  _glewInfo_GL_NV_point_sprite();
+#endif /* GL_NV_point_sprite */
+#ifdef GL_NV_present_video
+  _glewInfo_GL_NV_present_video();
+#endif /* GL_NV_present_video */
+#ifdef GL_NV_primitive_restart
+  _glewInfo_GL_NV_primitive_restart();
+#endif /* GL_NV_primitive_restart */
+#ifdef GL_NV_register_combiners
+  _glewInfo_GL_NV_register_combiners();
+#endif /* GL_NV_register_combiners */
+#ifdef GL_NV_register_combiners2
+  _glewInfo_GL_NV_register_combiners2();
+#endif /* GL_NV_register_combiners2 */
+#ifdef GL_NV_texgen_emboss
+  _glewInfo_GL_NV_texgen_emboss();
+#endif /* GL_NV_texgen_emboss */
+#ifdef GL_NV_texgen_reflection
+  _glewInfo_GL_NV_texgen_reflection();
+#endif /* GL_NV_texgen_reflection */
+#ifdef GL_NV_texture_compression_vtc
+  _glewInfo_GL_NV_texture_compression_vtc();
+#endif /* GL_NV_texture_compression_vtc */
+#ifdef GL_NV_texture_env_combine4
+  _glewInfo_GL_NV_texture_env_combine4();
+#endif /* GL_NV_texture_env_combine4 */
+#ifdef GL_NV_texture_expand_normal
+  _glewInfo_GL_NV_texture_expand_normal();
+#endif /* GL_NV_texture_expand_normal */
+#ifdef GL_NV_texture_rectangle
+  _glewInfo_GL_NV_texture_rectangle();
+#endif /* GL_NV_texture_rectangle */
+#ifdef GL_NV_texture_shader
+  _glewInfo_GL_NV_texture_shader();
+#endif /* GL_NV_texture_shader */
+#ifdef GL_NV_texture_shader2
+  _glewInfo_GL_NV_texture_shader2();
+#endif /* GL_NV_texture_shader2 */
+#ifdef GL_NV_texture_shader3
+  _glewInfo_GL_NV_texture_shader3();
+#endif /* GL_NV_texture_shader3 */
+#ifdef GL_NV_transform_feedback
+  _glewInfo_GL_NV_transform_feedback();
+#endif /* GL_NV_transform_feedback */
+#ifdef GL_NV_vertex_array_range
+  _glewInfo_GL_NV_vertex_array_range();
+#endif /* GL_NV_vertex_array_range */
+#ifdef GL_NV_vertex_array_range2
+  _glewInfo_GL_NV_vertex_array_range2();
+#endif /* GL_NV_vertex_array_range2 */
+#ifdef GL_NV_vertex_program
+  _glewInfo_GL_NV_vertex_program();
+#endif /* GL_NV_vertex_program */
+#ifdef GL_NV_vertex_program1_1
+  _glewInfo_GL_NV_vertex_program1_1();
+#endif /* GL_NV_vertex_program1_1 */
+#ifdef GL_NV_vertex_program2
+  _glewInfo_GL_NV_vertex_program2();
+#endif /* GL_NV_vertex_program2 */
+#ifdef GL_NV_vertex_program2_option
+  _glewInfo_GL_NV_vertex_program2_option();
+#endif /* GL_NV_vertex_program2_option */
+#ifdef GL_NV_vertex_program3
+  _glewInfo_GL_NV_vertex_program3();
+#endif /* GL_NV_vertex_program3 */
+#ifdef GL_NV_vertex_program4
+  _glewInfo_GL_NV_vertex_program4();
+#endif /* GL_NV_vertex_program4 */
+#ifdef GL_OES_byte_coordinates
+  _glewInfo_GL_OES_byte_coordinates();
+#endif /* GL_OES_byte_coordinates */
+#ifdef GL_OES_compressed_paletted_texture
+  _glewInfo_GL_OES_compressed_paletted_texture();
+#endif /* GL_OES_compressed_paletted_texture */
+#ifdef GL_OES_read_format
+  _glewInfo_GL_OES_read_format();
+#endif /* GL_OES_read_format */
+#ifdef GL_OES_single_precision
+  _glewInfo_GL_OES_single_precision();
+#endif /* GL_OES_single_precision */
+#ifdef GL_OML_interlace
+  _glewInfo_GL_OML_interlace();
+#endif /* GL_OML_interlace */
+#ifdef GL_OML_resample
+  _glewInfo_GL_OML_resample();
+#endif /* GL_OML_resample */
+#ifdef GL_OML_subsample
+  _glewInfo_GL_OML_subsample();
+#endif /* GL_OML_subsample */
+#ifdef GL_PGI_misc_hints
+  _glewInfo_GL_PGI_misc_hints();
+#endif /* GL_PGI_misc_hints */
+#ifdef GL_PGI_vertex_hints
+  _glewInfo_GL_PGI_vertex_hints();
+#endif /* GL_PGI_vertex_hints */
+#ifdef GL_REND_screen_coordinates
+  _glewInfo_GL_REND_screen_coordinates();
+#endif /* GL_REND_screen_coordinates */
+#ifdef GL_S3_s3tc
+  _glewInfo_GL_S3_s3tc();
+#endif /* GL_S3_s3tc */
+#ifdef GL_SGIS_color_range
+  _glewInfo_GL_SGIS_color_range();
+#endif /* GL_SGIS_color_range */
+#ifdef GL_SGIS_detail_texture
+  _glewInfo_GL_SGIS_detail_texture();
+#endif /* GL_SGIS_detail_texture */
+#ifdef GL_SGIS_fog_function
+  _glewInfo_GL_SGIS_fog_function();
+#endif /* GL_SGIS_fog_function */
+#ifdef GL_SGIS_generate_mipmap
+  _glewInfo_GL_SGIS_generate_mipmap();
+#endif /* GL_SGIS_generate_mipmap */
+#ifdef GL_SGIS_multisample
+  _glewInfo_GL_SGIS_multisample();
+#endif /* GL_SGIS_multisample */
+#ifdef GL_SGIS_pixel_texture
+  _glewInfo_GL_SGIS_pixel_texture();
+#endif /* GL_SGIS_pixel_texture */
+#ifdef GL_SGIS_point_line_texgen
+  _glewInfo_GL_SGIS_point_line_texgen();
+#endif /* GL_SGIS_point_line_texgen */
+#ifdef GL_SGIS_sharpen_texture
+  _glewInfo_GL_SGIS_sharpen_texture();
+#endif /* GL_SGIS_sharpen_texture */
+#ifdef GL_SGIS_texture4D
+  _glewInfo_GL_SGIS_texture4D();
+#endif /* GL_SGIS_texture4D */
+#ifdef GL_SGIS_texture_border_clamp
+  _glewInfo_GL_SGIS_texture_border_clamp();
+#endif /* GL_SGIS_texture_border_clamp */
+#ifdef GL_SGIS_texture_edge_clamp
+  _glewInfo_GL_SGIS_texture_edge_clamp();
+#endif /* GL_SGIS_texture_edge_clamp */
+#ifdef GL_SGIS_texture_filter4
+  _glewInfo_GL_SGIS_texture_filter4();
+#endif /* GL_SGIS_texture_filter4 */
+#ifdef GL_SGIS_texture_lod
+  _glewInfo_GL_SGIS_texture_lod();
+#endif /* GL_SGIS_texture_lod */
+#ifdef GL_SGIS_texture_select
+  _glewInfo_GL_SGIS_texture_select();
+#endif /* GL_SGIS_texture_select */
+#ifdef GL_SGIX_async
+  _glewInfo_GL_SGIX_async();
+#endif /* GL_SGIX_async */
+#ifdef GL_SGIX_async_histogram
+  _glewInfo_GL_SGIX_async_histogram();
+#endif /* GL_SGIX_async_histogram */
+#ifdef GL_SGIX_async_pixel
+  _glewInfo_GL_SGIX_async_pixel();
+#endif /* GL_SGIX_async_pixel */
+#ifdef GL_SGIX_blend_alpha_minmax
+  _glewInfo_GL_SGIX_blend_alpha_minmax();
+#endif /* GL_SGIX_blend_alpha_minmax */
+#ifdef GL_SGIX_clipmap
+  _glewInfo_GL_SGIX_clipmap();
+#endif /* GL_SGIX_clipmap */
+#ifdef GL_SGIX_convolution_accuracy
+  _glewInfo_GL_SGIX_convolution_accuracy();
+#endif /* GL_SGIX_convolution_accuracy */
+#ifdef GL_SGIX_depth_texture
+  _glewInfo_GL_SGIX_depth_texture();
+#endif /* GL_SGIX_depth_texture */
+#ifdef GL_SGIX_flush_raster
+  _glewInfo_GL_SGIX_flush_raster();
+#endif /* GL_SGIX_flush_raster */
+#ifdef GL_SGIX_fog_offset
+  _glewInfo_GL_SGIX_fog_offset();
+#endif /* GL_SGIX_fog_offset */
+#ifdef GL_SGIX_fog_texture
+  _glewInfo_GL_SGIX_fog_texture();
+#endif /* GL_SGIX_fog_texture */
+#ifdef GL_SGIX_fragment_specular_lighting
+  _glewInfo_GL_SGIX_fragment_specular_lighting();
+#endif /* GL_SGIX_fragment_specular_lighting */
+#ifdef GL_SGIX_framezoom
+  _glewInfo_GL_SGIX_framezoom();
+#endif /* GL_SGIX_framezoom */
+#ifdef GL_SGIX_interlace
+  _glewInfo_GL_SGIX_interlace();
+#endif /* GL_SGIX_interlace */
+#ifdef GL_SGIX_ir_instrument1
+  _glewInfo_GL_SGIX_ir_instrument1();
+#endif /* GL_SGIX_ir_instrument1 */
+#ifdef GL_SGIX_list_priority
+  _glewInfo_GL_SGIX_list_priority();
+#endif /* GL_SGIX_list_priority */
+#ifdef GL_SGIX_pixel_texture
+  _glewInfo_GL_SGIX_pixel_texture();
+#endif /* GL_SGIX_pixel_texture */
+#ifdef GL_SGIX_pixel_texture_bits
+  _glewInfo_GL_SGIX_pixel_texture_bits();
+#endif /* GL_SGIX_pixel_texture_bits */
+#ifdef GL_SGIX_reference_plane
+  _glewInfo_GL_SGIX_reference_plane();
+#endif /* GL_SGIX_reference_plane */
+#ifdef GL_SGIX_resample
+  _glewInfo_GL_SGIX_resample();
+#endif /* GL_SGIX_resample */
+#ifdef GL_SGIX_shadow
+  _glewInfo_GL_SGIX_shadow();
+#endif /* GL_SGIX_shadow */
+#ifdef GL_SGIX_shadow_ambient
+  _glewInfo_GL_SGIX_shadow_ambient();
+#endif /* GL_SGIX_shadow_ambient */
+#ifdef GL_SGIX_sprite
+  _glewInfo_GL_SGIX_sprite();
+#endif /* GL_SGIX_sprite */
+#ifdef GL_SGIX_tag_sample_buffer
+  _glewInfo_GL_SGIX_tag_sample_buffer();
+#endif /* GL_SGIX_tag_sample_buffer */
+#ifdef GL_SGIX_texture_add_env
+  _glewInfo_GL_SGIX_texture_add_env();
+#endif /* GL_SGIX_texture_add_env */
+#ifdef GL_SGIX_texture_coordinate_clamp
+  _glewInfo_GL_SGIX_texture_coordinate_clamp();
+#endif /* GL_SGIX_texture_coordinate_clamp */
+#ifdef GL_SGIX_texture_lod_bias
+  _glewInfo_GL_SGIX_texture_lod_bias();
+#endif /* GL_SGIX_texture_lod_bias */
+#ifdef GL_SGIX_texture_multi_buffer
+  _glewInfo_GL_SGIX_texture_multi_buffer();
+#endif /* GL_SGIX_texture_multi_buffer */
+#ifdef GL_SGIX_texture_range
+  _glewInfo_GL_SGIX_texture_range();
+#endif /* GL_SGIX_texture_range */
+#ifdef GL_SGIX_texture_scale_bias
+  _glewInfo_GL_SGIX_texture_scale_bias();
+#endif /* GL_SGIX_texture_scale_bias */
+#ifdef GL_SGIX_vertex_preclip
+  _glewInfo_GL_SGIX_vertex_preclip();
+#endif /* GL_SGIX_vertex_preclip */
+#ifdef GL_SGIX_vertex_preclip_hint
+  _glewInfo_GL_SGIX_vertex_preclip_hint();
+#endif /* GL_SGIX_vertex_preclip_hint */
+#ifdef GL_SGIX_ycrcb
+  _glewInfo_GL_SGIX_ycrcb();
+#endif /* GL_SGIX_ycrcb */
+#ifdef GL_SGI_color_matrix
+  _glewInfo_GL_SGI_color_matrix();
+#endif /* GL_SGI_color_matrix */
+#ifdef GL_SGI_color_table
+  _glewInfo_GL_SGI_color_table();
+#endif /* GL_SGI_color_table */
+#ifdef GL_SGI_texture_color_table
+  _glewInfo_GL_SGI_texture_color_table();
+#endif /* GL_SGI_texture_color_table */
+#ifdef GL_SUNX_constant_data
+  _glewInfo_GL_SUNX_constant_data();
+#endif /* GL_SUNX_constant_data */
+#ifdef GL_SUN_convolution_border_modes
+  _glewInfo_GL_SUN_convolution_border_modes();
+#endif /* GL_SUN_convolution_border_modes */
+#ifdef GL_SUN_global_alpha
+  _glewInfo_GL_SUN_global_alpha();
+#endif /* GL_SUN_global_alpha */
+#ifdef GL_SUN_mesh_array
+  _glewInfo_GL_SUN_mesh_array();
+#endif /* GL_SUN_mesh_array */
+#ifdef GL_SUN_read_video_pixels
+  _glewInfo_GL_SUN_read_video_pixels();
+#endif /* GL_SUN_read_video_pixels */
+#ifdef GL_SUN_slice_accum
+  _glewInfo_GL_SUN_slice_accum();
+#endif /* GL_SUN_slice_accum */
+#ifdef GL_SUN_triangle_list
+  _glewInfo_GL_SUN_triangle_list();
+#endif /* GL_SUN_triangle_list */
+#ifdef GL_SUN_vertex
+  _glewInfo_GL_SUN_vertex();
+#endif /* GL_SUN_vertex */
+#ifdef GL_WIN_phong_shading
+  _glewInfo_GL_WIN_phong_shading();
+#endif /* GL_WIN_phong_shading */
+#ifdef GL_WIN_specular_fog
+  _glewInfo_GL_WIN_specular_fog();
+#endif /* GL_WIN_specular_fog */
+#ifdef GL_WIN_swap_hint
+  _glewInfo_GL_WIN_swap_hint();
+#endif /* GL_WIN_swap_hint */
+}
+
+/* ------------------------------------------------------------------------ */
+
+#ifdef _WIN32
+
+static void wglewInfo ()
+{
+#ifdef WGL_3DFX_multisample
+  _glewInfo_WGL_3DFX_multisample();
+#endif /* WGL_3DFX_multisample */
+#ifdef WGL_3DL_stereo_control
+  _glewInfo_WGL_3DL_stereo_control();
+#endif /* WGL_3DL_stereo_control */
+#ifdef WGL_ARB_buffer_region
+  _glewInfo_WGL_ARB_buffer_region();
+#endif /* WGL_ARB_buffer_region */
+#ifdef WGL_ARB_create_context
+  _glewInfo_WGL_ARB_create_context();
+#endif /* WGL_ARB_create_context */
+#ifdef WGL_ARB_extensions_string
+  _glewInfo_WGL_ARB_extensions_string();
+#endif /* WGL_ARB_extensions_string */
+#ifdef WGL_ARB_framebuffer_sRGB
+  _glewInfo_WGL_ARB_framebuffer_sRGB();
+#endif /* WGL_ARB_framebuffer_sRGB */
+#ifdef WGL_ARB_make_current_read
+  _glewInfo_WGL_ARB_make_current_read();
+#endif /* WGL_ARB_make_current_read */
+#ifdef WGL_ARB_multisample
+  _glewInfo_WGL_ARB_multisample();
+#endif /* WGL_ARB_multisample */
+#ifdef WGL_ARB_pbuffer
+  _glewInfo_WGL_ARB_pbuffer();
+#endif /* WGL_ARB_pbuffer */
+#ifdef WGL_ARB_pixel_format
+  _glewInfo_WGL_ARB_pixel_format();
+#endif /* WGL_ARB_pixel_format */
+#ifdef WGL_ARB_pixel_format_float
+  _glewInfo_WGL_ARB_pixel_format_float();
+#endif /* WGL_ARB_pixel_format_float */
+#ifdef WGL_ARB_render_texture
+  _glewInfo_WGL_ARB_render_texture();
+#endif /* WGL_ARB_render_texture */
+#ifdef WGL_ATI_pixel_format_float
+  _glewInfo_WGL_ATI_pixel_format_float();
+#endif /* WGL_ATI_pixel_format_float */
+#ifdef WGL_ATI_render_texture_rectangle
+  _glewInfo_WGL_ATI_render_texture_rectangle();
+#endif /* WGL_ATI_render_texture_rectangle */
+#ifdef WGL_EXT_depth_float
+  _glewInfo_WGL_EXT_depth_float();
+#endif /* WGL_EXT_depth_float */
+#ifdef WGL_EXT_display_color_table
+  _glewInfo_WGL_EXT_display_color_table();
+#endif /* WGL_EXT_display_color_table */
+#ifdef WGL_EXT_extensions_string
+  _glewInfo_WGL_EXT_extensions_string();
+#endif /* WGL_EXT_extensions_string */
+#ifdef WGL_EXT_framebuffer_sRGB
+  _glewInfo_WGL_EXT_framebuffer_sRGB();
+#endif /* WGL_EXT_framebuffer_sRGB */
+#ifdef WGL_EXT_make_current_read
+  _glewInfo_WGL_EXT_make_current_read();
+#endif /* WGL_EXT_make_current_read */
+#ifdef WGL_EXT_multisample
+  _glewInfo_WGL_EXT_multisample();
+#endif /* WGL_EXT_multisample */
+#ifdef WGL_EXT_pbuffer
+  _glewInfo_WGL_EXT_pbuffer();
+#endif /* WGL_EXT_pbuffer */
+#ifdef WGL_EXT_pixel_format
+  _glewInfo_WGL_EXT_pixel_format();
+#endif /* WGL_EXT_pixel_format */
+#ifdef WGL_EXT_pixel_format_packed_float
+  _glewInfo_WGL_EXT_pixel_format_packed_float();
+#endif /* WGL_EXT_pixel_format_packed_float */
+#ifdef WGL_EXT_swap_control
+  _glewInfo_WGL_EXT_swap_control();
+#endif /* WGL_EXT_swap_control */
+#ifdef WGL_I3D_digital_video_control
+  _glewInfo_WGL_I3D_digital_video_control();
+#endif /* WGL_I3D_digital_video_control */
+#ifdef WGL_I3D_gamma
+  _glewInfo_WGL_I3D_gamma();
+#endif /* WGL_I3D_gamma */
+#ifdef WGL_I3D_genlock
+  _glewInfo_WGL_I3D_genlock();
+#endif /* WGL_I3D_genlock */
+#ifdef WGL_I3D_image_buffer
+  _glewInfo_WGL_I3D_image_buffer();
+#endif /* WGL_I3D_image_buffer */
+#ifdef WGL_I3D_swap_frame_lock
+  _glewInfo_WGL_I3D_swap_frame_lock();
+#endif /* WGL_I3D_swap_frame_lock */
+#ifdef WGL_I3D_swap_frame_usage
+  _glewInfo_WGL_I3D_swap_frame_usage();
+#endif /* WGL_I3D_swap_frame_usage */
+#ifdef WGL_NV_float_buffer
+  _glewInfo_WGL_NV_float_buffer();
+#endif /* WGL_NV_float_buffer */
+#ifdef WGL_NV_gpu_affinity
+  _glewInfo_WGL_NV_gpu_affinity();
+#endif /* WGL_NV_gpu_affinity */
+#ifdef WGL_NV_present_video
+  _glewInfo_WGL_NV_present_video();
+#endif /* WGL_NV_present_video */
+#ifdef WGL_NV_render_depth_texture
+  _glewInfo_WGL_NV_render_depth_texture();
+#endif /* WGL_NV_render_depth_texture */
+#ifdef WGL_NV_render_texture_rectangle
+  _glewInfo_WGL_NV_render_texture_rectangle();
+#endif /* WGL_NV_render_texture_rectangle */
+#ifdef WGL_NV_swap_group
+  _glewInfo_WGL_NV_swap_group();
+#endif /* WGL_NV_swap_group */
+#ifdef WGL_NV_vertex_array_range
+  _glewInfo_WGL_NV_vertex_array_range();
+#endif /* WGL_NV_vertex_array_range */
+#ifdef WGL_NV_video_output
+  _glewInfo_WGL_NV_video_output();
+#endif /* WGL_NV_video_output */
+#ifdef WGL_OML_sync_control
+  _glewInfo_WGL_OML_sync_control();
+#endif /* WGL_OML_sync_control */
+}
+
+#else /* _UNIX */
+
+static void glxewInfo ()
+{
+#ifdef GLX_VERSION_1_2
+  _glewInfo_GLX_VERSION_1_2();
+#endif /* GLX_VERSION_1_2 */
+#ifdef GLX_VERSION_1_3
+  _glewInfo_GLX_VERSION_1_3();
+#endif /* GLX_VERSION_1_3 */
+#ifdef GLX_VERSION_1_4
+  _glewInfo_GLX_VERSION_1_4();
+#endif /* GLX_VERSION_1_4 */
+#ifdef GLX_3DFX_multisample
+  _glewInfo_GLX_3DFX_multisample();
+#endif /* GLX_3DFX_multisample */
+#ifdef GLX_ARB_create_context
+  _glewInfo_GLX_ARB_create_context();
+#endif /* GLX_ARB_create_context */
+#ifdef GLX_ARB_fbconfig_float
+  _glewInfo_GLX_ARB_fbconfig_float();
+#endif /* GLX_ARB_fbconfig_float */
+#ifdef GLX_ARB_framebuffer_sRGB
+  _glewInfo_GLX_ARB_framebuffer_sRGB();
+#endif /* GLX_ARB_framebuffer_sRGB */
+#ifdef GLX_ARB_get_proc_address
+  _glewInfo_GLX_ARB_get_proc_address();
+#endif /* GLX_ARB_get_proc_address */
+#ifdef GLX_ARB_multisample
+  _glewInfo_GLX_ARB_multisample();
+#endif /* GLX_ARB_multisample */
+#ifdef GLX_ATI_pixel_format_float
+  _glewInfo_GLX_ATI_pixel_format_float();
+#endif /* GLX_ATI_pixel_format_float */
+#ifdef GLX_ATI_render_texture
+  _glewInfo_GLX_ATI_render_texture();
+#endif /* GLX_ATI_render_texture */
+#ifdef GLX_EXT_fbconfig_packed_float
+  _glewInfo_GLX_EXT_fbconfig_packed_float();
+#endif /* GLX_EXT_fbconfig_packed_float */
+#ifdef GLX_EXT_framebuffer_sRGB
+  _glewInfo_GLX_EXT_framebuffer_sRGB();
+#endif /* GLX_EXT_framebuffer_sRGB */
+#ifdef GLX_EXT_import_context
+  _glewInfo_GLX_EXT_import_context();
+#endif /* GLX_EXT_import_context */
+#ifdef GLX_EXT_scene_marker
+  _glewInfo_GLX_EXT_scene_marker();
+#endif /* GLX_EXT_scene_marker */
+#ifdef GLX_EXT_texture_from_pixmap
+  _glewInfo_GLX_EXT_texture_from_pixmap();
+#endif /* GLX_EXT_texture_from_pixmap */
+#ifdef GLX_EXT_visual_info
+  _glewInfo_GLX_EXT_visual_info();
+#endif /* GLX_EXT_visual_info */
+#ifdef GLX_EXT_visual_rating
+  _glewInfo_GLX_EXT_visual_rating();
+#endif /* GLX_EXT_visual_rating */
+#ifdef GLX_MESA_agp_offset
+  _glewInfo_GLX_MESA_agp_offset();
+#endif /* GLX_MESA_agp_offset */
+#ifdef GLX_MESA_copy_sub_buffer
+  _glewInfo_GLX_MESA_copy_sub_buffer();
+#endif /* GLX_MESA_copy_sub_buffer */
+#ifdef GLX_MESA_pixmap_colormap
+  _glewInfo_GLX_MESA_pixmap_colormap();
+#endif /* GLX_MESA_pixmap_colormap */
+#ifdef GLX_MESA_release_buffers
+  _glewInfo_GLX_MESA_release_buffers();
+#endif /* GLX_MESA_release_buffers */
+#ifdef GLX_MESA_set_3dfx_mode
+  _glewInfo_GLX_MESA_set_3dfx_mode();
+#endif /* GLX_MESA_set_3dfx_mode */
+#ifdef GLX_NV_float_buffer
+  _glewInfo_GLX_NV_float_buffer();
+#endif /* GLX_NV_float_buffer */
+#ifdef GLX_NV_present_video
+  _glewInfo_GLX_NV_present_video();
+#endif /* GLX_NV_present_video */
+#ifdef GLX_NV_swap_group
+  _glewInfo_GLX_NV_swap_group();
+#endif /* GLX_NV_swap_group */
+#ifdef GLX_NV_vertex_array_range
+  _glewInfo_GLX_NV_vertex_array_range();
+#endif /* GLX_NV_vertex_array_range */
+#ifdef GLX_NV_video_output
+  _glewInfo_GLX_NV_video_output();
+#endif /* GLX_NV_video_output */
+#ifdef GLX_OML_swap_method
+  _glewInfo_GLX_OML_swap_method();
+#endif /* GLX_OML_swap_method */
+#if defined(GLX_OML_sync_control) && defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)
+#include <inttypes.h>
+  _glewInfo_GLX_OML_sync_control();
+#endif /* GLX_OML_sync_control */
+#ifdef GLX_SGIS_blended_overlay
+  _glewInfo_GLX_SGIS_blended_overlay();
+#endif /* GLX_SGIS_blended_overlay */
+#ifdef GLX_SGIS_color_range
+  _glewInfo_GLX_SGIS_color_range();
+#endif /* GLX_SGIS_color_range */
+#ifdef GLX_SGIS_multisample
+  _glewInfo_GLX_SGIS_multisample();
+#endif /* GLX_SGIS_multisample */
+#ifdef GLX_SGIS_shared_multisample
+  _glewInfo_GLX_SGIS_shared_multisample();
+#endif /* GLX_SGIS_shared_multisample */
+#ifdef GLX_SGIX_fbconfig
+  _glewInfo_GLX_SGIX_fbconfig();
+#endif /* GLX_SGIX_fbconfig */
+#ifdef GLX_SGIX_hyperpipe
+  _glewInfo_GLX_SGIX_hyperpipe();
+#endif /* GLX_SGIX_hyperpipe */
+#ifdef GLX_SGIX_pbuffer
+  _glewInfo_GLX_SGIX_pbuffer();
+#endif /* GLX_SGIX_pbuffer */
+#ifdef GLX_SGIX_swap_barrier
+  _glewInfo_GLX_SGIX_swap_barrier();
+#endif /* GLX_SGIX_swap_barrier */
+#ifdef GLX_SGIX_swap_group
+  _glewInfo_GLX_SGIX_swap_group();
+#endif /* GLX_SGIX_swap_group */
+#ifdef GLX_SGIX_video_resize
+  _glewInfo_GLX_SGIX_video_resize();
+#endif /* GLX_SGIX_video_resize */
+#ifdef GLX_SGIX_visual_select_group
+  _glewInfo_GLX_SGIX_visual_select_group();
+#endif /* GLX_SGIX_visual_select_group */
+#ifdef GLX_SGI_cushion
+  _glewInfo_GLX_SGI_cushion();
+#endif /* GLX_SGI_cushion */
+#ifdef GLX_SGI_make_current_read
+  _glewInfo_GLX_SGI_make_current_read();
+#endif /* GLX_SGI_make_current_read */
+#ifdef GLX_SGI_swap_control
+  _glewInfo_GLX_SGI_swap_control();
+#endif /* GLX_SGI_swap_control */
+#ifdef GLX_SGI_video_sync
+  _glewInfo_GLX_SGI_video_sync();
+#endif /* GLX_SGI_video_sync */
+#ifdef GLX_SUN_get_transparent_index
+  _glewInfo_GLX_SUN_get_transparent_index();
+#endif /* GLX_SUN_get_transparent_index */
+#ifdef GLX_SUN_video_resize
+  _glewInfo_GLX_SUN_video_resize();
+#endif /* GLX_SUN_video_resize */
+}
+
+#endif /* _WIN32 */
+
+/* ------------------------------------------------------------------------ */
+
+#if defined(_WIN32) || !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+int main (int argc, char** argv)
+#else
+int main (void)
+#endif
+{
+  GLuint err;
+
+#if defined(_WIN32) || !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+  char* display = NULL;
+  int visual = -1;
+
+  if (glewParseArgs(argc-1, argv+1, &display, &visual))
+  {
+#if defined(_WIN32)
+    fprintf(stderr, "Usage: glewinfo [-pf <id>]\n");
+#else
+    fprintf(stderr, "Usage: glewinfo [-display <display>] [-visual <id>]\n");
+#endif
+    return 1;
+  }
+#endif
+
+#if defined(_WIN32)
+  if (GL_TRUE == glewCreateContext(&visual))
+#elif defined(__APPLE__) && !defined(GLEW_APPLE_GLX)
+  if (GL_TRUE == glewCreateContext())
+#else
+  if (GL_TRUE == glewCreateContext(display, &visual))
+#endif
+  {
+    fprintf(stderr, "Error: glewCreateContext failed\n");
+    glewDestroyContext();
+    return 1;
+  }
+  glewExperimental = GL_TRUE;
+#ifdef GLEW_MX
+  err = glewContextInit(glewGetContext());
+#ifdef _WIN32
+  err = err || wglewContextInit(wglewGetContext());
+#elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+  err = err || glxewContextInit(glxewGetContext());
+#endif
+
+#else
+  err = glewInit();
+#endif
+  if (GLEW_OK != err)
+  {
+    fprintf(stderr, "Error [main]: glewInit failed: %s\n", glewGetErrorString(err));
+    glewDestroyContext();
+    return 1;
+  }
+#if defined(_WIN32)
+  f = fopen("glewinfo.txt", "w");
+  if (f == NULL) f = stdout;
+#else
+  f = stdout;
+#endif
+  fprintf(f, "---------------------------\n");
+  fprintf(f, "    GLEW Extension Info\n");
+  fprintf(f, "---------------------------\n\n");
+  fprintf(f, "GLEW version %s\n", glewGetString(GLEW_VERSION));
+#if defined(_WIN32)
+  fprintf(f, "Reporting capabilities of pixelformat %d\n", visual);
+#elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+  fprintf(f, "Reporting capabilities of display %s, visual 0x%x\n", 
+    display == NULL ? getenv("DISPLAY") : display, visual);
+#endif
+  fprintf(f, "Running on a %s from %s\n", 
+	  glGetString(GL_RENDERER), glGetString(GL_VENDOR));
+  fprintf(f, "OpenGL version %s is supported\n", glGetString(GL_VERSION));
+  glewInfo();
+#if defined(_WIN32)
+  wglewInfo();
+#else
+  glxewInfo();
+#endif
+  if (f != stdout) fclose(f);
+  glewDestroyContext();
+  return 0;
+}
+
+/* ------------------------------------------------------------------------ */
+
+#if defined(_WIN32) || !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+GLboolean glewParseArgs (int argc, char** argv, char** display, int* visual)
+{
+  int p = 0;
+  while (p < argc)
+  {
+#if defined(_WIN32)
+    if (!strcmp(argv[p], "-pf") || !strcmp(argv[p], "-pixelformat"))
+    {
+      if (++p >= argc) return GL_TRUE;
+      *display = 0;
+      *visual = strtol(argv[p++], NULL, 0);
+    }
+    else
+      return GL_TRUE;
+#else
+    if (!strcmp(argv[p], "-display"))
+    {
+      if (++p >= argc) return GL_TRUE;
+      *display = argv[p++];
+    }
+    else if (!strcmp(argv[p], "-visual"))
+    {
+      if (++p >= argc) return GL_TRUE;
+      *visual = (int)strtol(argv[p++], NULL, 0);
+    }
+    else
+      return GL_TRUE;
+#endif
+  }
+  return GL_FALSE;
+}
+#endif
+
+/* ------------------------------------------------------------------------ */
+
+#if defined(_WIN32)
+
+HWND wnd = NULL;
+HDC dc = NULL;
+HGLRC rc = NULL;
+
+GLboolean glewCreateContext (int* pixelformat)
+{
+  WNDCLASS wc;
+  PIXELFORMATDESCRIPTOR pfd;
+  /* register window class */
+  ZeroMemory(&wc, sizeof(WNDCLASS));
+  wc.hInstance = GetModuleHandle(NULL);
+  wc.lpfnWndProc = DefWindowProc;
+  wc.lpszClassName = "GLEW";
+  if (0 == RegisterClass(&wc)) return GL_TRUE;
+  /* create window */
+  wnd = CreateWindow("GLEW", "GLEW", 0, CW_USEDEFAULT, CW_USEDEFAULT, CW_USEDEFAULT, 
+                     CW_USEDEFAULT, NULL, NULL, GetModuleHandle(NULL), NULL);
+  if (NULL == wnd) return GL_TRUE;
+  /* get the device context */
+  dc = GetDC(wnd);
+  if (NULL == dc) return GL_TRUE;
+  /* find pixel format */
+  ZeroMemory(&pfd, sizeof(PIXELFORMATDESCRIPTOR));
+  if (*pixelformat == -1) /* find default */
+  {
+    pfd.nSize = sizeof(PIXELFORMATDESCRIPTOR);
+    pfd.nVersion = 1;
+    pfd.dwFlags = PFD_DRAW_TO_WINDOW | PFD_SUPPORT_OPENGL;
+    *pixelformat = ChoosePixelFormat(dc, &pfd);
+    if (*pixelformat == 0) return GL_TRUE;
+  }
+  /* set the pixel format for the dc */
+  if (FALSE == SetPixelFormat(dc, *pixelformat, &pfd)) return GL_TRUE;
+  /* create rendering context */
+  rc = wglCreateContext(dc);
+  if (NULL == rc) return GL_TRUE;
+  if (FALSE == wglMakeCurrent(dc, rc)) return GL_TRUE;
+  return GL_FALSE;
+}
+
+void glewDestroyContext ()
+{
+  if (NULL != rc) wglMakeCurrent(NULL, NULL);
+  if (NULL != rc) wglDeleteContext(wglGetCurrentContext());
+  if (NULL != wnd && NULL != dc) ReleaseDC(wnd, dc);
+  if (NULL != wnd) DestroyWindow(wnd);
+  UnregisterClass("GLEW", GetModuleHandle(NULL));
+}
+
+/* ------------------------------------------------------------------------ */
+
+#elif defined(__APPLE__) && !defined(GLEW_APPLE_GLX)
+
+#include <AGL/agl.h>
+
+AGLContext ctx, octx;
+
+GLboolean glewCreateContext ()
+{
+  int attrib[] = { AGL_RGBA, AGL_NONE };
+  AGLPixelFormat pf;
+  /*int major, minor;
+  SetPortWindowPort(wnd);
+  aglGetVersion(&major, &minor);
+  fprintf(stderr, "GL %d.%d\n", major, minor);*/
+  pf = aglChoosePixelFormat(NULL, 0, attrib);
+  if (NULL == pf) return GL_TRUE;
+  ctx = aglCreateContext(pf, NULL);
+  if (NULL == ctx || AGL_NO_ERROR != aglGetError()) return GL_TRUE;
+  aglDestroyPixelFormat(pf);
+  /*aglSetDrawable(ctx, GetWindowPort(wnd));*/
+  octx = aglGetCurrentContext();
+  if (NULL == aglSetCurrentContext(ctx)) return GL_TRUE;
+  return GL_FALSE;
+}
+
+void glewDestroyContext ()
+{
+  aglSetCurrentContext(octx);
+  if (NULL != ctx) aglDestroyContext(ctx);
+}
+
+/* ------------------------------------------------------------------------ */
+
+#else /* __UNIX || (__APPLE__ && GLEW_APPLE_GLX) */
+
+Display* dpy = NULL;
+XVisualInfo* vi = NULL;
+XVisualInfo* vis = NULL;
+GLXContext ctx = NULL;
+Window wnd = 0;
+Colormap cmap = 0;
+
+GLboolean glewCreateContext (const char* display, int* visual)
+{
+  int attrib[] = { GLX_RGBA, GLX_DOUBLEBUFFER, None };
+  int erb, evb;
+  XSetWindowAttributes swa;
+  /* open display */
+  dpy = XOpenDisplay(display);
+  if (NULL == dpy) return GL_TRUE;
+  /* query for glx */
+  if (!glXQueryExtension(dpy, &erb, &evb)) return GL_TRUE;
+  /* choose visual */
+  if (*visual == -1)
+  {
+    vi = glXChooseVisual(dpy, DefaultScreen(dpy), attrib);
+    if (NULL == vi) return GL_TRUE;
+    *visual = (int)XVisualIDFromVisual(vi->visual);
+  }
+  else
+  {
+    int n_vis, i;
+    vis = XGetVisualInfo(dpy, 0, NULL, &n_vis);
+    for (i=0; i<n_vis; i++)
+    {
+      if ((int)XVisualIDFromVisual(vis[i].visual) == *visual)
+        vi = &vis[i];
+    }
+    if (vi == NULL) return GL_TRUE;
+  }
+  /* create context */
+  ctx = glXCreateContext(dpy, vi, None, True);
+  if (NULL == ctx) return GL_TRUE;
+  /* create window */
+  /*wnd = XCreateSimpleWindow(dpy, RootWindow(dpy, vi->screen), 0, 0, 1, 1, 1, 0, 0);*/
+  cmap = XCreateColormap(dpy, RootWindow(dpy, vi->screen), vi->visual, AllocNone);
+  swa.border_pixel = 0;
+  swa.colormap = cmap;
+  wnd = XCreateWindow(dpy, RootWindow(dpy, vi->screen), 
+                      0, 0, 1, 1, 0, vi->depth, InputOutput, vi->visual, 
+                      CWBorderPixel | CWColormap, &swa);
+  /* make context current */
+  if (!glXMakeCurrent(dpy, wnd, ctx)) return GL_TRUE;
+  return GL_FALSE;
+}
+
+void glewDestroyContext ()
+{
+  if (NULL != dpy && NULL != ctx) glXDestroyContext(dpy, ctx);
+  if (NULL != dpy && 0 != wnd) XDestroyWindow(dpy, wnd);
+  if (NULL != dpy && 0 != cmap) XFreeColormap(dpy, cmap);
+  if (NULL != vis)
+    XFree(vis);
+  else if (NULL != vi)
+    XFree(vi);
+  if (NULL != dpy) XCloseDisplay(dpy);
+}
+
+#endif /* __UNIX || (__APPLE__ && GLEW_APPLE_GLX) */
diff --git a/src/glew/visualinfo.c b/src/glew/visualinfo.c
new file mode 100644
index 0000000000..384313df22
--- /dev/null
+++ b/src/glew/visualinfo.c
@@ -0,0 +1,1173 @@
+/*
+** visualinfo.c
+**
+** Copyright (C) Nate Robins, 1997
+**               Michael Wimmer, 1999
+**               Milan Ikits, 2002-2008
+**
+** visualinfo is a small utility that displays all available visuals,
+** aka. pixelformats, in an OpenGL system along with renderer version
+** information. It shows a table of all the visuals that support OpenGL
+** along with their capabilities. The format of the table is similar to
+** that of glxinfo on Unix systems:
+**
+** visual ~= pixel format descriptor
+** id       = visual id (integer from 1 - max visuals)
+** tp       = type (wn: window, pb: pbuffer, wp: window & pbuffer, bm: bitmap)
+** ac	    = acceleration (ge: generic, fu: full, no: none)
+** fm	    = format (i: integer, f: float, c: color index)
+** db	    = double buffer (y = yes)
+** sw       = swap method (x: exchange, c: copy, u: undefined)
+** st	    = stereo (y = yes)
+** sz       = total # bits
+** r        = # bits of red
+** g        = # bits of green
+** b        = # bits of blue
+** a        = # bits of alpha
+** axbf     = # aux buffers
+** dpth     = # bits of depth
+** stcl     = # bits of stencil
+*/
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <GL/glew.h>
+#if defined(_WIN32)
+#include <GL/wglew.h>
+#elif defined(__APPLE__) && !defined(GLEW_APPLE_GLX)
+#include <AGL/agl.h>
+#else
+#include <GL/glxew.h>
+#endif
+
+#ifdef GLEW_MX
+GLEWContext _glewctx;
+#  define glewGetContext() (&_glewctx)
+#  ifdef _WIN32
+WGLEWContext _wglewctx;
+#    define wglewGetContext() (&_wglewctx)
+#  elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+GLXEWContext _glxewctx;
+#    define glxewGetContext() (&_glxewctx)
+#  endif
+#endif /* GLEW_MX */
+
+typedef struct GLContextStruct
+{
+#ifdef _WIN32
+  HWND wnd;
+  HDC dc;
+  HGLRC rc;
+#elif defined(__APPLE__) && !defined(GLEW_APPLE_GLX)
+  AGLContext ctx, octx;
+#else
+  Display* dpy;
+  XVisualInfo* vi;
+  GLXContext ctx;
+  Window wnd;
+  Colormap cmap;
+#endif
+} GLContext;
+
+void InitContext (GLContext* ctx);
+GLboolean CreateContext (GLContext* ctx);
+void DestroyContext (GLContext* ctx);
+void VisualInfo (GLContext* ctx);
+void PrintExtensions (const char* s);
+GLboolean ParseArgs (int argc, char** argv);
+
+int showall = 0;
+int displaystdout = 0;
+int verbose = 0;
+int drawableonly = 0;
+
+char* display = NULL;
+int visual = -1;
+
+FILE* file = 0;
+GLContext ctx;
+
+int 
+main (int argc, char** argv)
+{
+  GLenum err;
+
+  /* ---------------------------------------------------------------------- */
+  /* parse arguments */
+  if (GL_TRUE == ParseArgs(argc-1, argv+1))
+  {
+#if defined(_WIN32)
+    fprintf(stderr, "Usage: visualinfo [-a] [-s] [-h] [-pf <id>]\n");
+    fprintf(stderr, "        -a: show all visuals\n");
+    fprintf(stderr, "        -s: display to stdout instead of visualinfo.txt\n");
+    fprintf(stderr, "        -pf <id>: use given pixelformat\n");
+    fprintf(stderr, "        -h: this screen\n");
+#else
+    fprintf(stderr, "Usage: visualinfo [-h] [-display <display>] [-visual <id>]\n");
+    fprintf(stderr, "        -h: this screen\n");
+    fprintf(stderr, "        -display <display>: use given display\n");
+    fprintf(stderr, "        -visual <id>: use given visual\n");
+#endif
+    return 1;
+  }
+
+  /* ---------------------------------------------------------------------- */
+  /* create OpenGL rendering context */
+  InitContext(&ctx);
+  if (GL_TRUE == CreateContext(&ctx))
+  {
+    fprintf(stderr, "Error: CreateContext failed\n");
+    DestroyContext(&ctx);
+    return 1;
+  }
+
+  /* ---------------------------------------------------------------------- */
+  /* initialize GLEW */
+  glewExperimental = GL_TRUE;
+#ifdef GLEW_MX
+  err = glewContextInit(glewGetContext());
+#  ifdef _WIN32
+  err = err || wglewContextInit(wglewGetContext());
+#  elif !defined(__APPLE__) || defined(GLEW_APPLE_GLX)
+  err = err || glxewContextInit(glxewGetContext());
+#  endif
+#else
+  err = glewInit();
+#endif
+  if (GLEW_OK != err)
+  {
+    fprintf(stderr, "Error [main]: glewInit failed: %s\n", glewGetErrorString(err));
+    DestroyContext(&ctx);
+    return 1;
+  }
+
+  /* ---------------------------------------------------------------------- */
+  /* open file */
+#if defined(_WIN32)
+  if (!displaystdout) 
+    file = fopen("visualinfo.txt", "w");
+  if (file == NULL)
+    file = stdout;
+#else
+  file = stdout;
+#endif
+
+  /* ---------------------------------------------------------------------- */
+  /* output header information */
+  /* OpenGL extensions */
+  fprintf(file, "OpenGL vendor string: %s\n", glGetString(GL_VENDOR));
+  fprintf(file, "OpenGL renderer string: %s\n", glGetString(GL_RENDERER));
+  fprintf(file, "OpenGL version string: %s\n", glGetString(GL_VERSION));
+  fprintf(file, "OpenGL extensions (GL_): \n");
+  PrintExtensions((char*)glGetString(GL_EXTENSIONS));
+  /* GLU extensions */
+  fprintf(file, "GLU version string: %s\n", gluGetString(GLU_VERSION));
+  fprintf(file, "GLU extensions (GLU_): \n");
+  PrintExtensions((char*)gluGetString(GLU_EXTENSIONS));
+
+  /* ---------------------------------------------------------------------- */
+  /* extensions string */
+#if defined(_WIN32)
+  /* WGL extensions */
+  if (WGLEW_ARB_extensions_string || WGLEW_EXT_extensions_string)
+  {
+    fprintf(file, "WGL extensions (WGL_): \n");
+    PrintExtensions(wglGetExtensionsStringARB ? 
+                    (char*)wglGetExtensionsStringARB(ctx.dc) :
+		    (char*)wglGetExtensionsStringEXT());
+  }
+#elif defined(__APPLE__) && !defined(GLEW_APPLE_GLX)
+  
+#else
+  /* GLX extensions */
+  fprintf(file, "GLX extensions (GLX_): \n");
+  PrintExtensions(glXQueryExtensionsString(glXGetCurrentDisplay(), 
+                                           DefaultScreen(glXGetCurrentDisplay())));
+#endif
+
+  /* ---------------------------------------------------------------------- */
+  /* enumerate all the formats */
+  VisualInfo(&ctx);
+
+  /* ---------------------------------------------------------------------- */
+  /* release resources */
+  DestroyContext(&ctx);
+  if (file != stdout)
+    fclose(file);
+  return 0;
+}
+
+/* do the magic to separate all extensions with comma's, except
+   for the last one that _may_ terminate in a space. */
+void PrintExtensions (const char* s)
+{
+  char t[80];
+  int i=0;
+  char* p=0;
+
+  t[79] = '\0';
+  while (*s)
+  {
+    t[i++] = *s;
+    if(*s == ' ')
+    {
+      if (*(s+1) != '\0') {
+	t[i-1] = ',';
+	t[i] = ' ';
+	p = &t[i++];
+      }
+      else /* zoinks! last one terminated in a space! */
+      {
+	t[i-1] = '\0';
+      }
+    }
+    if(i > 80 - 5)
+    {
+      *p = t[i] = '\0';
+      fprintf(file, "    %s\n", t);
+      p++;
+      i = (int)strlen(p);
+      strcpy(t, p);
+    }
+    s++;
+  }
+  t[i] = '\0';
+  fprintf(file, "    %s.\n", t);
+}
+
+/* ---------------------------------------------------------------------- */
+
+#if defined(_WIN32)
+
+void
+VisualInfoARB (GLContext* ctx)
+{
+  int attrib[32], value[32], n_attrib, n_pbuffer=0, n_float=0;
+  int i, pf, maxpf;
+  unsigned int c;
+
+  /* to get pbuffer capable pixel formats */
+  attrib[0] = WGL_DRAW_TO_PBUFFER_ARB;
+  attrib[1] = GL_TRUE;
+  attrib[2] = 0;
+  wglChoosePixelFormatARB(ctx->dc, attrib, 0, 1, &pf, &c);
+  /* query number of pixel formats */
+  attrib[0] = WGL_NUMBER_PIXEL_FORMATS_ARB;
+  wglGetPixelFormatAttribivARB(ctx->dc, 0, 0, 1, attrib, value);
+  maxpf = value[0];
+  for (i=0; i<32; i++)
+    value[i] = 0;
+
+  attrib[0] = WGL_SUPPORT_OPENGL_ARB;
+  attrib[1] = WGL_DRAW_TO_WINDOW_ARB;
+  attrib[2] = WGL_DRAW_TO_BITMAP_ARB;
+  attrib[3] = WGL_ACCELERATION_ARB;
+  /* WGL_NO_ACCELERATION_ARB, WGL_GENERIC_ACCELERATION_ARB, WGL_FULL_ACCELERATION_ARB */
+  attrib[4] = WGL_SWAP_METHOD_ARB;
+  /* WGL_SWAP_EXCHANGE_ARB, WGL_SWAP_COPY_ARB, WGL_SWAP_UNDEFINED_ARB */
+  attrib[5] = WGL_DOUBLE_BUFFER_ARB;
+  attrib[6] = WGL_STEREO_ARB;
+  attrib[7] = WGL_PIXEL_TYPE_ARB;
+  /* WGL_TYPE_RGBA_ARB, WGL_TYPE_COLORINDEX_ARB,
+     WGL_TYPE_RGBA_FLOAT_ATI (WGL_ATI_pixel_format_float) */
+  /* Color buffer information */
+  attrib[8] = WGL_COLOR_BITS_ARB;
+  attrib[9] = WGL_RED_BITS_ARB;
+  attrib[10] = WGL_GREEN_BITS_ARB;
+  attrib[11] = WGL_BLUE_BITS_ARB;
+  attrib[12] = WGL_ALPHA_BITS_ARB;
+  /* Accumulation buffer information */
+  attrib[13] = WGL_ACCUM_BITS_ARB;
+  attrib[14] = WGL_ACCUM_RED_BITS_ARB;
+  attrib[15] = WGL_ACCUM_GREEN_BITS_ARB;
+  attrib[16] = WGL_ACCUM_BLUE_BITS_ARB;
+  attrib[17] = WGL_ACCUM_ALPHA_BITS_ARB;
+  /* Depth, stencil, and aux buffer information */
+  attrib[18] = WGL_DEPTH_BITS_ARB;
+  attrib[19] = WGL_STENCIL_BITS_ARB;
+  attrib[20] = WGL_AUX_BUFFERS_ARB;
+  /* Layer information */
+  attrib[21] = WGL_NUMBER_OVERLAYS_ARB;
+  attrib[22] = WGL_NUMBER_UNDERLAYS_ARB;
+  attrib[23] = WGL_SWAP_LAYER_BUFFERS_ARB;
+  attrib[24] = WGL_SAMPLES_ARB;
+  attrib[25] = WGL_SUPPORT_GDI_ARB;
+  n_attrib = 26;
+  if (WGLEW_ARB_pbuffer)
+  {
+    attrib[n_attrib] = WGL_DRAW_TO_PBUFFER_ARB;
+    n_pbuffer = n_attrib;
+    n_attrib++;
+  }
+  if (WGLEW_NV_float_buffer)
+  {
+    attrib[n_attrib] = WGL_FLOAT_COMPONENTS_NV;
+    n_float = n_attrib;
+    n_attrib++;
+  }
+  
+  if (!verbose)
+  {
+    /* print table header */
+    fprintf(file, " +-----+-------------------------+-----------------+----------+-----------------+----------+\n");
+    fprintf(file, " |     |          visual         |      color      | ax dp st |      accum      |   layer  |\n");
+    fprintf(file, " |  id | tp ac gd fm db sw st ms |  sz  r  g  b  a | bf th cl |  sz  r  g  b  a | ov un sw |\n");
+    fprintf(file, " +-----+-------------------------+-----------------+----------+-----------------+----------+\n");
+    /* loop through all the pixel formats */
+    for(i = 1; i <= maxpf; i++)
+    {
+      wglGetPixelFormatAttribivARB(ctx->dc, i, 0, n_attrib, attrib, value);
+      /* only describe this format if it supports OpenGL */
+      if (!value[0]) continue;
+      /* by default show only fully accelerated window or pbuffer capable visuals */
+      if (!showall
+	  && ((value[2] && !value[1])
+	  || (!WGLEW_ARB_pbuffer || !value[n_pbuffer])
+	  || (value[3] != WGL_FULL_ACCELERATION_ARB))) continue;
+      /* print out the information for this visual */
+      /* visual id */
+      fprintf(file, " |% 4d | ", i);
+      /* visual type */
+      if (value[1])
+      {
+	if (WGLEW_ARB_pbuffer && value[n_pbuffer]) fprintf(file, "wp ");
+	else fprintf(file, "wn ");
+      }
+      else
+      {
+	if (value[2]) fprintf(file, "bm ");
+	else if (WGLEW_ARB_pbuffer && value[n_pbuffer]) fprintf(file, "pb ");
+      }
+      /* acceleration */
+      fprintf(file, "%s ", value[3] == WGL_FULL_ACCELERATION_ARB ? "fu" : 
+	      value[3] == WGL_GENERIC_ACCELERATION_ARB ? "ge" :
+	      value[3] == WGL_NO_ACCELERATION_ARB ? "no" : ". ");
+      /* gdi support */
+      fprintf(file, " %c ", value[25] ? 'y' : '.');
+      /* format */
+      if (WGLEW_NV_float_buffer && value[n_float]) fprintf(file, " f ");
+      else if (WGLEW_ATI_pixel_format_float && value[7] == WGL_TYPE_RGBA_FLOAT_ATI) fprintf(file, " f ");
+      else if (value[7] == WGL_TYPE_RGBA_ARB) fprintf(file, " i ");
+      else if (value[7] == WGL_TYPE_COLORINDEX_ARB) fprintf(file, " c ");
+      /* double buffer */
+      fprintf(file, " %c ", value[5] ? 'y' : '.');
+      /* swap method */
+      if (value[4] == WGL_SWAP_EXCHANGE_ARB) fprintf(file, " x ");
+      else if (value[4] == WGL_SWAP_COPY_ARB) fprintf(file, " c ");
+      else if (value[4] == WGL_SWAP_UNDEFINED_ARB) fprintf(file, " . ");
+      else fprintf(file, " . ");
+      /* stereo */
+      fprintf(file, " %c ", value[6] ? 'y' : '.');
+      /* multisample */
+      if (value[24] > 0)
+	fprintf(file, "%2d | ", value[24]);
+      else
+	fprintf(file, " . | ");
+      /* color size */
+      if (value[8]) fprintf(file, "%3d ", value[8]);
+      else fprintf(file, "  . ");
+      /* red */
+      if (value[9]) fprintf(file, "%2d ", value[9]); 
+      else fprintf(file, " . ");
+      /* green */
+      if (value[10]) fprintf(file, "%2d ", value[10]); 
+      else fprintf(file, " . ");
+      /* blue */
+      if (value[11]) fprintf(file, "%2d ", value[11]);
+      else fprintf(file, " . ");
+      /* alpha */
+      if (value[12]) fprintf(file, "%2d | ", value[12]); 
+      else fprintf(file, " . | ");
+      /* aux buffers */
+      if (value[20]) fprintf(file, "%2d ", value[20]);
+      else fprintf(file, " . ");
+      /* depth */
+      if (value[18]) fprintf(file, "%2d ", value[18]);
+      else fprintf(file, " . ");
+      /* stencil */
+      if (value[19]) fprintf(file, "%2d | ", value[19]);
+      else fprintf(file, " . | ");
+      /* accum size */
+      if (value[13]) fprintf(file, "%3d ", value[13]);
+      else fprintf(file, "  . ");
+      /* accum red */
+      if (value[14]) fprintf(file, "%2d ", value[14]);
+      else fprintf(file, " . ");
+      /* accum green */
+      if (value[15]) fprintf(file, "%2d ", value[15]);
+      else fprintf(file, " . ");
+      /* accum blue */
+      if (value[16]) fprintf(file, "%2d ", value[16]);
+      else fprintf(file, " . ");
+      /* accum alpha */
+      if (value[17]) fprintf(file, "%2d | ", value[17]);
+      else fprintf(file, " . | ");
+      /* overlay */
+      if (value[21]) fprintf(file, "%2d ", value[21]);
+      else fprintf(file, " . ");
+      /* underlay */
+      if (value[22]) fprintf(file, "%2d ", value[22]);
+      else fprintf(file, " . ");
+      /* layer swap */
+      if (value[23]) fprintf(file, "y ");
+      else fprintf(file, " . ");
+      fprintf(file, "|\n");
+    }
+    /* print table footer */
+    fprintf(file, " +-----+-------------------------+-----------------+----------+-----------------+----------+\n");
+    fprintf(file, " |     |          visual         |      color      | ax dp st |      accum      |   layer  |\n");
+    fprintf(file, " |  id | tp ac gd fm db sw st ms |  sz  r  g  b  a | bf th cl |  sz  r  g  b  a | ov un sw |\n");
+    fprintf(file, " +-----+-------------------------+-----------------+----------+-----------------+----------+\n");
+  }
+  else /* verbose */
+  {
+#if 0
+    fprintf(file, "\n");
+    /* loop through all the pixel formats */
+    for(i = 1; i <= maxpf; i++)
+    {	    
+      DescribePixelFormat(ctx->dc, i, sizeof(PIXELFORMATDESCRIPTOR), &pfd);
+      /* only describe this format if it supports OpenGL */
+      if(!(pfd.dwFlags & PFD_SUPPORT_OPENGL)
+	 || (drawableonly && !(pfd.dwFlags & PFD_DRAW_TO_WINDOW))) continue;
+      fprintf(file, "Visual ID: %2d  depth=%d  class=%s\n", i, pfd.cDepthBits, 
+	     pfd.cColorBits <= 8 ? "PseudoColor" : "TrueColor");
+      fprintf(file, "    bufferSize=%d level=%d renderType=%s doubleBuffer=%d stereo=%d\n", pfd.cColorBits, pfd.bReserved, pfd.iPixelType == PFD_TYPE_RGBA ? "rgba" : "ci", pfd.dwFlags & PFD_DOUBLEBUFFER, pfd.dwFlags & PFD_STEREO);
+      fprintf(file, "    generic=%d generic accelerated=%d\n", (pfd.dwFlags & PFD_GENERIC_FORMAT) == PFD_GENERIC_FORMAT, (pfd.dwFlags & PFD_GENERIC_ACCELERATED) == PFD_GENERIC_ACCELERATED);
+      fprintf(file, "    rgba: redSize=%d greenSize=%d blueSize=%d alphaSize=%d\n", pfd.cRedBits, pfd.cGreenBits, pfd.cBlueBits, pfd.cAlphaBits);
+      fprintf(file, "    auxBuffers=%d depthSize=%d stencilSize=%d\n", pfd.cAuxBuffers, pfd.cDepthBits, pfd.cStencilBits);
+      fprintf(file, "    accum: redSize=%d greenSize=%d blueSize=%d alphaSize=%d\n", pfd.cAccumRedBits, pfd.cAccumGreenBits, pfd.cAccumBlueBits, pfd.cAccumAlphaBits);
+      fprintf(file, "    multiSample=%d multisampleBuffers=%d\n", 0, 0);
+      fprintf(file, "    Opaque.\n");
+    }
+#endif
+  }
+}
+
+void
+VisualInfoGDI (GLContext* ctx)
+{
+  int i, maxpf;
+  PIXELFORMATDESCRIPTOR pfd;
+
+  /* calling DescribePixelFormat() with NULL pfd (!!!) return maximum
+     number of pixel formats */
+  maxpf = DescribePixelFormat(ctx->dc, 1, 0, NULL);
+
+  if (!verbose)
+  {
+    fprintf(file, "-----------------------------------------------------------------------------\n");
+    fprintf(file, "   visual   x  bf  lv rg d st ge ge  r  g  b a  ax dp st   accum buffs    ms \n");
+    fprintf(file, " id  dep tp sp sz  l  ci b ro ne ac sz sz sz sz bf th cl  sz  r  g  b  a ns b\n");
+    fprintf(file, "-----------------------------------------------------------------------------\n");
+
+    /* loop through all the pixel formats */
+    for(i = 1; i <= maxpf; i++)
+    {
+      DescribePixelFormat(ctx->dc, i, sizeof(PIXELFORMATDESCRIPTOR), &pfd);
+      /* only describe this format if it supports OpenGL */
+      if(!(pfd.dwFlags & PFD_SUPPORT_OPENGL)
+	 || (drawableonly && (pfd.dwFlags & PFD_DRAW_TO_BITMAP))) continue;
+      /* other criteria could be tested here for actual pixel format
+	 choosing in an application:
+	   
+	 for (...each pixel format...) {
+	 if (pfd.dwFlags & PFD_SUPPORT_OPENGL &&
+	 pfd.dwFlags & PFD_DOUBLEBUFFER &&
+	 pfd.cDepthBits >= 24 &&
+	 pfd.cColorBits >= 24)
+	 {
+	 goto found;
+	 }
+	 }
+	 ... not found so exit ...
+	 found:
+	 ... found so use it ...
+      */
+      /* print out the information for this pixel format */
+      fprintf(file, "0x%02x ", i);
+      fprintf(file, "%3d ", pfd.cColorBits);
+      if(pfd.dwFlags & PFD_DRAW_TO_WINDOW) fprintf(file, "wn ");
+      else if(pfd.dwFlags & PFD_DRAW_TO_BITMAP) fprintf(file, "bm ");
+      else fprintf(file, "pb ");
+      /* should find transparent pixel from LAYERPLANEDESCRIPTOR */
+      fprintf(file, " . "); 
+      fprintf(file, "%3d ", pfd.cColorBits);
+      /* bReserved field indicates number of over/underlays */
+      if(pfd.bReserved) fprintf(file, " %d ", pfd.bReserved);
+      else fprintf(file, " . "); 
+      fprintf(file, " %c ", pfd.iPixelType == PFD_TYPE_RGBA ? 'r' : 'c');
+      fprintf(file, "%c ", pfd.dwFlags & PFD_DOUBLEBUFFER ? 'y' : '.');
+      fprintf(file, " %c ", pfd.dwFlags & PFD_STEREO ? 'y' : '.');
+      /* added: */
+      fprintf(file, " %c ", pfd.dwFlags & PFD_GENERIC_FORMAT ? 'y' : '.');
+      fprintf(file, " %c ", pfd.dwFlags & PFD_GENERIC_ACCELERATED ? 'y' : '.');
+      if(pfd.cRedBits && pfd.iPixelType == PFD_TYPE_RGBA) 
+	fprintf(file, "%2d ", pfd.cRedBits);
+      else fprintf(file, " . ");
+      if(pfd.cGreenBits && pfd.iPixelType == PFD_TYPE_RGBA) 
+	fprintf(file, "%2d ", pfd.cGreenBits);
+      else fprintf(file, " . ");
+      if(pfd.cBlueBits && pfd.iPixelType == PFD_TYPE_RGBA) 
+	fprintf(file, "%2d ", pfd.cBlueBits);
+      else fprintf(file, " . ");
+      if(pfd.cAlphaBits && pfd.iPixelType == PFD_TYPE_RGBA) 
+	fprintf(file, "%2d ", pfd.cAlphaBits);
+      else fprintf(file, " . ");
+      if(pfd.cAuxBuffers)     fprintf(file, "%2d ", pfd.cAuxBuffers);
+      else fprintf(file, " . ");
+      if(pfd.cDepthBits)      fprintf(file, "%2d ", pfd.cDepthBits);
+      else fprintf(file, " . ");
+      if(pfd.cStencilBits)    fprintf(file, "%2d ", pfd.cStencilBits);
+      else fprintf(file, " . ");
+      if(pfd.cAccumBits)   fprintf(file, "%3d ", pfd.cAccumBits);
+      else fprintf(file, "  . ");
+      if(pfd.cAccumRedBits)   fprintf(file, "%2d ", pfd.cAccumRedBits);
+      else fprintf(file, " . ");
+      if(pfd.cAccumGreenBits) fprintf(file, "%2d ", pfd.cAccumGreenBits);
+      else fprintf(file, " . ");
+      if(pfd.cAccumBlueBits)  fprintf(file, "%2d ", pfd.cAccumBlueBits);
+      else fprintf(file, " . ");
+      if(pfd.cAccumAlphaBits) fprintf(file, "%2d ", pfd.cAccumAlphaBits);
+      else fprintf(file, " . ");
+      /* no multisample in win32 */
+      fprintf(file, " . .\n");
+    }
+    /* print table footer */
+    fprintf(file, "-----------------------------------------------------------------------------\n");
+    fprintf(file, "   visual   x  bf  lv rg d st ge ge  r  g  b a  ax dp st   accum buffs    ms \n");
+    fprintf(file, " id  dep tp sp sz  l  ci b ro ne ac sz sz sz sz bf th cl  sz  r  g  b  a ns b\n");
+    fprintf(file, "-----------------------------------------------------------------------------\n");
+  }
+  else /* verbose */
+  {
+    fprintf(file, "\n");
+    /* loop through all the pixel formats */
+    for(i = 1; i <= maxpf; i++)
+    {	    
+      DescribePixelFormat(ctx->dc, i, sizeof(PIXELFORMATDESCRIPTOR), &pfd);
+      /* only describe this format if it supports OpenGL */
+      if(!(pfd.dwFlags & PFD_SUPPORT_OPENGL)
+	 || (drawableonly && !(pfd.dwFlags & PFD_DRAW_TO_WINDOW))) continue;
+      fprintf(file, "Visual ID: %2d  depth=%d  class=%s\n", i, pfd.cDepthBits, 
+	     pfd.cColorBits <= 8 ? "PseudoColor" : "TrueColor");
+      fprintf(file, "    bufferSize=%d level=%d renderType=%s doubleBuffer=%ld stereo=%ld\n", pfd.cColorBits, pfd.bReserved, pfd.iPixelType == PFD_TYPE_RGBA ? "rgba" : "ci", pfd.dwFlags & PFD_DOUBLEBUFFER, pfd.dwFlags & PFD_STEREO);
+      fprintf(file, "    generic=%d generic accelerated=%d\n", (pfd.dwFlags & PFD_GENERIC_FORMAT) == PFD_GENERIC_FORMAT, (pfd.dwFlags & PFD_GENERIC_ACCELERATED) == PFD_GENERIC_ACCELERATED);
+      fprintf(file, "    rgba: redSize=%d greenSize=%d blueSize=%d alphaSize=%d\n", pfd.cRedBits, pfd.cGreenBits, pfd.cBlueBits, pfd.cAlphaBits);
+      fprintf(file, "    auxBuffers=%d depthSize=%d stencilSize=%d\n", pfd.cAuxBuffers, pfd.cDepthBits, pfd.cStencilBits);
+      fprintf(file, "    accum: redSize=%d greenSize=%d blueSize=%d alphaSize=%d\n", pfd.cAccumRedBits, pfd.cAccumGreenBits, pfd.cAccumBlueBits, pfd.cAccumAlphaBits);
+      fprintf(file, "    multiSample=%d multisampleBuffers=%d\n", 0, 0);
+      fprintf(file, "    Opaque.\n");
+    }
+  }
+}
+
+void
+VisualInfo (GLContext* ctx)
+{
+  if (WGLEW_ARB_pixel_format)
+    VisualInfoARB(ctx);
+  else
+    VisualInfoGDI(ctx);
+}
+
+/* ---------------------------------------------------------------------- */
+
+#elif defined(__APPLE__) && !defined(GLEW_APPLE_GLX)
+
+void
+VisualInfo (GLContext* ctx)
+{
+/*
+  int attrib[] = { AGL_RGBA, AGL_NONE };
+  AGLPixelFormat pf;
+  GLint value;
+  pf = aglChoosePixelFormat(NULL, 0, attrib);
+  while (pf != NULL)
+  {
+    aglDescribePixelFormat(pf, GL_RGBA, &value);
+    fprintf(stderr, "%d\n", value);
+    pf = aglNextPixelFormat(pf);
+  }
+*/
+}
+
+#else /* GLX */
+
+void
+VisualInfo (GLContext* ctx)
+{
+  int n_fbc;
+  GLXFBConfig* fbc;
+  int value, ret, i;
+
+  fbc = glXGetFBConfigs(ctx->dpy, DefaultScreen(ctx->dpy), &n_fbc);
+
+  if (fbc)
+  {
+    if (!verbose)
+    {
+      /* print table header */
+      fprintf(file, " +-----+-------------------------+-----------------+----------+-------------+-------+------+\n");
+      fprintf(file, " |     |        visual           |      color      | ax dp st |    accum    |   ms  |  cav |\n");
+      fprintf(file, " |  id | tp xr cl fm db st lv xp |  sz  r  g  b  a | bf th cl | r  g  b  a  | ns  b |  eat |\n");
+      fprintf(file, " +-----+-------------------------+-----------------+----------+-------------+-------+------+\n");
+      /* loop through all the fbcs */
+      for (i=0; i<n_fbc; i++)
+      {
+        /* print out the information for this fbc */
+        /* visual id */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_FBCONFIG_ID, &value);
+        if (ret != Success)
+        {
+          fprintf(file, "|  ?  |");
+        }
+        else
+        {
+          fprintf(file, " |% 4d | ", value);
+        }
+        /* visual type */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_DRAWABLE_TYPE, &value);
+        if (ret != Success)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (value & GLX_WINDOW_BIT)
+          {
+            if (value & GLX_PBUFFER_BIT)
+            {
+              fprintf(file, "wp ");
+            }
+            else
+            {
+              fprintf(file, "wn ");
+            }
+          }
+          else
+          {
+            if (value & GLX_PBUFFER_BIT)
+            {
+              fprintf(file, "pb ");
+            }
+            else if (value & GLX_PIXMAP_BIT)
+            {
+              fprintf(file, "pm ");
+            }
+            else
+            {
+              fprintf(file, " ? ");
+            }
+          }
+        }
+        /* x renderable */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_X_RENDERABLE, &value);
+        if (ret != Success)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          fprintf(file, value ? " y " : " n ");
+        }
+        /* class */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_X_VISUAL_TYPE, &value);
+        if (ret != Success)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (GLX_TRUE_COLOR == value)
+            fprintf(file, "tc ");
+          else if (GLX_DIRECT_COLOR == value)
+            fprintf(file, "dc ");
+          else if (GLX_PSEUDO_COLOR == value)
+            fprintf(file, "pc ");
+          else if (GLX_STATIC_COLOR == value)
+            fprintf(file, "sc ");
+          else if (GLX_GRAY_SCALE == value)
+            fprintf(file, "gs ");
+          else if (GLX_STATIC_GRAY == value)
+            fprintf(file, "sg ");
+          else if (GLX_X_VISUAL_TYPE == value)
+            fprintf(file, " . ");
+          else
+            fprintf(file, " ? ");
+        }
+        /* format */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_RENDER_TYPE, &value);
+        if (ret != Success)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (GLXEW_NV_float_buffer)
+          {
+            int ret2, value2;
+            ret2 = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_FLOAT_COMPONENTS_NV, &value2);
+            if (Success == ret2 && GL_TRUE == value2)
+            {
+              fprintf(file, " f ");
+            }
+            else if (value & GLX_RGBA_BIT)
+              fprintf(file, " i ");
+            else if (value & GLX_COLOR_INDEX_BIT)
+              fprintf(file, " c ");
+            else
+              fprintf(file, " ? ");
+          }
+          else
+          {
+            if (value & GLX_RGBA_FLOAT_ATI_BIT)
+              fprintf(file, " f ");
+            else if (value & GLX_RGBA_BIT)
+              fprintf(file, " i ");
+            else if (value & GLX_COLOR_INDEX_BIT)
+              fprintf(file, " c ");
+            else
+              fprintf(file, " ? ");
+          }
+        }
+        /* double buffer */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_DOUBLEBUFFER, &value);
+        fprintf(file, " %c ", Success != ret ? '?' : (value ? 'y' : '.'));
+        /* stereo */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_STEREO, &value);
+        fprintf(file, " %c ", Success != ret ? '?' : (value ? 'y' : '.'));
+        /* level */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_LEVEL, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          fprintf(file, "%2d ", value);
+        }
+        /* transparency */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_TRANSPARENT_TYPE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? | ");
+        }
+        else
+        {
+          if (GLX_TRANSPARENT_RGB == value)
+            fprintf(file, " r | ");
+          else if (GLX_TRANSPARENT_INDEX == value)
+            fprintf(file, " i | ");
+          else if (GLX_NONE == value)
+            fprintf(file, " . | ");
+          else
+            fprintf(file, " ? | ");
+        }
+        /* color size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_BUFFER_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, "  ? ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%3d ", value);
+          else
+            fprintf(file, "  . ");
+        }
+        /* red size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_RED_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d ", value);
+          else
+            fprintf(file, " . ");
+        }
+        /* green size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_GREEN_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d ", value);
+          else
+            fprintf(file, " . ");
+        }
+        /* blue size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_BLUE_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d ", value);
+          else
+            fprintf(file, " . ");
+        }
+        /* alpha size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_ALPHA_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? | ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d | ", value);
+          else
+            fprintf(file, " . | ");
+        }
+        /* aux buffers */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_AUX_BUFFERS, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d ", value);
+          else
+            fprintf(file, " . ");
+        }
+        /* depth size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_DEPTH_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d ", value);
+          else
+            fprintf(file, " . ");
+        }
+        /* stencil size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_STENCIL_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? | ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d | ", value);
+          else
+            fprintf(file, " . | ");
+        }
+        /* accum red size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_ACCUM_RED_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d ", value);
+          else
+            fprintf(file, " . ");
+        }
+        /* accum green size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_ACCUM_GREEN_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d ", value);
+          else
+            fprintf(file, " . ");
+        }
+        /* accum blue size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_ACCUM_BLUE_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d ", value);
+          else
+            fprintf(file, " . ");
+        }
+        /* accum alpha size */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_ACCUM_ALPHA_SIZE, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? | ");
+        }
+        else
+        {
+          if (value)
+            fprintf(file, "%2d | ", value);
+          else
+            fprintf(file, " . | ");
+        }
+        /* multisample */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_SAMPLES, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? ");
+        }
+        else
+        {
+          fprintf(file, "%2d ", value);
+        }
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_SAMPLE_BUFFERS, &value);
+        if (Success != ret)
+        {
+          fprintf(file, " ? | ");
+        }
+        else
+        {
+          fprintf(file, "%2d | ", value);
+        }
+        /* caveat */
+        ret = glXGetFBConfigAttrib(ctx->dpy, fbc[i], GLX_CONFIG_CAVEAT, &value);
+        if (Success != ret)
+        {
+          fprintf(file, "???? |");
+        }
+        else
+        {
+          if (GLX_NONE == value)
+            fprintf(file, "none |\n");
+          else if (GLX_SLOW_CONFIG == value)
+            fprintf(file, "slow |\n");
+          else if (GLX_NON_CONFORMANT_CONFIG == value)
+            fprintf(file, "ncft |\n");
+          else
+            fprintf(file, "???? |\n");
+        }
+      }
+      /* print table footer */
+      fprintf(file, " +-----+-------------------------+-----------------+----------+-------------+-------+------+\n");
+      fprintf(file, " |  id | tp xr cl fm db st lv xp |  sz  r  g  b  a | bf th cl | r  g  b  a  | ns  b |  eat |\n");
+      fprintf(file, " |     |        visual           |      color      | ax dp st |    accum    |   ms  |  cav |\n");
+      fprintf(file, " +-----+-------------------------+-----------------+----------+-------------+-------+------+\n");
+    }
+  }
+}
+
+#endif
+
+/* ------------------------------------------------------------------------ */
+
+#if defined(_WIN32)
+
+void InitContext (GLContext* ctx)
+{
+  ctx->wnd = NULL;
+  ctx->dc = NULL;
+  ctx->rc = NULL;
+}
+
+GLboolean CreateContext (GLContext* ctx)
+{
+  WNDCLASS wc;
+  PIXELFORMATDESCRIPTOR pfd;
+  /* check for input */
+  if (NULL == ctx) return GL_TRUE;
+  /* register window class */
+  ZeroMemory(&wc, sizeof(WNDCLASS));
+  wc.hInstance = GetModuleHandle(NULL);
+  wc.lpfnWndProc = DefWindowProc;
+  wc.lpszClassName = "GLEW";
+  if (0 == RegisterClass(&wc)) return GL_TRUE;
+  /* create window */
+  ctx->wnd = CreateWindow("GLEW", "GLEW", 0, CW_USEDEFAULT, CW_USEDEFAULT, 
+                          CW_USEDEFAULT, CW_USEDEFAULT, NULL, NULL, 
+                          GetModuleHandle(NULL), NULL);
+  if (NULL == ctx->wnd) return GL_TRUE;
+  /* get the device context */
+  ctx->dc = GetDC(ctx->wnd);
+  if (NULL == ctx->dc) return GL_TRUE;
+  /* find pixel format */
+  ZeroMemory(&pfd, sizeof(PIXELFORMATDESCRIPTOR));
+  if (visual == -1) /* find default */
+  {
+    pfd.nSize = sizeof(PIXELFORMATDESCRIPTOR);
+    pfd.nVersion = 1;
+    pfd.dwFlags = PFD_DRAW_TO_WINDOW | PFD_SUPPORT_OPENGL;
+    visual = ChoosePixelFormat(ctx->dc, &pfd);
+    if (0 == visual) return GL_TRUE;
+  }
+  /* set the pixel format for the dc */
+  if (FALSE == SetPixelFormat(ctx->dc, visual, &pfd)) return GL_TRUE;
+  /* create rendering context */
+  ctx->rc = wglCreateContext(ctx->dc);
+  if (NULL == ctx->rc) return GL_TRUE;
+  if (FALSE == wglMakeCurrent(ctx->dc, ctx->rc)) return GL_TRUE;
+  return GL_FALSE;
+}
+
+void DestroyContext (GLContext* ctx)
+{
+  if (NULL == ctx) return;
+  if (NULL != ctx->rc) wglMakeCurrent(NULL, NULL);
+  if (NULL != ctx->rc) wglDeleteContext(wglGetCurrentContext());
+  if (NULL != ctx->wnd && NULL != ctx->dc) ReleaseDC(ctx->wnd, ctx->dc);
+  if (NULL != ctx->wnd) DestroyWindow(ctx->wnd);
+  UnregisterClass("GLEW", GetModuleHandle(NULL));
+}
+
+/* ------------------------------------------------------------------------ */
+
+#elif defined(__APPLE__) && !defined(GLEW_APPLE_GLX)
+
+void InitContext (GLContext* ctx)
+{
+  ctx->ctx = NULL;
+  ctx->octx = NULL;
+}
+
+GLboolean CreateContext (GLContext* ctx)
+{
+  int attrib[] = { AGL_RGBA, AGL_NONE };
+  AGLPixelFormat pf;
+  /* check input */
+  if (NULL == ctx) return GL_TRUE;
+  /*int major, minor;
+  SetPortWindowPort(wnd);
+  aglGetVersion(&major, &minor);
+  fprintf(stderr, "GL %d.%d\n", major, minor);*/
+  pf = aglChoosePixelFormat(NULL, 0, attrib);
+  if (NULL == pf) return GL_TRUE;
+  ctx->ctx = aglCreateContext(pf, NULL);
+  if (NULL == ctx->ctx || AGL_NO_ERROR != aglGetError()) return GL_TRUE;
+  aglDestroyPixelFormat(pf);
+  /*aglSetDrawable(ctx, GetWindowPort(wnd));*/
+  ctx->octx = aglGetCurrentContext();
+  if (NULL == aglSetCurrentContext(ctx->ctx)) return GL_TRUE;
+  return GL_FALSE;
+}
+
+void DestroyContext (GLContext* ctx)
+{
+  if (NULL == ctx) return;
+  aglSetCurrentContext(ctx->octx);
+  if (NULL != ctx->ctx) aglDestroyContext(ctx->ctx);
+}
+
+/* ------------------------------------------------------------------------ */
+
+#else /* __UNIX || (__APPLE__ && GLEW_APPLE_GLX) */
+
+void InitContext (GLContext* ctx)
+{
+  ctx->dpy = NULL;
+  ctx->vi = NULL;
+  ctx->ctx = NULL;
+  ctx->wnd = 0;
+  ctx->cmap = 0;
+}
+
+GLboolean CreateContext (GLContext* ctx)
+{
+  int attrib[] = { GLX_RGBA, GLX_DOUBLEBUFFER, None };
+  int erb, evb;
+  XSetWindowAttributes swa;
+  /* check input */
+  if (NULL == ctx) return GL_TRUE;
+  /* open display */
+  ctx->dpy = XOpenDisplay(display);
+  if (NULL == ctx->dpy) return GL_TRUE;
+  /* query for glx */
+  if (!glXQueryExtension(ctx->dpy, &erb, &evb)) return GL_TRUE;
+  /* choose visual */
+  ctx->vi = glXChooseVisual(ctx->dpy, DefaultScreen(ctx->dpy), attrib);
+  if (NULL == ctx->vi) return GL_TRUE;
+  /* create context */
+  ctx->ctx = glXCreateContext(ctx->dpy, ctx->vi, None, True);
+  if (NULL == ctx->ctx) return GL_TRUE;
+  /* create window */
+  /*wnd = XCreateSimpleWindow(dpy, RootWindow(dpy, vi->screen), 0, 0, 1, 1, 1, 0, 0);*/
+  ctx->cmap = XCreateColormap(ctx->dpy, RootWindow(ctx->dpy, ctx->vi->screen),
+                              ctx->vi->visual, AllocNone);
+  swa.border_pixel = 0;
+  swa.colormap = ctx->cmap;
+  ctx->wnd = XCreateWindow(ctx->dpy, RootWindow(ctx->dpy, ctx->vi->screen), 
+                           0, 0, 1, 1, 0, ctx->vi->depth, InputOutput, ctx->vi->visual, 
+                           CWBorderPixel | CWColormap, &swa);
+  /* make context current */
+  if (!glXMakeCurrent(ctx->dpy, ctx->wnd, ctx->ctx)) return GL_TRUE;
+  return GL_FALSE;
+}
+
+void DestroyContext (GLContext* ctx)
+{
+  if (NULL != ctx->dpy && NULL != ctx->ctx) glXDestroyContext(ctx->dpy, ctx->ctx);
+  if (NULL != ctx->dpy && 0 != ctx->wnd) XDestroyWindow(ctx->dpy, ctx->wnd);
+  if (NULL != ctx->dpy && 0 != ctx->cmap) XFreeColormap(ctx->dpy, ctx->cmap);
+  if (NULL != ctx->vi) XFree(ctx->vi);
+  if (NULL != ctx->dpy) XCloseDisplay(ctx->dpy);
+}
+
+#endif /* __UNIX || (__APPLE__ && GLEW_APPLE_GLX) */
+
+GLboolean ParseArgs (int argc, char** argv)
+{
+  int p = 0;
+  while (p < argc)
+  {
+#if defined(_WIN32)
+    if (!strcmp(argv[p], "-pf") || !strcmp(argv[p], "-pixelformat"))
+    {
+      if (++p >= argc) return GL_TRUE;
+      display = NULL;
+      visual = strtol(argv[p], NULL, 0);
+    }
+    else if (!strcmp(argv[p], "-a"))
+    {
+      showall = 1;
+    }
+    else if (!strcmp(argv[p], "-s"))
+    {
+      displaystdout = 1;
+    }
+    else if (!strcmp(argv[p], "-h"))
+    {
+      return GL_TRUE;
+    }
+    else
+      return GL_TRUE;
+#else
+    if (!strcmp(argv[p], "-display"))
+    {
+      if (++p >= argc) return GL_TRUE;
+      display = argv[p];
+    }
+    else if (!strcmp(argv[p], "-visual"))
+    {
+      if (++p >= argc) return GL_TRUE;
+      visual = (int)strtol(argv[p], NULL, 0);
+    }
+    else if (!strcmp(argv[p], "-h"))
+    {
+      return GL_TRUE;
+    }
+    else
+      return GL_TRUE;
+#endif
+    p++;
+  }
+  return GL_FALSE;
+}
diff --git a/src/glu/.gitignore b/src/glu/.gitignore
new file mode 100644
index 0000000000..279ea7d434
--- /dev/null
+++ b/src/glu/.gitignore
@@ -0,0 +1 @@
+exptmp
diff --git a/src/glu/Makefile b/src/glu/Makefile
index b025a90b67..e519dfeec4 100644
--- a/src/glu/Makefile
+++ b/src/glu/Makefile
@@ -18,7 +18,11 @@ pcedit = sed \
 	-e 's,@INSTALL_DIR@,$(INSTALL_DIR),' \
 	-e 's,@INSTALL_LIB_DIR@,$(INSTALL_LIB_DIR),' \
 	-e 's,@INSTALL_INC_DIR@,$(INSTALL_INC_DIR),' \
-	-e 's,@VERSION@,$(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY),'
+	-e 's,@VERSION@,$(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY),' \
+	-e 's,@GLU_PC_REQ@,$(GLU_PC_REQ),' \
+	-e 's,@GLU_PC_REQ_PRIV@,$(GLU_PC_REQ_PRIV),' \
+	-e 's,@GLU_PC_LIB_PRIV@,$(GLU_PC_LIB_PRIV),' \
+	-e 's,@GLU_PC_CFLAGS@,$(GLU_PC_CFLAGS),'
 glu.pc: glu.pc.in
 	$(pcedit) $< > $@
 
diff --git a/src/glu/descrip.mms b/src/glu/descrip.mms
deleted file mode 100644
index 6d5cd858da..0000000000
--- a/src/glu/descrip.mms
+++ /dev/null
@@ -1,9 +0,0 @@
-# Makefile for Mesa for VMS
-# contributed by Jouk Jansen  joukj@hrem.stm.tudelft.nl
-
-all :
-# PIPE is avalailable on VMS7.0 and higher. For lower versions split the
-#command in two conditional command.   JJ
-	if f$search("SYS$SYSTEM:CXX$COMPILER.EXE") .nes. "" then pipe set default [.sgi] ; $(MMS)$(MMSQUALIFIERS)
-	if f$search("SYS$SYSTEM:CXX$COMPILER.EXE") .eqs. "" then pipe set default [.mesa] ; $(MMS)$(MMSQUALIFIERS)
-	set default [-]
diff --git a/src/glu/glu.pc.in b/src/glu/glu.pc.in
index 8606b9b222..bc2517e90e 100644
--- a/src/glu/glu.pc.in
+++ b/src/glu/glu.pc.in
@@ -5,7 +5,9 @@ includedir=@INSTALL_INC_DIR@
 
 Name: glu
 Description: Mesa OpenGL Utility library
-Requires: gl
+Requires: @GLU_PC_REQ@
+Requires.private: @GLU_PC_REQ_PRIV@
 Version: @VERSION@
 Libs: -L${libdir} -lGLU
-Cflags: -I${includedir}
+Libs.private: @GLU_PC_LIB_PRIV@
+Cflags: -I${includedir} @GLU_PC_CFLAGS@
diff --git a/src/glu/mesa/Makefile.DJ b/src/glu/mesa/Makefile.DJ
deleted file mode 100644
index 92bcdaae94..0000000000
--- a/src/glu/mesa/Makefile.DJ
+++ /dev/null
@@ -1,100 +0,0 @@
-# Mesa 3-D graphics library
-# Version:  4.0
-# 
-# Copyright (C) 1999  Brian Paul   All Rights Reserved.
-# 
-# Permission is hereby granted, free of charge, to any person obtaining a
-# copy of this software and associated documentation files (the "Software"),
-# to deal in the Software without restriction, including without limitation
-# the rights to use, copy, modify, merge, publish, distribute, sublicense,
-# and/or sell copies of the Software, and to permit persons to whom the
-# Software is furnished to do so, subject to the following conditions:
-# 
-# The above copyright notice and this permission notice shall be included
-# in all copies or substantial portions of the Software.
-# 
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
-# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
-# BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
-# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
-# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-# DOS/DJGPP glu makefile v1.5 for Mesa
-#
-#  Copyright (C) 2002 - Daniel Borca
-#  Email : dborca@users.sourceforge.net
-#  Web   : http://www.geocities.com/dborca
-
-
-#
-#  Available options:
-#
-#    Environment variables:
-#	CFLAGS
-#
-#    Targets:
-#	all:		build GLU
-#	clean:		remove object files
-#
-
-
-
-.PHONY: all clean
-
-TOP = ../../..
-LIBDIR = $(TOP)/lib
-GLU_LIB = libglu.a
-GLU_DXE = glu.dxe
-GLU_IMP = libiglu.a
-
-export LD_LIBRARY_PATH := $(LD_LIBRARY_PATH);$(LIBDIR);$(GLIDE)/lib
-
-CC = gcc
-CFLAGS += -I$(TOP)/include
-
-AR = ar
-ARFLAGS = crus
-
-HAVEDXE3 = $(wildcard $(DJDIR)/bin/dxe3gen.exe)
-
-ifeq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),)
-UNLINK = del $(subst /,\,$(1))
-else
-UNLINK = $(RM) $(1)
-endif
-
-CORE_SOURCES = \
-	glu.c \
-	mipmap.c \
-	nurbs.c \
-	nurbscrv.c \
-	nurbssrf.c \
-	nurbsutl.c \
-	polytest.c \
-	project.c \
-	quadric.c \
-	tess.c \
-	tesselat.c
-
-SOURCES = $(CORE_SOURCES)
-
-OBJECTS = $(SOURCES:.c=.o)
-
-.c.o:
-	$(CC) -o $@ $(CFLAGS) -c $<
-
-all: $(LIBDIR)/$(GLU_LIB) $(LIBDIR)/$(GLU_DXE) $(LIBDIR)/$(GLU_IMP)
-
-$(LIBDIR)/$(GLU_LIB): $(OBJECTS)
-	$(AR) $(ARFLAGS) $@ $^
-
-$(LIBDIR)/$(GLU_DXE) $(LIBDIR)/$(GLU_IMP): $(OBJECTS)
-ifeq ($(HAVEDXE3),)
-	$(warning Missing DXE3 package... Skipping $(GLU_DXE))
-else
-	-dxe3gen -o $(LIBDIR)/$(GLU_DXE) -Y $(LIBDIR)/$(GLU_IMP) -D "MesaGLU DJGPP" -E _glu -P gl.dxe -U $^
-endif
-
-clean:
-	-$(call UNLINK,*.o)
diff --git a/src/glu/mesa/descrip.mms b/src/glu/mesa/descrip.mms
deleted file mode 100644
index 2b3f64d8bc..0000000000
--- a/src/glu/mesa/descrip.mms
+++ /dev/null
@@ -1,61 +0,0 @@
-# Makefile for GLU for VMS
-# contributed by Jouk Jansen  joukj@hrem.stm.tudelft.nl
-
-.first
-	define gl [-.include.gl]
-
-.include [-]mms-config.
-
-##### MACROS #####
-
-VPATH = RCS
-
-INCDIR = $disk2:[-.include]
-LIBDIR = [-.lib]
-CFLAGS = /include=$(INCDIR)/define=(FBIND=1)/name=(as_is,short)/float=ieee/ieee=denorm
-
-SOURCES = glu.c mipmap.c nurbs.c nurbscrv.c nurbssrf.c nurbsutl.c \
-	polytest.c project.c quadric.c tess.c tesselat.c
-
-OBJECTS =glu.obj,mipmap.obj,nurbs.obj,nurbscrv.obj,nurbssrf.obj,nurbsutl.obj,\
-	polytest.obj,project.obj,quadric.obj,tess.obj,tesselat.obj
-
-
-##### RULES #####
-
-VERSION=MesaGlu V3.2
-
-##### TARGETS #####
-
-# Make the library:
-$(LIBDIR)$(GLU_LIB) : $(OBJECTS)
-.ifdef SHARE
-  @ WRITE_ SYS$OUTPUT "  generating mesagl1.opt"
-  @ OPEN_/WRITE FILE  mesagl1.opt
-  @ WRITE_ FILE "!"
-  @ WRITE_ FILE "! mesagl1.opt generated by DESCRIP.$(MMS_EXT)" 
-  @ WRITE_ FILE "!"
-  @ WRITE_ FILE "IDENTIFICATION=""$(VERSION)"""
-  @ WRITE_ FILE "GSMATCH=LEQUAL,3,2
-  @ WRITE_ FILE "$(OBJECTS)"
-  @ WRITE_ FILE "[-.lib]libmesagl.exe/SHARE"
-  @ WRITE_ FILE "SYS$SHARE:DECW$XEXTLIBSHR/SHARE"
-  @ WRITE_ FILE "SYS$SHARE:DECW$XLIBSHR/SHARE"
-  @ CLOSE_ FILE
-  @ WRITE_ SYS$OUTPUT "  generating mesagl.map ..."
-  @ LINK_/NODEB/NOSHARE/NOEXE/MAP=mesagl.map/FULL mesagl1.opt/OPT
-  @ WRITE_ SYS$OUTPUT "  analyzing mesagl.map ..."
-  @ @[-.vms]ANALYZE_MAP.COM mesagl.map mesagl.opt
-  @ WRITE_ SYS$OUTPUT "  linking $(GLU_LIB) ..."
-  @ LINK_/noinform/NODEB/SHARE=$(GLU_LIB)/MAP=mesagl.map/FULL mesagl1.opt/opt,mesagl.opt/opt
-.else
-  @ $(MAKELIB) $(GLU_LIB) $(OBJECTS)
-.endif
-  @ rename $(GLU_LIB)* $(LIBDIR)
-
-clean :
-	delete *.obj;*
-	purge
-
-include mms_depend.
-
diff --git a/src/glu/mesa/mms_depend b/src/glu/mesa/mms_depend
deleted file mode 100644
index ed59ca9de8..0000000000
--- a/src/glu/mesa/mms_depend
+++ /dev/null
@@ -1,15 +0,0 @@
-# DO NOT DELETE THIS LINE -- make depend depends on it.
-
-glu.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h
-mipmap.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h
-nurbs.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h nurbs.h
-nurbscrv.obj : nurbs.h gluP.h [-.include.gl]gl.h [-.include.gl]glu.h
-nurbssrf.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h nurbs.h
-nurbsutl.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h nurbs.h
-project.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h
-quadric.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h
-tess.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h tess.h
-tess_fist.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h tess.h
-tess_hash.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h tess.h
-tess_heap.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h tess.h
-tess_clip.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h tess.h
diff --git a/src/glu/sgi/.gitignore b/src/glu/sgi/.gitignore
new file mode 100644
index 0000000000..279ea7d434
--- /dev/null
+++ b/src/glu/sgi/.gitignore
@@ -0,0 +1 @@
+exptmp
diff --git a/src/glu/sgi/Makefile.DJ b/src/glu/sgi/Makefile.DJ
deleted file mode 100644
index b5df3e846a..0000000000
--- a/src/glu/sgi/Makefile.DJ
+++ /dev/null
@@ -1,188 +0,0 @@
-# Mesa 3-D graphics library
-# Version:  4.0
-# 
-# Copyright (C) 1999  Brian Paul   All Rights Reserved.
-# 
-# Permission is hereby granted, free of charge, to any person obtaining a
-# copy of this software and associated documentation files (the "Software"),
-# to deal in the Software without restriction, including without limitation
-# the rights to use, copy, modify, merge, publish, distribute, sublicense,
-# and/or sell copies of the Software, and to permit persons to whom the
-# Software is furnished to do so, subject to the following conditions:
-# 
-# The above copyright notice and this permission notice shall be included
-# in all copies or substantial portions of the Software.
-# 
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
-# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
-# BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
-# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
-# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-# DOS/DJGPP glu makefile v1.5 for Mesa
-#
-#  Copyright (C) 2002 - Daniel Borca
-#  Email : dborca@users.sourceforge.net
-#  Web   : http://www.geocities.com/dborca
-
-
-#
-#  Available options:
-#
-#    Environment variables:
-#	CFLAGS
-#
-#    Targets:
-#	all:		build GLU
-#	clean:		remove object files
-#
-
-
-
-.PHONY: all clean
-
-TOP = ../../..
-LIBDIR = $(TOP)/lib
-GLU_LIB = libglu.a
-GLU_DXE = glu.dxe
-GLU_IMP = libiglu.a
-
-export LD_LIBRARY_PATH := $(LD_LIBRARY_PATH);$(LIBDIR);$(GLIDE)/lib
-
-CC = gcc
-CFLAGS += -DNDEBUG -DLIBRARYBUILD -I$(TOP)/include -Iinclude
-CXX = gpp
-CXXFLAGS = $(CFLAGS) -Ilibnurbs/internals -Ilibnurbs/interface -Ilibnurbs/nurbtess
-
-AR = ar
-ARFLAGS = crus
-
-HAVEDXE3 = $(wildcard $(DJDIR)/bin/dxe3gen.exe)
-
-ifeq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),)
-UNLINK = del $(subst /,\,$(1))
-else
-UNLINK = $(RM) $(1)
-endif
-
-C_SOURCES = \
-	libutil/error.c \
-	libutil/glue.c \
-	libutil/mipmap.c \
-	libutil/project.c \
-	libutil/quad.c \
-	libutil/registry.c \
-	libtess/dict.c \
-	libtess/geom.c \
-	libtess/memalloc.c \
-	libtess/mesh.c \
-	libtess/normal.c \
-	libtess/priorityq.c \
-	libtess/render.c \
-	libtess/sweep.c \
-	libtess/tess.c \
-	libtess/tessmono.c
-
-CC_SOURCES = \
-	libnurbs/interface/bezierEval.cc \
-	libnurbs/interface/bezierPatch.cc \
-	libnurbs/interface/bezierPatchMesh.cc \
-	libnurbs/interface/glcurveval.cc \
-	libnurbs/interface/glinterface.cc \
-	libnurbs/interface/glrenderer.cc \
-	libnurbs/interface/glsurfeval.cc \
-	libnurbs/interface/incurveeval.cc \
-	libnurbs/interface/insurfeval.cc \
-	libnurbs/internals/arc.cc \
-	libnurbs/internals/arcsorter.cc \
-	libnurbs/internals/arctess.cc \
-	libnurbs/internals/backend.cc \
-	libnurbs/internals/basiccrveval.cc \
-	libnurbs/internals/basicsurfeval.cc \
-	libnurbs/internals/bin.cc \
-	libnurbs/internals/bufpool.cc \
-	libnurbs/internals/cachingeval.cc \
-	libnurbs/internals/ccw.cc \
-	libnurbs/internals/coveandtiler.cc \
-	libnurbs/internals/curve.cc \
-	libnurbs/internals/curvelist.cc \
-	libnurbs/internals/curvesub.cc \
-	libnurbs/internals/dataTransform.cc \
-	libnurbs/internals/displaylist.cc \
-	libnurbs/internals/flist.cc \
-	libnurbs/internals/flistsorter.cc \
-	libnurbs/internals/hull.cc \
-	libnurbs/internals/intersect.cc \
-	libnurbs/internals/knotvector.cc \
-	libnurbs/internals/mapdesc.cc \
-	libnurbs/internals/mapdescv.cc \
-	libnurbs/internals/maplist.cc \
-	libnurbs/internals/mesher.cc \
-	libnurbs/internals/monoTriangulationBackend.cc \
-	libnurbs/internals/monotonizer.cc \
-	libnurbs/internals/mycode.cc \
-	libnurbs/internals/nurbsinterfac.cc \
-	libnurbs/internals/nurbstess.cc \
-	libnurbs/internals/patch.cc \
-	libnurbs/internals/patchlist.cc \
-	libnurbs/internals/quilt.cc \
-	libnurbs/internals/reader.cc \
-	libnurbs/internals/renderhints.cc \
-	libnurbs/internals/slicer.cc \
-	libnurbs/internals/sorter.cc \
-	libnurbs/internals/splitarcs.cc \
-	libnurbs/internals/subdivider.cc \
-	libnurbs/internals/tobezier.cc \
-	libnurbs/internals/trimline.cc \
-	libnurbs/internals/trimregion.cc \
-	libnurbs/internals/trimvertpool.cc \
-	libnurbs/internals/uarray.cc \
-	libnurbs/internals/varray.cc \
-	libnurbs/nurbtess/directedLine.cc \
-	libnurbs/nurbtess/gridWrap.cc \
-	libnurbs/nurbtess/monoChain.cc \
-	libnurbs/nurbtess/monoPolyPart.cc \
-	libnurbs/nurbtess/monoTriangulation.cc \
-	libnurbs/nurbtess/partitionX.cc \
-	libnurbs/nurbtess/partitionY.cc \
-	libnurbs/nurbtess/polyDBG.cc \
-	libnurbs/nurbtess/polyUtil.cc \
-	libnurbs/nurbtess/primitiveStream.cc \
-	libnurbs/nurbtess/quicksort.cc \
-	libnurbs/nurbtess/rectBlock.cc \
-	libnurbs/nurbtess/sampleComp.cc \
-	libnurbs/nurbtess/sampleCompBot.cc \
-	libnurbs/nurbtess/sampleCompRight.cc \
-	libnurbs/nurbtess/sampleCompTop.cc \
-	libnurbs/nurbtess/sampleMonoPoly.cc \
-	libnurbs/nurbtess/sampledLine.cc \
-	libnurbs/nurbtess/searchTree.cc
-
-SOURCES = $(C_SOURCES) $(CC_SOURCES)
-
-OBJECTS = $(addsuffix .o,$(basename $(SOURCES)))
-
-.c.o:
-	$(CC) -o $@ $(CFLAGS) -c $<
-.cc.o:
-	$(CXX) -o $@ $(CXXFLAGS) -c $<
-
-all: $(LIBDIR)/$(GLU_LIB) $(LIBDIR)/$(GLU_DXE) $(LIBDIR)/$(GLU_IMP)
-
-$(LIBDIR)/$(GLU_LIB): $(OBJECTS)
-	$(AR) $(ARFLAGS) $@ $^
-
-$(LIBDIR)/$(GLU_DXE) $(LIBDIR)/$(GLU_IMP): $(OBJECTS)
-ifeq ($(HAVEDXE3),)
-	$(warning Missing DXE3 package... Skipping $(GLU_DXE))
-else
-	-dxe3gen -o $(LIBDIR)/$(GLU_DXE) -Y $(LIBDIR)/$(GLU_IMP) -D "MesaGLU/SGI DJGPP" -E _glu -P gl.dxe -U $^
-endif
-
-clean:
-	-$(call UNLINK,libutil/*.o)
-	-$(call UNLINK,libtess/*.o)
-	-$(call UNLINK,libnurbs/interface/*.o)
-	-$(call UNLINK,libnurbs/internals/*.o)
-	-$(call UNLINK,libnurbs/nurbtess/*.o)
diff --git a/src/glu/sgi/descrip.mms b/src/glu/sgi/descrip.mms
deleted file mode 100644
index 5abc8b2e04..0000000000
--- a/src/glu/sgi/descrip.mms
+++ /dev/null
@@ -1,451 +0,0 @@
-# Makefile for GLU for VMS
-# contributed by Jouk Jansen  joukj@hrem.stm.tudelft.nl
-
-.first
-	define gl [---.include.gl]
-
-.include [---]mms-config.
-
-##### MACROS #####
-
-VPATH = RCS
-
-INCDIR =([-.include],[.include],[.internals],[.libnurbs.internals],\
-       [.libnurbs.interface],[.libnurbs.nurbtess])
-LIBDIR = [---.lib]
-CFLAGS = /include=$(INCDIR)/name=(as_is,short)/float=ieee/ieee=denorm
-
-LU_OBJECTS=\
-	[.libutil]error.obj,	\
-	[.libutil]glue.obj,	\
-	[.libutil]mipmap.obj,\
-	[.libutil]project.obj,\
-	[.libutil]quad.obj,	\
-	[.libutil]registry.obj
-
-LT_OBJECTS=[.libtess]dict.obj,	\
-	[.libtess]geom.obj,	\
-	[.libtess]memalloc.obj,\
-	[.libtess]mesh.obj,	\
-	[.libtess]normal.obj,\
-	[.libtess]priorityq.obj,\
-	[.libtess]render.obj,\
-	[.libtess]sweep.obj,	\
-	[.libtess]tess.obj,	\
-	[.libtess]tessmono.obj
-
-LI_OBJECTS=[.libnurbs.interface]bezierEval.obj,	\
-	[.libnurbs.interface]bezierPatch.obj,	\
-	[.libnurbs.interface]bezierPatchMesh.obj,	\
-	[.libnurbs.interface]glcurveval.obj,	\
-	[.libnurbs.interface]glinterface.obj
-
-LI_OBJECTS1=[.libnurbs.interface]glrenderer.obj,	\
-	[.libnurbs.interface]glsurfeval.obj,	\
-	[.libnurbs.interface]incurveeval.obj,	\
-	[.libnurbs.interface]insurfeval.obj
-
-LI2_OBJECTS=[.libnurbs.internals]arc.obj,		\
-	[.libnurbs.internals]arcsorter.obj,		\
-	[.libnurbs.internals]arctess.obj,		\
-	[.libnurbs.internals]backend.obj,		\
-	[.libnurbs.internals]basiccrveval.obj,	\
-	[.libnurbs.internals]basicsurfeval.obj
-
-LI2_OBJECTS1=[.libnurbs.internals]bin.obj,		\
-	[.libnurbs.internals]bufpool.obj,		\
-	[.libnurbs.internals]cachingeval.obj,	\
-	[.libnurbs.internals]ccw.obj,		\
-	[.libnurbs.internals]coveandtiler.obj,	\
-	[.libnurbs.internals]curve.obj,		\
-	[.libnurbs.internals]curvelist.obj
-
-LI2_OBJECTS2=[.libnurbs.internals]curvesub.obj,		\
-	[.libnurbs.internals]dataTransform.obj,	\
-	[.libnurbs.internals]displaylist.obj,	\
-	[.libnurbs.internals]flist.obj,		\
-	[.libnurbs.internals]flistsorter.obj
-
-LI2_OBJECTS3=[.libnurbs.internals]hull.obj,		\
-	[.libnurbs.internals]intersect.obj,		\
-	[.libnurbs.internals]knotvector.obj,	\
-	[.libnurbs.internals]mapdesc.obj
-
-LI2_OBJECTS4=[.libnurbs.internals]mapdescv.obj,		\
-	[.libnurbs.internals]maplist.obj,		\
-	[.libnurbs.internals]mesher.obj,		\
-	[.libnurbs.internals]monoTriangulationBackend.obj,\
-	[.libnurbs.internals]monotonizer.obj
-
-LI2_OBJECTS5=[.libnurbs.internals]mycode.obj,		\
-	[.libnurbs.internals]nurbsinterfac.obj,	\
-	[.libnurbs.internals]nurbstess.obj,		\
-	[.libnurbs.internals]patch.obj
-
-LI2_OBJECTS6=[.libnurbs.internals]patchlist.obj,		\
-	[.libnurbs.internals]quilt.obj,		\
-	[.libnurbs.internals]reader.obj,		\
-	[.libnurbs.internals]renderhints.obj,	\
-	[.libnurbs.internals]slicer.obj
-
-LI2_OBJECTS7=[.libnurbs.internals]sorter.obj,		\
-	[.libnurbs.internals]splitarcs.obj,		\
-	[.libnurbs.internals]subdivider.obj,	\
-	[.libnurbs.internals]tobezier.obj
-
-LI2_OBJECTS8=[.libnurbs.internals]trimline.obj,		\
-	[.libnurbs.internals]trimregion.obj,	\
-	[.libnurbs.internals]trimvertpool.obj,	\
-	[.libnurbs.internals]uarray.obj,		\
-	[.libnurbs.internals]varray.obj
-
-LN_OBJECTS=[.libnurbs.nurbtess]directedLine.obj,	\
-	[.libnurbs.nurbtess]gridWrap.obj,		\
-	[.libnurbs.nurbtess]monoChain.obj,		\
-	[.libnurbs.nurbtess]monoPolyPart.obj,	\
-	[.libnurbs.nurbtess]monoTriangulation.obj
-
-LN_OBJECTS1=[.libnurbs.nurbtess]partitionX.obj,		\
-	[.libnurbs.nurbtess]partitionY.obj,		\
-	[.libnurbs.nurbtess]polyDBG.obj
-
-LN_OBJECTS2=[.libnurbs.nurbtess]polyUtil.obj,		\
-	[.libnurbs.nurbtess]primitiveStream.obj,	\
-	[.libnurbs.nurbtess]quicksort.obj,		\
-	[.libnurbs.nurbtess]rectBlock.obj
-
-LN_OBJECTS3=[.libnurbs.nurbtess]sampleComp.obj,		\
-	[.libnurbs.nurbtess]sampleCompBot.obj,	\
-	[.libnurbs.nurbtess]sampleCompRight.obj
-
-LN_OBJECTS4=[.libnurbs.nurbtess]sampleCompTop.obj,	\
-	[.libnurbs.nurbtess]sampleMonoPoly.obj,\
-	[.libnurbs.nurbtess]sampledLine.obj,	\
-	[.libnurbs.nurbtess]searchTree.obj
-
-##### RULES #####
-
-VERSION=MesaGlu V3.5
-
-##### TARGETS #####
-
-# Make the library:
-$(LIBDIR)$(GLU_LIB) : $(LU_OBJECTS) $(LT_OBJECTS) $(LI_OBJECTS) $(LI_OBJECTS1)\
-		    $(LI2_OBJECTS) $(LI2_OBJECTS1) $(LI2_OBJECTS2)\
-		    $(LI2_OBJECTS3) $(LI2_OBJECTS4) $(LI2_OBJECTS5)\
-		    $(LI2_OBJECTS6) $(LI2_OBJECTS7) $(LI2_OBJECTS8)\
-		    $(LN_OBJECTS) $(LN_OBJECTS1) $(LN_OBJECTS2)\
-		    $(LN_OBJECTS3) $(LN_OBJECTS4)
-  @ $(MAKELIB) $(GLU_LIB)  $(LU_OBJECTS),$(LT_OBJECTS),$(LI_OBJECTS),\
-	$(LI2_OBJECTS),$(LN_OBJECTS)
-  @ rename $(GLU_LIB)* $(LIBDIR)
-.ifdef SHARE
-  @ WRITE_ SYS$OUTPUT "  generating mesagl1.opt"
-  @ OPEN_/WRITE FILE  mesagl1.opt
-  @ WRITE_ FILE "!"
-  @ WRITE_ FILE "! mesagl1.opt generated by DESCRIP.$(MMS_EXT)" 
-  @ WRITE_ FILE "!"
-  @ WRITE_ FILE "IDENTIFICATION=""$(VERSION)"""
-  @ WRITE_ FILE "GSMATCH=LEQUAL,3,5
-  @ WRITE_ FILE "$(LU_OBJECTS)"
-  @ WRITE_ FILE "$(LT_OBJECTS)"
-  @ WRITE_ FILE "$(LI_OBJECTS)"
-  @ WRITE_ FILE "$(LI_OBJECTS1)"
-  @ WRITE_ FILE "$(LI2_OBJECTS)"
-  @ WRITE_ FILE "$(LI2_OBJECTS1)"
-  @ WRITE_ FILE "$(LI2_OBJECTS2)"
-  @ WRITE_ FILE "$(LI2_OBJECTS3)"
-  @ WRITE_ FILE "$(LI2_OBJECTS4)"
-  @ WRITE_ FILE "$(LI2_OBJECTS5)"
-  @ WRITE_ FILE "$(LI2_OBJECTS6)"
-  @ WRITE_ FILE "$(LI2_OBJECTS7)"
-  @ WRITE_ FILE "$(LI2_OBJECTS8)"
-  @ WRITE_ FILE "$(LN_OBJECTS)"
-  @ WRITE_ FILE "$(LN_OBJECTS1)"
-  @ WRITE_ FILE "$(LN_OBJECTS2)"
-  @ WRITE_ FILE "$(LN_OBJECTS3)"
-  @ WRITE_ FILE "$(LN_OBJECTS4)"
-  @ WRITE_ FILE "[---.lib]libmesagl.exe/SHARE"
-  @ WRITE_ FILE "SYS$SHARE:DECW$XEXTLIBSHR/SHARE"
-  @ WRITE_ FILE "SYS$SHARE:DECW$XLIBSHR/SHARE"
-  @ CLOSE_ FILE
-#  @ WRITE_ SYS$OUTPUT "  generating mesagl.map ..."
-#  @ CXXLINK_/NODEB/NOSHARE/NOEXE/MAP=mesagl.map/FULL mesagl1.opt/OPT
-#  @ WRITE_ SYS$OUTPUT "  analyzing mesagl.map ..."
-#  @ @[-.vms]ANALYZE_MAP.COM mesagl.map mesagl.opt
-  @ WRITE_ SYS$OUTPUT "  linking $(GLU_SHAR) ..."
-#  @ CXXLINK_/noinform/NODEB/SHARE=$(GLU_SHAR)/MAP=mesagl.map/FULL mesagl1.opt/opt,mesagl.opt/opt
-  @ CXXLINK_/noinform/NODEB/SHARE=$(GLU_SHAR)/MAP=mesagl.map/FULL mesagl1.opt/opt,mesaglu.opt/opt
-  @ rename $(GLU_SHAR)* $(LIBDIR)
-.endif
-
-clean :
-	delete [...]*.obj;*
-	purge
-
-[.libutil]error.obj : [.libutil]error.c
-	$(CC) $(CFLAGS) /obj=[.libutil]error.obj [.libutil]error.c
-
-[.libutil]glue.obj : [.libutil]glue.c
-	$(CC) $(CFLAGS) /obj=[.libutil]glue.obj [.libutil]glue.c
-
-[.libutil]mipmap.obj : [.libutil]mipmap.c
-	$(CC) $(CFLAGS) /obj=[.libutil]mipmap.obj [.libutil]mipmap.c
-
-[.libutil]project.obj : [.libutil]project.c
-	$(CC) $(CFLAGS) /obj=[.libutil]project.obj [.libutil]project.c
-
-[.libutil]quad.obj : [.libutil]quad.c
-	$(CC) $(CFLAGS) /obj=[.libutil]quad.obj [.libutil]quad.c
-
-[.libutil]registry.obj : [.libutil]registry.c
-	$(CC) $(CFLAGS) /obj=[.libutil]registry.obj [.libutil]registry.c
-
-[.libtess]dict.obj : [.libtess]dict.c
-	$(CC) $(CFLAGS) /obj=[.libtess]dict.obj [.libtess]dict.c
-
-[.libtess]geom.obj : [.libtess]geom.c
-	$(CC) $(CFLAGS) /obj=[.libtess]geom.obj [.libtess]geom.c
-
-[.libtess]memalloc.obj : [.libtess]memalloc.c
-	$(CC) $(CFLAGS) /obj=[.libtess]memalloc.obj [.libtess]memalloc.c
-
-[.libtess]mesh.obj : [.libtess]mesh.c
-	$(CC) $(CFLAGS) /obj=[.libtess]mesh.obj [.libtess]mesh.c
-
-[.libtess]normal.obj : [.libtess]normal.c
-	$(CC) $(CFLAGS) /obj=[.libtess]normal.obj [.libtess]normal.c
-
-[.libtess]priorityq.obj : [.libtess]priorityq.c
-	$(CC) $(CFLAGS) /obj=[.libtess]priorityq.obj [.libtess]priorityq.c
-
-[.libtess]render.obj : [.libtess]render.c
-	$(CC) $(CFLAGS) /obj=[.libtess]render.obj [.libtess]render.c
-
-[.libtess]sweep.obj : [.libtess]sweep.c
-	$(CC) $(CFLAGS) /obj=[.libtess]sweep.obj [.libtess]sweep.c
-
-[.libtess]tess.obj : [.libtess]tess.c
-	$(CC) $(CFLAGS) /obj=[.libtess]tess.obj [.libtess]tess.c
-
-[.libtess]tessmono.obj : [.libtess]tessmono.c
-	$(CC) $(CFLAGS) /obj=[.libtess]tessmono.obj [.libtess]tessmono.c
-
-[.libnurbs.interface]bezierEval.obj : [.libnurbs.interface]bezierEval.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.interface]bezierEval.obj [.libnurbs.interface]bezierEval.cc
-
-[.libnurbs.interface]bezierPatch.obj : [.libnurbs.interface]bezierPatch.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.interface]bezierPatch.obj [.libnurbs.interface]bezierPatch.cc
-
-[.libnurbs.interface]bezierPatchMesh.obj : [.libnurbs.interface]bezierPatchMesh.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.interface]bezierPatchMesh.obj [.libnurbs.interface]bezierPatchMesh.cc
-
-[.libnurbs.interface]glcurveval.obj : [.libnurbs.interface]glcurveval.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.interface]glcurveval.obj [.libnurbs.interface]glcurveval.cc
-
-[.libnurbs.interface]glinterface.obj : [.libnurbs.interface]glinterface.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.interface]glinterface.obj [.libnurbs.interface]glinterface.cc
-
-[.libnurbs.interface]glrenderer.obj : [.libnurbs.interface]glrenderer.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.interface]glrenderer.obj [.libnurbs.interface]glrenderer.cc
-
-[.libnurbs.interface]glsurfeval.obj : [.libnurbs.interface]glsurfeval.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.interface]glsurfeval.obj [.libnurbs.interface]glsurfeval.cc
-
-[.libnurbs.interface]incurveeval.obj : [.libnurbs.interface]incurveeval.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.interface]incurveeval.obj [.libnurbs.interface]incurveeval.cc
-
-[.libnurbs.interface]insurfeval.obj : [.libnurbs.interface]insurfeval.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.interface]insurfeval.obj [.libnurbs.interface]insurfeval.cc
-
-[.libnurbs.internals]arc.obj : [.libnurbs.internals]arc.cc
-	$(CXX) $(CFLAGS)/list/show=all /obj=[.libnurbs.internals]arc.obj [.libnurbs.internals]arc.cc
-
-[.libnurbs.internals]arcsorter.obj : [.libnurbs.internals]arcsorter.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]arcsorter.obj [.libnurbs.internals]arcsorter.cc
-
-[.libnurbs.internals]arctess.obj : [.libnurbs.internals]arctess.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]arctess.obj [.libnurbs.internals]arctess.cc
-
-[.libnurbs.internals]backend.obj : [.libnurbs.internals]backend.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]backend.obj [.libnurbs.internals]backend.cc
-
-[.libnurbs.internals]basiccrveval.obj : [.libnurbs.internals]basiccrveval.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]basiccrveval.obj [.libnurbs.internals]basiccrveval.cc
-
-[.libnurbs.internals]basicsurfeval.obj : [.libnurbs.internals]basicsurfeval.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]basicsurfeval.obj [.libnurbs.internals]basicsurfeval.cc
-
-[.libnurbs.internals]bin.obj : [.libnurbs.internals]bin.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]bin.obj [.libnurbs.internals]bin.cc
-
-[.libnurbs.internals]bufpool.obj : [.libnurbs.internals]bufpool.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]bufpool.obj [.libnurbs.internals]bufpool.cc
-
-[.libnurbs.internals]cachingeval.obj : [.libnurbs.internals]cachingeval.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]cachingeval.obj [.libnurbs.internals]cachingeval.cc
-
-[.libnurbs.internals]ccw.obj : [.libnurbs.internals]ccw.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]ccw.obj [.libnurbs.internals]ccw.cc
-
-[.libnurbs.internals]coveandtiler.obj : [.libnurbs.internals]coveandtiler.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]coveandtiler.obj [.libnurbs.internals]coveandtiler.cc
-
-[.libnurbs.internals]curve.obj : [.libnurbs.internals]curve.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]curve.obj [.libnurbs.internals]curve.cc
-
-[.libnurbs.internals]curvelist.obj : [.libnurbs.internals]curvelist.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]curvelist.obj [.libnurbs.internals]curvelist.cc
-
-[.libnurbs.internals]curvesub.obj : [.libnurbs.internals]curvesub.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]curvesub.obj [.libnurbs.internals]curvesub.cc
-
-[.libnurbs.internals]dataTransform.obj : [.libnurbs.internals]dataTransform.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]dataTransform.obj [.libnurbs.internals]dataTransform.cc
-
-[.libnurbs.internals]displaylist.obj : [.libnurbs.internals]displaylist.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]displaylist.obj [.libnurbs.internals]displaylist.cc
-
-[.libnurbs.internals]flist.obj : [.libnurbs.internals]flist.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]flist.obj [.libnurbs.internals]flist.cc
-
-[.libnurbs.internals]flistsorter.obj : [.libnurbs.internals]flistsorter.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]flistsorter.obj [.libnurbs.internals]flistsorter.cc
-
-[.libnurbs.internals]hull.obj : [.libnurbs.internals]hull.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]hull.obj [.libnurbs.internals]hull.cc
-
-[.libnurbs.internals]intersect.obj : [.libnurbs.internals]intersect.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]intersect.obj [.libnurbs.internals]intersect.cc
-
-[.libnurbs.internals]knotvector.obj : [.libnurbs.internals]knotvector.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]knotvector.obj [.libnurbs.internals]knotvector.cc
-
-[.libnurbs.internals]mapdesc.obj : [.libnurbs.internals]mapdesc.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]mapdesc.obj [.libnurbs.internals]mapdesc.cc
-
-[.libnurbs.internals]mapdescv.obj : [.libnurbs.internals]mapdescv.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]mapdescv.obj [.libnurbs.internals]mapdescv.cc
-
-[.libnurbs.internals]maplist.obj : [.libnurbs.internals]maplist.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]maplist.obj [.libnurbs.internals]maplist.cc
-
-[.libnurbs.internals]mesher.obj : [.libnurbs.internals]mesher.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]mesher.obj [.libnurbs.internals]mesher.cc
-
-[.libnurbs.internals]monoTriangulationBackend.obj : [.libnurbs.internals]monoTriangulationBackend.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]monoTriangulationBackend.obj [.libnurbs.internals]monoTriangulationBackend.cc
-
-[.libnurbs.internals]monotonizer.obj : [.libnurbs.internals]monotonizer.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]monotonizer.obj [.libnurbs.internals]monotonizer.cc
-
-[.libnurbs.internals]mycode.obj : [.libnurbs.internals]mycode.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]mycode.obj [.libnurbs.internals]mycode.cc
-
-[.libnurbs.internals]nurbsinterfac.obj : [.libnurbs.internals]nurbsinterfac.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]nurbsinterfac.obj [.libnurbs.internals]nurbsinterfac.cc
-
-[.libnurbs.internals]nurbstess.obj : [.libnurbs.internals]nurbstess.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]nurbstess.obj [.libnurbs.internals]nurbstess.cc
-
-[.libnurbs.internals]patch.obj : [.libnurbs.internals]patch.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]patch.obj [.libnurbs.internals]patch.cc
-
-[.libnurbs.internals]patchlist.obj : [.libnurbs.internals]patchlist.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]patchlist.obj [.libnurbs.internals]patchlist.cc
-
-[.libnurbs.internals]quilt.obj : [.libnurbs.internals]quilt.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]quilt.obj [.libnurbs.internals]quilt.cc
-
-[.libnurbs.internals]reader.obj : [.libnurbs.internals]reader.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]reader.obj [.libnurbs.internals]reader.cc
-
-[.libnurbs.internals]renderhints.obj : [.libnurbs.internals]renderhints.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]renderhints.obj [.libnurbs.internals]renderhints.cc
-
-[.libnurbs.internals]slicer.obj : [.libnurbs.internals]slicer.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]slicer.obj [.libnurbs.internals]slicer.cc
-
-[.libnurbs.internals]sorter.obj : [.libnurbs.internals]sorter.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]sorter.obj [.libnurbs.internals]sorter.cc
-
-[.libnurbs.internals]splitarcs.obj : [.libnurbs.internals]splitarcs.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]splitarcs.obj [.libnurbs.internals]splitarcs.cc
-
-[.libnurbs.internals]subdivider.obj : [.libnurbs.internals]subdivider.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]subdivider.obj [.libnurbs.internals]subdivider.cc
-
-[.libnurbs.internals]tobezier.obj : [.libnurbs.internals]tobezier.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]tobezier.obj [.libnurbs.internals]tobezier.cc
-
-[.libnurbs.internals]trimline.obj : [.libnurbs.internals]trimline.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]trimline.obj [.libnurbs.internals]trimline.cc
-
-[.libnurbs.internals]trimregion.obj : [.libnurbs.internals]trimregion.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]trimregion.obj [.libnurbs.internals]trimregion.cc
-
-[.libnurbs.internals]trimvertpool.obj : [.libnurbs.internals]trimvertpool.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]trimvertpool.obj [.libnurbs.internals]trimvertpool.cc
-
-[.libnurbs.internals]uarray.obj : [.libnurbs.internals]uarray.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]uarray.obj [.libnurbs.internals]uarray.cc
-
-[.libnurbs.internals]varray.obj : [.libnurbs.internals]varray.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.internals]varray.obj [.libnurbs.internals]varray.cc
-
-[.libnurbs.nurbtess]directedLine.obj : [.libnurbs.nurbtess]directedLine.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]directedLine.obj [.libnurbs.nurbtess]directedLine.cc
-
-[.libnurbs.nurbtess]gridWrap.obj : [.libnurbs.nurbtess]gridWrap.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]gridWrap.obj [.libnurbs.nurbtess]gridWrap.cc
-
-[.libnurbs.nurbtess]monoChain.obj : [.libnurbs.nurbtess]monoChain.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]monoChain.obj [.libnurbs.nurbtess]monoChain.cc
-
-[.libnurbs.nurbtess]monoPolyPart.obj : [.libnurbs.nurbtess]monoPolyPart.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]monoPolyPart.obj [.libnurbs.nurbtess]monoPolyPart.cc
-
-[.libnurbs.nurbtess]monoTriangulation.obj : [.libnurbs.nurbtess]monoTriangulation.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]monoTriangulation.obj [.libnurbs.nurbtess]monoTriangulation.cc
-
-[.libnurbs.nurbtess]partitionX.obj : [.libnurbs.nurbtess]partitionX.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]partitionX.obj [.libnurbs.nurbtess]partitionX.cc
-
-[.libnurbs.nurbtess]partitionY.obj : [.libnurbs.nurbtess]partitionY.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]partitionY.obj [.libnurbs.nurbtess]partitionY.cc
-
-[.libnurbs.nurbtess]polyDBG.obj : [.libnurbs.nurbtess]polyDBG.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]polyDBG.obj [.libnurbs.nurbtess]polyDBG.cc
-
-[.libnurbs.nurbtess]polyUtil.obj : [.libnurbs.nurbtess]polyUtil.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]polyUtil.obj [.libnurbs.nurbtess]polyUtil.cc
-
-[.libnurbs.nurbtess]primitiveStream.obj : [.libnurbs.nurbtess]primitiveStream.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]primitiveStream.obj [.libnurbs.nurbtess]primitiveStream.cc
-
-[.libnurbs.nurbtess]quicksort.obj : [.libnurbs.nurbtess]quicksort.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]quicksort.obj [.libnurbs.nurbtess]quicksort.cc
-
-[.libnurbs.nurbtess]rectBlock.obj : [.libnurbs.nurbtess]rectBlock.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]rectBlock.obj [.libnurbs.nurbtess]rectBlock.cc
-
-[.libnurbs.nurbtess]sampleComp.obj : [.libnurbs.nurbtess]sampleComp.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampleComp.obj [.libnurbs.nurbtess]sampleComp.cc
-
-[.libnurbs.nurbtess]sampleCompBot.obj : [.libnurbs.nurbtess]sampleCompBot.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampleCompBot.obj [.libnurbs.nurbtess]sampleCompBot.cc
-
-[.libnurbs.nurbtess]sampleCompRight.obj : [.libnurbs.nurbtess]sampleCompRight.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampleCompRight.obj [.libnurbs.nurbtess]sampleCompRight.cc
-
-[.libnurbs.nurbtess]sampleCompTop.obj : [.libnurbs.nurbtess]sampleCompTop.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampleCompTop.obj [.libnurbs.nurbtess]sampleCompTop.cc
-
-[.libnurbs.nurbtess]sampleMonoPoly.obj : [.libnurbs.nurbtess]sampleMonoPoly.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampleMonoPoly.obj [.libnurbs.nurbtess]sampleMonoPoly.cc
-
-[.libnurbs.nurbtess]sampledLine.obj : [.libnurbs.nurbtess]sampledLine.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampledLine.obj [.libnurbs.nurbtess]sampledLine.cc
-
-[.libnurbs.nurbtess]searchTree.obj : [.libnurbs.nurbtess]searchTree.cc
-	$(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]searchTree.obj [.libnurbs.nurbtess]searchTree.cc
diff --git a/src/glu/sgi/glu.exports b/src/glu/sgi/glu.exports
index 1d1b6da24f..aeb7272916 100644
--- a/src/glu/sgi/glu.exports
+++ b/src/glu/sgi/glu.exports
@@ -57,3 +57,62 @@
 	gluTessVertex
 	gluUnProject
 	gluUnProject4
+	mgluBeginCurve
+	mgluBeginPolygon
+	mgluBeginSurface
+	mgluBeginTrim
+	mgluBuild1DMipmapLevels
+	mgluBuild1DMipmaps
+	mgluBuild2DMipmapLevels
+	mgluBuild2DMipmaps
+	mgluBuild3DMipmapLevels
+	mgluBuild3DMipmaps
+	mgluCheckExtension
+	mgluCylinder
+	mgluDeleteNurbsRenderer
+	mgluDeleteQuadric
+	mgluDeleteTess
+	mgluDisk
+	mgluEndCurve
+	mgluEndPolygon
+	mgluEndSurface
+	mgluEndTrim
+	mgluErrorString
+	mgluGetNurbsProperty
+	mgluGetString
+	mgluGetTessProperty
+	mgluLoadSamplingMatrices
+	mgluLookAt
+	mgluNewNurbsRenderer
+	mgluNewQuadric
+	mgluNewTess
+	mgluNextContour
+	mgluNurbsCallback
+	mgluNurbsCallbackData
+	mgluNurbsCallbackDataEXT
+	mgluNurbsCurve
+	mgluNurbsProperty
+	mgluNurbsSurface
+	mgluOrtho2D
+	mgluPartialDisk
+	mgluPerspective
+	mgluPickMatrix
+	mgluProject
+	mgluPwlCurve
+	mgluQuadricCallback
+	mgluQuadricDrawStyle
+	mgluQuadricNormals
+	mgluQuadricOrientation
+	mgluQuadricTexture
+	mgluScaleImage
+	mgluSphere
+	mgluTessBeginContour
+	mgluTessBeginPolygon
+	mgluTessCallback
+	mgluTessEndContour
+	mgluTessEndPolygon
+	mgluTessNormal
+	mgluTessProperty
+	mgluTessVertex
+	mgluUnProject
+	mgluUnProject4
diff --git a/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc b/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc
index 3dc16313ff..ac7ff84fc4 100644
--- a/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc
+++ b/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc
@@ -103,23 +103,22 @@ bezierPatchMesh *bezierPatchMeshMake(int maptype, float umin, float umax, int us
   int dimension;
   int the_ustride;
   int the_vstride;
-  
-  bezierPatchMesh *ret = (bezierPatchMesh*) malloc(sizeof(bezierPatchMesh));
-  assert(ret);
 
-  ret->bpatch = NULL;
-  ret->bpatch_normal = NULL;
-  ret->bpatch_color  = NULL;
-  ret->bpatch_texcoord = NULL;
- 
   if(maptype == GL_MAP2_VERTEX_3) dimension = 3;
   else if (maptype==GL_MAP2_VERTEX_4) dimension = 4;
   else {
     fprintf(stderr, "error in inMap2f, maptype=%i is wrong, maptype,map is invalid\n", maptype);
     return NULL;
   }
-  
+
+  bezierPatchMesh *ret = (bezierPatchMesh*) malloc(sizeof(bezierPatchMesh));
+  assert(ret);
+
+  ret->bpatch_normal = NULL;
+  ret->bpatch_color  = NULL;
+  ret->bpatch_texcoord = NULL;
   ret->bpatch = bezierPatchMake(umin, vmin, umax, vmax, uorder, vorder, dimension);
+
   /*copy the control points there*/
   the_ustride = vorder * dimension;
   the_vstride = dimension;
diff --git a/src/glut/beos/Makefile b/src/glut/beos/Makefile
index 14f4f43f84..a9f8684c9b 100644
--- a/src/glut/beos/Makefile
+++ b/src/glut/beos/Makefile
@@ -92,6 +92,7 @@ clean:
 	-rm -f *.o
 
 depend: $(SOURCES)
+	rm -f depend
 	touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) > /dev/null 
 
diff --git a/src/glut/directfb/Makefile b/src/glut/directfb/Makefile
index 34c3743ef4..bf93d63e30 100644
--- a/src/glut/directfb/Makefile
+++ b/src/glut/directfb/Makefile
@@ -70,6 +70,7 @@ $(TOP)/$(LIB_DIR)/$(GLUT_LIB_NAME): depend $(OBJECTS)
 # Run 'make -f Makefile.solo dep' to update the dependencies if you change
 # what's included by any source file.
 depend: $(SOURCES) 
+	rm -f depend
 	touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) > /dev/null 
 
diff --git a/src/glut/dos/Makefile.DJ b/src/glut/dos/Makefile.DJ
deleted file mode 100644
index 7e4e0b8576..0000000000
--- a/src/glut/dos/Makefile.DJ
+++ /dev/null
@@ -1,126 +0,0 @@
-# DOS/DJGPP Mesa Utility Toolkit
-# Version:  1.0
-#
-# Copyright (C) 2005  Daniel Borca   All Rights Reserved.
-#
-# Permission is hereby granted, free of charge, to any person obtaining a
-# copy of this software and associated documentation files (the "Software"),
-# to deal in the Software without restriction, including without limitation
-# the rights to use, copy, modify, merge, publish, distribute, sublicense,
-# and/or sell copies of the Software, and to permit persons to whom the
-# Software is furnished to do so, subject to the following conditions:
-#
-# The above copyright notice and this permission notice shall be included
-# in all copies or substantial portions of the Software.
-#
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
-# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
-# DANIEL BORCA BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
-# IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
-# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-
-#
-#  Available options:
-#
-#    Environment variables:
-#	CFLAGS
-#
-#	GLIDE		path to Glide3 SDK; used to resolve DXEs.
-#			default = $(TOP)/glide3
-#
-#    Targets:
-#	all:		build GLUT
-#	clean:		remove object files
-#
-
-
-
-.PHONY: all clean
-
-TOP = ../../..
-GLIDE ?= $(TOP)/glide3
-LIBDIR = $(TOP)/lib
-GLUT_LIB = libglut.a
-GLUT_DXE = glut.dxe
-GLUT_IMP = libiglut.a
-
-export LD_LIBRARY_PATH := $(LD_LIBRARY_PATH);$(LIBDIR);$(GLIDE)/lib
-
-CC = gcc
-CFLAGS += -I$(TOP)/include -I. -IPC_HW
-CFLAGS += -DGLUT_IMPORT_LIB
-
-AR = ar
-ARFLAGS = crus
-
-HAVEDXE3 = $(wildcard $(DJDIR)/bin/dxe3gen.exe)
-
-ifeq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),)
-UNLINK = del $(subst /,\,$(1))
-else
-UNLINK = $(RM) $(1)
-endif
-
-CORE_SOURCES = \
-	loop.c \
-	callback.c \
-	color.c \
-	extens.c \
-	init.c \
-	menu.c \
-	mouse.c \
-	overlay.c \
-	state.c \
-	util.c \
-	window.c \
-	f8x13.c \
-	f9x15.c \
-	hel10.c \
-	hel12.c \
-	hel18.c \
-	tr10.c \
-	tr24.c \
-	mroman.c \
-	roman.c \
-	bitmap.c \
-	stroke.c \
-	teapot.c \
-	shapes.c
-
-PC_HW_SOURCES = \
-	PC_HW/pc_hw.c \
-	PC_HW/pc_keyb.c \
-	PC_HW/pc_mouse.c \
-	PC_HW/pc_timer.c \
-	PC_HW/pc_irq.S
-
-SOURCES = $(CORE_SOURCES) $(PC_HW_SOURCES)
-
-OBJECTS = $(addsuffix .o,$(basename $(SOURCES)))
-
-.c.o:
-	$(CC) -o $@ $(CFLAGS) -c $<
-.S.o:
-	$(CC) -o $@ $(CFLAGS) -c $<
-.s.o:
-	$(CC) -o $@ $(CFLAGS) -x assembler-with-cpp -c $<
-
-all: $(LIBDIR)/$(GLUT_LIB) $(LIBDIR)/$(GLUT_DXE) $(LIBDIR)/$(GLUT_IMP)
-
-$(LIBDIR)/$(GLUT_LIB): $(OBJECTS)
-	$(AR) $(ARFLAGS) $@ $^
-
-$(LIBDIR)/$(GLUT_DXE) $(LIBDIR)/$(GLUT_IMP): $(OBJECTS)
-ifeq ($(HAVEDXE3),)
-	$(warning Missing DXE3 package... Skipping $(GLUT_DXE))
-else
-	-dxe3gen -o $(LIBDIR)/$(GLUT_DXE) -Y $(LIBDIR)/$(GLUT_IMP) -D "MesaGLUT DJGPP" -E _glut -P gl.dxe -U $^
-endif
-
-clean:
-	-$(call UNLINK,*.o)
-	-$(call UNLINK,PC_HW/*.o)
-
--include depend
diff --git a/src/glut/fbdev/Makefile b/src/glut/fbdev/Makefile
index 4081ccec74..199d8c390a 100644
--- a/src/glut/fbdev/Makefile
+++ b/src/glut/fbdev/Makefile
@@ -77,6 +77,7 @@ install:
 # Run 'make -f Makefile.solo dep' to update the dependencies if you change
 # what's included by any source file.
 depend: $(SOURCES) 
+	rm -f depend
 	touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) > /dev/null 
 
diff --git a/src/glut/ggi/Makefile b/src/glut/ggi/Makefile
index 1ae1ba0d70..c23dfcac10 100644
--- a/src/glut/ggi/Makefile
+++ b/src/glut/ggi/Makefile
@@ -46,6 +46,7 @@ clean:
 
 
 depend: $(SOURCES)
+	rm -f depend
 	touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) -I$(TOP)/include $(SOURCES)
 
diff --git a/src/glut/glx/Makefile b/src/glut/glx/Makefile
index 7f886f775a..9a975bbc60 100644
--- a/src/glut/glx/Makefile
+++ b/src/glut/glx/Makefile
@@ -56,6 +56,7 @@ SOURCES = \
 	glut_modifier.c \
 	glut_mroman.c \
 	glut_overlay.c \
+	glut_ppm.c \
 	glut_roman.c \
 	glut_shapes.c \
 	glut_space.c \
@@ -103,7 +104,10 @@ pcedit = sed \
 	-e 's,@INSTALL_DIR@,$(INSTALL_DIR),' \
 	-e 's,@INSTALL_LIB_DIR@,$(INSTALL_LIB_DIR),' \
 	-e 's,@INSTALL_INC_DIR@,$(INSTALL_INC_DIR),' \
-	-e 's,@VERSION@,$(GLUT_MAJOR).$(GLUT_MINOR).$(GLUT_TINY),'
+	-e 's,@VERSION@,$(GLUT_MAJOR).$(GLUT_MINOR).$(GLUT_TINY),' \
+	-e 's,@GLUT_PC_REQ_PRIV@,$(GLUT_PC_REQ_PRIV),' \
+	-e 's,@GLUT_PC_LIB_PRIV@,$(GLUT_PC_LIB_PRIV),' \
+	-e 's,@GLUT_PC_CFLAGS@,$(GLUT_PC_CFLAGS),'
 glut.pc: glut.pc.in
 	$(pcedit) $< > $@
 
@@ -127,6 +131,7 @@ clean:
 
 depend: $(SOURCES)
 	@ echo "running $(MKDEP)"
+	@ rm -f depend
 	@ touch depend
 	@ $(MKDEP) $(MKDEP_OPTIONS) -I$(TOP)/include $(SOURCES) \
 		$(X11_INCLUDES) > /dev/null 
diff --git a/src/glut/glx/SConscript b/src/glut/glx/SConscript
new file mode 100644
index 0000000000..938fec03df
--- /dev/null
+++ b/src/glut/glx/SConscript
@@ -0,0 +1,87 @@
+Import('*')
+
+env = env.Clone()
+
+if env['platform'] != 'windows':
+    Return()
+
+target = 'glut32'
+
+env.Replace(CPPDEFINES = [
+    'BUILD_GLUT32', 
+    'GLUT_BUILDING_LIB', 
+    'NDEBUG',
+    'GLUT_NO_WARNING_DISABLE',
+])
+
+env.AppendUnique(CPPPATH = [
+    '#/include',
+])
+
+env.PrependUnique(LIBS = [
+    'winmm', 
+    'gdi32', 
+    'user32', 
+    'opengl32', 
+    'glu32',
+])
+
+sources = [
+    'glut_bitmap.c',
+    'glut_bwidth.c',
+    'glut_cindex.c',
+    'glut_cmap.c',
+    'glut_cursor.c',
+    'glut_dials.c',
+    'glut_dstr.c',
+    'glut_event.c',
+    'glut_ext.c',
+    'glut_fullscrn.c',
+    'glut_gamemode.c',
+    'glut_get.c',
+    'glut_init.c',
+    'glut_input.c',
+    'glut_joy.c',
+    'glut_key.c',
+    'glut_keyctrl.c',
+    'glut_keyup.c',
+    'glut_mesa.c',
+    'glut_modifier.c',
+    'glut_overlay.c',
+    'glut_ppm.c',
+    'glut_shapes.c',
+    'glut_space.c',
+    'glut_stroke.c',
+    'glut_swap.c',
+    'glut_swidth.c',
+    'glut_tablet.c',
+    'glut_teapot.c',
+    'glut_util.c',
+    'glut_vidresize.c',
+    'glut_warp.c',
+    'glut_win.c',
+    'glut_winmisc.c',
+
+    'win32_glx.c',
+    'win32_menu.c',
+    'win32_util.c',
+    'win32_winproc.c',
+    'win32_x11.c',
+    
+    'glut_8x13.c',
+    'glut_9x15.c',
+    'glut_hel10.c',
+    'glut_hel12.c',
+    'glut_hel18.c',
+    'glut_mroman.c',
+    'glut_roman.c',
+    'glut_tr10.c',
+    'glut_tr24.c',
+
+    'glut.def',
+]
+
+env.SharedLibrary(
+    target = target,
+    source = sources,
+)
diff --git a/src/glut/glx/descrip.mms b/src/glut/glx/descrip.mms
deleted file mode 100644
index 2e858309b6..0000000000
--- a/src/glut/glx/descrip.mms
+++ /dev/null
@@ -1,208 +0,0 @@
-# Makefile for GLUT for VMS
-# contributed by Jouk Jansen  joukj@hrem.stm.tudelft.nl
-
-.first
-	define gl [---.include.gl]
-
-.include [---]mms-config.
-
-##### MACROS #####
-GLUT_MAJOR = 3
-GLUT_MINOR = 7
-
-VPATH = RCS
-
-INCDIR = [---.include]
-LIBDIR = [---.lib]
-CFLAGS = /nowarn/include=$(INCDIR)/prefix=all/name=(as_is,short)/float=ieee/ieee=denorm
-
-SOURCES = \
-glut_8x13.c \
-glut_9x15.c \
-glut_bitmap.c \
-glut_bwidth.c \
-glut_cindex.c \
-glut_cmap.c \
-glut_cursor.c \
-glut_dials.c \
-glut_dstr.c \
-glut_event.c \
-glut_ext.c \
-glut_fullscrn.c \
-glut_gamemode.c \
-glut_get.c \
-glut_glxext.c \
-glut_hel10.c \
-glut_hel12.c \
-glut_hel18.c \
-glut_init.c \
-glut_input.c \
-glut_joy.c \
-glut_key.c \
-glut_keyctrl.c \
-glut_keyup.c \
-glut_menu.c \
-glut_menu2.c \
-glut_mesa.c \
-glut_modifier.c \
-glut_mroman.c \
-glut_overlay.c \
-glut_roman.c \
-glut_shapes.c \
-glut_space.c \
-glut_stroke.c \
-glut_swap.c \
-glut_swidth.c \
-glut_tablet.c \
-glut_teapot.c \
-glut_tr10.c \
-glut_tr24.c \
-glut_util.c \
-glut_vidresize.c \
-glut_warp.c \
-glut_win.c \
-glut_winmisc.c \
-layerutil.c
-
-OBJECTS0=glut_8x13.obj,\
-glut_9x15.obj,\
-glut_bitmap.obj,\
-glut_bwidth.obj,\
-glut_cindex.obj,\
-glut_cmap.obj,\
-glut_cursor.obj,\
-glut_dials.obj,\
-glut_dstr.obj,\
-glut_event.obj,\
-glut_ext.obj,\
-glut_fullscrn.obj,\
-glut_gamemode.obj
-
-OBJECTS1=glut_get.obj,\
-glut_glxext.obj,\
-glut_hel10.obj,\
-glut_hel12.obj,\
-glut_hel18.obj,\
-glut_init.obj,\
-glut_input.obj,\
-glut_joy.obj,\
-glut_key.obj,\
-glut_keyctrl.obj,\
-glut_keyup.obj,\
-glut_menu.obj,\
-glut_menu2.obj,\
-glut_mesa.obj,\
-glut_modifier.obj
-
-OBJECTS2=glut_mroman.obj,\
-glut_overlay.obj,\
-glut_roman.obj,\
-glut_shapes.obj,\
-glut_space.obj,\
-glut_stroke.obj,\
-glut_swap.obj,\
-glut_swidth.obj,\
-glut_tablet.obj,\
-glut_teapot.obj,\
-glut_tr10.obj,\
-glut_tr24.obj,\
-glut_util.obj,\
-glut_vidresize.obj
-
-OBJECTS3=glut_warp.obj,\
-glut_win.obj,\
-glut_winmisc.obj,\
-layerutil.obj
-
-##### RULES #####
-
-VERSION=Glut V3.7
-
-##### TARGETS #####
-
-# Make the library
-$(LIBDIR)$(GLUT_LIB) : $(OBJECTS0) $(OBJECTS1) $(OBJECTS2) $(OBJECTS3)
-  @ $(MAKELIB) $(GLUT_LIB) $(OBJECTS0)
-  @ library $(GLUT_LIB) $(OBJECTS1)
-  @ library $(GLUT_LIB) $(OBJECTS2)
-  @ library $(GLUT_LIB) $(OBJECTS3)
-  @ rename $(GLUT_LIB)* $(LIBDIR)
-.ifdef SHARE
-  @ WRITE_ SYS$OUTPUT "  generating mesagl1.opt"
-  @ OPEN_/WRITE FILE  mesagl1.opt
-  @ WRITE_ FILE "!"
-  @ WRITE_ FILE "! mesagl1.opt generated by DESCRIP.$(MMS_EXT)" 
-  @ WRITE_ FILE "!"
-  @ WRITE_ FILE "IDENTIFICATION=""$(VERSION)"""
-  @ WRITE_ FILE "GSMATCH=LEQUAL,3,7
-  @ WRITE_ FILE "$(OBJECTS0)"
-  @ WRITE_ FILE "$(OBJECTS1)"
-  @ WRITE_ FILE "$(OBJECTS2)"
-  @ WRITE_ FILE "$(OBJECTS3)"
-  @ WRITE_ FILE "[---.lib]libmesaglu.exe/SHARE"
-  @ WRITE_ FILE "[---.lib]libmesagl.exe/SHARE"
-  @ write file "sys$library:decw$xmulibshr.exe/share"
-  @ WRITE_ FILE "SYS$SHARE:DECW$XEXTLIBSHR/SHARE"
-  @ WRITE_ FILE "SYS$SHARE:DECW$XLIBSHR/SHARE"
-  @ CLOSE_ FILE
-  @ WRITE_ SYS$OUTPUT "  generating mesagl.map ..."
-  @ CXXLINK_/NODEB/NOSHARE/NOEXE/MAP=mesagl.map/FULL mesagl1.opt/OPT
-  @ WRITE_ SYS$OUTPUT "  analyzing mesagl.map ..."
-  @ @[---.vms]ANALYZE_MAP.COM mesagl.map mesagl.opt
-  @ WRITE_ SYS$OUTPUT "  linking $(GLUT_SHAR) ..."
-  @ CXXLINK_/NODEB/SHARE=$(GLUT_SHAR)/MAP=mesagl.map/FULL mesagl1.opt/opt,mesagl.opt/opt
-  @ rename $(GLUT_SHAR)* $(LIBDIR)
-.endif
-
-clean :
-	delete *.obj;*
-	purge
-
-include mms_depend.
-
-glut_8x13.obj : glut_8x13.c
-glut_9x15.obj : glut_9x15.c
-glut_bitmap.obj : glut_bitmap.c
-glut_bwidth.obj : glut_bwidth.c
-glut_cindex.obj : glut_cindex.c
-glut_cmap.obj : glut_cmap.c
-glut_cursor.obj : glut_cursor.c
-glut_dials.obj : glut_dials.c
-glut_dstr.obj : glut_dstr.c
-glut_event.obj : glut_event.c
-glut_ext.obj : glut_ext.c
-glut_fullscrn.obj : glut_fullscrn.c
-glut_gamemode.obj : glut_gamemode.c
-glut_get.obj : glut_get.c
-glut_glxext.obj : glut_glxext.c
-glut_hel10.obj : glut_hel10.c
-glut_hel12.obj : glut_hel12.c
-glut_hel18.obj : glut_hel18.c
-glut_init.obj : glut_init.c
-glut_input.obj : glut_input.c
-glut_joy.obj : glut_joy.c
-glut_key.obj : glut_key.c
-glut_keyctrl.obj : glut_keyctrl.c
-glut_keyup.obj : glut_keyup.c
-glut_menu.obj : glut_menu.c
-glut_menu2.obj : glut_menu2.c
-glut_mesa.obj : glut_mesa.c
-glut_modifier.obj : glut_modifier.c
-glut_mroman.obj : glut_mroman.c
-glut_overlay.obj : glut_overlay.c
-glut_roman.obj : glut_roman.c
-glut_shapes.obj : glut_shapes.c
-glut_space.obj : glut_space.c
-glut_stroke.obj : glut_stroke.c
-glut_swap.obj : glut_swap.c
-glut_swidth.obj : glut_swidth.c
-glut_tablet.obj : glut_tablet.c
-glut_teapot.obj : glut_teapot.c
-glut_tr10.obj : glut_tr10.c
-glut_tr24.obj : glut_tr24.c
-glut_util.obj : glut_util.c
-glut_vidresize.obj : glut_vidresize.c
-glut_warp.obj : glut_warp.c
-glut_win.obj : glut_win.c
-glut_winmisc.obj : glut_winmisc.c
-layerutil.obj : layerutil.c
diff --git a/src/glut/glx/glut.def b/src/glut/glx/glut.def
index d5e0e0cf2a..90010815e4 100644
--- a/src/glut/glx/glut.def
+++ b/src/glut/glx/glut.def
@@ -1,9 +1,4 @@
-DESCRIPTION 'OpenGL Utility Toolkit for Win32'
-
-VERSION 3.7
-
 EXPORTS
-
 	glutAddMenuEntry
 	glutAddSubMenu
 	glutAttachMenu
@@ -128,4 +123,3 @@ EXPORTS
 	glutBitmapTimesRoman24
 ;    __glutSetFCB
 ;    __glutGetFCB
-
diff --git a/src/glut/glx/glut.pc.in b/src/glut/glx/glut.pc.in
index e8638fe148..ae0689d7e8 100644
--- a/src/glut/glx/glut.pc.in
+++ b/src/glut/glx/glut.pc.in
@@ -6,6 +6,8 @@ includedir=@INSTALL_INC_DIR@
 Name: glut
 Description: Mesa OpenGL Utility Toolkit library
 Requires: gl glu
+Requires.private: @GLUT_PC_REQ_PRIV@
 Version: @VERSION@
 Libs: -L${libdir} -lglut
-Cflags: -I${includedir}
+Libs.private: @GLUT_PC_LIB_PRIV@
+Cflags: -I${includedir} @GLUT_PC_CFLAGS@
diff --git a/src/glut/glx/glut_event.c b/src/glut/glx/glut_event.c
index b5df7b2311..d6d963896a 100644
--- a/src/glut/glx/glut_event.c
+++ b/src/glut/glx/glut_event.c
@@ -1321,7 +1321,12 @@ processWindowWorkList(GLUTwindow * window)
        is where the finish works gets queued for indirect
        contexts. */
     __glutSetWindow(window);
-    glFinish();
+#if !defined(_WIN32)
+    if (!window->isDirect)
+#endif
+    {
+       glFinish();
+    }
   }
   if (workMask & GLUT_DEBUG_WORK) {
     __glutSetWindow(window);
@@ -1338,9 +1343,7 @@ processWindowWorkList(GLUTwindow * window)
   }
 }
 
-#ifndef _WIN32
 static  /* X11 implementations do not need this global. */
-#endif
 void
 __glutProcessWindowWorkLists(void)
 {
diff --git a/src/glut/glx/glut_init.c b/src/glut/glx/glut_init.c
index b1a42a328d..78843e93bc 100644
--- a/src/glut/glx/glut_init.c
+++ b/src/glut/glx/glut_init.c
@@ -51,6 +51,7 @@ int __glutInitX = -1, __glutInitY = -1;
 GLboolean __glutForceDirect = GL_FALSE,
   __glutTryDirect = GL_TRUE;
 Atom __glutWMDeleteWindow;
+char *__glutPPMFile = NULL;
 /* *INDENT-ON* */
 
 #ifdef _WIN32
@@ -341,6 +342,9 @@ glutInit(int *argcp, char **argv)
            __glutFPS = 5000;  /* 5000 milliseconds */
      }
   }
+
+  /* check if GLUT_PPM_FILE env var is set */
+  __glutPPMFile = getenv("GLUT_PPM_FILE");
 }
 
 #ifdef _WIN32
diff --git a/src/glut/glx/glut_input.c b/src/glut/glx/glut_input.c
index a76ff9a435..7d2eda00f7 100644
--- a/src/glut/glx/glut_input.c
+++ b/src/glut/glx/glut_input.c
@@ -295,6 +295,7 @@ __glutProcessDeviceEvents(XEvent * event)
 static GLUTeventParser eventParser =
 {__glutProcessDeviceEvents, NULL};
 
+#if !defined(_WIN32)
 static void
 addDeviceEventParser(void)
 {
@@ -305,6 +306,7 @@ addDeviceEventParser(void)
   been_here = True;
   __glutRegisterEventParser(&eventParser);
 }
+#endif
 
 static int
 probeDevices(void)
diff --git a/src/glut/glx/glut_ppm.c b/src/glut/glx/glut_ppm.c
new file mode 100644
index 0000000000..49dca8532f
--- /dev/null
+++ b/src/glut/glx/glut_ppm.c
@@ -0,0 +1,80 @@
+/*
+ * PPM file output
+ * Brian Paul
+ * 8 Dec 2008
+ */
+
+#include <assert.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include "glutint.h"
+
+
+static void
+write_ppm_file(const char *filename, const GLubyte *buffer,
+               int width, int height)
+{
+   const int binary = 1;
+   FILE *f = fopen( filename, "w" );
+   if (f) {
+      const GLubyte *ptr = buffer;
+      int i, x, y;
+      if (binary) {
+         fprintf(f,"P6\n");
+         fprintf(f,"# ppm-file created by GLUT\n");
+         fprintf(f,"%i %i\n", width, height);
+         fprintf(f,"255\n");
+         fclose(f);
+         f = fopen( filename, "ab" );  /* reopen in binary append mode */
+         for (y = height - 1; y >= 0; y--) {
+            for (x = 0; x < width; x++) {
+               i = (y * width + x) * 4;
+               fputc(ptr[i], f);   /* write red */
+               fputc(ptr[i+1], f); /* write green */
+               fputc(ptr[i+2], f); /* write blue */
+            }
+         }
+      }
+      else {
+         /*ASCII*/
+         int counter = 0;
+         fprintf(f,"P3\n");
+         fprintf(f,"# ascii ppm file created by GLUT\n");
+         fprintf(f,"%i %i\n", width, height);
+         fprintf(f,"255\n");
+         for (y = height - 1; y >= 0; y--) {
+            for (x = 0; x < width; x++) {
+               i = (y * width + x) * 4;
+               fprintf(f, " %3d %3d %3d", ptr[i], ptr[i+1], ptr[i+2]);
+               counter++;
+               if (counter % 5 == 0)
+                  fprintf(f, "\n");
+            }
+         }
+      }
+      fclose(f);
+   }
+}
+
+
+/**
+ * Called from SwapBuffers if the GLUT_PPM_FILE env var is set.
+ */
+void __glutWritePPMFile(void)
+{
+   int w = glutGet(GLUT_WINDOW_WIDTH);
+   int h = glutGet(GLUT_WINDOW_HEIGHT);
+   GLubyte *buf;
+
+   assert(__glutPPMFile);
+
+   buf = (GLubyte *) malloc(w * h * 4);
+   if (buf) {
+      /* XXX save/restore pixel packing */
+      glReadPixels(0, 0, w, h, GL_RGBA, GL_UNSIGNED_BYTE, buf);
+      write_ppm_file(__glutPPMFile, buf, w, h);
+      free(buf);
+   }
+
+   __glutPPMFile = NULL; /* only write one file */
+}
diff --git a/src/glut/glx/glut_swap.c b/src/glut/glx/glut_swap.c
index 4831b8eb35..07a2da88a2 100644
--- a/src/glut/glx/glut_swap.c
+++ b/src/glut/glx/glut_swap.c
@@ -18,6 +18,10 @@ glutSwapBuffers(void)
 {
   GLUTwindow *window = __glutCurrentWindow;
 
+  if (__glutPPMFile) {
+     __glutWritePPMFile();
+  }
+
   if (window->renderWin == window->win) {
     if (__glutCurrentWindow->treatAsSingle) {
       /* Pretend the double buffered window is single buffered,
diff --git a/src/glut/glx/glutint.h b/src/glut/glx/glutint.h
index a962c78023..ba30af60f0 100644
--- a/src/glut/glx/glutint.h
+++ b/src/glut/glx/glutint.h
@@ -32,10 +32,6 @@
 
 #include <GL/glut.h>
 
-#if defined(MESA) && defined(_WIN32) && !defined(__CYGWIN32__)
-#include <gl/mesa_wgl.h>
-#endif
-
 #ifndef _WIN32
 /* added by BrianP: */
 #ifndef APIENTRY
@@ -48,24 +44,6 @@
 /* GLUT_BUILDING_LIB is used by <GL/glut.h> to 1) not #pragma link
    with the GLUT library, and 2) avoid the Win32 atexit hack. */
 
-/* This must be done after <GL/gl.h> is included.  MESA is defined
-   if the <GL/gl.h> is supplied by Brian Paul's Mesa library. */
-#if defined(MESA) && defined(_WIN32)
-/* Mesa implements "wgl" versions of GDI entry points needed for
-   using OpenGL.  Map these "wgl" versions to the GDI names via
-   macros. */
-#define ChoosePixelFormat   wglChoosePixelFormat
-#define DescribePixelFormat wglDescribePixelFormat
-#define GetPixelFormat      wglGetPixelFormat
-#define SetPixelFormat      wglSetPixelFormat
-#define SwapBuffers         wglSwapBuffers
-#define GetCurrentContext   wglGetCurrentContext
-#define GetCurrentDC        wglGetCurrentDC
-#define MakeCurrent         wglMakeCurrent
-#define CreateContext       wglCreateContext
-#define DeleteContext       wglDeleteContext
-#endif /* MESA */
-
 #ifdef SUPPORT_FORTRAN
 #include <GL/glutf90.h>
 #endif
@@ -572,27 +550,27 @@ typedef struct {
 #ifdef _WIN32
 #define MAKE_CURRENT_LAYER(window)                                    \
   {                                                                   \
-    HGLRC currentContext = GetCurrentContext();                       \
-    HDC currentDc = GetCurrentDC();                                   \
+    HGLRC currentContext = wglGetCurrentContext();                    \
+    HDC currentDc = wglGetCurrentDC();                                \
                                                                       \
     if (currentContext != window->renderCtx                           \
       || currentDc != window->renderDc) {                             \
-      MakeCurrent(window->renderDc, window->renderCtx);               \
+      wglMakeCurrent(window->renderDc, window->renderCtx);            \
     }                                                                 \
   }
 #define MAKE_CURRENT_WINDOW(window)                                   \
   {                                                                   \
-    HGLRC currentContext = GetCurrentContext();                       \
-    HDC currentDc = GetCurrentDC();                                   \
+    HGLRC currentContext = wglGetCurrentContext();                    \
+    HDC currentDc = wglGetCurrentDC();                                \
                                                                       \
     if (currentContext != window->ctx || currentDc != window->hdc) {  \
-      MakeCurrent(window->hdc, window->ctx);                          \
+      wglMakeCurrent(window->hdc, window->ctx);                       \
     }                                                                 \
   }
 #define MAKE_CURRENT_OVERLAY(overlay) \
-  MakeCurrent(overlay->hdc, overlay->ctx)
+  wglMakeCurrent(overlay->hdc, overlay->ctx)
 #define UNMAKE_CURRENT() \
-  MakeCurrent(NULL, NULL)
+  wglMakeCurrent(NULL, NULL)
 #define SWAP_BUFFERS_WINDOW(window) \
   SwapBuffers(window->hdc)
 #define SWAP_BUFFERS_LAYER(window) \
@@ -664,6 +642,7 @@ extern unsigned int __glutModifierMask;
 #ifdef _WIN32
 extern void (__cdecl *__glutExitFunc)(int retval);
 #endif
+extern char *__glutPPMFile;
 
 /* private variables from glut_menu.c */
 extern GLUTmenuItem *__glutItemSelected;
@@ -684,6 +663,9 @@ extern void __glutFreeOverlay(GLUToverlay * overlay);
 extern XVisualInfo *__glutDetermineWindowVisual(Bool * treatAsSingle,
   Bool * visAlloced, void **fbc);
 
+/* private variables from glut_ppm.c */
+extern void __glutWritePPMFile(void);
+
 /* private variables from glut_mesa.c */
 extern int __glutMesaSwapHackSupport;
 
@@ -812,4 +794,5 @@ extern LONG WINAPI __glutWindowProc(HWND win, UINT msg, WPARAM w, LPARAM l);
 extern HDC XHDC;
 #endif
 
+
 #endif /* __glutint_h__ */
diff --git a/src/glut/glx/glutstroke.h b/src/glut/glx/glutstroke.h
index cbc9e156fe..74b60314aa 100644
--- a/src/glut/glx/glutstroke.h
+++ b/src/glut/glx/glutstroke.h
@@ -7,7 +7,7 @@
    and is provided without guarantee or warrantee expressed or 
    implied. This program is -not- in the public domain. */
 
-#if defined(_WIN32)
+#if defined(_MSC_VER)
 #pragma warning (disable:4244)  /* disable bogus conversion warnings */
 #pragma warning (disable:4305)  /* VC++ 5.0 version of above warning. */
 #endif
diff --git a/src/glut/glx/glutwin32.h b/src/glut/glx/glutwin32.h
index 304eeb20a2..01db99fad6 100644
--- a/src/glut/glx/glutwin32.h
+++ b/src/glut/glx/glutwin32.h
@@ -41,6 +41,7 @@
 
 /* Private routines from win32_util.c */
 #ifndef __CYGWIN32__
+struct timeval;
 extern int gettimeofday(struct timeval* tp, void* tzp);
 #endif
 extern void *__glutFont(void *font);
diff --git a/src/glut/glx/mms_depend b/src/glut/glx/mms_depend
deleted file mode 100644
index 98f87c29e2..0000000000
--- a/src/glut/glx/mms_depend
+++ /dev/null
@@ -1,72 +0,0 @@
-# DO NOT DELETE
-
-glut_8x13.obj : glutbitmap.h [---.include.gl]gl.h
-glut_9x15.obj : glutbitmap.h [---.include.gl]gl.h
-glut_bitmap.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_bitmap.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_bitmap.obj : glutbitmap.h
-glut_bwidth.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_bwidth.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_bwidth.obj : glutbitmap.h
-glut_cindex.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_cindex.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_cindex.obj : layerutil.h
-glut_cursor.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_cursor.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_dials.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h
-glut_dials.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h [---.include.gl]glu.h
-glut_dstr.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h
-glut_dstr.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h [---.include.gl]glu.h
-glut_event.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_event.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_ext.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_ext.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_fullscrn.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h
-glut_fullscrn.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h
-glut_fullscrn.obj : [---.include.gl]glu.h
-glut_get.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_get.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_hel10.obj : glutbitmap.h [---.include.gl]gl.h
-glut_hel12.obj : glutbitmap.h [---.include.gl]gl.h
-glut_hel18.obj : glutbitmap.h [---.include.gl]gl.h
-glut_init.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_init.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_menu.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_menu.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h layerutil.h
-glut_mesa.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h
-glut_mesa.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h [---.include.gl]glu.h
-glut_modifier.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h
-glut_modifier.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h
-glut_modifier.obj : [---.include.gl]glu.h
-glut_mroman.obj : glutstroke.h
-glut_overlay.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_overlay.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_overlay.obj : layerutil.h
-glut_roman.obj : glutstroke.h
-glut_shapes.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_shapes.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_space.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h
-glut_space.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h [---.include.gl]glu.h
-glut_stroke.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_stroke.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_stroke.obj : glutstroke.h
-glut_swidth.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_swidth.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_swidth.obj : glutstroke.h
-glut_tablet.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h
-glut_tablet.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h [---.include.gl]glu.h
-glut_teapot.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_tr10.obj : glutbitmap.h [---.include.gl]gl.h
-glut_tr24.obj : glutbitmap.h [---.include.gl]gl.h
-glut_util.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_util.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_vidresize.obj : [---.include.gl]glx.h [---.include.gl]gl.h
-glut_vidresize.obj : [---.include.gl]xmesa.h glutint.h [---.include.gl]glut.h
-glut_vidresize.obj : [---.include.gl]glu.h
-glut_warp.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_warp.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_win.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_win.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-glut_winmisc.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h
-glut_winmisc.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h
-layerutil.obj : layerutil.h
diff --git a/src/glut/glx/win32_glx.c b/src/glut/glx/win32_glx.c
index ecfff44b93..3522e606ec 100644
--- a/src/glut/glx/win32_glx.c
+++ b/src/glut/glx/win32_glx.c
@@ -22,7 +22,7 @@ glXCreateContext(Display * display, XVisualInfo * visinfo,
      routine. */
   HGLRC context;
 
-  context = CreateContext(XHDC);
+  context = wglCreateContext(XHDC);
 
 #if 0
   /* XXX GLUT doesn't support it now, so don't worry about display list
diff --git a/src/glut/glx/win32_winproc.c b/src/glut/glx/win32_winproc.c
index e1fc785ebb..4a9dc3781c 100644
--- a/src/glut/glx/win32_winproc.c
+++ b/src/glut/glx/win32_winproc.c
@@ -20,10 +20,10 @@
 extern unsigned __glutMenuButton;
 extern GLUTidleCB __glutIdleFunc;
 extern GLUTtimer *__glutTimerList;
-extern GLUTmenuItem *__glutGetUniqueMenuItem(GLUTmenu * menu, int unique);
+extern GLUTmenuItem *__glutGetUniqueMenuItem(GLUTmenu * menu, UINT unique);
 static HMENU __glutHMenu;
 
-void
+static void
 updateWindowState(GLUTwindow *window, int visState)
 {
   GLUTwindow* child;
diff --git a/src/glut/glx/win32_x11.h b/src/glut/glx/win32_x11.h
index 6f5c3a9aea..daf065e27b 100644
--- a/src/glut/glx/win32_x11.h
+++ b/src/glut/glx/win32_x11.h
@@ -228,7 +228,7 @@ typedef struct {
   SetWindowPos(window, HWND_BOTTOM, 0, 0, 0, 0, SWP_NOMOVE | SWP_NOSIZE)
 
 #define XSetWMName(display, window, tp) \
-  SetWindowText(window, (tp)->value)
+  SetWindowText(window, (const char *)(tp)->value)
 
 /* There really isn't a way to set the icon name separate from the
    windows name in Win32, so, just set the windows name. */
diff --git a/src/glut/mini/Makefile b/src/glut/mini/Makefile
index e47d09edb7..841a473f89 100644
--- a/src/glut/mini/Makefile
+++ b/src/glut/mini/Makefile
@@ -64,6 +64,7 @@ $(TOP)/$(LIB_DIR)/$(GLUT_LIB_NAME): depend $(OBJECTS)
 # Run 'make -f Makefile.solo dep' to update the dependencies if you change
 # what's included by any source file.
 depend: $(SOURCES) 
+	rm -f depend
 	touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) > /dev/null 
 
@@ -77,7 +78,10 @@ pcedit = sed \
 	-e 's,@INSTALL_DIR@,$(INSTALL_DIR),' \
 	-e 's,@INSTALL_LIB_DIR@,$(INSTALL_LIB_DIR),' \
 	-e 's,@INSTALL_INC_DIR@,$(INSTALL_INC_DIR),' \
-	-e 's,@VERSION@,$(GLUT_MAJOR).$(GLUT_MINOR).$(GLUT_TINY),'
+	-e 's,@VERSION@,$(GLUT_MAJOR).$(GLUT_MINOR).$(GLUT_TINY),' \
+	-e 's,@GLUT_PC_REQ_PRIV@,$(GLUT_PC_REQ_PRIV),' \
+	-e 's,@GLUT_PC_LIB_PRIV@,$(GLUT_PC_LIB_PRIV),' \
+	-e 's,@GLUT_PC_CFLAGS@,$(GLUT_PC_CFLAGS),'
 glut.pc: glut.pc.in
 	$(pcedit) $< > $@
 
diff --git a/src/glut/mini/glut.pc.in b/src/glut/mini/glut.pc.in
index e8638fe148..ae0689d7e8 100644
--- a/src/glut/mini/glut.pc.in
+++ b/src/glut/mini/glut.pc.in
@@ -6,6 +6,8 @@ includedir=@INSTALL_INC_DIR@
 Name: glut
 Description: Mesa OpenGL Utility Toolkit library
 Requires: gl glu
+Requires.private: @GLUT_PC_REQ_PRIV@
 Version: @VERSION@
 Libs: -L${libdir} -lglut
-Cflags: -I${includedir}
+Libs.private: @GLUT_PC_LIB_PRIV@
+Cflags: -I${includedir} @GLUT_PC_CFLAGS@
diff --git a/src/glw/Makefile b/src/glw/Makefile
index 753c4b74d4..b153a6df75 100644
--- a/src/glw/Makefile
+++ b/src/glw/Makefile
@@ -30,7 +30,10 @@ pcedit = sed \
 	-e 's,@INSTALL_DIR@,$(INSTALL_DIR),' \
 	-e 's,@INSTALL_LIB_DIR@,$(INSTALL_LIB_DIR),' \
 	-e 's,@INSTALL_INC_DIR@,$(INSTALL_INC_DIR),' \
-	-e 's,@VERSION@,$(MAJOR).$(MINOR).$(TINY),'
+	-e 's,@VERSION@,$(MAJOR).$(MINOR).$(TINY),' \
+	-e 's,@GLW_PC_REQ_PRIV@,$(GLW_PC_REQ_PRIV),' \
+	-e 's,@GLW_PC_LIB_PRIV@,$(GLW_PC_LIB_PRIV),' \
+	-e 's,@GLW_PC_CFLAGS@,$(GLW_PC_CFLAGS),'
 glw.pc: glw.pc.in
 	$(pcedit) $< > $@
 
@@ -61,6 +64,7 @@ $(TOP)/$(LIB_DIR)/$(GLW_LIB_NAME): $(OBJECTS)
 # by any source file.
 # 
 depend: $(GLW_SOURCES)
+	rm -f depend
 	touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) -I$(TOP)/include $(GLW_SOURCES) \
 		$(X11_INCLUDES) > /dev/null
diff --git a/src/glw/glw.pc.in b/src/glw/glw.pc.in
index 25f3f73b8c..5493093be1 100644
--- a/src/glw/glw.pc.in
+++ b/src/glw/glw.pc.in
@@ -6,6 +6,8 @@ includedir=@INSTALL_INC_DIR@
 Name: glw
 Description: Mesa OpenGL widget library
 Requires: gl
+Requires.private: @GLW_PC_REQ_PRIV@
 Version: @VERSION@
 Libs: -L${libdir} -lGLw
-Cflags: -I${includedir}
+Libs.private: @GLW_PC_LIB_PRIV@
+Cflags: -I${includedir} @GLW_PC_CFLAGS@
diff --git a/src/glx/mini/Makefile b/src/glx/mini/Makefile
index aabc1d2be3..6b5a3c76d7 100644
--- a/src/glx/mini/Makefile
+++ b/src/glx/mini/Makefile
@@ -65,6 +65,7 @@ drmtest: xf86drm.o drmtest.o
 
 
 depend: $(C_SOURCES) $(ASM_SOURCES)
+	rm -f depend
 	touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(C_SOURCES) $(ASM_SOURCES) \
 		> /dev/null 
diff --git a/src/glx/x11/Makefile b/src/glx/x11/Makefile
index 419cb2853b..86d84d4b9f 100644
--- a/src/glx/x11/Makefile
+++ b/src/glx/x11/Makefile
@@ -39,9 +39,9 @@ SOURCES = \
 	  dri2_glx.c \
 	  dri2.c
 
-include $(TOP)/src/mesa/sources
+include $(TOP)/src/mesa/sources.mak
 
-MESA_ASM_API = $(addprefix $(TOP)/src/mesa/, $(ASM_API))
+MESA_GLAPI_ASM_SOURCES = $(addprefix $(TOP)/src/mesa/, $(GLAPI_ASM_SOURCES))
 MESA_GLAPI_SOURCES = $(addprefix $(TOP)/src/mesa/, $(GLAPI_SOURCES))
 MESA_GLAPI_OBJECTS = $(addprefix $(TOP)/src/mesa/, $(GLAPI_OBJECTS))
 
@@ -77,10 +77,11 @@ $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME):  $(OBJECTS) Makefile
 		$(GL_LIB_DEPS) $(OBJECTS)
 
 
-depend: $(SOURCES) $(MESA_GLAPI_SOURCES) $(MESA_ASM_API) Makefile
+depend: $(SOURCES) $(MESA_GLAPI_SOURCES) $(MESA_GLAPI_ASM_SOURCES) Makefile
+	rm -f depend
 	touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) \
-		$(MESA_GLAPI_SOURCES) $(MESA_ASM_API) 
+		$(MESA_GLAPI_SOURCES) $(MESA_GLAPI_ASM_SOURCES) 
 
 
 # Emacs tags
diff --git a/src/glx/x11/dri2.c b/src/glx/x11/dri2.c
index 3b48cd9c1c..f967432b99 100644
--- a/src/glx/x11/dri2.c
+++ b/src/glx/x11/dri2.c
@@ -253,7 +253,7 @@ DRI2Buffer *DRI2GetBuffers(Display *dpy, XID drawable,
     *height = rep.height;
     *outCount = rep.count;
 
-    buffers = Xmalloc(count * sizeof buffers[0]);
+    buffers = Xmalloc(rep.count * sizeof buffers[0]);
     if (buffers == NULL) {
 	_XEatData(dpy, rep.count * sizeof repBuffer);
 	UnlockDisplay(dpy);
diff --git a/src/glx/x11/dri2_glx.c b/src/glx/x11/dri2_glx.c
index 2bee67780b..b878f05dda 100644
--- a/src/glx/x11/dri2_glx.c
+++ b/src/glx/x11/dri2_glx.c
@@ -60,6 +60,9 @@ struct __GLXDRIdisplayPrivateRec {
     int driMajor;
     int driMinor;
     int driPatch;
+
+    unsigned long configureSeqno;
+    Bool (*oldConfigProc)(Display *, XEvent *, xEvent *);
 };
 
 struct __GLXDRIcontextPrivateRec {
@@ -73,6 +76,7 @@ struct __GLXDRIdrawablePrivateRec {
     __DRIbuffer buffers[5];
     int bufferCount;
     int width, height;
+    unsigned long configureSeqno;
 };
 
 static void dri2DestroyContext(__GLXDRIcontext *context,
@@ -166,6 +170,8 @@ static __GLXDRIdrawable *dri2CreateDrawable(__GLXscreenConfigs *psc,
     pdraw->base.xDrawable = xDrawable;
     pdraw->base.drawable = drawable;
     pdraw->base.psc = psc;
+    pdraw->bufferCount = 0;
+    pdraw->configureSeqno = ~0;
 
     DRI2CreateDrawable(psc->dpy, xDrawable);
 
@@ -223,9 +229,30 @@ dri2GetBuffers(__DRIdrawable *driDrawable,
 	       int *out_count, void *loaderPrivate)
 {
     __GLXDRIdrawablePrivate *pdraw = loaderPrivate;
+    __GLXdisplayPrivate *dpyPriv = __glXInitialize(pdraw->base.psc->dpy);
+    __GLXDRIdisplayPrivate *pdp = (__GLXDRIdisplayPrivate *)dpyPriv->dri2Display;
     DRI2Buffer *buffers;
     int i;
 
+    /**
+     * Check if a ConfigureNotify has come in since we last asked for the
+     * buffers associated with this drawable.  If not, we can assume that they're
+     * the same set at glViewport time, and save a synchronous round-trip to the
+     * X Server.
+     */
+    if (pdraw->configureSeqno == pdp->configureSeqno &&
+	count == pdraw->bufferCount) {
+	for (i = 0; i < count; i++) {
+	    if (pdraw->buffers[i].attachment != attachments[i])
+		break;
+	}
+	if (i == count) {
+	    *out_count = pdraw->bufferCount;
+	    return pdraw->buffers;
+	}
+    }
+    pdraw->configureSeqno = pdp->configureSeqno;
+
     buffers = DRI2GetBuffers(pdraw->base.psc->dpy, pdraw->base.xDrawable,
 			     width, height, attachments, count, out_count);
     if (buffers == NULL)
@@ -233,6 +260,7 @@ dri2GetBuffers(__DRIdrawable *driDrawable,
 
     pdraw->width = *width;
     pdraw->height = *height;
+    pdraw->bufferCount = *out_count;
 
     /* This assumes the DRI2 buffer attachment tokens matches the
      * __DRIbuffer tokens. */
@@ -282,8 +310,10 @@ static __GLXDRIscreen *dri2CreateScreen(__GLXscreenConfigs *psc, int screen,
 	return NULL;
 
     psc->driver = driOpenDriver(driverName);
-    if (psc->driver == NULL)
+    if (psc->driver == NULL) {
+	ErrorMessageF("driver pointer missing\n");
 	goto handle_error;
+    }
 
     extensions = dlsym(psc->driver, __DRI_DRIVER_EXTENSIONS);
     if (extensions == NULL) {
@@ -309,11 +339,15 @@ static __GLXDRIscreen *dri2CreateScreen(__GLXscreenConfigs *psc, int screen,
 	return NULL;
     }
 
-    if (drmGetMagic(psc->fd, &magic))
+    if (drmGetMagic(psc->fd, &magic)) {
+	ErrorMessageF("failed to get magic\n");
 	return NULL;
+    }
 
-    if (!DRI2Authenticate(psc->dpy, RootWindow(psc->dpy, screen), magic))
+    if (!DRI2Authenticate(psc->dpy, RootWindow(psc->dpy, screen), magic)) {
+	ErrorMessageF("failed to authenticate magic %d\n", magic);
 	return NULL;
+    }
 
     psc->__driScreen = 
 	psc->dri2->createNewScreen(screen, psc->fd,
@@ -359,6 +393,28 @@ static void dri2DestroyDisplay(__GLXDRIdisplay *dpy)
     Xfree(dpy);
 }
 
+/**
+ * Makes a note on receiving ConfigureNotify that we need to re-check the
+ * DRI2 buffers, as window sizes may have resulted in reallocation.
+ */
+static Bool dri2ConfigureNotifyProc(Display *dpy, XEvent *re, xEvent *event)
+{
+    __GLXdisplayPrivate *dpyPriv = __glXInitialize(dpy);
+    __GLXDRIdisplayPrivate *pdp;
+    Bool ret;
+
+    /* We should always be able to find our pdp, as it only gets torn down
+     * when the Display is torn down.
+     */
+    pdp = (__GLXDRIdisplayPrivate *)dpyPriv->dri2Display;
+
+    ret = pdp->oldConfigProc(dpy, re, event);
+
+    pdp->configureSeqno = re->xconfigure.serial;
+
+    return ret;
+}
+
 /*
  * Allocate, initialize and return a __DRIdisplayPrivate object.
  * This is called from __glXInitialize() when we are given a new
@@ -381,7 +437,11 @@ _X_HIDDEN __GLXDRIdisplay *dri2CreateDisplay(Display *dpy)
 	return NULL;
     }
 
+    pdp->oldConfigProc = XESetWireToEvent(dpy, ConfigureNotify,
+					  dri2ConfigureNotifyProc);
+
     pdp->driPatch = 0;
+    pdp->configureSeqno = 0;
 
     pdp->base.destroyDisplay = dri2DestroyDisplay;
     pdp->base.createScreen = dri2CreateScreen;
diff --git a/src/glx/x11/drisw_glx.c b/src/glx/x11/drisw_glx.c
index fee45952b3..35bbd9151c 100644
--- a/src/glx/x11/drisw_glx.c
+++ b/src/glx/x11/drisw_glx.c
@@ -113,7 +113,7 @@ swrastGetDrawableInfo(__DRIdrawable * draw,
                       int *x, int *y, int *w, int *h, void *loaderPrivate)
 {
    __GLXDRIdrawablePrivate *pdp = loaderPrivate;
-   __GLXDRIdrawable *pdraw = &(pdp->base);;
+   __GLXDRIdrawable *pdraw = &(pdp->base);
    Display *dpy = pdraw->psc->dpy;
    Drawable drawable;
 
@@ -141,7 +141,7 @@ swrastPutImage(__DRIdrawable * draw, int op,
                int x, int y, int w, int h, char *data, void *loaderPrivate)
 {
    __GLXDRIdrawablePrivate *pdp = loaderPrivate;
-   __GLXDRIdrawable *pdraw = &(pdp->base);;
+   __GLXDRIdrawable *pdraw = &(pdp->base);
    Display *dpy = pdraw->psc->dpy;
    Drawable drawable;
    XImage *ximage;
@@ -176,7 +176,7 @@ swrastGetImage(__DRIdrawable * draw,
                int x, int y, int w, int h, char *data, void *loaderPrivate)
 {
    __GLXDRIdrawablePrivate *pdp = loaderPrivate;
-   __GLXDRIdrawable *pdraw = &(pdp->base);;
+   __GLXDRIdrawable *pdraw = &(pdp->base);
    Display *dpy = pdraw->psc->dpy;
    Drawable drawable;
    XImage *ximage;
diff --git a/src/glx/x11/glx_pbuffer.c b/src/glx/x11/glx_pbuffer.c
index c63d53439d..a602cd2881 100644
--- a/src/glx/x11/glx_pbuffer.c
+++ b/src/glx/x11/glx_pbuffer.c
@@ -220,14 +220,14 @@ GetDrawableAttribute(Display * dpy, GLXDrawable drawable,
    unsigned int length;
    unsigned int i;
    unsigned int num_attributes;
+   GLboolean use_glx_1_3;
 
    if ((dpy == NULL) || (drawable == 0)) {
       return 0;
    }
 
    priv = __glXInitialize(dpy);
-   GLboolean use_glx_1_3 = ((priv->majorVersion > 1)
-                            || (priv->minorVersion >= 3));
+   use_glx_1_3 = ((priv->majorVersion > 1) || (priv->minorVersion >= 3));
 
    *value = 0;
 
diff --git a/src/glx/x11/glx_query.c b/src/glx/x11/glx_query.c
index 56f3495186..2789b841b1 100644
--- a/src/glx/x11/glx_query.c
+++ b/src/glx/x11/glx_query.c
@@ -31,6 +31,70 @@
 
 #include "glxclient.h"
 
+#if defined(USE_XCB)
+# include <X11/Xlib-xcb.h>
+# include <xcb/xcb.h>
+# include <xcb/glx.h>
+#endif
+
+#ifdef USE_XCB
+
+/**
+ * Exchange a protocol request for glXQueryServerString.
+ */
+char *
+__glXQueryServerString(Display* dpy,
+                       int opcode,
+                       CARD32 screen,
+                       CARD32 name)
+{
+   xcb_connection_t *c = XGetXCBConnection(dpy);
+   xcb_glx_query_server_string_reply_t* reply =
+      xcb_glx_query_server_string_reply(c,
+                                        xcb_glx_query_server_string(c,
+                                                                    screen,
+                                                                    name),
+                                        NULL);
+
+   /* The spec doesn't mention this, but the Xorg server replies with
+    * a string already terminated with '\0'. */
+   uint32_t len = xcb_glx_query_server_string_string_length(reply);
+   char* buf = Xmalloc(len);
+   memcpy(buf, xcb_glx_query_server_string_string(reply), len);
+   free(reply);
+
+   return buf;
+}
+
+/**
+ * Exchange a protocol request for glGetString.
+ */
+char *
+__glXGetString(Display* dpy,
+               int opcode,
+               CARD32 contextTag,
+               CARD32 name)
+{
+   xcb_connection_t *c = XGetXCBConnection(dpy);
+   xcb_glx_get_string_reply_t* reply =
+      xcb_glx_get_string_reply(c,
+                               xcb_glx_get_string(c,
+                                                  contextTag,
+                                                  name),
+                               NULL);
+
+   /* The spec doesn't mention this, but the Xorg server replies with
+    * a string already terminated with '\0'. */
+   uint32_t len = xcb_glx_get_string_string_length(reply);
+   char* buf = Xmalloc(len);
+   memcpy(buf, xcb_glx_get_string_string(reply), len);
+   free(reply);
+
+   return buf;
+}
+
+#else
+
 /**
  * GLX protocol structure for the ficticious "GXLGenericGetString" request.
  * 
@@ -57,7 +121,7 @@ typedef struct GLXGenericGetString
  * Query the Server GLX string.
  * This routine will allocate the necessay space for the string.
  */
-char *
+static char *
 __glXGetStringFromServer(Display * dpy, int opcode, CARD32 glxCode,
                          CARD32 for_whom, CARD32 name)
 {
@@ -101,3 +165,27 @@ __glXGetStringFromServer(Display * dpy, int opcode, CARD32 glxCode,
 
    return buf;
 }
+
+char *
+__glXQueryServerString(Display* dpy,
+                       int opcode,
+                       CARD32 screen,
+                       CARD32 name)
+{
+   return __glXGetStringFromServer(dpy, opcode,
+                                   X_GLXQueryServerString,
+                                   screen, name);
+}
+
+char *
+__glXGetString(Display* dpy,
+               int opcode,
+               CARD32 contextTag,
+               CARD32 name)
+{
+   return __glXGetStringFromServer(dpy, opcode, X_GLsop_GetString,
+                                   contextTag, name);
+}
+
+#endif /* USE_XCB */
+
diff --git a/src/glx/x11/glxclient.h b/src/glx/x11/glxclient.h
index 02e8750c97..9332eb64d3 100644
--- a/src/glx/x11/glxclient.h
+++ b/src/glx/x11/glxclient.h
@@ -602,8 +602,11 @@ extern void __glXSendLargeCommand(__GLXcontext *, const GLvoid *, GLint,
 				  const GLvoid *, GLint);
 
 /* Initialize the GLX extension for dpy */
+extern __GLXdisplayPrivate * __glXGetPrivateFromDisplay(Display *dpy);
 extern __GLXdisplayPrivate *__glXInitialize(Display*);
 
+extern void __glXPreferEGL(int state);
+
 /************************************************************************/
 
 extern int __glXDebug;
@@ -739,8 +742,10 @@ extern void _XSend(Display*, const void*, long);
 extern void __glXInitializeVisualConfigFromTags( __GLcontextModes *config,
     int count, const INT32 *bp, Bool tagged_only, Bool fbconfig_style_tags );
 
-extern char * __glXGetStringFromServer( Display * dpy, int opcode,
-    CARD32 glxCode, CARD32 for_whom, CARD32 name );
+extern char * __glXQueryServerString(Display* dpy, int opcode,
+                                     CARD32 screen, CARD32 name);
+extern char * __glXGetString(Display* dpy, int opcode,
+                             CARD32 screen, CARD32 name);
 
 extern char *__glXstrdup(const char *str);
 
diff --git a/src/glx/x11/glxcmds.c b/src/glx/x11/glxcmds.c
index 391e0be05e..c68b6ac4ef 100644
--- a/src/glx/x11/glxcmds.c
+++ b/src/glx/x11/glxcmds.c
@@ -44,6 +44,12 @@
 #include "xf86dri.h"
 #endif
 
+#if defined(USE_XCB)
+#include <X11/Xlib-xcb.h>
+#include <xcb/xcb.h>
+#include <xcb/glx.h>
+#endif
+
 static const char __glXGLXClientVendorName[] = "SGI";
 static const char __glXGLXClientVersion[] = "1.4";
 
@@ -749,8 +755,10 @@ PUBLIC void glXCopyContext(Display *dpy, GLXContext source,
  */
 static Bool __glXIsDirect(Display *dpy, GLXContextID contextID)
 {
+#if !defined(USE_XCB)
     xGLXIsDirectReq *req;
     xGLXIsDirectReply reply;
+#endif
     CARD8 opcode;
 
     opcode = __glXSetupForCommand(dpy);
@@ -758,6 +766,18 @@ static Bool __glXIsDirect(Display *dpy, GLXContextID contextID)
 	return GL_FALSE;
     }
 
+#ifdef USE_XCB
+   xcb_connection_t* c = XGetXCBConnection(dpy);
+   xcb_glx_is_direct_reply_t* reply =
+      xcb_glx_is_direct_reply(c,
+                              xcb_glx_is_direct(c, contextID),
+                              NULL);
+
+   const Bool is_direct = reply->is_direct ? True : False;
+   free(reply);
+
+   return is_direct;
+#else
     /* Send the glXIsDirect request */
     LockDisplay(dpy);
     GetReq(GLXIsDirect,req);
@@ -769,6 +789,7 @@ static Bool __glXIsDirect(Display *dpy, GLXContextID contextID)
     SyncHandle();
 
     return reply.isDirect;
+#endif /* USE_XCB */
 }
 
 /**
@@ -840,10 +861,15 @@ PUBLIC void glXDestroyGLXPixmap(Display *dpy, GLXPixmap glxpixmap)
 
 PUBLIC void glXSwapBuffers(Display *dpy, GLXDrawable drawable)
 {
-    xGLXSwapBuffersReq *req;
     GLXContext gc;
     GLXContextTag tag;
     CARD8 opcode;
+#ifdef USE_XCB
+    xcb_connection_t *c;
+#else
+    xGLXSwapBuffersReq *req;
+#endif
+
 #ifdef GLX_DIRECT_RENDERING
     __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL);
 
@@ -871,6 +897,11 @@ PUBLIC void glXSwapBuffers(Display *dpy, GLXDrawable drawable)
 	tag = 0;
     }
 
+#ifdef USE_XCB
+    c = XGetXCBConnection(dpy);
+    xcb_glx_swap_buffers(c, tag, drawable);
+    xcb_flush(c);
+#else
     /* Send the glXSwapBuffers request */
     LockDisplay(dpy);
     GetReq(GLXSwapBuffers,req);
@@ -881,6 +912,7 @@ PUBLIC void glXSwapBuffers(Display *dpy, GLXDrawable drawable)
     UnlockDisplay(dpy);
     SyncHandle();
     XFlush(dpy);
+#endif /* USE_XCB */
 }
 
 
@@ -1314,9 +1346,8 @@ PUBLIC const char *glXQueryExtensionsString( Display *dpy, int screen )
 
     if (!psc->effectiveGLXexts) {
         if (!psc->serverGLXexts) {
-	    psc->serverGLXexts = __glXGetStringFromServer(dpy, priv->majorOpcode,
-					  	   X_GLXQueryServerString,
-					  	   screen, GLX_EXTENSIONS);
+           psc->serverGLXexts =
+              __glXQueryServerString(dpy, priv->majorOpcode, screen, GLX_EXTENSIONS);
 	}
 
 	__glXCalculateUsableExtensions(psc,
@@ -1371,8 +1402,7 @@ PUBLIC const char *glXQueryServerString( Display *dpy, int screen, int name )
     }
 
     if ( *str == NULL ) {
-	*str = __glXGetStringFromServer(dpy, priv->majorOpcode,
-					X_GLXQueryServerString, screen, name);
+       *str = __glXQueryServerString(dpy, priv->majorOpcode, screen, name);
     }
     
     return *str;
@@ -1380,9 +1410,18 @@ PUBLIC const char *glXQueryServerString( Display *dpy, int screen, int name )
 
 void __glXClientInfo (  Display *dpy, int opcode  )
 {
-    xGLXClientInfoReq *req;
-    int size;
     char * ext_str = __glXGetClientGLExtensionString();
+    int size = strlen( ext_str ) + 1;
+
+#ifdef USE_XCB
+   xcb_connection_t *c = XGetXCBConnection(dpy);
+   xcb_glx_client_info(c,
+                       GLX_MAJOR_VERSION,
+                       GLX_MINOR_VERSION,
+                       size,
+                       (const uint8_t *)ext_str);
+#else
+    xGLXClientInfoReq *req;
 
     /* Send the glXClientInfo request */
     LockDisplay(dpy);
@@ -1392,14 +1431,14 @@ void __glXClientInfo (  Display *dpy, int opcode  )
     req->major = GLX_MAJOR_VERSION;
     req->minor = GLX_MINOR_VERSION;
 
-    size = strlen( ext_str ) + 1;
     req->length += (size + 3) >> 2;
     req->numbytes = size;
     Data(dpy, ext_str, size);
 
     UnlockDisplay(dpy);
     SyncHandle();
-    
+#endif /* USE_XCB */
+
     Xfree( ext_str );
 }
 
diff --git a/src/glx/x11/glxcurrent.c b/src/glx/x11/glxcurrent.c
index f3eb045d9f..4d0a7c65eb 100644
--- a/src/glx/x11/glxcurrent.c
+++ b/src/glx/x11/glxcurrent.c
@@ -312,13 +312,11 @@ SendMakeCurrentRequest(Display * dpy, CARD8 opcode,
 
 #ifdef GLX_DIRECT_RENDERING
 static __GLXDRIdrawable *
-FetchDRIDrawable(Display * dpy,
-                 GLXDrawable glxDrawable, GLXContext gc, Bool pre13)
+FetchDRIDrawable(Display * dpy, GLXDrawable glxDrawable, GLXContext gc)
 {
    __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
    __GLXDRIdrawable *pdraw;
    __GLXscreenConfigs *psc;
-   XID drawable;
 
    if (priv == NULL)
       return NULL;
@@ -330,15 +328,7 @@ FetchDRIDrawable(Display * dpy,
    if (__glxHashLookup(psc->drawHash, glxDrawable, (void *) &pdraw) == 0)
       return pdraw;
 
-   /* If this is glXMakeCurrent (pre GLX 1.3) we allow creating the
-    * GLX drawable on the fly.  Otherwise we pass None as the X
-    * drawable */
-   if (pre13)
-      drawable = glxDrawable;
-   else
-      drawable = None;
-
-   pdraw = psc->driScreen->createDrawable(psc, drawable,
+   pdraw = psc->driScreen->createDrawable(psc, glxDrawable,
                                           glxDrawable, gc->mode);
    if (__glxHashInsert(psc->drawHash, glxDrawable, pdraw)) {
       (*pdraw->destroyDrawable) (pdraw);
@@ -357,7 +347,7 @@ FetchDRIDrawable(Display * dpy,
  */
 static Bool
 MakeContextCurrent(Display * dpy, GLXDrawable draw,
-                   GLXDrawable read, GLXContext gc, Bool pre13)
+                   GLXDrawable read, GLXContext gc)
 {
    xGLXMakeCurrentReply reply;
    const GLXContext oldGC = __glXGetCurrentContext();
@@ -365,7 +355,7 @@ MakeContextCurrent(Display * dpy, GLXDrawable draw,
    const CARD8 oldOpcode = ((gc == oldGC) || (oldGC == &dummyContext))
       ? opcode : __glXSetupForCommand(oldGC->currentDpy);
    Bool bindReturnValue;
-
+   __GLXattribute *state;
 
    if (!opcode || !oldOpcode) {
       return GL_FALSE;
@@ -384,8 +374,21 @@ MakeContextCurrent(Display * dpy, GLXDrawable draw,
 #ifdef GLX_DIRECT_RENDERING
    /* Bind the direct rendering context to the drawable */
    if (gc && gc->driContext) {
-      __GLXDRIdrawable *pdraw = FetchDRIDrawable(dpy, draw, gc, pre13);
-      __GLXDRIdrawable *pread = FetchDRIDrawable(dpy, read, gc, pre13);
+      __GLXDRIdrawable *pdraw = FetchDRIDrawable(dpy, draw, gc);
+      __GLXDRIdrawable *pread = FetchDRIDrawable(dpy, read, gc);
+
+      if ((pdraw == NULL) || (pread == NULL)) {
+         xError error;
+
+         error.errorCode = GLXBadDrawable;
+         error.resourceID = (pdraw == NULL) ? draw : read;
+         error.sequenceNumber = dpy->request;
+         error.type = X_Error;
+         error.majorCode = gc->majorOpcode;
+         error.minorCode = X_GLXMakeContextCurrent;
+         _XError(dpy, &error);
+         return False;
+      }
 
       bindReturnValue =
          (gc->driContext->bindContext) (gc->driContext, pdraw, pread);
@@ -489,8 +492,7 @@ MakeContextCurrent(Display * dpy, GLXDrawable draw,
             } while (0);
 #endif
 
-            __GLXattribute *state =
-               (__GLXattribute *) (gc->client_state_private);
+            state = (__GLXattribute *) (gc->client_state_private);
 
             gc->currentContextTag = reply.contextTag;
             if (state->array_state == NULL) {
@@ -517,15 +519,16 @@ MakeContextCurrent(Display * dpy, GLXDrawable draw,
 PUBLIC Bool
 glXMakeCurrent(Display * dpy, GLXDrawable draw, GLXContext gc)
 {
-   return MakeContextCurrent(dpy, draw, draw, gc, True);
+   return MakeContextCurrent(dpy, draw, draw, gc);
 }
 
 PUBLIC
 GLX_ALIAS(Bool, glXMakeCurrentReadSGI,
           (Display * dpy, GLXDrawable d, GLXDrawable r, GLXContext ctx),
-          (dpy, d, r, ctx, False), MakeContextCurrent)
+          (dpy, d, r, ctx), MakeContextCurrent)
 
-     PUBLIC GLX_ALIAS(Bool, glXMakeContextCurrent,
-                      (Display * dpy, GLXDrawable d, GLXDrawable r,
-                       GLXContext ctx), (dpy, d, r, ctx, False),
-                      MakeContextCurrent)
+PUBLIC
+GLX_ALIAS(Bool, glXMakeContextCurrent,
+	  (Display * dpy, GLXDrawable d, GLXDrawable r,
+	   GLXContext ctx), (dpy, d, r, ctx),
+	  MakeContextCurrent)
diff --git a/src/glx/x11/glxext.c b/src/glx/x11/glxext.c
index b6bdbd52bb..be6edf9b19 100644
--- a/src/glx/x11/glxext.c
+++ b/src/glx/x11/glxext.c
@@ -139,7 +139,8 @@ XEXT_GENERATE_FIND_DISPLAY(__glXFindDisplay, __glXExtensionInfo,
 ** Free the per screen configs data as well as the array of
 ** __glXScreenConfigs.
 */
-     static void FreeScreenConfigs(__GLXdisplayPrivate * priv)
+static void
+FreeScreenConfigs(__GLXdisplayPrivate * priv)
 {
    __GLXscreenConfigs *psc;
    GLint i, screens;
@@ -221,6 +222,25 @@ __glXFreeDisplayPrivate(XExtData * extension)
 static Bool
 QueryVersion(Display * dpy, int opcode, int *major, int *minor)
 {
+#ifdef USE_XCB
+   xcb_connection_t *c = XGetXCBConnection(dpy);
+   xcb_glx_query_version_reply_t* reply =
+      xcb_glx_query_version_reply(c,
+                                  xcb_glx_query_version(c,
+                                                        GLX_MAJOR_VERSION,
+                                                        GLX_MINOR_VERSION),
+                                  NULL);
+
+   if(reply->major_version != GLX_MAJOR_VERSION)
+   {
+      free(reply);
+      return GL_FALSE;
+   }
+   *major = reply->major_version;
+   *minor = min(reply->minor_version, GLX_MINOR_VERSION);
+   free(reply);
+   return GL_TRUE;
+#else
    xGLXQueryVersionReq *req;
    xGLXQueryVersionReply reply;
 
@@ -245,6 +265,7 @@ QueryVersion(Display * dpy, int opcode, int *major, int *minor)
    *major = reply.majorVersion;
    *minor = min(reply.minorVersion, GLX_MINOR_VERSION);
    return GL_TRUE;
+#endif /* USE_XCB */
 }
 
 
@@ -529,9 +550,7 @@ getFBConfigs(Display * dpy, __GLXdisplayPrivate * priv, int screen)
    __GLXscreenConfigs *psc;
 
    psc = priv->screenConfigs + screen;
-   psc->serverGLXexts = __glXGetStringFromServer(dpy, priv->majorOpcode,
-                                                 X_GLXQueryServerString,
-                                                 screen, GLX_EXTENSIONS);
+   psc->serverGLXexts = __glXQueryServerString(dpy, priv->majorOpcode, screen, GLX_EXTENSIONS);
 
    LockDisplay(dpy);
 
@@ -589,9 +608,7 @@ AllocAndFetchScreenConfigs(Display * dpy, __GLXdisplayPrivate * priv)
    memset(psc, 0, screens * sizeof(__GLXscreenConfigs));
    priv->screenConfigs = psc;
 
-   priv->serverGLXversion = __glXGetStringFromServer(dpy, priv->majorOpcode,
-                                                     X_GLXQueryServerString,
-                                                     0, GLX_VERSION);
+   priv->serverGLXversion = __glXQueryServerString(dpy, priv->majorOpcode, 0, GLX_VERSION);
    if (priv->serverGLXversion == NULL) {
       FreeScreenConfigs(priv);
       return GL_FALSE;
diff --git a/src/glx/x11/glxextensions.c b/src/glx/x11/glxextensions.c
index d061fa9333..051433a006 100644
--- a/src/glx/x11/glxextensions.c
+++ b/src/glx/x11/glxextensions.c
@@ -99,7 +99,7 @@ static const struct extension_info known_glx_extensions[] = {
    { GLX(SGIS_color_range),            VER(0,0), N, N, N, N },
    { GLX(SGIS_multisample),            VER(0,0), Y, Y, N, N },
    { GLX(SGIX_fbconfig),               VER(1,3), Y, Y, N, N },
-   { GLX(SGIX_pbuffer),                VER(1,3), Y, N, N, N },
+   { GLX(SGIX_pbuffer),                VER(1,3), Y, Y, N, N },
    { GLX(SGIX_swap_barrier),           VER(0,0), N, N, N, N },
    { GLX(SGIX_swap_group),             VER(0,0), N, N, N, N },
    { GLX(SGIX_visual_select_group),    VER(0,0), Y, Y, N, N },
diff --git a/src/glx/x11/indirect.c b/src/glx/x11/indirect.c
index 1fcd5ca1bb..fb87abc0a5 100644
--- a/src/glx/x11/indirect.c
+++ b/src/glx/x11/indirect.c
@@ -5397,8 +5397,7 @@ __glx_TexSubImage_1D2D(unsigned opcode, unsigned dim, GLenum target,
             (void) memcpy((void *) (gc->pc + 44), (void *) (&height), 4);
             (void) memcpy((void *) (gc->pc + 48), (void *) (&format), 4);
             (void) memcpy((void *) (gc->pc + 52), (void *) (&type), 4);
-            (void) memcpy((void *) (gc->pc + 56),
-                          (void *) ((pixels == NULL) ? one : zero), 4);
+            (void) memset((void *) (gc->pc + 56), 0, 4);
             if (compsize > 0) {
                 (*gc->fillImage) (gc, dim, width, height, 1, format, type,
                                   pixels, gc->pc + 60, gc->pc + 4);
@@ -5424,7 +5423,7 @@ __glx_TexSubImage_1D2D(unsigned opcode, unsigned dim, GLenum target,
             (void) memcpy((void *) (pc + 48), (void *) (&height), 4);
             (void) memcpy((void *) (pc + 52), (void *) (&format), 4);
             (void) memcpy((void *) (pc + 56), (void *) (&type), 4);
-            (void) memcpy((void *) (pc + 60), zero, 4);
+            (void) memset((void *) (pc + 60), 0, 4);
             __glXSendLargeImage(gc, compsize, dim, width, height, 1, format,
                                 type, pixels, pc + 64, pc + 8);
         }
@@ -6869,8 +6868,7 @@ __glx_TexSubImage_3D4D(unsigned opcode, unsigned dim, GLenum target,
             (void) memcpy((void *) (gc->pc + 76), (void *) (&extent), 4);
             (void) memcpy((void *) (gc->pc + 80), (void *) (&format), 4);
             (void) memcpy((void *) (gc->pc + 84), (void *) (&type), 4);
-            (void) memcpy((void *) (gc->pc + 88),
-                          (void *) ((pixels == NULL) ? one : zero), 4);
+            (void) memset((void *) (gc->pc + 88), 0, 4);
             if (compsize > 0) {
                 (*gc->fillImage) (gc, dim, width, height, depth, format, type,
                                   pixels, gc->pc + 92, gc->pc + 4);
@@ -6900,7 +6898,7 @@ __glx_TexSubImage_3D4D(unsigned opcode, unsigned dim, GLenum target,
             (void) memcpy((void *) (pc + 80), (void *) (&extent), 4);
             (void) memcpy((void *) (pc + 84), (void *) (&format), 4);
             (void) memcpy((void *) (pc + 88), (void *) (&type), 4);
-            (void) memcpy((void *) (pc + 92), zero, 4);
+            (void) memset((void *) (pc + 92), 0, 4);
             __glXSendLargeImage(gc, compsize, dim, width, height, depth,
                                 format, type, pixels, pc + 96, pc + 8);
         }
@@ -7525,26 +7523,6 @@ __indirect_glGetProgramivARB(GLenum target, GLenum pname, GLint * params)
 
 #define X_GLrop_ProgramEnvParameter4dvARB 4185
 void
-__indirect_glProgramEnvParameter4dARB(GLenum target, GLuint index, GLdouble x,
-                                      GLdouble y, GLdouble z, GLdouble w)
-{
-    __GLXcontext *const gc = __glXGetCurrentContext();
-    const GLuint cmdlen = 44;
-    emit_header(gc->pc, X_GLrop_ProgramEnvParameter4dvARB, cmdlen);
-    (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
-    (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4);
-    (void) memcpy((void *) (gc->pc + 12), (void *) (&x), 8);
-    (void) memcpy((void *) (gc->pc + 20), (void *) (&y), 8);
-    (void) memcpy((void *) (gc->pc + 28), (void *) (&z), 8);
-    (void) memcpy((void *) (gc->pc + 36), (void *) (&w), 8);
-    gc->pc += cmdlen;
-    if (__builtin_expect(gc->pc > gc->limit, 0)) {
-        (void) __glXFlushRenderBuffer(gc, gc->pc);
-    }
-}
-
-#define X_GLrop_ProgramEnvParameter4dvARB 4185
-void
 __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index,
                                        const GLdouble * params)
 {
@@ -7562,26 +7540,6 @@ __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index,
 
 #define X_GLrop_ProgramEnvParameter4fvARB 4184
 void
-__indirect_glProgramEnvParameter4fARB(GLenum target, GLuint index, GLfloat x,
-                                      GLfloat y, GLfloat z, GLfloat w)
-{
-    __GLXcontext *const gc = __glXGetCurrentContext();
-    const GLuint cmdlen = 28;
-    emit_header(gc->pc, X_GLrop_ProgramEnvParameter4fvARB, cmdlen);
-    (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
-    (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4);
-    (void) memcpy((void *) (gc->pc + 12), (void *) (&x), 4);
-    (void) memcpy((void *) (gc->pc + 16), (void *) (&y), 4);
-    (void) memcpy((void *) (gc->pc + 20), (void *) (&z), 4);
-    (void) memcpy((void *) (gc->pc + 24), (void *) (&w), 4);
-    gc->pc += cmdlen;
-    if (__builtin_expect(gc->pc > gc->limit, 0)) {
-        (void) __glXFlushRenderBuffer(gc, gc->pc);
-    }
-}
-
-#define X_GLrop_ProgramEnvParameter4fvARB 4184
-void
 __indirect_glProgramEnvParameter4fvARB(GLenum target, GLuint index,
                                        const GLfloat * params)
 {
diff --git a/src/glx/x11/indirect.h b/src/glx/x11/indirect.h
index 0719a1b302..f8c88b36bb 100644
--- a/src/glx/x11/indirect.h
+++ b/src/glx/x11/indirect.h
@@ -517,9 +517,7 @@ extern HIDDEN void __indirect_glGetProgramivARB(GLenum target, GLenum pname, GLi
 extern HIDDEN void __indirect_glGetVertexAttribdvARB(GLuint index, GLenum pname, GLdouble * params);
 extern HIDDEN void __indirect_glGetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat * params);
 extern HIDDEN void __indirect_glGetVertexAttribivARB(GLuint index, GLenum pname, GLint * params);
-extern HIDDEN void __indirect_glProgramEnvParameter4dARB(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w);
 extern HIDDEN void __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index, const GLdouble * params);
-extern HIDDEN void __indirect_glProgramEnvParameter4fARB(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w);
 extern HIDDEN void __indirect_glProgramEnvParameter4fvARB(GLenum target, GLuint index, const GLfloat * params);
 extern HIDDEN void __indirect_glProgramLocalParameter4dARB(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w);
 extern HIDDEN void __indirect_glProgramLocalParameter4dvARB(GLenum target, GLuint index, const GLdouble * params);
diff --git a/src/glx/x11/indirect_init.c b/src/glx/x11/indirect_init.c
index 852fe712c6..479184337c 100644
--- a/src/glx/x11/indirect_init.c
+++ b/src/glx/x11/indirect_init.c
@@ -526,9 +526,7 @@ __GLapi * __glXNewIndirectAPI( void )
     glAPI->GetVertexAttribdvARB = __indirect_glGetVertexAttribdvARB;
     glAPI->GetVertexAttribfvARB = __indirect_glGetVertexAttribfvARB;
     glAPI->GetVertexAttribivARB = __indirect_glGetVertexAttribivARB;
-    glAPI->ProgramEnvParameter4dARB = __indirect_glProgramEnvParameter4dARB;
     glAPI->ProgramEnvParameter4dvARB = __indirect_glProgramEnvParameter4dvARB;
-    glAPI->ProgramEnvParameter4fARB = __indirect_glProgramEnvParameter4fARB;
     glAPI->ProgramEnvParameter4fvARB = __indirect_glProgramEnvParameter4fvARB;
     glAPI->ProgramLocalParameter4dARB = __indirect_glProgramLocalParameter4dARB;
     glAPI->ProgramLocalParameter4dvARB = __indirect_glProgramLocalParameter4dvARB;
diff --git a/src/glx/x11/single2.c b/src/glx/x11/single2.c
index b008c6996c..0b2d5113cc 100644
--- a/src/glx/x11/single2.c
+++ b/src/glx/x11/single2.c
@@ -40,6 +40,7 @@
 #ifdef USE_XCB
 #include <xcb/xcb.h>
 #include <xcb/glx.h>
+#include <X11/Xlib-xcb.h>
 #endif /* USE_XCB */
 
 
@@ -678,9 +679,7 @@ __indirect_glGetString(GLenum name)
     */
 
    (void) __glXFlushRenderBuffer(gc, gc->pc);
-   s = (GLubyte *) __glXGetStringFromServer(dpy, gc->majorOpcode,
-                                            X_GLsop_GetString,
-                                            gc->currentContextTag, name);
+   s = (GLubyte *) __glXGetString(dpy, gc->majorOpcode, gc->currentContextTag, name);
    if (!s) {
       /* Throw data on the floor */
       __glXSetError(gc, GL_OUT_OF_MEMORY);
diff --git a/src/mesa/Makefile b/src/mesa/Makefile
index ba65ce695f..78b71d4676 100644
--- a/src/mesa/Makefile
+++ b/src/mesa/Makefile
@@ -3,8 +3,7 @@
 TOP = ../..
 include $(TOP)/configs/current
 
-include sources
-
+include sources.mak
 
 .SUFFIXES : .cpp
 
@@ -19,9 +18,11 @@ include sources
 
 
 
+
 # Default: build dependencies, then asm_subdirs, then convenience
 # libs (.a) and finally the device drivers:
-default: depend asm_subdirs libmesa.a libglapi.a driver_subdirs
+default: depend asm_subdirs libmesa.a libmesagallium.a libglapi.a \
+	driver_subdirs
 
 
 
@@ -32,10 +33,17 @@ default: depend asm_subdirs libmesa.a libglapi.a driver_subdirs
 libmesa.a: $(MESA_OBJECTS)
 	@ $(MKLIB) -o mesa -static $(MESA_OBJECTS)
 
+# Make archive of subset of core mesa object files for gallium
+libmesagallium.a: $(MESA_GALLIUM_OBJECTS)
+	@ $(MKLIB) -o mesagallium -static $(MESA_GALLIUM_OBJECTS)
+
 # Make archive of gl* API dispatcher functions only
 libglapi.a: $(GLAPI_OBJECTS)
-	@ $(MKLIB) -o glapi -static $(GLAPI_OBJECTS)
-
+	@if [ "${WINDOW_SYSTEM}" = "dri" ] ; then \
+		touch libglapi.a ; \
+	else \
+		$(MKLIB) -o glapi -static $(GLAPI_OBJECTS) ; \
+	fi
 
 ######################################################################
 # Device drivers
@@ -88,14 +96,29 @@ install: default
 	done
 
 
-pcedit = sed \
+gl_pcedit = sed \
 	-e 's,@INSTALL_DIR@,$(INSTALL_DIR),' \
 	-e 's,@INSTALL_LIB_DIR@,$(INSTALL_LIB_DIR),' \
 	-e 's,@INSTALL_INC_DIR@,$(INSTALL_INC_DIR),' \
-	-e 's,@VERSION@,$(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY),'
+	-e 's,@VERSION@,$(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY),' \
+	-e 's,@GL_PC_REQ_PRIV@,$(GL_PC_REQ_PRIV),' \
+	-e 's,@GL_PC_LIB_PRIV@,$(GL_PC_LIB_PRIV),' \
+	-e 's,@GL_PC_CFLAGS@,$(GL_PC_CFLAGS),'
 
 gl.pc: gl.pc.in
-	$(pcedit) $< > $@
+	$(gl_pcedit) $< > $@
+
+osmesa_pcedit = sed \
+	-e 's,@INSTALL_DIR@,$(INSTALL_DIR),' \
+	-e 's,@INSTALL_LIB_DIR@,$(INSTALL_LIB_DIR),' \
+	-e 's,@INSTALL_INC_DIR@,$(INSTALL_INC_DIR),' \
+	-e 's,@VERSION@,$(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY),' \
+	-e 's,@OSMESA_LIB@,$(OSMESA_LIB),' \
+	-e 's,@OSMESA_PC_REQ@,$(OSMESA_PC_REQ),' \
+	-e 's,@OSMESA_PC_LIB_PRIV@,$(OSMESA_PC_LIB_PRIV),'
+
+osmesa.pc: osmesa.pc.in
+	$(osmesa_pcedit) $< > $@
 
 install-headers:
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_INC_DIR)/GL
@@ -109,10 +132,12 @@ install-libgl: default gl.pc install-headers
 		$(DESTDIR)$(INSTALL_LIB_DIR)
 	$(INSTALL) -m 644 gl.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
-install-osmesa: default
+install-osmesa: default osmesa.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)
+	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 	$(INSTALL) $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_GLOB) \
 		$(DESTDIR)$(INSTALL_LIB_DIR)
+	$(INSTALL) -m 644 osmesa.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
 install-dri: default
 	cd drivers/dri && $(MAKE) install
diff --git a/src/mesa/Makefile.DJ b/src/mesa/Makefile.DJ
deleted file mode 100644
index 06a13fb1ab..0000000000
--- a/src/mesa/Makefile.DJ
+++ /dev/null
@@ -1,166 +0,0 @@
-# Mesa 3-D graphics library
-# Version:  5.1
-# 
-# Copyright (C) 1999-2003  Brian Paul   All Rights Reserved.
-# 
-# Permission is hereby granted, free of charge, to any person obtaining a
-# copy of this software and associated documentation files (the "Software"),
-# to deal in the Software without restriction, including without limitation
-# the rights to use, copy, modify, merge, publish, distribute, sublicense,
-# and/or sell copies of the Software, and to permit persons to whom the
-# Software is furnished to do so, subject to the following conditions:
-# 
-# The above copyright notice and this permission notice shall be included
-# in all copies or substantial portions of the Software.
-# 
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
-# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
-# BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
-# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
-# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-
-# DOS/DJGPP core makefile v1.7 for Mesa
-#
-#  Copyright (C) 2002 - Daniel Borca
-#  Email : dborca@users.sourceforge.net
-#  Web   : http://www.geocities.com/dborca
-
-
-#
-#  Available options:
-#
-#    Environment variables:
-#	CFLAGS
-#
-#	GLIDE		path to Glide3 SDK; used with FX.
-#			default = $(TOP)/glide3
-#	FX=1		build for 3dfx Glide3. Note that this disables
-#			compilation of most DMesa code and requires fxMesa.
-#			As a consequence, you'll need the DJGPP Glide3
-#			library to build any application.
-#			default = no
-#	X86=1		optimize for x86 (if possible, use MMX, SSE, 3DNow).
-#			default = no
-#
-#    Targets:
-#	all:		build GL
-#	clean:		remove object files
-#
-
-
-
-.PHONY: all clean
-.INTERMEDIATE: x86/gen_matypes.exe
-
-TOP = ../..
-GLIDE ?= $(TOP)/glide3
-LIBDIR = $(TOP)/lib
-GL_LIB = libgl.a
-GL_DXE = gl.dxe
-GL_IMP = libigl.a
-
-export LD_LIBRARY_PATH := $(LD_LIBRARY_PATH);$(LIBDIR);$(GLIDE)/lib
-
-CC = gcc
-CFLAGS += $(INCLUDE_DIRS)
-CFLAGS += -DUSE_EXTERNAL_DXTN_LIB=1
-ifeq ($(FX),1)
-CFLAGS += -D__DOS__
-CFLAGS += -I$(GLIDE)/include -DFX
-LIBNAME = "Mesa/FX DJGPP"
-else
-LIBNAME = "Mesa DJGPP"
-endif
-
-AR = ar
-ARFLAGS = crus
-
-HAVEDXE3 = $(wildcard $(DJDIR)/bin/dxe3gen.exe)
-
-ifeq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),)
-UNLINK = del $(subst /,\,$(1))
-else
-UNLINK = $(RM) $(1)
-endif
-
-include sources
-
-ifeq ($(X86),1)
-CFLAGS += -DUSE_X86_ASM
-CFLAGS += -DUSE_MMX_ASM
-CFLAGS += -DUSE_SSE_ASM
-CFLAGS += -DUSE_3DNOW_ASM
-X86_SOURCES += $(X86_API)
-else
-X86_SOURCES =
-endif
-
-DRIVER_SOURCES = \
-	drivers/dos/dmesa.c
-ifeq ($(FX),1)
-DRIVER_SOURCES += \
-	$(GLIDE_DRIVER_SOURCES)
-else
-DRIVER_SOURCES += \
-	$(OSMESA_DRIVER_SOURCES) \
-	drivers/dos/video.c \
-	drivers/dos/vesa.c \
-	drivers/dos/blit.S \
-	drivers/dos/vga.c \
-	drivers/dos/null.c \
-	drivers/dos/dpmi.c
-endif
-
-SOURCES = $(CORE_SOURCES) $(X86_SOURCES) $(COMMON_DRIVER_SOURCES) $(DRIVER_SOURCES)
-
-OBJECTS = $(addsuffix .o,$(basename $(SOURCES)))
-
-X86_OBJECTS = $(addsuffix .o,$(basename $(X86_SOURCES)))
-
-.c.o:
-	$(CC) -o $@ $(CFLAGS) -c $<
-.S.o:
-	$(CC) -o $@ $(CFLAGS) -c $<
-.s.o:
-	$(CC) -o $@ $(CFLAGS) -x assembler-with-cpp -c $<
-
-all: $(LIBDIR)/$(GL_LIB) $(LIBDIR)/$(GL_DXE) $(LIBDIR)/$(GL_IMP)
-
-$(LIBDIR)/$(GL_LIB): $(OBJECTS)
-	$(AR) $(ARFLAGS) $@ $^
-
-$(LIBDIR)/$(GL_DXE) $(LIBDIR)/$(GL_IMP): $(OBJECTS)
-ifeq ($(HAVEDXE3),)
-	$(warning Missing DXE3 package... Skipping $(GL_DXE))
-else
-ifeq ($(FX),1)
-	-dxe3gen -o $(LIBDIR)/$(GL_DXE) -Y $(LIBDIR)/$(GL_IMP) -D $(LIBNAME) -E _gl -E _DMesa -P glide3x.dxe -U $^
-else
-	-dxe3gen -o $(LIBDIR)/$(GL_DXE) -Y $(LIBDIR)/$(GL_IMP) -D $(LIBNAME) -E _gl -E _DMesa -U $^
-endif
-endif
-
-$(X86_OBJECTS): x86/matypes.h
-
-x86/matypes.h: x86/gen_matypes.exe
-	$< > $@
-
-x86/gen_matypes.exe: x86/gen_matypes.c
-	$(CC) -o $@ $(CFLAGS) -s $<
-
-clean:
-	-$(call UNLINK,array_cache/*.o)
-	-$(call UNLINK,glapi/*.o)
-	-$(call UNLINK,main/*.o)
-	-$(call UNLINK,math/*.o)
-	-$(call UNLINK,shader/*.o)
-	-$(call UNLINK,sparc/*.o)
-	-$(call UNLINK,ppc/*.o)
-	-$(call UNLINK,swrast/*.o)
-	-$(call UNLINK,swrast_setup/*.o)
-	-$(call UNLINK,tnl/*.o)
-	-$(call UNLINK,x86/*.o)
-	-$(call UNLINK,drivers/common/*.o)
-	-$(call UNLINK,drivers/dos/*.o)
-	-$(call UNLINK,drivers/glide/*.o)
diff --git a/src/mesa/Makefile.mgw b/src/mesa/Makefile.mgw
index 886b344ec9..097c390a83 100644
--- a/src/mesa/Makefile.mgw
+++ b/src/mesa/Makefile.mgw
@@ -124,7 +124,7 @@ ifneq ($(wildcard $(addsuffix /rm,$(subst :, ,$(PATH)))),)
 UNLINK = $(RM) $(1)
 endif
 
-include sources
+include sources.mak
 
 CFLAGS += $(INCLUDE_DIRS) 
 
diff --git a/src/mesa/SConscript b/src/mesa/SConscript
new file mode 100644
index 0000000000..a878d31dbb
--- /dev/null
+++ b/src/mesa/SConscript
@@ -0,0 +1,343 @@
+#######################################################################
+# SConscript for Mesa
+
+
+Import('*')
+
+if env['platform'] != 'winddk':
+
+	env = env.Clone()
+	
+	env.Append(CPPPATH = [
+		'#/src/mesa',
+	])
+	
+	if env['platform'] == 'windows':
+		env.Append(CPPDEFINES = [
+			'_GDI32_', # prevent gl* being declared __declspec(dllimport) in MS headers
+			'BUILD_GL32', # declare gl* as __declspec(dllexport) in Mesa headers 
+		])	 
+
+	#
+	# Source files
+	#
+	
+	main_sources = [
+		'main/api_arrayelt.c',
+		'main/api_exec.c',
+		'main/api_loopback.c',
+		'main/api_noop.c',
+		'main/api_validate.c',
+		'main/accum.c',
+		'main/attrib.c',
+		'main/arrayobj.c',
+		'main/blend.c',
+		'main/bufferobj.c',
+		'main/buffers.c',
+		'main/clear.c',
+		'main/clip.c',
+		'main/colortab.c',
+		'main/context.c',
+		'main/convolve.c',
+		'main/debug.c',
+		'main/depth.c',
+		'main/depthstencil.c',
+		'main/dlist.c',
+		'main/dlopen.c',
+		'main/drawpix.c',
+		'main/enable.c',
+		'main/enums.c',
+		'main/eval.c',
+		'main/execmem.c',
+		'main/extensions.c',
+		'main/fbobject.c',
+		'main/feedback.c',
+		'main/ffvertex_prog.c',
+		'main/fog.c',
+		'main/framebuffer.c',
+		'main/get.c',
+		'main/getstring.c',
+		'main/hash.c',
+		'main/hint.c',
+		'main/histogram.c',
+		'main/image.c',
+		'main/imports.c',
+		'main/light.c',
+		'main/lines.c',
+		'main/matrix.c',
+		'main/mipmap.c',
+		'main/mm.c',
+		'main/multisample.c',
+		'main/pixel.c',
+		'main/pixelstore.c',
+		'main/points.c',
+		'main/polygon.c',
+		'main/queryobj.c',
+		'main/rastpos.c',
+		'main/rbadaptors.c',
+		'main/readpix.c',
+		'main/renderbuffer.c',
+		'main/scissor.c',
+		'main/shaders.c',
+		'main/state.c',
+		'main/stencil.c',
+		'main/texcompress.c',
+		'main/texcompress_s3tc.c',
+		'main/texcompress_fxt1.c',
+		'main/texenv.c',
+		'main/texenvprogram.c',
+		'main/texformat.c',
+		'main/texgen.c',
+		'main/teximage.c',
+		'main/texobj.c',
+		'main/texparam.c',
+		'main/texrender.c',
+		'main/texstate.c',
+		'main/texstore.c',
+		'main/varray.c',
+		'main/vtxfmt.c',
+	]
+	
+	math_sources = [
+		'math/m_debug_clip.c',
+		'math/m_debug_norm.c',
+		'math/m_debug_xform.c',
+		'math/m_eval.c',
+		'math/m_matrix.c',
+		'math/m_translate.c',
+		'math/m_vector.c',
+		'math/m_xform.c',
+	]
+	
+	vbo_sources = [
+		'vbo/vbo_context.c',
+		'vbo/vbo_exec.c',
+		'vbo/vbo_exec_api.c',
+		'vbo/vbo_exec_array.c',
+		'vbo/vbo_exec_draw.c',
+		'vbo/vbo_exec_eval.c',
+		'vbo/vbo_rebase.c',
+		'vbo/vbo_split.c',
+		'vbo/vbo_split_copy.c',
+		'vbo/vbo_split_inplace.c',
+		'vbo/vbo_save.c',
+		'vbo/vbo_save_api.c',
+		'vbo/vbo_save_draw.c',
+		'vbo/vbo_save_loopback.c',
+	]
+	
+	vf_sources = [
+		'vf/vf.c',
+		'vf/vf_generic.c',
+		'vf/vf_sse.c',
+	]
+	
+	statetracker_sources = [
+		'state_tracker/st_atom.c',
+		'state_tracker/st_atom_blend.c',
+		'state_tracker/st_atom_clip.c',
+		'state_tracker/st_atom_constbuf.c',
+		'state_tracker/st_atom_depth.c',
+		'state_tracker/st_atom_framebuffer.c',
+		'state_tracker/st_atom_pixeltransfer.c',
+		'state_tracker/st_atom_sampler.c',
+		'state_tracker/st_atom_scissor.c',
+		'state_tracker/st_atom_shader.c',
+		'state_tracker/st_atom_rasterizer.c',
+		'state_tracker/st_atom_stipple.c',
+		'state_tracker/st_atom_texture.c',
+		'state_tracker/st_atom_viewport.c',
+		'state_tracker/st_cb_accum.c',
+		'state_tracker/st_cb_bitmap.c',
+		'state_tracker/st_cb_blit.c',
+		'state_tracker/st_cb_bufferobjects.c',
+		'state_tracker/st_cb_clear.c',
+		'state_tracker/st_cb_flush.c',
+		'state_tracker/st_cb_drawpixels.c',
+		'state_tracker/st_cb_fbo.c',
+		'state_tracker/st_cb_get.c',
+		'state_tracker/st_cb_feedback.c',
+		'state_tracker/st_cb_program.c',
+		'state_tracker/st_cb_queryobj.c',
+		'state_tracker/st_cb_rasterpos.c',
+		'state_tracker/st_cb_readpixels.c',
+		'state_tracker/st_cb_strings.c',
+		'state_tracker/st_cb_texture.c',
+		'state_tracker/st_context.c',
+		'state_tracker/st_debug.c',
+		'state_tracker/st_draw.c',
+		'state_tracker/st_draw_feedback.c',
+		'state_tracker/st_extensions.c',
+		'state_tracker/st_format.c',
+		'state_tracker/st_framebuffer.c',
+		'state_tracker/st_gen_mipmap.c',
+		'state_tracker/st_mesa_to_tgsi.c',
+		'state_tracker/st_program.c',
+		'state_tracker/st_texture.c',
+	]
+	
+	shader_sources = [
+		'shader/arbprogparse.c',
+		'shader/arbprogram.c',
+		'shader/atifragshader.c',
+		'shader/grammar/grammar_mesa.c',
+		'shader/nvfragparse.c',
+		'shader/nvprogram.c',
+		'shader/nvvertparse.c',
+		'shader/program.c',
+		'shader/prog_cache.c',
+		'shader/prog_debug.c',
+		'shader/prog_execute.c',
+		'shader/prog_instruction.c',
+		'shader/prog_noise.c',
+		'shader/prog_parameter.c',
+		'shader/prog_print.c',
+		'shader/prog_statevars.c',
+		'shader/prog_uniform.c',
+		'shader/programopt.c',
+		'shader/shader_api.c',
+	]
+	
+	slang_sources = [
+		'shader/slang/slang_builtin.c',
+		'shader/slang/slang_codegen.c',
+		'shader/slang/slang_compile.c',
+		'shader/slang/slang_compile_function.c',
+		'shader/slang/slang_compile_operation.c',
+		'shader/slang/slang_compile_struct.c',
+		'shader/slang/slang_compile_variable.c',
+		'shader/slang/slang_emit.c',
+		'shader/slang/slang_ir.c',
+		'shader/slang/slang_label.c',
+		'shader/slang/slang_link.c',
+		'shader/slang/slang_log.c',
+		'shader/slang/slang_mem.c',
+		'shader/slang/slang_preprocess.c',
+		'shader/slang/slang_print.c',
+		'shader/slang/slang_simplify.c',
+		'shader/slang/slang_storage.c',
+		'shader/slang/slang_typeinfo.c',
+		'shader/slang/slang_vartable.c',
+		'shader/slang/slang_utility.c',
+	]
+	
+	mesa_sources = (
+		main_sources +
+		math_sources +
+		vbo_sources +
+		vf_sources +
+		statetracker_sources +
+		shader_sources +
+		slang_sources
+	)
+
+	glapi_sources = [
+		'main/dispatch.c',
+		'glapi/glapi.c',
+		'glapi/glapi_getproc.c',
+		'glapi/glthread.c',
+	]
+	
+	#
+	# Assembly sources
+	#
+	if gcc and env['machine'] == 'x86':
+		env.Append(CPPDEFINES = [
+			'USE_X86_ASM', 
+			'USE_MMX_ASM',
+			'USE_3DNOW_ASM',
+			'USE_SSE_ASM',
+		])
+		mesa_sources += [
+			'x86/common_x86.c',
+			'x86/x86.c',
+			'x86/3dnow.c',
+			'x86/sse.c',
+			'x86/common_x86_asm.S',
+			'x86/x86_xform2.S',
+			'x86/x86_xform3.S',
+			'x86/x86_xform4.S',
+			'x86/x86_cliptest.S',
+			'x86/mmx_blend.S',
+			'x86/3dnow_xform1.S',
+			'x86/3dnow_xform2.S',
+			'x86/3dnow_xform3.S',
+			'x86/3dnow_xform4.S',
+			'x86/3dnow_normal.S',
+			'x86/sse_xform1.S',
+			'x86/sse_xform2.S',
+			'x86/sse_xform3.S',
+			'x86/sse_xform4.S',
+			'x86/sse_normal.S',
+			'x86/read_rgba_span_x86.S',
+		]
+		glapi_sources += [
+			'x86/glapi_x86.S',
+		]
+	elif gcc and env['machine'] == 'x86_64':
+		env.Append(CPPDEFINES = [
+			'USE_X86_64_ASM', 
+		])
+		mesa_sources += [
+			'x86-64/x86-64.c',
+			'x86-64/xform4.S',
+		]
+		glapi_sources += [
+			'x86-64/glapi_x86-64.S'
+		]
+	elif gcc and env['machine'] == 'ppc':
+		env.Append(CPPDEFINES = [
+			'USE_PPC_ASM', 
+			'USE_VMX_ASM', 
+		])
+		mesa_sources += [
+			'ppc/common_ppc.c',
+		]
+		glapi_sources += [
+		]
+	elif gcc and env['machine'] == 'sparc':
+		mesa_sources += [
+			'sparc/sparc.c',
+			'sparc/clip.S',
+			'sparc/norm.S',
+			'sparc/xform.S',
+		]
+		glapi_sources += [
+			'sparc/glapi_sparc.S'
+		]
+	else:
+		pass
+	
+	# Generate matypes.h
+	if gcc and env['machine'] in ('x86', 'x86_64'):
+		# See http://www.scons.org/wiki/UsingCodeGenerators
+		gen_matypes = env.Program(
+			target = 'gen_matypes',
+			source = 'x86/gen_matypes.c',
+		)
+		matypes = env.Command(
+			'matypes.h',
+			gen_matypes,
+			gen_matypes[0].abspath + ' > $TARGET',
+		)
+		# Add the dir containing the generated header (somewhere inside  the
+		# build dir) to the include path  
+		env.Append(CPPPATH = [matypes[0].dir])
+	
+	#
+	# Libraries
+	# 
+
+	mesa = env.ConvenienceLibrary(
+		target = 'mesa',
+		source = mesa_sources,
+	)
+	Export('mesa')
+	
+	if not dri:
+		glapi = env.ConvenienceLibrary(
+			target = 'glapi',
+			source = glapi_sources,
+		)
+		Export('glapi')
+
diff --git a/src/mesa/drivers/allegro/amesa.c b/src/mesa/drivers/allegro/amesa.c
index ade6251848..0744677d2b 100644
--- a/src/mesa/drivers/allegro/amesa.c
+++ b/src/mesa/drivers/allegro/amesa.c
@@ -26,7 +26,7 @@
 #include "main/imports.h"
 #include "main/matrix.h"
 #include "main/mtypes.h"
-#include "GL/amesa.h"
+#include "amesa.h"
 
 
 struct amesa_visual
@@ -338,7 +338,7 @@ void AMesaDestroyBuffer(AMesaBuffer buffer)
 {
    if (buffer->Screen)     destroy_bitmap(buffer->Screen);
    if (buffer->Background) destroy_bitmap(buffer->Background);
-   _mesa_unreference_framebuffer(&buffer->GLBuffer);
+   _mesa_reference_framebuffer(&buffer->GLBuffer, NULL);
    free(buffer);
 }
 
diff --git a/src/mesa/drivers/allegro/amesa.h b/src/mesa/drivers/allegro/amesa.h
new file mode 100644
index 0000000000..852d34cf4f
--- /dev/null
+++ b/src/mesa/drivers/allegro/amesa.h
@@ -0,0 +1,65 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.3
+ * 
+ * Copyright (C) 1999-2000  Brian Paul   All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/* Allegro (DJGPP) driver by Bernhard Tschirren (bernie-t@geocities.com) */
+
+
+#ifndef AMESA_H
+#define AMESA_H
+
+
+#define AMESA_MAJOR_VERSION 3
+#define AMESA_MINOR_VERSION 3
+
+
+typedef struct amesa_visual  *AMesaVisual;
+typedef struct amesa_buffer  *AMesaBuffer;
+typedef struct amesa_context *AMesaContext;
+
+
+extern AMesaVisual AMesaCreateVisual(GLboolean dbFlag, GLint depth,
+                                     GLint depthSize,
+                                     GLint stencilSize,
+                                     GLint accumSize);
+
+extern void AMesaDestroyVisual(AMesaVisual visual);
+
+extern AMesaBuffer AMesaCreateBuffer(AMesaVisual visual,
+                                     GLint width, GLint height);
+
+extern void AMesaDestroyBuffer(AMesaBuffer buffer);
+
+
+extern AMesaContext AMesaCreateContext(AMesaVisual visual,
+                                       AMesaContext sharelist);
+
+extern void AMesaDestroyContext(AMesaContext context);
+
+extern GLboolean AMesaMakeCurrent(AMesaContext context, AMesaBuffer buffer);
+
+extern void AMesaSwapBuffers(AMesaBuffer buffer);
+
+
+#endif /* AMESA_H */
diff --git a/src/mesa/drivers/dri/Makefile b/src/mesa/drivers/dri/Makefile
index eef68825bc..32db097861 100644
--- a/src/mesa/drivers/dri/Makefile
+++ b/src/mesa/drivers/dri/Makefile
@@ -6,7 +6,7 @@ include $(TOP)/configs/current
 
 
 
-default: $(TOP)/$(LIB_DIR) subdirs
+default: $(TOP)/$(LIB_DIR) subdirs dri.pc
 
 
 $(TOP)/$(LIB_DIR):
@@ -25,7 +25,8 @@ pcedit = sed \
 	-e 's,@INSTALL_LIB_DIR@,$(INSTALL_LIB_DIR),' \
 	-e 's,@INSTALL_INC_DIR@,$(INSTALL_INC_DIR),' \
 	-e 's,@VERSION@,$(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY),' \
-	-e 's,@DRI_DRIVER_DIR@,$(DRI_DRIVER_SEARCH_DIR),'
+	-e 's,@DRI_DRIVER_DIR@,$(DRI_DRIVER_SEARCH_DIR),' \
+	-e 's,@DRI_PC_REQ_PRIV@,$(DRI_PC_REQ_PRIV),'
 
 dri.pc: dri.pc.in
 	$(pcedit) $< > $@
diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template
index 2dc3664cc6..2fa36bab3f 100644
--- a/src/mesa/drivers/dri/Makefile.template
+++ b/src/mesa/drivers/dri/Makefile.template
@@ -2,13 +2,16 @@
 
 MESA_MODULES = $(TOP)/src/mesa/libmesa.a
 
-COMMON_SOURCES = \
+COMMON_GALLIUM_SOURCES = \
         ../common/utils.c \
-        ../common/texmem.c \
         ../common/vblank.c \
         ../common/dri_util.c \
-        ../common/xmlconfig.c \
-        ../common/drirenderbuffer.c 
+        ../common/xmlconfig.c
+
+COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \
+        ../../common/driverfuncs.c \
+        ../common/texmem.c \
+        ../common/drirenderbuffer.c
 
 ifeq ($(WINDOW_SYSTEM),dri)
 WINOBJ=
@@ -59,9 +62,9 @@ SHARED_INCLUDES = \
 default: symlinks depend $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME)
 
 
-$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template
+$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template
 	$(MKLIB) -o $@ -noprefix -linker '$(CC)' -ldflags '$(LDFLAGS)' \
-		$(OBJECTS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS)
+		$(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS)
 
 
 $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME)
@@ -69,6 +72,7 @@ $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME)
 
 
 depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS)
+	rm -f depend
 	touch depend
 	$(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \
 		$(ASM_SOURCES)
diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c
index e9b1a80b91..ae79055405 100644
--- a/src/mesa/drivers/dri/common/dri_util.c
+++ b/src/mesa/drivers/dri/common/dri_util.c
@@ -314,13 +314,28 @@ static void driReportDamage(__DRIdrawable *pdp,
 static void driSwapBuffers(__DRIdrawable *dPriv)
 {
     __DRIscreen *psp = dPriv->driScreenPriv;
-
+    drm_clip_rect_t *rects;
+    int i;
+    
     if (!dPriv->numClipRects)
         return;
 
     psp->DriverAPI.SwapBuffers(dPriv);
 
-    driReportDamage(dPriv, dPriv->pClipRects, dPriv->numClipRects);
+    rects = _mesa_malloc(sizeof(*rects) * dPriv->numClipRects);
+
+    if (!rects)
+        return;
+
+    for (i = 0; i < dPriv->numClipRects; i++) {
+        rects[i].x1 = dPriv->pClipRects[i].x1 - dPriv->x;
+        rects[i].y1 = dPriv->pClipRects[i].y1 - dPriv->y;
+        rects[i].x2 = dPriv->pClipRects[i].x2 - dPriv->x;
+        rects[i].y2 = dPriv->pClipRects[i].y2 - dPriv->y;
+    }
+
+    driReportDamage(dPriv, rects, dPriv->numClipRects);
+    _mesa_free(rects);
 }
 
 static int driDrawableGetMSC( __DRIscreen *sPriv, __DRIdrawable *dPriv,
diff --git a/src/mesa/drivers/dri/common/dri_util.h b/src/mesa/drivers/dri/common/dri_util.h
index c0e1bea5e0..c95a5c8299 100644
--- a/src/mesa/drivers/dri/common/dri_util.h
+++ b/src/mesa/drivers/dri/common/dri_util.h
@@ -1,25 +1,3 @@
-/* $XFree86: xc/lib/GL/dri/dri_util.h,v 1.1 2002/02/22 21:32:52 dawes Exp $ */
-/**
- * \file dri_util.h
- * DRI utility functions definitions.
- *
- * This module acts as glue between GLX and the actual hardware driver.  A DRI
- * driver doesn't really \e have to use any of this - it's optional.  But, some
- * useful stuff is done here that otherwise would have to be duplicated in most
- * drivers.
- * 
- * Basically, these utility functions take care of some of the dirty details of
- * screen initialization, context creation, context binding, DRM setup, etc.
- *
- * These functions are compiled into each DRI driver so libGL.so knows nothing
- * about them.
- *
- * \sa dri_util.c.
- * 
- * \author Kevin E. Martin <kevin@precisioninsight.com>
- * \author Brian Paul <brian@precisioninsight.com>
- */
-
 /*
  * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
  * All Rights Reserved.
@@ -45,6 +23,26 @@
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
+/**
+ * \file dri_util.h
+ * DRI utility functions definitions.
+ *
+ * This module acts as glue between GLX and the actual hardware driver.  A DRI
+ * driver doesn't really \e have to use any of this - it's optional.  But, some
+ * useful stuff is done here that otherwise would have to be duplicated in most
+ * drivers.
+ * 
+ * Basically, these utility functions take care of some of the dirty details of
+ * screen initialization, context creation, context binding, DRM setup, etc.
+ *
+ * These functions are compiled into each DRI driver so libGL.so knows nothing
+ * about them.
+ *
+ * \sa dri_util.c.
+ * 
+ * \author Kevin E. Martin <kevin@precisioninsight.com>
+ * \author Brian Paul <brian@precisioninsight.com>
+ */
 
 #ifndef _DRI_UTIL_H_
 #define _DRI_UTIL_H_
@@ -108,6 +106,28 @@ do {                                                                    \
     }                                                                   \
 } while (0)
 
+/**
+ * Same as above, but for two drawables simultaneously.
+ *
+ */
+
+#define DRI_VALIDATE_TWO_DRAWABLES_INFO(psp, pdp, prp)			\
+do {								\
+    while (*((pdp)->pStamp) != (pdp)->lastStamp ||			\
+	   *((prp)->pStamp) != (prp)->lastStamp) {			\
+        register unsigned int hwContext = (psp)->pSAREA->lock.lock &	\
+	    ~(DRM_LOCK_HELD | DRM_LOCK_CONT);				\
+	DRM_UNLOCK((psp)->fd, &(psp)->pSAREA->lock, hwContext);		\
+									\
+	DRM_SPINLOCK(&(psp)->pSAREA->drawable_lock, (psp)->drawLockID);	\
+	DRI_VALIDATE_DRAWABLE_INFO_ONCE(pdp);                           \
+	DRI_VALIDATE_DRAWABLE_INFO_ONCE(prp);				\
+	DRM_SPINUNLOCK(&(psp)->pSAREA->drawable_lock, (psp)->drawLockID); \
+									\
+	DRM_LIGHT_LOCK((psp)->fd, &(psp)->pSAREA->lock, hwContext);	\
+    }                                                                   \
+} while (0)
+
 
 /**
  * Driver callback functions.
diff --git a/src/mesa/drivers/dri/common/drirenderbuffer.c b/src/mesa/drivers/dri/common/drirenderbuffer.c
index b99bf2033b..15af99136c 100644
--- a/src/mesa/drivers/dri/common/drirenderbuffer.c
+++ b/src/mesa/drivers/dri/common/drirenderbuffer.c
@@ -209,6 +209,8 @@ driUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv)
    struct gl_framebuffer *fb = (struct gl_framebuffer *) dPriv->driverPrivate;
    if (fb && (dPriv->w != fb->Width || dPriv->h != fb->Height)) {
       ctx->Driver.ResizeBuffers(ctx, fb, dPriv->w, dPriv->h);
+      /* if the driver needs the hw lock for ResizeBuffers, the drawable
+         might have changed again by now */
       assert(fb->Width == dPriv->w);
       assert(fb->Height == dPriv->h);
    }
diff --git a/src/mesa/drivers/dri/common/extension_helper.h b/src/mesa/drivers/dri/common/extension_helper.h
index b977ebf015..0d641f25f7 100644
--- a/src/mesa/drivers/dri/common/extension_helper.h
+++ b/src/mesa/drivers/dri/common/extension_helper.h
@@ -119,9 +119,10 @@ static const char ActiveTextureARB_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_blit)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_blit)
 static const char BlitFramebufferEXT_names[] = 
     "iiiiiiiiii\0" /* Parameter signature */
+    "glBlitFramebuffer\0"
     "glBlitFramebufferEXT\0"
     "";
 #endif
@@ -433,17 +434,18 @@ static const char Color4ubVertex3fvSUN_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_texture_array)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_texture_array)
 static const char FramebufferTextureLayerEXT_names[] = 
     "iiiii\0" /* Parameter signature */
+    "glFramebufferTextureLayer\0"
     "glFramebufferTextureLayerEXT\0"
     "";
 #endif
 
 #if defined(need_GL_SGIX_list_priority)
-static const char GetListParameterivSGIX_names[] = 
+static const char GetListParameterfvSGIX_names[] = 
     "iip\0" /* Parameter signature */
-    "glGetListParameterivSGIX\0"
+    "glGetListParameterfvSGIX\0"
     "";
 #endif
 
@@ -777,6 +779,13 @@ static const char VertexAttribs4fvNV_names[] =
     "";
 #endif
 
+#if defined(need_GL_NV_vertex_array_range)
+static const char VertexArrayRangeNV_names[] = 
+    "ip\0" /* Parameter signature */
+    "glVertexArrayRangeNV\0"
+    "";
+#endif
+
 #if defined(need_GL_SGIX_fragment_lighting)
 static const char FragmentLightiSGIX_names[] = 
     "iii\0" /* Parameter signature */
@@ -946,9 +955,10 @@ static const char VertexAttribs3fvNV_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char GenerateMipmapEXT_names[] = 
     "i\0" /* Parameter signature */
+    "glGenerateMipmap\0"
     "glGenerateMipmapEXT\0"
     "";
 #endif
@@ -1077,13 +1087,6 @@ static const char AreTexturesResident_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
-static const char IsRenderbufferEXT_names[] = 
-    "i\0" /* Parameter signature */
-    "glIsRenderbufferEXT\0"
-    "";
-#endif
-
 #if defined(need_GL_VERSION_2_0) || defined(need_GL_ATI_separate_stencil)
 static const char StencilOpSeparate_names[] = 
     "iiii\0" /* Parameter signature */
@@ -1171,9 +1174,10 @@ static const char TexCoord2fNormal3fVertex3fSUN_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char RenderbufferStorageEXT_names[] = 
     "iiii\0" /* Parameter signature */
+    "glRenderbufferStorage\0"
     "glRenderbufferStorageEXT\0"
     "";
 #endif
@@ -1364,9 +1368,10 @@ static const char SpriteParameterfvSGIX_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char CheckFramebufferStatusEXT_names[] = 
     "i\0" /* Parameter signature */
+    "glCheckFramebufferStatus\0"
     "glCheckFramebufferStatusEXT\0"
     "";
 #endif
@@ -1554,13 +1559,6 @@ static const char FragmentLightivSGIX_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
-static const char DeleteRenderbuffersEXT_names[] = 
-    "ip\0" /* Parameter signature */
-    "glDeleteRenderbuffersEXT\0"
-    "";
-#endif
-
 #if defined(need_GL_EXT_pixel_transform)
 static const char PixelTransformParameterfvEXT_names[] = 
     "iip\0" /* Parameter signature */
@@ -1621,9 +1619,10 @@ static const char WindowPos2dMESA_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char FramebufferTexture3DEXT_names[] = 
     "iiiiii\0" /* Parameter signature */
+    "glFramebufferTexture3D\0"
     "glFramebufferTexture3DEXT\0"
     "";
 #endif
@@ -1673,10 +1672,11 @@ static const char IndexFuncEXT_names[] =
     "";
 #endif
 
-#if defined(need_GL_SGIX_list_priority)
-static const char GetListParameterfvSGIX_names[] = 
-    "iip\0" /* Parameter signature */
-    "glGetListParameterfvSGIX\0"
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
+static const char FramebufferTexture2DEXT_names[] = 
+    "iiiii\0" /* Parameter signature */
+    "glFramebufferTexture2D\0"
+    "glFramebufferTexture2DEXT\0"
     "";
 #endif
 
@@ -1763,13 +1763,6 @@ static const char DeleteFencesNV_names[] =
     "";
 #endif
 
-#if defined(need_GL_SGIX_polynomial_ffd)
-static const char DeformationMap3dSGIX_names[] = 
-    "iddiiddiiddiip\0" /* Parameter signature */
-    "glDeformationMap3dSGIX\0"
-    "";
-#endif
-
 #if defined(need_GL_VERSION_2_0)
 static const char IsShader_names[] = 
     "i\0" /* Parameter signature */
@@ -1894,6 +1887,13 @@ static const char MatrixIndexuivARB_names[] =
     "";
 #endif
 
+#if defined(need_GL_ARB_framebuffer_object)
+static const char RenderbufferStorageMultisample_names[] = 
+    "iiiii\0" /* Parameter signature */
+    "glRenderbufferStorageMultisample\0"
+    "";
+#endif
+
 #if defined(need_GL_EXT_coordinate_frame)
 static const char Tangent3sEXT_names[] = 
     "iii\0" /* Parameter signature */
@@ -1944,9 +1944,10 @@ static const char VertexAttrib2dvNV_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char FramebufferRenderbufferEXT_names[] = 
     "iiii\0" /* Parameter signature */
+    "glFramebufferRenderbuffer\0"
     "glFramebufferRenderbufferEXT\0"
     "";
 #endif
@@ -1973,9 +1974,10 @@ static const char SetFenceNV_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char FramebufferTexture1DEXT_names[] = 
     "iiiii\0" /* Parameter signature */
+    "glFramebufferTexture1D\0"
     "glFramebufferTexture1DEXT\0"
     "";
 #endif
@@ -2105,9 +2107,10 @@ static const char Tangent3fvEXT_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char BindFramebufferEXT_names[] = 
     "ii\0" /* Parameter signature */
+    "glBindFramebuffer\0"
     "glBindFramebufferEXT\0"
     "";
 #endif
@@ -2488,9 +2491,10 @@ static const char MapParameterivNV_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char GenRenderbuffersEXT_names[] = 
     "ip\0" /* Parameter signature */
+    "glGenRenderbuffers\0"
     "glGenRenderbuffersEXT\0"
     "";
 #endif
@@ -2952,9 +2956,10 @@ static const char WindowPos2ivMESA_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char IsFramebufferEXT_names[] = 
     "i\0" /* Parameter signature */
+    "glIsFramebuffer\0"
     "glIsFramebufferEXT\0"
     "";
 #endif
@@ -3125,13 +3130,6 @@ static const char ReplacementCodeuiSUN_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
-static const char FramebufferTexture2DEXT_names[] = 
-    "iiiii\0" /* Parameter signature */
-    "glFramebufferTexture2DEXT\0"
-    "";
-#endif
-
 #if defined(need_GL_NV_vertex_program)
 static const char VertexAttribPointerNV_names[] = 
     "iiiip\0" /* Parameter signature */
@@ -3139,13 +3137,21 @@ static const char VertexAttribPointerNV_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char GetFramebufferAttachmentParameterivEXT_names[] = 
     "iiip\0" /* Parameter signature */
+    "glGetFramebufferAttachmentParameteriv\0"
     "glGetFramebufferAttachmentParameterivEXT\0"
     "";
 #endif
 
+#if defined(need_GL_EXT_pixel_transform)
+static const char PixelTransformParameterfEXT_names[] = 
+    "iif\0" /* Parameter signature */
+    "glPixelTransformParameterfEXT\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_1_3)
 static const char MultiTexCoord4dvARB_names[] = 
     "ip\0" /* Parameter signature */
@@ -3264,9 +3270,10 @@ static const char VertexAttrib3fvARB_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char DeleteFramebuffersEXT_names[] = 
     "ip\0" /* Parameter signature */
+    "glDeleteFramebuffers\0"
     "glDeleteFramebuffersEXT\0"
     "";
 #endif
@@ -3360,9 +3367,10 @@ static const char GetPixelTexGenParameterfvSGIS_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char GenFramebuffersEXT_names[] = 
     "ip\0" /* Parameter signature */
+    "glGenFramebuffers\0"
     "glGenFramebuffersEXT\0"
     "";
 #endif
@@ -3374,10 +3382,10 @@ static const char GetProgramParameterdvNV_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_pixel_transform)
-static const char PixelTransformParameterfEXT_names[] = 
-    "iif\0" /* Parameter signature */
-    "glPixelTransformParameterfEXT\0"
+#if defined(need_GL_SGIX_instruments)
+static const char PollInstrumentsSGIX_names[] = 
+    "p\0" /* Parameter signature */
+    "glPollInstrumentsSGIX\0"
     "";
 #endif
 
@@ -3709,9 +3717,10 @@ static const char MultiDrawArraysEXT_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char BindRenderbufferEXT_names[] = 
     "ii\0" /* Parameter signature */
+    "glBindRenderbuffer\0"
     "glBindRenderbufferEXT\0"
     "";
 #endif
@@ -4530,6 +4539,13 @@ static const char Minmax_names[] =
     "";
 #endif
 
+#if defined(need_GL_SGIX_polynomial_ffd)
+static const char DeformationMap3dSGIX_names[] = 
+    "iddiiddiiddiip\0" /* Parameter signature */
+    "glDeformationMap3dSGIX\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord)
 static const char FogCoorddvEXT_names[] = 
     "p\0" /* Parameter signature */
@@ -4605,10 +4621,11 @@ static const char WeightdvARB_names[] =
     "";
 #endif
 
-#if defined(need_GL_SGIX_instruments)
-static const char PollInstrumentsSGIX_names[] = 
-    "p\0" /* Parameter signature */
-    "glPollInstrumentsSGIX\0"
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
+static const char DeleteRenderbuffersEXT_names[] = 
+    "ip\0" /* Parameter signature */
+    "glDeleteRenderbuffers\0"
+    "glDeleteRenderbuffersEXT\0"
     "";
 #endif
 
@@ -4684,9 +4701,10 @@ static const char MultiTexCoord2fvARB_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_framebuffer_object)
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
 static const char GetRenderbufferParameterivEXT_names[] = 
     "iip\0" /* Parameter signature */
+    "glGetRenderbufferParameteriv\0"
     "glGetRenderbufferParameterivEXT\0"
     "";
 #endif
@@ -4742,6 +4760,13 @@ static const char Tangent3svEXT_names[] =
     "";
 #endif
 
+#if defined(need_GL_SGIX_list_priority)
+static const char GetListParameterivSGIX_names[] = 
+    "iip\0" /* Parameter signature */
+    "glGetListParameterivSGIX\0"
+    "";
+#endif
+
 #if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object)
 static const char BindBufferARB_names[] = 
     "ii\0" /* Parameter signature */
@@ -4823,10 +4848,11 @@ static const char FragmentMaterialivSGIX_names[] =
     "";
 #endif
 
-#if defined(need_GL_NV_vertex_array_range)
-static const char VertexArrayRangeNV_names[] = 
-    "ip\0" /* Parameter signature */
-    "glVertexArrayRangeNV\0"
+#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object)
+static const char IsRenderbufferEXT_names[] = 
+    "i\0" /* Parameter signature */
+    "glIsRenderbuffer\0"
+    "glIsRenderbufferEXT\0"
     "";
 #endif
 
@@ -4892,6 +4918,32 @@ static const struct dri_extension_function GL_ARB_draw_buffers_functions[] = {
 };
 #endif
 
+#if defined(need_GL_ARB_framebuffer_object)
+static const struct dri_extension_function GL_ARB_framebuffer_object_functions[] = {
+    { BlitFramebufferEXT_names, BlitFramebufferEXT_remap_index, -1 },
+    { FramebufferTextureLayerEXT_names, FramebufferTextureLayerEXT_remap_index, -1 },
+    { GenerateMipmapEXT_names, GenerateMipmapEXT_remap_index, -1 },
+    { RenderbufferStorageEXT_names, RenderbufferStorageEXT_remap_index, -1 },
+    { CheckFramebufferStatusEXT_names, CheckFramebufferStatusEXT_remap_index, -1 },
+    { FramebufferTexture3DEXT_names, FramebufferTexture3DEXT_remap_index, -1 },
+    { FramebufferTexture2DEXT_names, FramebufferTexture2DEXT_remap_index, -1 },
+    { RenderbufferStorageMultisample_names, RenderbufferStorageMultisample_remap_index, -1 },
+    { FramebufferRenderbufferEXT_names, FramebufferRenderbufferEXT_remap_index, -1 },
+    { FramebufferTexture1DEXT_names, FramebufferTexture1DEXT_remap_index, -1 },
+    { BindFramebufferEXT_names, BindFramebufferEXT_remap_index, -1 },
+    { GenRenderbuffersEXT_names, GenRenderbuffersEXT_remap_index, -1 },
+    { IsFramebufferEXT_names, IsFramebufferEXT_remap_index, -1 },
+    { GetFramebufferAttachmentParameterivEXT_names, GetFramebufferAttachmentParameterivEXT_remap_index, -1 },
+    { DeleteFramebuffersEXT_names, DeleteFramebuffersEXT_remap_index, -1 },
+    { GenFramebuffersEXT_names, GenFramebuffersEXT_remap_index, -1 },
+    { BindRenderbufferEXT_names, BindRenderbufferEXT_remap_index, -1 },
+    { DeleteRenderbuffersEXT_names, DeleteRenderbuffersEXT_remap_index, -1 },
+    { GetRenderbufferParameterivEXT_names, GetRenderbufferParameterivEXT_remap_index, -1 },
+    { IsRenderbufferEXT_names, IsRenderbufferEXT_remap_index, -1 },
+    { NULL, 0, 0 }
+};
+#endif
+
 #if defined(need_GL_ARB_matrix_palette)
 static const struct dri_extension_function GL_ARB_matrix_palette_functions[] = {
     { MatrixIndexusvARB_names, MatrixIndexusvARB_remap_index, -1 },
@@ -5319,22 +5371,22 @@ static const struct dri_extension_function GL_EXT_framebuffer_blit_functions[] =
 #if defined(need_GL_EXT_framebuffer_object)
 static const struct dri_extension_function GL_EXT_framebuffer_object_functions[] = {
     { GenerateMipmapEXT_names, GenerateMipmapEXT_remap_index, -1 },
-    { IsRenderbufferEXT_names, IsRenderbufferEXT_remap_index, -1 },
     { RenderbufferStorageEXT_names, RenderbufferStorageEXT_remap_index, -1 },
     { CheckFramebufferStatusEXT_names, CheckFramebufferStatusEXT_remap_index, -1 },
-    { DeleteRenderbuffersEXT_names, DeleteRenderbuffersEXT_remap_index, -1 },
     { FramebufferTexture3DEXT_names, FramebufferTexture3DEXT_remap_index, -1 },
+    { FramebufferTexture2DEXT_names, FramebufferTexture2DEXT_remap_index, -1 },
     { FramebufferRenderbufferEXT_names, FramebufferRenderbufferEXT_remap_index, -1 },
     { FramebufferTexture1DEXT_names, FramebufferTexture1DEXT_remap_index, -1 },
     { BindFramebufferEXT_names, BindFramebufferEXT_remap_index, -1 },
     { GenRenderbuffersEXT_names, GenRenderbuffersEXT_remap_index, -1 },
     { IsFramebufferEXT_names, IsFramebufferEXT_remap_index, -1 },
-    { FramebufferTexture2DEXT_names, FramebufferTexture2DEXT_remap_index, -1 },
     { GetFramebufferAttachmentParameterivEXT_names, GetFramebufferAttachmentParameterivEXT_remap_index, -1 },
     { DeleteFramebuffersEXT_names, DeleteFramebuffersEXT_remap_index, -1 },
     { GenFramebuffersEXT_names, GenFramebuffersEXT_remap_index, -1 },
     { BindRenderbufferEXT_names, BindRenderbufferEXT_remap_index, -1 },
+    { DeleteRenderbuffersEXT_names, DeleteRenderbuffersEXT_remap_index, -1 },
     { GetRenderbufferParameterivEXT_names, GetRenderbufferParameterivEXT_remap_index, -1 },
+    { IsRenderbufferEXT_names, IsRenderbufferEXT_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
@@ -5415,8 +5467,8 @@ static const struct dri_extension_function GL_EXT_paletted_texture_functions[] =
 #if defined(need_GL_EXT_pixel_transform)
 static const struct dri_extension_function GL_EXT_pixel_transform_functions[] = {
     { PixelTransformParameterfvEXT_names, PixelTransformParameterfvEXT_remap_index, -1 },
-    { PixelTransformParameteriEXT_names, PixelTransformParameteriEXT_remap_index, -1 },
     { PixelTransformParameterfEXT_names, PixelTransformParameterfEXT_remap_index, -1 },
+    { PixelTransformParameteriEXT_names, PixelTransformParameteriEXT_remap_index, -1 },
     { PixelTransformParameterivEXT_names, PixelTransformParameterivEXT_remap_index, -1 },
     { NULL, 0, 0 }
 };
@@ -5716,8 +5768,8 @@ static const struct dri_extension_function GL_NV_register_combiners2_functions[]
 
 #if defined(need_GL_NV_vertex_array_range)
 static const struct dri_extension_function GL_NV_vertex_array_range_functions[] = {
-    { FlushVertexArrayRangeNV_names, FlushVertexArrayRangeNV_remap_index, -1 },
     { VertexArrayRangeNV_names, VertexArrayRangeNV_remap_index, -1 },
+    { FlushVertexArrayRangeNV_names, FlushVertexArrayRangeNV_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
@@ -5946,11 +5998,11 @@ static const struct dri_extension_function GL_SGIX_instruments_functions[] = {
 #if defined(need_GL_SGIX_list_priority)
 static const struct dri_extension_function GL_SGIX_list_priority_functions[] = {
     { ListParameterfSGIX_names, ListParameterfSGIX_remap_index, -1 },
-    { GetListParameterivSGIX_names, GetListParameterivSGIX_remap_index, -1 },
     { GetListParameterfvSGIX_names, GetListParameterfvSGIX_remap_index, -1 },
     { ListParameteriSGIX_names, ListParameteriSGIX_remap_index, -1 },
     { ListParameterfvSGIX_names, ListParameterfvSGIX_remap_index, -1 },
     { ListParameterivSGIX_names, ListParameterivSGIX_remap_index, -1 },
+    { GetListParameterivSGIX_names, GetListParameterivSGIX_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
@@ -5965,9 +6017,9 @@ static const struct dri_extension_function GL_SGIX_pixel_texture_functions[] = {
 #if defined(need_GL_SGIX_polynomial_ffd)
 static const struct dri_extension_function GL_SGIX_polynomial_ffd_functions[] = {
     { LoadIdentityDeformationMapSGIX_names, LoadIdentityDeformationMapSGIX_remap_index, -1 },
-    { DeformationMap3dSGIX_names, DeformationMap3dSGIX_remap_index, -1 },
     { DeformSGIX_names, DeformSGIX_remap_index, -1 },
     { DeformationMap3fSGIX_names, DeformationMap3fSGIX_remap_index, -1 },
+    { DeformationMap3dSGIX_names, DeformationMap3dSGIX_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
diff --git a/src/mesa/drivers/dri/common/utils.c b/src/mesa/drivers/dri/common/utils.c
index 30c860b96c..6b44ed9a67 100644
--- a/src/mesa/drivers/dri/common/utils.c
+++ b/src/mesa/drivers/dri/common/utils.c
@@ -179,13 +179,18 @@ driGetRendererString( char * buffer, const char * hardware_name,
 
 
 
+#define need_GL_ARB_draw_buffers
 #define need_GL_ARB_multisample
+#define need_GL_ARB_texture_compression
 #define need_GL_ARB_transpose_matrix
+#define need_GL_ARB_vertex_buffer_object
 #define need_GL_ARB_window_pos
 #define need_GL_EXT_compiled_vertex_array
+#define need_GL_EXT_multi_draw_arrays
 #define need_GL_EXT_polygon_offset
 #define need_GL_EXT_texture_object
 #define need_GL_EXT_vertex_array
+#define need_GL_IBM_multimode_draw_arrays
 #define need_GL_MESA_window_pos
 
 /* These are needed in *all* drivers because Mesa internally implements
@@ -198,14 +203,19 @@ driGetRendererString( char * buffer, const char * hardware_name,
 #include "extension_helper.h"
 
 static const struct dri_extension all_mesa_extensions[] = {
+   { "GL_ARB_draw_buffers",          GL_ARB_draw_buffers_functions },
    { "GL_ARB_multisample",           GL_ARB_multisample_functions },
+   { "GL_ARB_texture_compression",   GL_ARB_texture_compression_functions },
    { "GL_ARB_transpose_matrix",      GL_ARB_transpose_matrix_functions },
+   { "GL_ARB_vertex_buffer_object",  GL_ARB_vertex_buffer_object_functions},
    { "GL_ARB_window_pos",            GL_ARB_window_pos_functions },
    { "GL_EXT_blend_func_separate",   GL_EXT_blend_func_separate_functions },
    { "GL_EXT_compiled_vertex_array", GL_EXT_compiled_vertex_array_functions },
+   { "GL_EXT_multi_draw_arrays",     GL_EXT_multi_draw_arrays_functions },
    { "GL_EXT_polygon_offset",        GL_EXT_polygon_offset_functions },
    { "GL_EXT_texture_object",        GL_EXT_texture_object_functions },
    { "GL_EXT_vertex_array",          GL_EXT_vertex_array_functions },
+   { "GL_IBM_multimode_draw_arrays", GL_IBM_multimode_draw_arrays_functions },
    { "GL_MESA_window_pos",           GL_MESA_window_pos_functions },
    { "GL_NV_vertex_program",         GL_NV_vertex_program_functions },
    { NULL,                           NULL }
@@ -310,8 +320,10 @@ void driInitSingleExtension( GLcontext * ctx,
 	     */
 	    offset = _glapi_add_dispatch( functions, parameter_signature );
 	    if (offset == -1) {
+#if 0 /* this causes noise with egl */
 		fprintf(stderr, "DISPATCH ERROR! _glapi_add_dispatch failed "
 			"to add %s!\n", functions[0]);
+#endif
 	    }
 	    else if (ext->functions[i].remap_index != -1) {
 		driDispatchRemapTable[ ext->functions[i].remap_index ] = 
@@ -504,6 +516,9 @@ GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer,
  *                      \c GLX_SWAP_UNDEFINED_OML.  See the
  *                      GLX_OML_swap_method extension spec for more details.
  * \param num_db_modes  Number of entries in \c db_modes.
+ * \param msaa_samples  Array of msaa sample count. 0 represents a visual
+ *                      without a multisample buffer.
+ * \param num_msaa_modes Number of entries in \c msaa_samples.
  * \param visType       GLX visual type.  Usually either \c GLX_TRUE_COLOR or
  *                      \c GLX_DIRECT_COLOR.
  * 
@@ -523,7 +538,8 @@ __DRIconfig **
 driCreateConfigs(GLenum fb_format, GLenum fb_type,
 		 const uint8_t * depth_bits, const uint8_t * stencil_bits,
 		 unsigned num_depth_stencil_bits,
-		 const GLenum * db_modes, unsigned num_db_modes)
+		 const GLenum * db_modes, unsigned num_db_modes,
+	 	 const u_int8_t * msaa_samples, unsigned num_msaa_modes)
 {
    static const uint8_t bits_table[4][4] = {
      /* R  G  B  A */
@@ -583,9 +599,7 @@ driCreateConfigs(GLenum fb_format, GLenum fb_type,
    int index;
    __DRIconfig **configs, **c;
    __GLcontextModes *modes;
-   unsigned i;
-   unsigned j;
-   unsigned k;
+   unsigned i, j, k, h;
    unsigned num_modes;
    unsigned num_accum_bits = 2;
 
@@ -658,7 +672,7 @@ driCreateConfigs(GLenum fb_format, GLenum fb_type,
 	 break;
    }
 
-   num_modes = num_depth_stencil_bits * num_db_modes * num_accum_bits;
+   num_modes = num_depth_stencil_bits * num_db_modes * num_accum_bits * num_msaa_modes;
    configs = _mesa_calloc((num_modes + 1) * sizeof *configs);
    if (configs == NULL)
        return NULL;
@@ -666,66 +680,72 @@ driCreateConfigs(GLenum fb_format, GLenum fb_type,
     c = configs;
     for ( k = 0 ; k < num_depth_stencil_bits ; k++ ) {
 	for ( i = 0 ; i < num_db_modes ; i++ ) {
-	    for ( j = 0 ; j < num_accum_bits ; j++ ) {
-		*c = _mesa_malloc (sizeof **c);
-		modes = &(*c)->modes;
-		c++;
-
-		memset(modes, 0, sizeof *modes);
-		modes->redBits   = bits[0];
-		modes->greenBits = bits[1];
-		modes->blueBits  = bits[2];
-		modes->alphaBits = bits[3];
-		modes->redMask   = masks[0];
-		modes->greenMask = masks[1];
-		modes->blueMask  = masks[2];
-		modes->alphaMask = masks[3];
-		modes->rgbBits   = modes->redBits + modes->greenBits
-		    + modes->blueBits + modes->alphaBits;
-
-		modes->accumRedBits   = 16 * j;
-		modes->accumGreenBits = 16 * j;
-		modes->accumBlueBits  = 16 * j;
-		modes->accumAlphaBits = (masks[3] != 0) ? 16 * j : 0;
-		modes->visualRating = (j == 0) ? GLX_NONE : GLX_SLOW_CONFIG;
-
-		modes->stencilBits = stencil_bits[k];
-		modes->depthBits = depth_bits[k];
-
-		modes->transparentPixel = GLX_NONE;
-		modes->transparentRed = GLX_DONT_CARE;
-		modes->transparentGreen = GLX_DONT_CARE;
-		modes->transparentBlue = GLX_DONT_CARE;
-		modes->transparentAlpha = GLX_DONT_CARE;
-		modes->transparentIndex = GLX_DONT_CARE;
-		modes->visualType = GLX_DONT_CARE;
-		modes->renderType = GLX_RGBA_BIT;
-		modes->drawableType = GLX_WINDOW_BIT;
-		modes->rgbMode = GL_TRUE;
-
-		if ( db_modes[i] == GLX_NONE ) {
-		    modes->doubleBufferMode = GL_FALSE;
-		}
-		else {
-		    modes->doubleBufferMode = GL_TRUE;
-		    modes->swapMethod = db_modes[i];
-		}
-
-		modes->haveAccumBuffer = ((modes->accumRedBits +
+	    for ( h = 0 ; h < num_msaa_modes; h++ ) {
+	    	for ( j = 0 ; j < num_accum_bits ; j++ ) {
+		    *c = _mesa_malloc (sizeof **c);
+		    modes = &(*c)->modes;
+		    c++;
+
+		    memset(modes, 0, sizeof *modes);
+		    modes->redBits   = bits[0];
+		    modes->greenBits = bits[1];
+		    modes->blueBits  = bits[2];
+		    modes->alphaBits = bits[3];
+		    modes->redMask   = masks[0];
+		    modes->greenMask = masks[1];
+		    modes->blueMask  = masks[2];
+		    modes->alphaMask = masks[3];
+		    modes->rgbBits   = modes->redBits + modes->greenBits
+		    	+ modes->blueBits + modes->alphaBits;
+
+		    modes->accumRedBits   = 16 * j;
+		    modes->accumGreenBits = 16 * j;
+		    modes->accumBlueBits  = 16 * j;
+		    modes->accumAlphaBits = (masks[3] != 0) ? 16 * j : 0;
+		    modes->visualRating = (j == 0) ? GLX_NONE : GLX_SLOW_CONFIG;
+
+		    modes->stencilBits = stencil_bits[k];
+		    modes->depthBits = depth_bits[k];
+
+		    modes->transparentPixel = GLX_NONE;
+		    modes->transparentRed = GLX_DONT_CARE;
+		    modes->transparentGreen = GLX_DONT_CARE;
+		    modes->transparentBlue = GLX_DONT_CARE;
+		    modes->transparentAlpha = GLX_DONT_CARE;
+		    modes->transparentIndex = GLX_DONT_CARE;
+		    modes->visualType = GLX_DONT_CARE;
+		    modes->renderType = GLX_RGBA_BIT;
+		    modes->drawableType = GLX_WINDOW_BIT;
+		    modes->rgbMode = GL_TRUE;
+
+		    if ( db_modes[i] == GLX_NONE ) {
+		    	modes->doubleBufferMode = GL_FALSE;
+		    }
+		    else {
+		    	modes->doubleBufferMode = GL_TRUE;
+		    	modes->swapMethod = db_modes[i];
+		    }
+
+		    modes->samples = msaa_samples[h];
+		    modes->sampleBuffers = modes->samples ? 1 : 0;
+
+
+		    modes->haveAccumBuffer = ((modes->accumRedBits +
 					   modes->accumGreenBits +
 					   modes->accumBlueBits +
 					   modes->accumAlphaBits) > 0);
-		modes->haveDepthBuffer = (modes->depthBits > 0);
-		modes->haveStencilBuffer = (modes->stencilBits > 0);
-
-		modes->bindToTextureRgb = GL_TRUE;
-		modes->bindToTextureRgba = GL_TRUE;
-		modes->bindToMipmapTexture = GL_FALSE;
-		modes->bindToTextureTargets = modes->rgbMode ?
-		    __DRI_ATTRIB_TEXTURE_1D_BIT |
-		    __DRI_ATTRIB_TEXTURE_2D_BIT |
-		    __DRI_ATTRIB_TEXTURE_RECTANGLE_BIT :
-		    0;
+		    modes->haveDepthBuffer = (modes->depthBits > 0);
+		    modes->haveStencilBuffer = (modes->stencilBits > 0);
+
+		    modes->bindToTextureRgb = GL_TRUE;
+		    modes->bindToTextureRgba = GL_TRUE;
+		    modes->bindToMipmapTexture = GL_FALSE;
+		    modes->bindToTextureTargets = modes->rgbMode ?
+		    	__DRI_ATTRIB_TEXTURE_1D_BIT |
+		    	__DRI_ATTRIB_TEXTURE_2D_BIT |
+		    	__DRI_ATTRIB_TEXTURE_RECTANGLE_BIT :
+		    	0;
+		}
 	    }
 	}
     }
@@ -734,9 +754,10 @@ driCreateConfigs(GLenum fb_format, GLenum fb_type,
     return configs;
 }
 
-const __DRIconfig **driConcatConfigs(__DRIconfig **a, __DRIconfig **b)
+__DRIconfig **driConcatConfigs(__DRIconfig **a,
+			       __DRIconfig **b)
 {
-    const __DRIconfig **all;
+    __DRIconfig **all;
     int i, j, index;
 
     i = 0;
diff --git a/src/mesa/drivers/dri/common/utils.h b/src/mesa/drivers/dri/common/utils.h
index 0c974dbff3..9e9e5bc224 100644
--- a/src/mesa/drivers/dri/common/utils.h
+++ b/src/mesa/drivers/dri/common/utils.h
@@ -131,9 +131,11 @@ extern __DRIconfig **
 driCreateConfigs(GLenum fb_format, GLenum fb_type,
 		 const uint8_t * depth_bits, const uint8_t * stencil_bits,
 		 unsigned num_depth_stencil_bits,
-		 const GLenum * db_modes, unsigned num_db_modes);
+		 const GLenum * db_modes, unsigned num_db_modes,
+    		 const uint8_t * msaa_samples, unsigned num_msaa_modes);
 
-const __DRIconfig **driConcatConfigs(__DRIconfig **a, __DRIconfig **b);
+__DRIconfig **driConcatConfigs(__DRIconfig **a,
+			       __DRIconfig **b);
 
 int
 driGetConfigAttrib(const __DRIconfig *config,
diff --git a/src/mesa/drivers/dri/common/vblank.c b/src/mesa/drivers/dri/common/vblank.c
index d610253fe6..12aeaa108f 100644
--- a/src/mesa/drivers/dri/common/vblank.c
+++ b/src/mesa/drivers/dri/common/vblank.c
@@ -130,9 +130,8 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv,
 
 
    if ( divisor != 0 ) {
-      unsigned int target = (unsigned int)target_msc;
-      unsigned int next = target;
-      unsigned int r;
+      int64_t next = target_msc;
+      int64_t r;
       int dont_wait = (target_msc == 0);
 
       do {
@@ -154,9 +153,9 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv,
 
 	 *msc = vblank_to_msc(priv, vbl.reply.sequence);
 
-         dont_wait = 0;
-         if (target_msc != 0 && *msc == target)
+         if (!dont_wait && *msc == next)
             break;
+         dont_wait = 0;
 
          /* Assuming the wait-done test fails, the next refresh to wait for
           * will be one that satisfies (MSC % divisor) == remainder.  The
@@ -165,11 +164,12 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv,
           * If this refresh has already happened, we add divisor to obtain 
           * the next refresh after the current one that will satisfy it.
           */
-         r = (*msc % (unsigned int)divisor);
-         next = (*msc - r + (unsigned int)remainder);
-         if (next <= *msc) next += (unsigned int)divisor;
+         r = ((uint64_t)*msc % divisor);
+         next = (*msc - r + remainder);
+         if (next <= *msc)
+	    next += divisor;
 
-      } while ( r != (unsigned int)remainder );
+      } while (r != remainder);
    }
    else {
       /* If the \c divisor is zero, just wait until the MSC is greater
diff --git a/src/mesa/drivers/dri/dri.pc.in b/src/mesa/drivers/dri/dri.pc.in
index c47ee9c7e7..695aa6cfd6 100644
--- a/src/mesa/drivers/dri/dri.pc.in
+++ b/src/mesa/drivers/dri/dri.pc.in
@@ -7,4 +7,5 @@ dridriverdir=@DRI_DRIVER_DIR@
 Name: dri
 Description: Direct Rendering Infrastructure
 Version: @VERSION@
+Requires.private: @DRI_PC_REQ_PRIV@
 Cflags: -I${includedir}
diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.c b/src/mesa/drivers/dri/ffb/ffb_xmesa.c
index 679f8561d2..00bdcec7a6 100644
--- a/src/mesa/drivers/dri/ffb/ffb_xmesa.c
+++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.c
@@ -260,6 +260,8 @@ ffbCreateContext(const __GLcontextModes *mesaVis,
 	ctx->Const.MaxLineWidthAA = 1.0;
 	ctx->Const.LineWidthGranularity = 1.0;
 
+	ctx->Const.MaxDrawBuffers = 1;
+
 	/* Instead of having GCC emit these constants a zillion times
 	 * everywhere in the driver, put them here.
 	 */
@@ -626,6 +628,7 @@ ffbFillInModes( __DRIscreenPrivate *psp,
 
    uint8_t depth_bits_array[3];
    uint8_t stencil_bits_array[3];
+   uint8_t msaa_samples_array[1];
 
    depth_bits_array[0] = 0;
    depth_bits_array[1] = depth_bits;
@@ -639,6 +642,8 @@ ffbFillInModes( __DRIscreenPrivate *psp,
    stencil_bits_array[1] = 0;
    stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+   msaa_samples_array[0] = 0;
+
    depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
    back_buffer_factor  = (have_back_buffer) ? 3 : 1;
 
@@ -654,9 +659,10 @@ ffbFillInModes( __DRIscreenPrivate *psp,
    configs = driCreateConfigs(fb_format, fb_type,
 			      depth_bits_array, stencil_bits_array,
 			      depth_buffer_factor, back_buffer_modes,
-			      back_buffer_factor);
+			      back_buffer_factor,
+                               msaa_samples_array, 1);
    if (configs == NULL) {
-    fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
+      fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
               __LINE__);
       return NULL;
    }
diff --git a/src/mesa/drivers/dri/gamma/gamma_context.c b/src/mesa/drivers/dri/gamma/gamma_context.c
index c91bedce3a..b0ac299daa 100644
--- a/src/mesa/drivers/dri/gamma/gamma_context.c
+++ b/src/mesa/drivers/dri/gamma/gamma_context.c
@@ -133,6 +133,8 @@ GLboolean gammaCreateContext( const __GLcontextModes *glVisual,
    ctx->Const.MaxPointSizeAA = 16.0; 
    ctx->Const.PointSizeGranularity = 0.25;
 
+   ctx->Const.MaxDrawBuffers = 1;
+
    gmesa->texHeap = mmInit( 0, gmesa->gammaScreen->textureSize );
 
    make_empty_list(&gmesa->TexObjList);
diff --git a/src/mesa/drivers/dri/glcore/Makefile b/src/mesa/drivers/dri/glcore/Makefile
new file mode 100644
index 0000000000..ac7e1de928
--- /dev/null
+++ b/src/mesa/drivers/dri/glcore/Makefile
@@ -0,0 +1,84 @@
+# src/mesa/drivers/dri/glcore/Makefile
+
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = glcore_dri.so
+
+DRIVER_SOURCES = glcore_driver.c \
+		 $(TOP)/src/mesa/drivers/common/driverfuncs.c \
+		 ../common/dri_util.c
+
+C_SOURCES = \
+	$(DRIVER_SOURCES) \
+	$(DRI_SOURCES) 
+
+
+# Include directories
+INCLUDE_DIRS = \
+	-I. \
+	-I../common \
+	-I../dri_client \
+	-I../dri_client/imports \
+	-Iserver \
+	-I$(TOP)/include \
+	-I$(DRM_SOURCE_PATH)/shared-core \
+	-I$(TOP)/src/mesa \
+	-I$(TOP)/src/mesa/main \
+	-I$(TOP)/src/mesa/glapi \
+	-I$(TOP)/src/mesa/math \
+	-I$(TOP)/src/mesa/transform \
+	-I$(TOP)/src/mesa/shader \
+	-I$(TOP)/src/mesa/swrast \
+	-I$(TOP)/src/mesa/swrast_setup
+
+# Core Mesa objects
+MESA_MODULES = $(TOP)/src/mesa/libmesa.a
+
+# Libraries that the driver shared lib depends on
+LIB_DEPS = -lm -lpthread -lc
+# LIB_DEPS = -lGL -lm -lpthread -lc
+
+
+ASM_SOURCES = 
+
+OBJECTS = $(C_SOURCES:.c=.o) \
+	  $(ASM_SOURCES:.S=.o) 
+
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(DEFINES) $< -o $@
+
+.S.o:
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(DEFINES)  $< -o $@
+
+
+##### TARGETS #####
+
+default: depend $(TOP)/$(LIB_DIR)/$(LIBNAME)
+
+
+$(TOP)/$(LIB_DIR)/$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(WINOBJ) Makefile
+	CC="$(CC)" CXX="$(CXX)" $(TOP)/bin/mklib -o $(LIBNAME) -noprefix -install $(TOP)/$(LIB_DIR) \
+		$(OBJECTS) $(WINLIB) $(LIB_DEPS) $(WINOBJ) $(MESA_MODULES)
+
+
+depend: $(C_SOURCES) $(ASM_SOURCES)
+	rm -f depend
+	touch depend
+	$(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(C_SOURCES) $(ASM_SOURCES) \
+		> /dev/null 
+
+
+# Emacs tags
+tags:
+	etags `find . -name \*.[ch]` `find ../include`
+
+
+clean:
+	-rm -f *.o server/*.o
+
+
+include depend
diff --git a/src/mesa/drivers/dri/i810/i810context.c b/src/mesa/drivers/dri/i810/i810context.c
index c281a4990e..6785655686 100644
--- a/src/mesa/drivers/dri/i810/i810context.c
+++ b/src/mesa/drivers/dri/i810/i810context.c
@@ -63,11 +63,6 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "drirenderbuffer.h"
 #include "utils.h"
 
-#define need_GL_ARB_multisample
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
-#include "extension_helper.h"
-
 #ifndef I810_DEBUG
 int I810_DEBUG = (0);
 #endif
@@ -123,14 +118,11 @@ static void i810BufferSize(GLframebuffer *buffer, GLuint *width, GLuint *height)
  */
 const struct dri_extension card_extensions[] =
 {
-    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
     { "GL_ARB_multitexture",               NULL },
-    { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
     { "GL_ARB_texture_env_add",            NULL },
     { "GL_ARB_texture_env_combine",        NULL },
     { "GL_ARB_texture_env_crossbar",       NULL },
     { "GL_ARB_texture_mirrored_repeat",    NULL },
-    { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
     { "GL_EXT_stencil_wrap",               NULL },
     { "GL_EXT_texture_edge_clamp",         NULL },
     { "GL_EXT_texture_env_combine",        NULL },
diff --git a/src/mesa/drivers/dri/i810/i810screen.c b/src/mesa/drivers/dri/i810/i810screen.c
index 48603f5d79..9a5a39c1db 100644
--- a/src/mesa/drivers/dri/i810/i810screen.c
+++ b/src/mesa/drivers/dri/i810/i810screen.c
@@ -77,6 +77,7 @@ i810FillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[2];
     uint8_t stencil_bits_array[2];
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = depth_bits;
     depth_bits_array[1] = depth_bits;
@@ -88,13 +89,16 @@ i810FillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[0] = 0;
     stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 2 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
     configs = driCreateConfigs(GL_RGB, GL_UNSIGNED_SHORT_5_6_5,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor,
-			       back_buffer_modes, back_buffer_factor);
+			       back_buffer_modes, back_buffer_factor,
+                               msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
 		 __func__, __LINE__ );
diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile
index 5858e0ee9f..954a7e2af1 100644
--- a/src/mesa/drivers/dri/i915/Makefile
+++ b/src/mesa/drivers/dri/i915/Makefile
@@ -18,8 +18,9 @@ DRIVER_SOURCES = \
 	intel_regions.c \
 	intel_buffer_objects.c \
 	intel_batchbuffer.c \
+	intel_clear.c \
+	intel_extensions.c \
 	intel_mipmap_tree.c \
-	i915_tex_layout.c \
 	intel_tex_layout.c \
 	intel_tex_image.c \
 	intel_tex_subimage.c \
@@ -34,7 +35,9 @@ DRIVER_SOURCES = \
 	intel_pixel_read.c \
 	intel_buffers.c \
 	intel_blit.c \
+	intel_swapbuffers.c \
 	i915_tex.c \
+	i915_tex_layout.c \
 	i915_texstate.c \
 	i915_context.c \
 	i915_debug.c \
@@ -50,8 +53,7 @@ DRIVER_SOURCES = \
 	intel_span.c \
 	intel_state.c \
 	intel_tris.c \
-	intel_fbo.c \
-	intel_depthstencil.c
+	intel_fbo.c
 
 C_SOURCES = \
 	$(COMMON_SOURCES) \
diff --git a/src/mesa/drivers/dri/i915/i830_context.c b/src/mesa/drivers/dri/i915/i830_context.c
index 09b1ec922f..fdce8af31f 100644
--- a/src/mesa/drivers/dri/i915/i830_context.c
+++ b/src/mesa/drivers/dri/i915/i830_context.c
@@ -97,6 +97,8 @@ i830CreateContext(const __GLcontextModes * mesaVis,
    ctx->Const.MaxTextureRectSize = (1 << 11);
    ctx->Const.MaxTextureUnits = I830_TEX_UNITS;
 
+   ctx->Const.MaxDrawBuffers = 1;
+
    _tnl_init_vertices(ctx, ctx->Const.MaxArrayLockSize + 12,
                       18 * sizeof(GLfloat));
 
diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c
index 3b3ff2bced..8fc8aa5f90 100644
--- a/src/mesa/drivers/dri/i915/i830_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i830_vtbl.c
@@ -297,7 +297,7 @@ i830_emit_invarient_state(struct intel_context *intel)
 {
    BATCH_LOCALS;
 
-   BEGIN_BATCH(40, IGNORE_CLIPRECTS);
+   BEGIN_BATCH(30, IGNORE_CLIPRECTS);
 
    OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD);
    OUT_BATCH(0);
@@ -491,8 +491,17 @@ i830_emit_state(struct intel_context *intel)
    }
 
    if (dirty & I830_UPLOAD_BUFFERS) {
+      GLuint count = 9; 
+
       DBG("I830_UPLOAD_BUFFERS:\n");
-      BEGIN_BATCH(I830_DEST_SETUP_SIZE + 2, IGNORE_CLIPRECTS);
+
+      if (state->depth_region)
+          count += 3;
+
+      if (intel->constant_cliprect)
+          count += 6;
+
+      BEGIN_BATCH(count, IGNORE_CLIPRECTS);
       OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR0]);
       OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR1]);
       OUT_RELOC(state->draw_region->buffer,
@@ -557,6 +566,8 @@ i830_emit_state(struct intel_context *intel)
          OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S4]);
          OUT_BATCH(state->Tex[i][I830_TEXREG_MCS]);
          OUT_BATCH(state->Tex[i][I830_TEXREG_CUBE]);
+
+         ADVANCE_BATCH();
       }
 
       if (dirty & I830_UPLOAD_TEXBLEND(i)) {
diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c
index e0ddc7fd61..6e2d41e19a 100644
--- a/src/mesa/drivers/dri/i915/i915_context.c
+++ b/src/mesa/drivers/dri/i915/i915_context.c
@@ -50,15 +50,6 @@
  * Mesa's Driver Functions
  ***************************************/
 
-static const struct dri_extension i915_extensions[] = {
-   {"GL_ARB_depth_texture", NULL},
-   {"GL_ARB_fragment_program", NULL},
-   {"GL_ARB_shadow", NULL},
-   {"GL_ARB_texture_non_power_of_two", NULL},
-   {"GL_EXT_shadow_funcs", NULL},
-   {NULL, NULL}
-};
-
 /* Override intel default.
  */
 static void
@@ -170,10 +161,8 @@ i915CreateContext(const __GLcontextModes * mesaVis,
    ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; /* I don't think we have one */
 
    ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
-   ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE;
-
-   driInitExtensions(ctx, i915_extensions, GL_FALSE);
 
+   ctx->Const.MaxDrawBuffers = 1;
 
    _tnl_init_vertices(ctx, ctx->Const.MaxArrayLockSize + 12,
                       36 * sizeof(GLfloat));
diff --git a/src/mesa/drivers/dri/i915/i915_fragprog.c b/src/mesa/drivers/dri/i915/i915_fragprog.c
index 8bd761ec6a..f091d600c3 100644
--- a/src/mesa/drivers/dri/i915/i915_fragprog.c
+++ b/src/mesa/drivers/dri/i915/i915_fragprog.c
@@ -33,6 +33,7 @@
 #include "shader/prog_parameter.h"
 #include "shader/program.h"
 #include "shader/programopt.h"
+#include "shader/prog_print.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
@@ -1048,6 +1049,9 @@ i915ProgramStringNotify(GLcontext * ctx,
          _mesa_append_fog_code(ctx, &p->FragProg);
          p->FragProg.FogOption = GL_NONE;
       }
+
+      if (INTEL_DEBUG & DEBUG_STATE)
+	 _mesa_print_program(prog);
    }
 
    _tnl_program_string(ctx, target, prog);
@@ -1105,30 +1109,14 @@ i915ValidateFragmentProgram(struct i915_context *i915)
       EMIT_ATTR(_TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, S4_VFMT_COLOR, 4);
    }
 
-   if ((inputsRead & (FRAG_BIT_COL1 | FRAG_BIT_FOGC)) ||
-       i915->vertex_fog != I915_FOG_NONE) {
-
-      if (inputsRead & FRAG_BIT_COL1) {
-         intel->specoffset = offset / 4;
-         EMIT_ATTR(_TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, S4_VFMT_SPEC_FOG, 3);
-      }
-      else
-         EMIT_PAD(3);
-
-      if ((inputsRead & FRAG_BIT_FOGC) || i915->vertex_fog != I915_FOG_NONE)
-         EMIT_ATTR(_TNL_ATTRIB_FOG, EMIT_1UB_1F, S4_VFMT_SPEC_FOG, 1);
-      else
-         EMIT_PAD(1);
+   if (inputsRead & FRAG_BIT_COL1) {
+       intel->specoffset = offset / 4;
+       EMIT_ATTR(_TNL_ATTRIB_COLOR1, EMIT_4UB_4F_BGRA, S4_VFMT_SPEC_FOG, 4);
    }
 
-   /* XXX this was disabled, but enabling this code helped fix the Glean
-    * tfragprog1 fog tests.
-    */
-#if 1
    if ((inputsRead & FRAG_BIT_FOGC) || i915->vertex_fog != I915_FOG_NONE) {
       EMIT_ATTR(_TNL_ATTRIB_FOG, EMIT_1F, S4_VFMT_FOG_PARAM, 4);
    }
-#endif
 
    for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) {
       if (inputsRead & FRAG_BIT_TEX(i)) {
diff --git a/src/mesa/drivers/dri/i915/i915_state.c b/src/mesa/drivers/dri/i915/i915_state.c
index 9d04358e4f..814fb59fd3 100644
--- a/src/mesa/drivers/dri/i915/i915_state.c
+++ b/src/mesa/drivers/dri/i915/i915_state.c
@@ -41,6 +41,7 @@
 #include "intel_fbo.h"
 #include "intel_screen.h"
 #include "intel_batchbuffer.h"
+#include "intel_buffers.h"
 
 #include "i915_context.h"
 #include "i915_reg.h"
@@ -301,6 +302,65 @@ i915DepthMask(GLcontext * ctx, GLboolean flag)
       i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_DEPTH_WRITE_ENABLE;
 }
 
+
+
+/**
+ * Update the viewport transformation matrix.  Depends on:
+ *  - viewport pos/size
+ *  - depthrange
+ *  - window pos/size or FBO size
+ */
+void
+intelCalcViewport(GLcontext * ctx)
+{
+   struct intel_context *intel = intel_context(ctx);
+   const GLfloat *v = ctx->Viewport._WindowMap.m;
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   GLfloat *m = intel->ViewportMatrix.m;
+   GLfloat yScale, yBias;
+
+   if (ctx->DrawBuffer->Name) {
+      /* User created FBO */
+      /* y=0=bottom */
+      yScale = 1.0;
+      yBias = 0.0;
+   }
+   else {
+      /* window buffer, y=0=top */
+      yScale = -1.0;
+      yBias = (intel->driDrawable) ? intel->driDrawable->h : 0.0F;
+   }
+
+   m[MAT_SX] = v[MAT_SX];
+   m[MAT_TX] = v[MAT_TX];
+
+   m[MAT_SY] = v[MAT_SY] * yScale;
+   m[MAT_TY] = v[MAT_TY] * yScale + yBias;
+
+   m[MAT_SZ] = v[MAT_SZ] * depthScale;
+   m[MAT_TZ] = v[MAT_TZ] * depthScale;
+}
+
+
+/** Called from ctx->Driver.Viewport() */
+static void
+i915Viewport(GLcontext * ctx,
+              GLint x, GLint y, GLsizei width, GLsizei height)
+{
+   intelCalcViewport(ctx);
+
+   intel_viewport(ctx, x, y, width, height);
+}
+
+
+/** Called from ctx->Driver.DepthRange() */
+static void
+i915DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval)
+{
+   intelCalcViewport(ctx);
+}
+
+
 /* =============================================================
  * Polygon stipple
  *
@@ -569,7 +629,7 @@ i915_update_fog(GLcontext * ctx)
    GLboolean enabled;
    GLboolean try_pixel_fog;
 
-   if (ctx->FragmentProgram._Active) {
+   if (ctx->FragmentProgram._Current) {
       /* Pull in static fog state from program */
       mode = ctx->FragmentProgram._Current->FogOption;
       enabled = (mode != GL_NONE);
@@ -964,6 +1024,8 @@ i915InitStateFunctions(struct dd_function_table *functions)
    functions->StencilFuncSeparate = i915StencilFuncSeparate;
    functions->StencilMaskSeparate = i915StencilMaskSeparate;
    functions->StencilOpSeparate = i915StencilOpSeparate;
+   functions->DepthRange = i915DepthRange;
+   functions->Viewport = i915Viewport;
 }
 
 
diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c
index d1b0dcdf31..adbb52a3a3 100644
--- a/src/mesa/drivers/dri/i915/i915_texstate.c
+++ b/src/mesa/drivers/dri/i915/i915_texstate.c
@@ -295,6 +295,13 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
            wt == GL_CLAMP_TO_BORDER || wr == GL_CLAMP_TO_BORDER))
          return GL_FALSE;
 
+      /* Only support TEXCOORDMODE_CLAMP_EDGE and TEXCOORDMODE_CUBE (not 
+       * used) when using cube map texture coordinates
+       */
+      if (tObj->Target == GL_TEXTURE_CUBE_MAP_ARB &&
+          (((ws != GL_CLAMP) && (ws != GL_CLAMP_TO_EDGE)) ||
+           ((wt != GL_CLAMP) && (wt != GL_CLAMP_TO_EDGE))))
+          return GL_FALSE;
 
       state[I915_TEXREG_SS3] = ss3;     /* SS3_NORMALIZED_COORDS */
 
diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c
index e79c955d64..3f6d282d34 100644
--- a/src/mesa/drivers/dri/i915/i915_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i915_vtbl.c
@@ -173,7 +173,7 @@ i915_emit_invarient_state(struct intel_context *intel)
 {
    BATCH_LOCALS;
 
-   BEGIN_BATCH(200, IGNORE_CLIPRECTS);
+   BEGIN_BATCH(20, IGNORE_CLIPRECTS);
 
    OUT_BATCH(_3DSTATE_AA_CMD |
              AA_LINE_ECAAR_WIDTH_ENABLE |
@@ -376,9 +376,18 @@ i915_emit_state(struct intel_context *intel)
    }
 
    if (dirty & I915_UPLOAD_BUFFERS) {
+      GLuint count = 9;
+
       if (INTEL_DEBUG & DEBUG_STATE)
          fprintf(stderr, "I915_UPLOAD_BUFFERS:\n");
-      BEGIN_BATCH(I915_DEST_SETUP_SIZE + 2, IGNORE_CLIPRECTS);
+
+      if (state->depth_region)
+          count += 3;
+
+      if (intel->constant_cliprect)
+          count += 6;
+
+      BEGIN_BATCH(count, IGNORE_CLIPRECTS);
       OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR0]);
       OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR1]);
       OUT_RELOC(state->draw_region->buffer,
diff --git a/src/mesa/drivers/dri/i915/intel_clear.c b/src/mesa/drivers/dri/i915/intel_clear.c
new file mode 120000
index 0000000000..9a2a742a0d
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_clear.c
@@ -0,0 +1 @@
+../intel/intel_clear.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i915/intel_depthstencil.c b/src/mesa/drivers/dri/i915/intel_depthstencil.c
deleted file mode 120000
index 4ac4ae690a..0000000000
--- a/src/mesa/drivers/dri/i915/intel_depthstencil.c
+++ /dev/null
@@ -1 +0,0 @@
-../intel/intel_depthstencil.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i915/intel_extensions.c b/src/mesa/drivers/dri/i915/intel_extensions.c
new file mode 120000
index 0000000000..a2f3e8cd20
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_extensions.c
@@ -0,0 +1 @@
+../intel/intel_extensions.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i915/intel_render.c b/src/mesa/drivers/dri/i915/intel_render.c
index 467abe4d7c..410052b3c2 100644
--- a/src/mesa/drivers/dri/i915/intel_render.c
+++ b/src/mesa/drivers/dri/i915/intel_render.c
@@ -117,6 +117,26 @@ intelDmaPrimitive(struct intel_context *intel, GLenum prim)
    intel_set_prim(intel, hw_prim[prim]);
 }
 
+static inline GLuint intel_get_vb_max(struct intel_context *intel)
+{
+   GLuint ret;
+
+   if (intel->intelScreen->no_vbo)
+      ret = intel->batch->size - 1500;
+   else
+      ret = INTEL_VB_SIZE;
+   ret /= (intel->vertex_size * 4);
+   return ret;
+}
+
+static inline GLuint intel_get_current_max(struct intel_context *intel)
+{
+
+   if (intel->intelScreen->no_vbo)
+      return intel_get_vb_max(intel);
+   else
+      return (INTEL_VB_SIZE - intel->prim.current_offset) / (intel->vertex_size * 4);
+}
 
 #define LOCAL_VARS struct intel_context *intel = intel_context(ctx)
 #define INIT( prim ) 				\
@@ -126,9 +146,8 @@ do {						\
 
 #define FLUSH() INTEL_FIREVERTICES(intel)
 
-#define GET_SUBSEQUENT_VB_MAX_VERTS() (INTEL_VB_SIZE / (intel->vertex_size * 4))
-#define GET_CURRENT_VB_MAX_VERTS() \
-   ((INTEL_VB_SIZE - intel->prim.current_offset) / (intel->vertex_size * 4))
+#define GET_SUBSEQUENT_VB_MAX_VERTS() intel_get_vb_max(intel)
+#define GET_CURRENT_VB_MAX_VERTS() intel_get_current_max(intel)
 
 #define ALLOC_VERTS(nr) intel_get_prim_space(intel, nr)
 
diff --git a/src/mesa/drivers/dri/i915/intel_state.c b/src/mesa/drivers/dri/i915/intel_state.c
index 09aa62da36..519672fc35 100644..120000
--- a/src/mesa/drivers/dri/i915/intel_state.c
+++ b/src/mesa/drivers/dri/i915/intel_state.c
@@ -1,295 +1 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-
-#include "main/glheader.h"
-#include "main/context.h"
-#include "main/macros.h"
-#include "main/enums.h"
-#include "main/colormac.h"
-#include "main/dd.h"
-
-#include "intel_screen.h"
-#include "intel_context.h"
-#include "intel_fbo.h"
-#include "intel_regions.h"
-#include "swrast/swrast.h"
-
-int 
-intel_translate_shadow_compare_func( GLenum func )
-{
-   switch(func) {
-   case GL_NEVER: 
-       return COMPAREFUNC_ALWAYS; 
-   case GL_LESS: 
-       return COMPAREFUNC_LEQUAL; 
-   case GL_LEQUAL: 
-       return COMPAREFUNC_LESS;
-   case GL_GREATER: 
-       return COMPAREFUNC_GEQUAL; 
-   case GL_GEQUAL: 
-      return COMPAREFUNC_GREATER; 
-   case GL_NOTEQUAL: 
-      return COMPAREFUNC_EQUAL; 
-   case GL_EQUAL: 
-      return COMPAREFUNC_NOTEQUAL; 
-   case GL_ALWAYS: 
-       return COMPAREFUNC_NEVER; 
-   }
-
-   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func);
-   return COMPAREFUNC_NEVER; 
-}
-
-int
-intel_translate_compare_func(GLenum func)
-{
-   switch (func) {
-   case GL_NEVER:
-      return COMPAREFUNC_NEVER;
-   case GL_LESS:
-      return COMPAREFUNC_LESS;
-   case GL_LEQUAL:
-      return COMPAREFUNC_LEQUAL;
-   case GL_GREATER:
-      return COMPAREFUNC_GREATER;
-   case GL_GEQUAL:
-      return COMPAREFUNC_GEQUAL;
-   case GL_NOTEQUAL:
-      return COMPAREFUNC_NOTEQUAL;
-   case GL_EQUAL:
-      return COMPAREFUNC_EQUAL;
-   case GL_ALWAYS:
-      return COMPAREFUNC_ALWAYS;
-   }
-
-   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func);
-   return COMPAREFUNC_ALWAYS;
-}
-
-int
-intel_translate_stencil_op(GLenum op)
-{
-   switch (op) {
-   case GL_KEEP:
-      return STENCILOP_KEEP;
-   case GL_ZERO:
-      return STENCILOP_ZERO;
-   case GL_REPLACE:
-      return STENCILOP_REPLACE;
-   case GL_INCR:
-      return STENCILOP_INCRSAT;
-   case GL_DECR:
-      return STENCILOP_DECRSAT;
-   case GL_INCR_WRAP:
-      return STENCILOP_INCR;
-   case GL_DECR_WRAP:
-      return STENCILOP_DECR;
-   case GL_INVERT:
-      return STENCILOP_INVERT;
-   default:
-      return STENCILOP_ZERO;
-   }
-}
-
-int
-intel_translate_blend_factor(GLenum factor)
-{
-   switch (factor) {
-   case GL_ZERO:
-      return BLENDFACT_ZERO;
-   case GL_SRC_ALPHA:
-      return BLENDFACT_SRC_ALPHA;
-   case GL_ONE:
-      return BLENDFACT_ONE;
-   case GL_SRC_COLOR:
-      return BLENDFACT_SRC_COLR;
-   case GL_ONE_MINUS_SRC_COLOR:
-      return BLENDFACT_INV_SRC_COLR;
-   case GL_DST_COLOR:
-      return BLENDFACT_DST_COLR;
-   case GL_ONE_MINUS_DST_COLOR:
-      return BLENDFACT_INV_DST_COLR;
-   case GL_ONE_MINUS_SRC_ALPHA:
-      return BLENDFACT_INV_SRC_ALPHA;
-   case GL_DST_ALPHA:
-      return BLENDFACT_DST_ALPHA;
-   case GL_ONE_MINUS_DST_ALPHA:
-      return BLENDFACT_INV_DST_ALPHA;
-   case GL_SRC_ALPHA_SATURATE:
-      return BLENDFACT_SRC_ALPHA_SATURATE;
-   case GL_CONSTANT_COLOR:
-      return BLENDFACT_CONST_COLOR;
-   case GL_ONE_MINUS_CONSTANT_COLOR:
-      return BLENDFACT_INV_CONST_COLOR;
-   case GL_CONSTANT_ALPHA:
-      return BLENDFACT_CONST_ALPHA;
-   case GL_ONE_MINUS_CONSTANT_ALPHA:
-      return BLENDFACT_INV_CONST_ALPHA;
-   }
-
-   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, factor);
-   return BLENDFACT_ZERO;
-}
-
-int
-intel_translate_logic_op(GLenum opcode)
-{
-   switch (opcode) {
-   case GL_CLEAR:
-      return LOGICOP_CLEAR;
-   case GL_AND:
-      return LOGICOP_AND;
-   case GL_AND_REVERSE:
-      return LOGICOP_AND_RVRSE;
-   case GL_COPY:
-      return LOGICOP_COPY;
-   case GL_COPY_INVERTED:
-      return LOGICOP_COPY_INV;
-   case GL_AND_INVERTED:
-      return LOGICOP_AND_INV;
-   case GL_NOOP:
-      return LOGICOP_NOOP;
-   case GL_XOR:
-      return LOGICOP_XOR;
-   case GL_OR:
-      return LOGICOP_OR;
-   case GL_OR_INVERTED:
-      return LOGICOP_OR_INV;
-   case GL_NOR:
-      return LOGICOP_NOR;
-   case GL_EQUIV:
-      return LOGICOP_EQUIV;
-   case GL_INVERT:
-      return LOGICOP_INV;
-   case GL_OR_REVERSE:
-      return LOGICOP_OR_RVRSE;
-   case GL_NAND:
-      return LOGICOP_NAND;
-   case GL_SET:
-      return LOGICOP_SET;
-   default:
-      return LOGICOP_SET;
-   }
-}
-
-
-static void
-intelClearColor(GLcontext * ctx, const GLfloat color[4])
-{
-   struct intel_context *intel = intel_context(ctx);
-   GLubyte clear[4];
-
-   CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]);
-   CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]);
-   CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]);
-   CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]);
-
-   /* compute both 32 and 16-bit clear values */
-   intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1],
-                                               clear[2], clear[3]);
-   intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]);
-}
-
-
-/**
- * Update the viewport transformation matrix.  Depends on:
- *  - viewport pos/size
- *  - depthrange
- *  - window pos/size or FBO size
- */
-static void
-intelCalcViewport(GLcontext * ctx)
-{
-   struct intel_context *intel = intel_context(ctx);
-   const GLfloat *v = ctx->Viewport._WindowMap.m;
-   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
-   GLfloat *m = intel->ViewportMatrix.m;
-   GLfloat yScale, yBias;
-
-   if (ctx->DrawBuffer->Name) {
-      /* User created FBO */
-      struct intel_renderbuffer *irb
-         = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
-      if (irb && !irb->RenderToTexture) {
-         /* y=0=top */
-         yScale = -1.0;
-         yBias = irb->Base.Height;
-      }
-      else {
-         /* y=0=bottom */
-         yScale = 1.0;
-         yBias = 0.0;
-      }
-   }
-   else {
-      /* window buffer, y=0=top */
-      yScale = -1.0;
-      yBias = (intel->driDrawable) ? intel->driDrawable->h : 0.0F;
-   }
-
-   m[MAT_SX] = v[MAT_SX];
-   m[MAT_TX] = v[MAT_TX];
-
-   m[MAT_SY] = v[MAT_SY] * yScale;
-   m[MAT_TY] = v[MAT_TY] * yScale + yBias;
-
-   m[MAT_SZ] = v[MAT_SZ] * depthScale;
-   m[MAT_TZ] = v[MAT_TZ] * depthScale;
-}
-
-static void
-intelViewport(GLcontext * ctx,
-              GLint x, GLint y, GLsizei width, GLsizei height)
-{
-   intelCalcViewport(ctx);
-}
-
-static void
-intelDepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval)
-{
-   intelCalcViewport(ctx);
-}
-
-/* Fallback to swrast for select and feedback.
- */
-static void
-intelRenderMode(GLcontext * ctx, GLenum mode)
-{
-   struct intel_context *intel = intel_context(ctx);
-   FALLBACK(intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER));
-}
-
-
-void
-intelInitStateFuncs(struct dd_function_table *functions)
-{
-   functions->RenderMode = intelRenderMode;
-   functions->Viewport = intelViewport;
-   functions->DepthRange = intelDepthRange;
-   functions->ClearColor = intelClearColor;
-}
+../intel/intel_state.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i915/intel_swapbuffers.c b/src/mesa/drivers/dri/i915/intel_swapbuffers.c
new file mode 120000
index 0000000000..148d5215aa
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_swapbuffers.c
@@ -0,0 +1 @@
+../intel/intel_swapbuffers.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c
index 06210035af..e80996580c 100644
--- a/src/mesa/drivers/dri/i915/intel_tris.c
+++ b/src/mesa/drivers/dri/i915/intel_tris.c
@@ -61,9 +61,100 @@ static void intelRenderPrimitive(GLcontext * ctx, GLenum prim);
 static void intelRasterPrimitive(GLcontext * ctx, GLenum rprim,
                                  GLuint hwprim);
 
+static void
+intel_flush_inline_primitive(struct intel_context *intel)
+{
+   GLuint used = intel->batch->ptr - intel->prim.start_ptr;
+
+   assert(intel->prim.primitive != ~0);
+
+/*    _mesa_printf("/\n"); */
+
+   if (used < 8)
+      goto do_discard;
+
+   *(int *) intel->prim.start_ptr = (_3DPRIMITIVE |
+                                     intel->prim.primitive | (used / 4 - 2));
+
+   goto finished;
+
+ do_discard:
+   intel->batch->ptr -= used;
+
+ finished:
+   intel->prim.primitive = ~0;
+   intel->prim.start_ptr = 0;
+   intel->prim.flush = 0;
+}
+
+static void intel_start_inline(struct intel_context *intel, uint32_t prim)
+{
+   BATCH_LOCALS;
+   uint32_t batch_flags = LOOP_CLIPRECTS;
+
+   intel->vtbl.emit_state(intel);
+
+   intel->no_batch_wrap = GL_TRUE;
+
+   /*_mesa_printf("%s *", __progname);*/
+
+   /* Emit a slot which will be filled with the inline primitive
+    * command later.
+    */
+   BEGIN_BATCH(2, batch_flags);
+   OUT_BATCH(0);
+
+   assert((intel->batch->dirty_state & (1<<1)) == 0);
+
+   intel->prim.start_ptr = intel->batch->ptr;
+   intel->prim.primitive = prim;
+   intel->prim.flush = intel_flush_inline_primitive;
+
+   OUT_BATCH(0);
+   ADVANCE_BATCH();
+
+   intel->no_batch_wrap = GL_FALSE;
+/*    _mesa_printf(">"); */
+}
+
+static void intel_wrap_inline(struct intel_context *intel)
+{
+   GLuint prim = intel->prim.primitive;
+
+   intel_flush_inline_primitive(intel);
+   intel_batchbuffer_flush(intel->batch);
+   intel_start_inline(intel, prim);  /* ??? */
+}
+
+static GLuint *intel_extend_inline(struct intel_context *intel, GLuint dwords)
+{
+   GLuint sz = dwords * sizeof(GLuint);
+   GLuint *ptr;
+
+   assert(intel->prim.flush == intel_flush_inline_primitive);
+
+   if (intel_batchbuffer_space(intel->batch) < sz)
+      intel_wrap_inline(intel);
+
+/*    _mesa_printf("."); */
+
+   intel->vtbl.assert_not_dirty(intel);
+
+   ptr = (GLuint *) intel->batch->ptr;
+   intel->batch->ptr += sz;
+
+   return ptr;
+}
+
 /** Sets the primitive type for a primitive sequence, flushing as needed. */
 void intel_set_prim(struct intel_context *intel, uint32_t prim)
 {
+   /* if we have no VBOs */
+
+   if (intel->intelScreen->no_vbo) {
+      intel_start_inline(intel, prim);
+      return;
+   }
    if (prim != intel->prim.primitive) {
       INTEL_FIREVERTICES(intel);
       intel->prim.primitive = prim;
@@ -75,6 +166,10 @@ uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count)
 {
    uint32_t *addr;
 
+   if (intel->intelScreen->no_vbo) {
+      return intel_extend_inline(intel, count * intel->vertex_size);
+   }
+
    /* Check for space in the existing VB */
    if (intel->prim.vb_bo == NULL ||
        (intel->prim.current_offset +
@@ -118,8 +213,7 @@ void intel_flush_prim(struct intel_context *intel)
       return;
 
    /* Clear the current prims out of the context state so that a batch flush
-    * flush triggered by wait_flips or emit_state doesn't loop back to
-    * flush_prim again.
+    * flush triggered by emit_state doesn't loop back to flush_prim again.
     */
    vb_bo = intel->prim.vb_bo;
    dri_bo_reference(vb_bo);
@@ -131,8 +225,6 @@ void intel_flush_prim(struct intel_context *intel)
       intel->prim.start_offset = ALIGN(intel->prim.start_offset, 128);
    intel->prim.flush = NULL;
 
-   intel_wait_flips(intel);
-
    intel->vtbl.emit_state(intel);
 
    aper_array[0] = intel->batch->buf;
@@ -155,7 +247,7 @@ void intel_flush_prim(struct intel_context *intel)
 
 #if 0
    printf("emitting %d..%d=%d vertices size %d\n", offset,
-	  intel->prim.current_offset, intel->prim.count,
+	  intel->prim.current_offset, count,
 	  intel->vertex_size * 4);
 #endif
 
@@ -186,7 +278,7 @@ void intel_flush_prim(struct intel_context *intel)
       OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0,
 		offset | (intel->vertex_size << S0_VB_PITCH_SHIFT_830) |
 		S0_VB_ENABLE_830);
-      /* S1
+      /* S2
        * This is somewhat unfortunate -- VB width is tied up with
        * vertex format data that we've already uploaded through
        * _3DSTATE_VFT[01]_CMD.  We may want to replace emits of VFT state with
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile
index 005460f354..2934414d99 100644
--- a/src/mesa/drivers/dri/i965/Makefile
+++ b/src/mesa/drivers/dri/i965/Makefile
@@ -9,9 +9,10 @@ DRIVER_SOURCES = \
 	intel_blit.c \
 	intel_buffer_objects.c \
 	intel_buffers.c \
+	intel_clear.c \
 	intel_context.c \
 	intel_decode.c \
-	intel_depthstencil.c \
+	intel_extensions.c \
 	intel_fbo.c \
 	intel_mipmap_tree.c \
 	intel_regions.c \
@@ -22,6 +23,7 @@ DRIVER_SOURCES = \
 	intel_pixel_copy.c \
 	intel_pixel_draw.c \
 	intel_state.c \
+	intel_swapbuffers.c \
 	intel_tex.c \
 	intel_tex_copy.c \
 	intel_tex_format.c \
@@ -49,7 +51,6 @@ DRIVER_SOURCES = \
 	brw_gs.c \
 	brw_gs_emit.c \
 	brw_gs_state.c \
-	brw_metaops.c \
 	brw_misc_state.c \
 	brw_program.c \
 	brw_queryobj.c \
@@ -68,7 +69,6 @@ DRIVER_SOURCES = \
 	brw_vs_constval.c \
 	brw_vs_emit.c \
 	brw_vs_state.c \
-	brw_vs_tnl.c \
 	brw_vtbl.c \
 	brw_wm.c \
 	brw_wm_debug.c \
diff --git a/src/mesa/drivers/dri/i965/brw_cc.c b/src/mesa/drivers/dri/i965/brw_cc.c
index fa8121e02d..82370162f5 100644
--- a/src/mesa/drivers/dri/i965/brw_cc.c
+++ b/src/mesa/drivers/dri/i965/brw_cc.c
@@ -83,59 +83,60 @@ struct brw_cc_unit_key {
 static void
 cc_unit_populate_key(struct brw_context *brw, struct brw_cc_unit_key *key)
 {
-   struct gl_stencil_attrib *stencil = brw->attribs.Stencil;
+   GLcontext *ctx = &brw->intel.ctx;
+   const unsigned back = ctx->Stencil._BackFace;
 
    memset(key, 0, sizeof(*key));
 
-   key->stencil = stencil->Enabled;
-   key->stencil_two_side = stencil->_TestTwoSide;
+   key->stencil = ctx->Stencil.Enabled;
+   key->stencil_two_side = ctx->Stencil._TestTwoSide;
 
    if (key->stencil) {
-      key->stencil_func[0] = stencil->Function[0];
-      key->stencil_fail_op[0] = stencil->FailFunc[0];
-      key->stencil_pass_depth_fail_op[0] = stencil->ZFailFunc[0];
-      key->stencil_pass_depth_pass_op[0] = stencil->ZPassFunc[0];
-      key->stencil_ref[0] = stencil->Ref[0];
-      key->stencil_write_mask[0] = stencil->WriteMask[0];
-      key->stencil_test_mask[0] = stencil->ValueMask[0];
+      key->stencil_func[0] = ctx->Stencil.Function[0];
+      key->stencil_fail_op[0] = ctx->Stencil.FailFunc[0];
+      key->stencil_pass_depth_fail_op[0] = ctx->Stencil.ZFailFunc[0];
+      key->stencil_pass_depth_pass_op[0] = ctx->Stencil.ZPassFunc[0];
+      key->stencil_ref[0] = ctx->Stencil.Ref[0];
+      key->stencil_write_mask[0] = ctx->Stencil.WriteMask[0];
+      key->stencil_test_mask[0] = ctx->Stencil.ValueMask[0];
    }
    if (key->stencil_two_side) {
-      key->stencil_func[1] = stencil->Function[1];
-      key->stencil_fail_op[1] = stencil->FailFunc[1];
-      key->stencil_pass_depth_fail_op[1] = stencil->ZFailFunc[1];
-      key->stencil_pass_depth_pass_op[1] = stencil->ZPassFunc[1];
-      key->stencil_ref[1] = stencil->Ref[1];
-      key->stencil_write_mask[1] = stencil->WriteMask[1];
-      key->stencil_test_mask[1] = stencil->ValueMask[1];
+      key->stencil_func[1] = ctx->Stencil.Function[back];
+      key->stencil_fail_op[1] = ctx->Stencil.FailFunc[back];
+      key->stencil_pass_depth_fail_op[1] = ctx->Stencil.ZFailFunc[back];
+      key->stencil_pass_depth_pass_op[1] = ctx->Stencil.ZPassFunc[back];
+      key->stencil_ref[1] = ctx->Stencil.Ref[back];
+      key->stencil_write_mask[1] = ctx->Stencil.WriteMask[back];
+      key->stencil_test_mask[1] = ctx->Stencil.ValueMask[back];
    }
 
-   if (brw->attribs.Color->_LogicOpEnabled)
-      key->logic_op = brw->attribs.Color->LogicOp;
+   if (ctx->Color._LogicOpEnabled)
+      key->logic_op = ctx->Color.LogicOp;
    else
       key->logic_op = GL_COPY;
 
-   key->color_blend = brw->attribs.Color->BlendEnabled;
+   key->color_blend = ctx->Color.BlendEnabled;
    if (key->color_blend) {
-      key->blend_eq_rgb = brw->attribs.Color->BlendEquationRGB;
-      key->blend_eq_a = brw->attribs.Color->BlendEquationA;
-      key->blend_src_rgb = brw->attribs.Color->BlendSrcRGB;
-      key->blend_dst_rgb = brw->attribs.Color->BlendDstRGB;
-      key->blend_src_a = brw->attribs.Color->BlendSrcA;
-      key->blend_dst_a = brw->attribs.Color->BlendDstA;
+      key->blend_eq_rgb = ctx->Color.BlendEquationRGB;
+      key->blend_eq_a = ctx->Color.BlendEquationA;
+      key->blend_src_rgb = ctx->Color.BlendSrcRGB;
+      key->blend_dst_rgb = ctx->Color.BlendDstRGB;
+      key->blend_src_a = ctx->Color.BlendSrcA;
+      key->blend_dst_a = ctx->Color.BlendDstA;
    }
 
-   key->alpha_enabled = brw->attribs.Color->AlphaEnabled;
+   key->alpha_enabled = ctx->Color.AlphaEnabled;
    if (key->alpha_enabled) {
-      key->alpha_func = brw->attribs.Color->AlphaFunc;
-      key->alpha_ref = brw->attribs.Color->AlphaRef;
+      key->alpha_func = ctx->Color.AlphaFunc;
+      key->alpha_ref = ctx->Color.AlphaRef;
    }
 
-   key->dither = brw->attribs.Color->DitherFlag;
+   key->dither = ctx->Color.DitherFlag;
 
-   key->depth_test = brw->attribs.Depth->Test;
+   key->depth_test = ctx->Depth.Test;
    if (key->depth_test) {
-      key->depth_func = brw->attribs.Depth->Func;
-      key->depth_write = brw->attribs.Depth->Mask;
+      key->depth_func = ctx->Depth.Func;
+      key->depth_write = ctx->Depth.Mask;
    }
 }
 
diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c
index 38d8b704d7..5cffcebde4 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.c
+++ b/src/mesa/drivers/dri/i965/brw_clip.c
@@ -145,14 +145,14 @@ static void upload_clip_prog(struct brw_context *brw)
    /* CACHE_NEW_VS_PROG */
    key.attrs = brw->vs.prog_data->outputs_written;
    /* _NEW_LIGHT */
-   key.do_flat_shading = (brw->attribs.Light->ShadeModel == GL_FLAT);
+   key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT);
    /* _NEW_TRANSFORM */
-   key.nr_userclip = brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled);
+   key.nr_userclip = brw_count_bits(ctx->Transform.ClipPlanesEnabled);
    key.clip_mode = BRW_CLIPMODE_NORMAL;
 
    /* _NEW_POLYGON */
    if (key.primitive == GL_TRIANGLES) {
-      if (brw->attribs.Polygon->CullFaceMode == GL_FRONT_AND_BACK) 
+      if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) 
 	 key.clip_mode = BRW_CLIPMODE_REJECT_ALL;
       else {
 	 GLuint fill_front = CLIP_CULL;
@@ -160,44 +160,44 @@ static void upload_clip_prog(struct brw_context *brw)
 	 GLuint offset_front = 0;
 	 GLuint offset_back = 0;
 
-	 if (!brw->attribs.Polygon->CullFlag ||
-	     brw->attribs.Polygon->CullFaceMode != GL_FRONT) {
-	    switch (brw->attribs.Polygon->FrontMode) {
+	 if (!ctx->Polygon.CullFlag ||
+	     ctx->Polygon.CullFaceMode != GL_FRONT) {
+	    switch (ctx->Polygon.FrontMode) {
 	    case GL_FILL: 
 	       fill_front = CLIP_FILL; 
 	       offset_front = 0;
 	       break;
 	    case GL_LINE:
 	       fill_front = CLIP_LINE;
-	       offset_front = brw->attribs.Polygon->OffsetLine;
+	       offset_front = ctx->Polygon.OffsetLine;
 	       break;
 	    case GL_POINT:
 	       fill_front = CLIP_POINT;
-	       offset_front = brw->attribs.Polygon->OffsetPoint;
+	       offset_front = ctx->Polygon.OffsetPoint;
 	       break;
 	    }
 	 }
 
-	 if (!brw->attribs.Polygon->CullFlag ||
-	     brw->attribs.Polygon->CullFaceMode != GL_BACK) {
-	    switch (brw->attribs.Polygon->BackMode) {
+	 if (!ctx->Polygon.CullFlag ||
+	     ctx->Polygon.CullFaceMode != GL_BACK) {
+	    switch (ctx->Polygon.BackMode) {
 	    case GL_FILL: 
 	       fill_back = CLIP_FILL; 
 	       offset_back = 0;
 	       break;
 	    case GL_LINE:
 	       fill_back = CLIP_LINE;
-	       offset_back = brw->attribs.Polygon->OffsetLine;
+	       offset_back = ctx->Polygon.OffsetLine;
 	       break;
 	    case GL_POINT:
 	       fill_back = CLIP_POINT;
-	       offset_back = brw->attribs.Polygon->OffsetPoint;
+	       offset_back = ctx->Polygon.OffsetPoint;
 	       break;
 	    }
 	 }
 
-	 if (brw->attribs.Polygon->BackMode != GL_FILL ||
-	     brw->attribs.Polygon->FrontMode != GL_FILL) {
+	 if (ctx->Polygon.BackMode != GL_FILL ||
+	     ctx->Polygon.FrontMode != GL_FILL) {
 	    key.do_unfilled = 1;
 
 	    /* Most cases the fixed function units will handle.  Cases where
@@ -207,17 +207,17 @@ static void upload_clip_prog(struct brw_context *brw)
 
 	    if (offset_back || offset_front) {
 	       /* _NEW_POLYGON, _NEW_BUFFERS */
-	       key.offset_units = brw->attribs.Polygon->OffsetUnits * brw->intel.polygon_offset_scale;
-	       key.offset_factor = brw->attribs.Polygon->OffsetFactor * ctx->DrawBuffer->_MRD;
+	       key.offset_units = ctx->Polygon.OffsetUnits * brw->intel.polygon_offset_scale;
+	       key.offset_factor = ctx->Polygon.OffsetFactor * ctx->DrawBuffer->_MRD;
 	    }
 
-	    switch (brw->attribs.Polygon->FrontFace) {
+	    switch (ctx->Polygon.FrontFace) {
 	    case GL_CCW:
 	       key.fill_ccw = fill_front;
 	       key.fill_cw = fill_back;
 	       key.offset_ccw = offset_front;
 	       key.offset_cw = offset_back;
-	       if (brw->attribs.Light->Model.TwoSide &&
+	       if (ctx->Light.Model.TwoSide &&
 		   key.fill_cw != CLIP_CULL) 
 		  key.copy_bfc_cw = 1;
 	       break;
@@ -226,7 +226,7 @@ static void upload_clip_prog(struct brw_context *brw)
 	       key.fill_ccw = fill_back;
 	       key.offset_cw = offset_front;
 	       key.offset_ccw = offset_back;
-	       if (brw->attribs.Light->Model.TwoSide &&
+	       if (ctx->Light.Model.TwoSide &&
 		   key.fill_ccw != CLIP_CULL) 
 		  key.copy_bfc_ccw = 1;
 	       break;
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index e2bc08a6cb..eaac6224f6 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -32,6 +32,7 @@
 
 #include "main/imports.h"
 #include "main/api_noop.h"
+#include "main/macros.h"
 #include "main/vtxfmt.h"
 #include "main/simple_list.h"
 #include "shader/shader_api.h"
@@ -71,32 +72,10 @@ static void brwInitDriverFunctions( struct dd_function_table *functions )
    brwInitFragProgFuncs( functions );
    brwInitProgFuncs( functions );
    brw_init_queryobj_functions(functions);
-}
-
 
-static void brw_init_attribs( struct brw_context *brw )
-{
-   GLcontext *ctx = &brw->intel.ctx;
-
-   brw->attribs.Color = &ctx->Color;
-   brw->attribs.Depth = &ctx->Depth;
-   brw->attribs.Fog = &ctx->Fog;
-   brw->attribs.Hint = &ctx->Hint;
-   brw->attribs.Light = &ctx->Light;
-   brw->attribs.Line = &ctx->Line;
-   brw->attribs.Point = &ctx->Point;
-   brw->attribs.Polygon = &ctx->Polygon;
-   brw->attribs.Scissor = &ctx->Scissor;
-   brw->attribs.Stencil = &ctx->Stencil;
-   brw->attribs.Texture = &ctx->Texture;
-   brw->attribs.Transform = &ctx->Transform;
-   brw->attribs.Viewport = &ctx->Viewport;
-   brw->attribs.VertexProgram = &ctx->VertexProgram;
-   brw->attribs.FragmentProgram = &ctx->FragmentProgram;
-   brw->attribs.PolygonStipple = &ctx->PolygonStipple[0];
+   functions->Viewport = intel_viewport;
 }
 
-
 GLboolean brwCreateContext( const __GLcontextModes *mesaVis,
 			    __DRIcontextPrivate *driContextPriv,
 			    void *sharedContextPrivate)
@@ -126,23 +105,24 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis,
 
    TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline;
 
-   ctx->Const.MaxTextureUnits = BRW_MAX_TEX_UNIT;
    ctx->Const.MaxTextureImageUnits = BRW_MAX_TEX_UNIT;
-   ctx->Const.MaxTextureCoordUnits = BRW_MAX_TEX_UNIT;
+   ctx->Const.MaxTextureCoordUnits = 8; /* Mesa limit */
+   ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureCoordUnits,
+                                     ctx->Const.MaxTextureImageUnits);
    ctx->Const.MaxVertexTextureImageUnits = 0; /* no vertex shader textures */
 
-   /* Advertise the full hardware capabilities.  The new memory
-    * manager should cope much better with overload situations:
+   /* Mesa limits textures to 4kx4k; it would be nice to fix that someday
     */
-   ctx->Const.MaxTextureLevels = 12;
+   ctx->Const.MaxTextureLevels = 13;
    ctx->Const.Max3DTextureLevels = 9;
    ctx->Const.MaxCubeTextureLevels = 12;
-   ctx->Const.MaxTextureRectSize = (1<<11);
+   ctx->Const.MaxTextureRectSize = (1<<12);
    
+   /* if conformance mode is set, swrast can handle any size AA point */
+   ctx->Const.MaxPointSizeAA = 255.0;
+
 /*    ctx->Const.MaxNativeVertexProgramTemps = 32; */
 
-   brw_init_attribs( brw );
-   brw_init_metaops( brw );
    brw_init_state( brw );
 
    brw->state.dirty.mesa = ~0;
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index e3904be977..df90c2027f 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -131,18 +131,17 @@ struct brw_context;
 #define BRW_NEW_WM_INPUT_DIMENSIONS     0x100
 #define BRW_NEW_INPUT_VARYING           0x200
 #define BRW_NEW_PSP                     0x800
-#define BRW_NEW_METAOPS                 0x1000
 #define BRW_NEW_FENCE                   0x2000
-#define BRW_NEW_LOCK                    0x4000
-#define BRW_NEW_INDICES			0x8000
-#define BRW_NEW_VERTICES		0x10000
+#define BRW_NEW_INDICES			0x4000
+#define BRW_NEW_VERTICES		0x8000
 /**
  * Used for any batch entry with a relocated pointer that will be used
  * by any 3D rendering.
  */
-#define BRW_NEW_BATCH			0x8000
+#define BRW_NEW_BATCH			0x10000
 /** brw->depth_region updated */
-#define BRW_NEW_DEPTH_BUFFER		0x10000
+#define BRW_NEW_DEPTH_BUFFER		0x20000
+#define BRW_NEW_NR_SURFACES		0x40000
 
 struct brw_state_flags {
    /** State update flags signalled by mesa internals */
@@ -158,7 +157,6 @@ struct brw_state_flags {
 struct brw_vertex_program {
    struct gl_vertex_program program;
    GLuint id;
-   GLuint param_state;		/* flags indicating state tracked by params */
 };
 
 
@@ -166,7 +164,6 @@ struct brw_vertex_program {
 struct brw_fragment_program {
    struct gl_fragment_program program;
    GLuint id;
-   GLuint param_state;		/* flags indicating state tracked by params */
 };
 
 
@@ -240,7 +237,7 @@ struct brw_vs_ouput_sizes {
 };
 
 
-#define BRW_MAX_TEX_UNIT 8
+#define BRW_MAX_TEX_UNIT 16
 #define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + MAX_DRAW_BUFFERS
 
 enum brw_cache_id {
@@ -305,26 +302,6 @@ struct brw_cache {
 };
 
 
-
-struct brw_state_pointers {
-   struct gl_colorbuffer_attrib	*Color;
-   struct gl_depthbuffer_attrib	*Depth;
-   struct gl_fog_attrib		*Fog;
-   struct gl_hint_attrib	*Hint;
-   struct gl_light_attrib	*Light;
-   struct gl_line_attrib	*Line;
-   struct gl_point_attrib	*Point;
-   struct gl_polygon_attrib	*Polygon;
-   GLuint                       *PolygonStipple;
-   struct gl_scissor_attrib	*Scissor;
-   struct gl_stencil_attrib	*Stencil;
-   struct gl_texture_attrib	*Texture;
-   struct gl_transform_attrib	*Transform;
-   struct gl_viewport_attrib	*Viewport;
-   struct gl_vertex_program_state *VertexProgram; 
-   struct gl_fragment_program_state *FragmentProgram;
-};
-
 /* Considered adding a member to this struct to document which flags
  * an update might raise so that ordering of the state atoms can be
  * checked or derived at runtime.  Dropped the idea in favor of having
@@ -459,7 +436,6 @@ struct brw_context
       int validated_bo_count;
    } state;
 
-   struct brw_state_pointers attribs;
    struct brw_cache cache;
    struct brw_cached_batch_item *cached_batch_items;
 
@@ -493,28 +469,6 @@ struct brw_context
       unsigned int offset;
    } ib;
 
-   struct {
-      /* Will be allocated on demand if needed.   
-       */
-      struct brw_state_pointers attribs;
-      struct gl_vertex_program *vp;
-      struct gl_fragment_program *fp, *fp_tex;
-
-      struct gl_buffer_object *vbo;
-
-      struct intel_region *saved_draw_region;
-      GLuint saved_nr_draw_regions;
-      struct intel_region *saved_depth_region;
-
-      GLuint restore_draw_buffers[MAX_DRAW_BUFFERS];
-      GLuint restore_num_draw_buffers;
-
-      struct gl_fragment_program *restore_fp;
-      
-      GLboolean active;
-   } metaops;
-
-
    /* Active vertex program: 
     */
    const struct gl_vertex_program *vertex_program;
@@ -705,13 +659,6 @@ void brw_FrameBufferTexInit( struct brw_context *brw,
 void brw_FrameBufferTexDestroy( struct brw_context *brw );
 void brw_validate_textures( struct brw_context *brw );
 
-/*======================================================================
- * brw_metaops.c
- */
-
-void brw_init_metaops( struct brw_context *brw );
-void brw_destroy_metaops( struct brw_context *brw );
-
 
 /*======================================================================
  * brw_program.c
diff --git a/src/mesa/drivers/dri/i965/brw_curbe.c b/src/mesa/drivers/dri/i965/brw_curbe.c
index c7bac7b0c5..4eaaa5f871 100644
--- a/src/mesa/drivers/dri/i965/brw_curbe.c
+++ b/src/mesa/drivers/dri/i965/brw_curbe.c
@@ -48,6 +48,7 @@
  */
 static void calculate_curbe_offsets( struct brw_context *brw )
 {
+   GLcontext *ctx = &brw->intel.ctx;
    /* CACHE_NEW_WM_PROG */
    GLuint nr_fp_regs = (brw->wm.prog_data->nr_params + 15) / 16;
    
@@ -58,8 +59,8 @@ static void calculate_curbe_offsets( struct brw_context *brw )
    GLuint total_regs;
 
    /* _NEW_TRANSFORM */
-   if (brw->attribs.Transform->ClipPlanesEnabled) {
-      GLuint nr_planes = 6 + brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled);
+   if (ctx->Transform.ClipPlanesEnabled) {
+      GLuint nr_planes = 6 + brw_count_bits(ctx->Transform.ClipPlanesEnabled);
       nr_clip_regs = (nr_planes * 4 + 15) / 16;
    }
 
@@ -184,8 +185,8 @@ static void prepare_constant_buffer(struct brw_context *brw)
     * function will also be called whenever fp or vp changes.
     */
    brw->curbe.tracked_state.dirty.mesa = (_NEW_TRANSFORM|_NEW_PROJECTION);
-   brw->curbe.tracked_state.dirty.mesa |= vp->param_state;
-   brw->curbe.tracked_state.dirty.mesa |= fp->param_state;
+   brw->curbe.tracked_state.dirty.mesa |= vp->program.Base.Parameters->StateFlags;
+   brw->curbe.tracked_state.dirty.mesa |= fp->program.Base.Parameters->StateFlags;
 
    if (sz == 0) {
 
@@ -233,11 +234,11 @@ static void prepare_constant_buffer(struct brw_context *brw)
        */
       assert(MAX_CLIP_PLANES == 6);
       for (j = 0; j < MAX_CLIP_PLANES; j++) {
-	 if (brw->attribs.Transform->ClipPlanesEnabled & (1<<j)) {
-	    buf[offset + i * 4 + 0] = brw->attribs.Transform->_ClipUserPlane[j][0];
-	    buf[offset + i * 4 + 1] = brw->attribs.Transform->_ClipUserPlane[j][1];
-	    buf[offset + i * 4 + 2] = brw->attribs.Transform->_ClipUserPlane[j][2];
-	    buf[offset + i * 4 + 3] = brw->attribs.Transform->_ClipUserPlane[j][3];
+	 if (ctx->Transform.ClipPlanesEnabled & (1<<j)) {
+	    buf[offset + i * 4 + 0] = ctx->Transform._ClipUserPlane[j][0];
+	    buf[offset + i * 4 + 1] = ctx->Transform._ClipUserPlane[j][1];
+	    buf[offset + i * 4 + 2] = ctx->Transform._ClipUserPlane[j][2];
+	    buf[offset + i * 4 + 3] = ctx->Transform._ClipUserPlane[j][3];
 	    i++;
 	 }
       }
diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
index f893dd6742..99fd587e9f 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.c
+++ b/src/mesa/drivers/dri/i965/brw_draw.c
@@ -49,7 +49,7 @@
 
 #define FILE_DEBUG_FLAG DEBUG_BATCH
 
-static GLuint hw_prim[GL_POLYGON+1] = {
+static GLuint prim_to_hw_prim[GL_POLYGON+1] = {
    _3DPRIM_POINTLIST,
    _3DPRIM_LINELIST,
    _3DPRIM_LINELOOP,
@@ -84,15 +84,17 @@ static const GLenum reduced_prim[GL_POLYGON+1] = {
  */
 static GLuint brw_set_prim(struct brw_context *brw, GLenum prim)
 {
+   GLcontext *ctx = &brw->intel.ctx;
+
    if (INTEL_DEBUG & DEBUG_PRIMS)
       _mesa_printf("PRIM: %s\n", _mesa_lookup_enum_by_nr(prim));
    
    /* Slight optimization to avoid the GS program when not needed:
     */
    if (prim == GL_QUAD_STRIP &&
-       brw->attribs.Light->ShadeModel != GL_FLAT &&
-       brw->attribs.Polygon->FrontMode == GL_FILL &&
-       brw->attribs.Polygon->BackMode == GL_FILL)
+       ctx->Light.ShadeModel != GL_FLAT &&
+       ctx->Polygon.FrontMode == GL_FILL &&
+       ctx->Polygon.BackMode == GL_FILL)
       prim = GL_TRIANGLE_STRIP;
 
    if (prim != brw->primitive) {
@@ -103,12 +105,9 @@ static GLuint brw_set_prim(struct brw_context *brw, GLenum prim)
 	 brw->intel.reduced_primitive = reduced_prim[prim];
 	 brw->state.dirty.brw |= BRW_NEW_REDUCED_PRIMITIVE;
       }
-
-      brw_validate_state(brw);
-      brw_upload_state(brw);
    }
 
-   return hw_prim[prim];
+   return prim_to_hw_prim[prim];
 }
 
 
@@ -123,9 +122,9 @@ static GLuint trim(GLenum prim, GLuint length)
 }
 
 
-static void brw_emit_prim( struct brw_context *brw, 
-			   const struct _mesa_prim *prim )
-
+static void brw_emit_prim(struct brw_context *brw,
+			  const struct _mesa_prim *prim,
+			  uint32_t hw_prim)
 {
    struct brw_3d_primitive prim_packet;
 
@@ -136,7 +135,7 @@ static void brw_emit_prim( struct brw_context *brw,
    prim_packet.header.opcode = CMD_3D_PRIM;
    prim_packet.header.length = sizeof(prim_packet)/4 - 2;
    prim_packet.header.pad = 0;
-   prim_packet.header.topology = brw_set_prim(brw, prim->mode);
+   prim_packet.header.topology = hw_prim;
    prim_packet.header.indexed = prim->indexed;
 
    prim_packet.verts_per_instance = trim(prim->mode, prim->count);
@@ -169,14 +168,11 @@ static void brw_merge_inputs( struct brw_context *brw,
    for (i = 0; i < VERT_ATTRIB_MAX; i++) {
       brw->vb.inputs[i].glarray = arrays[i];
 
-      /* XXX: metaops passes null arrays */
-      if (arrays[i]) {
-	 if (arrays[i]->StrideB != 0)
-	    brw->vb.info.varying |= 1 << i;
+      if (arrays[i]->StrideB != 0)
+	 brw->vb.info.varying |= 1 << i;
 
 	 brw->vb.info.sizes[i/16] |= (brw->vb.inputs[i].glarray->Size - 1) <<
 	    ((i%16) * 2);
-      }
    }
 
    /* Raise statechanges if input sizes and varying have changed: 
@@ -195,12 +191,13 @@ static GLboolean check_fallbacks( struct brw_context *brw,
 				  const struct _mesa_prim *prim,
 				  GLuint nr_prims )
 {
+   GLcontext *ctx = &brw->intel.ctx;
    GLuint i;
 
    if (!brw->intel.strict_conformance)
       return GL_FALSE;
 
-   if (brw->attribs.Polygon->SmoothFlag) {
+   if (ctx->Polygon.SmoothFlag) {
       for (i = 0; i < nr_prims; i++)
 	 if (reduced_prim[prim[i].mode] == GL_TRIANGLES) 
 	    return GL_TRUE;
@@ -209,7 +206,7 @@ static GLboolean check_fallbacks( struct brw_context *brw,
    /* BRW hardware will do AA lines, but they are non-conformant it
     * seems.  TBD whether we keep this fallback:
     */
-   if (brw->attribs.Line->SmoothFlag) {
+   if (ctx->Line.SmoothFlag) {
       for (i = 0; i < nr_prims; i++)
 	 if (reduced_prim[prim[i].mode] == GL_LINES) 
 	    return GL_TRUE;
@@ -218,7 +215,7 @@ static GLboolean check_fallbacks( struct brw_context *brw,
    /* Stipple -- these fallbacks could be resolved with a little
     * bit of work?
     */
-   if (brw->attribs.Line->StippleFlag) {
+   if (ctx->Line.StippleFlag) {
       for (i = 0; i < nr_prims; i++) {
 	 /* GS doesn't get enough information to know when to reset
 	  * the stipple counter?!?
@@ -227,14 +224,14 @@ static GLboolean check_fallbacks( struct brw_context *brw,
 	    return GL_TRUE;
 	    
 	 if (prim[i].mode == GL_POLYGON &&
-	     (brw->attribs.Polygon->FrontMode == GL_LINE ||
-	      brw->attribs.Polygon->BackMode == GL_LINE))
+	     (ctx->Polygon.FrontMode == GL_LINE ||
+	      ctx->Polygon.BackMode == GL_LINE))
 	    return GL_TRUE;
       }
    }
 
 
-   if (brw->attribs.Point->SmoothFlag) {
+   if (ctx->Point.SmoothFlag) {
       for (i = 0; i < nr_prims; i++)
 	 if (prim[i].mode == GL_POINTS) 
 	    return GL_TRUE;
@@ -258,11 +255,15 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
    struct brw_context *brw = brw_context(ctx);
    GLboolean retval = GL_FALSE;
    GLboolean warn = GL_FALSE;
+   GLboolean first_time = GL_TRUE;
    GLuint i;
 
    if (ctx->NewState)
       _mesa_update_state( ctx );
 
+   if (check_fallbacks(brw, prim, nr_prims))
+      return GL_FALSE;
+
    brw_validate_textures( brw );
 
    /* Bind all inputs, derive varying and size information:
@@ -275,6 +276,7 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
    brw->vb.min_index = min_index;
    brw->vb.max_index = max_index;
    brw->state.dirty.brw |= BRW_NEW_VERTICES;
+
    /* Have to validate state quite late.  Will rebuild tnl_program,
     * which depends on varying information.  
     * 
@@ -289,59 +291,58 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
       return GL_TRUE;
    }
 
-   /* Flush the batch if it's approaching full, so that we don't wrap while
-    * we've got validated state that needs to be in the same batch as the
-    * primitives.  This fraction is just a guess (minimal full state plus
-    * a primitive is around 512 bytes), and would be better if we had
-    * an upper bound of how much we might emit in a single
-    * brw_try_draw_prims().
-    */
-   intel_batchbuffer_require_space(intel->batch, intel->batch->size / 4,
-				   LOOP_CLIPRECTS);
-   {
-      /* Set the first primitive early, ahead of validate_state:
+   for (i = 0; i < nr_prims; i++) {
+      uint32_t hw_prim;
+
+      /* Flush the batch if it's approaching full, so that we don't wrap while
+       * we've got validated state that needs to be in the same batch as the
+       * primitives.  This fraction is just a guess (minimal full state plus
+       * a primitive is around 512 bytes), and would be better if we had
+       * an upper bound of how much we might emit in a single
+       * brw_try_draw_prims().
        */
-      brw_set_prim(brw, prim[0].mode);
+      intel_batchbuffer_require_space(intel->batch, intel->batch->size / 4,
+				      LOOP_CLIPRECTS);
 
-      brw_validate_state( brw );
+      hw_prim = brw_set_prim(brw, prim[i].mode);
 
-      /* Various fallback checks:
-       */
-      if (brw->intel.Fallback) 
-	 goto out;
+      if (first_time || (brw->state.dirty.brw & BRW_NEW_PRIMITIVE)) {
+	 first_time = GL_FALSE;
 
-      if (check_fallbacks( brw, prim, nr_prims ))
-	 goto out;
+	 brw_validate_state(brw);
 
-      /* Check that we can fit our state in with our existing batchbuffer, or
-       * flush otherwise.
-       */
-      if (dri_bufmgr_check_aperture_space(brw->state.validated_bos,
-					  brw->state.validated_bo_count)) {
-	 static GLboolean warned;
-	 intel_batchbuffer_flush(intel->batch);
-
-	 /* Validate the state after we flushed the batch (which would have
-	  * changed the set of dirty state).  If we still fail to
-	  * check_aperture, warn of what's happening, but attempt to continue
-	  * on since it may succeed anyway, and the user would probably rather
-	  * see a failure and a warning than a fallback.
+	 /* Various fallback checks:  */
+	 if (brw->intel.Fallback)
+	    goto out;
+
+	 /* Check that we can fit our state in with our existing batchbuffer, or
+	  * flush otherwise.
 	  */
-	 brw_validate_state(brw);
-	 if (!warned &&
-	     dri_bufmgr_check_aperture_space(brw->state.validated_bos,
+	 if (dri_bufmgr_check_aperture_space(brw->state.validated_bos,
 					     brw->state.validated_bo_count)) {
-	    warn = GL_TRUE;
-	    warned = GL_TRUE;
+	    static GLboolean warned;
+	    intel_batchbuffer_flush(intel->batch);
+
+	    /* Validate the state after we flushed the batch (which would have
+	     * changed the set of dirty state).  If we still fail to
+	     * check_aperture, warn of what's happening, but attempt to continue
+	     * on since it may succeed anyway, and the user would probably rather
+	     * see a failure and a warning than a fallback.
+	     */
+	    brw_validate_state(brw);
+	    if (!warned &&
+		dri_bufmgr_check_aperture_space(brw->state.validated_bos,
+						brw->state.validated_bo_count)) {
+	       warn = GL_TRUE;
+	       warned = GL_TRUE;
+	    }
 	 }
-      }
-
-      brw_upload_state(brw);
 
-      for (i = 0; i < nr_prims; i++) {
-	 brw_emit_prim(brw, &prim[i]);
+	 brw_upload_state(brw);
       }
 
+      brw_emit_prim(brw, &prim[i], hw_prim);
+
       retval = GL_TRUE;
    }
 
diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c
index 73d6dea01e..02998d5957 100644
--- a/src/mesa/drivers/dri/i965/brw_draw_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c
@@ -156,7 +156,13 @@ static GLuint byte_types_scale[5] = {
 };
 
 
-static GLuint get_surface_type( GLenum type, GLuint size, GLboolean normalized )
+/**
+ * Given vertex array type/size/format/normalized info, return
+ * the appopriate hardware surface type.
+ * Format will be GL_RGBA or possibly GL_BGRA for GLubyte[4] color arrays.
+ */
+static GLuint get_surface_type( GLenum type, GLuint size,
+                                GLenum format, GLboolean normalized )
 {
    if (INTEL_DEBUG & DEBUG_VERTS)
       _mesa_printf("type %s size %d normalized %d\n", 
@@ -171,11 +177,20 @@ static GLuint get_surface_type( GLenum type, GLuint size, GLboolean normalized )
       case GL_BYTE: return byte_types_norm[size];
       case GL_UNSIGNED_INT: return uint_types_norm[size];
       case GL_UNSIGNED_SHORT: return ushort_types_norm[size];
-      case GL_UNSIGNED_BYTE: return ubyte_types_norm[size];
+      case GL_UNSIGNED_BYTE:
+         if (format == GL_BGRA) {
+            /* See GL_EXT_vertex_array_bgra */
+            assert(size == 4);
+            return BRW_SURFACEFORMAT_B8G8R8A8_UNORM;
+         }
+         else {
+            return ubyte_types_norm[size];
+         }
       default: assert(0); return 0;
       }      
    }
    else {
+      assert(format == GL_RGBA); /* sanity check */
       switch (type) {
       case GL_DOUBLE: return double_types[size];
       case GL_FLOAT: return float_types[size];
@@ -281,7 +296,7 @@ copy_array_to_vbo_array( struct brw_context *brw,
    } else {
       void *data;
       char *dest;
-      const char *src = element->glarray->Ptr;
+      const unsigned char *src = element->glarray->Ptr;
       int i;
 
       data = _mesa_malloc(dst_stride * element->count);
@@ -484,6 +499,7 @@ static void brw_emit_vertices(struct brw_context *brw)
       struct brw_vertex_element *input = enabled[i];
       uint32_t format = get_surface_type(input->glarray->Type,
 					 input->glarray->Size,
+					 input->glarray->Format,
 					 input->glarray->Normalized);
       uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
       uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
@@ -549,7 +565,7 @@ static void brw_prepare_indices(struct brw_context *brw)
        */
       dri_bo_subdata(bo, offset, ib_size, index_buffer->ptr);
    } else {
-      offset = (GLuint)index_buffer->ptr;
+      offset = (GLuint) (unsigned long) index_buffer->ptr;
 
       /* If the index buffer isn't aligned to its element size, we have to
        * rebase it into a temporary.
diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h
index 49b422ee2f..9e2b39af9b 100644
--- a/src/mesa/drivers/dri/i965/brw_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_eu.h
@@ -129,17 +129,28 @@ static INLINE int type_sz( GLuint type )
    }
 }
 
+/**
+ * Construct a brw_reg.
+ * \param file  one of the BRW_x_REGISTER_FILE values
+ * \param nr  register number/index
+ * \param subnr  register sub number
+ * \param type  one of BRW_REGISTER_TYPE_x
+ * \param vstride  one of BRW_VERTICAL_STRIDE_x
+ * \param width  one of BRW_WIDTH_x
+ * \param hstride  one of BRW_HORIZONTAL_STRIDE_x
+ * \param swizzle  one of BRW_SWIZZLE_x
+ * \param writemask  WRITEMASK_X/Y/Z/W bitfield
+ */
 static INLINE struct brw_reg brw_reg( GLuint file,
-					GLuint nr,
-					GLuint subnr,
-					GLuint type,
-					GLuint vstride,
-					GLuint width,
-					GLuint hstride,
-					GLuint swizzle,
-					GLuint writemask)
-{
-      
+                                      GLuint nr,
+                                      GLuint subnr,
+                                      GLuint type,
+                                      GLuint vstride,
+                                      GLuint width,
+                                      GLuint hstride,
+                                      GLuint swizzle,
+                                      GLuint writemask )
+{
    struct brw_reg reg;
    reg.type = type;
    reg.file = file;
@@ -166,6 +177,7 @@ static INLINE struct brw_reg brw_reg( GLuint file,
    return reg;
 }
 
+/** Construct float[16] register */
 static INLINE struct brw_reg brw_vec16_reg( GLuint file,
 					      GLuint nr,
 					      GLuint subnr )
@@ -181,6 +193,7 @@ static INLINE struct brw_reg brw_vec16_reg( GLuint file,
 		  WRITEMASK_XYZW);
 }
 
+/** Construct float[8] register */
 static INLINE struct brw_reg brw_vec8_reg( GLuint file,
 					     GLuint nr,
 					     GLuint subnr )
@@ -196,7 +209,7 @@ static INLINE struct brw_reg brw_vec8_reg( GLuint file,
 		  WRITEMASK_XYZW);
 }
 
-
+/** Construct float[4] register */
 static INLINE struct brw_reg brw_vec4_reg( GLuint file,
 					      GLuint nr,
 					      GLuint subnr )
@@ -212,7 +225,7 @@ static INLINE struct brw_reg brw_vec4_reg( GLuint file,
 		  WRITEMASK_XYZW);
 }
 
-
+/** Construct float[2] register */
 static INLINE struct brw_reg brw_vec2_reg( GLuint file,
 					      GLuint nr,
 					      GLuint subnr )
@@ -228,6 +241,7 @@ static INLINE struct brw_reg brw_vec2_reg( GLuint file,
 		  WRITEMASK_XY);
 }
 
+/** Construct float[1] register */
 static INLINE struct brw_reg brw_vec1_reg( GLuint file,
 					     GLuint nr,
 					     GLuint subnr )
@@ -277,6 +291,7 @@ static INLINE struct brw_reg byte_offset( struct brw_reg reg,
 }
    
 
+/** Construct unsigned word[16] register */
 static INLINE struct brw_reg brw_uw16_reg( GLuint file,
 					     GLuint nr,
 					     GLuint subnr )
@@ -284,6 +299,7 @@ static INLINE struct brw_reg brw_uw16_reg( GLuint file,
    return suboffset(retype(brw_vec16_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr);
 }
 
+/** Construct unsigned word[8] register */
 static INLINE struct brw_reg brw_uw8_reg( GLuint file,
 					    GLuint nr,
 					    GLuint subnr )
@@ -291,6 +307,7 @@ static INLINE struct brw_reg brw_uw8_reg( GLuint file,
    return suboffset(retype(brw_vec8_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr);
 }
 
+/** Construct unsigned word[1] register */
 static INLINE struct brw_reg brw_uw1_reg( GLuint file,
 					    GLuint nr,
 					    GLuint subnr )
@@ -311,6 +328,7 @@ static INLINE struct brw_reg brw_imm_reg( GLuint type )
 		   0);      
 }
 
+/** Construct float immediate register */
 static INLINE struct brw_reg brw_imm_f( GLfloat f )
 {
    struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_F);
@@ -318,6 +336,7 @@ static INLINE struct brw_reg brw_imm_f( GLfloat f )
    return imm;
 }
 
+/** Construct integer immediate register */
 static INLINE struct brw_reg brw_imm_d( GLint d )
 {
    struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_D);
@@ -325,6 +344,7 @@ static INLINE struct brw_reg brw_imm_d( GLint d )
    return imm;
 }
 
+/** Construct uint immediate register */
 static INLINE struct brw_reg brw_imm_ud( GLuint ud )
 {
    struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UD);
@@ -332,6 +352,7 @@ static INLINE struct brw_reg brw_imm_ud( GLuint ud )
    return imm;
 }
 
+/** Construct ushort immediate register */
 static INLINE struct brw_reg brw_imm_uw( GLushort uw )
 {
    struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UW);
@@ -339,6 +360,7 @@ static INLINE struct brw_reg brw_imm_uw( GLushort uw )
    return imm;
 }
 
+/** Construct short immediate register */
 static INLINE struct brw_reg brw_imm_w( GLshort w )
 {
    struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_W);
@@ -350,8 +372,7 @@ static INLINE struct brw_reg brw_imm_w( GLshort w )
  * numbers alias with _V and _VF below:
  */
 
-/* Vector of eight signed half-byte values: 
- */
+/** Construct vector of eight signed half-byte values */
 static INLINE struct brw_reg brw_imm_v( GLuint v )
 {
    struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_V);
@@ -362,8 +383,7 @@ static INLINE struct brw_reg brw_imm_v( GLuint v )
    return imm;
 }
 
-/* Vector of four 8-bit float values:
- */
+/** Construct vector of four 8-bit float values */
 static INLINE struct brw_reg brw_imm_vf( GLuint v )
 {
    struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF);
@@ -400,44 +420,43 @@ static INLINE struct brw_reg brw_address( struct brw_reg reg )
    return brw_imm_uw(reg.nr * REG_SIZE + reg.subnr);
 }
 
-
-static INLINE struct brw_reg brw_vec1_grf( GLuint nr,
-					       GLuint subnr )
+/** Construct float[1] general-purpose register */
+static INLINE struct brw_reg brw_vec1_grf( GLuint nr, GLuint subnr )
 {
    return brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
 }
 
-static INLINE struct brw_reg brw_vec8_grf( GLuint nr,
-					     GLuint subnr )
+/** Construct float[2] general-purpose register */
+static INLINE struct brw_reg brw_vec2_grf( GLuint nr, GLuint subnr )
 {
-   return brw_vec8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
+   return brw_vec2_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
 }
 
-static INLINE struct brw_reg brw_vec4_grf( GLuint nr,
-					     GLuint subnr )
+/** Construct float[4] general-purpose register */
+static INLINE struct brw_reg brw_vec4_grf( GLuint nr, GLuint subnr )
 {
    return brw_vec4_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
 }
 
-
-static INLINE struct brw_reg brw_vec2_grf( GLuint nr,
-					     GLuint subnr )
+/** Construct float[8] general-purpose register */
+static INLINE struct brw_reg brw_vec8_grf( GLuint nr, GLuint subnr )
 {
-   return brw_vec2_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
+   return brw_vec8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
 }
 
-static INLINE struct brw_reg brw_uw8_grf( GLuint nr,
-					    GLuint subnr )
+
+static INLINE struct brw_reg brw_uw8_grf( GLuint nr, GLuint subnr )
 {
    return brw_uw8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
 }
 
-static INLINE struct brw_reg brw_uw16_grf( GLuint nr,
-					    GLuint subnr )
+static INLINE struct brw_reg brw_uw16_grf( GLuint nr, GLuint subnr )
 {
    return brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr);
 }
 
+
+/** Construct null register (usually used for setting condition codes) */
 static INLINE struct brw_reg brw_null_reg( void )
 {
    return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, 
@@ -524,13 +543,13 @@ static INLINE struct brw_reg stride( struct brw_reg reg,
 				       GLuint width,
 				       GLuint hstride )
 {
-   
    reg.vstride = cvt(vstride);
    reg.width = cvt(width) - 1;
    reg.hstride = cvt(hstride);
    return reg;
 }
 
+
 static INLINE struct brw_reg vec16( struct brw_reg reg )
 {
    return stride(reg, 16,16,1);
@@ -556,6 +575,7 @@ static INLINE struct brw_reg vec1( struct brw_reg reg )
    return stride(reg, 0,1,0);
 }
 
+
 static INLINE struct brw_reg get_element( struct brw_reg reg, GLuint elt )
 {
    return vec1(suboffset(reg, elt));
@@ -687,7 +707,7 @@ static INLINE struct brw_indirect brw_indirect( GLuint addr_subnr, GLint offset
 
 static INLINE struct brw_instruction *current_insn( struct brw_compile *p)
 {
-	return &p->store[p->nr_insn];
+   return &p->store[p->nr_insn];
 }
 
 void brw_pop_insn_state( struct brw_compile *p );
@@ -733,6 +753,7 @@ ALU2(ADD)
 ALU2(MUL)
 ALU1(FRC)
 ALU1(RNDD)
+ALU1(RNDZ)
 ALU2(MAC)
 ALU2(MACH)
 ALU1(LZD)
diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index ce4cf46cfa..4e099b5945 100644
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -439,6 +439,7 @@ ALU2(ADD)
 ALU2(MUL)
 ALU1(FRC)
 ALU1(RNDD)
+ALU1(RNDZ)
 ALU2(MAC)
 ALU2(MACH)
 ALU1(LZD)
diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c
index 4ea660a51a..e63098fdd4 100644
--- a/src/mesa/drivers/dri/i965/brw_fallback.c
+++ b/src/mesa/drivers/dri/i965/brw_fallback.c
@@ -47,20 +47,12 @@ static GLboolean do_check_fallback(struct brw_context *brw)
    GLcontext *ctx = &brw->intel.ctx;
    GLuint i;
 
-   /* BRW_NEW_METAOPS
-    */
-   if (brw->metaops.active)
-      return GL_FALSE;
-
    if (brw->intel.no_rast) {
       DBG("FALLBACK: rasterization disabled\n");
       return GL_TRUE;
    }
 
    /* _NEW_RENDERMODE
-    *
-    * XXX: need to save/restore RenderMode in metaops state, or
-    * somehow move to a new attribs pointer:
     */
    if (ctx->RenderMode != GL_RENDER) {
       DBG("FALLBACK: render mode\n");
@@ -70,7 +62,7 @@ static GLboolean do_check_fallback(struct brw_context *brw)
    /* _NEW_TEXTURE:
     */
    for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
-      struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[i];
+      struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
       if (texUnit->_ReallyEnabled) {
 	 struct intel_texture_object *intelObj = intel_texture_object(texUnit->_Current);
 	 struct gl_texture_image *texImage = intelObj->base.Image[0][intelObj->firstLevel];
@@ -83,7 +75,7 @@ static GLboolean do_check_fallback(struct brw_context *brw)
    
    /* _NEW_STENCIL 
     */
-   if (brw->attribs.Stencil->Enabled && 
+   if (ctx->Stencil.Enabled && 
        !brw->intel.hw_stencil) {
       DBG("FALLBACK: stencil\n");
       return GL_TRUE;
@@ -101,7 +93,7 @@ static void check_fallback(struct brw_context *brw)
 const struct brw_tracked_state brw_check_fallback = {
    .dirty = {
       .mesa = _NEW_BUFFERS | _NEW_RENDERMODE | _NEW_TEXTURE | _NEW_STENCIL,
-      .brw  = BRW_NEW_METAOPS,
+      .brw  = 0,
       .cache = 0
    },
    .prepare = check_fallback
diff --git a/src/mesa/drivers/dri/i965/brw_metaops.c b/src/mesa/drivers/dri/i965/brw_metaops.c
deleted file mode 100644
index 41bfa2e256..0000000000
--- a/src/mesa/drivers/dri/i965/brw_metaops.c
+++ /dev/null
@@ -1,583 +0,0 @@
-/*
- Copyright (C) Intel Corp.  2006.  All Rights Reserved.
- Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
- develop this 3D driver.
- 
- Permission is hereby granted, free of charge, to any person obtaining
- a copy of this software and associated documentation files (the
- "Software"), to deal in the Software without restriction, including
- without limitation the rights to use, copy, modify, merge, publish,
- distribute, sublicense, and/or sell copies of the Software, and to
- permit persons to whom the Software is furnished to do so, subject to
- the following conditions:
- 
- The above copyright notice and this permission notice (including the
- next paragraph) shall be included in all copies or substantial
- portions of the Software.
- 
- THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
- IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
- LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
- OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
- WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- 
- **********************************************************************/
- /*
-  * Authors:
-  *   Keith Whitwell <keith@tungstengraphics.com>
-  *   frame buffer texture by Gary Wong <gtw@gnu.org>
-  */
- 
-
-
-#include "main/glheader.h"
-#include "main/context.h"
-#include "main/macros.h"
-
-#include "shader/arbprogparse.h"
-
-#include "intel_screen.h"
-#include "intel_batchbuffer.h"
-#include "intel_regions.h"
-#include "intel_buffers.h"
-
-#include "brw_context.h"
-#include "brw_defines.h"
-#include "brw_draw.h"
-#include "brw_fallback.h"
-
-#define INIT(brw, STRUCT, ATTRIB) 		\
-do {						\
-   brw->attribs.ATTRIB = &ctx->ATTRIB;		\
-} while (0)
-
-#define DUP(brw, STRUCT, ATTRIB) 		\
-do {						\
-   brw->metaops.attribs.ATTRIB = MALLOC_STRUCT(STRUCT);	\
-   memcpy(brw->metaops.attribs.ATTRIB, 			\
-	  brw->attribs.ATTRIB,			\
-	  sizeof(struct STRUCT));		\
-} while (0)
-
-
-#define INSTALL(brw, ATTRIB, STATE)		\
-do {						\
-   brw->attribs.ATTRIB = brw->metaops.attribs.ATTRIB;	\
-   brw->state.dirty.mesa |= STATE;		\
-} while (0)
-
-#define RESTORE(brw, ATTRIB, STATE)			\
-do {							\
-   brw->attribs.ATTRIB = &brw->intel.ctx.ATTRIB;	\
-   brw->state.dirty.mesa |= STATE;			\
-} while (0)
-
-static void init_attribs( struct brw_context *brw )
-{
-   DUP(brw, gl_colorbuffer_attrib, Color);
-   DUP(brw, gl_depthbuffer_attrib, Depth);
-   DUP(brw, gl_fog_attrib, Fog);
-   DUP(brw, gl_hint_attrib, Hint);
-   DUP(brw, gl_light_attrib, Light);
-   DUP(brw, gl_line_attrib, Line);
-   DUP(brw, gl_point_attrib, Point);
-   DUP(brw, gl_polygon_attrib, Polygon);
-   DUP(brw, gl_scissor_attrib, Scissor);
-   DUP(brw, gl_stencil_attrib, Stencil);
-   DUP(brw, gl_texture_attrib, Texture);
-   DUP(brw, gl_transform_attrib, Transform);
-   DUP(brw, gl_viewport_attrib, Viewport);
-   DUP(brw, gl_vertex_program_state, VertexProgram);
-   DUP(brw, gl_fragment_program_state, FragmentProgram);
-}
-
-static void install_attribs( struct brw_context *brw )
-{
-   INSTALL(brw, Color, _NEW_COLOR);
-   INSTALL(brw, Depth, _NEW_DEPTH);
-   INSTALL(brw, Fog, _NEW_FOG);
-   INSTALL(brw, Hint, _NEW_HINT);
-   INSTALL(brw, Light, _NEW_LIGHT);
-   INSTALL(brw, Line, _NEW_LINE);
-   INSTALL(brw, Point, _NEW_POINT);
-   INSTALL(brw, Polygon, _NEW_POLYGON);
-   INSTALL(brw, Scissor, _NEW_SCISSOR);
-   INSTALL(brw, Stencil, _NEW_STENCIL);
-   INSTALL(brw, Texture, _NEW_TEXTURE);
-   INSTALL(brw, Transform, _NEW_TRANSFORM);
-   INSTALL(brw, Viewport, _NEW_VIEWPORT);
-   INSTALL(brw, VertexProgram, _NEW_PROGRAM);
-   INSTALL(brw, FragmentProgram, _NEW_PROGRAM);
-}
-
-static void restore_attribs( struct brw_context *brw )
-{
-   RESTORE(brw, Color, _NEW_COLOR);
-   RESTORE(brw, Depth, _NEW_DEPTH);
-   RESTORE(brw, Fog, _NEW_FOG);
-   RESTORE(brw, Hint, _NEW_HINT);
-   RESTORE(brw, Light, _NEW_LIGHT);
-   RESTORE(brw, Line, _NEW_LINE);
-   RESTORE(brw, Point, _NEW_POINT);
-   RESTORE(brw, Polygon, _NEW_POLYGON);
-   RESTORE(brw, Scissor, _NEW_SCISSOR);
-   RESTORE(brw, Stencil, _NEW_STENCIL);
-   RESTORE(brw, Texture, _NEW_TEXTURE);
-   RESTORE(brw, Transform, _NEW_TRANSFORM);
-   RESTORE(brw, Viewport, _NEW_VIEWPORT);
-   RESTORE(brw, VertexProgram, _NEW_PROGRAM);
-   RESTORE(brw, FragmentProgram, _NEW_PROGRAM);
-}
-
-
-static const char *vp_prog =
-      "!!ARBvp1.0\n"
-      "MOV  result.color, vertex.color;\n"
-      "MOV  result.position, vertex.position;\n"
-      "END\n";
-
-static const char *fp_prog =
-      "!!ARBfp1.0\n"
-      "MOV result.color, fragment.color;\n"
-      "END\n";
-
-static const char *fp_tex_prog =
-      "!!ARBfp1.0\n"
-      "TEMP a;\n"
-      "ADD a, fragment.position, program.local[0];\n"
-      "MUL a, a, program.local[1];\n"
-      "TEX result.color, a, texture[0], 2D;\n"
-      "MOV result.depth.z, fragment.position;\n"
-      "END\n";
-
-/* Derived values of importance:
- *
- *   FragmentProgram->_Current
- *   VertexProgram->_Enabled
- *   brw->vertex_program
- *   DrawBuffer->_ColorDrawBufferIndexes[0]
- * 
- *
- * More if drawpixels-through-texture is added.  
- */
-static void init_metaops_state( struct brw_context *brw )
-{
-   GLcontext *ctx = &brw->intel.ctx;
-
-   brw->metaops.vbo = ctx->Driver.NewBufferObject(ctx, 1, GL_ARRAY_BUFFER_ARB);
-
-   ctx->Driver.BufferData(ctx,
-			  GL_ARRAY_BUFFER_ARB,
-			  4096,
-			  NULL,
-			  GL_DYNAMIC_DRAW_ARB,
-			  brw->metaops.vbo);
-
-   brw->metaops.fp = (struct gl_fragment_program *)
-      ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 1 );
-
-   brw->metaops.fp_tex = (struct gl_fragment_program *)
-      ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 1 );
-
-   brw->metaops.vp = (struct gl_vertex_program *)
-      ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 1 );
-
-   _mesa_parse_arb_fragment_program(ctx, GL_FRAGMENT_PROGRAM_ARB, 
-				    fp_prog, strlen(fp_prog),
-				    brw->metaops.fp);
-
-   _mesa_parse_arb_fragment_program(ctx, GL_FRAGMENT_PROGRAM_ARB, 
-				    fp_tex_prog, strlen(fp_tex_prog),
-				    brw->metaops.fp_tex);
-
-   _mesa_parse_arb_vertex_program(ctx, GL_VERTEX_PROGRAM_ARB, 
-				  vp_prog, strlen(vp_prog),
-				  brw->metaops.vp);
-
-   brw->metaops.attribs.VertexProgram->_Current = brw->metaops.vp;
-   brw->metaops.attribs.VertexProgram->_Enabled = GL_TRUE;
-
-   brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp;
-}
-
-static void meta_flat_shade( struct intel_context *intel )
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-
-   brw->metaops.attribs.Light->ShadeModel = GL_FLAT;
-   brw->state.dirty.mesa |= _NEW_LIGHT;
-}
-
-
-static void meta_no_stencil_write( struct intel_context *intel )
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-
-   brw->metaops.attribs.Stencil->Enabled = GL_FALSE;
-   brw->metaops.attribs.Stencil->WriteMask[0] = GL_FALSE; 
-   brw->state.dirty.mesa |= _NEW_STENCIL;
-}
-
-static void meta_no_depth_write( struct intel_context *intel )
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-
-   brw->metaops.attribs.Depth->Test = GL_FALSE;
-   brw->metaops.attribs.Depth->Mask = GL_FALSE;
-   brw->state.dirty.mesa |= _NEW_DEPTH;
-}
-
-
-static void meta_depth_replace( struct intel_context *intel )
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-
-   /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_TRUE )
-    * ctx->Driver.DepthMask( ctx, GL_TRUE )
-    */
-   brw->metaops.attribs.Depth->Test = GL_TRUE;
-   brw->metaops.attribs.Depth->Mask = GL_TRUE;
-   brw->state.dirty.mesa |= _NEW_DEPTH;
-
-   /* ctx->Driver.DepthFunc( ctx, GL_ALWAYS )
-    */
-   brw->metaops.attribs.Depth->Func = GL_ALWAYS;
-
-   brw->state.dirty.mesa |= _NEW_DEPTH;
-}
-
-
-static void meta_stencil_replace( struct intel_context *intel,
-				 GLuint s_mask,
-				 GLuint s_clear)
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-
-   brw->metaops.attribs.Stencil->Enabled = GL_TRUE;
-   brw->metaops.attribs.Stencil->WriteMask[0] = s_mask;
-   brw->metaops.attribs.Stencil->ValueMask[0] = 0xff;
-   brw->metaops.attribs.Stencil->Ref[0] = s_clear;
-   brw->metaops.attribs.Stencil->Function[0] = GL_ALWAYS;
-   brw->metaops.attribs.Stencil->FailFunc[0] = GL_REPLACE;
-   brw->metaops.attribs.Stencil->ZPassFunc[0] = GL_REPLACE;
-   brw->metaops.attribs.Stencil->ZFailFunc[0] = GL_REPLACE;
-   brw->state.dirty.mesa |= _NEW_STENCIL;
-}
-
-
-static void meta_color_mask( struct intel_context *intel, GLboolean state )
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-
-   if (state)
-      COPY_4V(brw->metaops.attribs.Color->ColorMask, 
-	      brw->intel.ctx.Color.ColorMask); 
-   else
-      ASSIGN_4V(brw->metaops.attribs.Color->ColorMask, 0, 0, 0, 0);
-
-   brw->state.dirty.mesa |= _NEW_COLOR;
-}
-
-static void meta_no_texture( struct intel_context *intel )
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-   
-   brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp;
-   
-   brw->metaops.attribs.Texture->CurrentUnit = 0;
-   brw->metaops.attribs.Texture->_EnabledUnits = 0;
-   brw->metaops.attribs.Texture->_EnabledCoordUnits = 0;
-   brw->metaops.attribs.Texture->Unit[ 0 ].Enabled = 0;
-   brw->metaops.attribs.Texture->Unit[ 0 ]._ReallyEnabled = 0;
-
-   brw->state.dirty.mesa |= _NEW_TEXTURE | _NEW_PROGRAM;
-}
-
-static void meta_texture_blend_replace(struct intel_context *intel)
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-
-   brw->metaops.attribs.Texture->CurrentUnit = 0;
-   brw->metaops.attribs.Texture->_EnabledUnits = 1;
-   brw->metaops.attribs.Texture->_EnabledCoordUnits = 1;
-   brw->metaops.attribs.Texture->Unit[ 0 ].Enabled = TEXTURE_2D_BIT;
-   brw->metaops.attribs.Texture->Unit[ 0 ]._ReallyEnabled = TEXTURE_2D_BIT;
-   brw->metaops.attribs.Texture->Unit[ 0 ].Current2D =
-      intel->frame_buffer_texobj;
-   brw->metaops.attribs.Texture->Unit[ 0 ]._Current =
-      intel->frame_buffer_texobj;
-
-   brw->state.dirty.mesa |= _NEW_TEXTURE | _NEW_PROGRAM;
-}
-
-static void meta_import_pixel_state(struct intel_context *intel)
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-   
-   RESTORE(brw, Color, _NEW_COLOR);
-   RESTORE(brw, Depth, _NEW_DEPTH);
-   RESTORE(brw, Fog, _NEW_FOG);
-   RESTORE(brw, Scissor, _NEW_SCISSOR);
-   RESTORE(brw, Stencil, _NEW_STENCIL);
-   RESTORE(brw, Texture, _NEW_TEXTURE);
-   RESTORE(brw, FragmentProgram, _NEW_PROGRAM);
-}
-
-static void meta_frame_buffer_texture( struct intel_context *intel,
-				       GLint xoff, GLint yoff )
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-   struct intel_region *region = intel_drawbuf_region( intel );
-   
-   INSTALL(brw, FragmentProgram, _NEW_PROGRAM);
-
-   brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp_tex;
-   /* This is unfortunate, but seems to be necessary, since later on we
-      will end up calling _mesa_load_state_parameters to lookup the
-      local params (below), and that will want to look in ctx.FragmentProgram
-      instead of brw->attribs.FragmentProgram. */
-   intel->ctx.FragmentProgram.Current = brw->metaops.fp_tex;
-
-   brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 0 ] = xoff;
-   brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 1 ] = yoff;
-   brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 2 ] = 0.0;
-   brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 3 ] = 0.0;
-   brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 0 ] =
-      1.0 / region->pitch;
-   brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 1 ] =
-      -1.0 / region->height;
-   brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 2 ] = 0.0;
-   brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 3 ] = 1.0;
-   
-   brw->state.dirty.mesa |= _NEW_PROGRAM;
-}
-
-
-static void meta_draw_region( struct intel_context *intel,
-			     struct intel_region *draw_region,
-			     struct intel_region *depth_region )
-{
-   struct brw_context *brw = brw_context(&intel->ctx);
-
-   if (!brw->metaops.saved_draw_region) {
-      brw->metaops.saved_draw_region = brw->state.draw_regions[0];
-      brw->metaops.saved_nr_draw_regions = brw->state.nr_draw_regions;
-      brw->metaops.saved_depth_region = brw->state.depth_region;
-   }
-
-   brw->state.draw_regions[0] = draw_region;
-   brw->state.nr_draw_regions = 1;
-   brw->state.depth_region = depth_region;
-
-   if (intel->frame_buffer_texobj != NULL)
-      brw_FrameBufferTexDestroy(brw);
-
-   if (draw_region)
-       brw_FrameBufferTexInit(brw, draw_region);
-
-   brw->state.dirty.mesa |= _NEW_BUFFERS;
-}
-
-
-static void meta_draw_quad(struct intel_context *intel, 
-			   GLfloat x0, GLfloat x1,
-			   GLfloat y0, GLfloat y1, 
-			   GLfloat z,
-			   GLuint color,
-			   GLfloat s0, GLfloat s1,
-			   GLfloat t0, GLfloat t1)
-{
-   GLcontext *ctx = &intel->ctx;
-   struct brw_context *brw = brw_context(&intel->ctx);
-   struct gl_client_array pos_array;
-   struct gl_client_array color_array;
-   struct gl_client_array *attribs[VERT_ATTRIB_MAX];
-   struct _mesa_prim prim[1];
-   GLfloat pos[4][3];
-
-   ctx->Driver.BufferData(ctx,
-			  GL_ARRAY_BUFFER_ARB,
-			  sizeof(pos) + sizeof(color),
-			  NULL,
-			  GL_DYNAMIC_DRAW_ARB,
-			  brw->metaops.vbo);
-
-   pos[0][0] = x0;
-   pos[0][1] = y0;
-   pos[0][2] = z;
-
-   pos[1][0] = x1;
-   pos[1][1] = y0;
-   pos[1][2] = z;
-
-   pos[2][0] = x1;
-   pos[2][1] = y1;
-   pos[2][2] = z;
-
-   pos[3][0] = x0;
-   pos[3][1] = y1;
-   pos[3][2] = z;
-
-   ctx->Driver.BufferSubData(ctx,
-			     GL_ARRAY_BUFFER_ARB,
-			     0,
-			     sizeof(pos),
-			     pos,
-			     brw->metaops.vbo);
-
-   /* Convert incoming ARGB to required RGBA */
-   /* Note this color is stored as GL_UNSIGNED_BYTE */
-   color = (color & 0xff00ff00) | (((color >> 16) | (color << 16)) & 0xff00ff);
-
-   ctx->Driver.BufferSubData(ctx,
-			     GL_ARRAY_BUFFER_ARB,
-			     sizeof(pos),
-			     sizeof(color),
-			     &color,
-			     brw->metaops.vbo);
-
-   /* Ignoring texture coords. 
-    */
-
-   memset(attribs, 0, VERT_ATTRIB_MAX * sizeof(*attribs));
-
-   attribs[VERT_ATTRIB_POS] = &pos_array;
-   attribs[VERT_ATTRIB_POS]->Ptr = 0;
-   attribs[VERT_ATTRIB_POS]->Type = GL_FLOAT;
-   attribs[VERT_ATTRIB_POS]->Enabled = 1;
-   attribs[VERT_ATTRIB_POS]->Size = 3;
-   attribs[VERT_ATTRIB_POS]->StrideB = 3 * sizeof(GLfloat);
-   attribs[VERT_ATTRIB_POS]->Stride = 3 * sizeof(GLfloat);
-   attribs[VERT_ATTRIB_POS]->_MaxElement = 4;
-   attribs[VERT_ATTRIB_POS]->Normalized = 0;
-   attribs[VERT_ATTRIB_POS]->BufferObj = brw->metaops.vbo;
-
-   attribs[VERT_ATTRIB_COLOR0] = &color_array;
-   attribs[VERT_ATTRIB_COLOR0]->Ptr = (const GLubyte *)sizeof(pos);
-   attribs[VERT_ATTRIB_COLOR0]->Type = GL_UNSIGNED_BYTE;
-   attribs[VERT_ATTRIB_COLOR0]->Enabled = 1;
-   attribs[VERT_ATTRIB_COLOR0]->Size = 4;
-   attribs[VERT_ATTRIB_COLOR0]->StrideB = 0;
-   attribs[VERT_ATTRIB_COLOR0]->Stride = 0;
-   attribs[VERT_ATTRIB_COLOR0]->_MaxElement = 1;
-   attribs[VERT_ATTRIB_COLOR0]->Normalized = 1;
-   attribs[VERT_ATTRIB_COLOR0]->BufferObj = brw->metaops.vbo;
-   
-   /* Just ignoring texture coordinates for now. 
-    */
-
-   memset(prim, 0, sizeof(*prim));
-
-   prim[0].mode = GL_TRIANGLE_FAN;
-   prim[0].begin = 1;
-   prim[0].end = 1;
-   prim[0].weak = 0;
-   prim[0].pad = 0;
-   prim[0].start = 0;
-   prim[0].count = 4;
-
-   brw_draw_prims(&brw->intel.ctx, 
-		  (const struct gl_client_array **)attribs,
-		  prim, 1,
-		  NULL,
-		  0,
-		  3 );
-}
-
-
-static void install_meta_state( struct intel_context *intel )
-{
-   GLcontext *ctx = &intel->ctx;
-   struct brw_context *brw = brw_context(ctx);
-   GLuint i;
-
-   if (!brw->metaops.vbo) {
-      init_metaops_state(brw);
-   }
-
-   install_attribs(brw);
-   
-   meta_no_texture(&brw->intel);
-   meta_flat_shade(&brw->intel);
-   for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) {
-      brw->metaops.restore_draw_buffers[i]
-         = ctx->DrawBuffer->_ColorDrawBufferIndexes[i];
-   }
-   brw->metaops.restore_num_draw_buffers = ctx->DrawBuffer->_NumColorDrawBuffers;
-
-   brw->metaops.restore_fp = ctx->FragmentProgram.Current;
-
-   /* This works without adjusting refcounts.  Fix later? 
-    */
-   brw->metaops.saved_draw_region = brw->state.draw_regions[0];
-   brw->metaops.saved_nr_draw_regions = brw->state.nr_draw_regions;
-   brw->metaops.saved_depth_region = brw->state.depth_region;
-   brw->metaops.active = 1;
-   
-   brw->state.dirty.brw |= BRW_NEW_METAOPS;
-}
-
-static void leave_meta_state( struct intel_context *intel )
-{
-   GLcontext *ctx = &intel->ctx;
-   struct brw_context *brw = brw_context(ctx);
-   GLuint i;
-
-   restore_attribs(brw);
-
-   for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) {
-      ctx->DrawBuffer->_ColorDrawBufferIndexes[i]
-         = brw->metaops.restore_draw_buffers[i];
-   }
-   ctx->DrawBuffer->_NumColorDrawBuffers = brw->metaops.restore_num_draw_buffers;
-
-   ctx->FragmentProgram.Current = brw->metaops.restore_fp;
-
-   brw->state.draw_regions[0] = brw->metaops.saved_draw_region;
-   brw->state.nr_draw_regions = brw->metaops.saved_nr_draw_regions;
-   brw->state.depth_region = brw->metaops.saved_depth_region;
-   brw->metaops.saved_draw_region = NULL;
-   brw->metaops.saved_depth_region = NULL;
-   brw->metaops.active = 0;
-
-   brw->state.dirty.mesa |= _NEW_BUFFERS;
-   brw->state.dirty.brw |= BRW_NEW_METAOPS;
-}
-
-
-
-void brw_init_metaops( struct brw_context *brw )
-{
-   init_attribs(brw);
-
-
-   brw->intel.vtbl.install_meta_state = install_meta_state;
-   brw->intel.vtbl.leave_meta_state = leave_meta_state;
-   brw->intel.vtbl.meta_no_depth_write = meta_no_depth_write;
-   brw->intel.vtbl.meta_no_stencil_write = meta_no_stencil_write;
-   brw->intel.vtbl.meta_stencil_replace = meta_stencil_replace;
-   brw->intel.vtbl.meta_depth_replace = meta_depth_replace;
-   brw->intel.vtbl.meta_color_mask = meta_color_mask;
-   brw->intel.vtbl.meta_no_texture = meta_no_texture;
-   brw->intel.vtbl.meta_import_pixel_state = meta_import_pixel_state;
-   brw->intel.vtbl.meta_frame_buffer_texture = meta_frame_buffer_texture;
-   brw->intel.vtbl.meta_draw_region = meta_draw_region;
-   brw->intel.vtbl.meta_draw_quad = meta_draw_quad;
-   brw->intel.vtbl.meta_texture_blend_replace = meta_texture_blend_replace;
-/*    brw->intel.vtbl.meta_tex_rect_source = meta_tex_rect_source; */
-/*    brw->intel.vtbl.meta_draw_format = set_draw_format; */
-}
-
-void brw_destroy_metaops( struct brw_context *brw )
-{
-   GLcontext *ctx = &brw->intel.ctx;
-
-   if (brw->metaops.vbo)
-      ctx->Driver.DeleteBuffer( ctx, brw->metaops.vbo );
-
-/*    ctx->Driver.DeleteProgram( ctx, brw->metaops.fp ); */
-/*    ctx->Driver.DeleteProgram( ctx, brw->metaops.fp_tex ); */
-/*    ctx->Driver.DeleteProgram( ctx, brw->metaops.vp ); */
-}
diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c
index 627705fa9b..9dcdad7b4e 100644
--- a/src/mesa/drivers/dri/i965/brw_misc_state.c
+++ b/src/mesa/drivers/dri/i965/brw_misc_state.c
@@ -48,15 +48,16 @@
 
 static void upload_blend_constant_color(struct brw_context *brw)
 {
+   GLcontext *ctx = &brw->intel.ctx;
    struct brw_blend_constant_color bcc;
 
    memset(&bcc, 0, sizeof(bcc));      
    bcc.header.opcode = CMD_BLEND_CONSTANT_COLOR;
    bcc.header.length = sizeof(bcc)/4-2;
-   bcc.blend_constant_color[0] = brw->attribs.Color->BlendColor[0];
-   bcc.blend_constant_color[1] = brw->attribs.Color->BlendColor[1];
-   bcc.blend_constant_color[2] = brw->attribs.Color->BlendColor[2];
-   bcc.blend_constant_color[3] = brw->attribs.Color->BlendColor[3];
+   bcc.blend_constant_color[0] = ctx->Color.BlendColor[0];
+   bcc.blend_constant_color[1] = ctx->Color.BlendColor[1];
+   bcc.blend_constant_color[2] = ctx->Color.BlendColor[2];
+   bcc.blend_constant_color[3] = ctx->Color.BlendColor[3];
 
    BRW_CACHED_BATCH_STRUCT(brw, &bcc);
 }
@@ -154,10 +155,7 @@ static void upload_pipelined_state_pointers(struct brw_context *brw )
       OUT_RELOC(brw->gs.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 1);
    else
       OUT_BATCH(0);
-   if (!brw->metaops.active)
-      OUT_RELOC(brw->clip.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 1);
-   else
-      OUT_BATCH(0);
+   OUT_RELOC(brw->clip.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 1);
    OUT_RELOC(brw->sf.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
    OUT_RELOC(brw->wm.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
    OUT_RELOC(brw->cc.state_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
@@ -186,7 +184,7 @@ static void upload_psp_urb_cbs(struct brw_context *brw )
 const struct brw_tracked_state brw_psp_urb_cbs = {
    .dirty = {
       .mesa = 0,
-      .brw = BRW_NEW_URB_FENCE | BRW_NEW_METAOPS | BRW_NEW_BATCH,
+      .brw = BRW_NEW_URB_FENCE | BRW_NEW_BATCH,
       .cache = (CACHE_NEW_VS_UNIT | 
 		CACHE_NEW_GS_UNIT | 
 		CACHE_NEW_GS_PROG | 
@@ -284,6 +282,7 @@ const struct brw_tracked_state brw_depthbuffer = {
 
 static void upload_polygon_stipple(struct brw_context *brw)
 {
+   GLcontext *ctx = &brw->intel.ctx;
    struct brw_polygon_stipple bps;
    GLuint i;
 
@@ -292,7 +291,7 @@ static void upload_polygon_stipple(struct brw_context *brw)
    bps.header.length = sizeof(bps)/4-2;
 
    for (i = 0; i < 32; i++)
-      bps.stipple[i] = brw->attribs.PolygonStipple[31 - i]; /* invert */
+      bps.stipple[i] = ctx->PolygonStipple[31 - i]; /* invert */
 
    BRW_CACHED_BATCH_STRUCT(brw, &bps);
 }
@@ -370,6 +369,7 @@ const struct brw_tracked_state brw_aa_line_parameters = {
 
 static void upload_line_stipple(struct brw_context *brw)
 {
+   GLcontext *ctx = &brw->intel.ctx;
    struct brw_line_stipple bls;
    GLfloat tmp;
    GLint tmpi;
@@ -378,10 +378,10 @@ static void upload_line_stipple(struct brw_context *brw)
    bls.header.opcode = CMD_LINE_STIPPLE_PATTERN;
    bls.header.length = sizeof(bls)/4 - 2;
 
-   bls.bits0.pattern = brw->attribs.Line->StipplePattern;
-   bls.bits1.repeat_count = brw->attribs.Line->StippleFactor;
+   bls.bits0.pattern = ctx->Line.StipplePattern;
+   bls.bits1.repeat_count = ctx->Line.StippleFactor;
 
-   tmp = 1.0 / (GLfloat) brw->attribs.Line->StippleFactor;
+   tmp = 1.0 / (GLfloat) ctx->Line.StippleFactor;
    tmpi = tmp * (1<<13);
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_program.c b/src/mesa/drivers/dri/i965/brw_program.c
index c38610b24e..0c86911044 100644
--- a/src/mesa/drivers/dri/i965/brw_program.c
+++ b/src/mesa/drivers/dri/i965/brw_program.c
@@ -111,13 +111,18 @@ static void brwProgramStringNotify( GLcontext *ctx,
 				    struct gl_program *prog )
 {
    if (target == GL_FRAGMENT_PROGRAM_ARB) {
+      struct gl_fragment_program *fprog = (struct gl_fragment_program *) prog;
       struct brw_context *brw = brw_context(ctx);
       struct brw_fragment_program *p = (struct brw_fragment_program *)prog;
       struct brw_fragment_program *fp = (struct brw_fragment_program *)brw->fragment_program;
+      if (fprog->FogOption) {
+         _mesa_append_fog_code(ctx, fprog);
+         fprog->FogOption = GL_NONE;
+      }
+
       if (p == fp)
 	 brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM;
       p->id = brw->program_id++;      
-      p->param_state = p->program.Base.Parameters->StateFlags;
    }
    else if (target == GL_VERTEX_PROGRAM_ARB) {
       struct brw_context *brw = brw_context(ctx);
@@ -129,7 +134,6 @@ static void brwProgramStringNotify( GLcontext *ctx,
 	 _mesa_insert_mvp_code(ctx, &p->program);
       }
       p->id = brw->program_id++;      
-      p->param_state = p->program.Base.Parameters->StateFlags;
 
       /* Also tell tnl about it:
        */
diff --git a/src/mesa/drivers/dri/i965/brw_sf.c b/src/mesa/drivers/dri/i965/brw_sf.c
index 9dce6cd8e6..8c1711538a 100644
--- a/src/mesa/drivers/dri/i965/brw_sf.c
+++ b/src/mesa/drivers/dri/i965/brw_sf.c
@@ -46,6 +46,7 @@
 static void compile_sf_prog( struct brw_context *brw,
 			     struct brw_sf_prog_key *key )
 {
+   GLcontext *ctx = &brw->intel.ctx;
    struct brw_sf_compile c;
    const GLuint *program;
    GLuint program_size;
@@ -73,10 +74,12 @@ static void compile_sf_prog( struct brw_context *brw,
 	 c.attr_to_idx[i] = idx;
 	 c.idx_to_attr[idx] = i;
 	 if (i >= VERT_RESULT_TEX0 && i <= VERT_RESULT_TEX7) {
-		 c.point_attrs[i].CoordReplace = 
-			brw->attribs.Point->CoordReplace[i - VERT_RESULT_TEX0];
-	 } else
-		 c.point_attrs[i].CoordReplace = GL_FALSE;
+            c.point_attrs[i].CoordReplace = 
+               ctx->Point.CoordReplace[i - VERT_RESULT_TEX0];
+	 }
+         else {
+            c.point_attrs[i].CoordReplace = GL_FALSE;
+         }
 	 idx++;
       }
    
@@ -106,7 +109,6 @@ static void compile_sf_prog( struct brw_context *brw,
       assert(0);
       return;
    }
-	 
 
    /* get the program
     */
@@ -127,6 +129,7 @@ static void compile_sf_prog( struct brw_context *brw,
  */
 static void upload_sf_prog(struct brw_context *brw)
 {
+   GLcontext *ctx = &brw->intel.ctx;
    struct brw_sf_prog_key key;
 
    memset(&key, 0, sizeof(key));
@@ -157,15 +160,15 @@ static void upload_sf_prog(struct brw_context *brw)
       break;
    }
 
-   key.do_point_sprite = brw->attribs.Point->PointSprite;
-   key.SpriteOrigin = brw->attribs.Point->SpriteOrigin;
+   key.do_point_sprite = ctx->Point.PointSprite;
+   key.SpriteOrigin = ctx->Point.SpriteOrigin;
    /* _NEW_LIGHT */
-   key.do_flat_shading = (brw->attribs.Light->ShadeModel == GL_FLAT);
-   key.do_twoside_color = (brw->attribs.Light->Enabled && brw->attribs.Light->Model.TwoSide);
+   key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT);
+   key.do_twoside_color = (ctx->Light.Enabled && ctx->Light.Model.TwoSide);
 
    /* _NEW_POLYGON */
    if (key.do_twoside_color)
-      key.frontface_ccw = (brw->attribs.Polygon->FrontFace == GL_CCW);
+      key.frontface_ccw = (ctx->Polygon.FrontFace == GL_CCW);
 
    dri_bo_unreference(brw->sf.prog_bo);
    brw->sf.prog_bo = brw_search_cache(&brw->cache, BRW_SF_PROG,
diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c
index 506126fcfb..e96d5354b3 100644
--- a/src/mesa/drivers/dri/i965/brw_sf_state.c
+++ b/src/mesa/drivers/dri/i965/brw_sf_state.c
@@ -42,45 +42,30 @@ static void upload_sf_vp(struct brw_context *brw)
    GLcontext *ctx = &brw->intel.ctx;
    const GLfloat depth_scale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
    struct brw_sf_viewport sfv;
-   struct intel_renderbuffer *irb =
-      intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
    GLfloat y_scale, y_bias;
+   const GLboolean render_to_fbo = (ctx->DrawBuffer->Name != 0);
 
    memset(&sfv, 0, sizeof(sfv));
 
-   if (ctx->DrawBuffer->Name) {
-      /* User-created FBO */
-      if (irb && !irb->RenderToTexture) {
-	 y_scale = -1.0;
-	 y_bias = ctx->DrawBuffer->Height;
-      } else {
-	 y_scale = 1.0;
-	 y_bias = 0;
-      }
-   } else {
+   if (render_to_fbo) {
+      y_scale = 1.0;
+      y_bias = 0;
+   }
+   else {
       y_scale = -1.0;
       y_bias = ctx->DrawBuffer->Height;
    }
 
-   /* _NEW_VIEWPORT, BRW_NEW_METAOPS */
-
-   if (!brw->metaops.active) {
-      const GLfloat *v = ctx->Viewport._WindowMap.m;
-
-      sfv.viewport.m00 = v[MAT_SX];
-      sfv.viewport.m11 = v[MAT_SY] * y_scale;
-      sfv.viewport.m22 = v[MAT_SZ] * depth_scale;
-      sfv.viewport.m30 = v[MAT_TX];
-      sfv.viewport.m31 = v[MAT_TY] * y_scale + y_bias;
-      sfv.viewport.m32 = v[MAT_TZ] * depth_scale;
-   } else {
-      sfv.viewport.m00 =   1;
-      sfv.viewport.m11 = - 1;
-      sfv.viewport.m22 =   1;
-      sfv.viewport.m30 =   0;
-      sfv.viewport.m31 =   ctx->DrawBuffer->Height;
-      sfv.viewport.m32 =   0;
-   }
+   /* _NEW_VIEWPORT */
+
+   const GLfloat *v = ctx->Viewport._WindowMap.m;
+
+   sfv.viewport.m00 = v[MAT_SX];
+   sfv.viewport.m11 = v[MAT_SY] * y_scale;
+   sfv.viewport.m22 = v[MAT_SZ] * depth_scale;
+   sfv.viewport.m30 = v[MAT_TX];
+   sfv.viewport.m31 = v[MAT_TY] * y_scale + y_bias;
+   sfv.viewport.m32 = v[MAT_TZ] * depth_scale;
 
    /* _NEW_SCISSOR */
 
@@ -91,10 +76,20 @@ static void upload_sf_vp(struct brw_context *brw)
     * Note that the hardware's coordinates are inclusive, while Mesa's min is
     * inclusive but max is exclusive.
     */
-   sfv.scissor.xmin = ctx->DrawBuffer->_Xmin;
-   sfv.scissor.xmax = ctx->DrawBuffer->_Xmax - 1;
-   sfv.scissor.ymin = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax;
-   sfv.scissor.ymax = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin - 1;
+   if (render_to_fbo) {
+      /* texmemory: Y=0=bottom */
+      sfv.scissor.xmin = ctx->DrawBuffer->_Xmin;
+      sfv.scissor.xmax = ctx->DrawBuffer->_Xmax - 1;
+      sfv.scissor.ymin = ctx->DrawBuffer->_Ymin;
+      sfv.scissor.ymax = ctx->DrawBuffer->_Ymax - 1;
+   }
+   else {
+      /* memory: Y=0=top */
+      sfv.scissor.xmin = ctx->DrawBuffer->_Xmin;
+      sfv.scissor.xmax = ctx->DrawBuffer->_Xmax - 1;
+      sfv.scissor.ymin = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax;
+      sfv.scissor.ymax = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin - 1;
+   }
 
    dri_bo_unreference(brw->sf.vp_bo);
    brw->sf.vp_bo = brw_cache_data( &brw->cache, BRW_SF_VP, &sfv, NULL, 0 );
@@ -104,7 +99,7 @@ const struct brw_tracked_state brw_sf_vp = {
    .dirty = {
       .mesa  = (_NEW_VIEWPORT | 
 		_NEW_SCISSOR),
-      .brw   = BRW_NEW_METAOPS,
+      .brw   = 0,
       .cache = 0
    },
    .prepare = upload_sf_vp
@@ -117,7 +112,11 @@ struct brw_sf_unit_key {
    unsigned int nr_urb_entries, urb_size, sfsize;
 
    GLenum front_face, cull_face;
-   GLboolean scissor, line_smooth, point_sprite, point_attenuated;
+   unsigned scissor:1;
+   unsigned line_smooth:1;
+   unsigned point_sprite:1;
+   unsigned point_attenuated:1;
+   unsigned render_to_fbo:1;
    float line_width;
    float point_size;
 };
@@ -125,6 +124,7 @@ struct brw_sf_unit_key {
 static void
 sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key)
 {
+   GLcontext *ctx = &brw->intel.ctx;
    memset(key, 0, sizeof(*key));
 
    /* CACHE_NEW_SF_PROG */
@@ -136,20 +136,22 @@ sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key)
    key->urb_size = brw->urb.vsize;
    key->sfsize = brw->urb.sfsize;
 
-   key->scissor = brw->attribs.Scissor->Enabled;
-   key->front_face = brw->attribs.Polygon->FrontFace;
+   key->scissor = ctx->Scissor.Enabled;
+   key->front_face = ctx->Polygon.FrontFace;
 
-   if (brw->attribs.Polygon->CullFlag)
-      key->cull_face = brw->attribs.Polygon->CullFaceMode;
+   if (ctx->Polygon.CullFlag)
+      key->cull_face = ctx->Polygon.CullFaceMode;
    else
       key->cull_face = GL_NONE;
 
-   key->line_width = brw->attribs.Line->Width;
-   key->line_smooth = brw->attribs.Line->SmoothFlag;
+   key->line_width = ctx->Line.Width;
+   key->line_smooth = ctx->Line.SmoothFlag;
+
+   key->point_sprite = ctx->Point.PointSprite;
+   key->point_size = ctx->Point.Size;
+   key->point_attenuated = ctx->Point._Attenuated;
 
-   key->point_sprite = brw->attribs.Point->PointSprite;
-   key->point_size = brw->attribs.Point->Size;
-   key->point_attenuated = brw->attribs.Point->_Attenuated;
+   key->render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0;
 }
 
 static dri_bo *
@@ -196,6 +198,11 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
    else
       sf.sf5.front_winding = BRW_FRONTWINDING_CW;
 
+   /* The viewport is inverted for rendering to a FBO, and that inverts
+    * polygon front/back orientation.
+    */
+   sf.sf5.front_winding ^= key->render_to_fbo;
+
    switch (key->cull_face) {
    case GL_FRONT:
       sf.sf6.cull_mode = BRW_CULLMODE_FRONT;
@@ -229,7 +236,7 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
    /* XXX clamp max depends on AA vs. non-AA */
 
    sf.sf7.sprite_point = key->point_sprite;
-   sf.sf7.point_size = CLAMP(nearbyint(key->point_size), 1, 255) * (1<<3);
+   sf.sf7.point_size = CLAMP(rint(key->point_size), 1, 255) * (1<<3);
    sf.sf7.use_point_size_state = !key->point_attenuated;
    sf.sf7.aa_line_distance_mode = 0;
 
@@ -294,8 +301,7 @@ const struct brw_tracked_state brw_sf_unit = {
 		_NEW_LINE | 
 		_NEW_POINT | 
 		_NEW_SCISSOR),
-      .brw   = (BRW_NEW_URB_FENCE |
-		BRW_NEW_METAOPS),
+      .brw   = BRW_NEW_URB_FENCE,
       .cache = (CACHE_NEW_SF_VP |
 		CACHE_NEW_SF_PROG)
    },
diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h
index bb22c03eeb..df839c5b30 100644
--- a/src/mesa/drivers/dri/i965/brw_state.h
+++ b/src/mesa/drivers/dri/i965/brw_state.h
@@ -83,7 +83,6 @@ const struct brw_tracked_state brw_wm_unit;
 
 const struct brw_tracked_state brw_psp_urb_cbs;
 
-const struct brw_tracked_state brw_active_vertprog;
 const struct brw_tracked_state brw_pipe_control;
 
 const struct brw_tracked_state brw_clear_surface_cache;
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index 7a642bd2a8..5de1450e61 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -181,11 +181,123 @@ brw_clear_validated_bos(struct brw_context *brw)
    brw->state.validated_bo_count = 0;
 }
 
+struct dirty_bit_map {
+   uint32_t bit;
+   char *name;
+   uint32_t count;
+};
+
+#define DEFINE_BIT(name) {name, #name, 0}
+
+static struct dirty_bit_map mesa_bits[] = {
+   DEFINE_BIT(_NEW_MODELVIEW),
+   DEFINE_BIT(_NEW_PROJECTION),
+   DEFINE_BIT(_NEW_TEXTURE_MATRIX),
+   DEFINE_BIT(_NEW_COLOR_MATRIX),
+   DEFINE_BIT(_NEW_ACCUM),
+   DEFINE_BIT(_NEW_COLOR),
+   DEFINE_BIT(_NEW_DEPTH),
+   DEFINE_BIT(_NEW_EVAL),
+   DEFINE_BIT(_NEW_FOG),
+   DEFINE_BIT(_NEW_HINT),
+   DEFINE_BIT(_NEW_LIGHT),
+   DEFINE_BIT(_NEW_LINE),
+   DEFINE_BIT(_NEW_PIXEL),
+   DEFINE_BIT(_NEW_POINT),
+   DEFINE_BIT(_NEW_POLYGON),
+   DEFINE_BIT(_NEW_POLYGONSTIPPLE),
+   DEFINE_BIT(_NEW_SCISSOR),
+   DEFINE_BIT(_NEW_STENCIL),
+   DEFINE_BIT(_NEW_TEXTURE),
+   DEFINE_BIT(_NEW_TRANSFORM),
+   DEFINE_BIT(_NEW_VIEWPORT),
+   DEFINE_BIT(_NEW_PACKUNPACK),
+   DEFINE_BIT(_NEW_ARRAY),
+   DEFINE_BIT(_NEW_RENDERMODE),
+   DEFINE_BIT(_NEW_BUFFERS),
+   DEFINE_BIT(_NEW_MULTISAMPLE),
+   DEFINE_BIT(_NEW_TRACK_MATRIX),
+   DEFINE_BIT(_NEW_PROGRAM),
+   {0, 0, 0}
+};
+
+static struct dirty_bit_map brw_bits[] = {
+   DEFINE_BIT(BRW_NEW_URB_FENCE),
+   DEFINE_BIT(BRW_NEW_FRAGMENT_PROGRAM),
+   DEFINE_BIT(BRW_NEW_VERTEX_PROGRAM),
+   DEFINE_BIT(BRW_NEW_INPUT_DIMENSIONS),
+   DEFINE_BIT(BRW_NEW_CURBE_OFFSETS),
+   DEFINE_BIT(BRW_NEW_REDUCED_PRIMITIVE),
+   DEFINE_BIT(BRW_NEW_PRIMITIVE),
+   DEFINE_BIT(BRW_NEW_CONTEXT),
+   DEFINE_BIT(BRW_NEW_WM_INPUT_DIMENSIONS),
+   DEFINE_BIT(BRW_NEW_INPUT_VARYING),
+   DEFINE_BIT(BRW_NEW_PSP),
+   DEFINE_BIT(BRW_NEW_FENCE),
+   DEFINE_BIT(BRW_NEW_INDICES),
+   DEFINE_BIT(BRW_NEW_VERTICES),
+   DEFINE_BIT(BRW_NEW_BATCH),
+   DEFINE_BIT(BRW_NEW_DEPTH_BUFFER),
+   {0, 0, 0}
+};
+
+static struct dirty_bit_map cache_bits[] = {
+   DEFINE_BIT(CACHE_NEW_CC_VP),
+   DEFINE_BIT(CACHE_NEW_CC_UNIT),
+   DEFINE_BIT(CACHE_NEW_WM_PROG),
+   DEFINE_BIT(CACHE_NEW_SAMPLER_DEFAULT_COLOR),
+   DEFINE_BIT(CACHE_NEW_SAMPLER),
+   DEFINE_BIT(CACHE_NEW_WM_UNIT),
+   DEFINE_BIT(CACHE_NEW_SF_PROG),
+   DEFINE_BIT(CACHE_NEW_SF_VP),
+   DEFINE_BIT(CACHE_NEW_SF_UNIT),
+   DEFINE_BIT(CACHE_NEW_VS_UNIT),
+   DEFINE_BIT(CACHE_NEW_VS_PROG),
+   DEFINE_BIT(CACHE_NEW_GS_UNIT),
+   DEFINE_BIT(CACHE_NEW_GS_PROG),
+   DEFINE_BIT(CACHE_NEW_CLIP_VP),
+   DEFINE_BIT(CACHE_NEW_CLIP_UNIT),
+   DEFINE_BIT(CACHE_NEW_CLIP_PROG),
+   DEFINE_BIT(CACHE_NEW_SURFACE),
+   DEFINE_BIT(CACHE_NEW_SURF_BIND),
+   {0, 0, 0}
+};
+
+
+static void
+brw_update_dirty_count(struct dirty_bit_map *bit_map, int32_t bits)
+{
+   int i;
+
+   for (i = 0; i < 32; i++) {
+      if (bit_map[i].bit == 0)
+	 return;
+
+      if (bit_map[i].bit & bits)
+	 bit_map[i].count++;
+   }
+}
+
+static void
+brw_print_dirty_count(struct dirty_bit_map *bit_map, int32_t bits)
+{
+   int i;
+
+   for (i = 0; i < 32; i++) {
+      if (bit_map[i].bit == 0)
+	 return;
+
+      fprintf(stderr, "0x%08x: %12d (%s)\n",
+	      bit_map[i].bit, bit_map[i].count, bit_map[i].name);
+   }
+}
+
 /***********************************************************************
  * Emit all state:
  */
 void brw_validate_state( struct brw_context *brw )
 {
+   GLcontext *ctx = &brw->intel.ctx;
    struct intel_context *intel = &brw->intel;
    struct brw_state_flags *state = &brw->state.dirty;
    GLuint i;
@@ -202,17 +314,13 @@ void brw_validate_state( struct brw_context *brw )
       state->brw |= ~0;
    }
 
-   /* texenv program needs to notify us somehow when this happens: 
-    * Some confusion about which state flag should represent this change.
-    */
-   if (brw->fragment_program != brw->attribs.FragmentProgram->_Current) {
-      brw->fragment_program = brw->attribs.FragmentProgram->_Current;
-      brw->state.dirty.mesa |= _NEW_PROGRAM;
+   if (brw->fragment_program != ctx->FragmentProgram._Current) {
+      brw->fragment_program = ctx->FragmentProgram._Current;
       brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM;
    }
 
-   if (brw->vertex_program != brw->attribs.VertexProgram->_Current) {
-      brw->vertex_program = brw->attribs.VertexProgram->_Current;
+   if (brw->vertex_program != ctx->VertexProgram._Current) {
+      brw->vertex_program = ctx->VertexProgram._Current;
       brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM;
    }
 
@@ -246,6 +354,7 @@ void brw_upload_state(struct brw_context *brw)
 {
    struct brw_state_flags *state = &brw->state.dirty;
    int i;
+   static int dirty_count = 0;
 
    brw_clear_validated_bos(brw);
 
@@ -301,6 +410,18 @@ void brw_upload_state(struct brw_context *brw)
       }
    }
 
+   if (INTEL_DEBUG & DEBUG_STATE) {
+      brw_update_dirty_count(mesa_bits, state->mesa);
+      brw_update_dirty_count(brw_bits, state->brw);
+      brw_update_dirty_count(cache_bits, state->cache);
+      if (dirty_count++ % 1000 == 0) {
+	 brw_print_dirty_count(mesa_bits, state->mesa);
+	 brw_print_dirty_count(brw_bits, state->brw);
+	 brw_print_dirty_count(cache_bits, state->cache);
+	 fprintf(stderr, "\n");
+      }
+   }
+
    if (!brw->intel.Fallback)
       memset(state, 0, sizeof(*state));
 }
diff --git a/src/mesa/drivers/dri/i965/brw_structs.h b/src/mesa/drivers/dri/i965/brw_structs.h
index 4e577d0f6a..d97ff27f0a 100644
--- a/src/mesa/drivers/dri/i965/brw_structs.h
+++ b/src/mesa/drivers/dri/i965/brw_structs.h
@@ -1031,10 +1031,10 @@ struct brw_surface_state
       GLuint writedisable_green:1; 
       GLuint writedisable_red:1; 
       GLuint writedisable_alpha:1; 
-      GLuint surface_format:9; 
+      GLuint surface_format:9;     /**< BRW_SURFACEFORMAT_x */
       GLuint data_return_format:1; 
       GLuint pad0:1;
-      GLuint surface_type:3; 
+      GLuint surface_type:3;       /**< BRW_SURFACE_1D/2D/3D/CUBE */
    } ss0;
    
    struct {
diff --git a/src/mesa/drivers/dri/i965/brw_tex.c b/src/mesa/drivers/dri/i965/brw_tex.c
index 0bb6f176a0..ef99e9c1ae 100644
--- a/src/mesa/drivers/dri/i965/brw_tex.c
+++ b/src/mesa/drivers/dri/i965/brw_tex.c
@@ -86,11 +86,12 @@ void brw_FrameBufferTexDestroy( struct brw_context *brw )
  */
 void brw_validate_textures( struct brw_context *brw )
 {
+   GLcontext *ctx = &brw->intel.ctx;
    struct intel_context *intel = &brw->intel;
    int i;
 
    for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
-      struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[i];
+      struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
 
       if (texUnit->_ReallyEnabled) {
 	 intel_finalize_mipmap_tree(intel, i);
diff --git a/src/mesa/drivers/dri/i965/brw_vs.c b/src/mesa/drivers/dri/i965/brw_vs.c
index 1db7ceebcf..e3111c6680 100644
--- a/src/mesa/drivers/dri/i965/brw_vs.c
+++ b/src/mesa/drivers/dri/i965/brw_vs.c
@@ -85,6 +85,7 @@ static void do_vs_prog( struct brw_context *brw,
 
 static void brw_upload_vs_prog(struct brw_context *brw)
 {
+   GLcontext *ctx = &brw->intel.ctx;
    struct brw_vs_prog_key key;
    struct brw_vertex_program *vp = 
       (struct brw_vertex_program *)brw->vertex_program;
@@ -97,14 +98,9 @@ static void brw_upload_vs_prog(struct brw_context *brw)
     * the inputs it asks for, whether they are varying or not.
     */
    key.program_string_id = vp->id;
-   key.nr_userclip = brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled);
-   key.copy_edgeflag = (brw->attribs.Polygon->FrontMode != GL_FILL ||
-			brw->attribs.Polygon->BackMode != GL_FILL);
-
-   /* BRW_NEW_METAOPS
-    */
-   if (brw->metaops.active)
-      key.know_w_is_one = 1;
+   key.nr_userclip = brw_count_bits(ctx->Transform.ClipPlanesEnabled);
+   key.copy_edgeflag = (ctx->Polygon.FrontMode != GL_FILL ||
+			ctx->Polygon.BackMode != GL_FILL);
 
    /* Make an early check for the key.
     */
@@ -123,7 +119,7 @@ static void brw_upload_vs_prog(struct brw_context *brw)
 const struct brw_tracked_state brw_vs_prog = {
    .dirty = {
       .mesa  = _NEW_TRANSFORM | _NEW_POLYGON,
-      .brw   = BRW_NEW_VERTEX_PROGRAM | BRW_NEW_METAOPS,
+      .brw   = BRW_NEW_VERTEX_PROGRAM,
       .cache = 0
    },
    .prepare = brw_upload_vs_prog
diff --git a/src/mesa/drivers/dri/i965/brw_vs.h b/src/mesa/drivers/dri/i965/brw_vs.h
index 22388ec99d..99d0e93722 100644
--- a/src/mesa/drivers/dri/i965/brw_vs.h
+++ b/src/mesa/drivers/dri/i965/brw_vs.h
@@ -43,7 +43,6 @@ struct brw_vs_prog_key {
    GLuint program_string_id;
    GLuint nr_userclip:4;
    GLuint copy_edgeflag:1;
-   GLuint know_w_is_one:1;
    GLuint pad:26;
 };
 
diff --git a/src/mesa/drivers/dri/i965/brw_vs_constval.c b/src/mesa/drivers/dri/i965/brw_vs_constval.c
index 6fbac02de6..9977677fd7 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_constval.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_constval.c
@@ -168,6 +168,7 @@ static GLuint get_input_size(struct brw_context *brw,
  */
 static void calc_wm_input_sizes( struct brw_context *brw )
 {
+   GLcontext *ctx = &brw->intel.ctx;
    /* BRW_NEW_VERTEX_PROGRAM */
    struct brw_vertex_program *vp = 
       (struct brw_vertex_program *)brw->vertex_program;
@@ -179,7 +180,7 @@ static void calc_wm_input_sizes( struct brw_context *brw )
    memset(&t, 0, sizeof(t));
 
    /* _NEW_LIGHT */
-   if (brw->attribs.Light->Model.TwoSide)
+   if (ctx->Light.Model.TwoSide)
       t.twoside = 1;
 
    for (i = 0; i < VERT_ATTRIB_MAX; i++) 
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index 25b4ee85cb..24b7dc30fe 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -73,8 +73,6 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 
    c->prog_data.curb_read_length = reg - 1;
 
-
-
    /* Allocate input regs:  
     */
    c->nr_inputs = 0;
@@ -84,8 +82,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
 	 c->regs[PROGRAM_INPUT][i] = brw_vec8_grf(reg, 0);
 	 reg++;
       }
-   }     
-
+   }
 
    /* Allocate outputs: TODO: could organize the non-position outputs
     * to go straight into message regs.
@@ -196,6 +193,7 @@ static void unalias1( struct brw_vs_compile *c,
       struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask);
       func(c, tmp, arg0);
       brw_MOV(p, dst, tmp);
+      release_tmp(c, tmp);
    }
    else {
       func(c, dst, arg0);
@@ -217,12 +215,38 @@ static void unalias2( struct brw_vs_compile *c,
       struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask);
       func(c, tmp, arg0, arg1);
       brw_MOV(p, dst, tmp);
+      release_tmp(c, tmp);
    }
    else {
       func(c, dst, arg0, arg1);
    }
 }
 
+static void unalias3( struct brw_vs_compile *c,
+		      struct brw_reg dst,
+		      struct brw_reg arg0,
+		      struct brw_reg arg1,
+		      struct brw_reg arg2,
+		      void (*func)( struct brw_vs_compile *,
+				    struct brw_reg,
+				    struct brw_reg,
+				    struct brw_reg,
+				    struct brw_reg ))
+{
+   if ((dst.file == arg0.file && dst.nr == arg0.nr) ||
+       (dst.file == arg1.file && dst.nr == arg1.nr) ||
+       (dst.file == arg2.file && dst.nr == arg2.nr)) {
+      struct brw_compile *p = &c->func;
+      struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask);
+      func(c, tmp, arg0, arg1, arg2);
+      brw_MOV(p, dst, tmp);
+      release_tmp(c, tmp);
+   }
+   else {
+      func(c, dst, arg0, arg1, arg2);
+   }
+}
+
 static void emit_sop( struct brw_compile *p,
                       struct brw_reg dst,
                       struct brw_reg arg0,
@@ -339,6 +363,7 @@ static void emit_math1( struct brw_vs_compile *c,
    }
 }
 
+
 static void emit_math2( struct brw_vs_compile *c, 
 			GLuint function,
 			struct brw_reg dst,
@@ -370,7 +395,6 @@ static void emit_math2( struct brw_vs_compile *c,
       release_tmp(c, tmp);
    }
 }
-		     
 
 
 static void emit_exp_noalias( struct brw_vs_compile *c,
@@ -420,7 +444,7 @@ static void emit_exp_noalias( struct brw_vs_compile *c,
 		 BRW_MATH_FUNCTION_EXP, 
 		 brw_writemask(dst, WRITEMASK_Z),
 		 brw_swizzle1(arg0, 0), 
-		 BRW_MATH_PRECISION_PARTIAL);
+		 BRW_MATH_PRECISION_FULL);
    }  
 
    if (dst.dw1.bits.writemask & WRITEMASK_W) {
@@ -521,8 +545,6 @@ static void emit_log_noalias( struct brw_vs_compile *c,
 }
 
 
-      
-
 /* Need to unalias - consider swizzles:   r0 = DST r0.xxxx r1
  */
 static void emit_dst_noalias( struct brw_vs_compile *c, 
@@ -544,6 +566,7 @@ static void emit_dst_noalias( struct brw_vs_compile *c,
       brw_MOV(p, brw_writemask(dst, WRITEMASK_W), arg1);
 }
 
+
 static void emit_xpd( struct brw_compile *p,
 		      struct brw_reg dst,
 		      struct brw_reg t,
@@ -554,7 +577,6 @@ static void emit_xpd( struct brw_compile *p,
 }
 
 
-
 static void emit_lit_noalias( struct brw_vs_compile *c, 
 			      struct brw_reg dst,
 			      struct brw_reg arg0 )
@@ -595,8 +617,42 @@ static void emit_lit_noalias( struct brw_vs_compile *c,
    brw_ENDIF(p, if_insn);
 }
 
+static void emit_lrp_noalias(struct brw_vs_compile *c,
+			     struct brw_reg dst,
+			     struct brw_reg arg0,
+			     struct brw_reg arg1,
+			     struct brw_reg arg2)
+{
+   struct brw_compile *p = &c->func;
+
+   brw_ADD(p, dst, negate(arg0), brw_imm_f(1.0));
+   brw_MUL(p, brw_null_reg(), dst, arg2);
+   brw_MAC(p, dst, arg0, arg1);
+}
+
+/** 3 or 4-component vector normalization */
+static void emit_nrm( struct brw_vs_compile *c, 
+                      struct brw_reg dst,
+                      struct brw_reg arg0,
+                      int num_comps)
+{
+   struct brw_compile *p = &c->func;
+   struct brw_reg tmp = get_tmp(c);
+
+   /* tmp = dot(arg0, arg0) */
+   if (num_comps == 3)
+      brw_DP3(p, tmp, arg0, arg0);
+   else
+      brw_DP4(p, tmp, arg0, arg0);
+
+   /* tmp = 1 / sqrt(tmp) */
+   emit_math1(c, BRW_MATH_FUNCTION_RSQ, tmp, tmp, BRW_MATH_PRECISION_FULL);
 
+   /* dst = arg0 * tmp */
+   brw_MUL(p, dst, arg0, tmp);
 
+   release_tmp(c, tmp);
+}
 
 
 /* TODO: relative addressing!
@@ -634,7 +690,6 @@ static struct brw_reg get_reg( struct brw_vs_compile *c,
 }
 
 
-
 static struct brw_reg deref( struct brw_vs_compile *c,
 			     struct brw_reg arg,
 			     GLint offset)
@@ -728,8 +783,6 @@ static struct brw_reg get_dst( struct brw_vs_compile *c,
 }
 
 
-
-
 static void emit_swz( struct brw_vs_compile *c, 
 		      struct brw_reg dst,
 		      struct prog_src_register src )
@@ -801,8 +854,8 @@ static void emit_swz( struct brw_vs_compile *c,
 }
 
 
-
-/* Post-vertex-program processing.  Send the results to the URB.
+/**
+ * Post-vertex-program processing.  Send the results to the URB.
  */
 static void emit_vertex_write( struct brw_vs_compile *c)
 {
@@ -817,23 +870,16 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 	      get_reg(c, PROGRAM_INPUT, VERT_ATTRIB_EDGEFLAG));
    }
 
-
    /* Build ndc coords */
-   if (!c->key.know_w_is_one) {
-      ndc = get_tmp(c);
-      emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL);
-      brw_MUL(p, brw_writemask(ndc, WRITEMASK_XYZ), pos, ndc);
-   }
-   else {
-      ndc = pos;
-   }
+   ndc = get_tmp(c);
+   emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL);
+   brw_MUL(p, brw_writemask(ndc, WRITEMASK_XYZ), pos, ndc);
 
    /* Update the header for point size, user clipping flags, and -ve rhw
     * workaround.
     */
    if ((c->prog_data.outputs_written & (1<<VERT_RESULT_PSIZ)) ||
-       c->key.nr_userclip ||
-       (!BRW_IS_G4X(p->brw) && !c->key.know_w_is_one))
+       c->key.nr_userclip || !BRW_IS_G4X(p->brw))
    {
       struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD);
       GLuint i;
@@ -848,7 +894,6 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 	 brw_AND(p, brw_writemask(header1, WRITEMASK_W), header1, brw_imm_ud(0x7ff<<8));
       }
 
-
       for (i = 0; i < c->key.nr_userclip; i++) {
 	 brw_set_conditionalmod(p, BRW_CONDITIONAL_L);
 	 brw_DP4(p, brw_null_reg(), pos, c->userplane[i]);
@@ -856,7 +901,6 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 	 brw_set_predicate_control(p, BRW_PREDICATE_NONE);
       }
 
-
       /* i965 clipping workaround: 
        * 1) Test for -ve rhw
        * 2) If set, 
@@ -866,7 +910,7 @@ static void emit_vertex_write( struct brw_vs_compile *c)
        * Later, clipping will detect ucp[6] and ensure the primitive is
        * clipped against all fixed planes.
        */
-      if (!BRW_IS_G4X(p->brw) && !c->key.know_w_is_one) {
+      if (!BRW_IS_G4X(p->brw)) {
 	 brw_CMP(p,
 		 vec8(brw_null_reg()),
 		 BRW_CONDITIONAL_L,
@@ -888,14 +932,12 @@ static void emit_vertex_write( struct brw_vs_compile *c)
       brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), brw_imm_ud(0));
    }
 
-
    /* Emit the (interleaved) headers for the two vertices - an 8-reg
     * of zeros followed by two sets of NDC coordinates:
     */
    brw_set_access_mode(p, BRW_ALIGN_1);
    brw_MOV(p, offset(m0, 2), ndc);
    brw_MOV(p, offset(m0, 3), pos);
-   
 
    brw_urb_WRITE(p, 
 		 brw_null_reg(), /* dest */
@@ -909,9 +951,9 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 		 1, 		/* writes complete */
 		 0, 		/* urb destination offset */
 		 BRW_URB_SWIZZLE_INTERLEAVE);
-
 }
 
+
 static void 
 post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst )
 {
@@ -958,7 +1000,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
    GLuint file;
 
    if (INTEL_DEBUG & DEBUG_VS) {
-      _mesa_printf("\n\n\nvs-emit:\n");
+      _mesa_printf("vs-emit:\n");
       _mesa_print_program(&c->vp->program.Base); 
       _mesa_printf("\n");
    }
@@ -1016,6 +1058,11 @@ void brw_vs_emit(struct brw_vs_compile *c )
       else
 	  dst = get_dst(c, inst->DstReg);
 
+      if (inst->SaturateMode != SATURATE_OFF) {
+	 _mesa_problem(NULL, "Unsupported saturate %d in vertex shader",
+                       inst->SaturateMode);
+      }
+
       switch (inst->Opcode) {
       case OPCODE_ABS:
 	 brw_MOV(p, dst, brw_abs(args[0]));
@@ -1035,6 +1082,12 @@ void brw_vs_emit(struct brw_vs_compile *c )
       case OPCODE_DPH:
 	 brw_DPH(p, dst, args[0], args[1]);
 	 break;
+      case OPCODE_NRM3:
+	 emit_nrm(c, dst, args[0], 3);
+	 break;
+      case OPCODE_NRM4:
+	 emit_nrm(c, dst, args[0], 4);
+	 break;
       case OPCODE_DST:
 	 unalias2(c, dst, args[0], args[1], emit_dst_noalias); 
 	 break;
@@ -1062,6 +1115,9 @@ void brw_vs_emit(struct brw_vs_compile *c )
       case OPCODE_LIT:
 	 unalias1(c, dst, args[0], emit_lit_noalias);
 	 break;
+      case OPCODE_LRP:
+	 unalias3(c, dst, args[0], args[1], args[2], emit_lrp_noalias);
+	 break;
       case OPCODE_MAD:
 	 brw_MOV(p, brw_acc_reg(), args[2]);
 	 brw_MAC(p, dst, args[0], args[1]);
@@ -1102,7 +1158,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
 	 break;
       case OPCODE_SGT:
          emit_sgt(p, dst, args[0], args[1]);
-        break;
+         break;
       case OPCODE_SLT:
 	 emit_slt(p, dst, args[0], args[1]);
 	 break;
@@ -1118,6 +1174,10 @@ void brw_vs_emit(struct brw_vs_compile *c )
 	  */
 	 emit_swz(c, dst, inst->SrcReg[0] );
 	 break;
+      case OPCODE_TRUNC:
+         /* round toward zero */
+	 brw_RNDZ(p, dst, args[0]);
+	 break;
       case OPCODE_XPD:
 	 emit_xpd(p, dst, args[0], args[1]);
 	 break;
@@ -1136,7 +1196,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
          brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
          brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
          brw_set_predicate_control_flag_value(p, 0xff);
-        break;
+         break;
       case OPCODE_CAL:
 	 brw_set_access_mode(p, BRW_ALIGN_1);
 	 brw_ADD(p, deref_1d(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16));
@@ -1145,7 +1205,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
 			 get_addr_reg(stack_index), brw_imm_d(4));
 	 inst->Data = &p->store[p->nr_insn];
 	 brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
-        break;
+         break;
       case OPCODE_RET:
 	 brw_ADD(p, get_addr_reg(stack_index),
 			 get_addr_reg(stack_index), brw_imm_d(-4));
@@ -1154,17 +1214,17 @@ void brw_vs_emit(struct brw_vs_compile *c )
 	 brw_set_access_mode(p, BRW_ALIGN_16);
       case OPCODE_END:	
          brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
-        break;
+         break;
       case OPCODE_PRINT:
       case OPCODE_BGNSUB:
       case OPCODE_ENDSUB:
+         /* no-op instructions */
 	 break;
       default:
-	 _mesa_printf("Unsupported opcode %i (%s) in vertex shader\n",
-		      inst->Opcode, inst->Opcode < MAX_OPCODE ?
+	 _mesa_problem(NULL, "Unsupported opcode %i (%s) in vertex shader",
+                       inst->Opcode, inst->Opcode < MAX_OPCODE ?
 				    _mesa_opcode_string(inst->Opcode) :
 				    "unknown");
-	 break;
       }
 
       if ((inst->DstReg.File == PROGRAM_OUTPUT)
diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c
index 942581696d..1a63766ea1 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_state.c
@@ -49,6 +49,8 @@ struct brw_vs_unit_key {
 static void
 vs_unit_populate_key(struct brw_context *brw, struct brw_vs_unit_key *key)
 {
+   GLcontext *ctx = &brw->intel.ctx;
+
    memset(key, 0, sizeof(*key));
 
    /* CACHE_NEW_VS_PROG */
@@ -61,7 +63,7 @@ vs_unit_populate_key(struct brw_context *brw, struct brw_vs_unit_key *key)
    key->urb_size = brw->urb.vsize;
 
    /* BRW_NEW_CURBE_OFFSETS, _NEW_TRANSFORM */
-   if (brw->attribs.Transform->ClipPlanesEnabled) {
+   if (ctx->Transform.ClipPlanesEnabled) {
       /* Note that we read in the userclip planes as well, hence
        * clip_start:
        */
diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c
index 2d4c81274e..b501a59ccd 100644
--- a/src/mesa/drivers/dri/i965/brw_vtbl.c
+++ b/src/mesa/drivers/dri/i965/brw_vtbl.c
@@ -23,14 +23,12 @@
  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  
- **********************************************************************/
- /*
-  * Authors:
-  *   Keith Whitwell <keith@tungstengraphics.com>
-  */
-            
-
+**********************************************************************/
 
+/*
+ * Authors:
+ *   Keith Whitwell <keith@tungstengraphics.com>
+ */
 
 #include "main/glheader.h"
 #include "main/mtypes.h"
@@ -44,12 +42,11 @@
 #include "brw_context.h"
 #include "brw_defines.h"
 #include "brw_state.h"
-
 #include "brw_draw.h"
 #include "brw_state.h"
 #include "brw_fallback.h"
 #include "brw_vs.h"
-#include <stdarg.h>
+
 
 static void
 dri_bo_release(dri_bo **bo)
@@ -58,21 +55,22 @@ dri_bo_release(dri_bo **bo)
    *bo = NULL;
 }
 
-/* called from intelDestroyContext()
+
+/**
+ * called from intelDestroyContext()
  */
 static void brw_destroy_context( struct intel_context *intel )
 {
    struct brw_context *brw = brw_context(&intel->ctx);
    int i;
 
-   brw_destroy_metaops(brw);
    brw_destroy_state(brw);
    brw_draw_destroy( brw );
 
    brw_FrameBufferTexDestroy( brw );
 
    for (i = 0; i < brw->state.nr_draw_regions; i++)
-       intel_region_release(&brw->state.draw_regions[i]);
+      intel_region_release(&brw->state.draw_regions[i]);
    brw->state.nr_draw_regions = 0;
    intel_region_release(&brw->state.depth_region);
 
@@ -99,37 +97,46 @@ static void brw_destroy_context( struct intel_context *intel )
    dri_bo_release(&brw->cc.vp_bo);
 }
 
-/* called from intelDrawBuffer()
+
+/**
+ * called from intelDrawBuffer()
  */
 static void brw_set_draw_region( struct intel_context *intel, 
-				  struct intel_region *draw_regions[],
-				  struct intel_region *depth_region,
-				GLuint num_regions)
+                                 struct intel_region *draw_regions[],
+                                 struct intel_region *depth_region,
+                                 GLuint num_regions)
 {
    struct brw_context *brw = brw_context(&intel->ctx);
    int i;
+
+   /* release old color/depth regions */
    if (brw->state.depth_region != depth_region)
       brw->state.dirty.brw |= BRW_NEW_DEPTH_BUFFER;
    for (i = 0; i < brw->state.nr_draw_regions; i++)
        intel_region_release(&brw->state.draw_regions[i]);
    intel_region_release(&brw->state.depth_region);
+
+   /* reference new color/depth regions */
    for (i = 0; i < num_regions; i++)
        intel_region_reference(&brw->state.draw_regions[i], draw_regions[i]);
    intel_region_reference(&brw->state.depth_region, depth_region);
    brw->state.nr_draw_regions = num_regions;
 }
 
-/* called from intel_batchbuffer_flush and children before sending a
+
+/**
+ * called from intel_batchbuffer_flush and children before sending a
  * batchbuffer off.
  */
 static void brw_finish_batch(struct intel_context *intel)
 {
    struct brw_context *brw = brw_context(&intel->ctx);
-
    brw_emit_query_end(brw);
 }
 
-/* called from intelFlushBatchLocked
+
+/**
+ * called from intelFlushBatchLocked
  */
 static void brw_new_batch( struct intel_context *intel )
 {
@@ -160,24 +167,21 @@ static void brw_new_batch( struct intel_context *intel )
    }
 }
 
-static void brw_note_fence( struct intel_context *intel, 
-			    GLuint fence )
+
+static void brw_note_fence( struct intel_context *intel, GLuint fence )
 {
    brw_context(&intel->ctx)->state.dirty.brw |= BRW_NEW_FENCE;
 }
- 
+
+
 static void brw_note_unlock( struct intel_context *intel )
 {
    struct brw_context *brw = brw_context(&intel->ctx);
-
    brw_state_cache_check_size(brw);
-
-   brw_context(&intel->ctx)->state.dirty.brw |= BRW_NEW_LOCK;
 }
 
 
-void brw_do_flush( struct brw_context *brw, 
-		   GLuint flags )
+void brw_do_flush( struct brw_context *brw, GLuint flags )
 {
    struct brw_mi_flush flush;
    memset(&flush, 0, sizeof(flush));      
@@ -187,8 +191,7 @@ void brw_do_flush( struct brw_context *brw,
 }
 
 
-static void brw_emit_flush( struct intel_context *intel,
-			GLuint unused )
+static void brw_emit_flush( struct intel_context *intel, GLuint unused )
 {
    brw_do_flush(brw_context(&intel->ctx),
 		BRW_FLUSH_STATE_CACHE|BRW_FLUSH_READ_CACHE);
@@ -208,6 +211,7 @@ static GLuint brw_flush_cmd( void )
    return *(GLuint *)&flush;
 }
 
+
 static void brw_invalidate_state( struct intel_context *intel, GLuint new_state )
 {
    /* nothing */
@@ -217,14 +221,14 @@ static void brw_invalidate_state( struct intel_context *intel, GLuint new_state
 void brwInitVtbl( struct brw_context *brw )
 {
    brw->intel.vtbl.check_vertex_size = 0;
-   brw->intel.vtbl.emit_state = 0; 
-   brw->intel.vtbl.reduced_primitive_state = 0;	
+   brw->intel.vtbl.emit_state = 0;
+   brw->intel.vtbl.reduced_primitive_state = 0;
    brw->intel.vtbl.render_start = 0;
-   brw->intel.vtbl.update_texture_state = 0; 
+   brw->intel.vtbl.update_texture_state = 0;
 
-   brw->intel.vtbl.invalidate_state = brw_invalidate_state; 
-   brw->intel.vtbl.note_fence = brw_note_fence; 
-   brw->intel.vtbl.note_unlock = brw_note_unlock; 
+   brw->intel.vtbl.invalidate_state = brw_invalidate_state;
+   brw->intel.vtbl.note_fence = brw_note_fence;
+   brw->intel.vtbl.note_unlock = brw_note_unlock;
    brw->intel.vtbl.new_batch = brw_new_batch;
    brw->intel.vtbl.finish_batch = brw_finish_batch;
    brw->intel.vtbl.destroy = brw_destroy_context;
@@ -233,4 +237,3 @@ void brwInitVtbl( struct brw_context *brw )
    brw->intel.vtbl.emit_flush = brw_emit_flush;
    brw->intel.vtbl.debug_batch = brw_debug_batch;
 }
-
diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c
index 361312c2ca..ea708a0681 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.c
+++ b/src/mesa/drivers/dri/i965/brw_wm.c
@@ -36,63 +36,24 @@
 #include "brw_state.h"
 
 
+/** Return number of src args for given instruction */
 GLuint brw_wm_nr_args( GLuint opcode )
 {
    switch (opcode) {
-
    case WM_PIXELXY:
-   case OPCODE_ABS:
-   case OPCODE_FLR:
-   case OPCODE_FRC:
-   case OPCODE_SWZ:
-   case OPCODE_MOV:
-   case OPCODE_COS:
-   case OPCODE_EX2:
-   case OPCODE_LG2:
-   case OPCODE_RCP:
-   case OPCODE_RSQ:
-   case OPCODE_SIN:
-   case OPCODE_SCS:
-   case OPCODE_TEX:
-   case OPCODE_TXB:
-   case OPCODE_TXP:	
-   case OPCODE_KIL:
-   case OPCODE_LIT: 
-   case WM_CINTERP: 
-   case WM_WPOSXY: 
+   case WM_CINTERP:
+   case WM_WPOSXY:
       return 1;
-
-   case OPCODE_POW:
-   case OPCODE_SUB:
-   case OPCODE_SGE:
-   case OPCODE_SGT:
-   case OPCODE_SLE:
-   case OPCODE_SLT:
-   case OPCODE_SEQ:
-   case OPCODE_SNE:
-   case OPCODE_ADD:
-   case OPCODE_MAX:
-   case OPCODE_MIN:
-   case OPCODE_MUL:
-   case OPCODE_XPD:
-   case OPCODE_DP3:	
-   case OPCODE_DP4:
-   case OPCODE_DPH:
-   case OPCODE_DST:
-   case WM_LINTERP: 
+   case WM_LINTERP:
    case WM_DELTAXY:
    case WM_PIXELW:
       return 2;
-
    case WM_FB_WRITE:
-   case WM_PINTERP: 
-   case OPCODE_MAD:	
-   case OPCODE_CMP:
-   case OPCODE_LRP:
+   case WM_PINTERP:
       return 3;
-      
    default:
-      return 0;
+      assert(opcode < MAX_OPCODE);
+      return _mesa_num_inst_src_regs(opcode);
    }
 }
 
@@ -175,6 +136,9 @@ static void do_wm_prog( struct brw_context *brw,
 	*/
        brw_wm_emit(c);
    }
+   if (INTEL_DEBUG & DEBUG_WM)
+      fprintf(stderr, "\n");
+
    /* get the program
     */
    program = brw_get_program(&c->func, &program_size);
@@ -193,6 +157,7 @@ static void do_wm_prog( struct brw_context *brw,
 static void brw_wm_populate_key( struct brw_context *brw,
 				 struct brw_wm_prog_key *key )
 {
+   GLcontext *ctx = &brw->intel.ctx;
    /* BRW_NEW_FRAGMENT_PROGRAM */
    struct brw_fragment_program *fp = 
       (struct brw_fragment_program *)brw->fragment_program;
@@ -206,57 +171,50 @@ static void brw_wm_populate_key( struct brw_context *brw,
     */
    /* _NEW_COLOR */
    if (fp->program.UsesKill ||
-       brw->attribs.Color->AlphaEnabled)
+       ctx->Color.AlphaEnabled)
       lookup |= IZ_PS_KILL_ALPHATEST_BIT;
 
    if (fp->program.Base.OutputsWritten & (1<<FRAG_RESULT_DEPR))
       lookup |= IZ_PS_COMPUTES_DEPTH_BIT;
 
    /* _NEW_DEPTH */
-   if (brw->attribs.Depth->Test)
+   if (ctx->Depth.Test)
       lookup |= IZ_DEPTH_TEST_ENABLE_BIT;
 
-   if (brw->attribs.Depth->Test &&  
-       brw->attribs.Depth->Mask) /* ?? */
+   if (ctx->Depth.Test &&  
+       ctx->Depth.Mask) /* ?? */
       lookup |= IZ_DEPTH_WRITE_ENABLE_BIT;
 
    /* _NEW_STENCIL */
-   if (brw->attribs.Stencil->Enabled) {
+   if (ctx->Stencil.Enabled) {
       lookup |= IZ_STENCIL_TEST_ENABLE_BIT;
 
-      if (brw->attribs.Stencil->WriteMask[0] ||
-	  (brw->attribs.Stencil->_TestTwoSide &&
-	   brw->attribs.Stencil->WriteMask[1]))
+      if (ctx->Stencil.WriteMask[0] ||
+	  ctx->Stencil.WriteMask[ctx->Stencil._BackFace])
 	 lookup |= IZ_STENCIL_WRITE_ENABLE_BIT;
    }
 
-   /* XXX: when should this be disabled?
-    */
-   if (1)
-      lookup |= IZ_EARLY_DEPTH_TEST_BIT;
-
-   
    line_aa = AA_NEVER;
 
    /* _NEW_LINE, _NEW_POLYGON, BRW_NEW_REDUCED_PRIMITIVE */
-   if (brw->attribs.Line->SmoothFlag) {
+   if (ctx->Line.SmoothFlag) {
       if (brw->intel.reduced_primitive == GL_LINES) {
 	 line_aa = AA_ALWAYS;
       }
       else if (brw->intel.reduced_primitive == GL_TRIANGLES) {
-	 if (brw->attribs.Polygon->FrontMode == GL_LINE) {
+	 if (ctx->Polygon.FrontMode == GL_LINE) {
 	    line_aa = AA_SOMETIMES;
 
-	    if (brw->attribs.Polygon->BackMode == GL_LINE ||
-		(brw->attribs.Polygon->CullFlag &&
-		 brw->attribs.Polygon->CullFaceMode == GL_BACK))
+	    if (ctx->Polygon.BackMode == GL_LINE ||
+		(ctx->Polygon.CullFlag &&
+		 ctx->Polygon.CullFaceMode == GL_BACK))
 	       line_aa = AA_ALWAYS;
 	 }
-	 else if (brw->attribs.Polygon->BackMode == GL_LINE) {
+	 else if (ctx->Polygon.BackMode == GL_LINE) {
 	    line_aa = AA_SOMETIMES;
 
-	    if ((brw->attribs.Polygon->CullFlag &&
-		 brw->attribs.Polygon->CullFaceMode == GL_FRONT))
+	    if ((ctx->Polygon.CullFlag &&
+		 ctx->Polygon.CullFaceMode == GL_FRONT))
 	       line_aa = AA_ALWAYS;
 	 }
       }
@@ -271,20 +229,25 @@ static void brw_wm_populate_key( struct brw_context *brw,
    key->projtex_mask = brw->wm.input_size_masks[4-1] >> (FRAG_ATTRIB_TEX0 - FRAG_ATTRIB_WPOS); 
 
    /* _NEW_LIGHT */
-   key->flat_shade = (brw->attribs.Light->ShadeModel == GL_FLAT);
+   key->flat_shade = (ctx->Light.ShadeModel == GL_FLAT);
 
    /* _NEW_TEXTURE */
    for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
-      const struct gl_texture_unit *unit = &brw->attribs.Texture->Unit[i];
-      const struct gl_texture_object *t = unit->_Current;
+      const struct gl_texture_unit *unit = &ctx->Texture.Unit[i];
 
       if (unit->_ReallyEnabled) {
-	 if (t->Image[0][t->BaseLevel]->InternalFormat == GL_YCBCR_MESA) {
-	    key->yuvtex_mask |= 1<<i;
-	    if (t->Image[0][t->BaseLevel]->TexFormat->MesaFormat == 
-		    MESA_FORMAT_YCBCR)
-		key->yuvtex_swap_mask |= 1<< i;
+         const struct gl_texture_object *t = unit->_Current;
+         const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
+	 if (img->InternalFormat == GL_YCBCR_MESA) {
+	    key->yuvtex_mask |= 1 << i;
+	    if (img->TexFormat->MesaFormat == MESA_FORMAT_YCBCR)
+		key->yuvtex_swap_mask |= 1 << i;
 	 }
+
+         key->tex_swizzles[i] = t->_Swizzle;
+      }
+      else {
+         key->tex_swizzles[i] = SWIZZLE_NOOP;
       }
    }
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h
index 896390c17b..0f46a25b1a 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.h
+++ b/src/mesa/drivers/dri/i965/brw_wm.h
@@ -49,8 +49,7 @@
 #define IZ_DEPTH_TEST_ENABLE_BIT    0x8
 #define IZ_STENCIL_WRITE_ENABLE_BIT 0x10
 #define IZ_STENCIL_TEST_ENABLE_BIT  0x20
-#define IZ_EARLY_DEPTH_TEST_BIT     0x40
-#define IZ_BIT_MAX                  0x80
+#define IZ_BIT_MAX                  0x40
 
 #define AA_NEVER     0
 #define AA_SOMETIMES 1
@@ -61,16 +60,17 @@ struct brw_wm_prog_key {
    GLuint aa_dest_stencil_reg:3;
    GLuint dest_depth_reg:3;
    GLuint nr_depth_regs:3;
-   GLuint projtex_mask:8;
-   GLuint shadowtex_mask:8;
    GLuint computes_depth:1;	/* could be derived from program string */
    GLuint source_depth_to_render_target:1;
    GLuint flat_shade:1;
    GLuint runtime_check_aads_emit:1;
    
-   GLuint yuvtex_mask:8;
-   GLuint yuvtex_swap_mask:8;	/* UV swaped */
-   GLuint pad1:16;
+   GLuint projtex_mask:16;
+   GLuint shadowtex_mask:16;
+   GLuint yuvtex_mask:16;
+   GLuint yuvtex_swap_mask:16;	/* UV swaped */
+
+   GLuint tex_swizzles[BRW_MAX_TEX_UNIT];
 
    GLuint program_string_id:32;
    GLuint origin_x, origin_y;
diff --git a/src/mesa/drivers/dri/i965/brw_wm_debug.c b/src/mesa/drivers/dri/i965/brw_wm_debug.c
index f31d0974ec..8f07f89ebc 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_debug.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_debug.c
@@ -163,9 +163,9 @@ void brw_wm_print_program( struct brw_wm_compile *c,
 {
    GLuint insn;
 
-   _mesa_printf("\n\n\n%s:\n", stage);
+   _mesa_printf("%s:\n", stage);
    for (insn = 0; insn < c->nr_insns; insn++)
       brw_wm_print_insn(c, &c->instruction[insn]);
-   _mesa_printf("\n\n\n");
+   _mesa_printf("\n");
 }
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c
index 58c78c4b2c..b5050a3e40 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c
@@ -194,7 +194,7 @@ static void emit_linterp( struct brw_compile *p,
    interp[2] = brw_vec1_grf(nr+1, 0);
    interp[3] = brw_vec1_grf(nr+1, 4);
 
-   for(i = 0; i < 4; i++ ) {
+   for (i = 0; i < 4; i++) {
       if (mask & (1<<i)) {
 	 brw_LINE(p, brw_null_reg(), interp[i], deltas[0]);
 	 brw_MAC(p, dst[i], suboffset(interp[i],1), deltas[1]);
@@ -219,42 +219,40 @@ static void emit_pinterp( struct brw_compile *p,
    interp[2] = brw_vec1_grf(nr+1, 0);
    interp[3] = brw_vec1_grf(nr+1, 4);
 
-   for(i = 0; i < 4; i++ ) {
+   for (i = 0; i < 4; i++) {
       if (mask & (1<<i)) {
 	 brw_LINE(p, brw_null_reg(), interp[i], deltas[0]);
 	 brw_MAC(p, dst[i], suboffset(interp[i],1), deltas[1]);
       }
    }
-   for(i = 0; i < 4; i++ ) {
+   for (i = 0; i < 4; i++) {
       if (mask & (1<<i)) {
 	 brw_MUL(p, dst[i], dst[i], w[3]);
       }
    }
 }
 
+
 static void emit_cinterp( struct brw_compile *p, 
 			 const struct brw_reg *dst,
 			 GLuint mask,
 			 const struct brw_reg *arg0 )
 {
-	struct brw_reg interp[4];
-	GLuint nr = arg0[0].nr;
-	GLuint i;
-
-	interp[0] = brw_vec1_grf(nr, 0);
-	interp[1] = brw_vec1_grf(nr, 4);
-	interp[2] = brw_vec1_grf(nr+1, 0);
-	interp[3] = brw_vec1_grf(nr+1, 4);
-
-	for(i = 0; i < 4; i++ ) {
-		if (mask & (1<<i)) {
-			brw_MOV(p, dst[i], suboffset(interp[i],3));	/* TODO: optimize away like other moves */
-		}
-	}
-}
-
+   struct brw_reg interp[4];
+   GLuint nr = arg0[0].nr;
+   GLuint i;
 
+   interp[0] = brw_vec1_grf(nr, 0);
+   interp[1] = brw_vec1_grf(nr, 4);
+   interp[2] = brw_vec1_grf(nr+1, 0);
+   interp[3] = brw_vec1_grf(nr+1, 4);
 
+   for (i = 0; i < 4; i++) {
+      if (mask & (1<<i)) {
+         brw_MOV(p, dst[i], suboffset(interp[i],3));	/* TODO: optimize away like other moves */
+      }
+   }
+}
 
 
 static void emit_alu1( struct brw_compile *p, 
@@ -280,6 +278,7 @@ static void emit_alu1( struct brw_compile *p,
       brw_set_saturate(p, 0);
 }
 
+
 static void emit_alu2( struct brw_compile *p, 
 		       struct brw_instruction *(*func)(struct brw_compile *, 
 						       struct brw_reg, 
@@ -351,6 +350,7 @@ static void emit_lrp( struct brw_compile *p,
       }
    }
 }
+
 static void emit_sop( struct brw_compile *p, 
 		      const struct brw_reg *dst,
 		      GLuint mask,
@@ -376,7 +376,7 @@ static void emit_slt( struct brw_compile *p,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
-	 emit_sop(p, dst, mask, BRW_CONDITIONAL_L, arg0, arg1);
+   emit_sop(p, dst, mask, BRW_CONDITIONAL_L, arg0, arg1);
 }
 
 static void emit_sle( struct brw_compile *p, 
@@ -385,7 +385,7 @@ static void emit_sle( struct brw_compile *p,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
-	 emit_sop(p, dst, mask, BRW_CONDITIONAL_LE, arg0, arg1);
+   emit_sop(p, dst, mask, BRW_CONDITIONAL_LE, arg0, arg1);
 }
 
 static void emit_sgt( struct brw_compile *p, 
@@ -394,7 +394,7 @@ static void emit_sgt( struct brw_compile *p,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
-	 emit_sop(p, dst, mask, BRW_CONDITIONAL_G, arg0, arg1);
+   emit_sop(p, dst, mask, BRW_CONDITIONAL_G, arg0, arg1);
 }
 
 static void emit_sge( struct brw_compile *p, 
@@ -403,7 +403,7 @@ static void emit_sge( struct brw_compile *p,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
-	 emit_sop(p, dst, mask, BRW_CONDITIONAL_GE, arg0, arg1);
+   emit_sop(p, dst, mask, BRW_CONDITIONAL_GE, arg0, arg1);
 }
 
 static void emit_seq( struct brw_compile *p, 
@@ -412,7 +412,7 @@ static void emit_seq( struct brw_compile *p,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
-	 emit_sop(p, dst, mask, BRW_CONDITIONAL_EQ, arg0, arg1);
+   emit_sop(p, dst, mask, BRW_CONDITIONAL_EQ, arg0, arg1);
 }
 
 static void emit_sne( struct brw_compile *p, 
@@ -421,7 +421,7 @@ static void emit_sne( struct brw_compile *p,
 		      const struct brw_reg *arg0,
 		      const struct brw_reg *arg1 )
 {
-	 emit_sop(p, dst, mask, BRW_CONDITIONAL_NEQ, arg0, arg1);
+   emit_sop(p, dst, mask, BRW_CONDITIONAL_NEQ, arg0, arg1);
 }
 
 static void emit_cmp( struct brw_compile *p, 
@@ -505,7 +505,7 @@ static void emit_dp3( struct brw_compile *p,
 		      const struct brw_reg *arg1 )
 {
    if (!(mask & WRITEMASK_XYZW))
-      return; /* Do not emit dead code*/
+      return; /* Do not emit dead code */
 
    assert((mask & WRITEMASK_XYZW) == WRITEMASK_X);
 
@@ -525,7 +525,7 @@ static void emit_dp4( struct brw_compile *p,
 		      const struct brw_reg *arg1 )
 {
    if (!(mask & WRITEMASK_XYZW))
-      return; /* Do not emit dead code*/
+      return; /* Do not emit dead code */
 
    assert((mask & WRITEMASK_XYZW) == WRITEMASK_X);
 
@@ -546,7 +546,7 @@ static void emit_dph( struct brw_compile *p,
 		      const struct brw_reg *arg1 )
 {
    if (!(mask & WRITEMASK_XYZW))
-      return; /* Do not emit dead code*/
+      return; /* Do not emit dead code */
 
    assert((mask & WRITEMASK_XYZW) == WRITEMASK_X);
 
@@ -592,7 +592,7 @@ static void emit_math1( struct brw_compile *p,
 			const struct brw_reg *arg0 )
 {
    if (!(mask & WRITEMASK_XYZW))
-      return; /* Do not emit dead code*/
+      return; /* Do not emit dead code */
 
    //assert((mask & WRITEMASK_XYZW) == WRITEMASK_X ||
    //	  function == BRW_MATH_FUNCTION_SINCOS);
@@ -619,7 +619,7 @@ static void emit_math2( struct brw_compile *p,
 			const struct brw_reg *arg1)
 {
    if (!(mask & WRITEMASK_XYZW))
-      return; /* Do not emit dead code*/
+      return; /* Do not emit dead code */
 
    assert((mask & WRITEMASK_XYZW) == WRITEMASK_X);
 
@@ -760,7 +760,6 @@ static void emit_txb( struct brw_wm_compile *c,
    brw_MOV(p, brw_message_reg(8), arg[3]);
    msgLength = 9;
 
-
    brw_SAMPLE(p, 
 	      retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW),
 	      1,
@@ -772,7 +771,6 @@ static void emit_txb( struct brw_wm_compile *c,
 	      8,		/* responseLength */
 	      msgLength,
 	      0);	
-
 }
 
 
@@ -823,7 +821,6 @@ static void emit_kil( struct brw_wm_compile *c,
    struct brw_reg r0uw = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW);
    GLuint i;
    
-
    /* XXX - usually won't need 4 compares!
     */
    for (i = 0; i < 4; i++) {
@@ -836,6 +833,7 @@ static void emit_kil( struct brw_wm_compile *c,
    }
 }
 
+
 static void fire_fb_write( struct brw_wm_compile *c,
 			   GLuint base_reg,
 			   GLuint nr,
@@ -869,6 +867,7 @@ static void fire_fb_write( struct brw_wm_compile *c,
 		eot);
 }
 
+
 static void emit_aa( struct brw_wm_compile *c,
 		     struct brw_reg *arg1,
 		     GLuint reg )
@@ -962,7 +961,6 @@ static void emit_fb_write( struct brw_wm_compile *c,
       nr += 2;
    }
 
-
    if (!c->key.runtime_check_aads_emit) {
       if (c->key.aa_dest_stencil_reg)
 	 emit_aa(c, arg1, 2);
@@ -996,8 +994,6 @@ static void emit_fb_write( struct brw_wm_compile *c,
 }
 
 
-
-
 /* Post-fragment-program processing.  Send the results to the
  * framebuffer.
  */
@@ -1022,6 +1018,7 @@ static void emit_spill( struct brw_wm_compile *c,
 		   slot);
 }
 
+
 static void emit_unspill( struct brw_wm_compile *c,
 			  struct brw_reg reg,
 			  GLuint slot )
@@ -1047,7 +1044,6 @@ static void emit_unspill( struct brw_wm_compile *c,
 }
 
 
-
 /**
  * Retrieve upto 4 GEN4 register pairs for the given wm reg:
  */
@@ -1073,6 +1069,7 @@ static void get_argument_regs( struct brw_wm_compile *c,
    }
 }
 
+
 static void spill_values( struct brw_wm_compile *c,
 			  struct brw_wm_value *values,
 			  GLuint nr )
@@ -1085,7 +1082,6 @@ static void spill_values( struct brw_wm_compile *c,
 }
 
 
-
 /* Emit the fragment program instructions here.
  */
 void brw_wm_emit( struct brw_wm_compile *c )
@@ -1176,7 +1172,7 @@ void brw_wm_emit( struct brw_wm_compile *c )
 	 emit_alu1(p, brw_RNDD, dst, dst_flags, args[0]);
 	 break;
 
-      case OPCODE_DP3:	/*  */
+      case OPCODE_DP3:
 	 emit_dp3(p, dst, dst_flags, args[0], args[1]);
 	 break;
 
@@ -1188,7 +1184,7 @@ void brw_wm_emit( struct brw_wm_compile *c )
 	 emit_dph(p, dst, dst_flags, args[0], args[1]);
 	 break;
 
-      case OPCODE_LRP:	/*  */
+      case OPCODE_LRP:
 	 emit_lrp(p, dst, dst_flags, args[0], args[1], args[2]);
 	 break;
 
@@ -1315,8 +1311,3 @@ void brw_wm_emit( struct brw_wm_compile *c )
 		      inst->dst[i]->spill_slot);
    }
 }
-
-
-
-
-
diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c
index d3e926c328..ea3f3fc678 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_fp.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c
@@ -111,6 +111,12 @@ static struct prog_src_register src_swizzle1( struct prog_src_register reg, int
    return src_swizzle(reg, x, x, x, x);
 }
 
+static struct prog_src_register src_swizzle4( struct prog_src_register reg, uint swizzle )
+{
+   reg.Swizzle = swizzle;
+   return reg;
+}
+
 
 /***********************************************************************
  * Dest regs
@@ -122,10 +128,11 @@ static struct prog_dst_register dst_reg(GLuint file, GLuint idx)
    reg.File = file;
    reg.Index = idx;
    reg.WriteMask = WRITEMASK_XYZW;
+   reg.RelAddr = 0;
    reg.CondMask = 0;
    reg.CondSwizzle = 0;
-   reg.pad = 0;
    reg.CondSrc = 0;
+   reg.pad = 0;
    return reg;
 }
 
@@ -426,10 +433,6 @@ static struct prog_src_register search_or_add_param5(struct brw_wm_compile *c,
 
    idx = _mesa_add_state_reference( paramList, tokens );
 
-   /* Recalculate state dependency: 
-    */
-   c->fp->param_state = paramList->StateFlags;
-
    return src_reg(PROGRAM_STATE_VAR, idx);
 }
 
@@ -557,12 +560,19 @@ static void precalc_lit( struct brw_wm_compile *c,
    }
 }
 
+
+/**
+ * Some TEX instructions require extra code, cube map coordinate
+ * normalization, or coordinate scaling for RECT textures, etc.
+ * This function emits those extra instructions and the TEX
+ * instruction itself.
+ */
 static void precalc_tex( struct brw_wm_compile *c,
 			 const struct prog_instruction *inst )
 {
    struct prog_src_register coord;
    struct prog_dst_register tmpcoord;
-   GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit];
+   const GLuint unit = c->fp->program.Base.SamplerUnits[inst->TexSrcUnit];
 
    if (inst->TexSrcTarget == TEXTURE_CUBE_INDEX) {
        struct prog_instruction *out;
@@ -572,9 +582,11 @@ static void precalc_tex( struct brw_wm_compile *c,
        struct prog_src_register tmp1src = src_reg_from_dst(tmp1);
        struct prog_src_register src0 = inst->SrcReg[0];
 
+       /* find longest component of coord vector and normalize it */
        tmpcoord = get_temp(c);
        coord = src_reg_from_dst(tmpcoord);
 
+       /* tmpcoord = src0 (i.e.: coord = src0) */
        out = emit_op(c, OPCODE_MOV,
                      tmpcoord,
                      0, 0, 0,
@@ -584,6 +596,7 @@ static void precalc_tex( struct brw_wm_compile *c,
        out->SrcReg[0].NegateBase = 0;
        out->SrcReg[0].Abs = 1;
 
+       /* tmp0 = MAX(coord.X, coord.Y) */
        emit_op(c, OPCODE_MAX,
                tmp0,
                0, 0, 0,
@@ -591,6 +604,7 @@ static void precalc_tex( struct brw_wm_compile *c,
                src_swizzle1(coord, Y),
                src_undef());
 
+       /* tmp1 = MAX(tmp0, coord.Z) */
        emit_op(c, OPCODE_MAX,
                tmp1,
                0, 0, 0,
@@ -598,6 +612,7 @@ static void precalc_tex( struct brw_wm_compile *c,
                src_swizzle1(coord, Z),
                src_undef());
 
+       /* tmp0 = 1 / tmp1 */
        emit_op(c, OPCODE_RCP,
                tmp0,
                0, 0, 0,
@@ -605,6 +620,7 @@ static void precalc_tex( struct brw_wm_compile *c,
                src_undef(),
                src_undef());
 
+       /* tmpCoord = src0 * tmp0 */
        emit_op(c, OPCODE_MUL,
                tmpcoord,
                0, 0, 0,
@@ -614,7 +630,8 @@ static void precalc_tex( struct brw_wm_compile *c,
 
        release_temp(c, tmp0);
        release_temp(c, tmp1);
-   } else if (inst->TexSrcTarget == TEXTURE_RECT_INDEX) {
+   }
+   else if (inst->TexSrcTarget == TEXTURE_RECT_INDEX) {
       struct prog_src_register scale = 
 	 search_or_add_param5( c, 
 			       STATE_INTERNAL, 
@@ -645,19 +662,9 @@ static void precalc_tex( struct brw_wm_compile *c,
     * conversion requires allocating a temporary variable which we
     * don't have the facility to do that late in the compilation.
     */
-   if (!(c->key.yuvtex_mask & (1<<unit))) {
-      emit_op(c, 
-	      OPCODE_TEX,
-	      inst->DstReg,
-	      inst->SaturateMode,
-	      unit,
-	      inst->TexSrcTarget,
-	      coord,
-	      src_undef(),
-	      src_undef());
-   }
-   else {
-       GLboolean  swap_uv = c->key.yuvtex_swap_mask & (1<<unit);
+   if (c->key.yuvtex_mask & (1 << unit)) {
+      /* convert ycbcr to RGBA */
+      GLboolean  swap_uv = c->key.yuvtex_swap_mask & (1<<unit);
 
       /* 
 	 CONST C0 = { -.5, -.0625,  -.5, 1.164 }
@@ -737,6 +744,31 @@ static void precalc_tex( struct brw_wm_compile *c,
 
       release_temp(c, tmp);
    }
+   else {
+      /* ordinary RGBA tex instruction */
+      emit_op(c, 
+	      OPCODE_TEX,
+	      inst->DstReg,
+	      inst->SaturateMode,
+	      unit,
+	      inst->TexSrcTarget,
+	      coord,
+	      src_undef(),
+	      src_undef());
+   }
+
+   /* For GL_EXT_texture_swizzle: */
+   if (c->key.tex_swizzles[unit] != SWIZZLE_NOOP) {
+      /* swizzle the result of the TEX instruction */
+      struct prog_src_register tmpsrc = src_reg_from_dst(inst->DstReg);
+      emit_op(c, OPCODE_SWZ,
+              inst->DstReg,
+              SATURATE_OFF, /* saturate already done above */
+              0, 0,   /* tex unit, target N/A */
+              src_swizzle4(tmpsrc, c->key.tex_swizzles[unit]),
+              src_undef(),
+              src_undef());
+   }
 
    if ((inst->TexSrcTarget == TEXTURE_RECT_INDEX) ||
        (inst->TexSrcTarget == TEXTURE_CUBE_INDEX))
@@ -814,62 +846,11 @@ static void precalc_txp( struct brw_wm_compile *c,
 
 
 
-
-
-/***********************************************************************
- * Add instructions to perform fog blending
- */
-
-static void fog_blend( struct brw_wm_compile *c,
-			     struct prog_src_register fog_factor )
-{
-   struct prog_dst_register outcolor = dst_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLR);
-   struct prog_src_register fogcolor = search_or_add_param5( c, STATE_FOG_COLOR, 0,0,0,0 );
-
-   /* color.xyz = LRP fog_factor.xxxx, output_color, fog_color */
-   
-   emit_op(c, 
-	   OPCODE_LRP,
-	   dst_mask(outcolor, WRITEMASK_XYZ),
-	   0, 0, 0,
-	   fog_factor,
-	   src_reg_from_dst(outcolor),
-	   fogcolor);
-}
-
-
-
-/* This one is simple - just take the interpolated fog coordinate and
- * use it as the fog blend factor.
- */
-static void fog_interpolated( struct brw_wm_compile *c )
-{
-   struct prog_src_register fogc = src_reg(PROGRAM_INPUT, FRAG_ATTRIB_FOGC);
-   
-   if (!(c->fp_interp_emitted & (1<<FRAG_ATTRIB_FOGC))) 
-      emit_interp(c, FRAG_ATTRIB_FOGC);
-
-   fog_blend( c, src_swizzle1(fogc, GET_SWZ(fogc.Swizzle,X)));
-}
-
-static void emit_fog( struct brw_wm_compile *c ) 
-{
-   if (!c->fp->program.FogOption)
-      return;
-
-   if (1) 
-      fog_interpolated( c );
-   else {
-      /* TODO: per-pixel fog */
-      assert(0);
-   }
-}
-
 static void emit_fb_write( struct brw_wm_compile *c )
 {
-   struct prog_src_register outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLR);
    struct prog_src_register payload_r0_depth = src_reg(PROGRAM_PAYLOAD, PAYLOAD_DEPTH);
    struct prog_src_register outdepth = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_DEPR);
+   struct prog_src_register outcolor;
    GLuint i;
 
    struct prog_instruction *inst, *last_inst;
@@ -893,7 +874,14 @@ static void emit_fb_write( struct brw_wm_compile *c )
 	   }
        }
        last_inst->Sampler |= 1; //eot
-   }else {
+   }
+   else {
+      /* if gl_FragData[0] is written, use it, else use gl_FragColor */
+      if (c->fp->program.Base.OutputsWritten & (1 << FRAG_RESULT_DATA0))
+         outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_DATA0);
+      else 
+         outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLR);
+
        inst = emit_op(c, WM_FB_WRITE, dst_mask(dst_undef(),0),
 	       0, 0, 0, outcolor, payload_r0_depth, outdepth);
        inst->Sampler = 1|(0<<1);
@@ -960,7 +948,7 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
    GLuint insn;
 
    if (INTEL_DEBUG & DEBUG_WM) {
-      _mesa_printf("\n\n\npre-fp:\n");
+      _mesa_printf("pre-fp:\n");
       _mesa_print_program(&fp->program.Base); 
       _mesa_printf("\n");
    }
@@ -1055,7 +1043,6 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
          emit_ddy(c, inst);
 	break;
       case OPCODE_END:
-	 emit_fog(c);
 	 emit_fb_write(c);
 	 break;
       case OPCODE_PRINT:
@@ -1068,7 +1055,7 @@ void brw_wm_pass_fp( struct brw_wm_compile *c )
    }
 
    if (INTEL_DEBUG & DEBUG_WM) {
-	   _mesa_printf("\n\n\npass_fp:\n");
+	   _mesa_printf("pass_fp:\n");
 	   print_insns( c->prog_instructions, c->nr_fp_insns );
 	   _mesa_printf("\n");
    }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
index cb728190f5..8fd776ac39 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
@@ -267,7 +267,7 @@ static void emit_trunc( struct brw_wm_compile *c,
 	    struct brw_reg src, dst;
 	    dst = get_dst_reg(c, inst, i, 1) ;
 	    src = get_src_reg(c, &inst->SrcReg[0], i, 1);
-	    brw_RNDD(p, dst, src);
+	    brw_RNDZ(p, dst, src);
 	}
     }
     brw_set_saturate(p, 0);
@@ -623,7 +623,7 @@ static void emit_dph(struct brw_wm_compile *c,
     brw_MAC(p, brw_null_reg(), src0[1], src1[1]);
     brw_MAC(p, dst, src0[2], src1[2]);
     brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0);
-    brw_ADD(p, dst, src0[3], src1[3]);
+    brw_ADD(p, dst, dst, src1[3]);
     brw_set_saturate(p, 0);
 }
 
@@ -892,15 +892,19 @@ static void emit_lrp(struct brw_wm_compile *c,
     }
 }
 
+/**
+ * For GLSL shaders, this KIL will be unconditional.
+ * It may be contained inside an IF/ENDIF structure of course.
+ */
 static void emit_kil(struct brw_wm_compile *c)
 {
-	struct brw_compile *p = &c->func;
-	struct brw_reg depth = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW);
-	brw_push_insn_state(p);
-	brw_set_mask_control(p, BRW_MASK_DISABLE);
-	brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); //IMASK
-	brw_AND(p, depth, c->emit_mask_reg, depth);
-	brw_pop_insn_state(p);
+    struct brw_compile *p = &c->func;
+    struct brw_reg depth = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW);
+    brw_push_insn_state(p);
+    brw_set_mask_control(p, BRW_MASK_DISABLE);
+    brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); //IMASK
+    brw_AND(p, depth, c->emit_mask_reg, depth);
+    brw_pop_insn_state(p);
 }
 
 static void emit_mad(struct brw_wm_compile *c,
@@ -1095,7 +1099,7 @@ static void noise1_sub( struct brw_wm_compile *c ) {
     /* Arrange the two end coordinates into scalars (itmp0/itmp1) to
        be hashed.  Also compute the remainder (offset within the unit
        length), interleaved to reduce register dependency penalties. */
-    brw_RNDD( p, itmp[ 0 ], param );
+    brw_RNDD( p, retype( itmp[ 0 ], BRW_REGISTER_TYPE_D ), param );
     brw_FRC( p, param, param );
     brw_ADD( p, itmp[ 1 ], itmp[ 0 ], brw_imm_ud( 1 ) );
     brw_MOV( p, itmp[ 3 ], brw_imm_ud( 0x79D9 ) ); /* constant used later */
@@ -1220,8 +1224,8 @@ static void noise2_sub( struct brw_wm_compile *c ) {
     /* Arrange the four corner coordinates into scalars (itmp0..itmp3) to
        be hashed.  Also compute the remainders (offsets within the unit
        square), interleaved to reduce register dependency penalties. */
-    brw_RNDD( p, itmp[ 0 ], param0 );
-    brw_RNDD( p, itmp[ 1 ], param1 );
+    brw_RNDD( p, retype( itmp[ 0 ], BRW_REGISTER_TYPE_D ), param0 );
+    brw_RNDD( p, retype( itmp[ 1 ], BRW_REGISTER_TYPE_D ), param1 );
     brw_FRC( p, param0, param0 );
     brw_FRC( p, param1, param1 );
     brw_MOV( p, itmp[ 4 ], brw_imm_ud( 0xBA97 ) ); /* constant used later */
@@ -1400,21 +1404,19 @@ static void noise3_sub( struct brw_wm_compile *c ) {
     /* Arrange the eight corner coordinates into scalars (itmp0..itmp3) to
        be hashed.  Also compute the remainders (offsets within the unit
        cube), interleaved to reduce register dependency penalties. */
-    brw_RNDD( p, itmp[ 0 ], param0 );
-    brw_RNDD( p, itmp[ 1 ], param1 );
-    brw_RNDD( p, itmp[ 2 ], param2 );
-    brw_MOV( p, itmp[ 4 ], brw_imm_ud( 0xBC8F ) ); /* constant used later */
-    brw_MOV( p, itmp[ 5 ], brw_imm_ud( 0xD0BD ) ); /* constant used later */
-    brw_MOV( p, itmp[ 6 ], brw_imm_ud( 0x9B93 ) ); /* constant used later */
+    brw_RNDD( p, retype( itmp[ 0 ], BRW_REGISTER_TYPE_D ), param0 );
+    brw_RNDD( p, retype( itmp[ 1 ], BRW_REGISTER_TYPE_D ), param1 );
+    brw_RNDD( p, retype( itmp[ 2 ], BRW_REGISTER_TYPE_D ), param2 );
     brw_FRC( p, param0, param0 );
     brw_FRC( p, param1, param1 );
     brw_FRC( p, param2, param2 );
     /* Since we now have only 16 bits of precision in the hash, we must
        be more careful about thorough mixing to maintain entropy as we
        squash the input vector into a small scalar. */
-    brw_MUL( p, brw_acc_reg(), itmp[ 4 ], itmp[ 0 ] );
-    brw_MAC( p, brw_acc_reg(), itmp[ 5 ], itmp[ 1 ] );
-    brw_MAC( p, itmp[ 0 ], itmp[ 6 ], itmp[ 2 ] );
+    brw_MUL( p, brw_null_reg(), low_words( itmp[ 0 ] ), brw_imm_uw( 0xBC8F ) );
+    brw_MAC( p, brw_null_reg(), low_words( itmp[ 1 ] ), brw_imm_uw( 0xD0BD ) );
+    brw_MAC( p, low_words( itmp[ 0 ] ), low_words( itmp[ 2 ] ),
+	     brw_imm_uw( 0x9B93 ) );
     brw_ADD( p, high_words( itmp[ 0 ] ), low_words( itmp[ 0 ] ),
 	     brw_imm_uw( 0xBC8F ) );
 
@@ -1668,6 +1670,430 @@ static void emit_noise3( struct brw_wm_compile *c,
     release_tmps( c, mark );
 }
     
+/* For the four-dimensional case, the little micro-optimisation benefits
+   we obtain by unrolling all the loops aren't worth the massive bloat it
+   now causes.  Instead, we loop twice around performing a similar operation
+   to noise3, once for the w=0 cube and once for the w=1, with a bit more
+   code to glue it all together. */
+static void noise4_sub( struct brw_wm_compile *c ) {
+
+    struct brw_compile *p = &c->func;
+    struct brw_reg param[ 4 ],
+	x0y0, x0y1, x1y0, x1y1, /* gradients at four of the corners */
+	w0, /* noise for the w=0 cube */
+	floors[ 2 ], /* integer coordinates of base corner of hypercube */
+	interp[ 4 ], /* interpolation coefficients */
+	t, tmp[ 8 ], /* float temporaries */
+	itmp[ 8 ], /* unsigned integer temporaries (aliases of floats above) */
+	wtmp[ 8 ]; /* 16-way unsigned word temporaries (aliases of above) */
+    int i, j;
+    int mark = mark_tmps( c );
+    GLuint loop, origin;
+    
+    x0y0 = alloc_tmp( c );
+    x0y1 = alloc_tmp( c );
+    x1y0 = alloc_tmp( c );
+    x1y1 = alloc_tmp( c );
+    t = alloc_tmp( c );
+    w0 = alloc_tmp( c );    
+    floors[ 0 ] = retype( alloc_tmp( c ), BRW_REGISTER_TYPE_UD );
+    floors[ 1 ] = retype( alloc_tmp( c ), BRW_REGISTER_TYPE_UD );
+
+    for( i = 0; i < 4; i++ ) {
+	param[ i ] = lookup_tmp( c, mark - 5 + i );
+	interp[ i ] = alloc_tmp( c );
+    }
+    
+    for( i = 0; i < 8; i++ ) {
+	tmp[ i ] = alloc_tmp( c );
+	itmp[ i ] = retype( tmp[ i ], BRW_REGISTER_TYPE_UD );
+	wtmp[ i ] = brw_uw16_grf( tmp[ i ].nr, 0 );
+    }
+
+    brw_set_access_mode( p, BRW_ALIGN_1 );
+
+    /* We only want 16 bits of precision from the integral part of each
+       co-ordinate, but unfortunately the RNDD semantics would saturate
+       at 16 bits if we performed the operation directly to a 16-bit
+       destination.  Therefore, we round to 32-bit temporaries where
+       appropriate, and then store only the lower 16 bits. */
+    brw_RNDD( p, retype( floors[ 0 ], BRW_REGISTER_TYPE_D ), param[ 0 ] );
+    brw_RNDD( p, retype( itmp[ 0 ], BRW_REGISTER_TYPE_D ), param[ 1 ] );
+    brw_RNDD( p, retype( floors[ 1 ], BRW_REGISTER_TYPE_D ), param[ 2 ] );
+    brw_RNDD( p, retype( itmp[ 1 ], BRW_REGISTER_TYPE_D ), param[ 3 ] );
+    brw_MOV( p, high_words( floors[ 0 ] ), low_words( itmp[ 0 ] ) );
+    brw_MOV( p, high_words( floors[ 1 ] ), low_words( itmp[ 1 ] ) );
+
+    /* Modify the flag register here, because the side effect is useful
+       later (see below).  We know for certain that all flags will be
+       cleared, since the FRC instruction cannot possibly generate
+       negative results.  Even for exceptional inputs (infinities, denormals,
+       NaNs), the architecture guarantees that the L conditional is false. */
+    brw_set_conditionalmod( p, BRW_CONDITIONAL_L );
+    brw_FRC( p, param[ 0 ], param[ 0 ] );
+    brw_set_predicate_control( p, BRW_PREDICATE_NONE );
+    for( i = 1; i < 4; i++ )	
+	brw_FRC( p, param[ i ], param[ i ] );
+    
+    /* Calculate the interpolation coefficients (6t^5 - 15t^4 + 10t^3) first
+       of all. */
+    for( i = 0; i < 4; i++ )
+	brw_MUL( p, interp[ i ], param[ i ], brw_imm_f( 6.0 ) );
+    for( i = 0; i < 4; i++ )
+	brw_ADD( p, interp[ i ], interp[ i ], brw_imm_f( -15.0 ) );
+    for( i = 0; i < 4; i++ )
+	brw_MUL( p, interp[ i ], interp[ i ], param[ i ] );
+    for( i = 0; i < 4; i++ )
+	brw_ADD( p, interp[ i ], interp[ i ], brw_imm_f( 10.0 ) );
+    for( j = 0; j < 3; j++ )
+	for( i = 0; i < 4; i++ )
+	    brw_MUL( p, interp[ i ], interp[ i ], param[ i ] );
+
+    /* Mark the current address, as it will be a jump destination.  The
+       following code will be executed twice: first, with the flag
+       register clear indicating the w=0 case, and second with flags
+       set for w=1. */
+    loop = p->nr_insn;
+    
+    /* Arrange the eight corner coordinates into scalars (itmp0..itmp3) to
+       be hashed.  Since we have only 16 bits of precision in the hash, we
+       must be careful about thorough mixing to maintain entropy as we
+       squash the input vector into a small scalar. */
+    brw_MUL( p, brw_null_reg(), low_words( floors[ 0 ] ),
+	     brw_imm_uw( 0xBC8F ) );
+    brw_MAC( p, brw_null_reg(), high_words( floors[ 0 ] ),
+	     brw_imm_uw( 0xD0BD ) );
+    brw_MAC( p, brw_null_reg(), low_words( floors[ 1 ] ),
+	     brw_imm_uw( 0x9B93 ) );
+    brw_MAC( p, low_words( itmp[ 0 ] ), high_words( floors[ 1 ] ),
+	     brw_imm_uw( 0xA359 ) );
+    brw_ADD( p, high_words( itmp[ 0 ] ), low_words( itmp[ 0 ] ),
+	     brw_imm_uw( 0xBC8F ) );
+
+    /* Temporarily disable the execution mask while we work with ExecSize=16
+       channels (the mask is set for ExecSize=8 and is probably incorrect).
+       Although this might cause execution of unwanted channels, the code
+       writes only to temporary registers and has no side effects, so
+       disabling the mask is harmless. */
+    brw_push_insn_state( p );
+    brw_set_mask_control( p, BRW_MASK_DISABLE );
+    brw_ADD( p, wtmp[ 1 ], wtmp[ 0 ], brw_imm_uw( 0xD0BD ) );
+    brw_ADD( p, wtmp[ 2 ], wtmp[ 0 ], brw_imm_uw( 0x9B93 ) );
+    brw_ADD( p, wtmp[ 3 ], wtmp[ 1 ], brw_imm_uw( 0x9B93 ) );
+
+    /* We're now ready to perform the hashing.  The eight hashes are
+       interleaved for performance.  The hash function used is
+       designed to rapidly achieve avalanche and require only 16x16
+       bit multiplication, and 8-bit swizzles (which we get for
+       free). */
+    for( i = 0; i < 4; i++ )
+	brw_MUL( p, wtmp[ i ], wtmp[ i ], brw_imm_uw( 0x28D9 ) );
+    for( i = 0; i < 4; i++ )
+	brw_XOR( p, even_bytes( wtmp[ i ] ), even_bytes( wtmp[ i ] ),
+		 odd_bytes( wtmp[ i ] ) );
+    for( i = 0; i < 4; i++ )
+	brw_MUL( p, wtmp[ i ], wtmp[ i ], brw_imm_uw( 0xC6D5 ) );
+    for( i = 0; i < 4; i++ )
+	brw_XOR( p, even_bytes( wtmp[ i ] ), even_bytes( wtmp[ i ] ),
+		 odd_bytes( wtmp[ i ] ) );
+    brw_pop_insn_state( p );
+
+    /* Now we want to initialise the four rear gradients based on the
+       hashes.  Format conversion from signed integer to float leaves
+       everything scaled too high by a factor of pow( 2, 15 ), but
+       we correct for that right at the end. */
+    /* x component */
+    brw_ADD( p, t, param[ 0 ], brw_imm_f( -1.0 ) );
+    brw_MOV( p, x0y0, low_words( tmp[ 0 ] ) );
+    brw_MOV( p, x0y1, low_words( tmp[ 1 ] ) );
+    brw_MOV( p, x1y0, high_words( tmp[ 0 ] ) );
+    brw_MOV( p, x1y1, high_words( tmp[ 1 ] ) );
+
+    brw_push_insn_state( p );
+    brw_set_mask_control( p, BRW_MASK_DISABLE );
+    brw_SHL( p, wtmp[ 0 ], wtmp[ 0 ], brw_imm_uw( 4 ) );
+    brw_SHL( p, wtmp[ 1 ], wtmp[ 1 ], brw_imm_uw( 4 ) );
+    brw_pop_insn_state( p );
+    
+    brw_MUL( p, x1y0, x1y0, t );
+    brw_MUL( p, x1y1, x1y1, t );
+    brw_ADD( p, t, param[ 1 ], brw_imm_f( -1.0 ) );
+    brw_MUL( p, x0y0, x0y0, param[ 0 ] );
+    brw_MUL( p, x0y1, x0y1, param[ 0 ] );
+
+    /* y component */
+    brw_MOV( p, tmp[ 5 ], low_words( tmp[ 1 ] ) );
+    brw_MOV( p, tmp[ 7 ], high_words( tmp[ 1 ] ) );
+    brw_MOV( p, tmp[ 4 ], low_words( tmp[ 0 ] ) );
+    brw_MOV( p, tmp[ 6 ], high_words( tmp[ 0 ] ) );
+    
+    brw_push_insn_state( p );
+    brw_set_mask_control( p, BRW_MASK_DISABLE );
+    brw_SHL( p, wtmp[ 0 ], wtmp[ 0 ], brw_imm_uw( 4 ) );
+    brw_SHL( p, wtmp[ 1 ], wtmp[ 1 ], brw_imm_uw( 4 ) );
+    brw_pop_insn_state( p );
+
+    brw_MUL( p, tmp[ 5 ], tmp[ 5 ], t );
+    brw_MUL( p, tmp[ 7 ], tmp[ 7 ], t );    
+    /* prepare t for the w component (used below): w the first time through
+       the loop; w - 1 the second time) */
+    brw_set_predicate_control( p, BRW_PREDICATE_NORMAL );
+    brw_ADD( p, t, param[ 3 ], brw_imm_f( -1.0 ) );
+    p->current->header.predicate_inverse = 1;
+    brw_MOV( p, t, param[ 3 ] );
+    p->current->header.predicate_inverse = 0;
+    brw_set_predicate_control( p, BRW_PREDICATE_NONE );
+    brw_MUL( p, tmp[ 4 ], tmp[ 4 ], param[ 1 ] );
+    brw_MUL( p, tmp[ 6 ], tmp[ 6 ], param[ 1 ] );
+    
+    brw_ADD( p, x0y1, x0y1, tmp[ 5 ] );
+    brw_ADD( p, x1y1, x1y1, tmp[ 7 ] );
+    brw_ADD( p, x0y0, x0y0, tmp[ 4 ] );
+    brw_ADD( p, x1y0, x1y0, tmp[ 6 ] );
+    
+    /* z component */
+    brw_MOV( p, tmp[ 4 ], low_words( tmp[ 0 ] ) );
+    brw_MOV( p, tmp[ 5 ], low_words( tmp[ 1 ] ) );
+    brw_MOV( p, tmp[ 6 ], high_words( tmp[ 0 ] ) );
+    brw_MOV( p, tmp[ 7 ], high_words( tmp[ 1 ] ) );
+
+    brw_push_insn_state( p );
+    brw_set_mask_control( p, BRW_MASK_DISABLE );
+    brw_SHL( p, wtmp[ 0 ], wtmp[ 0 ], brw_imm_uw( 4 ) );
+    brw_SHL( p, wtmp[ 1 ], wtmp[ 1 ], brw_imm_uw( 4 ) );
+    brw_pop_insn_state( p );
+
+    brw_MUL( p, tmp[ 4 ], tmp[ 4 ], param[ 2 ] );
+    brw_MUL( p, tmp[ 5 ], tmp[ 5 ], param[ 2 ] );
+    brw_MUL( p, tmp[ 6 ], tmp[ 6 ], param[ 2 ] );
+    brw_MUL( p, tmp[ 7 ], tmp[ 7 ], param[ 2 ] );
+    
+    brw_ADD( p, x0y0, x0y0, tmp[ 4 ] );
+    brw_ADD( p, x0y1, x0y1, tmp[ 5 ] );
+    brw_ADD( p, x1y0, x1y0, tmp[ 6 ] );
+    brw_ADD( p, x1y1, x1y1, tmp[ 7 ] );
+
+    /* w component */
+    brw_MOV( p, tmp[ 4 ], low_words( tmp[ 0 ] ) );
+    brw_MOV( p, tmp[ 5 ], low_words( tmp[ 1 ] ) );
+    brw_MOV( p, tmp[ 6 ], high_words( tmp[ 0 ] ) );
+    brw_MOV( p, tmp[ 7 ], high_words( tmp[ 1 ] ) );
+
+    brw_MUL( p, tmp[ 4 ], tmp[ 4 ], t );
+    brw_MUL( p, tmp[ 5 ], tmp[ 5 ], t );
+    brw_MUL( p, tmp[ 6 ], tmp[ 6 ], t );
+    brw_MUL( p, tmp[ 7 ], tmp[ 7 ], t );
+    brw_ADD( p, t, param[ 0 ], brw_imm_f( -1.0 ) );
+    
+    brw_ADD( p, x0y0, x0y0, tmp[ 4 ] );
+    brw_ADD( p, x0y1, x0y1, tmp[ 5 ] );
+    brw_ADD( p, x1y0, x1y0, tmp[ 6 ] );
+    brw_ADD( p, x1y1, x1y1, tmp[ 7 ] );
+
+    /* Here we interpolate in the y dimension... */
+    brw_ADD( p, x0y1, x0y1, negate( x0y0 ) );
+    brw_ADD( p, x1y1, x1y1, negate( x1y0 ) );
+    brw_MUL( p, x0y1, x0y1, interp[ 1 ] );
+    brw_MUL( p, x1y1, x1y1, interp[ 1 ] );
+    brw_ADD( p, x0y0, x0y0, x0y1 );
+    brw_ADD( p, x1y0, x1y0, x1y1 );
+
+    /* And now in x.  Leave the result in tmp[ 0 ] (see below)... */
+    brw_ADD( p, x1y0, x1y0, negate( x0y0 ) );
+    brw_MUL( p, x1y0, x1y0, interp[ 0 ] );
+    brw_ADD( p, tmp[ 0 ], x0y0, x1y0 );
+
+    /* Now do the same thing for the front four gradients... */
+    /* x component */
+    brw_MOV( p, x0y0, low_words( tmp[ 2 ] ) );
+    brw_MOV( p, x0y1, low_words( tmp[ 3 ] ) );
+    brw_MOV( p, x1y0, high_words( tmp[ 2 ] ) );
+    brw_MOV( p, x1y1, high_words( tmp[ 3 ] ) );
+
+    brw_push_insn_state( p );
+    brw_set_mask_control( p, BRW_MASK_DISABLE );
+    brw_SHL( p, wtmp[ 2 ], wtmp[ 2 ], brw_imm_uw( 4 ) );
+    brw_SHL( p, wtmp[ 3 ], wtmp[ 3 ], brw_imm_uw( 4 ) );
+    brw_pop_insn_state( p );
+
+    brw_MUL( p, x1y0, x1y0, t );
+    brw_MUL( p, x1y1, x1y1, t );
+    brw_ADD( p, t, param[ 1 ], brw_imm_f( -1.0 ) );
+    brw_MUL( p, x0y0, x0y0, param[ 0 ] );
+    brw_MUL( p, x0y1, x0y1, param[ 0 ] );
+
+    /* y component */
+    brw_MOV( p, tmp[ 5 ], low_words( tmp[ 3 ] ) );
+    brw_MOV( p, tmp[ 7 ], high_words( tmp[ 3 ] ) );
+    brw_MOV( p, tmp[ 4 ], low_words( tmp[ 2 ] ) );
+    brw_MOV( p, tmp[ 6 ], high_words( tmp[ 2 ] ) );
+    
+    brw_push_insn_state( p );
+    brw_set_mask_control( p, BRW_MASK_DISABLE );
+    brw_SHL( p, wtmp[ 2 ], wtmp[ 2 ], brw_imm_uw( 4 ) );
+    brw_SHL( p, wtmp[ 3 ], wtmp[ 3 ], brw_imm_uw( 4 ) );
+    brw_pop_insn_state( p );
+
+    brw_MUL( p, tmp[ 5 ], tmp[ 5 ], t );
+    brw_MUL( p, tmp[ 7 ], tmp[ 7 ], t );
+    brw_ADD( p, t, param[ 2 ], brw_imm_f( -1.0 ) );
+    brw_MUL( p, tmp[ 4 ], tmp[ 4 ], param[ 1 ] );
+    brw_MUL( p, tmp[ 6 ], tmp[ 6 ], param[ 1 ] );
+    
+    brw_ADD( p, x0y1, x0y1, tmp[ 5 ] );
+    brw_ADD( p, x1y1, x1y1, tmp[ 7 ] );
+    brw_ADD( p, x0y0, x0y0, tmp[ 4 ] );
+    brw_ADD( p, x1y0, x1y0, tmp[ 6 ] );
+    
+    /* z component */
+    brw_MOV( p, tmp[ 4 ], low_words( tmp[ 2 ] ) );
+    brw_MOV( p, tmp[ 5 ], low_words( tmp[ 3 ] ) );
+    brw_MOV( p, tmp[ 6 ], high_words( tmp[ 2 ] ) );
+    brw_MOV( p, tmp[ 7 ], high_words( tmp[ 3 ] ) );
+
+    brw_push_insn_state( p );
+    brw_set_mask_control( p, BRW_MASK_DISABLE );
+    brw_SHL( p, wtmp[ 2 ], wtmp[ 2 ], brw_imm_uw( 4 ) );
+    brw_SHL( p, wtmp[ 3 ], wtmp[ 3 ], brw_imm_uw( 4 ) );
+    brw_pop_insn_state( p );
+
+    brw_MUL( p, tmp[ 4 ], tmp[ 4 ], t );
+    brw_MUL( p, tmp[ 5 ], tmp[ 5 ], t );
+    brw_MUL( p, tmp[ 6 ], tmp[ 6 ], t );
+    brw_MUL( p, tmp[ 7 ], tmp[ 7 ], t );
+    /* prepare t for the w component (used below): w the first time through
+       the loop; w - 1 the second time) */
+    brw_set_predicate_control( p, BRW_PREDICATE_NORMAL );
+    brw_ADD( p, t, param[ 3 ], brw_imm_f( -1.0 ) );
+    p->current->header.predicate_inverse = 1;
+    brw_MOV( p, t, param[ 3 ] );
+    p->current->header.predicate_inverse = 0;
+    brw_set_predicate_control( p, BRW_PREDICATE_NONE );
+    
+    brw_ADD( p, x0y0, x0y0, tmp[ 4 ] );
+    brw_ADD( p, x0y1, x0y1, tmp[ 5 ] );
+    brw_ADD( p, x1y0, x1y0, tmp[ 6 ] );
+    brw_ADD( p, x1y1, x1y1, tmp[ 7 ] );
+
+    /* w component */
+    brw_MOV( p, tmp[ 4 ], low_words( tmp[ 2 ] ) );
+    brw_MOV( p, tmp[ 5 ], low_words( tmp[ 3 ] ) );
+    brw_MOV( p, tmp[ 6 ], high_words( tmp[ 2 ] ) );
+    brw_MOV( p, tmp[ 7 ], high_words( tmp[ 3 ] ) );
+
+    brw_MUL( p, tmp[ 4 ], tmp[ 4 ], t );
+    brw_MUL( p, tmp[ 5 ], tmp[ 5 ], t );
+    brw_MUL( p, tmp[ 6 ], tmp[ 6 ], t );
+    brw_MUL( p, tmp[ 7 ], tmp[ 7 ], t );
+    
+    brw_ADD( p, x0y0, x0y0, tmp[ 4 ] );
+    brw_ADD( p, x0y1, x0y1, tmp[ 5 ] );
+    brw_ADD( p, x1y0, x1y0, tmp[ 6 ] );
+    brw_ADD( p, x1y1, x1y1, tmp[ 7 ] );
+
+    /* Interpolate in the y dimension: */
+    brw_ADD( p, x0y1, x0y1, negate( x0y0 ) );
+    brw_ADD( p, x1y1, x1y1, negate( x1y0 ) );
+    brw_MUL( p, x0y1, x0y1, interp[ 1 ] );
+    brw_MUL( p, x1y1, x1y1, interp[ 1 ] );
+    brw_ADD( p, x0y0, x0y0, x0y1 );
+    brw_ADD( p, x1y0, x1y0, x1y1 );
+
+    /* And now in x.  The rear face is in tmp[ 0 ] (see above), so this
+       time put the front face in tmp[ 1 ] and we're nearly there... */
+    brw_ADD( p, x1y0, x1y0, negate( x0y0 ) );
+    brw_MUL( p, x1y0, x1y0, interp[ 0 ] );
+    brw_ADD( p, tmp[ 1 ], x0y0, x1y0 );
+
+    /* Another interpolation, in the z dimension: */
+    brw_ADD( p, tmp[ 1 ], tmp[ 1 ], negate( tmp[ 0 ] ) );    
+    brw_MUL( p, tmp[ 1 ], tmp[ 1 ], interp[ 2 ] );
+    brw_ADD( p, tmp[ 0 ], tmp[ 0 ], tmp[ 1 ] );
+
+    /* Exit the loop if we've computed both cubes... */
+    origin = p->nr_insn;
+    brw_push_insn_state( p );
+    brw_set_predicate_control( p, BRW_PREDICATE_NORMAL );
+    brw_set_mask_control( p, BRW_MASK_DISABLE );
+    brw_ADD( p, brw_ip_reg(), brw_ip_reg(), brw_imm_d( 0 ) );
+    brw_pop_insn_state( p );
+
+    /* Save the result for the w=0 case, and increment the w coordinate: */
+    brw_MOV( p, w0, tmp[ 0 ] );
+    brw_ADD( p, high_words( floors[ 1 ] ), high_words( floors[ 1 ] ),
+	     brw_imm_uw( 1 ) );
+
+    /* Loop around for the other cube.  Explicitly set the flag register
+       (unfortunately we must spend an extra instruction to do this: we
+       can't rely on a side effect of the previous MOV or ADD because
+       conditional modifiers which are normally true might be false in
+       exceptional circumstances, e.g. given a NaN input; the add to
+       brw_ip_reg() is not suitable because the IP is not an 8-vector). */
+    brw_push_insn_state( p );
+    brw_set_mask_control( p, BRW_MASK_DISABLE );
+    brw_MOV( p, brw_flag_reg(), brw_imm_uw( 0xFF ) );
+    brw_ADD( p, brw_ip_reg(), brw_ip_reg(),
+	     brw_imm_d( ( loop - p->nr_insn ) << 4 ) );
+    brw_pop_insn_state( p );
+
+    /* Patch the previous conditional branch now that we know the
+       destination address. */
+    brw_set_src1( p->store + origin,
+		  brw_imm_d( ( p->nr_insn - origin ) << 4 ) );
+
+    /* The very last interpolation. */
+    brw_ADD( p, tmp[ 0 ], tmp[ 0 ], negate( w0 ) );    
+    brw_MUL( p, tmp[ 0 ], tmp[ 0 ], interp[ 3 ] );
+    brw_ADD( p, tmp[ 0 ], tmp[ 0 ], w0 );
+
+    /* scale by pow( 2, -15 ), as described above */
+    brw_MUL( p, param[ 0 ], tmp[ 0 ], brw_imm_f( 0.000030517578125 ) );
+
+    release_tmps( c, mark );
+}
+
+static void emit_noise4( struct brw_wm_compile *c,
+			 struct prog_instruction *inst )
+{
+    struct brw_compile *p = &c->func;
+    struct brw_reg src0, src1, src2, src3, param0, param1, param2, param3, dst;
+    GLuint mask = inst->DstReg.WriteMask;
+    int i;
+    int mark = mark_tmps( c );
+
+    assert( mark == 0 );
+    
+    src0 = get_src_reg( c, inst->SrcReg, 0, 1 );
+    src1 = get_src_reg( c, inst->SrcReg, 1, 1 );
+    src2 = get_src_reg( c, inst->SrcReg, 2, 1 );
+    src3 = get_src_reg( c, inst->SrcReg, 3, 1 );
+
+    param0 = alloc_tmp( c );
+    param1 = alloc_tmp( c );
+    param2 = alloc_tmp( c );
+    param3 = alloc_tmp( c );
+
+    brw_MOV( p, param0, src0 );
+    brw_MOV( p, param1, src1 );
+    brw_MOV( p, param2, src2 );
+    brw_MOV( p, param3, src3 );
+
+    invoke_subroutine( c, SUB_NOISE4, noise4_sub );
+    
+    /* Fill in the result: */
+    brw_set_saturate( p, inst->SaturateMode == SATURATE_ZERO_ONE );
+    for (i = 0 ; i < 4; i++) {
+	if (mask & (1<<i)) {
+	    dst = get_dst_reg(c, inst, i, 1);
+	    brw_MOV( p, dst, param0 );
+	}
+    }
+    if( inst->SaturateMode == SATURATE_ZERO_ONE )
+	brw_set_saturate( p, 0 );
+    
+    release_tmps( c, mark );
+}
+    
 static void emit_wpos_xy(struct brw_wm_compile *c,
                 struct prog_instruction *inst)
 {
@@ -1996,8 +2422,9 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c)
 	    case OPCODE_NOISE3:
 		emit_noise3(c, inst);
 		break;
-	    /* case OPCODE_NOISE4: */
-		/* not yet implemented */
+	    case OPCODE_NOISE4:
+		emit_noise4(c, inst);
+		break;
 	    case OPCODE_TEX:
 		emit_tex(c, inst);
 		break;
diff --git a/src/mesa/drivers/dri/i965/brw_wm_iz.c b/src/mesa/drivers/dri/i965/brw_wm_iz.c
index 0bb5d5ba83..bd60ac9b31 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_iz.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_iz.c
@@ -52,70 +52,6 @@ const struct {
 {
  { P, 0, 0, 0, 0 }, 
  { P, 0, 0, 0, 0 }, 
- { C, 0, 1, 0, 0 }, 
- { C, 0, 1, 0, 0 }, 
- { C, 1, 1, 0, 0 }, 
- { C, 1, 1, 0, 0 }, 
- { C, 0, 1, 0, 0 }, 
- { C, 0, 1, 0, 0 }, 
- { C, 1, 1, 1, 0 }, 
- { C, 1, 1, 1, 0 }, 
- { C, 0, 1, 1, 0 }, 
- { C, 0, 1, 1, 0 }, 
- { C, 1, 1, 1, 0 }, 
- { C, 1, 1, 1, 0 }, 
- { C, 0, 1, 1, 0 }, 
- { C, 0, 1, 1, 0 }, 
- { P, 0, 0, 0, 0 }, 
- { P, 0, 0, 0, 0 }, 
- { C, 0, 1, 0, 0 }, 
- { C, 0, 1, 0, 0 }, 
- { C, 1, 1, 0, 0 }, 
- { C, 1, 1, 0, 0 }, 
- { C, 0, 1, 0, 0 }, 
- { C, 0, 1, 0, 0 }, 
- { C, 1, 1, 1, 0 }, 
- { C, 1, 1, 1, 0 }, 
- { C, 0, 1, 1, 0 }, 
- { C, 0, 1, 1, 0 }, 
- { C, 1, 1, 1, 0 }, 
- { C, 1, 1, 1, 0 }, 
- { C, 0, 1, 1, 0 }, 
- { C, 0, 1, 1, 0 }, 
- { C, 0, 0, 0, 1 }, 
- { C, 0, 0, 0, 1 }, 
- { C, 0, 1, 0, 1 }, 
- { C, 0, 1, 0, 1 }, 
- { C, 1, 1, 0, 1 }, 
- { C, 1, 1, 0, 1 }, 
- { C, 0, 1, 0, 1 }, 
- { C, 0, 1, 0, 1 }, 
- { C, 1, 1, 1, 1 }, 
- { C, 1, 1, 1, 1 }, 
- { C, 0, 1, 1, 1 }, 
- { C, 0, 1, 1, 1 }, 
- { C, 1, 1, 1, 1 }, 
- { C, 1, 1, 1, 1 }, 
- { C, 0, 1, 1, 1 }, 
- { C, 0, 1, 1, 1 }, 
- { C, 0, 0, 0, 1 }, 
- { C, 0, 0, 0, 1 }, 
- { C, 0, 1, 0, 1 }, 
- { C, 0, 1, 0, 1 }, 
- { C, 1, 1, 0, 1 }, 
- { C, 1, 1, 0, 1 }, 
- { C, 0, 1, 0, 1 }, 
- { C, 0, 1, 0, 1 }, 
- { C, 1, 1, 1, 1 }, 
- { C, 1, 1, 1, 1 }, 
- { C, 0, 1, 1, 1 }, 
- { C, 0, 1, 1, 1 }, 
- { C, 1, 1, 1, 1 }, 
- { C, 1, 1, 1, 1 }, 
- { C, 0, 1, 1, 1 }, 
- { C, 0, 1, 1, 1 }, 
- { P, 0, 0, 0, 0 }, 
- { P, 0, 0, 0, 0 }, 
  { P, 0, 0, 0, 0 }, 
  { P, 0, 0, 0, 0 }, 
  { P, 0, 0, 0, 0 }, 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass0.c b/src/mesa/drivers/dri/i965/brw_wm_pass0.c
index 205a7160d3..590cd946ec 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass0.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass0.c
@@ -51,6 +51,7 @@ static struct brw_wm_value *get_value( struct brw_wm_compile *c)
    return &c->vreg[c->nr_vreg++];
 }
 
+/** return pointer to a newly allocated instruction */
 static struct brw_wm_instruction *get_instruction( struct brw_wm_compile *c )
 {
    assert(c->nr_insns < BRW_WM_MAX_INSN);
@@ -60,6 +61,7 @@ static struct brw_wm_instruction *get_instruction( struct brw_wm_compile *c )
 /***********************************************************************
  */
 
+/** Init the "undef" register */
 static void pass0_init_undef( struct brw_wm_compile *c)
 {
    struct brw_wm_ref *ref = &c->undef_ref;
@@ -69,6 +71,7 @@ static void pass0_init_undef( struct brw_wm_compile *c)
    ref->prevuse = NULL;
 }
 
+/** Set a FP register to a value */
 static void pass0_set_fpreg_value( struct brw_wm_compile *c,
 				   GLuint file,
 				   GLuint idx,
@@ -83,6 +86,7 @@ static void pass0_set_fpreg_value( struct brw_wm_compile *c,
    c->pass0_fp_reg[file][idx][component] = ref;
 }
 
+/** Set a FP register to a ref */
 static void pass0_set_fpreg_ref( struct brw_wm_compile *c,
 				 GLuint file,
 				 GLuint idx,
@@ -115,12 +119,13 @@ static const struct brw_wm_ref *get_param_ref( struct brw_wm_compile *c,
       ref->value = &c->creg[i/16];
       ref->insn = 0;
       ref->prevuse = NULL;
-      
+
       return ref;
    }
 }
 
 
+/** Return a ref to a constant/literal value */
 static const struct brw_wm_ref *get_const_ref( struct brw_wm_compile *c,
 					       const GLfloat *constval )
 {
@@ -142,7 +147,7 @@ static const struct brw_wm_ref *get_const_ref( struct brw_wm_compile *c,
        */
       c->constref[i].constval = *constval;
       c->constref[i].ref = get_param_ref(c, constval);
-   
+
       return c->constref[i].ref;
    }
    else {
@@ -187,7 +192,7 @@ static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c,
 	 
 	 /* There's something really hokey about parameters parsed in
 	  * arb programs - they all end up in here, whether they be
-	  * state values, paramters or constants.  This duplicates the
+	  * state values, parameters or constants.  This duplicates the
 	  * structure above & also seems to subvert the limits set for
 	  * each type of constant/param.
 	  */ 
@@ -198,7 +203,7 @@ static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c,
 	     */
 	    ref = get_const_ref(c, &plist->ParameterValues[idx][component]);
 	    break;
-	    
+
 	 case PROGRAM_STATE_VAR:
 	 case PROGRAM_UNIFORM:
 	    /* These may change from run to run:
@@ -229,14 +234,13 @@ static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c,
 
 
 
-
 /***********************************************************************
  * Straight translation to internal instruction format
  */
 
 static void pass0_set_dst( struct brw_wm_compile *c,
-			   struct brw_wm_instruction *out,		     
-			   const struct prog_instruction *inst,		     
+			   struct brw_wm_instruction *out,
+			   const struct prog_instruction *inst,
 			   GLuint writemask )
 {
    const struct prog_dst_register *dst = &inst->DstReg;
@@ -245,18 +249,17 @@ static void pass0_set_dst( struct brw_wm_compile *c,
    for (i = 0; i < 4; i++) {
       if (writemask & (1<<i)) {
 	 out->dst[i] = get_value(c);
-
 	 pass0_set_fpreg_value(c, dst->File, dst->Index, i, out->dst[i]);
       }
    }
-   
+
    out->writemask = writemask;
 }
 
 
 static void pass0_set_dst_scalar( struct brw_wm_compile *c,
-				  struct brw_wm_instruction *out,		     
-				  const struct prog_instruction *inst,		     
+				  struct brw_wm_instruction *out,
+                                  const struct prog_instruction *inst,
 				  GLuint writemask )
 {
    if (writemask) {
@@ -282,7 +285,6 @@ static void pass0_set_dst_scalar( struct brw_wm_compile *c,
 }
 
 
-
 static const struct brw_wm_ref *get_fp_src_reg_ref( struct brw_wm_compile *c,
 						    struct prog_src_register src,
 						    GLuint i )
@@ -292,14 +294,13 @@ static const struct brw_wm_ref *get_fp_src_reg_ref( struct brw_wm_compile *c,
    static const GLfloat const_zero = 0.0;
    static const GLfloat const_one = 1.0;
 
-	 
    if (component == SWIZZLE_ZERO) 
       src_ref = get_const_ref(c, &const_zero);
    else if (component == SWIZZLE_ONE) 
       src_ref = get_const_ref(c, &const_one);
    else 
       src_ref = pass0_get_reg(c, src.File, src.Index, component);
-	 
+
    return src_ref;
 }
 
@@ -311,19 +312,19 @@ static struct brw_wm_ref *get_new_ref( struct brw_wm_compile *c,
 {
    const struct brw_wm_ref *ref = get_fp_src_reg_ref(c, src, i);
    struct brw_wm_ref *newref = get_ref(c);
-      
+
    newref->value = ref->value;
    newref->hw_reg = ref->hw_reg;
 
-   if (insn) { 
+   if (insn) {
       newref->insn = insn - c->instruction;
       newref->prevuse = newref->value->lastuse;
       newref->value->lastuse = newref;
    }
 
-   if (src.NegateBase & (1<<i)) 
+   if (src.NegateBase & (1<<i))
       newref->hw_reg.negate ^= 1;
-	    
+
    if (src.Abs) {
       newref->hw_reg.negate = 0;
       newref->hw_reg.abs = 1;
@@ -333,7 +334,6 @@ static struct brw_wm_ref *get_new_ref( struct brw_wm_compile *c,
 }
 
 
-
 static struct brw_wm_instruction *translate_insn( struct brw_wm_compile *c,
 						  const struct prog_instruction *inst )
 {
@@ -379,14 +379,22 @@ static void pass0_precalc_mov( struct brw_wm_compile *c,
 {
    const struct prog_dst_register *dst = &inst->DstReg;
    GLuint writemask = inst->DstReg.WriteMask;
+   struct brw_wm_ref *refs[4];
    GLuint i;
 
    /* Get the effect of a MOV by manipulating our register table:
+    * First get all refs, then assign refs.  This ensures that "in-place"
+    * swizzles such as:
+    *   MOV t, t.xxyx
+    * are handled correctly.  Previously, these two steps were done in
+    * one loop and the above case was incorrectly handled.
     */
    for (i = 0; i < 4; i++) {
-      if (writemask & (1<<i)) {	    
-	 pass0_set_fpreg_ref( c, dst->File, dst->Index, i, 
-			      get_new_ref(c, inst->SrcReg[0], i, NULL));
+      refs[i] = get_new_ref(c, inst->SrcReg[0], i, NULL);
+   }
+   for (i = 0; i < 4; i++) {
+      if (writemask & (1 << i)) {	    
+         pass0_set_fpreg_ref( c, dst->File, dst->Index, i, refs[i]);
       }
    }
 }
@@ -418,6 +426,7 @@ static void pass0_init_payload( struct brw_wm_compile *c )
 			     &c->payload.input_interp[i] );      
 }
 
+
 /***********************************************************************
  * PASS 0
  *
@@ -440,7 +449,6 @@ void brw_wm_pass0( struct brw_wm_compile *c )
    for (insn = 0; insn < c->nr_fp_insns; insn++) {
       const struct prog_instruction *inst = &c->prog_instructions[insn];
 
-
       /* Optimize away moves, otherwise emit translated instruction:
        */      
       switch (inst->Opcode) {
@@ -453,8 +461,6 @@ void brw_wm_pass0( struct brw_wm_compile *c )
 	    translate_insn(c, inst);
 	 }
 	 break;
-	 
-
       default:
 	 translate_insn(c, inst);
 	 break;
@@ -465,4 +471,3 @@ void brw_wm_pass0( struct brw_wm_compile *c )
       brw_wm_print_program(c, "pass0");
    }
 }
-
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass1.c b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
index f6f3a38e9e..6eaed8a665 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass1.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass1.c
@@ -58,7 +58,8 @@ static void unlink_ref(struct brw_wm_ref *ref)
 
    if (ref == value->lastuse) {
       value->lastuse = ref->prevuse;
-   } else {
+   }
+   else {
       struct brw_wm_ref *i = value->lastuse;
       while (i->prevuse != ref) i = i->prevuse;
       i->prevuse = ref->prevuse;
@@ -75,8 +76,9 @@ static void track_arg(struct brw_wm_compile *c,
    for (i = 0; i < 4; i++) {
       struct brw_wm_ref *ref = inst->src[arg][i];
       if (ref) {
-	 if (readmask & (1<<i)) 
+	 if (readmask & (1<<i)) {
 	    ref->value->contributes_to_output = 1;
+         }
 	 else {
 	    unlink_ref(ref);
 	    inst->src[arg][i] = NULL;
@@ -88,15 +90,21 @@ static void track_arg(struct brw_wm_compile *c,
 static GLuint get_texcoord_mask( GLuint tex_idx )
 {
    switch (tex_idx) {
-   case TEXTURE_1D_INDEX: return WRITEMASK_X;
-   case TEXTURE_2D_INDEX: return WRITEMASK_XY;
-   case TEXTURE_3D_INDEX: return WRITEMASK_XYZ;
-   case TEXTURE_CUBE_INDEX: return WRITEMASK_XYZ;
-   case TEXTURE_RECT_INDEX: return WRITEMASK_XY;
+   case TEXTURE_1D_INDEX:
+      return WRITEMASK_X;
+   case TEXTURE_2D_INDEX:
+      return WRITEMASK_XY;
+   case TEXTURE_3D_INDEX:
+      return WRITEMASK_XYZ;
+   case TEXTURE_CUBE_INDEX:
+      return WRITEMASK_XYZ;
+   case TEXTURE_RECT_INDEX:
+      return WRITEMASK_XY;
    default: return 0;
    }
 }
 
+
 /* Step two: Basically this is dead code elimination.  
  *
  * Iterate backwards over instructions, noting which values
@@ -273,6 +281,3 @@ void brw_wm_pass1( struct brw_wm_compile *c )
       brw_wm_print_program(c, "pass1");
    }
 }
-
-
-
diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass2.c b/src/mesa/drivers/dri/i965/brw_wm_pass2.c
index 6fca9ad220..780edbc42e 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_pass2.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_pass2.c
@@ -84,7 +84,7 @@ static void init_registers( struct brw_wm_compile *c )
    for (j = 0; j < c->nr_creg; j++) 
       prealloc_reg(c, &c->creg[j], i++);
 
-   for (j = 0; j < FRAG_ATTRIB_MAX; j++) 
+   for (j = 0; j < FRAG_ATTRIB_MAX; j++) {
       if (inputs & (1<<j)) {
 	 /* index for vs output and ps input are not the same 
 	    in shader varying */
@@ -96,6 +96,7 @@ static void init_registers( struct brw_wm_compile *c )
 	 nr_interp_regs++;
 	 prealloc_reg(c, &c->payload.input_interp[index], i++);
       }
+   }
 
    assert(nr_interp_regs >= 1);
 
@@ -120,7 +121,7 @@ static void update_register_usage(struct brw_wm_compile *c,
       /* Only search those which can change:
        */
       if (grf->nextuse < thisinsn) {
-	 struct brw_wm_ref *ref = grf->value->lastuse;
+	 const struct brw_wm_ref *ref = grf->value->lastuse;
 
 	 /* Has last use of value been passed?
 	  */
@@ -148,7 +149,7 @@ static void spill_value(struct brw_wm_compile *c,
    /* Allocate a spill slot.  Note that allocations start from 0x40 -
     * the first slot is reserved to mean "undef" in brw_wm_emit.c
     */
-   if (!value->spill_slot) {  
+   if (!value->spill_slot) {
       c->last_scratch += 0x40;	
       value->spill_slot = c->last_scratch;
    }
@@ -189,7 +190,7 @@ static GLuint search_contiguous_regs(struct brw_wm_compile *c,
 	 if (grf[i+j].nextuse < group_nextuse)
 	    group_nextuse = grf[i+j].nextuse;
       }
-	 
+
       if (group_nextuse > furthest) {
 	 furthest = group_nextuse;
 	 reg = i;
@@ -197,7 +198,7 @@ static GLuint search_contiguous_regs(struct brw_wm_compile *c,
    }
 
    assert(furthest != thisinsn);
-   
+
    /* Any non-empty regs will need to be spilled:
     */
    for (j = 0; j < nr; j++) 
@@ -243,7 +244,7 @@ static void alloc_contiguous_dest(struct brw_wm_compile *c,
 
 static void load_args(struct brw_wm_compile *c, 
 		      struct brw_wm_instruction *inst)
-{   
+{
    GLuint thisinsn = inst - c->instruction;
    GLuint i,j;
 
@@ -258,17 +259,17 @@ static void load_args(struct brw_wm_compile *c,
 		* register allocation and mark the ref as requiring a fill.
 		*/
 	       GLuint reg = search_contiguous_regs(c, 1, thisinsn);
-            
+
 	       c->pass2_grf[reg].value = ref->value;
 	       c->pass2_grf[reg].nextuse = thisinsn;
-	    
+
 	       ref->value->resident = &c->pass2_grf[reg];
 
 	       /* Note that a fill is required:
 		*/
 	       ref->unspill_reg = reg*2;
 	    }
-	    
+
 	    /* Adjust the hw_reg to point at the value's current location:
 	     */
 	    assert(ref->value == ref->value->resident->value);
@@ -294,7 +295,7 @@ void brw_wm_pass2( struct brw_wm_compile *c )
 
    for (insn = 0; insn < c->nr_insns; insn++) {
       struct brw_wm_instruction *inst = &c->instruction[insn];
-      
+
       /* Update registers' nextuse values:
        */
       update_register_usage(c, insn);
@@ -322,11 +323,11 @@ void brw_wm_pass2( struct brw_wm_compile *c )
 	 break;
       }
 
-      if (TEST_DST_SPILLS && inst->opcode != WM_PIXELXY)
+      if (TEST_DST_SPILLS && inst->opcode != WM_PIXELXY) {
 	 for (i = 0; i < 4; i++)	
 	    if (inst->dst[i])
 	       spill_value(c, inst->dst[i]);
-
+      }
    }
 
    if (INTEL_DEBUG & DEBUG_WM) {
@@ -339,6 +340,3 @@ void brw_wm_pass2( struct brw_wm_compile *c )
        brw_wm_print_program(c, "pass2/done");
    }
 }
-
-
-
diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
index f12ef47a7d..b6dac0d698 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
@@ -220,14 +220,15 @@ static void
 brw_wm_sampler_populate_key(struct brw_context *brw,
 			    struct wm_sampler_key *key)
 {
+   GLcontext *ctx = &brw->intel.ctx;
    int unit;
 
    memset(key, 0, sizeof(*key));
 
    for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) {
-      if (brw->attribs.Texture->Unit[unit]._ReallyEnabled) {
+      if (ctx->Texture.Unit[unit]._ReallyEnabled) {
 	 struct wm_sampler_entry *entry = &key->sampler[unit];
-	 struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[unit];
+	 struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
 	 struct gl_texture_object *texObj = texUnit->_Current;
 	 struct intel_texture_object *intelObj = intel_texture_object(texObj);
 	 struct gl_texture_image *firstImage =
@@ -244,7 +245,7 @@ brw_wm_sampler_populate_key(struct brw_context *brw,
 	 entry->minfilter = texObj->MinFilter;
 	 entry->magfilter = texObj->MagFilter;
 	 entry->comparemode = texObj->CompareMode;
-    entry->comparefunc = texObj->CompareFunc;
+         entry->comparefunc = texObj->CompareFunc;
 
 	 dri_bo_unreference(brw->wm.sdc_bo[unit]);
 	 if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) {
@@ -274,6 +275,7 @@ brw_wm_sampler_populate_key(struct brw_context *brw,
  */
 static void upload_wm_samplers( struct brw_context *brw )
 {
+   GLcontext *ctx = &brw->intel.ctx;
    struct wm_sampler_key key;
    int i;
 
@@ -317,7 +319,7 @@ static void upload_wm_samplers( struct brw_context *brw )
 
       /* Emit SDC relocations */
       for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
-	 if (!brw->attribs.Texture->Unit[i]._ReallyEnabled)
+	 if (!ctx->Texture.Unit[i]._ReallyEnabled)
 	    continue;
 
 	 dri_bo_emit_reloc(brw->wm.sampler_bo,
diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c
index fd461618bc..3c3b3473d6 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_state.c
@@ -60,6 +60,7 @@ struct brw_wm_unit_key {
 static void
 wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key)
 {
+   GLcontext *ctx = &brw->intel.ctx;
    const struct gl_fragment_program *fp = brw->fragment_program;
    struct intel_context *intel = &brw->intel;
 
@@ -88,14 +89,14 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key)
    /* BRW_NEW_CURBE_OFFSETS */
    key->curbe_offset = brw->curbe.wm_start;
 
-   /* CACHE_NEW_SURFACE */
+   /* BRW_NEW_NR_SURFACEs */
    key->nr_surfaces = brw->wm.nr_surfaces;
 
    /* CACHE_NEW_SAMPLER */
    key->sampler_count = brw->wm.sampler_count;
 
    /* _NEW_POLYGONSTIPPLE */
-   key->polygon_stipple = brw->attribs.Polygon->StippleFlag;
+   key->polygon_stipple = ctx->Polygon.StippleFlag;
 
    /* BRW_NEW_FRAGMENT_PROGRAM */
    key->uses_depth = (fp->Base.InputsRead & (1 << FRAG_ATTRIB_WPOS)) != 0;
@@ -105,19 +106,19 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key)
       (fp->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR)) != 0;
 
    /* _NEW_COLOR */
-   key->uses_kill = fp->UsesKill || brw->attribs.Color->AlphaEnabled;
+   key->uses_kill = fp->UsesKill || ctx->Color.AlphaEnabled;
    key->is_glsl = brw_wm_is_glsl(fp);
 
    /* XXX: This needs a flag to indicate when it changes. */
    key->stats_wm = intel->stats_wm;
 
    /* _NEW_LINE */
-   key->line_stipple = brw->attribs.Line->StippleFlag;
+   key->line_stipple = ctx->Line.StippleFlag;
 
    /* _NEW_POLYGON */
-   key->offset_enable = brw->attribs.Polygon->OffsetFill;
-   key->offset_units = brw->attribs.Polygon->OffsetUnits;
-   key->offset_factor = brw->attribs.Polygon->OffsetFactor;
+   key->offset_enable = ctx->Polygon.OffsetFill;
+   key->offset_units = ctx->Polygon.OffsetUnits;
+   key->offset_factor = ctx->Polygon.OffsetFactor;
 }
 
 static dri_bo *
@@ -281,10 +282,9 @@ const struct brw_tracked_state brw_wm_unit = {
 
       .brw = (BRW_NEW_FRAGMENT_PROGRAM | 
 	      BRW_NEW_CURBE_OFFSETS |
-	      BRW_NEW_LOCK),
+	      BRW_NEW_NR_SURFACES),
 
-      .cache = (CACHE_NEW_SURFACE | 
-		CACHE_NEW_WM_PROG | 
+      .cache = (CACHE_NEW_WM_PROG |
 		CACHE_NEW_SAMPLER)
    },
    .prepare = upload_wm_unit,
diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
index 3790b50c97..d70f9c646c 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
@@ -192,21 +192,27 @@ brw_create_texture_surface( struct brw_context *brw,
    if (key->bo) 
       surf.ss0.surface_format = translate_tex_format(key->format, key->depthmode);
    else {
-     switch(key->depth) {
-     case 32: surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; break;
-     default:
-     case 24: surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8X8_UNORM; break;
-     case 16: surf.ss0.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM; break;
-     }
+      switch (key->depth) {
+      case 32:
+         surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM;
+         break;
+      default:
+      case 24:
+         surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8X8_UNORM;
+         break;
+      case 16:
+         surf.ss0.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM;
+         break;
+      }
    }
 
    /* This is ok for all textures with channel width 8bit or less:
     */
 /*    surf.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */
    if (key->bo)
-     surf.ss1.base_addr = key->bo->offset; /* reloc */
+      surf.ss1.base_addr = key->bo->offset; /* reloc */
    else
-     surf.ss1.base_addr = key->offset;
+      surf.ss1.base_addr = key->offset;
 
    surf.ss2.mip_count = key->last_level - key->first_level;
    surf.ss2.width = key->width - 1;
@@ -247,7 +253,7 @@ static void
 brw_update_texture_surface( GLcontext *ctx, GLuint unit )
 {
    struct brw_context *brw = brw_context(ctx);
-   struct gl_texture_object *tObj = brw->attribs.Texture->Unit[unit]._Current;
+   struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current;
    struct intel_texture_object *intelObj = intel_texture_object(tObj);
    struct gl_texture_image *firstImage = tObj->Image[0][intelObj->firstLevel];
    struct brw_wm_surface_key key;
@@ -295,6 +301,7 @@ static void
 brw_update_region_surface(struct brw_context *brw, struct intel_region *region,
 			  unsigned int unit, GLboolean cached)
 {
+   GLcontext *ctx = &brw->intel.ctx;
    dri_bo *region_bo = NULL;
    struct {
       unsigned int surface_type;
@@ -327,10 +334,10 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region,
       key.height = 1;
       key.cpp = 4;
    }
-   memcpy(key.color_mask, brw->attribs.Color->ColorMask,
+   memcpy(key.color_mask, ctx->Color.ColorMask,
 	  sizeof(key.color_mask));
-   key.color_blend = (!brw->attribs.Color->_LogicOpEnabled &&
-		      brw->attribs.Color->BlendEnabled);
+   key.color_blend = (!ctx->Color._LogicOpEnabled &&
+		      ctx->Color.BlendEnabled);
 
    dri_bo_unreference(brw->wm.surf_bo[unit]);
    brw->wm.surf_bo[unit] = NULL;
@@ -438,6 +445,7 @@ static void prepare_wm_surfaces(struct brw_context *brw )
    GLcontext *ctx = &brw->intel.ctx;
    struct intel_context *intel = &brw->intel;
    GLuint i;
+   int old_nr_surfaces;
 
    if (brw->state.nr_draw_regions  > 1) {
       for (i = 0; i < brw->state.nr_draw_regions; i++) {
@@ -448,10 +456,11 @@ static void prepare_wm_surfaces(struct brw_context *brw )
       brw_update_region_surface(brw, brw->state.draw_regions[0], 0, GL_TRUE);
    }
 
+   old_nr_surfaces = brw->wm.nr_surfaces;
    brw->wm.nr_surfaces = MAX_DRAW_BUFFERS;
 
    for (i = 0; i < BRW_MAX_TEX_UNIT; i++) {
-      struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[i];
+      struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
 
       /* _NEW_TEXTURE, BRW_NEW_TEXDATA */
       if(texUnit->_ReallyEnabled) {
@@ -473,6 +482,9 @@ static void prepare_wm_surfaces(struct brw_context *brw )
 
    dri_bo_unreference(brw->wm.bind_bo);
    brw->wm.bind_bo = brw_wm_get_binding_table(brw);
+
+   if (brw->wm.nr_surfaces != old_nr_surfaces)
+      brw->state.dirty.brw |= BRW_NEW_NR_SURFACES;
 }
 
 
diff --git a/src/mesa/drivers/dri/i965/intel_clear.c b/src/mesa/drivers/dri/i965/intel_clear.c
new file mode 120000
index 0000000000..9a2a742a0d
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/intel_clear.c
@@ -0,0 +1 @@
+../intel/intel_clear.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i965/intel_depthstencil.c b/src/mesa/drivers/dri/i965/intel_depthstencil.c
deleted file mode 120000
index 4ac4ae690a..0000000000
--- a/src/mesa/drivers/dri/i965/intel_depthstencil.c
+++ /dev/null
@@ -1 +0,0 @@
-../intel/intel_depthstencil.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i965/intel_extensions.c b/src/mesa/drivers/dri/i965/intel_extensions.c
new file mode 120000
index 0000000000..a2f3e8cd20
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/intel_extensions.c
@@ -0,0 +1 @@
+../intel/intel_extensions.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i965/intel_state.c b/src/mesa/drivers/dri/i965/intel_state.c
index 67ef5f78c1..519672fc35 100644..120000
--- a/src/mesa/drivers/dri/i965/intel_state.c
+++ b/src/mesa/drivers/dri/i965/intel_state.c
@@ -1,225 +1 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-
-#include "main/glheader.h"
-#include "main/context.h"
-#include "main/macros.h"
-#include "main/enums.h"
-#include "main/colormac.h"
-#include "main/dd.h"
-
-#include "intel_screen.h"
-#include "intel_context.h"
-#include "intel_regions.h"
-#include "swrast/swrast.h"
-
-int intel_translate_shadow_compare_func( GLenum func )
-{
-   switch(func) {
-   case GL_NEVER: 
-       return COMPAREFUNC_ALWAYS; 
-   case GL_LESS: 
-       return COMPAREFUNC_LEQUAL; 
-   case GL_LEQUAL: 
-       return COMPAREFUNC_LESS;
-   case GL_GREATER: 
-       return COMPAREFUNC_GEQUAL; 
-   case GL_GEQUAL: 
-      return COMPAREFUNC_GREATER; 
-   case GL_NOTEQUAL: 
-      return COMPAREFUNC_EQUAL; 
-   case GL_EQUAL: 
-      return COMPAREFUNC_NOTEQUAL; 
-   case GL_ALWAYS: 
-       return COMPAREFUNC_NEVER; 
-   }
-
-   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func);
-   return COMPAREFUNC_NEVER; 
-}
-
-int intel_translate_compare_func( GLenum func )
-{
-   switch(func) {
-   case GL_NEVER: 
-      return COMPAREFUNC_NEVER; 
-   case GL_LESS: 
-      return COMPAREFUNC_LESS; 
-   case GL_LEQUAL: 
-      return COMPAREFUNC_LEQUAL; 
-   case GL_GREATER: 
-      return COMPAREFUNC_GREATER; 
-   case GL_GEQUAL: 
-      return COMPAREFUNC_GEQUAL; 
-   case GL_NOTEQUAL: 
-      return COMPAREFUNC_NOTEQUAL; 
-   case GL_EQUAL: 
-      return COMPAREFUNC_EQUAL; 
-   case GL_ALWAYS: 
-      return COMPAREFUNC_ALWAYS; 
-   }
-
-   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func);
-   return COMPAREFUNC_ALWAYS; 
-}
-
-int intel_translate_stencil_op( GLenum op )
-{
-   switch(op) {
-   case GL_KEEP: 
-      return STENCILOP_KEEP; 
-   case GL_ZERO: 
-      return STENCILOP_ZERO; 
-   case GL_REPLACE: 
-      return STENCILOP_REPLACE; 
-   case GL_INCR: 
-      return STENCILOP_INCRSAT;
-   case GL_DECR: 
-      return STENCILOP_DECRSAT;
-   case GL_INCR_WRAP:
-      return STENCILOP_INCR; 
-   case GL_DECR_WRAP:
-      return STENCILOP_DECR; 
-   case GL_INVERT: 
-      return STENCILOP_INVERT; 
-   default: 
-      return STENCILOP_ZERO;
-   }
-}
-
-int intel_translate_blend_factor( GLenum factor )
-{
-   switch(factor) {
-   case GL_ZERO: 
-      return BLENDFACT_ZERO; 
-   case GL_SRC_ALPHA: 
-      return BLENDFACT_SRC_ALPHA; 
-   case GL_ONE: 
-      return BLENDFACT_ONE; 
-   case GL_SRC_COLOR: 
-      return BLENDFACT_SRC_COLR; 
-   case GL_ONE_MINUS_SRC_COLOR: 
-      return BLENDFACT_INV_SRC_COLR; 
-   case GL_DST_COLOR: 
-      return BLENDFACT_DST_COLR; 
-   case GL_ONE_MINUS_DST_COLOR: 
-      return BLENDFACT_INV_DST_COLR; 
-   case GL_ONE_MINUS_SRC_ALPHA:
-      return BLENDFACT_INV_SRC_ALPHA; 
-   case GL_DST_ALPHA: 
-      return BLENDFACT_DST_ALPHA; 
-   case GL_ONE_MINUS_DST_ALPHA:
-      return BLENDFACT_INV_DST_ALPHA; 
-   case GL_SRC_ALPHA_SATURATE: 
-      return BLENDFACT_SRC_ALPHA_SATURATE;
-   case GL_CONSTANT_COLOR:
-      return BLENDFACT_CONST_COLOR; 
-   case GL_ONE_MINUS_CONSTANT_COLOR:
-      return BLENDFACT_INV_CONST_COLOR;
-   case GL_CONSTANT_ALPHA:
-      return BLENDFACT_CONST_ALPHA; 
-   case GL_ONE_MINUS_CONSTANT_ALPHA:
-      return BLENDFACT_INV_CONST_ALPHA;
-   }
-   
-   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, factor);
-   return BLENDFACT_ZERO;
-}
-
-int intel_translate_logic_op( GLenum opcode )
-{
-   switch(opcode) {
-   case GL_CLEAR: 
-      return LOGICOP_CLEAR; 
-   case GL_AND: 
-      return LOGICOP_AND; 
-   case GL_AND_REVERSE: 
-      return LOGICOP_AND_RVRSE; 
-   case GL_COPY: 
-      return LOGICOP_COPY; 
-   case GL_COPY_INVERTED: 
-      return LOGICOP_COPY_INV; 
-   case GL_AND_INVERTED: 
-      return LOGICOP_AND_INV; 
-   case GL_NOOP: 
-      return LOGICOP_NOOP; 
-   case GL_XOR: 
-      return LOGICOP_XOR; 
-   case GL_OR: 
-      return LOGICOP_OR; 
-   case GL_OR_INVERTED: 
-      return LOGICOP_OR_INV; 
-   case GL_NOR: 
-      return LOGICOP_NOR; 
-   case GL_EQUIV: 
-      return LOGICOP_EQUIV; 
-   case GL_INVERT: 
-      return LOGICOP_INV; 
-   case GL_OR_REVERSE: 
-      return LOGICOP_OR_RVRSE; 
-   case GL_NAND: 
-      return LOGICOP_NAND; 
-   case GL_SET: 
-      return LOGICOP_SET; 
-   default:
-      return LOGICOP_SET;
-   }
-}
-
-
-static void intelClearColor(GLcontext *ctx, const GLfloat color[4])
-{
-   struct intel_context *intel = intel_context(ctx);
-
-   UNCLAMPED_FLOAT_TO_RGBA_CHAN(intel->clear_chan, color);
-
-   intel->ClearColor8888 = INTEL_PACKCOLOR8888(intel->clear_chan[0],
-					       intel->clear_chan[1],
-					       intel->clear_chan[2],
-					       intel->clear_chan[3]);
-   intel->ClearColor565 = INTEL_PACKCOLOR565(intel->clear_chan[0],
-					     intel->clear_chan[1],
-					     intel->clear_chan[2]);
-}
-
-
-
-/* Fallback to swrast for select and feedback.
- */
-static void intelRenderMode( GLcontext *ctx, GLenum mode )
-{
-   struct intel_context *intel = intel_context(ctx);
-   FALLBACK( intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER) );
-}
-
-
-void intelInitStateFuncs( struct dd_function_table *functions )
-{
-   functions->RenderMode = intelRenderMode;
-   functions->ClearColor = intelClearColor;
-}
+../intel/intel_state.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i965/intel_swapbuffers.c b/src/mesa/drivers/dri/i965/intel_swapbuffers.c
new file mode 120000
index 0000000000..148d5215aa
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/intel_swapbuffers.c
@@ -0,0 +1 @@
+../intel/intel_swapbuffers.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.h b/src/mesa/drivers/dri/intel/intel_batchbuffer.h
index 8129996979..51579df09e 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.h
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.h
@@ -55,6 +55,12 @@ struct intel_batchbuffer
 
    GLuint size;
 
+   /** Tracking of BEGIN_BATCH()/OUT_BATCH()/ADVANCE_BATCH() debugging */
+   struct {
+      GLuint total;
+      GLubyte *start_ptr;
+   } emit;
+
    GLuint dirty_state;
 };
 
@@ -143,9 +149,12 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch,
 
 #define BEGIN_BATCH(n, cliprect_mode) do {				\
    intel_batchbuffer_require_space(intel->batch, (n)*4, cliprect_mode); \
+   assert(intel->batch->emit.start_ptr == NULL);			\
+   intel->batch->emit.total = (n) * 4;					\
+   intel->batch->emit.start_ptr = intel->batch->ptr;			\
 } while (0)
 
-#define OUT_BATCH(d)  intel_batchbuffer_emit_dword(intel->batch, d)
+#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d)
 
 #define OUT_RELOC(buf, read_domains, write_domain, delta) do {		\
    assert((delta) >= 0);						\
@@ -153,7 +162,16 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch,
 				read_domains, write_domain, delta);	\
 } while (0)
 
-#define ADVANCE_BATCH() do { } while(0)
+#define ADVANCE_BATCH() do {						\
+   unsigned int _n = intel->batch->ptr - intel->batch->emit.start_ptr;	\
+   assert(intel->batch->emit.start_ptr != NULL);			\
+   if (_n != intel->batch->emit.total) {				\
+      fprintf(stderr, "ADVANCE_BATCH: %d of %d dwords emitted\n",	\
+	      _n, intel->batch->emit.total);				\
+      abort();								\
+   }									\
+   intel->batch->emit.start_ptr = NULL;					\
+} while(0)
 
 
 static INLINE void
diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c
index e1046f4a5d..208f90c0ab 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.c
+++ b/src/mesa/drivers/dri/intel/intel_blit.c
@@ -40,6 +40,7 @@
 #include "intel_reg.h"
 #include "intel_regions.h"
 #include "intel_batchbuffer.h"
+#include "intel_chipset.h"
 
 #define FILE_DEBUG_FLAG DEBUG_BLIT
 
@@ -359,51 +360,24 @@ intelEmitCopyBlit(struct intel_context *intel,
       return;
    }
 
-   /* Initial y values don't seem to work with negative pitches.  If
-    * we adjust the offsets manually (below), it seems to work fine.
-    *
-    * On the other hand, if we always adjust, the hardware doesn't
-    * know which blit directions to use, so overlapping copypixels get
-    * the wrong result.
-    */
-   if (dst_pitch > 0 && src_pitch > 0) {
-      assert(dst_x < dst_x2);
-      assert(dst_y < dst_y2);
-
-      BEGIN_BATCH(8, NO_LOOP_CLIPRECTS);
-      OUT_BATCH(CMD);
-      OUT_BATCH(BR13 | dst_pitch);
-      OUT_BATCH((dst_y << 16) | dst_x);
-      OUT_BATCH((dst_y2 << 16) | dst_x2);
-      OUT_RELOC(dst_buffer,
-		I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-		dst_offset);
-      OUT_BATCH((src_y << 16) | src_x);
-      OUT_BATCH(src_pitch);
-      OUT_RELOC(src_buffer,
-		I915_GEM_DOMAIN_RENDER, 0,
-		src_offset);
-      ADVANCE_BATCH();
-   }
-   else {
-      assert(dst_x < dst_x2);
-      assert(h > 0);
-
-      BEGIN_BATCH(8, NO_LOOP_CLIPRECTS);
-      OUT_BATCH(CMD);
-      OUT_BATCH(BR13 | ((uint16_t)dst_pitch));
-      OUT_BATCH((0 << 16) | dst_x);
-      OUT_BATCH((h << 16) | dst_x2);
-      OUT_RELOC(dst_buffer,
-		I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-		dst_offset + dst_y * dst_pitch);
-      OUT_BATCH((0 << 16) | src_x);
-      OUT_BATCH(src_pitch);
-      OUT_RELOC(src_buffer,
-		I915_GEM_DOMAIN_RENDER, 0,
-		src_offset + src_y * src_pitch);
-      ADVANCE_BATCH();
-   }
+   assert(dst_x < dst_x2);
+   assert(dst_y < dst_y2);
+
+   BEGIN_BATCH(8, NO_LOOP_CLIPRECTS);
+   OUT_BATCH(CMD);
+   OUT_BATCH(BR13 | (uint16_t)dst_pitch);
+   OUT_BATCH((dst_y << 16) | dst_x);
+   OUT_BATCH((dst_y2 << 16) | dst_x2);
+   OUT_RELOC(dst_buffer,
+	     I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+	     dst_offset);
+   OUT_BATCH((src_y << 16) | src_x);
+   OUT_BATCH((uint16_t)src_pitch);
+   OUT_RELOC(src_buffer,
+	     I915_GEM_DOMAIN_RENDER, 0,
+	     src_offset);
+   ADVANCE_BATCH();
+
    intel_batchbuffer_emit_mi_flush(intel->batch);
 }
 
@@ -570,7 +544,6 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
                   _mesa_debug(ctx, "hardware blit clear buf %d rb id %d\n",
                   buf, irb->Base.Name);
                 */
-	       intel_wait_flips(intel);
 
                assert(b.x1 < b.x2);
                assert(b.y1 < b.y2);
@@ -626,7 +599,7 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
    intel_batchbuffer_require_space( intel->batch,
 				    (8 * 4) +
 				    (3 * 4) +
-				    dwords,
+				    dwords * 4,
 				    REFERENCES_CLIPRECTS );
 
    opcode = XY_SETUP_BLT_CMD;
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c
index f8f009c6a3..7f2144abd4 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.c
+++ b/src/mesa/drivers/dri/intel/intel_buffers.c
@@ -25,41 +25,14 @@
  * 
  **************************************************************************/
 
-#include "intel_screen.h"
 #include "intel_context.h"
-#include "intel_blit.h"
 #include "intel_buffers.h"
-#include "intel_chipset.h"
-#include "intel_depthstencil.h"
 #include "intel_fbo.h"
 #include "intel_regions.h"
 #include "intel_batchbuffer.h"
-#include "intel_reg.h"
-#include "main/context.h"
 #include "main/framebuffer.h"
-#include "swrast/swrast.h"
-#include "utils.h"
 #include "drirenderbuffer.h"
-#include "vblank.h"
-#include "i915_drm.h"
 
-/* This block can be removed when libdrm >= 2.3.1 is required */
-
-#ifndef DRM_IOCTL_I915_FLIP
-
-#define DRM_VBLANK_FLIP 0x8000000
-
-typedef struct drm_i915_flip {
-   int pipes;
-} drm_i915_flip_t;
-
-#undef DRM_IOCTL_I915_FLIP
-#define DRM_IOCTL_I915_FLIP DRM_IOW(DRM_COMMAND_BASE + DRM_I915_FLIP, \
-				    drm_i915_flip_t)
-
-#endif
-
-#define FILE_DEBUG_FLAG DEBUG_BLIT
 
 /**
  * XXX move this into a new dri/common/cliprects.c file.
@@ -130,7 +103,6 @@ intel_get_cliprects(struct intel_context *intel,
 		    int *x_off, int *y_off)
 {
    __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
 
    if (intel->constant_cliprect) {
       /* FBO or DRI2 rendering, which can just use the fb's size. */
@@ -143,8 +115,7 @@ intel_get_cliprects(struct intel_context *intel,
       *num_cliprects = 1;
       *x_off = 0;
       *y_off = 0;
-   } else if (intel->front_cliprects ||
-       intel_fb->pf_active || dPriv->numBackClipRects == 0) {
+   } else if (intel->front_cliprects || dPriv->numBackClipRects == 0) {
       /* use the front clip rects */
       *cliprects = dPriv->pClipRects;
       *num_cliprects = dPriv->numClipRects;
@@ -160,641 +131,6 @@ intel_get_cliprects(struct intel_context *intel,
    }
 }
 
-static void
-intelUpdatePageFlipping(struct intel_context *intel,
-			GLint areaA, GLint areaB)
-{
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
-   GLboolean pf_active;
-   GLint pf_planes;
-
-   /* Update page flipping info */
-   pf_planes = 0;
-
-   if (areaA > 0)
-      pf_planes |= 1;
-
-   if (areaB > 0)
-      pf_planes |= 2;
-
-   intel_fb->pf_current_page = (intel->sarea->pf_current_page >>
-				(intel_fb->pf_planes & 0x2)) & 0x3;
-
-   intel_fb->pf_num_pages = intel->intelScreen->third.handle ? 3 : 2;
-
-   pf_active = pf_planes && (pf_planes & intel->sarea->pf_active) == pf_planes;
-
-   if (INTEL_DEBUG & DEBUG_LOCK)
-      if (pf_active != intel_fb->pf_active)
-	 _mesa_printf("%s - Page flipping %sactive\n", __progname,
-		      pf_active ? "" : "in");
-
-   if (pf_active) {
-      /* Sync pages between planes if flipping on both at the same time */
-      if (pf_planes == 0x3 && pf_planes != intel_fb->pf_planes &&
-	  (intel->sarea->pf_current_page & 0x3) !=
-	  (((intel->sarea->pf_current_page) >> 2) & 0x3)) {
-	 drm_i915_flip_t flip;
-
-	 if (intel_fb->pf_current_page ==
-	     (intel->sarea->pf_current_page & 0x3)) {
-	    /* XXX: This is ugly, but emitting two flips 'in a row' can cause
-	     * lockups for unknown reasons.
-	     */
-	    intel->sarea->pf_current_page =
-	       intel->sarea->pf_current_page & 0x3;
-	    intel->sarea->pf_current_page |=
-	       ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) %
-		intel_fb->pf_num_pages) << 2;
-
-	    flip.pipes = 0x2;
-	 } else {
-	    intel->sarea->pf_current_page =
-	       intel->sarea->pf_current_page & (0x3 << 2);
-	    intel->sarea->pf_current_page |=
-	       (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) %
-	       intel_fb->pf_num_pages;
-
-	    flip.pipes = 0x1;
-	 }
-
-	 drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip));
-      }
-
-      intel_fb->pf_planes = pf_planes;
-   }
-
-   intel_fb->pf_active = pf_active;
-   intel_flip_renderbuffers(intel_fb);
-   intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer);
-}
-
-/**
- * This will be called whenever the currently bound window is moved/resized.
- * XXX: actually, it seems to NOT be called when the window is only moved (BP).
- */
-void
-intelWindowMoved(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
-
-   if (!intel->intelScreen->driScrnPriv->dri2.enabled &&
-       intel->intelScreen->driScrnPriv->ddx_version.minor >= 7) {
-      volatile struct drm_i915_sarea *sarea = intel->sarea;
-      drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w,
-				   .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h };
-      drm_clip_rect_t planeA_rect = { .x1 = sarea->planeA_x, .y1 = sarea->planeA_y,
-				     .x2 = sarea->planeA_x + sarea->planeA_w,
-				     .y2 = sarea->planeA_y + sarea->planeA_h };
-      drm_clip_rect_t planeB_rect = { .x1 = sarea->planeB_x, .y1 = sarea->planeB_y,
-				     .x2 = sarea->planeB_x + sarea->planeB_w,
-				     .y2 = sarea->planeB_y + sarea->planeB_h };
-      GLint areaA = driIntersectArea( drw_rect, planeA_rect );
-      GLint areaB = driIntersectArea( drw_rect, planeB_rect );
-      GLuint flags = dPriv->vblFlags;
-
-      intelUpdatePageFlipping(intel, areaA, areaB);
-
-      /* Update vblank info
-       */
-      if (areaB > areaA || (areaA == areaB && areaB > 0)) {
-	 flags = dPriv->vblFlags | VBLANK_FLAG_SECONDARY;
-      } else {
-	 flags = dPriv->vblFlags & ~VBLANK_FLAG_SECONDARY;
-      }
-
-      /* Check to see if we changed pipes */
-      if (flags != dPriv->vblFlags && dPriv->vblFlags &&
-	  !(dPriv->vblFlags & VBLANK_FLAG_NO_IRQ)) {
-	 int64_t count;
-	 drmVBlank vbl;
-	 int i;
-
-	 /*
-	  * Deal with page flipping
-	  */
-	 vbl.request.type = DRM_VBLANK_ABSOLUTE;
-
-	 if ( dPriv->vblFlags & VBLANK_FLAG_SECONDARY ) {
-	    vbl.request.type |= DRM_VBLANK_SECONDARY;
-	 }
-
-	 for (i = 0; i < intel_fb->pf_num_pages; i++) {
-	    if (!intel_fb->color_rb[i] ||
-		(intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <=
-		(1<<23))
-	       continue;
-
-	    vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending;
-	    drmWaitVBlank(intel->driFd, &vbl);
-	 }
-
-	 /*
-	  * Update msc_base from old pipe
-	  */
-	 driDrawableGetMSC32(dPriv->driScreenPriv, dPriv, &count);
-	 dPriv->msc_base = count;
-	 /*
-	  * Then get new vblank_base and vblSeq values
-	  */
-	 dPriv->vblFlags = flags;
-	 driGetCurrentVBlank(dPriv);
-	 dPriv->vblank_base = dPriv->vblSeq;
-
-	 intel_fb->vbl_waited = dPriv->vblSeq;
-
-	 for (i = 0; i < intel_fb->pf_num_pages; i++) {
-	    if (intel_fb->color_rb[i])
-	       intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited;
-	 }
-      }
-   } else {
-      dPriv->vblFlags &= ~VBLANK_FLAG_SECONDARY;
-   }
-
-   /* Update Mesa's notion of window size */
-   driUpdateFramebufferSize(ctx, dPriv);
-   intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */
-
-   /* Update hardware scissor */
-   if (ctx->Driver.Scissor != NULL) {
-      ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
-			  ctx->Scissor.Width, ctx->Scissor.Height);
-   }
-
-   /* Re-calculate viewport related state */
-   if (ctx->Driver.DepthRange != NULL)
-      ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far );
-}
-
-
-
-/* A true meta version of this would be very simple and additionally
- * machine independent.  Maybe we'll get there one day.
- */
-static void
-intelClearWithTris(struct intel_context *intel, GLbitfield mask)
-{
-   GLcontext *ctx = &intel->ctx;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   GLuint buf;
-
-   intel->vtbl.install_meta_state(intel);
-
-   /* Back and stencil cliprects are the same.  Try and do both
-    * buffers at once:
-    */
-   if (mask & (BUFFER_BIT_BACK_LEFT | BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH)) {
-      struct intel_region *backRegion =
-	 intel_get_rb_region(fb, BUFFER_BACK_LEFT);
-      struct intel_region *depthRegion =
-	 intel_get_rb_region(fb, BUFFER_DEPTH);
-
-      intel->vtbl.meta_draw_region(intel, backRegion, depthRegion);
-
-      if (mask & BUFFER_BIT_BACK_LEFT)
-	 intel->vtbl.meta_color_mask(intel, GL_TRUE);
-      else
-	 intel->vtbl.meta_color_mask(intel, GL_FALSE);
-
-      if (mask & BUFFER_BIT_STENCIL)
-	 intel->vtbl.meta_stencil_replace(intel,
-					  intel->ctx.Stencil.WriteMask[0],
-					  intel->ctx.Stencil.Clear);
-      else
-	 intel->vtbl.meta_no_stencil_write(intel);
-
-      if (mask & BUFFER_BIT_DEPTH)
-	 intel->vtbl.meta_depth_replace(intel);
-      else
-	 intel->vtbl.meta_no_depth_write(intel);
-
-      intel->vtbl.meta_draw_quad(intel,
-				 fb->_Xmin,
-				 fb->_Xmax,
-				 fb->_Ymin,
-				 fb->_Ymax,
-				 intel->ctx.Depth.Clear,
-				 intel->ClearColor8888,
-				 0, 0, 0, 0);   /* texcoords */
-
-      mask &= ~(BUFFER_BIT_BACK_LEFT | BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH);
-   }
-
-   /* clear the remaining (color) renderbuffers */
-   for (buf = 0; buf < BUFFER_COUNT && mask; buf++) {
-      const GLuint bufBit = 1 << buf;
-      if (mask & bufBit) {
-	 struct intel_renderbuffer *irbColor =
-	    intel_renderbuffer(fb->Attachment[buf].Renderbuffer);
-
-	 ASSERT(irbColor);
-
-	 intel->vtbl.meta_no_depth_write(intel);
-	 intel->vtbl.meta_no_stencil_write(intel);
-	 intel->vtbl.meta_color_mask(intel, GL_TRUE);
-	 intel->vtbl.meta_draw_region(intel, irbColor->region, NULL);
-
-	 intel->vtbl.meta_draw_quad(intel,
-				    fb->_Xmin,
-				    fb->_Xmax,
-				    fb->_Ymin,
-				    fb->_Ymax,
-				    0, intel->ClearColor8888,
-				    0, 0, 0, 0);   /* texcoords */
-
-	 mask &= ~bufBit;
-      }
-   }
-
-   intel->vtbl.leave_meta_state(intel);
-}
-
-static const char *buffer_names[] = {
-   [BUFFER_FRONT_LEFT] = "front",
-   [BUFFER_BACK_LEFT] = "back",
-   [BUFFER_FRONT_RIGHT] = "front right",
-   [BUFFER_BACK_RIGHT] = "back right",
-   [BUFFER_AUX0] = "aux0",
-   [BUFFER_AUX1] = "aux1",
-   [BUFFER_AUX2] = "aux2",
-   [BUFFER_AUX3] = "aux3",
-   [BUFFER_DEPTH] = "depth",
-   [BUFFER_STENCIL] = "stencil",
-   [BUFFER_ACCUM] = "accum",
-   [BUFFER_COLOR0] = "color0",
-   [BUFFER_COLOR1] = "color1",
-   [BUFFER_COLOR2] = "color2",
-   [BUFFER_COLOR3] = "color3",
-   [BUFFER_COLOR4] = "color4",
-   [BUFFER_COLOR5] = "color5",
-   [BUFFER_COLOR6] = "color6",
-   [BUFFER_COLOR7] = "color7",
-};
-
-/**
- * Called by ctx->Driver.Clear.
- */
-static void
-intelClear(GLcontext *ctx, GLbitfield mask)
-{
-   struct intel_context *intel = intel_context(ctx);
-   const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
-   GLbitfield tri_mask = 0;
-   GLbitfield blit_mask = 0;
-   GLbitfield swrast_mask = 0;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   GLuint i;
-
-   if (0)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   /* HW color buffers (front, back, aux, generic FBO, etc) */
-   if (colorMask == ~0) {
-      /* clear all R,G,B,A */
-      /* XXX FBO: need to check if colorbuffers are software RBOs! */
-      blit_mask |= (mask & BUFFER_BITS_COLOR);
-   }
-   else {
-      /* glColorMask in effect */
-      tri_mask |= (mask & BUFFER_BITS_COLOR);
-   }
-
-   /* HW stencil */
-   if (mask & BUFFER_BIT_STENCIL) {
-      const struct intel_region *stencilRegion
-         = intel_get_rb_region(fb, BUFFER_STENCIL);
-      if (stencilRegion) {
-         /* have hw stencil */
-         if (IS_965(intel->intelScreen->deviceID) ||
-	     (ctx->Stencil.WriteMask[0] & 0xff) != 0xff) {
-	    /* We have to use the 3D engine if we're clearing a partial mask
-	     * of the stencil buffer, or if we're on a 965 which has a tiled
-	     * depth/stencil buffer in a layout we can't blit to.
-	     */
-            tri_mask |= BUFFER_BIT_STENCIL;
-         }
-         else {
-            /* clearing all stencil bits, use blitting */
-            blit_mask |= BUFFER_BIT_STENCIL;
-         }
-      }
-   }
-
-   /* HW depth */
-   if (mask & BUFFER_BIT_DEPTH) {
-      /* clear depth with whatever method is used for stencil (see above) */
-      if (IS_965(intel->intelScreen->deviceID) ||
-	  tri_mask & BUFFER_BIT_STENCIL)
-         tri_mask |= BUFFER_BIT_DEPTH;
-      else
-         blit_mask |= BUFFER_BIT_DEPTH;
-   }
-
-   /* SW fallback clearing */
-   swrast_mask = mask & ~tri_mask & ~blit_mask;
-
-   for (i = 0; i < BUFFER_COUNT; i++) {
-      GLuint bufBit = 1 << i;
-      if ((blit_mask | tri_mask) & bufBit) {
-         if (!fb->Attachment[i].Renderbuffer->ClassID) {
-            blit_mask &= ~bufBit;
-            tri_mask &= ~bufBit;
-            swrast_mask |= bufBit;
-         }
-      }
-   }
-
-   if (blit_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("blit clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (blit_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
-      intelClearWithBlit(ctx, blit_mask);
-   }
-
-   if (tri_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("tri clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (tri_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
-      intelClearWithTris(intel, tri_mask);
-   }
-
-   if (swrast_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("swrast clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (swrast_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
-      _swrast_Clear(ctx, swrast_mask);
-   }
-}
-
-
-/* Emit wait for pending flips */
-void
-intel_wait_flips(struct intel_context *intel)
-{
-   struct intel_framebuffer *intel_fb =
-      (struct intel_framebuffer *) intel->ctx.DrawBuffer;
-   struct intel_renderbuffer *intel_rb =
-      intel_get_renderbuffer(&intel_fb->Base,
-			     intel_fb->Base._ColorDrawBufferIndexes[0] ==
-			     BUFFER_FRONT_LEFT ? BUFFER_FRONT_LEFT :
-			     BUFFER_BACK_LEFT);
-
-   if (intel->intelScreen->driScrnPriv->dri2.enabled)
-      return;
-
-   if (intel_fb->Base.Name == 0 && intel_rb &&
-       intel_rb->pf_pending == intel_fb->pf_seq) {
-      GLint pf_planes = intel_fb->pf_planes;
-      BATCH_LOCALS;
-
-      /* Wait for pending flips to take effect */
-      BEGIN_BATCH(2, NO_LOOP_CLIPRECTS);
-      OUT_BATCH(pf_planes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP)
-		: 0);
-      OUT_BATCH(pf_planes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP)
-		: 0);
-      ADVANCE_BATCH();
-
-      intel_rb->pf_pending--;
-   }
-}
-
-
-/* Flip the front & back buffers
- */
-static GLboolean
-intelPageFlip(const __DRIdrawablePrivate * dPriv)
-{
-   struct intel_context *intel;
-   int ret;
-   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
-
-   if (INTEL_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate;
-
-   if (intel->intelScreen->drmMinor < 9)
-      return GL_FALSE;
-
-   intelFlush(&intel->ctx);
-
-   ret = 0;
-
-   LOCK_HARDWARE(intel);
-
-   if (dPriv->numClipRects && intel_fb->pf_active) {
-      drm_i915_flip_t flip;
-
-      flip.pipes = intel_fb->pf_planes;
-
-      ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip));
-   }
-
-   UNLOCK_HARDWARE(intel);
-
-   if (ret || !intel_fb->pf_active)
-      return GL_FALSE;
-
-   if (!dPriv->numClipRects) {
-      usleep(10000);	/* throttle invisible client 10ms */
-   }
-
-   intel_fb->pf_current_page = (intel->sarea->pf_current_page >>
-				(intel_fb->pf_planes & 0x2)) & 0x3;
-
-   if (dPriv->numClipRects != 0) {
-      intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->pf_pending =
-      intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->pf_pending =
-	 ++intel_fb->pf_seq;
-   }
-
-   intel_flip_renderbuffers(intel_fb);
-   intel_draw_buffer(&intel->ctx, &intel_fb->Base);
-
-   return GL_TRUE;
-}
-
-static GLboolean
-intelScheduleSwap(__DRIdrawablePrivate * dPriv, GLboolean *missed_target)
-{
-   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
-   unsigned int interval;
-   struct intel_context *intel =
-      intelScreenContext(dPriv->driScreenPriv->private);
-   const intelScreenPrivate *intelScreen = intel->intelScreen;
-   unsigned int target;
-   drm_i915_vblank_swap_t swap;
-   GLboolean ret;
-
-   if (!dPriv->vblFlags ||
-       (dPriv->vblFlags & VBLANK_FLAG_NO_IRQ) ||
-       intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6))
-      return GL_FALSE;
-
-   interval = driGetVBlankInterval(dPriv);
-
-   swap.seqtype = DRM_VBLANK_ABSOLUTE;
-
-   if (dPriv->vblFlags & VBLANK_FLAG_SYNC) {
-      swap.seqtype |= DRM_VBLANK_NEXTONMISS;
-   } else if (interval == 0)
-      return GL_FALSE;
-
-   swap.drawable = dPriv->hHWDrawable;
-   target = swap.sequence = dPriv->vblSeq + interval;
-
-   if ( dPriv->vblFlags & VBLANK_FLAG_SECONDARY ) {
-      swap.seqtype |= DRM_VBLANK_SECONDARY;
-   }
-
-   LOCK_HARDWARE(intel);
-
-   intel_batchbuffer_flush(intel->batch);
-
-   if ( intel_fb->pf_active ) {
-      swap.seqtype |= DRM_VBLANK_FLIP;
-
-      intel_fb->pf_current_page = (((intel->sarea->pf_current_page >>
-				     (intel_fb->pf_planes & 0x2)) & 0x3) + 1) %
-				  intel_fb->pf_num_pages;
-   }
-
-   if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap,
-			    sizeof(swap))) {
-      dPriv->vblSeq = swap.sequence;
-      swap.sequence -= target;
-      *missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23);
-
-      intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->vbl_pending =
-	 intel_get_renderbuffer(&intel_fb->Base,
-				BUFFER_FRONT_LEFT)->vbl_pending =
-	 dPriv->vblSeq;
-
-      if (swap.seqtype & DRM_VBLANK_FLIP) {
-	 intel_flip_renderbuffers(intel_fb);
-	 intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer);
-      }
-
-      ret = GL_TRUE;
-   } else {
-      if (swap.seqtype & DRM_VBLANK_FLIP) {
-	 intel_fb->pf_current_page = ((intel->sarea->pf_current_page >>
-					(intel_fb->pf_planes & 0x2)) & 0x3) %
-				     intel_fb->pf_num_pages;
-      }
-
-      ret = GL_FALSE;
-   }
-
-   UNLOCK_HARDWARE(intel);
-
-   return ret;
-}
-  
-void
-intelSwapBuffers(__DRIdrawablePrivate * dPriv)
-{
-   __DRIscreenPrivate *psp = dPriv->driScreenPriv;
-
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      GET_CURRENT_CONTEXT(ctx);
-      struct intel_context *intel;
-
-      if (ctx == NULL)
-	 return;
-
-      intel = intel_context(ctx);
-
-      if (ctx->Visual.doubleBufferMode) {
-	 GLboolean missed_target;
-	 struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
-	 int64_t ust;
-         
-	 _mesa_notifySwapBuffers(ctx);  /* flush pending rendering comands */
-
-         if (!intelScheduleSwap(dPriv, &missed_target)) {
-	    driWaitForVBlank(dPriv, &missed_target);
-
-	    /*
-	     * Update each buffer's vbl_pending so we don't get too out of
-	     * sync
-	     */
-	    intel_get_renderbuffer(&intel_fb->Base,
-				   BUFFER_BACK_LEFT)->vbl_pending = 
-		    intel_get_renderbuffer(&intel_fb->Base,
-					   BUFFER_FRONT_LEFT)->vbl_pending =
-		    dPriv->vblSeq;
-	    if (!intelPageFlip(dPriv)) {
-	       intelCopyBuffer(dPriv, NULL);
-	    }
-	 }
-
-	 intel_fb->swap_count++;
-	 (*psp->systemTime->getUST) (&ust);
-	 if (missed_target) {
-	    intel_fb->swap_missed_count++;
-	    intel_fb->swap_missed_ust = ust - intel_fb->swap_ust;
-	 }
-
-	 intel_fb->swap_ust = ust;
-      }
-      drmCommandNone(intel->driFd, DRM_I915_GEM_THROTTLE);
-
-   }
-   else {
-      /* XXX this shouldn't be an error but we can't handle it for now */
-      fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__);
-   }
-}
-
-void
-intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h)
-{
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      struct intel_context *intel =
-         (struct intel_context *) dPriv->driContextPriv->driverPrivate;
-      GLcontext *ctx = &intel->ctx;
-
-      if (ctx->Visual.doubleBufferMode) {
-         drm_clip_rect_t rect;
-         rect.x1 = x + dPriv->x;
-         rect.y1 = (dPriv->h - y - h) + dPriv->y;
-         rect.x2 = rect.x1 + w;
-         rect.y2 = rect.y1 + h;
-         _mesa_notifySwapBuffers(ctx);  /* flush pending rendering comands */
-         intelCopyBuffer(dPriv, &rect);
-      }
-   }
-   else {
-      /* XXX this shouldn't be an error but we can't handle it for now */
-      fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__);
-   }
-}
-
 
 /**
  * Update the hardware state for drawing into a window or framebuffer object.
@@ -818,7 +154,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       return;
    }
 
-   /* Do this here, note core Mesa, since this function is called from
+   /* Do this here, not core Mesa, since this function is called from
     * many places within the driver.
     */
    if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
@@ -836,9 +172,6 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       return;
    }
 
-   if (fb->Name)
-      intel_validate_paired_depth_stencil(ctx, fb);
-
    /*
     * How many color buffers are we drawing into?
     */
@@ -846,7 +179,8 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       /* writing to 0  */
       colorRegions[0] = NULL;
       intel->constant_cliprect = GL_TRUE;
-   } else if (fb->_NumColorDrawBuffers > 1) {
+   }
+   else if (fb->_NumColorDrawBuffers > 1) {
        int i;
        struct intel_renderbuffer *irb;
 
@@ -885,14 +219,6 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       }
    }
 
-   /* Update culling direction which changes depending on the
-    * orientation of the buffer:
-    */
-   if (ctx->Driver.FrontFace)
-      ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace);
-   else
-      ctx->NewState |= _NEW_POLYGON;
-
    if (!colorRegions[0]) {
       FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE);
    }
@@ -924,20 +250,13 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
    /***
     *** Stencil buffer
     *** This can only be hardware accelerated if we're using a
-    *** combined DEPTH_STENCIL buffer (for now anyway).
+    *** combined DEPTH_STENCIL buffer.
     ***/
    if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) {
       irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped);
       if (irbStencil && irbStencil->region) {
          ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
          FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE);
-         /* need to re-compute stencil hw state */
-	 if (ctx->Driver.Enable != NULL)
-	    ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled);
-	 else
-	    ctx->NewState |= _NEW_STENCIL;
-         if (!depthRegion)
-            depthRegion = irbStencil->region;
       }
       else {
          FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_TRUE);
@@ -946,37 +265,30 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
    else {
       /* XXX FBO: instead of FALSE, pass ctx->Stencil.Enabled ??? */
       FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE);
-      /* need to re-compute stencil hw state */
-      if (ctx->Driver.Enable != NULL)
-	 ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled);
-      else
-	 ctx->NewState |= _NEW_STENCIL;
    }
 
    /*
-    * Update depth test state
+    * Update depth and stencil test state
     */
    if (ctx->Driver.Enable) {
-      if (ctx->Depth.Test && fb->Visual.depthBits > 0) {
-	 ctx->Driver.Enable(ctx, GL_DEPTH_TEST, GL_TRUE);
-      } else {
-	 ctx->Driver.Enable(ctx, GL_DEPTH_TEST, GL_FALSE);
-      }
-   } else {
-      ctx->NewState |= _NEW_DEPTH;
+      ctx->Driver.Enable(ctx, GL_DEPTH_TEST,
+                         (ctx->Depth.Test && fb->Visual.depthBits > 0));
+      ctx->Driver.Enable(ctx, GL_STENCIL_TEST,
+                         (ctx->Stencil.Enabled && fb->Visual.stencilBits > 0));
+   }
+   else {
+      ctx->NewState |= (_NEW_DEPTH | _NEW_STENCIL);
    }
 
    intel->vtbl.set_draw_region(intel, colorRegions, depthRegion, 
-	fb->_NumColorDrawBuffers);
+                               fb->_NumColorDrawBuffers);
 
    /* update viewport since it depends on window size */
-   if (ctx->Driver.Viewport) {
-      ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y,
-			   ctx->Viewport.Width, ctx->Viewport.Height);
-   } else {
-      ctx->NewState |= _NEW_VIEWPORT;
-   }
-
+#ifdef I915
+   intelCalcViewport(ctx);
+#else
+   ctx->NewState |= _NEW_VIEWPORT;
+#endif
    /* Set state we know depends on drawable parameters:
     */
    if (ctx->Driver.Scissor)
@@ -988,6 +300,14 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       ctx->Driver.DepthRange(ctx,
 			     ctx->Viewport.Near,
 			     ctx->Viewport.Far);
+
+   /* Update culling direction which changes depending on the
+    * orientation of the buffer:
+    */
+   if (ctx->Driver.FrontFace)
+      ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace);
+   else
+      ctx->NewState |= _NEW_POLYGON;
 }
 
 
@@ -1018,7 +338,6 @@ intelReadBuffer(GLcontext * ctx, GLenum mode)
 void
 intelInitBufferFuncs(struct dd_function_table *functions)
 {
-   functions->Clear = intelClear;
    functions->DrawBuffer = intelDrawBuffer;
    functions->ReadBuffer = intelReadBuffer;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.h b/src/mesa/drivers/dri/intel/intel_buffers.h
index e5afb37dd1..6069d38e9e 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.h
+++ b/src/mesa/drivers/dri/intel/intel_buffers.h
@@ -45,12 +45,6 @@ extern struct intel_region *intel_readbuf_region(struct intel_context *intel);
 
 extern struct intel_region *intel_drawbuf_region(struct intel_context *intel);
 
-extern void intel_wait_flips(struct intel_context *intel);
-
-extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv);
-
-extern void intelWindowMoved(struct intel_context *intel);
-
 extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb);
 
 extern void intelInitBufferFuncs(struct dd_function_table *functions);
@@ -59,5 +53,8 @@ void intel_get_cliprects(struct intel_context *intel,
 			 struct drm_clip_rect **cliprects,
 			 unsigned int *num_cliprects,
 			 int *x_off, int *y_off);
+#ifdef I915
+void intelCalcViewport(GLcontext * ctx);
+#endif
 
 #endif /* INTEL_BUFFERS_H */
diff --git a/src/mesa/drivers/dri/intel/intel_clear.c b/src/mesa/drivers/dri/intel/intel_clear.c
new file mode 100644
index 0000000000..b229136316
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_clear.c
@@ -0,0 +1,376 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 Intel Corporation.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "main/glheader.h"
+#include "main/enums.h"
+#include "main/image.h"
+#include "main/mtypes.h"
+#include "main/attrib.h"
+#include "main/blend.h"
+#include "main/bufferobj.h"
+#include "main/buffers.h"
+#include "main/depth.h"
+#include "main/enable.h"
+#include "main/macros.h"
+#include "main/matrix.h"
+#include "main/texstate.h"
+#include "main/shaders.h"
+#include "main/stencil.h"
+#include "main/varray.h"
+#include "glapi/dispatch.h"
+#include "swrast/swrast.h"
+
+#include "intel_context.h"
+#include "intel_blit.h"
+#include "intel_chipset.h"
+#include "intel_clear.h"
+#include "intel_fbo.h"
+#include "intel_pixel.h"
+
+#define FILE_DEBUG_FLAG DEBUG_BLIT
+
+/**
+ * Perform glClear where mask contains only color, depth, and/or stencil.
+ *
+ * The implementation is based on calling into Mesa to set GL state and
+ * performing normal triangle rendering.  The intent of this path is to
+ * have as generic a path as possible, so that any driver could make use of
+ * it.
+ */
+void
+intel_clear_tris(GLcontext *ctx, GLbitfield mask)
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLfloat vertices[4][3];
+   GLfloat color[4][4];
+   GLfloat dst_z;
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   int i;
+   GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE;
+   GLboolean saved_shader_program = 0;
+   unsigned int saved_active_texture;
+
+   assert((mask & ~(BUFFER_BIT_BACK_LEFT | BUFFER_BIT_FRONT_LEFT |
+		    BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) == 0);
+
+   _mesa_PushAttrib(GL_COLOR_BUFFER_BIT |
+		    GL_CURRENT_BIT |
+		    GL_DEPTH_BUFFER_BIT |
+		    GL_ENABLE_BIT |
+		    GL_STENCIL_BUFFER_BIT |
+		    GL_TRANSFORM_BIT |
+		    GL_CURRENT_BIT);
+   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
+   saved_active_texture = ctx->Texture.CurrentUnit;
+
+   /* Disable existing GL state we don't want to apply to a clear. */
+   _mesa_Disable(GL_ALPHA_TEST);
+   _mesa_Disable(GL_BLEND);
+   _mesa_Disable(GL_CULL_FACE);
+   _mesa_Disable(GL_FOG);
+   _mesa_Disable(GL_POLYGON_SMOOTH);
+   _mesa_Disable(GL_POLYGON_STIPPLE);
+   _mesa_Disable(GL_POLYGON_OFFSET_FILL);
+   _mesa_Disable(GL_LIGHTING);
+   _mesa_Disable(GL_CLIP_PLANE0);
+   _mesa_Disable(GL_CLIP_PLANE1);
+   _mesa_Disable(GL_CLIP_PLANE2);
+   _mesa_Disable(GL_CLIP_PLANE3);
+   _mesa_Disable(GL_CLIP_PLANE4);
+   _mesa_Disable(GL_CLIP_PLANE5);
+   if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) {
+      saved_fp_enable = GL_TRUE;
+      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
+   }
+   if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) {
+      saved_vp_enable = GL_TRUE;
+      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
+   }
+   if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) {
+      saved_shader_program = ctx->Shader.CurrentProgram->Name;
+      _mesa_UseProgramObjectARB(0);
+   }
+
+   if (ctx->Texture._EnabledUnits != 0) {
+      int i;
+
+      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
+	 _mesa_ActiveTextureARB(GL_TEXTURE0 + i);
+	 _mesa_Disable(GL_TEXTURE_1D);
+	 _mesa_Disable(GL_TEXTURE_2D);
+	 _mesa_Disable(GL_TEXTURE_3D);
+	 if (ctx->Extensions.ARB_texture_cube_map)
+	    _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB);
+	 if (ctx->Extensions.NV_texture_rectangle)
+	    _mesa_Disable(GL_TEXTURE_RECTANGLE_NV);
+	 if (ctx->Extensions.MESA_texture_array) {
+	    _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT);
+	    _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT);
+	 }
+      }
+   }
+
+   intel_meta_set_passthrough_transform(intel);
+
+   for (i = 0; i < 4; i++) {
+      color[i][0] = ctx->Color.ClearColor[0];
+      color[i][1] = ctx->Color.ClearColor[1];
+      color[i][2] = ctx->Color.ClearColor[2];
+      color[i][3] = ctx->Color.ClearColor[3];
+   }
+
+   /* convert clear Z from [0,1] to NDC coord in [-1,1] */
+   dst_z = -1.0 + 2.0 * ctx->Depth.Clear;
+
+   /* Prepare the vertices, which are the same regardless of which buffer we're
+    * drawing to.
+    */
+   vertices[0][0] = fb->_Xmin;
+   vertices[0][1] = fb->_Ymin;
+   vertices[0][2] = dst_z;
+   vertices[1][0] = fb->_Xmax;
+   vertices[1][1] = fb->_Ymin;
+   vertices[1][2] = dst_z;
+   vertices[2][0] = fb->_Xmax;
+   vertices[2][1] = fb->_Ymax;
+   vertices[2][2] = dst_z;
+   vertices[3][0] = fb->_Xmin;
+   vertices[3][1] = fb->_Ymax;
+   vertices[3][2] = dst_z;
+
+   _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &color);
+   _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), &vertices);
+   _mesa_Enable(GL_COLOR_ARRAY);
+   _mesa_Enable(GL_VERTEX_ARRAY);
+
+   while (mask != 0) {
+      GLuint this_mask = 0;
+
+      if (mask & BUFFER_BIT_BACK_LEFT)
+	 this_mask = BUFFER_BIT_BACK_LEFT;
+      else if (mask & BUFFER_BIT_FRONT_LEFT)
+	 this_mask = BUFFER_BIT_FRONT_LEFT;
+
+      /* Clear depth/stencil in the same pass as color. */
+      this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL));
+
+      /* Select the current color buffer and use the color write mask if
+       * we have one, otherwise don't write any color channels.
+       */
+      if (this_mask & BUFFER_BIT_FRONT_LEFT)
+	 _mesa_DrawBuffer(GL_FRONT_LEFT);
+      else if (this_mask & BUFFER_BIT_BACK_LEFT)
+	 _mesa_DrawBuffer(GL_BACK_LEFT);
+      else
+	 _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
+
+      /* Control writing of the depth clear value to depth. */
+      if (this_mask & BUFFER_BIT_DEPTH) {
+	 _mesa_DepthFunc(GL_ALWAYS);
+	 _mesa_Enable(GL_DEPTH_TEST);
+      } else {
+	 _mesa_Disable(GL_DEPTH_TEST);
+	 _mesa_DepthMask(GL_FALSE);
+      }
+
+      /* Control writing of the stencil clear value to stencil. */
+      if (this_mask & BUFFER_BIT_STENCIL) {
+	 _mesa_Enable(GL_STENCIL_TEST);
+	 _mesa_StencilOp(GL_REPLACE, GL_REPLACE, GL_REPLACE);
+	 _mesa_StencilFuncSeparate(GL_FRONT, GL_ALWAYS, ctx->Stencil.Clear,
+				   ctx->Stencil.WriteMask[0]);
+      } else {
+	 _mesa_Disable(GL_STENCIL_TEST);
+      }
+
+      CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
+
+      mask &= ~this_mask;
+   }
+
+   intel_meta_restore_transform(intel);
+
+   _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture);
+   if (saved_fp_enable)
+      _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
+   if (saved_vp_enable)
+      _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
+
+   if (saved_shader_program)
+      _mesa_UseProgramObjectARB(saved_shader_program);
+
+   _mesa_PopClientAttrib();
+   _mesa_PopAttrib();
+}
+
+static const char *buffer_names[] = {
+   [BUFFER_FRONT_LEFT] = "front",
+   [BUFFER_BACK_LEFT] = "back",
+   [BUFFER_FRONT_RIGHT] = "front right",
+   [BUFFER_BACK_RIGHT] = "back right",
+   [BUFFER_AUX0] = "aux0",
+   [BUFFER_AUX1] = "aux1",
+   [BUFFER_AUX2] = "aux2",
+   [BUFFER_AUX3] = "aux3",
+   [BUFFER_DEPTH] = "depth",
+   [BUFFER_STENCIL] = "stencil",
+   [BUFFER_ACCUM] = "accum",
+   [BUFFER_COLOR0] = "color0",
+   [BUFFER_COLOR1] = "color1",
+   [BUFFER_COLOR2] = "color2",
+   [BUFFER_COLOR3] = "color3",
+   [BUFFER_COLOR4] = "color4",
+   [BUFFER_COLOR5] = "color5",
+   [BUFFER_COLOR6] = "color6",
+   [BUFFER_COLOR7] = "color7",
+};
+
+/**
+ * Called by ctx->Driver.Clear.
+ */
+static void
+intelClear(GLcontext *ctx, GLbitfield mask)
+{
+   struct intel_context *intel = intel_context(ctx);
+   const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
+   GLbitfield tri_mask = 0;
+   GLbitfield blit_mask = 0;
+   GLbitfield swrast_mask = 0;
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   GLuint i;
+
+   if (0)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   /* HW color buffers (front, back, aux, generic FBO, etc) */
+   if (colorMask == ~0) {
+      /* clear all R,G,B,A */
+      /* XXX FBO: need to check if colorbuffers are software RBOs! */
+      blit_mask |= (mask & BUFFER_BITS_COLOR);
+   }
+   else {
+      /* glColorMask in effect */
+      tri_mask |= (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT));
+   }
+
+   /* HW stencil */
+   if (mask & BUFFER_BIT_STENCIL) {
+      const struct intel_region *stencilRegion
+         = intel_get_rb_region(fb, BUFFER_STENCIL);
+      if (stencilRegion) {
+         /* have hw stencil */
+         if (IS_965(intel->intelScreen->deviceID) ||
+	     (ctx->Stencil.WriteMask[0] & 0xff) != 0xff) {
+	    /* We have to use the 3D engine if we're clearing a partial mask
+	     * of the stencil buffer, or if we're on a 965 which has a tiled
+	     * depth/stencil buffer in a layout we can't blit to.
+	     */
+            tri_mask |= BUFFER_BIT_STENCIL;
+         }
+         else {
+            /* clearing all stencil bits, use blitting */
+            blit_mask |= BUFFER_BIT_STENCIL;
+         }
+      }
+   }
+
+   /* HW depth */
+   if (mask & BUFFER_BIT_DEPTH) {
+      /* clear depth with whatever method is used for stencil (see above) */
+      if (IS_965(intel->intelScreen->deviceID) ||
+	  tri_mask & BUFFER_BIT_STENCIL)
+         tri_mask |= BUFFER_BIT_DEPTH;
+      else
+         blit_mask |= BUFFER_BIT_DEPTH;
+   }
+
+   /* If we're doing a tri pass for depth/stencil, include a likely color
+    * buffer with it.
+    */
+   if (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) {
+      tri_mask |= blit_mask & BUFFER_BIT_BACK_LEFT;
+      blit_mask &= ~BUFFER_BIT_BACK_LEFT;
+   }
+
+   /* SW fallback clearing */
+   swrast_mask = mask & ~tri_mask & ~blit_mask;
+
+   for (i = 0; i < BUFFER_COUNT; i++) {
+      GLuint bufBit = 1 << i;
+      if ((blit_mask | tri_mask) & bufBit) {
+         if (!fb->Attachment[i].Renderbuffer->ClassID) {
+            blit_mask &= ~bufBit;
+            tri_mask &= ~bufBit;
+            swrast_mask |= bufBit;
+         }
+      }
+   }
+
+   if (blit_mask) {
+      if (INTEL_DEBUG & DEBUG_BLIT) {
+	 DBG("blit clear:");
+	 for (i = 0; i < BUFFER_COUNT; i++) {
+	    if (blit_mask & (1 << i))
+	       DBG(" %s", buffer_names[i]);
+	 }
+	 DBG("\n");
+      }
+      intelClearWithBlit(ctx, blit_mask);
+   }
+
+   if (tri_mask) {
+      if (INTEL_DEBUG & DEBUG_BLIT) {
+	 DBG("tri clear:");
+	 for (i = 0; i < BUFFER_COUNT; i++) {
+	    if (tri_mask & (1 << i))
+	       DBG(" %s", buffer_names[i]);
+	 }
+	 DBG("\n");
+      }
+      intel_clear_tris(ctx, tri_mask);
+   }
+
+   if (swrast_mask) {
+      if (INTEL_DEBUG & DEBUG_BLIT) {
+	 DBG("swrast clear:");
+	 for (i = 0; i < BUFFER_COUNT; i++) {
+	    if (swrast_mask & (1 << i))
+	       DBG(" %s", buffer_names[i]);
+	 }
+	 DBG("\n");
+      }
+      _swrast_Clear(ctx, swrast_mask);
+   }
+}
+
+
+void
+intelInitClearFuncs(struct dd_function_table *functions)
+{
+   functions->Clear = intelClear;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_clear.h b/src/mesa/drivers/dri/intel/intel_clear.h
new file mode 100644
index 0000000000..7fd6b310a9
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_clear.h
@@ -0,0 +1,38 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef INTEL_CLEAR_H
+#define INTEL_CLEAR_H
+
+struct dd_function_table;
+
+extern void
+intelInitClearFuncs(struct dd_function_table *functions);
+
+
+#endif /* INTEL_CLEAR_H */
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 9ac18e6960..d7ccfa0605 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -28,8 +28,6 @@
 
 #include "main/glheader.h"
 #include "main/context.h"
-#include "main/matrix.h"
-#include "main/simple_list.h"
 #include "main/extensions.h"
 #include "main/framebuffer.h"
 #include "main/imports.h"
@@ -38,66 +36,40 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-
-#include "tnl/t_pipeline.h"
-#include "tnl/t_vertex.h"
-
 #include "drivers/common/driverfuncs.h"
 
-#include "intel_screen.h"
-
 #include "i830_dri.h"
 
 #include "intel_chipset.h"
 #include "intel_buffers.h"
 #include "intel_tex.h"
 #include "intel_batchbuffer.h"
-#include "intel_blit.h"
+#include "intel_clear.h"
+#include "intel_extensions.h"
 #include "intel_pixel.h"
 #include "intel_regions.h"
 #include "intel_buffer_objects.h"
 #include "intel_fbo.h"
 #include "intel_decode.h"
 #include "intel_bufmgr.h"
+#include "intel_screen.h"
+#include "intel_swapbuffers.h"
 
 #include "drirenderbuffer.h"
 #include "vblank.h"
 #include "utils.h"
 #include "xmlpool.h"            /* for symbolic values of enum-type options */
+
+
 #ifndef INTEL_DEBUG
 int INTEL_DEBUG = (0);
 #endif
 
-#define need_GL_ARB_multisample
-#define need_GL_ARB_occlusion_query
-#define need_GL_ARB_point_parameters
-#define need_GL_ARB_shader_objects
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
-#define need_GL_ARB_vertex_program
-#define need_GL_ARB_vertex_shader
-#define need_GL_ARB_window_pos
-#define need_GL_EXT_blend_color
-#define need_GL_EXT_blend_equation_separate
-#define need_GL_EXT_blend_func_separate
-#define need_GL_EXT_blend_minmax
-#define need_GL_EXT_cull_vertex
-#define need_GL_EXT_fog_coord
-#define need_GL_EXT_framebuffer_object
-#define need_GL_EXT_multi_draw_arrays
-#define need_GL_EXT_point_parameters
-#define need_GL_EXT_secondary_color
-#define need_GL_ATI_separate_stencil
-#define need_GL_NV_point_sprite
-#define need_GL_NV_vertex_program
-#define need_GL_VERSION_2_0
-#define need_GL_VERSION_2_1
-
-#include "extension_helper.h"
-
-#define DRIVER_DATE                     "20080716"
+
+#define DRIVER_DATE                     "20090114"
 #define DRIVER_DATE_GEM                 "GEM " DRIVER_DATE
 
+
 static const GLubyte *
 intelGetString(GLcontext * ctx, GLenum name)
 {
@@ -282,6 +254,9 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 	   return;
        }
 
+       if (rb == NULL)
+	  continue;
+
        if (rb->region) {
 	  dri_bo_flink(rb->region->buffer, &name);
 	  if (name == buffers[i].name)
@@ -317,11 +292,13 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
    driUpdateFramebufferSize(&intel->ctx, drawable);
 }
 
-static void
+void
 intel_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
 {
     struct intel_context *intel = intel_context(ctx);
     __DRIcontext *driContext = intel->driContext;
+    void (*old_viewport)(GLcontext *ctx, GLint x, GLint y,
+			 GLsizei w, GLsizei h);
 
     if (!driContext->driScreenPriv->dri2.enabled)
 	return;
@@ -330,118 +307,14 @@ intel_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
     if (driContext->driDrawablePriv != driContext->driReadablePriv)
 	intel_update_renderbuffers(driContext, driContext->driReadablePriv);
 
+    old_viewport = ctx->Driver.Viewport;
     ctx->Driver.Viewport = NULL;
     intel->driDrawable = driContext->driDrawablePriv;
     intelWindowMoved(intel);
     intel_draw_buffer(ctx, intel->ctx.DrawBuffer);
-    ctx->Driver.Viewport = intel_viewport;
+    ctx->Driver.Viewport = old_viewport;
 }
 
-/**
- * Extension strings exported by the intel driver.
- *
- * Extensions supported by all chips supported by i830_dri, i915_dri, or
- * i965_dri.
- */
-static const struct dri_extension card_extensions[] = {
-   { "GL_ARB_multisample",                GL_ARB_multisample_functions },
-   { "GL_ARB_multitexture",               NULL },
-   { "GL_ARB_point_parameters",           GL_ARB_point_parameters_functions },
-   { "GL_ARB_texture_border_clamp",       NULL },
-   { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
-   { "GL_ARB_texture_cube_map",           NULL },
-   { "GL_ARB_texture_env_add",            NULL },
-   { "GL_ARB_texture_env_combine",        NULL },
-   { "GL_ARB_texture_env_crossbar",       NULL },
-   { "GL_ARB_texture_env_dot3",           NULL },
-   { "GL_ARB_texture_mirrored_repeat",    NULL },
-   { "GL_ARB_texture_rectangle",          NULL },
-   { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
-   { "GL_ARB_vertex_program",             GL_ARB_vertex_program_functions },
-   { "GL_ARB_window_pos",                 GL_ARB_window_pos_functions },
-   { "GL_EXT_blend_color",                GL_EXT_blend_color_functions },
-   { "GL_EXT_blend_equation_separate",    GL_EXT_blend_equation_separate_functions },
-   { "GL_EXT_blend_func_separate",        GL_EXT_blend_func_separate_functions },
-   { "GL_EXT_blend_minmax",               GL_EXT_blend_minmax_functions },
-   { "GL_EXT_blend_logic_op",             NULL },
-   { "GL_EXT_blend_subtract",             NULL },
-   { "GL_EXT_cull_vertex",                GL_EXT_cull_vertex_functions },
-   { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
-   { "GL_EXT_multi_draw_arrays",          GL_EXT_multi_draw_arrays_functions },
-   { "GL_EXT_packed_depth_stencil",       NULL },
-   { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
-   { "GL_EXT_stencil_wrap",               NULL },
-   { "GL_EXT_texture_edge_clamp",         NULL },
-   { "GL_EXT_texture_env_combine",        NULL },
-   { "GL_EXT_texture_env_dot3",           NULL },
-   { "GL_EXT_texture_filter_anisotropic", NULL },
-   { "GL_EXT_texture_lod_bias",           NULL },
-   { "GL_3DFX_texture_compression_FXT1",  NULL },
-   { "GL_APPLE_client_storage",           NULL },
-   { "GL_MESA_pack_invert",               NULL },
-   { "GL_MESA_ycbcr_texture",             NULL },
-   { "GL_NV_blend_square",                NULL },
-   { "GL_NV_point_sprite",                GL_NV_point_sprite_functions },
-   { "GL_NV_vertex_program",              GL_NV_vertex_program_functions },
-   { "GL_NV_vertex_program1_1",           NULL },
-   { "GL_SGIS_generate_mipmap",           NULL },
-   { NULL, NULL }
-};
-
-static const struct dri_extension brw_extensions[] = {
-   { "GL_ARB_depth_texture",              NULL },
-   { "GL_ARB_draw_buffers",               NULL },
-   { "GL_ARB_fragment_program",           NULL },
-   { "GL_ARB_fragment_program_shadow",    NULL },
-   { "GL_ARB_fragment_shader",            NULL },
-   { "GL_ARB_occlusion_query",            GL_ARB_occlusion_query_functions },
-   { "GL_ARB_point_sprite", 		  NULL },
-   { "GL_ARB_shader_objects",             GL_ARB_shader_objects_functions },
-   { "GL_ARB_shading_language_100",       GL_VERSION_2_0_functions },
-#if 0
-   /* Support for GLSL 1.20 is currently broken in core Mesa.
-    */
-   { "GL_ARB_shading_language_120",       GL_VERSION_2_1_functions },
-#endif
-   { "GL_ARB_shadow",                     NULL },
-   { "GL_ARB_texture_non_power_of_two",   NULL },
-   { "GL_ARB_vertex_shader",              GL_ARB_vertex_shader_functions },
-   { "GL_EXT_shadow_funcs",               NULL },
-   { "GL_EXT_texture_sRGB",		  NULL },
-   { "GL_ATI_separate_stencil",           GL_ATI_separate_stencil_functions },
-   { NULL,                                NULL }
-};
-
-static const struct dri_extension arb_oq_extensions[] = {
-   { NULL, NULL }
-};
-
-static const struct dri_extension ttm_extensions[] = {
-   { "GL_ARB_pixel_buffer_object",        NULL },
-   { "GL_EXT_framebuffer_object",         GL_EXT_framebuffer_object_functions },
-   { NULL, NULL }
-};
-
-/**
- * Initializes potential list of extensions if ctx == NULL, or actually enables
- * extensions for a context.
- */
-void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging)
-{
-   struct intel_context *intel = ctx?intel_context(ctx):NULL;
-
-   /* Disable imaging extension until convolution is working in teximage paths.
-    */
-   enable_imaging = GL_FALSE;
-
-   driInitExtensions(ctx, card_extensions, enable_imaging);
-
-   if (intel == NULL || intel->ttm)
-      driInitExtensions(ctx, ttm_extensions, GL_FALSE);
-
-   if (intel == NULL || IS_965(intel->intelScreen->deviceID))
-      driInitExtensions(ctx, brw_extensions, GL_FALSE);
-}
 
 static const struct dri_debug_control debug_control[] = {
    { "tex",   DEBUG_TEXTURE},
@@ -492,9 +365,8 @@ intelInvalidateState(GLcontext * ctx, GLuint new_state)
       intel->vtbl.invalidate_state( intel, new_state );
 }
 
-
-void
-intelFlush(GLcontext * ctx)
+static void
+intel_flush(GLcontext *ctx, GLboolean needs_mi_flush)
 {
    struct intel_context *intel = intel_context(ctx);
 
@@ -508,13 +380,26 @@ intelFlush(GLcontext * ctx)
     * lands onscreen in a timely manner, even if the X Server doesn't trigger
     * a flush for us.
     */
-   intel_batchbuffer_emit_mi_flush(intel->batch);
+   if (needs_mi_flush)
+      intel_batchbuffer_emit_mi_flush(intel->batch);
 
    if (intel->batch->map != intel->batch->ptr)
       intel_batchbuffer_flush(intel->batch);
 }
 
 void
+intelFlush(GLcontext * ctx)
+{
+   intel_flush(ctx, GL_FALSE);
+}
+
+static void
+intel_glFlush(GLcontext *ctx)
+{
+   intel_flush(ctx, GL_TRUE);
+}
+
+void
 intelFinish(GLcontext * ctx)
 {
    struct gl_framebuffer *fb = ctx->DrawBuffer;
@@ -540,11 +425,10 @@ intelInitDriverFunctions(struct dd_function_table *functions)
 {
    _mesa_init_driver_functions(functions);
 
-   functions->Flush = intelFlush;
+   functions->Flush = intel_glFlush;
    functions->Finish = intelFinish;
    functions->GetString = intelGetString;
    functions->UpdateState = intelInvalidateState;
-   functions->Viewport = intel_viewport;
 
    functions->CopyColorTable = _swrast_CopyColorTable;
    functions->CopyColorSubTable = _swrast_CopyColorSubTable;
@@ -553,6 +437,7 @@ intelInitDriverFunctions(struct dd_function_table *functions)
 
    intelInitTextureFuncs(functions);
    intelInitStateFuncs(functions);
+   intelInitClearFuncs(functions);
    intelInitBufferFuncs(functions);
    intelInitPixelFuncs(functions);
 }
@@ -773,7 +658,6 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
 
       intel_region_release(&intel->front_region);
       intel_region_release(&intel->back_region);
-      intel_region_release(&intel->third_region);
       intel_region_release(&intel->depth_region);
 
       driDestroyOptionCache(&intel->optionCache);
@@ -823,12 +707,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
 	    intel_renderbuffer_set_region(intel_fb->color_rb[1],
 					  intel->back_region);
          }
-#if 0
-         if (intel_fb->color_rb[2]) {
-	    intel_renderbuffer_set_region(intel_fb->color_rb[2],
-					  intel->third_region);
-         }
-#endif
+
          if (irbDepth) {
 	    intel_renderbuffer_set_region(irbDepth, intel->depth_region);
          }
@@ -861,11 +740,16 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
 		  ? driGetDefaultVBlankFlags(&intel->optionCache)
 		 : VBLANK_FLAG_NO_IRQ;
 
+	       /* Prevent error printf if one crtc is disabled, this will
+		* be properly calculated in intelWindowMoved() next.
+		*/
+		driDrawPriv->vblFlags = intelFixupVblank(intel, driDrawPriv);
+
 	       (*psp->systemTime->getUST) (&intel_fb->swap_ust);
 	       driDrawableInitVBlank(driDrawPriv);
 	       intel_fb->vbl_waited = driDrawPriv->vblSeq;
 
-	       for (i = 0; i < (intel->intelScreen->third.handle ? 3 : 2); i++) {
+	       for (i = 0; i < 2; i++) {
 		  if (intel_fb->color_rb[i])
 		     intel_fb->color_rb[i]->vbl_pending = driDrawPriv->vblSeq;
 	       }
@@ -889,7 +773,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags)
 {
    __DRIdrawablePrivate *dPriv = intel->driDrawable;
    __DRIscreenPrivate *sPriv = intel->driScreen;
-   volatile struct drm_i915_sarea *sarea = intel->sarea;
+   volatile drm_i915_sarea_t *sarea = intel->sarea;
    int me = intel->hHWContext;
 
    drmGetLock(intel->driFd, intel->hHWContext, flags);
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index 3938af4c72..18dc43c4a4 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -157,6 +157,20 @@ struct intel_context
       void (*debug_batch)(struct intel_context *intel);
    } vtbl;
 
+   struct {
+      struct gl_fragment_program *bitmap_fp;
+      struct gl_vertex_program *passthrough_vp;
+
+      struct gl_fragment_program *saved_fp;
+      GLboolean saved_fp_enable;
+      struct gl_vertex_program *saved_vp;
+      GLboolean saved_vp_enable;
+
+      GLint saved_vp_x, saved_vp_y;
+      GLsizei saved_vp_width, saved_vp_height;
+      GLenum saved_matrix_mode;
+   } meta;
+
    GLint refcount;
    GLuint Fallback;
    GLuint NewGLState;
@@ -166,7 +180,6 @@ struct intel_context
 
    struct intel_region *front_region;
    struct intel_region *back_region;
-   struct intel_region *third_region;
    struct intel_region *depth_region;
 
    /**
@@ -184,6 +197,7 @@ struct intel_context
       GLuint id;
       uint32_t primitive;	/**< Current hardware primitive type */
       void (*flush) (struct intel_context *);
+      GLubyte *start_ptr; /**< for i8xx */
       dri_bo *vb_bo;
       uint8_t *vb;
       unsigned int start_offset; /**< Byte offset of primitive sequence */
@@ -196,7 +210,6 @@ struct intel_context
    char *prevLockFile;
    int prevLockLine;
 
-   GLubyte clear_chan[4];
    GLuint ClearColor565;
    GLuint ClearColor8888;
 
@@ -424,7 +437,6 @@ extern void intelFinish(GLcontext * ctx);
 extern void intelFlush(GLcontext * ctx);
 
 extern void intelInitDriverFunctions(struct dd_function_table *functions);
-extern void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging);
 
 
 /* ================================================================
@@ -495,6 +507,9 @@ extern int intel_translate_stencil_op(GLenum op);
 extern int intel_translate_blend_factor(GLenum factor);
 extern int intel_translate_logic_op(GLenum opcode);
 
+void intel_viewport(GLcontext * ctx, GLint x, GLint y,
+		    GLsizei width, GLsizei height);
+
 void intel_update_renderbuffers(__DRIcontext *context,
 				__DRIdrawable *drawable);
 
diff --git a/src/mesa/drivers/dri/intel/intel_decode.c b/src/mesa/drivers/dri/intel/intel_decode.c
index 9c105013c0..136221c37f 100644
--- a/src/mesa/drivers/dri/intel/intel_decode.c
+++ b/src/mesa/drivers/dri/intel/intel_decode.c
@@ -37,6 +37,7 @@
 
 #include <stdio.h>
 #include <stdarg.h>
+#include <string.h>
 #include <inttypes.h>
 
 #include "intel_decode.h"
@@ -86,27 +87,28 @@ decode_mi(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 
     struct {
 	uint32_t opcode;
+	int len_mask;
 	int min_len;
 	int max_len;
 	char *name;
     } opcodes_mi[] = {
-	{ 0x08, 1, 1, "MI_ARB_ON_OFF" },
-	{ 0x0a, 1, 1, "MI_BATCH_BUFFER_END" },
-	{ 0x31, 2, 2, "MI_BATCH_BUFFER_START" },
-	{ 0x14, 3, 3, "MI_DISPLAY_BUFFER_INFO" },
-	{ 0x04, 1, 1, "MI_FLUSH" },
-	{ 0x22, 3, 3, "MI_LOAD_REGISTER_IMM" },
-	{ 0x13, 2, 2, "MI_LOAD_SCAN_LINES_EXCL" },
-	{ 0x12, 2, 2, "MI_LOAD_SCAN_LINES_INCL" },
-	{ 0x00, 1, 1, "MI_NOOP" },
-	{ 0x11, 2, 2, "MI_OVERLAY_FLIP" },
-	{ 0x07, 1, 1, "MI_REPORT_HEAD" },
-	{ 0x18, 2, 2, "MI_SET_CONTEXT" },
-	{ 0x20, 3, 4, "MI_STORE_DATA_IMM" },
-	{ 0x21, 3, 4, "MI_STORE_DATA_INDEX" },
-	{ 0x24, 3, 3, "MI_STORE_REGISTER_MEM" },
-	{ 0x02, 1, 1, "MI_USER_INTERRUPT" },
-	{ 0x03, 1, 1, "MI_WAIT_FOR_EVENT" },
+	{ 0x08, 0, 1, 1, "MI_ARB_ON_OFF" },
+	{ 0x0a, 0, 1, 1, "MI_BATCH_BUFFER_END" },
+	{ 0x31, 0x3f, 2, 2, "MI_BATCH_BUFFER_START" },
+	{ 0x14, 0x3f, 3, 3, "MI_DISPLAY_BUFFER_INFO" },
+	{ 0x04, 0, 1, 1, "MI_FLUSH" },
+	{ 0x22, 0, 3, 3, "MI_LOAD_REGISTER_IMM" },
+	{ 0x13, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_EXCL" },
+	{ 0x12, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_INCL" },
+	{ 0x00, 0, 1, 1, "MI_NOOP" },
+	{ 0x11, 0x3f, 2, 2, "MI_OVERLAY_FLIP" },
+	{ 0x07, 0, 1, 1, "MI_REPORT_HEAD" },
+	{ 0x18, 0x3f, 2, 2, "MI_SET_CONTEXT" },
+	{ 0x20, 0x3f, 3, 4, "MI_STORE_DATA_IMM" },
+	{ 0x21, 0x3f, 3, 4, "MI_STORE_DATA_INDEX" },
+	{ 0x24, 0x3f, 3, 3, "MI_STORE_REGISTER_MEM" },
+	{ 0x02, 0, 1, 1, "MI_USER_INTERRUPT" },
+	{ 0x03, 0, 1, 1, "MI_WAIT_FOR_EVENT" },
     };
 
 
@@ -117,12 +119,14 @@ decode_mi(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 
 	    instr_out(data, hw_offset, 0, "%s\n", opcodes_mi[opcode].name);
 	    if (opcodes_mi[opcode].max_len > 1) {
-		len = (data[0] & 0x000000ff) + 2;
+		len = (data[0] & opcodes_mi[opcode].len_mask) + 2;
 		if (len < opcodes_mi[opcode].min_len ||
 		    len > opcodes_mi[opcode].max_len)
 		{
-		    fprintf(out, "Bad length in %s\n",
-			    opcodes_mi[opcode].name);
+		    fprintf(out, "Bad length (%d) in %s, [%d, %d]\n",
+			    len, opcodes_mi[opcode].name,
+			    opcodes_mi[opcode].min_len,
+			    opcodes_mi[opcode].max_len);
 		}
 	    }
 
@@ -308,6 +312,15 @@ decode_3d_1c(uint32_t *data, int count, uint32_t hw_offset, int *failures)
     case 0x10:
 	instr_out(data, hw_offset, 0, "3DSTATE_SCISSOR_ENABLE\n");
 	return 1;
+    case 0x01:
+	instr_out(data, hw_offset, 0, "3DSTATE_MAP_COORD_SET_I830\n");
+	return 1;
+    case 0x0a:
+	instr_out(data, hw_offset, 0, "3DSTATE_MAP_CUBE_I830\n");
+	return 1;
+    case 0x05:
+	instr_out(data, hw_offset, 0, "3DSTATE_MAP_TEX_STREAM_I830\n");
+	return 1;
     }
 
     instr_out(data, hw_offset, 0, "3D UNKNOWN\n");
@@ -315,34 +328,508 @@ decode_3d_1c(uint32_t *data, int count, uint32_t hw_offset, int *failures)
     return 1;
 }
 
+/** Sets the string dstname to describe the destination of the PS instruction */
+static void
+i915_get_instruction_dst(uint32_t *data, int i, char *dstname, int do_mask)
+{
+    uint32_t a0 = data[i];
+    int dst_nr = (a0 >> 14) & 0xf;
+    char dstmask[8];
+    char *sat;
+
+    if (do_mask) {
+	if (((a0 >> 10) & 0xf) == 0xf) {
+	    dstmask[0] = 0;
+	} else {
+	    int dstmask_index = 0;
+
+	    dstmask[dstmask_index++] = '.';
+	    if (a0 & (1 << 10))
+		dstmask[dstmask_index++] = 'x';
+	    if (a0 & (1 << 11))
+		dstmask[dstmask_index++] = 'y';
+	    if (a0 & (1 << 12))
+		dstmask[dstmask_index++] = 'z';
+	    if (a0 & (1 << 13))
+		dstmask[dstmask_index++] = 'w';
+	    dstmask[dstmask_index++] = 0;
+	}
+
+	if (a0 & (1 << 22))
+	    sat = ".sat";
+	else
+	    sat = "";
+    } else {
+	dstmask[0] = 0;
+	sat = "";
+    }
+
+    switch ((a0 >> 19) & 0x7) {
+    case 0:
+	if (dst_nr > 15)
+	    fprintf(out, "bad destination reg R%d\n", dst_nr);
+	sprintf(dstname, "R%d%s%s", dst_nr, dstmask, sat);
+	break;
+    case 4:
+	if (dst_nr > 0)
+	    fprintf(out, "bad destination reg oC%d\n", dst_nr);
+	sprintf(dstname, "oC%s%s", dstmask, sat);
+	break;
+    case 5:
+	if (dst_nr > 0)
+	    fprintf(out, "bad destination reg oD%d\n", dst_nr);
+	sprintf(dstname, "oD%s%s",  dstmask, sat);
+	break;
+    case 6:
+	if (dst_nr > 2)
+	    fprintf(out, "bad destination reg U%d\n", dst_nr);
+	sprintf(dstname, "U%d%s%s", dst_nr, dstmask, sat);
+	break;
+    default:
+	sprintf(dstname, "RESERVED");
+	break;
+    }
+}
+
+static char *
+i915_get_channel_swizzle(uint32_t select)
+{
+    switch (select & 0x7) {
+    case 0:
+	return (select & 8) ? "-x" : "x";
+    case 1:
+	return (select & 8) ? "-y" : "y";
+    case 2:
+	return (select & 8) ? "-z" : "z";
+    case 3:
+	return (select & 8) ? "-w" : "w";
+    case 4:
+	return (select & 8) ? "-0" : "0";
+    case 5:
+	return (select & 8) ? "-1" : "1";
+    default:
+	return (select & 8) ? "-bad" : "bad";
+    }
+}
+
+static void
+i915_get_instruction_src_name(uint32_t src_type, uint32_t src_nr, char *name)
+{
+    switch (src_type) {
+    case 0:
+	sprintf(name, "R%d", src_nr);
+	if (src_nr > 15)
+	    fprintf(out, "bad src reg %s\n", name);
+	break;
+    case 1:
+	if (src_nr < 8)
+	    sprintf(name, "T%d", src_nr);
+	else if (src_nr == 8)
+	    sprintf(name, "DIFFUSE");
+	else if (src_nr == 9)
+	    sprintf(name, "SPECULAR");
+	else if (src_nr == 10)
+	    sprintf(name, "FOG");
+	else {
+	    fprintf(out, "bad src reg T%d\n", src_nr);
+	    sprintf(name, "RESERVED");
+	}
+	break;
+    case 2:
+	sprintf(name, "C%d", src_nr);
+	if (src_nr > 31)
+	    fprintf(out, "bad src reg %s\n", name);
+	break;
+    case 4:
+	sprintf(name, "oC");
+	if (src_nr > 0)
+	    fprintf(out, "bad src reg oC%d\n", src_nr);
+	break;
+    case 5:
+	sprintf(name, "oD");
+	if (src_nr > 0)
+	    fprintf(out, "bad src reg oD%d\n", src_nr);
+	break;
+    case 6:
+	sprintf(name, "U%d", src_nr);
+	if (src_nr > 2)
+	    fprintf(out, "bad src reg %s\n", name);
+	break;
+    default:
+	fprintf(out, "bad src reg type %d\n", src_type);
+	sprintf(name, "RESERVED");
+	break;
+    }
+}
+
+static void
+i915_get_instruction_src0(uint32_t *data, int i, char *srcname)
+{
+    uint32_t a0 = data[i];
+    uint32_t a1 = data[i + 1];
+    int src_nr = (a0 >> 2) & 0x1f;
+    char *swizzle_x = i915_get_channel_swizzle((a1 >> 28) & 0xf);
+    char *swizzle_y = i915_get_channel_swizzle((a1 >> 24) & 0xf);
+    char *swizzle_z = i915_get_channel_swizzle((a1 >> 20) & 0xf);
+    char *swizzle_w = i915_get_channel_swizzle((a1 >> 16) & 0xf);
+    char swizzle[100];
+
+    i915_get_instruction_src_name((a0 >> 7) & 0x7, src_nr, srcname);
+    sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
+    if (strcmp(swizzle, ".xyzw") != 0)
+	strcat(srcname, swizzle);
+}
+
+static void
+i915_get_instruction_src1(uint32_t *data, int i, char *srcname)
+{
+    uint32_t a1 = data[i + 1];
+    uint32_t a2 = data[i + 2];
+    int src_nr = (a1 >> 8) & 0x1f;
+    char *swizzle_x = i915_get_channel_swizzle((a1 >> 4) & 0xf);
+    char *swizzle_y = i915_get_channel_swizzle((a1 >> 0) & 0xf);
+    char *swizzle_z = i915_get_channel_swizzle((a2 >> 28) & 0xf);
+    char *swizzle_w = i915_get_channel_swizzle((a2 >> 24) & 0xf);
+    char swizzle[100];
+
+    i915_get_instruction_src_name((a1 >> 13) & 0x7, src_nr, srcname);
+    sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
+    if (strcmp(swizzle, ".xyzw") != 0)
+	strcat(srcname, swizzle);
+}
+
+static void
+i915_get_instruction_src2(uint32_t *data, int i, char *srcname)
+{
+    uint32_t a2 = data[i + 2];
+    int src_nr = (a2 >> 16) & 0x1f;
+    char *swizzle_x = i915_get_channel_swizzle((a2 >> 12) & 0xf);
+    char *swizzle_y = i915_get_channel_swizzle((a2 >> 8) & 0xf);
+    char *swizzle_z = i915_get_channel_swizzle((a2 >> 4) & 0xf);
+    char *swizzle_w = i915_get_channel_swizzle((a2 >> 0) & 0xf);
+    char swizzle[100];
+
+    i915_get_instruction_src_name((a2 >> 21) & 0x7, src_nr, srcname);
+    sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
+    if (strcmp(swizzle, ".xyzw") != 0)
+	strcat(srcname, swizzle);
+}
+
+static void
+i915_get_instruction_addr(uint32_t src_type, uint32_t src_nr, char *name)
+{
+    switch (src_type) {
+    case 0:
+	sprintf(name, "R%d", src_nr);
+	if (src_nr > 15)
+	    fprintf(out, "bad src reg %s\n", name);
+	break;
+    case 1:
+	if (src_nr < 8)
+	    sprintf(name, "T%d", src_nr);
+	else if (src_nr == 8)
+	    sprintf(name, "DIFFUSE");
+	else if (src_nr == 9)
+	    sprintf(name, "SPECULAR");
+	else if (src_nr == 10)
+	    sprintf(name, "FOG");
+	else {
+	    fprintf(out, "bad src reg T%d\n", src_nr);
+	    sprintf(name, "RESERVED");
+	}
+	break;
+    case 4:
+	sprintf(name, "oC");
+	if (src_nr > 0)
+	    fprintf(out, "bad src reg oC%d\n", src_nr);
+	break;
+    case 5:
+	sprintf(name, "oD");
+	if (src_nr > 0)
+	    fprintf(out, "bad src reg oD%d\n", src_nr);
+	break;
+    default:
+	fprintf(out, "bad src reg type %d\n", src_type);
+	sprintf(name, "RESERVED");
+	break;
+    }
+}
+
+static void
+i915_decode_alu1(uint32_t *data, uint32_t hw_offset,
+		 int i, char *instr_prefix, char *op_name)
+{
+    char dst[100], src0[100];
+
+    i915_get_instruction_dst(data, i, dst, 1);
+    i915_get_instruction_src0(data, i, src0);
+
+    instr_out(data, hw_offset, i++, "%s: %s %s, %s\n", instr_prefix,
+	      op_name, dst, src0);
+    instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+    instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+}
+
+static void
+i915_decode_alu2(uint32_t *data, uint32_t hw_offset,
+		 int i, char *instr_prefix, char *op_name)
+{
+    char dst[100], src0[100], src1[100];
+
+    i915_get_instruction_dst(data, i, dst, 1);
+    i915_get_instruction_src0(data, i, src0);
+    i915_get_instruction_src1(data, i, src1);
+
+    instr_out(data, hw_offset, i++, "%s: %s %s, %s, %s\n", instr_prefix,
+	      op_name, dst, src0, src1);
+    instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+    instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+}
+
+static void
+i915_decode_alu3(uint32_t *data, uint32_t hw_offset,
+		 int i, char *instr_prefix, char *op_name)
+{
+    char dst[100], src0[100], src1[100], src2[100];
+
+    i915_get_instruction_dst(data, i, dst, 1);
+    i915_get_instruction_src0(data, i, src0);
+    i915_get_instruction_src1(data, i, src1);
+    i915_get_instruction_src2(data, i, src2);
+
+    instr_out(data, hw_offset, i++, "%s: %s %s, %s, %s, %s\n", instr_prefix,
+	      op_name, dst, src0, src1, src2);
+    instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+    instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+}
+
+static void
+i915_decode_tex(uint32_t *data, uint32_t hw_offset, int i, char *instr_prefix,
+		char *tex_name)
+{
+    uint32_t t0 = data[i];
+    uint32_t t1 = data[i + 1];
+    char dst_name[100];
+    char addr_name[100];
+    int sampler_nr;
+
+    i915_get_instruction_dst(data, i, dst_name, 0);
+    i915_get_instruction_addr((t1 >> 24) & 0x7,
+			      (t1 >> 17) & 0xf,
+			      addr_name);
+    sampler_nr = t0 & 0xf;
+
+    instr_out(data, hw_offset, i++, "%s: %s %s, S%d, %s\n", instr_prefix,
+	      tex_name, dst_name, sampler_nr, addr_name);
+    instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+    instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+}
+
+static void
+i915_decode_dcl(uint32_t *data, uint32_t hw_offset, int i, char *instr_prefix)
+{
+    uint32_t d0 = data[i];
+    char *sampletype;
+    int dcl_nr = (d0 >> 14) & 0xf;
+    char *dcl_x = d0 & (1 << 10) ? "x" : "";
+    char *dcl_y = d0 & (1 << 11) ? "y" : "";
+    char *dcl_z = d0 & (1 << 12) ? "z" : "";
+    char *dcl_w = d0 & (1 << 13) ? "w" : "";
+    char dcl_mask[10];
+
+    switch ((d0 >> 19) & 0x3) {
+    case 1:
+	sprintf(dcl_mask, ".%s%s%s%s", dcl_x, dcl_y, dcl_z, dcl_w);
+	if (strcmp(dcl_mask, ".") == 0)
+	    fprintf(out, "bad (empty) dcl mask\n");
+
+	if (dcl_nr > 10)
+	    fprintf(out, "bad T%d dcl register number\n", dcl_nr);
+	if (dcl_nr < 8) {
+	    if (strcmp(dcl_mask, ".x") != 0 &&
+		strcmp(dcl_mask, ".xy") != 0 &&
+		strcmp(dcl_mask, ".xz") != 0 &&
+		strcmp(dcl_mask, ".w") != 0 &&
+		strcmp(dcl_mask, ".xyzw") != 0) {
+		fprintf(out, "bad T%d.%s dcl mask\n", dcl_nr, dcl_mask);
+	    }
+	    instr_out(data, hw_offset, i++, "%s: DCL T%d%s\n", instr_prefix,
+		      dcl_nr, dcl_mask);
+	} else {
+	    if (strcmp(dcl_mask, ".xz") == 0)
+		fprintf(out, "errataed bad dcl mask %s\n", dcl_mask);
+	    else if (strcmp(dcl_mask, ".xw") == 0)
+		fprintf(out, "errataed bad dcl mask %s\n", dcl_mask);
+	    else if (strcmp(dcl_mask, ".xzw") == 0)
+		fprintf(out, "errataed bad dcl mask %s\n", dcl_mask);
+
+	    if (dcl_nr == 8) {
+		instr_out(data, hw_offset, i++, "%s: DCL DIFFUSE%s\n", instr_prefix,
+			  dcl_mask);
+	    } else if (dcl_nr == 9) {
+		instr_out(data, hw_offset, i++, "%s: DCL SPECULAR%s\n", instr_prefix,
+			  dcl_mask);
+	    } else if (dcl_nr == 10) {
+		instr_out(data, hw_offset, i++, "%s: DCL FOG%s\n", instr_prefix,
+			  dcl_mask);
+	    }
+	}
+	instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+	instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+	break;
+    case 3:
+	switch ((d0 >> 22) & 0x3) {
+	case 0:
+	    sampletype = "2D";
+	    break;
+	case 1:
+	    sampletype = "CUBE";
+	    break;
+	case 2:
+	    sampletype = "3D";
+	    break;
+	default:
+	    sampletype = "RESERVED";
+	    break;
+	}
+	if (dcl_nr > 15)
+	    fprintf(out, "bad S%d dcl register number\n", dcl_nr);
+	instr_out(data, hw_offset, i++, "%s: DCL S%d %s\n", instr_prefix,
+		  dcl_nr, sampletype);
+	instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+	instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+	break;
+    default:
+	instr_out(data, hw_offset, i++, "%s: DCL RESERVED%d\n", instr_prefix, dcl_nr);
+	instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+	instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+    }
+}
+
+static void
+i915_decode_instruction(uint32_t *data, uint32_t hw_offset,
+			int i, char *instr_prefix)
+{
+    switch ((data[i] >> 24) & 0x1f) {
+    case 0x0:
+	instr_out(data, hw_offset, i++, "%s: NOP\n", instr_prefix);
+	instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+	instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+	break;
+    case 0x01:
+	i915_decode_alu2(data, hw_offset, i, instr_prefix, "ADD");
+	break;
+    case 0x02:
+	i915_decode_alu1(data, hw_offset, i, instr_prefix, "MOV");
+	break;
+    case 0x03:
+	i915_decode_alu2(data, hw_offset, i, instr_prefix, "MUL");
+	break;
+    case 0x04:
+	i915_decode_alu3(data, hw_offset, i, instr_prefix, "MAD");
+	break;
+    case 0x05:
+	i915_decode_alu3(data, hw_offset, i, instr_prefix, "DP2ADD");
+	break;
+    case 0x06:
+	i915_decode_alu2(data, hw_offset, i, instr_prefix, "DP3");
+	break;
+    case 0x07:
+	i915_decode_alu2(data, hw_offset, i, instr_prefix, "DP4");
+	break;
+    case 0x08:
+	i915_decode_alu1(data, hw_offset, i, instr_prefix, "FRC");
+	break;
+    case 0x09:
+	i915_decode_alu1(data, hw_offset, i, instr_prefix, "RCP");
+	break;
+    case 0x0a:
+	i915_decode_alu1(data, hw_offset, i, instr_prefix, "RSQ");
+	break;
+    case 0x0b:
+	i915_decode_alu1(data, hw_offset, i, instr_prefix, "EXP");
+	break;
+    case 0x0c:
+	i915_decode_alu1(data, hw_offset, i, instr_prefix, "LOG");
+	break;
+    case 0x0d:
+	i915_decode_alu2(data, hw_offset, i, instr_prefix, "CMP");
+	break;
+    case 0x0e:
+	i915_decode_alu2(data, hw_offset, i, instr_prefix, "MIN");
+	break;
+    case 0x0f:
+	i915_decode_alu2(data, hw_offset, i, instr_prefix, "MAX");
+	break;
+    case 0x10:
+	i915_decode_alu1(data, hw_offset, i, instr_prefix, "FLR");
+	break;
+    case 0x11:
+	i915_decode_alu1(data, hw_offset, i, instr_prefix, "MOD");
+	break;
+    case 0x12:
+	i915_decode_alu1(data, hw_offset, i, instr_prefix, "TRC");
+	break;
+    case 0x13:
+	i915_decode_alu2(data, hw_offset, i, instr_prefix, "SGE");
+	break;
+    case 0x14:
+	i915_decode_alu2(data, hw_offset, i, instr_prefix, "SLT");
+	break;
+    case 0x15:
+	i915_decode_tex(data, hw_offset, i, instr_prefix, "TEXLD");
+	break;
+    case 0x16:
+	i915_decode_tex(data, hw_offset, i, instr_prefix, "TEXLDP");
+	break;
+    case 0x17:
+	i915_decode_tex(data, hw_offset, i, instr_prefix, "TEXLDB");
+	break;
+    case 0x19:
+	i915_decode_dcl(data, hw_offset, i, instr_prefix);
+	break;
+    default:
+	instr_out(data, hw_offset, i++, "%s: unknown\n", instr_prefix);
+	instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+	instr_out(data, hw_offset, i++, "%s\n", instr_prefix);
+	break;
+    }
+}
+
 static int
-decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures)
+decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures, int i830)
 {
     unsigned int len, i, c, opcode, word, map, sampler, instr;
 
     struct {
 	uint32_t opcode;
+	int i830_only;
 	int min_len;
 	int max_len;
 	char *name;
     } opcodes_3d_1d[] = {
-	{ 0x8e, 3, 3, "3DSTATE_BUFFER_INFO" },
-	{ 0x86, 4, 4, "3DSTATE_CHROMA_KEY" },
-	{ 0x9c, 1, 1, "3DSTATE_CLEAR_PARAMETERS" },
-	{ 0x88, 2, 2, "3DSTATE_CONSTANT_BLEND_COLOR" },
-	{ 0x99, 2, 2, "3DSTATE_DEFAULT_DIFFUSE" },
-	{ 0x9a, 2, 2, "3DSTATE_DEFAULT_SPECULAR" },
-	{ 0x98, 2, 2, "3DSTATE_DEFAULT_Z" },
-	{ 0x97, 2, 2, "3DSTATE_DEPTH_OFFSET_SCALE" },
-	{ 0x85, 2, 2, "3DSTATE_DEST_BUFFER_VARIABLES" },
-	{ 0x80, 5, 5, "3DSTATE_DRAWING_RECTANGLE" },
-	{ 0x8e, 3, 3, "3DSTATE_BUFFER_INFO" },
-	{ 0x9d, 65, 65, "3DSTATE_FILTER_COEFFICIENTS_4X4" },
-	{ 0x9e, 4, 4, "3DSTATE_MONO_FILTER" },
-	{ 0x89, 4, 4, "3DSTATE_FOG_MODE" },
-	{ 0x8f, 2, 16, "3DSTATE_MAP_PALLETE_LOAD_32" },
-	{ 0x81, 3, 3, "3DSTATE_SCISSOR_RECTANGLE" },
-	{ 0x83, 2, 2, "3DSTATE_SPAN_STIPPLE" },
+	{ 0x8e, 0, 3, 3, "3DSTATE_BUFFER_INFO" },
+	{ 0x86, 0, 4, 4, "3DSTATE_CHROMA_KEY" },
+	{ 0x9c, 0, 1, 1, "3DSTATE_CLEAR_PARAMETERS" },
+	{ 0x88, 0, 2, 2, "3DSTATE_CONSTANT_BLEND_COLOR" },
+	{ 0x99, 0, 2, 2, "3DSTATE_DEFAULT_DIFFUSE" },
+	{ 0x9a, 0, 2, 2, "3DSTATE_DEFAULT_SPECULAR" },
+	{ 0x98, 0, 2, 2, "3DSTATE_DEFAULT_Z" },
+	{ 0x97, 0, 2, 2, "3DSTATE_DEPTH_OFFSET_SCALE" },
+	{ 0x85, 0, 2, 2, "3DSTATE_DEST_BUFFER_VARIABLES" },
+	{ 0x80, 0, 5, 5, "3DSTATE_DRAWING_RECTANGLE" },
+	{ 0x8e, 0, 3, 3, "3DSTATE_BUFFER_INFO" },
+	{ 0x9d, 0, 65, 65, "3DSTATE_FILTER_COEFFICIENTS_4X4" },
+	{ 0x9e, 0, 4, 4, "3DSTATE_MONO_FILTER" },
+	{ 0x89, 0, 4, 4, "3DSTATE_FOG_MODE" },
+	{ 0x8f, 0, 2, 16, "3DSTATE_MAP_PALLETE_LOAD_32" },
+	{ 0x81, 0, 3, 3, "3DSTATE_SCISSOR_RECTANGLE" },
+	{ 0x83, 0, 2, 2, "3DSTATE_SPAN_STIPPLE" },
+	{ 0x8c, 1, 2, 2, "3DSTATE_MAP_COORD_TRANSFORM_I830" },
+	{ 0x8b, 1, 2, 2, "3DSTATE_MAP_VERTEX_TRANSFORM_I830" },
+	{ 0x8d, 1, 3, 3, "3DSTATE_W_STATE_I830" },
+	{ 0x01, 1, 2, 2, "3DSTATE_COLOR_FACTOR_I830" },
+	{ 0x02, 1, 2, 2, "3DSTATE_MAP_COORD_SETBIND_I830" },
     };
 
     switch ((data[0] & 0x00ff0000) >> 16) {
@@ -426,8 +913,9 @@ decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures)
     case 0x00:
 	instr_out(data, hw_offset, 0, "3DSTATE_MAP_STATE\n");
 	len = (data[0] & 0x0000003f) + 2;
+	instr_out(data, hw_offset, 1, "mask\n");
 
-	i = 1;
+	i = 2;
 	for (map = 0; map <= 15; map++) {
 	    if (data[1] & (1 << map)) {
 		if (i + 3 >= count)
@@ -447,7 +935,7 @@ decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 	instr_out(data, hw_offset, 0, "3DSTATE_PIXEL_SHADER_CONSTANTS\n");
 	len = (data[0] & 0x000000ff) + 2;
 
-	i = 1;
+	i = 2;
 	for (c = 0; c <= 31; c++) {
 	    if (data[1] & (1 << c)) {
 		if (i + 4 >= count)
@@ -467,7 +955,7 @@ decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 	    }
 	}
 	if (len != i) {
-	    fprintf(out, "Bad count in 3DSTATE_MAP_STATE\n");
+	    fprintf(out, "Bad count in 3DSTATE_PIXEL_SHADER_CONSTANTS\n");
 	    (*failures)++;
 	}
 	return len;
@@ -480,17 +968,22 @@ decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 	}
 	i = 1;
 	for (instr = 0; instr < (len - 1) / 3; instr++) {
+	    char instr_prefix[10];
+
 	    if (i + 3 >= count)
-		BUFFER_FAIL(count, len, "3DSTATE_MAP_STATE");
-	    instr_out(data, hw_offset, i++, "PS%03x\n", instr);
-	    instr_out(data, hw_offset, i++, "PS%03x\n", instr);
-	    instr_out(data, hw_offset, i++, "PS%03x\n", instr);
+		BUFFER_FAIL(count, len, "3DSTATE_PIXEL_SHADER_PROGRAM");
+	    sprintf(instr_prefix, "PS%03d", instr);
+	    i915_decode_instruction(data, hw_offset, i, instr_prefix);
+	    i += 3;
 	}
 	return len;
     case 0x01:
+	if (i830)
+	    break;
 	instr_out(data, hw_offset, 0, "3DSTATE_SAMPLER_STATE\n");
+	instr_out(data, hw_offset, 1, "mask\n");
 	len = (data[0] & 0x0000003f) + 2;
-	i = 1;
+	i = 2;
 	for (sampler = 0; sampler <= 15; sampler++) {
 	    if (data[1] & (1 << sampler)) {
 		if (i + 3 >= count)
@@ -513,6 +1006,9 @@ decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures)
     for (opcode = 0; opcode < sizeof(opcodes_3d_1d) / sizeof(opcodes_3d_1d[0]);
 	 opcode++)
     {
+	if (opcodes_3d_1d[opcode].i830_only && !i830)
+	    continue;
+
 	if (((data[0] & 0x00ff0000) >> 16) == opcodes_3d_1d[opcode].opcode) {
 	    len = 1;
 
@@ -756,7 +1252,7 @@ decode_3d(uint32_t *data, int count, uint32_t hw_offset, int *failures)
     case 0x1f:
 	return decode_3d_primitive(data, count, hw_offset, failures);
     case 0x1d:
-	return decode_3d_1d(data, count, hw_offset, failures);
+	return decode_3d_1d(data, count, hw_offset, failures, 0);
     case 0x1c:
 	return decode_3d_1c(data, count, hw_offset, failures);
     }
@@ -816,10 +1312,71 @@ get_965_depthformat(unsigned int depthformat)
     }
 }
 
+static const char *
+get_965_element_component(uint32_t data, int component)
+{
+    uint32_t component_control = (data >> (16 + (3 - component) * 4)) & 0x7;
+
+    switch (component_control) {
+    case 0:
+	return "nostore";
+    case 1:
+	switch (component) {
+	case 0: return "X";
+	case 1: return "Y";
+	case 2: return "Z";
+	case 3: return "W";
+	default: return "fail";
+	}
+    case 2:
+	return "0.0";
+    case 3:
+	return "1.0";
+    case 4:
+	return "0x1";
+    case 5:
+	return "VID";
+    default:
+	return "fail";
+    }
+}
+
+static const char *
+get_965_prim_type(uint32_t data)
+{
+    uint32_t primtype = (data >> 10) & 0x1f;
+
+    switch (primtype) {
+    case 0x01: return "point list";
+    case 0x02: return "line list";
+    case 0x03: return "line strip";
+    case 0x04: return "tri list";
+    case 0x05: return "tri strip";
+    case 0x06: return "tri fan";
+    case 0x07: return "quad list";
+    case 0x08: return "quad strip";
+    case 0x09: return "line list adj";
+    case 0x0a: return "line strip adj";
+    case 0x0b: return "tri list adj";
+    case 0x0c: return "tri strip adj";
+    case 0x0d: return "tri strip reverse";
+    case 0x0e: return "polygon";
+    case 0x0f: return "rect list";
+    case 0x10: return "line loop";
+    case 0x11: return "point list bf";
+    case 0x12: return "line strip cont";
+    case 0x13: return "line strip bf";
+    case 0x14: return "line strip cont bf";
+    case 0x15: return "tri fan no stipple";
+    default: return "fail";
+    }
+}
+
 static int
 decode_3d_965(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 {
     unsigned int opcode, len;
+    int i;
 
     struct {
 	uint32_t opcode;
@@ -840,8 +1397,7 @@ decode_3d_965(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 	{ 0x780b, 1, 1, "3DSTATE_VF_STATISTICS" },
 	{ 0x7808, 5, 257, "3DSTATE_VERTEX_BUFFERS" },
 	{ 0x7809, 3, 256, "3DSTATE_VERTEX_ELEMENTS" },
-	/* 0x7808: 3DSTATE_VERTEX_BUFFERS */
-	/* 0x7809: 3DSTATE_VERTEX_ELEMENTS */
+	{ 0x780a, 3, 3, "3DSTATE_INDEX_BUFFER" },
 	{ 0x7900, 4, 4, "3DSTATE_DRAWING_RECTANGLE" },
 	{ 0x7901, 5, 5, "3DSTATE_CONSTANT_COLOR" },
 	{ 0x7905, 5, 7, "3DSTATE_DEPTH_BUFFER" },
@@ -927,6 +1483,64 @@ decode_3d_965(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 
 	return len;
 
+    case 0x7808:
+	len = (data[0] & 0xff) + 2;
+	if ((len - 1) % 4 != 0)
+	    fprintf(out, "Bad count in 3DSTATE_VERTEX_BUFFERS\n");
+	if (count < len)
+	    BUFFER_FAIL(count, len, "3DSTATE_VERTEX_BUFFERS");
+	instr_out(data, hw_offset, 0, "3DSTATE_VERTEX_BUFFERS\n");
+
+	for (i = 1; i < len;) {
+	    instr_out(data, hw_offset, i, "buffer %d: %s, pitch %db\n",
+		      data[i] >> 27,
+		      data[i] & (1 << 26) ? "random" : "sequential",
+		      data[i] & 0x07ff);
+	    i++;
+	    instr_out(data, hw_offset, i++, "buffer address\n");
+	    instr_out(data, hw_offset, i++, "max index\n");
+	    instr_out(data, hw_offset, i++, "mbz\n");
+	}
+	return len;
+
+    case 0x7809:
+	len = (data[0] & 0xff) + 2;
+	if ((len + 1) % 2 != 0)
+	    fprintf(out, "Bad count in 3DSTATE_VERTEX_ELEMENTS\n");
+	if (count < len)
+	    BUFFER_FAIL(count, len, "3DSTATE_VERTEX_ELEMENTS");
+	instr_out(data, hw_offset, 0, "3DSTATE_VERTEX_ELEMENTS\n");
+
+	for (i = 1; i < len;) {
+	    instr_out(data, hw_offset, i, "buffer %d: %svalid, type 0x%04x, "
+		      "src offset 0x%04xd bytes\n",
+		      data[i] >> 27,
+		      data[i] & (1 << 26) ? "" : "in",
+		      (data[i] >> 16) & 0x1ff,
+		      data[i] & 0x07ff);
+	    i++;
+	    instr_out(data, hw_offset, i, "(%s, %s, %s, %s), "
+		      "dst offset 0x%02x bytes\n",
+		      get_965_element_component(data[i], 0),
+		      get_965_element_component(data[i], 1),
+		      get_965_element_component(data[i], 2),
+		      get_965_element_component(data[i], 3),
+		      (data[i] & 0xff) * 4);
+	    i++;
+	}
+	return len;
+
+    case 0x780a:
+	len = (data[0] & 0xff) + 2;
+	if (len != 3)
+	    fprintf(out, "Bad count in 3DSTATE_INDEX_BUFFER\n");
+	if (count < len)
+	    BUFFER_FAIL(count, len, "3DSTATE_INDEX_BUFFER");
+	instr_out(data, hw_offset, 0, "3DSTATE_INDEX_BUFFER\n");
+	instr_out(data, hw_offset, 1, "beginning buffer address\n");
+	instr_out(data, hw_offset, 2, "ending buffer address\n");
+	return len;
+
     case 0x7900:
 	if (len != 4)
 	    fprintf(out, "Bad count in 3DSTATE_DRAWING_RECTANGLE\n");
@@ -948,9 +1562,9 @@ decode_3d_965(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 	return len;
 
     case 0x7905:
-	if (len != 5)
+	if (len != 5 && len != 6)
 	    fprintf(out, "Bad count in 3DSTATE_DEPTH_BUFFER\n");
-	if (count < 5)
+	if (count < len)
 	    BUFFER_FAIL(count, len, "3DSTATE_DEPTH_BUFFER");
 
 	instr_out(data, hw_offset, 0,
@@ -965,8 +1579,28 @@ decode_3d_965(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 		  ((data[3] & 0x0007ffc0) >> 6) + 1,
 		  ((data[3] & 0xfff80000) >> 19) + 1);
 	instr_out(data, hw_offset, 4, "volume depth\n");
+	if (len == 6)
+	    instr_out(data, hw_offset, 5, "\n");
 
 	return len;
+
+    case 0x7b00:
+	len = (data[0] & 0xff) + 2;
+	if (len != 6)
+	    fprintf(out, "Bad count in 3DPRIMITIVE\n");
+	if (count < len)
+	    BUFFER_FAIL(count, len, "3DPRIMITIVE");
+
+	instr_out(data, hw_offset, 0,
+		  "3DPRIMITIVE: %s %s\n",
+		  get_965_prim_type(data[0]),
+		  (data[0] & (1 << 15)) ? "random" : "sequential");
+	instr_out(data, hw_offset, 1, "primitive count\n");
+	instr_out(data, hw_offset, 2, "start vertex\n");
+	instr_out(data, hw_offset, 3, "instance count\n");
+	instr_out(data, hw_offset, 4, "start instance\n");
+	instr_out(data, hw_offset, 5, "index bias\n");
+	return len;
     }
 
     for (opcode = 0; opcode < sizeof(opcodes_3d) / sizeof(opcodes_3d[0]);
@@ -999,6 +1633,73 @@ decode_3d_965(uint32_t *data, int count, uint32_t hw_offset, int *failures)
     return 1;
 }
 
+static int
+decode_3d_i830(uint32_t *data, int count, uint32_t hw_offset, int *failures)
+{
+    unsigned int opcode;
+
+    struct {
+	uint32_t opcode;
+	int min_len;
+	int max_len;
+	char *name;
+    } opcodes_3d[] = {
+	{ 0x02, 1, 1, "3DSTATE_MODES_3" },
+	{ 0x03, 1, 1, "3DSTATE_ENABLES_1"},
+	{ 0x04, 1, 1, "3DSTATE_ENABLES_2"},
+	{ 0x05, 1, 1, "3DSTATE_VFT0"},
+	{ 0x06, 1, 1, "3DSTATE_AA"},
+	{ 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES" },
+	{ 0x08, 1, 1, "3DSTATE_MODES_1" },
+	{ 0x09, 1, 1, "3DSTATE_STENCIL_TEST" },
+	{ 0x0a, 1, 1, "3DSTATE_VFT1"},
+	{ 0x0b, 1, 1, "3DSTATE_INDPT_ALPHA_BLEND" },
+	{ 0x0c, 1, 1, "3DSTATE_MODES_5" },
+	{ 0x0d, 1, 1, "3DSTATE_MAP_BLEND_OP" },
+	{ 0x0e, 1, 1, "3DSTATE_MAP_BLEND_ARG" },
+	{ 0x0f, 1, 1, "3DSTATE_MODES_2" },
+	{ 0x15, 1, 1, "3DSTATE_FOG_COLOR" },
+	{ 0x16, 1, 1, "3DSTATE_MODES_4" },
+    };
+
+    switch ((data[0] & 0x1f000000) >> 24) {
+    case 0x1f:
+	return decode_3d_primitive(data, count, hw_offset, failures);
+    case 0x1d:
+	return decode_3d_1d(data, count, hw_offset, failures, 1);
+    case 0x1c:
+	return decode_3d_1c(data, count, hw_offset, failures);
+    }
+
+    for (opcode = 0; opcode < sizeof(opcodes_3d) / sizeof(opcodes_3d[0]);
+	 opcode++) {
+	if ((data[0] & 0x1f000000) >> 24 == opcodes_3d[opcode].opcode) {
+	    unsigned int len = 1, i;
+
+	    instr_out(data, hw_offset, 0, "%s\n", opcodes_3d[opcode].name);
+	    if (opcodes_3d[opcode].max_len > 1) {
+		len = (data[0] & 0xff) + 2;
+		if (len < opcodes_3d[opcode].min_len ||
+		    len > opcodes_3d[opcode].max_len)
+		{
+		    fprintf(out, "Bad count in %s\n", opcodes_3d[opcode].name);
+		}
+	    }
+
+	    for (i = 1; i < len; i++) {
+		if (i >= count)
+		    BUFFER_FAIL(count, len, opcodes_3d[opcode].name);
+		instr_out(data, hw_offset, i, "dword %d\n", i);
+	    }
+	    return len;
+	}
+    }
+
+    instr_out(data, hw_offset, 0, "3D UNKNOWN\n");
+    (*failures)++;
+    return 1;
+}
+
 /**
  * Decodes an i830-i915 batch buffer, writing the output to stdout.
  *
@@ -1028,9 +1729,12 @@ intel_decode(uint32_t *data, int count, uint32_t hw_offset, uint32_t devid)
 	    if (IS_965(devid)) {
 		index += decode_3d_965(data + index, count - index,
 				       hw_offset + index * 4, &failures);
-	    } else {
+	    } else if (IS_9XX(devid)) {
 		index += decode_3d(data + index, count - index,
 				   hw_offset + index * 4, &failures);
+	    } else {
+		index += decode_3d_i830(data + index, count - index,
+					hw_offset + index * 4, &failures);
 	    }
 	    break;
 	default:
diff --git a/src/mesa/drivers/dri/intel/intel_depthstencil.c b/src/mesa/drivers/dri/intel/intel_depthstencil.c
deleted file mode 100644
index c2b4d7728b..0000000000
--- a/src/mesa/drivers/dri/intel/intel_depthstencil.c
+++ /dev/null
@@ -1,252 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/context.h"
-#include "main/depthstencil.h"
-#include "main/fbobject.h"
-#include "main/framebuffer.h"
-#include "main/hash.h"
-#include "main/mtypes.h"
-#include "main/renderbuffer.h"
-
-#include "intel_context.h"
-#include "intel_fbo.h"
-#include "intel_depthstencil.h"
-#include "intel_regions.h"
-#include "intel_span.h"
-
-/**
- * The GL_EXT_framebuffer_object allows the user to create their own
- * framebuffer objects consisting of color renderbuffers (0 or more),
- * depth renderbuffers (0 or 1) and stencil renderbuffers (0 or 1).
- *
- * The spec considers depth and stencil renderbuffers to be totally independent
- * buffers.  In reality, most graphics hardware today uses a combined
- * depth+stencil buffer (one 32-bit pixel = 24 bits of Z + 8 bits of stencil).
- *
- * This causes difficulty because the user may create some number of depth
- * renderbuffers and some number of stencil renderbuffers and bind them
- * together in framebuffers in any combination.
- *
- * This code manages all that.
- *
- * 1. Depth renderbuffers are always allocated in hardware as 32bpp
- *    GL_DEPTH24_STENCIL8 buffers.
- *
- * 2. Stencil renderbuffers are initially allocated in software as 8bpp
- *    GL_STENCIL_INDEX8 buffers.
- *
- * 3. Depth and Stencil renderbuffers use the PairedStencil and PairedDepth
- *    fields (respectively) to indicate if the buffer's currently paired
- *    with another stencil or depth buffer (respectively).
- *
- * 4. When a depth and stencil buffer are initially both attached to the
- *    current framebuffer, we merge the stencil buffer values into the
- *    depth buffer (really a depth+stencil buffer).  The then hardware uses
- *    the combined buffer.
- *
- * 5. Whenever a depth or stencil buffer is reallocated (with
- *    glRenderbufferStorage) we undo the pairing and copy the stencil values
- *    from the combined depth/stencil buffer back to the stencil-only buffer.
- *
- * 6. We also undo the pairing when we find a change in buffer bindings.
- *
- * 7. If a framebuffer is only using a depth renderbuffer (no stencil), we
- *    just use the combined depth/stencil buffer and ignore the stencil values.
- *
- * 8. If a framebuffer is only using a stencil renderbuffer (no depth) we have
- *    to promote the 8bpp software stencil buffer to a 32bpp hardware
- *    depth+stencil buffer.
- *
- */
-
-/**
- * Undo the pairing/interleaving between depth and stencil buffers.
- * irb should be a depth/stencil or stencil renderbuffer.
- */
-void
-intel_unpair_depth_stencil(GLcontext *ctx, struct intel_renderbuffer *irb)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct gl_renderbuffer *rb = &irb->Base;
-
-   if (irb->PairedStencil) {
-      /* irb is a depth/stencil buffer */
-      struct gl_renderbuffer *stencilRb;
-      struct intel_renderbuffer *stencilIrb;
-
-      ASSERT(rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-
-      stencilRb = _mesa_lookup_renderbuffer(ctx, irb->PairedStencil);
-      stencilIrb = intel_renderbuffer(stencilRb);
-      if (stencilIrb) {
-         /* need to extract stencil values from the depth buffer */
-	 ASSERT(stencilIrb->PairedDepth == rb->Name);
-	 intel_renderbuffer_map(intel, rb);
-	 intel_renderbuffer_map(intel, stencilRb);
-	 _mesa_extract_stencil(ctx, rb, stencilRb);
-	 intel_renderbuffer_unmap(intel, stencilRb);
-	 intel_renderbuffer_unmap(intel, rb);
-         stencilIrb->PairedDepth = 0;
-      }
-      irb->PairedStencil = 0;
-   }
-   else if (irb->PairedDepth) {
-      /* irb is a stencil buffer */
-      struct gl_renderbuffer *depthRb;
-      struct intel_renderbuffer *depthIrb;
-
-      ASSERT(rb->_ActualFormat == GL_STENCIL_INDEX8_EXT ||
-             rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-
-      depthRb = _mesa_lookup_renderbuffer(ctx, irb->PairedDepth);
-      depthIrb = intel_renderbuffer(depthRb);
-      if (depthIrb) {
-         /* need to extract stencil values from the depth buffer */
-	 ASSERT(depthIrb->PairedStencil == rb->Name);
-	 intel_renderbuffer_map(intel, rb);
-	 intel_renderbuffer_map(intel, depthRb);
-	 _mesa_extract_stencil(ctx, depthRb, rb);
-	 intel_renderbuffer_unmap(intel, depthRb);
-	 intel_renderbuffer_unmap(intel, rb);
-         depthIrb->PairedStencil = 0;
-      }
-      irb->PairedDepth = 0;
-   }
-   else {
-      _mesa_problem(ctx, "Problem in undo_depth_stencil_pairing");
-   }
-
-   ASSERT(irb->PairedStencil == 0);
-   ASSERT(irb->PairedDepth == 0);
-}
-
-
-/**
- * Examine the depth and stencil renderbuffers which are attached to the
- * framebuffer.  If both depth and stencil are attached, make sure that the
- * renderbuffers are 'paired' (combined).  If only depth or only stencil is
- * attached, undo any previous pairing.
- *
- * Must be called if NewState & _NEW_BUFFER (when renderbuffer attachments
- * change, for example).
- */
-void
-intel_validate_paired_depth_stencil(GLcontext * ctx,
-                                    struct gl_framebuffer *fb)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_renderbuffer *depthRb, *stencilRb;
-
-   depthRb = intel_get_renderbuffer(fb, BUFFER_DEPTH);
-   stencilRb = intel_get_renderbuffer(fb, BUFFER_STENCIL);
-
-   if (depthRb && stencilRb) {
-      if (depthRb == stencilRb) {
-         /* Using a user-created combined depth/stencil buffer.
-          * Nothing to do.
-          */
-         ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_STENCIL_EXT);
-         ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-      }
-      else {
-         /* Separate depth/stencil buffers, need to interleave now */
-         ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_COMPONENT);
-         ASSERT(stencilRb->Base._BaseFormat == GL_STENCIL_INDEX);
-         /* may need to interleave depth/stencil now */
-         if (depthRb->PairedStencil == stencilRb->Base.Name) {
-            /* OK, the depth and stencil buffers are already interleaved */
-            ASSERT(stencilRb->PairedDepth == depthRb->Base.Name);
-         }
-         else {
-            /* need to setup new pairing/interleaving */
-            if (depthRb->PairedStencil) {
-               intel_unpair_depth_stencil(ctx, depthRb);
-            }
-            if (stencilRb->PairedDepth) {
-               intel_unpair_depth_stencil(ctx, stencilRb);
-            }
-
-            ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-            ASSERT(stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT ||
-                   stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-
-            /* establish new pairing: interleave stencil into depth buffer */
-	    intel_renderbuffer_map(intel, &depthRb->Base);
-	    intel_renderbuffer_map(intel, &stencilRb->Base);
-            _mesa_insert_stencil(ctx, &depthRb->Base, &stencilRb->Base);
-	    intel_renderbuffer_unmap(intel, &stencilRb->Base);
-	    intel_renderbuffer_unmap(intel, &depthRb->Base);
-            depthRb->PairedStencil = stencilRb->Base.Name;
-            stencilRb->PairedDepth = depthRb->Base.Name;
-         }
-
-      }
-   }
-   else if (depthRb) {
-      /* Depth buffer but no stencil buffer.
-       * We'll use a GL_DEPTH24_STENCIL8 buffer and ignore the stencil bits.
-       */
-      /* can't assert this until storage is allocated:
-         ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-       */
-      /* intel_undo any previous pairing */
-      if (depthRb->PairedStencil) {
-         intel_unpair_depth_stencil(ctx, depthRb);
-      }
-   }
-   else if (stencilRb) {
-      /* Stencil buffer but no depth buffer.
-       * Since h/w doesn't typically support just 8bpp stencil w/out Z,
-       * we'll use a GL_DEPTH24_STENCIL8 buffer and ignore the depth bits.
-       */
-      /* undo any previous pairing */
-      if (stencilRb->PairedDepth) {
-         intel_unpair_depth_stencil(ctx, stencilRb);
-      }
-      if (stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT) {
-         /* promote buffer to GL_DEPTH24_STENCIL8 for hw rendering */
-         _mesa_promote_stencil(ctx, &stencilRb->Base);
-         ASSERT(stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-      }
-   }
-
-   /* Finally, update the fb->_DepthBuffer and fb->_StencilBuffer fields */
-   _mesa_update_depth_buffer(ctx, fb, BUFFER_DEPTH);
-   if (depthRb && depthRb->PairedStencil)
-      _mesa_update_stencil_buffer(ctx, fb, BUFFER_DEPTH);
-   else
-      _mesa_update_stencil_buffer(ctx, fb, BUFFER_STENCIL);
-
-
-   /* The hardware should use fb->Attachment[BUFFER_DEPTH].Renderbuffer
-    * first, if present, then fb->Attachment[BUFFER_STENCIL].Renderbuffer
-    * if present.
-    */
-}
diff --git a/src/mesa/drivers/dri/intel/intel_depthstencil.h b/src/mesa/drivers/dri/intel/intel_depthstencil.h
deleted file mode 100644
index 740eb0d989..0000000000
--- a/src/mesa/drivers/dri/intel/intel_depthstencil.h
+++ /dev/null
@@ -1,15 +0,0 @@
-
-#ifndef INTEL_DEPTH_STENCIL_H
-#define INTEL_DEPTH_STENCIL_H
-
-#include "intel_fbo.h"
-
-extern void
-intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb);
-
-extern void
-intel_validate_paired_depth_stencil(GLcontext * ctx,
-                                    struct gl_framebuffer *fb);
-
-
-#endif /* INTEL_DEPTH_STENCIL_H */
diff --git a/src/mesa/drivers/dri/intel/intel_extensions.c b/src/mesa/drivers/dri/intel/intel_extensions.c
new file mode 100644
index 0000000000..28223ca141
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_extensions.c
@@ -0,0 +1,189 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "intel_chipset.h"
+#include "intel_context.h"
+#include "intel_extensions.h"
+
+
+#define need_GL_ARB_framebuffer_object
+#define need_GL_ARB_occlusion_query
+#define need_GL_ARB_point_parameters
+#define need_GL_ARB_shader_objects
+#define need_GL_ARB_vertex_program
+#define need_GL_ARB_vertex_shader
+#define need_GL_ARB_window_pos
+#define need_GL_EXT_blend_color
+#define need_GL_EXT_blend_equation_separate
+#define need_GL_EXT_blend_func_separate
+#define need_GL_EXT_blend_minmax
+#define need_GL_EXT_cull_vertex
+#define need_GL_EXT_fog_coord
+#define need_GL_EXT_framebuffer_object
+#define need_GL_EXT_framebuffer_blit
+#define need_GL_EXT_point_parameters
+#define need_GL_EXT_secondary_color
+#define need_GL_EXT_stencil_two_side
+#define need_GL_ATI_separate_stencil
+#define need_GL_NV_point_sprite
+#define need_GL_NV_vertex_program
+#define need_GL_VERSION_2_0
+#define need_GL_VERSION_2_1
+
+#include "extension_helper.h"
+
+
+/**
+ * Extension strings exported by the intel driver.
+ *
+ * Extensions supported by all chips supported by i830_dri, i915_dri, or
+ * i965_dri.
+ */
+static const struct dri_extension card_extensions[] = {
+   { "GL_ARB_multitexture",               NULL },
+   { "GL_ARB_point_parameters",           GL_ARB_point_parameters_functions },
+   { "GL_ARB_texture_border_clamp",       NULL },
+   { "GL_ARB_texture_cube_map",           NULL },
+   { "GL_ARB_texture_env_add",            NULL },
+   { "GL_ARB_texture_env_combine",        NULL },
+   { "GL_ARB_texture_env_crossbar",       NULL },
+   { "GL_ARB_texture_env_dot3",           NULL },
+   { "GL_ARB_texture_mirrored_repeat",    NULL },
+   { "GL_ARB_texture_rectangle",          NULL },
+   { "GL_ARB_vertex_program",             GL_ARB_vertex_program_functions },
+   { "GL_ARB_window_pos",                 GL_ARB_window_pos_functions },
+   { "GL_EXT_blend_color",                GL_EXT_blend_color_functions },
+   { "GL_EXT_blend_equation_separate",    GL_EXT_blend_equation_separate_functions },
+   { "GL_EXT_blend_func_separate",        GL_EXT_blend_func_separate_functions },
+   { "GL_EXT_blend_minmax",               GL_EXT_blend_minmax_functions },
+   { "GL_EXT_blend_logic_op",             NULL },
+   { "GL_EXT_blend_subtract",             NULL },
+   { "GL_EXT_cull_vertex",                GL_EXT_cull_vertex_functions },
+   { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+   { "GL_EXT_packed_depth_stencil",       NULL },
+   { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
+   { "GL_EXT_stencil_wrap",               NULL },
+   { "GL_EXT_texture_edge_clamp",         NULL },
+   { "GL_EXT_texture_env_combine",        NULL },
+   { "GL_EXT_texture_env_dot3",           NULL },
+   { "GL_EXT_texture_filter_anisotropic", NULL },
+   { "GL_EXT_texture_lod_bias",           NULL },
+   { "GL_3DFX_texture_compression_FXT1",  NULL },
+   { "GL_APPLE_client_storage",           NULL },
+   { "GL_MESA_pack_invert",               NULL },
+   { "GL_MESA_ycbcr_texture",             NULL },
+   { "GL_NV_blend_square",                NULL },
+   { "GL_NV_point_sprite",                GL_NV_point_sprite_functions },
+   { "GL_NV_vertex_program",              GL_NV_vertex_program_functions },
+   { "GL_NV_vertex_program1_1",           NULL },
+   { "GL_SGIS_generate_mipmap",           NULL },
+   { NULL, NULL }
+};
+
+
+/** i915 / i945-only extensions */
+static const struct dri_extension i915_extensions[] = {
+   { "GL_ARB_depth_texture",              NULL },
+   { "GL_ARB_fragment_program",           NULL },
+   { "GL_ARB_shadow",                     NULL },
+   { "GL_ARB_texture_non_power_of_two",   NULL },
+   { "GL_ATI_texture_env_combine3",       NULL },
+   { "GL_EXT_shadow_funcs",               NULL },
+   { "GL_NV_texture_env_combine4",        NULL },
+   { NULL,                                NULL }
+};
+
+
+/** i965-only extensions */
+static const struct dri_extension brw_extensions[] = {
+   { "GL_ARB_depth_texture",              NULL },
+   { "GL_ARB_fragment_program",           NULL },
+   { "GL_ARB_fragment_program_shadow",    NULL },
+   { "GL_ARB_fragment_shader",            NULL },
+   { "GL_ARB_framebuffer_object",         GL_ARB_framebuffer_object_functions},
+   { "GL_ARB_occlusion_query",            GL_ARB_occlusion_query_functions },
+   { "GL_ARB_point_sprite", 		  NULL },
+   { "GL_ARB_shader_objects",             GL_ARB_shader_objects_functions },
+   { "GL_ARB_shading_language_100",       GL_VERSION_2_0_functions },
+#if 0
+   /* Support for GLSL 1.20 is currently broken in core Mesa.
+    */
+   { "GL_ARB_shading_language_120",       GL_VERSION_2_1_functions },
+#endif
+   { "GL_ARB_shadow",                     NULL },
+   { "GL_ARB_texture_non_power_of_two",   NULL },
+   { "GL_ARB_vertex_shader",              GL_ARB_vertex_shader_functions },
+   { "GL_EXT_shadow_funcs",               NULL },
+   { "GL_EXT_stencil_two_side",           GL_EXT_stencil_two_side_functions },
+   { "GL_EXT_texture_sRGB",		  NULL },
+   { "GL_EXT_texture_swizzle",		  NULL },
+   { "GL_EXT_vertex_array_bgra",	  NULL },
+   { "GL_ATI_separate_stencil",           GL_ATI_separate_stencil_functions },
+   { "GL_ATI_texture_env_combine3",       NULL },
+   { "GL_NV_texture_env_combine4",        NULL },
+   { NULL,                                NULL }
+};
+
+
+static const struct dri_extension arb_oq_extensions[] = {
+   { NULL, NULL }
+};
+
+
+static const struct dri_extension ttm_extensions[] = {
+   { "GL_ARB_pixel_buffer_object",      NULL },
+   { "GL_EXT_framebuffer_blit",         GL_EXT_framebuffer_blit_functions },
+   { "GL_EXT_framebuffer_object",       GL_EXT_framebuffer_object_functions },
+   { NULL, NULL }
+};
+
+
+/**
+ * Initializes potential list of extensions if ctx == NULL, or actually enables
+ * extensions for a context.
+ */
+void
+intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging)
+{
+   struct intel_context *intel = ctx?intel_context(ctx):NULL;
+
+   /* Disable imaging extension until convolution is working in teximage paths.
+    */
+   enable_imaging = GL_FALSE;
+
+   driInitExtensions(ctx, card_extensions, enable_imaging);
+
+   if (intel == NULL || intel->ttm)
+      driInitExtensions(ctx, ttm_extensions, GL_FALSE);
+
+   if (intel == NULL || IS_965(intel->intelScreen->deviceID))
+      driInitExtensions(ctx, brw_extensions, GL_FALSE);
+
+   if (intel == NULL || IS_915(intel->intelScreen->deviceID)
+       || IS_945(intel->intelScreen->deviceID))
+      driInitExtensions(ctx, i915_extensions, GL_FALSE);
+}
diff --git a/src/mesa/drivers/dri/intel/intel_extensions.h b/src/mesa/drivers/dri/intel/intel_extensions.h
new file mode 100644
index 0000000000..97147ecdb0
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_extensions.h
@@ -0,0 +1,36 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef INTEL_EXTENSIONS_H
+#define INTEL_EXTENSIONS_H
+
+
+extern void
+intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging);
+
+
+#endif
diff --git a/src/mesa/drivers/dri/intel/intel_fbo.c b/src/mesa/drivers/dri/intel/intel_fbo.c
index fce5e36b9d..54d6044ad3 100644
--- a/src/mesa/drivers/dri/intel/intel_fbo.c
+++ b/src/mesa/drivers/dri/intel/intel_fbo.c
@@ -27,6 +27,7 @@
 
 
 #include "main/imports.h"
+#include "main/macros.h"
 #include "main/mtypes.h"
 #include "main/fbobject.h"
 #include "main/framebuffer.h"
@@ -37,95 +38,13 @@
 
 #include "intel_context.h"
 #include "intel_buffers.h"
-#include "intel_depthstencil.h"
 #include "intel_fbo.h"
 #include "intel_mipmap_tree.h"
 #include "intel_regions.h"
-#include "intel_span.h"
 
 
 #define FILE_DEBUG_FLAG DEBUG_FBO
 
-#define INTEL_RB_CLASS 0x12345678
-
-
-/* XXX FBO: move this to intel_context.h (inlined) */
-/**
- * Return a gl_renderbuffer ptr casted to intel_renderbuffer.
- * NULL will be returned if the rb isn't really an intel_renderbuffer.
- * This is determiend by checking the ClassID.
- */
-struct intel_renderbuffer *
-intel_renderbuffer(struct gl_renderbuffer *rb)
-{
-   struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb;
-   if (irb && irb->Base.ClassID == INTEL_RB_CLASS) {
-      /*_mesa_warning(NULL, "Returning non-intel Rb\n");*/
-      return irb;
-   }
-   else
-      return NULL;
-}
-
-
-struct intel_renderbuffer *
-intel_get_renderbuffer(struct gl_framebuffer *fb, int attIndex)
-{
-   if (attIndex >= 0)
-      return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer);
-   else
-      return NULL;
-}
-
-
-void
-intel_flip_renderbuffers(struct intel_framebuffer *intel_fb)
-{
-   int current_page = intel_fb->pf_current_page;
-   int next_page = (current_page + 1) % intel_fb->pf_num_pages;
-   struct gl_renderbuffer *tmp_rb;
-
-   /* Exchange renderbuffers if necessary but make sure their reference counts
-    * are preserved.
-    */
-   if (intel_fb->color_rb[current_page] &&
-       intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer !=
-       &intel_fb->color_rb[current_page]->Base) {
-      tmp_rb = NULL;
-      _mesa_reference_renderbuffer(&tmp_rb,
-	 intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
-      tmp_rb = &intel_fb->color_rb[current_page]->Base;
-      _mesa_reference_renderbuffer(
-	 &intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer, tmp_rb);
-      _mesa_reference_renderbuffer(&tmp_rb, NULL);
-   }
-
-   if (intel_fb->color_rb[next_page] &&
-       intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer !=
-       &intel_fb->color_rb[next_page]->Base) {
-      tmp_rb = NULL;
-      _mesa_reference_renderbuffer(&tmp_rb,
-	 intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer);
-      tmp_rb = &intel_fb->color_rb[next_page]->Base;
-      _mesa_reference_renderbuffer(
-	 &intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer, tmp_rb);
-      _mesa_reference_renderbuffer(&tmp_rb, NULL);
-   }
-}
-
-
-struct intel_region *
-intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex)
-{
-   struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex);
-
-   if (irb)
-      return irb->region;
-   else
-      return NULL;
-}
-
-
 
 /**
  * Create a new framebuffer object.
@@ -140,6 +59,7 @@ intel_new_framebuffer(GLcontext * ctx, GLuint name)
 }
 
 
+/** Called by gl_renderbuffer::Delete() */
 static void
 intel_delete_renderbuffer(struct gl_renderbuffer *rb)
 {
@@ -149,10 +69,6 @@ intel_delete_renderbuffer(struct gl_renderbuffer *rb)
 
    ASSERT(irb);
 
-   if (irb->PairedStencil || irb->PairedDepth) {
-      intel_unpair_depth_stencil(ctx, irb);
-   }
-
    if (irb->span_cache != NULL)
       _mesa_free(irb->span_cache);
 
@@ -164,7 +80,6 @@ intel_delete_renderbuffer(struct gl_renderbuffer *rb)
 }
 
 
-
 /**
  * Return a pointer to a specific pixel in a renderbuffer.
  */
@@ -179,7 +94,6 @@ intel_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb,
 }
 
 
-
 /**
  * Called via glRenderbufferStorageEXT() to set the format and allocate
  * storage for a user-created renderbuffer.
@@ -310,7 +224,6 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
 }
 
 
-
 /**
  * Called for each hardware renderbuffer when a _window_ is resized.
  * Just update fields.
@@ -328,6 +241,7 @@ intel_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
    return GL_TRUE;
 }
 
+
 static void
 intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb,
 		     GLuint width, GLuint height)
@@ -344,7 +258,7 @@ intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb,
    }
 
    /* Make sure all window system renderbuffers are up to date */
-   for (i = 0; i < 3; i++) {
+   for (i = 0; i < 2; i++) {
       struct gl_renderbuffer *rb = &intel_fb->color_rb[i]->Base;
 
       /* only resize if size is changing */
@@ -354,6 +268,8 @@ intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb,
    }
 }
 
+
+/** Dummy function for gl_renderbuffer::AllocStorage() */
 static GLboolean
 intel_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
                         GLenum internalFormat, GLuint width, GLuint height)
@@ -373,10 +289,9 @@ intel_renderbuffer_set_region(struct intel_renderbuffer *rb,
    rb->region = NULL;
    intel_region_reference(&rb->region, region);
    intel_region_release(&old);
-
-   rb->pfPitch = region->pitch;
 }
 
+
 /**
  * Create a new intel_renderbuffer which corresponds to an on-screen window,
  * not a user-created renderbuffer.
@@ -496,9 +411,6 @@ intel_bind_framebuffer(GLcontext * ctx, GLenum target,
 {
    if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) {
       intel_draw_buffer(ctx, fb);
-      /* Integer depth range depends on depth buffer bits */
-      if (ctx->Driver.DepthRange != NULL)
-	 ctx->Driver.DepthRange(ctx, ctx->Viewport.Near, ctx->Viewport.Far);
    }
    else {
       /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */
@@ -522,6 +434,7 @@ intel_framebuffer_renderbuffer(GLcontext * ctx,
    intel_draw_buffer(ctx, fb);
 }
 
+
 static GLboolean
 intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb, 
 		     struct gl_texture_image *texImage)
@@ -529,20 +442,25 @@ intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb,
    if (texImage->TexFormat == &_mesa_texformat_argb8888) {
       irb->Base._ActualFormat = GL_RGBA8;
       irb->Base._BaseFormat = GL_RGBA;
+      irb->Base.DataType = GL_UNSIGNED_BYTE;
       DBG("Render to RGBA8 texture OK\n");
    }
    else if (texImage->TexFormat == &_mesa_texformat_rgb565) {
       irb->Base._ActualFormat = GL_RGB5;
       irb->Base._BaseFormat = GL_RGB;
+      irb->Base.DataType = GL_UNSIGNED_SHORT;
       DBG("Render to RGB5 texture OK\n");
    }
    else if (texImage->TexFormat == &_mesa_texformat_z16) {
       irb->Base._ActualFormat = GL_DEPTH_COMPONENT16;
       irb->Base._BaseFormat = GL_DEPTH_COMPONENT;
+      irb->Base.DataType = GL_UNSIGNED_SHORT;
       DBG("Render to DEPTH16 texture OK\n");
-   } else if (texImage->TexFormat == &_mesa_texformat_s8_z24) {
+   }
+   else if (texImage->TexFormat == &_mesa_texformat_s8_z24) {
       irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
       irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT;
+      irb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT;
       DBG("Render to DEPTH_STENCIL texture OK\n");
    }
    else {
@@ -554,7 +472,6 @@ intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb,
    irb->Base.InternalFormat = irb->Base._ActualFormat;
    irb->Base.Width = texImage->Width;
    irb->Base.Height = texImage->Height;
-   irb->Base.DataType = GL_UNSIGNED_BYTE;       /* FBO XXX fix */
    irb->Base.RedBits = texImage->TexFormat->RedBits;
    irb->Base.GreenBits = texImage->TexFormat->GreenBits;
    irb->Base.BlueBits = texImage->TexFormat->BlueBits;
@@ -564,11 +481,10 @@ intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb,
    irb->Base.Delete = intel_delete_renderbuffer;
    irb->Base.AllocStorage = intel_nop_alloc_storage;
 
-   irb->RenderToTexture = GL_TRUE;
-
    return GL_TRUE;
 }
 
+
 /**
  * When glFramebufferTexture[123]D is called this function sets up the
  * gl_renderbuffer wrapper around the texture image.
@@ -577,7 +493,7 @@ intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb,
 static struct intel_renderbuffer *
 intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage)
 {
-   const GLuint name = ~0;      /* not significant, but distinct for debugging */
+   const GLuint name = ~0;   /* not significant, but distinct for debugging */
    struct intel_renderbuffer *irb;
 
    /* make an intel_renderbuffer to wrap the texture image */
@@ -620,7 +536,15 @@ intel_render_texture(GLcontext * ctx,
 
    ASSERT(newImage);
 
-   if (!irb) {
+   if (newImage->Border != 0) {
+      /* Fallback on drawing to a texture with a border, which won't have a
+       * miptree.
+       */
+      _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
+      _mesa_render_texture(ctx, fb, att);
+      return;
+   }
+   else if (!irb) {
       irb = intel_wrap_texture(ctx, newImage);
       if (irb) {
          /* bind the wrapper to the attachment point */
@@ -631,7 +555,9 @@ intel_render_texture(GLcontext * ctx,
          _mesa_render_texture(ctx, fb, att);
          return;
       }
-   } if (!intel_update_wrapper(ctx, irb, newImage)) {
+   }
+
+   if (!intel_update_wrapper(ctx, irb, newImage)) {
        _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
        _mesa_render_texture(ctx, fb, att);
        return;
@@ -693,6 +619,94 @@ intel_finish_render_texture(GLcontext * ctx,
 
 
 /**
+ * Do additional "completeness" testing of a framebuffer object.
+ */
+static void
+intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+   const struct intel_renderbuffer *depthRb =
+      intel_get_renderbuffer(fb, BUFFER_DEPTH);
+   const struct intel_renderbuffer *stencilRb =
+      intel_get_renderbuffer(fb, BUFFER_STENCIL);
+
+   if (stencilRb && stencilRb != depthRb) {
+      /* we only support combined depth/stencil buffers, not separate
+       * stencil buffers.
+       */
+      fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT;
+   }
+}
+
+
+/**
+ * Called from glBlitFramebuffer().
+ * For now, we're doing an approximation with glCopyPixels().
+ * XXX we need to bypass all the per-fragment operations, except scissor.
+ */
+static void
+intel_blit_framebuffer(GLcontext *ctx,
+                       GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                       GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                       GLbitfield mask, GLenum filter)
+{
+   const GLfloat xZoomSave = ctx->Pixel.ZoomX;
+   const GLfloat yZoomSave = ctx->Pixel.ZoomY;
+   GLsizei width, height;
+   GLfloat xFlip = 1.0F, yFlip = 1.0F;
+
+   if (srcX1 < srcX0) {
+      GLint tmp = srcX1;
+      srcX1 = srcX0;
+      srcX0 = tmp;
+      xFlip = -1.0F;
+   }
+
+   if (srcY1 < srcY0) {
+      GLint tmp = srcY1;
+      srcY1 = srcY0;
+      srcY0 = tmp;
+      yFlip = -1.0F;
+   }
+
+   width = srcX1 - srcX0;
+   height = srcY1 - srcY0;
+
+   ctx->Pixel.ZoomX = xFlip * (dstX1 - dstX0) / (srcX1 - srcY0);
+   ctx->Pixel.ZoomY = yFlip * (dstY1 - dstY0) / (srcY1 - srcY0);
+
+   if (ctx->Pixel.ZoomX < 0.0F) {
+      dstX0 = MAX2(dstX0, dstX1);
+   }
+   else {
+      dstX0 = MIN2(dstX0, dstX1);
+   }
+
+   if (ctx->Pixel.ZoomY < 0.0F) {
+      dstY0 = MAX2(dstY0, dstY1);
+   }
+   else {
+      dstY0 = MIN2(dstY0, dstY1);
+   }
+
+   if (mask & GL_COLOR_BUFFER_BIT) {
+      ctx->Driver.CopyPixels(ctx, srcX0, srcY0, width, height,
+                             dstX0, dstY0, GL_COLOR);
+   }
+   if (mask & GL_DEPTH_BUFFER_BIT) {
+      ctx->Driver.CopyPixels(ctx, srcX0, srcY0, width, height,
+                             dstX0, dstY0, GL_DEPTH);
+   }
+   if (mask & GL_STENCIL_BUFFER_BIT) {
+      ctx->Driver.CopyPixels(ctx, srcX0, srcY0, width, height,
+                             dstX0, dstY0, GL_STENCIL);
+   }
+      
+   ctx->Pixel.ZoomX = xZoomSave;
+   ctx->Pixel.ZoomY = yZoomSave;
+}
+
+
+/**
  * Do one-time context initializations related to GL_EXT_framebuffer_object.
  * Hook in device driver functions.
  */
@@ -706,4 +720,6 @@ intel_fbo_init(struct intel_context *intel)
    intel->ctx.Driver.RenderTexture = intel_render_texture;
    intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture;
    intel->ctx.Driver.ResizeBuffers = intel_resize_buffers;
+   intel->ctx.Driver.ValidateFramebuffer = intel_validate_framebuffer;
+   intel->ctx.Driver.BlitFramebuffer = intel_blit_framebuffer;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_fbo.h b/src/mesa/drivers/dri/intel/intel_fbo.h
index 9d15582d78..7226ee026f 100644
--- a/src/mesa/drivers/dri/intel/intel_fbo.h
+++ b/src/mesa/drivers/dri/intel/intel_fbo.h
@@ -39,14 +39,7 @@ struct intel_framebuffer
 {
    struct gl_framebuffer Base;
 
-   struct intel_renderbuffer *color_rb[3];
-
-   /* Drawable page flipping state */
-   GLboolean pf_active;
-   GLuint pf_seq;
-   GLint pf_planes;
-   GLint pf_current_page;
-   GLint pf_num_pages;
+   struct intel_renderbuffer *color_rb[2];
 
    /* VBI
     */
@@ -62,21 +55,11 @@ struct intel_framebuffer
 
 /**
  * Intel renderbuffer, derived from gl_renderbuffer.
- * Note: The PairedDepth and PairedStencil fields use renderbuffer IDs,
- * not pointers because in some circumstances a deleted renderbuffer could
- * result in a dangling pointer here.
  */
 struct intel_renderbuffer
 {
    struct gl_renderbuffer Base;
    struct intel_region *region;
-   GLuint pfPitch;              /* possibly paged flipped pitch */
-   GLboolean RenderToTexture;   /* RTT? */
-
-   GLuint PairedDepth;   /**< only used if this is a depth renderbuffer */
-   GLuint PairedStencil; /**< only used if this is a stencil renderbuffer */
-
-   GLuint pf_pending;  /**< sequence number of pending flip */
 
    GLuint vbl_pending;   /**< vblank sequence number of pending flip */
 
@@ -84,32 +67,71 @@ struct intel_renderbuffer
    unsigned long span_cache_offset;
 };
 
-extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer
-                                                     *rb);
+
+/**
+ * gl_renderbuffer is a base class which we subclass.  The Class field
+ * is used for simple run-time type checking.
+ */
+#define INTEL_RB_CLASS 0x12345678
+
+
+/**
+ * Return a gl_renderbuffer ptr casted to intel_renderbuffer.
+ * NULL will be returned if the rb isn't really an intel_renderbuffer.
+ * This is determined by checking the ClassID.
+ */
+static INLINE struct intel_renderbuffer *
+intel_renderbuffer(struct gl_renderbuffer *rb)
+{
+   struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb;
+   if (irb && irb->Base.ClassID == INTEL_RB_CLASS) {
+      /*_mesa_warning(NULL, "Returning non-intel Rb\n");*/
+      return irb;
+   }
+   else
+      return NULL;
+}
+
+
+/**
+ * Return a framebuffer's renderbuffer, named by a BUFFER_x index.
+ */
+static INLINE struct intel_renderbuffer *
+intel_get_renderbuffer(struct gl_framebuffer *fb, int attIndex)
+{
+   if (attIndex >= 0)
+      return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer);
+   else
+      return NULL;
+}
+
 
 extern void
 intel_renderbuffer_set_region(struct intel_renderbuffer *irb,
 			      struct intel_region *region);
 
+
 extern struct intel_renderbuffer *
 intel_create_renderbuffer(GLenum intFormat);
 
-extern void intel_fbo_init(struct intel_context *intel);
 
-
-/* XXX make inline or macro */
-extern struct intel_renderbuffer *intel_get_renderbuffer(struct gl_framebuffer
-                                                         *fb,
-                                                         int attIndex);
-
-extern void intel_flip_renderbuffers(struct intel_framebuffer *intel_fb);
+extern void
+intel_fbo_init(struct intel_context *intel);
 
 
-/* XXX make inline or macro */
-extern struct intel_region *intel_get_rb_region(struct gl_framebuffer *fb,
-                                                GLuint attIndex);
+extern void
+intel_flip_renderbuffers(struct intel_framebuffer *intel_fb);
 
 
+static INLINE struct intel_region *
+intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex)
+{
+   struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex);
+   if (irb)
+      return irb->region;
+   else
+      return NULL;
+}
 
 
 #endif /* INTEL_FBO_H */
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
index b96ba72853..bf1c3f03f0 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
@@ -111,9 +111,9 @@ intel_miptree_create(struct intel_context *intel,
 				      first_level, last_level, width0,
 				      height0, depth0, cpp, compress_byte);
    /*
-    * pitch == 0 indicates the null texture
+    * pitch == 0 || height == 0  indicates the null texture
     */
-   if (!mt || !mt->pitch)
+   if (!mt || !mt->pitch || !mt->total_height)
       return NULL;
 
    mt->region = intel_region_alloc(intel,
@@ -163,7 +163,7 @@ intel_miptree_create_for_region(struct intel_context *intel,
    mt->pitch = region->pitch;
 #endif
 
-   mt->region = region;
+   intel_region_reference(&mt->region, region);
 
    return mt;
  }
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c
index b267ffd890..5e32288844 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel.c
@@ -27,7 +27,12 @@
 
 #include "main/enums.h"
 #include "main/state.h"
+#include "main/context.h"
+#include "main/enable.h"
+#include "main/matrix.h"
 #include "swrast/swrast.h"
+#include "shader/arbprogram.h"
+#include "shader/program.h"
 
 #include "intel_context.h"
 #include "intel_pixel.h"
@@ -112,12 +117,6 @@ intel_check_blit_fragment_ops(GLcontext * ctx, GLboolean src_alpha_is_one)
       return GL_FALSE;
    }
 
-   if (ctx->Scissor.Enabled) {
-      /* XXX Note: Scissor could be done with the blitter */
-      DBG("fallback due to image scissor\n");
-      return GL_FALSE;
-   }
-
    if (ctx->RenderMode != GL_RENDER) {
       DBG("fallback due to render mode\n");
       return GL_FALSE;
@@ -173,6 +172,162 @@ intel_check_blit_format(struct intel_region * region,
    return GL_FALSE;
 }
 
+void
+intel_meta_set_passthrough_transform(struct intel_context *intel)
+{
+   GLcontext *ctx = &intel->ctx;
+
+   intel->meta.saved_vp_x = ctx->Viewport.X;
+   intel->meta.saved_vp_y = ctx->Viewport.Y;
+   intel->meta.saved_vp_width = ctx->Viewport.Width;
+   intel->meta.saved_vp_height = ctx->Viewport.Height;
+   intel->meta.saved_matrix_mode = ctx->Transform.MatrixMode;
+
+   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
+
+   _mesa_MatrixMode(GL_PROJECTION);
+   _mesa_PushMatrix();
+   _mesa_LoadIdentity();
+   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
+
+   _mesa_MatrixMode(GL_MODELVIEW);
+   _mesa_PushMatrix();
+   _mesa_LoadIdentity();
+}
+
+void
+intel_meta_restore_transform(struct intel_context *intel)
+{
+   _mesa_MatrixMode(GL_PROJECTION);
+   _mesa_PopMatrix();
+   _mesa_MatrixMode(GL_MODELVIEW);
+   _mesa_PopMatrix();
+
+   _mesa_MatrixMode(intel->meta.saved_matrix_mode);
+
+   _mesa_Viewport(intel->meta.saved_vp_x, intel->meta.saved_vp_y,
+		  intel->meta.saved_vp_width, intel->meta.saved_vp_height);
+}
+
+/**
+ * Set up a vertex program to pass through the position and first texcoord
+ * for pixel path.
+ */
+void
+intel_meta_set_passthrough_vertex_program(struct intel_context *intel)
+{
+   GLcontext *ctx = &intel->ctx;
+   static const char *vp =
+      "!!ARBvp1.0\n"
+      "TEMP vertexClip;\n"
+      "DP4 vertexClip.x, state.matrix.mvp.row[0], vertex.position;\n"
+      "DP4 vertexClip.y, state.matrix.mvp.row[1], vertex.position;\n"
+      "DP4 vertexClip.z, state.matrix.mvp.row[2], vertex.position;\n"
+      "DP4 vertexClip.w, state.matrix.mvp.row[3], vertex.position;\n"
+      "MOV result.position, vertexClip;\n"
+      "MOV result.texcoord[0], vertex.texcoord[0];\n"
+      "MOV result.color, vertex.color;\n"
+      "END\n";
+
+   assert(intel->meta.saved_vp == NULL);
+
+   _mesa_reference_vertprog(ctx, &intel->meta.saved_vp,
+			    ctx->VertexProgram.Current);
+   if (intel->meta.passthrough_vp == NULL) {
+      GLuint prog_name;
+      _mesa_GenPrograms(1, &prog_name);
+      _mesa_BindProgram(GL_VERTEX_PROGRAM_ARB, prog_name);
+      _mesa_ProgramStringARB(GL_VERTEX_PROGRAM_ARB,
+			     GL_PROGRAM_FORMAT_ASCII_ARB,
+			     strlen(vp), (const GLubyte *)vp);
+      _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp,
+			       ctx->VertexProgram.Current);
+      _mesa_DeletePrograms(1, &prog_name);
+   }
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
+			    intel->meta.passthrough_vp);
+   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
+			   &intel->meta.passthrough_vp->Base);
+
+   intel->meta.saved_vp_enable = ctx->VertexProgram.Enabled;
+   _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
+}
+
+/**
+ * Restores the previous vertex program after
+ * intel_meta_set_passthrough_vertex_program()
+ */
+void
+intel_meta_restore_vertex_program(struct intel_context *intel)
+{
+   GLcontext *ctx = &intel->ctx;
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
+			    intel->meta.saved_vp);
+   _mesa_reference_vertprog(ctx, &intel->meta.saved_vp, NULL);
+   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
+			   &ctx->VertexProgram.Current->Base);
+
+   if (!intel->meta.saved_vp_enable)
+      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
+}
+
+/**
+ * Binds the given program string to GL_FRAGMENT_PROGRAM_ARB, caching the
+ * program object.
+ */
+void
+intel_meta_set_fragment_program(struct intel_context *intel,
+				struct gl_fragment_program **prog,
+				const char *prog_string)
+{
+   GLcontext *ctx = &intel->ctx;
+   assert(intel->meta.saved_fp == NULL);
+
+   _mesa_reference_fragprog(ctx, &intel->meta.saved_fp,
+			    ctx->FragmentProgram.Current);
+   if (*prog == NULL) {
+      GLuint prog_name;
+      _mesa_GenPrograms(1, &prog_name);
+      _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, prog_name);
+      _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB,
+			     GL_PROGRAM_FORMAT_ASCII_ARB,
+			     strlen(prog_string), (const GLubyte *)prog_string);
+      _mesa_reference_fragprog(ctx, prog, ctx->FragmentProgram.Current);
+      /* Note that DeletePrograms unbinds the program on us */
+      _mesa_DeletePrograms(1, &prog_name);
+   }
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, *prog);
+   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, &((*prog)->Base));
+
+   intel->meta.saved_fp_enable = ctx->FragmentProgram.Enabled;
+   _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
+}
+
+/**
+ * Restores the previous fragment program after
+ * intel_meta_set_fragment_program()
+ */
+void
+intel_meta_restore_fragment_program(struct intel_context *intel)
+{
+   GLcontext *ctx = &intel->ctx;
+
+   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
+   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
+			    intel->meta.saved_fp);
+   _mesa_reference_fragprog(ctx, &intel->meta.saved_fp, NULL);
+   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
+			   &ctx->FragmentProgram.Current->Base);
+
+   if (!intel->meta.saved_fp_enable)
+      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
+}
 
 void
 intelInitPixelFuncs(struct dd_function_table *functions)
@@ -187,3 +342,13 @@ intelInitPixelFuncs(struct dd_function_table *functions)
 #endif
    }
 }
+
+void
+intel_free_pixel_state(struct intel_context *intel)
+{
+   GLcontext *ctx = &intel->ctx;
+
+   _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp, NULL);
+   _mesa_reference_fragprog(ctx, &intel->meta.bitmap_fp, NULL);
+}
+
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.h b/src/mesa/drivers/dri/intel/intel_pixel.h
index 6fa6effe83..cb41fa182c 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.h
+++ b/src/mesa/drivers/dri/intel/intel_pixel.h
@@ -31,6 +31,15 @@
 #include "main/mtypes.h"
 
 void intelInitPixelFuncs(struct dd_function_table *functions);
+void intel_meta_set_passthrough_transform(struct intel_context *intel);
+void intel_meta_restore_transform(struct intel_context *intel);
+void intel_meta_set_passthrough_vertex_program(struct intel_context *intel);
+void intel_meta_restore_vertex_program(struct intel_context *intel);
+void intel_meta_set_fragment_program(struct intel_context *intel,
+				     struct gl_fragment_program **prog,
+				     const char *prog_string);
+void intel_meta_restore_fragment_program(struct intel_context *intel);
+void intel_free_pixel_state(struct intel_context *intel);
 
 GLboolean intel_check_blit_fragment_ops(GLcontext * ctx,
 					GLboolean src_alpha_is_one);
@@ -67,4 +76,6 @@ void intelBitmap(GLcontext * ctx,
 		 const struct gl_pixelstore_attrib *unpack,
 		 const GLubyte * pixels);
 
+void intel_clear_tris(GLcontext *ctx, GLbitfield mask);
+
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index 0565197ea0..1db7f5594e 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -32,6 +32,18 @@
 #include "main/mtypes.h"
 #include "main/macros.h"
 #include "main/bufferobj.h"
+#include "main/pixelstore.h"
+#include "main/state.h"
+#include "main/teximage.h"
+#include "main/texenv.h"
+#include "main/texobj.h"
+#include "main/texstate.h"
+#include "main/texparam.h"
+#include "main/varray.h"
+#include "main/attrib.h"
+#include "main/enable.h"
+#include "shader/arbprogram.h"
+#include "glapi/dispatch.h"
 #include "swrast/swrast.h"
 
 #include "intel_screen.h"
@@ -86,6 +98,11 @@ static GLboolean test_bit( const GLubyte *src,
    return (src[bit/8] & (1<<(bit % 8))) ? 1 : 0;
 }
 
+static GLboolean test_msb_bit(const GLubyte *src, GLuint bit)
+{
+   return (src[bit/8] & (1<<(7 - (bit % 8)))) ? 1 : 0;
+}
+
 static void set_bit( GLubyte *dest,
 			  GLuint bit )
 {
@@ -149,8 +166,18 @@ static GLuint get_bitmap_rect(GLsizei width, GLsizei height,
    return count;
 }
 
-
-
+/**
+ * Returns the low Y value of the vertical range given, flipped according to
+ * whether the framebuffer is or not.
+ */
+static inline int
+y_flip(struct gl_framebuffer *fb, int y, int height)
+{
+   if (fb->Name != 0)
+      return y;
+   else
+      return fb->Height - y - height;
+}
 
 /*
  * Render a bitmap.
@@ -164,9 +191,26 @@ do_blit_bitmap( GLcontext *ctx,
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_region *dst = intel_drawbuf_region(intel);
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
    GLfloat tmpColor[4];
    GLubyte ubcolor[4];
    GLuint color8888, color565;
+   unsigned int num_cliprects;
+   drm_clip_rect_t *cliprects;
+   int x_off, y_off;
+   GLsizei bitmap_width = width;
+   GLsizei bitmap_height = height;
+
+   /* Update draw buffer bounds */
+   _mesa_update_state(ctx);
+
+   if (ctx->Depth.Test) {
+      /* The blit path produces incorrect results when depth testing is on.
+       * It seems the blit Z coord is always 1.0 (the far plane) so fragments
+       * will likely be obscured by other, closer geometry.
+       */
+      return GL_FALSE;
+   }
 
    if (!dst)
        return GL_FALSE;
@@ -196,59 +240,38 @@ do_blit_bitmap( GLcontext *ctx,
 
    LOCK_HARDWARE(intel);
 
-   if (intel->driDrawable->numClipRects) {
-      __DRIdrawablePrivate *dPriv = intel->driDrawable;
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      drm_clip_rect_t dest_rect;
-      GLint nbox = dPriv->numClipRects;
-      GLint srcx = 0, srcy = 0;
-      GLint orig_screen_x1, orig_screen_y2;
+   intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off);
+   if (num_cliprects != 0) {
       GLuint i;
+      GLint orig_dstx = dstx;
+      GLint orig_dsty = dsty;
 
-
-      orig_screen_x1 = dPriv->x + dstx;
-      orig_screen_y2 = dPriv->y + (dPriv->h - dsty);
-
-      /* Do scissoring in GL coordinates:
-       */
-      if (ctx->Scissor.Enabled)
-      {
-	 GLint x = ctx->Scissor.X;
-	 GLint y = ctx->Scissor.Y;
-	 GLuint w = ctx->Scissor.Width;
-	 GLuint h = ctx->Scissor.Height;
-
-         if (!_mesa_clip_to_region(x, y, x+w-1, y+h-1, &dstx, &dsty, &width, &height))
+      /* Clip to buffer bounds and scissor. */
+      if (!_mesa_clip_to_region(fb->_Xmin, fb->_Ymin,
+				fb->_Xmax, fb->_Ymax,
+				&dstx, &dsty, &width, &height))
             goto out;
-      }
-
-      /* Convert from GL to hardware coordinates:
-       */
-      dsty = dPriv->y + (dPriv->h - dsty - height);  
-      dstx = dPriv->x + dstx;
 
-      dest_rect.x1 = dstx < 0 ? 0 : dstx;
-      dest_rect.y1 = dsty < 0 ? 0 : dsty;
-      dest_rect.x2 = dstx + width < 0 ? 0 : dstx + width;
-      dest_rect.y2 = dsty + height < 0 ? 0 : dsty + height;
+      dstx = x_off + dstx;
+      dsty = y_off + y_flip(fb, dsty, height);
 
-      for (i = 0; i < nbox; i++) {
-         drm_clip_rect_t rect;
-	 int box_w, box_h;
+      for (i = 0; i < num_cliprects; i++) {
+	 int box_x, box_y, box_w, box_h;
 	 GLint px, py;
 	 GLuint stipple[32];  
 
-         if (!intel_intersect_cliprects(&rect, &dest_rect, &box[i]))
-            continue;
-
-	 /* Now go back to GL coordinates to figure out what subset of
-	  * the bitmap we are uploading for this cliprect:
-	  */
-	 box_w = rect.x2 - rect.x1;
-	 box_h = rect.y2 - rect.y1;
-	 srcx = rect.x1 - orig_screen_x1;
-	 srcy = orig_screen_y2 - rect.y2;
+	 box_x = dstx;
+	 box_y = dsty;
+	 box_w = width;
+	 box_h = height;
 
+	 /* Clip to drawable cliprect */
+         if (!_mesa_clip_to_region(cliprects[i].x1,
+				   cliprects[i].y1,
+				   cliprects[i].x2,
+				   cliprects[i].y2,
+				   &box_x, &box_y, &box_w, &box_h))
+	    continue;
 
 #define DY 32
 #define DX 32
@@ -269,13 +292,19 @@ do_blit_bitmap( GLcontext *ctx,
 
 	       /* May need to adjust this when padding has been introduced in
 		* sz above:
+		*
+		* Have to translate destination coordinates back into source
+		* coordinates.
 		*/
-	       if (get_bitmap_rect(width, height, unpack, 
+	       if (get_bitmap_rect(bitmap_width, bitmap_height, unpack,
 				   bitmap,
-				   srcx + px, srcy + py, w, h,
+				   -orig_dstx + (box_x + px - x_off),
+				   -orig_dsty + y_flip(fb,
+						       box_y + py - y_off, h),
+				   w, h,
 				   (GLubyte *)stipple,
 				   8,
-				   GL_TRUE) == 0)
+				   fb->Name == 0 ? GL_TRUE : GL_FALSE) == 0)
 		  continue;
 
 	       /* 
@@ -289,8 +318,8 @@ do_blit_bitmap( GLcontext *ctx,
 						  dst->buffer,
 						  0,
 						  dst->tiling,
-						  rect.x1 + px,
-						  rect.y2 - (py + h),
+						  box_x + px,
+						  box_y + py,
 						  w, h,
 						  logic_op);
 	    } 
@@ -300,6 +329,8 @@ do_blit_bitmap( GLcontext *ctx,
 out:
    UNLOCK_HARDWARE(intel);
 
+   if (INTEL_DEBUG & DEBUG_SYNC)
+      intel_batchbuffer_flush(intel->batch);
 
    if (unpack->BufferObj->Name) {
       /* done with PBO so unmap it now */
@@ -310,9 +341,179 @@ out:
    return GL_TRUE;
 }
 
+static GLboolean
+intel_texture_bitmap(GLcontext * ctx,
+		     GLint dst_x, GLint dst_y,
+		     GLsizei width, GLsizei height,
+		     const struct gl_pixelstore_attrib *unpack,
+		     const GLubyte *bitmap)
+{
+   struct intel_context *intel = intel_context(ctx);
+   static const char *fp =
+      "!!ARBfp1.0\n"
+      "TEMP val;\n"
+      "PARAM color=program.local[0];\n"
+      "TEX val, fragment.texcoord[0], texture[0], 2D;\n"
+      "ADD val, val.wwww, {-.5, -.5, -.5, -.5};\n"
+      "KIL val;\n"
+      "MOV result.color, color;\n"
+      "END\n";
+   GLuint texname;
+   GLfloat vertices[4][4];
+   GLfloat texcoords[4][2];
+   GLint old_active_texture;
+   GLubyte *unpacked_bitmap;
+   GLubyte *a8_bitmap;
+   int x, y;
+   GLfloat dst_z;
+
+   /* We need a fragment program for the KIL effect */
+   if (!ctx->Extensions.ARB_fragment_program ||
+       !ctx->Extensions.ARB_vertex_program) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr,
+		 "glBitmap fallback: No fragment/vertex program support\n");
+      return GL_FALSE;
+   }
+
+   /* We're going to mess with texturing with no regard to existing texture
+    * state, so if there is some set up we have to bail.
+    */
+   if (ctx->Texture._EnabledUnits != 0) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glBitmap fallback: texturing enabled\n");
+      return GL_FALSE;
+   }
+
+   /* Can't do textured DrawPixels with a fragment program, unless we were
+    * to generate a new program that sampled our texture and put the results
+    * in the fragment color before the user's program started.
+    */
+   if (ctx->FragmentProgram.Enabled) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glBitmap fallback: fragment program enabled\n");
+      return GL_FALSE;
+   }
 
+   if (ctx->VertexProgram.Enabled) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glBitmap fallback: vertex program enabled\n");
+      return GL_FALSE;
+   }
 
+   /* Check that we can load in a texture this big. */
+   if (width > (1 << (ctx->Const.MaxTextureLevels - 1)) ||
+       height > (1 << (ctx->Const.MaxTextureLevels - 1))) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glBitmap fallback: bitmap too large (%dx%d)\n",
+		 width, height);
+      return GL_FALSE;
+   }
 
+   /* Convert the A1 bitmap to an A8 format suitable for glTexImage */
+   if (unpack->BufferObj->Name) {
+      bitmap = map_pbo(ctx, width, height, unpack, bitmap);
+      if (bitmap == NULL)
+	 return GL_TRUE;	/* even though this is an error, we're done */
+   }
+   unpacked_bitmap = _mesa_unpack_bitmap(width, height, bitmap,
+					 unpack);
+   a8_bitmap = _mesa_calloc(width * height);
+   for (y = 0; y < height; y++) {
+      for (x = 0; x < width; x++) {
+	 if (test_msb_bit(unpacked_bitmap, ALIGN(width, 8) * y + x))
+	    a8_bitmap[y * width + x] = 0xff;
+      }
+   }
+   _mesa_free(unpacked_bitmap);
+   if (unpack->BufferObj->Name) {
+      /* done with PBO so unmap it now */
+      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
+                              unpack->BufferObj);
+   }
+
+   /* Save GL state before we start setting up our drawing */
+   _mesa_PushAttrib(GL_ENABLE_BIT | GL_CURRENT_BIT |
+		    GL_VIEWPORT_BIT);
+   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT |
+			  GL_CLIENT_PIXEL_STORE_BIT);
+   old_active_texture = ctx->Texture.CurrentUnit;
+
+   _mesa_Disable(GL_POLYGON_STIPPLE);
+
+   /* Upload our bitmap data to an alpha texture */
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB);
+   _mesa_Enable(GL_TEXTURE_2D);
+   _mesa_GenTextures(1, &texname);
+   _mesa_BindTexture(GL_TEXTURE_2D, texname);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+
+   _mesa_PixelStorei(GL_UNPACK_SWAP_BYTES, GL_FALSE);
+   _mesa_PixelStorei(GL_UNPACK_LSB_FIRST, GL_FALSE);
+   _mesa_PixelStorei(GL_UNPACK_ROW_LENGTH, 0);
+   _mesa_PixelStorei(GL_UNPACK_SKIP_PIXELS, 0);
+   _mesa_PixelStorei(GL_UNPACK_SKIP_ROWS, 0);
+   _mesa_PixelStorei(GL_UNPACK_ALIGNMENT, 1);
+   _mesa_TexImage2D(GL_TEXTURE_2D, 0, GL_ALPHA, width, height, 0,
+		    GL_ALPHA, GL_UNSIGNED_BYTE, a8_bitmap);
+   _mesa_free(a8_bitmap);
+
+   intel_meta_set_fragment_program(intel, &intel->meta.bitmap_fp, fp);
+   _mesa_ProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0,
+				     ctx->Current.RasterColor);
+   intel_meta_set_passthrough_vertex_program(intel);
+   intel_meta_set_passthrough_transform(intel);
+
+   /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */
+   dst_z = -1.0 + 2.0 * ctx->Current.RasterPos[2];
+
+   vertices[0][0] = dst_x;
+   vertices[0][1] = dst_y;
+   vertices[0][2] = dst_z;
+   vertices[0][3] = 1.0;
+   vertices[1][0] = dst_x + width;
+   vertices[1][1] = dst_y;
+   vertices[1][2] = dst_z;
+   vertices[1][3] = 1.0;
+   vertices[2][0] = dst_x + width;
+   vertices[2][1] = dst_y + height;
+   vertices[2][2] = dst_z;
+   vertices[2][3] = 1.0;
+   vertices[3][0] = dst_x;
+   vertices[3][1] = dst_y + height;
+   vertices[3][2] = dst_z;
+   vertices[3][3] = 1.0;
+
+   texcoords[0][0] = 0.0;
+   texcoords[0][1] = 0.0;
+   texcoords[1][0] = 1.0;
+   texcoords[1][1] = 0.0;
+   texcoords[2][0] = 1.0;
+   texcoords[2][1] = 1.0;
+   texcoords[3][0] = 0.0;
+   texcoords[3][1] = 1.0;
+
+   _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
+   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
+   _mesa_Enable(GL_VERTEX_ARRAY);
+   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
+   CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
+
+   intel_meta_restore_transform(intel);
+   intel_meta_restore_fragment_program(intel);
+   intel_meta_restore_vertex_program(intel);
+
+   _mesa_PopClientAttrib();
+   _mesa_Disable(GL_TEXTURE_2D); /* asserted that it was disabled at entry */
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
+   _mesa_PopAttrib();
+
+   _mesa_DeleteTextures(1, &texname);
+
+   return GL_TRUE;
+}
 
 /* There are a large number of possible ways to implement bitmap on
  * this hardware, most of them have some sort of drawback.  Here are a
@@ -345,6 +546,10 @@ intelBitmap(GLcontext * ctx,
                           unpack, pixels))
       return;
 
+   if (intel_texture_bitmap(ctx, x, y, width, height,
+			    unpack, pixels))
+      return;
+
    if (INTEL_DEBUG & DEBUG_PIXEL)
       _mesa_printf("%s: fallback to swrast\n", __FUNCTION__);
 
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_copy.c b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
index 1b3cb5adcb..7c7aa6097c 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
@@ -119,6 +119,12 @@ do_texture_copypixels(GLcontext * ctx,
    if (!src || !dst || type != GL_COLOR)
       return GL_FALSE;
 
+   if (ctx->_ImageTransferState) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         fprintf(stderr, "%s: check_color failed\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
    /* Can't handle overlapping regions.  Don't have sufficient control
     * over rasterization to pull it off in-place.  Punt on these for
     * now.
@@ -260,6 +266,14 @@ do_blit_copypixels(GLcontext * ctx,
    struct intel_context *intel = intel_context(ctx);
    struct intel_region *dst = intel_drawbuf_region(intel);
    struct intel_region *src = copypix_src_region(intel, type);
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   struct gl_framebuffer *read_fb = ctx->ReadBuffer;
+   unsigned int num_cliprects;
+   drm_clip_rect_t *cliprects;
+   int x_off, y_off;
+
+   /* Update draw buffer bounds */
+   _mesa_update_state(ctx);
 
    /* Copypixels can be more than a straight copy.  Ensure all the
     * extra operations are disabled:
@@ -277,71 +291,74 @@ do_blit_copypixels(GLcontext * ctx,
 
    LOCK_HARDWARE(intel);
 
-   if (intel->driDrawable->numClipRects) {
-      __DRIdrawablePrivate *dPriv = intel->driDrawable;
-      __DRIdrawablePrivate *dReadPriv = intel->driReadDrawable;
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      GLint nbox = dPriv->numClipRects;
-      GLint delta_x = 0;
-      GLint delta_y = 0;
+   intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off);
+   if (num_cliprects != 0) {
+      GLint delta_x;
+      GLint delta_y;
+      GLint orig_dstx;
+      GLint orig_dsty;
+      GLint orig_srcx;
+      GLint orig_srcy;
       GLuint i;
 
-      /* Do scissoring in GL coordinates:
-       */
-      if (ctx->Scissor.Enabled)
-      {
-	 GLint x = ctx->Scissor.X;
-	 GLint y = ctx->Scissor.Y;
-	 GLuint w = ctx->Scissor.Width;
-	 GLuint h = ctx->Scissor.Height;
-	 GLint dx = dstx - srcx;
-         GLint dy = dsty - srcy;
-
-         if (!_mesa_clip_to_region(x, y, x+w-1, y+h-1, &dstx, &dsty, &width, &height))
-            goto out;
-	 
-         srcx = dstx - dx;
-         srcy = dsty - dy;
-      }
+      /* XXX: We fail to handle different inversion between read and draw framebuffer. */
+
+      /* Clip to destination buffer. */
+      orig_dstx = dstx;
+      orig_dsty = dsty;
+      if (!_mesa_clip_to_region(fb->_Xmin, fb->_Ymin,
+				fb->_Xmax, fb->_Ymax,
+				&dstx, &dsty, &width, &height))
+	 goto out;
+      /* Adjust src coords for our post-clipped destination origin */
+      srcx += dstx - orig_dstx;
+      srcy += dsty - orig_dsty;
+
+      /* Clip to source buffer. */
+      orig_srcx = srcx;
+      orig_srcy = srcy;
+      if (!_mesa_clip_to_region(0, 0,
+				read_fb->Width, read_fb->Height,
+				&srcx, &srcy, &width, &height))
+	 goto out;
+      /* Adjust dst coords for our post-clipped source origin */
+      dstx += srcx - orig_srcx;
+      dsty += srcy - orig_srcy;
 
       /* Convert from GL to hardware coordinates:
        */
-      dsty = dPriv->h - dsty - height;  
-      srcy = dPriv->h - srcy - height;  
-      dstx += dPriv->x;
-      dsty += dPriv->y;
-      srcx += dReadPriv->x;
-      srcy += dReadPriv->y;
-
-      /* Clip against the source region.  This is the only source
-       * clipping we do.  Dst is clipped with cliprects below.
-       */
-      {
-         delta_x = srcx - dstx;
-         delta_y = srcy - dsty;
-
-         if (!_mesa_clip_to_region(0, 0, src->pitch, src->height,
-                                   &srcx, &srcy, &width, &height))
-            goto out;
+      if (fb->Name == 0) {
+	 /* copypixels to a system framebuffer */
+	 dstx = x_off + dstx;
+	 dsty = y_off + (fb->Height - dsty - height);
+      } else {
+	 /* copypixels to a user framebuffer object */
+	 dstx = x_off + dstx;
+	 dsty = y_off + dsty;
+      }
 
-         dstx = srcx - delta_x;
-         dsty = srcy - delta_y;
+      /* Flip source Y if it's a system framebuffer. */
+      if (read_fb->Name == 0) {
+	 srcx = intel->driReadDrawable->x + srcx;
+	 srcy = intel->driReadDrawable->y + (fb->Height - srcy - height);
       }
 
+      delta_x = srcx - dstx;
+      delta_y = srcy - dsty;
       /* Could do slightly more clipping: Eg, take the intersection of
-       * the existing set of cliprects and those cliprects translated
-       * by delta_x, delta_y:
-       * 
+       * the destination cliprects and the read drawable cliprects
+       *
        * This code will not overwrite other windows, but will
        * introduce garbage when copying from obscured window regions.
        */
-      for (i = 0; i < nbox; i++) {
+      for (i = 0; i < num_cliprects; i++) {
 	 GLint clip_x = dstx;
 	 GLint clip_y = dsty;
 	 GLint clip_w = width;
 	 GLint clip_h = height;
 
-         if (!_mesa_clip_to_region(box[i].x1, box[i].y1, box[i].x2, box[i].y2,
+         if (!_mesa_clip_to_region(cliprects[i].x1, cliprects[i].y1,
+				   cliprects[i].x2, cliprects[i].y2,
 				   &clip_x, &clip_y, &clip_w, &clip_h))
             continue;
 
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
index 8ebbc95a1d..7be7ea82b3 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
@@ -36,7 +36,6 @@
 #include "main/texobj.h"
 #include "main/texstate.h"
 #include "main/texparam.h"
-#include "main/matrix.h"
 #include "main/varray.h"
 #include "main/attrib.h"
 #include "main/enable.h"
@@ -68,9 +67,13 @@ intel_texture_drawpixels(GLcontext * ctx,
 			 const struct gl_pixelstore_attrib *unpack,
 			 const GLvoid *pixels)
 {
+   struct intel_context *intel = intel_context(ctx);
    GLuint texname;
    GLfloat vertices[4][4];
    GLfloat texcoords[4][2];
+   GLfloat z;
+   GLint old_active_texture;
+   GLenum internalFormat;
 
    /* We're going to mess with texturing with no regard to existing texture
     * state, so if there is some set up we have to bail.
@@ -91,7 +94,7 @@ intel_texture_drawpixels(GLcontext * ctx,
       return GL_FALSE;
    }
 
-   /* We don't have a way to generate fragments with stencil values which *
+   /* We don't have a way to generate fragments with stencil values which
     * will set the resulting stencil value.
     */
    if (format == GL_STENCIL_INDEX)
@@ -117,13 +120,14 @@ intel_texture_drawpixels(GLcontext * ctx,
       return GL_FALSE;
    }
 
-   _mesa_PushAttrib(GL_ENABLE_BIT | GL_TRANSFORM_BIT | GL_TEXTURE_BIT |
+   _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT |
 		    GL_CURRENT_BIT);
    _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
 
    /* XXX: pixel store stuff */
    _mesa_Disable(GL_POLYGON_STIPPLE);
 
+   old_active_texture = ctx->Texture.CurrentUnit;
    _mesa_ActiveTextureARB(GL_TEXTURE0_ARB);
    _mesa_Enable(GL_TEXTURE_2D);
    _mesa_GenTextures(1, &texname);
@@ -131,21 +135,17 @@ intel_texture_drawpixels(GLcontext * ctx,
    _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
    _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
    _mesa_TexEnvf(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE);
-   /*
-   _mesa_TexEnvf(GL_TEXTURE_ENV, GL_COMBINE_RGB, GL_REPLACE);
-   _mesa_TexEnvf(GL_TEXTURE_ENV, GL_COMBINE_ALPHA, GL_REPLACE);
-   */
-   _mesa_TexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, width, height, 0, format,
+   if (type == GL_ALPHA)
+      internalFormat = GL_ALPHA;
+   else
+      internalFormat = GL_RGBA;
+   _mesa_TexImage2D(GL_TEXTURE_2D, 0, internalFormat, width, height, 0, format,
 		    type, pixels);
 
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
+   intel_meta_set_passthrough_transform(intel);
 
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
+   /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */
+   z = -1.0 + 2.0 * ctx->Current.RasterPos[2];
 
    /* Create the vertex buffer based on the current raster pos.  The x and y
     * we're handed are ctx->Current.RasterPos[0,1] rounded to integers.
@@ -154,19 +154,19 @@ intel_texture_drawpixels(GLcontext * ctx,
     */
    vertices[0][0] = x;
    vertices[0][1] = y;
-   vertices[0][2] = ctx->Current.RasterPos[2];
+   vertices[0][2] = z;
    vertices[0][3] = 1.0;
    vertices[1][0] = x + width * ctx->Pixel.ZoomX;
    vertices[1][1] = y;
-   vertices[1][2] = ctx->Current.RasterPos[2];
+   vertices[1][2] = z;
    vertices[1][3] = 1.0;
    vertices[2][0] = x + width * ctx->Pixel.ZoomX;
    vertices[2][1] = y + height * ctx->Pixel.ZoomY;
-   vertices[2][2] = ctx->Current.RasterPos[2];
+   vertices[2][2] = z;
    vertices[2][3] = 1.0;
    vertices[3][0] = x;
    vertices[3][1] = y + height * ctx->Pixel.ZoomY;
-   vertices[3][2] = ctx->Current.RasterPos[2];
+   vertices[3][2] = z;
    vertices[3][3] = 1.0;
 
    texcoords[0][0] = 0.0;
@@ -179,15 +179,15 @@ intel_texture_drawpixels(GLcontext * ctx,
    texcoords[3][1] = 1.0;
 
    _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
    _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
    _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
    CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
 
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PopMatrix();
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PopMatrix();
+   intel_meta_restore_transform(intel);
+
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
    _mesa_PopClientAttrib();
    _mesa_PopAttrib();
 
@@ -205,6 +205,7 @@ intel_stencil_drawpixels(GLcontext * ctx,
 			 const struct gl_pixelstore_attrib *unpack,
 			 const GLvoid *pixels)
 {
+   struct intel_context *intel = intel_context(ctx);
    GLuint texname, rb_name, fb_name, old_fb_name;
    GLfloat vertices[4][2];
    GLfloat texcoords[4][2];
@@ -214,6 +215,7 @@ intel_stencil_drawpixels(GLcontext * ctx,
    struct gl_pixelstore_attrib old_unpack;
    GLstencil *stencil_pixels;
    int row;
+   GLint old_active_texture;
 
    if (format != GL_STENCIL_INDEX)
       return GL_FALSE;
@@ -230,6 +232,10 @@ intel_stencil_drawpixels(GLcontext * ctx,
       return GL_FALSE;
    }
 
+   /* We don't support stencil testing/ops here */
+   if (ctx->Stencil.Enabled)
+      return GL_FALSE;
+
    /* We use FBOs for our wrapping of the depthbuffer into a color
     * destination.
     */
@@ -267,10 +273,11 @@ intel_stencil_drawpixels(GLcontext * ctx,
       return GL_FALSE;
    }
 
-   _mesa_PushAttrib(GL_ENABLE_BIT | GL_TRANSFORM_BIT | GL_TEXTURE_BIT |
+   _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT |
 		    GL_CURRENT_BIT | GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
    _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
    old_fb_name = ctx->DrawBuffer->Name;
+   old_active_texture = ctx->Texture.CurrentUnit;
 
    _mesa_Disable(GL_POLYGON_STIPPLE);
    _mesa_Disable(GL_DEPTH_TEST);
@@ -335,14 +342,7 @@ intel_stencil_drawpixels(GLcontext * ctx,
    ctx->Unpack = old_unpack;
    _mesa_free(stencil_pixels);
 
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
-
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
+   intel_meta_set_passthrough_transform(intel);
 
    vertices[0][0] = x;
    vertices[0][1] = y;
@@ -363,17 +363,17 @@ intel_stencil_drawpixels(GLcontext * ctx,
    texcoords[3][1] = 1.0;
 
    _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
+   _mesa_ClientActiveTextureARB(GL_TEXTURE0);
    _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
    _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
    CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
 
+   intel_meta_restore_transform(intel);
+
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
    _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, old_fb_name);
 
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PopMatrix();
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PopMatrix();
    _mesa_PopClientAttrib();
    _mesa_PopAttrib();
 
@@ -404,27 +404,6 @@ intelDrawPixels(GLcontext * ctx,
    if (INTEL_DEBUG & DEBUG_PIXEL)
       _mesa_printf("%s: fallback to swrast\n", __FUNCTION__);
 
-   if (ctx->FragmentProgram._Current == ctx->FragmentProgram._TexEnvProgram) {
-      /*
-       * We don't want the i915 texenv program to be applied to DrawPixels.
-       * This is really just a performance optimization (mesa will other-
-       * wise happily run the fragment program on each pixel in the image).
-       */
-      struct gl_fragment_program *fpSave = ctx->FragmentProgram._Current;
-   /* can't just set current frag prog to 0 here as on buffer resize
-      we'll get new state checks which will segfault. Remains a hack. */
-      ctx->FragmentProgram._Current = NULL;
-      ctx->FragmentProgram._UseTexEnvProgram = GL_FALSE;
-      ctx->FragmentProgram._Active = GL_FALSE;
-      _swrast_DrawPixels( ctx, x, y, width, height, format, type,
-                          unpack, pixels );
-      ctx->FragmentProgram._Current = fpSave;
-      ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE;
-      ctx->FragmentProgram._Active = GL_TRUE;
-      _swrast_InvalidateState(ctx, _NEW_PROGRAM);
-   }
-   else {
-      _swrast_DrawPixels( ctx, x, y, width, height, format, type,
-                          unpack, pixels );
-   }
+   _swrast_DrawPixels(ctx, x, y, width, height, format, type,
+		      unpack, pixels);
 }
diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h
index 81a7386e42..57ac8f0cc1 100644
--- a/src/mesa/drivers/dri/intel/intel_reg.h
+++ b/src/mesa/drivers/dri/intel/intel_reg.h
@@ -71,16 +71,16 @@
 /** @{
  * 915 definitions
  */
-#define S0_VB_OFFSET_MASK		0xffffffc
+#define S0_VB_OFFSET_MASK		0xffffffc0
 #define S0_AUTO_CACHE_INV_DISABLE	(1<<0)
 /** @} */
 
 /** @{
  * 830 definitions
  */
-#define S0_VB_OFFSET_MASK_830		0xffffff8
+#define S0_VB_OFFSET_MASK_830		0xffffff80
 #define S0_VB_PITCH_SHIFT_830		1
-#define S0_VB_ENABLE_830		0
+#define S0_VB_ENABLE_830		(1<<0)
 /** @} */
 
 #define S1_VERTEX_WIDTH_SHIFT          24
@@ -100,8 +100,8 @@
 #define S2_TEXCOORD_FMT(unit, type)    ((type)<<(unit*4))
 #define S2_TEXCOORD_NONE               (~0)
 #define S2_TEX_COUNT_SHIFT_830		12
-#define S2_VERTEX_0_WIDTH_SHIFT_830	0
-#define S2_VERTEX_1_WIDTH_SHIFT_830	6
+#define S2_VERTEX_1_WIDTH_SHIFT_830	0
+#define S2_VERTEX_0_WIDTH_SHIFT_830	6
 /* S3 not interesting */
 
 #define S4_POINT_WIDTH_SHIFT           23
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index 8dbcc3050e..ec85c4131a 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -152,7 +152,7 @@ void
 intel_region_reference(struct intel_region **dst, struct intel_region *src)
 {
    if (src)
-      DBG("%s %d\n", __FUNCTION__, src->refcount);
+      DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);
 
    assert(*dst == NULL);
    if (src) {
@@ -169,7 +169,7 @@ intel_region_release(struct intel_region **region_handle)
    if (region == NULL)
       return;
 
-   DBG("%s %d\n", __FUNCTION__, region->refcount - 1);
+   DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);
 
    ASSERT(region->refcount > 0);
    region->refcount--;
@@ -550,15 +550,6 @@ intel_recreate_static_regions(struct intel_context *intel)
 			    intel->back_region,
 			    &intelScreen->back);
 
-#ifdef I915
-   if (intelScreen->third.handle) {
-      intel->third_region =
-	 intel_recreate_static(intel, "third",
-			       intel->third_region,
-			       &intelScreen->third);
-   }
-#endif /* I915 */
-
    /* Still assumes front.cpp == depth.cpp.  We can kill this when we move to
     * private buffers.
     */
diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c
index cf09fad745..a52271158c 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.c
+++ b/src/mesa/drivers/dri/intel/intel_screen.c
@@ -28,27 +28,27 @@
 #include "main/glheader.h"
 #include "main/context.h"
 #include "main/framebuffer.h"
-#include "main/matrix.h"
 #include "main/renderbuffer.h"
-#include "main/simple_list.h"
+
 #include "utils.h"
 #include "vblank.h"
 #include "xmlpool.h"
 
-
-#include "intel_screen.h"
-
+#include "intel_batchbuffer.h"
 #include "intel_buffers.h"
-#include "intel_tex.h"
-#include "intel_span.h"
-#include "intel_fbo.h"
+#include "intel_bufmgr.h"
 #include "intel_chipset.h"
+#include "intel_extensions.h"
+#include "intel_fbo.h"
+#include "intel_regions.h"
+#include "intel_swapbuffers.h"
+#include "intel_screen.h"
+#include "intel_span.h"
+#include "intel_tex.h"
 
 #include "i915_drm.h"
 #include "i830_dri.h"
-#include "intel_regions.h"
-#include "intel_batchbuffer.h"
-#include "intel_bufmgr.h"
+
 
 PUBLIC const char __driConfigOptions[] =
    DRI_CONF_BEGIN
@@ -134,7 +134,7 @@ intelPrintDRIInfo(intelScreenPrivate * intelScreen,
 
 
 static void
-intelPrintSAREA(const struct drm_i915_sarea * sarea)
+intelPrintSAREA(const drm_i915_sarea_t * sarea)
 {
    fprintf(stderr, "SAREA: sarea width %d  height %d\n", sarea->width,
            sarea->height);
@@ -159,9 +159,9 @@ intelPrintSAREA(const struct drm_i915_sarea * sarea)
  * A number of the screen parameters are obtained/computed from
  * information in the SAREA.  This function updates those parameters.
  */
-void
+static void
 intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen,
-                           struct drm_i915_sarea * sarea)
+                           drm_i915_sarea_t * sarea)
 {
    intelScreen->width = sarea->width;
    intelScreen->height = sarea->height;
@@ -177,13 +177,6 @@ intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen,
    intelScreen->back.size = sarea->back_size;
    intelScreen->back.tiled = sarea->back_tiled;
 
-   if (intelScreen->driScrnPriv->ddx_version.minor >= 8) {
-      intelScreen->third.offset = sarea->third_offset;
-      intelScreen->third.handle = sarea->third_handle;
-      intelScreen->third.size = sarea->third_size;
-      intelScreen->third.tiled = sarea->third_tiled;
-   }
-
    intelScreen->depth.offset = sarea->depth_offset;
    intelScreen->depth.handle = sarea->depth_handle;
    intelScreen->depth.size = sarea->depth_size;
@@ -192,12 +185,10 @@ intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen,
    if (intelScreen->driScrnPriv->ddx_version.minor >= 9) {
       intelScreen->front.bo_handle = sarea->front_bo_handle;
       intelScreen->back.bo_handle = sarea->back_bo_handle;
-      intelScreen->third.bo_handle = sarea->third_bo_handle;
       intelScreen->depth.bo_handle = sarea->depth_bo_handle;
    } else {
       intelScreen->front.bo_handle = -1;
       intelScreen->back.bo_handle = -1;
-      intelScreen->third.bo_handle = -1;
       intelScreen->depth.bo_handle = -1;
    }
 
@@ -253,7 +244,7 @@ static GLboolean intelInitDriver(__DRIscreenPrivate *sPriv)
 {
    intelScreenPrivate *intelScreen;
    I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv;
-   struct drm_i915_sarea *sarea;
+   drm_i915_sarea_t *sarea;
 
    if (sPriv->devPrivSize != sizeof(I830DRIRec)) {
       fprintf(stderr,
@@ -273,7 +264,7 @@ static GLboolean intelInitDriver(__DRIscreenPrivate *sPriv)
 
    intelScreen->driScrnPriv = sPriv;
    sPriv->private = (void *) intelScreen;
-   sarea = (struct drm_i915_sarea *)
+   sarea = (drm_i915_sarea_t *)
       (((GLubyte *) sPriv->pSAREA) + gDRIPriv->sarea_priv_offset);
    intelScreen->sarea = sarea;
 
@@ -325,8 +316,6 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv,
                   __DRIdrawablePrivate * driDrawPriv,
                   const __GLcontextModes * mesaVis, GLboolean isPixmap)
 {
-   intelScreenPrivate *screen = (intelScreenPrivate *) driScrnPriv->private;
-
    if (isPixmap) {
       return GL_FALSE;          /* not implemented */
    }
@@ -353,12 +342,6 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv,
          _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT,
 				&intel_fb->color_rb[1]->Base);
 
-	 if (screen->third.handle) {
-	    struct gl_renderbuffer *tmp_rb = NULL;
-
-	    intel_fb->color_rb[2] = intel_create_renderbuffer(rgbFormat);
-	    _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base);
-	 }
       }
 
       if (mesaVis->depthBits == 24) {
@@ -461,6 +444,7 @@ intelCreateContext(const __GLcontextModes * mesaVis,
 				  sharedContextPrivate);
       }
    } else {
+      intelScreen->no_vbo = GL_TRUE;
       return i830CreateContext(mesaVis, driContextPriv, sharedContextPrivate);
    }
 #else
@@ -481,8 +465,6 @@ intelFillInModes(__DRIscreenPrivate *psp,
    __GLcontextModes *m;
    unsigned depth_buffer_factor;
    unsigned back_buffer_factor;
-   GLenum fb_format;
-   GLenum fb_type;
    int i;
 
    /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't
@@ -494,6 +476,7 @@ intelFillInModes(__DRIscreenPrivate *psp,
 
    uint8_t depth_bits_array[3];
    uint8_t stencil_bits_array[3];
+   uint8_t msaa_samples_array[1];
 
    depth_bits_array[0] = 0;
    depth_bits_array[1] = depth_bits;
@@ -510,22 +493,39 @@ intelFillInModes(__DRIscreenPrivate *psp,
 
    stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+   msaa_samples_array[0] = 0;
+
    depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
    back_buffer_factor = (have_back_buffer) ? 3 : 1;
 
    if (pixel_bits == 16) {
-      fb_format = GL_RGB;
-      fb_type = GL_UNSIGNED_SHORT_5_6_5;
+      configs = driCreateConfigs(GL_RGB, GL_UNSIGNED_SHORT_5_6_5,
+				 depth_bits_array, stencil_bits_array,
+				 depth_buffer_factor, back_buffer_modes,
+				 back_buffer_factor,
+				 msaa_samples_array, 1);
    }
    else {
-      fb_format = GL_BGRA;
-      fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+      __DRIconfig **configs_a8r8g8b8;
+      __DRIconfig **configs_x8r8g8b8;
+
+      configs_a8r8g8b8 = driCreateConfigs(GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV,
+					  depth_bits_array,
+					  stencil_bits_array,
+					  depth_buffer_factor,
+					  back_buffer_modes,
+					  back_buffer_factor,
+					  msaa_samples_array, 1);
+      configs_x8r8g8b8 = driCreateConfigs(GL_BGR, GL_UNSIGNED_INT_8_8_8_8_REV,
+					  depth_bits_array,
+					  stencil_bits_array,
+					  depth_buffer_factor,
+					  back_buffer_modes,
+					  back_buffer_factor,
+					  msaa_samples_array, 1);
+      configs = driConcatConfigs(configs_a8r8g8b8, configs_x8r8g8b8);
    }
 
-   configs = driCreateConfigs(fb_format, fb_type,
-			      depth_bits_array, stencil_bits_array,
-			      depth_buffer_factor, back_buffer_modes,
-			      back_buffer_factor);
    if (configs == NULL) {
     fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
               __LINE__);
@@ -687,6 +687,17 @@ static const
 __DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp)
 {
    intelScreenPrivate *intelScreen;
+   GLenum fb_format[3];
+   GLenum fb_type[3];
+   /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't
+    * support pageflipping at all.
+    */
+   static const GLenum back_buffer_modes[] = {
+      GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML
+   };
+   uint8_t depth_bits[4], stencil_bits[4], msaa_samples_array[1];
+   int color;
+   __DRIconfig **configs = NULL;
 
    /* Calling driInitExtensions here, with a NULL context pointer,
     * does not actually enable the extensions.  It just makes sure
@@ -726,8 +737,50 @@ __DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp)
    intelScreen->irq_active = 1;
    psp->extensions = intelScreenExtensions;
 
-   return driConcatConfigs(intelFillInModes(psp, 16, 16, 0, 1),
-			   intelFillInModes(psp, 32, 24, 8, 1));
+   depth_bits[0] = 0;
+   stencil_bits[0] = 0;
+   depth_bits[1] = 16;
+   stencil_bits[1] = 0;
+   depth_bits[2] = 24;
+   stencil_bits[2] = 0;
+   depth_bits[3] = 24;
+   stencil_bits[3] = 8;
+
+   msaa_samples_array[0] = 0;
+
+   fb_format[0] = GL_RGB;
+   fb_type[0] = GL_UNSIGNED_SHORT_5_6_5;
+
+   fb_format[1] = GL_BGR;
+   fb_type[1] = GL_UNSIGNED_INT_8_8_8_8_REV;
+
+   fb_format[2] = GL_BGRA;
+   fb_type[2] = GL_UNSIGNED_INT_8_8_8_8_REV;
+
+   for (color = 0; color < ARRAY_SIZE(fb_format); color++) {
+      __DRIconfig **new_configs;
+
+      new_configs = driCreateConfigs(fb_format[color], fb_type[color],
+				     depth_bits,
+				     stencil_bits,
+				     ARRAY_SIZE(depth_bits),
+				     back_buffer_modes,
+				     ARRAY_SIZE(back_buffer_modes),
+				     msaa_samples_array,
+				     ARRAY_SIZE(msaa_samples_array));
+      if (configs == NULL)
+	 configs = new_configs;
+      else
+	 configs = driConcatConfigs(configs, new_configs);
+   }
+
+   if (configs == NULL) {
+      fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
+              __LINE__);
+      return NULL;
+   }
+
+   return (const __DRIconfig **)configs;
 }
 
 const struct __DriverAPIRec driDriverAPI = {
diff --git a/src/mesa/drivers/dri/intel/intel_screen.h b/src/mesa/drivers/dri/intel/intel_screen.h
index fc913da5bf..e1036de4db 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.h
+++ b/src/mesa/drivers/dri/intel/intel_screen.h
@@ -56,7 +56,6 @@ typedef struct
 {
    intelRegion front;
    intelRegion back;
-   intelRegion third;
    intelRegion depth;
    intelRegion tex;
 
@@ -69,7 +68,7 @@ typedef struct
 
    __DRIscreenPrivate *driScrnPriv;
 
-   volatile struct drm_i915_sarea *sarea;
+   volatile drm_i915_sarea_t *sarea;
 
    int drmMinor;
 
@@ -77,6 +76,7 @@ typedef struct
 
    GLboolean no_hw;
 
+   GLboolean no_vbo;
    int ttm;
    dri_bufmgr *bufmgr;
 
@@ -92,10 +92,6 @@ extern GLboolean intelMapScreenRegions(__DRIscreenPrivate * sPriv);
 
 extern void intelUnmapScreenRegions(intelScreenPrivate * intelScreen);
 
-extern void
-intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen,
-                           struct drm_i915_sarea * sarea);
-
 extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv);
 
 extern GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv);
@@ -105,11 +101,6 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
                  __DRIdrawablePrivate * driDrawPriv,
                  __DRIdrawablePrivate * driReadPriv);
 
-extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv);
-
-extern void
-intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h);
-
 extern struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen);
 
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c
index 8f4e681ffe..bdd2fd9e85 100644
--- a/src/mesa/drivers/dri/intel/intel_span.c
+++ b/src/mesa/drivers/dri/intel/intel_span.c
@@ -150,7 +150,7 @@ static uint32_t x_tile_swizzle(struct intel_renderbuffer *irb,
 	int	x_tile_number, y_tile_number;
 	int	tile_off, tile_base;
 	
-	tile_stride = (irb->pfPitch * irb->region->cpp) << 3;
+	tile_stride = (irb->region->pitch * irb->region->cpp) << 3;
 
 	xbyte = x * irb->region->cpp;
 
@@ -190,7 +190,7 @@ static uint32_t x_tile_swizzle(struct intel_renderbuffer *irb,
 	printf("(%d,%d) -> %d + %d = %d (pitch = %d, tstride = %d)\n",
 	       x, y, tile_off, tile_base,
 	       tile_off + tile_base,
-	       irb->pfPitch, tile_stride);
+	       irb->region->pitch, tile_stride);
 #endif
 
 	return tile_base + tile_off;
@@ -205,7 +205,7 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 	int	x_tile_number, y_tile_number;
 	int	tile_off, tile_base;
 	
-	tile_stride = (irb->pfPitch * irb->region->cpp) << 5;
+	tile_stride = (irb->region->pitch * irb->region->cpp) << 5;
 
 	xbyte = x * irb->region->cpp;
 
@@ -255,8 +255,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 #define LOCAL_VARS							\
    struct intel_context *intel = intel_context(ctx);			\
    struct intel_renderbuffer *irb = intel_renderbuffer(rb);		\
-   const GLint yScale = irb->RenderToTexture ? 1 : -1;			\
-   const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1;	\
+   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
+   const GLint yBias = ctx->DrawBuffer->Name ? 0 : irb->Base.Height - 1;\
    unsigned int num_cliprects;						\
    struct drm_clip_rect *cliprects;					\
    int x_off, y_off;							\
@@ -392,8 +392,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 #define LOCAL_DEPTH_VARS						\
    struct intel_context *intel = intel_context(ctx);			\
    struct intel_renderbuffer *irb = intel_renderbuffer(rb);		\
-   const GLint yScale = irb->RenderToTexture ? 1 : -1;			\
-   const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \
+   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
+   const GLint yBias = ctx->DrawBuffer->Name ? 0 : irb->Base.Height - 1;\
    unsigned int num_cliprects;						\
    struct drm_clip_rect *cliprects;					\
    int x_off, y_off;							\
@@ -528,8 +528,6 @@ intel_renderbuffer_map(struct intel_context *intel, struct gl_renderbuffer *rb)
    if (irb == NULL || irb->region == NULL)
       return;
 
-   irb->pfPitch = irb->region->pitch;
-
    intel_set_span_functions(intel, rb);
 }
 
@@ -543,7 +541,6 @@ intel_renderbuffer_unmap(struct intel_context *intel,
       return;
 
    clear_span_cache(irb);
-   irb->pfPitch = 0;
 
    rb->GetRow = NULL;
    rb->PutRow = NULL;
@@ -633,7 +630,7 @@ intelSpanRenderStart(GLcontext * ctx)
    intelFlush(&intel->ctx);
    LOCK_HARDWARE(intel);
 
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+   for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
       if (ctx->Texture.Unit[i]._ReallyEnabled) {
          struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current;
          intel_tex_map_images(intel, intel_texture_object(texObj));
@@ -655,7 +652,7 @@ intelSpanRenderFinish(GLcontext * ctx)
 
    _swrast_flush(ctx);
 
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+   for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
       if (ctx->Texture.Unit[i]._ReallyEnabled) {
          struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current;
          intel_tex_unmap_images(intel, intel_texture_object(texObj));
diff --git a/src/mesa/drivers/dri/intel/intel_state.c b/src/mesa/drivers/dri/intel/intel_state.c
new file mode 100644
index 0000000000..4ee742377d
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_state.c
@@ -0,0 +1,233 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/enums.h"
+#include "main/colormac.h"
+#include "main/dd.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_regions.h"
+#include "swrast/swrast.h"
+
+int
+intel_translate_shadow_compare_func(GLenum func)
+{
+   switch (func) {
+   case GL_NEVER: 
+       return COMPAREFUNC_ALWAYS;
+   case GL_LESS: 
+       return COMPAREFUNC_LEQUAL;
+   case GL_LEQUAL: 
+       return COMPAREFUNC_LESS;
+   case GL_GREATER: 
+       return COMPAREFUNC_GEQUAL;
+   case GL_GEQUAL: 
+      return COMPAREFUNC_GREATER;
+   case GL_NOTEQUAL: 
+      return COMPAREFUNC_EQUAL;
+   case GL_EQUAL: 
+      return COMPAREFUNC_NOTEQUAL;
+   case GL_ALWAYS: 
+       return COMPAREFUNC_NEVER;
+   }
+
+   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func);
+   return COMPAREFUNC_NEVER;
+}
+
+int
+intel_translate_compare_func(GLenum func)
+{
+   switch (func) {
+   case GL_NEVER:
+      return COMPAREFUNC_NEVER;
+   case GL_LESS:
+      return COMPAREFUNC_LESS;
+   case GL_LEQUAL:
+      return COMPAREFUNC_LEQUAL;
+   case GL_GREATER:
+      return COMPAREFUNC_GREATER;
+   case GL_GEQUAL:
+      return COMPAREFUNC_GEQUAL;
+   case GL_NOTEQUAL:
+      return COMPAREFUNC_NOTEQUAL;
+   case GL_EQUAL:
+      return COMPAREFUNC_EQUAL;
+   case GL_ALWAYS:
+      return COMPAREFUNC_ALWAYS;
+   }
+
+   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func);
+   return COMPAREFUNC_ALWAYS;
+}
+
+int
+intel_translate_stencil_op(GLenum op)
+{
+   switch (op) {
+   case GL_KEEP:
+      return STENCILOP_KEEP;
+   case GL_ZERO:
+      return STENCILOP_ZERO;
+   case GL_REPLACE:
+      return STENCILOP_REPLACE;
+   case GL_INCR:
+      return STENCILOP_INCRSAT;
+   case GL_DECR:
+      return STENCILOP_DECRSAT;
+   case GL_INCR_WRAP:
+      return STENCILOP_INCR;
+   case GL_DECR_WRAP:
+      return STENCILOP_DECR;
+   case GL_INVERT:
+      return STENCILOP_INVERT;
+   default:
+      return STENCILOP_ZERO;
+   }
+}
+
+int
+intel_translate_blend_factor(GLenum factor)
+{
+   switch (factor) {
+   case GL_ZERO:
+      return BLENDFACT_ZERO;
+   case GL_SRC_ALPHA:
+      return BLENDFACT_SRC_ALPHA;
+   case GL_ONE:
+      return BLENDFACT_ONE;
+   case GL_SRC_COLOR:
+      return BLENDFACT_SRC_COLR;
+   case GL_ONE_MINUS_SRC_COLOR:
+      return BLENDFACT_INV_SRC_COLR;
+   case GL_DST_COLOR:
+      return BLENDFACT_DST_COLR;
+   case GL_ONE_MINUS_DST_COLOR:
+      return BLENDFACT_INV_DST_COLR;
+   case GL_ONE_MINUS_SRC_ALPHA:
+      return BLENDFACT_INV_SRC_ALPHA;
+   case GL_DST_ALPHA:
+      return BLENDFACT_DST_ALPHA;
+   case GL_ONE_MINUS_DST_ALPHA:
+      return BLENDFACT_INV_DST_ALPHA;
+   case GL_SRC_ALPHA_SATURATE:
+      return BLENDFACT_SRC_ALPHA_SATURATE;
+   case GL_CONSTANT_COLOR:
+      return BLENDFACT_CONST_COLOR;
+   case GL_ONE_MINUS_CONSTANT_COLOR:
+      return BLENDFACT_INV_CONST_COLOR;
+   case GL_CONSTANT_ALPHA:
+      return BLENDFACT_CONST_ALPHA;
+   case GL_ONE_MINUS_CONSTANT_ALPHA:
+      return BLENDFACT_INV_CONST_ALPHA;
+   }
+
+   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, factor);
+   return BLENDFACT_ZERO;
+}
+
+int
+intel_translate_logic_op(GLenum opcode)
+{
+   switch (opcode) {
+   case GL_CLEAR:
+      return LOGICOP_CLEAR;
+   case GL_AND:
+      return LOGICOP_AND;
+   case GL_AND_REVERSE:
+      return LOGICOP_AND_RVRSE;
+   case GL_COPY:
+      return LOGICOP_COPY;
+   case GL_COPY_INVERTED:
+      return LOGICOP_COPY_INV;
+   case GL_AND_INVERTED:
+      return LOGICOP_AND_INV;
+   case GL_NOOP:
+      return LOGICOP_NOOP;
+   case GL_XOR:
+      return LOGICOP_XOR;
+   case GL_OR:
+      return LOGICOP_OR;
+   case GL_OR_INVERTED:
+      return LOGICOP_OR_INV;
+   case GL_NOR:
+      return LOGICOP_NOR;
+   case GL_EQUIV:
+      return LOGICOP_EQUIV;
+   case GL_INVERT:
+      return LOGICOP_INV;
+   case GL_OR_REVERSE:
+      return LOGICOP_OR_RVRSE;
+   case GL_NAND:
+      return LOGICOP_NAND;
+   case GL_SET:
+      return LOGICOP_SET;
+   default:
+      return LOGICOP_SET;
+   }
+}
+
+
+static void
+intelClearColor(GLcontext *ctx, const GLfloat color[4])
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLubyte clear[4];
+
+   CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]);
+   CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]);
+   CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]);
+   CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]);
+
+   /* compute both 32 and 16-bit clear values */
+   intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1],
+                                               clear[2], clear[3]);
+   intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]);
+}
+
+
+/* Fallback to swrast for select and feedback.
+ */
+static void
+intelRenderMode(GLcontext *ctx, GLenum mode)
+{
+   struct intel_context *intel = intel_context(ctx);
+   FALLBACK(intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER));
+}
+
+
+void
+intelInitStateFuncs(struct dd_function_table *functions)
+{
+   functions->RenderMode = intelRenderMode;
+   functions->ClearColor = intelClearColor;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_swapbuffers.c b/src/mesa/drivers/dri/intel/intel_swapbuffers.c
new file mode 100644
index 0000000000..7d035b9f6e
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_swapbuffers.c
@@ -0,0 +1,248 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "intel_blit.h"
+#include "intel_buffers.h"
+#include "intel_swapbuffers.h"
+#include "intel_fbo.h"
+#include "intel_batchbuffer.h"
+#include "drirenderbuffer.h"
+#include "vblank.h"
+#include "i915_drm.h"
+
+
+
+/*
+ * Correct a drawablePrivate's set of vblank flags WRT the current context.
+ * When considering multiple crtcs.
+ */
+GLuint
+intelFixupVblank(struct intel_context *intel, __DRIdrawablePrivate *dPriv)
+{
+   if (!intel->intelScreen->driScrnPriv->dri2.enabled &&
+       intel->intelScreen->driScrnPriv->ddx_version.minor >= 7) {
+      volatile drm_i915_sarea_t *sarea = intel->sarea;
+      drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w,
+				   .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h };
+      drm_clip_rect_t planeA_rect = { .x1 = sarea->planeA_x, .y1 = sarea->planeA_y,
+				     .x2 = sarea->planeA_x + sarea->planeA_w,
+				     .y2 = sarea->planeA_y + sarea->planeA_h };
+      drm_clip_rect_t planeB_rect = { .x1 = sarea->planeB_x, .y1 = sarea->planeB_y,
+				     .x2 = sarea->planeB_x + sarea->planeB_w,
+				     .y2 = sarea->planeB_y + sarea->planeB_h };
+      GLint areaA = driIntersectArea( drw_rect, planeA_rect );
+      GLint areaB = driIntersectArea( drw_rect, planeB_rect );
+      GLuint flags = dPriv->vblFlags;
+
+      /* Update vblank info
+       */
+      if (areaB > areaA || (areaA == areaB && areaB > 0)) {
+	 flags = dPriv->vblFlags | VBLANK_FLAG_SECONDARY;
+      } else {
+	 flags = dPriv->vblFlags & ~VBLANK_FLAG_SECONDARY;
+      }
+
+      /* Do the stupid test: Is one of them actually disabled?
+       */
+      if (sarea->planeA_w == 0 || sarea->planeA_h == 0) {
+	 flags = dPriv->vblFlags | VBLANK_FLAG_SECONDARY;
+      } else if (sarea->planeB_w == 0 || sarea->planeB_h == 0) {
+	 flags = dPriv->vblFlags & ~VBLANK_FLAG_SECONDARY;
+      }
+
+      return flags;
+   } else {
+      return dPriv->vblFlags & ~VBLANK_FLAG_SECONDARY;
+   }
+}
+
+
+/**
+ * Called from driSwapBuffers()
+ */
+void
+intelSwapBuffers(__DRIdrawablePrivate * dPriv)
+{
+   __DRIscreenPrivate *psp = dPriv->driScreenPriv;
+
+   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+      GET_CURRENT_CONTEXT(ctx);
+      struct intel_context *intel;
+
+      if (ctx == NULL)
+	 return;
+
+      intel = intel_context(ctx);
+
+      if (ctx->Visual.doubleBufferMode) {
+	 GLboolean missed_target;
+	 struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
+	 int64_t ust;
+         
+	 _mesa_notifySwapBuffers(ctx);  /* flush pending rendering comands */
+
+	/*
+	 * The old swapping ioctl was incredibly racy, just wait for vblank
+	 * and do the swap ourselves.
+	 */
+	 driWaitForVBlank(dPriv, &missed_target);
+
+	 /*
+	  * Update each buffer's vbl_pending so we don't get too out of
+	  * sync
+	  */
+	 intel_get_renderbuffer(&intel_fb->Base,
+		   		BUFFER_BACK_LEFT)->vbl_pending = dPriv->vblSeq;
+         intel_get_renderbuffer(&intel_fb->Base,
+		   		BUFFER_FRONT_LEFT)->vbl_pending = dPriv->vblSeq;
+
+	 intelCopyBuffer(dPriv, NULL);
+
+	 intel_fb->swap_count++;
+	 (*psp->systemTime->getUST) (&ust);
+	 if (missed_target) {
+	    intel_fb->swap_missed_count++;
+	    intel_fb->swap_missed_ust = ust - intel_fb->swap_ust;
+	 }
+
+	 intel_fb->swap_ust = ust;
+      }
+      drmCommandNone(intel->driFd, DRM_I915_GEM_THROTTLE);
+   }
+   else {
+      /* XXX this shouldn't be an error but we can't handle it for now */
+      fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__);
+   }
+}
+
+
+/**
+ * Called from driCopySubBuffer()
+ */
+void
+intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h)
+{
+   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+      struct intel_context *intel =
+         (struct intel_context *) dPriv->driContextPriv->driverPrivate;
+      GLcontext *ctx = &intel->ctx;
+
+      if (ctx->Visual.doubleBufferMode) {
+         drm_clip_rect_t rect;
+         rect.x1 = x + dPriv->x;
+         rect.y1 = (dPriv->h - y - h) + dPriv->y;
+         rect.x2 = rect.x1 + w;
+         rect.y2 = rect.y1 + h;
+         _mesa_notifySwapBuffers(ctx);  /* flush pending rendering comands */
+         intelCopyBuffer(dPriv, &rect);
+      }
+   }
+   else {
+      /* XXX this shouldn't be an error but we can't handle it for now */
+      fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__);
+   }
+}
+
+
+/**
+ * This will be called whenever the currently bound window is moved/resized.
+ * XXX: actually, it seems to NOT be called when the window is only moved (BP).
+ */
+void
+intelWindowMoved(struct intel_context *intel)
+{
+   GLcontext *ctx = &intel->ctx;
+   __DRIdrawablePrivate *dPriv = intel->driDrawable;
+   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
+
+   if (!intel->intelScreen->driScrnPriv->dri2.enabled &&
+       intel->intelScreen->driScrnPriv->ddx_version.minor >= 7) {
+      GLuint flags = intelFixupVblank(intel, dPriv);
+
+      /* Check to see if we changed pipes */
+      if (flags != dPriv->vblFlags && dPriv->vblFlags &&
+	  !(dPriv->vblFlags & VBLANK_FLAG_NO_IRQ)) {
+	 int64_t count;
+	 drmVBlank vbl;
+	 int i;
+
+	 /*
+	  * Deal with page flipping
+	  */
+	 vbl.request.type = DRM_VBLANK_ABSOLUTE;
+
+	 if ( dPriv->vblFlags & VBLANK_FLAG_SECONDARY ) {
+	    vbl.request.type |= DRM_VBLANK_SECONDARY;
+	 }
+
+	 for (i = 0; i < 2; i++) {
+	    if (!intel_fb->color_rb[i] ||
+		(intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <=
+		(1<<23))
+	       continue;
+
+	    vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending;
+	    drmWaitVBlank(intel->driFd, &vbl);
+	 }
+
+	 /*
+	  * Update msc_base from old pipe
+	  */
+	 driDrawableGetMSC32(dPriv->driScreenPriv, dPriv, &count);
+	 dPriv->msc_base = count;
+	 /*
+	  * Then get new vblank_base and vblSeq values
+	  */
+	 dPriv->vblFlags = flags;
+	 driGetCurrentVBlank(dPriv);
+	 dPriv->vblank_base = dPriv->vblSeq;
+
+	 intel_fb->vbl_waited = dPriv->vblSeq;
+
+	 for (i = 0; i < 2; i++) {
+	    if (intel_fb->color_rb[i])
+	       intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited;
+	 }
+      }
+   } else {
+      dPriv->vblFlags &= ~VBLANK_FLAG_SECONDARY;
+   }
+
+   /* Update Mesa's notion of window size */
+   driUpdateFramebufferSize(ctx, dPriv);
+   intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */
+
+   /* Update hardware scissor */
+   if (ctx->Driver.Scissor != NULL) {
+      ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
+			  ctx->Scissor.Width, ctx->Scissor.Height);
+   }
+
+   /* Re-calculate viewport related state */
+   if (ctx->Driver.DepthRange != NULL)
+      ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far );
+}
diff --git a/src/mesa/drivers/dri/intel/intel_swapbuffers.h b/src/mesa/drivers/dri/intel/intel_swapbuffers.h
new file mode 100644
index 0000000000..75bb6242ff
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_swapbuffers.h
@@ -0,0 +1,52 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef INTEL_SWAPBUFFERS_H
+#define INTEL_SWAPBUFFERS_H
+
+#include "dri_util.h"
+#include "drm.h"
+
+struct intel_context;
+struct intel_framebuffer;
+
+
+extern void
+intelSwapBuffers(__DRIdrawablePrivate * dPriv);
+
+extern void
+intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h);
+
+extern GLuint
+intelFixupVblank(struct intel_context *intel, __DRIdrawablePrivate *dPriv);
+
+extern void
+intelWindowMoved(struct intel_context *intel);
+
+
+#endif /* INTEL_SWAPBUFFERS_H */
diff --git a/src/mesa/drivers/dri/intel/intel_tex.c b/src/mesa/drivers/dri/intel/intel_tex.c
index 23455a4c74..ae0994b183 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.c
+++ b/src/mesa/drivers/dri/intel/intel_tex.c
@@ -93,7 +93,7 @@ intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage)
 static void *
 do_memcpy(void *dest, const void *src, size_t n)
 {
-   if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) {
+   if ((((unsigned long) src) & 63) || (((unsigned long) dest) & 63)) {
       return __memcpy(dest, src, n);
    }
    else
@@ -222,22 +222,16 @@ intelInitTextureFuncs(struct dd_function_table *functions)
    functions->TexSubImage1D = intelTexSubImage1D;
    functions->TexSubImage2D = intelTexSubImage2D;
    functions->TexSubImage3D = intelTexSubImage3D;
-#ifdef I915
    functions->CopyTexImage1D = intelCopyTexImage1D;
    functions->CopyTexImage2D = intelCopyTexImage2D;
    functions->CopyTexSubImage1D = intelCopyTexSubImage1D;
    functions->CopyTexSubImage2D = intelCopyTexSubImage2D;
-#else
-   functions->CopyTexImage1D = _swrast_copy_teximage1d;
-   functions->CopyTexImage2D = _swrast_copy_teximage2d;
-   functions->CopyTexSubImage1D = _swrast_copy_texsubimage1d;
-   functions->CopyTexSubImage2D = _swrast_copy_texsubimage2d;
-#endif
    functions->GetTexImage = intelGetTexImage;
    functions->GenerateMipmap = intelGenerateMipmap;
 
    /* compressed texture functions */
    functions->CompressedTexImage2D = intelCompressedTexImage2D;
+   functions->CompressedTexSubImage2D = intelCompressedTexSubImage2D;
    functions->GetCompressedTexImage = intelGetCompressedTexImage;
 
    functions->NewTextureObject = intelNewTextureObject;
diff --git a/src/mesa/drivers/dri/intel/intel_tex.h b/src/mesa/drivers/dri/intel/intel_tex.h
index 6219c1c953..742ccc043a 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.h
+++ b/src/mesa/drivers/dri/intel/intel_tex.h
@@ -130,6 +130,16 @@ void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
 				struct gl_texture_object *texObj,
 				struct gl_texture_image *texImage );
 
+void intelCompressedTexSubImage2D(GLcontext * ctx,
+				  GLenum target,
+				  GLint level,
+				  GLint xoffset, GLint yoffset,
+				  GLsizei width, GLsizei height,
+				  GLenum format, GLsizei imageSize,
+				  const GLvoid * pixels,
+				  struct gl_texture_object *texObj,
+				  struct gl_texture_image *texImage);
+
 void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
 				GLvoid *pixels,
 				struct gl_texture_object *texObj,
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index f4cb4a781c..08437aa0e2 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -98,7 +98,9 @@ do_copy_texsubimage(struct intel_context *intel,
       get_teximage_source(intel, internalFormat);
 
    if (!intelImage->mt || !src) {
-      DBG("%s fail %p %p\n", __FUNCTION__, intelImage->mt, src);
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "%s fail %p %p\n",
+		 __FUNCTION__, intelImage->mt, src);
       return GL_FALSE;
    }
 
@@ -110,56 +112,53 @@ do_copy_texsubimage(struct intel_context *intel,
                                                        intelImage->level);
       const GLint orig_x = x;
       const GLint orig_y = y;
-      const struct gl_framebuffer *fb = ctx->DrawBuffer;
-
-      if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax,
-                               &x, &y, &width, &height)) {
-         /* Update dst for clipped src.  Need to also clip the source rect.
-          */
-         dstx += x - orig_x;
-         dsty += y - orig_y;
-
-         if (ctx->ReadBuffer->Name == 0) {
-            /* reading from a window, adjust x, y */
-            __DRIdrawablePrivate *dPriv = intel->driDrawable;
-            GLuint window_y;
-            /* window_y = position of window on screen if y=0=bottom */
-            window_y = intel->intelScreen->height - (dPriv->y + dPriv->h);
-            y = window_y + y;
-            x += dPriv->x;
-         }
-         else {
-            /* reading from a FBO */
-            /* invert Y */
-            y = ctx->ReadBuffer->Height - y - 1;
-         }
-
-
-         /* A bit of fiddling to get the blitter to work with -ve
-          * pitches.  But we get a nice inverted blit this way, so it's
-          * worth it:
-          */
-         intelEmitCopyBlit(intel,
-                           intelImage->mt->cpp,
-                           -src->pitch,
-                           src->buffer,
-                           src->height * src->pitch * src->cpp,
-			   src->tiling,
-                           intelImage->mt->pitch,
-                           intelImage->mt->region->buffer,
-                           image_offset,
-			   intelImage->mt->region->tiling,
-                           x, y + height, dstx, dsty, width, height,
-			   GL_COPY); /* ? */
+      GLshort src_pitch;
+
+      /* Update dst for clipped src.  Need to also clip the source rect. */
+      dstx += x - orig_x;
+      dsty += y - orig_y;
+
+      /* image_offset may be non-page-aligned, but that's illegal for tiling. */
+      assert(intelImage->mt->region->tiling == I915_TILING_NONE);
+
+      if (ctx->ReadBuffer->Name == 0) {
+	 /* reading from a window, adjust x, y */
+	 __DRIdrawablePrivate *dPriv = intel->driDrawable;
+	 y = dPriv->y + (dPriv->h - (y + height));
+	 x += dPriv->x;
+
+	 /* Invert the data coming from the source rectangle due to GL
+	  * and hardware disagreeing on where y=0 is.
+	  *
+	  * It appears that our offsets and pitches get mangled
+	  * appropriately by the hardware, and we don't need to adjust them
+	  * on our own.
+	  */
+	 src_pitch = -src->pitch;
+      } else {
+	 /* reading from a FBO, y is already oriented the way we like */
+	 src_pitch = src->pitch;
       }
-   }
 
+      intelEmitCopyBlit(intel,
+			intelImage->mt->cpp,
+			src_pitch,
+			src->buffer,
+			0,
+			src->tiling,
+			intelImage->mt->pitch,
+			intelImage->mt->region->buffer,
+			image_offset,
+			intelImage->mt->region->tiling,
+			x, y, dstx, dsty, width, height,
+			GL_COPY);
+   }
 
    UNLOCK_HARDWARE(intel);
 
    /* GL_SGIS_generate_mipmap */
    if (intelImage->level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      intel_generate_mipmap(ctx, target, texObj);
+      ctx->Driver.GenerateMipmap(ctx, target, texObj);
    }
 
    return GL_TRUE;
@@ -180,6 +179,7 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
       _mesa_select_tex_object(ctx, texUnit, target);
    struct gl_texture_image *texImage =
       _mesa_select_tex_image(ctx, texObj, target, level);
+   int srcx, srcy, dstx, dsty, height;
 
    if (border)
       goto fail;
@@ -191,10 +191,20 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
                           width, border,
                           GL_RGBA, CHAN_TYPE, NULL,
                           &ctx->DefaultPacking, texObj, texImage);
+   srcx = x;
+   srcy = y;
+   dstx = 0;
+   dsty = 0;
+   height = 1;
+   if (!_mesa_clip_copytexsubimage(ctx,
+				   &dstx, &dsty,
+				   &srcx, &srcy,
+				   &width, &height))
+      return;
 
    if (!do_copy_texsubimage(intel_context(ctx), target,
                             intel_texture_image(texImage),
-                            internalFormat, 0, 0, x, y, width, 1))
+                            internalFormat, 0, 0, x, y, width, height))
       goto fail;
 
    return;
@@ -216,10 +226,21 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
       _mesa_select_tex_object(ctx, texUnit, target);
    struct gl_texture_image *texImage =
       _mesa_select_tex_image(ctx, texObj, target, level);
+   int srcx, srcy, dstx, dsty;
 
    if (border)
       goto fail;
 
+   srcx = x;
+   srcy = y;
+   dstx = 0;
+   dsty = 0;
+   if (!_mesa_clip_copytexsubimage(ctx,
+				   &dstx, &dsty,
+				   &srcx, &srcy,
+				   &width, &height))
+      return;
+
    /* Setup or redefine the texture object, mipmap tree and texture
     * image.  Don't populate yet.  
     */
diff --git a/src/mesa/drivers/dri/intel/intel_tex_format.c b/src/mesa/drivers/dri/intel/intel_tex_format.c
index 2be060dd3e..2715a540d0 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_format.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_format.c
@@ -16,7 +16,7 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
                          GLenum format, GLenum type)
 {
    struct intel_context *intel = intel_context(ctx);
-   const GLboolean do32bpt = (intel->ctx.Visual.rgbBits == 32);
+   const GLboolean do32bpt = (intel->ctx.Visual.rgbBits >= 24);
 
    switch (internalFormat) {
    case 4:
@@ -134,8 +134,14 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
    case GL_DEPTH_COMPONENT16:
    case GL_DEPTH_COMPONENT24:
    case GL_DEPTH_COMPONENT32:
+#if 0
       return &_mesa_texformat_z16;
-
+#else
+      /* fall-through.
+       * 16bpp depth texture can't be paired with a stencil buffer so
+       * always used combined depth/stencil format.
+       */
+#endif
    case GL_DEPTH_STENCIL_EXT:
    case GL_DEPTH24_STENCIL8_EXT:
       return &_mesa_texformat_s8_z24;
@@ -158,7 +164,7 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
    case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
    case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
    case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
-     return &_mesa_texformat_srgb_dxt1;
+      return &_mesa_texformat_srgb_dxt1;
 #endif
 
    default:
diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c
index 2ac7dceb0f..866022d0ce 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_image.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_image.c
@@ -209,7 +209,8 @@ try_pbo_upload(struct intel_context *intel,
       return GL_FALSE;
    }
 
-   src_offset = (GLuint) pixels;
+   /* note: potential 64-bit ptr to 32-bit int cast */
+   src_offset = (GLuint) (unsigned long) pixels;
 
    if (unpack->RowLength > 0)
       src_stride = unpack->RowLength;
@@ -264,7 +265,8 @@ try_pbo_zcopy(struct intel_context *intel,
       return GL_FALSE;
    }
 
-   src_offset = (GLuint) pixels;
+   /* note: potential 64-bit ptr to 32-bit int cast */
+   src_offset = (GLuint) (unsigned long) pixels;
 
    if (unpack->RowLength > 0)
       src_stride = unpack->RowLength;
diff --git a/src/mesa/drivers/dri/intel/intel_tex_subimage.c b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
index b752361886..f86de56897 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_subimage.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
@@ -184,3 +184,18 @@ intelTexSubImage1D(GLcontext * ctx,
                     format, type, pixels, packing, texObj, texImage);
 
 }
+
+void
+intelCompressedTexSubImage2D(GLcontext * ctx,
+			     GLenum target,
+			     GLint level,
+			     GLint xoffset, GLint yoffset,
+			     GLsizei width, GLsizei height,
+			     GLenum format, GLsizei imageSize,
+			     const GLvoid * pixels,
+			     struct gl_texture_object *texObj,
+			     struct gl_texture_image *texImage)
+{
+   fprintf(stderr, "stubbed CompressedTexSubImage2D: %dx%d@%dx%d\n",
+	   width, height, xoffset, yoffset);
+}
diff --git a/src/mesa/drivers/dri/mach64/mach64_context.c b/src/mesa/drivers/dri/mach64/mach64_context.c
index 99abd209b6..9c7f513c6f 100644
--- a/src/mesa/drivers/dri/mach64/mach64_context.c
+++ b/src/mesa/drivers/dri/mach64/mach64_context.c
@@ -57,10 +57,6 @@
 #include "utils.h"
 #include "vblank.h"
 
-#define need_GL_ARB_multisample
-#define need_GL_ARB_vertex_buffer_object
-#include "extension_helper.h"
-
 #ifndef MACH64_DEBUG
 int MACH64_DEBUG = (0);
 #endif
@@ -82,9 +78,7 @@ static const struct dri_debug_control debug_control[] =
 
 const struct dri_extension card_extensions[] =
 {
-    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
     { "GL_ARB_multitexture",               NULL },
-    { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
     { "GL_EXT_texture_edge_clamp",         NULL },
     { "GL_MESA_ycbcr_texture",             NULL },
     { "GL_SGIS_generate_mipmap",           NULL },
@@ -196,6 +190,7 @@ GLboolean mach64CreateContext( const __GLcontextModes *glVisual,
    ctx->Const.MaxTextureUnits = 2;
    ctx->Const.MaxTextureImageUnits = 2;
    ctx->Const.MaxTextureCoordUnits = 2;
+   ctx->Const.MaxDrawBuffers = 1;
 
    heap = mach64Screen->IsPCI ? MACH64_CARD_HEAP : MACH64_AGP_HEAP;
 
diff --git a/src/mesa/drivers/dri/mach64/mach64_context.h b/src/mesa/drivers/dri/mach64/mach64_context.h
index 55e0618ff8..854751626d 100644
--- a/src/mesa/drivers/dri/mach64/mach64_context.h
+++ b/src/mesa/drivers/dri/mach64/mach64_context.h
@@ -294,7 +294,13 @@ extern GLboolean mach64UnbindContext( __DRIcontextPrivate *driContextPriv );
 #define LE32_OUT( x, y )	do { *(GLuint *)(x) = (y); } while (0)
 #define LE32_OUT_FLOAT( x, y )	do { *(GLfloat *)(x) = (y); } while (0)
 #else
+#ifndef __OpenBSD__
 #include <byteswap.h>
+#else
+#include <machine/endian.h>
+#define bswap_32 bswap32
+#endif
+
 #define LE32_IN( x )		bswap_32( *(GLuint *)(x) )
 #define LE32_IN_FLOAT( x )						\
 ({									\
diff --git a/src/mesa/drivers/dri/mach64/mach64_screen.c b/src/mesa/drivers/dri/mach64/mach64_screen.c
index 6bfb4c32b1..43e59594dd 100644
--- a/src/mesa/drivers/dri/mach64/mach64_screen.c
+++ b/src/mesa/drivers/dri/mach64/mach64_screen.c
@@ -93,6 +93,7 @@ mach64FillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[2];
     uint8_t stencil_bits_array[2];
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = depth_bits;
     depth_bits_array[1] = depth_bits;
@@ -104,6 +105,8 @@ mach64FillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[0] = 0;
     stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 2 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
@@ -119,7 +122,8 @@ mach64FillInModes( __DRIscreenPrivate *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor, back_buffer_modes,
-			       back_buffer_factor);
+			       back_buffer_factor,
+                               msaa_samples_array, 1);
     if (configs == NULL) {
        fprintf(stderr, "[%s:%u] Error creating FBConfig!\n",
 	       __func__, __LINE__);
diff --git a/src/mesa/drivers/dri/mach64/mach64_tex.c b/src/mesa/drivers/dri/mach64/mach64_tex.c
index 1f9d3c57eb..9fe267eafd 100644
--- a/src/mesa/drivers/dri/mach64/mach64_tex.c
+++ b/src/mesa/drivers/dri/mach64/mach64_tex.c
@@ -152,6 +152,7 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
    case GL_ALPHA8:
    case GL_ALPHA12:
    case GL_ALPHA16:
+   case GL_COMPRESSED_ALPHA:
    case 2:
    case GL_LUMINANCE_ALPHA:
    case GL_LUMINANCE4_ALPHA4:
@@ -160,9 +161,11 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
    case GL_LUMINANCE12_ALPHA4:
    case GL_LUMINANCE12_ALPHA12:
    case GL_LUMINANCE16_ALPHA16:
+   case GL_COMPRESSED_LUMINANCE_ALPHA:
    case 4:
    case GL_RGBA:
    case GL_RGBA2:
+   case GL_COMPRESSED_RGBA:
       if (mmesa->mach64Screen->cpp == 4)
          return &_mesa_texformat_argb8888;
       else
@@ -193,6 +196,7 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
    case GL_RGB10:
    case GL_RGB12:
    case GL_RGB16:
+   case GL_COMPRESSED_RGB:
       if (mmesa->mach64Screen->cpp == 4)
          return &_mesa_texformat_argb8888;
       else
@@ -204,6 +208,7 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
    case GL_LUMINANCE8:
    case GL_LUMINANCE12:
    case GL_LUMINANCE16:
+   case GL_COMPRESSED_LUMINANCE:
       if (mmesa->mach64Screen->cpp == 4)
          return &_mesa_texformat_argb8888; /* inefficient but accurate */
       else
@@ -214,6 +219,7 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
    case GL_INTENSITY8:
    case GL_INTENSITY12:
    case GL_INTENSITY16:
+   case GL_COMPRESSED_INTENSITY:
       if (mmesa->mach64Screen->cpp == 4)
          return &_mesa_texformat_argb8888; /* inefficient but accurate */
       else
diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.c b/src/mesa/drivers/dri/mga/mga_xmesa.c
index 86da3a2cac..926534d6df 100644
--- a/src/mesa/drivers/dri/mga/mga_xmesa.c
+++ b/src/mesa/drivers/dri/mga/mga_xmesa.c
@@ -69,13 +69,9 @@
 
 #include "GL/internal/dri_interface.h"
 
-#define need_GL_ARB_multisample
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
 #define need_GL_ARB_vertex_program
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_gpu_program_parameters
-#define need_GL_EXT_multi_draw_arrays
 #define need_GL_EXT_secondary_color
 #if 0
 #define need_GL_EXT_paletted_texture
@@ -133,6 +129,7 @@ mgaFillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[3];
     uint8_t stencil_bits_array[3];
+    uint8_t msaa_samples_array[1];
 
 
     depth_bits_array[0] = 0;
@@ -147,6 +144,8 @@ mgaFillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[1] = 0;
     stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
@@ -162,7 +161,8 @@ mgaFillInModes( __DRIscreenPrivate *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor,
-			       back_buffer_modes, back_buffer_factor);
+			       back_buffer_modes, back_buffer_factor,
+                               msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
 		 __func__, __LINE__ );
@@ -385,13 +385,9 @@ static const struct dri_extension g400_extensions[] =
 
 static const struct dri_extension card_extensions[] =
 {
-   { "GL_ARB_multisample",            GL_ARB_multisample_functions },
-   { "GL_ARB_texture_compression",    GL_ARB_texture_compression_functions },
    { "GL_ARB_texture_rectangle",      NULL },
-   { "GL_ARB_vertex_buffer_object",   GL_ARB_vertex_buffer_object_functions },
    { "GL_EXT_blend_logic_op",         NULL },
    { "GL_EXT_fog_coord",              GL_EXT_fog_coord_functions },
-   { "GL_EXT_multi_draw_arrays",      GL_EXT_multi_draw_arrays_functions },
    /* paletted_textures currently doesn't work, but we could fix them later */
 #if defined( need_GL_EXT_paletted_texture )
    { "GL_EXT_shared_texture_palette", NULL },
@@ -539,6 +535,8 @@ mgaCreateContext( const __GLcontextModes *mesaVis,
    ctx->Const.MaxLineWidthAA = 10.0;
    ctx->Const.LineWidthGranularity = 1.0;
 
+   ctx->Const.MaxDrawBuffers = 1;
+
    mmesa->texture_depth = driQueryOptioni (&mmesa->optionCache,
 					   "texture_depth");
    if (mmesa->texture_depth == DRI_CONF_TEXTURE_DEPTH_FB)
diff --git a/src/mesa/drivers/dri/r128/r128_context.c b/src/mesa/drivers/dri/r128/r128_context.c
index 535a98cc01..f511a67bad 100644
--- a/src/mesa/drivers/dri/r128/r128_context.c
+++ b/src/mesa/drivers/dri/r128/r128_context.c
@@ -65,9 +65,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 int R128_DEBUG = 0;
 #endif
 
-#define need_GL_ARB_multisample
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
 #define need_GL_EXT_blend_minmax
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_secondary_color
@@ -75,12 +72,9 @@ int R128_DEBUG = 0;
 
 const struct dri_extension card_extensions[] =
 {
-    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
     { "GL_ARB_multitexture",               NULL },
-    { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
     { "GL_ARB_texture_env_add",            NULL },
     { "GL_ARB_texture_mirrored_repeat",    NULL },
-    { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
     { "GL_EXT_blend_subtract",             GL_EXT_blend_minmax_functions },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
     { "GL_EXT_texture_edge_clamp",         NULL },
@@ -229,6 +223,8 @@ GLboolean r128CreateContext( const __GLcontextModes *glVisual,
    ctx->Const.MaxLineWidthAA = 1.0;
    ctx->Const.LineWidthGranularity = 1.0;
 
+   ctx->Const.MaxDrawBuffers = 1;
+
 #if ENABLE_PERF_BOXES
    rmesa->boxes = driQueryOptionb(&rmesa->optionCache, "performance_boxes");
 #endif
diff --git a/src/mesa/drivers/dri/r128/r128_screen.c b/src/mesa/drivers/dri/r128/r128_screen.c
index cb3a147dba..7cda4ca5d3 100644
--- a/src/mesa/drivers/dri/r128/r128_screen.c
+++ b/src/mesa/drivers/dri/r128/r128_screen.c
@@ -422,7 +422,7 @@ r128FillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[2];
     uint8_t stencil_bits_array[2];
-
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = depth_bits;
     depth_bits_array[1] = depth_bits;
@@ -434,6 +434,8 @@ r128FillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[0] = 0;
     stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 2 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
@@ -446,26 +448,27 @@ r128FillInModes( __DRIscreenPrivate *psp,
         fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
     }
 
-   configs = driCreateConfigs(fb_format, fb_type,
-			      depth_bits_array, stencil_bits_array,
-			      depth_buffer_factor, back_buffer_modes,
-			      back_buffer_factor);
-   if (configs == NULL) {
-    fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
-              __LINE__);
-      return NULL;
-   }
+    configs = driCreateConfigs(fb_format, fb_type,
+                               depth_bits_array, stencil_bits_array,
+                               depth_buffer_factor, back_buffer_modes,
+                               back_buffer_factor,
+                               msaa_samples_array, 1);
+    if (configs == NULL) {
+        fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
+                __LINE__);
+        return NULL;
+    }
 
-   /* Mark the visual as slow if there are "fake" stencil bits.
-    */
-   for (i = 0; configs[i]; i++) {
-      m = &configs[i]->modes;
-      if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) {
-         m->visualRating = GLX_SLOW_CONFIG;
-      }
-   }
+    /* Mark the visual as slow if there are "fake" stencil bits.
+     */
+    for (i = 0; configs[i]; i++) {
+        m = &configs[i]->modes;
+        if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) {
+            m->visualRating = GLX_SLOW_CONFIG;
+        }
+    }
 
-   return (const __DRIconfig **) configs;
+    return (const __DRIconfig **) configs;
 }
 
 
diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c
index 1f5a9f2427..431bf8235c 100644
--- a/src/mesa/drivers/dri/r200/r200_context.c
+++ b/src/mesa/drivers/dri/r200/r200_context.c
@@ -63,14 +63,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "radeon_span.h"
 
-#define need_GL_ARB_multisample
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
 #define need_GL_ARB_vertex_program
 #define need_GL_ATI_fragment_shader
 #define need_GL_EXT_blend_minmax
 #define need_GL_EXT_fog_coord
-#define need_GL_EXT_multi_draw_arrays
 #define need_GL_EXT_secondary_color
 #define need_GL_EXT_blend_equation_separate
 #define need_GL_EXT_blend_func_separate
@@ -118,20 +114,16 @@ static const GLubyte *r200GetString( GLcontext *ctx, GLenum name )
  */
 const struct dri_extension card_extensions[] =
 {
-    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
     { "GL_ARB_multitexture",               NULL },
     { "GL_ARB_texture_border_clamp",       NULL },
-    { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
     { "GL_ARB_texture_env_add",            NULL },
     { "GL_ARB_texture_env_combine",        NULL },
     { "GL_ARB_texture_env_dot3",           NULL },
     { "GL_ARB_texture_env_crossbar",       NULL },
     { "GL_ARB_texture_mirrored_repeat",    NULL },
-    { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
     { "GL_EXT_blend_minmax",               GL_EXT_blend_minmax_functions },
     { "GL_EXT_blend_subtract",             NULL },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
-    { "GL_EXT_multi_draw_arrays",          GL_EXT_multi_draw_arrays_functions },
     { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
     { "GL_EXT_stencil_wrap",               NULL },
     { "GL_EXT_texture_edge_clamp",         NULL },
@@ -426,6 +418,8 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    ctx->Const.VertexProgram.MaxNativeParameters = R200_VSF_MAX_PARAM;
    ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
 
+   ctx->Const.MaxDrawBuffers = 1;
+
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 8bad6467f2..6d68309d3f 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -76,19 +76,17 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 int future_hw_tcl_on = 1;
 int hw_tcl_on = 1;
 
-#define need_GL_EXT_stencil_two_side
-#define need_GL_ARB_multisample
+#define need_GL_VERSION_2_0
 #define need_GL_ARB_point_parameters
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
 #define need_GL_ARB_vertex_program
-#define need_GL_EXT_blend_minmax
-//#define need_GL_EXT_fog_coord
-#define need_GL_EXT_multi_draw_arrays
-#define need_GL_EXT_secondary_color
 #define need_GL_EXT_blend_equation_separate
 #define need_GL_EXT_blend_func_separate
+#define need_GL_EXT_blend_minmax
+//#define need_GL_EXT_fog_coord
 #define need_GL_EXT_gpu_program_parameters
+#define need_GL_EXT_secondary_color
+#define need_GL_EXT_stencil_two_side
+#define need_GL_ATI_separate_stencil
 #define need_GL_NV_vertex_program
 #include "extension_helper.h"
 
@@ -96,27 +94,23 @@ const struct dri_extension card_extensions[] = {
   /* *INDENT-OFF* */
   {"GL_ARB_depth_texture",		NULL},
   {"GL_ARB_fragment_program",		NULL},
-  {"GL_ARB_multisample",		GL_ARB_multisample_functions},
   {"GL_ARB_multitexture",		NULL},
   {"GL_ARB_point_parameters",		GL_ARB_point_parameters_functions},
   {"GL_ARB_shadow",			NULL},
   {"GL_ARB_shadow_ambient",		NULL},
   {"GL_ARB_texture_border_clamp",	NULL},
-  {"GL_ARB_texture_compression",	GL_ARB_texture_compression_functions},
   {"GL_ARB_texture_cube_map",		NULL},
   {"GL_ARB_texture_env_add",		NULL},
   {"GL_ARB_texture_env_combine",	NULL},
   {"GL_ARB_texture_env_crossbar",	NULL},
   {"GL_ARB_texture_env_dot3",		NULL},
   {"GL_ARB_texture_mirrored_repeat",	NULL},
-  {"GL_ARB_vertex_buffer_object",	GL_ARB_vertex_buffer_object_functions},
   {"GL_ARB_vertex_program",		GL_ARB_vertex_program_functions},
   {"GL_EXT_blend_equation_separate",	GL_EXT_blend_equation_separate_functions},
   {"GL_EXT_blend_func_separate",	GL_EXT_blend_func_separate_functions},
   {"GL_EXT_blend_minmax",		GL_EXT_blend_minmax_functions},
   {"GL_EXT_blend_subtract",		NULL},
 //  {"GL_EXT_fog_coord",			GL_EXT_fog_coord_functions },
-  {"GL_EXT_multi_draw_arrays",		GL_EXT_multi_draw_arrays_functions},
   {"GL_EXT_gpu_program_parameters",     GL_EXT_gpu_program_parameters_functions},
   {"GL_EXT_secondary_color", 		GL_EXT_secondary_color_functions},
   {"GL_EXT_shadow_funcs",		NULL},
@@ -129,6 +123,7 @@ const struct dri_extension card_extensions[] = {
   {"GL_EXT_texture_lod_bias",		NULL},
   {"GL_EXT_texture_mirror_clamp",	NULL},
   {"GL_EXT_texture_rectangle",		NULL},
+  {"GL_ATI_separate_stencil",		GL_ATI_separate_stencil_functions},
   {"GL_ATI_texture_env_combine3",	NULL},
   {"GL_ATI_texture_mirror_once",	NULL},
   {"GL_MESA_pack_invert",		NULL},
@@ -141,6 +136,16 @@ const struct dri_extension card_extensions[] = {
   /* *INDENT-ON* */
 };
 
+
+/**
+ * The GL 2.0 functions are needed to make display lists work with
+ * functions added by GL_ATI_separate_stencil.
+ */
+const struct dri_extension gl_20_extension[] = {
+  {"GL_VERSION_2_0",			GL_VERSION_2_0_functions },
+};
+
+
 extern struct tnl_pipeline_stage _r300_render_stage;
 extern const struct tnl_pipeline_stage _r300_tcl_stage;
 
@@ -346,6 +351,8 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	    ( /*512 */ RADEON_BUFFER_SIZE * 16 * 1024) / (4 * 4);
 #endif
 
+	ctx->Const.MaxDrawBuffers = 1;
+
 	/* Initialize the software rasterizer and helper modules.
 	 */
 	_swrast_CreateContext(ctx);
diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h
index 8b3fe431ab..5f344be116 100644
--- a/src/mesa/drivers/dri/r300/r300_reg.h
+++ b/src/mesa/drivers/dri/r300/r300_reg.h
@@ -64,7 +64,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R300_SE_VPORT_ZSCALE                0x1DA8
 #define R300_SE_VPORT_ZOFFSET               0x1DAC
 
-
+#define R300_VAP_PORT_IDX0		    0x2040
 /*
  * Vertex Array Processing (VAP) Control
  */
@@ -656,7 +656,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_GB_FOG_SELECT_C3A           (3 << 0)
 #	define R300_GB_FOG_SELECT_1_1_W         (4 << 0)
 #	define R300_GB_FOG_SELECT_Z		(5 << 0)
-#	define R300_GB_DEPTH_SELECT_Z		(0 << 3
+#	define R300_GB_DEPTH_SELECT_Z		(0 << 3)
 #	define R300_GB_DEPTH_SELECT_1_1_W	(1 << 3)
 #	define R300_GB_W_SELECT_1_W		(0 << 4)
 #	define R300_GB_W_SELECT_1		(1 << 4)
@@ -730,8 +730,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R500_RS_IP_TEX_PTR_Q_SHIFT 			18
 #define R500_RS_IP_COL_PTR_SHIFT 			24
 #define R500_RS_IP_COL_FMT_SHIFT 			27
-#	define R500_RS_COL_PTR(x)		        (x << 24)
-#       define R500_RS_COL_FMT(x)                       (x << 27)
+#	define R500_RS_COL_PTR(x)		        ((x) << 24)
+#       define R500_RS_COL_FMT(x)                       ((x) << 27)
 /* gap */
 #define R500_RS_IP_OFFSET_DIS 				(0 << 31)
 #define R500_RS_IP_OFFSET_EN 				(1 << 31)
@@ -1172,9 +1172,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R300_RS_IP_3				        0x431C
 #       define R300_RS_INTERP_SRC_SHIFT          2 /* TODO: check for removal */
 #       define R300_RS_INTERP_SRC_MASK           (7 << 2) /* TODO: check for removal */
-#	define R300_RS_TEX_PTR(x)		        (x << 0)
-#	define R300_RS_COL_PTR(x)		        (x << 6)
-#	define R300_RS_COL_FMT(x)		        (x << 9)
+#	define R300_RS_TEX_PTR(x)		        ((x) << 0)
+#	define R300_RS_COL_PTR(x)		        ((x) << 6)
+#	define R300_RS_COL_FMT(x)		        ((x) << 9)
 #	define R300_RS_COL_FMT_RGBA		        0
 #	define R300_RS_COL_FMT_RGB0		        1
 #	define R300_RS_COL_FMT_RGB1		        2
@@ -1184,10 +1184,10 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_RS_COL_FMT_111A		        8
 #	define R300_RS_COL_FMT_1110		        9
 #	define R300_RS_COL_FMT_1111		        10
-#	define R300_RS_SEL_S(x)		                (x << 13)
-#	define R300_RS_SEL_T(x)		                (x << 16)
-#	define R300_RS_SEL_R(x)		                (x << 19)
-#	define R300_RS_SEL_Q(x)		                (x << 22)
+#	define R300_RS_SEL_S(x)		                ((x) << 13)
+#	define R300_RS_SEL_T(x)		                ((x) << 16)
+#	define R300_RS_SEL_R(x)		                ((x) << 19)
+#	define R300_RS_SEL_Q(x)		                ((x) << 22)
 #	define R300_RS_SEL_C0		                0
 #	define R300_RS_SEL_C1		                1
 #	define R300_RS_SEL_C2		                2
@@ -2712,7 +2712,7 @@ enum {
 #   define R500_ALPHA_OP_COS				13
 #   define R500_ALPHA_OP_MDH				14
 #   define R500_ALPHA_OP_MDV				15
-#   define R500_ALPHA_ADDRD(x)				(x << 4)
+#   define R500_ALPHA_ADDRD(x)				((x) << 4)
 #   define R500_ALPHA_ADDRD_REL				(1 << 11)
 #  define R500_ALPHA_SEL_A_SHIFT			12
 #   define R500_ALPHA_SEL_A_SRC0			(0 << 12)
@@ -2756,16 +2756,16 @@ enum {
 #   define R500_ALPHA_OMOD_DIV_4			(5 << 26)
 #   define R500_ALPHA_OMOD_DIV_8			(6 << 26)
 #   define R500_ALPHA_OMOD_DISABLE			(7 << 26)
-#   define R500_ALPHA_TARGET(x)				(x << 29)
+#   define R500_ALPHA_TARGET(x)				((x) << 29)
 #   define R500_ALPHA_W_OMASK				(1 << 31)
 #define R500_US_ALU_ALPHA_ADDR_0			0x9800
-#   define R500_ALPHA_ADDR0(x)				(x << 0)
+#   define R500_ALPHA_ADDR0(x)				((x) << 0)
 #   define R500_ALPHA_ADDR0_CONST			(1 << 8)
 #   define R500_ALPHA_ADDR0_REL				(1 << 9)
-#   define R500_ALPHA_ADDR1(x)				(x << 10)
+#   define R500_ALPHA_ADDR1(x)				((x) << 10)
 #   define R500_ALPHA_ADDR1_CONST			(1 << 18)
 #   define R500_ALPHA_ADDR1_REL				(1 << 19)
-#   define R500_ALPHA_ADDR2(x)				(x << 20)
+#   define R500_ALPHA_ADDR2(x)				((x) << 20)
 #   define R500_ALPHA_ADDR2_CONST			(1 << 28)
 #   define R500_ALPHA_ADDR2_REL				(1 << 29)
 #   define R500_ALPHA_SRCP_OP_1_MINUS_2A0		(0 << 30)
@@ -2786,7 +2786,7 @@ enum {
 #   define R500_ALU_RGBA_OP_SOP				(10 << 0)
 #   define R500_ALU_RGBA_OP_MDH				(11 << 0)
 #   define R500_ALU_RGBA_OP_MDV				(12 << 0)
-#   define R500_ALU_RGBA_ADDRD(x)			(x << 4)
+#   define R500_ALU_RGBA_ADDRD(x)			((x) << 4)
 #   define R500_ALU_RGBA_ADDRD_REL			(1 << 11)
 #  define R500_ALU_RGBA_SEL_C_SHIFT			12
 #   define R500_ALU_RGBA_SEL_C_SRC0			(0 << 12)
@@ -2913,16 +2913,16 @@ enum {
 #   define R500_ALU_RGB_OMOD_DIV_4			(5 << 26)
 #   define R500_ALU_RGB_OMOD_DIV_8			(6 << 26)
 #   define R500_ALU_RGB_OMOD_DISABLE			(7 << 26)
-#   define R500_ALU_RGB_TARGET(x)			(x << 29)
+#   define R500_ALU_RGB_TARGET(x)			((x) << 29)
 #   define R500_ALU_RGB_WMASK				(1 << 31)
 #define R500_US_ALU_RGB_ADDR_0				0x9000
-#   define R500_RGB_ADDR0(x)				(x << 0)
+#   define R500_RGB_ADDR0(x)				((x) << 0)
 #   define R500_RGB_ADDR0_CONST				(1 << 8)
 #   define R500_RGB_ADDR0_REL				(1 << 9)
-#   define R500_RGB_ADDR1(x)				(x << 10)
+#   define R500_RGB_ADDR1(x)				((x) << 10)
 #   define R500_RGB_ADDR1_CONST				(1 << 18)
 #   define R500_RGB_ADDR1_REL				(1 << 19)
-#   define R500_RGB_ADDR2(x)				(x << 20)
+#   define R500_RGB_ADDR2(x)				((x) << 20)
 #   define R500_RGB_ADDR2_CONST				(1 << 28)
 #   define R500_RGB_ADDR2_REL				(1 << 29)
 #   define R500_RGB_SRCP_OP_1_MINUS_2RGB0		(0 << 30)
@@ -2977,19 +2977,19 @@ enum {
 
 /* note that these are 8 bit lengths, despite the offsets, at least for R500 */
 #define R500_US_CODE_ADDR				0x4630
-#   define R500_US_CODE_START_ADDR(x)			(x << 0)
-#   define R500_US_CODE_END_ADDR(x)			(x << 16)
+#   define R500_US_CODE_START_ADDR(x)			((x) << 0)
+#   define R500_US_CODE_END_ADDR(x)			((x) << 16)
 #define R500_US_CODE_OFFSET				0x4638
-#   define R500_US_CODE_OFFSET_ADDR(x)			(x << 0)
+#   define R500_US_CODE_OFFSET_ADDR(x)			((x) << 0)
 #define R500_US_CODE_RANGE				0x4634
-#   define R500_US_CODE_RANGE_ADDR(x)			(x << 0)
-#   define R500_US_CODE_RANGE_SIZE(x)			(x << 16)
+#   define R500_US_CODE_RANGE_ADDR(x)			((x) << 0)
+#   define R500_US_CODE_RANGE_SIZE(x)			((x) << 16)
 #define R500_US_CONFIG					0x4600
 #   define R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO		(1 << 1)
 #define R500_US_FC_ADDR_0				0xa000
-#   define R500_FC_BOOL_ADDR(x)				(x << 0)
-#   define R500_FC_INT_ADDR(x)				(x << 8)
-#   define R500_FC_JUMP_ADDR(x)				(x << 16)
+#   define R500_FC_BOOL_ADDR(x)				((x) << 0)
+#   define R500_FC_INT_ADDR(x)				((x) << 8)
+#   define R500_FC_JUMP_ADDR(x)				((x) << 16)
 #   define R500_FC_JUMP_GLOBAL				(1 << 31)
 #define R500_US_FC_BOOL_CONST				0x4620
 #   define R500_FC_KBOOL(x)				(x)
@@ -3010,8 +3010,8 @@ enum {
 #   define R500_FC_A_OP_NONE				(0 << 6)
 #   define R500_FC_A_OP_POP				(1 << 6)
 #   define R500_FC_A_OP_PUSH				(2 << 6)
-#   define R500_FC_JUMP_FUNC(x)				(x << 8)
-#   define R500_FC_B_POP_CNT(x)				(x << 16)
+#   define R500_FC_JUMP_FUNC(x)				((x) << 8)
+#   define R500_FC_B_POP_CNT(x)				((x) << 16)
 #   define R500_FC_B_OP0_NONE				(0 << 24)
 #   define R500_FC_B_OP0_DECR				(1 << 24)
 #   define R500_FC_B_OP0_INCR				(2 << 24)
@@ -3020,14 +3020,14 @@ enum {
 #   define R500_FC_B_OP1_INCR				(2 << 26)
 #   define R500_FC_IGNORE_UNCOVERED			(1 << 28)
 #define R500_US_FC_INT_CONST_0				0x4c00
-#   define R500_FC_INT_CONST_KR(x)			(x << 0)
-#   define R500_FC_INT_CONST_KG(x)			(x << 8)
-#   define R500_FC_INT_CONST_KB(x)			(x << 16)
+#   define R500_FC_INT_CONST_KR(x)			((x) << 0)
+#   define R500_FC_INT_CONST_KG(x)			((x) << 8)
+#   define R500_FC_INT_CONST_KB(x)			((x) << 16)
 /* _0 through _15 */
 #define R500_US_FORMAT0_0				0x4640
-#   define R500_FORMAT_TXWIDTH(x)			(x << 0)
-#   define R500_FORMAT_TXHEIGHT(x)			(x << 11)
-#   define R500_FORMAT_TXDEPTH(x)			(x << 22)
+#   define R500_FORMAT_TXWIDTH(x)			((x) << 0)
+#   define R500_FORMAT_TXHEIGHT(x)			((x) << 11)
+#   define R500_FORMAT_TXDEPTH(x)			((x) << 22)
 /* _0 through _3 */
 #define R500_US_OUT_FMT_0				0x46a4
 #   define R500_OUT_FMT_C4_8				(0 << 0)
@@ -3068,12 +3068,12 @@ enum {
 #   define R500_C3_SEL_R				(1 << 14)
 #   define R500_C3_SEL_G				(2 << 14)
 #   define R500_C3_SEL_B				(3 << 14)
-#   define R500_OUT_SIGN(x)				(x << 16)
+#   define R500_OUT_SIGN(x)				((x) << 16)
 #   define R500_ROUND_ADJ				(1 << 20)
 #define R500_US_PIXSIZE					0x4604
 #   define R500_PIX_SIZE(x)				(x)
 #define R500_US_TEX_ADDR_0				0x9800
-#   define R500_TEX_SRC_ADDR(x)				(x << 0)
+#   define R500_TEX_SRC_ADDR(x)				((x) << 0)
 #   define R500_TEX_SRC_ADDR_REL			(1 << 7)
 #   define R500_TEX_SRC_S_SWIZ_R			(0 << 8)
 #   define R500_TEX_SRC_S_SWIZ_G			(1 << 8)
@@ -3091,7 +3091,7 @@ enum {
 #   define R500_TEX_SRC_Q_SWIZ_G			(1 << 14)
 #   define R500_TEX_SRC_Q_SWIZ_B			(2 << 14)
 #   define R500_TEX_SRC_Q_SWIZ_A			(3 << 14)
-#   define R500_TEX_DST_ADDR(x)				(x << 16)
+#   define R500_TEX_DST_ADDR(x)				((x) << 16)
 #   define R500_TEX_DST_ADDR_REL			(1 << 23)
 #   define R500_TEX_DST_R_SWIZ_R			(0 << 24)
 #   define R500_TEX_DST_R_SWIZ_G			(1 << 24)
@@ -3110,7 +3110,7 @@ enum {
 #   define R500_TEX_DST_A_SWIZ_B			(2 << 30)
 #   define R500_TEX_DST_A_SWIZ_A			(3 << 30)
 #define R500_US_TEX_ADDR_DXDY_0				0xa000
-#   define R500_DX_ADDR(x)				(x << 0)
+#   define R500_DX_ADDR(x)				((x) << 0)
 #   define R500_DX_ADDR_REL				(1 << 7)
 #   define R500_DX_S_SWIZ_R				(0 << 8)
 #   define R500_DX_S_SWIZ_G				(1 << 8)
@@ -3128,7 +3128,7 @@ enum {
 #   define R500_DX_Q_SWIZ_G				(1 << 14)
 #   define R500_DX_Q_SWIZ_B				(2 << 14)
 #   define R500_DX_Q_SWIZ_A				(3 << 14)
-#   define R500_DY_ADDR(x)				(x << 16)
+#   define R500_DY_ADDR(x)				((x) << 16)
 #   define R500_DY_ADDR_REL				(1 << 17)
 #   define R500_DY_S_SWIZ_R				(0 << 24)
 #   define R500_DY_S_SWIZ_G				(1 << 24)
@@ -3147,7 +3147,7 @@ enum {
 #   define R500_DY_Q_SWIZ_B				(2 << 30)
 #   define R500_DY_Q_SWIZ_A				(3 << 30)
 #define R500_US_TEX_INST_0				0x9000
-#   define R500_TEX_ID(x)				(x << 16)
+#   define R500_TEX_ID(x)				((x) << 16)
 #   define R500_TEX_INST_NOP				(0 << 22)
 #   define R500_TEX_INST_LD				(1 << 22)
 #   define R500_TEX_INST_TEXKILL			(2 << 22)
@@ -3208,9 +3208,9 @@ enum {
 #define R300_PACKET3_3D_LOAD_VBPNTR         0x00002F00
 
 #define R300_PACKET3_INDX_BUFFER            0x00003300
-#    define R300_EB_UNK1_SHIFT                      24
-#    define R300_EB_UNK1                    (0x80<<24)
-#    define R300_EB_UNK2                        0x0810
+#    define R300_INDX_BUFFER_DST_SHIFT          0
+#    define R300_INDX_BUFFER_SKIP_SHIFT         16
+#    define R300_INDX_BUFFER_ONE_REG_WR		(1<<31)
 
 /* Same as R300_PACKET3_3D_DRAW_VBUF but without VAP_VTX_FMT */
 #define R300_PACKET3_3D_DRAW_VBUF_2         0x00003400
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 8055952a76..f46477f0f6 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -197,7 +197,8 @@ static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type)
 		
 		if (!rmesa->radeon.radeonScreen->kernel_mm) {
 			OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
-			OUT_BATCH(R300_EB_UNK1 | (0 << 16) | R300_EB_UNK2);
+			OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
+	    			 (R300_VAP_PORT_IDX0 >> 2));
 			OUT_BATCH_RELOC(rmesa->state.elt_dma_offset,
 					rmesa->state.elt_dma_bo,
 					rmesa->state.elt_dma_offset,
@@ -205,7 +206,8 @@ static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type)
 			OUT_BATCH(vertex_count);
 		} else {
 			OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
-			OUT_BATCH(R300_EB_UNK1 | (0 << 16) | R300_EB_UNK2);
+			OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
+	    			 (R300_VAP_PORT_IDX0 >> 2));
 			OUT_BATCH(rmesa->state.elt_dma_offset);
 			OUT_BATCH(vertex_count);
 			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
@@ -424,6 +426,8 @@ static GLboolean r300RunRender(GLcontext * ctx,
 static int r300Fallback(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	const unsigned back = ctx->Stencil._BackFace;
+
 	/* Do we need to use new-style shaders?
 	 * Also is there a better way to do this? */
 	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
@@ -448,12 +452,14 @@ static int r300Fallback(GLcontext * ctx)
 
 	FALLBACK_IF(ctx->RenderMode != GL_RENDER);
 
-	FALLBACK_IF(ctx->Stencil._TestTwoSide
-		    && (ctx->Stencil.Ref[0] != ctx->Stencil.Ref[1]
-			|| ctx->Stencil.ValueMask[0] !=
-			ctx->Stencil.ValueMask[1]
-			|| ctx->Stencil.WriteMask[0] !=
-			ctx->Stencil.WriteMask[1]));
+	/* If GL_EXT_stencil_two_side is disabled, this fallback check can
+	 * be removed.
+	 */
+	FALLBACK_IF(ctx->Stencil.Ref[0] != ctx->Stencil.Ref[back]
+		    || ctx->Stencil.ValueMask[0] !=
+		    ctx->Stencil.ValueMask[back]
+		    || ctx->Stencil.WriteMask[0] !=
+		    ctx->Stencil.WriteMask[back]);
 
 	if (ctx->Extensions.NV_point_sprite || ctx->Extensions.ARB_point_sprite)
 		FALLBACK_IF(ctx->Point.PointSprite);
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index fe58ac00a8..93ef06fb26 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -781,6 +781,7 @@ static void r300Fogfv(GLcontext * ctx, GLenum pname, const GLfloat * param)
 			    R300_FG_FOG_BLEND_FN_EXP2;
 			fogScale.f = 0.3 * ctx->Fog.Density;
 			fogStart.f = 0.0;
+                        break;
 		default:
 			return;
 		}
@@ -972,15 +973,9 @@ static void r300StencilFuncSeparate(GLcontext * ctx, GLenum face,
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	GLuint refmask =
-	    (((ctx->Stencil.
-	       Ref[0] & 0xff) << R300_STENCILREF_SHIFT) | ((ctx->
-							    Stencil.
-							    ValueMask
-							    [0] &
-							    0xff)
-							   <<
-							   R300_STENCILMASK_SHIFT));
-
+	    ((ctx->Stencil.Ref[0] & 0xff) << R300_STENCILREF_SHIFT)
+	     | ((ctx->Stencil.ValueMask[0] & 0xff) << R300_STENCILMASK_SHIFT);
+	const unsigned back = ctx->Stencil._BackFace;
 	GLuint flag;
 
 	R300_STATECHANGE(rmesa, zs);
@@ -998,8 +993,7 @@ static void r300StencilFuncSeparate(GLcontext * ctx, GLenum face,
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |=
 	    (flag << R300_S_FRONT_FUNC_SHIFT);
 
-	if (ctx->Stencil._TestTwoSide)
-		flag = translate_func(ctx->Stencil.Function[1]);
+	flag = translate_func(ctx->Stencil.Function[back]);
 
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |=
 	    (flag << R300_S_BACK_FUNC_SHIFT);
@@ -1024,6 +1018,7 @@ static void r300StencilOpSeparate(GLcontext * ctx, GLenum face,
 				  GLenum fail, GLenum zfail, GLenum zpass)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	const unsigned back = ctx->Stencil._BackFace;
 
 	R300_STATECHANGE(rmesa, zs);
 	/* It is easier to mask what's left.. */
@@ -1040,23 +1035,13 @@ static void r300StencilOpSeparate(GLcontext * ctx, GLenum face,
 	    | (translate_stencil_op(ctx->Stencil.ZPassFunc[0]) <<
 	       R300_S_FRONT_ZPASS_OP_SHIFT);
 
-	if (ctx->Stencil._TestTwoSide) {
-		rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |=
-		    (translate_stencil_op(ctx->Stencil.FailFunc[1]) <<
-		     R300_S_BACK_SFAIL_OP_SHIFT)
-		    | (translate_stencil_op(ctx->Stencil.ZFailFunc[1]) <<
-		       R300_S_BACK_ZFAIL_OP_SHIFT)
-		    | (translate_stencil_op(ctx->Stencil.ZPassFunc[1]) <<
-		       R300_S_BACK_ZPASS_OP_SHIFT);
-	} else {
-		rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |=
-		    (translate_stencil_op(ctx->Stencil.FailFunc[0]) <<
-		     R300_S_BACK_SFAIL_OP_SHIFT)
-		    | (translate_stencil_op(ctx->Stencil.ZFailFunc[0]) <<
-		       R300_S_BACK_ZFAIL_OP_SHIFT)
-		    | (translate_stencil_op(ctx->Stencil.ZPassFunc[0]) <<
-		       R300_S_BACK_ZPASS_OP_SHIFT);
-	}
+	rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |=
+	    (translate_stencil_op(ctx->Stencil.FailFunc[back]) <<
+	     R300_S_BACK_SFAIL_OP_SHIFT)
+	    | (translate_stencil_op(ctx->Stencil.ZFailFunc[back]) <<
+	       R300_S_BACK_ZFAIL_OP_SHIFT)
+	    | (translate_stencil_op(ctx->Stencil.ZPassFunc[back]) <<
+	       R300_S_BACK_ZPASS_OP_SHIFT);
 }
 
 /* =============================================================
@@ -1647,6 +1632,13 @@ static void r300SetupRSUnit(GLcontext * ctx)
 		rs_col_count += count;
 	}
 
+	if (InputsRead & FRAG_BIT_FOGC) {
+		/* XXX FIX THIS
+		 * Just turn off the bit for now.
+		 * Need to do something similar to the color/texcoord inputs.
+		 */
+		InputsRead &= ~FRAG_BIT_FOGC;
+	}
 
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
 		int swiz;
diff --git a/src/mesa/drivers/dri/r300/radeon_program_pair.c b/src/mesa/drivers/dri/r300/radeon_program_pair.c
index a6da9f56b5..8a945d8537 100644
--- a/src/mesa/drivers/dri/r300/radeon_program_pair.c
+++ b/src/mesa/drivers/dri/r300/radeon_program_pair.c
@@ -473,6 +473,11 @@ static void allocate_input_registers(struct pair_state *s)
 		alloc_hw_reg(s, PROGRAM_INPUT, FRAG_ATTRIB_COL1, hwindex++);
 	InputsRead &= ~FRAG_BIT_COL1;
 
+	/* Fog coordinate */
+	if (InputsRead & FRAG_BIT_FOGC)
+		alloc_hw_reg(s, PROGRAM_INPUT, FRAG_ATTRIB_FOGC, hwindex++);
+	InputsRead &= ~FRAG_BIT_FOGC;
+
 	/* Anything else */
 	if (InputsRead)
 		error("Don't know how to handle inputs 0x%x\n", InputsRead);
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c
index 0296cffc20..7db62be7af 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_context.c
@@ -63,9 +63,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_tcl.h"
 #include "radeon_maos.h"
 
-#define need_GL_ARB_multisample
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
 #define need_GL_EXT_blend_minmax
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_secondary_color
@@ -81,16 +78,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 const struct dri_extension card_extensions[] =
 {
-    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
     { "GL_ARB_multitexture",               NULL },
     { "GL_ARB_texture_border_clamp",       NULL },
-    { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
     { "GL_ARB_texture_env_add",            NULL },
     { "GL_ARB_texture_env_combine",        NULL },
     { "GL_ARB_texture_env_crossbar",       NULL },
     { "GL_ARB_texture_env_dot3",           NULL },
     { "GL_ARB_texture_mirrored_repeat",    NULL },
-    { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
     { "GL_EXT_blend_logic_op",             NULL },
     { "GL_EXT_blend_subtract",             GL_EXT_blend_minmax_functions },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
@@ -344,6 +338,8 @@ radeonCreateContext( const __GLcontextModes *glVisual,
 
    rmesa->boxes = 0;
 
+   ctx->Const.MaxDrawBuffers = 1;
+
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c
index 8b06fb4d3e..21b94d2e52 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.c
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.c
@@ -212,6 +212,8 @@ DRI_CONF_BEGIN
 DRI_CONF_END;
 static const GLuint __driNConfigOptions = 17;
 
+extern const struct dri_extension gl_20_extension[];
+
 #ifndef RADEON_DEBUG
 
 static const struct dri_debug_control debug_control[] = {
@@ -279,7 +281,7 @@ radeonFillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[2];
     uint8_t stencil_bits_array[2];
-
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = depth_bits;
     depth_bits_array[1] = depth_bits;
@@ -291,6 +293,8 @@ radeonFillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[0] = 0;
     stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 2 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
@@ -306,7 +310,8 @@ radeonFillInModes( __DRIscreenPrivate *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor,
-			       back_buffer_modes, back_buffer_factor);
+			       back_buffer_modes, back_buffer_factor,
+			       msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
 		 __func__, __LINE__ );
@@ -1368,6 +1373,7 @@ static void radeonDestroyContext(__DRIcontextPrivate * driContextPriv)
 
 #endif
 
+
 /**
  * This is the driver specific part of the createNewScreen entry point.
  *
@@ -1420,6 +1426,8 @@ radeonInitScreen(__DRIscreenPrivate *psp)
    driInitSingleExtension( NULL, NV_vp_extension );
    driInitSingleExtension( NULL, ATI_fs_extension );
    driInitExtensions( NULL, point_extensions, GL_FALSE );
+#elif defined(RADEON_COMMON_FOR_R300)
+   driInitSingleExtension( NULL, gl_20_extension );
 #endif
 
    if (!radeonInitDriver(psp))
diff --git a/src/mesa/drivers/dri/s3v/s3v_context.c b/src/mesa/drivers/dri/s3v/s3v_context.c
index 14502f95ae..0a3bf7258d 100644
--- a/src/mesa/drivers/dri/s3v/s3v_context.c
+++ b/src/mesa/drivers/dri/s3v/s3v_context.c
@@ -108,6 +108,8 @@ GLboolean s3vCreateContext(const __GLcontextModes *glVisual,
 	ctx->Const.MaxLineWidthAA = 1.0;
 	ctx->Const.LineWidthGranularity = 1.0;
 
+	ctx->Const.MaxDrawBuffers = 1;
+
 	vmesa->texHeap = mmInit( 0, vmesa->s3vScreen->textureSize );
 	DEBUG(("vmesa->s3vScreen->textureSize = 0x%x\n",
 		vmesa->s3vScreen->textureSize));
diff --git a/src/mesa/drivers/dri/savage/savage_xmesa.c b/src/mesa/drivers/dri/savage/savage_xmesa.c
index a344aab71b..326d595352 100644
--- a/src/mesa/drivers/dri/savage/savage_xmesa.c
+++ b/src/mesa/drivers/dri/savage/savage_xmesa.c
@@ -59,9 +59,6 @@
 #include "drirenderbuffer.h"
 #include "texmem.h"
 
-#define need_GL_ARB_multisample
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
 #define need_GL_EXT_secondary_color
 #include "extension_helper.h"
 
@@ -133,10 +130,7 @@ struct timeval tv_s1,tv_f1;
 
 static const struct dri_extension card_extensions[] =
 {
-    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
     { "GL_ARB_multitexture",               NULL },
-    { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
-    { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
     { "GL_EXT_stencil_wrap",               NULL },
     { "GL_EXT_texture_lod_bias",           NULL },
     { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
@@ -350,7 +344,9 @@ savageCreateContext( const __GLcontextModes *mesaVis,
    ctx->Const.MaxLineWidthAA = 3.0;
    ctx->Const.LineWidthGranularity = 1.0;
 #endif
-   
+
+   ctx->Const.MaxDrawBuffers = 1;
+
    /* Dri stuff
     */
    imesa->hHWContext = driContextPriv->hHWContext;
@@ -914,7 +910,7 @@ savageFillInModes( __DRIscreenPrivate *psp,
 
     uint8_t depth_bits_array[2];
     uint8_t stencil_bits_array[2];
-
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = depth_bits;
     depth_bits_array[1] = depth_bits;
@@ -926,6 +922,8 @@ savageFillInModes( __DRIscreenPrivate *psp,
     stencil_bits_array[0] = 0;
     stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 2 : 1;
     back_buffer_factor  = (have_back_buffer) ? 2 : 1;
 
@@ -941,7 +939,8 @@ savageFillInModes( __DRIscreenPrivate *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor,
-			       back_buffer_modes, back_buffer_factor);
+			       back_buffer_modes, back_buffer_factor,
+                               msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
 		 __func__, __LINE__ );
diff --git a/src/mesa/drivers/dri/sis/sis_alloc.c b/src/mesa/drivers/dri/sis/sis_alloc.c
index 4ca4052803..ce34e44da2 100644
--- a/src/mesa/drivers/dri/sis/sis_alloc.c
+++ b/src/mesa/drivers/dri/sis/sis_alloc.c
@@ -137,7 +137,7 @@ sisAllocZStencilBuffer( sisContextPtr smesa )
 {
    int cpp = ( smesa->glCtx->Visual.depthBits +
                smesa->glCtx->Visual.stencilBits ) / 8;
-   unsigned char *addr;
+   char *addr;
 
    smesa->depth.bpp = cpp * 8;
    smesa->depth.pitch = ALIGNMENT(smesa->driDrawable->w * cpp, 4);
@@ -150,7 +150,7 @@ sisAllocZStencilBuffer( sisContextPtr smesa )
    addr = (char *)ALIGNMENT((unsigned long)addr, Z_BUFFER_HW_ALIGNMENT);
 
    smesa->depth.map = addr;
-   smesa->depth.offset = addr - smesa->FbBase;
+   smesa->depth.offset = addr - (char *)smesa->FbBase;
 
    /* stencil buffer is same as depth buffer */
    smesa->stencil.size = smesa->depth.size;
@@ -173,7 +173,7 @@ void
 sisAllocBackbuffer( sisContextPtr smesa )
 {
    int cpp = smesa->bytesPerPixel;
-   unsigned char *addr;
+   char *addr;
 
    smesa->back.bpp = smesa->bytesPerPixel * 8;
    smesa->back.pitch = ALIGNMENT(smesa->driDrawable->w * cpp, 4);
@@ -186,7 +186,7 @@ sisAllocBackbuffer( sisContextPtr smesa )
    addr = (char *)ALIGNMENT((unsigned long)addr, DRAW_BUFFER_HW_ALIGNMENT);
 
    smesa->back.map = addr;
-   smesa->back.offset = addr - smesa->FbBase;
+   smesa->back.offset = addr - (char *)smesa->FbBase;
 }
 
 void
diff --git a/src/mesa/drivers/dri/sis/sis_context.c b/src/mesa/drivers/dri/sis/sis_context.c
index 00d17da3ba..a070fe3d79 100644
--- a/src/mesa/drivers/dri/sis/sis_context.c
+++ b/src/mesa/drivers/dri/sis/sis_context.c
@@ -57,9 +57,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 
-#define need_GL_ARB_multisample
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_secondary_color
 #include "extension_helper.h"
@@ -74,12 +71,9 @@ int GlobalCmdQueueLen = 0;
 
 struct dri_extension card_extensions[] =
 {
-    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
     { "GL_ARB_multitexture",               NULL },
     { "GL_ARB_texture_border_clamp",       NULL },
-    { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
     { "GL_ARB_texture_mirrored_repeat",    NULL },
-    { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
     /*{ "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },*/
     { "GL_EXT_texture_lod_bias",           NULL },
     { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
@@ -91,9 +85,7 @@ struct dri_extension card_extensions[] =
 
 struct dri_extension card_extensions_6326[] =
 {
-    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
     /*{ "GL_ARB_texture_border_clamp",       NULL },*/
-    { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
     /*{ "GL_ARB_texture_mirrored_repeat",    NULL },*/
     /*{ "GL_MESA_ycbcr_texture",             NULL },*/
     { NULL,                                NULL }
diff --git a/src/mesa/drivers/dri/sis/sis_screen.c b/src/mesa/drivers/dri/sis/sis_screen.c
index b1a5d15236..9eb27fef08 100644
--- a/src/mesa/drivers/dri/sis/sis_screen.c
+++ b/src/mesa/drivers/dri/sis/sis_screen.c
@@ -77,6 +77,7 @@ sisFillInModes(__DRIscreenPrivate *psp, int bpp)
    };
    uint8_t depth_bits_array[4];
    uint8_t stencil_bits_array[4];
+   uint8_t msaa_samples_array[1];
 
    depth_bits_array[0] = 0;
    stencil_bits_array[0] = 0;
@@ -87,6 +88,8 @@ sisFillInModes(__DRIscreenPrivate *psp, int bpp)
    depth_bits_array[3] = 32;
    stencil_bits_array[3] = 0;
 
+   msaa_samples_array[0] = 0;
+
    depth_buffer_factor = 4;
    back_buffer_factor = 2;
 
@@ -100,7 +103,8 @@ sisFillInModes(__DRIscreenPrivate *psp, int bpp)
 
    configs = driCreateConfigs(fb_format, fb_type, depth_bits_array,
 			      stencil_bits_array, depth_buffer_factor,
-			      back_buffer_modes, back_buffer_factor);
+			      back_buffer_modes, back_buffer_factor,
+                              msaa_samples_array, 1);
    if (configs == NULL) {
       fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, __LINE__);
       return NULL;
diff --git a/src/mesa/drivers/dri/sis/sis_tris.c b/src/mesa/drivers/dri/sis/sis_tris.c
index 095941aea2..76d12d07b3 100644
--- a/src/mesa/drivers/dri/sis/sis_tris.c
+++ b/src/mesa/drivers/dri/sis/sis_tris.c
@@ -994,7 +994,7 @@ sisFlushPrimsLocked(sisContextPtr smesa)
 	 MMIO(REG_3D_PrimitiveSet, smesa->dwPrimitiveSet);
       }
       while (smesa->vb_last < smesa->vb_cur) {
-	 sis_emit_func(smesa, smesa->vb_last);
+	 sis_emit_func(smesa, (char *)smesa->vb_last);
 	 smesa->vb_last += incr;
       }
       mWait3DCmdQueue(1);
diff --git a/src/mesa/drivers/dri/swrast/swrast.c b/src/mesa/drivers/dri/swrast/swrast.c
index 49f1b8b9ee..2e7f11327e 100644
--- a/src/mesa/drivers/dri/swrast/swrast.c
+++ b/src/mesa/drivers/dri/swrast/swrast.c
@@ -66,12 +66,13 @@
 #define need_GL_ARB_vertex_program
 #define need_GL_APPLE_vertex_array_object
 #define need_GL_ATI_fragment_shader
+#define need_GL_ATI_separate_stencil
 #define need_GL_EXT_depth_bounds_test
 #define need_GL_EXT_framebuffer_object
 #define need_GL_EXT_framebuffer_blit
 #define need_GL_EXT_gpu_program_parameters
 #define need_GL_EXT_paletted_texture
-#define need_GL_IBM_multimode_draw_arrays
+#define need_GL_EXT_stencil_two_side
 #define need_GL_MESA_resize_buffers
 #define need_GL_NV_vertex_program
 #define need_GL_NV_fragment_program
@@ -96,12 +97,13 @@ const struct dri_extension card_extensions[] =
     { "GL_ARB_vertex_program",		GL_ARB_vertex_program_functions },
     { "GL_APPLE_vertex_array_object",	GL_APPLE_vertex_array_object_functions },
     { "GL_ATI_fragment_shader",		GL_ATI_fragment_shader_functions },
+    { "GL_ATI_separate_stencil",	GL_ATI_separate_stencil_functions },
     { "GL_EXT_depth_bounds_test",	GL_EXT_depth_bounds_test_functions },
     { "GL_EXT_framebuffer_object",	GL_EXT_framebuffer_object_functions },
     { "GL_EXT_framebuffer_blit",	GL_EXT_framebuffer_blit_functions },
     { "GL_EXT_gpu_program_parameters",	GL_EXT_gpu_program_parameters_functions },
     { "GL_EXT_paletted_texture",	GL_EXT_paletted_texture_functions },
-    { "GL_IBM_multimode_draw_arrays",	GL_IBM_multimode_draw_arrays_functions },
+    { "GL_EXT_stencil_two_side",	GL_EXT_stencil_two_side_functions },
     { "GL_MESA_resize_buffers",		GL_MESA_resize_buffers_functions },
     { "GL_NV_vertex_program",		GL_NV_vertex_program_functions },
     { "GL_NV_fragment_program",		GL_NV_fragment_program_functions },
@@ -145,6 +147,7 @@ swrastFillInModes(__DRIscreen *psp,
 
     uint8_t depth_bits_array[4];
     uint8_t stencil_bits_array[4];
+    uint8_t msaa_samples_array[1];
 
     depth_bits_array[0] = 0;
     depth_bits_array[1] = 0;
@@ -159,26 +162,38 @@ swrastFillInModes(__DRIscreen *psp,
     stencil_bits_array[2] = 0;
     stencil_bits_array[3] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+    msaa_samples_array[0] = 0;
+
     depth_buffer_factor = 4;
     back_buffer_factor = 2;
 
-    if (pixel_bits == 8) {
+    switch (pixel_bits) {
+    case 8:
 	fb_format = GL_RGB;
 	fb_type = GL_UNSIGNED_BYTE_2_3_3_REV;
-    }
-    else if (pixel_bits == 16) {
+	break;
+    case 16:
 	fb_format = GL_RGB;
 	fb_type = GL_UNSIGNED_SHORT_5_6_5;
-    }
-    else {
+	break;
+    case 24:
+	fb_format = GL_BGR;
+	fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+	break;
+    case 32:
 	fb_format = GL_BGRA;
 	fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+	break;
+    default:
+	fprintf(stderr, "[%s:%u] bad depth %d\n", __func__, __LINE__,
+		pixel_bits);
+	return NULL;
     }
 
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor, back_buffer_modes,
-			       back_buffer_factor);
+			       back_buffer_factor, msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
 		__LINE__);
@@ -194,7 +209,7 @@ driCreateNewScreen(int scrn, const __DRIextension **extensions,
 {
     static const __DRIextension *emptyExtensionList[] = { NULL };
     __DRIscreen *psp;
-    __DRIconfig **configs8, **configs16, **configs32;
+    __DRIconfig **configs8, **configs16, **configs24, **configs32;
 
     (void) data;
 
@@ -211,11 +226,13 @@ driCreateNewScreen(int scrn, const __DRIextension **extensions,
 
     configs8  = swrastFillInModes(psp,  8,  8, 0, 1);
     configs16 = swrastFillInModes(psp, 16, 16, 0, 1);
+    configs24 = swrastFillInModes(psp, 24, 24, 8, 1);
     configs32 = swrastFillInModes(psp, 32, 24, 8, 1);
 
-    configs16 = (__DRIconfig **)driConcatConfigs(configs8, configs16);
-
-    *driver_configs = driConcatConfigs(configs16, configs32);
+    configs16 = driConcatConfigs(configs8, configs16);
+    configs24 = driConcatConfigs(configs16, configs24);
+    *driver_configs = (const __DRIconfig **)
+       driConcatConfigs(configs24, configs32);
 
     driInitExtensions( NULL, card_extensions, GL_FALSE );
 
@@ -247,19 +264,24 @@ static GLuint
 choose_pixel_format(const GLvisual *v)
 {
     if (v->rgbMode) {
-	int bpp = v->rgbBits;
+	int depth = v->rgbBits;
 
-	if (bpp == 32
+	if (depth == 32
 	    && v->redMask   == 0xff0000
 	    && v->greenMask == 0x00ff00
 	    && v->blueMask  == 0x0000ff)
 	    return PF_A8R8G8B8;
-	else if (bpp == 16
+	else if (depth == 24
+	    && v->redMask   == 0xff0000
+	    && v->greenMask == 0x00ff00
+	    && v->blueMask  == 0x0000ff)
+	    return PF_X8R8G8B8;
+	else if (depth == 16
 	    && v->redMask   == 0xf800
 	    && v->greenMask == 0x07e0
 	    && v->blueMask  == 0x001f)
 	    return PF_R5G6B5;
-	else if (bpp == 8
+	else if (depth == 8
 	    && v->redMask   == 0x07
 	    && v->greenMask == 0x38
 	    && v->blueMask  == 0xc0)
@@ -288,7 +310,6 @@ swrast_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb,
 			   GLenum internalFormat, GLuint width, GLuint height)
 {
     struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
-    int bpp;
     unsigned mask = PITCH_ALIGN_BITS - 1;
 
     TRACE;
@@ -297,23 +318,8 @@ swrast_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb,
     rb->Width = width;
     rb->Height = height;
 
-    switch (internalFormat) {
-    case GL_RGB:
-	bpp = rb->RedBits + rb->GreenBits + rb->BlueBits;
-	break;
-    case GL_RGBA:
-	bpp = rb->RedBits + rb->GreenBits + rb->BlueBits + rb->AlphaBits;
-	break;
-    case GL_COLOR_INDEX8_EXT:
-	bpp = rb->IndexBits;
-	break;
-    default:
-	_mesa_problem( NULL, "unexpected format in %s", __FUNCTION__ );
-	return GL_FALSE;
-    }
-
     /* always pad to PITCH_ALIGN_BITS */
-    xrb->pitch = ((width * bpp + mask) & ~mask) / 8;
+    xrb->pitch = ((width * xrb->bpp + mask) & ~mask) / 8;
 
     return GL_TRUE;
 }
@@ -369,6 +375,17 @@ swrast_new_renderbuffer(const GLvisual *visual, GLboolean front)
 	xrb->Base.GreenBits = 8 * sizeof(GLubyte);
 	xrb->Base.BlueBits  = 8 * sizeof(GLubyte);
 	xrb->Base.AlphaBits = 8 * sizeof(GLubyte);
+	xrb->bpp = 32;
+	break;
+    case PF_X8R8G8B8:
+	xrb->Base.InternalFormat = GL_RGB;
+	xrb->Base._BaseFormat = GL_RGB;
+	xrb->Base.DataType = GL_UNSIGNED_BYTE;
+	xrb->Base.RedBits   = 8 * sizeof(GLubyte);
+	xrb->Base.GreenBits = 8 * sizeof(GLubyte);
+	xrb->Base.BlueBits  = 8 * sizeof(GLubyte);
+	xrb->Base.AlphaBits = 0;
+	xrb->bpp = 32;
 	break;
     case PF_R5G6B5:
 	xrb->Base.InternalFormat = GL_RGB;
@@ -378,6 +395,7 @@ swrast_new_renderbuffer(const GLvisual *visual, GLboolean front)
 	xrb->Base.GreenBits = 6 * sizeof(GLubyte);
 	xrb->Base.BlueBits  = 5 * sizeof(GLubyte);
 	xrb->Base.AlphaBits = 0;
+	xrb->bpp = 16;
 	break;
     case PF_R3G3B2:
 	xrb->Base.InternalFormat = GL_RGB;
@@ -387,12 +405,14 @@ swrast_new_renderbuffer(const GLvisual *visual, GLboolean front)
 	xrb->Base.GreenBits = 3 * sizeof(GLubyte);
 	xrb->Base.BlueBits  = 2 * sizeof(GLubyte);
 	xrb->Base.AlphaBits = 0;
+	xrb->bpp = 8;
 	break;
     case PF_CI8:
 	xrb->Base.InternalFormat = GL_COLOR_INDEX8_EXT;
 	xrb->Base._BaseFormat = GL_COLOR_INDEX;
 	xrb->Base.DataType = GL_UNSIGNED_BYTE;
 	xrb->Base.IndexBits = 8 * sizeof(GLubyte);
+	xrb->bpp = 8;
 	break;
     default:
 	return NULL;
diff --git a/src/mesa/drivers/dri/swrast/swrast_priv.h b/src/mesa/drivers/dri/swrast/swrast_priv.h
index a707ffc40a..1a5fb31d5a 100644
--- a/src/mesa/drivers/dri/swrast/swrast_priv.h
+++ b/src/mesa/drivers/dri/swrast/swrast_priv.h
@@ -90,6 +90,8 @@ struct swrast_renderbuffer {
 
     /* renderbuffer pitch (in bytes) */
     GLuint pitch;
+   /* bits per pixel of storage */
+    GLuint bpp;
 };
 
 static INLINE __DRIcontext *
@@ -115,10 +117,10 @@ swrast_renderbuffer(struct gl_renderbuffer *rb)
  * Pixel formats we support
  */
 #define PF_CI8        1		/**< Color Index mode */
-#define PF_A8R8G8B8   2		/**< 32-bit TrueColor:  8-A, 8-R, 8-G, 8-B bits */
-#define PF_R5G6B5     3		/**< 16-bit TrueColor:  5-R, 6-G, 5-B bits */
-#define PF_R3G3B2     4		/**<  8-bit TrueColor:  3-R, 3-G, 2-B bits */
-
+#define PF_A8R8G8B8   2		/**< 32bpp TrueColor:  8-A, 8-R, 8-G, 8-B bits */
+#define PF_R5G6B5     3		/**< 16bpp TrueColor:  5-R, 6-G, 5-B bits */
+#define PF_R3G3B2     4		/**<  8bpp TrueColor:  3-R, 3-G, 2-B bits */
+#define PF_X8R8G8B8   5		/**< 32bpp TrueColor:  8-R, 8-G, 8-B bits */
 
 /**
  * Renderbuffer pitch alignment (in bits).
diff --git a/src/mesa/drivers/dri/swrast/swrast_span.c b/src/mesa/drivers/dri/swrast/swrast_span.c
index 5e990368b2..2d3c25dcbe 100644
--- a/src/mesa/drivers/dri/swrast/swrast_span.c
+++ b/src/mesa/drivers/dri/swrast/swrast_span.c
@@ -79,6 +79,24 @@ static const GLubyte kernel[16] = {
    DST[BCOMP] = SRC[0]
 
 
+/* 32-bit BGRX */
+#define STORE_PIXEL_X8R8G8B8(DST, X, Y, VALUE) \
+   DST[3] = 0xff; \
+   DST[2] = VALUE[RCOMP]; \
+   DST[1] = VALUE[GCOMP]; \
+   DST[0] = VALUE[BCOMP]
+#define STORE_PIXEL_RGB_X8R8G8B8(DST, X, Y, VALUE) \
+   DST[3] = 0xff; \
+   DST[2] = VALUE[RCOMP]; \
+   DST[1] = VALUE[GCOMP]; \
+   DST[0] = VALUE[BCOMP]
+#define FETCH_PIXEL_X8R8G8B8(DST, SRC) \
+   DST[ACOMP] = 0xff; \
+   DST[RCOMP] = SRC[2]; \
+   DST[GCOMP] = SRC[1]; \
+   DST[BCOMP] = SRC[0]
+
+
 /* 16-bit BGR */
 #define STORE_PIXEL_R5G6B5(DST, X, Y, VALUE) \
    do { \
@@ -139,6 +157,24 @@ static const GLubyte kernel[16] = {
 #include "swrast/s_spantemp.h"
 
 
+/* 32-bit BGRX */
+#define NAME(FUNC) FUNC##_X8R8G8B8
+#define RB_TYPE GLubyte
+#define SPAN_VARS \
+   struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
+#define INIT_PIXEL_PTR(P, X, Y) \
+   GLubyte *P = (GLubyte *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch + (X) * 4;
+#define INC_PIXEL_PTR(P) P += 4
+#define STORE_PIXEL(DST, X, Y, VALUE) \
+   STORE_PIXEL_X8R8G8B8(DST, X, Y, VALUE)
+#define STORE_PIXEL_RGB(DST, X, Y, VALUE) \
+   STORE_PIXEL_RGB_X8R8G8B8(DST, X, Y, VALUE)
+#define FETCH_PIXEL(DST, SRC) \
+   FETCH_PIXEL_X8R8G8B8(DST, SRC)
+
+#include "swrast/s_spantemp.h"
+
+
 /* 16-bit BGR */
 #define NAME(FUNC) FUNC##_R5G6B5
 #define RB_TYPE GLubyte
@@ -210,6 +246,24 @@ static const GLubyte kernel[16] = {
 #include "swrast_spantemp.h"
 
 
+/* 32-bit BGRX */
+#define NAME(FUNC) FUNC##_X8R8G8B8_front
+#define RB_TYPE GLubyte
+#define SPAN_VARS \
+   struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
+#define INIT_PIXEL_PTR(P, X, Y) \
+   GLubyte *P = (GLubyte *)row;
+#define INC_PIXEL_PTR(P) P += 4
+#define STORE_PIXEL(DST, X, Y, VALUE) \
+   STORE_PIXEL_X8R8G8B8(DST, X, Y, VALUE)
+#define STORE_PIXEL_RGB(DST, X, Y, VALUE) \
+   STORE_PIXEL_RGB_X8R8G8B8(DST, X, Y, VALUE)
+#define FETCH_PIXEL(DST, SRC) \
+   FETCH_PIXEL_X8R8G8B8(DST, SRC)
+
+#include "swrast_spantemp.h"
+
+
 /* 16-bit BGR */
 #define NAME(FUNC) FUNC##_R5G6B5_front
 #define RB_TYPE GLubyte
@@ -279,6 +333,15 @@ swrast_set_span_funcs_back(struct swrast_renderbuffer *xrb,
 	xrb->Base.PutValues = put_values_A8R8G8B8;
 	xrb->Base.PutMonoValues = put_mono_values_A8R8G8B8;
 	break;
+    case PF_X8R8G8B8:
+	xrb->Base.GetRow = get_row_X8R8G8B8;
+	xrb->Base.GetValues = get_values_X8R8G8B8;
+	xrb->Base.PutRow = put_row_X8R8G8B8;
+	xrb->Base.PutRowRGB = put_row_rgb_X8R8G8B8;
+	xrb->Base.PutMonoRow = put_mono_row_X8R8G8B8;
+	xrb->Base.PutValues = put_values_X8R8G8B8;
+	xrb->Base.PutMonoValues = put_mono_values_X8R8G8B8;
+	break;
     case PF_R5G6B5:
 	xrb->Base.GetRow = get_row_R5G6B5;
 	xrb->Base.GetValues = get_values_R5G6B5;
@@ -334,6 +397,15 @@ swrast_set_span_funcs_front(struct swrast_renderbuffer *xrb,
 	xrb->Base.PutValues = put_values_A8R8G8B8_front;
 	xrb->Base.PutMonoValues = put_mono_values_A8R8G8B8_front;
 	break;
+    case PF_X8R8G8B8:
+	xrb->Base.GetRow = get_row_X8R8G8B8_front;
+	xrb->Base.GetValues = get_values_X8R8G8B8_front;
+	xrb->Base.PutRow = put_row_X8R8G8B8_front;
+	xrb->Base.PutRowRGB = put_row_rgb_X8R8G8B8_front;
+	xrb->Base.PutMonoRow = put_mono_row_X8R8G8B8_front;
+	xrb->Base.PutValues = put_values_X8R8G8B8_front;
+	xrb->Base.PutMonoValues = put_mono_values_X8R8G8B8_front;
+	break;
     case PF_R5G6B5:
 	xrb->Base.GetRow = get_row_R5G6B5_front;
 	xrb->Base.GetValues = get_values_R5G6B5_front;
diff --git a/src/mesa/drivers/dri/tdfx/tdfx_context.c b/src/mesa/drivers/dri/tdfx/tdfx_context.c
index ef688d103d..20046fcb3a 100644
--- a/src/mesa/drivers/dri/tdfx/tdfx_context.c
+++ b/src/mesa/drivers/dri/tdfx/tdfx_context.c
@@ -58,20 +58,15 @@
 
 #include "utils.h"
 
-#define need_GL_ARB_multisample
 /* #define need_GL_ARB_point_parameters */
 #define need_GL_ARB_occlusion_query
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
 /* #define need_GL_ARB_vertex_program */
 #define need_GL_EXT_blend_equation_separate
 #define need_GL_EXT_blend_func_separate
 #define need_GL_EXT_blend_minmax
 #define need_GL_EXT_fog_coord
-#define need_GL_EXT_multi_draw_arrays
 #define need_GL_EXT_paletted_texture
 /* #define need_GL_EXT_secondary_color */
-#define need_GL_IBM_multimode_draw_arrays
 /* #define need_GL_MESA_program_debug */
 /* #define need_GL_NV_vertex_program */
 #include "extension_helper.h"
@@ -82,20 +77,16 @@
  */
 const struct dri_extension card_extensions[] =
 {
-    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
     { "GL_ARB_occlusion_query",            GL_ARB_occlusion_query_functions },
     { "GL_ARB_texture_mirrored_repeat",    NULL },
-    { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
 
     { "GL_EXT_blend_func_separate",        GL_EXT_blend_func_separate_functions },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
-    { "GL_EXT_multi_draw_arrays",          GL_EXT_multi_draw_arrays_functions },
     { "GL_EXT_paletted_texture",           GL_EXT_paletted_texture_functions },
     { "GL_EXT_shared_texture_palette",     NULL },
     { "GL_EXT_stencil_wrap",               NULL },
     { "GL_EXT_texture_env_add",            NULL },
     { "GL_EXT_texture_lod_bias",           NULL },
-    { "GL_IBM_multimode_draw_arrays",      GL_IBM_multimode_draw_arrays_functions },
 
 #ifdef need_GL_ARB_point_parameters
     { "GL_ARB_point_parameters",           GL_ARB_point_parameters_functions },
@@ -122,7 +113,6 @@ const struct dri_extension card_extensions[] =
  */
 const struct dri_extension napalm_extensions[] =
 {
-    { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
     { "GL_ARB_texture_env_combine",        NULL },
     { "GL_EXT_blend_equation_separate",    GL_EXT_blend_equation_separate_functions },
     { "GL_EXT_blend_subtract",             GL_EXT_blend_minmax_functions },
@@ -319,6 +309,8 @@ GLboolean tdfxCreateContext( const __GLcontextModes *mesaVis,
    ctx->Const.MaxLineWidthAA = 1.0;
    ctx->Const.LineWidthGranularity = 1.0;
 
+   ctx->Const.MaxDrawBuffers = 1;
+
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
diff --git a/src/mesa/drivers/dri/tdfx/tdfx_dd.c b/src/mesa/drivers/dri/tdfx/tdfx_dd.c
index 2cef079515..8472df607a 100644
--- a/src/mesa/drivers/dri/tdfx/tdfx_dd.c
+++ b/src/mesa/drivers/dri/tdfx/tdfx_dd.c
@@ -110,13 +110,13 @@ static const GLubyte *tdfxDDGetString( GLcontext *ctx, GLenum name )
 
 
 static void
-tdfxBeginQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q)
+tdfxBeginQuery(GLcontext *ctx, struct gl_query_object *q)
 {
    tdfxContextPtr fxMesa = TDFX_CONTEXT(ctx);
 
    (void) q;
 
-   if (target == GL_SAMPLES_PASSED_ARB) {
+   if (q->Target == GL_SAMPLES_PASSED_ARB) {
       LOCK_HARDWARE(fxMesa);
       fxMesa->Glide.grFinish();
       fxMesa->Glide.grReset(GR_STATS_PIXELS);
@@ -126,14 +126,14 @@ tdfxBeginQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q)
 
 
 static void
-tdfxEndQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q)
+tdfxEndQuery(GLcontext *ctx, struct gl_query_object *q)
 {
    tdfxContextPtr fxMesa = TDFX_CONTEXT(ctx);
    FxI32 total_pixels;
    FxI32 z_fail_pixels;
 
 
-   if (target == GL_SAMPLES_PASSED_ARB) {
+   if (q->Target == GL_SAMPLES_PASSED_ARB) {
       LOCK_HARDWARE(fxMesa);
       fxMesa->Glide.grFinish();
 
diff --git a/src/mesa/drivers/dri/tdfx/tdfx_screen.c b/src/mesa/drivers/dri/tdfx/tdfx_screen.c
index cd22b84951..5f2f5cfff5 100644
--- a/src/mesa/drivers/dri/tdfx/tdfx_screen.c
+++ b/src/mesa/drivers/dri/tdfx/tdfx_screen.c
@@ -361,6 +361,7 @@ tdfxFillInModes(__DRIscreenPrivate *psp,
 	static const GLenum db_modes[2] = { GLX_NONE, GLX_SWAP_UNDEFINED_OML };
 	uint8_t depth_bits_array[4];
 	uint8_t stencil_bits_array[4];
+        uint8_t msaa_samples_array[1];
 	if(deep) {
 		depth_bits_array[0] = 0;
 		depth_bits_array[1] = 24;
@@ -377,13 +378,17 @@ tdfxFillInModes(__DRIscreenPrivate *psp,
 		stencil_bits_array[3] = 8;
 	}
 
-	return driCreateConfigs(
-		deep ? GL_RGBA : GL_RGB,
-		deep ? GL_UNSIGNED_INT_8_8_8_8 : GL_UNSIGNED_SHORT_5_6_5,
-		depth_bits_array,
-		stencil_bits_array,
-		deep ? 2 : 4,
-		db_modes, 2);
+	msaa_samples_array[0] = 0;
+
+	return (const __DRIconfig **)
+	   driCreateConfigs(deep ? GL_RGBA : GL_RGB,
+			    deep ? GL_UNSIGNED_INT_8_8_8_8 :
+				   GL_UNSIGNED_SHORT_5_6_5,
+			    depth_bits_array,
+			    stencil_bits_array,
+			    deep ? 2 : 4,
+			    db_modes, 2,
+			    msaa_samples_array, 1);
 }
 
 /**
diff --git a/src/mesa/drivers/dri/trident/trident_context.c b/src/mesa/drivers/dri/trident/trident_context.c
index e134cfcf8e..aefd8a243f 100644
--- a/src/mesa/drivers/dri/trident/trident_context.c
+++ b/src/mesa/drivers/dri/trident/trident_context.c
@@ -128,6 +128,8 @@ tridentCreateContext( const __GLcontextModes *glVisual,
    ctx->Const.MaxPointSizeAA = 16.0; 
    ctx->Const.PointSizeGranularity = 0.25;
 
+   ctx->Const.MaxDrawBuffers = 1;
+
 #if 0
    tmesa->texHeap = mmInit( 0, tmesa->tridentScreen->textureSize );
 
diff --git a/src/mesa/drivers/dri/trident/trident_vb.c b/src/mesa/drivers/dri/trident/trident_vb.c
index b231f5ef15..055a914595 100644
--- a/src/mesa/drivers/dri/trident/trident_vb.c
+++ b/src/mesa/drivers/dri/trident/trident_vb.c
@@ -402,7 +402,7 @@ void tridentInitVB( GLcontext *ctx )
    tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx);
    GLuint size = TNL_CONTEXT(ctx)->vb.Size;
 
-   tmesa->verts = (char *)ALIGN_MALLOC( size * 16 * 4, 32 );
+   tmesa->verts = (GLubyte *)ALIGN_MALLOC( size * 16 * 4, 32 );
 
    {
       static int firsttime = 1;
diff --git a/src/mesa/drivers/dri/unichrome/via_context.c b/src/mesa/drivers/dri/unichrome/via_context.c
index f5bdb65eb0..6eb19ac079 100644
--- a/src/mesa/drivers/dri/unichrome/via_context.c
+++ b/src/mesa/drivers/dri/unichrome/via_context.c
@@ -62,9 +62,7 @@
 #include "main/macros.h"
 #include "drirenderbuffer.h"
 
-#define need_GL_ARB_multisample
 #define need_GL_ARB_point_parameters
-#define need_GL_ARB_vertex_buffer_object
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_secondary_color
 #include "extension_helper.h"
@@ -366,14 +364,12 @@ void viaReAllocateBuffers(GLcontext *ctx, GLframebuffer *drawbuffer,
  */
 const struct dri_extension card_extensions[] =
 {
-    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
     { "GL_ARB_multitexture",               NULL },
     { "GL_ARB_point_parameters",           GL_ARB_point_parameters_functions },
     { "GL_ARB_texture_env_add",            NULL },
     { "GL_ARB_texture_env_combine",        NULL },
 /*    { "GL_ARB_texture_env_dot3",           NULL }, */
     { "GL_ARB_texture_mirrored_repeat",    NULL },
-    { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
     { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
     { "GL_EXT_stencil_wrap",               NULL },
@@ -577,6 +573,8 @@ viaCreateContext(const __GLcontextModes *visual,
     ctx->Const.MaxPointSizeAA = 1.0;
     ctx->Const.PointSizeGranularity = 1.0;
 
+    ctx->Const.MaxDrawBuffers = 1;
+
     ctx->Driver.GetString = viaGetString;
 
     ctx->DriverCtx = (void *)vmesa;
diff --git a/src/mesa/drivers/dri/unichrome/via_screen.c b/src/mesa/drivers/dri/unichrome/via_screen.c
index 988f9935ac..e9f566161c 100644
--- a/src/mesa/drivers/dri/unichrome/via_screen.c
+++ b/src/mesa/drivers/dri/unichrome/via_screen.c
@@ -342,6 +342,7 @@ viaFillInModes( __DRIscreenPrivate *psp,
      */
     static const uint8_t depth_bits_array[4]   = { 0, 16, 24, 32 };
     static const uint8_t stencil_bits_array[4] = { 0,  0,  8,  0 };
+    uint8_t msaa_samples_array[1] = { 0 };
     const unsigned depth_buffer_factor = 3;
 
     if ( pixel_bits == 16 ) {
@@ -356,7 +357,8 @@ viaFillInModes( __DRIscreenPrivate *psp,
     configs = driCreateConfigs(fb_format, fb_type,
 			       depth_bits_array, stencil_bits_array,
 			       depth_buffer_factor, back_buffer_modes,
-			       back_buffer_factor);
+			       back_buffer_factor,
+                               msaa_samples_array, 1);
     if (configs == NULL) {
 	fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
 		__LINE__);
diff --git a/src/mesa/drivers/fbdev/glfbdev.c b/src/mesa/drivers/fbdev/glfbdev.c
index 3c874ba57a..531558dc4d 100644
--- a/src/mesa/drivers/fbdev/glfbdev.c
+++ b/src/mesa/drivers/fbdev/glfbdev.c
@@ -685,7 +685,7 @@ glFBDevDestroyBuffer( GLFBDevBufferPtr buffer )
       }
       {
          struct gl_framebuffer *fb = &buffer->glframebuffer;
-         _mesa_unreference_framebuffer(&fb);
+         _mesa_reference_framebuffer(&fb, NULL);
       }
    }
 }
diff --git a/src/mesa/drivers/glide/fxapi.c b/src/mesa/drivers/glide/fxapi.c
index b7708fd636..238f491599 100644
--- a/src/mesa/drivers/glide/fxapi.c
+++ b/src/mesa/drivers/glide/fxapi.c
@@ -728,7 +728,7 @@ errorhandler:
        FREE(fxMesa->fogTable);
     }
     if (fxMesa->glBuffer) {
-       _mesa_unreference_framebuffer(&fxMesa->glBuffer);
+       _mesa_reference_framebuffer(&fxMesa->glBuffer, NULL);
     }
     if (fxMesa->glVis) {
        _mesa_destroy_visual(fxMesa->glVis);
@@ -828,7 +828,7 @@ fxMesaDestroyContext(fxMesaContext fxMesa)
    fxDDDestroyFxMesaContext(fxMesa); /* must be before _mesa_destroy_context */
    _mesa_destroy_visual(fxMesa->glVis);
    _mesa_destroy_context(fxMesa->glCtx);
-   _mesa_unreference_framebuffer(&fxMesa->glBuffer);
+   _mesa_reference_framebuffer(&fxMesa->glBuffer, NULL);
    fxTMClose(fxMesa); /* must be after _mesa_destroy_context */
 
    FREE(fxMesa);
diff --git a/src/mesa/drivers/glide/fxdd.c b/src/mesa/drivers/glide/fxdd.c
index 213ef2382f..338cd37382 100644
--- a/src/mesa/drivers/glide/fxdd.c
+++ b/src/mesa/drivers/glide/fxdd.c
@@ -1773,6 +1773,8 @@ fxDDInitFxMesaContext(fxMesaContext fxMesa)
    ctx->Const.MaxTextureImageUnits = fxMesa->haveTwoTMUs ? 2 : 1;
    ctx->Const.MaxTextureUnits = MAX2(ctx->Const.MaxTextureImageUnits, ctx->Const.MaxTextureCoordUnits);
 
+   ctx->Const.MaxDrawBuffers = 1;
+
    fxMesa->new_state = _NEW_ALL;
    if (!fxMesa->haveHwStencil) {
       /* don't touch stencil if there is none */
@@ -1862,7 +1864,6 @@ fxDDInitExtensions(GLcontext * ctx)
    }
 
    if (fxMesa->type >= GR_SSTTYPE_Voodoo4) {
-      _mesa_enable_extension(ctx, "GL_ARB_texture_compression");
       _mesa_enable_extension(ctx, "GL_3DFX_texture_compression_FXT1");
       _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
       _mesa_enable_extension(ctx, "GL_S3_s3tc");
@@ -1894,7 +1895,6 @@ fxDDInitExtensions(GLcontext * ctx)
        * fxSetupSingleTMU_NoLock/fxSetupDoubleTMU_NoLock:
        *    grTexDownloadTable(GR_TEXTABLE_NCC0, &(ti->palette));
        */
-      /*_mesa_enable_extension(ctx, "GL_ARB_texture_compression");*/
       _mesa_enable_extension(ctx, "GL_SGIS_generate_mipmap");
    }
 
@@ -1917,9 +1917,6 @@ fxDDInitExtensions(GLcontext * ctx)
    }
 
    /* core-level extensions */
-   _mesa_enable_extension(ctx, "GL_EXT_multi_draw_arrays");
-   _mesa_enable_extension(ctx, "GL_IBM_multimode_draw_arrays");
-   _mesa_enable_extension(ctx, "GL_ARB_vertex_buffer_object");
    /* dangerous */
    if (getenv("MESA_FX_ALLOW_VP")) {
       _mesa_enable_extension(ctx, "GL_ARB_vertex_program");
diff --git a/src/mesa/drivers/glslcompiler/Makefile b/src/mesa/drivers/glslcompiler/Makefile
index dc4abd44d4..ffe4e0a234 100644
--- a/src/mesa/drivers/glslcompiler/Makefile
+++ b/src/mesa/drivers/glslcompiler/Makefile
@@ -33,7 +33,7 @@ default: $(PROGRAM)
 
 
 glslcompiler: $(OBJECTS)
-	$(CC) $(OBJECTS) -lm -lpthread -o $@
+	$(CC) $(OBJECTS) $(GL_LIB_DEPS) -o $@
 
 
 glslcompiler.o: glslcompiler.c
diff --git a/src/mesa/drivers/glslcompiler/glslcompiler.c b/src/mesa/drivers/glslcompiler/glslcompiler.c
index 34cce977c8..918ec41856 100644
--- a/src/mesa/drivers/glslcompiler/glslcompiler.c
+++ b/src/mesa/drivers/glslcompiler/glslcompiler.c
@@ -227,7 +227,7 @@ PrintShaderInstructions(GLuint shader, FILE *f)
    GET_CURRENT_CONTEXT(ctx);
    struct gl_shader *sh = _mesa_lookup_shader(ctx, shader);
    struct gl_program *prog = sh->Program;
-   _mesa_print_program_opt(prog, Options.Mode, Options.LineNumbers);
+   _mesa_fprint_program_opt(stdout, prog, Options.Mode, Options.LineNumbers);
 }
 
 
diff --git a/src/mesa/drivers/osmesa/Makefile b/src/mesa/drivers/osmesa/Makefile
index d6a18b1d75..3b3984200a 100644
--- a/src/mesa/drivers/osmesa/Makefile
+++ b/src/mesa/drivers/osmesa/Makefile
@@ -30,8 +30,7 @@ CORE_MESA = $(TOP)/src/mesa/libmesa.a $(TOP)/src/mesa/libglapi.a
 	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
 
 
-default:
-# $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME)
+default: $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME)
 	@ if [ "${DRIVER_DIRS}" = "osmesa" ] ; then \
 		$(MAKE) osmesa16 ; \
 	else \
@@ -42,9 +41,9 @@ default:
 
 
 # The normal libOSMesa is used in conjuction with libGL
-osmesa8: $(TOP)/lib/$(OSMESA_LIB_NAME)
+osmesa8: $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME)
 
-$(TOP)/lib/$(OSMESA_LIB_NAME): $(OBJECTS)
+$(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OBJECTS)
 	$(MKLIB) -o $(OSMESA_LIB) -linker '$(CC)' -ldflags '$(LDFLAGS)' \
 		-major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \
 		-install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \
diff --git a/src/mesa/drivers/osmesa/osmesa.c b/src/mesa/drivers/osmesa/osmesa.c
index f2367bbbb7..904659e345 100644
--- a/src/mesa/drivers/osmesa/osmesa.c
+++ b/src/mesa/drivers/osmesa/osmesa.c
@@ -1310,7 +1310,7 @@ OSMesaDestroyContext( OSMesaContext osmesa )
       _swrast_DestroyContext( &osmesa->mesa );
 
       _mesa_destroy_visual( osmesa->gl_visual );
-      _mesa_unreference_framebuffer( &osmesa->gl_buffer );
+      _mesa_reference_framebuffer( &osmesa->gl_buffer, NULL );
 
       _mesa_free_context_data( &osmesa->mesa );
       _mesa_free( osmesa );
diff --git a/src/mesa/drivers/windows/gdi/mesa.def b/src/mesa/drivers/windows/gdi/mesa.def
index 3f2d644e86..b386e34aad 100644
--- a/src/mesa/drivers/windows/gdi/mesa.def
+++ b/src/mesa/drivers/windows/gdi/mesa.def
@@ -867,6 +867,7 @@ EXPORTS
 	_glapi_get_proc_address
 	_mesa_add_soft_renderbuffers
 	_mesa_add_renderbuffer
+	_mesa_begin_query
 	_mesa_buffer_data
 	_mesa_buffer_get_subdata
 	_mesa_buffer_map
@@ -881,6 +882,7 @@ EXPORTS
 	_mesa_delete_array_object
 	_mesa_delete_buffer_object
 	_mesa_delete_program
+	_mesa_delete_query
 	_mesa_delete_texture_object
 	_mesa_destroy_framebuffer
 	_mesa_destroy_visual
@@ -890,6 +892,7 @@ EXPORTS
 	_mesa_enable_2_0_extensions
 	_mesa_enable_2_1_extensions
 	_mesa_enable_sw_extensions
+	_mesa_end_query
 	_mesa_error
 	_mesa_finish_render_texture
 	_mesa_framebuffer_renderbuffer
@@ -941,6 +944,7 @@ EXPORTS
 	_mesa_update_framebuffer_visual
 	_mesa_use_program
 	_mesa_Viewport
+	_mesa_wait_query
 	_swrast_Accum
 	_swrast_Bitmap
 	_swrast_BlitFramebuffer
@@ -973,3 +977,4 @@ EXPORTS
 	_tnl_InvalidateState
 	_tnl_run_pipeline
 	_tnl_program_string
+	_tnl_RasterPos
\ No newline at end of file
diff --git a/src/mesa/drivers/windows/gdi/wgl.c b/src/mesa/drivers/windows/gdi/wgl.c
index 9f0bb9122a..8d8087067f 100644
--- a/src/mesa/drivers/windows/gdi/wgl.c
+++ b/src/mesa/drivers/windows/gdi/wgl.c
@@ -601,8 +601,9 @@ WINGDIAPI BOOL GLAPIENTRY wglCopyContext(HGLRC hglrcSrc,
 WINGDIAPI HGLRC GLAPIENTRY wglCreateLayerContext(HDC hdc,
 						 int iLayerPlane)
 {
-    (void) hdc; (void) iLayerPlane;
     SetLastError(0);
+    if (iLayerPlane == 0)
+        return wglCreateContext( hdc );
     return(NULL);
 }
 
diff --git a/src/mesa/drivers/windows/gdi/wmesa.c b/src/mesa/drivers/windows/gdi/wmesa.c
index 9dafb74723..80746950c7 100644
--- a/src/mesa/drivers/windows/gdi/wmesa.c
+++ b/src/mesa/drivers/windows/gdi/wmesa.c
@@ -56,11 +56,13 @@ wmesa_free_framebuffer(HDC hdc)
 	prev = pwfb;
     }
     if (pwfb) {
+        struct gl_framebuffer *fb;
 	if (pwfb == FirstFramebuffer)
 	    FirstFramebuffer = pwfb->next;
 	else
 	    prev->next = pwfb->next;
-	free(pwfb);
+        fb = &pwfb->Base;
+        _mesa_unreference_framebuffer(&fb); 
     }
 }
 
@@ -1503,6 +1505,9 @@ WMesaContext WMesaCreateContext(HDC hDC,
     ctx = &c->gl_ctx;
     _mesa_initialize_context(ctx, visual, NULL, &functions, (void *)c);
 
+    /* visual no longer needed - it was copied by _mesa_initialize_context() */
+    _mesa_destroy_visual(visual);
+
     _mesa_enable_sw_extensions(ctx);
     _mesa_enable_1_3_extensions(ctx);
     _mesa_enable_1_4_extensions(ctx);
diff --git a/src/mesa/drivers/windows/gldirect/dglcontext.c b/src/mesa/drivers/windows/gldirect/dglcontext.c
index 4ad7a76e67..e9c23d1ccb 100644
--- a/src/mesa/drivers/windows/gldirect/dglcontext.c
+++ b/src/mesa/drivers/windows/gldirect/dglcontext.c
@@ -1482,6 +1482,7 @@ SkipPrimaryCreate:
 #else
 	lpCtx->glCtx->Const.MaxTextureSize = 1024;
 #endif
+	lpCtx->glCtx->Const.MaxDrawBuffers = 1;
 
 	// Setup the Display Driver pointers
 	dglSetupDDPointers(lpCtx->glCtx);
diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_wgl_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_wgl_dx7.c
index 0f8fe33eb1..fa44a952a0 100644
--- a/src/mesa/drivers/windows/gldirect/dx7/gld_wgl_dx7.c
+++ b/src/mesa/drivers/windows/gldirect/dx7/gld_wgl_dx7.c
@@ -1422,6 +1422,8 @@ BOOL gldInitialiseMesa_DX(
 		lpCtx->glCtx->Const.MaxTextureUnits = 1;
 	}
 
+	lpCtx->glCtx->Const.MaxDrawBuffers = 1;
+
 	// max texture size
 //	MaxTextureSize = min(gld->d3dCaps8.MaxTextureHeight, gld->d3dCaps8.MaxTextureWidth);
 	MaxTextureSize = min(gld->d3dCaps.dwMaxTextureHeight, gld->d3dCaps.dwMaxTextureWidth);
diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_wgl_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_wgl_dx8.c
index 690f68b68f..011d810e97 100644
--- a/src/mesa/drivers/windows/gldirect/dx8/gld_wgl_dx8.c
+++ b/src/mesa/drivers/windows/gldirect/dx8/gld_wgl_dx8.c
@@ -1204,6 +1204,7 @@ BOOL gldInitialiseMesa_DX(
 		MaxTextureSize >>= 1;
 	}
 	lpCtx->glCtx->Const.MaxTextureLevels = (TextureLevels) ? TextureLevels : 8;
+	lpCtx->glCtx->Const.MaxDrawBuffers = 1;
 
 	IDirect3DDevice8_SetRenderState(gld->pDev, D3DRS_LIGHTING, FALSE);
 	IDirect3DDevice8_SetRenderState(gld->pDev, D3DRS_CULLMODE, D3DCULL_NONE);
diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_wgl_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_wgl_dx9.c
index dc465c5418..a03b865bb4 100644
--- a/src/mesa/drivers/windows/gldirect/dx9/gld_wgl_dx9.c
+++ b/src/mesa/drivers/windows/gldirect/dx9/gld_wgl_dx9.c
@@ -1206,6 +1206,7 @@ BOOL gldInitialiseMesa_DX(
 		MaxTextureSize >>= 1;
 	}
 	lpCtx->glCtx->Const.MaxTextureLevels = (TextureLevels) ? TextureLevels : 8;
+	lpCtx->glCtx->Const.MaxDrawBuffers = 1;
 
 	IDirect3DDevice9_SetRenderState(gld->pDev, D3DRS_LIGHTING, FALSE);
 	IDirect3DDevice9_SetRenderState(gld->pDev, D3DRS_CULLMODE, D3DCULL_NONE);
diff --git a/src/mesa/drivers/windows/gldirect/mesasw/gld_wgl_mesasw.c b/src/mesa/drivers/windows/gldirect/mesasw/gld_wgl_mesasw.c
index b590dc795a..342a742867 100644
--- a/src/mesa/drivers/windows/gldirect/mesasw/gld_wgl_mesasw.c
+++ b/src/mesa/drivers/windows/gldirect/mesasw/gld_wgl_mesasw.c
@@ -1625,6 +1625,7 @@ BOOL gldInitialiseMesa_MesaSW(
 
 	// Added this to force max texture diminsion to 256. KeithH
 	ctx->Const.MaxTextureLevels = 8;
+	ctx->Const.MaxDrawBuffers = 1;
 
 	_mesa_enable_sw_extensions(ctx);
 	_mesa_enable_imaging_extensions(ctx);
diff --git a/src/mesa/drivers/x11/fakeglx.c b/src/mesa/drivers/x11/fakeglx.c
index 827d39f995..3b004a3ee2 100644
--- a/src/mesa/drivers/x11/fakeglx.c
+++ b/src/mesa/drivers/x11/fakeglx.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.5
  *
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -42,7 +43,6 @@
 
 #include "glxheader.h"
 #include "glxapi.h"
-#include "GL/xmesa.h"
 #include "main/context.h"
 #include "main/config.h"
 #include "main/macros.h"
@@ -1393,6 +1393,25 @@ Fake_glXChooseVisual( Display *dpy, int screen, int *list )
 }
 
 
+/**
+ * Init basic fields of a new fake_glx_context.
+ */
+static void
+init_glx_context(struct fake_glx_context *glxCtx, Display *dpy)
+{
+   /* Always return True.  See if anyone's confused... */
+   GLboolean direct = GL_TRUE;
+
+   glxCtx->xmesaContext->direct = direct;
+   glxCtx->glxContext.isDirect = direct;
+   glxCtx->glxContext.currentDpy = dpy;
+   glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
+
+   assert((void *) glxCtx == (void *) &(glxCtx->glxContext));
+}
+
+
+
 static GLXContext
 Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo,
                        GLXContext share_list, Bool direct )
@@ -1431,12 +1450,7 @@ Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo,
       return NULL;
    }
 
-   glxCtx->xmesaContext->direct = GL_FALSE;
-   glxCtx->glxContext.isDirect = GL_FALSE;
-   glxCtx->glxContext.currentDpy = dpy;
-   glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
-
-   assert((void *) glxCtx == (void *) &(glxCtx->glxContext));
+   init_glx_context(glxCtx, dpy);
 
    return (GLXContext) glxCtx;
 }
@@ -2442,12 +2456,7 @@ Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config,
       return NULL;
    }
 
-   glxCtx->xmesaContext->direct = GL_FALSE;
-   glxCtx->glxContext.isDirect = GL_FALSE;
-   glxCtx->glxContext.currentDpy = dpy;
-   glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
-
-   assert((void *) glxCtx == (void *) &(glxCtx->glxContext));
+   init_glx_context(glxCtx, dpy);
 
    return (GLXContext) glxCtx;
 }
@@ -2665,12 +2674,7 @@ Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int re
       return NULL;
    }
 
-   glxCtx->xmesaContext->direct = GL_FALSE;
-   glxCtx->glxContext.isDirect = GL_FALSE;
-   glxCtx->glxContext.currentDpy = dpy;
-   glxCtx->glxContext.xid = (XID) glxCtx;  /* self pointer */
-
-   assert((void *) glxCtx == (void *) &(glxCtx->glxContext));
+   init_glx_context(glxCtx, dpy);
 
    return (GLXContext) glxCtx;
 }
diff --git a/src/mesa/drivers/x11/fxmesa.h b/src/mesa/drivers/x11/fxmesa.h
new file mode 100644
index 0000000000..f8e9661f9c
--- /dev/null
+++ b/src/mesa/drivers/x11/fxmesa.h
@@ -0,0 +1,103 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  4.0
+ * Copyright (C) 1995-2001  Brian Paul
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Library General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Library General Public License for more details.
+ *
+ * You should have received a copy of the GNU Library General Public
+ * License along with this library; if not, write to the Free
+ * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+
+/*
+ * FXMesa - 3Dfx Glide driver for Mesa.  Contributed by David Bucciarelli
+ *
+ * NOTE: This version requires Glide3 (http://sourceforge.net/projects/glide)
+ */
+
+
+#ifndef FXMESA_H
+#define FXMESA_H
+
+
+#include <glide.h>
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+#define FXMESA_MAJOR_VERSION 6
+#define FXMESA_MINOR_VERSION 3
+
+
+/*
+ * Values for attribList parameter to fxMesaCreateContext():
+ */
+#define FXMESA_NONE		0	/* to terminate attribList */
+#define FXMESA_DOUBLEBUFFER	10
+#define FXMESA_ALPHA_SIZE	11      /* followed by an integer */
+#define FXMESA_DEPTH_SIZE	12      /* followed by an integer */
+#define FXMESA_STENCIL_SIZE	13      /* followed by an integer */
+#define FXMESA_ACCUM_SIZE	14      /* followed by an integer */
+#define FXMESA_COLORDEPTH	20      /* followed by an integer */
+#define FXMESA_SHARE_CONTEXT 990099	/* keep in sync with xmesa1.c! */
+
+
+
+typedef struct tfxMesaContext *fxMesaContext;
+
+
+#if defined (__BEOS__)
+#pragma export on
+#endif
+
+
+GLAPI fxMesaContext GLAPIENTRY fxMesaCreateContext(GLuint win, GrScreenResolution_t,
+						  GrScreenRefresh_t,
+						  const GLint attribList[]);
+
+GLAPI fxMesaContext GLAPIENTRY fxMesaCreateBestContext(GLuint win,
+						      GLint width, GLint height,
+						      const GLint attribList[]);
+GLAPI void GLAPIENTRY fxMesaDestroyContext(fxMesaContext ctx);
+
+GLAPI GLint GLAPIENTRY fxMesaSelectCurrentBoard(int n);
+
+GLAPI void GLAPIENTRY fxMesaMakeCurrent(fxMesaContext ctx);
+
+GLAPI fxMesaContext GLAPIENTRY fxMesaGetCurrentContext(void);
+
+GLAPI void GLAPIENTRY fxMesaSwapBuffers(void);
+
+GLAPI void GLAPIENTRY fxMesaSetNearFar(GLfloat nearVal, GLfloat farVal);
+
+GLAPI void GLAPIENTRY fxMesaUpdateScreenSize(fxMesaContext ctx);
+
+GLAPI void GLAPIENTRY fxCloseHardware(void);
+
+GLAPI void GLAPIENTRY fxGetScreenGeometry (GLint *w, GLint *h);
+
+
+#if defined (__BEOS__)
+#pragma export off
+#endif
+
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif
diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c
index c9009bad03..2a8c3a4fd3 100644
--- a/src/mesa/drivers/x11/xm_api.c
+++ b/src/mesa/drivers/x11/xm_api.c
@@ -63,7 +63,6 @@
 #endif
 
 #include "glxheader.h"
-#include "GL/xmesa.h"
 #include "xmesaP.h"
 #include "main/context.h"
 #include "main/extensions.h"
@@ -493,7 +492,7 @@ xmesa_free_buffer(XMesaBuffer buffer)
          b->frontxrb->drawable = 0;
 
          /* Unreference.  If count = zero we'll really delete the buffer */
-         _mesa_unreference_framebuffer(&fb);
+         _mesa_reference_framebuffer(&fb, NULL);
 
          return;
       }
@@ -1326,7 +1325,6 @@ xmesa_convert_from_x_visual_type( int visualType )
 #define need_GL_EXT_framebuffer_blit
 #define need_GL_EXT_gpu_program_parameters
 #define need_GL_EXT_paletted_texture
-#define need_GL_IBM_multimode_draw_arrays
 #define need_GL_MESA_resize_buffers
 #define need_GL_NV_vertex_program
 #define need_GL_NV_fragment_program
@@ -1356,7 +1354,6 @@ const struct dri_extension card_extensions[] =
    { "GL_EXT_framebuffer_blit",		GL_EXT_framebuffer_blit_functions },
    { "GL_EXT_gpu_program_parameters",	GL_EXT_gpu_program_parameters_functions },
    { "GL_EXT_paletted_texture",		GL_EXT_paletted_texture_functions },
-   { "GL_IBM_multimode_draw_arrays",	GL_IBM_multimode_draw_arrays_functions },
    { "GL_MESA_resize_buffers",		GL_MESA_resize_buffers_functions },
    { "GL_NV_vertex_program",		GL_NV_vertex_program_functions },
    { "GL_NV_fragment_program",		GL_NV_fragment_program_functions },
diff --git a/src/mesa/drivers/x11/xm_buffer.c b/src/mesa/drivers/x11/xm_buffer.c
index f104d44d05..7ad67bc34d 100644
--- a/src/mesa/drivers/x11/xm_buffer.c
+++ b/src/mesa/drivers/x11/xm_buffer.c
@@ -30,7 +30,6 @@
 
 
 #include "glxheader.h"
-#include "GL/xmesa.h"
 #include "xmesaP.h"
 #include "main/imports.h"
 #include "main/framebuffer.h"
diff --git a/src/mesa/drivers/x11/xmesa.h b/src/mesa/drivers/x11/xmesa.h
new file mode 100644
index 0000000000..98139af833
--- /dev/null
+++ b/src/mesa/drivers/x11/xmesa.h
@@ -0,0 +1,424 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ * 
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ * 
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ * Mesa/X11 interface.  This header file serves as the documentation for
+ * the Mesa/X11 interface functions.
+ *
+ * Note: this interface isn't intended for user programs.  It's primarily
+ * just for implementing the pseudo-GLX interface.
+ */
+
+
+/* Sample Usage:
+
+In addition to the usual X calls to select a visual, create a colormap
+and create a window, you must do the following to use the X/Mesa interface:
+
+1. Call XMesaCreateVisual() to make an XMesaVisual from an XVisualInfo.
+
+2. Call XMesaCreateContext() to create an X/Mesa rendering context, given
+   the XMesaVisual.
+
+3. Call XMesaCreateWindowBuffer() to create an XMesaBuffer from an X window
+   and XMesaVisual.
+
+4. Call XMesaMakeCurrent() to bind the XMesaBuffer to an XMesaContext and
+   to make the context the current one.
+
+5. Make gl* calls to render your graphics.
+
+6. Use XMesaSwapBuffers() when double buffering to swap front/back buffers.
+
+7. Before the X window is destroyed, call XMesaDestroyBuffer().
+
+8. Before exiting, call XMesaDestroyVisual and XMesaDestroyContext.
+
+*/
+
+
+
+
+#ifndef XMESA_H
+#define XMESA_H
+
+#ifdef __VMS
+#include <GL/vms_x_fix.h>
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef XFree86Server
+#include "xmesa_xf86.h"
+#else
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+#include "xmesa_x.h"
+#endif
+#include "GL/gl.h"
+
+#ifdef AMIWIN
+#include <pragmas/xlib_pragmas.h>
+extern struct Library *XLibBase;
+#endif
+
+
+#define XMESA_MAJOR_VERSION 6
+#define XMESA_MINOR_VERSION 3
+
+
+
+/*
+ * Values passed to XMesaGetString:
+ */
+#define XMESA_VERSION 1
+#define XMESA_EXTENSIONS 2
+
+
+/*
+ * Values passed to XMesaSetFXmode:
+ */
+#define XMESA_FX_WINDOW       1
+#define XMESA_FX_FULLSCREEN   2
+
+
+
+typedef struct xmesa_context *XMesaContext;
+
+typedef struct xmesa_visual *XMesaVisual;
+
+typedef struct xmesa_buffer *XMesaBuffer;
+
+
+
+/*
+ * Create a new X/Mesa visual.
+ * Input:  display - X11 display
+ *         visinfo - an XVisualInfo pointer
+ *         rgb_flag - GL_TRUE = RGB mode,
+ *                    GL_FALSE = color index mode
+ *         alpha_flag - alpha buffer requested?
+ *         db_flag - GL_TRUE = double-buffered,
+ *                   GL_FALSE = single buffered
+ *         stereo_flag - stereo visual?
+ *         ximage_flag - GL_TRUE = use an XImage for back buffer,
+ *                       GL_FALSE = use an off-screen pixmap for back buffer
+ *         depth_size - requested bits/depth values, or zero
+ *         stencil_size - requested bits/stencil values, or zero
+ *         accum_red_size - requested bits/red accum values, or zero
+ *         accum_green_size - requested bits/green accum values, or zero
+ *         accum_blue_size - requested bits/blue accum values, or zero
+ *         accum_alpha_size - requested bits/alpha accum values, or zero
+ *         num_samples - number of samples/pixel if multisampling, or zero
+ *         level - visual level, usually 0
+ *         visualCaveat - ala the GLX extension, usually GLX_NONE_EXT
+ * Return;  a new XMesaVisual or 0 if error.
+ */
+extern XMesaVisual XMesaCreateVisual( XMesaDisplay *display,
+                                      XMesaVisualInfo visinfo,
+                                      GLboolean rgb_flag,
+                                      GLboolean alpha_flag,
+                                      GLboolean db_flag,
+                                      GLboolean stereo_flag,
+                                      GLboolean ximage_flag,
+                                      GLint depth_size,
+                                      GLint stencil_size,
+                                      GLint accum_red_size,
+                                      GLint accum_green_size,
+                                      GLint accum_blue_size,
+                                      GLint accum_alpha_size,
+                                      GLint num_samples,
+                                      GLint level,
+                                      GLint visualCaveat );
+
+/*
+ * Destroy an XMesaVisual, but not the associated XVisualInfo.
+ */
+extern void XMesaDestroyVisual( XMesaVisual v );
+
+
+
+/*
+ * Create a new XMesaContext for rendering into an X11 window.
+ *
+ * Input:  visual - an XMesaVisual
+ *         share_list - another XMesaContext with which to share display
+ *                      lists or NULL if no sharing is wanted.
+ * Return:  an XMesaContext or NULL if error.
+ */
+extern XMesaContext XMesaCreateContext( XMesaVisual v,
+					XMesaContext share_list );
+
+
+/*
+ * Destroy a rendering context as returned by XMesaCreateContext()
+ */
+extern void XMesaDestroyContext( XMesaContext c );
+
+
+#ifdef XFree86Server
+/*
+ * These are the extra routines required for integration with XFree86.
+ * None of these routines should be user visible. -KEM
+ */
+extern GLboolean XMesaForceCurrent( XMesaContext c );
+
+extern GLboolean XMesaLoseCurrent( XMesaContext c );
+
+extern GLboolean XMesaCopyContext( XMesaContext src,
+				   XMesaContext dst,
+				   GLuint mask );
+#endif /* XFree86Server */
+
+
+/*
+ * Create an XMesaBuffer from an X window.
+ */
+extern XMesaBuffer XMesaCreateWindowBuffer( XMesaVisual v, XMesaWindow w );
+
+
+/*
+ * Create an XMesaBuffer from an X pixmap.
+ */
+extern XMesaBuffer XMesaCreatePixmapBuffer( XMesaVisual v,
+					    XMesaPixmap p,
+					    XMesaColormap cmap );
+
+
+/*
+ * Destroy an XMesaBuffer, but not the corresponding window or pixmap.
+ */
+extern void XMesaDestroyBuffer( XMesaBuffer b );
+
+
+/*
+ * Return the XMesaBuffer handle which corresponds to an X drawable, if any.
+ *
+ * New in Mesa 2.3.
+ */
+extern XMesaBuffer XMesaFindBuffer( XMesaDisplay *dpy,
+				    XMesaDrawable d );
+
+
+
+/*
+ * Bind a buffer to a context and make the context the current one.
+ */
+extern GLboolean XMesaMakeCurrent( XMesaContext c,
+				   XMesaBuffer b );
+
+
+/*
+ * Bind two buffers (read and draw) to a context and make the
+ * context the current one.
+ * New in Mesa 3.3
+ */
+extern GLboolean XMesaMakeCurrent2( XMesaContext c,
+                                    XMesaBuffer drawBuffer,
+                                    XMesaBuffer readBuffer );
+
+
+/*
+ * Unbind the current context from its buffer.
+ */
+extern GLboolean XMesaUnbindContext( XMesaContext c );
+
+
+/*
+ * Return a handle to the current context.
+ */
+extern XMesaContext XMesaGetCurrentContext( void );
+
+
+/*
+ * Return handle to the current (draw) buffer.
+ */
+extern XMesaBuffer XMesaGetCurrentBuffer( void );
+
+
+/*
+ * Return handle to the current read buffer.
+ * New in Mesa 3.3
+ */
+extern XMesaBuffer XMesaGetCurrentReadBuffer( void );
+
+
+/*
+ * Swap the front and back buffers for the given buffer.  No action is
+ * taken if the buffer is not double buffered.
+ */
+extern void XMesaSwapBuffers( XMesaBuffer b );
+
+
+/*
+ * Copy a sub-region of the back buffer to the front buffer.
+ *
+ * New in Mesa 2.6
+ */
+extern void XMesaCopySubBuffer( XMesaBuffer b,
+				int x,
+				int y,
+				int width,
+				int height );
+
+
+/*
+ * Return a pointer to the the Pixmap or XImage being used as the back
+ * color buffer of an XMesaBuffer.  This function is a way to get "under
+ * the hood" of X/Mesa so one can manipulate the back buffer directly.
+ * Input:  b - the XMesaBuffer
+ * Output:  pixmap - pointer to back buffer's Pixmap, or 0
+ *          ximage - pointer to back buffer's XImage, or NULL
+ * Return:  GL_TRUE = context is double buffered
+ *          GL_FALSE = context is single buffered
+ */
+extern GLboolean XMesaGetBackBuffer( XMesaBuffer b,
+				     XMesaPixmap *pixmap,
+				     XMesaImage **ximage );
+
+
+
+/*
+ * Return the depth buffer associated with an XMesaBuffer.
+ * Input:  b - the XMesa buffer handle
+ * Output:  width, height - size of buffer in pixels
+ *          bytesPerValue - bytes per depth value (2 or 4)
+ *          buffer - pointer to depth buffer values
+ * Return:  GL_TRUE or GL_FALSE to indicate success or failure.
+ *
+ * New in Mesa 2.4.
+ */
+extern GLboolean XMesaGetDepthBuffer( XMesaBuffer b,
+				      GLint *width,
+				      GLint *height,
+				      GLint *bytesPerValue,
+				      void **buffer );
+
+
+
+/*
+ * Flush/sync a context
+ */
+extern void XMesaFlush( XMesaContext c );
+
+
+
+/*
+ * Get an X/Mesa-specific string.
+ * Input:  name - either XMESA_VERSION or XMESA_EXTENSIONS
+ */
+extern const char *XMesaGetString( XMesaContext c, int name );
+
+
+
+/*
+ * Scan for XMesaBuffers whose window/pixmap has been destroyed, then free
+ * any memory used by that buffer.
+ *
+ * New in Mesa 2.3.
+ */
+extern void XMesaGarbageCollect( void );
+
+
+
+/*
+ * Return a dithered pixel value.
+ * Input:  c - XMesaContext
+ *         x, y - window coordinate
+ *         red, green, blue, alpha - color components in [0,1]
+ * Return:  pixel value
+ *
+ * New in Mesa 2.3.
+ */
+extern unsigned long XMesaDitherColor( XMesaContext xmesa,
+				       GLint x,
+				       GLint y,
+				       GLfloat red,
+				       GLfloat green,
+				       GLfloat blue,
+				       GLfloat alpha );
+
+
+
+/*
+ * 3Dfx Glide driver only!
+ * Set 3Dfx/Glide full-screen or window rendering mode.
+ * Input:  mode - either XMESA_FX_WINDOW (window rendering mode) or
+ *                XMESA_FX_FULLSCREEN (full-screen rendering mode)
+ * Return:  GL_TRUE if success
+ *          GL_FALSE if invalid mode or if not using 3Dfx driver
+ *
+ * New in Mesa 2.6.
+ */
+extern GLboolean XMesaSetFXmode( GLint mode );
+
+
+
+/*
+ * Reallocate the back/depth/stencil/accum/etc/ buffers associated with
+ * buffer <b> if its size has changed.
+ *
+ * New in Mesa 4.0.2
+ */
+extern void XMesaResizeBuffers( XMesaBuffer b );
+
+
+
+/*
+ * Create a pbuffer.
+ * New in Mesa 4.1
+ */
+extern XMesaBuffer XMesaCreatePBuffer(XMesaVisual v, XMesaColormap cmap,
+                                      unsigned int width, unsigned int height);
+
+
+
+/*
+ * Texture from Pixmap
+ * New in Mesa 7.1
+ */
+extern void
+XMesaBindTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer,
+                  const int *attrib_list);
+
+extern void
+XMesaReleaseTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer);
+
+
+extern XMesaBuffer
+XMesaCreatePixmapTextureBuffer(XMesaVisual v, XMesaPixmap p,
+                               XMesaColormap cmap,
+                               int format, int target, int mipmap);
+
+
+
+#ifdef __cplusplus
+}
+#endif
+
+
+#endif
diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h
index 98867ac710..65e747d7b9 100644
--- a/src/mesa/drivers/x11/xmesaP.h
+++ b/src/mesa/drivers/x11/xmesaP.h
@@ -27,10 +27,10 @@
 #define XMESAP_H
 
 
-#include "GL/xmesa.h"
+#include "xmesa.h"
 #include "main/mtypes.h"
 #if defined(FX)
-#include "GL/fxmesa.h"
+#include "fxmesa.h"
 #include "xm_glide.h"
 #endif
 #ifdef XFree86Server
diff --git a/src/mesa/drivers/x11/xmesa_x.h b/src/mesa/drivers/x11/xmesa_x.h
new file mode 100644
index 0000000000..865bab4313
--- /dev/null
+++ b/src/mesa/drivers/x11/xmesa_x.h
@@ -0,0 +1,86 @@
+
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ * When we're building the XMesa driver for stand-alone Mesa we
+ * include this file when building the xm_*.c files.
+ * We need to define some types and macros differently when building
+ * in the Xserver vs. stand-alone Mesa.
+ */
+
+#ifndef _XMESA_X_H_
+#define _XMESA_X_H_
+
+typedef Display      XMesaDisplay;
+typedef Pixmap       XMesaPixmap;
+typedef Colormap     XMesaColormap;
+typedef Drawable     XMesaDrawable;
+typedef Window       XMesaWindow;
+typedef GC           XMesaGC;
+typedef XVisualInfo *XMesaVisualInfo;
+typedef XImage       XMesaImage;
+typedef XPoint       XMesaPoint;
+typedef XColor       XMesaColor;
+
+#define XMesaDestroyImage      XDestroyImage
+
+#define XMesaPutPixel          XPutPixel
+#define XMesaGetPixel          XGetPixel
+
+#define XMesaSetForeground     XSetForeground
+#define XMesaSetBackground     XSetBackground
+#define XMesaSetPlaneMask      XSetPlaneMask
+#define XMesaSetFunction       XSetFunction
+#define XMesaSetFillStyle      XSetFillStyle
+#define XMesaSetTile           XSetTile
+
+#define XMesaDrawPoint         XDrawPoint
+#define XMesaDrawPoints        XDrawPoints
+#define XMesaDrawLine          XDrawLine
+#define XMesaFillRectangle     XFillRectangle
+#define XMesaGetImage          XGetImage
+#define XMesaPutImage          XPutImage
+#define XMesaCopyArea          XCopyArea
+
+#define XMesaCreatePixmap      XCreatePixmap
+#define XMesaFreePixmap        XFreePixmap
+#define XMesaFreeGC            XFreeGC
+
+#define GET_COLORMAP_SIZE(__v)  __v->visinfo->colormap_size
+#define GET_REDMASK(__v)        __v->mesa_visual.redMask
+#define GET_GREENMASK(__v)      __v->mesa_visual.greenMask
+#define GET_BLUEMASK(__v)       __v->mesa_visual.blueMask
+#define GET_VISUAL_DEPTH(__v)   __v->visinfo->depth
+#define GET_BLACK_PIXEL(__v)    BlackPixel(__v->display, __v->mesa_visual.screen)
+#define CHECK_BYTE_ORDER(__v)   host_byte_order()==ImageByteOrder(__v->display)
+#define CHECK_FOR_HPCR(__v)     XInternAtom(__v->display, "_HP_RGB_SMOOTH_MAP_LIST", True)
+
+#endif
diff --git a/src/mesa/drivers/x11/xmesa_xf86.h b/src/mesa/drivers/x11/xmesa_xf86.h
new file mode 100644
index 0000000000..10f93c3ab6
--- /dev/null
+++ b/src/mesa/drivers/x11/xmesa_xf86.h
@@ -0,0 +1,198 @@
+
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ * When we're building the XMesa driver for use in the X server (as the
+ * indirect render) we include this file when building the xm_*.c files.
+ * We need to define some types and macros differently when building
+ * in the Xserver vs. stand-alone Mesa.
+ */
+
+#ifndef _XMESA_XF86_H_
+#define _XMESA_XF86_H_
+
+#include "GL/glxtokens.h"
+#include "scrnintstr.h"
+#include "pixmapstr.h"
+#include "gcstruct.h"
+#include "servermd.h"
+
+
+typedef struct _XMesaImageRec {
+    int width, height;
+    char *data;
+    int bytes_per_line; /* Padded to 32 bits */
+    int bits_per_pixel;
+} XMesaImage;
+
+typedef ScreenRec   XMesaDisplay;
+typedef PixmapPtr   XMesaPixmap;
+typedef ColormapPtr XMesaColormap;
+typedef DrawablePtr XMesaDrawable;
+typedef WindowPtr   XMesaWindow;
+typedef GCPtr       XMesaGC;
+typedef VisualPtr   XMesaVisualInfo;
+typedef DDXPointRec XMesaPoint;
+typedef xColorItem  XMesaColor;
+
+#define XMesaSetGeneric(__d,__gc,__val,__mask) \
+do { \
+    CARD32 __v[1]; \
+    (void) __d; \
+    __v[0] = __val; \
+    dixChangeGC(NullClient, __gc, __mask, __v, NULL); \
+} while (0)
+
+#define XMesaSetGenericPtr(__d,__gc,__pval,__mask) \
+do { \
+    ChangeGCVal __v[1]; \
+    (void) __d; \
+    __v[0].ptr = __pval; \
+    dixChangeGC(NullClient, __gc, __mask, NULL, __v); \
+} while (0)
+
+#define XMesaSetForeground(d,gc,v) XMesaSetGeneric(d,gc,v,GCForeground)
+#define XMesaSetBackground(d,gc,v) XMesaSetGeneric(d,gc,v,GCBackground)
+#define XMesaSetPlaneMask(d,gc,v)  XMesaSetGeneric(d,gc,v,GCPlaneMask)
+#define XMesaSetFunction(d,gc,v)   XMesaSetGeneric(d,gc,v,GCFunction)
+#define XMesaSetFillStyle(d,gc,v)  XMesaSetGeneric(d,gc,v,GCFillStyle)
+
+#define XMesaSetTile(d,gc,v)       XMesaSetGenericPtr(d,gc,v,GCTile)
+
+#define XMesaDrawPoint(__d,__b,__gc,__x,__y) \
+do { \
+    XMesaPoint __p[1]; \
+    (void) __d; \
+    __p[0].x = __x; \
+    __p[0].y = __y; \
+    ValidateGC(__b, __gc); \
+    (*gc->ops->PolyPoint)(__b, __gc, CoordModeOrigin, 1, __p); \
+} while (0)
+
+#define XMesaDrawPoints(__d,__b,__gc,__p,__n,__m) \
+do { \
+    (void) __d; \
+    ValidateGC(__b, __gc); \
+    (*gc->ops->PolyPoint)(__b, __gc, __m, __n, __p); \
+} while (0)
+
+#define XMesaDrawLine(__d, __b, __gc, __x0, __y0, __x1, __y1) \
+do { \
+    XMesaPoint __p[2]; \
+    (void) __d; \
+    __p[0].x = __x0; \
+    __p[0].y = __y0; \
+    __p[1].x = __x1; \
+    __p[1].y = __y1; \
+    ValidateGC(__b, __gc); \
+    (*gc->ops->PolyLines)(__b, __gc, CoordModeOrigin, 2, __p); \
+} while (0)
+
+#define XMesaFillRectangle(__d,__b,__gc,__x,__y,__w,__h) \
+do { \
+    xRectangle __r[1]; \
+    (void) __d; \
+    ValidateGC((DrawablePtr)__b, __gc); \
+    __r[0].x = __x; \
+    __r[0].y = __y; \
+    __r[0].width = __w; \
+    __r[0].height = __h; \
+    (*__gc->ops->PolyFillRect)((DrawablePtr)__b, __gc, 1, __r); \
+} while (0)
+
+static _X_INLINE XMesaImage *XMesaGetImage(XMesaDisplay *dpy, PixmapPtr p, int x,
+					int y, unsigned int width,
+					unsigned int height,
+					unsigned long plane_mask, int format)
+{
+    XMesaImage *img = Xcalloc(sizeof(*img));
+
+    img->width = p->drawable.width;
+    img->height = p->drawable.height;
+    img->bits_per_pixel = p->drawable.bitsPerPixel;
+    img->bytes_per_line = PixmapBytePad(width, p->drawable.depth);
+    img->data = malloc(height * img->bytes_per_line);
+
+    /* Assumes: Images are always in ZPixmap format */
+    (*p->drawable.pScreen->GetImage)(&p->drawable, x, y, width, height,
+				     plane_mask, ZPixmap, img->data);
+
+    return img;
+}
+
+#define XMesaPutImage(__d,__b,__gc,__i,__sx,__sy,__x,__y,__w,__h) \
+do { \
+    /* Assumes: Images are always in ZPixmap format */ \
+    (void) __d; \
+    ASSERT(!__sx && !__sy); /* The SubImage case */     \
+    ValidateGC(__b, __gc); \
+    (*__gc->ops->PutImage)(__b, __gc, ((XMesaDrawable)(__b))->depth, \
+			   __x, __y, __w, __h, 0, ZPixmap, \
+			   ((XMesaImage *)(__i))->data); \
+} while (0)
+
+#define XMesaCopyArea(__d,__sb,__db,__gc,__sx,__sy,__w,__h,__x,__y) \
+do { \
+    (void) __d; \
+    ValidateGC(__db, __gc); \
+    (*__gc->ops->CopyArea)((DrawablePtr)__sb, __db, __gc, \
+			   __sx, __sy, __w, __h, __x, __y); \
+} while (0)
+
+
+/* CreatePixmap returns a PixmapPtr; so, it cannot be inside braces */
+#ifdef CREATE_PIXMAP_USAGE_SCRATCH
+#define XMesaCreatePixmap(__d,__b,__w,__h,__depth) \
+  (*__d->CreatePixmap)(__d, __w, __h, __depth, 0)
+#else
+#define XMesaCreatePixmap(__d,__b,__w,__h,__depth) \
+    (*__d->CreatePixmap)(__d, __w, __h, __depth)
+#endif
+
+#define XMesaFreePixmap(__d,__b) \
+    (*__d->DestroyPixmap)(__b)
+
+#define XMesaFreeGC(__d,__gc) \
+do { \
+    (void) __d; \
+    FreeScratchGC(__gc); \
+} while (0)
+
+#define GET_COLORMAP_SIZE(__v)  __v->ColormapEntries
+#define GET_REDMASK(__v)        __v->mesa_visual.redMask
+#define GET_GREENMASK(__v)      __v->mesa_visual.greenMask
+#define GET_BLUEMASK(__v)       __v->mesa_visual.blueMask
+#define GET_VISUAL_DEPTH(__v)   __v->nplanes
+#define GET_BLACK_PIXEL(__v)    __v->display->blackPixel
+#define CHECK_BYTE_ORDER(__v)   GL_TRUE
+#define CHECK_FOR_HPCR(__v)     GL_FALSE
+
+#endif
diff --git a/src/mesa/gl.pc.in b/src/mesa/gl.pc.in
index 1927880d5f..0462b9fca2 100644
--- a/src/mesa/gl.pc.in
+++ b/src/mesa/gl.pc.in
@@ -5,7 +5,8 @@ includedir=@INSTALL_INC_DIR@
 
 Name: gl
 Description: Mesa OpenGL library
-Requires:
+Requires.private: @GL_PC_REQ_PRIV@
 Version: @VERSION@
 Libs: -L${libdir} -lGL
-Cflags: -I${includedir}
+Libs.private: @GL_PC_LIB_PRIV@
+Cflags: -I${includedir} @GL_PC_CFLAGS@
diff --git a/src/mesa/glapi/ARB_framebuffer_object.xml b/src/mesa/glapi/ARB_framebuffer_object.xml
new file mode 100644
index 0000000000..b700e7e77d
--- /dev/null
+++ b/src/mesa/glapi/ARB_framebuffer_object.xml
@@ -0,0 +1,269 @@
+<?xml version="1.0"?>
+<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
+
+<!-- Note: no GLX protocol info yet. -->
+
+
+<OpenGLAPI>
+<category name="GL_ARB_framebuffer_object" number="45">
+    <enum name="FRAMEBUFFER"                              value="0x8D40"/>
+    <enum name="READ_FRAMEBUFFER"                         value="0x8CA8"/>
+    <enum name="DRAW_FRAMEBUFFER"                         value="0x8CA9"/>
+
+    <enum name="RENDERBUFFER"                             value="0x8D41"/>
+
+    <enum name="STENCIL_INDEX1_EXT"                       value="0x8D46"/>
+    <enum name="STENCIL_INDEX4_EXT"                       value="0x8D47"/>
+    <enum name="STENCIL_INDEX8_EXT"                       value="0x8D48"/>
+    <enum name="STENCIL_INDEX16_EXT"                      value="0x8D49"/>
+
+    <enum name="RENDERBUFFER_WIDTH"              value="0x8D42"/>
+    <enum name="RENDERBUFFER_HEIGHT"             value="0x8D43"/>
+    <enum name="RENDERBUFFER_INTERNAL_FORMAT"    value="0x8D44"/>
+    <enum name="RENDERBUFFER_RED_SIZE"           value="0x8D50"/>
+    <enum name="RENDERBUFFER_GREEN_SIZE"         value="0x8D51"/>
+    <enum name="RENDERBUFFER_BLUE_SIZE"          value="0x8D52"/>
+    <enum name="RENDERBUFFER_ALPHA_SIZE"         value="0x8D53"/>
+    <enum name="RENDERBUFFER_DEPTH_SIZE"         value="0x8D54"/>
+    <enum name="RENDERBUFFER_STENCIL_SIZE"       value="0x8D55"/>
+    <enum name="RENDERBUFFER_SAMPLES"            value="0x8CAB"/>
+
+    <enum name="FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE" count="1" value="0x8CD0">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_OBJECT_NAME" count="1" value="0x8CD1">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL" count="1" value="0x8CD2">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE" count="1" value="0x8CD3">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER" count="1" value="0x8CD4">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING" count="1" value="0x8210">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE" count="1" value="0x8211">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_RED_SIZE" count="1" value="0x8212">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_GREEN_SIZE" count="1" value="0x8213">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_BLUE_SIZE" count="1" value="0x8214">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE" count="1" value="0x8215">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE" count="1" value="0x8216">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+    <enum name="FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE" count="1" value="0x8217">
+        <size name="GetFramebufferAttachmentParameteriv" mode="get"/>
+    </enum>
+
+    <enum name="SRGB"                                      value="0x8C40"/>
+    <enum name="UNSIGNED_NORMALIZED"                       value="0x8C17"/>
+    <enum name="FRAMEBUFFER_DEFAULT"                       value="0x8218"/>
+    <enum name="INDEX"                                     value="0x8222"/>
+
+    <enum name="COLOR_ATTACHMENT0"                         value="0x8CE0"/>
+    <enum name="COLOR_ATTACHMENT1"                         value="0x8CE1"/>
+    <enum name="COLOR_ATTACHMENT2"                         value="0x8CE2"/>
+    <enum name="COLOR_ATTACHMENT3"                         value="0x8CE3"/>
+    <enum name="COLOR_ATTACHMENT4"                         value="0x8CE4"/>
+    <enum name="COLOR_ATTACHMENT5"                         value="0x8CE5"/>
+    <enum name="COLOR_ATTACHMENT6"                         value="0x8CE6"/>
+    <enum name="COLOR_ATTACHMENT7"                         value="0x8CE7"/>
+    <enum name="COLOR_ATTACHMENT8"                         value="0x8CE8"/>
+    <enum name="COLOR_ATTACHMENT9"                         value="0x8CE9"/>
+    <enum name="COLOR_ATTACHMENT10"                        value="0x8CEA"/>
+    <enum name="COLOR_ATTACHMENT11"                        value="0x8CEB"/>
+    <enum name="COLOR_ATTACHMENT12"                        value="0x8CEC"/>
+    <enum name="COLOR_ATTACHMENT13"                        value="0x8CED"/>
+    <enum name="COLOR_ATTACHMENT14"                        value="0x8CEE"/>
+    <enum name="COLOR_ATTACHMENT15"                        value="0x8CEF"/>
+    <enum name="DEPTH_ATTACHMENT"                          value="0x8D00"/>
+    <enum name="STENCIL_ATTACHMENT"                        value="0x8D20"/>
+    <enum name="DEPTH_STENCIL_ATTACHMENT"                  value="0x821A"/>
+
+    <enum name="MAX_SAMPLES"                               value="0x8D57"/>
+
+    <enum name="FRAMEBUFFER_COMPLETE"                      value="0x8CD5"/>
+    <enum name="FRAMEBUFFER_INCOMPLETE_ATTACHMENT"         value="0x8CD6"/>
+    <enum name="FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT" value="0x8CD7"/>
+    <enum name="FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT"    value="0x8CDB"/>
+    <enum name="FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT"    value="0x8CDC"/>
+    <enum name="FRAMEBUFFER_UNSUPPORTED"                   value="0x8CDD"/>
+    <enum name="FRAMEBUFFER_INCOMPLETE_MULTISAMPLE"        value="0x8D56"/>
+    <enum name="FRAMEBUFFER_UNDEFINED"                     value="0x8219"/>
+
+    <enum name="FRAMEBUFFER_BINDING_EXT"         count="1" value="0x8CA6">
+        <size name="Get" mode="get"/>
+    </enum>
+    <enum name="DRAW_FRAMEBUFFER_BINDING_EXT"    count="1" value="0x8CA6">
+        <size name="Get" mode="get"/>
+    </enum>
+    <enum name="READ_FRAMEBUFFER_BINDING_EXT"    count="1" value="0x8CAA">
+        <size name="Get" mode="get"/>
+    </enum>
+    <enum name="RENDERBUFFER_BINDING_EXT"        count="1" value="0x8CA7">
+        <size name="Get" mode="get"/>
+    </enum>
+    <enum name="MAX_COLOR_ATTACHMENTS_EXT"       count="1" value="0x8CDF">
+        <size name="Get" mode="get"/>
+    </enum>
+    <enum name="MAX_RENDERBUFFER_SIZE_EXT"       count="1" value="0x84E8">
+        <size name="Get" mode="get"/>
+    </enum>
+
+    <enum name="INVALID_FRAMEBUFFER_OPERATION"            value="0x0506"/>
+
+    <enum name="DEPTH_STENCIL"                            value="0x84F9"/>
+
+    <enum name="UNSIGNED_INT_24_8"                        value="0x84FA"/>
+
+    <enum name="DEPTH24_STENCIL8"                         value="0x88F0"/>
+
+    <enum name="TEXTURE_STENCIL_SIZE"                     value="0x88F1"/>
+
+
+
+    <function name="IsRenderbuffer" alias="IsRenderbufferEXT">
+        <param name="renderbuffer" type="GLuint"/>
+	<return type="GLboolean"/>
+    </function>
+
+    <function name="BindRenderbuffer" alias="BindRenderbufferEXT">
+        <param name="target" type="GLenum"/>
+        <param name="renderbuffer" type="GLuint"/>
+    </function>
+
+    <function name="DeleteRenderbuffers" alias="DeleteRenderbuffersEXT">
+        <param name="n" type="GLsizei" counter="true"/>
+        <param name="renderbuffers" type="const GLuint *" count="n"/>
+    </function>
+
+    <function name="GenRenderbuffers" alias="GenRenderbuffersEXT">
+        <param name="n" type="GLsizei" counter="true"/>
+        <param name="renderbuffers" type="GLuint *" count="n" output="true"/>
+    </function>
+
+    <function name="RenderbufferStorage" alias="RenderbufferStorageEXT">
+        <param name="target" type="GLenum"/>
+        <param name="internalformat" type="GLenum"/>
+        <param name="width" type="GLsizei"/>
+        <param name="height" type="GLsizei"/>
+    </function>
+
+    <function name="RenderbufferStorageMultisample" offset="assign">
+        <param name="target" type="GLenum"/>
+        <param name="samples" type="GLsizei"/>
+        <param name="internalformat" type="GLenum"/>
+        <param name="width" type="GLsizei"/>
+        <param name="height" type="GLsizei"/>
+    </function>
+
+    <function name="GetRenderbufferParameteriv" alias="GetRenderbufferParameterivEXT">
+        <param name="target" type="GLenum"/>
+        <param name="pname" type="GLenum"/>
+        <param name="params" type="GLint *" output="true"/>
+    </function>
+
+    <function name="IsFramebuffer" alias="IsFramebufferEXT">
+        <param name="framebuffer" type="GLuint"/>
+	<return type="GLboolean"/>
+    </function>
+
+    <function name="BindFramebuffer" alias="BindFramebufferEXT">
+        <param name="target" type="GLenum"/>
+        <param name="framebuffer" type="GLuint"/>
+    </function>
+
+    <function name="DeleteFramebuffers" alias="DeleteFramebuffersEXT">
+        <param name="n" type="GLsizei" counter="true"/>
+        <param name="framebuffers" type="const GLuint *" count="n"/>
+    </function>
+
+    <function name="GenFramebuffers" alias="GenFramebuffersEXT">
+        <param name="n" type="GLsizei" counter="true"/>
+        <param name="framebuffers" type="GLuint *" count="n" output="true"/>
+    </function>
+
+    <function name="CheckFramebufferStatus" alias="CheckFramebufferStatusEXT">
+        <param name="target" type="GLenum"/>
+	<return type="GLenum"/>
+    </function>
+
+    <function name="FramebufferTexture1D" alias="FramebufferTexture1DEXT">
+        <param name="target" type="GLenum"/>
+        <param name="attachment" type="GLenum"/>
+        <param name="textarget" type="GLenum"/>
+        <param name="texture" type="GLuint"/>
+        <param name="level" type="GLint"/>
+    </function>
+
+    <function name="FramebufferTexture2D" alias="FramebufferTexture2DEXT">
+        <param name="target" type="GLenum"/>
+        <param name="attachment" type="GLenum"/>
+        <param name="textarget" type="GLenum"/>
+        <param name="texture" type="GLuint"/>
+        <param name="level" type="GLint"/>
+    </function>
+
+    <function name="FramebufferTexture3D" alias="FramebufferTexture3DEXT">
+        <param name="target" type="GLenum"/>
+        <param name="attachment" type="GLenum"/>
+        <param name="textarget" type="GLenum"/>
+        <param name="texture" type="GLuint"/>
+        <param name="level" type="GLint"/>
+        <param name="zoffset" type="GLint"/>
+    </function>
+
+    <function name="FramebufferTextureLayer" alias="FramebufferTextureLayerEXT">
+        <param name="target" type="GLenum"/>
+        <param name="attachment" type="GLenum"/>
+        <param name="texture" type="GLuint"/>
+        <param name="level" type="GLint"/>
+        <param name="layer" type="GLint"/>
+    </function>
+
+    <function name="FramebufferRenderbuffer" alias="FramebufferRenderbufferEXT">
+        <param name="target" type="GLenum"/>
+        <param name="attachment" type="GLenum"/>
+        <param name="renderbuffertarget" type="GLenum"/>
+        <param name="renderbuffer" type="GLuint"/>
+    </function>
+
+    <function name="GetFramebufferAttachmentParameteriv" alias="GetFramebufferAttachmentParameterivEXT">
+        <param name="target" type="GLenum"/>
+        <param name="attachment" type="GLenum"/>
+        <param name="pname" type="GLenum"/>
+        <param name="params" type="GLint *" output="true"/>
+    </function>
+
+    <function name="BlitFramebuffer" alias="BlitFramebufferEXT">
+        <param name="srcX0" type="GLint"/>
+        <param name="srcY0" type="GLint"/>
+        <param name="srcX1" type="GLint"/>
+        <param name="srcY1" type="GLint"/>
+        <param name="dstX0" type="GLint"/>
+        <param name="dstY0" type="GLint"/>
+        <param name="dstX1" type="GLint"/>
+        <param name="dstY1" type="GLint"/>
+        <param name="mask" type="GLbitfield"/>
+        <param name="filter" type="GLenum"/>
+    </function>
+
+    <function name="GenerateMipmap" alias="GenerateMipmapEXT">
+        <param name="target" type="GLenum"/>
+    </function>
+</category>
+
+</OpenGLAPI>
diff --git a/src/mesa/glapi/Makefile b/src/mesa/glapi/Makefile
index adc53d9249..f524167a47 100644
--- a/src/mesa/glapi/Makefile
+++ b/src/mesa/glapi/Makefile
@@ -47,6 +47,7 @@ SERVER_OUTPUTS = \
 
 API_XML = gl_API.xml \
 	EXT_framebuffer_object.xml \
+	ARB_framebuffer_object.xml \
 	APPLE_vertex_array_object.xml
 
 COMMON = gl_XML.py glX_XML.py license.py $(API_XML) typeexpr.py
diff --git a/src/mesa/glapi/descrip.mms b/src/mesa/glapi/descrip.mms
deleted file mode 100644
index ee96baa6fc..0000000000
--- a/src/mesa/glapi/descrip.mms
+++ /dev/null
@@ -1,39 +0,0 @@
-# Makefile for core library for VMS
-# contributed by Jouk Jansen  joukj@hrem.nano.tudelft.nl
-# Last revision : 29 September 2008
-
-.first
-	define gl [---.include.gl]
-	define main [-.main]
-
-.include [---]mms-config.
-
-##### MACROS #####
-
-VPATH = RCS
-
-INCDIR = [---.include],[-.main]
-LIBDIR = [---.lib]
-CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm
-
-SOURCES = glapi.c glthread.c glapi_getproc.c
-
-OBJECTS =  glapi.obj,glthread.obj,glapi_getproc.obj
-
-##### RULES #####
-
-VERSION=Mesa V3.4
-
-##### TARGETS #####
-# Make the library
-$(LIBDIR)$(GL_LIB) : $(OBJECTS)
-  @ library $(LIBDIR)$(GL_LIB) $(OBJECTS)
-
-clean :
-	purge
-	delete *.obj;*
-
-glapi.obj : glapi.c
-
-glthread.obj : glthread.c
-glapi_getproc.obj : glapi_getproc.c
diff --git a/src/mesa/glapi/dispatch.h b/src/mesa/glapi/dispatch.h
index f019aa6aaf..d7f21692ed 100644
--- a/src/mesa/glapi/dispatch.h
+++ b/src/mesa/glapi/dispatch.h
@@ -1743,6 +1743,9 @@
 #define CALL_DrawBuffersARB(disp, parameters) (*((disp)->DrawBuffersARB)) parameters
 #define GET_DrawBuffersARB(disp) ((disp)->DrawBuffersARB)
 #define SET_DrawBuffersARB(disp, fn) ((disp)->DrawBuffersARB = fn)
+#define CALL_RenderbufferStorageMultisample(disp, parameters) (*((disp)->RenderbufferStorageMultisample)) parameters
+#define GET_RenderbufferStorageMultisample(disp) ((disp)->RenderbufferStorageMultisample)
+#define SET_RenderbufferStorageMultisample(disp, fn) ((disp)->RenderbufferStorageMultisample = fn)
 #define CALL_PolygonOffsetEXT(disp, parameters) (*((disp)->PolygonOffsetEXT)) parameters
 #define GET_PolygonOffsetEXT(disp) ((disp)->PolygonOffsetEXT)
 #define SET_PolygonOffsetEXT(disp, fn) ((disp)->PolygonOffsetEXT = fn)
@@ -2373,7 +2376,7 @@
 
 #else
 
-#define driDispatchRemapTable_size 362
+#define driDispatchRemapTable_size 363
 extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 
 #define AttachShader_remap_index 0
@@ -2529,215 +2532,216 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 #define GetActiveAttribARB_remap_index 150
 #define GetAttribLocationARB_remap_index 151
 #define DrawBuffersARB_remap_index 152
-#define PolygonOffsetEXT_remap_index 153
-#define GetPixelTexGenParameterfvSGIS_remap_index 154
-#define GetPixelTexGenParameterivSGIS_remap_index 155
-#define PixelTexGenParameterfSGIS_remap_index 156
-#define PixelTexGenParameterfvSGIS_remap_index 157
-#define PixelTexGenParameteriSGIS_remap_index 158
-#define PixelTexGenParameterivSGIS_remap_index 159
-#define SampleMaskSGIS_remap_index 160
-#define SamplePatternSGIS_remap_index 161
-#define ColorPointerEXT_remap_index 162
-#define EdgeFlagPointerEXT_remap_index 163
-#define IndexPointerEXT_remap_index 164
-#define NormalPointerEXT_remap_index 165
-#define TexCoordPointerEXT_remap_index 166
-#define VertexPointerEXT_remap_index 167
-#define PointParameterfEXT_remap_index 168
-#define PointParameterfvEXT_remap_index 169
-#define LockArraysEXT_remap_index 170
-#define UnlockArraysEXT_remap_index 171
-#define CullParameterdvEXT_remap_index 172
-#define CullParameterfvEXT_remap_index 173
-#define SecondaryColor3bEXT_remap_index 174
-#define SecondaryColor3bvEXT_remap_index 175
-#define SecondaryColor3dEXT_remap_index 176
-#define SecondaryColor3dvEXT_remap_index 177
-#define SecondaryColor3fEXT_remap_index 178
-#define SecondaryColor3fvEXT_remap_index 179
-#define SecondaryColor3iEXT_remap_index 180
-#define SecondaryColor3ivEXT_remap_index 181
-#define SecondaryColor3sEXT_remap_index 182
-#define SecondaryColor3svEXT_remap_index 183
-#define SecondaryColor3ubEXT_remap_index 184
-#define SecondaryColor3ubvEXT_remap_index 185
-#define SecondaryColor3uiEXT_remap_index 186
-#define SecondaryColor3uivEXT_remap_index 187
-#define SecondaryColor3usEXT_remap_index 188
-#define SecondaryColor3usvEXT_remap_index 189
-#define SecondaryColorPointerEXT_remap_index 190
-#define MultiDrawArraysEXT_remap_index 191
-#define MultiDrawElementsEXT_remap_index 192
-#define FogCoordPointerEXT_remap_index 193
-#define FogCoorddEXT_remap_index 194
-#define FogCoorddvEXT_remap_index 195
-#define FogCoordfEXT_remap_index 196
-#define FogCoordfvEXT_remap_index 197
-#define PixelTexGenSGIX_remap_index 198
-#define BlendFuncSeparateEXT_remap_index 199
-#define FlushVertexArrayRangeNV_remap_index 200
-#define VertexArrayRangeNV_remap_index 201
-#define CombinerInputNV_remap_index 202
-#define CombinerOutputNV_remap_index 203
-#define CombinerParameterfNV_remap_index 204
-#define CombinerParameterfvNV_remap_index 205
-#define CombinerParameteriNV_remap_index 206
-#define CombinerParameterivNV_remap_index 207
-#define FinalCombinerInputNV_remap_index 208
-#define GetCombinerInputParameterfvNV_remap_index 209
-#define GetCombinerInputParameterivNV_remap_index 210
-#define GetCombinerOutputParameterfvNV_remap_index 211
-#define GetCombinerOutputParameterivNV_remap_index 212
-#define GetFinalCombinerInputParameterfvNV_remap_index 213
-#define GetFinalCombinerInputParameterivNV_remap_index 214
-#define ResizeBuffersMESA_remap_index 215
-#define WindowPos2dMESA_remap_index 216
-#define WindowPos2dvMESA_remap_index 217
-#define WindowPos2fMESA_remap_index 218
-#define WindowPos2fvMESA_remap_index 219
-#define WindowPos2iMESA_remap_index 220
-#define WindowPos2ivMESA_remap_index 221
-#define WindowPos2sMESA_remap_index 222
-#define WindowPos2svMESA_remap_index 223
-#define WindowPos3dMESA_remap_index 224
-#define WindowPos3dvMESA_remap_index 225
-#define WindowPos3fMESA_remap_index 226
-#define WindowPos3fvMESA_remap_index 227
-#define WindowPos3iMESA_remap_index 228
-#define WindowPos3ivMESA_remap_index 229
-#define WindowPos3sMESA_remap_index 230
-#define WindowPos3svMESA_remap_index 231
-#define WindowPos4dMESA_remap_index 232
-#define WindowPos4dvMESA_remap_index 233
-#define WindowPos4fMESA_remap_index 234
-#define WindowPos4fvMESA_remap_index 235
-#define WindowPos4iMESA_remap_index 236
-#define WindowPos4ivMESA_remap_index 237
-#define WindowPos4sMESA_remap_index 238
-#define WindowPos4svMESA_remap_index 239
-#define MultiModeDrawArraysIBM_remap_index 240
-#define MultiModeDrawElementsIBM_remap_index 241
-#define DeleteFencesNV_remap_index 242
-#define FinishFenceNV_remap_index 243
-#define GenFencesNV_remap_index 244
-#define GetFenceivNV_remap_index 245
-#define IsFenceNV_remap_index 246
-#define SetFenceNV_remap_index 247
-#define TestFenceNV_remap_index 248
-#define AreProgramsResidentNV_remap_index 249
-#define BindProgramNV_remap_index 250
-#define DeleteProgramsNV_remap_index 251
-#define ExecuteProgramNV_remap_index 252
-#define GenProgramsNV_remap_index 253
-#define GetProgramParameterdvNV_remap_index 254
-#define GetProgramParameterfvNV_remap_index 255
-#define GetProgramStringNV_remap_index 256
-#define GetProgramivNV_remap_index 257
-#define GetTrackMatrixivNV_remap_index 258
-#define GetVertexAttribPointervNV_remap_index 259
-#define GetVertexAttribdvNV_remap_index 260
-#define GetVertexAttribfvNV_remap_index 261
-#define GetVertexAttribivNV_remap_index 262
-#define IsProgramNV_remap_index 263
-#define LoadProgramNV_remap_index 264
-#define ProgramParameters4dvNV_remap_index 265
-#define ProgramParameters4fvNV_remap_index 266
-#define RequestResidentProgramsNV_remap_index 267
-#define TrackMatrixNV_remap_index 268
-#define VertexAttrib1dNV_remap_index 269
-#define VertexAttrib1dvNV_remap_index 270
-#define VertexAttrib1fNV_remap_index 271
-#define VertexAttrib1fvNV_remap_index 272
-#define VertexAttrib1sNV_remap_index 273
-#define VertexAttrib1svNV_remap_index 274
-#define VertexAttrib2dNV_remap_index 275
-#define VertexAttrib2dvNV_remap_index 276
-#define VertexAttrib2fNV_remap_index 277
-#define VertexAttrib2fvNV_remap_index 278
-#define VertexAttrib2sNV_remap_index 279
-#define VertexAttrib2svNV_remap_index 280
-#define VertexAttrib3dNV_remap_index 281
-#define VertexAttrib3dvNV_remap_index 282
-#define VertexAttrib3fNV_remap_index 283
-#define VertexAttrib3fvNV_remap_index 284
-#define VertexAttrib3sNV_remap_index 285
-#define VertexAttrib3svNV_remap_index 286
-#define VertexAttrib4dNV_remap_index 287
-#define VertexAttrib4dvNV_remap_index 288
-#define VertexAttrib4fNV_remap_index 289
-#define VertexAttrib4fvNV_remap_index 290
-#define VertexAttrib4sNV_remap_index 291
-#define VertexAttrib4svNV_remap_index 292
-#define VertexAttrib4ubNV_remap_index 293
-#define VertexAttrib4ubvNV_remap_index 294
-#define VertexAttribPointerNV_remap_index 295
-#define VertexAttribs1dvNV_remap_index 296
-#define VertexAttribs1fvNV_remap_index 297
-#define VertexAttribs1svNV_remap_index 298
-#define VertexAttribs2dvNV_remap_index 299
-#define VertexAttribs2fvNV_remap_index 300
-#define VertexAttribs2svNV_remap_index 301
-#define VertexAttribs3dvNV_remap_index 302
-#define VertexAttribs3fvNV_remap_index 303
-#define VertexAttribs3svNV_remap_index 304
-#define VertexAttribs4dvNV_remap_index 305
-#define VertexAttribs4fvNV_remap_index 306
-#define VertexAttribs4svNV_remap_index 307
-#define VertexAttribs4ubvNV_remap_index 308
-#define AlphaFragmentOp1ATI_remap_index 309
-#define AlphaFragmentOp2ATI_remap_index 310
-#define AlphaFragmentOp3ATI_remap_index 311
-#define BeginFragmentShaderATI_remap_index 312
-#define BindFragmentShaderATI_remap_index 313
-#define ColorFragmentOp1ATI_remap_index 314
-#define ColorFragmentOp2ATI_remap_index 315
-#define ColorFragmentOp3ATI_remap_index 316
-#define DeleteFragmentShaderATI_remap_index 317
-#define EndFragmentShaderATI_remap_index 318
-#define GenFragmentShadersATI_remap_index 319
-#define PassTexCoordATI_remap_index 320
-#define SampleMapATI_remap_index 321
-#define SetFragmentShaderConstantATI_remap_index 322
-#define PointParameteriNV_remap_index 323
-#define PointParameterivNV_remap_index 324
-#define ActiveStencilFaceEXT_remap_index 325
-#define BindVertexArrayAPPLE_remap_index 326
-#define DeleteVertexArraysAPPLE_remap_index 327
-#define GenVertexArraysAPPLE_remap_index 328
-#define IsVertexArrayAPPLE_remap_index 329
-#define GetProgramNamedParameterdvNV_remap_index 330
-#define GetProgramNamedParameterfvNV_remap_index 331
-#define ProgramNamedParameter4dNV_remap_index 332
-#define ProgramNamedParameter4dvNV_remap_index 333
-#define ProgramNamedParameter4fNV_remap_index 334
-#define ProgramNamedParameter4fvNV_remap_index 335
-#define DepthBoundsEXT_remap_index 336
-#define BlendEquationSeparateEXT_remap_index 337
-#define BindFramebufferEXT_remap_index 338
-#define BindRenderbufferEXT_remap_index 339
-#define CheckFramebufferStatusEXT_remap_index 340
-#define DeleteFramebuffersEXT_remap_index 341
-#define DeleteRenderbuffersEXT_remap_index 342
-#define FramebufferRenderbufferEXT_remap_index 343
-#define FramebufferTexture1DEXT_remap_index 344
-#define FramebufferTexture2DEXT_remap_index 345
-#define FramebufferTexture3DEXT_remap_index 346
-#define GenFramebuffersEXT_remap_index 347
-#define GenRenderbuffersEXT_remap_index 348
-#define GenerateMipmapEXT_remap_index 349
-#define GetFramebufferAttachmentParameterivEXT_remap_index 350
-#define GetRenderbufferParameterivEXT_remap_index 351
-#define IsFramebufferEXT_remap_index 352
-#define IsRenderbufferEXT_remap_index 353
-#define RenderbufferStorageEXT_remap_index 354
-#define BlitFramebufferEXT_remap_index 355
-#define FramebufferTextureLayerEXT_remap_index 356
-#define StencilFuncSeparateATI_remap_index 357
-#define ProgramEnvParameters4fvEXT_remap_index 358
-#define ProgramLocalParameters4fvEXT_remap_index 359
-#define GetQueryObjecti64vEXT_remap_index 360
-#define GetQueryObjectui64vEXT_remap_index 361
+#define RenderbufferStorageMultisample_remap_index 153
+#define PolygonOffsetEXT_remap_index 154
+#define GetPixelTexGenParameterfvSGIS_remap_index 155
+#define GetPixelTexGenParameterivSGIS_remap_index 156
+#define PixelTexGenParameterfSGIS_remap_index 157
+#define PixelTexGenParameterfvSGIS_remap_index 158
+#define PixelTexGenParameteriSGIS_remap_index 159
+#define PixelTexGenParameterivSGIS_remap_index 160
+#define SampleMaskSGIS_remap_index 161
+#define SamplePatternSGIS_remap_index 162
+#define ColorPointerEXT_remap_index 163
+#define EdgeFlagPointerEXT_remap_index 164
+#define IndexPointerEXT_remap_index 165
+#define NormalPointerEXT_remap_index 166
+#define TexCoordPointerEXT_remap_index 167
+#define VertexPointerEXT_remap_index 168
+#define PointParameterfEXT_remap_index 169
+#define PointParameterfvEXT_remap_index 170
+#define LockArraysEXT_remap_index 171
+#define UnlockArraysEXT_remap_index 172
+#define CullParameterdvEXT_remap_index 173
+#define CullParameterfvEXT_remap_index 174
+#define SecondaryColor3bEXT_remap_index 175
+#define SecondaryColor3bvEXT_remap_index 176
+#define SecondaryColor3dEXT_remap_index 177
+#define SecondaryColor3dvEXT_remap_index 178
+#define SecondaryColor3fEXT_remap_index 179
+#define SecondaryColor3fvEXT_remap_index 180
+#define SecondaryColor3iEXT_remap_index 181
+#define SecondaryColor3ivEXT_remap_index 182
+#define SecondaryColor3sEXT_remap_index 183
+#define SecondaryColor3svEXT_remap_index 184
+#define SecondaryColor3ubEXT_remap_index 185
+#define SecondaryColor3ubvEXT_remap_index 186
+#define SecondaryColor3uiEXT_remap_index 187
+#define SecondaryColor3uivEXT_remap_index 188
+#define SecondaryColor3usEXT_remap_index 189
+#define SecondaryColor3usvEXT_remap_index 190
+#define SecondaryColorPointerEXT_remap_index 191
+#define MultiDrawArraysEXT_remap_index 192
+#define MultiDrawElementsEXT_remap_index 193
+#define FogCoordPointerEXT_remap_index 194
+#define FogCoorddEXT_remap_index 195
+#define FogCoorddvEXT_remap_index 196
+#define FogCoordfEXT_remap_index 197
+#define FogCoordfvEXT_remap_index 198
+#define PixelTexGenSGIX_remap_index 199
+#define BlendFuncSeparateEXT_remap_index 200
+#define FlushVertexArrayRangeNV_remap_index 201
+#define VertexArrayRangeNV_remap_index 202
+#define CombinerInputNV_remap_index 203
+#define CombinerOutputNV_remap_index 204
+#define CombinerParameterfNV_remap_index 205
+#define CombinerParameterfvNV_remap_index 206
+#define CombinerParameteriNV_remap_index 207
+#define CombinerParameterivNV_remap_index 208
+#define FinalCombinerInputNV_remap_index 209
+#define GetCombinerInputParameterfvNV_remap_index 210
+#define GetCombinerInputParameterivNV_remap_index 211
+#define GetCombinerOutputParameterfvNV_remap_index 212
+#define GetCombinerOutputParameterivNV_remap_index 213
+#define GetFinalCombinerInputParameterfvNV_remap_index 214
+#define GetFinalCombinerInputParameterivNV_remap_index 215
+#define ResizeBuffersMESA_remap_index 216
+#define WindowPos2dMESA_remap_index 217
+#define WindowPos2dvMESA_remap_index 218
+#define WindowPos2fMESA_remap_index 219
+#define WindowPos2fvMESA_remap_index 220
+#define WindowPos2iMESA_remap_index 221
+#define WindowPos2ivMESA_remap_index 222
+#define WindowPos2sMESA_remap_index 223
+#define WindowPos2svMESA_remap_index 224
+#define WindowPos3dMESA_remap_index 225
+#define WindowPos3dvMESA_remap_index 226
+#define WindowPos3fMESA_remap_index 227
+#define WindowPos3fvMESA_remap_index 228
+#define WindowPos3iMESA_remap_index 229
+#define WindowPos3ivMESA_remap_index 230
+#define WindowPos3sMESA_remap_index 231
+#define WindowPos3svMESA_remap_index 232
+#define WindowPos4dMESA_remap_index 233
+#define WindowPos4dvMESA_remap_index 234
+#define WindowPos4fMESA_remap_index 235
+#define WindowPos4fvMESA_remap_index 236
+#define WindowPos4iMESA_remap_index 237
+#define WindowPos4ivMESA_remap_index 238
+#define WindowPos4sMESA_remap_index 239
+#define WindowPos4svMESA_remap_index 240
+#define MultiModeDrawArraysIBM_remap_index 241
+#define MultiModeDrawElementsIBM_remap_index 242
+#define DeleteFencesNV_remap_index 243
+#define FinishFenceNV_remap_index 244
+#define GenFencesNV_remap_index 245
+#define GetFenceivNV_remap_index 246
+#define IsFenceNV_remap_index 247
+#define SetFenceNV_remap_index 248
+#define TestFenceNV_remap_index 249
+#define AreProgramsResidentNV_remap_index 250
+#define BindProgramNV_remap_index 251
+#define DeleteProgramsNV_remap_index 252
+#define ExecuteProgramNV_remap_index 253
+#define GenProgramsNV_remap_index 254
+#define GetProgramParameterdvNV_remap_index 255
+#define GetProgramParameterfvNV_remap_index 256
+#define GetProgramStringNV_remap_index 257
+#define GetProgramivNV_remap_index 258
+#define GetTrackMatrixivNV_remap_index 259
+#define GetVertexAttribPointervNV_remap_index 260
+#define GetVertexAttribdvNV_remap_index 261
+#define GetVertexAttribfvNV_remap_index 262
+#define GetVertexAttribivNV_remap_index 263
+#define IsProgramNV_remap_index 264
+#define LoadProgramNV_remap_index 265
+#define ProgramParameters4dvNV_remap_index 266
+#define ProgramParameters4fvNV_remap_index 267
+#define RequestResidentProgramsNV_remap_index 268
+#define TrackMatrixNV_remap_index 269
+#define VertexAttrib1dNV_remap_index 270
+#define VertexAttrib1dvNV_remap_index 271
+#define VertexAttrib1fNV_remap_index 272
+#define VertexAttrib1fvNV_remap_index 273
+#define VertexAttrib1sNV_remap_index 274
+#define VertexAttrib1svNV_remap_index 275
+#define VertexAttrib2dNV_remap_index 276
+#define VertexAttrib2dvNV_remap_index 277
+#define VertexAttrib2fNV_remap_index 278
+#define VertexAttrib2fvNV_remap_index 279
+#define VertexAttrib2sNV_remap_index 280
+#define VertexAttrib2svNV_remap_index 281
+#define VertexAttrib3dNV_remap_index 282
+#define VertexAttrib3dvNV_remap_index 283
+#define VertexAttrib3fNV_remap_index 284
+#define VertexAttrib3fvNV_remap_index 285
+#define VertexAttrib3sNV_remap_index 286
+#define VertexAttrib3svNV_remap_index 287
+#define VertexAttrib4dNV_remap_index 288
+#define VertexAttrib4dvNV_remap_index 289
+#define VertexAttrib4fNV_remap_index 290
+#define VertexAttrib4fvNV_remap_index 291
+#define VertexAttrib4sNV_remap_index 292
+#define VertexAttrib4svNV_remap_index 293
+#define VertexAttrib4ubNV_remap_index 294
+#define VertexAttrib4ubvNV_remap_index 295
+#define VertexAttribPointerNV_remap_index 296
+#define VertexAttribs1dvNV_remap_index 297
+#define VertexAttribs1fvNV_remap_index 298
+#define VertexAttribs1svNV_remap_index 299
+#define VertexAttribs2dvNV_remap_index 300
+#define VertexAttribs2fvNV_remap_index 301
+#define VertexAttribs2svNV_remap_index 302
+#define VertexAttribs3dvNV_remap_index 303
+#define VertexAttribs3fvNV_remap_index 304
+#define VertexAttribs3svNV_remap_index 305
+#define VertexAttribs4dvNV_remap_index 306
+#define VertexAttribs4fvNV_remap_index 307
+#define VertexAttribs4svNV_remap_index 308
+#define VertexAttribs4ubvNV_remap_index 309
+#define AlphaFragmentOp1ATI_remap_index 310
+#define AlphaFragmentOp2ATI_remap_index 311
+#define AlphaFragmentOp3ATI_remap_index 312
+#define BeginFragmentShaderATI_remap_index 313
+#define BindFragmentShaderATI_remap_index 314
+#define ColorFragmentOp1ATI_remap_index 315
+#define ColorFragmentOp2ATI_remap_index 316
+#define ColorFragmentOp3ATI_remap_index 317
+#define DeleteFragmentShaderATI_remap_index 318
+#define EndFragmentShaderATI_remap_index 319
+#define GenFragmentShadersATI_remap_index 320
+#define PassTexCoordATI_remap_index 321
+#define SampleMapATI_remap_index 322
+#define SetFragmentShaderConstantATI_remap_index 323
+#define PointParameteriNV_remap_index 324
+#define PointParameterivNV_remap_index 325
+#define ActiveStencilFaceEXT_remap_index 326
+#define BindVertexArrayAPPLE_remap_index 327
+#define DeleteVertexArraysAPPLE_remap_index 328
+#define GenVertexArraysAPPLE_remap_index 329
+#define IsVertexArrayAPPLE_remap_index 330
+#define GetProgramNamedParameterdvNV_remap_index 331
+#define GetProgramNamedParameterfvNV_remap_index 332
+#define ProgramNamedParameter4dNV_remap_index 333
+#define ProgramNamedParameter4dvNV_remap_index 334
+#define ProgramNamedParameter4fNV_remap_index 335
+#define ProgramNamedParameter4fvNV_remap_index 336
+#define DepthBoundsEXT_remap_index 337
+#define BlendEquationSeparateEXT_remap_index 338
+#define BindFramebufferEXT_remap_index 339
+#define BindRenderbufferEXT_remap_index 340
+#define CheckFramebufferStatusEXT_remap_index 341
+#define DeleteFramebuffersEXT_remap_index 342
+#define DeleteRenderbuffersEXT_remap_index 343
+#define FramebufferRenderbufferEXT_remap_index 344
+#define FramebufferTexture1DEXT_remap_index 345
+#define FramebufferTexture2DEXT_remap_index 346
+#define FramebufferTexture3DEXT_remap_index 347
+#define GenFramebuffersEXT_remap_index 348
+#define GenRenderbuffersEXT_remap_index 349
+#define GenerateMipmapEXT_remap_index 350
+#define GetFramebufferAttachmentParameterivEXT_remap_index 351
+#define GetRenderbufferParameterivEXT_remap_index 352
+#define IsFramebufferEXT_remap_index 353
+#define IsRenderbufferEXT_remap_index 354
+#define RenderbufferStorageEXT_remap_index 355
+#define BlitFramebufferEXT_remap_index 356
+#define FramebufferTextureLayerEXT_remap_index 357
+#define StencilFuncSeparateATI_remap_index 358
+#define ProgramEnvParameters4fvEXT_remap_index 359
+#define ProgramLocalParameters4fvEXT_remap_index 360
+#define GetQueryObjecti64vEXT_remap_index 361
+#define GetQueryObjectui64vEXT_remap_index 362
 
 #define CALL_AttachShader(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLuint, GLuint)), driDispatchRemapTable[AttachShader_remap_index], parameters)
 #define GET_AttachShader(disp) GET_by_offset(disp, driDispatchRemapTable[AttachShader_remap_index])
@@ -3198,6 +3202,9 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ];
 #define CALL_DrawBuffersARB(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLsizei, const GLenum *)), driDispatchRemapTable[DrawBuffersARB_remap_index], parameters)
 #define GET_DrawBuffersARB(disp) GET_by_offset(disp, driDispatchRemapTable[DrawBuffersARB_remap_index])
 #define SET_DrawBuffersARB(disp, fn) SET_by_offset(disp, driDispatchRemapTable[DrawBuffersARB_remap_index], fn)
+#define CALL_RenderbufferStorageMultisample(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLsizei, GLenum, GLsizei, GLsizei)), driDispatchRemapTable[RenderbufferStorageMultisample_remap_index], parameters)
+#define GET_RenderbufferStorageMultisample(disp) GET_by_offset(disp, driDispatchRemapTable[RenderbufferStorageMultisample_remap_index])
+#define SET_RenderbufferStorageMultisample(disp, fn) SET_by_offset(disp, driDispatchRemapTable[RenderbufferStorageMultisample_remap_index], fn)
 #define CALL_PolygonOffsetEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLfloat, GLfloat)), driDispatchRemapTable[PolygonOffsetEXT_remap_index], parameters)
 #define GET_PolygonOffsetEXT(disp) GET_by_offset(disp, driDispatchRemapTable[PolygonOffsetEXT_remap_index])
 #define SET_PolygonOffsetEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[PolygonOffsetEXT_remap_index], fn)
diff --git a/src/mesa/glapi/extension_helper.py b/src/mesa/glapi/extension_helper.py
index 375e3ea59e..64f64a2fd8 100644
--- a/src/mesa/glapi/extension_helper.py
+++ b/src/mesa/glapi/extension_helper.py
@@ -174,6 +174,9 @@ class PrintGlExtensionGlue(gl_XML.gl_print_base):
 
 				parameter_signature = ''
 				for p in f.parameterIterator():
+					if p.is_padding:
+						continue
+
 					# FIXME: This is a *really* ugly hack. :(
 
 					tn = p.type_expr.get_base_type_node()
diff --git a/src/mesa/glapi/glX_proto_recv.py b/src/mesa/glapi/glX_proto_recv.py
index 20f75575cf..923c1958f0 100644
--- a/src/mesa/glapi/glX_proto_recv.py
+++ b/src/mesa/glapi/glX_proto_recv.py
@@ -89,8 +89,8 @@ class PrintGlxDispatchFunctions(glX_proto_common.glx_print_proto):
 		print '#include "glxbyteorder.h"'
 		print '#include "indirect_util.h"'
 		print '#include "singlesize.h"'
-		print '#include "glapitable.h"'
 		print '#include "glapi.h"'
+		print '#include "glapitable.h"'
 		print '#include "glthread.h"'
 		print '#include "dispatch.h"'
 		print ''
@@ -225,6 +225,8 @@ class PrintGlxDispatchFunctions(glX_proto_common.glx_print_proto):
 		list = []
 
 		for param in f.parameterIterator():
+			if param.is_padding:
+				continue
 
 			if param.is_counter or param.is_image() or param.is_output or param.name in f.count_parameter_list or len(param.count_parameter_list):
 				location = param.name
diff --git a/src/mesa/glapi/glX_proto_send.py b/src/mesa/glapi/glX_proto_send.py
index b00b8a1ba6..501706acc7 100644
--- a/src/mesa/glapi/glX_proto_send.py
+++ b/src/mesa/glapi/glX_proto_send.py
@@ -333,7 +333,7 @@ const GLuint __glXDefaultPixelStore[9] = { 0, 0, 0, 0, 0, 0, 0, 0, 1 };
 					if image.img_pad_dimensions:
 						do_it = 1
 						break
-			
+
 
 				if do_it:
 					[h, n] = hash_pixel_function(func)
@@ -422,7 +422,10 @@ generic_%u_byte( GLint rop, const void * ptr )
 		else:
 			src_ptr = "&" + p.name
 
-		if not extra_offset:
+		if p.is_padding:
+			print '(void) memset((void *)(%s + %u), 0, %s);' \
+			    % (pc, p.offset + adjust, p.size_string() )
+		elif not extra_offset:
 			print '(void) memcpy((void *)(%s + %u), (void *)(%s), %s);' \
 			    % (pc, p.offset + adjust, src_ptr, p.size_string() )
 		else:
@@ -472,6 +475,10 @@ generic_%u_byte( GLint rop, const void * ptr )
 				else:
 					dim_str = str(dim)
 
+				if param.is_padding:
+					print '(void) memset((void *)(%s + %u), 0, %s);' \
+					% (pc, (param.offset - 4) + adjust, param.size_string() )
+
 				if param.img_null_flag:
 					if large:
 						print '(void) memcpy((void *)(%s + %u), zero, 4);' % (pc, (param.offset - 4) + adjust)
@@ -739,6 +746,9 @@ generic_%u_byte( GLint rop, const void * ptr )
 
 			p_string = ""
 			for param in f.parameterIterateGlxSend():
+				if param.is_padding:
+					continue
+
 				p_string += ", " + param.name
 
 				if param.is_image():
diff --git a/src/mesa/glapi/glX_proto_size.py b/src/mesa/glapi/glX_proto_size.py
index 2b9a643362..95cb5110cc 100644
--- a/src/mesa/glapi/glX_proto_size.py
+++ b/src/mesa/glapi/glX_proto_size.py
@@ -581,6 +581,11 @@ class PrintGlxReqSize_c(PrintGlxReqSize_common):
 
 		self.common_emit_fixups(fixup)
 
+		if img.img_null_flag:
+			print ''
+			print '	   if (*(CARD32 *) (pc + %s))' % (img.offset - 4)
+			print '	       return 0;'
+
 		print ''
 		print '    return __glXImageSize(%s, %s, %s, %s, %s, %s,' % (img.img_format, img.img_type, img.img_target, w, h, d )
 		print '                          image_height, row_length, skip_images,'
diff --git a/src/mesa/glapi/gl_API.dtd b/src/mesa/glapi/gl_API.dtd
index f89d381866..30c646c924 100644
--- a/src/mesa/glapi/gl_API.dtd
+++ b/src/mesa/glapi/gl_API.dtd
@@ -45,6 +45,7 @@
                    counter             (true | false) "false"
                    count_scale         NMTOKEN "1"
                    output              (true | false) "false"
+                   padding             (true | false) "false"
                    img_width           NMTOKEN #IMPLIED
                    img_height          NMTOKEN #IMPLIED
                    img_depth           NMTOKEN #IMPLIED
diff --git a/src/mesa/glapi/gl_API.xml b/src/mesa/glapi/gl_API.xml
index 6c0367aad7..8212fc2477 100644
--- a/src/mesa/glapi/gl_API.xml
+++ b/src/mesa/glapi/gl_API.xml
@@ -3267,7 +3267,8 @@
         <param name="width" type="GLsizei"/>
         <param name="format" type="GLenum"/>
         <param name="type" type="GLenum"/>
-        <param name="pixels" type="const GLvoid *" img_width="width" img_xoff="xoffset" img_format="format" img_type="type" img_target="target" img_null_flag="true" img_pad_dimensions="true"/>
+        <param name="UNUSED" type="GLuint" padding="true"/>
+        <param name="pixels" type="const GLvoid *" img_width="width" img_xoff="xoffset" img_format="format" img_type="type" img_target="target" img_pad_dimensions="true"/>
         <glx rop="4099" large="true"/>
     </function>
 
@@ -3280,7 +3281,8 @@
         <param name="height" type="GLsizei"/>
         <param name="format" type="GLenum"/>
         <param name="type" type="GLenum"/>
-        <param name="pixels" type="const GLvoid *" img_width="width" img_height="height" img_xoff="xoffset" img_yoff="yoffset" img_format="format" img_type="type" img_target="target" img_null_flag="true" img_pad_dimensions="true"/>
+        <param name="UNUSED" type="GLuint" padding="true"/>
+        <param name="pixels" type="const GLvoid *" img_width="width" img_height="height" img_xoff="xoffset" img_yoff="yoffset" img_format="format" img_type="type" img_target="target" img_pad_dimensions="true"/>
         <glx rop="4100" large="true"/>
     </function>
 
@@ -3994,7 +3996,8 @@
         <param name="depth" type="GLsizei"/>
         <param name="format" type="GLenum"/>
         <param name="type" type="GLenum"/>
-        <param name="pixels" type="const GLvoid *" img_width="width" img_height="height" img_depth="depth" img_xoff="xoffset" img_yoff="yoffset" img_zoff="zoffset" img_format="format" img_type="type" img_target="target" img_null_flag="true" img_pad_dimensions="true"/>
+        <param name="UNUSED" type="GLuint" padding="true"/>
+        <param name="pixels" type="const GLvoid *" img_width="width" img_height="height" img_depth="depth" img_xoff="xoffset" img_yoff="yoffset" img_zoff="zoffset" img_format="format" img_type="type" img_target="target" img_pad_dimensions="true"/>
         <glx rop="4115" large="true"/>
     </function>
 
@@ -7943,6 +7946,9 @@
     </enum>
 </category>
 
+<xi:include href="ARB_framebuffer_object.xml" xmlns:xi="http://www.w3.org/2001/XInclude"/>
+
+
 <!-- Non-ARB extensions sorted by extension number. -->
 
 <category name="GL_EXT_blend_color" number="2">
@@ -8061,6 +8067,7 @@
         <param name="depth" type="GLsizei"/>
         <param name="format" type="GLenum"/>
         <param name="type" type="GLenum"/>
+        <param name="UNUSED" type="GLuint" padding="true"/>
         <param name="pixels" type="const GLvoid *"/>
     </function>
 </category>
@@ -8092,6 +8099,7 @@
         <param name="width" type="GLsizei"/>
         <param name="format" type="GLenum"/>
         <param name="type" type="GLenum"/>
+        <param name="UNUSED" type="GLuint" padding="true"/>
         <param name="pixels" type="const GLvoid *"/>
     </function>
 
@@ -8104,6 +8112,7 @@
         <param name="height" type="GLsizei"/>
         <param name="format" type="GLenum"/>
         <param name="type" type="GLenum"/>
+        <param name="UNUSED" type="GLuint" padding="true"/>
         <param name="pixels" type="const GLvoid *"/>
     </function>
 </category>
@@ -8627,7 +8636,8 @@
         <param name="size4d" type="GLsizei"/>
         <param name="format" type="GLenum"/>
         <param name="type" type="GLenum"/>
-        <param name="pixels" type="const GLvoid *" img_width="width" img_height="height" img_depth="depth" img_extent="size4d" img_xoff="xoffset" img_yoff="yoffset" img_zoff="zoffset" img_woff="woffset" img_format="format" img_type="type" img_target="target" img_null_flag="true" img_pad_dimensions="true"/>
+        <param name="UNUSED" type="GLuint" padding="true"/>
+        <param name="pixels" type="const GLvoid *" img_width="width" img_height="height" img_depth="depth" img_extent="size4d" img_xoff="xoffset" img_yoff="yoffset" img_zoff="zoffset" img_woff="woffset" img_format="format" img_type="type" img_target="target" img_pad_dimensions="true"/>
         <glx rop="2058" ignore="true"/>
     </function>
 </category>
diff --git a/src/mesa/glapi/gl_XML.py b/src/mesa/glapi/gl_XML.py
index b7a7388400..b98919134f 100644
--- a/src/mesa/glapi/gl_XML.py
+++ b/src/mesa/glapi/gl_XML.py
@@ -309,6 +309,9 @@ def create_parameter_string(parameters, include_names):
 
 	list = []
 	for p in parameters:
+		if p.is_padding:
+			continue
+
 		if include_names:
 			list.append( p.string() )
 		else:
@@ -463,6 +466,7 @@ class gl_parameter:
 		self.img_null_flag      = is_attr_true( element, 'img_null_flag' )
 		self.img_send_null      = is_attr_true( element, 'img_send_null' )
 
+		self.is_padding = is_attr_true( element, 'padding' )
 		return
 
 
diff --git a/src/mesa/glapi/gl_apitemp.py b/src/mesa/glapi/gl_apitemp.py
index 6e35571e14..a37c08d6ce 100644
--- a/src/mesa/glapi/gl_apitemp.py
+++ b/src/mesa/glapi/gl_apitemp.py
@@ -63,6 +63,9 @@ class PrintGlOffsets(gl_XML.gl_print_base):
 		n = f.static_name(name)
 
 		for p in f.parameterIterator():
+			if p.is_padding:
+				continue
+
 			if p.is_pointer():
 				cast = "(const void *) "
 			else:
diff --git a/src/mesa/glapi/gl_x86_asm.py b/src/mesa/glapi/gl_x86_asm.py
index 651cb03f14..0dbf3ebe0a 100644
--- a/src/mesa/glapi/gl_x86_asm.py
+++ b/src/mesa/glapi/gl_x86_asm.py
@@ -44,6 +44,9 @@ class PrintGenericStubs(gl_XML.gl_print_base):
 	def get_stack_size(self, f):
 		size = 0
 		for p in f.parameterIterator():
+			if p.is_padding:
+				continue
+
 			size += p.get_stack_size()
 
 		return size
diff --git a/src/mesa/glapi/glapi.c b/src/mesa/glapi/glapi.c
index c3ebf60719..9b5144a88b 100644
--- a/src/mesa/glapi/glapi.c
+++ b/src/mesa/glapi/glapi.c
@@ -98,6 +98,7 @@ _glapi_set_warning_func( _glapi_warning_func func )
 static GLboolean
 warn(void)
 {
+#if !defined(_WIN32_WCE)
    if ((WarnFlag || getenv("MESA_DEBUG") || getenv("LIBGL_DEBUG"))
        && warning_func) {
       return GL_TRUE;
@@ -105,6 +106,9 @@ warn(void)
    else {
       return GL_FALSE;
    }
+#else
+   return GL_FALSE;
+#endif
 }
 
 
diff --git a/src/mesa/glapi/glapioffsets.h b/src/mesa/glapi/glapioffsets.h
index 43c265924b..f412234ac5 100644
--- a/src/mesa/glapi/glapioffsets.h
+++ b/src/mesa/glapi/glapioffsets.h
@@ -593,216 +593,217 @@
 #define _gloffset_GetActiveAttribARB 558
 #define _gloffset_GetAttribLocationARB 559
 #define _gloffset_DrawBuffersARB 560
-#define _gloffset_PolygonOffsetEXT 561
-#define _gloffset_GetPixelTexGenParameterfvSGIS 562
-#define _gloffset_GetPixelTexGenParameterivSGIS 563
-#define _gloffset_PixelTexGenParameterfSGIS 564
-#define _gloffset_PixelTexGenParameterfvSGIS 565
-#define _gloffset_PixelTexGenParameteriSGIS 566
-#define _gloffset_PixelTexGenParameterivSGIS 567
-#define _gloffset_SampleMaskSGIS 568
-#define _gloffset_SamplePatternSGIS 569
-#define _gloffset_ColorPointerEXT 570
-#define _gloffset_EdgeFlagPointerEXT 571
-#define _gloffset_IndexPointerEXT 572
-#define _gloffset_NormalPointerEXT 573
-#define _gloffset_TexCoordPointerEXT 574
-#define _gloffset_VertexPointerEXT 575
-#define _gloffset_PointParameterfEXT 576
-#define _gloffset_PointParameterfvEXT 577
-#define _gloffset_LockArraysEXT 578
-#define _gloffset_UnlockArraysEXT 579
-#define _gloffset_CullParameterdvEXT 580
-#define _gloffset_CullParameterfvEXT 581
-#define _gloffset_SecondaryColor3bEXT 582
-#define _gloffset_SecondaryColor3bvEXT 583
-#define _gloffset_SecondaryColor3dEXT 584
-#define _gloffset_SecondaryColor3dvEXT 585
-#define _gloffset_SecondaryColor3fEXT 586
-#define _gloffset_SecondaryColor3fvEXT 587
-#define _gloffset_SecondaryColor3iEXT 588
-#define _gloffset_SecondaryColor3ivEXT 589
-#define _gloffset_SecondaryColor3sEXT 590
-#define _gloffset_SecondaryColor3svEXT 591
-#define _gloffset_SecondaryColor3ubEXT 592
-#define _gloffset_SecondaryColor3ubvEXT 593
-#define _gloffset_SecondaryColor3uiEXT 594
-#define _gloffset_SecondaryColor3uivEXT 595
-#define _gloffset_SecondaryColor3usEXT 596
-#define _gloffset_SecondaryColor3usvEXT 597
-#define _gloffset_SecondaryColorPointerEXT 598
-#define _gloffset_MultiDrawArraysEXT 599
-#define _gloffset_MultiDrawElementsEXT 600
-#define _gloffset_FogCoordPointerEXT 601
-#define _gloffset_FogCoorddEXT 602
-#define _gloffset_FogCoorddvEXT 603
-#define _gloffset_FogCoordfEXT 604
-#define _gloffset_FogCoordfvEXT 605
-#define _gloffset_PixelTexGenSGIX 606
-#define _gloffset_BlendFuncSeparateEXT 607
-#define _gloffset_FlushVertexArrayRangeNV 608
-#define _gloffset_VertexArrayRangeNV 609
-#define _gloffset_CombinerInputNV 610
-#define _gloffset_CombinerOutputNV 611
-#define _gloffset_CombinerParameterfNV 612
-#define _gloffset_CombinerParameterfvNV 613
-#define _gloffset_CombinerParameteriNV 614
-#define _gloffset_CombinerParameterivNV 615
-#define _gloffset_FinalCombinerInputNV 616
-#define _gloffset_GetCombinerInputParameterfvNV 617
-#define _gloffset_GetCombinerInputParameterivNV 618
-#define _gloffset_GetCombinerOutputParameterfvNV 619
-#define _gloffset_GetCombinerOutputParameterivNV 620
-#define _gloffset_GetFinalCombinerInputParameterfvNV 621
-#define _gloffset_GetFinalCombinerInputParameterivNV 622
-#define _gloffset_ResizeBuffersMESA 623
-#define _gloffset_WindowPos2dMESA 624
-#define _gloffset_WindowPos2dvMESA 625
-#define _gloffset_WindowPos2fMESA 626
-#define _gloffset_WindowPos2fvMESA 627
-#define _gloffset_WindowPos2iMESA 628
-#define _gloffset_WindowPos2ivMESA 629
-#define _gloffset_WindowPos2sMESA 630
-#define _gloffset_WindowPos2svMESA 631
-#define _gloffset_WindowPos3dMESA 632
-#define _gloffset_WindowPos3dvMESA 633
-#define _gloffset_WindowPos3fMESA 634
-#define _gloffset_WindowPos3fvMESA 635
-#define _gloffset_WindowPos3iMESA 636
-#define _gloffset_WindowPos3ivMESA 637
-#define _gloffset_WindowPos3sMESA 638
-#define _gloffset_WindowPos3svMESA 639
-#define _gloffset_WindowPos4dMESA 640
-#define _gloffset_WindowPos4dvMESA 641
-#define _gloffset_WindowPos4fMESA 642
-#define _gloffset_WindowPos4fvMESA 643
-#define _gloffset_WindowPos4iMESA 644
-#define _gloffset_WindowPos4ivMESA 645
-#define _gloffset_WindowPos4sMESA 646
-#define _gloffset_WindowPos4svMESA 647
-#define _gloffset_MultiModeDrawArraysIBM 648
-#define _gloffset_MultiModeDrawElementsIBM 649
-#define _gloffset_DeleteFencesNV 650
-#define _gloffset_FinishFenceNV 651
-#define _gloffset_GenFencesNV 652
-#define _gloffset_GetFenceivNV 653
-#define _gloffset_IsFenceNV 654
-#define _gloffset_SetFenceNV 655
-#define _gloffset_TestFenceNV 656
-#define _gloffset_AreProgramsResidentNV 657
-#define _gloffset_BindProgramNV 658
-#define _gloffset_DeleteProgramsNV 659
-#define _gloffset_ExecuteProgramNV 660
-#define _gloffset_GenProgramsNV 661
-#define _gloffset_GetProgramParameterdvNV 662
-#define _gloffset_GetProgramParameterfvNV 663
-#define _gloffset_GetProgramStringNV 664
-#define _gloffset_GetProgramivNV 665
-#define _gloffset_GetTrackMatrixivNV 666
-#define _gloffset_GetVertexAttribPointervNV 667
-#define _gloffset_GetVertexAttribdvNV 668
-#define _gloffset_GetVertexAttribfvNV 669
-#define _gloffset_GetVertexAttribivNV 670
-#define _gloffset_IsProgramNV 671
-#define _gloffset_LoadProgramNV 672
-#define _gloffset_ProgramParameters4dvNV 673
-#define _gloffset_ProgramParameters4fvNV 674
-#define _gloffset_RequestResidentProgramsNV 675
-#define _gloffset_TrackMatrixNV 676
-#define _gloffset_VertexAttrib1dNV 677
-#define _gloffset_VertexAttrib1dvNV 678
-#define _gloffset_VertexAttrib1fNV 679
-#define _gloffset_VertexAttrib1fvNV 680
-#define _gloffset_VertexAttrib1sNV 681
-#define _gloffset_VertexAttrib1svNV 682
-#define _gloffset_VertexAttrib2dNV 683
-#define _gloffset_VertexAttrib2dvNV 684
-#define _gloffset_VertexAttrib2fNV 685
-#define _gloffset_VertexAttrib2fvNV 686
-#define _gloffset_VertexAttrib2sNV 687
-#define _gloffset_VertexAttrib2svNV 688
-#define _gloffset_VertexAttrib3dNV 689
-#define _gloffset_VertexAttrib3dvNV 690
-#define _gloffset_VertexAttrib3fNV 691
-#define _gloffset_VertexAttrib3fvNV 692
-#define _gloffset_VertexAttrib3sNV 693
-#define _gloffset_VertexAttrib3svNV 694
-#define _gloffset_VertexAttrib4dNV 695
-#define _gloffset_VertexAttrib4dvNV 696
-#define _gloffset_VertexAttrib4fNV 697
-#define _gloffset_VertexAttrib4fvNV 698
-#define _gloffset_VertexAttrib4sNV 699
-#define _gloffset_VertexAttrib4svNV 700
-#define _gloffset_VertexAttrib4ubNV 701
-#define _gloffset_VertexAttrib4ubvNV 702
-#define _gloffset_VertexAttribPointerNV 703
-#define _gloffset_VertexAttribs1dvNV 704
-#define _gloffset_VertexAttribs1fvNV 705
-#define _gloffset_VertexAttribs1svNV 706
-#define _gloffset_VertexAttribs2dvNV 707
-#define _gloffset_VertexAttribs2fvNV 708
-#define _gloffset_VertexAttribs2svNV 709
-#define _gloffset_VertexAttribs3dvNV 710
-#define _gloffset_VertexAttribs3fvNV 711
-#define _gloffset_VertexAttribs3svNV 712
-#define _gloffset_VertexAttribs4dvNV 713
-#define _gloffset_VertexAttribs4fvNV 714
-#define _gloffset_VertexAttribs4svNV 715
-#define _gloffset_VertexAttribs4ubvNV 716
-#define _gloffset_AlphaFragmentOp1ATI 717
-#define _gloffset_AlphaFragmentOp2ATI 718
-#define _gloffset_AlphaFragmentOp3ATI 719
-#define _gloffset_BeginFragmentShaderATI 720
-#define _gloffset_BindFragmentShaderATI 721
-#define _gloffset_ColorFragmentOp1ATI 722
-#define _gloffset_ColorFragmentOp2ATI 723
-#define _gloffset_ColorFragmentOp3ATI 724
-#define _gloffset_DeleteFragmentShaderATI 725
-#define _gloffset_EndFragmentShaderATI 726
-#define _gloffset_GenFragmentShadersATI 727
-#define _gloffset_PassTexCoordATI 728
-#define _gloffset_SampleMapATI 729
-#define _gloffset_SetFragmentShaderConstantATI 730
-#define _gloffset_PointParameteriNV 731
-#define _gloffset_PointParameterivNV 732
-#define _gloffset_ActiveStencilFaceEXT 733
-#define _gloffset_BindVertexArrayAPPLE 734
-#define _gloffset_DeleteVertexArraysAPPLE 735
-#define _gloffset_GenVertexArraysAPPLE 736
-#define _gloffset_IsVertexArrayAPPLE 737
-#define _gloffset_GetProgramNamedParameterdvNV 738
-#define _gloffset_GetProgramNamedParameterfvNV 739
-#define _gloffset_ProgramNamedParameter4dNV 740
-#define _gloffset_ProgramNamedParameter4dvNV 741
-#define _gloffset_ProgramNamedParameter4fNV 742
-#define _gloffset_ProgramNamedParameter4fvNV 743
-#define _gloffset_DepthBoundsEXT 744
-#define _gloffset_BlendEquationSeparateEXT 745
-#define _gloffset_BindFramebufferEXT 746
-#define _gloffset_BindRenderbufferEXT 747
-#define _gloffset_CheckFramebufferStatusEXT 748
-#define _gloffset_DeleteFramebuffersEXT 749
-#define _gloffset_DeleteRenderbuffersEXT 750
-#define _gloffset_FramebufferRenderbufferEXT 751
-#define _gloffset_FramebufferTexture1DEXT 752
-#define _gloffset_FramebufferTexture2DEXT 753
-#define _gloffset_FramebufferTexture3DEXT 754
-#define _gloffset_GenFramebuffersEXT 755
-#define _gloffset_GenRenderbuffersEXT 756
-#define _gloffset_GenerateMipmapEXT 757
-#define _gloffset_GetFramebufferAttachmentParameterivEXT 758
-#define _gloffset_GetRenderbufferParameterivEXT 759
-#define _gloffset_IsFramebufferEXT 760
-#define _gloffset_IsRenderbufferEXT 761
-#define _gloffset_RenderbufferStorageEXT 762
-#define _gloffset_BlitFramebufferEXT 763
-#define _gloffset_FramebufferTextureLayerEXT 764
-#define _gloffset_StencilFuncSeparateATI 765
-#define _gloffset_ProgramEnvParameters4fvEXT 766
-#define _gloffset_ProgramLocalParameters4fvEXT 767
-#define _gloffset_GetQueryObjecti64vEXT 768
-#define _gloffset_GetQueryObjectui64vEXT 769
-#define _gloffset_FIRST_DYNAMIC 770
+#define _gloffset_RenderbufferStorageMultisample 561
+#define _gloffset_PolygonOffsetEXT 562
+#define _gloffset_GetPixelTexGenParameterfvSGIS 563
+#define _gloffset_GetPixelTexGenParameterivSGIS 564
+#define _gloffset_PixelTexGenParameterfSGIS 565
+#define _gloffset_PixelTexGenParameterfvSGIS 566
+#define _gloffset_PixelTexGenParameteriSGIS 567
+#define _gloffset_PixelTexGenParameterivSGIS 568
+#define _gloffset_SampleMaskSGIS 569
+#define _gloffset_SamplePatternSGIS 570
+#define _gloffset_ColorPointerEXT 571
+#define _gloffset_EdgeFlagPointerEXT 572
+#define _gloffset_IndexPointerEXT 573
+#define _gloffset_NormalPointerEXT 574
+#define _gloffset_TexCoordPointerEXT 575
+#define _gloffset_VertexPointerEXT 576
+#define _gloffset_PointParameterfEXT 577
+#define _gloffset_PointParameterfvEXT 578
+#define _gloffset_LockArraysEXT 579
+#define _gloffset_UnlockArraysEXT 580
+#define _gloffset_CullParameterdvEXT 581
+#define _gloffset_CullParameterfvEXT 582
+#define _gloffset_SecondaryColor3bEXT 583
+#define _gloffset_SecondaryColor3bvEXT 584
+#define _gloffset_SecondaryColor3dEXT 585
+#define _gloffset_SecondaryColor3dvEXT 586
+#define _gloffset_SecondaryColor3fEXT 587
+#define _gloffset_SecondaryColor3fvEXT 588
+#define _gloffset_SecondaryColor3iEXT 589
+#define _gloffset_SecondaryColor3ivEXT 590
+#define _gloffset_SecondaryColor3sEXT 591
+#define _gloffset_SecondaryColor3svEXT 592
+#define _gloffset_SecondaryColor3ubEXT 593
+#define _gloffset_SecondaryColor3ubvEXT 594
+#define _gloffset_SecondaryColor3uiEXT 595
+#define _gloffset_SecondaryColor3uivEXT 596
+#define _gloffset_SecondaryColor3usEXT 597
+#define _gloffset_SecondaryColor3usvEXT 598
+#define _gloffset_SecondaryColorPointerEXT 599
+#define _gloffset_MultiDrawArraysEXT 600
+#define _gloffset_MultiDrawElementsEXT 601
+#define _gloffset_FogCoordPointerEXT 602
+#define _gloffset_FogCoorddEXT 603
+#define _gloffset_FogCoorddvEXT 604
+#define _gloffset_FogCoordfEXT 605
+#define _gloffset_FogCoordfvEXT 606
+#define _gloffset_PixelTexGenSGIX 607
+#define _gloffset_BlendFuncSeparateEXT 608
+#define _gloffset_FlushVertexArrayRangeNV 609
+#define _gloffset_VertexArrayRangeNV 610
+#define _gloffset_CombinerInputNV 611
+#define _gloffset_CombinerOutputNV 612
+#define _gloffset_CombinerParameterfNV 613
+#define _gloffset_CombinerParameterfvNV 614
+#define _gloffset_CombinerParameteriNV 615
+#define _gloffset_CombinerParameterivNV 616
+#define _gloffset_FinalCombinerInputNV 617
+#define _gloffset_GetCombinerInputParameterfvNV 618
+#define _gloffset_GetCombinerInputParameterivNV 619
+#define _gloffset_GetCombinerOutputParameterfvNV 620
+#define _gloffset_GetCombinerOutputParameterivNV 621
+#define _gloffset_GetFinalCombinerInputParameterfvNV 622
+#define _gloffset_GetFinalCombinerInputParameterivNV 623
+#define _gloffset_ResizeBuffersMESA 624
+#define _gloffset_WindowPos2dMESA 625
+#define _gloffset_WindowPos2dvMESA 626
+#define _gloffset_WindowPos2fMESA 627
+#define _gloffset_WindowPos2fvMESA 628
+#define _gloffset_WindowPos2iMESA 629
+#define _gloffset_WindowPos2ivMESA 630
+#define _gloffset_WindowPos2sMESA 631
+#define _gloffset_WindowPos2svMESA 632
+#define _gloffset_WindowPos3dMESA 633
+#define _gloffset_WindowPos3dvMESA 634
+#define _gloffset_WindowPos3fMESA 635
+#define _gloffset_WindowPos3fvMESA 636
+#define _gloffset_WindowPos3iMESA 637
+#define _gloffset_WindowPos3ivMESA 638
+#define _gloffset_WindowPos3sMESA 639
+#define _gloffset_WindowPos3svMESA 640
+#define _gloffset_WindowPos4dMESA 641
+#define _gloffset_WindowPos4dvMESA 642
+#define _gloffset_WindowPos4fMESA 643
+#define _gloffset_WindowPos4fvMESA 644
+#define _gloffset_WindowPos4iMESA 645
+#define _gloffset_WindowPos4ivMESA 646
+#define _gloffset_WindowPos4sMESA 647
+#define _gloffset_WindowPos4svMESA 648
+#define _gloffset_MultiModeDrawArraysIBM 649
+#define _gloffset_MultiModeDrawElementsIBM 650
+#define _gloffset_DeleteFencesNV 651
+#define _gloffset_FinishFenceNV 652
+#define _gloffset_GenFencesNV 653
+#define _gloffset_GetFenceivNV 654
+#define _gloffset_IsFenceNV 655
+#define _gloffset_SetFenceNV 656
+#define _gloffset_TestFenceNV 657
+#define _gloffset_AreProgramsResidentNV 658
+#define _gloffset_BindProgramNV 659
+#define _gloffset_DeleteProgramsNV 660
+#define _gloffset_ExecuteProgramNV 661
+#define _gloffset_GenProgramsNV 662
+#define _gloffset_GetProgramParameterdvNV 663
+#define _gloffset_GetProgramParameterfvNV 664
+#define _gloffset_GetProgramStringNV 665
+#define _gloffset_GetProgramivNV 666
+#define _gloffset_GetTrackMatrixivNV 667
+#define _gloffset_GetVertexAttribPointervNV 668
+#define _gloffset_GetVertexAttribdvNV 669
+#define _gloffset_GetVertexAttribfvNV 670
+#define _gloffset_GetVertexAttribivNV 671
+#define _gloffset_IsProgramNV 672
+#define _gloffset_LoadProgramNV 673
+#define _gloffset_ProgramParameters4dvNV 674
+#define _gloffset_ProgramParameters4fvNV 675
+#define _gloffset_RequestResidentProgramsNV 676
+#define _gloffset_TrackMatrixNV 677
+#define _gloffset_VertexAttrib1dNV 678
+#define _gloffset_VertexAttrib1dvNV 679
+#define _gloffset_VertexAttrib1fNV 680
+#define _gloffset_VertexAttrib1fvNV 681
+#define _gloffset_VertexAttrib1sNV 682
+#define _gloffset_VertexAttrib1svNV 683
+#define _gloffset_VertexAttrib2dNV 684
+#define _gloffset_VertexAttrib2dvNV 685
+#define _gloffset_VertexAttrib2fNV 686
+#define _gloffset_VertexAttrib2fvNV 687
+#define _gloffset_VertexAttrib2sNV 688
+#define _gloffset_VertexAttrib2svNV 689
+#define _gloffset_VertexAttrib3dNV 690
+#define _gloffset_VertexAttrib3dvNV 691
+#define _gloffset_VertexAttrib3fNV 692
+#define _gloffset_VertexAttrib3fvNV 693
+#define _gloffset_VertexAttrib3sNV 694
+#define _gloffset_VertexAttrib3svNV 695
+#define _gloffset_VertexAttrib4dNV 696
+#define _gloffset_VertexAttrib4dvNV 697
+#define _gloffset_VertexAttrib4fNV 698
+#define _gloffset_VertexAttrib4fvNV 699
+#define _gloffset_VertexAttrib4sNV 700
+#define _gloffset_VertexAttrib4svNV 701
+#define _gloffset_VertexAttrib4ubNV 702
+#define _gloffset_VertexAttrib4ubvNV 703
+#define _gloffset_VertexAttribPointerNV 704
+#define _gloffset_VertexAttribs1dvNV 705
+#define _gloffset_VertexAttribs1fvNV 706
+#define _gloffset_VertexAttribs1svNV 707
+#define _gloffset_VertexAttribs2dvNV 708
+#define _gloffset_VertexAttribs2fvNV 709
+#define _gloffset_VertexAttribs2svNV 710
+#define _gloffset_VertexAttribs3dvNV 711
+#define _gloffset_VertexAttribs3fvNV 712
+#define _gloffset_VertexAttribs3svNV 713
+#define _gloffset_VertexAttribs4dvNV 714
+#define _gloffset_VertexAttribs4fvNV 715
+#define _gloffset_VertexAttribs4svNV 716
+#define _gloffset_VertexAttribs4ubvNV 717
+#define _gloffset_AlphaFragmentOp1ATI 718
+#define _gloffset_AlphaFragmentOp2ATI 719
+#define _gloffset_AlphaFragmentOp3ATI 720
+#define _gloffset_BeginFragmentShaderATI 721
+#define _gloffset_BindFragmentShaderATI 722
+#define _gloffset_ColorFragmentOp1ATI 723
+#define _gloffset_ColorFragmentOp2ATI 724
+#define _gloffset_ColorFragmentOp3ATI 725
+#define _gloffset_DeleteFragmentShaderATI 726
+#define _gloffset_EndFragmentShaderATI 727
+#define _gloffset_GenFragmentShadersATI 728
+#define _gloffset_PassTexCoordATI 729
+#define _gloffset_SampleMapATI 730
+#define _gloffset_SetFragmentShaderConstantATI 731
+#define _gloffset_PointParameteriNV 732
+#define _gloffset_PointParameterivNV 733
+#define _gloffset_ActiveStencilFaceEXT 734
+#define _gloffset_BindVertexArrayAPPLE 735
+#define _gloffset_DeleteVertexArraysAPPLE 736
+#define _gloffset_GenVertexArraysAPPLE 737
+#define _gloffset_IsVertexArrayAPPLE 738
+#define _gloffset_GetProgramNamedParameterdvNV 739
+#define _gloffset_GetProgramNamedParameterfvNV 740
+#define _gloffset_ProgramNamedParameter4dNV 741
+#define _gloffset_ProgramNamedParameter4dvNV 742
+#define _gloffset_ProgramNamedParameter4fNV 743
+#define _gloffset_ProgramNamedParameter4fvNV 744
+#define _gloffset_DepthBoundsEXT 745
+#define _gloffset_BlendEquationSeparateEXT 746
+#define _gloffset_BindFramebufferEXT 747
+#define _gloffset_BindRenderbufferEXT 748
+#define _gloffset_CheckFramebufferStatusEXT 749
+#define _gloffset_DeleteFramebuffersEXT 750
+#define _gloffset_DeleteRenderbuffersEXT 751
+#define _gloffset_FramebufferRenderbufferEXT 752
+#define _gloffset_FramebufferTexture1DEXT 753
+#define _gloffset_FramebufferTexture2DEXT 754
+#define _gloffset_FramebufferTexture3DEXT 755
+#define _gloffset_GenFramebuffersEXT 756
+#define _gloffset_GenRenderbuffersEXT 757
+#define _gloffset_GenerateMipmapEXT 758
+#define _gloffset_GetFramebufferAttachmentParameterivEXT 759
+#define _gloffset_GetRenderbufferParameterivEXT 760
+#define _gloffset_IsFramebufferEXT 761
+#define _gloffset_IsRenderbufferEXT 762
+#define _gloffset_RenderbufferStorageEXT 763
+#define _gloffset_BlitFramebufferEXT 764
+#define _gloffset_FramebufferTextureLayerEXT 765
+#define _gloffset_StencilFuncSeparateATI 766
+#define _gloffset_ProgramEnvParameters4fvEXT 767
+#define _gloffset_ProgramLocalParameters4fvEXT 768
+#define _gloffset_GetQueryObjecti64vEXT 769
+#define _gloffset_GetQueryObjectui64vEXT 770
+#define _gloffset_FIRST_DYNAMIC 771
 
 #else
 
@@ -959,6 +960,7 @@
 #define _gloffset_GetActiveAttribARB driDispatchRemapTable[GetActiveAttribARB_remap_index]
 #define _gloffset_GetAttribLocationARB driDispatchRemapTable[GetAttribLocationARB_remap_index]
 #define _gloffset_DrawBuffersARB driDispatchRemapTable[DrawBuffersARB_remap_index]
+#define _gloffset_RenderbufferStorageMultisample driDispatchRemapTable[RenderbufferStorageMultisample_remap_index]
 #define _gloffset_PolygonOffsetEXT driDispatchRemapTable[PolygonOffsetEXT_remap_index]
 #define _gloffset_GetPixelTexGenParameterfvSGIS driDispatchRemapTable[GetPixelTexGenParameterfvSGIS_remap_index]
 #define _gloffset_GetPixelTexGenParameterivSGIS driDispatchRemapTable[GetPixelTexGenParameterivSGIS_remap_index]
diff --git a/src/mesa/glapi/glapitable.h b/src/mesa/glapi/glapitable.h
index 5d9d40a8a2..2ae9bf6738 100644
--- a/src/mesa/glapi/glapitable.h
+++ b/src/mesa/glapi/glapitable.h
@@ -601,215 +601,216 @@ struct _glapi_table
    void (GLAPIENTRYP GetActiveAttribARB)(GLhandleARB program, GLuint index, GLsizei bufSize, GLsizei * length, GLint * size, GLenum * type, GLcharARB * name); /* 558 */
    GLint (GLAPIENTRYP GetAttribLocationARB)(GLhandleARB program, const GLcharARB * name); /* 559 */
    void (GLAPIENTRYP DrawBuffersARB)(GLsizei n, const GLenum * bufs); /* 560 */
-   void (GLAPIENTRYP PolygonOffsetEXT)(GLfloat factor, GLfloat bias); /* 561 */
-   void (GLAPIENTRYP GetPixelTexGenParameterfvSGIS)(GLenum pname, GLfloat * params); /* 562 */
-   void (GLAPIENTRYP GetPixelTexGenParameterivSGIS)(GLenum pname, GLint * params); /* 563 */
-   void (GLAPIENTRYP PixelTexGenParameterfSGIS)(GLenum pname, GLfloat param); /* 564 */
-   void (GLAPIENTRYP PixelTexGenParameterfvSGIS)(GLenum pname, const GLfloat * params); /* 565 */
-   void (GLAPIENTRYP PixelTexGenParameteriSGIS)(GLenum pname, GLint param); /* 566 */
-   void (GLAPIENTRYP PixelTexGenParameterivSGIS)(GLenum pname, const GLint * params); /* 567 */
-   void (GLAPIENTRYP SampleMaskSGIS)(GLclampf value, GLboolean invert); /* 568 */
-   void (GLAPIENTRYP SamplePatternSGIS)(GLenum pattern); /* 569 */
-   void (GLAPIENTRYP ColorPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 570 */
-   void (GLAPIENTRYP EdgeFlagPointerEXT)(GLsizei stride, GLsizei count, const GLboolean * pointer); /* 571 */
-   void (GLAPIENTRYP IndexPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 572 */
-   void (GLAPIENTRYP NormalPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 573 */
-   void (GLAPIENTRYP TexCoordPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 574 */
-   void (GLAPIENTRYP VertexPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 575 */
-   void (GLAPIENTRYP PointParameterfEXT)(GLenum pname, GLfloat param); /* 576 */
-   void (GLAPIENTRYP PointParameterfvEXT)(GLenum pname, const GLfloat * params); /* 577 */
-   void (GLAPIENTRYP LockArraysEXT)(GLint first, GLsizei count); /* 578 */
-   void (GLAPIENTRYP UnlockArraysEXT)(void); /* 579 */
-   void (GLAPIENTRYP CullParameterdvEXT)(GLenum pname, GLdouble * params); /* 580 */
-   void (GLAPIENTRYP CullParameterfvEXT)(GLenum pname, GLfloat * params); /* 581 */
-   void (GLAPIENTRYP SecondaryColor3bEXT)(GLbyte red, GLbyte green, GLbyte blue); /* 582 */
-   void (GLAPIENTRYP SecondaryColor3bvEXT)(const GLbyte * v); /* 583 */
-   void (GLAPIENTRYP SecondaryColor3dEXT)(GLdouble red, GLdouble green, GLdouble blue); /* 584 */
-   void (GLAPIENTRYP SecondaryColor3dvEXT)(const GLdouble * v); /* 585 */
-   void (GLAPIENTRYP SecondaryColor3fEXT)(GLfloat red, GLfloat green, GLfloat blue); /* 586 */
-   void (GLAPIENTRYP SecondaryColor3fvEXT)(const GLfloat * v); /* 587 */
-   void (GLAPIENTRYP SecondaryColor3iEXT)(GLint red, GLint green, GLint blue); /* 588 */
-   void (GLAPIENTRYP SecondaryColor3ivEXT)(const GLint * v); /* 589 */
-   void (GLAPIENTRYP SecondaryColor3sEXT)(GLshort red, GLshort green, GLshort blue); /* 590 */
-   void (GLAPIENTRYP SecondaryColor3svEXT)(const GLshort * v); /* 591 */
-   void (GLAPIENTRYP SecondaryColor3ubEXT)(GLubyte red, GLubyte green, GLubyte blue); /* 592 */
-   void (GLAPIENTRYP SecondaryColor3ubvEXT)(const GLubyte * v); /* 593 */
-   void (GLAPIENTRYP SecondaryColor3uiEXT)(GLuint red, GLuint green, GLuint blue); /* 594 */
-   void (GLAPIENTRYP SecondaryColor3uivEXT)(const GLuint * v); /* 595 */
-   void (GLAPIENTRYP SecondaryColor3usEXT)(GLushort red, GLushort green, GLushort blue); /* 596 */
-   void (GLAPIENTRYP SecondaryColor3usvEXT)(const GLushort * v); /* 597 */
-   void (GLAPIENTRYP SecondaryColorPointerEXT)(GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 598 */
-   void (GLAPIENTRYP MultiDrawArraysEXT)(GLenum mode, GLint * first, GLsizei * count, GLsizei primcount); /* 599 */
-   void (GLAPIENTRYP MultiDrawElementsEXT)(GLenum mode, const GLsizei * count, GLenum type, const GLvoid ** indices, GLsizei primcount); /* 600 */
-   void (GLAPIENTRYP FogCoordPointerEXT)(GLenum type, GLsizei stride, const GLvoid * pointer); /* 601 */
-   void (GLAPIENTRYP FogCoorddEXT)(GLdouble coord); /* 602 */
-   void (GLAPIENTRYP FogCoorddvEXT)(const GLdouble * coord); /* 603 */
-   void (GLAPIENTRYP FogCoordfEXT)(GLfloat coord); /* 604 */
-   void (GLAPIENTRYP FogCoordfvEXT)(const GLfloat * coord); /* 605 */
-   void (GLAPIENTRYP PixelTexGenSGIX)(GLenum mode); /* 606 */
-   void (GLAPIENTRYP BlendFuncSeparateEXT)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha); /* 607 */
-   void (GLAPIENTRYP FlushVertexArrayRangeNV)(void); /* 608 */
-   void (GLAPIENTRYP VertexArrayRangeNV)(GLsizei length, const GLvoid * pointer); /* 609 */
-   void (GLAPIENTRYP CombinerInputNV)(GLenum stage, GLenum portion, GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 610 */
-   void (GLAPIENTRYP CombinerOutputNV)(GLenum stage, GLenum portion, GLenum abOutput, GLenum cdOutput, GLenum sumOutput, GLenum scale, GLenum bias, GLboolean abDotProduct, GLboolean cdDotProduct, GLboolean muxSum); /* 611 */
-   void (GLAPIENTRYP CombinerParameterfNV)(GLenum pname, GLfloat param); /* 612 */
-   void (GLAPIENTRYP CombinerParameterfvNV)(GLenum pname, const GLfloat * params); /* 613 */
-   void (GLAPIENTRYP CombinerParameteriNV)(GLenum pname, GLint param); /* 614 */
-   void (GLAPIENTRYP CombinerParameterivNV)(GLenum pname, const GLint * params); /* 615 */
-   void (GLAPIENTRYP FinalCombinerInputNV)(GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 616 */
-   void (GLAPIENTRYP GetCombinerInputParameterfvNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLfloat * params); /* 617 */
-   void (GLAPIENTRYP GetCombinerInputParameterivNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLint * params); /* 618 */
-   void (GLAPIENTRYP GetCombinerOutputParameterfvNV)(GLenum stage, GLenum portion, GLenum pname, GLfloat * params); /* 619 */
-   void (GLAPIENTRYP GetCombinerOutputParameterivNV)(GLenum stage, GLenum portion, GLenum pname, GLint * params); /* 620 */
-   void (GLAPIENTRYP GetFinalCombinerInputParameterfvNV)(GLenum variable, GLenum pname, GLfloat * params); /* 621 */
-   void (GLAPIENTRYP GetFinalCombinerInputParameterivNV)(GLenum variable, GLenum pname, GLint * params); /* 622 */
-   void (GLAPIENTRYP ResizeBuffersMESA)(void); /* 623 */
-   void (GLAPIENTRYP WindowPos2dMESA)(GLdouble x, GLdouble y); /* 624 */
-   void (GLAPIENTRYP WindowPos2dvMESA)(const GLdouble * v); /* 625 */
-   void (GLAPIENTRYP WindowPos2fMESA)(GLfloat x, GLfloat y); /* 626 */
-   void (GLAPIENTRYP WindowPos2fvMESA)(const GLfloat * v); /* 627 */
-   void (GLAPIENTRYP WindowPos2iMESA)(GLint x, GLint y); /* 628 */
-   void (GLAPIENTRYP WindowPos2ivMESA)(const GLint * v); /* 629 */
-   void (GLAPIENTRYP WindowPos2sMESA)(GLshort x, GLshort y); /* 630 */
-   void (GLAPIENTRYP WindowPos2svMESA)(const GLshort * v); /* 631 */
-   void (GLAPIENTRYP WindowPos3dMESA)(GLdouble x, GLdouble y, GLdouble z); /* 632 */
-   void (GLAPIENTRYP WindowPos3dvMESA)(const GLdouble * v); /* 633 */
-   void (GLAPIENTRYP WindowPos3fMESA)(GLfloat x, GLfloat y, GLfloat z); /* 634 */
-   void (GLAPIENTRYP WindowPos3fvMESA)(const GLfloat * v); /* 635 */
-   void (GLAPIENTRYP WindowPos3iMESA)(GLint x, GLint y, GLint z); /* 636 */
-   void (GLAPIENTRYP WindowPos3ivMESA)(const GLint * v); /* 637 */
-   void (GLAPIENTRYP WindowPos3sMESA)(GLshort x, GLshort y, GLshort z); /* 638 */
-   void (GLAPIENTRYP WindowPos3svMESA)(const GLshort * v); /* 639 */
-   void (GLAPIENTRYP WindowPos4dMESA)(GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 640 */
-   void (GLAPIENTRYP WindowPos4dvMESA)(const GLdouble * v); /* 641 */
-   void (GLAPIENTRYP WindowPos4fMESA)(GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 642 */
-   void (GLAPIENTRYP WindowPos4fvMESA)(const GLfloat * v); /* 643 */
-   void (GLAPIENTRYP WindowPos4iMESA)(GLint x, GLint y, GLint z, GLint w); /* 644 */
-   void (GLAPIENTRYP WindowPos4ivMESA)(const GLint * v); /* 645 */
-   void (GLAPIENTRYP WindowPos4sMESA)(GLshort x, GLshort y, GLshort z, GLshort w); /* 646 */
-   void (GLAPIENTRYP WindowPos4svMESA)(const GLshort * v); /* 647 */
-   void (GLAPIENTRYP MultiModeDrawArraysIBM)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride); /* 648 */
-   void (GLAPIENTRYP MultiModeDrawElementsIBM)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride); /* 649 */
-   void (GLAPIENTRYP DeleteFencesNV)(GLsizei n, const GLuint * fences); /* 650 */
-   void (GLAPIENTRYP FinishFenceNV)(GLuint fence); /* 651 */
-   void (GLAPIENTRYP GenFencesNV)(GLsizei n, GLuint * fences); /* 652 */
-   void (GLAPIENTRYP GetFenceivNV)(GLuint fence, GLenum pname, GLint * params); /* 653 */
-   GLboolean (GLAPIENTRYP IsFenceNV)(GLuint fence); /* 654 */
-   void (GLAPIENTRYP SetFenceNV)(GLuint fence, GLenum condition); /* 655 */
-   GLboolean (GLAPIENTRYP TestFenceNV)(GLuint fence); /* 656 */
-   GLboolean (GLAPIENTRYP AreProgramsResidentNV)(GLsizei n, const GLuint * ids, GLboolean * residences); /* 657 */
-   void (GLAPIENTRYP BindProgramNV)(GLenum target, GLuint program); /* 658 */
-   void (GLAPIENTRYP DeleteProgramsNV)(GLsizei n, const GLuint * programs); /* 659 */
-   void (GLAPIENTRYP ExecuteProgramNV)(GLenum target, GLuint id, const GLfloat * params); /* 660 */
-   void (GLAPIENTRYP GenProgramsNV)(GLsizei n, GLuint * programs); /* 661 */
-   void (GLAPIENTRYP GetProgramParameterdvNV)(GLenum target, GLuint index, GLenum pname, GLdouble * params); /* 662 */
-   void (GLAPIENTRYP GetProgramParameterfvNV)(GLenum target, GLuint index, GLenum pname, GLfloat * params); /* 663 */
-   void (GLAPIENTRYP GetProgramStringNV)(GLuint id, GLenum pname, GLubyte * program); /* 664 */
-   void (GLAPIENTRYP GetProgramivNV)(GLuint id, GLenum pname, GLint * params); /* 665 */
-   void (GLAPIENTRYP GetTrackMatrixivNV)(GLenum target, GLuint address, GLenum pname, GLint * params); /* 666 */
-   void (GLAPIENTRYP GetVertexAttribPointervNV)(GLuint index, GLenum pname, GLvoid ** pointer); /* 667 */
-   void (GLAPIENTRYP GetVertexAttribdvNV)(GLuint index, GLenum pname, GLdouble * params); /* 668 */
-   void (GLAPIENTRYP GetVertexAttribfvNV)(GLuint index, GLenum pname, GLfloat * params); /* 669 */
-   void (GLAPIENTRYP GetVertexAttribivNV)(GLuint index, GLenum pname, GLint * params); /* 670 */
-   GLboolean (GLAPIENTRYP IsProgramNV)(GLuint program); /* 671 */
-   void (GLAPIENTRYP LoadProgramNV)(GLenum target, GLuint id, GLsizei len, const GLubyte * program); /* 672 */
-   void (GLAPIENTRYP ProgramParameters4dvNV)(GLenum target, GLuint index, GLuint num, const GLdouble * params); /* 673 */
-   void (GLAPIENTRYP ProgramParameters4fvNV)(GLenum target, GLuint index, GLuint num, const GLfloat * params); /* 674 */
-   void (GLAPIENTRYP RequestResidentProgramsNV)(GLsizei n, const GLuint * ids); /* 675 */
-   void (GLAPIENTRYP TrackMatrixNV)(GLenum target, GLuint address, GLenum matrix, GLenum transform); /* 676 */
-   void (GLAPIENTRYP VertexAttrib1dNV)(GLuint index, GLdouble x); /* 677 */
-   void (GLAPIENTRYP VertexAttrib1dvNV)(GLuint index, const GLdouble * v); /* 678 */
-   void (GLAPIENTRYP VertexAttrib1fNV)(GLuint index, GLfloat x); /* 679 */
-   void (GLAPIENTRYP VertexAttrib1fvNV)(GLuint index, const GLfloat * v); /* 680 */
-   void (GLAPIENTRYP VertexAttrib1sNV)(GLuint index, GLshort x); /* 681 */
-   void (GLAPIENTRYP VertexAttrib1svNV)(GLuint index, const GLshort * v); /* 682 */
-   void (GLAPIENTRYP VertexAttrib2dNV)(GLuint index, GLdouble x, GLdouble y); /* 683 */
-   void (GLAPIENTRYP VertexAttrib2dvNV)(GLuint index, const GLdouble * v); /* 684 */
-   void (GLAPIENTRYP VertexAttrib2fNV)(GLuint index, GLfloat x, GLfloat y); /* 685 */
-   void (GLAPIENTRYP VertexAttrib2fvNV)(GLuint index, const GLfloat * v); /* 686 */
-   void (GLAPIENTRYP VertexAttrib2sNV)(GLuint index, GLshort x, GLshort y); /* 687 */
-   void (GLAPIENTRYP VertexAttrib2svNV)(GLuint index, const GLshort * v); /* 688 */
-   void (GLAPIENTRYP VertexAttrib3dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z); /* 689 */
-   void (GLAPIENTRYP VertexAttrib3dvNV)(GLuint index, const GLdouble * v); /* 690 */
-   void (GLAPIENTRYP VertexAttrib3fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z); /* 691 */
-   void (GLAPIENTRYP VertexAttrib3fvNV)(GLuint index, const GLfloat * v); /* 692 */
-   void (GLAPIENTRYP VertexAttrib3sNV)(GLuint index, GLshort x, GLshort y, GLshort z); /* 693 */
-   void (GLAPIENTRYP VertexAttrib3svNV)(GLuint index, const GLshort * v); /* 694 */
-   void (GLAPIENTRYP VertexAttrib4dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 695 */
-   void (GLAPIENTRYP VertexAttrib4dvNV)(GLuint index, const GLdouble * v); /* 696 */
-   void (GLAPIENTRYP VertexAttrib4fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 697 */
-   void (GLAPIENTRYP VertexAttrib4fvNV)(GLuint index, const GLfloat * v); /* 698 */
-   void (GLAPIENTRYP VertexAttrib4sNV)(GLuint index, GLshort x, GLshort y, GLshort z, GLshort w); /* 699 */
-   void (GLAPIENTRYP VertexAttrib4svNV)(GLuint index, const GLshort * v); /* 700 */
-   void (GLAPIENTRYP VertexAttrib4ubNV)(GLuint index, GLubyte x, GLubyte y, GLubyte z, GLubyte w); /* 701 */
-   void (GLAPIENTRYP VertexAttrib4ubvNV)(GLuint index, const GLubyte * v); /* 702 */
-   void (GLAPIENTRYP VertexAttribPointerNV)(GLuint index, GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 703 */
-   void (GLAPIENTRYP VertexAttribs1dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 704 */
-   void (GLAPIENTRYP VertexAttribs1fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 705 */
-   void (GLAPIENTRYP VertexAttribs1svNV)(GLuint index, GLsizei n, const GLshort * v); /* 706 */
-   void (GLAPIENTRYP VertexAttribs2dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 707 */
-   void (GLAPIENTRYP VertexAttribs2fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 708 */
-   void (GLAPIENTRYP VertexAttribs2svNV)(GLuint index, GLsizei n, const GLshort * v); /* 709 */
-   void (GLAPIENTRYP VertexAttribs3dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 710 */
-   void (GLAPIENTRYP VertexAttribs3fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 711 */
-   void (GLAPIENTRYP VertexAttribs3svNV)(GLuint index, GLsizei n, const GLshort * v); /* 712 */
-   void (GLAPIENTRYP VertexAttribs4dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 713 */
-   void (GLAPIENTRYP VertexAttribs4fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 714 */
-   void (GLAPIENTRYP VertexAttribs4svNV)(GLuint index, GLsizei n, const GLshort * v); /* 715 */
-   void (GLAPIENTRYP VertexAttribs4ubvNV)(GLuint index, GLsizei n, const GLubyte * v); /* 716 */
-   void (GLAPIENTRYP AlphaFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 717 */
-   void (GLAPIENTRYP AlphaFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 718 */
-   void (GLAPIENTRYP AlphaFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 719 */
-   void (GLAPIENTRYP BeginFragmentShaderATI)(void); /* 720 */
-   void (GLAPIENTRYP BindFragmentShaderATI)(GLuint id); /* 721 */
-   void (GLAPIENTRYP ColorFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 722 */
-   void (GLAPIENTRYP ColorFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 723 */
-   void (GLAPIENTRYP ColorFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 724 */
-   void (GLAPIENTRYP DeleteFragmentShaderATI)(GLuint id); /* 725 */
-   void (GLAPIENTRYP EndFragmentShaderATI)(void); /* 726 */
-   GLuint (GLAPIENTRYP GenFragmentShadersATI)(GLuint range); /* 727 */
-   void (GLAPIENTRYP PassTexCoordATI)(GLuint dst, GLuint coord, GLenum swizzle); /* 728 */
-   void (GLAPIENTRYP SampleMapATI)(GLuint dst, GLuint interp, GLenum swizzle); /* 729 */
-   void (GLAPIENTRYP SetFragmentShaderConstantATI)(GLuint dst, const GLfloat * value); /* 730 */
-   void (GLAPIENTRYP PointParameteriNV)(GLenum pname, GLint param); /* 731 */
-   void (GLAPIENTRYP PointParameterivNV)(GLenum pname, const GLint * params); /* 732 */
-   void (GLAPIENTRYP ActiveStencilFaceEXT)(GLenum face); /* 733 */
-   void (GLAPIENTRYP BindVertexArrayAPPLE)(GLuint array); /* 734 */
-   void (GLAPIENTRYP DeleteVertexArraysAPPLE)(GLsizei n, const GLuint * arrays); /* 735 */
-   void (GLAPIENTRYP GenVertexArraysAPPLE)(GLsizei n, GLuint * arrays); /* 736 */
-   GLboolean (GLAPIENTRYP IsVertexArrayAPPLE)(GLuint array); /* 737 */
-   void (GLAPIENTRYP GetProgramNamedParameterdvNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble * params); /* 738 */
-   void (GLAPIENTRYP GetProgramNamedParameterfvNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat * params); /* 739 */
-   void (GLAPIENTRYP ProgramNamedParameter4dNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 740 */
-   void (GLAPIENTRYP ProgramNamedParameter4dvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLdouble * v); /* 741 */
-   void (GLAPIENTRYP ProgramNamedParameter4fNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 742 */
-   void (GLAPIENTRYP ProgramNamedParameter4fvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLfloat * v); /* 743 */
-   void (GLAPIENTRYP DepthBoundsEXT)(GLclampd zmin, GLclampd zmax); /* 744 */
-   void (GLAPIENTRYP BlendEquationSeparateEXT)(GLenum modeRGB, GLenum modeA); /* 745 */
-   void (GLAPIENTRYP BindFramebufferEXT)(GLenum target, GLuint framebuffer); /* 746 */
-   void (GLAPIENTRYP BindRenderbufferEXT)(GLenum target, GLuint renderbuffer); /* 747 */
-   GLenum (GLAPIENTRYP CheckFramebufferStatusEXT)(GLenum target); /* 748 */
-   void (GLAPIENTRYP DeleteFramebuffersEXT)(GLsizei n, const GLuint * framebuffers); /* 749 */
-   void (GLAPIENTRYP DeleteRenderbuffersEXT)(GLsizei n, const GLuint * renderbuffers); /* 750 */
-   void (GLAPIENTRYP FramebufferRenderbufferEXT)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer); /* 751 */
-   void (GLAPIENTRYP FramebufferTexture1DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 752 */
-   void (GLAPIENTRYP FramebufferTexture2DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 753 */
-   void (GLAPIENTRYP FramebufferTexture3DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset); /* 754 */
-   void (GLAPIENTRYP GenFramebuffersEXT)(GLsizei n, GLuint * framebuffers); /* 755 */
-   void (GLAPIENTRYP GenRenderbuffersEXT)(GLsizei n, GLuint * renderbuffers); /* 756 */
-   void (GLAPIENTRYP GenerateMipmapEXT)(GLenum target); /* 757 */
-   void (GLAPIENTRYP GetFramebufferAttachmentParameterivEXT)(GLenum target, GLenum attachment, GLenum pname, GLint * params); /* 758 */
-   void (GLAPIENTRYP GetRenderbufferParameterivEXT)(GLenum target, GLenum pname, GLint * params); /* 759 */
-   GLboolean (GLAPIENTRYP IsFramebufferEXT)(GLuint framebuffer); /* 760 */
-   GLboolean (GLAPIENTRYP IsRenderbufferEXT)(GLuint renderbuffer); /* 761 */
-   void (GLAPIENTRYP RenderbufferStorageEXT)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height); /* 762 */
-   void (GLAPIENTRYP BlitFramebufferEXT)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); /* 763 */
-   void (GLAPIENTRYP FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer); /* 764 */
-   void (GLAPIENTRYP StencilFuncSeparateATI)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); /* 765 */
-   void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 766 */
-   void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 767 */
-   void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 768 */
-   void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 769 */
+   void (GLAPIENTRYP RenderbufferStorageMultisample)(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height); /* 561 */
+   void (GLAPIENTRYP PolygonOffsetEXT)(GLfloat factor, GLfloat bias); /* 562 */
+   void (GLAPIENTRYP GetPixelTexGenParameterfvSGIS)(GLenum pname, GLfloat * params); /* 563 */
+   void (GLAPIENTRYP GetPixelTexGenParameterivSGIS)(GLenum pname, GLint * params); /* 564 */
+   void (GLAPIENTRYP PixelTexGenParameterfSGIS)(GLenum pname, GLfloat param); /* 565 */
+   void (GLAPIENTRYP PixelTexGenParameterfvSGIS)(GLenum pname, const GLfloat * params); /* 566 */
+   void (GLAPIENTRYP PixelTexGenParameteriSGIS)(GLenum pname, GLint param); /* 567 */
+   void (GLAPIENTRYP PixelTexGenParameterivSGIS)(GLenum pname, const GLint * params); /* 568 */
+   void (GLAPIENTRYP SampleMaskSGIS)(GLclampf value, GLboolean invert); /* 569 */
+   void (GLAPIENTRYP SamplePatternSGIS)(GLenum pattern); /* 570 */
+   void (GLAPIENTRYP ColorPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 571 */
+   void (GLAPIENTRYP EdgeFlagPointerEXT)(GLsizei stride, GLsizei count, const GLboolean * pointer); /* 572 */
+   void (GLAPIENTRYP IndexPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 573 */
+   void (GLAPIENTRYP NormalPointerEXT)(GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 574 */
+   void (GLAPIENTRYP TexCoordPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 575 */
+   void (GLAPIENTRYP VertexPointerEXT)(GLint size, GLenum type, GLsizei stride, GLsizei count, const GLvoid * pointer); /* 576 */
+   void (GLAPIENTRYP PointParameterfEXT)(GLenum pname, GLfloat param); /* 577 */
+   void (GLAPIENTRYP PointParameterfvEXT)(GLenum pname, const GLfloat * params); /* 578 */
+   void (GLAPIENTRYP LockArraysEXT)(GLint first, GLsizei count); /* 579 */
+   void (GLAPIENTRYP UnlockArraysEXT)(void); /* 580 */
+   void (GLAPIENTRYP CullParameterdvEXT)(GLenum pname, GLdouble * params); /* 581 */
+   void (GLAPIENTRYP CullParameterfvEXT)(GLenum pname, GLfloat * params); /* 582 */
+   void (GLAPIENTRYP SecondaryColor3bEXT)(GLbyte red, GLbyte green, GLbyte blue); /* 583 */
+   void (GLAPIENTRYP SecondaryColor3bvEXT)(const GLbyte * v); /* 584 */
+   void (GLAPIENTRYP SecondaryColor3dEXT)(GLdouble red, GLdouble green, GLdouble blue); /* 585 */
+   void (GLAPIENTRYP SecondaryColor3dvEXT)(const GLdouble * v); /* 586 */
+   void (GLAPIENTRYP SecondaryColor3fEXT)(GLfloat red, GLfloat green, GLfloat blue); /* 587 */
+   void (GLAPIENTRYP SecondaryColor3fvEXT)(const GLfloat * v); /* 588 */
+   void (GLAPIENTRYP SecondaryColor3iEXT)(GLint red, GLint green, GLint blue); /* 589 */
+   void (GLAPIENTRYP SecondaryColor3ivEXT)(const GLint * v); /* 590 */
+   void (GLAPIENTRYP SecondaryColor3sEXT)(GLshort red, GLshort green, GLshort blue); /* 591 */
+   void (GLAPIENTRYP SecondaryColor3svEXT)(const GLshort * v); /* 592 */
+   void (GLAPIENTRYP SecondaryColor3ubEXT)(GLubyte red, GLubyte green, GLubyte blue); /* 593 */
+   void (GLAPIENTRYP SecondaryColor3ubvEXT)(const GLubyte * v); /* 594 */
+   void (GLAPIENTRYP SecondaryColor3uiEXT)(GLuint red, GLuint green, GLuint blue); /* 595 */
+   void (GLAPIENTRYP SecondaryColor3uivEXT)(const GLuint * v); /* 596 */
+   void (GLAPIENTRYP SecondaryColor3usEXT)(GLushort red, GLushort green, GLushort blue); /* 597 */
+   void (GLAPIENTRYP SecondaryColor3usvEXT)(const GLushort * v); /* 598 */
+   void (GLAPIENTRYP SecondaryColorPointerEXT)(GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 599 */
+   void (GLAPIENTRYP MultiDrawArraysEXT)(GLenum mode, GLint * first, GLsizei * count, GLsizei primcount); /* 600 */
+   void (GLAPIENTRYP MultiDrawElementsEXT)(GLenum mode, const GLsizei * count, GLenum type, const GLvoid ** indices, GLsizei primcount); /* 601 */
+   void (GLAPIENTRYP FogCoordPointerEXT)(GLenum type, GLsizei stride, const GLvoid * pointer); /* 602 */
+   void (GLAPIENTRYP FogCoorddEXT)(GLdouble coord); /* 603 */
+   void (GLAPIENTRYP FogCoorddvEXT)(const GLdouble * coord); /* 604 */
+   void (GLAPIENTRYP FogCoordfEXT)(GLfloat coord); /* 605 */
+   void (GLAPIENTRYP FogCoordfvEXT)(const GLfloat * coord); /* 606 */
+   void (GLAPIENTRYP PixelTexGenSGIX)(GLenum mode); /* 607 */
+   void (GLAPIENTRYP BlendFuncSeparateEXT)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha); /* 608 */
+   void (GLAPIENTRYP FlushVertexArrayRangeNV)(void); /* 609 */
+   void (GLAPIENTRYP VertexArrayRangeNV)(GLsizei length, const GLvoid * pointer); /* 610 */
+   void (GLAPIENTRYP CombinerInputNV)(GLenum stage, GLenum portion, GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 611 */
+   void (GLAPIENTRYP CombinerOutputNV)(GLenum stage, GLenum portion, GLenum abOutput, GLenum cdOutput, GLenum sumOutput, GLenum scale, GLenum bias, GLboolean abDotProduct, GLboolean cdDotProduct, GLboolean muxSum); /* 612 */
+   void (GLAPIENTRYP CombinerParameterfNV)(GLenum pname, GLfloat param); /* 613 */
+   void (GLAPIENTRYP CombinerParameterfvNV)(GLenum pname, const GLfloat * params); /* 614 */
+   void (GLAPIENTRYP CombinerParameteriNV)(GLenum pname, GLint param); /* 615 */
+   void (GLAPIENTRYP CombinerParameterivNV)(GLenum pname, const GLint * params); /* 616 */
+   void (GLAPIENTRYP FinalCombinerInputNV)(GLenum variable, GLenum input, GLenum mapping, GLenum componentUsage); /* 617 */
+   void (GLAPIENTRYP GetCombinerInputParameterfvNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLfloat * params); /* 618 */
+   void (GLAPIENTRYP GetCombinerInputParameterivNV)(GLenum stage, GLenum portion, GLenum variable, GLenum pname, GLint * params); /* 619 */
+   void (GLAPIENTRYP GetCombinerOutputParameterfvNV)(GLenum stage, GLenum portion, GLenum pname, GLfloat * params); /* 620 */
+   void (GLAPIENTRYP GetCombinerOutputParameterivNV)(GLenum stage, GLenum portion, GLenum pname, GLint * params); /* 621 */
+   void (GLAPIENTRYP GetFinalCombinerInputParameterfvNV)(GLenum variable, GLenum pname, GLfloat * params); /* 622 */
+   void (GLAPIENTRYP GetFinalCombinerInputParameterivNV)(GLenum variable, GLenum pname, GLint * params); /* 623 */
+   void (GLAPIENTRYP ResizeBuffersMESA)(void); /* 624 */
+   void (GLAPIENTRYP WindowPos2dMESA)(GLdouble x, GLdouble y); /* 625 */
+   void (GLAPIENTRYP WindowPos2dvMESA)(const GLdouble * v); /* 626 */
+   void (GLAPIENTRYP WindowPos2fMESA)(GLfloat x, GLfloat y); /* 627 */
+   void (GLAPIENTRYP WindowPos2fvMESA)(const GLfloat * v); /* 628 */
+   void (GLAPIENTRYP WindowPos2iMESA)(GLint x, GLint y); /* 629 */
+   void (GLAPIENTRYP WindowPos2ivMESA)(const GLint * v); /* 630 */
+   void (GLAPIENTRYP WindowPos2sMESA)(GLshort x, GLshort y); /* 631 */
+   void (GLAPIENTRYP WindowPos2svMESA)(const GLshort * v); /* 632 */
+   void (GLAPIENTRYP WindowPos3dMESA)(GLdouble x, GLdouble y, GLdouble z); /* 633 */
+   void (GLAPIENTRYP WindowPos3dvMESA)(const GLdouble * v); /* 634 */
+   void (GLAPIENTRYP WindowPos3fMESA)(GLfloat x, GLfloat y, GLfloat z); /* 635 */
+   void (GLAPIENTRYP WindowPos3fvMESA)(const GLfloat * v); /* 636 */
+   void (GLAPIENTRYP WindowPos3iMESA)(GLint x, GLint y, GLint z); /* 637 */
+   void (GLAPIENTRYP WindowPos3ivMESA)(const GLint * v); /* 638 */
+   void (GLAPIENTRYP WindowPos3sMESA)(GLshort x, GLshort y, GLshort z); /* 639 */
+   void (GLAPIENTRYP WindowPos3svMESA)(const GLshort * v); /* 640 */
+   void (GLAPIENTRYP WindowPos4dMESA)(GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 641 */
+   void (GLAPIENTRYP WindowPos4dvMESA)(const GLdouble * v); /* 642 */
+   void (GLAPIENTRYP WindowPos4fMESA)(GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 643 */
+   void (GLAPIENTRYP WindowPos4fvMESA)(const GLfloat * v); /* 644 */
+   void (GLAPIENTRYP WindowPos4iMESA)(GLint x, GLint y, GLint z, GLint w); /* 645 */
+   void (GLAPIENTRYP WindowPos4ivMESA)(const GLint * v); /* 646 */
+   void (GLAPIENTRYP WindowPos4sMESA)(GLshort x, GLshort y, GLshort z, GLshort w); /* 647 */
+   void (GLAPIENTRYP WindowPos4svMESA)(const GLshort * v); /* 648 */
+   void (GLAPIENTRYP MultiModeDrawArraysIBM)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride); /* 649 */
+   void (GLAPIENTRYP MultiModeDrawElementsIBM)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride); /* 650 */
+   void (GLAPIENTRYP DeleteFencesNV)(GLsizei n, const GLuint * fences); /* 651 */
+   void (GLAPIENTRYP FinishFenceNV)(GLuint fence); /* 652 */
+   void (GLAPIENTRYP GenFencesNV)(GLsizei n, GLuint * fences); /* 653 */
+   void (GLAPIENTRYP GetFenceivNV)(GLuint fence, GLenum pname, GLint * params); /* 654 */
+   GLboolean (GLAPIENTRYP IsFenceNV)(GLuint fence); /* 655 */
+   void (GLAPIENTRYP SetFenceNV)(GLuint fence, GLenum condition); /* 656 */
+   GLboolean (GLAPIENTRYP TestFenceNV)(GLuint fence); /* 657 */
+   GLboolean (GLAPIENTRYP AreProgramsResidentNV)(GLsizei n, const GLuint * ids, GLboolean * residences); /* 658 */
+   void (GLAPIENTRYP BindProgramNV)(GLenum target, GLuint program); /* 659 */
+   void (GLAPIENTRYP DeleteProgramsNV)(GLsizei n, const GLuint * programs); /* 660 */
+   void (GLAPIENTRYP ExecuteProgramNV)(GLenum target, GLuint id, const GLfloat * params); /* 661 */
+   void (GLAPIENTRYP GenProgramsNV)(GLsizei n, GLuint * programs); /* 662 */
+   void (GLAPIENTRYP GetProgramParameterdvNV)(GLenum target, GLuint index, GLenum pname, GLdouble * params); /* 663 */
+   void (GLAPIENTRYP GetProgramParameterfvNV)(GLenum target, GLuint index, GLenum pname, GLfloat * params); /* 664 */
+   void (GLAPIENTRYP GetProgramStringNV)(GLuint id, GLenum pname, GLubyte * program); /* 665 */
+   void (GLAPIENTRYP GetProgramivNV)(GLuint id, GLenum pname, GLint * params); /* 666 */
+   void (GLAPIENTRYP GetTrackMatrixivNV)(GLenum target, GLuint address, GLenum pname, GLint * params); /* 667 */
+   void (GLAPIENTRYP GetVertexAttribPointervNV)(GLuint index, GLenum pname, GLvoid ** pointer); /* 668 */
+   void (GLAPIENTRYP GetVertexAttribdvNV)(GLuint index, GLenum pname, GLdouble * params); /* 669 */
+   void (GLAPIENTRYP GetVertexAttribfvNV)(GLuint index, GLenum pname, GLfloat * params); /* 670 */
+   void (GLAPIENTRYP GetVertexAttribivNV)(GLuint index, GLenum pname, GLint * params); /* 671 */
+   GLboolean (GLAPIENTRYP IsProgramNV)(GLuint program); /* 672 */
+   void (GLAPIENTRYP LoadProgramNV)(GLenum target, GLuint id, GLsizei len, const GLubyte * program); /* 673 */
+   void (GLAPIENTRYP ProgramParameters4dvNV)(GLenum target, GLuint index, GLuint num, const GLdouble * params); /* 674 */
+   void (GLAPIENTRYP ProgramParameters4fvNV)(GLenum target, GLuint index, GLuint num, const GLfloat * params); /* 675 */
+   void (GLAPIENTRYP RequestResidentProgramsNV)(GLsizei n, const GLuint * ids); /* 676 */
+   void (GLAPIENTRYP TrackMatrixNV)(GLenum target, GLuint address, GLenum matrix, GLenum transform); /* 677 */
+   void (GLAPIENTRYP VertexAttrib1dNV)(GLuint index, GLdouble x); /* 678 */
+   void (GLAPIENTRYP VertexAttrib1dvNV)(GLuint index, const GLdouble * v); /* 679 */
+   void (GLAPIENTRYP VertexAttrib1fNV)(GLuint index, GLfloat x); /* 680 */
+   void (GLAPIENTRYP VertexAttrib1fvNV)(GLuint index, const GLfloat * v); /* 681 */
+   void (GLAPIENTRYP VertexAttrib1sNV)(GLuint index, GLshort x); /* 682 */
+   void (GLAPIENTRYP VertexAttrib1svNV)(GLuint index, const GLshort * v); /* 683 */
+   void (GLAPIENTRYP VertexAttrib2dNV)(GLuint index, GLdouble x, GLdouble y); /* 684 */
+   void (GLAPIENTRYP VertexAttrib2dvNV)(GLuint index, const GLdouble * v); /* 685 */
+   void (GLAPIENTRYP VertexAttrib2fNV)(GLuint index, GLfloat x, GLfloat y); /* 686 */
+   void (GLAPIENTRYP VertexAttrib2fvNV)(GLuint index, const GLfloat * v); /* 687 */
+   void (GLAPIENTRYP VertexAttrib2sNV)(GLuint index, GLshort x, GLshort y); /* 688 */
+   void (GLAPIENTRYP VertexAttrib2svNV)(GLuint index, const GLshort * v); /* 689 */
+   void (GLAPIENTRYP VertexAttrib3dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z); /* 690 */
+   void (GLAPIENTRYP VertexAttrib3dvNV)(GLuint index, const GLdouble * v); /* 691 */
+   void (GLAPIENTRYP VertexAttrib3fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z); /* 692 */
+   void (GLAPIENTRYP VertexAttrib3fvNV)(GLuint index, const GLfloat * v); /* 693 */
+   void (GLAPIENTRYP VertexAttrib3sNV)(GLuint index, GLshort x, GLshort y, GLshort z); /* 694 */
+   void (GLAPIENTRYP VertexAttrib3svNV)(GLuint index, const GLshort * v); /* 695 */
+   void (GLAPIENTRYP VertexAttrib4dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 696 */
+   void (GLAPIENTRYP VertexAttrib4dvNV)(GLuint index, const GLdouble * v); /* 697 */
+   void (GLAPIENTRYP VertexAttrib4fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 698 */
+   void (GLAPIENTRYP VertexAttrib4fvNV)(GLuint index, const GLfloat * v); /* 699 */
+   void (GLAPIENTRYP VertexAttrib4sNV)(GLuint index, GLshort x, GLshort y, GLshort z, GLshort w); /* 700 */
+   void (GLAPIENTRYP VertexAttrib4svNV)(GLuint index, const GLshort * v); /* 701 */
+   void (GLAPIENTRYP VertexAttrib4ubNV)(GLuint index, GLubyte x, GLubyte y, GLubyte z, GLubyte w); /* 702 */
+   void (GLAPIENTRYP VertexAttrib4ubvNV)(GLuint index, const GLubyte * v); /* 703 */
+   void (GLAPIENTRYP VertexAttribPointerNV)(GLuint index, GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 704 */
+   void (GLAPIENTRYP VertexAttribs1dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 705 */
+   void (GLAPIENTRYP VertexAttribs1fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 706 */
+   void (GLAPIENTRYP VertexAttribs1svNV)(GLuint index, GLsizei n, const GLshort * v); /* 707 */
+   void (GLAPIENTRYP VertexAttribs2dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 708 */
+   void (GLAPIENTRYP VertexAttribs2fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 709 */
+   void (GLAPIENTRYP VertexAttribs2svNV)(GLuint index, GLsizei n, const GLshort * v); /* 710 */
+   void (GLAPIENTRYP VertexAttribs3dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 711 */
+   void (GLAPIENTRYP VertexAttribs3fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 712 */
+   void (GLAPIENTRYP VertexAttribs3svNV)(GLuint index, GLsizei n, const GLshort * v); /* 713 */
+   void (GLAPIENTRYP VertexAttribs4dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 714 */
+   void (GLAPIENTRYP VertexAttribs4fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 715 */
+   void (GLAPIENTRYP VertexAttribs4svNV)(GLuint index, GLsizei n, const GLshort * v); /* 716 */
+   void (GLAPIENTRYP VertexAttribs4ubvNV)(GLuint index, GLsizei n, const GLubyte * v); /* 717 */
+   void (GLAPIENTRYP AlphaFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 718 */
+   void (GLAPIENTRYP AlphaFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 719 */
+   void (GLAPIENTRYP AlphaFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 720 */
+   void (GLAPIENTRYP BeginFragmentShaderATI)(void); /* 721 */
+   void (GLAPIENTRYP BindFragmentShaderATI)(GLuint id); /* 722 */
+   void (GLAPIENTRYP ColorFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 723 */
+   void (GLAPIENTRYP ColorFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 724 */
+   void (GLAPIENTRYP ColorFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 725 */
+   void (GLAPIENTRYP DeleteFragmentShaderATI)(GLuint id); /* 726 */
+   void (GLAPIENTRYP EndFragmentShaderATI)(void); /* 727 */
+   GLuint (GLAPIENTRYP GenFragmentShadersATI)(GLuint range); /* 728 */
+   void (GLAPIENTRYP PassTexCoordATI)(GLuint dst, GLuint coord, GLenum swizzle); /* 729 */
+   void (GLAPIENTRYP SampleMapATI)(GLuint dst, GLuint interp, GLenum swizzle); /* 730 */
+   void (GLAPIENTRYP SetFragmentShaderConstantATI)(GLuint dst, const GLfloat * value); /* 731 */
+   void (GLAPIENTRYP PointParameteriNV)(GLenum pname, GLint param); /* 732 */
+   void (GLAPIENTRYP PointParameterivNV)(GLenum pname, const GLint * params); /* 733 */
+   void (GLAPIENTRYP ActiveStencilFaceEXT)(GLenum face); /* 734 */
+   void (GLAPIENTRYP BindVertexArrayAPPLE)(GLuint array); /* 735 */
+   void (GLAPIENTRYP DeleteVertexArraysAPPLE)(GLsizei n, const GLuint * arrays); /* 736 */
+   void (GLAPIENTRYP GenVertexArraysAPPLE)(GLsizei n, GLuint * arrays); /* 737 */
+   GLboolean (GLAPIENTRYP IsVertexArrayAPPLE)(GLuint array); /* 738 */
+   void (GLAPIENTRYP GetProgramNamedParameterdvNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble * params); /* 739 */
+   void (GLAPIENTRYP GetProgramNamedParameterfvNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat * params); /* 740 */
+   void (GLAPIENTRYP ProgramNamedParameter4dNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 741 */
+   void (GLAPIENTRYP ProgramNamedParameter4dvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLdouble * v); /* 742 */
+   void (GLAPIENTRYP ProgramNamedParameter4fNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 743 */
+   void (GLAPIENTRYP ProgramNamedParameter4fvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLfloat * v); /* 744 */
+   void (GLAPIENTRYP DepthBoundsEXT)(GLclampd zmin, GLclampd zmax); /* 745 */
+   void (GLAPIENTRYP BlendEquationSeparateEXT)(GLenum modeRGB, GLenum modeA); /* 746 */
+   void (GLAPIENTRYP BindFramebufferEXT)(GLenum target, GLuint framebuffer); /* 747 */
+   void (GLAPIENTRYP BindRenderbufferEXT)(GLenum target, GLuint renderbuffer); /* 748 */
+   GLenum (GLAPIENTRYP CheckFramebufferStatusEXT)(GLenum target); /* 749 */
+   void (GLAPIENTRYP DeleteFramebuffersEXT)(GLsizei n, const GLuint * framebuffers); /* 750 */
+   void (GLAPIENTRYP DeleteRenderbuffersEXT)(GLsizei n, const GLuint * renderbuffers); /* 751 */
+   void (GLAPIENTRYP FramebufferRenderbufferEXT)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer); /* 752 */
+   void (GLAPIENTRYP FramebufferTexture1DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 753 */
+   void (GLAPIENTRYP FramebufferTexture2DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 754 */
+   void (GLAPIENTRYP FramebufferTexture3DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset); /* 755 */
+   void (GLAPIENTRYP GenFramebuffersEXT)(GLsizei n, GLuint * framebuffers); /* 756 */
+   void (GLAPIENTRYP GenRenderbuffersEXT)(GLsizei n, GLuint * renderbuffers); /* 757 */
+   void (GLAPIENTRYP GenerateMipmapEXT)(GLenum target); /* 758 */
+   void (GLAPIENTRYP GetFramebufferAttachmentParameterivEXT)(GLenum target, GLenum attachment, GLenum pname, GLint * params); /* 759 */
+   void (GLAPIENTRYP GetRenderbufferParameterivEXT)(GLenum target, GLenum pname, GLint * params); /* 760 */
+   GLboolean (GLAPIENTRYP IsFramebufferEXT)(GLuint framebuffer); /* 761 */
+   GLboolean (GLAPIENTRYP IsRenderbufferEXT)(GLuint renderbuffer); /* 762 */
+   void (GLAPIENTRYP RenderbufferStorageEXT)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height); /* 763 */
+   void (GLAPIENTRYP BlitFramebufferEXT)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); /* 764 */
+   void (GLAPIENTRYP FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer); /* 765 */
+   void (GLAPIENTRYP StencilFuncSeparateATI)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); /* 766 */
+   void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 767 */
+   void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 768 */
+   void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 769 */
+   void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 770 */
 };
 
 #endif /* !defined( _GLAPI_TABLE_H_ ) */
diff --git a/src/mesa/glapi/glapitemp.h b/src/mesa/glapi/glapitemp.h
index 09259f4704..960d70e081 100644
--- a/src/mesa/glapi/glapitemp.h
+++ b/src/mesa/glapi/glapitemp.h
@@ -4006,63 +4006,68 @@ KEYWORD1 void KEYWORD2 NAME(DrawBuffersATI)(GLsizei n, const GLenum * bufs)
    DISPATCH(DrawBuffersARB, (n, bufs), (F, "glDrawBuffersATI(%d, %p);\n", n, (const void *) bufs));
 }
 
+KEYWORD1 void KEYWORD2 NAME(RenderbufferStorageMultisample)(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height)
+{
+   DISPATCH(RenderbufferStorageMultisample, (target, samples, internalformat, width, height), (F, "glRenderbufferStorageMultisample(0x%x, %d, 0x%x, %d, %d);\n", target, samples, internalformat, width, height));
+}
+
 KEYWORD1 void KEYWORD2 NAME(PolygonOffsetEXT)(GLfloat factor, GLfloat bias)
 {
    DISPATCH(PolygonOffsetEXT, (factor, bias), (F, "glPolygonOffsetEXT(%f, %f);\n", factor, bias));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_562)(GLenum pname, GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_563)(GLenum pname, GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_562)(GLenum pname, GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_563)(GLenum pname, GLfloat * params)
 {
    DISPATCH(GetPixelTexGenParameterfvSGIS, (pname, params), (F, "glGetPixelTexGenParameterfvSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_563)(GLenum pname, GLint * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_564)(GLenum pname, GLint * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_563)(GLenum pname, GLint * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_564)(GLenum pname, GLint * params)
 {
    DISPATCH(GetPixelTexGenParameterivSGIS, (pname, params), (F, "glGetPixelTexGenParameterivSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_564)(GLenum pname, GLfloat param);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_565)(GLenum pname, GLfloat param);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_564)(GLenum pname, GLfloat param)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_565)(GLenum pname, GLfloat param)
 {
    DISPATCH(PixelTexGenParameterfSGIS, (pname, param), (F, "glPixelTexGenParameterfSGIS(0x%x, %f);\n", pname, param));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_565)(GLenum pname, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_566)(GLenum pname, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_565)(GLenum pname, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_566)(GLenum pname, const GLfloat * params)
 {
    DISPATCH(PixelTexGenParameterfvSGIS, (pname, params), (F, "glPixelTexGenParameterfvSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_566)(GLenum pname, GLint param);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_567)(GLenum pname, GLint param);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_566)(GLenum pname, GLint param)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_567)(GLenum pname, GLint param)
 {
    DISPATCH(PixelTexGenParameteriSGIS, (pname, param), (F, "glPixelTexGenParameteriSGIS(0x%x, %d);\n", pname, param));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_567)(GLenum pname, const GLint * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_568)(GLenum pname, const GLint * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_567)(GLenum pname, const GLint * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_568)(GLenum pname, const GLint * params)
 {
    DISPATCH(PixelTexGenParameterivSGIS, (pname, params), (F, "glPixelTexGenParameterivSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_568)(GLclampf value, GLboolean invert);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_569)(GLclampf value, GLboolean invert);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_568)(GLclampf value, GLboolean invert)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_569)(GLclampf value, GLboolean invert)
 {
    DISPATCH(SampleMaskSGIS, (value, invert), (F, "glSampleMaskSGIS(%f, %d);\n", value, invert));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_569)(GLenum pattern);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_570)(GLenum pattern);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_569)(GLenum pattern)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_570)(GLenum pattern)
 {
    DISPATCH(SamplePatternSGIS, (pattern), (F, "glSamplePatternSGIS(0x%x);\n", pattern));
 }
@@ -4112,9 +4117,9 @@ KEYWORD1 void KEYWORD2 NAME(PointParameterfEXT)(GLenum pname, GLfloat param)
    DISPATCH(PointParameterfEXT, (pname, param), (F, "glPointParameterfEXT(0x%x, %f);\n", pname, param));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_576)(GLenum pname, GLfloat param);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_577)(GLenum pname, GLfloat param);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_576)(GLenum pname, GLfloat param)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_577)(GLenum pname, GLfloat param)
 {
    DISPATCH(PointParameterfEXT, (pname, param), (F, "glPointParameterfSGIS(0x%x, %f);\n", pname, param));
 }
@@ -4134,9 +4139,9 @@ KEYWORD1 void KEYWORD2 NAME(PointParameterfvEXT)(GLenum pname, const GLfloat * p
    DISPATCH(PointParameterfvEXT, (pname, params), (F, "glPointParameterfvEXT(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_577)(GLenum pname, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_578)(GLenum pname, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_577)(GLenum pname, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_578)(GLenum pname, const GLfloat * params)
 {
    DISPATCH(PointParameterfvEXT, (pname, params), (F, "glPointParameterfvSGIS(0x%x, %p);\n", pname, (const void *) params));
 }
@@ -4151,16 +4156,16 @@ KEYWORD1 void KEYWORD2 NAME(UnlockArraysEXT)(void)
    DISPATCH(UnlockArraysEXT, (), (F, "glUnlockArraysEXT();\n"));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_580)(GLenum pname, GLdouble * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_581)(GLenum pname, GLdouble * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_580)(GLenum pname, GLdouble * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_581)(GLenum pname, GLdouble * params)
 {
    DISPATCH(CullParameterdvEXT, (pname, params), (F, "glCullParameterdvEXT(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_581)(GLenum pname, GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_582)(GLenum pname, GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_581)(GLenum pname, GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_582)(GLenum pname, GLfloat * params)
 {
    DISPATCH(CullParameterfvEXT, (pname, params), (F, "glCullParameterfvEXT(0x%x, %p);\n", pname, (const void *) params));
 }
@@ -4405,9 +4410,9 @@ KEYWORD1 void KEYWORD2 NAME(FogCoordfvEXT)(const GLfloat * coord)
    DISPATCH(FogCoordfvEXT, (coord), (F, "glFogCoordfvEXT(%p);\n", (const void *) coord));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_606)(GLenum mode);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_607)(GLenum mode);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_606)(GLenum mode)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_607)(GLenum mode)
 {
    DISPATCH(PixelTexGenSGIX, (mode), (F, "glPixelTexGenSGIX(0x%x);\n", mode));
 }
@@ -4422,9 +4427,9 @@ KEYWORD1 void KEYWORD2 NAME(BlendFuncSeparateEXT)(GLenum sfactorRGB, GLenum dfac
    DISPATCH(BlendFuncSeparateEXT, (sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha), (F, "glBlendFuncSeparateEXT(0x%x, 0x%x, 0x%x, 0x%x);\n", sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_607)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_608)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_607)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_608)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha)
 {
    DISPATCH(BlendFuncSeparateEXT, (sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha), (F, "glBlendFuncSeparateINGR(0x%x, 0x%x, 0x%x, 0x%x);\n", sfactorRGB, dfactorRGB, sfactorAlpha, dfactorAlpha));
 }
@@ -4789,65 +4794,65 @@ KEYWORD1 void KEYWORD2 NAME(WindowPos4svMESA)(const GLshort * v)
    DISPATCH(WindowPos4svMESA, (v), (F, "glWindowPos4svMESA(%p);\n", (const void *) v));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_648)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_649)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_648)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_649)(const GLenum * mode, const GLint * first, const GLsizei * count, GLsizei primcount, GLint modestride)
 {
    DISPATCH(MultiModeDrawArraysIBM, (mode, first, count, primcount, modestride), (F, "glMultiModeDrawArraysIBM(%p, %p, %p, %d, %d);\n", (const void *) mode, (const void *) first, (const void *) count, primcount, modestride));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_649)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_650)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_649)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_650)(const GLenum * mode, const GLsizei * count, GLenum type, const GLvoid * const * indices, GLsizei primcount, GLint modestride)
 {
    DISPATCH(MultiModeDrawElementsIBM, (mode, count, type, indices, primcount, modestride), (F, "glMultiModeDrawElementsIBM(%p, %p, 0x%x, %p, %d, %d);\n", (const void *) mode, (const void *) count, type, (const void *) indices, primcount, modestride));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_650)(GLsizei n, const GLuint * fences);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_651)(GLsizei n, const GLuint * fences);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_650)(GLsizei n, const GLuint * fences)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_651)(GLsizei n, const GLuint * fences)
 {
    DISPATCH(DeleteFencesNV, (n, fences), (F, "glDeleteFencesNV(%d, %p);\n", n, (const void *) fences));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_651)(GLuint fence);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_652)(GLuint fence);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_651)(GLuint fence)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_652)(GLuint fence)
 {
    DISPATCH(FinishFenceNV, (fence), (F, "glFinishFenceNV(%d);\n", fence));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_652)(GLsizei n, GLuint * fences);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_653)(GLsizei n, GLuint * fences);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_652)(GLsizei n, GLuint * fences)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_653)(GLsizei n, GLuint * fences)
 {
    DISPATCH(GenFencesNV, (n, fences), (F, "glGenFencesNV(%d, %p);\n", n, (const void *) fences));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_653)(GLuint fence, GLenum pname, GLint * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_654)(GLuint fence, GLenum pname, GLint * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_653)(GLuint fence, GLenum pname, GLint * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_654)(GLuint fence, GLenum pname, GLint * params)
 {
    DISPATCH(GetFenceivNV, (fence, pname, params), (F, "glGetFenceivNV(%d, 0x%x, %p);\n", fence, pname, (const void *) params));
 }
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_654)(GLuint fence);
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_655)(GLuint fence);
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_654)(GLuint fence)
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_655)(GLuint fence)
 {
    RETURN_DISPATCH(IsFenceNV, (fence), (F, "glIsFenceNV(%d);\n", fence));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_655)(GLuint fence, GLenum condition);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_656)(GLuint fence, GLenum condition);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_655)(GLuint fence, GLenum condition)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_656)(GLuint fence, GLenum condition)
 {
    DISPATCH(SetFenceNV, (fence, condition), (F, "glSetFenceNV(%d, 0x%x);\n", fence, condition));
 }
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_656)(GLuint fence);
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_657)(GLuint fence);
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_656)(GLuint fence)
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_657)(GLuint fence)
 {
    RETURN_DISPATCH(TestFenceNV, (fence), (F, "glTestFenceNV(%d);\n", fence));
 }
@@ -5272,37 +5277,37 @@ KEYWORD1 void KEYWORD2 NAME(PointParameterivNV)(GLenum pname, const GLint * para
    DISPATCH(PointParameterivNV, (pname, params), (F, "glPointParameterivNV(0x%x, %p);\n", pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_733)(GLenum face);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_734)(GLenum face);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_733)(GLenum face)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_734)(GLenum face)
 {
    DISPATCH(ActiveStencilFaceEXT, (face), (F, "glActiveStencilFaceEXT(0x%x);\n", face));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_734)(GLuint array);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_735)(GLuint array);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_734)(GLuint array)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_735)(GLuint array)
 {
    DISPATCH(BindVertexArrayAPPLE, (array), (F, "glBindVertexArrayAPPLE(%d);\n", array));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_735)(GLsizei n, const GLuint * arrays);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_736)(GLsizei n, const GLuint * arrays);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_735)(GLsizei n, const GLuint * arrays)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_736)(GLsizei n, const GLuint * arrays)
 {
    DISPATCH(DeleteVertexArraysAPPLE, (n, arrays), (F, "glDeleteVertexArraysAPPLE(%d, %p);\n", n, (const void *) arrays));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_736)(GLsizei n, GLuint * arrays);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_737)(GLsizei n, GLuint * arrays);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_736)(GLsizei n, GLuint * arrays)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_737)(GLsizei n, GLuint * arrays)
 {
    DISPATCH(GenVertexArraysAPPLE, (n, arrays), (F, "glGenVertexArraysAPPLE(%d, %p);\n", n, (const void *) arrays));
 }
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_737)(GLuint array);
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_738)(GLuint array);
 
-KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_737)(GLuint array)
+KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_738)(GLuint array)
 {
    RETURN_DISPATCH(IsVertexArrayAPPLE, (array), (F, "glIsVertexArrayAPPLE(%d);\n", array));
 }
@@ -5337,9 +5342,9 @@ KEYWORD1 void KEYWORD2 NAME(ProgramNamedParameter4fvNV)(GLuint id, GLsizei len,
    DISPATCH(ProgramNamedParameter4fvNV, (id, len, name, v), (F, "glProgramNamedParameter4fvNV(%d, %d, %p, %p);\n", id, len, (const void *) name, (const void *) v));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_744)(GLclampd zmin, GLclampd zmax);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_745)(GLclampd zmin, GLclampd zmax);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_744)(GLclampd zmin, GLclampd zmax)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_745)(GLclampd zmin, GLclampd zmax)
 {
    DISPATCH(DepthBoundsEXT, (zmin, zmax), (F, "glDepthBoundsEXT(%f, %f);\n", zmin, zmax));
 }
@@ -5349,141 +5354,236 @@ KEYWORD1 void KEYWORD2 NAME(BlendEquationSeparate)(GLenum modeRGB, GLenum modeA)
    DISPATCH(BlendEquationSeparateEXT, (modeRGB, modeA), (F, "glBlendEquationSeparate(0x%x, 0x%x);\n", modeRGB, modeA));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_745)(GLenum modeRGB, GLenum modeA);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_746)(GLenum modeRGB, GLenum modeA);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_745)(GLenum modeRGB, GLenum modeA)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_746)(GLenum modeRGB, GLenum modeA)
 {
    DISPATCH(BlendEquationSeparateEXT, (modeRGB, modeA), (F, "glBlendEquationSeparateEXT(0x%x, 0x%x);\n", modeRGB, modeA));
 }
 
+KEYWORD1 void KEYWORD2 NAME(BindFramebuffer)(GLenum target, GLuint framebuffer)
+{
+   DISPATCH(BindFramebufferEXT, (target, framebuffer), (F, "glBindFramebuffer(0x%x, %d);\n", target, framebuffer));
+}
+
 KEYWORD1 void KEYWORD2 NAME(BindFramebufferEXT)(GLenum target, GLuint framebuffer)
 {
    DISPATCH(BindFramebufferEXT, (target, framebuffer), (F, "glBindFramebufferEXT(0x%x, %d);\n", target, framebuffer));
 }
 
+KEYWORD1 void KEYWORD2 NAME(BindRenderbuffer)(GLenum target, GLuint renderbuffer)
+{
+   DISPATCH(BindRenderbufferEXT, (target, renderbuffer), (F, "glBindRenderbuffer(0x%x, %d);\n", target, renderbuffer));
+}
+
 KEYWORD1 void KEYWORD2 NAME(BindRenderbufferEXT)(GLenum target, GLuint renderbuffer)
 {
    DISPATCH(BindRenderbufferEXT, (target, renderbuffer), (F, "glBindRenderbufferEXT(0x%x, %d);\n", target, renderbuffer));
 }
 
+KEYWORD1 GLenum KEYWORD2 NAME(CheckFramebufferStatus)(GLenum target)
+{
+   RETURN_DISPATCH(CheckFramebufferStatusEXT, (target), (F, "glCheckFramebufferStatus(0x%x);\n", target));
+}
+
 KEYWORD1 GLenum KEYWORD2 NAME(CheckFramebufferStatusEXT)(GLenum target)
 {
    RETURN_DISPATCH(CheckFramebufferStatusEXT, (target), (F, "glCheckFramebufferStatusEXT(0x%x);\n", target));
 }
 
+KEYWORD1 void KEYWORD2 NAME(DeleteFramebuffers)(GLsizei n, const GLuint * framebuffers)
+{
+   DISPATCH(DeleteFramebuffersEXT, (n, framebuffers), (F, "glDeleteFramebuffers(%d, %p);\n", n, (const void *) framebuffers));
+}
+
 KEYWORD1 void KEYWORD2 NAME(DeleteFramebuffersEXT)(GLsizei n, const GLuint * framebuffers)
 {
    DISPATCH(DeleteFramebuffersEXT, (n, framebuffers), (F, "glDeleteFramebuffersEXT(%d, %p);\n", n, (const void *) framebuffers));
 }
 
+KEYWORD1 void KEYWORD2 NAME(DeleteRenderbuffers)(GLsizei n, const GLuint * renderbuffers)
+{
+   DISPATCH(DeleteRenderbuffersEXT, (n, renderbuffers), (F, "glDeleteRenderbuffers(%d, %p);\n", n, (const void *) renderbuffers));
+}
+
 KEYWORD1 void KEYWORD2 NAME(DeleteRenderbuffersEXT)(GLsizei n, const GLuint * renderbuffers)
 {
    DISPATCH(DeleteRenderbuffersEXT, (n, renderbuffers), (F, "glDeleteRenderbuffersEXT(%d, %p);\n", n, (const void *) renderbuffers));
 }
 
+KEYWORD1 void KEYWORD2 NAME(FramebufferRenderbuffer)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer)
+{
+   DISPATCH(FramebufferRenderbufferEXT, (target, attachment, renderbuffertarget, renderbuffer), (F, "glFramebufferRenderbuffer(0x%x, 0x%x, 0x%x, %d);\n", target, attachment, renderbuffertarget, renderbuffer));
+}
+
 KEYWORD1 void KEYWORD2 NAME(FramebufferRenderbufferEXT)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer)
 {
    DISPATCH(FramebufferRenderbufferEXT, (target, attachment, renderbuffertarget, renderbuffer), (F, "glFramebufferRenderbufferEXT(0x%x, 0x%x, 0x%x, %d);\n", target, attachment, renderbuffertarget, renderbuffer));
 }
 
+KEYWORD1 void KEYWORD2 NAME(FramebufferTexture1D)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level)
+{
+   DISPATCH(FramebufferTexture1DEXT, (target, attachment, textarget, texture, level), (F, "glFramebufferTexture1D(0x%x, 0x%x, 0x%x, %d, %d);\n", target, attachment, textarget, texture, level));
+}
+
 KEYWORD1 void KEYWORD2 NAME(FramebufferTexture1DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level)
 {
    DISPATCH(FramebufferTexture1DEXT, (target, attachment, textarget, texture, level), (F, "glFramebufferTexture1DEXT(0x%x, 0x%x, 0x%x, %d, %d);\n", target, attachment, textarget, texture, level));
 }
 
+KEYWORD1 void KEYWORD2 NAME(FramebufferTexture2D)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level)
+{
+   DISPATCH(FramebufferTexture2DEXT, (target, attachment, textarget, texture, level), (F, "glFramebufferTexture2D(0x%x, 0x%x, 0x%x, %d, %d);\n", target, attachment, textarget, texture, level));
+}
+
 KEYWORD1 void KEYWORD2 NAME(FramebufferTexture2DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level)
 {
    DISPATCH(FramebufferTexture2DEXT, (target, attachment, textarget, texture, level), (F, "glFramebufferTexture2DEXT(0x%x, 0x%x, 0x%x, %d, %d);\n", target, attachment, textarget, texture, level));
 }
 
+KEYWORD1 void KEYWORD2 NAME(FramebufferTexture3D)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset)
+{
+   DISPATCH(FramebufferTexture3DEXT, (target, attachment, textarget, texture, level, zoffset), (F, "glFramebufferTexture3D(0x%x, 0x%x, 0x%x, %d, %d, %d);\n", target, attachment, textarget, texture, level, zoffset));
+}
+
 KEYWORD1 void KEYWORD2 NAME(FramebufferTexture3DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset)
 {
    DISPATCH(FramebufferTexture3DEXT, (target, attachment, textarget, texture, level, zoffset), (F, "glFramebufferTexture3DEXT(0x%x, 0x%x, 0x%x, %d, %d, %d);\n", target, attachment, textarget, texture, level, zoffset));
 }
 
+KEYWORD1 void KEYWORD2 NAME(GenFramebuffers)(GLsizei n, GLuint * framebuffers)
+{
+   DISPATCH(GenFramebuffersEXT, (n, framebuffers), (F, "glGenFramebuffers(%d, %p);\n", n, (const void *) framebuffers));
+}
+
 KEYWORD1 void KEYWORD2 NAME(GenFramebuffersEXT)(GLsizei n, GLuint * framebuffers)
 {
    DISPATCH(GenFramebuffersEXT, (n, framebuffers), (F, "glGenFramebuffersEXT(%d, %p);\n", n, (const void *) framebuffers));
 }
 
+KEYWORD1 void KEYWORD2 NAME(GenRenderbuffers)(GLsizei n, GLuint * renderbuffers)
+{
+   DISPATCH(GenRenderbuffersEXT, (n, renderbuffers), (F, "glGenRenderbuffers(%d, %p);\n", n, (const void *) renderbuffers));
+}
+
 KEYWORD1 void KEYWORD2 NAME(GenRenderbuffersEXT)(GLsizei n, GLuint * renderbuffers)
 {
    DISPATCH(GenRenderbuffersEXT, (n, renderbuffers), (F, "glGenRenderbuffersEXT(%d, %p);\n", n, (const void *) renderbuffers));
 }
 
+KEYWORD1 void KEYWORD2 NAME(GenerateMipmap)(GLenum target)
+{
+   DISPATCH(GenerateMipmapEXT, (target), (F, "glGenerateMipmap(0x%x);\n", target));
+}
+
 KEYWORD1 void KEYWORD2 NAME(GenerateMipmapEXT)(GLenum target)
 {
    DISPATCH(GenerateMipmapEXT, (target), (F, "glGenerateMipmapEXT(0x%x);\n", target));
 }
 
+KEYWORD1 void KEYWORD2 NAME(GetFramebufferAttachmentParameteriv)(GLenum target, GLenum attachment, GLenum pname, GLint * params)
+{
+   DISPATCH(GetFramebufferAttachmentParameterivEXT, (target, attachment, pname, params), (F, "glGetFramebufferAttachmentParameteriv(0x%x, 0x%x, 0x%x, %p);\n", target, attachment, pname, (const void *) params));
+}
+
 KEYWORD1 void KEYWORD2 NAME(GetFramebufferAttachmentParameterivEXT)(GLenum target, GLenum attachment, GLenum pname, GLint * params)
 {
    DISPATCH(GetFramebufferAttachmentParameterivEXT, (target, attachment, pname, params), (F, "glGetFramebufferAttachmentParameterivEXT(0x%x, 0x%x, 0x%x, %p);\n", target, attachment, pname, (const void *) params));
 }
 
+KEYWORD1 void KEYWORD2 NAME(GetRenderbufferParameteriv)(GLenum target, GLenum pname, GLint * params)
+{
+   DISPATCH(GetRenderbufferParameterivEXT, (target, pname, params), (F, "glGetRenderbufferParameteriv(0x%x, 0x%x, %p);\n", target, pname, (const void *) params));
+}
+
 KEYWORD1 void KEYWORD2 NAME(GetRenderbufferParameterivEXT)(GLenum target, GLenum pname, GLint * params)
 {
    DISPATCH(GetRenderbufferParameterivEXT, (target, pname, params), (F, "glGetRenderbufferParameterivEXT(0x%x, 0x%x, %p);\n", target, pname, (const void *) params));
 }
 
+KEYWORD1 GLboolean KEYWORD2 NAME(IsFramebuffer)(GLuint framebuffer)
+{
+   RETURN_DISPATCH(IsFramebufferEXT, (framebuffer), (F, "glIsFramebuffer(%d);\n", framebuffer));
+}
+
 KEYWORD1 GLboolean KEYWORD2 NAME(IsFramebufferEXT)(GLuint framebuffer)
 {
    RETURN_DISPATCH(IsFramebufferEXT, (framebuffer), (F, "glIsFramebufferEXT(%d);\n", framebuffer));
 }
 
+KEYWORD1 GLboolean KEYWORD2 NAME(IsRenderbuffer)(GLuint renderbuffer)
+{
+   RETURN_DISPATCH(IsRenderbufferEXT, (renderbuffer), (F, "glIsRenderbuffer(%d);\n", renderbuffer));
+}
+
 KEYWORD1 GLboolean KEYWORD2 NAME(IsRenderbufferEXT)(GLuint renderbuffer)
 {
    RETURN_DISPATCH(IsRenderbufferEXT, (renderbuffer), (F, "glIsRenderbufferEXT(%d);\n", renderbuffer));
 }
 
+KEYWORD1 void KEYWORD2 NAME(RenderbufferStorage)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height)
+{
+   DISPATCH(RenderbufferStorageEXT, (target, internalformat, width, height), (F, "glRenderbufferStorage(0x%x, 0x%x, %d, %d);\n", target, internalformat, width, height));
+}
+
 KEYWORD1 void KEYWORD2 NAME(RenderbufferStorageEXT)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height)
 {
    DISPATCH(RenderbufferStorageEXT, (target, internalformat, width, height), (F, "glRenderbufferStorageEXT(0x%x, 0x%x, %d, %d);\n", target, internalformat, width, height));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_763)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
+KEYWORD1 void KEYWORD2 NAME(BlitFramebuffer)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter)
+{
+   DISPATCH(BlitFramebufferEXT, (srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter), (F, "glBlitFramebuffer(%d, %d, %d, %d, %d, %d, %d, %d, %d, 0x%x);\n", srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter));
+}
+
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_764)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_763)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_764)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter)
 {
    DISPATCH(BlitFramebufferEXT, (srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter), (F, "glBlitFramebufferEXT(%d, %d, %d, %d, %d, %d, %d, %d, %d, 0x%x);\n", srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter));
 }
 
+KEYWORD1 void KEYWORD2 NAME(FramebufferTextureLayer)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer)
+{
+   DISPATCH(FramebufferTextureLayerEXT, (target, attachment, texture, level, layer), (F, "glFramebufferTextureLayer(0x%x, 0x%x, %d, %d, %d);\n", target, attachment, texture, level, layer));
+}
+
 KEYWORD1 void KEYWORD2 NAME(FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer)
 {
    DISPATCH(FramebufferTextureLayerEXT, (target, attachment, texture, level, layer), (F, "glFramebufferTextureLayerEXT(0x%x, 0x%x, %d, %d, %d);\n", target, attachment, texture, level, layer));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_765)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_766)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_765)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_766)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask)
 {
    DISPATCH(StencilFuncSeparateATI, (frontfunc, backfunc, ref, mask), (F, "glStencilFuncSeparateATI(0x%x, 0x%x, %d, %d);\n", frontfunc, backfunc, ref, mask));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_766)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_767)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_766)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_767)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
 {
    DISPATCH(ProgramEnvParameters4fvEXT, (target, index, count, params), (F, "glProgramEnvParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_767)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_768)(GLenum target, GLuint index, GLsizei count, const GLfloat * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_767)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_768)(GLenum target, GLuint index, GLsizei count, const GLfloat * params)
 {
    DISPATCH(ProgramLocalParameters4fvEXT, (target, index, count, params), (F, "glProgramLocalParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_768)(GLuint id, GLenum pname, GLint64EXT * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLuint id, GLenum pname, GLint64EXT * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_768)(GLuint id, GLenum pname, GLint64EXT * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLuint id, GLenum pname, GLint64EXT * params)
 {
    DISPATCH(GetQueryObjecti64vEXT, (id, pname, params), (F, "glGetQueryObjecti64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params));
 }
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLuint id, GLenum pname, GLuint64EXT * params);
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLuint id, GLenum pname, GLuint64EXT * params);
 
-KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLuint id, GLenum pname, GLuint64EXT * params)
+KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLuint id, GLenum pname, GLuint64EXT * params)
 {
    DISPATCH(GetQueryObjectui64vEXT, (id, pname, params), (F, "glGetQueryObjectui64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params));
 }
@@ -6063,8 +6163,8 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(GetActiveAttribARB),
    TABLE_ENTRY(GetAttribLocationARB),
    TABLE_ENTRY(DrawBuffersARB),
+   TABLE_ENTRY(RenderbufferStorageMultisample),
    TABLE_ENTRY(PolygonOffsetEXT),
-   TABLE_ENTRY(_dispatch_stub_562),
    TABLE_ENTRY(_dispatch_stub_563),
    TABLE_ENTRY(_dispatch_stub_564),
    TABLE_ENTRY(_dispatch_stub_565),
@@ -6072,6 +6172,7 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(_dispatch_stub_567),
    TABLE_ENTRY(_dispatch_stub_568),
    TABLE_ENTRY(_dispatch_stub_569),
+   TABLE_ENTRY(_dispatch_stub_570),
    TABLE_ENTRY(ColorPointerEXT),
    TABLE_ENTRY(EdgeFlagPointerEXT),
    TABLE_ENTRY(IndexPointerEXT),
@@ -6082,8 +6183,8 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(PointParameterfvEXT),
    TABLE_ENTRY(LockArraysEXT),
    TABLE_ENTRY(UnlockArraysEXT),
-   TABLE_ENTRY(_dispatch_stub_580),
    TABLE_ENTRY(_dispatch_stub_581),
+   TABLE_ENTRY(_dispatch_stub_582),
    TABLE_ENTRY(SecondaryColor3bEXT),
    TABLE_ENTRY(SecondaryColor3bvEXT),
    TABLE_ENTRY(SecondaryColor3dEXT),
@@ -6108,7 +6209,7 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(FogCoorddvEXT),
    TABLE_ENTRY(FogCoordfEXT),
    TABLE_ENTRY(FogCoordfvEXT),
-   TABLE_ENTRY(_dispatch_stub_606),
+   TABLE_ENTRY(_dispatch_stub_607),
    TABLE_ENTRY(BlendFuncSeparateEXT),
    TABLE_ENTRY(FlushVertexArrayRangeNV),
    TABLE_ENTRY(VertexArrayRangeNV),
@@ -6150,7 +6251,6 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(WindowPos4ivMESA),
    TABLE_ENTRY(WindowPos4sMESA),
    TABLE_ENTRY(WindowPos4svMESA),
-   TABLE_ENTRY(_dispatch_stub_648),
    TABLE_ENTRY(_dispatch_stub_649),
    TABLE_ENTRY(_dispatch_stub_650),
    TABLE_ENTRY(_dispatch_stub_651),
@@ -6159,6 +6259,7 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(_dispatch_stub_654),
    TABLE_ENTRY(_dispatch_stub_655),
    TABLE_ENTRY(_dispatch_stub_656),
+   TABLE_ENTRY(_dispatch_stub_657),
    TABLE_ENTRY(AreProgramsResidentNV),
    TABLE_ENTRY(BindProgramNV),
    TABLE_ENTRY(DeleteProgramsNV),
@@ -6235,19 +6336,19 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(SetFragmentShaderConstantATI),
    TABLE_ENTRY(PointParameteriNV),
    TABLE_ENTRY(PointParameterivNV),
-   TABLE_ENTRY(_dispatch_stub_733),
    TABLE_ENTRY(_dispatch_stub_734),
    TABLE_ENTRY(_dispatch_stub_735),
    TABLE_ENTRY(_dispatch_stub_736),
    TABLE_ENTRY(_dispatch_stub_737),
+   TABLE_ENTRY(_dispatch_stub_738),
    TABLE_ENTRY(GetProgramNamedParameterdvNV),
    TABLE_ENTRY(GetProgramNamedParameterfvNV),
    TABLE_ENTRY(ProgramNamedParameter4dNV),
    TABLE_ENTRY(ProgramNamedParameter4dvNV),
    TABLE_ENTRY(ProgramNamedParameter4fNV),
    TABLE_ENTRY(ProgramNamedParameter4fvNV),
-   TABLE_ENTRY(_dispatch_stub_744),
    TABLE_ENTRY(_dispatch_stub_745),
+   TABLE_ENTRY(_dispatch_stub_746),
    TABLE_ENTRY(BindFramebufferEXT),
    TABLE_ENTRY(BindRenderbufferEXT),
    TABLE_ENTRY(CheckFramebufferStatusEXT),
@@ -6265,13 +6366,13 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = {
    TABLE_ENTRY(IsFramebufferEXT),
    TABLE_ENTRY(IsRenderbufferEXT),
    TABLE_ENTRY(RenderbufferStorageEXT),
-   TABLE_ENTRY(_dispatch_stub_763),
+   TABLE_ENTRY(_dispatch_stub_764),
    TABLE_ENTRY(FramebufferTextureLayerEXT),
-   TABLE_ENTRY(_dispatch_stub_765),
    TABLE_ENTRY(_dispatch_stub_766),
    TABLE_ENTRY(_dispatch_stub_767),
    TABLE_ENTRY(_dispatch_stub_768),
    TABLE_ENTRY(_dispatch_stub_769),
+   TABLE_ENTRY(_dispatch_stub_770),
    /* A whole bunch of no-op functions.  These might be called
     * when someone tries to call a dynamically-registered
     * extension function without a current rendering context.
@@ -6639,6 +6740,25 @@ static _glapi_proc UNUSED_TABLE_NAME[] = {
    TABLE_ENTRY(PointParameteri),
    TABLE_ENTRY(PointParameteriv),
    TABLE_ENTRY(BlendEquationSeparate),
+   TABLE_ENTRY(BindFramebuffer),
+   TABLE_ENTRY(BindRenderbuffer),
+   TABLE_ENTRY(CheckFramebufferStatus),
+   TABLE_ENTRY(DeleteFramebuffers),
+   TABLE_ENTRY(DeleteRenderbuffers),
+   TABLE_ENTRY(FramebufferRenderbuffer),
+   TABLE_ENTRY(FramebufferTexture1D),
+   TABLE_ENTRY(FramebufferTexture2D),
+   TABLE_ENTRY(FramebufferTexture3D),
+   TABLE_ENTRY(GenFramebuffers),
+   TABLE_ENTRY(GenRenderbuffers),
+   TABLE_ENTRY(GenerateMipmap),
+   TABLE_ENTRY(GetFramebufferAttachmentParameteriv),
+   TABLE_ENTRY(GetRenderbufferParameteriv),
+   TABLE_ENTRY(IsFramebuffer),
+   TABLE_ENTRY(IsRenderbuffer),
+   TABLE_ENTRY(RenderbufferStorage),
+   TABLE_ENTRY(BlitFramebuffer),
+   TABLE_ENTRY(FramebufferTextureLayer),
 };
 #endif /*UNUSED_TABLE_NAME*/
 
diff --git a/src/mesa/glapi/glprocs.h b/src/mesa/glapi/glprocs.h
index 5a99d505ec..ba652bc3c6 100644
--- a/src/mesa/glapi/glprocs.h
+++ b/src/mesa/glapi/glprocs.h
@@ -613,6 +613,7 @@ static const char gl_string_table[] =
     "glGetActiveAttribARB\0"
     "glGetAttribLocationARB\0"
     "glDrawBuffersARB\0"
+    "glRenderbufferStorageMultisample\0"
     "glPolygonOffsetEXT\0"
     "glGetPixelTexGenParameterfvSGIS\0"
     "glGetPixelTexGenParameterivSGIS\0"
@@ -1101,6 +1102,25 @@ static const char gl_string_table[] =
     "glPointParameteriv\0"
     "glBlendEquationSeparate\0"
     "glBlendEquationSeparateATI\0"
+    "glBindFramebuffer\0"
+    "glBindRenderbuffer\0"
+    "glCheckFramebufferStatus\0"
+    "glDeleteFramebuffers\0"
+    "glDeleteRenderbuffers\0"
+    "glFramebufferRenderbuffer\0"
+    "glFramebufferTexture1D\0"
+    "glFramebufferTexture2D\0"
+    "glFramebufferTexture3D\0"
+    "glGenFramebuffers\0"
+    "glGenRenderbuffers\0"
+    "glGenerateMipmap\0"
+    "glGetFramebufferAttachmentParameteriv\0"
+    "glGetRenderbufferParameteriv\0"
+    "glIsFramebuffer\0"
+    "glIsRenderbuffer\0"
+    "glRenderbufferStorage\0"
+    "glBlitFramebuffer\0"
+    "glFramebufferTextureLayer\0"
     ;
 
 
@@ -1118,7 +1138,6 @@ static const char gl_string_table[] =
 #define gl_dispatch_stub_364 mgl_dispatch_stub_364
 #define gl_dispatch_stub_365 mgl_dispatch_stub_365
 #define gl_dispatch_stub_366 mgl_dispatch_stub_366
-#define gl_dispatch_stub_562 mgl_dispatch_stub_562
 #define gl_dispatch_stub_563 mgl_dispatch_stub_563
 #define gl_dispatch_stub_564 mgl_dispatch_stub_564
 #define gl_dispatch_stub_565 mgl_dispatch_stub_565
@@ -1126,10 +1145,10 @@ static const char gl_string_table[] =
 #define gl_dispatch_stub_567 mgl_dispatch_stub_567
 #define gl_dispatch_stub_568 mgl_dispatch_stub_568
 #define gl_dispatch_stub_569 mgl_dispatch_stub_569
-#define gl_dispatch_stub_580 mgl_dispatch_stub_580
+#define gl_dispatch_stub_570 mgl_dispatch_stub_570
 #define gl_dispatch_stub_581 mgl_dispatch_stub_581
-#define gl_dispatch_stub_606 mgl_dispatch_stub_606
-#define gl_dispatch_stub_648 mgl_dispatch_stub_648
+#define gl_dispatch_stub_582 mgl_dispatch_stub_582
+#define gl_dispatch_stub_607 mgl_dispatch_stub_607
 #define gl_dispatch_stub_649 mgl_dispatch_stub_649
 #define gl_dispatch_stub_650 mgl_dispatch_stub_650
 #define gl_dispatch_stub_651 mgl_dispatch_stub_651
@@ -1138,19 +1157,20 @@ static const char gl_string_table[] =
 #define gl_dispatch_stub_654 mgl_dispatch_stub_654
 #define gl_dispatch_stub_655 mgl_dispatch_stub_655
 #define gl_dispatch_stub_656 mgl_dispatch_stub_656
-#define gl_dispatch_stub_733 mgl_dispatch_stub_733
+#define gl_dispatch_stub_657 mgl_dispatch_stub_657
 #define gl_dispatch_stub_734 mgl_dispatch_stub_734
 #define gl_dispatch_stub_735 mgl_dispatch_stub_735
 #define gl_dispatch_stub_736 mgl_dispatch_stub_736
 #define gl_dispatch_stub_737 mgl_dispatch_stub_737
-#define gl_dispatch_stub_744 mgl_dispatch_stub_744
+#define gl_dispatch_stub_738 mgl_dispatch_stub_738
 #define gl_dispatch_stub_745 mgl_dispatch_stub_745
-#define gl_dispatch_stub_763 mgl_dispatch_stub_763
-#define gl_dispatch_stub_765 mgl_dispatch_stub_765
+#define gl_dispatch_stub_746 mgl_dispatch_stub_746
+#define gl_dispatch_stub_764 mgl_dispatch_stub_764
 #define gl_dispatch_stub_766 mgl_dispatch_stub_766
 #define gl_dispatch_stub_767 mgl_dispatch_stub_767
 #define gl_dispatch_stub_768 mgl_dispatch_stub_768
 #define gl_dispatch_stub_769 mgl_dispatch_stub_769
+#define gl_dispatch_stub_770 mgl_dispatch_stub_770
 #endif /* USE_MGL_NAMESPACE */
 
 
@@ -1169,7 +1189,6 @@ extern void gl_dispatch_stub_363(void);
 extern void gl_dispatch_stub_364(void);
 extern void gl_dispatch_stub_365(void);
 extern void gl_dispatch_stub_366(void);
-extern void gl_dispatch_stub_562(void);
 extern void gl_dispatch_stub_563(void);
 extern void gl_dispatch_stub_564(void);
 extern void gl_dispatch_stub_565(void);
@@ -1177,10 +1196,10 @@ extern void gl_dispatch_stub_566(void);
 extern void gl_dispatch_stub_567(void);
 extern void gl_dispatch_stub_568(void);
 extern void gl_dispatch_stub_569(void);
-extern void gl_dispatch_stub_580(void);
+extern void gl_dispatch_stub_570(void);
 extern void gl_dispatch_stub_581(void);
-extern void gl_dispatch_stub_606(void);
-extern void gl_dispatch_stub_648(void);
+extern void gl_dispatch_stub_582(void);
+extern void gl_dispatch_stub_607(void);
 extern void gl_dispatch_stub_649(void);
 extern void gl_dispatch_stub_650(void);
 extern void gl_dispatch_stub_651(void);
@@ -1189,19 +1208,20 @@ extern void gl_dispatch_stub_653(void);
 extern void gl_dispatch_stub_654(void);
 extern void gl_dispatch_stub_655(void);
 extern void gl_dispatch_stub_656(void);
-extern void gl_dispatch_stub_733(void);
+extern void gl_dispatch_stub_657(void);
 extern void gl_dispatch_stub_734(void);
 extern void gl_dispatch_stub_735(void);
 extern void gl_dispatch_stub_736(void);
 extern void gl_dispatch_stub_737(void);
-extern void gl_dispatch_stub_744(void);
+extern void gl_dispatch_stub_738(void);
 extern void gl_dispatch_stub_745(void);
-extern void gl_dispatch_stub_763(void);
-extern void gl_dispatch_stub_765(void);
+extern void gl_dispatch_stub_746(void);
+extern void gl_dispatch_stub_764(void);
 extern void gl_dispatch_stub_766(void);
 extern void gl_dispatch_stub_767(void);
 extern void gl_dispatch_stub_768(void);
 extern void gl_dispatch_stub_769(void);
+extern void gl_dispatch_stub_770(void);
 #endif /* defined(NEED_FUNCTION_POINTER) || defined(GLX_INDIRECT_RENDERING) */
 
 static const glprocs_table_t static_functions[] = {
@@ -1766,494 +1786,514 @@ static const glprocs_table_t static_functions[] = {
     NAME_FUNC_OFFSET( 8890, glGetActiveAttribARB, glGetActiveAttribARB, NULL, _gloffset_GetActiveAttribARB),
     NAME_FUNC_OFFSET( 8911, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB),
     NAME_FUNC_OFFSET( 8934, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
-    NAME_FUNC_OFFSET( 8951, glPolygonOffsetEXT, glPolygonOffsetEXT, NULL, _gloffset_PolygonOffsetEXT),
-    NAME_FUNC_OFFSET( 8970, gl_dispatch_stub_562, gl_dispatch_stub_562, NULL, _gloffset_GetPixelTexGenParameterfvSGIS),
-    NAME_FUNC_OFFSET( 9002, gl_dispatch_stub_563, gl_dispatch_stub_563, NULL, _gloffset_GetPixelTexGenParameterivSGIS),
-    NAME_FUNC_OFFSET( 9034, gl_dispatch_stub_564, gl_dispatch_stub_564, NULL, _gloffset_PixelTexGenParameterfSGIS),
-    NAME_FUNC_OFFSET( 9062, gl_dispatch_stub_565, gl_dispatch_stub_565, NULL, _gloffset_PixelTexGenParameterfvSGIS),
-    NAME_FUNC_OFFSET( 9091, gl_dispatch_stub_566, gl_dispatch_stub_566, NULL, _gloffset_PixelTexGenParameteriSGIS),
-    NAME_FUNC_OFFSET( 9119, gl_dispatch_stub_567, gl_dispatch_stub_567, NULL, _gloffset_PixelTexGenParameterivSGIS),
-    NAME_FUNC_OFFSET( 9148, gl_dispatch_stub_568, gl_dispatch_stub_568, NULL, _gloffset_SampleMaskSGIS),
-    NAME_FUNC_OFFSET( 9165, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_SamplePatternSGIS),
-    NAME_FUNC_OFFSET( 9185, glColorPointerEXT, glColorPointerEXT, NULL, _gloffset_ColorPointerEXT),
-    NAME_FUNC_OFFSET( 9203, glEdgeFlagPointerEXT, glEdgeFlagPointerEXT, NULL, _gloffset_EdgeFlagPointerEXT),
-    NAME_FUNC_OFFSET( 9224, glIndexPointerEXT, glIndexPointerEXT, NULL, _gloffset_IndexPointerEXT),
-    NAME_FUNC_OFFSET( 9242, glNormalPointerEXT, glNormalPointerEXT, NULL, _gloffset_NormalPointerEXT),
-    NAME_FUNC_OFFSET( 9261, glTexCoordPointerEXT, glTexCoordPointerEXT, NULL, _gloffset_TexCoordPointerEXT),
-    NAME_FUNC_OFFSET( 9282, glVertexPointerEXT, glVertexPointerEXT, NULL, _gloffset_VertexPointerEXT),
-    NAME_FUNC_OFFSET( 9301, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET( 9322, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET( 9344, glLockArraysEXT, glLockArraysEXT, NULL, _gloffset_LockArraysEXT),
-    NAME_FUNC_OFFSET( 9360, glUnlockArraysEXT, glUnlockArraysEXT, NULL, _gloffset_UnlockArraysEXT),
-    NAME_FUNC_OFFSET( 9378, gl_dispatch_stub_580, gl_dispatch_stub_580, NULL, _gloffset_CullParameterdvEXT),
-    NAME_FUNC_OFFSET( 9399, gl_dispatch_stub_581, gl_dispatch_stub_581, NULL, _gloffset_CullParameterfvEXT),
-    NAME_FUNC_OFFSET( 9420, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
-    NAME_FUNC_OFFSET( 9442, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
-    NAME_FUNC_OFFSET( 9465, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
-    NAME_FUNC_OFFSET( 9487, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
-    NAME_FUNC_OFFSET( 9510, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
-    NAME_FUNC_OFFSET( 9532, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
-    NAME_FUNC_OFFSET( 9555, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
-    NAME_FUNC_OFFSET( 9577, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
-    NAME_FUNC_OFFSET( 9600, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
-    NAME_FUNC_OFFSET( 9622, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
-    NAME_FUNC_OFFSET( 9645, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
-    NAME_FUNC_OFFSET( 9668, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
-    NAME_FUNC_OFFSET( 9692, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
-    NAME_FUNC_OFFSET( 9715, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
-    NAME_FUNC_OFFSET( 9739, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
-    NAME_FUNC_OFFSET( 9762, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
-    NAME_FUNC_OFFSET( 9786, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
-    NAME_FUNC_OFFSET( 9813, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
-    NAME_FUNC_OFFSET( 9834, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
-    NAME_FUNC_OFFSET( 9857, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
-    NAME_FUNC_OFFSET( 9878, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
-    NAME_FUNC_OFFSET( 9893, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
-    NAME_FUNC_OFFSET( 9909, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
-    NAME_FUNC_OFFSET( 9924, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
-    NAME_FUNC_OFFSET( 9940, gl_dispatch_stub_606, gl_dispatch_stub_606, NULL, _gloffset_PixelTexGenSGIX),
-    NAME_FUNC_OFFSET( 9958, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
-    NAME_FUNC_OFFSET( 9981, glFlushVertexArrayRangeNV, glFlushVertexArrayRangeNV, NULL, _gloffset_FlushVertexArrayRangeNV),
-    NAME_FUNC_OFFSET(10007, glVertexArrayRangeNV, glVertexArrayRangeNV, NULL, _gloffset_VertexArrayRangeNV),
-    NAME_FUNC_OFFSET(10028, glCombinerInputNV, glCombinerInputNV, NULL, _gloffset_CombinerInputNV),
-    NAME_FUNC_OFFSET(10046, glCombinerOutputNV, glCombinerOutputNV, NULL, _gloffset_CombinerOutputNV),
-    NAME_FUNC_OFFSET(10065, glCombinerParameterfNV, glCombinerParameterfNV, NULL, _gloffset_CombinerParameterfNV),
-    NAME_FUNC_OFFSET(10088, glCombinerParameterfvNV, glCombinerParameterfvNV, NULL, _gloffset_CombinerParameterfvNV),
-    NAME_FUNC_OFFSET(10112, glCombinerParameteriNV, glCombinerParameteriNV, NULL, _gloffset_CombinerParameteriNV),
-    NAME_FUNC_OFFSET(10135, glCombinerParameterivNV, glCombinerParameterivNV, NULL, _gloffset_CombinerParameterivNV),
-    NAME_FUNC_OFFSET(10159, glFinalCombinerInputNV, glFinalCombinerInputNV, NULL, _gloffset_FinalCombinerInputNV),
-    NAME_FUNC_OFFSET(10182, glGetCombinerInputParameterfvNV, glGetCombinerInputParameterfvNV, NULL, _gloffset_GetCombinerInputParameterfvNV),
-    NAME_FUNC_OFFSET(10214, glGetCombinerInputParameterivNV, glGetCombinerInputParameterivNV, NULL, _gloffset_GetCombinerInputParameterivNV),
-    NAME_FUNC_OFFSET(10246, glGetCombinerOutputParameterfvNV, glGetCombinerOutputParameterfvNV, NULL, _gloffset_GetCombinerOutputParameterfvNV),
-    NAME_FUNC_OFFSET(10279, glGetCombinerOutputParameterivNV, glGetCombinerOutputParameterivNV, NULL, _gloffset_GetCombinerOutputParameterivNV),
-    NAME_FUNC_OFFSET(10312, glGetFinalCombinerInputParameterfvNV, glGetFinalCombinerInputParameterfvNV, NULL, _gloffset_GetFinalCombinerInputParameterfvNV),
-    NAME_FUNC_OFFSET(10349, glGetFinalCombinerInputParameterivNV, glGetFinalCombinerInputParameterivNV, NULL, _gloffset_GetFinalCombinerInputParameterivNV),
-    NAME_FUNC_OFFSET(10386, glResizeBuffersMESA, glResizeBuffersMESA, NULL, _gloffset_ResizeBuffersMESA),
-    NAME_FUNC_OFFSET(10406, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
-    NAME_FUNC_OFFSET(10424, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
-    NAME_FUNC_OFFSET(10443, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
-    NAME_FUNC_OFFSET(10461, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
-    NAME_FUNC_OFFSET(10480, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
-    NAME_FUNC_OFFSET(10498, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
-    NAME_FUNC_OFFSET(10517, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
-    NAME_FUNC_OFFSET(10535, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
-    NAME_FUNC_OFFSET(10554, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
-    NAME_FUNC_OFFSET(10572, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
-    NAME_FUNC_OFFSET(10591, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
-    NAME_FUNC_OFFSET(10609, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
-    NAME_FUNC_OFFSET(10628, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
-    NAME_FUNC_OFFSET(10646, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
-    NAME_FUNC_OFFSET(10665, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
-    NAME_FUNC_OFFSET(10683, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
-    NAME_FUNC_OFFSET(10702, glWindowPos4dMESA, glWindowPos4dMESA, NULL, _gloffset_WindowPos4dMESA),
-    NAME_FUNC_OFFSET(10720, glWindowPos4dvMESA, glWindowPos4dvMESA, NULL, _gloffset_WindowPos4dvMESA),
-    NAME_FUNC_OFFSET(10739, glWindowPos4fMESA, glWindowPos4fMESA, NULL, _gloffset_WindowPos4fMESA),
-    NAME_FUNC_OFFSET(10757, glWindowPos4fvMESA, glWindowPos4fvMESA, NULL, _gloffset_WindowPos4fvMESA),
-    NAME_FUNC_OFFSET(10776, glWindowPos4iMESA, glWindowPos4iMESA, NULL, _gloffset_WindowPos4iMESA),
-    NAME_FUNC_OFFSET(10794, glWindowPos4ivMESA, glWindowPos4ivMESA, NULL, _gloffset_WindowPos4ivMESA),
-    NAME_FUNC_OFFSET(10813, glWindowPos4sMESA, glWindowPos4sMESA, NULL, _gloffset_WindowPos4sMESA),
-    NAME_FUNC_OFFSET(10831, glWindowPos4svMESA, glWindowPos4svMESA, NULL, _gloffset_WindowPos4svMESA),
-    NAME_FUNC_OFFSET(10850, gl_dispatch_stub_648, gl_dispatch_stub_648, NULL, _gloffset_MultiModeDrawArraysIBM),
-    NAME_FUNC_OFFSET(10875, gl_dispatch_stub_649, gl_dispatch_stub_649, NULL, _gloffset_MultiModeDrawElementsIBM),
-    NAME_FUNC_OFFSET(10902, gl_dispatch_stub_650, gl_dispatch_stub_650, NULL, _gloffset_DeleteFencesNV),
-    NAME_FUNC_OFFSET(10919, gl_dispatch_stub_651, gl_dispatch_stub_651, NULL, _gloffset_FinishFenceNV),
-    NAME_FUNC_OFFSET(10935, gl_dispatch_stub_652, gl_dispatch_stub_652, NULL, _gloffset_GenFencesNV),
-    NAME_FUNC_OFFSET(10949, gl_dispatch_stub_653, gl_dispatch_stub_653, NULL, _gloffset_GetFenceivNV),
-    NAME_FUNC_OFFSET(10964, gl_dispatch_stub_654, gl_dispatch_stub_654, NULL, _gloffset_IsFenceNV),
-    NAME_FUNC_OFFSET(10976, gl_dispatch_stub_655, gl_dispatch_stub_655, NULL, _gloffset_SetFenceNV),
-    NAME_FUNC_OFFSET(10989, gl_dispatch_stub_656, gl_dispatch_stub_656, NULL, _gloffset_TestFenceNV),
-    NAME_FUNC_OFFSET(11003, glAreProgramsResidentNV, glAreProgramsResidentNV, NULL, _gloffset_AreProgramsResidentNV),
-    NAME_FUNC_OFFSET(11027, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
-    NAME_FUNC_OFFSET(11043, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
-    NAME_FUNC_OFFSET(11062, glExecuteProgramNV, glExecuteProgramNV, NULL, _gloffset_ExecuteProgramNV),
-    NAME_FUNC_OFFSET(11081, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
-    NAME_FUNC_OFFSET(11097, glGetProgramParameterdvNV, glGetProgramParameterdvNV, NULL, _gloffset_GetProgramParameterdvNV),
-    NAME_FUNC_OFFSET(11123, glGetProgramParameterfvNV, glGetProgramParameterfvNV, NULL, _gloffset_GetProgramParameterfvNV),
-    NAME_FUNC_OFFSET(11149, glGetProgramStringNV, glGetProgramStringNV, NULL, _gloffset_GetProgramStringNV),
-    NAME_FUNC_OFFSET(11170, glGetProgramivNV, glGetProgramivNV, NULL, _gloffset_GetProgramivNV),
-    NAME_FUNC_OFFSET(11187, glGetTrackMatrixivNV, glGetTrackMatrixivNV, NULL, _gloffset_GetTrackMatrixivNV),
-    NAME_FUNC_OFFSET(11208, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
-    NAME_FUNC_OFFSET(11236, glGetVertexAttribdvNV, glGetVertexAttribdvNV, NULL, _gloffset_GetVertexAttribdvNV),
-    NAME_FUNC_OFFSET(11258, glGetVertexAttribfvNV, glGetVertexAttribfvNV, NULL, _gloffset_GetVertexAttribfvNV),
-    NAME_FUNC_OFFSET(11280, glGetVertexAttribivNV, glGetVertexAttribivNV, NULL, _gloffset_GetVertexAttribivNV),
-    NAME_FUNC_OFFSET(11302, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
-    NAME_FUNC_OFFSET(11316, glLoadProgramNV, glLoadProgramNV, NULL, _gloffset_LoadProgramNV),
-    NAME_FUNC_OFFSET(11332, glProgramParameters4dvNV, glProgramParameters4dvNV, NULL, _gloffset_ProgramParameters4dvNV),
-    NAME_FUNC_OFFSET(11357, glProgramParameters4fvNV, glProgramParameters4fvNV, NULL, _gloffset_ProgramParameters4fvNV),
-    NAME_FUNC_OFFSET(11382, glRequestResidentProgramsNV, glRequestResidentProgramsNV, NULL, _gloffset_RequestResidentProgramsNV),
-    NAME_FUNC_OFFSET(11410, glTrackMatrixNV, glTrackMatrixNV, NULL, _gloffset_TrackMatrixNV),
-    NAME_FUNC_OFFSET(11426, glVertexAttrib1dNV, glVertexAttrib1dNV, NULL, _gloffset_VertexAttrib1dNV),
-    NAME_FUNC_OFFSET(11445, glVertexAttrib1dvNV, glVertexAttrib1dvNV, NULL, _gloffset_VertexAttrib1dvNV),
-    NAME_FUNC_OFFSET(11465, glVertexAttrib1fNV, glVertexAttrib1fNV, NULL, _gloffset_VertexAttrib1fNV),
-    NAME_FUNC_OFFSET(11484, glVertexAttrib1fvNV, glVertexAttrib1fvNV, NULL, _gloffset_VertexAttrib1fvNV),
-    NAME_FUNC_OFFSET(11504, glVertexAttrib1sNV, glVertexAttrib1sNV, NULL, _gloffset_VertexAttrib1sNV),
-    NAME_FUNC_OFFSET(11523, glVertexAttrib1svNV, glVertexAttrib1svNV, NULL, _gloffset_VertexAttrib1svNV),
-    NAME_FUNC_OFFSET(11543, glVertexAttrib2dNV, glVertexAttrib2dNV, NULL, _gloffset_VertexAttrib2dNV),
-    NAME_FUNC_OFFSET(11562, glVertexAttrib2dvNV, glVertexAttrib2dvNV, NULL, _gloffset_VertexAttrib2dvNV),
-    NAME_FUNC_OFFSET(11582, glVertexAttrib2fNV, glVertexAttrib2fNV, NULL, _gloffset_VertexAttrib2fNV),
-    NAME_FUNC_OFFSET(11601, glVertexAttrib2fvNV, glVertexAttrib2fvNV, NULL, _gloffset_VertexAttrib2fvNV),
-    NAME_FUNC_OFFSET(11621, glVertexAttrib2sNV, glVertexAttrib2sNV, NULL, _gloffset_VertexAttrib2sNV),
-    NAME_FUNC_OFFSET(11640, glVertexAttrib2svNV, glVertexAttrib2svNV, NULL, _gloffset_VertexAttrib2svNV),
-    NAME_FUNC_OFFSET(11660, glVertexAttrib3dNV, glVertexAttrib3dNV, NULL, _gloffset_VertexAttrib3dNV),
-    NAME_FUNC_OFFSET(11679, glVertexAttrib3dvNV, glVertexAttrib3dvNV, NULL, _gloffset_VertexAttrib3dvNV),
-    NAME_FUNC_OFFSET(11699, glVertexAttrib3fNV, glVertexAttrib3fNV, NULL, _gloffset_VertexAttrib3fNV),
-    NAME_FUNC_OFFSET(11718, glVertexAttrib3fvNV, glVertexAttrib3fvNV, NULL, _gloffset_VertexAttrib3fvNV),
-    NAME_FUNC_OFFSET(11738, glVertexAttrib3sNV, glVertexAttrib3sNV, NULL, _gloffset_VertexAttrib3sNV),
-    NAME_FUNC_OFFSET(11757, glVertexAttrib3svNV, glVertexAttrib3svNV, NULL, _gloffset_VertexAttrib3svNV),
-    NAME_FUNC_OFFSET(11777, glVertexAttrib4dNV, glVertexAttrib4dNV, NULL, _gloffset_VertexAttrib4dNV),
-    NAME_FUNC_OFFSET(11796, glVertexAttrib4dvNV, glVertexAttrib4dvNV, NULL, _gloffset_VertexAttrib4dvNV),
-    NAME_FUNC_OFFSET(11816, glVertexAttrib4fNV, glVertexAttrib4fNV, NULL, _gloffset_VertexAttrib4fNV),
-    NAME_FUNC_OFFSET(11835, glVertexAttrib4fvNV, glVertexAttrib4fvNV, NULL, _gloffset_VertexAttrib4fvNV),
-    NAME_FUNC_OFFSET(11855, glVertexAttrib4sNV, glVertexAttrib4sNV, NULL, _gloffset_VertexAttrib4sNV),
-    NAME_FUNC_OFFSET(11874, glVertexAttrib4svNV, glVertexAttrib4svNV, NULL, _gloffset_VertexAttrib4svNV),
-    NAME_FUNC_OFFSET(11894, glVertexAttrib4ubNV, glVertexAttrib4ubNV, NULL, _gloffset_VertexAttrib4ubNV),
-    NAME_FUNC_OFFSET(11914, glVertexAttrib4ubvNV, glVertexAttrib4ubvNV, NULL, _gloffset_VertexAttrib4ubvNV),
-    NAME_FUNC_OFFSET(11935, glVertexAttribPointerNV, glVertexAttribPointerNV, NULL, _gloffset_VertexAttribPointerNV),
-    NAME_FUNC_OFFSET(11959, glVertexAttribs1dvNV, glVertexAttribs1dvNV, NULL, _gloffset_VertexAttribs1dvNV),
-    NAME_FUNC_OFFSET(11980, glVertexAttribs1fvNV, glVertexAttribs1fvNV, NULL, _gloffset_VertexAttribs1fvNV),
-    NAME_FUNC_OFFSET(12001, glVertexAttribs1svNV, glVertexAttribs1svNV, NULL, _gloffset_VertexAttribs1svNV),
-    NAME_FUNC_OFFSET(12022, glVertexAttribs2dvNV, glVertexAttribs2dvNV, NULL, _gloffset_VertexAttribs2dvNV),
-    NAME_FUNC_OFFSET(12043, glVertexAttribs2fvNV, glVertexAttribs2fvNV, NULL, _gloffset_VertexAttribs2fvNV),
-    NAME_FUNC_OFFSET(12064, glVertexAttribs2svNV, glVertexAttribs2svNV, NULL, _gloffset_VertexAttribs2svNV),
-    NAME_FUNC_OFFSET(12085, glVertexAttribs3dvNV, glVertexAttribs3dvNV, NULL, _gloffset_VertexAttribs3dvNV),
-    NAME_FUNC_OFFSET(12106, glVertexAttribs3fvNV, glVertexAttribs3fvNV, NULL, _gloffset_VertexAttribs3fvNV),
-    NAME_FUNC_OFFSET(12127, glVertexAttribs3svNV, glVertexAttribs3svNV, NULL, _gloffset_VertexAttribs3svNV),
-    NAME_FUNC_OFFSET(12148, glVertexAttribs4dvNV, glVertexAttribs4dvNV, NULL, _gloffset_VertexAttribs4dvNV),
-    NAME_FUNC_OFFSET(12169, glVertexAttribs4fvNV, glVertexAttribs4fvNV, NULL, _gloffset_VertexAttribs4fvNV),
-    NAME_FUNC_OFFSET(12190, glVertexAttribs4svNV, glVertexAttribs4svNV, NULL, _gloffset_VertexAttribs4svNV),
-    NAME_FUNC_OFFSET(12211, glVertexAttribs4ubvNV, glVertexAttribs4ubvNV, NULL, _gloffset_VertexAttribs4ubvNV),
-    NAME_FUNC_OFFSET(12233, glAlphaFragmentOp1ATI, glAlphaFragmentOp1ATI, NULL, _gloffset_AlphaFragmentOp1ATI),
-    NAME_FUNC_OFFSET(12255, glAlphaFragmentOp2ATI, glAlphaFragmentOp2ATI, NULL, _gloffset_AlphaFragmentOp2ATI),
-    NAME_FUNC_OFFSET(12277, glAlphaFragmentOp3ATI, glAlphaFragmentOp3ATI, NULL, _gloffset_AlphaFragmentOp3ATI),
-    NAME_FUNC_OFFSET(12299, glBeginFragmentShaderATI, glBeginFragmentShaderATI, NULL, _gloffset_BeginFragmentShaderATI),
-    NAME_FUNC_OFFSET(12324, glBindFragmentShaderATI, glBindFragmentShaderATI, NULL, _gloffset_BindFragmentShaderATI),
-    NAME_FUNC_OFFSET(12348, glColorFragmentOp1ATI, glColorFragmentOp1ATI, NULL, _gloffset_ColorFragmentOp1ATI),
-    NAME_FUNC_OFFSET(12370, glColorFragmentOp2ATI, glColorFragmentOp2ATI, NULL, _gloffset_ColorFragmentOp2ATI),
-    NAME_FUNC_OFFSET(12392, glColorFragmentOp3ATI, glColorFragmentOp3ATI, NULL, _gloffset_ColorFragmentOp3ATI),
-    NAME_FUNC_OFFSET(12414, glDeleteFragmentShaderATI, glDeleteFragmentShaderATI, NULL, _gloffset_DeleteFragmentShaderATI),
-    NAME_FUNC_OFFSET(12440, glEndFragmentShaderATI, glEndFragmentShaderATI, NULL, _gloffset_EndFragmentShaderATI),
-    NAME_FUNC_OFFSET(12463, glGenFragmentShadersATI, glGenFragmentShadersATI, NULL, _gloffset_GenFragmentShadersATI),
-    NAME_FUNC_OFFSET(12487, glPassTexCoordATI, glPassTexCoordATI, NULL, _gloffset_PassTexCoordATI),
-    NAME_FUNC_OFFSET(12505, glSampleMapATI, glSampleMapATI, NULL, _gloffset_SampleMapATI),
-    NAME_FUNC_OFFSET(12520, glSetFragmentShaderConstantATI, glSetFragmentShaderConstantATI, NULL, _gloffset_SetFragmentShaderConstantATI),
-    NAME_FUNC_OFFSET(12551, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
-    NAME_FUNC_OFFSET(12571, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
-    NAME_FUNC_OFFSET(12592, gl_dispatch_stub_733, gl_dispatch_stub_733, NULL, _gloffset_ActiveStencilFaceEXT),
-    NAME_FUNC_OFFSET(12615, gl_dispatch_stub_734, gl_dispatch_stub_734, NULL, _gloffset_BindVertexArrayAPPLE),
-    NAME_FUNC_OFFSET(12638, gl_dispatch_stub_735, gl_dispatch_stub_735, NULL, _gloffset_DeleteVertexArraysAPPLE),
-    NAME_FUNC_OFFSET(12664, gl_dispatch_stub_736, gl_dispatch_stub_736, NULL, _gloffset_GenVertexArraysAPPLE),
-    NAME_FUNC_OFFSET(12687, gl_dispatch_stub_737, gl_dispatch_stub_737, NULL, _gloffset_IsVertexArrayAPPLE),
-    NAME_FUNC_OFFSET(12708, glGetProgramNamedParameterdvNV, glGetProgramNamedParameterdvNV, NULL, _gloffset_GetProgramNamedParameterdvNV),
-    NAME_FUNC_OFFSET(12739, glGetProgramNamedParameterfvNV, glGetProgramNamedParameterfvNV, NULL, _gloffset_GetProgramNamedParameterfvNV),
-    NAME_FUNC_OFFSET(12770, glProgramNamedParameter4dNV, glProgramNamedParameter4dNV, NULL, _gloffset_ProgramNamedParameter4dNV),
-    NAME_FUNC_OFFSET(12798, glProgramNamedParameter4dvNV, glProgramNamedParameter4dvNV, NULL, _gloffset_ProgramNamedParameter4dvNV),
-    NAME_FUNC_OFFSET(12827, glProgramNamedParameter4fNV, glProgramNamedParameter4fNV, NULL, _gloffset_ProgramNamedParameter4fNV),
-    NAME_FUNC_OFFSET(12855, glProgramNamedParameter4fvNV, glProgramNamedParameter4fvNV, NULL, _gloffset_ProgramNamedParameter4fvNV),
-    NAME_FUNC_OFFSET(12884, gl_dispatch_stub_744, gl_dispatch_stub_744, NULL, _gloffset_DepthBoundsEXT),
-    NAME_FUNC_OFFSET(12901, gl_dispatch_stub_745, gl_dispatch_stub_745, NULL, _gloffset_BlendEquationSeparateEXT),
-    NAME_FUNC_OFFSET(12928, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
-    NAME_FUNC_OFFSET(12949, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
-    NAME_FUNC_OFFSET(12971, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
-    NAME_FUNC_OFFSET(12999, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
-    NAME_FUNC_OFFSET(13023, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
-    NAME_FUNC_OFFSET(13048, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
-    NAME_FUNC_OFFSET(13077, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
-    NAME_FUNC_OFFSET(13103, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
-    NAME_FUNC_OFFSET(13129, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
-    NAME_FUNC_OFFSET(13155, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
-    NAME_FUNC_OFFSET(13176, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
-    NAME_FUNC_OFFSET(13198, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
-    NAME_FUNC_OFFSET(13218, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
-    NAME_FUNC_OFFSET(13259, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
-    NAME_FUNC_OFFSET(13291, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
-    NAME_FUNC_OFFSET(13310, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
-    NAME_FUNC_OFFSET(13330, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
-    NAME_FUNC_OFFSET(13355, gl_dispatch_stub_763, gl_dispatch_stub_763, NULL, _gloffset_BlitFramebufferEXT),
-    NAME_FUNC_OFFSET(13376, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
-    NAME_FUNC_OFFSET(13405, gl_dispatch_stub_765, gl_dispatch_stub_765, NULL, _gloffset_StencilFuncSeparateATI),
-    NAME_FUNC_OFFSET(13430, gl_dispatch_stub_766, gl_dispatch_stub_766, NULL, _gloffset_ProgramEnvParameters4fvEXT),
-    NAME_FUNC_OFFSET(13459, gl_dispatch_stub_767, gl_dispatch_stub_767, NULL, _gloffset_ProgramLocalParameters4fvEXT),
-    NAME_FUNC_OFFSET(13490, gl_dispatch_stub_768, gl_dispatch_stub_768, NULL, _gloffset_GetQueryObjecti64vEXT),
-    NAME_FUNC_OFFSET(13514, gl_dispatch_stub_769, gl_dispatch_stub_769, NULL, _gloffset_GetQueryObjectui64vEXT),
-    NAME_FUNC_OFFSET(13539, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement),
-    NAME_FUNC_OFFSET(13557, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture),
-    NAME_FUNC_OFFSET(13574, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays),
-    NAME_FUNC_OFFSET(13590, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident),
-    NAME_FUNC_OFFSET(13615, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D),
-    NAME_FUNC_OFFSET(13635, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D),
-    NAME_FUNC_OFFSET(13655, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D),
-    NAME_FUNC_OFFSET(13678, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D),
-    NAME_FUNC_OFFSET(13701, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures),
-    NAME_FUNC_OFFSET(13721, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures),
-    NAME_FUNC_OFFSET(13738, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv),
-    NAME_FUNC_OFFSET(13755, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture),
-    NAME_FUNC_OFFSET(13770, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures),
-    NAME_FUNC_OFFSET(13794, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D),
-    NAME_FUNC_OFFSET(13813, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D),
-    NAME_FUNC_OFFSET(13832, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor),
-    NAME_FUNC_OFFSET(13848, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation),
-    NAME_FUNC_OFFSET(13867, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements),
-    NAME_FUNC_OFFSET(13890, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
-    NAME_FUNC_OFFSET(13906, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
-    NAME_FUNC_OFFSET(13922, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv),
-    NAME_FUNC_OFFSET(13949, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv),
-    NAME_FUNC_OFFSET(13976, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable),
-    NAME_FUNC_OFFSET(13996, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
-    NAME_FUNC_OFFSET(14015, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
-    NAME_FUNC_OFFSET(14034, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
-    NAME_FUNC_OFFSET(14064, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
-    NAME_FUNC_OFFSET(14094, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
-    NAME_FUNC_OFFSET(14124, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
-    NAME_FUNC_OFFSET(14154, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable),
-    NAME_FUNC_OFFSET(14173, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable),
-    NAME_FUNC_OFFSET(14196, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D),
-    NAME_FUNC_OFFSET(14221, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D),
-    NAME_FUNC_OFFSET(14246, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf),
-    NAME_FUNC_OFFSET(14273, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv),
-    NAME_FUNC_OFFSET(14301, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri),
-    NAME_FUNC_OFFSET(14328, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv),
-    NAME_FUNC_OFFSET(14356, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D),
-    NAME_FUNC_OFFSET(14385, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D),
-    NAME_FUNC_OFFSET(14414, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter),
-    NAME_FUNC_OFFSET(14440, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv),
-    NAME_FUNC_OFFSET(14471, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv),
-    NAME_FUNC_OFFSET(14502, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter),
-    NAME_FUNC_OFFSET(14526, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D),
-    NAME_FUNC_OFFSET(14549, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram),
-    NAME_FUNC_OFFSET(14567, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv),
-    NAME_FUNC_OFFSET(14596, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv),
-    NAME_FUNC_OFFSET(14625, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax),
-    NAME_FUNC_OFFSET(14640, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv),
-    NAME_FUNC_OFFSET(14666, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv),
-    NAME_FUNC_OFFSET(14692, glHistogram, glHistogram, NULL, _gloffset_Histogram),
-    NAME_FUNC_OFFSET(14707, glMinmax, glMinmax, NULL, _gloffset_Minmax),
-    NAME_FUNC_OFFSET(14719, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram),
-    NAME_FUNC_OFFSET(14739, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax),
-    NAME_FUNC_OFFSET(14756, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D),
-    NAME_FUNC_OFFSET(14772, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D),
-    NAME_FUNC_OFFSET(14791, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D),
-    NAME_FUNC_OFFSET(14814, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB),
-    NAME_FUNC_OFFSET(14830, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB),
-    NAME_FUNC_OFFSET(14852, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB),
-    NAME_FUNC_OFFSET(14870, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB),
-    NAME_FUNC_OFFSET(14889, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB),
-    NAME_FUNC_OFFSET(14907, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB),
-    NAME_FUNC_OFFSET(14926, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB),
-    NAME_FUNC_OFFSET(14944, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB),
-    NAME_FUNC_OFFSET(14963, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB),
-    NAME_FUNC_OFFSET(14981, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB),
-    NAME_FUNC_OFFSET(15000, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB),
-    NAME_FUNC_OFFSET(15018, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB),
-    NAME_FUNC_OFFSET(15037, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB),
-    NAME_FUNC_OFFSET(15055, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB),
-    NAME_FUNC_OFFSET(15074, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB),
-    NAME_FUNC_OFFSET(15092, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB),
-    NAME_FUNC_OFFSET(15111, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB),
-    NAME_FUNC_OFFSET(15129, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB),
-    NAME_FUNC_OFFSET(15148, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB),
-    NAME_FUNC_OFFSET(15166, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB),
-    NAME_FUNC_OFFSET(15185, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB),
-    NAME_FUNC_OFFSET(15203, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB),
-    NAME_FUNC_OFFSET(15222, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB),
-    NAME_FUNC_OFFSET(15240, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB),
-    NAME_FUNC_OFFSET(15259, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB),
-    NAME_FUNC_OFFSET(15277, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB),
-    NAME_FUNC_OFFSET(15296, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB),
-    NAME_FUNC_OFFSET(15314, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB),
-    NAME_FUNC_OFFSET(15333, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB),
-    NAME_FUNC_OFFSET(15351, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB),
-    NAME_FUNC_OFFSET(15370, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB),
-    NAME_FUNC_OFFSET(15388, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB),
-    NAME_FUNC_OFFSET(15407, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB),
-    NAME_FUNC_OFFSET(15425, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB),
-    NAME_FUNC_OFFSET(15444, glStencilOpSeparate, glStencilOpSeparate, NULL, _gloffset_StencilOpSeparate),
-    NAME_FUNC_OFFSET(15467, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB),
-    NAME_FUNC_OFFSET(15490, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB),
-    NAME_FUNC_OFFSET(15513, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB),
-    NAME_FUNC_OFFSET(15536, glMultTransposeMatrixfARB, glMultTransposeMatrixfARB, NULL, _gloffset_MultTransposeMatrixfARB),
-    NAME_FUNC_OFFSET(15559, glSampleCoverageARB, glSampleCoverageARB, NULL, _gloffset_SampleCoverageARB),
-    NAME_FUNC_OFFSET(15576, glCompressedTexImage1DARB, glCompressedTexImage1DARB, NULL, _gloffset_CompressedTexImage1DARB),
-    NAME_FUNC_OFFSET(15599, glCompressedTexImage2DARB, glCompressedTexImage2DARB, NULL, _gloffset_CompressedTexImage2DARB),
-    NAME_FUNC_OFFSET(15622, glCompressedTexImage3DARB, glCompressedTexImage3DARB, NULL, _gloffset_CompressedTexImage3DARB),
-    NAME_FUNC_OFFSET(15645, glCompressedTexSubImage1DARB, glCompressedTexSubImage1DARB, NULL, _gloffset_CompressedTexSubImage1DARB),
-    NAME_FUNC_OFFSET(15671, glCompressedTexSubImage2DARB, glCompressedTexSubImage2DARB, NULL, _gloffset_CompressedTexSubImage2DARB),
-    NAME_FUNC_OFFSET(15697, glCompressedTexSubImage3DARB, glCompressedTexSubImage3DARB, NULL, _gloffset_CompressedTexSubImage3DARB),
-    NAME_FUNC_OFFSET(15723, glGetCompressedTexImageARB, glGetCompressedTexImageARB, NULL, _gloffset_GetCompressedTexImageARB),
-    NAME_FUNC_OFFSET(15747, glDisableVertexAttribArrayARB, glDisableVertexAttribArrayARB, NULL, _gloffset_DisableVertexAttribArrayARB),
-    NAME_FUNC_OFFSET(15774, glEnableVertexAttribArrayARB, glEnableVertexAttribArrayARB, NULL, _gloffset_EnableVertexAttribArrayARB),
-    NAME_FUNC_OFFSET(15800, glGetVertexAttribdvARB, glGetVertexAttribdvARB, NULL, _gloffset_GetVertexAttribdvARB),
-    NAME_FUNC_OFFSET(15820, glGetVertexAttribfvARB, glGetVertexAttribfvARB, NULL, _gloffset_GetVertexAttribfvARB),
-    NAME_FUNC_OFFSET(15840, glGetVertexAttribivARB, glGetVertexAttribivARB, NULL, _gloffset_GetVertexAttribivARB),
-    NAME_FUNC_OFFSET(15860, glProgramEnvParameter4dARB, glProgramEnvParameter4dARB, NULL, _gloffset_ProgramEnvParameter4dARB),
-    NAME_FUNC_OFFSET(15883, glProgramEnvParameter4dvARB, glProgramEnvParameter4dvARB, NULL, _gloffset_ProgramEnvParameter4dvARB),
-    NAME_FUNC_OFFSET(15907, glProgramEnvParameter4fARB, glProgramEnvParameter4fARB, NULL, _gloffset_ProgramEnvParameter4fARB),
-    NAME_FUNC_OFFSET(15930, glProgramEnvParameter4fvARB, glProgramEnvParameter4fvARB, NULL, _gloffset_ProgramEnvParameter4fvARB),
-    NAME_FUNC_OFFSET(15954, glVertexAttrib1dARB, glVertexAttrib1dARB, NULL, _gloffset_VertexAttrib1dARB),
-    NAME_FUNC_OFFSET(15971, glVertexAttrib1dvARB, glVertexAttrib1dvARB, NULL, _gloffset_VertexAttrib1dvARB),
-    NAME_FUNC_OFFSET(15989, glVertexAttrib1fARB, glVertexAttrib1fARB, NULL, _gloffset_VertexAttrib1fARB),
-    NAME_FUNC_OFFSET(16006, glVertexAttrib1fvARB, glVertexAttrib1fvARB, NULL, _gloffset_VertexAttrib1fvARB),
-    NAME_FUNC_OFFSET(16024, glVertexAttrib1sARB, glVertexAttrib1sARB, NULL, _gloffset_VertexAttrib1sARB),
-    NAME_FUNC_OFFSET(16041, glVertexAttrib1svARB, glVertexAttrib1svARB, NULL, _gloffset_VertexAttrib1svARB),
-    NAME_FUNC_OFFSET(16059, glVertexAttrib2dARB, glVertexAttrib2dARB, NULL, _gloffset_VertexAttrib2dARB),
-    NAME_FUNC_OFFSET(16076, glVertexAttrib2dvARB, glVertexAttrib2dvARB, NULL, _gloffset_VertexAttrib2dvARB),
-    NAME_FUNC_OFFSET(16094, glVertexAttrib2fARB, glVertexAttrib2fARB, NULL, _gloffset_VertexAttrib2fARB),
-    NAME_FUNC_OFFSET(16111, glVertexAttrib2fvARB, glVertexAttrib2fvARB, NULL, _gloffset_VertexAttrib2fvARB),
-    NAME_FUNC_OFFSET(16129, glVertexAttrib2sARB, glVertexAttrib2sARB, NULL, _gloffset_VertexAttrib2sARB),
-    NAME_FUNC_OFFSET(16146, glVertexAttrib2svARB, glVertexAttrib2svARB, NULL, _gloffset_VertexAttrib2svARB),
-    NAME_FUNC_OFFSET(16164, glVertexAttrib3dARB, glVertexAttrib3dARB, NULL, _gloffset_VertexAttrib3dARB),
-    NAME_FUNC_OFFSET(16181, glVertexAttrib3dvARB, glVertexAttrib3dvARB, NULL, _gloffset_VertexAttrib3dvARB),
-    NAME_FUNC_OFFSET(16199, glVertexAttrib3fARB, glVertexAttrib3fARB, NULL, _gloffset_VertexAttrib3fARB),
-    NAME_FUNC_OFFSET(16216, glVertexAttrib3fvARB, glVertexAttrib3fvARB, NULL, _gloffset_VertexAttrib3fvARB),
-    NAME_FUNC_OFFSET(16234, glVertexAttrib3sARB, glVertexAttrib3sARB, NULL, _gloffset_VertexAttrib3sARB),
-    NAME_FUNC_OFFSET(16251, glVertexAttrib3svARB, glVertexAttrib3svARB, NULL, _gloffset_VertexAttrib3svARB),
-    NAME_FUNC_OFFSET(16269, glVertexAttrib4NbvARB, glVertexAttrib4NbvARB, NULL, _gloffset_VertexAttrib4NbvARB),
-    NAME_FUNC_OFFSET(16288, glVertexAttrib4NivARB, glVertexAttrib4NivARB, NULL, _gloffset_VertexAttrib4NivARB),
-    NAME_FUNC_OFFSET(16307, glVertexAttrib4NsvARB, glVertexAttrib4NsvARB, NULL, _gloffset_VertexAttrib4NsvARB),
-    NAME_FUNC_OFFSET(16326, glVertexAttrib4NubARB, glVertexAttrib4NubARB, NULL, _gloffset_VertexAttrib4NubARB),
-    NAME_FUNC_OFFSET(16345, glVertexAttrib4NubvARB, glVertexAttrib4NubvARB, NULL, _gloffset_VertexAttrib4NubvARB),
-    NAME_FUNC_OFFSET(16365, glVertexAttrib4NuivARB, glVertexAttrib4NuivARB, NULL, _gloffset_VertexAttrib4NuivARB),
-    NAME_FUNC_OFFSET(16385, glVertexAttrib4NusvARB, glVertexAttrib4NusvARB, NULL, _gloffset_VertexAttrib4NusvARB),
-    NAME_FUNC_OFFSET(16405, glVertexAttrib4bvARB, glVertexAttrib4bvARB, NULL, _gloffset_VertexAttrib4bvARB),
-    NAME_FUNC_OFFSET(16423, glVertexAttrib4dARB, glVertexAttrib4dARB, NULL, _gloffset_VertexAttrib4dARB),
-    NAME_FUNC_OFFSET(16440, glVertexAttrib4dvARB, glVertexAttrib4dvARB, NULL, _gloffset_VertexAttrib4dvARB),
-    NAME_FUNC_OFFSET(16458, glVertexAttrib4fARB, glVertexAttrib4fARB, NULL, _gloffset_VertexAttrib4fARB),
-    NAME_FUNC_OFFSET(16475, glVertexAttrib4fvARB, glVertexAttrib4fvARB, NULL, _gloffset_VertexAttrib4fvARB),
-    NAME_FUNC_OFFSET(16493, glVertexAttrib4ivARB, glVertexAttrib4ivARB, NULL, _gloffset_VertexAttrib4ivARB),
-    NAME_FUNC_OFFSET(16511, glVertexAttrib4sARB, glVertexAttrib4sARB, NULL, _gloffset_VertexAttrib4sARB),
-    NAME_FUNC_OFFSET(16528, glVertexAttrib4svARB, glVertexAttrib4svARB, NULL, _gloffset_VertexAttrib4svARB),
-    NAME_FUNC_OFFSET(16546, glVertexAttrib4ubvARB, glVertexAttrib4ubvARB, NULL, _gloffset_VertexAttrib4ubvARB),
-    NAME_FUNC_OFFSET(16565, glVertexAttrib4uivARB, glVertexAttrib4uivARB, NULL, _gloffset_VertexAttrib4uivARB),
-    NAME_FUNC_OFFSET(16584, glVertexAttrib4usvARB, glVertexAttrib4usvARB, NULL, _gloffset_VertexAttrib4usvARB),
-    NAME_FUNC_OFFSET(16603, glVertexAttribPointerARB, glVertexAttribPointerARB, NULL, _gloffset_VertexAttribPointerARB),
-    NAME_FUNC_OFFSET(16625, glBindBufferARB, glBindBufferARB, NULL, _gloffset_BindBufferARB),
-    NAME_FUNC_OFFSET(16638, glBufferDataARB, glBufferDataARB, NULL, _gloffset_BufferDataARB),
-    NAME_FUNC_OFFSET(16651, glBufferSubDataARB, glBufferSubDataARB, NULL, _gloffset_BufferSubDataARB),
-    NAME_FUNC_OFFSET(16667, glDeleteBuffersARB, glDeleteBuffersARB, NULL, _gloffset_DeleteBuffersARB),
-    NAME_FUNC_OFFSET(16683, glGenBuffersARB, glGenBuffersARB, NULL, _gloffset_GenBuffersARB),
-    NAME_FUNC_OFFSET(16696, glGetBufferParameterivARB, glGetBufferParameterivARB, NULL, _gloffset_GetBufferParameterivARB),
-    NAME_FUNC_OFFSET(16719, glGetBufferPointervARB, glGetBufferPointervARB, NULL, _gloffset_GetBufferPointervARB),
-    NAME_FUNC_OFFSET(16739, glGetBufferSubDataARB, glGetBufferSubDataARB, NULL, _gloffset_GetBufferSubDataARB),
-    NAME_FUNC_OFFSET(16758, glIsBufferARB, glIsBufferARB, NULL, _gloffset_IsBufferARB),
-    NAME_FUNC_OFFSET(16769, glMapBufferARB, glMapBufferARB, NULL, _gloffset_MapBufferARB),
-    NAME_FUNC_OFFSET(16781, glUnmapBufferARB, glUnmapBufferARB, NULL, _gloffset_UnmapBufferARB),
-    NAME_FUNC_OFFSET(16795, glBeginQueryARB, glBeginQueryARB, NULL, _gloffset_BeginQueryARB),
-    NAME_FUNC_OFFSET(16808, glDeleteQueriesARB, glDeleteQueriesARB, NULL, _gloffset_DeleteQueriesARB),
-    NAME_FUNC_OFFSET(16824, glEndQueryARB, glEndQueryARB, NULL, _gloffset_EndQueryARB),
-    NAME_FUNC_OFFSET(16835, glGenQueriesARB, glGenQueriesARB, NULL, _gloffset_GenQueriesARB),
-    NAME_FUNC_OFFSET(16848, glGetQueryObjectivARB, glGetQueryObjectivARB, NULL, _gloffset_GetQueryObjectivARB),
-    NAME_FUNC_OFFSET(16867, glGetQueryObjectuivARB, glGetQueryObjectuivARB, NULL, _gloffset_GetQueryObjectuivARB),
-    NAME_FUNC_OFFSET(16887, glGetQueryivARB, glGetQueryivARB, NULL, _gloffset_GetQueryivARB),
-    NAME_FUNC_OFFSET(16900, glIsQueryARB, glIsQueryARB, NULL, _gloffset_IsQueryARB),
-    NAME_FUNC_OFFSET(16910, glCompileShaderARB, glCompileShaderARB, NULL, _gloffset_CompileShaderARB),
-    NAME_FUNC_OFFSET(16926, glGetActiveUniformARB, glGetActiveUniformARB, NULL, _gloffset_GetActiveUniformARB),
-    NAME_FUNC_OFFSET(16945, glGetShaderSourceARB, glGetShaderSourceARB, NULL, _gloffset_GetShaderSourceARB),
-    NAME_FUNC_OFFSET(16963, glGetUniformLocationARB, glGetUniformLocationARB, NULL, _gloffset_GetUniformLocationARB),
-    NAME_FUNC_OFFSET(16984, glGetUniformfvARB, glGetUniformfvARB, NULL, _gloffset_GetUniformfvARB),
-    NAME_FUNC_OFFSET(16999, glGetUniformivARB, glGetUniformivARB, NULL, _gloffset_GetUniformivARB),
-    NAME_FUNC_OFFSET(17014, glLinkProgramARB, glLinkProgramARB, NULL, _gloffset_LinkProgramARB),
-    NAME_FUNC_OFFSET(17028, glShaderSourceARB, glShaderSourceARB, NULL, _gloffset_ShaderSourceARB),
-    NAME_FUNC_OFFSET(17043, glUniform1fARB, glUniform1fARB, NULL, _gloffset_Uniform1fARB),
-    NAME_FUNC_OFFSET(17055, glUniform1fvARB, glUniform1fvARB, NULL, _gloffset_Uniform1fvARB),
-    NAME_FUNC_OFFSET(17068, glUniform1iARB, glUniform1iARB, NULL, _gloffset_Uniform1iARB),
-    NAME_FUNC_OFFSET(17080, glUniform1ivARB, glUniform1ivARB, NULL, _gloffset_Uniform1ivARB),
-    NAME_FUNC_OFFSET(17093, glUniform2fARB, glUniform2fARB, NULL, _gloffset_Uniform2fARB),
-    NAME_FUNC_OFFSET(17105, glUniform2fvARB, glUniform2fvARB, NULL, _gloffset_Uniform2fvARB),
-    NAME_FUNC_OFFSET(17118, glUniform2iARB, glUniform2iARB, NULL, _gloffset_Uniform2iARB),
-    NAME_FUNC_OFFSET(17130, glUniform2ivARB, glUniform2ivARB, NULL, _gloffset_Uniform2ivARB),
-    NAME_FUNC_OFFSET(17143, glUniform3fARB, glUniform3fARB, NULL, _gloffset_Uniform3fARB),
-    NAME_FUNC_OFFSET(17155, glUniform3fvARB, glUniform3fvARB, NULL, _gloffset_Uniform3fvARB),
-    NAME_FUNC_OFFSET(17168, glUniform3iARB, glUniform3iARB, NULL, _gloffset_Uniform3iARB),
-    NAME_FUNC_OFFSET(17180, glUniform3ivARB, glUniform3ivARB, NULL, _gloffset_Uniform3ivARB),
-    NAME_FUNC_OFFSET(17193, glUniform4fARB, glUniform4fARB, NULL, _gloffset_Uniform4fARB),
-    NAME_FUNC_OFFSET(17205, glUniform4fvARB, glUniform4fvARB, NULL, _gloffset_Uniform4fvARB),
-    NAME_FUNC_OFFSET(17218, glUniform4iARB, glUniform4iARB, NULL, _gloffset_Uniform4iARB),
-    NAME_FUNC_OFFSET(17230, glUniform4ivARB, glUniform4ivARB, NULL, _gloffset_Uniform4ivARB),
-    NAME_FUNC_OFFSET(17243, glUniformMatrix2fvARB, glUniformMatrix2fvARB, NULL, _gloffset_UniformMatrix2fvARB),
-    NAME_FUNC_OFFSET(17262, glUniformMatrix3fvARB, glUniformMatrix3fvARB, NULL, _gloffset_UniformMatrix3fvARB),
-    NAME_FUNC_OFFSET(17281, glUniformMatrix4fvARB, glUniformMatrix4fvARB, NULL, _gloffset_UniformMatrix4fvARB),
-    NAME_FUNC_OFFSET(17300, glUseProgramObjectARB, glUseProgramObjectARB, NULL, _gloffset_UseProgramObjectARB),
-    NAME_FUNC_OFFSET(17313, glValidateProgramARB, glValidateProgramARB, NULL, _gloffset_ValidateProgramARB),
-    NAME_FUNC_OFFSET(17331, glBindAttribLocationARB, glBindAttribLocationARB, NULL, _gloffset_BindAttribLocationARB),
-    NAME_FUNC_OFFSET(17352, glGetActiveAttribARB, glGetActiveAttribARB, NULL, _gloffset_GetActiveAttribARB),
-    NAME_FUNC_OFFSET(17370, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB),
-    NAME_FUNC_OFFSET(17390, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
-    NAME_FUNC_OFFSET(17404, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
-    NAME_FUNC_OFFSET(17421, gl_dispatch_stub_568, gl_dispatch_stub_568, NULL, _gloffset_SampleMaskSGIS),
-    NAME_FUNC_OFFSET(17437, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_SamplePatternSGIS),
-    NAME_FUNC_OFFSET(17456, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET(17474, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET(17495, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
-    NAME_FUNC_OFFSET(17517, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET(17536, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET(17558, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
-    NAME_FUNC_OFFSET(17581, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
-    NAME_FUNC_OFFSET(17600, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
-    NAME_FUNC_OFFSET(17620, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
-    NAME_FUNC_OFFSET(17639, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
-    NAME_FUNC_OFFSET(17659, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
-    NAME_FUNC_OFFSET(17678, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
-    NAME_FUNC_OFFSET(17698, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
-    NAME_FUNC_OFFSET(17717, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
-    NAME_FUNC_OFFSET(17737, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
-    NAME_FUNC_OFFSET(17756, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
-    NAME_FUNC_OFFSET(17776, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
-    NAME_FUNC_OFFSET(17796, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
-    NAME_FUNC_OFFSET(17817, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
-    NAME_FUNC_OFFSET(17837, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
-    NAME_FUNC_OFFSET(17858, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
-    NAME_FUNC_OFFSET(17878, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
-    NAME_FUNC_OFFSET(17899, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
-    NAME_FUNC_OFFSET(17923, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
-    NAME_FUNC_OFFSET(17941, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
-    NAME_FUNC_OFFSET(17961, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
-    NAME_FUNC_OFFSET(17979, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
-    NAME_FUNC_OFFSET(17991, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
-    NAME_FUNC_OFFSET(18004, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
-    NAME_FUNC_OFFSET(18016, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
-    NAME_FUNC_OFFSET(18029, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
-    NAME_FUNC_OFFSET(18049, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
-    NAME_FUNC_OFFSET(18073, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
-    NAME_FUNC_OFFSET(18087, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
-    NAME_FUNC_OFFSET(18104, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
-    NAME_FUNC_OFFSET(18119, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
-    NAME_FUNC_OFFSET(18137, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
-    NAME_FUNC_OFFSET(18151, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
-    NAME_FUNC_OFFSET(18168, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
-    NAME_FUNC_OFFSET(18183, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
-    NAME_FUNC_OFFSET(18201, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
-    NAME_FUNC_OFFSET(18215, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
-    NAME_FUNC_OFFSET(18232, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
-    NAME_FUNC_OFFSET(18247, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
-    NAME_FUNC_OFFSET(18265, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
-    NAME_FUNC_OFFSET(18279, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
-    NAME_FUNC_OFFSET(18296, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
-    NAME_FUNC_OFFSET(18311, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
-    NAME_FUNC_OFFSET(18329, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
-    NAME_FUNC_OFFSET(18343, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
-    NAME_FUNC_OFFSET(18360, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
-    NAME_FUNC_OFFSET(18375, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
-    NAME_FUNC_OFFSET(18393, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
-    NAME_FUNC_OFFSET(18407, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
-    NAME_FUNC_OFFSET(18424, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
-    NAME_FUNC_OFFSET(18439, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
-    NAME_FUNC_OFFSET(18457, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
-    NAME_FUNC_OFFSET(18471, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
-    NAME_FUNC_OFFSET(18488, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
-    NAME_FUNC_OFFSET(18503, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
-    NAME_FUNC_OFFSET(18521, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
-    NAME_FUNC_OFFSET(18535, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
-    NAME_FUNC_OFFSET(18552, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
-    NAME_FUNC_OFFSET(18567, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
-    NAME_FUNC_OFFSET(18585, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
-    NAME_FUNC_OFFSET(18602, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
-    NAME_FUNC_OFFSET(18622, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
-    NAME_FUNC_OFFSET(18639, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
-    NAME_FUNC_OFFSET(18665, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
-    NAME_FUNC_OFFSET(18694, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
-    NAME_FUNC_OFFSET(18709, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
-    NAME_FUNC_OFFSET(18727, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
-    NAME_FUNC_OFFSET(18746, gl_dispatch_stub_745, gl_dispatch_stub_745, NULL, _gloffset_BlendEquationSeparateEXT),
-    NAME_FUNC_OFFSET(18770, gl_dispatch_stub_745, gl_dispatch_stub_745, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET( 8951, glRenderbufferStorageMultisample, glRenderbufferStorageMultisample, NULL, _gloffset_RenderbufferStorageMultisample),
+    NAME_FUNC_OFFSET( 8984, glPolygonOffsetEXT, glPolygonOffsetEXT, NULL, _gloffset_PolygonOffsetEXT),
+    NAME_FUNC_OFFSET( 9003, gl_dispatch_stub_563, gl_dispatch_stub_563, NULL, _gloffset_GetPixelTexGenParameterfvSGIS),
+    NAME_FUNC_OFFSET( 9035, gl_dispatch_stub_564, gl_dispatch_stub_564, NULL, _gloffset_GetPixelTexGenParameterivSGIS),
+    NAME_FUNC_OFFSET( 9067, gl_dispatch_stub_565, gl_dispatch_stub_565, NULL, _gloffset_PixelTexGenParameterfSGIS),
+    NAME_FUNC_OFFSET( 9095, gl_dispatch_stub_566, gl_dispatch_stub_566, NULL, _gloffset_PixelTexGenParameterfvSGIS),
+    NAME_FUNC_OFFSET( 9124, gl_dispatch_stub_567, gl_dispatch_stub_567, NULL, _gloffset_PixelTexGenParameteriSGIS),
+    NAME_FUNC_OFFSET( 9152, gl_dispatch_stub_568, gl_dispatch_stub_568, NULL, _gloffset_PixelTexGenParameterivSGIS),
+    NAME_FUNC_OFFSET( 9181, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_SampleMaskSGIS),
+    NAME_FUNC_OFFSET( 9198, gl_dispatch_stub_570, gl_dispatch_stub_570, NULL, _gloffset_SamplePatternSGIS),
+    NAME_FUNC_OFFSET( 9218, glColorPointerEXT, glColorPointerEXT, NULL, _gloffset_ColorPointerEXT),
+    NAME_FUNC_OFFSET( 9236, glEdgeFlagPointerEXT, glEdgeFlagPointerEXT, NULL, _gloffset_EdgeFlagPointerEXT),
+    NAME_FUNC_OFFSET( 9257, glIndexPointerEXT, glIndexPointerEXT, NULL, _gloffset_IndexPointerEXT),
+    NAME_FUNC_OFFSET( 9275, glNormalPointerEXT, glNormalPointerEXT, NULL, _gloffset_NormalPointerEXT),
+    NAME_FUNC_OFFSET( 9294, glTexCoordPointerEXT, glTexCoordPointerEXT, NULL, _gloffset_TexCoordPointerEXT),
+    NAME_FUNC_OFFSET( 9315, glVertexPointerEXT, glVertexPointerEXT, NULL, _gloffset_VertexPointerEXT),
+    NAME_FUNC_OFFSET( 9334, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET( 9355, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET( 9377, glLockArraysEXT, glLockArraysEXT, NULL, _gloffset_LockArraysEXT),
+    NAME_FUNC_OFFSET( 9393, glUnlockArraysEXT, glUnlockArraysEXT, NULL, _gloffset_UnlockArraysEXT),
+    NAME_FUNC_OFFSET( 9411, gl_dispatch_stub_581, gl_dispatch_stub_581, NULL, _gloffset_CullParameterdvEXT),
+    NAME_FUNC_OFFSET( 9432, gl_dispatch_stub_582, gl_dispatch_stub_582, NULL, _gloffset_CullParameterfvEXT),
+    NAME_FUNC_OFFSET( 9453, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
+    NAME_FUNC_OFFSET( 9475, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
+    NAME_FUNC_OFFSET( 9498, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
+    NAME_FUNC_OFFSET( 9520, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
+    NAME_FUNC_OFFSET( 9543, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
+    NAME_FUNC_OFFSET( 9565, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
+    NAME_FUNC_OFFSET( 9588, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
+    NAME_FUNC_OFFSET( 9610, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
+    NAME_FUNC_OFFSET( 9633, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
+    NAME_FUNC_OFFSET( 9655, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
+    NAME_FUNC_OFFSET( 9678, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
+    NAME_FUNC_OFFSET( 9701, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
+    NAME_FUNC_OFFSET( 9725, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
+    NAME_FUNC_OFFSET( 9748, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
+    NAME_FUNC_OFFSET( 9772, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
+    NAME_FUNC_OFFSET( 9795, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
+    NAME_FUNC_OFFSET( 9819, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
+    NAME_FUNC_OFFSET( 9846, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
+    NAME_FUNC_OFFSET( 9867, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
+    NAME_FUNC_OFFSET( 9890, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
+    NAME_FUNC_OFFSET( 9911, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
+    NAME_FUNC_OFFSET( 9926, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
+    NAME_FUNC_OFFSET( 9942, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
+    NAME_FUNC_OFFSET( 9957, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
+    NAME_FUNC_OFFSET( 9973, gl_dispatch_stub_607, gl_dispatch_stub_607, NULL, _gloffset_PixelTexGenSGIX),
+    NAME_FUNC_OFFSET( 9991, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
+    NAME_FUNC_OFFSET(10014, glFlushVertexArrayRangeNV, glFlushVertexArrayRangeNV, NULL, _gloffset_FlushVertexArrayRangeNV),
+    NAME_FUNC_OFFSET(10040, glVertexArrayRangeNV, glVertexArrayRangeNV, NULL, _gloffset_VertexArrayRangeNV),
+    NAME_FUNC_OFFSET(10061, glCombinerInputNV, glCombinerInputNV, NULL, _gloffset_CombinerInputNV),
+    NAME_FUNC_OFFSET(10079, glCombinerOutputNV, glCombinerOutputNV, NULL, _gloffset_CombinerOutputNV),
+    NAME_FUNC_OFFSET(10098, glCombinerParameterfNV, glCombinerParameterfNV, NULL, _gloffset_CombinerParameterfNV),
+    NAME_FUNC_OFFSET(10121, glCombinerParameterfvNV, glCombinerParameterfvNV, NULL, _gloffset_CombinerParameterfvNV),
+    NAME_FUNC_OFFSET(10145, glCombinerParameteriNV, glCombinerParameteriNV, NULL, _gloffset_CombinerParameteriNV),
+    NAME_FUNC_OFFSET(10168, glCombinerParameterivNV, glCombinerParameterivNV, NULL, _gloffset_CombinerParameterivNV),
+    NAME_FUNC_OFFSET(10192, glFinalCombinerInputNV, glFinalCombinerInputNV, NULL, _gloffset_FinalCombinerInputNV),
+    NAME_FUNC_OFFSET(10215, glGetCombinerInputParameterfvNV, glGetCombinerInputParameterfvNV, NULL, _gloffset_GetCombinerInputParameterfvNV),
+    NAME_FUNC_OFFSET(10247, glGetCombinerInputParameterivNV, glGetCombinerInputParameterivNV, NULL, _gloffset_GetCombinerInputParameterivNV),
+    NAME_FUNC_OFFSET(10279, glGetCombinerOutputParameterfvNV, glGetCombinerOutputParameterfvNV, NULL, _gloffset_GetCombinerOutputParameterfvNV),
+    NAME_FUNC_OFFSET(10312, glGetCombinerOutputParameterivNV, glGetCombinerOutputParameterivNV, NULL, _gloffset_GetCombinerOutputParameterivNV),
+    NAME_FUNC_OFFSET(10345, glGetFinalCombinerInputParameterfvNV, glGetFinalCombinerInputParameterfvNV, NULL, _gloffset_GetFinalCombinerInputParameterfvNV),
+    NAME_FUNC_OFFSET(10382, glGetFinalCombinerInputParameterivNV, glGetFinalCombinerInputParameterivNV, NULL, _gloffset_GetFinalCombinerInputParameterivNV),
+    NAME_FUNC_OFFSET(10419, glResizeBuffersMESA, glResizeBuffersMESA, NULL, _gloffset_ResizeBuffersMESA),
+    NAME_FUNC_OFFSET(10439, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
+    NAME_FUNC_OFFSET(10457, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
+    NAME_FUNC_OFFSET(10476, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
+    NAME_FUNC_OFFSET(10494, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
+    NAME_FUNC_OFFSET(10513, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
+    NAME_FUNC_OFFSET(10531, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
+    NAME_FUNC_OFFSET(10550, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
+    NAME_FUNC_OFFSET(10568, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
+    NAME_FUNC_OFFSET(10587, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
+    NAME_FUNC_OFFSET(10605, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
+    NAME_FUNC_OFFSET(10624, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
+    NAME_FUNC_OFFSET(10642, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
+    NAME_FUNC_OFFSET(10661, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
+    NAME_FUNC_OFFSET(10679, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
+    NAME_FUNC_OFFSET(10698, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
+    NAME_FUNC_OFFSET(10716, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
+    NAME_FUNC_OFFSET(10735, glWindowPos4dMESA, glWindowPos4dMESA, NULL, _gloffset_WindowPos4dMESA),
+    NAME_FUNC_OFFSET(10753, glWindowPos4dvMESA, glWindowPos4dvMESA, NULL, _gloffset_WindowPos4dvMESA),
+    NAME_FUNC_OFFSET(10772, glWindowPos4fMESA, glWindowPos4fMESA, NULL, _gloffset_WindowPos4fMESA),
+    NAME_FUNC_OFFSET(10790, glWindowPos4fvMESA, glWindowPos4fvMESA, NULL, _gloffset_WindowPos4fvMESA),
+    NAME_FUNC_OFFSET(10809, glWindowPos4iMESA, glWindowPos4iMESA, NULL, _gloffset_WindowPos4iMESA),
+    NAME_FUNC_OFFSET(10827, glWindowPos4ivMESA, glWindowPos4ivMESA, NULL, _gloffset_WindowPos4ivMESA),
+    NAME_FUNC_OFFSET(10846, glWindowPos4sMESA, glWindowPos4sMESA, NULL, _gloffset_WindowPos4sMESA),
+    NAME_FUNC_OFFSET(10864, glWindowPos4svMESA, glWindowPos4svMESA, NULL, _gloffset_WindowPos4svMESA),
+    NAME_FUNC_OFFSET(10883, gl_dispatch_stub_649, gl_dispatch_stub_649, NULL, _gloffset_MultiModeDrawArraysIBM),
+    NAME_FUNC_OFFSET(10908, gl_dispatch_stub_650, gl_dispatch_stub_650, NULL, _gloffset_MultiModeDrawElementsIBM),
+    NAME_FUNC_OFFSET(10935, gl_dispatch_stub_651, gl_dispatch_stub_651, NULL, _gloffset_DeleteFencesNV),
+    NAME_FUNC_OFFSET(10952, gl_dispatch_stub_652, gl_dispatch_stub_652, NULL, _gloffset_FinishFenceNV),
+    NAME_FUNC_OFFSET(10968, gl_dispatch_stub_653, gl_dispatch_stub_653, NULL, _gloffset_GenFencesNV),
+    NAME_FUNC_OFFSET(10982, gl_dispatch_stub_654, gl_dispatch_stub_654, NULL, _gloffset_GetFenceivNV),
+    NAME_FUNC_OFFSET(10997, gl_dispatch_stub_655, gl_dispatch_stub_655, NULL, _gloffset_IsFenceNV),
+    NAME_FUNC_OFFSET(11009, gl_dispatch_stub_656, gl_dispatch_stub_656, NULL, _gloffset_SetFenceNV),
+    NAME_FUNC_OFFSET(11022, gl_dispatch_stub_657, gl_dispatch_stub_657, NULL, _gloffset_TestFenceNV),
+    NAME_FUNC_OFFSET(11036, glAreProgramsResidentNV, glAreProgramsResidentNV, NULL, _gloffset_AreProgramsResidentNV),
+    NAME_FUNC_OFFSET(11060, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
+    NAME_FUNC_OFFSET(11076, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
+    NAME_FUNC_OFFSET(11095, glExecuteProgramNV, glExecuteProgramNV, NULL, _gloffset_ExecuteProgramNV),
+    NAME_FUNC_OFFSET(11114, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
+    NAME_FUNC_OFFSET(11130, glGetProgramParameterdvNV, glGetProgramParameterdvNV, NULL, _gloffset_GetProgramParameterdvNV),
+    NAME_FUNC_OFFSET(11156, glGetProgramParameterfvNV, glGetProgramParameterfvNV, NULL, _gloffset_GetProgramParameterfvNV),
+    NAME_FUNC_OFFSET(11182, glGetProgramStringNV, glGetProgramStringNV, NULL, _gloffset_GetProgramStringNV),
+    NAME_FUNC_OFFSET(11203, glGetProgramivNV, glGetProgramivNV, NULL, _gloffset_GetProgramivNV),
+    NAME_FUNC_OFFSET(11220, glGetTrackMatrixivNV, glGetTrackMatrixivNV, NULL, _gloffset_GetTrackMatrixivNV),
+    NAME_FUNC_OFFSET(11241, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
+    NAME_FUNC_OFFSET(11269, glGetVertexAttribdvNV, glGetVertexAttribdvNV, NULL, _gloffset_GetVertexAttribdvNV),
+    NAME_FUNC_OFFSET(11291, glGetVertexAttribfvNV, glGetVertexAttribfvNV, NULL, _gloffset_GetVertexAttribfvNV),
+    NAME_FUNC_OFFSET(11313, glGetVertexAttribivNV, glGetVertexAttribivNV, NULL, _gloffset_GetVertexAttribivNV),
+    NAME_FUNC_OFFSET(11335, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
+    NAME_FUNC_OFFSET(11349, glLoadProgramNV, glLoadProgramNV, NULL, _gloffset_LoadProgramNV),
+    NAME_FUNC_OFFSET(11365, glProgramParameters4dvNV, glProgramParameters4dvNV, NULL, _gloffset_ProgramParameters4dvNV),
+    NAME_FUNC_OFFSET(11390, glProgramParameters4fvNV, glProgramParameters4fvNV, NULL, _gloffset_ProgramParameters4fvNV),
+    NAME_FUNC_OFFSET(11415, glRequestResidentProgramsNV, glRequestResidentProgramsNV, NULL, _gloffset_RequestResidentProgramsNV),
+    NAME_FUNC_OFFSET(11443, glTrackMatrixNV, glTrackMatrixNV, NULL, _gloffset_TrackMatrixNV),
+    NAME_FUNC_OFFSET(11459, glVertexAttrib1dNV, glVertexAttrib1dNV, NULL, _gloffset_VertexAttrib1dNV),
+    NAME_FUNC_OFFSET(11478, glVertexAttrib1dvNV, glVertexAttrib1dvNV, NULL, _gloffset_VertexAttrib1dvNV),
+    NAME_FUNC_OFFSET(11498, glVertexAttrib1fNV, glVertexAttrib1fNV, NULL, _gloffset_VertexAttrib1fNV),
+    NAME_FUNC_OFFSET(11517, glVertexAttrib1fvNV, glVertexAttrib1fvNV, NULL, _gloffset_VertexAttrib1fvNV),
+    NAME_FUNC_OFFSET(11537, glVertexAttrib1sNV, glVertexAttrib1sNV, NULL, _gloffset_VertexAttrib1sNV),
+    NAME_FUNC_OFFSET(11556, glVertexAttrib1svNV, glVertexAttrib1svNV, NULL, _gloffset_VertexAttrib1svNV),
+    NAME_FUNC_OFFSET(11576, glVertexAttrib2dNV, glVertexAttrib2dNV, NULL, _gloffset_VertexAttrib2dNV),
+    NAME_FUNC_OFFSET(11595, glVertexAttrib2dvNV, glVertexAttrib2dvNV, NULL, _gloffset_VertexAttrib2dvNV),
+    NAME_FUNC_OFFSET(11615, glVertexAttrib2fNV, glVertexAttrib2fNV, NULL, _gloffset_VertexAttrib2fNV),
+    NAME_FUNC_OFFSET(11634, glVertexAttrib2fvNV, glVertexAttrib2fvNV, NULL, _gloffset_VertexAttrib2fvNV),
+    NAME_FUNC_OFFSET(11654, glVertexAttrib2sNV, glVertexAttrib2sNV, NULL, _gloffset_VertexAttrib2sNV),
+    NAME_FUNC_OFFSET(11673, glVertexAttrib2svNV, glVertexAttrib2svNV, NULL, _gloffset_VertexAttrib2svNV),
+    NAME_FUNC_OFFSET(11693, glVertexAttrib3dNV, glVertexAttrib3dNV, NULL, _gloffset_VertexAttrib3dNV),
+    NAME_FUNC_OFFSET(11712, glVertexAttrib3dvNV, glVertexAttrib3dvNV, NULL, _gloffset_VertexAttrib3dvNV),
+    NAME_FUNC_OFFSET(11732, glVertexAttrib3fNV, glVertexAttrib3fNV, NULL, _gloffset_VertexAttrib3fNV),
+    NAME_FUNC_OFFSET(11751, glVertexAttrib3fvNV, glVertexAttrib3fvNV, NULL, _gloffset_VertexAttrib3fvNV),
+    NAME_FUNC_OFFSET(11771, glVertexAttrib3sNV, glVertexAttrib3sNV, NULL, _gloffset_VertexAttrib3sNV),
+    NAME_FUNC_OFFSET(11790, glVertexAttrib3svNV, glVertexAttrib3svNV, NULL, _gloffset_VertexAttrib3svNV),
+    NAME_FUNC_OFFSET(11810, glVertexAttrib4dNV, glVertexAttrib4dNV, NULL, _gloffset_VertexAttrib4dNV),
+    NAME_FUNC_OFFSET(11829, glVertexAttrib4dvNV, glVertexAttrib4dvNV, NULL, _gloffset_VertexAttrib4dvNV),
+    NAME_FUNC_OFFSET(11849, glVertexAttrib4fNV, glVertexAttrib4fNV, NULL, _gloffset_VertexAttrib4fNV),
+    NAME_FUNC_OFFSET(11868, glVertexAttrib4fvNV, glVertexAttrib4fvNV, NULL, _gloffset_VertexAttrib4fvNV),
+    NAME_FUNC_OFFSET(11888, glVertexAttrib4sNV, glVertexAttrib4sNV, NULL, _gloffset_VertexAttrib4sNV),
+    NAME_FUNC_OFFSET(11907, glVertexAttrib4svNV, glVertexAttrib4svNV, NULL, _gloffset_VertexAttrib4svNV),
+    NAME_FUNC_OFFSET(11927, glVertexAttrib4ubNV, glVertexAttrib4ubNV, NULL, _gloffset_VertexAttrib4ubNV),
+    NAME_FUNC_OFFSET(11947, glVertexAttrib4ubvNV, glVertexAttrib4ubvNV, NULL, _gloffset_VertexAttrib4ubvNV),
+    NAME_FUNC_OFFSET(11968, glVertexAttribPointerNV, glVertexAttribPointerNV, NULL, _gloffset_VertexAttribPointerNV),
+    NAME_FUNC_OFFSET(11992, glVertexAttribs1dvNV, glVertexAttribs1dvNV, NULL, _gloffset_VertexAttribs1dvNV),
+    NAME_FUNC_OFFSET(12013, glVertexAttribs1fvNV, glVertexAttribs1fvNV, NULL, _gloffset_VertexAttribs1fvNV),
+    NAME_FUNC_OFFSET(12034, glVertexAttribs1svNV, glVertexAttribs1svNV, NULL, _gloffset_VertexAttribs1svNV),
+    NAME_FUNC_OFFSET(12055, glVertexAttribs2dvNV, glVertexAttribs2dvNV, NULL, _gloffset_VertexAttribs2dvNV),
+    NAME_FUNC_OFFSET(12076, glVertexAttribs2fvNV, glVertexAttribs2fvNV, NULL, _gloffset_VertexAttribs2fvNV),
+    NAME_FUNC_OFFSET(12097, glVertexAttribs2svNV, glVertexAttribs2svNV, NULL, _gloffset_VertexAttribs2svNV),
+    NAME_FUNC_OFFSET(12118, glVertexAttribs3dvNV, glVertexAttribs3dvNV, NULL, _gloffset_VertexAttribs3dvNV),
+    NAME_FUNC_OFFSET(12139, glVertexAttribs3fvNV, glVertexAttribs3fvNV, NULL, _gloffset_VertexAttribs3fvNV),
+    NAME_FUNC_OFFSET(12160, glVertexAttribs3svNV, glVertexAttribs3svNV, NULL, _gloffset_VertexAttribs3svNV),
+    NAME_FUNC_OFFSET(12181, glVertexAttribs4dvNV, glVertexAttribs4dvNV, NULL, _gloffset_VertexAttribs4dvNV),
+    NAME_FUNC_OFFSET(12202, glVertexAttribs4fvNV, glVertexAttribs4fvNV, NULL, _gloffset_VertexAttribs4fvNV),
+    NAME_FUNC_OFFSET(12223, glVertexAttribs4svNV, glVertexAttribs4svNV, NULL, _gloffset_VertexAttribs4svNV),
+    NAME_FUNC_OFFSET(12244, glVertexAttribs4ubvNV, glVertexAttribs4ubvNV, NULL, _gloffset_VertexAttribs4ubvNV),
+    NAME_FUNC_OFFSET(12266, glAlphaFragmentOp1ATI, glAlphaFragmentOp1ATI, NULL, _gloffset_AlphaFragmentOp1ATI),
+    NAME_FUNC_OFFSET(12288, glAlphaFragmentOp2ATI, glAlphaFragmentOp2ATI, NULL, _gloffset_AlphaFragmentOp2ATI),
+    NAME_FUNC_OFFSET(12310, glAlphaFragmentOp3ATI, glAlphaFragmentOp3ATI, NULL, _gloffset_AlphaFragmentOp3ATI),
+    NAME_FUNC_OFFSET(12332, glBeginFragmentShaderATI, glBeginFragmentShaderATI, NULL, _gloffset_BeginFragmentShaderATI),
+    NAME_FUNC_OFFSET(12357, glBindFragmentShaderATI, glBindFragmentShaderATI, NULL, _gloffset_BindFragmentShaderATI),
+    NAME_FUNC_OFFSET(12381, glColorFragmentOp1ATI, glColorFragmentOp1ATI, NULL, _gloffset_ColorFragmentOp1ATI),
+    NAME_FUNC_OFFSET(12403, glColorFragmentOp2ATI, glColorFragmentOp2ATI, NULL, _gloffset_ColorFragmentOp2ATI),
+    NAME_FUNC_OFFSET(12425, glColorFragmentOp3ATI, glColorFragmentOp3ATI, NULL, _gloffset_ColorFragmentOp3ATI),
+    NAME_FUNC_OFFSET(12447, glDeleteFragmentShaderATI, glDeleteFragmentShaderATI, NULL, _gloffset_DeleteFragmentShaderATI),
+    NAME_FUNC_OFFSET(12473, glEndFragmentShaderATI, glEndFragmentShaderATI, NULL, _gloffset_EndFragmentShaderATI),
+    NAME_FUNC_OFFSET(12496, glGenFragmentShadersATI, glGenFragmentShadersATI, NULL, _gloffset_GenFragmentShadersATI),
+    NAME_FUNC_OFFSET(12520, glPassTexCoordATI, glPassTexCoordATI, NULL, _gloffset_PassTexCoordATI),
+    NAME_FUNC_OFFSET(12538, glSampleMapATI, glSampleMapATI, NULL, _gloffset_SampleMapATI),
+    NAME_FUNC_OFFSET(12553, glSetFragmentShaderConstantATI, glSetFragmentShaderConstantATI, NULL, _gloffset_SetFragmentShaderConstantATI),
+    NAME_FUNC_OFFSET(12584, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
+    NAME_FUNC_OFFSET(12604, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
+    NAME_FUNC_OFFSET(12625, gl_dispatch_stub_734, gl_dispatch_stub_734, NULL, _gloffset_ActiveStencilFaceEXT),
+    NAME_FUNC_OFFSET(12648, gl_dispatch_stub_735, gl_dispatch_stub_735, NULL, _gloffset_BindVertexArrayAPPLE),
+    NAME_FUNC_OFFSET(12671, gl_dispatch_stub_736, gl_dispatch_stub_736, NULL, _gloffset_DeleteVertexArraysAPPLE),
+    NAME_FUNC_OFFSET(12697, gl_dispatch_stub_737, gl_dispatch_stub_737, NULL, _gloffset_GenVertexArraysAPPLE),
+    NAME_FUNC_OFFSET(12720, gl_dispatch_stub_738, gl_dispatch_stub_738, NULL, _gloffset_IsVertexArrayAPPLE),
+    NAME_FUNC_OFFSET(12741, glGetProgramNamedParameterdvNV, glGetProgramNamedParameterdvNV, NULL, _gloffset_GetProgramNamedParameterdvNV),
+    NAME_FUNC_OFFSET(12772, glGetProgramNamedParameterfvNV, glGetProgramNamedParameterfvNV, NULL, _gloffset_GetProgramNamedParameterfvNV),
+    NAME_FUNC_OFFSET(12803, glProgramNamedParameter4dNV, glProgramNamedParameter4dNV, NULL, _gloffset_ProgramNamedParameter4dNV),
+    NAME_FUNC_OFFSET(12831, glProgramNamedParameter4dvNV, glProgramNamedParameter4dvNV, NULL, _gloffset_ProgramNamedParameter4dvNV),
+    NAME_FUNC_OFFSET(12860, glProgramNamedParameter4fNV, glProgramNamedParameter4fNV, NULL, _gloffset_ProgramNamedParameter4fNV),
+    NAME_FUNC_OFFSET(12888, glProgramNamedParameter4fvNV, glProgramNamedParameter4fvNV, NULL, _gloffset_ProgramNamedParameter4fvNV),
+    NAME_FUNC_OFFSET(12917, gl_dispatch_stub_745, gl_dispatch_stub_745, NULL, _gloffset_DepthBoundsEXT),
+    NAME_FUNC_OFFSET(12934, gl_dispatch_stub_746, gl_dispatch_stub_746, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(12961, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
+    NAME_FUNC_OFFSET(12982, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
+    NAME_FUNC_OFFSET(13004, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
+    NAME_FUNC_OFFSET(13032, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
+    NAME_FUNC_OFFSET(13056, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
+    NAME_FUNC_OFFSET(13081, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
+    NAME_FUNC_OFFSET(13110, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
+    NAME_FUNC_OFFSET(13136, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
+    NAME_FUNC_OFFSET(13162, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
+    NAME_FUNC_OFFSET(13188, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
+    NAME_FUNC_OFFSET(13209, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
+    NAME_FUNC_OFFSET(13231, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
+    NAME_FUNC_OFFSET(13251, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
+    NAME_FUNC_OFFSET(13292, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
+    NAME_FUNC_OFFSET(13324, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
+    NAME_FUNC_OFFSET(13343, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
+    NAME_FUNC_OFFSET(13363, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
+    NAME_FUNC_OFFSET(13388, gl_dispatch_stub_764, gl_dispatch_stub_764, NULL, _gloffset_BlitFramebufferEXT),
+    NAME_FUNC_OFFSET(13409, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
+    NAME_FUNC_OFFSET(13438, gl_dispatch_stub_766, gl_dispatch_stub_766, NULL, _gloffset_StencilFuncSeparateATI),
+    NAME_FUNC_OFFSET(13463, gl_dispatch_stub_767, gl_dispatch_stub_767, NULL, _gloffset_ProgramEnvParameters4fvEXT),
+    NAME_FUNC_OFFSET(13492, gl_dispatch_stub_768, gl_dispatch_stub_768, NULL, _gloffset_ProgramLocalParameters4fvEXT),
+    NAME_FUNC_OFFSET(13523, gl_dispatch_stub_769, gl_dispatch_stub_769, NULL, _gloffset_GetQueryObjecti64vEXT),
+    NAME_FUNC_OFFSET(13547, gl_dispatch_stub_770, gl_dispatch_stub_770, NULL, _gloffset_GetQueryObjectui64vEXT),
+    NAME_FUNC_OFFSET(13572, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement),
+    NAME_FUNC_OFFSET(13590, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture),
+    NAME_FUNC_OFFSET(13607, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays),
+    NAME_FUNC_OFFSET(13623, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident),
+    NAME_FUNC_OFFSET(13648, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D),
+    NAME_FUNC_OFFSET(13668, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D),
+    NAME_FUNC_OFFSET(13688, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D),
+    NAME_FUNC_OFFSET(13711, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D),
+    NAME_FUNC_OFFSET(13734, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures),
+    NAME_FUNC_OFFSET(13754, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures),
+    NAME_FUNC_OFFSET(13771, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv),
+    NAME_FUNC_OFFSET(13788, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture),
+    NAME_FUNC_OFFSET(13803, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures),
+    NAME_FUNC_OFFSET(13827, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D),
+    NAME_FUNC_OFFSET(13846, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D),
+    NAME_FUNC_OFFSET(13865, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor),
+    NAME_FUNC_OFFSET(13881, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation),
+    NAME_FUNC_OFFSET(13900, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements),
+    NAME_FUNC_OFFSET(13923, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
+    NAME_FUNC_OFFSET(13939, glColorTable, glColorTable, NULL, _gloffset_ColorTable),
+    NAME_FUNC_OFFSET(13955, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv),
+    NAME_FUNC_OFFSET(13982, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv),
+    NAME_FUNC_OFFSET(14009, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable),
+    NAME_FUNC_OFFSET(14029, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
+    NAME_FUNC_OFFSET(14048, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable),
+    NAME_FUNC_OFFSET(14067, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
+    NAME_FUNC_OFFSET(14097, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv),
+    NAME_FUNC_OFFSET(14127, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
+    NAME_FUNC_OFFSET(14157, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv),
+    NAME_FUNC_OFFSET(14187, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable),
+    NAME_FUNC_OFFSET(14206, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable),
+    NAME_FUNC_OFFSET(14229, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D),
+    NAME_FUNC_OFFSET(14254, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D),
+    NAME_FUNC_OFFSET(14279, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf),
+    NAME_FUNC_OFFSET(14306, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv),
+    NAME_FUNC_OFFSET(14334, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri),
+    NAME_FUNC_OFFSET(14361, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv),
+    NAME_FUNC_OFFSET(14389, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D),
+    NAME_FUNC_OFFSET(14418, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D),
+    NAME_FUNC_OFFSET(14447, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter),
+    NAME_FUNC_OFFSET(14473, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv),
+    NAME_FUNC_OFFSET(14504, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv),
+    NAME_FUNC_OFFSET(14535, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter),
+    NAME_FUNC_OFFSET(14559, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D),
+    NAME_FUNC_OFFSET(14582, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram),
+    NAME_FUNC_OFFSET(14600, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv),
+    NAME_FUNC_OFFSET(14629, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv),
+    NAME_FUNC_OFFSET(14658, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax),
+    NAME_FUNC_OFFSET(14673, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv),
+    NAME_FUNC_OFFSET(14699, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv),
+    NAME_FUNC_OFFSET(14725, glHistogram, glHistogram, NULL, _gloffset_Histogram),
+    NAME_FUNC_OFFSET(14740, glMinmax, glMinmax, NULL, _gloffset_Minmax),
+    NAME_FUNC_OFFSET(14752, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram),
+    NAME_FUNC_OFFSET(14772, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax),
+    NAME_FUNC_OFFSET(14789, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D),
+    NAME_FUNC_OFFSET(14805, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D),
+    NAME_FUNC_OFFSET(14824, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D),
+    NAME_FUNC_OFFSET(14847, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB),
+    NAME_FUNC_OFFSET(14863, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB),
+    NAME_FUNC_OFFSET(14885, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB),
+    NAME_FUNC_OFFSET(14903, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB),
+    NAME_FUNC_OFFSET(14922, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB),
+    NAME_FUNC_OFFSET(14940, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB),
+    NAME_FUNC_OFFSET(14959, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB),
+    NAME_FUNC_OFFSET(14977, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB),
+    NAME_FUNC_OFFSET(14996, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB),
+    NAME_FUNC_OFFSET(15014, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB),
+    NAME_FUNC_OFFSET(15033, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB),
+    NAME_FUNC_OFFSET(15051, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB),
+    NAME_FUNC_OFFSET(15070, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB),
+    NAME_FUNC_OFFSET(15088, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB),
+    NAME_FUNC_OFFSET(15107, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB),
+    NAME_FUNC_OFFSET(15125, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB),
+    NAME_FUNC_OFFSET(15144, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB),
+    NAME_FUNC_OFFSET(15162, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB),
+    NAME_FUNC_OFFSET(15181, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB),
+    NAME_FUNC_OFFSET(15199, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB),
+    NAME_FUNC_OFFSET(15218, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB),
+    NAME_FUNC_OFFSET(15236, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB),
+    NAME_FUNC_OFFSET(15255, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB),
+    NAME_FUNC_OFFSET(15273, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB),
+    NAME_FUNC_OFFSET(15292, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB),
+    NAME_FUNC_OFFSET(15310, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB),
+    NAME_FUNC_OFFSET(15329, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB),
+    NAME_FUNC_OFFSET(15347, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB),
+    NAME_FUNC_OFFSET(15366, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB),
+    NAME_FUNC_OFFSET(15384, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB),
+    NAME_FUNC_OFFSET(15403, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB),
+    NAME_FUNC_OFFSET(15421, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB),
+    NAME_FUNC_OFFSET(15440, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB),
+    NAME_FUNC_OFFSET(15458, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB),
+    NAME_FUNC_OFFSET(15477, glStencilOpSeparate, glStencilOpSeparate, NULL, _gloffset_StencilOpSeparate),
+    NAME_FUNC_OFFSET(15500, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB),
+    NAME_FUNC_OFFSET(15523, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB),
+    NAME_FUNC_OFFSET(15546, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB),
+    NAME_FUNC_OFFSET(15569, glMultTransposeMatrixfARB, glMultTransposeMatrixfARB, NULL, _gloffset_MultTransposeMatrixfARB),
+    NAME_FUNC_OFFSET(15592, glSampleCoverageARB, glSampleCoverageARB, NULL, _gloffset_SampleCoverageARB),
+    NAME_FUNC_OFFSET(15609, glCompressedTexImage1DARB, glCompressedTexImage1DARB, NULL, _gloffset_CompressedTexImage1DARB),
+    NAME_FUNC_OFFSET(15632, glCompressedTexImage2DARB, glCompressedTexImage2DARB, NULL, _gloffset_CompressedTexImage2DARB),
+    NAME_FUNC_OFFSET(15655, glCompressedTexImage3DARB, glCompressedTexImage3DARB, NULL, _gloffset_CompressedTexImage3DARB),
+    NAME_FUNC_OFFSET(15678, glCompressedTexSubImage1DARB, glCompressedTexSubImage1DARB, NULL, _gloffset_CompressedTexSubImage1DARB),
+    NAME_FUNC_OFFSET(15704, glCompressedTexSubImage2DARB, glCompressedTexSubImage2DARB, NULL, _gloffset_CompressedTexSubImage2DARB),
+    NAME_FUNC_OFFSET(15730, glCompressedTexSubImage3DARB, glCompressedTexSubImage3DARB, NULL, _gloffset_CompressedTexSubImage3DARB),
+    NAME_FUNC_OFFSET(15756, glGetCompressedTexImageARB, glGetCompressedTexImageARB, NULL, _gloffset_GetCompressedTexImageARB),
+    NAME_FUNC_OFFSET(15780, glDisableVertexAttribArrayARB, glDisableVertexAttribArrayARB, NULL, _gloffset_DisableVertexAttribArrayARB),
+    NAME_FUNC_OFFSET(15807, glEnableVertexAttribArrayARB, glEnableVertexAttribArrayARB, NULL, _gloffset_EnableVertexAttribArrayARB),
+    NAME_FUNC_OFFSET(15833, glGetVertexAttribdvARB, glGetVertexAttribdvARB, NULL, _gloffset_GetVertexAttribdvARB),
+    NAME_FUNC_OFFSET(15853, glGetVertexAttribfvARB, glGetVertexAttribfvARB, NULL, _gloffset_GetVertexAttribfvARB),
+    NAME_FUNC_OFFSET(15873, glGetVertexAttribivARB, glGetVertexAttribivARB, NULL, _gloffset_GetVertexAttribivARB),
+    NAME_FUNC_OFFSET(15893, glProgramEnvParameter4dARB, glProgramEnvParameter4dARB, NULL, _gloffset_ProgramEnvParameter4dARB),
+    NAME_FUNC_OFFSET(15916, glProgramEnvParameter4dvARB, glProgramEnvParameter4dvARB, NULL, _gloffset_ProgramEnvParameter4dvARB),
+    NAME_FUNC_OFFSET(15940, glProgramEnvParameter4fARB, glProgramEnvParameter4fARB, NULL, _gloffset_ProgramEnvParameter4fARB),
+    NAME_FUNC_OFFSET(15963, glProgramEnvParameter4fvARB, glProgramEnvParameter4fvARB, NULL, _gloffset_ProgramEnvParameter4fvARB),
+    NAME_FUNC_OFFSET(15987, glVertexAttrib1dARB, glVertexAttrib1dARB, NULL, _gloffset_VertexAttrib1dARB),
+    NAME_FUNC_OFFSET(16004, glVertexAttrib1dvARB, glVertexAttrib1dvARB, NULL, _gloffset_VertexAttrib1dvARB),
+    NAME_FUNC_OFFSET(16022, glVertexAttrib1fARB, glVertexAttrib1fARB, NULL, _gloffset_VertexAttrib1fARB),
+    NAME_FUNC_OFFSET(16039, glVertexAttrib1fvARB, glVertexAttrib1fvARB, NULL, _gloffset_VertexAttrib1fvARB),
+    NAME_FUNC_OFFSET(16057, glVertexAttrib1sARB, glVertexAttrib1sARB, NULL, _gloffset_VertexAttrib1sARB),
+    NAME_FUNC_OFFSET(16074, glVertexAttrib1svARB, glVertexAttrib1svARB, NULL, _gloffset_VertexAttrib1svARB),
+    NAME_FUNC_OFFSET(16092, glVertexAttrib2dARB, glVertexAttrib2dARB, NULL, _gloffset_VertexAttrib2dARB),
+    NAME_FUNC_OFFSET(16109, glVertexAttrib2dvARB, glVertexAttrib2dvARB, NULL, _gloffset_VertexAttrib2dvARB),
+    NAME_FUNC_OFFSET(16127, glVertexAttrib2fARB, glVertexAttrib2fARB, NULL, _gloffset_VertexAttrib2fARB),
+    NAME_FUNC_OFFSET(16144, glVertexAttrib2fvARB, glVertexAttrib2fvARB, NULL, _gloffset_VertexAttrib2fvARB),
+    NAME_FUNC_OFFSET(16162, glVertexAttrib2sARB, glVertexAttrib2sARB, NULL, _gloffset_VertexAttrib2sARB),
+    NAME_FUNC_OFFSET(16179, glVertexAttrib2svARB, glVertexAttrib2svARB, NULL, _gloffset_VertexAttrib2svARB),
+    NAME_FUNC_OFFSET(16197, glVertexAttrib3dARB, glVertexAttrib3dARB, NULL, _gloffset_VertexAttrib3dARB),
+    NAME_FUNC_OFFSET(16214, glVertexAttrib3dvARB, glVertexAttrib3dvARB, NULL, _gloffset_VertexAttrib3dvARB),
+    NAME_FUNC_OFFSET(16232, glVertexAttrib3fARB, glVertexAttrib3fARB, NULL, _gloffset_VertexAttrib3fARB),
+    NAME_FUNC_OFFSET(16249, glVertexAttrib3fvARB, glVertexAttrib3fvARB, NULL, _gloffset_VertexAttrib3fvARB),
+    NAME_FUNC_OFFSET(16267, glVertexAttrib3sARB, glVertexAttrib3sARB, NULL, _gloffset_VertexAttrib3sARB),
+    NAME_FUNC_OFFSET(16284, glVertexAttrib3svARB, glVertexAttrib3svARB, NULL, _gloffset_VertexAttrib3svARB),
+    NAME_FUNC_OFFSET(16302, glVertexAttrib4NbvARB, glVertexAttrib4NbvARB, NULL, _gloffset_VertexAttrib4NbvARB),
+    NAME_FUNC_OFFSET(16321, glVertexAttrib4NivARB, glVertexAttrib4NivARB, NULL, _gloffset_VertexAttrib4NivARB),
+    NAME_FUNC_OFFSET(16340, glVertexAttrib4NsvARB, glVertexAttrib4NsvARB, NULL, _gloffset_VertexAttrib4NsvARB),
+    NAME_FUNC_OFFSET(16359, glVertexAttrib4NubARB, glVertexAttrib4NubARB, NULL, _gloffset_VertexAttrib4NubARB),
+    NAME_FUNC_OFFSET(16378, glVertexAttrib4NubvARB, glVertexAttrib4NubvARB, NULL, _gloffset_VertexAttrib4NubvARB),
+    NAME_FUNC_OFFSET(16398, glVertexAttrib4NuivARB, glVertexAttrib4NuivARB, NULL, _gloffset_VertexAttrib4NuivARB),
+    NAME_FUNC_OFFSET(16418, glVertexAttrib4NusvARB, glVertexAttrib4NusvARB, NULL, _gloffset_VertexAttrib4NusvARB),
+    NAME_FUNC_OFFSET(16438, glVertexAttrib4bvARB, glVertexAttrib4bvARB, NULL, _gloffset_VertexAttrib4bvARB),
+    NAME_FUNC_OFFSET(16456, glVertexAttrib4dARB, glVertexAttrib4dARB, NULL, _gloffset_VertexAttrib4dARB),
+    NAME_FUNC_OFFSET(16473, glVertexAttrib4dvARB, glVertexAttrib4dvARB, NULL, _gloffset_VertexAttrib4dvARB),
+    NAME_FUNC_OFFSET(16491, glVertexAttrib4fARB, glVertexAttrib4fARB, NULL, _gloffset_VertexAttrib4fARB),
+    NAME_FUNC_OFFSET(16508, glVertexAttrib4fvARB, glVertexAttrib4fvARB, NULL, _gloffset_VertexAttrib4fvARB),
+    NAME_FUNC_OFFSET(16526, glVertexAttrib4ivARB, glVertexAttrib4ivARB, NULL, _gloffset_VertexAttrib4ivARB),
+    NAME_FUNC_OFFSET(16544, glVertexAttrib4sARB, glVertexAttrib4sARB, NULL, _gloffset_VertexAttrib4sARB),
+    NAME_FUNC_OFFSET(16561, glVertexAttrib4svARB, glVertexAttrib4svARB, NULL, _gloffset_VertexAttrib4svARB),
+    NAME_FUNC_OFFSET(16579, glVertexAttrib4ubvARB, glVertexAttrib4ubvARB, NULL, _gloffset_VertexAttrib4ubvARB),
+    NAME_FUNC_OFFSET(16598, glVertexAttrib4uivARB, glVertexAttrib4uivARB, NULL, _gloffset_VertexAttrib4uivARB),
+    NAME_FUNC_OFFSET(16617, glVertexAttrib4usvARB, glVertexAttrib4usvARB, NULL, _gloffset_VertexAttrib4usvARB),
+    NAME_FUNC_OFFSET(16636, glVertexAttribPointerARB, glVertexAttribPointerARB, NULL, _gloffset_VertexAttribPointerARB),
+    NAME_FUNC_OFFSET(16658, glBindBufferARB, glBindBufferARB, NULL, _gloffset_BindBufferARB),
+    NAME_FUNC_OFFSET(16671, glBufferDataARB, glBufferDataARB, NULL, _gloffset_BufferDataARB),
+    NAME_FUNC_OFFSET(16684, glBufferSubDataARB, glBufferSubDataARB, NULL, _gloffset_BufferSubDataARB),
+    NAME_FUNC_OFFSET(16700, glDeleteBuffersARB, glDeleteBuffersARB, NULL, _gloffset_DeleteBuffersARB),
+    NAME_FUNC_OFFSET(16716, glGenBuffersARB, glGenBuffersARB, NULL, _gloffset_GenBuffersARB),
+    NAME_FUNC_OFFSET(16729, glGetBufferParameterivARB, glGetBufferParameterivARB, NULL, _gloffset_GetBufferParameterivARB),
+    NAME_FUNC_OFFSET(16752, glGetBufferPointervARB, glGetBufferPointervARB, NULL, _gloffset_GetBufferPointervARB),
+    NAME_FUNC_OFFSET(16772, glGetBufferSubDataARB, glGetBufferSubDataARB, NULL, _gloffset_GetBufferSubDataARB),
+    NAME_FUNC_OFFSET(16791, glIsBufferARB, glIsBufferARB, NULL, _gloffset_IsBufferARB),
+    NAME_FUNC_OFFSET(16802, glMapBufferARB, glMapBufferARB, NULL, _gloffset_MapBufferARB),
+    NAME_FUNC_OFFSET(16814, glUnmapBufferARB, glUnmapBufferARB, NULL, _gloffset_UnmapBufferARB),
+    NAME_FUNC_OFFSET(16828, glBeginQueryARB, glBeginQueryARB, NULL, _gloffset_BeginQueryARB),
+    NAME_FUNC_OFFSET(16841, glDeleteQueriesARB, glDeleteQueriesARB, NULL, _gloffset_DeleteQueriesARB),
+    NAME_FUNC_OFFSET(16857, glEndQueryARB, glEndQueryARB, NULL, _gloffset_EndQueryARB),
+    NAME_FUNC_OFFSET(16868, glGenQueriesARB, glGenQueriesARB, NULL, _gloffset_GenQueriesARB),
+    NAME_FUNC_OFFSET(16881, glGetQueryObjectivARB, glGetQueryObjectivARB, NULL, _gloffset_GetQueryObjectivARB),
+    NAME_FUNC_OFFSET(16900, glGetQueryObjectuivARB, glGetQueryObjectuivARB, NULL, _gloffset_GetQueryObjectuivARB),
+    NAME_FUNC_OFFSET(16920, glGetQueryivARB, glGetQueryivARB, NULL, _gloffset_GetQueryivARB),
+    NAME_FUNC_OFFSET(16933, glIsQueryARB, glIsQueryARB, NULL, _gloffset_IsQueryARB),
+    NAME_FUNC_OFFSET(16943, glCompileShaderARB, glCompileShaderARB, NULL, _gloffset_CompileShaderARB),
+    NAME_FUNC_OFFSET(16959, glGetActiveUniformARB, glGetActiveUniformARB, NULL, _gloffset_GetActiveUniformARB),
+    NAME_FUNC_OFFSET(16978, glGetShaderSourceARB, glGetShaderSourceARB, NULL, _gloffset_GetShaderSourceARB),
+    NAME_FUNC_OFFSET(16996, glGetUniformLocationARB, glGetUniformLocationARB, NULL, _gloffset_GetUniformLocationARB),
+    NAME_FUNC_OFFSET(17017, glGetUniformfvARB, glGetUniformfvARB, NULL, _gloffset_GetUniformfvARB),
+    NAME_FUNC_OFFSET(17032, glGetUniformivARB, glGetUniformivARB, NULL, _gloffset_GetUniformivARB),
+    NAME_FUNC_OFFSET(17047, glLinkProgramARB, glLinkProgramARB, NULL, _gloffset_LinkProgramARB),
+    NAME_FUNC_OFFSET(17061, glShaderSourceARB, glShaderSourceARB, NULL, _gloffset_ShaderSourceARB),
+    NAME_FUNC_OFFSET(17076, glUniform1fARB, glUniform1fARB, NULL, _gloffset_Uniform1fARB),
+    NAME_FUNC_OFFSET(17088, glUniform1fvARB, glUniform1fvARB, NULL, _gloffset_Uniform1fvARB),
+    NAME_FUNC_OFFSET(17101, glUniform1iARB, glUniform1iARB, NULL, _gloffset_Uniform1iARB),
+    NAME_FUNC_OFFSET(17113, glUniform1ivARB, glUniform1ivARB, NULL, _gloffset_Uniform1ivARB),
+    NAME_FUNC_OFFSET(17126, glUniform2fARB, glUniform2fARB, NULL, _gloffset_Uniform2fARB),
+    NAME_FUNC_OFFSET(17138, glUniform2fvARB, glUniform2fvARB, NULL, _gloffset_Uniform2fvARB),
+    NAME_FUNC_OFFSET(17151, glUniform2iARB, glUniform2iARB, NULL, _gloffset_Uniform2iARB),
+    NAME_FUNC_OFFSET(17163, glUniform2ivARB, glUniform2ivARB, NULL, _gloffset_Uniform2ivARB),
+    NAME_FUNC_OFFSET(17176, glUniform3fARB, glUniform3fARB, NULL, _gloffset_Uniform3fARB),
+    NAME_FUNC_OFFSET(17188, glUniform3fvARB, glUniform3fvARB, NULL, _gloffset_Uniform3fvARB),
+    NAME_FUNC_OFFSET(17201, glUniform3iARB, glUniform3iARB, NULL, _gloffset_Uniform3iARB),
+    NAME_FUNC_OFFSET(17213, glUniform3ivARB, glUniform3ivARB, NULL, _gloffset_Uniform3ivARB),
+    NAME_FUNC_OFFSET(17226, glUniform4fARB, glUniform4fARB, NULL, _gloffset_Uniform4fARB),
+    NAME_FUNC_OFFSET(17238, glUniform4fvARB, glUniform4fvARB, NULL, _gloffset_Uniform4fvARB),
+    NAME_FUNC_OFFSET(17251, glUniform4iARB, glUniform4iARB, NULL, _gloffset_Uniform4iARB),
+    NAME_FUNC_OFFSET(17263, glUniform4ivARB, glUniform4ivARB, NULL, _gloffset_Uniform4ivARB),
+    NAME_FUNC_OFFSET(17276, glUniformMatrix2fvARB, glUniformMatrix2fvARB, NULL, _gloffset_UniformMatrix2fvARB),
+    NAME_FUNC_OFFSET(17295, glUniformMatrix3fvARB, glUniformMatrix3fvARB, NULL, _gloffset_UniformMatrix3fvARB),
+    NAME_FUNC_OFFSET(17314, glUniformMatrix4fvARB, glUniformMatrix4fvARB, NULL, _gloffset_UniformMatrix4fvARB),
+    NAME_FUNC_OFFSET(17333, glUseProgramObjectARB, glUseProgramObjectARB, NULL, _gloffset_UseProgramObjectARB),
+    NAME_FUNC_OFFSET(17346, glValidateProgramARB, glValidateProgramARB, NULL, _gloffset_ValidateProgramARB),
+    NAME_FUNC_OFFSET(17364, glBindAttribLocationARB, glBindAttribLocationARB, NULL, _gloffset_BindAttribLocationARB),
+    NAME_FUNC_OFFSET(17385, glGetActiveAttribARB, glGetActiveAttribARB, NULL, _gloffset_GetActiveAttribARB),
+    NAME_FUNC_OFFSET(17403, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB),
+    NAME_FUNC_OFFSET(17423, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
+    NAME_FUNC_OFFSET(17437, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB),
+    NAME_FUNC_OFFSET(17454, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_SampleMaskSGIS),
+    NAME_FUNC_OFFSET(17470, gl_dispatch_stub_570, gl_dispatch_stub_570, NULL, _gloffset_SamplePatternSGIS),
+    NAME_FUNC_OFFSET(17489, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET(17507, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET(17528, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT),
+    NAME_FUNC_OFFSET(17550, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET(17569, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET(17591, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT),
+    NAME_FUNC_OFFSET(17614, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT),
+    NAME_FUNC_OFFSET(17633, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT),
+    NAME_FUNC_OFFSET(17653, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT),
+    NAME_FUNC_OFFSET(17672, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT),
+    NAME_FUNC_OFFSET(17692, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT),
+    NAME_FUNC_OFFSET(17711, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT),
+    NAME_FUNC_OFFSET(17731, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT),
+    NAME_FUNC_OFFSET(17750, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT),
+    NAME_FUNC_OFFSET(17770, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT),
+    NAME_FUNC_OFFSET(17789, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT),
+    NAME_FUNC_OFFSET(17809, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT),
+    NAME_FUNC_OFFSET(17829, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT),
+    NAME_FUNC_OFFSET(17850, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT),
+    NAME_FUNC_OFFSET(17870, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT),
+    NAME_FUNC_OFFSET(17891, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT),
+    NAME_FUNC_OFFSET(17911, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT),
+    NAME_FUNC_OFFSET(17932, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT),
+    NAME_FUNC_OFFSET(17956, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT),
+    NAME_FUNC_OFFSET(17974, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT),
+    NAME_FUNC_OFFSET(17994, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT),
+    NAME_FUNC_OFFSET(18012, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT),
+    NAME_FUNC_OFFSET(18024, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT),
+    NAME_FUNC_OFFSET(18037, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT),
+    NAME_FUNC_OFFSET(18049, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT),
+    NAME_FUNC_OFFSET(18062, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
+    NAME_FUNC_OFFSET(18082, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT),
+    NAME_FUNC_OFFSET(18106, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
+    NAME_FUNC_OFFSET(18120, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA),
+    NAME_FUNC_OFFSET(18137, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
+    NAME_FUNC_OFFSET(18152, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA),
+    NAME_FUNC_OFFSET(18170, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
+    NAME_FUNC_OFFSET(18184, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA),
+    NAME_FUNC_OFFSET(18201, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
+    NAME_FUNC_OFFSET(18216, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA),
+    NAME_FUNC_OFFSET(18234, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
+    NAME_FUNC_OFFSET(18248, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA),
+    NAME_FUNC_OFFSET(18265, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
+    NAME_FUNC_OFFSET(18280, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA),
+    NAME_FUNC_OFFSET(18298, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
+    NAME_FUNC_OFFSET(18312, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA),
+    NAME_FUNC_OFFSET(18329, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
+    NAME_FUNC_OFFSET(18344, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA),
+    NAME_FUNC_OFFSET(18362, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
+    NAME_FUNC_OFFSET(18376, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA),
+    NAME_FUNC_OFFSET(18393, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
+    NAME_FUNC_OFFSET(18408, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA),
+    NAME_FUNC_OFFSET(18426, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
+    NAME_FUNC_OFFSET(18440, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA),
+    NAME_FUNC_OFFSET(18457, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
+    NAME_FUNC_OFFSET(18472, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA),
+    NAME_FUNC_OFFSET(18490, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
+    NAME_FUNC_OFFSET(18504, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA),
+    NAME_FUNC_OFFSET(18521, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
+    NAME_FUNC_OFFSET(18536, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA),
+    NAME_FUNC_OFFSET(18554, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
+    NAME_FUNC_OFFSET(18568, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA),
+    NAME_FUNC_OFFSET(18585, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
+    NAME_FUNC_OFFSET(18600, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA),
+    NAME_FUNC_OFFSET(18618, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV),
+    NAME_FUNC_OFFSET(18635, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
+    NAME_FUNC_OFFSET(18655, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
+    NAME_FUNC_OFFSET(18672, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
+    NAME_FUNC_OFFSET(18698, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
+    NAME_FUNC_OFFSET(18727, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
+    NAME_FUNC_OFFSET(18742, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
+    NAME_FUNC_OFFSET(18760, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
+    NAME_FUNC_OFFSET(18779, gl_dispatch_stub_746, gl_dispatch_stub_746, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(18803, gl_dispatch_stub_746, gl_dispatch_stub_746, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(18830, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT),
+    NAME_FUNC_OFFSET(18848, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT),
+    NAME_FUNC_OFFSET(18867, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT),
+    NAME_FUNC_OFFSET(18892, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT),
+    NAME_FUNC_OFFSET(18913, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT),
+    NAME_FUNC_OFFSET(18935, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT),
+    NAME_FUNC_OFFSET(18961, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT),
+    NAME_FUNC_OFFSET(18984, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT),
+    NAME_FUNC_OFFSET(19007, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT),
+    NAME_FUNC_OFFSET(19030, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT),
+    NAME_FUNC_OFFSET(19048, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT),
+    NAME_FUNC_OFFSET(19067, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT),
+    NAME_FUNC_OFFSET(19084, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT),
+    NAME_FUNC_OFFSET(19122, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT),
+    NAME_FUNC_OFFSET(19151, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT),
+    NAME_FUNC_OFFSET(19167, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT),
+    NAME_FUNC_OFFSET(19184, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT),
+    NAME_FUNC_OFFSET(19206, gl_dispatch_stub_764, gl_dispatch_stub_764, NULL, _gloffset_BlitFramebufferEXT),
+    NAME_FUNC_OFFSET(19224, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT),
     NAME_FUNC_OFFSET(-1, NULL, NULL, NULL, 0)
 };
 
diff --git a/src/mesa/glapi/glthread.c b/src/mesa/glapi/glthread.c
index b818f4e26e..56ddf7c25a 100644
--- a/src/mesa/glapi/glthread.c
+++ b/src/mesa/glapi/glthread.c
@@ -25,16 +25,14 @@
 
 /*
  * XXX There's probably some work to do in order to make this file
- * truly reusable outside of Mesa.
+ * truly reusable outside of Mesa.  First, the glheader.h include must go.
  */
 
-
 #ifdef HAVE_DIX_CONFIG_H
 #include <dix-config.h>
 #endif
 
-#include <stdlib.h>
-#include <stdio.h>
+#include "main/glheader.h"
 #include "glthread.h"
 
 
diff --git a/src/mesa/glapi/glthread.h b/src/mesa/glapi/glthread.h
index e2765cebb1..ae2f79f37d 100644
--- a/src/mesa/glapi/glthread.h
+++ b/src/mesa/glapi/glthread.h
@@ -116,9 +116,49 @@ typedef pthread_mutex_t _glthread_Mutex;
 #define _glthread_UNLOCK_MUTEX(name) \
    (void) pthread_mutex_unlock(&(name))
 
-#endif /* PTHREADS */
+typedef pthread_cond_t _glthread_Cond;
 
+#define _glthread_DECLARE_STATIC_COND(name) \
+   static _glthread_Cond name = PTHREAD_COND_INITIALIZER
 
+#define _glthread_INIT_COND(cond)			\
+   pthread_cond_init(&(cond), NULL)
+
+#define _glthread_DESTROY_COND(name) \
+   pthread_cond_destroy(&(name))
+
+#define _glthread_COND_WAIT(cond, mutex) \
+  pthread_cond_wait(&(cond), &(mutex))
+
+#define _glthread_COND_SIGNAL(cond) \
+  pthread_cond_signal(&(cond))
+
+#define _glthread_COND_BROADCAST(cond) \
+  pthread_cond_broadcast(&(cond))
+
+
+#else /* PTHREADS */
+
+typedef unsigned int _glthread_Cond;
+#define _glthread_DECLARE_STATIC_COND(name) \
+//  #warning Condition variables not implemented.
+
+#define _glthread_INIT_COND(cond)	    \
+  ASSERT(0);
+
+#define _glthread_DESTROY_COND(name) \
+  ASSERT(0);
+
+#define _glthread_COND_WAIT(cond, mutex) \
+  ASSERT(0);
+
+#define _glthread_COND_SIGNAL(cond) \
+  ASSERT(0);
+
+#define _glthread_COND_BROADCAST(cond) \
+  ASSERT(0);
+
+#endif
 
 
 /*
@@ -259,11 +299,11 @@ typedef benaphore _glthread_Mutex;
  * THREADS not defined
  */
 
-typedef int _glthread_TSD;
+typedef GLuint _glthread_TSD;
 
-typedef int _glthread_Thread;
+typedef GLuint _glthread_Thread;
 
-typedef int _glthread_Mutex;
+typedef GLuint _glthread_Mutex;
 
 #define _glthread_DECLARE_STATIC_MUTEX(name)  static _glthread_Mutex name = 0
 
@@ -298,6 +338,10 @@ _glthread_GetTSD(_glthread_TSD *);
 extern void
 _glthread_SetTSD(_glthread_TSD *, void *);
 
+#if !defined __GNUC__ || __GNUC__ < 3
+#  define __builtin_expect(x, y) x
+#endif
+
 #if defined(GLX_USE_TLS)
 
 extern __thread struct _glapi_table * _glapi_tls_Dispatch
diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c
index bae3bf11cb..f8ee0d59dd 100644
--- a/src/mesa/main/api_exec.c
+++ b/src/mesa/main/api_exec.c
@@ -880,5 +880,12 @@ _mesa_init_exec_table(struct _glapi_table *exec)
 
    /* GL_ATI_separate_stencil */
    SET_StencilFuncSeparateATI(exec, _mesa_StencilFuncSeparateATI);
+
+#if FEATURE_ARB_framebuffer_object
+   /* The ARB_fbo functions are the union of
+    * GL_EXT_fbo, GL_EXT_framebuffer_blit, GL_EXT_texture_array
+    */
+   SET_RenderbufferStorageMultisample(exec, _mesa_RenderbufferStorageMultisample);
+#endif
 }
 
diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c
index 98dfbb105f..5c8955d7c8 100644
--- a/src/mesa/main/api_validate.c
+++ b/src/mesa/main/api_validate.c
@@ -78,6 +78,23 @@ max_buffer_index(GLcontext *ctx, GLuint count, GLenum type,
    return max;
 }
 
+static GLboolean
+check_valid_to_render(GLcontext *ctx, char *function)
+{
+   if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
+      _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
+                  "glDraw%s(incomplete framebuffer)", function);
+      return GL_FALSE;
+   }
+
+   /* Always need vertex positions, unless a vertex program is in use */
+   if (!ctx->VertexProgram._Current &&
+       !ctx->Array.ArrayObj->Vertex.Enabled &&
+       !ctx->Array.ArrayObj->VertexAttrib[0].Enabled)
+      return GL_FALSE;
+
+   return GL_TRUE;
+}
 
 GLboolean
 _mesa_validate_DrawElements(GLcontext *ctx,
@@ -108,10 +125,7 @@ _mesa_validate_DrawElements(GLcontext *ctx,
    if (ctx->NewState)
       _mesa_update_state(ctx);
 
-   /* Always need vertex positions, unless a vertex program is in use */
-   if (!ctx->VertexProgram._Current &&
-       !ctx->Array.ArrayObj->Vertex.Enabled &&
-       !ctx->Array.ArrayObj->VertexAttrib[0].Enabled)
+   if (!check_valid_to_render(ctx, "Elements"))
       return GL_FALSE;
 
    /* Vertex buffer object tests */
@@ -119,6 +133,12 @@ _mesa_validate_DrawElements(GLcontext *ctx,
       /* use indices in the buffer object */
       GLuint indexBytes;
 
+      if (!ctx->Array.ElementArrayBufferObj->Size) {
+         _mesa_warning(ctx,
+                       "glDrawElements called with empty array elements buffer");
+         return GL_FALSE;
+      }
+
       if (type == GL_UNSIGNED_INT) {
          indexBytes = count * sizeof(GLuint);
       }
@@ -155,7 +175,6 @@ _mesa_validate_DrawElements(GLcontext *ctx,
    return GL_TRUE;
 }
 
-
 GLboolean
 _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
 				 GLuint start, GLuint end,
@@ -190,10 +209,7 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
    if (ctx->NewState)
       _mesa_update_state(ctx);
 
-   /* Always need vertex positions, unless a vertex program is in use */
-   if (!ctx->VertexProgram._Current &&
-       !ctx->Array.ArrayObj->Vertex.Enabled &&
-       !ctx->Array.ArrayObj->VertexAttrib[0].Enabled)
+   if (!check_valid_to_render(ctx, "RangeElements"))
       return GL_FALSE;
 
    /* Vertex buffer object tests */
@@ -261,10 +277,7 @@ _mesa_validate_DrawArrays(GLcontext *ctx,
    if (ctx->NewState)
       _mesa_update_state(ctx);
 
-   /* Always need vertex positions, unless a vertex program is in use */
-   if (!ctx->VertexProgram._Current &&
-       !ctx->Array.ArrayObj->Vertex.Enabled &&
-       !ctx->Array.ArrayObj->VertexAttrib[0].Enabled)
+   if (!check_valid_to_render(ctx, "Arrays"))
       return GL_FALSE;
 
    if (ctx->Const.CheckArrayBounds) {
diff --git a/src/mesa/main/arrayobj.c b/src/mesa/main/arrayobj.c
index 1461239317..b04095fd16 100644
--- a/src/mesa/main/arrayobj.c
+++ b/src/mesa/main/arrayobj.c
@@ -98,6 +98,28 @@ _mesa_delete_array_object( GLcontext *ctx, struct gl_array_object *obj )
 }
 
 
+static void
+init_array(GLcontext *ctx,
+           struct gl_client_array *array, GLint size, GLint type)
+{
+   array->Size = size;
+   array->Type = type;
+   array->Format = GL_RGBA; /* only significant for GL_EXT_vertex_array_bgra */
+   array->Stride = 0;
+   array->StrideB = 0;
+   array->Ptr = NULL;
+   array->Enabled = GL_FALSE;
+   array->Normalized = GL_FALSE;
+#if FEATURE_ARB_vertex_buffer_object
+   /* Vertex array buffers */
+   array->BufferObj = ctx->Array.NullBufferObj;
+#endif
+}
+
+
+/**
+ * Initialize a gl_array_object's arrays.
+ */
 void
 _mesa_initialize_array_object( GLcontext *ctx,
 			       struct gl_array_object *obj,
@@ -107,87 +129,23 @@ _mesa_initialize_array_object( GLcontext *ctx,
 
    obj->Name = name;
 
-   /* Vertex arrays */
-   obj->Vertex.Size = 4;
-   obj->Vertex.Type = GL_FLOAT;
-   obj->Vertex.Stride = 0;
-   obj->Vertex.StrideB = 0;
-   obj->Vertex.Ptr = NULL;
-   obj->Vertex.Enabled = GL_FALSE;
-   obj->Normal.Type = GL_FLOAT;
-   obj->Normal.Stride = 0;
-   obj->Normal.StrideB = 0;
-   obj->Normal.Ptr = NULL;
-   obj->Normal.Enabled = GL_FALSE;
-   obj->Color.Size = 4;
-   obj->Color.Type = GL_FLOAT;
-   obj->Color.Stride = 0;
-   obj->Color.StrideB = 0;
-   obj->Color.Ptr = NULL;
-   obj->Color.Enabled = GL_FALSE;
-   obj->SecondaryColor.Size = 4;
-   obj->SecondaryColor.Type = GL_FLOAT;
-   obj->SecondaryColor.Stride = 0;
-   obj->SecondaryColor.StrideB = 0;
-   obj->SecondaryColor.Ptr = NULL;
-   obj->SecondaryColor.Enabled = GL_FALSE;
-   obj->FogCoord.Size = 1;
-   obj->FogCoord.Type = GL_FLOAT;
-   obj->FogCoord.Stride = 0;
-   obj->FogCoord.StrideB = 0;
-   obj->FogCoord.Ptr = NULL;
-   obj->FogCoord.Enabled = GL_FALSE;
-   obj->Index.Type = GL_FLOAT;
-   obj->Index.Stride = 0;
-   obj->Index.StrideB = 0;
-   obj->Index.Ptr = NULL;
-   obj->Index.Enabled = GL_FALSE;
-   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
-      obj->TexCoord[i].Size = 4;
-      obj->TexCoord[i].Type = GL_FLOAT;
-      obj->TexCoord[i].Stride = 0;
-      obj->TexCoord[i].StrideB = 0;
-      obj->TexCoord[i].Ptr = NULL;
-      obj->TexCoord[i].Enabled = GL_FALSE;
+   /* Init the individual arrays */
+   init_array(ctx, &obj->Vertex, 4, GL_FLOAT);
+   init_array(ctx, &obj->Normal, 3, GL_FLOAT);
+   init_array(ctx, &obj->Color, 4, GL_FLOAT);
+   init_array(ctx, &obj->SecondaryColor, 4, GL_FLOAT);
+   init_array(ctx, &obj->FogCoord, 1, GL_FLOAT);
+   init_array(ctx, &obj->Index, 1, GL_FLOAT);
+   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
+      init_array(ctx, &obj->TexCoord[i], 4, GL_FLOAT);
    }
-   obj->EdgeFlag.Stride = 0;
-   obj->EdgeFlag.StrideB = 0;
-   obj->EdgeFlag.Ptr = NULL;
-   obj->EdgeFlag.Enabled = GL_FALSE;
+   init_array(ctx, &obj->EdgeFlag, 1, GL_BOOL);
    for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-      obj->VertexAttrib[i].Size = 4;
-      obj->VertexAttrib[i].Type = GL_FLOAT;
-      obj->VertexAttrib[i].Stride = 0;
-      obj->VertexAttrib[i].StrideB = 0;
-      obj->VertexAttrib[i].Ptr = NULL;
-      obj->VertexAttrib[i].Enabled = GL_FALSE;
-      obj->VertexAttrib[i].Normalized = GL_FALSE;
+      init_array(ctx, &obj->VertexAttrib[i], 4, GL_FLOAT);
    }
 
 #if FEATURE_point_size_array
-   obj->PointSize.Type = GL_FLOAT;
-   obj->PointSize.Stride = 0;
-   obj->PointSize.StrideB = 0;
-   obj->PointSize.Ptr = NULL;
-   obj->PointSize.Enabled = GL_FALSE;
-   obj->PointSize.BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-#if FEATURE_ARB_vertex_buffer_object
-   /* Vertex array buffers */
-   obj->Vertex.BufferObj = ctx->Array.NullBufferObj;
-   obj->Normal.BufferObj = ctx->Array.NullBufferObj;
-   obj->Color.BufferObj = ctx->Array.NullBufferObj;
-   obj->SecondaryColor.BufferObj = ctx->Array.NullBufferObj;
-   obj->FogCoord.BufferObj = ctx->Array.NullBufferObj;
-   obj->Index.BufferObj = ctx->Array.NullBufferObj;
-   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
-      obj->TexCoord[i].BufferObj = ctx->Array.NullBufferObj;
-   }
-   obj->EdgeFlag.BufferObj = ctx->Array.NullBufferObj;
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-      obj->VertexAttrib[i].BufferObj = ctx->Array.NullBufferObj;
-   }
+   init_array(ctx, &obj->PointSize, 1, GL_FLOAT);
 #endif
 }
 
@@ -335,7 +293,7 @@ _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids)
 	 unbind_buffer_object( ctx, obj->SecondaryColor.BufferObj );
 	 unbind_buffer_object( ctx, obj->FogCoord.BufferObj );
 	 unbind_buffer_object( ctx, obj->Index.BufferObj );
-	 for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
+	 for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
 	    unbind_buffer_object( ctx, obj->TexCoord[i].BufferObj );
 	 }
 	 unbind_buffer_object( ctx, obj->EdgeFlag.BufferObj );
diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c
index dc85da2518..218e0aeb6b 100644
--- a/src/mesa/main/attrib.c
+++ b/src/mesa/main/attrib.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.2
+ * Version:  7.3
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -224,7 +225,7 @@ _mesa_PushAttrib(GLbitfield mask)
       attr->SampleAlphaToOne = ctx->Multisample.SampleAlphaToOne;
       attr->SampleCoverage = ctx->Multisample.SampleCoverage;
       attr->SampleCoverageInvert = ctx->Multisample.SampleCoverageInvert;
-      for (i=0; i<MAX_TEXTURE_UNITS; i++) {
+      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
          attr->Texture[i] = ctx->Texture.Unit[i].Enabled;
          attr->TexGen[i] = ctx->Texture.Unit[i].TexGenEnabled;
          attr->TextureColorTable[i] = ctx->Texture.Unit[i].ColorTableEnabled;
@@ -819,15 +820,9 @@ pop_texture_group(GLcontext *ctx, struct texture_state *texstate)
             _mesa_TexParameterf(target, GL_TEXTURE_MAX_ANISOTROPY_EXT,
                                 obj->MaxAnisotropy);
          }
-         if (ctx->Extensions.SGIX_shadow) {
-            _mesa_TexParameteri(target, GL_TEXTURE_COMPARE_SGIX,
-                                obj->CompareFlag);
-            _mesa_TexParameteri(target, GL_TEXTURE_COMPARE_OPERATOR_SGIX,
-                                obj->CompareOperator);
-         }
-         if (ctx->Extensions.SGIX_shadow_ambient) {
-            _mesa_TexParameterf(target, GL_SHADOW_AMBIENT_SGIX,
-                                obj->ShadowAmbient);
+         if (ctx->Extensions.ARB_shadow_ambient) {
+            _mesa_TexParameterf(target, GL_TEXTURE_COMPARE_FAIL_VALUE_ARB,
+                                obj->CompareFailValue);
          }
       }
 
@@ -908,14 +903,13 @@ _mesa_PopAttrib(void)
                    * function, but legal for the later.
                    */
                   GLboolean multipleBuffers = GL_FALSE;
-                  if (ctx->Extensions.ARB_draw_buffers) {
-                     GLuint i;
-                     for (i = 1; i < ctx->Const.MaxDrawBuffers; i++) {
-                        if (color->DrawBuffer[i] != GL_NONE) {
-                           multipleBuffers = GL_TRUE;
-                           break;
-                        }
-                     }
+		  GLuint i;
+
+		  for (i = 1; i < ctx->Const.MaxDrawBuffers; i++) {
+		     if (color->DrawBuffer[i] != GL_NONE) {
+			multipleBuffers = GL_TRUE;
+			break;
+		     }
                   }
                   /* Call the API_level functions, not _mesa_drawbuffers()
                    * since we need to do error checking on the pop'd
@@ -1011,9 +1005,8 @@ _mesa_PopAttrib(void)
                _mesa_Hint(GL_FOG_HINT, hint->Fog);
                _mesa_Hint(GL_CLIP_VOLUME_CLIPPING_HINT_EXT,
                           hint->ClipVolumeClipping);
-               if (ctx->Extensions.ARB_texture_compression)
-                  _mesa_Hint(GL_TEXTURE_COMPRESSION_HINT_ARB,
-                             hint->TextureCompression);
+	       _mesa_Hint(GL_TEXTURE_COMPRESSION_HINT_ARB,
+			  hint->TextureCompression);
             }
             break;
          case GL_LIGHTING_BIT:
@@ -1274,9 +1267,6 @@ adjust_buffer_object_ref_counts(struct gl_array_attrib *array, GLint step)
       array->ArrayObj->TexCoord[i].BufferObj->RefCount += step;
    for (i = 0; i < VERT_ATTRIB_MAX; i++)
       array->ArrayObj->VertexAttrib[i].BufferObj->RefCount += step;
-
-   array->ArrayBufferObj->RefCount += step;
-   array->ElementArrayBufferObj->RefCount += step;
 }
 
 
diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c
index 918e87246f..016543da01 100644
--- a/src/mesa/main/bufferobj.c
+++ b/src/mesa/main/bufferobj.c
@@ -38,6 +38,13 @@
 #include "bufferobj.h"
 
 
+#ifdef FEATURE_OES_mapbuffer
+#define DEFAULT_ACCESS GL_WRITE_ONLY;
+#else
+#define DEFAULT_ACCESS GL_READ_WRITE;
+#endif
+
+
 /**
  * Get the buffer object bound to the specified target in a GL context.
  *
@@ -255,7 +262,7 @@ _mesa_initialize_buffer_object( struct gl_buffer_object *obj,
    obj->RefCount = 1;
    obj->Name = name;
    obj->Usage = GL_STATIC_DRAW_ARB;
-   obj->Access = GL_READ_WRITE_ARB;
+   obj->Access = DEFAULT_ACCESS;
 }
 
 
@@ -664,8 +671,8 @@ _mesa_map_drawpix_pbo(GLcontext *ctx,
  * \sa _mesa_unmap_bitmap_pbo
  */
 void
-_mesa_unmap_drapix_pbo(GLcontext *ctx,
-                       const struct gl_pixelstore_attrib *unpack)
+_mesa_unmap_drawpix_pbo(GLcontext *ctx,
+                        const struct gl_pixelstore_attrib *unpack)
 {
    if (unpack->BufferObj->Name) {
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
@@ -801,7 +808,7 @@ _mesa_DeleteBuffersARB(GLsizei n, const GLuint *ids)
          unbind(ctx, &ctx->Array.ArrayObj->FogCoord.BufferObj, bufObj);
          unbind(ctx, &ctx->Array.ArrayObj->Index.BufferObj, bufObj);
          unbind(ctx, &ctx->Array.ArrayObj->EdgeFlag.BufferObj, bufObj);
-         for (j = 0; j < MAX_TEXTURE_UNITS; j++) {
+         for (j = 0; j < MAX_TEXTURE_COORD_UNITS; j++) {
             unbind(ctx, &ctx->Array.ArrayObj->TexCoord[j].BufferObj, bufObj);
          }
          for (j = 0; j < VERT_ATTRIB_MAX; j++) {
@@ -1065,7 +1072,7 @@ _mesa_UnmapBufferARB(GLenum target)
       status = ctx->Driver.UnmapBuffer( ctx, target, bufObj );
    }
 
-   bufObj->Access = GL_READ_WRITE_ARB; /* initial value, OK? */
+   bufObj->Access = DEFAULT_ACCESS;
    bufObj->Pointer = NULL;
 
    return status;
diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h
index 163fe241aa..3c08f0083c 100644
--- a/src/mesa/main/bufferobj.h
+++ b/src/mesa/main/bufferobj.h
@@ -103,9 +103,8 @@ _mesa_map_drawpix_pbo(GLcontext *ctx,
                       const GLvoid *pixels);
 
 extern void
-_mesa_unmap_drapix_pbo(GLcontext *ctx,
-                       const struct gl_pixelstore_attrib *unpack);
-
+_mesa_unmap_drawpix_pbo(GLcontext *ctx,
+                        const struct gl_pixelstore_attrib *unpack);
 
 extern void *
 _mesa_map_readpix_pbo(GLcontext *ctx,
diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c
index b23d2f612b..818d068a12 100644
--- a/src/mesa/main/buffers.c
+++ b/src/mesa/main/buffers.c
@@ -247,13 +247,14 @@ _mesa_DrawBuffer(GLenum buffer)
       destMask = draw_buffer_enum_to_bitmask(buffer);
       if (destMask == BAD_MASK) {
          /* totally bogus buffer */
-         _mesa_error(ctx, GL_INVALID_ENUM, "glDrawBuffer(buffer)");
+         _mesa_error(ctx, GL_INVALID_ENUM, "glDrawBuffer(buffer=0x%x)", buffer);
          return;
       }
       destMask &= supportedMask;
       if (destMask == 0x0) {
          /* none of the named color buffers exist! */
-         _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawBuffer(buffer)");
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "glDrawBuffer(buffer=0x%x)", buffer);
          return;
       }
    }
@@ -289,10 +290,6 @@ _mesa_DrawBuffersARB(GLsizei n, const GLenum *buffers)
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (!ctx->Extensions.ARB_draw_buffers) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawBuffersARB");
-      return;
-   }
    if (n < 1 || n > (GLsizei) ctx->Const.MaxDrawBuffers) {
       _mesa_error(ctx, GL_INVALID_VALUE, "glDrawBuffersARB(n)");
       return;
@@ -419,7 +416,7 @@ _mesa_drawbuffers(GLcontext *ctx, GLuint n, const GLenum *buffers,
       }
    }
 
-   ctx->NewState |= _NEW_COLOR;
+   ctx->NewState |= _NEW_BUFFERS;
 }
 
 
@@ -494,6 +491,7 @@ _mesa_ReadBuffer(GLenum buffer)
    /* OK, all error checking has been completed now */
 
    _mesa_readbuffer(ctx, buffer, srcBuffer);
+   ctx->NewState |= _NEW_BUFFERS;
 
    /*
     * Call device driver function.
diff --git a/src/mesa/main/colormac.h b/src/mesa/main/colormac.h
index a34bd2ed38..74692e9a98 100644
--- a/src/mesa/main/colormac.h
+++ b/src/mesa/main/colormac.h
@@ -195,6 +195,10 @@ do {						\
 #define PACK_COLOR_565_REV( X, Y, Z ) \
    (((X) & 0xf8) | ((Y) & 0xe0) >> 5 | (((Y) & 0x1c) << 11) | (((Z) & 0xf8) << 5))
 
+#define PACK_COLOR_5551( R, G, B, A )					\
+   ((((R) & 0xf8) << 8) | (((G) & 0xf8) << 3) | (((B) & 0xf8) >> 2) |	\
+    ((A) ? 1 : 0))
+
 #define PACK_COLOR_1555( A, B, G, R )					\
    ((((B) & 0xf8) << 7) | (((G) & 0xf8) << 2) | (((R) & 0xf8) >> 3) |	\
     ((A) ? 0x8000 : 0))
diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h
index 882e2f224a..c3feffda98 100644
--- a/src/mesa/main/config.h
+++ b/src/mesa/main/config.h
@@ -1,13 +1,9 @@
-/**
- * \file config.h
- * Tunable configuration parameters.
- */
-
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.3
  *
  * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2008  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -27,6 +23,10 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
+/**
+ * \file config.h
+ * Tunable configuration parameters.
+ */
 
 #ifndef MESA_CONFIG_H_INCLUDED
 #define MESA_CONFIG_H_INCLUDED
@@ -115,27 +115,28 @@
 /** Maximum number of layers in a 1D or 2D array texture - GL_MESA_texture_array */
 #define MAX_ARRAY_TEXTURE_LAYERS 64
 
-/** Number of texture units - GL_ARB_multitexture
- * This needs to be the larger of MAX_TEXTURE_COORD_UNITS and
- * MAX_TEXTURE_IMAGE_UNITS seen below, since MAX_TEXTURE_UNITS is used
- * to dimension some arrays that store both coord and image data.
-*/
-#define MAX_TEXTURE_UNITS 8
-
-/*@}*/
+/**
+ * Max number of texture coordinate units.  This mainly just applies to
+ * the fixed-function vertex code.  This will be difficult to raise above
+ * eight because of various vertex attribute bitvectors.
+ */
+#define MAX_TEXTURE_COORD_UNITS 8
 
+/**
+ * Max number of texture image units.  Also determines number of texture
+ * samplers in shaders.
+ */
+#define MAX_TEXTURE_IMAGE_UNITS 16
 
 /**
- * \name Separate numbers of texture coordinates and texture image units.
- *
- * These values will eventually replace most instances of MAX_TEXTURE_UNITS.
- * We should always have MAX_TEXTURE_COORD_UNITS <= MAX_TEXTURE_IMAGE_UNITS.
- * And, GL_MAX_TEXTURE_UNITS <= MAX_TEXTURE_COORD_UNITS.
+ * Larger of MAX_TEXTURE_COORD_UNITS and MAX_TEXTURE_IMAGE_UNITS.
+ * This value is only used for dimensioning arrays.
+ * Either MAX_TEXTURE_COORD_UNITS or MAX_TEXTURE_IMAGE_UNITS (or the
+ * corresponding ctx->Const.MaxTextureCoord/ImageUnits fields) should be
+ * used almost everywhere else.
  */
-/*@{*/
-#define MAX_TEXTURE_COORD_UNITS 8
-#define MAX_TEXTURE_IMAGE_UNITS 8
-/*@}*/
+#define MAX_TEXTURE_UNITS ((MAX_TEXTURE_COORD_UNITS > MAX_TEXTURE_IMAGE_UNITS) ? MAX_TEXTURE_COORD_UNITS : MAX_TEXTURE_IMAGE_UNITS)
+
 
 /** 
  * Maximum viewport/image width. Must accomodate all texture sizes too. 
@@ -176,23 +177,23 @@
 /** For GL_ARB_fragment_program */
 /*@{*/
 #define MAX_FRAGMENT_PROGRAM_ADDRESS_REGS 0
-#define MAX_FRAGMENT_PROGRAM_ALU_INSTRUCTIONS 48
-#define MAX_FRAGMENT_PROGRAM_TEX_INSTRUCTIONS 24
-#define MAX_FRAGMENT_PROGRAM_TEX_INDIRECTIONS  4
 /*@}*/
 
 /** For any program target/extension */
 /*@{*/
-#define MAX_PROGRAM_LOCAL_PARAMS 128 /* KW: power of two */
+#define MAX_PROGRAM_INSTRUCTIONS  (16 * 1024)
+#define MAX_PROGRAM_LOCAL_PARAMS 256 /**< per-program constants (power of two) */
 #define MAX_PROGRAM_ENV_PARAMS 128
 #define MAX_PROGRAM_MATRICES 8
 #define MAX_PROGRAM_MATRIX_STACK_DEPTH 4
 #define MAX_PROGRAM_CALL_DEPTH 8
 #define MAX_PROGRAM_TEMPS 128
 #define MAX_PROGRAM_ADDRESS_REGS 2
-#define MAX_UNIFORMS 128   /**< number of float components */
+#define MAX_UNIFORMS 256   /**< number of vec4 uniforms */
 #define MAX_VARYING 8      /**< number of float[4] vectors */
-#define MAX_SAMPLERS 8
+#define MAX_SAMPLERS MAX_TEXTURE_IMAGE_UNITS
+#define MAX_PROGRAM_INPUTS 32
+#define MAX_PROGRAM_OUTPUTS 32
 /*@}*/
 
 /** For GL_NV_vertex_program */
@@ -218,8 +219,8 @@
 /** For GL_ARB_vertex_shader */
 /*@{*/
 #define MAX_VERTEX_ATTRIBS 16
-#define MAX_VERTEX_TEXTURE_IMAGE_UNITS MAX_TEXTURE_UNITS
-#define MAX_COMBINED_TEXTURE_IMAGE_UNITS (MAX_TEXTURE_IMAGE_UNITS + MAX_VERTEX_TEXTURE_IMAGE_UNITS)
+#define MAX_VERTEX_TEXTURE_IMAGE_UNITS MAX_TEXTURE_IMAGE_UNITS
+#define MAX_COMBINED_TEXTURE_IMAGE_UNITS MAX_TEXTURE_IMAGE_UNITS
 /*@}*/
 
 
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index a41cb42952..c5b90d9cda 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -1,14 +1,9 @@
-/**
- * \file context.c
- * Mesa context/visual/framebuffer management functions.
- * \author Brian Paul
- */
-
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.3
  *
  * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2008  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -28,6 +23,11 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
+/**
+ * \file context.c
+ * Mesa context/visual/framebuffer management functions.
+ * \author Brian Paul
+ */
 
 /**
  * \mainpage Mesa Main Module
@@ -145,18 +145,13 @@
 #include "glapi/glthread.h"
 #include "glapi/glapioffsets.h"
 #include "glapi/glapitable.h"
-#if FEATURE_NV_vertex_program || FEATURE_NV_fragment_program
 #include "shader/program.h"
-#endif
 #include "shader/shader_api.h"
 #if FEATURE_ATI_fragment_shader
 #include "shader/atifragshader.h"
 #endif
 #if _HAVE_FULL_GL
-#include "math/m_translate.h"
 #include "math/m_matrix.h"
-#include "math/m_xform.h"
-#include "math/mathmod.h"
 #endif
 
 #ifdef USE_SPARC_ASM
@@ -186,9 +181,11 @@ GLfloat _mesa_ubyte_to_float_color_tab[256];
  * We have to finish any pending rendering.
  */
 void
-_mesa_notifySwapBuffers(__GLcontext *gc)
+_mesa_notifySwapBuffers(__GLcontext *ctx)
 {
-   FLUSH_VERTICES( gc, 0 );
+   if (ctx->Driver.Flush) {
+      ctx->Driver.Flush(ctx);
+   }
 }
 
 
@@ -387,13 +384,9 @@ one_time_init( GLcontext *ctx )
 
       _mesa_init_sqrt_table();
 
-#if _HAVE_FULL_GL
-      _math_init();
-
       for (i = 0; i < 256; i++) {
          _mesa_ubyte_to_float_color_tab[i] = (float) i / 255.0F;
       }
-#endif
 
 #ifdef USE_SPARC_ASM
       _mesa_init_sparc_glapi_relocs();
@@ -581,7 +574,7 @@ static void
 delete_displaylist_cb(GLuint id, void *data, void *userData)
 {
 #if FEATURE_dlist
-   struct mesa_display_list *list = (struct mesa_display_list *) data;
+   struct gl_display_list *list = (struct gl_display_list *) data;
    GLcontext *ctx = (GLcontext *) userData;
    _mesa_delete_list(ctx, list);
 #endif
@@ -611,6 +604,7 @@ delete_program_cb(GLuint id, void *data, void *userData)
    ctx->Driver.DeleteProgram(ctx, prog);
 }
 
+#if FEATURE_ATI_fragment_shader
 /**
  * Callback for deleting an ATI fragment shader object.
  * Called by _mesa_HashDeleteAll().
@@ -622,6 +616,7 @@ delete_fragshader_cb(GLuint id, void *data, void *userData)
    GLcontext *ctx = (GLcontext *) userData;
    _mesa_delete_ati_fragment_shader(ctx, shader);
 }
+#endif
 
 /**
  * Callback for deleting a buffer object.  Called by _mesa_HashDeleteAll().
@@ -712,7 +707,6 @@ delete_renderbuffer_cb(GLuint id, void *data, void *userData)
 }
 
 
-
 /**
  * Deallocate a shared state object and all children structures.
  *
@@ -818,11 +812,33 @@ _mesa_init_current(GLcontext *ctx)
 
 
 /**
- * Init vertex/fragment program native limits from logical limits.
+ * Init vertex/fragment program limits.
+ * Important: drivers should override these with actual limits.
  */
 static void
-init_natives(struct gl_program_constants *prog)
+init_program_limits(GLenum type, struct gl_program_constants *prog)
 {
+   prog->MaxInstructions = MAX_PROGRAM_INSTRUCTIONS;
+   prog->MaxAluInstructions = MAX_PROGRAM_INSTRUCTIONS;
+   prog->MaxTexInstructions = MAX_PROGRAM_INSTRUCTIONS;
+   prog->MaxTexIndirections = MAX_PROGRAM_INSTRUCTIONS;
+   prog->MaxTemps = MAX_PROGRAM_TEMPS;
+   prog->MaxEnvParams = MAX_PROGRAM_ENV_PARAMS;
+   prog->MaxLocalParams = MAX_PROGRAM_LOCAL_PARAMS;
+   prog->MaxUniformComponents = 4 * MAX_UNIFORMS;
+
+   if (type == GL_VERTEX_PROGRAM_ARB) {
+      prog->MaxParameters = MAX_NV_VERTEX_PROGRAM_PARAMS;
+      prog->MaxAttribs = MAX_NV_VERTEX_PROGRAM_INPUTS;
+      prog->MaxAddressRegs = MAX_VERTEX_PROGRAM_ADDRESS_REGS;
+   }
+   else {
+      prog->MaxParameters = MAX_NV_FRAGMENT_PROGRAM_PARAMS;
+      prog->MaxAttribs = MAX_NV_FRAGMENT_PROGRAM_INPUTS;
+      prog->MaxAddressRegs = MAX_FRAGMENT_PROGRAM_ADDRESS_REGS;
+   }
+
+   /* copy the above limits to init native limits */
    prog->MaxNativeInstructions = prog->MaxInstructions;
    prog->MaxNativeAluInstructions = prog->MaxAluInstructions;
    prog->MaxNativeTexInstructions = prog->MaxTexInstructions;
@@ -847,9 +863,6 @@ _mesa_init_constants(GLcontext *ctx)
    assert(MAX_TEXTURE_LEVELS >= MAX_3D_TEXTURE_LEVELS);
    assert(MAX_TEXTURE_LEVELS >= MAX_CUBE_TEXTURE_LEVELS);
 
-   assert(MAX_TEXTURE_UNITS >= MAX_TEXTURE_COORD_UNITS);
-   assert(MAX_TEXTURE_UNITS >= MAX_TEXTURE_IMAGE_UNITS);
-
    /* Constants, may be overriden (usually only reduced) by device drivers */
    ctx->Const.MaxTextureLevels = MAX_TEXTURE_LEVELS;
    ctx->Const.Max3DTextureLevels = MAX_3D_TEXTURE_LEVELS;
@@ -884,33 +897,10 @@ _mesa_init_constants(GLcontext *ctx)
    ctx->Const.MaxViewportWidth = MAX_WIDTH;
    ctx->Const.MaxViewportHeight = MAX_HEIGHT;
 #if FEATURE_ARB_vertex_program
-   ctx->Const.VertexProgram.MaxInstructions = MAX_NV_VERTEX_PROGRAM_INSTRUCTIONS;
-   ctx->Const.VertexProgram.MaxAluInstructions = 0;
-   ctx->Const.VertexProgram.MaxTexInstructions = 0;
-   ctx->Const.VertexProgram.MaxTexIndirections = 0;
-   ctx->Const.VertexProgram.MaxAttribs = MAX_NV_VERTEX_PROGRAM_INPUTS;
-   ctx->Const.VertexProgram.MaxTemps = MAX_PROGRAM_TEMPS;
-   ctx->Const.VertexProgram.MaxParameters = MAX_NV_VERTEX_PROGRAM_PARAMS;
-   ctx->Const.VertexProgram.MaxLocalParams = MAX_PROGRAM_LOCAL_PARAMS;
-   ctx->Const.VertexProgram.MaxEnvParams = MAX_PROGRAM_ENV_PARAMS;
-   ctx->Const.VertexProgram.MaxAddressRegs = MAX_VERTEX_PROGRAM_ADDRESS_REGS;
-   ctx->Const.VertexProgram.MaxUniformComponents = 4 * MAX_UNIFORMS;
-   init_natives(&ctx->Const.VertexProgram);
+   init_program_limits(GL_VERTEX_PROGRAM_ARB, &ctx->Const.VertexProgram);
 #endif
-
 #if FEATURE_ARB_fragment_program
-   ctx->Const.FragmentProgram.MaxInstructions = MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS;
-   ctx->Const.FragmentProgram.MaxAluInstructions = MAX_FRAGMENT_PROGRAM_ALU_INSTRUCTIONS;
-   ctx->Const.FragmentProgram.MaxTexInstructions = MAX_FRAGMENT_PROGRAM_TEX_INSTRUCTIONS;
-   ctx->Const.FragmentProgram.MaxTexIndirections = MAX_FRAGMENT_PROGRAM_TEX_INDIRECTIONS;
-   ctx->Const.FragmentProgram.MaxAttribs = MAX_NV_FRAGMENT_PROGRAM_INPUTS;
-   ctx->Const.FragmentProgram.MaxTemps = MAX_PROGRAM_TEMPS;
-   ctx->Const.FragmentProgram.MaxParameters = MAX_NV_FRAGMENT_PROGRAM_PARAMS;
-   ctx->Const.FragmentProgram.MaxLocalParams = MAX_PROGRAM_LOCAL_PARAMS;
-   ctx->Const.FragmentProgram.MaxEnvParams = MAX_PROGRAM_ENV_PARAMS;
-   ctx->Const.FragmentProgram.MaxAddressRegs = MAX_FRAGMENT_PROGRAM_ADDRESS_REGS;
-   ctx->Const.FragmentProgram.MaxUniformComponents = 4 * MAX_UNIFORMS;
-   init_natives(&ctx->Const.FragmentProgram);
+   init_program_limits(GL_FRAGMENT_PROGRAM_ARB, &ctx->Const.FragmentProgram);
 #endif
    ctx->Const.MaxProgramMatrices = MAX_PROGRAM_MATRICES;
    ctx->Const.MaxProgramMatrixStackDepth = MAX_PROGRAM_MATRIX_STACK_DEPTH;
@@ -935,6 +925,9 @@ _mesa_init_constants(GLcontext *ctx)
    ctx->Const.MaxVarying = MAX_VARYING;
 #endif
 
+   /* GL_ARB_framebuffer_object */
+   ctx->Const.MaxSamples = 0;
+
    /* sanity checks */
    ASSERT(ctx->Const.MaxTextureUnits == MIN2(ctx->Const.MaxTextureImageUnits,
                                              ctx->Const.MaxTextureCoordUnits));
@@ -963,6 +956,9 @@ check_context_limits(GLcontext *ctx)
    assert(ctx->Const.MaxTextureUnits <= MAX_TEXTURE_IMAGE_UNITS);
    assert(ctx->Const.MaxTextureUnits <= MAX_TEXTURE_COORD_UNITS);
 
+   /* number of coord units cannot be greater than number of image units */
+   assert(ctx->Const.MaxTextureCoordUnits <= ctx->Const.MaxTextureImageUnits);
+
    assert(ctx->Const.MaxViewportWidth <= MAX_WIDTH);
    assert(ctx->Const.MaxViewportHeight <= MAX_WIDTH);
 
@@ -1063,6 +1059,7 @@ init_attrib_groups(GLcontext *ctx)
    /* Miscellaneous */
    ctx->NewState = _NEW_ALL;
    ctx->ErrorValue = (GLenum) GL_NO_ERROR;
+   ctx->varying_vp_inputs = ~0;
 
    return GL_TRUE;
 }
@@ -1162,7 +1159,7 @@ _mesa_initialize_context(GLcontext *ctx,
                          const struct dd_function_table *driverFunctions,
                          void *driverContext)
 {
-   ASSERT(driverContext);
+   /*ASSERT(driverContext);*/
    assert(driverFunctions->NewTextureObject);
    assert(driverFunctions->FreeTexImageData);
 
@@ -1225,7 +1222,6 @@ _mesa_initialize_context(GLcontext *ctx,
 
    ctx->FragmentProgram._MaintainTexEnvProgram
       = (_mesa_getenv("MESA_TEX_PROG") != NULL);
-   ctx->FragmentProgram._UseTexEnvProgram = ctx->FragmentProgram._MaintainTexEnvProgram;
 
    ctx->VertexProgram._MaintainTnlProgram
       = (_mesa_getenv("MESA_TNL_PROG") != NULL);
@@ -1267,7 +1263,7 @@ _mesa_create_context(const GLvisual *visual,
    GLcontext *ctx;
 
    ASSERT(visual);
-   ASSERT(driverContext);
+   /*ASSERT(driverContext);*/
 
    ctx = (GLcontext *) _mesa_calloc(sizeof(GLcontext));
    if (!ctx)
@@ -1302,10 +1298,10 @@ _mesa_free_context_data( GLcontext *ctx )
    }
 
    /* unreference WinSysDraw/Read buffers */
-   _mesa_unreference_framebuffer(&ctx->WinSysDrawBuffer);
-   _mesa_unreference_framebuffer(&ctx->WinSysReadBuffer);
-   _mesa_unreference_framebuffer(&ctx->DrawBuffer);
-   _mesa_unreference_framebuffer(&ctx->ReadBuffer);
+   _mesa_reference_framebuffer(&ctx->WinSysDrawBuffer, NULL);
+   _mesa_reference_framebuffer(&ctx->WinSysReadBuffer, NULL);
+   _mesa_reference_framebuffer(&ctx->DrawBuffer, NULL);
+   _mesa_reference_framebuffer(&ctx->ReadBuffer, NULL);
 
    _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, NULL);
    _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, NULL);
@@ -1315,7 +1311,9 @@ _mesa_free_context_data( GLcontext *ctx )
    _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL);
    _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram, NULL);
 
+#if FEATURE_attrib_stack
    _mesa_free_attrib_data(ctx);
+#endif
    _mesa_free_lighting_data( ctx );
 #if FEATURE_evaluators
    _mesa_free_eval_data( ctx );
@@ -1632,7 +1630,22 @@ _mesa_make_current( GLcontext *newCtx, GLframebuffer *drawBuffer,
           * or not bound to a user-created FBO.
           */
          if (!newCtx->DrawBuffer || newCtx->DrawBuffer->Name == 0) {
+            /* KW: merge conflict here, revisit. 
+             */
+            /* fix up the fb fields - these will end up wrong otherwise
+             * if the DRIdrawable changes, and everything relies on them.
+             * This is a bit messy (same as needed in _mesa_BindFramebufferEXT)
+             */
+            unsigned int i;
+            GLenum buffers[MAX_DRAW_BUFFERS];
+
             _mesa_reference_framebuffer(&newCtx->DrawBuffer, drawBuffer);
+
+            for(i = 0; i < newCtx->Const.MaxDrawBuffers; i++) {
+               buffers[i] = newCtx->Color.DrawBuffer[i];
+            }
+
+            _mesa_drawbuffers(newCtx, newCtx->Const.MaxDrawBuffers, buffers, NULL);
          }
          if (!newCtx->ReadBuffer || newCtx->ReadBuffer->Name == 0) {
             _mesa_reference_framebuffer(&newCtx->ReadBuffer, readBuffer);
diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h
index ddb38030bf..411b6a7b21 100644
--- a/src/mesa/main/dd.h
+++ b/src/mesa/main/dd.h
@@ -34,7 +34,7 @@
 /* THIS FILE ONLY INCLUDED BY mtypes.h !!!!! */
 
 struct gl_pixelstore_attrib;
-struct mesa_display_list;
+struct gl_display_list;
 
 /**
  * Device driver function table.
@@ -808,6 +808,8 @@ struct dd_function_table {
                          struct gl_renderbuffer_attachment *att);
    void (*FinishRenderTexture)(GLcontext *ctx,
                                struct gl_renderbuffer_attachment *att);
+   void (*ValidateFramebuffer)(GLcontext *ctx,
+                               struct gl_framebuffer *fb);
    /*@}*/
 #endif
 #if FEATURE_EXT_framebuffer_blit
@@ -997,7 +999,7 @@ struct dd_function_table {
     * Notify the T&L component before and after calling a display list.
     */
    void (*BeginCallList)( GLcontext *ctx, 
-			  struct mesa_display_list *dlist );
+			  struct gl_display_list *dlist );
    /**
     * Called by glEndCallList().
     *
diff --git a/src/mesa/main/debug.c b/src/mesa/main/debug.c
index 77fef32558..fcef093ac3 100644
--- a/src/mesa/main/debug.c
+++ b/src/mesa/main/debug.c
@@ -24,9 +24,13 @@
 
 #include "mtypes.h"
 #include "context.h"
+#include "hash.h"
 #include "imports.h"
 #include "debug.h"
 #include "get.h"
+#include "texobj.h"
+#include "texformat.h"
+
 
 /**
  * Primitive names
@@ -164,9 +168,7 @@ static void add_debug_flags( const char *debug )
       { "api",       VERBOSE_API },
       { "list",      VERBOSE_DISPLAY_LIST },
       { "lighting",  VERBOSE_LIGHTING },
-      { "disassem",  VERBOSE_DISASSEM },
-      { "glsl",      VERBOSE_GLSL },     /* report GLSL compile/link errors */
-      { "glsl_dump", VERBOSE_GLSL_DUMP } /* print shader GPU instructions */
+      { "disassem",  VERBOSE_DISASSEM }
    };
    GLuint i;
 
@@ -221,3 +223,98 @@ _mesa_init_debug( GLcontext *ctx )
       add_debug_flags(c);
 }
 
+
+/*
+ * Write ppm file
+ */
+static void
+write_ppm(const char *filename, const GLubyte *buffer, int width, int height,
+          int comps, int rcomp, int gcomp, int bcomp)
+{
+   FILE *f = fopen( filename, "w" );
+   if (f) {
+      int i, x, y;
+      const GLubyte *ptr = buffer;
+      fprintf(f,"P6\n");
+      fprintf(f,"# ppm-file created by osdemo.c\n");
+      fprintf(f,"%i %i\n", width,height);
+      fprintf(f,"255\n");
+      fclose(f);
+      f = fopen( filename, "ab" );  /* reopen in binary append mode */
+      for (y=height-1; y>=0; y--) {
+         for (x=0; x<width; x++) {
+            i = (y*width + x) * comps;
+            fputc(ptr[i+rcomp], f);   /* write red */
+            fputc(ptr[i+gcomp], f); /* write green */
+            fputc(ptr[i+bcomp], f); /* write blue */
+         }
+      }
+      fclose(f);
+   }
+}
+
+
+/**
+ * Write level[0] image to a ppm file.
+ */
+static void
+write_texture_image(struct gl_texture_object *texObj)
+{
+   const struct gl_texture_image *img = texObj->Image[0][0];
+   if (img) {
+      char s[100];
+
+      /* make filename */
+      sprintf(s, "/tmp/teximage%u.ppm", texObj->Name);
+
+      switch (img->TexFormat->MesaFormat) {
+      case MESA_FORMAT_RGBA8888:
+         write_ppm(s, img->Data, img->Width, img->Height, 4, 3, 2, 1);
+         break;
+      case MESA_FORMAT_ARGB8888:
+         write_ppm(s, img->Data, img->Width, img->Height, 4, 2, 1, 0);
+         break;
+      default:
+         printf("XXXX unsupported mesa tex format %d in %s\n",
+                img->TexFormat->MesaFormat, __FUNCTION__);
+      }
+   }
+}
+
+
+static GLboolean DumpImages;
+
+
+static void
+dump_texture_cb(GLuint id, void *data, void *userData)
+{
+   struct gl_texture_object *texObj = (struct gl_texture_object *) data;
+   int i;
+   (void) userData;
+
+   printf("Texture %u\n", texObj->Name);
+   printf("  Target 0x%x\n", texObj->Target);
+   for (i = 0; i < MAX_TEXTURE_LEVELS; i++) {
+      struct gl_texture_image *texImg = texObj->Image[0][i];
+      if (texImg) {
+         printf("  Image %u: %d x %d x %d at %p\n", i,
+                texImg->Width, texImg->Height, texImg->Depth, texImg->Data);
+         if (DumpImages && i == 0) {
+            write_texture_image(texObj);
+         }
+      }
+   }
+}
+
+
+/**
+ * Print basic info about all texture objext to stdout.
+ * If dumpImages is true, write PPM of level[0] image to a file.
+ */
+void
+_mesa_dump_textures(GLboolean dumpImages)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   DumpImages = dumpImages;
+   _mesa_HashDeleteAll(ctx->Shared->TexObjects, dump_texture_cb, ctx);
+}
diff --git a/src/mesa/main/debug.h b/src/mesa/main/debug.h
index 94d99c384b..1862ec75b7 100644
--- a/src/mesa/main/debug.h
+++ b/src/mesa/main/debug.h
@@ -57,4 +57,7 @@ extern void _mesa_init_debug( GLcontext *ctx );
 
 #endif
 
+extern void
+_mesa_dump_textures(GLboolean dumpImages);
+
 #endif
diff --git a/src/mesa/main/depthstencil.c b/src/mesa/main/depthstencil.c
index 9d208e2997..7be2aacaf2 100644
--- a/src/mesa/main/depthstencil.c
+++ b/src/mesa/main/depthstencil.c
@@ -62,15 +62,9 @@ nop_get_pointer(GLcontext *ctx, struct gl_renderbuffer *rb, GLint x, GLint y)
 static void
 delete_wrapper(struct gl_renderbuffer *rb)
 {
-   struct gl_renderbuffer *dsrb = rb->Wrapped;
-   ASSERT(dsrb);
    ASSERT(rb->_ActualFormat == GL_DEPTH_COMPONENT24 ||
           rb->_ActualFormat == GL_STENCIL_INDEX8_EXT);
-   /* decrement refcount on the wrapped buffer and delete it if necessary */
-   dsrb->RefCount--;
-   if (dsrb->RefCount <= 0) {
-      dsrb->Delete(dsrb);
-   }
+   _mesa_reference_renderbuffer(&rb->Wrapped, NULL);
    _mesa_free(rb);
 }
 
diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c
index c7db435506..db1415bc06 100644
--- a/src/mesa/main/dlist.c
+++ b/src/mesa/main/dlist.c
@@ -370,7 +370,7 @@ typedef enum
  * contiguous nodes in memory.
  * Each node is the union of a variety of data types.
  */
-union node
+union gl_dlist_node
 {
    OpCode opcode;
    GLboolean b;
@@ -387,6 +387,9 @@ union node
 };
 
 
+typedef union gl_dlist_node Node;
+
+
 /**
  * How many nodes to allocate at a time.
  *
@@ -414,13 +417,13 @@ void mesa_print_display_list(GLuint list);
  * Make an empty display list.  This is used by glGenLists() to
  * reserve display list IDs.
  */
-static struct mesa_display_list *
-make_list(GLuint list, GLuint count)
+static struct gl_display_list *
+make_list(GLuint name, GLuint count)
 {
-   struct mesa_display_list *dlist = CALLOC_STRUCT(mesa_display_list);
-   dlist->id = list;
-   dlist->node = (Node *) _mesa_malloc(sizeof(Node) * count);
-   dlist->node[0].opcode = OPCODE_END_OF_LIST;
+   struct gl_display_list *dlist = CALLOC_STRUCT(gl_display_list);
+   dlist->Name = name;
+   dlist->Head = (Node *) _mesa_malloc(sizeof(Node) * count);
+   dlist->Head[0].opcode = OPCODE_END_OF_LIST;
    return dlist;
 }
 
@@ -428,10 +431,10 @@ make_list(GLuint list, GLuint count)
 /**
  * Lookup function to just encapsulate casting.
  */
-static INLINE struct mesa_display_list *
+static INLINE struct gl_display_list *
 lookup_list(GLcontext *ctx, GLuint list)
 {
-   return (struct mesa_display_list *)
+   return (struct gl_display_list *)
       _mesa_HashLookup(ctx->Shared->DisplayList, list);
 }
 
@@ -442,12 +445,12 @@ lookup_list(GLcontext *ctx, GLuint list)
  * \param dlist - display list pointer
  */
 void
-_mesa_delete_list(GLcontext *ctx, struct mesa_display_list *dlist)
+_mesa_delete_list(GLcontext *ctx, struct gl_display_list *dlist)
 {
    Node *n, *block;
    GLboolean done;
 
-   n = block = dlist->node;
+   n = block = dlist->Head;
 
    done = block ? GL_FALSE : GL_TRUE;
    while (!done) {
@@ -596,7 +599,7 @@ _mesa_delete_list(GLcontext *ctx, struct mesa_display_list *dlist)
 static void
 destroy_list(GLcontext *ctx, GLuint list)
 {
-   struct mesa_display_list *dlist;
+   struct gl_display_list *dlist;
 
    if (list == 0)
       return;
@@ -2004,7 +2007,7 @@ save_Lightfv(GLenum light, GLenum pname, const GLfloat *params)
    Node *n;
    ASSERT_OUTSIDE_SAVE_BEGIN_END_AND_FLUSH(ctx);
    n = ALLOC_INSTRUCTION(ctx, OPCODE_LIGHT, 6);
-   if (OPCODE_LIGHT) {
+   if (n) {
       GLint i, nParams;
       n[1].e = light;
       n[2].e = pname;
@@ -2521,12 +2524,12 @@ save_MultMatrixd(const GLdouble * m)
 
 
 static void GLAPIENTRY
-save_NewList(GLuint list, GLenum mode)
+save_NewList(GLuint name, GLenum mode)
 {
    GET_CURRENT_CONTEXT(ctx);
    /* It's an error to call this function while building a display list */
    _mesa_error(ctx, GL_INVALID_OPERATION, "glNewList");
-   (void) list;
+   (void) name;
    (void) mode;
 }
 
@@ -5731,7 +5734,7 @@ islist(GLcontext *ctx, GLuint list)
 static void
 execute_list(GLcontext *ctx, GLuint list)
 {
-   struct mesa_display_list *dlist;
+   struct gl_display_list *dlist;
    Node *n;
    GLboolean done;
 
@@ -5752,7 +5755,7 @@ execute_list(GLcontext *ctx, GLuint list)
    if (ctx->Driver.BeginCallList)
       ctx->Driver.BeginCallList(ctx, dlist);
 
-   n = dlist->node;
+   n = dlist->Head;
 
    done = GL_FALSE;
    while (!done) {
@@ -6723,7 +6726,7 @@ _mesa_GenLists(GLsizei range)
  * Begin a new display list.
  */
 void GLAPIENTRY
-_mesa_NewList(GLuint list, GLenum mode)
+_mesa_NewList(GLuint name, GLenum mode)
 {
    GET_CURRENT_CONTEXT(ctx);
    GLint i;
@@ -6732,10 +6735,10 @@ _mesa_NewList(GLuint list, GLenum mode)
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
    if (MESA_VERBOSE & VERBOSE_API)
-      _mesa_debug(ctx, "glNewList %u %s\n", list,
+      _mesa_debug(ctx, "glNewList %u %s\n", name,
                   _mesa_lookup_enum_by_nr(mode));
 
-   if (list == 0) {
+   if (name == 0) {
       _mesa_error(ctx, GL_INVALID_VALUE, "glNewList");
       return;
    }
@@ -6745,7 +6748,7 @@ _mesa_NewList(GLuint list, GLenum mode)
       return;
    }
 
-   if (ctx->ListState.CurrentListPtr) {
+   if (ctx->ListState.CurrentList) {
       /* already compiling a display list */
       _mesa_error(ctx, GL_INVALID_OPERATION, "glNewList");
       return;
@@ -6755,10 +6758,8 @@ _mesa_NewList(GLuint list, GLenum mode)
    ctx->ExecuteFlag = (mode == GL_COMPILE_AND_EXECUTE);
 
    /* Allocate new display list */
-   ctx->ListState.CurrentListNum = list;
-   ctx->ListState.CurrentList = make_list(list, BLOCK_SIZE);
-   ctx->ListState.CurrentBlock = ctx->ListState.CurrentList->node;
-   ctx->ListState.CurrentListPtr = ctx->ListState.CurrentBlock;
+   ctx->ListState.CurrentList = make_list(name, BLOCK_SIZE);
+   ctx->ListState.CurrentBlock = ctx->ListState.CurrentList->Head;
    ctx->ListState.CurrentPos = 0;
 
    /* Reset acumulated list state:
@@ -6770,7 +6771,7 @@ _mesa_NewList(GLuint list, GLenum mode)
       ctx->ListState.ActiveMaterialSize[i] = 0;
 
    ctx->Driver.CurrentSavePrimitive = PRIM_UNKNOWN;
-   ctx->Driver.NewList(ctx, list, mode);
+   ctx->Driver.NewList(ctx, name, mode);
 
    ctx->CurrentDispatch = ctx->Save;
    _glapi_set_dispatch(ctx->CurrentDispatch);
@@ -6791,7 +6792,7 @@ _mesa_EndList(void)
       _mesa_debug(ctx, "glEndList\n");
 
    /* Check that a list is under construction */
-   if (!ctx->ListState.CurrentListPtr) {
+   if (!ctx->ListState.CurrentList) {
       _mesa_error(ctx, GL_INVALID_OPERATION, "glEndList");
       return;
    }
@@ -6804,18 +6805,18 @@ _mesa_EndList(void)
    (void) ALLOC_INSTRUCTION(ctx, OPCODE_END_OF_LIST, 0);
 
    /* Destroy old list, if any */
-   destroy_list(ctx, ctx->ListState.CurrentListNum);
-   /* Install the list */
-   _mesa_HashInsert(ctx->Shared->DisplayList, ctx->ListState.CurrentListNum,
+   destroy_list(ctx, ctx->ListState.CurrentList->Name);
+
+   /* Install the new list */
+   _mesa_HashInsert(ctx->Shared->DisplayList,
+                    ctx->ListState.CurrentList->Name,
                     ctx->ListState.CurrentList);
 
 
    if (MESA_VERBOSE & VERBOSE_DISPLAY_LIST)
-      mesa_print_display_list(ctx->ListState.CurrentListNum);
+      mesa_print_display_list(ctx->ListState.CurrentList->Name);
 
    ctx->ListState.CurrentList = NULL;
-   ctx->ListState.CurrentListNum = 0;
-   ctx->ListState.CurrentListPtr = NULL;
    ctx->ExecuteFlag = GL_TRUE;
    ctx->CompileFlag = GL_FALSE;
 
@@ -8209,7 +8210,7 @@ enum_string(GLenum k)
 static void GLAPIENTRY
 print_list(GLcontext *ctx, GLuint list)
 {
-   struct mesa_display_list *dlist;
+   struct gl_display_list *dlist;
    Node *n;
    GLboolean done;
 
@@ -8222,7 +8223,7 @@ print_list(GLcontext *ctx, GLuint list)
    if (!dlist)
       return;
 
-   n = dlist->node;
+   n = dlist->Head;
 
    _mesa_printf("START-LIST %u, address %p\n", list, (void *) n);
 
@@ -8584,9 +8585,7 @@ _mesa_init_display_list(GLcontext *ctx)
    ctx->ListState.CallDepth = 0;
    ctx->ExecuteFlag = GL_TRUE;
    ctx->CompileFlag = GL_FALSE;
-   ctx->ListState.CurrentListPtr = NULL;
    ctx->ListState.CurrentBlock = NULL;
-   ctx->ListState.CurrentListNum = 0;
    ctx->ListState.CurrentPos = 0;
 
    /* Display List group */
diff --git a/src/mesa/main/dlist.h b/src/mesa/main/dlist.h
index ef6a10af83..ab7ec2c8db 100644
--- a/src/mesa/main/dlist.h
+++ b/src/mesa/main/dlist.h
@@ -39,7 +39,7 @@
 #if _HAVE_FULL_GL
 
 extern void
-_mesa_delete_list(GLcontext *ctx, struct mesa_display_list *dlist);
+_mesa_delete_list(GLcontext *ctx, struct gl_display_list *dlist);
 
 extern void GLAPIENTRY _mesa_CallList( GLuint list );
 
diff --git a/src/mesa/main/dlopen.c b/src/mesa/main/dlopen.c
new file mode 100644
index 0000000000..8bc83c094f
--- /dev/null
+++ b/src/mesa/main/dlopen.c
@@ -0,0 +1,99 @@
+/*
+ * Mesa 3-D graphics library
+ *
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/**
+ * Wrapper functions for dlopen(), dlsym(), dlclose().
+ * Note that the #ifdef tests for various environments should be expanded.
+ */
+
+
+#include "glheader.h"
+#include "imports.h"
+#include "dlopen.h"
+
+#if defined(_GNU_SOURCE) && !defined(__MINGW32__)
+#include <dlfcn.h>
+#endif
+
+
+/**
+ * Wrapper for dlopen().
+ * Note that 'flags' isn't used at this time.
+ */
+void *
+_mesa_dlopen(const char *libname, int flags)
+{
+#if defined(_GNU_SOURCE)
+   flags = RTLD_LAZY | RTLD_GLOBAL; /* Overriding flags at this time */
+   return dlopen(libname, flags);
+#elif defined(__MINGW32__)
+   return LoadLibraryA(libname);
+#else
+   return NULL;
+#endif
+}
+
+
+/**
+ * Wrapper for dlsym() that does a cast to a generic function type,
+ * rather than a void *.  This reduces the number of warnings that are
+ * generated.
+ */
+GenericFunc
+_mesa_dlsym(void *handle, const char *fname)
+{
+#if defined(__DJGPP__)
+   /* need '_' prefix on symbol names */
+   char fname2[1000];
+   fname2[0] = '_';
+   _mesa_strncpy(fname2 + 1, fname, 998);
+   fname2[999] = 0;
+   return (GenericFunc) dlsym(handle, fname2);
+#elif defined(_GNU_SOURCE)
+   return (GenericFunc) dlsym(handle, fname);
+#elif defined(__MINGW32__)
+   return (GenericFunc) GetProcAddress(handle, fname);
+#else
+   return (GenericFunc) NULL;
+#endif
+}
+
+
+/**
+ * Wrapper for dlclose().
+ */
+void
+_mesa_dlclose(void *handle)
+{
+#if defined(_GNU_SOURCE)
+   dlclose(handle);
+#elif defined(__MINGW32__)
+   FreeLibrary(handle);
+#else
+   (void) handle;
+#endif
+}
+
+
+
diff --git a/src/mesa/main/dlopen.h b/src/mesa/main/dlopen.h
new file mode 100644
index 0000000000..9895a22549
--- /dev/null
+++ b/src/mesa/main/dlopen.h
@@ -0,0 +1,42 @@
+/*
+ * Mesa 3-D graphics library
+ *
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#ifndef DLOPEN_H
+#define DLOPEN_H
+
+
+typedef void (*GenericFunc)(void);
+
+
+extern void *
+_mesa_dlopen(const char *libname, int flags);
+
+extern GenericFunc
+_mesa_dlsym(void *handle, const char *fname);
+
+extern void
+_mesa_dlclose(void *handle);
+
+
+#endif
diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c
index ac97bc1ff9..13cfa0e756 100644
--- a/src/mesa/main/drawpix.c
+++ b/src/mesa/main/drawpix.c
@@ -153,7 +153,7 @@ _mesa_CopyPixels( GLint srcx, GLint srcy, GLsizei width, GLsizei height,
       return;
    }
 
-   if (!ctx->Current.RasterPosValid) {
+   if (!ctx->Current.RasterPosValid || width ==0 || height == 0) {
       return;
    }
 
@@ -218,7 +218,7 @@ _mesa_Bitmap( GLsizei width, GLsizei height,
 
    if (ctx->RenderMode == GL_RENDER) {
       /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */
-      const GLfloat epsilon = (const GLfloat)0.0001;
+      const GLfloat epsilon = 0.0001F;
       GLint x = IFLOOR(ctx->Current.RasterPos[0] + epsilon - xorig);
       GLint y = IFLOOR(ctx->Current.RasterPos[1] + epsilon - yorig);
 
diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c
index e0483a72e8..2b54fac694 100644
--- a/src/mesa/main/enable.c
+++ b/src/mesa/main/enable.c
@@ -201,6 +201,26 @@ _mesa_DisableClientState( GLenum cap )
    }
 
 
+
+/**
+ * Return pointer to current texture unit for setting/getting coordinate
+ * state.
+ * Note that we'll set GL_INVALID_OPERATION if the active texture unit is
+ * higher than the number of supported coordinate units.  And we'll return NULL.
+ */
+static struct gl_texture_unit *
+get_texcoord_unit(GLcontext *ctx)
+{
+   if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureCoordUnits) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glEnable/Disable(texcoord unit)");
+      return NULL;
+   }
+   else {
+      return &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   }
+}
+
+
 /**
  * Helper function to enable or disable a texture target.
  */
@@ -321,10 +341,6 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state)
          ctx->Transform.CullVertexFlag = state;
          break;
       case GL_DEPTH_TEST:
-         if (state && ctx->DrawBuffer->Visual.depthBits == 0) {
-            _mesa_warning(ctx,"glEnable(GL_DEPTH_TEST) but no depth buffer");
-            return;
-         }
          if (ctx->Depth.Test == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_DEPTH);
@@ -612,54 +628,62 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state)
             return;
          }
          break;
-      case GL_TEXTURE_GEN_Q: {
-         GLuint unit = ctx->Texture.CurrentUnit;
-         struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-         GLuint newenabled = texUnit->TexGenEnabled & ~Q_BIT;
-         if (state)
-            newenabled |= Q_BIT;
-         if (texUnit->TexGenEnabled == newenabled)
-            return;
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         texUnit->TexGenEnabled = newenabled;
+      case GL_TEXTURE_GEN_Q:
+         {
+            struct gl_texture_unit *texUnit = get_texcoord_unit(ctx);
+            if (texUnit) {
+               GLuint newenabled = texUnit->TexGenEnabled & ~Q_BIT;
+               if (state)
+                  newenabled |= Q_BIT;
+               if (texUnit->TexGenEnabled == newenabled)
+                  return;
+               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+               texUnit->TexGenEnabled = newenabled;
+            }
+         }
          break;
-      }
-      case GL_TEXTURE_GEN_R: {
-         GLuint unit = ctx->Texture.CurrentUnit;
-         struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-         GLuint newenabled = texUnit->TexGenEnabled & ~R_BIT;
-         if (state)
-            newenabled |= R_BIT;
-         if (texUnit->TexGenEnabled == newenabled)
-            return;
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         texUnit->TexGenEnabled = newenabled;
+      case GL_TEXTURE_GEN_R:
+         {
+            struct gl_texture_unit *texUnit = get_texcoord_unit(ctx);
+            if (texUnit) {
+               GLuint newenabled = texUnit->TexGenEnabled & ~R_BIT;
+               if (state)
+                  newenabled |= R_BIT;
+               if (texUnit->TexGenEnabled == newenabled)
+                  return;
+               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+               texUnit->TexGenEnabled = newenabled;
+            }
+         }
          break;
-      }
-      case GL_TEXTURE_GEN_S: {
-         GLuint unit = ctx->Texture.CurrentUnit;
-         struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-         GLuint newenabled = texUnit->TexGenEnabled & ~S_BIT;
-         if (state)
-            newenabled |= S_BIT;
-         if (texUnit->TexGenEnabled == newenabled)
-            return;
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         texUnit->TexGenEnabled = newenabled;
+      case GL_TEXTURE_GEN_S:
+         {
+            struct gl_texture_unit *texUnit = get_texcoord_unit(ctx);
+            if (texUnit) {
+               GLuint newenabled = texUnit->TexGenEnabled & ~S_BIT;
+               if (state)
+                  newenabled |= S_BIT;
+               if (texUnit->TexGenEnabled == newenabled)
+                  return;
+               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+               texUnit->TexGenEnabled = newenabled;
+            }
+         }
          break;
-      }
-      case GL_TEXTURE_GEN_T: {
-         GLuint unit = ctx->Texture.CurrentUnit;
-         struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-         GLuint newenabled = texUnit->TexGenEnabled & ~T_BIT;
-         if (state)
-            newenabled |= T_BIT;
-         if (texUnit->TexGenEnabled == newenabled)
-            return;
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         texUnit->TexGenEnabled = newenabled;
+      case GL_TEXTURE_GEN_T:
+         {
+            struct gl_texture_unit *texUnit = get_texcoord_unit(ctx);
+            if (texUnit) {
+               GLuint newenabled = texUnit->TexGenEnabled & ~T_BIT;
+               if (state)
+                  newenabled |= T_BIT;
+               if (texUnit->TexGenEnabled == newenabled)
+                  return;
+               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+               texUnit->TexGenEnabled = newenabled;
+            }
+         }
          break;
-      }
 
       /*
        * CLIENT STATE!!!
@@ -748,35 +772,30 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state)
 
       /* GL_ARB_multisample */
       case GL_MULTISAMPLE_ARB:
-         CHECK_EXTENSION(ARB_multisample, cap);
          if (ctx->Multisample.Enabled == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_MULTISAMPLE);
          ctx->Multisample.Enabled = state;
          break;
       case GL_SAMPLE_ALPHA_TO_COVERAGE_ARB:
-         CHECK_EXTENSION(ARB_multisample, cap);
          if (ctx->Multisample.SampleAlphaToCoverage == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_MULTISAMPLE);
          ctx->Multisample.SampleAlphaToCoverage = state;
          break;
       case GL_SAMPLE_ALPHA_TO_ONE_ARB:
-         CHECK_EXTENSION(ARB_multisample, cap);
          if (ctx->Multisample.SampleAlphaToOne == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_MULTISAMPLE);
          ctx->Multisample.SampleAlphaToOne = state;
          break;
       case GL_SAMPLE_COVERAGE_ARB:
-         CHECK_EXTENSION(ARB_multisample, cap);
          if (ctx->Multisample.SampleCoverage == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_MULTISAMPLE);
          ctx->Multisample.SampleCoverage = state;
          break;
       case GL_SAMPLE_COVERAGE_INVERT_ARB:
-         CHECK_EXTENSION(ARB_multisample, cap);
          if (ctx->Multisample.SampleCoverageInvert == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_MULTISAMPLE);
@@ -898,10 +917,13 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state)
             return;
          FLUSH_VERTICES(ctx, _NEW_STENCIL);
          ctx->Stencil.TestTwoSide = state;
-         if (state)
+         if (state) {
+            ctx->Stencil._BackFace = 2;
             ctx->_TriangleCaps |= DD_TRI_TWOSTENCIL;
-         else
+         } else {
+            ctx->Stencil._BackFace = 1;
             ctx->_TriangleCaps &= ~DD_TRI_TWOSTENCIL;
+         }
          break;
 
 #if FEATURE_ARB_fragment_program
@@ -929,6 +951,7 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state)
          break;
 
       /* GL_MESA_program_debug */
+#if FEATURE_MESA_program_debug
       case GL_FRAGMENT_PROGRAM_CALLBACK_MESA:
          CHECK_EXTENSION(MESA_program_debug, cap);
          ctx->FragmentProgram.CallbackEnabled = state;
@@ -937,6 +960,7 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state)
          CHECK_EXTENSION(MESA_program_debug, cap);
          ctx->VertexProgram.CallbackEnabled = state;
          break;
+#endif
 
 #if FEATURE_ATI_fragment_shader
       case GL_FRAGMENT_SHADER_ATI:
@@ -1153,28 +1177,36 @@ _mesa_IsEnabled( GLenum cap )
          return is_texture_enabled(ctx, TEXTURE_3D_BIT);
       case GL_TEXTURE_GEN_Q:
          {
-            const struct gl_texture_unit *texUnit;
-            texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
-            return (texUnit->TexGenEnabled & Q_BIT) ? GL_TRUE : GL_FALSE;
+            const struct gl_texture_unit *texUnit = get_texcoord_unit(ctx);
+            if (texUnit) {
+               return (texUnit->TexGenEnabled & Q_BIT) ? GL_TRUE : GL_FALSE;
+            }
          }
+         return GL_FALSE;
       case GL_TEXTURE_GEN_R:
          {
-            const struct gl_texture_unit *texUnit;
-            texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
-            return (texUnit->TexGenEnabled & R_BIT) ? GL_TRUE : GL_FALSE;
+            const struct gl_texture_unit *texUnit = get_texcoord_unit(ctx);
+            if (texUnit) {
+               return (texUnit->TexGenEnabled & R_BIT) ? GL_TRUE : GL_FALSE;
+            }
          }
+         return GL_FALSE;
       case GL_TEXTURE_GEN_S:
          {
-            const struct gl_texture_unit *texUnit;
-            texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
-            return (texUnit->TexGenEnabled & S_BIT) ? GL_TRUE : GL_FALSE;
+            const struct gl_texture_unit *texUnit = get_texcoord_unit(ctx);
+            if (texUnit) {
+               return (texUnit->TexGenEnabled & S_BIT) ? GL_TRUE : GL_FALSE;
+            }
          }
+         return GL_FALSE;
       case GL_TEXTURE_GEN_T:
          {
-            const struct gl_texture_unit *texUnit;
-            texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
-            return (texUnit->TexGenEnabled & T_BIT) ? GL_TRUE : GL_FALSE;
+            const struct gl_texture_unit *texUnit = get_texcoord_unit(ctx);
+            if (texUnit) {
+               return (texUnit->TexGenEnabled & T_BIT) ? GL_TRUE : GL_FALSE;
+            }
          }
+         return GL_FALSE;
 
       /*
        * CLIENT STATE!!!
@@ -1249,19 +1281,14 @@ _mesa_IsEnabled( GLenum cap )
 
       /* GL_ARB_multisample */
       case GL_MULTISAMPLE_ARB:
-         CHECK_EXTENSION(ARB_multisample);
          return ctx->Multisample.Enabled;
       case GL_SAMPLE_ALPHA_TO_COVERAGE_ARB:
-         CHECK_EXTENSION(ARB_multisample);
          return ctx->Multisample.SampleAlphaToCoverage;
       case GL_SAMPLE_ALPHA_TO_ONE_ARB:
-         CHECK_EXTENSION(ARB_multisample);
          return ctx->Multisample.SampleAlphaToOne;
       case GL_SAMPLE_COVERAGE_ARB:
-         CHECK_EXTENSION(ARB_multisample);
          return ctx->Multisample.SampleCoverage;
       case GL_SAMPLE_COVERAGE_INVERT_ARB:
-         CHECK_EXTENSION(ARB_multisample);
          return ctx->Multisample.SampleCoverageInvert;
 
       /* GL_IBM_rasterpos_clip */
@@ -1378,12 +1405,15 @@ _mesa_IsEnabled( GLenum cap )
          return ctx->Depth.BoundsTest;
 
       /* GL_MESA_program_debug */
+#if FEATURE_MESA_program_debug
       case GL_FRAGMENT_PROGRAM_CALLBACK_MESA:
          CHECK_EXTENSION(MESA_program_debug);
          return ctx->FragmentProgram.CallbackEnabled;
       case GL_VERTEX_PROGRAM_CALLBACK_MESA:
          CHECK_EXTENSION(MESA_program_debug);
          return ctx->VertexProgram.CallbackEnabled;
+#endif
+
 #if FEATURE_ATI_fragment_shader
       case GL_FRAGMENT_SHADER_ATI:
 	 CHECK_EXTENSION(ATI_fragment_shader);
diff --git a/src/mesa/main/enums.c b/src/mesa/main/enums.c
index 4796f3027a..a864f5a070 100644
--- a/src/mesa/main/enums.c
+++ b/src/mesa/main/enums.c
@@ -180,21 +180,37 @@ LONGSTRING static const char enum_string_table[] =
    "GL_COLOR_ARRAY_SIZE\0"
    "GL_COLOR_ARRAY_STRIDE\0"
    "GL_COLOR_ARRAY_TYPE\0"
+   "GL_COLOR_ATTACHMENT0\0"
    "GL_COLOR_ATTACHMENT0_EXT\0"
+   "GL_COLOR_ATTACHMENT1\0"
+   "GL_COLOR_ATTACHMENT10\0"
    "GL_COLOR_ATTACHMENT10_EXT\0"
+   "GL_COLOR_ATTACHMENT11\0"
    "GL_COLOR_ATTACHMENT11_EXT\0"
+   "GL_COLOR_ATTACHMENT12\0"
    "GL_COLOR_ATTACHMENT12_EXT\0"
+   "GL_COLOR_ATTACHMENT13\0"
    "GL_COLOR_ATTACHMENT13_EXT\0"
+   "GL_COLOR_ATTACHMENT14\0"
    "GL_COLOR_ATTACHMENT14_EXT\0"
+   "GL_COLOR_ATTACHMENT15\0"
    "GL_COLOR_ATTACHMENT15_EXT\0"
    "GL_COLOR_ATTACHMENT1_EXT\0"
+   "GL_COLOR_ATTACHMENT2\0"
    "GL_COLOR_ATTACHMENT2_EXT\0"
+   "GL_COLOR_ATTACHMENT3\0"
    "GL_COLOR_ATTACHMENT3_EXT\0"
+   "GL_COLOR_ATTACHMENT4\0"
    "GL_COLOR_ATTACHMENT4_EXT\0"
+   "GL_COLOR_ATTACHMENT5\0"
    "GL_COLOR_ATTACHMENT5_EXT\0"
+   "GL_COLOR_ATTACHMENT6\0"
    "GL_COLOR_ATTACHMENT6_EXT\0"
+   "GL_COLOR_ATTACHMENT7\0"
    "GL_COLOR_ATTACHMENT7_EXT\0"
+   "GL_COLOR_ATTACHMENT8\0"
    "GL_COLOR_ATTACHMENT8_EXT\0"
+   "GL_COLOR_ATTACHMENT9\0"
    "GL_COLOR_ATTACHMENT9_EXT\0"
    "GL_COLOR_BUFFER_BIT\0"
    "GL_COLOR_CLEAR_VALUE\0"
@@ -348,6 +364,8 @@ LONGSTRING static const char enum_string_table[] =
    "GL_DECR_WRAP_EXT\0"
    "GL_DELETE_STATUS\0"
    "GL_DEPTH\0"
+   "GL_DEPTH24_STENCIL8\0"
+   "GL_DEPTH_ATTACHMENT\0"
    "GL_DEPTH_ATTACHMENT_EXT\0"
    "GL_DEPTH_BIAS\0"
    "GL_DEPTH_BITS\0"
@@ -369,6 +387,8 @@ LONGSTRING static const char enum_string_table[] =
    "GL_DEPTH_FUNC\0"
    "GL_DEPTH_RANGE\0"
    "GL_DEPTH_SCALE\0"
+   "GL_DEPTH_STENCIL\0"
+   "GL_DEPTH_STENCIL_ATTACHMENT\0"
    "GL_DEPTH_STENCIL_NV\0"
    "GL_DEPTH_STENCIL_TO_BGRA_NV\0"
    "GL_DEPTH_STENCIL_TO_RGBA_NV\0"
@@ -437,6 +457,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_DRAW_BUFFER9\0"
    "GL_DRAW_BUFFER9_ARB\0"
    "GL_DRAW_BUFFER9_ATI\0"
+   "GL_DRAW_FRAMEBUFFER\0"
    "GL_DRAW_FRAMEBUFFER_BINDING_EXT\0"
    "GL_DRAW_FRAMEBUFFER_EXT\0"
    "GL_DRAW_PIXEL_TOKEN\0"
@@ -522,23 +543,44 @@ LONGSTRING static const char enum_string_table[] =
    "GL_FRAGMENT_SHADER\0"
    "GL_FRAGMENT_SHADER_ARB\0"
    "GL_FRAGMENT_SHADER_DERIVATIVE_HINT\0"
+   "GL_FRAMEBUFFER\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME\0"
    "GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME_EXT\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE\0"
    "GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE_EXT\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE\0"
    "GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_3D_ZOFFSET_EXT\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE\0"
    "GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE_EXT\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER\0"
    "GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER_EXT\0"
+   "GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL\0"
    "GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL_EXT\0"
    "GL_FRAMEBUFFER_BINDING_EXT\0"
+   "GL_FRAMEBUFFER_COMPLETE\0"
    "GL_FRAMEBUFFER_COMPLETE_EXT\0"
+   "GL_FRAMEBUFFER_DEFAULT\0"
    "GL_FRAMEBUFFER_EXT\0"
+   "GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT\0"
    "GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT_EXT\0"
    "GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT\0"
    "GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT\0"
    "GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT\0"
    "GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT\0"
+   "GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT\0"
    "GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT_EXT\0"
+   "GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE\0"
    "GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT\0"
    "GL_FRAMEBUFFER_STATUS_ERROR_EXT\0"
+   "GL_FRAMEBUFFER_UNDEFINED\0"
+   "GL_FRAMEBUFFER_UNSUPPORTED\0"
    "GL_FRAMEBUFFER_UNSUPPORTED_EXT\0"
    "GL_FRONT\0"
    "GL_FRONT_AND_BACK\0"
@@ -611,6 +653,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_INCR\0"
    "GL_INCR_WRAP\0"
    "GL_INCR_WRAP_EXT\0"
+   "GL_INDEX\0"
    "GL_INDEX_ARRAY\0"
    "GL_INDEX_ARRAY_BUFFER_BINDING\0"
    "GL_INDEX_ARRAY_BUFFER_BINDING_ARB\0"
@@ -646,6 +689,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_INT_VEC4\0"
    "GL_INT_VEC4_ARB\0"
    "GL_INVALID_ENUM\0"
+   "GL_INVALID_FRAMEBUFFER_OPERATION\0"
    "GL_INVALID_FRAMEBUFFER_OPERATION_EXT\0"
    "GL_INVALID_OPERATION\0"
    "GL_INVALID_VALUE\0"
@@ -890,6 +934,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB\0"
    "GL_MAX_RECTANGLE_TEXTURE_SIZE_NV\0"
    "GL_MAX_RENDERBUFFER_SIZE_EXT\0"
+   "GL_MAX_SAMPLES\0"
    "GL_MAX_SHININESS_NV\0"
    "GL_MAX_SPOT_EXPONENT_NV\0"
    "GL_MAX_TEXTURE_COORDS\0"
@@ -1257,6 +1302,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_R3_G3_B2\0"
    "GL_RASTER_POSITION_UNCLIPPED_IBM\0"
    "GL_READ_BUFFER\0"
+   "GL_READ_FRAMEBUFFER\0"
    "GL_READ_FRAMEBUFFER_BINDING_EXT\0"
    "GL_READ_FRAMEBUFFER_EXT\0"
    "GL_READ_ONLY\0"
@@ -1273,10 +1319,21 @@ LONGSTRING static const char enum_string_table[] =
    "GL_REFLECTION_MAP_ARB\0"
    "GL_REFLECTION_MAP_NV\0"
    "GL_RENDER\0"
+   "GL_RENDERBUFFER\0"
+   "GL_RENDERBUFFER_ALPHA_SIZE\0"
    "GL_RENDERBUFFER_BINDING_EXT\0"
+   "GL_RENDERBUFFER_BLUE_SIZE\0"
+   "GL_RENDERBUFFER_DEPTH_SIZE\0"
    "GL_RENDERBUFFER_EXT\0"
+   "GL_RENDERBUFFER_GREEN_SIZE\0"
+   "GL_RENDERBUFFER_HEIGHT\0"
    "GL_RENDERBUFFER_HEIGHT_EXT\0"
+   "GL_RENDERBUFFER_INTERNAL_FORMAT\0"
    "GL_RENDERBUFFER_INTERNAL_FORMAT_EXT\0"
+   "GL_RENDERBUFFER_RED_SIZE\0"
+   "GL_RENDERBUFFER_SAMPLES\0"
+   "GL_RENDERBUFFER_STENCIL_SIZE\0"
+   "GL_RENDERBUFFER_WIDTH\0"
    "GL_RENDERBUFFER_WIDTH_EXT\0"
    "GL_RENDERER\0"
    "GL_RENDER_MODE\0"
@@ -1420,6 +1477,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_SRC_ALPHA\0"
    "GL_SRC_ALPHA_SATURATE\0"
    "GL_SRC_COLOR\0"
+   "GL_SRGB\0"
    "GL_STACK_OVERFLOW\0"
    "GL_STACK_UNDERFLOW\0"
    "GL_STATIC_COPY\0"
@@ -1429,6 +1487,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_STATIC_READ\0"
    "GL_STATIC_READ_ARB\0"
    "GL_STENCIL\0"
+   "GL_STENCIL_ATTACHMENT\0"
    "GL_STENCIL_ATTACHMENT_EXT\0"
    "GL_STENCIL_BACK_FAIL\0"
    "GL_STENCIL_BACK_FAIL_ATI\0"
@@ -1654,6 +1713,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_TEXTURE_RED_SIZE_EXT\0"
    "GL_TEXTURE_RESIDENT\0"
    "GL_TEXTURE_STACK_DEPTH\0"
+   "GL_TEXTURE_STENCIL_SIZE\0"
    "GL_TEXTURE_TOO_LARGE_EXT\0"
    "GL_TEXTURE_UNSIGNED_REMAP_MODE_NV\0"
    "GL_TEXTURE_WIDTH\0"
@@ -1692,10 +1752,12 @@ LONGSTRING static const char enum_string_table[] =
    "GL_UNSIGNED_BYTE_3_3_2\0"
    "GL_UNSIGNED_INT\0"
    "GL_UNSIGNED_INT_10_10_10_2\0"
+   "GL_UNSIGNED_INT_24_8\0"
    "GL_UNSIGNED_INT_24_8_NV\0"
    "GL_UNSIGNED_INT_2_10_10_10_REV\0"
    "GL_UNSIGNED_INT_8_8_8_8\0"
    "GL_UNSIGNED_INT_8_8_8_8_REV\0"
+   "GL_UNSIGNED_NORMALIZED\0"
    "GL_UNSIGNED_SHORT\0"
    "GL_UNSIGNED_SHORT_1_5_5_5_REV\0"
    "GL_UNSIGNED_SHORT_4_4_4_4\0"
@@ -1785,7 +1847,7 @@ LONGSTRING static const char enum_string_table[] =
    "GL_ZOOM_Y\0"
    ;
 
-static const enum_elt all_enums[1748] =
+static const enum_elt all_enums[1810] =
 {
    {     0, 0x00000600 }, /* GL_2D */
    {     6, 0x00001407 }, /* GL_2_BYTES */
@@ -1932,2896 +1994,2982 @@ static const enum_elt all_enums[1748] =
    {  2414, 0x00008081 }, /* GL_COLOR_ARRAY_SIZE */
    {  2434, 0x00008083 }, /* GL_COLOR_ARRAY_STRIDE */
    {  2456, 0x00008082 }, /* GL_COLOR_ARRAY_TYPE */
-   {  2476, 0x00008CE0 }, /* GL_COLOR_ATTACHMENT0_EXT */
-   {  2501, 0x00008CEA }, /* GL_COLOR_ATTACHMENT10_EXT */
-   {  2527, 0x00008CEB }, /* GL_COLOR_ATTACHMENT11_EXT */
-   {  2553, 0x00008CEC }, /* GL_COLOR_ATTACHMENT12_EXT */
-   {  2579, 0x00008CED }, /* GL_COLOR_ATTACHMENT13_EXT */
-   {  2605, 0x00008CEE }, /* GL_COLOR_ATTACHMENT14_EXT */
-   {  2631, 0x00008CEF }, /* GL_COLOR_ATTACHMENT15_EXT */
-   {  2657, 0x00008CE1 }, /* GL_COLOR_ATTACHMENT1_EXT */
-   {  2682, 0x00008CE2 }, /* GL_COLOR_ATTACHMENT2_EXT */
-   {  2707, 0x00008CE3 }, /* GL_COLOR_ATTACHMENT3_EXT */
-   {  2732, 0x00008CE4 }, /* GL_COLOR_ATTACHMENT4_EXT */
-   {  2757, 0x00008CE5 }, /* GL_COLOR_ATTACHMENT5_EXT */
-   {  2782, 0x00008CE6 }, /* GL_COLOR_ATTACHMENT6_EXT */
-   {  2807, 0x00008CE7 }, /* GL_COLOR_ATTACHMENT7_EXT */
-   {  2832, 0x00008CE8 }, /* GL_COLOR_ATTACHMENT8_EXT */
-   {  2857, 0x00008CE9 }, /* GL_COLOR_ATTACHMENT9_EXT */
-   {  2882, 0x00004000 }, /* GL_COLOR_BUFFER_BIT */
-   {  2902, 0x00000C22 }, /* GL_COLOR_CLEAR_VALUE */
-   {  2923, 0x00001900 }, /* GL_COLOR_INDEX */
-   {  2938, 0x00001603 }, /* GL_COLOR_INDEXES */
-   {  2955, 0x00000BF2 }, /* GL_COLOR_LOGIC_OP */
-   {  2973, 0x00000B57 }, /* GL_COLOR_MATERIAL */
-   {  2991, 0x00000B55 }, /* GL_COLOR_MATERIAL_FACE */
-   {  3014, 0x00000B56 }, /* GL_COLOR_MATERIAL_PARAMETER */
-   {  3042, 0x000080B1 }, /* GL_COLOR_MATRIX */
-   {  3058, 0x000080B1 }, /* GL_COLOR_MATRIX_SGI */
-   {  3078, 0x000080B2 }, /* GL_COLOR_MATRIX_STACK_DEPTH */
-   {  3106, 0x000080B2 }, /* GL_COLOR_MATRIX_STACK_DEPTH_SGI */
-   {  3138, 0x00008458 }, /* GL_COLOR_SUM */
-   {  3151, 0x00008458 }, /* GL_COLOR_SUM_ARB */
-   {  3168, 0x000080D0 }, /* GL_COLOR_TABLE */
-   {  3183, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE */
-   {  3209, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE_EXT */
-   {  3239, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE_SGI */
-   {  3269, 0x000080D7 }, /* GL_COLOR_TABLE_BIAS */
-   {  3289, 0x000080D7 }, /* GL_COLOR_TABLE_BIAS_SGI */
-   {  3313, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE */
-   {  3338, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE_EXT */
-   {  3367, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE_SGI */
-   {  3396, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT */
-   {  3418, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT_EXT */
-   {  3444, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT_SGI */
-   {  3470, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE */
-   {  3496, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE_EXT */
-   {  3526, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE_SGI */
-   {  3556, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE */
-   {  3586, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE_EXT */
-   {  3620, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE_SGI */
-   {  3654, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE */
-   {  3684, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE_EXT */
-   {  3718, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE_SGI */
-   {  3752, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE */
-   {  3776, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE_EXT */
-   {  3804, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE_SGI */
-   {  3832, 0x000080D6 }, /* GL_COLOR_TABLE_SCALE */
-   {  3853, 0x000080D6 }, /* GL_COLOR_TABLE_SCALE_SGI */
-   {  3878, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH */
-   {  3899, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH_EXT */
-   {  3924, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH_SGI */
-   {  3949, 0x00000C23 }, /* GL_COLOR_WRITEMASK */
-   {  3968, 0x00008570 }, /* GL_COMBINE */
-   {  3979, 0x00008503 }, /* GL_COMBINE4 */
-   {  3991, 0x00008572 }, /* GL_COMBINE_ALPHA */
-   {  4008, 0x00008572 }, /* GL_COMBINE_ALPHA_ARB */
-   {  4029, 0x00008572 }, /* GL_COMBINE_ALPHA_EXT */
-   {  4050, 0x00008570 }, /* GL_COMBINE_ARB */
-   {  4065, 0x00008570 }, /* GL_COMBINE_EXT */
-   {  4080, 0x00008571 }, /* GL_COMBINE_RGB */
-   {  4095, 0x00008571 }, /* GL_COMBINE_RGB_ARB */
-   {  4114, 0x00008571 }, /* GL_COMBINE_RGB_EXT */
-   {  4133, 0x0000884E }, /* GL_COMPARE_REF_DEPTH_TO_TEXTURE_EXT */
-   {  4169, 0x0000884E }, /* GL_COMPARE_R_TO_TEXTURE */
-   {  4193, 0x0000884E }, /* GL_COMPARE_R_TO_TEXTURE_ARB */
-   {  4221, 0x00001300 }, /* GL_COMPILE */
-   {  4232, 0x00001301 }, /* GL_COMPILE_AND_EXECUTE */
-   {  4255, 0x00008B81 }, /* GL_COMPILE_STATUS */
-   {  4273, 0x000084E9 }, /* GL_COMPRESSED_ALPHA */
-   {  4293, 0x000084E9 }, /* GL_COMPRESSED_ALPHA_ARB */
-   {  4317, 0x000084EC }, /* GL_COMPRESSED_INTENSITY */
-   {  4341, 0x000084EC }, /* GL_COMPRESSED_INTENSITY_ARB */
-   {  4369, 0x000084EA }, /* GL_COMPRESSED_LUMINANCE */
-   {  4393, 0x000084EB }, /* GL_COMPRESSED_LUMINANCE_ALPHA */
-   {  4423, 0x000084EB }, /* GL_COMPRESSED_LUMINANCE_ALPHA_ARB */
-   {  4457, 0x000084EA }, /* GL_COMPRESSED_LUMINANCE_ARB */
-   {  4485, 0x000084ED }, /* GL_COMPRESSED_RGB */
-   {  4503, 0x000084EE }, /* GL_COMPRESSED_RGBA */
-   {  4522, 0x000084EE }, /* GL_COMPRESSED_RGBA_ARB */
-   {  4545, 0x000086B1 }, /* GL_COMPRESSED_RGBA_FXT1_3DFX */
-   {  4574, 0x000083F1 }, /* GL_COMPRESSED_RGBA_S3TC_DXT1_EXT */
-   {  4607, 0x000083F2 }, /* GL_COMPRESSED_RGBA_S3TC_DXT3_EXT */
-   {  4640, 0x000083F3 }, /* GL_COMPRESSED_RGBA_S3TC_DXT5_EXT */
-   {  4673, 0x000084ED }, /* GL_COMPRESSED_RGB_ARB */
-   {  4695, 0x000086B0 }, /* GL_COMPRESSED_RGB_FXT1_3DFX */
-   {  4723, 0x000083F0 }, /* GL_COMPRESSED_RGB_S3TC_DXT1_EXT */
-   {  4755, 0x000086A3 }, /* GL_COMPRESSED_TEXTURE_FORMATS */
-   {  4785, 0x00008576 }, /* GL_CONSTANT */
-   {  4797, 0x00008003 }, /* GL_CONSTANT_ALPHA */
-   {  4815, 0x00008003 }, /* GL_CONSTANT_ALPHA_EXT */
-   {  4837, 0x00008576 }, /* GL_CONSTANT_ARB */
-   {  4853, 0x00001207 }, /* GL_CONSTANT_ATTENUATION */
-   {  4877, 0x00008151 }, /* GL_CONSTANT_BORDER_HP */
-   {  4899, 0x00008001 }, /* GL_CONSTANT_COLOR */
-   {  4917, 0x00008001 }, /* GL_CONSTANT_COLOR_EXT */
-   {  4939, 0x00008576 }, /* GL_CONSTANT_EXT */
-   {  4955, 0x00008010 }, /* GL_CONVOLUTION_1D */
-   {  4973, 0x00008011 }, /* GL_CONVOLUTION_2D */
-   {  4991, 0x00008154 }, /* GL_CONVOLUTION_BORDER_COLOR */
-   {  5019, 0x00008154 }, /* GL_CONVOLUTION_BORDER_COLOR_HP */
-   {  5050, 0x00008013 }, /* GL_CONVOLUTION_BORDER_MODE */
-   {  5077, 0x00008013 }, /* GL_CONVOLUTION_BORDER_MODE_EXT */
-   {  5108, 0x00008015 }, /* GL_CONVOLUTION_FILTER_BIAS */
-   {  5135, 0x00008015 }, /* GL_CONVOLUTION_FILTER_BIAS_EXT */
-   {  5166, 0x00008014 }, /* GL_CONVOLUTION_FILTER_SCALE */
-   {  5194, 0x00008014 }, /* GL_CONVOLUTION_FILTER_SCALE_EXT */
-   {  5226, 0x00008017 }, /* GL_CONVOLUTION_FORMAT */
-   {  5248, 0x00008017 }, /* GL_CONVOLUTION_FORMAT_EXT */
-   {  5274, 0x00008019 }, /* GL_CONVOLUTION_HEIGHT */
-   {  5296, 0x00008019 }, /* GL_CONVOLUTION_HEIGHT_EXT */
-   {  5322, 0x00008018 }, /* GL_CONVOLUTION_WIDTH */
-   {  5343, 0x00008018 }, /* GL_CONVOLUTION_WIDTH_EXT */
-   {  5368, 0x00008862 }, /* GL_COORD_REPLACE */
-   {  5385, 0x00008862 }, /* GL_COORD_REPLACE_ARB */
-   {  5406, 0x00008862 }, /* GL_COORD_REPLACE_NV */
-   {  5426, 0x00001503 }, /* GL_COPY */
-   {  5434, 0x0000150C }, /* GL_COPY_INVERTED */
-   {  5451, 0x00000706 }, /* GL_COPY_PIXEL_TOKEN */
-   {  5471, 0x00000B44 }, /* GL_CULL_FACE */
-   {  5484, 0x00000B45 }, /* GL_CULL_FACE_MODE */
-   {  5502, 0x000081AA }, /* GL_CULL_VERTEX_EXT */
-   {  5521, 0x000081AC }, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
-   {  5553, 0x000081AB }, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
-   {  5588, 0x00008626 }, /* GL_CURRENT_ATTRIB_NV */
-   {  5609, 0x00000001 }, /* GL_CURRENT_BIT */
-   {  5624, 0x00000B00 }, /* GL_CURRENT_COLOR */
-   {  5641, 0x00008453 }, /* GL_CURRENT_FOG_COORD */
-   {  5662, 0x00008453 }, /* GL_CURRENT_FOG_COORDINATE */
-   {  5688, 0x00000B01 }, /* GL_CURRENT_INDEX */
-   {  5705, 0x00008641 }, /* GL_CURRENT_MATRIX_ARB */
-   {  5727, 0x00008845 }, /* GL_CURRENT_MATRIX_INDEX_ARB */
-   {  5755, 0x00008641 }, /* GL_CURRENT_MATRIX_NV */
-   {  5776, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
-   {  5810, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_NV */
-   {  5843, 0x00000B02 }, /* GL_CURRENT_NORMAL */
-   {  5861, 0x00008843 }, /* GL_CURRENT_PALETTE_MATRIX_ARB */
-   {  5891, 0x00008B8D }, /* GL_CURRENT_PROGRAM */
-   {  5910, 0x00008865 }, /* GL_CURRENT_QUERY */
-   {  5927, 0x00008865 }, /* GL_CURRENT_QUERY_ARB */
-   {  5948, 0x00000B04 }, /* GL_CURRENT_RASTER_COLOR */
-   {  5972, 0x00000B09 }, /* GL_CURRENT_RASTER_DISTANCE */
-   {  5999, 0x00000B05 }, /* GL_CURRENT_RASTER_INDEX */
-   {  6023, 0x00000B07 }, /* GL_CURRENT_RASTER_POSITION */
-   {  6050, 0x00000B08 }, /* GL_CURRENT_RASTER_POSITION_VALID */
-   {  6083, 0x00000B06 }, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
-   {  6116, 0x00008459 }, /* GL_CURRENT_SECONDARY_COLOR */
-   {  6143, 0x00000B03 }, /* GL_CURRENT_TEXTURE_COORDS */
-   {  6169, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB */
-   {  6194, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB_ARB */
-   {  6223, 0x000086A8 }, /* GL_CURRENT_WEIGHT_ARB */
-   {  6245, 0x00000900 }, /* GL_CW */
-   {  6251, 0x0000875B }, /* GL_DEBUG_ASSERT_MESA */
-   {  6272, 0x00008759 }, /* GL_DEBUG_OBJECT_MESA */
-   {  6293, 0x0000875A }, /* GL_DEBUG_PRINT_MESA */
-   {  6313, 0x00002101 }, /* GL_DECAL */
-   {  6322, 0x00001E03 }, /* GL_DECR */
-   {  6330, 0x00008508 }, /* GL_DECR_WRAP */
-   {  6343, 0x00008508 }, /* GL_DECR_WRAP_EXT */
-   {  6360, 0x00008B80 }, /* GL_DELETE_STATUS */
-   {  6377, 0x00001801 }, /* GL_DEPTH */
-   {  6386, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT_EXT */
-   {  6410, 0x00000D1F }, /* GL_DEPTH_BIAS */
-   {  6424, 0x00000D56 }, /* GL_DEPTH_BITS */
-   {  6438, 0x00008891 }, /* GL_DEPTH_BOUNDS_EXT */
-   {  6458, 0x00008890 }, /* GL_DEPTH_BOUNDS_TEST_EXT */
-   {  6483, 0x00000100 }, /* GL_DEPTH_BUFFER_BIT */
-   {  6503, 0x0000864F }, /* GL_DEPTH_CLAMP_NV */
-   {  6521, 0x00000B73 }, /* GL_DEPTH_CLEAR_VALUE */
-   {  6542, 0x00001902 }, /* GL_DEPTH_COMPONENT */
-   {  6561, 0x000081A5 }, /* GL_DEPTH_COMPONENT16 */
-   {  6582, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_ARB */
-   {  6607, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_SGIX */
-   {  6633, 0x000081A6 }, /* GL_DEPTH_COMPONENT24 */
-   {  6654, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_ARB */
-   {  6679, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_SGIX */
-   {  6705, 0x000081A7 }, /* GL_DEPTH_COMPONENT32 */
-   {  6726, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_ARB */
-   {  6751, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_SGIX */
-   {  6777, 0x00000B74 }, /* GL_DEPTH_FUNC */
-   {  6791, 0x00000B70 }, /* GL_DEPTH_RANGE */
-   {  6806, 0x00000D1E }, /* GL_DEPTH_SCALE */
-   {  6821, 0x000084F9 }, /* GL_DEPTH_STENCIL_NV */
-   {  6841, 0x0000886F }, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
-   {  6869, 0x0000886E }, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
-   {  6897, 0x00000B71 }, /* GL_DEPTH_TEST */
-   {  6911, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE */
-   {  6933, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE_ARB */
-   {  6959, 0x00000B72 }, /* GL_DEPTH_WRITEMASK */
-   {  6978, 0x00001201 }, /* GL_DIFFUSE */
-   {  6989, 0x00000BD0 }, /* GL_DITHER */
-   {  6999, 0x00000A02 }, /* GL_DOMAIN */
-   {  7009, 0x00001100 }, /* GL_DONT_CARE */
-   {  7022, 0x000086AE }, /* GL_DOT3_RGB */
-   {  7034, 0x000086AF }, /* GL_DOT3_RGBA */
-   {  7047, 0x000086AF }, /* GL_DOT3_RGBA_ARB */
-   {  7064, 0x00008741 }, /* GL_DOT3_RGBA_EXT */
-   {  7081, 0x000086AE }, /* GL_DOT3_RGB_ARB */
-   {  7097, 0x00008740 }, /* GL_DOT3_RGB_EXT */
-   {  7113, 0x0000140A }, /* GL_DOUBLE */
-   {  7123, 0x00000C32 }, /* GL_DOUBLEBUFFER */
-   {  7139, 0x00000C01 }, /* GL_DRAW_BUFFER */
-   {  7154, 0x00008825 }, /* GL_DRAW_BUFFER0 */
-   {  7170, 0x00008825 }, /* GL_DRAW_BUFFER0_ARB */
-   {  7190, 0x00008825 }, /* GL_DRAW_BUFFER0_ATI */
-   {  7210, 0x00008826 }, /* GL_DRAW_BUFFER1 */
-   {  7226, 0x0000882F }, /* GL_DRAW_BUFFER10 */
-   {  7243, 0x0000882F }, /* GL_DRAW_BUFFER10_ARB */
-   {  7264, 0x0000882F }, /* GL_DRAW_BUFFER10_ATI */
-   {  7285, 0x00008830 }, /* GL_DRAW_BUFFER11 */
-   {  7302, 0x00008830 }, /* GL_DRAW_BUFFER11_ARB */
-   {  7323, 0x00008830 }, /* GL_DRAW_BUFFER11_ATI */
-   {  7344, 0x00008831 }, /* GL_DRAW_BUFFER12 */
-   {  7361, 0x00008831 }, /* GL_DRAW_BUFFER12_ARB */
-   {  7382, 0x00008831 }, /* GL_DRAW_BUFFER12_ATI */
-   {  7403, 0x00008832 }, /* GL_DRAW_BUFFER13 */
-   {  7420, 0x00008832 }, /* GL_DRAW_BUFFER13_ARB */
-   {  7441, 0x00008832 }, /* GL_DRAW_BUFFER13_ATI */
-   {  7462, 0x00008833 }, /* GL_DRAW_BUFFER14 */
-   {  7479, 0x00008833 }, /* GL_DRAW_BUFFER14_ARB */
-   {  7500, 0x00008833 }, /* GL_DRAW_BUFFER14_ATI */
-   {  7521, 0x00008834 }, /* GL_DRAW_BUFFER15 */
-   {  7538, 0x00008834 }, /* GL_DRAW_BUFFER15_ARB */
-   {  7559, 0x00008834 }, /* GL_DRAW_BUFFER15_ATI */
-   {  7580, 0x00008826 }, /* GL_DRAW_BUFFER1_ARB */
-   {  7600, 0x00008826 }, /* GL_DRAW_BUFFER1_ATI */
-   {  7620, 0x00008827 }, /* GL_DRAW_BUFFER2 */
-   {  7636, 0x00008827 }, /* GL_DRAW_BUFFER2_ARB */
-   {  7656, 0x00008827 }, /* GL_DRAW_BUFFER2_ATI */
-   {  7676, 0x00008828 }, /* GL_DRAW_BUFFER3 */
-   {  7692, 0x00008828 }, /* GL_DRAW_BUFFER3_ARB */
-   {  7712, 0x00008828 }, /* GL_DRAW_BUFFER3_ATI */
-   {  7732, 0x00008829 }, /* GL_DRAW_BUFFER4 */
-   {  7748, 0x00008829 }, /* GL_DRAW_BUFFER4_ARB */
-   {  7768, 0x00008829 }, /* GL_DRAW_BUFFER4_ATI */
-   {  7788, 0x0000882A }, /* GL_DRAW_BUFFER5 */
-   {  7804, 0x0000882A }, /* GL_DRAW_BUFFER5_ARB */
-   {  7824, 0x0000882A }, /* GL_DRAW_BUFFER5_ATI */
-   {  7844, 0x0000882B }, /* GL_DRAW_BUFFER6 */
-   {  7860, 0x0000882B }, /* GL_DRAW_BUFFER6_ARB */
-   {  7880, 0x0000882B }, /* GL_DRAW_BUFFER6_ATI */
-   {  7900, 0x0000882C }, /* GL_DRAW_BUFFER7 */
-   {  7916, 0x0000882C }, /* GL_DRAW_BUFFER7_ARB */
-   {  7936, 0x0000882C }, /* GL_DRAW_BUFFER7_ATI */
-   {  7956, 0x0000882D }, /* GL_DRAW_BUFFER8 */
-   {  7972, 0x0000882D }, /* GL_DRAW_BUFFER8_ARB */
-   {  7992, 0x0000882D }, /* GL_DRAW_BUFFER8_ATI */
-   {  8012, 0x0000882E }, /* GL_DRAW_BUFFER9 */
-   {  8028, 0x0000882E }, /* GL_DRAW_BUFFER9_ARB */
-   {  8048, 0x0000882E }, /* GL_DRAW_BUFFER9_ATI */
-   {  8068, 0x00008CA6 }, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
-   {  8100, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER_EXT */
-   {  8124, 0x00000705 }, /* GL_DRAW_PIXEL_TOKEN */
-   {  8144, 0x00000304 }, /* GL_DST_ALPHA */
-   {  8157, 0x00000306 }, /* GL_DST_COLOR */
-   {  8170, 0x000088EA }, /* GL_DYNAMIC_COPY */
-   {  8186, 0x000088EA }, /* GL_DYNAMIC_COPY_ARB */
-   {  8206, 0x000088E8 }, /* GL_DYNAMIC_DRAW */
-   {  8222, 0x000088E8 }, /* GL_DYNAMIC_DRAW_ARB */
-   {  8242, 0x000088E9 }, /* GL_DYNAMIC_READ */
-   {  8258, 0x000088E9 }, /* GL_DYNAMIC_READ_ARB */
-   {  8278, 0x00000B43 }, /* GL_EDGE_FLAG */
-   {  8291, 0x00008079 }, /* GL_EDGE_FLAG_ARRAY */
-   {  8310, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
-   {  8344, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB */
-   {  8382, 0x00008093 }, /* GL_EDGE_FLAG_ARRAY_POINTER */
-   {  8409, 0x0000808C }, /* GL_EDGE_FLAG_ARRAY_STRIDE */
-   {  8435, 0x00008893 }, /* GL_ELEMENT_ARRAY_BUFFER */
-   {  8459, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
-   {  8491, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB */
-   {  8527, 0x00001600 }, /* GL_EMISSION */
-   {  8539, 0x00002000 }, /* GL_ENABLE_BIT */
-   {  8553, 0x00000202 }, /* GL_EQUAL */
-   {  8562, 0x00001509 }, /* GL_EQUIV */
-   {  8571, 0x00010000 }, /* GL_EVAL_BIT */
-   {  8583, 0x00000800 }, /* GL_EXP */
-   {  8590, 0x00000801 }, /* GL_EXP2 */
-   {  8598, 0x00001F03 }, /* GL_EXTENSIONS */
-   {  8612, 0x00002400 }, /* GL_EYE_LINEAR */
-   {  8626, 0x00002502 }, /* GL_EYE_PLANE */
-   {  8639, 0x0000855C }, /* GL_EYE_PLANE_ABSOLUTE_NV */
-   {  8664, 0x0000855B }, /* GL_EYE_RADIAL_NV */
-   {  8681, 0x00000000 }, /* GL_FALSE */
-   {  8690, 0x00001101 }, /* GL_FASTEST */
-   {  8701, 0x00001C01 }, /* GL_FEEDBACK */
-   {  8713, 0x00000DF0 }, /* GL_FEEDBACK_BUFFER_POINTER */
-   {  8740, 0x00000DF1 }, /* GL_FEEDBACK_BUFFER_SIZE */
-   {  8764, 0x00000DF2 }, /* GL_FEEDBACK_BUFFER_TYPE */
-   {  8788, 0x00001B02 }, /* GL_FILL */
-   {  8796, 0x00001D00 }, /* GL_FLAT */
-   {  8804, 0x00001406 }, /* GL_FLOAT */
-   {  8813, 0x00008B5A }, /* GL_FLOAT_MAT2 */
-   {  8827, 0x00008B5A }, /* GL_FLOAT_MAT2_ARB */
-   {  8845, 0x00008B5B }, /* GL_FLOAT_MAT3 */
-   {  8859, 0x00008B5B }, /* GL_FLOAT_MAT3_ARB */
-   {  8877, 0x00008B5C }, /* GL_FLOAT_MAT4 */
-   {  8891, 0x00008B5C }, /* GL_FLOAT_MAT4_ARB */
-   {  8909, 0x00008B50 }, /* GL_FLOAT_VEC2 */
-   {  8923, 0x00008B50 }, /* GL_FLOAT_VEC2_ARB */
-   {  8941, 0x00008B51 }, /* GL_FLOAT_VEC3 */
-   {  8955, 0x00008B51 }, /* GL_FLOAT_VEC3_ARB */
-   {  8973, 0x00008B52 }, /* GL_FLOAT_VEC4 */
-   {  8987, 0x00008B52 }, /* GL_FLOAT_VEC4_ARB */
-   {  9005, 0x00000B60 }, /* GL_FOG */
-   {  9012, 0x00000080 }, /* GL_FOG_BIT */
-   {  9023, 0x00000B66 }, /* GL_FOG_COLOR */
-   {  9036, 0x00008451 }, /* GL_FOG_COORD */
-   {  9049, 0x00008451 }, /* GL_FOG_COORDINATE */
-   {  9067, 0x00008457 }, /* GL_FOG_COORDINATE_ARRAY */
-   {  9091, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
-   {  9130, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB */
-   {  9173, 0x00008456 }, /* GL_FOG_COORDINATE_ARRAY_POINTER */
-   {  9205, 0x00008455 }, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
-   {  9236, 0x00008454 }, /* GL_FOG_COORDINATE_ARRAY_TYPE */
-   {  9265, 0x00008450 }, /* GL_FOG_COORDINATE_SOURCE */
-   {  9290, 0x00008457 }, /* GL_FOG_COORD_ARRAY */
-   {  9309, 0x0000889D }, /* GL_FOG_COORD_ARRAY_BUFFER_BINDING */
-   {  9343, 0x00008456 }, /* GL_FOG_COORD_ARRAY_POINTER */
-   {  9370, 0x00008455 }, /* GL_FOG_COORD_ARRAY_STRIDE */
-   {  9396, 0x00008454 }, /* GL_FOG_COORD_ARRAY_TYPE */
-   {  9420, 0x00008450 }, /* GL_FOG_COORD_SRC */
-   {  9437, 0x00000B62 }, /* GL_FOG_DENSITY */
-   {  9452, 0x0000855A }, /* GL_FOG_DISTANCE_MODE_NV */
-   {  9476, 0x00000B64 }, /* GL_FOG_END */
-   {  9487, 0x00000C54 }, /* GL_FOG_HINT */
-   {  9499, 0x00000B61 }, /* GL_FOG_INDEX */
-   {  9512, 0x00000B65 }, /* GL_FOG_MODE */
-   {  9524, 0x00008198 }, /* GL_FOG_OFFSET_SGIX */
-   {  9543, 0x00008199 }, /* GL_FOG_OFFSET_VALUE_SGIX */
-   {  9568, 0x00000B63 }, /* GL_FOG_START */
-   {  9581, 0x00008452 }, /* GL_FRAGMENT_DEPTH */
-   {  9599, 0x00008804 }, /* GL_FRAGMENT_PROGRAM_ARB */
-   {  9623, 0x00008B30 }, /* GL_FRAGMENT_SHADER */
-   {  9642, 0x00008B30 }, /* GL_FRAGMENT_SHADER_ARB */
-   {  9665, 0x00008B8B }, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
-   {  9700, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME_EXT */
-   {  9742, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE_EXT */
-   {  9784, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_3D_ZOFFSET_EXT */
-   {  9833, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE_EXT */
-   {  9885, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER_EXT */
-   {  9929, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL_EXT */
-   {  9973, 0x00008CA6 }, /* GL_FRAMEBUFFER_BINDING_EXT */
-   { 10000, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE_EXT */
-   { 10028, 0x00008D40 }, /* GL_FRAMEBUFFER_EXT */
-   { 10047, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT_EXT */
-   { 10088, 0x00008CD9 }, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
-   { 10129, 0x00008CDB }, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
-   { 10171, 0x00008CD8 }, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
-   { 10222, 0x00008CDA }, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
-   { 10260, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT_EXT */
-   { 10309, 0x00008CDC }, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
-   { 10351, 0x00008CDE }, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
-   { 10383, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED_EXT */
-   { 10414, 0x00000404 }, /* GL_FRONT */
-   { 10423, 0x00000408 }, /* GL_FRONT_AND_BACK */
-   { 10441, 0x00000B46 }, /* GL_FRONT_FACE */
-   { 10455, 0x00000400 }, /* GL_FRONT_LEFT */
-   { 10469, 0x00000401 }, /* GL_FRONT_RIGHT */
-   { 10484, 0x00008006 }, /* GL_FUNC_ADD */
-   { 10496, 0x00008006 }, /* GL_FUNC_ADD_EXT */
-   { 10512, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT */
-   { 10537, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT_EXT */
-   { 10566, 0x0000800A }, /* GL_FUNC_SUBTRACT */
-   { 10583, 0x0000800A }, /* GL_FUNC_SUBTRACT_EXT */
-   { 10604, 0x00008191 }, /* GL_GENERATE_MIPMAP */
-   { 10623, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT */
-   { 10647, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT_SGIS */
-   { 10676, 0x00008191 }, /* GL_GENERATE_MIPMAP_SGIS */
-   { 10700, 0x00000206 }, /* GL_GEQUAL */
-   { 10710, 0x00008009 }, /* GL_GL_BLEND_EQUATION_RGB */
-   { 10735, 0x00008C4A }, /* GL_GL_COMPRESSED_SLUMINANCE */
-   { 10763, 0x00008C4B }, /* GL_GL_COMPRESSED_SLUMINANCE_ALPHA */
-   { 10797, 0x00008C48 }, /* GL_GL_COMPRESSED_SRGB */
-   { 10819, 0x00008C49 }, /* GL_GL_COMPRESSED_SRGB_ALPHA */
-   { 10847, 0x0000845F }, /* GL_GL_CURRENT_RASTER_SECONDARY_COLOR */
-   { 10884, 0x00008B65 }, /* GL_GL_FLOAT_MAT2x3 */
-   { 10903, 0x00008B66 }, /* GL_GL_FLOAT_MAT2x4 */
-   { 10922, 0x00008B67 }, /* GL_GL_FLOAT_MAT3x2 */
-   { 10941, 0x00008B68 }, /* GL_GL_FLOAT_MAT3x4 */
-   { 10960, 0x00008B69 }, /* GL_GL_FLOAT_MAT4x2 */
-   { 10979, 0x00008B6A }, /* GL_GL_FLOAT_MAT4x3 */
-   { 10998, 0x000088EB }, /* GL_GL_PIXEL_PACK_BUFFER */
-   { 11022, 0x000088ED }, /* GL_GL_PIXEL_PACK_BUFFER_BINDING */
-   { 11054, 0x000088EC }, /* GL_GL_PIXEL_UNPACK_BUFFER */
-   { 11080, 0x000088EF }, /* GL_GL_PIXEL_UNPACK_BUFFER_BINDING */
-   { 11114, 0x00008C46 }, /* GL_GL_SLUMINANCE */
-   { 11131, 0x00008C47 }, /* GL_GL_SLUMINANCE8 */
-   { 11149, 0x00008C45 }, /* GL_GL_SLUMINANCE8_ALPHA8 */
-   { 11174, 0x00008C44 }, /* GL_GL_SLUMINANCE_ALPHA */
-   { 11197, 0x00008C40 }, /* GL_GL_SRGB */
-   { 11208, 0x00008C41 }, /* GL_GL_SRGB8 */
-   { 11220, 0x00008C43 }, /* GL_GL_SRGB8_ALPHA8 */
-   { 11239, 0x00008C42 }, /* GL_GL_SRGB_ALPHA */
-   { 11256, 0x00000204 }, /* GL_GREATER */
-   { 11267, 0x00001904 }, /* GL_GREEN */
-   { 11276, 0x00000D19 }, /* GL_GREEN_BIAS */
-   { 11290, 0x00000D53 }, /* GL_GREEN_BITS */
-   { 11304, 0x00000D18 }, /* GL_GREEN_SCALE */
-   { 11319, 0x00008000 }, /* GL_HINT_BIT */
-   { 11331, 0x00008024 }, /* GL_HISTOGRAM */
-   { 11344, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE */
-   { 11368, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE_EXT */
-   { 11396, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE */
-   { 11419, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE_EXT */
-   { 11446, 0x00008024 }, /* GL_HISTOGRAM_EXT */
-   { 11463, 0x00008027 }, /* GL_HISTOGRAM_FORMAT */
-   { 11483, 0x00008027 }, /* GL_HISTOGRAM_FORMAT_EXT */
-   { 11507, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE */
-   { 11531, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE_EXT */
-   { 11559, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE */
-   { 11587, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE_EXT */
-   { 11619, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE */
-   { 11641, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE_EXT */
-   { 11667, 0x0000802D }, /* GL_HISTOGRAM_SINK */
-   { 11685, 0x0000802D }, /* GL_HISTOGRAM_SINK_EXT */
-   { 11707, 0x00008026 }, /* GL_HISTOGRAM_WIDTH */
-   { 11726, 0x00008026 }, /* GL_HISTOGRAM_WIDTH_EXT */
-   { 11749, 0x0000862A }, /* GL_IDENTITY_NV */
-   { 11764, 0x00008150 }, /* GL_IGNORE_BORDER_HP */
-   { 11784, 0x00008B9B }, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
-   { 11824, 0x00008B9A }, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
-   { 11862, 0x00001E02 }, /* GL_INCR */
-   { 11870, 0x00008507 }, /* GL_INCR_WRAP */
-   { 11883, 0x00008507 }, /* GL_INCR_WRAP_EXT */
-   { 11900, 0x00008077 }, /* GL_INDEX_ARRAY */
-   { 11915, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING */
-   { 11945, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING_ARB */
-   { 11979, 0x00008091 }, /* GL_INDEX_ARRAY_POINTER */
-   { 12002, 0x00008086 }, /* GL_INDEX_ARRAY_STRIDE */
-   { 12024, 0x00008085 }, /* GL_INDEX_ARRAY_TYPE */
-   { 12044, 0x00000D51 }, /* GL_INDEX_BITS */
-   { 12058, 0x00000C20 }, /* GL_INDEX_CLEAR_VALUE */
-   { 12079, 0x00000BF1 }, /* GL_INDEX_LOGIC_OP */
-   { 12097, 0x00000C30 }, /* GL_INDEX_MODE */
-   { 12111, 0x00000D13 }, /* GL_INDEX_OFFSET */
-   { 12127, 0x00000D12 }, /* GL_INDEX_SHIFT */
-   { 12142, 0x00000C21 }, /* GL_INDEX_WRITEMASK */
-   { 12161, 0x00008B84 }, /* GL_INFO_LOG_LENGTH */
-   { 12180, 0x00001404 }, /* GL_INT */
-   { 12187, 0x00008049 }, /* GL_INTENSITY */
-   { 12200, 0x0000804C }, /* GL_INTENSITY12 */
-   { 12215, 0x0000804C }, /* GL_INTENSITY12_EXT */
-   { 12234, 0x0000804D }, /* GL_INTENSITY16 */
-   { 12249, 0x0000804D }, /* GL_INTENSITY16_EXT */
-   { 12268, 0x0000804A }, /* GL_INTENSITY4 */
-   { 12282, 0x0000804A }, /* GL_INTENSITY4_EXT */
-   { 12300, 0x0000804B }, /* GL_INTENSITY8 */
-   { 12314, 0x0000804B }, /* GL_INTENSITY8_EXT */
-   { 12332, 0x00008049 }, /* GL_INTENSITY_EXT */
-   { 12349, 0x00008575 }, /* GL_INTERPOLATE */
-   { 12364, 0x00008575 }, /* GL_INTERPOLATE_ARB */
-   { 12383, 0x00008575 }, /* GL_INTERPOLATE_EXT */
-   { 12402, 0x00008B53 }, /* GL_INT_VEC2 */
-   { 12414, 0x00008B53 }, /* GL_INT_VEC2_ARB */
-   { 12430, 0x00008B54 }, /* GL_INT_VEC3 */
-   { 12442, 0x00008B54 }, /* GL_INT_VEC3_ARB */
-   { 12458, 0x00008B55 }, /* GL_INT_VEC4 */
-   { 12470, 0x00008B55 }, /* GL_INT_VEC4_ARB */
-   { 12486, 0x00000500 }, /* GL_INVALID_ENUM */
-   { 12502, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION_EXT */
-   { 12539, 0x00000502 }, /* GL_INVALID_OPERATION */
-   { 12560, 0x00000501 }, /* GL_INVALID_VALUE */
-   { 12577, 0x0000862B }, /* GL_INVERSE_NV */
-   { 12591, 0x0000862D }, /* GL_INVERSE_TRANSPOSE_NV */
-   { 12615, 0x0000150A }, /* GL_INVERT */
-   { 12625, 0x00001E00 }, /* GL_KEEP */
-   { 12633, 0x00000406 }, /* GL_LEFT */
-   { 12641, 0x00000203 }, /* GL_LEQUAL */
-   { 12651, 0x00000201 }, /* GL_LESS */
-   { 12659, 0x00004000 }, /* GL_LIGHT0 */
-   { 12669, 0x00004001 }, /* GL_LIGHT1 */
-   { 12679, 0x00004002 }, /* GL_LIGHT2 */
-   { 12689, 0x00004003 }, /* GL_LIGHT3 */
-   { 12699, 0x00004004 }, /* GL_LIGHT4 */
-   { 12709, 0x00004005 }, /* GL_LIGHT5 */
-   { 12719, 0x00004006 }, /* GL_LIGHT6 */
-   { 12729, 0x00004007 }, /* GL_LIGHT7 */
-   { 12739, 0x00000B50 }, /* GL_LIGHTING */
-   { 12751, 0x00000040 }, /* GL_LIGHTING_BIT */
-   { 12767, 0x00000B53 }, /* GL_LIGHT_MODEL_AMBIENT */
-   { 12790, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL */
-   { 12819, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL_EXT */
-   { 12852, 0x00000B51 }, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
-   { 12880, 0x00000B52 }, /* GL_LIGHT_MODEL_TWO_SIDE */
-   { 12904, 0x00001B01 }, /* GL_LINE */
-   { 12912, 0x00002601 }, /* GL_LINEAR */
-   { 12922, 0x00001208 }, /* GL_LINEAR_ATTENUATION */
-   { 12944, 0x00008170 }, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
-   { 12974, 0x0000844F }, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
-   { 13005, 0x00002703 }, /* GL_LINEAR_MIPMAP_LINEAR */
-   { 13029, 0x00002701 }, /* GL_LINEAR_MIPMAP_NEAREST */
-   { 13054, 0x00000001 }, /* GL_LINES */
-   { 13063, 0x00000004 }, /* GL_LINE_BIT */
-   { 13075, 0x00000002 }, /* GL_LINE_LOOP */
-   { 13088, 0x00000707 }, /* GL_LINE_RESET_TOKEN */
-   { 13108, 0x00000B20 }, /* GL_LINE_SMOOTH */
-   { 13123, 0x00000C52 }, /* GL_LINE_SMOOTH_HINT */
-   { 13143, 0x00000B24 }, /* GL_LINE_STIPPLE */
-   { 13159, 0x00000B25 }, /* GL_LINE_STIPPLE_PATTERN */
-   { 13183, 0x00000B26 }, /* GL_LINE_STIPPLE_REPEAT */
-   { 13206, 0x00000003 }, /* GL_LINE_STRIP */
-   { 13220, 0x00000702 }, /* GL_LINE_TOKEN */
-   { 13234, 0x00000B21 }, /* GL_LINE_WIDTH */
-   { 13248, 0x00000B23 }, /* GL_LINE_WIDTH_GRANULARITY */
-   { 13274, 0x00000B22 }, /* GL_LINE_WIDTH_RANGE */
-   { 13294, 0x00008B82 }, /* GL_LINK_STATUS */
-   { 13309, 0x00000B32 }, /* GL_LIST_BASE */
-   { 13322, 0x00020000 }, /* GL_LIST_BIT */
-   { 13334, 0x00000B33 }, /* GL_LIST_INDEX */
-   { 13348, 0x00000B30 }, /* GL_LIST_MODE */
-   { 13361, 0x00000101 }, /* GL_LOAD */
-   { 13369, 0x00000BF1 }, /* GL_LOGIC_OP */
-   { 13381, 0x00000BF0 }, /* GL_LOGIC_OP_MODE */
-   { 13398, 0x00008CA1 }, /* GL_LOWER_LEFT */
-   { 13412, 0x00001909 }, /* GL_LUMINANCE */
-   { 13425, 0x00008041 }, /* GL_LUMINANCE12 */
-   { 13440, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12 */
-   { 13463, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12_EXT */
-   { 13490, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4 */
-   { 13512, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4_EXT */
-   { 13538, 0x00008041 }, /* GL_LUMINANCE12_EXT */
-   { 13557, 0x00008042 }, /* GL_LUMINANCE16 */
-   { 13572, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16 */
-   { 13595, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16_EXT */
-   { 13622, 0x00008042 }, /* GL_LUMINANCE16_EXT */
-   { 13641, 0x0000803F }, /* GL_LUMINANCE4 */
-   { 13655, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4 */
-   { 13676, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4_EXT */
-   { 13701, 0x0000803F }, /* GL_LUMINANCE4_EXT */
-   { 13719, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2 */
-   { 13740, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2_EXT */
-   { 13765, 0x00008040 }, /* GL_LUMINANCE8 */
-   { 13779, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8 */
-   { 13800, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8_EXT */
-   { 13825, 0x00008040 }, /* GL_LUMINANCE8_EXT */
-   { 13843, 0x0000190A }, /* GL_LUMINANCE_ALPHA */
-   { 13862, 0x00000D90 }, /* GL_MAP1_COLOR_4 */
-   { 13878, 0x00000DD0 }, /* GL_MAP1_GRID_DOMAIN */
-   { 13898, 0x00000DD1 }, /* GL_MAP1_GRID_SEGMENTS */
-   { 13920, 0x00000D91 }, /* GL_MAP1_INDEX */
-   { 13934, 0x00000D92 }, /* GL_MAP1_NORMAL */
-   { 13949, 0x00000D93 }, /* GL_MAP1_TEXTURE_COORD_1 */
-   { 13973, 0x00000D94 }, /* GL_MAP1_TEXTURE_COORD_2 */
-   { 13997, 0x00000D95 }, /* GL_MAP1_TEXTURE_COORD_3 */
-   { 14021, 0x00000D96 }, /* GL_MAP1_TEXTURE_COORD_4 */
-   { 14045, 0x00000D97 }, /* GL_MAP1_VERTEX_3 */
-   { 14062, 0x00000D98 }, /* GL_MAP1_VERTEX_4 */
-   { 14079, 0x00008660 }, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
-   { 14107, 0x0000866A }, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
-   { 14136, 0x0000866B }, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
-   { 14165, 0x0000866C }, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
-   { 14194, 0x0000866D }, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
-   { 14223, 0x0000866E }, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
-   { 14252, 0x0000866F }, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
-   { 14281, 0x00008661 }, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
-   { 14309, 0x00008662 }, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
-   { 14337, 0x00008663 }, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
-   { 14365, 0x00008664 }, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
-   { 14393, 0x00008665 }, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
-   { 14421, 0x00008666 }, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
-   { 14449, 0x00008667 }, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
-   { 14477, 0x00008668 }, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
-   { 14505, 0x00008669 }, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
-   { 14533, 0x00000DB0 }, /* GL_MAP2_COLOR_4 */
-   { 14549, 0x00000DD2 }, /* GL_MAP2_GRID_DOMAIN */
-   { 14569, 0x00000DD3 }, /* GL_MAP2_GRID_SEGMENTS */
-   { 14591, 0x00000DB1 }, /* GL_MAP2_INDEX */
-   { 14605, 0x00000DB2 }, /* GL_MAP2_NORMAL */
-   { 14620, 0x00000DB3 }, /* GL_MAP2_TEXTURE_COORD_1 */
-   { 14644, 0x00000DB4 }, /* GL_MAP2_TEXTURE_COORD_2 */
-   { 14668, 0x00000DB5 }, /* GL_MAP2_TEXTURE_COORD_3 */
-   { 14692, 0x00000DB6 }, /* GL_MAP2_TEXTURE_COORD_4 */
-   { 14716, 0x00000DB7 }, /* GL_MAP2_VERTEX_3 */
-   { 14733, 0x00000DB8 }, /* GL_MAP2_VERTEX_4 */
-   { 14750, 0x00008670 }, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
-   { 14778, 0x0000867A }, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
-   { 14807, 0x0000867B }, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
-   { 14836, 0x0000867C }, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
-   { 14865, 0x0000867D }, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
-   { 14894, 0x0000867E }, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
-   { 14923, 0x0000867F }, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
-   { 14952, 0x00008671 }, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
-   { 14980, 0x00008672 }, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
-   { 15008, 0x00008673 }, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
-   { 15036, 0x00008674 }, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
-   { 15064, 0x00008675 }, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
-   { 15092, 0x00008676 }, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
-   { 15120, 0x00008677 }, /* GL_MAP2_VERTEX_ATTRIB7_4_NV */
-   { 15148, 0x00008678 }, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
-   { 15176, 0x00008679 }, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
-   { 15204, 0x00000D10 }, /* GL_MAP_COLOR */
-   { 15217, 0x00000D11 }, /* GL_MAP_STENCIL */
-   { 15232, 0x000088C0 }, /* GL_MATRIX0_ARB */
-   { 15247, 0x00008630 }, /* GL_MATRIX0_NV */
-   { 15261, 0x000088CA }, /* GL_MATRIX10_ARB */
-   { 15277, 0x000088CB }, /* GL_MATRIX11_ARB */
-   { 15293, 0x000088CC }, /* GL_MATRIX12_ARB */
-   { 15309, 0x000088CD }, /* GL_MATRIX13_ARB */
-   { 15325, 0x000088CE }, /* GL_MATRIX14_ARB */
-   { 15341, 0x000088CF }, /* GL_MATRIX15_ARB */
-   { 15357, 0x000088D0 }, /* GL_MATRIX16_ARB */
-   { 15373, 0x000088D1 }, /* GL_MATRIX17_ARB */
-   { 15389, 0x000088D2 }, /* GL_MATRIX18_ARB */
-   { 15405, 0x000088D3 }, /* GL_MATRIX19_ARB */
-   { 15421, 0x000088C1 }, /* GL_MATRIX1_ARB */
-   { 15436, 0x00008631 }, /* GL_MATRIX1_NV */
-   { 15450, 0x000088D4 }, /* GL_MATRIX20_ARB */
-   { 15466, 0x000088D5 }, /* GL_MATRIX21_ARB */
-   { 15482, 0x000088D6 }, /* GL_MATRIX22_ARB */
-   { 15498, 0x000088D7 }, /* GL_MATRIX23_ARB */
-   { 15514, 0x000088D8 }, /* GL_MATRIX24_ARB */
-   { 15530, 0x000088D9 }, /* GL_MATRIX25_ARB */
-   { 15546, 0x000088DA }, /* GL_MATRIX26_ARB */
-   { 15562, 0x000088DB }, /* GL_MATRIX27_ARB */
-   { 15578, 0x000088DC }, /* GL_MATRIX28_ARB */
-   { 15594, 0x000088DD }, /* GL_MATRIX29_ARB */
-   { 15610, 0x000088C2 }, /* GL_MATRIX2_ARB */
-   { 15625, 0x00008632 }, /* GL_MATRIX2_NV */
-   { 15639, 0x000088DE }, /* GL_MATRIX30_ARB */
-   { 15655, 0x000088DF }, /* GL_MATRIX31_ARB */
-   { 15671, 0x000088C3 }, /* GL_MATRIX3_ARB */
-   { 15686, 0x00008633 }, /* GL_MATRIX3_NV */
-   { 15700, 0x000088C4 }, /* GL_MATRIX4_ARB */
-   { 15715, 0x00008634 }, /* GL_MATRIX4_NV */
-   { 15729, 0x000088C5 }, /* GL_MATRIX5_ARB */
-   { 15744, 0x00008635 }, /* GL_MATRIX5_NV */
-   { 15758, 0x000088C6 }, /* GL_MATRIX6_ARB */
-   { 15773, 0x00008636 }, /* GL_MATRIX6_NV */
-   { 15787, 0x000088C7 }, /* GL_MATRIX7_ARB */
-   { 15802, 0x00008637 }, /* GL_MATRIX7_NV */
-   { 15816, 0x000088C8 }, /* GL_MATRIX8_ARB */
-   { 15831, 0x000088C9 }, /* GL_MATRIX9_ARB */
-   { 15846, 0x00008844 }, /* GL_MATRIX_INDEX_ARRAY_ARB */
-   { 15872, 0x00008849 }, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
-   { 15906, 0x00008846 }, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
-   { 15937, 0x00008848 }, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
-   { 15970, 0x00008847 }, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
-   { 16001, 0x00000BA0 }, /* GL_MATRIX_MODE */
-   { 16016, 0x00008840 }, /* GL_MATRIX_PALETTE_ARB */
-   { 16038, 0x00008008 }, /* GL_MAX */
-   { 16045, 0x00008073 }, /* GL_MAX_3D_TEXTURE_SIZE */
-   { 16068, 0x000088FF }, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
-   { 16100, 0x00000D35 }, /* GL_MAX_ATTRIB_STACK_DEPTH */
-   { 16126, 0x00000D3B }, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
-   { 16159, 0x00008177 }, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
-   { 16185, 0x00008178 }, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-   { 16219, 0x00000D32 }, /* GL_MAX_CLIP_PLANES */
-   { 16238, 0x00008CDF }, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
-   { 16267, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
-   { 16299, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH_SGI */
-   { 16335, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
-   { 16371, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB */
-   { 16411, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT */
-   { 16437, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT_EXT */
-   { 16467, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH */
-   { 16492, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH_EXT */
-   { 16521, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
-   { 16550, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE_ARB */
-   { 16583, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS */
-   { 16603, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ARB */
-   { 16627, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ATI */
-   { 16651, 0x000080E9 }, /* GL_MAX_ELEMENTS_INDICES */
-   { 16675, 0x000080E8 }, /* GL_MAX_ELEMENTS_VERTICES */
-   { 16700, 0x00000D30 }, /* GL_MAX_EVAL_ORDER */
-   { 16718, 0x00008008 }, /* GL_MAX_EXT */
-   { 16729, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
-   { 16764, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB */
-   { 16803, 0x00000D31 }, /* GL_MAX_LIGHTS */
-   { 16817, 0x00000B31 }, /* GL_MAX_LIST_NESTING */
-   { 16837, 0x00008841 }, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
-   { 16875, 0x00000D36 }, /* GL_MAX_MODELVIEW_STACK_DEPTH */
-   { 16904, 0x00000D37 }, /* GL_MAX_NAME_STACK_DEPTH */
-   { 16928, 0x00008842 }, /* GL_MAX_PALETTE_MATRICES_ARB */
-   { 16956, 0x00000D34 }, /* GL_MAX_PIXEL_MAP_TABLE */
-   { 16979, 0x000088B1 }, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
-   { 17016, 0x0000880B }, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
-   { 17052, 0x000088AD }, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
-   { 17079, 0x000088F5 }, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
-   { 17108, 0x000088B5 }, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
-   { 17142, 0x000088F4 }, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
-   { 17178, 0x000088F6 }, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
-   { 17205, 0x000088A1 }, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
-   { 17237, 0x000088B4 }, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
-   { 17273, 0x000088F8 }, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
-   { 17302, 0x000088F7 }, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
-   { 17331, 0x0000862F }, /* GL_MAX_PROGRAM_MATRICES_ARB */
-   { 17359, 0x0000862E }, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
-   { 17397, 0x000088B3 }, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-   { 17441, 0x0000880E }, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-   { 17484, 0x000088AF }, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
-   { 17518, 0x000088A3 }, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-   { 17557, 0x000088AB }, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
-   { 17594, 0x000088A7 }, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
-   { 17632, 0x00008810 }, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-   { 17675, 0x0000880F }, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-   { 17718, 0x000088A9 }, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
-   { 17748, 0x000088A5 }, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
-   { 17779, 0x0000880D }, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
-   { 17815, 0x0000880C }, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
-   { 17851, 0x00000D38 }, /* GL_MAX_PROJECTION_STACK_DEPTH */
-   { 17881, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
-   { 17915, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_NV */
-   { 17948, 0x000084E8 }, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
-   { 17977, 0x00008504 }, /* GL_MAX_SHININESS_NV */
-   { 17997, 0x00008505 }, /* GL_MAX_SPOT_EXPONENT_NV */
-   { 18021, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS */
-   { 18043, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS_ARB */
-   { 18069, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS */
-   { 18096, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS_ARB */
-   { 18127, 0x000084FD }, /* GL_MAX_TEXTURE_LOD_BIAS */
-   { 18151, 0x000084FF }, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
-   { 18185, 0x00000D33 }, /* GL_MAX_TEXTURE_SIZE */
-   { 18205, 0x00000D39 }, /* GL_MAX_TEXTURE_STACK_DEPTH */
-   { 18232, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS */
-   { 18253, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS_ARB */
-   { 18278, 0x0000862F }, /* GL_MAX_TRACK_MATRICES_NV */
-   { 18303, 0x0000862E }, /* GL_MAX_TRACK_MATRIX_STACK_DEPTH_NV */
-   { 18338, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS */
-   { 18360, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS_ARB */
-   { 18386, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS */
-   { 18408, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS_ARB */
-   { 18434, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
-   { 18468, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS_ARB */
-   { 18506, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
-   { 18539, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS_ARB */
-   { 18576, 0x000086A4 }, /* GL_MAX_VERTEX_UNITS_ARB */
-   { 18600, 0x00000D3A }, /* GL_MAX_VIEWPORT_DIMS */
-   { 18621, 0x00008007 }, /* GL_MIN */
-   { 18628, 0x0000802E }, /* GL_MINMAX */
-   { 18638, 0x0000802E }, /* GL_MINMAX_EXT */
-   { 18652, 0x0000802F }, /* GL_MINMAX_FORMAT */
-   { 18669, 0x0000802F }, /* GL_MINMAX_FORMAT_EXT */
-   { 18690, 0x00008030 }, /* GL_MINMAX_SINK */
-   { 18705, 0x00008030 }, /* GL_MINMAX_SINK_EXT */
-   { 18724, 0x00008007 }, /* GL_MIN_EXT */
-   { 18735, 0x00008370 }, /* GL_MIRRORED_REPEAT */
-   { 18754, 0x00008370 }, /* GL_MIRRORED_REPEAT_ARB */
-   { 18777, 0x00008370 }, /* GL_MIRRORED_REPEAT_IBM */
-   { 18800, 0x00008742 }, /* GL_MIRROR_CLAMP_ATI */
-   { 18820, 0x00008742 }, /* GL_MIRROR_CLAMP_EXT */
-   { 18840, 0x00008912 }, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
-   { 18870, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_ATI */
-   { 18898, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
-   { 18926, 0x00001700 }, /* GL_MODELVIEW */
-   { 18939, 0x00001700 }, /* GL_MODELVIEW0_ARB */
-   { 18957, 0x0000872A }, /* GL_MODELVIEW10_ARB */
-   { 18976, 0x0000872B }, /* GL_MODELVIEW11_ARB */
-   { 18995, 0x0000872C }, /* GL_MODELVIEW12_ARB */
-   { 19014, 0x0000872D }, /* GL_MODELVIEW13_ARB */
-   { 19033, 0x0000872E }, /* GL_MODELVIEW14_ARB */
-   { 19052, 0x0000872F }, /* GL_MODELVIEW15_ARB */
-   { 19071, 0x00008730 }, /* GL_MODELVIEW16_ARB */
-   { 19090, 0x00008731 }, /* GL_MODELVIEW17_ARB */
-   { 19109, 0x00008732 }, /* GL_MODELVIEW18_ARB */
-   { 19128, 0x00008733 }, /* GL_MODELVIEW19_ARB */
-   { 19147, 0x0000850A }, /* GL_MODELVIEW1_ARB */
-   { 19165, 0x00008734 }, /* GL_MODELVIEW20_ARB */
-   { 19184, 0x00008735 }, /* GL_MODELVIEW21_ARB */
-   { 19203, 0x00008736 }, /* GL_MODELVIEW22_ARB */
-   { 19222, 0x00008737 }, /* GL_MODELVIEW23_ARB */
-   { 19241, 0x00008738 }, /* GL_MODELVIEW24_ARB */
-   { 19260, 0x00008739 }, /* GL_MODELVIEW25_ARB */
-   { 19279, 0x0000873A }, /* GL_MODELVIEW26_ARB */
-   { 19298, 0x0000873B }, /* GL_MODELVIEW27_ARB */
-   { 19317, 0x0000873C }, /* GL_MODELVIEW28_ARB */
-   { 19336, 0x0000873D }, /* GL_MODELVIEW29_ARB */
-   { 19355, 0x00008722 }, /* GL_MODELVIEW2_ARB */
-   { 19373, 0x0000873E }, /* GL_MODELVIEW30_ARB */
-   { 19392, 0x0000873F }, /* GL_MODELVIEW31_ARB */
-   { 19411, 0x00008723 }, /* GL_MODELVIEW3_ARB */
-   { 19429, 0x00008724 }, /* GL_MODELVIEW4_ARB */
-   { 19447, 0x00008725 }, /* GL_MODELVIEW5_ARB */
-   { 19465, 0x00008726 }, /* GL_MODELVIEW6_ARB */
-   { 19483, 0x00008727 }, /* GL_MODELVIEW7_ARB */
-   { 19501, 0x00008728 }, /* GL_MODELVIEW8_ARB */
-   { 19519, 0x00008729 }, /* GL_MODELVIEW9_ARB */
-   { 19537, 0x00000BA6 }, /* GL_MODELVIEW_MATRIX */
-   { 19557, 0x00008629 }, /* GL_MODELVIEW_PROJECTION_NV */
-   { 19584, 0x00000BA3 }, /* GL_MODELVIEW_STACK_DEPTH */
-   { 19609, 0x00002100 }, /* GL_MODULATE */
-   { 19621, 0x00008744 }, /* GL_MODULATE_ADD_ATI */
-   { 19641, 0x00008745 }, /* GL_MODULATE_SIGNED_ADD_ATI */
-   { 19668, 0x00008746 }, /* GL_MODULATE_SUBTRACT_ATI */
-   { 19693, 0x00000103 }, /* GL_MULT */
-   { 19701, 0x0000809D }, /* GL_MULTISAMPLE */
-   { 19716, 0x000086B2 }, /* GL_MULTISAMPLE_3DFX */
-   { 19736, 0x0000809D }, /* GL_MULTISAMPLE_ARB */
-   { 19755, 0x20000000 }, /* GL_MULTISAMPLE_BIT */
-   { 19774, 0x20000000 }, /* GL_MULTISAMPLE_BIT_3DFX */
-   { 19798, 0x20000000 }, /* GL_MULTISAMPLE_BIT_ARB */
-   { 19821, 0x00008534 }, /* GL_MULTISAMPLE_FILTER_HINT_NV */
-   { 19851, 0x00002A25 }, /* GL_N3F_V3F */
-   { 19862, 0x00000D70 }, /* GL_NAME_STACK_DEPTH */
-   { 19882, 0x0000150E }, /* GL_NAND */
-   { 19890, 0x00002600 }, /* GL_NEAREST */
-   { 19901, 0x0000844E }, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
-   { 19932, 0x0000844D }, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
-   { 19964, 0x00002702 }, /* GL_NEAREST_MIPMAP_LINEAR */
-   { 19989, 0x00002700 }, /* GL_NEAREST_MIPMAP_NEAREST */
-   { 20015, 0x00000200 }, /* GL_NEVER */
-   { 20024, 0x00001102 }, /* GL_NICEST */
-   { 20034, 0x00000000 }, /* GL_NONE */
-   { 20042, 0x00001505 }, /* GL_NOOP */
-   { 20050, 0x00001508 }, /* GL_NOR */
-   { 20057, 0x00000BA1 }, /* GL_NORMALIZE */
-   { 20070, 0x00008075 }, /* GL_NORMAL_ARRAY */
-   { 20086, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
-   { 20117, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING_ARB */
-   { 20152, 0x0000808F }, /* GL_NORMAL_ARRAY_POINTER */
-   { 20176, 0x0000807F }, /* GL_NORMAL_ARRAY_STRIDE */
-   { 20199, 0x0000807E }, /* GL_NORMAL_ARRAY_TYPE */
-   { 20220, 0x00008511 }, /* GL_NORMAL_MAP */
-   { 20234, 0x00008511 }, /* GL_NORMAL_MAP_ARB */
-   { 20252, 0x00008511 }, /* GL_NORMAL_MAP_NV */
-   { 20269, 0x00000205 }, /* GL_NOTEQUAL */
-   { 20281, 0x00000000 }, /* GL_NO_ERROR */
-   { 20293, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
-   { 20327, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB */
-   { 20365, 0x00008B89 }, /* GL_OBJECT_ACTIVE_ATTRIBUTES_ARB */
-   { 20397, 0x00008B8A }, /* GL_OBJECT_ACTIVE_ATTRIBUTE_MAX_LENGTH_ARB */
-   { 20439, 0x00008B86 }, /* GL_OBJECT_ACTIVE_UNIFORMS_ARB */
-   { 20469, 0x00008B87 }, /* GL_OBJECT_ACTIVE_UNIFORM_MAX_LENGTH_ARB */
-   { 20509, 0x00008B85 }, /* GL_OBJECT_ATTACHED_OBJECTS_ARB */
-   { 20540, 0x00008B81 }, /* GL_OBJECT_COMPILE_STATUS_ARB */
-   { 20569, 0x00008B80 }, /* GL_OBJECT_DELETE_STATUS_ARB */
-   { 20597, 0x00008B84 }, /* GL_OBJECT_INFO_LOG_LENGTH_ARB */
-   { 20627, 0x00002401 }, /* GL_OBJECT_LINEAR */
-   { 20644, 0x00008B82 }, /* GL_OBJECT_LINK_STATUS_ARB */
-   { 20670, 0x00002501 }, /* GL_OBJECT_PLANE */
-   { 20686, 0x00008B88 }, /* GL_OBJECT_SHADER_SOURCE_LENGTH_ARB */
-   { 20721, 0x00008B4F }, /* GL_OBJECT_SUBTYPE_ARB */
-   { 20743, 0x00008B4E }, /* GL_OBJECT_TYPE_ARB */
-   { 20762, 0x00008B83 }, /* GL_OBJECT_VALIDATE_STATUS_ARB */
-   { 20792, 0x00008165 }, /* GL_OCCLUSION_TEST_HP */
-   { 20813, 0x00008166 }, /* GL_OCCLUSION_TEST_RESULT_HP */
-   { 20841, 0x00000001 }, /* GL_ONE */
-   { 20848, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA */
-   { 20876, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA_EXT */
-   { 20908, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR */
-   { 20936, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR_EXT */
-   { 20968, 0x00000305 }, /* GL_ONE_MINUS_DST_ALPHA */
-   { 20991, 0x00000307 }, /* GL_ONE_MINUS_DST_COLOR */
-   { 21014, 0x00000303 }, /* GL_ONE_MINUS_SRC_ALPHA */
-   { 21037, 0x00000301 }, /* GL_ONE_MINUS_SRC_COLOR */
-   { 21060, 0x00008598 }, /* GL_OPERAND0_ALPHA */
-   { 21078, 0x00008598 }, /* GL_OPERAND0_ALPHA_ARB */
-   { 21100, 0x00008598 }, /* GL_OPERAND0_ALPHA_EXT */
-   { 21122, 0x00008590 }, /* GL_OPERAND0_RGB */
-   { 21138, 0x00008590 }, /* GL_OPERAND0_RGB_ARB */
-   { 21158, 0x00008590 }, /* GL_OPERAND0_RGB_EXT */
-   { 21178, 0x00008599 }, /* GL_OPERAND1_ALPHA */
-   { 21196, 0x00008599 }, /* GL_OPERAND1_ALPHA_ARB */
-   { 21218, 0x00008599 }, /* GL_OPERAND1_ALPHA_EXT */
-   { 21240, 0x00008591 }, /* GL_OPERAND1_RGB */
-   { 21256, 0x00008591 }, /* GL_OPERAND1_RGB_ARB */
-   { 21276, 0x00008591 }, /* GL_OPERAND1_RGB_EXT */
-   { 21296, 0x0000859A }, /* GL_OPERAND2_ALPHA */
-   { 21314, 0x0000859A }, /* GL_OPERAND2_ALPHA_ARB */
-   { 21336, 0x0000859A }, /* GL_OPERAND2_ALPHA_EXT */
-   { 21358, 0x00008592 }, /* GL_OPERAND2_RGB */
-   { 21374, 0x00008592 }, /* GL_OPERAND2_RGB_ARB */
-   { 21394, 0x00008592 }, /* GL_OPERAND2_RGB_EXT */
-   { 21414, 0x0000859B }, /* GL_OPERAND3_ALPHA_NV */
-   { 21435, 0x00008593 }, /* GL_OPERAND3_RGB_NV */
-   { 21454, 0x00001507 }, /* GL_OR */
-   { 21460, 0x00000A01 }, /* GL_ORDER */
-   { 21469, 0x0000150D }, /* GL_OR_INVERTED */
-   { 21484, 0x0000150B }, /* GL_OR_REVERSE */
-   { 21498, 0x00000505 }, /* GL_OUT_OF_MEMORY */
-   { 21515, 0x00000D05 }, /* GL_PACK_ALIGNMENT */
-   { 21533, 0x0000806C }, /* GL_PACK_IMAGE_HEIGHT */
-   { 21554, 0x00008758 }, /* GL_PACK_INVERT_MESA */
-   { 21574, 0x00000D01 }, /* GL_PACK_LSB_FIRST */
-   { 21592, 0x00000D02 }, /* GL_PACK_ROW_LENGTH */
-   { 21611, 0x0000806B }, /* GL_PACK_SKIP_IMAGES */
-   { 21631, 0x00000D04 }, /* GL_PACK_SKIP_PIXELS */
-   { 21651, 0x00000D03 }, /* GL_PACK_SKIP_ROWS */
-   { 21669, 0x00000D00 }, /* GL_PACK_SWAP_BYTES */
-   { 21688, 0x00008B92 }, /* GL_PALETTE4_R5_G6_B5_OES */
-   { 21713, 0x00008B94 }, /* GL_PALETTE4_RGB5_A1_OES */
-   { 21737, 0x00008B90 }, /* GL_PALETTE4_RGB8_OES */
-   { 21758, 0x00008B93 }, /* GL_PALETTE4_RGBA4_OES */
-   { 21780, 0x00008B91 }, /* GL_PALETTE4_RGBA8_OES */
-   { 21802, 0x00008B97 }, /* GL_PALETTE8_R5_G6_B5_OES */
-   { 21827, 0x00008B99 }, /* GL_PALETTE8_RGB5_A1_OES */
-   { 21851, 0x00008B95 }, /* GL_PALETTE8_RGB8_OES */
-   { 21872, 0x00008B98 }, /* GL_PALETTE8_RGBA4_OES */
-   { 21894, 0x00008B96 }, /* GL_PALETTE8_RGBA8_OES */
-   { 21916, 0x00000700 }, /* GL_PASS_THROUGH_TOKEN */
-   { 21938, 0x00000C50 }, /* GL_PERSPECTIVE_CORRECTION_HINT */
-   { 21969, 0x00000C79 }, /* GL_PIXEL_MAP_A_TO_A */
-   { 21989, 0x00000CB9 }, /* GL_PIXEL_MAP_A_TO_A_SIZE */
-   { 22014, 0x00000C78 }, /* GL_PIXEL_MAP_B_TO_B */
-   { 22034, 0x00000CB8 }, /* GL_PIXEL_MAP_B_TO_B_SIZE */
-   { 22059, 0x00000C77 }, /* GL_PIXEL_MAP_G_TO_G */
-   { 22079, 0x00000CB7 }, /* GL_PIXEL_MAP_G_TO_G_SIZE */
-   { 22104, 0x00000C75 }, /* GL_PIXEL_MAP_I_TO_A */
-   { 22124, 0x00000CB5 }, /* GL_PIXEL_MAP_I_TO_A_SIZE */
-   { 22149, 0x00000C74 }, /* GL_PIXEL_MAP_I_TO_B */
-   { 22169, 0x00000CB4 }, /* GL_PIXEL_MAP_I_TO_B_SIZE */
-   { 22194, 0x00000C73 }, /* GL_PIXEL_MAP_I_TO_G */
-   { 22214, 0x00000CB3 }, /* GL_PIXEL_MAP_I_TO_G_SIZE */
-   { 22239, 0x00000C70 }, /* GL_PIXEL_MAP_I_TO_I */
-   { 22259, 0x00000CB0 }, /* GL_PIXEL_MAP_I_TO_I_SIZE */
-   { 22284, 0x00000C72 }, /* GL_PIXEL_MAP_I_TO_R */
-   { 22304, 0x00000CB2 }, /* GL_PIXEL_MAP_I_TO_R_SIZE */
-   { 22329, 0x00000C76 }, /* GL_PIXEL_MAP_R_TO_R */
-   { 22349, 0x00000CB6 }, /* GL_PIXEL_MAP_R_TO_R_SIZE */
-   { 22374, 0x00000C71 }, /* GL_PIXEL_MAP_S_TO_S */
-   { 22394, 0x00000CB1 }, /* GL_PIXEL_MAP_S_TO_S_SIZE */
-   { 22419, 0x00000020 }, /* GL_PIXEL_MODE_BIT */
-   { 22437, 0x000088ED }, /* GL_PIXEL_PACK_BUFFER_BINDING_EXT */
-   { 22470, 0x000088EB }, /* GL_PIXEL_PACK_BUFFER_EXT */
-   { 22495, 0x000088EF }, /* GL_PIXEL_UNPACK_BUFFER_BINDING_EXT */
-   { 22530, 0x000088EC }, /* GL_PIXEL_UNPACK_BUFFER_EXT */
-   { 22557, 0x00001B00 }, /* GL_POINT */
-   { 22566, 0x00000000 }, /* GL_POINTS */
-   { 22576, 0x00000002 }, /* GL_POINT_BIT */
-   { 22589, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION */
-   { 22619, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_ARB */
-   { 22653, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_EXT */
-   { 22687, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_SGIS */
-   { 22722, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE */
-   { 22751, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_ARB */
-   { 22784, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_EXT */
-   { 22817, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_SGIS */
-   { 22851, 0x00000B11 }, /* GL_POINT_SIZE */
-   { 22865, 0x00000B13 }, /* GL_POINT_SIZE_GRANULARITY */
-   { 22891, 0x00008127 }, /* GL_POINT_SIZE_MAX */
-   { 22909, 0x00008127 }, /* GL_POINT_SIZE_MAX_ARB */
-   { 22931, 0x00008127 }, /* GL_POINT_SIZE_MAX_EXT */
-   { 22953, 0x00008127 }, /* GL_POINT_SIZE_MAX_SGIS */
-   { 22976, 0x00008126 }, /* GL_POINT_SIZE_MIN */
-   { 22994, 0x00008126 }, /* GL_POINT_SIZE_MIN_ARB */
-   { 23016, 0x00008126 }, /* GL_POINT_SIZE_MIN_EXT */
-   { 23038, 0x00008126 }, /* GL_POINT_SIZE_MIN_SGIS */
-   { 23061, 0x00000B12 }, /* GL_POINT_SIZE_RANGE */
-   { 23081, 0x00000B10 }, /* GL_POINT_SMOOTH */
-   { 23097, 0x00000C51 }, /* GL_POINT_SMOOTH_HINT */
-   { 23118, 0x00008861 }, /* GL_POINT_SPRITE */
-   { 23134, 0x00008861 }, /* GL_POINT_SPRITE_ARB */
-   { 23154, 0x00008CA0 }, /* GL_POINT_SPRITE_COORD_ORIGIN */
-   { 23183, 0x00008861 }, /* GL_POINT_SPRITE_NV */
-   { 23202, 0x00008863 }, /* GL_POINT_SPRITE_R_MODE_NV */
-   { 23228, 0x00000701 }, /* GL_POINT_TOKEN */
-   { 23243, 0x00000009 }, /* GL_POLYGON */
-   { 23254, 0x00000008 }, /* GL_POLYGON_BIT */
-   { 23269, 0x00000B40 }, /* GL_POLYGON_MODE */
-   { 23285, 0x00008039 }, /* GL_POLYGON_OFFSET_BIAS */
-   { 23308, 0x00008038 }, /* GL_POLYGON_OFFSET_FACTOR */
-   { 23333, 0x00008037 }, /* GL_POLYGON_OFFSET_FILL */
-   { 23356, 0x00002A02 }, /* GL_POLYGON_OFFSET_LINE */
-   { 23379, 0x00002A01 }, /* GL_POLYGON_OFFSET_POINT */
-   { 23403, 0x00002A00 }, /* GL_POLYGON_OFFSET_UNITS */
-   { 23427, 0x00000B41 }, /* GL_POLYGON_SMOOTH */
-   { 23445, 0x00000C53 }, /* GL_POLYGON_SMOOTH_HINT */
-   { 23468, 0x00000B42 }, /* GL_POLYGON_STIPPLE */
-   { 23487, 0x00000010 }, /* GL_POLYGON_STIPPLE_BIT */
-   { 23510, 0x00000703 }, /* GL_POLYGON_TOKEN */
-   { 23527, 0x00001203 }, /* GL_POSITION */
-   { 23539, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
-   { 23571, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS_SGI */
-   { 23607, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
-   { 23640, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE_SGI */
-   { 23677, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
-   { 23708, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS_SGI */
-   { 23743, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
-   { 23775, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE_SGI */
-   { 23811, 0x000080D2 }, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
-   { 23844, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
-   { 23876, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS_SGI */
-   { 23912, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
-   { 23945, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE_SGI */
-   { 23982, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS */
-   { 24012, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS_SGI */
-   { 24046, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE */
-   { 24077, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE_SGI */
-   { 24112, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
-   { 24143, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS_EXT */
-   { 24178, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
-   { 24210, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE_EXT */
-   { 24246, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS */
-   { 24276, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS_EXT */
-   { 24310, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE */
-   { 24341, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE_EXT */
-   { 24376, 0x000080D1 }, /* GL_POST_CONVOLUTION_COLOR_TABLE */
-   { 24408, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS */
-   { 24439, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS_EXT */
-   { 24474, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE */
-   { 24506, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE_EXT */
-   { 24542, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS */
-   { 24571, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS_EXT */
-   { 24604, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE */
-   { 24634, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE_EXT */
-   { 24668, 0x0000817B }, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
-   { 24707, 0x00008179 }, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
-   { 24740, 0x0000817C }, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
-   { 24780, 0x0000817A }, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
-   { 24814, 0x00008578 }, /* GL_PREVIOUS */
-   { 24826, 0x00008578 }, /* GL_PREVIOUS_ARB */
-   { 24842, 0x00008578 }, /* GL_PREVIOUS_EXT */
-   { 24858, 0x00008577 }, /* GL_PRIMARY_COLOR */
-   { 24875, 0x00008577 }, /* GL_PRIMARY_COLOR_ARB */
-   { 24896, 0x00008577 }, /* GL_PRIMARY_COLOR_EXT */
-   { 24917, 0x000088B0 }, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
-   { 24950, 0x00008805 }, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
-   { 24982, 0x000088AC }, /* GL_PROGRAM_ATTRIBS_ARB */
-   { 25005, 0x00008677 }, /* GL_PROGRAM_BINDING_ARB */
-   { 25028, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_ARB */
-   { 25058, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_NV */
-   { 25087, 0x00008874 }, /* GL_PROGRAM_ERROR_STRING_ARB */
-   { 25115, 0x00008876 }, /* GL_PROGRAM_FORMAT_ARB */
-   { 25137, 0x00008875 }, /* GL_PROGRAM_FORMAT_ASCII_ARB */
-   { 25165, 0x000088A0 }, /* GL_PROGRAM_INSTRUCTIONS_ARB */
-   { 25193, 0x00008627 }, /* GL_PROGRAM_LENGTH_ARB */
-   { 25215, 0x00008627 }, /* GL_PROGRAM_LENGTH_NV */
-   { 25236, 0x000088B2 }, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-   { 25276, 0x00008808 }, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-   { 25315, 0x000088AE }, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
-   { 25345, 0x000088A2 }, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-   { 25380, 0x000088AA }, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
-   { 25413, 0x000088A6 }, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
-   { 25447, 0x0000880A }, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-   { 25486, 0x00008809 }, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-   { 25525, 0x00008B40 }, /* GL_PROGRAM_OBJECT_ARB */
-   { 25547, 0x000088A8 }, /* GL_PROGRAM_PARAMETERS_ARB */
-   { 25573, 0x00008644 }, /* GL_PROGRAM_PARAMETER_NV */
-   { 25597, 0x00008647 }, /* GL_PROGRAM_RESIDENT_NV */
-   { 25620, 0x00008628 }, /* GL_PROGRAM_STRING_ARB */
-   { 25642, 0x00008628 }, /* GL_PROGRAM_STRING_NV */
-   { 25663, 0x00008646 }, /* GL_PROGRAM_TARGET_NV */
-   { 25684, 0x000088A4 }, /* GL_PROGRAM_TEMPORARIES_ARB */
-   { 25711, 0x00008807 }, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
-   { 25743, 0x00008806 }, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
-   { 25775, 0x000088B6 }, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
-   { 25810, 0x00001701 }, /* GL_PROJECTION */
-   { 25824, 0x00000BA7 }, /* GL_PROJECTION_MATRIX */
-   { 25845, 0x00000BA4 }, /* GL_PROJECTION_STACK_DEPTH */
-   { 25871, 0x000080D3 }, /* GL_PROXY_COLOR_TABLE */
-   { 25892, 0x00008025 }, /* GL_PROXY_HISTOGRAM */
-   { 25911, 0x00008025 }, /* GL_PROXY_HISTOGRAM_EXT */
-   { 25934, 0x000080D5 }, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
-   { 25973, 0x000080D4 }, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
-   { 26011, 0x00008063 }, /* GL_PROXY_TEXTURE_1D */
-   { 26031, 0x00008C19 }, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
-   { 26061, 0x00008063 }, /* GL_PROXY_TEXTURE_1D_EXT */
-   { 26085, 0x00008064 }, /* GL_PROXY_TEXTURE_2D */
-   { 26105, 0x00008C1B }, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
-   { 26135, 0x00008064 }, /* GL_PROXY_TEXTURE_2D_EXT */
-   { 26159, 0x00008070 }, /* GL_PROXY_TEXTURE_3D */
-   { 26179, 0x000080BD }, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
-   { 26212, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP */
-   { 26238, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP_ARB */
-   { 26268, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
-   { 26299, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_NV */
-   { 26329, 0x00002003 }, /* GL_Q */
-   { 26334, 0x00001209 }, /* GL_QUADRATIC_ATTENUATION */
-   { 26359, 0x00000007 }, /* GL_QUADS */
-   { 26368, 0x00008614 }, /* GL_QUAD_MESH_SUN */
-   { 26385, 0x00000008 }, /* GL_QUAD_STRIP */
-   { 26399, 0x00008864 }, /* GL_QUERY_COUNTER_BITS */
-   { 26421, 0x00008864 }, /* GL_QUERY_COUNTER_BITS_ARB */
-   { 26447, 0x00008866 }, /* GL_QUERY_RESULT */
-   { 26463, 0x00008866 }, /* GL_QUERY_RESULT_ARB */
-   { 26483, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE */
-   { 26509, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE_ARB */
-   { 26539, 0x00002002 }, /* GL_R */
-   { 26544, 0x00002A10 }, /* GL_R3_G3_B2 */
-   { 26556, 0x00019262 }, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
-   { 26589, 0x00000C02 }, /* GL_READ_BUFFER */
-   { 26604, 0x00008CAA }, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
-   { 26636, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER_EXT */
-   { 26660, 0x000088B8 }, /* GL_READ_ONLY */
-   { 26673, 0x000088B8 }, /* GL_READ_ONLY_ARB */
-   { 26690, 0x000088BA }, /* GL_READ_WRITE */
-   { 26704, 0x000088BA }, /* GL_READ_WRITE_ARB */
-   { 26722, 0x00001903 }, /* GL_RED */
-   { 26729, 0x00008016 }, /* GL_REDUCE */
-   { 26739, 0x00008016 }, /* GL_REDUCE_EXT */
-   { 26753, 0x00000D15 }, /* GL_RED_BIAS */
-   { 26765, 0x00000D52 }, /* GL_RED_BITS */
-   { 26777, 0x00000D14 }, /* GL_RED_SCALE */
-   { 26790, 0x00008512 }, /* GL_REFLECTION_MAP */
-   { 26808, 0x00008512 }, /* GL_REFLECTION_MAP_ARB */
-   { 26830, 0x00008512 }, /* GL_REFLECTION_MAP_NV */
-   { 26851, 0x00001C00 }, /* GL_RENDER */
-   { 26861, 0x00008CA7 }, /* GL_RENDERBUFFER_BINDING_EXT */
-   { 26889, 0x00008D41 }, /* GL_RENDERBUFFER_EXT */
-   { 26909, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT_EXT */
-   { 26936, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT_EXT */
-   { 26972, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH_EXT */
-   { 26998, 0x00001F01 }, /* GL_RENDERER */
-   { 27010, 0x00000C40 }, /* GL_RENDER_MODE */
-   { 27025, 0x00002901 }, /* GL_REPEAT */
-   { 27035, 0x00001E01 }, /* GL_REPLACE */
-   { 27046, 0x00008062 }, /* GL_REPLACE_EXT */
-   { 27061, 0x00008153 }, /* GL_REPLICATE_BORDER_HP */
-   { 27084, 0x0000803A }, /* GL_RESCALE_NORMAL */
-   { 27102, 0x0000803A }, /* GL_RESCALE_NORMAL_EXT */
-   { 27124, 0x00000102 }, /* GL_RETURN */
-   { 27134, 0x00001907 }, /* GL_RGB */
-   { 27141, 0x00008052 }, /* GL_RGB10 */
-   { 27150, 0x00008059 }, /* GL_RGB10_A2 */
-   { 27162, 0x00008059 }, /* GL_RGB10_A2_EXT */
-   { 27178, 0x00008052 }, /* GL_RGB10_EXT */
-   { 27191, 0x00008053 }, /* GL_RGB12 */
-   { 27200, 0x00008053 }, /* GL_RGB12_EXT */
-   { 27213, 0x00008054 }, /* GL_RGB16 */
-   { 27222, 0x00008054 }, /* GL_RGB16_EXT */
-   { 27235, 0x0000804E }, /* GL_RGB2_EXT */
-   { 27247, 0x0000804F }, /* GL_RGB4 */
-   { 27255, 0x0000804F }, /* GL_RGB4_EXT */
-   { 27267, 0x000083A1 }, /* GL_RGB4_S3TC */
-   { 27280, 0x00008050 }, /* GL_RGB5 */
-   { 27288, 0x00008057 }, /* GL_RGB5_A1 */
-   { 27299, 0x00008057 }, /* GL_RGB5_A1_EXT */
-   { 27314, 0x00008050 }, /* GL_RGB5_EXT */
-   { 27326, 0x00008051 }, /* GL_RGB8 */
-   { 27334, 0x00008051 }, /* GL_RGB8_EXT */
-   { 27346, 0x00001908 }, /* GL_RGBA */
-   { 27354, 0x0000805A }, /* GL_RGBA12 */
-   { 27364, 0x0000805A }, /* GL_RGBA12_EXT */
-   { 27378, 0x0000805B }, /* GL_RGBA16 */
-   { 27388, 0x0000805B }, /* GL_RGBA16_EXT */
-   { 27402, 0x00008055 }, /* GL_RGBA2 */
-   { 27411, 0x00008055 }, /* GL_RGBA2_EXT */
-   { 27424, 0x00008056 }, /* GL_RGBA4 */
-   { 27433, 0x000083A5 }, /* GL_RGBA4_DXT5_S3TC */
-   { 27452, 0x00008056 }, /* GL_RGBA4_EXT */
-   { 27465, 0x000083A3 }, /* GL_RGBA4_S3TC */
-   { 27479, 0x00008058 }, /* GL_RGBA8 */
-   { 27488, 0x00008058 }, /* GL_RGBA8_EXT */
-   { 27501, 0x000083A4 }, /* GL_RGBA_DXT5_S3TC */
-   { 27519, 0x00000C31 }, /* GL_RGBA_MODE */
-   { 27532, 0x000083A2 }, /* GL_RGBA_S3TC */
-   { 27545, 0x000083A0 }, /* GL_RGB_S3TC */
-   { 27557, 0x00008573 }, /* GL_RGB_SCALE */
-   { 27570, 0x00008573 }, /* GL_RGB_SCALE_ARB */
-   { 27587, 0x00008573 }, /* GL_RGB_SCALE_EXT */
-   { 27604, 0x00000407 }, /* GL_RIGHT */
-   { 27613, 0x00002000 }, /* GL_S */
-   { 27618, 0x00008B5D }, /* GL_SAMPLER_1D */
-   { 27632, 0x00008B61 }, /* GL_SAMPLER_1D_SHADOW */
-   { 27653, 0x00008B5E }, /* GL_SAMPLER_2D */
-   { 27667, 0x00008B62 }, /* GL_SAMPLER_2D_SHADOW */
-   { 27688, 0x00008B5F }, /* GL_SAMPLER_3D */
-   { 27702, 0x00008B60 }, /* GL_SAMPLER_CUBE */
-   { 27718, 0x000080A9 }, /* GL_SAMPLES */
-   { 27729, 0x000086B4 }, /* GL_SAMPLES_3DFX */
-   { 27745, 0x000080A9 }, /* GL_SAMPLES_ARB */
-   { 27760, 0x00008914 }, /* GL_SAMPLES_PASSED */
-   { 27778, 0x00008914 }, /* GL_SAMPLES_PASSED_ARB */
-   { 27800, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
-   { 27828, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE_ARB */
-   { 27860, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE */
-   { 27883, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE_ARB */
-   { 27910, 0x000080A8 }, /* GL_SAMPLE_BUFFERS */
-   { 27928, 0x000086B3 }, /* GL_SAMPLE_BUFFERS_3DFX */
-   { 27951, 0x000080A8 }, /* GL_SAMPLE_BUFFERS_ARB */
-   { 27973, 0x000080A0 }, /* GL_SAMPLE_COVERAGE */
-   { 27992, 0x000080A0 }, /* GL_SAMPLE_COVERAGE_ARB */
-   { 28015, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT */
-   { 28041, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT_ARB */
-   { 28071, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE */
-   { 28096, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE_ARB */
-   { 28125, 0x00080000 }, /* GL_SCISSOR_BIT */
-   { 28140, 0x00000C10 }, /* GL_SCISSOR_BOX */
-   { 28155, 0x00000C11 }, /* GL_SCISSOR_TEST */
-   { 28171, 0x0000845E }, /* GL_SECONDARY_COLOR_ARRAY */
-   { 28196, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
-   { 28236, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB */
-   { 28280, 0x0000845D }, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
-   { 28313, 0x0000845A }, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
-   { 28343, 0x0000845C }, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
-   { 28375, 0x0000845B }, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
-   { 28405, 0x00001C02 }, /* GL_SELECT */
-   { 28415, 0x00000DF3 }, /* GL_SELECTION_BUFFER_POINTER */
-   { 28443, 0x00000DF4 }, /* GL_SELECTION_BUFFER_SIZE */
-   { 28468, 0x00008012 }, /* GL_SEPARABLE_2D */
-   { 28484, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR */
-   { 28511, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR_EXT */
-   { 28542, 0x0000150F }, /* GL_SET */
-   { 28549, 0x00008B48 }, /* GL_SHADER_OBJECT_ARB */
-   { 28570, 0x00008B88 }, /* GL_SHADER_SOURCE_LENGTH */
-   { 28594, 0x00008B4F }, /* GL_SHADER_TYPE */
-   { 28609, 0x00000B54 }, /* GL_SHADE_MODEL */
-   { 28624, 0x00008B8C }, /* GL_SHADING_LANGUAGE_VERSION */
-   { 28652, 0x000080BF }, /* GL_SHADOW_AMBIENT_SGIX */
-   { 28675, 0x000081FB }, /* GL_SHARED_TEXTURE_PALETTE_EXT */
-   { 28705, 0x00001601 }, /* GL_SHININESS */
-   { 28718, 0x00001402 }, /* GL_SHORT */
-   { 28727, 0x000081F9 }, /* GL_SINGLE_COLOR */
-   { 28743, 0x000081F9 }, /* GL_SINGLE_COLOR_EXT */
-   { 28763, 0x000085CC }, /* GL_SLICE_ACCUM_SUN */
-   { 28782, 0x00001D01 }, /* GL_SMOOTH */
-   { 28792, 0x00000B23 }, /* GL_SMOOTH_LINE_WIDTH_GRANULARITY */
-   { 28825, 0x00000B22 }, /* GL_SMOOTH_LINE_WIDTH_RANGE */
-   { 28852, 0x00000B13 }, /* GL_SMOOTH_POINT_SIZE_GRANULARITY */
-   { 28885, 0x00000B12 }, /* GL_SMOOTH_POINT_SIZE_RANGE */
-   { 28912, 0x00008588 }, /* GL_SOURCE0_ALPHA */
-   { 28929, 0x00008588 }, /* GL_SOURCE0_ALPHA_ARB */
-   { 28950, 0x00008588 }, /* GL_SOURCE0_ALPHA_EXT */
-   { 28971, 0x00008580 }, /* GL_SOURCE0_RGB */
-   { 28986, 0x00008580 }, /* GL_SOURCE0_RGB_ARB */
-   { 29005, 0x00008580 }, /* GL_SOURCE0_RGB_EXT */
-   { 29024, 0x00008589 }, /* GL_SOURCE1_ALPHA */
-   { 29041, 0x00008589 }, /* GL_SOURCE1_ALPHA_ARB */
-   { 29062, 0x00008589 }, /* GL_SOURCE1_ALPHA_EXT */
-   { 29083, 0x00008581 }, /* GL_SOURCE1_RGB */
-   { 29098, 0x00008581 }, /* GL_SOURCE1_RGB_ARB */
-   { 29117, 0x00008581 }, /* GL_SOURCE1_RGB_EXT */
-   { 29136, 0x0000858A }, /* GL_SOURCE2_ALPHA */
-   { 29153, 0x0000858A }, /* GL_SOURCE2_ALPHA_ARB */
-   { 29174, 0x0000858A }, /* GL_SOURCE2_ALPHA_EXT */
-   { 29195, 0x00008582 }, /* GL_SOURCE2_RGB */
-   { 29210, 0x00008582 }, /* GL_SOURCE2_RGB_ARB */
-   { 29229, 0x00008582 }, /* GL_SOURCE2_RGB_EXT */
-   { 29248, 0x0000858B }, /* GL_SOURCE3_ALPHA_NV */
-   { 29268, 0x00008583 }, /* GL_SOURCE3_RGB_NV */
-   { 29286, 0x00001202 }, /* GL_SPECULAR */
-   { 29298, 0x00002402 }, /* GL_SPHERE_MAP */
-   { 29312, 0x00001206 }, /* GL_SPOT_CUTOFF */
-   { 29327, 0x00001204 }, /* GL_SPOT_DIRECTION */
-   { 29345, 0x00001205 }, /* GL_SPOT_EXPONENT */
-   { 29362, 0x00008588 }, /* GL_SRC0_ALPHA */
-   { 29376, 0x00008580 }, /* GL_SRC0_RGB */
-   { 29388, 0x00008589 }, /* GL_SRC1_ALPHA */
-   { 29402, 0x00008581 }, /* GL_SRC1_RGB */
-   { 29414, 0x0000858A }, /* GL_SRC2_ALPHA */
-   { 29428, 0x00008582 }, /* GL_SRC2_RGB */
-   { 29440, 0x00000302 }, /* GL_SRC_ALPHA */
-   { 29453, 0x00000308 }, /* GL_SRC_ALPHA_SATURATE */
-   { 29475, 0x00000300 }, /* GL_SRC_COLOR */
-   { 29488, 0x00000503 }, /* GL_STACK_OVERFLOW */
-   { 29506, 0x00000504 }, /* GL_STACK_UNDERFLOW */
-   { 29525, 0x000088E6 }, /* GL_STATIC_COPY */
-   { 29540, 0x000088E6 }, /* GL_STATIC_COPY_ARB */
-   { 29559, 0x000088E4 }, /* GL_STATIC_DRAW */
-   { 29574, 0x000088E4 }, /* GL_STATIC_DRAW_ARB */
-   { 29593, 0x000088E5 }, /* GL_STATIC_READ */
-   { 29608, 0x000088E5 }, /* GL_STATIC_READ_ARB */
-   { 29627, 0x00001802 }, /* GL_STENCIL */
-   { 29638, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT_EXT */
-   { 29664, 0x00008801 }, /* GL_STENCIL_BACK_FAIL */
-   { 29685, 0x00008801 }, /* GL_STENCIL_BACK_FAIL_ATI */
-   { 29710, 0x00008800 }, /* GL_STENCIL_BACK_FUNC */
-   { 29731, 0x00008800 }, /* GL_STENCIL_BACK_FUNC_ATI */
-   { 29756, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
-   { 29788, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL_ATI */
-   { 29824, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
-   { 29856, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS_ATI */
-   { 29892, 0x00008CA3 }, /* GL_STENCIL_BACK_REF */
-   { 29912, 0x00008CA4 }, /* GL_STENCIL_BACK_VALUE_MASK */
-   { 29939, 0x00008CA5 }, /* GL_STENCIL_BACK_WRITEMASK */
-   { 29965, 0x00000D57 }, /* GL_STENCIL_BITS */
-   { 29981, 0x00000400 }, /* GL_STENCIL_BUFFER_BIT */
-   { 30003, 0x00000B91 }, /* GL_STENCIL_CLEAR_VALUE */
-   { 30026, 0x00000B94 }, /* GL_STENCIL_FAIL */
-   { 30042, 0x00000B92 }, /* GL_STENCIL_FUNC */
-   { 30058, 0x00001901 }, /* GL_STENCIL_INDEX */
-   { 30075, 0x00008D49 }, /* GL_STENCIL_INDEX16_EXT */
-   { 30098, 0x00008D46 }, /* GL_STENCIL_INDEX1_EXT */
-   { 30120, 0x00008D47 }, /* GL_STENCIL_INDEX4_EXT */
-   { 30142, 0x00008D48 }, /* GL_STENCIL_INDEX8_EXT */
-   { 30164, 0x00008D45 }, /* GL_STENCIL_INDEX_EXT */
-   { 30185, 0x00000B95 }, /* GL_STENCIL_PASS_DEPTH_FAIL */
-   { 30212, 0x00000B96 }, /* GL_STENCIL_PASS_DEPTH_PASS */
-   { 30239, 0x00000B97 }, /* GL_STENCIL_REF */
-   { 30254, 0x00000B90 }, /* GL_STENCIL_TEST */
-   { 30270, 0x00008910 }, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
-   { 30299, 0x00000B93 }, /* GL_STENCIL_VALUE_MASK */
-   { 30321, 0x00000B98 }, /* GL_STENCIL_WRITEMASK */
-   { 30342, 0x00000C33 }, /* GL_STEREO */
-   { 30352, 0x000088E2 }, /* GL_STREAM_COPY */
-   { 30367, 0x000088E2 }, /* GL_STREAM_COPY_ARB */
-   { 30386, 0x000088E0 }, /* GL_STREAM_DRAW */
-   { 30401, 0x000088E0 }, /* GL_STREAM_DRAW_ARB */
-   { 30420, 0x000088E1 }, /* GL_STREAM_READ */
-   { 30435, 0x000088E1 }, /* GL_STREAM_READ_ARB */
-   { 30454, 0x00000D50 }, /* GL_SUBPIXEL_BITS */
-   { 30471, 0x000084E7 }, /* GL_SUBTRACT */
-   { 30483, 0x000084E7 }, /* GL_SUBTRACT_ARB */
-   { 30499, 0x00002001 }, /* GL_T */
-   { 30504, 0x00002A2A }, /* GL_T2F_C3F_V3F */
-   { 30519, 0x00002A2C }, /* GL_T2F_C4F_N3F_V3F */
-   { 30538, 0x00002A29 }, /* GL_T2F_C4UB_V3F */
-   { 30554, 0x00002A2B }, /* GL_T2F_N3F_V3F */
-   { 30569, 0x00002A27 }, /* GL_T2F_V3F */
-   { 30580, 0x00002A2D }, /* GL_T4F_C4F_N3F_V4F */
-   { 30599, 0x00002A28 }, /* GL_T4F_V4F */
-   { 30610, 0x00008031 }, /* GL_TABLE_TOO_LARGE_EXT */
-   { 30633, 0x00001702 }, /* GL_TEXTURE */
-   { 30644, 0x000084C0 }, /* GL_TEXTURE0 */
-   { 30656, 0x000084C0 }, /* GL_TEXTURE0_ARB */
-   { 30672, 0x000084C1 }, /* GL_TEXTURE1 */
-   { 30684, 0x000084CA }, /* GL_TEXTURE10 */
-   { 30697, 0x000084CA }, /* GL_TEXTURE10_ARB */
-   { 30714, 0x000084CB }, /* GL_TEXTURE11 */
-   { 30727, 0x000084CB }, /* GL_TEXTURE11_ARB */
-   { 30744, 0x000084CC }, /* GL_TEXTURE12 */
-   { 30757, 0x000084CC }, /* GL_TEXTURE12_ARB */
-   { 30774, 0x000084CD }, /* GL_TEXTURE13 */
-   { 30787, 0x000084CD }, /* GL_TEXTURE13_ARB */
-   { 30804, 0x000084CE }, /* GL_TEXTURE14 */
-   { 30817, 0x000084CE }, /* GL_TEXTURE14_ARB */
-   { 30834, 0x000084CF }, /* GL_TEXTURE15 */
-   { 30847, 0x000084CF }, /* GL_TEXTURE15_ARB */
-   { 30864, 0x000084D0 }, /* GL_TEXTURE16 */
-   { 30877, 0x000084D0 }, /* GL_TEXTURE16_ARB */
-   { 30894, 0x000084D1 }, /* GL_TEXTURE17 */
-   { 30907, 0x000084D1 }, /* GL_TEXTURE17_ARB */
-   { 30924, 0x000084D2 }, /* GL_TEXTURE18 */
-   { 30937, 0x000084D2 }, /* GL_TEXTURE18_ARB */
-   { 30954, 0x000084D3 }, /* GL_TEXTURE19 */
-   { 30967, 0x000084D3 }, /* GL_TEXTURE19_ARB */
-   { 30984, 0x000084C1 }, /* GL_TEXTURE1_ARB */
-   { 31000, 0x000084C2 }, /* GL_TEXTURE2 */
-   { 31012, 0x000084D4 }, /* GL_TEXTURE20 */
-   { 31025, 0x000084D4 }, /* GL_TEXTURE20_ARB */
-   { 31042, 0x000084D5 }, /* GL_TEXTURE21 */
-   { 31055, 0x000084D5 }, /* GL_TEXTURE21_ARB */
-   { 31072, 0x000084D6 }, /* GL_TEXTURE22 */
-   { 31085, 0x000084D6 }, /* GL_TEXTURE22_ARB */
-   { 31102, 0x000084D7 }, /* GL_TEXTURE23 */
-   { 31115, 0x000084D7 }, /* GL_TEXTURE23_ARB */
-   { 31132, 0x000084D8 }, /* GL_TEXTURE24 */
-   { 31145, 0x000084D8 }, /* GL_TEXTURE24_ARB */
-   { 31162, 0x000084D9 }, /* GL_TEXTURE25 */
-   { 31175, 0x000084D9 }, /* GL_TEXTURE25_ARB */
-   { 31192, 0x000084DA }, /* GL_TEXTURE26 */
-   { 31205, 0x000084DA }, /* GL_TEXTURE26_ARB */
-   { 31222, 0x000084DB }, /* GL_TEXTURE27 */
-   { 31235, 0x000084DB }, /* GL_TEXTURE27_ARB */
-   { 31252, 0x000084DC }, /* GL_TEXTURE28 */
-   { 31265, 0x000084DC }, /* GL_TEXTURE28_ARB */
-   { 31282, 0x000084DD }, /* GL_TEXTURE29 */
-   { 31295, 0x000084DD }, /* GL_TEXTURE29_ARB */
-   { 31312, 0x000084C2 }, /* GL_TEXTURE2_ARB */
-   { 31328, 0x000084C3 }, /* GL_TEXTURE3 */
-   { 31340, 0x000084DE }, /* GL_TEXTURE30 */
-   { 31353, 0x000084DE }, /* GL_TEXTURE30_ARB */
-   { 31370, 0x000084DF }, /* GL_TEXTURE31 */
-   { 31383, 0x000084DF }, /* GL_TEXTURE31_ARB */
-   { 31400, 0x000084C3 }, /* GL_TEXTURE3_ARB */
-   { 31416, 0x000084C4 }, /* GL_TEXTURE4 */
-   { 31428, 0x000084C4 }, /* GL_TEXTURE4_ARB */
-   { 31444, 0x000084C5 }, /* GL_TEXTURE5 */
-   { 31456, 0x000084C5 }, /* GL_TEXTURE5_ARB */
-   { 31472, 0x000084C6 }, /* GL_TEXTURE6 */
-   { 31484, 0x000084C6 }, /* GL_TEXTURE6_ARB */
-   { 31500, 0x000084C7 }, /* GL_TEXTURE7 */
-   { 31512, 0x000084C7 }, /* GL_TEXTURE7_ARB */
-   { 31528, 0x000084C8 }, /* GL_TEXTURE8 */
-   { 31540, 0x000084C8 }, /* GL_TEXTURE8_ARB */
-   { 31556, 0x000084C9 }, /* GL_TEXTURE9 */
-   { 31568, 0x000084C9 }, /* GL_TEXTURE9_ARB */
-   { 31584, 0x00000DE0 }, /* GL_TEXTURE_1D */
-   { 31598, 0x00008C18 }, /* GL_TEXTURE_1D_ARRAY_EXT */
-   { 31622, 0x00000DE1 }, /* GL_TEXTURE_2D */
-   { 31636, 0x00008C1A }, /* GL_TEXTURE_2D_ARRAY_EXT */
-   { 31660, 0x0000806F }, /* GL_TEXTURE_3D */
-   { 31674, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE */
-   { 31696, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE_EXT */
-   { 31722, 0x0000813C }, /* GL_TEXTURE_BASE_LEVEL */
-   { 31744, 0x00008068 }, /* GL_TEXTURE_BINDING_1D */
-   { 31766, 0x00008C1C }, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
-   { 31798, 0x00008069 }, /* GL_TEXTURE_BINDING_2D */
-   { 31820, 0x00008C1D }, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
-   { 31852, 0x0000806A }, /* GL_TEXTURE_BINDING_3D */
-   { 31874, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP */
-   { 31902, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP_ARB */
-   { 31934, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
-   { 31967, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_NV */
-   { 31999, 0x00040000 }, /* GL_TEXTURE_BIT */
-   { 32014, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE */
-   { 32035, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE_EXT */
-   { 32060, 0x00001005 }, /* GL_TEXTURE_BORDER */
-   { 32078, 0x00001004 }, /* GL_TEXTURE_BORDER_COLOR */
-   { 32102, 0x00008171 }, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
-   { 32133, 0x00008176 }, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
-   { 32163, 0x00008172 }, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
-   { 32193, 0x00008175 }, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
-   { 32228, 0x00008173 }, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
-   { 32259, 0x00008174 }, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-   { 32297, 0x000080BC }, /* GL_TEXTURE_COLOR_TABLE_SGI */
-   { 32324, 0x000081EF }, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
-   { 32356, 0x000080BF }, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
-   { 32390, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC */
-   { 32414, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC_ARB */
-   { 32442, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE */
-   { 32466, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE_ARB */
-   { 32494, 0x0000819B }, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
-   { 32527, 0x0000819A }, /* GL_TEXTURE_COMPARE_SGIX */
-   { 32551, 0x00001003 }, /* GL_TEXTURE_COMPONENTS */
-   { 32573, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED */
-   { 32595, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED_ARB */
-   { 32621, 0x000086A3 }, /* GL_TEXTURE_COMPRESSED_FORMATS_ARB */
-   { 32655, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
-   { 32688, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE_ARB */
-   { 32725, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT */
-   { 32753, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT_ARB */
-   { 32785, 0x00008078 }, /* GL_TEXTURE_COORD_ARRAY */
-   { 32808, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
-   { 32846, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB */
-   { 32888, 0x00008092 }, /* GL_TEXTURE_COORD_ARRAY_POINTER */
-   { 32919, 0x00008088 }, /* GL_TEXTURE_COORD_ARRAY_SIZE */
-   { 32947, 0x0000808A }, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
-   { 32977, 0x00008089 }, /* GL_TEXTURE_COORD_ARRAY_TYPE */
-   { 33005, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP */
-   { 33025, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP_ARB */
-   { 33049, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
-   { 33080, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB */
-   { 33115, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
-   { 33146, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB */
-   { 33181, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
-   { 33212, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB */
-   { 33247, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
-   { 33278, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB */
-   { 33313, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
-   { 33344, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB */
-   { 33379, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
-   { 33410, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB */
-   { 33445, 0x00008071 }, /* GL_TEXTURE_DEPTH */
-   { 33462, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE */
-   { 33484, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE_ARB */
-   { 33510, 0x00002300 }, /* GL_TEXTURE_ENV */
-   { 33525, 0x00002201 }, /* GL_TEXTURE_ENV_COLOR */
-   { 33546, 0x00002200 }, /* GL_TEXTURE_ENV_MODE */
-   { 33566, 0x00008500 }, /* GL_TEXTURE_FILTER_CONTROL */
-   { 33592, 0x00002500 }, /* GL_TEXTURE_GEN_MODE */
-   { 33612, 0x00000C63 }, /* GL_TEXTURE_GEN_Q */
-   { 33629, 0x00000C62 }, /* GL_TEXTURE_GEN_R */
-   { 33646, 0x00000C60 }, /* GL_TEXTURE_GEN_S */
-   { 33663, 0x00000C61 }, /* GL_TEXTURE_GEN_T */
-   { 33680, 0x0000819D }, /* GL_TEXTURE_GEQUAL_R_SGIX */
-   { 33705, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE */
-   { 33727, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE_EXT */
-   { 33753, 0x00001001 }, /* GL_TEXTURE_HEIGHT */
-   { 33771, 0x000080ED }, /* GL_TEXTURE_INDEX_SIZE_EXT */
-   { 33797, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE */
-   { 33823, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE_EXT */
-   { 33853, 0x00001003 }, /* GL_TEXTURE_INTERNAL_FORMAT */
-   { 33880, 0x0000819C }, /* GL_TEXTURE_LEQUAL_R_SGIX */
-   { 33905, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS */
-   { 33925, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS_EXT */
-   { 33949, 0x00008190 }, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
-   { 33976, 0x0000818E }, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
-   { 34003, 0x0000818F }, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
-   { 34030, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE */
-   { 34056, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE_EXT */
-   { 34086, 0x00002800 }, /* GL_TEXTURE_MAG_FILTER */
-   { 34108, 0x00000BA8 }, /* GL_TEXTURE_MATRIX */
-   { 34126, 0x000084FE }, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
-   { 34156, 0x0000836B }, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
-   { 34184, 0x00008369 }, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
-   { 34212, 0x0000836A }, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
-   { 34240, 0x0000813D }, /* GL_TEXTURE_MAX_LEVEL */
-   { 34261, 0x0000813B }, /* GL_TEXTURE_MAX_LOD */
-   { 34280, 0x00002801 }, /* GL_TEXTURE_MIN_FILTER */
-   { 34302, 0x0000813A }, /* GL_TEXTURE_MIN_LOD */
-   { 34321, 0x00008066 }, /* GL_TEXTURE_PRIORITY */
-   { 34341, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_ARB */
-   { 34366, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_NV */
-   { 34390, 0x0000805C }, /* GL_TEXTURE_RED_SIZE */
-   { 34410, 0x0000805C }, /* GL_TEXTURE_RED_SIZE_EXT */
-   { 34434, 0x00008067 }, /* GL_TEXTURE_RESIDENT */
-   { 34454, 0x00000BA5 }, /* GL_TEXTURE_STACK_DEPTH */
-   { 34477, 0x00008065 }, /* GL_TEXTURE_TOO_LARGE_EXT */
-   { 34502, 0x0000888F }, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
-   { 34536, 0x00001000 }, /* GL_TEXTURE_WIDTH */
-   { 34553, 0x00008072 }, /* GL_TEXTURE_WRAP_R */
-   { 34571, 0x00002802 }, /* GL_TEXTURE_WRAP_S */
-   { 34589, 0x00002803 }, /* GL_TEXTURE_WRAP_T */
-   { 34607, 0x000088BF }, /* GL_TIME_ELAPSED_EXT */
-   { 34627, 0x00008648 }, /* GL_TRACK_MATRIX_NV */
-   { 34646, 0x00008649 }, /* GL_TRACK_MATRIX_TRANSFORM_NV */
-   { 34675, 0x00001000 }, /* GL_TRANSFORM_BIT */
-   { 34692, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX */
-   { 34718, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX_ARB */
-   { 34748, 0x000088B7 }, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
-   { 34780, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
-   { 34810, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX_ARB */
-   { 34844, 0x0000862C }, /* GL_TRANSPOSE_NV */
-   { 34860, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX */
-   { 34891, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX_ARB */
-   { 34926, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX */
-   { 34954, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX_ARB */
-   { 34986, 0x00000004 }, /* GL_TRIANGLES */
-   { 34999, 0x00000006 }, /* GL_TRIANGLE_FAN */
-   { 35015, 0x00008615 }, /* GL_TRIANGLE_MESH_SUN */
-   { 35036, 0x00000005 }, /* GL_TRIANGLE_STRIP */
-   { 35054, 0x00000001 }, /* GL_TRUE */
-   { 35062, 0x00000CF5 }, /* GL_UNPACK_ALIGNMENT */
-   { 35082, 0x0000806E }, /* GL_UNPACK_IMAGE_HEIGHT */
-   { 35105, 0x00000CF1 }, /* GL_UNPACK_LSB_FIRST */
-   { 35125, 0x00000CF2 }, /* GL_UNPACK_ROW_LENGTH */
-   { 35146, 0x0000806D }, /* GL_UNPACK_SKIP_IMAGES */
-   { 35168, 0x00000CF4 }, /* GL_UNPACK_SKIP_PIXELS */
-   { 35190, 0x00000CF3 }, /* GL_UNPACK_SKIP_ROWS */
-   { 35210, 0x00000CF0 }, /* GL_UNPACK_SWAP_BYTES */
-   { 35231, 0x00001401 }, /* GL_UNSIGNED_BYTE */
-   { 35248, 0x00008362 }, /* GL_UNSIGNED_BYTE_2_3_3_REV */
-   { 35275, 0x00008032 }, /* GL_UNSIGNED_BYTE_3_3_2 */
-   { 35298, 0x00001405 }, /* GL_UNSIGNED_INT */
-   { 35314, 0x00008036 }, /* GL_UNSIGNED_INT_10_10_10_2 */
-   { 35341, 0x000084FA }, /* GL_UNSIGNED_INT_24_8_NV */
-   { 35365, 0x00008368 }, /* GL_UNSIGNED_INT_2_10_10_10_REV */
-   { 35396, 0x00008035 }, /* GL_UNSIGNED_INT_8_8_8_8 */
-   { 35420, 0x00008367 }, /* GL_UNSIGNED_INT_8_8_8_8_REV */
-   { 35448, 0x00001403 }, /* GL_UNSIGNED_SHORT */
-   { 35466, 0x00008366 }, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
-   { 35496, 0x00008033 }, /* GL_UNSIGNED_SHORT_4_4_4_4 */
-   { 35522, 0x00008365 }, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
-   { 35552, 0x00008034 }, /* GL_UNSIGNED_SHORT_5_5_5_1 */
-   { 35578, 0x00008363 }, /* GL_UNSIGNED_SHORT_5_6_5 */
-   { 35602, 0x00008364 }, /* GL_UNSIGNED_SHORT_5_6_5_REV */
-   { 35630, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_APPLE */
-   { 35658, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_MESA */
-   { 35685, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
-   { 35717, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_MESA */
-   { 35748, 0x00008CA2 }, /* GL_UPPER_LEFT */
-   { 35762, 0x00002A20 }, /* GL_V2F */
-   { 35769, 0x00002A21 }, /* GL_V3F */
-   { 35776, 0x00008B83 }, /* GL_VALIDATE_STATUS */
-   { 35795, 0x00001F00 }, /* GL_VENDOR */
-   { 35805, 0x00001F02 }, /* GL_VERSION */
-   { 35816, 0x00008074 }, /* GL_VERTEX_ARRAY */
-   { 35832, 0x000085B5 }, /* GL_VERTEX_ARRAY_BINDING_APPLE */
-   { 35862, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
-   { 35893, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING_ARB */
-   { 35928, 0x0000808E }, /* GL_VERTEX_ARRAY_POINTER */
-   { 35952, 0x0000807A }, /* GL_VERTEX_ARRAY_SIZE */
-   { 35973, 0x0000807C }, /* GL_VERTEX_ARRAY_STRIDE */
-   { 35996, 0x0000807B }, /* GL_VERTEX_ARRAY_TYPE */
-   { 36017, 0x00008650 }, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
-   { 36044, 0x0000865A }, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
-   { 36072, 0x0000865B }, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
-   { 36100, 0x0000865C }, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
-   { 36128, 0x0000865D }, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
-   { 36156, 0x0000865E }, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
-   { 36184, 0x0000865F }, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
-   { 36212, 0x00008651 }, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
-   { 36239, 0x00008652 }, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
-   { 36266, 0x00008653 }, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
-   { 36293, 0x00008654 }, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
-   { 36320, 0x00008655 }, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
-   { 36347, 0x00008656 }, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
-   { 36374, 0x00008657 }, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
-   { 36401, 0x00008658 }, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
-   { 36428, 0x00008659 }, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
-   { 36455, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
-   { 36493, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB */
-   { 36535, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
-   { 36566, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB */
-   { 36601, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
-   { 36635, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB */
-   { 36673, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
-   { 36704, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER_ARB */
-   { 36739, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
-   { 36767, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB */
-   { 36799, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
-   { 36829, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB */
-   { 36863, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
-   { 36891, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB */
-   { 36923, 0x000086A7 }, /* GL_VERTEX_BLEND_ARB */
-   { 36943, 0x00008620 }, /* GL_VERTEX_PROGRAM_ARB */
-   { 36965, 0x0000864A }, /* GL_VERTEX_PROGRAM_BINDING_NV */
-   { 36994, 0x00008620 }, /* GL_VERTEX_PROGRAM_NV */
-   { 37015, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE */
-   { 37044, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_ARB */
-   { 37077, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_NV */
-   { 37109, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE */
-   { 37136, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_ARB */
-   { 37167, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_NV */
-   { 37197, 0x00008B31 }, /* GL_VERTEX_SHADER */
-   { 37214, 0x00008B31 }, /* GL_VERTEX_SHADER_ARB */
-   { 37235, 0x00008621 }, /* GL_VERTEX_STATE_PROGRAM_NV */
-   { 37262, 0x00000BA2 }, /* GL_VIEWPORT */
-   { 37274, 0x00000800 }, /* GL_VIEWPORT_BIT */
-   { 37290, 0x000086AD }, /* GL_WEIGHT_ARRAY_ARB */
-   { 37310, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
-   { 37341, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING_ARB */
-   { 37376, 0x000086AC }, /* GL_WEIGHT_ARRAY_POINTER_ARB */
-   { 37404, 0x000086AB }, /* GL_WEIGHT_ARRAY_SIZE_ARB */
-   { 37429, 0x000086AA }, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
-   { 37456, 0x000086A9 }, /* GL_WEIGHT_ARRAY_TYPE_ARB */
-   { 37481, 0x000086A6 }, /* GL_WEIGHT_SUM_UNITY_ARB */
-   { 37505, 0x000081D4 }, /* GL_WRAP_BORDER_SUN */
-   { 37524, 0x000088B9 }, /* GL_WRITE_ONLY */
-   { 37538, 0x000088B9 }, /* GL_WRITE_ONLY_ARB */
-   { 37556, 0x00001506 }, /* GL_XOR */
-   { 37563, 0x000085B9 }, /* GL_YCBCR_422_APPLE */
-   { 37582, 0x00008757 }, /* GL_YCBCR_MESA */
-   { 37596, 0x00000000 }, /* GL_ZERO */
-   { 37604, 0x00000D16 }, /* GL_ZOOM_X */
-   { 37614, 0x00000D17 }, /* GL_ZOOM_Y */
+   {  2476, 0x00008CE0 }, /* GL_COLOR_ATTACHMENT0 */
+   {  2497, 0x00008CE0 }, /* GL_COLOR_ATTACHMENT0_EXT */
+   {  2522, 0x00008CE1 }, /* GL_COLOR_ATTACHMENT1 */
+   {  2543, 0x00008CEA }, /* GL_COLOR_ATTACHMENT10 */
+   {  2565, 0x00008CEA }, /* GL_COLOR_ATTACHMENT10_EXT */
+   {  2591, 0x00008CEB }, /* GL_COLOR_ATTACHMENT11 */
+   {  2613, 0x00008CEB }, /* GL_COLOR_ATTACHMENT11_EXT */
+   {  2639, 0x00008CEC }, /* GL_COLOR_ATTACHMENT12 */
+   {  2661, 0x00008CEC }, /* GL_COLOR_ATTACHMENT12_EXT */
+   {  2687, 0x00008CED }, /* GL_COLOR_ATTACHMENT13 */
+   {  2709, 0x00008CED }, /* GL_COLOR_ATTACHMENT13_EXT */
+   {  2735, 0x00008CEE }, /* GL_COLOR_ATTACHMENT14 */
+   {  2757, 0x00008CEE }, /* GL_COLOR_ATTACHMENT14_EXT */
+   {  2783, 0x00008CEF }, /* GL_COLOR_ATTACHMENT15 */
+   {  2805, 0x00008CEF }, /* GL_COLOR_ATTACHMENT15_EXT */
+   {  2831, 0x00008CE1 }, /* GL_COLOR_ATTACHMENT1_EXT */
+   {  2856, 0x00008CE2 }, /* GL_COLOR_ATTACHMENT2 */
+   {  2877, 0x00008CE2 }, /* GL_COLOR_ATTACHMENT2_EXT */
+   {  2902, 0x00008CE3 }, /* GL_COLOR_ATTACHMENT3 */
+   {  2923, 0x00008CE3 }, /* GL_COLOR_ATTACHMENT3_EXT */
+   {  2948, 0x00008CE4 }, /* GL_COLOR_ATTACHMENT4 */
+   {  2969, 0x00008CE4 }, /* GL_COLOR_ATTACHMENT4_EXT */
+   {  2994, 0x00008CE5 }, /* GL_COLOR_ATTACHMENT5 */
+   {  3015, 0x00008CE5 }, /* GL_COLOR_ATTACHMENT5_EXT */
+   {  3040, 0x00008CE6 }, /* GL_COLOR_ATTACHMENT6 */
+   {  3061, 0x00008CE6 }, /* GL_COLOR_ATTACHMENT6_EXT */
+   {  3086, 0x00008CE7 }, /* GL_COLOR_ATTACHMENT7 */
+   {  3107, 0x00008CE7 }, /* GL_COLOR_ATTACHMENT7_EXT */
+   {  3132, 0x00008CE8 }, /* GL_COLOR_ATTACHMENT8 */
+   {  3153, 0x00008CE8 }, /* GL_COLOR_ATTACHMENT8_EXT */
+   {  3178, 0x00008CE9 }, /* GL_COLOR_ATTACHMENT9 */
+   {  3199, 0x00008CE9 }, /* GL_COLOR_ATTACHMENT9_EXT */
+   {  3224, 0x00004000 }, /* GL_COLOR_BUFFER_BIT */
+   {  3244, 0x00000C22 }, /* GL_COLOR_CLEAR_VALUE */
+   {  3265, 0x00001900 }, /* GL_COLOR_INDEX */
+   {  3280, 0x00001603 }, /* GL_COLOR_INDEXES */
+   {  3297, 0x00000BF2 }, /* GL_COLOR_LOGIC_OP */
+   {  3315, 0x00000B57 }, /* GL_COLOR_MATERIAL */
+   {  3333, 0x00000B55 }, /* GL_COLOR_MATERIAL_FACE */
+   {  3356, 0x00000B56 }, /* GL_COLOR_MATERIAL_PARAMETER */
+   {  3384, 0x000080B1 }, /* GL_COLOR_MATRIX */
+   {  3400, 0x000080B1 }, /* GL_COLOR_MATRIX_SGI */
+   {  3420, 0x000080B2 }, /* GL_COLOR_MATRIX_STACK_DEPTH */
+   {  3448, 0x000080B2 }, /* GL_COLOR_MATRIX_STACK_DEPTH_SGI */
+   {  3480, 0x00008458 }, /* GL_COLOR_SUM */
+   {  3493, 0x00008458 }, /* GL_COLOR_SUM_ARB */
+   {  3510, 0x000080D0 }, /* GL_COLOR_TABLE */
+   {  3525, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE */
+   {  3551, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE_EXT */
+   {  3581, 0x000080DD }, /* GL_COLOR_TABLE_ALPHA_SIZE_SGI */
+   {  3611, 0x000080D7 }, /* GL_COLOR_TABLE_BIAS */
+   {  3631, 0x000080D7 }, /* GL_COLOR_TABLE_BIAS_SGI */
+   {  3655, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE */
+   {  3680, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE_EXT */
+   {  3709, 0x000080DC }, /* GL_COLOR_TABLE_BLUE_SIZE_SGI */
+   {  3738, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT */
+   {  3760, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT_EXT */
+   {  3786, 0x000080D8 }, /* GL_COLOR_TABLE_FORMAT_SGI */
+   {  3812, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE */
+   {  3838, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE_EXT */
+   {  3868, 0x000080DB }, /* GL_COLOR_TABLE_GREEN_SIZE_SGI */
+   {  3898, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE */
+   {  3928, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE_EXT */
+   {  3962, 0x000080DF }, /* GL_COLOR_TABLE_INTENSITY_SIZE_SGI */
+   {  3996, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE */
+   {  4026, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE_EXT */
+   {  4060, 0x000080DE }, /* GL_COLOR_TABLE_LUMINANCE_SIZE_SGI */
+   {  4094, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE */
+   {  4118, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE_EXT */
+   {  4146, 0x000080DA }, /* GL_COLOR_TABLE_RED_SIZE_SGI */
+   {  4174, 0x000080D6 }, /* GL_COLOR_TABLE_SCALE */
+   {  4195, 0x000080D6 }, /* GL_COLOR_TABLE_SCALE_SGI */
+   {  4220, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH */
+   {  4241, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH_EXT */
+   {  4266, 0x000080D9 }, /* GL_COLOR_TABLE_WIDTH_SGI */
+   {  4291, 0x00000C23 }, /* GL_COLOR_WRITEMASK */
+   {  4310, 0x00008570 }, /* GL_COMBINE */
+   {  4321, 0x00008503 }, /* GL_COMBINE4 */
+   {  4333, 0x00008572 }, /* GL_COMBINE_ALPHA */
+   {  4350, 0x00008572 }, /* GL_COMBINE_ALPHA_ARB */
+   {  4371, 0x00008572 }, /* GL_COMBINE_ALPHA_EXT */
+   {  4392, 0x00008570 }, /* GL_COMBINE_ARB */
+   {  4407, 0x00008570 }, /* GL_COMBINE_EXT */
+   {  4422, 0x00008571 }, /* GL_COMBINE_RGB */
+   {  4437, 0x00008571 }, /* GL_COMBINE_RGB_ARB */
+   {  4456, 0x00008571 }, /* GL_COMBINE_RGB_EXT */
+   {  4475, 0x0000884E }, /* GL_COMPARE_REF_DEPTH_TO_TEXTURE_EXT */
+   {  4511, 0x0000884E }, /* GL_COMPARE_R_TO_TEXTURE */
+   {  4535, 0x0000884E }, /* GL_COMPARE_R_TO_TEXTURE_ARB */
+   {  4563, 0x00001300 }, /* GL_COMPILE */
+   {  4574, 0x00001301 }, /* GL_COMPILE_AND_EXECUTE */
+   {  4597, 0x00008B81 }, /* GL_COMPILE_STATUS */
+   {  4615, 0x000084E9 }, /* GL_COMPRESSED_ALPHA */
+   {  4635, 0x000084E9 }, /* GL_COMPRESSED_ALPHA_ARB */
+   {  4659, 0x000084EC }, /* GL_COMPRESSED_INTENSITY */
+   {  4683, 0x000084EC }, /* GL_COMPRESSED_INTENSITY_ARB */
+   {  4711, 0x000084EA }, /* GL_COMPRESSED_LUMINANCE */
+   {  4735, 0x000084EB }, /* GL_COMPRESSED_LUMINANCE_ALPHA */
+   {  4765, 0x000084EB }, /* GL_COMPRESSED_LUMINANCE_ALPHA_ARB */
+   {  4799, 0x000084EA }, /* GL_COMPRESSED_LUMINANCE_ARB */
+   {  4827, 0x000084ED }, /* GL_COMPRESSED_RGB */
+   {  4845, 0x000084EE }, /* GL_COMPRESSED_RGBA */
+   {  4864, 0x000084EE }, /* GL_COMPRESSED_RGBA_ARB */
+   {  4887, 0x000086B1 }, /* GL_COMPRESSED_RGBA_FXT1_3DFX */
+   {  4916, 0x000083F1 }, /* GL_COMPRESSED_RGBA_S3TC_DXT1_EXT */
+   {  4949, 0x000083F2 }, /* GL_COMPRESSED_RGBA_S3TC_DXT3_EXT */
+   {  4982, 0x000083F3 }, /* GL_COMPRESSED_RGBA_S3TC_DXT5_EXT */
+   {  5015, 0x000084ED }, /* GL_COMPRESSED_RGB_ARB */
+   {  5037, 0x000086B0 }, /* GL_COMPRESSED_RGB_FXT1_3DFX */
+   {  5065, 0x000083F0 }, /* GL_COMPRESSED_RGB_S3TC_DXT1_EXT */
+   {  5097, 0x000086A3 }, /* GL_COMPRESSED_TEXTURE_FORMATS */
+   {  5127, 0x00008576 }, /* GL_CONSTANT */
+   {  5139, 0x00008003 }, /* GL_CONSTANT_ALPHA */
+   {  5157, 0x00008003 }, /* GL_CONSTANT_ALPHA_EXT */
+   {  5179, 0x00008576 }, /* GL_CONSTANT_ARB */
+   {  5195, 0x00001207 }, /* GL_CONSTANT_ATTENUATION */
+   {  5219, 0x00008151 }, /* GL_CONSTANT_BORDER_HP */
+   {  5241, 0x00008001 }, /* GL_CONSTANT_COLOR */
+   {  5259, 0x00008001 }, /* GL_CONSTANT_COLOR_EXT */
+   {  5281, 0x00008576 }, /* GL_CONSTANT_EXT */
+   {  5297, 0x00008010 }, /* GL_CONVOLUTION_1D */
+   {  5315, 0x00008011 }, /* GL_CONVOLUTION_2D */
+   {  5333, 0x00008154 }, /* GL_CONVOLUTION_BORDER_COLOR */
+   {  5361, 0x00008154 }, /* GL_CONVOLUTION_BORDER_COLOR_HP */
+   {  5392, 0x00008013 }, /* GL_CONVOLUTION_BORDER_MODE */
+   {  5419, 0x00008013 }, /* GL_CONVOLUTION_BORDER_MODE_EXT */
+   {  5450, 0x00008015 }, /* GL_CONVOLUTION_FILTER_BIAS */
+   {  5477, 0x00008015 }, /* GL_CONVOLUTION_FILTER_BIAS_EXT */
+   {  5508, 0x00008014 }, /* GL_CONVOLUTION_FILTER_SCALE */
+   {  5536, 0x00008014 }, /* GL_CONVOLUTION_FILTER_SCALE_EXT */
+   {  5568, 0x00008017 }, /* GL_CONVOLUTION_FORMAT */
+   {  5590, 0x00008017 }, /* GL_CONVOLUTION_FORMAT_EXT */
+   {  5616, 0x00008019 }, /* GL_CONVOLUTION_HEIGHT */
+   {  5638, 0x00008019 }, /* GL_CONVOLUTION_HEIGHT_EXT */
+   {  5664, 0x00008018 }, /* GL_CONVOLUTION_WIDTH */
+   {  5685, 0x00008018 }, /* GL_CONVOLUTION_WIDTH_EXT */
+   {  5710, 0x00008862 }, /* GL_COORD_REPLACE */
+   {  5727, 0x00008862 }, /* GL_COORD_REPLACE_ARB */
+   {  5748, 0x00008862 }, /* GL_COORD_REPLACE_NV */
+   {  5768, 0x00001503 }, /* GL_COPY */
+   {  5776, 0x0000150C }, /* GL_COPY_INVERTED */
+   {  5793, 0x00000706 }, /* GL_COPY_PIXEL_TOKEN */
+   {  5813, 0x00000B44 }, /* GL_CULL_FACE */
+   {  5826, 0x00000B45 }, /* GL_CULL_FACE_MODE */
+   {  5844, 0x000081AA }, /* GL_CULL_VERTEX_EXT */
+   {  5863, 0x000081AC }, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
+   {  5895, 0x000081AB }, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
+   {  5930, 0x00008626 }, /* GL_CURRENT_ATTRIB_NV */
+   {  5951, 0x00000001 }, /* GL_CURRENT_BIT */
+   {  5966, 0x00000B00 }, /* GL_CURRENT_COLOR */
+   {  5983, 0x00008453 }, /* GL_CURRENT_FOG_COORD */
+   {  6004, 0x00008453 }, /* GL_CURRENT_FOG_COORDINATE */
+   {  6030, 0x00000B01 }, /* GL_CURRENT_INDEX */
+   {  6047, 0x00008641 }, /* GL_CURRENT_MATRIX_ARB */
+   {  6069, 0x00008845 }, /* GL_CURRENT_MATRIX_INDEX_ARB */
+   {  6097, 0x00008641 }, /* GL_CURRENT_MATRIX_NV */
+   {  6118, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
+   {  6152, 0x00008640 }, /* GL_CURRENT_MATRIX_STACK_DEPTH_NV */
+   {  6185, 0x00000B02 }, /* GL_CURRENT_NORMAL */
+   {  6203, 0x00008843 }, /* GL_CURRENT_PALETTE_MATRIX_ARB */
+   {  6233, 0x00008B8D }, /* GL_CURRENT_PROGRAM */
+   {  6252, 0x00008865 }, /* GL_CURRENT_QUERY */
+   {  6269, 0x00008865 }, /* GL_CURRENT_QUERY_ARB */
+   {  6290, 0x00000B04 }, /* GL_CURRENT_RASTER_COLOR */
+   {  6314, 0x00000B09 }, /* GL_CURRENT_RASTER_DISTANCE */
+   {  6341, 0x00000B05 }, /* GL_CURRENT_RASTER_INDEX */
+   {  6365, 0x00000B07 }, /* GL_CURRENT_RASTER_POSITION */
+   {  6392, 0x00000B08 }, /* GL_CURRENT_RASTER_POSITION_VALID */
+   {  6425, 0x00000B06 }, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
+   {  6458, 0x00008459 }, /* GL_CURRENT_SECONDARY_COLOR */
+   {  6485, 0x00000B03 }, /* GL_CURRENT_TEXTURE_COORDS */
+   {  6511, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB */
+   {  6536, 0x00008626 }, /* GL_CURRENT_VERTEX_ATTRIB_ARB */
+   {  6565, 0x000086A8 }, /* GL_CURRENT_WEIGHT_ARB */
+   {  6587, 0x00000900 }, /* GL_CW */
+   {  6593, 0x0000875B }, /* GL_DEBUG_ASSERT_MESA */
+   {  6614, 0x00008759 }, /* GL_DEBUG_OBJECT_MESA */
+   {  6635, 0x0000875A }, /* GL_DEBUG_PRINT_MESA */
+   {  6655, 0x00002101 }, /* GL_DECAL */
+   {  6664, 0x00001E03 }, /* GL_DECR */
+   {  6672, 0x00008508 }, /* GL_DECR_WRAP */
+   {  6685, 0x00008508 }, /* GL_DECR_WRAP_EXT */
+   {  6702, 0x00008B80 }, /* GL_DELETE_STATUS */
+   {  6719, 0x00001801 }, /* GL_DEPTH */
+   {  6728, 0x000088F0 }, /* GL_DEPTH24_STENCIL8 */
+   {  6748, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT */
+   {  6768, 0x00008D00 }, /* GL_DEPTH_ATTACHMENT_EXT */
+   {  6792, 0x00000D1F }, /* GL_DEPTH_BIAS */
+   {  6806, 0x00000D56 }, /* GL_DEPTH_BITS */
+   {  6820, 0x00008891 }, /* GL_DEPTH_BOUNDS_EXT */
+   {  6840, 0x00008890 }, /* GL_DEPTH_BOUNDS_TEST_EXT */
+   {  6865, 0x00000100 }, /* GL_DEPTH_BUFFER_BIT */
+   {  6885, 0x0000864F }, /* GL_DEPTH_CLAMP_NV */
+   {  6903, 0x00000B73 }, /* GL_DEPTH_CLEAR_VALUE */
+   {  6924, 0x00001902 }, /* GL_DEPTH_COMPONENT */
+   {  6943, 0x000081A5 }, /* GL_DEPTH_COMPONENT16 */
+   {  6964, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_ARB */
+   {  6989, 0x000081A5 }, /* GL_DEPTH_COMPONENT16_SGIX */
+   {  7015, 0x000081A6 }, /* GL_DEPTH_COMPONENT24 */
+   {  7036, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_ARB */
+   {  7061, 0x000081A6 }, /* GL_DEPTH_COMPONENT24_SGIX */
+   {  7087, 0x000081A7 }, /* GL_DEPTH_COMPONENT32 */
+   {  7108, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_ARB */
+   {  7133, 0x000081A7 }, /* GL_DEPTH_COMPONENT32_SGIX */
+   {  7159, 0x00000B74 }, /* GL_DEPTH_FUNC */
+   {  7173, 0x00000B70 }, /* GL_DEPTH_RANGE */
+   {  7188, 0x00000D1E }, /* GL_DEPTH_SCALE */
+   {  7203, 0x000084F9 }, /* GL_DEPTH_STENCIL */
+   {  7220, 0x0000821A }, /* GL_DEPTH_STENCIL_ATTACHMENT */
+   {  7248, 0x000084F9 }, /* GL_DEPTH_STENCIL_NV */
+   {  7268, 0x0000886F }, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
+   {  7296, 0x0000886E }, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
+   {  7324, 0x00000B71 }, /* GL_DEPTH_TEST */
+   {  7338, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE */
+   {  7360, 0x0000884B }, /* GL_DEPTH_TEXTURE_MODE_ARB */
+   {  7386, 0x00000B72 }, /* GL_DEPTH_WRITEMASK */
+   {  7405, 0x00001201 }, /* GL_DIFFUSE */
+   {  7416, 0x00000BD0 }, /* GL_DITHER */
+   {  7426, 0x00000A02 }, /* GL_DOMAIN */
+   {  7436, 0x00001100 }, /* GL_DONT_CARE */
+   {  7449, 0x000086AE }, /* GL_DOT3_RGB */
+   {  7461, 0x000086AF }, /* GL_DOT3_RGBA */
+   {  7474, 0x000086AF }, /* GL_DOT3_RGBA_ARB */
+   {  7491, 0x00008741 }, /* GL_DOT3_RGBA_EXT */
+   {  7508, 0x000086AE }, /* GL_DOT3_RGB_ARB */
+   {  7524, 0x00008740 }, /* GL_DOT3_RGB_EXT */
+   {  7540, 0x0000140A }, /* GL_DOUBLE */
+   {  7550, 0x00000C32 }, /* GL_DOUBLEBUFFER */
+   {  7566, 0x00000C01 }, /* GL_DRAW_BUFFER */
+   {  7581, 0x00008825 }, /* GL_DRAW_BUFFER0 */
+   {  7597, 0x00008825 }, /* GL_DRAW_BUFFER0_ARB */
+   {  7617, 0x00008825 }, /* GL_DRAW_BUFFER0_ATI */
+   {  7637, 0x00008826 }, /* GL_DRAW_BUFFER1 */
+   {  7653, 0x0000882F }, /* GL_DRAW_BUFFER10 */
+   {  7670, 0x0000882F }, /* GL_DRAW_BUFFER10_ARB */
+   {  7691, 0x0000882F }, /* GL_DRAW_BUFFER10_ATI */
+   {  7712, 0x00008830 }, /* GL_DRAW_BUFFER11 */
+   {  7729, 0x00008830 }, /* GL_DRAW_BUFFER11_ARB */
+   {  7750, 0x00008830 }, /* GL_DRAW_BUFFER11_ATI */
+   {  7771, 0x00008831 }, /* GL_DRAW_BUFFER12 */
+   {  7788, 0x00008831 }, /* GL_DRAW_BUFFER12_ARB */
+   {  7809, 0x00008831 }, /* GL_DRAW_BUFFER12_ATI */
+   {  7830, 0x00008832 }, /* GL_DRAW_BUFFER13 */
+   {  7847, 0x00008832 }, /* GL_DRAW_BUFFER13_ARB */
+   {  7868, 0x00008832 }, /* GL_DRAW_BUFFER13_ATI */
+   {  7889, 0x00008833 }, /* GL_DRAW_BUFFER14 */
+   {  7906, 0x00008833 }, /* GL_DRAW_BUFFER14_ARB */
+   {  7927, 0x00008833 }, /* GL_DRAW_BUFFER14_ATI */
+   {  7948, 0x00008834 }, /* GL_DRAW_BUFFER15 */
+   {  7965, 0x00008834 }, /* GL_DRAW_BUFFER15_ARB */
+   {  7986, 0x00008834 }, /* GL_DRAW_BUFFER15_ATI */
+   {  8007, 0x00008826 }, /* GL_DRAW_BUFFER1_ARB */
+   {  8027, 0x00008826 }, /* GL_DRAW_BUFFER1_ATI */
+   {  8047, 0x00008827 }, /* GL_DRAW_BUFFER2 */
+   {  8063, 0x00008827 }, /* GL_DRAW_BUFFER2_ARB */
+   {  8083, 0x00008827 }, /* GL_DRAW_BUFFER2_ATI */
+   {  8103, 0x00008828 }, /* GL_DRAW_BUFFER3 */
+   {  8119, 0x00008828 }, /* GL_DRAW_BUFFER3_ARB */
+   {  8139, 0x00008828 }, /* GL_DRAW_BUFFER3_ATI */
+   {  8159, 0x00008829 }, /* GL_DRAW_BUFFER4 */
+   {  8175, 0x00008829 }, /* GL_DRAW_BUFFER4_ARB */
+   {  8195, 0x00008829 }, /* GL_DRAW_BUFFER4_ATI */
+   {  8215, 0x0000882A }, /* GL_DRAW_BUFFER5 */
+   {  8231, 0x0000882A }, /* GL_DRAW_BUFFER5_ARB */
+   {  8251, 0x0000882A }, /* GL_DRAW_BUFFER5_ATI */
+   {  8271, 0x0000882B }, /* GL_DRAW_BUFFER6 */
+   {  8287, 0x0000882B }, /* GL_DRAW_BUFFER6_ARB */
+   {  8307, 0x0000882B }, /* GL_DRAW_BUFFER6_ATI */
+   {  8327, 0x0000882C }, /* GL_DRAW_BUFFER7 */
+   {  8343, 0x0000882C }, /* GL_DRAW_BUFFER7_ARB */
+   {  8363, 0x0000882C }, /* GL_DRAW_BUFFER7_ATI */
+   {  8383, 0x0000882D }, /* GL_DRAW_BUFFER8 */
+   {  8399, 0x0000882D }, /* GL_DRAW_BUFFER8_ARB */
+   {  8419, 0x0000882D }, /* GL_DRAW_BUFFER8_ATI */
+   {  8439, 0x0000882E }, /* GL_DRAW_BUFFER9 */
+   {  8455, 0x0000882E }, /* GL_DRAW_BUFFER9_ARB */
+   {  8475, 0x0000882E }, /* GL_DRAW_BUFFER9_ATI */
+   {  8495, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER */
+   {  8515, 0x00008CA6 }, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
+   {  8547, 0x00008CA9 }, /* GL_DRAW_FRAMEBUFFER_EXT */
+   {  8571, 0x00000705 }, /* GL_DRAW_PIXEL_TOKEN */
+   {  8591, 0x00000304 }, /* GL_DST_ALPHA */
+   {  8604, 0x00000306 }, /* GL_DST_COLOR */
+   {  8617, 0x000088EA }, /* GL_DYNAMIC_COPY */
+   {  8633, 0x000088EA }, /* GL_DYNAMIC_COPY_ARB */
+   {  8653, 0x000088E8 }, /* GL_DYNAMIC_DRAW */
+   {  8669, 0x000088E8 }, /* GL_DYNAMIC_DRAW_ARB */
+   {  8689, 0x000088E9 }, /* GL_DYNAMIC_READ */
+   {  8705, 0x000088E9 }, /* GL_DYNAMIC_READ_ARB */
+   {  8725, 0x00000B43 }, /* GL_EDGE_FLAG */
+   {  8738, 0x00008079 }, /* GL_EDGE_FLAG_ARRAY */
+   {  8757, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
+   {  8791, 0x0000889B }, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB */
+   {  8829, 0x00008093 }, /* GL_EDGE_FLAG_ARRAY_POINTER */
+   {  8856, 0x0000808C }, /* GL_EDGE_FLAG_ARRAY_STRIDE */
+   {  8882, 0x00008893 }, /* GL_ELEMENT_ARRAY_BUFFER */
+   {  8906, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
+   {  8938, 0x00008895 }, /* GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB */
+   {  8974, 0x00001600 }, /* GL_EMISSION */
+   {  8986, 0x00002000 }, /* GL_ENABLE_BIT */
+   {  9000, 0x00000202 }, /* GL_EQUAL */
+   {  9009, 0x00001509 }, /* GL_EQUIV */
+   {  9018, 0x00010000 }, /* GL_EVAL_BIT */
+   {  9030, 0x00000800 }, /* GL_EXP */
+   {  9037, 0x00000801 }, /* GL_EXP2 */
+   {  9045, 0x00001F03 }, /* GL_EXTENSIONS */
+   {  9059, 0x00002400 }, /* GL_EYE_LINEAR */
+   {  9073, 0x00002502 }, /* GL_EYE_PLANE */
+   {  9086, 0x0000855C }, /* GL_EYE_PLANE_ABSOLUTE_NV */
+   {  9111, 0x0000855B }, /* GL_EYE_RADIAL_NV */
+   {  9128, 0x00000000 }, /* GL_FALSE */
+   {  9137, 0x00001101 }, /* GL_FASTEST */
+   {  9148, 0x00001C01 }, /* GL_FEEDBACK */
+   {  9160, 0x00000DF0 }, /* GL_FEEDBACK_BUFFER_POINTER */
+   {  9187, 0x00000DF1 }, /* GL_FEEDBACK_BUFFER_SIZE */
+   {  9211, 0x00000DF2 }, /* GL_FEEDBACK_BUFFER_TYPE */
+   {  9235, 0x00001B02 }, /* GL_FILL */
+   {  9243, 0x00001D00 }, /* GL_FLAT */
+   {  9251, 0x00001406 }, /* GL_FLOAT */
+   {  9260, 0x00008B5A }, /* GL_FLOAT_MAT2 */
+   {  9274, 0x00008B5A }, /* GL_FLOAT_MAT2_ARB */
+   {  9292, 0x00008B5B }, /* GL_FLOAT_MAT3 */
+   {  9306, 0x00008B5B }, /* GL_FLOAT_MAT3_ARB */
+   {  9324, 0x00008B5C }, /* GL_FLOAT_MAT4 */
+   {  9338, 0x00008B5C }, /* GL_FLOAT_MAT4_ARB */
+   {  9356, 0x00008B50 }, /* GL_FLOAT_VEC2 */
+   {  9370, 0x00008B50 }, /* GL_FLOAT_VEC2_ARB */
+   {  9388, 0x00008B51 }, /* GL_FLOAT_VEC3 */
+   {  9402, 0x00008B51 }, /* GL_FLOAT_VEC3_ARB */
+   {  9420, 0x00008B52 }, /* GL_FLOAT_VEC4 */
+   {  9434, 0x00008B52 }, /* GL_FLOAT_VEC4_ARB */
+   {  9452, 0x00000B60 }, /* GL_FOG */
+   {  9459, 0x00000080 }, /* GL_FOG_BIT */
+   {  9470, 0x00000B66 }, /* GL_FOG_COLOR */
+   {  9483, 0x00008451 }, /* GL_FOG_COORD */
+   {  9496, 0x00008451 }, /* GL_FOG_COORDINATE */
+   {  9514, 0x00008457 }, /* GL_FOG_COORDINATE_ARRAY */
+   {  9538, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
+   {  9577, 0x0000889D }, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB */
+   {  9620, 0x00008456 }, /* GL_FOG_COORDINATE_ARRAY_POINTER */
+   {  9652, 0x00008455 }, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
+   {  9683, 0x00008454 }, /* GL_FOG_COORDINATE_ARRAY_TYPE */
+   {  9712, 0x00008450 }, /* GL_FOG_COORDINATE_SOURCE */
+   {  9737, 0x00008457 }, /* GL_FOG_COORD_ARRAY */
+   {  9756, 0x0000889D }, /* GL_FOG_COORD_ARRAY_BUFFER_BINDING */
+   {  9790, 0x00008456 }, /* GL_FOG_COORD_ARRAY_POINTER */
+   {  9817, 0x00008455 }, /* GL_FOG_COORD_ARRAY_STRIDE */
+   {  9843, 0x00008454 }, /* GL_FOG_COORD_ARRAY_TYPE */
+   {  9867, 0x00008450 }, /* GL_FOG_COORD_SRC */
+   {  9884, 0x00000B62 }, /* GL_FOG_DENSITY */
+   {  9899, 0x0000855A }, /* GL_FOG_DISTANCE_MODE_NV */
+   {  9923, 0x00000B64 }, /* GL_FOG_END */
+   {  9934, 0x00000C54 }, /* GL_FOG_HINT */
+   {  9946, 0x00000B61 }, /* GL_FOG_INDEX */
+   {  9959, 0x00000B65 }, /* GL_FOG_MODE */
+   {  9971, 0x00008198 }, /* GL_FOG_OFFSET_SGIX */
+   {  9990, 0x00008199 }, /* GL_FOG_OFFSET_VALUE_SGIX */
+   { 10015, 0x00000B63 }, /* GL_FOG_START */
+   { 10028, 0x00008452 }, /* GL_FRAGMENT_DEPTH */
+   { 10046, 0x00008804 }, /* GL_FRAGMENT_PROGRAM_ARB */
+   { 10070, 0x00008B30 }, /* GL_FRAGMENT_SHADER */
+   { 10089, 0x00008B30 }, /* GL_FRAGMENT_SHADER_ARB */
+   { 10112, 0x00008B8B }, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
+   { 10147, 0x00008D40 }, /* GL_FRAMEBUFFER */
+   { 10162, 0x00008215 }, /* GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE */
+   { 10199, 0x00008214 }, /* GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE */
+   { 10235, 0x00008210 }, /* GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING */
+   { 10276, 0x00008211 }, /* GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE */
+   { 10317, 0x00008216 }, /* GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE */
+   { 10354, 0x00008213 }, /* GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE */
+   { 10391, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME */
+   { 10429, 0x00008CD1 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME_EXT */
+   { 10471, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE */
+   { 10509, 0x00008CD0 }, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE_EXT */
+   { 10551, 0x00008212 }, /* GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE */
+   { 10586, 0x00008217 }, /* GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE */
+   { 10625, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_3D_ZOFFSET_EXT */
+   { 10674, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE */
+   { 10722, 0x00008CD3 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE_EXT */
+   { 10774, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER */
+   { 10814, 0x00008CD4 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER_EXT */
+   { 10858, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL */
+   { 10898, 0x00008CD2 }, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL_EXT */
+   { 10942, 0x00008CA6 }, /* GL_FRAMEBUFFER_BINDING_EXT */
+   { 10969, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE */
+   { 10993, 0x00008CD5 }, /* GL_FRAMEBUFFER_COMPLETE_EXT */
+   { 11021, 0x00008218 }, /* GL_FRAMEBUFFER_DEFAULT */
+   { 11044, 0x00008D40 }, /* GL_FRAMEBUFFER_EXT */
+   { 11063, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT */
+   { 11100, 0x00008CD6 }, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT_EXT */
+   { 11141, 0x00008CD9 }, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
+   { 11182, 0x00008CDB }, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
+   { 11224, 0x00008CD8 }, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
+   { 11275, 0x00008CDA }, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
+   { 11313, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT */
+   { 11358, 0x00008CD7 }, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT_EXT */
+   { 11407, 0x00008D56 }, /* GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE */
+   { 11445, 0x00008CDC }, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
+   { 11487, 0x00008CDE }, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
+   { 11519, 0x00008219 }, /* GL_FRAMEBUFFER_UNDEFINED */
+   { 11544, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED */
+   { 11571, 0x00008CDD }, /* GL_FRAMEBUFFER_UNSUPPORTED_EXT */
+   { 11602, 0x00000404 }, /* GL_FRONT */
+   { 11611, 0x00000408 }, /* GL_FRONT_AND_BACK */
+   { 11629, 0x00000B46 }, /* GL_FRONT_FACE */
+   { 11643, 0x00000400 }, /* GL_FRONT_LEFT */
+   { 11657, 0x00000401 }, /* GL_FRONT_RIGHT */
+   { 11672, 0x00008006 }, /* GL_FUNC_ADD */
+   { 11684, 0x00008006 }, /* GL_FUNC_ADD_EXT */
+   { 11700, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT */
+   { 11725, 0x0000800B }, /* GL_FUNC_REVERSE_SUBTRACT_EXT */
+   { 11754, 0x0000800A }, /* GL_FUNC_SUBTRACT */
+   { 11771, 0x0000800A }, /* GL_FUNC_SUBTRACT_EXT */
+   { 11792, 0x00008191 }, /* GL_GENERATE_MIPMAP */
+   { 11811, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT */
+   { 11835, 0x00008192 }, /* GL_GENERATE_MIPMAP_HINT_SGIS */
+   { 11864, 0x00008191 }, /* GL_GENERATE_MIPMAP_SGIS */
+   { 11888, 0x00000206 }, /* GL_GEQUAL */
+   { 11898, 0x00008009 }, /* GL_GL_BLEND_EQUATION_RGB */
+   { 11923, 0x00008C4A }, /* GL_GL_COMPRESSED_SLUMINANCE */
+   { 11951, 0x00008C4B }, /* GL_GL_COMPRESSED_SLUMINANCE_ALPHA */
+   { 11985, 0x00008C48 }, /* GL_GL_COMPRESSED_SRGB */
+   { 12007, 0x00008C49 }, /* GL_GL_COMPRESSED_SRGB_ALPHA */
+   { 12035, 0x0000845F }, /* GL_GL_CURRENT_RASTER_SECONDARY_COLOR */
+   { 12072, 0x00008B65 }, /* GL_GL_FLOAT_MAT2x3 */
+   { 12091, 0x00008B66 }, /* GL_GL_FLOAT_MAT2x4 */
+   { 12110, 0x00008B67 }, /* GL_GL_FLOAT_MAT3x2 */
+   { 12129, 0x00008B68 }, /* GL_GL_FLOAT_MAT3x4 */
+   { 12148, 0x00008B69 }, /* GL_GL_FLOAT_MAT4x2 */
+   { 12167, 0x00008B6A }, /* GL_GL_FLOAT_MAT4x3 */
+   { 12186, 0x000088EB }, /* GL_GL_PIXEL_PACK_BUFFER */
+   { 12210, 0x000088ED }, /* GL_GL_PIXEL_PACK_BUFFER_BINDING */
+   { 12242, 0x000088EC }, /* GL_GL_PIXEL_UNPACK_BUFFER */
+   { 12268, 0x000088EF }, /* GL_GL_PIXEL_UNPACK_BUFFER_BINDING */
+   { 12302, 0x00008C46 }, /* GL_GL_SLUMINANCE */
+   { 12319, 0x00008C47 }, /* GL_GL_SLUMINANCE8 */
+   { 12337, 0x00008C45 }, /* GL_GL_SLUMINANCE8_ALPHA8 */
+   { 12362, 0x00008C44 }, /* GL_GL_SLUMINANCE_ALPHA */
+   { 12385, 0x00008C40 }, /* GL_GL_SRGB */
+   { 12396, 0x00008C41 }, /* GL_GL_SRGB8 */
+   { 12408, 0x00008C43 }, /* GL_GL_SRGB8_ALPHA8 */
+   { 12427, 0x00008C42 }, /* GL_GL_SRGB_ALPHA */
+   { 12444, 0x00000204 }, /* GL_GREATER */
+   { 12455, 0x00001904 }, /* GL_GREEN */
+   { 12464, 0x00000D19 }, /* GL_GREEN_BIAS */
+   { 12478, 0x00000D53 }, /* GL_GREEN_BITS */
+   { 12492, 0x00000D18 }, /* GL_GREEN_SCALE */
+   { 12507, 0x00008000 }, /* GL_HINT_BIT */
+   { 12519, 0x00008024 }, /* GL_HISTOGRAM */
+   { 12532, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE */
+   { 12556, 0x0000802B }, /* GL_HISTOGRAM_ALPHA_SIZE_EXT */
+   { 12584, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE */
+   { 12607, 0x0000802A }, /* GL_HISTOGRAM_BLUE_SIZE_EXT */
+   { 12634, 0x00008024 }, /* GL_HISTOGRAM_EXT */
+   { 12651, 0x00008027 }, /* GL_HISTOGRAM_FORMAT */
+   { 12671, 0x00008027 }, /* GL_HISTOGRAM_FORMAT_EXT */
+   { 12695, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE */
+   { 12719, 0x00008029 }, /* GL_HISTOGRAM_GREEN_SIZE_EXT */
+   { 12747, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE */
+   { 12775, 0x0000802C }, /* GL_HISTOGRAM_LUMINANCE_SIZE_EXT */
+   { 12807, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE */
+   { 12829, 0x00008028 }, /* GL_HISTOGRAM_RED_SIZE_EXT */
+   { 12855, 0x0000802D }, /* GL_HISTOGRAM_SINK */
+   { 12873, 0x0000802D }, /* GL_HISTOGRAM_SINK_EXT */
+   { 12895, 0x00008026 }, /* GL_HISTOGRAM_WIDTH */
+   { 12914, 0x00008026 }, /* GL_HISTOGRAM_WIDTH_EXT */
+   { 12937, 0x0000862A }, /* GL_IDENTITY_NV */
+   { 12952, 0x00008150 }, /* GL_IGNORE_BORDER_HP */
+   { 12972, 0x00008B9B }, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
+   { 13012, 0x00008B9A }, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
+   { 13050, 0x00001E02 }, /* GL_INCR */
+   { 13058, 0x00008507 }, /* GL_INCR_WRAP */
+   { 13071, 0x00008507 }, /* GL_INCR_WRAP_EXT */
+   { 13088, 0x00008222 }, /* GL_INDEX */
+   { 13097, 0x00008077 }, /* GL_INDEX_ARRAY */
+   { 13112, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING */
+   { 13142, 0x00008899 }, /* GL_INDEX_ARRAY_BUFFER_BINDING_ARB */
+   { 13176, 0x00008091 }, /* GL_INDEX_ARRAY_POINTER */
+   { 13199, 0x00008086 }, /* GL_INDEX_ARRAY_STRIDE */
+   { 13221, 0x00008085 }, /* GL_INDEX_ARRAY_TYPE */
+   { 13241, 0x00000D51 }, /* GL_INDEX_BITS */
+   { 13255, 0x00000C20 }, /* GL_INDEX_CLEAR_VALUE */
+   { 13276, 0x00000BF1 }, /* GL_INDEX_LOGIC_OP */
+   { 13294, 0x00000C30 }, /* GL_INDEX_MODE */
+   { 13308, 0x00000D13 }, /* GL_INDEX_OFFSET */
+   { 13324, 0x00000D12 }, /* GL_INDEX_SHIFT */
+   { 13339, 0x00000C21 }, /* GL_INDEX_WRITEMASK */
+   { 13358, 0x00008B84 }, /* GL_INFO_LOG_LENGTH */
+   { 13377, 0x00001404 }, /* GL_INT */
+   { 13384, 0x00008049 }, /* GL_INTENSITY */
+   { 13397, 0x0000804C }, /* GL_INTENSITY12 */
+   { 13412, 0x0000804C }, /* GL_INTENSITY12_EXT */
+   { 13431, 0x0000804D }, /* GL_INTENSITY16 */
+   { 13446, 0x0000804D }, /* GL_INTENSITY16_EXT */
+   { 13465, 0x0000804A }, /* GL_INTENSITY4 */
+   { 13479, 0x0000804A }, /* GL_INTENSITY4_EXT */
+   { 13497, 0x0000804B }, /* GL_INTENSITY8 */
+   { 13511, 0x0000804B }, /* GL_INTENSITY8_EXT */
+   { 13529, 0x00008049 }, /* GL_INTENSITY_EXT */
+   { 13546, 0x00008575 }, /* GL_INTERPOLATE */
+   { 13561, 0x00008575 }, /* GL_INTERPOLATE_ARB */
+   { 13580, 0x00008575 }, /* GL_INTERPOLATE_EXT */
+   { 13599, 0x00008B53 }, /* GL_INT_VEC2 */
+   { 13611, 0x00008B53 }, /* GL_INT_VEC2_ARB */
+   { 13627, 0x00008B54 }, /* GL_INT_VEC3 */
+   { 13639, 0x00008B54 }, /* GL_INT_VEC3_ARB */
+   { 13655, 0x00008B55 }, /* GL_INT_VEC4 */
+   { 13667, 0x00008B55 }, /* GL_INT_VEC4_ARB */
+   { 13683, 0x00000500 }, /* GL_INVALID_ENUM */
+   { 13699, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION */
+   { 13732, 0x00000506 }, /* GL_INVALID_FRAMEBUFFER_OPERATION_EXT */
+   { 13769, 0x00000502 }, /* GL_INVALID_OPERATION */
+   { 13790, 0x00000501 }, /* GL_INVALID_VALUE */
+   { 13807, 0x0000862B }, /* GL_INVERSE_NV */
+   { 13821, 0x0000862D }, /* GL_INVERSE_TRANSPOSE_NV */
+   { 13845, 0x0000150A }, /* GL_INVERT */
+   { 13855, 0x00001E00 }, /* GL_KEEP */
+   { 13863, 0x00000406 }, /* GL_LEFT */
+   { 13871, 0x00000203 }, /* GL_LEQUAL */
+   { 13881, 0x00000201 }, /* GL_LESS */
+   { 13889, 0x00004000 }, /* GL_LIGHT0 */
+   { 13899, 0x00004001 }, /* GL_LIGHT1 */
+   { 13909, 0x00004002 }, /* GL_LIGHT2 */
+   { 13919, 0x00004003 }, /* GL_LIGHT3 */
+   { 13929, 0x00004004 }, /* GL_LIGHT4 */
+   { 13939, 0x00004005 }, /* GL_LIGHT5 */
+   { 13949, 0x00004006 }, /* GL_LIGHT6 */
+   { 13959, 0x00004007 }, /* GL_LIGHT7 */
+   { 13969, 0x00000B50 }, /* GL_LIGHTING */
+   { 13981, 0x00000040 }, /* GL_LIGHTING_BIT */
+   { 13997, 0x00000B53 }, /* GL_LIGHT_MODEL_AMBIENT */
+   { 14020, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL */
+   { 14049, 0x000081F8 }, /* GL_LIGHT_MODEL_COLOR_CONTROL_EXT */
+   { 14082, 0x00000B51 }, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
+   { 14110, 0x00000B52 }, /* GL_LIGHT_MODEL_TWO_SIDE */
+   { 14134, 0x00001B01 }, /* GL_LINE */
+   { 14142, 0x00002601 }, /* GL_LINEAR */
+   { 14152, 0x00001208 }, /* GL_LINEAR_ATTENUATION */
+   { 14174, 0x00008170 }, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
+   { 14204, 0x0000844F }, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
+   { 14235, 0x00002703 }, /* GL_LINEAR_MIPMAP_LINEAR */
+   { 14259, 0x00002701 }, /* GL_LINEAR_MIPMAP_NEAREST */
+   { 14284, 0x00000001 }, /* GL_LINES */
+   { 14293, 0x00000004 }, /* GL_LINE_BIT */
+   { 14305, 0x00000002 }, /* GL_LINE_LOOP */
+   { 14318, 0x00000707 }, /* GL_LINE_RESET_TOKEN */
+   { 14338, 0x00000B20 }, /* GL_LINE_SMOOTH */
+   { 14353, 0x00000C52 }, /* GL_LINE_SMOOTH_HINT */
+   { 14373, 0x00000B24 }, /* GL_LINE_STIPPLE */
+   { 14389, 0x00000B25 }, /* GL_LINE_STIPPLE_PATTERN */
+   { 14413, 0x00000B26 }, /* GL_LINE_STIPPLE_REPEAT */
+   { 14436, 0x00000003 }, /* GL_LINE_STRIP */
+   { 14450, 0x00000702 }, /* GL_LINE_TOKEN */
+   { 14464, 0x00000B21 }, /* GL_LINE_WIDTH */
+   { 14478, 0x00000B23 }, /* GL_LINE_WIDTH_GRANULARITY */
+   { 14504, 0x00000B22 }, /* GL_LINE_WIDTH_RANGE */
+   { 14524, 0x00008B82 }, /* GL_LINK_STATUS */
+   { 14539, 0x00000B32 }, /* GL_LIST_BASE */
+   { 14552, 0x00020000 }, /* GL_LIST_BIT */
+   { 14564, 0x00000B33 }, /* GL_LIST_INDEX */
+   { 14578, 0x00000B30 }, /* GL_LIST_MODE */
+   { 14591, 0x00000101 }, /* GL_LOAD */
+   { 14599, 0x00000BF1 }, /* GL_LOGIC_OP */
+   { 14611, 0x00000BF0 }, /* GL_LOGIC_OP_MODE */
+   { 14628, 0x00008CA1 }, /* GL_LOWER_LEFT */
+   { 14642, 0x00001909 }, /* GL_LUMINANCE */
+   { 14655, 0x00008041 }, /* GL_LUMINANCE12 */
+   { 14670, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12 */
+   { 14693, 0x00008047 }, /* GL_LUMINANCE12_ALPHA12_EXT */
+   { 14720, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4 */
+   { 14742, 0x00008046 }, /* GL_LUMINANCE12_ALPHA4_EXT */
+   { 14768, 0x00008041 }, /* GL_LUMINANCE12_EXT */
+   { 14787, 0x00008042 }, /* GL_LUMINANCE16 */
+   { 14802, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16 */
+   { 14825, 0x00008048 }, /* GL_LUMINANCE16_ALPHA16_EXT */
+   { 14852, 0x00008042 }, /* GL_LUMINANCE16_EXT */
+   { 14871, 0x0000803F }, /* GL_LUMINANCE4 */
+   { 14885, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4 */
+   { 14906, 0x00008043 }, /* GL_LUMINANCE4_ALPHA4_EXT */
+   { 14931, 0x0000803F }, /* GL_LUMINANCE4_EXT */
+   { 14949, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2 */
+   { 14970, 0x00008044 }, /* GL_LUMINANCE6_ALPHA2_EXT */
+   { 14995, 0x00008040 }, /* GL_LUMINANCE8 */
+   { 15009, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8 */
+   { 15030, 0x00008045 }, /* GL_LUMINANCE8_ALPHA8_EXT */
+   { 15055, 0x00008040 }, /* GL_LUMINANCE8_EXT */
+   { 15073, 0x0000190A }, /* GL_LUMINANCE_ALPHA */
+   { 15092, 0x00000D90 }, /* GL_MAP1_COLOR_4 */
+   { 15108, 0x00000DD0 }, /* GL_MAP1_GRID_DOMAIN */
+   { 15128, 0x00000DD1 }, /* GL_MAP1_GRID_SEGMENTS */
+   { 15150, 0x00000D91 }, /* GL_MAP1_INDEX */
+   { 15164, 0x00000D92 }, /* GL_MAP1_NORMAL */
+   { 15179, 0x00000D93 }, /* GL_MAP1_TEXTURE_COORD_1 */
+   { 15203, 0x00000D94 }, /* GL_MAP1_TEXTURE_COORD_2 */
+   { 15227, 0x00000D95 }, /* GL_MAP1_TEXTURE_COORD_3 */
+   { 15251, 0x00000D96 }, /* GL_MAP1_TEXTURE_COORD_4 */
+   { 15275, 0x00000D97 }, /* GL_MAP1_VERTEX_3 */
+   { 15292, 0x00000D98 }, /* GL_MAP1_VERTEX_4 */
+   { 15309, 0x00008660 }, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
+   { 15337, 0x0000866A }, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
+   { 15366, 0x0000866B }, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
+   { 15395, 0x0000866C }, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
+   { 15424, 0x0000866D }, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
+   { 15453, 0x0000866E }, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
+   { 15482, 0x0000866F }, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
+   { 15511, 0x00008661 }, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
+   { 15539, 0x00008662 }, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
+   { 15567, 0x00008663 }, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
+   { 15595, 0x00008664 }, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
+   { 15623, 0x00008665 }, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
+   { 15651, 0x00008666 }, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
+   { 15679, 0x00008667 }, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
+   { 15707, 0x00008668 }, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
+   { 15735, 0x00008669 }, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
+   { 15763, 0x00000DB0 }, /* GL_MAP2_COLOR_4 */
+   { 15779, 0x00000DD2 }, /* GL_MAP2_GRID_DOMAIN */
+   { 15799, 0x00000DD3 }, /* GL_MAP2_GRID_SEGMENTS */
+   { 15821, 0x00000DB1 }, /* GL_MAP2_INDEX */
+   { 15835, 0x00000DB2 }, /* GL_MAP2_NORMAL */
+   { 15850, 0x00000DB3 }, /* GL_MAP2_TEXTURE_COORD_1 */
+   { 15874, 0x00000DB4 }, /* GL_MAP2_TEXTURE_COORD_2 */
+   { 15898, 0x00000DB5 }, /* GL_MAP2_TEXTURE_COORD_3 */
+   { 15922, 0x00000DB6 }, /* GL_MAP2_TEXTURE_COORD_4 */
+   { 15946, 0x00000DB7 }, /* GL_MAP2_VERTEX_3 */
+   { 15963, 0x00000DB8 }, /* GL_MAP2_VERTEX_4 */
+   { 15980, 0x00008670 }, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
+   { 16008, 0x0000867A }, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
+   { 16037, 0x0000867B }, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
+   { 16066, 0x0000867C }, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
+   { 16095, 0x0000867D }, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
+   { 16124, 0x0000867E }, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
+   { 16153, 0x0000867F }, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
+   { 16182, 0x00008671 }, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
+   { 16210, 0x00008672 }, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
+   { 16238, 0x00008673 }, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
+   { 16266, 0x00008674 }, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
+   { 16294, 0x00008675 }, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
+   { 16322, 0x00008676 }, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
+   { 16350, 0x00008677 }, /* GL_MAP2_VERTEX_ATTRIB7_4_NV */
+   { 16378, 0x00008678 }, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
+   { 16406, 0x00008679 }, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
+   { 16434, 0x00000D10 }, /* GL_MAP_COLOR */
+   { 16447, 0x00000D11 }, /* GL_MAP_STENCIL */
+   { 16462, 0x000088C0 }, /* GL_MATRIX0_ARB */
+   { 16477, 0x00008630 }, /* GL_MATRIX0_NV */
+   { 16491, 0x000088CA }, /* GL_MATRIX10_ARB */
+   { 16507, 0x000088CB }, /* GL_MATRIX11_ARB */
+   { 16523, 0x000088CC }, /* GL_MATRIX12_ARB */
+   { 16539, 0x000088CD }, /* GL_MATRIX13_ARB */
+   { 16555, 0x000088CE }, /* GL_MATRIX14_ARB */
+   { 16571, 0x000088CF }, /* GL_MATRIX15_ARB */
+   { 16587, 0x000088D0 }, /* GL_MATRIX16_ARB */
+   { 16603, 0x000088D1 }, /* GL_MATRIX17_ARB */
+   { 16619, 0x000088D2 }, /* GL_MATRIX18_ARB */
+   { 16635, 0x000088D3 }, /* GL_MATRIX19_ARB */
+   { 16651, 0x000088C1 }, /* GL_MATRIX1_ARB */
+   { 16666, 0x00008631 }, /* GL_MATRIX1_NV */
+   { 16680, 0x000088D4 }, /* GL_MATRIX20_ARB */
+   { 16696, 0x000088D5 }, /* GL_MATRIX21_ARB */
+   { 16712, 0x000088D6 }, /* GL_MATRIX22_ARB */
+   { 16728, 0x000088D7 }, /* GL_MATRIX23_ARB */
+   { 16744, 0x000088D8 }, /* GL_MATRIX24_ARB */
+   { 16760, 0x000088D9 }, /* GL_MATRIX25_ARB */
+   { 16776, 0x000088DA }, /* GL_MATRIX26_ARB */
+   { 16792, 0x000088DB }, /* GL_MATRIX27_ARB */
+   { 16808, 0x000088DC }, /* GL_MATRIX28_ARB */
+   { 16824, 0x000088DD }, /* GL_MATRIX29_ARB */
+   { 16840, 0x000088C2 }, /* GL_MATRIX2_ARB */
+   { 16855, 0x00008632 }, /* GL_MATRIX2_NV */
+   { 16869, 0x000088DE }, /* GL_MATRIX30_ARB */
+   { 16885, 0x000088DF }, /* GL_MATRIX31_ARB */
+   { 16901, 0x000088C3 }, /* GL_MATRIX3_ARB */
+   { 16916, 0x00008633 }, /* GL_MATRIX3_NV */
+   { 16930, 0x000088C4 }, /* GL_MATRIX4_ARB */
+   { 16945, 0x00008634 }, /* GL_MATRIX4_NV */
+   { 16959, 0x000088C5 }, /* GL_MATRIX5_ARB */
+   { 16974, 0x00008635 }, /* GL_MATRIX5_NV */
+   { 16988, 0x000088C6 }, /* GL_MATRIX6_ARB */
+   { 17003, 0x00008636 }, /* GL_MATRIX6_NV */
+   { 17017, 0x000088C7 }, /* GL_MATRIX7_ARB */
+   { 17032, 0x00008637 }, /* GL_MATRIX7_NV */
+   { 17046, 0x000088C8 }, /* GL_MATRIX8_ARB */
+   { 17061, 0x000088C9 }, /* GL_MATRIX9_ARB */
+   { 17076, 0x00008844 }, /* GL_MATRIX_INDEX_ARRAY_ARB */
+   { 17102, 0x00008849 }, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
+   { 17136, 0x00008846 }, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
+   { 17167, 0x00008848 }, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
+   { 17200, 0x00008847 }, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
+   { 17231, 0x00000BA0 }, /* GL_MATRIX_MODE */
+   { 17246, 0x00008840 }, /* GL_MATRIX_PALETTE_ARB */
+   { 17268, 0x00008008 }, /* GL_MAX */
+   { 17275, 0x00008073 }, /* GL_MAX_3D_TEXTURE_SIZE */
+   { 17298, 0x000088FF }, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
+   { 17330, 0x00000D35 }, /* GL_MAX_ATTRIB_STACK_DEPTH */
+   { 17356, 0x00000D3B }, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
+   { 17389, 0x00008177 }, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
+   { 17415, 0x00008178 }, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+   { 17449, 0x00000D32 }, /* GL_MAX_CLIP_PLANES */
+   { 17468, 0x00008CDF }, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
+   { 17497, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
+   { 17529, 0x000080B3 }, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH_SGI */
+   { 17565, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
+   { 17601, 0x00008B4D }, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS_ARB */
+   { 17641, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT */
+   { 17667, 0x0000801B }, /* GL_MAX_CONVOLUTION_HEIGHT_EXT */
+   { 17697, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH */
+   { 17722, 0x0000801A }, /* GL_MAX_CONVOLUTION_WIDTH_EXT */
+   { 17751, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
+   { 17780, 0x0000851C }, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE_ARB */
+   { 17813, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS */
+   { 17833, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ARB */
+   { 17857, 0x00008824 }, /* GL_MAX_DRAW_BUFFERS_ATI */
+   { 17881, 0x000080E9 }, /* GL_MAX_ELEMENTS_INDICES */
+   { 17905, 0x000080E8 }, /* GL_MAX_ELEMENTS_VERTICES */
+   { 17930, 0x00000D30 }, /* GL_MAX_EVAL_ORDER */
+   { 17948, 0x00008008 }, /* GL_MAX_EXT */
+   { 17959, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
+   { 17994, 0x00008B49 }, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS_ARB */
+   { 18033, 0x00000D31 }, /* GL_MAX_LIGHTS */
+   { 18047, 0x00000B31 }, /* GL_MAX_LIST_NESTING */
+   { 18067, 0x00008841 }, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
+   { 18105, 0x00000D36 }, /* GL_MAX_MODELVIEW_STACK_DEPTH */
+   { 18134, 0x00000D37 }, /* GL_MAX_NAME_STACK_DEPTH */
+   { 18158, 0x00008842 }, /* GL_MAX_PALETTE_MATRICES_ARB */
+   { 18186, 0x00000D34 }, /* GL_MAX_PIXEL_MAP_TABLE */
+   { 18209, 0x000088B1 }, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
+   { 18246, 0x0000880B }, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
+   { 18282, 0x000088AD }, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
+   { 18309, 0x000088F5 }, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
+   { 18338, 0x000088B5 }, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
+   { 18372, 0x000088F4 }, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
+   { 18408, 0x000088F6 }, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
+   { 18435, 0x000088A1 }, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
+   { 18467, 0x000088B4 }, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
+   { 18503, 0x000088F8 }, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
+   { 18532, 0x000088F7 }, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
+   { 18561, 0x0000862F }, /* GL_MAX_PROGRAM_MATRICES_ARB */
+   { 18589, 0x0000862E }, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
+   { 18627, 0x000088B3 }, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+   { 18671, 0x0000880E }, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+   { 18714, 0x000088AF }, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
+   { 18748, 0x000088A3 }, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+   { 18787, 0x000088AB }, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
+   { 18824, 0x000088A7 }, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
+   { 18862, 0x00008810 }, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+   { 18905, 0x0000880F }, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+   { 18948, 0x000088A9 }, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
+   { 18978, 0x000088A5 }, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
+   { 19009, 0x0000880D }, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
+   { 19045, 0x0000880C }, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
+   { 19081, 0x00000D38 }, /* GL_MAX_PROJECTION_STACK_DEPTH */
+   { 19111, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
+   { 19145, 0x000084F8 }, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_NV */
+   { 19178, 0x000084E8 }, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
+   { 19207, 0x00008D57 }, /* GL_MAX_SAMPLES */
+   { 19222, 0x00008504 }, /* GL_MAX_SHININESS_NV */
+   { 19242, 0x00008505 }, /* GL_MAX_SPOT_EXPONENT_NV */
+   { 19266, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS */
+   { 19288, 0x00008871 }, /* GL_MAX_TEXTURE_COORDS_ARB */
+   { 19314, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS */
+   { 19341, 0x00008872 }, /* GL_MAX_TEXTURE_IMAGE_UNITS_ARB */
+   { 19372, 0x000084FD }, /* GL_MAX_TEXTURE_LOD_BIAS */
+   { 19396, 0x000084FF }, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
+   { 19430, 0x00000D33 }, /* GL_MAX_TEXTURE_SIZE */
+   { 19450, 0x00000D39 }, /* GL_MAX_TEXTURE_STACK_DEPTH */
+   { 19477, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS */
+   { 19498, 0x000084E2 }, /* GL_MAX_TEXTURE_UNITS_ARB */
+   { 19523, 0x0000862F }, /* GL_MAX_TRACK_MATRICES_NV */
+   { 19548, 0x0000862E }, /* GL_MAX_TRACK_MATRIX_STACK_DEPTH_NV */
+   { 19583, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS */
+   { 19605, 0x00008B4B }, /* GL_MAX_VARYING_FLOATS_ARB */
+   { 19631, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS */
+   { 19653, 0x00008869 }, /* GL_MAX_VERTEX_ATTRIBS_ARB */
+   { 19679, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
+   { 19713, 0x00008B4C }, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS_ARB */
+   { 19751, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
+   { 19784, 0x00008B4A }, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS_ARB */
+   { 19821, 0x000086A4 }, /* GL_MAX_VERTEX_UNITS_ARB */
+   { 19845, 0x00000D3A }, /* GL_MAX_VIEWPORT_DIMS */
+   { 19866, 0x00008007 }, /* GL_MIN */
+   { 19873, 0x0000802E }, /* GL_MINMAX */
+   { 19883, 0x0000802E }, /* GL_MINMAX_EXT */
+   { 19897, 0x0000802F }, /* GL_MINMAX_FORMAT */
+   { 19914, 0x0000802F }, /* GL_MINMAX_FORMAT_EXT */
+   { 19935, 0x00008030 }, /* GL_MINMAX_SINK */
+   { 19950, 0x00008030 }, /* GL_MINMAX_SINK_EXT */
+   { 19969, 0x00008007 }, /* GL_MIN_EXT */
+   { 19980, 0x00008370 }, /* GL_MIRRORED_REPEAT */
+   { 19999, 0x00008370 }, /* GL_MIRRORED_REPEAT_ARB */
+   { 20022, 0x00008370 }, /* GL_MIRRORED_REPEAT_IBM */
+   { 20045, 0x00008742 }, /* GL_MIRROR_CLAMP_ATI */
+   { 20065, 0x00008742 }, /* GL_MIRROR_CLAMP_EXT */
+   { 20085, 0x00008912 }, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
+   { 20115, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_ATI */
+   { 20143, 0x00008743 }, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
+   { 20171, 0x00001700 }, /* GL_MODELVIEW */
+   { 20184, 0x00001700 }, /* GL_MODELVIEW0_ARB */
+   { 20202, 0x0000872A }, /* GL_MODELVIEW10_ARB */
+   { 20221, 0x0000872B }, /* GL_MODELVIEW11_ARB */
+   { 20240, 0x0000872C }, /* GL_MODELVIEW12_ARB */
+   { 20259, 0x0000872D }, /* GL_MODELVIEW13_ARB */
+   { 20278, 0x0000872E }, /* GL_MODELVIEW14_ARB */
+   { 20297, 0x0000872F }, /* GL_MODELVIEW15_ARB */
+   { 20316, 0x00008730 }, /* GL_MODELVIEW16_ARB */
+   { 20335, 0x00008731 }, /* GL_MODELVIEW17_ARB */
+   { 20354, 0x00008732 }, /* GL_MODELVIEW18_ARB */
+   { 20373, 0x00008733 }, /* GL_MODELVIEW19_ARB */
+   { 20392, 0x0000850A }, /* GL_MODELVIEW1_ARB */
+   { 20410, 0x00008734 }, /* GL_MODELVIEW20_ARB */
+   { 20429, 0x00008735 }, /* GL_MODELVIEW21_ARB */
+   { 20448, 0x00008736 }, /* GL_MODELVIEW22_ARB */
+   { 20467, 0x00008737 }, /* GL_MODELVIEW23_ARB */
+   { 20486, 0x00008738 }, /* GL_MODELVIEW24_ARB */
+   { 20505, 0x00008739 }, /* GL_MODELVIEW25_ARB */
+   { 20524, 0x0000873A }, /* GL_MODELVIEW26_ARB */
+   { 20543, 0x0000873B }, /* GL_MODELVIEW27_ARB */
+   { 20562, 0x0000873C }, /* GL_MODELVIEW28_ARB */
+   { 20581, 0x0000873D }, /* GL_MODELVIEW29_ARB */
+   { 20600, 0x00008722 }, /* GL_MODELVIEW2_ARB */
+   { 20618, 0x0000873E }, /* GL_MODELVIEW30_ARB */
+   { 20637, 0x0000873F }, /* GL_MODELVIEW31_ARB */
+   { 20656, 0x00008723 }, /* GL_MODELVIEW3_ARB */
+   { 20674, 0x00008724 }, /* GL_MODELVIEW4_ARB */
+   { 20692, 0x00008725 }, /* GL_MODELVIEW5_ARB */
+   { 20710, 0x00008726 }, /* GL_MODELVIEW6_ARB */
+   { 20728, 0x00008727 }, /* GL_MODELVIEW7_ARB */
+   { 20746, 0x00008728 }, /* GL_MODELVIEW8_ARB */
+   { 20764, 0x00008729 }, /* GL_MODELVIEW9_ARB */
+   { 20782, 0x00000BA6 }, /* GL_MODELVIEW_MATRIX */
+   { 20802, 0x00008629 }, /* GL_MODELVIEW_PROJECTION_NV */
+   { 20829, 0x00000BA3 }, /* GL_MODELVIEW_STACK_DEPTH */
+   { 20854, 0x00002100 }, /* GL_MODULATE */
+   { 20866, 0x00008744 }, /* GL_MODULATE_ADD_ATI */
+   { 20886, 0x00008745 }, /* GL_MODULATE_SIGNED_ADD_ATI */
+   { 20913, 0x00008746 }, /* GL_MODULATE_SUBTRACT_ATI */
+   { 20938, 0x00000103 }, /* GL_MULT */
+   { 20946, 0x0000809D }, /* GL_MULTISAMPLE */
+   { 20961, 0x000086B2 }, /* GL_MULTISAMPLE_3DFX */
+   { 20981, 0x0000809D }, /* GL_MULTISAMPLE_ARB */
+   { 21000, 0x20000000 }, /* GL_MULTISAMPLE_BIT */
+   { 21019, 0x20000000 }, /* GL_MULTISAMPLE_BIT_3DFX */
+   { 21043, 0x20000000 }, /* GL_MULTISAMPLE_BIT_ARB */
+   { 21066, 0x00008534 }, /* GL_MULTISAMPLE_FILTER_HINT_NV */
+   { 21096, 0x00002A25 }, /* GL_N3F_V3F */
+   { 21107, 0x00000D70 }, /* GL_NAME_STACK_DEPTH */
+   { 21127, 0x0000150E }, /* GL_NAND */
+   { 21135, 0x00002600 }, /* GL_NEAREST */
+   { 21146, 0x0000844E }, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
+   { 21177, 0x0000844D }, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
+   { 21209, 0x00002702 }, /* GL_NEAREST_MIPMAP_LINEAR */
+   { 21234, 0x00002700 }, /* GL_NEAREST_MIPMAP_NEAREST */
+   { 21260, 0x00000200 }, /* GL_NEVER */
+   { 21269, 0x00001102 }, /* GL_NICEST */
+   { 21279, 0x00000000 }, /* GL_NONE */
+   { 21287, 0x00001505 }, /* GL_NOOP */
+   { 21295, 0x00001508 }, /* GL_NOR */
+   { 21302, 0x00000BA1 }, /* GL_NORMALIZE */
+   { 21315, 0x00008075 }, /* GL_NORMAL_ARRAY */
+   { 21331, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
+   { 21362, 0x00008897 }, /* GL_NORMAL_ARRAY_BUFFER_BINDING_ARB */
+   { 21397, 0x0000808F }, /* GL_NORMAL_ARRAY_POINTER */
+   { 21421, 0x0000807F }, /* GL_NORMAL_ARRAY_STRIDE */
+   { 21444, 0x0000807E }, /* GL_NORMAL_ARRAY_TYPE */
+   { 21465, 0x00008511 }, /* GL_NORMAL_MAP */
+   { 21479, 0x00008511 }, /* GL_NORMAL_MAP_ARB */
+   { 21497, 0x00008511 }, /* GL_NORMAL_MAP_NV */
+   { 21514, 0x00000205 }, /* GL_NOTEQUAL */
+   { 21526, 0x00000000 }, /* GL_NO_ERROR */
+   { 21538, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
+   { 21572, 0x000086A2 }, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB */
+   { 21610, 0x00008B89 }, /* GL_OBJECT_ACTIVE_ATTRIBUTES_ARB */
+   { 21642, 0x00008B8A }, /* GL_OBJECT_ACTIVE_ATTRIBUTE_MAX_LENGTH_ARB */
+   { 21684, 0x00008B86 }, /* GL_OBJECT_ACTIVE_UNIFORMS_ARB */
+   { 21714, 0x00008B87 }, /* GL_OBJECT_ACTIVE_UNIFORM_MAX_LENGTH_ARB */
+   { 21754, 0x00008B85 }, /* GL_OBJECT_ATTACHED_OBJECTS_ARB */
+   { 21785, 0x00008B81 }, /* GL_OBJECT_COMPILE_STATUS_ARB */
+   { 21814, 0x00008B80 }, /* GL_OBJECT_DELETE_STATUS_ARB */
+   { 21842, 0x00008B84 }, /* GL_OBJECT_INFO_LOG_LENGTH_ARB */
+   { 21872, 0x00002401 }, /* GL_OBJECT_LINEAR */
+   { 21889, 0x00008B82 }, /* GL_OBJECT_LINK_STATUS_ARB */
+   { 21915, 0x00002501 }, /* GL_OBJECT_PLANE */
+   { 21931, 0x00008B88 }, /* GL_OBJECT_SHADER_SOURCE_LENGTH_ARB */
+   { 21966, 0x00008B4F }, /* GL_OBJECT_SUBTYPE_ARB */
+   { 21988, 0x00008B4E }, /* GL_OBJECT_TYPE_ARB */
+   { 22007, 0x00008B83 }, /* GL_OBJECT_VALIDATE_STATUS_ARB */
+   { 22037, 0x00008165 }, /* GL_OCCLUSION_TEST_HP */
+   { 22058, 0x00008166 }, /* GL_OCCLUSION_TEST_RESULT_HP */
+   { 22086, 0x00000001 }, /* GL_ONE */
+   { 22093, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA */
+   { 22121, 0x00008004 }, /* GL_ONE_MINUS_CONSTANT_ALPHA_EXT */
+   { 22153, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR */
+   { 22181, 0x00008002 }, /* GL_ONE_MINUS_CONSTANT_COLOR_EXT */
+   { 22213, 0x00000305 }, /* GL_ONE_MINUS_DST_ALPHA */
+   { 22236, 0x00000307 }, /* GL_ONE_MINUS_DST_COLOR */
+   { 22259, 0x00000303 }, /* GL_ONE_MINUS_SRC_ALPHA */
+   { 22282, 0x00000301 }, /* GL_ONE_MINUS_SRC_COLOR */
+   { 22305, 0x00008598 }, /* GL_OPERAND0_ALPHA */
+   { 22323, 0x00008598 }, /* GL_OPERAND0_ALPHA_ARB */
+   { 22345, 0x00008598 }, /* GL_OPERAND0_ALPHA_EXT */
+   { 22367, 0x00008590 }, /* GL_OPERAND0_RGB */
+   { 22383, 0x00008590 }, /* GL_OPERAND0_RGB_ARB */
+   { 22403, 0x00008590 }, /* GL_OPERAND0_RGB_EXT */
+   { 22423, 0x00008599 }, /* GL_OPERAND1_ALPHA */
+   { 22441, 0x00008599 }, /* GL_OPERAND1_ALPHA_ARB */
+   { 22463, 0x00008599 }, /* GL_OPERAND1_ALPHA_EXT */
+   { 22485, 0x00008591 }, /* GL_OPERAND1_RGB */
+   { 22501, 0x00008591 }, /* GL_OPERAND1_RGB_ARB */
+   { 22521, 0x00008591 }, /* GL_OPERAND1_RGB_EXT */
+   { 22541, 0x0000859A }, /* GL_OPERAND2_ALPHA */
+   { 22559, 0x0000859A }, /* GL_OPERAND2_ALPHA_ARB */
+   { 22581, 0x0000859A }, /* GL_OPERAND2_ALPHA_EXT */
+   { 22603, 0x00008592 }, /* GL_OPERAND2_RGB */
+   { 22619, 0x00008592 }, /* GL_OPERAND2_RGB_ARB */
+   { 22639, 0x00008592 }, /* GL_OPERAND2_RGB_EXT */
+   { 22659, 0x0000859B }, /* GL_OPERAND3_ALPHA_NV */
+   { 22680, 0x00008593 }, /* GL_OPERAND3_RGB_NV */
+   { 22699, 0x00001507 }, /* GL_OR */
+   { 22705, 0x00000A01 }, /* GL_ORDER */
+   { 22714, 0x0000150D }, /* GL_OR_INVERTED */
+   { 22729, 0x0000150B }, /* GL_OR_REVERSE */
+   { 22743, 0x00000505 }, /* GL_OUT_OF_MEMORY */
+   { 22760, 0x00000D05 }, /* GL_PACK_ALIGNMENT */
+   { 22778, 0x0000806C }, /* GL_PACK_IMAGE_HEIGHT */
+   { 22799, 0x00008758 }, /* GL_PACK_INVERT_MESA */
+   { 22819, 0x00000D01 }, /* GL_PACK_LSB_FIRST */
+   { 22837, 0x00000D02 }, /* GL_PACK_ROW_LENGTH */
+   { 22856, 0x0000806B }, /* GL_PACK_SKIP_IMAGES */
+   { 22876, 0x00000D04 }, /* GL_PACK_SKIP_PIXELS */
+   { 22896, 0x00000D03 }, /* GL_PACK_SKIP_ROWS */
+   { 22914, 0x00000D00 }, /* GL_PACK_SWAP_BYTES */
+   { 22933, 0x00008B92 }, /* GL_PALETTE4_R5_G6_B5_OES */
+   { 22958, 0x00008B94 }, /* GL_PALETTE4_RGB5_A1_OES */
+   { 22982, 0x00008B90 }, /* GL_PALETTE4_RGB8_OES */
+   { 23003, 0x00008B93 }, /* GL_PALETTE4_RGBA4_OES */
+   { 23025, 0x00008B91 }, /* GL_PALETTE4_RGBA8_OES */
+   { 23047, 0x00008B97 }, /* GL_PALETTE8_R5_G6_B5_OES */
+   { 23072, 0x00008B99 }, /* GL_PALETTE8_RGB5_A1_OES */
+   { 23096, 0x00008B95 }, /* GL_PALETTE8_RGB8_OES */
+   { 23117, 0x00008B98 }, /* GL_PALETTE8_RGBA4_OES */
+   { 23139, 0x00008B96 }, /* GL_PALETTE8_RGBA8_OES */
+   { 23161, 0x00000700 }, /* GL_PASS_THROUGH_TOKEN */
+   { 23183, 0x00000C50 }, /* GL_PERSPECTIVE_CORRECTION_HINT */
+   { 23214, 0x00000C79 }, /* GL_PIXEL_MAP_A_TO_A */
+   { 23234, 0x00000CB9 }, /* GL_PIXEL_MAP_A_TO_A_SIZE */
+   { 23259, 0x00000C78 }, /* GL_PIXEL_MAP_B_TO_B */
+   { 23279, 0x00000CB8 }, /* GL_PIXEL_MAP_B_TO_B_SIZE */
+   { 23304, 0x00000C77 }, /* GL_PIXEL_MAP_G_TO_G */
+   { 23324, 0x00000CB7 }, /* GL_PIXEL_MAP_G_TO_G_SIZE */
+   { 23349, 0x00000C75 }, /* GL_PIXEL_MAP_I_TO_A */
+   { 23369, 0x00000CB5 }, /* GL_PIXEL_MAP_I_TO_A_SIZE */
+   { 23394, 0x00000C74 }, /* GL_PIXEL_MAP_I_TO_B */
+   { 23414, 0x00000CB4 }, /* GL_PIXEL_MAP_I_TO_B_SIZE */
+   { 23439, 0x00000C73 }, /* GL_PIXEL_MAP_I_TO_G */
+   { 23459, 0x00000CB3 }, /* GL_PIXEL_MAP_I_TO_G_SIZE */
+   { 23484, 0x00000C70 }, /* GL_PIXEL_MAP_I_TO_I */
+   { 23504, 0x00000CB0 }, /* GL_PIXEL_MAP_I_TO_I_SIZE */
+   { 23529, 0x00000C72 }, /* GL_PIXEL_MAP_I_TO_R */
+   { 23549, 0x00000CB2 }, /* GL_PIXEL_MAP_I_TO_R_SIZE */
+   { 23574, 0x00000C76 }, /* GL_PIXEL_MAP_R_TO_R */
+   { 23594, 0x00000CB6 }, /* GL_PIXEL_MAP_R_TO_R_SIZE */
+   { 23619, 0x00000C71 }, /* GL_PIXEL_MAP_S_TO_S */
+   { 23639, 0x00000CB1 }, /* GL_PIXEL_MAP_S_TO_S_SIZE */
+   { 23664, 0x00000020 }, /* GL_PIXEL_MODE_BIT */
+   { 23682, 0x000088ED }, /* GL_PIXEL_PACK_BUFFER_BINDING_EXT */
+   { 23715, 0x000088EB }, /* GL_PIXEL_PACK_BUFFER_EXT */
+   { 23740, 0x000088EF }, /* GL_PIXEL_UNPACK_BUFFER_BINDING_EXT */
+   { 23775, 0x000088EC }, /* GL_PIXEL_UNPACK_BUFFER_EXT */
+   { 23802, 0x00001B00 }, /* GL_POINT */
+   { 23811, 0x00000000 }, /* GL_POINTS */
+   { 23821, 0x00000002 }, /* GL_POINT_BIT */
+   { 23834, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION */
+   { 23864, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_ARB */
+   { 23898, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_EXT */
+   { 23932, 0x00008129 }, /* GL_POINT_DISTANCE_ATTENUATION_SGIS */
+   { 23967, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE */
+   { 23996, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_ARB */
+   { 24029, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_EXT */
+   { 24062, 0x00008128 }, /* GL_POINT_FADE_THRESHOLD_SIZE_SGIS */
+   { 24096, 0x00000B11 }, /* GL_POINT_SIZE */
+   { 24110, 0x00000B13 }, /* GL_POINT_SIZE_GRANULARITY */
+   { 24136, 0x00008127 }, /* GL_POINT_SIZE_MAX */
+   { 24154, 0x00008127 }, /* GL_POINT_SIZE_MAX_ARB */
+   { 24176, 0x00008127 }, /* GL_POINT_SIZE_MAX_EXT */
+   { 24198, 0x00008127 }, /* GL_POINT_SIZE_MAX_SGIS */
+   { 24221, 0x00008126 }, /* GL_POINT_SIZE_MIN */
+   { 24239, 0x00008126 }, /* GL_POINT_SIZE_MIN_ARB */
+   { 24261, 0x00008126 }, /* GL_POINT_SIZE_MIN_EXT */
+   { 24283, 0x00008126 }, /* GL_POINT_SIZE_MIN_SGIS */
+   { 24306, 0x00000B12 }, /* GL_POINT_SIZE_RANGE */
+   { 24326, 0x00000B10 }, /* GL_POINT_SMOOTH */
+   { 24342, 0x00000C51 }, /* GL_POINT_SMOOTH_HINT */
+   { 24363, 0x00008861 }, /* GL_POINT_SPRITE */
+   { 24379, 0x00008861 }, /* GL_POINT_SPRITE_ARB */
+   { 24399, 0x00008CA0 }, /* GL_POINT_SPRITE_COORD_ORIGIN */
+   { 24428, 0x00008861 }, /* GL_POINT_SPRITE_NV */
+   { 24447, 0x00008863 }, /* GL_POINT_SPRITE_R_MODE_NV */
+   { 24473, 0x00000701 }, /* GL_POINT_TOKEN */
+   { 24488, 0x00000009 }, /* GL_POLYGON */
+   { 24499, 0x00000008 }, /* GL_POLYGON_BIT */
+   { 24514, 0x00000B40 }, /* GL_POLYGON_MODE */
+   { 24530, 0x00008039 }, /* GL_POLYGON_OFFSET_BIAS */
+   { 24553, 0x00008038 }, /* GL_POLYGON_OFFSET_FACTOR */
+   { 24578, 0x00008037 }, /* GL_POLYGON_OFFSET_FILL */
+   { 24601, 0x00002A02 }, /* GL_POLYGON_OFFSET_LINE */
+   { 24624, 0x00002A01 }, /* GL_POLYGON_OFFSET_POINT */
+   { 24648, 0x00002A00 }, /* GL_POLYGON_OFFSET_UNITS */
+   { 24672, 0x00000B41 }, /* GL_POLYGON_SMOOTH */
+   { 24690, 0x00000C53 }, /* GL_POLYGON_SMOOTH_HINT */
+   { 24713, 0x00000B42 }, /* GL_POLYGON_STIPPLE */
+   { 24732, 0x00000010 }, /* GL_POLYGON_STIPPLE_BIT */
+   { 24755, 0x00000703 }, /* GL_POLYGON_TOKEN */
+   { 24772, 0x00001203 }, /* GL_POSITION */
+   { 24784, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
+   { 24816, 0x000080BB }, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS_SGI */
+   { 24852, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
+   { 24885, 0x000080B7 }, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE_SGI */
+   { 24922, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
+   { 24953, 0x000080BA }, /* GL_POST_COLOR_MATRIX_BLUE_BIAS_SGI */
+   { 24988, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
+   { 25020, 0x000080B6 }, /* GL_POST_COLOR_MATRIX_BLUE_SCALE_SGI */
+   { 25056, 0x000080D2 }, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
+   { 25089, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
+   { 25121, 0x000080B9 }, /* GL_POST_COLOR_MATRIX_GREEN_BIAS_SGI */
+   { 25157, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
+   { 25190, 0x000080B5 }, /* GL_POST_COLOR_MATRIX_GREEN_SCALE_SGI */
+   { 25227, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS */
+   { 25257, 0x000080B8 }, /* GL_POST_COLOR_MATRIX_RED_BIAS_SGI */
+   { 25291, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE */
+   { 25322, 0x000080B4 }, /* GL_POST_COLOR_MATRIX_RED_SCALE_SGI */
+   { 25357, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
+   { 25388, 0x00008023 }, /* GL_POST_CONVOLUTION_ALPHA_BIAS_EXT */
+   { 25423, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
+   { 25455, 0x0000801F }, /* GL_POST_CONVOLUTION_ALPHA_SCALE_EXT */
+   { 25491, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS */
+   { 25521, 0x00008022 }, /* GL_POST_CONVOLUTION_BLUE_BIAS_EXT */
+   { 25555, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE */
+   { 25586, 0x0000801E }, /* GL_POST_CONVOLUTION_BLUE_SCALE_EXT */
+   { 25621, 0x000080D1 }, /* GL_POST_CONVOLUTION_COLOR_TABLE */
+   { 25653, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS */
+   { 25684, 0x00008021 }, /* GL_POST_CONVOLUTION_GREEN_BIAS_EXT */
+   { 25719, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE */
+   { 25751, 0x0000801D }, /* GL_POST_CONVOLUTION_GREEN_SCALE_EXT */
+   { 25787, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS */
+   { 25816, 0x00008020 }, /* GL_POST_CONVOLUTION_RED_BIAS_EXT */
+   { 25849, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE */
+   { 25879, 0x0000801C }, /* GL_POST_CONVOLUTION_RED_SCALE_EXT */
+   { 25913, 0x0000817B }, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
+   { 25952, 0x00008179 }, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
+   { 25985, 0x0000817C }, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
+   { 26025, 0x0000817A }, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
+   { 26059, 0x00008578 }, /* GL_PREVIOUS */
+   { 26071, 0x00008578 }, /* GL_PREVIOUS_ARB */
+   { 26087, 0x00008578 }, /* GL_PREVIOUS_EXT */
+   { 26103, 0x00008577 }, /* GL_PRIMARY_COLOR */
+   { 26120, 0x00008577 }, /* GL_PRIMARY_COLOR_ARB */
+   { 26141, 0x00008577 }, /* GL_PRIMARY_COLOR_EXT */
+   { 26162, 0x000088B0 }, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
+   { 26195, 0x00008805 }, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
+   { 26227, 0x000088AC }, /* GL_PROGRAM_ATTRIBS_ARB */
+   { 26250, 0x00008677 }, /* GL_PROGRAM_BINDING_ARB */
+   { 26273, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_ARB */
+   { 26303, 0x0000864B }, /* GL_PROGRAM_ERROR_POSITION_NV */
+   { 26332, 0x00008874 }, /* GL_PROGRAM_ERROR_STRING_ARB */
+   { 26360, 0x00008876 }, /* GL_PROGRAM_FORMAT_ARB */
+   { 26382, 0x00008875 }, /* GL_PROGRAM_FORMAT_ASCII_ARB */
+   { 26410, 0x000088A0 }, /* GL_PROGRAM_INSTRUCTIONS_ARB */
+   { 26438, 0x00008627 }, /* GL_PROGRAM_LENGTH_ARB */
+   { 26460, 0x00008627 }, /* GL_PROGRAM_LENGTH_NV */
+   { 26481, 0x000088B2 }, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+   { 26521, 0x00008808 }, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+   { 26560, 0x000088AE }, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
+   { 26590, 0x000088A2 }, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+   { 26625, 0x000088AA }, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
+   { 26658, 0x000088A6 }, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
+   { 26692, 0x0000880A }, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+   { 26731, 0x00008809 }, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+   { 26770, 0x00008B40 }, /* GL_PROGRAM_OBJECT_ARB */
+   { 26792, 0x000088A8 }, /* GL_PROGRAM_PARAMETERS_ARB */
+   { 26818, 0x00008644 }, /* GL_PROGRAM_PARAMETER_NV */
+   { 26842, 0x00008647 }, /* GL_PROGRAM_RESIDENT_NV */
+   { 26865, 0x00008628 }, /* GL_PROGRAM_STRING_ARB */
+   { 26887, 0x00008628 }, /* GL_PROGRAM_STRING_NV */
+   { 26908, 0x00008646 }, /* GL_PROGRAM_TARGET_NV */
+   { 26929, 0x000088A4 }, /* GL_PROGRAM_TEMPORARIES_ARB */
+   { 26956, 0x00008807 }, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
+   { 26988, 0x00008806 }, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
+   { 27020, 0x000088B6 }, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
+   { 27055, 0x00001701 }, /* GL_PROJECTION */
+   { 27069, 0x00000BA7 }, /* GL_PROJECTION_MATRIX */
+   { 27090, 0x00000BA4 }, /* GL_PROJECTION_STACK_DEPTH */
+   { 27116, 0x000080D3 }, /* GL_PROXY_COLOR_TABLE */
+   { 27137, 0x00008025 }, /* GL_PROXY_HISTOGRAM */
+   { 27156, 0x00008025 }, /* GL_PROXY_HISTOGRAM_EXT */
+   { 27179, 0x000080D5 }, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
+   { 27218, 0x000080D4 }, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
+   { 27256, 0x00008063 }, /* GL_PROXY_TEXTURE_1D */
+   { 27276, 0x00008C19 }, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
+   { 27306, 0x00008063 }, /* GL_PROXY_TEXTURE_1D_EXT */
+   { 27330, 0x00008064 }, /* GL_PROXY_TEXTURE_2D */
+   { 27350, 0x00008C1B }, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
+   { 27380, 0x00008064 }, /* GL_PROXY_TEXTURE_2D_EXT */
+   { 27404, 0x00008070 }, /* GL_PROXY_TEXTURE_3D */
+   { 27424, 0x000080BD }, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
+   { 27457, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP */
+   { 27483, 0x0000851B }, /* GL_PROXY_TEXTURE_CUBE_MAP_ARB */
+   { 27513, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
+   { 27544, 0x000084F7 }, /* GL_PROXY_TEXTURE_RECTANGLE_NV */
+   { 27574, 0x00002003 }, /* GL_Q */
+   { 27579, 0x00001209 }, /* GL_QUADRATIC_ATTENUATION */
+   { 27604, 0x00000007 }, /* GL_QUADS */
+   { 27613, 0x00008614 }, /* GL_QUAD_MESH_SUN */
+   { 27630, 0x00000008 }, /* GL_QUAD_STRIP */
+   { 27644, 0x00008864 }, /* GL_QUERY_COUNTER_BITS */
+   { 27666, 0x00008864 }, /* GL_QUERY_COUNTER_BITS_ARB */
+   { 27692, 0x00008866 }, /* GL_QUERY_RESULT */
+   { 27708, 0x00008866 }, /* GL_QUERY_RESULT_ARB */
+   { 27728, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE */
+   { 27754, 0x00008867 }, /* GL_QUERY_RESULT_AVAILABLE_ARB */
+   { 27784, 0x00002002 }, /* GL_R */
+   { 27789, 0x00002A10 }, /* GL_R3_G3_B2 */
+   { 27801, 0x00019262 }, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
+   { 27834, 0x00000C02 }, /* GL_READ_BUFFER */
+   { 27849, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER */
+   { 27869, 0x00008CAA }, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
+   { 27901, 0x00008CA8 }, /* GL_READ_FRAMEBUFFER_EXT */
+   { 27925, 0x000088B8 }, /* GL_READ_ONLY */
+   { 27938, 0x000088B8 }, /* GL_READ_ONLY_ARB */
+   { 27955, 0x000088BA }, /* GL_READ_WRITE */
+   { 27969, 0x000088BA }, /* GL_READ_WRITE_ARB */
+   { 27987, 0x00001903 }, /* GL_RED */
+   { 27994, 0x00008016 }, /* GL_REDUCE */
+   { 28004, 0x00008016 }, /* GL_REDUCE_EXT */
+   { 28018, 0x00000D15 }, /* GL_RED_BIAS */
+   { 28030, 0x00000D52 }, /* GL_RED_BITS */
+   { 28042, 0x00000D14 }, /* GL_RED_SCALE */
+   { 28055, 0x00008512 }, /* GL_REFLECTION_MAP */
+   { 28073, 0x00008512 }, /* GL_REFLECTION_MAP_ARB */
+   { 28095, 0x00008512 }, /* GL_REFLECTION_MAP_NV */
+   { 28116, 0x00001C00 }, /* GL_RENDER */
+   { 28126, 0x00008D41 }, /* GL_RENDERBUFFER */
+   { 28142, 0x00008D53 }, /* GL_RENDERBUFFER_ALPHA_SIZE */
+   { 28169, 0x00008CA7 }, /* GL_RENDERBUFFER_BINDING_EXT */
+   { 28197, 0x00008D52 }, /* GL_RENDERBUFFER_BLUE_SIZE */
+   { 28223, 0x00008D54 }, /* GL_RENDERBUFFER_DEPTH_SIZE */
+   { 28250, 0x00008D41 }, /* GL_RENDERBUFFER_EXT */
+   { 28270, 0x00008D51 }, /* GL_RENDERBUFFER_GREEN_SIZE */
+   { 28297, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT */
+   { 28320, 0x00008D43 }, /* GL_RENDERBUFFER_HEIGHT_EXT */
+   { 28347, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT */
+   { 28379, 0x00008D44 }, /* GL_RENDERBUFFER_INTERNAL_FORMAT_EXT */
+   { 28415, 0x00008D50 }, /* GL_RENDERBUFFER_RED_SIZE */
+   { 28440, 0x00008CAB }, /* GL_RENDERBUFFER_SAMPLES */
+   { 28464, 0x00008D55 }, /* GL_RENDERBUFFER_STENCIL_SIZE */
+   { 28493, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH */
+   { 28515, 0x00008D42 }, /* GL_RENDERBUFFER_WIDTH_EXT */
+   { 28541, 0x00001F01 }, /* GL_RENDERER */
+   { 28553, 0x00000C40 }, /* GL_RENDER_MODE */
+   { 28568, 0x00002901 }, /* GL_REPEAT */
+   { 28578, 0x00001E01 }, /* GL_REPLACE */
+   { 28589, 0x00008062 }, /* GL_REPLACE_EXT */
+   { 28604, 0x00008153 }, /* GL_REPLICATE_BORDER_HP */
+   { 28627, 0x0000803A }, /* GL_RESCALE_NORMAL */
+   { 28645, 0x0000803A }, /* GL_RESCALE_NORMAL_EXT */
+   { 28667, 0x00000102 }, /* GL_RETURN */
+   { 28677, 0x00001907 }, /* GL_RGB */
+   { 28684, 0x00008052 }, /* GL_RGB10 */
+   { 28693, 0x00008059 }, /* GL_RGB10_A2 */
+   { 28705, 0x00008059 }, /* GL_RGB10_A2_EXT */
+   { 28721, 0x00008052 }, /* GL_RGB10_EXT */
+   { 28734, 0x00008053 }, /* GL_RGB12 */
+   { 28743, 0x00008053 }, /* GL_RGB12_EXT */
+   { 28756, 0x00008054 }, /* GL_RGB16 */
+   { 28765, 0x00008054 }, /* GL_RGB16_EXT */
+   { 28778, 0x0000804E }, /* GL_RGB2_EXT */
+   { 28790, 0x0000804F }, /* GL_RGB4 */
+   { 28798, 0x0000804F }, /* GL_RGB4_EXT */
+   { 28810, 0x000083A1 }, /* GL_RGB4_S3TC */
+   { 28823, 0x00008050 }, /* GL_RGB5 */
+   { 28831, 0x00008057 }, /* GL_RGB5_A1 */
+   { 28842, 0x00008057 }, /* GL_RGB5_A1_EXT */
+   { 28857, 0x00008050 }, /* GL_RGB5_EXT */
+   { 28869, 0x00008051 }, /* GL_RGB8 */
+   { 28877, 0x00008051 }, /* GL_RGB8_EXT */
+   { 28889, 0x00001908 }, /* GL_RGBA */
+   { 28897, 0x0000805A }, /* GL_RGBA12 */
+   { 28907, 0x0000805A }, /* GL_RGBA12_EXT */
+   { 28921, 0x0000805B }, /* GL_RGBA16 */
+   { 28931, 0x0000805B }, /* GL_RGBA16_EXT */
+   { 28945, 0x00008055 }, /* GL_RGBA2 */
+   { 28954, 0x00008055 }, /* GL_RGBA2_EXT */
+   { 28967, 0x00008056 }, /* GL_RGBA4 */
+   { 28976, 0x000083A5 }, /* GL_RGBA4_DXT5_S3TC */
+   { 28995, 0x00008056 }, /* GL_RGBA4_EXT */
+   { 29008, 0x000083A3 }, /* GL_RGBA4_S3TC */
+   { 29022, 0x00008058 }, /* GL_RGBA8 */
+   { 29031, 0x00008058 }, /* GL_RGBA8_EXT */
+   { 29044, 0x000083A4 }, /* GL_RGBA_DXT5_S3TC */
+   { 29062, 0x00000C31 }, /* GL_RGBA_MODE */
+   { 29075, 0x000083A2 }, /* GL_RGBA_S3TC */
+   { 29088, 0x000083A0 }, /* GL_RGB_S3TC */
+   { 29100, 0x00008573 }, /* GL_RGB_SCALE */
+   { 29113, 0x00008573 }, /* GL_RGB_SCALE_ARB */
+   { 29130, 0x00008573 }, /* GL_RGB_SCALE_EXT */
+   { 29147, 0x00000407 }, /* GL_RIGHT */
+   { 29156, 0x00002000 }, /* GL_S */
+   { 29161, 0x00008B5D }, /* GL_SAMPLER_1D */
+   { 29175, 0x00008B61 }, /* GL_SAMPLER_1D_SHADOW */
+   { 29196, 0x00008B5E }, /* GL_SAMPLER_2D */
+   { 29210, 0x00008B62 }, /* GL_SAMPLER_2D_SHADOW */
+   { 29231, 0x00008B5F }, /* GL_SAMPLER_3D */
+   { 29245, 0x00008B60 }, /* GL_SAMPLER_CUBE */
+   { 29261, 0x000080A9 }, /* GL_SAMPLES */
+   { 29272, 0x000086B4 }, /* GL_SAMPLES_3DFX */
+   { 29288, 0x000080A9 }, /* GL_SAMPLES_ARB */
+   { 29303, 0x00008914 }, /* GL_SAMPLES_PASSED */
+   { 29321, 0x00008914 }, /* GL_SAMPLES_PASSED_ARB */
+   { 29343, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
+   { 29371, 0x0000809E }, /* GL_SAMPLE_ALPHA_TO_COVERAGE_ARB */
+   { 29403, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE */
+   { 29426, 0x0000809F }, /* GL_SAMPLE_ALPHA_TO_ONE_ARB */
+   { 29453, 0x000080A8 }, /* GL_SAMPLE_BUFFERS */
+   { 29471, 0x000086B3 }, /* GL_SAMPLE_BUFFERS_3DFX */
+   { 29494, 0x000080A8 }, /* GL_SAMPLE_BUFFERS_ARB */
+   { 29516, 0x000080A0 }, /* GL_SAMPLE_COVERAGE */
+   { 29535, 0x000080A0 }, /* GL_SAMPLE_COVERAGE_ARB */
+   { 29558, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT */
+   { 29584, 0x000080AB }, /* GL_SAMPLE_COVERAGE_INVERT_ARB */
+   { 29614, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE */
+   { 29639, 0x000080AA }, /* GL_SAMPLE_COVERAGE_VALUE_ARB */
+   { 29668, 0x00080000 }, /* GL_SCISSOR_BIT */
+   { 29683, 0x00000C10 }, /* GL_SCISSOR_BOX */
+   { 29698, 0x00000C11 }, /* GL_SCISSOR_TEST */
+   { 29714, 0x0000845E }, /* GL_SECONDARY_COLOR_ARRAY */
+   { 29739, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
+   { 29779, 0x0000889C }, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB */
+   { 29823, 0x0000845D }, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
+   { 29856, 0x0000845A }, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
+   { 29886, 0x0000845C }, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
+   { 29918, 0x0000845B }, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
+   { 29948, 0x00001C02 }, /* GL_SELECT */
+   { 29958, 0x00000DF3 }, /* GL_SELECTION_BUFFER_POINTER */
+   { 29986, 0x00000DF4 }, /* GL_SELECTION_BUFFER_SIZE */
+   { 30011, 0x00008012 }, /* GL_SEPARABLE_2D */
+   { 30027, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR */
+   { 30054, 0x000081FA }, /* GL_SEPARATE_SPECULAR_COLOR_EXT */
+   { 30085, 0x0000150F }, /* GL_SET */
+   { 30092, 0x00008B48 }, /* GL_SHADER_OBJECT_ARB */
+   { 30113, 0x00008B88 }, /* GL_SHADER_SOURCE_LENGTH */
+   { 30137, 0x00008B4F }, /* GL_SHADER_TYPE */
+   { 30152, 0x00000B54 }, /* GL_SHADE_MODEL */
+   { 30167, 0x00008B8C }, /* GL_SHADING_LANGUAGE_VERSION */
+   { 30195, 0x000080BF }, /* GL_SHADOW_AMBIENT_SGIX */
+   { 30218, 0x000081FB }, /* GL_SHARED_TEXTURE_PALETTE_EXT */
+   { 30248, 0x00001601 }, /* GL_SHININESS */
+   { 30261, 0x00001402 }, /* GL_SHORT */
+   { 30270, 0x000081F9 }, /* GL_SINGLE_COLOR */
+   { 30286, 0x000081F9 }, /* GL_SINGLE_COLOR_EXT */
+   { 30306, 0x000085CC }, /* GL_SLICE_ACCUM_SUN */
+   { 30325, 0x00001D01 }, /* GL_SMOOTH */
+   { 30335, 0x00000B23 }, /* GL_SMOOTH_LINE_WIDTH_GRANULARITY */
+   { 30368, 0x00000B22 }, /* GL_SMOOTH_LINE_WIDTH_RANGE */
+   { 30395, 0x00000B13 }, /* GL_SMOOTH_POINT_SIZE_GRANULARITY */
+   { 30428, 0x00000B12 }, /* GL_SMOOTH_POINT_SIZE_RANGE */
+   { 30455, 0x00008588 }, /* GL_SOURCE0_ALPHA */
+   { 30472, 0x00008588 }, /* GL_SOURCE0_ALPHA_ARB */
+   { 30493, 0x00008588 }, /* GL_SOURCE0_ALPHA_EXT */
+   { 30514, 0x00008580 }, /* GL_SOURCE0_RGB */
+   { 30529, 0x00008580 }, /* GL_SOURCE0_RGB_ARB */
+   { 30548, 0x00008580 }, /* GL_SOURCE0_RGB_EXT */
+   { 30567, 0x00008589 }, /* GL_SOURCE1_ALPHA */
+   { 30584, 0x00008589 }, /* GL_SOURCE1_ALPHA_ARB */
+   { 30605, 0x00008589 }, /* GL_SOURCE1_ALPHA_EXT */
+   { 30626, 0x00008581 }, /* GL_SOURCE1_RGB */
+   { 30641, 0x00008581 }, /* GL_SOURCE1_RGB_ARB */
+   { 30660, 0x00008581 }, /* GL_SOURCE1_RGB_EXT */
+   { 30679, 0x0000858A }, /* GL_SOURCE2_ALPHA */
+   { 30696, 0x0000858A }, /* GL_SOURCE2_ALPHA_ARB */
+   { 30717, 0x0000858A }, /* GL_SOURCE2_ALPHA_EXT */
+   { 30738, 0x00008582 }, /* GL_SOURCE2_RGB */
+   { 30753, 0x00008582 }, /* GL_SOURCE2_RGB_ARB */
+   { 30772, 0x00008582 }, /* GL_SOURCE2_RGB_EXT */
+   { 30791, 0x0000858B }, /* GL_SOURCE3_ALPHA_NV */
+   { 30811, 0x00008583 }, /* GL_SOURCE3_RGB_NV */
+   { 30829, 0x00001202 }, /* GL_SPECULAR */
+   { 30841, 0x00002402 }, /* GL_SPHERE_MAP */
+   { 30855, 0x00001206 }, /* GL_SPOT_CUTOFF */
+   { 30870, 0x00001204 }, /* GL_SPOT_DIRECTION */
+   { 30888, 0x00001205 }, /* GL_SPOT_EXPONENT */
+   { 30905, 0x00008588 }, /* GL_SRC0_ALPHA */
+   { 30919, 0x00008580 }, /* GL_SRC0_RGB */
+   { 30931, 0x00008589 }, /* GL_SRC1_ALPHA */
+   { 30945, 0x00008581 }, /* GL_SRC1_RGB */
+   { 30957, 0x0000858A }, /* GL_SRC2_ALPHA */
+   { 30971, 0x00008582 }, /* GL_SRC2_RGB */
+   { 30983, 0x00000302 }, /* GL_SRC_ALPHA */
+   { 30996, 0x00000308 }, /* GL_SRC_ALPHA_SATURATE */
+   { 31018, 0x00000300 }, /* GL_SRC_COLOR */
+   { 31031, 0x00008C40 }, /* GL_SRGB */
+   { 31039, 0x00000503 }, /* GL_STACK_OVERFLOW */
+   { 31057, 0x00000504 }, /* GL_STACK_UNDERFLOW */
+   { 31076, 0x000088E6 }, /* GL_STATIC_COPY */
+   { 31091, 0x000088E6 }, /* GL_STATIC_COPY_ARB */
+   { 31110, 0x000088E4 }, /* GL_STATIC_DRAW */
+   { 31125, 0x000088E4 }, /* GL_STATIC_DRAW_ARB */
+   { 31144, 0x000088E5 }, /* GL_STATIC_READ */
+   { 31159, 0x000088E5 }, /* GL_STATIC_READ_ARB */
+   { 31178, 0x00001802 }, /* GL_STENCIL */
+   { 31189, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT */
+   { 31211, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT_EXT */
+   { 31237, 0x00008801 }, /* GL_STENCIL_BACK_FAIL */
+   { 31258, 0x00008801 }, /* GL_STENCIL_BACK_FAIL_ATI */
+   { 31283, 0x00008800 }, /* GL_STENCIL_BACK_FUNC */
+   { 31304, 0x00008800 }, /* GL_STENCIL_BACK_FUNC_ATI */
+   { 31329, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
+   { 31361, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL_ATI */
+   { 31397, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
+   { 31429, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS_ATI */
+   { 31465, 0x00008CA3 }, /* GL_STENCIL_BACK_REF */
+   { 31485, 0x00008CA4 }, /* GL_STENCIL_BACK_VALUE_MASK */
+   { 31512, 0x00008CA5 }, /* GL_STENCIL_BACK_WRITEMASK */
+   { 31538, 0x00000D57 }, /* GL_STENCIL_BITS */
+   { 31554, 0x00000400 }, /* GL_STENCIL_BUFFER_BIT */
+   { 31576, 0x00000B91 }, /* GL_STENCIL_CLEAR_VALUE */
+   { 31599, 0x00000B94 }, /* GL_STENCIL_FAIL */
+   { 31615, 0x00000B92 }, /* GL_STENCIL_FUNC */
+   { 31631, 0x00001901 }, /* GL_STENCIL_INDEX */
+   { 31648, 0x00008D49 }, /* GL_STENCIL_INDEX16_EXT */
+   { 31671, 0x00008D46 }, /* GL_STENCIL_INDEX1_EXT */
+   { 31693, 0x00008D47 }, /* GL_STENCIL_INDEX4_EXT */
+   { 31715, 0x00008D48 }, /* GL_STENCIL_INDEX8_EXT */
+   { 31737, 0x00008D45 }, /* GL_STENCIL_INDEX_EXT */
+   { 31758, 0x00000B95 }, /* GL_STENCIL_PASS_DEPTH_FAIL */
+   { 31785, 0x00000B96 }, /* GL_STENCIL_PASS_DEPTH_PASS */
+   { 31812, 0x00000B97 }, /* GL_STENCIL_REF */
+   { 31827, 0x00000B90 }, /* GL_STENCIL_TEST */
+   { 31843, 0x00008910 }, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
+   { 31872, 0x00000B93 }, /* GL_STENCIL_VALUE_MASK */
+   { 31894, 0x00000B98 }, /* GL_STENCIL_WRITEMASK */
+   { 31915, 0x00000C33 }, /* GL_STEREO */
+   { 31925, 0x000088E2 }, /* GL_STREAM_COPY */
+   { 31940, 0x000088E2 }, /* GL_STREAM_COPY_ARB */
+   { 31959, 0x000088E0 }, /* GL_STREAM_DRAW */
+   { 31974, 0x000088E0 }, /* GL_STREAM_DRAW_ARB */
+   { 31993, 0x000088E1 }, /* GL_STREAM_READ */
+   { 32008, 0x000088E1 }, /* GL_STREAM_READ_ARB */
+   { 32027, 0x00000D50 }, /* GL_SUBPIXEL_BITS */
+   { 32044, 0x000084E7 }, /* GL_SUBTRACT */
+   { 32056, 0x000084E7 }, /* GL_SUBTRACT_ARB */
+   { 32072, 0x00002001 }, /* GL_T */
+   { 32077, 0x00002A2A }, /* GL_T2F_C3F_V3F */
+   { 32092, 0x00002A2C }, /* GL_T2F_C4F_N3F_V3F */
+   { 32111, 0x00002A29 }, /* GL_T2F_C4UB_V3F */
+   { 32127, 0x00002A2B }, /* GL_T2F_N3F_V3F */
+   { 32142, 0x00002A27 }, /* GL_T2F_V3F */
+   { 32153, 0x00002A2D }, /* GL_T4F_C4F_N3F_V4F */
+   { 32172, 0x00002A28 }, /* GL_T4F_V4F */
+   { 32183, 0x00008031 }, /* GL_TABLE_TOO_LARGE_EXT */
+   { 32206, 0x00001702 }, /* GL_TEXTURE */
+   { 32217, 0x000084C0 }, /* GL_TEXTURE0 */
+   { 32229, 0x000084C0 }, /* GL_TEXTURE0_ARB */
+   { 32245, 0x000084C1 }, /* GL_TEXTURE1 */
+   { 32257, 0x000084CA }, /* GL_TEXTURE10 */
+   { 32270, 0x000084CA }, /* GL_TEXTURE10_ARB */
+   { 32287, 0x000084CB }, /* GL_TEXTURE11 */
+   { 32300, 0x000084CB }, /* GL_TEXTURE11_ARB */
+   { 32317, 0x000084CC }, /* GL_TEXTURE12 */
+   { 32330, 0x000084CC }, /* GL_TEXTURE12_ARB */
+   { 32347, 0x000084CD }, /* GL_TEXTURE13 */
+   { 32360, 0x000084CD }, /* GL_TEXTURE13_ARB */
+   { 32377, 0x000084CE }, /* GL_TEXTURE14 */
+   { 32390, 0x000084CE }, /* GL_TEXTURE14_ARB */
+   { 32407, 0x000084CF }, /* GL_TEXTURE15 */
+   { 32420, 0x000084CF }, /* GL_TEXTURE15_ARB */
+   { 32437, 0x000084D0 }, /* GL_TEXTURE16 */
+   { 32450, 0x000084D0 }, /* GL_TEXTURE16_ARB */
+   { 32467, 0x000084D1 }, /* GL_TEXTURE17 */
+   { 32480, 0x000084D1 }, /* GL_TEXTURE17_ARB */
+   { 32497, 0x000084D2 }, /* GL_TEXTURE18 */
+   { 32510, 0x000084D2 }, /* GL_TEXTURE18_ARB */
+   { 32527, 0x000084D3 }, /* GL_TEXTURE19 */
+   { 32540, 0x000084D3 }, /* GL_TEXTURE19_ARB */
+   { 32557, 0x000084C1 }, /* GL_TEXTURE1_ARB */
+   { 32573, 0x000084C2 }, /* GL_TEXTURE2 */
+   { 32585, 0x000084D4 }, /* GL_TEXTURE20 */
+   { 32598, 0x000084D4 }, /* GL_TEXTURE20_ARB */
+   { 32615, 0x000084D5 }, /* GL_TEXTURE21 */
+   { 32628, 0x000084D5 }, /* GL_TEXTURE21_ARB */
+   { 32645, 0x000084D6 }, /* GL_TEXTURE22 */
+   { 32658, 0x000084D6 }, /* GL_TEXTURE22_ARB */
+   { 32675, 0x000084D7 }, /* GL_TEXTURE23 */
+   { 32688, 0x000084D7 }, /* GL_TEXTURE23_ARB */
+   { 32705, 0x000084D8 }, /* GL_TEXTURE24 */
+   { 32718, 0x000084D8 }, /* GL_TEXTURE24_ARB */
+   { 32735, 0x000084D9 }, /* GL_TEXTURE25 */
+   { 32748, 0x000084D9 }, /* GL_TEXTURE25_ARB */
+   { 32765, 0x000084DA }, /* GL_TEXTURE26 */
+   { 32778, 0x000084DA }, /* GL_TEXTURE26_ARB */
+   { 32795, 0x000084DB }, /* GL_TEXTURE27 */
+   { 32808, 0x000084DB }, /* GL_TEXTURE27_ARB */
+   { 32825, 0x000084DC }, /* GL_TEXTURE28 */
+   { 32838, 0x000084DC }, /* GL_TEXTURE28_ARB */
+   { 32855, 0x000084DD }, /* GL_TEXTURE29 */
+   { 32868, 0x000084DD }, /* GL_TEXTURE29_ARB */
+   { 32885, 0x000084C2 }, /* GL_TEXTURE2_ARB */
+   { 32901, 0x000084C3 }, /* GL_TEXTURE3 */
+   { 32913, 0x000084DE }, /* GL_TEXTURE30 */
+   { 32926, 0x000084DE }, /* GL_TEXTURE30_ARB */
+   { 32943, 0x000084DF }, /* GL_TEXTURE31 */
+   { 32956, 0x000084DF }, /* GL_TEXTURE31_ARB */
+   { 32973, 0x000084C3 }, /* GL_TEXTURE3_ARB */
+   { 32989, 0x000084C4 }, /* GL_TEXTURE4 */
+   { 33001, 0x000084C4 }, /* GL_TEXTURE4_ARB */
+   { 33017, 0x000084C5 }, /* GL_TEXTURE5 */
+   { 33029, 0x000084C5 }, /* GL_TEXTURE5_ARB */
+   { 33045, 0x000084C6 }, /* GL_TEXTURE6 */
+   { 33057, 0x000084C6 }, /* GL_TEXTURE6_ARB */
+   { 33073, 0x000084C7 }, /* GL_TEXTURE7 */
+   { 33085, 0x000084C7 }, /* GL_TEXTURE7_ARB */
+   { 33101, 0x000084C8 }, /* GL_TEXTURE8 */
+   { 33113, 0x000084C8 }, /* GL_TEXTURE8_ARB */
+   { 33129, 0x000084C9 }, /* GL_TEXTURE9 */
+   { 33141, 0x000084C9 }, /* GL_TEXTURE9_ARB */
+   { 33157, 0x00000DE0 }, /* GL_TEXTURE_1D */
+   { 33171, 0x00008C18 }, /* GL_TEXTURE_1D_ARRAY_EXT */
+   { 33195, 0x00000DE1 }, /* GL_TEXTURE_2D */
+   { 33209, 0x00008C1A }, /* GL_TEXTURE_2D_ARRAY_EXT */
+   { 33233, 0x0000806F }, /* GL_TEXTURE_3D */
+   { 33247, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE */
+   { 33269, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE_EXT */
+   { 33295, 0x0000813C }, /* GL_TEXTURE_BASE_LEVEL */
+   { 33317, 0x00008068 }, /* GL_TEXTURE_BINDING_1D */
+   { 33339, 0x00008C1C }, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
+   { 33371, 0x00008069 }, /* GL_TEXTURE_BINDING_2D */
+   { 33393, 0x00008C1D }, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
+   { 33425, 0x0000806A }, /* GL_TEXTURE_BINDING_3D */
+   { 33447, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP */
+   { 33475, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP_ARB */
+   { 33507, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
+   { 33540, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_NV */
+   { 33572, 0x00040000 }, /* GL_TEXTURE_BIT */
+   { 33587, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE */
+   { 33608, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE_EXT */
+   { 33633, 0x00001005 }, /* GL_TEXTURE_BORDER */
+   { 33651, 0x00001004 }, /* GL_TEXTURE_BORDER_COLOR */
+   { 33675, 0x00008171 }, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
+   { 33706, 0x00008176 }, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
+   { 33736, 0x00008172 }, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
+   { 33766, 0x00008175 }, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
+   { 33801, 0x00008173 }, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
+   { 33832, 0x00008174 }, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+   { 33870, 0x000080BC }, /* GL_TEXTURE_COLOR_TABLE_SGI */
+   { 33897, 0x000081EF }, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
+   { 33929, 0x000080BF }, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
+   { 33963, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC */
+   { 33987, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC_ARB */
+   { 34015, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE */
+   { 34039, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE_ARB */
+   { 34067, 0x0000819B }, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
+   { 34100, 0x0000819A }, /* GL_TEXTURE_COMPARE_SGIX */
+   { 34124, 0x00001003 }, /* GL_TEXTURE_COMPONENTS */
+   { 34146, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED */
+   { 34168, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED_ARB */
+   { 34194, 0x000086A3 }, /* GL_TEXTURE_COMPRESSED_FORMATS_ARB */
+   { 34228, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
+   { 34261, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE_ARB */
+   { 34298, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT */
+   { 34326, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT_ARB */
+   { 34358, 0x00008078 }, /* GL_TEXTURE_COORD_ARRAY */
+   { 34381, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
+   { 34419, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB */
+   { 34461, 0x00008092 }, /* GL_TEXTURE_COORD_ARRAY_POINTER */
+   { 34492, 0x00008088 }, /* GL_TEXTURE_COORD_ARRAY_SIZE */
+   { 34520, 0x0000808A }, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
+   { 34550, 0x00008089 }, /* GL_TEXTURE_COORD_ARRAY_TYPE */
+   { 34578, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP */
+   { 34598, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP_ARB */
+   { 34622, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
+   { 34653, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB */
+   { 34688, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
+   { 34719, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB */
+   { 34754, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
+   { 34785, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB */
+   { 34820, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
+   { 34851, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB */
+   { 34886, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
+   { 34917, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB */
+   { 34952, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
+   { 34983, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB */
+   { 35018, 0x00008071 }, /* GL_TEXTURE_DEPTH */
+   { 35035, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE */
+   { 35057, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE_ARB */
+   { 35083, 0x00002300 }, /* GL_TEXTURE_ENV */
+   { 35098, 0x00002201 }, /* GL_TEXTURE_ENV_COLOR */
+   { 35119, 0x00002200 }, /* GL_TEXTURE_ENV_MODE */
+   { 35139, 0x00008500 }, /* GL_TEXTURE_FILTER_CONTROL */
+   { 35165, 0x00002500 }, /* GL_TEXTURE_GEN_MODE */
+   { 35185, 0x00000C63 }, /* GL_TEXTURE_GEN_Q */
+   { 35202, 0x00000C62 }, /* GL_TEXTURE_GEN_R */
+   { 35219, 0x00000C60 }, /* GL_TEXTURE_GEN_S */
+   { 35236, 0x00000C61 }, /* GL_TEXTURE_GEN_T */
+   { 35253, 0x0000819D }, /* GL_TEXTURE_GEQUAL_R_SGIX */
+   { 35278, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE */
+   { 35300, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE_EXT */
+   { 35326, 0x00001001 }, /* GL_TEXTURE_HEIGHT */
+   { 35344, 0x000080ED }, /* GL_TEXTURE_INDEX_SIZE_EXT */
+   { 35370, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE */
+   { 35396, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE_EXT */
+   { 35426, 0x00001003 }, /* GL_TEXTURE_INTERNAL_FORMAT */
+   { 35453, 0x0000819C }, /* GL_TEXTURE_LEQUAL_R_SGIX */
+   { 35478, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS */
+   { 35498, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS_EXT */
+   { 35522, 0x00008190 }, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
+   { 35549, 0x0000818E }, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
+   { 35576, 0x0000818F }, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
+   { 35603, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE */
+   { 35629, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE_EXT */
+   { 35659, 0x00002800 }, /* GL_TEXTURE_MAG_FILTER */
+   { 35681, 0x00000BA8 }, /* GL_TEXTURE_MATRIX */
+   { 35699, 0x000084FE }, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
+   { 35729, 0x0000836B }, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
+   { 35757, 0x00008369 }, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
+   { 35785, 0x0000836A }, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
+   { 35813, 0x0000813D }, /* GL_TEXTURE_MAX_LEVEL */
+   { 35834, 0x0000813B }, /* GL_TEXTURE_MAX_LOD */
+   { 35853, 0x00002801 }, /* GL_TEXTURE_MIN_FILTER */
+   { 35875, 0x0000813A }, /* GL_TEXTURE_MIN_LOD */
+   { 35894, 0x00008066 }, /* GL_TEXTURE_PRIORITY */
+   { 35914, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_ARB */
+   { 35939, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_NV */
+   { 35963, 0x0000805C }, /* GL_TEXTURE_RED_SIZE */
+   { 35983, 0x0000805C }, /* GL_TEXTURE_RED_SIZE_EXT */
+   { 36007, 0x00008067 }, /* GL_TEXTURE_RESIDENT */
+   { 36027, 0x00000BA5 }, /* GL_TEXTURE_STACK_DEPTH */
+   { 36050, 0x000088F1 }, /* GL_TEXTURE_STENCIL_SIZE */
+   { 36074, 0x00008065 }, /* GL_TEXTURE_TOO_LARGE_EXT */
+   { 36099, 0x0000888F }, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
+   { 36133, 0x00001000 }, /* GL_TEXTURE_WIDTH */
+   { 36150, 0x00008072 }, /* GL_TEXTURE_WRAP_R */
+   { 36168, 0x00002802 }, /* GL_TEXTURE_WRAP_S */
+   { 36186, 0x00002803 }, /* GL_TEXTURE_WRAP_T */
+   { 36204, 0x000088BF }, /* GL_TIME_ELAPSED_EXT */
+   { 36224, 0x00008648 }, /* GL_TRACK_MATRIX_NV */
+   { 36243, 0x00008649 }, /* GL_TRACK_MATRIX_TRANSFORM_NV */
+   { 36272, 0x00001000 }, /* GL_TRANSFORM_BIT */
+   { 36289, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX */
+   { 36315, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX_ARB */
+   { 36345, 0x000088B7 }, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
+   { 36377, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
+   { 36407, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX_ARB */
+   { 36441, 0x0000862C }, /* GL_TRANSPOSE_NV */
+   { 36457, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX */
+   { 36488, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX_ARB */
+   { 36523, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX */
+   { 36551, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX_ARB */
+   { 36583, 0x00000004 }, /* GL_TRIANGLES */
+   { 36596, 0x00000006 }, /* GL_TRIANGLE_FAN */
+   { 36612, 0x00008615 }, /* GL_TRIANGLE_MESH_SUN */
+   { 36633, 0x00000005 }, /* GL_TRIANGLE_STRIP */
+   { 36651, 0x00000001 }, /* GL_TRUE */
+   { 36659, 0x00000CF5 }, /* GL_UNPACK_ALIGNMENT */
+   { 36679, 0x0000806E }, /* GL_UNPACK_IMAGE_HEIGHT */
+   { 36702, 0x00000CF1 }, /* GL_UNPACK_LSB_FIRST */
+   { 36722, 0x00000CF2 }, /* GL_UNPACK_ROW_LENGTH */
+   { 36743, 0x0000806D }, /* GL_UNPACK_SKIP_IMAGES */
+   { 36765, 0x00000CF4 }, /* GL_UNPACK_SKIP_PIXELS */
+   { 36787, 0x00000CF3 }, /* GL_UNPACK_SKIP_ROWS */
+   { 36807, 0x00000CF0 }, /* GL_UNPACK_SWAP_BYTES */
+   { 36828, 0x00001401 }, /* GL_UNSIGNED_BYTE */
+   { 36845, 0x00008362 }, /* GL_UNSIGNED_BYTE_2_3_3_REV */
+   { 36872, 0x00008032 }, /* GL_UNSIGNED_BYTE_3_3_2 */
+   { 36895, 0x00001405 }, /* GL_UNSIGNED_INT */
+   { 36911, 0x00008036 }, /* GL_UNSIGNED_INT_10_10_10_2 */
+   { 36938, 0x000084FA }, /* GL_UNSIGNED_INT_24_8 */
+   { 36959, 0x000084FA }, /* GL_UNSIGNED_INT_24_8_NV */
+   { 36983, 0x00008368 }, /* GL_UNSIGNED_INT_2_10_10_10_REV */
+   { 37014, 0x00008035 }, /* GL_UNSIGNED_INT_8_8_8_8 */
+   { 37038, 0x00008367 }, /* GL_UNSIGNED_INT_8_8_8_8_REV */
+   { 37066, 0x00008C17 }, /* GL_UNSIGNED_NORMALIZED */
+   { 37089, 0x00001403 }, /* GL_UNSIGNED_SHORT */
+   { 37107, 0x00008366 }, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
+   { 37137, 0x00008033 }, /* GL_UNSIGNED_SHORT_4_4_4_4 */
+   { 37163, 0x00008365 }, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
+   { 37193, 0x00008034 }, /* GL_UNSIGNED_SHORT_5_5_5_1 */
+   { 37219, 0x00008363 }, /* GL_UNSIGNED_SHORT_5_6_5 */
+   { 37243, 0x00008364 }, /* GL_UNSIGNED_SHORT_5_6_5_REV */
+   { 37271, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_APPLE */
+   { 37299, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_MESA */
+   { 37326, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
+   { 37358, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_MESA */
+   { 37389, 0x00008CA2 }, /* GL_UPPER_LEFT */
+   { 37403, 0x00002A20 }, /* GL_V2F */
+   { 37410, 0x00002A21 }, /* GL_V3F */
+   { 37417, 0x00008B83 }, /* GL_VALIDATE_STATUS */
+   { 37436, 0x00001F00 }, /* GL_VENDOR */
+   { 37446, 0x00001F02 }, /* GL_VERSION */
+   { 37457, 0x00008074 }, /* GL_VERTEX_ARRAY */
+   { 37473, 0x000085B5 }, /* GL_VERTEX_ARRAY_BINDING_APPLE */
+   { 37503, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
+   { 37534, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING_ARB */
+   { 37569, 0x0000808E }, /* GL_VERTEX_ARRAY_POINTER */
+   { 37593, 0x0000807A }, /* GL_VERTEX_ARRAY_SIZE */
+   { 37614, 0x0000807C }, /* GL_VERTEX_ARRAY_STRIDE */
+   { 37637, 0x0000807B }, /* GL_VERTEX_ARRAY_TYPE */
+   { 37658, 0x00008650 }, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
+   { 37685, 0x0000865A }, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
+   { 37713, 0x0000865B }, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
+   { 37741, 0x0000865C }, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
+   { 37769, 0x0000865D }, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
+   { 37797, 0x0000865E }, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
+   { 37825, 0x0000865F }, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
+   { 37853, 0x00008651 }, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
+   { 37880, 0x00008652 }, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
+   { 37907, 0x00008653 }, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
+   { 37934, 0x00008654 }, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
+   { 37961, 0x00008655 }, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
+   { 37988, 0x00008656 }, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
+   { 38015, 0x00008657 }, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
+   { 38042, 0x00008658 }, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
+   { 38069, 0x00008659 }, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
+   { 38096, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
+   { 38134, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB */
+   { 38176, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
+   { 38207, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB */
+   { 38242, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
+   { 38276, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB */
+   { 38314, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
+   { 38345, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER_ARB */
+   { 38380, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
+   { 38408, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB */
+   { 38440, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
+   { 38470, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB */
+   { 38504, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
+   { 38532, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB */
+   { 38564, 0x000086A7 }, /* GL_VERTEX_BLEND_ARB */
+   { 38584, 0x00008620 }, /* GL_VERTEX_PROGRAM_ARB */
+   { 38606, 0x0000864A }, /* GL_VERTEX_PROGRAM_BINDING_NV */
+   { 38635, 0x00008620 }, /* GL_VERTEX_PROGRAM_NV */
+   { 38656, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE */
+   { 38685, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_ARB */
+   { 38718, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_NV */
+   { 38750, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE */
+   { 38777, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_ARB */
+   { 38808, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_NV */
+   { 38838, 0x00008B31 }, /* GL_VERTEX_SHADER */
+   { 38855, 0x00008B31 }, /* GL_VERTEX_SHADER_ARB */
+   { 38876, 0x00008621 }, /* GL_VERTEX_STATE_PROGRAM_NV */
+   { 38903, 0x00000BA2 }, /* GL_VIEWPORT */
+   { 38915, 0x00000800 }, /* GL_VIEWPORT_BIT */
+   { 38931, 0x000086AD }, /* GL_WEIGHT_ARRAY_ARB */
+   { 38951, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
+   { 38982, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING_ARB */
+   { 39017, 0x000086AC }, /* GL_WEIGHT_ARRAY_POINTER_ARB */
+   { 39045, 0x000086AB }, /* GL_WEIGHT_ARRAY_SIZE_ARB */
+   { 39070, 0x000086AA }, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
+   { 39097, 0x000086A9 }, /* GL_WEIGHT_ARRAY_TYPE_ARB */
+   { 39122, 0x000086A6 }, /* GL_WEIGHT_SUM_UNITY_ARB */
+   { 39146, 0x000081D4 }, /* GL_WRAP_BORDER_SUN */
+   { 39165, 0x000088B9 }, /* GL_WRITE_ONLY */
+   { 39179, 0x000088B9 }, /* GL_WRITE_ONLY_ARB */
+   { 39197, 0x00001506 }, /* GL_XOR */
+   { 39204, 0x000085B9 }, /* GL_YCBCR_422_APPLE */
+   { 39223, 0x00008757 }, /* GL_YCBCR_MESA */
+   { 39237, 0x00000000 }, /* GL_ZERO */
+   { 39245, 0x00000D16 }, /* GL_ZOOM_X */
+   { 39255, 0x00000D17 }, /* GL_ZOOM_Y */
 };
 
-static const unsigned reduced_enums[1284] =
+static const unsigned reduced_enums[1308] =
 {
-       434, /* GL_FALSE */
-       643, /* GL_LINES */
-       645, /* GL_LINE_LOOP */
-       652, /* GL_LINE_STRIP */
-      1639, /* GL_TRIANGLES */
-      1642, /* GL_TRIANGLE_STRIP */
-      1640, /* GL_TRIANGLE_FAN */
-      1209, /* GL_QUADS */
-      1211, /* GL_QUAD_STRIP */
-      1097, /* GL_POLYGON */
-      1109, /* GL_POLYGON_STIPPLE_BIT */
-      1062, /* GL_PIXEL_MODE_BIT */
-       630, /* GL_LIGHTING_BIT */
-       456, /* GL_FOG_BIT */
+       455, /* GL_FALSE */
+       687, /* GL_LINES */
+       689, /* GL_LINE_LOOP */
+       696, /* GL_LINE_STRIP */
+      1699, /* GL_TRIANGLES */
+      1702, /* GL_TRIANGLE_STRIP */
+      1700, /* GL_TRIANGLE_FAN */
+      1254, /* GL_QUADS */
+      1256, /* GL_QUAD_STRIP */
+      1142, /* GL_POLYGON */
+      1154, /* GL_POLYGON_STIPPLE_BIT */
+      1107, /* GL_PIXEL_MODE_BIT */
+       674, /* GL_LIGHTING_BIT */
+       477, /* GL_FOG_BIT */
          8, /* GL_ACCUM */
-       662, /* GL_LOAD */
-      1251, /* GL_RETURN */
-       935, /* GL_MULT */
+       706, /* GL_LOAD */
+      1308, /* GL_RETURN */
+       980, /* GL_MULT */
         23, /* GL_ADD */
-       951, /* GL_NEVER */
-       620, /* GL_LESS */
-       424, /* GL_EQUAL */
-       619, /* GL_LEQUAL */
-       545, /* GL_GREATER */
-       966, /* GL_NOTEQUAL */
-       520, /* GL_GEQUAL */
+       996, /* GL_NEVER */
+       664, /* GL_LESS */
+       445, /* GL_EQUAL */
+       663, /* GL_LEQUAL */
+       587, /* GL_GREATER */
+      1011, /* GL_NOTEQUAL */
+       562, /* GL_GEQUAL */
         46, /* GL_ALWAYS */
-      1384, /* GL_SRC_COLOR */
-       995, /* GL_ONE_MINUS_SRC_COLOR */
-      1382, /* GL_SRC_ALPHA */
-       994, /* GL_ONE_MINUS_SRC_ALPHA */
-       405, /* GL_DST_ALPHA */
-       992, /* GL_ONE_MINUS_DST_ALPHA */
-       406, /* GL_DST_COLOR */
-       993, /* GL_ONE_MINUS_DST_COLOR */
-      1383, /* GL_SRC_ALPHA_SATURATE */
-       508, /* GL_FRONT_LEFT */
-       509, /* GL_FRONT_RIGHT */
+      1441, /* GL_SRC_COLOR */
+      1040, /* GL_ONE_MINUS_SRC_COLOR */
+      1439, /* GL_SRC_ALPHA */
+      1039, /* GL_ONE_MINUS_SRC_ALPHA */
+       426, /* GL_DST_ALPHA */
+      1037, /* GL_ONE_MINUS_DST_ALPHA */
+       427, /* GL_DST_COLOR */
+      1038, /* GL_ONE_MINUS_DST_COLOR */
+      1440, /* GL_SRC_ALPHA_SATURATE */
+       550, /* GL_FRONT_LEFT */
+       551, /* GL_FRONT_RIGHT */
         68, /* GL_BACK_LEFT */
         69, /* GL_BACK_RIGHT */
-       505, /* GL_FRONT */
+       547, /* GL_FRONT */
         67, /* GL_BACK */
-       618, /* GL_LEFT */
-      1291, /* GL_RIGHT */
-       506, /* GL_FRONT_AND_BACK */
+       662, /* GL_LEFT */
+      1348, /* GL_RIGHT */
+       548, /* GL_FRONT_AND_BACK */
         62, /* GL_AUX0 */
         63, /* GL_AUX1 */
         64, /* GL_AUX2 */
         65, /* GL_AUX3 */
-       610, /* GL_INVALID_ENUM */
-       613, /* GL_INVALID_VALUE */
-       612, /* GL_INVALID_OPERATION */
-      1385, /* GL_STACK_OVERFLOW */
-      1386, /* GL_STACK_UNDERFLOW */
-      1020, /* GL_OUT_OF_MEMORY */
-       611, /* GL_INVALID_FRAMEBUFFER_OPERATION_EXT */
+       653, /* GL_INVALID_ENUM */
+       657, /* GL_INVALID_VALUE */
+       656, /* GL_INVALID_OPERATION */
+      1443, /* GL_STACK_OVERFLOW */
+      1444, /* GL_STACK_UNDERFLOW */
+      1065, /* GL_OUT_OF_MEMORY */
+       654, /* GL_INVALID_FRAMEBUFFER_OPERATION */
          0, /* GL_2D */
          2, /* GL_3D */
          3, /* GL_3D_COLOR */
          4, /* GL_3D_COLOR_TEXTURE */
          6, /* GL_4D_COLOR_TEXTURE */
-      1040, /* GL_PASS_THROUGH_TOKEN */
-      1096, /* GL_POINT_TOKEN */
-       653, /* GL_LINE_TOKEN */
-      1110, /* GL_POLYGON_TOKEN */
+      1085, /* GL_PASS_THROUGH_TOKEN */
+      1141, /* GL_POINT_TOKEN */
+       697, /* GL_LINE_TOKEN */
+      1155, /* GL_POLYGON_TOKEN */
         73, /* GL_BITMAP_TOKEN */
-       404, /* GL_DRAW_PIXEL_TOKEN */
-       270, /* GL_COPY_PIXEL_TOKEN */
-       646, /* GL_LINE_RESET_TOKEN */
-       427, /* GL_EXP */
-       428, /* GL_EXP2 */
-       303, /* GL_CW */
+       425, /* GL_DRAW_PIXEL_TOKEN */
+       286, /* GL_COPY_PIXEL_TOKEN */
+       690, /* GL_LINE_RESET_TOKEN */
+       448, /* GL_EXP */
+       449, /* GL_EXP2 */
+       319, /* GL_CW */
        115, /* GL_CCW */
        136, /* GL_COEFF */
-      1017, /* GL_ORDER */
-       343, /* GL_DOMAIN */
-       278, /* GL_CURRENT_COLOR */
-       281, /* GL_CURRENT_INDEX */
-       287, /* GL_CURRENT_NORMAL */
-       299, /* GL_CURRENT_TEXTURE_COORDS */
-       292, /* GL_CURRENT_RASTER_COLOR */
-       294, /* GL_CURRENT_RASTER_INDEX */
-       297, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
-       295, /* GL_CURRENT_RASTER_POSITION */
-       296, /* GL_CURRENT_RASTER_POSITION_VALID */
-       293, /* GL_CURRENT_RASTER_DISTANCE */
-      1089, /* GL_POINT_SMOOTH */
-      1078, /* GL_POINT_SIZE */
-      1088, /* GL_POINT_SIZE_RANGE */
-      1079, /* GL_POINT_SIZE_GRANULARITY */
-       647, /* GL_LINE_SMOOTH */
-       654, /* GL_LINE_WIDTH */
-       656, /* GL_LINE_WIDTH_RANGE */
-       655, /* GL_LINE_WIDTH_GRANULARITY */
-       649, /* GL_LINE_STIPPLE */
-       650, /* GL_LINE_STIPPLE_PATTERN */
-       651, /* GL_LINE_STIPPLE_REPEAT */
-       661, /* GL_LIST_MODE */
-       820, /* GL_MAX_LIST_NESTING */
-       658, /* GL_LIST_BASE */
-       660, /* GL_LIST_INDEX */
-      1099, /* GL_POLYGON_MODE */
-      1106, /* GL_POLYGON_SMOOTH */
-      1108, /* GL_POLYGON_STIPPLE */
-       413, /* GL_EDGE_FLAG */
-       271, /* GL_CULL_FACE */
-       272, /* GL_CULL_FACE_MODE */
-       507, /* GL_FRONT_FACE */
-       629, /* GL_LIGHTING */
-       634, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
-       635, /* GL_LIGHT_MODEL_TWO_SIDE */
-       631, /* GL_LIGHT_MODEL_AMBIENT */
-      1337, /* GL_SHADE_MODEL */
-       167, /* GL_COLOR_MATERIAL_FACE */
-       168, /* GL_COLOR_MATERIAL_PARAMETER */
-       166, /* GL_COLOR_MATERIAL */
-       455, /* GL_FOG */
-       477, /* GL_FOG_INDEX */
-       473, /* GL_FOG_DENSITY */
-       481, /* GL_FOG_START */
-       475, /* GL_FOG_END */
-       478, /* GL_FOG_MODE */
-       457, /* GL_FOG_COLOR */
-       332, /* GL_DEPTH_RANGE */
-       337, /* GL_DEPTH_TEST */
-       340, /* GL_DEPTH_WRITEMASK */
-       320, /* GL_DEPTH_CLEAR_VALUE */
-       331, /* GL_DEPTH_FUNC */
+      1062, /* GL_ORDER */
+       363, /* GL_DOMAIN */
+       294, /* GL_CURRENT_COLOR */
+       297, /* GL_CURRENT_INDEX */
+       303, /* GL_CURRENT_NORMAL */
+       315, /* GL_CURRENT_TEXTURE_COORDS */
+       308, /* GL_CURRENT_RASTER_COLOR */
+       310, /* GL_CURRENT_RASTER_INDEX */
+       313, /* GL_CURRENT_RASTER_TEXTURE_COORDS */
+       311, /* GL_CURRENT_RASTER_POSITION */
+       312, /* GL_CURRENT_RASTER_POSITION_VALID */
+       309, /* GL_CURRENT_RASTER_DISTANCE */
+      1134, /* GL_POINT_SMOOTH */
+      1123, /* GL_POINT_SIZE */
+      1133, /* GL_POINT_SIZE_RANGE */
+      1124, /* GL_POINT_SIZE_GRANULARITY */
+       691, /* GL_LINE_SMOOTH */
+       698, /* GL_LINE_WIDTH */
+       700, /* GL_LINE_WIDTH_RANGE */
+       699, /* GL_LINE_WIDTH_GRANULARITY */
+       693, /* GL_LINE_STIPPLE */
+       694, /* GL_LINE_STIPPLE_PATTERN */
+       695, /* GL_LINE_STIPPLE_REPEAT */
+       705, /* GL_LIST_MODE */
+       864, /* GL_MAX_LIST_NESTING */
+       702, /* GL_LIST_BASE */
+       704, /* GL_LIST_INDEX */
+      1144, /* GL_POLYGON_MODE */
+      1151, /* GL_POLYGON_SMOOTH */
+      1153, /* GL_POLYGON_STIPPLE */
+       434, /* GL_EDGE_FLAG */
+       287, /* GL_CULL_FACE */
+       288, /* GL_CULL_FACE_MODE */
+       549, /* GL_FRONT_FACE */
+       673, /* GL_LIGHTING */
+       678, /* GL_LIGHT_MODEL_LOCAL_VIEWER */
+       679, /* GL_LIGHT_MODEL_TWO_SIDE */
+       675, /* GL_LIGHT_MODEL_AMBIENT */
+      1394, /* GL_SHADE_MODEL */
+       183, /* GL_COLOR_MATERIAL_FACE */
+       184, /* GL_COLOR_MATERIAL_PARAMETER */
+       182, /* GL_COLOR_MATERIAL */
+       476, /* GL_FOG */
+       498, /* GL_FOG_INDEX */
+       494, /* GL_FOG_DENSITY */
+       502, /* GL_FOG_START */
+       496, /* GL_FOG_END */
+       499, /* GL_FOG_MODE */
+       478, /* GL_FOG_COLOR */
+       350, /* GL_DEPTH_RANGE */
+       357, /* GL_DEPTH_TEST */
+       360, /* GL_DEPTH_WRITEMASK */
+       338, /* GL_DEPTH_CLEAR_VALUE */
+       349, /* GL_DEPTH_FUNC */
         12, /* GL_ACCUM_CLEAR_VALUE */
-      1420, /* GL_STENCIL_TEST */
-      1408, /* GL_STENCIL_CLEAR_VALUE */
-      1410, /* GL_STENCIL_FUNC */
-      1422, /* GL_STENCIL_VALUE_MASK */
-      1409, /* GL_STENCIL_FAIL */
-      1417, /* GL_STENCIL_PASS_DEPTH_FAIL */
-      1418, /* GL_STENCIL_PASS_DEPTH_PASS */
-      1419, /* GL_STENCIL_REF */
-      1423, /* GL_STENCIL_WRITEMASK */
-       789, /* GL_MATRIX_MODE */
-       956, /* GL_NORMALIZE */
-      1729, /* GL_VIEWPORT */
-       930, /* GL_MODELVIEW_STACK_DEPTH */
-      1189, /* GL_PROJECTION_STACK_DEPTH */
-      1618, /* GL_TEXTURE_STACK_DEPTH */
-       928, /* GL_MODELVIEW_MATRIX */
-      1188, /* GL_PROJECTION_MATRIX */
-      1603, /* GL_TEXTURE_MATRIX */
+      1479, /* GL_STENCIL_TEST */
+      1467, /* GL_STENCIL_CLEAR_VALUE */
+      1469, /* GL_STENCIL_FUNC */
+      1481, /* GL_STENCIL_VALUE_MASK */
+      1468, /* GL_STENCIL_FAIL */
+      1476, /* GL_STENCIL_PASS_DEPTH_FAIL */
+      1477, /* GL_STENCIL_PASS_DEPTH_PASS */
+      1478, /* GL_STENCIL_REF */
+      1482, /* GL_STENCIL_WRITEMASK */
+       833, /* GL_MATRIX_MODE */
+      1001, /* GL_NORMALIZE */
+      1791, /* GL_VIEWPORT */
+       975, /* GL_MODELVIEW_STACK_DEPTH */
+      1234, /* GL_PROJECTION_STACK_DEPTH */
+      1677, /* GL_TEXTURE_STACK_DEPTH */
+       973, /* GL_MODELVIEW_MATRIX */
+      1233, /* GL_PROJECTION_MATRIX */
+      1662, /* GL_TEXTURE_MATRIX */
         60, /* GL_ATTRIB_STACK_DEPTH */
        126, /* GL_CLIENT_ATTRIB_STACK_DEPTH */
         43, /* GL_ALPHA_TEST */
         44, /* GL_ALPHA_TEST_FUNC */
         45, /* GL_ALPHA_TEST_REF */
-       342, /* GL_DITHER */
+       362, /* GL_DITHER */
         77, /* GL_BLEND_DST */
         85, /* GL_BLEND_SRC */
         74, /* GL_BLEND */
-       664, /* GL_LOGIC_OP_MODE */
-       584, /* GL_INDEX_LOGIC_OP */
-       165, /* GL_COLOR_LOGIC_OP */
+       708, /* GL_LOGIC_OP_MODE */
+       627, /* GL_INDEX_LOGIC_OP */
+       181, /* GL_COLOR_LOGIC_OP */
         66, /* GL_AUX_BUFFERS */
-       353, /* GL_DRAW_BUFFER */
-      1221, /* GL_READ_BUFFER */
-      1318, /* GL_SCISSOR_BOX */
-      1319, /* GL_SCISSOR_TEST */
-       583, /* GL_INDEX_CLEAR_VALUE */
-       588, /* GL_INDEX_WRITEMASK */
-       162, /* GL_COLOR_CLEAR_VALUE */
-       204, /* GL_COLOR_WRITEMASK */
-       585, /* GL_INDEX_MODE */
-      1285, /* GL_RGBA_MODE */
-       352, /* GL_DOUBLEBUFFER */
-      1424, /* GL_STEREO */
-      1244, /* GL_RENDER_MODE */
-      1041, /* GL_PERSPECTIVE_CORRECTION_HINT */
-      1090, /* GL_POINT_SMOOTH_HINT */
-       648, /* GL_LINE_SMOOTH_HINT */
-      1107, /* GL_POLYGON_SMOOTH_HINT */
-       476, /* GL_FOG_HINT */
-      1584, /* GL_TEXTURE_GEN_S */
-      1585, /* GL_TEXTURE_GEN_T */
-      1583, /* GL_TEXTURE_GEN_R */
-      1582, /* GL_TEXTURE_GEN_Q */
-      1054, /* GL_PIXEL_MAP_I_TO_I */
-      1060, /* GL_PIXEL_MAP_S_TO_S */
-      1056, /* GL_PIXEL_MAP_I_TO_R */
-      1052, /* GL_PIXEL_MAP_I_TO_G */
-      1050, /* GL_PIXEL_MAP_I_TO_B */
-      1048, /* GL_PIXEL_MAP_I_TO_A */
-      1058, /* GL_PIXEL_MAP_R_TO_R */
-      1046, /* GL_PIXEL_MAP_G_TO_G */
-      1044, /* GL_PIXEL_MAP_B_TO_B */
-      1042, /* GL_PIXEL_MAP_A_TO_A */
-      1055, /* GL_PIXEL_MAP_I_TO_I_SIZE */
-      1061, /* GL_PIXEL_MAP_S_TO_S_SIZE */
-      1057, /* GL_PIXEL_MAP_I_TO_R_SIZE */
-      1053, /* GL_PIXEL_MAP_I_TO_G_SIZE */
-      1051, /* GL_PIXEL_MAP_I_TO_B_SIZE */
-      1049, /* GL_PIXEL_MAP_I_TO_A_SIZE */
-      1059, /* GL_PIXEL_MAP_R_TO_R_SIZE */
-      1047, /* GL_PIXEL_MAP_G_TO_G_SIZE */
-      1045, /* GL_PIXEL_MAP_B_TO_B_SIZE */
-      1043, /* GL_PIXEL_MAP_A_TO_A_SIZE */
-      1651, /* GL_UNPACK_SWAP_BYTES */
-      1646, /* GL_UNPACK_LSB_FIRST */
-      1647, /* GL_UNPACK_ROW_LENGTH */
-      1650, /* GL_UNPACK_SKIP_ROWS */
-      1649, /* GL_UNPACK_SKIP_PIXELS */
-      1644, /* GL_UNPACK_ALIGNMENT */
-      1029, /* GL_PACK_SWAP_BYTES */
-      1024, /* GL_PACK_LSB_FIRST */
-      1025, /* GL_PACK_ROW_LENGTH */
-      1028, /* GL_PACK_SKIP_ROWS */
-      1027, /* GL_PACK_SKIP_PIXELS */
-      1021, /* GL_PACK_ALIGNMENT */
-       742, /* GL_MAP_COLOR */
-       743, /* GL_MAP_STENCIL */
-       587, /* GL_INDEX_SHIFT */
-       586, /* GL_INDEX_OFFSET */
-      1233, /* GL_RED_SCALE */
-      1231, /* GL_RED_BIAS */
-      1746, /* GL_ZOOM_X */
-      1747, /* GL_ZOOM_Y */
-       549, /* GL_GREEN_SCALE */
-       547, /* GL_GREEN_BIAS */
+       373, /* GL_DRAW_BUFFER */
+      1266, /* GL_READ_BUFFER */
+      1375, /* GL_SCISSOR_BOX */
+      1376, /* GL_SCISSOR_TEST */
+       626, /* GL_INDEX_CLEAR_VALUE */
+       631, /* GL_INDEX_WRITEMASK */
+       178, /* GL_COLOR_CLEAR_VALUE */
+       220, /* GL_COLOR_WRITEMASK */
+       628, /* GL_INDEX_MODE */
+      1342, /* GL_RGBA_MODE */
+       372, /* GL_DOUBLEBUFFER */
+      1483, /* GL_STEREO */
+      1301, /* GL_RENDER_MODE */
+      1086, /* GL_PERSPECTIVE_CORRECTION_HINT */
+      1135, /* GL_POINT_SMOOTH_HINT */
+       692, /* GL_LINE_SMOOTH_HINT */
+      1152, /* GL_POLYGON_SMOOTH_HINT */
+       497, /* GL_FOG_HINT */
+      1643, /* GL_TEXTURE_GEN_S */
+      1644, /* GL_TEXTURE_GEN_T */
+      1642, /* GL_TEXTURE_GEN_R */
+      1641, /* GL_TEXTURE_GEN_Q */
+      1099, /* GL_PIXEL_MAP_I_TO_I */
+      1105, /* GL_PIXEL_MAP_S_TO_S */
+      1101, /* GL_PIXEL_MAP_I_TO_R */
+      1097, /* GL_PIXEL_MAP_I_TO_G */
+      1095, /* GL_PIXEL_MAP_I_TO_B */
+      1093, /* GL_PIXEL_MAP_I_TO_A */
+      1103, /* GL_PIXEL_MAP_R_TO_R */
+      1091, /* GL_PIXEL_MAP_G_TO_G */
+      1089, /* GL_PIXEL_MAP_B_TO_B */
+      1087, /* GL_PIXEL_MAP_A_TO_A */
+      1100, /* GL_PIXEL_MAP_I_TO_I_SIZE */
+      1106, /* GL_PIXEL_MAP_S_TO_S_SIZE */
+      1102, /* GL_PIXEL_MAP_I_TO_R_SIZE */
+      1098, /* GL_PIXEL_MAP_I_TO_G_SIZE */
+      1096, /* GL_PIXEL_MAP_I_TO_B_SIZE */
+      1094, /* GL_PIXEL_MAP_I_TO_A_SIZE */
+      1104, /* GL_PIXEL_MAP_R_TO_R_SIZE */
+      1092, /* GL_PIXEL_MAP_G_TO_G_SIZE */
+      1090, /* GL_PIXEL_MAP_B_TO_B_SIZE */
+      1088, /* GL_PIXEL_MAP_A_TO_A_SIZE */
+      1711, /* GL_UNPACK_SWAP_BYTES */
+      1706, /* GL_UNPACK_LSB_FIRST */
+      1707, /* GL_UNPACK_ROW_LENGTH */
+      1710, /* GL_UNPACK_SKIP_ROWS */
+      1709, /* GL_UNPACK_SKIP_PIXELS */
+      1704, /* GL_UNPACK_ALIGNMENT */
+      1074, /* GL_PACK_SWAP_BYTES */
+      1069, /* GL_PACK_LSB_FIRST */
+      1070, /* GL_PACK_ROW_LENGTH */
+      1073, /* GL_PACK_SKIP_ROWS */
+      1072, /* GL_PACK_SKIP_PIXELS */
+      1066, /* GL_PACK_ALIGNMENT */
+       786, /* GL_MAP_COLOR */
+       787, /* GL_MAP_STENCIL */
+       630, /* GL_INDEX_SHIFT */
+       629, /* GL_INDEX_OFFSET */
+      1279, /* GL_RED_SCALE */
+      1277, /* GL_RED_BIAS */
+      1808, /* GL_ZOOM_X */
+      1809, /* GL_ZOOM_Y */
+       591, /* GL_GREEN_SCALE */
+       589, /* GL_GREEN_BIAS */
         91, /* GL_BLUE_SCALE */
         89, /* GL_BLUE_BIAS */
         42, /* GL_ALPHA_SCALE */
         40, /* GL_ALPHA_BIAS */
-       333, /* GL_DEPTH_SCALE */
-       314, /* GL_DEPTH_BIAS */
-       815, /* GL_MAX_EVAL_ORDER */
-       819, /* GL_MAX_LIGHTS */
-       798, /* GL_MAX_CLIP_PLANES */
-       863, /* GL_MAX_TEXTURE_SIZE */
-       825, /* GL_MAX_PIXEL_MAP_TABLE */
-       794, /* GL_MAX_ATTRIB_STACK_DEPTH */
-       822, /* GL_MAX_MODELVIEW_STACK_DEPTH */
-       823, /* GL_MAX_NAME_STACK_DEPTH */
-       851, /* GL_MAX_PROJECTION_STACK_DEPTH */
-       864, /* GL_MAX_TEXTURE_STACK_DEPTH */
-       878, /* GL_MAX_VIEWPORT_DIMS */
-       795, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
-      1431, /* GL_SUBPIXEL_BITS */
-       582, /* GL_INDEX_BITS */
-      1232, /* GL_RED_BITS */
-       548, /* GL_GREEN_BITS */
+       351, /* GL_DEPTH_SCALE */
+       332, /* GL_DEPTH_BIAS */
+       859, /* GL_MAX_EVAL_ORDER */
+       863, /* GL_MAX_LIGHTS */
+       842, /* GL_MAX_CLIP_PLANES */
+       908, /* GL_MAX_TEXTURE_SIZE */
+       869, /* GL_MAX_PIXEL_MAP_TABLE */
+       838, /* GL_MAX_ATTRIB_STACK_DEPTH */
+       866, /* GL_MAX_MODELVIEW_STACK_DEPTH */
+       867, /* GL_MAX_NAME_STACK_DEPTH */
+       895, /* GL_MAX_PROJECTION_STACK_DEPTH */
+       909, /* GL_MAX_TEXTURE_STACK_DEPTH */
+       923, /* GL_MAX_VIEWPORT_DIMS */
+       839, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */
+      1490, /* GL_SUBPIXEL_BITS */
+       625, /* GL_INDEX_BITS */
+      1278, /* GL_RED_BITS */
+       590, /* GL_GREEN_BITS */
         90, /* GL_BLUE_BITS */
         41, /* GL_ALPHA_BITS */
-       315, /* GL_DEPTH_BITS */
-      1406, /* GL_STENCIL_BITS */
+       333, /* GL_DEPTH_BITS */
+      1465, /* GL_STENCIL_BITS */
         14, /* GL_ACCUM_RED_BITS */
         13, /* GL_ACCUM_GREEN_BITS */
         10, /* GL_ACCUM_BLUE_BITS */
          9, /* GL_ACCUM_ALPHA_BITS */
-       944, /* GL_NAME_STACK_DEPTH */
+       989, /* GL_NAME_STACK_DEPTH */
         61, /* GL_AUTO_NORMAL */
-       688, /* GL_MAP1_COLOR_4 */
-       691, /* GL_MAP1_INDEX */
-       692, /* GL_MAP1_NORMAL */
-       693, /* GL_MAP1_TEXTURE_COORD_1 */
-       694, /* GL_MAP1_TEXTURE_COORD_2 */
-       695, /* GL_MAP1_TEXTURE_COORD_3 */
-       696, /* GL_MAP1_TEXTURE_COORD_4 */
-       697, /* GL_MAP1_VERTEX_3 */
-       698, /* GL_MAP1_VERTEX_4 */
-       715, /* GL_MAP2_COLOR_4 */
-       718, /* GL_MAP2_INDEX */
-       719, /* GL_MAP2_NORMAL */
-       720, /* GL_MAP2_TEXTURE_COORD_1 */
-       721, /* GL_MAP2_TEXTURE_COORD_2 */
-       722, /* GL_MAP2_TEXTURE_COORD_3 */
-       723, /* GL_MAP2_TEXTURE_COORD_4 */
-       724, /* GL_MAP2_VERTEX_3 */
-       725, /* GL_MAP2_VERTEX_4 */
-       689, /* GL_MAP1_GRID_DOMAIN */
-       690, /* GL_MAP1_GRID_SEGMENTS */
-       716, /* GL_MAP2_GRID_DOMAIN */
-       717, /* GL_MAP2_GRID_SEGMENTS */
-      1508, /* GL_TEXTURE_1D */
-      1510, /* GL_TEXTURE_2D */
-       437, /* GL_FEEDBACK_BUFFER_POINTER */
-       438, /* GL_FEEDBACK_BUFFER_SIZE */
-       439, /* GL_FEEDBACK_BUFFER_TYPE */
-      1328, /* GL_SELECTION_BUFFER_POINTER */
-      1329, /* GL_SELECTION_BUFFER_SIZE */
-      1621, /* GL_TEXTURE_WIDTH */
-      1589, /* GL_TEXTURE_HEIGHT */
-      1545, /* GL_TEXTURE_COMPONENTS */
-      1529, /* GL_TEXTURE_BORDER_COLOR */
-      1528, /* GL_TEXTURE_BORDER */
-       344, /* GL_DONT_CARE */
-       435, /* GL_FASTEST */
-       952, /* GL_NICEST */
+       732, /* GL_MAP1_COLOR_4 */
+       735, /* GL_MAP1_INDEX */
+       736, /* GL_MAP1_NORMAL */
+       737, /* GL_MAP1_TEXTURE_COORD_1 */
+       738, /* GL_MAP1_TEXTURE_COORD_2 */
+       739, /* GL_MAP1_TEXTURE_COORD_3 */
+       740, /* GL_MAP1_TEXTURE_COORD_4 */
+       741, /* GL_MAP1_VERTEX_3 */
+       742, /* GL_MAP1_VERTEX_4 */
+       759, /* GL_MAP2_COLOR_4 */
+       762, /* GL_MAP2_INDEX */
+       763, /* GL_MAP2_NORMAL */
+       764, /* GL_MAP2_TEXTURE_COORD_1 */
+       765, /* GL_MAP2_TEXTURE_COORD_2 */
+       766, /* GL_MAP2_TEXTURE_COORD_3 */
+       767, /* GL_MAP2_TEXTURE_COORD_4 */
+       768, /* GL_MAP2_VERTEX_3 */
+       769, /* GL_MAP2_VERTEX_4 */
+       733, /* GL_MAP1_GRID_DOMAIN */
+       734, /* GL_MAP1_GRID_SEGMENTS */
+       760, /* GL_MAP2_GRID_DOMAIN */
+       761, /* GL_MAP2_GRID_SEGMENTS */
+      1567, /* GL_TEXTURE_1D */
+      1569, /* GL_TEXTURE_2D */
+       458, /* GL_FEEDBACK_BUFFER_POINTER */
+       459, /* GL_FEEDBACK_BUFFER_SIZE */
+       460, /* GL_FEEDBACK_BUFFER_TYPE */
+      1385, /* GL_SELECTION_BUFFER_POINTER */
+      1386, /* GL_SELECTION_BUFFER_SIZE */
+      1681, /* GL_TEXTURE_WIDTH */
+      1648, /* GL_TEXTURE_HEIGHT */
+      1604, /* GL_TEXTURE_COMPONENTS */
+      1588, /* GL_TEXTURE_BORDER_COLOR */
+      1587, /* GL_TEXTURE_BORDER */
+       364, /* GL_DONT_CARE */
+       456, /* GL_FASTEST */
+       997, /* GL_NICEST */
         47, /* GL_AMBIENT */
-       341, /* GL_DIFFUSE */
-      1371, /* GL_SPECULAR */
-      1111, /* GL_POSITION */
-      1374, /* GL_SPOT_DIRECTION */
-      1375, /* GL_SPOT_EXPONENT */
-      1373, /* GL_SPOT_CUTOFF */
-       244, /* GL_CONSTANT_ATTENUATION */
-       638, /* GL_LINEAR_ATTENUATION */
-      1208, /* GL_QUADRATIC_ATTENUATION */
-       218, /* GL_COMPILE */
-       219, /* GL_COMPILE_AND_EXECUTE */
+       361, /* GL_DIFFUSE */
+      1428, /* GL_SPECULAR */
+      1156, /* GL_POSITION */
+      1431, /* GL_SPOT_DIRECTION */
+      1432, /* GL_SPOT_EXPONENT */
+      1430, /* GL_SPOT_CUTOFF */
+       260, /* GL_CONSTANT_ATTENUATION */
+       682, /* GL_LINEAR_ATTENUATION */
+      1253, /* GL_QUADRATIC_ATTENUATION */
+       234, /* GL_COMPILE */
+       235, /* GL_COMPILE_AND_EXECUTE */
        110, /* GL_BYTE */
-      1652, /* GL_UNSIGNED_BYTE */
-      1342, /* GL_SHORT */
-      1661, /* GL_UNSIGNED_SHORT */
-       590, /* GL_INT */
-      1655, /* GL_UNSIGNED_INT */
-       442, /* GL_FLOAT */
+      1712, /* GL_UNSIGNED_BYTE */
+      1399, /* GL_SHORT */
+      1723, /* GL_UNSIGNED_SHORT */
+       633, /* GL_INT */
+      1715, /* GL_UNSIGNED_INT */
+       463, /* GL_FLOAT */
          1, /* GL_2_BYTES */
          5, /* GL_3_BYTES */
          7, /* GL_4_BYTES */
-       351, /* GL_DOUBLE */
+       371, /* GL_DOUBLE */
        122, /* GL_CLEAR */
         49, /* GL_AND */
         51, /* GL_AND_REVERSE */
-       268, /* GL_COPY */
+       284, /* GL_COPY */
         50, /* GL_AND_INVERTED */
-       954, /* GL_NOOP */
-      1742, /* GL_XOR */
-      1016, /* GL_OR */
-       955, /* GL_NOR */
-       425, /* GL_EQUIV */
-       616, /* GL_INVERT */
-      1019, /* GL_OR_REVERSE */
-       269, /* GL_COPY_INVERTED */
-      1018, /* GL_OR_INVERTED */
-       945, /* GL_NAND */
-      1333, /* GL_SET */
-       422, /* GL_EMISSION */
-      1341, /* GL_SHININESS */
+       999, /* GL_NOOP */
+      1804, /* GL_XOR */
+      1061, /* GL_OR */
+      1000, /* GL_NOR */
+       446, /* GL_EQUIV */
+       660, /* GL_INVERT */
+      1064, /* GL_OR_REVERSE */
+       285, /* GL_COPY_INVERTED */
+      1063, /* GL_OR_INVERTED */
+       990, /* GL_NAND */
+      1390, /* GL_SET */
+       443, /* GL_EMISSION */
+      1398, /* GL_SHININESS */
         48, /* GL_AMBIENT_AND_DIFFUSE */
-       164, /* GL_COLOR_INDEXES */
-       895, /* GL_MODELVIEW */
-      1187, /* GL_PROJECTION */
-      1443, /* GL_TEXTURE */
+       180, /* GL_COLOR_INDEXES */
+       940, /* GL_MODELVIEW */
+      1232, /* GL_PROJECTION */
+      1502, /* GL_TEXTURE */
        137, /* GL_COLOR */
-       312, /* GL_DEPTH */
-      1393, /* GL_STENCIL */
-       163, /* GL_COLOR_INDEX */
-      1411, /* GL_STENCIL_INDEX */
-       321, /* GL_DEPTH_COMPONENT */
-      1228, /* GL_RED */
-       546, /* GL_GREEN */
+       328, /* GL_DEPTH */
+      1451, /* GL_STENCIL */
+       179, /* GL_COLOR_INDEX */
+      1470, /* GL_STENCIL_INDEX */
+       339, /* GL_DEPTH_COMPONENT */
+      1274, /* GL_RED */
+       588, /* GL_GREEN */
         88, /* GL_BLUE */
         31, /* GL_ALPHA */
-      1252, /* GL_RGB */
-      1271, /* GL_RGBA */
-       666, /* GL_LUMINANCE */
-       687, /* GL_LUMINANCE_ALPHA */
+      1309, /* GL_RGB */
+      1328, /* GL_RGBA */
+       710, /* GL_LUMINANCE */
+       731, /* GL_LUMINANCE_ALPHA */
         72, /* GL_BITMAP */
-      1067, /* GL_POINT */
-       636, /* GL_LINE */
-       440, /* GL_FILL */
-      1237, /* GL_RENDER */
-       436, /* GL_FEEDBACK */
-      1327, /* GL_SELECT */
-       441, /* GL_FLAT */
-      1346, /* GL_SMOOTH */
-       617, /* GL_KEEP */
-      1246, /* GL_REPLACE */
-       573, /* GL_INCR */
-       308, /* GL_DECR */
-      1676, /* GL_VENDOR */
-      1243, /* GL_RENDERER */
-      1677, /* GL_VERSION */
-       429, /* GL_EXTENSIONS */
-      1292, /* GL_S */
-      1434, /* GL_T */
-      1218, /* GL_R */
-      1207, /* GL_Q */
-       931, /* GL_MODULATE */
-       307, /* GL_DECAL */
-      1579, /* GL_TEXTURE_ENV_MODE */
-      1578, /* GL_TEXTURE_ENV_COLOR */
-      1577, /* GL_TEXTURE_ENV */
-       430, /* GL_EYE_LINEAR */
-       978, /* GL_OBJECT_LINEAR */
-      1372, /* GL_SPHERE_MAP */
-      1581, /* GL_TEXTURE_GEN_MODE */
-       980, /* GL_OBJECT_PLANE */
-       431, /* GL_EYE_PLANE */
-       946, /* GL_NEAREST */
-       637, /* GL_LINEAR */
-       950, /* GL_NEAREST_MIPMAP_NEAREST */
-       642, /* GL_LINEAR_MIPMAP_NEAREST */
-       949, /* GL_NEAREST_MIPMAP_LINEAR */
-       641, /* GL_LINEAR_MIPMAP_LINEAR */
-      1602, /* GL_TEXTURE_MAG_FILTER */
-      1610, /* GL_TEXTURE_MIN_FILTER */
-      1623, /* GL_TEXTURE_WRAP_S */
-      1624, /* GL_TEXTURE_WRAP_T */
+      1112, /* GL_POINT */
+       680, /* GL_LINE */
+       461, /* GL_FILL */
+      1283, /* GL_RENDER */
+       457, /* GL_FEEDBACK */
+      1384, /* GL_SELECT */
+       462, /* GL_FLAT */
+      1403, /* GL_SMOOTH */
+       661, /* GL_KEEP */
+      1303, /* GL_REPLACE */
+       615, /* GL_INCR */
+       324, /* GL_DECR */
+      1738, /* GL_VENDOR */
+      1300, /* GL_RENDERER */
+      1739, /* GL_VERSION */
+       450, /* GL_EXTENSIONS */
+      1349, /* GL_S */
+      1493, /* GL_T */
+      1263, /* GL_R */
+      1252, /* GL_Q */
+       976, /* GL_MODULATE */
+       323, /* GL_DECAL */
+      1638, /* GL_TEXTURE_ENV_MODE */
+      1637, /* GL_TEXTURE_ENV_COLOR */
+      1636, /* GL_TEXTURE_ENV */
+       451, /* GL_EYE_LINEAR */
+      1023, /* GL_OBJECT_LINEAR */
+      1429, /* GL_SPHERE_MAP */
+      1640, /* GL_TEXTURE_GEN_MODE */
+      1025, /* GL_OBJECT_PLANE */
+       452, /* GL_EYE_PLANE */
+       991, /* GL_NEAREST */
+       681, /* GL_LINEAR */
+       995, /* GL_NEAREST_MIPMAP_NEAREST */
+       686, /* GL_LINEAR_MIPMAP_NEAREST */
+       994, /* GL_NEAREST_MIPMAP_LINEAR */
+       685, /* GL_LINEAR_MIPMAP_LINEAR */
+      1661, /* GL_TEXTURE_MAG_FILTER */
+      1669, /* GL_TEXTURE_MIN_FILTER */
+      1683, /* GL_TEXTURE_WRAP_S */
+      1684, /* GL_TEXTURE_WRAP_T */
        116, /* GL_CLAMP */
-      1245, /* GL_REPEAT */
-      1105, /* GL_POLYGON_OFFSET_UNITS */
-      1104, /* GL_POLYGON_OFFSET_POINT */
-      1103, /* GL_POLYGON_OFFSET_LINE */
-      1219, /* GL_R3_G3_B2 */
-      1673, /* GL_V2F */
-      1674, /* GL_V3F */
+      1302, /* GL_REPEAT */
+      1150, /* GL_POLYGON_OFFSET_UNITS */
+      1149, /* GL_POLYGON_OFFSET_POINT */
+      1148, /* GL_POLYGON_OFFSET_LINE */
+      1264, /* GL_R3_G3_B2 */
+      1735, /* GL_V2F */
+      1736, /* GL_V3F */
        113, /* GL_C4UB_V2F */
        114, /* GL_C4UB_V3F */
        111, /* GL_C3F_V3F */
-       943, /* GL_N3F_V3F */
+       988, /* GL_N3F_V3F */
        112, /* GL_C4F_N3F_V3F */
-      1439, /* GL_T2F_V3F */
-      1441, /* GL_T4F_V4F */
-      1437, /* GL_T2F_C4UB_V3F */
-      1435, /* GL_T2F_C3F_V3F */
-      1438, /* GL_T2F_N3F_V3F */
-      1436, /* GL_T2F_C4F_N3F_V3F */
-      1440, /* GL_T4F_C4F_N3F_V4F */
+      1498, /* GL_T2F_V3F */
+      1500, /* GL_T4F_V4F */
+      1496, /* GL_T2F_C4UB_V3F */
+      1494, /* GL_T2F_C3F_V3F */
+      1497, /* GL_T2F_N3F_V3F */
+      1495, /* GL_T2F_C4F_N3F_V3F */
+      1499, /* GL_T4F_C4F_N3F_V4F */
        129, /* GL_CLIP_PLANE0 */
        130, /* GL_CLIP_PLANE1 */
        131, /* GL_CLIP_PLANE2 */
        132, /* GL_CLIP_PLANE3 */
        133, /* GL_CLIP_PLANE4 */
        134, /* GL_CLIP_PLANE5 */
-       621, /* GL_LIGHT0 */
-       622, /* GL_LIGHT1 */
-       623, /* GL_LIGHT2 */
-       624, /* GL_LIGHT3 */
-       625, /* GL_LIGHT4 */
-       626, /* GL_LIGHT5 */
-       627, /* GL_LIGHT6 */
-       628, /* GL_LIGHT7 */
-       550, /* GL_HINT_BIT */
-       246, /* GL_CONSTANT_COLOR */
-       990, /* GL_ONE_MINUS_CONSTANT_COLOR */
-       241, /* GL_CONSTANT_ALPHA */
-       988, /* GL_ONE_MINUS_CONSTANT_ALPHA */
+       665, /* GL_LIGHT0 */
+       666, /* GL_LIGHT1 */
+       667, /* GL_LIGHT2 */
+       668, /* GL_LIGHT3 */
+       669, /* GL_LIGHT4 */
+       670, /* GL_LIGHT5 */
+       671, /* GL_LIGHT6 */
+       672, /* GL_LIGHT7 */
+       592, /* GL_HINT_BIT */
+       262, /* GL_CONSTANT_COLOR */
+      1035, /* GL_ONE_MINUS_CONSTANT_COLOR */
+       257, /* GL_CONSTANT_ALPHA */
+      1033, /* GL_ONE_MINUS_CONSTANT_ALPHA */
         75, /* GL_BLEND_COLOR */
-       510, /* GL_FUNC_ADD */
-       879, /* GL_MIN */
-       791, /* GL_MAX */
+       552, /* GL_FUNC_ADD */
+       924, /* GL_MIN */
+       835, /* GL_MAX */
         80, /* GL_BLEND_EQUATION */
-       514, /* GL_FUNC_SUBTRACT */
-       512, /* GL_FUNC_REVERSE_SUBTRACT */
-       249, /* GL_CONVOLUTION_1D */
-       250, /* GL_CONVOLUTION_2D */
-      1330, /* GL_SEPARABLE_2D */
-       253, /* GL_CONVOLUTION_BORDER_MODE */
-       257, /* GL_CONVOLUTION_FILTER_SCALE */
-       255, /* GL_CONVOLUTION_FILTER_BIAS */
-      1229, /* GL_REDUCE */
-       259, /* GL_CONVOLUTION_FORMAT */
-       263, /* GL_CONVOLUTION_WIDTH */
-       261, /* GL_CONVOLUTION_HEIGHT */
-       806, /* GL_MAX_CONVOLUTION_WIDTH */
-       804, /* GL_MAX_CONVOLUTION_HEIGHT */
-      1144, /* GL_POST_CONVOLUTION_RED_SCALE */
-      1140, /* GL_POST_CONVOLUTION_GREEN_SCALE */
-      1135, /* GL_POST_CONVOLUTION_BLUE_SCALE */
-      1131, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
-      1142, /* GL_POST_CONVOLUTION_RED_BIAS */
-      1138, /* GL_POST_CONVOLUTION_GREEN_BIAS */
-      1133, /* GL_POST_CONVOLUTION_BLUE_BIAS */
-      1129, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
-       551, /* GL_HISTOGRAM */
-      1191, /* GL_PROXY_HISTOGRAM */
-       567, /* GL_HISTOGRAM_WIDTH */
-       557, /* GL_HISTOGRAM_FORMAT */
-       563, /* GL_HISTOGRAM_RED_SIZE */
-       559, /* GL_HISTOGRAM_GREEN_SIZE */
-       554, /* GL_HISTOGRAM_BLUE_SIZE */
-       552, /* GL_HISTOGRAM_ALPHA_SIZE */
-       561, /* GL_HISTOGRAM_LUMINANCE_SIZE */
-       565, /* GL_HISTOGRAM_SINK */
-       880, /* GL_MINMAX */
-       882, /* GL_MINMAX_FORMAT */
-       884, /* GL_MINMAX_SINK */
-      1442, /* GL_TABLE_TOO_LARGE_EXT */
-      1654, /* GL_UNSIGNED_BYTE_3_3_2 */
-      1663, /* GL_UNSIGNED_SHORT_4_4_4_4 */
-      1665, /* GL_UNSIGNED_SHORT_5_5_5_1 */
-      1659, /* GL_UNSIGNED_INT_8_8_8_8 */
-      1656, /* GL_UNSIGNED_INT_10_10_10_2 */
-      1102, /* GL_POLYGON_OFFSET_FILL */
-      1101, /* GL_POLYGON_OFFSET_FACTOR */
-      1100, /* GL_POLYGON_OFFSET_BIAS */
-      1249, /* GL_RESCALE_NORMAL */
+       556, /* GL_FUNC_SUBTRACT */
+       554, /* GL_FUNC_REVERSE_SUBTRACT */
+       265, /* GL_CONVOLUTION_1D */
+       266, /* GL_CONVOLUTION_2D */
+      1387, /* GL_SEPARABLE_2D */
+       269, /* GL_CONVOLUTION_BORDER_MODE */
+       273, /* GL_CONVOLUTION_FILTER_SCALE */
+       271, /* GL_CONVOLUTION_FILTER_BIAS */
+      1275, /* GL_REDUCE */
+       275, /* GL_CONVOLUTION_FORMAT */
+       279, /* GL_CONVOLUTION_WIDTH */
+       277, /* GL_CONVOLUTION_HEIGHT */
+       850, /* GL_MAX_CONVOLUTION_WIDTH */
+       848, /* GL_MAX_CONVOLUTION_HEIGHT */
+      1189, /* GL_POST_CONVOLUTION_RED_SCALE */
+      1185, /* GL_POST_CONVOLUTION_GREEN_SCALE */
+      1180, /* GL_POST_CONVOLUTION_BLUE_SCALE */
+      1176, /* GL_POST_CONVOLUTION_ALPHA_SCALE */
+      1187, /* GL_POST_CONVOLUTION_RED_BIAS */
+      1183, /* GL_POST_CONVOLUTION_GREEN_BIAS */
+      1178, /* GL_POST_CONVOLUTION_BLUE_BIAS */
+      1174, /* GL_POST_CONVOLUTION_ALPHA_BIAS */
+       593, /* GL_HISTOGRAM */
+      1236, /* GL_PROXY_HISTOGRAM */
+       609, /* GL_HISTOGRAM_WIDTH */
+       599, /* GL_HISTOGRAM_FORMAT */
+       605, /* GL_HISTOGRAM_RED_SIZE */
+       601, /* GL_HISTOGRAM_GREEN_SIZE */
+       596, /* GL_HISTOGRAM_BLUE_SIZE */
+       594, /* GL_HISTOGRAM_ALPHA_SIZE */
+       603, /* GL_HISTOGRAM_LUMINANCE_SIZE */
+       607, /* GL_HISTOGRAM_SINK */
+       925, /* GL_MINMAX */
+       927, /* GL_MINMAX_FORMAT */
+       929, /* GL_MINMAX_SINK */
+      1501, /* GL_TABLE_TOO_LARGE_EXT */
+      1714, /* GL_UNSIGNED_BYTE_3_3_2 */
+      1725, /* GL_UNSIGNED_SHORT_4_4_4_4 */
+      1727, /* GL_UNSIGNED_SHORT_5_5_5_1 */
+      1720, /* GL_UNSIGNED_INT_8_8_8_8 */
+      1716, /* GL_UNSIGNED_INT_10_10_10_2 */
+      1147, /* GL_POLYGON_OFFSET_FILL */
+      1146, /* GL_POLYGON_OFFSET_FACTOR */
+      1145, /* GL_POLYGON_OFFSET_BIAS */
+      1306, /* GL_RESCALE_NORMAL */
         36, /* GL_ALPHA4 */
         38, /* GL_ALPHA8 */
         32, /* GL_ALPHA12 */
         34, /* GL_ALPHA16 */
-       677, /* GL_LUMINANCE4 */
-       683, /* GL_LUMINANCE8 */
-       667, /* GL_LUMINANCE12 */
-       673, /* GL_LUMINANCE16 */
-       678, /* GL_LUMINANCE4_ALPHA4 */
-       681, /* GL_LUMINANCE6_ALPHA2 */
-       684, /* GL_LUMINANCE8_ALPHA8 */
-       670, /* GL_LUMINANCE12_ALPHA4 */
-       668, /* GL_LUMINANCE12_ALPHA12 */
-       674, /* GL_LUMINANCE16_ALPHA16 */
-       591, /* GL_INTENSITY */
-       596, /* GL_INTENSITY4 */
-       598, /* GL_INTENSITY8 */
-       592, /* GL_INTENSITY12 */
-       594, /* GL_INTENSITY16 */
-      1261, /* GL_RGB2_EXT */
-      1262, /* GL_RGB4 */
-      1265, /* GL_RGB5 */
-      1269, /* GL_RGB8 */
-      1253, /* GL_RGB10 */
-      1257, /* GL_RGB12 */
-      1259, /* GL_RGB16 */
-      1276, /* GL_RGBA2 */
-      1278, /* GL_RGBA4 */
-      1266, /* GL_RGB5_A1 */
-      1282, /* GL_RGBA8 */
-      1254, /* GL_RGB10_A2 */
-      1272, /* GL_RGBA12 */
-      1274, /* GL_RGBA16 */
-      1615, /* GL_TEXTURE_RED_SIZE */
-      1587, /* GL_TEXTURE_GREEN_SIZE */
-      1526, /* GL_TEXTURE_BLUE_SIZE */
-      1513, /* GL_TEXTURE_ALPHA_SIZE */
-      1600, /* GL_TEXTURE_LUMINANCE_SIZE */
-      1591, /* GL_TEXTURE_INTENSITY_SIZE */
-      1247, /* GL_REPLACE_EXT */
-      1195, /* GL_PROXY_TEXTURE_1D */
-      1198, /* GL_PROXY_TEXTURE_2D */
-      1619, /* GL_TEXTURE_TOO_LARGE_EXT */
-      1612, /* GL_TEXTURE_PRIORITY */
-      1617, /* GL_TEXTURE_RESIDENT */
-      1516, /* GL_TEXTURE_BINDING_1D */
-      1518, /* GL_TEXTURE_BINDING_2D */
-      1520, /* GL_TEXTURE_BINDING_3D */
-      1026, /* GL_PACK_SKIP_IMAGES */
-      1022, /* GL_PACK_IMAGE_HEIGHT */
-      1648, /* GL_UNPACK_SKIP_IMAGES */
-      1645, /* GL_UNPACK_IMAGE_HEIGHT */
-      1512, /* GL_TEXTURE_3D */
-      1201, /* GL_PROXY_TEXTURE_3D */
-      1574, /* GL_TEXTURE_DEPTH */
-      1622, /* GL_TEXTURE_WRAP_R */
-       792, /* GL_MAX_3D_TEXTURE_SIZE */
-      1678, /* GL_VERTEX_ARRAY */
-       957, /* GL_NORMAL_ARRAY */
+       721, /* GL_LUMINANCE4 */
+       727, /* GL_LUMINANCE8 */
+       711, /* GL_LUMINANCE12 */
+       717, /* GL_LUMINANCE16 */
+       722, /* GL_LUMINANCE4_ALPHA4 */
+       725, /* GL_LUMINANCE6_ALPHA2 */
+       728, /* GL_LUMINANCE8_ALPHA8 */
+       714, /* GL_LUMINANCE12_ALPHA4 */
+       712, /* GL_LUMINANCE12_ALPHA12 */
+       718, /* GL_LUMINANCE16_ALPHA16 */
+       634, /* GL_INTENSITY */
+       639, /* GL_INTENSITY4 */
+       641, /* GL_INTENSITY8 */
+       635, /* GL_INTENSITY12 */
+       637, /* GL_INTENSITY16 */
+      1318, /* GL_RGB2_EXT */
+      1319, /* GL_RGB4 */
+      1322, /* GL_RGB5 */
+      1326, /* GL_RGB8 */
+      1310, /* GL_RGB10 */
+      1314, /* GL_RGB12 */
+      1316, /* GL_RGB16 */
+      1333, /* GL_RGBA2 */
+      1335, /* GL_RGBA4 */
+      1323, /* GL_RGB5_A1 */
+      1339, /* GL_RGBA8 */
+      1311, /* GL_RGB10_A2 */
+      1329, /* GL_RGBA12 */
+      1331, /* GL_RGBA16 */
+      1674, /* GL_TEXTURE_RED_SIZE */
+      1646, /* GL_TEXTURE_GREEN_SIZE */
+      1585, /* GL_TEXTURE_BLUE_SIZE */
+      1572, /* GL_TEXTURE_ALPHA_SIZE */
+      1659, /* GL_TEXTURE_LUMINANCE_SIZE */
+      1650, /* GL_TEXTURE_INTENSITY_SIZE */
+      1304, /* GL_REPLACE_EXT */
+      1240, /* GL_PROXY_TEXTURE_1D */
+      1243, /* GL_PROXY_TEXTURE_2D */
+      1679, /* GL_TEXTURE_TOO_LARGE_EXT */
+      1671, /* GL_TEXTURE_PRIORITY */
+      1676, /* GL_TEXTURE_RESIDENT */
+      1575, /* GL_TEXTURE_BINDING_1D */
+      1577, /* GL_TEXTURE_BINDING_2D */
+      1579, /* GL_TEXTURE_BINDING_3D */
+      1071, /* GL_PACK_SKIP_IMAGES */
+      1067, /* GL_PACK_IMAGE_HEIGHT */
+      1708, /* GL_UNPACK_SKIP_IMAGES */
+      1705, /* GL_UNPACK_IMAGE_HEIGHT */
+      1571, /* GL_TEXTURE_3D */
+      1246, /* GL_PROXY_TEXTURE_3D */
+      1633, /* GL_TEXTURE_DEPTH */
+      1682, /* GL_TEXTURE_WRAP_R */
+       836, /* GL_MAX_3D_TEXTURE_SIZE */
+      1740, /* GL_VERTEX_ARRAY */
+      1002, /* GL_NORMAL_ARRAY */
        138, /* GL_COLOR_ARRAY */
-       576, /* GL_INDEX_ARRAY */
-      1553, /* GL_TEXTURE_COORD_ARRAY */
-       414, /* GL_EDGE_FLAG_ARRAY */
-      1683, /* GL_VERTEX_ARRAY_SIZE */
-      1685, /* GL_VERTEX_ARRAY_TYPE */
-      1684, /* GL_VERTEX_ARRAY_STRIDE */
-       962, /* GL_NORMAL_ARRAY_TYPE */
-       961, /* GL_NORMAL_ARRAY_STRIDE */
+       619, /* GL_INDEX_ARRAY */
+      1612, /* GL_TEXTURE_COORD_ARRAY */
+       435, /* GL_EDGE_FLAG_ARRAY */
+      1745, /* GL_VERTEX_ARRAY_SIZE */
+      1747, /* GL_VERTEX_ARRAY_TYPE */
+      1746, /* GL_VERTEX_ARRAY_STRIDE */
+      1007, /* GL_NORMAL_ARRAY_TYPE */
+      1006, /* GL_NORMAL_ARRAY_STRIDE */
        142, /* GL_COLOR_ARRAY_SIZE */
        144, /* GL_COLOR_ARRAY_TYPE */
        143, /* GL_COLOR_ARRAY_STRIDE */
-       581, /* GL_INDEX_ARRAY_TYPE */
-       580, /* GL_INDEX_ARRAY_STRIDE */
-      1557, /* GL_TEXTURE_COORD_ARRAY_SIZE */
-      1559, /* GL_TEXTURE_COORD_ARRAY_TYPE */
-      1558, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
-       418, /* GL_EDGE_FLAG_ARRAY_STRIDE */
-      1682, /* GL_VERTEX_ARRAY_POINTER */
-       960, /* GL_NORMAL_ARRAY_POINTER */
+       624, /* GL_INDEX_ARRAY_TYPE */
+       623, /* GL_INDEX_ARRAY_STRIDE */
+      1616, /* GL_TEXTURE_COORD_ARRAY_SIZE */
+      1618, /* GL_TEXTURE_COORD_ARRAY_TYPE */
+      1617, /* GL_TEXTURE_COORD_ARRAY_STRIDE */
+       439, /* GL_EDGE_FLAG_ARRAY_STRIDE */
+      1744, /* GL_VERTEX_ARRAY_POINTER */
+      1005, /* GL_NORMAL_ARRAY_POINTER */
        141, /* GL_COLOR_ARRAY_POINTER */
-       579, /* GL_INDEX_ARRAY_POINTER */
-      1556, /* GL_TEXTURE_COORD_ARRAY_POINTER */
-       417, /* GL_EDGE_FLAG_ARRAY_POINTER */
-       936, /* GL_MULTISAMPLE */
-      1304, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
-      1306, /* GL_SAMPLE_ALPHA_TO_ONE */
-      1311, /* GL_SAMPLE_COVERAGE */
-      1308, /* GL_SAMPLE_BUFFERS */
-      1299, /* GL_SAMPLES */
-      1315, /* GL_SAMPLE_COVERAGE_VALUE */
-      1313, /* GL_SAMPLE_COVERAGE_INVERT */
-       169, /* GL_COLOR_MATRIX */
-       171, /* GL_COLOR_MATRIX_STACK_DEPTH */
-       800, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
-      1127, /* GL_POST_COLOR_MATRIX_RED_SCALE */
-      1123, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
-      1118, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
-      1114, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
-      1125, /* GL_POST_COLOR_MATRIX_RED_BIAS */
-      1121, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
-      1116, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
-      1112, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
-      1536, /* GL_TEXTURE_COLOR_TABLE_SGI */
-      1202, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
-      1538, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
+       622, /* GL_INDEX_ARRAY_POINTER */
+      1615, /* GL_TEXTURE_COORD_ARRAY_POINTER */
+       438, /* GL_EDGE_FLAG_ARRAY_POINTER */
+       981, /* GL_MULTISAMPLE */
+      1361, /* GL_SAMPLE_ALPHA_TO_COVERAGE */
+      1363, /* GL_SAMPLE_ALPHA_TO_ONE */
+      1368, /* GL_SAMPLE_COVERAGE */
+      1365, /* GL_SAMPLE_BUFFERS */
+      1356, /* GL_SAMPLES */
+      1372, /* GL_SAMPLE_COVERAGE_VALUE */
+      1370, /* GL_SAMPLE_COVERAGE_INVERT */
+       185, /* GL_COLOR_MATRIX */
+       187, /* GL_COLOR_MATRIX_STACK_DEPTH */
+       844, /* GL_MAX_COLOR_MATRIX_STACK_DEPTH */
+      1172, /* GL_POST_COLOR_MATRIX_RED_SCALE */
+      1168, /* GL_POST_COLOR_MATRIX_GREEN_SCALE */
+      1163, /* GL_POST_COLOR_MATRIX_BLUE_SCALE */
+      1159, /* GL_POST_COLOR_MATRIX_ALPHA_SCALE */
+      1170, /* GL_POST_COLOR_MATRIX_RED_BIAS */
+      1166, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */
+      1161, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */
+      1157, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */
+      1595, /* GL_TEXTURE_COLOR_TABLE_SGI */
+      1247, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */
+      1597, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
         79, /* GL_BLEND_DST_RGB */
         87, /* GL_BLEND_SRC_RGB */
         78, /* GL_BLEND_DST_ALPHA */
         86, /* GL_BLEND_SRC_ALPHA */
-       175, /* GL_COLOR_TABLE */
-      1137, /* GL_POST_CONVOLUTION_COLOR_TABLE */
-      1120, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
-      1190, /* GL_PROXY_COLOR_TABLE */
-      1194, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
-      1193, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
-       199, /* GL_COLOR_TABLE_SCALE */
-       179, /* GL_COLOR_TABLE_BIAS */
-       184, /* GL_COLOR_TABLE_FORMAT */
-       201, /* GL_COLOR_TABLE_WIDTH */
-       196, /* GL_COLOR_TABLE_RED_SIZE */
-       187, /* GL_COLOR_TABLE_GREEN_SIZE */
-       181, /* GL_COLOR_TABLE_BLUE_SIZE */
-       176, /* GL_COLOR_TABLE_ALPHA_SIZE */
-       193, /* GL_COLOR_TABLE_LUMINANCE_SIZE */
-       190, /* GL_COLOR_TABLE_INTENSITY_SIZE */
+       191, /* GL_COLOR_TABLE */
+      1182, /* GL_POST_CONVOLUTION_COLOR_TABLE */
+      1165, /* GL_POST_COLOR_MATRIX_COLOR_TABLE */
+      1235, /* GL_PROXY_COLOR_TABLE */
+      1239, /* GL_PROXY_POST_CONVOLUTION_COLOR_TABLE */
+      1238, /* GL_PROXY_POST_COLOR_MATRIX_COLOR_TABLE */
+       215, /* GL_COLOR_TABLE_SCALE */
+       195, /* GL_COLOR_TABLE_BIAS */
+       200, /* GL_COLOR_TABLE_FORMAT */
+       217, /* GL_COLOR_TABLE_WIDTH */
+       212, /* GL_COLOR_TABLE_RED_SIZE */
+       203, /* GL_COLOR_TABLE_GREEN_SIZE */
+       197, /* GL_COLOR_TABLE_BLUE_SIZE */
+       192, /* GL_COLOR_TABLE_ALPHA_SIZE */
+       209, /* GL_COLOR_TABLE_LUMINANCE_SIZE */
+       206, /* GL_COLOR_TABLE_INTENSITY_SIZE */
         70, /* GL_BGR */
         71, /* GL_BGRA */
-       814, /* GL_MAX_ELEMENTS_VERTICES */
-       813, /* GL_MAX_ELEMENTS_INDICES */
-      1590, /* GL_TEXTURE_INDEX_SIZE_EXT */
+       858, /* GL_MAX_ELEMENTS_VERTICES */
+       857, /* GL_MAX_ELEMENTS_INDICES */
+      1649, /* GL_TEXTURE_INDEX_SIZE_EXT */
        135, /* GL_CLIP_VOLUME_CLIPPING_HINT_EXT */
-      1084, /* GL_POINT_SIZE_MIN */
-      1080, /* GL_POINT_SIZE_MAX */
-      1074, /* GL_POINT_FADE_THRESHOLD_SIZE */
-      1070, /* GL_POINT_DISTANCE_ATTENUATION */
+      1129, /* GL_POINT_SIZE_MIN */
+      1125, /* GL_POINT_SIZE_MAX */
+      1119, /* GL_POINT_FADE_THRESHOLD_SIZE */
+      1115, /* GL_POINT_DISTANCE_ATTENUATION */
        117, /* GL_CLAMP_TO_BORDER */
        120, /* GL_CLAMP_TO_EDGE */
-      1611, /* GL_TEXTURE_MIN_LOD */
-      1609, /* GL_TEXTURE_MAX_LOD */
-      1515, /* GL_TEXTURE_BASE_LEVEL */
-      1608, /* GL_TEXTURE_MAX_LEVEL */
-       570, /* GL_IGNORE_BORDER_HP */
-       245, /* GL_CONSTANT_BORDER_HP */
-      1248, /* GL_REPLICATE_BORDER_HP */
-       251, /* GL_CONVOLUTION_BORDER_COLOR */
-       985, /* GL_OCCLUSION_TEST_HP */
-       986, /* GL_OCCLUSION_TEST_RESULT_HP */
-       639, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
-      1530, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
-      1532, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
-      1534, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
-      1535, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-      1533, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
-      1531, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
-       796, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
-       797, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
-      1147, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
-      1149, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
-      1146, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
-      1148, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
-      1598, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
-      1599, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
-      1597, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
-       516, /* GL_GENERATE_MIPMAP */
-       517, /* GL_GENERATE_MIPMAP_HINT */
-       479, /* GL_FOG_OFFSET_SGIX */
-       480, /* GL_FOG_OFFSET_VALUE_SGIX */
-      1544, /* GL_TEXTURE_COMPARE_SGIX */
-      1543, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
-      1594, /* GL_TEXTURE_LEQUAL_R_SGIX */
-      1586, /* GL_TEXTURE_GEQUAL_R_SGIX */
-       322, /* GL_DEPTH_COMPONENT16 */
-       325, /* GL_DEPTH_COMPONENT24 */
-       328, /* GL_DEPTH_COMPONENT32 */
-       273, /* GL_CULL_VERTEX_EXT */
-       275, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
-       274, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
-      1739, /* GL_WRAP_BORDER_SUN */
-      1537, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
-       632, /* GL_LIGHT_MODEL_COLOR_CONTROL */
-      1343, /* GL_SINGLE_COLOR */
-      1331, /* GL_SEPARATE_SPECULAR_COLOR */
-      1340, /* GL_SHARED_TEXTURE_PALETTE_EXT */
-      1653, /* GL_UNSIGNED_BYTE_2_3_3_REV */
-      1666, /* GL_UNSIGNED_SHORT_5_6_5 */
-      1667, /* GL_UNSIGNED_SHORT_5_6_5_REV */
-      1664, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
-      1662, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
-      1660, /* GL_UNSIGNED_INT_8_8_8_8_REV */
-      1658, /* GL_UNSIGNED_INT_2_10_10_10_REV */
-      1606, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
-      1607, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
-      1605, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
-       887, /* GL_MIRRORED_REPEAT */
-      1287, /* GL_RGB_S3TC */
-      1264, /* GL_RGB4_S3TC */
-      1286, /* GL_RGBA_S3TC */
-      1281, /* GL_RGBA4_S3TC */
-      1284, /* GL_RGBA_DXT5_S3TC */
-      1279, /* GL_RGBA4_DXT5_S3TC */
-       238, /* GL_COMPRESSED_RGB_S3TC_DXT1_EXT */
-       233, /* GL_COMPRESSED_RGBA_S3TC_DXT1_EXT */
-       234, /* GL_COMPRESSED_RGBA_S3TC_DXT3_EXT */
-       235, /* GL_COMPRESSED_RGBA_S3TC_DXT5_EXT */
-       948, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
-       947, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
-       640, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
-       466, /* GL_FOG_COORDINATE_SOURCE */
-       458, /* GL_FOG_COORD */
-       482, /* GL_FRAGMENT_DEPTH */
-       279, /* GL_CURRENT_FOG_COORD */
-       465, /* GL_FOG_COORDINATE_ARRAY_TYPE */
-       464, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
-       463, /* GL_FOG_COORDINATE_ARRAY_POINTER */
-       460, /* GL_FOG_COORDINATE_ARRAY */
-       173, /* GL_COLOR_SUM */
-       298, /* GL_CURRENT_SECONDARY_COLOR */
-      1324, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
-      1326, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
-      1325, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
-      1323, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
-      1320, /* GL_SECONDARY_COLOR_ARRAY */
-       526, /* GL_GL_CURRENT_RASTER_SECONDARY_COLOR */
+      1670, /* GL_TEXTURE_MIN_LOD */
+      1668, /* GL_TEXTURE_MAX_LOD */
+      1574, /* GL_TEXTURE_BASE_LEVEL */
+      1667, /* GL_TEXTURE_MAX_LEVEL */
+       612, /* GL_IGNORE_BORDER_HP */
+       261, /* GL_CONSTANT_BORDER_HP */
+      1305, /* GL_REPLICATE_BORDER_HP */
+       267, /* GL_CONVOLUTION_BORDER_COLOR */
+      1030, /* GL_OCCLUSION_TEST_HP */
+      1031, /* GL_OCCLUSION_TEST_RESULT_HP */
+       683, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */
+      1589, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */
+      1591, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */
+      1593, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */
+      1594, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+      1592, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */
+      1590, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */
+       840, /* GL_MAX_CLIPMAP_DEPTH_SGIX */
+       841, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */
+      1192, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */
+      1194, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */
+      1191, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */
+      1193, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */
+      1657, /* GL_TEXTURE_LOD_BIAS_S_SGIX */
+      1658, /* GL_TEXTURE_LOD_BIAS_T_SGIX */
+      1656, /* GL_TEXTURE_LOD_BIAS_R_SGIX */
+       558, /* GL_GENERATE_MIPMAP */
+       559, /* GL_GENERATE_MIPMAP_HINT */
+       500, /* GL_FOG_OFFSET_SGIX */
+       501, /* GL_FOG_OFFSET_VALUE_SGIX */
+      1603, /* GL_TEXTURE_COMPARE_SGIX */
+      1602, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */
+      1653, /* GL_TEXTURE_LEQUAL_R_SGIX */
+      1645, /* GL_TEXTURE_GEQUAL_R_SGIX */
+       340, /* GL_DEPTH_COMPONENT16 */
+       343, /* GL_DEPTH_COMPONENT24 */
+       346, /* GL_DEPTH_COMPONENT32 */
+       289, /* GL_CULL_VERTEX_EXT */
+       291, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */
+       290, /* GL_CULL_VERTEX_EYE_POSITION_EXT */
+      1801, /* GL_WRAP_BORDER_SUN */
+      1596, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */
+       676, /* GL_LIGHT_MODEL_COLOR_CONTROL */
+      1400, /* GL_SINGLE_COLOR */
+      1388, /* GL_SEPARATE_SPECULAR_COLOR */
+      1397, /* GL_SHARED_TEXTURE_PALETTE_EXT */
+       511, /* GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING */
+       512, /* GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE */
+       519, /* GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE */
+       514, /* GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE */
+       510, /* GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE */
+       509, /* GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE */
+       513, /* GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE */
+       520, /* GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE */
+       531, /* GL_FRAMEBUFFER_DEFAULT */
+       544, /* GL_FRAMEBUFFER_UNDEFINED */
+       353, /* GL_DEPTH_STENCIL_ATTACHMENT */
+       618, /* GL_INDEX */
+      1713, /* GL_UNSIGNED_BYTE_2_3_3_REV */
+      1728, /* GL_UNSIGNED_SHORT_5_6_5 */
+      1729, /* GL_UNSIGNED_SHORT_5_6_5_REV */
+      1726, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */
+      1724, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */
+      1721, /* GL_UNSIGNED_INT_8_8_8_8_REV */
+      1719, /* GL_UNSIGNED_INT_2_10_10_10_REV */
+      1665, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */
+      1666, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */
+      1664, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */
+       932, /* GL_MIRRORED_REPEAT */
+      1344, /* GL_RGB_S3TC */
+      1321, /* GL_RGB4_S3TC */
+      1343, /* GL_RGBA_S3TC */
+      1338, /* GL_RGBA4_S3TC */
+      1341, /* GL_RGBA_DXT5_S3TC */
+      1336, /* GL_RGBA4_DXT5_S3TC */
+       254, /* GL_COMPRESSED_RGB_S3TC_DXT1_EXT */
+       249, /* GL_COMPRESSED_RGBA_S3TC_DXT1_EXT */
+       250, /* GL_COMPRESSED_RGBA_S3TC_DXT3_EXT */
+       251, /* GL_COMPRESSED_RGBA_S3TC_DXT5_EXT */
+       993, /* GL_NEAREST_CLIPMAP_NEAREST_SGIX */
+       992, /* GL_NEAREST_CLIPMAP_LINEAR_SGIX */
+       684, /* GL_LINEAR_CLIPMAP_NEAREST_SGIX */
+       487, /* GL_FOG_COORDINATE_SOURCE */
+       479, /* GL_FOG_COORD */
+       503, /* GL_FRAGMENT_DEPTH */
+       295, /* GL_CURRENT_FOG_COORD */
+       486, /* GL_FOG_COORDINATE_ARRAY_TYPE */
+       485, /* GL_FOG_COORDINATE_ARRAY_STRIDE */
+       484, /* GL_FOG_COORDINATE_ARRAY_POINTER */
+       481, /* GL_FOG_COORDINATE_ARRAY */
+       189, /* GL_COLOR_SUM */
+       314, /* GL_CURRENT_SECONDARY_COLOR */
+      1381, /* GL_SECONDARY_COLOR_ARRAY_SIZE */
+      1383, /* GL_SECONDARY_COLOR_ARRAY_TYPE */
+      1382, /* GL_SECONDARY_COLOR_ARRAY_STRIDE */
+      1380, /* GL_SECONDARY_COLOR_ARRAY_POINTER */
+      1377, /* GL_SECONDARY_COLOR_ARRAY */
+       568, /* GL_GL_CURRENT_RASTER_SECONDARY_COLOR */
         28, /* GL_ALIASED_POINT_SIZE_RANGE */
         27, /* GL_ALIASED_LINE_WIDTH_RANGE */
-      1444, /* GL_TEXTURE0 */
-      1446, /* GL_TEXTURE1 */
-      1468, /* GL_TEXTURE2 */
-      1490, /* GL_TEXTURE3 */
-      1496, /* GL_TEXTURE4 */
-      1498, /* GL_TEXTURE5 */
-      1500, /* GL_TEXTURE6 */
-      1502, /* GL_TEXTURE7 */
-      1504, /* GL_TEXTURE8 */
-      1506, /* GL_TEXTURE9 */
-      1447, /* GL_TEXTURE10 */
-      1449, /* GL_TEXTURE11 */
-      1451, /* GL_TEXTURE12 */
-      1453, /* GL_TEXTURE13 */
-      1455, /* GL_TEXTURE14 */
-      1457, /* GL_TEXTURE15 */
-      1459, /* GL_TEXTURE16 */
-      1461, /* GL_TEXTURE17 */
-      1463, /* GL_TEXTURE18 */
-      1465, /* GL_TEXTURE19 */
-      1469, /* GL_TEXTURE20 */
-      1471, /* GL_TEXTURE21 */
-      1473, /* GL_TEXTURE22 */
-      1475, /* GL_TEXTURE23 */
-      1477, /* GL_TEXTURE24 */
-      1479, /* GL_TEXTURE25 */
-      1481, /* GL_TEXTURE26 */
-      1483, /* GL_TEXTURE27 */
-      1485, /* GL_TEXTURE28 */
-      1487, /* GL_TEXTURE29 */
-      1491, /* GL_TEXTURE30 */
-      1493, /* GL_TEXTURE31 */
+      1503, /* GL_TEXTURE0 */
+      1505, /* GL_TEXTURE1 */
+      1527, /* GL_TEXTURE2 */
+      1549, /* GL_TEXTURE3 */
+      1555, /* GL_TEXTURE4 */
+      1557, /* GL_TEXTURE5 */
+      1559, /* GL_TEXTURE6 */
+      1561, /* GL_TEXTURE7 */
+      1563, /* GL_TEXTURE8 */
+      1565, /* GL_TEXTURE9 */
+      1506, /* GL_TEXTURE10 */
+      1508, /* GL_TEXTURE11 */
+      1510, /* GL_TEXTURE12 */
+      1512, /* GL_TEXTURE13 */
+      1514, /* GL_TEXTURE14 */
+      1516, /* GL_TEXTURE15 */
+      1518, /* GL_TEXTURE16 */
+      1520, /* GL_TEXTURE17 */
+      1522, /* GL_TEXTURE18 */
+      1524, /* GL_TEXTURE19 */
+      1528, /* GL_TEXTURE20 */
+      1530, /* GL_TEXTURE21 */
+      1532, /* GL_TEXTURE22 */
+      1534, /* GL_TEXTURE23 */
+      1536, /* GL_TEXTURE24 */
+      1538, /* GL_TEXTURE25 */
+      1540, /* GL_TEXTURE26 */
+      1542, /* GL_TEXTURE27 */
+      1544, /* GL_TEXTURE28 */
+      1546, /* GL_TEXTURE29 */
+      1550, /* GL_TEXTURE30 */
+      1552, /* GL_TEXTURE31 */
         18, /* GL_ACTIVE_TEXTURE */
        123, /* GL_CLIENT_ACTIVE_TEXTURE */
-       865, /* GL_MAX_TEXTURE_UNITS */
-      1632, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
-      1635, /* GL_TRANSPOSE_PROJECTION_MATRIX */
-      1637, /* GL_TRANSPOSE_TEXTURE_MATRIX */
-      1629, /* GL_TRANSPOSE_COLOR_MATRIX */
-      1432, /* GL_SUBTRACT */
-       854, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
-       221, /* GL_COMPRESSED_ALPHA */
-       225, /* GL_COMPRESSED_LUMINANCE */
-       226, /* GL_COMPRESSED_LUMINANCE_ALPHA */
-       223, /* GL_COMPRESSED_INTENSITY */
-       229, /* GL_COMPRESSED_RGB */
-       230, /* GL_COMPRESSED_RGBA */
-      1551, /* GL_TEXTURE_COMPRESSION_HINT */
-      1613, /* GL_TEXTURE_RECTANGLE_ARB */
-      1523, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
-      1205, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
-       852, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
-       334, /* GL_DEPTH_STENCIL_NV */
-      1657, /* GL_UNSIGNED_INT_24_8_NV */
-       861, /* GL_MAX_TEXTURE_LOD_BIAS */
-      1604, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
-       862, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
-      1580, /* GL_TEXTURE_FILTER_CONTROL */
-      1595, /* GL_TEXTURE_LOD_BIAS */
-       206, /* GL_COMBINE4 */
-       855, /* GL_MAX_SHININESS_NV */
-       856, /* GL_MAX_SPOT_EXPONENT_NV */
-       574, /* GL_INCR_WRAP */
-       309, /* GL_DECR_WRAP */
-       907, /* GL_MODELVIEW1_ARB */
-       963, /* GL_NORMAL_MAP */
-      1234, /* GL_REFLECTION_MAP */
-      1560, /* GL_TEXTURE_CUBE_MAP */
-      1521, /* GL_TEXTURE_BINDING_CUBE_MAP */
-      1568, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
-      1562, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
-      1570, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
-      1564, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
-      1572, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
-      1566, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
-      1203, /* GL_PROXY_TEXTURE_CUBE_MAP */
-       808, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
-       942, /* GL_MULTISAMPLE_FILTER_HINT_NV */
-       474, /* GL_FOG_DISTANCE_MODE_NV */
-       433, /* GL_EYE_RADIAL_NV */
-       432, /* GL_EYE_PLANE_ABSOLUTE_NV */
-       205, /* GL_COMBINE */
-       212, /* GL_COMBINE_RGB */
-       207, /* GL_COMBINE_ALPHA */
-      1288, /* GL_RGB_SCALE */
+       910, /* GL_MAX_TEXTURE_UNITS */
+      1692, /* GL_TRANSPOSE_MODELVIEW_MATRIX */
+      1695, /* GL_TRANSPOSE_PROJECTION_MATRIX */
+      1697, /* GL_TRANSPOSE_TEXTURE_MATRIX */
+      1689, /* GL_TRANSPOSE_COLOR_MATRIX */
+      1491, /* GL_SUBTRACT */
+       898, /* GL_MAX_RENDERBUFFER_SIZE_EXT */
+       237, /* GL_COMPRESSED_ALPHA */
+       241, /* GL_COMPRESSED_LUMINANCE */
+       242, /* GL_COMPRESSED_LUMINANCE_ALPHA */
+       239, /* GL_COMPRESSED_INTENSITY */
+       245, /* GL_COMPRESSED_RGB */
+       246, /* GL_COMPRESSED_RGBA */
+      1610, /* GL_TEXTURE_COMPRESSION_HINT */
+      1672, /* GL_TEXTURE_RECTANGLE_ARB */
+      1582, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */
+      1250, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */
+       896, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */
+       352, /* GL_DEPTH_STENCIL */
+      1717, /* GL_UNSIGNED_INT_24_8 */
+       906, /* GL_MAX_TEXTURE_LOD_BIAS */
+      1663, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */
+       907, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */
+      1639, /* GL_TEXTURE_FILTER_CONTROL */
+      1654, /* GL_TEXTURE_LOD_BIAS */
+       222, /* GL_COMBINE4 */
+       900, /* GL_MAX_SHININESS_NV */
+       901, /* GL_MAX_SPOT_EXPONENT_NV */
+       616, /* GL_INCR_WRAP */
+       325, /* GL_DECR_WRAP */
+       952, /* GL_MODELVIEW1_ARB */
+      1008, /* GL_NORMAL_MAP */
+      1280, /* GL_REFLECTION_MAP */
+      1619, /* GL_TEXTURE_CUBE_MAP */
+      1580, /* GL_TEXTURE_BINDING_CUBE_MAP */
+      1627, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */
+      1621, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */
+      1629, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */
+      1623, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */
+      1631, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */
+      1625, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */
+      1248, /* GL_PROXY_TEXTURE_CUBE_MAP */
+       852, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */
+       987, /* GL_MULTISAMPLE_FILTER_HINT_NV */
+       495, /* GL_FOG_DISTANCE_MODE_NV */
+       454, /* GL_EYE_RADIAL_NV */
+       453, /* GL_EYE_PLANE_ABSOLUTE_NV */
+       221, /* GL_COMBINE */
+       228, /* GL_COMBINE_RGB */
+       223, /* GL_COMBINE_ALPHA */
+      1345, /* GL_RGB_SCALE */
         24, /* GL_ADD_SIGNED */
-       601, /* GL_INTERPOLATE */
-       240, /* GL_CONSTANT */
-      1153, /* GL_PRIMARY_COLOR */
-      1150, /* GL_PREVIOUS */
-      1354, /* GL_SOURCE0_RGB */
-      1360, /* GL_SOURCE1_RGB */
-      1366, /* GL_SOURCE2_RGB */
-      1370, /* GL_SOURCE3_RGB_NV */
-      1351, /* GL_SOURCE0_ALPHA */
-      1357, /* GL_SOURCE1_ALPHA */
-      1363, /* GL_SOURCE2_ALPHA */
-      1369, /* GL_SOURCE3_ALPHA_NV */
-       999, /* GL_OPERAND0_RGB */
-      1005, /* GL_OPERAND1_RGB */
-      1011, /* GL_OPERAND2_RGB */
-      1015, /* GL_OPERAND3_RGB_NV */
-       996, /* GL_OPERAND0_ALPHA */
-      1002, /* GL_OPERAND1_ALPHA */
-      1008, /* GL_OPERAND2_ALPHA */
-      1014, /* GL_OPERAND3_ALPHA_NV */
-      1679, /* GL_VERTEX_ARRAY_BINDING_APPLE */
-      1743, /* GL_YCBCR_422_APPLE */
-      1668, /* GL_UNSIGNED_SHORT_8_8_APPLE */
-      1670, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
-      1345, /* GL_SLICE_ACCUM_SUN */
-      1210, /* GL_QUAD_MESH_SUN */
-      1641, /* GL_TRIANGLE_MESH_SUN */
-      1717, /* GL_VERTEX_PROGRAM_ARB */
-      1728, /* GL_VERTEX_STATE_PROGRAM_NV */
-      1704, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
-      1710, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
-      1712, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
-      1714, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
-       300, /* GL_CURRENT_VERTEX_ATTRIB */
-      1166, /* GL_PROGRAM_LENGTH_ARB */
-      1180, /* GL_PROGRAM_STRING_ARB */
-       929, /* GL_MODELVIEW_PROJECTION_NV */
-       569, /* GL_IDENTITY_NV */
-       614, /* GL_INVERSE_NV */
-      1634, /* GL_TRANSPOSE_NV */
-       615, /* GL_INVERSE_TRANSPOSE_NV */
-       838, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
-       837, /* GL_MAX_PROGRAM_MATRICES_ARB */
-       745, /* GL_MATRIX0_NV */
-       757, /* GL_MATRIX1_NV */
-       769, /* GL_MATRIX2_NV */
-       773, /* GL_MATRIX3_NV */
-       775, /* GL_MATRIX4_NV */
-       777, /* GL_MATRIX5_NV */
-       779, /* GL_MATRIX6_NV */
-       781, /* GL_MATRIX7_NV */
-       285, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
-       282, /* GL_CURRENT_MATRIX_ARB */
-      1720, /* GL_VERTEX_PROGRAM_POINT_SIZE */
-      1723, /* GL_VERTEX_PROGRAM_TWO_SIDE */
-      1178, /* GL_PROGRAM_PARAMETER_NV */
-      1708, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
-      1182, /* GL_PROGRAM_TARGET_NV */
-      1179, /* GL_PROGRAM_RESIDENT_NV */
-      1626, /* GL_TRACK_MATRIX_NV */
-      1627, /* GL_TRACK_MATRIX_TRANSFORM_NV */
-      1718, /* GL_VERTEX_PROGRAM_BINDING_NV */
-      1160, /* GL_PROGRAM_ERROR_POSITION_ARB */
-       319, /* GL_DEPTH_CLAMP_NV */
-      1686, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
-      1693, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
-      1694, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
-      1695, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
-      1696, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
-      1697, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
-      1698, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
-      1699, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
-      1700, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
-      1701, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
-      1687, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
-      1688, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
-      1689, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
-      1690, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
-      1691, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
-      1692, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
-       699, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
-       706, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
-       707, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
-       708, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
-       709, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
-       710, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
-       711, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
-       712, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
-       713, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
-       714, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
-       700, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
-       701, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
-       702, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
-       703, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
-       704, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
-       705, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
-       726, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
-       733, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
-       734, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
-       735, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
-       736, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
-       737, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
-       738, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
-      1159, /* GL_PROGRAM_BINDING_ARB */
-       740, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
-       741, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
-       727, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
-       728, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
-       729, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
-       730, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
-       731, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
-       732, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
-      1549, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
-      1546, /* GL_TEXTURE_COMPRESSED */
-       968, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
-       239, /* GL_COMPRESSED_TEXTURE_FORMATS */
-       877, /* GL_MAX_VERTEX_UNITS_ARB */
+       644, /* GL_INTERPOLATE */
+       256, /* GL_CONSTANT */
+      1198, /* GL_PRIMARY_COLOR */
+      1195, /* GL_PREVIOUS */
+      1411, /* GL_SOURCE0_RGB */
+      1417, /* GL_SOURCE1_RGB */
+      1423, /* GL_SOURCE2_RGB */
+      1427, /* GL_SOURCE3_RGB_NV */
+      1408, /* GL_SOURCE0_ALPHA */
+      1414, /* GL_SOURCE1_ALPHA */
+      1420, /* GL_SOURCE2_ALPHA */
+      1426, /* GL_SOURCE3_ALPHA_NV */
+      1044, /* GL_OPERAND0_RGB */
+      1050, /* GL_OPERAND1_RGB */
+      1056, /* GL_OPERAND2_RGB */
+      1060, /* GL_OPERAND3_RGB_NV */
+      1041, /* GL_OPERAND0_ALPHA */
+      1047, /* GL_OPERAND1_ALPHA */
+      1053, /* GL_OPERAND2_ALPHA */
+      1059, /* GL_OPERAND3_ALPHA_NV */
+      1741, /* GL_VERTEX_ARRAY_BINDING_APPLE */
+      1805, /* GL_YCBCR_422_APPLE */
+      1730, /* GL_UNSIGNED_SHORT_8_8_APPLE */
+      1732, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */
+      1402, /* GL_SLICE_ACCUM_SUN */
+      1255, /* GL_QUAD_MESH_SUN */
+      1701, /* GL_TRIANGLE_MESH_SUN */
+      1779, /* GL_VERTEX_PROGRAM_ARB */
+      1790, /* GL_VERTEX_STATE_PROGRAM_NV */
+      1766, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */
+      1772, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */
+      1774, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */
+      1776, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */
+       316, /* GL_CURRENT_VERTEX_ATTRIB */
+      1211, /* GL_PROGRAM_LENGTH_ARB */
+      1225, /* GL_PROGRAM_STRING_ARB */
+       974, /* GL_MODELVIEW_PROJECTION_NV */
+       611, /* GL_IDENTITY_NV */
+       658, /* GL_INVERSE_NV */
+      1694, /* GL_TRANSPOSE_NV */
+       659, /* GL_INVERSE_TRANSPOSE_NV */
+       882, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */
+       881, /* GL_MAX_PROGRAM_MATRICES_ARB */
+       789, /* GL_MATRIX0_NV */
+       801, /* GL_MATRIX1_NV */
+       813, /* GL_MATRIX2_NV */
+       817, /* GL_MATRIX3_NV */
+       819, /* GL_MATRIX4_NV */
+       821, /* GL_MATRIX5_NV */
+       823, /* GL_MATRIX6_NV */
+       825, /* GL_MATRIX7_NV */
+       301, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */
+       298, /* GL_CURRENT_MATRIX_ARB */
+      1782, /* GL_VERTEX_PROGRAM_POINT_SIZE */
+      1785, /* GL_VERTEX_PROGRAM_TWO_SIDE */
+      1223, /* GL_PROGRAM_PARAMETER_NV */
+      1770, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */
+      1227, /* GL_PROGRAM_TARGET_NV */
+      1224, /* GL_PROGRAM_RESIDENT_NV */
+      1686, /* GL_TRACK_MATRIX_NV */
+      1687, /* GL_TRACK_MATRIX_TRANSFORM_NV */
+      1780, /* GL_VERTEX_PROGRAM_BINDING_NV */
+      1205, /* GL_PROGRAM_ERROR_POSITION_ARB */
+       337, /* GL_DEPTH_CLAMP_NV */
+      1748, /* GL_VERTEX_ATTRIB_ARRAY0_NV */
+      1755, /* GL_VERTEX_ATTRIB_ARRAY1_NV */
+      1756, /* GL_VERTEX_ATTRIB_ARRAY2_NV */
+      1757, /* GL_VERTEX_ATTRIB_ARRAY3_NV */
+      1758, /* GL_VERTEX_ATTRIB_ARRAY4_NV */
+      1759, /* GL_VERTEX_ATTRIB_ARRAY5_NV */
+      1760, /* GL_VERTEX_ATTRIB_ARRAY6_NV */
+      1761, /* GL_VERTEX_ATTRIB_ARRAY7_NV */
+      1762, /* GL_VERTEX_ATTRIB_ARRAY8_NV */
+      1763, /* GL_VERTEX_ATTRIB_ARRAY9_NV */
+      1749, /* GL_VERTEX_ATTRIB_ARRAY10_NV */
+      1750, /* GL_VERTEX_ATTRIB_ARRAY11_NV */
+      1751, /* GL_VERTEX_ATTRIB_ARRAY12_NV */
+      1752, /* GL_VERTEX_ATTRIB_ARRAY13_NV */
+      1753, /* GL_VERTEX_ATTRIB_ARRAY14_NV */
+      1754, /* GL_VERTEX_ATTRIB_ARRAY15_NV */
+       743, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */
+       750, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */
+       751, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */
+       752, /* GL_MAP1_VERTEX_ATTRIB3_4_NV */
+       753, /* GL_MAP1_VERTEX_ATTRIB4_4_NV */
+       754, /* GL_MAP1_VERTEX_ATTRIB5_4_NV */
+       755, /* GL_MAP1_VERTEX_ATTRIB6_4_NV */
+       756, /* GL_MAP1_VERTEX_ATTRIB7_4_NV */
+       757, /* GL_MAP1_VERTEX_ATTRIB8_4_NV */
+       758, /* GL_MAP1_VERTEX_ATTRIB9_4_NV */
+       744, /* GL_MAP1_VERTEX_ATTRIB10_4_NV */
+       745, /* GL_MAP1_VERTEX_ATTRIB11_4_NV */
+       746, /* GL_MAP1_VERTEX_ATTRIB12_4_NV */
+       747, /* GL_MAP1_VERTEX_ATTRIB13_4_NV */
+       748, /* GL_MAP1_VERTEX_ATTRIB14_4_NV */
+       749, /* GL_MAP1_VERTEX_ATTRIB15_4_NV */
+       770, /* GL_MAP2_VERTEX_ATTRIB0_4_NV */
+       777, /* GL_MAP2_VERTEX_ATTRIB1_4_NV */
+       778, /* GL_MAP2_VERTEX_ATTRIB2_4_NV */
+       779, /* GL_MAP2_VERTEX_ATTRIB3_4_NV */
+       780, /* GL_MAP2_VERTEX_ATTRIB4_4_NV */
+       781, /* GL_MAP2_VERTEX_ATTRIB5_4_NV */
+       782, /* GL_MAP2_VERTEX_ATTRIB6_4_NV */
+      1204, /* GL_PROGRAM_BINDING_ARB */
+       784, /* GL_MAP2_VERTEX_ATTRIB8_4_NV */
+       785, /* GL_MAP2_VERTEX_ATTRIB9_4_NV */
+       771, /* GL_MAP2_VERTEX_ATTRIB10_4_NV */
+       772, /* GL_MAP2_VERTEX_ATTRIB11_4_NV */
+       773, /* GL_MAP2_VERTEX_ATTRIB12_4_NV */
+       774, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */
+       775, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */
+       776, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */
+      1608, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */
+      1605, /* GL_TEXTURE_COMPRESSED */
+      1013, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */
+       255, /* GL_COMPRESSED_TEXTURE_FORMATS */
+       922, /* GL_MAX_VERTEX_UNITS_ARB */
         22, /* GL_ACTIVE_VERTEX_UNITS_ARB */
-      1738, /* GL_WEIGHT_SUM_UNITY_ARB */
-      1716, /* GL_VERTEX_BLEND_ARB */
-       302, /* GL_CURRENT_WEIGHT_ARB */
-      1737, /* GL_WEIGHT_ARRAY_TYPE_ARB */
-      1736, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
-      1735, /* GL_WEIGHT_ARRAY_SIZE_ARB */
-      1734, /* GL_WEIGHT_ARRAY_POINTER_ARB */
-      1731, /* GL_WEIGHT_ARRAY_ARB */
-       345, /* GL_DOT3_RGB */
-       346, /* GL_DOT3_RGBA */
-       237, /* GL_COMPRESSED_RGB_FXT1_3DFX */
-       232, /* GL_COMPRESSED_RGBA_FXT1_3DFX */
-       937, /* GL_MULTISAMPLE_3DFX */
-      1309, /* GL_SAMPLE_BUFFERS_3DFX */
-      1300, /* GL_SAMPLES_3DFX */
-       918, /* GL_MODELVIEW2_ARB */
-       921, /* GL_MODELVIEW3_ARB */
-       922, /* GL_MODELVIEW4_ARB */
-       923, /* GL_MODELVIEW5_ARB */
-       924, /* GL_MODELVIEW6_ARB */
-       925, /* GL_MODELVIEW7_ARB */
-       926, /* GL_MODELVIEW8_ARB */
-       927, /* GL_MODELVIEW9_ARB */
-       897, /* GL_MODELVIEW10_ARB */
-       898, /* GL_MODELVIEW11_ARB */
-       899, /* GL_MODELVIEW12_ARB */
-       900, /* GL_MODELVIEW13_ARB */
-       901, /* GL_MODELVIEW14_ARB */
-       902, /* GL_MODELVIEW15_ARB */
-       903, /* GL_MODELVIEW16_ARB */
-       904, /* GL_MODELVIEW17_ARB */
-       905, /* GL_MODELVIEW18_ARB */
-       906, /* GL_MODELVIEW19_ARB */
-       908, /* GL_MODELVIEW20_ARB */
-       909, /* GL_MODELVIEW21_ARB */
-       910, /* GL_MODELVIEW22_ARB */
-       911, /* GL_MODELVIEW23_ARB */
-       912, /* GL_MODELVIEW24_ARB */
-       913, /* GL_MODELVIEW25_ARB */
-       914, /* GL_MODELVIEW26_ARB */
-       915, /* GL_MODELVIEW27_ARB */
-       916, /* GL_MODELVIEW28_ARB */
-       917, /* GL_MODELVIEW29_ARB */
-       919, /* GL_MODELVIEW30_ARB */
-       920, /* GL_MODELVIEW31_ARB */
-       350, /* GL_DOT3_RGB_EXT */
-       348, /* GL_DOT3_RGBA_EXT */
-       891, /* GL_MIRROR_CLAMP_EXT */
-       894, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
-       932, /* GL_MODULATE_ADD_ATI */
-       933, /* GL_MODULATE_SIGNED_ADD_ATI */
-       934, /* GL_MODULATE_SUBTRACT_ATI */
-      1744, /* GL_YCBCR_MESA */
-      1023, /* GL_PACK_INVERT_MESA */
-       305, /* GL_DEBUG_OBJECT_MESA */
-       306, /* GL_DEBUG_PRINT_MESA */
-       304, /* GL_DEBUG_ASSERT_MESA */
+      1800, /* GL_WEIGHT_SUM_UNITY_ARB */
+      1778, /* GL_VERTEX_BLEND_ARB */
+       318, /* GL_CURRENT_WEIGHT_ARB */
+      1799, /* GL_WEIGHT_ARRAY_TYPE_ARB */
+      1798, /* GL_WEIGHT_ARRAY_STRIDE_ARB */
+      1797, /* GL_WEIGHT_ARRAY_SIZE_ARB */
+      1796, /* GL_WEIGHT_ARRAY_POINTER_ARB */
+      1793, /* GL_WEIGHT_ARRAY_ARB */
+       365, /* GL_DOT3_RGB */
+       366, /* GL_DOT3_RGBA */
+       253, /* GL_COMPRESSED_RGB_FXT1_3DFX */
+       248, /* GL_COMPRESSED_RGBA_FXT1_3DFX */
+       982, /* GL_MULTISAMPLE_3DFX */
+      1366, /* GL_SAMPLE_BUFFERS_3DFX */
+      1357, /* GL_SAMPLES_3DFX */
+       963, /* GL_MODELVIEW2_ARB */
+       966, /* GL_MODELVIEW3_ARB */
+       967, /* GL_MODELVIEW4_ARB */
+       968, /* GL_MODELVIEW5_ARB */
+       969, /* GL_MODELVIEW6_ARB */
+       970, /* GL_MODELVIEW7_ARB */
+       971, /* GL_MODELVIEW8_ARB */
+       972, /* GL_MODELVIEW9_ARB */
+       942, /* GL_MODELVIEW10_ARB */
+       943, /* GL_MODELVIEW11_ARB */
+       944, /* GL_MODELVIEW12_ARB */
+       945, /* GL_MODELVIEW13_ARB */
+       946, /* GL_MODELVIEW14_ARB */
+       947, /* GL_MODELVIEW15_ARB */
+       948, /* GL_MODELVIEW16_ARB */
+       949, /* GL_MODELVIEW17_ARB */
+       950, /* GL_MODELVIEW18_ARB */
+       951, /* GL_MODELVIEW19_ARB */
+       953, /* GL_MODELVIEW20_ARB */
+       954, /* GL_MODELVIEW21_ARB */
+       955, /* GL_MODELVIEW22_ARB */
+       956, /* GL_MODELVIEW23_ARB */
+       957, /* GL_MODELVIEW24_ARB */
+       958, /* GL_MODELVIEW25_ARB */
+       959, /* GL_MODELVIEW26_ARB */
+       960, /* GL_MODELVIEW27_ARB */
+       961, /* GL_MODELVIEW28_ARB */
+       962, /* GL_MODELVIEW29_ARB */
+       964, /* GL_MODELVIEW30_ARB */
+       965, /* GL_MODELVIEW31_ARB */
+       370, /* GL_DOT3_RGB_EXT */
+       368, /* GL_DOT3_RGBA_EXT */
+       936, /* GL_MIRROR_CLAMP_EXT */
+       939, /* GL_MIRROR_CLAMP_TO_EDGE_EXT */
+       977, /* GL_MODULATE_ADD_ATI */
+       978, /* GL_MODULATE_SIGNED_ADD_ATI */
+       979, /* GL_MODULATE_SUBTRACT_ATI */
+      1806, /* GL_YCBCR_MESA */
+      1068, /* GL_PACK_INVERT_MESA */
+       321, /* GL_DEBUG_OBJECT_MESA */
+       322, /* GL_DEBUG_PRINT_MESA */
+       320, /* GL_DEBUG_ASSERT_MESA */
        106, /* GL_BUFFER_SIZE */
        108, /* GL_BUFFER_USAGE */
-      1397, /* GL_STENCIL_BACK_FUNC */
-      1395, /* GL_STENCIL_BACK_FAIL */
-      1399, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
-      1401, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
-       483, /* GL_FRAGMENT_PROGRAM_ARB */
-      1157, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
-      1185, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
-      1184, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
-      1169, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-      1175, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-      1174, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-       827, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
-       850, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
-       849, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
-       840, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
-       846, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
-       845, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
-       810, /* GL_MAX_DRAW_BUFFERS */
-       354, /* GL_DRAW_BUFFER0 */
-       357, /* GL_DRAW_BUFFER1 */
-       378, /* GL_DRAW_BUFFER2 */
-       381, /* GL_DRAW_BUFFER3 */
-       384, /* GL_DRAW_BUFFER4 */
-       387, /* GL_DRAW_BUFFER5 */
-       390, /* GL_DRAW_BUFFER6 */
-       393, /* GL_DRAW_BUFFER7 */
-       396, /* GL_DRAW_BUFFER8 */
-       399, /* GL_DRAW_BUFFER9 */
-       358, /* GL_DRAW_BUFFER10 */
-       361, /* GL_DRAW_BUFFER11 */
-       364, /* GL_DRAW_BUFFER12 */
-       367, /* GL_DRAW_BUFFER13 */
-       370, /* GL_DRAW_BUFFER14 */
-       373, /* GL_DRAW_BUFFER15 */
+      1456, /* GL_STENCIL_BACK_FUNC */
+      1454, /* GL_STENCIL_BACK_FAIL */
+      1458, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */
+      1460, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */
+       504, /* GL_FRAGMENT_PROGRAM_ARB */
+      1202, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */
+      1230, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */
+      1229, /* GL_PROGRAM_TEX_INDIRECTIONS_ARB */
+      1214, /* GL_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+      1220, /* GL_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+      1219, /* GL_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+       871, /* GL_MAX_PROGRAM_ALU_INSTRUCTIONS_ARB */
+       894, /* GL_MAX_PROGRAM_TEX_INSTRUCTIONS_ARB */
+       893, /* GL_MAX_PROGRAM_TEX_INDIRECTIONS_ARB */
+       884, /* GL_MAX_PROGRAM_NATIVE_ALU_INSTRUCTIONS_ARB */
+       890, /* GL_MAX_PROGRAM_NATIVE_TEX_INSTRUCTIONS_ARB */
+       889, /* GL_MAX_PROGRAM_NATIVE_TEX_INDIRECTIONS_ARB */
+       854, /* GL_MAX_DRAW_BUFFERS */
+       374, /* GL_DRAW_BUFFER0 */
+       377, /* GL_DRAW_BUFFER1 */
+       398, /* GL_DRAW_BUFFER2 */
+       401, /* GL_DRAW_BUFFER3 */
+       404, /* GL_DRAW_BUFFER4 */
+       407, /* GL_DRAW_BUFFER5 */
+       410, /* GL_DRAW_BUFFER6 */
+       413, /* GL_DRAW_BUFFER7 */
+       416, /* GL_DRAW_BUFFER8 */
+       419, /* GL_DRAW_BUFFER9 */
+       378, /* GL_DRAW_BUFFER10 */
+       381, /* GL_DRAW_BUFFER11 */
+       384, /* GL_DRAW_BUFFER12 */
+       387, /* GL_DRAW_BUFFER13 */
+       390, /* GL_DRAW_BUFFER14 */
+       393, /* GL_DRAW_BUFFER15 */
         81, /* GL_BLEND_EQUATION_ALPHA */
-       790, /* GL_MATRIX_PALETTE_ARB */
-       821, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
-       824, /* GL_MAX_PALETTE_MATRICES_ARB */
-       288, /* GL_CURRENT_PALETTE_MATRIX_ARB */
-       784, /* GL_MATRIX_INDEX_ARRAY_ARB */
-       283, /* GL_CURRENT_MATRIX_INDEX_ARB */
-       786, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
-       788, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
-       787, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
-       785, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
-      1575, /* GL_TEXTURE_DEPTH_SIZE */
-       338, /* GL_DEPTH_TEXTURE_MODE */
-      1541, /* GL_TEXTURE_COMPARE_MODE */
-      1539, /* GL_TEXTURE_COMPARE_FUNC */
-       216, /* GL_COMPARE_R_TO_TEXTURE */
-      1091, /* GL_POINT_SPRITE */
-       265, /* GL_COORD_REPLACE */
-      1095, /* GL_POINT_SPRITE_R_MODE_NV */
-      1212, /* GL_QUERY_COUNTER_BITS */
-       290, /* GL_CURRENT_QUERY */
-      1214, /* GL_QUERY_RESULT */
-      1216, /* GL_QUERY_RESULT_AVAILABLE */
-       871, /* GL_MAX_VERTEX_ATTRIBS */
-      1706, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
-       336, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
-       335, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
-       857, /* GL_MAX_TEXTURE_COORDS */
-       859, /* GL_MAX_TEXTURE_IMAGE_UNITS */
-      1162, /* GL_PROGRAM_ERROR_STRING_ARB */
-      1164, /* GL_PROGRAM_FORMAT_ASCII_ARB */
-      1163, /* GL_PROGRAM_FORMAT_ARB */
-      1620, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
-       317, /* GL_DEPTH_BOUNDS_TEST_EXT */
-       316, /* GL_DEPTH_BOUNDS_EXT */
+       834, /* GL_MATRIX_PALETTE_ARB */
+       865, /* GL_MAX_MATRIX_PALETTE_STACK_DEPTH_ARB */
+       868, /* GL_MAX_PALETTE_MATRICES_ARB */
+       304, /* GL_CURRENT_PALETTE_MATRIX_ARB */
+       828, /* GL_MATRIX_INDEX_ARRAY_ARB */
+       299, /* GL_CURRENT_MATRIX_INDEX_ARB */
+       830, /* GL_MATRIX_INDEX_ARRAY_SIZE_ARB */
+       832, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */
+       831, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */
+       829, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */
+      1634, /* GL_TEXTURE_DEPTH_SIZE */
+       358, /* GL_DEPTH_TEXTURE_MODE */
+      1600, /* GL_TEXTURE_COMPARE_MODE */
+      1598, /* GL_TEXTURE_COMPARE_FUNC */
+       232, /* GL_COMPARE_R_TO_TEXTURE */
+      1136, /* GL_POINT_SPRITE */
+       281, /* GL_COORD_REPLACE */
+      1140, /* GL_POINT_SPRITE_R_MODE_NV */
+      1257, /* GL_QUERY_COUNTER_BITS */
+       306, /* GL_CURRENT_QUERY */
+      1259, /* GL_QUERY_RESULT */
+      1261, /* GL_QUERY_RESULT_AVAILABLE */
+       916, /* GL_MAX_VERTEX_ATTRIBS */
+      1768, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */
+       356, /* GL_DEPTH_STENCIL_TO_RGBA_NV */
+       355, /* GL_DEPTH_STENCIL_TO_BGRA_NV */
+       902, /* GL_MAX_TEXTURE_COORDS */
+       904, /* GL_MAX_TEXTURE_IMAGE_UNITS */
+      1207, /* GL_PROGRAM_ERROR_STRING_ARB */
+      1209, /* GL_PROGRAM_FORMAT_ASCII_ARB */
+      1208, /* GL_PROGRAM_FORMAT_ARB */
+      1680, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */
+       335, /* GL_DEPTH_BOUNDS_TEST_EXT */
+       334, /* GL_DEPTH_BOUNDS_EXT */
         52, /* GL_ARRAY_BUFFER */
-       419, /* GL_ELEMENT_ARRAY_BUFFER */
+       440, /* GL_ELEMENT_ARRAY_BUFFER */
         53, /* GL_ARRAY_BUFFER_BINDING */
-       420, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
-      1680, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
-       958, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
+       441, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */
+      1742, /* GL_VERTEX_ARRAY_BUFFER_BINDING */
+      1003, /* GL_NORMAL_ARRAY_BUFFER_BINDING */
        139, /* GL_COLOR_ARRAY_BUFFER_BINDING */
-       577, /* GL_INDEX_ARRAY_BUFFER_BINDING */
-      1554, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
-       415, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
-      1321, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
-       461, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
-      1732, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
-      1702, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
-      1165, /* GL_PROGRAM_INSTRUCTIONS_ARB */
-       833, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
-      1171, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-       842, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
-      1183, /* GL_PROGRAM_TEMPORARIES_ARB */
-       848, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
-      1173, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
-       844, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
-      1177, /* GL_PROGRAM_PARAMETERS_ARB */
-       847, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
-      1172, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
-       843, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
-      1158, /* GL_PROGRAM_ATTRIBS_ARB */
-       828, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
-      1170, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
-       841, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
-      1156, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
-       826, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
-      1168, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-       839, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
-       834, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
-       830, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
-      1186, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
-      1631, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
-      1224, /* GL_READ_ONLY */
-      1740, /* GL_WRITE_ONLY */
-      1226, /* GL_READ_WRITE */
+       620, /* GL_INDEX_ARRAY_BUFFER_BINDING */
+      1613, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */
+       436, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */
+      1378, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */
+       482, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */
+      1794, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */
+      1764, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */
+      1210, /* GL_PROGRAM_INSTRUCTIONS_ARB */
+       877, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */
+      1216, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+       886, /* GL_MAX_PROGRAM_NATIVE_INSTRUCTIONS_ARB */
+      1228, /* GL_PROGRAM_TEMPORARIES_ARB */
+       892, /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
+      1218, /* GL_PROGRAM_NATIVE_TEMPORARIES_ARB */
+       888, /* GL_MAX_PROGRAM_NATIVE_TEMPORARIES_ARB */
+      1222, /* GL_PROGRAM_PARAMETERS_ARB */
+       891, /* GL_MAX_PROGRAM_PARAMETERS_ARB */
+      1217, /* GL_PROGRAM_NATIVE_PARAMETERS_ARB */
+       887, /* GL_MAX_PROGRAM_NATIVE_PARAMETERS_ARB */
+      1203, /* GL_PROGRAM_ATTRIBS_ARB */
+       872, /* GL_MAX_PROGRAM_ATTRIBS_ARB */
+      1215, /* GL_PROGRAM_NATIVE_ATTRIBS_ARB */
+       885, /* GL_MAX_PROGRAM_NATIVE_ATTRIBS_ARB */
+      1201, /* GL_PROGRAM_ADDRESS_REGISTERS_ARB */
+       870, /* GL_MAX_PROGRAM_ADDRESS_REGISTERS_ARB */
+      1213, /* GL_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+       883, /* GL_MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB */
+       878, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */
+       874, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */
+      1231, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */
+      1691, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */
+      1270, /* GL_READ_ONLY */
+      1802, /* GL_WRITE_ONLY */
+      1272, /* GL_READ_WRITE */
        100, /* GL_BUFFER_ACCESS */
        102, /* GL_BUFFER_MAPPED */
        104, /* GL_BUFFER_MAP_POINTER */
-      1625, /* GL_TIME_ELAPSED_EXT */
-       744, /* GL_MATRIX0_ARB */
-       756, /* GL_MATRIX1_ARB */
-       768, /* GL_MATRIX2_ARB */
-       772, /* GL_MATRIX3_ARB */
-       774, /* GL_MATRIX4_ARB */
-       776, /* GL_MATRIX5_ARB */
-       778, /* GL_MATRIX6_ARB */
-       780, /* GL_MATRIX7_ARB */
-       782, /* GL_MATRIX8_ARB */
-       783, /* GL_MATRIX9_ARB */
-       746, /* GL_MATRIX10_ARB */
-       747, /* GL_MATRIX11_ARB */
-       748, /* GL_MATRIX12_ARB */
-       749, /* GL_MATRIX13_ARB */
-       750, /* GL_MATRIX14_ARB */
-       751, /* GL_MATRIX15_ARB */
-       752, /* GL_MATRIX16_ARB */
-       753, /* GL_MATRIX17_ARB */
-       754, /* GL_MATRIX18_ARB */
-       755, /* GL_MATRIX19_ARB */
-       758, /* GL_MATRIX20_ARB */
-       759, /* GL_MATRIX21_ARB */
-       760, /* GL_MATRIX22_ARB */
-       761, /* GL_MATRIX23_ARB */
-       762, /* GL_MATRIX24_ARB */
-       763, /* GL_MATRIX25_ARB */
-       764, /* GL_MATRIX26_ARB */
-       765, /* GL_MATRIX27_ARB */
-       766, /* GL_MATRIX28_ARB */
-       767, /* GL_MATRIX29_ARB */
-       770, /* GL_MATRIX30_ARB */
-       771, /* GL_MATRIX31_ARB */
-      1427, /* GL_STREAM_DRAW */
-      1429, /* GL_STREAM_READ */
-      1425, /* GL_STREAM_COPY */
-      1389, /* GL_STATIC_DRAW */
-      1391, /* GL_STATIC_READ */
-      1387, /* GL_STATIC_COPY */
-       409, /* GL_DYNAMIC_DRAW */
-       411, /* GL_DYNAMIC_READ */
-       407, /* GL_DYNAMIC_COPY */
-       533, /* GL_GL_PIXEL_PACK_BUFFER */
-       535, /* GL_GL_PIXEL_UNPACK_BUFFER */
-       534, /* GL_GL_PIXEL_PACK_BUFFER_BINDING */
-       536, /* GL_GL_PIXEL_UNPACK_BUFFER_BINDING */
-       831, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
-       829, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
-       832, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
-       836, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
-       835, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
-       793, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
-      1421, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
+      1685, /* GL_TIME_ELAPSED_EXT */
+       788, /* GL_MATRIX0_ARB */
+       800, /* GL_MATRIX1_ARB */
+       812, /* GL_MATRIX2_ARB */
+       816, /* GL_MATRIX3_ARB */
+       818, /* GL_MATRIX4_ARB */
+       820, /* GL_MATRIX5_ARB */
+       822, /* GL_MATRIX6_ARB */
+       824, /* GL_MATRIX7_ARB */
+       826, /* GL_MATRIX8_ARB */
+       827, /* GL_MATRIX9_ARB */
+       790, /* GL_MATRIX10_ARB */
+       791, /* GL_MATRIX11_ARB */
+       792, /* GL_MATRIX12_ARB */
+       793, /* GL_MATRIX13_ARB */
+       794, /* GL_MATRIX14_ARB */
+       795, /* GL_MATRIX15_ARB */
+       796, /* GL_MATRIX16_ARB */
+       797, /* GL_MATRIX17_ARB */
+       798, /* GL_MATRIX18_ARB */
+       799, /* GL_MATRIX19_ARB */
+       802, /* GL_MATRIX20_ARB */
+       803, /* GL_MATRIX21_ARB */
+       804, /* GL_MATRIX22_ARB */
+       805, /* GL_MATRIX23_ARB */
+       806, /* GL_MATRIX24_ARB */
+       807, /* GL_MATRIX25_ARB */
+       808, /* GL_MATRIX26_ARB */
+       809, /* GL_MATRIX27_ARB */
+       810, /* GL_MATRIX28_ARB */
+       811, /* GL_MATRIX29_ARB */
+       814, /* GL_MATRIX30_ARB */
+       815, /* GL_MATRIX31_ARB */
+      1486, /* GL_STREAM_DRAW */
+      1488, /* GL_STREAM_READ */
+      1484, /* GL_STREAM_COPY */
+      1447, /* GL_STATIC_DRAW */
+      1449, /* GL_STATIC_READ */
+      1445, /* GL_STATIC_COPY */
+       430, /* GL_DYNAMIC_DRAW */
+       432, /* GL_DYNAMIC_READ */
+       428, /* GL_DYNAMIC_COPY */
+       575, /* GL_GL_PIXEL_PACK_BUFFER */
+       577, /* GL_GL_PIXEL_UNPACK_BUFFER */
+       576, /* GL_GL_PIXEL_PACK_BUFFER_BINDING */
+       578, /* GL_GL_PIXEL_UNPACK_BUFFER_BINDING */
+       329, /* GL_DEPTH24_STENCIL8 */
+      1678, /* GL_TEXTURE_STENCIL_SIZE */
+       875, /* GL_MAX_PROGRAM_EXEC_INSTRUCTIONS_NV */
+       873, /* GL_MAX_PROGRAM_CALL_DEPTH_NV */
+       876, /* GL_MAX_PROGRAM_IF_DEPTH_NV */
+       880, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */
+       879, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */
+       837, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */
+      1480, /* GL_STENCIL_TEST_TWO_SIDE_EXT */
         17, /* GL_ACTIVE_STENCIL_FACE_EXT */
-       892, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
-      1302, /* GL_SAMPLES_PASSED */
-       484, /* GL_FRAGMENT_SHADER */
-      1726, /* GL_VERTEX_SHADER */
-      1176, /* GL_PROGRAM_OBJECT_ARB */
-      1334, /* GL_SHADER_OBJECT_ARB */
-       817, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
-       875, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
-       869, /* GL_MAX_VARYING_FLOATS */
-       873, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
-       802, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
-       983, /* GL_OBJECT_TYPE_ARB */
-      1336, /* GL_SHADER_TYPE */
-       449, /* GL_FLOAT_VEC2 */
-       451, /* GL_FLOAT_VEC3 */
-       453, /* GL_FLOAT_VEC4 */
-       604, /* GL_INT_VEC2 */
-       606, /* GL_INT_VEC3 */
-       608, /* GL_INT_VEC4 */
+       937, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */
+      1359, /* GL_SAMPLES_PASSED */
+       505, /* GL_FRAGMENT_SHADER */
+      1788, /* GL_VERTEX_SHADER */
+      1221, /* GL_PROGRAM_OBJECT_ARB */
+      1391, /* GL_SHADER_OBJECT_ARB */
+       861, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */
+       920, /* GL_MAX_VERTEX_UNIFORM_COMPONENTS */
+       914, /* GL_MAX_VARYING_FLOATS */
+       918, /* GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS */
+       846, /* GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS */
+      1028, /* GL_OBJECT_TYPE_ARB */
+      1393, /* GL_SHADER_TYPE */
+       470, /* GL_FLOAT_VEC2 */
+       472, /* GL_FLOAT_VEC3 */
+       474, /* GL_FLOAT_VEC4 */
+       647, /* GL_INT_VEC2 */
+       649, /* GL_INT_VEC3 */
+       651, /* GL_INT_VEC4 */
         92, /* GL_BOOL */
         94, /* GL_BOOL_VEC2 */
         96, /* GL_BOOL_VEC3 */
         98, /* GL_BOOL_VEC4 */
-       443, /* GL_FLOAT_MAT2 */
-       445, /* GL_FLOAT_MAT3 */
-       447, /* GL_FLOAT_MAT4 */
-      1293, /* GL_SAMPLER_1D */
-      1295, /* GL_SAMPLER_2D */
-      1297, /* GL_SAMPLER_3D */
-      1298, /* GL_SAMPLER_CUBE */
-      1294, /* GL_SAMPLER_1D_SHADOW */
-      1296, /* GL_SAMPLER_2D_SHADOW */
-       527, /* GL_GL_FLOAT_MAT2x3 */
-       528, /* GL_GL_FLOAT_MAT2x4 */
-       529, /* GL_GL_FLOAT_MAT3x2 */
-       530, /* GL_GL_FLOAT_MAT3x4 */
-       531, /* GL_GL_FLOAT_MAT4x2 */
-       532, /* GL_GL_FLOAT_MAT4x3 */
-       311, /* GL_DELETE_STATUS */
-       220, /* GL_COMPILE_STATUS */
-       657, /* GL_LINK_STATUS */
-      1675, /* GL_VALIDATE_STATUS */
-       589, /* GL_INFO_LOG_LENGTH */
+       464, /* GL_FLOAT_MAT2 */
+       466, /* GL_FLOAT_MAT3 */
+       468, /* GL_FLOAT_MAT4 */
+      1350, /* GL_SAMPLER_1D */
+      1352, /* GL_SAMPLER_2D */
+      1354, /* GL_SAMPLER_3D */
+      1355, /* GL_SAMPLER_CUBE */
+      1351, /* GL_SAMPLER_1D_SHADOW */
+      1353, /* GL_SAMPLER_2D_SHADOW */
+       569, /* GL_GL_FLOAT_MAT2x3 */
+       570, /* GL_GL_FLOAT_MAT2x4 */
+       571, /* GL_GL_FLOAT_MAT3x2 */
+       572, /* GL_GL_FLOAT_MAT3x4 */
+       573, /* GL_GL_FLOAT_MAT4x2 */
+       574, /* GL_GL_FLOAT_MAT4x3 */
+       327, /* GL_DELETE_STATUS */
+       236, /* GL_COMPILE_STATUS */
+       701, /* GL_LINK_STATUS */
+      1737, /* GL_VALIDATE_STATUS */
+       632, /* GL_INFO_LOG_LENGTH */
         55, /* GL_ATTACHED_SHADERS */
         20, /* GL_ACTIVE_UNIFORMS */
         21, /* GL_ACTIVE_UNIFORM_MAX_LENGTH */
-      1335, /* GL_SHADER_SOURCE_LENGTH */
+      1392, /* GL_SHADER_SOURCE_LENGTH */
         15, /* GL_ACTIVE_ATTRIBUTES */
         16, /* GL_ACTIVE_ATTRIBUTE_MAX_LENGTH */
-       486, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
-      1338, /* GL_SHADING_LANGUAGE_VERSION */
-       289, /* GL_CURRENT_PROGRAM */
-      1032, /* GL_PALETTE4_RGB8_OES */
-      1034, /* GL_PALETTE4_RGBA8_OES */
-      1030, /* GL_PALETTE4_R5_G6_B5_OES */
-      1033, /* GL_PALETTE4_RGBA4_OES */
-      1031, /* GL_PALETTE4_RGB5_A1_OES */
-      1037, /* GL_PALETTE8_RGB8_OES */
-      1039, /* GL_PALETTE8_RGBA8_OES */
-      1035, /* GL_PALETTE8_R5_G6_B5_OES */
-      1038, /* GL_PALETTE8_RGBA4_OES */
-      1036, /* GL_PALETTE8_RGB5_A1_OES */
-       572, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
-       571, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
-      1509, /* GL_TEXTURE_1D_ARRAY_EXT */
-      1196, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
-      1511, /* GL_TEXTURE_2D_ARRAY_EXT */
-      1199, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
-      1517, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
-      1519, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
-       541, /* GL_GL_SRGB */
-       542, /* GL_GL_SRGB8 */
-       544, /* GL_GL_SRGB_ALPHA */
-       543, /* GL_GL_SRGB8_ALPHA8 */
-       540, /* GL_GL_SLUMINANCE_ALPHA */
-       539, /* GL_GL_SLUMINANCE8_ALPHA8 */
-       537, /* GL_GL_SLUMINANCE */
-       538, /* GL_GL_SLUMINANCE8 */
-       524, /* GL_GL_COMPRESSED_SRGB */
-       525, /* GL_GL_COMPRESSED_SRGB_ALPHA */
-       522, /* GL_GL_COMPRESSED_SLUMINANCE */
-       523, /* GL_GL_COMPRESSED_SLUMINANCE_ALPHA */
-      1093, /* GL_POINT_SPRITE_COORD_ORIGIN */
-       665, /* GL_LOWER_LEFT */
-      1672, /* GL_UPPER_LEFT */
-      1403, /* GL_STENCIL_BACK_REF */
-      1404, /* GL_STENCIL_BACK_VALUE_MASK */
-      1405, /* GL_STENCIL_BACK_WRITEMASK */
-       402, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
-      1238, /* GL_RENDERBUFFER_BINDING_EXT */
-      1223, /* GL_READ_FRAMEBUFFER_EXT */
-       403, /* GL_DRAW_FRAMEBUFFER_EXT */
-      1222, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
-       488, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE_EXT */
-       487, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME_EXT */
-       492, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL_EXT */
-       490, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE_EXT */
-       489, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_3D_ZOFFSET_EXT */
-       494, /* GL_FRAMEBUFFER_COMPLETE_EXT */
-       496, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT_EXT */
-       501, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT_EXT */
-       499, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
-       497, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
-       500, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
-       498, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
-       502, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
-       504, /* GL_FRAMEBUFFER_UNSUPPORTED_EXT */
-       503, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
-       799, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
-       145, /* GL_COLOR_ATTACHMENT0_EXT */
-       152, /* GL_COLOR_ATTACHMENT1_EXT */
-       153, /* GL_COLOR_ATTACHMENT2_EXT */
-       154, /* GL_COLOR_ATTACHMENT3_EXT */
-       155, /* GL_COLOR_ATTACHMENT4_EXT */
-       156, /* GL_COLOR_ATTACHMENT5_EXT */
-       157, /* GL_COLOR_ATTACHMENT6_EXT */
-       158, /* GL_COLOR_ATTACHMENT7_EXT */
-       159, /* GL_COLOR_ATTACHMENT8_EXT */
-       160, /* GL_COLOR_ATTACHMENT9_EXT */
-       146, /* GL_COLOR_ATTACHMENT10_EXT */
-       147, /* GL_COLOR_ATTACHMENT11_EXT */
-       148, /* GL_COLOR_ATTACHMENT12_EXT */
-       149, /* GL_COLOR_ATTACHMENT13_EXT */
-       150, /* GL_COLOR_ATTACHMENT14_EXT */
-       151, /* GL_COLOR_ATTACHMENT15_EXT */
-       313, /* GL_DEPTH_ATTACHMENT_EXT */
-      1394, /* GL_STENCIL_ATTACHMENT_EXT */
-       495, /* GL_FRAMEBUFFER_EXT */
-      1239, /* GL_RENDERBUFFER_EXT */
-      1242, /* GL_RENDERBUFFER_WIDTH_EXT */
-      1240, /* GL_RENDERBUFFER_HEIGHT_EXT */
-      1241, /* GL_RENDERBUFFER_INTERNAL_FORMAT_EXT */
-      1416, /* GL_STENCIL_INDEX_EXT */
-      1413, /* GL_STENCIL_INDEX1_EXT */
-      1414, /* GL_STENCIL_INDEX4_EXT */
-      1415, /* GL_STENCIL_INDEX8_EXT */
-      1412, /* GL_STENCIL_INDEX16_EXT */
-       426, /* GL_EVAL_BIT */
-      1220, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
-       659, /* GL_LIST_BIT */
-      1525, /* GL_TEXTURE_BIT */
-      1317, /* GL_SCISSOR_BIT */
+       507, /* GL_FRAGMENT_SHADER_DERIVATIVE_HINT */
+      1395, /* GL_SHADING_LANGUAGE_VERSION */
+       305, /* GL_CURRENT_PROGRAM */
+      1077, /* GL_PALETTE4_RGB8_OES */
+      1079, /* GL_PALETTE4_RGBA8_OES */
+      1075, /* GL_PALETTE4_R5_G6_B5_OES */
+      1078, /* GL_PALETTE4_RGBA4_OES */
+      1076, /* GL_PALETTE4_RGB5_A1_OES */
+      1082, /* GL_PALETTE8_RGB8_OES */
+      1084, /* GL_PALETTE8_RGBA8_OES */
+      1080, /* GL_PALETTE8_R5_G6_B5_OES */
+      1083, /* GL_PALETTE8_RGBA4_OES */
+      1081, /* GL_PALETTE8_RGB5_A1_OES */
+       614, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */
+       613, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */
+      1722, /* GL_UNSIGNED_NORMALIZED */
+      1568, /* GL_TEXTURE_1D_ARRAY_EXT */
+      1241, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */
+      1570, /* GL_TEXTURE_2D_ARRAY_EXT */
+      1244, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */
+      1576, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */
+      1578, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */
+       583, /* GL_GL_SRGB */
+       584, /* GL_GL_SRGB8 */
+       586, /* GL_GL_SRGB_ALPHA */
+       585, /* GL_GL_SRGB8_ALPHA8 */
+       582, /* GL_GL_SLUMINANCE_ALPHA */
+       581, /* GL_GL_SLUMINANCE8_ALPHA8 */
+       579, /* GL_GL_SLUMINANCE */
+       580, /* GL_GL_SLUMINANCE8 */
+       566, /* GL_GL_COMPRESSED_SRGB */
+       567, /* GL_GL_COMPRESSED_SRGB_ALPHA */
+       564, /* GL_GL_COMPRESSED_SLUMINANCE */
+       565, /* GL_GL_COMPRESSED_SLUMINANCE_ALPHA */
+      1138, /* GL_POINT_SPRITE_COORD_ORIGIN */
+       709, /* GL_LOWER_LEFT */
+      1734, /* GL_UPPER_LEFT */
+      1462, /* GL_STENCIL_BACK_REF */
+      1463, /* GL_STENCIL_BACK_VALUE_MASK */
+      1464, /* GL_STENCIL_BACK_WRITEMASK */
+       423, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */
+      1286, /* GL_RENDERBUFFER_BINDING_EXT */
+      1267, /* GL_READ_FRAMEBUFFER */
+       422, /* GL_DRAW_FRAMEBUFFER */
+      1268, /* GL_READ_FRAMEBUFFER_BINDING_EXT */
+      1296, /* GL_RENDERBUFFER_SAMPLES */
+       517, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE */
+       515, /* GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME */
+       526, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL */
+       522, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE */
+       524, /* GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER */
+       529, /* GL_FRAMEBUFFER_COMPLETE */
+       533, /* GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT */
+       539, /* GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT */
+       537, /* GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT */
+       535, /* GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT */
+       538, /* GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT */
+       536, /* GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER_EXT */
+       542, /* GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER_EXT */
+       545, /* GL_FRAMEBUFFER_UNSUPPORTED */
+       543, /* GL_FRAMEBUFFER_STATUS_ERROR_EXT */
+       843, /* GL_MAX_COLOR_ATTACHMENTS_EXT */
+       145, /* GL_COLOR_ATTACHMENT0 */
+       147, /* GL_COLOR_ATTACHMENT1 */
+       161, /* GL_COLOR_ATTACHMENT2 */
+       163, /* GL_COLOR_ATTACHMENT3 */
+       165, /* GL_COLOR_ATTACHMENT4 */
+       167, /* GL_COLOR_ATTACHMENT5 */
+       169, /* GL_COLOR_ATTACHMENT6 */
+       171, /* GL_COLOR_ATTACHMENT7 */
+       173, /* GL_COLOR_ATTACHMENT8 */
+       175, /* GL_COLOR_ATTACHMENT9 */
+       148, /* GL_COLOR_ATTACHMENT10 */
+       150, /* GL_COLOR_ATTACHMENT11 */
+       152, /* GL_COLOR_ATTACHMENT12 */
+       154, /* GL_COLOR_ATTACHMENT13 */
+       156, /* GL_COLOR_ATTACHMENT14 */
+       158, /* GL_COLOR_ATTACHMENT15 */
+       330, /* GL_DEPTH_ATTACHMENT */
+      1452, /* GL_STENCIL_ATTACHMENT */
+       508, /* GL_FRAMEBUFFER */
+      1284, /* GL_RENDERBUFFER */
+      1298, /* GL_RENDERBUFFER_WIDTH */
+      1291, /* GL_RENDERBUFFER_HEIGHT */
+      1293, /* GL_RENDERBUFFER_INTERNAL_FORMAT */
+      1475, /* GL_STENCIL_INDEX_EXT */
+      1472, /* GL_STENCIL_INDEX1_EXT */
+      1473, /* GL_STENCIL_INDEX4_EXT */
+      1474, /* GL_STENCIL_INDEX8_EXT */
+      1471, /* GL_STENCIL_INDEX16_EXT */
+      1295, /* GL_RENDERBUFFER_RED_SIZE */
+      1290, /* GL_RENDERBUFFER_GREEN_SIZE */
+      1287, /* GL_RENDERBUFFER_BLUE_SIZE */
+      1285, /* GL_RENDERBUFFER_ALPHA_SIZE */
+      1288, /* GL_RENDERBUFFER_DEPTH_SIZE */
+      1297, /* GL_RENDERBUFFER_STENCIL_SIZE */
+       541, /* GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE */
+       899, /* GL_MAX_SAMPLES */
+       447, /* GL_EVAL_BIT */
+      1265, /* GL_RASTER_POSITION_UNCLIPPED_IBM */
+       703, /* GL_LIST_BIT */
+      1584, /* GL_TEXTURE_BIT */
+      1374, /* GL_SCISSOR_BIT */
         29, /* GL_ALL_ATTRIB_BITS */
-       939, /* GL_MULTISAMPLE_BIT */
+       984, /* GL_MULTISAMPLE_BIT */
         30, /* GL_ALL_CLIENT_ATTRIB_BITS */
 };
 
diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
index de75325f15..1b31e283cd 100644
--- a/src/mesa/main/extensions.c
+++ b/src/mesa/main/extensions.c
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.3
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
@@ -45,13 +45,14 @@ static const struct {
    int flag_offset;
 } default_extensions[] = {
    { OFF, "GL_ARB_depth_texture",              F(ARB_depth_texture) },
-   { OFF, "GL_ARB_draw_buffers",               F(ARB_draw_buffers) },
+   { ON,  "GL_ARB_draw_buffers",               F(ARB_draw_buffers) },
    { OFF, "GL_ARB_fragment_program",           F(ARB_fragment_program) },
    { OFF, "GL_ARB_fragment_program_shadow",    F(ARB_fragment_program_shadow) },
    { OFF, "GL_ARB_fragment_shader",            F(ARB_fragment_shader) },
+   { OFF, "GL_ARB_framebuffer_object",         F(ARB_framebuffer_object) },
    { OFF, "GL_ARB_half_float_pixel",           F(ARB_half_float_pixel) },
    { OFF, "GL_ARB_imaging",                    F(ARB_imaging) },
-   { OFF, "GL_ARB_multisample",                F(ARB_multisample) },
+   { ON,  "GL_ARB_multisample",                F(ARB_multisample) },
    { OFF, "GL_ARB_multitexture",               F(ARB_multitexture) },
    { OFF, "GL_ARB_occlusion_query",            F(ARB_occlusion_query) },
    { OFF, "GL_ARB_pixel_buffer_object",        F(EXT_pixel_buffer_object) },
@@ -61,9 +62,9 @@ static const struct {
    { OFF, "GL_ARB_shading_language_100",       F(ARB_shading_language_100) },
    { OFF, "GL_ARB_shading_language_120",       F(ARB_shading_language_120) },
    { OFF, "GL_ARB_shadow",                     F(ARB_shadow) },
-   { OFF, "GL_ARB_shadow_ambient",             F(SGIX_shadow_ambient) },
+   { OFF, "GL_ARB_shadow_ambient",             F(ARB_shadow_ambient) },
    { OFF, "GL_ARB_texture_border_clamp",       F(ARB_texture_border_clamp) },
-   { OFF, "GL_ARB_texture_compression",        F(ARB_texture_compression) },
+   { ON,  "GL_ARB_texture_compression",        F(ARB_texture_compression) },
    { OFF, "GL_ARB_texture_cube_map",           F(ARB_texture_cube_map) },
    { OFF, "GL_ARB_texture_env_add",            F(EXT_texture_env_add) },
    { OFF, "GL_ARB_texture_env_combine",        F(ARB_texture_env_combine) },
@@ -74,7 +75,7 @@ static const struct {
    { OFF, "GL_ARB_texture_non_power_of_two",   F(ARB_texture_non_power_of_two)},
    { OFF, "GL_ARB_texture_rectangle",          F(NV_texture_rectangle) },
    { ON,  "GL_ARB_transpose_matrix",           F(ARB_transpose_matrix) },
-   { OFF, "GL_ARB_vertex_buffer_object",       F(ARB_vertex_buffer_object) },
+   { ON,  "GL_ARB_vertex_buffer_object",       F(ARB_vertex_buffer_object) },
    { OFF, "GL_ARB_vertex_program",             F(ARB_vertex_program) },
    { OFF, "GL_ARB_vertex_shader",              F(ARB_vertex_shader) },
    { ON,  "GL_ARB_window_pos",                 F(ARB_window_pos) },
@@ -86,7 +87,7 @@ static const struct {
    { OFF, "GL_EXT_blend_logic_op",             F(EXT_blend_logic_op) },
    { OFF, "GL_EXT_blend_minmax",               F(EXT_blend_minmax) },
    { OFF, "GL_EXT_blend_subtract",             F(EXT_blend_subtract) },
-   { ON,  "GL_EXT_clip_volume_hint",           F(EXT_clip_volume_hint) },
+   { OFF, "GL_EXT_clip_volume_hint",           F(EXT_clip_volume_hint) },
    { OFF, "GL_EXT_cull_vertex",                F(EXT_cull_vertex) },
    { ON,  "GL_EXT_compiled_vertex_array",      F(EXT_compiled_vertex_array) },
    { OFF, "GL_EXT_convolution",                F(EXT_convolution) },
@@ -98,7 +99,7 @@ static const struct {
    { OFF, "GL_EXT_fog_coord",                  F(EXT_fog_coord) },
    { OFF, "GL_EXT_gpu_program_parameters",     F(EXT_gpu_program_parameters) },
    { OFF, "GL_EXT_histogram",                  F(EXT_histogram) },
-   { OFF, "GL_EXT_multi_draw_arrays",          F(EXT_multi_draw_arrays) },
+   { ON,  "GL_EXT_multi_draw_arrays",          F(EXT_multi_draw_arrays) },
    { OFF, "GL_EXT_packed_depth_stencil",       F(EXT_packed_depth_stencil) },
    { ON,  "GL_EXT_packed_pixels",              F(EXT_packed_pixels) },
    { OFF, "GL_EXT_paletted_texture",           F(EXT_paletted_texture) },
@@ -126,8 +127,10 @@ static const struct {
    { ON,  "GL_EXT_texture_object",             F(EXT_texture_object) },
    { OFF, "GL_EXT_texture_rectangle",          F(NV_texture_rectangle) },
    { OFF, "GL_EXT_texture_sRGB",               F(EXT_texture_sRGB) },
+   { OFF, "GL_EXT_texture_swizzle",            F(EXT_texture_swizzle) },
    { OFF, "GL_EXT_timer_query",                F(EXT_timer_query) },
    { ON,  "GL_EXT_vertex_array",               F(EXT_vertex_array) },
+   { OFF, "GL_EXT_vertex_array_bgra",          F(EXT_vertex_array_bgra) },
    { OFF, "GL_EXT_vertex_array_set",           F(EXT_vertex_array_set) },
    { OFF, "GL_3DFX_texture_compression_FXT1",  F(TDFX_texture_compression_FXT1) },
    { OFF, "GL_APPLE_client_storage",           F(APPLE_client_storage) },
@@ -138,7 +141,7 @@ static const struct {
    { OFF, "GL_ATI_texture_mirror_once",        F(ATI_texture_mirror_once)},
    { OFF, "GL_ATI_fragment_shader",            F(ATI_fragment_shader)},
    { OFF, "GL_ATI_separate_stencil",           F(ATI_separate_stencil)},
-   { OFF, "GL_IBM_multimode_draw_arrays",      F(IBM_multimode_draw_arrays) },
+   { ON,  "GL_IBM_multimode_draw_arrays",      F(IBM_multimode_draw_arrays) },
    { ON,  "GL_IBM_rasterpos_clip",             F(IBM_rasterpos_clip) },
    { OFF, "GL_IBM_texture_mirrored_repeat",    F(ARB_texture_mirrored_repeat)},
    { OFF, "GL_INGR_blend_func_separate",       F(EXT_blend_func_separate) },
@@ -153,6 +156,7 @@ static const struct {
    { OFF, "GL_NV_fragment_program",            F(NV_fragment_program) },
    { ON,  "GL_NV_light_max_exponent",          F(NV_light_max_exponent) },
    { OFF, "GL_NV_point_sprite",                F(NV_point_sprite) },
+   { OFF, "GL_NV_texture_env_combine4",        F(NV_texture_env_combine4) },
    { OFF, "GL_NV_texture_rectangle",           F(NV_texture_rectangle) },
    { ON,  "GL_NV_texgen_reflection",           F(NV_texgen_reflection) },
    { OFF, "GL_NV_vertex_program",              F(NV_vertex_program) },
@@ -165,10 +169,7 @@ static const struct {
    { OFF, "GL_SGIS_texture_border_clamp",      F(ARB_texture_border_clamp) },
    { ON,  "GL_SGIS_texture_edge_clamp",        F(SGIS_texture_edge_clamp) },
    { ON,  "GL_SGIS_texture_lod",               F(SGIS_texture_lod) },
-   { OFF, "GL_SGIX_depth_texture",             F(ARB_depth_texture) },
-   { OFF, "GL_SGIX_shadow",                    F(SGIX_shadow) },
-   { OFF, "GL_SGIX_shadow_ambient",            F(SGIX_shadow_ambient) },
-   { OFF, "GL_SUN_multi_draw_arrays",          F(EXT_multi_draw_arrays) },
+   { ON,  "GL_SUN_multi_draw_arrays",          F(EXT_multi_draw_arrays) },
    { OFF, "GL_S3_s3tc",                        F(S3_s3tc) },
 };
 
@@ -182,7 +183,7 @@ void
 _mesa_enable_sw_extensions(GLcontext *ctx)
 {
    ctx->Extensions.ARB_depth_texture = GL_TRUE;
-   ctx->Extensions.ARB_draw_buffers = GL_TRUE;
+   /*ctx->Extensions.ARB_draw_buffers = GL_TRUE;*/
 #if FEATURE_ARB_fragment_program
    ctx->Extensions.ARB_fragment_program = GL_TRUE;
    ctx->Extensions.ARB_fragment_program_shadow = GL_TRUE;
@@ -190,6 +191,9 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
 #if FEATURE_ARB_fragment_shader
    ctx->Extensions.ARB_fragment_shader = GL_TRUE;
 #endif
+#if FEATURE_ARB_framebuffer_object
+   ctx->Extensions.ARB_framebuffer_object = GL_TRUE;
+#endif
    ctx->Extensions.ARB_half_float_pixel = GL_TRUE;
    ctx->Extensions.ARB_imaging = GL_TRUE;
    ctx->Extensions.ARB_multitexture = GL_TRUE;
@@ -207,6 +211,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    ctx->Extensions.ARB_shading_language_120 = GL_FALSE; /* not quite done */
 #endif
    ctx->Extensions.ARB_shadow = GL_TRUE;
+   ctx->Extensions.ARB_shadow_ambient = GL_TRUE;
    ctx->Extensions.ARB_texture_border_clamp = GL_TRUE;
    ctx->Extensions.ARB_texture_cube_map = GL_TRUE;
    ctx->Extensions.ARB_texture_env_combine = GL_TRUE;
@@ -222,7 +227,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    ctx->Extensions.ARB_vertex_shader = GL_TRUE;
 #endif
 #if FEATURE_ARB_vertex_buffer_object
-   ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE;
+   /*ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE;*/
 #endif
    ctx->Extensions.APPLE_vertex_array_object = GL_TRUE;
 #if FEATURE_ATI_fragment_shader
@@ -247,7 +252,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    ctx->Extensions.EXT_framebuffer_blit = GL_TRUE;
 #endif
    ctx->Extensions.EXT_histogram = GL_TRUE;
-   ctx->Extensions.EXT_multi_draw_arrays = GL_TRUE;
+   /*ctx->Extensions.EXT_multi_draw_arrays = GL_TRUE;*/
    ctx->Extensions.EXT_packed_depth_stencil = GL_TRUE;
    ctx->Extensions.EXT_paletted_texture = GL_TRUE;
 #if FEATURE_EXT_pixel_buffer_object
@@ -258,7 +263,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    ctx->Extensions.EXT_secondary_color = GL_TRUE;
    ctx->Extensions.EXT_shared_texture_palette = GL_TRUE;
    ctx->Extensions.EXT_stencil_wrap = GL_TRUE;
-   ctx->Extensions.EXT_stencil_two_side = GL_FALSE; /* obsolete */
+   ctx->Extensions.EXT_stencil_two_side = GL_TRUE;
    ctx->Extensions.EXT_texture_env_add = GL_TRUE;
    ctx->Extensions.EXT_texture_env_combine = GL_TRUE;
    ctx->Extensions.EXT_texture_env_dot3 = GL_TRUE;
@@ -267,7 +272,9 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
 #if FEATURE_EXT_texture_sRGB
    ctx->Extensions.EXT_texture_sRGB = GL_TRUE;
 #endif
-   ctx->Extensions.IBM_multimode_draw_arrays = GL_TRUE;
+   ctx->Extensions.EXT_texture_swizzle = GL_TRUE;
+   ctx->Extensions.EXT_vertex_array_bgra = GL_TRUE;
+   /*ctx->Extensions.IBM_multimode_draw_arrays = GL_TRUE;*/
    ctx->Extensions.MESA_pack_invert = GL_TRUE;
 #if FEATURE_MESA_program_debug
    ctx->Extensions.MESA_program_debug = GL_TRUE;
@@ -278,6 +285,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    ctx->Extensions.NV_blend_square = GL_TRUE;
    /*ctx->Extensions.NV_light_max_exponent = GL_TRUE;*/
    ctx->Extensions.NV_point_sprite = GL_TRUE;
+   ctx->Extensions.NV_texture_env_combine4 = GL_TRUE;
    ctx->Extensions.NV_texture_rectangle = GL_TRUE;
    /*ctx->Extensions.NV_texgen_reflection = GL_TRUE;*/
 #if FEATURE_NV_vertex_program
@@ -292,11 +300,18 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    ctx->Extensions.SGI_texture_color_table = GL_TRUE;
    ctx->Extensions.SGIS_generate_mipmap = GL_TRUE;
    ctx->Extensions.SGIS_texture_edge_clamp = GL_TRUE;
-   ctx->Extensions.SGIX_shadow = GL_TRUE;
-   ctx->Extensions.SGIX_shadow_ambient = GL_TRUE;
 #if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program
    ctx->Extensions.EXT_gpu_program_parameters = GL_TRUE;
 #endif
+#if FEATURE_texture_fxt1
+   _mesa_enable_extension(ctx, "GL_3DFX_texture_compression_FXT1");
+#endif
+#if FEATURE_texture_s3tc
+   if (ctx->Mesa_DXTn) {
+      _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+      _mesa_enable_extension(ctx, "GL_S3_s3tc");
+   }
+#endif
 }
 
 
@@ -326,10 +341,10 @@ _mesa_enable_imaging_extensions(GLcontext *ctx)
 void
 _mesa_enable_1_3_extensions(GLcontext *ctx)
 {
-   ctx->Extensions.ARB_multisample = GL_TRUE;
+   /*ctx->Extensions.ARB_multisample = GL_TRUE;*/
    ctx->Extensions.ARB_multitexture = GL_TRUE;
    ctx->Extensions.ARB_texture_border_clamp = GL_TRUE;
-   ctx->Extensions.ARB_texture_compression = GL_TRUE;
+   /*ctx->Extensions.ARB_texture_compression = GL_TRUE;*/
    ctx->Extensions.ARB_texture_cube_map = GL_TRUE;
    ctx->Extensions.ARB_texture_env_combine = GL_TRUE;
    ctx->Extensions.ARB_texture_env_dot3 = GL_TRUE;
@@ -356,7 +371,7 @@ _mesa_enable_1_4_extensions(GLcontext *ctx)
    ctx->Extensions.EXT_blend_minmax = GL_TRUE;
    ctx->Extensions.EXT_blend_subtract = GL_TRUE;
    ctx->Extensions.EXT_fog_coord = GL_TRUE;
-   ctx->Extensions.EXT_multi_draw_arrays = GL_TRUE;
+   /*ctx->Extensions.EXT_multi_draw_arrays = GL_TRUE;*/
    ctx->Extensions.EXT_point_parameters = GL_TRUE;
    ctx->Extensions.EXT_secondary_color = GL_TRUE;
    ctx->Extensions.EXT_stencil_wrap = GL_TRUE;
@@ -373,7 +388,7 @@ void
 _mesa_enable_1_5_extensions(GLcontext *ctx)
 {
    ctx->Extensions.ARB_occlusion_query = GL_TRUE;
-   ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE;
+   /*ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE;*/
    ctx->Extensions.EXT_shadow_funcs = GL_TRUE;
 }
 
@@ -385,7 +400,7 @@ _mesa_enable_1_5_extensions(GLcontext *ctx)
 void
 _mesa_enable_2_0_extensions(GLcontext *ctx)
 {
-   ctx->Extensions.ARB_draw_buffers = GL_TRUE;
+   /*ctx->Extensions.ARB_draw_buffers = GL_TRUE;*/
 #if FEATURE_ARB_fragment_shader
    ctx->Extensions.ARB_fragment_shader = GL_TRUE;
 #endif
@@ -397,7 +412,7 @@ _mesa_enable_2_0_extensions(GLcontext *ctx)
 #if FEATURE_ARB_shading_language_100
    ctx->Extensions.ARB_shading_language_100 = GL_TRUE;
 #endif
-   ctx->Extensions.EXT_stencil_two_side = GL_FALSE; /* obsolete */
+   ctx->Extensions.EXT_stencil_two_side = GL_TRUE;
 #if FEATURE_ARB_vertex_shader
    ctx->Extensions.ARB_vertex_shader = GL_TRUE;
 #endif
diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c
index 4c92d1fb5a..c3cdc11037 100644
--- a/src/mesa/main/fbobject.c
+++ b/src/mesa/main/fbobject.c
@@ -3,6 +3,7 @@
  * Version:  7.1
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -24,6 +25,8 @@
 
 
 /*
+ * GL_EXT/ARB_framebuffer_object extensions
+ *
  * Authors:
  *   Brian Paul
  */
@@ -34,6 +37,7 @@
 #include "fbobject.h"
 #include "framebuffer.h"
 #include "hash.h"
+#include "macros.h"
 #include "mipmap.h"
 #include "renderbuffer.h"
 #include "state.h"
@@ -122,8 +126,22 @@ _mesa_lookup_framebuffer(GLcontext *ctx, GLuint id)
 
 
 /**
+ * Mark the given framebuffer as invalid.  This will force the
+ * test for framebuffer completeness to be done before the framebuffer
+ * is used.
+ */
+static void
+invalidate_framebuffer(struct gl_framebuffer *fb)
+{
+   fb->_Status = 0; /* "indeterminate" */
+}
+
+
+/**
  * Given a GL_*_ATTACHMENTn token, return a pointer to the corresponding
  * gl_renderbuffer_attachment object.
+ * If \p attachment is GL_DEPTH_STENCIL_ATTACHMENT, return a pointer to
+ * the depth buffer attachment point.
  */
 struct gl_renderbuffer_attachment *
 _mesa_get_attachment(GLcontext *ctx, struct gl_framebuffer *fb,
@@ -153,6 +171,8 @@ _mesa_get_attachment(GLcontext *ctx, struct gl_framebuffer *fb,
 	 return NULL;
       }
       return &fb->Attachment[BUFFER_COLOR0 + i];
+   case GL_DEPTH_STENCIL_ATTACHMENT:
+      /* fall-through */
    case GL_DEPTH_ATTACHMENT_EXT:
       return &fb->Attachment[BUFFER_DEPTH];
    case GL_STENCIL_ATTACHMENT_EXT:
@@ -173,7 +193,7 @@ _mesa_remove_attachment(GLcontext *ctx, struct gl_renderbuffer_attachment *att)
    if (att->Type == GL_TEXTURE) {
       ASSERT(att->Texture);
       if (ctx->Driver.FinishRenderTexture) {
-         /* tell driver we're done rendering to this texobj */
+         /* tell driver that we're done rendering to this texture. */
          ctx->Driver.FinishRenderTexture(ctx, att);
       }
       _mesa_reference_texobj(&att->Texture, NULL); /* unbind */
@@ -226,6 +246,8 @@ _mesa_set_texture_attachment(GLcontext *ctx,
    if (att->Texture->Image[att->CubeMapFace][att->TextureLevel]) {
       ctx->Driver.RenderTexture(ctx, fb, att);
    }
+
+   invalidate_framebuffer(fb);
 }
 
 
@@ -263,11 +285,19 @@ _mesa_framebuffer_renderbuffer(GLcontext *ctx, struct gl_framebuffer *fb,
    ASSERT(att);
    if (rb) {
       _mesa_set_renderbuffer_attachment(ctx, att, rb);
+      if (attachment == GL_DEPTH_STENCIL_ATTACHMENT) {
+         /* do stencil attachment here (depth already done above) */
+         att = _mesa_get_attachment(ctx, fb, GL_STENCIL_ATTACHMENT_EXT);
+         assert(att);
+         _mesa_set_renderbuffer_attachment(ctx, att, rb);
+      }
    }
    else {
       _mesa_remove_attachment(ctx, att);
    }
 
+   invalidate_framebuffer(fb);
+
    _glthread_UNLOCK_MUTEX(fb->Mutex);
 }
 
@@ -410,15 +440,18 @@ fbo_incomplete(const char *msg, int index)
 /**
  * Test if the given framebuffer object is complete and update its
  * Status field with the results.
+ * Calls the ctx->Driver.ValidateFramebuffer() function to allow the
+ * driver to make hardware-specific validation/completeness checks.
  * Also update the framebuffer's Width and Height fields if the
  * framebuffer is complete.
  */
 void
 _mesa_test_framebuffer_completeness(GLcontext *ctx, struct gl_framebuffer *fb)
 {
-   GLuint numImages, width = 0, height = 0;
-   GLenum intFormat = GL_NONE;
-   GLuint w = 0, h = 0;
+   GLuint numImages;
+   GLenum intFormat = GL_NONE; /* color buffers' internal format */
+   GLuint minWidth = ~0, minHeight = ~0, maxWidth = 0, maxHeight = 0;
+   GLint numSamples = -1;
    GLint i;
    GLuint j;
 
@@ -428,11 +461,22 @@ _mesa_test_framebuffer_completeness(GLcontext *ctx, struct gl_framebuffer *fb)
    fb->Width = 0;
    fb->Height = 0;
 
-   /* Start at -2 to more easily loop over all attachment points */
+   /* Start at -2 to more easily loop over all attachment points.
+    *  -2: depth buffer
+    *  -1: stencil buffer
+    * >=0: color buffer
+    */
    for (i = -2; i < (GLint) ctx->Const.MaxColorAttachments; i++) {
       struct gl_renderbuffer_attachment *att;
       GLenum f;
 
+      /*
+       * XXX for ARB_fbo, only check color buffers that are named by
+       * GL_READ_BUFFER and GL_DRAW_BUFFERi.
+       */
+
+      /* check for attachment completeness
+       */
       if (i == -2) {
          att = &fb->Attachment[BUFFER_DEPTH];
          test_attachment_completeness(ctx, GL_DEPTH, att);
@@ -461,11 +505,15 @@ _mesa_test_framebuffer_completeness(GLcontext *ctx, struct gl_framebuffer *fb)
          }
       }
 
+      /* get width, height, format of the renderbuffer/texture
+       */
       if (att->Type == GL_TEXTURE) {
          const struct gl_texture_image *texImg
             = att->Texture->Image[att->CubeMapFace][att->TextureLevel];
-         w = texImg->Width;
-         h = texImg->Height;
+         minWidth = MIN2(minWidth, texImg->Width);
+         maxWidth = MAX2(maxWidth, texImg->Width);
+         minHeight = MIN2(minHeight, texImg->Height);
+         maxHeight = MAX2(maxHeight, texImg->Height);
          f = texImg->_BaseFormat;
          numImages++;
          if (f != GL_RGB && f != GL_RGBA && f != GL_DEPTH_COMPONENT
@@ -476,8 +524,10 @@ _mesa_test_framebuffer_completeness(GLcontext *ctx, struct gl_framebuffer *fb)
          }
       }
       else if (att->Type == GL_RENDERBUFFER_EXT) {
-         w = att->Renderbuffer->Width;
-         h = att->Renderbuffer->Height;
+         minWidth = MIN2(minWidth, att->Renderbuffer->Width);
+         maxWidth = MAX2(minWidth, att->Renderbuffer->Width);
+         minHeight = MIN2(minHeight, att->Renderbuffer->Height);
+         maxHeight = MAX2(minHeight, att->Renderbuffer->Height);
          f = att->Renderbuffer->InternalFormat;
          numImages++;
       }
@@ -486,25 +536,41 @@ _mesa_test_framebuffer_completeness(GLcontext *ctx, struct gl_framebuffer *fb)
          continue;
       }
 
+      if (numSamples < 0) {
+         /* first buffer */
+         numSamples = att->Renderbuffer->NumSamples;
+      }
+
+      /* Error-check width, height, format, samples
+       */
       if (numImages == 1) {
-         /* set required width, height and format */
-         width = w;
-         height = h;
-         if (i >= 0)
+         /* save format, num samples */
+         if (i >= 0) {
             intFormat = f;
+         }
       }
       else {
-         /* check that width, height, format are same */
-         if (w != width || h != height) {
-            fb->_Status = GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT;
-            fbo_incomplete("width or height mismatch", -1);
-            return;
+         if (!ctx->Extensions.ARB_framebuffer_object) {
+            /* check that width, height, format are same */
+            if (minWidth != maxWidth || minHeight != maxHeight) {
+               fb->_Status = GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS_EXT;
+               fbo_incomplete("width or height mismatch", -1);
+               return;
+            }
+            /* check that all color buffer have same format */
+            if (intFormat != GL_NONE && f != intFormat) {
+               fb->_Status = GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT;
+               fbo_incomplete("format mismatch", -1);
+               return;
+            }
          }
-         if (intFormat != GL_NONE && f != intFormat) {
-            fb->_Status = GL_FRAMEBUFFER_INCOMPLETE_FORMATS_EXT;
-            fbo_incomplete("format mismatch", -1);
+         if (att->Renderbuffer &&
+             att->Renderbuffer->NumSamples != numSamples) {
+            fb->_Status = GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE;
+            fbo_incomplete("inconsistant number of samples", i);
             return;
-         }
+         }            
+
       }
    }
 
@@ -542,12 +608,32 @@ _mesa_test_framebuffer_completeness(GLcontext *ctx, struct gl_framebuffer *fb)
       return;
    }
 
-   /*
-    * If we get here, the framebuffer is complete!
-    */
+   /* Provisionally set status = COMPLETE ... */
    fb->_Status = GL_FRAMEBUFFER_COMPLETE_EXT;
-   fb->Width = w;
-   fb->Height = h;
+
+   /* ... but the driver may say the FB is incomplete.
+    * Drivers will most likely set the status to GL_FRAMEBUFFER_UNSUPPORTED
+    * if anything.
+    */
+   if (ctx->Driver.ValidateFramebuffer) {
+      ctx->Driver.ValidateFramebuffer(ctx, fb);
+      if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
+         fbo_incomplete("driver marked FBO as incomplete", -1);
+      }
+   }
+
+   if (fb->_Status == GL_FRAMEBUFFER_COMPLETE_EXT) {
+      /*
+       * Note that if ARB_framebuffer_object is supported and the attached
+       * renderbuffers/textures are different sizes, the framebuffer
+       * width/height will be set to the smallest width/height.
+       */
+      fb->Width = minWidth;
+      fb->Height = minHeight;
+
+      /* finally, update the visual info for the framebuffer */
+      _mesa_update_framebuffer_visual(fb);
+   }
 }
 
 
@@ -593,6 +679,12 @@ _mesa_BindRenderbufferEXT(GLenum target, GLuint renderbuffer)
          /* ID was reserved, but no real renderbuffer object made yet */
          newRb = NULL;
       }
+      else if (!newRb && ctx->Extensions.ARB_framebuffer_object) {
+         /* All RB IDs must be Gen'd */
+         _mesa_error(ctx, GL_INVALID_OPERATION, "glBindRenderbuffer(buffer)");
+         return;
+      }
+
       if (!newRb) {
 	 /* create new renderbuffer object */
 	 newRb = ctx->Driver.NewRenderbuffer(ctx, renderbuffer);
@@ -615,6 +707,27 @@ _mesa_BindRenderbufferEXT(GLenum target, GLuint renderbuffer)
 }
 
 
+/**
+ * If the given renderbuffer is anywhere attached to the framebuffer, detach
+ * the renderbuffer.
+ * This is used when a renderbuffer object is deleted.
+ * The spec calls for unbinding.
+ */
+static void
+detach_renderbuffer(GLcontext *ctx,
+                    struct gl_framebuffer *fb,
+                    struct gl_renderbuffer *rb)
+{
+   GLuint i;
+   for (i = 0; i < BUFFER_COUNT; i++) {
+      if (fb->Attachment[i].Renderbuffer == rb) {
+         _mesa_remove_attachment(ctx, &fb->Attachment[i]);
+      }
+   }
+   invalidate_framebuffer(fb);
+}
+
+
 void GLAPIENTRY
 _mesa_DeleteRenderbuffersEXT(GLsizei n, const GLuint *renderbuffers)
 {
@@ -636,6 +749,13 @@ _mesa_DeleteRenderbuffersEXT(GLsizei n, const GLuint *renderbuffers)
                _mesa_BindRenderbufferEXT(GL_RENDERBUFFER_EXT, 0);
             }
 
+            if (ctx->DrawBuffer->Name) {
+               detach_renderbuffer(ctx, ctx->DrawBuffer, rb);
+            }
+            if (ctx->ReadBuffer->Name && ctx->ReadBuffer != ctx->DrawBuffer) {
+               detach_renderbuffer(ctx, ctx->ReadBuffer, rb);
+            }
+
 	    /* Remove from hash table immediately, to free the ID.
              * But the object will not be freed until it's no longer
              * referenced anywhere else.
@@ -737,10 +857,21 @@ _mesa_base_fbo_format(GLcontext *ctx, GLenum internalFormat)
 }
 
 
-void GLAPIENTRY
-_mesa_RenderbufferStorageEXT(GLenum target, GLenum internalFormat,
-                             GLsizei width, GLsizei height)
+/** sentinal value, see below */
+#define NO_SAMPLES 1000
+
+
+/**
+ * Helper function used by _mesa_RenderbufferStorageEXT() and 
+ * _mesa_RenderbufferStorageMultisample().
+ * samples will be NO_SAMPLES if called by _mesa_RenderbufferStorageEXT().
+ */
+static void
+renderbuffer_storage(GLenum target, GLenum internalFormat,
+                     GLsizei width, GLsizei height, GLsizei samples)
 {
+   const char *func = samples == NO_SAMPLES ?
+      "glRenderbufferStorage" : "RenderbufferStorageMultisample";
    struct gl_renderbuffer *rb;
    GLenum baseFormat;
    GET_CURRENT_CONTEXT(ctx);
@@ -748,31 +879,39 @@ _mesa_RenderbufferStorageEXT(GLenum target, GLenum internalFormat,
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
    if (target != GL_RENDERBUFFER_EXT) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glRenderbufferStorageEXT(target)");
+      _mesa_error(ctx, GL_INVALID_ENUM, "%s(target)", func);
       return;
    }
 
    baseFormat = _mesa_base_fbo_format(ctx, internalFormat);
    if (baseFormat == 0) {
-      _mesa_error(ctx, GL_INVALID_ENUM,
-                  "glRenderbufferStorageEXT(internalFormat)");
+      _mesa_error(ctx, GL_INVALID_ENUM, "%s(internalFormat)", func);
       return;
    }
 
    if (width < 1 || width > (GLsizei) ctx->Const.MaxRenderbufferSize) {
-      _mesa_error(ctx, GL_INVALID_VALUE, "glRenderbufferStorageEXT(width)");
+      _mesa_error(ctx, GL_INVALID_VALUE, "%s(width)", func);
       return;
    }
 
    if (height < 1 || height > (GLsizei) ctx->Const.MaxRenderbufferSize) {
-      _mesa_error(ctx, GL_INVALID_VALUE, "glRenderbufferStorageEXT(height)");
+      _mesa_error(ctx, GL_INVALID_VALUE, "%s(height)", func);
       return;
    }
 
-   rb = ctx->CurrentRenderbuffer;
+   if (samples == NO_SAMPLES) {
+      /* NumSamples == 0 indicates non-multisampling */
+      samples = 0;
+   }
+   else if (samples > ctx->Const.MaxSamples) {
+      /* note: driver may choose to use more samples than what's requested */
+      _mesa_error(ctx, GL_INVALID_VALUE, "%s(samples)", func);
+      return;
+   }
 
+   rb = ctx->CurrentRenderbuffer;
    if (!rb) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glRenderbufferStorageEXT");
+      _mesa_error(ctx, GL_INVALID_OPERATION, func);
       return;
    }
 
@@ -794,6 +933,7 @@ _mesa_RenderbufferStorageEXT(GLenum target, GLenum internalFormat,
    rb->IndexBits =
    rb->DepthBits =
    rb->StencilBits = 0;
+   rb->NumSamples = samples;
 
    /* Now allocate the storage */
    ASSERT(rb->AllocStorage);
@@ -820,7 +960,8 @@ _mesa_RenderbufferStorageEXT(GLenum target, GLenum internalFormat,
       rb->AlphaBits =
       rb->IndexBits =
       rb->DepthBits =
-      rb->StencilBits = 0;
+      rb->StencilBits =
+      rb->NumSamples = 0;
    }
 
    /*
@@ -833,8 +974,31 @@ _mesa_RenderbufferStorageEXT(GLenum target, GLenum internalFormat,
 
 
 void GLAPIENTRY
+_mesa_RenderbufferStorageEXT(GLenum target, GLenum internalFormat,
+                             GLsizei width, GLsizei height)
+{
+   /* GL_ARB_fbo says calling this function is equivalent to calling
+    * glRenderbufferStorageMultisample() with samples=0.  We pass in
+    * a token value here just for error reporting purposes.
+    */
+   renderbuffer_storage(target, internalFormat, width, height, NO_SAMPLES);
+}
+
+
+void GLAPIENTRY
+_mesa_RenderbufferStorageMultisample(GLenum target, GLsizei samples,
+                                     GLenum internalFormat,
+                                     GLsizei width, GLsizei height)
+{
+   renderbuffer_storage(target, internalFormat, width, height, samples);
+}
+
+
+
+void GLAPIENTRY
 _mesa_GetRenderbufferParameterivEXT(GLenum target, GLenum pname, GLint *params)
 {
+   struct gl_renderbuffer *rb;
    GET_CURRENT_CONTEXT(ctx);
 
    ASSERT_OUTSIDE_BEGIN_END(ctx);
@@ -845,7 +1009,8 @@ _mesa_GetRenderbufferParameterivEXT(GLenum target, GLenum pname, GLint *params)
       return;
    }
 
-   if (!ctx->CurrentRenderbuffer) {
+   rb = ctx->CurrentRenderbuffer;
+   if (!rb) {
       _mesa_error(ctx, GL_INVALID_OPERATION,
                   "glGetRenderbufferParameterivEXT");
       return;
@@ -855,32 +1020,38 @@ _mesa_GetRenderbufferParameterivEXT(GLenum target, GLenum pname, GLint *params)
 
    switch (pname) {
    case GL_RENDERBUFFER_WIDTH_EXT:
-      *params = ctx->CurrentRenderbuffer->Width;
+      *params = rb->Width;
       return;
    case GL_RENDERBUFFER_HEIGHT_EXT:
-      *params = ctx->CurrentRenderbuffer->Height;
+      *params = rb->Height;
       return;
    case GL_RENDERBUFFER_INTERNAL_FORMAT_EXT:
-      *params = ctx->CurrentRenderbuffer->InternalFormat;
+      *params = rb->InternalFormat;
       return;
    case GL_RENDERBUFFER_RED_SIZE_EXT:
-      *params = ctx->CurrentRenderbuffer->RedBits;
+      *params = rb->RedBits;
       break;
    case GL_RENDERBUFFER_GREEN_SIZE_EXT:
-      *params = ctx->CurrentRenderbuffer->GreenBits;
+      *params = rb->GreenBits;
       break;
    case GL_RENDERBUFFER_BLUE_SIZE_EXT:
-      *params = ctx->CurrentRenderbuffer->BlueBits;
+      *params = rb->BlueBits;
       break;
    case GL_RENDERBUFFER_ALPHA_SIZE_EXT:
-      *params = ctx->CurrentRenderbuffer->AlphaBits;
+      *params = rb->AlphaBits;
       break;
    case GL_RENDERBUFFER_DEPTH_SIZE_EXT:
-      *params = ctx->CurrentRenderbuffer->DepthBits;
+      *params = rb->DepthBits;
       break;
    case GL_RENDERBUFFER_STENCIL_SIZE_EXT:
-      *params = ctx->CurrentRenderbuffer->StencilBits;
+      *params = rb->StencilBits;
       break;
+   case GL_RENDERBUFFER_SAMPLES:
+      if (ctx->Extensions.ARB_framebuffer_object) {
+         *params = rb->NumSamples;
+         break;
+      }
+      /* fallthrough */
    default:
       _mesa_error(ctx, GL_INVALID_ENUM,
                   "glGetRenderbufferParameterivEXT(target)");
@@ -946,6 +1117,13 @@ _mesa_BindFramebufferEXT(GLenum target, GLuint framebuffer)
    GLboolean bindReadBuf, bindDrawBuf;
    GET_CURRENT_CONTEXT(ctx);
 
+#ifdef DEBUG
+   if (ctx->Extensions.ARB_framebuffer_object) {
+      ASSERT(ctx->Extensions.EXT_framebuffer_object);
+      ASSERT(ctx->Extensions.EXT_framebuffer_blit);
+   }
+#endif
+
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
    if (!ctx->Extensions.EXT_framebuffer_object) {
@@ -995,6 +1173,12 @@ _mesa_BindFramebufferEXT(GLenum target, GLuint framebuffer)
          /* ID was reserved, but no real framebuffer object made yet */
          newFb = NULL;
       }
+      else if (!newFb && ctx->Extensions.ARB_framebuffer_object) {
+         /* All FBO IDs must be Gen'd */
+         _mesa_error(ctx, GL_INVALID_OPERATION, "glBindFramebuffer(buffer)");
+         return;
+      }
+
       if (!newFb) {
 	 /* create new framebuffer object */
 	 newFb = ctx->Driver.NewFramebuffer(ctx, framebuffer);
@@ -1069,7 +1253,12 @@ _mesa_DeleteFramebuffersEXT(GLsizei n, const GLuint *framebuffers)
             if (fb == ctx->DrawBuffer) {
                /* bind default */
                ASSERT(fb->RefCount >= 2);
-               _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, 0);
+               _mesa_BindFramebufferEXT(GL_DRAW_FRAMEBUFFER_EXT, 0);
+            }
+            if (fb == ctx->ReadBuffer) {
+               /* bind default */
+               ASSERT(fb->RefCount >= 2);
+               _mesa_BindFramebufferEXT(GL_READ_FRAMEBUFFER_EXT, 0);
             }
 
 	    /* remove from hash table immediately, to free the ID */
@@ -1079,7 +1268,7 @@ _mesa_DeleteFramebuffersEXT(GLsizei n, const GLuint *framebuffers)
                /* But the object will not be freed until it's no longer
                 * bound in any context.
                 */
-               _mesa_unreference_framebuffer(&fb);
+               _mesa_reference_framebuffer(&fb, NULL);
 	    }
 	 }
       }
@@ -1158,7 +1347,10 @@ _mesa_CheckFramebufferStatusEXT(GLenum target)
 
    FLUSH_VERTICES(ctx, _NEW_BUFFERS);
 
-   _mesa_test_framebuffer_completeness(ctx, buffer);
+   if (buffer->_Status != GL_FRAMEBUFFER_COMPLETE) {
+      _mesa_test_framebuffer_completeness(ctx, buffer);
+   }
+
    return buffer->_Status;
 }
 
@@ -1175,16 +1367,31 @@ framebuffer_texture(GLcontext *ctx, const char *caller, GLenum target,
    struct gl_renderbuffer_attachment *att;
    struct gl_texture_object *texObj = NULL;
    struct gl_framebuffer *fb;
+   GLboolean error = GL_FALSE;
 
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   if (target != GL_FRAMEBUFFER_EXT) {
+   switch (target) {
+   case GL_READ_FRAMEBUFFER_EXT:
+      error = !ctx->Extensions.EXT_framebuffer_blit;
+      fb = ctx->ReadBuffer;
+      break;
+   case GL_DRAW_FRAMEBUFFER_EXT:
+      error = !ctx->Extensions.EXT_framebuffer_blit;
+      /* fall-through */
+   case GL_FRAMEBUFFER_EXT:
+      fb = ctx->DrawBuffer;
+      break;
+   default:
+      error = GL_TRUE;
+   }
+
+   if (error) {
       _mesa_error(ctx, GL_INVALID_ENUM,
-                  "glFramebufferTexture%sEXT(target)", caller);
+                  "glFramebufferTexture%sEXT(target=0x%x)", caller, target);
       return;
    }
 
-   fb = ctx->DrawBuffer;
    ASSERT(fb);
 
    /* check framebuffer binding */
@@ -1239,7 +1446,6 @@ framebuffer_texture(GLcontext *ctx, const char *caller, GLenum target,
          }
       }
 
-
       if ((level < 0) || 
           (level >= _mesa_max_texture_levels(ctx, texObj->Target))) {
          _mesa_error(ctx, GL_INVALID_VALUE,
@@ -1255,6 +1461,18 @@ framebuffer_texture(GLcontext *ctx, const char *caller, GLenum target,
       return;
    }
 
+   if (texObj && attachment == GL_DEPTH_STENCIL_ATTACHMENT) {
+      /* the texture format must be depth+stencil */
+      const struct gl_texture_image *texImg;
+      texImg = texObj->Image[0][texObj->BaseLevel];
+      if (!texImg || texImg->_BaseFormat != GL_DEPTH_STENCIL) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "glFramebufferTexture%sEXT(texture is not"
+                     " DEPTH_STENCIL format)", caller);
+         return;
+      }
+   }
+
    FLUSH_VERTICES(ctx, _NEW_BUFFERS);
    /* The above doesn't fully flush the drivers in the way that a
     * glFlush does, but that is required here:
@@ -1266,10 +1484,22 @@ framebuffer_texture(GLcontext *ctx, const char *caller, GLenum target,
    if (texObj) {
       _mesa_set_texture_attachment(ctx, fb, att, texObj, textarget,
                                    level, zoffset);
+      /* Set the render-to-texture flag.  We'll check this flag in
+       * glTexImage() and friends to determine if we need to revalidate
+       * any FBOs that might be rendering into this texture.
+       * This flag never gets cleared since it's non-trivial to determine
+       * when all FBOs might be done rendering to this texture.  That's OK
+       * though since it's uncommon to render to a texture then repeatedly
+       * call glTexImage() to change images in the texture.
+       */
+      texObj->_RenderToTexture = GL_TRUE;
    }
    else {
       _mesa_remove_attachment(ctx, att);
    }
+
+   invalidate_framebuffer(fb);
+
    _glthread_UNLOCK_MUTEX(fb->Mutex);
 }
 
@@ -1303,7 +1533,7 @@ _mesa_FramebufferTexture2DEXT(GLenum target, GLenum attachment,
        (textarget != GL_TEXTURE_RECTANGLE_ARB) &&
        (!IS_CUBE_FACE(textarget))) {
       _mesa_error(ctx, GL_INVALID_OPERATION,
-                  "glFramebufferTexture2DEXT(textarget)");
+                  "glFramebufferTexture2DEXT(textarget=0x%x)", textarget);
       return;
    }
 
@@ -1413,6 +1643,17 @@ _mesa_FramebufferRenderbufferEXT(GLenum target, GLenum attachment,
       rb = NULL;
    }
 
+   if (attachment == GL_DEPTH_STENCIL_ATTACHMENT) {
+      /* make sure the renderbuffer is a depth/stencil format */
+      if (rb->_BaseFormat != GL_DEPTH_STENCIL) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "glFramebufferRenderbufferEXT(renderbuffer"
+                     " is not DEPTH_STENCIL format)");
+         return;
+      }
+   }
+
+
    FLUSH_VERTICES(ctx, _NEW_BUFFERS);
    /* The above doesn't fully flush the drivers in the way that a
     * glFlush does, but that is required here:
@@ -1481,6 +1722,19 @@ _mesa_GetFramebufferAttachmentParameterivEXT(GLenum target, GLenum attachment,
       return;
    }
 
+   if (attachment == GL_DEPTH_STENCIL_ATTACHMENT) {
+      /* the depth and stencil attachments must point to the same buffer */
+      const struct gl_renderbuffer_attachment *depthAtt, *stencilAtt;
+      depthAtt = _mesa_get_attachment(ctx, buffer, GL_DEPTH_ATTACHMENT);
+      stencilAtt = _mesa_get_attachment(ctx, buffer, GL_STENCIL_ATTACHMENT);
+      if (depthAtt->Renderbuffer != stencilAtt->Renderbuffer) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "glGetFramebufferAttachmentParameterivEXT(DEPTH/STENCIL"
+                     " attachments differ)");
+         return;
+      }
+   }
+
    FLUSH_VERTICES(ctx, _NEW_BUFFERS);
    /* The above doesn't fully flush the drivers in the way that a
     * glFlush does, but that is required here:
@@ -1541,6 +1795,79 @@ _mesa_GetFramebufferAttachmentParameterivEXT(GLenum target, GLenum attachment,
 		     "glGetFramebufferAttachmentParameterivEXT(pname)");
       }
       return;
+   case GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING:
+      if (!ctx->Extensions.ARB_framebuffer_object) {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glGetFramebufferAttachmentParameterivEXT(pname)");
+      }
+      else {
+         *params = att->Renderbuffer->ColorEncoding;
+      }
+      return;
+   case GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE:
+      if (!ctx->Extensions.ARB_framebuffer_object) {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glGetFramebufferAttachmentParameterivEXT(pname)");
+         return;
+      }
+      else {
+         *params = att->Renderbuffer->ComponentType;
+      }
+      return;
+   case GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE:
+      if (!ctx->Extensions.ARB_framebuffer_object) {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glGetFramebufferAttachmentParameterivEXT(pname)");
+      }
+      else {
+         *params = att->Renderbuffer->RedBits;
+      }
+      return;
+   case GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE:
+      if (!ctx->Extensions.ARB_framebuffer_object) {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glGetFramebufferAttachmentParameterivEXT(pname)");
+      }
+      else {
+         *params = att->Renderbuffer->GreenBits;
+      }
+      return;
+   case GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE:
+      if (!ctx->Extensions.ARB_framebuffer_object) {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glGetFramebufferAttachmentParameterivEXT(pname)");
+      }
+      else {
+         *params = att->Renderbuffer->BlueBits;
+      }
+      return;
+   case GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE:
+      if (!ctx->Extensions.ARB_framebuffer_object) {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glGetFramebufferAttachmentParameterivEXT(pname)");
+      }
+      else {
+         *params = att->Renderbuffer->AlphaBits;
+      }
+      return;
+   case GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE:
+      if (!ctx->Extensions.ARB_framebuffer_object) {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glGetFramebufferAttachmentParameterivEXT(pname)");
+      }
+      else {
+         *params = att->Renderbuffer->DepthBits;
+      }
+      return;
+   case GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE:
+      if (!ctx->Extensions.ARB_framebuffer_object) {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glGetFramebufferAttachmentParameterivEXT(pname)");
+      }
+      else {
+         *params = att->Renderbuffer->StencilBits;
+      }
+      return;
    default:
       _mesa_error(ctx, GL_INVALID_ENUM,
                   "glGetFramebufferAttachmentParameterivEXT(pname)");
@@ -1574,19 +1901,38 @@ _mesa_GenerateMipmapEXT(GLenum target)
    texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
    texObj = _mesa_select_tex_object(ctx, texUnit, target);
 
-   /* XXX this might not handle cube maps correctly */
    _mesa_lock_texture(ctx, texObj);
-   ctx->Driver.GenerateMipmap(ctx, target, texObj);
+   if (target == GL_TEXTURE_CUBE_MAP) {
+      int face;
+
+      for (face = 0; face < 6; face++)
+	 ctx->Driver.GenerateMipmap(ctx,
+				    GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB + face,
+				    texObj);
+   } else {
+      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+   }
    _mesa_unlock_texture(ctx, texObj);
 }
 
 
 #if FEATURE_EXT_framebuffer_blit
+/**
+ * Blit rectangular region, optionally from one framebuffer to another.
+ *
+ * Note, if the src buffer is multisampled and the dest is not, this is
+ * when the samples must be resolved to a single color.
+ */
 void GLAPIENTRY
 _mesa_BlitFramebufferEXT(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
                          GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
                          GLbitfield mask, GLenum filter)
 {
+   const GLbitfield legalMaskBits = (GL_COLOR_BUFFER_BIT |
+                                     GL_DEPTH_BUFFER_BIT |
+                                     GL_STENCIL_BUFFER_BIT);
+   const struct gl_framebuffer *readFb, *drawFb;
+   const struct gl_renderbuffer *colorReadRb, *colorDrawRb;
    GET_CURRENT_CONTEXT(ctx);
 
    ASSERT_OUTSIDE_BEGIN_END(ctx);
@@ -1596,13 +1942,19 @@ _mesa_BlitFramebufferEXT(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
       _mesa_update_state(ctx);
    }
 
-   if (!ctx->ReadBuffer) {
-      /* XXX */
+   readFb = ctx->ReadBuffer;
+   drawFb = ctx->DrawBuffer;
+
+   if (!readFb || !drawFb) {
+      /* This will normally never happen but someday we may want to
+       * support MakeCurrent() with no drawables.
+       */
+      return;
    }
 
    /* check for complete framebuffers */
-   if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT ||
-       ctx->ReadBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
+   if (drawFb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT ||
+       readFb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
       _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
                   "glBlitFramebufferEXT(incomplete draw/read buffers)");
       return;
@@ -1613,9 +1965,7 @@ _mesa_BlitFramebufferEXT(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
       return;
    }
 
-   if (mask & ~(GL_COLOR_BUFFER_BIT |
-                GL_DEPTH_BUFFER_BIT |
-                GL_STENCIL_BUFFER_BIT)) {
+   if (mask & ~legalMaskBits) {
       _mesa_error( ctx, GL_INVALID_VALUE, "glBlitFramebufferEXT(mask)");
       return;
    }
@@ -1628,9 +1978,18 @@ _mesa_BlitFramebufferEXT(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
       return;
    }
 
+   /* get color read/draw renderbuffers */
+   if (mask & GL_COLOR_BUFFER_BIT) {
+      colorReadRb = readFb->_ColorReadBuffer;
+      colorDrawRb = drawFb->_ColorDrawBuffers[0];
+   }
+   else {
+      colorReadRb = colorDrawRb = NULL;
+   }
+
    if (mask & GL_STENCIL_BUFFER_BIT) {
-      struct gl_renderbuffer *readRb = ctx->ReadBuffer->_StencilBuffer;
-      struct gl_renderbuffer *drawRb = ctx->DrawBuffer->_StencilBuffer;
+      struct gl_renderbuffer *readRb = readFb->_StencilBuffer;
+      struct gl_renderbuffer *drawRb = drawFb->_StencilBuffer;
       if (readRb->StencilBits != drawRb->StencilBits) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "glBlitFramebufferEXT(stencil buffer size mismatch");
@@ -1639,8 +1998,8 @@ _mesa_BlitFramebufferEXT(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
    }
 
    if (mask & GL_DEPTH_BUFFER_BIT) {
-      struct gl_renderbuffer *readRb = ctx->ReadBuffer->_DepthBuffer;
-      struct gl_renderbuffer *drawRb = ctx->DrawBuffer->_DepthBuffer;
+      struct gl_renderbuffer *readRb = readFb->_DepthBuffer;
+      struct gl_renderbuffer *drawRb = drawFb->_DepthBuffer;
       if (readRb->DepthBits != drawRb->DepthBits) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "glBlitFramebufferEXT(depth buffer size mismatch");
@@ -1648,6 +2007,34 @@ _mesa_BlitFramebufferEXT(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
       }
    }
 
+   if (readFb->Visual.samples > 0 &&
+       drawFb->Visual.samples > 0 &&
+       readFb->Visual.samples != drawFb->Visual.samples) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glBlitFramebufferEXT(mismatched samples");
+      return;
+   }
+
+   /* extra checks for multisample copies... */
+   if (readFb->Visual.samples > 0 || drawFb->Visual.samples > 0) {
+      /* src and dest region sizes must be the same */
+      if (srcX1 - srcX0 != dstX1 - dstX0 ||
+          srcY1 - srcY0 != dstY1 - dstY0) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                "glBlitFramebufferEXT(bad src/dst multisample region sizes");
+         return;
+      }
+
+      /* color formats must match */
+      if (colorReadRb &&
+          colorDrawRb &&
+          colorReadRb->_ActualFormat != colorDrawRb->_ActualFormat) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                "glBlitFramebufferEXT(bad src/dst multisample pixel formats");
+         return;
+      }
+   }
+
    if (!ctx->Extensions.EXT_framebuffer_blit) {
       _mesa_error(ctx, GL_INVALID_OPERATION, "glBlitFramebufferEXT");
       return;
diff --git a/src/mesa/main/fbobject.h b/src/mesa/main/fbobject.h
index b6154719ab..33d16cc5a0 100644
--- a/src/mesa/main/fbobject.h
+++ b/src/mesa/main/fbobject.h
@@ -83,6 +83,11 @@ extern void GLAPIENTRY
 _mesa_RenderbufferStorageEXT(GLenum target, GLenum internalformat,
                              GLsizei width, GLsizei height);
 
+void GLAPIENTRY
+_mesa_RenderbufferStorageMultisample(GLenum target, GLsizei samples,
+                                     GLenum internalformat,
+                                     GLsizei width, GLsizei height);
+
 extern void GLAPIENTRY
 _mesa_GetRenderbufferParameterivEXT(GLenum target, GLenum pname,
                                     GLint *params);
diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c
index ebc02940de..d70b78f258 100644
--- a/src/mesa/main/ffvertex_prog.c
+++ b/src/mesa/main/ffvertex_prog.c
@@ -26,7 +26,7 @@
  **************************************************************************/
 
 /**
- * \file ffvertex_prog.
+ * \file ffvertex_prog.c
  *
  * Create a vertex program to execute the current fixed function T&L pipeline.
  * \author Keith Whitwell
@@ -47,17 +47,17 @@
 
 
 struct state_key {
+   unsigned light_color_material_mask:12;
+   unsigned light_material_mask:12;
    unsigned light_global_enabled:1;
    unsigned light_local_viewer:1;
    unsigned light_twoside:1;
    unsigned light_color_material:1;
-   unsigned light_color_material_mask:12;
-   unsigned light_material_mask:12;
    unsigned material_shininess_is_zero:1;
-
    unsigned need_eye_coords:1;
    unsigned normalize:1;
    unsigned rescale_normals:1;
+
    unsigned fog_source_is_depth:1;
    unsigned tnl_do_vertex_fog:1;
    unsigned separate_specular:1;
@@ -67,6 +67,8 @@ struct state_key {
    unsigned texture_enabled_global:1;
    unsigned fragprog_inputs_read:12;
 
+   unsigned varying_vp_inputs;
+
    struct {
       unsigned light_enabled:1;
       unsigned light_eyepos3_is_zero:1;
@@ -99,6 +101,7 @@ static GLuint translate_fog_mode( GLenum mode )
    }
 }
 
+
 #define TXG_NONE           0
 #define TXG_OBJ_LINEAR     1
 #define TXG_EYE_LINEAR     2
@@ -143,6 +146,7 @@ tnl_get_per_vertex_materials(GLcontext *ctx)
    return mask;
 }
 
+
 /**
  * Should fog be computed per-vertex?
  */
@@ -157,6 +161,7 @@ tnl_get_per_vertex_fog(GLcontext *ctx)
 #endif
 }
 
+
 static GLboolean check_active_shininess( GLcontext *ctx,
                                          const struct state_key *key,
                                          GLuint side )
@@ -174,8 +179,6 @@ static GLboolean check_active_shininess( GLcontext *ctx,
 
    return GL_FALSE;
 }
-     
-
 
 
 static void make_state_key( GLcontext *ctx, struct state_key *key )
@@ -193,6 +196,7 @@ static void make_state_key( GLcontext *ctx, struct state_key *key )
    key->need_eye_coords = ctx->_NeedEyeCoords;
 
    key->fragprog_inputs_read = fp->Base.InputsRead;
+   key->varying_vp_inputs = ctx->varying_vp_inputs;
 
    if (ctx->RenderMode == GL_FEEDBACK) {
       /* make sure the vertprog emits color and tex0 */
@@ -275,7 +279,7 @@ static void make_state_key( GLcontext *ctx, struct state_key *key )
        ctx->Texture._EnabledUnits)
       key->texture_enabled_global = 1;
       
-   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
+   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
       struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
 
       if (texUnit->_ReallyEnabled)
@@ -407,11 +411,13 @@ static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w )
    return reg;
 }
 
+
 static struct ureg swizzle1( struct ureg reg, int x )
 {
    return swizzle(reg, x, x, x, x);
 }
 
+
 static struct ureg get_temp( struct tnl_program *p )
 {
    int bit = _mesa_ffs( ~p->temp_in_use );
@@ -427,6 +433,7 @@ static struct ureg get_temp( struct tnl_program *p )
    return make_ureg(PROGRAM_TEMPORARY, bit-1);
 }
 
+
 static struct ureg reserve_temp( struct tnl_program *p )
 {
    struct ureg temp = get_temp( p );
@@ -434,6 +441,7 @@ static struct ureg reserve_temp( struct tnl_program *p )
    return temp;
 }
 
+
 static void release_temp( struct tnl_program *p, struct ureg reg )
 {
    if (reg.file == PROGRAM_TEMPORARY) {
@@ -448,16 +456,49 @@ static void release_temps( struct tnl_program *p )
 }
 
 
+static struct ureg register_param5(struct tnl_program *p, 
+				   GLint s0,
+				   GLint s1,
+				   GLint s2,
+				   GLint s3,
+                                   GLint s4)
+{
+   gl_state_index tokens[STATE_LENGTH];
+   GLint idx;
+   tokens[0] = s0;
+   tokens[1] = s1;
+   tokens[2] = s2;
+   tokens[3] = s3;
+   tokens[4] = s4;
+   idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens );
+   return make_ureg(PROGRAM_STATE_VAR, idx);
+}
+
+
+#define register_param1(p,s0)          register_param5(p,s0,0,0,0,0)
+#define register_param2(p,s0,s1)       register_param5(p,s0,s1,0,0,0)
+#define register_param3(p,s0,s1,s2)    register_param5(p,s0,s1,s2,0,0)
+#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
+
+
 
 /**
  * \param input  one of VERT_ATTRIB_x tokens.
  */
 static struct ureg register_input( struct tnl_program *p, GLuint input )
 {
-   p->program->Base.InputsRead |= (1<<input);
-   return make_ureg(PROGRAM_INPUT, input);
+   /* Material attribs are passed here as inputs >= 32
+    */
+   if (input >= 32 || (p->state->varying_vp_inputs & (1<<input))) {
+      p->program->Base.InputsRead |= (1<<input);
+      return make_ureg(PROGRAM_INPUT, input);
+   }
+   else {
+      return register_param3( p, STATE_INTERNAL, STATE_CURRENT_ATTRIB, input );
+   }
 }
 
+
 /**
  * \param input  one of VERT_RESULT_x tokens.
  */
@@ -467,6 +508,7 @@ static struct ureg register_output( struct tnl_program *p, GLuint output )
    return make_ureg(PROGRAM_OUTPUT, output);
 }
 
+
 static struct ureg register_const4f( struct tnl_program *p, 
 			      GLfloat s0,
 			      GLfloat s1,
@@ -496,6 +538,7 @@ static GLboolean is_undef( struct ureg reg )
    return reg.file == PROGRAM_UNDEFINED;
 }
 
+
 static struct ureg get_identity_param( struct tnl_program *p )
 {
    if (is_undef(p->identity)) 
@@ -504,31 +547,6 @@ static struct ureg get_identity_param( struct tnl_program *p )
    return p->identity;
 }
 
-static struct ureg register_param5(struct tnl_program *p, 
-				   GLint s0,
-				   GLint s1,
-				   GLint s2,
-				   GLint s3,
-                                   GLint s4)
-{
-   gl_state_index tokens[STATE_LENGTH];
-   GLint idx;
-   tokens[0] = s0;
-   tokens[1] = s1;
-   tokens[2] = s2;
-   tokens[3] = s3;
-   tokens[4] = s4;
-   idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens );
-   return make_ureg(PROGRAM_STATE_VAR, idx);
-}
-
-
-#define register_param1(p,s0)          register_param5(p,s0,0,0,0,0)
-#define register_param2(p,s0,s1)       register_param5(p,s0,s1,0,0,0)
-#define register_param3(p,s0,s1,s2)    register_param5(p,s0,s1,s2,0,0)
-#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
-
-
 static void register_matrix_param5( struct tnl_program *p,
 				    GLint s0, /* modelview, projection, etc */
 				    GLint s1, /* texture matrix number */
@@ -561,6 +579,7 @@ static void emit_arg( struct prog_src_register *src,
    ASSERT(src->Index == reg.idx);
 }
 
+
 static void emit_dst( struct prog_dst_register *dst,
 		      struct ureg reg, GLuint mask )
 {
@@ -576,6 +595,7 @@ static void emit_dst( struct prog_dst_register *dst,
    ASSERT(dst->Index == reg.idx);
 }
 
+
 static void debug_insn( struct prog_instruction *inst, const char *fn,
 			GLuint line )
 {
@@ -686,6 +706,7 @@ static void emit_matrix_transform_vec4( struct tnl_program *p,
    emit_op2(p, OPCODE_DP4, dest, WRITEMASK_W, src, mat[3]);
 }
 
+
 /* This version is much easier to implement if writemasks are not
  * supported natively on the target or (like SSE), the target doesn't
  * have a clean/obvious dotproduct implementation.
@@ -711,6 +732,7 @@ static void emit_transpose_matrix_transform_vec4( struct tnl_program *p,
       release_temp(p, tmp);
 }
 
+
 static void emit_matrix_transform_vec3( struct tnl_program *p,
 					struct ureg dest,
 					const struct ureg *mat,
@@ -738,6 +760,7 @@ static void emit_normalize_vec3( struct tnl_program *p,
 #endif
 }
 
+
 static void emit_passthrough( struct tnl_program *p, 
 			      GLuint input,
 			      GLuint output )
@@ -746,6 +769,7 @@ static void emit_passthrough( struct tnl_program *p,
    emit_op1(p, OPCODE_MOV, out, 0, register_input(p, input)); 
 }
 
+
 static struct ureg get_eye_position( struct tnl_program *p )
 {
    if (is_undef(p->eye_position)) {
@@ -793,7 +817,6 @@ static struct ureg get_eye_position_z( struct tnl_program *p )
 }
    
 
-
 static struct ureg get_eye_position_normalized( struct tnl_program *p )
 {
    if (is_undef(p->eye_position_normalized)) {
@@ -855,7 +878,6 @@ static struct ureg get_transformed_normal( struct tnl_program *p )
 }
 
 
-
 static void build_hpos( struct tnl_program *p )
 {
    struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
@@ -881,7 +903,9 @@ static GLuint material_attrib( GLuint side, GLuint property )
 	   side);
 }
 
-/* Get a bitmask of which material values vary on a per-vertex basis.
+
+/**
+ * Get a bitmask of which material values vary on a per-vertex basis.
  */
 static void set_material_flags( struct tnl_program *p )
 {
@@ -917,7 +941,9 @@ static struct ureg get_material( struct tnl_program *p, GLuint side,
 				   MAT_BIT_FRONT_AMBIENT | \
 				   MAT_BIT_FRONT_DIFFUSE) << (side))
 
-/* Either return a precalculated constant value or emit code to
+
+/**
+ * Either return a precalculated constant value or emit code to
  * calculate these values dynamically in the case where material calls
  * are present between begin/end pairs.
  *
@@ -960,6 +986,7 @@ static struct ureg get_lightprod( struct tnl_program *p, GLuint light,
       return register_param4(p, STATE_LIGHTPROD, light, side, property);
 }
 
+
 static struct ureg calculate_light_attenuation( struct tnl_program *p,
 						GLuint i, 
 						struct ureg VPpli,
@@ -1216,7 +1243,6 @@ static void build_lighting( struct tnl_program *p )
 	    struct ureg res0, res1;
 	    GLuint mask0, mask1;
 
-	    
 	    if (count == nr_lights) {
 	       if (separate) {
 		  mask0 = WRITEMASK_XYZ;
@@ -1237,7 +1263,6 @@ static void build_lighting( struct tnl_program *p )
 	       res1 = _col1;
 	    }
 
-
 	    if (!is_undef(att)) {
                /* light is attenuated by distance */
                emit_op1(p, OPCODE_LIT, lit, 0, dots);
@@ -1310,7 +1335,6 @@ static void build_lighting( struct tnl_program *p )
 
 	    emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0);
 	    emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1);
-
             /* restore negate flag for next lighting */
             dots = negate(dots);
 
@@ -1385,6 +1409,7 @@ static void build_fog( struct tnl_program *p )
       emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, WRITEMASK_X, input);
    }
 }
+
  
 static void build_reflect_texgen( struct tnl_program *p,
 				  struct ureg dest,
@@ -1404,6 +1429,7 @@ static void build_reflect_texgen( struct tnl_program *p,
    release_temp(p, tmp);
 }
 
+
 static void build_sphere_texgen( struct tnl_program *p,
 				 struct ureg dest,
 				 GLuint writemask )
@@ -1451,7 +1477,7 @@ static void build_texture_transform( struct tnl_program *p )
 {
    GLuint i, j;
 
-   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
+   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
 
       if (!(p->state->fragprog_inputs_read & FRAG_BIT_TEX(i)))
 	 continue;
@@ -1514,10 +1540,8 @@ static void build_texture_transform( struct tnl_program *p )
 	       case TXG_NONE:
 		  copy_mask |= WRITEMASK_X << j;
 	       }
-
 	    }
 
-	 
 	    if (sphere_mask) {
 	       build_sphere_texgen(p, out_texgen, sphere_mask);
 	    }
@@ -1600,6 +1624,7 @@ static void build_atten_pointsize( struct tnl_program *p )
    release_temp(p, ut);
 }
 
+
 /**
  * Emit constant point size.
  */
@@ -1610,6 +1635,7 @@ static void build_constant_pointsize( struct tnl_program *p )
    emit_op1(p, OPCODE_MOV, out, WRITEMASK_X, state_size);
 }
 
+
 /**
  * Pass-though per-vertex point size, from user's point size array.
  */
diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c
index af78363ad3..351bf6959a 100644
--- a/src/mesa/main/framebuffer.c
+++ b/src/mesa/main/framebuffer.c
@@ -35,6 +35,7 @@
 #include "buffers.h"
 #include "context.h"
 #include "depthstencil.h"
+#include "macros.h"
 #include "mtypes.h"
 #include "fbobject.h"
 #include "framebuffer.h"
@@ -222,19 +223,36 @@ _mesa_reference_framebuffer(struct gl_framebuffer **ptr,
       /* no change */
       return;
    }
+
    if (*ptr) {
-      _mesa_unreference_framebuffer(ptr);
+      /* unreference old renderbuffer */
+      GLboolean deleteFlag = GL_FALSE;
+      struct gl_framebuffer *oldFb = *ptr;
+
+      _glthread_LOCK_MUTEX(oldFb->Mutex);
+      ASSERT(oldFb->RefCount > 0);
+      oldFb->RefCount--;
+      deleteFlag = (oldFb->RefCount == 0);
+      _glthread_UNLOCK_MUTEX(oldFb->Mutex);
+      
+      if (deleteFlag)
+         oldFb->Delete(oldFb);
+
+      *ptr = NULL;
    }
    assert(!*ptr);
-   assert(fb);
-   _glthread_LOCK_MUTEX(fb->Mutex);
-   fb->RefCount++;
-   _glthread_UNLOCK_MUTEX(fb->Mutex);
-   *ptr = fb;
+
+   if (fb) {
+      _glthread_LOCK_MUTEX(fb->Mutex);
+      fb->RefCount++;
+      _glthread_UNLOCK_MUTEX(fb->Mutex);
+      *ptr = fb;
+   }
 }
 
 
 /**
+ * XXX this function is deprecated.
  * Undo/remove a reference to a framebuffer object.
  * Decrement the framebuffer object's reference count and delete it when
  * the refcount hits zero.
@@ -243,21 +261,7 @@ _mesa_reference_framebuffer(struct gl_framebuffer **ptr,
 void
 _mesa_unreference_framebuffer(struct gl_framebuffer **fb)
 {
-   assert(fb);
-   if (*fb) {
-      GLboolean deleteFlag = GL_FALSE;
-
-      _glthread_LOCK_MUTEX((*fb)->Mutex);
-      ASSERT((*fb)->RefCount > 0);
-      (*fb)->RefCount--;
-      deleteFlag = ((*fb)->RefCount == 0);
-      _glthread_UNLOCK_MUTEX((*fb)->Mutex);
-      
-      if (deleteFlag)
-         (*fb)->Delete(*fb);
-
-      *fb = NULL;
-   }
+   _mesa_reference_framebuffer(fb, NULL);
 }
 
 
@@ -418,14 +422,14 @@ _mesa_ResizeBuffersMESA( void )
 /**
  * Examine all the framebuffer's renderbuffers to update the Width/Height
  * fields of the framebuffer.  If we have renderbuffers with different
- * sizes, set the framebuffer's width and height to zero.
+ * sizes, set the framebuffer's width and height to the min size.
  * Note: this is only intended for user-created framebuffers, not
  * window-system framebuffes.
  */
 static void
-update_framebuffer_size(struct gl_framebuffer *fb)
+update_framebuffer_size(GLcontext *ctx, struct gl_framebuffer *fb)
 {
-   GLboolean haveSize = GL_FALSE;
+   GLuint minWidth = ~0, minHeight = ~0;
    GLuint i;
 
    /* user-created framebuffers only */
@@ -435,21 +439,19 @@ update_framebuffer_size(struct gl_framebuffer *fb)
       struct gl_renderbuffer_attachment *att = &fb->Attachment[i];
       const struct gl_renderbuffer *rb = att->Renderbuffer;
       if (rb) {
-         if (haveSize) {
-            if (rb->Width != fb->Width && rb->Height != fb->Height) {
-               /* size mismatch! */
-               fb->Width = 0;
-               fb->Height = 0;
-               return;
-            }
-         }
-         else {
-            fb->Width = rb->Width;
-            fb->Height = rb->Height;
-            haveSize = GL_TRUE;
-         }
+         minWidth = MIN2(minWidth, rb->Width);
+         minHeight = MIN2(minHeight, rb->Height);
       }
    }
+
+   if (minWidth != ~0) {
+      fb->Width = minWidth;
+      fb->Height = minHeight;
+   }
+   else {
+      fb->Width = 0;
+      fb->Height = 0;
+   }
 }
 
 
@@ -469,7 +471,7 @@ _mesa_update_draw_buffer_bounds(GLcontext *ctx)
 
    if (buffer->Name) {
       /* user-created framebuffer size depends on the renderbuffers */
-      update_framebuffer_size(buffer);
+      update_framebuffer_size(ctx, buffer);
    }
 
    buffer->_Xmin = 0;
@@ -544,6 +546,7 @@ _mesa_update_framebuffer_visual(struct gl_framebuffer *fb)
             fb->Visual.rgbBits = fb->Visual.redBits
                + fb->Visual.greenBits + fb->Visual.blueBits;
             fb->Visual.floatMode = GL_FALSE;
+            fb->Visual.samples = rb->NumSamples;
             break;
          }
          else if (rb->_BaseFormat == GL_COLOR_INDEX) {
@@ -793,8 +796,9 @@ update_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
       /* This is a user-created framebuffer.
        * Completeness only matters for user-created framebuffers.
        */
-      _mesa_test_framebuffer_completeness(ctx, fb);
-      _mesa_update_framebuffer_visual(fb);
+      if (fb->_Status != GL_FRAMEBUFFER_COMPLETE) {
+         _mesa_test_framebuffer_completeness(ctx, fb);
+      }
    }
 
    /* Strictly speaking, we don't need to update the draw-state
diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c
index f72aa6a288..87a821b12d 100644
--- a/src/mesa/main/get.c
+++ b/src/mesa/main/get.c
@@ -289,7 +289,7 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          params[0] = INT_TO_BOOLEAN(ctx->DrawBuffer->Visual.depthBits);
          break;
       case GL_DEPTH_CLEAR_VALUE:
-         params[0] = FLOAT_TO_BOOLEAN(ctx->Depth.Clear);
+         params[0] = FLOAT_TO_BOOLEAN(((GLfloat) ctx->Depth.Clear));
          break;
       case GL_DEPTH_FUNC:
          params[0] = ENUM_TO_BOOLEAN(ctx->Depth.Func);
@@ -460,7 +460,7 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          params[0] = INT_TO_BOOLEAN(ctx->List.ListBase);
          break;
       case GL_LIST_INDEX:
-         params[0] = INT_TO_BOOLEAN(ctx->ListState.CurrentListNum);
+         params[0] = INT_TO_BOOLEAN((ctx->ListState.CurrentList ? ctx->ListState.CurrentList->Name : 0));
          break;
       case GL_LIST_MODE:
          {
@@ -1078,15 +1078,12 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          params[0] = INT_TO_BOOLEAN((1 << (ctx->Const.MaxCubeTextureLevels - 1)));
          break;
       case GL_TEXTURE_COMPRESSION_HINT_ARB:
-         CHECK_EXT1(ARB_texture_compression, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Hint.TextureCompression);
          break;
       case GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB:
-         CHECK_EXT1(ARB_texture_compression, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(_mesa_get_compressed_formats(ctx, NULL, GL_FALSE));
          break;
       case GL_COMPRESSED_TEXTURE_FORMATS_ARB:
-         CHECK_EXT1(ARB_texture_compression, "GetBooleanv");
          {
          GLint formats[100];
          GLuint i, n = _mesa_get_compressed_formats(ctx, formats, GL_FALSE);
@@ -1368,35 +1365,27 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          params[0] = FLOAT_TO_BOOLEAN(ctx->Const.MaxTextureMaxAnisotropy);
          break;
       case GL_MULTISAMPLE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetBooleanv");
          params[0] = ctx->Multisample.Enabled;
          break;
       case GL_SAMPLE_ALPHA_TO_COVERAGE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetBooleanv");
          params[0] = ctx->Multisample.SampleAlphaToCoverage;
          break;
       case GL_SAMPLE_ALPHA_TO_ONE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetBooleanv");
          params[0] = ctx->Multisample.SampleAlphaToOne;
          break;
       case GL_SAMPLE_COVERAGE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetBooleanv");
          params[0] = ctx->Multisample.SampleCoverage;
          break;
       case GL_SAMPLE_COVERAGE_VALUE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetBooleanv");
          params[0] = FLOAT_TO_BOOLEAN(ctx->Multisample.SampleCoverageValue);
          break;
       case GL_SAMPLE_COVERAGE_INVERT_ARB:
-         CHECK_EXT1(ARB_multisample, "GetBooleanv");
          params[0] = ctx->Multisample.SampleCoverageInvert;
          break;
       case GL_SAMPLE_BUFFERS_ARB:
-         CHECK_EXT1(ARB_multisample, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->DrawBuffer->Visual.sampleBuffers);
          break;
       case GL_SAMPLES_ARB:
-         CHECK_EXT1(ARB_multisample, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->DrawBuffer->Visual.samples);
          break;
       case GL_RASTER_POSITION_UNCLIPPED_IBM:
@@ -1592,43 +1581,33 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          params[0] = FLOAT_TO_BOOLEAN(ctx->Const.MaxSpotExponent);
          break;
       case GL_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ArrayBufferObj->Name);
          break;
       case GL_VERTEX_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ArrayObj->Vertex.BufferObj->Name);
          break;
       case GL_NORMAL_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ArrayObj->Normal.BufferObj->Name);
          break;
       case GL_COLOR_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ArrayObj->Color.BufferObj->Name);
          break;
       case GL_INDEX_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ArrayObj->Index.BufferObj->Name);
          break;
       case GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ArrayObj->TexCoord[ctx->Array.ActiveTexture].BufferObj->Name);
          break;
       case GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ArrayObj->EdgeFlag.BufferObj->Name);
          break;
       case GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ArrayObj->SecondaryColor.BufferObj->Name);
          break;
       case GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ArrayObj->FogCoord.BufferObj->Name);
          break;
       case GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Array.ElementArrayBufferObj->Name);
          break;
       case GL_PIXEL_PACK_BUFFER_BINDING_EXT:
@@ -1753,15 +1732,12 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          params[0] = INT_TO_BOOLEAN(ctx->VertexProgram.CurrentPosition);
          break;
       case GL_MAX_DRAW_BUFFERS_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Const.MaxDrawBuffers);
          break;
       case GL_DRAW_BUFFER0_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetBooleanv");
          params[0] = ENUM_TO_BOOLEAN(ctx->DrawBuffer->ColorDrawBuffer[0]);
          break;
       case GL_DRAW_BUFFER1_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetBooleanv");
          {
          GLenum buffer;
          if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
@@ -1773,7 +1749,6 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          }
          break;
       case GL_DRAW_BUFFER2_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetBooleanv");
          {
          GLenum buffer;
          if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
@@ -1785,7 +1760,6 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          }
          break;
       case GL_DRAW_BUFFER3_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetBooleanv");
          {
          GLenum buffer;
          if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
@@ -1905,6 +1879,10 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
          CHECK_EXT1(ARB_shader_objects, "GetBooleanv");
          params[0] = INT_TO_BOOLEAN(ctx->Shader.CurrentProgram ? ctx->Shader.CurrentProgram->Name : 0);
          break;
+      case GL_MAX_SAMPLES:
+         CHECK_EXT1(ARB_framebuffer_object, "GetBooleanv");
+         params[0] = INT_TO_BOOLEAN(ctx->Const.MaxSamples);
+         break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetBooleanv(pname=0x%x)", pname);
    }
@@ -2137,7 +2115,7 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          params[0] = (GLfloat)(ctx->DrawBuffer->Visual.depthBits);
          break;
       case GL_DEPTH_CLEAR_VALUE:
-         params[0] = ctx->Depth.Clear;
+         params[0] = ((GLfloat) ctx->Depth.Clear);
          break;
       case GL_DEPTH_FUNC:
          params[0] = ENUM_TO_FLOAT(ctx->Depth.Func);
@@ -2308,7 +2286,7 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          params[0] = (GLfloat)(ctx->List.ListBase);
          break;
       case GL_LIST_INDEX:
-         params[0] = (GLfloat)(ctx->ListState.CurrentListNum);
+         params[0] = (GLfloat)((ctx->ListState.CurrentList ? ctx->ListState.CurrentList->Name : 0));
          break;
       case GL_LIST_MODE:
          {
@@ -2926,15 +2904,12 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          params[0] = (GLfloat)((1 << (ctx->Const.MaxCubeTextureLevels - 1)));
          break;
       case GL_TEXTURE_COMPRESSION_HINT_ARB:
-         CHECK_EXT1(ARB_texture_compression, "GetFloatv");
          params[0] = (GLfloat)(ctx->Hint.TextureCompression);
          break;
       case GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB:
-         CHECK_EXT1(ARB_texture_compression, "GetFloatv");
          params[0] = (GLfloat)(_mesa_get_compressed_formats(ctx, NULL, GL_FALSE));
          break;
       case GL_COMPRESSED_TEXTURE_FORMATS_ARB:
-         CHECK_EXT1(ARB_texture_compression, "GetFloatv");
          {
          GLint formats[100];
          GLuint i, n = _mesa_get_compressed_formats(ctx, formats, GL_FALSE);
@@ -3216,35 +3191,27 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          params[0] = ctx->Const.MaxTextureMaxAnisotropy;
          break;
       case GL_MULTISAMPLE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetFloatv");
          params[0] = BOOLEAN_TO_FLOAT(ctx->Multisample.Enabled);
          break;
       case GL_SAMPLE_ALPHA_TO_COVERAGE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetFloatv");
          params[0] = BOOLEAN_TO_FLOAT(ctx->Multisample.SampleAlphaToCoverage);
          break;
       case GL_SAMPLE_ALPHA_TO_ONE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetFloatv");
          params[0] = BOOLEAN_TO_FLOAT(ctx->Multisample.SampleAlphaToOne);
          break;
       case GL_SAMPLE_COVERAGE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetFloatv");
          params[0] = BOOLEAN_TO_FLOAT(ctx->Multisample.SampleCoverage);
          break;
       case GL_SAMPLE_COVERAGE_VALUE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetFloatv");
          params[0] = ctx->Multisample.SampleCoverageValue;
          break;
       case GL_SAMPLE_COVERAGE_INVERT_ARB:
-         CHECK_EXT1(ARB_multisample, "GetFloatv");
          params[0] = BOOLEAN_TO_FLOAT(ctx->Multisample.SampleCoverageInvert);
          break;
       case GL_SAMPLE_BUFFERS_ARB:
-         CHECK_EXT1(ARB_multisample, "GetFloatv");
          params[0] = (GLfloat)(ctx->DrawBuffer->Visual.sampleBuffers);
          break;
       case GL_SAMPLES_ARB:
-         CHECK_EXT1(ARB_multisample, "GetFloatv");
          params[0] = (GLfloat)(ctx->DrawBuffer->Visual.samples);
          break;
       case GL_RASTER_POSITION_UNCLIPPED_IBM:
@@ -3440,43 +3407,33 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          params[0] = ctx->Const.MaxSpotExponent;
          break;
       case GL_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ArrayBufferObj->Name);
          break;
       case GL_VERTEX_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ArrayObj->Vertex.BufferObj->Name);
          break;
       case GL_NORMAL_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ArrayObj->Normal.BufferObj->Name);
          break;
       case GL_COLOR_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ArrayObj->Color.BufferObj->Name);
          break;
       case GL_INDEX_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ArrayObj->Index.BufferObj->Name);
          break;
       case GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ArrayObj->TexCoord[ctx->Array.ActiveTexture].BufferObj->Name);
          break;
       case GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ArrayObj->EdgeFlag.BufferObj->Name);
          break;
       case GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ArrayObj->SecondaryColor.BufferObj->Name);
          break;
       case GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ArrayObj->FogCoord.BufferObj->Name);
          break;
       case GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetFloatv");
          params[0] = (GLfloat)(ctx->Array.ElementArrayBufferObj->Name);
          break;
       case GL_PIXEL_PACK_BUFFER_BINDING_EXT:
@@ -3601,15 +3558,12 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          params[0] = (GLfloat)(ctx->VertexProgram.CurrentPosition);
          break;
       case GL_MAX_DRAW_BUFFERS_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetFloatv");
          params[0] = (GLfloat)(ctx->Const.MaxDrawBuffers);
          break;
       case GL_DRAW_BUFFER0_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetFloatv");
          params[0] = ENUM_TO_FLOAT(ctx->DrawBuffer->ColorDrawBuffer[0]);
          break;
       case GL_DRAW_BUFFER1_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetFloatv");
          {
          GLenum buffer;
          if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
@@ -3621,7 +3575,6 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          }
          break;
       case GL_DRAW_BUFFER2_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetFloatv");
          {
          GLenum buffer;
          if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
@@ -3633,7 +3586,6 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          }
          break;
       case GL_DRAW_BUFFER3_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetFloatv");
          {
          GLenum buffer;
          if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
@@ -3753,6 +3705,10 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
          CHECK_EXT1(ARB_shader_objects, "GetFloatv");
          params[0] = (GLfloat)(ctx->Shader.CurrentProgram ? ctx->Shader.CurrentProgram->Name : 0);
          break;
+      case GL_MAX_SAMPLES:
+         CHECK_EXT1(ARB_framebuffer_object, "GetFloatv");
+         params[0] = (GLfloat)(ctx->Const.MaxSamples);
+         break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetFloatv(pname=0x%x)", pname);
    }
@@ -3985,7 +3941,7 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          params[0] = ctx->DrawBuffer->Visual.depthBits;
          break;
       case GL_DEPTH_CLEAR_VALUE:
-         params[0] = FLOAT_TO_INT(ctx->Depth.Clear);
+         params[0] = FLOAT_TO_INT(((GLfloat) ctx->Depth.Clear));
          break;
       case GL_DEPTH_FUNC:
          params[0] = ENUM_TO_INT(ctx->Depth.Func);
@@ -4156,7 +4112,7 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          params[0] = ctx->List.ListBase;
          break;
       case GL_LIST_INDEX:
-         params[0] = ctx->ListState.CurrentListNum;
+         params[0] = (ctx->ListState.CurrentList ? ctx->ListState.CurrentList->Name : 0);
          break;
       case GL_LIST_MODE:
          {
@@ -4774,15 +4730,12 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          params[0] = (1 << (ctx->Const.MaxCubeTextureLevels - 1));
          break;
       case GL_TEXTURE_COMPRESSION_HINT_ARB:
-         CHECK_EXT1(ARB_texture_compression, "GetIntegerv");
          params[0] = ctx->Hint.TextureCompression;
          break;
       case GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB:
-         CHECK_EXT1(ARB_texture_compression, "GetIntegerv");
          params[0] = _mesa_get_compressed_formats(ctx, NULL, GL_FALSE);
          break;
       case GL_COMPRESSED_TEXTURE_FORMATS_ARB:
-         CHECK_EXT1(ARB_texture_compression, "GetIntegerv");
          {
          GLint formats[100];
          GLuint i, n = _mesa_get_compressed_formats(ctx, formats, GL_FALSE);
@@ -5064,35 +5017,27 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          params[0] = IROUND(ctx->Const.MaxTextureMaxAnisotropy);
          break;
       case GL_MULTISAMPLE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetIntegerv");
          params[0] = BOOLEAN_TO_INT(ctx->Multisample.Enabled);
          break;
       case GL_SAMPLE_ALPHA_TO_COVERAGE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetIntegerv");
          params[0] = BOOLEAN_TO_INT(ctx->Multisample.SampleAlphaToCoverage);
          break;
       case GL_SAMPLE_ALPHA_TO_ONE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetIntegerv");
          params[0] = BOOLEAN_TO_INT(ctx->Multisample.SampleAlphaToOne);
          break;
       case GL_SAMPLE_COVERAGE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetIntegerv");
          params[0] = BOOLEAN_TO_INT(ctx->Multisample.SampleCoverage);
          break;
       case GL_SAMPLE_COVERAGE_VALUE_ARB:
-         CHECK_EXT1(ARB_multisample, "GetIntegerv");
          params[0] = IROUND(ctx->Multisample.SampleCoverageValue);
          break;
       case GL_SAMPLE_COVERAGE_INVERT_ARB:
-         CHECK_EXT1(ARB_multisample, "GetIntegerv");
          params[0] = BOOLEAN_TO_INT(ctx->Multisample.SampleCoverageInvert);
          break;
       case GL_SAMPLE_BUFFERS_ARB:
-         CHECK_EXT1(ARB_multisample, "GetIntegerv");
          params[0] = ctx->DrawBuffer->Visual.sampleBuffers;
          break;
       case GL_SAMPLES_ARB:
-         CHECK_EXT1(ARB_multisample, "GetIntegerv");
          params[0] = ctx->DrawBuffer->Visual.samples;
          break;
       case GL_RASTER_POSITION_UNCLIPPED_IBM:
@@ -5288,43 +5233,33 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          params[0] = IROUND(ctx->Const.MaxSpotExponent);
          break;
       case GL_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetIntegerv");
          params[0] = ctx->Array.ArrayBufferObj->Name;
          break;
       case GL_VERTEX_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetIntegerv");
          params[0] = ctx->Array.ArrayObj->Vertex.BufferObj->Name;
          break;
       case GL_NORMAL_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetIntegerv");
          params[0] = ctx->Array.ArrayObj->Normal.BufferObj->Name;
          break;
       case GL_COLOR_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetIntegerv");
          params[0] = ctx->Array.ArrayObj->Color.BufferObj->Name;
          break;
       case GL_INDEX_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetIntegerv");
          params[0] = ctx->Array.ArrayObj->Index.BufferObj->Name;
          break;
       case GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetIntegerv");
          params[0] = ctx->Array.ArrayObj->TexCoord[ctx->Array.ActiveTexture].BufferObj->Name;
          break;
       case GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetIntegerv");
          params[0] = ctx->Array.ArrayObj->EdgeFlag.BufferObj->Name;
          break;
       case GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetIntegerv");
          params[0] = ctx->Array.ArrayObj->SecondaryColor.BufferObj->Name;
          break;
       case GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetIntegerv");
          params[0] = ctx->Array.ArrayObj->FogCoord.BufferObj->Name;
          break;
       case GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB:
-         CHECK_EXT1(ARB_vertex_buffer_object, "GetIntegerv");
          params[0] = ctx->Array.ElementArrayBufferObj->Name;
          break;
       case GL_PIXEL_PACK_BUFFER_BINDING_EXT:
@@ -5449,15 +5384,12 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          params[0] = ctx->VertexProgram.CurrentPosition;
          break;
       case GL_MAX_DRAW_BUFFERS_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetIntegerv");
          params[0] = ctx->Const.MaxDrawBuffers;
          break;
       case GL_DRAW_BUFFER0_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetIntegerv");
          params[0] = ENUM_TO_INT(ctx->DrawBuffer->ColorDrawBuffer[0]);
          break;
       case GL_DRAW_BUFFER1_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetIntegerv");
          {
          GLenum buffer;
          if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
@@ -5469,7 +5401,6 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          }
          break;
       case GL_DRAW_BUFFER2_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetIntegerv");
          {
          GLenum buffer;
          if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
@@ -5481,7 +5412,6 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          }
          break;
       case GL_DRAW_BUFFER3_ARB:
-         CHECK_EXT1(ARB_draw_buffers, "GetIntegerv");
          {
          GLenum buffer;
          if (pname - GL_DRAW_BUFFER0_ARB >= ctx->Const.MaxDrawBuffers) {
@@ -5601,6 +5531,10 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
          CHECK_EXT1(ARB_shader_objects, "GetIntegerv");
          params[0] = ctx->Shader.CurrentProgram ? ctx->Shader.CurrentProgram->Name : 0;
          break;
+      case GL_MAX_SAMPLES:
+         CHECK_EXT1(ARB_framebuffer_object, "GetIntegerv");
+         params[0] = ctx->Const.MaxSamples;
+         break;
       default:
          _mesa_error(ctx, GL_INVALID_ENUM, "glGetIntegerv(pname=0x%x)", pname);
    }
diff --git a/src/mesa/main/get_gen.py b/src/mesa/main/get_gen.py
index 152e378b4f..3b2496c663 100644
--- a/src/mesa/main/get_gen.py
+++ b/src/mesa/main/get_gen.py
@@ -180,7 +180,7 @@ StateVars = [
 	( "GL_DEPTH_BIAS", GLfloat, ["ctx->Pixel.DepthBias"], "", None ),
 	( "GL_DEPTH_BITS", GLint, ["ctx->DrawBuffer->Visual.depthBits"],
 	  "", None ),
-	( "GL_DEPTH_CLEAR_VALUE", GLfloatN, ["ctx->Depth.Clear"], "", None ),
+	( "GL_DEPTH_CLEAR_VALUE", GLfloatN, ["((GLfloat) ctx->Depth.Clear)"], "", None ),
 	( "GL_DEPTH_FUNC", GLenum, ["ctx->Depth.Func"], "", None ),
 	( "GL_DEPTH_RANGE", GLfloatN,
 	  [ "ctx->Viewport.Near", "ctx->Viewport.Far" ], "", None ),
@@ -254,7 +254,7 @@ StateVars = [
 	  ["ctx->Const.MinLineWidth",
 	   "ctx->Const.MaxLineWidth"], "", None ),
 	( "GL_LIST_BASE", GLint, ["ctx->List.ListBase"], "", None ),
-	( "GL_LIST_INDEX", GLint, ["ctx->ListState.CurrentListNum"], "", None ),
+	( "GL_LIST_INDEX", GLint, ["(ctx->ListState.CurrentList ? ctx->ListState.CurrentList->Name : 0)"], "", None ),
 	( "GL_LIST_MODE", GLenum, ["mode"],
 	  """GLenum mode;
          if (!ctx->CompileFlag)
@@ -523,10 +523,10 @@ StateVars = [
 
 	# GL_ARB_texture_compression */
 	( "GL_TEXTURE_COMPRESSION_HINT_ARB", GLint,
-	  ["ctx->Hint.TextureCompression"], "", ["ARB_texture_compression"] ),
+	  ["ctx->Hint.TextureCompression"], "", None ),
 	( "GL_NUM_COMPRESSED_TEXTURE_FORMATS_ARB", GLint,
 	  ["_mesa_get_compressed_formats(ctx, NULL, GL_FALSE)"],
-	  "", ["ARB_texture_compression"] ),
+	  "", None ),
 	( "GL_COMPRESSED_TEXTURE_FORMATS_ARB", GLenum,
 	  [],
 	  """GLint formats[100];
@@ -534,7 +534,7 @@ StateVars = [
          ASSERT(n <= 100);
          for (i = 0; i < n; i++)
             params[i] = CONVERSION(formats[i]);""",
-	  ["ARB_texture_compression"] ),
+	  None ),
 
 	# GL_EXT_compiled_vertex_array
 	( "GL_ARRAY_ELEMENT_LOCK_FIRST_EXT", GLint, ["ctx->Array.LockFirst"],
@@ -681,21 +681,21 @@ StateVars = [
 
 	# GL_ARB_multisample
 	( "GL_MULTISAMPLE_ARB", GLboolean,
-	  ["ctx->Multisample.Enabled"], "", ["ARB_multisample"] ),
+	  ["ctx->Multisample.Enabled"], "", None ),
 	( "GL_SAMPLE_ALPHA_TO_COVERAGE_ARB", GLboolean,
-	  ["ctx->Multisample.SampleAlphaToCoverage"], "", ["ARB_multisample"] ),
+	  ["ctx->Multisample.SampleAlphaToCoverage"], "", None ),
 	( "GL_SAMPLE_ALPHA_TO_ONE_ARB", GLboolean,
-	  ["ctx->Multisample.SampleAlphaToOne"], "", ["ARB_multisample"] ),
+	  ["ctx->Multisample.SampleAlphaToOne"], "", None ),
 	( "GL_SAMPLE_COVERAGE_ARB", GLboolean,
-	  ["ctx->Multisample.SampleCoverage"], "", ["ARB_multisample"] ),
+	  ["ctx->Multisample.SampleCoverage"], "", None ),
 	( "GL_SAMPLE_COVERAGE_VALUE_ARB", GLfloat,
-	  ["ctx->Multisample.SampleCoverageValue"], "", ["ARB_multisample"] ),
+	  ["ctx->Multisample.SampleCoverageValue"], "", None ),
 	( "GL_SAMPLE_COVERAGE_INVERT_ARB", GLboolean,
-	  ["ctx->Multisample.SampleCoverageInvert"], "", ["ARB_multisample"] ),
+	  ["ctx->Multisample.SampleCoverageInvert"], "", None ),
 	( "GL_SAMPLE_BUFFERS_ARB", GLint,
-	  ["ctx->DrawBuffer->Visual.sampleBuffers"], "", ["ARB_multisample"] ),
+	  ["ctx->DrawBuffer->Visual.sampleBuffers"], "", None ),
 	( "GL_SAMPLES_ARB", GLint,
-	  ["ctx->DrawBuffer->Visual.samples"], "", ["ARB_multisample"] ),
+	  ["ctx->DrawBuffer->Visual.samples"], "", None ),
 
 	# GL_IBM_rasterpos_clip
 	( "GL_RASTER_POSITION_UNCLIPPED_IBM", GLboolean,
@@ -815,30 +815,30 @@ StateVars = [
 
 	# GL_ARB_vertex_buffer_object
 	( "GL_ARRAY_BUFFER_BINDING_ARB", GLint,
-	  ["ctx->Array.ArrayBufferObj->Name"], "", ["ARB_vertex_buffer_object"] ),
+	  ["ctx->Array.ArrayBufferObj->Name"], "", None ),
 	( "GL_VERTEX_ARRAY_BUFFER_BINDING_ARB", GLint,
-	  ["ctx->Array.ArrayObj->Vertex.BufferObj->Name"], "", ["ARB_vertex_buffer_object"] ),
+	  ["ctx->Array.ArrayObj->Vertex.BufferObj->Name"], "", None ),
 	( "GL_NORMAL_ARRAY_BUFFER_BINDING_ARB", GLint,
-	  ["ctx->Array.ArrayObj->Normal.BufferObj->Name"], "", ["ARB_vertex_buffer_object"] ),
+	  ["ctx->Array.ArrayObj->Normal.BufferObj->Name"], "", None ),
 	( "GL_COLOR_ARRAY_BUFFER_BINDING_ARB", GLint,
-	  ["ctx->Array.ArrayObj->Color.BufferObj->Name"], "", ["ARB_vertex_buffer_object"] ),
+	  ["ctx->Array.ArrayObj->Color.BufferObj->Name"], "", None ),
 	( "GL_INDEX_ARRAY_BUFFER_BINDING_ARB", GLint,
-	  ["ctx->Array.ArrayObj->Index.BufferObj->Name"], "", ["ARB_vertex_buffer_object"] ),
+	  ["ctx->Array.ArrayObj->Index.BufferObj->Name"], "", None ),
 	( "GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB", GLint,
 	  ["ctx->Array.ArrayObj->TexCoord[ctx->Array.ActiveTexture].BufferObj->Name"],
-	  "", ["ARB_vertex_buffer_object"] ),
+	  "", None ),
 	( "GL_EDGE_FLAG_ARRAY_BUFFER_BINDING_ARB", GLint,
-	  ["ctx->Array.ArrayObj->EdgeFlag.BufferObj->Name"], "", ["ARB_vertex_buffer_object"] ),
+	  ["ctx->Array.ArrayObj->EdgeFlag.BufferObj->Name"], "", None ),
 	( "GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING_ARB", GLint,
 	  ["ctx->Array.ArrayObj->SecondaryColor.BufferObj->Name"],
-	  "", ["ARB_vertex_buffer_object"] ),
+	  "", None ),
 	( "GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING_ARB", GLint,
 	  ["ctx->Array.ArrayObj->FogCoord.BufferObj->Name"],
-	  "", ["ARB_vertex_buffer_object"] ),
+	  "", None ),
 	# GL_WEIGHT_ARRAY_BUFFER_BINDING_ARB - not supported
 	( "GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB", GLint,
 	  ["ctx->Array.ElementArrayBufferObj->Name"],
-	  "", ["ARB_vertex_buffer_object"] ),
+	  "", None ),
 
 	# GL_EXT_pixel_buffer_object
 	( "GL_PIXEL_PACK_BUFFER_BINDING_EXT", GLint,
@@ -914,9 +914,9 @@ StateVars = [
 
 	# GL_ARB_draw_buffers
 	( "GL_MAX_DRAW_BUFFERS_ARB", GLint,
-	  ["ctx->Const.MaxDrawBuffers"], "", ["ARB_draw_buffers"] ),
+	  ["ctx->Const.MaxDrawBuffers"], "", None ),
 	( "GL_DRAW_BUFFER0_ARB", GLenum,
-	  ["ctx->DrawBuffer->ColorDrawBuffer[0]"], "", ["ARB_draw_buffers"] ),
+	  ["ctx->DrawBuffer->ColorDrawBuffer[0]"], "", None ),
 	( "GL_DRAW_BUFFER1_ARB", GLenum,
 	  ["buffer"],
 	  """GLenum buffer;
@@ -924,7 +924,7 @@ StateVars = [
             _mesa_error(ctx, GL_INVALID_ENUM, "glGet(GL_DRAW_BUFFERx_ARB)");
             return;
          }
-         buffer = ctx->DrawBuffer->ColorDrawBuffer[1];""", ["ARB_draw_buffers"] ),
+         buffer = ctx->DrawBuffer->ColorDrawBuffer[1];""", None ),
 	( "GL_DRAW_BUFFER2_ARB", GLenum,
 	  ["buffer"],
 	  """GLenum buffer;
@@ -932,7 +932,7 @@ StateVars = [
             _mesa_error(ctx, GL_INVALID_ENUM, "glGet(GL_DRAW_BUFFERx_ARB)");
             return;
          }
-         buffer = ctx->DrawBuffer->ColorDrawBuffer[2];""", ["ARB_draw_buffers"] ),
+         buffer = ctx->DrawBuffer->ColorDrawBuffer[2];""", None ),
 	( "GL_DRAW_BUFFER3_ARB", GLenum,
 	  ["buffer"],
 	  """GLenum buffer;
@@ -940,7 +940,7 @@ StateVars = [
             _mesa_error(ctx, GL_INVALID_ENUM, "glGet(GL_DRAW_BUFFERx_ARB)");
             return;
          }
-         buffer = ctx->DrawBuffer->ColorDrawBuffer[3];""", ["ARB_draw_buffers"] ),
+         buffer = ctx->DrawBuffer->ColorDrawBuffer[3];""", None ),
 	# XXX Add more GL_DRAW_BUFFERn_ARB entries as needed in the future
 
 	# GL_OES_read_format
@@ -1009,7 +1009,11 @@ StateVars = [
 	# close enough for now.
 	( "GL_CURRENT_PROGRAM", GLint,
 	  ["ctx->Shader.CurrentProgram ? ctx->Shader.CurrentProgram->Name : 0"],
-	  "", ["ARB_shader_objects"] )
+	  "", ["ARB_shader_objects"] ),
+
+	# GL_ARB_framebuffer_object
+	( "GL_MAX_SAMPLES", GLint, ["ctx->Const.MaxSamples"], "",
+	  ["ARB_framebuffer_object"] )
 ]
 
 
diff --git a/src/mesa/main/getstring.c b/src/mesa/main/getstring.c
index 94bf5de1e8..a9e22d340a 100644
--- a/src/mesa/main/getstring.c
+++ b/src/mesa/main/getstring.c
@@ -82,7 +82,16 @@ compute_version(const GLcontext *ctx)
                               ctx->Extensions.ARB_vertex_shader &&
                               ctx->Extensions.ARB_fragment_shader &&
                               ctx->Extensions.ARB_texture_non_power_of_two &&
-                              ctx->Extensions.EXT_blend_equation_separate);
+                              ctx->Extensions.EXT_blend_equation_separate &&
+
+			      /* Technically, 2.0 requires the functionality
+			       * of the EXT version.  Enable 2.0 if either
+			       * extension is available, and assume that a
+			       * driver that only exposes the ATI extension
+			       * will fallback to software when necessary.
+			       */
+			      (ctx->Extensions.EXT_stencil_two_side
+			       || ctx->Extensions.ATI_separate_stencil));
    const GLboolean ver_2_1 = (ver_2_0 &&
                               /*ctx->Extensions.ARB_shading_language_120 &&*/
                               ctx->Extensions.EXT_pixel_buffer_object &&
diff --git a/src/mesa/main/glheader.h b/src/mesa/main/glheader.h
index 1d0f178dc4..5657976711 100644
--- a/src/mesa/main/glheader.h
+++ b/src/mesa/main/glheader.h
@@ -64,22 +64,33 @@
 #include <stdarg.h>
 
 
-/* Get typedefs for uintptr_t and friends */
-#if defined(__MINGW32__) || defined(__NetBSD__)
-#  include <stdint.h>
-#elif defined(_WIN32)
-#  include <BaseTsd.h>
-#  if _MSC_VER == 1200
-     typedef UINT_PTR uintptr_t;
-#  endif 
-#elif defined(__INTERIX)
-/* Interix 3.x has a gcc that shadows this. */
-#  ifndef _UINTPTR_T_DEFINED
-     typedef unsigned long uintptr_t;
-#  define _UINTPTR_T_DEFINED
+/* Get standard integer types */
+#if defined(_MSC_VER)
+
+   typedef __int8             int8_t;
+   typedef unsigned __int8    uint8_t;
+   typedef __int16            int16_t;
+   typedef unsigned __int16   uint16_t;
+#  ifndef __eglplatform_h_
+     typedef __int32            int32_t;
+#  endif
+   typedef unsigned __int32   uint32_t;
+   typedef __int64            int64_t;
+   typedef unsigned __int64   uint64_t;
+
+#  if defined(_WIN64)
+     typedef __int64            intptr_t;
+     typedef unsigned __int64   uintptr_t;
+#  else
+     typedef __int32            intptr_t;
+     typedef unsigned __int32   uintptr_t;
 #  endif
+
+#  define INT64_C(__val) __val##i64
+#  define UINT64_C(__val) __val##ui64
+
 #else
-#  include <inttypes.h>
+#  include <stdint.h>
 #endif
 
 
@@ -146,7 +157,8 @@
 #include <byteswap.h>
 #define CPU_TO_LE32( x )	bswap_32( x )
 #else /*__linux__*/
-#define CPU_TO_LE32( x )	( x )  /* fix me for non-Linux big-endian! */
+#include <sys/endian.h>
+#define CPU_TO_LE32( x )	bswap32( x )
 #endif /*__linux__*/
 #define MESA_BIG_ENDIAN 1
 #else
diff --git a/src/mesa/main/hint.c b/src/mesa/main/hint.c
index dcfa9c7363..e2d4129a38 100644
--- a/src/mesa/main/hint.c
+++ b/src/mesa/main/hint.c
@@ -89,10 +89,6 @@ _mesa_Hint( GLenum target, GLenum mode )
 
       /* GL_ARB_texture_compression */
       case GL_TEXTURE_COMPRESSION_HINT_ARB:
-         if (!ctx->Extensions.ARB_texture_compression) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glHint(target)");
-	    return;
-         }
 	 if (ctx->Hint.TextureCompression == mode)
 	    return;
 	 FLUSH_VERTICES(ctx, _NEW_HINT);
diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c
index 1a6e864b98..4d86c54777 100644
--- a/src/mesa/main/image.c
+++ b/src/mesa/main/image.c
@@ -61,7 +61,7 @@
 /**
  * \return GL_TRUE if type is packed pixel type, GL_FALSE otherwise.
  */
-static GLboolean
+GLboolean
 _mesa_type_is_packed(GLenum type)
 {
    switch (type) {
@@ -1689,7 +1689,7 @@ _mesa_pack_rgba_span_float(GLcontext *ctx, GLuint n, GLfloat rgba[][4],
 
    if (dstFormat == GL_LUMINANCE || dstFormat == GL_LUMINANCE_ALPHA) {
       /* compute luminance values */
-      if (dstType != GL_FLOAT || ctx->Color.ClampReadColor == GL_TRUE) {
+      if (transferOps & IMAGE_CLAMP_BIT) {
          for (i = 0; i < n; i++) {
             GLfloat sum = rgba[i][RCOMP] + rgba[i][GCOMP] + rgba[i][BCOMP];
             luminance[i] = CLAMP(sum, 0.0F, 1.0F);
@@ -5119,7 +5119,7 @@ _mesa_clip_copytexsubimage(const GLcontext *ctx,
    const struct gl_framebuffer *fb = ctx->ReadBuffer;
    const GLint srcX0 = *srcX, srcY0 = *srcY;
 
-   if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax,
+   if (_mesa_clip_to_region(0, 0, fb->Width, fb->Height,
                             srcX, srcY, width, height)) {
       *destX = *destX + *srcX - srcX0;
       *destY = *destY + *srcY - srcY0;
@@ -5152,7 +5152,7 @@ _mesa_clip_to_region(GLint xmin, GLint ymin,
 
    /* right clipping */
    if (*x + *width > xmax)
-      *width -= (*x + *width - xmax - 1);
+      *width -= (*x + *width - xmax);
 
    if (*width <= 0)
       return GL_FALSE;
@@ -5165,7 +5165,7 @@ _mesa_clip_to_region(GLint xmin, GLint ymin,
 
    /* top (or bottom) clipping */
    if (*y + *height > ymax)
-      *height -= (*y + *height - ymax - 1);
+      *height -= (*y + *height - ymax);
 
    if (*height <= 0)
       return GL_FALSE;
diff --git a/src/mesa/main/image.h b/src/mesa/main/image.h
index 38e1374c20..0e0bbd96d8 100644
--- a/src/mesa/main/image.h
+++ b/src/mesa/main/image.h
@@ -36,6 +36,9 @@ _mesa_swap2( GLushort *p, GLuint n );
 extern void
 _mesa_swap4( GLuint *p, GLuint n );
 
+extern GLboolean
+_mesa_type_is_packed(GLenum type);
+
 extern GLint
 _mesa_sizeof_type( GLenum type );
 
diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c
index 13cb84ca4b..cb04594c1f 100644
--- a/src/mesa/main/imports.c
+++ b/src/mesa/main/imports.c
@@ -557,7 +557,7 @@ _mesa_pow(double x, double y)
  * Find the first bit set in a word.
  */
 int
-_mesa_ffs(int i)
+_mesa_ffs(int32_t i)
 {
 #if (defined(_WIN32) ) || defined(__IBMC__) || defined(__IBMCPP__)
    register int bit = 0;
@@ -594,11 +594,7 @@ _mesa_ffs(int i)
  *          if no bits set.
  */
 int
-#ifdef __MINGW32__
-_mesa_ffsll(long val)
-#else
-_mesa_ffsll(long long val)
-#endif
+_mesa_ffsll(int64_t val)
 {
 #ifdef ffsll
    return ffsll(val);
@@ -607,11 +603,11 @@ _mesa_ffsll(long long val)
 
    assert(sizeof(val) == 8);
 
-   bit = _mesa_ffs(val);
+   bit = _mesa_ffs((int32_t)val);
    if (bit != 0)
       return bit;
 
-   bit = _mesa_ffs(val >> 32);
+   bit = _mesa_ffs((int32_t)(val >> 32));
    if (bit != 0)
       return 32 + bit;
 
@@ -934,18 +930,41 @@ _mesa_sprintf( char *str, const char *fmt, ... )
    return r;
 }
 
+/** Wrapper around vsnprintf() */
+int
+_mesa_snprintf( char *str, size_t size, const char *fmt, ... )
+{
+   int r;
+   va_list args;
+   va_start( args, fmt );  
+   r = vsnprintf( str, size, fmt, args );
+   va_end( args );
+   return r;
+}
+
 /** Wrapper around printf(), using vsprintf() for the formatting. */
 void
 _mesa_printf( const char *fmtString, ... )
 {
+   va_list args;
+   va_start( args, fmtString );  
+   vfprintf(stderr, fmtString, args);
+   va_end( args );
+}
+
+/** Wrapper around fprintf(), using vsprintf() for the formatting. */
+void
+_mesa_fprintf( FILE *f, const char *fmtString, ... )
+{
    char s[MAXSTRING];
    va_list args;
    va_start( args, fmtString );  
    vsnprintf(s, MAXSTRING, fmtString, args);
    va_end( args );
-   fprintf(stderr,"%s", s);
+   fprintf(f, "%s", s);
 }
 
+
 /** Wrapper around vsprintf() */
 int
 _mesa_vsprintf( char *str, const char *fmt, va_list args )
diff --git a/src/mesa/main/imports.h b/src/mesa/main/imports.h
index 453151ce0b..a421eb5e75 100644
--- a/src/mesa/main/imports.h
+++ b/src/mesa/main/imports.h
@@ -708,14 +708,10 @@ extern double
 _mesa_pow(double x, double y);
 
 extern int
-_mesa_ffs(int i);
+_mesa_ffs(int32_t i);
 
 extern int
-#ifdef __MINGW32__
-_mesa_ffsll(long i);
-#else
-_mesa_ffsll(long long i);
-#endif
+_mesa_ffsll(int64_t i);
 
 extern unsigned int
 _mesa_bitcount(unsigned int n);
@@ -767,9 +763,15 @@ _mesa_strtod( const char *s, char **end );
 extern int
 _mesa_sprintf( char *str, const char *fmt, ... );
 
+extern int
+_mesa_snprintf( char *str, size_t size, const char *fmt, ... );
+
 extern void
 _mesa_printf( const char *fmtString, ... );
 
+extern void
+_mesa_fprintf( FILE *f, const char *fmtString, ... );
+
 extern int 
 _mesa_vsprintf( char *str, const char *fmt, va_list args );
 
diff --git a/src/mesa/main/light.c b/src/mesa/main/light.c
index 10ee088a2d..a15c1f0be0 100644
--- a/src/mesa/main/light.c
+++ b/src/mesa/main/light.c
@@ -208,7 +208,8 @@ _mesa_Lightfv( GLenum light, GLenum pname, const GLfloat *params )
       if (_math_matrix_is_dirty(ctx->ModelviewMatrixStack.Top)) {
 	 _math_matrix_analyse(ctx->ModelviewMatrixStack.Top);
       }
-      TRANSFORM_NORMAL(temp, params, ctx->ModelviewMatrixStack.Top->inv);
+      TRANSFORM_DIRECTION(temp, params, ctx->ModelviewMatrixStack.Top->m);
+      NORMALIZE_3FV(temp);
       params = temp;
       break;
    case GL_SPOT_EXPONENT:
@@ -1367,6 +1368,7 @@ _mesa_init_lighting( GLcontext *ctx )
    /* Miscellaneous */
    ctx->Light._NeedEyeCoords = GL_FALSE;
    ctx->_NeedEyeCoords = GL_FALSE;
+   ctx->_ForceEyeCoords = GL_FALSE;
    ctx->_ModelViewInvScale = 1.0;
 }
 
diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h
index 3819da3d68..8fb32dd7e9 100644
--- a/src/mesa/main/mfeatures.h
+++ b/src/mesa/main/mfeatures.h
@@ -60,6 +60,7 @@
 
 #define FEATURE_ARB_occlusion_query  _HAVE_FULL_GL
 #define FEATURE_ARB_fragment_program  _HAVE_FULL_GL
+#define FEATURE_ARB_framebuffer_object  _HAVE_FULL_GL
 #define FEATURE_ARB_vertex_buffer_object  _HAVE_FULL_GL
 #define FEATURE_ARB_vertex_program  _HAVE_FULL_GL
 #define FEATURE_ARB_vertex_shader _HAVE_FULL_GL
diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c
index 13d90e78fe..3dd4b3391b 100644
--- a/src/mesa/main/mipmap.c
+++ b/src/mesa/main/mipmap.c
@@ -41,11 +41,76 @@ bytes_per_pixel(GLenum datatype, GLuint comps)
 {
    GLint b = _mesa_sizeof_packed_type(datatype);
    assert(b >= 0);
-   return b * comps;
+
+   if (_mesa_type_is_packed(datatype))
+       return b;
+   else
+       return b * comps;
 }
 
 
 /**
+ * \name Support macros for do_row and do_row_3d
+ *
+ * The macro madness is here for two reasons.  First, it compacts the code
+ * slightly.  Second, it makes it much easier to adjust the specifics of the
+ * filter to tune the rounding characteristics.
+ */
+/*@{*/
+#define DECLARE_ROW_POINTERS(t, e) \
+      const t(*rowA)[e] = (const t(*)[e]) srcRowA; \
+      const t(*rowB)[e] = (const t(*)[e]) srcRowB; \
+      const t(*rowC)[e] = (const t(*)[e]) srcRowC; \
+      const t(*rowD)[e] = (const t(*)[e]) srcRowD; \
+      t(*dst)[e] = (t(*)[e]) dstRow
+
+#define DECLARE_ROW_POINTERS0(t) \
+      const t *rowA = (const t *) srcRowA; \
+      const t *rowB = (const t *) srcRowB; \
+      const t *rowC = (const t *) srcRowC; \
+      const t *rowD = (const t *) srcRowD; \
+      t *dst = (t *) dstRow
+
+#define FILTER_SUM_3D(Aj, Ak, Bj, Bk, Cj, Ck, Dj, Dk) \
+   ((unsigned) Aj + (unsigned) Ak \
+    + (unsigned) Bj + (unsigned) Bk \
+    + (unsigned) Cj + (unsigned) Ck \
+    + (unsigned) Dj + (unsigned) Dk \
+    + 4) >> 3
+
+#define FILTER_3D(e) \
+   do { \
+      dst[i][e] = FILTER_SUM_3D(rowA[j][e], rowA[k][e], \
+                                rowB[j][e], rowB[k][e], \
+                                rowC[j][e], rowC[k][e], \
+                                rowD[j][e], rowD[k][e]); \
+   } while(0)
+   
+#define FILTER_F_3D(e) \
+   do { \
+      dst[i][e] = (rowA[j][e] + rowA[k][e] \
+                   + rowB[j][e] + rowB[k][e] \
+                   + rowC[j][e] + rowC[k][e] \
+                   + rowD[j][e] + rowD[k][e]) * 0.125F; \
+   } while(0)
+
+#define FILTER_HF_3D(e) \
+   do { \
+      const GLfloat aj = _mesa_half_to_float(rowA[j][e]); \
+      const GLfloat ak = _mesa_half_to_float(rowA[k][e]); \
+      const GLfloat bj = _mesa_half_to_float(rowB[j][e]); \
+      const GLfloat bk = _mesa_half_to_float(rowB[k][e]); \
+      const GLfloat cj = _mesa_half_to_float(rowC[j][e]); \
+      const GLfloat ck = _mesa_half_to_float(rowC[k][e]); \
+      const GLfloat dj = _mesa_half_to_float(rowD[j][e]); \
+      const GLfloat dk = _mesa_half_to_float(rowD[k][e]); \
+      dst[i][e] = _mesa_float_to_half((aj + ak + bj + bk + cj + ck + dj + dk) \
+                                      * 0.125F); \
+   } while(0)
+/*@}*/
+
+
+/**
  * Average together two rows of a source image to produce a single new
  * row in the dest image.  It's legal for the two source rows to point
  * to the same data.  The source width must be equal to either the
@@ -361,7 +426,7 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth,
          const GLint rowAr0 = rowA[j] & 0x1f;
          const GLint rowAr1 = rowA[k] & 0x1f;
          const GLint rowBr0 = rowB[j] & 0x1f;
-         const GLint rowBr1 = rowB[k] & 0xf;
+         const GLint rowBr1 = rowB[k] & 0x1f;
          const GLint rowAg0 = (rowA[j] >> 5) & 0x1f;
          const GLint rowAg1 = (rowA[k] >> 5) & 0x1f;
          const GLint rowBg0 = (rowB[j] >> 5) & 0x1f;
@@ -412,6 +477,384 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth,
 }
 
 
+/**
+ * Average together four rows of a source image to produce a single new
+ * row in the dest image.  It's legal for the two source rows to point
+ * to the same data.  The source width must be equal to either the
+ * dest width or two times the dest width.
+ *
+ * \param datatype  GL pixel type \c GL_UNSIGNED_BYTE, \c GL_UNSIGNED_SHORT,
+ *                  \c GL_FLOAT, etc.
+ * \param comps     number of components per pixel (1..4)
+ * \param srcWidth  Width of a row in the source data
+ * \param srcRowA   Pointer to one of the rows of source data
+ * \param srcRowB   Pointer to one of the rows of source data
+ * \param srcRowC   Pointer to one of the rows of source data
+ * \param srcRowD   Pointer to one of the rows of source data
+ * \param dstWidth  Width of a row in the destination data
+ * \param srcRowA   Pointer to the row of destination data
+ */
+static void
+do_row_3D(GLenum datatype, GLuint comps, GLint srcWidth,
+          const GLvoid *srcRowA, const GLvoid *srcRowB,
+          const GLvoid *srcRowC, const GLvoid *srcRowD,
+          GLint dstWidth, GLvoid *dstRow)
+{
+   const GLuint k0 = (srcWidth == dstWidth) ? 0 : 1;
+   const GLuint colStride = (srcWidth == dstWidth) ? 1 : 2;
+   GLuint i, j, k;
+
+   ASSERT(comps >= 1);
+   ASSERT(comps <= 4);
+
+   if ((datatype == GL_UNSIGNED_BYTE) && (comps == 4)) {
+      DECLARE_ROW_POINTERS(GLubyte, 4);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+         FILTER_3D(2);
+         FILTER_3D(3);
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_BYTE) && (comps == 3)) {
+      DECLARE_ROW_POINTERS(GLubyte, 3);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+         FILTER_3D(2);
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_BYTE) && (comps == 2)) {
+      DECLARE_ROW_POINTERS(GLubyte, 2);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_BYTE) && (comps == 1)) {
+      DECLARE_ROW_POINTERS(GLubyte, 1);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_SHORT) && (comps == 4)) {
+      DECLARE_ROW_POINTERS(GLushort, 4);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+         FILTER_3D(2);
+         FILTER_3D(3);
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_SHORT) && (comps == 3)) {
+      DECLARE_ROW_POINTERS(GLushort, 3);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+         FILTER_3D(2);
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_SHORT) && (comps == 2)) {
+      DECLARE_ROW_POINTERS(GLushort, 2);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+         FILTER_3D(1);
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_SHORT) && (comps == 1)) {
+      DECLARE_ROW_POINTERS(GLushort, 1);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_3D(0);
+      }
+   }
+   else if ((datatype == GL_FLOAT) && (comps == 4)) {
+      DECLARE_ROW_POINTERS(GLfloat, 4);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_F_3D(0);
+         FILTER_F_3D(1);
+         FILTER_F_3D(2);
+         FILTER_F_3D(3);
+      }
+   }
+   else if ((datatype == GL_FLOAT) && (comps == 3)) {
+      DECLARE_ROW_POINTERS(GLfloat, 3);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_F_3D(0);
+         FILTER_F_3D(1);
+         FILTER_F_3D(2);
+      }
+   }
+   else if ((datatype == GL_FLOAT) && (comps == 2)) {
+      DECLARE_ROW_POINTERS(GLfloat, 2);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_F_3D(0);
+         FILTER_F_3D(1);
+      }
+   }
+   else if ((datatype == GL_FLOAT) && (comps == 1)) {
+      DECLARE_ROW_POINTERS(GLfloat, 1);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_F_3D(0);
+      }
+   }
+   else if ((datatype == GL_HALF_FLOAT_ARB) && (comps == 4)) {
+      DECLARE_ROW_POINTERS(GLhalfARB, 4);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_HF_3D(0);
+         FILTER_HF_3D(1);
+         FILTER_HF_3D(2);
+         FILTER_HF_3D(3);
+      }
+   }
+   else if ((datatype == GL_HALF_FLOAT_ARB) && (comps == 3)) {
+      DECLARE_ROW_POINTERS(GLhalfARB, 4);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_HF_3D(0);
+         FILTER_HF_3D(1);
+         FILTER_HF_3D(2);
+      }
+   }
+   else if ((datatype == GL_HALF_FLOAT_ARB) && (comps == 2)) {
+      DECLARE_ROW_POINTERS(GLhalfARB, 4);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_HF_3D(0);
+         FILTER_HF_3D(1);
+      }
+   }
+   else if ((datatype == GL_HALF_FLOAT_ARB) && (comps == 1)) {
+      DECLARE_ROW_POINTERS(GLhalfARB, 4);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         FILTER_HF_3D(0);
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_INT) && (comps == 1)) {
+      const GLuint *rowA = (const GLuint *) srcRowA;
+      const GLuint *rowB = (const GLuint *) srcRowB;
+      const GLuint *rowC = (const GLuint *) srcRowC;
+      const GLuint *rowD = (const GLuint *) srcRowD;
+      GLfloat *dst = (GLfloat *) dstRow;
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const uint64_t tmp = (((uint64_t) rowA[j] + (uint64_t) rowA[k])
+                               + ((uint64_t) rowB[j] + (uint64_t) rowB[k])
+                               + ((uint64_t) rowC[j] + (uint64_t) rowC[k])
+                               + ((uint64_t) rowD[j] + (uint64_t) rowD[k]));
+         dst[i] = (GLfloat)((double) tmp * 0.125);
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_SHORT_5_6_5) && (comps == 3)) {
+      DECLARE_ROW_POINTERS0(GLushort);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const GLint rowAr0 = rowA[j] & 0x1f;
+         const GLint rowAr1 = rowA[k] & 0x1f;
+         const GLint rowBr0 = rowB[j] & 0x1f;
+         const GLint rowBr1 = rowB[k] & 0x1f;
+         const GLint rowCr0 = rowC[j] & 0x1f;
+         const GLint rowCr1 = rowC[k] & 0x1f;
+         const GLint rowDr0 = rowD[j] & 0x1f;
+         const GLint rowDr1 = rowD[k] & 0x1f;
+         const GLint rowAg0 = (rowA[j] >> 5) & 0x3f;
+         const GLint rowAg1 = (rowA[k] >> 5) & 0x3f;
+         const GLint rowBg0 = (rowB[j] >> 5) & 0x3f;
+         const GLint rowBg1 = (rowB[k] >> 5) & 0x3f;
+         const GLint rowCg0 = (rowC[j] >> 5) & 0x3f;
+         const GLint rowCg1 = (rowC[k] >> 5) & 0x3f;
+         const GLint rowDg0 = (rowD[j] >> 5) & 0x3f;
+         const GLint rowDg1 = (rowD[k] >> 5) & 0x3f;
+         const GLint rowAb0 = (rowA[j] >> 11) & 0x1f;
+         const GLint rowAb1 = (rowA[k] >> 11) & 0x1f;
+         const GLint rowBb0 = (rowB[j] >> 11) & 0x1f;
+         const GLint rowBb1 = (rowB[k] >> 11) & 0x1f;
+         const GLint rowCb0 = (rowC[j] >> 11) & 0x1f;
+         const GLint rowCb1 = (rowC[k] >> 11) & 0x1f;
+         const GLint rowDb0 = (rowD[j] >> 11) & 0x1f;
+         const GLint rowDb1 = (rowD[k] >> 11) & 0x1f;
+         const GLint r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
+                                       rowCr0, rowCr1, rowDr0, rowDr1);
+         const GLint g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
+                                       rowCg0, rowCg1, rowDg0, rowDg1);
+         const GLint b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
+                                       rowCb0, rowCb1, rowDb0, rowDb1);
+         dst[i] = (b << 11) | (g << 5) | r;
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_SHORT_4_4_4_4) && (comps == 4)) {
+      DECLARE_ROW_POINTERS0(GLushort);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const GLint rowAr0 = rowA[j] & 0xf;
+         const GLint rowAr1 = rowA[k] & 0xf;
+         const GLint rowBr0 = rowB[j] & 0xf;
+         const GLint rowBr1 = rowB[k] & 0xf;
+         const GLint rowCr0 = rowC[j] & 0xf;
+         const GLint rowCr1 = rowC[k] & 0xf;
+         const GLint rowDr0 = rowD[j] & 0xf;
+         const GLint rowDr1 = rowD[k] & 0xf;
+         const GLint rowAg0 = (rowA[j] >> 4) & 0xf;
+         const GLint rowAg1 = (rowA[k] >> 4) & 0xf;
+         const GLint rowBg0 = (rowB[j] >> 4) & 0xf;
+         const GLint rowBg1 = (rowB[k] >> 4) & 0xf;
+         const GLint rowCg0 = (rowC[j] >> 4) & 0xf;
+         const GLint rowCg1 = (rowC[k] >> 4) & 0xf;
+         const GLint rowDg0 = (rowD[j] >> 4) & 0xf;
+         const GLint rowDg1 = (rowD[k] >> 4) & 0xf;
+         const GLint rowAb0 = (rowA[j] >> 8) & 0xf;
+         const GLint rowAb1 = (rowA[k] >> 8) & 0xf;
+         const GLint rowBb0 = (rowB[j] >> 8) & 0xf;
+         const GLint rowBb1 = (rowB[k] >> 8) & 0xf;
+         const GLint rowCb0 = (rowC[j] >> 8) & 0xf;
+         const GLint rowCb1 = (rowC[k] >> 8) & 0xf;
+         const GLint rowDb0 = (rowD[j] >> 8) & 0xf;
+         const GLint rowDb1 = (rowD[k] >> 8) & 0xf;
+         const GLint rowAa0 = (rowA[j] >> 12) & 0xf;
+         const GLint rowAa1 = (rowA[k] >> 12) & 0xf;
+         const GLint rowBa0 = (rowB[j] >> 12) & 0xf;
+         const GLint rowBa1 = (rowB[k] >> 12) & 0xf;
+         const GLint rowCa0 = (rowC[j] >> 12) & 0xf;
+         const GLint rowCa1 = (rowC[k] >> 12) & 0xf;
+         const GLint rowDa0 = (rowD[j] >> 12) & 0xf;
+         const GLint rowDa1 = (rowD[k] >> 12) & 0xf;
+         const GLint r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
+                                       rowCr0, rowCr1, rowDr0, rowDr1);
+         const GLint g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
+                                       rowCg0, rowCg1, rowDg0, rowDg1);
+         const GLint b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
+                                       rowCb0, rowCb1, rowDb0, rowDb1);
+         const GLint a = FILTER_SUM_3D(rowAa0, rowAa1, rowBa0, rowBa1,
+                                       rowCa0, rowCa1, rowDa0, rowDa1);
+
+         dst[i] = (a << 12) | (b << 8) | (g << 4) | r;
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_SHORT_1_5_5_5_REV) && (comps == 4)) {
+      DECLARE_ROW_POINTERS0(GLushort);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const GLint rowAr0 = rowA[j] & 0x1f;
+         const GLint rowAr1 = rowA[k] & 0x1f;
+         const GLint rowBr0 = rowB[j] & 0x1f;
+         const GLint rowBr1 = rowB[k] & 0x1f;
+         const GLint rowCr0 = rowC[j] & 0x1f;
+         const GLint rowCr1 = rowC[k] & 0x1f;
+         const GLint rowDr0 = rowD[j] & 0x1f;
+         const GLint rowDr1 = rowD[k] & 0x1f;
+         const GLint rowAg0 = (rowA[j] >> 5) & 0x1f;
+         const GLint rowAg1 = (rowA[k] >> 5) & 0x1f;
+         const GLint rowBg0 = (rowB[j] >> 5) & 0x1f;
+         const GLint rowBg1 = (rowB[k] >> 5) & 0x1f;
+         const GLint rowCg0 = (rowC[j] >> 5) & 0x1f;
+         const GLint rowCg1 = (rowC[k] >> 5) & 0x1f;
+         const GLint rowDg0 = (rowD[j] >> 5) & 0x1f;
+         const GLint rowDg1 = (rowD[k] >> 5) & 0x1f;
+         const GLint rowAb0 = (rowA[j] >> 10) & 0x1f;
+         const GLint rowAb1 = (rowA[k] >> 10) & 0x1f;
+         const GLint rowBb0 = (rowB[j] >> 10) & 0x1f;
+         const GLint rowBb1 = (rowB[k] >> 10) & 0x1f;
+         const GLint rowCb0 = (rowC[j] >> 10) & 0x1f;
+         const GLint rowCb1 = (rowC[k] >> 10) & 0x1f;
+         const GLint rowDb0 = (rowD[j] >> 10) & 0x1f;
+         const GLint rowDb1 = (rowD[k] >> 10) & 0x1f;
+         const GLint rowAa0 = (rowA[j] >> 15) & 0x1;
+         const GLint rowAa1 = (rowA[k] >> 15) & 0x1;
+         const GLint rowBa0 = (rowB[j] >> 15) & 0x1;
+         const GLint rowBa1 = (rowB[k] >> 15) & 0x1;
+         const GLint rowCa0 = (rowC[j] >> 15) & 0x1;
+         const GLint rowCa1 = (rowC[k] >> 15) & 0x1;
+         const GLint rowDa0 = (rowD[j] >> 15) & 0x1;
+         const GLint rowDa1 = (rowD[k] >> 15) & 0x1;
+         const GLint r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
+                                       rowCr0, rowCr1, rowDr0, rowDr1);
+         const GLint g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
+                                       rowCg0, rowCg1, rowDg0, rowDg1);
+         const GLint b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
+                                       rowCb0, rowCb1, rowDb0, rowDb1);
+         const GLint a = FILTER_SUM_3D(rowAa0, rowAa1, rowBa0, rowBa1,
+                                       rowCa0, rowCa1, rowDa0, rowDa1);
+
+         dst[i] = (a << 15) | (b << 10) | (g << 5) | r;
+      }
+   }
+   else if ((datatype == GL_UNSIGNED_BYTE_3_3_2) && (comps == 3)) {
+      DECLARE_ROW_POINTERS0(GLushort);
+
+      for (i = j = 0, k = k0; i < (GLuint) dstWidth;
+           i++, j += colStride, k += colStride) {
+         const GLint rowAr0 = rowA[j] & 0x3;
+         const GLint rowAr1 = rowA[k] & 0x3;
+         const GLint rowBr0 = rowB[j] & 0x3;
+         const GLint rowBr1 = rowB[k] & 0x3;
+         const GLint rowCr0 = rowC[j] & 0x3;
+         const GLint rowCr1 = rowC[k] & 0x3;
+         const GLint rowDr0 = rowD[j] & 0x3;
+         const GLint rowDr1 = rowD[k] & 0x3;
+         const GLint rowAg0 = (rowA[j] >> 2) & 0x7;
+         const GLint rowAg1 = (rowA[k] >> 2) & 0x7;
+         const GLint rowBg0 = (rowB[j] >> 2) & 0x7;
+         const GLint rowBg1 = (rowB[k] >> 2) & 0x7;
+         const GLint rowCg0 = (rowC[j] >> 2) & 0x7;
+         const GLint rowCg1 = (rowC[k] >> 2) & 0x7;
+         const GLint rowDg0 = (rowD[j] >> 2) & 0x7;
+         const GLint rowDg1 = (rowD[k] >> 2) & 0x7;
+         const GLint rowAb0 = (rowA[j] >> 5) & 0x7;
+         const GLint rowAb1 = (rowA[k] >> 5) & 0x7;
+         const GLint rowBb0 = (rowB[j] >> 5) & 0x7;
+         const GLint rowBb1 = (rowB[k] >> 5) & 0x7;
+         const GLint rowCb0 = (rowC[j] >> 5) & 0x7;
+         const GLint rowCb1 = (rowC[k] >> 5) & 0x7;
+         const GLint rowDb0 = (rowD[j] >> 5) & 0x7;
+         const GLint rowDb1 = (rowD[k] >> 5) & 0x7;
+         const GLint r = FILTER_SUM_3D(rowAr0, rowAr1, rowBr0, rowBr1,
+                                       rowCr0, rowCr1, rowDr0, rowDr1);
+         const GLint g = FILTER_SUM_3D(rowAg0, rowAg1, rowBg0, rowBg1,
+                                       rowCg0, rowCg1, rowDg0, rowDg1);
+         const GLint b = FILTER_SUM_3D(rowAb0, rowAb1, rowBb0, rowBb1,
+                                       rowCb0, rowCb1, rowDb0, rowDb1);
+         dst[i] = (b << 5) | (g << 2) | r;
+      }
+   }
+   else {
+      _mesa_problem(NULL, "bad format in do_row()");
+   }
+}
+
+
 /*
  * These functions generate a 1/2-size mipmap image from a source image.
  * Texture borders are handled by copying or averaging the source image's
@@ -544,7 +987,6 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border,
    const GLint dstWidthNB = dstWidth - 2 * border;
    const GLint dstHeightNB = dstHeight - 2 * border;
    const GLint dstDepthNB = dstDepth - 2 * border;
-   GLvoid *tmpRowA, *tmpRowB;
    GLint img, row;
    GLint bytesPerSrcImage, bytesPerDstImage;
    GLint bytesPerSrcRow, bytesPerDstRow;
@@ -552,15 +994,6 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border,
 
    (void) srcDepthNB; /* silence warnings */
 
-   /* Need two temporary row buffers */
-   tmpRowA = _mesa_malloc(srcWidth * bpt);
-   if (!tmpRowA)
-      return;
-   tmpRowB = _mesa_malloc(srcWidth * bpt);
-   if (!tmpRowB) {
-      _mesa_free(tmpRowA);
-      return;
-   }
 
    bytesPerSrcImage = srcWidth * srcHeight * bpt;
    bytesPerDstImage = dstWidth * dstHeight * bpt;
@@ -607,15 +1040,11 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border,
       GLubyte *dstImgRow = imgDst;
 
       for (row = 0; row < dstHeightNB; row++) {
-         /* Average together two rows from first src image */
-         do_row(datatype, comps, srcWidthNB, srcImgARowA, srcImgARowB,
-                srcWidthNB, tmpRowA);
-         /* Average together two rows from second src image */
-         do_row(datatype, comps, srcWidthNB, srcImgBRowA, srcImgBRowB,
-                srcWidthNB, tmpRowB);
-         /* Average together the temp rows to make the final row */
-         do_row(datatype, comps, srcWidthNB, tmpRowA, tmpRowB,
-                dstWidthNB, dstImgRow);
+         do_row_3D(datatype, comps, srcWidthNB, 
+                   srcImgARowA, srcImgARowB,
+                   srcImgBRowA, srcImgBRowB,
+                   dstWidthNB, dstImgRow);
+
          /* advance to next rows */
          srcImgARowA += bytesPerSrcRow + srcRowOffset;
          srcImgARowB += bytesPerSrcRow + srcRowOffset;
@@ -625,8 +1054,6 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border,
       }
    }
 
-   _mesa_free(tmpRowA);
-   _mesa_free(tmpRowB);
 
    /* Luckily we can leverage the make_2d_mipmap() function here! */
    if (border > 0) {
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index 284f81b7cf..be982afe39 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -1,15 +1,9 @@
-/**
- * \file mtypes.h
- * Main Mesa data structures.
- *
- * Please try to mark derived values with a leading underscore ('_').
- */
-
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.3
+ * Version:  7.3
  *
  * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -29,7 +23,12 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
-
+/**
+ * \file mtypes.h
+ * Main Mesa data structures.
+ *
+ * Please try to mark derived values with a leading underscore ('_').
+ */
 
 #ifndef TYPES_H
 #define TYPES_H
@@ -126,6 +125,8 @@ struct gl_program_cache;
 struct gl_texture_format;
 struct gl_texture_image;
 struct gl_texture_object;
+struct st_context;
+struct pipe_surface;
 typedef struct __GLcontextRec GLcontext;
 typedef struct __GLcontextModesRec GLvisual;
 typedef struct gl_framebuffer GLframebuffer;
@@ -639,7 +640,7 @@ struct gl_current_attrib
    GLfloat RasterColor[4];
    GLfloat RasterSecondaryColor[4];
    GLfloat RasterIndex;
-   GLfloat RasterTexCoords[MAX_TEXTURE_COORD_UNITS][4];
+   GLfloat RasterTexCoords[MAX_TEXTURE_UNITS][4];
    GLboolean RasterPosValid;
    /*@}*/
 };
@@ -723,10 +724,10 @@ struct gl_enable_attrib
    GLboolean SampleCoverage;          /* GL_ARB_multisample */
    GLboolean SampleCoverageInvert;    /* GL_ARB_multisample */
    GLboolean RasterPositionUnclipped; /* GL_IBM_rasterpos_clip */
-   GLuint Texture[MAX_TEXTURE_IMAGE_UNITS];
-   GLuint TexGen[MAX_TEXTURE_COORD_UNITS];
+   GLuint Texture[MAX_TEXTURE_UNITS];
+   GLuint TexGen[MAX_TEXTURE_UNITS];
    /* SGI_texture_color_table */
-   GLboolean TextureColorTable[MAX_TEXTURE_IMAGE_UNITS];
+   GLboolean TextureColorTable[MAX_TEXTURE_UNITS];
    /* GL_ARB_vertex_program / GL_NV_vertex_program */
    GLboolean VertexProgram;
    GLboolean VertexProgramPointSize;
@@ -902,7 +903,7 @@ struct gl_light_attrib
    /*@{*/
    GLboolean _NeedEyeCoords;		
    GLboolean _NeedVertices;		/**< Use fast shader? */
-   GLbitfield  _Flags;		        /**< LIGHT_* flags, see above */
+   GLbitfield _Flags;		        /**< LIGHT_* flags, see above */
    GLfloat _BaseColor[2][3];
    /*@}*/
 };
@@ -1069,7 +1070,7 @@ struct gl_point_attrib
    GLfloat Threshold;		/**< GL_EXT_point_parameters */
    GLboolean _Attenuated;	/**< True if Params != [1, 0, 0] */
    GLboolean PointSprite;	/**< GL_NV/ARB_point_sprite */
-   GLboolean CoordReplace[MAX_TEXTURE_COORD_UNITS]; /**< GL_ARB_point_sprite */
+   GLboolean CoordReplace[MAX_TEXTURE_UNITS]; /**< GL_ARB_point_sprite */
    GLenum SpriteRMode;		/**< GL_NV_point_sprite (only!) */
    GLenum SpriteOrigin;		/**< GL_ARB_point_sprite */
 };
@@ -1109,25 +1110,40 @@ struct gl_scissor_attrib
 
 /**
  * Stencil attribute group (GL_STENCIL_BUFFER_BIT).
+ *
+ * Three sets of stencil data are tracked so that OpenGL 2.0,
+ * GL_EXT_stencil_two_side, and GL_ATI_separate_stencil can all be supported
+ * simultaneously.  In each of the stencil state arrays, element 0 corresponds
+ * to GL_FRONT.  Element 1 corresponds to the OpenGL 2.0 /
+ * GL_ATI_separate_stencil GL_BACK state.  Element 2 corresponds to the
+ * GL_EXT_stencil_two_side GL_BACK state.
+ *
+ * The derived value \c _BackFace is either 1 or 2 depending on whether or
+ * not GL_STENCIL_TEST_TWO_SIDE_EXT is enabled.
+ *
+ * The derived value \c _TestTwoSide is set when the front-face and back-face
+ * stencil state are different.
  */
 struct gl_stencil_attrib
 {
    GLboolean Enabled;		/**< Enabled flag */
    GLboolean TestTwoSide;	/**< GL_EXT_stencil_two_side */
-   GLubyte ActiveFace;		/**< GL_EXT_stencil_two_side (0 or 1) */
+   GLubyte ActiveFace;		/**< GL_EXT_stencil_two_side (0 or 2) */
    GLboolean _TestTwoSide;
-   GLenum Function[2];		/**< Stencil function */
-   GLenum FailFunc[2];		/**< Fail function */
-   GLenum ZPassFunc[2];		/**< Depth buffer pass function */
-   GLenum ZFailFunc[2];		/**< Depth buffer fail function */
-   GLint Ref[2];		/**< Reference value */
-   GLuint ValueMask[2];		/**< Value mask */
-   GLuint WriteMask[2];		/**< Write mask */
+   GLubyte _BackFace;
+   GLenum Function[3];		/**< Stencil function */
+   GLenum FailFunc[3];		/**< Fail function */
+   GLenum ZPassFunc[3];		/**< Depth buffer pass function */
+   GLenum ZFailFunc[3];		/**< Depth buffer fail function */
+   GLint Ref[3];		/**< Reference value */
+   GLuint ValueMask[3];		/**< Value mask */
+   GLuint WriteMask[3];		/**< Write mask */
    GLuint Clear;		/**< Clear value */
 };
 
 
-#define NUM_TEXTURE_TARGETS 7   /* 1D, 2D, 3D, CUBE, RECT, 1D_STACK, and 2D_STACK */
+/** 1D, 2D, 3D, CUBE, RECT, 1D_ARRAY, and 2D_ARRAY targets */
+#define NUM_TEXTURE_TARGETS 7
 
 /**
  * An index for each type of texture object
@@ -1420,11 +1436,9 @@ struct gl_texture_object
    GLint BaseLevel;		/**< min mipmap level, OpenGL 1.2 */
    GLint MaxLevel;		/**< max mipmap level, OpenGL 1.2 */
    GLfloat MaxAnisotropy;	/**< GL_EXT_texture_filter_anisotropic */
-   GLboolean CompareFlag;	/**< GL_SGIX_shadow */
-   GLenum CompareOperator;	/**< GL_SGIX_shadow */
-   GLfloat ShadowAmbient;       /**< GL_ARB_shadow_ambient */
    GLenum CompareMode;		/**< GL_ARB_shadow */
    GLenum CompareFunc;		/**< GL_ARB_shadow */
+   GLfloat CompareFailValue;    /**< GL_ARB_shadow_ambient */
    GLenum _Function;		/**< Comparison function derived from 
 				 * \c CompareOperator, \c CompareMode, and
 				 * \c CompareFunc.
@@ -1433,8 +1447,11 @@ struct gl_texture_object
    GLint _MaxLevel;		/**< actual max mipmap level (q in the spec) */
    GLfloat _MaxLambda;		/**< = _MaxLevel - BaseLevel (q - b in spec) */
    GLint CropRect[4];           /**< GL_OES_draw_texture */
+   GLenum Swizzle[4];           /**< GL_EXT_texture_swizzle */
+   GLuint _Swizzle;             /**< same as Swizzle, but SWIZZLE_* format */
    GLboolean GenerateMipmap;    /**< GL_SGIS_generate_mipmap */
    GLboolean _Complete;		/**< Is texture object complete? */
+   GLboolean _RenderToTexture;  /**< Any rendering to this texture? */
 
    /** Actual texture images, indexed by [cube face] and [mipmap level] */
    struct gl_texture_image *Image[MAX_FACES][MAX_TEXTURE_LEVELS];
@@ -1455,23 +1472,20 @@ struct gl_texture_object
 
 /**
  * Texture combine environment state.
- * 
- * \todo
- * If GL_NV_texture_env_combine4 is ever supported, the arrays in this
- * structure will need to be expanded for 4 elements.
+ * Up to four combiner sources are possible with GL_NV_texture_env_combine4.
  */
 struct gl_tex_env_combine_state
 {
    GLenum ModeRGB;       /**< GL_REPLACE, GL_DECAL, GL_ADD, etc. */
    GLenum ModeA;         /**< GL_REPLACE, GL_DECAL, GL_ADD, etc. */
-   GLenum SourceRGB[3];  /**< GL_PRIMARY_COLOR, GL_TEXTURE, etc. */
-   GLenum SourceA[3];    /**< GL_PRIMARY_COLOR, GL_TEXTURE, etc. */
-   GLenum OperandRGB[3]; /**< SRC_COLOR, ONE_MINUS_SRC_COLOR, etc */
-   GLenum OperandA[3];   /**< SRC_ALPHA, ONE_MINUS_SRC_ALPHA, etc */
+   GLenum SourceRGB[4];  /**< GL_PRIMARY_COLOR, GL_TEXTURE, etc. */
+   GLenum SourceA[4];    /**< GL_PRIMARY_COLOR, GL_TEXTURE, etc. */
+   GLenum OperandRGB[4]; /**< SRC_COLOR, ONE_MINUS_SRC_COLOR, etc */
+   GLenum OperandA[4];   /**< SRC_ALPHA, ONE_MINUS_SRC_ALPHA, etc */
    GLuint ScaleShiftRGB; /**< 0, 1 or 2 */
    GLuint ScaleShiftA;   /**< 0, 1 or 2 */
-   GLuint _NumArgsRGB;   /**< Number of inputs used for the combine mode. */
-   GLuint _NumArgsA;     /**< Number of inputs used for the combine mode. */
+   GLuint _NumArgsRGB;   /**< Number of inputs used for the RGB combiner */
+   GLuint _NumArgsA;     /**< Number of inputs used for the A combiner */
 };
 
 
@@ -1557,7 +1571,7 @@ struct gl_texture_attrib
     * name multitexture 
     */
    /**@{*/
-   GLuint CurrentUnit;	       /**< Active texture unit */
+   GLuint CurrentUnit;   /**< Active texture unit [0, MaxTextureImageUnits-1] */
    GLbitfield _EnabledUnits;  /**< one bit set for each really-enabled unit */
    GLbitfield _EnabledCoordUnits;   /**< one bit per enabled coordinate unit */
    GLbitfield _GenFlags;            /**< for texgen */
@@ -1567,7 +1581,6 @@ struct gl_texture_attrib
 
    struct gl_texture_unit Unit[MAX_TEXTURE_UNITS];
 
-   /** Proxy texture objects */
    struct gl_texture_object *ProxyTex[NUM_TEXTURE_TARGETS];
 
    /** GL_EXT_shared_texture_palette */
@@ -1662,6 +1675,7 @@ struct gl_client_array
 {
    GLint Size;                  /**< components per element (1,2,3,4) */
    GLenum Type;                 /**< datatype: GL_FLOAT, GL_INT, etc */
+   GLenum Format;               /**< default: GL_RGBA, but may be GL_BGRA */
    GLsizei Stride;		/**< user-specified stride */
    GLsizei StrideB;		/**< actual stride in bytes */
    const GLubyte *Ptr;          /**< Points to array data */
@@ -1873,10 +1887,13 @@ struct gl_program
 
    GLbitfield InputsRead;     /**< Bitmask of which input regs are read */
    GLbitfield OutputsWritten; /**< Bitmask of which output regs are written to */
-   GLbitfield TexturesUsed[MAX_TEXTURE_IMAGE_UNITS];  /**< TEXTURE_x_BIT bitmask */
+   GLbitfield InputFlags[MAX_PROGRAM_INPUTS];   /**< PROG_PARAM_BIT_x flags */
+   GLbitfield OutputFlags[MAX_PROGRAM_OUTPUTS]; /**< PROG_PARAM_BIT_x flags */
+   GLbitfield TexturesUsed[MAX_TEXTURE_UNITS];  /**< TEXTURE_x_BIT bitmask */
    GLbitfield SamplersUsed;   /**< Bitfield of which samplers are used */
    GLbitfield ShadowSamplers; /**< Texture units used for shadow sampling. */
 
+
    /** Named parameters, constants, etc. from program text */
    struct gl_program_parameter_list *Parameters;
    /** Numbered local parameters */
@@ -1954,14 +1971,15 @@ struct gl_program_state
  */
 struct gl_vertex_program_state
 {
-   GLboolean Enabled;               /**< GL_VERTEX_PROGRAM_ARB/NV */
-   GLboolean _Enabled;              /**< Enabled and valid program? */
+   GLboolean Enabled;               /**< User-set GL_VERTEX_PROGRAM_ARB/NV flag */
+   GLboolean _Enabled;              /**< Enabled and _valid_ user program? */
    GLboolean PointSizeEnabled;      /**< GL_VERTEX_PROGRAM_POINT_SIZE_ARB/NV */
    GLboolean TwoSideEnabled;        /**< GL_VERTEX_PROGRAM_TWO_SIDE_ARB/NV */
-   struct gl_vertex_program *Current;  /**< user-bound vertex program */
+   struct gl_vertex_program *Current;  /**< User-bound vertex program */
 
-   /** Currently enabled and valid program (including internal programs
-    * and compiled shader programs).
+   /** Currently enabled and valid vertex program (including internal programs,
+    * user-defined vertex programs and GLSL vertex shaders).
+    * This is the program we must use when rendering.
     */
    struct gl_vertex_program *_Current;
 
@@ -1986,6 +2004,8 @@ struct gl_vertex_program_state
    GLboolean CallbackEnabled;
    GLuint CurrentPosition;
 #endif
+
+   GLboolean _Overriden;
 };
 
 
@@ -1995,12 +2015,12 @@ struct gl_vertex_program_state
 struct gl_fragment_program_state
 {
    GLboolean Enabled;     /**< User-set fragment program enable flag */
-   GLboolean _Enabled;    /**< Fragment program enabled and valid? */
-   GLboolean _Active;
+   GLboolean _Enabled;    /**< Enabled and _valid_ user program? */
    struct gl_fragment_program *Current;  /**< User-bound fragment program */
 
-   /** Currently enabled and valid program (including internal programs
-    * and compiled shader programs).
+   /** Currently enabled and valid fragment program (including internal programs,
+    * user-defined fragment programs and GLSL fragment shaders).
+    * This is the program we must use when rendering.
     */
    struct gl_fragment_program *_Current;
 
@@ -2008,7 +2028,6 @@ struct gl_fragment_program_state
 
    /** Should fixed-function texturing be implemented with a fragment prog? */
    GLboolean _MaintainTexEnvProgram;
-   GLboolean _UseTexEnvProgram;
 
    /** Program to emulate fixed-function texture env/combine (see above) */
    struct gl_fragment_program *_TexEnvProgram;
@@ -2092,6 +2111,13 @@ struct gl_query_state
 };
 
 
+/** Set by #pragma directives */
+struct gl_sl_pragmas
+{
+   GLboolean Optimize;  /**< defaults on */
+   GLboolean Debug;     /**< defaults off */
+};
+
 
 /**
  * A GLSL vertex or fragment shader object.
@@ -2102,12 +2128,12 @@ struct gl_shader
    GLuint Name;  /**< AKA the handle */
    GLint RefCount;  /**< Reference count */
    GLboolean DeletePending;
-
-   const GLchar *Source;  /**< Source code string */
    GLboolean CompileStatus;
+   GLboolean Main;  /**< shader defines main() */
+   const GLchar *Source;  /**< Source code string */
    struct gl_program *Program;  /**< Post-compile assembly code */
    GLchar *InfoLog;
-   GLboolean Main;  /**< shader defines main() */
+   struct gl_sl_pragmas Pragmas;
 };
 
 
@@ -2139,6 +2165,13 @@ struct gl_shader_program
 };   
 
 
+#define GLSL_DUMP      0x1  /**< Dump shaders to stdout */
+#define GLSL_LOG       0x2  /**< Write shaders to files */
+#define GLSL_OPT       0x4  /**< Force optimizations (override pragmas) */
+#define GLSL_NO_OPT    0x8  /**< Force no optimizations (override pragmas) */
+#define GLSL_UNIFORMS 0x10  /**< Print glUniform calls */
+
+
 /**
  * Context state for GLSL vertex/fragment shaders.
  */
@@ -2150,6 +2183,7 @@ struct gl_shader_state
    GLboolean EmitCondCodes;             /**< Use condition codes? */
    GLboolean EmitComments;              /**< Annotated instructions */
    void *MemPool;
+   GLbitfield Flags;                    /**< Mask of GLSL_x flags */
 };
 
 
@@ -2251,7 +2285,9 @@ struct gl_renderbuffer
    GLenum _ActualFormat;  /**< The driver-chosen format */
    GLenum _BaseFormat;    /**< Either GL_RGB, GL_RGBA, GL_DEPTH_COMPONENT or
                                GL_STENCIL_INDEX. */
-   GLenum DataType;      /**< Type of values passed to the Get/Put functions */
+   GLenum ColorEncoding; /**< GL_LINEAR or GL_SRGB */
+   GLenum ComponentType; /**< GL_FLOAT, GL_INT, GL_UNSIGNED_INT,
+                              GL_UNSIGNED_NORMALIZED or GL_INDEX */
    GLubyte RedBits;      /**< Bits of red per pixel */
    GLubyte GreenBits;
    GLubyte BlueBits;
@@ -2259,6 +2295,9 @@ struct gl_renderbuffer
    GLubyte IndexBits;
    GLubyte DepthBits;
    GLubyte StencilBits;
+   GLubyte NumSamples;
+
+   GLenum DataType;      /**< Type of values passed to the Get/Put functions */
    GLvoid *Data;        /**< This may not be used by some kinds of RBs */
 
    /* Used to wrap one renderbuffer around another: */
@@ -2464,9 +2503,9 @@ struct gl_constants
    GLint MaxTextureRectSize;            /* GL_NV_texture_rectangle */
    GLuint MaxTextureCoordUnits;
    GLuint MaxTextureImageUnits;
-   GLuint MaxTextureUnits;              /* = MIN(CoordUnits, ImageUnits) */
-   GLfloat MaxTextureMaxAnisotropy;	/* GL_EXT_texture_filter_anisotropic */
-   GLfloat MaxTextureLodBias;           /* GL_EXT_texture_lod_bias */
+   GLuint MaxTextureUnits;              /**< = MIN(CoordUnits, ImageUnits) */
+   GLfloat MaxTextureMaxAnisotropy;     /**< GL_EXT_texture_filter_anisotropic */
+   GLfloat MaxTextureLodBias;           /**< GL_EXT_texture_lod_bias */
    GLuint MaxArrayLockSize;
    GLint SubPixelBits;
    GLfloat MinPointSize, MaxPointSize;		/* aliased */
@@ -2501,6 +2540,8 @@ struct gl_constants
    /* GL_ARB_vertex_shader */
    GLuint MaxVertexTextureImageUnits;
    GLuint MaxVarying;  /**< Number of float[4] vectors */
+   /* GL_ARB_framebuffer_object */
+   GLuint MaxSamples;
 };
 
 
@@ -2522,6 +2563,7 @@ struct gl_extensions
    GLboolean ARB_fragment_program;
    GLboolean ARB_fragment_program_shadow;
    GLboolean ARB_fragment_shader;
+   GLboolean ARB_framebuffer_object;
    GLboolean ARB_half_float_pixel;
    GLboolean ARB_imaging;
    GLboolean ARB_multisample;
@@ -2532,6 +2574,7 @@ struct gl_extensions
    GLboolean ARB_shading_language_100;
    GLboolean ARB_shading_language_120;
    GLboolean ARB_shadow;
+   GLboolean ARB_shadow_ambient; /* or GL_ARB_shadow_ambient */
    GLboolean ARB_texture_border_clamp;
    GLboolean ARB_texture_compression;
    GLboolean ARB_texture_cube_map;
@@ -2592,8 +2635,10 @@ struct gl_extensions
    GLboolean EXT_texture_lod_bias;
    GLboolean EXT_texture_mirror_clamp;
    GLboolean EXT_texture_sRGB;
+   GLboolean EXT_texture_swizzle;
    GLboolean EXT_timer_query;
    GLboolean EXT_vertex_array;
+   GLboolean EXT_vertex_array_bgra;
    GLboolean EXT_vertex_array_set;
    /* vendor extensions */
    GLboolean APPLE_client_storage;
@@ -2616,6 +2661,7 @@ struct gl_extensions
    GLboolean NV_light_max_exponent;
    GLboolean NV_point_sprite;
    GLboolean NV_texgen_reflection;
+   GLboolean NV_texture_env_combine4;
    GLboolean NV_texture_rectangle;
    GLboolean NV_vertex_program;
    GLboolean NV_vertex_program1_1;
@@ -2626,8 +2672,6 @@ struct gl_extensions
    GLboolean SGIS_generate_mipmap;
    GLboolean SGIS_texture_edge_clamp;
    GLboolean SGIS_texture_lod;
-   GLboolean SGIX_shadow;
-   GLboolean SGIX_shadow_ambient; /* or GL_ARB_shadow_ambient */
    GLboolean TDFX_texture_compression_FXT1;
    GLboolean S3_s3tc;
    /*@}*/
@@ -2719,6 +2763,7 @@ struct gl_matrix_stack
 #define _NEW_MULTISAMPLE        0x2000000  /**< __GLcontextRec::Multisample */
 #define _NEW_TRACK_MATRIX       0x4000000  /**< __GLcontextRec::VertexProgram */
 #define _NEW_PROGRAM            0x8000000  /**< __GLcontextRec::VertexProgram */
+#define _NEW_CURRENT_ATTRIB     0x10000000  /**< __GLcontextRec::Current */
 #define _NEW_ALL ~0
 /*@}*/
 
@@ -2801,6 +2846,10 @@ struct gl_matrix_stack
 /*@}*/
 
 
+/**
+ * Composite state flags
+ */
+/*@{*/
 #define _MESA_NEW_NEED_EYE_COORDS         (_NEW_LIGHT |		\
                                            _NEW_TEXTURE |	\
                                            _NEW_POINT |		\
@@ -2810,16 +2859,11 @@ struct gl_matrix_stack
 #define _MESA_NEW_NEED_NORMALS            (_NEW_LIGHT |		\
                                            _NEW_TEXTURE)
 
-#define _IMAGE_NEW_TRANSFER_STATE         (_NEW_PIXEL | _NEW_COLOR_MATRIX)
-
-
+#define _MESA_NEW_TRANSFER_STATE          (_NEW_PIXEL |		\
+                                           _NEW_COLOR_MATRIX)
+/*@}*/
 
 
-/*
- * Forward declaration of display list data types:
- */
-union node;
-typedef union node Node;
 
 
 /* This has to be included here. */
@@ -2851,21 +2895,31 @@ struct gl_tnl_module
    /*@}*/
 };
 
-/* Strictly this is a tnl/ private concept, but it doesn't seem
+
+/**
+ * Display list flags.
+ * Strictly this is a tnl-private concept, but it doesn't seem
  * worthwhile adding a tnl private structure just to hold this one bit
  * of information:
  */
-#define MESA_DLIST_DANGLING_REFS     0x1 
+#define DLIST_DANGLING_REFS     0x1 
 
-/* Provide a location where information about a display list can be
+
+/** Opaque declaration of display list payload data type */
+union gl_dlist_node;
+
+
+/**
+ * Provide a location where information about a display list can be
  * collected.  Could be extended with driverPrivate structures,
  * etc. in the future.
  */
-struct mesa_display_list
+struct gl_display_list
 {
-   Node *node;
-   GLuint id;
-   GLbitfield flags;
+   GLuint Name;
+   GLbitfield Flags;  /**< DLIST_x flags */
+   /** The dlist commands are in a linked list of nodes */
+   union gl_dlist_node *Head;
 };
 
 
@@ -2876,10 +2930,8 @@ struct gl_dlist_state
 {
    GLuint CallDepth;		/**< Current recursion calling depth */
 
-   struct mesa_display_list *CurrentList;
-   Node *CurrentListPtr;	/**< Head of list being compiled */
-   GLuint CurrentListNum;	/**< Number of the list being compiled */
-   Node *CurrentBlock;		/**< Pointer to current block of nodes */
+   struct gl_display_list *CurrentList; /**< List currently being compiled */
+   union gl_dlist_node *CurrentBlock; /**< Pointer to current block of nodes */
    GLuint CurrentPos;		/**< Index into current block of nodes */
 
    GLvertexformat ListVtxfmt;
@@ -2941,7 +2993,7 @@ struct __GLcontextRec
    struct gl_matrix_stack ModelviewMatrixStack;
    struct gl_matrix_stack ProjectionMatrixStack;
    struct gl_matrix_stack ColorMatrixStack;
-   struct gl_matrix_stack TextureMatrixStack[MAX_TEXTURE_COORD_UNITS];
+   struct gl_matrix_stack TextureMatrixStack[MAX_TEXTURE_UNITS];
    struct gl_matrix_stack ProgramMatrixStack[MAX_PROGRAM_MATRICES];
    struct gl_matrix_stack *CurrentStack; /**< Points to one of the above stacks */
    /*@}*/
@@ -3046,6 +3098,8 @@ struct __GLcontextRec
    GLenum RenderMode;        /**< either GL_RENDER, GL_SELECT, GL_FEEDBACK */
    GLbitfield NewState;      /**< bitwise-or of _NEW_* flags */
 
+   GLbitfield varying_vp_inputs;  /**< mask of VERT_BIT_* flags */
+
    /** \name Derived state */
    /*@{*/
    /** Bitwise-or of DD_* flags.  Note that this bitfield may be used before
@@ -3091,7 +3145,7 @@ struct __GLcontextRec
    void *swsetup_context;
    void *swtnl_context;
    void *swtnl_im;
-   void *acache_context;
+   struct st_context *st;
    void *aelt_context;
    /*@}*/
 };
@@ -3129,8 +3183,6 @@ enum _verbose
    VERBOSE_PRIMS		= 0x0400,
    VERBOSE_VERTS		= 0x0800,
    VERBOSE_DISASSEM		= 0x1000,
-   VERBOSE_GLSL			= 0x2000,
-   VERBOSE_GLSL_DUMP		= 0x4000
 };
 
 
diff --git a/src/mesa/main/multisample.c b/src/mesa/main/multisample.c
index b9cfad9216..01b68df7af 100644
--- a/src/mesa/main/multisample.c
+++ b/src/mesa/main/multisample.c
@@ -37,11 +37,6 @@ _mesa_SampleCoverageARB(GLclampf value, GLboolean invert)
 {
    GET_CURRENT_CONTEXT(ctx);
 
-   if (!ctx->Extensions.ARB_multisample) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glSampleCoverageARB");
-      return;
-   }
-
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH( ctx );
 
    ctx->Multisample.SampleCoverageValue = (GLfloat) CLAMP(value, 0.0, 1.0);
diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c
index 8d24a201f0..52781e048e 100644
--- a/src/mesa/main/pixel.c
+++ b/src/mesa/main/pixel.c
@@ -828,7 +828,7 @@ void _mesa_update_pixel( GLcontext *ctx, GLuint new_state )
 
    /* References ColorMatrix.type (derived above).
     */
-   if (new_state & _IMAGE_NEW_TRANSFER_STATE)
+   if (new_state & _MESA_NEW_TRANSFER_STATE)
       update_image_transfer_state(ctx);
 }
 
diff --git a/src/mesa/main/points.c b/src/mesa/main/points.c
index 1fe697033f..4c8fc1f72e 100644
--- a/src/mesa/main/points.c
+++ b/src/mesa/main/points.c
@@ -257,7 +257,7 @@ _mesa_init_point(GLcontext *ctx)
    ctx->Point.PointSprite = GL_FALSE; /* GL_ARB/NV_point_sprite */
    ctx->Point.SpriteRMode = GL_ZERO; /* GL_NV_point_sprite (only!) */
    ctx->Point.SpriteOrigin = GL_UPPER_LEFT; /* GL_ARB_point_sprite */
-   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
+   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
       ctx->Point.CoordReplace[i] = GL_FALSE; /* GL_ARB/NV_point_sprite */
    }
 }
diff --git a/src/mesa/main/rastpos.c b/src/mesa/main/rastpos.c
index 9842172f46..9f309d6ab8 100644
--- a/src/mesa/main/rastpos.c
+++ b/src/mesa/main/rastpos.c
@@ -500,7 +500,7 @@ void _mesa_init_rastpos( GLcontext * ctx )
    ASSIGN_4V( ctx->Current.RasterColor, 1.0, 1.0, 1.0, 1.0 );
    ASSIGN_4V( ctx->Current.RasterSecondaryColor, 0.0, 0.0, 0.0, 1.0 );
    ctx->Current.RasterIndex = 1.0;
-   for (i=0; i<MAX_TEXTURE_UNITS; i++)
+   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
       ASSIGN_4V( ctx->Current.RasterTexCoords[i], 0.0, 0.0, 0.0, 1.0 );
    ctx->Current.RasterPosValid = GL_TRUE;
 }
diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c
index 6f1d7c3960..38be8266e0 100644
--- a/src/mesa/main/renderbuffer.c
+++ b/src/mesa/main/renderbuffer.c
@@ -1199,7 +1199,8 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb,
 
    if (width > 0 && height > 0) {
       /* allocate new buffer storage */
-      rb->Data = _mesa_malloc(width * height * pixelSize);
+      rb->Data = malloc(width * height * pixelSize);
+
       if (rb->Data == NULL) {
          rb->Width = 0;
          rb->Height = 0;
@@ -1487,11 +1488,16 @@ _mesa_init_renderbuffer(struct gl_renderbuffer *rb, GLuint name)
    rb->InternalFormat = GL_NONE;
    rb->_ActualFormat = GL_NONE;
    rb->_BaseFormat = GL_NONE;
-   rb->DataType = GL_NONE;
+
+   rb->ComponentType = GL_UNSIGNED_NORMALIZED; /* ARB_fbo */
+   rb->ColorEncoding = GL_LINEAR; /* ARB_fbo */
+
    rb->RedBits = rb->GreenBits = rb->BlueBits = rb->AlphaBits = 0;
    rb->IndexBits = 0;
    rb->DepthBits = 0;
    rb->StencilBits = 0;
+
+   rb->DataType = GL_NONE;
    rb->Data = NULL;
 
    /* Point back to ourself so that we don't have to check for Wrapped==NULL
diff --git a/src/mesa/main/sources b/src/mesa/main/sources
index eb8cd900ee..4e58702c74 100644
--- a/src/mesa/main/sources
+++ b/src/mesa/main/sources
@@ -20,6 +20,7 @@ debug.c \
 depth.c \
 depthstencil.c \
 dlist.c \
+dlopen.c \
 drawpix.c \
 enable.c \
 enums.c \
@@ -78,7 +79,6 @@ vsnprintf.c
 MESA_MAIN_HEADERS = \
 accum.h \
 api_arrayelt.h \
-api_eval.h \
 api_exec.h \
 api_loopback.h \
 api_noop.h \
@@ -101,6 +101,7 @@ debug.h \
 depth.h \
 depthstencil.h \
 dlist.h \
+dlopen.h \
 drawpix.h \
 enable.h \
 enums.h \
diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c
index e87a62f59d..0a39279bff 100644
--- a/src/mesa/main/state.c
+++ b/src/mesa/main/state.c
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.3
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
@@ -173,13 +173,16 @@ update_arrays( GLcontext *ctx )
 }
 
 
+/**
+ * Update the following fields:
+ *   ctx->VertexProgram._Enabled
+ *   ctx->FragmentProgram._Enabled
+ *   ctx->ATIFragmentShader._Enabled
+ * This needs to be done before texture state validation.
+ */
 static void
-update_program(GLcontext *ctx)
+update_program_enables(GLcontext *ctx)
 {
-   const struct gl_shader_program *shProg = ctx->Shader.CurrentProgram;
-   const struct gl_vertex_program *prevVP = ctx->VertexProgram._Current;
-   const struct gl_fragment_program *prevFP = ctx->FragmentProgram._Current;
-
    /* These _Enabled flags indicate if the program is enabled AND valid. */
    ctx->VertexProgram._Enabled = ctx->VertexProgram.Enabled
       && ctx->VertexProgram.Current->Base.Instructions;
@@ -187,11 +190,30 @@ update_program(GLcontext *ctx)
       && ctx->FragmentProgram.Current->Base.Instructions;
    ctx->ATIFragmentShader._Enabled = ctx->ATIFragmentShader.Enabled
       && ctx->ATIFragmentShader.Current->Instructions[0];
+}
+
+
+/**
+ * Update vertex/fragment program state.  In particular, update these fields:
+ *   ctx->VertexProgram._Current
+ *   ctx->VertexProgram._TnlProgram,
+ * These point to the highest priority enabled vertex/fragment program or are
+ * NULL if fixed-function processing is to be done.
+ *
+ * This function needs to be called after texture state validation in case
+ * we're generating a fragment program from fixed-function texture state.
+ */
+static void
+update_program(GLcontext *ctx)
+{
+   const struct gl_shader_program *shProg = ctx->Shader.CurrentProgram;
+   const struct gl_vertex_program *prevVP = ctx->VertexProgram._Current;
+   const struct gl_fragment_program *prevFP = ctx->FragmentProgram._Current;
 
    /*
     * Set the ctx->VertexProgram._Current and ctx->FragmentProgram._Current
-    * pointers to the programs that should be enabled/used.  These will only
-    * be NULL if we need to use the fixed-function code.
+    * pointers to the programs that should be used for rendering.  If either
+    * is NULL, use fixed-function code paths.
     *
     * These programs may come from several sources.  The priority is as
     * follows:
@@ -204,8 +226,6 @@ update_program(GLcontext *ctx)
     * come up, or matter.
     */
 
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL);
-
    if (shProg && shProg->LinkStatus && shProg->FragmentProgram) {
       /* Use shader programs */
       _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current,
@@ -256,17 +276,6 @@ update_program(GLcontext *ctx)
       _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, NULL);
    }
 
-   /* XXX: get rid of _Active flag.
-    */
-#if 1
-   ctx->FragmentProgram._Active = ctx->FragmentProgram._Enabled;
-   if (ctx->FragmentProgram._MaintainTexEnvProgram &&
-       !ctx->FragmentProgram._Enabled) {
-      if (ctx->FragmentProgram._UseTexEnvProgram)
-	 ctx->FragmentProgram._Active = GL_TRUE;
-   }
-#endif
-
    /* Let the driver know what's happening:
     */
    if (ctx->FragmentProgram._Current != prevFP && ctx->Driver.BindProgram) {
@@ -439,16 +448,37 @@ _mesa_update_state_locked( GLcontext *ctx )
    GLbitfield new_state = ctx->NewState;
    GLbitfield prog_flags = _NEW_PROGRAM;
 
+   if (new_state == _NEW_CURRENT_ATTRIB) 
+      goto out;
+
    if (MESA_VERBOSE & VERBOSE_STATE)
       _mesa_print_state("_mesa_update_state", new_state);
 
+   /* Determine which state flags effect vertex/fragment program state */
+   if (ctx->FragmentProgram._MaintainTexEnvProgram) {
+      prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR);
+   }
+   if (ctx->VertexProgram._MaintainTnlProgram) {
+      prog_flags |= (_NEW_ARRAY | _NEW_TEXTURE | _NEW_TEXTURE_MATRIX |
+                     _NEW_TRANSFORM | _NEW_POINT |
+                     _NEW_FOG | _NEW_LIGHT |
+                     _MESA_NEW_NEED_EYE_COORDS);
+   }
+
+   /*
+    * Now update derived state info
+    */
+
+   if (new_state & prog_flags)
+      update_program_enables( ctx );
+
    if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION))
       _mesa_update_modelview_project( ctx, new_state );
 
    if (new_state & (_NEW_PROGRAM|_NEW_TEXTURE|_NEW_TEXTURE_MATRIX))
       _mesa_update_texture( ctx, new_state );
 
-   if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL))
+   if (new_state & _NEW_BUFFERS)
       _mesa_update_framebuffer(ctx);
 
    if (new_state & (_NEW_SCISSOR | _NEW_BUFFERS | _NEW_VIEWPORT))
@@ -464,7 +494,7 @@ _mesa_update_state_locked( GLcontext *ctx )
       _mesa_update_stencil( ctx );
 
 #if FEATURE_pixel_transfer
-   if (new_state & _IMAGE_NEW_TRANSFER_STATE)
+   if (new_state & _MESA_NEW_TRANSFER_STATE)
       _mesa_update_pixel( ctx, new_state );
 #endif
 
@@ -501,20 +531,9 @@ _mesa_update_state_locked( GLcontext *ctx )
    if (new_state & _MESA_NEW_NEED_EYE_COORDS) 
       _mesa_update_tnl_spaces( ctx, new_state );
 
-   if (ctx->FragmentProgram._MaintainTexEnvProgram) {
-      prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR);
-   }
-   if (ctx->VertexProgram._MaintainTnlProgram) {
-      prog_flags |= (_NEW_ARRAY | _NEW_TEXTURE | _NEW_TEXTURE_MATRIX |
-                     _NEW_TRANSFORM | _NEW_POINT |
-                     _NEW_FOG | _NEW_LIGHT |
-                     _MESA_NEW_NEED_EYE_COORDS);
-   }
    if (new_state & prog_flags)
       update_program( ctx );
 
-
-
    /*
     * Give the driver a chance to act upon the new_state flags.
     * The driver might plug in different span functions, for example.
@@ -524,6 +543,7 @@ _mesa_update_state_locked( GLcontext *ctx )
     * Set ctx->NewState to zero to avoid recursion if
     * Driver.UpdateState() has to call FLUSH_VERTICES().  (fixed?)
     */
+ out:
    new_state = ctx->NewState;
    ctx->NewState = 0;
    ctx->Driver.UpdateState(ctx, new_state);
@@ -540,3 +560,59 @@ _mesa_update_state( GLcontext *ctx )
    _mesa_update_state_locked(ctx);
    _mesa_unlock_context_textures(ctx);
 }
+
+
+
+
+/**
+ * Want to figure out which fragment program inputs are actually
+ * constant/current values from ctx->Current.  These should be
+ * referenced as a tracked state variable rather than a fragment
+ * program input, to save the overhead of putting a constant value in
+ * every submitted vertex, transferring it to hardware, interpolating
+ * it across the triangle, etc...
+ *
+ * When there is a VP bound, just use vp->outputs.  But when we're
+ * generating vp from fixed function state, basically want to
+ * calculate:
+ *
+ * vp_out_2_fp_in( vp_in_2_vp_out( varying_inputs ) | 
+ *                 potential_vp_outputs )
+ *
+ * Where potential_vp_outputs is calculated by looking at enabled
+ * texgen, etc.
+ * 
+ * The generated fragment program should then only declare inputs that
+ * may vary or otherwise differ from the ctx->Current values.
+ * Otherwise, the fp should track them as state values instead.
+ */
+void
+_mesa_set_varying_vp_inputs( GLcontext *ctx,
+                             GLbitfield varying_inputs )
+{
+   if (ctx->varying_vp_inputs != varying_inputs) {
+      ctx->varying_vp_inputs = varying_inputs;
+      ctx->NewState |= _NEW_ARRAY;
+      /*_mesa_printf("%s %x\n", __FUNCTION__, varying_inputs);*/
+   }
+}
+
+
+/**
+ * Used by drivers to tell core Mesa that the driver is going to
+ * install/ use its own vertex program.  In particular, this will
+ * prevent generated fragment programs from using state vars instead
+ * of ordinary varyings/inputs.
+ */
+void
+_mesa_set_vp_override(GLcontext *ctx, GLboolean flag)
+{
+   if (ctx->VertexProgram._Overriden != flag) {
+      ctx->VertexProgram._Overriden = flag;
+
+      /* Set one of the bits which will trigger fragment program
+       * regeneration:
+       */
+      ctx->NewState |= _NEW_ARRAY; 
+   }
+}
diff --git a/src/mesa/main/state.h b/src/mesa/main/state.h
index bb7cb8f32a..29db08a0b9 100644
--- a/src/mesa/main/state.h
+++ b/src/mesa/main/state.h
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.3
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
@@ -29,13 +29,21 @@
 #include "mtypes.h"
 
 extern void
-_mesa_update_state( GLcontext *ctx );
+_mesa_update_state(GLcontext *ctx);
 
 /* As above but can only be called between _mesa_lock_context_textures() and 
  * _mesa_unlock_context_textures().
  */
 extern void
-_mesa_update_state_locked( GLcontext *ctx );
+_mesa_update_state_locked(GLcontext *ctx);
+
+
+extern void
+_mesa_set_varying_vp_inputs(GLcontext *ctx, GLbitfield varying_inputs);
+
+
+extern void
+_mesa_set_vp_override(GLcontext *ctx, GLboolean flag);
 
 
 #endif
diff --git a/src/mesa/main/stencil.c b/src/mesa/main/stencil.c
index 2a4c38b1f2..b4ea9978d8 100644
--- a/src/mesa/main/stencil.c
+++ b/src/mesa/main/stencil.c
@@ -27,21 +27,6 @@
  * \file stencil.c
  * Stencil operations.
  *
- * Note: There's an incompatibility between GL_EXT_stencil_two_side and
- * OpenGL 2.0's two-sided stencil feature.
- *
- * With GL_EXT_stencil_two_side, calling glStencilOp/Func/Mask() only the
- * front OR back face state (as set by glActiveStencilFaceEXT) is set.
- *
- * But with OpenGL 2.0, calling glStencilOp/Func/Mask() sets BOTH the
- * front AND back state.
- *
- * So either we advertise the GL_EXT_stencil_two_side extension, or OpenGL
- * 2.0, but not both.
- *
- * Also, note that GL_ATI_separate_stencil is different as well:
- * glStencilFuncSeparateATI(GLenum frontfunc, GLenum backfunc, ...)  vs.
- * glStencilFuncSeparate(GLenum face, GLenum func, ...).
  */
 
 
@@ -198,6 +183,7 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask )
 {
    GET_CURRENT_CONTEXT(ctx);
    const GLint stencilMax = (1 << ctx->DrawBuffer->Visual.stencilBits) - 1;
+   const GLint face = ctx->Stencil.ActiveFace;
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
    if (!validate_stencil_func(ctx, func)) {
@@ -207,9 +193,7 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask )
 
    ref = CLAMP( ref, 0, stencilMax );
 
-   if (ctx->Extensions.EXT_stencil_two_side) {
-      /* only set active face state */
-      const GLint face = ctx->Stencil.ActiveFace;
+   if (face != 0) {
       if (ctx->Stencil.Function[face] == func &&
           ctx->Stencil.ValueMask[face] == mask &&
           ctx->Stencil.Ref[face] == ref)
@@ -218,9 +202,12 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask )
       ctx->Stencil.Function[face] = func;
       ctx->Stencil.Ref[face] = ref;
       ctx->Stencil.ValueMask[face] = mask;
-      if (ctx->Driver.StencilFuncSeparate) {
-         ctx->Driver.StencilFuncSeparate(ctx, face ? GL_BACK : GL_FRONT,
-                                         func, ref, mask);
+
+      /* Only propagate the change to the driver if EXT_stencil_two_side
+       * is enabled.
+       */
+      if (ctx->Driver.StencilFuncSeparate && ctx->Stencil.TestTwoSide) {
+         ctx->Driver.StencilFuncSeparate(ctx, GL_BACK, func, ref, mask);
       }
    }
    else {
@@ -237,7 +224,9 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask )
       ctx->Stencil.Ref[0]       = ctx->Stencil.Ref[1]       = ref;
       ctx->Stencil.ValueMask[0] = ctx->Stencil.ValueMask[1] = mask;
       if (ctx->Driver.StencilFuncSeparate) {
-         ctx->Driver.StencilFuncSeparate(ctx, GL_FRONT_AND_BACK,
+         ctx->Driver.StencilFuncSeparate(ctx,
+					 ((ctx->Stencil.TestTwoSide)
+					  ? GL_FRONT : GL_FRONT_AND_BACK),
                                          func, ref, mask);
       }
    }
@@ -259,17 +248,23 @@ void GLAPIENTRY
 _mesa_StencilMask( GLuint mask )
 {
    GET_CURRENT_CONTEXT(ctx);
+   const GLint face = ctx->Stencil.ActiveFace;
+
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   if (ctx->Extensions.EXT_stencil_two_side) {
-      /* only set active face state */
-      const GLint face = ctx->Stencil.ActiveFace;
+   if (face != 0) {
+      /* Only modify the EXT_stencil_two_side back-face state.
+       */
       if (ctx->Stencil.WriteMask[face] == mask)
          return;
       FLUSH_VERTICES(ctx, _NEW_STENCIL);
       ctx->Stencil.WriteMask[face] = mask;
-      if (ctx->Driver.StencilMaskSeparate) {
-         ctx->Driver.StencilMaskSeparate(ctx, face ? GL_BACK : GL_FRONT, mask);
+
+      /* Only propagate the change to the driver if EXT_stencil_two_side
+       * is enabled.
+       */
+      if (ctx->Driver.StencilMaskSeparate && ctx->Stencil.TestTwoSide) {
+         ctx->Driver.StencilMaskSeparate(ctx, GL_BACK, mask);
       }
    }
    else {
@@ -280,7 +275,10 @@ _mesa_StencilMask( GLuint mask )
       FLUSH_VERTICES(ctx, _NEW_STENCIL);
       ctx->Stencil.WriteMask[0] = ctx->Stencil.WriteMask[1] = mask;
       if (ctx->Driver.StencilMaskSeparate) {
-         ctx->Driver.StencilMaskSeparate(ctx, GL_FRONT_AND_BACK, mask);
+         ctx->Driver.StencilMaskSeparate(ctx,
+					 ((ctx->Stencil.TestTwoSide)
+					  ? GL_FRONT : GL_FRONT_AND_BACK),
+					  mask);
       }
    }
 }
@@ -304,6 +302,8 @@ void GLAPIENTRY
 _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass)
 {
    GET_CURRENT_CONTEXT(ctx);
+   const GLint face = ctx->Stencil.ActiveFace;
+
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
    if (!validate_stencil_op(ctx, fail)) {
@@ -319,9 +319,8 @@ _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass)
       return;
    }
 
-   if (ctx->Extensions.EXT_stencil_two_side) {
+   if (face != 0) {
       /* only set active face state */
-      const GLint face = ctx->Stencil.ActiveFace;
       if (ctx->Stencil.ZFailFunc[face] == zfail &&
           ctx->Stencil.ZPassFunc[face] == zpass &&
           ctx->Stencil.FailFunc[face] == fail)
@@ -330,9 +329,12 @@ _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass)
       ctx->Stencil.ZFailFunc[face] = zfail;
       ctx->Stencil.ZPassFunc[face] = zpass;
       ctx->Stencil.FailFunc[face] = fail;
-      if (ctx->Driver.StencilOpSeparate) {
-         ctx->Driver.StencilOpSeparate(ctx, face ? GL_BACK : GL_FRONT,
-                                       fail, zfail, zpass);
+
+      /* Only propagate the change to the driver if EXT_stencil_two_side
+       * is enabled.
+       */
+      if (ctx->Driver.StencilOpSeparate && ctx->Stencil.TestTwoSide) {
+         ctx->Driver.StencilOpSeparate(ctx, GL_BACK, fail, zfail, zpass);
       }
    }
    else {
@@ -349,7 +351,9 @@ _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass)
       ctx->Stencil.ZPassFunc[0] = ctx->Stencil.ZPassFunc[1] = zpass;
       ctx->Stencil.FailFunc[0]  = ctx->Stencil.FailFunc[1]  = fail;
       if (ctx->Driver.StencilOpSeparate) {
-         ctx->Driver.StencilOpSeparate(ctx, GL_FRONT_AND_BACK,
+         ctx->Driver.StencilOpSeparate(ctx,
+				       ((ctx->Stencil.TestTwoSide)
+					? GL_FRONT : GL_FRONT_AND_BACK),
                                        fail, zfail, zpass);
       }
    }
@@ -372,7 +376,7 @@ _mesa_ActiveStencilFaceEXT(GLenum face)
 
    if (face == GL_FRONT || face == GL_BACK) {
       FLUSH_VERTICES(ctx, _NEW_STENCIL);
-      ctx->Stencil.ActiveFace = (face == GL_FRONT) ? 0 : 1;
+      ctx->Stencil.ActiveFace = (face == GL_FRONT) ? 0 : 2;
    }
    else {
       _mesa_error(ctx, GL_INVALID_ENUM, "glActiveStencilFaceEXT(face)");
@@ -513,19 +517,16 @@ _mesa_StencilMaskSeparate(GLenum face, GLuint mask)
 void
 _mesa_update_stencil(GLcontext *ctx)
 {
-   if (ctx->Extensions.EXT_stencil_two_side) {
-      ctx->Stencil._TestTwoSide = ctx->Stencil.TestTwoSide;
-   }
-   else {
-      ctx->Stencil._TestTwoSide = 
-         (ctx->Stencil.Function[0] != ctx->Stencil.Function[1] ||
-          ctx->Stencil.FailFunc[0] != ctx->Stencil.FailFunc[1] ||
-          ctx->Stencil.ZPassFunc[0] != ctx->Stencil.ZPassFunc[1] ||
-          ctx->Stencil.ZFailFunc[0] != ctx->Stencil.ZFailFunc[1] ||
-          ctx->Stencil.Ref[0] != ctx->Stencil.Ref[1] ||
-          ctx->Stencil.ValueMask[0] != ctx->Stencil.ValueMask[1] ||
-          ctx->Stencil.WriteMask[0] != ctx->Stencil.WriteMask[1]);
-   }
+   const GLint face = ctx->Stencil._BackFace;
+
+   ctx->Stencil._TestTwoSide =
+       (ctx->Stencil.Function[0] != ctx->Stencil.Function[face] ||
+	ctx->Stencil.FailFunc[0] != ctx->Stencil.FailFunc[face] ||
+	ctx->Stencil.ZPassFunc[0] != ctx->Stencil.ZPassFunc[face] ||
+	ctx->Stencil.ZFailFunc[0] != ctx->Stencil.ZFailFunc[face] ||
+	ctx->Stencil.Ref[0] != ctx->Stencil.Ref[face] ||
+	ctx->Stencil.ValueMask[0] != ctx->Stencil.ValueMask[face] ||
+	ctx->Stencil.WriteMask[0] != ctx->Stencil.WriteMask[face]);
 }
 
 
@@ -544,17 +545,24 @@ _mesa_init_stencil(GLcontext *ctx)
    ctx->Stencil.ActiveFace = 0;  /* 0 = GL_FRONT, 1 = GL_BACK */
    ctx->Stencil.Function[0] = GL_ALWAYS;
    ctx->Stencil.Function[1] = GL_ALWAYS;
+   ctx->Stencil.Function[2] = GL_ALWAYS;
    ctx->Stencil.FailFunc[0] = GL_KEEP;
    ctx->Stencil.FailFunc[1] = GL_KEEP;
+   ctx->Stencil.FailFunc[2] = GL_KEEP;
    ctx->Stencil.ZPassFunc[0] = GL_KEEP;
    ctx->Stencil.ZPassFunc[1] = GL_KEEP;
+   ctx->Stencil.ZPassFunc[2] = GL_KEEP;
    ctx->Stencil.ZFailFunc[0] = GL_KEEP;
    ctx->Stencil.ZFailFunc[1] = GL_KEEP;
+   ctx->Stencil.ZFailFunc[2] = GL_KEEP;
    ctx->Stencil.Ref[0] = 0;
    ctx->Stencil.Ref[1] = 0;
+   ctx->Stencil.Ref[2] = 0;
    ctx->Stencil.ValueMask[0] = ~0U;
    ctx->Stencil.ValueMask[1] = ~0U;
+   ctx->Stencil.ValueMask[2] = ~0U;
    ctx->Stencil.WriteMask[0] = ~0U;
    ctx->Stencil.WriteMask[1] = ~0U;
+   ctx->Stencil.WriteMask[2] = ~0U;
    ctx->Stencil.Clear = 0;
 }
diff --git a/src/mesa/main/texcompress.c b/src/mesa/main/texcompress.c
index c44d594d68..c1b8c7675a 100644
--- a/src/mesa/main/texcompress.c
+++ b/src/mesa/main/texcompress.c
@@ -3,6 +3,7 @@
  * Version:  6.5.1
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -56,60 +57,58 @@ GLuint
 _mesa_get_compressed_formats(GLcontext *ctx, GLint *formats, GLboolean all)
 {
    GLuint n = 0;
-   if (ctx->Extensions.ARB_texture_compression) {
-      if (ctx->Extensions.TDFX_texture_compression_FXT1) {
-         if (formats) {
-            formats[n++] = GL_COMPRESSED_RGB_FXT1_3DFX;
-            formats[n++] = GL_COMPRESSED_RGBA_FXT1_3DFX;
-         }
-         else {
-            n += 2;
-         }
+   if (ctx->Extensions.TDFX_texture_compression_FXT1) {
+      if (formats) {
+         formats[n++] = GL_COMPRESSED_RGB_FXT1_3DFX;
+         formats[n++] = GL_COMPRESSED_RGBA_FXT1_3DFX;
       }
-      if (ctx->Extensions.EXT_texture_compression_s3tc) {
-         if (formats) {
-            formats[n++] = GL_COMPRESSED_RGB_S3TC_DXT1_EXT;
-            /* This format has some restrictions/limitations and so should
-             * not be returned via the GL_COMPRESSED_TEXTURE_FORMATS query.
-             * Specifically, all transparent pixels become black.  NVIDIA
-             * omits this format too.
-             */
-            if (all)
-               formats[n++] = GL_COMPRESSED_RGBA_S3TC_DXT1_EXT;
-            formats[n++] = GL_COMPRESSED_RGBA_S3TC_DXT3_EXT;
-            formats[n++] = GL_COMPRESSED_RGBA_S3TC_DXT5_EXT;
-         }
-         else {
-            n += 3;
-            if (all)
-               n += 1;
-         }
+      else {
+         n += 2;
       }
-      if (ctx->Extensions.S3_s3tc) {
-         if (formats) {
-            formats[n++] = GL_RGB_S3TC;
-            formats[n++] = GL_RGB4_S3TC;
-            formats[n++] = GL_RGBA_S3TC;
-            formats[n++] = GL_RGBA4_S3TC;
-         }
-         else {
-            n += 4;
-         }
+   }
+   if (ctx->Extensions.EXT_texture_compression_s3tc) {
+      if (formats) {
+         formats[n++] = GL_COMPRESSED_RGB_S3TC_DXT1_EXT;
+         /* This format has some restrictions/limitations and so should
+          * not be returned via the GL_COMPRESSED_TEXTURE_FORMATS query.
+          * Specifically, all transparent pixels become black.  NVIDIA
+          * omits this format too.
+          */
+         if (all)
+             formats[n++] = GL_COMPRESSED_RGBA_S3TC_DXT1_EXT;
+         formats[n++] = GL_COMPRESSED_RGBA_S3TC_DXT3_EXT;
+         formats[n++] = GL_COMPRESSED_RGBA_S3TC_DXT5_EXT;
+      }
+      else {
+         n += 3;
+         if (all)
+             n += 1;
+      }
+   }
+   if (ctx->Extensions.S3_s3tc) {
+      if (formats) {
+         formats[n++] = GL_RGB_S3TC;
+         formats[n++] = GL_RGB4_S3TC;
+         formats[n++] = GL_RGBA_S3TC;
+         formats[n++] = GL_RGBA4_S3TC;
+      }
+      else {
+         n += 4;
       }
+   }
 #if FEATURE_EXT_texture_sRGB
-      if (ctx->Extensions.EXT_texture_sRGB) {
-         if (formats) {
-            formats[n++] = GL_COMPRESSED_SRGB_S3TC_DXT1_EXT;
-            formats[n++] = GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT;
-            formats[n++] = GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT;
-            formats[n++] = GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT;
-         }
-         else {
-            n += 4;
-         }
+   if (ctx->Extensions.EXT_texture_sRGB) {
+      if (formats) {
+         formats[n++] = GL_COMPRESSED_SRGB_S3TC_DXT1_EXT;
+         formats[n++] = GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT;
+         formats[n++] = GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT;
+         formats[n++] = GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT;
+      }
+      else {
+         n += 4;
       }
-#endif /* FEATURE_EXT_texture_sRGB */
    }
+#endif /* FEATURE_EXT_texture_sRGB */
    return n;
 }
 
@@ -137,8 +136,10 @@ _mesa_compressed_texture_size( GLcontext *ctx,
 
    ASSERT(depth == 1);
    (void) depth;
+   (void) size;
 
    switch (mesaFormat) {
+#if FEATURE_texture_fxt1
    case MESA_FORMAT_RGB_FXT1:
    case MESA_FORMAT_RGBA_FXT1:
       /* round up width to next multiple of 8, height to next multiple of 4 */
@@ -149,11 +150,15 @@ _mesa_compressed_texture_size( GLcontext *ctx,
       /* Textures smaller than 8x4 will effectively be made into 8x4 and
        * take 16 bytes.
        */
-      if (size < 16)
-         size = 16;
       return size;
+#endif
+#if FEATURE_texture_s3tc
    case MESA_FORMAT_RGB_DXT1:
    case MESA_FORMAT_RGBA_DXT1:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+#endif
       /* round up width, height to next multiple of 4 */
       width = (width + 3) & ~3;
       height = (height + 3) & ~3;
@@ -162,11 +167,13 @@ _mesa_compressed_texture_size( GLcontext *ctx,
       /* Textures smaller than 4x4 will effectively be made into 4x4 and
        * take 8 bytes.
        */
-      if (size < 8)
-         size = 8;
       return size;
    case MESA_FORMAT_RGBA_DXT3:
    case MESA_FORMAT_RGBA_DXT5:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
+#endif
       /* round up width, height to next multiple of 4 */
       width = (width + 3) & ~3;
       height = (height + 3) & ~3;
@@ -175,9 +182,8 @@ _mesa_compressed_texture_size( GLcontext *ctx,
       /* Textures smaller than 4x4 will effectively be made into 4x4 and
        * take 16 bytes.
        */
-      if (size < 16)
-         size = 16;
       return size;
+#endif
    default:
       _mesa_problem(ctx, "bad mesaFormat in _mesa_compressed_texture_size");
       return 0;
@@ -202,12 +208,15 @@ _mesa_compressed_texture_size_glenum(GLcontext *ctx,
    GLuint mesaFormat;
 
    switch (glformat) {
+#if FEATURE_texture_fxt1
    case GL_COMPRESSED_RGB_FXT1_3DFX:
       mesaFormat = MESA_FORMAT_RGB_FXT1;
       break;
    case GL_COMPRESSED_RGBA_FXT1_3DFX:
       mesaFormat = MESA_FORMAT_RGBA_FXT1;
       break;
+#endif
+#if FEATURE_texture_s3tc
    case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
    case GL_RGB_S3TC:
       mesaFormat = MESA_FORMAT_RGB_DXT1;
@@ -224,6 +233,21 @@ _mesa_compressed_texture_size_glenum(GLcontext *ctx,
    case GL_RGBA4_S3TC:
       mesaFormat = MESA_FORMAT_RGBA_DXT5;
       break;
+#if FEATURE_EXT_texture_sRGB
+   case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
+      mesaFormat = MESA_FORMAT_SRGB_DXT1;
+      break;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
+      mesaFormat = MESA_FORMAT_SRGBA_DXT1;
+      break;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
+      mesaFormat = MESA_FORMAT_SRGBA_DXT3;
+      break;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
+      mesaFormat = MESA_FORMAT_SRGBA_DXT5;
+      break;
+#endif
+#endif
    default:
       return 0;
    }
@@ -245,18 +269,30 @@ _mesa_compressed_row_stride(GLuint mesaFormat, GLsizei width)
    GLint stride;
 
    switch (mesaFormat) {
+#if FEATURE_texture_fxt1
    case MESA_FORMAT_RGB_FXT1:
    case MESA_FORMAT_RGBA_FXT1:
       stride = ((width + 7) / 8) * 16; /* 16 bytes per 8x4 tile */
       break;
+#endif
+#if FEATURE_texture_s3tc
    case MESA_FORMAT_RGB_DXT1:
    case MESA_FORMAT_RGBA_DXT1:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+#endif
       stride = ((width + 3) / 4) * 8; /* 8 bytes per 4x4 tile */
       break;
    case MESA_FORMAT_RGBA_DXT3:
    case MESA_FORMAT_RGBA_DXT5:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
+#endif
       stride = ((width + 3) / 4) * 16; /* 16 bytes per 4x4 tile */
       break;
+#endif
    default:
       _mesa_problem(NULL, "bad mesaFormat in _mesa_compressed_row_stride");
       return 0;
@@ -293,18 +329,30 @@ _mesa_compressed_image_address(GLint col, GLint row, GLint img,
     */
 
    switch (mesaFormat) {
+#if FEATURE_texture_fxt1
    case MESA_FORMAT_RGB_FXT1:
    case MESA_FORMAT_RGBA_FXT1:
       addr = (GLubyte *) image + 16 * (((width + 7) / 8) * (row / 4) + col / 8);
       break;
+#endif
+#if FEATURE_texture_s3tc
    case MESA_FORMAT_RGB_DXT1:
    case MESA_FORMAT_RGBA_DXT1:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+#endif
       addr = (GLubyte *) image + 8 * (((width + 3) / 4) * (row / 4) + col / 4);
       break;
    case MESA_FORMAT_RGBA_DXT3:
    case MESA_FORMAT_RGBA_DXT5:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
+#endif
       addr = (GLubyte *) image + 16 * (((width + 3) / 4) * (row / 4) + col / 4);
       break;
+#endif
    default:
       _mesa_problem(NULL, "bad mesaFormat in _mesa_compressed_image_address");
       addr = NULL;
diff --git a/src/mesa/main/texcompress_fxt1.c b/src/mesa/main/texcompress_fxt1.c
index 45f344b0c5..fc151605c9 100644
--- a/src/mesa/main/texcompress_fxt1.c
+++ b/src/mesa/main/texcompress_fxt1.c
@@ -298,17 +298,17 @@ const struct gl_texture_format _mesa_texformat_rgba_fxt1 = {
 /*
  * Define a 64-bit unsigned integer type and macros
  */
-#ifdef GL_EXT_timer_query  /* this extensions defines the GLuint64EXT type */
+#if 1
 
 #define FX64_NATIVE 1
 
-typedef GLuint64EXT Fx64;
+typedef uint64_t Fx64;
 
 #define FX64_MOV32(a, b) a = b
 #define FX64_OR32(a, b)  a |= b
 #define FX64_SHL(a, c)   a <<= c
 
-#else  /* !GL_EXT_timer_query */
+#else
 
 #define FX64_NATIVE 0
 
@@ -330,7 +330,7 @@ typedef struct {
        }                                               \
    } while (0)
 
-#endif  /* !GL_EXT_timer_query */
+#endif
 
 
 #define F(i) (GLfloat)1 /* can be used to obtain an oblong metric: 0.30 / 0.59 / 0.11 */
diff --git a/src/mesa/main/texcompress_s3tc.c b/src/mesa/main/texcompress_s3tc.c
index 4f329cdf59..d17e18da6b 100644
--- a/src/mesa/main/texcompress_s3tc.c
+++ b/src/mesa/main/texcompress_s3tc.c
@@ -3,6 +3,7 @@
  * Version:  6.5.3
  *
  * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -29,7 +30,7 @@
  */
 
 #ifndef USE_EXTERNAL_DXTN_LIB
-#define USE_EXTERNAL_DXTN_LIB 0
+#define USE_EXTERNAL_DXTN_LIB 1
 #endif
 
 #include "glheader.h"
@@ -37,15 +38,12 @@
 #include "colormac.h"
 #include "context.h"
 #include "convolve.h"
+#include "dlopen.h"
 #include "image.h"
 #include "texcompress.h"
 #include "texformat.h"
 #include "texstore.h"
 
-#if USE_EXTERNAL_DXTN_LIB && !defined(__MINGW32__)
-#include <dlfcn.h>
-#endif
-
 #ifdef __MINGW32__
 #define DXTN_LIBNAME "dxtn.dll"
 #define RTLD_LAZY 0
@@ -56,6 +54,34 @@
 #define DXTN_LIBNAME "libtxc_dxtn.so"
 #endif
 
+#if FEATURE_EXT_texture_sRGB
+/**
+ * Convert an 8-bit sRGB value from non-linear space to a
+ * linear RGB value in [0, 1].
+ * Implemented with a 256-entry lookup table.
+ */
+static INLINE GLfloat
+nonlinear_to_linear(GLubyte cs8)
+{
+   static GLfloat table[256];
+   static GLboolean tableReady = GL_FALSE;
+   if (!tableReady) {
+      /* compute lookup table now */
+      GLuint i;
+      for (i = 0; i < 256; i++) {
+         const GLfloat cs = UBYTE_TO_FLOAT(i);
+         if (cs <= 0.04045) {
+            table[i] = cs / 12.92f;
+         }
+         else {
+            table[i] = (GLfloat) _mesa_pow((cs + 0.055) / 1.055, 2.4);
+         }
+      }
+      tableReady = GL_TRUE;
+   }
+   return table[cs8];
+}
+#endif /* FEATURE_EXT_texture_sRGB */
 
 typedef void (*dxtFetchTexelFuncExt)( GLint srcRowstride, GLubyte *pixdata, GLint col, GLint row, GLvoid *texelOut );
 
@@ -74,72 +100,6 @@ static dxtCompressTexFuncExt ext_tx_compress_dxtn = NULL;
 static void *dxtlibhandle = NULL;
 
 
-typedef void (*GenericFunc)(void);
-
-
-/**
- * Wrapper for dlopen().
- * XXX Probably move this and the following wrappers into imports.h someday.
- */
-static void *
-_mesa_dlopen(const char *libname, int flags)
-{
-#if USE_EXTERNAL_DXTN_LIB
-#ifdef __MINGW32__
-   return LoadLibrary(libname);
-#else
-   return dlopen(libname, flags);
-#endif
-#else
-   return NULL;
-#endif /* USE_EXTERNAL_DXTN_LIB */
-}
-
-
-/**
- * Wrapper for dlsym() that does a cast to a generic function type,
- * rather than a void *.  This reduces the number of warnings that are
- * generated.
- */
-static GenericFunc
-_mesa_dlsym(void *handle, const char *fname)
-{
-#if USE_EXTERNAL_DXTN_LIB
-#ifdef __MINGW32__
-   return (GenericFunc) GetProcAddress(handle, fname);
-#elif defined(__DJGPP__)
-   /* need '_' prefix on symbol names */
-   char fname2[1000];
-   fname2[0] = '_';
-   _mesa_strncpy(fname2 + 1, fname, 998);
-   fname2[999] = 0;
-   return (GenericFunc) dlsym(handle, fname2);
-#else
-   return (GenericFunc) dlsym(handle, fname);
-#endif
-#else
-   return (GenericFunc) NULL;
-#endif /* USE_EXTERNAL_DXTN_LIB */
-}
-
-
-/**
- * Wrapper for dlclose().
- */
-static void
-_mesa_dlclose(void *handle)
-{
-#if USE_EXTERNAL_DXTN_LIB
-#ifdef __MINGW32__
-   FreeLibrary(handle);
-#else
-   dlclose(handle);
-#endif
-#endif
-}
-
-
-
 void
 _mesa_init_texture_s3tc( GLcontext *ctx )
 {
@@ -147,7 +107,7 @@ _mesa_init_texture_s3tc( GLcontext *ctx )
    ctx->Mesa_DXTn = GL_FALSE;
 #if USE_EXTERNAL_DXTN_LIB
    if (!dxtlibhandle) {
-      dxtlibhandle = _mesa_dlopen(DXTN_LIBNAME, RTLD_LAZY | RTLD_GLOBAL);
+      dxtlibhandle = _mesa_dlopen(DXTN_LIBNAME, 0);
       if (!dxtlibhandle) {
 	 _mesa_warning(ctx, "couldn't open " DXTN_LIBNAME ", software DXTn "
 	    "compression/decompression unavailable");
@@ -552,6 +512,59 @@ fetch_texel_2d_f_rgba_dxt5( const struct gl_texture_image *texImage,
    texel[ACOMP] = CHAN_TO_FLOAT(rgba[ACOMP]);
 }
 
+#if FEATURE_EXT_texture_sRGB
+static void
+fetch_texel_2d_f_srgb_dxt1( const struct gl_texture_image *texImage,
+                            GLint i, GLint j, GLint k, GLfloat *texel )
+{
+   /* just sample as GLchan and convert to float here */
+   GLchan rgba[4];
+   fetch_texel_2d_rgb_dxt1(texImage, i, j, k, rgba);
+   texel[RCOMP] = nonlinear_to_linear(rgba[RCOMP]);
+   texel[GCOMP] = nonlinear_to_linear(rgba[GCOMP]);
+   texel[BCOMP] = nonlinear_to_linear(rgba[BCOMP]);
+   texel[ACOMP] = CHAN_TO_FLOAT(rgba[ACOMP]);
+}
+
+static void
+fetch_texel_2d_f_srgba_dxt1( const struct gl_texture_image *texImage,
+                             GLint i, GLint j, GLint k, GLfloat *texel )
+{
+   /* just sample as GLchan and convert to float here */
+   GLchan rgba[4];
+   fetch_texel_2d_rgba_dxt1(texImage, i, j, k, rgba);
+   texel[RCOMP] = nonlinear_to_linear(rgba[RCOMP]);
+   texel[GCOMP] = nonlinear_to_linear(rgba[GCOMP]);
+   texel[BCOMP] = nonlinear_to_linear(rgba[BCOMP]);
+   texel[ACOMP] = CHAN_TO_FLOAT(rgba[ACOMP]);
+}
+
+static void
+fetch_texel_2d_f_srgba_dxt3( const struct gl_texture_image *texImage,
+                             GLint i, GLint j, GLint k, GLfloat *texel )
+{
+   /* just sample as GLchan and convert to float here */
+   GLchan rgba[4];
+   fetch_texel_2d_rgba_dxt3(texImage, i, j, k, rgba);
+   texel[RCOMP] = nonlinear_to_linear(rgba[RCOMP]);
+   texel[GCOMP] = nonlinear_to_linear(rgba[GCOMP]);
+   texel[BCOMP] = nonlinear_to_linear(rgba[BCOMP]);
+   texel[ACOMP] = CHAN_TO_FLOAT(rgba[ACOMP]);
+}
+
+static void
+fetch_texel_2d_f_srgba_dxt5( const struct gl_texture_image *texImage,
+                             GLint i, GLint j, GLint k, GLfloat *texel )
+{
+   /* just sample as GLchan and convert to float here */
+   GLchan rgba[4];
+   fetch_texel_2d_rgba_dxt5(texImage, i, j, k, rgba);
+   texel[RCOMP] = nonlinear_to_linear(rgba[RCOMP]);
+   texel[GCOMP] = nonlinear_to_linear(rgba[GCOMP]);
+   texel[BCOMP] = nonlinear_to_linear(rgba[BCOMP]);
+   texel[ACOMP] = CHAN_TO_FLOAT(rgba[ACOMP]);
+}
+#endif
 
 const struct gl_texture_format _mesa_texformat_rgb_dxt1 = {
    MESA_FORMAT_RGB_DXT1,		/* MesaFormat */
@@ -577,6 +590,78 @@ const struct gl_texture_format _mesa_texformat_rgb_dxt1 = {
    NULL					/* StoreTexel */
 };
 
+const struct gl_texture_format _mesa_texformat_rgba_dxt1 = {
+   MESA_FORMAT_RGBA_DXT1,		/* MesaFormat */
+   GL_RGBA,				/* BaseFormat */
+   GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
+   4, /*approx*/			/* RedBits */
+   4, /*approx*/			/* GreenBits */
+   4, /*approx*/			/* BlueBits */
+   1, /*approx*/			/* AlphaBits */
+   0,					/* LuminanceBits */
+   0,					/* IntensityBits */
+   0,					/* IndexBits */
+   0,					/* DepthBits */
+   0,					/* StencilBits */
+   0,					/* TexelBytes */
+   texstore_rgba_dxt1,			/* StoreTexImageFunc */
+   NULL, /*impossible*/ 		/* FetchTexel1D */
+   fetch_texel_2d_rgba_dxt1, 		/* FetchTexel2D */
+   NULL, /*impossible*/ 		/* FetchTexel3D */
+   NULL, /*impossible*/ 		/* FetchTexel1Df */
+   fetch_texel_2d_f_rgba_dxt1, 		/* FetchTexel2Df */
+   NULL, /*impossible*/ 		/* FetchTexel3Df */
+   NULL					/* StoreTexel */
+};
+
+const struct gl_texture_format _mesa_texformat_rgba_dxt3 = {
+   MESA_FORMAT_RGBA_DXT3,		/* MesaFormat */
+   GL_RGBA,				/* BaseFormat */
+   GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
+   4, /*approx*/			/* RedBits */
+   4, /*approx*/			/* GreenBits */
+   4, /*approx*/			/* BlueBits */
+   4, /*approx*/			/* AlphaBits */
+   0,					/* LuminanceBits */
+   0,					/* IntensityBits */
+   0,					/* IndexBits */
+   0,					/* DepthBits */
+   0,					/* StencilBits */
+   0,					/* TexelBytes */
+   texstore_rgba_dxt3,			/* StoreTexImageFunc */
+   NULL, /*impossible*/ 		/* FetchTexel1D */
+   fetch_texel_2d_rgba_dxt3, 		/* FetchTexel2D */
+   NULL, /*impossible*/ 		/* FetchTexel3D */
+   NULL, /*impossible*/ 		/* FetchTexel1Df */
+   fetch_texel_2d_f_rgba_dxt3, 		/* FetchTexel2Df */
+   NULL, /*impossible*/ 		/* FetchTexel3Df */
+   NULL					/* StoreTexel */
+};
+
+const struct gl_texture_format _mesa_texformat_rgba_dxt5 = {
+   MESA_FORMAT_RGBA_DXT5,		/* MesaFormat */
+   GL_RGBA,				/* BaseFormat */
+   GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
+   4,/*approx*/				/* RedBits */
+   4,/*approx*/				/* GreenBits */
+   4,/*approx*/				/* BlueBits */
+   4,/*approx*/				/* AlphaBits */
+   0,					/* LuminanceBits */
+   0,					/* IntensityBits */
+   0,					/* IndexBits */
+   0,					/* DepthBits */
+   0,					/* StencilBits */
+   0,					/* TexelBytes */
+   texstore_rgba_dxt5,			/* StoreTexImageFunc */
+   NULL, /*impossible*/ 		/* FetchTexel1D */
+   fetch_texel_2d_rgba_dxt5, 		/* FetchTexel2D */
+   NULL, /*impossible*/ 		/* FetchTexel3D */
+   NULL, /*impossible*/ 		/* FetchTexel1Df */
+   fetch_texel_2d_f_rgba_dxt5, 		/* FetchTexel2Df */
+   NULL, /*impossible*/ 		/* FetchTexel3Df */
+   NULL					/* StoreTexel */
+};
+
 #if FEATURE_EXT_texture_sRGB
 const struct gl_texture_format _mesa_texformat_srgb_dxt1 = {
    MESA_FORMAT_SRGB_DXT1,		/* MesaFormat */
@@ -594,17 +679,16 @@ const struct gl_texture_format _mesa_texformat_srgb_dxt1 = {
    0,					/* TexelBytes */
    texstore_rgb_dxt1,			/* StoreTexImageFunc */
    NULL, /*impossible*/ 		/* FetchTexel1D */
-   fetch_texel_2d_rgb_dxt1, 		/* FetchTexel2D */
+   NULL, 				/* FetchTexel2D */
    NULL, /*impossible*/ 		/* FetchTexel3D */
    NULL, /*impossible*/ 		/* FetchTexel1Df */
-   fetch_texel_2d_f_rgb_dxt1, 		/* FetchTexel2Df */
+   fetch_texel_2d_f_srgb_dxt1, 		/* FetchTexel2Df */
    NULL, /*impossible*/ 		/* FetchTexel3Df */
    NULL					/* StoreTexel */
 };
-#endif
 
-const struct gl_texture_format _mesa_texformat_rgba_dxt1 = {
-   MESA_FORMAT_RGBA_DXT1,		/* MesaFormat */
+const struct gl_texture_format _mesa_texformat_srgba_dxt1 = {
+   MESA_FORMAT_SRGBA_DXT1,		/* MesaFormat */
    GL_RGBA,				/* BaseFormat */
    GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
    4, /*approx*/			/* RedBits */
@@ -619,16 +703,16 @@ const struct gl_texture_format _mesa_texformat_rgba_dxt1 = {
    0,					/* TexelBytes */
    texstore_rgba_dxt1,			/* StoreTexImageFunc */
    NULL, /*impossible*/ 		/* FetchTexel1D */
-   fetch_texel_2d_rgba_dxt1, 		/* FetchTexel2D */
+   NULL, 				/* FetchTexel2D */
    NULL, /*impossible*/ 		/* FetchTexel3D */
    NULL, /*impossible*/ 		/* FetchTexel1Df */
-   fetch_texel_2d_f_rgba_dxt1, 		/* FetchTexel2Df */
+   fetch_texel_2d_f_srgba_dxt1, 	/* FetchTexel2Df */
    NULL, /*impossible*/ 		/* FetchTexel3Df */
    NULL					/* StoreTexel */
 };
 
-const struct gl_texture_format _mesa_texformat_rgba_dxt3 = {
-   MESA_FORMAT_RGBA_DXT3,		/* MesaFormat */
+const struct gl_texture_format _mesa_texformat_srgba_dxt3 = {
+   MESA_FORMAT_SRGBA_DXT3,		/* MesaFormat */
    GL_RGBA,				/* BaseFormat */
    GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
    4, /*approx*/			/* RedBits */
@@ -643,16 +727,16 @@ const struct gl_texture_format _mesa_texformat_rgba_dxt3 = {
    0,					/* TexelBytes */
    texstore_rgba_dxt3,			/* StoreTexImageFunc */
    NULL, /*impossible*/ 		/* FetchTexel1D */
-   fetch_texel_2d_rgba_dxt3, 		/* FetchTexel2D */
+   NULL, 				/* FetchTexel2D */
    NULL, /*impossible*/ 		/* FetchTexel3D */
    NULL, /*impossible*/ 		/* FetchTexel1Df */
-   fetch_texel_2d_f_rgba_dxt3, 		/* FetchTexel2Df */
+   fetch_texel_2d_f_srgba_dxt3, 	/* FetchTexel2Df */
    NULL, /*impossible*/ 		/* FetchTexel3Df */
    NULL					/* StoreTexel */
 };
 
-const struct gl_texture_format _mesa_texformat_rgba_dxt5 = {
-   MESA_FORMAT_RGBA_DXT5,		/* MesaFormat */
+const struct gl_texture_format _mesa_texformat_srgba_dxt5 = {
+   MESA_FORMAT_SRGBA_DXT5,		/* MesaFormat */
    GL_RGBA,				/* BaseFormat */
    GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
    4,/*approx*/				/* RedBits */
@@ -667,10 +751,11 @@ const struct gl_texture_format _mesa_texformat_rgba_dxt5 = {
    0,					/* TexelBytes */
    texstore_rgba_dxt5,			/* StoreTexImageFunc */
    NULL, /*impossible*/ 		/* FetchTexel1D */
-   fetch_texel_2d_rgba_dxt5, 		/* FetchTexel2D */
+   NULL, 				/* FetchTexel2D */
    NULL, /*impossible*/ 		/* FetchTexel3D */
    NULL, /*impossible*/ 		/* FetchTexel1Df */
-   fetch_texel_2d_f_rgba_dxt5, 		/* FetchTexel2Df */
+   fetch_texel_2d_f_srgba_dxt5, 	/* FetchTexel2Df */
    NULL, /*impossible*/ 		/* FetchTexel3Df */
    NULL					/* StoreTexel */
 };
+#endif
diff --git a/src/mesa/main/texenv.c b/src/mesa/main/texenv.c
index e072cea136..097923182a 100644
--- a/src/mesa/main/texenv.c
+++ b/src/mesa/main/texenv.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.5
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -37,6 +38,415 @@
 #include "math/m_xform.h"
 
 
+#define TE_ERROR(errCode, msg, value)				\
+   _mesa_error(ctx, errCode, msg, _mesa_lookup_enum_by_nr(value));
+
+
+/** Set texture env mode */
+static void
+set_env_mode(GLcontext *ctx,
+             struct gl_texture_unit *texUnit,
+             GLenum mode)
+{
+   GLboolean legal;
+
+   if (texUnit->EnvMode == mode)
+      return;
+
+   switch (mode) {
+   case GL_MODULATE:
+   case GL_BLEND:
+   case GL_DECAL:
+   case GL_REPLACE:
+      legal = GL_TRUE;
+      break;
+   case GL_REPLACE_EXT:
+      mode = GL_REPLACE; /* GL_REPLACE_EXT != GL_REPLACE */
+      legal = GL_TRUE;
+      break;
+   case GL_ADD:
+      legal = ctx->Extensions.EXT_texture_env_add;
+      break;
+   case GL_COMBINE:
+      legal = (ctx->Extensions.EXT_texture_env_combine ||
+               ctx->Extensions.ARB_texture_env_combine);
+      break;
+   case GL_COMBINE4_NV:
+      legal = ctx->Extensions.NV_texture_env_combine4;
+      break;
+   default:
+      legal = GL_FALSE;
+   }
+
+   if (legal) {
+      FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+      texUnit->EnvMode = mode;
+   }
+   else {
+      TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
+   }
+}
+
+
+static void
+set_env_color(GLcontext *ctx,
+              struct gl_texture_unit *texUnit,
+              const GLfloat *color)
+{
+   GLfloat tmp[4];
+   tmp[0] = CLAMP(color[0], 0.0F, 1.0F);
+   tmp[1] = CLAMP(color[1], 0.0F, 1.0F);
+   tmp[2] = CLAMP(color[2], 0.0F, 1.0F);
+   tmp[3] = CLAMP(color[3], 0.0F, 1.0F);
+   if (TEST_EQ_4V(tmp, texUnit->EnvColor))
+      return;
+   FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+   COPY_4FV(texUnit->EnvColor, tmp);
+}
+
+
+/** Set an RGB or A combiner mode/function */
+static void
+set_combiner_mode(GLcontext *ctx,
+                  struct gl_texture_unit *texUnit,
+                  GLenum pname, GLenum mode)
+{
+   GLboolean legal;
+
+   if (!ctx->Extensions.EXT_texture_env_combine &&
+       !ctx->Extensions.ARB_texture_env_combine) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glTexEnv(pname)");
+      return;
+   }
+
+   switch (mode) {
+   case GL_REPLACE:
+   case GL_MODULATE:
+   case GL_ADD:
+   case GL_ADD_SIGNED:
+   case GL_INTERPOLATE:
+      legal = GL_TRUE;
+      break;
+   case GL_SUBTRACT:
+      legal = ctx->Extensions.ARB_texture_env_combine;
+      break;
+   case GL_DOT3_RGB_EXT:
+   case GL_DOT3_RGBA_EXT:
+      legal = (ctx->Extensions.EXT_texture_env_dot3 &&
+               pname == GL_COMBINE_RGB);
+      break;
+   case GL_DOT3_RGB:
+   case GL_DOT3_RGBA:
+      legal = (ctx->Extensions.ARB_texture_env_dot3 &&
+               pname == GL_COMBINE_RGB);
+      break;
+   case GL_MODULATE_ADD_ATI:
+   case GL_MODULATE_SIGNED_ADD_ATI:
+   case GL_MODULATE_SUBTRACT_ATI:
+      legal =ctx->Extensions.ATI_texture_env_combine3;
+      break;
+   default:
+      legal = GL_FALSE;
+   }
+
+   if (!legal) {
+      TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
+      return;
+   }
+
+   switch (pname) {
+   case GL_COMBINE_RGB:
+      if (texUnit->Combine.ModeRGB == mode)
+         return;
+      FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+      texUnit->Combine.ModeRGB = mode;
+      break;
+
+   case GL_COMBINE_ALPHA:
+      if (texUnit->Combine.ModeA == mode)
+         return;
+      FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+      texUnit->Combine.ModeA = mode;
+      break;
+   default:
+      TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
+   }
+}
+
+
+
+/** Set an RGB or A combiner source term */
+static void
+set_combiner_source(GLcontext *ctx,
+                    struct gl_texture_unit *texUnit,
+                    GLenum pname, GLenum param)
+{
+   GLuint src;
+   GLboolean alpha, legal;
+
+   if (!ctx->Extensions.EXT_texture_env_combine &&
+       !ctx->Extensions.ARB_texture_env_combine) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glTexEnv(pname)");
+      return;
+   }
+
+   /*
+    * Translate pname to (src, alpha).
+    */
+   switch (pname) {
+   case GL_SOURCE0_RGB:
+      src = 0;
+      alpha = GL_FALSE;
+      break;
+   case GL_SOURCE1_RGB:
+      src = 1;
+      alpha = GL_FALSE;
+      break;
+   case GL_SOURCE2_RGB:
+      src = 2;
+      alpha = GL_FALSE;
+      break;
+   case GL_SOURCE3_RGB_NV:
+      if (ctx->Extensions.NV_texture_env_combine4) {
+         src = 3;
+         alpha = GL_FALSE;
+      }
+      else {
+         TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
+         return;
+      }
+      break;
+   case GL_SOURCE0_ALPHA:
+      src = 0;
+      alpha = GL_TRUE;
+      break;
+   case GL_SOURCE1_ALPHA:
+      src = 1;
+      alpha = GL_TRUE;
+      break;
+   case GL_SOURCE2_ALPHA:
+      src = 2;
+      alpha = GL_TRUE;
+      break;
+   case GL_SOURCE3_ALPHA_NV:
+      if (ctx->Extensions.NV_texture_env_combine4) {
+         src = 3;
+         alpha = GL_TRUE;
+      }
+      else {
+         TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
+         return;
+      }
+      break;
+   default:
+      TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
+      return;
+   }
+
+   assert(src < 4);
+
+   /*
+    * Error-check param (the source term)
+    */
+   switch (param) {
+   case GL_TEXTURE:
+   case GL_CONSTANT:
+   case GL_PRIMARY_COLOR:
+   case GL_PREVIOUS:
+      legal = GL_TRUE;
+      break;
+   case GL_TEXTURE0:
+   case GL_TEXTURE1:
+   case GL_TEXTURE2:
+   case GL_TEXTURE3:
+   case GL_TEXTURE4:
+   case GL_TEXTURE5:
+   case GL_TEXTURE6:
+   case GL_TEXTURE7:
+      legal = (ctx->Extensions.ARB_texture_env_crossbar &&
+               param - GL_TEXTURE0 < ctx->Const.MaxTextureUnits);
+      break;
+   case GL_ZERO:
+      legal = (ctx->Extensions.ATI_texture_env_combine3 ||
+               ctx->Extensions.NV_texture_env_combine4);
+      break;
+   case GL_ONE:
+      legal = ctx->Extensions.ATI_texture_env_combine3;
+      break;
+   default:
+      legal = GL_FALSE;
+   }
+
+   if (!legal) {
+      TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", param);
+      return;
+   }
+
+   FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+
+   if (alpha)
+      texUnit->Combine.SourceA[src] = param;
+   else
+      texUnit->Combine.SourceRGB[src] = param;
+}
+
+
+/** Set an RGB or A combiner operand term */
+static void
+set_combiner_operand(GLcontext *ctx,
+                     struct gl_texture_unit *texUnit,
+                     GLenum pname, GLenum param)
+{
+   GLuint op;
+   GLboolean alpha, legal;
+
+   if (!ctx->Extensions.EXT_texture_env_combine &&
+       !ctx->Extensions.ARB_texture_env_combine) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glTexEnv(pname)");
+      return;
+   }
+
+   switch (pname) {
+   case GL_OPERAND0_RGB:
+      op = 0;
+      alpha = GL_FALSE;
+      break;
+   case GL_OPERAND1_RGB:
+      op = 1;
+      alpha = GL_FALSE;
+      break;
+   case GL_OPERAND2_RGB:
+      if (ctx->Extensions.ARB_texture_env_combine) {
+         op = 2;
+         alpha = GL_FALSE;
+      }
+      else {
+         TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
+         return;
+      }
+      break;
+   case GL_OPERAND3_RGB_NV:
+      if (ctx->Extensions.NV_texture_env_combine4) {
+         op = 3;
+         alpha = GL_FALSE;
+      }
+      else {
+         TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
+         return;
+      }
+      break;
+   case GL_OPERAND0_ALPHA:
+      op = 0;
+      alpha = GL_TRUE;
+      break;
+   case GL_OPERAND1_ALPHA:
+      op = 1;
+      alpha = GL_TRUE;
+      break;
+   case GL_OPERAND2_ALPHA:
+      if (ctx->Extensions.ARB_texture_env_combine) {
+         op = 2;
+         alpha = GL_TRUE;
+      }
+      else {
+         TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
+         return;
+      }
+      break;
+   case GL_OPERAND3_ALPHA_NV:
+      if (ctx->Extensions.NV_texture_env_combine4) {
+         op = 3;
+         alpha = GL_TRUE;
+      }
+      else {
+         TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
+         return;
+      }
+      break;
+   default:
+      TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
+      return;
+   }
+
+   assert(op < 4);
+
+   /*
+    * Error-check param (the source operand)
+    */
+   switch (param) {
+   case GL_SRC_COLOR:
+   case GL_ONE_MINUS_SRC_COLOR:
+      legal = !alpha;
+      break;
+   case GL_SRC_ALPHA:
+   case GL_ONE_MINUS_SRC_ALPHA:
+      legal = GL_TRUE;
+      break;
+   default:
+      legal = GL_FALSE;
+   }
+
+   if (!legal) {
+      TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", param);
+      return;
+   }
+
+   FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+
+   if (alpha)
+      texUnit->Combine.OperandA[op] = param;
+   else
+      texUnit->Combine.OperandRGB[op] = param;
+}
+
+
+static void
+set_combiner_scale(GLcontext *ctx,
+                   struct gl_texture_unit *texUnit,
+                   GLenum pname, GLfloat scale)
+{
+   GLuint shift;
+
+   if (!ctx->Extensions.EXT_texture_env_combine &&
+       !ctx->Extensions.ARB_texture_env_combine) {
+      _mesa_error(ctx, GL_INVALID_ENUM, "glTexEnv(pname)");
+      return;
+   }
+
+   if (scale == 1.0F) {
+      shift = 0;
+   }
+   else if (scale == 2.0F) {
+      shift = 1;
+   }
+   else if (scale == 4.0F) {
+      shift = 2;
+   }
+   else {
+      _mesa_error( ctx, GL_INVALID_VALUE,
+                   "glTexEnv(GL_RGB_SCALE not 1, 2 or 4)" );
+      return;
+   }
+
+   switch (pname) {
+   case GL_RGB_SCALE:
+      if (texUnit->Combine.ScaleShiftRGB == shift)
+         return;
+      FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+      texUnit->Combine.ScaleShiftRGB = shift;
+      break;
+   case GL_ALPHA_SCALE:
+      if (texUnit->Combine.ScaleShiftA == shift)
+         return;
+      FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+      texUnit->Combine.ScaleShiftA = shift;
+      break;
+   default:
+      TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
+   }
+}
+
+
 
 void GLAPIENTRY
 _mesa_TexEnvfv( GLenum target, GLenum pname, const GLfloat *param )
@@ -55,377 +465,41 @@ _mesa_TexEnvfv( GLenum target, GLenum pname, const GLfloat *param )
 
    texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
 
-#define TE_ERROR(errCode, msg, value)				\
-   _mesa_error(ctx, errCode, msg, _mesa_lookup_enum_by_nr(value));
-
    if (target == GL_TEXTURE_ENV) {
       switch (pname) {
       case GL_TEXTURE_ENV_MODE:
-         {
-            GLenum mode = (GLenum) (GLint) *param;
-            if (mode == GL_REPLACE_EXT)
-               mode = GL_REPLACE;
-	    if (texUnit->EnvMode == mode)
-	       return;
-            if (mode == GL_MODULATE ||
-                mode == GL_BLEND ||
-                mode == GL_DECAL ||
-                mode == GL_REPLACE ||
-                (mode == GL_ADD && ctx->Extensions.EXT_texture_env_add) ||
-                (mode == GL_COMBINE &&
-                 (ctx->Extensions.EXT_texture_env_combine ||
-                  ctx->Extensions.ARB_texture_env_combine))) {
-               /* legal */
-               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-               texUnit->EnvMode = mode;
-            }
-            else {
-               TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
-               return;
-            }
-         }
+         set_env_mode(ctx, texUnit, (GLenum) (GLint) param[0]);
          break;
       case GL_TEXTURE_ENV_COLOR:
-         {
-            GLfloat tmp[4];
-            tmp[0] = CLAMP( param[0], 0.0F, 1.0F );
-            tmp[1] = CLAMP( param[1], 0.0F, 1.0F );
-            tmp[2] = CLAMP( param[2], 0.0F, 1.0F );
-            tmp[3] = CLAMP( param[3], 0.0F, 1.0F );
-            if (TEST_EQ_4V(tmp, texUnit->EnvColor))
-               return;
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            COPY_4FV(texUnit->EnvColor, tmp);
-         }
+         set_env_color(ctx, texUnit, param);
          break;
       case GL_COMBINE_RGB:
-	 if (ctx->Extensions.EXT_texture_env_combine ||
-             ctx->Extensions.ARB_texture_env_combine) {
-	    const GLenum mode = (GLenum) (GLint) *param;
-	    if (texUnit->Combine.ModeRGB == mode)
-	       return;
-	    switch (mode) {
-	    case GL_REPLACE:
-	    case GL_MODULATE:
-	    case GL_ADD:
-	    case GL_ADD_SIGNED:
-	    case GL_INTERPOLATE:
-               /* OK */
-	       break;
-            case GL_SUBTRACT:
-               if (!ctx->Extensions.ARB_texture_env_combine) {
-                  TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
-                  return;
-               }
-               break;
-	    case GL_DOT3_RGB_EXT:
-	    case GL_DOT3_RGBA_EXT:
-	       if (!ctx->Extensions.EXT_texture_env_dot3) {
-                  TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
-		  return;
-	       }
-	       break;
-	    case GL_DOT3_RGB:
-	    case GL_DOT3_RGBA:
-	       if (!ctx->Extensions.ARB_texture_env_dot3) {
-                  TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
-		  return;
-	       }
-	       break;
-	    case GL_MODULATE_ADD_ATI:
-	    case GL_MODULATE_SIGNED_ADD_ATI:
-	    case GL_MODULATE_SUBTRACT_ATI:
-	       if (!ctx->Extensions.ATI_texture_env_combine3) {
-                  TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
-		  return;
-	       }
-	       break;
-	    default:
-               TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
-	       return;
-	    }
-	    FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-	    texUnit->Combine.ModeRGB = mode;
-	 }
-	 else {
-            TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
-	    return;
-	 }
-         break;
       case GL_COMBINE_ALPHA:
-	 if (ctx->Extensions.EXT_texture_env_combine ||
-             ctx->Extensions.ARB_texture_env_combine) {
-	    const GLenum mode = (GLenum) (GLint) *param;
-	    if (texUnit->Combine.ModeA == mode)
-	       return;
-            switch (mode) {
-	    case GL_REPLACE:
-	    case GL_MODULATE:
-	    case GL_ADD:
-	    case GL_ADD_SIGNED:
-	    case GL_INTERPOLATE:
-	       /* OK */
-	       break;
-	    case GL_SUBTRACT:
-	       if (!ctx->Extensions.ARB_texture_env_combine) {
-		  TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
-		  return;
-	       }
-	       break;
-	    case GL_MODULATE_ADD_ATI:
-	    case GL_MODULATE_SIGNED_ADD_ATI:
-	    case GL_MODULATE_SUBTRACT_ATI:
-	       if (!ctx->Extensions.ATI_texture_env_combine3) {
-                  TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
-		  return;
-	       }
-	       break;
-	    default:
-	       TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode);
-	       return;
-	    }
-	    FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-	    texUnit->Combine.ModeA = mode;
-	 }
-	 else {
-            TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
-	    return;
-	 }
+         set_combiner_mode(ctx, texUnit, pname, (GLenum) (GLint) param[0]);
 	 break;
       case GL_SOURCE0_RGB:
       case GL_SOURCE1_RGB:
       case GL_SOURCE2_RGB:
-	 if (ctx->Extensions.EXT_texture_env_combine ||
-	     ctx->Extensions.ARB_texture_env_combine) {
-	    const GLenum source = (GLenum) (GLint) *param;
-	    const GLuint s = pname - GL_SOURCE0_RGB;
-	    if (texUnit->Combine.SourceRGB[s] == source)
-	       return;
-            if (source == GL_TEXTURE ||
-                source == GL_CONSTANT ||
-                source == GL_PRIMARY_COLOR ||
-                source == GL_PREVIOUS ||
-                (ctx->Extensions.ARB_texture_env_crossbar &&
-                 source >= GL_TEXTURE0 &&
-                 source < GL_TEXTURE0 + ctx->Const.MaxTextureUnits) ||
-                (ctx->Extensions.ATI_texture_env_combine3 &&
-                 (source == GL_ZERO || source == GL_ONE))) {
-               /* legal */
-	       FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-	       texUnit->Combine.SourceRGB[s] = source;
-            }
-            else {
-               TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", source);
-	       return;
-	    }
-	 }
-	 else {
-            TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
-	    return;
-	 }
-	 break;
+      case GL_SOURCE3_RGB_NV:
       case GL_SOURCE0_ALPHA:
       case GL_SOURCE1_ALPHA:
       case GL_SOURCE2_ALPHA:
-	 if (ctx->Extensions.EXT_texture_env_combine ||
-             ctx->Extensions.ARB_texture_env_combine) {
-	    const GLenum source = (GLenum) (GLint) *param;
-	    const GLuint s = pname - GL_SOURCE0_ALPHA;
-	    if (texUnit->Combine.SourceA[s] == source)
-	       return;
-            if (source == GL_TEXTURE ||
-                source == GL_CONSTANT ||
-                source == GL_PRIMARY_COLOR ||
-                source == GL_PREVIOUS ||
-                (ctx->Extensions.ARB_texture_env_crossbar &&
-                 source >= GL_TEXTURE0 &&
-                 source < GL_TEXTURE0 + ctx->Const.MaxTextureUnits) ||
-		(ctx->Extensions.ATI_texture_env_combine3 &&
-                 (source == GL_ZERO || source == GL_ONE))) {
-               /* legal */
-	       FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-	       texUnit->Combine.SourceA[s] = source;
-            }
-            else {
-               TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", source);
-	       return;
-	    }
-	 }
-	 else {
-            TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
-	    return;
-	 }
+      case GL_SOURCE3_ALPHA_NV:
+         set_combiner_source(ctx, texUnit, pname, (GLenum) (GLint) param[0]);
 	 break;
       case GL_OPERAND0_RGB:
       case GL_OPERAND1_RGB:
-	 if (ctx->Extensions.EXT_texture_env_combine ||
-	     ctx->Extensions.ARB_texture_env_combine) {
-	    const GLenum operand = (GLenum) (GLint) *param;
-	    const GLuint s = pname - GL_OPERAND0_RGB;
-	    if (texUnit->Combine.OperandRGB[s] == operand)
-	       return;
-	    switch (operand) {
-	    case GL_SRC_COLOR:
-	    case GL_ONE_MINUS_SRC_COLOR:
-	    case GL_SRC_ALPHA:
-	    case GL_ONE_MINUS_SRC_ALPHA:
-	       FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-	       texUnit->Combine.OperandRGB[s] = operand;
-	       break;
-	    default:
-               TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand);
-	       return;
-	    }
-	 }
-	 else {
-            TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
-	    return;
-	 }
-	 break;
+      case GL_OPERAND2_RGB:
+      case GL_OPERAND3_RGB_NV:
       case GL_OPERAND0_ALPHA:
       case GL_OPERAND1_ALPHA:
-	 if (ctx->Extensions.EXT_texture_env_combine ||
-             ctx->Extensions.ARB_texture_env_combine) {
-	    const GLenum operand = (GLenum) (GLint) *param;
-	    if (texUnit->Combine.OperandA[pname-GL_OPERAND0_ALPHA] == operand)
-	       return;
-	    switch (operand) {
-	    case GL_SRC_ALPHA:
-	    case GL_ONE_MINUS_SRC_ALPHA:
-	       FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-	       texUnit->Combine.OperandA[pname-GL_OPERAND0_ALPHA] = operand;
-	       break;
-	    default:
-               TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand);
-	       return;
-	    }
-	 }
-	 else {
-            TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
-	    return;
-	 }
-	 break;
-      case GL_OPERAND2_RGB:
-	 if (ctx->Extensions.ARB_texture_env_combine) {
-	    const GLenum operand = (GLenum) (GLint) *param;
-	    if (texUnit->Combine.OperandRGB[2] == operand)
-	       return;
-	    switch (operand) {
-	    case GL_SRC_COLOR:           /* ARB combine only */
-	    case GL_ONE_MINUS_SRC_COLOR: /* ARB combine only */
-	    case GL_SRC_ALPHA:
-	    case GL_ONE_MINUS_SRC_ALPHA: /* ARB combine only */
-	       FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-	       texUnit->Combine.OperandRGB[2] = operand;
-               break;
-	    default:
-               TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand);
-	       return;
-	    }
-	 }
-	 else if (ctx->Extensions.EXT_texture_env_combine) {
-	    const GLenum operand = (GLenum) (GLint) *param;
-	    if (texUnit->Combine.OperandRGB[2] == operand)
-	       return;
-	    /* operand must be GL_SRC_ALPHA which is the initial value - thus
-	       don't need to actually compare the operand to the possible value */
-	    else {
-               TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand);
-	       return;
-	    }
-	 }
-	 else {
-            TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
-	    return;
-	 }
-	 break;
       case GL_OPERAND2_ALPHA:
-	 if (ctx->Extensions.ARB_texture_env_combine) {
-	    const GLenum operand = (GLenum) (GLint) *param;
-	    if (texUnit->Combine.OperandA[2] == operand)
-	       return;
-	    switch (operand) {
-	    case GL_SRC_ALPHA:
-	    case GL_ONE_MINUS_SRC_ALPHA: /* ARB combine only */
-	       FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-	       texUnit->Combine.OperandA[2] = operand;
-	       break;
-	    default:
-               TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand);
-	       return;
-	    }
-	 }
-	 else if (ctx->Extensions.EXT_texture_env_combine) {
-	    const GLenum operand = (GLenum) (GLint) *param;
-	    if (texUnit->Combine.OperandA[2] == operand)
-	       return;
-	    /* operand must be GL_SRC_ALPHA which is the initial value - thus
-	       don't need to actually compare the operand to the possible value */
-	    else {
-               TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand);
-	       return;
-	    }
-	 }
-	 else {
-            TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
-	    return;
-	 }
+      case GL_OPERAND3_ALPHA_NV:
+         set_combiner_operand(ctx, texUnit, pname, (GLenum) (GLint) param[0]);
 	 break;
       case GL_RGB_SCALE:
-	 if (ctx->Extensions.EXT_texture_env_combine ||
-             ctx->Extensions.ARB_texture_env_combine) {
-	    GLuint newshift;
-	    if (*param == 1.0) {
-	       newshift = 0;
-	    }
-	    else if (*param == 2.0) {
-	       newshift = 1;
-	    }
-	    else if (*param == 4.0) {
-	       newshift = 2;
-	    }
-	    else {
-	       _mesa_error( ctx, GL_INVALID_VALUE,
-                            "glTexEnv(GL_RGB_SCALE not 1, 2 or 4)" );
-	       return;
-	    }
-	    if (texUnit->Combine.ScaleShiftRGB == newshift)
-	       return;
-	    FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-	    texUnit->Combine.ScaleShiftRGB = newshift;
-	 }
-	 else {
-            TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
-	    return;
-	 }
-	 break;
       case GL_ALPHA_SCALE:
-	 if (ctx->Extensions.EXT_texture_env_combine ||
-             ctx->Extensions.ARB_texture_env_combine) {
-	    GLuint newshift;
-	    if (*param == 1.0) {
-	       newshift = 0;
-	    }
-	    else if (*param == 2.0) {
-	       newshift = 1;
-	    }
-	    else if (*param == 4.0) {
-	       newshift = 2;
-	    }
-	    else {
-	       _mesa_error( ctx, GL_INVALID_VALUE,
-                            "glTexEnv(GL_ALPHA_SCALE not 1, 2 or 4)" );
-	       return;
-	    }
-	    if (texUnit->Combine.ScaleShiftA == newshift)
-	       return;
-	    FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-	    texUnit->Combine.ScaleShiftA = newshift;
-	 }
-	 else {
-            TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname);
-	    return;
-	 }
+         set_combiner_scale(ctx, texUnit, pname, param[0]);
 	 break;
       default:
 	 _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(pname)" );
@@ -533,6 +607,144 @@ _mesa_TexEnviv( GLenum target, GLenum pname, const GLint *param )
 }
 
 
+
+/**
+ * Helper for glGetTexEnvi/f()
+ * \return  value of queried pname or -1 if error.
+ */
+static GLint
+get_texenvi(GLcontext *ctx, const struct gl_texture_unit *texUnit,
+            GLenum pname)
+{
+   switch (pname) {
+   case GL_TEXTURE_ENV_MODE:
+      return texUnit->EnvMode;
+      break;
+   case GL_COMBINE_RGB:
+      if (ctx->Extensions.EXT_texture_env_combine ||
+          ctx->Extensions.ARB_texture_env_combine) {
+         return texUnit->Combine.ModeRGB;
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_COMBINE_ALPHA:
+      if (ctx->Extensions.EXT_texture_env_combine ||
+          ctx->Extensions.ARB_texture_env_combine) {
+         return texUnit->Combine.ModeA;
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_SOURCE0_RGB:
+   case GL_SOURCE1_RGB:
+   case GL_SOURCE2_RGB:
+      if (ctx->Extensions.EXT_texture_env_combine ||
+          ctx->Extensions.ARB_texture_env_combine) {
+         const unsigned rgb_idx = pname - GL_SOURCE0_RGB;
+         return texUnit->Combine.SourceRGB[rgb_idx];
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_SOURCE3_RGB_NV:
+      if (ctx->Extensions.NV_texture_env_combine4) {
+         return texUnit->Combine.SourceRGB[3];
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_SOURCE0_ALPHA:
+   case GL_SOURCE1_ALPHA:
+   case GL_SOURCE2_ALPHA:
+      if (ctx->Extensions.EXT_texture_env_combine ||
+          ctx->Extensions.ARB_texture_env_combine) {
+         const unsigned alpha_idx = pname - GL_SOURCE0_ALPHA;
+         return texUnit->Combine.SourceA[alpha_idx];
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_SOURCE3_ALPHA_NV:
+      if (ctx->Extensions.NV_texture_env_combine4) {
+         return texUnit->Combine.SourceA[3];
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_OPERAND0_RGB:
+   case GL_OPERAND1_RGB:
+   case GL_OPERAND2_RGB:
+      if (ctx->Extensions.EXT_texture_env_combine ||
+          ctx->Extensions.ARB_texture_env_combine) {
+         const unsigned op_rgb = pname - GL_OPERAND0_RGB;
+         return texUnit->Combine.OperandRGB[op_rgb];
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_OPERAND3_RGB_NV:
+      if (ctx->Extensions.NV_texture_env_combine4) {
+         return texUnit->Combine.OperandRGB[3];
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_OPERAND0_ALPHA:
+   case GL_OPERAND1_ALPHA:
+   case GL_OPERAND2_ALPHA:
+      if (ctx->Extensions.EXT_texture_env_combine ||
+          ctx->Extensions.ARB_texture_env_combine) {
+         const unsigned op_alpha = pname - GL_OPERAND0_ALPHA;
+         return texUnit->Combine.OperandA[op_alpha];
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_OPERAND3_ALPHA_NV:
+      if (ctx->Extensions.NV_texture_env_combine4) {
+         return texUnit->Combine.OperandA[3];
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_RGB_SCALE:
+      if (ctx->Extensions.EXT_texture_env_combine ||
+          ctx->Extensions.ARB_texture_env_combine) {
+         return 1 << texUnit->Combine.ScaleShiftRGB;
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   case GL_ALPHA_SCALE:
+      if (ctx->Extensions.EXT_texture_env_combine ||
+          ctx->Extensions.ARB_texture_env_combine) {
+         return 1 << texUnit->Combine.ScaleShiftA;
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
+      }
+      break;
+   default:
+      ;
+   }
+
+   return -1; /* error */
+}
+
+
+
 void GLAPIENTRY
 _mesa_GetTexEnvfv( GLenum target, GLenum pname, GLfloat *params )
 {
@@ -551,111 +763,14 @@ _mesa_GetTexEnvfv( GLenum target, GLenum pname, GLfloat *params )
    texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
 
    if (target == GL_TEXTURE_ENV) {
-      switch (pname) {
-         case GL_TEXTURE_ENV_MODE:
-            *params = ENUM_TO_FLOAT(texUnit->EnvMode);
-            break;
-         case GL_TEXTURE_ENV_COLOR:
-            COPY_4FV( params, texUnit->EnvColor );
-            break;
-         case GL_COMBINE_RGB:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-               *params = (GLfloat) texUnit->Combine.ModeRGB;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
-            }
-            break;
-         case GL_COMBINE_ALPHA:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-               *params = (GLfloat) texUnit->Combine.ModeA;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
-            }
-            break;
-         case GL_SOURCE0_RGB:
-         case GL_SOURCE1_RGB:
-         case GL_SOURCE2_RGB:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-	       const unsigned rgb_idx = pname - GL_SOURCE0_RGB;
-               *params = (GLfloat) texUnit->Combine.SourceRGB[rgb_idx];
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
-            }
-            break;
-         case GL_SOURCE0_ALPHA:
-         case GL_SOURCE1_ALPHA:
-         case GL_SOURCE2_ALPHA:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-	       const unsigned alpha_idx = pname - GL_SOURCE0_ALPHA;
-               *params = (GLfloat) texUnit->Combine.SourceA[alpha_idx];
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
-            }
-            break;
-         case GL_OPERAND0_RGB:
-         case GL_OPERAND1_RGB:
-         case GL_OPERAND2_RGB:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-	       const unsigned op_rgb = pname - GL_OPERAND0_RGB;
-               *params = (GLfloat) texUnit->Combine.OperandRGB[op_rgb];
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
-            }
-            break;
-         case GL_OPERAND0_ALPHA:
-         case GL_OPERAND1_ALPHA:
-         case GL_OPERAND2_ALPHA:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-	       const unsigned op_alpha = pname - GL_OPERAND0_ALPHA;
-               *params = (GLfloat) texUnit->Combine.OperandA[op_alpha];
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
-            }
-            break;
-         case GL_RGB_SCALE:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-               if (texUnit->Combine.ScaleShiftRGB == 0)
-                  *params = 1.0;
-               else if (texUnit->Combine.ScaleShiftRGB == 1)
-                  *params = 2.0;
-               else
-                  *params = 4.0;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
-               return;
-            }
-            break;
-         case GL_ALPHA_SCALE:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-               if (texUnit->Combine.ScaleShiftA == 0)
-                  *params = 1.0;
-               else if (texUnit->Combine.ScaleShiftA == 1)
-                  *params = 2.0;
-               else
-                  *params = 4.0;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)");
-               return;
-            }
-            break;
-         default:
-            _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname=0x%x)", pname);
+      if (pname == GL_TEXTURE_ENV_COLOR) {
+         COPY_4FV( params, texUnit->EnvColor );
+      }
+      else {
+         GLint val = get_texenvi(ctx, texUnit, pname);
+         if (val >= 0) {
+            *params = (GLfloat) val;
+         }
       }
    }
    else if (target == GL_TEXTURE_FILTER_CONTROL_EXT) {
@@ -712,115 +827,17 @@ _mesa_GetTexEnviv( GLenum target, GLenum pname, GLint *params )
    texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
 
    if (target == GL_TEXTURE_ENV) {
-      switch (pname) {
-         case GL_TEXTURE_ENV_MODE:
-            *params = (GLint) texUnit->EnvMode;
-            break;
-         case GL_TEXTURE_ENV_COLOR:
-            params[0] = FLOAT_TO_INT( texUnit->EnvColor[0] );
-            params[1] = FLOAT_TO_INT( texUnit->EnvColor[1] );
-            params[2] = FLOAT_TO_INT( texUnit->EnvColor[2] );
-            params[3] = FLOAT_TO_INT( texUnit->EnvColor[3] );
-            break;
-         case GL_COMBINE_RGB:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-               *params = (GLint) texUnit->Combine.ModeRGB;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)");
-            }
-            break;
-         case GL_COMBINE_ALPHA:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-               *params = (GLint) texUnit->Combine.ModeA;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)");
-            }
-            break;
-         case GL_SOURCE0_RGB:
-         case GL_SOURCE1_RGB:
-         case GL_SOURCE2_RGB:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-	       const unsigned rgb_idx = pname - GL_SOURCE0_RGB;
-               *params = (GLint) texUnit->Combine.SourceRGB[rgb_idx];
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)");
-            }
-            break;
-         case GL_SOURCE0_ALPHA:
-         case GL_SOURCE1_ALPHA:
-         case GL_SOURCE2_ALPHA:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-	       const unsigned alpha_idx = pname - GL_SOURCE0_ALPHA;
-               *params = (GLint) texUnit->Combine.SourceA[alpha_idx];
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)");
-            }
-            break;
-         case GL_OPERAND0_RGB:
-         case GL_OPERAND1_RGB:
-         case GL_OPERAND2_RGB:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-	       const unsigned op_rgb = pname - GL_OPERAND0_RGB;
-               *params = (GLint) texUnit->Combine.OperandRGB[op_rgb];
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)");
-            }
-            break;
-         case GL_OPERAND0_ALPHA:
-         case GL_OPERAND1_ALPHA:
-         case GL_OPERAND2_ALPHA:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-	       const unsigned op_alpha = pname - GL_OPERAND0_ALPHA;
-               *params = (GLint) texUnit->Combine.OperandA[op_alpha];
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)");
-            }
-            break;
-         case GL_RGB_SCALE:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-               if (texUnit->Combine.ScaleShiftRGB == 0)
-                  *params = 1;
-               else if (texUnit->Combine.ScaleShiftRGB == 1)
-                  *params = 2;
-               else
-                  *params = 4;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)");
-               return;
-            }
-            break;
-         case GL_ALPHA_SCALE:
-            if (ctx->Extensions.EXT_texture_env_combine ||
-                ctx->Extensions.ARB_texture_env_combine) {
-               if (texUnit->Combine.ScaleShiftA == 0)
-                  *params = 1;
-               else if (texUnit->Combine.ScaleShiftA == 1)
-                  *params = 2;
-               else
-                  *params = 4;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)");
-               return;
-            }
-            break;
-         default:
-            _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname=0x%x)",
-                        pname);
+      if (pname == GL_TEXTURE_ENV_COLOR) {
+         params[0] = FLOAT_TO_INT( texUnit->EnvColor[0] );
+         params[1] = FLOAT_TO_INT( texUnit->EnvColor[1] );
+         params[2] = FLOAT_TO_INT( texUnit->EnvColor[2] );
+         params[3] = FLOAT_TO_INT( texUnit->EnvColor[3] );
+      }
+      else {
+         GLint val = get_texenvi(ctx, texUnit, pname);
+         if (val >= 0) {
+            *params = val;
+         }
       }
    }
    else if (target == GL_TEXTURE_FILTER_CONTROL_EXT) {
diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c
index dcd7f90567..af51a206a5 100644
--- a/src/mesa/main/texenvprogram.c
+++ b/src/mesa/main/texenvprogram.c
@@ -2,6 +2,7 @@
  * 
  * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
+ * Copyright 2009 VMware, Inc.  All Rights Reserved.
  * 
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the
@@ -38,6 +39,24 @@
 #include "texenvprogram.h"
 
 
+#define MAX_TERMS 4
+
+
+/*
+ * Note on texture units:
+ *
+ * The number of texture units supported by fixed-function fragment
+ * processing is MAX_TEXTURE_COORD_UNITS, not MAX_TEXTURE_IMAGE_UNITS.
+ * That's because there's a one-to-one correspondence between texture
+ * coordinates and samplers in fixed-function processing.
+ *
+ * Since fixed-function vertex processing is limited to MAX_TEXTURE_COORD_UNITS
+ * sets of texcoords, so is fixed-function fragment processing.
+ *
+ * We can safely use ctx->Const.MaxTextureUnits for loop bounds.
+ */
+
+
 struct texenvprog_cache_item
 {
    GLuint hash;
@@ -48,11 +67,9 @@ struct texenvprog_cache_item
 
 
 /**
- * This MAX is probably a bit generous, but that's OK.  There can be
- * up to four instructions per texture unit (TEX + 3 for combine),
- * then there's fog and specular add.
+ * Up to nine instructions per tex unit, plus fog, specular color.
  */
-#define MAX_INSTRUCTIONS ((MAX_TEXTURE_UNITS * 4) + 12)
+#define MAX_INSTRUCTIONS ((MAX_TEXTURE_COORD_UNITS * 9) + 12)
 
 #define DISASSEM (MESA_VERBOSE & VERBOSE_DISASSEM)
 
@@ -62,10 +79,12 @@ struct mode_opt {
 };
 
 struct state_key {
-   GLbitfield enabled_units;
+   GLuint nr_enabled_units:8;
+   GLuint enabled_units:8;
    GLuint separate_specular:1;
    GLuint fog_enabled:1;
    GLuint fog_mode:2;
+   GLuint inputs_available:12;
 
    struct {
       GLuint enabled:1;
@@ -74,13 +93,13 @@ struct state_key {
       GLuint ScaleShiftRGB:2;
       GLuint ScaleShiftA:2;
 
-      GLuint NumArgsRGB:2;
+      GLuint NumArgsRGB:3;
       GLuint ModeRGB:4;
-      struct mode_opt OptRGB[3];
+      struct mode_opt OptRGB[MAX_TERMS];
 
-      GLuint NumArgsA:2;
+      GLuint NumArgsA:3;
       GLuint ModeA:4;
-      struct mode_opt OptA[3];
+      struct mode_opt OptA[MAX_TERMS];
    } unit[8];
 };
 
@@ -116,7 +135,9 @@ static GLuint translate_operand( GLenum operand )
    case GL_ONE_MINUS_SRC_ALPHA: return OPR_ONE_MINUS_SRC_ALPHA;
    case GL_ZERO: return OPR_ZERO;
    case GL_ONE: return OPR_ONE;
-   default:	return OPR_UNKNOWN;
+   default:
+      assert(0);
+      return OPR_UNKNOWN;
    }
 }
 
@@ -132,6 +153,7 @@ static GLuint translate_operand( GLenum operand )
 #define SRC_CONSTANT 9
 #define SRC_PRIMARY_COLOR 10
 #define SRC_PREVIOUS 11
+#define SRC_ZERO     12
 #define SRC_UNKNOWN  15
 
 static GLuint translate_source( GLenum src )
@@ -149,32 +171,49 @@ static GLuint translate_source( GLenum src )
    case GL_CONSTANT: return SRC_CONSTANT;
    case GL_PRIMARY_COLOR: return SRC_PRIMARY_COLOR;
    case GL_PREVIOUS: return SRC_PREVIOUS;
-   default: return SRC_UNKNOWN;
+   case GL_ZERO:
+      return SRC_ZERO;
+   default:
+      assert(0);
+      return SRC_UNKNOWN;
    }
 }
 
-#define MODE_REPLACE       0
-#define MODE_MODULATE      1
-#define MODE_ADD           2
-#define MODE_ADD_SIGNED    3
-#define MODE_INTERPOLATE   4
-#define MODE_SUBTRACT      5
-#define MODE_DOT3_RGB      6
-#define MODE_DOT3_RGB_EXT  7
-#define MODE_DOT3_RGBA     8
-#define MODE_DOT3_RGBA_EXT 9
-#define MODE_MODULATE_ADD_ATI           10
-#define MODE_MODULATE_SIGNED_ADD_ATI    11
-#define MODE_MODULATE_SUBTRACT_ATI      12
-#define MODE_UNKNOWN       15
-
-static GLuint translate_mode( GLenum mode )
+#define MODE_REPLACE                     0  /* r = a0 */
+#define MODE_MODULATE                    1  /* r = a0 * a1 */
+#define MODE_ADD                         2  /* r = a0 + a1 */
+#define MODE_ADD_SIGNED                  3  /* r = a0 + a1 - 0.5 */
+#define MODE_INTERPOLATE                 4  /* r = a0 * a2 + a1 * (1 - a2) */
+#define MODE_SUBTRACT                    5  /* r = a0 - a1 */
+#define MODE_DOT3_RGB                    6  /* r = a0 . a1 */
+#define MODE_DOT3_RGB_EXT                7  /* r = a0 . a1 */
+#define MODE_DOT3_RGBA                   8  /* r = a0 . a1 */
+#define MODE_DOT3_RGBA_EXT               9  /* r = a0 . a1 */
+#define MODE_MODULATE_ADD_ATI           10  /* r = a0 * a2 + a1 */
+#define MODE_MODULATE_SIGNED_ADD_ATI    11  /* r = a0 * a2 + a1 - 0.5 */
+#define MODE_MODULATE_SUBTRACT_ATI      12  /* r = a0 * a2 - a1 */
+#define MODE_ADD_PRODUCTS               13  /* r = a0 * a1 + a2 * a3 */
+#define MODE_ADD_PRODUCTS_SIGNED        14  /* r = a0 * a1 + a2 * a3 - 0.5 */
+#define MODE_UNKNOWN                    15
+
+/**
+ * Translate GL combiner state into a MODE_x value
+ */
+static GLuint translate_mode( GLenum envMode, GLenum mode )
 {
    switch (mode) {
    case GL_REPLACE: return MODE_REPLACE;
    case GL_MODULATE: return MODE_MODULATE;
-   case GL_ADD: return MODE_ADD;
-   case GL_ADD_SIGNED: return MODE_ADD_SIGNED;
+   case GL_ADD:
+      if (envMode == GL_COMBINE4_NV)
+         return MODE_ADD_PRODUCTS;
+      else
+         return MODE_ADD;
+   case GL_ADD_SIGNED:
+      if (envMode == GL_COMBINE4_NV)
+         return MODE_ADD_PRODUCTS_SIGNED;
+      else
+         return MODE_ADD_SIGNED;
    case GL_INTERPOLATE: return MODE_INTERPOLATE;
    case GL_SUBTRACT: return MODE_SUBTRACT;
    case GL_DOT3_RGB: return MODE_DOT3_RGB;
@@ -184,23 +223,127 @@ static GLuint translate_mode( GLenum mode )
    case GL_MODULATE_ADD_ATI: return MODE_MODULATE_ADD_ATI;
    case GL_MODULATE_SIGNED_ADD_ATI: return MODE_MODULATE_SIGNED_ADD_ATI;
    case GL_MODULATE_SUBTRACT_ATI: return MODE_MODULATE_SUBTRACT_ATI;
-   default: return MODE_UNKNOWN;
+   default:
+      assert(0);
+      return MODE_UNKNOWN;
    }
 }
 
 #define TEXTURE_UNKNOWN_INDEX 7
 static GLuint translate_tex_src_bit( GLbitfield bit )
 {
+   /* make sure number of switch cases is correct */
+   assert(NUM_TEXTURE_TARGETS == 7);
    switch (bit) {
    case TEXTURE_1D_BIT:   return TEXTURE_1D_INDEX;
    case TEXTURE_2D_BIT:   return TEXTURE_2D_INDEX;
    case TEXTURE_RECT_BIT: return TEXTURE_RECT_INDEX;
    case TEXTURE_3D_BIT:   return TEXTURE_3D_INDEX;
    case TEXTURE_CUBE_BIT: return TEXTURE_CUBE_INDEX;
-   default: return TEXTURE_UNKNOWN_INDEX;
+   case TEXTURE_1D_ARRAY_BIT: return TEXTURE_1D_ARRAY_INDEX;
+   case TEXTURE_2D_ARRAY_BIT: return TEXTURE_2D_ARRAY_INDEX;
+   default:
+      assert(0);
+      return TEXTURE_UNKNOWN_INDEX;
+   }
+}
+
+#define VERT_BIT_TEX_ANY    (0xff << VERT_ATTRIB_TEX0)
+#define VERT_RESULT_TEX_ANY (0xff << VERT_RESULT_TEX0)
+
+/**
+ * Identify all possible varying inputs.  The fragment program will
+ * never reference non-varying inputs, but will track them via state
+ * constants instead.
+ *
+ * This function figures out all the inputs that the fragment program
+ * has access to.  The bitmask is later reduced to just those which
+ * are actually referenced.
+ */
+static GLbitfield get_fp_input_mask( GLcontext *ctx )
+{
+   const GLboolean vertexShader = (ctx->Shader.CurrentProgram &&
+                                   ctx->Shader.CurrentProgram->VertexProgram);
+   const GLboolean vertexProgram = ctx->VertexProgram._Enabled;
+   GLbitfield fp_inputs = 0x0;
+
+   if (ctx->VertexProgram._Overriden) {
+      /* Somebody's messing with the vertex program and we don't have
+       * a clue what's happening.  Assume that it could be producing
+       * all possible outputs.
+       */
+      fp_inputs = ~0;
+   }
+   else if (ctx->RenderMode == GL_FEEDBACK) {
+      fp_inputs = (FRAG_BIT_COL0 | FRAG_BIT_TEX0);
+   }
+   else if (!(vertexProgram || vertexShader) ||
+            !ctx->VertexProgram._Current) {
+      /* Fixed function vertex logic */
+      GLbitfield varying_inputs = ctx->varying_vp_inputs;
+
+      /* These get generated in the setup routine regardless of the
+       * vertex program:
+       */
+      if (ctx->Point.PointSprite)
+         varying_inputs |= FRAG_BITS_TEX_ANY;
+
+      /* First look at what values may be computed by the generated
+       * vertex program:
+       */
+      if (ctx->Light.Enabled) {
+         fp_inputs |= FRAG_BIT_COL0;
+
+         if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR)
+            fp_inputs |= FRAG_BIT_COL1;
+      }
+
+      fp_inputs |= (ctx->Texture._TexGenEnabled |
+                    ctx->Texture._TexMatEnabled) << FRAG_ATTRIB_TEX0;
+
+      /* Then look at what might be varying as a result of enabled
+       * arrays, etc:
+       */
+      if (varying_inputs & VERT_BIT_COLOR0) fp_inputs |= FRAG_BIT_COL0;
+      if (varying_inputs & VERT_BIT_COLOR1) fp_inputs |= FRAG_BIT_COL1;
+
+      fp_inputs |= (((varying_inputs & VERT_BIT_TEX_ANY) >> VERT_ATTRIB_TEX0) 
+                    << FRAG_ATTRIB_TEX0);
+
+   }
+   else {
+      /* calculate from vp->outputs */
+      struct gl_vertex_program *vprog;
+      GLbitfield vp_outputs;
+
+      /* Choose GLSL vertex shader over ARB vertex program.  Need this
+       * since vertex shader state validation comes after fragment state
+       * validation (see additional comments in state.c).
+       */
+      if (vertexShader)
+         vprog = ctx->Shader.CurrentProgram->VertexProgram;
+      else
+         vprog = ctx->VertexProgram._Current;
+
+      vp_outputs = vprog->Base.OutputsWritten;
+
+      /* These get generated in the setup routine regardless of the
+       * vertex program:
+       */
+      if (ctx->Point.PointSprite)
+         vp_outputs |= FRAG_BITS_TEX_ANY;
+
+      if (vp_outputs & (1 << VERT_RESULT_COL0)) fp_inputs |= FRAG_BIT_COL0;
+      if (vp_outputs & (1 << VERT_RESULT_COL1)) fp_inputs |= FRAG_BIT_COL1;
+
+      fp_inputs |= (((vp_outputs & VERT_RESULT_TEX_ANY) >> VERT_RESULT_TEX0) 
+                    << FRAG_ATTRIB_TEX0);
    }
+   
+   return fp_inputs;
 }
 
+
 /**
  * Examine current texture environment state and generate a unique
  * key to identify it.
@@ -208,34 +351,43 @@ static GLuint translate_tex_src_bit( GLbitfield bit )
 static void make_state_key( GLcontext *ctx,  struct state_key *key )
 {
    GLuint i, j;
-	
+   GLbitfield inputs_referenced = FRAG_BIT_COL0;
+   GLbitfield inputs_available = get_fp_input_mask( ctx );
+
    memset(key, 0, sizeof(*key));
 
-   for (i=0;i<MAX_TEXTURE_UNITS;i++) {
+   for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
       const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
+      GLenum format;
 
       if (!texUnit->_ReallyEnabled || !texUnit->Enabled)
          continue;
 
+      format = texUnit->_Current->Image[0][texUnit->_Current->BaseLevel]->_BaseFormat;
+
       key->unit[i].enabled = 1;
       key->enabled_units |= (1<<i);
+      key->nr_enabled_units = i+1;
+      inputs_referenced |= FRAG_BIT_TEX(i);
 
       key->unit[i].source_index = 
 	 translate_tex_src_bit(texUnit->_ReallyEnabled);		
-      key->unit[i].shadow = texUnit->_Current->CompareMode == GL_COMPARE_R_TO_TEXTURE;
+      key->unit[i].shadow = ((texUnit->_Current->CompareMode == GL_COMPARE_R_TO_TEXTURE) && 
+                             ((format == GL_DEPTH_COMPONENT) || 
+                              (format == GL_DEPTH_STENCIL_EXT)));
 
       key->unit[i].NumArgsRGB = texUnit->_CurrentCombine->_NumArgsRGB;
       key->unit[i].NumArgsA = texUnit->_CurrentCombine->_NumArgsA;
 
       key->unit[i].ModeRGB =
-	 translate_mode(texUnit->_CurrentCombine->ModeRGB);
+	 translate_mode(texUnit->EnvMode, texUnit->_CurrentCombine->ModeRGB);
       key->unit[i].ModeA =
-	 translate_mode(texUnit->_CurrentCombine->ModeA);
+	 translate_mode(texUnit->EnvMode, texUnit->_CurrentCombine->ModeA);
 		
       key->unit[i].ScaleShiftRGB = texUnit->_CurrentCombine->ScaleShiftRGB;
       key->unit[i].ScaleShiftA = texUnit->_CurrentCombine->ScaleShiftA;
 
-      for (j=0;j<3;j++) {
+      for (j = 0; j < MAX_TERMS; j++) {
          key->unit[i].OptRGB[j].Operand =
 	    translate_operand(texUnit->_CurrentCombine->OperandRGB[j]);
          key->unit[i].OptA[j].Operand =
@@ -247,16 +399,22 @@ static void make_state_key( GLcontext *ctx,  struct state_key *key )
       }
    }
 	
-   if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR)
+   if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
       key->separate_specular = 1;
+      inputs_referenced |= FRAG_BIT_COL1;
+   }
 
    if (ctx->Fog.Enabled) {
       key->fog_enabled = 1;
       key->fog_mode = translate_fog_mode(ctx->Fog.Mode);
+      inputs_referenced |= FRAG_BIT_FOGC; /* maybe */
    }
+
+   key->inputs_available = (inputs_available & inputs_referenced);
 }
 
-/* Use uregs to represent registers internally, translate to Mesa's
+/**
+ * Use uregs to represent registers internally, translate to Mesa's
  * expected formats on emit.  
  *
  * NOTE: These are passed by value extensively in this file rather
@@ -289,28 +447,28 @@ static const struct ureg undef = {
 };
 
 
-/* State used to build the fragment program:
+/** State used to build the fragment program:
  */
 struct texenv_fragment_program {
    struct gl_fragment_program *program;
    GLcontext *ctx;
    struct state_key *state;
 
-   GLbitfield alu_temps;	/* Track texture indirections, see spec. */
-   GLbitfield temps_output;	/* Track texture indirections, see spec. */
-   GLbitfield temp_in_use;	/* Tracks temporary regs which are in use. */
+   GLbitfield alu_temps;	/**< Track texture indirections, see spec. */
+   GLbitfield temps_output;	/**< Track texture indirections, see spec. */
+   GLbitfield temp_in_use;	/**< Tracks temporary regs which are in use. */
    GLboolean error;
 
-   struct ureg src_texture[MAX_TEXTURE_UNITS];   
+   struct ureg src_texture[MAX_TEXTURE_COORD_UNITS];   
    /* Reg containing each texture unit's sampled texture color,
     * else undef.
     */
 
-   struct ureg src_previous;	/* Reg containing color from previous 
+   struct ureg src_previous;	/**< Reg containing color from previous 
 				 * stage.  May need to be decl'd.
 				 */
 
-   GLuint last_tex_stage;	/* Number of last enabled texture unit */
+   GLuint last_tex_stage;	/**< Number of last enabled texture unit */
 
    struct ureg half;
    struct ureg one;
@@ -388,7 +546,7 @@ static struct ureg get_tex_temp( struct texenv_fragment_program *p )
 {
    int bit;
    
-   /* First try to find availble temp not previously used (to avoid
+   /* First try to find available temp not previously used (to avoid
     * starting a new texture indirection).  According to the spec, the
     * ~p->temps_output isn't necessary, but will keep it there for
     * now:
@@ -413,6 +571,14 @@ static struct ureg get_tex_temp( struct texenv_fragment_program *p )
 }
 
 
+/** Mark a temp reg as being no longer allocatable. */
+static void reserve_temp( struct texenv_fragment_program *p, struct ureg r )
+{
+   if (r.file == PROGRAM_TEMPORARY)
+      p->temps_output |= (1 << r.idx);
+}
+
+
 static void release_temps(GLcontext *ctx, struct texenv_fragment_program *p )
 {
    GLuint max_temp = ctx->Const.FragmentProgram.MaxTemps;
@@ -451,11 +617,29 @@ static struct ureg register_param5( struct texenv_fragment_program *p,
 #define register_param3(p,s0,s1,s2)    register_param5(p,s0,s1,s2,0,0)
 #define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
 
+static GLuint frag_to_vert_attrib( GLuint attrib )
+{
+   switch (attrib) {
+   case FRAG_ATTRIB_COL0: return VERT_ATTRIB_COLOR0;
+   case FRAG_ATTRIB_COL1: return VERT_ATTRIB_COLOR1;
+   default:
+      assert(attrib >= FRAG_ATTRIB_TEX0);
+      assert(attrib <= FRAG_ATTRIB_TEX7);
+      return attrib - FRAG_ATTRIB_TEX0 + VERT_ATTRIB_TEX0;
+   }
+}
+
 
 static struct ureg register_input( struct texenv_fragment_program *p, GLuint input )
 {
-   p->program->Base.InputsRead |= (1 << input);
-   return make_ureg(PROGRAM_INPUT, input);
+   if (p->state->inputs_available & (1<<input)) {
+      p->program->Base.InputsRead |= (1 << input);
+      return make_ureg(PROGRAM_INPUT, input);
+   }
+   else {
+      GLuint idx = frag_to_vert_attrib( input );
+      return register_param3( p, STATE_INTERNAL, STATE_CURRENT_ATTRIB, idx );
+   }
 }
 
 
@@ -506,10 +690,12 @@ emit_op(struct texenv_fragment_program *p,
 
    emit_dst( &inst->DstReg, dest, mask );
 
+#if 0
    /* Accounting for indirection tracking:
     */
    if (dest.file == PROGRAM_TEMPORARY)
       p->temps_output |= 1 << dest.idx;
+#endif
 
    return inst;
 }
@@ -564,6 +750,10 @@ static struct ureg emit_texld( struct texenv_fragment_program *p,
 
    p->program->Base.NumTexInstructions++;
 
+   /* Accounting for indirection tracking:
+    */
+   reserve_temp(p, dest);
+
    /* Is this a texture indirection?
     */
    if ((coord.file == PROGRAM_TEMPORARY &&
@@ -588,14 +778,16 @@ static struct ureg register_const4f( struct texenv_fragment_program *p,
 {
    GLfloat values[4];
    GLuint idx, swizzle;
+   struct ureg r;
    values[0] = s0;
    values[1] = s1;
    values[2] = s2;
    values[3] = s3;
    idx = _mesa_add_unnamed_constant( p->program->Base.Parameters, values, 4,
                                      &swizzle );
-   ASSERT(swizzle == SWIZZLE_NOOP);
-   return make_ureg(PROGRAM_CONSTANT, idx);
+   r = make_ureg(PROGRAM_CONSTANT, idx);
+   r.swz = swizzle;
+   return r;
 }
 
 #define register_scalar_const(p, s0)    register_const4f(p, s0, s0, s0, s0)
@@ -657,12 +849,17 @@ static struct ureg get_source( struct texenv_fragment_program *p,
    case SRC_PRIMARY_COLOR:
       return register_input(p, FRAG_ATTRIB_COL0);
 
+   case SRC_ZERO:
+      return get_zero(p);
+
    case SRC_PREVIOUS:
-   default:
       if (is_undef(p->src_previous))
 	 return register_input(p, FRAG_ATTRIB_COL0);
       else
 	 return p->src_previous;
+
+   default:
+      assert(0);
    }
 }
 
@@ -703,7 +900,9 @@ static struct ureg emit_combine_source( struct texenv_fragment_program *p,
    case OPR_ONE:
       return get_one(p);
    case OPR_SRC_COLOR: 
+      return src;
    default:
+      assert(0);
       return src;
    }
 }
@@ -752,10 +951,12 @@ static struct ureg emit_combine( struct texenv_fragment_program *p,
 				 GLuint mode,
 				 const struct mode_opt *opt)
 {
-   struct ureg src[3];
+   struct ureg src[MAX_TERMS];
    struct ureg tmp, half;
    GLuint i;
 
+   assert(nr <= MAX_TERMS);
+
    tmp = undef; /* silence warning (bug 5318) */
 
    for (i = 0; i < nr; i++)
@@ -831,7 +1032,26 @@ static struct ureg emit_combine( struct texenv_fragment_program *p,
       /* Arg0 * Arg2 - Arg1 */
       emit_arith( p, OPCODE_MAD, dest, mask, 0, src[0], src[2], negate(src[1]) );
       return dest;
+   case MODE_ADD_PRODUCTS:
+      /* Arg0 * Arg1 + Arg2 * Arg3 */
+      {
+         struct ureg tmp0 = get_temp(p);
+         emit_arith( p, OPCODE_MUL, tmp0, mask, 0, src[0], src[1], undef );
+         emit_arith( p, OPCODE_MAD, dest, mask, saturate, src[2], src[3], tmp0 );
+      }
+      return dest;
+   case MODE_ADD_PRODUCTS_SIGNED:
+      /* Arg0 * Arg1 + Arg2 * Arg3 - 0.5 */
+      {
+         struct ureg tmp0 = get_temp(p);
+         half = get_half(p);
+         emit_arith( p, OPCODE_MUL, tmp0, mask, 0, src[0], src[1], undef );
+         emit_arith( p, OPCODE_MAD, tmp0, mask, 0, src[2], src[3], tmp0 );
+         emit_arith( p, OPCODE_SUB, dest, mask, saturate, tmp0, half, undef );
+      }
+      return dest;
    default: 
+      assert(0);
       return src[0];
    }
 }
@@ -953,6 +1173,7 @@ static void load_texture( struct texenv_fragment_program *p, GLuint unit )
 	 p->src_texture[unit] = emit_texld( p, OPCODE_TXP,
 					    tmp, WRITEMASK_XYZW, 
 					    unit, dim, texcoord );
+
 	 if (p->state->unit[unit].shadow)
 	    p->program->Base.ShadowSamplers |= 1 << unit;
 
@@ -987,6 +1208,7 @@ static GLboolean load_texenv_source( struct texenv_fragment_program *p,
       break;
       
    default:
+      /* not a texture src - do nothing */
       break;
    }
  
@@ -1051,7 +1273,7 @@ create_new_program(GLcontext *ctx, struct state_key *key,
    p.program->Base.InputsRead = 0;
    p.program->Base.OutputsWritten = 1 << FRAG_RESULT_COLR;
 
-   for (unit = 0; unit < MAX_TEXTURE_UNITS; unit++)
+   for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++)
       p.src_texture[unit] = undef;
 
    p.src_previous = undef;
@@ -1078,6 +1300,7 @@ create_new_program(GLcontext *ctx, struct state_key *key,
       for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++)
 	 if (key->enabled_units & (1<<unit)) {
 	    p.src_previous = emit_texenv( &p, unit );
+            reserve_temp(&p, p.src_previous); /* don't re-use this temp reg */
 	    release_temps(ctx, &p);	/* release all temps */
 	 }
    }
@@ -1182,39 +1405,3 @@ _mesa_get_fixed_func_fragment_program(GLcontext *ctx)
 
    return prog;
 }
-
-
-
-/**
- * If _MaintainTexEnvProgram is set we'll generate a fragment program that
- * implements the current texture env/combine mode.
- * This function generates that program and puts it into effect.
- */
-void
-_mesa_UpdateTexEnvProgram( GLcontext *ctx )
-{
-   const struct gl_fragment_program *prev = ctx->FragmentProgram._Current;
-	
-   ASSERT(ctx->FragmentProgram._MaintainTexEnvProgram);
-
-   /* If a conventional fragment program/shader isn't in effect... */
-   if (!ctx->FragmentProgram._Enabled &&
-       (!ctx->Shader.CurrentProgram ||
-        !ctx->Shader.CurrentProgram->FragmentProgram) ) 
-   {
-      struct gl_fragment_program *newProg;
-
-      newProg = _mesa_get_fixed_func_fragment_program(ctx);
-
-      _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, newProg);
-      _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram, newProg);
-   } 
-
-   /* Tell the driver about the change.  Could define a new target for
-    * this?
-    */
-   if (ctx->FragmentProgram._Current != prev && ctx->Driver.BindProgram) {
-      ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
-                         (struct gl_program *) ctx->FragmentProgram._Current);
-   }
-}
diff --git a/src/mesa/main/texenvprogram.h b/src/mesa/main/texenvprogram.h
index a7aa60cf37..0a162d2e7a 100644
--- a/src/mesa/main/texenvprogram.h
+++ b/src/mesa/main/texenvprogram.h
@@ -32,6 +32,4 @@
 extern struct gl_fragment_program *
 _mesa_get_fixed_func_fragment_program(GLcontext *ctx);
 
-extern void _mesa_UpdateTexEnvProgram( GLcontext *ctx );
-
 #endif
diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c
index 4442ce39a4..16d05cc7d0 100644
--- a/src/mesa/main/texformat.c
+++ b/src/mesa/main/texformat.c
@@ -3,6 +3,7 @@
  * Version:  6.5.1
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -333,6 +334,30 @@ const struct gl_texture_format _mesa_texformat_srgba8 = {
    store_texel_srgba8			/* StoreTexel */
 };
 
+const struct gl_texture_format _mesa_texformat_sargb8 = {
+   MESA_FORMAT_SARGB8,			/* MesaFormat */
+   GL_RGBA,				/* BaseFormat */
+   GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
+   8,					/* RedBits */
+   8,					/* GreenBits */
+   8,					/* BlueBits */
+   8,					/* AlphaBits */
+   0,					/* LuminanceBits */
+   0,					/* IntensityBits */
+   0,					/* IndexBits */
+   0,					/* DepthBits */
+   0,					/* StencilBits */
+   4,					/* TexelBytes */
+   _mesa_texstore_sargb8,		/* StoreTexImageFunc */
+   NULL,				/* FetchTexel1D */
+   NULL,				/* FetchTexel2D */
+   NULL,				/* FetchTexel3D */
+   fetch_texel_1d_sargb8,		/* FetchTexel1Df */
+   fetch_texel_2d_sargb8,		/* FetchTexel2Df */
+   fetch_texel_3d_sargb8,		/* FetchTexel3Df */
+   store_texel_sargb8			/* StoreTexel */
+};
+
 const struct gl_texture_format _mesa_texformat_sl8 = {
    MESA_FORMAT_SL8,			/* MesaFormat */
    GL_LUMINANCE,			/* BaseFormat */
@@ -871,6 +896,30 @@ const struct gl_texture_format _mesa_texformat_rgb565_rev = {
    store_texel_rgb565_rev		/* StoreTexel */
 };
 
+const struct gl_texture_format _mesa_texformat_rgba4444 = {
+   MESA_FORMAT_RGBA4444,		/* MesaFormat */
+   GL_RGBA,				/* BaseFormat */
+   GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
+   4,					/* RedBits */
+   4,					/* GreenBits */
+   4,					/* BlueBits */
+   4,					/* AlphaBits */
+   0,					/* LuminanceBits */
+   0,					/* IntensityBits */
+   0,					/* IndexBits */
+   0,					/* DepthBits */
+   0,					/* StencilBits */
+   2,					/* TexelBytes */
+   _mesa_texstore_rgba4444,		/* StoreTexImageFunc */
+   fetch_texel_1d_rgba4444,		/* FetchTexel1D */
+   fetch_texel_2d_rgba4444,		/* FetchTexel2D */
+   fetch_texel_3d_rgba4444,		/* FetchTexel3D */
+   NULL,				/* FetchTexel1Df */
+   NULL,				/* FetchTexel2Df */
+   NULL,				/* FetchTexel3Df */
+   store_texel_rgba4444			/* StoreTexel */
+};
+
 const struct gl_texture_format _mesa_texformat_argb4444 = {
    MESA_FORMAT_ARGB4444,		/* MesaFormat */
    GL_RGBA,				/* BaseFormat */
@@ -919,6 +968,30 @@ const struct gl_texture_format _mesa_texformat_argb4444_rev = {
    store_texel_argb4444_rev		/* StoreTexel */
 };
 
+const struct gl_texture_format _mesa_texformat_rgba5551 = {
+   MESA_FORMAT_RGBA5551,		/* MesaFormat */
+   GL_RGBA,				/* BaseFormat */
+   GL_UNSIGNED_NORMALIZED_ARB,		/* DataType */
+   5,					/* RedBits */
+   5,					/* GreenBits */
+   5,					/* BlueBits */
+   1,					/* AlphaBits */
+   0,					/* LuminanceBits */
+   0,					/* IntensityBits */
+   0,					/* IndexBits */
+   0,					/* DepthBits */
+   0,					/* StencilBits */
+   2,					/* TexelBytes */
+   _mesa_texstore_rgba5551,		/* StoreTexImageFunc */
+   fetch_texel_1d_rgba5551,		/* FetchTexel1D */
+   fetch_texel_2d_rgba5551,		/* FetchTexel2D */
+   fetch_texel_3d_rgba5551,		/* FetchTexel3D */
+   NULL,				/* FetchTexel1Df */
+   NULL,				/* FetchTexel2Df */
+   NULL,				/* FetchTexel3Df */
+   store_texel_rgba5551			/* StoreTexel */
+};
+
 const struct gl_texture_format _mesa_texformat_argb1555 = {
    MESA_FORMAT_ARGB1555,		/* MesaFormat */
    GL_RGBA,				/* BaseFormat */
@@ -1433,41 +1506,39 @@ _mesa_choose_tex_format( GLcontext *ctx, GLint internalFormat,
       }
    }
 
-   if (ctx->Extensions.ARB_texture_compression) {
-      switch (internalFormat) {
-         case GL_COMPRESSED_ALPHA_ARB:
-            return &_mesa_texformat_alpha;
-         case GL_COMPRESSED_LUMINANCE_ARB:
-            return &_mesa_texformat_luminance;
-         case GL_COMPRESSED_LUMINANCE_ALPHA_ARB:
-            return &_mesa_texformat_luminance_alpha;
-         case GL_COMPRESSED_INTENSITY_ARB:
-            return &_mesa_texformat_intensity;
-         case GL_COMPRESSED_RGB_ARB:
+   switch (internalFormat) {
+      case GL_COMPRESSED_ALPHA_ARB:
+         return &_mesa_texformat_alpha;
+      case GL_COMPRESSED_LUMINANCE_ARB:
+         return &_mesa_texformat_luminance;
+      case GL_COMPRESSED_LUMINANCE_ALPHA_ARB:
+         return &_mesa_texformat_luminance_alpha;
+      case GL_COMPRESSED_INTENSITY_ARB:
+         return &_mesa_texformat_intensity;
+      case GL_COMPRESSED_RGB_ARB:
 #if FEATURE_texture_fxt1
-            if (ctx->Extensions.TDFX_texture_compression_FXT1)
-               return &_mesa_texformat_rgb_fxt1;
+         if (ctx->Extensions.TDFX_texture_compression_FXT1)
+            return &_mesa_texformat_rgb_fxt1;
 #endif
 #if FEATURE_texture_s3tc
-            if (ctx->Extensions.EXT_texture_compression_s3tc ||
-                ctx->Extensions.S3_s3tc)
-               return &_mesa_texformat_rgb_dxt1;
+         if (ctx->Extensions.EXT_texture_compression_s3tc ||
+             ctx->Extensions.S3_s3tc)
+            return &_mesa_texformat_rgb_dxt1;
 #endif
-            return &_mesa_texformat_rgb;
-         case GL_COMPRESSED_RGBA_ARB:
+         return &_mesa_texformat_rgb;
+      case GL_COMPRESSED_RGBA_ARB:
 #if FEATURE_texture_fxt1
-            if (ctx->Extensions.TDFX_texture_compression_FXT1)
-               return &_mesa_texformat_rgba_fxt1;
+         if (ctx->Extensions.TDFX_texture_compression_FXT1)
+            return &_mesa_texformat_rgba_fxt1;
 #endif
 #if FEATURE_texture_s3tc
-            if (ctx->Extensions.EXT_texture_compression_s3tc ||
-                ctx->Extensions.S3_s3tc)
-               return &_mesa_texformat_rgba_dxt3; /* Not rgba_dxt1, see spec */
+         if (ctx->Extensions.EXT_texture_compression_s3tc ||
+             ctx->Extensions.S3_s3tc)
+            return &_mesa_texformat_rgba_dxt3; /* Not rgba_dxt1, see spec */
 #endif
-            return &_mesa_texformat_rgba;
-         default:
-            ; /* fallthrough */
-      }
+         return &_mesa_texformat_rgba;
+      default:
+         ; /* fallthrough */
    }
 
    if (ctx->Extensions.MESA_ycbcr_texture) {
@@ -1578,21 +1649,40 @@ _mesa_choose_tex_format( GLcontext *ctx, GLint internalFormat,
          case GL_SLUMINANCE_ALPHA_EXT:
          case GL_SLUMINANCE8_ALPHA8_EXT:
             return &_mesa_texformat_sla8;
-         /* NOTE: not supporting any compression of sRGB at this time */
-         case GL_COMPRESSED_SRGB_EXT:
-            return &_mesa_texformat_srgb8;
-         case GL_COMPRESSED_SRGB_ALPHA_EXT:
-            return &_mesa_texformat_srgba8;
          case GL_COMPRESSED_SLUMINANCE_EXT:
             return &_mesa_texformat_sl8;
          case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT:
             return &_mesa_texformat_sla8;
-         case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
+         case GL_COMPRESSED_SRGB_EXT:
+#if FEATURE_texture_s3tc
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgb_dxt1;
+#endif
             return &_mesa_texformat_srgb8;
+         case GL_COMPRESSED_SRGB_ALPHA_EXT:
+#if FEATURE_texture_s3tc
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgba_dxt3; /* Not srgba_dxt1, see spec */
+#endif
+            return &_mesa_texformat_srgba8;
+#if FEATURE_texture_s3tc
+         case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgb_dxt1;
+            break;
          case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgba_dxt1;
+            break;
          case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgba_dxt3;
+            break;
          case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
-            return &_mesa_texformat_srgba8;
+            if (ctx->Extensions.EXT_texture_compression_s3tc)
+               return &_mesa_texformat_srgba_dxt5;
+            break;
+#endif
          default:
             ; /* fallthrough */
       }
@@ -1641,7 +1731,7 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format,
    case MESA_FORMAT_ARGB1555:
    case MESA_FORMAT_ARGB1555_REV:
       *datatype = GL_UNSIGNED_SHORT_1_5_5_5_REV;
-      *comps = 3;
+      *comps = 4;
       return;
 
    case MESA_FORMAT_AL88:
@@ -1673,6 +1763,11 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format,
       *comps = 1; /* XXX OK? */
       return;
 
+   case MESA_FORMAT_S8_Z24:
+      *datatype = GL_UNSIGNED_INT;
+      *comps = 1; /* XXX OK? */
+      return;
+
    case MESA_FORMAT_Z16:
       *datatype = GL_UNSIGNED_SHORT;
       *comps = 1;
@@ -1683,11 +1778,13 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format,
       *comps = 1;
       return;
 
+#if FEATURE_EXT_texture_sRGB
    case MESA_FORMAT_SRGB8:
       *datatype = GL_UNSIGNED_BYTE;
       *comps = 3;
       return;
    case MESA_FORMAT_SRGBA8:
+   case MESA_FORMAT_SARGB8:
       *datatype = GL_UNSIGNED_BYTE;
       *comps = 4;
       return;
@@ -1699,17 +1796,28 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format,
       *datatype = GL_UNSIGNED_BYTE;
       *comps = 2;
       return;
+#endif
 
+#if FEATURE_texture_fxt1
    case MESA_FORMAT_RGB_FXT1:
    case MESA_FORMAT_RGBA_FXT1:
+#endif
+#if FEATURE_texture_s3tc
    case MESA_FORMAT_RGB_DXT1:
    case MESA_FORMAT_RGBA_DXT1:
    case MESA_FORMAT_RGBA_DXT3:
    case MESA_FORMAT_RGBA_DXT5:
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
+#endif
       /* XXX generate error instead? */
       *datatype = GL_UNSIGNED_BYTE;
       *comps = 0;
       return;
+#endif
 
    case MESA_FORMAT_RGBA:
       *datatype = CHAN_TYPE;
diff --git a/src/mesa/main/texformat.h b/src/mesa/main/texformat.h
index 18900fe08b..31364c36b1 100644
--- a/src/mesa/main/texformat.h
+++ b/src/mesa/main/texformat.h
@@ -3,6 +3,7 @@
  * Version:  6.5.1
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -70,8 +71,10 @@ enum _format {
    MESA_FORMAT_BGR888,		/*           BBBB BBBB GGGG GGGG RRRR RRRR */
    MESA_FORMAT_RGB565,		/*                     RRRR RGGG GGGB BBBB */
    MESA_FORMAT_RGB565_REV,	/*                     GGGB BBBB RRRR RGGG */
+   MESA_FORMAT_RGBA4444,        /*                     RRRR GGGG BBBB AAAA */
    MESA_FORMAT_ARGB4444,	/*                     AAAA RRRR GGGG BBBB */
    MESA_FORMAT_ARGB4444_REV,	/*                     GGGG BBBB AAAA RRRR */
+   MESA_FORMAT_RGBA5551,        /*                     RRRR RGGG GGBB BBBA */
    MESA_FORMAT_ARGB1555,	/*                     ARRR RRGG GGGB BBBB */
    MESA_FORMAT_ARGB1555_REV,	/*                     GGGB BBBB ARRR RRGG */
    MESA_FORMAT_AL88,		/*                     AAAA AAAA LLLL LLLL */
@@ -96,9 +99,15 @@ enum _format {
    /*@{*/
    MESA_FORMAT_SRGB8,
    MESA_FORMAT_SRGBA8,
+   MESA_FORMAT_SARGB8,
    MESA_FORMAT_SL8,
    MESA_FORMAT_SLA8,
+#if FEATURE_texture_s3tc
    MESA_FORMAT_SRGB_DXT1,
+   MESA_FORMAT_SRGBA_DXT1,
+   MESA_FORMAT_SRGBA_DXT3,
+   MESA_FORMAT_SRGBA_DXT5,
+#endif
    /*@}*/
 #endif
 
@@ -106,12 +115,16 @@ enum _format {
     * \name Compressed texture formats.
     */
    /*@{*/
+#if FEATURE_texture_fxt1
    MESA_FORMAT_RGB_FXT1,
    MESA_FORMAT_RGBA_FXT1,
+#endif
+#if FEATURE_texture_s3tc
    MESA_FORMAT_RGB_DXT1,
    MESA_FORMAT_RGBA_DXT1,
    MESA_FORMAT_RGBA_DXT3,
    MESA_FORMAT_RGBA_DXT5,
+#endif
    /*@}*/
 
    /**
@@ -168,9 +181,15 @@ extern const struct gl_texture_format _mesa_texformat_intensity;
 /*@{*/
 extern const struct gl_texture_format _mesa_texformat_srgb8;
 extern const struct gl_texture_format _mesa_texformat_srgba8;
+extern const struct gl_texture_format _mesa_texformat_sargb8;
 extern const struct gl_texture_format _mesa_texformat_sl8;
 extern const struct gl_texture_format _mesa_texformat_sla8;
+#if FEATURE_texture_s3tc
 extern const struct gl_texture_format _mesa_texformat_srgb_dxt1;
+extern const struct gl_texture_format _mesa_texformat_srgba_dxt1;
+extern const struct gl_texture_format _mesa_texformat_srgba_dxt3;
+extern const struct gl_texture_format _mesa_texformat_srgba_dxt5;
+#endif
 /*@}*/
 #endif
 
@@ -200,10 +219,12 @@ extern const struct gl_texture_format _mesa_texformat_rgb888;
 extern const struct gl_texture_format _mesa_texformat_bgr888;
 extern const struct gl_texture_format _mesa_texformat_rgb565;
 extern const struct gl_texture_format _mesa_texformat_rgb565_rev;
+extern const struct gl_texture_format _mesa_texformat_rgba4444;
 extern const struct gl_texture_format _mesa_texformat_argb4444;
 extern const struct gl_texture_format _mesa_texformat_argb4444_rev;
 extern const struct gl_texture_format _mesa_texformat_argb1555;
 extern const struct gl_texture_format _mesa_texformat_argb1555_rev;
+extern const struct gl_texture_format _mesa_texformat_rgba5551;
 extern const struct gl_texture_format _mesa_texformat_al88;
 extern const struct gl_texture_format _mesa_texformat_al88_rev;
 extern const struct gl_texture_format _mesa_texformat_rgb332;
@@ -225,12 +246,16 @@ extern const struct gl_texture_format _mesa_texformat_ycbcr_rev;
 
 /** \name Compressed formats */
 /*@{*/
+#if FEATURE_texture_fxt1
 extern const struct gl_texture_format _mesa_texformat_rgb_fxt1;
 extern const struct gl_texture_format _mesa_texformat_rgba_fxt1;
+#endif
+#if FEATURE_texture_s3tc
 extern const struct gl_texture_format _mesa_texformat_rgb_dxt1;
 extern const struct gl_texture_format _mesa_texformat_rgba_dxt1;
 extern const struct gl_texture_format _mesa_texformat_rgba_dxt3;
 extern const struct gl_texture_format _mesa_texformat_rgba_dxt5;
+#endif
 /*@}*/
 
 /** \name The null format */
diff --git a/src/mesa/main/texformat_tmp.h b/src/mesa/main/texformat_tmp.h
index 7499ba7b36..275340cabd 100644
--- a/src/mesa/main/texformat_tmp.h
+++ b/src/mesa/main/texformat_tmp.h
@@ -3,6 +3,7 @@
  * Version:  6.5.1
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -694,7 +695,7 @@ static void store_texel_argb8888_rev(struct gl_texture_image *texImage,
 {
    const GLubyte *rgba = (const GLubyte *) texel;
    GLuint *dst = TEXEL_ADDR(GLuint, texImage, i, j, k, 1);
-   *dst = PACK_COLOR_8888(rgba[ACOMP], rgba[RCOMP], rgba[GCOMP], rgba[BCOMP]);
+   *dst = PACK_COLOR_8888(rgba[BCOMP], rgba[GCOMP], rgba[RCOMP], rgba[ACOMP]);
 }
 #endif
 
@@ -803,6 +804,30 @@ static void store_texel_rgb565_rev(struct gl_texture_image *texImage,
 }
 #endif
 
+/* MESA_FORMAT_RGBA4444 ******************************************************/
+
+/* Fetch texel from 1D, 2D or 3D argb444 texture, return 4 GLchans */
+static void FETCH(rgba4444)( const struct gl_texture_image *texImage,
+			     GLint i, GLint j, GLint k, GLchan *texel )
+{
+   const GLushort *src = TEXEL_ADDR(GLushort, texImage, i, j, k, 1);
+   const GLushort s = *src;
+   texel[RCOMP] = UBYTE_TO_CHAN( ((s >> 12) & 0xf) | ((s >> 8) & 0xf0) );
+   texel[GCOMP] = UBYTE_TO_CHAN( ((s >>  8) & 0xf) | ((s >> 4) & 0xf0) );
+   texel[BCOMP] = UBYTE_TO_CHAN( ((s >>  4) & 0xf) | ((s     ) & 0xf0) );
+   texel[ACOMP] = UBYTE_TO_CHAN( ((s      ) & 0xf) | ((s << 4) & 0xf0) );
+}
+
+#if DIM == 3
+static void store_texel_rgba4444(struct gl_texture_image *texImage,
+                                 GLint i, GLint j, GLint k, const void *texel)
+{
+   const GLubyte *rgba = (const GLubyte *) texel;
+   GLushort *dst = TEXEL_ADDR(GLushort, texImage, i, j, k, 1);
+   *dst = PACK_COLOR_4444(rgba[RCOMP], rgba[GCOMP], rgba[BCOMP], rgba[ACOMP]);
+}
+#endif
+
 
 /* MESA_FORMAT_ARGB4444 ******************************************************/
 
@@ -824,7 +849,7 @@ static void store_texel_argb4444(struct gl_texture_image *texImage,
 {
    const GLubyte *rgba = (const GLubyte *) texel;
    GLushort *dst = TEXEL_ADDR(GLushort, texImage, i, j, k, 1);
-   *dst = PACK_COLOR_4444(rgba[RCOMP], rgba[GCOMP], rgba[BCOMP], rgba[ACOMP]);
+   *dst = PACK_COLOR_4444(rgba[ACOMP], rgba[RCOMP], rgba[GCOMP], rgba[BCOMP]);
 }
 #endif
 
@@ -852,6 +877,29 @@ static void store_texel_argb4444_rev(struct gl_texture_image *texImage,
 }
 #endif
 
+/* MESA_FORMAT_RGBA5551 ******************************************************/
+
+/* Fetch texel from 1D, 2D or 3D argb1555 texture, return 4 GLchans */
+static void FETCH(rgba5551)( const struct gl_texture_image *texImage,
+			     GLint i, GLint j, GLint k, GLchan *texel )
+{
+   const GLushort *src = TEXEL_ADDR(GLushort, texImage, i, j, k, 1);
+   const GLushort s = *src;
+   texel[RCOMP] = UBYTE_TO_CHAN( ((s >>  8) & 0xf8) | ((s >> 13) & 0x7) );
+   texel[GCOMP] = UBYTE_TO_CHAN( ((s >>  3) & 0xf8) | ((s >>  8) & 0x7) );
+   texel[BCOMP] = UBYTE_TO_CHAN( ((s <<  2) & 0xf8) | ((s >>  3) & 0x7) );
+   texel[ACOMP] = UBYTE_TO_CHAN( ((s) & 0x01) ? 255 : 0);
+}
+
+#if DIM == 3
+static void store_texel_rgba5551(struct gl_texture_image *texImage,
+                                 GLint i, GLint j, GLint k, const void *texel)
+{
+   const GLubyte *rgba = (const GLubyte *) texel;
+   GLushort *dst = TEXEL_ADDR(GLushort, texImage, i, j, k, 1);
+   *dst = PACK_COLOR_5551(rgba[RCOMP], rgba[GCOMP], rgba[BCOMP], rgba[ACOMP]);
+}
+#endif
 
 /* MESA_FORMAT_ARGB1555 ******************************************************/
 
@@ -1175,11 +1223,11 @@ static void store_texel_srgb8(struct gl_texture_image *texImage,
 static void FETCH(srgba8)(const struct gl_texture_image *texImage,
                           GLint i, GLint j, GLint k, GLfloat *texel )
 {
-   const GLubyte *src = TEXEL_ADDR(GLubyte, texImage, i, j, k, 4);
-   texel[RCOMP] = nonlinear_to_linear(src[0]);
-   texel[GCOMP] = nonlinear_to_linear(src[1]);
-   texel[BCOMP] = nonlinear_to_linear(src[2]);
-   texel[ACOMP] = UBYTE_TO_FLOAT(src[3]); /* linear! */
+   const GLuint s = *TEXEL_ADDR(GLuint, texImage, i, j, k, 1);
+   texel[RCOMP] = nonlinear_to_linear( (s >> 24) );
+   texel[GCOMP] = nonlinear_to_linear( (s >> 16) & 0xff );
+   texel[BCOMP] = nonlinear_to_linear( (s >>  8) & 0xff );
+   texel[ACOMP] = UBYTE_TO_FLOAT( (s      ) & 0xff ); /* linear! */
 }
 
 #if DIM == 3
@@ -1187,10 +1235,29 @@ static void store_texel_srgba8(struct gl_texture_image *texImage,
                                GLint i, GLint j, GLint k, const void *texel)
 {
    const GLubyte *rgba = (const GLubyte *) texel;
-   GLubyte *dst = TEXEL_ADDR(GLubyte, texImage, i, j, k, 4);
-   dst[0] = rgba[RCOMP];
-   dst[1] = rgba[GCOMP];
-   dst[2] = rgba[BCOMP];
+   GLuint *dst = TEXEL_ADDR(GLuint, texImage, i, j, k, 1);
+   *dst = PACK_COLOR_8888(rgba[RCOMP], rgba[GCOMP], rgba[BCOMP], rgba[ACOMP]);
+}
+#endif
+
+/* Fetch texel from 1D, 2D or 3D sargb8 texture, return 4 GLfloats */
+static void FETCH(sargb8)(const struct gl_texture_image *texImage,
+                          GLint i, GLint j, GLint k, GLfloat *texel )
+{
+   const GLuint s = *TEXEL_ADDR(GLuint, texImage, i, j, k, 1);
+   texel[RCOMP] = nonlinear_to_linear( (s >> 16) & 0xff );
+   texel[GCOMP] = nonlinear_to_linear( (s >>  8) & 0xff );
+   texel[BCOMP] = nonlinear_to_linear( (s      ) & 0xff );
+   texel[ACOMP] = UBYTE_TO_FLOAT( (s >> 24) ); /* linear! */
+}
+
+#if DIM == 3
+static void store_texel_sargb8(struct gl_texture_image *texImage,
+                               GLint i, GLint j, GLint k, const void *texel)
+{
+   const GLubyte *rgba = (const GLubyte *) texel;
+   GLuint *dst = TEXEL_ADDR(GLuint, texImage, i, j, k, 1);
+   *dst = PACK_COLOR_8888(rgba[ACOMP], rgba[RCOMP], rgba[GCOMP], rgba[BCOMP]);
 }
 #endif
 
diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c
index 9e968ba743..8fd69c4f6a 100644
--- a/src/mesa/main/teximage.c
+++ b/src/mesa/main/teximage.c
@@ -1,8 +1,9 @@
 /*
  * mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.5
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -37,6 +38,7 @@
 #endif
 #include "fbobject.h"
 #include "framebuffer.h"
+#include "hash.h"
 #include "image.h"
 #include "imports.h"
 #include "macros.h"
@@ -253,23 +255,21 @@ _mesa_base_tex_format( GLcontext *ctx, GLint internalFormat )
       }
    }
 
-   if (ctx->Extensions.ARB_texture_compression) {
-      switch (internalFormat) {
-         case GL_COMPRESSED_ALPHA:
-            return GL_ALPHA;
-         case GL_COMPRESSED_LUMINANCE:
-            return GL_LUMINANCE;
-         case GL_COMPRESSED_LUMINANCE_ALPHA:
-            return GL_LUMINANCE_ALPHA;
-         case GL_COMPRESSED_INTENSITY:
-            return GL_INTENSITY;
-         case GL_COMPRESSED_RGB:
-            return GL_RGB;
-         case GL_COMPRESSED_RGBA:
-            return GL_RGBA;
-         default:
-            ; /* fallthrough */
-      }
+   switch (internalFormat) {
+   case GL_COMPRESSED_ALPHA:
+      return GL_ALPHA;
+   case GL_COMPRESSED_LUMINANCE:
+      return GL_LUMINANCE;
+   case GL_COMPRESSED_LUMINANCE_ALPHA:
+      return GL_LUMINANCE_ALPHA;
+   case GL_COMPRESSED_INTENSITY:
+      return GL_INTENSITY;
+   case GL_COMPRESSED_RGB:
+      return GL_RGB;
+   case GL_COMPRESSED_RGBA:
+      return GL_RGBA;
+   default:
+      ; /* fallthrough */
    }
          
    if (ctx->Extensions.TDFX_texture_compression_FXT1) {
@@ -925,6 +925,7 @@ struct gl_texture_image *
 _mesa_get_proxy_tex_image(GLcontext *ctx, GLenum target, GLint level)
 {
    struct gl_texture_image *texImage;
+   GLuint texIndex;
 
    if (level < 0 )
       return NULL;
@@ -933,111 +934,54 @@ _mesa_get_proxy_tex_image(GLcontext *ctx, GLenum target, GLint level)
    case GL_PROXY_TEXTURE_1D:
       if (level >= ctx->Const.MaxTextureLevels)
          return NULL;
-      texImage = ctx->Texture.ProxyTex[TEXTURE_1D_INDEX]->Image[0][level];
-      if (!texImage) {
-         texImage = ctx->Driver.NewTextureImage(ctx);
-         if (!texImage) {
-            _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation");
-            return NULL;
-         }
-         ctx->Texture.ProxyTex[TEXTURE_1D_INDEX]->Image[0][level] = texImage;
-         /* Set the 'back' pointer */
-         texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_1D_INDEX];
-      }
-      return texImage;
+      texIndex = TEXTURE_1D_INDEX;
+      break;
    case GL_PROXY_TEXTURE_2D:
       if (level >= ctx->Const.MaxTextureLevels)
          return NULL;
-      texImage = ctx->Texture.ProxyTex[TEXTURE_2D_INDEX]->Image[0][level];
-      if (!texImage) {
-         texImage = ctx->Driver.NewTextureImage(ctx);
-         if (!texImage) {
-            _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation");
-            return NULL;
-         }
-         ctx->Texture.ProxyTex[TEXTURE_2D_INDEX]->Image[0][level] = texImage;
-         /* Set the 'back' pointer */
-         texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_2D_INDEX];
-      }
-      return texImage;
+      texIndex = TEXTURE_2D_INDEX;
+      break;
    case GL_PROXY_TEXTURE_3D:
       if (level >= ctx->Const.Max3DTextureLevels)
          return NULL;
-      texImage = ctx->Texture.ProxyTex[TEXTURE_3D_INDEX]->Image[0][level];
-      if (!texImage) {
-         texImage = ctx->Driver.NewTextureImage(ctx);
-         if (!texImage) {
-            _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation");
-            return NULL;
-         }
-         ctx->Texture.ProxyTex[TEXTURE_3D_INDEX]->Image[0][level] = texImage;
-         /* Set the 'back' pointer */
-         texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_3D_INDEX];
-      }
-      return texImage;
+      texIndex = TEXTURE_3D_INDEX;
+      break;
    case GL_PROXY_TEXTURE_CUBE_MAP:
       if (level >= ctx->Const.MaxCubeTextureLevels)
          return NULL;
-      texImage = ctx->Texture.ProxyTex[TEXTURE_CUBE_INDEX]->Image[0][level];
-      if (!texImage) {
-         texImage = ctx->Driver.NewTextureImage(ctx);
-         if (!texImage) {
-            _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation");
-            return NULL;
-         }
-         ctx->Texture.ProxyTex[TEXTURE_CUBE_INDEX]->Image[0][level] = texImage;
-         /* Set the 'back' pointer */
-         texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_CUBE_INDEX];
-      }
-      return texImage;
+      texIndex = TEXTURE_CUBE_INDEX;
+      break;
    case GL_PROXY_TEXTURE_RECTANGLE_NV:
       if (level > 0)
          return NULL;
-      texImage = ctx->Texture.ProxyTex[TEXTURE_RECT_INDEX]->Image[0][level];
-      if (!texImage) {
-         texImage = ctx->Driver.NewTextureImage(ctx);
-         if (!texImage) {
-            _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation");
-            return NULL;
-         }
-         ctx->Texture.ProxyTex[TEXTURE_RECT_INDEX]->Image[0][level] = texImage;
-         /* Set the 'back' pointer */
-         texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_RECT_INDEX];
-      }
-      return texImage;
+      texIndex = TEXTURE_RECT_INDEX;
+      break;
    case GL_PROXY_TEXTURE_1D_ARRAY_EXT:
       if (level >= ctx->Const.MaxTextureLevels)
          return NULL;
-      texImage = ctx->Texture.ProxyTex[TEXTURE_1D_ARRAY_INDEX]->Image[0][level];
-      if (!texImage) {
-         texImage = ctx->Driver.NewTextureImage(ctx);
-         if (!texImage) {
-            _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation");
-            return NULL;
-         }
-         ctx->Texture.ProxyTex[TEXTURE_1D_ARRAY_INDEX]->Image[0][level] = texImage;
-         /* Set the 'back' pointer */
-         texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_1D_ARRAY_INDEX];
-      }
-      return texImage;
+      texIndex = TEXTURE_1D_ARRAY_INDEX;
+      break;
    case GL_PROXY_TEXTURE_2D_ARRAY_EXT:
       if (level >= ctx->Const.MaxTextureLevels)
          return NULL;
-      texImage = ctx->Texture.ProxyTex[TEXTURE_2D_ARRAY_INDEX]->Image[0][level];
-      if (!texImage) {
-         texImage = ctx->Driver.NewTextureImage(ctx);
-         if (!texImage) {
-            _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation");
-            return NULL;
-         }
-         ctx->Texture.ProxyTex[TEXTURE_2D_ARRAY_INDEX]->Image[0][level] = texImage;
-         /* Set the 'back' pointer */
-         texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_2D_ARRAY_INDEX];
-      }
-      return texImage;
+      texIndex = TEXTURE_2D_ARRAY_INDEX;
+      break;
    default:
       return NULL;
    }
+
+   texImage = ctx->Texture.ProxyTex[texIndex]->Image[0][level];
+   if (!texImage) {
+      texImage = ctx->Driver.NewTextureImage(ctx);
+      if (!texImage) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation");
+         return NULL;
+      }
+      ctx->Texture.ProxyTex[texIndex]->Image[0][level] = texImage;
+      /* Set the 'back' pointer */
+      texImage->TexObject = ctx->Texture.ProxyTex[texIndex];
+   }
+   return texImage;
 }
 
 
@@ -1217,19 +1161,30 @@ _mesa_init_teximage_fields(GLcontext *ctx, GLenum target,
    img->Width = width;
    img->Height = height;
    img->Depth = depth;
+
    img->Width2 = width - 2 * border;   /* == 1 << img->WidthLog2; */
-   img->Height2 = height - 2 * border; /* == 1 << img->HeightLog2; */
-   img->Depth2 = depth - 2 * border;   /* == 1 << img->DepthLog2; */
    img->WidthLog2 = logbase2(img->Width2);
-   if (height == 1)  /* 1-D texture */
+
+   if (height == 1) { /* 1-D texture */
+      img->Height2 = 1;
       img->HeightLog2 = 0;
-   else
+   }
+   else {
+      img->Height2 = height - 2 * border; /* == 1 << img->HeightLog2; */
       img->HeightLog2 = logbase2(img->Height2);
-   if (depth == 1)   /* 2-D texture */
+   }
+
+   if (depth == 1) {  /* 2-D texture */
+      img->Depth2 = 1;
       img->DepthLog2 = 0;
-   else
+   }
+   else {
+      img->Depth2 = depth - 2 * border;   /* == 1 << img->DepthLog2; */
       img->DepthLog2 = logbase2(img->Depth2);
+   }
+
    img->MaxLog2 = MAX2(img->WidthLog2, img->HeightLog2);
+
    img->IsCompressed = GL_FALSE;
    img->CompressedSize = 0;
 
@@ -2224,7 +2179,8 @@ copytexsubimage_error_check2( GLcontext *ctx, GLuint dimensions,
 
    if (!_mesa_source_buffer_exists(ctx, teximage->_BaseFormat)) {
       _mesa_error(ctx, GL_INVALID_OPERATION,
-               "glCopyTexSubImage%dD(missing readbuffer)", dimensions);
+                  "glCopyTexSubImage%dD(missing readbuffer, format=0x%x)",
+                  dimensions, teximage->_BaseFormat);
       return GL_TRUE;
    }
 
@@ -2380,23 +2336,33 @@ _mesa_GetTexImage( GLenum target, GLint level, GLenum format,
 }
 
 
+/** Callback info for walking over FBO hash table */
+struct cb_info
+{
+   GLcontext *ctx;
+   struct gl_texture_object *texObj;
+   GLuint level, face;
+};
+
 
 /**
- * Check if the given texture image is bound to any framebuffer objects
- * and update/invalidate them.
- * XXX We're only checking the currently bound framebuffer object for now.
- * In the future, perhaps struct gl_texture_image should have a pointer (or
- * list of pointers (yikes)) to the gl_framebuffer(s) which it's bound to.
+ * Check render to texture callback.  Called from _mesa_HashWalk().
  */
 static void
-update_fbo_texture(GLcontext *ctx, struct gl_texture_object *texObj,
-                   GLuint face, GLuint level)
+check_rtt_cb(GLuint key, void *data, void *userData)
 {
-   if (ctx->DrawBuffer->Name) {
+   struct gl_framebuffer *fb = (struct gl_framebuffer *) data;
+   const struct cb_info *info = (struct cb_info *) userData;
+   GLcontext *ctx = info->ctx;
+   const struct gl_texture_object *texObj = info->texObj;
+   const GLuint level = info->level, face = info->face;
+
+   /* If this is a user-created FBO */
+   if (fb->Name) {
       GLuint i;
+      /* check if any of the FBO's attachments point to 'texObj' */
       for (i = 0; i < BUFFER_COUNT; i++) {
-         struct gl_renderbuffer_attachment *att = 
-            ctx->DrawBuffer->Attachment + i;
+         struct gl_renderbuffer_attachment *att = fb->Attachment + i;
          if (att->Type == GL_TEXTURE &&
              att->Texture == texObj &&
              att->TextureLevel == level &&
@@ -2404,12 +2370,36 @@ update_fbo_texture(GLcontext *ctx, struct gl_texture_object *texObj,
             ASSERT(att->Texture->Image[att->CubeMapFace][att->TextureLevel]);
             /* Tell driver about the new renderbuffer texture */
             ctx->Driver.RenderTexture(ctx, ctx->DrawBuffer, att);
+            /* Mark fb status as indeterminate to force re-validation */
+            fb->_Status = 0;
          }
       }
    }
 }
 
 
+/**
+ * When a texture image is specified we have to check if it's bound to
+ * any framebuffer objects (render to texture) in order to detect changes
+ * in size or format since that effects FBO completeness.
+ * Any FBOs rendering into the texture must be re-validated.
+ */
+static void
+update_fbo_texture(GLcontext *ctx, struct gl_texture_object *texObj,
+                   GLuint face, GLuint level)
+{
+   /* Only check this texture if it's been marked as RenderToTexture */
+   if (texObj->_RenderToTexture) {
+      struct cb_info info;
+      info.ctx = ctx;
+      info.texObj = texObj;
+      info.level = level;
+      info.face = face;
+      _mesa_HashWalk(ctx->Shared->FrameBuffers, check_rtt_cb, &info);
+   }
+}
+
+
 
 /*
  * Called from the API.  Note that width includes the border.
@@ -2441,7 +2431,7 @@ _mesa_TexImage1D( GLenum target, GLint level, GLint internalFormat,
          return;   /* error was recorded */
       }
 
-      if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+      if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
 	 _mesa_update_state(ctx);
 
       texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
@@ -2547,7 +2537,7 @@ _mesa_TexImage2D( GLenum target, GLint level, GLint internalFormat,
          return;   /* error was recorded */
       }
 
-      if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+      if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
 	 _mesa_update_state(ctx);
 
       texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
@@ -2603,7 +2593,7 @@ _mesa_TexImage2D( GLenum target, GLint level, GLint internalFormat,
                               1, border)) {
          /* when error, clear all proxy texture image parameters */
          if (texImage)
-            clear_teximage_fields(ctx->Texture.ProxyTex[TEXTURE_2D_INDEX]->Image[0][level]);
+            clear_teximage_fields(texImage);
       }
       else {
          /* no error, set the tex image parameters */
@@ -2648,7 +2638,7 @@ _mesa_TexImage3D( GLenum target, GLint level, GLint internalFormat,
          return;   /* error was recorded */
       }
 
-      if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+      if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
 	 _mesa_update_state(ctx);
 
       texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
@@ -2741,7 +2731,7 @@ _mesa_TexSubImage1D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
       _mesa_update_state(ctx);
 
 #if FEATURE_convolve
@@ -2801,7 +2791,7 @@ _mesa_TexSubImage2D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
       _mesa_update_state(ctx);
 
 #if FEATURE_convolve
@@ -2861,7 +2851,7 @@ _mesa_TexSubImage3D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
       _mesa_update_state(ctx);
 
    if (subtexture_error_check(ctx, 3, target, level, xoffset, yoffset, zoffset,
@@ -2917,7 +2907,7 @@ _mesa_CopyTexImage1D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
       _mesa_update_state(ctx);
 
 #if FEATURE_convolve
@@ -2982,7 +2972,7 @@ _mesa_CopyTexImage2D( GLenum target, GLint level, GLenum internalFormat,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
       _mesa_update_state(ctx);
 
 #if FEATURE_convolve
@@ -3050,7 +3040,7 @@ _mesa_CopyTexSubImage1D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
       _mesa_update_state(ctx);
 
    if (copytexsubimage_error_check1(ctx, 1, target, level))
@@ -3105,7 +3095,7 @@ _mesa_CopyTexSubImage2D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
       _mesa_update_state(ctx);
 
    if (copytexsubimage_error_check1(ctx, 2, target, level))
@@ -3160,7 +3150,7 @@ _mesa_CopyTexSubImage3D( GLenum target, GLint level,
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
-   if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE)
+   if (ctx->NewState & _MESA_NEW_TRANSFER_STATE)
       _mesa_update_state(ctx);
 
    if (copytexsubimage_error_check1(ctx, 3, target, level))
diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c
index 2a54ff7ff9..b4e30c8bb7 100644
--- a/src/mesa/main/texobj.c
+++ b/src/mesa/main/texobj.c
@@ -42,6 +42,8 @@
 #include "texstate.h"
 #include "texobj.h"
 #include "mtypes.h"
+#include "shader/prog_instruction.h"
+
 
 
 /**********************************************************************/
@@ -132,12 +134,15 @@ _mesa_initialize_texture_object( struct gl_texture_object *obj,
    obj->BaseLevel = 0;
    obj->MaxLevel = 1000;
    obj->MaxAnisotropy = 1.0;
-   obj->CompareFlag = GL_FALSE;                      /* SGIX_shadow */
-   obj->CompareOperator = GL_TEXTURE_LEQUAL_R_SGIX;  /* SGIX_shadow */
    obj->CompareMode = GL_NONE;         /* ARB_shadow */
    obj->CompareFunc = GL_LEQUAL;       /* ARB_shadow */
+   obj->CompareFailValue = 0.0F;       /* ARB_shadow_ambient */
    obj->DepthMode = GL_LUMINANCE;      /* ARB_depth_texture */
-   obj->ShadowAmbient = 0.0F;          /* ARB/SGIX_shadow_ambient */
+   obj->Swizzle[0] = GL_RED;
+   obj->Swizzle[1] = GL_GREEN;
+   obj->Swizzle[2] = GL_BLUE;
+   obj->Swizzle[3] = GL_ALPHA;
+   obj->_Swizzle = SWIZZLE_NOOP;
 }
 
 
@@ -241,17 +246,17 @@ _mesa_copy_texture_object( struct gl_texture_object *dest,
    dest->BaseLevel = src->BaseLevel;
    dest->MaxLevel = src->MaxLevel;
    dest->MaxAnisotropy = src->MaxAnisotropy;
-   dest->CompareFlag = src->CompareFlag;
-   dest->CompareOperator = src->CompareOperator;
-   dest->ShadowAmbient = src->ShadowAmbient;
    dest->CompareMode = src->CompareMode;
    dest->CompareFunc = src->CompareFunc;
+   dest->CompareFailValue = src->CompareFailValue;
    dest->DepthMode = src->DepthMode;
    dest->_MaxLevel = src->_MaxLevel;
    dest->_MaxLambda = src->_MaxLambda;
    dest->GenerateMipmap = src->GenerateMipmap;
    dest->Palette = src->Palette;
    dest->_Complete = src->_Complete;
+   COPY_4V(dest->Swizzle, src->Swizzle);
+   dest->_Swizzle = src->_Swizzle;
 }
 
 
@@ -383,11 +388,21 @@ _mesa_test_texobj_completeness( const GLcontext *ctx,
 
    t->_Complete = GL_TRUE;  /* be optimistic */
 
+   /* Detect cases where the application set the base level to an invalid
+    * value.
+    */
+   if ((baseLevel < 0) || (baseLevel > MAX_TEXTURE_LEVELS)) {
+      char s[100];
+      _mesa_sprintf(s, "base level = %d is invalid", baseLevel);
+      incomplete(t, s);
+      t->_Complete = GL_FALSE;
+      return;
+   }
+
    /* Always need the base level image */
    if (!t->Image[0][baseLevel]) {
       char s[100];
-      _mesa_sprintf(s, "obj %p (%d) Image[baseLevel=%d] == NULL",
-              (void *) t, t->Name, baseLevel);
+      _mesa_sprintf(s, "Image[baseLevel=%d] == NULL", baseLevel);
       incomplete(t, s);
       t->_Complete = GL_FALSE;
       return;
@@ -750,24 +765,31 @@ unbind_texobj_from_texunits(GLcontext *ctx, struct gl_texture_object *texObj)
       struct gl_texture_unit *unit = &ctx->Texture.Unit[u];
       if (texObj == unit->Current1D) {
          _mesa_reference_texobj(&unit->Current1D, ctx->Shared->Default1D);
+         ASSERT(unit->Current1D);
       }
       else if (texObj == unit->Current2D) {
          _mesa_reference_texobj(&unit->Current2D, ctx->Shared->Default2D);
+         ASSERT(unit->Current2D);
       }
       else if (texObj == unit->Current3D) {
          _mesa_reference_texobj(&unit->Current3D, ctx->Shared->Default3D);
+         ASSERT(unit->Current3D);
       }
       else if (texObj == unit->CurrentCubeMap) {
          _mesa_reference_texobj(&unit->CurrentCubeMap, ctx->Shared->DefaultCubeMap);
+         ASSERT(unit->CurrentCubeMap);
       }
       else if (texObj == unit->CurrentRect) {
          _mesa_reference_texobj(&unit->CurrentRect, ctx->Shared->DefaultRect);
+         ASSERT(unit->CurrentRect);
       }
       else if (texObj == unit->Current1DArray) {
          _mesa_reference_texobj(&unit->Current1DArray, ctx->Shared->Default1DArray);
+         ASSERT(unit->Current1DArray);
       }
       else if (texObj == unit->Current2DArray) {
          _mesa_reference_texobj(&unit->Current2DArray, ctx->Shared->Default2DArray);
+         ASSERT(unit->Current2DArray);
       }
    }
 }
@@ -941,24 +963,31 @@ _mesa_BindTexture( GLenum target, GLuint texName )
    switch (target) {
       case GL_TEXTURE_1D:
          _mesa_reference_texobj(&texUnit->Current1D, newTexObj);
+         ASSERT(texUnit->Current1D);
          break;
       case GL_TEXTURE_2D:
          _mesa_reference_texobj(&texUnit->Current2D, newTexObj);
+         ASSERT(texUnit->Current2D);
          break;
       case GL_TEXTURE_3D:
          _mesa_reference_texobj(&texUnit->Current3D, newTexObj);
+         ASSERT(texUnit->Current3D);
          break;
       case GL_TEXTURE_CUBE_MAP_ARB:
          _mesa_reference_texobj(&texUnit->CurrentCubeMap, newTexObj);
+         ASSERT(texUnit->CurrentCubeMap);
          break;
       case GL_TEXTURE_RECTANGLE_NV:
          _mesa_reference_texobj(&texUnit->CurrentRect, newTexObj);
+         ASSERT(texUnit->CurrentRect);
          break;
       case GL_TEXTURE_1D_ARRAY_EXT:
-         texUnit->Current1DArray = newTexObj;
+         _mesa_reference_texobj(&texUnit->Current1DArray, newTexObj);
+         ASSERT(texUnit->Current1DArray);
          break;
       case GL_TEXTURE_2D_ARRAY_EXT:
-         texUnit->Current2DArray = newTexObj;
+         _mesa_reference_texobj(&texUnit->Current2DArray, newTexObj);
+         ASSERT(texUnit->Current2DArray);
          break;
       default:
          /* Bad target should be caught above */
diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c
index a9e752a637..865aae8627 100644
--- a/src/mesa/main/texparam.c
+++ b/src/mesa/main/texparam.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.5
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -37,6 +38,7 @@
 #include "main/texcompress.h"
 #include "main/texparam.h"
 #include "main/teximage.h"
+#include "shader/prog_instruction.h"
 
 
 /**
@@ -72,404 +74,609 @@ validate_texture_wrap_mode(GLcontext * ctx, GLenum target, GLenum wrap)
 }
 
 
-void GLAPIENTRY
-_mesa_TexParameterf( GLenum target, GLenum pname, GLfloat param )
+/**
+ * Get current texture object for given target.
+ * Return NULL if any error.
+ */
+static struct gl_texture_object *
+get_texobj(GLcontext *ctx, GLenum target)
 {
-   _mesa_TexParameterfv(target, pname, &param);
+   struct gl_texture_unit *texUnit;
+
+   if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureImageUnits) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glTexParameter(current unit)");
+      return NULL;
+   }
+
+   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+
+   switch (target) {
+   case GL_TEXTURE_1D:
+      return texUnit->Current1D;
+   case GL_TEXTURE_2D:
+      return texUnit->Current2D;
+   case GL_TEXTURE_3D:
+      return texUnit->Current3D;
+   case GL_TEXTURE_CUBE_MAP:
+      if (ctx->Extensions.ARB_texture_cube_map) {
+         return texUnit->CurrentCubeMap;
+      }
+      break;
+   case GL_TEXTURE_RECTANGLE_NV:
+      if (ctx->Extensions.NV_texture_rectangle) {
+         return texUnit->CurrentRect;
+      }
+      break;
+   case GL_TEXTURE_1D_ARRAY_EXT:
+      if (ctx->Extensions.MESA_texture_array) {
+         return texUnit->Current1DArray;
+      }
+      break;
+   case GL_TEXTURE_2D_ARRAY_EXT:
+      if (ctx->Extensions.MESA_texture_array) {
+         return texUnit->Current2DArray;
+      }
+      break;
+   default:
+      ;
+   }
+
+   _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(target)");
+   return NULL;
 }
 
 
-void GLAPIENTRY
-_mesa_TexParameterfv( GLenum target, GLenum pname, const GLfloat *params )
+/**
+ * Convert GL_RED/GREEN/BLUE/ALPHA/ZERO/ONE to SWIZZLE_X/Y/Z/W/ZERO/ONE.
+ * \return -1 if error.
+ */
+static GLint
+comp_to_swizzle(GLenum comp)
 {
-   const GLenum eparam = (GLenum) (GLint) params[0];
-   struct gl_texture_unit *texUnit;
-   struct gl_texture_object *texObj;
-   GET_CURRENT_CONTEXT(ctx);
-   ASSERT_OUTSIDE_BEGIN_END(ctx);
+   switch (comp) {
+   case GL_RED:
+      return SWIZZLE_X;
+   case GL_GREEN:
+      return SWIZZLE_Y;
+   case GL_BLUE:
+      return SWIZZLE_Z;
+   case GL_ALPHA:
+      return SWIZZLE_W;
+   case GL_ZERO:
+      return SWIZZLE_ZERO;
+   case GL_ONE:
+      return SWIZZLE_ONE;
+   default:
+      return -1;
+   }
+}
 
-   if (MESA_VERBOSE&(VERBOSE_API|VERBOSE_TEXTURE))
-      _mesa_debug(ctx, "glTexParameter %s %s %.1f(%s)...\n",
-                  _mesa_lookup_enum_by_nr(target),
-                  _mesa_lookup_enum_by_nr(pname),
-                  *params,
-		  _mesa_lookup_enum_by_nr(eparam));
 
-   if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureImageUnits) {
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glTexParameterfv(current unit)");
-      return;
+static void
+set_swizzle_component(GLuint *swizzle, GLuint comp, GLuint swz)
+{
+   ASSERT(comp < 4);
+   ASSERT(swz <= SWIZZLE_NIL);
+   {
+      GLuint mask = 0x7 << (3 * comp);
+      GLuint s = (*swizzle & ~mask) | (swz << (3 * comp));
+      *swizzle = s;
    }
+}
 
-   texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
 
-   switch (target) {
-      case GL_TEXTURE_1D:
-         texObj = texUnit->Current1D;
-         break;
-      case GL_TEXTURE_2D:
-         texObj = texUnit->Current2D;
-         break;
-      case GL_TEXTURE_3D:
-         texObj = texUnit->Current3D;
-         break;
-      case GL_TEXTURE_CUBE_MAP:
-         if (!ctx->Extensions.ARB_texture_cube_map) {
-            _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" );
-            return;
-         }
-         texObj = texUnit->CurrentCubeMap;
-         break;
-      case GL_TEXTURE_RECTANGLE_NV:
-         if (!ctx->Extensions.NV_texture_rectangle) {
-            _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" );
-            return;
-         }
-         texObj = texUnit->CurrentRect;
-         break;
-      case GL_TEXTURE_1D_ARRAY_EXT:
-         if (!ctx->Extensions.MESA_texture_array) {
-            _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" );
-            return;
-         }
-         texObj = texUnit->Current1DArray;
-         break;
-      case GL_TEXTURE_2D_ARRAY_EXT:
-         if (!ctx->Extensions.MESA_texture_array) {
-            _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" );
+/**
+ * This is called just prior to changing any texture object state.
+ * Any pending rendering will be flushed out, we'll set the _NEW_TEXTURE
+ * state flag and then mark the texture object as 'incomplete' so that any
+ * per-texture derived state gets recomputed.
+ */
+static INLINE void
+flush(GLcontext *ctx, struct gl_texture_object *texObj)
+{
+   FLUSH_VERTICES(ctx, _NEW_TEXTURE);
+   texObj->_Complete = GL_FALSE;
+}
+
+
+/** Set an integer-valued texture parameter */
+static void
+set_tex_parameteri(GLcontext *ctx,
+                   struct gl_texture_object *texObj,
+                   GLenum pname, const GLint *params)
+{
+   switch (pname) {
+   case GL_TEXTURE_MIN_FILTER:
+      if (texObj->MinFilter == params[0])
+         return;
+      switch (params[0]) {
+      case GL_NEAREST:
+      case GL_LINEAR:
+         flush(ctx, texObj);
+         texObj->MinFilter = params[0];
+         return;
+      case GL_NEAREST_MIPMAP_NEAREST:
+      case GL_LINEAR_MIPMAP_NEAREST:
+      case GL_NEAREST_MIPMAP_LINEAR:
+      case GL_LINEAR_MIPMAP_LINEAR:
+         if (texObj->Target != GL_TEXTURE_RECTANGLE_NV) {
+            flush(ctx, texObj);
+            texObj->MinFilter = params[0];
             return;
          }
-         texObj = texUnit->Current2DArray;
-         break;
+         /* fall-through */
       default:
-         _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" );
+         _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" );
+      }
+      return;
+
+   case GL_TEXTURE_MAG_FILTER:
+      if (texObj->MagFilter == params[0])
          return;
-   }
+      switch (params[0]) {
+      case GL_NEAREST:
+      case GL_LINEAR:
+         flush(ctx, texObj);
+         texObj->MagFilter = params[0];
+         return;
+      default:
+         _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" );
+      }
+      return;
 
-   switch (pname) {
-      case GL_TEXTURE_MIN_FILTER:
-         /* A small optimization */
-         if (texObj->MinFilter == eparam)
-            return;
-         if (eparam==GL_NEAREST || eparam==GL_LINEAR) {
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            texObj->MinFilter = eparam;
-         }
-         else if ((eparam==GL_NEAREST_MIPMAP_NEAREST ||
-                   eparam==GL_LINEAR_MIPMAP_NEAREST ||
-                   eparam==GL_NEAREST_MIPMAP_LINEAR ||
-                   eparam==GL_LINEAR_MIPMAP_LINEAR) &&
-                  texObj->Target != GL_TEXTURE_RECTANGLE_NV) {
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            texObj->MinFilter = eparam;
-         }
-         else {
-            _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" );
-            return;
-         }
-         break;
-      case GL_TEXTURE_MAG_FILTER:
-         /* A small optimization */
-         if (texObj->MagFilter == eparam)
-            return;
+   case GL_TEXTURE_WRAP_S:
+      if (texObj->WrapS == params[0])
+         return;
+      if (validate_texture_wrap_mode(ctx, texObj->Target, params[0])) {
+         flush(ctx, texObj);
+         texObj->WrapS = params[0];
+      }
+      return;
 
-         if (eparam==GL_NEAREST || eparam==GL_LINEAR) {
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            texObj->MagFilter = eparam;
-         }
-         else {
-            _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" );
-            return;
-         }
-         break;
-      case GL_TEXTURE_WRAP_S:
-         if (texObj->WrapS == eparam)
-            return;
-         if (validate_texture_wrap_mode(ctx, texObj->Target, eparam)) {
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            texObj->WrapS = eparam;
-         }
-         else {
-            return;
-         }
-         break;
-      case GL_TEXTURE_WRAP_T:
-         if (texObj->WrapT == eparam)
-            return;
-         if (validate_texture_wrap_mode(ctx, texObj->Target, eparam)) {
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            texObj->WrapT = eparam;
-         }
-         else {
-            return;
-         }
-         break;
-      case GL_TEXTURE_WRAP_R:
-         if (texObj->WrapR == eparam)
-            return;
-         if (validate_texture_wrap_mode(ctx, texObj->Target, eparam)) {
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            texObj->WrapR = eparam;
-         }
-         else {
-	    return;
-         }
-         break;
-      case GL_TEXTURE_BORDER_COLOR:
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         texObj->BorderColor[RCOMP] = params[0];
-         texObj->BorderColor[GCOMP] = params[1];
-         texObj->BorderColor[BCOMP] = params[2];
-         texObj->BorderColor[ACOMP] = params[3];
-         UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[RCOMP], params[0]);
-         UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[GCOMP], params[1]);
-         UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[BCOMP], params[2]);
-         UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[ACOMP], params[3]);
-         break;
-      case GL_TEXTURE_MIN_LOD:
-         if (texObj->MinLod == params[0])
-            return;
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         texObj->MinLod = params[0];
-         break;
-      case GL_TEXTURE_MAX_LOD:
-         if (texObj->MaxLod == params[0])
-            return;
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         texObj->MaxLod = params[0];
-         break;
-      case GL_TEXTURE_BASE_LEVEL:
-         if (params[0] < 0.0) {
-            _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)");
-            return;
-         }
-         if (target == GL_TEXTURE_RECTANGLE_ARB && params[0] != 0.0) {
-            _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)");
-            return;
-         }
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         texObj->BaseLevel = (GLint) params[0];
-         break;
-      case GL_TEXTURE_MAX_LEVEL:
-         if (params[0] < 0.0) {
-            _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)");
-            return;
-         }
-         if (target == GL_TEXTURE_RECTANGLE_ARB) {
-            _mesa_error(ctx, GL_INVALID_OPERATION, "glTexParameter(param)");
-            return;
-         }
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         texObj->MaxLevel = (GLint) params[0];
-         break;
-      case GL_TEXTURE_PRIORITY:
-         FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-         texObj->Priority = CLAMP( params[0], 0.0F, 1.0F );
-         break;
-      case GL_TEXTURE_MAX_ANISOTROPY_EXT:
-         if (ctx->Extensions.EXT_texture_filter_anisotropic) {
-	    if (params[0] < 1.0) {
-	       _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)" );
-	       return;
-	    }
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            /* clamp to max, that's what NVIDIA does */
-            texObj->MaxAnisotropy = MIN2(params[0],
-                                         ctx->Const.MaxTextureMaxAnisotropy);
+   case GL_TEXTURE_WRAP_T:
+      if (texObj->WrapT == params[0])
+         return;
+      if (validate_texture_wrap_mode(ctx, texObj->Target, params[0])) {
+         flush(ctx, texObj);
+         texObj->WrapT = params[0];
+      }
+      return;
+
+   case GL_TEXTURE_WRAP_R:
+      if (texObj->WrapR == params[0])
+         return;
+      if (validate_texture_wrap_mode(ctx, texObj->Target, params[0])) {
+         flush(ctx, texObj);
+         texObj->WrapR = params[0];
+      }
+      return;
+
+   case GL_TEXTURE_BASE_LEVEL:
+      if (texObj->BaseLevel == params[0])
+         return;
+      if (params[0] < 0 ||
+          (texObj->Target == GL_TEXTURE_RECTANGLE_ARB && params[0] != 0)) {
+         _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)");
+         return;
+      }
+      flush(ctx, texObj);
+      texObj->BaseLevel = params[0];
+      return;
+
+   case GL_TEXTURE_MAX_LEVEL:
+      if (texObj->MaxLevel == params[0])
+         return;
+      if (params[0] < 0 || texObj->Target == GL_TEXTURE_RECTANGLE_ARB) {
+         _mesa_error(ctx, GL_INVALID_OPERATION, "glTexParameter(param)");
+         return;
+      }
+      flush(ctx, texObj);
+      texObj->MaxLevel = params[0];
+      return;
+
+   case GL_GENERATE_MIPMAP_SGIS:
+      if (ctx->Extensions.SGIS_generate_mipmap) {
+         if (texObj->GenerateMipmap != params[0]) {
+            flush(ctx, texObj);
+            texObj->GenerateMipmap = params[0] ? GL_TRUE : GL_FALSE;
          }
-         else {
-            _mesa_error(ctx, GL_INVALID_ENUM,
-                        "glTexParameter(pname=GL_TEXTURE_MAX_ANISOTROPY_EXT)");
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glTexParameter(pname=GL_GENERATE_MIPMAP_SGIS)");
+      }
+      return;
+
+   case GL_TEXTURE_COMPARE_MODE_ARB:
+      if (ctx->Extensions.ARB_shadow &&
+          (params[0] == GL_NONE ||
+           params[0] == GL_COMPARE_R_TO_TEXTURE_ARB)) {
+         if (texObj->CompareMode != params[0]) {
+            flush(ctx, texObj);
+            texObj->CompareMode = params[0];
+         }
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glTexParameter(GL_TEXTURE_COMPARE_MODE_ARB)");
+      }
+      return;
+
+   case GL_TEXTURE_COMPARE_FUNC_ARB:
+      if (ctx->Extensions.ARB_shadow) {
+         if (texObj->CompareFunc == params[0])
             return;
-         }
-         break;
-      case GL_TEXTURE_COMPARE_SGIX:
-         if (ctx->Extensions.SGIX_shadow) {
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            texObj->CompareFlag = params[0] ? GL_TRUE : GL_FALSE;
-         }
-         else {
-            _mesa_error(ctx, GL_INVALID_ENUM,
-                        "glTexParameter(pname=GL_TEXTURE_COMPARE_SGIX)");
+         switch (params[0]) {
+         case GL_LEQUAL:
+         case GL_GEQUAL:
+            flush(ctx, texObj);
+            texObj->CompareFunc = params[0];
             return;
-         }
-         break;
-      case GL_TEXTURE_COMPARE_OPERATOR_SGIX:
-         if (ctx->Extensions.SGIX_shadow) {
-            GLenum op = (GLenum) params[0];
-            if (op == GL_TEXTURE_LEQUAL_R_SGIX ||
-                op == GL_TEXTURE_GEQUAL_R_SGIX) {
-               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-               texObj->CompareOperator = op;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(param)");
+         case GL_EQUAL:
+         case GL_NOTEQUAL:
+         case GL_LESS:
+         case GL_GREATER:
+         case GL_ALWAYS:
+         case GL_NEVER:
+            if (ctx->Extensions.EXT_shadow_funcs) {
+               flush(ctx, texObj);
+               texObj->CompareFunc = params[0];
+               return;
             }
-         }
-         else {
-            _mesa_error(ctx, GL_INVALID_ENUM,
-                    "glTexParameter(pname=GL_TEXTURE_COMPARE_OPERATOR_SGIX)");
-            return;
-         }
-         break;
-      case GL_SHADOW_AMBIENT_SGIX: /* aka GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
-         if (ctx->Extensions.SGIX_shadow_ambient) {
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            texObj->ShadowAmbient = CLAMP(params[0], 0.0F, 1.0F);
-         }
-         else {
+            /* fall-through */
+         default:
             _mesa_error(ctx, GL_INVALID_ENUM,
-                        "glTexParameter(pname=GL_SHADOW_AMBIENT_SGIX)");
-            return;
+                        "glTexParameter(GL_TEXTURE_COMPARE_FUNC_ARB)");
          }
-         break;
-      case GL_GENERATE_MIPMAP_SGIS:
-         if (ctx->Extensions.SGIS_generate_mipmap) {
-            FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-            texObj->GenerateMipmap = params[0] ? GL_TRUE : GL_FALSE;
-         }
-         else {
-            _mesa_error(ctx, GL_INVALID_ENUM,
-                        "glTexParameter(pname=GL_GENERATE_MIPMAP_SGIS)");
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(param)");
+      }
+      return;
+
+   case GL_DEPTH_TEXTURE_MODE_ARB:
+      if (ctx->Extensions.ARB_depth_texture &&
+          (params[0] == GL_LUMINANCE ||
+           params[0] == GL_INTENSITY ||
+           params[0] == GL_ALPHA)) {
+         if (texObj->DepthMode != params[0]) {
+            flush(ctx, texObj);
+            texObj->DepthMode = params[0];
+         }
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glTexParameter(GL_DEPTH_TEXTURE_MODE_ARB)");
+      }
+      return;
+
+#ifdef FEATURE_OES_draw_texture
+   case GL_TEXTURE_CROP_RECT_OES:
+      texObj->CropRect[0] = params[0];
+      texObj->CropRect[1] = params[1];
+      texObj->CropRect[2] = params[2];
+      texObj->CropRect[3] = params[3];
+      return;
+#endif
+
+   case GL_TEXTURE_SWIZZLE_R_EXT:
+   case GL_TEXTURE_SWIZZLE_G_EXT:
+   case GL_TEXTURE_SWIZZLE_B_EXT:
+   case GL_TEXTURE_SWIZZLE_A_EXT:
+      if (ctx->Extensions.EXT_texture_swizzle) {
+         const GLuint comp = pname - GL_TEXTURE_SWIZZLE_R_EXT;
+         const GLint swz = comp_to_swizzle(params[0]);
+         if (swz < 0) {
+            _mesa_error(ctx, GL_INVALID_OPERATION,
+                        "glTexParameter(swizzle 0x%x)", params[0]);
             return;
          }
-         break;
-      case GL_TEXTURE_COMPARE_MODE_ARB:
-         if (ctx->Extensions.ARB_shadow) {
-            const GLenum mode = (GLenum) params[0];
-            if (mode == GL_NONE || mode == GL_COMPARE_R_TO_TEXTURE_ARB) {
-               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-               texObj->CompareMode = mode;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM,
-                           "glTexParameter(bad GL_TEXTURE_COMPARE_MODE_ARB: 0x%x)", mode);
-               return;
-            }
-         }
-         else {
-            _mesa_error(ctx, GL_INVALID_ENUM,
-                        "glTexParameter(pname=GL_TEXTURE_COMPARE_MODE_ARB)");
+         ASSERT(comp < 4);
+         if (swz >= 0) {
+            flush(ctx, texObj);
+            texObj->Swizzle[comp] = params[0];
+            set_swizzle_component(&texObj->_Swizzle, comp, swz);
             return;
          }
-         break;
-      case GL_TEXTURE_COMPARE_FUNC_ARB:
-         if (ctx->Extensions.ARB_shadow) {
-            const GLenum func = (GLenum) params[0];
-            if (func == GL_LEQUAL || func == GL_GEQUAL) {
-               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-               texObj->CompareFunc = func;
-            }
-            else if (ctx->Extensions.EXT_shadow_funcs &&
-                     (func == GL_EQUAL ||
-                      func == GL_NOTEQUAL ||
-                      func == GL_LESS ||
-                      func == GL_GREATER ||
-                      func == GL_ALWAYS ||
-                      func == GL_NEVER)) {
-               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-               texObj->CompareFunc = func;
+      }
+      _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(pname=0x%x)", pname);
+      return;
+
+   case GL_TEXTURE_SWIZZLE_RGBA_EXT:
+      if (ctx->Extensions.EXT_texture_swizzle) {
+         GLuint comp;
+         flush(ctx, texObj);
+         for (comp = 0; comp < 4; comp++) {
+            const GLint swz = comp_to_swizzle(params[comp]);
+            if (swz >= 0) {
+               texObj->Swizzle[comp] = params[comp];
+               set_swizzle_component(&texObj->_Swizzle, comp, swz);
             }
             else {
-               _mesa_error(ctx, GL_INVALID_ENUM,
-                           "glTexParameter(bad GL_TEXTURE_COMPARE_FUNC_ARB)");
+               _mesa_error(ctx, GL_INVALID_OPERATION,
+                           "glTexParameter(swizzle 0x%x)", params[comp]);
                return;
             }
          }
-         else {
-            _mesa_error(ctx, GL_INVALID_ENUM,
-                        "glTexParameter(pname=GL_TEXTURE_COMPARE_FUNC_ARB)");
+         return;
+      }
+      _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(pname=0x%x)", pname);
+      return;
+
+   default:
+      _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(pname=0x%x)", pname);
+   }
+}
+
+
+/** Set a float-valued texture parameter */
+static void
+set_tex_parameterf(GLcontext *ctx,
+                   struct gl_texture_object *texObj,
+                   GLenum pname, const GLfloat *params)
+{
+   switch (pname) {
+   case GL_TEXTURE_MIN_LOD:
+      if (texObj->MinLod == params[0])
+         return;
+      flush(ctx, texObj);
+      texObj->MinLod = params[0];
+      return;
+
+   case GL_TEXTURE_MAX_LOD:
+      if (texObj->MaxLod == params[0])
+         return;
+      flush(ctx, texObj);
+      texObj->MaxLod = params[0];
+      return;
+
+   case GL_TEXTURE_PRIORITY:
+      flush(ctx, texObj);
+      texObj->Priority = CLAMP(params[0], 0.0F, 1.0F);
+      return;
+
+   case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+      if (ctx->Extensions.EXT_texture_filter_anisotropic) {
+         if (texObj->MaxAnisotropy == params[0])
             return;
-         }
-         break;
-      case GL_DEPTH_TEXTURE_MODE_ARB:
-         if (ctx->Extensions.ARB_depth_texture) {
-            const GLenum result = (GLenum) params[0];
-            if (result == GL_LUMINANCE || result == GL_INTENSITY
-                || result == GL_ALPHA) {
-               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-               texObj->DepthMode = result;
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_ENUM,
-                          "glTexParameter(bad GL_DEPTH_TEXTURE_MODE_ARB)");
-               return;
-            }
-         }
-         else {
-            _mesa_error(ctx, GL_INVALID_ENUM,
-                        "glTexParameter(pname=GL_DEPTH_TEXTURE_MODE_ARB)");
+         if (params[0] < 1.0) {
+            _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)" );
             return;
          }
-         break;
-      case GL_TEXTURE_LOD_BIAS:
-         /* NOTE: this is really part of OpenGL 1.4, not EXT_texture_lod_bias*/
-         if (ctx->Extensions.EXT_texture_lod_bias) {
-            if (texObj->LodBias != params[0]) {
-               FLUSH_VERTICES(ctx, _NEW_TEXTURE);
-               texObj->LodBias = params[0];
-            }
-         }
-         break;
-#ifdef FEATURE_OES_draw_texture
-      case GL_TEXTURE_CROP_RECT_OES:
-         texObj->CropRect[0] = (GLint) params[0];
-         texObj->CropRect[1] = (GLint) params[1];
-         texObj->CropRect[2] = (GLint) params[2];
-         texObj->CropRect[3] = (GLint) params[3];
-         break;
-#endif
+         flush(ctx, texObj);
+         /* clamp to max, that's what NVIDIA does */
+         texObj->MaxAnisotropy = MIN2(params[0],
+                                      ctx->Const.MaxTextureMaxAnisotropy);
+      }
+      else {
+         _mesa_error(ctx, GL_INVALID_ENUM,
+                     "glTexParameter(pname=GL_TEXTURE_MAX_ANISOTROPY_EXT)");
+      }
+      return;
 
-      default:
+   case GL_TEXTURE_COMPARE_FAIL_VALUE_ARB:
+      if (ctx->Extensions.ARB_shadow_ambient) {
+         if (texObj->CompareFailValue != params[0]) {
+            flush(ctx, texObj);
+            texObj->CompareFailValue = CLAMP(params[0], 0.0F, 1.0F);
+         }
+      }
+      else {
          _mesa_error(ctx, GL_INVALID_ENUM,
-                     "glTexParameter(pname=0x%x)", pname);
-         return;
+                    "glTexParameter(pname=GL_TEXTURE_COMPARE_FAIL_VALUE_ARB)");
+      }
+      return;
+
+   case GL_TEXTURE_LOD_BIAS:
+      /* NOTE: this is really part of OpenGL 1.4, not EXT_texture_lod_bias */
+      if (ctx->Extensions.EXT_texture_lod_bias) {
+         if (texObj->LodBias != params[0]) {
+            flush(ctx, texObj);
+            texObj->LodBias = params[0];
+         }
+      }
+      break;
+
+   case GL_TEXTURE_BORDER_COLOR:
+      flush(ctx, texObj);
+      texObj->BorderColor[RCOMP] = params[0];
+      texObj->BorderColor[GCOMP] = params[1];
+      texObj->BorderColor[BCOMP] = params[2];
+      texObj->BorderColor[ACOMP] = params[3];
+      UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[RCOMP], params[0]);
+      UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[GCOMP], params[1]);
+      UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[BCOMP], params[2]);
+      UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[ACOMP], params[3]);
+      return;
+
+   default:
+      _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(pname=0x%x)", pname);
    }
+}
 
-   texObj->_Complete = GL_FALSE;
 
-   if (ctx->Driver.TexParameter) {
-      (*ctx->Driver.TexParameter)( ctx, target, texObj, pname, params );
+void GLAPIENTRY
+_mesa_TexParameterf(GLenum target, GLenum pname, GLfloat param)
+{
+   struct gl_texture_object *texObj;
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   texObj = get_texobj(ctx, target);
+   if (!texObj)
+      return;
+
+   switch (pname) {
+   case GL_TEXTURE_MIN_FILTER:
+   case GL_TEXTURE_MAG_FILTER:
+   case GL_TEXTURE_WRAP_S:
+   case GL_TEXTURE_WRAP_T:
+   case GL_TEXTURE_WRAP_R:
+   case GL_TEXTURE_BASE_LEVEL:
+   case GL_TEXTURE_MAX_LEVEL:
+   case GL_GENERATE_MIPMAP_SGIS:
+   case GL_TEXTURE_COMPARE_MODE_ARB:
+   case GL_TEXTURE_COMPARE_FUNC_ARB:
+   case GL_DEPTH_TEXTURE_MODE_ARB:
+      {
+         /* convert float param to int */
+         GLint p = (GLint) param;
+         set_tex_parameteri(ctx, texObj, pname, &p);
+      }
+      return;
+   default:
+      /* this will generate an error if pname is illegal */
+      set_tex_parameterf(ctx, texObj, pname, &param);
+   }
+
+   if (ctx->Driver.TexParameter && ctx->ErrorValue == GL_NO_ERROR) {
+      ctx->Driver.TexParameter(ctx, target, texObj, pname, &param);
    }
 }
 
 
 void GLAPIENTRY
-_mesa_TexParameteri( GLenum target, GLenum pname, GLint param )
+_mesa_TexParameterfv(GLenum target, GLenum pname, const GLfloat *params)
 {
-   GLfloat fparam[4];
-   if (pname == GL_TEXTURE_PRIORITY)
-      fparam[0] = INT_TO_FLOAT(param);
-   else
-      fparam[0] = (GLfloat) param;
-   fparam[1] = fparam[2] = fparam[3] = 0.0;
-   _mesa_TexParameterfv(target, pname, fparam);
+   struct gl_texture_object *texObj;
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   texObj = get_texobj(ctx, target);
+   if (!texObj)
+      return;
+
+   switch (pname) {
+   case GL_TEXTURE_MIN_FILTER:
+   case GL_TEXTURE_MAG_FILTER:
+   case GL_TEXTURE_WRAP_S:
+   case GL_TEXTURE_WRAP_T:
+   case GL_TEXTURE_WRAP_R:
+   case GL_TEXTURE_BASE_LEVEL:
+   case GL_TEXTURE_MAX_LEVEL:
+   case GL_GENERATE_MIPMAP_SGIS:
+   case GL_TEXTURE_COMPARE_MODE_ARB:
+   case GL_TEXTURE_COMPARE_FUNC_ARB:
+   case GL_DEPTH_TEXTURE_MODE_ARB:
+      {
+         /* convert float param to int */
+         GLint p = (GLint) params[0];
+         set_tex_parameteri(ctx, texObj, pname, &p);
+      }
+      break;
+
+#ifdef FEATURE_OES_draw_texture
+   case GL_TEXTURE_CROP_RECT_OES:
+      {
+         /* convert float params to int */
+         GLint iparams[4];
+         iparams[0] = (GLint) params[0];
+         iparams[1] = (GLint) params[1];
+         iparams[2] = (GLint) params[2];
+         iparams[3] = (GLint) params[3];
+         set_tex_parameteri(ctx, target, iparams);
+      }
+      break;
+#endif
+
+   default:
+      /* this will generate an error if pname is illegal */
+      set_tex_parameterf(ctx, texObj, pname, params);
+   }
+
+   if (ctx->Driver.TexParameter && ctx->ErrorValue == GL_NO_ERROR) {
+      ctx->Driver.TexParameter(ctx, target, texObj, pname, params);
+   }
 }
 
 
 void GLAPIENTRY
-_mesa_TexParameteriv( GLenum target, GLenum pname, const GLint *params )
+_mesa_TexParameteri(GLenum target, GLenum pname, GLint param)
 {
-   GLfloat fparam[4];
-   if (pname == GL_TEXTURE_BORDER_COLOR) {
-      fparam[0] = INT_TO_FLOAT(params[0]);
-      fparam[1] = INT_TO_FLOAT(params[1]);
-      fparam[2] = INT_TO_FLOAT(params[2]);
-      fparam[3] = INT_TO_FLOAT(params[3]);
+   struct gl_texture_object *texObj;
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   texObj = get_texobj(ctx, target);
+   if (!texObj)
+      return;
+
+   switch (pname) {
+   case GL_TEXTURE_MIN_LOD:
+   case GL_TEXTURE_MAX_LOD:
+   case GL_TEXTURE_PRIORITY:
+   case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+   case GL_TEXTURE_LOD_BIAS:
+   case GL_TEXTURE_COMPARE_FAIL_VALUE_ARB:
+      {
+         GLfloat fparam = (GLfloat) param;
+         /* convert int param to float */
+         set_tex_parameterf(ctx, texObj, pname, &fparam);
+      }
+      break;
+   default:
+      /* this will generate an error if pname is illegal */
+      set_tex_parameteri(ctx, texObj, pname, &param);
    }
-   else if (pname == GL_TEXTURE_CROP_RECT_OES) {
-      fparam[0] = (GLfloat) params[0];
-      fparam[1] = (GLfloat) params[1];
-      fparam[2] = (GLfloat) params[2];
-      fparam[3] = (GLfloat) params[3];
+
+   if (ctx->Driver.TexParameter && ctx->ErrorValue == GL_NO_ERROR) {
+      GLfloat fparam = (GLfloat) param;
+      ctx->Driver.TexParameter(ctx, target, texObj, pname, &fparam);
    }
-   else {
-      if (pname == GL_TEXTURE_PRIORITY)
-         fparam[0] = INT_TO_FLOAT(params[0]);
-      else
-         fparam[0] = (GLfloat) params[0];
-      fparam[1] = fparam[2] = fparam[3] = 0.0F;
+}
+
+
+void GLAPIENTRY
+_mesa_TexParameteriv(GLenum target, GLenum pname, const GLint *params)
+{
+   struct gl_texture_object *texObj;
+   GET_CURRENT_CONTEXT(ctx);
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
+
+   texObj = get_texobj(ctx, target);
+   if (!texObj)
+      return;
+
+   switch (pname) {
+   case GL_TEXTURE_BORDER_COLOR:
+      {
+         /* convert int params to float */
+         GLfloat fparams[4];
+         fparams[0] = INT_TO_FLOAT(params[0]);
+         fparams[1] = INT_TO_FLOAT(params[1]);
+         fparams[2] = INT_TO_FLOAT(params[2]);
+         fparams[3] = INT_TO_FLOAT(params[3]);
+         set_tex_parameterf(ctx, texObj, pname, fparams);
+      }
+      break;
+   case GL_TEXTURE_MIN_LOD:
+   case GL_TEXTURE_MAX_LOD:
+   case GL_TEXTURE_PRIORITY:
+   case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+   case GL_TEXTURE_LOD_BIAS:
+   case GL_TEXTURE_COMPARE_FAIL_VALUE_ARB:
+      {
+         /* convert int param to float */
+         GLfloat fparam = (GLfloat) params[0];
+         set_tex_parameterf(ctx, texObj, pname, &fparam);
+      }
+      break;
+   default:
+      /* this will generate an error if pname is illegal */
+      set_tex_parameteri(ctx, texObj, pname, params);
+   }
+
+   if (ctx->Driver.TexParameter && ctx->ErrorValue == GL_NO_ERROR) {
+      GLfloat fparams[4];
+      fparams[0] = INT_TO_FLOAT(params[0]);
+      if (pname == GL_TEXTURE_BORDER_COLOR ||
+          pname == GL_TEXTURE_CROP_RECT_OES) {
+         fparams[1] = INT_TO_FLOAT(params[1]);
+         fparams[2] = INT_TO_FLOAT(params[2]);
+         fparams[3] = INT_TO_FLOAT(params[3]);
+      }
+      ctx->Driver.TexParameter(ctx, target, texObj, pname, fparams);
    }
-   _mesa_TexParameterfv(target, pname, fparam);
 }
 
 
@@ -660,33 +867,21 @@ _mesa_GetTexLevelParameteriv( GLenum target, GLint level,
 
       /* GL_ARB_texture_compression */
       case GL_TEXTURE_COMPRESSED_IMAGE_SIZE:
-         if (ctx->Extensions.ARB_texture_compression) {
-            if (img->IsCompressed && !isProxy) {
-               /* Don't use ctx->Driver.CompressedTextureSize() since that
-                * may returned a padded hardware size.
-                */
-               *params = _mesa_compressed_texture_size(ctx, img->Width,
-                                                   img->Height, img->Depth,
-                                                   img->TexFormat->MesaFormat);
-            }
-            else {
-               _mesa_error(ctx, GL_INVALID_OPERATION,
-                           "glGetTexLevelParameter[if]v(pname)");
-            }
-         }
-         else {
-            _mesa_error(ctx, GL_INVALID_ENUM,
-                        "glGetTexLevelParameter[if]v(pname)");
-         }
+	 if (img->IsCompressed && !isProxy) {
+	    /* Don't use ctx->Driver.CompressedTextureSize() since that
+	     * may returned a padded hardware size.
+	     */
+	    *params = _mesa_compressed_texture_size(ctx, img->Width,
+						    img->Height, img->Depth,
+						    img->TexFormat->MesaFormat);
+	 }
+	 else {
+	    _mesa_error(ctx, GL_INVALID_OPERATION,
+			"glGetTexLevelParameter[if]v(pname)");
+	 }
          break;
       case GL_TEXTURE_COMPRESSED:
-         if (ctx->Extensions.ARB_texture_compression) {
-            *params = (GLint) img->IsCompressed;
-         }
-         else {
-            _mesa_error(ctx, GL_INVALID_ENUM,
-                        "glGetTexLevelParameter[if]v(pname)");
-         }
+         *params = (GLint) img->IsCompressed;
          break;
 
       /* GL_ARB_texture_float */
@@ -841,63 +1036,49 @@ _mesa_GetTexParameterfv( GLenum target, GLenum pname, GLfloat *params )
             *params = obj->MaxAnisotropy;
          }
 	 else
-	    error = 1;
-         break;
-      case GL_TEXTURE_COMPARE_SGIX:
-         if (ctx->Extensions.SGIX_shadow) {
-            *params = (GLfloat) obj->CompareFlag;
-         }
-	 else 
-	    error = 1;
-         break;
-      case GL_TEXTURE_COMPARE_OPERATOR_SGIX:
-         if (ctx->Extensions.SGIX_shadow) {
-            *params = (GLfloat) obj->CompareOperator;
-         }
-	 else 
-	    error = 1;
+	    error = GL_TRUE;
          break;
-      case GL_SHADOW_AMBIENT_SGIX: /* aka GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
-         if (ctx->Extensions.SGIX_shadow_ambient) {
-            *params = obj->ShadowAmbient;
+      case GL_TEXTURE_COMPARE_FAIL_VALUE_ARB:
+         if (ctx->Extensions.ARB_shadow_ambient) {
+            *params = obj->CompareFailValue;
          }
 	 else 
-	    error = 1;
+	    error = GL_TRUE;
          break;
       case GL_GENERATE_MIPMAP_SGIS:
          if (ctx->Extensions.SGIS_generate_mipmap) {
             *params = (GLfloat) obj->GenerateMipmap;
          }
 	 else 
-	    error = 1;
+	    error = GL_TRUE;
          break;
       case GL_TEXTURE_COMPARE_MODE_ARB:
          if (ctx->Extensions.ARB_shadow) {
             *params = (GLfloat) obj->CompareMode;
          }
 	 else 
-	    error = 1;
+	    error = GL_TRUE;
          break;
       case GL_TEXTURE_COMPARE_FUNC_ARB:
          if (ctx->Extensions.ARB_shadow) {
             *params = (GLfloat) obj->CompareFunc;
          }
 	 else 
-	    error = 1;
+	    error = GL_TRUE;
          break;
       case GL_DEPTH_TEXTURE_MODE_ARB:
          if (ctx->Extensions.ARB_depth_texture) {
             *params = (GLfloat) obj->DepthMode;
          }
 	 else 
-	    error = 1;
+	    error = GL_TRUE;
          break;
       case GL_TEXTURE_LOD_BIAS:
          if (ctx->Extensions.EXT_texture_lod_bias) {
             *params = obj->LodBias;
          }
 	 else 
-	    error = 1;
+	    error = GL_TRUE;
          break;
 #ifdef FEATURE_OES_draw_texture
       case GL_TEXTURE_CROP_RECT_OES:
@@ -907,10 +1088,37 @@ _mesa_GetTexParameterfv( GLenum target, GLenum pname, GLfloat *params )
          params[3] = obj->CropRect[3];
          break;
 #endif
+
+      case GL_TEXTURE_SWIZZLE_R_EXT:
+      case GL_TEXTURE_SWIZZLE_G_EXT:
+      case GL_TEXTURE_SWIZZLE_B_EXT:
+      case GL_TEXTURE_SWIZZLE_A_EXT:
+         if (ctx->Extensions.EXT_texture_swizzle) {
+            GLuint comp = pname - GL_TEXTURE_SWIZZLE_R_EXT;
+            *params = (GLfloat) obj->Swizzle[comp];
+         }
+         else {
+            error = GL_TRUE;
+         }
+         break;
+
+      case GL_TEXTURE_SWIZZLE_RGBA_EXT:
+         if (ctx->Extensions.EXT_texture_swizzle) {
+            GLuint comp;
+            for (comp = 0; comp < 4; comp++) {
+               params[comp] = (GLfloat) obj->Swizzle[comp];
+            }
+         }
+         else {
+            error = GL_TRUE;
+         }
+         break;
+
       default:
-	 error = 1;
+	 error = GL_TRUE;
 	 break;
    }
+
    if (error)
       _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameterfv(pname=0x%x)",
 		  pname);
@@ -924,6 +1132,7 @@ _mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params )
 {
    struct gl_texture_unit *texUnit;
    struct gl_texture_object *obj;
+   GLboolean error = GL_FALSE;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
@@ -944,19 +1153,19 @@ _mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params )
    switch (pname) {
       case GL_TEXTURE_MAG_FILTER:
          *params = (GLint) obj->MagFilter;
-         return;
+         break;;
       case GL_TEXTURE_MIN_FILTER:
          *params = (GLint) obj->MinFilter;
-         return;
+         break;;
       case GL_TEXTURE_WRAP_S:
          *params = (GLint) obj->WrapS;
-         return;
+         break;;
       case GL_TEXTURE_WRAP_T:
          *params = (GLint) obj->WrapT;
-         return;
+         break;;
       case GL_TEXTURE_WRAP_R:
          *params = (GLint) obj->WrapR;
-         return;
+         break;;
       case GL_TEXTURE_BORDER_COLOR:
          {
             GLfloat b[4];
@@ -969,7 +1178,7 @@ _mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params )
             params[2] = FLOAT_TO_INT(b[2]);
             params[3] = FLOAT_TO_INT(b[3]);
          }
-         return;
+         break;;
       case GL_TEXTURE_RESIDENT:
          {
             GLboolean resident;
@@ -979,74 +1188,76 @@ _mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params )
                resident = GL_TRUE;
             *params = (GLint) resident;
          }
-         return;
+         break;;
       case GL_TEXTURE_PRIORITY:
          *params = FLOAT_TO_INT(obj->Priority);
-         return;
+         break;;
       case GL_TEXTURE_MIN_LOD:
          *params = (GLint) obj->MinLod;
-         return;
+         break;;
       case GL_TEXTURE_MAX_LOD:
          *params = (GLint) obj->MaxLod;
-         return;
+         break;;
       case GL_TEXTURE_BASE_LEVEL:
          *params = obj->BaseLevel;
-         return;
+         break;;
       case GL_TEXTURE_MAX_LEVEL:
          *params = obj->MaxLevel;
-         return;
+         break;;
       case GL_TEXTURE_MAX_ANISOTROPY_EXT:
          if (ctx->Extensions.EXT_texture_filter_anisotropic) {
             *params = (GLint) obj->MaxAnisotropy;
-            return;
          }
-         break;
-      case GL_TEXTURE_COMPARE_SGIX:
-         if (ctx->Extensions.SGIX_shadow) {
-            *params = (GLint) obj->CompareFlag;
-            return;
+         else {
+            error = GL_TRUE;
          }
          break;
-      case GL_TEXTURE_COMPARE_OPERATOR_SGIX:
-         if (ctx->Extensions.SGIX_shadow) {
-            *params = (GLint) obj->CompareOperator;
-            return;
+      case GL_TEXTURE_COMPARE_FAIL_VALUE_ARB:
+         if (ctx->Extensions.ARB_shadow_ambient) {
+            *params = (GLint) FLOAT_TO_INT(obj->CompareFailValue);
          }
-         break;
-      case GL_SHADOW_AMBIENT_SGIX: /* aka GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */
-         if (ctx->Extensions.SGIX_shadow_ambient) {
-            *params = (GLint) FLOAT_TO_INT(obj->ShadowAmbient);
-            return;
+         else {
+            error = GL_TRUE;
          }
          break;
       case GL_GENERATE_MIPMAP_SGIS:
          if (ctx->Extensions.SGIS_generate_mipmap) {
             *params = (GLint) obj->GenerateMipmap;
-            return;
+         }
+         else {
+            error = GL_TRUE;
          }
          break;
       case GL_TEXTURE_COMPARE_MODE_ARB:
          if (ctx->Extensions.ARB_shadow) {
             *params = (GLint) obj->CompareMode;
-            return;
+         }
+         else {
+            error = GL_TRUE;
          }
          break;
       case GL_TEXTURE_COMPARE_FUNC_ARB:
          if (ctx->Extensions.ARB_shadow) {
             *params = (GLint) obj->CompareFunc;
-            return;
+         }
+         else {
+            error = GL_TRUE;
          }
          break;
       case GL_DEPTH_TEXTURE_MODE_ARB:
          if (ctx->Extensions.ARB_depth_texture) {
             *params = (GLint) obj->DepthMode;
-            return;
+         }
+         else {
+            error = GL_TRUE;
          }
          break;
       case GL_TEXTURE_LOD_BIAS:
          if (ctx->Extensions.EXT_texture_lod_bias) {
             *params = (GLint) obj->LodBias;
-            return;
+         }
+         else {
+            error = GL_TRUE;
          }
          break;
 #ifdef FEATURE_OES_draw_texture
@@ -1057,9 +1268,35 @@ _mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params )
          params[3] = obj->CropRect[3];
          break;
 #endif
+      case GL_TEXTURE_SWIZZLE_R_EXT:
+      case GL_TEXTURE_SWIZZLE_G_EXT:
+      case GL_TEXTURE_SWIZZLE_B_EXT:
+      case GL_TEXTURE_SWIZZLE_A_EXT:
+         if (ctx->Extensions.EXT_texture_swizzle) {
+            GLuint comp = pname - GL_TEXTURE_SWIZZLE_R_EXT;
+            *params = obj->Swizzle[comp];
+         }
+         else {
+            error = GL_TRUE;
+         }
+         break;
+
+      case GL_TEXTURE_SWIZZLE_RGBA_EXT:
+         if (ctx->Extensions.EXT_texture_swizzle) {
+            COPY_4V(params, obj->Swizzle);
+         }
+         else {
+            error = GL_TRUE;
+         }
+         break;
+
       default:
          ; /* silence warnings */
    }
-   /* If we get here, pname was an unrecognized enum */
-   _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameteriv(pname=0x%x)", pname);
+
+   if (error)
+      _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameteriv(pname=0x%x)",
+		  pname);
+
+   _mesa_unlock_texture(ctx, obj);
 }
diff --git a/src/mesa/main/texrender.c b/src/mesa/main/texrender.c
index 163bda4501..49de6f5b8a 100644
--- a/src/mesa/main/texrender.c
+++ b/src/mesa/main/texrender.c
@@ -49,6 +49,14 @@ texture_get_row(GLcontext *ctx, struct gl_renderbuffer *rb, GLuint count,
          trb->TexImage->FetchTexelc(trb->TexImage, x + i, y, z, rgbaOut + 4 * i);
       }
    }
+   else if (rb->DataType == GL_UNSIGNED_SHORT) {
+      GLushort *zValues = (GLushort *) values;
+      for (i = 0; i < count; i++) {
+         GLfloat flt;
+         trb->TexImage->FetchTexelf(trb->TexImage, x + i, y, z, &flt);
+         zValues[i] = (GLushort) (flt * 0xffff);
+      }
+   }
    else if (rb->DataType == GL_UNSIGNED_INT) {
       GLuint *zValues = (GLuint *) values;
       /*
@@ -96,6 +104,15 @@ texture_get_values(GLcontext *ctx, struct gl_renderbuffer *rb, GLuint count,
 				    z, rgbaOut + 4 * i);
       }
    }
+   else if (rb->DataType == GL_UNSIGNED_SHORT) {
+      GLushort *zValues = (GLushort *) values;
+      for (i = 0; i < count; i++) {
+         GLfloat flt;
+         trb->TexImage->FetchTexelf(trb->TexImage, x[i], y[i] + trb->Yoffset,
+				    z, &flt);
+         zValues[i] = (GLushort) (flt * 0xffff);
+      }
+   }
    else if (rb->DataType == GL_UNSIGNED_INT) {
       GLuint *zValues = (GLuint *) values;
       for (i = 0; i < count; i++) {
@@ -147,6 +164,67 @@ texture_put_row(GLcontext *ctx, struct gl_renderbuffer *rb, GLuint count,
          rgba += 4;
       }
    }
+   else if (rb->DataType == GL_UNSIGNED_SHORT) {
+      const GLushort *zValues = (const GLushort *) values;
+      for (i = 0; i < count; i++) {
+         if (!mask || mask[i]) {
+            trb->Store(trb->TexImage, x + i, y, z, zValues + i);
+         }
+      }
+   }
+   else if (rb->DataType == GL_UNSIGNED_INT) {
+      const GLuint *zValues = (const GLuint *) values;
+      for (i = 0; i < count; i++) {
+         if (!mask || mask[i]) {
+            trb->Store(trb->TexImage, x + i, y, z, zValues + i);
+         }
+      }
+   }
+   else if (rb->DataType == GL_UNSIGNED_INT_24_8_EXT) {
+      const GLuint *zValues = (const GLuint *) values;
+      for (i = 0; i < count; i++) {
+         if (!mask || mask[i]) {
+            GLfloat flt = (GLfloat) ((zValues[i] >> 8) * (1.0 / 0xffffff));
+            trb->Store(trb->TexImage, x + i, y, z, &flt);
+         }
+      }
+   }
+   else {
+      _mesa_problem(ctx, "invalid rb->DataType in texture_put_row");
+   }
+}
+
+/**
+ * Put row of RGB values into a renderbuffer that wraps a texture image.
+ */
+static void
+texture_put_row_rgb(GLcontext *ctx, struct gl_renderbuffer *rb, GLuint count,
+                GLint x, GLint y, const void *values, const GLubyte *mask)
+{
+   const struct texture_renderbuffer *trb
+      = (const struct texture_renderbuffer *) rb;
+   const GLint z = trb->Zoffset;
+   GLuint i;
+
+   y += trb->Yoffset;
+
+   if (rb->DataType == CHAN_TYPE) {
+      const GLchan *rgb = (const GLchan *) values;
+      for (i = 0; i < count; i++) {
+         if (!mask || mask[i]) {
+            trb->Store(trb->TexImage, x + i, y, z, rgb);
+         }
+         rgb += 3;
+      }
+   }
+   else if (rb->DataType == GL_UNSIGNED_SHORT) {
+      const GLushort *zValues = (const GLushort *) values;
+      for (i = 0; i < count; i++) {
+         if (!mask || mask[i]) {
+            trb->Store(trb->TexImage, x + i, y, z, zValues + i);
+         }
+      }
+   }
    else if (rb->DataType == GL_UNSIGNED_INT) {
       const GLuint *zValues = (const GLuint *) values;
       for (i = 0; i < count; i++) {
@@ -189,6 +267,14 @@ texture_put_mono_row(GLcontext *ctx, struct gl_renderbuffer *rb, GLuint count,
          }
       }
    }
+   else if (rb->DataType == GL_UNSIGNED_SHORT) {
+      const GLushort zValue = *((const GLushort *) value);
+      for (i = 0; i < count; i++) {
+         if (!mask || mask[i]) {
+            trb->Store(trb->TexImage, x + i, y, z, &zValue);
+         }
+      }
+   }
    else if (rb->DataType == GL_UNSIGNED_INT) {
       const GLuint zValue = *((const GLuint *) value);
       for (i = 0; i < count; i++) {
@@ -231,12 +317,19 @@ texture_put_values(GLcontext *ctx, struct gl_renderbuffer *rb, GLuint count,
          rgba += 4;
       }
    }
+   else if (rb->DataType == GL_UNSIGNED_SHORT) {
+      const GLushort *zValues = (const GLushort *) values;
+      for (i = 0; i < count; i++) {
+         if (!mask || mask[i]) {
+            trb->Store(trb->TexImage, x[i], y[i] + trb->Yoffset, z, zValues + i);
+         }
+      }
+   }
    else if (rb->DataType == GL_UNSIGNED_INT) {
       const GLuint *zValues = (const GLuint *) values;
       for (i = 0; i < count; i++) {
          if (!mask || mask[i]) {
-            trb->Store(trb->TexImage, x[i], y[i] + trb->Yoffset, z,
-		       zValues + i);
+            trb->Store(trb->TexImage, x[i], y[i] + trb->Yoffset, z, zValues + i);
          }
       }
    }
@@ -281,6 +374,14 @@ texture_put_mono_values(GLcontext *ctx, struct gl_renderbuffer *rb,
          }
       }
    }
+   else if (rb->DataType == GL_UNSIGNED_SHORT) {
+      const GLushort zValue = *((const GLushort *) value);
+      for (i = 0; i < count; i++) {
+         if (!mask || mask[i]) {
+            trb->Store(trb->TexImage, x[i], y[i] + trb->Yoffset, z, &zValue);
+         }
+      }
+   }
    else if (rb->DataType == GL_UNSIGNED_INT_24_8_EXT) {
       const GLuint zValue = *((const GLuint *) value);
       const GLfloat flt = (GLfloat) ((zValue >> 8) * (1.0 / 0xffffff));
@@ -332,6 +433,7 @@ wrap_texture(GLcontext *ctx, struct gl_renderbuffer_attachment *att)
    trb->Base.GetRow = texture_get_row;
    trb->Base.GetValues = texture_get_values;
    trb->Base.PutRow = texture_put_row;
+   trb->Base.PutRowRGB = texture_put_row_rgb;
    trb->Base.PutMonoRow = texture_put_mono_row;
    trb->Base.PutValues = texture_put_values;
    trb->Base.PutMonoValues = texture_put_mono_values;
diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c
index f019377041..7761af7589 100644
--- a/src/mesa/main/texstate.c
+++ b/src/mesa/main/texstate.c
@@ -53,10 +53,10 @@
  */
 static const struct gl_tex_env_combine_state default_combine_state = {
    GL_MODULATE, GL_MODULATE,
-   { GL_TEXTURE, GL_PREVIOUS, GL_CONSTANT },
-   { GL_TEXTURE, GL_PREVIOUS, GL_CONSTANT },
-   { GL_SRC_COLOR, GL_SRC_COLOR, GL_SRC_ALPHA },
-   { GL_SRC_ALPHA, GL_SRC_ALPHA, GL_SRC_ALPHA },
+   { GL_TEXTURE, GL_PREVIOUS, GL_CONSTANT, GL_CONSTANT },
+   { GL_TEXTURE, GL_PREVIOUS, GL_CONSTANT, GL_CONSTANT },
+   { GL_SRC_COLOR, GL_SRC_COLOR, GL_SRC_ALPHA, GL_SRC_ALPHA },
+   { GL_SRC_ALPHA, GL_SRC_ALPHA, GL_SRC_ALPHA, GL_SRC_ALPHA },
    0, 0,
    2, 2
 };
@@ -81,7 +81,7 @@ _mesa_copy_texture_state( const GLcontext *src, GLcontext *dst )
    dst->Texture.SharedPalette = src->Texture.SharedPalette;
 
    /* per-unit state */
-   for (i = 0; i < src->Const.MaxTextureUnits; i++) {
+   for (i = 0; i < src->Const.MaxTextureImageUnits; i++) {
       dst->Texture.Unit[i].Enabled = src->Texture.Unit[i].Enabled;
       dst->Texture.Unit[i].EnvMode = src->Texture.Unit[i].EnvMode;
       COPY_4V(dst->Texture.Unit[i].EnvColor, src->Texture.Unit[i].EnvColor);
@@ -307,8 +307,7 @@ _mesa_ActiveTextureARB(GLenum texture)
       _mesa_debug(ctx, "glActiveTexture %s\n",
                   _mesa_lookup_enum_by_nr(texture));
 
-   /* XXX error-check against max(coordunits, imageunits) */
-   if (texUnit >= ctx->Const.MaxTextureUnits) {
+   if (texUnit >= ctx->Const.MaxTextureImageUnits) {
       _mesa_error(ctx, GL_INVALID_ENUM, "glActiveTexture(texture)");
       return;
    }
@@ -369,7 +368,7 @@ update_texture_matrices( GLcontext *ctx )
 
    ctx->Texture._TexMatEnabled = 0;
 
-   for (i=0; i < ctx->Const.MaxTextureUnits; i++) {
+   for (i=0; i < ctx->Const.MaxTextureCoordUnits; i++) {
       if (_math_matrix_is_dirty(ctx->TextureMatrixStack[i].Top)) {
 	 _math_matrix_analyse( ctx->TextureMatrixStack[i].Top );
 
@@ -405,16 +404,6 @@ update_texture_compare_function(GLcontext *ctx,
        */
       tObj->_Function = GL_NONE;
    }
-   else if (tObj->CompareFlag) {
-      /* GL_SGIX_shadow */
-      if (tObj->CompareOperator == GL_TEXTURE_LEQUAL_R_SGIX) {
-         tObj->_Function = GL_LEQUAL;
-      }
-      else {
-         ASSERT(tObj->CompareOperator == GL_TEXTURE_GEQUAL_R_SGIX);
-         tObj->_Function = GL_GEQUAL;
-      }
-   }
    else if (tObj->CompareMode == GL_COMPARE_R_TO_TEXTURE_ARB) {
       /* GL_ARB_shadow */
       tObj->_Function = tObj->CompareFunc;
@@ -448,6 +437,96 @@ texture_override(GLcontext *ctx,
 
 
 /**
+ * Examine texture unit's combine/env state to update derived state.
+ */
+static void
+update_tex_combine(GLcontext *ctx, struct gl_texture_unit *texUnit)
+{
+   /* Set the texUnit->_CurrentCombine field to point to the user's combiner
+    * state, or the combiner state which is derived from traditional texenv
+    * mode.
+    */
+   if (texUnit->EnvMode == GL_COMBINE ||
+       texUnit->EnvMode == GL_COMBINE4_NV) {
+      texUnit->_CurrentCombine = & texUnit->Combine;
+   }
+   else {
+      const struct gl_texture_object *texObj = texUnit->_Current;
+      GLenum format = texObj->Image[0][texObj->BaseLevel]->_BaseFormat;
+      if (format == GL_COLOR_INDEX) {
+         format = GL_RGBA;  /* a bit of a hack */
+      }
+      else if (format == GL_DEPTH_COMPONENT ||
+               format == GL_DEPTH_STENCIL_EXT) {
+         format = texObj->DepthMode;
+      }
+      calculate_derived_texenv(&texUnit->_EnvMode, texUnit->EnvMode, format);
+      texUnit->_CurrentCombine = & texUnit->_EnvMode;
+   }
+
+   /* Determine number of source RGB terms in the combiner function */
+   switch (texUnit->_CurrentCombine->ModeRGB) {
+   case GL_REPLACE:
+      texUnit->_CurrentCombine->_NumArgsRGB = 1;
+      break;
+   case GL_ADD:
+   case GL_ADD_SIGNED:
+      if (texUnit->EnvMode == GL_COMBINE4_NV)
+         texUnit->_CurrentCombine->_NumArgsRGB = 4;
+      else
+         texUnit->_CurrentCombine->_NumArgsRGB = 2;
+      break;
+   case GL_MODULATE:
+   case GL_SUBTRACT:
+   case GL_DOT3_RGB:
+   case GL_DOT3_RGBA:
+   case GL_DOT3_RGB_EXT:
+   case GL_DOT3_RGBA_EXT:
+      texUnit->_CurrentCombine->_NumArgsRGB = 2;
+      break;
+   case GL_INTERPOLATE:
+   case GL_MODULATE_ADD_ATI:
+   case GL_MODULATE_SIGNED_ADD_ATI:
+   case GL_MODULATE_SUBTRACT_ATI:
+      texUnit->_CurrentCombine->_NumArgsRGB = 3;
+      break;
+   default:
+      texUnit->_CurrentCombine->_NumArgsRGB = 0;
+      _mesa_problem(ctx, "invalid RGB combine mode in update_texture_state");
+      return;
+   }
+
+   /* Determine number of source Alpha terms in the combiner function */
+   switch (texUnit->_CurrentCombine->ModeA) {
+   case GL_REPLACE:
+      texUnit->_CurrentCombine->_NumArgsA = 1;
+      break;
+   case GL_ADD:
+   case GL_ADD_SIGNED:
+      if (texUnit->EnvMode == GL_COMBINE4_NV)
+         texUnit->_CurrentCombine->_NumArgsA = 4;
+      else
+         texUnit->_CurrentCombine->_NumArgsA = 2;
+      break;
+   case GL_MODULATE:
+   case GL_SUBTRACT:
+      texUnit->_CurrentCombine->_NumArgsA = 2;
+      break;
+   case GL_INTERPOLATE:
+   case GL_MODULATE_ADD_ATI:
+   case GL_MODULATE_SIGNED_ADD_ATI:
+   case GL_MODULATE_SUBTRACT_ATI:
+      texUnit->_CurrentCombine->_NumArgsA = 3;
+      break;
+   default:
+      texUnit->_CurrentCombine->_NumArgsA = 0;
+      _mesa_problem(ctx, "invalid Alpha combine mode in update_texture_state");
+      break;
+   }
+}
+
+
+/**
  * \note This routine refers to derived texture matrix values to
  * compute the ENABLE_TEXMAT flags, but is only called on
  * _NEW_TEXTURE.  On changes to _NEW_TEXTURE_MATRIX, the ENABLE_TEXMAT
@@ -479,9 +558,8 @@ update_texture_state( GLcontext *ctx )
       }
    }
 
-   ctx->NewState |= _NEW_TEXTURE; /* TODO: only set this if there are 
-				   * actual changes. 
-				   */
+   /* TODO: only set this if there are actual changes */
+   ctx->NewState |= _NEW_TEXTURE;
 
    ctx->Texture._EnabledUnits = 0;
    ctx->Texture._GenFlags = 0;
@@ -491,7 +569,7 @@ update_texture_state( GLcontext *ctx )
    /*
     * Update texture unit state.
     */
-   for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
+   for (unit = 0; unit < ctx->Const.MaxTextureImageUnits; unit++) {
       struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
       GLbitfield enableBits;
 
@@ -499,25 +577,35 @@ update_texture_state( GLcontext *ctx )
       texUnit->_ReallyEnabled = 0;
       texUnit->_GenFlags = 0;
 
-      /* Get the bitmask of texture enables.
+      /* Get the bitmask of texture target enables.
        * enableBits will be a mask of the TEXTURE_*_BIT flags indicating
        * which texture targets are enabled (fixed function) or referenced
        * by a fragment shader/program.  When multiple flags are set, we'll
        * settle on the one with highest priority (see texture_override below).
        */
-      if (fprog || vprog) {
-         enableBits = 0x0;
-         if (fprog)
-            enableBits |= fprog->Base.TexturesUsed[unit];
-         if (vprog)
-            enableBits |= vprog->Base.TexturesUsed[unit];
+      enableBits = 0x0;
+      if (vprog) {
+         enableBits |= vprog->Base.TexturesUsed[unit];
+      }
+      if (fprog) {
+         enableBits |= fprog->Base.TexturesUsed[unit];
       }
       else {
-         if (!texUnit->Enabled)
-            continue;
-         enableBits = texUnit->Enabled;
+         /* fixed-function fragment program */
+         enableBits |= texUnit->Enabled;
       }
 
+      if (enableBits == 0x0)
+         continue;
+
+      ASSERT(texUnit->Current1D);
+      ASSERT(texUnit->Current2D);
+      ASSERT(texUnit->Current3D);
+      ASSERT(texUnit->CurrentCubeMap);
+      ASSERT(texUnit->CurrentRect);
+      ASSERT(texUnit->Current1DArray);
+      ASSERT(texUnit->Current2DArray);
+
       /* Look for the highest-priority texture target that's enabled and
        * complete.  That's the one we'll use for texturing.  If we're using
        * a fragment program we're guaranteed that bitcount(enabledBits) <= 1.
@@ -541,75 +629,14 @@ update_texture_state( GLcontext *ctx )
          continue;
       }
 
-      if (texUnit->_ReallyEnabled)
-         ctx->Texture._EnabledUnits |= (1 << unit);
+      /* if we get here, we know this texture unit is enabled */
 
-      if (texUnit->EnvMode == GL_COMBINE) {
-	 texUnit->_CurrentCombine = & texUnit->Combine;
-      }
-      else {
-         const struct gl_texture_object *texObj = texUnit->_Current;
-         GLenum format = texObj->Image[0][texObj->BaseLevel]->_BaseFormat;
-         if (format == GL_COLOR_INDEX) {
-            format = GL_RGBA;  /* a bit of a hack */
-         }
-         else if (format == GL_DEPTH_COMPONENT
-                  || format == GL_DEPTH_STENCIL_EXT) {
-            format = texObj->DepthMode;
-         }
-	 calculate_derived_texenv(&texUnit->_EnvMode, texUnit->EnvMode, format);
-	 texUnit->_CurrentCombine = & texUnit->_EnvMode;
-      }
-
-      switch (texUnit->_CurrentCombine->ModeRGB) {
-      case GL_REPLACE:
-	 texUnit->_CurrentCombine->_NumArgsRGB = 1;
-	 break;
-      case GL_MODULATE:
-      case GL_ADD:
-      case GL_ADD_SIGNED:
-      case GL_SUBTRACT:
-      case GL_DOT3_RGB:
-      case GL_DOT3_RGBA:
-      case GL_DOT3_RGB_EXT:
-      case GL_DOT3_RGBA_EXT:
-	 texUnit->_CurrentCombine->_NumArgsRGB = 2;
-	 break;
-      case GL_INTERPOLATE:
-      case GL_MODULATE_ADD_ATI:
-      case GL_MODULATE_SIGNED_ADD_ATI:
-      case GL_MODULATE_SUBTRACT_ATI:
-	 texUnit->_CurrentCombine->_NumArgsRGB = 3;
-	 break;
-      default:
-	 texUnit->_CurrentCombine->_NumArgsRGB = 0;
-         _mesa_problem(ctx, "invalid RGB combine mode in update_texture_state");
-         return;
-      }
+      ctx->Texture._EnabledUnits |= (1 << unit);
 
-      switch (texUnit->_CurrentCombine->ModeA) {
-      case GL_REPLACE:
-	 texUnit->_CurrentCombine->_NumArgsA = 1;
-	 break;
-      case GL_MODULATE:
-      case GL_ADD:
-      case GL_ADD_SIGNED:
-      case GL_SUBTRACT:
-	 texUnit->_CurrentCombine->_NumArgsA = 2;
-	 break;
-      case GL_INTERPOLATE:
-      case GL_MODULATE_ADD_ATI:
-      case GL_MODULATE_SIGNED_ADD_ATI:
-      case GL_MODULATE_SUBTRACT_ATI:
-	 texUnit->_CurrentCombine->_NumArgsA = 3;
-	 break;
-      default:
-	 texUnit->_CurrentCombine->_NumArgsA = 0;
-         _mesa_problem(ctx, "invalid Alpha combine mode in update_texture_state");
-	 break;
-      }
+      update_tex_combine(ctx, texUnit);
    }
 
+
    /* Determine which texture coordinate sets are actually needed */
    if (fprog) {
       const GLuint coordMask = (1 << MAX_TEXTURE_COORD_UNITS) - 1;
@@ -621,7 +648,7 @@ update_texture_state( GLcontext *ctx )
    }
 
    /* Setup texgen for those texture coordinate sets that are in use */
-   for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
+   for (unit = 0; unit < ctx->Const.MaxTextureCoordUnits; unit++) {
       struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
 
       if (!(ctx->Texture._EnabledCoordUnits & (1 << unit)))
diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c
index abeed3baa1..6360ca15f8 100644
--- a/src/mesa/main/texstore.c
+++ b/src/mesa/main/texstore.c
@@ -3,6 +3,7 @@
  * Version:  7.3
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -885,8 +886,8 @@ _mesa_swizzle_ubyte_image(GLcontext *ctx,
 
 /*    _mesa_printf("map %d %d %d %d\n", map[0], map[1], map[2], map[3]);  */
 
-   if (srcRowStride == dstRowStride &&
-       srcComponents == dstComponents &&
+   if (srcComponents == dstComponents &&
+       srcRowStride == dstRowStride &&
        srcRowStride == srcWidth * srcComponents &&
        dimensions < 3) {
       /* 1 and 2D images only */
@@ -1898,6 +1899,60 @@ _mesa_texstore_bgr888(TEXSTORE_PARAMS)
    return GL_TRUE;
 }
 
+GLboolean
+_mesa_texstore_rgba4444(TEXSTORE_PARAMS)
+{
+   ASSERT(dstFormat == &_mesa_texformat_rgba4444);
+   ASSERT(dstFormat->TexelBytes == 2);
+
+   if (!ctx->_ImageTransferState &&
+       !srcPacking->SwapBytes &&
+       dstFormat == &_mesa_texformat_rgba4444 &&
+       baseInternalFormat == GL_RGBA &&
+       srcFormat == GL_RGBA &&
+       srcType == GL_UNSIGNED_SHORT_4_4_4_4){
+      /* simple memcpy path */
+      memcpy_texture(ctx, dims,
+                     dstFormat, dstAddr, dstXoffset, dstYoffset, dstZoffset,
+                     dstRowStride,
+                     dstImageOffsets,
+                     srcWidth, srcHeight, srcDepth, srcFormat, srcType,
+                     srcAddr, srcPacking);
+   }
+   else {
+      /* general path */
+      const GLchan *tempImage = _mesa_make_temp_chan_image(ctx, dims,
+                                                 baseInternalFormat,
+                                                 dstFormat->BaseFormat,
+                                                 srcWidth, srcHeight, srcDepth,
+                                                 srcFormat, srcType, srcAddr,
+                                                 srcPacking);
+      const GLchan *src = tempImage;
+      GLint img, row, col;
+      if (!tempImage)
+         return GL_FALSE;
+      _mesa_adjust_image_for_convolution(ctx, dims, &srcWidth, &srcHeight);
+      for (img = 0; img < srcDepth; img++) {
+         GLubyte *dstRow = (GLubyte *) dstAddr
+            + dstImageOffsets[dstZoffset + img] * dstFormat->TexelBytes
+            + dstYoffset * dstRowStride
+            + dstXoffset * dstFormat->TexelBytes;
+         for (row = 0; row < srcHeight; row++) {
+            GLushort *dstUS = (GLushort *) dstRow;
+	    for (col = 0; col < srcWidth; col++) {
+	      dstUS[col] = PACK_COLOR_4444( CHAN_TO_UBYTE(src[RCOMP]),
+					    CHAN_TO_UBYTE(src[GCOMP]),
+					    CHAN_TO_UBYTE(src[BCOMP]),
+					    CHAN_TO_UBYTE(src[ACOMP]) );
+	      src += 4;
+            }
+            dstRow += dstRowStride;
+         }
+      }
+      _mesa_free((void *) tempImage);
+   }
+   return GL_TRUE;
+}
 
 GLboolean
 _mesa_texstore_argb4444(TEXSTORE_PARAMS)
@@ -1966,7 +2021,60 @@ _mesa_texstore_argb4444(TEXSTORE_PARAMS)
    return GL_TRUE;
 }
 
+GLboolean
+_mesa_texstore_rgba5551(TEXSTORE_PARAMS)
+{
+   ASSERT(dstFormat == &_mesa_texformat_rgba5551);
+   ASSERT(dstFormat->TexelBytes == 2);
 
+   if (!ctx->_ImageTransferState &&
+       !srcPacking->SwapBytes &&
+       dstFormat == &_mesa_texformat_rgba5551 &&
+       baseInternalFormat == GL_RGBA &&
+       srcFormat == GL_RGBA &&
+       srcType == GL_UNSIGNED_SHORT_5_5_5_1) {
+      /* simple memcpy path */
+      memcpy_texture(ctx, dims,
+                     dstFormat, dstAddr, dstXoffset, dstYoffset, dstZoffset,
+                     dstRowStride,
+                     dstImageOffsets,
+                     srcWidth, srcHeight, srcDepth, srcFormat, srcType,
+                     srcAddr, srcPacking);
+   }
+   else {
+      /* general path */
+      const GLchan *tempImage = _mesa_make_temp_chan_image(ctx, dims,
+                                                 baseInternalFormat,
+                                                 dstFormat->BaseFormat,
+                                                 srcWidth, srcHeight, srcDepth,
+                                                 srcFormat, srcType, srcAddr,
+                                                 srcPacking);
+      const GLchan *src =tempImage;
+      GLint img, row, col;
+      if (!tempImage)
+         return GL_FALSE;
+      _mesa_adjust_image_for_convolution(ctx, dims, &srcWidth, &srcHeight);
+      for (img = 0; img < srcDepth; img++) {
+         GLubyte *dstRow = (GLubyte *) dstAddr
+            + dstImageOffsets[dstZoffset + img] * dstFormat->TexelBytes
+            + dstYoffset * dstRowStride
+            + dstXoffset * dstFormat->TexelBytes;
+         for (row = 0; row < srcHeight; row++) {
+            GLushort *dstUS = (GLushort *) dstRow;
+	    for (col = 0; col < srcWidth; col++) {
+	       dstUS[col] = PACK_COLOR_5551( CHAN_TO_UBYTE(src[RCOMP]),
+					     CHAN_TO_UBYTE(src[GCOMP]),
+					     CHAN_TO_UBYTE(src[BCOMP]),
+					     CHAN_TO_UBYTE(src[ACOMP]) );
+	      src += 4;
+	    }
+            dstRow += dstRowStride;
+         }
+      }
+      _mesa_free((void *) tempImage);
+   }
+   return GL_TRUE;
+}
 
 GLboolean
 _mesa_texstore_argb1555(TEXSTORE_PARAMS)
@@ -2662,7 +2770,6 @@ _mesa_texstore_rgba_float16(TEXSTORE_PARAMS)
 GLboolean
 _mesa_texstore_srgb8(TEXSTORE_PARAMS)
 {
-   const GLboolean littleEndian = _mesa_little_endian();
    const struct gl_texture_format *newDstFormat;
    StoreTexImageFunc store;
    GLboolean k;
@@ -2670,14 +2777,8 @@ _mesa_texstore_srgb8(TEXSTORE_PARAMS)
    ASSERT(dstFormat == &_mesa_texformat_srgb8);
 
    /* reuse normal rgb texstore code */
-   if (littleEndian) {
-      newDstFormat = &_mesa_texformat_bgr888;
-      store = _mesa_texstore_bgr888;
-   }
-   else {
-      newDstFormat = &_mesa_texformat_rgb888;
-      store = _mesa_texstore_rgb888;
-   }
+   newDstFormat = &_mesa_texformat_rgb888;
+   store = _mesa_texstore_rgb888;
 
    k = store(ctx, dims, baseInternalFormat,
              newDstFormat, dstAddr,
@@ -2693,17 +2794,13 @@ _mesa_texstore_srgb8(TEXSTORE_PARAMS)
 GLboolean
 _mesa_texstore_srgba8(TEXSTORE_PARAMS)
 {
-   const GLboolean littleEndian = _mesa_little_endian();
    const struct gl_texture_format *newDstFormat;
    GLboolean k;
 
    ASSERT(dstFormat == &_mesa_texformat_srgba8);
 
    /* reuse normal rgba texstore code */
-   if (littleEndian)
-      newDstFormat = &_mesa_texformat_rgba8888_rev;
-   else
-      newDstFormat = &_mesa_texformat_rgba8888;
+   newDstFormat = &_mesa_texformat_rgba8888;
 
    k = _mesa_texstore_rgba8888(ctx, dims, baseInternalFormat,
                                newDstFormat, dstAddr,
@@ -2717,6 +2814,28 @@ _mesa_texstore_srgba8(TEXSTORE_PARAMS)
 
 
 GLboolean
+_mesa_texstore_sargb8(TEXSTORE_PARAMS)
+{
+   const struct gl_texture_format *newDstFormat;
+   GLboolean k;
+
+   ASSERT(dstFormat == &_mesa_texformat_sargb8);
+
+   /* reuse normal rgba texstore code */
+   newDstFormat = &_mesa_texformat_argb8888;
+
+   k = _mesa_texstore_argb8888(ctx, dims, baseInternalFormat,
+                               newDstFormat, dstAddr,
+                               dstXoffset, dstYoffset, dstZoffset,
+                               dstRowStride, dstImageOffsets,
+                               srcWidth, srcHeight, srcDepth,
+                               srcFormat, srcType,
+                               srcAddr, srcPacking);
+   return k;
+}
+
+
+GLboolean
 _mesa_texstore_sl8(TEXSTORE_PARAMS)
 {
    const struct gl_texture_format *newDstFormat;
@@ -2741,17 +2860,13 @@ _mesa_texstore_sl8(TEXSTORE_PARAMS)
 GLboolean
 _mesa_texstore_sla8(TEXSTORE_PARAMS)
 {
-   const GLboolean littleEndian = _mesa_little_endian();
    const struct gl_texture_format *newDstFormat;
    GLboolean k;
 
    ASSERT(dstFormat == &_mesa_texformat_sla8);
 
    /* reuse normal luminance/alpha texstore code */
-   if (littleEndian)
-      newDstFormat = &_mesa_texformat_al88;
-   else
-      newDstFormat = &_mesa_texformat_al88_rev;
+   newDstFormat = &_mesa_texformat_al88;
 
    k = _mesa_texstore_al88(ctx, dims, baseInternalFormat,
                            newDstFormat, dstAddr,
@@ -3581,16 +3696,40 @@ is_srgb_teximage(const struct gl_texture_image *texImage)
    switch (texImage->TexFormat->MesaFormat) {
    case MESA_FORMAT_SRGB8:
    case MESA_FORMAT_SRGBA8:
+   case MESA_FORMAT_SARGB8:
    case MESA_FORMAT_SL8:
    case MESA_FORMAT_SLA8:
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
       return GL_TRUE;
    default:
       return GL_FALSE;
    }
 }
 
-#endif /* FEATURE_EXT_texture_sRGB */
 
+/**
+ * Convert a float value from linear space to a
+ * non-linear sRGB value in [0, 255].
+ * Not terribly efficient.
+ */
+static INLINE GLfloat
+linear_to_nonlinear(GLfloat cl)
+{
+   /* can't have values outside [0, 1] */
+   GLfloat cs;
+   if (cl < 0.0031308) {
+      cs = 12.92 * cl;
+   }
+   else {
+      cs = 1.055 * _mesa_pow(cl, 0.41666) - 0.055;
+   }
+   return cs;
+}
+
+#endif /* FEATURE_EXT_texture_sRGB */
 
 /**
  * This is the software fallback for Driver.GetTexImage().
@@ -3707,18 +3846,48 @@ _mesa_get_teximage(GLcontext *ctx, GLenum target, GLint level,
             }
 #if FEATURE_EXT_texture_sRGB
             else if (is_srgb_teximage(texImage)) {
-               /* no pixel transfer and no non-linear to linear conversion */
-               const GLint comps = texImage->TexFormat->TexelBytes;
-               const GLint rowstride = comps * texImage->RowStride;
-               MEMCPY(dest,
-                      (const GLubyte *) texImage->Data + row * rowstride,
-                      comps * width * sizeof(GLubyte));
+               /* special case this since need to backconvert values */
+               /* convert row to RGBA format */
+               GLfloat rgba[MAX_WIDTH][4];
+               GLint col;
+               GLbitfield transferOps = 0x0;
+
+               for (col = 0; col < width; col++) {
+                  (*texImage->FetchTexelf)(texImage, col, row, img, rgba[col]);
+                  if (texImage->TexFormat->BaseFormat == GL_LUMINANCE) {
+                     rgba[col][RCOMP] = linear_to_nonlinear(rgba[col][RCOMP]);
+                     rgba[col][GCOMP] = 0.0;
+                     rgba[col][BCOMP] = 0.0;
+                  }
+                  else if (texImage->TexFormat->BaseFormat == GL_LUMINANCE_ALPHA) {
+                     rgba[col][RCOMP] = linear_to_nonlinear(rgba[col][RCOMP]);
+                     rgba[col][GCOMP] = 0.0;
+                     rgba[col][BCOMP] = 0.0;
+                  }
+                  else if (texImage->TexFormat->BaseFormat == GL_RGB ||
+                     texImage->TexFormat->BaseFormat == GL_RGBA) {
+                     rgba[col][RCOMP] = linear_to_nonlinear(rgba[col][RCOMP]);
+                     rgba[col][GCOMP] = linear_to_nonlinear(rgba[col][GCOMP]);
+                     rgba[col][BCOMP] = linear_to_nonlinear(rgba[col][BCOMP]);
+                  }
+               }
+               _mesa_pack_rgba_span_float(ctx, width, (GLfloat (*)[4]) rgba,
+                                          format, type, dest,
+                                          &ctx->Pack, transferOps /*image xfer ops*/);
             }
 #endif /* FEATURE_EXT_texture_sRGB */
             else {
                /* general case:  convert row to RGBA format */
                GLfloat rgba[MAX_WIDTH][4];
                GLint col;
+               GLbitfield transferOps = 0x0;
+
+               if (type == GL_FLOAT && 
+                   ((ctx->Color.ClampReadColor == GL_TRUE) ||
+                    (ctx->Color.ClampReadColor == GL_FIXED_ONLY_ARB &&
+                     texImage->TexFormat->DataType != GL_FLOAT)))
+                  transferOps |= IMAGE_CLAMP_BIT;
+
                for (col = 0; col < width; col++) {
                   (*texImage->FetchTexelf)(texImage, col, row, img, rgba[col]);
                   if (texImage->TexFormat->BaseFormat == GL_ALPHA) {
@@ -3743,7 +3912,7 @@ _mesa_get_teximage(GLcontext *ctx, GLenum target, GLint level,
                }
                _mesa_pack_rgba_span_float(ctx, width, (GLfloat (*)[4]) rgba,
                                           format, type, dest,
-                                          &ctx->Pack, 0x0 /*image xfer ops*/);
+                                          &ctx->Pack, transferOps /*image xfer ops*/);
             } /* format */
          } /* row */
       } /* img */
diff --git a/src/mesa/main/texstore.h b/src/mesa/main/texstore.h
index c9edf14dbc..b03386b2ac 100644
--- a/src/mesa/main/texstore.h
+++ b/src/mesa/main/texstore.h
@@ -3,6 +3,7 @@
  * Version:  6.5.1
  *
  * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 VMware, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -46,8 +47,10 @@ extern GLboolean _mesa_texstore_rgb888(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_bgr888(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_rgb565(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_rgb565_rev(TEXSTORE_PARAMS);
+extern GLboolean _mesa_texstore_rgba4444(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_argb4444(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_argb4444_rev(TEXSTORE_PARAMS);
+extern GLboolean _mesa_texstore_rgba5551(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_argb1555(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_argb1555_rev(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_al88(TEXSTORE_PARAMS);
@@ -71,6 +74,7 @@ extern GLboolean _mesa_texstore_rgba_dxt5(TEXSTORE_PARAMS);
 #if FEATURE_EXT_texture_sRGB
 extern GLboolean _mesa_texstore_srgb8(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_srgba8(TEXSTORE_PARAMS);
+extern GLboolean _mesa_texstore_sargb8(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_sl8(TEXSTORE_PARAMS);
 extern GLboolean _mesa_texstore_sla8(TEXSTORE_PARAMS);
 #endif
diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c
index 9d9b28b518..106252e460 100644
--- a/src/mesa/main/varray.c
+++ b/src/mesa/main/varray.c
@@ -52,11 +52,13 @@
 static void
 update_array(GLcontext *ctx, struct gl_client_array *array,
              GLbitfield dirtyBit, GLsizei elementSize,
-             GLint size, GLenum type,
+             GLint size, GLenum type, GLenum format,
              GLsizei stride, GLboolean normalized, const GLvoid *ptr)
 {
+   ASSERT(format == GL_RGBA || format == GL_BGRA);
    array->Size = size;
    array->Type = type;
+   array->Format = format;
    array->Stride = stride;
    array->StrideB = stride ? stride : elementSize;
    array->Normalized = normalized;
@@ -132,7 +134,7 @@ _mesa_VertexPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr)
    }
 
    update_array(ctx, &ctx->Array.ArrayObj->Vertex, _NEW_ARRAY_VERTEX,
-                elementSize, size, type, stride, GL_FALSE, ptr);
+                elementSize, size, type, GL_RGBA, stride, GL_FALSE, ptr);
 
    if (ctx->Driver.VertexPointer)
       ctx->Driver.VertexPointer( ctx, size, type, stride, ptr );
@@ -182,7 +184,7 @@ _mesa_NormalPointer(GLenum type, GLsizei stride, const GLvoid *ptr )
    }
 
    update_array(ctx, &ctx->Array.ArrayObj->Normal, _NEW_ARRAY_NORMAL,
-                elementSize, 3, type, stride, GL_TRUE, ptr);
+                elementSize, 3, type, GL_RGBA, stride, GL_TRUE, ptr);
 
    if (ctx->Driver.NormalPointer)
       ctx->Driver.NormalPointer( ctx, type, stride, ptr );
@@ -193,12 +195,15 @@ void GLAPIENTRY
 _mesa_ColorPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr)
 {
    GLsizei elementSize;
+   GLenum format;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
    if (size < 3 || size > 4) {
-      _mesa_error( ctx, GL_INVALID_VALUE, "glColorPointer(size)" );
-      return;
+      if (!ctx->Extensions.EXT_vertex_array_bgra || size != GL_BGRA) {
+         _mesa_error(ctx, GL_INVALID_VALUE, "glColorPointer(size)");
+         return;
+      }
    }
    if (stride < 0) {
       _mesa_error( ctx, GL_INVALID_VALUE, "glColorPointer(stride)" );
@@ -209,6 +214,18 @@ _mesa_ColorPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr)
       _mesa_debug(ctx, "glColorPointer( sz %d type %s stride %d )\n", size,
                   _mesa_lookup_enum_by_nr( type ), stride);
 
+   if (size == GL_BGRA) {
+      if (type != GL_UNSIGNED_BYTE) {
+         _mesa_error(ctx, GL_INVALID_VALUE, "glColorPointer(GL_BGRA/GLubyte)");
+         return;
+      }
+      format = GL_BGRA;
+      size = 4;
+   }
+   else {
+      format = GL_RGBA;
+   }
+
    switch (type) {
       case GL_BYTE:
          elementSize = size * sizeof(GLbyte);
@@ -245,7 +262,7 @@ _mesa_ColorPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr)
    }
 
    update_array(ctx, &ctx->Array.ArrayObj->Color, _NEW_ARRAY_COLOR0,
-                elementSize, size, type, stride, GL_TRUE, ptr);
+                elementSize, size, type, format, stride, GL_TRUE, ptr);
 
    if (ctx->Driver.ColorPointer)
       ctx->Driver.ColorPointer( ctx, size, type, stride, ptr );
@@ -277,7 +294,7 @@ _mesa_FogCoordPointerEXT(GLenum type, GLsizei stride, const GLvoid *ptr)
    }
 
    update_array(ctx, &ctx->Array.ArrayObj->FogCoord, _NEW_ARRAY_FOGCOORD,
-                elementSize, 1, type, stride, GL_FALSE, ptr);
+                elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr);
 
    if (ctx->Driver.FogCoordPointer)
       ctx->Driver.FogCoordPointer( ctx, type, stride, ptr );
@@ -318,7 +335,7 @@ _mesa_IndexPointer(GLenum type, GLsizei stride, const GLvoid *ptr)
    }
 
    update_array(ctx, &ctx->Array.ArrayObj->Index, _NEW_ARRAY_INDEX,
-                elementSize, 1, type, stride, GL_FALSE, ptr);
+                elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr);
 
    if (ctx->Driver.IndexPointer)
       ctx->Driver.IndexPointer( ctx, type, stride, ptr );
@@ -330,12 +347,15 @@ _mesa_SecondaryColorPointerEXT(GLint size, GLenum type,
 			       GLsizei stride, const GLvoid *ptr)
 {
    GLsizei elementSize;
+   GLenum format;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
 
    if (size != 3 && size != 4) {
-      _mesa_error( ctx, GL_INVALID_VALUE, "glSecondaryColorPointer(size)" );
-      return;
+      if (!ctx->Extensions.EXT_vertex_array_bgra || size != GL_BGRA) {
+         _mesa_error(ctx, GL_INVALID_VALUE, "glSecondaryColorPointer(size)");
+         return;
+      }
    }
    if (stride < 0) {
       _mesa_error( ctx, GL_INVALID_VALUE, "glSecondaryColorPointer(stride)" );
@@ -346,6 +366,18 @@ _mesa_SecondaryColorPointerEXT(GLint size, GLenum type,
       _mesa_debug(ctx, "glSecondaryColorPointer( sz %d type %s stride %d )\n",
                   size, _mesa_lookup_enum_by_nr( type ), stride);
 
+   if (size == GL_BGRA) {
+      if (type != GL_UNSIGNED_BYTE) {
+         _mesa_error(ctx, GL_INVALID_VALUE, "glColorPointer(GL_BGRA/GLubyte)");
+         return;
+      }
+      format = GL_BGRA;
+      size = 4;
+   }
+   else {
+      format = GL_RGBA;
+   }
+
    switch (type) {
       case GL_BYTE:
          elementSize = size * sizeof(GLbyte);
@@ -377,7 +409,7 @@ _mesa_SecondaryColorPointerEXT(GLint size, GLenum type,
    }
 
    update_array(ctx, &ctx->Array.ArrayObj->SecondaryColor, _NEW_ARRAY_COLOR1,
-                elementSize, size, type, stride, GL_TRUE, ptr);
+                elementSize, size, type, format, stride, GL_TRUE, ptr);
 
    if (ctx->Driver.SecondaryColorPointer)
       ctx->Driver.SecondaryColorPointer( ctx, size, type, stride, ptr );
@@ -437,7 +469,7 @@ _mesa_TexCoordPointer(GLint size, GLenum type, GLsizei stride,
 
    update_array(ctx, &ctx->Array.ArrayObj->TexCoord[unit],
                 _NEW_ARRAY_TEXCOORD(unit),
-                elementSize, size, type, stride, GL_FALSE, ptr);
+                elementSize, size, type, GL_RGBA, stride, GL_FALSE, ptr);
 
    if (ctx->Driver.TexCoordPointer)
       ctx->Driver.TexCoordPointer( ctx, size, type, stride, ptr );
@@ -456,7 +488,8 @@ _mesa_EdgeFlagPointer(GLsizei stride, const GLvoid *ptr)
    }
 
    update_array(ctx, &ctx->Array.ArrayObj->EdgeFlag, _NEW_ARRAY_EDGEFLAG,
-                sizeof(GLboolean), 1, GL_UNSIGNED_BYTE, stride, GL_FALSE, ptr);
+                sizeof(GLboolean), 1, GL_UNSIGNED_BYTE, GL_RGBA,
+                stride, GL_FALSE, ptr);
 
    if (ctx->Driver.EdgeFlagPointer)
       ctx->Driver.EdgeFlagPointer( ctx, stride, ptr );
@@ -490,7 +523,7 @@ _mesa_PointSizePointer(GLenum type, GLsizei stride, const GLvoid *ptr)
    }
 
    update_array(ctx, &ctx->Array.ArrayObj->PointSize, _NEW_ARRAY_POINT_SIZE,
-                elementSize, 1, type, stride, GL_FALSE, ptr);
+                elementSize, 1, type, GL_RGBA, stride, GL_FALSE, ptr);
 }
 
 
@@ -546,7 +579,7 @@ _mesa_VertexAttribPointerNV(GLuint index, GLint size, GLenum type,
 
    update_array(ctx, &ctx->Array.ArrayObj->VertexAttrib[index],
                 _NEW_ARRAY_ATTRIB(index),
-                elementSize, size, type, stride, normalized, ptr);
+                elementSize, size, type, GL_RGBA, stride, normalized, ptr);
 
    if (ctx->Driver.VertexAttribPointer)
       ctx->Driver.VertexAttribPointer( ctx, index, size, type, stride, ptr );
@@ -561,6 +594,7 @@ _mesa_VertexAttribPointerARB(GLuint index, GLint size, GLenum type,
                              GLsizei stride, const GLvoid *ptr)
 {
    GLsizei elementSize;
+   GLenum format;
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
@@ -570,8 +604,10 @@ _mesa_VertexAttribPointerARB(GLuint index, GLint size, GLenum type,
    }
 
    if (size < 1 || size > 4) {
-      _mesa_error(ctx, GL_INVALID_VALUE, "glVertexAttribPointerARB(size)");
-      return;
+      if (!ctx->Extensions.EXT_vertex_array_bgra || size != GL_BGRA) {
+         _mesa_error(ctx, GL_INVALID_VALUE, "glVertexAttribPointerARB(size)");
+         return;
+      }
    }
 
    if (stride < 0) {
@@ -579,6 +615,20 @@ _mesa_VertexAttribPointerARB(GLuint index, GLint size, GLenum type,
       return;
    }
 
+   if (size == GL_BGRA) {
+      if (type != GL_UNSIGNED_BYTE) {
+         _mesa_error(ctx, GL_INVALID_VALUE,
+                     "glVertexAttribPointerARB(GL_BGRA/type)");
+         return;
+      }
+      format = GL_BGRA;
+      size = 4;
+      normalized = GL_TRUE;
+   }
+   else {
+      format = GL_RGBA;
+   }
+
    /* check for valid 'type' and compute StrideB right away */
    /* NOTE: more types are supported here than in the NV extension */
    switch (type) {
@@ -618,7 +668,7 @@ _mesa_VertexAttribPointerARB(GLuint index, GLint size, GLenum type,
 
    update_array(ctx, &ctx->Array.ArrayObj->VertexAttrib[index],
                 _NEW_ARRAY_ATTRIB(index),
-                elementSize, size, type, stride, normalized, ptr);
+                elementSize, size, type, GL_RGBA, stride, normalized, ptr);
 
    if (ctx->Driver.VertexAttribPointer)
       ctx->Driver.VertexAttribPointer(ctx, index, size, type, stride, ptr);
diff --git a/src/mesa/main/varray.h b/src/mesa/main/varray.h
index 6385689fcc..97d5c8219d 100644
--- a/src/mesa/main/varray.h
+++ b/src/mesa/main/varray.h
@@ -156,6 +156,20 @@ _mesa_LockArraysEXT(GLint first, GLsizei count);
 extern void GLAPIENTRY
 _mesa_UnlockArraysEXT( void );
 
+
+extern void GLAPIENTRY
+_mesa_DrawArrays(GLenum mode, GLint first, GLsizei count);
+
+extern void GLAPIENTRY
+_mesa_DrawElements(GLenum mode, GLsizei count, GLenum type,
+                   const GLvoid *indices);
+
+extern void GLAPIENTRY
+_mesa_DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count,
+                        GLenum type, const GLvoid *indices);
+
+
+
 extern void
 _mesa_init_varray( GLcontext * ctx );
 
diff --git a/src/mesa/main/version.h b/src/mesa/main/version.h
index 81034a35d1..9287e8515c 100644
--- a/src/mesa/main/version.h
+++ b/src/mesa/main/version.h
@@ -1,6 +1,10 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.3
+<<<<<<< HEAD:src/mesa/main/version.h
+ * Version:  7.5
+=======
+ * Version:  7.4
+>>>>>>> origin/gallium-0.2:src/mesa/main/version.h
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
@@ -29,9 +33,9 @@
 
 /* Mesa version */
 #define MESA_MAJOR 7
-#define MESA_MINOR 3
+#define MESA_MINOR 5
 #define MESA_PATCH 0
-#define MESA_VERSION_STRING "7.3-devel"
+#define MESA_VERSION_STRING "7.5-devel"
 
 /* To make version comparison easy */
 #define MESA_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))
diff --git a/src/mesa/math/m_debug_xform.c b/src/mesa/math/m_debug_xform.c
index ec22c7052d..df8cc066b6 100644
--- a/src/mesa/math/m_debug_xform.c
+++ b/src/mesa/math/m_debug_xform.c
@@ -206,7 +206,8 @@ static int test_transform_function( transform_func func, int psize,
          case VAR:
             break;
          default:
-            abort();
+            ASSERT(0);
+            return 0;
          }
       }
    }
diff --git a/src/mesa/math/m_matrix.h b/src/mesa/math/m_matrix.h
index e8303f3ac5..a8d9000e89 100644
--- a/src/mesa/math/m_matrix.h
+++ b/src/mesa/math/m_matrix.h
@@ -189,6 +189,18 @@ do {								\
 } while (0)
 
 
+/**
+ * Transform a direction by a matrix.
+ */
+#define TRANSFORM_DIRECTION( TO, DIR, MAT )			\
+do {								\
+   TO[0] = DIR[0] * MAT[0] + DIR[1] * MAT[4] + DIR[2] * MAT[8];	\
+   TO[1] = DIR[0] * MAT[1] + DIR[1] * MAT[5] + DIR[2] * MAT[9];	\
+   TO[2] = DIR[0] * MAT[2] + DIR[1] * MAT[6] + DIR[2] * MAT[10];\
+} while (0)
+
+
+
 /*@}*/
 
 
diff --git a/src/mesa/math/m_xform.c b/src/mesa/math/m_xform.c
index 901ae5b416..fdc8abd434 100644
--- a/src/mesa/math/m_xform.c
+++ b/src/mesa/math/m_xform.c
@@ -220,11 +220,3 @@ _math_init_transformation( void )
    _mesa_init_all_x86_64_transform_asm();
 #endif
 }
-
-void
-_math_init( void )
-{
-   _math_init_transformation();
-   _math_init_translate();
-   _math_init_eval();
-}
diff --git a/src/mesa/math/mathmod.h b/src/mesa/math/mathmod.h
index 6fbaaea94b..fb0862b5b1 100644
--- a/src/mesa/math/mathmod.h
+++ b/src/mesa/math/mathmod.h
@@ -36,6 +36,4 @@
 #ifndef _MESA_MATH_H_
 #define _MESA_MATH_H_
 
-extern void _math_init( void );
-
 #endif
diff --git a/src/mesa/osmesa.pc.in b/src/mesa/osmesa.pc.in
new file mode 100644
index 0000000000..05327f40aa
--- /dev/null
+++ b/src/mesa/osmesa.pc.in
@@ -0,0 +1,12 @@
+prefix=@INSTALL_DIR@
+exec_prefix=${prefix}
+libdir=@INSTALL_LIB_DIR@
+includedir=@INSTALL_INC_DIR@
+
+Name: osmesa
+Description: Mesa Off-screen Rendering library
+Requires: @OSMESA_PC_REQ@
+Version: @VERSION@
+Libs: -L${libdir} -l@OSMESA_LIB@
+Libs.private: @OSMESA_PC_LIB_PRIV@
+Cflags: -I${includedir}
diff --git a/src/mesa/ppc/common_ppc.c b/src/mesa/ppc/common_ppc.c
index 786e1817ac..c7cec96e5d 100644
--- a/src/mesa/ppc/common_ppc.c
+++ b/src/mesa/ppc/common_ppc.c
@@ -85,7 +85,7 @@ void _mesa_init_all_ppc_transform_asm( void )
    }
    
 # ifndef USE_VMX_ASM
-   _mesa_ppc_cpu_features &= ~PPC_FEATURES_HAS_ALTIVEC;
+   _mesa_ppc_cpu_features &= ~PPC_FEATURE_HAS_ALTIVEC;
 # endif
 #endif
 }
diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c
index 39988b5fca..2e0fc3694a 100644
--- a/src/mesa/shader/arbprogparse.c
+++ b/src/mesa/shader/arbprogparse.c
@@ -30,6 +30,27 @@
  * \author Karl Rasche
  */
 
+/**
+Notes on program parameters, etc.
+
+The instructions we emit will use six kinds of source registers:
+
+  PROGRAM_INPUT      - input registers
+  PROGRAM_TEMPORARY  - temp registers
+  PROGRAM_ADDRESS    - address/indirect register
+  PROGRAM_SAMPLER    - texture sampler
+  PROGRAM_CONSTANT   - indexes into program->Parameters, a known constant/literal
+  PROGRAM_STATE_VAR  - indexes into program->Parameters, and may actually be:
+                       + a state variable, like "state.fog.color", or
+                       + a pointer to a "program.local[k]" parameter, or
+                       + a pointer to a "program.env[k]" parameter
+
+Basically, all the program.local[] and program.env[] values will get mapped
+into the unified gl_program->Parameters array.  This solves the problem of
+having three separate program parameter arrays.
+*/
+
+
 #include "main/glheader.h"
 #include "main/imports.h"
 #include "main/context.h"
@@ -38,16 +59,11 @@
 #include "shader/grammar/grammar_mesa.h"
 #include "arbprogparse.h"
 #include "program.h"
+#include "programopt.h"
 #include "prog_parameter.h"
 #include "prog_statevars.h"
 #include "prog_instruction.h"
 
-
-/* For ARB programs, use the NV instruction limits */
-#define MAX_INSTRUCTIONS MAX2(MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS, \
-                              MAX_NV_VERTEX_PROGRAM_INSTRUCTIONS)
-
-
 /**
  * This is basically a union of the vertex_program and fragment_program
  * structs that we can use to parse the program into
@@ -963,6 +979,8 @@ parse_output_color_num (GLcontext * ctx, const GLubyte ** inst,
 
 
 /**
+ * Validate the index of a texture coordinate
+ *
  * \param coord The texture unit index
  * \return 0 on sucess, 1 on error
  */
@@ -972,8 +990,8 @@ parse_texcoord_num (GLcontext * ctx, const GLubyte ** inst,
 {
    GLint i = parse_integer (inst, Program);
 
-   if ((i < 0) || (i >= (int)ctx->Const.MaxTextureUnits)) {
-      program_error(ctx, Program->Position, "Invalid texture unit index");
+   if ((i < 0) || (i >= (int)ctx->Const.MaxTextureCoordUnits)) {
+      program_error(ctx, Program->Position, "Invalid texture coordinate index");
       return 1;
    }
 
@@ -981,6 +999,29 @@ parse_texcoord_num (GLcontext * ctx, const GLubyte ** inst,
    return 0;
 }
 
+
+/**
+ * Validate the index of a texture image unit
+ *
+ * \param coord The texture unit index
+ * \return 0 on sucess, 1 on error
+ */
+static GLuint
+parse_teximage_num (GLcontext * ctx, const GLubyte ** inst,
+                    struct arb_program *Program, GLuint * coord)
+{
+   GLint i = parse_integer (inst, Program);
+
+   if ((i < 0) || (i >= (int)ctx->Const.MaxTextureImageUnits)) {
+      program_error(ctx, Program->Position, "Invalid texture image index");
+      return 1;
+   }
+
+   *coord = (GLuint) i;
+   return 0;
+}
+
+
 /**
  * \param coord The weight index
  * \return 0 on sucess, 1 on error
@@ -1870,7 +1911,11 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst,
                                         const_values, 4);
          if (param_var->param_binding_begin == ~0U)
             param_var->param_binding_begin = idx;
-         param_var->param_binding_type = PROGRAM_CONSTANT;
+         param_var->param_binding_type = PROGRAM_STATE_VAR;
+         /* Note: when we reference this parameter in an instruction later,
+          * we'll check if it's really a constant/immediate and set the
+          * instruction register type appropriately.
+          */
          param_var->param_binding_length++;
          Program->Base.NumParameters++;
          break;
@@ -2577,6 +2622,18 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst,
          return 1;
    }
 
+   if (*File == PROGRAM_STATE_VAR) {
+      enum register_file file;
+
+      /* If we're referencing the Program->Parameters[] array, check if the
+       * parameter is really a constant/literal.  If so, set File to CONSTANT.
+       */
+      assert(*Index < (GLint) Program->Base.Parameters->NumParameters);
+      file = Program->Base.Parameters->Parameters[*Index].Type;
+      if (file == PROGRAM_CONSTANT)
+         *File = PROGRAM_CONSTANT;
+   }
+
    /* Add attributes to InputsRead only if they are used the program.
     * This avoids the handling of unused ATTRIB declarations in the drivers. */
    if (*File == PROGRAM_INPUT)
@@ -3003,7 +3060,7 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst,
             return 1;
 
          /* texImageUnit */
-         if (parse_texcoord_num (ctx, inst, Program, &texcoord))
+         if (parse_teximage_num (ctx, inst, Program, &texcoord))
             return 1;
          fp->TexSrcUnit = texcoord;
 
@@ -3353,11 +3410,11 @@ debug_variables (GLcontext * ctx, struct var_cache *vc_head,
                fprintf (stderr, "%s\n",
                         Program->Base.Parameters->Parameters[a + b].Name);
                if (Program->Base.Parameters->Parameters[a + b].Type == PROGRAM_STATE_VAR) {
-                  const char *s;
+                  char *s;
                   s = _mesa_program_state_string(Program->Base.Parameters->Parameters
                                                  [a + b].StateIndexes);
                   fprintf(stderr, "%s\n", s);
-                  _mesa_free((char *) s);
+                  _mesa_free(s);
                }
                else
                   fprintf (stderr, "%f %f %f %f\n",
@@ -3405,7 +3462,7 @@ parse_instructions(GLcontext * ctx, const GLubyte * inst,
       : ctx->Const.VertexProgram.MaxInstructions;
    GLint err = 0;
 
-   ASSERT(MAX_INSTRUCTIONS >= maxInst);
+   ASSERT(MAX_PROGRAM_INSTRUCTIONS >= maxInst);
 
    Program->MajorVersion = (GLuint) * inst++;
    Program->MinorVersion = (GLuint) * inst++;
@@ -3591,8 +3648,7 @@ enable_parser_extensions(GLcontext *ctx, grammar id)
    if (ctx->Extensions.NV_texture_rectangle
        && !enable_ext(ctx, id, "texture_rectangle"))
       return GL_FALSE;
-   if (ctx->Extensions.ARB_draw_buffers
-       && !enable_ext(ctx, id, "draw_buffers"))
+   if (!enable_ext(ctx, id, "draw_buffers"))
       return GL_FALSE;
    if (ctx->Extensions.MESA_texture_array
        && !enable_ext(ctx, id, "texture_array"))
@@ -3760,7 +3816,7 @@ _mesa_parse_arb_program(GLcontext *ctx, GLenum target,
 
    /* Initialize the arb_program struct */
    program->Base.String = strz;
-   program->Base.Instructions = _mesa_alloc_instructions(MAX_INSTRUCTIONS);
+   program->Base.Instructions = _mesa_alloc_instructions(MAX_PROGRAM_INSTRUCTIONS);
    program->Base.NumInstructions =
    program->Base.NumTemporaries =
    program->Base.NumParameters =
@@ -3805,12 +3861,12 @@ _mesa_parse_arb_program(GLcontext *ctx, GLenum target,
 
    _mesa_free (parsed);
 
-   /* Reallocate the instruction array from size [MAX_INSTRUCTIONS]
+   /* Reallocate the instruction array from size [MAX_PROGRAM_INSTRUCTIONS]
     * to size [ap.Base.NumInstructions].
     */
    program->Base.Instructions
       = _mesa_realloc_instructions(program->Base.Instructions,
-                                   MAX_INSTRUCTIONS,
+                                   MAX_PROGRAM_INSTRUCTIONS,
                                    program->Base.NumInstructions);
 
    return !err;
@@ -3874,6 +3930,16 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target,
       _mesa_free_parameter_list(program->Base.Parameters);
    program->Base.Parameters    = ap.Base.Parameters;
 
+   /* Append fog instructions now if the program has "OPTION ARB_fog_exp"
+    * or similar.  We used to leave this up to drivers, but it appears
+    * there's no hardware that wants to do fog in a discrete stage separate
+    * from the fragment shader.
+    */
+   if (program->FogOption != GL_NONE) {
+      _mesa_append_fog_code(ctx, program);
+      program->FogOption = GL_NONE;
+   }
+
 #if DEBUG_FP
    _mesa_printf("____________Fragment program %u ________\n", program->Base.Id);
    _mesa_print_program(&program->Base);
diff --git a/src/mesa/shader/arbprogram.c b/src/mesa/shader/arbprogram.c
index 760dac2399..7831e0faf9 100644
--- a/src/mesa/shader/arbprogram.c
+++ b/src/mesa/shader/arbprogram.c
@@ -335,10 +335,6 @@ _mesa_GetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat *params)
          COPY_4V(params, ctx->Current.Attrib[VERT_ATTRIB_GENERIC0 + index]);
          break;
       case GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB:
-         if (!ctx->Extensions.ARB_vertex_buffer_object) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glGetVertexAttribfvARB(pname)");
-            return;
-         }
          params[0] = (GLfloat) ctx->Array.ArrayObj->VertexAttrib[index].BufferObj->Name;
          break;
       default:
@@ -807,6 +803,7 @@ _mesa_GetProgramLocalParameterdvARB(GLenum target, GLuint index,
 {
    GET_CURRENT_CONTEXT(ctx);
    GLfloat floatParams[4];
+   ASSIGN_4V(floatParams, 0.0F, 0.0F, 0.0F, 0.0F);
    _mesa_GetProgramLocalParameterfvARB(target, index, floatParams);
    if (ctx->ErrorValue == GL_NO_ERROR) {
       COPY_4V(params, floatParams);
diff --git a/src/mesa/shader/atifragshader.h b/src/mesa/shader/atifragshader.h
index 32fb3a8019..e1dc20e606 100644
--- a/src/mesa/shader/atifragshader.h
+++ b/src/mesa/shader/atifragshader.h
@@ -8,6 +8,8 @@
 #ifndef ATIFRAGSHADER_H
 #define ATIFRAGSHADER_H
 
+#include "main/mtypes.h"
+
 #define MAX_NUM_INSTRUCTIONS_PER_PASS_ATI 8
 #define MAX_NUM_PASSES_ATI                2
 #define MAX_NUM_FRAGMENT_REGISTERS_ATI    6
diff --git a/src/mesa/shader/grammar/descrip.mms b/src/mesa/shader/grammar/descrip.mms
deleted file mode 100644
index 6976b70d6a..0000000000
--- a/src/mesa/shader/grammar/descrip.mms
+++ /dev/null
@@ -1,42 +0,0 @@
-# Makefile for core library for VMS
-# contributed by Jouk Jansen  joukj@hrem.nano.tudelft.nl
-# Last revision : 29 September 2008
-
-.first
-	define gl [----.include.gl]
-	define math [--.math]
-	define swrast [--.swrast]
-	define array_cache [--.array_cache]
-	define main [--.main]
-
-.include [----]mms-config.
-
-##### MACROS #####
-
-VPATH = RCS
-
-INCDIR = [----.include],[],[--.main],[--.glapi],[-.slang]
-LIBDIR = [----.lib]
-CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm
-
-SOURCES = grammar_mesa.c
-
-OBJECTS = grammar_mesa.obj
-
-##### RULES #####
-
-VERSION=Mesa V3.4
-
-##### TARGETS #####
-all : 
-	$(MMS)$(MMSQUALIFIERS) $(LIBDIR)$(GL_LIB)
-
-# Make the library
-$(LIBDIR)$(GL_LIB) : $(OBJECTS)
-  @ library $(LIBDIR)$(GL_LIB) $(OBJECTS)
-
-clean :
-	purge
-	delete *.obj;*
-
-grammar_mesa.obj : grammar_mesa.c grammar.c
diff --git a/src/mesa/shader/nvprogram.c b/src/mesa/shader/nvprogram.c
index 88272fff3f..16116c4339 100644
--- a/src/mesa/shader/nvprogram.c
+++ b/src/mesa/shader/nvprogram.c
@@ -469,10 +469,6 @@ _mesa_GetVertexAttribivNV(GLuint index, GLenum pname, GLint *params)
          params[3] = (GLint) ctx->Current.Attrib[index][3];
          break;
       case GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB:
-         if (!ctx->Extensions.ARB_vertex_buffer_object) {
-            _mesa_error(ctx, GL_INVALID_ENUM, "glGetVertexAttribdvNV");
-            return;
-         }
          params[0] = ctx->Array.ArrayObj->VertexAttrib[index].BufferObj->Name;
          break;
       default:
diff --git a/src/mesa/shader/prog_cache.c b/src/mesa/shader/prog_cache.c
index 36a25377c5..9437e59613 100644
--- a/src/mesa/shader/prog_cache.c
+++ b/src/mesa/shader/prog_cache.c
@@ -44,6 +44,7 @@ struct cache_item
 struct gl_program_cache
 {
    struct cache_item **items;
+   struct cache_item *last;
    GLuint size, n_items;
 };
 
@@ -83,6 +84,8 @@ rehash(struct gl_program_cache *cache)
    struct cache_item *c, *next;
    GLuint size, i;
 
+   cache->last = NULL;
+
    size = cache->size * 3;
    items = (struct cache_item**) _mesa_malloc(size * sizeof(*items));
    _mesa_memset(items, 0, size * sizeof(*items));
@@ -105,6 +108,8 @@ clear_cache(GLcontext *ctx, struct gl_program_cache *cache)
 {
    struct cache_item *c, *next;
    GLuint i;
+   
+   cache->last = NULL;
 
    for (i = 0; i < cache->size; i++) {
       for (c = cache->items[i]; c; c = next) {
@@ -149,18 +154,26 @@ _mesa_delete_program_cache(GLcontext *ctx, struct gl_program_cache *cache)
 
 
 struct gl_program *
-_mesa_search_program_cache(const struct gl_program_cache *cache,
+_mesa_search_program_cache(struct gl_program_cache *cache,
                            const void *key, GLuint keysize)
 {
-   const GLuint hash = hash_key(key, keysize);
-   struct cache_item *c;
-
-   for (c = cache->items[hash % cache->size]; c; c = c->next) {
-      if (c->hash == hash && memcmp(c->key, key, keysize) == 0)
-	 return c->program;
+   if (cache->last && 
+       memcmp(cache->last->key, key, keysize) == 0) {
+      return cache->last->program;
    }
+   else {
+      const GLuint hash = hash_key(key, keysize);
+      struct cache_item *c;
+
+      for (c = cache->items[hash % cache->size]; c; c = c->next) {
+         if (c->hash == hash && memcmp(c->key, key, keysize) == 0) {
+            cache->last = c;
+            return c->program;
+         }
+      }
 
-   return NULL;
+      return NULL;
+   }
 }
 
 
diff --git a/src/mesa/shader/prog_cache.h b/src/mesa/shader/prog_cache.h
index a8c91fba01..4e1ccac03f 100644
--- a/src/mesa/shader/prog_cache.h
+++ b/src/mesa/shader/prog_cache.h
@@ -42,7 +42,7 @@ _mesa_delete_program_cache(GLcontext *ctx, struct gl_program_cache *pc);
 
 
 extern struct gl_program *
-_mesa_search_program_cache(const struct gl_program_cache *cache,
+_mesa_search_program_cache(struct gl_program_cache *cache,
                            const void *key, GLuint keysize);
 
 extern void
diff --git a/src/mesa/shader/prog_execute.c b/src/mesa/shader/prog_execute.c
index 923611b797..a93733c085 100644
--- a/src/mesa/shader/prog_execute.c
+++ b/src/mesa/shader/prog_execute.c
@@ -43,7 +43,7 @@
 #include "prog_instruction.h"
 #include "prog_parameter.h"
 #include "prog_print.h"
-#include "shader/slang/slang_library_noise.h"
+#include "prog_noise.h"
 
 
 /* debug predicate */
@@ -705,7 +705,7 @@ _mesa_execute_program(GLcontext * ctx,
          {
             GLfloat t[4];
             fetch_vector4(&inst->SrcReg[0], machine, t);
-            machine->AddressReg[0][0] = (GLint) FLOORF(t[0]);
+            machine->AddressReg[0][0] = IFLOOR(t[0]);
          }
          break;
       case OPCODE_BGNLOOP:
@@ -963,7 +963,10 @@ _mesa_execute_program(GLcontext * ctx,
          {
             GLfloat a[4], result[4];
             fetch_vector1(&inst->SrcReg[0], machine, a);
-            result[0] = result[1] = result[2] = result[3] = LOG2(a[0]);
+	    /* The fast LOG2 macro doesn't meet the precision requirements.
+	     */
+            result[0] = result[1] = result[2] = result[3] =
+		(log(a[0]) * 1.442695F);
             store_vector4(inst, machine, result);
          }
          break;
@@ -1022,7 +1025,11 @@ _mesa_execute_program(GLcontext * ctx,
                   GLfloat mantissa = FREXPF(t[0], &exponent);
                   q[0] = (GLfloat) (exponent - 1);
                   q[1] = (GLfloat) (2.0 * mantissa); /* map [.5, 1) -> [1, 2) */
-                  q[2] = (GLfloat) (q[0] + LOG2(q[1]));
+
+		  /* The fast LOG2 macro doesn't meet the precision
+		   * requirements.
+		   */
+                  q[2] = (log(t[0]) * 1.442695F);
                }
             }
             else {
@@ -1137,7 +1144,8 @@ _mesa_execute_program(GLcontext * ctx,
             fetch_vector1(&inst->SrcReg[0], machine, a);
             result[0] =
                result[1] =
-               result[2] = result[3] = _slang_library_noise1(a[0]);
+               result[2] =
+               result[3] = _mesa_noise1(a[0]);
             store_vector4(inst, machine, result);
          }
          break;
@@ -1147,7 +1155,7 @@ _mesa_execute_program(GLcontext * ctx,
             fetch_vector4(&inst->SrcReg[0], machine, a);
             result[0] =
                result[1] =
-               result[2] = result[3] = _slang_library_noise2(a[0], a[1]);
+               result[2] = result[3] = _mesa_noise2(a[0], a[1]);
             store_vector4(inst, machine, result);
          }
          break;
@@ -1158,7 +1166,7 @@ _mesa_execute_program(GLcontext * ctx,
             result[0] =
                result[1] =
                result[2] =
-               result[3] = _slang_library_noise3(a[0], a[1], a[2]);
+               result[3] = _mesa_noise3(a[0], a[1], a[2]);
             store_vector4(inst, machine, result);
          }
          break;
@@ -1169,7 +1177,7 @@ _mesa_execute_program(GLcontext * ctx,
             result[0] =
                result[1] =
                result[2] =
-               result[3] = _slang_library_noise4(a[0], a[1], a[2], a[3]);
+               result[3] = _mesa_noise4(a[0], a[1], a[2], a[3]);
             store_vector4(inst, machine, result);
          }
          break;
diff --git a/src/mesa/shader/prog_execute.h b/src/mesa/shader/prog_execute.h
index 18b13e11a4..8ceb7b092e 100644
--- a/src/mesa/shader/prog_execute.h
+++ b/src/mesa/shader/prog_execute.h
@@ -25,6 +25,8 @@
 #ifndef PROG_EXECUTE_H
 #define PROG_EXECUTE_H
 
+#include "main/config.h"
+
 
 typedef void (*FetchTexelLodFunc)(GLcontext *ctx, const GLfloat texcoord[4],
                                   GLfloat lambda, GLuint unit, GLfloat color[4]);
@@ -36,10 +38,6 @@ typedef void (*FetchTexelDerivFunc)(GLcontext *ctx, const GLfloat texcoord[4],
                                     GLuint unit, GLfloat color[4]);
 
 
-/** The larger of VERT_RESULT_MAX, FRAG_RESULT_MAX */
-#define MAX_PROGRAM_OUTPUTS VERT_RESULT_MAX
-
-
 /**
  * Virtual machine state used during execution of vertex/fragment programs.
  */
diff --git a/src/mesa/shader/prog_instruction.c b/src/mesa/shader/prog_instruction.c
index f5c0a498fb..4a6d0d670a 100644
--- a/src/mesa/shader/prog_instruction.c
+++ b/src/mesa/shader/prog_instruction.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.3
+ * Version:  7.3
  *
- * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -157,7 +158,7 @@ static const struct instruction_info InstInfo[MAX_OPCODE] = {
    { OPCODE_AND,    "AND",     2, 1 },
    { OPCODE_ARA,    "ARA",     1, 1 },
    { OPCODE_ARL,    "ARL",     1, 1 },
-   { OPCODE_ARL_NV, "ARL",     1, 1 },
+   { OPCODE_ARL_NV, "ARL_NV",  1, 1 },
    { OPCODE_ARR,    "ARL",     1, 1 },
    { OPCODE_BGNLOOP,"BGNLOOP", 0, 0 },
    { OPCODE_BGNSUB, "BGNSUB",  0, 0 },
@@ -186,7 +187,7 @@ static const struct instruction_info InstInfo[MAX_OPCODE] = {
    { OPCODE_FRC,    "FRC",     1, 1 },
    { OPCODE_IF,     "IF",      1, 0 },
    { OPCODE_KIL,    "KIL",     1, 0 },
-   { OPCODE_KIL_NV, "KIL",     0, 0 },
+   { OPCODE_KIL_NV, "KIL_NV",  0, 0 },
    { OPCODE_LG2,    "LG2",     1, 1 },
    { OPCODE_LIT,    "LIT",     1, 1 },
    { OPCODE_LOG,    "LOG",     1, 1 },
@@ -235,7 +236,7 @@ static const struct instruction_info InstInfo[MAX_OPCODE] = {
    { OPCODE_TXD,    "TXD",     3, 1 },
    { OPCODE_TXL,    "TXL",     1, 1 },
    { OPCODE_TXP,    "TXP",     1, 1 },
-   { OPCODE_TXP_NV, "TXP",     1, 1 },
+   { OPCODE_TXP_NV, "TXP_NV",  1, 1 },
    { OPCODE_TRUNC,  "TRUNC",   1, 1 },
    { OPCODE_UP2H,   "UP2H",    1, 1 },
    { OPCODE_UP2US,  "UP2US",   1, 1 },
diff --git a/src/mesa/shader/prog_instruction.h b/src/mesa/shader/prog_instruction.h
index 268afc528c..c649b3db5e 100644
--- a/src/mesa/shader/prog_instruction.h
+++ b/src/mesa/shader/prog_instruction.h
@@ -418,11 +418,13 @@ struct prog_instruction
     */
    GLint BranchTarget;
 
+#if 01 /* XXX just use this for i965 driver for now! */
    /**
     * For TEX instructions in shaders, the sampler to use for the
     * texture lookup.
     */
    GLint Sampler;
+#endif
 
    const char *Comment;
 };
diff --git a/src/mesa/shader/prog_noise.c b/src/mesa/shader/prog_noise.c
new file mode 100644
index 0000000000..1713ddb5f3
--- /dev/null
+++ b/src/mesa/shader/prog_noise.c
@@ -0,0 +1,638 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 2006  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/*
+ * SimplexNoise1234
+ * Copyright (c) 2003-2005, Stefan Gustavson
+ *
+ * Contact: stegu@itn.liu.se
+ */
+
+/**
+ * \file
+ * \brief C implementation of Perlin Simplex Noise over 1, 2, 3 and 4 dims.
+ * \author Stefan Gustavson (stegu@itn.liu.se)
+ *
+ *
+ * This implementation is "Simplex Noise" as presented by
+ * Ken Perlin at a relatively obscure and not often cited course
+ * session "Real-Time Shading" at Siggraph 2001 (before real
+ * time shading actually took on), under the title "hardware noise".
+ * The 3D function is numerically equivalent to his Java reference
+ * code available in the PDF course notes, although I re-implemented
+ * it from scratch to get more readable code. The 1D, 2D and 4D cases
+ * were implemented from scratch by me from Ken Perlin's text.
+ *
+ * This file has no dependencies on any other file, not even its own
+ * header file. The header file is made for use by external code only.
+ */
+
+
+#include "main/imports.h"
+#include "prog_noise.h"
+
+#define FASTFLOOR(x) ( ((x)>0) ? ((int)x) : (((int)x)-1) )
+
+/*
+ * ---------------------------------------------------------------------
+ * Static data
+ */
+
+/**
+ * Permutation table. This is just a random jumble of all numbers 0-255,
+ * repeated twice to avoid wrapping the index at 255 for each lookup.
+ * This needs to be exactly the same for all instances on all platforms,
+ * so it's easiest to just keep it as static explicit data.
+ * This also removes the need for any initialisation of this class.
+ *
+ * Note that making this an int[] instead of a char[] might make the
+ * code run faster on platforms with a high penalty for unaligned single
+ * byte addressing. Intel x86 is generally single-byte-friendly, but
+ * some other CPUs are faster with 4-aligned reads.
+ * However, a char[] is smaller, which avoids cache trashing, and that
+ * is probably the most important aspect on most architectures.
+ * This array is accessed a *lot* by the noise functions.
+ * A vector-valued noise over 3D accesses it 96 times, and a
+ * float-valued 4D noise 64 times. We want this to fit in the cache!
+ */
+unsigned char perm[512] = { 151, 160, 137, 91, 90, 15,
+   131, 13, 201, 95, 96, 53, 194, 233, 7, 225, 140, 36, 103, 30, 69, 142, 8,
+      99, 37, 240, 21, 10, 23,
+   190, 6, 148, 247, 120, 234, 75, 0, 26, 197, 62, 94, 252, 219, 203, 117, 35,
+      11, 32, 57, 177, 33,
+   88, 237, 149, 56, 87, 174, 20, 125, 136, 171, 168, 68, 175, 74, 165, 71,
+      134, 139, 48, 27, 166,
+   77, 146, 158, 231, 83, 111, 229, 122, 60, 211, 133, 230, 220, 105, 92, 41,
+      55, 46, 245, 40, 244,
+   102, 143, 54, 65, 25, 63, 161, 1, 216, 80, 73, 209, 76, 132, 187, 208, 89,
+      18, 169, 200, 196,
+   135, 130, 116, 188, 159, 86, 164, 100, 109, 198, 173, 186, 3, 64, 52, 217,
+      226, 250, 124, 123,
+   5, 202, 38, 147, 118, 126, 255, 82, 85, 212, 207, 206, 59, 227, 47, 16, 58,
+      17, 182, 189, 28, 42,
+   223, 183, 170, 213, 119, 248, 152, 2, 44, 154, 163, 70, 221, 153, 101, 155,
+      167, 43, 172, 9,
+   129, 22, 39, 253, 19, 98, 108, 110, 79, 113, 224, 232, 178, 185, 112, 104,
+      218, 246, 97, 228,
+   251, 34, 242, 193, 238, 210, 144, 12, 191, 179, 162, 241, 81, 51, 145, 235,
+      249, 14, 239, 107,
+   49, 192, 214, 31, 181, 199, 106, 157, 184, 84, 204, 176, 115, 121, 50, 45,
+      127, 4, 150, 254,
+   138, 236, 205, 93, 222, 114, 67, 29, 24, 72, 243, 141, 128, 195, 78, 66,
+      215, 61, 156, 180,
+   151, 160, 137, 91, 90, 15,
+   131, 13, 201, 95, 96, 53, 194, 233, 7, 225, 140, 36, 103, 30, 69, 142, 8,
+      99, 37, 240, 21, 10, 23,
+   190, 6, 148, 247, 120, 234, 75, 0, 26, 197, 62, 94, 252, 219, 203, 117, 35,
+      11, 32, 57, 177, 33,
+   88, 237, 149, 56, 87, 174, 20, 125, 136, 171, 168, 68, 175, 74, 165, 71,
+      134, 139, 48, 27, 166,
+   77, 146, 158, 231, 83, 111, 229, 122, 60, 211, 133, 230, 220, 105, 92, 41,
+      55, 46, 245, 40, 244,
+   102, 143, 54, 65, 25, 63, 161, 1, 216, 80, 73, 209, 76, 132, 187, 208, 89,
+      18, 169, 200, 196,
+   135, 130, 116, 188, 159, 86, 164, 100, 109, 198, 173, 186, 3, 64, 52, 217,
+      226, 250, 124, 123,
+   5, 202, 38, 147, 118, 126, 255, 82, 85, 212, 207, 206, 59, 227, 47, 16, 58,
+      17, 182, 189, 28, 42,
+   223, 183, 170, 213, 119, 248, 152, 2, 44, 154, 163, 70, 221, 153, 101, 155,
+      167, 43, 172, 9,
+   129, 22, 39, 253, 19, 98, 108, 110, 79, 113, 224, 232, 178, 185, 112, 104,
+      218, 246, 97, 228,
+   251, 34, 242, 193, 238, 210, 144, 12, 191, 179, 162, 241, 81, 51, 145, 235,
+      249, 14, 239, 107,
+   49, 192, 214, 31, 181, 199, 106, 157, 184, 84, 204, 176, 115, 121, 50, 45,
+      127, 4, 150, 254,
+   138, 236, 205, 93, 222, 114, 67, 29, 24, 72, 243, 141, 128, 195, 78, 66,
+      215, 61, 156, 180
+};
+
+/*
+ * ---------------------------------------------------------------------
+ */
+
+/*
+ * Helper functions to compute gradients-dot-residualvectors (1D to 4D)
+ * Note that these generate gradients of more than unit length. To make
+ * a close match with the value range of classic Perlin noise, the final
+ * noise values need to be rescaled to fit nicely within [-1,1].
+ * (The simplex noise functions as such also have different scaling.)
+ * Note also that these noise functions are the most practical and useful
+ * signed version of Perlin noise. To return values according to the
+ * RenderMan specification from the SL noise() and pnoise() functions,
+ * the noise values need to be scaled and offset to [0,1], like this:
+ * float SLnoise = (SimplexNoise1234::noise(x,y,z) + 1.0) * 0.5;
+ */
+
+static float
+grad1(int hash, float x)
+{
+   int h = hash & 15;
+   float grad = 1.0f + (h & 7); /* Gradient value 1.0, 2.0, ..., 8.0 */
+   if (h & 8)
+      grad = -grad;             /* Set a random sign for the gradient */
+   return (grad * x);           /* Multiply the gradient with the distance */
+}
+
+static float
+grad2(int hash, float x, float y)
+{
+   int h = hash & 7;            /* Convert low 3 bits of hash code */
+   float u = h < 4 ? x : y;     /* into 8 simple gradient directions, */
+   float v = h < 4 ? y : x;     /* and compute the dot product with (x,y). */
+   return ((h & 1) ? -u : u) + ((h & 2) ? -2.0f * v : 2.0f * v);
+}
+
+static float
+grad3(int hash, float x, float y, float z)
+{
+   int h = hash & 15;           /* Convert low 4 bits of hash code into 12 simple */
+   float u = h < 8 ? x : y;     /* gradient directions, and compute dot product. */
+   float v = h < 4 ? y : h == 12 || h == 14 ? x : z;    /* Fix repeats at h = 12 to 15 */
+   return ((h & 1) ? -u : u) + ((h & 2) ? -v : v);
+}
+
+static float
+grad4(int hash, float x, float y, float z, float t)
+{
+   int h = hash & 31;           /* Convert low 5 bits of hash code into 32 simple */
+   float u = h < 24 ? x : y;    /* gradient directions, and compute dot product. */
+   float v = h < 16 ? y : z;
+   float w = h < 8 ? z : t;
+   return ((h & 1) ? -u : u) + ((h & 2) ? -v : v) + ((h & 4) ? -w : w);
+}
+
+/**
+ * A lookup table to traverse the simplex around a given point in 4D.
+ * Details can be found where this table is used, in the 4D noise method.
+ * TODO: This should not be required, backport it from Bill's GLSL code!
+ */
+static unsigned char simplex[64][4] = {
+   {0, 1, 2, 3}, {0, 1, 3, 2}, {0, 0, 0, 0}, {0, 2, 3, 1},
+   {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0}, {1, 2, 3, 0},
+   {0, 2, 1, 3}, {0, 0, 0, 0}, {0, 3, 1, 2}, {0, 3, 2, 1},
+   {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0}, {1, 3, 2, 0},
+   {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0},
+   {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0},
+   {1, 2, 0, 3}, {0, 0, 0, 0}, {1, 3, 0, 2}, {0, 0, 0, 0},
+   {0, 0, 0, 0}, {0, 0, 0, 0}, {2, 3, 0, 1}, {2, 3, 1, 0},
+   {1, 0, 2, 3}, {1, 0, 3, 2}, {0, 0, 0, 0}, {0, 0, 0, 0},
+   {0, 0, 0, 0}, {2, 0, 3, 1}, {0, 0, 0, 0}, {2, 1, 3, 0},
+   {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0},
+   {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0},
+   {2, 0, 1, 3}, {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0},
+   {3, 0, 1, 2}, {3, 0, 2, 1}, {0, 0, 0, 0}, {3, 1, 2, 0},
+   {2, 1, 0, 3}, {0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0},
+   {3, 1, 0, 2}, {0, 0, 0, 0}, {3, 2, 0, 1}, {3, 2, 1, 0}
+};
+
+
+/** 1D simplex noise */
+GLfloat
+_mesa_noise1(GLfloat x)
+{
+   int i0 = FASTFLOOR(x);
+   int i1 = i0 + 1;
+   float x0 = x - i0;
+   float x1 = x0 - 1.0f;
+   float t1 = 1.0f - x1 * x1;
+   float n0, n1;
+
+   float t0 = 1.0f - x0 * x0;
+/*  if(t0 < 0.0f) t0 = 0.0f; // this never happens for the 1D case */
+   t0 *= t0;
+   n0 = t0 * t0 * grad1(perm[i0 & 0xff], x0);
+
+/*  if(t1 < 0.0f) t1 = 0.0f; // this never happens for the 1D case */
+   t1 *= t1;
+   n1 = t1 * t1 * grad1(perm[i1 & 0xff], x1);
+   /* The maximum value of this noise is 8*(3/4)^4 = 2.53125 */
+   /* A factor of 0.395 would scale to fit exactly within [-1,1], but */
+   /* we want to match PRMan's 1D noise, so we scale it down some more. */
+   return 0.25f * (n0 + n1);
+}
+
+
+/** 2D simplex noise */
+GLfloat
+_mesa_noise2(GLfloat x, GLfloat y)
+{
+#define F2 0.366025403f         /* F2 = 0.5*(sqrt(3.0)-1.0) */
+#define G2 0.211324865f         /* G2 = (3.0-Math.sqrt(3.0))/6.0 */
+
+   float n0, n1, n2;            /* Noise contributions from the three corners */
+
+   /* Skew the input space to determine which simplex cell we're in */
+   float s = (x + y) * F2;      /* Hairy factor for 2D */
+   float xs = x + s;
+   float ys = y + s;
+   int i = FASTFLOOR(xs);
+   int j = FASTFLOOR(ys);
+
+   float t = (float) (i + j) * G2;
+   float X0 = i - t;            /* Unskew the cell origin back to (x,y) space */
+   float Y0 = j - t;
+   float x0 = x - X0;           /* The x,y distances from the cell origin */
+   float y0 = y - Y0;
+
+   float x1, y1, x2, y2;
+   int ii, jj;
+   float t0, t1, t2;
+
+   /* For the 2D case, the simplex shape is an equilateral triangle. */
+   /* Determine which simplex we are in. */
+   int i1, j1;                  /* Offsets for second (middle) corner of simplex in (i,j) coords */
+   if (x0 > y0) {
+      i1 = 1;
+      j1 = 0;
+   }                            /* lower triangle, XY order: (0,0)->(1,0)->(1,1) */
+   else {
+      i1 = 0;
+      j1 = 1;
+   }                            /* upper triangle, YX order: (0,0)->(0,1)->(1,1) */
+
+   /* A step of (1,0) in (i,j) means a step of (1-c,-c) in (x,y), and */
+   /* a step of (0,1) in (i,j) means a step of (-c,1-c) in (x,y), where */
+   /* c = (3-sqrt(3))/6 */
+
+   x1 = x0 - i1 + G2;           /* Offsets for middle corner in (x,y) unskewed coords */
+   y1 = y0 - j1 + G2;
+   x2 = x0 - 1.0f + 2.0f * G2;  /* Offsets for last corner in (x,y) unskewed coords */
+   y2 = y0 - 1.0f + 2.0f * G2;
+
+   /* Wrap the integer indices at 256, to avoid indexing perm[] out of bounds */
+   ii = i % 256;
+   jj = j % 256;
+
+   /* Calculate the contribution from the three corners */
+   t0 = 0.5f - x0 * x0 - y0 * y0;
+   if (t0 < 0.0f)
+      n0 = 0.0f;
+   else {
+      t0 *= t0;
+      n0 = t0 * t0 * grad2(perm[ii + perm[jj]], x0, y0);
+   }
+
+   t1 = 0.5f - x1 * x1 - y1 * y1;
+   if (t1 < 0.0f)
+      n1 = 0.0f;
+   else {
+      t1 *= t1;
+      n1 = t1 * t1 * grad2(perm[ii + i1 + perm[jj + j1]], x1, y1);
+   }
+
+   t2 = 0.5f - x2 * x2 - y2 * y2;
+   if (t2 < 0.0f)
+      n2 = 0.0f;
+   else {
+      t2 *= t2;
+      n2 = t2 * t2 * grad2(perm[ii + 1 + perm[jj + 1]], x2, y2);
+   }
+
+   /* Add contributions from each corner to get the final noise value. */
+   /* The result is scaled to return values in the interval [-1,1]. */
+   return 40.0f * (n0 + n1 + n2);       /* TODO: The scale factor is preliminary! */
+}
+
+
+/** 3D simplex noise */
+GLfloat
+_mesa_noise3(GLfloat x, GLfloat y, GLfloat z)
+{
+/* Simple skewing factors for the 3D case */
+#define F3 0.333333333f
+#define G3 0.166666667f
+
+   float n0, n1, n2, n3;        /* Noise contributions from the four corners */
+
+   /* Skew the input space to determine which simplex cell we're in */
+   float s = (x + y + z) * F3;  /* Very nice and simple skew factor for 3D */
+   float xs = x + s;
+   float ys = y + s;
+   float zs = z + s;
+   int i = FASTFLOOR(xs);
+   int j = FASTFLOOR(ys);
+   int k = FASTFLOOR(zs);
+
+   float t = (float) (i + j + k) * G3;
+   float X0 = i - t;            /* Unskew the cell origin back to (x,y,z) space */
+   float Y0 = j - t;
+   float Z0 = k - t;
+   float x0 = x - X0;           /* The x,y,z distances from the cell origin */
+   float y0 = y - Y0;
+   float z0 = z - Z0;
+
+   float x1, y1, z1, x2, y2, z2, x3, y3, z3;
+   int ii, jj, kk;
+   float t0, t1, t2, t3;
+
+   /* For the 3D case, the simplex shape is a slightly irregular tetrahedron. */
+   /* Determine which simplex we are in. */
+   int i1, j1, k1;              /* Offsets for second corner of simplex in (i,j,k) coords */
+   int i2, j2, k2;              /* Offsets for third corner of simplex in (i,j,k) coords */
+
+/* This code would benefit from a backport from the GLSL version! */
+   if (x0 >= y0) {
+      if (y0 >= z0) {
+         i1 = 1;
+         j1 = 0;
+         k1 = 0;
+         i2 = 1;
+         j2 = 1;
+         k2 = 0;
+      }                         /* X Y Z order */
+      else if (x0 >= z0) {
+         i1 = 1;
+         j1 = 0;
+         k1 = 0;
+         i2 = 1;
+         j2 = 0;
+         k2 = 1;
+      }                         /* X Z Y order */
+      else {
+         i1 = 0;
+         j1 = 0;
+         k1 = 1;
+         i2 = 1;
+         j2 = 0;
+         k2 = 1;
+      }                         /* Z X Y order */
+   }
+   else {                       /* x0<y0 */
+      if (y0 < z0) {
+         i1 = 0;
+         j1 = 0;
+         k1 = 1;
+         i2 = 0;
+         j2 = 1;
+         k2 = 1;
+      }                         /* Z Y X order */
+      else if (x0 < z0) {
+         i1 = 0;
+         j1 = 1;
+         k1 = 0;
+         i2 = 0;
+         j2 = 1;
+         k2 = 1;
+      }                         /* Y Z X order */
+      else {
+         i1 = 0;
+         j1 = 1;
+         k1 = 0;
+         i2 = 1;
+         j2 = 1;
+         k2 = 0;
+      }                         /* Y X Z order */
+   }
+
+   /* A step of (1,0,0) in (i,j,k) means a step of (1-c,-c,-c) in
+    * (x,y,z), a step of (0,1,0) in (i,j,k) means a step of
+    * (-c,1-c,-c) in (x,y,z), and a step of (0,0,1) in (i,j,k) means a
+    * step of (-c,-c,1-c) in (x,y,z), where c = 1/6.
+    */
+
+   x1 = x0 - i1 + G3;         /* Offsets for second corner in (x,y,z) coords */
+   y1 = y0 - j1 + G3;
+   z1 = z0 - k1 + G3;
+   x2 = x0 - i2 + 2.0f * G3;  /* Offsets for third corner in (x,y,z) coords */
+   y2 = y0 - j2 + 2.0f * G3;
+   z2 = z0 - k2 + 2.0f * G3;
+   x3 = x0 - 1.0f + 3.0f * G3;/* Offsets for last corner in (x,y,z) coords */
+   y3 = y0 - 1.0f + 3.0f * G3;
+   z3 = z0 - 1.0f + 3.0f * G3;
+
+   /* Wrap the integer indices at 256 to avoid indexing perm[] out of bounds */
+   ii = i % 256;
+   jj = j % 256;
+   kk = k % 256;
+
+   /* Calculate the contribution from the four corners */
+   t0 = 0.6f - x0 * x0 - y0 * y0 - z0 * z0;
+   if (t0 < 0.0f)
+      n0 = 0.0f;
+   else {
+      t0 *= t0;
+      n0 = t0 * t0 * grad3(perm[ii + perm[jj + perm[kk]]], x0, y0, z0);
+   }
+
+   t1 = 0.6f - x1 * x1 - y1 * y1 - z1 * z1;
+   if (t1 < 0.0f)
+      n1 = 0.0f;
+   else {
+      t1 *= t1;
+      n1 =
+         t1 * t1 * grad3(perm[ii + i1 + perm[jj + j1 + perm[kk + k1]]], x1,
+                         y1, z1);
+   }
+
+   t2 = 0.6f - x2 * x2 - y2 * y2 - z2 * z2;
+   if (t2 < 0.0f)
+      n2 = 0.0f;
+   else {
+      t2 *= t2;
+      n2 =
+         t2 * t2 * grad3(perm[ii + i2 + perm[jj + j2 + perm[kk + k2]]], x2,
+                         y2, z2);
+   }
+
+   t3 = 0.6f - x3 * x3 - y3 * y3 - z3 * z3;
+   if (t3 < 0.0f)
+      n3 = 0.0f;
+   else {
+      t3 *= t3;
+      n3 =
+         t3 * t3 * grad3(perm[ii + 1 + perm[jj + 1 + perm[kk + 1]]], x3, y3,
+                         z3);
+   }
+
+   /* Add contributions from each corner to get the final noise value.
+    * The result is scaled to stay just inside [-1,1]
+    */
+   return 32.0f * (n0 + n1 + n2 + n3);  /* TODO: The scale factor is preliminary! */
+}
+
+
+/** 4D simplex noise */
+GLfloat
+_mesa_noise4(GLfloat x, GLfloat y, GLfloat z, GLfloat w)
+{
+   /* The skewing and unskewing factors are hairy again for the 4D case */
+#define F4 0.309016994f         /* F4 = (Math.sqrt(5.0)-1.0)/4.0 */
+#define G4 0.138196601f         /* G4 = (5.0-Math.sqrt(5.0))/20.0 */
+
+   float n0, n1, n2, n3, n4;    /* Noise contributions from the five corners */
+
+   /* Skew the (x,y,z,w) space to determine which cell of 24 simplices we're in */
+   float s = (x + y + z + w) * F4;      /* Factor for 4D skewing */
+   float xs = x + s;
+   float ys = y + s;
+   float zs = z + s;
+   float ws = w + s;
+   int i = FASTFLOOR(xs);
+   int j = FASTFLOOR(ys);
+   int k = FASTFLOOR(zs);
+   int l = FASTFLOOR(ws);
+
+   float t = (i + j + k + l) * G4;      /* Factor for 4D unskewing */
+   float X0 = i - t;            /* Unskew the cell origin back to (x,y,z,w) space */
+   float Y0 = j - t;
+   float Z0 = k - t;
+   float W0 = l - t;
+
+   float x0 = x - X0;           /* The x,y,z,w distances from the cell origin */
+   float y0 = y - Y0;
+   float z0 = z - Z0;
+   float w0 = w - W0;
+
+   /* For the 4D case, the simplex is a 4D shape I won't even try to describe.
+    * To find out which of the 24 possible simplices we're in, we need to
+    * determine the magnitude ordering of x0, y0, z0 and w0.
+    * The method below is a good way of finding the ordering of x,y,z,w and
+    * then find the correct traversal order for the simplex we're in.
+    * First, six pair-wise comparisons are performed between each possible pair
+    * of the four coordinates, and the results are used to add up binary bits
+    * for an integer index.
+    */
+   int c1 = (x0 > y0) ? 32 : 0;
+   int c2 = (x0 > z0) ? 16 : 0;
+   int c3 = (y0 > z0) ? 8 : 0;
+   int c4 = (x0 > w0) ? 4 : 0;
+   int c5 = (y0 > w0) ? 2 : 0;
+   int c6 = (z0 > w0) ? 1 : 0;
+   int c = c1 + c2 + c3 + c4 + c5 + c6;
+
+   int i1, j1, k1, l1;  /* The integer offsets for the second simplex corner */
+   int i2, j2, k2, l2;  /* The integer offsets for the third simplex corner */
+   int i3, j3, k3, l3;  /* The integer offsets for the fourth simplex corner */
+
+   float x1, y1, z1, w1, x2, y2, z2, w2, x3, y3, z3, w3, x4, y4, z4, w4;
+   int ii, jj, kk, ll;
+   float t0, t1, t2, t3, t4;
+
+   /*
+    * simplex[c] is a 4-vector with the numbers 0, 1, 2 and 3 in some
+    * order.  Many values of c will never occur, since e.g. x>y>z>w
+    * makes x<z, y<w and x<w impossible. Only the 24 indices which
+    * have non-zero entries make any sense.  We use a thresholding to
+    * set the coordinates in turn from the largest magnitude.  The
+    * number 3 in the "simplex" array is at the position of the
+    * largest coordinate.
+    */
+   i1 = simplex[c][0] >= 3 ? 1 : 0;
+   j1 = simplex[c][1] >= 3 ? 1 : 0;
+   k1 = simplex[c][2] >= 3 ? 1 : 0;
+   l1 = simplex[c][3] >= 3 ? 1 : 0;
+   /* The number 2 in the "simplex" array is at the second largest coordinate. */
+   i2 = simplex[c][0] >= 2 ? 1 : 0;
+   j2 = simplex[c][1] >= 2 ? 1 : 0;
+   k2 = simplex[c][2] >= 2 ? 1 : 0;
+   l2 = simplex[c][3] >= 2 ? 1 : 0;
+   /* The number 1 in the "simplex" array is at the second smallest coordinate. */
+   i3 = simplex[c][0] >= 1 ? 1 : 0;
+   j3 = simplex[c][1] >= 1 ? 1 : 0;
+   k3 = simplex[c][2] >= 1 ? 1 : 0;
+   l3 = simplex[c][3] >= 1 ? 1 : 0;
+   /* The fifth corner has all coordinate offsets = 1, so no need to look that up. */
+
+   x1 = x0 - i1 + G4;           /* Offsets for second corner in (x,y,z,w) coords */
+   y1 = y0 - j1 + G4;
+   z1 = z0 - k1 + G4;
+   w1 = w0 - l1 + G4;
+   x2 = x0 - i2 + 2.0f * G4;    /* Offsets for third corner in (x,y,z,w) coords */
+   y2 = y0 - j2 + 2.0f * G4;
+   z2 = z0 - k2 + 2.0f * G4;
+   w2 = w0 - l2 + 2.0f * G4;
+   x3 = x0 - i3 + 3.0f * G4;    /* Offsets for fourth corner in (x,y,z,w) coords */
+   y3 = y0 - j3 + 3.0f * G4;
+   z3 = z0 - k3 + 3.0f * G4;
+   w3 = w0 - l3 + 3.0f * G4;
+   x4 = x0 - 1.0f + 4.0f * G4;  /* Offsets for last corner in (x,y,z,w) coords */
+   y4 = y0 - 1.0f + 4.0f * G4;
+   z4 = z0 - 1.0f + 4.0f * G4;
+   w4 = w0 - 1.0f + 4.0f * G4;
+
+   /* Wrap the integer indices at 256, to avoid indexing perm[] out of bounds */
+   ii = i % 256;
+   jj = j % 256;
+   kk = k % 256;
+   ll = l % 256;
+
+   /* Calculate the contribution from the five corners */
+   t0 = 0.6f - x0 * x0 - y0 * y0 - z0 * z0 - w0 * w0;
+   if (t0 < 0.0f)
+      n0 = 0.0f;
+   else {
+      t0 *= t0;
+      n0 =
+         t0 * t0 * grad4(perm[ii + perm[jj + perm[kk + perm[ll]]]], x0, y0,
+                         z0, w0);
+   }
+
+   t1 = 0.6f - x1 * x1 - y1 * y1 - z1 * z1 - w1 * w1;
+   if (t1 < 0.0f)
+      n1 = 0.0f;
+   else {
+      t1 *= t1;
+      n1 =
+         t1 * t1 *
+         grad4(perm[ii + i1 + perm[jj + j1 + perm[kk + k1 + perm[ll + l1]]]],
+               x1, y1, z1, w1);
+   }
+
+   t2 = 0.6f - x2 * x2 - y2 * y2 - z2 * z2 - w2 * w2;
+   if (t2 < 0.0f)
+      n2 = 0.0f;
+   else {
+      t2 *= t2;
+      n2 =
+         t2 * t2 *
+         grad4(perm[ii + i2 + perm[jj + j2 + perm[kk + k2 + perm[ll + l2]]]],
+               x2, y2, z2, w2);
+   }
+
+   t3 = 0.6f - x3 * x3 - y3 * y3 - z3 * z3 - w3 * w3;
+   if (t3 < 0.0f)
+      n3 = 0.0f;
+   else {
+      t3 *= t3;
+      n3 =
+         t3 * t3 *
+         grad4(perm[ii + i3 + perm[jj + j3 + perm[kk + k3 + perm[ll + l3]]]],
+               x3, y3, z3, w3);
+   }
+
+   t4 = 0.6f - x4 * x4 - y4 * y4 - z4 * z4 - w4 * w4;
+   if (t4 < 0.0f)
+      n4 = 0.0f;
+   else {
+      t4 *= t4;
+      n4 =
+         t4 * t4 *
+         grad4(perm[ii + 1 + perm[jj + 1 + perm[kk + 1 + perm[ll + 1]]]], x4,
+               y4, z4, w4);
+   }
+
+   /* Sum up and scale the result to cover the range [-1,1] */
+   return 27.0f * (n0 + n1 + n2 + n3 + n4);     /* TODO: The scale factor is preliminary! */
+}
diff --git a/src/mesa/shader/slang/slang_library_noise.h b/src/mesa/shader/prog_noise.h
index da7367c1ce..c4779479f9 100644
--- a/src/mesa/shader/slang/slang_library_noise.h
+++ b/src/mesa/shader/prog_noise.h
@@ -22,21 +22,13 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
-#if !defined SLANG_LIBRARY_NOISE_H
-#define SLANG_LIBRARY_NOISE_H
+#ifndef PROG_NOISE
+#define PROG_NOISE
 
-#if defined __cplusplus
-extern "C" {
-#endif
-
-GLfloat _slang_library_noise1 (GLfloat);
-GLfloat _slang_library_noise2 (GLfloat, GLfloat);
-GLfloat _slang_library_noise3 (GLfloat, GLfloat, GLfloat);
-GLfloat _slang_library_noise4 (GLfloat, GLfloat, GLfloat, GLfloat);
-
-#ifdef __cplusplus
-}
-#endif
+extern GLfloat _mesa_noise1(GLfloat);
+extern GLfloat _mesa_noise2(GLfloat, GLfloat);
+extern GLfloat _mesa_noise3(GLfloat, GLfloat, GLfloat);
+extern GLfloat _mesa_noise4(GLfloat, GLfloat, GLfloat, GLfloat);
 
 #endif
 
diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c
index 29c5b33f6c..8ae961241f 100644
--- a/src/mesa/shader/prog_parameter.c
+++ b/src/mesa/shader/prog_parameter.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.3
+ * Version:  7.3
  *
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -80,7 +80,8 @@ GLint
 _mesa_add_parameter(struct gl_program_parameter_list *paramList,
                     enum register_file type, const char *name,
                     GLuint size, GLenum datatype, const GLfloat *values,
-                    const gl_state_index state[STATE_LENGTH])
+                    const gl_state_index state[STATE_LENGTH],
+                    GLbitfield flags)
 {
    const GLuint oldNum = paramList->NumParameters;
    const GLuint sz4 = (size + 3) / 4; /* no. of new param slots needed */
@@ -125,6 +126,7 @@ _mesa_add_parameter(struct gl_program_parameter_list *paramList,
          p->Type = type;
          p->Size = size;
          p->DataType = datatype;
+         p->Flags = flags;
          if (values) {
             COPY_4V(paramList->ParameterValues[oldNum + i], values);
             values += 4;
@@ -156,7 +158,7 @@ _mesa_add_named_parameter(struct gl_program_parameter_list *paramList,
                           const char *name, const GLfloat values[4])
 {
    return _mesa_add_parameter(paramList, PROGRAM_NAMED_PARAM, name,
-                              4, GL_NONE, values, NULL);
+                              4, GL_NONE, values, NULL, 0x0);
                               
 }
 
@@ -185,9 +187,8 @@ _mesa_add_named_constant(struct gl_program_parameter_list *paramList,
       return pos;
    }
 #endif
-   size = 4; /** XXX fix */
    return _mesa_add_parameter(paramList, PROGRAM_CONSTANT, name,
-                              size, GL_NONE, values, NULL);
+                              size, GL_NONE, values, NULL, 0x0);
 }
 
 
@@ -239,7 +240,7 @@ _mesa_add_unnamed_constant(struct gl_program_parameter_list *paramList,
 
    /* add a new parameter to store this constant */
    pos = _mesa_add_parameter(paramList, PROGRAM_CONSTANT, NULL,
-                             size, GL_NONE, values, NULL);
+                             size, GL_NONE, values, NULL, 0x0);
    if (pos >= 0 && swizzleOut) {
       if (size == 1)
          *swizzleOut = SWIZZLE_XXXX;
@@ -273,7 +274,7 @@ _mesa_add_uniform(struct gl_program_parameter_list *paramList,
    }
    else {
       i = _mesa_add_parameter(paramList, PROGRAM_UNIFORM, name,
-                              size, datatype, values, NULL);
+                              size, datatype, values, NULL, 0x0);
       return i;
    }
 }
@@ -289,7 +290,8 @@ _mesa_use_uniform(struct gl_program_parameter_list *paramList,
    GLuint i;
    for (i = 0; i < paramList->NumParameters; i++) {
       struct gl_program_parameter *p = paramList->Parameters + i;
-      if (p->Type == PROGRAM_UNIFORM && _mesa_strcmp(p->Name, name) == 0) {
+      if ((p->Type == PROGRAM_UNIFORM || p->Type == PROGRAM_SAMPLER) &&
+          _mesa_strcmp(p->Name, name) == 0) {
          p->Used = GL_TRUE;
          /* Note that large uniforms may occupy several slots so we're
           * not done searching yet.
@@ -328,7 +330,7 @@ _mesa_add_sampler(struct gl_program_parameter_list *paramList,
       }
       value = (GLfloat) numSamplers;
       (void) _mesa_add_parameter(paramList, PROGRAM_SAMPLER, name,
-                                 size, datatype, &value, NULL);
+                                 size, datatype, &value, NULL, 0x0);
       return numSamplers;
    }
 }
@@ -339,7 +341,7 @@ _mesa_add_sampler(struct gl_program_parameter_list *paramList,
  */
 GLint
 _mesa_add_varying(struct gl_program_parameter_list *paramList,
-                  const char *name, GLuint size)
+                  const char *name, GLuint size, GLbitfield flags)
 {
    GLint i = _mesa_lookup_parameter_index(paramList, -1, name);
    if (i >= 0 && paramList->Parameters[i].Type == PROGRAM_VARYING) {
@@ -349,7 +351,7 @@ _mesa_add_varying(struct gl_program_parameter_list *paramList,
    else {
       /*assert(size == 4);*/
       i = _mesa_add_parameter(paramList, PROGRAM_VARYING, name,
-                              size, GL_NONE, NULL, NULL);
+                              size, GL_NONE, NULL, NULL, flags);
       return i;
    }
 }
@@ -378,7 +380,7 @@ _mesa_add_attribute(struct gl_program_parameter_list *paramList,
       if (size < 0)
          size = 4;
       i = _mesa_add_parameter(paramList, PROGRAM_INPUT, name,
-                              size, datatype, NULL, state);
+                              size, datatype, NULL, state, 0x0);
    }
    return i;
 }
@@ -422,7 +424,7 @@ _mesa_add_state_reference(struct gl_program_parameter_list *paramList,
                           const gl_state_index stateTokens[STATE_LENGTH])
 {
    const GLuint size = 4; /* XXX fix */
-   const char *name;
+   char *name;
    GLint index;
 
    /* Check if the state reference is already in the list */
@@ -445,11 +447,11 @@ _mesa_add_state_reference(struct gl_program_parameter_list *paramList,
    name = _mesa_program_state_string(stateTokens);
    index = _mesa_add_parameter(paramList, PROGRAM_STATE_VAR, name,
                                size, GL_NONE,
-                               NULL, (gl_state_index *) stateTokens);
+                               NULL, (gl_state_index *) stateTokens, 0x0);
    paramList->StateFlags |= _mesa_program_state_flags(stateTokens);
 
    /* free name string here since we duplicated it in add_parameter() */
-   _mesa_free((void *) name);
+   _mesa_free(name);
 
    return index;
 }
@@ -616,10 +618,11 @@ _mesa_clone_parameter_list(const struct gl_program_parameter_list *list)
       struct gl_program_parameter *pCopy;
       GLuint size = MIN2(p->Size, 4);
       GLint j = _mesa_add_parameter(clone, p->Type, p->Name, size, p->DataType,
-                                    list->ParameterValues[i], NULL);
+                                    list->ParameterValues[i], NULL, 0x0);
       ASSERT(j >= 0);
       pCopy = clone->Parameters + j;
       pCopy->Used = p->Used;
+      pCopy->Flags = p->Flags;
       /* copy state indexes */
       if (p->Type == PROGRAM_STATE_VAR) {
          GLint k;
@@ -657,7 +660,8 @@ _mesa_combine_parameter_lists(const struct gl_program_parameter_list *listA,
             _mesa_add_parameter(list, param->Type, param->Name, param->Size,
                                 param->DataType,
                                 listB->ParameterValues[i],
-                                param->StateIndexes);
+                                param->StateIndexes,
+                                param->Flags);
          }
       }
    }
diff --git a/src/mesa/shader/prog_parameter.h b/src/mesa/shader/prog_parameter.h
index 0c35198de7..200f2c0045 100644
--- a/src/mesa/shader/prog_parameter.h
+++ b/src/mesa/shader/prog_parameter.h
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.3
+ * Version:  7.3
  *
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -36,11 +36,20 @@
 
 
 /**
+ * Program parameter flags
+ */
+/*@{*/
+#define PROG_PARAM_BIT_CENTROID   0x1  /**< for varying vars (GLSL 1.20) */
+#define PROG_PARAM_BIT_INVARIANT  0x2  /**< for varying vars (GLSL 1.20) */
+#define PROG_PARAM_BIT_FLAT       0x4  /**< for varying vars (GLSL 1.30) */
+#define PROG_PARAM_BIT_LINEAR     0x8  /**< for varying vars (GLSL 1.30) */
+/*@}*/
+
+
+
+/**
  * Program parameter.
- * Used for NV_fragment_program for "DEFINE"d constants and "DECLARE"d
- * parameters.
- * Also used by ARB_vertex/fragment_programs for state variables, etc.
- * Used by shaders for uniforms, constants, varying vars, etc.
+ * Used by shaders/programs for uniforms, constants, varying vars, etc.
  */
 struct gl_program_parameter
 {
@@ -50,6 +59,7 @@ struct gl_program_parameter
    GLuint Size;             /**< Number of components (1..4) */
    GLboolean Used;          /**< Helper flag for GLSL uniform tracking */
    GLboolean Initialized;   /**< Has the ParameterValue[] been set? */
+   GLbitfield Flags;        /**< Bitmask of PROG_PARAM_*_BIT */
    /**
     * A sequence of STATE_* tokens and integers to identify GL state.
     */
@@ -94,7 +104,8 @@ extern GLint
 _mesa_add_parameter(struct gl_program_parameter_list *paramList,
                     enum register_file type, const char *name,
                     GLuint size, GLenum datatype, const GLfloat *values,
-                    const gl_state_index state[STATE_LENGTH]);
+                    const gl_state_index state[STATE_LENGTH],
+                    GLbitfield flags);
 
 extern GLint
 _mesa_add_named_parameter(struct gl_program_parameter_list *paramList,
@@ -125,7 +136,7 @@ _mesa_add_sampler(struct gl_program_parameter_list *paramList,
 
 extern GLint
 _mesa_add_varying(struct gl_program_parameter_list *paramList,
-                  const char *name, GLuint size);
+                  const char *name, GLuint size, GLbitfield flags);
 
 extern GLint
 _mesa_add_attribute(struct gl_program_parameter_list *paramList,
diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c
index baa4ef85e9..ce48767a87 100644
--- a/src/mesa/shader/prog_print.c
+++ b/src/mesa/shader/prog_print.c
@@ -3,6 +3,7 @@
  * Version:  7.3
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -71,6 +72,8 @@ file_string(enum register_file f, gl_prog_print_mode mode)
       return "ADDR";
    case PROGRAM_SAMPLER:
       return "SAMPLER";
+   case PROGRAM_UNDEFINED:
+      return "UNDEFINED";
    default:
       return "Unknown program file!";
    }
@@ -215,46 +218,48 @@ reg_string(enum register_file f, GLint index, gl_prog_print_mode mode,
    switch (mode) {
    case PROG_PRINT_DEBUG:
       if (relAddr)
-         sprintf(str, "%s[ADDR+%d]", file_string(f, mode), index);
+         _mesa_sprintf(str, "%s[ADDR+%d]", file_string(f, mode), index);
       else
-         sprintf(str, "%s[%d]", file_string(f, mode), index);
+         _mesa_sprintf(str, "%s[%d]", file_string(f, mode), index);
       break;
 
    case PROG_PRINT_ARB:
       switch (f) {
       case PROGRAM_INPUT:
-         sprintf(str, "%s", arb_input_attrib_string(index, prog->Target));
+         _mesa_sprintf(str, "%s", arb_input_attrib_string(index, prog->Target));
          break;
       case PROGRAM_OUTPUT:
-         sprintf(str, "%s", arb_output_attrib_string(index, prog->Target));
+         _mesa_sprintf(str, "%s", arb_output_attrib_string(index, prog->Target));
          break;
       case PROGRAM_TEMPORARY:
-         sprintf(str, "temp%d", index);
+         _mesa_sprintf(str, "temp%d", index);
          break;
       case PROGRAM_ENV_PARAM:
-         sprintf(str, "program.env[%d]", index);
+         _mesa_sprintf(str, "program.env[%d]", index);
          break;
       case PROGRAM_LOCAL_PARAM:
-         sprintf(str, "program.local[%d]", index);
+         _mesa_sprintf(str, "program.local[%d]", index);
          break;
       case PROGRAM_VARYING: /* extension */
-         sprintf(str, "varying[%d]", index);
+         _mesa_sprintf(str, "varying[%d]", index);
          break;
       case PROGRAM_CONSTANT: /* extension */
-         sprintf(str, "constant[%d]", index);
+         _mesa_sprintf(str, "constant[%d]", index);
          break;
       case PROGRAM_UNIFORM: /* extension */
-         sprintf(str, "uniform[%d]", index);
+         _mesa_sprintf(str, "uniform[%d]", index);
          break;
       case PROGRAM_STATE_VAR:
          {
             struct gl_program_parameter *param
                = prog->Parameters->Parameters + index;
-            sprintf(str, _mesa_program_state_string(param->StateIndexes));
+            char *state = _mesa_program_state_string(param->StateIndexes);
+            _mesa_sprintf(str, state);
+            _mesa_free(state);
          }
          break;
       case PROGRAM_ADDRESS:
-         sprintf(str, "A%d", index);
+         _mesa_sprintf(str, "A%d", index);
          break;
       default:
          _mesa_problem(NULL, "bad file in reg_string()");
@@ -265,30 +270,30 @@ reg_string(enum register_file f, GLint index, gl_prog_print_mode mode,
       switch (f) {
       case PROGRAM_INPUT:
          if (prog->Target == GL_VERTEX_PROGRAM_ARB)
-            sprintf(str, "v[%d]", index);
+            _mesa_sprintf(str, "v[%d]", index);
          else
-            sprintf(str, "f[%d]", index);
+            _mesa_sprintf(str, "f[%d]", index);
          break;
       case PROGRAM_OUTPUT:
-         sprintf(str, "o[%d]", index);
+         _mesa_sprintf(str, "o[%d]", index);
          break;
       case PROGRAM_TEMPORARY:
-         sprintf(str, "R%d", index);
+         _mesa_sprintf(str, "R%d", index);
          break;
       case PROGRAM_ENV_PARAM:
-         sprintf(str, "c[%d]", index);
+         _mesa_sprintf(str, "c[%d]", index);
          break;
       case PROGRAM_VARYING: /* extension */
-         sprintf(str, "varying[%d]", index);
+         _mesa_sprintf(str, "varying[%d]", index);
          break;
       case PROGRAM_UNIFORM: /* extension */
-         sprintf(str, "uniform[%d]", index);
+         _mesa_sprintf(str, "uniform[%d]", index);
          break;
       case PROGRAM_CONSTANT: /* extension */
-         sprintf(str, "constant[%d]", index);
+         _mesa_sprintf(str, "constant[%d]", index);
          break;
       case PROGRAM_STATE_VAR: /* extension */
-         sprintf(str, "state[%d]", index);
+         _mesa_sprintf(str, "state[%d]", index);
          break;
       default:
          _mesa_problem(NULL, "bad file in reg_string()");
@@ -356,6 +361,19 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended)
 }
 
 
+void
+_mesa_print_swizzle(GLuint swizzle)
+{
+   if (swizzle == SWIZZLE_XYZW) {
+      _mesa_printf(".xyzw\n");
+   }
+   else {
+      const char *s = _mesa_swizzle_string(swizzle, 0, 0);
+      _mesa_printf("%s\n", s);
+   }
+}
+
+
 const char *
 _mesa_writemask_string(GLuint writeMask)
 {
@@ -399,90 +417,98 @@ _mesa_condcode_string(GLuint condcode)
 
 
 static void
-print_dst_reg(const struct prog_dst_register *dstReg, gl_prog_print_mode mode,
-              const struct gl_program *prog)
+fprint_dst_reg(FILE * f,
+               const struct prog_dst_register *dstReg,
+               gl_prog_print_mode mode,
+               const struct gl_program *prog)
 {
-   _mesa_printf("%s%s",
-                reg_string((enum register_file) dstReg->File,
-                           dstReg->Index, mode, dstReg->RelAddr, prog),
-                _mesa_writemask_string(dstReg->WriteMask));
+   _mesa_fprintf(f, "%s%s",
+                 reg_string((enum register_file) dstReg->File,
+                            dstReg->Index, mode, dstReg->RelAddr, prog),
+                 _mesa_writemask_string(dstReg->WriteMask));
 
    if (dstReg->CondMask != COND_TR) {
-      _mesa_printf(" (%s.%s)",
-                   _mesa_condcode_string(dstReg->CondMask),
-                   _mesa_swizzle_string(dstReg->CondSwizzle, GL_FALSE, GL_FALSE));
+      _mesa_fprintf(f, " (%s.%s)",
+                    _mesa_condcode_string(dstReg->CondMask),
+                    _mesa_swizzle_string(dstReg->CondSwizzle,
+                                         GL_FALSE, GL_FALSE));
    }
 
 #if 0
-   _mesa_printf("%s[%d]%s",
+   _mesa_fprintf(f, "%s[%d]%s",
                 file_string((enum register_file) dstReg->File, mode),
                 dstReg->Index,
                 _mesa_writemask_string(dstReg->WriteMask));
 #endif
 }
 
+
 static void
-print_src_reg(const struct prog_src_register *srcReg, gl_prog_print_mode mode,
-              const struct gl_program *prog)
+fprint_src_reg(FILE *f,
+               const struct prog_src_register *srcReg, 
+               gl_prog_print_mode mode,
+               const struct gl_program *prog)
 {
-   _mesa_printf("%s%s",
-                reg_string((enum register_file) srcReg->File,
-                           srcReg->Index, mode, srcReg->RelAddr, prog),
-                _mesa_swizzle_string(srcReg->Swizzle,
-                                     srcReg->NegateBase, GL_FALSE));
+   _mesa_fprintf(f, "%s%s",
+                 reg_string((enum register_file) srcReg->File,
+                            srcReg->Index, mode, srcReg->RelAddr, prog),
+                 _mesa_swizzle_string(srcReg->Swizzle,
+                                      srcReg->NegateBase, GL_FALSE));
 #if 0
-   _mesa_printf("%s[%d]%s",
-                file_string((enum register_file) srcReg->File, mode),
-                srcReg->Index,
-                _mesa_swizzle_string(srcReg->Swizzle,
-                               srcReg->NegateBase, GL_FALSE));
+   _mesa_fprintf(f, "%s[%d]%s",
+                 file_string((enum register_file) srcReg->File, mode),
+                 srcReg->Index,
+                 _mesa_swizzle_string(srcReg->Swizzle,
+                                      srcReg->NegateBase, GL_FALSE));
 #endif
 }
 
+
 static void
-print_comment(const struct prog_instruction *inst)
+fprint_comment(FILE *f, const struct prog_instruction *inst)
 {
    if (inst->Comment)
-      _mesa_printf(";  # %s\n", inst->Comment);
+      _mesa_fprintf(f, ";  # %s\n", inst->Comment);
    else
-      _mesa_printf(";\n");
+      _mesa_fprintf(f, ";\n");
 }
 
 
 static void
-print_alu_instruction(const struct prog_instruction *inst,
-                      const char *opcode_string, GLuint numRegs,
-                      gl_prog_print_mode mode,
-                      const struct gl_program *prog)
+fprint_alu_instruction(FILE *f,
+                       const struct prog_instruction *inst,
+                       const char *opcode_string, GLuint numRegs,
+                       gl_prog_print_mode mode,
+                       const struct gl_program *prog)
 {
    GLuint j;
 
-   _mesa_printf("%s", opcode_string);
+   _mesa_fprintf(f, "%s", opcode_string);
    if (inst->CondUpdate)
-      _mesa_printf(".C");
+      _mesa_fprintf(f, ".C");
 
    /* frag prog only */
    if (inst->SaturateMode == SATURATE_ZERO_ONE)
-      _mesa_printf("_SAT");
+      _mesa_fprintf(f, "_SAT");
 
-   _mesa_printf(" ");
+   _mesa_fprintf(f, " ");
    if (inst->DstReg.File != PROGRAM_UNDEFINED) {
-      print_dst_reg(&inst->DstReg, mode, prog);
+      fprint_dst_reg(f, &inst->DstReg, mode, prog);
    }
    else {
-      _mesa_printf(" ???");
+      _mesa_fprintf(f, " ???");
    }
 
    if (numRegs > 0)
-      _mesa_printf(", ");
+      _mesa_fprintf(f, ", ");
 
    for (j = 0; j < numRegs; j++) {
-      print_src_reg(inst->SrcReg + j, mode, prog);
+      fprint_src_reg(f, inst->SrcReg + j, mode, prog);
       if (j + 1 < numRegs)
-	 _mesa_printf(", ");
+	 _mesa_fprintf(f, ", ");
    }
 
-   print_comment(inst);
+   fprint_comment(f, inst);
 }
 
 
@@ -490,23 +516,18 @@ void
 _mesa_print_alu_instruction(const struct prog_instruction *inst,
                             const char *opcode_string, GLuint numRegs)
 {
-   print_alu_instruction(inst, opcode_string, numRegs, PROG_PRINT_DEBUG, NULL);
-}
-
-
-void
-_mesa_print_instruction(const struct prog_instruction *inst)
-{
-   /* note: 4th param should be ignored for PROG_PRINT_DEBUG */
-   _mesa_print_instruction_opt(inst, 0, PROG_PRINT_DEBUG, NULL);
+   fprint_alu_instruction(stdout, inst, opcode_string,
+                          numRegs, PROG_PRINT_DEBUG, NULL);
 }
 
 
 /**
  * Print a single vertex/fragment program instruction.
  */
-GLint
-_mesa_print_instruction_opt(const struct prog_instruction *inst, GLint indent,
+static GLint
+_mesa_fprint_instruction_opt(FILE *f,
+                            const struct prog_instruction *inst,
+                            GLint indent,
                             gl_prog_print_mode mode,
                             const struct gl_program *prog)
 {
@@ -519,15 +540,15 @@ _mesa_print_instruction_opt(const struct prog_instruction *inst, GLint indent,
       indent -= 3;
    }
    for (i = 0; i < indent; i++) {
-      _mesa_printf(" ");
+      _mesa_fprintf(f, " ");
    }
 
    switch (inst->Opcode) {
    case OPCODE_PRINT:
-      _mesa_printf("PRINT '%s'", inst->Data);
+      _mesa_fprintf(f, "PRINT '%s'", inst->Data);
       if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
-         _mesa_printf(", ");
-         _mesa_printf("%s[%d]%s",
+         _mesa_fprintf(f, ", ");
+         _mesa_fprintf(f, "%s[%d]%s",
                       file_string((enum register_file) inst->SrcReg[0].File,
                                   mode),
                       inst->SrcReg[0].Index,
@@ -535,259 +556,295 @@ _mesa_print_instruction_opt(const struct prog_instruction *inst, GLint indent,
                                            inst->SrcReg[0].NegateBase, GL_FALSE));
       }
       if (inst->Comment)
-         _mesa_printf("  # %s", inst->Comment);
-      print_comment(inst);
+         _mesa_fprintf(f, "  # %s", inst->Comment);
+      fprint_comment(f, inst);
       break;
    case OPCODE_SWZ:
-      _mesa_printf("SWZ");
+      _mesa_fprintf(f, "SWZ");
       if (inst->SaturateMode == SATURATE_ZERO_ONE)
-         _mesa_printf("_SAT");
-      _mesa_printf(" ");
-      print_dst_reg(&inst->DstReg, mode, prog);
-      _mesa_printf(", %s[%d], %s",
+         _mesa_fprintf(f, "_SAT");
+      _mesa_fprintf(f, " ");
+      fprint_dst_reg(f, &inst->DstReg, mode, prog);
+      _mesa_fprintf(f, ", %s[%d], %s",
                    file_string((enum register_file) inst->SrcReg[0].File,
                                mode),
                    inst->SrcReg[0].Index,
                    _mesa_swizzle_string(inst->SrcReg[0].Swizzle,
                                         inst->SrcReg[0].NegateBase, GL_TRUE));
-      print_comment(inst);
+      fprint_comment(f, inst);
       break;
    case OPCODE_TEX:
    case OPCODE_TXP:
    case OPCODE_TXL:
    case OPCODE_TXB:
-      _mesa_printf("%s", _mesa_opcode_string(inst->Opcode));
+      _mesa_fprintf(f, "%s", _mesa_opcode_string(inst->Opcode));
       if (inst->SaturateMode == SATURATE_ZERO_ONE)
-         _mesa_printf("_SAT");
-      _mesa_printf(" ");
-      print_dst_reg(&inst->DstReg, mode, prog);
-      _mesa_printf(", ");
-      print_src_reg(&inst->SrcReg[0], mode, prog);
-      _mesa_printf(", texture[%d], ", inst->TexSrcUnit);
+         _mesa_fprintf(f, "_SAT");
+      _mesa_fprintf(f, " ");
+      fprint_dst_reg(f, &inst->DstReg, mode, prog);
+      _mesa_fprintf(f, ", ");
+      fprint_src_reg(f, &inst->SrcReg[0], mode, prog);
+      _mesa_fprintf(f, ", texture[%d], ", inst->TexSrcUnit);
       switch (inst->TexSrcTarget) {
-      case TEXTURE_1D_INDEX:   _mesa_printf("1D");    break;
-      case TEXTURE_2D_INDEX:   _mesa_printf("2D");    break;
-      case TEXTURE_3D_INDEX:   _mesa_printf("3D");    break;
-      case TEXTURE_CUBE_INDEX: _mesa_printf("CUBE");  break;
-      case TEXTURE_RECT_INDEX: _mesa_printf("RECT");  break;
+      case TEXTURE_1D_INDEX:   _mesa_fprintf(f, "1D");    break;
+      case TEXTURE_2D_INDEX:   _mesa_fprintf(f, "2D");    break;
+      case TEXTURE_3D_INDEX:   _mesa_fprintf(f, "3D");    break;
+      case TEXTURE_CUBE_INDEX: _mesa_fprintf(f, "CUBE");  break;
+      case TEXTURE_RECT_INDEX: _mesa_fprintf(f, "RECT");  break;
       default:
          ;
       }
-      print_comment(inst);
+      fprint_comment(f, inst);
       break;
 
    case OPCODE_KIL:
-      _mesa_printf("%s", _mesa_opcode_string(inst->Opcode));
-      _mesa_printf(" ");
-      print_src_reg(&inst->SrcReg[0], mode, prog);
-      print_comment(inst);
+      _mesa_fprintf(f, "%s", _mesa_opcode_string(inst->Opcode));
+      _mesa_fprintf(f, " ");
+      fprint_src_reg(f, &inst->SrcReg[0], mode, prog);
+      fprint_comment(f, inst);
       break;
    case OPCODE_KIL_NV:
-      _mesa_printf("%s", _mesa_opcode_string(inst->Opcode));
-      _mesa_printf(" ");
-      _mesa_printf("%s.%s",
+      _mesa_fprintf(f, "%s", _mesa_opcode_string(inst->Opcode));
+      _mesa_fprintf(f, " ");
+      _mesa_fprintf(f, "%s.%s",
                    _mesa_condcode_string(inst->DstReg.CondMask),
                    _mesa_swizzle_string(inst->DstReg.CondSwizzle,
                                         GL_FALSE, GL_FALSE));
-      print_comment(inst);
+      fprint_comment(f, inst);
       break;
 
    case OPCODE_ARL:
-      _mesa_printf("ARL ");
-      print_dst_reg(&inst->DstReg, mode, prog);
-      _mesa_printf(", ");
-      print_src_reg(&inst->SrcReg[0], mode, prog);
-      print_comment(inst);
+      _mesa_fprintf(f, "ARL ");
+      fprint_dst_reg(f, &inst->DstReg, mode, prog);
+      _mesa_fprintf(f, ", ");
+      fprint_src_reg(f, &inst->SrcReg[0], mode, prog);
+      fprint_comment(f, inst);
       break;
    case OPCODE_BRA:
-      _mesa_printf("BRA %d (%s%s)",
+      _mesa_fprintf(f, "BRA %d (%s%s)",
                    inst->BranchTarget,
                    _mesa_condcode_string(inst->DstReg.CondMask),
                    _mesa_swizzle_string(inst->DstReg.CondSwizzle, 0, GL_FALSE));
-      print_comment(inst);
+      fprint_comment(f, inst);
       break;
    case OPCODE_IF:
       if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
          /* Use ordinary register */
-         _mesa_printf("IF ");
-         print_src_reg(&inst->SrcReg[0], mode, prog);
-         _mesa_printf("; ");
+         _mesa_fprintf(f, "IF ");
+         fprint_src_reg(f, &inst->SrcReg[0], mode, prog);
+         _mesa_fprintf(f, "; ");
       }
       else {
          /* Use cond codes */
-         _mesa_printf("IF (%s%s);",
+         _mesa_fprintf(f, "IF (%s%s);",
                       _mesa_condcode_string(inst->DstReg.CondMask),
                       _mesa_swizzle_string(inst->DstReg.CondSwizzle,
                                            0, GL_FALSE));
       }
-      _mesa_printf(" # (if false, goto %d)", inst->BranchTarget);
-      print_comment(inst);
+      _mesa_fprintf(f, " # (if false, goto %d)", inst->BranchTarget);
+      fprint_comment(f, inst);
       return indent + 3;
    case OPCODE_ELSE:
-      _mesa_printf("ELSE; # (goto %d)\n", inst->BranchTarget);
+      _mesa_fprintf(f, "ELSE; # (goto %d)\n", inst->BranchTarget);
       return indent + 3;
    case OPCODE_ENDIF:
-      _mesa_printf("ENDIF;\n");
+      _mesa_fprintf(f, "ENDIF;\n");
       break;
    case OPCODE_BGNLOOP:
-      _mesa_printf("BGNLOOP; # (end at %d)\n", inst->BranchTarget);
+      _mesa_fprintf(f, "BGNLOOP; # (end at %d)\n", inst->BranchTarget);
       return indent + 3;
    case OPCODE_ENDLOOP:
-      _mesa_printf("ENDLOOP; # (goto %d)\n", inst->BranchTarget);
+      _mesa_fprintf(f, "ENDLOOP; # (goto %d)\n", inst->BranchTarget);
       break;
    case OPCODE_BRK:
    case OPCODE_CONT:
-      _mesa_printf("%s (%s%s); # (goto %d)",
+      _mesa_fprintf(f, "%s (%s%s); # (goto %d)",
                    _mesa_opcode_string(inst->Opcode),
                    _mesa_condcode_string(inst->DstReg.CondMask),
                    _mesa_swizzle_string(inst->DstReg.CondSwizzle, 0, GL_FALSE),
                    inst->BranchTarget);
-      print_comment(inst);
+      fprint_comment(f, inst);
       break;
 
    case OPCODE_BGNSUB:
       if (mode == PROG_PRINT_NV) {
-         _mesa_printf("%s:\n", inst->Comment); /* comment is label */
+         _mesa_fprintf(f, "%s:\n", inst->Comment); /* comment is label */
          return indent;
       }
       else {
-         _mesa_printf("BGNSUB");
-         print_comment(inst);
+         _mesa_fprintf(f, "BGNSUB");
+         fprint_comment(f, inst);
          return indent + 3;
       }
    case OPCODE_ENDSUB:
       if (mode == PROG_PRINT_DEBUG) {
-         _mesa_printf("ENDSUB");
-         print_comment(inst);
+         _mesa_fprintf(f, "ENDSUB");
+         fprint_comment(f, inst);
       }
       break;
    case OPCODE_CAL:
       if (mode == PROG_PRINT_NV) {
-         _mesa_printf("CAL %s;  # (goto %d)\n", inst->Comment, inst->BranchTarget);
+         _mesa_fprintf(f, "CAL %s;  # (goto %d)\n", inst->Comment, inst->BranchTarget);
       }
       else {
-         _mesa_printf("CAL %u", inst->BranchTarget);
-         print_comment(inst);
+         _mesa_fprintf(f, "CAL %u", inst->BranchTarget);
+         fprint_comment(f, inst);
       }
       break;
    case OPCODE_RET:
-      _mesa_printf("RET (%s%s)",
+      _mesa_fprintf(f, "RET (%s%s)",
                    _mesa_condcode_string(inst->DstReg.CondMask),
                    _mesa_swizzle_string(inst->DstReg.CondSwizzle, 0, GL_FALSE));
-      print_comment(inst);
+      fprint_comment(f, inst);
       break;
 
    case OPCODE_END:
-      _mesa_printf("END\n");
+      _mesa_fprintf(f, "END\n");
       break;
    case OPCODE_NOP:
       if (mode == PROG_PRINT_DEBUG) {
-         _mesa_printf("NOP");
-         print_comment(inst);
+         _mesa_fprintf(f, "NOP");
+         fprint_comment(f, inst);
       }
       else if (inst->Comment) {
          /* ARB/NV extensions don't have NOP instruction */
-         _mesa_printf("# %s\n", inst->Comment);
+         _mesa_fprintf(f, "# %s\n", inst->Comment);
       }
       break;
    /* XXX may need other special-case instructions */
    default:
       /* typical alu instruction */
-      print_alu_instruction(inst,
-                            _mesa_opcode_string(inst->Opcode),
-                            _mesa_num_inst_src_regs(inst->Opcode),
-                            mode, prog);
+      fprint_alu_instruction(f, inst,
+                             _mesa_opcode_string(inst->Opcode),
+                             _mesa_num_inst_src_regs(inst->Opcode),
+                             mode, prog);
       break;
    }
    return indent;
 }
 
 
-/**
- * Print program to stdout, default options.
- */
+GLint
+_mesa_print_instruction_opt(const struct prog_instruction *inst,
+                            GLint indent,
+                            gl_prog_print_mode mode,
+                            const struct gl_program *prog)
+{
+   return _mesa_fprint_instruction_opt(stdout, inst, indent, mode, prog);
+}
+
+
 void
-_mesa_print_program(const struct gl_program *prog)
+_mesa_print_instruction(const struct prog_instruction *inst)
 {
-   _mesa_print_program_opt(prog, PROG_PRINT_DEBUG, GL_TRUE);
+   /* note: 4th param should be ignored for PROG_PRINT_DEBUG */
+   _mesa_fprint_instruction_opt(stdout, inst, 0, PROG_PRINT_DEBUG, NULL);
 }
 
 
+
 /**
  * Print program, with options.
  */
 void
-_mesa_print_program_opt(const struct gl_program *prog,
-                        gl_prog_print_mode mode,
-                        GLboolean lineNumbers)
+_mesa_fprint_program_opt(FILE *f,
+                         const struct gl_program *prog,
+                         gl_prog_print_mode mode,
+                         GLboolean lineNumbers)
 {
    GLuint i, indent = 0;
 
    switch (prog->Target) {
    case GL_VERTEX_PROGRAM_ARB:
       if (mode == PROG_PRINT_ARB)
-         _mesa_printf("!!ARBvp1.0\n");
+         _mesa_fprintf(f, "!!ARBvp1.0\n");
       else if (mode == PROG_PRINT_NV)
-         _mesa_printf("!!VP1.0\n");
+         _mesa_fprintf(f, "!!VP1.0\n");
       else
-         _mesa_printf("# Vertex Program/Shader\n");
+         _mesa_fprintf(f, "# Vertex Program/Shader\n");
       break;
    case GL_FRAGMENT_PROGRAM_ARB:
    case GL_FRAGMENT_PROGRAM_NV:
       if (mode == PROG_PRINT_ARB)
-         _mesa_printf("!!ARBfp1.0\n");
+         _mesa_fprintf(f, "!!ARBfp1.0\n");
       else if (mode == PROG_PRINT_NV)
-         _mesa_printf("!!FP1.0\n");
+         _mesa_fprintf(f, "!!FP1.0\n");
       else
-         _mesa_printf("# Fragment Program/Shader\n");
+         _mesa_fprintf(f, "# Fragment Program/Shader\n");
       break;
    }
 
    for (i = 0; i < prog->NumInstructions; i++) {
       if (lineNumbers)
-         _mesa_printf("%3d: ", i);
-      indent = _mesa_print_instruction_opt(prog->Instructions + i,
+         _mesa_fprintf(f, "%3d: ", i);
+      indent = _mesa_fprint_instruction_opt(f, prog->Instructions + i,
                                            indent, mode, prog);
    }
 }
 
 
 /**
- * Print all of a program's parameters.
+ * Print program to stdout, default options.
  */
 void
-_mesa_print_program_parameters(GLcontext *ctx, const struct gl_program *prog)
+_mesa_print_program(const struct gl_program *prog)
+{
+   _mesa_fprint_program_opt(stdout, prog, PROG_PRINT_DEBUG, GL_TRUE);
+}
+
+
+/**
+ * Print all of a program's parameters/fields to given file.
+ */
+static void
+_mesa_fprint_program_parameters(FILE *f,
+                                GLcontext *ctx,
+                                const struct gl_program *prog)
 {
    GLuint i;
 
-   _mesa_printf("InputsRead: 0x%x\n", prog->InputsRead);
-   _mesa_printf("OutputsWritten: 0x%x\n", prog->OutputsWritten);
-   _mesa_printf("NumInstructions=%d\n", prog->NumInstructions);
-   _mesa_printf("NumTemporaries=%d\n", prog->NumTemporaries);
-   _mesa_printf("NumParameters=%d\n", prog->NumParameters);
-   _mesa_printf("NumAttributes=%d\n", prog->NumAttributes);
-   _mesa_printf("NumAddressRegs=%d\n", prog->NumAddressRegs);
-   _mesa_printf("Samplers=[ ");
+   _mesa_fprintf(f, "InputsRead: 0x%x\n", prog->InputsRead);
+   _mesa_fprintf(f, "OutputsWritten: 0x%x\n", prog->OutputsWritten);
+   _mesa_fprintf(f, "NumInstructions=%d\n", prog->NumInstructions);
+   _mesa_fprintf(f, "NumTemporaries=%d\n", prog->NumTemporaries);
+   _mesa_fprintf(f, "NumParameters=%d\n", prog->NumParameters);
+   _mesa_fprintf(f, "NumAttributes=%d\n", prog->NumAttributes);
+   _mesa_fprintf(f, "NumAddressRegs=%d\n", prog->NumAddressRegs);
+   _mesa_fprintf(f, "Samplers=[ ");
    for (i = 0; i < MAX_SAMPLERS; i++) {
-      _mesa_printf("%d ", prog->SamplerUnits[i]);
+      _mesa_fprintf(f, "%d ", prog->SamplerUnits[i]);
    }
-   _mesa_printf("]\n");
+   _mesa_fprintf(f, "]\n");
 
    _mesa_load_state_parameters(ctx, prog->Parameters);
 
 #if 0
-   _mesa_printf("Local Params:\n");
+   _mesa_fprintf(f, "Local Params:\n");
    for (i = 0; i < MAX_PROGRAM_LOCAL_PARAMS; i++){
       const GLfloat *p = prog->LocalParams[i];
-      _mesa_printf("%2d: %f, %f, %f, %f\n", i, p[0], p[1], p[2], p[3]);
+      _mesa_fprintf(f, "%2d: %f, %f, %f, %f\n", i, p[0], p[1], p[2], p[3]);
    }
 #endif	
    _mesa_print_parameter_list(prog->Parameters);
 }
 
 
+/**
+ * Print all of a program's parameters/fields to stdout.
+ */
 void
-_mesa_print_parameter_list(const struct gl_program_parameter_list *list)
+_mesa_print_program_parameters(GLcontext *ctx, const struct gl_program *prog)
+{
+   _mesa_fprint_program_parameters(stdout, ctx, prog);
+}
+
+
+/**
+ * Print a program parameter list to given file.
+ */
+static void
+_mesa_fprint_parameter_list(FILE *f,
+                            const struct gl_program_parameter_list *list)
 {
    const gl_prog_print_mode mode = PROG_PRINT_DEBUG;
    GLuint i;
@@ -795,13 +852,77 @@ _mesa_print_parameter_list(const struct gl_program_parameter_list *list)
    if (!list)
       return;
 
-   _mesa_printf("param list %p\n", (void *) list);
+   _mesa_fprintf(f, "param list %p\n", (void *) list);
    for (i = 0; i < list->NumParameters; i++){
       struct gl_program_parameter *param = list->Parameters + i;
       const GLfloat *v = list->ParameterValues[i];
-      _mesa_printf("param[%d] sz=%d %s %s = {%.3g, %.3g, %.3g, %.3g};\n",
+      _mesa_fprintf(f, "param[%d] sz=%d %s %s = {%.3g, %.3g, %.3g, %.3g}",
                    i, param->Size,
                    file_string(list->Parameters[i].Type, mode),
                    param->Name, v[0], v[1], v[2], v[3]);
+      if (param->Flags & PROG_PARAM_BIT_CENTROID)
+         _mesa_fprintf(f, " Centroid");
+      if (param->Flags & PROG_PARAM_BIT_INVARIANT)
+         _mesa_fprintf(f, " Invariant");
+      if (param->Flags & PROG_PARAM_BIT_FLAT)
+         _mesa_fprintf(f, " Flat");
+      if (param->Flags & PROG_PARAM_BIT_LINEAR)
+         _mesa_fprintf(f, " Linear");
+      _mesa_fprintf(f, "\n");
    }
 }
+
+
+/**
+ * Print a program parameter list to stdout.
+ */
+void
+_mesa_print_parameter_list(const struct gl_program_parameter_list *list)
+{
+   _mesa_fprint_parameter_list(stdout, list);
+}
+
+
+/**
+ * Write shader and associated info to a file.
+ */
+void
+_mesa_write_shader_to_file(const struct gl_shader *shader)
+{
+   const char *type;
+   char filename[100];
+   FILE *f;
+
+   if (shader->Type == GL_FRAGMENT_SHADER)
+      type = "frag";
+   else
+      type = "vert";
+
+   _mesa_snprintf(filename, strlen(filename), "shader_%u.%s", shader->Name, type);
+   f = fopen(filename, "w");
+   if (!f) {
+      fprintf(stderr, "Unable to open %s for writing\n", filename);
+      return;
+   }
+
+   fprintf(f, "/* Shader %u source */\n", shader->Name);
+   fputs(shader->Source, f);
+   fprintf(f, "\n");
+
+   fprintf(f, "/* Compile status: %s */\n",
+           shader->CompileStatus ? "ok" : "fail");
+   if (!shader->CompileStatus) {
+      fprintf(f, "/* Log Info: */\n");
+      fputs(shader->InfoLog, f);
+   }
+   else {
+      fprintf(f, "/* GPU code */\n");
+      fprintf(f, "/*\n");
+      _mesa_fprint_program_opt(f, shader->Program, PROG_PRINT_DEBUG, GL_TRUE);
+      fprintf(f, "*/\n");
+   }
+
+   fclose(f);
+}
+
+
diff --git a/src/mesa/shader/prog_print.h b/src/mesa/shader/prog_print.h
index cd9e388e82..d55661cebb 100644
--- a/src/mesa/shader/prog_print.h
+++ b/src/mesa/shader/prog_print.h
@@ -47,6 +47,9 @@ const char *
 _mesa_writemask_string(GLuint writeMask);
 
 extern void
+_mesa_print_swizzle(GLuint swizzle);
+
+extern void
 _mesa_print_alu_instruction(const struct prog_instruction *inst,
                             const char *opcode_string, GLuint numRegs);
 
@@ -62,8 +65,9 @@ extern void
 _mesa_print_program(const struct gl_program *prog);
 
 extern void
-_mesa_print_program_opt(const struct gl_program *prog, gl_prog_print_mode mode,
-                        GLboolean lineNumbers);
+_mesa_fprint_program_opt(FILE *f,
+                         const struct gl_program *prog, gl_prog_print_mode mode,
+                         GLboolean lineNumbers);
 
 extern void
 _mesa_print_program_parameters(GLcontext *ctx, const struct gl_program *prog);
@@ -72,4 +76,8 @@ extern void
 _mesa_print_parameter_list(const struct gl_program_parameter_list *list);
 
 
+extern void
+_mesa_write_shader_to_file(const struct gl_shader *shader);
+
+
 #endif /* PROG_PRINT_H */
diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c
index 971eb25a49..e1db30b78f 100644
--- a/src/mesa/shader/prog_statevars.c
+++ b/src/mesa/shader/prog_statevars.c
@@ -395,6 +395,12 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[],
 
    case STATE_INTERNAL:
       switch (state[1]) {
+      case STATE_CURRENT_ATTRIB: {
+         const GLuint idx = (GLuint) state[2];
+         COPY_4V(value, ctx->Current.Attrib[idx]);
+         return;
+      }						  
+
       case STATE_NORMAL_SCALE:
          ASSIGN_4V(value, 
                    ctx->_ModelViewInvScale, 
@@ -493,14 +499,17 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[],
             const struct gl_texture_object *texObj
                = ctx->Texture.Unit[unit]._Current;
             if (texObj) {
-               value[0] = texObj->ShadowAmbient;
-               value[1] = texObj->ShadowAmbient;
-               value[2] = texObj->ShadowAmbient;
-               value[3] = texObj->ShadowAmbient;
+               value[0] =
+               value[1] =
+               value[2] =
+               value[3] = texObj->CompareFailValue;
             }
          }
          return;
 
+         /* XXX: make sure new tokens added here are also handled in the 
+          * _mesa_program_state_flags() switch, below.
+          */
       default:
          /* unknown state indexes are silently ignored
           *  should be handled by the driver.
@@ -574,11 +583,29 @@ _mesa_program_state_flags(const gl_state_index state[STATE_LENGTH])
 
    case STATE_INTERNAL:
       switch (state[1]) {
+      case STATE_CURRENT_ATTRIB:
+         return _NEW_CURRENT_ATTRIB;
+
+      case STATE_NORMAL_SCALE:
+         return _NEW_MODELVIEW;
+
       case STATE_TEXRECT_SCALE:
       case STATE_SHADOW_AMBIENT:
 	 return _NEW_TEXTURE;
       case STATE_FOG_PARAMS_OPTIMIZED:
 	 return _NEW_FOG;
+      case STATE_LIGHT_SPOT_DIR_NORMALIZED:
+      case STATE_LIGHT_POSITION:
+      case STATE_LIGHT_POSITION_NORMALIZED:
+      case STATE_LIGHT_HALF_VECTOR:
+         return _NEW_LIGHT;
+
+      case STATE_PT_SCALE:
+      case STATE_PT_BIAS:
+      case STATE_PCM_SCALE:
+      case STATE_PCM_BIAS:
+         return _NEW_PIXEL;
+
       default:
          /* unknown state indexes are silently ignored and
          *  no flag set, since it is handled by the driver.
@@ -604,6 +631,9 @@ append(char *dst, const char *src)
 }
 
 
+/**
+ * Convert token 'k' to a string, append it onto 'dst' string.
+ */
 static void
 append_token(char *dst, gl_state_index k)
 {
@@ -736,11 +766,30 @@ append_token(char *dst, gl_state_index k)
    case STATE_LOCAL:
       append(dst, "local");
       break;
+   /* BEGIN internal state vars */
+   case STATE_INTERNAL:
+      append(dst, "(internal)");
+      break;
    case STATE_NORMAL_SCALE:
       append(dst, "normalScale");
       break;
-   case STATE_INTERNAL:
-      append(dst, "(internal)");
+   case STATE_TEXRECT_SCALE:
+      append(dst, "texrectScale");
+      break;
+   case STATE_FOG_PARAMS_OPTIMIZED:
+      append(dst, "fogParamsOptimized");
+      break;
+   case STATE_LIGHT_SPOT_DIR_NORMALIZED:
+      append(dst, "lightSpotDirNormalized");
+      break;
+   case STATE_LIGHT_POSITION:
+      append(dst, "lightPosition");
+      break;
+   case STATE_LIGHT_POSITION_NORMALIZED:
+      append(dst, "light.position.normalized");
+      break;
+   case STATE_LIGHT_HALF_VECTOR:
+      append(dst, "lightHalfVector");
       break;
    case STATE_PT_SCALE:
       append(dst, "PTscale");
@@ -755,10 +804,11 @@ append_token(char *dst, gl_state_index k)
       append(dst, "PCMbias");
       break;
    case STATE_SHADOW_AMBIENT:
-      append(dst, "ShadowAmbient");
+      append(dst, "CompareFailValue");
       break;
    default:
-      ;
+      /* probably STATE_INTERNAL_DRIVER+i (driver private state) */
+      append(dst, "driverState");
    }
 }
 
@@ -784,23 +834,23 @@ append_index(char *dst, GLint index)
  * For example, return "state.matrix.texture[2].inverse".
  * Use _mesa_free() to deallocate the string.
  */
-const char *
+char *
 _mesa_program_state_string(const gl_state_index state[STATE_LENGTH])
 {
    char str[1000] = "";
    char tmp[30];
 
    append(str, "state.");
-   append_token(str, (gl_state_index) state[0]);
+   append_token(str, state[0]);
 
    switch (state[0]) {
    case STATE_MATERIAL:
       append_face(str, state[1]);
-      append_token(str, (gl_state_index) state[2]);
+      append_token(str, state[2]);
       break;
    case STATE_LIGHT:
       append_index(str, state[1]); /* light number [i]. */
-      append_token(str, (gl_state_index) state[2]); /* coefficients */
+      append_token(str, state[2]); /* coefficients */
       break;
    case STATE_LIGHTMODEL_AMBIENT:
       append(str, "lightmodel.ambient");
@@ -816,11 +866,11 @@ _mesa_program_state_string(const gl_state_index state[STATE_LENGTH])
    case STATE_LIGHTPROD:
       append_index(str, state[1]); /* light number [i]. */
       append_face(str, state[2]);
-      append_token(str, (gl_state_index) state[3]);
+      append_token(str, state[3]);
       break;
    case STATE_TEXGEN:
       append_index(str, state[1]); /* tex unit [i] */
-      append_token(str, (gl_state_index) state[2]); /* plane coef */
+      append_token(str, state[2]); /* plane coef */
       break;
    case STATE_TEXENV_COLOR:
       append_index(str, state[1]); /* tex unit [i] */
@@ -842,11 +892,11 @@ _mesa_program_state_string(const gl_state_index state[STATE_LENGTH])
          /* state[2] = first row to fetch */
          /* state[3] = last row to fetch */
          /* state[4] = transpose, inverse or invtrans */
-         const gl_state_index mat = (gl_state_index) state[0];
+         const gl_state_index mat = state[0];
          const GLuint index = (GLuint) state[1];
          const GLuint firstRow = (GLuint) state[2];
          const GLuint lastRow = (GLuint) state[3];
-         const gl_state_index modifier = (gl_state_index) state[4];
+         const gl_state_index modifier = state[4];
          if (index ||
              mat == STATE_TEXTURE_MATRIX ||
              mat == STATE_PROGRAM_MATRIX)
@@ -874,10 +924,11 @@ _mesa_program_state_string(const gl_state_index state[STATE_LENGTH])
    case STATE_VERTEX_PROGRAM:
       /* state[1] = {STATE_ENV, STATE_LOCAL} */
       /* state[2] = parameter index          */
-      append_token(str, (gl_state_index) state[1]);
+      append_token(str, state[1]);
       append_index(str, state[2]);
       break;
    case STATE_INTERNAL:
+      append_token(str, state[1]);
       break;
    default:
       _mesa_problem(NULL, "Invalid state in _mesa_program_state_string");
diff --git a/src/mesa/shader/prog_statevars.h b/src/mesa/shader/prog_statevars.h
index d3091147f8..d5358a1d04 100644
--- a/src/mesa/shader/prog_statevars.h
+++ b/src/mesa/shader/prog_statevars.h
@@ -79,10 +79,10 @@ typedef enum gl_state_index_ {
    STATE_SHININESS,
    STATE_HALF_VECTOR,
 
-   STATE_POSITION,
-   STATE_ATTENUATION,
-   STATE_SPOT_DIRECTION,
-   STATE_SPOT_CUTOFF,
+   STATE_POSITION,       /**< xyzw = position */
+   STATE_ATTENUATION,    /**< xyz = attenuation, w = spot exponent */
+   STATE_SPOT_DIRECTION, /**< xyz = direction, w = cos(cutoff) */
+   STATE_SPOT_CUTOFF,    /**< x = cutoff, yzw = undefined */
 
    STATE_TEXGEN_EYE_S,
    STATE_TEXGEN_EYE_T,
@@ -104,6 +104,7 @@ typedef enum gl_state_index_ {
    STATE_LOCAL,
 
    STATE_INTERNAL,		/* Mesa additions */
+   STATE_CURRENT_ATTRIB,        /* ctx->Current vertex attrib value */
    STATE_NORMAL_SCALE,
    STATE_TEXRECT_SCALE,
    STATE_FOG_PARAMS_OPTIMIZED,  /* for faster fog calc */
@@ -130,7 +131,7 @@ extern GLbitfield
 _mesa_program_state_flags(const gl_state_index state[STATE_LENGTH]);
 
 
-extern const char *
+extern char *
 _mesa_program_state_string(const gl_state_index state[STATE_LENGTH]);
 
 
diff --git a/src/mesa/shader/prog_uniform.c b/src/mesa/shader/prog_uniform.c
index dc76be8e46..0642713148 100644
--- a/src/mesa/shader/prog_uniform.c
+++ b/src/mesa/shader/prog_uniform.c
@@ -142,8 +142,8 @@ _mesa_longest_uniform_name(const struct gl_uniform_list *list)
    GLint max = 0;
    GLuint i;
    for (i = 0; list && i < list->NumUniforms; i++) {
-      GLuint len = _mesa_strlen(list->Uniforms[i].Name);
-      if (len > (GLuint)max)
+      GLint len = (GLint)_mesa_strlen(list->Uniforms[i].Name);
+      if (len > max)
          max = len;
    }
    return max;
diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c
index 738891a029..7a3b827352 100644
--- a/src/mesa/shader/program.c
+++ b/src/mesa/shader/program.c
@@ -554,7 +554,6 @@ _mesa_insert_instructions(struct gl_program *prog, GLuint start, GLuint count)
    return GL_TRUE;
 }
 
-
 /**
  * Delete 'count' instructions at 'start' in the given program.
  * Adjust branch targets accordingly.
@@ -571,7 +570,7 @@ _mesa_delete_instructions(struct gl_program *prog, GLuint start, GLuint count)
    for (i = 0; i < prog->NumInstructions; i++) {
       struct prog_instruction *inst = prog->Instructions + i;
       if (inst->BranchTarget > 0) {
-         if (inst->BranchTarget >= start) {
+         if (inst->BranchTarget > start) {
             inst->BranchTarget -= count;
          }
       }
@@ -691,17 +690,47 @@ _mesa_combine_programs(GLcontext *ctx,
 
    if (newProg->Target == GL_FRAGMENT_PROGRAM_ARB) {
       struct gl_fragment_program *fprogA, *fprogB, *newFprog;
+      GLbitfield progB_inputsRead = progB->InputsRead;
+      GLint progB_colorFile, progB_colorIndex;
+
       fprogA = (struct gl_fragment_program *) progA;
       fprogB = (struct gl_fragment_program *) progB;
       newFprog = (struct gl_fragment_program *) newProg;
 
       newFprog->UsesKill = fprogA->UsesKill || fprogB->UsesKill;
 
+      /* We'll do a search and replace for instances
+       * of progB_colorFile/progB_colorIndex below...
+       */
+      progB_colorFile = PROGRAM_INPUT;
+      progB_colorIndex = FRAG_ATTRIB_COL0;
+
+      /*
+       * The fragment program may get color from a state var rather than
+       * a fragment input (vertex output) if it's constant.
+       * See the texenvprogram.c code.
+       * So, search the program's parameter list now to see if the program
+       * gets color from a state var instead of a conventional fragment
+       * input register.
+       */
+      for (i = 0; i < progB->Parameters->NumParameters; i++) {
+         struct gl_program_parameter *p = &progB->Parameters->Parameters[i];
+         if (p->Type == PROGRAM_STATE_VAR &&
+             p->StateIndexes[0] == STATE_INTERNAL &&
+             p->StateIndexes[1] == STATE_CURRENT_ATTRIB &&
+             p->StateIndexes[2] == VERT_ATTRIB_COLOR0) {
+            progB_inputsRead |= FRAG_BIT_COL0;
+            progB_colorFile = PROGRAM_STATE_VAR;
+            progB_colorIndex = i;
+            break;
+         }
+      }
+
       /* Connect color outputs of fprogA to color inputs of fprogB, via a
        * new temporary register.
        */
       if ((progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) &&
-          (progB->InputsRead & (1 << FRAG_ATTRIB_COL0))) {
+          (progB_inputsRead & FRAG_BIT_COL0)) {
          GLint tempReg = _mesa_find_free_register(newProg, PROGRAM_TEMPORARY);
          if (tempReg < 0) {
             _mesa_problem(ctx, "No free temp regs found in "
@@ -712,13 +741,14 @@ _mesa_combine_programs(GLcontext *ctx,
          replace_registers(newInst, lenA,
                            PROGRAM_OUTPUT, FRAG_RESULT_COLR,
                            PROGRAM_TEMPORARY, tempReg);
-         /* replace reads from input.color[0] with tempReg */
+         /* replace reads from the input color with tempReg */
          replace_registers(newInst + lenA, lenB,
-                           PROGRAM_INPUT, FRAG_ATTRIB_COL0,
-                           PROGRAM_TEMPORARY, tempReg);
+                           progB_colorFile, progB_colorIndex, /* search for */
+                           PROGRAM_TEMPORARY, tempReg  /* replace with */ );
       }
 
-      inputsB = progB->InputsRead;
+      /* compute combined program's InputsRead */
+      inputsB = progB_inputsRead;
       if (progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) {
          inputsB &= ~(1 << FRAG_ATTRIB_COL0);
       }
diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c
index b3d66c5bab..013e912e5d 100644
--- a/src/mesa/shader/shader_api.c
+++ b/src/mesa/shader/shader_api.c
@@ -370,6 +370,31 @@ _mesa_lookup_shader_err(GLcontext *ctx, GLuint name, const char *caller)
 }
 
 
+/**
+ * Return mask of GLSL_x flags by examining the MESA_GLSL env var.
+ */
+static GLbitfield
+get_shader_flags(void)
+{
+   GLbitfield flags = 0x0;
+   const char *env = _mesa_getenv("MESA_GLSL");
+
+   if (env) {
+      if (_mesa_strstr(env, "dump"))
+         flags |= GLSL_DUMP;
+      if (_mesa_strstr(env, "log"))
+         flags |= GLSL_LOG;
+      if (_mesa_strstr(env, "nopt"))
+         flags |= GLSL_NO_OPT;
+      else if (_mesa_strstr(env, "opt"))
+         flags |= GLSL_OPT;
+      if (_mesa_strstr(env, "uniform"))
+         flags |= GLSL_UNIFORMS;
+   }
+
+   return flags;
+}
+
 
 /**
  * Initialize context's shader state.
@@ -381,8 +406,9 @@ _mesa_init_shader_state(GLcontext * ctx)
     * are generated by the GLSL compiler.
     */
    ctx->Shader.EmitHighLevelInstructions = GL_TRUE;
-   ctx->Shader.EmitCondCodes = GL_TRUE; /* XXX probably want GL_FALSE... */
+   ctx->Shader.EmitCondCodes = GL_FALSE;/*GL_TRUE;*/ /* XXX probably want GL_FALSE... */
    ctx->Shader.EmitComments = GL_FALSE;
+   ctx->Shader.Flags = get_shader_flags();
 }
 
 
@@ -866,6 +892,7 @@ _mesa_get_active_uniform(GLcontext *ctx, GLuint program, GLuint index,
 {
    const struct gl_shader_program *shProg;
    const struct gl_program *prog;
+   const struct gl_program_parameter *param;
    GLint progPos;
 
    shProg = _mesa_lookup_shader_program_err(ctx, program, "glGetActiveUniform");
@@ -891,14 +918,30 @@ _mesa_get_active_uniform(GLcontext *ctx, GLuint program, GLuint index,
    if (!prog || progPos < 0)
       return; /* should never happen */
 
-   if (nameOut)
-      copy_string(nameOut, maxLength, length,
-                  prog->Parameters->Parameters[progPos].Name);
-   if (size)
-      *size = prog->Parameters->Parameters[progPos].Size
-         / sizeof_glsl_type(prog->Parameters->Parameters[progPos].DataType);
-   if (type)
-      *type = prog->Parameters->Parameters[progPos].DataType;
+   ASSERT(progPos < prog->Parameters->NumParameters);
+   param = &prog->Parameters->Parameters[progPos];
+
+   if (nameOut) {
+      copy_string(nameOut, maxLength, length, param->Name);
+   }
+
+   if (size) {
+      GLint typeSize = sizeof_glsl_type(param->DataType);
+      if (param->Size > typeSize) {
+         /* This is an array.
+          * Array elements are placed on vector[4] boundaries so they're
+          * a multiple of four floats.  We round typeSize up to next multiple
+          * of four to get the right size below.
+          */
+         typeSize = (typeSize + 3) & ~3;
+      }
+      /* Note that the returned size is in units of the <type>, not bytes */
+      *size = param->Size / typeSize;
+   }
+
+   if (type) {
+      *type = param->DataType;
+   }
 }
 
 
@@ -1117,7 +1160,8 @@ get_matrix_dims(GLenum type, GLint *rows, GLint *cols)
 
 /**
  * Determine the number of rows and columns occupied by a uniform
- * according to its datatype.
+ * according to its datatype.  For non-matrix types (such as GL_FLOAT_VEC4),
+ * the number of rows = 1 and cols = number of elements in the vector.
  */
 static void
 get_uniform_rows_cols(const struct gl_program_parameter *p,
@@ -1126,33 +1170,45 @@ get_uniform_rows_cols(const struct gl_program_parameter *p,
    get_matrix_dims(p->DataType, rows, cols);
    if (*rows == 0 && *cols == 0) {
       /* not a matrix type, probably a float or vector */
-      *rows = p->Size / 4 + 1;
-      if (p->Size % 4 == 0)
-         *cols = 4;
-      else
-         *cols = p->Size % 4;
+      if (p->Size <= 4) {
+         *rows = 1;
+         *cols = p->Size;
+      }
+      else {
+         *rows = p->Size / 4 + 1;
+         if (p->Size % 4 == 0)
+            *cols = 4;
+         else
+            *cols = p->Size % 4;
+      }
    }
 }
 
 
-#define MAX_UNIFORM_ELEMENTS 16
-
 /**
- * Helper for GetUniformfv(), GetUniformiv()
- * Returns number of elements written to 'params' output.
+ * Helper for get_uniform[fi]v() functions.
+ * Given a shader program name and uniform location, return a pointer
+ * to the shader program and return the program parameter position.
  */
-static GLuint
-get_uniformfv(GLcontext *ctx, GLuint program, GLint location,
-              GLfloat *params)
+static void
+lookup_uniform_parameter(GLcontext *ctx, GLuint program, GLint location,
+                         struct gl_program **progOut, GLint *paramPosOut)
 {
    struct gl_shader_program *shProg
       = _mesa_lookup_shader_program_err(ctx, program, "glGetUniform[if]v");
-   if (shProg) {
-      if (shProg->Uniforms &&
-          location >= 0 && location < (GLint) shProg->Uniforms->NumUniforms) {
-         GLint progPos;
-         const struct gl_program *prog = NULL;
+   struct gl_program *prog = NULL;
+   GLint progPos = -1;
 
+   /* if shProg is NULL, we'll have already recorded an error */
+
+   if (shProg) {
+      if (!shProg->Uniforms ||
+          location < 0 ||
+          location >= (GLint) shProg->Uniforms->NumUniforms) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,  "glGetUniformfv(location)");
+      }
+      else {
+         /* OK, find the gl_program and program parameter location */
          progPos = shProg->Uniforms->Uniforms[location].VertPos;
          if (progPos >= 0) {
             prog = &shProg->VertexProgram->Base;
@@ -1163,33 +1219,11 @@ get_uniformfv(GLcontext *ctx, GLuint program, GLint location,
                prog = &shProg->FragmentProgram->Base;
             }
          }
-
-         ASSERT(prog);
-         if (prog) {
-            const struct gl_program_parameter *p =
-               &prog->Parameters->Parameters[progPos];
-            GLint rows, cols, i, j, k;
-
-            /* See uniformiv() below */                    
-            assert(p->Size <= MAX_UNIFORM_ELEMENTS);
-
-            get_uniform_rows_cols(p, &rows, &cols);
-
-            k = 0;
-            for (i = 0; i < rows; i++) {
-               for (j = 0; j < cols; j++ ) {
-                  params[k++] = prog->Parameters->ParameterValues[progPos+i][j];
-               }
-            }
-
-            return p->Size;
-         }
-      }
-      else {
-         _mesa_error(ctx, GL_INVALID_OPERATION, "glGetUniformfv(location)");
       }
    }
-   return 0;
+
+   *progOut = prog;
+   *paramPosOut = progPos;
 }
 
 
@@ -1200,23 +1234,54 @@ static void
 _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location,
                     GLfloat *params)
 {
-   (void) get_uniformfv(ctx, program, location, params);
+   struct gl_program *prog;
+   GLint paramPos;
+
+   lookup_uniform_parameter(ctx, program, location, &prog, &paramPos);
+
+   if (prog) {
+      const struct gl_program_parameter *p =
+         &prog->Parameters->Parameters[paramPos];
+      GLint rows, cols, i, j, k;
+
+      get_uniform_rows_cols(p, &rows, &cols);
+
+      k = 0;
+      for (i = 0; i < rows; i++) {
+         for (j = 0; j < cols; j++ ) {
+            params[k++] = prog->Parameters->ParameterValues[paramPos+i][j];
+         }
+      }
+   }
 }
 
 
 /**
  * Called via ctx->Driver.GetUniformiv().
+ * \sa _mesa_get_uniformfv, only difference is a cast.
  */
 static void
 _mesa_get_uniformiv(GLcontext *ctx, GLuint program, GLint location,
                     GLint *params)
 {
-   GLfloat fparams[MAX_UNIFORM_ELEMENTS];
-   GLuint n = get_uniformfv(ctx, program, location, fparams);
-   GLuint i;
-   assert(n <= MAX_UNIFORM_ELEMENTS);
-   for (i = 0; i < n; i++) {
-      params[i] = (GLint) fparams[i];
+   struct gl_program *prog;
+   GLint paramPos;
+
+   lookup_uniform_parameter(ctx, program, location, &prog, &paramPos);
+
+   if (prog) {
+      const struct gl_program_parameter *p =
+         &prog->Parameters->Parameters[paramPos];
+      GLint rows, cols, i, j, k;
+
+      get_uniform_rows_cols(p, &rows, &cols);
+
+      k = 0;
+      for (i = 0; i < rows; i++) {
+         for (j = 0; j < cols; j++ ) {
+            params[k++] = (GLint) prog->Parameters->ParameterValues[paramPos+i][j];
+         }
+      }
    }
 }
 
@@ -1357,7 +1422,10 @@ _mesa_compile_shader(GLcontext *ctx, GLuint shaderObj)
    if (!sh)
       return;
 
-   sh->CompileStatus = _slang_compile(ctx, sh);
+   /* this call will set the sh->CompileStatus field to indicate if
+    * compilation was successful.
+    */
+   (void) _slang_compile(ctx, sh);
 }
 
 
@@ -1527,7 +1595,7 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program,
       return;
    }
 
-   if (index + offset > program->Parameters->Size) {
+   if (index + offset > (GLint) program->Parameters->Size) {
       /* out of bounds! */
       return;
    }
@@ -1535,27 +1603,36 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program,
    if (param->Type == PROGRAM_SAMPLER) {
       /* This controls which texture unit which is used by a sampler */
       GLuint texUnit, sampler;
+      GLint i;
 
       /* data type for setting samplers must be int */
-      if (type != GL_INT || count != 1) {
+      if (type != GL_INT) {
          _mesa_error(ctx, GL_INVALID_OPERATION,
                      "glUniform(only glUniform1i can be used "
                      "to set sampler uniforms)");
          return;
       }
 
-      sampler = (GLuint) program->Parameters->ParameterValues[index][0];
-      texUnit = ((GLuint *) values)[0];
+      /* XXX arrays of samplers haven't been tested much, but it's not a
+       * common thing...
+       */
+      for (i = 0; i < count; i++) {
+         sampler = (GLuint) program->Parameters->ParameterValues[index + i][0];
+         texUnit = ((GLuint *) values)[i];
+
+         /* check that the sampler (tex unit index) is legal */
+         if (texUnit >= ctx->Const.MaxTextureImageUnits) {
+            _mesa_error(ctx, GL_INVALID_VALUE,
+                        "glUniform1(invalid sampler/tex unit index)");
+            return;
+         }
 
-      /* check that the sampler (tex unit index) is legal */
-      if (texUnit >= ctx->Const.MaxTextureImageUnits) {
-         _mesa_error(ctx, GL_INVALID_VALUE,
-                     "glUniform1(invalid sampler/tex unit index)");
-         return;
+         /* This maps a sampler to a texture unit: */
+         if (sampler < MAX_SAMPLERS) {
+            program->SamplerUnits[sampler] = texUnit;
+         }
       }
 
-      /* This maps a sampler to a texture unit: */
-      program->SamplerUnits[sampler] = texUnit;
       _mesa_update_shader_textures_used(program);
 
       FLUSH_VERTICES(ctx, _NEW_TEXTURE);
@@ -1563,19 +1640,31 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program,
    else {
       /* ordinary uniform variable */
       GLsizei k, i;
-      GLint slots = (param->Size + 3) / 4;
+      const GLint slots = (param->Size + 3) / 4;
+      const GLint typeSize = sizeof_glsl_type(param->DataType);
 
-      if (count * elems > (GLint) param->Size) {
-         _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(count too large)");
-         return;
+      if (param->Size > typeSize) {
+         /* an array */
+         /* we'll ignore extra data below */
+      }
+      else {
+         /* non-array: count must be one */
+         if (count != 1) {
+            _mesa_error(ctx, GL_INVALID_OPERATION,
+                        "glUniform(uniform is not an array)");
+            return;
+         }
       }
-
-      if (count > slots)
-         count = slots;
 
       for (k = 0; k < count; k++) {
-         GLfloat *uniformVal =
-            program->Parameters->ParameterValues[index + offset + k];
+         GLfloat *uniformVal;
+
+         if (offset + k > slots) {
+            /* Extra array data is ignored */
+            break;
+         }
+
+         uniformVal = program->Parameters->ParameterValues[index + offset + k];
          if (is_integer_type(type)) {
             const GLint *iValues = ((const GLint *) values) + k * elems;
             for (i = 0; i < elems; i++) {
@@ -1592,7 +1681,7 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program,
          /* if the uniform is bool-valued, convert to 1.0 or 0.0 */
          if (is_boolean_type(param->DataType)) {
             for (i = 0; i < elems; i++) {
-               uniformVal[i] = uniformVal[i] ? 1.0 : 0.0;
+               uniformVal[i] = uniformVal[i] ? 1.0f : 0.0f;
             }
          }
       }
@@ -1610,6 +1699,7 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count,
    struct gl_shader_program *shProg = ctx->Shader.CurrentProgram;
    struct gl_uniform *uniform;
    GLint elems, offset;
+   GLenum basicType;
 
    if (!shProg || !shProg->LinkStatus) {
       _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(program not linked)");
@@ -1619,6 +1709,11 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count,
    if (location == -1)
       return;   /* The standard specifies this as a no-op */
 
+   if (location < -1) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(location)");
+      return;
+   }
+
    split_location_offset(&location, &offset);
 
    if (location < 0 || location >= (GLint) shProg->Uniforms->NumUniforms) {
@@ -1633,19 +1728,35 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count,
 
    switch (type) {
    case GL_FLOAT:
+      basicType = GL_FLOAT;
+      elems = 1;
+      break;
    case GL_INT:
+      basicType = GL_INT;
       elems = 1;
       break;
    case GL_FLOAT_VEC2:
+      basicType = GL_FLOAT;
+      elems = 2;
+      break;
    case GL_INT_VEC2:
+      basicType = GL_INT;
       elems = 2;
       break;
    case GL_FLOAT_VEC3:
+      basicType = GL_FLOAT;
+      elems = 3;
+      break;
    case GL_INT_VEC3:
+      basicType = GL_INT;
       elems = 3;
       break;
    case GL_FLOAT_VEC4:
+      basicType = GL_FLOAT;
+      elems = 4;
+      break;
    case GL_INT_VEC4:
+      basicType = GL_INT;
       elems = 4;
       break;
    default:
@@ -1657,6 +1768,25 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count,
 
    uniform = &shProg->Uniforms->Uniforms[location];
 
+   if (ctx->Shader.Flags & GLSL_UNIFORMS) {
+      GLint i;
+      _mesa_printf("Mesa: set program %u uniform %s (loc %d) to: ",
+                   shProg->Name, uniform->Name, location);
+      if (basicType == GL_INT) {
+         const GLint *v = (const GLint *) values;
+         for (i = 0; i < count * elems; i++) {
+            _mesa_printf("%d ", v[i]);
+         }
+      }
+      else {
+         const GLfloat *v = (const GLfloat *) values;
+         for (i = 0; i < count * elems; i++) {
+            _mesa_printf("%g ", v[i]);
+         }
+      }
+      _mesa_printf("\n");
+   }
+
    /* A uniform var may be used by both a vertex shader and a fragment
     * shader.  We may need to update one or both shader's uniform here:
     */
@@ -1758,6 +1888,11 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows,
    if (location == -1)
       return;   /* The standard specifies this as a no-op */
 
+   if (location < -1) {
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glUniformMatrix(location)");
+      return;
+   }
+
    split_location_offset(&location, &offset);
 
    if (location < 0 || location >= (GLint) shProg->Uniforms->NumUniforms) {
diff --git a/src/mesa/shader/slang/library/slang_120_core_gc.h b/src/mesa/shader/slang/library/slang_120_core_gc.h
index fb37f2f5ce..1fdbddf7c3 100644
--- a/src/mesa/shader/slang/library/slang_120_core_gc.h
+++ b/src/mesa/shader/slang/library/slang_120_core_gc.h
@@ -2,96 +2,99 @@
 /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */
 /* slang_120_core.gc */
 
-4,1,90,95,0,0,26,1,1,1,0,0,9,102,48,48,0,0,1,1,0,0,9,102,49,48,0,0,1,1,0,0,9,102,50,48,0,0,1,1,0,0,
-9,102,48,49,0,0,1,1,0,0,9,102,49,49,0,0,1,1,0,0,9,102,50,49,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,
-121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,102,50,48,0,
+5,1,90,95,0,0,26,0,1,1,1,0,0,9,0,102,48,48,0,0,1,1,0,0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,50,48,0,0,
+1,1,0,0,9,0,102,48,49,0,0,1,1,0,0,9,0,102,49,49,0,0,1,1,0,0,9,0,102,50,49,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,
+16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,
+18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,
+9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,9,0,102,0,
+0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,0,18,102,0,0,17,48,0,48,0,0,0,
+17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,
+5,0,105,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,9,18,95,95,
+114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,
+1,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,9,18,95,95,
+114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,
+11,0,99,48,0,0,1,1,0,0,11,0,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,
+0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,
+0,9,0,102,48,48,0,0,1,1,0,0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,50,48,0,0,1,1,0,0,9,0,102,51,48,0,0,
+1,1,0,0,9,0,102,48,49,0,0,1,1,0,0,9,0,102,49,49,0,0,1,1,0,0,9,0,102,50,49,0,0,1,1,0,0,9,0,102,51,
+49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,
+0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,
+0,18,102,51,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,
+0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
+10,49,0,57,59,119,0,18,102,51,49,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,
+17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,
+0,28,0,1,1,1,0,0,5,0,105,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,
+0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,102,0,0,0,20,0,0,1,90,95,0,
+0,28,0,1,1,1,0,0,1,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,
+0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,
+28,0,1,1,1,0,0,12,0,99,48,0,0,1,1,0,0,12,0,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,
+48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,90,95,
+0,0,27,0,1,1,1,0,0,9,0,102,48,48,0,0,1,1,0,0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,48,49,0,0,1,1,0,0,9,
+0,102,49,49,0,0,1,1,0,0,9,0,102,48,50,0,0,1,1,0,0,9,0,102,49,50,0,0,0,1,9,18,95,95,114,101,116,86,
+97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,
+59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,
+49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,
+95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,
+0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,5,0,105,0,0,0,
+1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,
+97,108,0,58,109,97,116,51,120,50,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,1,0,98,0,0,0,1,
+3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,
+108,0,58,109,97,116,51,120,50,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,10,0,99,48,0,0,1,
+1,0,0,10,0,99,49,0,0,1,1,0,0,10,0,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
+99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,9,0,102,48,48,0,0,1,1,0,
+0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,50,48,0,0,1,1,0,0,9,0,102,51,48,0,0,1,1,0,0,9,0,102,48,49,0,0,
+1,1,0,0,9,0,102,49,49,0,0,1,1,0,0,9,0,102,50,49,0,0,1,1,0,0,9,0,102,51,49,0,0,1,1,0,0,9,0,102,48,
+50,0,0,1,1,0,0,9,0,102,49,50,0,0,1,1,0,0,9,0,102,50,50,0,0,1,1,0,0,9,0,102,51,50,0,0,0,1,9,18,95,
+95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,
+122,0,18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,102,51,48,0,
 20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114,
 101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,
-16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,0,1,90,95,0,0,26,1,1,1,0,0,9,102,0,0,0,1,9,18,95,95,114,
-101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,
-48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,26,1,1,1,0,0,5,105,0,0,0,1,3,2,90,95,1,0,
-9,1,102,0,2,58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,
-116,50,120,51,0,18,102,0,0,0,20,0,0,1,90,95,0,0,26,1,1,1,0,0,1,98,0,0,0,1,3,2,90,95,1,0,9,1,102,0,
-2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,
-51,0,18,102,0,0,0,20,0,0,1,90,95,0,0,26,1,1,1,0,0,11,99,48,0,0,1,1,0,0,11,99,49,0,0,0,1,9,18,95,95,
+16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,
+0,18,102,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,
+0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,50,0,57,59,122,0,18,102,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
+10,50,0,57,59,119,0,18,102,51,50,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,
+17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,
+48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,5,0,105,0,0,0,1,3,2,
+90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,
+108,0,58,109,97,116,51,120,52,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,1,0,98,0,0,0,1,3,
+2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,
+108,0,58,109,97,116,51,120,52,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,12,0,99,48,0,0,1,
+1,0,0,12,0,99,49,0,0,1,1,0,0,12,0,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
+99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,9,0,102,48,48,0,0,1,1,0,
+0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,48,49,0,0,1,1,0,0,9,0,102,49,49,0,0,1,1,0,0,9,0,102,48,50,0,0,
+1,1,0,0,9,0,102,49,50,0,0,1,1,0,0,9,0,102,48,51,0,0,1,1,0,0,9,0,102,49,51,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,
+16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,
+0,18,102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,
+0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
+10,51,0,57,59,120,0,18,102,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,
+18,102,49,51,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
+0,58,109,97,116,52,120,50,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,16,10,52,0,0,17,48,0,48,
+0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,5,0,
+105,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,9,18,95,95,114,
+101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,1,0,
+98,0,0,0,1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,9,18,95,95,114,
+101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,10,
+0,99,48,0,0,1,1,0,0,10,0,99,49,0,0,1,1,0,0,10,0,99,50,0,0,1,1,0,0,10,0,99,51,0,0,0,1,9,18,95,95,
 114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
-57,18,99,49,0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,9,102,48,48,0,0,1,1,0,0,9,102,49,48,0,0,1,1,0,0,9,
-102,50,48,0,0,1,1,0,0,9,102,51,48,0,0,1,1,0,0,9,102,48,49,0,0,1,1,0,0,9,102,49,49,0,0,1,1,0,0,9,
-102,50,49,0,0,1,1,0,0,9,102,51,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,
-18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,
-18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,8,48,0,57,59,119,0,18,102,51,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,
-0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,
-49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0,18,102,51,49,0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,9,
-102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,102,0,0,17,48,0,48,0,0,
-0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,
-0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,5,105,0,0,0,1,3,2,90,95,1,0,9,1,102,0,2,58,102,108,111,97,116,0,
-18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,102,0,0,0,20,0,0,
-1,90,95,0,0,28,1,1,1,0,0,1,98,0,0,0,1,3,2,90,95,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18,98,0,0,
-0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,102,0,0,0,20,0,0,1,90,95,0,0,
-28,1,1,1,0,0,12,99,48,0,0,1,1,0,0,12,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,
-18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,90,95,0,0,27,
-1,1,1,0,0,9,102,48,48,0,0,1,1,0,0,9,102,49,48,0,0,1,1,0,0,9,102,48,49,0,0,1,1,0,0,9,102,49,49,0,0,
-1,1,0,0,9,102,48,50,0,0,1,1,0,0,9,102,49,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,
-57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,
-48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,
-59,121,0,18,102,49,50,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,58,109,97,116,51,120,50,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,
-0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,5,105,0,0,0,1,3,2,90,95,1,0,9,1,102,0,2,
-58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,
-50,0,18,102,0,0,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,1,98,0,0,0,1,3,2,90,95,1,0,9,1,102,0,2,58,102,
-108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,
-102,0,0,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,10,99,48,0,0,1,1,0,0,10,99,49,0,0,1,1,0,0,10,99,50,0,0,0,
-1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,
-0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,
-1,90,95,0,0,30,1,1,1,0,0,9,102,48,48,0,0,1,1,0,0,9,102,49,48,0,0,1,1,0,0,9,102,50,48,0,0,1,1,0,0,9,
-102,51,48,0,0,1,1,0,0,9,102,48,49,0,0,1,1,0,0,9,102,49,49,0,0,1,1,0,0,9,102,50,49,0,0,1,1,0,0,9,
-102,51,49,0,0,1,1,0,0,9,102,48,50,0,0,1,1,0,0,9,102,49,50,0,0,1,1,0,0,9,102,50,50,0,0,1,1,0,0,9,
-102,51,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,
-18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,
-57,59,119,0,18,102,51,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,
-48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,49,0,57,59,119,0,18,102,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,
-59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,102,49,
-50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,102,50,50,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,16,10,50,0,57,59,119,0,18,102,51,50,0,20,0,0,1,90,95,0,0,30,1,1,1,0,0,9,
-102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,102,0,0,17,48,0,48,0,0,
-0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,
-17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,30,1,1,1,0,0,5,
-105,0,0,0,1,3,2,90,95,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101,
-116,86,97,108,0,58,109,97,116,51,120,52,0,18,102,0,0,0,20,0,0,1,90,95,0,0,30,1,1,1,0,0,1,98,0,0,0,
-1,3,2,90,95,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,
-108,0,58,109,97,116,51,120,52,0,18,102,0,0,0,20,0,0,1,90,95,0,0,30,1,1,1,0,0,12,99,48,0,0,1,1,0,0,
-12,99,49,0,0,1,1,0,0,12,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,
-20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,90,95,0,0,29,1,1,1,0,0,9,102,48,48,0,0,1,1,0,0,9,102,49,48,
-0,0,1,1,0,0,9,102,48,49,0,0,1,1,0,0,9,102,49,49,0,0,1,1,0,0,9,102,48,50,0,0,1,1,0,0,9,102,49,50,0,
-0,1,1,0,0,9,102,48,51,0,0,1,1,0,0,9,102,49,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,
-57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,
-48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,
-59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0,18,102,48,
-51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,102,49,51,0,20,0,0,1,90,95,
-0,0,29,1,1,1,0,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,102,
-0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,16,10,52,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,
-0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,29,1,1,1,0,0,5,105,0,0,0,1,3,2,90,95,1,0,9,1,102,0,2,58,
-102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,
-18,102,0,0,0,20,0,0,1,90,95,0,0,29,1,1,1,0,0,1,98,0,0,0,1,3,2,90,95,1,0,9,1,102,0,2,58,102,108,111,
-97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,102,0,0,0,
-20,0,0,1,90,95,0,0,29,1,1,1,0,0,10,99,48,0,0,1,1,0,0,10,99,49,0,0,1,1,0,0,10,99,50,0,0,1,1,0,0,10,
-99,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,
-116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
-99,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,90,95,0,0,31,1,1,
-1,0,0,9,102,48,48,0,0,1,1,0,0,9,102,49,48,0,0,1,1,0,0,9,102,50,48,0,0,1,1,0,0,9,102,48,49,0,0,1,1,
-0,0,9,102,49,49,0,0,1,1,0,0,9,102,50,49,0,0,1,1,0,0,9,102,48,50,0,0,1,1,0,0,9,102,49,50,0,0,1,1,0,
-0,9,102,50,50,0,0,1,1,0,0,9,102,48,51,0,0,1,1,0,0,9,102,49,51,0,0,1,1,0,0,9,102,50,51,0,0,0,1,9,18,
+57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114,
+101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,9,0,102,48,48,0,0,1,
+1,0,0,9,0,102,49,48,0,0,1,1,0,0,9,0,102,50,48,0,0,1,1,0,0,9,0,102,48,49,0,0,1,1,0,0,9,0,102,49,49,
+0,0,1,1,0,0,9,0,102,50,49,0,0,1,1,0,0,9,0,102,48,50,0,0,1,1,0,0,9,0,102,49,50,0,0,1,1,0,0,9,0,102,
+50,50,0,0,1,1,0,0,9,0,102,48,51,0,0,1,1,0,0,9,0,102,49,51,0,0,1,1,0,0,9,0,102,50,51,0,0,0,1,9,18,
 95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,
 97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,
 59,122,0,18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,
@@ -101,645 +104,661 @@
 59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,102,50,
 50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0,18,102,48,51,0,20,0,9,18,95,95,
 114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,102,49,51,0,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,51,0,57,59,122,0,18,102,50,51,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,9,102,0,0,0,1,9,18,95,
-95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,
-17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,
-48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,5,105,0,0,0,1,3,2,90,
-95,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,
-109,97,116,52,120,51,0,18,102,0,0,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,1,98,0,0,0,1,3,2,90,95,1,0,9,1,
-102,0,2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,
-52,120,51,0,18,102,0,0,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,11,99,48,0,0,1,1,0,0,11,99,49,0,0,1,1,0,0,
-11,99,50,0,0,1,1,0,0,11,99,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,
-20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,
-0,0,1,90,95,0,0,13,1,1,1,0,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,
-90,95,0,0,13,1,1,1,0,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,
-0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,13,1,1,1,0,0,29,109,0,0,0,1,9,18,95,
-95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,
-20,0,0,1,90,95,0,0,13,1,1,1,0,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,
+108,0,16,10,51,0,57,59,122,0,18,102,50,51,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,
+95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,
+0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,
+48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,5,0,105,0,0,0,
+1,3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,
+97,108,0,58,109,97,116,52,120,51,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,1,0,98,0,0,0,1,
+3,2,90,95,1,0,9,0,1,102,0,2,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,
+108,0,58,109,97,116,52,120,51,0,0,18,102,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,11,0,99,48,0,0,1,
+1,0,0,11,0,99,49,0,0,1,1,0,0,11,0,99,50,0,0,1,1,0,0,11,0,99,51,0,0,0,1,9,18,95,95,114,101,116,86,
+97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,
+20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,
+116,86,97,108,0,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,
+90,95,0,0,13,0,1,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,0,
+18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,26,0,109,0,0,
+0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,
+18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,
+95,114,101,116,86,97,108,0,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,
+10,49,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,
+116,86,97,108,0,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,
+59,120,121,0,0,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,
+0,0,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,
+97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,
+1,90,95,0,0,13,0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,
 0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,
-13,1,1,1,0,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,
-0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,13,1,1,1,0,0,14,109,
-0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,
-18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,13,1,1,1,0,0,30,109,0,0,0,1,9,18,95,95,
-114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,
-0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,13,1,1,1,0,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,
-0,0,20,0,0,1,90,95,0,0,13,1,1,1,0,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,
-116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,90,
-95,0,0,26,1,1,1,0,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,
-26,1,1,1,0,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,
-16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,26,1,1,1,0,0,31,109,0,0,0,1,9,18,95,
-95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,
-57,0,0,20,0,0,1,90,95,0,0,26,1,1,1,0,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,
-116,50,120,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,
-0,20,0,0,1,90,95,0,0,26,1,1,1,0,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,
-50,120,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,0,
-20,0,0,1,90,95,0,0,26,1,1,1,0,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,
-120,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,0,20,0,
-0,1,90,95,0,0,26,1,1,1,0,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,
-51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0,0,18,109,0,
-16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,
-26,1,1,1,0,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,
-16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,
-59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,26,1,1,1,0,0,29,
-109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,59,
-120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,
-0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,28,109,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,30,109,0,0,0,1,9,18,95,95,
-114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,
-0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,
-50,120,52,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,26,
-109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,
-120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109,
-0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17,
-48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
-58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,
-0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,
-57,59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,
-31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,
-59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,
-109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,
-17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
-0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,
-48,0,48,0,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,
-17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,27,109,0,0,0,1,9,18,95,95,114,
-101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,
-57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,
-49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,1,1,1,0,0,29,109,0,0,
-0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,
+26,0,1,1,1,0,0,26,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,26,
+0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,0,18,109,
+0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,31,0,109,0,0,0,1,9,
+18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,
+10,49,0,57,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
+0,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,
+120,121,122,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,
+59,120,121,122,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,
+97,108,0,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,
+57,59,120,121,122,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,
+86,97,108,0,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,
+121,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,
+0,48,0,0,0,0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
+58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,
+48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,
+0,20,0,0,1,90,95,0,0,26,0,1,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,
+116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0,
+0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,0,20,0,0,1,
+90,95,0,0,28,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,
+95,0,0,28,0,1,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,
+0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,15,0,109,
+0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,0,18,
+109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,26,0,109,0,0,0,1,9,18,95,95,114,101,116,
+86,97,108,0,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,
+121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,
+0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,
+28,0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,
+109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,
+48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,
+49,0,57,59,122,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,
+95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,
+16,8,48,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,
+59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17,48,0,48,0,0,0,0,
+20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,
+116,50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,
+0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,
+48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,
+116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,
+59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,
+57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,28,0,1,1,1,0,0,29,0,109,0,0,0,
+1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,
 18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,
-0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,27,1,1,
-1,0,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,
-29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,
-0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,14,109,0,0,0,
-1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,
-10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,30,109,0,0,0,1,9,18,95,95,
-114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,
-48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,
-10,50,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,121,0,0,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,31,109,0,
-0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,
-0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,
-0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,121,0,0,0,20,0,0,1,90,95,0,0,27,1,
-1,1,0,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,
-48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,
-10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,13,109,0,
-0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,0,18,109,
-0,16,10,49,0,57,0,58,118,101,99,50,0,17,48,0,48,0,0,0,0,0,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,26,109,
-0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,
+0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,27,0,1,
+1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,27,0,1,1,1,
+0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,0,18,109,0,16,8,
+48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,14,
+0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,
+0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,30,0,109,0,
+0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,59,120,
 0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,
-121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,27,1,1,1,0,0,28,109,0,0,0,1,9,18,95,
-95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,
-8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,
-48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,0,0,14,1,1,1,0,0,14,109,0,0,0,1,9,18,95,95,114,101,116,
-86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,14,1,1,1,0,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,
-0,20,0,0,1,90,95,0,0,14,1,1,1,0,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,
-51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,
-16,10,50,0,57,59,120,121,122,0,0,0,20,0,0,1,90,95,0,0,14,1,1,1,0,0,15,109,0,0,0,1,9,18,95,95,114,
-101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,
-0,57,59,120,121,122,0,0,18,109,0,16,10,50,0,57,59,120,121,122,0,0,0,20,0,0,1,90,95,0,0,14,1,1,1,0,
-0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,18,
-109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,14,1,1,1,0,0,
-28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,59,120,
-121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,
-20,0,0,1,90,95,0,0,14,1,1,1,0,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,
-0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,
-57,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,14,1,1,1,0,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,
-0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,14,1,1,1,0,0,13,109,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,
-10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,30,1,1,1,
-0,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,30,1,1,1,0,0,15,
-109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,
-18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,30,1,1,1,0,0,14,109,0,0,0,1,
-9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,
-18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,
-30,1,1,1,0,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,
-16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,
-0,0,0,0,0,20,0,0,1,90,95,0,0,30,1,1,1,0,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,
-97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,
-49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,30,1,1,1,0,0,26,109,0,0,0,1,9,18,95,95,114,101,116,
-86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,
-0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,30,1,
-1,1,0,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,
-48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,
-0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,30,1,1,1,0,0,29,109,0,0,0,1,9,
-18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,
-17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,
-0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,30,1,1,1,0,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,
-97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,
-10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,0,
-20,0,0,1,90,95,0,0,29,1,1,1,0,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,
-1,90,95,0,0,29,1,1,1,0,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,
-0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,
-57,59,120,121,0,0,18,109,0,16,10,51,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,29,1,1,1,0,0,15,109,0,
-0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,
-121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,18,109,0,16,10,
-51,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,29,1,1,1,0,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,
-97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,
-17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,1,1,1,0,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
-58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,
-0,18,109,0,16,10,50,0,57,59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,1,1,1,
-0,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,
+121,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,121,0,0,0,20,0,0,1,90,95,0,0,
+27,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,0,18,
+109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,
+18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,121,
+0,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,
+97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,
+10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,
+1,90,95,0,0,27,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,
+120,50,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,58,118,101,99,50,0,0,17,48,0,48,0,0,0,
+0,0,0,20,0,0,1,90,95,0,0,27,0,1,1,1,0,0,26,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,
+97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,
+10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,
+1,90,95,0,0,27,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,
+120,50,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,
+57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,90,95,
+0,0,14,0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,
+14,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,109,0,
+16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,
+0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,
+59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10,50,0,57,59,120,121,122,
+0,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,
+97,116,51,0,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,
+18,109,0,16,10,50,0,57,59,120,121,122,0,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,26,0,109,0,0,0,1,9,
+18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,
+57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,28,0,109,0,0,0,
+1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,
+18,109,0,16,10,49,0,57,59,120,121,122,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,
+95,0,0,14,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,
+109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,
+17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,
+0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,13,0,109,0,0,0,1,
+9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,
+109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,
+30,0,1,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,30,
+0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,109,
+0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,
+0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,109,0,16,8,
+48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,
+0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,
+97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,
+18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,
+95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,
+49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,
+0,0,26,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,109,0,16,8,
+48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,
+0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,
+97,108,0,58,109,97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,
+16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,0,
+20,0,0,1,90,95,0,0,30,0,1,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,
+116,51,120,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,
+48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,
+30,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,0,18,
+109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,
+0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,29,
+0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,31,0,
+109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,
+59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,18,109,
+0,16,10,51,0,57,59,120,121,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,
+10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,18,109,0,16,10,51,0,57,59,120,121,
+0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,
+97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,
+0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,
+116,52,120,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,
+109,0,16,10,50,0,57,59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,
+30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,
 57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,17,
-48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,1,1,1,0,0,13,109,0,0,0,1,9,18,95,95,114,101,116,
-86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,
-0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,1,1,1,0,0,26,109,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,
-109,0,16,10,49,0,57,59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,
-0,1,90,95,0,0,29,1,1,1,0,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,
-50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,17,48,0,0,0,0,17,
-48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,31,109,0,0,0,1,9,18,95,95,
-114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,15,109,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,
-49,0,57,59,120,121,122,0,0,18,109,0,16,10,50,0,57,59,120,121,122,0,0,18,109,0,16,10,51,0,57,59,120,
-121,122,0,0,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,
-109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,
-17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,30,109,0,0,0,1,9,18,95,
-95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,
-109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10,50,0,57,59,120,121,122,0,0,17,48,0,0,0,0,17,
-48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,29,109,0,0,0,1,9,18,95,95,114,101,116,
-86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,
-0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,18,109,0,16,10,51,0,57,0,17,48,0,0,0,0,0,20,
-0,0,1,90,95,0,0,31,1,1,1,0,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,
-120,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,
-0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,27,109,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,
-109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,
-0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,
-120,121,122,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,
-0,0,20,0,0,1,90,95,0,0,31,1,1,1,0,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,
-116,52,120,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,
-0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,
-15,1,1,1,0,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,15,1,1,
-1,0,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,
-18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,
-0,0,0,0,0,20,0,0,1,90,95,0,0,15,1,1,1,0,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,
-97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,
-16,10,50,0,57,0,17,48,0,0,0,0,18,109,0,16,10,51,0,57,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,1,1,1,
-0,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,
-18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,
-48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,1,1,1,0,0,29,109,0,0,0,1,9,18,95,95,
-114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,
-109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,
-0,0,18,109,0,16,10,51,0,57,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,1,1,1,0,0,14,109,
-0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,
-0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,
-48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,1,1,1,0,0,26,109,0,0,0,1,9,18,95,95,
-114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,
-0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,
-0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,1,1,1,0,0,27,109,0,0,0,1,9,18,95,95,114,
-101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,
-0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,
-17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,1,1,1,0,0,13,109,0,
-0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,
-17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,
+48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,
+116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,
+0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,26,0,109,0,0,
+0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,59,120,
+121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,
+0,0,0,20,0,0,1,90,95,0,0,29,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,
+97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,17,
+48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,31,0,109,
+0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,15,0,109,0,0,
+0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,59,120,
+121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10,50,0,57,59,120,121,122,0,0,18,
+109,0,16,10,51,0,57,59,120,121,122,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,14,0,109,0,0,0,1,9,18,
+95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,
+49,0,57,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,
+31,0,1,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,
+109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10,50,
+0,57,59,120,121,122,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,
+0,0,29,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,109,0,16,8,
+48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,
+0,0,18,109,0,16,10,51,0,57,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,26,0,109,0,0,0,1,9,
+18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,
+10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,
+20,0,0,1,90,95,0,0,31,0,1,1,1,0,0,27,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,
+116,52,120,51,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,
+109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,
+31,0,1,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,
+109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,17,48,0,0,0,0,17,
+48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,31,0,1,1,1,
+0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,0,18,109,0,16,8,
+48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,
+0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,15,0,109,0,0,0,
+1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,30,0,109,0,0,0,1,9,
+18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,
+57,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,
+90,95,0,0,15,0,1,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,
+18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,
+57,0,17,48,0,0,0,0,18,109,0,16,10,51,0,57,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,28,0,
+109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,18,109,
+0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,
+0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,
+18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,
+0,0,0,0,18,109,0,16,10,51,0,57,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,
+14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,
+17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,
+0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,26,0,109,0,0,
+0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,
+18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,
+48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,27,0,109,
+0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,0,17,48,0,0,
+0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,
 49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,90,95,
-0,0,0,2,1,1,0,2,0,26,109,0,0,1,1,0,0,26,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,
-21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,90,95,0,0,0,2,1,1,0,2,0,28,109,0,0,1,
-1,0,0,28,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,
-18,110,0,16,10,49,0,57,21,0,0,1,90,95,0,0,0,2,1,1,0,2,0,27,109,0,0,1,1,0,0,27,110,0,0,0,1,9,18,109,
-0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,
-18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,90,95,0,0,0,2,1,1,0,2,0,30,109,0,0,1,1,0,0,
-30,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,
-0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,90,95,0,0,0,2,1,1,0,
-2,0,29,109,0,0,1,1,0,0,29,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,
-0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,
-18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,90,95,0,0,0,2,1,1,0,2,0,31,109,0,0,1,1,0,0,
-31,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,
-0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10,51,0,57,
-18,110,0,16,10,51,0,57,21,0,0,1,90,95,0,0,0,2,2,1,0,2,0,26,109,0,0,1,1,0,0,26,110,0,0,0,1,9,18,109,
-0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,
-90,95,0,0,0,2,2,1,0,2,0,28,109,0,0,1,1,0,0,28,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,
-0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,90,95,0,0,0,2,2,1,0,2,0,27,109,
-0,0,1,1,0,0,27,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,
-0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,90,95,0,
-0,0,2,2,1,0,2,0,30,109,0,0,1,1,0,0,30,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,
-0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,
-0,57,22,0,0,1,90,95,0,0,0,2,2,1,0,2,0,29,109,0,0,1,1,0,0,29,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,
-110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,
-57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,22,0,0,1,90,95,0,0,
-0,2,2,1,0,2,0,31,109,0,0,1,1,0,0,31,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,
+0,0,15,0,1,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,0,18,109,
+0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,
+17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,
+17,49,0,0,0,0,0,20,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,109,
+0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,
+90,95,0,0,0,0,2,1,1,0,2,0,28,0,109,0,0,1,1,0,0,28,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,
+16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,
+0,27,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,
+109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,
+0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,30,0,109,0,0,1,1,0,0,30,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,
+110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,
+57,18,110,0,16,10,50,0,57,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,29,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,
+9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,
+21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,
+51,0,57,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,31,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,109,0,16,8,
+48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,
+16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,
+90,95,0,0,0,0,2,2,1,0,2,0,26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,
+16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,
+0,28,0,109,0,0,1,1,0,0,28,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,
+109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,27,0,109,0,0,1,1,0,0,
+27,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,
+110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,90,95,0,0,0,0,2,
+2,1,0,2,0,30,0,109,0,0,1,1,0,0,30,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,
 9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,
-57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,22,0,0,1,90,95,0,0,0,2,4,1,0,2,0,26,109,0,
-0,1,1,0,0,26,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,
-57,18,110,0,16,10,49,0,57,24,0,0,1,90,95,0,0,0,2,4,1,0,2,0,28,109,0,0,1,1,0,0,28,110,0,0,0,1,9,18,
-109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,
-0,1,90,95,0,0,0,2,4,1,0,2,0,27,109,0,0,1,1,0,0,27,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,
-8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,
-110,0,16,10,50,0,57,24,0,0,1,90,95,0,0,0,2,4,1,0,2,0,30,109,0,0,1,1,0,0,30,110,0,0,0,1,9,18,109,0,
-16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,
-109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,90,95,0,0,0,2,4,1,0,2,0,29,109,0,0,1,1,0,0,29,
-110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,
-16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,
-18,110,0,16,10,51,0,57,24,0,0,1,90,95,0,0,0,2,4,1,0,2,0,31,109,0,0,1,1,0,0,31,110,0,0,0,1,9,18,109,
-0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,
-18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,
-24,0,0,1,90,95,0,0,11,2,21,1,1,0,0,26,109,0,0,1,1,0,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,
-10,49,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,
-109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,20,0,9,18,
-95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118,
-0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,20,0,0,1,90,95,0,0,12,2,21,1,1,0,0,28,109,0,0,1,1,
-0,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,48,
-0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,
-116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,
-109,0,16,10,49,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,
+57,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,29,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,9,18,109,0,16,8,48,0,
+57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,
+10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,22,0,0,1,90,
+95,0,0,0,0,2,2,1,0,2,0,31,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,
+48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,
+0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,22,0,0,1,90,95,0,0,0,0,2,4,1,
+0,2,0,26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,
+18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,28,0,109,0,0,1,1,
+0,0,28,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,
+18,110,0,16,10,49,0,57,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,27,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,9,
+18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,
+24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,30,0,109,0,
+0,1,1,0,0,30,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,
+0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,90,95,0,
+0,0,0,2,4,1,0,2,0,29,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,
+57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,
+10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,
+31,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,
+0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,9,
+18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,90,95,0,0,11,0,2,21,1,1,0,0,26,0,109,0,0,1,
+1,0,0,10,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,
+48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,20,0,9,18,95,95,114,
+101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,
+18,109,0,16,10,49,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,
 120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,20,
-0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,119,0,48,
-18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,119,0,48,46,20,0,0,1,90,95,0,0,10,2,21,1,1,0,0,27,109,
-0,0,1,1,0,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,
-16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,18,118,0,59,122,0,
-18,109,0,16,10,50,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,
-120,0,18,109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,
-118,0,59,122,0,18,109,0,16,10,50,0,57,59,121,0,48,46,20,0,0,1,90,95,0,0,12,2,21,1,1,0,0,30,109,0,0,
-1,1,0,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,
-48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,18,118,0,59,122,0,18,
-109,0,16,10,50,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,
-120,0,18,109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,
-118,0,59,122,0,18,109,0,16,10,50,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,
-122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,
-59,122,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,122,0,48,46,20,0,9,18,95,95,114,101,116,
-86,97,108,0,59,119,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,119,0,48,18,118,0,59,121,0,18,109,
-0,16,10,49,0,57,59,119,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,119,0,48,46,20,0,0,1,90,
-95,0,0,10,2,21,1,1,0,0,29,109,0,0,1,1,0,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,
-0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,
-120,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48,46,18,118,0,59,119,0,18,109,0,16,
-10,51,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,
-109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,
-122,0,18,109,0,16,10,50,0,57,59,121,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,121,0,48,
-46,20,0,0,1,90,95,0,0,11,2,21,1,1,0,0,31,109,0,0,1,1,0,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,
+0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,28,0,109,0,0,1,1,0,0,10,0,118,0,0,0,1,9,18,95,95,114,101,116,86,
 97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,
-16,10,49,0,57,59,120,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48,46,18,118,0,59,
-119,0,18,109,0,16,10,51,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,
-118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,
-48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,121,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,
-0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,
-8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,18,118,0,59,122,0,18,
-109,0,16,10,50,0,57,59,122,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,122,0,48,46,20,0,0,
-1,90,95,0,0,27,2,21,1,1,0,0,13,109,0,0,1,1,0,0,27,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
-16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
-57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,
-0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,29,2,21,1,1,0,0,13,109,0,0,1,1,0,0,29,110,0,0,0,1,9,
-18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,
-114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,
-116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,2,21,1,1,0,0,26,109,0,
-0,1,1,0,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,
-0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,
-20,0,0,1,90,95,0,0,14,2,21,1,1,0,0,26,109,0,0,1,1,0,0,27,110,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,
-49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,
-18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,31,2,21,1,1,0,0,26,109,0,0,1,1,0,0,29,110,0,
-0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,28,2,21,1,1,0,0,28,
-109,0,0,1,1,0,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,
-16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,
-57,48,20,0,0,1,90,95,0,0,30,2,21,1,1,0,0,28,109,0,0,1,1,0,0,27,110,0,0,0,1,9,18,95,95,114,101,116,
-86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,
-16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,
-0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,15,2,21,1,1,0,0,28,109,0,0,1,1,0,0,29,
-110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,
-9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,13,2,21,1,
-1,0,0,27,109,0,0,1,1,0,0,26,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,
+16,10,49,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,
+109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,20,0,9,18,
+95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118,
+0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,
+18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,119,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,
+119,0,48,46,20,0,0,1,90,95,0,0,10,0,2,21,1,1,0,0,27,0,109,0,0,1,1,0,0,11,0,118,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,
+121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48,
+46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,
+0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,
+57,59,121,0,48,46,20,0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,30,0,109,0,0,1,1,0,0,11,0,118,0,0,0,1,9,18,
+95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,
+0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,
+48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,
+121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,122,0,18,109,0,16,10,
+50,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,
+16,8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,18,118,0,59,122,0,
+18,109,0,16,10,50,0,57,59,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,118,0,59,
+120,0,18,109,0,16,8,48,0,57,59,119,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,119,0,48,46,18,
+118,0,59,122,0,18,109,0,16,10,50,0,57,59,119,0,48,46,20,0,0,1,90,95,0,0,10,0,2,21,1,1,0,0,29,0,109,
+0,0,1,1,0,0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,
+0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,18,118,0,59,122,
+0,18,109,0,16,10,50,0,57,59,120,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,120,0,48,46,20,
+0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,48,
+18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,
+59,121,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,121,0,48,46,20,0,0,1,90,95,0,0,11,0,2,
+21,1,1,0,0,31,0,109,0,0,1,1,0,0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,
+118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,
+48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,
+0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,
+8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,122,0,18,
+109,0,16,10,50,0,57,59,121,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,121,0,48,46,20,0,9,
+18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,
+118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,
+122,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,122,0,48,46,20,0,0,1,90,95,0,0,27,0,2,21,1,
+1,0,0,13,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,
+0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,
+16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,
+0,57,48,20,0,0,1,90,95,0,0,29,0,2,21,1,1,0,0,13,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,
+97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,
+16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,
+0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,0,2,21,1,1,0,0,26,0,109,0,0,1,1,0,0,
+13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,
+20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,
+90,95,0,0,14,0,2,21,1,1,0,0,26,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
+0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,
+0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
+109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,31,0,2,21,1,1,0,0,26,0,109,0,0,1,1,0,0,29,0,110,
+0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,
+18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,28,0,2,21,1,1,
+0,0,28,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,
 18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,
-16,10,49,0,57,48,20,0,0,1,90,95,0,0,27,2,21,1,1,0,0,27,109,0,0,1,1,0,0,14,110,0,0,0,1,9,18,95,95,
-114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,
-0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,29,2,21,1,1,0,0,27,109,0,0,1,
-1,0,0,31,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,
-57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,
-0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,
-2,21,1,1,0,0,14,109,0,0,1,1,0,0,26,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
-109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,
-110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,31,2,21,1,1,0,0,14,109,0,0,1,1,0,0,31,110,0,0,0,1,9,18,
+16,10,49,0,57,48,20,0,0,1,90,95,0,0,30,0,2,21,1,1,0,0,28,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,9,18,
 95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,
 101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,
-0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,28,2,21,1,1,0,0,30,109,0,0,1,
-1,0,0,26,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,
+86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,15,0,2,21,1,1,0,0,
+28,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,
+110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,
+10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,
+57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,
+0,0,1,90,95,0,0,13,0,2,21,1,1,0,0,27,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,95,95,114,101,116,86,
+97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
+10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,27,0,2,21,1,1,0,0,27,0,109,0,0,1,
+1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,
 57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,
-0,0,1,90,95,0,0,30,2,21,1,1,0,0,30,109,0,0,1,1,0,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
-0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,
-0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
-109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,15,2,21,1,1,0,0,30,109,0,0,1,1,0,0,31,110,0,0,0,
-1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,13,2,21,1,1,0,0,29,
-109,0,0,1,1,0,0,28,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,
-16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,
-57,48,20,0,0,1,90,95,0,0,27,2,21,1,1,0,0,29,109,0,0,1,1,0,0,30,110,0,0,0,1,9,18,95,95,114,101,116,
+0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,
+95,0,0,29,0,2,21,1,1,0,0,27,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
+16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
+57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,
+0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,
+0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,0,2,21,1,1,0,0,14,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,
+95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,
+101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,31,0,2,21,
+1,1,0,0,14,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
+109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,
+110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,
+10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,
+57,48,20,0,0,1,90,95,0,0,28,0,2,21,1,1,0,0,30,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,
+97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,30,0,2,21,1,1,0,0,30,
+0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,
+0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,
+0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,
+20,0,0,1,90,95,0,0,15,0,2,21,1,1,0,0,30,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,9,18,95,95,114,101,116,
 86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,
 16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,
-0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,29,2,21,1,1,0,0,29,109,0,0,1,1,0,0,15,
-110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,
-9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,2,21,1,
-1,0,0,31,109,0,0,1,1,0,0,28,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,
-18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,
-16,10,49,0,57,48,20,0,0,1,90,95,0,0,14,2,21,1,1,0,0,31,109,0,0,1,1,0,0,30,110,0,0,0,1,9,18,95,95,
-114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,
-0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,31,2,21,1,1,0,0,31,109,0,0,1,
-1,0,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,
-57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,
-0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,28,
-2,21,1,1,0,0,15,109,0,0,1,1,0,0,28,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
-109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,
-110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,30,2,21,1,1,0,0,15,109,0,0,1,1,0,0,30,110,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,0,2,3,1,0,2,0,26,
-109,0,0,1,1,0,0,13,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90,95,0,0,0,2,3,1,0,2,0,28,
-109,0,0,1,1,0,0,13,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90,95,0,0,0,2,3,1,0,2,0,27,
-109,0,0,1,1,0,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90,95,0,0,0,2,3,1,0,2,0,30,
-109,0,0,1,1,0,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90,95,0,0,0,2,3,1,0,2,0,29,
-109,0,0,1,1,0,0,15,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90,95,0,0,0,2,3,1,0,2,0,31,
-109,0,0,1,1,0,0,15,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90,95,0,0,11,2,21,1,1,0,0,
-10,118,0,0,1,1,0,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,
-118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,
-0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,
-111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,12,2,21,1,1,0,0,10,118,0,0,1,1,
-0,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,
-0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,
-109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,
-0,18,109,0,16,10,50,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,
-118,0,0,18,109,0,16,10,51,0,57,0,0,20,0,0,1,90,95,0,0,10,2,21,1,1,0,0,11,118,0,0,1,1,0,0,26,109,0,
-0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,
-57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,
-49,0,57,0,0,20,0,0,1,90,95,0,0,12,2,21,1,1,0,0,11,118,0,0,1,1,0,0,31,109,0,0,0,1,9,18,95,95,114,
-101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,
-95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,
-9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,
-20,0,0,1,90,95,0,0,10,2,21,1,1,0,0,12,118,0,0,1,1,0,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,
-86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,11,
-2,21,1,1,0,0,12,118,0,0,1,1,0,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,
-111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,
-100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,
-0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,0,2,1,1,0,2,0,26,109,0,
-0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,
-0,1,90,95,0,0,0,2,1,1,0,2,0,28,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,
-18,109,0,16,10,49,0,57,18,97,0,21,0,0,1,90,95,0,0,0,2,1,1,0,2,0,27,109,0,0,1,1,0,0,9,97,0,0,0,1,9,
-18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,
-18,97,0,21,0,0,1,90,95,0,0,0,2,1,1,0,2,0,30,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,
-18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,90,95,
-0,0,0,2,1,1,0,2,0,29,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,
-16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,
-21,0,0,1,90,95,0,0,0,2,1,1,0,2,0,31,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,
-21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10,
-51,0,57,18,97,0,21,0,0,1,90,95,0,0,0,2,2,1,0,2,0,26,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,
-48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,90,95,0,0,0,2,2,1,0,2,0,28,109,0,0,
-1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,
-1,90,95,0,0,0,2,2,1,0,2,0,27,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,
-18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,90,95,0,0,0,2,2,1,0,
-2,0,30,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,
-18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,90,95,0,0,0,2,2,1,0,2,0,29,109,0,0,1,1,0,0,
-9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,
-16,10,50,0,57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,90,95,0,0,0,2,2,1,0,2,0,31,
-109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,
-22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,90,95,0,0,0,2,
-3,1,0,2,0,26,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,
-0,57,18,97,0,23,0,0,1,90,95,0,0,0,2,3,1,0,2,0,28,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,
-57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,90,95,0,0,0,2,3,1,0,2,0,27,109,0,0,1,1,0,
-0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,
-0,16,10,50,0,57,18,97,0,23,0,0,1,90,95,0,0,0,2,3,1,0,2,0,30,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,
-0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,
-23,0,0,1,90,95,0,0,0,2,3,1,0,2,0,29,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,
-23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,9,18,109,0,16,10,
-51,0,57,18,97,0,23,0,0,1,90,95,0,0,0,2,3,1,0,2,0,31,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,
-48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,9,
-18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,90,95,0,0,0,2,4,1,0,2,0,26,109,0,0,1,1,0,0,9,97,0,0,0,1,9,
-18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,90,95,0,0,0,2,4,1,0,2,
-0,28,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,
-97,0,24,0,0,1,90,95,0,0,0,2,4,1,0,2,0,27,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,
-97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,90,95,0,
-0,0,2,4,1,0,2,0,30,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,
-10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,90,95,0,0,0,2,4,1,0,2,0,29,109,0,
-0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,
-9,18,109,0,16,10,50,0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,90,95,0,0,0,2,4,1,
-0,2,0,31,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,
-57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,90,
-95,0,0,26,2,26,1,1,0,0,26,109,0,0,1,1,0,0,26,110,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,
-48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,0,0,0,1,90,
-95,0,0,28,2,26,1,1,0,0,28,109,0,0,1,1,0,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,
-48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,0,0,0,1,90,
-95,0,0,27,2,26,1,1,0,0,27,109,0,0,1,1,0,0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,
-48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,0,16,
-10,50,0,57,18,110,0,16,10,50,0,57,46,0,0,0,0,1,90,95,0,0,30,2,26,1,1,0,0,30,109,0,0,1,1,0,0,30,110,
-0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,
-10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,0,0,0,1,
-90,95,0,0,29,2,26,1,1,0,0,29,109,0,0,1,1,0,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,
-16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,
-0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,0,0,0,
-0,1,90,95,0,0,31,2,26,1,1,0,0,31,109,0,0,1,1,0,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,
-0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,
-109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,0,
-0,0,0,1,90,95,0,0,26,2,27,1,1,0,0,26,109,0,0,1,1,0,0,26,110,0,0,0,1,8,58,109,97,116,50,120,51,0,18,
-109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,0,
-0,0,1,90,95,0,0,28,2,27,1,1,0,0,28,109,0,0,1,1,0,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,
-109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,0,
-0,0,1,90,95,0,0,27,2,27,1,1,0,0,27,109,0,0,1,1,0,0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18,
-109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,
-18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,0,0,0,1,90,95,0,0,30,2,27,1,1,0,0,30,109,0,0,1,
-1,0,0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,
-18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,
-47,0,0,0,0,1,90,95,0,0,29,2,27,1,1,0,0,29,109,0,0,1,1,0,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,
+0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,
+109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,13,0,2,21,1,1,0,0,29,0,109,0,0,1,1,0,0,28,0,110,
+0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,
+18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,
+0,27,0,2,21,1,1,0,0,29,0,109,0,0,1,1,0,0,30,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,
+48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,
+109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,
+110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,29,0,2,21,1,1,0,0,29,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,
+9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,0,2,21,1,1,0,0,31,0,
+109,0,0,1,1,0,0,28,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,
+16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,
+57,48,20,0,0,1,90,95,0,0,14,0,2,21,1,1,0,0,31,0,109,0,0,1,1,0,0,30,0,110,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,
+97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,
+16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,31,0,2,21,1,1,0,0,31,0,109,0,0,
+1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,
+0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,
+20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,
+95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,
+28,0,2,21,1,1,0,0,15,0,109,0,0,1,1,0,0,28,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,
+0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,
+109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,30,0,2,21,1,1,0,0,15,0,109,0,0,1,1,0,0,30,0,110,
+0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,
+18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,0,0,2,
+3,1,0,2,0,26,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90,95,0,0,
+0,0,2,3,1,0,2,0,28,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90,
+95,0,0,0,0,2,3,1,0,2,0,27,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,
+0,1,90,95,0,0,0,0,2,3,1,0,2,0,30,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,
+48,20,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,29,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,109,0,18,109,0,18,
+110,0,48,20,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,31,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,109,0,18,
+109,0,18,110,0,48,20,0,0,1,90,95,0,0,11,0,2,21,1,1,0,0,10,0,118,0,0,1,1,0,0,27,0,109,0,0,0,1,9,18,
+95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,
+0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,
+0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,
+50,0,57,0,0,20,0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,10,0,118,0,0,1,1,0,0,29,0,109,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,
+95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,
+0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,50,0,57,
+0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,
+51,0,57,0,0,20,0,0,1,90,95,0,0,10,0,2,21,1,1,0,0,11,0,118,0,0,1,1,0,0,26,0,109,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,
+95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,
+0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,11,0,118,0,0,1,1,0,0,31,0,109,0,0,0,1,9,18,95,95,114,101,116,86,
+97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,
+114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,
+18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,
+20,0,0,1,90,95,0,0,10,0,2,21,1,1,0,0,12,0,118,0,0,1,1,0,0,28,0,109,0,0,0,1,9,18,95,95,114,101,116,
+86,97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,
+101,116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,
+95,0,0,11,0,2,21,1,1,0,0,12,0,118,0,0,1,1,0,0,30,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
+59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,59,122,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,
+0,0,0,2,1,1,0,2,0,26,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,
+109,0,16,10,49,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,28,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,
+9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,2,1,1,
+0,2,0,27,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,
+0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,30,0,109,0,0,
+1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,
+9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,29,0,109,0,0,1,1,0,0,9,0,97,0,
+0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,
+50,0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,2,1,1,0,2,0,31,0,109,
+0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,
+21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,90,95,0,0,0,0,
+2,2,1,0,2,0,26,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,
+10,49,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,28,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,
+0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,27,
+0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,
+97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,30,0,109,0,0,1,1,0,0,
+9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,
+0,16,10,50,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,29,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,
+18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,
+18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,2,1,0,2,0,31,0,109,0,0,1,1,
+0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,
+109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,90,95,0,0,0,0,2,3,1,0,2,
+0,26,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,
+57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,28,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,
+48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,27,0,109,
+0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,
+23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,30,0,109,0,0,1,1,0,0,9,0,
+97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,
+16,10,50,0,57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,29,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,
+109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,
+97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,3,1,0,2,0,31,0,109,0,0,1,1,0,0,
+9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,
+0,16,10,50,0,57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,
+26,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,
+18,97,0,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,28,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,
+0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,27,0,109,0,0,
+1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,
+9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,30,0,109,0,0,1,1,0,0,9,0,97,0,
+0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,
+50,0,57,18,97,0,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,29,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,109,0,
+16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,
+24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,90,95,0,0,0,0,2,4,1,0,2,0,31,0,109,0,0,1,1,0,0,9,0,
+97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,
+16,10,50,0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,90,95,0,0,26,0,2,26,1,1,0,0,
+26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,
+16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,0,0,0,1,90,95,0,0,28,0,2,26,1,
+1,0,0,28,0,109,0,0,1,1,0,0,28,0,110,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,18,
+110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,0,0,0,1,90,95,0,0,27,0,
+2,26,1,1,0,0,27,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,
+0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,0,16,10,
+50,0,57,18,110,0,16,10,50,0,57,46,0,0,0,0,1,90,95,0,0,30,0,2,26,1,1,0,0,30,0,109,0,0,1,1,0,0,30,0,
+110,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,
+0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,0,0,
+0,1,90,95,0,0,29,0,2,26,1,1,0,0,29,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,8,58,109,97,116,52,120,50,0,
+0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,
+46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,
+57,46,0,0,0,0,1,90,95,0,0,31,0,2,26,1,1,0,0,31,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,8,58,109,97,116,
+52,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,
+10,49,0,57,46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,18,109,0,16,10,51,0,57,18,110,0,
+16,10,51,0,57,46,0,0,0,0,1,90,95,0,0,26,0,2,27,1,1,0,0,26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,8,58,
+109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,
+18,110,0,16,10,49,0,57,47,0,0,0,0,1,90,95,0,0,28,0,2,27,1,1,0,0,28,0,109,0,0,1,1,0,0,28,0,110,0,0,
+0,1,8,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,
+49,0,57,18,110,0,16,10,49,0,57,47,0,0,0,0,1,90,95,0,0,27,0,2,27,1,1,0,0,27,0,109,0,0,1,1,0,0,27,0,
+110,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,
+0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,0,0,
+0,1,90,95,0,0,30,0,2,27,1,1,0,0,30,0,109,0,0,1,1,0,0,30,0,110,0,0,0,1,8,58,109,97,116,51,120,52,0,
 0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,
-47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,
-57,47,0,0,0,0,1,90,95,0,0,31,2,27,1,1,0,0,31,109,0,0,1,1,0,0,31,110,0,0,0,1,8,58,109,97,116,52,120,
-51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,
-57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,18,109,0,16,10,51,0,57,18,110,0,16,10,
-51,0,57,47,0,0,0,0,1,90,95,0,0,26,2,22,1,1,0,0,26,109,0,0,1,1,0,0,26,110,0,0,0,1,8,58,109,97,116,
-50,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,
-49,0,57,49,0,0,0,0,1,90,95,0,0,28,2,22,1,1,0,0,28,109,0,0,1,1,0,0,28,110,0,0,0,1,8,58,109,97,116,
-50,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,
-49,0,57,49,0,0,0,0,1,90,95,0,0,27,2,22,1,1,0,0,27,109,0,0,1,1,0,0,27,110,0,0,0,1,8,58,109,97,116,
-51,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,
-49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,0,0,0,1,90,95,0,0,30,2,22,1,1,0,0,
-30,109,0,0,1,1,0,0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,
+47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,0,0,0,1,90,95,0,0,29,0,2,27,1,1,0,0,29,0,
+109,0,0,1,1,0,0,29,0,110,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,
+48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,
+10,50,0,57,47,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,0,0,0,0,1,90,95,0,0,31,0,2,27,1,1,
+0,0,31,0,109,0,0,1,1,0,0,31,0,110,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,18,
+110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,
+18,110,0,16,10,50,0,57,47,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,0,0,0,0,1,90,95,0,0,
+26,0,2,22,1,1,0,0,26,0,109,0,0,1,1,0,0,26,0,110,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,
+8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,0,0,0,1,90,
+95,0,0,28,0,2,22,1,1,0,0,28,0,109,0,0,1,1,0,0,28,0,110,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18,
+109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,0,
+0,0,1,90,95,0,0,27,0,2,22,1,1,0,0,27,0,109,0,0,1,1,0,0,27,0,110,0,0,0,1,8,58,109,97,116,51,120,50,
+0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,
+49,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,0,0,0,1,90,95,0,0,30,0,2,22,1,1,0,0,30,0,
+109,0,0,1,1,0,0,30,0,110,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,
 48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16,
-10,50,0,57,49,0,0,0,0,1,90,95,0,0,29,2,22,1,1,0,0,29,109,0,0,1,1,0,0,29,110,0,0,0,1,8,58,109,97,
-116,52,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,
-16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,18,109,0,16,10,51,0,57,18,
-110,0,16,10,51,0,57,49,0,0,0,0,1,90,95,0,0,31,2,22,1,1,0,0,31,109,0,0,1,1,0,0,31,110,0,0,0,1,8,58,
-109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,
+10,50,0,57,49,0,0,0,0,1,90,95,0,0,29,0,2,22,1,1,0,0,29,0,109,0,0,1,1,0,0,29,0,110,0,0,0,1,8,58,109,
+97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,
 110,0,16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,18,109,0,16,10,51,0,57,
-18,110,0,16,10,51,0,57,49,0,0,0,0,1,90,95,0,0,26,2,26,1,1,0,0,9,97,0,0,1,1,0,0,26,110,0,0,0,1,8,58,
-109,97,116,50,120,51,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,0,0,
-0,1,90,95,0,0,26,2,26,1,1,0,0,26,109,0,0,1,1,0,0,9,98,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,
-16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,28,2,26,1,1,0,0,9,
-97,0,0,1,1,0,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,
-0,18,110,0,16,10,49,0,57,46,0,0,0,0,1,90,95,0,0,28,2,26,1,1,0,0,28,109,0,0,1,1,0,0,9,98,0,0,0,1,8,
-58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,0,
-0,0,1,90,95,0,0,27,2,26,1,1,0,0,9,97,0,0,1,1,0,0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18,97,
-0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,
-0,0,0,0,1,90,95,0,0,27,2,26,1,1,0,0,27,109,0,0,1,1,0,0,9,98,0,0,0,1,8,58,109,97,116,51,120,50,0,18,
+18,110,0,16,10,51,0,57,49,0,0,0,0,1,90,95,0,0,31,0,2,22,1,1,0,0,31,0,109,0,0,1,1,0,0,31,0,110,0,0,
+0,1,8,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,
+49,0,57,18,110,0,16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,18,109,0,16,
+10,51,0,57,18,110,0,16,10,51,0,57,49,0,0,0,0,1,90,95,0,0,26,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,26,0,
+110,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,
+10,49,0,57,46,0,0,0,0,1,90,95,0,0,26,0,2,26,1,1,0,0,26,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,
+97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,0,0,0,
+1,90,95,0,0,28,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,28,0,110,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18,
+97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,0,0,0,1,90,95,0,0,28,0,2,26,1,
+1,0,0,28,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,18,
+98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,27,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,
+0,27,0,110,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,
+0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,0,0,0,1,90,95,0,0,27,0,2,26,1,1,0,0,27,0,
+109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,
+109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,30,0,2,26,1,
+1,0,0,9,0,97,0,0,1,1,0,0,30,0,110,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,97,0,18,110,0,16,8,48,0,
+57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,0,0,0,1,90,95,0,0,
+30,0,2,26,1,1,0,0,30,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,109,0,16,8,
+48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46,0,0,0,0,
+1,90,95,0,0,29,0,2,26,1,1,0,0,29,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18,
 109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,
-46,0,0,0,0,1,90,95,0,0,30,2,26,1,1,0,0,9,97,0,0,1,1,0,0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,
-18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,
-57,46,0,0,0,0,1,90,95,0,0,30,2,26,1,1,0,0,30,109,0,0,1,1,0,0,9,98,0,0,0,1,8,58,109,97,116,51,120,
-52,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,
-18,98,0,46,0,0,0,0,1,90,95,0,0,29,2,26,1,1,0,0,29,109,0,0,1,1,0,0,9,98,0,0,0,1,8,58,109,97,116,52,
-120,50,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,
-0,57,18,98,0,46,0,18,109,0,16,10,51,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,29,2,26,1,1,0,0,9,97,0,0,1,
-1,0,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,
-0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,18,97,0,18,110,0,16,10,51,0,57,46,0,0,0,0,
-1,90,95,0,0,31,2,26,1,1,0,0,31,109,0,0,1,1,0,0,9,98,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,
-16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46,0,
-18,109,0,16,10,51,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,31,2,26,1,1,0,0,9,97,0,0,1,1,0,0,31,110,0,0,
-0,1,8,58,109,97,116,52,120,51,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,
-46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,18,97,0,18,110,0,16,10,51,0,57,46,0,0,0,0,1,90,95,0,0,26,
-2,27,1,1,0,0,9,97,0,0,1,1,0,0,26,110,0,0,0,1,8,58,109,97,116,50,120,51,0,18,97,0,18,110,0,16,8,48,
-0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,0,0,0,1,90,95,0,0,26,2,27,1,1,0,0,26,109,0,0,1,1,0,0,
-9,98,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,
-18,98,0,47,0,0,0,0,1,90,95,0,0,28,2,27,1,1,0,0,9,97,0,0,1,1,0,0,28,110,0,0,0,1,8,58,109,97,116,50,
-120,52,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,0,0,0,1,90,95,0,0,
-28,2,27,1,1,0,0,28,109,0,0,1,1,0,0,9,98,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,
-18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,0,0,0,1,90,95,0,0,27,2,27,1,1,0,0,9,97,0,0,1,1,0,
-0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,
-16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,47,0,0,0,0,1,90,95,0,0,27,2,27,1,1,0,0,27,109,0,
-0,1,1,0,0,9,98,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,
-10,49,0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,0,0,0,1,90,95,0,0,30,2,27,1,1,0,0,9,97,
-0,0,1,1,0,0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,
-18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,47,0,0,0,0,1,90,95,0,0,30,2,27,1,1,0,0,
-30,109,0,0,1,1,0,0,9,98,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,
-109,0,16,10,49,0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,0,0,0,1,90,95,0,0,29,2,27,1,1,
-0,0,29,109,0,0,1,1,0,0,9,98,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,18,98,0,47,0,
-18,109,0,16,10,49,0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,18,109,0,16,10,51,0,57,18,
-98,0,47,0,0,0,0,1,90,95,0,0,29,2,27,1,1,0,0,9,97,0,0,1,1,0,0,29,110,0,0,0,1,8,58,109,97,116,52,120,
-50,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,
-50,0,57,47,0,18,97,0,18,110,0,16,10,51,0,57,47,0,0,0,0,1,90,95,0,0,31,2,27,1,1,0,0,31,109,0,0,1,1,
-0,0,9,98,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,
+46,0,18,109,0,16,10,51,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,29,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,29,
+0,110,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,
+10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,18,97,0,18,110,0,16,10,51,0,57,46,0,0,0,0,1,90,
+95,0,0,31,0,2,26,1,1,0,0,31,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,109,
+0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46,
+0,18,109,0,16,10,51,0,57,18,98,0,46,0,0,0,0,1,90,95,0,0,31,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,31,0,
+110,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,
+10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,18,97,0,18,110,0,16,10,51,0,57,46,0,0,0,0,1,90,
+95,0,0,26,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,26,0,110,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,97,0,
+18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,0,0,0,1,90,95,0,0,26,0,2,27,1,1,0,0,
+26,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,18,98,0,47,
+0,18,109,0,16,10,49,0,57,18,98,0,47,0,0,0,0,1,90,95,0,0,28,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,28,0,
+110,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,
+10,49,0,57,47,0,0,0,0,1,90,95,0,0,28,0,2,27,1,1,0,0,28,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,
+97,116,50,120,52,0,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,0,0,0,
+1,90,95,0,0,27,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,27,0,110,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,
+97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,
+47,0,0,0,0,1,90,95,0,0,27,0,2,27,1,1,0,0,27,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,8,58,109,97,116,51,
+120,50,0,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,18,109,0,16,10,
+50,0,57,18,98,0,47,0,0,0,0,1,90,95,0,0,30,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,30,0,110,0,0,0,1,8,58,
+109,97,116,51,120,52,0,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,
+97,0,18,110,0,16,10,50,0,57,47,0,0,0,0,1,90,95,0,0,30,0,2,27,1,1,0,0,30,0,109,0,0,1,1,0,0,9,0,98,0,
+0,0,1,8,58,109,97,116,51,120,52,0,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,
+98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,0,0,0,1,90,95,0,0,29,0,2,27,1,1,0,0,29,0,109,0,0,1,1,
+0,0,9,0,98,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,
+49,0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,18,109,0,16,10,51,0,57,18,98,0,47,0,0,0,0,
+1,90,95,0,0,29,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,29,0,110,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18,
+97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,
+47,0,18,97,0,18,110,0,16,10,51,0,57,47,0,0,0,0,1,90,95,0,0,31,0,2,27,1,1,0,0,31,0,109,0,0,1,1,0,0,
+9,0,98,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,
 0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,18,109,0,16,10,51,0,57,18,98,0,47,0,0,0,0,1,
-90,95,0,0,31,2,27,1,1,0,0,9,97,0,0,1,1,0,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,97,0,18,
-110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,47,0,18,
-97,0,18,110,0,16,10,51,0,57,47,0,0,0,0,1,90,95,0,0,26,2,21,1,1,0,0,9,97,0,0,1,1,0,0,26,110,0,0,0,1,
+90,95,0,0,31,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,31,0,110,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,
+97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,
+47,0,18,97,0,18,110,0,16,10,51,0,57,47,0,0,0,0,1,90,95,0,0,26,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,26,
+0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,
+0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,
+95,0,0,26,0,2,21,1,1,0,0,26,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,
+8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,
+18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,90,95,0,0,28,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,28,0,110,
+0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,
+95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,
+28,0,2,21,1,1,0,0,28,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,
+57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,
+16,10,49,0,57,18,98,0,48,20,0,0,1,90,95,0,0,27,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,27,0,110,0,0,0,1,
 9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,
-114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,26,2,21,
-1,1,0,0,26,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,
+114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,27,0,2,21,1,1,0,
+0,27,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,
+8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,
+18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,
+20,0,0,1,90,95,0,0,30,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,30,0,110,0,0,0,1,9,18,95,95,114,101,116,86,
+97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
+10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,
+18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,30,0,2,21,1,1,0,0,30,0,109,0,0,1,1,0,0,9,0,98,
+0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,
+95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,
+101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,90,95,0,0,29,0,2,21,1,
+1,0,0,29,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,
 16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,
-57,18,98,0,48,20,0,0,1,90,95,0,0,28,2,21,1,1,0,0,9,97,0,0,1,1,0,0,28,110,0,0,0,1,9,18,95,95,114,
-101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,
-97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,28,2,21,1,1,0,0,28,109,
-0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,
-98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,
-0,0,1,90,95,0,0,27,2,21,1,1,0,0,9,97,0,0,1,1,0,0,27,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
-16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
+57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,
+48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,
+1,90,95,0,0,29,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,29,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
+0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
 57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,
-18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,27,2,21,1,1,0,0,27,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,
-97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,90,95,0,0,30,2,21,1,1,0,0,9,97,0,
-0,1,1,0,0,30,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,
-0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,
-0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,
-95,0,0,30,2,21,1,1,0,0,30,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,
-0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,
-0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,
-0,57,18,98,0,48,20,0,0,1,90,95,0,0,29,2,21,1,1,0,0,29,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,
-101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,
-97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,
-16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,
-57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,1,90,95,0,0,29,2,21,1,1,0,0,9,97,0,0,1,1,0,0,29,110,0,
-0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,
-97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,31,2,21,1,1,0,0,31,109,
-0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,
-98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,
-0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,1,90,95,0,0,31,2,
-21,1,1,0,0,9,97,0,0,1,1,0,0,31,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,
-18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,
-10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,
-48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0,
-1,90,95,0,0,26,2,22,1,1,0,0,9,97,0,0,1,1,0,0,26,110,0,0,0,1,3,2,90,95,1,0,9,1,105,110,118,0,2,17,
-49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,
-0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,
-16,10,49,0,57,48,20,0,0,1,90,95,0,0,26,2,22,1,1,0,0,26,109,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,1,0,
-9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,
-18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,
-109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,28,2,22,1,1,0,0,9,97,0,0,1,1,0,0,28,110,
-0,0,0,1,3,2,90,95,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,
-97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,28,2,22,1,1,0,0,
-28,109,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,
-18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,0,1,90,
-95,0,0,27,2,22,1,1,0,0,9,97,0,0,1,1,0,0,27,110,0,0,0,1,3,2,90,95,1,0,9,1,105,110,118,0,2,17,49,0,
-48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,
-8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,
-49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,
-50,0,57,48,20,0,0,1,90,95,0,0,27,2,22,1,1,0,0,27,109,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,1,0,9,1,
+18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,
+16,10,51,0,57,48,20,0,0,1,90,95,0,0,31,0,2,21,1,1,0,0,31,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,
+95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
+10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,1,90,95,0,0,31,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,
+0,31,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,
+48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,
+18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,0,2,
+22,1,1,0,0,9,0,97,0,0,1,1,0,0,26,0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,
+18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,
+57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,
+57,48,20,0,0,1,90,95,0,0,26,0,2,22,1,1,0,0,26,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,
 105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
 109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,
-109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
-109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,30,2,22,1,1,0,0,9,97,0,0,1,1,0,0,30,110,
-0,0,0,1,3,2,90,95,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,
-97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,30,2,22,1,1,0,0,
-30,109,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,
-18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,0,1,90,
-95,0,0,29,2,22,1,1,0,0,29,109,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,1,0,9,1,105,110,118,0,2,17,49,0,
-48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,
-110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,
-110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,
-110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,105,
-110,118,0,48,20,0,0,1,90,95,0,0,29,2,22,1,1,0,0,9,97,0,0,1,1,0,0,29,110,0,0,0,1,3,2,90,95,1,0,9,1,
+109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,28,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,28,
+0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,
+116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,
+86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,28,0,2,22,
+1,1,0,0,28,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,18,
+98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,
+48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,
+48,20,0,0,1,90,95,0,0,27,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,27,0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,
 105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
 105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,
 105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
-105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,
-105,110,118,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,31,2,22,1,1,0,0,31,109,0,0,1,1,0,0,9,98,
-0,0,0,1,3,2,90,95,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,
-97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,31,2,22,1,1,0,0,
-9,97,0,0,1,1,0,0,31,110,0,0,0,1,3,2,90,95,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,
-18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,105,110,118,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,
-95,0,0,26,2,27,1,1,0,0,26,109,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,54,0,18,
-109,0,16,10,49,0,57,54,0,0,0,0,1,90,95,0,0,28,2,27,1,1,0,0,28,109,0,0,0,1,8,58,109,97,116,50,120,
-52,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,0,0,0,1,90,95,0,0,27,2,27,1,1,0,0,27,
-109,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18,
-109,0,16,10,50,0,57,54,0,0,0,0,1,90,95,0,0,30,2,27,1,1,0,0,30,109,0,0,0,1,8,58,109,97,116,51,120,
-52,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,0,0,0,1,90,
-95,0,0,29,2,27,1,1,0,0,29,109,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,54,0,18,
-109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,18,109,0,16,10,51,0,57,54,0,0,0,0,1,90,95,0,0,
-31,2,27,1,1,0,0,31,109,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,
-10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,18,109,0,16,10,51,0,57,54,0,0,0,0,1,90,95,0,0,0,2,25,1,
-0,2,0,26,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,0,1,90,95,0,0,0,2,
-25,1,0,2,0,28,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,0,1,90,95,0,0,
-0,2,25,1,0,2,0,27,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,
-0,16,10,50,0,57,52,0,0,1,90,95,0,0,0,2,25,1,0,2,0,30,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,
-109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,0,1,90,95,0,0,0,2,25,1,0,2,0,29,109,0,0,0,1,
-9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,9,18,109,
-0,16,10,51,0,57,52,0,0,1,90,95,0,0,0,2,25,1,0,2,0,31,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,
-109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,9,18,109,0,16,10,51,0,57,52,0,0,1,90,95,0,0,
-0,2,24,1,0,2,0,26,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,0,1,90,95,
-0,0,0,2,24,1,0,2,0,28,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,0,1,
-90,95,0,0,0,2,24,1,0,2,0,27,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,
-9,18,109,0,16,10,50,0,57,51,0,0,1,90,95,0,0,0,2,24,1,0,2,0,30,109,0,0,0,1,9,18,109,0,16,8,48,0,57,
-51,0,9,18,109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,0,1,90,95,0,0,0,2,24,1,0,2,0,29,
-109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,
-0,9,18,109,0,16,10,51,0,57,51,0,0,1,90,95,0,0,0,2,24,1,0,2,0,31,109,0,0,0,1,9,18,109,0,16,8,48,0,
+105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,27,0,2,22,1,1,0,0,27,0,109,0,0,1,1,0,0,
+9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,
+101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,30,0,2,
+22,1,1,0,0,9,0,97,0,0,1,1,0,0,30,0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,
+18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,
+57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,
+57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0,
+57,48,20,0,0,1,90,95,0,0,30,0,2,22,1,1,0,0,30,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,
+105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
+109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,
+109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
+109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,29,0,2,22,1,1,0,0,29,0,109,0,0,1,1,0,0,
+9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,
+101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,29,0,2,
+22,1,1,0,0,9,0,97,0,0,1,1,0,0,29,0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,
+18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,
+57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,
+57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0,
+57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,105,110,118,0,18,110,0,16,10,51,0,
+57,48,20,0,0,1,90,95,0,0,31,0,2,22,1,1,0,0,31,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,1,0,9,0,1,
+105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
+109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,
+109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
+109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,
+109,0,16,10,51,0,57,18,105,110,118,0,48,20,0,0,1,90,95,0,0,31,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,31,
+0,110,0,0,0,1,3,2,90,95,1,0,9,0,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,
+116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,
+86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,
+86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,
+86,97,108,0,16,10,51,0,57,18,105,110,118,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,26,0,2,27,
+1,1,0,0,26,0,109,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,
+49,0,57,54,0,0,0,0,1,90,95,0,0,28,0,2,27,1,1,0,0,28,0,109,0,0,0,1,8,58,109,97,116,50,120,52,0,0,18,
+109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,0,0,0,1,90,95,0,0,27,0,2,27,1,1,0,0,27,0,109,0,
+0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18,109,
+0,16,10,50,0,57,54,0,0,0,0,1,90,95,0,0,30,0,2,27,1,1,0,0,30,0,109,0,0,0,1,8,58,109,97,116,51,120,
+52,0,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,0,0,0,1,
+90,95,0,0,29,0,2,27,1,1,0,0,29,0,109,0,0,0,1,8,58,109,97,116,52,120,50,0,0,18,109,0,16,8,48,0,57,
+54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,18,109,0,16,10,51,0,57,54,0,0,0,0,1,
+90,95,0,0,31,0,2,27,1,1,0,0,31,0,109,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,
+54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,18,109,0,16,10,51,0,57,54,0,0,0,0,1,
+90,95,0,0,0,0,2,25,1,0,2,0,26,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,
+52,0,0,1,90,95,0,0,0,0,2,25,1,0,2,0,28,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,
+49,0,57,52,0,0,1,90,95,0,0,0,0,2,25,1,0,2,0,27,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,
+0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,0,1,90,95,0,0,0,0,2,25,1,0,2,0,30,0,109,0,0,0,1,
+9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,0,1,90,95,
+0,0,0,0,2,25,1,0,2,0,29,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,
+18,109,0,16,10,50,0,57,52,0,9,18,109,0,16,10,51,0,57,52,0,0,1,90,95,0,0,0,0,2,25,1,0,2,0,31,0,109,
+0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,9,
+18,109,0,16,10,51,0,57,52,0,0,1,90,95,0,0,0,0,2,24,1,0,2,0,26,0,109,0,0,0,1,9,18,109,0,16,8,48,0,
+57,51,0,9,18,109,0,16,10,49,0,57,51,0,0,1,90,95,0,0,0,0,2,24,1,0,2,0,28,0,109,0,0,0,1,9,18,109,0,
+16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,0,1,90,95,0,0,0,0,2,24,1,0,2,0,27,0,109,0,0,0,1,9,
+18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,0,1,90,95,0,
+0,0,0,2,24,1,0,2,0,30,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,9,
+18,109,0,16,10,50,0,57,51,0,0,1,90,95,0,0,0,0,2,24,1,0,2,0,29,0,109,0,0,0,1,9,18,109,0,16,8,48,0,
 57,51,0,9,18,109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,9,18,109,0,16,10,51,0,57,51,0,
-0,0
+0,1,90,95,0,0,0,0,2,24,1,0,2,0,31,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,
+57,51,0,9,18,109,0,16,10,50,0,57,51,0,9,18,109,0,16,10,51,0,57,51,0,0,0
diff --git a/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h b/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h
index 1a572311c3..c397b9f0fa 100644
--- a/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h
+++ b/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h
@@ -2,105 +2,107 @@
 /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */
 /* slang_builtin_120_common.gc */
 
-4,1,90,95,0,0,26,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,26,109,0,0,1,0,0,
-0,26,110,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,
-109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,90,95,0,0,28,0,109,97,116,114,105,120,67,
-111,109,112,77,117,108,116,0,1,0,0,0,28,109,0,0,1,0,0,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,
-18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,
-0,0,0,0,1,90,95,0,0,27,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,27,109,0,0,
-1,0,0,0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,
-0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,
-48,0,0,0,0,1,90,95,0,0,30,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,30,109,
-0,0,1,0,0,0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,
-48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,
-57,48,0,0,0,0,1,90,95,0,0,29,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,29,
-109,0,0,1,0,0,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,
-0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,
-50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,90,95,0,0,31,0,109,97,116,
-114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,31,109,0,0,1,0,0,0,31,110,0,0,0,1,8,58,109,97,
-116,52,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,
-16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,
-110,0,16,10,51,0,57,48,0,0,0,0,1,90,95,0,0,13,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,
-0,0,0,10,99,0,0,1,0,0,0,10,114,0,0,0,1,8,58,109,97,116,50,0,18,99,0,59,120,0,18,114,0,59,120,0,48,
-0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,
-18,114,0,59,121,0,48,0,0,0,0,1,90,95,0,0,14,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,
-0,0,11,99,0,0,1,0,0,0,11,114,0,0,0,1,8,58,109,97,116,51,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,
-18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,
-18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,
-48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,
-122,0,18,114,0,59,122,0,48,0,0,0,0,1,90,95,0,0,15,0,111,117,116,101,114,80,114,111,100,117,99,116,
-0,1,0,0,0,12,99,0,0,1,0,0,0,12,114,0,0,0,1,8,58,109,97,116,52,0,18,99,0,59,120,0,18,114,0,59,120,0,
-48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,
-119,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,
-121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,0,18,99,0,
-59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,
-59,122,0,48,0,18,99,0,59,119,0,18,114,0,59,122,0,48,0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,
-99,0,59,121,0,18,114,0,59,119,0,48,0,18,99,0,59,122,0,18,114,0,59,119,0,48,0,18,99,0,59,119,0,18,
-114,0,59,119,0,48,0,0,0,0,1,90,95,0,0,26,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,
-11,99,0,0,1,0,0,0,10,114,0,0,0,1,8,58,109,97,116,50,120,51,0,18,99,0,59,120,0,18,114,0,59,120,0,48,
-0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,
-18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,
-48,0,0,0,0,1,90,95,0,0,27,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,10,99,0,0,1,0,
-0,0,11,114,0,0,0,1,8,58,109,97,116,51,120,50,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,
-121,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,
-121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,0,0,0,1,
-90,95,0,0,28,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,12,99,0,0,1,0,0,0,10,114,0,
-0,0,1,8,58,109,97,116,50,120,52,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,
-0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,
-99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,
-114,0,59,121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,0,0,0,0,1,90,95,0,0,29,0,111,117,116,101,
-114,80,114,111,100,117,99,116,0,1,0,0,0,10,99,0,0,1,0,0,0,12,114,0,0,0,1,8,58,109,97,116,52,120,50,
+5,1,90,95,0,0,26,0,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,26,0,109,0,0,1,
+0,0,0,26,0,110,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,
+48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,90,95,0,0,28,0,0,109,97,116,114,
+105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,28,0,109,0,0,1,0,0,0,28,0,110,0,0,0,1,8,58,109,97,
+116,50,120,52,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,
+16,10,49,0,57,48,0,0,0,0,1,90,95,0,0,27,0,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,
+1,0,0,0,27,0,109,0,0,1,0,0,0,27,0,110,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,16,8,48,0,57,
+18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,
+57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,90,95,0,0,30,0,0,109,97,116,114,105,120,67,111,109,112,77,
+117,108,116,0,1,0,0,0,30,0,109,0,0,1,0,0,0,30,0,110,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,109,0,
+16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,
+0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,90,95,0,0,29,0,0,109,97,116,114,105,120,67,111,
+109,112,77,117,108,116,0,1,0,0,0,29,0,109,0,0,1,0,0,0,29,0,110,0,0,0,1,8,58,109,97,116,52,120,50,0,
+0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,
+48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,
+57,48,0,0,0,0,1,90,95,0,0,31,0,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,31,
+0,109,0,0,1,0,0,0,31,0,110,0,0,0,1,8,58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,18,110,0,16,
+8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,
+16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,90,95,0,0,13,0,0,111,
+117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,10,0,99,0,0,1,0,0,0,10,0,114,0,0,0,1,8,58,109,
+97,116,50,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,
+99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,0,0,0,1,90,95,0,0,14,
+0,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,11,0,99,0,0,1,0,0,0,11,0,114,0,0,0,1,8,
+58,109,97,116,51,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,
+0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,
+18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,
+48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,0,0,0,1,90,95,
+0,0,15,0,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,12,0,99,0,0,1,0,0,0,12,0,114,0,
+0,0,1,8,58,109,97,116,52,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,
+120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,
+59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,
+59,121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,
+99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,18,99,0,59,119,0,18,
+114,0,59,122,0,48,0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0,48,
+0,18,99,0,59,122,0,18,114,0,59,119,0,48,0,18,99,0,59,119,0,18,114,0,59,119,0,48,0,0,0,0,1,90,95,0,
+0,26,0,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,11,0,99,0,0,1,0,0,0,10,0,114,0,0,
+0,1,8,58,109,97,116,50,120,51,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,
+0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,
+99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,0,0,0,1,90,95,0,0,27,
+0,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,10,0,99,0,0,1,0,0,0,11,0,114,0,0,0,1,8,
+58,109,97,116,51,120,50,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,
+120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,
+59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,0,0,0,1,90,95,0,0,28,0,0,
+111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,12,0,99,0,0,1,0,0,0,10,0,114,0,0,0,1,8,58,
+109,97,116,50,120,52,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,
+0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59,
+120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,
+121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,0,0,0,0,1,90,95,0,0,29,0,0,111,117,116,101,114,80,
+114,111,100,117,99,116,0,1,0,0,0,10,0,99,0,0,1,0,0,0,12,0,114,0,0,0,1,8,58,109,97,116,52,120,50,0,
 0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,
 18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,
 48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,
-121,0,18,114,0,59,119,0,48,0,0,0,0,1,90,95,0,0,30,0,111,117,116,101,114,80,114,111,100,117,99,116,
-0,1,0,0,0,12,99,0,0,1,0,0,0,11,114,0,0,0,1,8,58,109,97,116,51,120,52,0,18,99,0,59,120,0,18,114,0,
-59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,
-99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,
-114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,
-0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,
-18,114,0,59,122,0,48,0,18,99,0,59,119,0,18,114,0,59,122,0,48,0,0,0,0,1,90,95,0,0,31,0,111,117,116,
-101,114,80,114,111,100,117,99,116,0,1,0,0,0,11,99,0,0,1,0,0,0,12,114,0,0,0,1,8,58,109,97,116,52,
-120,51,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,
-59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,
-59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,
-99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,18,99,0,59,120,0,18,
-114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0,48,0,18,99,0,59,122,0,18,114,0,59,119,0,48,
-0,0,0,0,1,90,95,0,0,13,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,13,109,0,0,0,1,8,58,109,97,
-116,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,8,48,0,57,
-59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,0,0,0,1,90,95,0,0,14,0,116,114,97,110,115,112,111,115,
-101,0,1,0,0,0,14,109,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,
-49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,
-10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,
-16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,0,0,0,1,90,95,0,0,15,0,116,114,97,110,
-115,112,111,115,101,0,1,0,0,0,15,109,0,0,0,1,8,58,109,97,116,52,0,18,109,0,16,8,48,0,57,59,120,0,0,
-18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,120,
-0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,
-121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,
-59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,10,51,0,57,59,122,0,0,18,109,0,16,8,48,0,
-57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,18,109,0,16,10,50,0,57,59,119,0,0,18,109,0,16,10,
-51,0,57,59,119,0,0,0,0,0,1,90,95,0,0,26,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,27,109,0,0,
-0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,
-18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,
-0,18,109,0,16,10,50,0,57,59,121,0,0,0,0,0,1,90,95,0,0,27,0,116,114,97,110,115,112,111,115,101,0,1,
-0,0,0,26,109,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,
-49,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,
-8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,0,0,0,1,90,95,0,0,28,0,116,114,97,110,115,
-112,111,115,101,0,1,0,0,0,29,109,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,
-0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,
-120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,
-59,121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,0,0,0,1,90,95,0,0,29,0,116,114,97,110,115,112,111,115,
-101,0,1,0,0,0,28,109,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,
-16,10,49,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,
-0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,
-109,0,16,10,49,0,57,59,119,0,0,0,0,0,1,90,95,0,0,30,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,
-31,109,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,
-59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,
-57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,
-51,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,
-10,50,0,57,59,122,0,0,18,109,0,16,10,51,0,57,59,122,0,0,0,0,0,1,90,95,0,0,31,0,116,114,97,110,115,
-112,111,115,101,0,1,0,0,0,30,109,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,59,120,
-0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,
-121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,8,48,0,57,
-59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,8,48,0,
-57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,18,109,0,16,10,50,0,57,59,119,0,0,0,0,0,0
+121,0,18,114,0,59,119,0,48,0,0,0,0,1,90,95,0,0,30,0,0,111,117,116,101,114,80,114,111,100,117,99,
+116,0,1,0,0,0,12,0,99,0,0,1,0,0,0,11,0,114,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,99,0,59,120,0,
+18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,
+48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,
+121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,119,0,18,114,0,59,
+121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,
+59,122,0,18,114,0,59,122,0,48,0,18,99,0,59,119,0,18,114,0,59,122,0,48,0,0,0,0,1,90,95,0,0,31,0,0,
+111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,0,11,0,99,0,0,1,0,0,0,12,0,114,0,0,0,1,8,58,
+109,97,116,52,120,51,0,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,
+0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,
+121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,
+122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,18,99,0,
+59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0,48,0,18,99,0,59,122,0,18,114,0,
+59,119,0,48,0,0,0,0,1,90,95,0,0,13,0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,13,0,109,0,0,0,
+1,8,58,109,97,116,50,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,
+0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,0,0,0,1,90,95,0,0,14,0,0,116,114,97,
+110,115,112,111,115,101,0,1,0,0,0,14,0,109,0,0,0,1,8,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,59,
+120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,
+59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,8,48,0,
+57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,0,0,0,1,90,95,0,
+0,15,0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,15,0,109,0,0,0,1,8,58,109,97,116,52,0,0,18,
+109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,
+18,109,0,16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,
+0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,
+0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,10,51,0,57,59,
+122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,18,109,0,16,10,50,0,57,
+59,119,0,0,18,109,0,16,10,51,0,57,59,119,0,0,0,0,0,1,90,95,0,0,26,0,0,116,114,97,110,115,112,111,
+115,101,0,1,0,0,0,27,0,109,0,0,0,1,8,58,109,97,116,50,120,51,0,0,18,109,0,16,8,48,0,57,59,120,0,0,
+18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,
+0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,0,0,0,1,90,95,0,0,27,0,0,116,
+114,97,110,115,112,111,115,101,0,1,0,0,0,26,0,109,0,0,0,1,8,58,109,97,116,51,120,50,0,0,18,109,0,
+16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,
+0,16,10,49,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,0,0,
+0,1,90,95,0,0,28,0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,29,0,109,0,0,0,1,8,58,109,97,116,
+50,120,52,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,
+0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,
+49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,0,0,0,1,90,
+95,0,0,29,0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,28,0,109,0,0,0,1,8,58,109,97,116,52,120,
+50,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,
+121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,
+59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,0,0,0,1,90,95,0,0,30,
+0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,31,0,109,0,0,0,1,8,58,109,97,116,51,120,52,0,0,18,
+109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,
+18,109,0,16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,
+0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,
+0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,10,51,0,57,59,
+122,0,0,0,0,0,1,90,95,0,0,31,0,0,116,114,97,110,115,112,111,115,101,0,1,0,0,0,30,0,109,0,0,0,1,8,
+58,109,97,116,52,120,51,0,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,
+109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,
+18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,
+0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,
+0,0,18,109,0,16,10,50,0,57,59,119,0,0,0,0,0,0
diff --git a/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h b/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h
index 3382fce6ad..add3b5aeaa 100644
--- a/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h
+++ b/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h
@@ -2,4 +2,4 @@
 /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */
 /* slang_builtin_120_fragment.gc */
 
-4,2,2,90,95,3,0,10,1,103,108,95,80,111,105,110,116,67,111,111,114,100,0,0,0,0
+5,2,2,90,95,3,0,10,0,1,103,108,95,80,111,105,110,116,67,111,111,114,100,0,0,0,0
diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc
index a051c53eea..230c57cea8 100644
--- a/src/mesa/shader/slang/library/slang_common_builtin.gc
+++ b/src/mesa/shader/slang/library/slang_common_builtin.gc
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  7.3
  *
  * Copyright (C) 2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2008  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -26,8 +27,9 @@
 // From Shader Spec, ver. 1.10, rev. 59
 //
 
-//bp: XXX these will probably go away since the value needs to be
-//determined at runtime and may vary from one GLcontext to another...
+// Note: the values assigned to these constants here aren't actually used.
+// They're set by the compiler according to the GL context limits.
+// See slang_simplify.c
 const int gl_MaxLights = 8;
 const int gl_MaxClipPlanes = 6;
 const int gl_MaxTextureUnits = 8;
@@ -98,6 +100,9 @@ struct gl_MaterialParameters {
 uniform gl_MaterialParameters gl_FrontMaterial;
 uniform gl_MaterialParameters gl_BackMaterial;
 
+/* NOTE: the order of these fields is significant!
+ * See the definition of the lighting state vars such as STATE_SPOT_DIRECTION.
+ */
 struct gl_LightSourceParameters {
     vec4 ambient;
     vec4 diffuse;
@@ -105,12 +110,14 @@ struct gl_LightSourceParameters {
     vec4 position;
     vec4 halfVector;
     vec3 spotDirection;
-    float spotExponent;
-    float spotCutoff;
     float spotCosCutoff;
+
     float constantAttenuation;
     float linearAttenuation;
     float quadraticAttenuation;
+    float spotExponent;
+
+    float spotCutoff;
 };
 
 uniform gl_LightSourceParameters gl_LightSource[gl_MaxLights];
diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h
index dfe406c233..759bf247d8 100644
--- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h
+++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h
@@ -2,847 +2,867 @@
 /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */
 /* slang_common_builtin.gc */
 
-4,2,2,90,95,1,0,5,1,103,108,95,77,97,120,76,105,103,104,116,115,0,2,16,10,56,0,0,0,2,2,90,95,1,0,5,
-1,103,108,95,77,97,120,67,108,105,112,80,108,97,110,101,115,0,2,16,10,54,0,0,0,2,2,90,95,1,0,5,1,
-103,108,95,77,97,120,84,101,120,116,117,114,101,85,110,105,116,115,0,2,16,10,56,0,0,0,2,2,90,95,1,
-0,5,1,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,2,16,10,56,0,0,0,2,
-2,90,95,1,0,5,1,103,108,95,77,97,120,86,101,114,116,101,120,65,116,116,114,105,98,115,0,2,16,10,49,
-54,0,0,0,2,2,90,95,1,0,5,1,103,108,95,77,97,120,86,101,114,116,101,120,85,110,105,102,111,114,109,
-67,111,109,112,111,110,101,110,116,115,0,2,16,10,53,49,50,0,0,0,2,2,90,95,1,0,5,1,103,108,95,77,97,
-120,86,97,114,121,105,110,103,70,108,111,97,116,115,0,2,16,10,51,50,0,0,0,2,2,90,95,1,0,5,1,103,
-108,95,77,97,120,86,101,114,116,101,120,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105,
-116,115,0,2,16,8,48,0,0,0,2,2,90,95,1,0,5,1,103,108,95,77,97,120,67,111,109,98,105,110,101,100,84,
-101,120,116,117,114,101,73,109,97,103,101,85,110,105,116,115,0,2,16,10,50,0,0,0,2,2,90,95,1,0,5,1,
-103,108,95,77,97,120,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105,116,115,0,2,16,10,50,
-0,0,0,2,2,90,95,1,0,5,1,103,108,95,77,97,120,70,114,97,103,109,101,110,116,85,110,105,102,111,114,
-109,67,111,109,112,111,110,101,110,116,115,0,2,16,10,54,52,0,0,0,2,2,90,95,1,0,5,1,103,108,95,77,
-97,120,68,114,97,119,66,117,102,102,101,114,115,0,2,16,10,49,0,0,0,2,2,90,95,4,0,15,1,103,108,95,
-77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,0,0,0,2,2,90,95,4,0,15,1,103,108,95,80,114,
-111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,0,0,2,2,90,95,4,0,15,1,103,108,95,77,111,
-100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,0,0,2,2,
-90,95,4,0,15,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,120,0,3,18,103,108,95,77,97,
-120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,14,1,103,108,95,78,111,
-114,109,97,108,77,97,116,114,105,120,0,0,0,2,2,90,95,4,0,15,1,103,108,95,77,111,100,101,108,86,105,
-101,119,77,97,116,114,105,120,73,110,118,101,114,115,101,0,0,0,2,2,90,95,4,0,15,1,103,108,95,80,
-114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115,101,0,0,0,2,2,90,
-95,4,0,15,1,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,
-97,116,114,105,120,73,110,118,101,114,115,101,0,0,0,2,2,90,95,4,0,15,1,103,108,95,84,101,120,116,
-117,114,101,77,97,116,114,105,120,73,110,118,101,114,115,101,0,3,18,103,108,95,77,97,120,84,101,
-120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,15,1,103,108,95,77,111,100,101,108,
-86,105,101,119,77,97,116,114,105,120,84,114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,1,
-103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,84,114,97,110,115,112,111,
-115,101,0,0,0,2,2,90,95,4,0,15,1,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,
-99,116,105,111,110,77,97,116,114,105,120,84,114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,
-1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,120,84,114,97,110,115,112,111,115,101,0,
+5,2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,76,105,103,104,116,115,0,2,16,10,56,0,0,0,2,2,90,95,1,0,
+5,0,1,103,108,95,77,97,120,67,108,105,112,80,108,97,110,101,115,0,2,16,10,54,0,0,0,2,2,90,95,1,0,5,
+0,1,103,108,95,77,97,120,84,101,120,116,117,114,101,85,110,105,116,115,0,2,16,10,56,0,0,0,2,2,90,
+95,1,0,5,0,1,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,2,16,10,56,0,
+0,0,2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,86,101,114,116,101,120,65,116,116,114,105,98,115,0,2,
+16,10,49,54,0,0,0,2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,86,101,114,116,101,120,85,110,105,102,
+111,114,109,67,111,109,112,111,110,101,110,116,115,0,2,16,10,53,49,50,0,0,0,2,2,90,95,1,0,5,0,1,
+103,108,95,77,97,120,86,97,114,121,105,110,103,70,108,111,97,116,115,0,2,16,10,51,50,0,0,0,2,2,90,
+95,1,0,5,0,1,103,108,95,77,97,120,86,101,114,116,101,120,84,101,120,116,117,114,101,73,109,97,103,
+101,85,110,105,116,115,0,2,16,8,48,0,0,0,2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,67,111,109,98,
+105,110,101,100,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105,116,115,0,2,16,10,50,0,0,0,
+2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105,
+116,115,0,2,16,10,50,0,0,0,2,2,90,95,1,0,5,0,1,103,108,95,77,97,120,70,114,97,103,109,101,110,116,
+85,110,105,102,111,114,109,67,111,109,112,111,110,101,110,116,115,0,2,16,10,54,52,0,0,0,2,2,90,95,
+1,0,5,0,1,103,108,95,77,97,120,68,114,97,119,66,117,102,102,101,114,115,0,2,16,10,49,0,0,0,2,2,90,
+95,4,0,15,0,1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,0,0,0,2,2,90,95,4,
+0,15,0,1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,0,0,2,2,90,95,4,
+0,15,0,1,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,
+116,114,105,120,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,
+120,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,
+4,0,14,0,1,103,108,95,78,111,114,109,97,108,77,97,116,114,105,120,0,0,0,2,2,90,95,4,0,15,0,1,103,
+108,95,77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,73,110,118,101,114,115,101,0,0,0,2,
+2,90,95,4,0,15,0,1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,
+118,101,114,115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,77,111,100,101,108,86,105,101,119,80,114,
+111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115,101,0,0,0,2,2,90,95,4,
+0,15,0,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,120,73,110,118,101,114,115,101,0,
 3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,15,
-1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,73,110,118,101,114,115,101,84,
-114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,1,103,108,95,80,114,111,106,101,99,116,105,
-111,110,77,97,116,114,105,120,73,110,118,101,114,115,101,84,114,97,110,115,112,111,115,101,0,0,0,2,
-2,90,95,4,0,15,1,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,
-110,77,97,116,114,105,120,73,110,118,101,114,115,101,84,114,97,110,115,112,111,115,101,0,0,0,2,2,
-90,95,4,0,15,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,120,73,110,118,101,114,115,
-101,84,114,97,110,115,112,111,115,101,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,
-111,111,114,100,115,0,0,0,2,2,90,95,4,0,9,1,103,108,95,78,111,114,109,97,108,83,99,97,108,101,0,0,
-0,2,2,90,95,0,0,24,103,108,95,68,101,112,116,104,82,97,110,103,101,80,97,114,97,109,101,116,101,
-114,115,0,9,110,101,97,114,0,0,0,1,9,102,97,114,0,0,0,1,9,100,105,102,102,0,0,0,0,0,0,2,2,90,95,4,
-0,25,103,108,95,68,101,112,116,104,82,97,110,103,101,80,97,114,97,109,101,116,101,114,115,0,1,103,
-108,95,68,101,112,116,104,82,97,110,103,101,0,0,0,2,2,90,95,4,0,12,1,103,108,95,67,108,105,112,80,
-108,97,110,101,0,3,18,103,108,95,77,97,120,67,108,105,112,80,108,97,110,101,115,0,0,0,2,2,90,95,0,
-0,24,103,108,95,80,111,105,110,116,80,97,114,97,109,101,116,101,114,115,0,9,115,105,122,101,0,0,0,
-1,9,115,105,122,101,77,105,110,0,0,0,1,9,115,105,122,101,77,97,120,0,0,0,1,9,102,97,100,101,84,104,
-114,101,115,104,111,108,100,83,105,122,101,0,0,0,1,9,100,105,115,116,97,110,99,101,67,111,110,115,
-116,97,110,116,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,100,105,115,116,97,110,99,101,
-76,105,110,101,97,114,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,100,105,115,116,97,110,
-99,101,81,117,97,100,114,97,116,105,99,65,116,116,101,110,117,97,116,105,111,110,0,0,0,0,0,0,2,2,
-90,95,4,0,25,103,108,95,80,111,105,110,116,80,97,114,97,109,101,116,101,114,115,0,1,103,108,95,80,
-111,105,110,116,0,0,0,2,2,90,95,0,0,24,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109,
-101,116,101,114,115,0,12,101,109,105,115,115,105,111,110,0,0,0,1,12,97,109,98,105,101,110,116,0,0,
-0,1,12,100,105,102,102,117,115,101,0,0,0,1,12,115,112,101,99,117,108,97,114,0,0,0,1,9,115,104,105,
-110,105,110,101,115,115,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,77,97,116,101,114,105,97,108,80,97,
-114,97,109,101,116,101,114,115,0,1,103,108,95,70,114,111,110,116,77,97,116,101,114,105,97,108,0,0,
-0,2,2,90,95,4,0,25,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109,101,116,101,114,115,0,
-1,103,108,95,66,97,99,107,77,97,116,101,114,105,97,108,0,0,0,2,2,90,95,0,0,24,103,108,95,76,105,
-103,104,116,83,111,117,114,99,101,80,97,114,97,109,101,116,101,114,115,0,12,97,109,98,105,101,110,
-116,0,0,0,1,12,100,105,102,102,117,115,101,0,0,0,1,12,115,112,101,99,117,108,97,114,0,0,0,1,12,112,
-111,115,105,116,105,111,110,0,0,0,1,12,104,97,108,102,86,101,99,116,111,114,0,0,0,1,11,115,112,111,
-116,68,105,114,101,99,116,105,111,110,0,0,0,1,9,115,112,111,116,69,120,112,111,110,101,110,116,0,0,
-0,1,9,115,112,111,116,67,117,116,111,102,102,0,0,0,1,9,115,112,111,116,67,111,115,67,117,116,111,
-102,102,0,0,0,1,9,99,111,110,115,116,97,110,116,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,
-9,108,105,110,101,97,114,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,113,117,97,100,114,97,
-116,105,99,65,116,116,101,110,117,97,116,105,111,110,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,76,
-105,103,104,116,83,111,117,114,99,101,80,97,114,97,109,101,116,101,114,115,0,1,103,108,95,76,105,
-103,104,116,83,111,117,114,99,101,0,3,18,103,108,95,77,97,120,76,105,103,104,116,115,0,0,0,2,2,90,
-95,0,0,24,103,108,95,76,105,103,104,116,77,111,100,101,108,80,97,114,97,109,101,116,101,114,115,0,
-12,97,109,98,105,101,110,116,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,76,105,103,104,116,77,111,100,
-101,108,80,97,114,97,109,101,116,101,114,115,0,1,103,108,95,76,105,103,104,116,77,111,100,101,108,
-0,0,0,2,2,90,95,0,0,24,103,108,95,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,
-115,0,12,115,99,101,110,101,67,111,108,111,114,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,76,105,103,
-104,116,77,111,100,101,108,80,114,111,100,117,99,116,115,0,1,103,108,95,70,114,111,110,116,76,105,
-103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,0,0,0,2,2,90,95,4,0,25,103,108,95,76,105,
-103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,115,0,1,103,108,95,66,97,99,107,76,105,
-103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,0,0,0,2,2,90,95,0,0,24,103,108,95,76,105,
-103,104,116,80,114,111,100,117,99,116,115,0,12,97,109,98,105,101,110,116,0,0,0,1,12,100,105,102,
-102,117,115,101,0,0,0,1,12,115,112,101,99,117,108,97,114,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,
-76,105,103,104,116,80,114,111,100,117,99,116,115,0,1,103,108,95,70,114,111,110,116,76,105,103,104,
-116,80,114,111,100,117,99,116,0,3,18,103,108,95,77,97,120,76,105,103,104,116,115,0,0,0,2,2,90,95,4,
-0,25,103,108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,1,103,108,95,66,97,99,107,76,
-105,103,104,116,80,114,111,100,117,99,116,0,3,18,103,108,95,77,97,120,76,105,103,104,116,115,0,0,0,
-2,2,90,95,4,0,12,1,103,108,95,84,101,120,116,117,114,101,69,110,118,67,111,108,111,114,0,3,18,103,
-108,95,77,97,120,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105,116,115,0,0,0,2,2,90,95,4,
-0,12,1,103,108,95,69,121,101,80,108,97,110,101,83,0,3,18,103,108,95,77,97,120,84,101,120,116,117,
-114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,1,103,108,95,69,121,101,80,108,97,110,101,84,
-0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,
-12,1,103,108,95,69,121,101,80,108,97,110,101,82,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,
-101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,1,103,108,95,69,121,101,80,108,97,110,101,81,0,3,
-18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,1,
-103,108,95,79,98,106,101,99,116,80,108,97,110,101,83,0,3,18,103,108,95,77,97,120,84,101,120,116,
-117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,1,103,108,95,79,98,106,101,99,116,80,108,
-97,110,101,84,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,
-2,2,90,95,4,0,12,1,103,108,95,79,98,106,101,99,116,80,108,97,110,101,82,0,3,18,103,108,95,77,97,
-120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,1,103,108,95,79,98,
-106,101,99,116,80,108,97,110,101,81,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,
-111,114,100,115,0,0,0,2,2,90,95,0,0,24,103,108,95,70,111,103,80,97,114,97,109,101,116,101,114,115,
-0,12,99,111,108,111,114,0,0,0,1,9,100,101,110,115,105,116,121,0,0,0,1,9,115,116,97,114,116,0,0,0,1,
-9,101,110,100,0,0,0,1,9,115,99,97,108,101,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,70,111,103,80,97,
-114,97,109,101,116,101,114,115,0,1,103,108,95,70,111,103,0,0,0,1,90,95,0,0,9,0,114,97,100,105,97,
-110,115,0,1,1,0,0,9,100,101,103,0,0,0,1,3,2,90,95,1,0,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,
-17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,
-116,86,97,108,0,0,18,100,101,103,0,0,18,99,0,0,0,0,1,90,95,0,0,10,0,114,97,100,105,97,110,115,0,1,
-1,0,0,10,100,101,103,0,0,0,1,3,2,90,95,1,0,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,
-0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,
-108,0,59,120,121,0,0,18,100,101,103,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,90,95,0,0,11,0,
-114,97,100,105,97,110,115,0,1,1,0,0,11,100,101,103,0,0,0,1,3,2,90,95,1,0,9,1,99,0,2,17,51,0,49,52,
-49,53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,
-0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,100,101,103,0,59,120,121,122,0,0,18,99,0,
-59,120,120,120,0,0,0,0,1,90,95,0,0,12,0,114,97,100,105,97,110,115,0,1,1,0,0,12,100,101,103,0,0,0,1,
-3,2,90,95,1,0,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,
-52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,100,101,103,0,0,18,
-99,0,59,120,120,120,120,0,0,0,0,1,90,95,0,0,9,0,100,101,103,114,101,101,115,0,1,1,0,0,9,114,97,100,
-0,0,0,1,3,2,90,95,1,0,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,
-118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,
-100,0,0,18,99,0,0,0,0,1,90,95,0,0,10,0,100,101,103,114,101,101,115,0,1,1,0,0,10,114,97,100,0,0,0,1,
-3,2,90,95,1,0,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,
-52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,114,97,
-100,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,90,95,0,0,11,0,100,101,103,114,101,101,115,0,1,1,
-0,0,11,114,97,100,0,0,0,1,3,2,90,95,1,0,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,
-54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,
-0,59,120,121,122,0,0,18,114,97,100,0,59,120,121,122,0,0,18,99,0,59,120,120,120,0,0,0,0,1,90,95,0,0,
-12,0,100,101,103,114,101,101,115,0,1,1,0,0,12,114,97,100,0,0,0,1,3,2,90,95,1,0,9,1,99,0,2,17,49,56,
-48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,
-121,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,0,0,18,99,0,59,120,120,120,120,0,0,0,0,1,90,
-95,0,0,9,0,115,105,110,0,1,1,0,0,9,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,
-105,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,90,95,0,0,
-10,0,115,105,110,0,1,1,0,0,10,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,
-110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,
-4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,
-100,105,97,110,115,0,59,121,0,0,0,0,1,90,95,0,0,11,0,115,105,110,0,1,1,0,0,11,114,97,100,105,97,
-110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,
-0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,
-114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,
-116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,
-0,59,122,0,0,0,0,1,90,95,0,0,12,0,115,105,110,0,1,1,0,0,12,114,97,100,105,97,110,115,0,0,0,1,4,102,
-108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,
-97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,
-0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,
-18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,
-111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,
-110,115,0,59,119,0,0,0,0,1,90,95,0,0,9,0,99,111,115,0,1,1,0,0,9,114,97,100,105,97,110,115,0,0,0,1,
-4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,
-105,97,110,115,0,0,0,0,1,90,95,0,0,10,0,99,111,115,0,1,1,0,0,10,114,97,100,105,97,110,115,0,0,0,1,
-4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,
-114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,
-114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,0,1,90,95,0,0,11,0,
-99,111,115,0,1,1,0,0,11,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,
-110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,
-4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,
-114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,
-114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,90,95,0,0,12,0,
-99,111,115,0,1,1,0,0,12,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,
+0,1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,84,114,97,110,115,112,111,
+115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,
+105,120,84,114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,77,111,100,101,108,
+86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,84,114,97,110,115,112,
+111,115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,120,
+84,114,97,110,115,112,111,115,101,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,
+111,114,100,115,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116,
+114,105,120,73,110,118,101,114,115,101,84,114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,0,
+1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115,
+101,84,114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,77,111,100,101,108,86,
+105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115,101,
+84,114,97,110,115,112,111,115,101,0,0,0,2,2,90,95,4,0,15,0,1,103,108,95,84,101,120,116,117,114,101,
+77,97,116,114,105,120,73,110,118,101,114,115,101,84,114,97,110,115,112,111,115,101,0,3,18,103,108,
+95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,9,0,1,103,108,
+95,78,111,114,109,97,108,83,99,97,108,101,0,0,0,2,2,90,95,0,0,24,103,108,95,68,101,112,116,104,82,
+97,110,103,101,80,97,114,97,109,101,116,101,114,115,0,9,0,110,101,97,114,0,0,0,1,9,0,102,97,114,0,
+0,0,1,9,0,100,105,102,102,0,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,68,101,112,116,104,82,97,110,
+103,101,80,97,114,97,109,101,116,101,114,115,0,0,1,103,108,95,68,101,112,116,104,82,97,110,103,101,
+0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,67,108,105,112,80,108,97,110,101,0,3,18,103,108,95,77,97,120,
+67,108,105,112,80,108,97,110,101,115,0,0,0,2,2,90,95,0,0,24,103,108,95,80,111,105,110,116,80,97,
+114,97,109,101,116,101,114,115,0,9,0,115,105,122,101,0,0,0,1,9,0,115,105,122,101,77,105,110,0,0,0,
+1,9,0,115,105,122,101,77,97,120,0,0,0,1,9,0,102,97,100,101,84,104,114,101,115,104,111,108,100,83,
+105,122,101,0,0,0,1,9,0,100,105,115,116,97,110,99,101,67,111,110,115,116,97,110,116,65,116,116,101,
+110,117,97,116,105,111,110,0,0,0,1,9,0,100,105,115,116,97,110,99,101,76,105,110,101,97,114,65,116,
+116,101,110,117,97,116,105,111,110,0,0,0,1,9,0,100,105,115,116,97,110,99,101,81,117,97,100,114,97,
+116,105,99,65,116,116,101,110,117,97,116,105,111,110,0,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,80,
+111,105,110,116,80,97,114,97,109,101,116,101,114,115,0,0,1,103,108,95,80,111,105,110,116,0,0,0,2,2,
+90,95,0,0,24,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109,101,116,101,114,115,0,12,0,
+101,109,105,115,115,105,111,110,0,0,0,1,12,0,97,109,98,105,101,110,116,0,0,0,1,12,0,100,105,102,
+102,117,115,101,0,0,0,1,12,0,115,112,101,99,117,108,97,114,0,0,0,1,9,0,115,104,105,110,105,110,101,
+115,115,0,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109,
+101,116,101,114,115,0,0,1,103,108,95,70,114,111,110,116,77,97,116,101,114,105,97,108,0,0,0,2,2,90,
+95,4,0,25,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109,101,116,101,114,115,0,0,1,103,
+108,95,66,97,99,107,77,97,116,101,114,105,97,108,0,0,0,2,2,90,95,0,0,24,103,108,95,76,105,103,104,
+116,83,111,117,114,99,101,80,97,114,97,109,101,116,101,114,115,0,12,0,97,109,98,105,101,110,116,0,
+0,0,1,12,0,100,105,102,102,117,115,101,0,0,0,1,12,0,115,112,101,99,117,108,97,114,0,0,0,1,12,0,112,
+111,115,105,116,105,111,110,0,0,0,1,12,0,104,97,108,102,86,101,99,116,111,114,0,0,0,1,11,0,115,112,
+111,116,68,105,114,101,99,116,105,111,110,0,0,0,1,9,0,115,112,111,116,67,111,115,67,117,116,111,
+102,102,0,0,0,1,9,0,99,111,110,115,116,97,110,116,65,116,116,101,110,117,97,116,105,111,110,0,0,0,
+1,9,0,108,105,110,101,97,114,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,0,113,117,97,100,
+114,97,116,105,99,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,0,115,112,111,116,69,120,112,
+111,110,101,110,116,0,0,0,1,9,0,115,112,111,116,67,117,116,111,102,102,0,0,0,0,0,0,0,2,2,90,95,4,0,
+25,103,108,95,76,105,103,104,116,83,111,117,114,99,101,80,97,114,97,109,101,116,101,114,115,0,0,1,
+103,108,95,76,105,103,104,116,83,111,117,114,99,101,0,3,18,103,108,95,77,97,120,76,105,103,104,116,
+115,0,0,0,2,2,90,95,0,0,24,103,108,95,76,105,103,104,116,77,111,100,101,108,80,97,114,97,109,101,
+116,101,114,115,0,12,0,97,109,98,105,101,110,116,0,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,76,105,
+103,104,116,77,111,100,101,108,80,97,114,97,109,101,116,101,114,115,0,0,1,103,108,95,76,105,103,
+104,116,77,111,100,101,108,0,0,0,2,2,90,95,0,0,24,103,108,95,76,105,103,104,116,77,111,100,101,108,
+80,114,111,100,117,99,116,115,0,12,0,115,99,101,110,101,67,111,108,111,114,0,0,0,0,0,0,0,2,2,90,95,
+4,0,25,103,108,95,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,115,0,0,1,103,
+108,95,70,114,111,110,116,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,0,0,0,2,
+2,90,95,4,0,25,103,108,95,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,115,0,0,
+1,103,108,95,66,97,99,107,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,0,0,0,2,
+2,90,95,0,0,24,103,108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,12,0,97,109,98,105,
+101,110,116,0,0,0,1,12,0,100,105,102,102,117,115,101,0,0,0,1,12,0,115,112,101,99,117,108,97,114,0,
+0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,0,1,103,
+108,95,70,114,111,110,116,76,105,103,104,116,80,114,111,100,117,99,116,0,3,18,103,108,95,77,97,120,
+76,105,103,104,116,115,0,0,0,2,2,90,95,4,0,25,103,108,95,76,105,103,104,116,80,114,111,100,117,99,
+116,115,0,0,1,103,108,95,66,97,99,107,76,105,103,104,116,80,114,111,100,117,99,116,0,3,18,103,108,
+95,77,97,120,76,105,103,104,116,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,84,101,120,116,117,114,
+101,69,110,118,67,111,108,111,114,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,73,109,97,
+103,101,85,110,105,116,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,69,121,101,80,108,97,110,101,83,0,
+3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,
+0,1,103,108,95,69,121,101,80,108,97,110,101,84,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,
+101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,69,121,101,80,108,97,110,101,82,0,
+3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,
+0,1,103,108,95,69,121,101,80,108,97,110,101,81,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,
+101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,79,98,106,101,99,116,80,108,97,
+110,101,83,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,
+90,95,4,0,12,0,1,103,108,95,79,98,106,101,99,116,80,108,97,110,101,84,0,3,18,103,108,95,77,97,120,
+84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,79,98,106,
+101,99,116,80,108,97,110,101,82,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,
+114,100,115,0,0,0,2,2,90,95,4,0,12,0,1,103,108,95,79,98,106,101,99,116,80,108,97,110,101,81,0,3,18,
+103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,0,0,24,103,
+108,95,70,111,103,80,97,114,97,109,101,116,101,114,115,0,12,0,99,111,108,111,114,0,0,0,1,9,0,100,
+101,110,115,105,116,121,0,0,0,1,9,0,115,116,97,114,116,0,0,0,1,9,0,101,110,100,0,0,0,1,9,0,115,99,
+97,108,101,0,0,0,0,0,0,0,2,2,90,95,4,0,25,103,108,95,70,111,103,80,97,114,97,109,101,116,101,114,
+115,0,0,1,103,108,95,70,111,103,0,0,0,1,90,95,0,0,9,0,0,114,97,100,105,97,110,115,0,1,1,0,0,9,0,
+100,101,103,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48,0,
+0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,
+18,100,101,103,0,0,18,99,0,0,0,0,1,90,95,0,0,10,0,0,114,97,100,105,97,110,115,0,1,1,0,0,10,0,100,
+101,103,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48,0,0,
+49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,
+120,121,0,0,18,100,101,103,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,90,95,0,0,11,0,0,114,97,
+100,105,97,110,115,0,1,1,0,0,11,0,100,101,103,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,51,0,49,52,49,
+53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,
+18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,100,101,103,0,59,120,121,122,0,0,18,99,0,59,
+120,120,120,0,0,0,0,1,90,95,0,0,12,0,0,114,97,100,105,97,110,115,0,1,1,0,0,12,0,100,101,103,0,0,0,
+1,3,2,90,95,1,0,9,0,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,
+101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,100,101,103,0,
+0,18,99,0,59,120,120,120,120,0,0,0,0,1,90,95,0,0,9,0,0,100,101,103,114,101,101,115,0,1,1,0,0,9,0,
+114,97,100,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,
+0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,
+18,114,97,100,0,0,18,99,0,0,0,0,1,90,95,0,0,10,0,0,100,101,103,114,101,101,115,0,1,1,0,0,10,0,114,
+97,100,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,
+0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,
+121,0,0,18,114,97,100,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,90,95,0,0,11,0,0,100,101,103,
+114,101,101,115,0,1,1,0,0,11,0,114,97,100,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,49,56,48,0,48,0,0,
+17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,
+95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,114,97,100,0,59,120,121,122,0,0,18,99,0,59,120,
+120,120,0,0,0,0,1,90,95,0,0,12,0,0,100,101,103,114,101,101,115,0,1,1,0,0,12,0,114,97,100,0,0,0,1,3,
+2,90,95,1,0,9,0,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,
+52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,0,0,18,99,
+0,59,120,120,120,120,0,0,0,0,1,90,95,0,0,9,0,0,115,105,110,0,1,1,0,0,9,0,114,97,100,105,97,110,115,
+0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,
+105,97,110,115,0,0,0,0,1,90,95,0,0,10,0,0,115,105,110,0,1,1,0,0,10,0,114,97,100,105,97,110,115,0,0,
+0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,
+97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,
+116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,0,1,90,95,0,0,11,0,0,115,
+105,110,0,1,1,0,0,11,0,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,
+18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,
+111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,
+110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,
+59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,90,95,0,0,12,0,0,115,105,110,0,1,1,0,0,
+12,0,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,
+116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,
+105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,
+0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,
+97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,
+116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,110,115,0,59,119,0,0,0,0,1,90,95,0,0,9,0,0,99,111,
+115,0,1,1,0,0,9,0,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101,
+0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,90,95,0,0,10,0,0,99,
+111,115,0,1,1,0,0,10,0,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,
 110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,
 4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,
-114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,
-114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,
-116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,
-110,115,0,59,119,0,0,0,0,1,90,95,0,0,9,0,116,97,110,0,1,1,0,0,9,97,110,103,108,101,0,0,0,1,3,2,90,
-95,1,0,9,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,90,95,1,0,9,1,99,0,2,58,99,
-111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,90,95,0,0,10,0,116,97,110,0,
-1,1,0,0,10,97,110,103,108,101,0,0,0,1,3,2,90,95,1,0,10,1,115,0,2,58,115,105,110,0,18,97,110,103,
-108,101,0,0,0,0,0,3,2,90,95,1,0,10,1,99,0,2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,
-115,0,18,99,0,49,0,0,1,90,95,0,0,11,0,116,97,110,0,1,1,0,0,11,97,110,103,108,101,0,0,0,1,3,2,90,95,
-1,0,11,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,90,95,1,0,11,1,99,0,2,58,99,
-111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,90,95,0,0,12,0,116,97,110,0,
-1,1,0,0,12,97,110,103,108,101,0,0,0,1,3,2,90,95,1,0,12,1,115,0,2,58,115,105,110,0,18,97,110,103,
-108,101,0,0,0,0,0,3,2,90,95,1,0,12,1,99,0,2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,
-115,0,18,99,0,49,0,0,1,90,95,0,0,9,0,97,115,105,110,0,1,1,0,0,9,120,0,0,0,1,3,2,90,95,1,0,9,1,97,
-48,0,2,17,49,0,53,55,48,55,50,56,56,0,0,0,0,3,2,90,95,1,0,9,1,97,49,0,2,17,48,0,50,49,50,49,49,52,
-52,0,0,54,0,0,3,2,90,95,1,0,9,1,97,50,0,2,17,48,0,48,55,52,50,54,49,48,0,0,0,0,3,2,90,95,1,0,9,1,
-104,97,108,102,80,105,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,48,0,53,0,0,48,0,0,3,2,90,95,1,0,9,1,
-121,0,2,58,97,98,115,0,18,120,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,104,97,108,102,80,
-105,0,58,115,113,114,116,0,17,49,0,48,0,0,18,121,0,47,0,0,18,97,48,0,18,121,0,18,97,49,0,18,97,50,
-0,18,121,0,48,46,48,46,48,47,58,115,105,103,110,0,18,120,0,0,0,48,20,0,0,1,90,95,0,0,10,0,97,115,
-105,110,0,1,1,0,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,
-18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105,110,0,18,118,
-0,59,121,0,0,0,20,0,0,1,90,95,0,0,11,0,97,115,105,110,0,1,1,0,0,11,118,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,59,120,0,58,97,115,105,110,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,
-97,108,0,59,121,0,58,97,115,105,110,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,
-0,59,122,0,58,97,115,105,110,0,18,118,0,59,122,0,0,0,20,0,0,1,90,95,0,0,12,0,97,115,105,110,0,1,1,
-0,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,18,118,0,59,120,
-0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105,110,0,18,118,0,59,121,0,0,0,
-20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,115,105,110,0,18,118,0,59,122,0,0,0,20,0,9,
-18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,115,105,110,0,18,118,0,59,119,0,0,0,20,0,0,1,90,95,
-0,0,9,0,97,99,111,115,0,1,1,0,0,9,120,0,0,0,1,3,2,90,95,1,0,9,1,104,97,108,102,80,105,0,2,17,51,0,
-49,52,49,53,57,50,54,0,0,17,48,0,53,0,0,48,0,0,9,18,95,95,114,101,116,86,97,108,0,18,104,97,108,
-102,80,105,0,58,97,115,105,110,0,18,120,0,0,0,47,20,0,0,1,90,95,0,0,10,0,97,99,111,115,0,1,1,0,0,
-10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,
-0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,0,
-1,90,95,0,0,11,0,97,99,111,115,0,1,1,0,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,
-0,58,97,99,111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,
-99,111,115,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,
-115,0,18,118,0,59,122,0,0,0,20,0,0,1,90,95,0,0,12,0,97,99,111,115,0,1,1,0,0,12,118,0,0,0,1,9,18,95,
-95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,
-101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,
-86,97,108,0,59,122,0,58,97,99,111,115,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,
-108,0,59,119,0,58,97,99,111,115,0,18,118,0,59,119,0,0,0,20,0,0,1,90,95,0,0,9,0,97,116,97,110,0,1,1,
-0,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,97,115,105,110,0,18,120,0,58,105,110,118,
-101,114,115,101,115,113,114,116,0,18,120,0,18,120,0,48,17,49,0,48,0,0,46,0,0,48,0,0,20,0,0,1,90,95,
-0,0,10,0,97,116,97,110,0,1,1,0,0,10,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,
-86,97,108,0,59,120,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,
-95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,
-121,0,0,0,20,0,0,1,90,95,0,0,11,0,97,116,97,110,0,1,1,0,0,11,121,95,111,118,101,114,95,120,0,0,0,1,
-9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,
-59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,121,95,111,118,
-101,114,95,120,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,
-18,121,95,111,118,101,114,95,120,0,59,122,0,0,0,20,0,0,1,90,95,0,0,12,0,97,116,97,110,0,1,1,0,0,12,
-121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,
-0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,
-58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,
-97,108,0,59,122,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,122,0,0,0,20,0,9,18,95,
-95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,119,0,
-0,0,20,0,0,1,90,95,0,0,9,0,97,116,97,110,0,1,1,0,0,9,121,0,0,1,1,0,0,9,120,0,0,0,1,3,2,90,95,0,0,9,
-1,114,0,0,0,10,58,97,98,115,0,18,120,0,0,0,17,49,0,48,0,45,52,0,41,0,2,9,18,114,0,58,97,116,97,110,
-0,18,121,0,18,120,0,49,0,0,20,0,10,18,120,0,17,48,0,48,0,0,40,0,2,9,18,114,0,18,114,0,58,115,105,
-103,110,0,18,121,0,0,0,17,51,0,49,52,49,53,57,51,0,0,48,46,20,0,0,9,14,0,0,2,9,18,114,0,58,115,105,
-103,110,0,18,121,0,0,0,17,49,0,53,55,48,55,57,54,53,0,0,48,20,0,0,8,18,114,0,0,0,1,90,95,0,0,10,0,
-97,116,97,110,0,1,1,0,0,10,117,0,0,1,1,0,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,
-120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,
-97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,0,1,90,95,0,0,
-11,0,97,116,97,110,0,1,1,0,0,11,117,0,0,1,1,0,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
-59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,
-86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,
-0,1,90,95,0,0,12,0,97,116,97,110,0,1,1,0,0,12,117,0,0,1,1,0,0,12,118,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,
-95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,
-20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,
-59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,18,117,0,59,119,0,
-0,18,118,0,59,119,0,0,0,20,0,0,1,90,95,0,0,9,0,112,111,119,0,1,1,0,0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,
-4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,
-0,0,0,1,90,95,0,0,10,0,112,111,119,0,1,1,0,0,10,97,0,0,1,1,0,0,10,98,0,0,0,1,4,102,108,111,97,116,
-95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,
-120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,
-0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,0,1,90,95,0,0,11,0,112,111,119,0,1,1,0,0,11,97,0,0,1,1,0,
-0,11,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,
-120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,
-95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,
-116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,
-0,59,122,0,0,0,0,1,90,95,0,0,12,0,112,111,119,0,1,1,0,0,12,97,0,0,1,1,0,0,12,98,0,0,0,1,4,102,108,
-111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,
-18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,
-0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,
-18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,
-97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,18,
-98,0,59,119,0,0,0,0,1,90,95,0,0,9,0,101,120,112,0,1,1,0,0,9,97,0,0,0,1,3,2,90,95,0,0,9,1,116,0,2,
-18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,
-95,114,101,116,86,97,108,0,0,18,116,0,0,0,0,1,90,95,0,0,10,0,101,120,112,0,1,1,0,0,10,97,0,0,0,1,3,
-2,90,95,0,0,10,1,116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0,0,4,102,108,111,97,116,
-95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,116,0,59,120,0,0,0,4,102,108,
-111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,116,0,59,121,0,0,0,0,
-1,90,95,0,0,11,0,101,120,112,0,1,1,0,0,11,97,0,0,0,1,3,2,90,95,0,0,11,1,116,0,2,18,97,0,17,49,0,52,
-52,50,54,57,53,48,50,0,0,48,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,
-97,108,0,59,120,0,0,18,116,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,
-101,116,86,97,108,0,59,121,0,0,18,116,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,
-95,95,114,101,116,86,97,108,0,59,122,0,0,18,116,0,59,122,0,0,0,0,1,90,95,0,0,12,0,101,120,112,0,1,
-1,0,0,12,97,0,0,0,1,3,2,90,95,0,0,12,1,116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0,0,
+114,97,100,105,97,110,115,0,59,121,0,0,0,0,1,90,95,0,0,11,0,0,99,111,115,0,1,1,0,0,11,0,114,97,100,
+105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,
+108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,99,111,115,
+105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,
+0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,
+18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,90,95,0,0,12,0,0,99,111,115,0,1,1,0,0,12,0,114,97,
+100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,
+86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,99,111,
+115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,
+121,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,
+0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,
+18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,110,115,0,59,119,0,0,0,0,1,90,95,
+0,0,9,0,0,116,97,110,0,1,1,0,0,9,0,97,110,103,108,101,0,0,0,1,3,2,90,95,1,0,9,0,1,115,0,2,58,115,
+105,110,0,0,18,97,110,103,108,101,0,0,0,0,0,3,2,90,95,1,0,9,0,1,99,0,2,58,99,111,115,0,0,18,97,110,
+103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,90,95,0,0,10,0,0,116,97,110,0,1,1,0,0,10,0,97,
+110,103,108,101,0,0,0,1,3,2,90,95,1,0,10,0,1,115,0,2,58,115,105,110,0,0,18,97,110,103,108,101,0,0,
+0,0,0,3,2,90,95,1,0,10,0,1,99,0,2,58,99,111,115,0,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,
+99,0,49,0,0,1,90,95,0,0,11,0,0,116,97,110,0,1,1,0,0,11,0,97,110,103,108,101,0,0,0,1,3,2,90,95,1,0,
+11,0,1,115,0,2,58,115,105,110,0,0,18,97,110,103,108,101,0,0,0,0,0,3,2,90,95,1,0,11,0,1,99,0,2,58,
+99,111,115,0,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,90,95,0,0,12,0,0,116,97,
+110,0,1,1,0,0,12,0,97,110,103,108,101,0,0,0,1,3,2,90,95,1,0,12,0,1,115,0,2,58,115,105,110,0,0,18,
+97,110,103,108,101,0,0,0,0,0,3,2,90,95,1,0,12,0,1,99,0,2,58,99,111,115,0,0,18,97,110,103,108,101,0,
+0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,90,95,0,0,9,0,0,97,115,105,110,0,1,1,0,0,9,0,120,0,0,0,1,3,2,
+90,95,1,0,9,0,1,97,48,0,2,17,49,0,53,55,48,55,50,56,56,0,0,0,0,3,2,90,95,1,0,9,0,1,97,49,0,2,17,48,
+0,50,49,50,49,49,52,52,0,0,54,0,0,3,2,90,95,1,0,9,0,1,97,50,0,2,17,48,0,48,55,52,50,54,49,48,0,0,0,
+0,3,2,90,95,1,0,9,0,1,104,97,108,102,80,105,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,48,0,53,0,0,48,
+0,0,3,2,90,95,1,0,9,0,1,121,0,2,58,97,98,115,0,0,18,120,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,
+0,18,104,97,108,102,80,105,0,58,115,113,114,116,0,0,17,49,0,48,0,0,18,121,0,47,0,0,18,97,48,0,18,
+121,0,18,97,49,0,18,97,50,0,18,121,0,48,46,48,46,48,47,58,115,105,103,110,0,0,18,120,0,0,0,48,20,0,
+0,1,90,95,0,0,10,0,0,97,115,105,110,0,1,1,0,0,10,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
+59,120,0,58,97,115,105,110,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,
+121,0,58,97,115,105,110,0,0,18,118,0,59,121,0,0,0,20,0,0,1,90,95,0,0,11,0,0,97,115,105,110,0,1,1,0,
+0,11,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,0,18,118,0,59,
+120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105,110,0,0,18,118,0,59,121,0,
+0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,115,105,110,0,0,18,118,0,59,122,0,0,0,
+20,0,0,1,90,95,0,0,12,0,0,97,115,105,110,0,1,1,0,0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,59,120,0,58,97,115,105,110,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,
+59,121,0,58,97,115,105,110,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,
+122,0,58,97,115,105,110,0,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,
+58,97,115,105,110,0,0,18,118,0,59,119,0,0,0,20,0,0,1,90,95,0,0,9,0,0,97,99,111,115,0,1,1,0,0,9,0,
+120,0,0,0,1,3,2,90,95,1,0,9,0,1,104,97,108,102,80,105,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,48,0,
+53,0,0,48,0,0,9,18,95,95,114,101,116,86,97,108,0,18,104,97,108,102,80,105,0,58,97,115,105,110,0,0,
+18,120,0,0,0,47,20,0,0,1,90,95,0,0,10,0,0,97,99,111,115,0,1,1,0,0,10,0,118,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,59,120,0,58,97,99,111,115,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,59,121,0,58,97,99,111,115,0,0,18,118,0,59,121,0,0,0,20,0,0,1,90,95,0,0,11,0,0,97,
+99,111,115,0,1,1,0,0,11,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115,
+0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,0,18,
+118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,115,0,0,18,118,0,
+59,122,0,0,0,20,0,0,1,90,95,0,0,12,0,0,97,99,111,115,0,1,1,0,0,12,0,118,0,0,0,1,9,18,95,95,114,101,
+116,86,97,108,0,59,120,0,58,97,99,111,115,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,
+97,108,0,59,121,0,58,97,99,111,115,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,
+0,59,122,0,58,97,99,111,115,0,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,
+119,0,58,97,99,111,115,0,0,18,118,0,59,119,0,0,0,20,0,0,1,90,95,0,0,9,0,0,97,116,97,110,0,1,1,0,0,
+9,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,97,115,105,110,0,0,18,120,0,58,105,110,118,
+101,114,115,101,115,113,114,116,0,0,18,120,0,18,120,0,48,17,49,0,48,0,0,46,0,0,48,0,0,20,0,0,1,90,
+95,0,0,10,0,0,97,116,97,110,0,1,1,0,0,10,0,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,
+20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,0,18,121,95,111,118,101,114,95,
+120,0,59,121,0,0,0,20,0,0,1,90,95,0,0,11,0,0,97,116,97,110,0,1,1,0,0,11,0,121,95,111,118,101,114,
+95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,0,18,121,95,111,118,
+101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,
+0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,
+58,97,116,97,110,0,0,18,121,95,111,118,101,114,95,120,0,59,122,0,0,0,20,0,0,1,90,95,0,0,12,0,0,97,
+116,97,110,0,1,1,0,0,12,0,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
+59,120,0,58,97,116,97,110,0,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,
+101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,
+20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,0,18,121,95,111,118,101,114,95,
+120,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,0,18,121,95,
+111,118,101,114,95,120,0,59,119,0,0,0,20,0,0,1,90,95,0,0,9,0,0,97,116,97,110,0,1,1,0,0,9,0,121,0,0,
+1,1,0,0,9,0,120,0,0,0,1,3,2,90,95,0,0,9,0,1,114,0,0,0,10,58,97,98,115,0,0,18,120,0,0,0,17,49,0,48,
+0,45,52,0,41,0,2,9,18,114,0,58,97,116,97,110,0,0,18,121,0,18,120,0,49,0,0,20,0,10,18,120,0,17,48,0,
+48,0,0,40,0,2,9,18,114,0,18,114,0,58,115,105,103,110,0,0,18,121,0,0,0,17,51,0,49,52,49,53,57,51,0,
+0,48,46,20,0,0,9,14,0,0,2,9,18,114,0,58,115,105,103,110,0,0,18,121,0,0,0,17,49,0,53,55,48,55,57,54,
+53,0,0,48,20,0,0,8,18,114,0,0,0,1,90,95,0,0,10,0,0,97,116,97,110,0,1,1,0,0,10,0,117,0,0,1,1,0,0,10,
+0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,0,18,117,0,59,120,0,0,
+18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,0,18,117,
+0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,0,1,90,95,0,0,11,0,0,97,116,97,110,0,1,1,0,0,11,0,117,0,0,
+1,1,0,0,11,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,0,18,117,0,
+59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,
+0,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,
+97,116,97,110,0,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,0,1,90,95,0,0,12,0,0,97,116,97,
+110,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,
+97,116,97,110,0,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,
+0,59,121,0,58,97,116,97,110,0,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,59,122,0,58,97,116,97,110,0,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,9,18,
+95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,0,18,117,0,59,119,0,0,18,118,0,59,119,0,
+0,0,20,0,0,1,90,95,0,0,9,0,0,112,111,119,0,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,102,108,111,
+97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,
+0,0,10,0,0,112,111,119,0,1,1,0,0,10,0,97,0,0,1,1,0,0,10,0,98,0,0,0,1,4,102,108,111,97,116,95,112,
+111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,
+0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,
+97,0,59,121,0,0,18,98,0,59,121,0,0,0,0,1,90,95,0,0,11,0,0,112,111,119,0,1,1,0,0,11,0,97,0,0,1,1,0,
+0,11,0,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,
+59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,
+18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,
+97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,
+98,0,59,122,0,0,0,0,1,90,95,0,0,12,0,0,112,111,119,0,1,1,0,0,12,0,97,0,0,1,1,0,0,12,0,98,0,0,0,1,4,
+102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,
+120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,
+97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,
+101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,4,
+102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59,
+119,0,0,18,98,0,59,119,0,0,0,0,1,90,95,0,0,9,0,0,101,120,112,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,
+0,9,0,1,116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0,0,4,102,108,111,97,116,95,101,120,
+112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,116,0,0,0,0,1,90,95,0,0,10,0,0,101,120,112,0,1,1,0,
+0,10,0,97,0,0,0,1,3,2,90,95,0,0,10,0,1,116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0,0,
 4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,116,0,59,
 120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,
-116,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,
-122,0,0,18,116,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,
-108,0,59,119,0,0,18,116,0,59,119,0,0,0,0,1,90,95,0,0,9,0,108,111,103,50,0,1,1,0,0,9,120,0,0,0,1,4,
-102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,
-0,10,0,108,111,103,50,0,1,1,0,0,10,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,
-114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,
-18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,90,95,0,0,11,0,108,111,103,
-50,0,1,1,0,0,11,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,
-108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,
-116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,
-114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1,90,95,0,0,12,0,108,111,103,50,0,1,1,0,
-0,12,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,
-0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,
-0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,
-86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,
-101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,90,95,0,0,9,0,108,111,103,0,1,1,0,0,9,120,
-0,0,0,1,3,2,90,95,1,0,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,
-18,120,0,0,0,18,99,0,48,0,0,1,90,95,0,0,10,0,108,111,103,0,1,1,0,0,10,118,0,0,0,1,3,2,90,95,1,0,9,
-1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,
-0,0,1,90,95,0,0,11,0,108,111,103,0,1,1,0,0,11,118,0,0,0,1,3,2,90,95,1,0,9,1,99,0,2,17,48,0,54,57,
-51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,90,95,0,0,12,0,
-108,111,103,0,1,1,0,0,12,118,0,0,0,1,3,2,90,95,1,0,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,
-0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,90,95,0,0,9,0,101,120,112,50,0,1,1,0,0,9,
-97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,
-0,1,90,95,0,0,10,0,101,120,112,50,0,1,1,0,0,10,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,
-18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,
-112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,0,1,90,95,0,0,11,0,101,
-120,112,50,0,1,1,0,0,11,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,
-86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,
-101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,
-95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,0,1,90,95,0,0,12,0,101,120,112,50,0,1,1,
-0,0,12,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,
-0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,
-59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,
-97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,
-116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,0,0,1,90,95,0,0,9,0,115,113,114,116,0,1,1,0,0,9,120,
-0,0,0,1,3,2,90,95,0,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,120,0,0,0,
-4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,90,95,0,0,
-10,0,115,113,114,116,0,1,1,0,0,10,118,0,0,0,1,3,2,90,95,0,0,9,1,114,0,0,0,4,102,108,111,97,116,95,
-114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,
-101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,
-118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,
-0,18,114,0,0,0,0,1,90,95,0,0,11,0,115,113,114,116,0,1,1,0,0,11,118,0,0,0,1,3,2,90,95,0,0,9,1,114,0,
-0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,
-114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,
-115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,
-116,86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,
-59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,
-114,0,0,0,0,1,90,95,0,0,12,0,115,113,114,116,0,1,1,0,0,12,118,0,0,0,1,3,2,90,95,0,0,9,1,114,0,0,0,
-4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,
-99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,
-113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,
-86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,
-122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,
-0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,119,0,0,0,4,102,108,111,97,116,
-95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,0,0,0,0,1,90,95,0,0,9,0,105,110,
-118,101,114,115,101,115,113,114,116,0,1,1,0,0,9,120,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,
-18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0,105,110,118,101,114,
-115,101,115,113,114,116,0,1,1,0,0,10,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,
-114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,
-95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,90,95,0,0,11,0,105,110,118,101,
-114,115,101,115,113,114,116,0,1,1,0,0,11,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,
-95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,
-18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,
-113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1,90,95,0,0,12,0,105,110,
-118,101,114,115,101,115,113,114,116,0,1,1,0,0,12,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,
-18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,
-113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,
-114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,
-116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,90,95,0,
-0,9,0,110,111,114,109,97,108,105,122,101,0,1,1,0,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
-0,17,49,0,48,0,0,20,0,0,1,90,95,0,0,10,0,110,111,114,109,97,108,105,122,101,0,1,1,0,0,10,118,0,0,0,
-1,3,2,90,95,1,0,9,1,115,0,2,58,105,110,118,101,114,115,101,115,113,114,116,0,58,100,111,116,0,18,
-118,0,0,18,118,0,0,0,0,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,
-116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,115,0,0,0,0,1,90,95,0,0,11,0,110,111,114,109,97,108,
-105,122,101,0,1,1,0,0,11,118,0,0,0,1,3,2,90,95,0,0,9,1,116,109,112,0,0,0,4,118,101,99,51,95,100,
-111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,
-109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,
-101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,90,95,0,0,12,0,110,111,
-114,109,97,108,105,122,101,0,1,1,0,0,12,118,0,0,0,1,3,2,90,95,0,0,9,1,116,109,112,0,0,0,4,118,101,
-99,52,95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,
-113,0,18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,
-18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,90,95,0,0,
-9,0,97,98,115,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,
-0,0,18,97,0,0,0,0,1,90,95,0,0,10,0,97,98,115,0,1,1,0,0,10,97,0,0,0,1,4,118,101,99,52,95,97,98,115,
-0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,90,95,0,0,11,0,97,98,115,0,1,1,0,
-0,11,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,
-18,97,0,0,0,0,1,90,95,0,0,12,0,97,98,115,0,1,1,0,0,12,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,
-95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,9,0,115,105,103,110,0,1,1,0,0,9,120,0,0,
-0,1,3,2,90,95,0,0,9,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,0,18,120,0,0,
-17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0,0,0,18,120,0,0,0,4,118,
-101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,
-0,0,0,1,90,95,0,0,10,0,115,105,103,110,0,1,1,0,0,10,118,0,0,0,1,3,2,90,95,0,0,10,1,112,0,0,1,1,110,
-0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,
-101,99,52,95,115,103,116,0,18,110,0,59,120,121,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,
-95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,0,18,
-110,0,0,0,0,1,90,95,0,0,11,0,115,105,103,110,0,1,1,0,0,11,118,0,0,0,1,3,2,90,95,0,0,11,1,112,0,0,1,
-1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,
-0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,59,120,121,122,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,
-118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,
-0,18,112,0,0,18,110,0,0,0,0,1,90,95,0,0,12,0,115,105,103,110,0,1,1,0,0,12,118,0,0,0,1,3,2,90,95,0,
-0,12,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,0,18,118,0,0,17,48,0,48,0,0,
+116,0,59,121,0,0,0,0,1,90,95,0,0,11,0,0,101,120,112,0,1,1,0,0,11,0,97,0,0,0,1,3,2,90,95,0,0,11,0,1,
+116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0,0,4,102,108,111,97,116,95,101,120,112,50,
+0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,116,0,59,120,0,0,0,4,102,108,111,97,116,95,101,
+120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,116,0,59,121,0,0,0,4,102,108,111,97,
+116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,116,0,59,122,0,0,0,0,1,90,
+95,0,0,12,0,0,101,120,112,0,1,1,0,0,12,0,97,0,0,0,1,3,2,90,95,0,0,12,0,1,116,0,2,18,97,0,17,49,0,
+52,52,50,54,57,53,48,50,0,0,48,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,
+86,97,108,0,59,120,0,0,18,116,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,
+101,116,86,97,108,0,59,121,0,0,18,116,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,
+95,95,114,101,116,86,97,108,0,59,122,0,0,18,116,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120,112,
+50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,116,0,59,119,0,0,0,0,1,90,95,0,0,9,0,0,108,111,
+103,50,0,1,1,0,0,9,0,120,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,
+97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0,0,108,111,103,50,0,1,1,0,0,10,0,118,0,0,0,1,4,102,108,
+111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,
+102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,
+0,0,0,0,1,90,95,0,0,11,0,0,108,111,103,50,0,1,1,0,0,11,0,118,0,0,0,1,4,102,108,111,97,116,95,108,
+111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,
+116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,
+108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,
+0,0,1,90,95,0,0,12,0,0,108,111,103,50,0,1,1,0,0,12,0,118,0,0,0,1,4,102,108,111,97,116,95,108,111,
+103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,
+108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,
+97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,
+108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,
+0,0,1,90,95,0,0,9,0,0,108,111,103,0,1,1,0,0,9,0,120,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,48,0,54,
+57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,0,18,120,0,0,0,18,99,0,48,0,0,1,90,95,0,0,10,
+0,0,108,111,103,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,48,0,54,57,51,49,52,55,49,
+56,49,0,0,0,0,8,58,108,111,103,50,0,0,18,118,0,0,0,18,99,0,48,0,0,1,90,95,0,0,11,0,0,108,111,103,0,
+1,1,0,0,11,0,118,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,
+58,108,111,103,50,0,0,18,118,0,0,0,18,99,0,48,0,0,1,90,95,0,0,12,0,0,108,111,103,0,1,1,0,0,12,0,
+118,0,0,0,1,3,2,90,95,1,0,9,0,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,
+50,0,0,18,118,0,0,0,18,99,0,48,0,0,1,90,95,0,0,9,0,0,101,120,112,50,0,1,1,0,0,9,0,97,0,0,0,1,4,102,
+108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,10,
+0,0,101,120,112,50,0,1,1,0,0,10,0,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,
+101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,
+95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,0,1,90,95,0,0,11,0,0,101,120,112,50,0,1,
+1,0,0,11,0,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,
+120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,
+108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,
+116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,0,1,90,95,0,0,12,0,0,101,120,112,50,0,1,1,0,0,12,0,
+97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,
+97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,
+0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,
+59,122,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,
+97,108,0,59,119,0,0,18,97,0,59,119,0,0,0,0,1,90,95,0,0,9,0,0,115,113,114,116,0,1,1,0,0,9,0,120,0,0,
+0,1,3,2,90,95,0,0,9,0,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,120,0,0,0,4,
+102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,90,95,0,0,
+10,0,0,115,113,114,116,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,114,0,0,0,4,102,108,111,97,
+116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,
+95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,
+0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,
+121,0,0,18,114,0,0,0,0,1,90,95,0,0,11,0,0,115,113,114,116,0,1,1,0,0,11,0,118,0,0,0,1,3,2,90,95,0,0,
+9,0,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,
+111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,
+97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,
+95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,
+0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,
+122,0,0,18,114,0,0,0,0,1,90,95,0,0,12,0,0,115,113,114,116,0,1,1,0,0,12,0,118,0,0,0,1,3,2,90,95,0,0,
+9,0,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,
+111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,
+97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,
+95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,
+0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,
+122,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,119,0,0,0,4,102,
+108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,0,0,0,0,1,90,95,
+0,0,9,0,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,0,9,0,120,0,0,0,1,4,102,108,111,97,
+116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0,0,
+105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,0,10,0,118,0,0,0,1,4,102,108,111,97,116,95,114,
+115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,
+95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,90,95,0,0,
+11,0,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,0,11,0,118,0,0,0,1,4,102,108,111,97,116,
+95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,
+97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,
+108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,
+0,1,90,95,0,0,12,0,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,0,12,0,118,0,0,0,1,4,102,
+108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,
+4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,
+0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,
+59,122,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,
+118,0,59,119,0,0,0,0,1,90,95,0,0,9,0,0,110,111,114,109,97,108,105,122,101,0,1,1,0,0,9,0,120,0,0,0,
+1,9,18,95,95,114,101,116,86,97,108,0,17,49,0,48,0,0,20,0,0,1,90,95,0,0,10,0,0,110,111,114,109,97,
+108,105,122,101,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,1,0,9,0,1,115,0,2,58,105,110,118,101,114,115,
+101,115,113,114,116,0,0,58,100,111,116,0,0,18,118,0,0,18,118,0,0,0,0,0,0,0,4,118,101,99,52,95,109,
+117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,115,0,
+0,0,0,1,90,95,0,0,11,0,0,110,111,114,109,97,108,105,122,101,0,1,1,0,0,11,0,118,0,0,0,1,3,2,90,95,0,
+0,9,0,1,116,109,112,0,0,0,4,118,101,99,51,95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,
+0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,
+95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,
+0,18,116,109,112,0,0,0,0,1,90,95,0,0,12,0,0,110,111,114,109,97,108,105,122,101,0,1,1,0,0,12,0,118,
+0,0,0,1,3,2,90,95,0,0,9,0,1,116,109,112,0,0,0,4,118,101,99,52,95,100,111,116,0,18,116,109,112,0,0,
+18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112,0,0,18,116,109,112,0,
+0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,
+121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,90,95,0,0,9,0,0,97,98,115,0,1,1,0,0,9,0,97,0,0,0,1,
+4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,10,0,0,
+97,98,115,0,1,1,0,0,10,0,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,
+0,59,120,121,0,0,18,97,0,0,0,0,1,90,95,0,0,11,0,0,97,98,115,0,1,1,0,0,11,0,97,0,0,0,1,4,118,101,99,
+52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,90,95,0,0,12,
+0,0,97,98,115,0,1,1,0,0,12,0,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,
+108,0,0,18,97,0,0,0,0,1,90,95,0,0,9,0,0,115,105,103,110,0,1,1,0,0,9,0,120,0,0,0,1,3,2,90,95,0,0,9,
+0,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,0,18,120,0,0,17,48,0,48,0,0,0,
+0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0,0,0,18,120,0,0,0,4,118,101,99,52,95,115,
+117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,0,0,0,1,90,95,0,
+0,10,0,0,115,105,103,110,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,0,0,10,0,1,112,0,0,1,1,110,0,0,0,4,
+118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,
+52,95,115,103,116,0,18,110,0,59,120,121,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,
+117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,0,18,110,0,0,0,
+0,1,90,95,0,0,11,0,0,115,105,103,110,0,1,1,0,0,11,0,118,0,0,0,1,3,2,90,95,0,0,11,0,1,112,0,0,1,1,
+110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,
+0,4,118,101,99,52,95,115,103,116,0,18,110,0,59,120,121,122,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,
+101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,
+112,0,0,18,110,0,0,0,0,1,90,95,0,0,12,0,0,115,105,103,110,0,1,1,0,0,12,0,118,0,0,0,1,3,2,90,95,0,0,
+12,0,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,0,18,118,0,0,17,48,0,48,0,0,
 0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,
 115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,0,0,0,1,90,
-95,0,0,9,0,102,108,111,111,114,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,
-95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,10,0,102,108,111,111,114,0,1,1,0,0,10,97,
-0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,
-18,97,0,0,0,0,1,90,95,0,0,11,0,102,108,111,111,114,0,1,1,0,0,11,97,0,0,0,1,4,118,101,99,52,95,102,
-108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,90,95,0,0,12,
-0,102,108,111,111,114,0,1,1,0,0,12,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,
-114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,9,0,99,101,105,108,0,1,1,0,0,9,97,0,0,0,1,3,2,
-90,95,0,0,9,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,
-9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,90,95,0,0,10,0,99,101,105,108,0,1,1,0,0,10,
-97,0,0,0,1,3,2,90,95,0,0,10,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,
-0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,54,20,0,0,1,90,95,0,0,11,0,
-99,101,105,108,0,1,1,0,0,11,97,0,0,0,1,3,2,90,95,0,0,11,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,
-102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,
-98,0,54,20,0,0,1,90,95,0,0,12,0,99,101,105,108,0,1,1,0,0,12,97,0,0,0,1,3,2,90,95,0,0,12,1,98,0,2,
-18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,
-116,86,97,108,0,18,98,0,54,20,0,0,1,90,95,0,0,9,0,102,114,97,99,116,0,1,1,0,0,9,97,0,0,0,1,4,118,
-101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,10,0,102,
-114,97,99,116,0,1,1,0,0,10,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,
-97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,90,95,0,0,11,0,102,114,97,99,116,0,1,1,0,0,11,97,0,0,0,1,4,
-118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,
-1,90,95,0,0,12,0,102,114,97,99,116,0,1,1,0,0,12,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,
-95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,9,0,109,111,100,0,1,1,0,0,9,97,0,0,1,1,0,
-0,9,98,0,0,0,1,3,2,90,95,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,
-99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,
-0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,
-1,90,95,0,0,10,0,109,111,100,0,1,1,0,0,10,97,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,0,0,9,1,111,110,
-101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,
-0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,97,0,18,98,0,58,102,108,111,111,
-114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,11,0,109,111,100,
-0,1,1,0,0,11,97,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,
-102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,
-101,116,86,97,108,0,59,120,121,122,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,
-101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,12,0,109,111,100,0,1,1,0,0,12,97,0,0,1,1,0,
-0,9,98,0,0,0,1,3,2,90,95,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,
-99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,
-0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,
-1,90,95,0,0,10,0,109,111,100,0,1,1,0,0,10,97,0,0,1,1,0,0,10,98,0,0,0,1,3,2,90,95,0,0,10,1,111,110,
-101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,
-0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,
-114,66,0,59,121,0,0,18,98,0,59,121,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,
-108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,11,0,
-109,111,100,0,1,1,0,0,11,97,0,0,1,1,0,0,11,98,0,0,0,1,3,2,90,95,0,0,11,1,111,110,101,79,118,101,
-114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,120,0,0,18,
-98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,121,0,
-0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,
-122,0,0,18,98,0,59,122,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,
-114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,12,0,109,111,100,
-0,1,1,0,0,12,97,0,0,1,1,0,0,12,98,0,0,0,1,3,2,90,95,0,0,12,1,111,110,101,79,118,101,114,66,0,0,0,4,
-102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,120,0,0,18,98,0,59,120,0,
-0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,121,0,0,18,98,0,59,
-121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,122,0,0,18,98,
-0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,119,0,0,
-18,98,0,59,119,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,
-18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,9,0,109,105,110,0,1,1,0,
-0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,
-0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,109,105,110,0,1,1,0,0,10,97,0,0,1,1,0,0,10,98,0,0,0,1,4,
-118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,
-0,0,18,98,0,59,120,121,0,0,0,0,1,90,95,0,0,11,0,109,105,110,0,1,1,0,0,11,97,0,0,1,1,0,0,11,98,0,0,
-0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,
-59,120,121,122,0,0,18,98,0,59,120,121,122,0,0,0,0,1,90,95,0,0,12,0,109,105,110,0,1,1,0,0,12,97,0,0,
-1,1,0,0,12,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,
-0,18,98,0,0,0,0,1,90,95,0,0,10,0,109,105,110,0,1,1,0,0,10,97,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,
-52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,90,
-95,0,0,11,0,109,105,110,0,1,1,0,0,11,97,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,
-18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,109,
-105,110,0,1,1,0,0,12,97,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,
-116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,0,109,97,120,0,1,1,0,0,9,97,0,0,1,1,0,0,9,
-98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,
-0,0,1,90,95,0,0,10,0,109,97,120,0,1,1,0,0,10,97,0,0,1,1,0,0,10,98,0,0,0,1,4,118,101,99,52,95,109,
-97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,
-0,0,0,0,1,90,95,0,0,11,0,109,97,120,0,1,1,0,0,11,97,0,0,1,1,0,0,11,98,0,0,0,1,4,118,101,99,52,95,
-109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,
-0,59,120,121,122,0,0,0,0,1,90,95,0,0,12,0,109,97,120,0,1,1,0,0,12,97,0,0,1,1,0,0,12,98,0,0,0,1,4,
+95,0,0,9,0,0,102,108,111,111,114,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,
+18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,10,0,0,102,108,111,111,114,0,1,1,0,0,
+10,0,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,
+121,0,0,18,97,0,0,0,0,1,90,95,0,0,11,0,0,102,108,111,111,114,0,1,1,0,0,11,0,97,0,0,0,1,4,118,101,
+99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,
+90,95,0,0,12,0,0,102,108,111,111,114,0,1,1,0,0,12,0,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,
+114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,9,0,0,99,101,105,108,0,1,1,0,0,
+9,0,97,0,0,0,1,3,2,90,95,0,0,9,0,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,
+18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,90,95,0,0,10,0,0,99,
+101,105,108,0,1,1,0,0,10,0,97,0,0,0,1,3,2,90,95,0,0,10,0,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,
+95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,
+98,0,54,20,0,0,1,90,95,0,0,11,0,0,99,101,105,108,0,1,1,0,0,11,0,97,0,0,0,1,3,2,90,95,0,0,11,0,1,98,
+0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,
+101,116,86,97,108,0,59,120,121,122,0,18,98,0,54,20,0,0,1,90,95,0,0,12,0,0,99,101,105,108,0,1,1,0,0,
+12,0,97,0,0,0,1,3,2,90,95,0,0,12,0,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,
+0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,90,95,0,0,9,0,0,102,
+114,97,99,116,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,
+97,108,0,0,18,97,0,0,0,0,1,90,95,0,0,10,0,0,102,114,97,99,116,0,1,1,0,0,10,0,97,0,0,0,1,4,118,101,
+99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,90,95,0,0,
+11,0,0,102,114,97,99,116,0,1,1,0,0,11,0,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,
+101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,90,95,0,0,12,0,0,102,114,97,99,116,0,1,1,0,
+0,12,0,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,
+0,0,1,90,95,0,0,9,0,0,109,111,100,0,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,
+111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,
+114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,
+0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,10,0,0,109,111,100,0,
+1,1,0,0,10,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,111,110,101,79,118,101,114,66,0,0,0,
+4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,
+114,101,116,86,97,108,0,59,120,121,0,18,97,0,18,98,0,58,102,108,111,111,114,0,0,18,97,0,18,111,110,
+101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,11,0,0,109,111,100,0,1,1,0,0,11,0,97,0,0,1,
+1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,
+95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,
+0,59,120,121,122,0,18,97,0,18,98,0,58,102,108,111,111,114,0,0,18,97,0,18,111,110,101,79,118,101,
+114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,12,0,0,109,111,100,0,1,1,0,0,12,0,97,0,0,1,1,0,0,9,0,98,0,
+0,0,1,3,2,90,95,0,0,9,0,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,
+18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,
+0,58,102,108,111,111,114,0,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,
+95,0,0,10,0,0,109,111,100,0,1,1,0,0,10,0,97,0,0,1,1,0,0,10,0,98,0,0,0,1,3,2,90,95,0,0,10,0,1,111,
+110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,
+66,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,
+101,114,66,0,59,121,0,0,18,98,0,59,121,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,
+102,108,111,111,114,0,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,
+11,0,0,109,111,100,0,1,1,0,0,11,0,97,0,0,1,1,0,0,11,0,98,0,0,0,1,3,2,90,95,0,0,11,0,1,111,110,101,
+79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,
+120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,
+0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,
+114,66,0,59,122,0,0,18,98,0,59,122,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,
+108,111,111,114,0,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,95,0,0,12,
+0,0,109,111,100,0,1,1,0,0,12,0,97,0,0,1,1,0,0,12,0,98,0,0,0,1,3,2,90,95,0,0,12,0,1,111,110,101,79,
+118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,
+120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,
+0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,
+114,66,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,
+118,101,114,66,0,59,119,0,0,18,98,0,59,119,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,
+0,58,102,108,111,111,114,0,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,90,
+95,0,0,9,0,0,109,105,110,0,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,105,
+110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,0,109,105,110,0,
+1,1,0,0,10,0,97,0,0,1,1,0,0,10,0,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,
+86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0,1,90,95,0,0,11,0,0,
+109,105,110,0,1,1,0,0,11,0,97,0,0,1,1,0,0,11,0,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,
+95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0,
+0,0,0,1,90,95,0,0,12,0,0,109,105,110,0,1,1,0,0,12,0,97,0,0,1,1,0,0,12,0,98,0,0,0,1,4,118,101,99,52,
+95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,0,109,
+105,110,0,1,1,0,0,10,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,
+101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,0,109,105,110,0,1,1,0,
+0,11,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,
+108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,0,109,105,110,0,1,1,0,0,12,0,97,
+0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,
+97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,0,0,109,97,120,0,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,
 118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,
-0,10,0,109,97,120,0,1,1,0,0,10,97,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,
-95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,109,97,120,0,1,
-1,0,0,11,97,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,
-108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,109,97,120,0,1,1,0,0,12,97,0,0,1,
-1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,
-98,0,0,0,0,1,90,95,0,0,9,0,99,108,97,109,112,0,1,1,0,0,9,118,97,108,0,0,1,1,0,0,9,109,105,110,86,
-97,108,0,0,1,1,0,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,
-114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,
-0,0,0,1,90,95,0,0,10,0,99,108,97,109,112,0,1,1,0,0,10,118,97,108,0,0,1,1,0,0,9,109,105,110,86,97,
-108,0,0,1,1,0,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,
-101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,
-0,1,90,95,0,0,11,0,99,108,97,109,112,0,1,1,0,0,11,118,97,108,0,0,1,1,0,0,9,109,105,110,86,97,108,0,
-0,1,1,0,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,
-116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,
-90,95,0,0,12,0,99,108,97,109,112,0,1,1,0,0,12,118,97,108,0,0,1,1,0,0,9,109,105,110,86,97,108,0,0,1,
-1,0,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,
-86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,
-95,0,0,10,0,99,108,97,109,112,0,1,1,0,0,10,118,97,108,0,0,1,1,0,0,10,109,105,110,86,97,108,0,0,1,1,
-0,0,10,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,
-97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,
-0,0,11,0,99,108,97,109,112,0,1,1,0,0,11,118,97,108,0,0,1,1,0,0,11,109,105,110,86,97,108,0,0,1,1,0,
-0,11,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,
-97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,
-0,0,12,0,99,108,97,109,112,0,1,1,0,0,12,118,97,108,0,0,1,1,0,0,12,109,105,110,86,97,108,0,0,1,1,0,
-0,12,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,
-97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,
-0,0,9,0,109,105,120,0,1,1,0,0,9,120,0,0,1,1,0,0,9,121,0,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,
-108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,
-10,0,109,105,120,0,1,1,0,0,10,120,0,0,1,1,0,0,10,121,0,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,
-108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,
-11,0,109,105,120,0,1,1,0,0,11,120,0,0,1,1,0,0,11,121,0,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,
-108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,
-12,0,109,105,120,0,1,1,0,0,12,120,0,0,1,1,0,0,12,121,0,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,
-108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,
-10,0,109,105,120,0,1,1,0,0,10,120,0,0,1,1,0,0,10,121,0,0,1,1,0,0,10,97,0,0,0,1,4,118,101,99,52,95,
-108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,
-11,0,109,105,120,0,1,1,0,0,11,120,0,0,1,1,0,0,11,121,0,0,1,1,0,0,11,97,0,0,0,1,4,118,101,99,52,95,
-108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,
-12,0,109,105,120,0,1,1,0,0,12,120,0,0,1,1,0,0,12,121,0,0,1,1,0,0,12,97,0,0,0,1,4,118,101,99,52,95,
-108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,9,
-0,115,116,101,112,0,1,1,0,0,9,101,100,103,101,0,0,1,1,0,0,9,120,0,0,0,1,4,118,101,99,52,95,115,103,
-101,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,90,95,0,0,10,0,
-115,116,101,112,0,1,1,0,0,10,101,100,103,101,0,0,1,1,0,0,10,120,0,0,0,1,4,118,101,99,52,95,115,103,
-101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,90,
-95,0,0,11,0,115,116,101,112,0,1,1,0,0,11,101,100,103,101,0,0,1,1,0,0,11,120,0,0,0,1,4,118,101,99,
-52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,120,0,0,18,101,100,103,
-101,0,0,0,0,1,90,95,0,0,12,0,115,116,101,112,0,1,1,0,0,12,101,100,103,101,0,0,1,1,0,0,12,120,0,0,0,
-1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,
-101,0,0,0,0,1,90,95,0,0,10,0,115,116,101,112,0,1,1,0,0,9,101,100,103,101,0,0,1,1,0,0,10,118,0,0,0,
-1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,
-101,100,103,101,0,0,0,0,1,90,95,0,0,11,0,115,116,101,112,0,1,1,0,0,9,101,100,103,101,0,0,1,1,0,0,
-11,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,
-0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,90,95,0,0,12,0,115,116,101,112,0,1,1,0,0,9,101,100,103,
-101,0,0,1,1,0,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,
-18,118,0,0,18,101,100,103,101,0,0,0,0,1,90,95,0,0,9,0,115,109,111,111,116,104,115,116,101,112,0,1,
-1,0,0,9,101,100,103,101,48,0,0,1,1,0,0,9,101,100,103,101,49,0,0,1,1,0,0,9,120,0,0,0,1,3,2,90,95,0,
-0,9,1,116,0,2,58,99,108,97,109,112,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,
-18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,
-51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,10,0,115,109,111,111,116,104,115,116,
-101,112,0,1,1,0,0,10,101,100,103,101,48,0,0,1,1,0,0,10,101,100,103,101,49,0,0,1,1,0,0,10,118,0,0,0,
-1,3,2,90,95,0,0,10,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,
-103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,
-116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,11,0,115,109,111,111,116,
-104,115,116,101,112,0,1,1,0,0,11,101,100,103,101,48,0,0,1,1,0,0,11,101,100,103,101,49,0,0,1,1,0,0,
-11,118,0,0,0,1,3,2,90,95,0,0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,
-47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,
-8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,12,0,115,
-109,111,111,116,104,115,116,101,112,0,1,1,0,0,12,101,100,103,101,48,0,0,1,1,0,0,12,101,100,103,101,
-49,0,0,1,1,0,0,12,118,0,0,0,1,3,2,90,95,0,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,
-100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,
-0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,
-95,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,9,101,100,103,101,48,0,0,1,1,0,0,9,
-101,100,103,101,49,0,0,1,1,0,0,10,118,0,0,0,1,3,2,90,95,0,0,10,1,116,0,2,58,99,108,97,109,112,0,18,
-118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,
-48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,
-47,48,0,0,1,90,95,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,9,101,100,103,101,48,
-0,0,1,1,0,0,9,101,100,103,101,49,0,0,1,1,0,0,11,118,0,0,0,1,3,2,90,95,0,0,11,1,116,0,2,58,99,108,
-97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,
+0,10,0,0,109,97,120,0,1,1,0,0,10,0,97,0,0,1,1,0,0,10,0,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,
+18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0,
+1,90,95,0,0,11,0,0,109,97,120,0,1,1,0,0,11,0,97,0,0,1,1,0,0,11,0,98,0,0,0,1,4,118,101,99,52,95,109,
+97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,
+120,121,122,0,0,0,0,1,90,95,0,0,12,0,0,109,97,120,0,1,1,0,0,12,0,97,0,0,1,1,0,0,12,0,98,0,0,0,1,4,
+118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,
+0,10,0,0,109,97,120,0,1,1,0,0,10,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,
+18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,0,109,97,
+120,0,1,1,0,0,11,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,
+116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,0,109,97,120,0,1,1,0,0,
+12,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,
+0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,0,0,99,108,97,109,112,0,1,1,0,0,9,0,118,97,108,0,0,1,1,0,
+0,9,0,109,105,110,86,97,108,0,0,1,1,0,0,9,0,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,
+97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,
+109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,10,0,0,99,108,97,109,112,0,1,1,0,0,10,0,118,97,108,0,0,1,
+1,0,0,9,0,109,105,110,86,97,108,0,0,1,1,0,0,9,0,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,
+108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,
+18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,11,0,0,99,108,97,109,112,0,1,1,0,0,11,0,118,97,108,0,0,
+1,1,0,0,9,0,109,105,110,86,97,108,0,0,1,1,0,0,9,0,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,
+99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,
+0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,12,0,0,99,108,97,109,112,0,1,1,0,0,12,0,118,97,108,
+0,0,1,1,0,0,9,0,109,105,110,86,97,108,0,0,1,1,0,0,9,0,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,
+95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,
+108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,10,0,0,99,108,97,109,112,0,1,1,0,0,10,0,118,97,
+108,0,0,1,1,0,0,10,0,109,105,110,86,97,108,0,0,1,1,0,0,10,0,109,97,120,86,97,108,0,0,0,1,4,118,101,
+99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,
+86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,11,0,0,99,108,97,109,112,0,1,1,0,0,11,0,
+118,97,108,0,0,1,1,0,0,11,0,109,105,110,86,97,108,0,0,1,1,0,0,11,0,109,97,120,86,97,108,0,0,0,1,4,
+118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,
+105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,12,0,0,99,108,97,109,112,0,1,1,0,
+0,12,0,118,97,108,0,0,1,1,0,0,12,0,109,105,110,86,97,108,0,0,1,1,0,0,12,0,109,97,120,86,97,108,0,0,
+0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,
+109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,90,95,0,0,9,0,0,109,105,120,0,1,1,0,0,
+9,0,120,0,0,1,1,0,0,9,0,121,0,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,
+114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0,0,109,105,120,0,1,1,
+0,0,10,0,120,0,0,1,1,0,0,10,0,121,0,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,
+95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,11,0,0,109,105,120,
+0,1,1,0,0,11,0,120,0,0,1,1,0,0,11,0,121,0,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,108,114,112,
+0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,12,0,0,109,
+105,120,0,1,1,0,0,12,0,120,0,0,1,1,0,0,12,0,121,0,0,1,1,0,0,9,0,97,0,0,0,1,4,118,101,99,52,95,108,
+114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0,
+0,109,105,120,0,1,1,0,0,10,0,120,0,0,1,1,0,0,10,0,121,0,0,1,1,0,0,10,0,97,0,0,0,1,4,118,101,99,52,
+95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,95,0,
+0,11,0,0,109,105,120,0,1,1,0,0,11,0,120,0,0,1,1,0,0,11,0,121,0,0,1,1,0,0,11,0,97,0,0,0,1,4,118,101,
+99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,90,
+95,0,0,12,0,0,109,105,120,0,1,1,0,0,12,0,120,0,0,1,1,0,0,12,0,121,0,0,1,1,0,0,12,0,97,0,0,0,1,4,
+118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,
+0,0,1,90,95,0,0,9,0,0,115,116,101,112,0,1,1,0,0,9,0,101,100,103,101,0,0,1,1,0,0,9,0,120,0,0,0,1,4,
+118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,
+0,0,0,1,90,95,0,0,10,0,0,115,116,101,112,0,1,1,0,0,10,0,101,100,103,101,0,0,1,1,0,0,10,0,120,0,0,0,
+1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,120,0,0,18,
+101,100,103,101,0,0,0,0,1,90,95,0,0,11,0,0,115,116,101,112,0,1,1,0,0,11,0,101,100,103,101,0,0,1,1,
+0,0,11,0,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,
+122,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,90,95,0,0,12,0,0,115,116,101,112,0,1,1,0,0,12,0,
+101,100,103,101,0,0,1,1,0,0,12,0,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,
+86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,90,95,0,0,10,0,0,115,116,101,112,0,1,1,0,0,9,
+0,101,100,103,101,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,
+116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,90,95,0,0,11,0,0,115,116,
+101,112,0,1,1,0,0,9,0,101,100,103,101,0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,
+0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,90,
+95,0,0,12,0,0,115,116,101,112,0,1,1,0,0,9,0,101,100,103,101,0,0,1,1,0,0,12,0,118,0,0,0,1,4,118,101,
+99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,
+90,95,0,0,9,0,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,9,0,101,100,103,101,48,0,0,1,1,0,
+0,9,0,101,100,103,101,49,0,0,1,1,0,0,9,0,120,0,0,0,1,3,2,90,95,0,0,9,0,1,116,0,2,58,99,108,97,109,
+112,0,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,
+0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,
+116,0,48,47,48,0,0,1,90,95,0,0,10,0,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,10,0,101,
+100,103,101,48,0,0,1,1,0,0,10,0,101,100,103,101,49,0,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,0,0,10,0,
+1,116,0,2,58,99,108,97,109,112,0,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,
+101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,
+0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,11,0,0,115,109,111,111,116,104,115,116,
+101,112,0,1,1,0,0,11,0,101,100,103,101,48,0,0,1,1,0,0,11,0,101,100,103,101,49,0,0,1,1,0,0,11,0,118,
+0,0,0,1,3,2,90,95,0,0,11,0,1,116,0,2,58,99,108,97,109,112,0,0,18,118,0,18,101,100,103,101,48,0,47,
+18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,
+18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,12,0,0,115,
+109,111,111,116,104,115,116,101,112,0,1,1,0,0,12,0,101,100,103,101,48,0,0,1,1,0,0,12,0,101,100,103,
+101,49,0,0,1,1,0,0,12,0,118,0,0,0,1,3,2,90,95,0,0,12,0,1,116,0,2,58,99,108,97,109,112,0,0,18,118,0,
+18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,
+0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,
+0,1,90,95,0,0,10,0,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,9,0,101,100,103,101,48,0,0,
+1,1,0,0,9,0,101,100,103,101,49,0,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,0,0,10,0,1,116,0,2,58,99,108,
+97,109,112,0,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,
 47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,
-0,18,116,0,48,47,48,0,0,1,90,95,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,9,101,
-100,103,101,48,0,0,1,1,0,0,9,101,100,103,101,49,0,0,1,1,0,0,12,118,0,0,0,1,3,2,90,95,0,0,12,1,116,
-0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,
-103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,
-17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,9,0,108,101,110,103,116,104,0,1,1,0,0,9,120,0,0,0,
-1,8,58,97,98,115,0,18,120,0,0,0,0,0,1,90,95,0,0,9,0,108,101,110,103,116,104,0,1,1,0,0,10,118,0,0,0,
-1,3,2,90,95,0,0,9,1,114,0,0,0,3,2,90,95,1,0,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,
-0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,
-0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,0,1,90,95,0,0,9,0,108,101,110,103,116,
-104,0,1,1,0,0,11,118,0,0,0,1,3,2,90,95,0,0,9,1,114,0,0,0,3,2,90,95,1,0,9,1,112,0,2,58,100,111,116,
-0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,
-108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,90,95,0,0,9,0,
-108,101,110,103,116,104,0,1,1,0,0,12,118,0,0,0,1,3,2,90,95,0,0,9,1,114,0,0,0,3,2,90,95,1,0,9,1,112,
-0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,
-18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,
-0,1,90,95,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,0,9,120,0,0,1,1,0,0,9,121,0,0,0,1,3,2,90,
-95,1,0,9,1,100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,
-116,104,0,18,100,0,0,0,20,0,0,1,90,95,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,0,10,118,0,0,1,
-1,0,0,10,117,0,0,0,1,3,2,90,95,1,0,10,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,
-86,97,108,0,58,108,101,110,103,116,104,0,18,100,50,0,0,0,20,0,0,1,90,95,0,0,9,0,100,105,115,116,97,
-110,99,101,0,1,1,0,0,11,118,0,0,1,1,0,0,11,117,0,0,0,1,3,2,90,95,1,0,11,1,100,51,0,2,18,118,0,18,
-117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,
-0,1,90,95,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,0,12,118,0,0,1,1,0,0,12,117,0,0,0,1,3,2,90,
-95,1,0,12,1,100,52,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,
-103,116,104,0,18,100,52,0,0,0,20,0,0,1,90,95,0,0,11,0,99,114,111,115,115,0,1,1,0,0,11,118,0,0,1,1,
-0,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115,115,0,18,95,95,114,101,116,86,97,108,0,59,120,
-121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,9,0,102,97,99,101,102,111,114,119,97,114,100,0,1,
-1,0,0,9,78,0,0,1,1,0,0,9,73,0,0,1,1,0,0,9,78,114,101,102,0,0,0,1,3,2,90,95,1,0,9,1,100,0,2,58,100,
-111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,
-103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,
-115,0,0,0,0,0,1,90,95,0,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,0,10,78,0,0,1,1,0,
-0,10,73,0,0,1,1,0,0,10,78,114,101,102,0,0,0,1,3,2,90,95,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,
-101,102,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,
-0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,
-0,0,11,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,0,11,78,0,0,1,1,0,0,11,73,0,0,1,1,0,0,11,
-78,114,101,102,0,0,0,1,3,2,90,95,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,
-0,0,0,3,2,90,95,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,
-100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,0,0,12,0,102,97,99,
-101,102,111,114,119,97,114,100,0,1,1,0,0,12,78,0,0,1,1,0,0,12,73,0,0,1,1,0,0,12,78,114,101,102,0,0,
-0,1,3,2,90,95,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,
-9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,
-105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,0,0,9,0,114,101,102,108,101,99,116,0,1,1,
-0,0,9,73,0,0,1,1,0,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,
-48,18,78,0,48,47,0,0,1,90,95,0,0,10,0,114,101,102,108,101,99,116,0,1,1,0,0,10,73,0,0,1,1,0,0,10,78,
-0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,90,
-95,0,0,11,0,114,101,102,108,101,99,116,0,1,1,0,0,11,73,0,0,1,1,0,0,11,78,0,0,0,1,8,18,73,0,17,50,0,
-48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,90,95,0,0,12,0,114,101,102,
-108,101,99,116,0,1,1,0,0,12,73,0,0,1,1,0,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,
-18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,90,95,0,0,9,0,114,101,102,114,97,99,116,0,1,1,0,0,9,
-73,0,0,1,1,0,0,9,78,0,0,1,1,0,0,9,101,116,97,0,0,0,1,3,2,90,95,0,0,9,1,110,95,100,111,116,95,105,0,
-2,58,100,111,116,0,18,78,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,
-97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,
-0,48,47,48,47,0,0,3,2,90,95,0,0,9,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,
-18,114,101,116,118,97,108,0,17,48,0,48,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,
-73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,
-78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,90,95,0,0,10,0,114,101,102,114,97,99,116,0,1,1,
-0,0,10,73,0,0,1,1,0,0,10,78,0,0,1,1,0,0,9,101,116,97,0,0,0,1,3,2,90,95,0,0,9,1,110,95,100,111,116,
-95,105,0,2,58,100,111,116,0,18,78,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,1,107,0,2,17,49,0,48,0,0,18,
-101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,
-116,95,105,0,48,47,48,47,0,0,3,2,90,95,0,0,10,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,
-48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118,101,99,50,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,
-116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,
-115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,90,95,0,0,
-11,0,114,101,102,114,97,99,116,0,1,1,0,0,11,73,0,0,1,1,0,0,11,78,0,0,1,1,0,0,9,101,116,97,0,0,0,1,
-3,2,90,95,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78,0,0,18,73,0,0,0,0,0,3,2,90,
-95,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,
-111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,90,95,0,0,11,1,114,101,116,118,
-97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118,101,99,51,0,17,
-48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,
-110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,
-101,116,118,97,108,0,0,0,1,90,95,0,0,12,0,114,101,102,114,97,99,116,0,1,1,0,0,12,73,0,0,1,1,0,0,12,
-78,0,0,1,1,0,0,9,101,116,97,0,0,0,1,3,2,90,95,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,
-0,18,78,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,
+0,18,116,0,48,47,48,0,0,1,90,95,0,0,11,0,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,0,9,0,
+101,100,103,101,48,0,0,1,1,0,0,9,0,101,100,103,101,49,0,0,1,1,0,0,11,0,118,0,0,0,1,3,2,90,95,0,0,
+11,0,1,116,0,2,58,99,108,97,109,112,0,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,
+0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,
+17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,12,0,0,115,109,111,111,116,104,115,
+116,101,112,0,1,1,0,0,9,0,101,100,103,101,48,0,0,1,1,0,0,9,0,101,100,103,101,49,0,0,1,1,0,0,12,0,
+118,0,0,0,1,3,2,90,95,0,0,12,0,1,116,0,2,58,99,108,97,109,112,0,0,18,118,0,18,101,100,103,101,48,0,
+47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,
+8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,90,95,0,0,9,0,0,108,
+101,110,103,116,104,0,1,1,0,0,9,0,120,0,0,0,1,8,58,97,98,115,0,0,18,120,0,0,0,0,0,1,90,95,0,0,9,0,
+0,108,101,110,103,116,104,0,1,1,0,0,10,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,114,0,0,0,3,2,90,95,1,0,9,
+0,1,112,0,2,58,100,111,116,0,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,
+18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,
+120,0,0,18,114,0,0,0,0,1,90,95,0,0,9,0,0,108,101,110,103,116,104,0,1,1,0,0,11,0,118,0,0,0,1,3,2,90,
+95,0,0,9,0,1,114,0,0,0,3,2,90,95,1,0,9,0,1,112,0,2,58,100,111,116,0,0,18,118,0,0,18,118,0,0,0,0,0,
+4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,
+18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,90,95,0,0,9,0,0,108,101,110,103,116,104,0,1,1,
+0,0,12,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,114,0,0,0,3,2,90,95,1,0,9,0,1,112,0,2,58,100,111,116,0,0,
+18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,
+108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,90,95,0,0,9,0,0,
+100,105,115,116,97,110,99,101,0,1,1,0,0,9,0,120,0,0,1,1,0,0,9,0,121,0,0,0,1,3,2,90,95,1,0,9,0,1,
+100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,0,
+18,100,0,0,0,20,0,0,1,90,95,0,0,9,0,0,100,105,115,116,97,110,99,101,0,1,1,0,0,10,0,118,0,0,1,1,0,0,
+10,0,117,0,0,0,1,3,2,90,95,1,0,10,0,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,
+86,97,108,0,58,108,101,110,103,116,104,0,0,18,100,50,0,0,0,20,0,0,1,90,95,0,0,9,0,0,100,105,115,
+116,97,110,99,101,0,1,1,0,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,0,1,3,2,90,95,1,0,11,0,1,100,51,0,2,
+18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,0,18,100,
+51,0,0,0,20,0,0,1,90,95,0,0,9,0,0,100,105,115,116,97,110,99,101,0,1,1,0,0,12,0,118,0,0,1,1,0,0,12,
+0,117,0,0,0,1,3,2,90,95,1,0,12,0,1,100,52,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,
+97,108,0,58,108,101,110,103,116,104,0,0,18,100,52,0,0,0,20,0,0,1,90,95,0,0,11,0,0,99,114,111,115,
+115,0,1,1,0,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115,115,0,18,95,
+95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,9,0,0,102,97,
+99,101,102,111,114,119,97,114,100,0,1,1,0,0,9,0,78,0,0,1,1,0,0,9,0,73,0,0,1,1,0,0,9,0,78,114,101,
+102,0,0,0,1,3,2,90,95,1,0,9,0,1,100,0,2,58,100,111,116,0,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,
+2,90,95,0,0,9,0,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,
+0,0,8,58,109,105,120,0,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,0,0,10,0,0,102,97,99,101,
+102,111,114,119,97,114,100,0,1,1,0,0,10,0,78,0,0,1,1,0,0,10,0,73,0,0,1,1,0,0,10,0,78,114,101,102,0,
+0,0,1,3,2,90,95,1,0,9,0,1,100,0,2,58,100,111,116,0,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,90,
+95,0,0,9,0,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,
+58,109,105,120,0,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,0,0,11,0,0,102,97,99,101,102,
+111,114,119,97,114,100,0,1,1,0,0,11,0,78,0,0,1,1,0,0,11,0,73,0,0,1,1,0,0,11,0,78,114,101,102,0,0,0,
+1,3,2,90,95,1,0,9,0,1,100,0,2,58,100,111,116,0,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,90,95,0,
+0,9,0,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,
+109,105,120,0,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,0,0,12,0,0,102,97,99,101,102,111,
+114,119,97,114,100,0,1,1,0,0,12,0,78,0,0,1,1,0,0,12,0,73,0,0,1,1,0,0,12,0,78,114,101,102,0,0,0,1,3,
+2,90,95,1,0,9,0,1,100,0,2,58,100,111,116,0,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,
+0,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,
+105,120,0,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,90,95,0,0,9,0,0,114,101,102,108,101,99,116,0,
+1,1,0,0,9,0,73,0,0,1,1,0,0,9,0,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,0,18,78,0,0,18,
+73,0,0,0,48,18,78,0,48,47,0,0,1,90,95,0,0,10,0,0,114,101,102,108,101,99,116,0,1,1,0,0,10,0,73,0,0,
+1,1,0,0,10,0,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,0,18,78,0,0,18,73,0,0,0,48,18,78,
+0,48,47,0,0,1,90,95,0,0,11,0,0,114,101,102,108,101,99,116,0,1,1,0,0,11,0,73,0,0,1,1,0,0,11,0,78,0,
+0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,90,
+95,0,0,12,0,0,114,101,102,108,101,99,116,0,1,1,0,0,12,0,73,0,0,1,1,0,0,12,0,78,0,0,0,1,8,18,73,0,
+17,50,0,48,0,0,58,100,111,116,0,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,90,95,0,0,9,0,0,114,
+101,102,114,97,99,116,0,1,1,0,0,9,0,73,0,0,1,1,0,0,9,0,78,0,0,1,1,0,0,9,0,101,116,97,0,0,0,1,3,2,
+90,95,0,0,9,0,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,0,18,78,0,0,18,73,0,0,0,0,0,3,2,90,
+95,0,0,9,0,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95,
+100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,90,95,0,0,9,0,1,114,101,
+116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,17,48,0,48,0,0,
+20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,
+116,95,105,0,48,58,115,113,114,116,0,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,
+108,0,0,0,1,90,95,0,0,10,0,0,114,101,102,114,97,99,116,0,1,1,0,0,10,0,73,0,0,1,1,0,0,10,0,78,0,0,1,
+1,0,0,9,0,101,116,97,0,0,0,1,3,2,90,95,0,0,9,0,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,0,
+18,78,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,0,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,
 0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,
-3,2,90,95,0,0,12,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,
-118,97,108,0,58,118,101,99,52,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,
-97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,
-0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,90,95,0,0,13,0,109,97,116,114,105,120,
-67,111,109,112,77,117,108,116,0,1,0,0,0,13,109,0,0,1,0,0,0,13,110,0,0,0,1,8,58,109,97,116,50,0,18,
-109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,
-0,0,1,90,95,0,0,14,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,14,109,0,0,1,0,
-0,0,14,110,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,
-16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,
-1,90,95,0,0,15,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,15,109,0,0,1,0,0,0,
-15,110,0,0,0,1,8,58,109,97,116,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,
-10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,
-16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,90,95,0,0,2,0,108,101,115,115,84,104,97,110,0,1,
-1,0,0,10,117,0,0,1,1,0,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,
-97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,108,101,115,115,84,104,97,110,0,
-1,1,0,0,11,117,0,0,1,1,0,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,
-97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,108,101,115,115,84,104,97,
-110,0,1,1,0,0,12,117,0,0,1,1,0,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,
-116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,
-0,6,117,0,0,1,1,0,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,
-0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0,0,
-7,117,0,0,1,1,0,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,
-59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,
-0,8,117,0,0,1,1,0,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,
-0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,
-1,0,0,10,117,0,0,1,1,0,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,
-97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,108,101,115,115,84,104,97,110,69,
-113,117,97,108,0,1,1,0,0,11,117,0,0,1,1,0,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,
-95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,108,101,
-115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,0,12,117,0,0,1,1,0,0,12,118,0,0,0,1,4,118,101,99,
-52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,
-108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,0,6,117,0,0,1,1,0,0,6,118,0,0,0,1,4,118,
+3,2,90,95,0,0,10,0,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,
+118,97,108,0,58,118,101,99,50,0,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,
+97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,0,18,107,0,
+0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,90,95,0,0,11,0,0,114,101,102,114,97,
+99,116,0,1,1,0,0,11,0,73,0,0,1,1,0,0,11,0,78,0,0,1,1,0,0,9,0,101,116,97,0,0,0,1,3,2,90,95,0,0,9,0,
+1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,0,18,78,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,0,1,
+107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,
+105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,90,95,0,0,11,0,1,114,101,116,118,97,108,0,
+0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118,101,99,51,0,0,17,48,0,48,
+0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,
+100,111,116,95,105,0,48,58,115,113,114,116,0,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,
+118,97,108,0,0,0,1,90,95,0,0,12,0,0,114,101,102,114,97,99,116,0,1,1,0,0,12,0,73,0,0,1,1,0,0,12,0,
+78,0,0,1,1,0,0,9,0,101,116,97,0,0,0,1,3,2,90,95,0,0,9,0,1,110,95,100,111,116,95,105,0,2,58,100,111,
+116,0,0,18,78,0,0,18,73,0,0,0,0,0,3,2,90,95,0,0,9,0,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,
+101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,
+48,47,0,0,3,2,90,95,0,0,12,0,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,
+114,101,116,118,97,108,0,58,118,101,99,52,0,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,
+0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,
+0,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,90,95,0,0,13,0,0,109,97,
+116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,0,13,0,109,0,0,1,0,0,0,13,0,110,0,0,0,1,8,58,
+109,97,116,50,0,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,
+16,10,49,0,57,48,0,0,0,0,1,90,95,0,0,14,0,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,
+1,0,0,0,14,0,109,0,0,1,0,0,0,14,0,110,0,0,0,1,8,58,109,97,116,51,0,0,18,109,0,16,8,48,0,57,18,110,
+0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,
+110,0,16,10,50,0,57,48,0,0,0,0,1,90,95,0,0,15,0,0,109,97,116,114,105,120,67,111,109,112,77,117,108,
+116,0,1,0,0,0,15,0,109,0,0,1,0,0,0,15,0,110,0,0,0,1,8,58,109,97,116,52,0,0,18,109,0,16,8,48,0,57,
+18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,
+57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,90,95,0,
+0,2,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,10,0,117,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,
+52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,
+95,0,0,3,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,11,0,117,0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,
+101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,
+0,0,0,1,90,95,0,0,4,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,
+0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,
+1,90,95,0,0,2,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,118,0,0,0,1,4,
+118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,
+0,0,0,1,90,95,0,0,3,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,7,0,117,0,0,1,1,0,0,7,0,118,0,0,0,
+1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,
+18,118,0,0,0,0,1,90,95,0,0,4,0,0,108,101,115,115,84,104,97,110,0,1,1,0,0,8,0,117,0,0,1,1,0,0,8,0,
+118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,
+0,0,0,0,1,90,95,0,0,2,0,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,0,10,0,117,0,0,1,
+1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,
+121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,0,108,101,115,115,84,104,97,110,69,113,117,97,
+108,0,1,1,0,0,11,0,117,0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,
+101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,108,101,115,115,
+84,104,97,110,69,113,117,97,108,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,0,1,4,118,101,99,52,95,
+115,108,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,108,
+101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,118,0,0,0,1,4,118,
 101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,
-0,1,90,95,0,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,0,7,117,0,0,1,1,0,0,7,
-118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,
-18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,
-0,8,117,0,0,1,1,0,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,
-0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,0,
-10,117,0,0,1,1,0,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,
-0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,103,114,101,97,116,101,114,84,104,97,
-110,0,1,1,0,0,11,117,0,0,1,1,0,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,
-116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,103,114,101,97,116,
-101,114,84,104,97,110,0,1,1,0,0,12,117,0,0,1,1,0,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,
-18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,103,114,101,97,116,
-101,114,84,104,97,110,0,1,1,0,0,6,117,0,0,1,1,0,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,
-18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,
-0,1,90,95,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,0,7,117,0,0,1,1,0,0,7,118,0,0,0,
-1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,
-18,118,0,0,0,0,1,90,95,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,0,8,117,0,0,1,1,0,
-0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,
-118,0,0,0,0,1,90,95,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,0,
-10,117,0,0,1,1,0,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,
-0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,103,114,101,97,116,101,114,84,104,97,
-110,69,113,117,97,108,0,1,1,0,0,11,117,0,0,1,1,0,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,
-18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,103,
-114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,0,12,117,0,0,1,1,0,0,12,118,0,0,0,1,
-4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,
-95,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,0,6,117,0,0,1,1,0,0,
-6,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,
-117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,
-0,1,1,0,0,7,117,0,0,1,1,0,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,
-97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,103,114,101,97,116,101,114,
-84,104,97,110,69,113,117,97,108,0,1,1,0,0,8,117,0,0,1,1,0,0,8,118,0,0,0,1,4,118,101,99,52,95,115,
-103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,101,113,117,
-97,108,0,1,1,0,0,10,117,0,0,1,1,0,0,10,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,
-101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,101,113,117,97,108,0,
-1,1,0,0,11,117,0,0,1,1,0,0,11,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,
-97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,101,113,117,97,108,0,1,1,0,0,
-12,117,0,0,1,1,0,0,12,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,
-0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,101,113,117,97,108,0,1,1,0,0,6,117,0,0,1,1,0,0,6,118,
-0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,
-0,18,118,0,0,0,0,1,90,95,0,0,3,0,101,113,117,97,108,0,1,1,0,0,7,117,0,0,1,1,0,0,7,118,0,0,0,1,4,
-118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,
-118,0,0,0,0,1,90,95,0,0,4,0,101,113,117,97,108,0,1,1,0,0,8,117,0,0,1,1,0,0,8,118,0,0,0,1,4,118,101,
-99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,
-0,101,113,117,97,108,0,1,1,0,0,2,117,0,0,1,1,0,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,
-95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,101,113,117,
-97,108,0,1,1,0,0,3,117,0,0,1,1,0,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,
-116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,101,113,117,97,108,0,
-1,1,0,0,4,117,0,0,1,1,0,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,
-97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,0,10,
-117,0,0,1,1,0,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,
-59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,0,
-11,117,0,0,1,1,0,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,
-0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,
-0,0,12,117,0,0,1,1,0,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,
-108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,0,6,117,0,
-0,1,1,0,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,
-121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,0,7,117,0,
-0,1,1,0,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,
-121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,0,8,
-117,0,0,1,1,0,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,
-18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,0,2,117,0,0,1,1,0,
-0,2,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,
-18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,0,3,117,0,0,1,1,0,
-0,3,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,
-0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,0,4,117,0,0,1,1,
-0,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,
-18,118,0,0,0,0,1,90,95,0,0,1,0,97,110,121,0,1,1,0,0,2,118,0,0,0,1,3,2,90,95,0,0,9,1,115,117,109,0,
-0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,
-121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,
-109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,97,110,121,0,1,1,0,0,3,118,0,0,0,1,3,2,90,95,
-0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,
+0,1,90,95,0,0,3,0,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,0,7,0,117,0,0,1,1,0,0,
+7,0,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,
+0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,
+1,0,0,8,0,117,0,0,1,1,0,0,8,0,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,
+97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,103,114,101,97,116,101,114,84,104,97,110,0,
+1,1,0,0,10,0,117,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,
+116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,0,103,114,101,97,116,101,
+114,84,104,97,110,0,1,1,0,0,11,0,117,0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,
+18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,
+103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,0,1,4,118,101,
+99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,
+0,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,118,0,0,0,1,4,118,
+101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,59,120,121,0,0,
+18,118,0,59,120,121,0,0,0,0,1,90,95,0,0,3,0,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,0,7,
+0,117,0,0,1,1,0,0,7,0,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,
+0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,103,114,101,97,116,101,114,84,104,
+97,110,0,1,1,0,0,8,0,117,0,0,1,1,0,0,8,0,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,
+101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,103,114,101,97,116,101,114,84,
+104,97,110,69,113,117,97,108,0,1,1,0,0,10,0,117,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,
+115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,
+0,3,0,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,0,11,0,117,0,0,1,1,0,0,
+11,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,
+0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,103,114,101,97,116,101,114,84,104,97,110,69,113,
+117,97,108,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,
+95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,103,114,101,97,116,101,
+114,84,104,97,110,69,113,117,97,108,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,118,0,0,0,1,4,118,101,99,52,
+95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,
+0,0,3,0,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,0,7,0,117,0,0,1,1,0,0,
+7,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,
+0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,
+97,108,0,1,1,0,0,8,0,117,0,0,1,1,0,0,8,0,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,
+101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,101,113,117,97,108,0,1,1,0,0,10,
+0,117,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,
+0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,0,101,113,117,97,108,0,1,1,0,0,11,0,117,
+0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,
+120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,101,113,117,97,108,0,1,1,0,0,12,0,117,
+0,0,1,1,0,0,12,0,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,
+18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,101,113,117,97,108,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,
+118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,
+117,0,0,18,118,0,0,0,0,1,90,95,0,0,3,0,0,101,113,117,97,108,0,1,1,0,0,7,0,117,0,0,1,1,0,0,7,0,118,
+0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,
+117,0,0,18,118,0,0,0,0,1,90,95,0,0,4,0,0,101,113,117,97,108,0,1,1,0,0,8,0,117,0,0,1,1,0,0,8,0,118,
+0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,
+0,0,1,90,95,0,0,2,0,0,101,113,117,97,108,0,1,1,0,0,2,0,117,0,0,1,1,0,0,2,0,118,0,0,0,1,4,118,101,
+99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,
+90,95,0,0,3,0,0,101,113,117,97,108,0,1,1,0,0,3,0,117,0,0,1,1,0,0,3,0,118,0,0,0,1,4,118,101,99,52,
+95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,
+90,95,0,0,4,0,0,101,113,117,97,108,0,1,1,0,0,4,0,117,0,0,1,1,0,0,4,0,118,0,0,0,1,4,118,101,99,52,
+95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,0,2,0,0,
+110,111,116,69,113,117,97,108,0,1,1,0,0,10,0,117,0,0,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,
+115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,90,95,0,
+0,3,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,11,0,117,0,0,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,
+52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,
+1,90,95,0,0,4,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,12,0,117,0,0,1,1,0,0,12,0,118,0,0,0,1,4,
+118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,90,
+95,0,0,2,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,6,0,117,0,0,1,1,0,0,6,0,118,0,0,0,1,4,118,101,
+99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,
+90,95,0,0,3,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,7,0,117,0,0,1,1,0,0,7,0,118,0,0,0,1,4,118,
+101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,
+0,0,0,1,90,95,0,0,4,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,8,0,117,0,0,1,1,0,0,8,0,118,0,0,0,
+1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,
+90,95,0,0,2,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,2,0,117,0,0,1,1,0,0,2,0,118,0,0,0,1,4,118,
+101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,
+0,1,90,95,0,0,3,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,3,0,117,0,0,1,1,0,0,3,0,118,0,0,0,1,4,
+118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,
+118,0,0,0,0,1,90,95,0,0,4,0,0,110,111,116,69,113,117,97,108,0,1,1,0,0,4,0,117,0,0,1,1,0,0,4,0,118,
+0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,
+0,0,1,90,95,0,0,1,0,0,97,110,121,0,1,1,0,0,2,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,115,117,109,0,0,0,4,
+118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,
+0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,
+120,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,0,97,110,121,0,1,1,0,0,3,0,118,0,0,0,1,3,2,90,95,0,0,
+9,0,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,
 120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,
 117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,
-86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,97,110,121,
-0,1,1,0,0,4,118,0,0,0,1,3,2,90,95,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,
-117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,
-115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97,
-100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,
-99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,
-48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,97,108,108,0,1,1,0,0,2,118,0,0,0,1,3,2,90,95,0,0,9,1,112,114,111,
-100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,
-120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,
-18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,97,108,108,0,1,1,0,0,3,118,0,0,0,1,3,2,
-90,95,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,
-114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,
-112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,
+86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,0,97,110,
+121,0,1,1,0,0,4,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,
+0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,
+100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,
+52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,
+118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,
+0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,0,97,108,108,0,1,1,0,0,2,0,118,0,0,0,1,3,2,90,95,0,0,9,0,
+1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,
+0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,
+86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,0,97,108,108,0,1,1,0,0,3,
+0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,
+108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,
+117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,
+4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,
+48,0,0,0,0,0,1,90,95,0,0,1,0,0,97,108,108,0,1,1,0,0,4,0,118,0,0,0,1,3,2,90,95,0,0,9,0,1,112,114,
+111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,
+59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,
+111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,
+112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,
 95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,
-90,95,0,0,1,0,97,108,108,0,1,1,0,0,4,118,0,0,0,1,3,2,90,95,0,0,9,1,112,114,111,100,0,0,0,4,118,101,
-99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,
-121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,
-111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,
-111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,
-114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,2,0,110,111,116,
-0,1,1,0,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,
-121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,110,111,116,0,1,1,0,0,3,118,0,0,0,1,4,118,
-101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,
-48,0,0,0,0,0,1,90,95,0,0,4,0,110,111,116,0,1,1,0,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,
-18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,12,0,116,101,120,
-116,117,114,101,49,68,0,1,1,0,0,16,115,97,109,112,108,101,114,0,0,1,1,0,0,9,99,111,111,114,100,0,0,
-0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,
-108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,49,68,
-80,114,111,106,0,1,1,0,0,16,115,97,109,112,108,101,114,0,0,1,1,0,0,10,99,111,111,114,100,0,0,0,1,4,
-118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,
-101,114,0,0,18,99,111,111,114,100,0,59,120,121,121,121,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,
-117,114,101,49,68,80,114,111,106,0,1,1,0,0,16,115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,111,111,
-114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,
-115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,
-114,101,50,68,0,1,1,0,0,17,115,97,109,112,108,101,114,0,0,1,1,0,0,10,99,111,111,114,100,0,0,0,1,4,
-118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,
-114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,
-111,106,0,1,1,0,0,17,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,0,1,4,118,
-101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,
-114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,
-114,101,50,68,80,114,111,106,0,1,1,0,0,17,115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,111,111,114,
-100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,
-97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,
-101,51,68,0,1,1,0,0,18,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,0,1,4,118,
-101,99,52,95,116,101,120,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,
-0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111,
-106,0,1,1,0,0,18,115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,
-52,95,116,101,120,112,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,
-0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,67,117,98,101,0,1,1,0,
-0,19,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,
-101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,
-111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,115,104,97,100,111,119,49,68,0,1,1,0,0,20,115,97,109,112,
-108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,
+90,95,0,0,2,0,0,110,111,116,0,1,1,0,0,2,0,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,
+114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,0,110,111,
+116,0,1,1,0,0,3,0,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,
+120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,4,0,0,110,111,116,0,1,1,0,0,4,0,118,0,
+0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,
+0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,0,1,1,0,0,16,0,115,97,109,112,108,101,
+114,0,0,1,1,0,0,9,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,95,95,
+114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,
+0,0,12,0,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,0,16,0,115,97,109,112,108,101,
+114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,
+95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,
+121,121,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,0,16,0,
+115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,
+120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,
+111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,0,1,1,0,0,17,0,115,97,109,
+112,108,101,114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50,100,
+0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,
+0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,0,17,0,115,97,109,
+112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,
+100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,
+59,120,121,122,122,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,
+1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,
+95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,
+18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,51,68,0,1,1,0,0,18,0,
+115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,
+120,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,
+114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,51,68,80,114,111,106,0,1,1,0,0,18,0,
+115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,
+120,112,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,
+111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,67,117,98,101,0,1,1,0,0,19,0,
+115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,
+120,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,
+111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,49,68,0,1,1,0,0,20,0,115,97,109,112,
+108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,
 95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,
-90,95,0,0,12,0,115,104,97,100,111,119,49,68,80,114,111,106,0,1,1,0,0,20,115,97,109,112,108,101,114,
-0,0,1,1,0,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,
-101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,
-12,0,115,104,97,100,111,119,50,68,0,1,1,0,0,21,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,
-111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,
-115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,115,104,97,100,111,
-119,50,68,80,114,111,106,0,1,1,0,0,21,115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,111,111,114,100,
-0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,
-109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,
-50,68,82,101,99,116,0,1,1,0,0,22,115,97,109,112,108,101,114,0,0,1,1,0,0,10,99,111,111,114,100,0,0,
-0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,
-97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,
-101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,0,22,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,
-111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,
-116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,
-0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,0,22,
-115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,
-120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,
-18,99,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,0,1,1,
-0,0,23,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,
-101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,
-18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,80,114,
-111,106,0,1,1,0,0,23,115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,111,111,114,100,0,0,0,1,4,118,
-101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,
-112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,9,0,110,111,105,115,101,49,0,1,1,0,0,
-9,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,49,0,18,95,95,114,101,116,86,97,108,0,0,
-18,120,0,0,0,0,1,90,95,0,0,9,0,110,111,105,115,101,49,0,1,1,0,0,10,120,0,0,0,1,4,102,108,111,97,
-116,95,110,111,105,115,101,50,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,9,0,
-110,111,105,115,101,49,0,1,1,0,0,11,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,51,0,
-18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,9,0,110,111,105,115,101,49,0,1,1,0,0,
-12,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,52,0,18,95,95,114,101,116,86,97,108,0,0,
-18,120,0,0,0,0,1,90,95,0,0,10,0,110,111,105,115,101,50,0,1,1,0,0,9,120,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,
-97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,0,1,90,95,
-0,0,10,0,110,111,105,115,101,50,0,1,1,0,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,
-0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,
-111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,
-0,20,0,0,1,90,95,0,0,10,0,110,111,105,115,101,50,0,1,1,0,0,11,120,0,0,0,1,9,18,95,95,114,101,116,
-86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,
-108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,
-55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,10,0,110,111,105,115,101,50,0,1,1,
-0,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,
-0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,
-101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,
-46,0,0,20,0,0,1,90,95,0,0,11,0,110,111,105,115,101,51,0,1,1,0,0,9,120,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,
-97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,
-0,0,1,90,95,0,0,11,0,110,111,105,115,101,51,0,1,1,0,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,
-121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,
-0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,
-0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,11,0,
-110,111,105,115,101,51,0,1,1,0,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,
-111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,
-115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,
-0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,
-0,58,118,101,99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,
-0,0,1,90,95,0,0,11,0,110,111,105,115,101,51,0,1,1,0,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,
-121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,
-0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,
-122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,
-0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,12,0,110,111,105,
-115,101,52,0,1,1,0,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,
-101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,
-18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,
-105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,
-0,58,110,111,105,115,101,49,0,18,120,0,17,50,51,0,53,52,0,0,46,0,0,20,0,0,1,90,95,0,0,12,0,110,111,
-105,115,101,52,0,1,1,0,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,
-115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,
-49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,
-95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53,
-0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,
-110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,
-0,46,0,0,20,0,0,1,90,95,0,0,12,0,110,111,105,115,101,52,0,1,1,0,0,11,120,0,0,0,1,9,18,95,95,114,
-101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,
-86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,
-0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,
-0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,
-0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,
-115,101,49,0,18,120,0,58,118,101,99,51,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,
-57,49,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,12,0,110,111,105,115,101,52,0,1,1,0,0,12,120,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,
-0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,53,0,
-52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20,0,9,
-18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,
-17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,17,51,55,0,52,56,0,0,0,0,46,0,
-0,20,0,0,0
+90,95,0,0,12,0,0,115,104,97,100,111,119,49,68,80,114,111,106,0,1,1,0,0,20,0,115,97,109,112,108,101,
+114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,
+95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,
+95,0,0,12,0,0,115,104,97,100,111,119,50,68,0,1,1,0,0,21,0,115,97,109,112,108,101,114,0,0,1,1,0,0,
+11,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,
+108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,115,104,
+97,100,111,119,50,68,80,114,111,106,0,1,1,0,0,21,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,
+111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,
+0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,
+116,117,114,101,50,68,82,101,99,116,0,1,1,0,0,22,0,115,97,109,112,108,101,114,0,0,1,1,0,0,10,0,99,
+111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,
+97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,
+101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,0,22,0,115,97,109,112,108,101,
+114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,
+116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,
+59,120,121,122,122,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,
+114,111,106,0,1,1,0,0,22,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,0,1,
+4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,
+97,109,112,108,101,114,0,0,18,99,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,
+119,50,68,82,101,99,116,0,1,1,0,0,23,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,
+100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,
+18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,
+111,119,50,68,82,101,99,116,80,114,111,106,0,1,1,0,0,23,0,115,97,109,112,108,101,114,0,0,1,1,0,0,
+12,0,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,
+114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,90,95,
+0,0,9,0,0,110,111,105,115,101,49,0,1,1,0,0,9,0,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,
+101,49,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,9,0,0,110,111,105,115,101,
+49,0,1,1,0,0,10,0,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,50,0,18,95,95,114,101,
+116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,9,0,0,110,111,105,115,101,49,0,1,1,0,0,11,0,120,0,0,0,
+1,4,102,108,111,97,116,95,110,111,105,115,101,51,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,
+0,1,90,95,0,0,9,0,0,110,111,105,115,101,49,0,1,1,0,0,12,0,120,0,0,0,1,4,102,108,111,97,116,95,110,
+111,105,115,101,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,10,0,0,110,111,
+105,115,101,50,0,1,1,0,0,9,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,
+105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,
+115,101,49,0,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,0,1,90,95,0,0,10,0,0,110,111,105,115,101,
+50,0,1,1,0,0,10,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,
+49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,
+18,120,0,58,118,101,99,50,0,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,0,1,90,95,0,
+0,10,0,0,110,111,105,115,101,50,0,1,1,0,0,11,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,
+120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,
+58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,51,0,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,
+0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,10,0,0,110,111,105,115,101,50,0,1,1,0,0,12,0,
+120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,
+20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,
+99,52,0,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,
+0,0,20,0,0,1,90,95,0,0,11,0,0,110,111,105,115,101,51,0,1,1,0,0,9,0,120,0,0,0,1,9,18,95,95,114,101,
+116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,
+97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,
+95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,0,18,120,0,17,53,0,52,55,0,0,46,0,
+0,20,0,0,1,90,95,0,0,11,0,0,110,111,105,115,101,51,0,1,1,0,0,10,0,120,0,0,0,1,9,18,95,95,114,101,
+116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,
+97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,50,0,0,17,49,57,0,51,52,0,0,
+0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,
+101,49,0,0,18,120,0,58,118,101,99,50,0,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,
+0,1,90,95,0,0,11,0,0,110,111,105,115,101,51,0,1,1,0,0,11,0,120,0,0,0,1,9,18,95,95,114,101,116,86,
+97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,
+0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,51,0,0,17,49,57,0,51,52,0,0,0,17,
+55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,
+110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,51,0,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,
+0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,11,0,0,110,111,105,115,101,51,0,1,1,0,0,12,0,
+120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,
+20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,
+99,52,0,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,
+0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,
+101,99,52,0,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,
+0,0,0,46,0,0,20,0,0,1,90,95,0,0,12,0,0,110,111,105,115,101,52,0,1,1,0,0,9,0,120,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,
+101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,
+0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,0,18,120,0,17,53,0,52,55,
+0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,0,18,120,0,
+17,50,51,0,53,52,0,0,46,0,0,20,0,0,1,90,95,0,0,12,0,0,110,111,105,115,101,52,0,1,1,0,0,10,0,120,0,
+0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,
+9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,
+50,0,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,
+59,122,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,50,0,0,17,53,0,52,55,0,0,0,17,49,55,
+0,56,53,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,
+0,0,18,120,0,58,118,101,99,50,0,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0,1,
+90,95,0,0,12,0,0,110,111,105,115,101,52,0,1,1,0,0,11,0,120,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,
+59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,51,0,0,17,49,57,0,51,52,0,0,0,17,55,
+0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,
+111,105,115,101,49,0,0,18,120,0,58,118,101,99,51,0,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,
+49,49,0,48,52,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,
+101,49,0,0,18,120,0,58,118,101,99,51,0,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,
+57,49,0,0,0,0,46,0,0,20,0,0,1,90,95,0,0,12,0,0,110,111,105,115,101,52,0,1,1,0,0,12,0,120,0,0,0,1,9,
+18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,0,18,120,0,0,0,20,0,9,18,95,
+95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,52,0,0,17,
+49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18,
+95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,99,52,0,0,
+17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,
+20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,0,18,120,0,58,118,101,
+99,52,0,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,17,51,55,0,52,56,0,
+0,0,0,46,0,0,20,0,0,0
diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc
index 10a6bb5cb1..0a0d15903b 100644
--- a/src/mesa/shader/slang/library/slang_core.gc
+++ b/src/mesa/shader/slang/library/slang_core.gc
@@ -1207,370 +1207,381 @@ float dot(const vec4 a, const vec4 b)
 
 //// int assignment operators
 
-void __operator += (inout int a, const int b)
+int __operator += (inout int a, const int b)
 {
-   __asm vec4_add a, a, b;
+   a = a + b;
+   return a;
 }
 
-void __operator -= (inout int a, const int b)
+int __operator -= (inout int a, const int b)
 {
-   __asm vec4_subtract a, a, b;
+   a = a - b;
+   return a;
 }
 
-void __operator *= (inout int a, const int b)
+int __operator *= (inout int a, const int b)
 {
-   __asm vec4_multiply a, a, b;
+   a = a * b;
+   return a;
 }
 
-void __operator /= (inout int a, const int b)
+int __operator /= (inout int a, const int b)
 {
-   float invB;
-   __asm float_rcp invB, b;
-   __asm vec4_multiply a, a, invB;
-   __asm vec4_to_ivec4 a, a;
+   a = a / b;
+   return a;
 }
 
 
 //// ivec2 assignment operators
 
-void __operator += (inout ivec2 v, const ivec2 u)
+ivec2 __operator += (inout ivec2 v, const ivec2 u)
 {
-   __asm vec4_add v, v, u;
+   v = v + u;
+   return v;
 }
 
-void __operator -= (inout ivec2 v, const ivec2 u)
+ivec2 __operator -= (inout ivec2 v, const ivec2 u)
 {
-   __asm vec4_subtract v, v, u;
+   v = v - u;
+   return v;
 }
 
-void __operator *= (inout ivec2 v, const ivec2 u)
+ivec2 __operator *= (inout ivec2 v, const ivec2 u)
 {
-   __asm vec4_multiply v, v, u;
+   v = v * u;
+   return v;
 }
 
-void __operator /= (inout ivec2 v, const ivec2 u)
+ivec2 __operator /= (inout ivec2 v, const ivec2 u)
 {
-   ivec2 inv, z;
-   __asm float_rcp inv.x, u.x;
-   __asm float_rcp inv.y, u.y;
-   __asm vec4_multiply z, v, inv;
-   __asm vec4_to_ivec4 v, z;
+   v = v / u;
+   return v;
 }
 
 
 //// ivec3 assignment operators
 
-void __operator += (inout ivec3 v, const ivec3 u)
+ivec3 __operator += (inout ivec3 v, const ivec3 u)
 {
-   __asm vec4_add v, v, u;
+   v = v + u;
+   return v;
 }
 
-void __operator -= (inout ivec3 v, const ivec3 u)
+ivec3 __operator -= (inout ivec3 v, const ivec3 u)
 {
-   __asm vec4_subtract v, v, u;
+   v = v - u;
+   return v;
 }
 
-void __operator *= (inout ivec3 v, const ivec3 u)
+ivec3 __operator *= (inout ivec3 v, const ivec3 u)
 {
-   __asm vec4_multiply v, v, u;
+   v = v * u;
+   return v;
 }
 
-void __operator /= (inout ivec3 v, const ivec3 u)
+ivec3 __operator /= (inout ivec3 v, const ivec3 u)
 {
-   ivec3 inv, z;
-   __asm float_rcp inv.x, u.x;
-   __asm float_rcp inv.y, u.y;
-   __asm vec4_multiply z, v, inv;
-   __asm vec4_to_ivec4 v, z;
+   v = v / u;
+   return v;
 }
 
 
 //// ivec4 assignment operators
 
-void __operator += (inout ivec4 v, const ivec4 u)
+ivec4 __operator += (inout ivec4 v, const ivec4 u)
 {
-   __asm vec4_add v, v, u;
+   v = v + u;
+   return v;
 }
 
-void __operator -= (inout ivec4 v, const ivec4 u)
+ivec4 __operator -= (inout ivec4 v, const ivec4 u)
 {
-   __asm vec4_subtract v, v, u;
+   v = v - u;
+   return v;
 }
 
-void __operator *= (inout ivec4 v, const ivec4 u)
+ivec4 __operator *= (inout ivec4 v, const ivec4 u)
 {
-   __asm vec4_multiply v, v, u;
+   v = v * u;
+   return v;
 }
 
-void __operator /= (inout ivec4 v, const ivec4 u)
+ivec4 __operator /= (inout ivec4 v, const ivec4 u)
 {
-   ivec4 inv, z;
-   __asm float_rcp inv.x, u.x;
-   __asm float_rcp inv.y, u.y;
-   __asm vec4_multiply z, v, inv;
-   __asm vec4_to_ivec4 v, z;
+   v = v / u;
+   return v;
 }
 
 
 //// float assignment operators
 
-void __operator += (inout float a, const float b)
+float __operator += (inout float a, const float b)
 {
-    __asm vec4_add a.x, a.x, b.x;
+   a = a + b;
+   return a;
 }
 
-void __operator -= (inout float a, const float b)
+float __operator -= (inout float a, const float b)
 {
-    __asm vec4_subtract a.x, a, b;
+   a = a - b;
+   return a;
 }
 
-void __operator *= (inout float a, const float b)
+float __operator *= (inout float a, const float b)
 {
-    __asm vec4_multiply a.x, a, b;
+   a = a * b;
+   return a;
 }
 
-void __operator /= (inout float a, const float b)
+float __operator /= (inout float a, const float b)
 {
-   float w; // = 1 / b
-   __asm float_rcp w.x, b;
-   __asm vec4_multiply a.x, a, w;
+   a = a / b;
+   return a;
 }
 
 
 //// vec2 assignment operators
 
-void __operator += (inout vec2 v, const vec2 u)
+vec2 __operator += (inout vec2 v, const vec2 u)
 {
-   __asm vec4_add v.xy, v.xy, u.xy;
+   v = v + u;
+   return v;
 }
 
-void __operator -= (inout vec2 v, const vec2 u)
+vec2 __operator -= (inout vec2 v, const vec2 u)
 {
-   __asm vec4_subtract v.xy, v.xy, u.xy;
+   v = v - u;
+   return v;
 }
 
-void __operator *= (inout vec2 v, const vec2 u)
+vec2 __operator *= (inout vec2 v, const vec2 u)
 {
-   __asm vec4_multiply v.xy, v.xy, u.xy;
+   v = v * u;
+   return v;
 }
 
-void __operator /= (inout vec2 v, const vec2 u)
+vec2 __operator /= (inout vec2 v, const vec2 u)
 {
-   vec2 w;
-   __asm float_rcp w.x, u.x;
-   __asm float_rcp w.y, u.y;
-   __asm vec4_multiply v.xy, v.xy, w.xy;
+   v = v / u;
+   return v;
 }
 
 
 //// vec3 assignment operators
 
-void __operator += (inout vec3 v, const vec3 u)
+vec3 __operator += (inout vec3 v, const vec3 u)
 {
-   __asm vec4_add v.xyz, v, u;
+   v = v + u;
+   return v;
 }
 
-void __operator -= (inout vec3 v, const vec3 u)
+vec3 __operator -= (inout vec3 v, const vec3 u)
 {
-   __asm vec4_subtract v.xyz, v, u;
+   v = v - u;
+   return v;
 }
 
-void __operator *= (inout vec3 v, const vec3 u)
+vec3 __operator *= (inout vec3 v, const vec3 u)
 {
-   __asm vec4_multiply v.xyz, v, u;
+   v = v * u;
+   return v;
 }
 
-void __operator /= (inout vec3 v, const vec3 u)
+vec3 __operator /= (inout vec3 v, const vec3 u)
 {
-   vec3 w;
-   __asm float_rcp w.x, u.x;
-   __asm float_rcp w.y, u.y;
-   __asm float_rcp w.z, u.z;
-   __asm vec4_multiply v.xyz, v.xyz, w.xyz;
+   v = v / u;
+   return v;
 }
 
 
 //// vec4 assignment operators
 
-void __operator += (inout vec4 v, const vec4 u)
+vec4 __operator += (inout vec4 v, const vec4 u)
 {
-   __asm vec4_add v, v, u;
+   v = v + u;
+   return v;
 }
 
-void __operator -= (inout vec4 v, const vec4 u)
+vec4 __operator -= (inout vec4 v, const vec4 u)
 {
-   __asm vec4_subtract v, v, u;
+   v = v - u;
+   return v;
 }
 
-void __operator *= (inout vec4 v, const vec4 u)
+vec4 __operator *= (inout vec4 v, const vec4 u)
 {
-   __asm vec4_multiply v, v, u;
+   v = v * u;
+   return v;
 }
 
-void __operator /= (inout vec4 v, const vec4 u)
+vec4 __operator /= (inout vec4 v, const vec4 u)
 {
-   vec4 w;
-   __asm float_rcp w.x, u.x;
-   __asm float_rcp w.y, u.y;
-   __asm float_rcp w.z, u.z;
-   __asm float_rcp w.w, u.w;
-   __asm vec4_multiply v, v, w;
+   v = v / u;
+   return v;
 }
 
 
 
 //// ivec2/int assignment operators
 
-void __operator += (inout ivec2 v, const int a)
+ivec2 __operator += (inout ivec2 v, const int a)
 {
-   __asm vec4_add v.xy, v.xy, a;
+   v = v + ivec2(a);
+   return v;
 }
 
-void __operator -= (inout ivec2 v, const int a)
+ivec2 __operator -= (inout ivec2 v, const int a)
 {
-   __asm vec4_subtract v.xy, v.xy, a;
+   v = v - ivec2(a);
+   return v;
 }
 
-void __operator *= (inout ivec2 v, const int a)
+ivec2 __operator *= (inout ivec2 v, const int a)
 {
-   __asm vec4_multiply v.xy, v.xy, a;
-   v.x *= a;
-   v.y *= a;
+   v = v * ivec2(a);
+   return v;
 }
 
-void __operator /= (inout ivec2 v, const int a)
+ivec2 __operator /= (inout ivec2 v, const int a)
 {
-// XXX rcp
-    v.x /= a;
-    v.y /= a;
+   v = v / ivec2(a);
+   return v;
 }
 
 
 //// ivec3/int assignment operators
 
-void __operator += (inout ivec3 v, const int a)
+ivec3 __operator += (inout ivec3 v, const int a)
 {
-   __asm vec4_add v.xyz, v.xyz, a;
+   v = v + ivec3(a);
+   return v;
 }
 
-void __operator -= (inout ivec3 v, const int a)
+ivec3 __operator -= (inout ivec3 v, const int a)
 {
-   __asm vec4_subtract v.xyz, v.xyz, a;
+   v = v - ivec3(a);
+   return v;
 }
 
-void __operator *= (inout ivec3 v, const int a)
+ivec3 __operator *= (inout ivec3 v, const int a)
 {
-   __asm vec4_multiply v.xyz, v.xyz, a;
+   v = v * ivec3(a);
+   return v;
 }
 
-void __operator /= (inout ivec3 v, const int a)
+ivec4 __operator /= (inout ivec3 v, const int a)
 {
-   // XXX rcp
-    v.x /= a;
-    v.y /= a;
-    v.z /= a;
+   v = v / ivec3(a);
+   return v;
 }
 
 
 //// ivec4/int assignment operators
 
-void __operator += (inout ivec4 v, const int a)
+ivec4 __operator += (inout ivec4 v, const int a)
 {
-   __asm vec4_add v, v, a;
+   v = v + ivec4(a);
+   return v;
 }
 
-void __operator -= (inout ivec4 v, const int a)
+ivec4 __operator -= (inout ivec4 v, const int a)
 {
-   __asm vec4_subtract v, v, a;
+   v = v - ivec4(a);
+   return v;
 }
 
-void __operator *= (inout ivec4 v, const int a)
+ivec4 __operator *= (inout ivec4 v, const int a)
 {
-   __asm vec4_multiply v, v, a;
+   v = v * ivec4(a);
+   return v;
 }
 
-void __operator /= (inout ivec4 v, const int a)
+ivec4 __operator /= (inout ivec4 v, const int a)
 {
-    v.x /= a;
-    v.y /= a;
-    v.z /= a;
-    v.w /= a;
+   v = v / ivec4(a);
+   return v;
 }
 
 
 
 //// vec2/float assignment operators
 
-void __operator += (inout vec2 v, const float a)
+vec2 __operator += (inout vec2 v, const float a)
 {
-   __asm vec4_add v.xy, v, a;
+   v = v + vec2(a);
+   return v;
 }
 
-void __operator -= (inout vec2 v, const float a)
+vec2 __operator -= (inout vec2 v, const float a)
 {
-   __asm vec4_subtract v.xy, v, a;
+   v = v - vec2(a);
+   return v;
 }
 
-void __operator *= (inout vec2 v, const float a)
+vec2 __operator *= (inout vec2 v, const float a)
 {
-   __asm vec4_multiply v.xy, v, a;
+   v = v * vec2(a);
+   return v;
 }
 
-void __operator /= (inout vec2 v, const float a)
+vec2 __operator /= (inout vec2 v, const float a)
 {
-   float invA;
-   __asm float_rcp invA, a;
-   __asm vec4_multiply v.xy, v.xy, invA;
+   v = v / vec2(a);
+   return v;
 }
 
 
 //// vec3/float assignment operators
 
-void __operator += (inout vec3 v, const float a)
+vec3 __operator += (inout vec3 v, const float a)
 {
-   __asm vec4_add v.xyz, v, a;
+   v = v + vec3(a);
+   return v;
 }
 
-void __operator -= (inout vec3 v, const float a)
+vec3 __operator -= (inout vec3 v, const float a)
 {
-   __asm vec4_subtract v.xyz, v, a;
+   v = v - vec3(a);
+   return v;
 }
 
-void __operator *= (inout vec3 v, const float a)
+vec3 __operator *= (inout vec3 v, const float a)
 {
-   __asm vec4_multiply v.xyz, v, a;
+   v = v * vec3(a);
+   return v;
 }
 
-void __operator /= (inout vec3 v, const float a)
+vec3 __operator /= (inout vec3 v, const float a)
 {
-   float invA;
-   __asm float_rcp invA, a;
-   __asm vec4_multiply v.xyz, v.xyz, invA;
+   v = v / vec3(a);
+   return v;
 }
 
 
 //// vec4/float assignment operators
 
-void __operator += (inout vec4 v, const float a)
+vec4 __operator += (inout vec4 v, const float a)
 {
-   __asm vec4_add v, v, a;
+   v = v + vec4(a);
+   return v;
 }
 
-void __operator -= (inout vec4 v, const float a)
+vec4 __operator -= (inout vec4 v, const float a)
 {
-   __asm vec4_subtract v, v, a;
+   v = v - vec4(a);
+   return v;
 }
 
-void __operator *= (inout vec4 v, const float a)
+vec4 __operator *= (inout vec4 v, const float a)
 {
-   __asm vec4_multiply v, v, a;
+   v = v * vec4(a);
+   return v;
 }
 
-void __operator /= (inout vec4 v, const float a)
+vec4 __operator /= (inout vec4 v, const float a)
 {
-   float invA;
-   __asm float_rcp invA, a;
-   __asm vec4_multiply v, v, invA;
+   v = v / vec4(a);
+   return v;
 }
 
 
@@ -1896,187 +1907,239 @@ vec4 __operator * (const vec4 v, const mat4 m)
 
 //// mat2 assignment operators
 
-void __operator += (inout mat2 m, const mat2 n)
+mat2 __operator += (inout mat2 m, const mat2 n)
 {
-    m[0] += n[0];
-    m[1] += n[1];
+   m[0] = m[0] + n[0];
+   m[1] = m[1] + n[1];
+   return m;
 }
 
-void __operator -= (inout mat2 m, const mat2 n)
+mat2 __operator -= (inout mat2 m, const mat2 n)
 {
-    m[0] -= n[0];
-    m[1] -= n[1];
+   m[0] = m[0] - n[0];
+   m[1] = m[1] - n[1];
+   return m;
 }
 
-void __operator *= (inout mat2 m, const mat2 n)
+mat2 __operator *= (inout mat2 m, const mat2 n)
 {
-    m = m * n;
+   m = m * n;
+   return m;
 }
 
-void __operator /= (inout mat2 m, const mat2 n)
+mat2 __operator /= (inout mat2 m, const mat2 n)
 {
-    m[0] /= n[0];
-    m[1] /= n[1];
+   m[0] = m[0] / n[0];
+   m[1] = m[1] / n[1];
+   return m;
 }
 
 
 //// mat3 assignment operators
 
-void __operator += (inout mat3 m, const mat3 n)
+mat3 __operator += (inout mat3 m, const mat3 n)
 {
-    m[0] += n[0];
-    m[1] += n[1];
-    m[2] += n[2];
+   m[0] = m[0] + n[0];
+   m[1] = m[1] + n[1];
+   m[2] = m[2] + n[2];
+   return m;
 }
 
-void __operator -= (inout mat3 m, const mat3 n)
+mat3 __operator -= (inout mat3 m, const mat3 n)
 {
-    m[0] -= n[0];
-    m[1] -= n[1];
-    m[2] -= n[2];
+   m[0] = m[0] - n[0];
+   m[1] = m[1] - n[1];
+   m[2] = m[2] - n[2];
+   return m;
 }
 
-void __operator *= (inout mat3 m, const mat3 n)
+mat3 __operator *= (inout mat3 m, const mat3 n)
 {
-    m = m * n;
+   m = m * n;
+   return m;
 }
 
-void __operator /= (inout mat3 m, const mat3 n)
+mat3 __operator /= (inout mat3 m, const mat3 n)
 {
-    m[0] /= n[0];
-    m[1] /= n[1];
-    m[2] /= n[2];
+   m[0] = m[0] / n[0];
+   m[1] = m[1] / n[1];
+   m[2] = m[2] / n[2];
+   return m;
 }
 
 
 // mat4 assignment operators
 
-void __operator += (inout mat4 m, const mat4 n)
+mat4 __operator += (inout mat4 m, const mat4 n)
 {
-    m[0] += n[0];
-    m[1] += n[1];
-    m[2] += n[2];
-    m[3] += n[3];
+   m[0] = m[0] + n[0];
+   m[1] = m[1] + n[1];
+   m[2] = m[2] + n[2];
+   m[3] = m[3] + n[3];
+   return m;
 }
 
-void __operator -= (inout mat4 m, const mat4 n) {
-    m[0] -= n[0];
-    m[1] -= n[1];
-    m[2] -= n[2];
-    m[3] -= n[3];
+mat4 __operator -= (inout mat4 m, const mat4 n)
+{
+   m[0] = m[0] - n[0];
+   m[1] = m[1] - n[1];
+   m[2] = m[2] - n[2];
+   m[3] = m[3] - n[3];
+   return m;
 }
 
-void __operator *= (inout mat4 m, const mat4 n)
+mat4 __operator *= (inout mat4 m, const mat4 n)
 {
-    m = m * n;
+   m = m * n;
+   return m;
 }
 
-void __operator /= (inout mat4 m, const mat4 n)
+mat4 __operator /= (inout mat4 m, const mat4 n)
 {
-    m[0] /= n[0];
-    m[1] /= n[1];
-    m[2] /= n[2];
-    m[3] /= n[3];
+   m[0] = m[0] / n[0];
+   m[1] = m[1] / n[1];
+   m[2] = m[2] / n[2];
+   m[3] = m[3] / n[3];
+   return m;
 }
 
 
 //// mat2/float assignment operators
 
-void __operator += (inout mat2 m, const float a) {
-    m[0] += a;
-    m[1] += a;
+mat2 __operator += (inout mat2 m, const float a)
+{
+   vec2 v = vec2(a);
+   m[0] = m[0] + v;
+   m[1] = m[1] + v;
+   return m;
 }
 
-void __operator -= (inout mat2 m, const float a) {
-    m[0] -= a;
-    m[1] -= a;
+mat2 __operator -= (inout mat2 m, const float a)
+{
+   vec2 v = vec2(a);
+   m[0] = m[0] - v;
+   m[1] = m[1] - v;
+   return m;
 }
 
-void __operator *= (inout mat2 m, const float a) {
-    m[0] *= a;
-    m[1] *= a;
+mat2 __operator *= (inout mat2 m, const float a)
+{
+   vec2 v = vec2(a);
+   m[0] = m[0] * v;
+   m[1] = m[1] * v;
+   return m;
 }
 
-void __operator /= (inout mat2 m, const float a) {
-    m[0] /= a;
-    m[1] /= a;
+mat2 __operator /= (inout mat2 m, const float a)
+{
+   vec2 v = vec2(1.0 / a);
+   m[0] = m[0] * v;
+   m[1] = m[1] * v;
+   return m;
 }
 
 
 //// mat3/float assignment operators
 
-void __operator += (inout mat3 m, const float a) {
-    m[0] += a;
-    m[1] += a;
-    m[2] += a;
+mat3 __operator += (inout mat3 m, const float a)
+{
+   vec3 v = vec3(a);
+   m[0] = m[0] + v;
+   m[1] = m[1] + v;
+   m[2] = m[2] + v;
+   return m;
 }
 
-void __operator -= (inout mat3 m, const float a) {
-    m[0] -= a;
-    m[1] -= a;
-    m[2] -= a;
+mat3 __operator -= (inout mat3 m, const float a)
+{
+   vec3 v = vec3(a);
+   m[0] = m[0] - v;
+   m[1] = m[1] - v;
+   m[2] = m[2] - v;
+   return m;
 }
 
-void __operator *= (inout mat3 m, const float a) {
-    m[0] *= a;
-    m[1] *= a;
-    m[2] *= a;
+mat3 __operator *= (inout mat3 m, const float a)
+{
+   vec3 v = vec3(a);
+   m[0] = m[0] * v;
+   m[1] = m[1] * v;
+   m[2] = m[2] * v;
+   return m;
 }
 
-void __operator /= (inout mat3 m, const float a) {
-    m[0] /= a;
-    m[1] /= a;
-    m[2] /= a;
+mat3 __operator /= (inout mat3 m, const float a)
+{
+   vec3 v = vec3(1.0 / a);
+   m[0] = m[0] * v;
+   m[1] = m[1] * v;
+   m[2] = m[2] * v;
+   return m;
 }
 
 
 //// mat4/float assignment operators
 
-void __operator += (inout mat4 m, const float a) {
-    m[0] += a;
-    m[1] += a;
-    m[2] += a;
-    m[3] += a;
+mat4 __operator += (inout mat4 m, const float a)
+{
+   vec4 v = vec4(a);
+   m[0] = m[0] + v;
+   m[1] = m[1] + v;
+   m[2] = m[2] + v;
+   m[3] = m[3] + v;
+   return m;
 }
 
-void __operator -= (inout mat4 m, const float a) {
-    m[0] -= a;
-    m[1] -= a;
-    m[2] -= a;
-    m[3] -= a;
+mat4 __operator -= (inout mat4 m, const float a)
+{
+   vec4 v = vec4(a);
+   m[0] = m[0] - v;
+   m[1] = m[1] - v;
+   m[2] = m[2] - v;
+   m[3] = m[3] - v;
+   return m;
 }
 
-void __operator *= (inout mat4 m, const float a) {
-    m[0] *= a;
-    m[1] *= a;
-    m[2] *= a;
-    m[3] *= a;
+mat4 __operator *= (inout mat4 m, const float a)
+{
+   vec4 v = vec4(a);
+   m[0] = m[0] * v;
+   m[1] = m[1] * v;
+   m[2] = m[2] * v;
+   m[3] = m[3] * v;
+   return m;
 }
 
-void __operator /= (inout mat4 m, const float a) {
-    m[0] /= a;
-    m[1] /= a;
-    m[2] /= a;
-    m[3] /= a;
+mat4 __operator /= (inout mat4 m, const float a)
+{
+   vec4 v = vec4(1.0 / a);
+   m[0] = m[0] * v;
+   m[1] = m[1] * v;
+   m[2] = m[2] * v;
+   m[3] = m[3] * v;
+   return m;
 }
 
 
 
 //// vec/mat assignment operators
 
-void __operator *= (inout vec2 v, const mat2 m)
+vec2 __operator *= (inout vec2 v, const mat2 m)
 {
-    v = v * m;
+   v = v * m;
+   return v;
 }
 
-void __operator *= (inout vec3 v, const mat3 m)
+vec3 __operator *= (inout vec3 v, const mat3 m)
 {
-    v = v * m;
+   v = v * m;
+   return v;
 }
 
-void __operator *= (inout vec4 v, const mat4 m)
+vec4 __operator *= (inout vec4 v, const mat4 m)
 {
-    v = v * m;
+   v = v * m;
+   return v;
 }
 
 
diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h
index 03cacc4db6..b3d3e87cf4 100644
--- a/src/mesa/shader/slang/library/slang_core_gc.h
+++ b/src/mesa/shader/slang/library/slang_core_gc.h
@@ -2,857 +2,868 @@
 /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */
 /* slang_core.gc */
 
-4,1,90,95,0,0,5,1,1,1,0,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,
-114,101,116,86,97,108,0,0,18,102,0,0,0,0,1,90,95,0,0,5,1,1,1,0,0,1,98,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,18,98,0,20,0,0,1,90,95,0,0,5,1,1,1,0,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,18,105,0,20,0,0,1,90,95,0,0,1,1,1,1,0,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,
-95,114,101,116,86,97,108,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,1,1,1,0,0,9,102,0,0,0,1,
-4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,0,17,48,0,48,0,0,0,0,
-0,1,90,95,0,0,1,1,1,1,0,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,90,95,0,
-0,9,1,1,1,0,0,5,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,
-86,97,108,0,0,18,105,0,0,0,0,1,90,95,0,0,9,1,1,1,0,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,111,
-95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,90,95,0,0,9,1,1,1,0,0,9,102,
-0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,102,0,20,0,0,1,90,95,0,0,10,1,1,1,0,0,9,120,0,0,1,1,
-0,0,9,121,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,
-86,97,108,0,59,121,0,18,121,0,20,0,0,1,90,95,0,0,10,1,1,1,0,0,9,102,0,0,0,1,4,118,101,99,52,95,109,
-111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,0,0,1,90,95,0,0,10,1,1,1,
-0,0,5,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,
-0,59,120,121,0,0,18,105,0,0,0,0,1,90,95,0,0,10,1,1,1,0,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,
-111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,
-10,1,1,1,0,0,2,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,
-86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,10,1,1,1,0,0,11,118,0,0,0,1,4,118,101,99,52,
-95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,
-90,95,0,0,10,1,1,1,0,0,12,118,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,
-97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,90,95,0,0,11,1,1,1,0,0,9,120,0,0,1,1,0,0,9,
-121,0,0,1,1,0,0,9,122,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,
-0,20,0,0,1,90,95,0,0,11,1,1,1,0,0,9,102,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,
-101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,0,0,1,90,95,0,0,11,1,1,1,0,0,5,105,0,0,0,1,4,105,
-118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,
-105,0,0,0,0,1,90,95,0,0,11,1,1,1,0,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,
-0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,1,1,1,0,0,3,98,
-0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,
-121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,1,1,1,0,0,12,118,0,0,0,1,4,118,101,99,52,95,109,111,118,
-101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,0,0,1,90,95,0,0,12,1,1,1,0,0,
-9,120,0,0,1,1,0,0,9,121,0,0,1,1,0,0,9,122,0,0,1,1,0,0,9,119,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,
-0,20,0,0,1,90,95,0,0,12,1,1,1,0,0,9,102,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,
-101,116,86,97,108,0,0,18,102,0,0,0,0,1,90,95,0,0,12,1,1,1,0,0,5,105,0,0,0,1,4,105,118,101,99,52,95,
-116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,90,95,0,0,12,1,1,1,
-0,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,
-0,0,18,98,0,0,0,0,1,90,95,0,0,12,1,1,1,0,0,4,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,
-99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,90,95,0,0,12,1,1,1,0,0,8,105,0,0,0,1,4,
-105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,
-1,90,95,0,0,12,1,1,1,0,0,11,118,51,0,0,1,1,0,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,
-120,121,122,0,18,118,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,0,20,0,0,1,90,95,
-0,0,12,1,1,1,0,0,10,118,50,0,0,1,1,0,0,9,102,49,0,0,1,1,0,0,9,102,50,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,59,120,121,0,18,118,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,102,
-49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,50,0,20,0,0,1,90,95,0,0,6,1,1,1,0,0,5,
-105,0,0,1,1,0,0,5,106,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,0,1,90,95,0,0,6,1,1,1,0,0,5,105,0,0,0,1,4,118,101,
-99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,0,0,0,1,90,95,
-0,0,6,1,1,1,0,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,
-116,86,97,108,0,59,120,121,0,0,18,102,0,0,0,0,1,90,95,0,0,6,1,1,1,0,0,1,98,0,0,0,1,4,118,101,99,52,
-95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,
-90,95,0,0,7,1,1,1,0,0,5,105,0,0,1,1,0,0,5,106,0,0,1,1,0,0,5,107,0,0,0,1,9,18,95,95,114,101,116,86,
-97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,9,18,95,
-95,114,101,116,86,97,108,0,59,122,0,18,107,0,20,0,0,1,90,95,0,0,7,1,1,1,0,0,5,105,0,0,0,1,4,118,
-101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,0,0,0,
-1,90,95,0,0,7,1,1,1,0,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,
-114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,0,0,1,90,95,0,0,7,1,1,1,0,0,1,98,0,0,0,1,4,
-118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,
-0,1,90,95,0,0,8,1,1,1,0,0,5,120,0,0,1,1,0,0,5,121,0,0,1,1,0,0,5,122,0,0,1,1,0,0,5,119,0,0,0,1,9,18,
+5,1,90,95,0,0,5,0,1,1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,
+95,95,114,101,116,86,97,108,0,0,18,102,0,0,0,0,1,90,95,0,0,5,0,1,1,1,0,0,1,0,98,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,18,98,0,20,0,0,1,90,95,0,0,5,0,1,1,1,0,0,5,0,105,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,18,105,0,20,0,0,1,90,95,0,0,1,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,52,95,
+115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,1,
+1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,102,
+0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,1,0,1,1,1,0,0,1,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
+0,18,98,0,20,0,0,1,90,95,0,0,9,0,1,1,1,0,0,5,0,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,
+101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,90,95,0,0,9,0,1,1,1,0,0,1,0,98,0,0,
+0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,
+0,0,1,90,95,0,0,9,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,102,0,20,0,0,1,
+90,95,0,0,10,0,1,1,1,0,0,9,0,120,0,0,1,1,0,0,9,0,121,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,
+120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,0,1,90,95,0,0,10,0,1,
+1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,
+120,121,0,0,18,102,0,0,0,0,1,90,95,0,0,10,0,1,1,1,0,0,5,0,105,0,0,0,1,4,105,118,101,99,52,95,116,
+111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,0,0,0,1,90,95,0,0,
+10,0,1,1,1,0,0,1,0,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,
+116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,1,1,1,0,0,2,0,98,0,0,0,1,4,105,118,
+101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,
+0,0,1,90,95,0,0,10,0,1,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,
+101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,90,95,0,0,10,0,1,1,1,0,0,12,0,118,
+0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,
+118,0,59,120,121,0,0,0,0,1,90,95,0,0,11,0,1,1,1,0,0,9,0,120,0,0,1,1,0,0,9,0,121,0,0,1,1,0,0,9,0,
+122,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,0,1,90,95,0,
+0,11,0,1,1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,
+108,0,59,120,121,122,0,0,18,102,0,0,0,0,1,90,95,0,0,11,0,1,1,1,0,0,5,0,105,0,0,0,1,4,105,118,101,
+99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,0,
+0,0,1,90,95,0,0,11,0,1,1,1,0,0,1,0,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,
+95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,1,1,1,0,0,3,0,98,0,
+0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,
+121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,1,1,1,0,0,12,0,118,0,0,0,1,4,118,101,99,52,95,109,111,
+118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,0,0,1,90,95,0,0,12,0,1,1,
+1,0,0,9,0,120,0,0,1,1,0,0,9,0,121,0,0,1,1,0,0,9,0,122,0,0,1,1,0,0,9,0,119,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,
+0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,
+119,0,18,119,0,20,0,0,1,90,95,0,0,12,0,1,1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,109,111,118,
+101,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,0,0,0,1,90,95,0,0,12,0,1,1,1,0,0,5,0,105,0,0,0,1,
+4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,
+0,1,90,95,0,0,12,0,1,1,1,0,0,1,0,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,
+95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,1,1,1,0,0,4,0,98,0,0,0,1,4,105,118,
+101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,90,95,0,
+0,12,0,1,1,1,0,0,8,0,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,
+101,116,86,97,108,0,0,18,105,0,0,0,0,1,90,95,0,0,12,0,1,1,1,0,0,11,0,118,51,0,0,1,1,0,0,9,0,102,0,
+0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,118,51,0,20,0,9,18,95,95,114,101,116,
+86,97,108,0,59,119,0,18,102,0,20,0,0,1,90,95,0,0,12,0,1,1,1,0,0,10,0,118,50,0,0,1,1,0,0,9,0,102,49,
+0,0,1,1,0,0,9,0,102,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,118,50,0,20,0,9,
+18,95,95,114,101,116,86,97,108,0,59,122,0,18,102,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,
+119,0,18,102,50,0,20,0,0,1,90,95,0,0,6,0,1,1,1,0,0,5,0,105,0,0,1,1,0,0,5,0,106,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,
+0,20,0,0,1,90,95,0,0,6,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,
+114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,0,0,0,1,90,95,0,0,6,0,1,1,1,0,0,9,0,102,0,0,0,1,4,
+118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,
+102,0,0,0,0,1,90,95,0,0,6,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,
+52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,7,0,1,1,1,0,0,5,0,
+105,0,0,1,1,0,0,5,0,106,0,0,1,1,0,0,5,0,107,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,
+105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,59,122,0,18,107,0,20,0,0,1,90,95,0,0,7,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,52,95,109,
+111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,0,0,0,1,90,95,0,0,7,0,1,
+1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,
+97,108,0,59,120,121,122,0,0,18,102,0,0,0,0,1,90,95,0,0,7,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99,
+52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,
+0,0,8,0,1,1,1,0,0,5,0,120,0,0,1,1,0,0,5,0,121,0,0,1,1,0,0,5,0,122,0,0,1,1,0,0,5,0,119,0,0,0,1,9,18,
 95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,
 18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,
-97,108,0,59,119,0,18,119,0,20,0,0,1,90,95,0,0,8,1,1,1,0,0,5,105,0,0,0,1,4,118,101,99,52,95,109,111,
-118,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,90,95,0,0,8,1,1,1,0,0,9,102,0,0,0,1,
-4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,0,0,
-0,1,90,95,0,0,8,1,1,1,0,0,1,98,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,
-114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,90,95,0,0,2,1,1,1,0,0,1,98,49,0,0,1,1,0,0,1,98,50,0,0,0,
-1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,
-59,121,0,18,98,50,0,20,0,0,1,90,95,0,0,2,1,1,1,0,0,5,105,49,0,0,1,1,0,0,5,105,50,0,0,0,1,4,118,101,
-99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,105,49,0,0,17,48,0,48,0,0,0,
-0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,105,50,0,0,17,48,
-0,48,0,0,0,0,0,1,90,95,0,0,2,1,1,1,0,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,
-114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,2,1,1,1,0,0,9,102,0,0,0,1,4,118,
-101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,17,48,0,48,0,
-0,0,0,0,1,90,95,0,0,2,1,1,1,0,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,
-116,86,97,108,0,59,120,121,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,2,1,1,1,0,0,10,118,0,0,
-0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,
-48,0,48,0,0,0,0,0,1,90,95,0,0,2,1,1,1,0,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,
-114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,1,1,1,0,0,1,
-98,49,0,0,1,1,0,0,1,98,50,0,0,1,1,0,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,
-18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,
-86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,90,95,0,0,3,1,1,1,0,0,9,102,49,0,0,1,1,0,0,9,102,50,0,0,1,
-1,0,0,9,102,51,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,
-0,18,102,49,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,
-0,59,121,0,0,18,102,50,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,
-116,86,97,108,0,59,122,0,0,18,102,51,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,1,1,1,0,0,1,98,0,0,0,1,
-4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,
-0,0,1,90,95,0,0,3,1,1,1,0,0,9,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,
-97,108,0,59,120,121,122,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,1,1,1,0,0,5,105,0,0,0,1,
-4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,0,17,
-48,0,48,0,0,0,0,0,1,90,95,0,0,3,1,1,1,0,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,
-114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,1,1,1,0,0,
-7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,
-18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,4,1,1,1,0,0,1,98,49,0,0,1,1,0,0,1,98,50,0,0,1,1,0,0,1,
-98,51,0,0,1,1,0,0,1,98,52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,
-95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,
-18,98,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,90,95,0,0,4,1,1,1,
-0,0,9,102,49,0,0,1,1,0,0,9,102,50,0,0,1,1,0,0,9,102,51,0,0,1,1,0,0,9,102,52,0,0,0,1,3,2,90,95,1,0,
-9,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,
-86,97,108,0,59,120,0,0,18,102,49,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,
-95,95,114,101,116,86,97,108,0,59,121,0,0,18,102,50,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,
-115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,102,51,0,0,18,122,101,114,111,0,0,0,4,
-118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,102,52,0,0,18,122,
-101,114,111,0,0,0,0,1,90,95,0,0,4,1,1,1,0,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,
-95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,98,0,0,0,0,1,90,95,0,0,4,1,1,1,0,0,9,102,0,
-0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,
-102,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,4,1,1,1,0,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,
-0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,90,95,
-0,0,4,1,1,1,0,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,
-59,120,121,122,119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,4,1,1,1,0,0,8,118,0,0,0,1,4,118,
-101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,118,0,0,17,
-48,0,48,0,0,0,0,0,1,90,95,0,0,13,1,1,1,0,0,9,109,48,48,0,0,1,1,0,0,9,109,49,48,0,0,1,1,0,0,9,109,
-48,49,0,0,1,1,0,0,9,109,49,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,
-109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,
-97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1,90,95,0,0,13,1,1,1,0,0,9,102,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,0,20,0,9,18,95,95,114,101,116,86,97,108,
-0,16,8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,
-120,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,0,20,0,
-0,1,90,95,0,0,13,1,1,1,0,0,5,105,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,105,0,0,0,
-0,0,0,0,1,90,95,0,0,13,1,1,1,0,0,1,98,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,98,0,
-0,0,0,0,0,0,1,90,95,0,0,13,1,1,1,0,0,10,99,48,0,0,1,1,0,0,10,99,49,0,0,0,1,9,18,95,95,114,101,116,
-86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,
-0,20,0,0,1,90,95,0,0,14,1,1,1,0,0,9,109,48,48,0,0,1,1,0,0,9,109,49,48,0,0,1,1,0,0,9,109,50,48,0,0,
-1,1,0,0,9,109,48,49,0,0,1,1,0,0,9,109,49,49,0,0,1,1,0,0,9,109,50,49,0,0,1,1,0,0,9,109,48,50,0,0,1,
-1,0,0,9,109,49,50,0,0,1,1,0,0,9,109,50,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,
-59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,
-0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,
-16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,
-0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,
-0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101,
-116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,0,1,90,95,0,0,14,1,1,1,0,0,9,102,0,0,0,
-1,3,2,90,95,0,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,
-101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,
-16,10,49,0,57,18,118,0,59,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
-118,0,59,121,121,120,0,20,0,0,1,90,95,0,0,14,1,1,1,0,0,5,105,0,0,0,1,8,58,109,97,116,51,0,58,102,
-108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,90,95,0,0,14,1,1,1,0,0,1,98,0,0,0,1,8,58,109,97,116,51,0,
-58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,90,95,0,0,14,1,1,1,0,0,11,99,48,0,0,1,1,0,0,11,99,49,
-0,0,1,1,0,0,11,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
-10,50,0,57,18,99,50,0,20,0,0,1,90,95,0,0,15,1,1,1,0,0,9,109,48,48,0,0,1,1,0,0,9,109,49,48,0,0,1,1,
-0,0,9,109,50,48,0,0,1,1,0,0,9,109,51,48,0,0,1,1,0,0,9,109,48,49,0,0,1,1,0,0,9,109,49,49,0,0,1,1,0,
-0,9,109,50,49,0,0,1,1,0,0,9,109,51,49,0,0,1,1,0,0,9,109,48,50,0,0,1,1,0,0,9,109,49,50,0,0,1,1,0,0,
-9,109,50,50,0,0,1,1,0,0,9,109,51,50,0,0,1,1,0,0,9,109,48,51,0,0,1,1,0,0,9,109,49,51,0,0,1,1,0,0,9,
-109,50,51,0,0,1,1,0,0,9,109,51,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,
+97,108,0,59,119,0,18,119,0,20,0,0,1,90,95,0,0,8,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,52,95,109,
+111,118,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,90,95,0,0,8,0,1,1,1,0,0,9,0,102,
+0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,
+102,0,0,0,0,1,90,95,0,0,8,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,
+52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,90,95,0,0,2,0,1,1,1,0,0,1,0,98,49,0,0,1,1,
+0,0,1,0,98,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,
+101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,0,1,90,95,0,0,2,0,1,1,1,0,0,5,0,105,49,0,0,1,1,0,0,5,
+0,105,50,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,
+105,49,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,
+121,0,0,18,105,50,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,2,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99,
+52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,
+2,0,1,1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,
+120,121,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,2,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,
+52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,0,17,48,0,48,0,0,0,0,
+0,1,90,95,0,0,2,0,1,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,
+86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,2,0,1,1,1,0,0,6,0,118,0,0,0,
+1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,
+48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,1,1,1,0,0,1,0,98,49,0,0,1,1,0,0,1,0,98,50,0,0,1,1,0,0,1,0,98,51,
+0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,90,
+95,0,0,3,0,1,1,1,0,0,9,0,102,49,0,0,1,1,0,0,9,0,102,50,0,0,1,1,0,0,9,0,102,51,0,0,0,1,4,118,101,99,
+52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,102,49,0,0,17,48,0,48,0,0,0,0,4,
+118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,102,50,0,0,17,48,0,
+48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,102,51,
+0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99,52,95,109,111,118,
+101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,3,0,1,1,1,0,0,
+9,0,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,
+0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,101,99,52,95,115,
+110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,90,
+95,0,0,3,0,1,1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,
+108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,3,0,1,1,1,0,0,7,0,118,0,0,0,1,
+4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17,
+48,0,48,0,0,0,0,0,1,90,95,0,0,4,0,1,1,1,0,0,1,0,98,49,0,0,1,1,0,0,1,0,98,50,0,0,1,1,0,0,1,0,98,51,
+0,0,1,1,0,0,1,0,98,52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,
+95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,
+98,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,90,95,0,0,4,0,1,1,1,0,
+0,9,0,102,49,0,0,1,1,0,0,9,0,102,50,0,0,1,1,0,0,9,0,102,51,0,0,1,1,0,0,9,0,102,52,0,0,0,1,3,2,90,
+95,1,0,9,0,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,
+101,116,86,97,108,0,59,120,0,0,18,102,49,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,
+101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,102,50,0,0,18,122,101,114,111,0,0,0,4,118,101,
+99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,102,51,0,0,18,122,101,114,
+111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,102,52,0,
+0,18,122,101,114,111,0,0,0,0,1,90,95,0,0,4,0,1,1,1,0,0,1,0,98,0,0,0,1,4,118,101,99,52,95,109,111,
+118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,98,0,0,0,0,1,90,95,0,0,4,0,1,
+1,1,0,0,9,0,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,
+121,122,119,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,4,0,1,1,1,0,0,5,0,105,0,0,0,1,4,118,
+101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,105,0,0,17,
+48,0,48,0,0,0,0,0,1,90,95,0,0,4,0,1,1,1,0,0,12,0,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,
+95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,4,
+0,1,1,1,0,0,8,0,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,
+120,121,122,119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,90,95,0,0,13,0,1,1,1,0,0,9,0,109,48,48,0,0,1,
+1,0,0,9,0,109,49,48,0,0,1,1,0,0,9,0,109,48,49,0,0,1,1,0,0,9,0,109,49,49,0,0,0,1,9,18,95,95,114,101,
+116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,
+48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,
+109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1,
+90,95,0,0,13,0,1,1,1,0,0,9,0,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,
+18,102,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95,
+95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,49,0,57,59,121,0,18,102,0,20,0,0,1,90,95,0,0,13,0,1,1,1,0,0,5,0,105,0,0,0,1,8,58,109,
+97,116,50,0,0,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,0,0,1,90,95,0,0,13,0,1,1,1,0,0,1,0,98,0,0,
+0,1,8,58,109,97,116,50,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,0,0,1,90,95,0,0,13,0,1,1,1,0,
+0,10,0,99,48,0,0,1,1,0,0,10,0,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,
+48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,90,95,0,0,14,0,1,1,
+1,0,0,9,0,109,48,48,0,0,1,1,0,0,9,0,109,49,48,0,0,1,1,0,0,9,0,109,50,48,0,0,1,1,0,0,9,0,109,48,49,
+0,0,1,1,0,0,9,0,109,49,49,0,0,1,1,0,0,9,0,109,50,49,0,0,1,1,0,0,9,0,109,48,50,0,0,1,1,0,0,9,0,109,
+49,50,0,0,1,1,0,0,9,0,109,50,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,
 18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,
 18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,8,48,0,57,59,119,0,18,109,51,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,
-0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,
-49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0,18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,
-59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,
-50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,16,10,51,0,57,59,120,0,18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,51,0,57,59,121,0,18,109,49,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,
-59,122,0,18,109,50,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,
-51,0,20,0,0,1,90,95,0,0,15,1,1,1,0,0,9,102,0,0,0,1,3,2,90,95,0,0,10,1,118,0,2,58,118,101,99,50,0,
-18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,
-121,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,121,0,
-20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59,121,121,120,121,0,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,51,0,57,18,118,0,59,121,121,121,120,0,20,0,0,1,90,95,0,0,15,1,1,1,
-0,0,5,105,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,90,95,0,0,15,
-1,1,1,0,0,1,98,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,90,95,0,
-0,15,1,1,1,0,0,12,99,48,0,0,1,1,0,0,12,99,49,0,0,1,1,0,0,12,99,50,0,0,1,1,0,0,12,99,51,0,0,0,1,9,
-18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,
-16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,
-18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,90,95,0,0,5,2,26,1,1,0,0,5,97,0,
-0,1,1,0,0,5,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,
-0,18,98,0,0,0,0,1,90,95,0,0,5,2,27,1,1,0,0,5,97,0,0,1,1,0,0,5,98,0,0,0,1,4,118,101,99,52,95,115,
+86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,
+0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,
+50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,
+95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,9,0,102,0,0,0,1,3,2,
+90,95,0,0,10,0,1,118,0,2,58,118,101,99,50,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,
+116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
+10,49,0,57,18,118,0,59,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,
+0,59,121,121,120,0,20,0,0,1,90,95,0,0,14,0,1,1,1,0,0,5,0,105,0,0,0,1,8,58,109,97,116,51,0,0,58,102,
+108,111,97,116,0,0,18,105,0,0,0,0,0,0,0,1,90,95,0,0,14,0,1,1,1,0,0,1,0,98,0,0,0,1,8,58,109,97,116,
+51,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,0,0,1,90,95,0,0,14,0,1,1,1,0,0,11,0,99,48,0,0,1,1,
+0,0,11,0,99,49,0,0,1,1,0,0,11,0,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
+99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,9,0,109,48,48,0,0,1,1,0,
+0,9,0,109,49,48,0,0,1,1,0,0,9,0,109,50,48,0,0,1,1,0,0,9,0,109,51,48,0,0,1,1,0,0,9,0,109,48,49,0,0,
+1,1,0,0,9,0,109,49,49,0,0,1,1,0,0,9,0,109,50,49,0,0,1,1,0,0,9,0,109,51,49,0,0,1,1,0,0,9,0,109,48,
+50,0,0,1,1,0,0,9,0,109,49,50,0,0,1,1,0,0,9,0,109,50,50,0,0,1,1,0,0,9,0,109,51,50,0,0,1,1,0,0,9,0,
+109,48,51,0,0,1,1,0,0,9,0,109,49,51,0,0,1,1,0,0,9,0,109,50,51,0,0,1,1,0,0,9,0,109,51,51,0,0,0,1,9,
+18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,
+86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,
+57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,109,51,
+48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,
+59,119,0,18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,
+50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,
+59,120,0,18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,109,49,
+51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,51,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,9,
+0,102,0,0,0,1,3,2,90,95,0,0,10,0,1,118,0,2,58,118,101,99,50,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,
+9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,121,0,20,0,9,18,95,95,114,
+101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,50,0,57,18,118,0,59,121,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,
+0,57,18,118,0,59,121,121,121,120,0,20,0,0,1,90,95,0,0,15,0,1,1,1,0,0,5,0,105,0,0,0,1,8,58,109,97,
+116,52,0,0,58,102,108,111,97,116,0,0,18,105,0,0,0,0,0,0,0,1,90,95,0,0,15,0,1,1,1,0,0,1,0,98,0,0,0,
+1,8,58,109,97,116,52,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,0,0,0,0,1,90,95,0,0,15,0,1,1,1,0,0,
+12,0,99,48,0,0,1,1,0,0,12,0,99,49,0,0,1,1,0,0,12,0,99,50,0,0,1,1,0,0,12,0,99,51,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
+57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114,
+101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,90,95,0,0,5,0,2,26,1,1,0,0,5,0,97,0,0,1,1,0,
+0,5,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,
+98,0,0,0,0,1,90,95,0,0,5,0,2,27,1,1,0,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,4,118,101,99,52,95,115,
 117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,
-5,2,21,1,1,0,0,5,97,0,0,1,1,0,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,
-18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,5,2,22,1,1,0,0,5,97,0,0,1,1,
-0,0,5,98,0,0,0,1,3,2,90,95,0,0,9,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,
-112,0,18,98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,
-0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,
-114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,6,2,26,1,1,0,0,6,97,0,0,1,1,0,0,6,98,0,0,0,1,
-4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,
-0,0,6,2,27,1,1,0,0,6,97,0,0,1,1,0,0,6,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,
-18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,6,2,21,1,1,0,0,6,97,0,0,1,1,
-0,0,6,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,
-108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,6,2,22,1,1,0,0,6,97,0,0,1,1,0,0,6,98,0,0,0,1,3,2,90,95,
-0,0,10,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,
-59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,
-18,98,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,
-98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,
-108,0,0,18,120,0,0,0,0,1,90,95,0,0,7,2,26,1,1,0,0,7,97,0,0,1,1,0,0,7,98,0,0,0,1,4,118,101,99,52,95,
-97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,7,2,27,1,1,0,0,
-7,97,0,0,1,1,0,0,7,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,
-116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,7,2,21,1,1,0,0,7,97,0,0,1,1,0,0,7,98,0,0,0,1,
-4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,
-18,98,0,0,0,0,1,90,95,0,0,7,2,22,1,1,0,0,7,97,0,0,1,1,0,0,7,98,0,0,0,1,3,2,90,95,0,0,11,1,98,73,
-110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,
-0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,
-0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,118,
-101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,
-101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,
-95,0,0,8,2,26,1,1,0,0,8,97,0,0,1,1,0,0,8,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,
-101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,8,2,27,1,1,0,0,8,97,0,0,1,1,0,0,8,98,0,0,
-0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,
-18,98,0,0,0,0,1,90,95,0,0,8,2,21,1,1,0,0,8,97,0,0,1,1,0,0,8,98,0,0,0,1,4,118,101,99,52,95,109,117,
-108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,8,
-2,22,1,1,0,0,8,97,0,0,1,1,0,0,8,98,0,0,0,1,3,2,90,95,0,0,12,1,98,73,110,118,0,0,1,1,120,0,0,0,4,
+5,0,2,21,1,1,0,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,
+121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,5,0,2,22,1,1,0,0,5,0,
+97,0,0,1,1,0,0,5,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,
+116,95,114,99,112,0,18,98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,
+108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,
+52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,6,0,2,26,1,1,0,0,6,0,97,0,0,1,1,
+0,0,6,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,
+98,0,0,0,0,1,90,95,0,0,6,0,2,27,1,1,0,0,6,0,97,0,0,1,1,0,0,6,0,98,0,0,0,1,4,118,101,99,52,95,115,
+117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,
+6,0,2,21,1,1,0,0,6,0,97,0,0,1,1,0,0,6,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,
+121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,6,0,2,22,1,1,0,0,6,0,
+97,0,0,1,1,0,0,6,0,98,0,0,0,1,3,2,90,95,0,0,10,0,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,
+97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,
+114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,
+105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,
+118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,7,0,2,26,1,1,0,0,7,0,
+97,0,0,1,1,0,0,7,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,
+18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,7,0,2,27,1,1,0,0,7,0,97,0,0,1,1,0,0,7,0,98,0,0,0,1,4,118,101,
+99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,
+1,90,95,0,0,7,0,2,21,1,1,0,0,7,0,97,0,0,1,1,0,0,7,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,
+105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,7,0,2,22,
+1,1,0,0,7,0,97,0,0,1,1,0,0,7,0,98,0,0,0,1,3,2,90,95,0,0,11,0,1,98,73,110,118,0,0,1,1,120,0,0,0,4,
+102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,
+111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,
+95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,
+116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,
+105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,8,0,2,26,1,1,0,0,
+8,0,97,0,0,1,1,0,0,8,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,
+0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,8,0,2,27,1,1,0,0,8,0,97,0,0,1,1,0,0,8,0,98,0,0,0,1,4,118,101,
+99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,
+1,90,95,0,0,8,0,2,21,1,1,0,0,8,0,97,0,0,1,1,0,0,8,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,
+105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,8,0,2,22,
+1,1,0,0,8,0,97,0,0,1,1,0,0,8,0,98,0,0,0,1,3,2,90,95,0,0,12,0,1,98,73,110,118,0,0,1,1,120,0,0,0,4,
 102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,
 111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,
 95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,
 112,0,18,98,73,110,118,0,59,119,0,0,18,98,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,
 112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,
-101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,9,2,26,1,1,0,0,9,97,0,0,
-1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,
-18,98,0,0,0,0,1,90,95,0,0,9,2,27,1,1,0,0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,
-98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,2,
-21,1,1,0,0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,
-95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,2,22,1,1,0,0,9,97,0,0,1,1,0,
-0,9,98,0,0,0,1,3,2,90,95,0,0,9,1,98,73,110,118,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,
-110,118,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,
-101,116,86,97,108,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,90,95,0,0,10,2,26,1,1,0,0,10,118,0,0,1,
-1,0,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,
-0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,10,2,27,1,1,0,0,10,118,0,0,1,1,0,0,10,117,0,0,0,1,4,118,
-101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,
-0,0,18,117,0,0,0,0,1,90,95,0,0,10,2,21,1,1,0,0,10,118,0,0,1,1,0,0,10,117,0,0,0,1,4,118,101,99,52,
-95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,
-117,0,0,0,0,1,90,95,0,0,10,2,22,1,1,0,0,10,118,0,0,1,1,0,0,10,117,0,0,0,1,3,2,90,95,0,0,10,1,119,0,
-0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,
-97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,
-116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,119,0,0,0,0,1,
-90,95,0,0,11,2,26,1,1,0,0,11,118,0,0,1,1,0,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,
-95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,11,2,27,1,1,0,
-0,11,118,0,0,1,1,0,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,
-101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,11,2,21,1,1,0,0,11,
-118,0,0,1,1,0,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,
-101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,11,2,22,1,1,0,0,11,
-118,0,0,1,1,0,0,11,117,0,0,0,1,3,2,90,95,0,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,
-18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,
-18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,
-0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,
-122,0,0,18,118,0,0,18,119,0,0,0,0,1,90,95,0,0,12,2,26,1,1,0,0,12,118,0,0,1,1,0,0,12,117,0,0,0,1,4,
-118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,
-0,0,12,2,27,1,1,0,0,12,118,0,0,1,1,0,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,
-116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,12,2,21,1,1,0,0,12,
-118,0,0,1,1,0,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,
-101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,12,2,22,1,1,0,0,12,118,0,0,1,1,0,0,12,
-117,0,0,0,1,3,2,90,95,0,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,
-18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,
-0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,
-116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,
-105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,119,0,0,0,0,1,90,95,0,0,10,2,26,
-1,1,0,0,9,97,0,0,1,1,0,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,
-108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,90,95,0,0,10,2,26,1,1,0,0,10,118,0,0,
-1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,
-0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,10,2,27,1,1,0,0,9,97,0,0,1,1,0,0,10,117,0,0,0,
-1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,
-0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,90,95,0,0,10,2,27,1,1,0,0,10,118,0,0,1,1,0,0,9,98,0,0,0,
-1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,
-0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,10,2,21,1,1,0,0,9,97,0,0,1,1,0,0,10,117,0,0,0,
-1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,
-0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,90,95,0,0,10,2,21,1,1,0,0,10,118,0,0,1,1,0,0,9,98,0,0,
+101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,90,95,0,0,9,0,2,26,1,1,0,0,9,0,97,
+0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,
+97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,
+52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,
+90,95,0,0,9,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,
+105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,0,2,22,
+1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,98,73,110,118,0,0,0,4,102,108,111,97,
+116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,
+105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,90,95,0,
+0,10,0,2,26,1,1,0,0,10,0,118,0,0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,
+114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,10,0,2,27,1,1,0,0,10,
+0,118,0,0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,
+101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,10,0,2,21,1,1,0,0,10,0,
+118,0,0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,
+101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,10,0,2,22,1,1,0,0,10,0,
+118,0,0,1,1,0,0,10,0,117,0,0,0,1,3,2,90,95,0,0,10,0,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,
+0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,
+0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,
+86,97,108,0,59,120,121,0,0,18,118,0,0,18,119,0,0,0,0,1,90,95,0,0,11,0,2,26,1,1,0,0,11,0,118,0,0,1,
+1,0,0,11,0,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,
+122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,11,0,2,27,1,1,0,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,
+0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,
+122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,11,0,2,21,1,1,0,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,
 0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,
-121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,10,2,22,1,1,0,0,9,97,0,0,1,1,0,0,10,117,
-0,0,0,1,3,2,90,95,0,0,10,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,
-118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,
-59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,
-101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,105,110,118,85,0,59,120,121,0,0,0,0,1,90,95,0,0,10,
-2,22,1,1,0,0,10,118,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,0,0,9,1,105,110,118,66,0,0,0,4,102,108,111,
+121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,11,0,2,22,1,1,0,0,11,0,118,0,0,1,1,0,0,11,0,117,
+0,0,0,1,3,2,90,95,0,0,11,0,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,
+117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,
+4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,
+109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,
+18,119,0,0,0,0,1,90,95,0,0,12,0,2,26,1,1,0,0,12,0,118,0,0,1,1,0,0,12,0,117,0,0,0,1,4,118,101,99,52,
+95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,12,0,2,27,
+1,1,0,0,12,0,118,0,0,1,1,0,0,12,0,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,
+95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,12,0,118,0,
+0,1,1,0,0,12,0,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,
+116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,12,0,2,22,1,1,0,0,12,0,118,0,0,1,1,0,0,12,
+0,117,0,0,0,1,3,2,90,95,0,0,12,0,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,
+0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,
+0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,
+97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,
+116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,119,0,0,0,0,1,90,95,0,0,10,
+0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,
+101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,90,95,0,0,10,0,2,26,1,1,
+0,0,10,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,
+108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,2,27,1,1,0,0,9,0,97,0,
+0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,
+86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,90,95,0,0,10,0,2,27,1,1,0,0,10,
+0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,
+101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,0,10,0,2,21,1,1,
+0,0,9,0,97,0,0,1,1,0,0,10,0,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,
+95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,90,95,0,0,10,0,2,
+21,1,1,0,0,10,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,
+0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,90,95,0,
+0,10,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,10,0,117,0,0,0,1,3,2,90,95,0,0,10,0,1,105,110,118,85,0,0,0,
+4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,
+108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,
+52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,
+18,105,110,118,85,0,59,120,121,0,0,0,0,1,90,95,0,0,10,0,2,22,1,1,0,0,10,0,118,0,0,1,1,0,0,9,0,98,0,
+0,0,1,3,2,90,95,0,0,9,0,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,
+66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,
+108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,66,0,0,0,0,1,90,95,0,0,11,0,2,26,1,1,0,
+0,9,0,97,0,0,1,1,0,0,11,0,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,
+108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0,0,0,0,1,90,95,0,0,11,0,2,26,1,1,0,0,
+11,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,
+0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,2,27,1,1,0,0,9,0,
+97,0,0,1,1,0,0,11,0,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,
+116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0,0,0,0,1,90,95,0,0,11,0,2,27,
+1,1,0,0,11,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,
+95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,
+0,0,11,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,11,0,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,
+112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,
+122,0,0,0,0,1,90,95,0,0,11,0,2,21,1,1,0,0,11,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,
+109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,
+120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,11,0,117,0,0,0,1,3,
+2,90,95,0,0,11,0,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,
+59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,
+0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,
+0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,
+0,59,120,121,122,0,0,18,97,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0,0,1,90,95,0,0,11,0,2,22,1,
+1,0,0,11,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,105,110,118,66,0,0,0,4,102,108,111,
 97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,
-112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,
-118,66,0,0,0,0,1,90,95,0,0,11,2,26,1,1,0,0,9,97,0,0,1,1,0,0,11,117,0,0,0,1,4,118,101,99,52,95,97,
-100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0,
-0,0,0,1,90,95,0,0,11,2,26,1,1,0,0,11,118,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,
-18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,90,
-95,0,0,11,2,27,1,1,0,0,9,97,0,0,1,1,0,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,
-116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0,0,0,
-0,1,90,95,0,0,11,2,27,1,1,0,0,11,118,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,
-114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,
-98,0,0,0,0,1,90,95,0,0,11,2,21,1,1,0,0,9,97,0,0,1,1,0,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,
-108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,
-59,120,121,122,0,0,0,0,1,90,95,0,0,11,2,21,1,1,0,0,11,118,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,
-95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,
-59,120,121,122,0,0,18,98,0,0,0,0,1,90,95,0,0,11,2,22,1,1,0,0,9,97,0,0,1,1,0,0,11,117,0,0,0,1,3,2,
-90,95,0,0,11,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,
-120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,
-18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,
-59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,
-59,120,121,122,0,0,18,97,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0,0,1,90,95,0,0,11,2,22,1,1,0,
-0,11,118,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,0,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,
-114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,
-0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,
-66,0,0,0,0,1,90,95,0,0,12,2,26,1,1,0,0,9,97,0,0,1,1,0,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,
-100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,90,95,0,0,12,2,26,1,1,0,0,12,
-118,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,
-118,0,0,18,98,0,0,0,0,1,90,95,0,0,12,2,27,1,1,0,0,9,97,0,0,1,1,0,0,12,117,0,0,0,1,4,118,101,99,52,
-95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,90,
-95,0,0,12,2,27,1,1,0,0,12,118,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,
-116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1,90,95,0,0,12,2,21,1,1,0,0,9,97,
-0,0,1,1,0,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,
-116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,90,95,0,0,12,2,21,1,1,0,0,12,118,0,0,1,1,0,0,9,98,0,0,
-0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,
-0,0,18,98,0,0,0,0,1,90,95,0,0,12,2,22,1,1,0,0,9,97,0,0,1,1,0,0,12,117,0,0,0,1,3,2,90,95,0,0,12,1,
-105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,
-59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,
-0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,
-102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,
-99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,105,110,
-118,85,0,0,0,0,1,90,95,0,0,12,2,22,1,1,0,0,12,118,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,0,0,9,1,105,
-110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,
-99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,
-110,118,66,0,0,0,0,1,90,95,0,0,6,2,26,1,1,0,0,5,97,0,0,1,1,0,0,6,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,46,20,0,0,1,90,95,0,0,6,2,26,1,1,0,0,6,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,
-98,0,0,0,46,20,0,0,1,90,95,0,0,6,2,27,1,1,0,0,5,97,0,0,1,1,0,0,6,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,47,20,0,0,1,90,95,0,0,6,2,27,1,1,0,0,6,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,
-98,0,0,0,47,20,0,0,1,90,95,0,0,6,2,21,1,1,0,0,5,97,0,0,1,1,0,0,6,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,48,20,0,0,1,90,95,0,0,6,2,21,1,1,0,0,6,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,
-98,0,0,0,48,20,0,0,1,90,95,0,0,6,2,22,1,1,0,0,5,97,0,0,1,1,0,0,6,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,49,20,0,0,1,90,95,0,0,6,2,22,1,1,0,0,6,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,
-98,0,0,0,49,20,0,0,1,90,95,0,0,7,2,26,1,1,0,0,5,97,0,0,1,1,0,0,7,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,46,20,0,0,1,90,95,0,0,7,2,26,1,1,0,0,7,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,
-98,0,0,0,46,20,0,0,1,90,95,0,0,7,2,27,1,1,0,0,5,97,0,0,1,1,0,0,7,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,47,20,0,0,1,90,95,0,0,7,2,27,1,1,0,0,7,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,
-98,0,0,0,47,20,0,0,1,90,95,0,0,7,2,21,1,1,0,0,5,97,0,0,1,1,0,0,7,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,48,20,0,0,1,90,95,0,0,7,2,21,1,1,0,0,7,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,
-98,0,0,0,48,20,0,0,1,90,95,0,0,7,2,22,1,1,0,0,5,97,0,0,1,1,0,0,7,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,49,20,0,0,1,90,95,0,0,7,2,22,1,1,0,0,7,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,
-98,0,0,0,49,20,0,0,1,90,95,0,0,8,2,26,1,1,0,0,5,97,0,0,1,1,0,0,8,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,46,20,0,0,1,90,95,0,0,8,2,26,1,1,0,0,8,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,
-98,0,0,0,46,20,0,0,1,90,95,0,0,8,2,27,1,1,0,0,5,97,0,0,1,1,0,0,8,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,47,20,0,0,1,90,95,0,0,8,2,27,1,1,0,0,8,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,
-98,0,0,0,47,20,0,0,1,90,95,0,0,8,2,21,1,1,0,0,5,97,0,0,1,1,0,0,8,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,48,20,0,0,1,90,95,0,0,8,2,21,1,1,0,0,8,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,
-98,0,0,0,48,20,0,0,1,90,95,0,0,8,2,22,1,1,0,0,5,97,0,0,1,1,0,0,8,117,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,49,20,0,0,1,90,95,0,0,8,2,22,1,1,0,0,8,
-118,0,0,1,1,0,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,
-98,0,0,0,49,20,0,0,1,90,95,0,0,5,2,27,1,1,0,0,5,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,
-101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,90,95,0,0,6,2,27,1,1,0,0,6,118,0,
+112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,
+105,110,118,66,0,0,0,0,1,90,95,0,0,12,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,12,0,117,0,0,0,1,4,118,101,
+99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,90,95,0,0,12,0,
+2,26,1,1,0,0,12,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,
+116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,12,0,
+117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,
+18,97,0,0,18,117,0,0,0,0,1,90,95,0,0,12,0,2,27,1,1,0,0,12,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,
+101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,
+0,0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,12,0,117,0,0,0,1,4,118,101,99,52,95,109,117,
+108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,90,95,0,0,
+12,0,2,21,1,1,0,0,12,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,
+108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1,90,95,0,0,12,0,2,22,1,1,0,
+0,9,0,97,0,0,1,1,0,0,12,0,117,0,0,0,1,3,2,90,95,0,0,12,0,1,105,110,118,85,0,0,0,4,102,108,111,97,
+116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,
+114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,
+112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,
+105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,
+121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,105,110,118,85,0,0,0,0,1,90,95,0,0,12,0,2,22,
+1,1,0,0,12,0,118,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,9,0,1,105,110,118,66,0,0,0,4,102,108,111,
+97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,
+112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,110,118,66,0,0,0,0,1,90,95,0,0,
+6,0,2,26,1,1,0,0,5,0,97,0,0,1,1,0,0,6,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,
+101,99,50,0,0,18,97,0,0,0,18,117,0,46,20,0,0,1,90,95,0,0,6,0,2,26,1,1,0,0,6,0,118,0,0,1,1,0,0,5,0,
+98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,0,18,98,0,0,0,46,20,
+0,0,1,90,95,0,0,6,0,2,27,1,1,0,0,5,0,97,0,0,1,1,0,0,6,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,58,105,118,101,99,50,0,0,18,97,0,0,0,18,117,0,47,20,0,0,1,90,95,0,0,6,0,2,27,1,1,0,0,6,0,118,
+0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,0,18,
+98,0,0,0,47,20,0,0,1,90,95,0,0,6,0,2,21,1,1,0,0,5,0,97,0,0,1,1,0,0,6,0,117,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,58,105,118,101,99,50,0,0,18,97,0,0,0,18,117,0,48,20,0,0,1,90,95,0,0,6,0,2,21,1,
+1,0,0,6,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,
+101,99,50,0,0,18,98,0,0,0,48,20,0,0,1,90,95,0,0,6,0,2,22,1,1,0,0,5,0,97,0,0,1,1,0,0,6,0,117,0,0,0,
+1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,0,18,97,0,0,0,18,117,0,49,20,0,0,1,90,
+95,0,0,6,0,2,22,1,1,0,0,6,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,
+118,0,58,105,118,101,99,50,0,0,18,98,0,0,0,49,20,0,0,1,90,95,0,0,7,0,2,26,1,1,0,0,5,0,97,0,0,1,1,0,
+0,7,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,0,18,97,0,0,0,18,117,0,
+46,20,0,0,1,90,95,0,0,7,0,2,26,1,1,0,0,7,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,
+86,97,108,0,18,118,0,58,105,118,101,99,51,0,0,18,98,0,0,0,46,20,0,0,1,90,95,0,0,7,0,2,27,1,1,0,0,5,
+0,97,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,0,18,97,
+0,0,0,18,117,0,47,20,0,0,1,90,95,0,0,7,0,2,27,1,1,0,0,7,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,
+95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,0,18,98,0,0,0,47,20,0,0,1,90,95,0,0,7,0,
+2,21,1,1,0,0,5,0,97,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,
+99,51,0,0,18,97,0,0,0,18,117,0,48,20,0,0,1,90,95,0,0,7,0,2,21,1,1,0,0,7,0,118,0,0,1,1,0,0,5,0,98,0,
+0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,0,18,98,0,0,0,48,20,0,0,1,
+90,95,0,0,7,0,2,22,1,1,0,0,5,0,97,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
+58,105,118,101,99,51,0,0,18,97,0,0,0,18,117,0,49,20,0,0,1,90,95,0,0,7,0,2,22,1,1,0,0,7,0,118,0,0,1,
+1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,0,18,98,0,
+0,0,49,20,0,0,1,90,95,0,0,8,0,2,26,1,1,0,0,5,0,97,0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,95,95,114,101,
+116,86,97,108,0,58,105,118,101,99,52,0,0,18,97,0,0,0,18,117,0,46,20,0,0,1,90,95,0,0,8,0,2,26,1,1,0,
+0,8,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,
+52,0,0,18,98,0,0,0,46,20,0,0,1,90,95,0,0,8,0,2,27,1,1,0,0,5,0,97,0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,
+95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,0,18,97,0,0,0,18,117,0,47,20,0,0,1,90,95,0,0,
+8,0,2,27,1,1,0,0,8,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,
+105,118,101,99,52,0,0,18,98,0,0,0,47,20,0,0,1,90,95,0,0,8,0,2,21,1,1,0,0,5,0,97,0,0,1,1,0,0,8,0,
+117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,0,18,97,0,0,0,18,117,0,48,20,
+0,0,1,90,95,0,0,8,0,2,21,1,1,0,0,8,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,18,118,0,58,105,118,101,99,52,0,0,18,98,0,0,0,48,20,0,0,1,90,95,0,0,8,0,2,22,1,1,0,0,5,0,97,
+0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,0,18,97,0,0,
+0,18,117,0,49,20,0,0,1,90,95,0,0,8,0,2,22,1,1,0,0,8,0,118,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,0,18,98,0,0,0,49,20,0,0,1,90,95,0,0,5,0,2,
+27,1,1,0,0,5,0,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,
+108,0,59,120,0,0,18,97,0,0,0,0,1,90,95,0,0,6,0,2,27,1,1,0,0,6,0,118,0,0,0,1,4,118,101,99,52,95,110,
+101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,90,95,0,0,7,0,2,27,1,1,0,
+0,7,0,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,
+18,118,0,0,0,0,1,90,95,0,0,8,0,2,27,1,1,0,0,8,0,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,
+101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,90,95,0,0,9,0,2,27,1,1,0,0,9,0,97,0,0,0,
+1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,
+0,0,0,1,90,95,0,0,10,0,2,27,1,1,0,0,10,0,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,
+18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,90,95,0,0,11,0,2,27,
+1,1,0,0,11,0,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,
+108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,0,0,1,90,95,0,0,12,0,2,27,1,1,0,0,12,0,118,0,
 0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,
-0,1,90,95,0,0,7,2,27,1,1,0,0,7,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,
-114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,90,95,0,0,8,2,27,1,1,0,0,8,118,0,0,0,1,4,118,101,99,52,
-95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,90,95,0,0,9,2,27,1,
-1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,
-120,0,0,18,97,0,0,0,0,1,90,95,0,0,10,2,27,1,1,0,0,10,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,
-116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,90,95,0,0,
-11,2,27,1,1,0,0,11,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,
-97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,0,0,1,90,95,0,0,12,2,27,1,1,0,0,12,118,0,0,
-0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,
-1,90,95,0,0,13,2,27,1,1,0,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,
-0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,
-20,0,0,1,90,95,0,0,14,2,27,1,1,0,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,
+0,1,90,95,0,0,13,0,2,27,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,
 18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,
-57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,0,1,90,
-95,0,0,15,2,27,1,1,0,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,
-8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,
-9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,9,18,95,95,114,101,
-116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,54,20,0,0,1,90,95,0,0,9,0,100,111,116,0,1,1,0,
-0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,48,20,0,0,1,90,
-95,0,0,9,0,100,111,116,0,1,1,0,0,10,97,0,0,1,1,0,0,10,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
-0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0,59,121,0,18,98,0,59,121,0,48,46,20,0,0,1,90,95,0,0,
-9,0,100,111,116,0,1,1,0,0,11,97,0,0,1,1,0,0,11,98,0,0,0,1,4,118,101,99,51,95,100,111,116,0,18,95,
-95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,9,0,100,111,116,0,1,1,0,0,12,97,0,
-0,1,1,0,0,12,98,0,0,0,1,4,118,101,99,52,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,
-0,0,18,98,0,0,0,0,1,90,95,0,0,0,2,1,1,0,2,0,5,97,0,0,1,1,0,0,5,98,0,0,0,1,4,118,101,99,52,95,97,
-100,100,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,0,2,2,1,0,2,0,5,97,0,0,1,1,0,0,5,98,0,0,0,
-1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,
-0,2,3,1,0,2,0,5,97,0,0,1,1,0,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,
-18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,0,2,4,1,0,2,0,5,97,0,0,1,1,0,0,5,98,0,0,0,1,3,2,90,
-95,0,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,
-0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,105,110,118,66,0,0,
-0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,97,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,1,1,0,
-2,0,6,118,0,0,1,1,0,0,6,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,
-0,0,0,1,90,95,0,0,0,2,2,1,0,2,0,6,118,0,0,1,1,0,0,6,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,
-114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,6,118,0,0,1,1,0,0,6,
-117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,
-0,0,1,90,95,0,0,0,2,4,1,0,2,0,6,118,0,0,1,1,0,0,6,117,0,0,0,1,3,2,90,95,0,0,6,1,105,110,118,0,0,1,
-1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,
-4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,
-99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,118,101,99,
-52,95,116,111,95,105,118,101,99,52,0,18,118,0,0,18,122,0,0,0,0,1,90,95,0,0,0,2,1,1,0,2,0,7,118,0,0,
-1,1,0,0,7,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,
-0,0,0,2,2,1,0,2,0,7,118,0,0,1,1,0,0,7,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,
-0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,7,118,0,0,1,1,0,0,7,117,0,0,0,1,4,
-118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,
-0,0,2,4,1,0,2,0,7,118,0,0,1,1,0,0,7,117,0,0,0,1,3,2,90,95,0,0,7,1,105,110,118,0,0,1,1,122,0,0,0,4,
-102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,
-97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,
-117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,118,101,99,52,95,116,
-111,95,105,118,101,99,52,0,18,118,0,0,18,122,0,0,0,0,1,90,95,0,0,0,2,1,1,0,2,0,8,118,0,0,1,1,0,0,8,
-117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,0,2,2,
-1,0,2,0,8,118,0,0,1,1,0,0,8,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,
-0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,8,118,0,0,1,1,0,0,8,117,0,0,0,1,4,118,101,99,
-52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,0,2,4,1,0,
-2,0,8,118,0,0,1,1,0,0,8,117,0,0,0,1,3,2,90,95,0,0,8,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,
-97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,
-114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,
-105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,
-101,99,52,0,18,118,0,0,18,122,0,0,0,0,1,90,95,0,0,0,2,1,1,0,2,0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,4,
-118,101,99,52,95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,0,1,90,95,
-0,0,0,2,2,1,0,2,0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,
-18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,
-4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,
-90,95,0,0,0,2,4,1,0,2,0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,0,0,9,1,119,0,0,0,4,102,108,111,97,
-116,95,114,99,112,0,18,119,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,
-121,0,18,97,0,59,120,0,0,18,97,0,0,18,119,0,0,0,0,1,90,95,0,0,0,2,1,1,0,2,0,10,118,0,0,1,1,0,0,10,
-117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,
-0,59,120,121,0,0,0,0,1,90,95,0,0,0,2,2,1,0,2,0,10,118,0,0,1,1,0,0,10,117,0,0,0,1,4,118,101,99,52,
-95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,
-121,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,10,118,0,0,1,1,0,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,
-108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,
-0,0,1,90,95,0,0,0,2,4,1,0,2,0,10,118,0,0,1,1,0,0,10,117,0,0,0,1,3,2,90,95,0,0,10,1,119,0,0,0,4,102,
-108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,
-114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,
-108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,119,0,59,120,121,0,0,0,0,1,90,95,0,0,
-0,2,1,1,0,2,0,11,118,0,0,1,1,0,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,
-121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,0,2,2,1,0,2,0,11,118,0,0,1,1,0,0,11,117,0,0,0,1,
-4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,
-0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,11,118,0,0,1,1,0,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,
-116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,0,2,4,1,0,
-2,0,11,118,0,0,1,1,0,0,11,117,0,0,0,1,3,2,90,95,0,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,
-112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,
-121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,
-122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,
-0,59,120,121,122,0,0,18,119,0,59,120,121,122,0,0,0,0,1,90,95,0,0,0,2,1,1,0,2,0,12,118,0,0,1,1,0,0,
-12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,0,
-2,2,1,0,2,0,12,118,0,0,1,1,0,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,
-118,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,12,118,0,0,1,1,0,0,12,117,0,0,0,1,4,
-118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,90,95,0,
-0,0,2,4,1,0,2,0,12,118,0,0,1,1,0,0,12,117,0,0,0,1,3,2,90,95,0,0,12,1,119,0,0,0,4,102,108,111,97,
-116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,
-18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,
-18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,
-0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,119,0,0,0,0,1,90,
-95,0,0,0,2,1,1,0,2,0,6,118,0,0,1,1,0,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,
-120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,2,1,0,2,0,6,118,0,0,1,1,0,0,5,97,
-0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120,
-121,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,6,118,0,0,1,1,0,0,5,97,0,0,0,1,4,118,101,99,52,95,
-109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,0,0,9,18,
-118,0,59,120,0,18,97,0,23,0,9,18,118,0,59,121,0,18,97,0,23,0,0,1,90,95,0,0,0,2,4,1,0,2,0,6,118,0,0,
-1,1,0,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,0,1,90,95,0,
-0,0,2,1,1,0,2,0,7,118,0,0,1,1,0,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,
-122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,2,1,0,2,0,7,118,0,0,1,1,0,0,5,97,
-0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,59,
-120,121,122,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,7,118,0,0,1,1,0,0,5,97,0,0,0,1,4,118,101,
-99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,
-18,97,0,0,0,0,1,90,95,0,0,0,2,4,1,0,2,0,7,118,0,0,1,1,0,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,
-24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,0,1,90,95,0,0,0,2,1,1,0,2,0,
-8,118,0,0,1,1,0,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,
-1,90,95,0,0,0,2,2,1,0,2,0,8,118,0,0,1,1,0,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,
-99,116,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,8,118,0,0,1,1,0,0,5,97,0,0,
-0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,90,
-95,0,0,0,2,4,1,0,2,0,8,118,0,0,1,1,0,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,
-121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,9,18,118,0,59,119,0,18,97,0,24,0,0,1,90,95,0,0,
-0,2,1,1,0,2,0,10,118,0,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,
-0,0,18,118,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,2,1,0,2,0,10,118,0,0,1,1,0,0,9,97,0,0,0,1,4,118,101,
-99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,0,0,0,1,90,95,0,
-0,0,2,3,1,0,2,0,10,118,0,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,
-0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,4,1,0,2,0,10,118,0,0,1,1,0,0,9,
-97,0,0,0,1,3,2,90,95,0,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,
-118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,
-0,18,118,0,59,120,121,0,0,18,105,110,118,65,0,0,0,0,1,90,95,0,0,0,2,1,1,0,2,0,11,118,0,0,1,1,0,0,9,
-97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,0,0,0,1,
-90,95,0,0,0,2,2,1,0,2,0,11,118,0,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,
-99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,11,118,0,0,
-1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,
-0,0,18,118,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,4,1,0,2,0,11,118,0,0,1,1,0,0,9,97,0,0,0,1,3,2,90,95,0,
-0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,
-4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,
-121,122,0,0,18,105,110,118,65,0,0,0,0,1,90,95,0,0,0,2,1,1,0,2,0,12,118,0,0,1,1,0,0,9,97,0,0,0,1,4,
-118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,2,1,0,2,0,12,118,
-0,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,
-18,97,0,0,0,0,1,90,95,0,0,0,2,3,1,0,2,0,12,118,0,0,1,1,0,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,
-108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,90,95,0,0,0,2,4,1,0,2,0,12,118,0,0,
-1,1,0,0,9,97,0,0,0,1,3,2,90,95,0,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,
-18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,
-18,118,0,0,18,105,110,118,65,0,0,0,0,1,90,95,0,0,13,2,26,1,1,0,0,13,109,0,0,1,1,0,0,13,110,0,0,0,1,
-9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,
-0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,
-46,20,0,0,1,90,95,0,0,13,2,27,1,1,0,0,13,109,0,0,1,1,0,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,
-97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,90,95,0,0,13,2,
-21,1,1,0,0,13,109,0,0,1,1,0,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
-109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,
-57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,
-18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,0,
-48,46,20,0,0,1,90,95,0,0,13,2,22,1,1,0,0,13,109,0,0,1,1,0,0,13,110,0,0,0,1,9,18,95,95,114,101,116,
-86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,
-116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,0,1,90,95,0,0,
-14,2,26,1,1,0,0,14,109,0,0,1,1,0,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,
-18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
-57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,
-50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,90,95,0,0,14,2,27,1,1,0,0,14,109,
-0,0,1,1,0,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,
-18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,
-57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,
-50,0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,90,95,0,0,14,2,21,1,1,0,0,14,109,0,0,1,1,0,0,14,110,0,0,
-0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,
-120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,
-10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
-10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,
-18,110,0,16,10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,
-122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,
-0,16,10,50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0,
-48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,90,95,0,0,14,2,
-22,1,1,0,0,14,109,0,0,1,1,0,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
-109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,
-18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,
-0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,90,95,0,0,15,2,26,1,1,0,0,15,109,0,
-0,1,1,0,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,
-110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,
-18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,
-0,57,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,
-10,51,0,57,18,110,0,16,10,51,0,57,46,20,0,0,1,90,95,0,0,15,2,27,1,1,0,0,15,109,0,0,1,1,0,0,15,110,
+57,54,20,0,0,1,90,95,0,0,14,0,2,27,1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,
+8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,
+16,10,49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,
+20,0,0,1,90,95,0,0,15,0,2,27,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,
+57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,
+49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,9,
+18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,54,20,0,0,1,90,95,0,0,9,0,0,
+100,111,116,0,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,
+18,98,0,48,20,0,0,1,90,95,0,0,9,0,0,100,111,116,0,1,1,0,0,10,0,97,0,0,1,1,0,0,10,0,98,0,0,0,1,9,18,
+95,95,114,101,116,86,97,108,0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0,59,121,0,18,98,0,59,121,
+0,48,46,20,0,0,1,90,95,0,0,9,0,0,100,111,116,0,1,1,0,0,11,0,97,0,0,1,1,0,0,11,0,98,0,0,0,1,4,118,
+101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,
+9,0,0,100,111,116,0,1,1,0,0,12,0,97,0,0,1,1,0,0,12,0,98,0,0,0,1,4,118,101,99,52,95,100,111,116,0,
+18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,90,95,0,0,5,0,2,1,1,0,2,0,5,0,97,0,0,
+1,1,0,0,5,0,98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,46,20,0,8,18,97,0,0,0,1,90,95,0,0,5,0,2,2,1,0,2,0,
+5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,47,20,0,8,18,97,0,0,0,1,90,95,0,0,5,0,
+2,3,1,0,2,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,48,20,0,8,18,97,0,0,0,1,90,
+95,0,0,5,0,2,4,1,0,2,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,49,20,0,8,18,97,
+0,0,0,1,90,95,0,0,6,0,2,1,1,0,2,0,6,0,118,0,0,1,1,0,0,6,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,
+46,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,2,1,0,2,0,6,0,118,0,0,1,1,0,0,6,0,117,0,0,0,1,9,18,118,0,
+18,118,0,18,117,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,3,1,0,2,0,6,0,118,0,0,1,1,0,0,6,0,117,0,
+0,0,1,9,18,118,0,18,118,0,18,117,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,4,1,0,2,0,6,0,118,0,0,
+1,1,0,0,6,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,7,0,2,1,1,
+0,2,0,7,0,118,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,46,20,0,8,18,118,0,0,0,1,90,
+95,0,0,7,0,2,2,1,0,2,0,7,0,118,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,47,20,0,8,
+18,118,0,0,0,1,90,95,0,0,7,0,2,3,1,0,2,0,7,0,118,0,0,1,1,0,0,7,0,117,0,0,0,1,9,18,118,0,18,118,0,
+18,117,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,7,0,2,4,1,0,2,0,7,0,118,0,0,1,1,0,0,7,0,117,0,0,0,1,9,
+18,118,0,18,118,0,18,117,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,8,0,2,1,1,0,2,0,8,0,118,0,0,1,1,0,0,
+8,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,8,0,2,2,1,0,2,0,8,
+0,118,0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,
+8,0,2,3,1,0,2,0,8,0,118,0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,48,20,0,8,18,118,
+0,0,0,1,90,95,0,0,8,0,2,4,1,0,2,0,8,0,118,0,0,1,1,0,0,8,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,
+49,20,0,8,18,118,0,0,0,1,90,95,0,0,9,0,2,1,1,0,2,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,97,0,18,
+97,0,18,98,0,46,20,0,8,18,97,0,0,0,1,90,95,0,0,9,0,2,2,1,0,2,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,9,
+18,97,0,18,97,0,18,98,0,47,20,0,8,18,97,0,0,0,1,90,95,0,0,9,0,2,3,1,0,2,0,9,0,97,0,0,1,1,0,0,9,0,
+98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,48,20,0,8,18,97,0,0,0,1,90,95,0,0,9,0,2,4,1,0,2,0,9,0,97,0,0,
+1,1,0,0,9,0,98,0,0,0,1,9,18,97,0,18,97,0,18,98,0,49,20,0,8,18,97,0,0,0,1,90,95,0,0,10,0,2,1,1,0,2,
+0,10,0,118,0,0,1,1,0,0,10,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,46,20,0,8,18,118,0,0,0,1,90,
+95,0,0,10,0,2,2,1,0,2,0,10,0,118,0,0,1,1,0,0,10,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,47,20,0,
+8,18,118,0,0,0,1,90,95,0,0,10,0,2,3,1,0,2,0,10,0,118,0,0,1,1,0,0,10,0,117,0,0,0,1,9,18,118,0,18,
+118,0,18,117,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,10,0,2,4,1,0,2,0,10,0,118,0,0,1,1,0,0,10,0,117,0,
+0,0,1,9,18,118,0,18,118,0,18,117,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,11,0,2,1,1,0,2,0,11,0,118,0,
+0,1,1,0,0,11,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,11,0,2,
+2,1,0,2,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,47,20,0,8,18,118,0,0,
+0,1,90,95,0,0,11,0,2,3,1,0,2,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,
+48,20,0,8,18,118,0,0,0,1,90,95,0,0,11,0,2,4,1,0,2,0,11,0,118,0,0,1,1,0,0,11,0,117,0,0,0,1,9,18,118,
+0,18,118,0,18,117,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,12,0,2,1,1,0,2,0,12,0,118,0,0,1,1,0,0,12,0,
+117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,12,0,2,2,1,0,2,0,12,0,
+118,0,0,1,1,0,0,12,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,
+12,0,2,3,1,0,2,0,12,0,118,0,0,1,1,0,0,12,0,117,0,0,0,1,9,18,118,0,18,118,0,18,117,0,48,20,0,8,18,
+118,0,0,0,1,90,95,0,0,12,0,2,4,1,0,2,0,12,0,118,0,0,1,1,0,0,12,0,117,0,0,0,1,9,18,118,0,18,118,0,
+18,117,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,1,1,0,2,0,6,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,
+18,118,0,18,118,0,58,105,118,101,99,50,0,0,18,97,0,0,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,2,
+1,0,2,0,6,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,0,18,97,0,0,
+0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,3,1,0,2,0,6,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,
+18,118,0,58,105,118,101,99,50,0,0,18,97,0,0,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,6,0,2,4,1,0,2,0,6,
+0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,0,18,97,0,0,0,49,20,0,
+8,18,118,0,0,0,1,90,95,0,0,7,0,2,1,1,0,2,0,7,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,
+58,105,118,101,99,51,0,0,18,97,0,0,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,7,0,2,2,1,0,2,0,7,0,118,0,
+0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,0,18,97,0,0,0,47,20,0,8,18,118,
+0,0,0,1,90,95,0,0,7,0,2,3,1,0,2,0,7,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,
+118,101,99,51,0,0,18,97,0,0,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,8,0,2,4,1,0,2,0,7,0,118,0,0,1,1,0,
+0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,0,18,97,0,0,0,49,20,0,8,18,118,0,0,0,1,
+90,95,0,0,8,0,2,1,1,0,2,0,8,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,
+52,0,0,18,97,0,0,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,8,0,2,2,1,0,2,0,8,0,118,0,0,1,1,0,0,5,0,97,0,
+0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,0,18,97,0,0,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,
+8,0,2,3,1,0,2,0,8,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,0,18,
+97,0,0,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,8,0,2,4,1,0,2,0,8,0,118,0,0,1,1,0,0,5,0,97,0,0,0,1,9,
+18,118,0,18,118,0,58,105,118,101,99,52,0,0,18,97,0,0,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,10,0,2,1,
+1,0,2,0,10,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,0,18,97,0,0,0,
+46,20,0,8,18,118,0,0,0,1,90,95,0,0,10,0,2,2,1,0,2,0,10,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,
+18,118,0,58,118,101,99,50,0,0,18,97,0,0,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,10,0,2,3,1,0,2,0,10,0,
+118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,0,18,97,0,0,0,48,20,0,8,18,
+118,0,0,0,1,90,95,0,0,10,0,2,4,1,0,2,0,10,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,
+118,101,99,50,0,0,18,97,0,0,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,11,0,2,1,1,0,2,0,11,0,118,0,0,1,1,
+0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,0,18,97,0,0,0,46,20,0,8,18,118,0,0,0,1,
+90,95,0,0,11,0,2,2,1,0,2,0,11,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,
+51,0,0,18,97,0,0,0,47,20,0,8,18,118,0,0,0,1,90,95,0,0,11,0,2,3,1,0,2,0,11,0,118,0,0,1,1,0,0,9,0,97,
+0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,0,18,97,0,0,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,11,
+0,2,4,1,0,2,0,11,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,0,18,97,0,
+0,0,49,20,0,8,18,118,0,0,0,1,90,95,0,0,12,0,2,1,1,0,2,0,12,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,
+118,0,18,118,0,58,118,101,99,52,0,0,18,97,0,0,0,46,20,0,8,18,118,0,0,0,1,90,95,0,0,12,0,2,2,1,0,2,
+0,12,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,0,18,97,0,0,0,47,20,0,
+8,18,118,0,0,0,1,90,95,0,0,12,0,2,3,1,0,2,0,12,0,118,0,0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,
+0,58,118,101,99,52,0,0,18,97,0,0,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,12,0,2,4,1,0,2,0,12,0,118,0,
+0,1,1,0,0,9,0,97,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,0,18,97,0,0,0,49,20,0,8,18,118,0,0,
+0,1,90,95,0,0,13,0,2,26,1,1,0,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,
+97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,90,95,0,0,13,0,2,
+27,1,1,0,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,
+18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
+57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,90,95,0,0,13,0,2,21,1,1,0,0,13,0,109,
+0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,
+18,110,0,16,8,48,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48,
+46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,
+57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,90,95,
+0,0,13,0,2,22,1,1,0,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,
+8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,
+16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,0,1,90,95,0,0,14,0,2,26,1,1,0,
+0,14,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,
+16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,
+0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
+109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,90,95,0,0,14,0,2,27,1,1,0,0,14,0,109,0,0,1,
+1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,
+110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,
+18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,
+0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,90,95,0,0,14,0,2,21,1,1,0,0,14,0,109,0,0,1,1,0,0,14,0,110,
 0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,
-57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,
-49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,
-16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,
-110,0,16,10,51,0,57,47,20,0,0,1,90,95,0,0,15,2,21,1,1,0,0,15,109,0,0,1,1,0,0,15,110,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,
-120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,121,0,48,46,18,109,0,16,
-10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,8,
-48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,
-16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,
-49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122,
-122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,
-95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120,
-120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,121,0,48,46,18,109,0,16,
-10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,
-50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,
-16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,
-51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,51,0,57,59,122,122,122,
-122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,59,119,119,119,119,0,48,46,20,0,0,1,90,
-95,0,0,15,2,22,1,1,0,0,15,109,0,0,1,1,0,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,
-48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
-10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,
-0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49,20,0,0,1,90,95,0,0,13,2,26,1,
-1,0,0,9,97,0,0,1,1,0,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,
-110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,
-49,0,57,46,20,0,0,1,90,95,0,0,13,2,26,1,1,0,0,13,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,
-108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,0,1,90,95,0,0,13,2,27,1,1,0,0,9,97,0,0,
-1,1,0,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,
-57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,
-0,1,90,95,0,0,13,2,27,1,1,0,0,13,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
-16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
-57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,0,1,90,95,0,0,13,2,21,1,1,0,0,9,97,0,0,1,1,0,0,13,110,0,
-0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,
-13,2,21,1,1,0,0,13,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
-109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,
-49,0,57,18,98,0,48,20,0,0,1,90,95,0,0,13,2,22,1,1,0,0,9,97,0,0,1,1,0,0,13,110,0,0,0,1,9,18,95,95,
-114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,
-86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,0,1,90,95,0,0,13,2,22,1,1,0,0,13,
-109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,
-18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,
-20,0,0,1,90,95,0,0,14,2,26,1,1,0,0,9,97,0,0,1,1,0,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,
-49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
-97,0,18,110,0,16,10,50,0,57,46,20,0,0,1,90,95,0,0,14,2,26,1,1,0,0,14,109,0,0,1,1,0,0,9,98,0,0,0,1,
-9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,
-114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,
-116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,90,95,0,0,14,2,27,1,1,0,0,
-9,97,0,0,1,1,0,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,
-16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,
-57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,
-0,1,90,95,0,0,14,2,27,1,1,0,0,14,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
-16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
-57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,
-0,16,10,50,0,57,18,98,0,47,20,0,0,1,90,95,0,0,14,2,21,1,1,0,0,9,97,0,0,1,1,0,0,14,110,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,
-97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,14,2,21,1,1,0,0,14,109,
-0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,
-98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,
-0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,90,
-95,0,0,14,2,22,1,1,0,0,9,97,0,0,1,1,0,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,
+57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109,
+0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,
+0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,
+0,57,18,110,0,16,10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,
+59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,
+18,110,0,16,10,50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,
+121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,90,
+95,0,0,14,0,2,22,1,1,0,0,14,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
+16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,
+0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,90,95,0,0,15,0,2,26,
+1,1,0,0,15,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
+109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,
+18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,
+0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
+10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,20,0,0,1,90,95,0,0,15,0,2,27,1,1,0,0,
+15,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,
+8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,
+16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
+109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,
+57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,90,95,0,0,15,0,2,21,1,1,0,0,15,0,109,
+0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,
+18,110,0,16,8,48,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,
+121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0,48,46,18,109,
+0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120,0,48,18,109,0,
+16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,
+10,49,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57,59,119,119,119,
+119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,
+16,10,50,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,
+121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48,46,18,109,0,16,
+10,51,0,57,18,110,0,16,10,50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,
+0,16,10,51,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120,0,48,18,109,0,16,
+10,49,0,57,18,110,0,16,10,51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,
+51,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,59,119,119,119,
+119,0,48,46,20,0,0,1,90,95,0,0,15,0,2,22,1,1,0,0,15,0,109,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,
+114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,
+95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,
+18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,
+0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,
+49,20,0,0,1,90,95,0,0,13,0,2,26,1,1,0,0,9,0,97,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,
+86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,
+16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,0,1,90,95,0,0,13,0,2,26,1,1,0,0,13,0,109,0,0,
+1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,
+0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,
+0,1,90,95,0,0,13,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,
+49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,90,95,0,0,13,0,2,27,1,1,0,0,13,0,109,0,0,1,1,0,
+0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,
+20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,0,1,
+90,95,0,0,13,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
+16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
+57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,90,95,0,0,13,0,2,21,1,1,0,0,13,0,109,0,0,1,1,0,0,9,0,
+98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,
+18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,90,95,0,
+0,13,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,
 0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,
-0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,
-16,10,50,0,57,49,20,0,0,1,90,95,0,0,14,2,22,1,1,0,0,14,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,
-101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,
-97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,
-16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,90,95,0,0,15,2,26,1,1,0,0,9,97,0,0,1,1,0,
-0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,
-20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,90,95,0,0,15,2,26,1,1,
-0,0,15,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,
-48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,
-98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,
-0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,90,
-95,0,0,15,2,27,1,1,0,0,9,97,0,0,1,1,0,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,
-0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,
-0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,
-16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,
-57,47,20,0,0,1,90,95,0,0,15,2,27,1,1,0,0,15,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,
-97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
-10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,
-18,109,0,16,10,50,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,
-16,10,51,0,57,18,98,0,47,20,0,0,1,90,95,0,0,15,2,21,1,1,0,0,9,97,0,0,1,1,0,0,15,110,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,
+0,18,110,0,16,10,49,0,57,49,20,0,0,1,90,95,0,0,13,0,2,22,1,1,0,0,13,0,109,0,0,1,1,0,0,9,0,98,0,0,0,
+1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,0,1,90,95,0,0,14,0,2,
+26,1,1,0,0,9,0,97,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
+97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,
+0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,
+0,57,46,20,0,0,1,90,95,0,0,14,0,2,26,1,1,0,0,14,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,
+97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,
+16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,90,95,0,0,14,0,2,27,1,1,0,0,9,0,97,0,0,1,
+1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,
+57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,
+9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,90,95,
+0,0,14,0,2,27,1,1,0,0,14,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,
+48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,
+109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,
+10,50,0,57,18,98,0,47,20,0,0,1,90,95,0,0,14,0,2,21,1,1,0,0,9,0,97,0,0,1,1,0,0,14,0,110,0,0,0,1,9,
+18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,
 101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,
-97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,
-16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0,1,90,95,0,0,15,2,21,1,1,0,0,15,109,0,0,1,1,
-0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,
-20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,
-95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,
-101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,1,90,95,0,0,15,2,22,1,1,
-0,0,9,97,0,0,1,1,0,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,
-0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,
-0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49,20,
-0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,49,20,0,0,1,90,
-95,0,0,15,2,22,1,1,0,0,15,109,0,0,1,1,0,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,
-0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,
-0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,
-0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,
-0,49,20,0,0,1,90,95,0,0,10,2,21,1,1,0,0,13,109,0,0,1,1,0,0,10,118,0,0,0,1,9,18,95,95,114,101,116,
-86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,
-121,0,48,46,20,0,0,1,90,95,0,0,10,2,21,1,1,0,0,10,118,0,0,1,1,0,0,13,109,0,0,0,1,9,18,95,95,114,
-101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,
-114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,
-90,95,0,0,11,2,21,1,1,0,0,14,109,0,0,1,1,0,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,
-109,0,16,8,48,0,57,18,118,0,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,
-48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122,122,0,48,46,20,0,0,1,90,95,0,0,11,2,21,1,1,0,0,11,
-118,0,0,1,1,0,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,
-0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,
+97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,90,95,0,0,14,0,2,21,1,1,0,0,14,0,
+109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,
+57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,
+48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,
+1,90,95,0,0,14,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
+0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
+57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,
+18,110,0,16,10,50,0,57,49,20,0,0,1,90,95,0,0,14,0,2,22,1,1,0,0,14,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,
+9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,90,95,0,0,15,0,2,26,1,1,0,
+0,9,0,97,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,
+110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,
+49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,
+20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,
+90,95,0,0,15,0,2,26,1,1,0,0,15,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
+16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,
+57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,
+0,16,10,50,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,
+0,57,18,98,0,46,20,0,0,1,90,95,0,0,15,0,2,27,1,1,0,0,9,0,97,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,
+95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,
+116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,
+108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,
+10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,90,95,0,0,15,0,2,27,1,1,0,0,15,0,109,0,0,1,1,
+0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,
+47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,
+18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0,0,1,90,95,0,0,15,0,2,
+21,1,1,0,0,9,0,97,0,0,1,1,0,0,15,0,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,
+97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,
+0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,
+0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,
+0,0,1,90,95,0,0,15,0,2,21,1,1,0,0,15,0,109,0,0,1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,
+49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,
+109,0,16,10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,
+10,51,0,57,18,98,0,48,20,0,0,1,90,95,0,0,15,0,2,22,1,1,0,0,9,0,97,0,0,1,1,0,0,15,0,110,0,0,0,1,9,
+18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,
+101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,
+97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,
+16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,49,20,0,0,1,90,95,0,0,15,0,2,22,1,1,0,0,15,0,109,0,0,
+1,1,0,0,9,0,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,
+0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,
+9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,9,18,95,95,
+114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,49,20,0,0,1,90,95,0,0,10,0,2,
+21,1,1,0,0,13,0,109,0,0,1,1,0,0,10,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,
+48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,0,48,46,20,0,0,1,90,95,
+0,0,10,0,2,21,1,1,0,0,10,0,118,0,0,1,1,0,0,13,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,
+120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,
+108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,90,95,0,0,11,0,2,
+21,1,1,0,0,14,0,109,0,0,1,1,0,0,11,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,
+48,0,57,18,118,0,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,
+0,16,10,50,0,57,18,118,0,59,122,122,122,0,48,46,20,0,0,1,90,95,0,0,11,0,2,21,1,1,0,0,11,0,118,0,0,
+1,1,0,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,
+18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,0,18,
 118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,
-0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,12,2,21,1,1,0,0,15,109,0,0,1,1,0,0,12,
-118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,120,0,
-48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,
-122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,118,0,59,119,119,119,119,0,48,46,20,0,0,1,90,95,
-0,0,12,2,21,1,1,0,0,12,118,0,0,1,1,0,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,
-58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,
-121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,
-0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,18,95,95,114,101,116,86,
-97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,20,0,0,1,90,95,0,0,0,2,1,
-1,0,2,0,13,109,0,0,1,1,0,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,
-109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,90,95,0,0,0,2,2,1,0,2,0,13,109,0,0,1,1,0,0,13,
-110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,
-16,10,49,0,57,22,0,0,1,90,95,0,0,0,2,3,1,0,2,0,13,109,0,0,1,1,0,0,13,110,0,0,0,1,9,18,109,0,18,109,
-0,18,110,0,48,20,0,0,1,90,95,0,0,0,2,4,1,0,2,0,13,109,0,0,1,1,0,0,13,110,0,0,0,1,9,18,109,0,16,8,
-48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,90,95,
-0,0,0,2,1,1,0,2,0,14,109,0,0,1,1,0,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,
-21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,
-50,0,57,21,0,0,1,90,95,0,0,0,2,2,1,0,2,0,14,109,0,0,1,1,0,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,
-18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,
-50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,90,95,0,0,0,2,3,1,0,2,0,14,109,0,0,1,1,0,0,14,110,0,0,0,1,
-9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,90,95,0,0,0,2,4,1,0,2,0,14,109,0,0,1,1,0,0,14,110,0,0,0,1,
-9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,
-24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,90,95,0,0,0,2,1,1,0,2,0,15,109,0,0,1,
-1,0,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,
-18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10,
-51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,90,95,0,0,0,2,2,1,0,2,0,15,109,0,0,1,1,0,0,15,110,0,0,0,1,
-9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,
-22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,
-51,0,57,22,0,0,1,90,95,0,0,0,2,3,1,0,2,0,15,109,0,0,1,1,0,0,15,110,0,0,0,1,9,18,109,0,18,109,0,18,
-110,0,48,20,0,0,1,90,95,0,0,0,2,4,1,0,2,0,15,109,0,0,1,1,0,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,
-57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,
-10,50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,90,
-95,0,0,0,2,1,1,0,2,0,13,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,
-0,16,10,49,0,57,18,97,0,21,0,0,1,90,95,0,0,0,2,2,1,0,2,0,13,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,
-0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,90,95,0,0,0,2,3,1,0,2,0,13,
-109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,
-23,0,0,1,90,95,0,0,0,2,4,1,0,2,0,13,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,
-24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,90,95,0,0,0,2,1,1,0,2,0,14,109,0,0,1,1,0,0,9,97,0,0,
-0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,
-0,57,18,97,0,21,0,0,1,90,95,0,0,0,2,2,1,0,2,0,14,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,
-57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,90,
-95,0,0,0,2,3,1,0,2,0,14,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,
-0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,90,95,0,0,0,2,4,1,0,2,0,14,
-109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,
-24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,90,95,0,0,0,2,1,1,0,2,0,15,109,0,0,1,1,0,0,9,97,0,0,
-0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,
-0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,90,95,0,0,0,2,2,1,0,2,0,15,109,0,0,1,1,
-0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,
-109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,90,95,0,0,0,2,3,1,0,2,0,
-15,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,
-97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,90,95,0,
-0,0,2,4,1,0,2,0,15,109,0,0,1,1,0,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,
-10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,
-0,0,1,90,95,0,0,0,2,3,1,0,2,0,10,118,0,0,1,1,0,0,13,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,
-0,0,1,90,95,0,0,0,2,3,1,0,2,0,11,118,0,0,1,1,0,0,14,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,
-0,0,1,90,95,0,0,0,2,3,1,0,2,0,12,118,0,0,1,1,0,0,15,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,
-0,0,1,90,95,0,0,5,2,25,1,0,2,0,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18,95,95,114,
-101,116,86,97,108,0,18,97,0,20,0,0,1,90,95,0,0,6,2,25,1,0,2,0,6,118,0,0,0,1,9,18,118,0,18,118,0,58,
-105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,
-95,0,0,7,2,25,1,0,2,0,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,
-20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,8,2,25,1,0,2,0,8,118,0,0,0,1,9,
-18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,
-18,118,0,20,0,0,1,90,95,0,0,9,2,25,1,0,2,0,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,
-18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,90,95,0,0,10,2,25,1,0,2,0,10,118,0,0,0,1,9,18,
-118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,
-118,0,20,0,0,1,90,95,0,0,11,2,25,1,0,2,0,11,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,
-49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,12,2,25,1,0,
-2,0,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,
-114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,13,2,25,1,0,2,0,13,109,0,0,0,1,9,18,109,0,16,8,
-48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,
-57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,
-97,108,0,18,109,0,20,0,0,1,90,95,0,0,14,2,25,1,0,2,0,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,
-0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,
-10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,
-50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,
-20,0,0,1,90,95,0,0,15,2,25,1,0,2,0,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,
-118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,
-101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,
-99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,
-52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,5,2,
-24,1,0,2,0,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,
-97,0,20,0,0,1,90,95,0,0,6,2,24,1,0,2,0,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,
-10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,7,2,24,1,0,2,0,
-7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,
-116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,8,2,24,1,0,2,0,8,118,0,0,0,1,9,18,118,0,18,118,0,58,
-105,118,101,99,52,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,
-95,0,0,9,2,24,1,0,2,0,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,46,20,0,9,18,95,95,114,101,116,
-86,97,108,0,18,97,0,20,0,0,1,90,95,0,0,10,2,24,1,0,2,0,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,
-101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,
-0,0,11,2,24,1,0,2,0,11,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,
-0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,12,2,24,1,0,2,0,12,118,0,0,0,1,9,
-18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,
-18,118,0,20,0,0,1,90,95,0,0,13,2,24,1,0,2,0,13,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,
-48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,
-57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,
-0,1,90,95,0,0,14,2,24,1,0,2,0,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,
-101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,
-99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,
-51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,15,
-2,24,1,0,2,0,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,
-0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,
-0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,
-0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,
-46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,5,0,95,95,112,111,115,116,
-68,101,99,114,0,1,0,2,0,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,
-97,0,16,10,49,0,47,20,0,0,1,90,95,0,0,6,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,6,118,0,0,
-0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,
-10,49,0,0,0,47,20,0,0,1,90,95,0,0,7,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,7,118,0,0,0,1,
-9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,
-49,0,0,0,47,20,0,0,1,90,95,0,0,8,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,8,118,0,0,0,1,9,
-18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,
-0,0,0,47,20,0,0,1,90,95,0,0,9,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,9,97,0,0,0,1,9,18,95,
-95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,0,1,90,95,0,0,10,
-0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
-18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,0,1,90,95,0,0,11,0,
-95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,
-118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,0,1,90,95,0,0,12,0,95,
-95,112,111,115,116,68,101,99,114,0,1,0,2,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,
-118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,90,95,0,0,13,0,95,
-95,112,111,115,116,68,101,99,114,0,1,0,2,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,
-109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,
-20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,
-0,1,90,95,0,0,14,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,14,109,0,0,0,1,9,18,95,95,114,101,
-116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,
-49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,
-48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,
-0,0,0,47,20,0,0,1,90,95,0,0,15,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,15,109,0,0,0,1,9,18,
-95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,
-101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,
-99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,
-52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,
-17,49,0,48,0,0,0,0,47,20,0,0,1,90,95,0,0,9,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,9,97,0,
-0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,90,
-95,0,0,10,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,
-97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,90,95,
-0,0,11,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,0,1,90,95,0,
-0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,
-108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,0,1,90,95,0,
-0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
-0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,90,95,0,0,6,0,95,95,112,111,115,116,73,110,
-99,114,0,1,0,2,0,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,
-0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,90,95,0,0,7,0,95,95,112,111,115,116,73,110,99,
-114,0,1,0,2,0,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,
-58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,90,95,0,0,8,0,95,95,112,111,115,116,73,110,99,
-114,0,1,0,2,0,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,
-58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,90,95,0,0,13,0,95,95,112,111,115,116,73,110,99,
-114,0,1,0,2,0,13,109,0,0,0,1,3,2,90,95,0,0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,
-109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,
-16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,90,95,0,0,14,0,95,95,
-112,111,115,116,73,110,99,114,0,1,0,2,0,14,109,0,0,0,1,3,2,90,95,0,0,14,1,110,0,2,18,109,0,0,0,9,
-18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,
-0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,
-10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,
-90,95,0,0,15,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,15,109,0,0,0,1,3,2,90,95,0,0,15,1,110,
-0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,
-0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,
-20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,
-9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,8,18,
-110,0,0,0,1,90,95,0,0,1,2,15,1,1,0,0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,4,118,101,99,52,95,115,103,116,
-0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0,0,0,0,1,90,95,0,0,1,2,15,1,1,0,0,5,
-97,0,0,1,1,0,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,
-0,0,40,0,0,1,90,95,0,0,1,2,16,1,1,0,0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,0,0,1,1,99,0,0,0,4,
-102,108,111,97,116,95,108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,90,95,0,0,
-1,2,16,1,1,0,0,5,97,0,0,1,1,0,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,
-97,116,0,18,98,0,0,0,41,0,0,1,90,95,0,0,1,2,18,1,1,0,0,9,97,0,0,1,1,0,0,9,98,0,0,0,1,3,2,90,95,0,0,
-1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,
-0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,
-0,32,0,0,1,90,95,0,0,1,2,18,1,1,0,0,5,97,0,0,1,1,0,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,
-0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,43,0,0,1,90,95,0,0,1,2,17,1,1,0,0,9,97,0,0,1,1,0,0,9,98,
-0,0,0,1,3,2,90,95,0,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115,115,0,18,103,0,
-0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,
-0,0,8,18,103,0,18,101,0,32,0,0,1,90,95,0,0,1,2,17,1,1,0,0,5,97,0,0,1,1,0,0,5,98,0,0,0,1,8,58,102,
-108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0,0,1,90,95,0,0,0,0,112,114,
-105,110,116,77,69,83,65,0,1,1,0,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114,105,110,116,0,18,
-102,0,0,0,0,1,90,95,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,5,105,0,0,0,1,4,105,110,116,
-95,112,114,105,110,116,0,18,105,0,0,0,0,1,90,95,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,
-1,98,0,0,0,1,4,98,111,111,108,95,112,114,105,110,116,0,18,98,0,0,0,0,1,90,95,0,0,0,0,112,114,105,
-110,116,77,69,83,65,0,1,1,0,0,10,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,
-120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,90,95,0,0,0,0,112,
-114,105,110,116,77,69,83,65,0,1,1,0,0,11,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,
-0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,
-110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,90,95,0,0,0,0,112,114,105,110,116,77,69,83,65,0,
-1,1,0,0,12,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,
-105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,
-59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,90,95,0,0,0,0,
-112,114,105,110,116,77,69,83,65,0,1,1,0,0,6,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,
-118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,90,95,0,0,
-0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,7,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,
-18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,
-114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,90,95,0,0,0,0,112,114,105,110,116,77,69,
-83,65,0,1,1,0,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,
-112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,
-18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,90,95,0,
-0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,2,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,
-0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,90,95,
-0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,3,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,
-65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,
-112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,90,95,0,0,0,0,112,114,105,110,116,77,
-69,83,65,0,1,1,0,0,4,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,
-58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,
-0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,90,95,
-0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,
-65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,
-0,0,1,90,95,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,14,109,0,0,0,1,9,58,112,114,105,110,
-116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,
-10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,90,95,0,0,
-0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,15,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,
-0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,
-9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77,69,
-83,65,0,18,109,0,16,10,51,0,57,0,0,0,0,1,90,95,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,
-16,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,112,114,105,
-110,116,77,69,83,65,0,1,1,0,0,17,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,
-1,90,95,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,18,101,0,0,0,1,4,105,110,116,95,112,114,
-105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,19,101,0,0,
-0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,112,114,105,110,116,77,
-69,83,65,0,1,1,0,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,
-0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,21,101,0,0,0,1,4,105,110,116,95,112,114,105,110,
-116,0,18,101,0,0,0,0,0
+0,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,90,95,0,0,12,0,2,21,1,1,0,0,15,0,109,0,0,1,1,0,
+0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,
+120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,
+118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,118,0,59,119,119,119,119,0,48,46,20,0,0,
+1,90,95,0,0,12,0,2,21,1,1,0,0,12,0,118,0,0,1,1,0,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,59,120,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,
+86,97,108,0,59,121,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,
+101,116,86,97,108,0,59,122,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,18,95,
+95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,20,0,
+0,1,90,95,0,0,13,0,2,1,1,0,2,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,
+109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,
+18,110,0,16,10,49,0,57,46,20,0,8,18,109,0,0,0,1,90,95,0,0,13,0,2,2,1,0,2,0,13,0,109,0,0,1,1,0,0,13,
+0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,109,
+0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,8,18,109,0,0,0,1,90,95,0,0,
+13,0,2,3,1,0,2,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,8,18,
+109,0,0,0,1,90,95,0,0,13,0,2,4,1,0,2,0,13,0,109,0,0,1,1,0,0,13,0,110,0,0,0,1,9,18,109,0,16,8,48,0,
+57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,
+0,57,18,110,0,16,10,49,0,57,49,20,0,8,18,109,0,0,0,1,90,95,0,0,14,0,2,1,1,0,2,0,14,0,109,0,0,1,1,0,
+0,14,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,
+18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,109,0,16,10,50,0,
+57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,8,18,109,0,0,0,1,90,95,0,0,14,0,2,2,1,0,2,
+0,14,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,
+8,48,0,57,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,
+18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,8,18,109,0,0,0,1,90,
+95,0,0,14,0,2,3,1,0,2,0,14,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,
+8,18,109,0,0,0,1,90,95,0,0,14,0,2,4,1,0,2,0,14,0,109,0,0,1,1,0,0,14,0,110,0,0,0,1,9,18,109,0,16,8,
+48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,
+10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,
+16,10,50,0,57,49,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,1,1,0,2,0,15,0,109,0,0,1,1,0,0,15,0,110,0,
+0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,109,0,16,10,
+49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,
+10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,
+16,10,51,0,57,46,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,2,1,0,2,0,15,0,109,0,0,1,1,0,0,15,0,110,0,
+0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,109,0,16,10,
+49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,
+10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,
+16,10,51,0,57,47,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,3,1,0,2,0,15,0,109,0,0,1,1,0,0,15,0,110,0,
+0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,4,1,0,2,0,15,0,109,0,
+0,1,1,0,0,15,0,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,
+20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,109,0,16,
+10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,9,18,109,0,16,10,51,0,57,18,109,0,
+16,10,51,0,57,18,110,0,16,10,51,0,57,49,20,0,8,18,109,0,0,0,1,90,95,0,0,13,0,2,1,1,0,2,0,13,0,109,
+0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,10,0,1,118,0,2,58,118,101,99,50,0,0,18,97,0,0,0,0,0,9,18,
+109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,
+49,0,57,18,118,0,46,20,0,8,18,109,0,0,0,1,90,95,0,0,13,0,2,2,1,0,2,0,13,0,109,0,0,1,1,0,0,9,0,97,0,
+0,0,1,3,2,90,95,0,0,10,0,1,118,0,2,58,118,101,99,50,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,
+109,0,16,8,48,0,57,18,118,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,47,20,
+0,8,18,109,0,0,0,1,90,95,0,0,13,0,2,3,1,0,2,0,13,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,10,
+0,1,118,0,2,58,118,101,99,50,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,
+118,0,48,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,48,20,0,8,18,109,0,0,0,1,90,
+95,0,0,13,0,2,4,1,0,2,0,13,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,10,0,1,118,0,2,58,118,
+101,99,50,0,0,17,49,0,48,0,0,18,97,0,49,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,
+118,0,48,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,48,20,0,8,18,109,0,0,0,1,90,
+95,0,0,14,0,2,1,1,0,2,0,14,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,11,0,1,118,0,2,58,118,
+101,99,51,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,46,20,0,9,18,
+109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,
+10,50,0,57,18,118,0,46,20,0,8,18,109,0,0,0,1,90,95,0,0,14,0,2,2,1,0,2,0,14,0,109,0,0,1,1,0,0,9,0,
+97,0,0,0,1,3,2,90,95,0,0,11,0,1,118,0,2,58,118,101,99,51,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,
+57,18,109,0,16,8,48,0,57,18,118,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,
+47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,118,0,47,20,0,8,18,109,0,0,0,1,90,95,0,
+0,14,0,2,3,1,0,2,0,14,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,11,0,1,118,0,2,58,118,101,99,
+51,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,48,20,0,9,18,109,0,
+16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,48,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,
+57,18,118,0,48,20,0,8,18,109,0,0,0,1,90,95,0,0,14,0,2,4,1,0,2,0,14,0,109,0,0,1,1,0,0,9,0,97,0,0,0,
+1,3,2,90,95,0,0,11,0,1,118,0,2,58,118,101,99,51,0,0,17,49,0,48,0,0,18,97,0,49,0,0,0,0,9,18,109,0,
+16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,48,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,
+57,18,118,0,48,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,118,0,48,20,0,8,18,109,0,0,
+0,1,90,95,0,0,15,0,2,1,1,0,2,0,15,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,12,0,1,118,0,2,58,
+118,101,99,52,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,46,20,0,9,
+18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,
+16,10,50,0,57,18,118,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,118,0,46,20,0,8,
+18,109,0,0,0,1,90,95,0,0,15,0,2,2,1,0,2,0,15,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,95,0,0,12,0,1,
+118,0,2,58,118,101,99,52,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,
+0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,47,20,0,9,18,109,0,16,10,50,0,
+57,18,109,0,16,10,50,0,57,18,118,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,118,
+0,47,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,3,1,0,2,0,15,0,109,0,0,1,1,0,0,9,0,97,0,0,0,1,3,2,90,
+95,0,0,12,0,1,118,0,2,58,118,101,99,52,0,0,18,97,0,0,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,
+48,0,57,18,118,0,48,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,118,0,48,20,0,9,18,109,
+0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,118,0,48,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,
+0,57,18,118,0,48,20,0,8,18,109,0,0,0,1,90,95,0,0,15,0,2,4,1,0,2,0,15,0,109,0,0,1,1,0,0,9,0,97,0,0,
+0,1,3,2,90,95,0,0,12,0,1,118,0,2,58,118,101,99,52,0,0,17,49,0,48,0,0,18,97,0,49,0,0,0,0,9,18,109,0,
+16,8,48,0,57,18,109,0,16,8,48,0,57,18,118,0,48,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,
+57,18,118,0,48,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,118,0,48,20,0,9,18,109,0,16,
+10,51,0,57,18,109,0,16,10,51,0,57,18,118,0,48,20,0,8,18,109,0,0,0,1,90,95,0,0,10,0,2,3,1,0,2,0,10,
+0,118,0,0,1,1,0,0,13,0,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,
+11,0,2,3,1,0,2,0,11,0,118,0,0,1,1,0,0,14,0,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,8,18,
+118,0,0,0,1,90,95,0,0,12,0,2,3,1,0,2,0,12,0,118,0,0,1,1,0,0,15,0,109,0,0,0,1,9,18,118,0,18,118,0,
+18,109,0,48,20,0,8,18,118,0,0,0,1,90,95,0,0,5,0,2,25,1,0,2,0,5,0,97,0,0,0,1,9,18,97,0,18,97,0,16,
+10,49,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,90,95,0,0,6,0,2,25,1,0,2,0,6,0,
+118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,
+116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,7,0,2,25,1,0,2,0,7,0,118,0,0,0,1,9,18,118,0,18,118,0,
+58,105,118,101,99,51,0,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,
+1,90,95,0,0,8,0,2,25,1,0,2,0,8,0,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,0,16,10,49,
+0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,9,0,2,25,1,0,2,0,9,0,
+97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,
+0,0,1,90,95,0,0,10,0,2,25,1,0,2,0,10,0,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,0,17,49,
+0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,11,0,2,25,1,0,
+2,0,11,0,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,
+95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,12,0,2,25,1,0,2,0,12,0,118,0,0,0,1,9,18,118,
+0,18,118,0,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,
+118,0,20,0,0,1,90,95,0,0,13,0,2,25,1,0,2,0,13,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,
+48,0,57,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,
+0,57,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,
+20,0,0,1,90,95,0,0,14,0,2,25,1,0,2,0,14,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,
+57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,
+58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,
+118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,
+90,95,0,0,15,0,2,25,1,0,2,0,15,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,
+101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,
+101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,
+101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,
+101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,
+95,0,0,5,0,2,24,1,0,2,0,5,0,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,
+86,97,108,0,18,97,0,20,0,0,1,90,95,0,0,6,0,2,24,1,0,2,0,6,0,118,0,0,0,1,9,18,118,0,18,118,0,58,105,
+118,101,99,50,0,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,
+95,0,0,7,0,2,24,1,0,2,0,7,0,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,0,16,10,49,0,0,
+0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,8,0,2,24,1,0,2,0,8,0,118,
+0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,
+86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,9,0,2,24,1,0,2,0,9,0,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,
+48,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,90,95,0,0,10,0,2,24,1,0,2,0,10,
+0,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,
+101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,11,0,2,24,1,0,2,0,11,0,118,0,0,0,1,9,18,118,0,18,
+118,0,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,
+20,0,0,1,90,95,0,0,12,0,2,24,1,0,2,0,12,0,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,0,17,
+49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,90,95,0,0,13,0,2,24,1,
+0,2,0,13,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,0,17,49,0,
+48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,0,17,49,0,48,
+0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,14,0,2,24,1,0,2,0,
+14,0,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,
+0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,
+0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,
+46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,15,0,2,24,1,0,2,0,15,0,109,
+0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,
+20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,
+0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,
+9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,9,
+18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,90,95,0,0,5,0,0,95,95,112,111,115,116,68,101,99,
+114,0,1,0,2,0,5,0,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,
+10,49,0,47,20,0,0,1,90,95,0,0,6,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,6,0,118,0,0,0,1,
+9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,0,16,
+10,49,0,0,0,47,20,0,0,1,90,95,0,0,7,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,7,0,118,0,0,
+0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,0,
+16,10,49,0,0,0,47,20,0,0,1,90,95,0,0,8,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,8,0,118,0,
+0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0,
+0,16,10,49,0,0,0,47,20,0,0,1,90,95,0,0,9,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,9,0,97,
+0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,0,
+1,90,95,0,0,10,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,10,0,118,0,0,0,1,9,18,95,95,114,
+101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,47,
+20,0,0,1,90,95,0,0,11,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,11,0,118,0,0,0,1,9,18,95,
+95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,0,17,49,0,48,0,0,0,
+0,47,20,0,0,1,90,95,0,0,12,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,12,0,118,0,0,0,1,9,18,
+95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,0,17,49,0,48,0,
+0,0,0,47,20,0,0,1,90,95,0,0,13,0,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,0,13,0,109,0,0,0,1,
+9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,
+118,101,99,50,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,
+118,101,99,50,0,0,17,49,0,48,0,0,0,0,47,20,0,0,1,90,95,0,0,14,0,0,95,95,112,111,115,116,68,101,99,
+114,0,1,0,2,0,14,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,
+0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,
+57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,
+18,109,0,16,10,50,0,57,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,47,20,0,0,1,90,95,0,0,15,0,0,95,95,
+112,111,115,116,68,101,99,114,0,1,0,2,0,15,0,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,
+0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,
+0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,
+9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,9,
+18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,47,20,0,0,1,
+90,95,0,0,9,0,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,9,0,97,0,0,0,1,9,18,95,95,114,101,
+116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,90,95,0,0,10,0,0,95,95,112,
+111,115,116,73,110,99,114,0,1,0,2,0,10,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,
+20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,46,20,0,0,1,90,95,0,0,11,0,0,95,
+95,112,111,115,116,73,110,99,114,0,1,0,2,0,11,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,
+118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,0,17,49,0,48,0,0,0,0,46,20,0,0,1,90,95,0,0,12,0,
+0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,12,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,
+0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,0,1,90,95,0,0,
+5,0,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,5,0,97,0,0,0,1,9,18,95,95,114,101,116,86,97,
+108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,90,95,0,0,6,0,0,95,95,112,111,115,116,
+73,110,99,114,0,1,0,2,0,6,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,
+0,18,118,0,58,105,118,101,99,50,0,0,16,10,49,0,0,0,46,20,0,0,1,90,95,0,0,7,0,0,95,95,112,111,115,
+116,73,110,99,114,0,1,0,2,0,7,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,
+118,0,18,118,0,58,105,118,101,99,51,0,0,16,10,49,0,0,0,46,20,0,0,1,90,95,0,0,8,0,0,95,95,112,111,
+115,116,73,110,99,114,0,1,0,2,0,8,0,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,
+18,118,0,18,118,0,58,105,118,101,99,51,0,0,16,10,49,0,0,0,46,20,0,0,1,90,95,0,0,13,0,0,95,95,112,
+111,115,116,73,110,99,114,0,1,0,2,0,13,0,109,0,0,0,1,3,2,90,95,0,0,13,0,1,110,0,2,18,109,0,0,0,9,
+18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,
+109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,0,17,49,0,48,0,0,0,0,46,20,0,8,18,
+110,0,0,0,1,90,95,0,0,14,0,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,0,14,0,109,0,0,0,1,3,2,90,
+95,0,0,14,0,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,
+0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,
+0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,0,
+17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,90,95,0,0,15,0,0,95,95,112,111,115,116,73,110,99,114,0,
+1,0,2,0,15,0,109,0,0,0,1,3,2,90,95,0,0,15,0,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,
+0,16,8,48,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,
+16,10,49,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,
+10,50,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,
+51,0,57,58,118,101,99,52,0,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,90,95,0,0,1,0,2,15,1,1,0,
+0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,
+108,0,59,120,0,0,18,98,0,0,18,97,0,0,0,0,1,90,95,0,0,1,0,2,15,1,1,0,0,5,0,97,0,0,1,1,0,0,5,0,98,0,
+0,0,1,8,58,102,108,111,97,116,0,0,18,97,0,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,40,0,0,1,90,95,
+0,0,1,0,2,16,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,1,0,1,99,0,0,0,4,102,108,111,
+97,116,95,108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,90,95,0,0,1,0,2,16,1,1,
+0,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,8,58,102,108,111,97,116,0,0,18,97,0,0,0,58,102,108,111,97,
+116,0,0,18,98,0,0,0,41,0,0,1,90,95,0,0,1,0,2,18,1,1,0,0,9,0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,
+95,0,0,1,0,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,98,0,0,
+18,97,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,
+0,18,101,0,32,0,0,1,90,95,0,0,1,0,2,18,1,1,0,0,5,0,97,0,0,1,1,0,0,5,0,98,0,0,0,1,8,58,102,108,111,
+97,116,0,0,18,97,0,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,43,0,0,1,90,95,0,0,1,0,2,17,1,1,0,0,9,
+0,97,0,0,1,1,0,0,9,0,98,0,0,0,1,3,2,90,95,0,0,1,0,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,
+108,101,115,115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,
+101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,90,95,0,0,1,0,2,17,1,1,0,0,5,0,97,0,0,1,
+1,0,0,5,0,98,0,0,0,1,8,58,102,108,111,97,116,0,0,18,97,0,0,0,58,102,108,111,97,116,0,0,18,98,0,0,0,
+42,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,9,0,102,0,0,0,1,4,102,108,111,
+97,116,95,112,114,105,110,116,0,18,102,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,
+1,1,0,0,5,0,105,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,105,0,0,0,0,1,90,95,0,0,0,0,0,
+112,114,105,110,116,77,69,83,65,0,1,1,0,0,1,0,98,0,0,0,1,4,98,111,111,108,95,112,114,105,110,116,0,
+18,98,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,10,0,118,0,0,0,1,9,58,
+112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,
+0,18,118,0,59,121,0,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,11,0,118,0,
+0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,
+69,83,65,0,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,122,0,0,
+0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,12,0,118,0,0,0,1,9,58,112,114,
+105,110,116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,
+118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,122,0,0,0,0,9,58,112,114,
+105,110,116,77,69,83,65,0,0,18,118,0,59,119,0,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,
+83,65,0,1,1,0,0,6,0,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,
+58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,121,0,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,
+116,77,69,83,65,0,1,1,0,0,7,0,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,120,
+0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,
+77,69,83,65,0,0,18,118,0,59,122,0,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,
+0,0,8,0,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,
+105,110,116,77,69,83,65,0,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,
+118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,119,0,0,0,0,0,1,90,95,0,
+0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,2,0,118,0,0,0,1,9,58,112,114,105,110,116,77,69,
+83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,121,0,0,0,0,
+0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,3,0,118,0,0,0,1,9,58,112,114,105,110,
+116,77,69,83,65,0,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,
+121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,122,0,0,0,0,0,1,90,95,0,0,0,0,0,
+112,114,105,110,116,77,69,83,65,0,1,1,0,0,4,0,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,
+18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,118,0,59,121,0,0,0,0,9,58,112,
+114,105,110,116,77,69,83,65,0,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,
+18,118,0,59,119,0,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,13,0,109,0,0,
+0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,
+77,69,83,65,0,0,18,109,0,16,10,49,0,57,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,
+1,1,0,0,14,0,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,109,0,16,8,48,0,57,0,0,0,9,58,
+112,114,105,110,116,77,69,83,65,0,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,
+65,0,0,18,109,0,16,10,50,0,57,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,
+15,0,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,
+105,110,116,77,69,83,65,0,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,
+18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,0,18,109,0,16,10,51,0,57,0,0,0,
+0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,16,0,101,0,0,0,1,4,105,110,116,95,
+112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,
+17,0,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,0,112,114,
+105,110,116,77,69,83,65,0,1,1,0,0,18,0,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,
+0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,0,19,0,101,0,0,0,1,4,105,110,116,
+95,112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,
+0,20,0,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,90,95,0,0,0,0,0,112,114,
+105,110,116,77,69,83,65,0,1,1,0,0,21,0,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,
+0,0,0,0
diff --git a/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h b/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h
index b4c9abbafb..738a0f9295 100644
--- a/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h
+++ b/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h
@@ -2,104 +2,106 @@
 /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */
 /* slang_fragment_builtin.gc */
 
-4,2,2,90,95,6,0,12,1,103,108,95,70,114,97,103,67,111,111,114,100,0,0,0,2,2,90,95,6,0,1,1,103,108,
-95,70,114,111,110,116,70,97,99,105,110,103,0,0,0,2,2,90,95,5,0,12,1,103,108,95,70,114,97,103,67,
-111,108,111,114,0,0,0,2,2,90,95,5,0,12,1,103,108,95,70,114,97,103,68,97,116,97,0,3,18,103,108,95,
-77,97,120,68,114,97,119,66,117,102,102,101,114,115,0,0,0,2,2,90,95,5,0,9,1,103,108,95,70,114,97,
-103,68,101,112,116,104,0,0,0,2,2,90,95,3,0,12,1,103,108,95,67,111,108,111,114,0,0,0,2,2,90,95,3,0,
-12,1,103,108,95,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,1,103,
-108,95,84,101,120,67,111,111,114,100,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,
-111,114,100,115,0,0,0,2,2,90,95,3,0,9,1,103,108,95,70,111,103,70,114,97,103,67,111,111,114,100,0,0,
-0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,49,68,0,1,1,0,0,16,115,97,109,112,108,101,114,0,0,1,
-1,0,0,9,99,111,111,114,100,0,0,1,1,0,0,9,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,99,111,111,114,
-100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,
-114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,
-114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,
-95,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,0,16,115,97,109,112,108,101,
-114,0,0,1,1,0,0,10,99,111,111,114,100,0,0,1,1,0,0,9,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,112,
-99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,
-18,99,111,111,114,100,0,59,121,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,
-20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,
-112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,
-49,68,80,114,111,106,0,1,1,0,0,16,115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,111,111,114,100,0,0,
-1,1,0,0,9,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,
-111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,
-18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,49,
-100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,
-100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,50,68,0,1,1,0,0,17,115,97,109,112,108,101,
-114,0,0,1,1,0,0,10,99,111,111,114,100,0,0,1,1,0,0,9,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,99,
-111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,0,18,99,111,111,114,100,0,59,120,
-121,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,
-101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,
-111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,0,
-17,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,1,1,0,0,9,98,105,97,115,0,0,0,
-1,3,2,90,95,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,
-111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,
-100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,
-116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,
-12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,0,17,115,97,109,112,108,101,114,0,0,
-1,1,0,0,12,99,111,111,114,100,0,0,1,1,0,0,9,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,112,99,111,
-111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,
-18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,
-20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,
-112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,
-51,68,0,1,1,0,0,18,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,1,1,0,0,9,98,
-105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,
-120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,
-18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97,108,
-0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,116,101,
-120,116,117,114,101,51,68,80,114,111,106,0,1,1,0,0,18,115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,
-111,111,114,100,0,0,1,1,0,0,9,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,112,99,111,111,114,100,0,0,
-0,9,18,112,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,18,99,
-111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,
-4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,
-108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,67,
-117,98,101,0,1,1,0,0,19,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,1,1,0,0,9,
-98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,
-59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,
-115,0,20,0,4,118,101,99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,
-115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,115,104,97,100,
-111,119,49,68,0,1,1,0,0,20,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,1,1,0,
-0,9,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,
-52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,
-97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,
-97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,115,104,97,100,111,
-119,49,68,80,114,111,106,0,1,1,0,0,20,115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,111,111,114,100,
-0,0,1,1,0,0,9,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,
-111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,119,0,49,20,
-0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,
-114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,
-101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,
-0,0,12,0,115,104,97,100,111,119,50,68,0,1,1,0,0,21,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,
-111,111,114,100,0,0,1,1,0,0,9,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,99,111,111,114,100,52,0,0,0,
-9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,
-52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,
-116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,
-12,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0,0,21,115,97,109,112,108,101,114,0,0,1,1,0,
-0,12,99,111,111,114,100,0,0,1,1,0,0,9,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,1,112,99,111,111,114,
+5,2,2,90,95,6,0,12,0,1,103,108,95,70,114,97,103,67,111,111,114,100,0,0,0,2,2,90,95,6,0,1,0,1,103,
+108,95,70,114,111,110,116,70,97,99,105,110,103,0,0,0,2,2,90,95,5,0,12,0,1,103,108,95,70,114,97,103,
+67,111,108,111,114,0,0,0,2,2,90,95,5,0,12,0,1,103,108,95,70,114,97,103,68,97,116,97,0,3,18,103,108,
+95,77,97,120,68,114,97,119,66,117,102,102,101,114,115,0,0,0,2,2,90,95,5,0,9,0,1,103,108,95,70,114,
+97,103,68,101,112,116,104,0,0,0,2,2,90,95,3,0,12,0,1,103,108,95,67,111,108,111,114,0,0,0,2,2,90,95,
+3,0,12,0,1,103,108,95,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,0,
+1,103,108,95,84,101,120,67,111,111,114,100,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,
+67,111,111,114,100,115,0,0,0,2,2,90,95,3,0,9,0,1,103,108,95,70,111,103,70,114,97,103,67,111,111,
+114,100,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,0,1,1,0,0,16,0,115,97,109,112,
+108,101,114,0,0,1,1,0,0,9,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,
+12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,0,18,99,111,111,114,100,0,
+20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,
+98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,
+100,52,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,0,16,0,
+115,97,109,112,108,101,114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,
+1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,
+111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,121,0,49,20,0,9,18,112,99,111,111,114,100,0,
+59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,
+86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,
+0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,0,16,0,115,97,109,112,108,101,114,0,0,1,
+1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,
+111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,
+99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,
+0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,
+108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,
+50,68,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,1,1,0,0,9,
+0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,
+52,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,
+18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,
+0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,116,101,
+120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,
+0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,
 100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,
-111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,
-59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,
+111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,
+4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,
+108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,
+50,68,80,114,111,106,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,
+0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,
+99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,
+119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,
 116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,
-112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,9,0,100,70,100,120,0,1,1,0,0,9,112,0,0,0,1,4,118,101,99,
-52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,0,59,120,120,120,120,0,0,0,
-0,1,90,95,0,0,10,0,100,70,100,120,0,1,1,0,0,10,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,
-95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,59,120,121,121,121,0,0,0,0,1,90,95,0,0,11,0,100,
-70,100,120,0,1,1,0,0,11,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,
-108,0,59,120,121,122,0,0,18,112,0,59,120,121,122,122,0,0,0,0,1,90,95,0,0,12,0,100,70,100,120,0,1,1,
-0,0,12,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,
-0,0,1,90,95,0,0,9,0,100,70,100,121,0,1,1,0,0,9,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,
-95,114,101,116,86,97,108,0,59,120,0,0,18,112,0,59,120,120,120,120,0,0,0,0,1,90,95,0,0,10,0,100,70,
-100,121,0,1,1,0,0,10,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,
-59,120,121,0,0,18,112,0,59,120,121,121,121,0,0,0,0,1,90,95,0,0,11,0,100,70,100,121,0,1,1,0,0,11,
-112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,
-18,112,0,59,120,121,122,122,0,0,0,0,1,90,95,0,0,12,0,100,70,100,121,0,1,1,0,0,12,112,0,0,0,1,4,118,
-101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,0,0,1,90,95,0,0,9,0,102,
-119,105,100,116,104,0,1,1,0,0,9,112,0,0,0,1,8,58,97,98,115,0,58,100,70,100,120,0,18,112,0,0,0,0,0,
-58,97,98,115,0,58,100,70,100,121,0,18,112,0,0,0,0,0,46,0,0,1,90,95,0,0,10,0,102,119,105,100,116,
-104,0,1,1,0,0,10,112,0,0,0,1,8,58,97,98,115,0,58,100,70,100,120,0,18,112,0,0,0,0,0,58,97,98,115,0,
-58,100,70,100,121,0,18,112,0,0,0,0,0,46,0,0,1,90,95,0,0,11,0,102,119,105,100,116,104,0,1,1,0,0,11,
-112,0,0,0,1,8,58,97,98,115,0,58,100,70,100,120,0,18,112,0,0,0,0,0,58,97,98,115,0,58,100,70,100,121,
-0,18,112,0,0,0,0,0,46,0,0,1,90,95,0,0,12,0,102,119,105,100,116,104,0,1,1,0,0,12,112,0,0,0,1,8,58,
-97,98,115,0,58,100,70,100,120,0,18,112,0,0,0,0,0,58,97,98,115,0,58,100,70,100,121,0,18,112,0,0,0,0,
-0,46,0,0,0
+112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,51,68,0,1,1,0,0,18,0,
+115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,
+1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,
+18,99,111,111,114,100,0,59,120,121,122,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,
+115,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,
+109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,
+114,101,51,68,80,114,111,106,0,1,1,0,0,18,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,
+114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,
+18,112,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,18,99,111,
+111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,
+118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,
+101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,67,
+117,98,101,0,1,1,0,0,19,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,
+0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,
+100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,
+105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,
+0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,
+100,111,119,49,68,0,1,1,0,0,20,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,
+0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,
+111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,
+0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,
+108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,115,
+104,97,100,111,119,49,68,80,114,111,106,0,1,1,0,0,20,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,
+99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,
+100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,
+114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,
+20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,
+98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,
+114,100,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,50,68,0,1,1,0,0,21,0,115,97,109,112,108,
+101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,0,1,3,2,90,95,0,0,12,
+0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,
+100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,
+101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,
+111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0,0,21,
+0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,98,105,97,115,0,0,
+0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,
+18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,
+114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,
+98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,
+18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,9,0,0,100,70,100,
+120,0,1,1,0,0,9,0,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,59,
+120,0,0,18,112,0,59,120,120,120,120,0,0,0,0,1,90,95,0,0,10,0,0,100,70,100,120,0,1,1,0,0,10,0,112,0,
+0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,59,
+120,121,121,121,0,0,0,0,1,90,95,0,0,11,0,0,100,70,100,120,0,1,1,0,0,11,0,112,0,0,0,1,4,118,101,99,
+52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,59,120,121,122,
+122,0,0,0,0,1,90,95,0,0,12,0,0,100,70,100,120,0,1,1,0,0,12,0,112,0,0,0,1,4,118,101,99,52,95,100,
+100,120,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,0,0,1,90,95,0,0,9,0,0,100,70,100,121,0,1,1,
+0,0,9,0,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,
+18,112,0,59,120,120,120,120,0,0,0,0,1,90,95,0,0,10,0,0,100,70,100,121,0,1,1,0,0,10,0,112,0,0,0,1,4,
+118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,59,120,121,
+121,121,0,0,0,0,1,90,95,0,0,11,0,0,100,70,100,121,0,1,1,0,0,11,0,112,0,0,0,1,4,118,101,99,52,95,
+100,100,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,59,120,121,122,122,0,0,
+0,0,1,90,95,0,0,12,0,0,100,70,100,121,0,1,1,0,0,12,0,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,
+18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,0,0,1,90,95,0,0,9,0,0,102,119,105,100,116,104,0,1,1,
+0,0,9,0,112,0,0,0,1,8,58,97,98,115,0,0,58,100,70,100,120,0,0,18,112,0,0,0,0,0,58,97,98,115,0,0,58,
+100,70,100,121,0,0,18,112,0,0,0,0,0,46,0,0,1,90,95,0,0,10,0,0,102,119,105,100,116,104,0,1,1,0,0,10,
+0,112,0,0,0,1,8,58,97,98,115,0,0,58,100,70,100,120,0,0,18,112,0,0,0,0,0,58,97,98,115,0,0,58,100,70,
+100,121,0,0,18,112,0,0,0,0,0,46,0,0,1,90,95,0,0,11,0,0,102,119,105,100,116,104,0,1,1,0,0,11,0,112,
+0,0,0,1,8,58,97,98,115,0,0,58,100,70,100,120,0,0,18,112,0,0,0,0,0,58,97,98,115,0,0,58,100,70,100,
+121,0,0,18,112,0,0,0,0,0,46,0,0,1,90,95,0,0,12,0,0,102,119,105,100,116,104,0,1,1,0,0,12,0,112,0,0,
+0,1,8,58,97,98,115,0,0,58,100,70,100,120,0,0,18,112,0,0,0,0,0,58,97,98,115,0,0,58,100,70,100,121,0,
+0,18,112,0,0,0,0,0,46,0,0,0
diff --git a/src/mesa/shader/slang/library/slang_pp_directives.syn b/src/mesa/shader/slang/library/slang_pp_directives.syn
index d4a321034d..b51d168cc0 100644
--- a/src/mesa/shader/slang/library/slang_pp_directives.syn
+++ b/src/mesa/shader/slang/library/slang_pp_directives.syn
@@ -79,6 +79,12 @@
 .emtcode BEHAVIOR_WARN    3
 .emtcode BEHAVIOR_DISABLE 4
 
+/*
+ * The PRAGMA_* symbols follow TOKEN_PRAGMA
+ */
+.emtcode PRAGMA_NO_PARAM 0
+.emtcode PRAGMA_PARAM    1
+
 source
    optional_directive .and .loop source_element .and '\0' .emit ESCAPE_TOKEN .emit TOKEN_END;
 
@@ -153,6 +159,7 @@ directive
    dir_elif .emit TOKEN_ELIF .or
    dir_endif .emit TOKEN_ENDIF .or
    dir_ext .emit TOKEN_EXTENSION .or
+   dir_pragma .emit TOKEN_PRAGMA .or
    dir_line .emit TOKEN_LINE;
 
 dir_define
@@ -187,6 +194,19 @@ dir_ext
 dir_line
    optional_space .and '#' .and optional_space .and "line" .and expression;
 
+dir_pragma
+   optional_space .and '#' .and optional_space .and "pragma" .and symbol .and opt_pragma_param;
+
+
+opt_pragma_param
+   pragma_param .or .true .emit PRAGMA_NO_PARAM;
+
+pragma_param
+   optional_space .and '(' .emit PRAGMA_PARAM .and optional_space .and symbol_no_space .and optional_space .and ')';
+
+symbol_no_space
+   symbol_character .emit * .and .loop symbol_character2 .emit * .and .true .emit '\0';
+
 symbol
    space .and symbol_character .emit * .and .loop symbol_character2 .emit * .and .true .emit '\0';
 
diff --git a/src/mesa/shader/slang/library/slang_pp_directives_syn.h b/src/mesa/shader/slang/library/slang_pp_directives_syn.h
index 35e7bc2761..430f8d8217 100644
--- a/src/mesa/shader/slang/library/slang_pp_directives_syn.h
+++ b/src/mesa/shader/slang/library/slang_pp_directives_syn.h
@@ -20,6 +20,8 @@
 ".emtcode BEHAVIOR_ENABLE 2\n"
 ".emtcode BEHAVIOR_WARN 3\n"
 ".emtcode BEHAVIOR_DISABLE 4\n"
+".emtcode PRAGMA_NO_PARAM 0\n"
+".emtcode PRAGMA_PARAM 1\n"
 "source\n"
 " optional_directive .and .loop source_element .and '\\0' .emit ESCAPE_TOKEN .emit TOKEN_END;\n"
 "source_element\n"
@@ -76,6 +78,7 @@
 " dir_elif .emit TOKEN_ELIF .or\n"
 " dir_endif .emit TOKEN_ENDIF .or\n"
 " dir_ext .emit TOKEN_EXTENSION .or\n"
+" dir_pragma .emit TOKEN_PRAGMA .or\n"
 " dir_line .emit TOKEN_LINE;\n"
 "dir_define\n"
 " optional_space .and '#' .and optional_space .and \"define\" .and symbol .and opt_parameters .and\n"
@@ -99,6 +102,14 @@
 " optional_space .and ':' .and optional_space .and extension_behavior;\n"
 "dir_line\n"
 " optional_space .and '#' .and optional_space .and \"line\" .and expression;\n"
+"dir_pragma\n"
+" optional_space .and '#' .and optional_space .and \"pragma\" .and symbol .and opt_pragma_param;\n"
+"opt_pragma_param\n"
+" pragma_param .or .true .emit PRAGMA_NO_PARAM;\n"
+"pragma_param\n"
+" optional_space .and '(' .emit PRAGMA_PARAM .and optional_space .and symbol_no_space .and optional_space .and ')';\n"
+"symbol_no_space\n"
+" symbol_character .emit * .and .loop symbol_character2 .emit * .and .true .emit '\\0';\n"
 "symbol\n"
 " space .and symbol_character .emit * .and .loop symbol_character2 .emit * .and .true .emit '\\0';\n"
 "opt_parameters\n"
diff --git a/src/mesa/shader/slang/library/slang_shader.syn b/src/mesa/shader/slang/library/slang_shader.syn
index cbfbda651a..cc5c70a02f 100644
--- a/src/mesa/shader/slang/library/slang_shader.syn
+++ b/src/mesa/shader/slang/library/slang_shader.syn
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
  *
  * Copyright (C) 2004-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2008 VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -59,7 +59,7 @@
 .syntax translation_unit;
 
 /* revision number - increment after each change affecting emitted output */
-.emtcode REVISION                                   4
+.emtcode REVISION                                   5
 
 /* external declaration (or precision or invariant stmt) */
 .emtcode EXTERNAL_NULL                              0
@@ -83,6 +83,10 @@
 .emtcode FUNCTION_CONSTRUCTOR                       1
 .emtcode FUNCTION_OPERATOR                          2
 
+/* function call type */
+.emtcode FUNCTION_CALL_NONARRAY                     0
+.emtcode FUNCTION_CALL_ARRAY                        1
+
 /* operator type */
 .emtcode OPERATOR_ADDASSIGN                         1
 .emtcode OPERATOR_SUBASSIGN                         2
@@ -178,6 +182,9 @@
 .emtcode TYPE_SPECIFIER_MAT34                       30
 .emtcode TYPE_SPECIFIER_MAT43                       31
 
+/* type specifier array */
+.emtcode TYPE_SPECIFIER_NONARRAY                    0
+.emtcode TYPE_SPECIFIER_ARRAY                       1
 
 /* structure field */
 .emtcode FIELD_NONE                                 0
@@ -248,6 +255,7 @@
 .emtcode OP_POSTINCREMENT                           60
 .emtcode OP_POSTDECREMENT                           61
 .emtcode OP_PRECISION                               62
+.emtcode OP_METHOD                                  63
 
 /* parameter qualifier */
 .emtcode PARAM_QUALIFIER_IN                         0
@@ -339,6 +347,25 @@ integer_expression
  * <function_call> ::= <function_call_generic>
  */
 function_call
+    function_call_or_method;
+
+/*
+ * <function_call_or_method> ::= <regular_function_call>
+ *                             | <postfix_expression> "." <function_call_generic>
+ */
+function_call_or_method
+    regular_function_call .or method_call;
+
+/*
+ * <method_call> ::= <identifier> "." <function_call_generic>
+ */
+method_call
+    identifier .emit OP_METHOD .and dot .and function_call_generic .and .true .emit OP_END;
+
+/*
+ * <regular_function_call> ::= <function_call_generic>
+ */
+regular_function_call
     function_call_generic .emit OP_CALL .and .true .emit OP_END;
 
 /*
@@ -380,11 +407,17 @@ function_call_header
 /*
  * <function_identifier> ::= <constructor_identifier>
  *                         | <identifier>
+ *                         | <type_specifier>
  *
- * note: <constructor_identifier> has been deleted
+ * note: <constructor_identifier> and <type_specifier> have been deleted
  */
 function_identifier
-    identifier;
+    identifier .and function_identifier_opt_array;
+function_identifier_opt_array
+    function_identifier_array .emit FUNCTION_CALL_ARRAY .or
+    .true .emit FUNCTION_CALL_NONARRAY;
+function_identifier_array
+    lbracket .and constant_expression .and rbracket;
 
 /*
  * <unary_expression> ::= <postfix_expression>
@@ -837,10 +870,8 @@ parameter_type_specifier_3
  * <init_declarator_list> ::= <single_declaration>
  *                          | <init_declarator_list> "," <identifier>
  *                          | <init_declarator_list> "," <identifier> "[" "]"
- *                          | <init_declarator_list> "," <identifier> "["
- *                            <constant_expression> "]"
- *                          | <init_declarator_list> "," <identifier> "="
- *                            <initializer>
+ *                          | <init_declarator_list> "," <identifier> "[" <constant_expression> "]"
+ *                          | <init_declarator_list> "," <identifier> "=" <initializer>
  */
 init_declarator_list
     single_declaration .and .loop init_declarator_list_1 .emit DECLARATOR_NEXT .and
@@ -860,10 +891,9 @@ init_declarator_list_5
  * <single_declaration> ::= <fully_specified_type>
  *                        | <fully_specified_type> <identifier>
  *                        | <fully_specified_type> <identifier> "[" "]"
- *                        | <fully_specified_type> <identifier> "["
- *                          <constant_expression> "]"
+ *                        | <fully_specified_type> <identifier> "[" <constant_expression> "]"
  *                        | <fully_specified_type> <identifier> "=" <initializer>
-*/
+ */
 single_declaration
     single_declaration_nospace .or single_declaration_space;
 single_declaration_space
@@ -924,12 +954,12 @@ centroid_qualifier
 
 
 /*
- *  <type_qualifier> ::= "const"
- *                     | "attribute" // Vertex only.
- *                     | "varying"
- *                     | "uniform"
- *                     | "__fixed_output"
- *                     | "__fixed_input"
+ * <type_qualifier> ::= "const"
+ *                    | "attribute" // Vertex only.
+ *                    | "varying"
+ *                    | "uniform"
+ *                    | "__fixed_output"
+ *                    | "__fixed_input"
  *
  * note: this is an extension to the standard language specification,
  * normally slang disallows __fixed_output and __fixed_input type qualifiers
@@ -943,40 +973,40 @@ type_qualifier
     .if (parsing_builtin != 0) "__fixed_input" .emit TYPE_QUALIFIER_FIXEDINPUT;
 
 /*
- * <type_specifier> ::= "void"
- *                    | "float"
- *                    | "int"
- *                    | "bool"
- *                    | "vec2"
- *                    | "vec3"
- *                    | "vec4"
- *                    | "bvec2"
- *                    | "bvec3"
- *                    | "bvec4"
- *                    | "ivec2"
- *                    | "ivec3"
- *                    | "ivec4"
- *                    | "mat2"
- *                    | "mat3"
- *                    | "mat4"
- *                    | "mat2x3"
- *                    | "mat3x2"
- *                    | "mat2x4"
- *                    | "mat4x2"
- *                    | "mat3x4"
- *                    | "mat4x3"
- *                    | "sampler1D"
- *                    | "sampler2D"
- *                    | "sampler3D"
- *                    | "samplerCube"
- *                    | "sampler1DShadow"
- *                    | "sampler2DShadow"
- *                    | "sampler2DRect"
- *                    | "sampler2DRectShadow"
- *                    | <struct_specifier>
- *                    | <type_name>
- */
-type_specifier_space
+ * <type_specifier_nonarray> ::= "void"
+ *                             | "float"
+ *                             | "int"
+ *                             | "bool"
+ *                             | "vec2"
+ *                             | "vec3"
+ *                             | "vec4"
+ *                             | "bvec2"
+ *                             | "bvec3"
+ *                             | "bvec4"
+ *                             | "ivec2"
+ *                             | "ivec3"
+ *                             | "ivec4"
+ *                             | "mat2"
+ *                             | "mat3"
+ *                             | "mat4"
+ *                             | "mat2x3"
+ *                             | "mat3x2"
+ *                             | "mat2x4"
+ *                             | "mat4x2"
+ *                             | "mat3x4"
+ *                             | "mat4x3"
+ *                             | "sampler1D"
+ *                             | "sampler2D"
+ *                             | "sampler3D"
+ *                             | "samplerCube"
+ *                             | "sampler1DShadow"
+ *                             | "sampler2DShadow"
+ *                             | "sampler2DRect"
+ *                             | "sampler2DRectShadow"
+ *                             | <struct_specifier>
+ *                             | <type_name>
+ */
+type_specifier_nonarray_space
     "void" .emit TYPE_SPECIFIER_VOID .or
     "float" .emit TYPE_SPECIFIER_FLOAT .or
     "int" .emit TYPE_SPECIFIER_INT .or
@@ -1008,12 +1038,27 @@ type_specifier_space
     "sampler2DRect" .emit TYPE_SPECIFIER_SAMPLER2DRECT .or
     "sampler2DRectShadow" .emit TYPE_SPECIFIER_SAMPLER2DRECTSHADOW .or
     type_name .emit TYPE_SPECIFIER_TYPENAME;
-type_specifier_nospace
+type_specifier_nonarray_nospace
     struct_specifier .emit TYPE_SPECIFIER_STRUCT;
+type_specifier_nonarray
+    type_specifier_nonarray_nospace .or type_specifier_nonarray_space;
+
+/*
+ * <type_specifier> ::= <type_specifier_nonarray>
+ *                    | <type_specifier_nonarray> "[" <constant_expression> "]"
+ */
+type_specifier_space
+    type_specifier_nonarray_space .and .true .emit TYPE_SPECIFIER_NONARRAY;
+type_specifier_nospace
+    type_specifier_nospace_array .or type_specifier_nospace_1;
+type_specifier_nospace_1
+    type_specifier_nonarray_nospace .and .true .emit TYPE_SPECIFIER_NONARRAY;
+type_specifier_nospace_array
+    type_specifier_nonarray .and lbracket .emit TYPE_SPECIFIER_ARRAY .and constant_expression .and rbracket;
 
 /*
  * <struct_specifier> ::= "struct" <identifier> "{" <struct_declaration_list> "}"
- *                       | "struct" "{" <struct_declaration_list> "}"
+ *                      | "struct" "{" <struct_declaration_list> "}"
  */
 struct_specifier
     "struct" .and struct_specifier_1 .and optional_space .and lbrace .error LBRACE_EXPECTED .and
@@ -1025,7 +1070,7 @@ struct_specifier_2
 
 /*
  * <struct_declaration_list> ::= <struct_declaration>
- *                              | <struct_declaration_list> <struct_declaration>
+ *                             | <struct_declaration_list> <struct_declaration>
  */
 struct_declaration_list
     struct_declaration .and .loop struct_declaration .emit FIELD_NEXT;
@@ -1042,7 +1087,7 @@ struct_declaration_nospace
 
 /*
  * <struct_declarator_list> ::= <struct_declarator>
- *                             | <struct_declarator_list> "," <struct_declarator>
+ *                            | <struct_declarator_list> "," <struct_declarator>
  */
 struct_declarator_list
     struct_declarator .and .loop struct_declarator_list_1 .emit FIELD_NEXT;
@@ -1117,13 +1162,6 @@ compound_statement_3
     lbrace .and statement_list .and rbrace;
 
 /*
- * <statement_no_new_scope> ::= <compound_statement_no_new_scope>
- *                            | <simple_statement>
- */
-statement_no_new_scope
-    compound_statement_no_new_scope .or simple_statement;
-
-/*
  * <compound_statement_no_new_scope> ::= "{" "}"
  *                                     | "{" <statement_list> "}"
  */
@@ -1136,6 +1174,7 @@ compound_statement_no_new_scope_2
 compound_statement_no_new_scope_3
     lbrace .and statement_list .and rbrace;
 
+
 /*
  * <statement_list> ::= <statement>
  *                    | <statement_list> <statement>
@@ -1197,8 +1236,7 @@ condition_3
 /*
  * <iteration_statement> ::= "while" "(" <condition> ")" <statement>
  *                         | "do" <statement> "while" "(" <expression> ")" ";"
- *                         | "for" "(" <for_init_statement> <for_rest_statement> ")"
- *                            <statement_no_new_scope>
+ *                         | "for" "(" <for_init_statement> <for_rest_statement> ")" <statement>
  */
 iteration_statement
     iteration_statement_1 .or iteration_statement_2 .or iteration_statement_3;
@@ -1210,7 +1248,7 @@ iteration_statement_2
     expression .and rparen .error RPAREN_EXPECTED .emit OP_END .and semicolon;
 iteration_statement_3
     "for" .emit OP_FOR .and lparen .error LPAREN_EXPECTED .and for_init_statement .and
-    for_rest_statement .and rparen .error RPAREN_EXPECTED .and statement_no_new_scope;
+    for_rest_statement .and rparen .error RPAREN_EXPECTED .and statement;
 
 /*
  * <for_init_statement> ::= <expression_statement>
diff --git a/src/mesa/shader/slang/library/slang_shader_syn.h b/src/mesa/shader/slang/library/slang_shader_syn.h
index f1fc9ca883..6a382970e1 100644
--- a/src/mesa/shader/slang/library/slang_shader_syn.h
+++ b/src/mesa/shader/slang/library/slang_shader_syn.h
@@ -2,7 +2,7 @@
 /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE .syn FILE */
 
 ".syntax translation_unit;\n"
-".emtcode REVISION 4\n"
+".emtcode REVISION 5\n"
 ".emtcode EXTERNAL_NULL 0\n"
 ".emtcode EXTERNAL_FUNCTION_DEFINITION 1\n"
 ".emtcode EXTERNAL_DECLARATION 2\n"
@@ -17,6 +17,8 @@
 ".emtcode FUNCTION_ORDINARY 0\n"
 ".emtcode FUNCTION_CONSTRUCTOR 1\n"
 ".emtcode FUNCTION_OPERATOR 2\n"
+".emtcode FUNCTION_CALL_NONARRAY 0\n"
+".emtcode FUNCTION_CALL_ARRAY 1\n"
 ".emtcode OPERATOR_ADDASSIGN 1\n"
 ".emtcode OPERATOR_SUBASSIGN 2\n"
 ".emtcode OPERATOR_MULASSIGN 3\n"
@@ -83,6 +85,8 @@
 ".emtcode TYPE_SPECIFIER_MAT42 29\n"
 ".emtcode TYPE_SPECIFIER_MAT34 30\n"
 ".emtcode TYPE_SPECIFIER_MAT43 31\n"
+".emtcode TYPE_SPECIFIER_NONARRAY 0\n"
+".emtcode TYPE_SPECIFIER_ARRAY 1\n"
 ".emtcode FIELD_NONE 0\n"
 ".emtcode FIELD_NEXT 1\n"
 ".emtcode FIELD_ARRAY 2\n"
@@ -136,6 +140,7 @@
 ".emtcode OP_POSTINCREMENT 60\n"
 ".emtcode OP_POSTDECREMENT 61\n"
 ".emtcode OP_PRECISION 62\n"
+".emtcode OP_METHOD 63\n"
 ".emtcode PARAM_QUALIFIER_IN 0\n"
 ".emtcode PARAM_QUALIFIER_OUT 1\n"
 ".emtcode PARAM_QUALIFIER_INOUT 2\n"
@@ -173,6 +178,12 @@
 "integer_expression\n"
 " expression;\n"
 "function_call\n"
+" function_call_or_method;\n"
+"function_call_or_method\n"
+" regular_function_call .or method_call;\n"
+"method_call\n"
+" identifier .emit OP_METHOD .and dot .and function_call_generic .and .true .emit OP_END;\n"
+"regular_function_call\n"
 " function_call_generic .emit OP_CALL .and .true .emit OP_END;\n"
 "function_call_generic\n"
 " function_call_generic_1 .or function_call_generic_2;\n"
@@ -192,7 +203,12 @@
 "function_call_header\n"
 " function_identifier .and lparen;\n"
 "function_identifier\n"
-" identifier;\n"
+" identifier .and function_identifier_opt_array;\n"
+"function_identifier_opt_array\n"
+" function_identifier_array .emit FUNCTION_CALL_ARRAY .or\n"
+" .true .emit FUNCTION_CALL_NONARRAY;\n"
+"function_identifier_array\n"
+" lbracket .and constant_expression .and rbracket;\n"
 "unary_expression\n"
 " postfix_expression .or unary_expression_1 .or unary_expression_2 .or unary_expression_3 .or\n"
 " unary_expression_4 .or unary_expression_5;\n"
@@ -455,7 +471,7 @@
 " \"uniform\" .emit TYPE_QUALIFIER_UNIFORM .or\n"
 " .if (parsing_builtin != 0) \"__fixed_output\" .emit TYPE_QUALIFIER_FIXEDOUTPUT .or\n"
 " .if (parsing_builtin != 0) \"__fixed_input\" .emit TYPE_QUALIFIER_FIXEDINPUT;\n"
-"type_specifier_space\n"
+"type_specifier_nonarray_space\n"
 " \"void\" .emit TYPE_SPECIFIER_VOID .or\n"
 " \"float\" .emit TYPE_SPECIFIER_FLOAT .or\n"
 " \"int\" .emit TYPE_SPECIFIER_INT .or\n"
@@ -487,8 +503,18 @@
 " \"sampler2DRect\" .emit TYPE_SPECIFIER_SAMPLER2DRECT .or\n"
 " \"sampler2DRectShadow\" .emit TYPE_SPECIFIER_SAMPLER2DRECTSHADOW .or\n"
 " type_name .emit TYPE_SPECIFIER_TYPENAME;\n"
-"type_specifier_nospace\n"
+"type_specifier_nonarray_nospace\n"
 " struct_specifier .emit TYPE_SPECIFIER_STRUCT;\n"
+"type_specifier_nonarray\n"
+" type_specifier_nonarray_nospace .or type_specifier_nonarray_space;\n"
+"type_specifier_space\n"
+" type_specifier_nonarray_space .and .true .emit TYPE_SPECIFIER_NONARRAY;\n"
+"type_specifier_nospace\n"
+" type_specifier_nospace_array .or type_specifier_nospace_1;\n"
+"type_specifier_nospace_1\n"
+" type_specifier_nonarray_nospace .and .true .emit TYPE_SPECIFIER_NONARRAY;\n"
+"type_specifier_nospace_array\n"
+" type_specifier_nonarray .and lbracket .emit TYPE_SPECIFIER_ARRAY .and constant_expression .and rbracket;\n"
 "struct_specifier\n"
 " \"struct\" .and struct_specifier_1 .and optional_space .and lbrace .error LBRACE_EXPECTED .and\n"
 " struct_declaration_list .and rbrace .emit FIELD_NONE;\n"
@@ -540,8 +566,6 @@
 " lbrace .and rbrace;\n"
 "compound_statement_3\n"
 " lbrace .and statement_list .and rbrace;\n"
-"statement_no_new_scope\n"
-" compound_statement_no_new_scope .or simple_statement;\n"
 "compound_statement_no_new_scope\n"
 " compound_statement_no_new_scope_1 .emit OP_BLOCK_BEGIN_NO_NEW_SCOPE .and .true .emit OP_END;\n"
 "compound_statement_no_new_scope_1\n"
@@ -591,7 +615,7 @@
 " expression .and rparen .error RPAREN_EXPECTED .emit OP_END .and semicolon;\n"
 "iteration_statement_3\n"
 " \"for\" .emit OP_FOR .and lparen .error LPAREN_EXPECTED .and for_init_statement .and\n"
-" for_rest_statement .and rparen .error RPAREN_EXPECTED .and statement_no_new_scope;\n"
+" for_rest_statement .and rparen .error RPAREN_EXPECTED .and statement;\n"
 "for_init_statement\n"
 " expression_statement .emit OP_EXPRESSION .or declaration_statement .emit OP_DECLARE;\n"
 "conditionopt\n"
diff --git a/src/mesa/shader/slang/library/slang_version_syn.h b/src/mesa/shader/slang/library/slang_version_syn.h
deleted file mode 100644
index 3b94d85927..0000000000
--- a/src/mesa/shader/slang/library/slang_version_syn.h
+++ /dev/null
@@ -1,64 +0,0 @@
-".syntax version_directive;\n"
-"version_directive\n"
-"	version_directive_1 .and .loop version_directive_2;\n"
-"version_directive_1\n"
-"	prior_optional_spaces .and optional_version_directive .and .true .emit $;\n"
-"version_directive_2\n"
-"	prior_optional_spaces .and version_directive_body .and .true .emit $;\n"
-"optional_version_directive\n"
-"	version_directive_body .or .true .emit 10 .emit 1;\n"
-"version_directive_body\n"
-"	'#' .and optional_space .and \"version\" .and space .and version_number .and optional_space .and\n"
-"	new_line;\n"
-"version_number\n"
-"	version_number_110;\n"
-"version_number_110\n"
-"	leading_zeroes .and \"110\" .emit 10 .emit 1;\n"
-"leading_zeroes\n"
-"	.loop zero;\n"
-"zero\n"
-"	'0';\n"
-"space\n"
-" single_space .and .loop single_space;\n"
-"optional_space\n"
-" .loop single_space;\n"
-"single_space\n"
-" ' ' .or '\\t';\n"
-"prior_optional_spaces\n"
-"	.loop prior_space;\n"
-"prior_space\n"
-"	c_style_comment_block .or cpp_style_comment_block .or space .or new_line;\n"
-"c_style_comment_block\n"
-" '/' .and '*' .and c_style_comment_rest;\n"
-"c_style_comment_rest\n"
-" .loop c_style_comment_char_no_star .and c_style_comment_rest_1;\n"
-"c_style_comment_rest_1\n"
-" c_style_comment_end .or c_style_comment_rest_2;\n"
-"c_style_comment_rest_2\n"
-" '*' .and c_style_comment_rest;\n"
-"c_style_comment_char_no_star\n"
-" '\\x2B'-'\\xFF' .or '\\x01'-'\\x29';\n"
-"c_style_comment_end\n"
-" '*' .and '/';\n"
-"cpp_style_comment_block\n"
-" '/' .and '/' .and cpp_style_comment_block_1;\n"
-"cpp_style_comment_block_1\n"
-" cpp_style_comment_block_2 .or cpp_style_comment_block_3;\n"
-"cpp_style_comment_block_2\n"
-" .loop cpp_style_comment_char .and new_line;\n"
-"cpp_style_comment_block_3\n"
-" .loop cpp_style_comment_char;\n"
-"cpp_style_comment_char\n"
-" '\\x0E'-'\\xFF' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n"
-"new_line\n"
-" cr_lf .or lf_cr .or '\\n' .or '\\r';\n"
-"cr_lf\n"
-" '\\r' .and '\\n';\n"
-"lf_cr\n"
-"	'\\n' .and '\\r';\n"
-".string __string_filter;\n"
-"__string_filter\n"
-" .loop __identifier_char;\n"
-"__identifier_char\n"
-" 'a'-'z' .or 'A'-'Z' .or '_' .or '0'-'9';\n"
-""
diff --git a/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h b/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h
index c8e730072f..2cb91bf268 100644
--- a/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h
+++ b/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h
@@ -2,103 +2,105 @@
 /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */
 /* slang_vertex_builtin.gc */
 
-4,2,2,90,95,5,0,12,1,103,108,95,80,111,115,105,116,105,111,110,0,0,0,2,2,90,95,5,0,9,1,103,108,95,
-80,111,105,110,116,83,105,122,101,0,0,0,2,2,90,95,5,0,12,1,103,108,95,67,108,105,112,86,101,114,
-116,101,120,0,0,0,2,2,90,95,2,0,12,1,103,108,95,67,111,108,111,114,0,0,0,2,2,90,95,2,0,12,1,103,
-108,95,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,90,95,2,0,11,1,103,108,95,78,
-111,114,109,97,108,0,0,0,2,2,90,95,2,0,12,1,103,108,95,86,101,114,116,101,120,0,0,0,2,2,90,95,2,0,
-12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,48,0,0,0,2,2,90,95,2,0,12,1,103,
-108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,49,0,0,0,2,2,90,95,2,0,12,1,103,108,95,77,
-117,108,116,105,84,101,120,67,111,111,114,100,50,0,0,0,2,2,90,95,2,0,12,1,103,108,95,77,117,108,
-116,105,84,101,120,67,111,111,114,100,51,0,0,0,2,2,90,95,2,0,12,1,103,108,95,77,117,108,116,105,84,
-101,120,67,111,111,114,100,52,0,0,0,2,2,90,95,2,0,12,1,103,108,95,77,117,108,116,105,84,101,120,67,
-111,111,114,100,53,0,0,0,2,2,90,95,2,0,12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,
-114,100,54,0,0,0,2,2,90,95,2,0,12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,55,
-0,0,0,2,2,90,95,2,0,9,1,103,108,95,70,111,103,67,111,111,114,100,0,0,0,2,2,90,95,3,0,12,1,103,108,
-95,70,114,111,110,116,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,1,103,108,95,66,97,99,107,67,111,
-108,111,114,0,0,0,2,2,90,95,3,0,12,1,103,108,95,70,114,111,110,116,83,101,99,111,110,100,97,114,
-121,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,1,103,108,95,66,97,99,107,83,101,99,111,110,100,97,
-114,121,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,1,103,108,95,84,101,120,67,111,111,114,100,0,3,
-18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,90,95,3,0,9,1,
-103,108,95,70,111,103,70,114,97,103,67,111,111,114,100,0,0,0,1,90,95,0,0,12,0,102,116,114,97,110,
-115,102,111,114,109,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,103,108,95,77,111,100,101,108,86,
-105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,16,8,48,0,57,18,103,108,
-95,86,101,114,116,101,120,0,59,120,120,120,120,0,48,18,103,108,95,77,111,100,101,108,86,105,101,
-119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,16,10,49,0,57,18,103,108,95,86,
-101,114,116,101,120,0,59,121,121,121,121,0,48,46,18,103,108,95,77,111,100,101,108,86,105,101,119,
-80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,16,10,50,0,57,18,103,108,95,86,101,
-114,116,101,120,0,59,122,122,122,122,0,48,46,18,103,108,95,77,111,100,101,108,86,105,101,119,80,
-114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,16,10,51,0,57,18,103,108,95,86,101,114,
-116,101,120,0,59,119,119,119,119,0,48,46,20,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,49,68,
-76,111,100,0,1,1,0,0,16,115,97,109,112,108,101,114,0,0,1,1,0,0,9,99,111,111,114,100,0,0,1,1,0,0,9,
-108,111,100,0,0,0,1,3,2,90,95,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,
-120,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,
-118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,
-101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,49,68,80,
-114,111,106,76,111,100,0,1,1,0,0,16,115,97,109,112,108,101,114,0,0,1,1,0,0,10,99,111,111,114,100,0,
-0,1,1,0,0,9,108,111,100,0,0,0,1,3,2,90,95,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,
-111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,121,0,49,20,0,9,
-18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,
-100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,
-100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,76,111,100,0,1,1,0,0,
-16,115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,111,111,114,100,0,0,1,1,0,0,9,108,111,100,0,0,0,1,
-3,2,90,95,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,
-111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,
-119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,
-108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,
-101,120,116,117,114,101,50,68,76,111,100,0,1,1,0,0,17,115,97,109,112,108,101,114,0,0,1,1,0,0,10,99,
-111,111,114,100,0,0,1,1,0,0,9,108,111,100,0,0,0,1,3,2,90,95,0,0,12,1,99,111,111,114,100,52,0,0,0,9,
-18,99,111,111,114,100,52,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,20,0,9,18,99,111,111,
-114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,
+5,2,2,90,95,5,0,12,0,1,103,108,95,80,111,115,105,116,105,111,110,0,0,0,2,2,90,95,5,0,9,0,1,103,108,
+95,80,111,105,110,116,83,105,122,101,0,0,0,2,2,90,95,5,0,12,0,1,103,108,95,67,108,105,112,86,101,
+114,116,101,120,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,67,111,108,111,114,0,0,0,2,2,90,95,2,0,12,0,
+1,103,108,95,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,90,95,2,0,11,0,1,103,
+108,95,78,111,114,109,97,108,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,86,101,114,116,101,120,0,0,0,2,
+2,90,95,2,0,12,0,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,48,0,0,0,2,2,90,95,
+2,0,12,0,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,49,0,0,0,2,2,90,95,2,0,12,0,
+1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,50,0,0,0,2,2,90,95,2,0,12,0,1,103,
+108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,51,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,
+77,117,108,116,105,84,101,120,67,111,111,114,100,52,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,77,117,
+108,116,105,84,101,120,67,111,111,114,100,53,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,77,117,108,116,
+105,84,101,120,67,111,111,114,100,54,0,0,0,2,2,90,95,2,0,12,0,1,103,108,95,77,117,108,116,105,84,
+101,120,67,111,111,114,100,55,0,0,0,2,2,90,95,2,0,9,0,1,103,108,95,70,111,103,67,111,111,114,100,0,
+0,0,2,2,90,95,3,0,12,0,1,103,108,95,70,114,111,110,116,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,0,
+1,103,108,95,66,97,99,107,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,0,1,103,108,95,70,114,111,110,
+116,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,0,1,103,108,95,66,
+97,99,107,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,90,95,3,0,12,0,1,103,108,
+95,84,101,120,67,111,111,114,100,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,
+114,100,115,0,0,0,2,2,90,95,3,0,9,0,1,103,108,95,70,111,103,70,114,97,103,67,111,111,114,100,0,0,0,
+1,90,95,0,0,12,0,0,102,116,114,97,110,115,102,111,114,109,0,0,1,9,18,95,95,114,101,116,86,97,108,0,
+18,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,
+114,105,120,0,16,8,48,0,57,18,103,108,95,86,101,114,116,101,120,0,59,120,120,120,120,0,48,18,103,
+108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,
+120,0,16,10,49,0,57,18,103,108,95,86,101,114,116,101,120,0,59,121,121,121,121,0,48,46,18,103,108,
+95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,
+16,10,50,0,57,18,103,108,95,86,101,114,116,101,120,0,59,122,122,122,122,0,48,46,18,103,108,95,77,
+111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,16,10,
+51,0,57,18,103,108,95,86,101,114,116,101,120,0,59,119,119,119,119,0,48,46,20,0,0,1,90,95,0,0,12,0,
+0,116,101,120,116,117,114,101,49,68,76,111,100,0,1,1,0,0,16,0,115,97,109,112,108,101,114,0,0,1,1,0,
+0,9,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,
+100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,
+114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,
 114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,
-95,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,76,111,100,0,1,1,0,0,17,115,97,109,
-112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,1,1,0,0,9,108,111,100,0,0,0,1,3,2,90,95,0,0,
-12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,
-0,59,120,121,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,
-108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,
-18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,
-116,117,114,101,50,68,80,114,111,106,76,111,100,0,1,1,0,0,17,115,97,109,112,108,101,114,0,0,1,1,0,
-0,12,99,111,111,114,100,0,0,1,1,0,0,9,108,111,100,0,0,0,1,3,2,90,95,0,0,12,1,112,99,111,111,114,
-100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,
-111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,
-118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,
-101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,51,68,
-76,111,100,0,1,1,0,0,18,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,1,1,0,0,9,
-108,111,100,0,0,0,1,3,2,90,95,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,
-120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,
-18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97,108,0,
-0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,116,101,120,
-116,117,114,101,51,68,80,114,111,106,76,111,100,0,1,1,0,0,18,115,97,109,112,108,101,114,0,0,1,1,0,
-0,12,99,111,111,114,100,0,0,1,1,0,0,9,108,111,100,0,0,0,1,3,2,90,95,0,0,12,1,112,99,111,111,114,
-100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,
-18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,
-20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,
-112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,116,101,120,116,117,114,101,
-67,117,98,101,76,111,100,0,1,1,0,0,19,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,
-0,0,1,1,0,0,9,108,111,100,0,0,0,1,3,2,90,95,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,
-114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,
-18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,
-108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,115,
-104,97,100,111,119,49,68,76,111,100,0,1,1,0,0,20,115,97,109,112,108,101,114,0,0,1,1,0,0,11,99,111,
-111,114,100,0,0,1,1,0,0,9,108,111,100,0,0,0,1,3,2,90,95,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,
-99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,
-59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,
-97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,115,
-104,97,100,111,119,49,68,80,114,111,106,76,111,100,0,1,1,0,0,20,115,97,109,112,108,101,114,0,0,1,1,
-0,0,12,99,111,111,114,100,0,0,1,1,0,0,9,108,111,100,0,0,0,1,3,2,90,95,0,0,12,1,112,99,111,111,114,
-100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,
-114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,
-20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,
-98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,
-114,100,0,0,0,0,1,90,95,0,0,12,0,115,104,97,100,111,119,50,68,76,111,100,0,1,1,0,0,21,115,97,109,
-112,108,101,114,0,0,1,1,0,0,11,99,111,111,114,100,0,0,1,1,0,0,9,108,111,100,0,0,0,1,3,2,90,95,0,0,
-12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,
-100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,
-120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,
-114,100,52,0,0,0,0,1,90,95,0,0,12,0,115,104,97,100,111,119,50,68,80,114,111,106,76,111,100,0,1,1,0,
-0,21,115,97,109,112,108,101,114,0,0,1,1,0,0,12,99,111,111,114,100,0,0,1,1,0,0,9,108,111,100,0,0,0,
-1,3,2,90,95,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,
-111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,
-100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,
+95,0,0,12,0,0,116,101,120,116,117,114,101,49,68,80,114,111,106,76,111,100,0,1,1,0,0,16,0,115,97,
+109,112,108,101,114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,
+95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,
+114,100,0,59,120,0,18,99,111,111,114,100,0,59,121,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,
+18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,
+0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,
+120,116,117,114,101,49,68,80,114,111,106,76,111,100,0,1,1,0,0,16,0,115,97,109,112,108,101,114,0,0,
+1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,
+111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,
+99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,
+4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,
+108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,101,
+50,68,76,111,100,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,0,0,10,0,99,111,111,114,100,0,0,
+1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,
+114,100,52,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,20,0,9,18,99,111,111,114,100,52,0,
+59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,
+97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,
+116,101,120,116,117,114,101,50,68,80,114,111,106,76,111,100,0,1,1,0,0,17,0,115,97,109,112,108,101,
+114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,
+112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,
+120,121,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,
 111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,
-115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,0
+115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,
+116,117,114,101,50,68,80,114,111,106,76,111,100,0,1,1,0,0,17,0,115,97,109,112,108,101,114,0,0,1,1,
+0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,
+111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,
+18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,
+20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,
+112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,116,101,120,116,117,114,
+101,51,68,76,111,100,0,1,1,0,0,18,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,
+0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,
+111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,20,0,9,18,99,111,111,
+114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,
+114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,
+95,0,0,12,0,0,116,101,120,116,117,114,101,51,68,80,114,111,106,76,111,100,0,1,1,0,0,18,0,115,97,
+109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,
+95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,122,0,18,99,
+111,111,114,100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,
+114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,
+101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,
+0,0,12,0,0,116,101,120,116,117,114,101,67,117,98,101,76,111,100,0,1,1,0,0,19,0,115,97,109,112,108,
+101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,
+1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,
+0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,
+99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,
+114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,49,68,76,111,100,0,1,1,0,0,20,0,115,
+97,109,112,108,101,114,0,0,1,1,0,0,11,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,
+90,95,0,0,12,0,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,
+111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,
+95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,
+99,111,111,114,100,52,0,0,0,0,1,90,95,0,0,12,0,0,115,104,97,100,111,119,49,68,80,114,111,106,76,
+111,100,0,1,1,0,0,20,0,115,97,109,112,108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,
+9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,
+100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,
+99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,
+59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,
+97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,90,95,0,0,12,0,0,
+115,104,97,100,111,119,50,68,76,111,100,0,1,1,0,0,21,0,115,97,109,112,108,101,114,0,0,1,1,0,0,11,0,
+99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,12,0,1,99,111,111,114,100,52,
+0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,
+114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,
+114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,90,
+95,0,0,12,0,0,115,104,97,100,111,119,50,68,80,114,111,106,76,111,100,0,1,1,0,0,21,0,115,97,109,112,
+108,101,114,0,0,1,1,0,0,12,0,99,111,111,114,100,0,0,1,1,0,0,9,0,108,111,100,0,0,0,1,3,2,90,95,0,0,
+12,0,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,
+100,0,59,120,121,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,
+18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,
+4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,
+108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,0
diff --git a/src/mesa/shader/slang/slang_builtin.c b/src/mesa/shader/slang/slang_builtin.c
index db00c54b8a..c0f4c79e13 100644
--- a/src/mesa/shader/slang/slang_builtin.c
+++ b/src/mesa/shader/slang/slang_builtin.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.3
+ * Version:  7.3
  *
  * Copyright (C) 2005-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2008  VMware, Inc.   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -39,6 +40,10 @@
 #include "shader/slang/slang_builtin.h"
 
 
+/** special state token (see below) */
+#define STATE_ARRAY ((gl_state_index) 0xfffff)
+
+
 /**
  * Lookup GL state given a variable name, 0, 1 or 2 indexes and a field.
  * Allocate room for the state in the given param list and return position
@@ -132,6 +137,8 @@ lookup_statevar(const char *var, GLint index1, GLint index2, const char *field,
       }
    }
    else if (strcmp(var, "gl_ClipPlane") == 0) {
+      if (index1 < 0)
+         return -1;
       tokens[0] = STATE_CLIPPLANE;
       tokens[1] = index1;
    }
@@ -196,8 +203,12 @@ lookup_statevar(const char *var, GLint index1, GLint index2, const char *field,
       }
    }
    else if (strcmp(var, "gl_LightSource") == 0) {
+      if (!field || index1 < 0)
+         return -1;
+
       tokens[0] = STATE_LIGHT;
       tokens[1] = index1;
+
       if (strcmp(field, "ambient") == 0) {
          tokens[2] = STATE_AMBIENT;
       }
@@ -272,6 +283,9 @@ lookup_statevar(const char *var, GLint index1, GLint index2, const char *field,
    }
    else if (strcmp(var, "gl_FrontLightProduct") == 0 ||
             strcmp(var, "gl_BackLightProduct") == 0) {
+      if (index1 < 0 || !field)
+         return -1;
+
       tokens[0] = STATE_LIGHTPROD;
       tokens[1] = index1; /* light number */
       if (strcmp(var, "gl_FrontLightProduct") == 0) {
@@ -294,45 +308,63 @@ lookup_statevar(const char *var, GLint index1, GLint index2, const char *field,
       }
    }
    else if (strcmp(var, "gl_TextureEnvColor") == 0) {
+      if (index1 < 0)
+         return -1;
       tokens[0] = STATE_TEXENV_COLOR;
       tokens[1] = index1;
    }
    else if (strcmp(var, "gl_EyePlaneS") == 0) {
+      if (index1 < 0)
+         return -1;
       tokens[0] = STATE_TEXGEN;
       tokens[1] = index1; /* tex unit */
       tokens[2] = STATE_TEXGEN_EYE_S;
    }
    else if (strcmp(var, "gl_EyePlaneT") == 0) {
+      if (index1 < 0)
+         return -1;
       tokens[0] = STATE_TEXGEN;
       tokens[1] = index1; /* tex unit */
       tokens[2] = STATE_TEXGEN_EYE_T;
    }
    else if (strcmp(var, "gl_EyePlaneR") == 0) {
+      if (index1 < 0)
+         return -1;
       tokens[0] = STATE_TEXGEN;
       tokens[1] = index1; /* tex unit */
       tokens[2] = STATE_TEXGEN_EYE_R;
    }
    else if (strcmp(var, "gl_EyePlaneQ") == 0) {
+      if (index1 < 0)
+         return -1;
       tokens[0] = STATE_TEXGEN;
       tokens[1] = index1; /* tex unit */
       tokens[2] = STATE_TEXGEN_EYE_Q;
    }
    else if (strcmp(var, "gl_ObjectPlaneS") == 0) {
+      if (index1 < 0)
+         return -1;
       tokens[0] = STATE_TEXGEN;
       tokens[1] = index1; /* tex unit */
       tokens[2] = STATE_TEXGEN_OBJECT_S;
    }
    else if (strcmp(var, "gl_ObjectPlaneT") == 0) {
+      if (index1 < 0)
+         return -1;
       tokens[0] = STATE_TEXGEN;
       tokens[1] = index1; /* tex unit */
       tokens[2] = STATE_TEXGEN_OBJECT_T;
    }
    else if (strcmp(var, "gl_ObjectPlaneR") == 0) {
+      if (index1 < 0)
+         return -1;
       tokens[0] = STATE_TEXGEN;
       tokens[1] = index1; /* tex unit */
       tokens[2] = STATE_TEXGEN_OBJECT_R;
    }
    else if (strcmp(var, "gl_ObjectPlaneQ") == 0) {
+      if (index1 < 0)
+         return -1;
       tokens[0] = STATE_TEXGEN;
       tokens[1] = index1; /* tex unit */
       tokens[2] = STATE_TEXGEN_OBJECT_Q;
@@ -386,13 +418,222 @@ lookup_statevar(const char *var, GLint index1, GLint index2, const char *field,
 }
 
 
+
+/**
+ * Given a variable name and datatype, emit uniform/constant buffer
+ * entries which will store that state variable.
+ * For example, if name="gl_LightSource" we'll emit 64 state variable
+ * vectors/references and return position where that data starts.  This will
+ * allow run-time array indexing into the light source array.
+ *
+ * Note that this is a recursive function.
+ *
+ * \return -1 if error, else index of start of data in the program parameter list
+ */
+static GLint
+emit_statevars(const char *name, int array_len,
+               const slang_type_specifier *type,
+               gl_state_index tokens[STATE_LENGTH],
+               struct gl_program_parameter_list *paramList)
+{
+   if (type->type == SLANG_SPEC_ARRAY) {
+      GLint i, pos;
+      assert(array_len > 0);
+      if (strcmp(name, "gl_ClipPlane") == 0) {
+         tokens[0] = STATE_CLIPPLANE;
+      }
+      else if (strcmp(name, "gl_LightSource") == 0) {
+         tokens[0] = STATE_LIGHT;
+      }
+      else if (strcmp(name, "gl_FrontLightProduct") == 0) {
+         tokens[0] = STATE_LIGHTPROD;
+         tokens[2] = 0; /* front */
+      }
+      else if (strcmp(name, "gl_BackLightProduct") == 0) {
+         tokens[0] = STATE_LIGHTPROD;
+         tokens[2] = 1; /* back */
+      }
+      else if (strcmp(name, "gl_TextureEnvColor") == 0) {
+         tokens[0] = STATE_TEXENV_COLOR;
+      }
+      else if (strcmp(name, "gl_EyePlaneS") == 0) {
+         tokens[0] = STATE_TEXGEN_EYE_S;
+      }
+      else if (strcmp(name, "gl_EyePlaneT") == 0) {
+         tokens[0] = STATE_TEXGEN_EYE_T;
+      }
+      else if (strcmp(name, "gl_EyePlaneR") == 0) {
+         tokens[0] = STATE_TEXGEN_EYE_R;
+      }
+      else if (strcmp(name, "gl_EyePlaneQ") == 0) {
+         tokens[0] = STATE_TEXGEN_EYE_Q;
+      }
+      else if (strcmp(name, "gl_ObjectPlaneS") == 0) {
+         tokens[0] = STATE_TEXGEN_OBJECT_S;
+      }
+      else if (strcmp(name, "gl_ObjectPlaneT") == 0) {
+         tokens[0] = STATE_TEXGEN_OBJECT_T;
+      }
+      else if (strcmp(name, "gl_ObjectPlaneR") == 0) {
+         tokens[0] = STATE_TEXGEN_OBJECT_R;
+      }
+      else if (strcmp(name, "gl_ObjectPlaneQ") == 0) {
+         tokens[0] = STATE_TEXGEN_OBJECT_Q;
+      }
+      else {
+         return -1; /* invalid array name */
+      }
+      for (i = 0; i < array_len; i++) {
+         GLint p;
+         tokens[1] = i;
+         p = emit_statevars(NULL, 0, type->_array, tokens, paramList);
+         if (i == 0)
+            pos = p;
+      }
+      return pos;
+   }
+   else if (type->type == SLANG_SPEC_STRUCT) {
+      const slang_variable_scope *fields = type->_struct->fields;
+      GLuint i, pos;
+      for (i = 0; i < fields->num_variables; i++) {
+         const slang_variable *var = fields->variables[i];
+         GLint p = emit_statevars(var->a_name, 0, &var->type.specifier,
+                                  tokens, paramList);
+         if (i == 0)
+            pos = p;
+      }
+      return pos;
+   }
+   else {
+      GLint pos;
+      assert(type->type == SLANG_SPEC_VEC4 ||
+             type->type == SLANG_SPEC_VEC3 ||
+             type->type == SLANG_SPEC_VEC2 ||
+             type->type == SLANG_SPEC_FLOAT ||
+             type->type == SLANG_SPEC_IVEC4 ||
+             type->type == SLANG_SPEC_IVEC3 ||
+             type->type == SLANG_SPEC_IVEC2 ||
+             type->type == SLANG_SPEC_INT);
+      if (name) {
+         GLint t;
+
+         if (tokens[0] == STATE_LIGHT)
+            t = 2;
+         else if (tokens[0] == STATE_LIGHTPROD)
+            t = 3;
+         else
+            return -1; /* invalid array name */
+
+         if (strcmp(name, "ambient") == 0) {
+            tokens[t] = STATE_AMBIENT;
+         }
+         else if (strcmp(name, "diffuse") == 0) {
+            tokens[t] = STATE_DIFFUSE;
+         }
+         else if (strcmp(name, "specular") == 0) {
+            tokens[t] = STATE_SPECULAR;
+         }
+         else if (strcmp(name, "position") == 0) {
+            tokens[t] = STATE_POSITION;
+         }
+         else if (strcmp(name, "halfVector") == 0) {
+            tokens[t] = STATE_HALF_VECTOR;
+         }
+         else if (strcmp(name, "spotDirection") == 0) {
+            tokens[t] = STATE_SPOT_DIRECTION; /* xyz components */
+         }
+         else if (strcmp(name, "spotCosCutoff") == 0) {
+            tokens[t] = STATE_SPOT_DIRECTION; /* w component */
+         }
+
+         else if (strcmp(name, "constantAttenuation") == 0) {
+            tokens[t] = STATE_ATTENUATION; /* x component */
+         }
+         else if (strcmp(name, "linearAttenuation") == 0) {
+            tokens[t] = STATE_ATTENUATION; /* y component */
+         }
+         else if (strcmp(name, "quadraticAttenuation") == 0) {
+            tokens[t] = STATE_ATTENUATION; /* z component */
+         }
+         else if (strcmp(name, "spotExponent") == 0) {
+            tokens[t] = STATE_ATTENUATION; /* w = spot exponent */
+         }
+
+         else if (strcmp(name, "spotCutoff") == 0) {
+            tokens[t] = STATE_SPOT_CUTOFF; /* x component */
+         }
+
+         else {
+            return -1; /* invalid field name */
+         }
+      }
+
+      pos = _mesa_add_state_reference(paramList, tokens);
+      return pos;
+   }
+
+   return 1;
+}
+
+
+/**
+ * Unroll the named built-in uniform variable into a sequence of state
+ * vars in the given parameter list.
+ */
+static GLint
+alloc_state_var_array(const slang_variable *var,
+                      struct gl_program_parameter_list *paramList)
+{
+   gl_state_index tokens[STATE_LENGTH];
+   GLuint i;
+   GLint pos;
+
+   /* Initialize the state tokens array.  This is very important.
+    * When we call _mesa_add_state_reference() it'll searches the parameter
+    * list to see if the given statevar token sequence is already present.
+    * This is normally a good thing since it prevents redundant values in the
+    * constant buffer.
+    *
+    * But when we're building arrays of state this can be bad.  For example,
+    * consider this fragment of GLSL code:
+    *   foo = gl_LightSource[3].diffuse;
+    *   ...
+    *   bar = gl_LightSource[i].diffuse;
+    *
+    * When we unroll the gl_LightSource array (for "bar") we want to re-emit
+    * gl_LightSource[3].diffuse and not re-use the first instance (from "foo")
+    * since that would upset the array layout.  We handle this situation by
+    * setting the last token in the state var token array to the special
+    * value STATE_ARRAY.
+    * This token will only be set for array state.  We can hijack the last
+    * element in the array for this since it's never used for light, clipplane
+    * or texture env array state.
+    */
+   for (i = 0; i < STATE_LENGTH; i++)
+      tokens[i] = 0;
+   tokens[STATE_LENGTH - 1] = STATE_ARRAY;
+
+   pos = emit_statevars(var->a_name, var->array_len, &var->type.specifier,
+                        tokens, paramList);
+
+   return pos;
+}
+
+
+
 /**
  * Allocate storage for a pre-defined uniform (a GL state variable).
  * As a memory-saving optimization, we try to only allocate storage for
  * state vars that are actually used.
- * For example, the "gl_LightSource" uniform is huge.  If we only use
- * a handful of gl_LightSource fields, we don't want to allocate storage
- * for all of gl_LightSource.
+ *
+ * Arrays such as gl_LightSource are handled specially.  For an expression
+ * like "gl_LightSource[2].diffuse", we can allocate a single uniform/constant
+ * slot and return the index.  In this case, we return direct=TRUE.
+ *
+ * Buf for something like "gl_LightSource[i].diffuse" we don't know the value
+ * of 'i' at compile time so we need to "unroll" the gl_LightSource array
+ * into a consecutive sequence of uniform/constant slots so it can be indexed
+ * at runtime.  In this case, we return direct=FALSE.
  *
  * Currently, all pre-defined uniforms are in one of these forms:
  *   var
@@ -401,52 +642,62 @@ lookup_statevar(const char *var, GLint index1, GLint index2, const char *field,
  *   var[i].field
  *   var[i][j]
  *
- * \return -1 upon error, else position in paramList of the state var/data
+ * \return -1 upon error, else position in paramList of the state variable/data
  */
 GLint
 _slang_alloc_statevar(slang_ir_node *n,
-                      struct gl_program_parameter_list *paramList)
+                      struct gl_program_parameter_list *paramList,
+                      GLboolean *direct)
 {
    slang_ir_node *n0 = n;
-   const char *field = NULL, *var;
-   GLint index1 = -1, index2 = -1, pos;
+   const char *field = NULL;
+   GLint index1 = -1, index2 = -1;
    GLuint swizzle;
 
+   *direct = GL_TRUE;
+
    if (n->Opcode == IR_FIELD) {
       field = n->Field;
       n = n->Children[0];
    }
 
    if (n->Opcode == IR_ELEMENT) {
-      /* XXX can only handle constant indexes for now */
       if (n->Children[1]->Opcode == IR_FLOAT) {
          index1 = (GLint) n->Children[1]->Value[0];
-         n = n->Children[0];
       }
       else {
-         return -1;
+         *direct = GL_FALSE;
       }
+      n = n->Children[0];
    }
 
    if (n->Opcode == IR_ELEMENT) {
       /* XXX can only handle constant indexes for now */
-      assert(n->Children[1]->Opcode == IR_FLOAT);
-      index2 = (GLint) n->Children[1]->Value[0];
+      if (n->Children[1]->Opcode == IR_FLOAT) {
+         index2 = (GLint) n->Children[1]->Value[0];
+      }
+      else {
+         *direct = GL_FALSE;
+      }
       n = n->Children[0];
    }
 
    assert(n->Opcode == IR_VAR);
-   var = (char *) n->Var->a_name;
 
-   pos = lookup_statevar(var, index1, index2, field, &swizzle, paramList);
-   assert(pos >= 0);
-   if (pos >= 0) {
-      /* newly resolved storage for the statevar/constant/uniform */
-      n0->Store->File = PROGRAM_STATE_VAR;
-      n0->Store->Index = pos;
-      n0->Store->Swizzle = swizzle;
-      n0->Store->Parent = NULL;
+   if (*direct) {
+      const char *var = (const char *) n->Var->a_name;
+      GLint pos =
+         lookup_statevar(var, index1, index2, field, &swizzle, paramList);
+      if (pos >= 0) {
+         /* newly resolved storage for the statevar/constant/uniform */
+         n0->Store->File = PROGRAM_STATE_VAR;
+         n0->Store->Index = pos;
+         n0->Store->Swizzle = swizzle;
+         n0->Store->Parent = NULL;
+         return pos;
+      }
    }
-   return pos;
-}
 
+   *direct = GL_FALSE;
+   return alloc_state_var_array(n->Var, paramList);
+}
diff --git a/src/mesa/shader/slang/slang_builtin.h b/src/mesa/shader/slang/slang_builtin.h
index 58629f4f7f..7f6fe80fcc 100644
--- a/src/mesa/shader/slang/slang_builtin.h
+++ b/src/mesa/shader/slang/slang_builtin.h
@@ -33,7 +33,8 @@
 
 extern GLint
 _slang_alloc_statevar(slang_ir_node *n,
-                      struct gl_program_parameter_list *paramList);
+                      struct gl_program_parameter_list *paramList,
+                      GLboolean *direct);
 
 
 #endif /* SLANG_BUILTIN_H */
diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c
index d23ae4d8cb..11340d26e2 100644
--- a/src/mesa/shader/slang/slang_codegen.c
+++ b/src/mesa/shader/slang/slang_codegen.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
  *
  * Copyright (C) 2005-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2008 VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -57,10 +57,39 @@
 #include "slang_print.h"
 
 
+/** Max iterations to unroll */
+const GLuint MAX_FOR_LOOP_UNROLL_ITERATIONS = 32;
+
+/** Max for-loop body size (in slang operations) to unroll */
+const GLuint MAX_FOR_LOOP_UNROLL_BODY_SIZE = 50;
+
+/** Max for-loop body complexity to unroll.
+ * We'll compute complexity as the product of the number of iterations
+ * and the size of the body.  So long-ish loops with very simple bodies
+ * can be unrolled, as well as short loops with larger bodies.
+ */
+const GLuint MAX_FOR_LOOP_UNROLL_COMPLEXITY = 256;
+
+
+
 static slang_ir_node *
 _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper);
 
 
+/**
+ * Retrieves type information about an operation.
+ * Returns GL_TRUE on success.
+ * Returns GL_FALSE otherwise.
+ */
+static GLboolean
+typeof_operation(const struct slang_assemble_ctx_ *A,
+                 slang_operation *op,
+                 slang_typeinfo *ti)
+{
+   return _slang_typeof_operation(op, &A->space, ti, A->atoms, A->log);
+}
+
+
 static GLboolean
 is_sampler_type(const slang_fully_specified_type *t)
 {
@@ -199,7 +228,14 @@ _slang_sizeof_type_specifier(const slang_type_specifier *spec)
       break;
    case SLANG_SPEC_STRUCT:
       sz = _slang_field_offset(spec, 0); /* special use */
-      if (sz > 4) {
+      if (sz == 1) {
+         /* 1-float structs are actually troublesome to deal with since they
+          * might get placed at R.x, R.y, R.z or R.z.  Return size=2 to
+          * ensure the object is placed at R.x
+          */
+         sz = 2;
+      }
+      else if (sz > 4) {
          sz = (sz + 3) & ~0x3; /* round up to multiple of four */
       }
       break;
@@ -220,6 +256,48 @@ _slang_sizeof_type_specifier(const slang_type_specifier *spec)
 
 
 /**
+ * Query variable/array length (number of elements).
+ * This is slightly non-trivial because there are two ways to express
+ * arrays: "float x[3]" vs. "float[3] x".
+ * \return the length of the array for the given variable, or 0 if not an array
+ */
+static GLint
+_slang_array_length(const slang_variable *var)
+{
+   if (var->type.array_len > 0) {
+      /* Ex: float[4] x; */
+      return var->type.array_len;
+   }
+   if (var->array_len > 0) {
+      /* Ex: float x[4]; */
+      return var->array_len;
+   }
+   return 0;
+}
+
+
+/**
+ * Compute total size of array give size of element, number of elements.
+ * \return size in floats
+ */
+static GLint
+_slang_array_size(GLint elemSize, GLint arrayLen)
+{
+   GLint total;
+   assert(elemSize > 0);
+   if (arrayLen > 1) {
+      /* round up base type to multiple of 4 */
+      total = ((elemSize + 3) & ~0x3) * MAX2(arrayLen, 1);
+   }
+   else {
+      total = elemSize;
+   }
+   return total;
+}
+
+
+
+/**
  * Establish the binding between a slang_ir_node and a slang_variable.
  * Then, allocate/attach a slang_ir_storage object to the IR node if needed.
  * The IR node must be a IR_VAR or IR_VAR_DECL node.
@@ -238,9 +316,9 @@ _slang_attach_storage(slang_ir_node *n, slang_variable *var)
 
    if (!n->Store) {
       /* need to setup storage */
-      if (n->Var && n->Var->aux) {
+      if (n->Var && n->Var->store) {
          /* node storage info = var storage info */
-         n->Store = (slang_ir_storage *) n->Var->aux;
+         n->Store = n->Var->store;
       }
       else {
          /* alloc new storage info */
@@ -251,8 +329,8 @@ _slang_attach_storage(slang_ir_node *n, slang_variable *var)
                 (void*) n->Store, n->Store->Size);
 #endif
          if (n->Var)
-            n->Var->aux = n->Store;
-         assert(n->Var->aux);
+            n->Var->store = n->Store;
+         assert(n->Var->store);
       }
    }
 }
@@ -335,7 +413,7 @@ _slang_input_index(const char *name, GLenum target, GLuint *swizzleOut)
    const struct input_info *inputs
       = (target == GL_VERTEX_PROGRAM_ARB) ? vertInputs : fragInputs;
 
-   ASSERT(MAX_TEXTURE_UNITS == 8); /* if this fails, fix vertInputs above */
+   ASSERT(MAX_TEXTURE_COORD_UNITS == 8); /* if this fails, fix vertInputs above */
 
    for (i = 0; inputs[i].Name; i++) {
       if (strcmp(inputs[i].Name, name) == 0) {
@@ -653,23 +731,11 @@ new_if(slang_ir_node *cond, slang_ir_node *ifPart, slang_ir_node *elsePart)
  * New IR_VAR node - a reference to a previously declared variable.
  */
 static slang_ir_node *
-new_var(slang_assemble_ctx *A, slang_operation *oper, slang_atom name)
+new_var(slang_assemble_ctx *A, slang_variable *var)
 {
-   slang_ir_node *n;
-   slang_variable *var = _slang_locate_variable(oper->locals, name, GL_TRUE);
-   if (!var)
-      return NULL;
-
-   assert(var->declared);
-
-   assert(!oper->var || oper->var == var);
-
-   n = new_node0(IR_VAR);
+   slang_ir_node *n = new_node0(IR_VAR);
    if (n) {
       _slang_attach_storage(n, var);
-      /*
-      printf("new_var %s store=%p\n", (char*)name, (void*) n->Store);
-      */
    }
    return n;
 }
@@ -783,7 +849,7 @@ static void
 slang_resolve_variable(slang_operation *oper)
 {
    if (oper->type == SLANG_OPER_IDENTIFIER && !oper->var) {
-      oper->var = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE);
+      oper->var = _slang_variable_locate(oper->locals, oper->a_id, GL_TRUE);
    }
 }
 
@@ -799,7 +865,7 @@ slang_substitute(slang_assemble_ctx *A, slang_operation *oper,
    switch (oper->type) {
    case SLANG_OPER_VARIABLE_DECL:
       {
-         slang_variable *v = _slang_locate_variable(oper->locals,
+         slang_variable *v = _slang_variable_locate(oper->locals,
                                                     oper->a_id, GL_TRUE);
          assert(v);
          if (v->initializer && oper->num_children == 0) {
@@ -821,7 +887,7 @@ slang_substitute(slang_assemble_ctx *A, slang_operation *oper,
          slang_atom id = oper->a_id;
          slang_variable *v;
 	 GLuint i;
-         v = _slang_locate_variable(oper->locals, id, GL_TRUE);
+         v = _slang_variable_locate(oper->locals, id, GL_TRUE);
 	 if (!v) {
             _mesa_problem(NULL, "var %s not found!\n", (char *) oper->a_id);
             return;
@@ -1414,27 +1480,6 @@ slang_find_asm_info(const char *name)
 
 
 /**
- * Return the default swizzle mask for accessing a variable of the
- * given size (in floats).  If size = 1, comp is used to identify
- * which component [0..3] of the register holds the variable.
- */
-static GLuint
-_slang_var_swizzle(GLint size, GLint comp)
-{
-   switch (size) {
-   case 1:
-      return MAKE_SWIZZLE4(comp, comp, comp, comp);
-   case 2:
-      return MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_NIL, SWIZZLE_NIL);
-   case 3:
-      return MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_NIL);
-   default:
-      return SWIZZLE_XYZW;
-   }
-}
-
-
-/**
  * Some write-masked assignments are simple, but others are hard.
  * Simple example:
  *    vec3 v;
@@ -1477,6 +1522,8 @@ _slang_simple_writemask(GLuint writemask, GLuint swizzle)
  * Convert the given swizzle into a writemask.  In some cases this
  * is trivial, in other cases, we'll need to also swizzle the right
  * hand side to put components in the right places.
+ * See comment above for more info.
+ * XXX this function could be simplified and should probably be renamed.
  * \param swizzle  the incoming swizzle
  * \param writemaskOut  returns the writemask
  * \param swizzleOut  swizzle to apply to the right-hand-side
@@ -1545,6 +1592,7 @@ swizzle_to_writemask(slang_assemble_ctx *A, GLuint swizzle,
 }
 
 
+#if 0 /* not used, but don't remove just yet */
 /**
  * Recursively traverse 'oper' to produce a swizzle mask in the event
  * of any vector subscripts and swizzle suffixes.
@@ -1598,8 +1646,10 @@ resolve_swizzle(const slang_operation *oper)
       return SWIZZLE_XYZW;
    }
 }
+#endif
 
 
+#if 0
 /**
  * Recursively descend through swizzle nodes to find the node's storage info.
  */
@@ -1611,7 +1661,7 @@ get_store(const slang_ir_node *n)
    }
    return n->Store;
 }
-
+#endif
 
 
 /**
@@ -1684,6 +1734,7 @@ _slang_gen_asm(slang_assemble_ctx *A, slang_operation *oper,
 }
 
 
+#if 0
 static void
 print_funcs(struct slang_function_scope_ *scope, const char *name)
 {
@@ -1697,6 +1748,7 @@ print_funcs(struct slang_function_scope_ *scope, const char *name)
    if (scope->outer_scope)
       print_funcs(scope->outer_scope, name);
 }
+#endif
 
 
 /**
@@ -1758,18 +1810,12 @@ _slang_find_function_by_max_argc(slang_function_scope *scope,
  * struct type.
  */
 static slang_function *
-_slang_make_constructor(slang_assemble_ctx *A, slang_struct *str)
+_slang_make_struct_constructor(slang_assemble_ctx *A, slang_struct *str)
 {
    const GLint numFields = str->fields->num_variables;
-
-   slang_function *fun = (slang_function *) _mesa_malloc(sizeof(slang_function));
-   if (!fun)
-      return NULL;
-
-   slang_function_construct(fun);
+   slang_function *fun = slang_function_new(SLANG_FUNC_CONSTRUCTOR);
 
    /* function header (name, return type) */
-   fun->kind = SLANG_FUNC_CONSTRUCTOR;
    fun->header.a_name = str->a_name;
    fun->header.type.qualifier = SLANG_QUAL_NONE;
    fun->header.type.specifier.type = SLANG_SPEC_STRUCT;
@@ -1783,7 +1829,7 @@ _slang_make_constructor(slang_assemble_ctx *A, slang_struct *str)
          printf("Field %d: %s\n", i, (char*) str->fields->variables[i]->a_name);
          */
          slang_variable *p = slang_variable_scope_grow(fun->parameters);
-         *p = *str->fields->variables[i]; /* copy the type */
+         *p = *str->fields->variables[i]; /* copy the variable and type */
          p->type.qualifier = SLANG_QUAL_CONST;
       }
       fun->param_count = fun->parameters->num_variables;
@@ -1888,7 +1934,6 @@ _slang_make_constructor(slang_assemble_ctx *A, slang_struct *str)
          ret->children[0].type = SLANG_OPER_IDENTIFIER;
          ret->children[0].a_id = var->a_name;
          ret->children[0].locals = _slang_variable_scope_new(scope);
-
       }
    }
    /*
@@ -1911,7 +1956,7 @@ _slang_locate_struct_constructor(slang_assemble_ctx *A, const char *name)
          /* found a structure type that matches the function name */
          if (!str->constructor) {
             /* create the constructor function now */
-            str->constructor = _slang_make_constructor(A, str);
+            str->constructor = _slang_make_struct_constructor(A, str);
          }
          return str->constructor;
       }
@@ -1920,6 +1965,156 @@ _slang_locate_struct_constructor(slang_assemble_ctx *A, const char *name)
 }
 
 
+/**
+ * Generate a new slang_function to satisfy a call to an array constructor.
+ * Ex:  float[3](1., 2., 3.)
+ */
+static slang_function *
+_slang_make_array_constructor(slang_assemble_ctx *A, slang_operation *oper)
+{
+   slang_type_specifier_type baseType;
+   slang_function *fun;
+   int num_elements;
+
+   fun = slang_function_new(SLANG_FUNC_CONSTRUCTOR);
+   if (!fun)
+      return NULL;
+
+   baseType = slang_type_specifier_type_from_string((char *) oper->a_id);
+
+   num_elements = oper->num_children;
+
+   /* function header, return type */
+   {
+      fun->header.a_name = oper->a_id;
+      fun->header.type.qualifier = SLANG_QUAL_NONE;
+      fun->header.type.specifier.type = SLANG_SPEC_ARRAY;
+      fun->header.type.specifier._array =
+         slang_type_specifier_new(baseType, NULL, NULL);
+      fun->header.type.array_len = num_elements;
+   }
+
+   /* function parameters (= number of elements) */
+   {
+      GLint i;
+      for (i = 0; i < num_elements; i++) {
+         /*
+         printf("Field %d: %s\n", i, (char*) str->fields->variables[i]->a_name);
+         */
+         slang_variable *p = slang_variable_scope_grow(fun->parameters);
+         char name[10];
+         _mesa_snprintf(name, sizeof(name), "p%d", i);
+         p->a_name = slang_atom_pool_atom(A->atoms, name);
+         p->type.qualifier = SLANG_QUAL_CONST;
+         p->type.specifier.type = baseType;
+      }
+      fun->param_count = fun->parameters->num_variables;
+   }
+
+   /* Add __retVal to params */
+   {
+      slang_variable *p = slang_variable_scope_grow(fun->parameters);
+      slang_atom a_retVal = slang_atom_pool_atom(A->atoms, "__retVal");
+      assert(a_retVal);
+      p->a_name = a_retVal;
+      p->type = fun->header.type;
+      p->type.qualifier = SLANG_QUAL_OUT;
+      p->type.specifier.type = baseType;
+      fun->param_count++;
+   }
+
+   /* function body is:
+    *    block:
+    *       declare T;
+    *       T[0] = p0;
+    *       T[1] = p1;
+    *       ...
+    *       T[n] = pn;
+    *       return T;
+    */
+   {
+      slang_variable_scope *scope;
+      slang_variable *var;
+      GLint i;
+
+      fun->body = slang_operation_new(1);
+      fun->body->type = SLANG_OPER_BLOCK_NEW_SCOPE;
+      fun->body->num_children = num_elements + 2;
+      fun->body->children = slang_operation_new(num_elements + 2);
+
+      scope = fun->body->locals;
+      scope->outer_scope = fun->parameters;
+
+      /* create local var 't' */
+      var = slang_variable_scope_grow(scope);
+      var->a_name = slang_atom_pool_atom(A->atoms, "ttt");
+      var->type = fun->header.type;/*XXX copy*/
+
+      /* declare t */
+      {
+         slang_operation *decl;
+
+         decl = &fun->body->children[0];
+         decl->type = SLANG_OPER_VARIABLE_DECL;
+         decl->locals = _slang_variable_scope_new(scope);
+         decl->a_id = var->a_name;
+      }
+
+      /* assign params to elements of t */
+      for (i = 0; i < num_elements; i++) {
+         slang_operation *assign = &fun->body->children[1 + i];
+
+         assign->type = SLANG_OPER_ASSIGN;
+         assign->locals = _slang_variable_scope_new(scope);
+         assign->num_children = 2;
+         assign->children = slang_operation_new(2);
+         
+         {
+            slang_operation *lhs = &assign->children[0];
+
+            lhs->type = SLANG_OPER_SUBSCRIPT;
+            lhs->locals = _slang_variable_scope_new(scope);
+            lhs->num_children = 2;
+            lhs->children = slang_operation_new(2);
+ 
+            lhs->children[0].type = SLANG_OPER_IDENTIFIER;
+            lhs->children[0].a_id = var->a_name;
+            lhs->children[0].locals = _slang_variable_scope_new(scope);
+
+            lhs->children[1].type = SLANG_OPER_LITERAL_INT;
+            lhs->children[1].literal[0] = (GLfloat) i;
+         }
+
+         {
+            slang_operation *rhs = &assign->children[1];
+
+            rhs->type = SLANG_OPER_IDENTIFIER;
+            rhs->locals = _slang_variable_scope_new(scope);
+            rhs->a_id = fun->parameters->variables[i]->a_name;
+         }         
+      }
+
+      /* return t; */
+      {
+         slang_operation *ret = &fun->body->children[num_elements + 1];
+
+         ret->type = SLANG_OPER_RETURN;
+         ret->locals = _slang_variable_scope_new(scope);
+         ret->num_children = 1;
+         ret->children = slang_operation_new(1);
+         ret->children[0].type = SLANG_OPER_IDENTIFIER;
+         ret->children[0].a_id = var->a_name;
+         ret->children[0].locals = _slang_variable_scope_new(scope);
+      }
+   }
+
+   /*
+   slang_print_function(fun, 1);
+   */
+
+   return fun;
+}
+
 
 static GLboolean
 _slang_is_vec_mat_type(const char *name)
@@ -1953,24 +2148,27 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name,
    const GLuint param_count = oper->num_children;
    slang_atom atom;
    slang_function *fun;
-   GLboolean error;
    slang_ir_node *n;
 
    atom = slang_atom_pool_atom(A->atoms, name);
    if (atom == SLANG_ATOM_NULL)
       return NULL;
 
-   /*
-    * First, try to find function by name and exact argument type matching.
-    */
-   fun = _slang_locate_function(A->space.funcs, atom, params, param_count,
-				&A->space, A->atoms, A->log, &error);
-
-   if (error) {
-      slang_info_log_error(A->log,
-                           "Function '%s' not found (check argument types)",
-                           name);
-      return NULL;
+   if (oper->array_constructor) {
+      /* this needs special handling */
+      fun = _slang_make_array_constructor(A, oper);
+   }
+   else {
+      /* Try to find function by name and exact argument type matching */
+      GLboolean error = GL_FALSE;
+      fun = _slang_function_locate(A->space.funcs, atom, params, param_count,
+                                   &A->space, A->atoms, A->log, &error);
+      if (error) {
+         slang_info_log_error(A->log,
+                              "Function '%s' not found (check argument types)",
+                              name);
+         return NULL;
+      }
    }
 
    if (!fun) {
@@ -2018,6 +2216,28 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name,
                            name);
       return NULL;
    }
+   if (!fun->body) {
+      slang_info_log_error(A->log,
+                           "Function '%s' prototyped but not defined.  "
+                           "Separate compilation units not supported.",
+                           name);
+      return NULL;
+   }
+
+   /* type checking to be sure function's return type matches 'dest' type */
+   if (dest) {
+      slang_typeinfo t0;
+
+      slang_typeinfo_construct(&t0);
+      typeof_operation(A, dest, &t0);
+
+      if (!slang_type_specifier_equal(&t0.spec, &fun->header.type.specifier)) {
+         slang_info_log_error(A->log,
+                              "Incompatible type returned by call to '%s'",
+                              name);
+         return NULL;
+      }
+   }
 
    n = _slang_gen_function_call(A, fun, oper, dest);
 
@@ -2029,6 +2249,51 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name,
       /*printf("Alloc storage for function result, size %d \n", size);*/
    }
 
+   if (oper->array_constructor) {
+      /* free the temporary array constructor function now */
+      slang_function_destruct(fun);
+   }
+
+   return n;
+}
+
+
+static slang_ir_node *
+_slang_gen_method_call(slang_assemble_ctx *A, slang_operation *oper)
+{
+   slang_atom *a_length = slang_atom_pool_atom(A->atoms, "length");
+   slang_ir_node *n;
+   slang_variable *var;
+
+   /* NOTE: In GLSL 1.20, there's only one kind of method
+    * call: array.length().  Anything else is an error.
+    */
+   if (oper->a_id != a_length) {
+      slang_info_log_error(A->log,
+                           "Undefined method call '%s'", (char *) oper->a_id);
+      return NULL;
+   }
+
+   /* length() takes no arguments */
+   if (oper->num_children > 0) {
+      slang_info_log_error(A->log, "Invalid arguments to length() method");
+      return NULL;
+   }
+
+   /* lookup the object/variable */
+   var = _slang_variable_locate(oper->locals, oper->a_obj, GL_TRUE);
+   if (!var || var->type.specifier.type != SLANG_SPEC_ARRAY) {
+      slang_info_log_error(A->log,
+                           "Undefined object '%s'", (char *) oper->a_obj);
+      return NULL;
+   }
+
+   /* Create a float/literal IR node encoding the array length */
+   n = new_node0(IR_FLOAT);
+   if (n) {
+      n->Value[0] = (float) _slang_array_length(var);
+      n->Store = _slang_new_ir_storage(PROGRAM_CONSTANT, -1, 1);
+   }
    return n;
 }
 
@@ -2063,7 +2328,7 @@ _slang_is_scalar_or_boolean(slang_assemble_ctx *A, slang_operation *oper)
    GLint size;
 
    slang_typeinfo_construct(&type);
-   _slang_typeof_operation(A, oper, &type);
+   typeof_operation(A, oper, &type);
    size = _slang_sizeof_type_specifier(&type.spec);
    slang_typeinfo_destruct(&type);
    return size == 1;
@@ -2080,7 +2345,7 @@ _slang_is_boolean(slang_assemble_ctx *A, slang_operation *oper)
    GLboolean isBool;
 
    slang_typeinfo_construct(&type);
-   _slang_typeof_operation(A, oper, &type);
+   typeof_operation(A, oper, &type);
    isBool = (type.spec.type == SLANG_SPEC_BOOL);
    slang_typeinfo_destruct(&type);
    return isBool;
@@ -2201,40 +2466,259 @@ _slang_gen_do(slang_assemble_ctx * A, const slang_operation *oper)
 
 
 /**
- * Generate for-loop using high-level IR_LOOP instruction.
+ * Recursively count the number of operations rooted at 'oper'.
+ * This gives some kind of indication of the size/complexity of an operation.
+ */
+static GLuint
+sizeof_operation(const slang_operation *oper)
+{
+   if (oper) {
+      GLuint count = 1; /* me */
+      GLuint i;
+      for (i = 0; i < oper->num_children; i++) {
+         count += sizeof_operation(&oper->children[i]);
+      }
+      return count;
+   }
+   else {
+      return 0;
+   }
+}
+
+
+/**
+ * Determine if a for-loop can be unrolled.
+ * At this time, only a rather narrow class of for loops can be unrolled.
+ * See code for details.
+ * When a loop can't be unrolled because it's too large we'll emit a
+ * message to the log.
+ */
+static GLboolean
+_slang_can_unroll_for_loop(slang_assemble_ctx * A, const slang_operation *oper)
+{
+   GLuint bodySize;
+   GLint start, end;
+   const char *varName;
+   slang_atom varId;
+
+   assert(oper->type == SLANG_OPER_FOR);
+   assert(oper->num_children == 4);
+
+   /* children[0] must be either "int i=constant" or "i=constant" */
+   if (oper->children[0].type == SLANG_OPER_BLOCK_NO_NEW_SCOPE) {
+      slang_variable *var;
+
+      if (oper->children[0].children[0].type != SLANG_OPER_VARIABLE_DECL)
+         return GL_FALSE;
+
+      varId = oper->children[0].children[0].a_id;
+
+      var = _slang_variable_locate(oper->children[0].children[0].locals,
+                                   varId, GL_TRUE);
+      if (!var)
+         return GL_FALSE;
+      if (!var->initializer)
+         return GL_FALSE;
+      if (var->initializer->type != SLANG_OPER_LITERAL_INT)
+         return GL_FALSE;
+      start = (GLint) var->initializer->literal[0];
+   }
+   else if (oper->children[0].type == SLANG_OPER_EXPRESSION) {
+      if (oper->children[0].children[0].type != SLANG_OPER_ASSIGN)
+         return GL_FALSE;
+      if (oper->children[0].children[0].children[0].type != SLANG_OPER_IDENTIFIER)
+         return GL_FALSE;
+      if (oper->children[0].children[0].children[1].type != SLANG_OPER_LITERAL_INT)
+         return GL_FALSE;
+
+      varId = oper->children[0].children[0].children[0].a_id;
+
+      start = (GLint) oper->children[0].children[0].children[1].literal[0];
+   }
+   else {
+      return GL_FALSE;
+   }
+
+   /* children[1] must be "i<constant" */
+   if (oper->children[1].type != SLANG_OPER_EXPRESSION)
+      return GL_FALSE;
+   if (oper->children[1].children[0].type != SLANG_OPER_LESS)
+      return GL_FALSE;
+   if (oper->children[1].children[0].children[0].type != SLANG_OPER_IDENTIFIER)
+      return GL_FALSE;
+   if (oper->children[1].children[0].children[1].type != SLANG_OPER_LITERAL_INT)
+      return GL_FALSE;
+
+   end = (GLint) oper->children[1].children[0].children[1].literal[0];
+
+   /* children[2] must be "i++" or "++i" */
+   if (oper->children[2].type != SLANG_OPER_POSTINCREMENT &&
+       oper->children[2].type != SLANG_OPER_PREINCREMENT)
+      return GL_FALSE;
+   if (oper->children[2].children[0].type != SLANG_OPER_IDENTIFIER)
+      return GL_FALSE;
+
+   /* make sure the same variable name is used in all places */
+   if ((oper->children[1].children[0].children[0].a_id != varId) ||
+       (oper->children[2].children[0].a_id != varId))
+      return GL_FALSE;
+
+   varName = (const char *) varId;
+
+   /* children[3], the loop body, can't be too large */
+   bodySize = sizeof_operation(&oper->children[3]);
+   if (bodySize > MAX_FOR_LOOP_UNROLL_BODY_SIZE) {
+      slang_info_log_print(A->log,
+                           "Note: 'for (%s ... )' body is too large/complex"
+                           " to unroll",
+                           varName);
+      return GL_FALSE;
+   }
+
+   if (start >= end)
+      return GL_FALSE; /* degenerate case */
+
+   if (end - start > MAX_FOR_LOOP_UNROLL_ITERATIONS) {
+      slang_info_log_print(A->log,
+                           "Note: 'for (%s=%d; %s<%d; ++%s)' is too"
+                           " many iterations to unroll",
+                           varName, start, varName, end, varName);
+      return GL_FALSE;
+   }
+
+   if ((end - start) * bodySize > MAX_FOR_LOOP_UNROLL_COMPLEXITY) {
+      slang_info_log_print(A->log,
+                           "Note: 'for (%s=%d; %s<%d; ++%s)' will generate"
+                           " too much code to unroll",
+                           varName, start, varName, end, varName);
+      return GL_FALSE;
+   }
+
+   return GL_TRUE; /* we can unroll the loop */
+}
+
+
+/**
+ * Unroll a for-loop.
+ * First we determine the number of iterations to unroll.
+ * Then for each iteration:
+ *   make a copy of the loop body
+ *   replace instances of the loop variable with the current iteration value
+ *   generate IR code for the body
+ * \return pointer to generated IR code or NULL if error, out of memory, etc.
+ */
+static slang_ir_node *
+_slang_unroll_for_loop(slang_assemble_ctx * A, const slang_operation *oper)
+{
+   GLint start, end, iter;
+   slang_ir_node *n, *root = NULL;
+   slang_atom varId;
+
+   if (oper->children[0].type == SLANG_OPER_BLOCK_NO_NEW_SCOPE) {
+      /* for (int i=0; ... */
+      slang_variable *var;
+
+      varId = oper->children[0].children[0].a_id;
+      var = _slang_variable_locate(oper->children[0].children[0].locals,
+                                   varId, GL_TRUE);
+      start = (GLint) var->initializer->literal[0];
+   }
+   else {
+      /* for (i=0; ... */
+      varId = oper->children[0].children[0].children[0].a_id;
+      start = (GLint) oper->children[0].children[0].children[1].literal[0];
+   }
+
+   end = (GLint) oper->children[1].children[0].children[1].literal[0];
+
+   for (iter = start; iter < end; iter++) {
+      slang_operation *body;
+
+      /* make a copy of the loop body */
+      body = slang_operation_new(1);
+      if (!body)
+         return NULL;
+
+      if (!slang_operation_copy(body, &oper->children[3]))
+         return NULL;
+
+      /* in body, replace instances of 'varId' with literal 'iter' */
+      {
+         slang_variable *oldVar;
+         slang_operation *newOper;
+
+         oldVar = _slang_variable_locate(oper->locals, varId, GL_TRUE);
+         if (!oldVar) {
+            /* undeclared loop variable */
+            slang_operation_delete(body);
+            return NULL;
+         }
+
+         newOper = slang_operation_new(1);
+         newOper->type = SLANG_OPER_LITERAL_INT;
+         newOper->literal_size = 1;
+         newOper->literal[0] = iter;
+
+         /* replace instances of the loop variable with newOper */
+         slang_substitute(A, body, 1, &oldVar, &newOper, GL_FALSE);
+      }
+
+      /* do IR codegen for body */
+      n = _slang_gen_operation(A, body);
+      root = new_seq(root, n);
+
+      slang_operation_delete(body);
+   }
+
+   return root;
+}
+
+
+/**
+ * Generate IR for a for-loop.  Unrolling will be done when possible.
  */
 static slang_ir_node *
 _slang_gen_for(slang_assemble_ctx * A, const slang_operation *oper)
 {
-   /*
-    * init code (child[0])
-    * LOOP:
-    *    BREAK if !expr (child[1])
-    *    body code (child[3])
-    *    tail code:
-    *       incr code (child[2])   // XXX continue here
-    */
-   slang_ir_node *prevLoop, *loop, *cond, *breakIf, *body, *init, *incr;
+   GLboolean unroll = _slang_can_unroll_for_loop(A, oper);
 
-   init = _slang_gen_operation(A, &oper->children[0]);
-   loop = new_loop(NULL);
+   if (unroll) {
+      slang_ir_node *code = _slang_unroll_for_loop(A, oper);
+      if (code)
+         return code;
+   }
 
-   /* save old, push new loop */
-   prevLoop = A->CurLoop;
-   A->CurLoop = loop;
+   /* conventional for-loop code generation */
+   {
+      /*
+       * init code (child[0])
+       * LOOP:
+       *    BREAK if !expr (child[1])
+       *    body code (child[3])
+       *    tail code:
+       *       incr code (child[2])   // XXX continue here
+       */
+      slang_ir_node *prevLoop, *loop, *cond, *breakIf, *body, *init, *incr;
+      init = _slang_gen_operation(A, &oper->children[0]);
+      loop = new_loop(NULL);
 
-   cond = new_cond(new_not(_slang_gen_operation(A, &oper->children[1])));
-   breakIf = new_break_if_true(A->CurLoop, cond);
-   body = _slang_gen_operation(A, &oper->children[3]);
-   incr = _slang_gen_operation(A, &oper->children[2]);
+      /* save old, push new loop */
+      prevLoop = A->CurLoop;
+      A->CurLoop = loop;
 
-   loop->Children[0] = new_seq(breakIf, body);
-   loop->Children[1] = incr;  /* tail code */
+      cond = new_cond(new_not(_slang_gen_operation(A, &oper->children[1])));
+      breakIf = new_break_if_true(A->CurLoop, cond);
+      body = _slang_gen_operation(A, &oper->children[3]);
+      incr = _slang_gen_operation(A, &oper->children[2]);
 
-   /* pop loop, restore prev */
-   A->CurLoop = prevLoop;
+      loop->Children[0] = new_seq(breakIf, body);
+      loop->Children[1] = incr;  /* tail code */
+
+      /* pop loop, restore prev */
+      A->CurLoop = prevLoop;
 
-   return new_seq(init, loop);
+      return new_seq(init, loop);
+   }
 }
 
 
@@ -2415,80 +2899,237 @@ _slang_gen_temporary(GLint size)
 
 
 /**
- * Generate IR node for allocating/declaring a variable.
+ * Generate program constants for an array.
+ * Ex: const vec2[3] v = vec2[3](vec2(1,1), vec2(2,2), vec2(3,3));
+ * This will allocate and initialize three vector constants, storing
+ * the array in constant memory, not temporaries like a non-const array.
+ * This can also be used for uniform array initializers.
+ * \return GL_TRUE for success, GL_FALSE if failure (semantic error, etc).
+ */
+static GLboolean
+make_constant_array(slang_assemble_ctx *A,
+                    slang_variable *var,
+                    slang_operation *initializer)
+{
+   struct gl_program *prog = A->program;
+   const GLenum datatype = _slang_gltype_from_specifier(&var->type.specifier);
+   const char *varName = (char *) var->a_name;
+   const GLuint numElements = initializer->num_children;
+   GLint size;
+   GLuint i, j;
+   GLfloat *values;
+
+   if (!var->store) {
+      var->store = _slang_new_ir_storage(PROGRAM_UNDEFINED, -6, -6);
+   }
+   size = var->store->Size;
+
+   assert(var->type.qualifier == SLANG_QUAL_CONST ||
+          var->type.qualifier == SLANG_QUAL_UNIFORM);
+   assert(initializer->type == SLANG_OPER_CALL);
+   assert(initializer->array_constructor);
+
+   values = (GLfloat *) _mesa_malloc(numElements * 4 * sizeof(GLfloat));
+
+   /* convert constructor params into ordinary floats */
+   for (i = 0; i < numElements; i++) {
+      const slang_operation *op = &initializer->children[i];
+      if (op->type != SLANG_OPER_LITERAL_FLOAT) {
+         /* unsupported type for this optimization */
+         free(values);
+         return GL_FALSE;
+      }
+      for (j = 0; j < op->literal_size; j++) {
+         values[i * 4 + j] = op->literal[j];
+      }
+      for ( ; j < 4; j++) {
+         values[i * 4 + j] = 0.0f;
+      }
+   }
+
+   /* slightly different paths for constants vs. uniforms */
+   if (var->type.qualifier == SLANG_QUAL_UNIFORM) {
+      var->store->File = PROGRAM_UNIFORM;
+      var->store->Index = _mesa_add_uniform(prog->Parameters, varName,
+                                            size, datatype, values);
+   }
+   else {
+      var->store->File = PROGRAM_CONSTANT;
+      var->store->Index = _mesa_add_named_constant(prog->Parameters, varName,
+                                                   values, size);
+   }
+   assert(var->store->Size == size);
+
+   _mesa_free(values);
+
+   return GL_TRUE;
+}
+
+
+
+/**
+ * Generate IR node for allocating/declaring a variable (either a local or
+ * a global).
+ * Generally, this involves allocating an slang_ir_storage instance for the
+ * variable, choosing a register file (temporary, constant, etc).
+ * For ordinary variables we do not yet allocate storage though.  We do that
+ * when we find the first actual use of the variable to avoid allocating temp
+ * regs that will never get used.
+ * At this time, uniforms are always allocated space in this function.
+ *
+ * \param initializer  Optional initializer expression for the variable.
  */
 static slang_ir_node *
-_slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var)
+_slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var,
+                    slang_operation *initializer)
 {
-   slang_ir_node *n;
+   const char *varName = (const char *) var->a_name;
+   const GLenum datatype = _slang_gltype_from_specifier(&var->type.specifier);
+   slang_ir_node *varDecl, *n;
+   slang_ir_storage *store;
+   GLint arrayLen, size, totalSize;  /* if array then totalSize > size */
+   enum register_file file;
 
    /*assert(!var->declared);*/
    var->declared = GL_TRUE;
 
-   n = new_node0(IR_VAR_DECL);
-   if (n) {
-      _slang_attach_storage(n, var);
-      assert(var->aux);
-      assert(n->Store == var->aux);
-      assert(n->Store);
-      assert(n->Store->Index < 0);
+   /* determine GPU register file for simple cases */
+   if (is_sampler_type(&var->type)) {
+      file = PROGRAM_SAMPLER;
+   }
+   else if (var->type.qualifier == SLANG_QUAL_UNIFORM) {
+      file = PROGRAM_UNIFORM;
+   }
+   else {
+      file = PROGRAM_TEMPORARY;
+   }
 
-      if (is_sampler_type(&var->type)) {
-         n->Store->File = PROGRAM_SAMPLER;
-      }
-      else {
-         n->Store->File = PROGRAM_TEMPORARY;
-      }
+   totalSize = size = _slang_sizeof_type_specifier(&var->type.specifier);
+   if (size <= 0) {
+      slang_info_log_error(A->log, "invalid declaration for '%s'", varName);
+      return NULL;
+   }
 
-      n->Store->Size = _slang_sizeof_type_specifier(&n->Var->type.specifier);
+   arrayLen = _slang_array_length(var);
+   totalSize = _slang_array_size(size, arrayLen);
 
-      if (n->Store->Size <= 0) {
-         slang_info_log_error(A->log, "invalid declaration for '%s'",
-                              (char*) var->a_name);
-         return NULL;
-      }
+   /* Allocate IR node for the declaration */
+   varDecl = new_node0(IR_VAR_DECL);
+   if (!varDecl)
+      return NULL;
+
+   _slang_attach_storage(varDecl, var); /* undefined storage at first */
+   assert(var->store);
+   assert(varDecl->Store == var->store);
+   assert(varDecl->Store);
+   assert(varDecl->Store->Index < 0);
+   store = var->store;
+
+   assert(store == varDecl->Store);
+
+
+   /* Fill in storage fields which we now know.  store->Index/Swizzle may be
+    * set for some cases below.  Otherwise, store->Index/Swizzle will be set
+    * during code emit.
+    */
+   store->File = file;
+   store->Size = totalSize;
+
+   /* if there's an initializer, generate IR for the expression */
+   if (initializer) {
+      slang_ir_node *varRef, *init;
+
+      if (var->type.qualifier == SLANG_QUAL_CONST) {
+         /* if the variable is const, the initializer must be a const
+          * expression as well.
+          */
 #if 0
-      printf("%s var %p %s  store=%p index=%d size=%d\n",
-             __FUNCTION__, (void *) var, (char *) var->a_name,
-             (void *) n->Store, n->Store->Index, n->Store->Size);
+         if (!_slang_is_constant_expr(initializer)) {
+            slang_info_log_error(A->log,
+                                 "initializer for %s not constant", varName);
+            return NULL;
+         }
 #endif
+      }
 
-      if (var->array_len > 0) {
-         /* this is an array */
-         /* cannot be const-qualified */
-         if (var->type.qualifier == SLANG_QUAL_CONST) {
-            slang_info_log_error(A->log, "array '%s' cannot be const",
-                                 (char*) var->a_name);
-            return NULL;
+      /* IR for the variable we're initializing */
+      varRef = new_var(A, var);
+      if (!varRef) {
+         slang_info_log_error(A->log, "out of memory");
+         return NULL;
+      }
+
+      /* constant-folding, etc here */
+      _slang_simplify(initializer, &A->space, A->atoms); 
+
+      /* look for simple constant-valued variables and uniforms */
+      if (var->type.qualifier == SLANG_QUAL_CONST ||
+          var->type.qualifier == SLANG_QUAL_UNIFORM) {
+
+         if (initializer->type == SLANG_OPER_CALL &&
+             initializer->array_constructor) {
+            /* array initializer */
+            if (make_constant_array(A, var, initializer))
+               return varRef;
          }
-         else {
-            /* round up element size to mult of 4 */
-            GLint sz = (n->Store->Size + 3) & ~3;
-            /* mult by array size */
-            sz *= var->array_len;
-            n->Store->Size = sz;
+         else if (initializer->type == SLANG_OPER_LITERAL_FLOAT ||
+                  initializer->type == SLANG_OPER_LITERAL_INT) {
+            /* simple float/vector initializer */
+            if (store->File == PROGRAM_UNIFORM) {
+               store->Index = _mesa_add_uniform(A->program->Parameters,
+                                                varName,
+                                                totalSize, datatype,
+                                                initializer->literal);
+               store->Swizzle = _slang_var_swizzle(size, 0);
+               return varRef;
+            }
+#if 0
+            else {
+               store->File = PROGRAM_CONSTANT;
+               store->Index = _mesa_add_named_constant(A->program->Parameters,
+                                                       varName,
+                                                       initializer->literal,
+                                                       totalSize);
+               store->Swizzle = _slang_var_swizzle(size, 0);
+               return varRef;
+            }
+#endif
          }
       }
 
-      assert(n->Store->Size > 0);
+      /* IR for initializer */
+      init = _slang_gen_operation(A, initializer);
+      if (!init)
+         return NULL;
 
-      /* setup default swizzle for storing the variable */
-      switch (n->Store->Size) {
-      case 2:
-         n->Store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y,
-                                           SWIZZLE_NIL, SWIZZLE_NIL);
-         break;
-      case 3:
-         n->Store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y,
-                                           SWIZZLE_Z, SWIZZLE_NIL);
-         break;
-      default:
-         /* Note that float-sized vars may be allocated in any x/y/z/w
-          * slot, but that won't be determined until code emit time.
-          */
-         n->Store->Swizzle = SWIZZLE_NOOP;
+      /* XXX remove this when type checking is added above */
+      if (init->Store && init->Store->Size != totalSize) {
+         slang_info_log_error(A->log, "invalid assignment (wrong types)");
+         return NULL;
       }
+
+      /* assign RHS to LHS */
+      n = new_node2(IR_COPY, varRef, init);
+      n = new_seq(varDecl, n);
    }
+   else {
+      /* no initializer */
+      n = varDecl;
+   }
+
+   if (store->File == PROGRAM_UNIFORM && store->Index < 0) {
+      /* always need to allocate storage for uniforms at this point */
+      store->Index = _mesa_add_uniform(A->program->Parameters, varName,
+                                       totalSize, datatype, NULL);
+      store->Swizzle = _slang_var_swizzle(size, 0);
+   }
+
+#if 0
+   printf("%s var %p %s  store=%p index=%d size=%d\n",
+          __FUNCTION__, (void *) var, (char *) varName,
+          (void *) store, store->Index, store->Size);
+#endif
+
    return n;
 }
 
@@ -2512,7 +3153,7 @@ _slang_gen_select(slang_assemble_ctx *A, slang_operation *oper)
 
    /* type of children[0] must be boolean */
    slang_typeinfo_construct(&type0);
-   _slang_typeof_operation(A, &oper->children[0], &type0);
+   typeof_operation(A, &oper->children[0], &type0);
    isBool = (type0.spec.type == SLANG_SPEC_BOOL);
    slang_typeinfo_destruct(&type0);
    if (!isBool) {
@@ -2522,8 +3163,8 @@ _slang_gen_select(slang_assemble_ctx *A, slang_operation *oper)
 
    slang_typeinfo_construct(&type1);
    slang_typeinfo_construct(&type2);
-   _slang_typeof_operation(A, &oper->children[1], &type1);
-   _slang_typeof_operation(A, &oper->children[2], &type2);
+   typeof_operation(A, &oper->children[1], &type1);
+   typeof_operation(A, &oper->children[2], &type2);
    isEqual = slang_type_specifier_equal(&type1.spec, &type2.spec);
    slang_typeinfo_destruct(&type1);
    slang_typeinfo_destruct(&type2);
@@ -2663,8 +3304,8 @@ _slang_gen_return(slang_assemble_ctx * A, slang_operation *oper)
 
 #if 1 /* DEBUG */
       {
-         slang_variable *v
-            = _slang_locate_variable(oper->locals, a_retVal, GL_TRUE);
+         slang_variable *v =
+            _slang_variable_locate(oper->locals, a_retVal, GL_TRUE);
          if (!v) {
             /* trying to return a value in a void-valued function */
             return NULL;
@@ -2694,6 +3335,7 @@ _slang_gen_return(slang_assemble_ctx * A, slang_operation *oper)
 }
 
 
+#if 0
 /**
  * Determine if the given operation/expression is const-valued.
  */
@@ -2705,7 +3347,7 @@ _slang_is_constant_expr(const slang_operation *oper)
 
    switch (oper->type) {
    case SLANG_OPER_IDENTIFIER:
-      var = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE);
+      var = _slang_variable_locate(oper->locals, oper->a_id, GL_TRUE);
       if (var && var->type.qualifier == SLANG_QUAL_CONST)
          return GL_TRUE;
       return GL_FALSE;
@@ -2717,6 +3359,7 @@ _slang_is_constant_expr(const slang_operation *oper)
       return GL_TRUE;
    }
 }
+#endif
 
 
 /**
@@ -2737,10 +3380,10 @@ _slang_assignment_compatible(slang_assemble_ctx *A,
    }
 
    slang_typeinfo_construct(&t0);
-   _slang_typeof_operation(A, op0, &t0);
+   typeof_operation(A, op0, &t0);
 
    slang_typeinfo_construct(&t1);
-   _slang_typeof_operation(A, op1, &t1);
+   typeof_operation(A, op1, &t1);
 
    sz0 = _slang_sizeof_type_specifier(&t0.spec);
    sz1 = _slang_sizeof_type_specifier(&t1.spec);
@@ -2779,29 +3422,29 @@ _slang_assignment_compatible(slang_assemble_ctx *A,
 }
 
 
-
 /**
- * Generate IR tree for a variable declaration.
+ * Generate IR tree for a local variable declaration.
+ * Basically do some error checking and call _slang_gen_var_decl().
  */
 static slang_ir_node *
 _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper)
 {
-   slang_ir_node *n;
-   slang_ir_node *varDecl;
-   slang_variable *v;
    const char *varName = (char *) oper->a_id;
+   slang_variable *var;
+   slang_ir_node *varDecl;
    slang_operation *initializer;
 
    assert(oper->type == SLANG_OPER_VARIABLE_DECL);
    assert(oper->num_children <= 1);
 
-   v = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE);
-   if (!v)
+   /* lookup the variable by name */
+   var = _slang_variable_locate(oper->locals, oper->a_id, GL_TRUE);
+   if (!var)
       return NULL;  /* "shouldn't happen" */
 
-   if (v->type.qualifier == SLANG_QUAL_ATTRIBUTE ||
-       v->type.qualifier == SLANG_QUAL_VARYING ||
-       v->type.qualifier == SLANG_QUAL_UNIFORM) {
+   if (var->type.qualifier == SLANG_QUAL_ATTRIBUTE ||
+       var->type.qualifier == SLANG_QUAL_VARYING ||
+       var->type.qualifier == SLANG_QUAL_UNIFORM) {
       /* can't declare attribute/uniform vars inside functions */
       slang_info_log_error(A->log,
                 "local variable '%s' cannot be an attribute/uniform/varying",
@@ -2816,85 +3459,46 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper)
    }
 #endif
 
-   varDecl = _slang_gen_var_decl(A, v);
-   if (!varDecl)
-      return NULL;
-
    /* check if the var has an initializer */
    if (oper->num_children > 0) {
       assert(oper->num_children == 1);
       initializer = &oper->children[0];
    }
-   else if (v->initializer) {
-      initializer = v->initializer;
+   else if (var->initializer) {
+      initializer = var->initializer;
    }
    else {
       initializer = NULL;
    }
 
-   if (v->type.qualifier == SLANG_QUAL_CONST && !initializer) {
-      slang_info_log_error(A->log,
-                           "const-qualified variable '%s' requires initializer",
-                           varName);
-      return NULL;
-   }
-
-
    if (initializer) {
-      slang_ir_node *var, *init;
-
-      /* type check/compare var and initializer */
+      /* check/compare var type and initializer type */
       if (!_slang_assignment_compatible(A, oper, initializer)) {
          slang_info_log_error(A->log, "incompatible types in assignment");
          return NULL;
       }         
-
-      var = new_var(A, oper, oper->a_id);
-      if (!var) {
-         slang_info_log_error(A->log, "undefined variable '%s'", varName);
-         return NULL;
-      }
-
-      if (v->type.qualifier == SLANG_QUAL_CONST) {
-         /* if the variable is const, the initializer must be a const
-          * expression as well.
-          */
-#if 0
-         if (!_slang_is_constant_expr(initializer)) {
-            slang_info_log_error(A->log,
-                                 "initializer for %s not constant", varName);
-            return NULL;
-         }
-#endif
-      }
-
-      _slang_simplify(initializer, &A->space, A->atoms); 
-
-      init = _slang_gen_operation(A, initializer);
-      if (!init)
-         return NULL;
-
-      /*assert(init->Store);*/
-
-      /* XXX remove this when type checking is added above */
-      if (init->Store && var->Store->Size != init->Store->Size) {
-         slang_info_log_error(A->log, "invalid assignment (wrong types)");
-         return NULL;
-      }
-
-      n = new_node2(IR_COPY, var, init);
-      n = new_seq(varDecl, n);
    }
    else {
-      n = varDecl;
+      if (var->type.qualifier == SLANG_QUAL_CONST) {
+         slang_info_log_error(A->log,
+                       "const-qualified variable '%s' requires initializer",
+                       varName);
+         return NULL;
+      }
    }
 
-   return n;
+   /* Generate IR node */
+   varDecl = _slang_gen_var_decl(A, var, initializer);
+   if (!varDecl)
+      return NULL;
+
+   return varDecl;
 }
 
 
 /**
- * Generate IR tree for a variable (such as in an expression).
+ * Generate IR tree for a reference to a variable (such as in an expression).
+ * This is different from a variable declaration.
  */
 static slang_ir_node *
 _slang_gen_variable(slang_assemble_ctx * A, slang_operation *oper)
@@ -2902,12 +3506,15 @@ _slang_gen_variable(slang_assemble_ctx * A, slang_operation *oper)
    /* If there's a variable associated with this oper (from inlining)
     * use it.  Otherwise, use the oper's var id.
     */
-   slang_atom aVar = oper->var ? oper->var->a_name : oper->a_id;
-   slang_ir_node *n = new_var(A, oper, aVar);
-   if (!n) {
-      slang_info_log_error(A->log, "undefined variable '%s'", (char *) aVar);
+   slang_atom name = oper->var ? oper->var->a_name : oper->a_id;
+   slang_variable *var = _slang_variable_locate(oper->locals, name, GL_TRUE);
+   slang_ir_node *n;
+   if (!var) {
+      slang_info_log_error(A->log, "undefined variable '%s'", (char *) name);
       return NULL;
    }
+   assert(var->declared);
+   n = new_var(A, var);
    return n;
 }
 
@@ -2964,6 +3571,22 @@ is_store_writable(const slang_assemble_ctx *A, const slang_ir_storage *store)
 
 
 /**
+ * Walk up an IR storage path to compute the final swizzle.
+ * This is used when we find an expression such as "foo.xz.yx".
+ */
+static GLuint
+root_swizzle(const slang_ir_storage *st)
+{
+   GLuint swizzle = st->Swizzle;
+   while (st->Parent) {
+      st = st->Parent;
+      swizzle = _slang_swizzle_swizzle(st->Swizzle, swizzle);
+   }
+   return swizzle;
+}
+
+
+/**
  * Generate IR tree for an assignment (=).
  */
 static slang_ir_node *
@@ -2972,7 +3595,7 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper)
    if (oper->children[0].type == SLANG_OPER_IDENTIFIER) {
       /* Check that var is writeable */
       slang_variable *var
-         = _slang_locate_variable(oper->children[0].locals,
+         = _slang_variable_locate(oper->children[0].locals,
                                   oper->children[0].a_id, GL_TRUE);
       if (!var) {
          slang_info_log_error(A->log, "undefined variable '%s'",
@@ -3038,9 +3661,9 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper)
       rhs = _slang_gen_operation(A, &oper->children[1]);
       if (lhs && rhs) {
          /* convert lhs swizzle into writemask */
+         const GLuint swizzle = root_swizzle(lhs->Store);
          GLuint writemask, newSwizzle;
-         if (!swizzle_to_writemask(A, lhs->Store->Swizzle,
-                                   &writemask, &newSwizzle)) {
+         if (!swizzle_to_writemask(A, swizzle, &writemask, &newSwizzle)) {
             /* Non-simple writemask, need to swizzle right hand side in
              * order to put components into the right place.
              */
@@ -3066,7 +3689,7 @@ _slang_gen_struct_field(slang_assemble_ctx * A, slang_operation *oper)
 
    /* type of struct */
    slang_typeinfo_construct(&ti);
-   _slang_typeof_operation(A, &oper->children[0], &ti);
+   typeof_operation(A, &oper->children[0], &ti);
 
    if (_slang_type_is_vector(ti.spec.type)) {
       /* the field should be a swizzle */
@@ -3114,11 +3737,11 @@ _slang_gen_struct_field(slang_assemble_ctx * A, slang_operation *oper)
       /* oper->a_id is the field name */
       slang_ir_node *base, *n;
       slang_typeinfo field_ti;
-      GLint fieldSize, fieldOffset = -1, swz;
+      GLint fieldSize, fieldOffset = -1;
 
       /* type of field */
       slang_typeinfo_construct(&field_ti);
-      _slang_typeof_operation(A, oper, &field_ti);
+      typeof_operation(A, oper, &field_ti);
 
       fieldSize = _slang_sizeof_type_specifier(&field_ti.spec);
       if (fieldSize > 0)
@@ -3147,22 +3770,12 @@ _slang_gen_struct_field(slang_assemble_ctx * A, slang_operation *oper)
       if (!n)
          return NULL;
 
-
-      /* setup the storage info for this node */
-      swz = fieldOffset % 4;
-
       n->Field = (char *) oper->a_id;
-      n->Store = _slang_new_ir_storage_relative(fieldOffset / 4,
-                                                fieldSize,
-                                                base->Store);
-      if (fieldSize == 1)
-         n->Store->Swizzle = MAKE_SWIZZLE4(swz, swz, swz, swz);
-      else if (fieldSize == 2)
-         n->Store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y,
-                                           SWIZZLE_NIL, SWIZZLE_NIL);
-      else if (fieldSize == 3)
-         n->Store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y,
-                                           SWIZZLE_Z, SWIZZLE_NIL);
+
+      /* Store the field's offset in storage->Index */
+      n->Store = _slang_new_ir_storage(base->Store->File,
+                                       fieldOffset,
+                                       fieldSize);
 
       return n;
    }
@@ -3179,7 +3792,7 @@ _slang_gen_array_element(slang_assemble_ctx * A, slang_operation *oper)
 
    /* get array's type info */
    slang_typeinfo_construct(&array_ti);
-   _slang_typeof_operation(A, &oper->children[0], &array_ti);
+   typeof_operation(A, &oper->children[0], &array_ti);
 
    if (_slang_type_is_vector(array_ti.spec.type)) {
       /* indexing a simple vector type: "vec4 v; v[0]=p;" */
@@ -3191,8 +3804,16 @@ _slang_gen_array_element(slang_assemble_ctx * A, slang_operation *oper)
       index = (GLint) oper->children[1].literal[0];
       if (oper->children[1].type != SLANG_OPER_LITERAL_INT ||
           index >= (GLint) max) {
+#if 0
          slang_info_log_error(A->log, "Invalid array index for vector type");
+         printf("type = %d\n", oper->children[1].type);
+         printf("index = %d, max = %d\n", index, max);
+         printf("array = %s\n", (char*)oper->children[0].a_id);
+         printf("index = %s\n", (char*)oper->children[1].a_id);
          return NULL;
+#else
+         index = 0;
+#endif
       }
 
       n = _slang_gen_operation(A, &oper->children[0]);
@@ -3215,7 +3836,7 @@ _slang_gen_array_element(slang_assemble_ctx * A, slang_operation *oper)
 
       /* size of array element */
       slang_typeinfo_construct(&elem_ti);
-      _slang_typeof_operation(A, oper, &elem_ti);
+      typeof_operation(A, oper, &elem_ti);
       elemSize = _slang_sizeof_type_specifier(&elem_ti.spec);
 
       if (_slang_type_is_matrix(array_ti.spec.type))
@@ -3255,12 +3876,12 @@ _slang_gen_array_element(slang_assemble_ctx * A, slang_operation *oper)
          }
 
          elem = new_node2(IR_ELEMENT, array, index);
-         elem->Store = _slang_new_ir_storage_relative(constIndex,
-                                                      elemSize,
-                                                      array->Store);
 
-         assert(elem->Store->Parent);
-         /* XXX try to do some array bounds checking here */
+         /* The storage info here will be updated during code emit */
+         elem->Store = _slang_new_ir_storage(array->Store->File,
+                                             array->Store->Index,
+                                             elemSize);
+         elem->Store->Swizzle = _slang_var_swizzle(elemSize, 0);
          return elem;
       }
       else {
@@ -3280,10 +3901,10 @@ _slang_gen_compare(slang_assemble_ctx *A, slang_operation *oper,
    slang_ir_node *n;
    
    slang_typeinfo_construct(&t0);
-   _slang_typeof_operation(A, &oper->children[0], &t0);
+   typeof_operation(A, &oper->children[0], &t0);
 
    slang_typeinfo_construct(&t1);
-   _slang_typeof_operation(A, &oper->children[0], &t1);
+   typeof_operation(A, &oper->children[0], &t1);
 
    if (t0.spec.type == SLANG_SPEC_ARRAY ||
        t1.spec.type == SLANG_SPEC_ARRAY) {
@@ -3535,6 +4156,8 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper)
    case SLANG_OPER_CALL:
       return _slang_gen_function_call_name(A, (const char *) oper->a_id,
                                            oper, NULL);
+   case SLANG_OPER_METHOD:
+      return _slang_gen_method_call(A, oper);
    case SLANG_OPER_RETURN:
       return _slang_gen_return(A, oper);
    case SLANG_OPER_LABEL:
@@ -3616,23 +4239,20 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper)
 
 
 /**
- * Compute total size of array give size of element, number of elements.
+ * Check if the given type specifier is a rectangular texture sampler.
  */
-static GLint
-array_size(GLint baseSize, GLint arrayLen)
+static GLboolean
+is_rect_sampler_spec(const slang_type_specifier *spec)
 {
-   GLint total;
-   if (arrayLen > 1) {
-      /* round up base type to multiple of 4 */
-      total = ((baseSize + 3) & ~0x3) * MAX2(arrayLen, 1);
-   }
-   else {
-      total = baseSize;
+   while (spec->_array) {
+      spec = spec->_array;
    }
-   return total;
+   return spec->type == SLANG_SPEC_SAMPLER2DRECT ||
+          spec->type == SLANG_SPEC_SAMPLER2DRECTSHADOW;
 }
 
 
+
 /**
  * Called by compiler when a global variable has been parsed/compiled.
  * Here we examine the variable's type to determine what kind of register
@@ -3656,8 +4276,14 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var,
    slang_ir_storage *store = NULL;
    int dbg = 0;
    const GLenum datatype = _slang_gltype_from_specifier(&var->type.specifier);
-   const GLint texIndex = sampler_to_texture_index(var->type.specifier.type);
    const GLint size = _slang_sizeof_type_specifier(&var->type.specifier);
+   const GLint arrayLen = _slang_array_length(var);
+   const GLint totalSize = _slang_array_size(size, arrayLen);
+   GLint texIndex = sampler_to_texture_index(var->type.specifier.type);
+
+   /* check for sampler2D arrays */
+   if (texIndex == -1 && var->type.specifier._array)
+      texIndex = sampler_to_texture_index(var->type.specifier._array->type);
 
    if (texIndex != -1) {
       /* This is a texture sampler variable...
@@ -3671,21 +4297,37 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var,
       }
 #if FEATURE_es2_glsl /* XXX should use FEATURE_texture_rect */
       /* disallow rect samplers */
-      if (var->type.specifier.type == SLANG_SPEC_SAMPLER2DRECT ||
-          var->type.specifier.type == SLANG_SPEC_SAMPLER2DRECTSHADOW) {
+      if (is_rect_sampler_spec(&var->type.specifier)) {
          slang_info_log_error(A->log, "invalid sampler type for '%s'", varName);
          return GL_FALSE;
       }
+#else
+      (void) is_rect_sampler_spec; /* silence warning */
 #endif
       {
          GLint sampNum = _mesa_add_sampler(prog->Parameters, varName, datatype);
-         store = _slang_new_ir_storage(PROGRAM_SAMPLER, sampNum, texIndex);
+         store = _slang_new_ir_storage_sampler(sampNum, texIndex, totalSize);
+
+         /* If we have a sampler array, then we need to allocate the 
+	  * additional samplers to ensure we don't allocate them elsewhere.
+	  * We can't directly use _mesa_add_sampler() as that checks the
+	  * varName and gets a match, so we call _mesa_add_parameter()
+	  * directly and use the last sampler number from the call above.
+	  */
+	 if (arrayLen > 0) {
+	    GLint a = arrayLen - 1;
+	    GLint i;
+	    for (i = 0; i < a; i++) {
+               GLfloat value = (GLfloat)(i + sampNum + 1);
+               (void) _mesa_add_parameter(prog->Parameters, PROGRAM_SAMPLER,
+                                 varName, 1, datatype, &value, NULL, 0x0);
+	    }
+	 }
       }
       if (dbg) printf("SAMPLER ");
    }
    else if (var->type.qualifier == SLANG_QUAL_UNIFORM) {
       /* Uniform variable */
-      const GLint totalSize = array_size(size, var->array_len);
       const GLuint swizzle = _slang_var_swizzle(totalSize, 0);
 
       if (prog) {
@@ -3727,27 +4369,10 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var,
             }
          }
          else {
-            GLint uniformLoc;
-            const GLfloat *initialValues = NULL;
-            if (var->initializer) {
-               _slang_simplify(var->initializer, &A->space, A->atoms);
-               if (var->initializer->type == SLANG_OPER_LITERAL_FLOAT ||
-                   var->initializer->type == SLANG_OPER_LITERAL_INT) {
-                  /* simple float/vector initializer */
-                  initialValues = var->initializer->literal;
-               }
-               else {
-                  /* complex initializer */
-                  slang_info_log_error(A->log,
-                     "unsupported initializer for uniform '%s'", varName);
-                  return GL_FALSE;
-               }
-            }
-
-            uniformLoc = _mesa_add_uniform(prog->Parameters, varName,
-                                           totalSize, datatype, initialValues);
-            store = _slang_new_ir_storage_swz(PROGRAM_UNIFORM, uniformLoc,
-                                              totalSize, swizzle);
+            /* non-struct uniform */
+            if (!_slang_gen_var_decl(A, var, var->initializer))
+               return GL_FALSE;
+            store = var->store;
          }
       }
       else {
@@ -3761,8 +4386,6 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var,
       if (dbg) printf("UNIFORM (sz %d) ", totalSize);
    }
    else if (var->type.qualifier == SLANG_QUAL_VARYING) {
-      const GLint totalSize = array_size(size, var->array_len);
-
       /* varyings must be float, vec or mat */
       if (!_slang_type_is_float_vec_mat(var->type.specifier.type) &&
           var->type.specifier.type != SLANG_SPEC_ARRAY) {
@@ -3780,8 +4403,19 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var,
 
       if (prog) {
          /* user-defined varying */
-         GLint varyingLoc = _mesa_add_varying(prog->Varying, varName, totalSize);
-         GLuint swizzle = _slang_var_swizzle(size, 0);
+         GLbitfield flags;
+         GLint varyingLoc;
+         GLuint swizzle;
+
+         flags = 0x0;
+         if (var->type.centroid == SLANG_CENTROID)
+            flags |= PROG_PARAM_BIT_CENTROID;
+         if (var->type.variant == SLANG_INVARIANT)
+            flags |= PROG_PARAM_BIT_INVARIANT;
+
+         varyingLoc = _mesa_add_varying(prog->Varying, varName,
+                                        totalSize, flags);
+         swizzle = _slang_var_swizzle(size, 0);
          store = _slang_new_ir_storage_swz(PROGRAM_VARYING, varyingLoc,
                                            totalSize, swizzle);
       }
@@ -3869,27 +4503,10 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var,
       slang_ir_node *n;
 
       /* IR node to declare the variable */
-      n = _slang_gen_var_decl(A, var);
-
-      /* IR code for the var's initializer, if present */
-      if (var->initializer) {
-         slang_ir_node *lhs, *rhs, *init;
-
-         /* Generate IR_COPY instruction to initialize the variable */
-         lhs = new_node0(IR_VAR);
-         lhs->Var = var;
-         lhs->Store = n->Store;
-
-         /* constant folding, etc */
-         _slang_simplify(var->initializer, &A->space, A->atoms);
-
-         rhs = _slang_gen_operation(A, var->initializer);
-         assert(rhs);
-         init = new_node2(IR_COPY, lhs, rhs);
-         n = new_seq(n, init);
-      }
+      n = _slang_gen_var_decl(A, var, var->initializer);
 
-      success = _slang_emit_code(n, A->vartable, A->program, GL_FALSE, A->log);
+      /* emit GPU instructions */
+      success = _slang_emit_code(n, A->vartable, A->program, A->pragmas, GL_FALSE, A->log);
 
       _slang_free_ir_tree(n);
    }
@@ -3898,7 +4515,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var,
                    store ? store->Index : -2);
 
    if (store)
-      var->aux = store;  /* save var's storage info */
+      var->store = store;  /* save var's storage info */
 
    var->declared = GL_TRUE;
 
@@ -3999,7 +4616,7 @@ _slang_codegen_function(slang_assemble_ctx * A, slang_function * fun)
 #endif
 
    /* Emit program instructions */
-   success = _slang_emit_code(n, A->vartable, A->program, GL_TRUE, A->log);
+   success = _slang_emit_code(n, A->vartable, A->program, A->pragmas, GL_TRUE, A->log);
    _slang_free_ir_tree(n);
 
    /* free codegen context */
diff --git a/src/mesa/shader/slang/slang_codegen.h b/src/mesa/shader/slang/slang_codegen.h
index ba7ca4c142..f2daa034e4 100644
--- a/src/mesa/shader/slang/slang_codegen.h
+++ b/src/mesa/shader/slang/slang_codegen.h
@@ -31,6 +31,20 @@
 #include "slang_compile.h"
 
 
+typedef struct slang_assemble_ctx_
+{
+   slang_atom_pool *atoms;
+   slang_name_space space;
+   struct gl_program *program;
+   struct gl_sl_pragmas *pragmas;
+   slang_var_table *vartable;
+   slang_info_log *log;
+   struct slang_label_ *curFuncEndLabel;
+   struct slang_ir_node_ *CurLoop;
+   struct slang_function_ *CurFunction;
+} slang_assemble_ctx;
+
+
 extern GLuint
 _slang_sizeof_type_specifier(const slang_type_specifier *spec);
 
diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c
index 72b83c0422..cfed977905 100644
--- a/src/mesa/shader/slang/slang_compile.c
+++ b/src/mesa/shader/slang/slang_compile.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.2
  *
  * Copyright (C) 2005-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2008 VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -74,18 +74,6 @@ legal_identifier(slang_atom name)
 }
 
 
-/**
- * Allocate storage for a variable of 'size' bytes from given pool.
- * Return the allocated address for the variable.
- */
-static GLuint
-slang_var_pool_alloc(slang_var_pool * pool, unsigned int size)
-{
-   const GLuint addr = pool->next_addr;
-   pool->next_addr += size;
-   return addr;
-}
-
 /*
  * slang_code_unit
  */
@@ -120,7 +108,6 @@ _slang_code_object_ctr(slang_code_object * self)
    for (i = 0; i < SLANG_BUILTIN_TOTAL; i++)
       _slang_code_unit_ctr(&self->builtin[i], self);
    _slang_code_unit_ctr(&self->unit, self);
-   self->varpool.next_addr = 0;
    slang_atom_pool_construct(&self->atompool);
 }
 
@@ -156,17 +143,28 @@ typedef struct slang_output_ctx_
    slang_variable_scope *vars;
    slang_function_scope *funs;
    slang_struct_scope *structs;
-   slang_var_pool *global_pool;
    struct gl_program *program;
+   struct gl_sl_pragmas *pragmas;
    slang_var_table *vartable;
    GLuint default_precision[TYPE_SPECIFIER_COUNT];
    GLboolean allow_precision;
    GLboolean allow_invariant;
    GLboolean allow_centroid;
+   GLboolean allow_array_types;  /* float[] syntax */
 } slang_output_ctx;
 
 /* _slang_compile() */
 
+
+/* Debugging aid, print file/line where parsing error is detected */
+#define RETURN0 \
+   do { \
+      if (0) \
+         printf("slang error at %s:%d\n", __FILE__, __LINE__); \
+      return 0; \
+   } while (0)
+
+
 static void
 parse_identifier_str(slang_parse_ctx * C, char **id)
 {
@@ -223,7 +221,7 @@ parse_float(slang_parse_ctx * C, float *number)
                                   _mesa_strlen(exponent) + 3) * sizeof(char));
    if (whole == NULL) {
       slang_info_log_memory(C->L);
-      return 0;
+      RETURN0;
    }
 
    slang_string_copy(whole, integral);
@@ -240,14 +238,14 @@ parse_float(slang_parse_ctx * C, float *number)
 }
 
 /* revision number - increment after each change affecting emitted output */
-#define REVISION 4
+#define REVISION 5
 
 static int
 check_revision(slang_parse_ctx * C)
 {
    if (*C->I != REVISION) {
       slang_info_log_error(C->L, "Internal compiler error.");
-      return 0;
+      RETURN0;
    }
    C->I++;
    return 1;
@@ -259,6 +257,10 @@ static int parse_expression(slang_parse_ctx *, slang_output_ctx *,
                             slang_operation *);
 static int parse_type_specifier(slang_parse_ctx *, slang_output_ctx *,
                                 slang_type_specifier *);
+static int
+parse_type_array_size(slang_parse_ctx *C,
+                      slang_output_ctx *O,
+                      GLint *array_len);
 
 static GLboolean
 parse_array_len(slang_parse_ctx * C, slang_output_ctx * O, GLuint * len)
@@ -285,7 +287,7 @@ parse_array_len(slang_parse_ctx * C, slang_output_ctx * O, GLuint * len)
       result = GL_TRUE;
       *len = (GLint) array_size.literal[0];
    } else if (array_size.type == SLANG_OPER_IDENTIFIER) {
-      slang_variable *var = _slang_locate_variable(array_size.locals, array_size.a_id, GL_TRUE);
+      slang_variable *var = _slang_variable_locate(array_size.locals, array_size.a_id, GL_TRUE);
       if (!var) {
          slang_info_log_error(C->L, "undefined variable '%s'",
                               (char *) array_size.a_id);
@@ -330,12 +332,26 @@ calculate_var_size(slang_parse_ctx * C, slang_output_ctx * O,
    return GL_TRUE;
 }
 
+static void
+promote_type_to_array(slang_parse_ctx *C,
+                      slang_fully_specified_type *type,
+                      GLint array_len)
+{
+   slang_type_specifier *baseType =
+      slang_type_specifier_new(type->specifier.type, NULL, NULL);
+
+   type->specifier.type = SLANG_SPEC_ARRAY;
+   type->specifier._array = baseType;
+   type->array_len = array_len;
+}
+
+
 static GLboolean
 convert_to_array(slang_parse_ctx * C, slang_variable * var,
                  const slang_type_specifier * sp)
 {
-   /* sized array - mark it as array, copy the specifier to the array element and
-    * parse the expression */
+   /* sized array - mark it as array, copy the specifier to the array element
+    * and parse the expression */
    var->type.specifier.type = SLANG_SPEC_ARRAY;
    var->type.specifier._array = (slang_type_specifier *)
       _slang_alloc(sizeof(slang_type_specifier));
@@ -355,7 +371,8 @@ convert_to_array(slang_parse_ctx * C, slang_variable * var,
 static GLboolean
 parse_struct_field_var(slang_parse_ctx * C, slang_output_ctx * O,
                        slang_variable * var, slang_atom a_name,
-                       const slang_type_specifier * sp)
+                       const slang_type_specifier * sp,
+                       GLuint array_len)
 {
    var->a_name = a_name;
    if (var->a_name == SLANG_ATOM_NULL)
@@ -363,10 +380,19 @@ parse_struct_field_var(slang_parse_ctx * C, slang_output_ctx * O,
 
    switch (*C->I++) {
    case FIELD_NONE:
-      if (!slang_type_specifier_copy(&var->type.specifier, sp))
-         return GL_FALSE;
+      if (array_len != -1) {
+         if (!convert_to_array(C, var, sp))
+            return GL_FALSE;
+         var->array_len = array_len;
+      }
+      else {
+         if (!slang_type_specifier_copy(&var->type.specifier, sp))
+            return GL_FALSE;
+      }
       break;
    case FIELD_ARRAY:
+      if (array_len != -1)
+         return GL_FALSE;
       if (!convert_to_array(C, var, sp))
          return GL_FALSE;
       if (!parse_array_len(C, O, &var->array_len))
@@ -384,26 +410,29 @@ parse_struct_field(slang_parse_ctx * C, slang_output_ctx * O,
                    slang_struct * st, slang_type_specifier * sp)
 {
    slang_output_ctx o = *O;
+   GLint array_len;
 
    o.structs = st->structs;
    if (!parse_type_specifier(C, &o, sp))
-      return 0;
+      RETURN0;
+   if (!parse_type_array_size(C, &o, &array_len))
+      RETURN0;
 
    do {
       slang_atom a_name;
       slang_variable *var = slang_variable_scope_grow(st->fields);
       if (!var) {
          slang_info_log_memory(C->L);
-         return 0;
+         RETURN0;
       }
       a_name = parse_identifier(C);
-      if (_slang_locate_variable(st->fields, a_name, GL_FALSE)) {
+      if (_slang_variable_locate(st->fields, a_name, GL_FALSE)) {
          slang_info_log_error(C->L, "duplicate field '%s'", (char *) a_name);
-         return 0;
+         RETURN0;
       }
 
-      if (!parse_struct_field_var(C, &o, var, a_name, sp))
-         return 0;
+      if (!parse_struct_field_var(C, &o, var, a_name, sp, array_len))
+         RETURN0;
    }
    while (*C->I++ != FIELD_NONE);
 
@@ -419,26 +448,26 @@ parse_struct(slang_parse_ctx * C, slang_output_ctx * O, slang_struct ** st)
    /* parse struct name (if any) and make sure it is unique in current scope */
    a_name = parse_identifier(C);
    if (a_name == SLANG_ATOM_NULL)
-      return 0;
+      RETURN0;
 
    name = slang_atom_pool_id(C->atoms, a_name);
    if (name[0] != '\0'
        && slang_struct_scope_find(O->structs, a_name, 0) != NULL) {
       slang_info_log_error(C->L, "%s: duplicate type name.", name);
-      return 0;
+      RETURN0;
    }
 
    /* set-up a new struct */
    *st = (slang_struct *) _slang_alloc(sizeof(slang_struct));
    if (*st == NULL) {
       slang_info_log_memory(C->L);
-      return 0;
+      RETURN0;
    }
    if (!slang_struct_construct(*st)) {
       _slang_free(*st);
       *st = NULL;
       slang_info_log_memory(C->L);
-      return 0;
+      RETURN0;
    }
    (**st).a_name = a_name;
    (**st).structs->outer_scope = O->structs;
@@ -450,7 +479,7 @@ parse_struct(slang_parse_ctx * C, slang_output_ctx * O, slang_struct ** st)
       slang_type_specifier_ctr(&sp);
       if (!parse_struct_field(C, O, *st, &sp)) {
          slang_type_specifier_dtr(&sp);
-         return 0;
+         RETURN0;
       }
       slang_type_specifier_dtr(&sp);
    }
@@ -468,14 +497,14 @@ parse_struct(slang_parse_ctx * C, slang_output_ctx * O, slang_struct ** st)
                                          * sizeof(slang_struct));
       if (O->structs->structs == NULL) {
          slang_info_log_memory(C->L);
-         return 0;
+         RETURN0;
       }
       s = &O->structs->structs[O->structs->num_structs];
       if (!slang_struct_construct(s))
-         return 0;
+         RETURN0;
       O->structs->num_structs++;
       if (!slang_struct_copy(s, *st))
-         return 0;
+         RETURN0;
    }
 
    return 1;
@@ -498,7 +527,7 @@ parse_type_variant(slang_parse_ctx * C, slang_type_variant *variant)
       *variant = SLANG_INVARIANT;
       return 1;
    default:
-      return 0;
+      RETURN0;
    }
 }
 
@@ -519,7 +548,7 @@ parse_type_centroid(slang_parse_ctx * C, slang_type_centroid *centroid)
       *centroid = SLANG_CENTROID;
       return 1;
    default:
-      return 0;
+      RETURN0;
    }
 }
 
@@ -560,7 +589,7 @@ parse_type_qualifier(slang_parse_ctx * C, slang_type_qualifier * qual)
       *qual = SLANG_QUAL_FIXEDINPUT;
       break;
    default:
-      return 0;
+      RETURN0;
    }
    return 1;
 }
@@ -698,7 +727,7 @@ parse_type_specifier(slang_parse_ctx * C, slang_output_ctx * O,
    case TYPE_SPECIFIER_STRUCT:
       spec->type = SLANG_SPEC_STRUCT;
       if (!parse_struct(C, O, &spec->_struct))
-         return 0;
+         RETURN0;
       break;
    case TYPE_SPECIFIER_TYPENAME:
       spec->type = SLANG_SPEC_STRUCT;
@@ -708,35 +737,60 @@ parse_type_specifier(slang_parse_ctx * C, slang_output_ctx * O,
 
          a_name = parse_identifier(C);
          if (a_name == NULL)
-            return 0;
+            RETURN0;
 
          stru = slang_struct_scope_find(O->structs, a_name, 1);
          if (stru == NULL) {
             slang_info_log_error(C->L, "undeclared type name '%s'",
                                  slang_atom_pool_id(C->atoms, a_name));
-            return 0;
+            RETURN0;
          }
 
          spec->_struct = (slang_struct *) _slang_alloc(sizeof(slang_struct));
          if (spec->_struct == NULL) {
             slang_info_log_memory(C->L);
-            return 0;
+            RETURN0;
          }
          if (!slang_struct_construct(spec->_struct)) {
             _slang_free(spec->_struct);
             spec->_struct = NULL;
-            return 0;
+            RETURN0;
          }
          if (!slang_struct_copy(spec->_struct, stru))
-            return 0;
+            RETURN0;
       }
       break;
    default:
-      return 0;
+      RETURN0;
    }
    return 1;
 }
 
+#define TYPE_SPECIFIER_NONARRAY 0
+#define TYPE_SPECIFIER_ARRAY    1
+
+static int
+parse_type_array_size(slang_parse_ctx *C,
+                      slang_output_ctx *O,
+                      GLint *array_len)
+{
+   GLuint size;
+
+   switch (*C->I++) {
+   case TYPE_SPECIFIER_NONARRAY:
+      *array_len = -1; /* -1 = not an array */
+      break;
+   case TYPE_SPECIFIER_ARRAY:
+      if (!parse_array_len(C, O, &size))
+         RETURN0;
+      *array_len = (GLint) size;
+      break;
+   default:
+      assert(0);
+      RETURN0;
+   }
+   return 1;
+}
 
 #define PRECISION_DEFAULT 0
 #define PRECISION_LOW     1
@@ -762,7 +816,7 @@ parse_type_precision(slang_parse_ctx *C,
       *precision = SLANG_PREC_HIGH;
       return 1;
    default:
-      return 0;
+      RETURN0;
    }
 }
 
@@ -771,36 +825,39 @@ parse_fully_specified_type(slang_parse_ctx * C, slang_output_ctx * O,
                            slang_fully_specified_type * type)
 {
    if (!parse_type_variant(C, &type->variant))
-      return 0;
+      RETURN0;
   
    if (!parse_type_centroid(C, &type->centroid))
-      return 0;
+      RETURN0;
 
    if (!parse_type_qualifier(C, &type->qualifier))
-      return 0;
+      RETURN0;
 
    if (!parse_type_precision(C, &type->precision))
-      return 0;
+      RETURN0;
 
    if (!parse_type_specifier(C, O, &type->specifier))
-      return 0;
+      RETURN0;
+
+   if (!parse_type_array_size(C, O, &type->array_len))
+      RETURN0;
 
    if (!O->allow_invariant && type->variant == SLANG_INVARIANT) {
       slang_info_log_error(C->L,
          "'invariant' keyword not allowed (perhaps set #version 120)");
-      return 0;
+      RETURN0;
    }
 
    if (!O->allow_centroid && type->centroid == SLANG_CENTROID) {
       slang_info_log_error(C->L,
          "'centroid' keyword not allowed (perhaps set #version 120)");
-      return 0;
+      RETURN0;
    }
    else if (type->centroid == SLANG_CENTROID &&
             type->qualifier != SLANG_QUAL_VARYING) {
       slang_info_log_error(C->L,
          "'centroid' keyword only allowed for varying vars");
-      return 0;
+      RETURN0;
    }
 
 
@@ -809,7 +866,7 @@ parse_fully_specified_type(slang_parse_ctx * C, slang_output_ctx * O,
        type->variant == SLANG_INVARIANT) {
       slang_info_log_error(C->L,
                            "invariant qualifer only allowed for varying vars");
-      return 0;
+      RETURN0;
    }
    */
 
@@ -824,10 +881,20 @@ parse_fully_specified_type(slang_parse_ctx * C, slang_output_ctx * O,
       /* only default is allowed */
       if (type->precision != SLANG_PREC_DEFAULT) {
          slang_info_log_error(C->L, "precision qualifiers not allowed");
-         return 0;
+         RETURN0;
       }
    }
 
+   if (!O->allow_array_types && type->array_len >= 0) {
+      slang_info_log_error(C->L, "first-class array types not allowed");
+      RETURN0;
+   }
+
+   if (type->array_len >= 0) {
+      /* convert type to array type (ex: convert "int" to "array of int" */
+      promote_type_to_array(C, type, type->array_len);
+   }
+
    return 1;
 }
 
@@ -895,6 +962,7 @@ parse_fully_specified_type(slang_parse_ctx * C, slang_output_ctx * O,
 #define OP_POSTINCREMENT 60
 #define OP_POSTDECREMENT 61
 #define OP_PRECISION 62
+#define OP_METHOD 63
 
 
 /**
@@ -928,14 +996,18 @@ static int
 parse_statement(slang_parse_ctx * C, slang_output_ctx * O,
                 slang_operation * oper)
 {
+   int op;
+
    oper->locals->outer_scope = O->vars;
-   switch (*C->I++) {
+
+   op = *C->I++;
+   switch (op) {
    case OP_BLOCK_BEGIN_NO_NEW_SCOPE:
       /* parse child statements, do not create new variable scope */
       oper->type = SLANG_OPER_BLOCK_NO_NEW_SCOPE;
       while (*C->I != OP_END)
-         if (!parse_child_operation(C, O, oper, 1))
-            return 0;
+         if (!parse_child_operation(C, O, oper, GL_TRUE))
+            RETURN0;
       C->I++;
       break;
    case OP_BLOCK_BEGIN_NEW_SCOPE:
@@ -946,8 +1018,8 @@ parse_statement(slang_parse_ctx * C, slang_output_ctx * O,
          oper->type = SLANG_OPER_BLOCK_NEW_SCOPE;
          o.vars = oper->locals;
          while (*C->I != OP_END)
-            if (!parse_child_operation(C, &o, oper, 1))
-               return 0;
+            if (!parse_child_operation(C, &o, oper, GL_TRUE))
+               RETURN0;
          C->I++;
       }
       break;
@@ -963,7 +1035,7 @@ parse_statement(slang_parse_ctx * C, slang_output_ctx * O,
           * than one declarators
           */
          if (!parse_declaration(C, O))
-            return 0;
+            RETURN0;
          if (first_var < O->vars->num_variables) {
             const unsigned int num_vars = O->vars->num_variables - first_var;
             unsigned int i;
@@ -972,18 +1044,23 @@ parse_statement(slang_parse_ctx * C, slang_output_ctx * O,
             oper->children = slang_operation_new(num_vars);
             if (oper->children == NULL) {
                slang_info_log_memory(C->L);
-               return 0;
+               RETURN0;
             }
             for (i = first_var; i < O->vars->num_variables; i++) {
                slang_operation *o = &oper->children[i - first_var];
+               slang_variable *var = O->vars->variables[i];
                o->type = SLANG_OPER_VARIABLE_DECL;
                o->locals->outer_scope = O->vars;
-               o->a_id = O->vars->variables[i]->a_name;
+               o->a_id = var->a_name;
+
+               /* new/someday...
+               calculate_var_size(C, O, var);
+               */
 
                if (!legal_identifier(o->a_id)) {
                   slang_info_log_error(C->L, "illegal variable name '%s'",
                                        (char *) o->a_id);
-                  return 0;
+                  RETURN0;
                }
             }
          }
@@ -996,10 +1073,10 @@ parse_statement(slang_parse_ctx * C, slang_output_ctx * O,
       oper->type = SLANG_OPER_ASM;
       oper->a_id = parse_identifier(C);
       if (oper->a_id == SLANG_ATOM_NULL)
-         return 0;
+         RETURN0;
       while (*C->I != OP_END) {
-         if (!parse_child_operation(C, O, oper, 0))
-            return 0;
+         if (!parse_child_operation(C, O, oper, GL_FALSE))
+            RETURN0;
       }
       C->I++;
       break;
@@ -1014,22 +1091,22 @@ parse_statement(slang_parse_ctx * C, slang_output_ctx * O,
       break;
    case OP_RETURN:
       oper->type = SLANG_OPER_RETURN;
-      if (!parse_child_operation(C, O, oper, 0))
-         return 0;
+      if (!parse_child_operation(C, O, oper, GL_FALSE))
+         RETURN0;
       break;
    case OP_EXPRESSION:
       oper->type = SLANG_OPER_EXPRESSION;
-      if (!parse_child_operation(C, O, oper, 0))
-         return 0;
+      if (!parse_child_operation(C, O, oper, GL_FALSE))
+         RETURN0;
       break;
    case OP_IF:
       oper->type = SLANG_OPER_IF;
-      if (!parse_child_operation(C, O, oper, 0))
-         return 0;
-      if (!parse_child_operation(C, O, oper, 1))
-         return 0;
-      if (!parse_child_operation(C, O, oper, 1))
-         return 0;
+      if (!parse_child_operation(C, O, oper, GL_FALSE))
+         RETURN0;
+      if (!parse_child_operation(C, O, oper, GL_TRUE))
+         RETURN0;
+      if (!parse_child_operation(C, O, oper, GL_TRUE))
+         RETURN0;
       break;
    case OP_WHILE:
       {
@@ -1037,18 +1114,18 @@ parse_statement(slang_parse_ctx * C, slang_output_ctx * O,
 
          oper->type = SLANG_OPER_WHILE;
          o.vars = oper->locals;
-         if (!parse_child_operation(C, &o, oper, 1))
-            return 0;
-         if (!parse_child_operation(C, &o, oper, 1))
-            return 0;
+         if (!parse_child_operation(C, &o, oper, GL_TRUE))
+            RETURN0;
+         if (!parse_child_operation(C, &o, oper, GL_TRUE))
+            RETURN0;
       }
       break;
    case OP_DO:
       oper->type = SLANG_OPER_DO;
-      if (!parse_child_operation(C, O, oper, 1))
-         return 0;
-      if (!parse_child_operation(C, O, oper, 0))
-         return 0;
+      if (!parse_child_operation(C, O, oper, GL_TRUE))
+         RETURN0;
+      if (!parse_child_operation(C, O, oper, GL_FALSE))
+         RETURN0;
       break;
    case OP_FOR:
       {
@@ -1056,14 +1133,14 @@ parse_statement(slang_parse_ctx * C, slang_output_ctx * O,
 
          oper->type = SLANG_OPER_FOR;
          o.vars = oper->locals;
-         if (!parse_child_operation(C, &o, oper, 1))
-            return 0;
-         if (!parse_child_operation(C, &o, oper, 1))
-            return 0;
-         if (!parse_child_operation(C, &o, oper, 0))
-            return 0;
-         if (!parse_child_operation(C, &o, oper, 1))
-            return 0;
+         if (!parse_child_operation(C, &o, oper, GL_TRUE))
+            RETURN0;
+         if (!parse_child_operation(C, &o, oper, GL_TRUE))
+            RETURN0;
+         if (!parse_child_operation(C, &o, oper, GL_FALSE))
+            RETURN0;
+         if (!parse_child_operation(C, &o, oper, GL_TRUE))
+            RETURN0;
       }
       break;
    case OP_PRECISION:
@@ -1077,7 +1154,8 @@ parse_statement(slang_parse_ctx * C, slang_output_ctx * O,
       }
       break;
    default:
-      return 0;
+      /*printf("Unexpected operation %d\n", op);*/
+      RETURN0;
    }
    return 1;
 }
@@ -1092,7 +1170,7 @@ handle_nary_expression(slang_parse_ctx * C, slang_operation * op,
    op->children = slang_operation_new(n);
    if (op->children == NULL) {
       slang_info_log_memory(C->L);
-      return 0;
+      RETURN0;
    }
    op->num_children = n;
 
@@ -1110,7 +1188,7 @@ handle_nary_expression(slang_parse_ctx * C, slang_operation * op,
                      *total_ops * sizeof(slang_operation));
    if (*ops == NULL) {
       slang_info_log_memory(C->L);
-      return 0;
+      RETURN0;
    }
    return 1;
 }
@@ -1124,6 +1202,9 @@ is_constructor_name(const char *name, slang_atom a_name,
    return slang_struct_scope_find(structs, a_name, 1) != NULL;
 }
 
+#define FUNCTION_CALL_NONARRAY 0
+#define FUNCTION_CALL_ARRAY    1
+
 static int
 parse_expression(slang_parse_ctx * C, slang_output_ctx * O,
                  slang_operation * oper)
@@ -1143,12 +1224,12 @@ parse_expression(slang_parse_ctx * C, slang_output_ctx * O,
                         (num_ops + 1) * sizeof(slang_operation));
       if (ops == NULL) {
          slang_info_log_memory(C->L);
-         return 0;
+         RETURN0;
       }
       op = &ops[num_ops];
       if (!slang_operation_construct(op)) {
          slang_info_log_memory(C->L);
-         return 0;
+         RETURN0;
       }
       num_ops++;
       op->locals->outer_scope = O->vars;
@@ -1160,7 +1241,7 @@ parse_expression(slang_parse_ctx * C, slang_output_ctx * O,
       case OP_PUSH_BOOL:
          op->type = SLANG_OPER_LITERAL_BOOL;
          if (!parse_number(C, &number))
-            return 0;
+            RETURN0;
          op->literal[0] =
          op->literal[1] =
          op->literal[2] =
@@ -1170,7 +1251,7 @@ parse_expression(slang_parse_ctx * C, slang_output_ctx * O,
       case OP_PUSH_INT:
          op->type = SLANG_OPER_LITERAL_INT;
          if (!parse_number(C, &number))
-            return 0;
+            RETURN0;
          op->literal[0] =
          op->literal[1] =
          op->literal[2] =
@@ -1180,7 +1261,7 @@ parse_expression(slang_parse_ctx * C, slang_output_ctx * O,
       case OP_PUSH_FLOAT:
          op->type = SLANG_OPER_LITERAL_FLOAT;
          if (!parse_float(C, &op->literal[0]))
-            return 0;
+            RETURN0;
          op->literal[1] =
          op->literal[2] =
          op->literal[3] = op->literal[0];
@@ -1190,37 +1271,37 @@ parse_expression(slang_parse_ctx * C, slang_output_ctx * O,
          op->type = SLANG_OPER_IDENTIFIER;
          op->a_id = parse_identifier(C);
          if (op->a_id == SLANG_ATOM_NULL)
-            return 0;
+            RETURN0;
          break;
       case OP_SEQUENCE:
          op->type = SLANG_OPER_SEQUENCE;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_ASSIGN:
          op->type = SLANG_OPER_ASSIGN;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_ADDASSIGN:
          op->type = SLANG_OPER_ADDASSIGN;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_SUBASSIGN:
          op->type = SLANG_OPER_SUBASSIGN;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_MULASSIGN:
          op->type = SLANG_OPER_MULASSIGN;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_DIVASSIGN:
          op->type = SLANG_OPER_DIVASSIGN;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
          /*case OP_MODASSIGN: */
          /*case OP_LSHASSIGN: */
@@ -1231,22 +1312,22 @@ parse_expression(slang_parse_ctx * C, slang_output_ctx * O,
       case OP_SELECT:
          op->type = SLANG_OPER_SELECT;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 3))
-            return 0;
+            RETURN0;
          break;
       case OP_LOGICALOR:
          op->type = SLANG_OPER_LOGICALOR;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_LOGICALXOR:
          op->type = SLANG_OPER_LOGICALXOR;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_LOGICALAND:
          op->type = SLANG_OPER_LOGICALAND;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
          /*case OP_BITOR: */
          /*case OP_BITXOR: */
@@ -1254,97 +1335,106 @@ parse_expression(slang_parse_ctx * C, slang_output_ctx * O,
       case OP_EQUAL:
          op->type = SLANG_OPER_EQUAL;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_NOTEQUAL:
          op->type = SLANG_OPER_NOTEQUAL;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_LESS:
          op->type = SLANG_OPER_LESS;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_GREATER:
          op->type = SLANG_OPER_GREATER;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_LESSEQUAL:
          op->type = SLANG_OPER_LESSEQUAL;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_GREATEREQUAL:
          op->type = SLANG_OPER_GREATEREQUAL;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
          /*case OP_LSHIFT: */
          /*case OP_RSHIFT: */
       case OP_ADD:
          op->type = SLANG_OPER_ADD;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_SUBTRACT:
          op->type = SLANG_OPER_SUBTRACT;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_MULTIPLY:
          op->type = SLANG_OPER_MULTIPLY;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
       case OP_DIVIDE:
          op->type = SLANG_OPER_DIVIDE;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
          /*case OP_MODULUS: */
       case OP_PREINCREMENT:
          op->type = SLANG_OPER_PREINCREMENT;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 1))
-            return 0;
+            RETURN0;
          break;
       case OP_PREDECREMENT:
          op->type = SLANG_OPER_PREDECREMENT;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 1))
-            return 0;
+            RETURN0;
          break;
       case OP_PLUS:
          op->type = SLANG_OPER_PLUS;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 1))
-            return 0;
+            RETURN0;
          break;
       case OP_MINUS:
          op->type = SLANG_OPER_MINUS;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 1))
-            return 0;
+            RETURN0;
          break;
       case OP_NOT:
          op->type = SLANG_OPER_NOT;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 1))
-            return 0;
+            RETURN0;
          break;
          /*case OP_COMPLEMENT: */
       case OP_SUBSCRIPT:
          op->type = SLANG_OPER_SUBSCRIPT;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 2))
-            return 0;
+            RETURN0;
          break;
-      case OP_CALL:
-         op->type = SLANG_OPER_CALL;
+      case OP_METHOD:
+         op->type = SLANG_OPER_METHOD;
+         op->a_obj = parse_identifier(C);
+         if (op->a_obj == SLANG_ATOM_NULL)
+            RETURN0;
+
          op->a_id = parse_identifier(C);
          if (op->a_id == SLANG_ATOM_NULL)
-            return 0;
-         while (*C->I != OP_END)
-            if (!parse_child_operation(C, O, op, 0))
-               return 0;
+            RETURN0;
+
+         assert(*C->I == OP_END);
          C->I++;
 
+         while (*C->I != OP_END)
+            if (!parse_child_operation(C, O, op, GL_FALSE))
+               RETURN0;
+         C->I++;
+#if 0
+         /* don't lookup the method (not yet anyway) */
          if (!C->parsing_builtin
              && !slang_function_scope_find_by_name(O->funs, op->a_id, 1)) {
             const char *id;
@@ -1352,7 +1442,79 @@ parse_expression(slang_parse_ctx * C, slang_output_ctx * O,
             id = slang_atom_pool_id(C->atoms, op->a_id);
             if (!is_constructor_name(id, op->a_id, O->structs)) {
                slang_info_log_error(C->L, "%s: undeclared function name.", id);
-               return 0;
+               RETURN0;
+            }
+         }
+#endif
+         break;
+      case OP_CALL:
+         {
+            GLboolean array_constructor = GL_FALSE;
+            GLint array_constructor_size;
+
+            op->type = SLANG_OPER_CALL;
+            op->a_id = parse_identifier(C);
+            if (op->a_id == SLANG_ATOM_NULL)
+               RETURN0;
+            switch (*C->I++) {
+            case FUNCTION_CALL_NONARRAY:
+               /* Nothing to do. */
+               break;
+            case FUNCTION_CALL_ARRAY:
+               /* Calling an array constructor. For example:
+                *   float[3](1.1, 2.2, 3.3);
+                */
+               if (!O->allow_array_types) {
+                  slang_info_log_error(C->L,
+                                       "array constructors not allowed "
+                                       "in this GLSL version");
+                  RETURN0;
+               }
+               else {
+                  /* parse the array constructor size */
+                  slang_operation array_size;
+                  array_constructor = GL_TRUE;
+                  slang_operation_construct(&array_size);
+                  if (!parse_expression(C, O, &array_size)) {
+                     slang_operation_destruct(&array_size);
+                     return GL_FALSE;
+                  }
+                  if (array_size.type != SLANG_OPER_LITERAL_INT) {
+                     slang_info_log_error(C->L,
+                        "constructor array size is not an integer");
+                     slang_operation_destruct(&array_size);
+                     RETURN0;
+                  }
+                  array_constructor_size = (int) array_size.literal[0];
+                  op->array_constructor = GL_TRUE;
+                  slang_operation_destruct(&array_size);
+               }
+               break;
+            default:
+               assert(0);
+               RETURN0;
+            }
+            while (*C->I != OP_END)
+               if (!parse_child_operation(C, O, op, GL_FALSE))
+                  RETURN0;
+            C->I++;
+
+            if (array_constructor &&
+                array_constructor_size != op->num_children) {
+               slang_info_log_error(C->L, "number of parameters to array"
+                                    " constructor does not match array size");
+               RETURN0;
+            }
+
+            if (!C->parsing_builtin
+                && !slang_function_scope_find_by_name(O->funs, op->a_id, 1)) {
+               const char *id;
+
+               id = slang_atom_pool_id(C->atoms, op->a_id);
+               if (!is_constructor_name(id, op->a_id, O->structs)) {
+                  slang_info_log_error(C->L, "%s: undeclared function name.", id);
+                  RETURN0;
+               }
             }
          }
          break;
@@ -1360,22 +1522,22 @@ parse_expression(slang_parse_ctx * C, slang_output_ctx * O,
          op->type = SLANG_OPER_FIELD;
          op->a_id = parse_identifier(C);
          if (op->a_id == SLANG_ATOM_NULL)
-            return 0;
+            RETURN0;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 1))
-            return 0;
+            RETURN0;
          break;
       case OP_POSTINCREMENT:
          op->type = SLANG_OPER_POSTINCREMENT;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 1))
-            return 0;
+            RETURN0;
          break;
       case OP_POSTDECREMENT:
          op->type = SLANG_OPER_POSTDECREMENT;
          if (!handle_nary_expression(C, op, &ops, &num_ops, 1))
-            return 0;
+            RETURN0;
          break;
       default:
-         return 0;
+         RETURN0;
       }
    }
    C->I++;
@@ -1406,7 +1568,7 @@ parse_parameter_declaration(slang_parse_ctx * C, slang_output_ctx * O,
     * two at most) because not all combinations are valid
     */
    if (!parse_type_qualifier(C, &param->type.qualifier))
-      return 0;
+      RETURN0;
 
    param_qual = *C->I++;
    switch (param_qual) {
@@ -1414,7 +1576,7 @@ parse_parameter_declaration(slang_parse_ctx * C, slang_output_ctx * O,
       if (param->type.qualifier != SLANG_QUAL_CONST
           && param->type.qualifier != SLANG_QUAL_NONE) {
          slang_info_log_error(C->L, "Invalid type qualifier.");
-         return 0;
+         RETURN0;
       }
       break;
    case PARAM_QUALIFIER_OUT:
@@ -1422,7 +1584,7 @@ parse_parameter_declaration(slang_parse_ctx * C, slang_output_ctx * O,
          param->type.qualifier = SLANG_QUAL_OUT;
       else {
          slang_info_log_error(C->L, "Invalid type qualifier.");
-         return 0;
+         RETURN0;
       }
       break;
    case PARAM_QUALIFIER_INOUT:
@@ -1430,11 +1592,11 @@ parse_parameter_declaration(slang_parse_ctx * C, slang_output_ctx * O,
          param->type.qualifier = SLANG_QUAL_INOUT;
       else {
          slang_info_log_error(C->L, "Invalid type qualifier.");
-         return 0;
+         RETURN0;
       }
       break;
    default:
-      return 0;
+      RETURN0;
    }
 
    /* parse precision qualifier (lowp, mediump, highp */
@@ -1444,10 +1606,30 @@ parse_parameter_declaration(slang_parse_ctx * C, slang_output_ctx * O,
 
    /* parse parameter's type specifier and name */
    if (!parse_type_specifier(C, O, &param->type.specifier))
-      return 0;
+      RETURN0;
+   if (!parse_type_array_size(C, O, &param->type.array_len))
+      RETURN0;
    param->a_name = parse_identifier(C);
    if (param->a_name == SLANG_ATOM_NULL)
-      return 0;
+      RETURN0;
+
+   /* first-class array
+    */
+   if (param->type.array_len >= 0) {
+      slang_type_specifier p;
+
+      slang_type_specifier_ctr(&p);
+      if (!slang_type_specifier_copy(&p, &param->type.specifier)) {
+         slang_type_specifier_dtr(&p);
+         RETURN0;
+      }
+      if (!convert_to_array(C, param, &p)) {
+         slang_type_specifier_dtr(&p);
+         RETURN0;
+      }
+      slang_type_specifier_dtr(&p);
+      param->array_len = param->type.array_len;
+   }
 
    /* if the parameter is an array, parse its size (the size must be
     * explicitly defined
@@ -1455,24 +1637,29 @@ parse_parameter_declaration(slang_parse_ctx * C, slang_output_ctx * O,
    if (*C->I++ == PARAMETER_ARRAY_PRESENT) {
       slang_type_specifier p;
 
+      if (param->type.array_len >= 0) {
+         slang_info_log_error(C->L, "multi-dimensional arrays not allowed");
+         RETURN0;
+      }
       slang_type_specifier_ctr(&p);
       if (!slang_type_specifier_copy(&p, &param->type.specifier)) {
          slang_type_specifier_dtr(&p);
-         return GL_FALSE;
+         RETURN0;
       }
       if (!convert_to_array(C, param, &p)) {
          slang_type_specifier_dtr(&p);
-         return GL_FALSE;
+         RETURN0;
       }
       slang_type_specifier_dtr(&p);
       if (!parse_array_len(C, O, &param->array_len))
-         return GL_FALSE;
+         RETURN0;
    }
 
+#if 0
    /* calculate the parameter size */
    if (!calculate_var_size(C, O, param))
-      return GL_FALSE;
-
+      RETURN0;
+#endif
    /* TODO: allocate the local address here? */
    return 1;
 }
@@ -1564,13 +1751,13 @@ parse_operator_name(slang_parse_ctx * C)
             slang_atom_pool_atom(C->atoms, operator_names[i].o_name);
          if (atom == SLANG_ATOM_NULL) {
             slang_info_log_memory(C->L);
-            return 0;
+            RETURN0;
          }
          C->I++;
          return atom;
       }
    }
-   return 0;
+   RETURN0;
 }
 
 
@@ -1581,7 +1768,7 @@ parse_function_prototype(slang_parse_ctx * C, slang_output_ctx * O,
    GLuint functype;
    /* parse function type and name */
    if (!parse_fully_specified_type(C, O, &func->header.type))
-      return 0;
+      RETURN0;
 
    functype = *C->I++;
    switch (functype) {
@@ -1589,35 +1776,35 @@ parse_function_prototype(slang_parse_ctx * C, slang_output_ctx * O,
       func->kind = SLANG_FUNC_ORDINARY;
       func->header.a_name = parse_identifier(C);
       if (func->header.a_name == SLANG_ATOM_NULL)
-         return 0;
+         RETURN0;
       break;
    case FUNCTION_CONSTRUCTOR:
       func->kind = SLANG_FUNC_CONSTRUCTOR;
       if (func->header.type.specifier.type == SLANG_SPEC_STRUCT)
-         return 0;
+         RETURN0;
       func->header.a_name =
          slang_atom_pool_atom(C->atoms,
                               slang_type_specifier_type_to_string
                               (func->header.type.specifier.type));
       if (func->header.a_name == SLANG_ATOM_NULL) {
          slang_info_log_memory(C->L);
-         return 0;
+         RETURN0;
       }
       break;
    case FUNCTION_OPERATOR:
       func->kind = SLANG_FUNC_OPERATOR;
       func->header.a_name = parse_operator_name(C);
       if (func->header.a_name == SLANG_ATOM_NULL)
-         return 0;
+         RETURN0;
       break;
    default:
-      return 0;
+      RETURN0;
    }
 
    if (!legal_identifier(func->header.a_name)) {
       slang_info_log_error(C->L, "illegal function name '%s'",
                            (char *) func->header.a_name);
-      return 0;
+      RETURN0;
    }
 
    /* parse function parameters */
@@ -1625,10 +1812,10 @@ parse_function_prototype(slang_parse_ctx * C, slang_output_ctx * O,
       slang_variable *p = slang_variable_scope_grow(func->parameters);
       if (!p) {
          slang_info_log_memory(C->L);
-         return 0;
+         RETURN0;
       }
       if (!parse_parameter_declaration(C, O, p))
-         return 0;
+         RETURN0;
    }
 
    /* if the function returns a value, append a hidden __retVal 'out'
@@ -1662,19 +1849,19 @@ parse_function_definition(slang_parse_ctx * C, slang_output_ctx * O,
    slang_output_ctx o = *O;
 
    if (!parse_function_prototype(C, O, func))
-      return 0;
+      RETURN0;
 
    /* create function's body operation */
    func->body = (slang_operation *) _slang_alloc(sizeof(slang_operation));
    if (func->body == NULL) {
       slang_info_log_memory(C->L);
-      return 0;
+      RETURN0;
    }
    if (!slang_operation_construct(func->body)) {
       _slang_free(func->body);
       func->body = NULL;
       slang_info_log_memory(C->L);
-      return 0;
+      RETURN0;
    }
 
    /* to parse the body the parse context is modified in order to
@@ -1683,7 +1870,7 @@ parse_function_definition(slang_parse_ctx * C, slang_output_ctx * O,
    C->global_scope = GL_FALSE;
    o.vars = func->parameters;
    if (!parse_statement(C, &o, func->body))
-      return 0;
+      RETURN0;
 
    C->global_scope = GL_TRUE;
    return 1;
@@ -1775,66 +1962,86 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O,
    a_name = parse_identifier(C);
 
    /* check if name is already in this scope */
-   if (_slang_locate_variable(O->vars, a_name, GL_FALSE)) {
+   if (_slang_variable_locate(O->vars, a_name, GL_FALSE)) {
       slang_info_log_error(C->L,
                    "declaration of '%s' conflicts with previous declaration",
                    (char *) a_name);
-      return 0;
+      RETURN0;
    }
 
    /* make room for the new variable and initialize it */
    var = slang_variable_scope_grow(O->vars);
    if (!var) {
       slang_info_log_memory(C->L);
-      return 0;
+      RETURN0;
    }
 
-   /* copy the declarator qualifier type, parse the identifier */
+   /* copy the declarator type qualifier/etc info, parse the identifier */
    var->type.qualifier = type->qualifier;
+   var->type.centroid = type->centroid;
+   var->type.precision = type->precision;
+   var->type.variant = type->variant;
+   var->type.array_len = type->array_len;
    var->a_name = a_name;
    if (var->a_name == SLANG_ATOM_NULL)
-      return 0;
+      RETURN0;
 
    switch (*C->I++) {
    case VARIABLE_NONE:
       /* simple variable declarator - just copy the specifier */
       if (!slang_type_specifier_copy(&var->type.specifier, &type->specifier))
-         return 0;
+         RETURN0;
       break;
    case VARIABLE_INITIALIZER:
       /* initialized variable - copy the specifier and parse the expression */
-      if (!slang_type_specifier_copy(&var->type.specifier, &type->specifier))
-         return 0;
+      if (0 && type->array_len >= 0) {
+         /* The type was something like "float[4]" */
+         convert_to_array(C, var, &type->specifier);
+         var->array_len = type->array_len;
+      }
+      else {
+         if (!slang_type_specifier_copy(&var->type.specifier, &type->specifier))
+            RETURN0;
+      }
       var->initializer =
          (slang_operation *) _slang_alloc(sizeof(slang_operation));
       if (var->initializer == NULL) {
          slang_info_log_memory(C->L);
-         return 0;
+         RETURN0;
       }
       if (!slang_operation_construct(var->initializer)) {
          _slang_free(var->initializer);
          var->initializer = NULL;
          slang_info_log_memory(C->L);
-         return 0;
+         RETURN0;
       }
       if (!parse_expression(C, O, var->initializer))
-         return 0;
+         RETURN0;
       break;
    case VARIABLE_ARRAY_UNKNOWN:
       /* unsized array - mark it as array and copy the specifier to
-         the array element
-      */
+       * the array element
+       */
+      if (type->array_len >= 0) {
+         slang_info_log_error(C->L, "multi-dimensional arrays not allowed");
+         RETURN0;
+      }
       if (!convert_to_array(C, var, &type->specifier))
          return GL_FALSE;
       break;
    case VARIABLE_ARRAY_EXPLICIT:
+      if (type->array_len >= 0) {
+         /* the user is trying to do something like: float[2] x[3]; */
+         slang_info_log_error(C->L, "multi-dimensional arrays not allowed");
+         RETURN0;
+      }
       if (!convert_to_array(C, var, &type->specifier))
          return GL_FALSE;
       if (!parse_array_len(C, O, &var->array_len))
          return GL_FALSE;
       break;
    default:
-      return 0;
+      RETURN0;
    }
 
    /* allocate global address space for a variable with a known size */
@@ -1843,7 +2050,6 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O,
             && var->array_len == 0)) {
       if (!calculate_var_size(C, O, var))
          return GL_FALSE;
-      var->address = slang_var_pool_alloc(O->global_pool, var->size);
    }
 
    /* emit code for global var decl */
@@ -1854,11 +2060,12 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O,
       A.space.structs = O->structs;
       A.space.vars = O->vars;
       A.program = O->program;
+      A.pragmas = O->pragmas;
       A.vartable = O->vartable;
       A.log = C->L;
       A.curFuncEndLabel = NULL;
       if (!_slang_codegen_global_variable(&A, var, C->type))
-         return 0;
+         RETURN0;
    }
 
    /* initialize global variable */
@@ -1871,7 +2078,7 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O,
          A.space.structs = O->structs;
          A.space.vars = O->vars;
          if (!initialize_global(&A, var))
-            return 0;
+            RETURN0;
       }
    }
    return 1;
@@ -1888,17 +2095,17 @@ parse_init_declarator_list(slang_parse_ctx * C, slang_output_ctx * O)
 
    /* parse the fully specified type, common to all declarators */
    if (!slang_fully_specified_type_construct(&type))
-      return 0;
+      RETURN0;
    if (!parse_fully_specified_type(C, O, &type)) {
       slang_fully_specified_type_destruct(&type);
-      return 0;
+      RETURN0;
    }
 
    /* parse declarators, pass-in the parsed type */
    do {
       if (!parse_init_declarator(C, O, &type)) {
          slang_fully_specified_type_destruct(&type);
-         return 0;
+         RETURN0;
       }
    }
    while (*C->I++ == DECLARATOR_NEXT);
@@ -1976,10 +2183,8 @@ parse_function(slang_parse_ctx * C, slang_output_ctx * O, int definition,
          }
 
          /* destroy the existing function declaration and replace it
-          * with the new one, remember to save the fixup table
+          * with the new one
           */
-         parsed_func.fixups = found_func->fixups;
-         slang_fixup_table_init(&found_func->fixups);
          slang_function_destruct(found_func);
          *found_func = parsed_func;
       }
@@ -2005,18 +2210,18 @@ parse_declaration(slang_parse_ctx * C, slang_output_ctx * O)
    switch (*C->I++) {
    case DECLARATION_INIT_DECLARATOR_LIST:
       if (!parse_init_declarator_list(C, O))
-         return 0;
+         RETURN0;
       break;
    case DECLARATION_FUNCTION_PROTOTYPE:
       {
          slang_function *dummy_func;
 
          if (!parse_function(C, O, 0, &dummy_func))
-            return 0;
+            RETURN0;
       }
       break;
    default:
-      return 0;
+      RETURN0;
    }
    return 1;
 }
@@ -2028,7 +2233,7 @@ parse_default_precision(slang_parse_ctx * C, slang_output_ctx * O)
 
    if (!O->allow_precision) {
       slang_info_log_error(C->L, "syntax error at \"precision\"");
-      return 0;
+      RETURN0;
    }
 
    precision = *C->I++;
@@ -2041,7 +2246,7 @@ parse_default_precision(slang_parse_ctx * C, slang_output_ctx * O)
    default:
       _mesa_problem(NULL, "unexpected precision %d at %s:%d\n",
                     precision, __FILE__, __LINE__);
-      return 0;
+      RETURN0;
    }
 
    type = *C->I++;
@@ -2061,7 +2266,7 @@ parse_default_precision(slang_parse_ctx * C, slang_output_ctx * O)
    default:
       _mesa_problem(NULL, "unexpected type %d at %s:%d\n",
                     type, __FILE__, __LINE__);
-      return 0;
+      RETURN0;
    }
 
    assert(type < TYPE_SPECIFIER_COUNT);
@@ -2108,7 +2313,7 @@ parse_invariant(slang_parse_ctx * C, slang_output_ctx * O)
    }
    else {
       slang_info_log_error(C->L, "syntax error at \"invariant\"");
-      return 0;
+      RETURN0;
    }
 }
       
@@ -2145,8 +2350,8 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit,
    o.funs = &unit->funs;
    o.structs = &unit->structs;
    o.vars = &unit->vars;
-   o.global_pool = &unit->object->varpool;
    o.program = shader ? shader->Program : NULL;
+   o.pragmas = shader ? &shader->Pragmas : NULL;
    o.vartable = _slang_new_var_table(maxRegs);
    _slang_push_var_table(o.vartable);
 
@@ -2168,6 +2373,8 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit,
 #endif
    init_default_precision(&o, unit->type);
 
+   /* allow 'float[]' keyword? */
+   o.allow_array_types = (C->version >= 120) ? GL_TRUE : GL_FALSE;
 
    /* parse individual functions and declarations */
    while (*C->I != EXTERNAL_NULL) {
@@ -2213,6 +2420,7 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit,
       A.space.structs = o.structs;
       A.space.vars = o.vars;
       A.program = o.program;
+      A.pragmas = &shader->Pragmas;
       A.vartable = o.vartable;
       A.log = C->L;
 
@@ -2270,7 +2478,9 @@ static GLboolean
 compile_with_grammar(grammar id, const char *source, slang_code_unit * unit,
                      slang_unit_type type, slang_info_log * infolog,
                      slang_code_unit * builtin,
-                     struct gl_shader *shader)
+                     struct gl_shader *shader,
+                     const struct gl_extensions *extensions,
+                     struct gl_sl_pragmas *pragmas)
 {
    byte *prod;
    GLuint size, start, version;
@@ -2298,7 +2508,8 @@ compile_with_grammar(grammar id, const char *source, slang_code_unit * unit,
 
    /* Now preprocess the source string. */
    slang_string_init(&preprocessed);
-   if (!_slang_preprocess_directives(&preprocessed, &source[start], infolog)) {
+   if (!_slang_preprocess_directives(&preprocessed, &source[start],
+                                     infolog, extensions, pragmas)) {
       slang_string_free(&preprocessed);
       slang_info_log_error(infolog, "failed to preprocess the source.");
       return GL_FALSE;
@@ -2371,7 +2582,9 @@ static const byte slang_vertex_builtin_gc[] = {
 static GLboolean
 compile_object(grammar * id, const char *source, slang_code_object * object,
                slang_unit_type type, slang_info_log * infolog,
-               struct gl_shader *shader)
+               struct gl_shader *shader,
+               const struct gl_extensions *extensions,
+               struct gl_sl_pragmas *pragmas)
 {
    slang_code_unit *builtins = NULL;
    GLuint base_version = 110;
@@ -2470,7 +2683,7 @@ compile_object(grammar * id, const char *source, slang_code_object * object,
 
    /* compile the actual shader - pass-in built-in library for external shader */
    return compile_with_grammar(*id, source, &object->unit, type, infolog,
-                               builtins, shader);
+                               builtins, shader, extensions, pragmas);
 }
 
 
@@ -2493,7 +2706,8 @@ compile_shader(GLcontext *ctx, slang_code_object * object,
    _slang_code_object_dtr(object);
    _slang_code_object_ctr(object);
 
-   success = compile_object(&id, shader->Source, object, type, infolog, shader);
+   success = compile_object(&id, shader->Source, object, type, infolog, shader,
+                            &ctx->Extensions, &shader->Pragmas);
    if (id != 0)
       grammar_destroy(id);
    if (!success)
@@ -2580,6 +2794,12 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader)
    _mesa_print_program(shader->Program);
 #endif
 
+   shader->CompileStatus = success;
+
+   if (ctx->Shader.Flags & GLSL_LOG) {
+      _mesa_write_shader_to_file(shader);
+   }
+
    return success;
 }
 
diff --git a/src/mesa/shader/slang/slang_compile.h b/src/mesa/shader/slang/slang_compile.h
index 35e468b44e..7fb549d33d 100644
--- a/src/mesa/shader/slang/slang_compile.h
+++ b/src/mesa/shader/slang/slang_compile.h
@@ -37,6 +37,13 @@
 extern "C" {
 #endif
 
+typedef struct slang_name_space_
+{
+   struct slang_function_scope_ *funcs;
+   struct slang_struct_scope_ *structs;
+   struct slang_variable_scope_ *vars;
+} slang_name_space;
+
 typedef enum slang_unit_type_
 {
    SLANG_UNIT_FRAGMENT_SHADER,
@@ -46,12 +53,6 @@ typedef enum slang_unit_type_
 } slang_unit_type;
 
 
-typedef struct slang_var_pool_
-{
-   GLuint next_addr;
-} slang_var_pool;
-
-
 typedef struct slang_code_unit_
 {
    slang_variable_scope vars;
@@ -79,7 +80,6 @@ typedef struct slang_code_object_
 {
    slang_code_unit builtin[SLANG_BUILTIN_TOTAL];
    slang_code_unit unit;
-   slang_var_pool varpool;
    slang_atom_pool atompool;
 } slang_code_object;
 
diff --git a/src/mesa/shader/slang/slang_compile_function.c b/src/mesa/shader/slang/slang_compile_function.c
index 8405d7f778..4dd885176d 100644
--- a/src/mesa/shader/slang/slang_compile_function.c
+++ b/src/mesa/shader/slang/slang_compile_function.c
@@ -32,42 +32,6 @@
 #include "slang_compile.h"
 #include "slang_mem.h"
 
-/* slang_fixup_table */
-
-void
-slang_fixup_table_init(slang_fixup_table * fix)
-{
-   fix->table = NULL;
-   fix->count = 0;
-}
-
-void
-slang_fixup_table_free(slang_fixup_table * fix)
-{
-   _slang_free(fix->table);
-   slang_fixup_table_init(fix);
-}
-
-/**
- * Add a new fixup address to the table.
- */
-GLboolean
-slang_fixup_save(slang_fixup_table *fixups, GLuint address)
-{
-   fixups->table = (GLuint *)
-      _slang_realloc(fixups->table,
-                     fixups->count * sizeof(GLuint),
-                     (fixups->count + 1) * sizeof(GLuint));
-   if (fixups->table == NULL)
-      return GL_FALSE;
-   fixups->table[fixups->count] = address;
-   fixups->count++;
-   return GL_TRUE;
-}
-
-
-
-/* slang_function */
 
 int
 slang_function_construct(slang_function * func)
@@ -86,8 +50,6 @@ slang_function_construct(slang_function * func)
    _slang_variable_scope_ctr(func->parameters);
    func->param_count = 0;
    func->body = NULL;
-   func->address = ~0;
-   slang_fixup_table_init(&func->fixups);
    return 1;
 }
 
@@ -101,9 +63,22 @@ slang_function_destruct(slang_function * func)
       slang_operation_destruct(func->body);
       _slang_free(func->body);
    }
-   slang_fixup_table_free(&func->fixups);
 }
 
+
+slang_function *
+slang_function_new(slang_function_kind kind)
+{
+   slang_function *fun = (slang_function *)
+      _slang_alloc(sizeof(slang_function));
+   if (fun) {
+      slang_function_construct(fun);
+      fun->kind = kind;
+   }
+   return fun;
+}
+
+
 /*
  * slang_function_scope
  */
@@ -214,3 +189,74 @@ slang_function_scope_find(slang_function_scope * funcs, slang_function * fun,
       return slang_function_scope_find(funcs->outer_scope, fun, 1);
    return NULL;
 }
+
+
+/**
+ * Lookup a function according to name and parameter count/types.
+ */
+slang_function *
+_slang_function_locate(const slang_function_scope * funcs, slang_atom a_name,
+                       slang_operation * args, GLuint num_args,
+                       const slang_name_space * space, slang_atom_pool * atoms,
+                       slang_info_log *log, GLboolean *error)
+{
+   slang_typeinfo arg_ti[100];
+   GLuint i;
+
+   *error = GL_FALSE;
+
+   /* determine type of each argument */
+   assert(num_args < 100);
+   for (i = 0; i < num_args; i++) {
+      if (!slang_typeinfo_construct(&arg_ti[i]))
+         return NULL;
+      if (!_slang_typeof_operation(&args[i], space, &arg_ti[i], atoms, log)) {
+         return NULL;
+      }
+   }
+
+   /* loop over function scopes */
+   while (funcs) {
+
+      /* look for function with matching name and argument/param types */
+      for (i = 0; i < funcs->num_functions; i++) {
+         slang_function *f = &funcs->functions[i];
+         const GLuint haveRetValue = _slang_function_has_return_value(f);
+         GLuint j;
+
+         if (a_name != f->header.a_name)
+            continue;
+         if (f->param_count - haveRetValue != num_args)
+            continue;
+
+         /* compare parameter / argument types */
+         for (j = 0; j < num_args; j++) {
+            if (!slang_type_specifier_compatible(&arg_ti[j].spec,
+                              &f->parameters->variables[j]->type.specifier)) {
+               /* param/arg types don't match */
+               break;
+            }
+
+            /* "out" and "inout" formal parameter requires the actual
+             * argument to be an l-value.
+             */
+            if (!arg_ti[j].can_be_referenced &&
+                (f->parameters->variables[j]->type.qualifier == SLANG_QUAL_OUT ||
+                 f->parameters->variables[j]->type.qualifier == SLANG_QUAL_INOUT)) {
+               /* param is not an lvalue! */
+               *error = GL_TRUE;
+               return NULL;
+            }
+         }
+
+         if (j == num_args) {
+            /* name and args match! */
+            return f;
+         }
+      }
+
+      funcs = funcs->outer_scope;
+   }
+
+   return NULL;
+}
diff --git a/src/mesa/shader/slang/slang_compile_function.h b/src/mesa/shader/slang/slang_compile_function.h
index a59b094e08..a5445ec253 100644
--- a/src/mesa/shader/slang/slang_compile_function.h
+++ b/src/mesa/shader/slang/slang_compile_function.h
@@ -26,8 +26,6 @@
 #define SLANG_COMPILE_FUNCTION_H
 
 
-struct slang_code_unit_;
-
 /**
  * Types of functions.
  */
@@ -40,21 +38,6 @@ typedef enum slang_function_kind_
 
 
 /**
- * When we need to fill in addresses which we won't know until the future,
- * we keep track of them with a fix-up table.
- */
-typedef struct slang_fixup_table_
-{
-   GLuint *table;     /**< array[count] of addresses */
-   GLuint count;
-} slang_fixup_table;
-
-extern void slang_fixup_table_init(slang_fixup_table *);
-extern void slang_fixup_table_free(slang_fixup_table *);
-extern GLboolean slang_fixup_save(slang_fixup_table *fixups, GLuint address);
-
-
-/**
  * Description of a compiled shader function.
  */
 typedef struct slang_function_
@@ -64,12 +47,14 @@ typedef struct slang_function_
    slang_variable_scope *parameters; /**< formal parameters AND local vars */
    unsigned int param_count;   /**< number of formal params (no locals) */
    slang_operation *body;      /**< The instruction tree */
-   unsigned int address;       /**< Address of this func in memory */
-   slang_fixup_table fixups;   /**< Mem locations which need func's address */
 } slang_function;
 
 extern int slang_function_construct(slang_function *);
 extern void slang_function_destruct(slang_function *);
+extern slang_function *slang_function_new(slang_function_kind kind);
+
+extern GLboolean
+_slang_function_has_return_value(const slang_function *fun);
 
 
 /**
@@ -89,13 +74,19 @@ _slang_function_scope_ctr(slang_function_scope *);
 extern void
 slang_function_scope_destruct(slang_function_scope *);
 
-extern GLboolean
-_slang_function_has_return_value(const slang_function *fun);
-
 extern int
 slang_function_scope_find_by_name(slang_function_scope *, slang_atom, int);
 
 extern slang_function *
 slang_function_scope_find(slang_function_scope *, slang_function *, int);
 
+extern struct slang_function_ *
+_slang_function_locate(const struct slang_function_scope_ *funcs,
+                       slang_atom name, struct slang_operation_ *params,
+                       GLuint num_params,
+                       const struct slang_name_space_ *space,
+                       slang_atom_pool *atoms, slang_info_log *log,
+                       GLboolean *error);
+
+
 #endif /* SLANG_COMPILE_FUNCTION_H */
diff --git a/src/mesa/shader/slang/slang_compile_operation.c b/src/mesa/shader/slang/slang_compile_operation.c
index 53cf6faff9..e650616880 100644
--- a/src/mesa/shader/slang/slang_compile_operation.c
+++ b/src/mesa/shader/slang/slang_compile_operation.c
@@ -44,6 +44,7 @@ slang_operation_construct(slang_operation * oper)
    oper->num_children = 0;
    oper->literal[0] = 0.0;
    oper->literal_size = 1;
+   oper->array_constructor = GL_FALSE;
    oper->a_id = SLANG_ATOM_NULL;
    oper->locals = _slang_variable_scope_new(NULL);
    if (oper->locals == NULL)
@@ -90,7 +91,7 @@ slang_replace_scope(slang_operation *oper,
    if (oper->type == SLANG_OPER_VARIABLE_DECL) {
       /* search/replace in the initializer */
       slang_variable *var;
-      var = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE);
+      var = _slang_variable_locate(oper->locals, oper->a_id, GL_TRUE);
       if (var && var->initializer) {
          slang_replace_scope(var->initializer, oldScope, newScope);
       }
diff --git a/src/mesa/shader/slang/slang_compile_operation.h b/src/mesa/shader/slang/slang_compile_operation.h
index 4f92aa9a08..37af5d617c 100644
--- a/src/mesa/shader/slang/slang_compile_operation.h
+++ b/src/mesa/shader/slang/slang_compile_operation.h
@@ -94,6 +94,7 @@ typedef enum slang_operation_type_
    SLANG_OPER_SUBSCRIPT,        /* [expr] "[" [expr] "]" */
    SLANG_OPER_CALL,             /* [func name] [param] [param] [...] */
    SLANG_OPER_NON_INLINED_CALL, /* a real function call */
+   SLANG_OPER_METHOD,           /* method call, such as  v.length() */
    SLANG_OPER_FIELD,            /* i.e.: ".next" or ".xzy" or ".xxx" etc */
    SLANG_OPER_POSTINCREMENT,    /* [var] "++" */
    SLANG_OPER_POSTDECREMENT     /* [var] "--" */
@@ -115,10 +116,17 @@ typedef struct slang_operation_
    GLfloat literal[4];           /**< Used for float, int and bool values */
    GLuint literal_size;          /**< 1, 2, 3, or 4 */
    slang_atom a_id;              /**< type: asm, identifier, call, field */
+   slang_atom a_obj;             /**< object in a method call */
    slang_variable_scope *locals; /**< local vars for scope */
    struct slang_function_ *fun;  /**< If type == SLANG_OPER_CALL */
    struct slang_variable_ *var;  /**< If type == slang_oper_identier */
    struct slang_label_ *label;   /**< If type == SLANG_OPER_LABEL */
+   /** If type==SLANG_OPER_CALL and we're calling an array constructor,
+    * for which there's no real function, we need to have a flag to
+    * indicate such.  num_children indicates number of elements.
+    */
+   GLboolean array_constructor;
+   double x;
 } slang_operation;
 
 
diff --git a/src/mesa/shader/slang/slang_compile_variable.c b/src/mesa/shader/slang/slang_compile_variable.c
index 3428b49e16..eab912710e 100644
--- a/src/mesa/shader/slang/slang_compile_variable.c
+++ b/src/mesa/shader/slang/slang_compile_variable.c
@@ -33,105 +33,6 @@
 #include "slang_mem.h"
 
 
-typedef struct
-{
-   const char *name;
-   slang_type_specifier_type type;
-} type_specifier_type_name;
-
-static const type_specifier_type_name type_specifier_type_names[] = {
-   {"void", SLANG_SPEC_VOID},
-   {"bool", SLANG_SPEC_BOOL},
-   {"bvec2", SLANG_SPEC_BVEC2},
-   {"bvec3", SLANG_SPEC_BVEC3},
-   {"bvec4", SLANG_SPEC_BVEC4},
-   {"int", SLANG_SPEC_INT},
-   {"ivec2", SLANG_SPEC_IVEC2},
-   {"ivec3", SLANG_SPEC_IVEC3},
-   {"ivec4", SLANG_SPEC_IVEC4},
-   {"float", SLANG_SPEC_FLOAT},
-   {"vec2", SLANG_SPEC_VEC2},
-   {"vec3", SLANG_SPEC_VEC3},
-   {"vec4", SLANG_SPEC_VEC4},
-   {"mat2", SLANG_SPEC_MAT2},
-   {"mat3", SLANG_SPEC_MAT3},
-   {"mat4", SLANG_SPEC_MAT4},
-   {"mat2x3", SLANG_SPEC_MAT23},
-   {"mat3x2", SLANG_SPEC_MAT32},
-   {"mat2x4", SLANG_SPEC_MAT24},
-   {"mat4x2", SLANG_SPEC_MAT42},
-   {"mat3x4", SLANG_SPEC_MAT34},
-   {"mat4x3", SLANG_SPEC_MAT43},
-   {"sampler1D", SLANG_SPEC_SAMPLER1D},
-   {"sampler2D", SLANG_SPEC_SAMPLER2D},
-   {"sampler3D", SLANG_SPEC_SAMPLER3D},
-   {"samplerCube", SLANG_SPEC_SAMPLERCUBE},
-   {"sampler1DShadow", SLANG_SPEC_SAMPLER1DSHADOW},
-   {"sampler2DShadow", SLANG_SPEC_SAMPLER2DSHADOW},
-   {"sampler2DRect", SLANG_SPEC_SAMPLER2DRECT},
-   {"sampler2DRectShadow", SLANG_SPEC_SAMPLER2DRECTSHADOW},
-   {NULL, SLANG_SPEC_VOID}
-};
-
-slang_type_specifier_type
-slang_type_specifier_type_from_string(const char *name)
-{
-   const type_specifier_type_name *p = type_specifier_type_names;
-   while (p->name != NULL) {
-      if (slang_string_compare(p->name, name) == 0)
-         break;
-      p++;
-   }
-   return p->type;
-}
-
-const char *
-slang_type_specifier_type_to_string(slang_type_specifier_type type)
-{
-   const type_specifier_type_name *p = type_specifier_type_names;
-   while (p->name != NULL) {
-      if (p->type == type)
-         break;
-      p++;
-   }
-   return p->name;
-}
-
-/* slang_fully_specified_type */
-
-int
-slang_fully_specified_type_construct(slang_fully_specified_type * type)
-{
-   type->qualifier = SLANG_QUAL_NONE;
-   slang_type_specifier_ctr(&type->specifier);
-   return 1;
-}
-
-void
-slang_fully_specified_type_destruct(slang_fully_specified_type * type)
-{
-   slang_type_specifier_dtr(&type->specifier);
-}
-
-int
-slang_fully_specified_type_copy(slang_fully_specified_type * x,
-                                const slang_fully_specified_type * y)
-{
-   slang_fully_specified_type z;
-
-   if (!slang_fully_specified_type_construct(&z))
-      return 0;
-   z.qualifier = y->qualifier;
-   if (!slang_type_specifier_copy(&z.specifier, &y->specifier)) {
-      slang_fully_specified_type_destruct(&z);
-      return 0;
-   }
-   slang_fully_specified_type_destruct(x);
-   *x = z;
-   return 1;
-}
-
-
 static slang_variable *
 slang_variable_new(void)
 {
@@ -263,10 +164,9 @@ slang_variable_construct(slang_variable * var)
    var->a_name = SLANG_ATOM_NULL;
    var->array_len = 0;
    var->initializer = NULL;
-   var->address = ~0;
    var->size = 0;
    var->isTemp = GL_FALSE;
-   var->aux = NULL;
+   var->store = NULL;
    var->declared = 0;
    return 1;
 }
@@ -318,7 +218,6 @@ slang_variable_copy(slang_variable * x, const slang_variable * y)
          return 0;
       }
    }
-   z.address = y->address;
    z.size = y->size;
    slang_variable_destruct(x);
    *x = z;
@@ -331,7 +230,7 @@ slang_variable_copy(slang_variable * x, const slang_variable * y)
  * \param all  if true, search parent scopes too.
  */
 slang_variable *
-_slang_locate_variable(const slang_variable_scope * scope,
+_slang_variable_locate(const slang_variable_scope * scope,
                        const slang_atom a_name, GLboolean all)
 {
    while (scope) {
diff --git a/src/mesa/shader/slang/slang_compile_variable.h b/src/mesa/shader/slang/slang_compile_variable.h
index c6db16fcfa..a667d91b47 100644
--- a/src/mesa/shader/slang/slang_compile_variable.h
+++ b/src/mesa/shader/slang/slang_compile_variable.h
@@ -25,73 +25,8 @@
 #ifndef SLANG_COMPILE_VARIABLE_H
 #define SLANG_COMPILE_VARIABLE_H
 
-#if defined __cplusplus
-extern "C" {
-#endif
 
-
-typedef enum slang_type_variant_
-{
-   SLANG_VARIANT,    /* the default */
-   SLANG_INVARIANT   /* indicates the "invariant" keyword */
-} slang_type_variant;
-
-
-typedef enum slang_type_centroid_
-{
-   SLANG_CENTER,    /* the default */
-   SLANG_CENTROID   /* indicates the "centroid" keyword */
-} slang_type_centroid;
-
-
-typedef enum slang_type_qualifier_
-{
-   SLANG_QUAL_NONE,
-   SLANG_QUAL_CONST,
-   SLANG_QUAL_ATTRIBUTE,
-   SLANG_QUAL_VARYING,
-   SLANG_QUAL_UNIFORM,
-   SLANG_QUAL_OUT,
-   SLANG_QUAL_INOUT,
-   SLANG_QUAL_FIXEDOUTPUT,      /* internal */
-   SLANG_QUAL_FIXEDINPUT        /* internal */
-} slang_type_qualifier;
-
-extern slang_type_specifier_type
-slang_type_specifier_type_from_string(const char *);
-
-extern const char *
-slang_type_specifier_type_to_string(slang_type_specifier_type);
-
-
-
-typedef enum slang_type_precision_
-{
-   SLANG_PREC_DEFAULT,
-   SLANG_PREC_LOW,
-   SLANG_PREC_MEDIUM,
-   SLANG_PREC_HIGH
-} slang_type_precision;
-
-
-typedef struct slang_fully_specified_type_
-{
-   slang_type_qualifier qualifier;
-   slang_type_specifier specifier;
-   slang_type_precision precision;
-   slang_type_variant variant;
-   slang_type_centroid centroid;
-} slang_fully_specified_type;
-
-extern int
-slang_fully_specified_type_construct(slang_fully_specified_type *);
-
-extern void
-slang_fully_specified_type_destruct(slang_fully_specified_type *);
-
-extern int
-slang_fully_specified_type_copy(slang_fully_specified_type *,
-				const slang_fully_specified_type *);
+struct slang_ir_storage_;
 
 
 /**
@@ -103,11 +38,10 @@ typedef struct slang_variable_
    slang_atom a_name;               /**< The variable's name (char *) */
    GLuint array_len;                /**< only if type == SLANG_SPEC_ARRAy */
    struct slang_operation_ *initializer; /**< Optional initializer code */
-   GLuint address;                  /**< Storage location */
    GLuint size;                     /**< Variable's size in bytes */
    GLboolean isTemp;                /**< a named temporary (__resultTmp) */
    GLboolean declared;              /**< for debug */
-   void *aux;                       /**< Used during code gen */
+   struct slang_ir_storage_ *store; /**< Storage for this var */
 } slang_variable;
 
 
@@ -148,12 +82,8 @@ extern int
 slang_variable_copy(slang_variable *, const slang_variable *);
 
 extern slang_variable *
-_slang_locate_variable(const slang_variable_scope *, const slang_atom a_name,
+_slang_variable_locate(const slang_variable_scope *, const slang_atom a_name,
                        GLboolean all);
 
 
-#ifdef __cplusplus
-}
-#endif
-
 #endif /* SLANG_COMPILE_VARIABLE_H */
diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c
index b67cea7617..2dd122c9a5 100644
--- a/src/mesa/shader/slang/slang_emit.c
+++ b/src/mesa/shader/slang/slang_emit.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
  *
  * Copyright (C) 2005-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2008 VMware, Inc.   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -60,6 +60,8 @@ typedef struct
    struct gl_program **Subroutines;
    GLuint NumSubroutines;
 
+   GLuint MaxInstructions;  /**< size of prog->Instructions[] buffer */
+
    /* code-gen options */
    GLboolean EmitHighLevelInstructions;
    GLboolean EmitCondCodes;
@@ -153,6 +155,28 @@ _slang_swizzle_swizzle(GLuint swz1, GLuint swz2)
 
 
 /**
+ * Return the default swizzle mask for accessing a variable of the
+ * given size (in floats).  If size = 1, comp is used to identify
+ * which component [0..3] of the register holds the variable.
+ */
+GLuint
+_slang_var_swizzle(GLint size, GLint comp)
+{
+   switch (size) {
+   case 1:
+      return MAKE_SWIZZLE4(comp, comp, comp, comp);
+   case 2:
+      return MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_NIL, SWIZZLE_NIL);
+   case 3:
+      return MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_NIL);
+   default:
+      return SWIZZLE_XYZW;
+   }
+}
+
+
+
+/**
  * Allocate storage for the given node (if it hasn't already been allocated).
  *
  * Typically this is temporary storage for an intermediate result (such as
@@ -259,13 +283,16 @@ fix_swizzle(GLuint swizzle)
 static void
 storage_to_dst_reg(struct prog_dst_register *dst, const slang_ir_storage *st)
 {
+   const GLboolean relAddr = st->RelAddr;
    const GLint size = st->Size;
    GLint index = st->Index;
    GLuint swizzle = st->Swizzle;
 
+   assert(index >= 0);
    /* if this is storage relative to some parent storage, walk up the tree */
    while (st->Parent) {
       st = st->Parent;
+      assert(st->Index >= 0);
       index += st->Index;
       swizzle = _slang_swizzle_swizzle(st->Swizzle, swizzle);
    }
@@ -283,25 +310,25 @@ storage_to_dst_reg(struct prog_dst_register *dst, const slang_ir_storage *st)
       dst->WriteMask = swizzle_to_writemask(swizzle);
    }
    else {
-      GLuint writemask;
       switch (size) {
       case 1:
-         writemask = WRITEMASK_X << GET_SWZ(st->Swizzle, 0);
+         dst->WriteMask = WRITEMASK_X << GET_SWZ(st->Swizzle, 0);
          break;
       case 2:
-         writemask = WRITEMASK_XY;
+         dst->WriteMask = WRITEMASK_XY;
          break;
       case 3:
-         writemask = WRITEMASK_XYZ;
+         dst->WriteMask = WRITEMASK_XYZ;
          break;
       case 4:
-         writemask = WRITEMASK_XYZW;
+         dst->WriteMask = WRITEMASK_XYZW;
          break;
       default:
          ; /* error would have been caught above */
       }
-      dst->WriteMask = writemask;
    }
+
+   dst->RelAddr = relAddr;
 }
 
 
@@ -316,8 +343,14 @@ storage_to_src_reg(struct prog_src_register *src, const slang_ir_storage *st)
    GLuint swizzle = st->Swizzle;
 
    /* if this is storage relative to some parent storage, walk up the tree */
+   assert(index >= 0);
    while (st->Parent) {
       st = st->Parent;
+      if (st->Index < 0) {
+         /* an error should have been reported already */
+         return;
+      }
+      assert(st->Index >= 0);
       index += st->Index;
       swizzle = _slang_swizzle_swizzle(fix_swizzle(st->Swizzle), swizzle);
    }
@@ -387,9 +420,17 @@ new_instruction(slang_emit_info *emitInfo, gl_inst_opcode opcode)
       _mesa_print_instruction(prog->Instructions + prog->NumInstructions - 1);
    }
 #endif
-   prog->Instructions = _mesa_realloc_instructions(prog->Instructions,
-                                                   prog->NumInstructions,
-                                                   prog->NumInstructions + 1);
+   assert(prog->NumInstructions <= emitInfo->MaxInstructions);
+
+   if (prog->NumInstructions == emitInfo->MaxInstructions) {
+      /* grow the instruction buffer */
+      emitInfo->MaxInstructions += 20;
+      prog->Instructions =
+         _mesa_realloc_instructions(prog->Instructions,
+                                    prog->NumInstructions,
+                                    emitInfo->MaxInstructions);
+   }
+
    inst = prog->Instructions + prog->NumInstructions;
    prog->NumInstructions++;
    _mesa_init_instructions(inst, 1);
@@ -403,60 +444,149 @@ new_instruction(slang_emit_info *emitInfo, gl_inst_opcode opcode)
 }
 
 
+static struct prog_instruction *
+emit_arl_load(slang_emit_info *emitInfo,
+              enum register_file file, GLint index, GLuint swizzle)
+{
+   struct prog_instruction *inst = new_instruction(emitInfo, OPCODE_ARL);
+   inst->SrcReg[0].File = file;
+   inst->SrcReg[0].Index = index;
+   inst->SrcReg[0].Swizzle = swizzle;
+   inst->DstReg.File = PROGRAM_ADDRESS;
+   inst->DstReg.Index = 0;
+   inst->DstReg.WriteMask = WRITEMASK_X;
+   return inst;
+}
+
+
 /**
  * Emit a new instruction with given opcode, operands.
+ * At this point the instruction may have multiple indirect register
+ * loads/stores.  We convert those into ARL loads and address-relative
+ * operands.  See comments inside.
+ * At some point in the future we could directly emit indirectly addressed
+ * registers in Mesa GPU instructions.
  */
 static struct prog_instruction *
 emit_instruction(slang_emit_info *emitInfo,
                  gl_inst_opcode opcode,
                  const slang_ir_storage *dst,
+                 const slang_ir_storage *src0,
                  const slang_ir_storage *src1,
-                 const slang_ir_storage *src2,
-                 const slang_ir_storage *src3)
+                 const slang_ir_storage *src2)
 {
-   struct gl_program *prog = emitInfo->prog;
    struct prog_instruction *inst;
+   GLuint numIndirect = 0;
+   const slang_ir_storage *src[3];
+   slang_ir_storage newSrc[3], newDst;
+   GLuint i;
+   GLboolean isTemp[3];
+
+   isTemp[0] = isTemp[1] = isTemp[2] = GL_FALSE;
+
+   src[0] = src0;
+   src[1] = src1;
+   src[2] = src2;
+
+   /* count up how many operands are indirect loads */
+   for (i = 0; i < 3; i++) {
+      if (src[i] && src[i]->IsIndirect)
+         numIndirect++;
+   }
+   if (dst && dst->IsIndirect)
+      numIndirect++;
+
+   /* Take special steps for indirect register loads.
+    * If we had multiple address registers this would be simpler.
+    * For example, this GLSL code:
+    *    x[i] = y[j] + z[k];
+    * would translate into something like:
+    *    ARL ADDR.x, i;
+    *    ARL ADDR.y, j;
+    *    ARL ADDR.z, k;
+    *    ADD TEMP[ADDR.x+5], TEMP[ADDR.y+9], TEMP[ADDR.z+4];
+    * But since we currently only have one address register we have to do this:
+    *    ARL ADDR.x, i;
+    *    MOV t1, TEMP[ADDR.x+9];
+    *    ARL ADDR.x, j;
+    *    MOV t2, TEMP[ADDR.x+4];
+    *    ARL ADDR.x, k;
+    *    ADD TEMP[ADDR.x+5], t1, t2;
+    * The code here figures this out...
+    */
+   if (numIndirect > 0) {
+      for (i = 0; i < 3; i++) {
+         if (src[i] && src[i]->IsIndirect) {
+            /* load the ARL register with the indirect register */
+            emit_arl_load(emitInfo,
+                          src[i]->IndirectFile,
+                          src[i]->IndirectIndex,
+                          src[i]->IndirectSwizzle);
+
+            if (numIndirect > 1) {
+               /* Need to load src[i] into a temporary register */
+               slang_ir_storage srcRelAddr;
+               alloc_local_temp(emitInfo, &newSrc[i], src[i]->Size);
+               isTemp[i] = GL_TRUE;
+
+               /* set RelAddr flag on src register */
+               srcRelAddr = *src[i];
+               srcRelAddr.RelAddr = GL_TRUE;
+               srcRelAddr.IsIndirect = GL_FALSE; /* not really needed */
+
+               /* MOV newSrc, srcRelAddr; */
+               inst = emit_instruction(emitInfo,
+                                       OPCODE_MOV,
+                                       &newSrc[i],
+                                       &srcRelAddr,
+                                       NULL,
+                                       NULL);
+
+               src[i] = &newSrc[i];
+            }
+            else {
+               /* just rewrite the src[i] storage to be ARL-relative */
+               newSrc[i] = *src[i];
+               newSrc[i].RelAddr = GL_TRUE;
+               newSrc[i].IsIndirect = GL_FALSE; /* not really needed */
+               src[i] = &newSrc[i];
+            }
+         }
+      }
+   }
 
-   prog->Instructions = _mesa_realloc_instructions(prog->Instructions,
-                                                   prog->NumInstructions,
-                                                   prog->NumInstructions + 1);
-   inst = prog->Instructions + prog->NumInstructions;
-   prog->NumInstructions++;
+   /* Take special steps for indirect dest register write */
+   if (dst && dst->IsIndirect) {
+      /* load the ARL register with the indirect register */
+      emit_arl_load(emitInfo,
+                    dst->IndirectFile,
+                    dst->IndirectIndex,
+                    dst->IndirectSwizzle);
+      newDst = *dst;
+      newDst.RelAddr = GL_TRUE;
+      newDst.IsIndirect = GL_FALSE;
+      dst = &newDst;
+   }
 
-   _mesa_init_instructions(inst, 1);
-   inst->Opcode = opcode;
-   inst->BranchTarget = -1; /* invalid */
+   /* OK, emit the instruction and its dst, src regs */
+   inst = new_instruction(emitInfo, opcode);
+   if (!inst)
+      return NULL;
 
    if (dst)
       storage_to_dst_reg(&inst->DstReg, dst);
 
-   if (src1)
-      storage_to_src_reg(&inst->SrcReg[0], src1);
-   if (src2)
-      storage_to_src_reg(&inst->SrcReg[1], src2);
-   if (src3)
-      storage_to_src_reg(&inst->SrcReg[2], src3);   
-
-   return inst;
-}
-
+   for (i = 0; i < 3; i++) {
+      if (src[i])
+         storage_to_src_reg(&inst->SrcReg[i], src[i]);
+   }
 
-/**
- * Emit an ARL instruction.
- */
-static struct prog_instruction *
-emit_arl_instruction(slang_emit_info *emitInfo,
-                     GLint addrReg,
-                     const slang_ir_storage *src)
-{
-   struct prog_instruction *inst;
+   /* Free any temp registers that we allocated above */
+   for (i = 0; i < 3; i++) {
+      if (isTemp[i])
+         _slang_free_temp(emitInfo->vt, &newSrc[i]);
+   }
 
-   assert(addrReg == 0); /* only one addr reg at this time */
-   inst = new_instruction(emitInfo, OPCODE_ARL);
-   storage_to_src_reg(&inst->SrcReg[0], src);
-   inst->DstReg.File = PROGRAM_ADDRESS;
-   inst->DstReg.Index = addrReg;
-   inst->DstReg.WriteMask = WRITEMASK_X;
    return inst;
 }
 
@@ -1034,13 +1164,17 @@ emit_fcall(slang_emit_info *emitInfo, slang_ir_node *n)
    struct gl_program *progSave;
    struct prog_instruction *inst;
    GLuint subroutineId;
+   GLuint maxInstSave;
 
    assert(n->Opcode == IR_CALL);
    assert(n->Label);
 
    /* save/push cur program */
+   maxInstSave = emitInfo->MaxInstructions;
    progSave = emitInfo->prog;
+
    emitInfo->prog = new_subroutine(emitInfo, &subroutineId);
+   emitInfo->MaxInstructions = emitInfo->prog->NumInstructions;
 
    _slang_label_set_location(n->Label, emitInfo->prog->NumInstructions,
                              emitInfo->prog);
@@ -1072,6 +1206,7 @@ emit_fcall(slang_emit_info *emitInfo, slang_ir_node *n)
 
    /* pop/restore cur program */
    emitInfo->prog = progSave;
+   emitInfo->MaxInstructions = maxInstSave;
 
    /* emit the function call */
    inst = new_instruction(emitInfo, OPCODE_CAL);
@@ -1136,6 +1271,20 @@ emit_tex(slang_emit_info *emitInfo, slang_ir_node *n)
       opcode = OPCODE_TXP;
    }
 
+   if (n->Children[0]->Opcode == IR_ELEMENT) {
+      /* array is the sampler (a uniform which'll indicate the texture unit) */
+      assert(n->Children[0]->Children[0]->Store);
+      assert(n->Children[0]->Children[0]->Store->File == PROGRAM_SAMPLER);
+
+      emit(emitInfo, n->Children[0]);
+
+      n->Children[0]->Var = n->Children[0]->Children[0]->Var;
+   } else {
+      /* this is the sampler (a uniform which'll indicate the texture unit) */
+      assert(n->Children[0]->Store);
+      assert(n->Children[0]->Store->File == PROGRAM_SAMPLER);
+   }
+
    /* emit code for the texcoord operand */
    (void) emit(emitInfo, n->Children[1]);
 
@@ -1151,16 +1300,14 @@ emit_tex(slang_emit_info *emitInfo, slang_ir_node *n)
                            NULL,
                            NULL);
 
-   /* Child[0] is the sampler (a uniform which'll indicate the texture unit) */
-   assert(n->Children[0]->Store);
-   /* Store->Index is the sampler index */
+   /* Store->Index is the uniform/sampler index */
    assert(n->Children[0]->Store->Index >= 0);
-   /* Store->Size is the texture target */
-   assert(n->Children[0]->Store->Size >= TEXTURE_1D_INDEX);
-   assert(n->Children[0]->Store->Size <= TEXTURE_RECT_INDEX);
+   inst->TexSrcUnit = n->Children[0]->Store->Index;
+   inst->TexSrcTarget = n->Children[0]->Store->TexTarget;
 
-   inst->TexSrcTarget = n->Children[0]->Store->Size;
-   inst->TexSrcUnit = n->Children[0]->Store->Index; /* i.e. uniform's index */
+   /* mark the sampler as being used */
+   _mesa_use_uniform(emitInfo->prog->Parameters,
+                     (char *) n->Children[0]->Var->a_name);
 
    return inst;
 }
@@ -1202,9 +1349,10 @@ emit_copy(slang_emit_info *emitInfo, slang_ir_node *n)
 
    if (n->Store->File == PROGRAM_SAMPLER) {
       /* no code generated for sampler assignments,
-       * just copy the sampler index at compile time.
+       * just copy the sampler index/target at compile time.
        */
       n->Store->Index = n->Children[1]->Store->Index;
+      n->Store->TexTarget = n->Children[1]->Store->TexTarget;
       return NULL;
    }
 
@@ -1212,7 +1360,9 @@ emit_copy(slang_emit_info *emitInfo, slang_ir_node *n)
    if (inst &&
        _slang_is_temp(emitInfo->vt, n->Children[1]->Store) &&
        (inst->DstReg.File == n->Children[1]->Store->File) &&
-       (inst->DstReg.Index == n->Children[1]->Store->Index)) {
+       (inst->DstReg.Index == n->Children[1]->Store->Index) &&
+       !n->Children[0]->Store->IsIndirect &&
+       n->Children[0]->Store->Size <= 4) {
       /* Peephole optimization:
        * The Right-Hand-Side has its results in a temporary place.
        * Modify the RHS (and the prev instruction) to store its results
@@ -1670,153 +1820,149 @@ emit_swizzle(slang_emit_info *emitInfo, slang_ir_node *n)
 
    inst = emit(emitInfo, n->Children[0]);
 
-   /* setup storage info, if needed */
-   if (!n->Store->Parent)
-      n->Store->Parent = n->Children[0]->Store;
-
+#if 0
    assert(n->Store->Parent);
-
+   /* Apply this node's swizzle to parent's storage */
+   GLuint swizzle = n->Store->Swizzle;
+   _slang_copy_ir_storage(n->Store, n->Store->Parent);
+   n->Store->Swizzle = _slang_swizzle_swizzle(n->Store->Swizzle, swizzle);
+   assert(!n->Store->Parent);
+#endif
    return inst;
 }
 
 
 /**
- * Move a block registers from src to dst (or move a single register).
- * \param size  size of block, in floats (<=4 means one register)
+ * Dereference array element:  element == array[index]
+ * This basically involves emitting code for computing the array index
+ * and updating the node/element's storage info.
  */
 static struct prog_instruction *
-move_block(slang_emit_info *emitInfo,
-           GLuint size, GLboolean relAddr,
-           const slang_ir_storage *dst,
-           const slang_ir_storage *src)
+emit_array_element(slang_emit_info *emitInfo, slang_ir_node *n)
 {
+   slang_ir_storage *arrayStore, *indexStore;
+   const int elemSize = n->Store->Size;           /* number of floats */
+   const GLint elemSizeVec = (elemSize + 3) / 4;  /* number of vec4 */
    struct prog_instruction *inst;
 
-   if (size > 4) {
-      /* move matrix/struct etc (block of registers) */
-      slang_ir_storage dstStore = *dst;
-      slang_ir_storage srcStore = *src;
-
-      dstStore.Size = 4;
-      srcStore.Size = 4;
-      while (size >= 4) {
-         inst = emit_instruction(emitInfo, OPCODE_MOV,
-                                 &dstStore,
-                                 &srcStore,
-                                 NULL,
-                                 NULL);
-         inst->SrcReg[0].RelAddr = relAddr;
-         inst_comment(inst, "IR_COPY block");
-         srcStore.Index++;
-         dstStore.Index++;
-         size -= 4;
-      }
-   }
-   else {
-      /* single register move */
-      inst = emit_instruction(emitInfo,
-                              OPCODE_MOV,
-                              dst,
-                              src,
-                              NULL,
-                              NULL);
-      inst->SrcReg[0].RelAddr = relAddr;
-   }
-   return inst;
-}
-
-
-
-/**
- * Dereference array element.  Just resolve storage for the array
- * element represented by this node.
- * This is typically where Indirect addressing comes into play.
- * See comments on struct slang_ir_storage.
- */
-static struct prog_instruction *
-emit_array_element(slang_emit_info *emitInfo, slang_ir_node *n)
-{
    assert(n->Opcode == IR_ELEMENT);
-   assert(n->Store);
-   assert(n->Store->File == PROGRAM_UNDEFINED);
-   assert(n->Store->Parent);
-   assert(n->Store->Size > 0);
+   assert(elemSize > 0);
 
+   /* special case for built-in state variables, like light state */
    {
       slang_ir_storage *root = n->Store;
+      assert(!root->Parent);
       while (root->Parent)
          root = root->Parent;
 
       if (root->File == PROGRAM_STATE_VAR) {
-         GLint index = _slang_alloc_statevar(n, emitInfo->prog->Parameters);
-         assert(n->Store->Index == index);
-         return NULL;
+         GLboolean direct;
+         GLint index =
+            _slang_alloc_statevar(n, emitInfo->prog->Parameters, &direct);
+         if (index < 0) {
+            /* error */
+            return NULL;
+         }
+         if (direct) {
+            n->Store->Index = index;
+            return NULL; /* all done */
+         }
       }
    }
 
-   /* do codegen for array */
+   /* do codegen for array itself */
    emit(emitInfo, n->Children[0]);
+   arrayStore = n->Children[0]->Store;
+
+   /* The initial array element storage is the array's storage,
+    * then modified below.
+    */
+   _slang_copy_ir_storage(n->Store, arrayStore);
+
 
    if (n->Children[1]->Opcode == IR_FLOAT) {
-      /* Constant array index.
-       * Set Store's index to be the offset of the array element in
-       * the register file.
-       */
+      /* Constant array index */
       const GLint element = (GLint) n->Children[1]->Value[0];
-      const GLint sz = (n->Store->Size + 3) / 4; /* size in slots/registers */
 
-      n->Store->Index = sz * element;
-      assert(n->Store->Parent);
+      /* this element's storage is the array's storage, plus constant offset */
+      n->Store->Index += elemSizeVec * element;
    }
    else {
       /* Variable array index */
-      struct prog_instruction *inst;
 
       /* do codegen for array index expression */
       emit(emitInfo, n->Children[1]);
+      indexStore = n->Children[1]->Store;
+
+      if (indexStore->IsIndirect) {
+         /* need to put the array index into a temporary since we can't
+          * directly support a[b[i]] constructs.
+          */
+
 
-      /* allocate temp storage for the array element */
-      assert(n->Store->Index < 0);
-      n->Store->File = PROGRAM_TEMPORARY;
-      n->Store->Parent = NULL;
-      alloc_node_storage(emitInfo, n, -1);
+         /*indexStore = tempstore();*/
+      }
 
-      if (n->Store->Size > 4) {
-         /* need to multiply the index by the element size */
-         const GLint elemSize = (n->Store->Size + 3) / 4;
-         slang_ir_storage indexTemp, elemSizeStore;
 
-         /* constant containing the element size */
-         constant_to_storage(emitInfo, (float) elemSize, &elemSizeStore);
+      if (elemSize > 4) {
+         /* need to multiply array index by array element size */
+         struct prog_instruction *inst;
+         slang_ir_storage *indexTemp;
+         slang_ir_storage elemSizeStore;
 
          /* allocate 1 float indexTemp */
-         alloc_local_temp(emitInfo, &indexTemp, 1);
+         indexTemp = _slang_new_ir_storage(PROGRAM_TEMPORARY, -1, 1);
+         _slang_alloc_temp(emitInfo->vt, indexTemp);
 
-         /* MUL temp, index, elemSize */
-         inst = emit_instruction(emitInfo, OPCODE_MUL,
-                                 &indexTemp, /* dest */
-                                 n->Children[1]->Store, /* the index */
+         /* allocate a constant containing the element size */
+         constant_to_storage(emitInfo, (float) elemSizeVec, &elemSizeStore);
+
+         /* multiply array index by element size */
+         inst = emit_instruction(emitInfo,
+                                 OPCODE_MUL,
+                                 indexTemp, /* dest */
+                                 indexStore, /* the index */
                                  &elemSizeStore,
                                  NULL);
 
-         /* load ADDR[0].X = temp */
-         inst = emit_arl_instruction(emitInfo, 0, &indexTemp);
-
-         _slang_free_temp(emitInfo->vt, &indexTemp);
+         indexStore = indexTemp;
       }
-      else {
-         /* simply load address reg w/ array index */
-         inst = emit_arl_instruction(emitInfo, 0, n->Children[1]->Store);
+
+      if (arrayStore->IsIndirect) {
+         /* ex: in a[i][j], a[i] (the arrayStore) is indirect */
+         /* Need to add indexStore to arrayStore->Indirect store */
+         slang_ir_storage indirectArray;
+         slang_ir_storage *indexTemp;
+
+         _slang_init_ir_storage(&indirectArray,
+                                arrayStore->IndirectFile,
+                                arrayStore->IndirectIndex,
+                                1,
+                                arrayStore->IndirectSwizzle);
+
+         /* allocate 1 float indexTemp */
+         indexTemp = _slang_new_ir_storage(PROGRAM_TEMPORARY, -1, 1);
+         _slang_alloc_temp(emitInfo->vt, indexTemp);
+
+         inst = emit_instruction(emitInfo,
+                                 OPCODE_ADD,
+                                 indexTemp,      /* dest */
+                                 indexStore,     /* the index */
+                                 &indirectArray, /* indirect array base */
+                                 NULL);
+
+         indexStore = indexTemp;
       }
 
-      /* copy from array element to temp storage */
-      move_block(emitInfo, n->Store->Size, GL_TRUE,
-                 n->Store, n->Children[0]->Store);
+      /* update the array element storage info */
+      n->Store->IsIndirect = GL_TRUE;
+      n->Store->IndirectFile = indexStore->File;
+      n->Store->IndirectIndex = indexStore->Index;
+      n->Store->IndirectSwizzle = indexStore->Swizzle;
    }
 
-   /* if array element size is one, make sure we only access X */
-   if (n->Store->Size == 1)
-      n->Store->Swizzle = SWIZZLE_XXXX;
+   n->Store->Size = elemSize;
+   n->Store->Swizzle = _slang_var_swizzle(elemSize, 0);
 
    return NULL; /* no instruction */
 }
@@ -1829,9 +1975,11 @@ static struct prog_instruction *
 emit_struct_field(slang_emit_info *emitInfo, slang_ir_node *n)
 {
    slang_ir_storage *root = n->Store;
+   GLint fieldOffset, fieldSize;
 
    assert(n->Opcode == IR_FIELD);
 
+   assert(!root->Parent);
    while (root->Parent)
       root = root->Parent;
 
@@ -1842,17 +1990,52 @@ emit_struct_field(slang_emit_info *emitInfo, slang_ir_node *n)
     * space for the ones that we actually use!
     */
    if (root->File == PROGRAM_STATE_VAR) {
-      root->Index = _slang_alloc_statevar(n, emitInfo->prog->Parameters);
-      if (root->Index < 0) {
+      GLboolean direct;
+      GLint index = _slang_alloc_statevar(n, emitInfo->prog->Parameters, &direct);
+      if (index < 0) {
          slang_info_log_error(emitInfo->log, "Error parsing state variable");
          return NULL;
       }
+      if (direct) {
+         root->Index = index;
+         return NULL; /* all done */
+      }
    }
-   else {
-      /* do codegen for struct */
-      emit(emitInfo, n->Children[0]);
+
+   /* do codegen for struct */
+   emit(emitInfo, n->Children[0]);
+   assert(n->Children[0]->Store->Index >= 0);
+
+
+   fieldOffset = n->Store->Index;
+   fieldSize = n->Store->Size;
+
+   _slang_copy_ir_storage(n->Store, n->Children[0]->Store);
+
+   n->Store->Index = n->Children[0]->Store->Index + fieldOffset / 4;
+   n->Store->Size = fieldSize;
+
+   switch (fieldSize) {
+   case 1:
+      {
+         GLint swz = fieldOffset % 4;
+         n->Store->Swizzle = MAKE_SWIZZLE4(swz, swz, swz, swz);
+      }
+      break;
+   case 2:
+      n->Store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y,
+                                        SWIZZLE_NIL, SWIZZLE_NIL);
+      break;
+   case 3:
+      n->Store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y,
+                                        SWIZZLE_Z, SWIZZLE_NIL);
+      break;
+   default:
+      n->Store->Swizzle = SWIZZLE_XYZW;
    }
 
+   assert(n->Store->Index >= 0);
+
    return NULL; /* no instruction */
 }
 
@@ -1891,7 +2074,7 @@ emit_var_decl(slang_emit_info *emitInfo, slang_ir_node *n)
         printf("IR_VAR_DECL %s %d store %p\n",
         (char*) n->Var->a_name, n->Store->Index, (void*) n->Store);
       */
-      assert(n->Var->aux == n->Store);
+      assert(n->Var->store == n->Store);
    }
    if (emitInfo->EmitComments) {
       /* emit NOP with comment describing the variable's storage location */
@@ -1909,7 +2092,7 @@ emit_var_decl(slang_emit_info *emitInfo, slang_ir_node *n)
 
 /**
  * Emit code for a reference to a variable.
- * Actually, no code is generated but we may do some memory alloation.
+ * Actually, no code is generated but we may do some memory allocation.
  * In particular, state vars (uniforms) are allocated on an as-needed basis.
  */
 static struct prog_instruction *
@@ -1919,12 +2102,28 @@ emit_var_ref(slang_emit_info *emitInfo, slang_ir_node *n)
    assert(n->Store->File != PROGRAM_UNDEFINED);
 
    if (n->Store->File == PROGRAM_STATE_VAR && n->Store->Index < 0) {
-      n->Store->Index = _slang_alloc_statevar(n, emitInfo->prog->Parameters);
+      GLboolean direct;
+      GLint index = _slang_alloc_statevar(n, emitInfo->prog->Parameters, &direct);
+      if (index < 0) {
+         /* error */
+         char s[100];
+         _mesa_snprintf(s, sizeof(s), "Undefined variable '%s'",
+                        (char *) n->Var->a_name);
+         slang_info_log_error(emitInfo->log, s);
+         return NULL;
+      }
+
+      n->Store->Index = index;
    }
-   else if (n->Store->File == PROGRAM_UNIFORM) {
+   else if (n->Store->File == PROGRAM_UNIFORM ||
+            n->Store->File == PROGRAM_SAMPLER) {
       /* mark var as used */
       _mesa_use_uniform(emitInfo->prog->Parameters, (char *) n->Var->a_name);
    }
+   else if (n->Store->File == PROGRAM_INPUT) {
+      assert(n->Store->Index >= 0);
+      emitInfo->prog->InputsRead |= (1 << n->Store->Index);
+   }
 
    if (n->Store->Index < 0) {
       /* probably ran out of registers */
@@ -2130,7 +2329,7 @@ _slang_resolve_subroutines(slang_emit_info *emitInfo)
       total += emitInfo->Subroutines[i]->NumInstructions;
    }
 
-   /* adjust BrancTargets within the functions */
+   /* adjust BranchTargets within the functions */
    for (i = 0; i < emitInfo->NumSubroutines; i++) {
       struct gl_program *sub = emitInfo->Subroutines[i];
       GLuint j;
@@ -2183,10 +2382,20 @@ _slang_resolve_subroutines(slang_emit_info *emitInfo)
 
 
 
-
+/**
+ * Convert the IR tree into GPU instructions.
+ * \param n  root of IR tree
+ * \param vt  variable table
+ * \param prog  program to put GPU instructions into
+ * \param pragmas  controls codegen options
+ * \param withEnd  if true, emit END opcode at end
+ * \param log  log for emitting errors/warnings/info
+ */
 GLboolean
 _slang_emit_code(slang_ir_node *n, slang_var_table *vt,
-                 struct gl_program *prog, GLboolean withEnd,
+                 struct gl_program *prog,
+                 const struct gl_sl_pragmas *pragmas,
+                 GLboolean withEnd,
                  slang_info_log *log)
 {
    GET_CURRENT_CONTEXT(ctx);
@@ -2199,10 +2408,11 @@ _slang_emit_code(slang_ir_node *n, slang_var_table *vt,
    emitInfo.prog = prog;
    emitInfo.Subroutines = NULL;
    emitInfo.NumSubroutines = 0;
+   emitInfo.MaxInstructions = prog->NumInstructions;
 
    emitInfo.EmitHighLevelInstructions = ctx->Shader.EmitHighLevelInstructions;
    emitInfo.EmitCondCodes = ctx->Shader.EmitCondCodes;
-   emitInfo.EmitComments = ctx->Shader.EmitComments;
+   emitInfo.EmitComments = ctx->Shader.EmitComments || pragmas->Debug;
    emitInfo.EmitBeginEndSub = GL_TRUE;
 
    if (!emitInfo.EmitCondCodes) {
diff --git a/src/mesa/shader/slang/slang_emit.h b/src/mesa/shader/slang/slang_emit.h
index 4db4bbe562..ab4c202d67 100644
--- a/src/mesa/shader/slang/slang_emit.h
+++ b/src/mesa/shader/slang/slang_emit.h
@@ -32,17 +32,19 @@
 #include "main/mtypes.h"
 
 
-extern void
-slang_print_ir(const slang_ir_node *n, int indent);
+extern GLuint
+_slang_swizzle_swizzle(GLuint swz1, GLuint swz2);
 
 
 extern GLuint
-_slang_swizzle_swizzle(GLuint swz1, GLuint swz2);
+_slang_var_swizzle(GLint size, GLint comp);
 
 
 extern GLboolean
 _slang_emit_code(slang_ir_node *n, slang_var_table *vartable,
-                 struct gl_program *prog, GLboolean withEnd,
+                 struct gl_program *prog,
+                 const struct gl_sl_pragmas *pragmas,
+                 GLboolean withEnd,
                  slang_info_log *log);
 
 
diff --git a/src/mesa/shader/slang/slang_ir.c b/src/mesa/shader/slang/slang_ir.c
index 4fe3a2b85d..e4c6e0ea51 100644
--- a/src/mesa/shader/slang/slang_ir.c
+++ b/src/mesa/shader/slang/slang_ir.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
  *
  * Copyright (C) 2005-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -116,6 +116,20 @@ _slang_ir_info(slang_ir_opcode opcode)
 }
 
 
+void
+_slang_init_ir_storage(slang_ir_storage *st,
+                       enum register_file file, GLint index, GLint size,
+                       GLuint swizzle)
+{
+   st->File = file;
+   st->Index = index;
+   st->Size = size;
+   st->Swizzle = swizzle;
+   st->Parent = NULL;
+   st->IsIndirect = GL_FALSE;
+}
+
+
 /**
  * Return a new slang_ir_storage object.
  */
@@ -130,6 +144,7 @@ _slang_new_ir_storage(enum register_file file, GLint index, GLint size)
       st->Size = size;
       st->Swizzle = SWIZZLE_NOOP;
       st->Parent = NULL;
+      st->IsIndirect = GL_FALSE;
    }
    return st;
 }
@@ -150,6 +165,7 @@ _slang_new_ir_storage_swz(enum register_file file, GLint index, GLint size,
       st->Size = size;
       st->Swizzle = swizzle;
       st->Parent = NULL;
+      st->IsIndirect = GL_FALSE;
    }
    return st;
 }
@@ -170,12 +186,66 @@ _slang_new_ir_storage_relative(GLint index, GLint size,
       st->Size = size;
       st->Swizzle = SWIZZLE_NOOP;
       st->Parent = parent;
+      st->IsIndirect = GL_FALSE;
+   }
+   return st;
+}
+
+
+slang_ir_storage *
+_slang_new_ir_storage_indirect(enum register_file file,
+                               GLint index,
+                               GLint size,
+                               enum register_file indirectFile,
+                               GLint indirectIndex,
+                               GLuint indirectSwizzle)
+{
+   slang_ir_storage *st;
+   st = (slang_ir_storage *) _slang_alloc(sizeof(slang_ir_storage));
+   if (st) {
+      st->File = file;
+      st->Index = index;
+      st->Size = size;
+      st->Swizzle = SWIZZLE_NOOP;
+      st->IsIndirect = GL_TRUE;
+      st->IndirectFile = indirectFile;
+      st->IndirectIndex = indirectIndex;
+      st->IndirectSwizzle = indirectSwizzle;
+   }
+   return st;
+}
+
+
+/**
+ * Allocate IR storage for a texture sampler.
+ * \param sampNum  the sampler number/index
+ * \param texTarget  one of TEXTURE_x_INDEX values
+ * \param size  number of samplers (in case of sampler array)
+ */
+slang_ir_storage *
+_slang_new_ir_storage_sampler(GLint sampNum, GLuint texTarget, GLint size)
+{
+   slang_ir_storage *st;
+   assert(texTarget < NUM_TEXTURE_TARGETS);
+   st = _slang_new_ir_storage(PROGRAM_SAMPLER, sampNum, size);
+   if (st) {
+      st->TexTarget = texTarget;
    }
    return st;
 }
 
 
 
+/* XXX temporary function */
+void
+_slang_copy_ir_storage(slang_ir_storage *dst, const slang_ir_storage *src)
+{
+   *dst = *src;
+   dst->Parent = NULL;
+}
+
+
+
 static const char *
 _slang_ir_name(slang_ir_opcode opcode)
 {
diff --git a/src/mesa/shader/slang/slang_ir.h b/src/mesa/shader/slang/slang_ir.h
index a4552ae144..644269d491 100644
--- a/src/mesa/shader/slang/slang_ir.h
+++ b/src/mesa/shader/slang/slang_ir.h
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.3
  *
- * Copyright (C) 2005-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2005-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -140,24 +140,54 @@ typedef enum
 
 
 /**
- * Describes where data storage is allocated.
+ * Describes where data/variables are stored in the various register files.
+ *
+ * In the simple case, the File, Index and Size fields indicate where
+ * a variable is stored.  For example, a vec3 variable may be stored
+ * as (File=PROGRAM_TEMPORARY, Index=6, Size=3).  Or, File[Index].
+ * Or, a program input like color may be stored as
+ * (File=PROGRAM_INPUT,Index=3,Size=4);
+ *
+ * For single-float values, the Swizzle field indicates which component
+ * of the vector contains the float.
+ *
+ * If IsIndirect is set, the storage is accessed through an indirect
+ * register lookup.  The value in question will be located at:
+ *   File[Index + IndirectFile[IndirectIndex]]
+ *
+ * This is primary used for indexing arrays.  For example, consider this
+ * GLSL code:
+ *   uniform int i;
+ *   float a[10];
+ *   float x = a[i];
+ *
+ * here, storage for a[i] would be described by (File=PROGRAM_TEMPORAY,
+ * Index=aPos, IndirectFile=PROGRAM_UNIFORM, IndirectIndex=iPos), which
+ * would mean TEMP[aPos + UNIFORM[iPos]]
  */
-struct _slang_ir_storage
+struct slang_ir_storage_
 {
    enum register_file File;  /**< PROGRAM_TEMPORARY, PROGRAM_INPUT, etc */
-   GLint Index;  /**< -1 means unallocated */
-   GLint Size;  /**< number of floats */
-   GLuint Swizzle;
+   GLint Index;    /**< -1 means unallocated */
+   GLint Size;     /**< number of floats or ints */
+   GLuint Swizzle; /**< Swizzle AND writemask info */
    GLint RefCount; /**< Used during IR tree delete */
-   GLboolean RelAddr;
+
+   GLboolean RelAddr; /* we'll remove this eventually */
+
+   GLboolean IsIndirect;
+   enum register_file IndirectFile;
+   GLint IndirectIndex;
+   GLuint IndirectSwizzle;
+   GLuint TexTarget;  /**< If File==PROGRAM_SAMPLER, one of TEXTURE_x_INDEX */
 
    /** If Parent is non-null, Index is relative to parent.
     * The other fields are ignored.
     */
-   struct _slang_ir_storage *Parent;
+   struct slang_ir_storage_ *Parent;
 };
 
-typedef struct _slang_ir_storage slang_ir_storage;
+typedef struct slang_ir_storage_ slang_ir_storage;
 
 
 /**
@@ -199,6 +229,11 @@ extern const slang_ir_info *
 _slang_ir_info(slang_ir_opcode opcode);
 
 
+extern void
+_slang_init_ir_storage(slang_ir_storage *st,
+                       enum register_file file, GLint index, GLint size,
+                       GLuint swizzle);
+
 extern slang_ir_storage *
 _slang_new_ir_storage(enum register_file file, GLint index, GLint size);
 
@@ -212,6 +247,21 @@ _slang_new_ir_storage_relative(GLint index, GLint size,
                                slang_ir_storage *parent);
 
 
+extern slang_ir_storage *
+_slang_new_ir_storage_indirect(enum register_file file,
+                               GLint index,
+                               GLint size,
+                               enum register_file indirectFile,
+                               GLint indirectIndex,
+                               GLuint indirectSwizzle);
+
+extern slang_ir_storage *
+_slang_new_ir_storage_sampler(GLint sampNum, GLuint texTarget, GLint size);
+
+
+extern void
+_slang_copy_ir_storage(slang_ir_storage *dst, const slang_ir_storage *src);
+
 
 extern void
 _slang_free_ir_tree(slang_ir_node *n);
diff --git a/src/mesa/shader/slang/slang_library_noise.c b/src/mesa/shader/slang/slang_library_noise.c
deleted file mode 100644
index 25a05657cc..0000000000
--- a/src/mesa/shader/slang/slang_library_noise.c
+++ /dev/null
@@ -1,501 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5
- *
- * Copyright (C) 2006  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-/*
- * SimplexNoise1234
- * Copyright (c) 2003-2005, Stefan Gustavson
- *
- * Contact: stegu@itn.liu.se
- */
-
-/** \file
-    \brief C implementation of Perlin Simplex Noise over 1,2,3, and 4 dimensions.
-    \author Stefan Gustavson (stegu@itn.liu.se)
-*/
-
-/*
- * This implementation is "Simplex Noise" as presented by
- * Ken Perlin at a relatively obscure and not often cited course
- * session "Real-Time Shading" at Siggraph 2001 (before real
- * time shading actually took on), under the title "hardware noise".
- * The 3D function is numerically equivalent to his Java reference
- * code available in the PDF course notes, although I re-implemented
- * it from scratch to get more readable code. The 1D, 2D and 4D cases
- * were implemented from scratch by me from Ken Perlin's text.
- *
- * This file has no dependencies on any other file, not even its own
- * header file. The header file is made for use by external code only.
- */
-
-
-#include "main/imports.h"
-#include "slang_library_noise.h"
-
-#define FASTFLOOR(x) ( ((x)>0) ? ((int)x) : (((int)x)-1) )
-
-/*
- * ---------------------------------------------------------------------
- * Static data
- */
-
-/*
- * Permutation table. This is just a random jumble of all numbers 0-255,
- * repeated twice to avoid wrapping the index at 255 for each lookup.
- * This needs to be exactly the same for all instances on all platforms,
- * so it's easiest to just keep it as static explicit data.
- * This also removes the need for any initialisation of this class.
- *
- * Note that making this an int[] instead of a char[] might make the
- * code run faster on platforms with a high penalty for unaligned single
- * byte addressing. Intel x86 is generally single-byte-friendly, but
- * some other CPUs are faster with 4-aligned reads.
- * However, a char[] is smaller, which avoids cache trashing, and that
- * is probably the most important aspect on most architectures.
- * This array is accessed a *lot* by the noise functions.
- * A vector-valued noise over 3D accesses it 96 times, and a
- * float-valued 4D noise 64 times. We want this to fit in the cache!
- */
-unsigned char perm[512] = {151,160,137,91,90,15,
-  131,13,201,95,96,53,194,233,7,225,140,36,103,30,69,142,8,99,37,240,21,10,23,
-  190, 6,148,247,120,234,75,0,26,197,62,94,252,219,203,117,35,11,32,57,177,33,
-  88,237,149,56,87,174,20,125,136,171,168, 68,175,74,165,71,134,139,48,27,166,
-  77,146,158,231,83,111,229,122,60,211,133,230,220,105,92,41,55,46,245,40,244,
-  102,143,54, 65,25,63,161, 1,216,80,73,209,76,132,187,208, 89,18,169,200,196,
-  135,130,116,188,159,86,164,100,109,198,173,186, 3,64,52,217,226,250,124,123,
-  5,202,38,147,118,126,255,82,85,212,207,206,59,227,47,16,58,17,182,189,28,42,
-  223,183,170,213,119,248,152, 2,44,154,163, 70,221,153,101,155,167, 43,172,9,
-  129,22,39,253, 19,98,108,110,79,113,224,232,178,185, 112,104,218,246,97,228,
-  251,34,242,193,238,210,144,12,191,179,162,241, 81,51,145,235,249,14,239,107,
-  49,192,214, 31,181,199,106,157,184, 84,204,176,115,121,50,45,127, 4,150,254,
-  138,236,205,93,222,114,67,29,24,72,243,141,128,195,78,66,215,61,156,180,
-  151,160,137,91,90,15,
-  131,13,201,95,96,53,194,233,7,225,140,36,103,30,69,142,8,99,37,240,21,10,23,
-  190, 6,148,247,120,234,75,0,26,197,62,94,252,219,203,117,35,11,32,57,177,33,
-  88,237,149,56,87,174,20,125,136,171,168, 68,175,74,165,71,134,139,48,27,166,
-  77,146,158,231,83,111,229,122,60,211,133,230,220,105,92,41,55,46,245,40,244,
-  102,143,54, 65,25,63,161, 1,216,80,73,209,76,132,187,208, 89,18,169,200,196,
-  135,130,116,188,159,86,164,100,109,198,173,186, 3,64,52,217,226,250,124,123,
-  5,202,38,147,118,126,255,82,85,212,207,206,59,227,47,16,58,17,182,189,28,42,
-  223,183,170,213,119,248,152, 2,44,154,163, 70,221,153,101,155,167, 43,172,9,
-  129,22,39,253, 19,98,108,110,79,113,224,232,178,185, 112,104,218,246,97,228,
-  251,34,242,193,238,210,144,12,191,179,162,241, 81,51,145,235,249,14,239,107,
-  49,192,214, 31,181,199,106,157,184, 84,204,176,115,121,50,45,127, 4,150,254,
-  138,236,205,93,222,114,67,29,24,72,243,141,128,195,78,66,215,61,156,180 
-};
-
-/*
- * ---------------------------------------------------------------------
- */
-
-/*
- * Helper functions to compute gradients-dot-residualvectors (1D to 4D)
- * Note that these generate gradients of more than unit length. To make
- * a close match with the value range of classic Perlin noise, the final
- * noise values need to be rescaled to fit nicely within [-1,1].
- * (The simplex noise functions as such also have different scaling.)
- * Note also that these noise functions are the most practical and useful
- * signed version of Perlin noise. To return values according to the
- * RenderMan specification from the SL noise() and pnoise() functions,
- * the noise values need to be scaled and offset to [0,1], like this:
- * float SLnoise = (SimplexNoise1234::noise(x,y,z) + 1.0) * 0.5;
- */
-
-static float  grad1( int hash, float x ) {
-    int h = hash & 15;
-    float grad = 1.0f + (h & 7);   /* Gradient value 1.0, 2.0, ..., 8.0 */
-    if (h&8) grad = -grad;         /* Set a random sign for the gradient */
-    return ( grad * x );           /* Multiply the gradient with the distance */
-}
-
-static float  grad2( int hash, float x, float y ) {
-    int h = hash & 7;      /* Convert low 3 bits of hash code */
-    float u = h<4 ? x : y;  /* into 8 simple gradient directions, */
-    float v = h<4 ? y : x;  /* and compute the dot product with (x,y). */
-    return ((h&1)? -u : u) + ((h&2)? -2.0f*v : 2.0f*v);
-}
-
-static float  grad3( int hash, float x, float y , float z ) {
-    int h = hash & 15;     /* Convert low 4 bits of hash code into 12 simple */
-    float u = h<8 ? x : y; /* gradient directions, and compute dot product. */
-    float v = h<4 ? y : h==12||h==14 ? x : z; /* Fix repeats at h = 12 to 15 */
-    return ((h&1)? -u : u) + ((h&2)? -v : v);
-}
-
-static float  grad4( int hash, float x, float y, float z, float t ) {
-    int h = hash & 31;      /* Convert low 5 bits of hash code into 32 simple */
-    float u = h<24 ? x : y; /* gradient directions, and compute dot product. */
-    float v = h<16 ? y : z;
-    float w = h<8 ? z : t;
-    return ((h&1)? -u : u) + ((h&2)? -v : v) + ((h&4)? -w : w);
-}
-
-  /* A lookup table to traverse the simplex around a given point in 4D. */
-  /* Details can be found where this table is used, in the 4D noise method. */
-  /* TODO: This should not be required, backport it from Bill's GLSL code! */
-  static unsigned char simplex[64][4] = {
-    {0,1,2,3},{0,1,3,2},{0,0,0,0},{0,2,3,1},{0,0,0,0},{0,0,0,0},{0,0,0,0},{1,2,3,0},
-    {0,2,1,3},{0,0,0,0},{0,3,1,2},{0,3,2,1},{0,0,0,0},{0,0,0,0},{0,0,0,0},{1,3,2,0},
-    {0,0,0,0},{0,0,0,0},{0,0,0,0},{0,0,0,0},{0,0,0,0},{0,0,0,0},{0,0,0,0},{0,0,0,0},
-    {1,2,0,3},{0,0,0,0},{1,3,0,2},{0,0,0,0},{0,0,0,0},{0,0,0,0},{2,3,0,1},{2,3,1,0},
-    {1,0,2,3},{1,0,3,2},{0,0,0,0},{0,0,0,0},{0,0,0,0},{2,0,3,1},{0,0,0,0},{2,1,3,0},
-    {0,0,0,0},{0,0,0,0},{0,0,0,0},{0,0,0,0},{0,0,0,0},{0,0,0,0},{0,0,0,0},{0,0,0,0},
-    {2,0,1,3},{0,0,0,0},{0,0,0,0},{0,0,0,0},{3,0,1,2},{3,0,2,1},{0,0,0,0},{3,1,2,0},
-    {2,1,0,3},{0,0,0,0},{0,0,0,0},{0,0,0,0},{3,1,0,2},{0,0,0,0},{3,2,0,1},{3,2,1,0}};
-
-/* 1D simplex noise */
-GLfloat _slang_library_noise1 (GLfloat x)
-{
-  int i0 = FASTFLOOR(x);
-  int i1 = i0 + 1;
-  float x0 = x - i0;
-  float x1 = x0 - 1.0f;
-  float t1 = 1.0f - x1*x1;
-  float n0, n1;
-
-  float t0 = 1.0f - x0*x0;
-/*  if(t0 < 0.0f) t0 = 0.0f; // this never happens for the 1D case */
-  t0 *= t0;
-  n0 = t0 * t0 * grad1(perm[i0 & 0xff], x0);
-
-/*  if(t1 < 0.0f) t1 = 0.0f; // this never happens for the 1D case */
-  t1 *= t1;
-  n1 = t1 * t1 * grad1(perm[i1 & 0xff], x1);
-  /* The maximum value of this noise is 8*(3/4)^4 = 2.53125 */
-  /* A factor of 0.395 would scale to fit exactly within [-1,1], but */
-  /* we want to match PRMan's 1D noise, so we scale it down some more. */
-  return 0.25f * (n0 + n1);
-}
-
-/* 2D simplex noise */
-GLfloat _slang_library_noise2 (GLfloat x, GLfloat y)
-{
-#define F2 0.366025403f /* F2 = 0.5*(sqrt(3.0)-1.0) */
-#define G2 0.211324865f /* G2 = (3.0-Math.sqrt(3.0))/6.0 */
-
-    float n0, n1, n2; /* Noise contributions from the three corners */
-
-    /* Skew the input space to determine which simplex cell we're in */
-    float s = (x+y)*F2; /* Hairy factor for 2D */
-    float xs = x + s;
-    float ys = y + s;
-    int i = FASTFLOOR(xs);
-    int j = FASTFLOOR(ys);
-
-    float t = (float)(i+j)*G2;
-    float X0 = i-t; /* Unskew the cell origin back to (x,y) space */
-    float Y0 = j-t;
-    float x0 = x-X0; /* The x,y distances from the cell origin */
-    float y0 = y-Y0;
-
-    float x1, y1, x2, y2;
-    int ii, jj;
-    float t0, t1, t2;
-
-    /* For the 2D case, the simplex shape is an equilateral triangle. */
-    /* Determine which simplex we are in. */
-    int i1, j1; /* Offsets for second (middle) corner of simplex in (i,j) coords */
-    if(x0>y0) {i1=1; j1=0;} /* lower triangle, XY order: (0,0)->(1,0)->(1,1) */
-    else {i1=0; j1=1;}      /* upper triangle, YX order: (0,0)->(0,1)->(1,1) */
-
-    /* A step of (1,0) in (i,j) means a step of (1-c,-c) in (x,y), and */
-    /* a step of (0,1) in (i,j) means a step of (-c,1-c) in (x,y), where */
-    /* c = (3-sqrt(3))/6 */
-
-    x1 = x0 - i1 + G2; /* Offsets for middle corner in (x,y) unskewed coords */
-    y1 = y0 - j1 + G2;
-    x2 = x0 - 1.0f + 2.0f * G2; /* Offsets for last corner in (x,y) unskewed coords */
-    y2 = y0 - 1.0f + 2.0f * G2;
-
-    /* Wrap the integer indices at 256, to avoid indexing perm[] out of bounds */
-    ii = i % 256;
-    jj = j % 256;
-
-    /* Calculate the contribution from the three corners */
-    t0 = 0.5f - x0*x0-y0*y0;
-    if(t0 < 0.0f) n0 = 0.0f;
-    else {
-      t0 *= t0;
-      n0 = t0 * t0 * grad2(perm[ii+perm[jj]], x0, y0); 
-    }
-
-    t1 = 0.5f - x1*x1-y1*y1;
-    if(t1 < 0.0f) n1 = 0.0f;
-    else {
-      t1 *= t1;
-      n1 = t1 * t1 * grad2(perm[ii+i1+perm[jj+j1]], x1, y1);
-    }
-
-    t2 = 0.5f - x2*x2-y2*y2;
-    if(t2 < 0.0f) n2 = 0.0f;
-    else {
-      t2 *= t2;
-      n2 = t2 * t2 * grad2(perm[ii+1+perm[jj+1]], x2, y2);
-    }
-
-    /* Add contributions from each corner to get the final noise value. */
-    /* The result is scaled to return values in the interval [-1,1]. */
-    return 40.0f * (n0 + n1 + n2); /* TODO: The scale factor is preliminary! */
-}
-
-/* 3D simplex noise */
-GLfloat _slang_library_noise3 (GLfloat x, GLfloat y, GLfloat z)
-{
-/* Simple skewing factors for the 3D case */
-#define F3 0.333333333f
-#define G3 0.166666667f
-
-    float n0, n1, n2, n3; /* Noise contributions from the four corners */
-
-    /* Skew the input space to determine which simplex cell we're in */
-    float s = (x+y+z)*F3; /* Very nice and simple skew factor for 3D */
-    float xs = x+s;
-    float ys = y+s;
-    float zs = z+s;
-    int i = FASTFLOOR(xs);
-    int j = FASTFLOOR(ys);
-    int k = FASTFLOOR(zs);
-
-    float t = (float)(i+j+k)*G3; 
-    float X0 = i-t; /* Unskew the cell origin back to (x,y,z) space */
-    float Y0 = j-t;
-    float Z0 = k-t;
-    float x0 = x-X0; /* The x,y,z distances from the cell origin */
-    float y0 = y-Y0;
-    float z0 = z-Z0;
-
-    float x1, y1, z1, x2, y2, z2, x3, y3, z3;
-    int ii, jj, kk;
-    float t0, t1, t2, t3;
-
-    /* For the 3D case, the simplex shape is a slightly irregular tetrahedron. */
-    /* Determine which simplex we are in. */
-    int i1, j1, k1; /* Offsets for second corner of simplex in (i,j,k) coords */
-    int i2, j2, k2; /* Offsets for third corner of simplex in (i,j,k) coords */
-
-/* This code would benefit from a backport from the GLSL version! */
-    if(x0>=y0) {
-      if(y0>=z0)
-        { i1=1; j1=0; k1=0; i2=1; j2=1; k2=0; } /* X Y Z order */
-        else if(x0>=z0) { i1=1; j1=0; k1=0; i2=1; j2=0; k2=1; } /* X Z Y order */
-        else { i1=0; j1=0; k1=1; i2=1; j2=0; k2=1; } /* Z X Y order */
-      }
-    else { /* x0<y0 */
-      if(y0<z0) { i1=0; j1=0; k1=1; i2=0; j2=1; k2=1; } /* Z Y X order */
-      else if(x0<z0) { i1=0; j1=1; k1=0; i2=0; j2=1; k2=1; } /* Y Z X order */
-      else { i1=0; j1=1; k1=0; i2=1; j2=1; k2=0; } /* Y X Z order */
-    }
-
-    /* A step of (1,0,0) in (i,j,k) means a step of (1-c,-c,-c) in (x,y,z), */
-    /* a step of (0,1,0) in (i,j,k) means a step of (-c,1-c,-c) in (x,y,z), and */
-    /* a step of (0,0,1) in (i,j,k) means a step of (-c,-c,1-c) in (x,y,z), where */
-    /* c = 1/6. */
-
-    x1 = x0 - i1 + G3; /* Offsets for second corner in (x,y,z) coords */
-    y1 = y0 - j1 + G3;
-    z1 = z0 - k1 + G3;
-    x2 = x0 - i2 + 2.0f*G3; /* Offsets for third corner in (x,y,z) coords */
-    y2 = y0 - j2 + 2.0f*G3;
-    z2 = z0 - k2 + 2.0f*G3;
-    x3 = x0 - 1.0f + 3.0f*G3; /* Offsets for last corner in (x,y,z) coords */
-    y3 = y0 - 1.0f + 3.0f*G3;
-    z3 = z0 - 1.0f + 3.0f*G3;
-
-    /* Wrap the integer indices at 256, to avoid indexing perm[] out of bounds */
-    ii = i % 256;
-    jj = j % 256;
-    kk = k % 256;
-
-    /* Calculate the contribution from the four corners */
-    t0 = 0.6f - x0*x0 - y0*y0 - z0*z0;
-    if(t0 < 0.0f) n0 = 0.0f;
-    else {
-      t0 *= t0;
-      n0 = t0 * t0 * grad3(perm[ii+perm[jj+perm[kk]]], x0, y0, z0);
-    }
-
-    t1 = 0.6f - x1*x1 - y1*y1 - z1*z1;
-    if(t1 < 0.0f) n1 = 0.0f;
-    else {
-      t1 *= t1;
-      n1 = t1 * t1 * grad3(perm[ii+i1+perm[jj+j1+perm[kk+k1]]], x1, y1, z1);
-    }
-
-    t2 = 0.6f - x2*x2 - y2*y2 - z2*z2;
-    if(t2 < 0.0f) n2 = 0.0f;
-    else {
-      t2 *= t2;
-      n2 = t2 * t2 * grad3(perm[ii+i2+perm[jj+j2+perm[kk+k2]]], x2, y2, z2);
-    }
-
-    t3 = 0.6f - x3*x3 - y3*y3 - z3*z3;
-    if(t3<0.0f) n3 = 0.0f;
-    else {
-      t3 *= t3;
-      n3 = t3 * t3 * grad3(perm[ii+1+perm[jj+1+perm[kk+1]]], x3, y3, z3);
-    }
-
-    /* Add contributions from each corner to get the final noise value. */
-    /* The result is scaled to stay just inside [-1,1] */
-    return 32.0f * (n0 + n1 + n2 + n3); /* TODO: The scale factor is preliminary! */
-}
-
-/* 4D simplex noise */
-GLfloat _slang_library_noise4 (GLfloat x, GLfloat y, GLfloat z, GLfloat w)
-{
-  /* The skewing and unskewing factors are hairy again for the 4D case */
-#define F4 0.309016994f /* F4 = (Math.sqrt(5.0)-1.0)/4.0 */
-#define G4 0.138196601f /* G4 = (5.0-Math.sqrt(5.0))/20.0 */
-
-    float n0, n1, n2, n3, n4; /* Noise contributions from the five corners */
-
-    /* Skew the (x,y,z,w) space to determine which cell of 24 simplices we're in */
-    float s = (x + y + z + w) * F4; /* Factor for 4D skewing */
-    float xs = x + s;
-    float ys = y + s;
-    float zs = z + s;
-    float ws = w + s;
-    int i = FASTFLOOR(xs);
-    int j = FASTFLOOR(ys);
-    int k = FASTFLOOR(zs);
-    int l = FASTFLOOR(ws);
-
-    float t = (i + j + k + l) * G4; /* Factor for 4D unskewing */
-    float X0 = i - t; /* Unskew the cell origin back to (x,y,z,w) space */
-    float Y0 = j - t;
-    float Z0 = k - t;
-    float W0 = l - t;
-
-    float x0 = x - X0;  /* The x,y,z,w distances from the cell origin */
-    float y0 = y - Y0;
-    float z0 = z - Z0;
-    float w0 = w - W0;
-
-    /* For the 4D case, the simplex is a 4D shape I won't even try to describe. */
-    /* To find out which of the 24 possible simplices we're in, we need to */
-    /* determine the magnitude ordering of x0, y0, z0 and w0. */
-    /* The method below is a good way of finding the ordering of x,y,z,w and */
-    /* then find the correct traversal order for the simplex we're in. */
-    /* First, six pair-wise comparisons are performed between each possible pair */
-    /* of the four coordinates, and the results are used to add up binary bits */
-    /* for an integer index. */
-    int c1 = (x0 > y0) ? 32 : 0;
-    int c2 = (x0 > z0) ? 16 : 0;
-    int c3 = (y0 > z0) ? 8 : 0;
-    int c4 = (x0 > w0) ? 4 : 0;
-    int c5 = (y0 > w0) ? 2 : 0;
-    int c6 = (z0 > w0) ? 1 : 0;
-    int c = c1 + c2 + c3 + c4 + c5 + c6;
-
-    int i1, j1, k1, l1; /* The integer offsets for the second simplex corner */
-    int i2, j2, k2, l2; /* The integer offsets for the third simplex corner */
-    int i3, j3, k3, l3; /* The integer offsets for the fourth simplex corner */
-
-    float x1, y1, z1, w1, x2, y2, z2, w2, x3, y3, z3, w3, x4, y4, z4, w4;
-    int ii, jj, kk, ll;
-    float t0, t1, t2, t3, t4;
-
-    /* simplex[c] is a 4-vector with the numbers 0, 1, 2 and 3 in some order. */
-    /* Many values of c will never occur, since e.g. x>y>z>w makes x<z, y<w and x<w */
-    /* impossible. Only the 24 indices which have non-zero entries make any sense. */
-    /* We use a thresholding to set the coordinates in turn from the largest magnitude. */
-    /* The number 3 in the "simplex" array is at the position of the largest coordinate. */
-    i1 = simplex[c][0]>=3 ? 1 : 0;
-    j1 = simplex[c][1]>=3 ? 1 : 0;
-    k1 = simplex[c][2]>=3 ? 1 : 0;
-    l1 = simplex[c][3]>=3 ? 1 : 0;
-    /* The number 2 in the "simplex" array is at the second largest coordinate. */
-    i2 = simplex[c][0]>=2 ? 1 : 0;
-    j2 = simplex[c][1]>=2 ? 1 : 0;
-    k2 = simplex[c][2]>=2 ? 1 : 0;
-    l2 = simplex[c][3]>=2 ? 1 : 0;
-    /* The number 1 in the "simplex" array is at the second smallest coordinate. */
-    i3 = simplex[c][0]>=1 ? 1 : 0;
-    j3 = simplex[c][1]>=1 ? 1 : 0;
-    k3 = simplex[c][2]>=1 ? 1 : 0;
-    l3 = simplex[c][3]>=1 ? 1 : 0;
-    /* The fifth corner has all coordinate offsets = 1, so no need to look that up. */
-
-    x1 = x0 - i1 + G4; /* Offsets for second corner in (x,y,z,w) coords */
-    y1 = y0 - j1 + G4;
-    z1 = z0 - k1 + G4;
-    w1 = w0 - l1 + G4;
-    x2 = x0 - i2 + 2.0f*G4; /* Offsets for third corner in (x,y,z,w) coords */
-    y2 = y0 - j2 + 2.0f*G4;
-    z2 = z0 - k2 + 2.0f*G4;
-    w2 = w0 - l2 + 2.0f*G4;
-    x3 = x0 - i3 + 3.0f*G4; /* Offsets for fourth corner in (x,y,z,w) coords */
-    y3 = y0 - j3 + 3.0f*G4;
-    z3 = z0 - k3 + 3.0f*G4;
-    w3 = w0 - l3 + 3.0f*G4;
-    x4 = x0 - 1.0f + 4.0f*G4; /* Offsets for last corner in (x,y,z,w) coords */
-    y4 = y0 - 1.0f + 4.0f*G4;
-    z4 = z0 - 1.0f + 4.0f*G4;
-    w4 = w0 - 1.0f + 4.0f*G4;
-
-    /* Wrap the integer indices at 256, to avoid indexing perm[] out of bounds */
-    ii = i % 256;
-    jj = j % 256;
-    kk = k % 256;
-    ll = l % 256;
-
-    /* Calculate the contribution from the five corners */
-    t0 = 0.6f - x0*x0 - y0*y0 - z0*z0 - w0*w0;
-    if(t0 < 0.0f) n0 = 0.0f;
-    else {
-      t0 *= t0;
-      n0 = t0 * t0 * grad4(perm[ii+perm[jj+perm[kk+perm[ll]]]], x0, y0, z0, w0);
-    }
-
-   t1 = 0.6f - x1*x1 - y1*y1 - z1*z1 - w1*w1;
-    if(t1 < 0.0f) n1 = 0.0f;
-    else {
-      t1 *= t1;
-      n1 = t1 * t1 * grad4(perm[ii+i1+perm[jj+j1+perm[kk+k1+perm[ll+l1]]]], x1, y1, z1, w1);
-    }
-
-   t2 = 0.6f - x2*x2 - y2*y2 - z2*z2 - w2*w2;
-    if(t2 < 0.0f) n2 = 0.0f;
-    else {
-      t2 *= t2;
-      n2 = t2 * t2 * grad4(perm[ii+i2+perm[jj+j2+perm[kk+k2+perm[ll+l2]]]], x2, y2, z2, w2);
-    }
-
-   t3 = 0.6f - x3*x3 - y3*y3 - z3*z3 - w3*w3;
-    if(t3 < 0.0f) n3 = 0.0f;
-    else {
-      t3 *= t3;
-      n3 = t3 * t3 * grad4(perm[ii+i3+perm[jj+j3+perm[kk+k3+perm[ll+l3]]]], x3, y3, z3, w3);
-    }
-
-   t4 = 0.6f - x4*x4 - y4*y4 - z4*z4 - w4*w4;
-    if(t4 < 0.0f) n4 = 0.0f;
-    else {
-      t4 *= t4;
-      n4 = t4 * t4 * grad4(perm[ii+1+perm[jj+1+perm[kk+1+perm[ll+1]]]], x4, y4, z4, w4);
-    }
-
-    /* Sum up and scale the result to cover the range [-1,1] */
-    return 27.0f * (n0 + n1 + n2 + n3 + n4); /* TODO: The scale factor is preliminary! */
-}
-
diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c
index 79fd9a064f..99f2cbdcc0 100644
--- a/src/mesa/shader/slang/slang_link.c
+++ b/src/mesa/shader/slang/slang_link.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.2
+ * Version:  7.3
  *
  * Copyright (C) 2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -76,42 +77,93 @@ link_error(struct gl_shader_program *shProg, const char *msg)
 
 
 /**
+ * Check if the given bit is either set or clear in both bitfields.
+ */
+static GLboolean
+bits_agree(GLbitfield flags1, GLbitfield flags2, GLbitfield bit)
+{
+   return (flags1 & bit) == (flags2 & bit);
+}
+
+
+/**
  * Linking varying vars involves rearranging varying vars so that the
  * vertex program's output varyings matches the order of the fragment
  * program's input varyings.
+ * We'll then rewrite instructions to replace PROGRAM_VARYING with either
+ * PROGRAM_INPUT or PROGRAM_OUTPUT depending on whether it's a vertex or
+ * fragment shader.
+ * This is also where we set program Input/OutputFlags to indicate
+ * which inputs are centroid-sampled, invariant, etc.
  */
 static GLboolean
 link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog)
 {
    GLuint *map, i, firstVarying, newFile;
+   GLbitfield *inOutFlags;
 
    map = (GLuint *) malloc(prog->Varying->NumParameters * sizeof(GLuint));
    if (!map)
       return GL_FALSE;
 
+   /* Varying variables are treated like other vertex program outputs
+    * (and like other fragment program inputs).  The position of the
+    * first varying differs for vertex/fragment programs...
+    * Also, replace File=PROGRAM_VARYING with File=PROGRAM_INPUT/OUTPUT.
+    */
+   if (prog->Target == GL_VERTEX_PROGRAM_ARB) {
+      firstVarying = VERT_RESULT_VAR0;
+      newFile = PROGRAM_OUTPUT;
+      inOutFlags = prog->OutputFlags;
+   }
+   else {
+      assert(prog->Target == GL_FRAGMENT_PROGRAM_ARB);
+      firstVarying = FRAG_ATTRIB_VAR0;
+      newFile = PROGRAM_INPUT;
+      inOutFlags = prog->InputFlags;
+   }
+
    for (i = 0; i < prog->Varying->NumParameters; i++) {
       /* see if this varying is in the linked varying list */
       const struct gl_program_parameter *var = prog->Varying->Parameters + i;
       GLint j = _mesa_lookup_parameter_index(shProg->Varying, -1, var->Name);
       if (j >= 0) {
-         /* already in list, check size */
-         if (var->Size != shProg->Varying->Parameters[j].Size) {
-            /* error */
+         /* varying is already in list, do some error checking */
+         const struct gl_program_parameter *v =
+            &shProg->Varying->Parameters[j];
+         if (var->Size != v->Size) {
             link_error(shProg, "mismatched varying variable types");
             return GL_FALSE;
          }
+         if (!bits_agree(var->Flags, v->Flags, PROG_PARAM_BIT_CENTROID)) {
+            char msg[100];
+            _mesa_snprintf(msg, sizeof(msg),
+                           "centroid modifier mismatch for '%s'", var->Name);
+            link_error(shProg, msg);
+            return GL_FALSE;
+         }
+         if (!bits_agree(var->Flags, v->Flags, PROG_PARAM_BIT_INVARIANT)) {
+            char msg[100];
+            _mesa_snprintf(msg, sizeof(msg),
+                           "invariant modifier mismatch for '%s'", var->Name);
+            link_error(shProg, msg);
+            return GL_FALSE;
+         }
       }
       else {
          /* not already in linked list */
-         j = _mesa_add_varying(shProg->Varying, var->Name, var->Size);
+         j = _mesa_add_varying(shProg->Varying, var->Name, var->Size,
+                               var->Flags);
       }
 
-      /* map varying[i] to varying[j].
+      /* Map varying[i] to varying[j].
+       * Plus, set prog->Input/OutputFlags[] as described above.
        * Note: the loop here takes care of arrays or large (sz>4) vars.
        */
       {
          GLint sz = var->Size;
          while (sz > 0) {
+            inOutFlags[firstVarying + j] = var->Flags;
             /*printf("Link varying from %d to %d\n", i, j);*/
             map[i++] = j++;
             sz -= 4;
@@ -121,21 +173,6 @@ link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog)
    }
 
 
-   /* Varying variables are treated like other vertex program outputs
-    * (and like other fragment program inputs).  The position of the
-    * first varying differs for vertex/fragment programs...
-    * Also, replace File=PROGRAM_VARYING with File=PROGRAM_INPUT/OUTPUT.
-    */
-   if (prog->Target == GL_VERTEX_PROGRAM_ARB) {
-      firstVarying = VERT_RESULT_VAR0;
-      newFile = PROGRAM_OUTPUT;
-   }
-   else {
-      assert(prog->Target == GL_FRAGMENT_PROGRAM_ARB);
-      firstVarying = FRAG_ATTRIB_VAR0;
-      newFile = PROGRAM_INPUT;
-   }
-
    /* OK, now scan the program/shader instructions looking for varying vars,
     * replacing the old index with the new index.
     */
@@ -170,13 +207,27 @@ link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog)
  * Build the shProg->Uniforms list.
  * This is basically a list/index of all uniforms found in either/both of
  * the vertex and fragment shaders.
+ *
+ * About uniforms:
+ * Each uniform has two indexes, one that points into the vertex
+ * program's parameter array and another that points into the fragment
+ * program's parameter array.  When the user changes a uniform's value
+ * we have to change the value in the vertex and/or fragment program's
+ * parameter array.
+ *
+ * This function will be called twice to set up the two uniform->parameter
+ * mappings.
+ *
+ * If a uniform is only present in the vertex program OR fragment program
+ * then the fragment/vertex parameter index, respectively, will be -1.
  */
-static void
-link_uniform_vars(struct gl_shader_program *shProg,
+static GLboolean
+link_uniform_vars(GLcontext *ctx,
+                  struct gl_shader_program *shProg,
                   struct gl_program *prog,
                   GLuint *numSamplers)
 {
-   GLuint samplerMap[MAX_SAMPLERS];
+   GLuint samplerMap[200]; /* max number of samplers declared, not used */
    GLuint i;
 
    for (i = 0; i < prog->Parameters->NumParameters; i++) {
@@ -191,44 +242,64 @@ link_uniform_vars(struct gl_shader_program *shProg,
        * Furthermore, we'll need to fix the state-var's size/datatype info.
        */
 
-      if ((p->Type == PROGRAM_UNIFORM && p->Used) ||
-          p->Type == PROGRAM_SAMPLER) {
+      if ((p->Type == PROGRAM_UNIFORM || p->Type == PROGRAM_SAMPLER)
+          && p->Used) {
+         /* add this uniform, indexing into the target's Parameters list */
          struct gl_uniform *uniform =
             _mesa_append_uniform(shProg->Uniforms, p->Name, prog->Target, i);
          if (uniform)
             uniform->Initialized = p->Initialized;
       }
 
-      if (p->Type == PROGRAM_SAMPLER) {
+      /* The samplerMap[] table we build here is used to remap/re-index
+       * sampler references by TEX instructions.
+       */
+      if (p->Type == PROGRAM_SAMPLER && p->Used) {
          /* Allocate a new sampler index */
-         GLuint sampNum = *numSamplers;
          GLuint oldSampNum = (GLuint) prog->Parameters->ParameterValues[i][0];
-         assert(oldSampNum < MAX_SAMPLERS);
-         samplerMap[oldSampNum] = sampNum;
+         GLuint newSampNum = *numSamplers;
+         if (newSampNum >= ctx->Const.MaxTextureImageUnits) {
+            char s[100];
+            _mesa_sprintf(s, "Too many texture samplers (%u, max is %u)",
+                          newSampNum, ctx->Const.MaxTextureImageUnits);
+            link_error(shProg, s);
+            return GL_FALSE;
+         }
+         /* save old->new mapping in the table */
+         if (oldSampNum < Elements(samplerMap))
+            samplerMap[oldSampNum] = newSampNum;
+         /* update parameter's sampler index */
+         prog->Parameters->ParameterValues[i][0] = (GLfloat) newSampNum;
          (*numSamplers)++;
       }
    }
 
-
-   /* OK, now scan the program/shader instructions looking for sampler vars,
-    * replacing the old index with the new index.
+   /* OK, now scan the program/shader instructions looking for texture
+    * instructions using sampler vars.  Replace old sampler indexes with
+    * new ones.
     */
    prog->SamplersUsed = 0x0;
    for (i = 0; i < prog->NumInstructions; i++) {
       struct prog_instruction *inst = prog->Instructions + i;
       if (_mesa_is_tex_instruction(inst->Opcode)) {
-         /*
+         const GLint oldSampNum = inst->TexSrcUnit;
+
+#if 0
          printf("====== remap sampler from %d to %d\n",
-                inst->Sampler, map[ inst->Sampler ]);
-         */
+                inst->TexSrcUnit, samplerMap[ inst->TexSrcUnit ]);
+#endif
+
          /* here, texUnit is really samplerUnit */
-         assert(inst->TexSrcUnit < MAX_SAMPLERS);
-         inst->TexSrcUnit = samplerMap[inst->TexSrcUnit];
-         prog->SamplerTargets[inst->TexSrcUnit] = inst->TexSrcTarget;
-         prog->SamplersUsed |= (1 << inst->TexSrcUnit);
+         if (oldSampNum < Elements(samplerMap)) {
+            const GLuint newSampNum = samplerMap[oldSampNum];
+            inst->TexSrcUnit = newSampNum;
+            prog->SamplerTargets[newSampNum] = inst->TexSrcTarget;
+            prog->SamplersUsed |= (1 << newSampNum);
+         }
       }
    }
 
+   return GL_TRUE;
 }
 
 
@@ -247,7 +318,7 @@ _slang_resolve_attributes(struct gl_shader_program *shProg,
 {
    GLint attribMap[MAX_VERTEX_ATTRIBS];
    GLuint i, j;
-   GLbitfield usedAttributes;
+   GLbitfield usedAttributes; /* generics only, not legacy attributes */
 
    assert(origProg != linkedProg);
    assert(origProg->Target == GL_VERTEX_PROGRAM_ARB);
@@ -271,6 +342,15 @@ _slang_resolve_attributes(struct gl_shader_program *shProg,
       usedAttributes |= (1 << attr);
    }
 
+   /* If gl_Vertex is used, that actually counts against the limit
+    * on generic vertex attributes.  This avoids the ambiguity of
+    * whether glVertexAttrib4fv(0, v) sets legacy attribute 0 (vert pos)
+    * or generic attribute[0].  If gl_Vertex is used, we want the former.
+    */
+   if (origProg->InputsRead & VERT_BIT_POS) {
+      usedAttributes |= 0x1;
+   }
+
    /* initialize the generic attribute map entries to -1 */
    for (i = 0; i < MAX_VERTEX_ATTRIBS; i++) {
       attribMap[i] = -1;
@@ -313,7 +393,7 @@ _slang_resolve_attributes(struct gl_shader_program *shProg,
                    * Start at 1 since generic attribute 0 always aliases
                    * glVertex/position.
                    */
-                  for (attr = 1; attr < MAX_VERTEX_ATTRIBS; attr++) {
+                  for (attr = 0; attr < MAX_VERTEX_ATTRIBS; attr++) {
                      if (((1 << attr) & usedAttributes) == 0)
                         break;
                   }
@@ -422,7 +502,6 @@ _slang_update_inputs_outputs(struct gl_program *prog)
          maxAddrReg = MAX2(maxAddrReg, inst->DstReg.Index + 1);
       }
    }
-
    prog->NumAddressRegs = maxAddrReg;
 }
 
@@ -524,10 +603,18 @@ _slang_link(GLcontext *ctx,
    }
 
    /* link uniform vars */
-   if (shProg->VertexProgram)
-      link_uniform_vars(shProg, &shProg->VertexProgram->Base, &numSamplers);
-   if (shProg->FragmentProgram)
-      link_uniform_vars(shProg, &shProg->FragmentProgram->Base, &numSamplers);
+   if (shProg->VertexProgram) {
+      if (!link_uniform_vars(ctx, shProg, &shProg->VertexProgram->Base,
+                             &numSamplers)) {
+         return;
+      }
+   }
+   if (shProg->FragmentProgram) {
+      if (!link_uniform_vars(ctx, shProg, &shProg->FragmentProgram->Base,
+                             &numSamplers)) {
+         return;
+      }
+   }
 
    /*_mesa_print_uniforms(shProg->Uniforms);*/
 
@@ -568,6 +655,17 @@ _slang_link(GLcontext *ctx,
       }         
    }
 
+   /* check that gl_FragColor and gl_FragData are not both written to */
+   if (shProg->FragmentProgram) {
+      GLbitfield outputsWritten = shProg->FragmentProgram->Base.OutputsWritten;
+      if ((outputsWritten & ((1 << FRAG_RESULT_COLR))) &&
+          (outputsWritten >= (1 << FRAG_RESULT_DATA0))) {
+         link_error(shProg, "Fragment program cannot write both gl_FragColor"
+                    " and gl_FragData[].\n");
+         return;
+      }         
+   }
+
 
    if (fragProg && shProg->FragmentProgram) {
       /* Compute initial program's TexturesUsed info */
@@ -576,12 +674,12 @@ _slang_link(GLcontext *ctx,
       /* notify driver that a new fragment program has been compiled/linked */
       ctx->Driver.ProgramStringNotify(ctx, GL_FRAGMENT_PROGRAM_ARB,
                                       &shProg->FragmentProgram->Base);
-      if (MESA_VERBOSE & VERBOSE_GLSL_DUMP) {
-         printf("Mesa original fragment program:\n");
+      if (ctx->Shader.Flags & GLSL_DUMP) {
+         _mesa_printf("Mesa pre-link fragment program:\n");
          _mesa_print_program(&fragProg->Base);
          _mesa_print_program_parameters(ctx, &fragProg->Base);
 
-         printf("Mesa post-link fragment program:\n");
+         _mesa_printf("Mesa post-link fragment program:\n");
          _mesa_print_program(&shProg->FragmentProgram->Base);
          _mesa_print_program_parameters(ctx, &shProg->FragmentProgram->Base);
       }
@@ -594,17 +692,25 @@ _slang_link(GLcontext *ctx,
       /* notify driver that a new vertex program has been compiled/linked */
       ctx->Driver.ProgramStringNotify(ctx, GL_VERTEX_PROGRAM_ARB,
                                       &shProg->VertexProgram->Base);
-      if (MESA_VERBOSE & VERBOSE_GLSL_DUMP) {
-         printf("Mesa original vertex program:\n");
+      if (ctx->Shader.Flags & GLSL_DUMP) {
+         _mesa_printf("Mesa pre-link vertex program:\n");
          _mesa_print_program(&vertProg->Base);
          _mesa_print_program_parameters(ctx, &vertProg->Base);
 
-         printf("Mesa post-link vertex program:\n");
+         _mesa_printf("Mesa post-link vertex program:\n");
          _mesa_print_program(&shProg->VertexProgram->Base);
          _mesa_print_program_parameters(ctx, &shProg->VertexProgram->Base);
       }
    }
 
+   if (ctx->Shader.Flags & GLSL_DUMP) {
+      _mesa_printf("Varying vars:\n");
+      _mesa_print_parameter_list(shProg->Varying);
+      if (shProg->InfoLog) {
+         _mesa_printf("Info Log: %s\n", shProg->InfoLog);
+      }
+   }
+
    shProg->LinkStatus = (shProg->VertexProgram || shProg->FragmentProgram);
 }
 
diff --git a/src/mesa/shader/slang/slang_log.c b/src/mesa/shader/slang/slang_log.c
index 25f696f67e..d7d2b4fbfd 100644
--- a/src/mesa/shader/slang/slang_log.c
+++ b/src/mesa/shader/slang/slang_log.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.3
+ * Version:  7.3
  *
  * Copyright (C) 2005-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -43,11 +44,7 @@ void
 slang_info_log_destruct(slang_info_log * log)
 {
    if (!log->dont_free_text)
-#if 0
-      slang_alloc_free(log->text);
-#else
       _mesa_free(log->text);
-#endif
 }
 
 static int
@@ -64,18 +61,10 @@ slang_info_log_message(slang_info_log * log, const char *prefix,
    if (log->text != NULL) {
       GLuint old_len = slang_string_length(log->text);
       log->text = (char *)
-#if 0
-	 slang_alloc_realloc(log->text, old_len + 1, old_len + size);
-#else
 	 _mesa_realloc(log->text, old_len + 1, old_len + size);
-#endif
    }
    else {
-#if 0
-      log->text = (char *) (slang_alloc_malloc(size));
-#else
       log->text = (char *) (_mesa_malloc(size));
-#endif
       if (log->text != NULL)
          log->text[0] = '\0';
    }
@@ -88,10 +77,6 @@ slang_info_log_message(slang_info_log * log, const char *prefix,
    slang_string_concat(log->text, msg);
    slang_string_concat(log->text, "\n");
 
-   if (MESA_VERBOSE & VERBOSE_GLSL) {
-      _mesa_printf("Mesa: GLSL %s", log->text);
-   }
-
    return 1;
 }
 
diff --git a/src/mesa/shader/slang/slang_mem.c b/src/mesa/shader/slang/slang_mem.c
index 21d6bfce88..9224578edb 100644
--- a/src/mesa/shader/slang/slang_mem.c
+++ b/src/mesa/shader/slang/slang_mem.c
@@ -184,6 +184,7 @@ _slang_realloc(void *oldBuffer, GLuint oldSize, GLuint newSize)
 #else
    GET_CURRENT_CONTEXT(ctx);
    slang_mempool *pool = (slang_mempool *) ctx->Shader.MemPool;
+   (void) pool;
 
    if (newSize < oldSize) {
       return oldBuffer;
@@ -235,6 +236,7 @@ _slang_free(void *addr)
    if (addr) {
       GET_CURRENT_CONTEXT(ctx);
       slang_mempool *pool = (slang_mempool *) ctx->Shader.MemPool;
+      (void) pool;
       ASSERT(is_valid_address(pool, addr));
    }
 #endif
diff --git a/src/mesa/shader/slang/slang_preprocess.c b/src/mesa/shader/slang/slang_preprocess.c
index 786f6467c9..89aaa3a621 100644
--- a/src/mesa/shader/slang/slang_preprocess.c
+++ b/src/mesa/shader/slang/slang_preprocess.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.2
  *
- * Copyright (C) 2005-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2005-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009 VMware, Inc.   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -51,6 +51,9 @@ grammar_error_to_log (slang_info_log *log)
    GLint pos;
 
    grammar_get_last_error ((byte *) (buf), sizeof (buf), &pos);
+   if (buf[0] == 0) {
+      _mesa_snprintf(buf, sizeof(buf), "Preprocessor error");
+   }
    slang_info_log_error (log, buf);
 }
 
@@ -475,52 +478,112 @@ pp_cond_stack_reevaluate (pp_cond_stack *self)
    self->top->effective = self->top->current && self->top[1].effective;
 }
 
-/*
+
+/**
  * Extension enables through #extension directive.
  * NOTE: Currently, only enable/disable state is stored.
  */
-
 typedef struct
 {
-   GLboolean MESA_shader_debug;        /* GL_MESA_shader_debug enable */
-   GLboolean ARB_texture_rectangle; /* GL_ARB_texture_rectangle enable */
+   GLboolean ARB_draw_buffers;
+   GLboolean ARB_texture_rectangle;
 } pp_ext;
 
-/*
+
+/**
  * Disable all extensions. Called at startup and on #extension all: disable.
  */
 static GLvoid
-pp_ext_disable_all (pp_ext *self)
+pp_ext_disable_all(pp_ext *self)
 {
-   self->MESA_shader_debug = GL_FALSE;
+   _mesa_memset(self, 0, sizeof(self));
 }
 
+
+/**
+ * Called during preprocessor initialization to set the initial enable/disable
+ * state of extensions.
+ */
 static GLvoid
-pp_ext_init (pp_ext *self)
+pp_ext_init(pp_ext *self, const struct gl_extensions *extensions)
 {
    pp_ext_disable_all (self);
-   self->ARB_texture_rectangle = GL_TRUE;
-   /* Other initialization code goes here. */
+   self->ARB_draw_buffers = GL_TRUE;
+   if (extensions->NV_texture_rectangle)
+      self->ARB_texture_rectangle = GL_TRUE;
 }
 
+/**
+ * Called in response to #extension directives to enable/disable
+ * the named extension.
+ */
 static GLboolean
-pp_ext_set (pp_ext *self, const char *name, GLboolean enable)
+pp_ext_set(pp_ext *self, const char *name, GLboolean enable)
 {
-   if (_mesa_strcmp (name, "MESA_shader_debug") == 0)
-      self->MESA_shader_debug = enable;
+   if (_mesa_strcmp (name, "GL_ARB_draw_buffers") == 0)
+      self->ARB_draw_buffers = enable;
    else if (_mesa_strcmp (name, "GL_ARB_texture_rectangle") == 0)
       self->ARB_texture_rectangle = enable;
-   /* Next extension name tests go here. */
    else
       return GL_FALSE;
    return GL_TRUE;
 }
 
-/*
- * The state of preprocessor: current line, file and version number, list of all defined macros
- * and the #if/#endif context.
+
+static void
+pp_pragmas_init(struct gl_sl_pragmas *pragmas)
+{
+   pragmas->Optimize = GL_TRUE;
+   pragmas->Debug = GL_FALSE;
+}
+
+
+/**
+ * Called in response to #pragma.  For example, "#pragma debug(on)" would
+ * call this function as pp_pragma("debug", "on").
+ * \return GL_TRUE if pragma is valid, GL_FALSE if invalid
  */
+static GLboolean
+pp_pragma(struct gl_sl_pragmas *pragmas, const char *pragma, const char *param)
+{
+#if 0
+   printf("#pragma %s %s\n", pragma, param);
+#endif
+   if (_mesa_strcmp(pragma, "optimize") == 0) {
+      if (!param)
+         return GL_FALSE; /* missing required param */
+      if (_mesa_strcmp(param, "on") == 0) {
+         pragmas->Optimize = GL_TRUE;
+      }
+      else if (_mesa_strcmp(param, "off") == 0) {
+         pragmas->Optimize = GL_FALSE;
+      }
+      else {
+         return GL_FALSE; /* invalid param */
+      }
+   }
+   else if (_mesa_strcmp(pragma, "debug") == 0) {
+      if (!param)
+         return GL_FALSE; /* missing required param */
+      if (_mesa_strcmp(param, "on") == 0) {
+         pragmas->Debug = GL_TRUE;
+      }
+      else if (_mesa_strcmp(param, "off") == 0) {
+         pragmas->Debug = GL_FALSE;
+      }
+      else {
+         return GL_FALSE; /* invalid param */
+      }
+   }
+   /* all other pragmas are silently ignored */
+   return GL_TRUE;
+}
+
 
+/**
+ * The state of preprocessor: current line, file and version number, list
+ * of all defined macros and the #if/#endif context.
+ */
 typedef struct
 {
    GLint line;
@@ -533,7 +596,8 @@ typedef struct
 } pp_state;
 
 static GLvoid
-pp_state_init (pp_state *self, slang_info_log *elog)
+pp_state_init (pp_state *self, slang_info_log *elog,
+               const struct gl_extensions *extensions)
 {
    self->line = 0;
    self->file = 1;
@@ -543,7 +607,7 @@ pp_state_init (pp_state *self, slang_info_log *elog)
    self->version = 110;
 #endif
    pp_symbols_init (&self->symbols);
-   pp_ext_init (&self->ext);
+   pp_ext_init (&self->ext, extensions);
    self->elog = elog;
 
    /* Initialize condition stack and create the global context. */
@@ -641,8 +705,10 @@ expand_symbol (expand_state *e, pp_symbol *symbol)
       SKIP_WHITE(e->input);
 
       /* Parse macro actual parameters. This can be anything, separated by a colon.
-       * TODO: What about nested/grouped parameters by parenthesis? */
+       */
       for (i = 0; i < symbol->parameters.count; i++) {
+         GLuint nested_paren_count = 0; /* track number of nested parentheses */
+
          if (*e->input == ')') {
             slang_info_log_error (e->state->elog, "preprocess error: unexpected ')'.");
             return GL_FALSE;
@@ -650,8 +716,19 @@ expand_symbol (expand_state *e, pp_symbol *symbol)
 
          /* Eat all characters up to the comma or closing parentheses. */
          pp_symbol_reset (&symbol->parameters.symbols[i]);
-         while (!IS_NULL(*e->input) && *e->input != ',' && *e->input != ')')
+         while (!IS_NULL(*e->input)) {
+            /* Exit loop only when all nested parens have been eaten. */
+            if (nested_paren_count == 0 && (*e->input == ',' || *e->input == ')'))
+               break;
+
+            /* Actually count nested parens here. */
+            if (*e->input == '(')
+               nested_paren_count++;
+            else if (*e->input == ')')
+               nested_paren_count--;
+
             slang_string_pushc (&symbol->parameters.symbols[i].replacement, *e->input++);
+         }
 
          /* If it was not the last paremeter, skip the comma. Otherwise, skip the
           * closing parentheses. */
@@ -837,9 +914,16 @@ parse_if (slang_string *output, const byte *prod, GLuint *pi, GLint *result, pp_
 #define BEHAVIOR_WARN    3
 #define BEHAVIOR_DISABLE 4
 
+#define PRAGMA_NO_PARAM  0
+#define PRAGMA_PARAM     1
+
+
 static GLboolean
-preprocess_source (slang_string *output, const char *source, grammar pid, grammar eid,
-                   slang_info_log *elog)
+preprocess_source (slang_string *output, const char *source,
+                   grammar pid, grammar eid,
+                   slang_info_log *elog,
+                   const struct gl_extensions *extensions,
+                   struct gl_sl_pragmas *pragmas)
 {
    static const char *predefined[] = {
       "__FILE__",
@@ -860,7 +944,8 @@ preprocess_source (slang_string *output, const char *source, grammar pid, gramma
       return GL_FALSE;
    }
 
-   pp_state_init (&state, elog);
+   pp_state_init (&state, elog, extensions);
+   pp_pragmas_init (pragmas);
 
    /* add the predefined symbols to the symbol table */
    for (i = 0; predefined[i]; i++) {
@@ -913,9 +998,11 @@ preprocess_source (slang_string *output, const char *source, grammar pid, gramma
       else {
          const char *id;
          GLuint idlen;
+         GLubyte token;
 
          i++;
-         switch (prod[i++]) {
+         token = prod[i++];
+         switch (token) {
 
          case TOKEN_END:
             /* End of source string.
@@ -1132,6 +1219,25 @@ preprocess_source (slang_string *output, const char *source, grammar pid, gramma
             }
             break;
 
+         case TOKEN_PRAGMA:
+            {
+               GLint have_param;
+               const char *pragma, *param;
+
+               pragma = (const char *) (&prod[i]);
+               i += _mesa_strlen(pragma) + 1;
+               have_param = (prod[i++] == PRAGMA_PARAM);
+               if (have_param) {
+                  param = (const char *) (&prod[i]);
+                  i += _mesa_strlen(param) + 1;
+               }
+               else {
+                  param = NULL;
+               }
+               pp_pragma(pragmas, pragma, param);
+            }
+            break;
+
          case TOKEN_LINE:
             id = (const char *) (&prod[i]);
             i += _mesa_strlen (id) + 1;
@@ -1183,8 +1289,21 @@ error:
    return GL_FALSE;
 }
 
+
+/**
+ * Run preprocessor on source code.
+ * \param extensions  indicates which GL extensions are enabled
+ * \param output  the post-process results
+ * \param input  the input text
+ * \param elog  log to record warnings, errors
+ * \return GL_TRUE for success, GL_FALSE for error
+ */
 GLboolean
-_slang_preprocess_directives (slang_string *output, const char *input, slang_info_log *elog)
+_slang_preprocess_directives(slang_string *output,
+                             const char *input,
+                             slang_info_log *elog,
+                             const struct gl_extensions *extensions,
+                             struct gl_sl_pragmas *pragmas)
 {
    grammar pid, eid;
    GLboolean success;
@@ -1200,7 +1319,7 @@ _slang_preprocess_directives (slang_string *output, const char *input, slang_inf
       grammar_destroy (pid);
       return GL_FALSE;
    }
-   success = preprocess_source (output, input, pid, eid, elog);
+   success = preprocess_source (output, input, pid, eid, elog, extensions, pragmas);
    grammar_destroy (eid);
    grammar_destroy (pid);
    return success;
diff --git a/src/mesa/shader/slang/slang_preprocess.h b/src/mesa/shader/slang/slang_preprocess.h
index d8eb12e4ff..f344820dae 100644
--- a/src/mesa/shader/slang/slang_preprocess.h
+++ b/src/mesa/shader/slang/slang_preprocess.h
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.3
  *
- * Copyright (C) 2005-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2005-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009 VMware, Inc.   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -33,8 +33,9 @@ extern GLboolean
 _slang_preprocess_version (const char *, GLuint *, GLuint *, slang_info_log *);
 
 extern GLboolean
-_slang_preprocess_directives (slang_string *output, const char *input,
-                              slang_info_log *);
-
+_slang_preprocess_directives(slang_string *output, const char *input,
+                             slang_info_log *,
+                             const struct gl_extensions *extensions,
+                             struct gl_sl_pragmas *pragmas);
 
 #endif /* SLANG_PREPROCESS_H */
diff --git a/src/mesa/shader/slang/slang_print.c b/src/mesa/shader/slang/slang_print.c
index b88cefc283..bf1475f91d 100644
--- a/src/mesa/shader/slang/slang_print.c
+++ b/src/mesa/shader/slang/slang_print.c
@@ -278,7 +278,7 @@ slang_print_tree(const slang_operation *op, int indent)
       assert(op->num_children == 0 || op->num_children == 1);
       {
          slang_variable *v;
-         v = _slang_locate_variable(op->locals, op->a_id, GL_TRUE);
+         v = _slang_variable_locate(op->locals, op->a_id, GL_TRUE);
          if (v) {
             const slang_variable_scope *scope;
             spaces(indent);
@@ -636,7 +636,7 @@ slang_print_tree(const slang_operation *op, int indent)
    case SLANG_OPER_CALL:
 #if 0
          slang_function *fun
-            = _slang_locate_function(A->space.funcs, oper->a_id,
+            = _slang_function_locate(A->space.funcs, oper->a_id,
                                      oper->children,
                                      oper->num_children, &A->space, A->atoms);
 #endif
@@ -653,6 +653,11 @@ slang_print_tree(const slang_operation *op, int indent)
       printf(")\n");
       break;
 
+   case SLANG_OPER_METHOD:
+      spaces(indent);
+      printf("METHOD CALL %s.%s\n", (char *) op->a_obj, (char *) op->a_id);
+      break;
+
    case SLANG_OPER_FIELD:
       spaces(indent);
       printf("FIELD %s of\n", (char*) op->a_id);
diff --git a/src/mesa/shader/slang/slang_simplify.c b/src/mesa/shader/slang/slang_simplify.c
index 6ef95c48f9..b8a21f642c 100644
--- a/src/mesa/shader/slang/slang_simplify.c
+++ b/src/mesa/shader/slang/slang_simplify.c
@@ -135,7 +135,7 @@ _slang_simplify(slang_operation *oper,
       /* look for user-defined constant */
       {
          slang_variable *var;
-         var = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE);
+         var = _slang_variable_locate(oper->locals, oper->a_id, GL_TRUE);
          if (var) {
             if (var->type.qualifier == SLANG_QUAL_CONST &&
                 var->initializer &&
@@ -380,7 +380,7 @@ _slang_cast_func_params(slang_operation *callOper, const slang_function *fun,
       /* Get type of arg[i] */
       if (!slang_typeinfo_construct(&argType))
          return GL_FALSE;
-      if (!_slang_typeof_operation_(&callOper->children[i], space,
+      if (!_slang_typeof_operation(&callOper->children[i], space,
                                     &argType, atoms, log)) {
          slang_typeinfo_destruct(&argType);
          return GL_FALSE;
@@ -464,7 +464,7 @@ _slang_adapt_call(slang_operation *callOper, const slang_function *fun,
          /* Get type of arg[i] */
          if (!slang_typeinfo_construct(&argType))
             return GL_FALSE;
-         if (!_slang_typeof_operation_(&callOper->children[i], space,
+         if (!_slang_typeof_operation(&callOper->children[i], space,
                                        &argType, atoms, log)) {
             slang_typeinfo_destruct(&argType);
             return GL_FALSE;
diff --git a/src/mesa/shader/slang/slang_storage.h b/src/mesa/shader/slang/slang_storage.h
index d3047ee516..1876a36dd6 100644
--- a/src/mesa/shader/slang/slang_storage.h
+++ b/src/mesa/shader/slang/slang_storage.h
@@ -26,7 +26,6 @@
 #define SLANG_STORAGE_H
 
 #include "slang_compile.h"
-#include "slang_typeinfo.h"
 
 
 /*
diff --git a/src/mesa/shader/slang/slang_typeinfo.c b/src/mesa/shader/slang/slang_typeinfo.c
index b2f0f14995..1ef43f58c0 100644
--- a/src/mesa/shader/slang/slang_typeinfo.c
+++ b/src/mesa/shader/slang/slang_typeinfo.c
@@ -23,7 +23,7 @@
  */
 
 /**
- * \file slang_assemble_typeinfo.c
+ * \file slang_typeinfo.c
  * slang type info
  * \author Michal Krol
  */
@@ -130,7 +130,7 @@ _slang_is_swizzle(const char *field, GLuint rows, slang_swizzle * swz)
  * do not have duplicated fields.  Returns GL_TRUE if this is a
  * swizzle mask.  Returns GL_FALSE otherwise
  */
-GLboolean
+static GLboolean
 _slang_is_swizzle_mask(const slang_swizzle * swz, GLuint rows)
 {
    GLuint i, c = 0;
@@ -154,7 +154,7 @@ _slang_is_swizzle_mask(const slang_swizzle * swz, GLuint rows)
  * Combines (multiplies) two swizzles to form single swizzle.
  * Example: "vec.wzyx.yx" --> "vec.zw".
  */
-GLvoid
+static void
 _slang_multiply_swizzles(slang_swizzle * dst, const slang_swizzle * left,
                          const slang_swizzle * right)
 {
@@ -166,6 +166,110 @@ _slang_multiply_swizzles(slang_swizzle * dst, const slang_swizzle * left,
 }
 
 
+typedef struct
+{
+   const char *name;
+   slang_type_specifier_type type;
+} type_specifier_type_name;
+
+static const type_specifier_type_name type_specifier_type_names[] = {
+   {"void", SLANG_SPEC_VOID},
+   {"bool", SLANG_SPEC_BOOL},
+   {"bvec2", SLANG_SPEC_BVEC2},
+   {"bvec3", SLANG_SPEC_BVEC3},
+   {"bvec4", SLANG_SPEC_BVEC4},
+   {"int", SLANG_SPEC_INT},
+   {"ivec2", SLANG_SPEC_IVEC2},
+   {"ivec3", SLANG_SPEC_IVEC3},
+   {"ivec4", SLANG_SPEC_IVEC4},
+   {"float", SLANG_SPEC_FLOAT},
+   {"vec2", SLANG_SPEC_VEC2},
+   {"vec3", SLANG_SPEC_VEC3},
+   {"vec4", SLANG_SPEC_VEC4},
+   {"mat2", SLANG_SPEC_MAT2},
+   {"mat3", SLANG_SPEC_MAT3},
+   {"mat4", SLANG_SPEC_MAT4},
+   {"mat2x3", SLANG_SPEC_MAT23},
+   {"mat3x2", SLANG_SPEC_MAT32},
+   {"mat2x4", SLANG_SPEC_MAT24},
+   {"mat4x2", SLANG_SPEC_MAT42},
+   {"mat3x4", SLANG_SPEC_MAT34},
+   {"mat4x3", SLANG_SPEC_MAT43},
+   {"sampler1D", SLANG_SPEC_SAMPLER1D},
+   {"sampler2D", SLANG_SPEC_SAMPLER2D},
+   {"sampler3D", SLANG_SPEC_SAMPLER3D},
+   {"samplerCube", SLANG_SPEC_SAMPLERCUBE},
+   {"sampler1DShadow", SLANG_SPEC_SAMPLER1DSHADOW},
+   {"sampler2DShadow", SLANG_SPEC_SAMPLER2DSHADOW},
+   {"sampler2DRect", SLANG_SPEC_SAMPLER2DRECT},
+   {"sampler2DRectShadow", SLANG_SPEC_SAMPLER2DRECTSHADOW},
+   {NULL, SLANG_SPEC_VOID}
+};
+
+slang_type_specifier_type
+slang_type_specifier_type_from_string(const char *name)
+{
+   const type_specifier_type_name *p = type_specifier_type_names;
+   while (p->name != NULL) {
+      if (slang_string_compare(p->name, name) == 0)
+         break;
+      p++;
+   }
+   return p->type;
+}
+
+const char *
+slang_type_specifier_type_to_string(slang_type_specifier_type type)
+{
+   const type_specifier_type_name *p = type_specifier_type_names;
+   while (p->name != NULL) {
+      if (p->type == type)
+         break;
+      p++;
+   }
+   return p->name;
+}
+
+/* slang_fully_specified_type */
+
+int
+slang_fully_specified_type_construct(slang_fully_specified_type * type)
+{
+   type->qualifier = SLANG_QUAL_NONE;
+   slang_type_specifier_ctr(&type->specifier);
+   return 1;
+}
+
+void
+slang_fully_specified_type_destruct(slang_fully_specified_type * type)
+{
+   slang_type_specifier_dtr(&type->specifier);
+}
+
+int
+slang_fully_specified_type_copy(slang_fully_specified_type * x,
+                                const slang_fully_specified_type * y)
+{
+   slang_fully_specified_type z;
+
+   if (!slang_fully_specified_type_construct(&z))
+      return 0;
+   z.qualifier = y->qualifier;
+   z.precision = y->precision;
+   z.variant = y->variant;
+   z.centroid = y->centroid;
+   z.array_len = y->array_len;
+   if (!slang_type_specifier_copy(&z.specifier, &y->specifier)) {
+      slang_fully_specified_type_destruct(&z);
+      return 0;
+   }
+   slang_fully_specified_type_destruct(x);
+   *x = z;
+   return 1;
+}
+
+
+
 GLvoid
 slang_type_specifier_ctr(slang_type_specifier * self)
 {
@@ -187,6 +291,21 @@ slang_type_specifier_dtr(slang_type_specifier * self)
    }
 }
 
+slang_type_specifier *
+slang_type_specifier_new(slang_type_specifier_type type,
+                         struct slang_struct_ *_struct,
+                         struct slang_type_specifier_ *_array)
+{
+   slang_type_specifier *spec =
+      (slang_type_specifier *) _slang_alloc(sizeof(slang_type_specifier));
+   if (spec) {
+      spec->type = type;
+      spec->_struct = _struct;
+      spec->_array = _array;
+   }
+   return spec;
+}
+
 GLboolean
 slang_type_specifier_copy(slang_type_specifier * x,
                           const slang_type_specifier * y)
@@ -250,7 +369,7 @@ slang_type_specifier_equal(const slang_type_specifier * x,
 /**
  * As above, but allow float/int casting.
  */
-static GLboolean
+GLboolean
 slang_type_specifier_compatible(const slang_type_specifier * x,
                                 const slang_type_specifier * y)
 {
@@ -307,7 +426,7 @@ _slang_typeof_function(slang_atom a_name,
 {
    GLboolean error;
 
-   *funFound = _slang_locate_function(space->funcs, a_name, params,
+   *funFound = _slang_function_locate(space->funcs, a_name, params,
                                       num_params, space, atoms, log, &error);
    if (error)
       return GL_FALSE;
@@ -362,14 +481,6 @@ typeof_math_call(const char *name, slang_operation *call,
    }
 }
 
-GLboolean
-_slang_typeof_operation(const slang_assemble_ctx * A,
-                        slang_operation * op,
-                        slang_typeinfo * ti)
-{
-   return _slang_typeof_operation_(op, &A->space, ti, A->atoms, A->log);
-}
-
 
 /**
  * Determine the return type of an operation.
@@ -380,7 +491,7 @@ _slang_typeof_operation(const slang_assemble_ctx * A,
  * \return GL_TRUE for success, GL_FALSE if failure
  */
 GLboolean
-_slang_typeof_operation_(slang_operation * op,
+_slang_typeof_operation(slang_operation * op,
                          const slang_name_space * space,
                          slang_typeinfo * ti,
                          slang_atom_pool * atoms,
@@ -412,7 +523,7 @@ _slang_typeof_operation_(slang_operation * op,
    case SLANG_OPER_DIVASSIGN:
    case SLANG_OPER_PREINCREMENT:
    case SLANG_OPER_PREDECREMENT:
-      if (!_slang_typeof_operation_(op->children, space, ti, atoms, log))
+      if (!_slang_typeof_operation(op->children, space, ti, atoms, log))
          return GL_FALSE;
       break;
    case SLANG_OPER_LITERAL_BOOL:
@@ -479,7 +590,7 @@ _slang_typeof_operation_(slang_operation * op,
    case SLANG_OPER_VARIABLE_DECL:
       {
          slang_variable *var;
-         var = _slang_locate_variable(op->locals, op->a_id, GL_TRUE);
+         var = _slang_variable_locate(op->locals, op->a_id, GL_TRUE);
          if (!var) {
             slang_info_log_error(log, "undefined variable '%s'",
                                  (char *) op->a_id);
@@ -490,12 +601,20 @@ _slang_typeof_operation_(slang_operation * op,
             return GL_FALSE;
          }
          ti->can_be_referenced = GL_TRUE;
-         ti->array_len = var->array_len;
+         if (var->type.specifier.type == SLANG_SPEC_ARRAY &&
+             var->type.array_len >= 1) {
+            /* the datatype is an array, ex: float[3] x; */
+            ti->array_len = var->type.array_len;
+         }
+         else {
+            /* the variable is an array, ex: float x[3]; */
+            ti->array_len = var->array_len;
+         }
       }
       break;
    case SLANG_OPER_SEQUENCE:
       /* TODO: check [0] and [1] if they match */
-      if (!_slang_typeof_operation_(&op->children[1], space, ti, atoms, log)) {
+      if (!_slang_typeof_operation(&op->children[1], space, ti, atoms, log)) {
          return GL_FALSE;
       }
       ti->can_be_referenced = GL_FALSE;
@@ -509,7 +628,7 @@ _slang_typeof_operation_(slang_operation * op,
       /*case SLANG_OPER_ANDASSIGN: */
    case SLANG_OPER_SELECT:
       /* TODO: check [1] and [2] if they match */
-      if (!_slang_typeof_operation_(&op->children[1], space, ti, atoms, log)) {
+      if (!_slang_typeof_operation(&op->children[1], space, ti, atoms, log)) {
          return GL_FALSE;
       }
       ti->can_be_referenced = GL_FALSE;
@@ -542,7 +661,7 @@ _slang_typeof_operation_(slang_operation * op,
       break;
    /*case SLANG_OPER_MODULUS: */
    case SLANG_OPER_PLUS:
-      if (!_slang_typeof_operation_(op->children, space, ti, atoms, log))
+      if (!_slang_typeof_operation(op->children, space, ti, atoms, log))
          return GL_FALSE;
       ti->can_be_referenced = GL_FALSE;
       ti->is_swizzled = GL_FALSE;
@@ -559,7 +678,7 @@ _slang_typeof_operation_(slang_operation * op,
 
          if (!slang_typeinfo_construct(&_ti))
             return GL_FALSE;
-         if (!_slang_typeof_operation_(op->children, space, &_ti, atoms, log)) {
+         if (!_slang_typeof_operation(op->children, space, &_ti, atoms, log)) {
             slang_typeinfo_destruct(&_ti);
             return GL_FALSE;
          }
@@ -583,7 +702,18 @@ _slang_typeof_operation_(slang_operation * op,
       }
       break;
    case SLANG_OPER_CALL:
-      if (op->fun) {
+      if (op->array_constructor) {
+         /* build array typeinfo */
+         ti->spec.type = SLANG_SPEC_ARRAY;
+         ti->spec._array = (slang_type_specifier *)
+            _slang_alloc(sizeof(slang_type_specifier));
+         slang_type_specifier_ctr(ti->spec._array);
+
+         ti->spec._array->type =
+            slang_type_specifier_type_from_string((char *) op->a_id);
+         ti->array_len = op->num_children;
+      }
+      else if (op->fun) {
          /* we've resolved this call before */
          slang_type_specifier_copy(&ti->spec, &op->fun->header.type.specifier);
       }
@@ -630,20 +760,26 @@ _slang_typeof_operation_(slang_operation * op,
          }
       }
       break;
+   case SLANG_OPER_METHOD:
+      /* at this time, GLSL 1.20 only has one method: array.length()
+       * which returns an integer.
+       */
+      ti->spec.type = SLANG_SPEC_INT;
+      break;
    case SLANG_OPER_FIELD:
       {
          slang_typeinfo _ti;
 
          if (!slang_typeinfo_construct(&_ti))
             return GL_FALSE;
-         if (!_slang_typeof_operation_(op->children, space, &_ti, atoms, log)) {
+         if (!_slang_typeof_operation(op->children, space, &_ti, atoms, log)) {
             slang_typeinfo_destruct(&_ti);
             return GL_FALSE;
          }
          if (_ti.spec.type == SLANG_SPEC_STRUCT) {
             slang_variable *field;
 
-            field = _slang_locate_variable(_ti.spec._struct->fields, op->a_id,
+            field = _slang_variable_locate(_ti.spec._struct->fields, op->a_id,
                                            GL_FALSE);
             if (field == NULL) {
                slang_typeinfo_destruct(&_ti);
@@ -742,7 +878,7 @@ _slang_typeof_operation_(slang_operation * op,
       break;
    case SLANG_OPER_POSTINCREMENT:
    case SLANG_OPER_POSTDECREMENT:
-      if (!_slang_typeof_operation_(op->children, space, ti, atoms, log))
+      if (!_slang_typeof_operation(op->children, space, ti, atoms, log))
          return GL_FALSE;
       ti->can_be_referenced = GL_FALSE;
       ti->is_swizzled = GL_FALSE;
@@ -756,77 +892,6 @@ _slang_typeof_operation_(slang_operation * op,
 
 
 /**
- * Lookup a function according to name and parameter count/types.
- */
-slang_function *
-_slang_locate_function(const slang_function_scope * funcs, slang_atom a_name,
-                       slang_operation * args, GLuint num_args,
-                       const slang_name_space * space, slang_atom_pool * atoms,
-                       slang_info_log *log, GLboolean *error)
-{
-   slang_typeinfo arg_ti[100];
-   GLuint i;
-
-   *error = GL_FALSE;
-
-   /* determine type of each argument */
-   assert(num_args < 100);
-   for (i = 0; i < num_args; i++) {
-      if (!slang_typeinfo_construct(&arg_ti[i]))
-         return NULL;
-      if (!_slang_typeof_operation_(&args[i], space, &arg_ti[i], atoms, log)) {
-         return NULL;
-      }
-   }
-
-   /* loop over function scopes */
-   while (funcs) {
-
-      /* look for function with matching name and argument/param types */
-      for (i = 0; i < funcs->num_functions; i++) {
-         slang_function *f = &funcs->functions[i];
-         const GLuint haveRetValue = _slang_function_has_return_value(f);
-         GLuint j;
-
-         if (a_name != f->header.a_name)
-            continue;
-         if (f->param_count - haveRetValue != num_args)
-            continue;
-
-         /* compare parameter / argument types */
-         for (j = 0; j < num_args; j++) {
-            if (!slang_type_specifier_compatible(&arg_ti[j].spec,
-                              &f->parameters->variables[j]->type.specifier)) {
-               /* param/arg types don't match */
-               break;
-            }
-
-            /* "out" and "inout" formal parameter requires the actual
-             * argument to be an l-value.
-             */
-            if (!arg_ti[j].can_be_referenced &&
-                (f->parameters->variables[j]->type.qualifier == SLANG_QUAL_OUT ||
-                 f->parameters->variables[j]->type.qualifier == SLANG_QUAL_INOUT)) {
-               /* param is not an lvalue! */
-               *error = GL_TRUE;
-               return NULL;
-            }
-         }
-
-         if (j == num_args) {
-            /* name and args match! */
-            return f;
-         }
-      }
-
-      funcs = funcs->outer_scope;
-   }
-
-   return NULL;
-}
-
-
-/**
  * Determine if a type is a matrix.
  * \return GL_TRUE if is a matrix, GL_FALSE otherwise.
  */
diff --git a/src/mesa/shader/slang/slang_typeinfo.h b/src/mesa/shader/slang/slang_typeinfo.h
index b3ad06b65c..e6fecd350a 100644
--- a/src/mesa/shader/slang/slang_typeinfo.h
+++ b/src/mesa/shader/slang/slang_typeinfo.h
@@ -34,6 +34,9 @@
 
 struct slang_operation_;
 
+struct slang_name_space_;
+
+
 
 /**
  * Holds complete information about vector swizzle - the <swizzle>
@@ -47,45 +50,45 @@ typedef struct slang_swizzle_
    GLuint swizzle[4];
 } slang_swizzle;
 
-typedef struct slang_name_space_
-{
-   struct slang_function_scope_ *funcs;
-   struct slang_struct_scope_ *structs;
-   struct slang_variable_scope_ *vars;
-} slang_name_space;
+extern GLboolean
+_slang_is_swizzle(const char *field, GLuint rows, slang_swizzle *swz);
 
 
-typedef struct slang_assemble_ctx_
+typedef enum slang_type_variant_
 {
-   slang_atom_pool *atoms;
-   slang_name_space space;
-   struct gl_program *program;
-   slang_var_table *vartable;
-   slang_info_log *log;
-   struct slang_label_ *curFuncEndLabel;
-   struct slang_ir_node_ *CurLoop;
-   struct slang_function_ *CurFunction;
-} slang_assemble_ctx;
-
-
-extern struct slang_function_ *
-_slang_locate_function(const struct slang_function_scope_ *funcs,
-                       slang_atom name, struct slang_operation_ *params,
-                       GLuint num_params,
-                       const slang_name_space *space,
-                       slang_atom_pool *atoms, slang_info_log *log,
-                       GLboolean *error);
+   SLANG_VARIANT,    /* the default */
+   SLANG_INVARIANT   /* indicates the "invariant" keyword */
+} slang_type_variant;
 
 
-extern GLboolean
-_slang_is_swizzle(const char *field, GLuint rows, slang_swizzle *swz);
+typedef enum slang_type_centroid_
+{
+   SLANG_CENTER,    /* the default */
+   SLANG_CENTROID   /* indicates the "centroid" keyword */
+} slang_type_centroid;
 
-extern GLboolean
-_slang_is_swizzle_mask(const slang_swizzle *swz, GLuint rows);
 
-extern GLvoid
-_slang_multiply_swizzles(slang_swizzle *, const slang_swizzle *,
-                         const slang_swizzle *);
+typedef enum slang_type_qualifier_
+{
+   SLANG_QUAL_NONE,
+   SLANG_QUAL_CONST,
+   SLANG_QUAL_ATTRIBUTE,
+   SLANG_QUAL_VARYING,
+   SLANG_QUAL_UNIFORM,
+   SLANG_QUAL_OUT,
+   SLANG_QUAL_INOUT,
+   SLANG_QUAL_FIXEDOUTPUT,      /* internal */
+   SLANG_QUAL_FIXEDINPUT        /* internal */
+} slang_type_qualifier;
+
+
+typedef enum slang_type_precision_
+{
+   SLANG_PREC_DEFAULT,
+   SLANG_PREC_LOW,
+   SLANG_PREC_MEDIUM,
+   SLANG_PREC_HIGH
+} slang_type_precision;
 
 
 /**
@@ -128,14 +131,21 @@ typedef enum slang_type_specifier_type_
 } slang_type_specifier_type;
 
 
+extern slang_type_specifier_type
+slang_type_specifier_type_from_string(const char *);
+
+extern const char *
+slang_type_specifier_type_to_string(slang_type_specifier_type);
+
+
 /**
  * Describes more sophisticated types, like structs and arrays.
  */
 typedef struct slang_type_specifier_
 {
    slang_type_specifier_type type;
-   struct slang_struct_ *_struct;         /**< used if type == spec_struct */
-   struct slang_type_specifier_ *_array;  /**< used if type == spec_array */
+   struct slang_struct_ *_struct;         /**< if type == SLANG_SPEC_STRUCT */
+   struct slang_type_specifier_ *_array;  /**< if type == SLANG_SPEC_ARRAY */
 } slang_type_specifier;
 
 
@@ -145,6 +155,12 @@ slang_type_specifier_ctr(slang_type_specifier *);
 extern GLvoid
 slang_type_specifier_dtr(slang_type_specifier *);
 
+extern slang_type_specifier *
+slang_type_specifier_new(slang_type_specifier_type type,
+                         struct slang_struct_ *_struct,
+                         struct slang_type_specifier_ *_array);
+
+
 extern GLboolean
 slang_type_specifier_copy(slang_type_specifier *, const slang_type_specifier *);
 
@@ -153,6 +169,33 @@ slang_type_specifier_equal(const slang_type_specifier *,
                            const slang_type_specifier *);
 
 
+extern GLboolean
+slang_type_specifier_compatible(const slang_type_specifier * x,
+                                const slang_type_specifier * y);
+
+
+typedef struct slang_fully_specified_type_
+{
+   slang_type_qualifier qualifier;
+   slang_type_specifier specifier;
+   slang_type_precision precision;
+   slang_type_variant variant;
+   slang_type_centroid centroid;
+   GLint array_len;           /**< -1 if not an array type */
+} slang_fully_specified_type;
+
+extern int
+slang_fully_specified_type_construct(slang_fully_specified_type *);
+
+extern void
+slang_fully_specified_type_destruct(slang_fully_specified_type *);
+
+extern int
+slang_fully_specified_type_copy(slang_fully_specified_type *,
+				const slang_fully_specified_type *);
+
+
+
 typedef struct slang_typeinfo_
 {
    GLboolean can_be_referenced;
@@ -169,19 +212,9 @@ extern GLvoid
 slang_typeinfo_destruct(slang_typeinfo *);
 
 
-/**
- * Retrieves type information about an operation.
- * Returns GL_TRUE on success.
- * Returns GL_FALSE otherwise.
- */
-extern GLboolean
-_slang_typeof_operation(const slang_assemble_ctx *,
-                        struct slang_operation_ *,
-                        slang_typeinfo *);
-
 extern GLboolean
-_slang_typeof_operation_(struct slang_operation_ *,
-                         const slang_name_space *,
+_slang_typeof_operation(struct slang_operation_ *,
+                         const struct slang_name_space_ *,
                          slang_typeinfo *, slang_atom_pool *,
                          slang_info_log *log);
 
diff --git a/src/mesa/shader/slang/slang_vartable.c b/src/mesa/shader/slang/slang_vartable.c
index 95971a70a9..a4ebacc093 100644
--- a/src/mesa/shader/slang/slang_vartable.c
+++ b/src/mesa/shader/slang/slang_vartable.c
@@ -4,6 +4,7 @@
 #include "shader/prog_print.h"
 #include "slang_compile.h"
 #include "slang_compile_variable.h"
+#include "slang_emit.h"
 #include "slang_mem.h"
 #include "slang_vartable.h"
 #include "slang_ir.h"
@@ -72,9 +73,8 @@ _slang_delete_var_table(slang_var_table *vt)
 
 
 /**
- * Create new table, put at head, return ptr to it.
- * XXX we should take a maxTemps parameter to indicate how many temporaries
- * are available for the current shader/program target.
+ * Create new table on top of vartable stack.
+ * Used when we enter a {} block.
  */
 void
 _slang_push_var_table(slang_var_table *vt)
@@ -95,7 +95,8 @@ _slang_push_var_table(slang_var_table *vt)
 
 
 /**
- * Destroy given table, return ptr to Parent
+ * Pop top entry from variable table.
+ * Used when we leave a {} block.
  */
 void
 _slang_pop_var_table(slang_var_table *vt)
@@ -107,7 +108,7 @@ _slang_pop_var_table(slang_var_table *vt)
 
    /* free the storage allocated for each variable */
    for (i = 0; i < t->NumVars; i++) {
-      slang_ir_storage *store = (slang_ir_storage *) t->Vars[i]->aux;
+      slang_ir_storage *store = t->Vars[i]->store;
       GLint j;
       GLuint comp;
       if (dbg) printf("  Free var %s, size %d at %d.%s\n",
@@ -125,10 +126,12 @@ _slang_pop_var_table(slang_var_table *vt)
       else
          comp = 0;
 
-      assert(store->Index >= 0);
-      for (j = 0; j < store->Size; j++) {
-         assert(t->Temps[store->Index * 4 + j + comp] == VAR);
-         t->Temps[store->Index * 4 + j + comp] = FREE;
+      /* store->Index may be -1 if we run out of registers */
+      if (store->Index >= 0) {
+         for (j = 0; j < store->Size; j++) {
+            assert(t->Temps[store->Index * 4 + j + comp] == VAR);
+            t->Temps[store->Index * 4 + j + comp] = FREE;
+         }
       }
       store->Index = -1;
    }
@@ -156,7 +159,7 @@ _slang_pop_var_table(slang_var_table *vt)
 
 
 /**
- * Add a new variable to the given symbol table.
+ * Add a new variable to the given var/symbol table.
  */
 void
 _slang_add_variable(slang_var_table *vt, slang_variable *v)
@@ -165,7 +168,7 @@ _slang_add_variable(slang_var_table *vt, slang_variable *v)
    assert(vt);
    t = vt->Top;
    assert(t);
-   if (dbg) printf("Adding var %s, store %p\n", (char *) v->a_name, v->aux);
+   if (dbg) printf("Adding var %s, store %p\n", (char *) v->a_name, (void *) v->store);
    t->Vars = (slang_variable **)
       _slang_realloc(t->Vars,
                      t->NumVars * sizeof(slang_variable *),
@@ -214,6 +217,7 @@ alloc_reg(slang_var_table *vt, GLint size, GLboolean isTemp)
    for (i = 0; i <= vt->MaxRegisters * 4 - size; i += step) {
       GLuint found = 0;
       for (j = 0; j < (GLuint) size; j++) {
+         assert(i + j < 4 * MAX_PROGRAM_TEMPS);
          if (i + j < vt->MaxRegisters * 4 && t->Temps[i + j] == FREE) {
             found++;
          }
@@ -225,13 +229,17 @@ alloc_reg(slang_var_table *vt, GLint size, GLboolean isTemp)
          /* found block of size free regs */
          if (size > 1)
             assert(i % 4 == 0);
-         for (j = 0; j < (GLuint) size; j++)
+         for (j = 0; j < (GLuint) size; j++) {
+            assert(i + j < 4 * MAX_PROGRAM_TEMPS);
             t->Temps[i + j] = isTemp ? TEMP : VAR;
+         }
          assert(i < MAX_PROGRAM_TEMPS * 4);
          t->ValSize[i] = size;
          return i;
       }
    }
+
+   /* if we get here, we ran out of registers */
    return -1;
 }
 
@@ -259,21 +267,7 @@ _slang_alloc_var(slang_var_table *vt, slang_ir_storage *store)
       return GL_FALSE;
 
    store->Index = i / 4;
-   if (store->Size == 1) {
-      const GLuint comp = i % 4;
-      store->Swizzle = MAKE_SWIZZLE4(comp, comp, comp, comp);
-   }
-   else if (store->Size == 2) {
-      store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y,
-                                     SWIZZLE_NIL, SWIZZLE_NIL);
-   }
-   else if (store->Size == 3) {
-      store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y,
-                                     SWIZZLE_Z, SWIZZLE_NIL);
-   }
-   else {
-      store->Swizzle = SWIZZLE_NOOP;
-   }
+   store->Swizzle = _slang_var_swizzle(store->Size, i % 4);
 
    if (dbg)
       printf("Alloc var storage sz %d at %d.%s (level %d) store %p\n",
@@ -301,20 +295,7 @@ _slang_alloc_temp(slang_var_table *vt, slang_ir_storage *store)
    assert(store->Index < 0);
 
    store->Index = i / 4;
-   if (store->Size == 1) {
-      const GLuint comp = i % 4;
-      store->Swizzle = MAKE_SWIZZLE4(comp, comp, comp, comp);
-   }
-   else {
-      /* XXX improve swizzled for size=2/3, use for writemask... */
-#if 1
-      if (store->Size == 2) {
-         store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y,
-                                        SWIZZLE_NIL, SWIZZLE_NIL);
-      }
-#endif
-      store->Swizzle = SWIZZLE_NOOP;
-   }
+   store->Swizzle = _slang_var_swizzle(store->Size, i % 4);
 
    if (dbg) printf("Alloc temp sz %d at %d.%s (level %d) store %p\n",
                    store->Size, store->Index,
diff --git a/src/mesa/shader/slang/slang_vartable.h b/src/mesa/shader/slang/slang_vartable.h
index 8a3b992c96..94bcd63f45 100644
--- a/src/mesa/shader/slang/slang_vartable.h
+++ b/src/mesa/shader/slang/slang_vartable.h
@@ -2,7 +2,7 @@
 #ifndef SLANG_VARTABLE_H
 #define SLANG_VARTABLE_H
 
-struct _slang_ir_storage;
+struct slang_ir_storage_;
 
 typedef struct slang_var_table_ slang_var_table;
 
@@ -27,16 +27,16 @@ extern struct slang_variable_ *
 _slang_find_variable(const slang_var_table *t, slang_atom name);
 
 extern GLboolean
-_slang_alloc_var(slang_var_table *t, struct _slang_ir_storage *store);
+_slang_alloc_var(slang_var_table *t, struct slang_ir_storage_ *store);
 
 extern GLboolean
-_slang_alloc_temp(slang_var_table *t, struct _slang_ir_storage *store);
+_slang_alloc_temp(slang_var_table *t, struct slang_ir_storage_ *store);
 
 extern void
-_slang_free_temp(slang_var_table *t, struct _slang_ir_storage *store);
+_slang_free_temp(slang_var_table *t, struct slang_ir_storage_ *store);
 
 extern GLboolean
-_slang_is_temp(const slang_var_table *t, const struct _slang_ir_storage *store);
+_slang_is_temp(const slang_var_table *t, const struct slang_ir_storage_ *store);
 
 
 #endif /* SLANG_VARTABLE_H */
diff --git a/src/mesa/sources b/src/mesa/sources.mak
index 9b9eb991fd..1bd81506e0 100644
--- a/src/mesa/sources
+++ b/src/mesa/sources.mak
@@ -21,6 +21,7 @@ MAIN_SOURCES = \
 	main/depth.c \
 	main/depthstencil.c \
 	main/dlist.c \
+	main/dlopen.c \
 	main/drawpix.c \
 	main/enable.c \
 	main/enums.c \
@@ -87,7 +88,10 @@ MATH_SOURCES = \
 	math/m_eval.c \
 	math/m_matrix.c \
 	math/m_translate.c \
-	math/m_vector.c \
+	math/m_vector.c
+
+MATH_XFORM_SOURCES = \
+	$(MATH_SOURCES) \
 	math/m_xform.c
 
 SWRAST_SOURCES = \
@@ -161,6 +165,54 @@ VBO_SOURCES = \
 	vbo/vbo_save_draw.c \
 	vbo/vbo_save_loopback.c 
 
+VF_SOURCES = \
+	vf/vf.c \
+	vf/vf_generic.c \
+	vf/vf_sse.c
+
+STATETRACKER_SOURCES = \
+	state_tracker/st_atom.c \
+	state_tracker/st_atom_blend.c \
+	state_tracker/st_atom_clip.c \
+	state_tracker/st_atom_constbuf.c \
+	state_tracker/st_atom_depth.c \
+	state_tracker/st_atom_framebuffer.c \
+	state_tracker/st_atom_pixeltransfer.c \
+	state_tracker/st_atom_sampler.c \
+	state_tracker/st_atom_scissor.c \
+	state_tracker/st_atom_shader.c \
+	state_tracker/st_atom_rasterizer.c \
+	state_tracker/st_atom_stipple.c \
+	state_tracker/st_atom_texture.c \
+	state_tracker/st_atom_viewport.c \
+	state_tracker/st_cb_accum.c \
+	state_tracker/st_cb_bitmap.c \
+	state_tracker/st_cb_blit.c \
+	state_tracker/st_cb_bufferobjects.c \
+	state_tracker/st_cb_clear.c \
+	state_tracker/st_cb_flush.c \
+	state_tracker/st_cb_get.c \
+	state_tracker/st_cb_drawpixels.c \
+	state_tracker/st_cb_fbo.c \
+	state_tracker/st_cb_feedback.c \
+	state_tracker/st_cb_program.c \
+	state_tracker/st_cb_queryobj.c \
+	state_tracker/st_cb_rasterpos.c \
+	state_tracker/st_cb_readpixels.c \
+	state_tracker/st_cb_strings.c \
+	state_tracker/st_cb_texture.c \
+	state_tracker/st_api.c \
+	state_tracker/st_context.c \
+	state_tracker/st_debug.c \
+	state_tracker/st_draw.c \
+	state_tracker/st_draw_feedback.c \
+	state_tracker/st_extensions.c \
+	state_tracker/st_format.c \
+	state_tracker/st_framebuffer.c \
+	state_tracker/st_gen_mipmap.c \
+	state_tracker/st_mesa_to_tgsi.c \
+	state_tracker/st_program.c \
+	state_tracker/st_texture.c
 
 SHADER_SOURCES = \
 	shader/arbprogparse.c \
@@ -175,6 +227,7 @@ SHADER_SOURCES = \
 	shader/prog_debug.c \
 	shader/prog_execute.c \
 	shader/prog_instruction.c \
+	shader/prog_noise.c \
 	shader/prog_parameter.c \
 	shader/prog_print.c \
 	shader/prog_statevars.c \
@@ -193,7 +246,6 @@ SLANG_SOURCES =	\
 	shader/slang/slang_emit.c	\
 	shader/slang/slang_ir.c	\
 	shader/slang/slang_label.c	\
-	shader/slang/slang_library_noise.c	\
 	shader/slang/slang_link.c	\
 	shader/slang/slang_log.c	\
 	shader/slang/slang_mem.c	\
@@ -260,33 +312,43 @@ COMMON_DRIVER_SOURCES =			\
 
 MESA_SOURCES = \
 	$(MAIN_SOURCES)		\
-	$(MATH_SOURCES)		\
+	$(MATH_XFORM_SOURCES)	\
 	$(VBO_SOURCES)		\
 	$(TNL_SOURCES)		\
 	$(SHADER_SOURCES)	\
 	$(SWRAST_SOURCES)	\
 	$(SWRAST_SETUP_SOURCES)	\
-	$(COMMON_DRIVER_SOURCES) \
+	$(COMMON_DRIVER_SOURCES)\
 	$(ASM_C_SOURCES)	\
 	$(SLANG_SOURCES)
 
 ALL_SOURCES = \
 	$(MESA_SOURCES)		\
 	$(GLAPI_SOURCES)	\
-	$(ASM_SOURCES)		\
-	$(COMMON_DRIVER_SOURCES)
+	$(MESA_ASM_SOURCES)
 
+MESA_GALLIUM_SOURCES = \
+	$(MAIN_SOURCES)		\
+	$(MATH_SOURCES)		\
+	$(VBO_SOURCES)		\
+	$(STATETRACKER_SOURCES)	\
+	$(SHADER_SOURCES)	\
+	$(SLANG_SOURCES)
 
 
 ### Object files
 
 MESA_OBJECTS = \
 	$(MESA_SOURCES:.c=.o) \
-	$(ASM_SOURCES:.S=.o)
+	$(MESA_ASM_SOURCES:.S=.o)
+
+MESA_GALLIUM_OBJECTS = \
+	$(MESA_GALLIUM_SOURCES:.c=.o) \
+	$(MESA_ASM_SOURCES:.S=.o)
 
 GLAPI_OBJECTS = \
 	$(GLAPI_SOURCES:.c=.o) \
-	$(ASM_API:.S=.o)
+	$(GLAPI_ASM_SOURCES:.S=.o)
 
 
 COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o)
@@ -294,11 +356,10 @@ COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o)
 
 
 ### Include directories
-### XXX we should be able to trim this down to just -Iinclude/ and -Isrc/mesa/
-### since most #includes have been changed from #include "mtypes.h" to
-### #include "main/mtypes.h", etc.
 
 INCLUDE_DIRS = \
 	-I$(TOP)/include \
 	-I$(TOP)/src/mesa \
-	-I$(TOP)/src/mesa/main
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/drivers \
+	-I$(TOP)/src/gallium/auxiliary
diff --git a/src/mesa/sparc/glapi_sparc.S b/src/mesa/sparc/glapi_sparc.S
index 3d1d55fd76..7d215939b4 100644
--- a/src/mesa/sparc/glapi_sparc.S
+++ b/src/mesa/sparc/glapi_sparc.S
@@ -626,8 +626,8 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */
 		.globl glGetActiveAttribARB ; .type glGetActiveAttribARB,#function
 		.globl glGetAttribLocationARB ; .type glGetAttribLocationARB,#function
 		.globl glDrawBuffersARB ; .type glDrawBuffersARB,#function
+		.globl glRenderbufferStorageMultisample ; .type glRenderbufferStorageMultisample,#function
 		.globl glPolygonOffsetEXT ; .type glPolygonOffsetEXT,#function
-		.globl gl_dispatch_stub_562 ; .type gl_dispatch_stub_562,#function
 		.globl gl_dispatch_stub_563 ; .type gl_dispatch_stub_563,#function
 		.globl gl_dispatch_stub_564 ; .type gl_dispatch_stub_564,#function
 		.globl gl_dispatch_stub_565 ; .type gl_dispatch_stub_565,#function
@@ -635,6 +635,7 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */
 		.globl gl_dispatch_stub_567 ; .type gl_dispatch_stub_567,#function
 		.globl gl_dispatch_stub_568 ; .type gl_dispatch_stub_568,#function
 		.globl gl_dispatch_stub_569 ; .type gl_dispatch_stub_569,#function
+		.globl gl_dispatch_stub_570 ; .type gl_dispatch_stub_570,#function
 		.globl glColorPointerEXT ; .type glColorPointerEXT,#function
 		.globl glEdgeFlagPointerEXT ; .type glEdgeFlagPointerEXT,#function
 		.globl glIndexPointerEXT ; .type glIndexPointerEXT,#function
@@ -645,8 +646,8 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */
 		.globl glPointParameterfvEXT ; .type glPointParameterfvEXT,#function
 		.globl glLockArraysEXT ; .type glLockArraysEXT,#function
 		.globl glUnlockArraysEXT ; .type glUnlockArraysEXT,#function
-		.globl gl_dispatch_stub_580 ; .type gl_dispatch_stub_580,#function
 		.globl gl_dispatch_stub_581 ; .type gl_dispatch_stub_581,#function
+		.globl gl_dispatch_stub_582 ; .type gl_dispatch_stub_582,#function
 		.globl glSecondaryColor3bEXT ; .type glSecondaryColor3bEXT,#function
 		.globl glSecondaryColor3bvEXT ; .type glSecondaryColor3bvEXT,#function
 		.globl glSecondaryColor3dEXT ; .type glSecondaryColor3dEXT,#function
@@ -671,7 +672,7 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */
 		.globl glFogCoorddvEXT ; .type glFogCoorddvEXT,#function
 		.globl glFogCoordfEXT ; .type glFogCoordfEXT,#function
 		.globl glFogCoordfvEXT ; .type glFogCoordfvEXT,#function
-		.globl gl_dispatch_stub_606 ; .type gl_dispatch_stub_606,#function
+		.globl gl_dispatch_stub_607 ; .type gl_dispatch_stub_607,#function
 		.globl glBlendFuncSeparateEXT ; .type glBlendFuncSeparateEXT,#function
 		.globl glFlushVertexArrayRangeNV ; .type glFlushVertexArrayRangeNV,#function
 		.globl glVertexArrayRangeNV ; .type glVertexArrayRangeNV,#function
@@ -713,7 +714,6 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */
 		.globl glWindowPos4ivMESA ; .type glWindowPos4ivMESA,#function
 		.globl glWindowPos4sMESA ; .type glWindowPos4sMESA,#function
 		.globl glWindowPos4svMESA ; .type glWindowPos4svMESA,#function
-		.globl gl_dispatch_stub_648 ; .type gl_dispatch_stub_648,#function
 		.globl gl_dispatch_stub_649 ; .type gl_dispatch_stub_649,#function
 		.globl gl_dispatch_stub_650 ; .type gl_dispatch_stub_650,#function
 		.globl gl_dispatch_stub_651 ; .type gl_dispatch_stub_651,#function
@@ -722,6 +722,7 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */
 		.globl gl_dispatch_stub_654 ; .type gl_dispatch_stub_654,#function
 		.globl gl_dispatch_stub_655 ; .type gl_dispatch_stub_655,#function
 		.globl gl_dispatch_stub_656 ; .type gl_dispatch_stub_656,#function
+		.globl gl_dispatch_stub_657 ; .type gl_dispatch_stub_657,#function
 		.globl glAreProgramsResidentNV ; .type glAreProgramsResidentNV,#function
 		.globl glBindProgramNV ; .type glBindProgramNV,#function
 		.globl glDeleteProgramsNV ; .type glDeleteProgramsNV,#function
@@ -798,19 +799,19 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */
 		.globl glSetFragmentShaderConstantATI ; .type glSetFragmentShaderConstantATI,#function
 		.globl glPointParameteriNV ; .type glPointParameteriNV,#function
 		.globl glPointParameterivNV ; .type glPointParameterivNV,#function
-		.globl gl_dispatch_stub_733 ; .type gl_dispatch_stub_733,#function
 		.globl gl_dispatch_stub_734 ; .type gl_dispatch_stub_734,#function
 		.globl gl_dispatch_stub_735 ; .type gl_dispatch_stub_735,#function
 		.globl gl_dispatch_stub_736 ; .type gl_dispatch_stub_736,#function
 		.globl gl_dispatch_stub_737 ; .type gl_dispatch_stub_737,#function
+		.globl gl_dispatch_stub_738 ; .type gl_dispatch_stub_738,#function
 		.globl glGetProgramNamedParameterdvNV ; .type glGetProgramNamedParameterdvNV,#function
 		.globl glGetProgramNamedParameterfvNV ; .type glGetProgramNamedParameterfvNV,#function
 		.globl glProgramNamedParameter4dNV ; .type glProgramNamedParameter4dNV,#function
 		.globl glProgramNamedParameter4dvNV ; .type glProgramNamedParameter4dvNV,#function
 		.globl glProgramNamedParameter4fNV ; .type glProgramNamedParameter4fNV,#function
 		.globl glProgramNamedParameter4fvNV ; .type glProgramNamedParameter4fvNV,#function
-		.globl gl_dispatch_stub_744 ; .type gl_dispatch_stub_744,#function
 		.globl gl_dispatch_stub_745 ; .type gl_dispatch_stub_745,#function
+		.globl gl_dispatch_stub_746 ; .type gl_dispatch_stub_746,#function
 		.globl glBindFramebufferEXT ; .type glBindFramebufferEXT,#function
 		.globl glBindRenderbufferEXT ; .type glBindRenderbufferEXT,#function
 		.globl glCheckFramebufferStatusEXT ; .type glCheckFramebufferStatusEXT,#function
@@ -828,13 +829,13 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */
 		.globl glIsFramebufferEXT ; .type glIsFramebufferEXT,#function
 		.globl glIsRenderbufferEXT ; .type glIsRenderbufferEXT,#function
 		.globl glRenderbufferStorageEXT ; .type glRenderbufferStorageEXT,#function
-		.globl gl_dispatch_stub_763 ; .type gl_dispatch_stub_763,#function
+		.globl gl_dispatch_stub_764 ; .type gl_dispatch_stub_764,#function
 		.globl glFramebufferTextureLayerEXT ; .type glFramebufferTextureLayerEXT,#function
-		.globl gl_dispatch_stub_765 ; .type gl_dispatch_stub_765,#function
 		.globl gl_dispatch_stub_766 ; .type gl_dispatch_stub_766,#function
 		.globl gl_dispatch_stub_767 ; .type gl_dispatch_stub_767,#function
 		.globl gl_dispatch_stub_768 ; .type gl_dispatch_stub_768,#function
 		.globl gl_dispatch_stub_769 ; .type gl_dispatch_stub_769,#function
+		.globl gl_dispatch_stub_770 ; .type gl_dispatch_stub_770,#function
 		.globl _mesa_sparc_glapi_begin ; .type _mesa_sparc_glapi_begin,#function
 _mesa_sparc_glapi_begin:
 
@@ -1399,8 +1400,8 @@ _mesa_sparc_glapi_begin:
 	GL_STUB(glGetActiveAttribARB, _gloffset_GetActiveAttribARB)
 	GL_STUB(glGetAttribLocationARB, _gloffset_GetAttribLocationARB)
 	GL_STUB(glDrawBuffersARB, _gloffset_DrawBuffersARB)
+	GL_STUB(glRenderbufferStorageMultisample, _gloffset_RenderbufferStorageMultisample)
 	GL_STUB(glPolygonOffsetEXT, _gloffset_PolygonOffsetEXT)
-	GL_STUB(gl_dispatch_stub_562, _gloffset__dispatch_stub_562)
 	GL_STUB(gl_dispatch_stub_563, _gloffset__dispatch_stub_563)
 	GL_STUB(gl_dispatch_stub_564, _gloffset__dispatch_stub_564)
 	GL_STUB(gl_dispatch_stub_565, _gloffset__dispatch_stub_565)
@@ -1408,6 +1409,7 @@ _mesa_sparc_glapi_begin:
 	GL_STUB(gl_dispatch_stub_567, _gloffset__dispatch_stub_567)
 	GL_STUB(gl_dispatch_stub_568, _gloffset__dispatch_stub_568)
 	GL_STUB(gl_dispatch_stub_569, _gloffset__dispatch_stub_569)
+	GL_STUB(gl_dispatch_stub_570, _gloffset__dispatch_stub_570)
 	GL_STUB(glColorPointerEXT, _gloffset_ColorPointerEXT)
 	GL_STUB(glEdgeFlagPointerEXT, _gloffset_EdgeFlagPointerEXT)
 	GL_STUB(glIndexPointerEXT, _gloffset_IndexPointerEXT)
@@ -1418,8 +1420,8 @@ _mesa_sparc_glapi_begin:
 	GL_STUB(glPointParameterfvEXT, _gloffset_PointParameterfvEXT)
 	GL_STUB(glLockArraysEXT, _gloffset_LockArraysEXT)
 	GL_STUB(glUnlockArraysEXT, _gloffset_UnlockArraysEXT)
-	GL_STUB(gl_dispatch_stub_580, _gloffset__dispatch_stub_580)
 	GL_STUB(gl_dispatch_stub_581, _gloffset__dispatch_stub_581)
+	GL_STUB(gl_dispatch_stub_582, _gloffset__dispatch_stub_582)
 	GL_STUB(glSecondaryColor3bEXT, _gloffset_SecondaryColor3bEXT)
 	GL_STUB(glSecondaryColor3bvEXT, _gloffset_SecondaryColor3bvEXT)
 	GL_STUB(glSecondaryColor3dEXT, _gloffset_SecondaryColor3dEXT)
@@ -1444,7 +1446,7 @@ _mesa_sparc_glapi_begin:
 	GL_STUB(glFogCoorddvEXT, _gloffset_FogCoorddvEXT)
 	GL_STUB(glFogCoordfEXT, _gloffset_FogCoordfEXT)
 	GL_STUB(glFogCoordfvEXT, _gloffset_FogCoordfvEXT)
-	GL_STUB(gl_dispatch_stub_606, _gloffset__dispatch_stub_606)
+	GL_STUB(gl_dispatch_stub_607, _gloffset__dispatch_stub_607)
 	GL_STUB(glBlendFuncSeparateEXT, _gloffset_BlendFuncSeparateEXT)
 	GL_STUB(glFlushVertexArrayRangeNV, _gloffset_FlushVertexArrayRangeNV)
 	GL_STUB(glVertexArrayRangeNV, _gloffset_VertexArrayRangeNV)
@@ -1486,7 +1488,6 @@ _mesa_sparc_glapi_begin:
 	GL_STUB(glWindowPos4ivMESA, _gloffset_WindowPos4ivMESA)
 	GL_STUB(glWindowPos4sMESA, _gloffset_WindowPos4sMESA)
 	GL_STUB(glWindowPos4svMESA, _gloffset_WindowPos4svMESA)
-	GL_STUB(gl_dispatch_stub_648, _gloffset__dispatch_stub_648)
 	GL_STUB(gl_dispatch_stub_649, _gloffset__dispatch_stub_649)
 	GL_STUB(gl_dispatch_stub_650, _gloffset__dispatch_stub_650)
 	GL_STUB(gl_dispatch_stub_651, _gloffset__dispatch_stub_651)
@@ -1495,6 +1496,7 @@ _mesa_sparc_glapi_begin:
 	GL_STUB(gl_dispatch_stub_654, _gloffset__dispatch_stub_654)
 	GL_STUB(gl_dispatch_stub_655, _gloffset__dispatch_stub_655)
 	GL_STUB(gl_dispatch_stub_656, _gloffset__dispatch_stub_656)
+	GL_STUB(gl_dispatch_stub_657, _gloffset__dispatch_stub_657)
 	GL_STUB(glAreProgramsResidentNV, _gloffset_AreProgramsResidentNV)
 	GL_STUB(glBindProgramNV, _gloffset_BindProgramNV)
 	GL_STUB(glDeleteProgramsNV, _gloffset_DeleteProgramsNV)
@@ -1571,19 +1573,19 @@ _mesa_sparc_glapi_begin:
 	GL_STUB(glSetFragmentShaderConstantATI, _gloffset_SetFragmentShaderConstantATI)
 	GL_STUB(glPointParameteriNV, _gloffset_PointParameteriNV)
 	GL_STUB(glPointParameterivNV, _gloffset_PointParameterivNV)
-	GL_STUB(gl_dispatch_stub_733, _gloffset__dispatch_stub_733)
 	GL_STUB(gl_dispatch_stub_734, _gloffset__dispatch_stub_734)
 	GL_STUB(gl_dispatch_stub_735, _gloffset__dispatch_stub_735)
 	GL_STUB(gl_dispatch_stub_736, _gloffset__dispatch_stub_736)
 	GL_STUB(gl_dispatch_stub_737, _gloffset__dispatch_stub_737)
+	GL_STUB(gl_dispatch_stub_738, _gloffset__dispatch_stub_738)
 	GL_STUB(glGetProgramNamedParameterdvNV, _gloffset_GetProgramNamedParameterdvNV)
 	GL_STUB(glGetProgramNamedParameterfvNV, _gloffset_GetProgramNamedParameterfvNV)
 	GL_STUB(glProgramNamedParameter4dNV, _gloffset_ProgramNamedParameter4dNV)
 	GL_STUB(glProgramNamedParameter4dvNV, _gloffset_ProgramNamedParameter4dvNV)
 	GL_STUB(glProgramNamedParameter4fNV, _gloffset_ProgramNamedParameter4fNV)
 	GL_STUB(glProgramNamedParameter4fvNV, _gloffset_ProgramNamedParameter4fvNV)
-	GL_STUB(gl_dispatch_stub_744, _gloffset__dispatch_stub_744)
 	GL_STUB(gl_dispatch_stub_745, _gloffset__dispatch_stub_745)
+	GL_STUB(gl_dispatch_stub_746, _gloffset__dispatch_stub_746)
 	GL_STUB(glBindFramebufferEXT, _gloffset_BindFramebufferEXT)
 	GL_STUB(glBindRenderbufferEXT, _gloffset_BindRenderbufferEXT)
 	GL_STUB(glCheckFramebufferStatusEXT, _gloffset_CheckFramebufferStatusEXT)
@@ -1601,13 +1603,13 @@ _mesa_sparc_glapi_begin:
 	GL_STUB(glIsFramebufferEXT, _gloffset_IsFramebufferEXT)
 	GL_STUB(glIsRenderbufferEXT, _gloffset_IsRenderbufferEXT)
 	GL_STUB(glRenderbufferStorageEXT, _gloffset_RenderbufferStorageEXT)
-	GL_STUB(gl_dispatch_stub_763, _gloffset__dispatch_stub_763)
+	GL_STUB(gl_dispatch_stub_764, _gloffset__dispatch_stub_764)
 	GL_STUB(glFramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT)
-	GL_STUB(gl_dispatch_stub_765, _gloffset__dispatch_stub_765)
 	GL_STUB(gl_dispatch_stub_766, _gloffset__dispatch_stub_766)
 	GL_STUB(gl_dispatch_stub_767, _gloffset__dispatch_stub_767)
 	GL_STUB(gl_dispatch_stub_768, _gloffset__dispatch_stub_768)
 	GL_STUB(gl_dispatch_stub_769, _gloffset__dispatch_stub_769)
+	GL_STUB(gl_dispatch_stub_770, _gloffset__dispatch_stub_770)
 
 		.globl _mesa_sparc_glapi_end ; .type _mesa_sparc_glapi_end,#function
 _mesa_sparc_glapi_end:
@@ -1866,3 +1868,22 @@ _mesa_sparc_glapi_end:
 	.globl glPointParameteri ; .type glPointParameteri,#function ; glPointParameteri = glPointParameteriNV
 	.globl glPointParameteriv ; .type glPointParameteriv,#function ; glPointParameteriv = glPointParameterivNV
 	.globl glBlendEquationSeparate ; .type glBlendEquationSeparate,#function ; glBlendEquationSeparate = glBlendEquationSeparateEXT
+	.globl glBindFramebuffer ; .type glBindFramebuffer,#function ; glBindFramebuffer = glBindFramebufferEXT
+	.globl glBindRenderbuffer ; .type glBindRenderbuffer,#function ; glBindRenderbuffer = glBindRenderbufferEXT
+	.globl glCheckFramebufferStatus ; .type glCheckFramebufferStatus,#function ; glCheckFramebufferStatus = glCheckFramebufferStatusEXT
+	.globl glDeleteFramebuffers ; .type glDeleteFramebuffers,#function ; glDeleteFramebuffers = glDeleteFramebuffersEXT
+	.globl glDeleteRenderbuffers ; .type glDeleteRenderbuffers,#function ; glDeleteRenderbuffers = glDeleteRenderbuffersEXT
+	.globl glFramebufferRenderbuffer ; .type glFramebufferRenderbuffer,#function ; glFramebufferRenderbuffer = glFramebufferRenderbufferEXT
+	.globl glFramebufferTexture1D ; .type glFramebufferTexture1D,#function ; glFramebufferTexture1D = glFramebufferTexture1DEXT
+	.globl glFramebufferTexture2D ; .type glFramebufferTexture2D,#function ; glFramebufferTexture2D = glFramebufferTexture2DEXT
+	.globl glFramebufferTexture3D ; .type glFramebufferTexture3D,#function ; glFramebufferTexture3D = glFramebufferTexture3DEXT
+	.globl glGenFramebuffers ; .type glGenFramebuffers,#function ; glGenFramebuffers = glGenFramebuffersEXT
+	.globl glGenRenderbuffers ; .type glGenRenderbuffers,#function ; glGenRenderbuffers = glGenRenderbuffersEXT
+	.globl glGenerateMipmap ; .type glGenerateMipmap,#function ; glGenerateMipmap = glGenerateMipmapEXT
+	.globl glGetFramebufferAttachmentParameteriv ; .type glGetFramebufferAttachmentParameteriv,#function ; glGetFramebufferAttachmentParameteriv = glGetFramebufferAttachmentParameterivEXT
+	.globl glGetRenderbufferParameteriv ; .type glGetRenderbufferParameteriv,#function ; glGetRenderbufferParameteriv = glGetRenderbufferParameterivEXT
+	.globl glIsFramebuffer ; .type glIsFramebuffer,#function ; glIsFramebuffer = glIsFramebufferEXT
+	.globl glIsRenderbuffer ; .type glIsRenderbuffer,#function ; glIsRenderbuffer = glIsRenderbufferEXT
+	.globl glRenderbufferStorage ; .type glRenderbufferStorage,#function ; glRenderbufferStorage = glRenderbufferStorageEXT
+	.globl glBlitFramebuffer ; .type glBlitFramebuffer,#function ; glBlitFramebuffer = glBlitFramebufferEXT
+	.globl glFramebufferTextureLayer ; .type glFramebufferTextureLayer,#function ; glFramebufferTextureLayer = glFramebufferTextureLayerEXT
diff --git a/src/mesa/state_tracker/Makefile b/src/mesa/state_tracker/Makefile
new file mode 100644
index 0000000000..0ab1dc6e6b
--- /dev/null
+++ b/src/mesa/state_tracker/Makefile
@@ -0,0 +1,2 @@
+default:
+	cd ../.. ; make
\ No newline at end of file
diff --git a/src/mesa/state_tracker/st_api.c b/src/mesa/state_tracker/st_api.c
new file mode 100644
index 0000000000..fc0e9a2316
--- /dev/null
+++ b/src/mesa/state_tracker/st_api.c
@@ -0,0 +1,33 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Just a global symbol for EGL to look for to identify the supported
+ * graphics API.
+ */
+int st_api_OpenGL = 1;
diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c
new file mode 100644
index 0000000000..fc8587f459
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom.c
@@ -0,0 +1,218 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/glheader.h"
+#include "main/context.h"
+
+#include "pipe/p_defines.h"
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_bitmap.h"
+#include "st_program.h"
+
+       
+
+/* This is used to initialize st->atoms[].  We could use this list
+ * directly except for a single atom, st_update_constants, which has a
+ * .dirty value which changes according to the parameters of the
+ * current fragment and vertex programs, and so cannot be a static
+ * value.
+ */
+static const struct st_tracked_state *atoms[] =
+{
+   &st_update_depth_stencil_alpha,
+   &st_update_clip,
+
+   &st_finalize_textures,
+   &st_update_shader,
+
+   &st_update_rasterizer,
+   &st_update_polygon_stipple,
+   &st_update_viewport,
+   &st_update_scissor,
+   &st_update_blend,
+   &st_update_sampler,
+   &st_update_texture,
+   &st_update_framebuffer,
+   &st_update_vs_constants,
+   &st_update_fs_constants,
+   &st_update_pixel_transfer
+};
+
+
+void st_init_atoms( struct st_context *st )
+{
+   GLuint i;
+
+   st->atoms = malloc(sizeof(atoms));
+   st->nr_atoms = sizeof(atoms)/sizeof(*atoms);
+   memcpy(st->atoms, atoms, sizeof(atoms));
+
+   /* Patch in a pointer to the dynamic state atom:
+    */
+   for (i = 0; i < st->nr_atoms; i++) {
+      if (st->atoms[i] == &st_update_vs_constants) {
+	 st->atoms[i] = &st->constants.tracked_state[PIPE_SHADER_VERTEX];
+	 st->atoms[i][0] = st_update_vs_constants;
+      }
+
+      if (st->atoms[i] == &st_update_fs_constants) {
+	 st->atoms[i] = &st->constants.tracked_state[PIPE_SHADER_FRAGMENT];
+	 st->atoms[i][0] = st_update_fs_constants;
+      }
+   }
+}
+
+
+void st_destroy_atoms( struct st_context *st )
+{
+   if (st->atoms) {
+      free(st->atoms);
+      st->atoms = NULL;
+   }
+}
+
+
+/***********************************************************************
+ */
+
+static GLboolean check_state( const struct st_state_flags *a,
+			      const struct st_state_flags *b )
+{
+   return ((a->mesa & b->mesa) ||
+	   (a->st & b->st));
+}
+
+static void accumulate_state( struct st_state_flags *a,
+			      const struct st_state_flags *b )
+{
+   a->mesa |= b->mesa;
+   a->st |= b->st;
+}
+
+
+static void xor_states( struct st_state_flags *result,
+			     const struct st_state_flags *a,
+			      const struct st_state_flags *b )
+{
+   result->mesa = a->mesa ^ b->mesa;
+   result->st = a->st ^ b->st;
+}
+
+
+/* Too complex to figure out, just check every time:
+ */
+static void check_program_state( struct st_context *st )
+{
+   GLcontext *ctx = st->ctx;
+
+   if (ctx->VertexProgram._Current != &st->vp->Base)
+      st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+
+   if (ctx->FragmentProgram._Current != &st->fp->Base)
+      st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+
+}
+
+
+/***********************************************************************
+ * Update all derived state:
+ */
+
+void st_validate_state( struct st_context *st )
+{
+   struct st_state_flags *state = &st->dirty;
+   GLuint i;
+
+   /* The bitmap cache is immune to pixel unpack changes.
+    * Note that GLUT makes several calls to glPixelStore for each
+    * bitmap char it draws so this is an important check.
+    */
+   if (state->mesa & ~_NEW_PACKUNPACK)
+      st_flush_bitmap_cache(st);
+
+   check_program_state( st );
+
+   if (state->st == 0)
+      return;
+
+//   _mesa_printf("%s %x/%x\n", __FUNCTION__, state->mesa, state->st);
+
+   if (1) {
+      /* Debug version which enforces various sanity checks on the
+       * state flags which are generated and checked to help ensure
+       * state atoms are ordered correctly in the list.
+       */
+      struct st_state_flags examined, prev;      
+      memset(&examined, 0, sizeof(examined));
+      prev = *state;
+
+      for (i = 0; i < st->nr_atoms; i++) {	 
+	 const struct st_tracked_state *atom = st->atoms[i];
+	 struct st_state_flags generated;
+	 
+//	 _mesa_printf("atom %s %x/%x\n", atom->name, atom->dirty.mesa, atom->dirty.st);
+
+	 if (!(atom->dirty.mesa || atom->dirty.st) ||
+	     !atom->update) {
+	    _mesa_printf("malformed atom %s\n", atom->name);
+	    assert(0);
+	 }
+
+	 if (check_state(state, &atom->dirty)) {
+	    st->atoms[i]->update( st );
+//	    _mesa_printf("after: %x\n", atom->dirty.mesa);
+	 }
+
+	 accumulate_state(&examined, &atom->dirty);
+
+	 /* generated = (prev ^ state)
+	  * if (examined & generated)
+	  *     fail;
+	  */
+	 xor_states(&generated, &prev, state);
+	 assert(!check_state(&examined, &generated));
+	 prev = *state;
+      }
+//      _mesa_printf("\n");
+
+   }
+   else {
+      const GLuint nr = st->nr_atoms;
+
+      for (i = 0; i < nr; i++) {	 
+	 if (check_state(state, &st->atoms[i]->dirty))
+	    st->atoms[i]->update( st );
+      }
+   }
+
+   memset(state, 0, sizeof(*state));
+}
+
+
+
diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h
new file mode 100644
index 0000000000..c7cffd85c8
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom.h
@@ -0,0 +1,66 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+    
+
+#ifndef ST_ATOM_H
+#define ST_ATOM_H
+
+struct st_context;
+struct st_tracked_state;
+
+void st_init_atoms( struct st_context *st );
+void st_destroy_atoms( struct st_context *st );
+
+
+void st_validate_state( struct st_context *st );
+
+
+extern const struct st_tracked_state st_update_framebuffer;
+extern const struct st_tracked_state st_update_clip;
+extern const struct st_tracked_state st_update_depth_stencil_alpha;
+extern const struct st_tracked_state st_update_shader;
+extern const struct st_tracked_state st_update_rasterizer;
+extern const struct st_tracked_state st_update_polygon_stipple;
+extern const struct st_tracked_state st_update_viewport;
+extern const struct st_tracked_state st_update_scissor;
+extern const struct st_tracked_state st_update_blend;
+extern const struct st_tracked_state st_update_sampler;
+extern const struct st_tracked_state st_update_texture;
+extern const struct st_tracked_state st_finalize_textures;
+extern const struct st_tracked_state st_update_fs_constants;
+extern const struct st_tracked_state st_update_vs_constants;
+extern const struct st_tracked_state st_update_pixel_transfer;
+
+
+uint st_compare_func_to_pipe(GLenum func);
+
+#endif
diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c
new file mode 100644
index 0000000000..35c09c3e08
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_blend.c
@@ -0,0 +1,232 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+ 
+
+#include "st_context.h"
+#include "st_atom.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "cso_cache/cso_context.h"
+
+#include "main/macros.h"
+
+/**
+ * Convert GLenum blend tokens to pipe tokens.
+ * Both blend factors and blend funcs are accepted.
+ */
+static GLuint
+translate_blend(GLenum blend)
+{
+   switch (blend) {
+   /* blend functions */
+   case GL_FUNC_ADD:
+      return PIPE_BLEND_ADD;
+   case GL_FUNC_SUBTRACT:
+      return PIPE_BLEND_SUBTRACT;
+   case GL_FUNC_REVERSE_SUBTRACT:
+      return PIPE_BLEND_REVERSE_SUBTRACT;
+   case GL_MIN:
+      return PIPE_BLEND_MIN;
+   case GL_MAX:
+      return PIPE_BLEND_MAX;
+
+   /* blend factors */
+   case GL_ONE:
+      return PIPE_BLENDFACTOR_ONE;
+   case GL_SRC_COLOR:
+      return PIPE_BLENDFACTOR_SRC_COLOR;
+   case GL_SRC_ALPHA:
+      return PIPE_BLENDFACTOR_SRC_ALPHA;
+   case GL_DST_ALPHA:
+      return PIPE_BLENDFACTOR_DST_ALPHA;
+   case GL_DST_COLOR:
+      return PIPE_BLENDFACTOR_DST_COLOR;
+   case GL_SRC_ALPHA_SATURATE:
+      return PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE;
+   case GL_CONSTANT_COLOR:
+      return PIPE_BLENDFACTOR_CONST_COLOR;
+   case GL_CONSTANT_ALPHA:
+      return PIPE_BLENDFACTOR_CONST_ALPHA;
+      /*
+      return PIPE_BLENDFACTOR_SRC1_COLOR;
+      return PIPE_BLENDFACTOR_SRC1_ALPHA;
+      */
+   case GL_ZERO:
+      return PIPE_BLENDFACTOR_ZERO;
+   case GL_ONE_MINUS_SRC_COLOR:
+      return PIPE_BLENDFACTOR_INV_SRC_COLOR;
+   case GL_ONE_MINUS_SRC_ALPHA:
+      return PIPE_BLENDFACTOR_INV_SRC_ALPHA;
+   case GL_ONE_MINUS_DST_COLOR:
+      return PIPE_BLENDFACTOR_INV_DST_COLOR;
+   case GL_ONE_MINUS_DST_ALPHA:
+      return PIPE_BLENDFACTOR_INV_DST_ALPHA;
+   case GL_ONE_MINUS_CONSTANT_COLOR:
+      return PIPE_BLENDFACTOR_INV_CONST_COLOR;
+   case GL_ONE_MINUS_CONSTANT_ALPHA:
+      return PIPE_BLENDFACTOR_INV_CONST_ALPHA;
+      /*
+      return PIPE_BLENDFACTOR_INV_SRC1_COLOR;
+      return PIPE_BLENDFACTOR_INV_SRC1_ALPHA;
+      */
+   default:
+      assert("invalid GL token in translate_blend()" == NULL);
+      return 0;
+   }
+}
+
+
+/**
+ * Convert GLenum logicop tokens to pipe tokens.
+ */
+static GLuint
+translate_logicop(GLenum logicop)
+{
+   switch (logicop) {
+   case GL_CLEAR:
+      return PIPE_LOGICOP_CLEAR;
+   case GL_NOR:
+      return PIPE_LOGICOP_NOR;
+   case GL_AND_INVERTED:
+      return PIPE_LOGICOP_AND_INVERTED;
+   case GL_COPY_INVERTED:
+      return PIPE_LOGICOP_COPY_INVERTED;
+   case GL_AND_REVERSE:
+      return PIPE_LOGICOP_AND_REVERSE;
+   case GL_INVERT:
+      return PIPE_LOGICOP_INVERT;
+   case GL_XOR:
+      return PIPE_LOGICOP_XOR;
+   case GL_NAND:
+      return PIPE_LOGICOP_NAND;
+   case GL_AND:
+      return PIPE_LOGICOP_AND;
+   case GL_EQUIV:
+      return PIPE_LOGICOP_EQUIV;
+   case GL_NOOP:
+      return PIPE_LOGICOP_NOOP;
+   case GL_OR_INVERTED:
+      return PIPE_LOGICOP_OR_INVERTED;
+   case GL_COPY:
+      return PIPE_LOGICOP_COPY;
+   case GL_OR_REVERSE:
+      return PIPE_LOGICOP_OR_REVERSE;
+   case GL_OR:
+      return PIPE_LOGICOP_OR;
+   case GL_SET:
+      return PIPE_LOGICOP_SET;
+   default:
+      assert("invalid GL token in translate_logicop()" == NULL);
+      return 0;
+   }
+}
+
+
+static void 
+update_blend( struct st_context *st )
+{
+   struct pipe_blend_state *blend = &st->state.blend;
+
+   memset(blend, 0, sizeof(*blend));
+
+   if (st->ctx->Color.ColorLogicOpEnabled ||
+       (st->ctx->Color.BlendEnabled &&
+        st->ctx->Color.BlendEquationRGB == GL_LOGIC_OP)) {
+      /* logicop enabled */
+      blend->logicop_enable = 1;
+      blend->logicop_func = translate_logicop(st->ctx->Color.LogicOp);
+   }
+   else if (st->ctx->Color.BlendEnabled) {
+      /* blending enabled */
+      blend->blend_enable = 1;
+
+      blend->rgb_func = translate_blend(st->ctx->Color.BlendEquationRGB);
+      if (st->ctx->Color.BlendEquationRGB == GL_MIN ||
+          st->ctx->Color.BlendEquationRGB == GL_MAX) {
+         /* Min/max are special */
+         blend->rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->rgb_dst_factor = PIPE_BLENDFACTOR_ONE;
+      }
+      else {
+         blend->rgb_src_factor = translate_blend(st->ctx->Color.BlendSrcRGB);
+         blend->rgb_dst_factor = translate_blend(st->ctx->Color.BlendDstRGB);
+      }
+
+      blend->alpha_func = translate_blend(st->ctx->Color.BlendEquationA);
+      if (st->ctx->Color.BlendEquationA == GL_MIN ||
+          st->ctx->Color.BlendEquationA == GL_MAX) {
+         /* Min/max are special */
+         blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+         blend->alpha_dst_factor = PIPE_BLENDFACTOR_ONE;
+      }
+      else {
+         blend->alpha_src_factor = translate_blend(st->ctx->Color.BlendSrcA);
+         blend->alpha_dst_factor = translate_blend(st->ctx->Color.BlendDstA);
+      }
+   }
+   else {
+      /* no blending / logicop */
+   }
+
+   /* Colormask - maybe reverse these bits? */
+   if (st->ctx->Color.ColorMask[0])
+      blend->colormask |= PIPE_MASK_R;
+   if (st->ctx->Color.ColorMask[1])
+      blend->colormask |= PIPE_MASK_G;
+   if (st->ctx->Color.ColorMask[2])
+      blend->colormask |= PIPE_MASK_B;
+   if (st->ctx->Color.ColorMask[3])
+      blend->colormask |= PIPE_MASK_A;
+
+   if (st->ctx->Color.DitherFlag)
+      blend->dither = 1;
+
+   cso_set_blend(st->cso_context, blend);
+
+   {
+      struct pipe_blend_color bc;
+      COPY_4FV(bc.color, st->ctx->Color.BlendColor);
+      cso_set_blend_color(st->cso_context, &bc);
+   }
+}
+
+
+const struct st_tracked_state st_update_blend = {
+   "st_update_blend",					/* name */
+   {							/* dirty */
+      (_NEW_COLOR),  /* XXX _NEW_BLEND someday? */	/* mesa */
+      0,						/* st */
+   },
+   update_blend,					/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_clip.c b/src/mesa/state_tracker/st_atom_clip.c
new file mode 100644
index 0000000000..23d709b814
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_clip.c
@@ -0,0 +1,71 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+ 
+
+#include "st_context.h"
+#include "pipe/p_context.h"
+#include "st_atom.h"
+
+
+/* Second state atom for user clip planes:
+ */
+static void update_clip( struct st_context *st )
+{
+   struct pipe_clip_state clip;
+   GLuint i;
+
+   memset(&clip, 0, sizeof(clip));
+
+   for (i = 0; i < PIPE_MAX_CLIP_PLANES; i++) {
+      if (st->ctx->Transform.ClipPlanesEnabled & (1 << i)) {
+	 memcpy(clip.ucp[clip.nr], 
+		st->ctx->Transform._ClipUserPlane[i], 
+		sizeof(clip.ucp[0]));
+	 clip.nr++;
+      }
+   }
+      
+   if (memcmp(&clip, &st->state.clip, sizeof(clip)) != 0) {
+      st->state.clip = clip;
+      st->pipe->set_clip_state(st->pipe, &clip);
+   }
+}
+
+
+const struct st_tracked_state st_update_clip = {
+   "st_update_clip",					/* name */
+   {							/* dirty */
+      (_NEW_TRANSFORM),					/* mesa */
+      0,						/* st */
+   },
+   update_clip						/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c
new file mode 100644
index 0000000000..514b10cd02
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_constbuf.c
@@ -0,0 +1,140 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_atom_constbuf.h"
+#include "st_program.h"
+
+
+/**
+ * Pass the given program parameters to the graphics pipe as a
+ * constant buffer.
+ * \param id  either PIPE_SHADER_VERTEX or PIPE_SHADER_FRAGMENT
+ */
+void st_upload_constants( struct st_context *st,
+                          struct gl_program_parameter_list *params,
+                          unsigned id)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_constant_buffer *cbuf = &st->state.constants[id];
+
+   assert(id == PIPE_SHADER_VERTEX || id == PIPE_SHADER_FRAGMENT);
+
+   /* update constants */
+   if (params && params->NumParameters) {
+      const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4;
+
+      /* Update our own dependency flags.  This works because this
+       * function will also be called whenever the program changes.
+       */
+      st->constants.tracked_state[id].dirty.mesa =
+         (params->StateFlags | _NEW_PROGRAM);
+
+      _mesa_load_state_parameters(st->ctx, params);
+
+      /* We always need to get a new buffer, to keep the drivers simple and
+       * avoid gratuitous rendering synchronization.
+       */
+      pipe_buffer_reference(pipe->screen, &cbuf->buffer, NULL );
+      cbuf->buffer = pipe_buffer_create(pipe->screen, 16, PIPE_BUFFER_USAGE_CONSTANT,
+					paramBytes );
+
+      if (0)
+      {
+	 printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", 
+                __FUNCTION__, id, params->NumParameters, params->StateFlags);
+         _mesa_print_parameter_list(params);
+      }
+
+      /* load Mesa constants into the constant buffer */
+      if (cbuf->buffer) {
+         void *map = pipe_buffer_map(pipe->screen, cbuf->buffer,
+                                     PIPE_BUFFER_USAGE_CPU_WRITE);
+         memcpy(map, params->ParameterValues, paramBytes);
+         pipe_buffer_unmap(pipe->screen, cbuf->buffer);
+      }
+
+      st->pipe->set_constant_buffer(st->pipe, id, 0, cbuf);
+   }
+   else {
+      st->constants.tracked_state[id].dirty.mesa = 0;
+      //  st->pipe->set_constant_buffer(st->pipe, id, 0, NULL);
+   }
+}
+
+/* Vertex shader:
+ */
+static void update_vs_constants(struct st_context *st )
+{
+   struct st_vertex_program *vp = st->vp;
+   struct gl_program_parameter_list *params = vp->Base.Base.Parameters;
+
+   st_upload_constants( st, params, PIPE_SHADER_VERTEX );
+}
+
+const struct st_tracked_state st_update_vs_constants = {
+   "st_update_vs_constants",				/* name */
+   {							/* dirty */
+      0,  /* set dynamically above */			/* mesa */
+      ST_NEW_VERTEX_PROGRAM,				/* st */
+   },
+   update_vs_constants					/* update */
+};
+
+/* Fragment shader:
+ */
+static void update_fs_constants(struct st_context *st )
+{
+   struct st_fragment_program *fp = st->fp;
+   struct gl_program_parameter_list *params = fp->Base.Base.Parameters;
+
+   st_upload_constants( st, params, PIPE_SHADER_FRAGMENT );
+}
+
+const struct st_tracked_state st_update_fs_constants = {
+   "st_update_fs_constants",				/* name */
+   {							/* dirty */
+      0,  /* set dynamically above */			/* mesa */
+      ST_NEW_FRAGMENT_PROGRAM,				/* st */
+   },
+   update_fs_constants					/* update */
+};
+
diff --git a/src/mesa/state_tracker/st_atom_constbuf.h b/src/mesa/state_tracker/st_atom_constbuf.h
new file mode 100644
index 0000000000..f707534e2c
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_constbuf.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_ATOM_CONSTBUF_H
+#define ST_ATOM_CONSTBUF_H
+
+
+void st_upload_constants( struct st_context *st,
+                          struct gl_program_parameter_list *params,
+                          unsigned id);
+
+
+#endif /* ST_ATOM_CONSTBUF_H */
diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c
new file mode 100644
index 0000000000..2d617bd95d
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_depth.c
@@ -0,0 +1,151 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  *   Zack  Rusin
+  */
+ 
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "cso_cache/cso_context.h"
+
+
+/**
+ * Convert an OpenGL compare mode to a pipe tokens.
+ */
+GLuint
+st_compare_func_to_pipe(GLenum func)
+{
+   /* Same values, just biased */
+   assert(PIPE_FUNC_NEVER == GL_NEVER - GL_NEVER);
+   assert(PIPE_FUNC_LESS == GL_LESS - GL_NEVER);
+   assert(PIPE_FUNC_EQUAL == GL_EQUAL - GL_NEVER);
+   assert(PIPE_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER);
+   assert(PIPE_FUNC_GREATER == GL_GREATER - GL_NEVER);
+   assert(PIPE_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER);
+   assert(PIPE_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER);
+   assert(PIPE_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER);
+   assert(func >= GL_NEVER);
+   assert(func <= GL_ALWAYS);
+   return func - GL_NEVER;
+}
+
+
+/**
+ * Convert GLenum stencil op tokens to pipe tokens.
+ */
+static GLuint
+gl_stencil_op_to_pipe(GLenum func)
+{
+   switch (func) {
+   case GL_KEEP:
+      return PIPE_STENCIL_OP_KEEP;
+   case GL_ZERO:
+      return PIPE_STENCIL_OP_ZERO;
+   case GL_REPLACE:
+      return PIPE_STENCIL_OP_REPLACE;
+   case GL_INCR:
+      return PIPE_STENCIL_OP_INCR;
+   case GL_DECR:
+      return PIPE_STENCIL_OP_DECR;
+   case GL_INCR_WRAP:
+      return PIPE_STENCIL_OP_INCR_WRAP;
+   case GL_DECR_WRAP:
+      return PIPE_STENCIL_OP_DECR_WRAP;
+   case GL_INVERT:
+      return PIPE_STENCIL_OP_INVERT;
+   default:
+      assert("invalid GL token in gl_stencil_op_to_pipe()" == NULL);
+      return 0;
+   }
+}
+
+static void
+update_depth_stencil_alpha(struct st_context *st)
+{
+   struct pipe_depth_stencil_alpha_state *dsa = &st->state.depth_stencil;
+
+   memset(dsa, 0, sizeof(*dsa));
+
+   dsa->depth.enabled = st->ctx->Depth.Test;
+   dsa->depth.writemask = st->ctx->Depth.Mask;
+   dsa->depth.func = st_compare_func_to_pipe(st->ctx->Depth.Func);
+
+   if (st->ctx->Query.CurrentOcclusionObject &&
+       st->ctx->Query.CurrentOcclusionObject->Active)
+      dsa->depth.occlusion_count = 1;
+
+   if (st->ctx->Stencil.Enabled && st->ctx->Visual.stencilBits > 0) {
+      dsa->stencil[0].enabled = 1;
+      dsa->stencil[0].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[0]);
+      dsa->stencil[0].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[0]);
+      dsa->stencil[0].zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[0]);
+      dsa->stencil[0].zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[0]);
+      dsa->stencil[0].ref_value = st->ctx->Stencil.Ref[0] & 0xff;
+      dsa->stencil[0].valuemask = st->ctx->Stencil.ValueMask[0] & 0xff;
+      dsa->stencil[0].writemask = st->ctx->Stencil.WriteMask[0] & 0xff;
+
+      if (st->ctx->Stencil._TestTwoSide) {
+         dsa->stencil[1].enabled = 1;
+         dsa->stencil[1].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[1]);
+         dsa->stencil[1].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[1]);
+         dsa->stencil[1].zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[1]);
+         dsa->stencil[1].zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[1]);
+         dsa->stencil[1].ref_value = st->ctx->Stencil.Ref[1] & 0xff;
+         dsa->stencil[1].valuemask = st->ctx->Stencil.ValueMask[1] & 0xff;
+         dsa->stencil[1].writemask = st->ctx->Stencil.WriteMask[1] & 0xff;
+      }
+      else {
+         dsa->stencil[1] = dsa->stencil[0];
+         dsa->stencil[1].enabled = 0;
+      }
+   }
+
+   if (st->ctx->Color.AlphaEnabled) {
+      dsa->alpha.enabled = 1;
+      dsa->alpha.func = st_compare_func_to_pipe(st->ctx->Color.AlphaFunc);
+      dsa->alpha.ref_value = st->ctx->Color.AlphaRef;
+   }
+
+   cso_set_depth_stencil_alpha(st->cso_context, dsa);
+}
+
+
+const struct st_tracked_state st_update_depth_stencil_alpha = {
+   "st_update_depth_stencil",				/* name */
+   {							/* dirty */
+      (_NEW_DEPTH|_NEW_STENCIL|_NEW_COLOR),		/* mesa */
+      0,						/* st */
+   },
+   update_depth_stencil_alpha				/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c
new file mode 100644
index 0000000000..625efdd66b
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_framebuffer.c
@@ -0,0 +1,176 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+ 
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_fbo.h"
+#include "st_public.h"
+#include "st_texture.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "cso_cache/cso_context.h"
+
+
+
+/**
+ * When doing GL render to texture, we have to be sure that finalize_texture()
+ * didn't yank out the pipe_texture that we earlier created a surface for.
+ * Check for that here and create a new surface if needed.
+ */
+static void
+update_renderbuffer_surface(struct st_context *st,
+                            struct st_renderbuffer *strb)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+   struct pipe_texture *texture = strb->rtt->pt;
+   int rtt_width = strb->Base.Width;
+   int rtt_height = strb->Base.Height;
+
+   if (!strb->surface ||
+       strb->surface->texture != texture ||
+       strb->surface->width != rtt_width ||
+       strb->surface->height != rtt_height) {
+      GLuint level;
+      /* find matching mipmap level size */
+      for (level = 0; level <= texture->last_level; level++) {
+         if (texture->width[level] == rtt_width &&
+             texture->height[level] == rtt_height) {
+
+            pipe_surface_reference(&strb->surface, NULL);
+
+            strb->surface = screen->get_tex_surface(screen,
+                                              texture,
+                                              strb->rtt_face,
+                                              level,
+                                              strb->rtt_slice,
+                                              PIPE_BUFFER_USAGE_GPU_READ |
+                                              PIPE_BUFFER_USAGE_GPU_WRITE);
+#if 0
+            printf("-- alloc new surface %d x %d into tex %p\n",
+                   strb->surface->width, strb->surface->height,
+                   texture);
+#endif
+            break;
+         }
+      }
+   }
+}
+
+
+/**
+ * Update framebuffer state (color, depth, stencil, etc. buffers)
+ */
+static void
+update_framebuffer_state( struct st_context *st )
+{
+   struct pipe_framebuffer_state *framebuffer = &st->state.framebuffer;
+   struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+   struct st_renderbuffer *strb;
+   GLuint i;
+
+   memset(framebuffer, 0, sizeof(*framebuffer));
+
+   framebuffer->width = fb->Width;
+   framebuffer->height = fb->Height;
+
+   /*printf("------ fb size %d x %d\n", fb->Width, fb->Height);*/
+
+   /* Examine Mesa's ctx->DrawBuffer->_ColorDrawBuffers state
+    * to determine which surfaces to draw to
+    */
+   framebuffer->nr_cbufs = 0;
+   for (i = 0; i < fb->_NumColorDrawBuffers; i++) {
+      strb = st_renderbuffer(fb->_ColorDrawBuffers[i]);
+
+      /*printf("--------- framebuffer surface rtt %p\n", strb->rtt);*/
+      if (strb->rtt) {
+         /* rendering to a GL texture, may have to update surface */
+         update_renderbuffer_surface(st, strb);
+      }
+
+      if (strb->surface) {
+         framebuffer->cbufs[framebuffer->nr_cbufs] = strb->surface;
+         framebuffer->nr_cbufs++;
+      }
+   }
+
+   strb = st_renderbuffer(fb->Attachment[BUFFER_DEPTH].Renderbuffer);
+   if (strb) {
+      strb = st_renderbuffer(strb->Base.Wrapped);
+      if (strb->rtt) {
+         /* rendering to a GL texture, may have to update surface */
+         update_renderbuffer_surface(st, strb);
+      }
+
+      framebuffer->zsbuf = strb->surface;
+   }
+   else {
+      strb = st_renderbuffer(fb->Attachment[BUFFER_STENCIL].Renderbuffer);
+      if (strb) {
+         strb = st_renderbuffer(strb->Base.Wrapped);
+         assert(strb->surface);
+         framebuffer->zsbuf = strb->surface;
+      }
+   }
+
+   cso_set_framebuffer(st->cso_context, framebuffer);
+
+   if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
+      if (st->frontbuffer_status == FRONT_STATUS_COPY_OF_BACK) {
+         /* copy back color buffer to front color buffer */
+         struct st_framebuffer *stfb = (struct st_framebuffer *) fb;
+	 struct pipe_surface *surf_front, *surf_back;
+         (void) st_get_framebuffer_surface(stfb, ST_SURFACE_FRONT_LEFT, &surf_front);
+         (void) st_get_framebuffer_surface(stfb, ST_SURFACE_BACK_LEFT, &surf_back);
+
+         st->pipe->surface_copy(st->pipe,
+                                FALSE,
+                                surf_front, 0, 0,  /* dest */
+                                surf_back, 0, 0,   /* src */
+                                fb->Width, fb->Height);
+      }
+      /* we're assuming we'll really draw to the front buffer */
+      st->frontbuffer_status = FRONT_STATUS_DIRTY;
+   }
+}
+
+
+const struct st_tracked_state st_update_framebuffer = {
+   "st_update_framebuffer",				/* name */
+   {							/* dirty */
+      _NEW_BUFFERS,					/* mesa */
+      ST_NEW_FRAMEBUFFER,				/* st */
+   },
+   update_framebuffer_state				/* update */
+};
+
diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c
new file mode 100644
index 0000000000..a357b71677
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c
@@ -0,0 +1,474 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/*
+ * Generate fragment programs to implement pixel transfer ops, such as
+ * scale/bias, colormatrix, colortable, convolution...
+ *
+ * Authors:
+ *   Brian Paul
+ */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "shader/program.h"
+#include "shader/prog_instruction.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "st_context.h"
+#include "st_format.h"
+#include "st_program.h"
+#include "st_texture.h"
+
+#include "pipe/p_screen.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "util/u_pack_color.h"
+
+
+struct state_key
+{
+   GLuint scaleAndBias:1;
+   GLuint colorMatrix:1;
+   GLuint colorMatrixPostScaleBias:1;
+   GLuint pixelMaps:1;
+
+#if 0
+   GLfloat Maps[3][256][4];
+   int NumMaps;
+   GLint NumStages;
+   pipeline_stage Stages[STAGE_MAX];
+   GLboolean StagesUsed[STAGE_MAX];
+   GLfloat Scale1[4], Bias1[4];
+   GLfloat Scale2[4], Bias2[4];
+#endif
+};
+
+
+static GLboolean
+is_identity(const GLfloat m[16])
+{
+   GLuint i;
+   for (i = 0; i < 16; i++) {
+      const int row = i % 4, col = i / 4;
+      const float val = (GLfloat)(row == col);
+      if (m[i] != val)
+         return GL_FALSE;
+   }
+   return GL_TRUE;
+}
+
+
+static void
+make_state_key(GLcontext *ctx,  struct state_key *key)
+{
+   static const GLfloat zero[4] = { 0.0f, 0.0f, 0.0f, 0.0f };
+   static const GLfloat one[4] = { 1.0f, 1.0f, 1.0f, 1.0f };
+
+   memset(key, 0, sizeof(*key));
+
+   if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 ||
+       ctx->Pixel.GreenBias != 0.0 || ctx->Pixel.GreenScale != 1.0 ||
+       ctx->Pixel.BlueBias != 0.0 || ctx->Pixel.BlueScale != 1.0 ||
+       ctx->Pixel.AlphaBias != 0.0 || ctx->Pixel.AlphaScale != 1.0) {
+      key->scaleAndBias = 1;
+   }
+
+   if (!is_identity(ctx->ColorMatrixStack.Top->m)) {
+      key->colorMatrix = 1;
+   }
+
+   if (!TEST_EQ_4V(ctx->Pixel.PostColorMatrixScale, one) ||
+       !TEST_EQ_4V(ctx->Pixel.PostColorMatrixBias, zero)) {
+      key->colorMatrixPostScaleBias = 1;
+   }
+
+   key->pixelMaps = ctx->Pixel.MapColorFlag;
+}
+
+
+static struct pipe_texture *
+create_color_map_texture(GLcontext *ctx)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_texture *pt;
+   enum pipe_format format;
+   const uint texSize = 256; /* simple, and usually perfect */
+
+   /* find an RGBA texture format */
+   format = st_choose_format(pipe, GL_RGBA, PIPE_TEXTURE_2D, PIPE_TEXTURE_USAGE_SAMPLER);
+
+   /* create texture for color map/table */
+   pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0,
+                          texSize, texSize, 1, 0,
+                          PIPE_TEXTURE_USAGE_SAMPLER);
+   return pt;
+}
+
+
+/**
+ * Update the pixelmap texture with the contents of the R/G/B/A pixel maps.
+ */
+static void
+load_color_map_texture(GLcontext *ctx, struct pipe_texture *pt)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surface;
+   const GLuint rSize = ctx->PixelMaps.RtoR.Size;
+   const GLuint gSize = ctx->PixelMaps.GtoG.Size;
+   const GLuint bSize = ctx->PixelMaps.BtoB.Size;
+   const GLuint aSize = ctx->PixelMaps.AtoA.Size;
+   const uint texSize = pt->width[0];
+   uint *dest;
+   uint i, j;
+
+   surface = screen->get_tex_surface(screen, pt, 0, 0, 0, 
+                                     PIPE_BUFFER_USAGE_CPU_WRITE);
+   dest = (uint *) screen->surface_map(screen, surface,
+                                       PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* Pack four 1D maps into a 2D texture:
+    * R map is placed horizontally, indexed by S, in channel 0
+    * G map is placed vertically, indexed by T, in channel 1
+    * B map is placed horizontally, indexed by S, in channel 2
+    * A map is placed vertically, indexed by T, in channel 3
+    */
+   for (i = 0; i < texSize; i++) {
+      for (j = 0; j < texSize; j++) {
+         int k = (i * texSize + j);
+         ubyte r = ctx->PixelMaps.RtoR.Map8[j * rSize / texSize];
+         ubyte g = ctx->PixelMaps.GtoG.Map8[i * gSize / texSize];
+         ubyte b = ctx->PixelMaps.BtoB.Map8[j * bSize / texSize];
+         ubyte a = ctx->PixelMaps.AtoA.Map8[i * aSize / texSize];
+         util_pack_color_ub(r, g, b, a, pt->format, dest + k);
+      }
+   }
+
+   screen->surface_unmap(screen, surface);
+   pipe_surface_reference(&surface, NULL);
+}
+
+
+
+#define MAX_INST 100
+
+/**
+ * Returns a fragment program which implements the current pixel transfer ops.
+ */
+static struct gl_fragment_program *
+get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key)
+{
+   struct st_context *st = ctx->st;
+   struct prog_instruction inst[MAX_INST];
+   struct gl_program_parameter_list *params;
+   struct gl_fragment_program *fp;
+   GLuint ic = 0;
+   const GLuint colorTemp = 0;
+
+   fp = (struct gl_fragment_program *)
+      ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
+   if (!fp)
+      return NULL;
+
+   params = _mesa_new_parameter_list();
+
+   /*
+    * Get initial pixel color from the texture.
+    * TEX colorTemp, fragment.texcoord[0], texture[0], 2D;
+    */
+   _mesa_init_instructions(inst + ic, 1);
+   inst[ic].Opcode = OPCODE_TEX;
+   inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+   inst[ic].DstReg.Index = colorTemp;
+   inst[ic].SrcReg[0].File = PROGRAM_INPUT;
+   inst[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
+   inst[ic].TexSrcUnit = 0;
+   inst[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+   ic++;
+   fp->Base.InputsRead = (1 << FRAG_ATTRIB_TEX0);
+   fp->Base.OutputsWritten = (1 << FRAG_RESULT_COLR);
+   fp->Base.SamplersUsed = 0x1;  /* sampler 0 (bit 0) is used */
+
+   if (key->scaleAndBias) {
+      static const gl_state_index scale_state[STATE_LENGTH] =
+         { STATE_INTERNAL, STATE_PT_SCALE, 0, 0, 0 };
+      static const gl_state_index bias_state[STATE_LENGTH] =
+         { STATE_INTERNAL, STATE_PT_BIAS, 0, 0, 0 };
+      GLfloat scale[4], bias[4];
+      GLint scale_p, bias_p;
+
+      scale[0] = ctx->Pixel.RedScale;
+      scale[1] = ctx->Pixel.GreenScale;
+      scale[2] = ctx->Pixel.BlueScale;
+      scale[3] = ctx->Pixel.AlphaScale;
+      bias[0] = ctx->Pixel.RedBias;
+      bias[1] = ctx->Pixel.GreenBias;
+      bias[2] = ctx->Pixel.BlueBias;
+      bias[3] = ctx->Pixel.AlphaBias;
+
+      scale_p = _mesa_add_state_reference(params, scale_state);
+      bias_p = _mesa_add_state_reference(params, bias_state);
+
+      /* MAD colorTemp, colorTemp, scale, bias; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_MAD;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = colorTemp;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = scale_p;
+      inst[ic].SrcReg[2].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[2].Index = bias_p;
+      ic++;
+   }
+
+   if (key->pixelMaps) {
+      const GLuint temp = 1;
+
+      /* create the colormap/texture now if not already done */
+      if (!st->pixel_xfer.pixelmap_texture) {
+         st->pixel_xfer.pixelmap_texture = create_color_map_texture(ctx);
+      }
+
+      /* with a little effort, we can do four pixel map look-ups with
+       * two TEX instructions:
+       */
+
+      /* TEX temp.rg, colorTemp.rgba, texture[1], 2D; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_TEX;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_XY; /* write R,G */
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].TexSrcUnit = 1;
+      inst[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+      ic++;
+
+      /* TEX temp.ba, colorTemp.baba, texture[1], 2D; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_TEX;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_ZW; /* write B,A */
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[0].Swizzle = MAKE_SWIZZLE4(SWIZZLE_Z, SWIZZLE_W,
+                                                 SWIZZLE_Z, SWIZZLE_W);
+      inst[ic].TexSrcUnit = 1;
+      inst[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+      ic++;
+
+      /* MOV colorTemp, temp; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_MOV;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = colorTemp;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = temp;
+      ic++;
+
+      fp->Base.SamplersUsed |= (1 << 1);  /* sampler 1 is used */
+   }
+
+   if (key->colorMatrix) {
+      static const gl_state_index row0_state[STATE_LENGTH] =
+         { STATE_COLOR_MATRIX, 0, 0, 0, 0 };
+      static const gl_state_index row1_state[STATE_LENGTH] =
+         { STATE_COLOR_MATRIX, 0, 1, 1, 0 };
+      static const gl_state_index row2_state[STATE_LENGTH] =
+         { STATE_COLOR_MATRIX, 0, 2, 2, 0 };
+      static const gl_state_index row3_state[STATE_LENGTH] =
+         { STATE_COLOR_MATRIX, 0, 3, 3, 0 };
+
+      GLint row0_p = _mesa_add_state_reference(params, row0_state);
+      GLint row1_p = _mesa_add_state_reference(params, row1_state);
+      GLint row2_p = _mesa_add_state_reference(params, row2_state);
+      GLint row3_p = _mesa_add_state_reference(params, row3_state);
+      const GLuint temp = 1;
+
+      /* DP4 temp.x, colorTemp, matrow0; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_DP4;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_X;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = row0_p;
+      ic++;
+
+      /* DP4 temp.y, colorTemp, matrow1; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_DP4;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_Y;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = row1_p;
+      ic++;
+
+      /* DP4 temp.z, colorTemp, matrow2; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_DP4;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_Z;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = row2_p;
+      ic++;
+
+      /* DP4 temp.w, colorTemp, matrow3; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_DP4;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = temp;
+      inst[ic].DstReg.WriteMask = WRITEMASK_W;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = row3_p;
+      ic++;
+
+      /* MOV colorTemp, temp; */
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_MOV;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = colorTemp;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = temp;
+      ic++;
+   }
+
+   if (key->colorMatrixPostScaleBias) {
+      static const gl_state_index scale_state[STATE_LENGTH] =
+         { STATE_INTERNAL, STATE_PT_SCALE, 0, 0, 0 };
+      static const gl_state_index bias_state[STATE_LENGTH] =
+         { STATE_INTERNAL, STATE_PT_BIAS, 0, 0, 0 };
+      GLint scale_param, bias_param;
+
+      scale_param = _mesa_add_state_reference(params, scale_state);
+      bias_param = _mesa_add_state_reference(params, bias_state);
+
+      _mesa_init_instructions(inst + ic, 1);
+      inst[ic].Opcode = OPCODE_MAD;
+      inst[ic].DstReg.File = PROGRAM_TEMPORARY;
+      inst[ic].DstReg.Index = colorTemp;
+      inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+      inst[ic].SrcReg[0].Index = colorTemp;
+      inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[1].Index = scale_param;
+      inst[ic].SrcReg[2].File = PROGRAM_STATE_VAR;
+      inst[ic].SrcReg[2].Index = bias_param;
+      ic++;
+   }
+
+   /* Modify last instruction's dst reg to write to result.color */
+   {
+      struct prog_instruction *last = &inst[ic - 1];
+      last->DstReg.File = PROGRAM_OUTPUT;
+      last->DstReg.Index = FRAG_RESULT_COLR;
+   }
+
+   /* END; */
+   _mesa_init_instructions(inst + ic, 1);
+   inst[ic].Opcode = OPCODE_END;
+   ic++;
+
+   assert(ic <= MAX_INST);
+
+
+   fp->Base.Instructions = _mesa_alloc_instructions(ic);
+   if (!fp->Base.Instructions) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY,
+                  "generating pixel transfer program");
+      return NULL;
+   }
+
+   _mesa_copy_instructions(fp->Base.Instructions, inst, ic);
+   fp->Base.NumInstructions = ic;
+   fp->Base.Parameters = params;
+
+#if 0
+   printf("========= pixel transfer prog\n");
+   _mesa_print_program(&fp->Base);
+   _mesa_print_parameter_list(fp->Base.Parameters);
+#endif
+
+   return fp;
+}
+
+
+
+/**
+ * Update st->pixel_xfer.program in response to new pixel-transfer state.
+ */
+static void
+update_pixel_transfer(struct st_context *st)
+{
+   GLcontext *ctx = st->ctx;
+   struct state_key key;
+   struct gl_fragment_program *fp;
+
+   make_state_key(st->ctx, &key);
+
+   fp = (struct gl_fragment_program *)
+      _mesa_search_program_cache(st->pixel_xfer.cache, &key, sizeof(key));
+   if (!fp) {
+      fp = get_pixel_transfer_program(st->ctx, &key);
+      _mesa_program_cache_insert(st->ctx, st->pixel_xfer.cache,
+                                 &key, sizeof(key), &fp->Base);
+   }
+
+   if (ctx->Pixel.MapColorFlag) {
+      load_color_map_texture(ctx, st->pixel_xfer.pixelmap_texture);
+   }
+   st->pixel_xfer.pixelmap_enabled = ctx->Pixel.MapColorFlag;
+
+   st->pixel_xfer.program = (struct st_fragment_program *) fp;
+}
+
+
+
+const struct st_tracked_state st_update_pixel_transfer = {
+   "st_update_pixel_transfer",				/* name */
+   {							/* dirty */
+      _NEW_PIXEL | _NEW_COLOR_MATRIX,			/* mesa */
+      0,						/* st */
+   },
+   update_pixel_transfer				/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c
new file mode 100644
index 0000000000..ea76487bcf
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_rasterizer.c
@@ -0,0 +1,289 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+ 
+#include "main/macros.h"
+#include "st_context.h"
+#include "st_atom.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "cso_cache/cso_context.h"
+
+
+static GLuint translate_fill( GLenum mode )
+{
+   switch (mode) {
+   case GL_POINT:
+      return PIPE_POLYGON_MODE_POINT;
+   case GL_LINE:
+      return PIPE_POLYGON_MODE_LINE;
+   case GL_FILL:
+      return PIPE_POLYGON_MODE_FILL;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+static GLboolean get_offset_flag( GLuint fill_mode, 
+				  const struct gl_polygon_attrib *p )
+{
+   switch (fill_mode) {
+   case PIPE_POLYGON_MODE_POINT:
+      return p->OffsetPoint;
+   case PIPE_POLYGON_MODE_LINE:
+      return p->OffsetLine;
+   case PIPE_POLYGON_MODE_FILL:
+      return p->OffsetFill;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+
+static void update_raster_state( struct st_context *st )
+{
+   GLcontext *ctx = st->ctx;
+   struct pipe_rasterizer_state *raster = &st->state.rasterizer;
+   const struct gl_vertex_program *vertProg = ctx->VertexProgram._Current;
+   uint i;
+
+   memset(raster, 0, sizeof(*raster));
+
+   raster->origin_lower_left = 1; /* Always true for OpenGL */
+   
+   /* _NEW_POLYGON, _NEW_BUFFERS
+    */
+   {
+      if (ctx->Polygon.FrontFace == GL_CCW)
+         raster->front_winding = PIPE_WINDING_CCW;
+      else
+         raster->front_winding = PIPE_WINDING_CW;
+
+      /* XXX
+       * I think the intention here is that user-created framebuffer objects
+       * use Y=0=TOP layout instead of OpenGL's normal Y=0=bottom layout.
+       * Flipping Y changes CW to CCW and vice-versa.
+       * But this is an implementation/driver-specific artifact - remove...
+       */
+      if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0)
+         raster->front_winding ^= PIPE_WINDING_BOTH;
+   }
+
+   /* _NEW_LIGHT
+    */
+   if (ctx->Light.ShadeModel == GL_FLAT)
+      raster->flatshade = 1;
+
+   /* _NEW_LIGHT | _NEW_PROGRAM
+    *
+    * Back-face colors can come from traditional lighting (when
+    * GL_LIGHT_MODEL_TWO_SIDE is set) or from vertex programs/shaders (when
+    * GL_VERTEX_PROGRAM_TWO_SIDE is set).  Note the logic here.
+    */
+   if (ctx->VertexProgram._Current) {
+      if (ctx->VertexProgram._Enabled ||
+          (ctx->Shader.CurrentProgram &&
+           ctx->Shader.CurrentProgram->VertexProgram &&
+           ctx->Shader.CurrentProgram->LinkStatus)) {
+         /* user-defined vertex program or shader */
+         raster->light_twoside = ctx->VertexProgram.TwoSideEnabled;
+      }
+      else {
+         /* TNL-generated program */
+         raster->light_twoside = ctx->Light.Enabled && ctx->Light.Model.TwoSide;
+      }
+   }
+   else if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) {
+      raster->light_twoside = 1;
+   }
+
+   /* _NEW_POLYGON
+    */
+   if (ctx->Polygon.CullFlag) {
+      if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) {
+	 raster->cull_mode = PIPE_WINDING_BOTH;
+      }
+      else if (ctx->Polygon.CullFaceMode == GL_FRONT) {
+	 raster->cull_mode = raster->front_winding;
+      }
+      else {
+	 raster->cull_mode = raster->front_winding ^ PIPE_WINDING_BOTH;
+      }
+   }
+
+   /* _NEW_POLYGON
+    */
+   {
+      GLuint fill_front = translate_fill( ctx->Polygon.FrontMode );
+      GLuint fill_back = translate_fill( ctx->Polygon.BackMode );
+      
+      if (raster->front_winding == PIPE_WINDING_CW) {
+	 raster->fill_cw = fill_front;
+	 raster->fill_ccw = fill_back;
+      }
+      else {
+	 raster->fill_cw = fill_back;
+	 raster->fill_ccw = fill_front;
+      }
+
+      /* Simplify when culling is active:
+       */
+      if (raster->cull_mode & PIPE_WINDING_CW) {
+	 raster->fill_cw = raster->fill_ccw;
+      }
+      
+      if (raster->cull_mode & PIPE_WINDING_CCW) {
+	 raster->fill_ccw = raster->fill_cw;
+      }
+   }
+
+   /* _NEW_POLYGON 
+    */
+   if (ctx->Polygon.OffsetUnits != 0.0 ||
+       ctx->Polygon.OffsetFactor != 0.0) {
+      raster->offset_cw = get_offset_flag( raster->fill_cw, &ctx->Polygon );
+      raster->offset_ccw = get_offset_flag( raster->fill_ccw, &ctx->Polygon );
+      raster->offset_units = ctx->Polygon.OffsetUnits;
+      raster->offset_scale = ctx->Polygon.OffsetFactor;
+   }
+
+   if (ctx->Polygon.SmoothFlag)
+      raster->poly_smooth = 1;
+
+   if (ctx->Polygon.StippleFlag)
+      raster->poly_stipple_enable = 1;
+
+
+   /* _NEW_BUFFERS, _NEW_POLYGON
+    */
+   if (raster->fill_cw != PIPE_POLYGON_MODE_FILL ||
+       raster->fill_ccw != PIPE_POLYGON_MODE_FILL)
+   {
+      GLfloat mrd = (ctx->DrawBuffer ? 
+		     ctx->DrawBuffer->_MRD : 
+		     1.0f);
+
+      raster->offset_units = ctx->Polygon.OffsetFactor * mrd;
+      raster->offset_scale = (ctx->Polygon.OffsetUnits * mrd *
+			    st->polygon_offset_scale);
+   }
+      
+   /* _NEW_POINT
+    */
+   raster->point_size = ctx->Point.Size;
+
+   raster->point_size_min = 0;    /* temporary, will go away */
+   raster->point_size_max = 1000; /* temporary, will go away */
+
+   raster->point_smooth = ctx->Point.SmoothFlag;
+   raster->point_sprite = ctx->Point.PointSprite;
+   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
+      if (ctx->Point.CoordReplace[i]) {
+         if (ctx->Point.SpriteOrigin == GL_UPPER_LEFT)
+            raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_UPPER_LEFT;
+         else 
+            raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_LOWER_LEFT;
+      }
+      else {
+         raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_NONE;
+      }
+   }
+
+   /* ST_NEW_VERTEX_PROGRAM
+    */
+   if (vertProg) {
+      if (vertProg->Base.Id == 0) {
+         if (vertProg->Base.OutputsWritten & (1 << VERT_RESULT_PSIZ)) {
+            /* generated program which emits point size */
+            raster->point_size_per_vertex = TRUE;
+         }
+      }
+      else if (ctx->VertexProgram.PointSizeEnabled) {
+         /* user-defined program and GL_VERTEX_PROGRAM_POINT_SIZE set */
+         raster->point_size_per_vertex = ctx->VertexProgram.PointSizeEnabled;
+      }
+   }
+   if (!raster->point_size_per_vertex) {
+      /* clamp size now */
+      raster->point_size = CLAMP(ctx->Point.Size,
+                                 ctx->Point.MinSize,
+                                 ctx->Point.MaxSize);
+   }
+
+   /* _NEW_LINE
+    */
+   raster->line_smooth = ctx->Line.SmoothFlag;
+   if (ctx->Line.SmoothFlag) {
+      raster->line_width = CLAMP(ctx->Line.Width,
+                                ctx->Const.MinLineWidthAA,
+                                ctx->Const.MaxLineWidthAA);
+   }
+   else {
+      raster->line_width = CLAMP(ctx->Line.Width,
+                                ctx->Const.MinLineWidth,
+                                ctx->Const.MaxLineWidth);
+   }
+
+   raster->line_stipple_enable = ctx->Line.StippleFlag;
+   raster->line_stipple_pattern = ctx->Line.StipplePattern;
+   /* GL stipple factor is in [1,256], remap to [0, 255] here */
+   raster->line_stipple_factor = ctx->Line.StippleFactor - 1;
+
+   /* _NEW_MULTISAMPLE */
+   if (ctx->Multisample._Enabled || st->force_msaa)
+      raster->multisample = 1;
+
+   /* _NEW_SCISSOR */
+   if (ctx->Scissor.Enabled)
+      raster->scissor = 1;
+
+   raster->gl_rasterization_rules = 1;
+
+   cso_set_rasterizer(st->cso_context, raster);
+}
+
+const struct st_tracked_state st_update_rasterizer = {
+   "st_update_rasterizer",    /* name */
+   {
+      (_NEW_BUFFERS |
+       _NEW_LIGHT |
+       _NEW_LINE |
+       _NEW_MULTISAMPLE |
+       _NEW_POINT |
+       _NEW_POLYGON |
+       _NEW_PROGRAM |
+       _NEW_SCISSOR),      /* mesa state dependencies*/
+      ST_NEW_VERTEX_PROGRAM,  /* state tracker dependencies */
+   },
+   update_raster_state     /* update function */
+};
diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c
new file mode 100644
index 0000000000..d7b904354f
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_sampler.c
@@ -0,0 +1,217 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+ 
+
+#include "main/macros.h"
+
+#include "st_context.h"
+#include "st_cb_texture.h"
+#include "st_atom.h"
+#include "st_program.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+
+#include "cso_cache/cso_context.h"
+
+
+/**
+ * Convert GLenum texcoord wrap tokens to pipe tokens.
+ */
+static GLuint
+gl_wrap_to_sp(GLenum wrap)
+{
+   switch (wrap) {
+   case GL_REPEAT:
+      return PIPE_TEX_WRAP_REPEAT;
+   case GL_CLAMP:
+      return PIPE_TEX_WRAP_CLAMP;
+   case GL_CLAMP_TO_EDGE:
+      return PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   case GL_CLAMP_TO_BORDER:
+      return PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+   case GL_MIRRORED_REPEAT:
+      return PIPE_TEX_WRAP_MIRROR_REPEAT;
+   case GL_MIRROR_CLAMP_EXT:
+      return PIPE_TEX_WRAP_MIRROR_CLAMP;
+   case GL_MIRROR_CLAMP_TO_EDGE_EXT:
+      return PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE;
+   case GL_MIRROR_CLAMP_TO_BORDER_EXT:
+      return PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER;
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+
+static GLuint
+gl_filter_to_mip_filter(GLenum filter)
+{
+   switch (filter) {
+   case GL_NEAREST:
+   case GL_LINEAR:
+      return PIPE_TEX_MIPFILTER_NONE;
+
+   case GL_NEAREST_MIPMAP_NEAREST:
+   case GL_LINEAR_MIPMAP_NEAREST:
+      return PIPE_TEX_MIPFILTER_NEAREST;
+
+   case GL_NEAREST_MIPMAP_LINEAR:
+   case GL_LINEAR_MIPMAP_LINEAR:
+      return PIPE_TEX_MIPFILTER_LINEAR;
+
+   default:
+      assert(0);
+      return PIPE_TEX_MIPFILTER_NONE;
+   }
+}
+
+
+static GLuint
+gl_filter_to_img_filter(GLenum filter)
+{
+   switch (filter) {
+   case GL_NEAREST:
+   case GL_NEAREST_MIPMAP_NEAREST:
+   case GL_NEAREST_MIPMAP_LINEAR:
+      return PIPE_TEX_FILTER_NEAREST;
+
+   case GL_LINEAR:
+   case GL_LINEAR_MIPMAP_NEAREST:
+   case GL_LINEAR_MIPMAP_LINEAR:
+      return PIPE_TEX_FILTER_LINEAR;
+
+   default:
+      assert(0);
+      return PIPE_TEX_FILTER_NEAREST;
+   }
+}
+
+
+static void 
+update_samplers(struct st_context *st)
+{
+   struct gl_vertex_program *vprog = st->ctx->VertexProgram._Current;
+   struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current;
+   const GLbitfield samplersUsed = (vprog->Base.SamplersUsed |
+                                    fprog->Base.SamplersUsed);
+   GLuint su;
+
+   st->state.num_samplers = 0;
+
+   /* loop over sampler units (aka tex image units) */
+   for (su = 0; su < st->ctx->Const.MaxTextureImageUnits; su++) {
+      struct pipe_sampler_state *sampler = st->state.samplers + su;
+
+      memset(sampler, 0, sizeof(*sampler));
+
+      if (samplersUsed & (1 << su)) {
+         struct gl_texture_object *texobj;
+         GLuint texUnit;
+
+         if (fprog->Base.SamplersUsed & (1 << su))
+            texUnit = fprog->Base.SamplerUnits[su];
+         else
+            texUnit = vprog->Base.SamplerUnits[su];
+
+         texobj = st->ctx->Texture.Unit[texUnit]._Current;
+         if (!texobj) {
+            texobj = st_get_default_texture(st);
+         }
+
+         sampler->wrap_s = gl_wrap_to_sp(texobj->WrapS);
+         sampler->wrap_t = gl_wrap_to_sp(texobj->WrapT);
+         sampler->wrap_r = gl_wrap_to_sp(texobj->WrapR);
+
+         sampler->min_img_filter = gl_filter_to_img_filter(texobj->MinFilter);
+         sampler->min_mip_filter = gl_filter_to_mip_filter(texobj->MinFilter);
+         sampler->mag_img_filter = gl_filter_to_img_filter(texobj->MagFilter);
+
+         if (texobj->Target != GL_TEXTURE_RECTANGLE_ARB)
+            sampler->normalized_coords = 1;
+
+         sampler->lod_bias = st->ctx->Texture.Unit[su].LodBias;
+         sampler->min_lod = MAX2(0.0f, texobj->MinLod);
+         sampler->max_lod = MIN2(texobj->MaxLevel - texobj->BaseLevel,
+                                 texobj->MaxLod);
+         if (sampler->max_lod < sampler->min_lod) {
+            /* The GL spec doesn't seem to specify what to do in this case.
+             * Swap the values.
+             */
+            float tmp = sampler->max_lod;
+            sampler->max_lod = sampler->min_lod;
+            sampler->min_lod = tmp;
+            assert(sampler->min_lod <= sampler->max_lod);
+         }
+
+         sampler->border_color[0] = texobj->BorderColor[RCOMP];
+         sampler->border_color[1] = texobj->BorderColor[GCOMP];
+         sampler->border_color[2] = texobj->BorderColor[BCOMP];
+         sampler->border_color[3] = texobj->BorderColor[ACOMP];
+
+	 sampler->max_anisotropy = texobj->MaxAnisotropy;
+         if (sampler->max_anisotropy > 1.0) {
+            sampler->min_img_filter = PIPE_TEX_FILTER_ANISO;
+            sampler->mag_img_filter = PIPE_TEX_FILTER_ANISO;
+         }
+
+         /* only care about ARB_shadow, not SGI shadow */
+         if (texobj->CompareMode == GL_COMPARE_R_TO_TEXTURE) {
+            sampler->compare_mode = PIPE_TEX_COMPARE_R_TO_TEXTURE;
+            sampler->compare_func
+               = st_compare_func_to_pipe(texobj->CompareFunc);
+         }
+
+         st->state.num_samplers = su + 1;
+
+         /*printf("%s su=%u non-null\n", __FUNCTION__, su);*/
+         cso_single_sampler(st->cso_context, su, sampler);
+      }
+      else {
+         /*printf("%s su=%u null\n", __FUNCTION__, su);*/
+         cso_single_sampler(st->cso_context, su, NULL);
+      }
+   }
+
+   cso_single_sampler_done(st->cso_context);
+}
+
+
+const struct st_tracked_state st_update_sampler = {
+   "st_update_sampler",					/* name */
+   {							/* dirty */
+      _NEW_TEXTURE,					/* mesa */
+      0,						/* st */
+   },
+   update_samplers					/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_scissor.c b/src/mesa/state_tracker/st_atom_scissor.c
new file mode 100644
index 0000000000..3fd59e1945
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_scissor.c
@@ -0,0 +1,92 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+ 
+
+#include "st_context.h"
+#include "pipe/p_context.h"
+#include "st_atom.h"
+
+
+/**
+ * Scissor depends on the scissor box, and the framebuffer dimensions.
+ */
+static void
+update_scissor( struct st_context *st )
+{
+   struct pipe_scissor_state scissor;
+   const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+   GLint miny, maxy;
+
+   scissor.minx = 0;
+   scissor.miny = 0;
+   scissor.maxx = fb->Width;
+   scissor.maxy = fb->Height;
+
+   if (st->ctx->Scissor.Enabled) {
+      if ((GLuint)st->ctx->Scissor.X > scissor.minx)
+         scissor.minx = st->ctx->Scissor.X;
+      if ((GLuint)st->ctx->Scissor.Y > scissor.miny)
+         scissor.miny = st->ctx->Scissor.Y;
+
+      if ((GLuint)st->ctx->Scissor.X + st->ctx->Scissor.Width < scissor.maxx)
+         scissor.maxx = st->ctx->Scissor.X + st->ctx->Scissor.Width;
+      if ((GLuint)st->ctx->Scissor.Y + st->ctx->Scissor.Height < scissor.maxy)
+         scissor.maxy = st->ctx->Scissor.Y + st->ctx->Scissor.Height;
+
+      /* check for null space */
+      if (scissor.minx >= scissor.maxx || scissor.miny >= scissor.maxy)
+         scissor.minx = scissor.miny = scissor.maxx = scissor.maxy = 0;
+   }
+
+   /* Now invert Y.  Pipe drivers use the convention Y=0=top for surfaces
+    */
+   miny = fb->Height - scissor.maxy;
+   maxy = fb->Height - scissor.miny;
+   scissor.miny = miny;
+   scissor.maxy = maxy;
+
+   if (memcmp(&scissor, &st->state.scissor, sizeof(scissor)) != 0) {
+      /* state has changed */
+      st->state.scissor = scissor;  /* struct copy */
+      st->pipe->set_scissor_state(st->pipe, &scissor); /* activate */
+   }
+}
+
+
+const struct st_tracked_state st_update_scissor = {
+   "st_update_scissor",					/* name */
+   {							/* dirty */
+      (_NEW_SCISSOR | _NEW_BUFFERS),			/* mesa */
+      0,						/* st */
+   },
+   update_scissor					/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
new file mode 100644
index 0000000000..cbd414e2d3
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -0,0 +1,369 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * State validation for vertex/fragment shaders.
+ * Note that we have to delay most vertex/fragment shader translation
+ * until rendering time since the linkage between the vertex outputs and
+ * fragment inputs can vary depending on the pairing of shaders.
+ *
+ * Authors:
+ *   Brian Paul
+ */
+
+
+
+#include "main/imports.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+#include "shader/program.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_simple_shaders.h"
+
+#include "cso_cache/cso_context.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_program.h"
+#include "st_atom_shader.h"
+#include "st_mesa_to_tgsi.h"
+
+
+/**
+ * This represents a vertex program, especially translated to match
+ * the inputs of a particular fragment shader.
+ */
+struct translated_vertex_program
+{
+   struct st_vertex_program *master;
+
+   /** The fragment shader "signature" this vertex shader is meant for: */
+   GLbitfield frag_inputs;
+
+   /** Compared against master vertex program's serialNo: */
+   GLuint serialNo;
+
+   /** Maps VERT_RESULT_x to slot */
+   GLuint output_to_slot[VERT_RESULT_MAX];
+   ubyte output_to_semantic_name[VERT_RESULT_MAX];
+   ubyte output_to_semantic_index[VERT_RESULT_MAX];
+
+   /** Pointer to the translated vertex program */
+   struct st_vertex_program *vp;
+
+   struct translated_vertex_program *next;  /**< next in linked list */
+};
+
+
+
+/**
+ * Given a vertex program output attribute, return the corresponding
+ * fragment program input attribute.
+ * \return -1 for vertex outputs that have no corresponding fragment input
+ */
+static GLint
+vp_out_to_fp_in(GLuint vertResult)
+{
+   if (vertResult >= VERT_RESULT_TEX0 &&
+       vertResult < VERT_RESULT_TEX0 + MAX_TEXTURE_COORD_UNITS)
+      return FRAG_ATTRIB_TEX0 + (vertResult - VERT_RESULT_TEX0);
+
+   if (vertResult >= VERT_RESULT_VAR0 &&
+       vertResult < VERT_RESULT_VAR0 + MAX_VARYING)
+      return FRAG_ATTRIB_VAR0 + (vertResult - VERT_RESULT_VAR0);
+
+   switch (vertResult) {
+   case VERT_RESULT_HPOS:
+      return FRAG_ATTRIB_WPOS;
+   case VERT_RESULT_COL0:
+      return FRAG_ATTRIB_COL0;
+   case VERT_RESULT_COL1:
+      return FRAG_ATTRIB_COL1;
+   case VERT_RESULT_FOGC:
+      return FRAG_ATTRIB_FOGC;
+   default:
+      /* Back-face colors, edge flags, etc */
+      return -1;
+   }
+}
+
+
+/**
+ * Find a translated vertex program that corresponds to stvp and
+ * has outputs matched to stfp's inputs.
+ * This performs vertex and fragment translation (to TGSI) when needed.
+ */
+static struct translated_vertex_program *
+find_translated_vp(struct st_context *st,
+                   struct st_vertex_program *stvp,
+                   struct st_fragment_program *stfp)
+{
+   static const GLuint UNUSED = ~0;
+   struct translated_vertex_program *xvp;
+   const GLbitfield fragInputsRead = stfp->Base.Base.InputsRead;
+
+   /*
+    * Translate fragment program if needed.
+    */
+   if (!stfp->state.tokens) {
+      GLuint inAttr, numIn = 0;
+
+      for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) {
+         if (fragInputsRead & (1 << inAttr)) {
+            stfp->input_to_slot[inAttr] = numIn;
+            numIn++;
+         }
+         else {
+            stfp->input_to_slot[inAttr] = UNUSED;
+         }
+      }
+
+      stfp->num_input_slots = numIn;
+
+      assert(stfp->Base.Base.NumInstructions > 1);
+
+      st_translate_fragment_program(st, stfp, stfp->input_to_slot);
+   }
+
+
+   /* See if we've got a translated vertex program whose outputs match
+    * the fragment program's inputs.
+    * XXX This could be a hash lookup, using InputsRead as the key.
+    */
+   for (xvp = stfp->vertex_programs; xvp; xvp = xvp->next) {
+      if (xvp->master == stvp && xvp->frag_inputs == fragInputsRead) {
+         break;
+      }
+   }
+
+   /* No?  Allocate translated vp object now */
+   if (!xvp) {
+      xvp = CALLOC_STRUCT(translated_vertex_program);
+      xvp->frag_inputs = fragInputsRead;
+      xvp->master = stvp;
+
+      xvp->next = stfp->vertex_programs;
+      stfp->vertex_programs = xvp;
+   }
+
+   /* See if we need to translate vertex program to TGSI form */
+   if (xvp->serialNo != stvp->serialNo) {
+      GLuint outAttr, dummySlot;
+      const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten;
+      GLuint numVpOuts = 0;
+      GLboolean emitPntSize = GL_FALSE, emitBFC0 = GL_FALSE, emitBFC1 = GL_FALSE;
+      GLint maxGeneric;
+
+      /* Compute mapping of vertex program outputs to slots, which depends
+       * on the fragment program's input->slot mapping.
+       */
+      for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
+         /* set defaults: */
+         xvp->output_to_slot[outAttr] = UNUSED;
+         xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_COUNT;
+         xvp->output_to_semantic_index[outAttr] = 99;
+
+         if (outAttr == VERT_RESULT_HPOS) {
+            /* always put xformed position into slot zero */
+            xvp->output_to_slot[VERT_RESULT_HPOS] = 0;
+            xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_POSITION;
+            xvp->output_to_semantic_index[outAttr] = 0;
+            numVpOuts++;
+         }
+         else if (outputsWritten & (1 << outAttr)) {
+            /* see if the frag prog wants this vert output */
+            GLint fpInAttrib = vp_out_to_fp_in(outAttr);
+            if (fpInAttrib >= 0) {
+               GLuint fpInSlot = stfp->input_to_slot[fpInAttrib];
+               if (fpInSlot != ~0) {
+                  /* match this vp output to the fp input */
+                  GLuint vpOutSlot = stfp->input_map[fpInSlot];
+                  xvp->output_to_slot[outAttr] = vpOutSlot;
+                  xvp->output_to_semantic_name[outAttr] = stfp->input_semantic_name[fpInSlot];
+                  xvp->output_to_semantic_index[outAttr] = stfp->input_semantic_index[fpInSlot];
+                  numVpOuts++;
+               }
+            }
+            else if (outAttr == VERT_RESULT_PSIZ)
+               emitPntSize = GL_TRUE;
+            else if (outAttr == VERT_RESULT_BFC0)
+               emitBFC0 = GL_TRUE;
+            else if (outAttr == VERT_RESULT_BFC1)
+               emitBFC1 = GL_TRUE;
+         }
+#if 0 /*debug*/
+         printf("assign vp output_to_slot[%d] = %d\n", outAttr, 
+                xvp->output_to_slot[outAttr]);
+#endif
+      }
+
+      /* must do these last */
+      if (emitPntSize) {
+         xvp->output_to_slot[VERT_RESULT_PSIZ] = numVpOuts++;
+         xvp->output_to_semantic_name[VERT_RESULT_PSIZ] = TGSI_SEMANTIC_PSIZE;
+         xvp->output_to_semantic_index[VERT_RESULT_PSIZ] = 0;
+      }
+      if (emitBFC0) {
+         xvp->output_to_slot[VERT_RESULT_BFC0] = numVpOuts++;
+         xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR;
+         xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 0;
+      }
+      if (emitBFC1) {
+         xvp->output_to_slot[VERT_RESULT_BFC1] = numVpOuts++;
+         xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR;
+         xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 1;
+      }
+
+      /* Unneeded vertex program outputs will go to this slot.
+       * We could use this info to do dead code elimination in the
+       * vertex program.
+       */
+      dummySlot = numVpOuts;
+
+      /* find max GENERIC slot index */
+      maxGeneric = -1;
+      for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
+         if (xvp->output_to_semantic_name[outAttr] == TGSI_SEMANTIC_GENERIC) {
+            maxGeneric = MAX2(maxGeneric,
+                              xvp->output_to_semantic_index[outAttr]);
+         }
+      }
+
+      /* Map vert program outputs that aren't used to the dummy slot
+       * (and an unused generic attribute slot).
+       */
+      for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
+         if (outputsWritten & (1 << outAttr)) {
+            if (xvp->output_to_slot[outAttr] == UNUSED) {
+               xvp->output_to_slot[outAttr] = dummySlot;
+               xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_GENERIC;
+               xvp->output_to_semantic_index[outAttr] = maxGeneric + 1;
+            }
+         }
+
+#if 0 /*debug*/
+         printf("vp output_to_slot[%d] = %d\n", outAttr, 
+                xvp->output_to_slot[outAttr]);
+#endif
+      }
+
+      assert(stvp->Base.Base.NumInstructions > 1);
+
+      st_translate_vertex_program(st, stvp, xvp->output_to_slot,
+                                  xvp->output_to_semantic_name,
+                                  xvp->output_to_semantic_index);
+
+      xvp->vp = stvp;
+
+      /* translated VP is up to date now */
+      xvp->serialNo = stvp->serialNo;
+   }
+
+   return xvp;
+}
+
+
+void
+st_free_translated_vertex_programs(struct st_context *st,
+                                   struct translated_vertex_program *xvp)
+{
+   struct translated_vertex_program *next;
+
+   while (xvp) {
+      next = xvp->next;
+      free(xvp);
+      xvp = next;
+   }
+}
+
+
+static void *
+get_passthrough_fs(struct st_context *st)
+{
+   struct pipe_shader_state shader;
+
+   if (!st->passthrough_fs) {
+      st->passthrough_fs =
+         util_make_fragment_passthrough_shader(st->pipe, &shader);
+#if 0      /* We actually need to keep the tokens around at this time */
+      free((void *) shader.tokens);
+#endif
+   }
+
+   return st->passthrough_fs;
+}
+
+
+static void
+update_linkage( struct st_context *st )
+{
+   struct st_vertex_program *stvp;
+   struct st_fragment_program *stfp;
+   struct translated_vertex_program *xvp;
+
+   /* find active shader and params -- Should be covered by
+    * ST_NEW_VERTEX_PROGRAM
+    */
+   assert(st->ctx->VertexProgram._Current);
+   stvp = st_vertex_program(st->ctx->VertexProgram._Current);
+   assert(stvp->Base.Base.Target == GL_VERTEX_PROGRAM_ARB);
+
+   assert(st->ctx->FragmentProgram._Current);
+   stfp = st_fragment_program(st->ctx->FragmentProgram._Current);
+   assert(stfp->Base.Base.Target == GL_FRAGMENT_PROGRAM_ARB);
+
+   xvp = find_translated_vp(st, stvp, stfp);
+
+   st_reference_vertprog(st, &st->vp, stvp);
+   st_reference_fragprog(st, &st->fp, stfp);
+
+   cso_set_vertex_shader_handle(st->cso_context, stvp->driver_shader);
+
+   if (st->missing_textures) {
+      /* use a pass-through frag shader that uses no textures */
+      void *fs = get_passthrough_fs(st);
+      cso_set_fragment_shader_handle(st->cso_context, fs);
+   }
+   else {
+      cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader);
+   }
+
+   st->vertex_result_to_slot = xvp->output_to_slot;
+}
+
+
+const struct st_tracked_state st_update_shader = {
+   "st_update_shader",					/* name */
+   {							/* dirty */
+      0,						/* mesa */
+      ST_NEW_VERTEX_PROGRAM | ST_NEW_FRAGMENT_PROGRAM	/* st */
+   },
+   update_linkage					/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_shader.h b/src/mesa/state_tracker/st_atom_shader.h
new file mode 100644
index 0000000000..8403bc66c9
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_shader.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_ATOM_SHADER_H
+#define ST_ATOM_SHADER_H
+
+
+extern void
+st_free_translated_vertex_programs(struct st_context *st,
+                                   struct translated_vertex_program *xvp);
+
+
+#endif /* ST_ATOM_SHADER_H */
diff --git a/src/mesa/state_tracker/st_atom_stipple.c b/src/mesa/state_tracker/st_atom_stipple.c
new file mode 100644
index 0000000000..f395930ab4
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_stipple.c
@@ -0,0 +1,63 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * \brief polygon stipple state
+  *
+  * Authors:
+  *   Brian Paul
+  */
+ 
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+
+
+static void 
+update_stipple( struct st_context *st )
+{
+   const GLuint sz = sizeof(st->state.poly_stipple.stipple);
+   assert(sz == sizeof(st->ctx->PolygonStipple));
+
+   if (memcmp(&st->state.poly_stipple.stipple, st->ctx->PolygonStipple, sz)) {
+      /* state has changed */
+      memcpy(st->state.poly_stipple.stipple, st->ctx->PolygonStipple, sz);
+      st->pipe->set_polygon_stipple(st->pipe, &st->state.poly_stipple);
+   }
+}
+
+
+const struct st_tracked_state st_update_polygon_stipple = {
+   "st_update_polygon_stipple",				/* name */
+   {							/* dirty */
+      (_NEW_POLYGONSTIPPLE),				/* mesa */
+      0,						/* st */
+   },
+   update_stipple					/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c
new file mode 100644
index 0000000000..21f7321f97
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_texture.c
@@ -0,0 +1,162 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+ 
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_texture.h"
+#include "st_cb_texture.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "cso_cache/cso_context.h"
+
+
+static void 
+update_textures(struct st_context *st)
+{
+   struct gl_vertex_program *vprog = st->ctx->VertexProgram._Current;
+   struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current;
+   const GLbitfield samplersUsed = (vprog->Base.SamplersUsed |
+                                    fprog->Base.SamplersUsed);
+   GLuint su;
+
+   st->state.num_textures = 0;
+
+   /* loop over sampler units (aka tex image units) */
+   for (su = 0; su < st->ctx->Const.MaxTextureImageUnits; su++) {
+      struct pipe_texture *pt = NULL;
+
+      if (samplersUsed & (1 << su)) {
+         struct gl_texture_object *texObj;
+         struct st_texture_object *stObj;
+         GLboolean flush, retval;
+         GLuint texUnit;
+
+         if (fprog->Base.SamplersUsed & (1 << su))
+            texUnit = fprog->Base.SamplerUnits[su];
+         else
+            texUnit = vprog->Base.SamplerUnits[su];
+
+         texObj = st->ctx->Texture.Unit[texUnit]._Current;
+
+         if (!texObj) {
+            texObj = st_get_default_texture(st);
+         }
+         stObj = st_texture_object(texObj);
+
+         retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush);
+         if (!retval) {
+            /* out of mem */
+            continue;
+         }
+
+         st->state.num_textures = su + 1;
+
+         pt = st_get_stobj_texture(stObj);
+      }
+
+      /*
+      if (pt) {
+         printf("%s su=%u non-null\n", __FUNCTION__, su);
+      }
+      else {
+         printf("%s su=%u null\n", __FUNCTION__, su);
+      }
+      */
+
+      pipe_texture_reference(&st->state.sampler_texture[su], pt);
+   }
+
+   cso_set_sampler_textures(st->cso_context,
+                            st->state.num_textures,
+                            st->state.sampler_texture);
+}
+
+
+const struct st_tracked_state st_update_texture = {
+   "st_update_texture",					/* name */
+   {							/* dirty */
+      _NEW_TEXTURE,					/* mesa */
+      ST_NEW_FRAGMENT_PROGRAM,				/* st */
+   },
+   update_textures					/* update */
+};
+
+
+
+
+static void 
+finalize_textures(struct st_context *st)
+{
+   struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current;
+   const GLboolean prev_missing_textures = st->missing_textures;
+   GLuint su;
+
+   st->missing_textures = GL_FALSE;
+
+   for (su = 0; su < st->ctx->Const.MaxTextureCoordUnits; su++) {
+      if (fprog->Base.SamplersUsed & (1 << su)) {
+         const GLuint texUnit = fprog->Base.SamplerUnits[su];
+         struct gl_texture_object *texObj
+            = st->ctx->Texture.Unit[texUnit]._Current;
+         struct st_texture_object *stObj = st_texture_object(texObj);
+
+         if (texObj) {
+            GLboolean flush, retval;
+
+            retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush);
+            if (!retval) {
+               /* out of mem */
+               st->missing_textures = GL_TRUE;
+               continue;
+            }
+
+            stObj->teximage_realloc = TRUE;
+         }
+      }
+   }
+
+   if (prev_missing_textures != st->missing_textures)
+      st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+}
+
+
+
+const struct st_tracked_state st_finalize_textures = {
+   "st_finalize_textures",		/* name */
+   {					/* dirty */
+      _NEW_TEXTURE,			/* mesa */
+      0,				/* st */
+   },
+   finalize_textures			/* update */
+};
diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c
new file mode 100644
index 0000000000..27ec2eb033
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_viewport.c
@@ -0,0 +1,91 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/context.h"
+#include "main/colormac.h"
+#include "st_context.h"
+#include "st_atom.h"
+#include "pipe/p_context.h"
+#include "cso_cache/cso_context.h"
+
+/**
+ * Update the viewport transformation matrix.  Depends on:
+ *  - viewport pos/size
+ *  - depthrange
+ *  - window pos/size or FBO size
+ */
+static void
+update_viewport( struct st_context *st )
+{
+   GLcontext *ctx = st->ctx;
+   GLfloat yScale, yBias;
+
+   /* _NEW_BUFFERS
+    */
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+      yScale = -1;
+      yBias = (GLfloat)ctx->DrawBuffer->Height;
+   }
+   else {
+      yScale = 1.0;
+      yBias = 0.0;
+   }
+
+   /* _NEW_VIEWPORT 
+    */
+   {
+      GLfloat x = (GLfloat)ctx->Viewport.X;
+      GLfloat y = (GLfloat)ctx->Viewport.Y;
+      GLfloat z = ctx->Viewport.Near;
+      GLfloat half_width = (GLfloat)ctx->Viewport.Width / 2.0f;
+      GLfloat half_height = (GLfloat)ctx->Viewport.Height / 2.0f;
+      GLfloat half_depth = (GLfloat)(ctx->Viewport.Far - ctx->Viewport.Near) / 2.0f;
+      
+      st->state.viewport.scale[0] = half_width;
+      st->state.viewport.scale[1] = half_height * yScale;
+      st->state.viewport.scale[2] = half_depth;
+      st->state.viewport.scale[3] = 1.0;
+
+      st->state.viewport.translate[0] = half_width + x;
+      st->state.viewport.translate[1] = (half_height + y) * yScale + yBias;
+      st->state.viewport.translate[2] = half_depth + z;
+      st->state.viewport.translate[3] = 0.0;
+
+      cso_set_viewport(st->cso_context, &st->state.viewport);
+   }
+}
+
+
+const struct st_tracked_state st_update_viewport = {
+   "st_update_viewport",				/* name */
+   {							/* dirty */
+      _NEW_BUFFERS | _NEW_VIEWPORT,			/* mesa */
+      0,						/* st */
+   },
+   update_viewport					/* update */
+};
diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h
new file mode 100644
index 0000000000..b81de316ec
--- /dev/null
+++ b/src/mesa/state_tracker/st_cache.h
@@ -0,0 +1,68 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Zack Rusin <zack@tungstengraphics.com>
+  */
+
+#ifndef ST_CACHE_H
+#define ST_CACHE_H
+
+#include "cso_cache/cso_cache.h"
+
+struct pipe_blend_state;
+struct pipe_sampler_state;
+struct st_context;
+
+
+const struct cso_blend *
+st_cached_blend_state(struct st_context *st,
+                      const struct pipe_blend_state *blend);
+
+const struct cso_sampler *
+st_cached_sampler_state(struct st_context *st,
+                        const struct pipe_sampler_state *sampler);
+
+const struct cso_depth_stencil_alpha *
+st_cached_depth_stencil_alpha_state(struct st_context *st,
+                              const struct pipe_depth_stencil_alpha_state *depth_stencil);
+
+const struct cso_rasterizer *
+st_cached_rasterizer_state(struct st_context *st,
+                           const struct pipe_rasterizer_state *raster);
+
+const struct cso_fragment_shader *
+st_cached_fs_state(struct st_context *st,
+                   const struct pipe_shader_state *templ);
+
+
+const struct cso_vertex_shader *
+st_cached_vs_state(struct st_context *st,
+                   const struct pipe_shader_state *templ);
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c
new file mode 100644
index 0000000000..a4e72b48ed
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_accum.c
@@ -0,0 +1,362 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+
+#include "st_context.h"
+#include "st_cb_accum.h"
+#include "st_cb_fbo.h"
+#include "st_draw.h"
+#include "st_public.h"
+#include "st_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+
+
+#define UNCLAMPED_FLOAT_TO_SHORT(us, f)  \
+   us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) )
+
+
+/**
+ * For hardware that supports deep color buffers, we could accelerate
+ * most/all the accum operations with blending/texturing.
+ * For now, just use the get/put_tile() functions and do things in software.
+ */
+
+
+/**
+ * Wrapper for pipe_get_tile_rgba().  Do format/cpp override to make the
+ * tile util function think the surface is 16bit/channel, even if it's not.
+ * See also: st_renderbuffer_alloc_storage()
+ */
+static void
+acc_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps,
+                  uint x, uint y, uint w, uint h, float *p)
+{
+   const enum pipe_format f = acc_ps->format;
+   const struct pipe_format_block b = acc_ps->block;
+
+   acc_ps->format = DEFAULT_ACCUM_PIPE_FORMAT;
+   acc_ps->block.size = 8;
+   acc_ps->block.width = 1;
+   acc_ps->block.height = 1;
+
+   pipe_get_tile_rgba(acc_ps, x, y, w, h, p);
+
+   acc_ps->format = f;
+   acc_ps->block = b;
+}
+
+
+/**
+ * Wrapper for pipe_put_tile_rgba().  Do format/cpp override to make the
+ * tile util function think the surface is 16bit/channel, even if it's not.
+ * See also: st_renderbuffer_alloc_storage()
+ */
+static void
+acc_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps,
+                  uint x, uint y, uint w, uint h, const float *p)
+{
+   enum pipe_format f = acc_ps->format;
+   const struct pipe_format_block b = acc_ps->block;
+
+   acc_ps->format = DEFAULT_ACCUM_PIPE_FORMAT;
+   acc_ps->block.size = 8;
+   acc_ps->block.width = 1;
+   acc_ps->block.height = 1;
+
+   pipe_put_tile_rgba(acc_ps, x, y, w, h, p);
+
+   acc_ps->format = f;
+   acc_ps->block = b;
+}
+
+
+
+void
+st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *acc_strb = st_renderbuffer(rb);
+   struct pipe_surface *acc_ps;
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   const GLint xpos = ctx->DrawBuffer->_Xmin;
+   const GLint ypos = ctx->DrawBuffer->_Ymin;
+   const GLint width = ctx->DrawBuffer->_Xmax - xpos;
+   const GLint height = ctx->DrawBuffer->_Ymax - ypos;
+   GLubyte *map;
+
+   acc_ps = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0,
+                                    PIPE_BUFFER_USAGE_CPU_WRITE);
+   map = screen->surface_map(screen, acc_ps,
+                             PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* note acc_strb->format might not equal acc_ps->format */
+   switch (acc_strb->format) {
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      {
+         GLshort r = FLOAT_TO_SHORT(ctx->Accum.ClearColor[0]);
+         GLshort g = FLOAT_TO_SHORT(ctx->Accum.ClearColor[1]);
+         GLshort b = FLOAT_TO_SHORT(ctx->Accum.ClearColor[2]);
+         GLshort a = FLOAT_TO_SHORT(ctx->Accum.ClearColor[3]);
+         int i, j;
+         for (i = 0; i < height; i++) {
+            GLshort *dst = (GLshort *) (map + (ypos + i) * acc_ps->stride + xpos * 8);
+            for (j = 0; j < width; j++) {
+               dst[0] = r;
+               dst[1] = g;
+               dst[2] = b;
+               dst[3] = a;
+               dst += 4;
+            }
+         }
+      }
+      break;
+   default:
+      _mesa_problem(ctx, "unexpected format in st_clear_accum_buffer()");
+   }
+
+   screen->surface_unmap(screen, acc_ps);
+   pipe_surface_reference(&acc_ps, NULL);
+}
+
+
+/** For ADD/MULT */
+static void
+accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias,
+          GLint xpos, GLint ypos, GLint width, GLint height,
+          struct st_renderbuffer *acc_strb)
+{
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   struct pipe_surface *acc_ps = acc_strb->surface;
+   GLubyte *map;
+
+   map = screen->surface_map(screen, acc_ps, 
+                             PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* note acc_strb->format might not equal acc_ps->format */
+   switch (acc_strb->format) {
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      {
+         int i, j;
+         for (i = 0; i < height; i++) {
+            GLshort *acc = (GLshort *) (map + (ypos + i) * acc_ps->stride + xpos * 8);
+            for (j = 0; j < width * 4; j++) {
+               float val = SHORT_TO_FLOAT(acc[j]) * scale + bias;
+               acc[j] = FLOAT_TO_SHORT(val);
+            }
+         }
+      }
+      break;
+   default:
+      _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()");
+   }
+
+   screen->surface_unmap(screen, acc_ps);
+}
+
+
+static void
+accum_accum(struct pipe_context *pipe, GLfloat value,
+            GLint xpos, GLint ypos, GLint width, GLint height,
+            struct st_renderbuffer *acc_strb,
+            struct st_renderbuffer *color_strb)
+{
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *acc_surf, *color_surf;
+   GLfloat *colorBuf, *accBuf;
+   GLint i;
+
+   acc_surf = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0,
+                                      (PIPE_BUFFER_USAGE_CPU_WRITE |
+                                       PIPE_BUFFER_USAGE_CPU_READ));
+
+   color_surf = screen->get_tex_surface(screen, color_strb->texture, 0, 0, 0,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+
+   colorBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+   accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+
+   pipe_get_tile_rgba(color_surf, xpos, ypos, width, height, colorBuf);
+   acc_get_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, accBuf);
+
+   for (i = 0; i < 4 * width * height; i++) {
+      accBuf[i] = accBuf[i] + colorBuf[i] * value;
+   }
+
+   acc_put_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, accBuf);
+
+   free(colorBuf);
+   free(accBuf);
+   pipe_surface_reference(&acc_surf, NULL);
+   pipe_surface_reference(&color_surf, NULL);
+}
+
+
+static void
+accum_load(struct pipe_context *pipe, GLfloat value,
+           GLint xpos, GLint ypos, GLint width, GLint height,
+           struct st_renderbuffer *acc_strb,
+           struct st_renderbuffer *color_strb)
+{
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *acc_surf, *color_surf;
+   GLfloat *buf;
+   GLint i;
+
+   acc_surf = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0,
+                                      PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   color_surf = screen->get_tex_surface(screen, color_strb->texture, 0, 0, 0,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+
+   buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+
+   pipe_get_tile_rgba(color_surf, xpos, ypos, width, height, buf);
+
+   for (i = 0; i < 4 * width * height; i++) {
+      buf[i] = buf[i] * value;
+   }
+
+   acc_put_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, buf);
+
+   free(buf);
+   pipe_surface_reference(&acc_surf, NULL);
+   pipe_surface_reference(&color_surf, NULL);
+}
+
+
+static void
+accum_return(GLcontext *ctx, GLfloat value,
+             GLint xpos, GLint ypos, GLint width, GLint height,
+             struct st_renderbuffer *acc_strb,
+             struct st_renderbuffer *color_strb)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const GLubyte *colormask = ctx->Color.ColorMask;
+   struct pipe_surface *acc_surf, *color_surf;
+   GLfloat *abuf, *cbuf = NULL;
+   GLint i, ch;
+
+   abuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+
+   acc_surf = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0,
+                                      PIPE_BUFFER_USAGE_CPU_READ);
+
+   color_surf = screen->get_tex_surface(screen, color_strb->texture, 0, 0, 0,
+                                        (PIPE_BUFFER_USAGE_CPU_READ |
+                                         PIPE_BUFFER_USAGE_CPU_WRITE));
+
+   acc_get_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, abuf);
+
+   if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) {
+      cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+      pipe_get_tile_rgba(color_surf, xpos, ypos, width, height, cbuf);
+   }
+
+   for (i = 0; i < width * height; i++) {
+      for (ch = 0; ch < 4; ch++) {
+         if (colormask[ch]) {
+            GLfloat val = abuf[i * 4 + ch] * value;
+            abuf[i * 4 + ch] = CLAMP(val, 0.0f, 1.0f);
+         }
+         else {
+            abuf[i * 4 + ch] = cbuf[i * 4 + ch];
+         }
+      }
+   }
+
+   pipe_put_tile_rgba(color_surf, xpos, ypos, width, height, abuf);
+
+   free(abuf);
+   if (cbuf)
+      free(cbuf);
+   pipe_surface_reference(&acc_surf, NULL);
+   pipe_surface_reference(&color_surf, NULL);
+}
+
+
+static void
+st_Accum(GLcontext *ctx, GLenum op, GLfloat value)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   struct st_renderbuffer *acc_strb
+     = st_renderbuffer(ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer);
+   struct st_renderbuffer *color_strb
+      = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer);
+
+   const GLint xpos = ctx->DrawBuffer->_Xmin;
+   const GLint ypos = ctx->DrawBuffer->_Ymin;
+   const GLint width = ctx->DrawBuffer->_Xmax - xpos;
+   const GLint height = ctx->DrawBuffer->_Ymax - ypos;
+
+   /* make sure color bufs aren't cached */
+   st_flush( st, PIPE_FLUSH_RENDER_CACHE, NULL );
+
+   switch (op) {
+   case GL_ADD:
+      if (value != 0.0F) {
+         accum_mad(ctx, 1.0, value, xpos, ypos, width, height, acc_strb);
+      }
+      break;
+   case GL_MULT:
+      if (value != 1.0F) {
+         accum_mad(ctx, value, 0.0, xpos, ypos, width, height, acc_strb);
+      }
+      break;
+   case GL_ACCUM:
+      if (value != 0.0F) {
+         accum_accum(pipe, value, xpos, ypos, width, height, acc_strb, color_strb);
+      }
+      break;
+   case GL_LOAD:
+      accum_load(pipe, value, xpos, ypos, width, height, acc_strb, color_strb);
+      break;
+   case GL_RETURN:
+      accum_return(ctx, value, xpos, ypos, width, height, acc_strb, color_strb);
+      break;
+   default:
+      assert(0);
+   }
+}
+
+
+
+void st_init_accum_functions(struct dd_function_table *functions)
+{
+   functions->Accum = st_Accum;
+}
diff --git a/src/mesa/state_tracker/st_cb_accum.h b/src/mesa/state_tracker/st_cb_accum.h
new file mode 100644
index 0000000000..ed9b7dab94
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_accum.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_ACCUM_H
+#define ST_CB_ACCUM_H
+
+
+extern void
+st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb);
+
+extern void st_init_accum_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_ACCUM_H */
diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c
new file mode 100644
index 0000000000..f14e562400
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_bitmap.c
@@ -0,0 +1,833 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/bufferobj.h"
+#include "main/macros.h"
+#include "main/texformat.h"
+#include "shader/program.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_atom_constbuf.h"
+#include "st_program.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_program.h"
+#include "st_mesa_to_tgsi.h"
+#include "st_texture.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+#include "util/u_draw_quad.h"
+#include "util/u_simple_shaders.h"
+#include "shader/prog_instruction.h"
+#include "cso_cache/cso_context.h"
+
+
+
+/**
+ * glBitmaps are drawn as textured quads.  The user's bitmap pattern
+ * is stored in a texture image.  An alpha8 texture format is used.
+ * The fragment shader samples a bit (texel) from the texture, then
+ * discards the fragment if the bit is off.
+ *
+ * Note that we actually store the inverse image of the bitmap to
+ * simplify the fragment program.  An "on" bit gets stored as texel=0x0
+ * and an "off" bit is stored as texel=0xff.  Then we kill the
+ * fragment if the negated texel value is less than zero.
+ */
+
+
+/**
+ * The bitmap cache attempts to accumulate multiple glBitmap calls in a
+ * buffer which is then rendered en mass upon a flush, state change, etc.
+ * A wide, short buffer is used to target the common case of a series
+ * of glBitmap calls being used to draw text.
+ */
+static GLboolean UseBitmapCache = GL_TRUE;
+
+
+#define BITMAP_CACHE_WIDTH  512
+#define BITMAP_CACHE_HEIGHT 32
+
+struct bitmap_cache
+{
+   /** Window pos to render the cached image */
+   GLint xpos, ypos;
+   /** Bounds of region used in window coords */
+   GLint xmin, ymin, xmax, ymax;
+
+   GLfloat color[4];
+
+   struct pipe_texture *texture;
+   struct pipe_surface *surf;
+
+   GLboolean empty;
+
+   /** An I8 texture image: */
+   ubyte *buffer;
+};
+
+
+
+
+/**
+ * Make fragment program for glBitmap:
+ *   Sample the texture and kill the fragment if the bit is 0.
+ * This program will be combined with the user's fragment program.
+ */
+static struct st_fragment_program *
+make_bitmap_fragment_program(GLcontext *ctx, GLuint samplerIndex)
+{
+   struct st_fragment_program *stfp;
+   struct gl_program *p;
+   GLuint ic = 0;
+
+   p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
+   if (!p)
+      return NULL;
+
+   p->NumInstructions = 3;
+
+   p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
+   if (!p->Instructions) {
+      ctx->Driver.DeleteProgram(ctx, p);
+      return NULL;
+   }
+   _mesa_init_instructions(p->Instructions, p->NumInstructions);
+
+   /* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */
+   p->Instructions[ic].Opcode = OPCODE_TEX;
+   p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].DstReg.Index = 0;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
+   p->Instructions[ic].TexSrcUnit = samplerIndex;
+   p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+   ic++;
+
+   /* KIL if -tmp0 < 0 # texel=0 -> keep / texel=0 -> discard */
+   p->Instructions[ic].Opcode = OPCODE_KIL;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].SrcReg[0].Index = 0;
+   p->Instructions[ic].SrcReg[0].NegateBase = NEGATE_XYZW;
+   ic++;
+
+   /* END; */
+   p->Instructions[ic++].Opcode = OPCODE_END;
+
+   assert(ic == p->NumInstructions);
+
+   p->InputsRead = FRAG_BIT_TEX0;
+   p->OutputsWritten = 0x0;
+   p->SamplersUsed = (1 << samplerIndex);
+
+   stfp = (struct st_fragment_program *) p;
+   stfp->Base.UsesKill = GL_TRUE;
+   st_translate_fragment_program(ctx->st, stfp, NULL);
+
+   return stfp;
+}
+
+
+static int
+find_free_bit(uint bitfield)
+{
+   int i;
+   for (i = 0; i < 32; i++) {
+      if ((bitfield & (1 << i)) == 0) {
+         return i;
+      }
+   }
+   return -1;
+}
+
+
+/**
+ * Combine basic bitmap fragment program with the user-defined program.
+ */
+static struct st_fragment_program *
+combined_bitmap_fragment_program(GLcontext *ctx)
+{
+   struct st_context *st = ctx->st;
+   struct st_fragment_program *stfp = st->fp;
+
+   if (!stfp->bitmap_program) {
+      /*
+       * Generate new program which is the user-defined program prefixed
+       * with the bitmap sampler/kill instructions.
+       */
+      struct st_fragment_program *bitmap_prog;
+      uint sampler;
+
+      sampler = find_free_bit(st->fp->Base.Base.SamplersUsed);
+      bitmap_prog = make_bitmap_fragment_program(ctx, sampler);
+
+      stfp->bitmap_program = (struct st_fragment_program *)
+         _mesa_combine_programs(ctx,
+                                &bitmap_prog->Base.Base, &stfp->Base.Base);
+      stfp->bitmap_program->bitmap_sampler = sampler;
+
+      /* done with this after combining */
+      st_reference_fragprog(st, &bitmap_prog, NULL);
+
+#if 0
+      {
+         struct gl_program *p = &stfp->bitmap_program->Base.Base;
+         printf("Combined bitmap program:\n");
+         _mesa_print_program(p);
+         printf("InputsRead: 0x%x\n", p->InputsRead);
+         printf("OutputsWritten: 0x%x\n", p->OutputsWritten);
+         _mesa_print_parameter_list(p->Parameters);
+      }
+#endif
+
+      /* translate to TGSI tokens */
+      st_translate_fragment_program(st, stfp->bitmap_program, NULL);
+   }
+
+   return stfp->bitmap_program;
+}
+
+
+/**
+ * Copy user-provide bitmap bits into texture buffer, expanding
+ * bits into texels.
+ * "On" bits will set texels to 0xff.
+ * "Off" bits will not modify texels.
+ * Note that the image is actually going to be upside down in
+ * the texture.  We deal with that with texcoords.
+ */
+static void
+unpack_bitmap(struct st_context *st,
+              GLint px, GLint py, GLsizei width, GLsizei height,
+              const struct gl_pixelstore_attrib *unpack,
+              const GLubyte *bitmap,
+              ubyte *destBuffer, uint destStride)
+{
+   GLint row, col;
+
+#define SET_PIXEL(COL, ROW) \
+   destBuffer[(py + (ROW)) * destStride + px + (COL)] = 0x0;
+
+   for (row = 0; row < height; row++) {
+      const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack,
+                 bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0);
+
+      if (unpack->LsbFirst) {
+         /* Lsb first */
+         GLubyte mask = 1U << (unpack->SkipPixels & 0x7);
+         for (col = 0; col < width; col++) {
+
+            if (*src & mask) {
+               SET_PIXEL(col, row);
+            }
+
+            if (mask == 128U) {
+               src++;
+               mask = 1U;
+            }
+            else {
+               mask = mask << 1;
+            }
+         }
+
+         /* get ready for next row */
+         if (mask != 1)
+            src++;
+      }
+      else {
+         /* Msb first */
+         GLubyte mask = 128U >> (unpack->SkipPixels & 0x7);
+         for (col = 0; col < width; col++) {
+
+            if (*src & mask) {
+               SET_PIXEL(col, row);
+            }
+
+            if (mask == 1U) {
+               src++;
+               mask = 128U;
+            }
+            else {
+               mask = mask >> 1;
+            }
+         }
+
+         /* get ready for next row */
+         if (mask != 128)
+            src++;
+      }
+
+   } /* row */
+
+#undef SET_PIXEL
+}
+
+
+/**
+ * Create a texture which represents a bitmap image.
+ */
+static struct pipe_texture *
+make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height,
+                    const struct gl_pixelstore_attrib *unpack,
+                    const GLubyte *bitmap)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surface;
+   ubyte *dest;
+   struct pipe_texture *pt;
+
+   /* PBO source... */
+   bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap);
+   if (!bitmap) {
+      return NULL;
+   }
+
+   /**
+    * Create texture to hold bitmap pattern.
+    */
+   pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, ctx->st->bitmap.tex_format,
+                          0, width, height, 1, 0,
+                          PIPE_TEXTURE_USAGE_SAMPLER);
+   if (!pt) {
+      _mesa_unmap_bitmap_pbo(ctx, unpack);
+      return NULL;
+   }
+
+   surface = screen->get_tex_surface(screen, pt, 0, 0, 0,
+                                     PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* map texture surface */
+   dest = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* Put image into texture surface */
+   memset(dest, 0xff, height * surface->stride);
+   unpack_bitmap(ctx->st, 0, 0, width, height, unpack, bitmap,
+                 dest, surface->stride);
+
+   _mesa_unmap_bitmap_pbo(ctx, unpack);
+
+   /* Release surface */
+   screen->surface_unmap(screen, surface);
+   pipe_surface_reference(&surface, NULL);
+
+   return pt;
+}
+
+static GLuint
+setup_bitmap_vertex_data(struct st_context *st,
+                         int x, int y, int width, int height,
+                         float z, const float color[4])
+{
+   struct pipe_context *pipe = st->pipe;
+   const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+   const GLfloat fb_width = (GLfloat)fb->Width;
+   const GLfloat fb_height = (GLfloat)fb->Height;
+   const GLfloat x0 = (GLfloat)x;
+   const GLfloat x1 = (GLfloat)(x + width);
+   const GLfloat y0 = (GLfloat)y;
+   const GLfloat y1 = (GLfloat)(y + height);
+   const GLfloat sLeft = (GLfloat)0.0, sRight = (GLfloat)1.0;
+   const GLfloat tTop = (GLfloat)0.0, tBot = (GLfloat)1.0 - tTop;
+   const GLfloat clip_x0 = (GLfloat)(x0 / fb_width * 2.0 - 1.0);
+   const GLfloat clip_y0 = (GLfloat)(y0 / fb_height * 2.0 - 1.0);
+   const GLfloat clip_x1 = (GLfloat)(x1 / fb_width * 2.0 - 1.0);
+   const GLfloat clip_y1 = (GLfloat)(y1 / fb_height * 2.0 - 1.0);
+   const GLuint max_slots = 4096 / sizeof(st->bitmap.vertices);
+   GLuint i;
+
+   if (st->bitmap.vbuf_slot >= max_slots) {
+      pipe_buffer_reference(pipe->screen, &st->bitmap.vbuf, NULL);
+      st->bitmap.vbuf_slot = 0;
+   }
+
+   if (!st->bitmap.vbuf) {
+      st->bitmap.vbuf = pipe_buffer_create(pipe->screen, 32, 
+                                           PIPE_BUFFER_USAGE_VERTEX,
+                                           max_slots * sizeof(st->bitmap.vertices));
+   }
+
+   /* Positions are in clip coords since we need to do clipping in case
+    * the bitmap quad goes beyond the window bounds.
+    */
+   st->bitmap.vertices[0][0][0] = clip_x0;
+   st->bitmap.vertices[0][0][1] = clip_y0;
+   st->bitmap.vertices[0][2][0] = sLeft;
+   st->bitmap.vertices[0][2][1] = tTop;
+
+   st->bitmap.vertices[1][0][0] = clip_x1;
+   st->bitmap.vertices[1][0][1] = clip_y0;
+   st->bitmap.vertices[1][2][0] = sRight;
+   st->bitmap.vertices[1][2][1] = tTop;
+   
+   st->bitmap.vertices[2][0][0] = clip_x1;
+   st->bitmap.vertices[2][0][1] = clip_y1;
+   st->bitmap.vertices[2][2][0] = sRight;
+   st->bitmap.vertices[2][2][1] = tBot;
+   
+   st->bitmap.vertices[3][0][0] = clip_x0;
+   st->bitmap.vertices[3][0][1] = clip_y1;
+   st->bitmap.vertices[3][2][0] = sLeft;
+   st->bitmap.vertices[3][2][1] = tBot;
+   
+   /* same for all verts: */
+   for (i = 0; i < 4; i++) {
+      st->bitmap.vertices[i][0][2] = z;
+      st->bitmap.vertices[i][0][3] = 1.0;
+      st->bitmap.vertices[i][1][0] = color[0];
+      st->bitmap.vertices[i][1][1] = color[1];
+      st->bitmap.vertices[i][1][2] = color[2];
+      st->bitmap.vertices[i][1][3] = color[3];
+      st->bitmap.vertices[i][2][2] = 0.0; /*R*/
+      st->bitmap.vertices[i][2][3] = 1.0; /*Q*/
+   }
+
+   /* put vertex data into vbuf */
+   {
+      char *buf = pipe_buffer_map(pipe->screen, 
+                                  st->bitmap.vbuf, 
+                                  PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      memcpy(buf + st->bitmap.vbuf_slot * sizeof st->bitmap.vertices, 
+             st->bitmap.vertices, 
+             sizeof st->bitmap.vertices);
+
+      pipe_buffer_unmap(pipe->screen, st->bitmap.vbuf);
+   }
+
+   return st->bitmap.vbuf_slot++ * sizeof st->bitmap.vertices;
+}
+
+
+
+/**
+ * Render a glBitmap by drawing a textured quad
+ */
+static void
+draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
+                 GLsizei width, GLsizei height,
+                 struct pipe_texture *pt,
+                 const GLfloat *color)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct cso_context *cso = ctx->st->cso_context;
+   struct st_fragment_program *stfp;
+   GLuint maxSize;
+   GLuint offset;
+
+   stfp = combined_bitmap_fragment_program(ctx);
+
+   /* As an optimization, Mesa's fragment programs will sometimes get the
+    * primary color from a statevar/constant rather than a varying variable.
+    * when that's the case, we need to ensure that we use the 'color'
+    * parameter and not the current attribute color (which may have changed
+    * through glRasterPos and state validation.
+    * So, we force the proper color here.  Not elegant, but it works.
+    */
+   {
+      GLfloat colorSave[4];
+      COPY_4V(colorSave, ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
+      COPY_4V(ctx->Current.Attrib[VERT_ATTRIB_COLOR0], color);
+      st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT);
+      COPY_4V(ctx->Current.Attrib[VERT_ATTRIB_COLOR0], colorSave);
+   }
+
+
+   /* limit checks */
+   /* XXX if the bitmap is larger than the max texture size, break
+    * it up into chunks.
+    */
+   maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
+   assert(width <= (GLsizei)maxSize);
+   assert(height <= (GLsizei)maxSize);
+
+   cso_save_rasterizer(cso);
+   cso_save_samplers(cso);
+   cso_save_sampler_textures(cso);
+   cso_save_viewport(cso);
+   cso_save_fragment_shader(cso);
+   cso_save_vertex_shader(cso);
+
+   /* rasterizer state: just scissor */
+   st->bitmap.rasterizer.scissor = ctx->Scissor.Enabled;
+   cso_set_rasterizer(cso, &st->bitmap.rasterizer);
+
+   /* fragment shader state: TEX lookup program */
+   cso_set_fragment_shader_handle(cso, stfp->driver_shader);
+
+   /* vertex shader state: position + texcoord pass-through */
+   cso_set_vertex_shader_handle(cso, st->bitmap.vs);
+
+   /* user samplers, plus our bitmap sampler */
+   {
+      struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+      uint num = MAX2(stfp->bitmap_sampler + 1, st->state.num_samplers);
+      uint i;
+      for (i = 0; i < st->state.num_samplers; i++) {
+         samplers[i] = &st->state.samplers[i];
+      }
+      samplers[stfp->bitmap_sampler] = &st->bitmap.sampler;
+      cso_set_samplers(cso, num, (const struct pipe_sampler_state **) samplers);
+   }
+
+   /* user textures, plus the bitmap texture */
+   {
+      struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+      uint num = MAX2(stfp->bitmap_sampler + 1, st->state.num_textures);
+      memcpy(textures, st->state.sampler_texture, sizeof(textures));
+      textures[stfp->bitmap_sampler] = pt;
+      cso_set_sampler_textures(cso, num, textures);
+   }
+
+   /* viewport state: viewport matching window dims */
+   {
+      const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+      const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP);
+      const GLfloat width = (GLfloat)fb->Width;
+      const GLfloat height = (GLfloat)fb->Height;
+      struct pipe_viewport_state vp;
+      vp.scale[0] =  0.5f * width;
+      vp.scale[1] = height * (invert ? -0.5f : 0.5f);
+      vp.scale[2] = 1.0f;
+      vp.scale[3] = 1.0f;
+      vp.translate[0] = 0.5f * width;
+      vp.translate[1] = 0.5f * height;
+      vp.translate[2] = 0.0f;
+      vp.translate[3] = 0.0f;
+      cso_set_viewport(cso, &vp);
+   }
+
+   /* draw textured quad */
+   offset = setup_bitmap_vertex_data(st, x, y, width, height,
+                                     ctx->Current.RasterPos[2],
+                                     color);
+
+   util_draw_vertex_buffer(pipe, st->bitmap.vbuf, offset,
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           3); /* attribs/vert */
+
+
+   /* restore state */
+   cso_restore_rasterizer(cso);
+   cso_restore_samplers(cso);
+   cso_restore_sampler_textures(cso);
+   cso_restore_viewport(cso);
+   cso_restore_fragment_shader(cso);
+   cso_restore_vertex_shader(cso);
+}
+
+
+static void
+reset_cache(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct bitmap_cache *cache = st->bitmap.cache;
+
+   //memset(cache->buffer, 0xff, sizeof(cache->buffer));
+   cache->empty = GL_TRUE;
+
+   cache->xmin = 1000000;
+   cache->xmax = -1000000;
+   cache->ymin = 1000000;
+   cache->ymax = -1000000;
+
+   if (cache->surf)
+      screen->tex_surface_release(screen, &cache->surf);
+
+   assert(!cache->texture);
+
+   /* allocate a new texture */
+   cache->texture = st_texture_create(st, PIPE_TEXTURE_2D,
+                                      st->bitmap.tex_format, 0,
+                                      BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT,
+                                      1, 0,
+                                      PIPE_TEXTURE_USAGE_SAMPLER);
+
+   /* Map the texture surface.
+    * Subsequent glBitmap calls will write into the texture image.
+    */
+   cache->surf = screen->get_tex_surface(screen, cache->texture, 0, 0, 0,
+                                         PIPE_BUFFER_USAGE_CPU_WRITE);
+   cache->buffer = screen->surface_map(screen, cache->surf,
+                                       PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* init image to all 0xff */
+   memset(cache->buffer, 0xff, BITMAP_CACHE_WIDTH * BITMAP_CACHE_HEIGHT);
+}
+
+
+/**
+ * If there's anything in the bitmap cache, draw/flush it now.
+ */
+void
+st_flush_bitmap_cache(struct st_context *st)
+{
+   if (!st->bitmap.cache->empty) {
+      struct bitmap_cache *cache = st->bitmap.cache;
+
+      if (st->ctx->DrawBuffer) {
+         struct pipe_context *pipe = st->pipe;
+         struct pipe_screen *screen = pipe->screen;
+
+         assert(cache->xmin <= cache->xmax);
+ 
+/*         printf("flush size %d x %d  at %d, %d\n",
+                cache->xmax - cache->xmin,
+                cache->ymax - cache->ymin,
+                cache->xpos, cache->ypos);
+*/
+
+         /* The texture surface has been mapped until now.
+          * So unmap and release the texture surface before drawing.
+          */
+         screen->surface_unmap(screen, cache->surf);
+         cache->buffer = NULL;
+
+         screen->tex_surface_release(screen, &cache->surf);
+
+         draw_bitmap_quad(st->ctx,
+                          cache->xpos,
+                          cache->ypos,
+                          st->ctx->Current.RasterPos[2],
+                          BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT,
+                          cache->texture,
+                          cache->color);
+      }
+
+      /* release/free the texture */
+      pipe_texture_reference(&cache->texture, NULL);
+
+      reset_cache(st);
+   }
+}
+
+/* Flush bitmap cache and release vertex buffer.
+ */
+void
+st_flush_bitmap( struct st_context *st )
+{
+   st_flush_bitmap_cache(st);
+
+   /* Release vertex buffer to avoid synchronous rendering if we were
+    * to map it in the next frame.
+    */
+   pipe_buffer_reference(st->pipe->screen, &st->bitmap.vbuf, NULL);
+   st->bitmap.vbuf_slot = 0;
+}
+
+
+/**
+ * Try to accumulate this glBitmap call in the bitmap cache.
+ * \return  GL_TRUE for success, GL_FALSE if bitmap is too large, etc.
+ */
+static GLboolean
+accum_bitmap(struct st_context *st,
+             GLint x, GLint y, GLsizei width, GLsizei height,
+             const struct gl_pixelstore_attrib *unpack,
+             const GLubyte *bitmap )
+{
+   struct bitmap_cache *cache = st->bitmap.cache;
+   int px = -999, py;
+
+   if (width > BITMAP_CACHE_WIDTH ||
+       height > BITMAP_CACHE_HEIGHT)
+      return GL_FALSE; /* too big to cache */
+
+   if (!cache->empty) {
+      px = x - cache->xpos;  /* pos in buffer */
+      py = y - cache->ypos;
+      if (px < 0 || px + width > BITMAP_CACHE_WIDTH ||
+          py < 0 || py + height > BITMAP_CACHE_HEIGHT ||
+          !TEST_EQ_4V(st->ctx->Current.RasterColor, cache->color)) {
+         /* This bitmap would extend beyond cache bounds, or the bitmap
+          * color is changing
+          * so flush and continue.
+          */
+         st_flush_bitmap_cache(st);
+      }
+   }
+
+   if (cache->empty) {
+      /* Initialize.  Center bitmap vertically in the buffer. */
+      px = 0;
+      py = (BITMAP_CACHE_HEIGHT - height) / 2;
+      cache->xpos = x;
+      cache->ypos = y - py;
+      cache->empty = GL_FALSE;
+      COPY_4FV(cache->color, st->ctx->Current.RasterColor);
+   }
+
+   assert(px != -999);
+
+   if (x < cache->xmin)
+      cache->xmin = x;
+   if (y < cache->ymin)
+      cache->ymin = y;
+   if (x + width > cache->xmax)
+      cache->xmax = x + width;
+   if (y + height > cache->ymax)
+      cache->ymax = y + height;
+
+   unpack_bitmap(st, px, py, width, height, unpack, bitmap,
+                 cache->buffer, BITMAP_CACHE_WIDTH);
+
+   return GL_TRUE; /* accumulated */
+}
+
+
+
+/**
+ * Called via ctx->Driver.Bitmap()
+ */
+static void
+st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
+          const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap )
+{
+   struct st_context *st = ctx->st;
+   struct pipe_texture *pt;
+
+   if (width == 0 || height == 0)
+      return;
+
+   st_validate_state(st);
+
+   if (!st->bitmap.vs) {
+      /* create pass-through vertex shader now */
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_COLOR,
+                                      TGSI_SEMANTIC_GENERIC };
+      const uint semantic_indexes[] = { 0, 0, 0 };
+      st->bitmap.vs = util_make_vertex_passthrough_shader(st->pipe, 3,
+                                                          semantic_names,
+                                                          semantic_indexes,
+                                                          &st->bitmap.vert_shader);
+   }
+
+   if (UseBitmapCache && accum_bitmap(st, x, y, width, height, unpack, bitmap))
+      return;
+
+   pt = make_bitmap_texture(ctx, width, height, unpack, bitmap);
+   if (pt) {
+      assert(pt->target == PIPE_TEXTURE_2D);
+      draw_bitmap_quad(ctx, x, y, ctx->Current.RasterPos[2],
+                       width, height, pt,
+                       st->ctx->Current.RasterColor);
+      /* release/free the texture */
+      pipe_texture_reference(&pt, NULL);
+   }
+}
+
+
+/** Per-context init */
+void
+st_init_bitmap_functions(struct dd_function_table *functions)
+{
+   functions->Bitmap = st_Bitmap;
+}
+
+
+/** Per-context init */
+void
+st_init_bitmap(struct st_context *st)
+{
+   struct pipe_sampler_state *sampler = &st->bitmap.sampler;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+
+   /* init sampler state once */
+   memset(sampler, 0, sizeof(*sampler));
+   sampler->wrap_s = PIPE_TEX_WRAP_CLAMP;
+   sampler->wrap_t = PIPE_TEX_WRAP_CLAMP;
+   sampler->wrap_r = PIPE_TEX_WRAP_CLAMP;
+   sampler->min_img_filter = PIPE_TEX_FILTER_NEAREST;
+   sampler->min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   sampler->mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+   sampler->normalized_coords = 1;
+
+   /* init baseline rasterizer state once */
+   memset(&st->bitmap.rasterizer, 0, sizeof(st->bitmap.rasterizer));
+   st->bitmap.rasterizer.gl_rasterization_rules = 1;
+   st->bitmap.rasterizer.bypass_vs = 1;
+
+   /* find a usable texture format */
+   if (screen->is_format_supported(screen, PIPE_FORMAT_I8_UNORM, PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      st->bitmap.tex_format = PIPE_FORMAT_I8_UNORM;
+   }
+   else {
+      /* XXX support more formats */
+      assert(0);
+   }
+
+   /* alloc bitmap cache object */
+   st->bitmap.cache = CALLOC_STRUCT(bitmap_cache);
+
+   reset_cache(st);
+}
+
+
+/** Per-context tear-down */
+void
+st_destroy_bitmap(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct bitmap_cache *cache = st->bitmap.cache;
+
+   screen->surface_unmap(screen, cache->surf);
+   screen->tex_surface_release(screen, &cache->surf);
+
+   if (st->bitmap.vs) {
+      cso_delete_vertex_shader(st->cso_context, st->bitmap.vs);
+      st->bitmap.vs = NULL;
+   }
+
+   if (st->bitmap.vbuf) {
+      pipe_buffer_reference(pipe->screen, &st->bitmap.vbuf, NULL);
+      st->bitmap.vbuf = NULL;
+   }
+
+   if (st->bitmap.cache) {
+      pipe_texture_release(&st->bitmap.cache->texture);
+      FREE(st->bitmap.cache);
+      st->bitmap.cache = NULL;
+   }
+}
diff --git a/src/mesa/state_tracker/st_cb_bitmap.h b/src/mesa/state_tracker/st_cb_bitmap.h
new file mode 100644
index 0000000000..81cf61981d
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_bitmap.h
@@ -0,0 +1,52 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_BITMAP_H
+#define ST_CB_BITMAP_H
+
+
+extern void
+st_init_bitmap_functions(struct dd_function_table *functions);
+
+extern void
+st_init_bitmap(struct st_context *st);
+
+extern void
+st_destroy_bitmap(struct st_context *st);
+
+extern void
+st_flush_bitmap_cache(struct st_context *st);
+
+/* Flush bitmap cache and release vertex buffer.  Needed at end of
+ * frame to avoid synchronous rendering.
+ */
+extern void
+st_flush_bitmap(struct st_context *st);
+
+
+#endif /* ST_CB_BITMAP_H */
diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c
new file mode 100644
index 0000000000..2852623472
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_blit.c
@@ -0,0 +1,111 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "main/texformat.h"
+#include "shader/program.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "st_context.h"
+#include "st_program.h"
+#include "st_cb_blit.h"
+#include "st_cb_fbo.h"
+
+#include "util/u_blit.h"
+
+#include "cso_cache/cso_context.h"
+
+
+void
+st_init_blit(struct st_context *st)
+{
+   st->blit = util_create_blit(st->pipe, st->cso_context);
+}
+
+
+void
+st_destroy_blit(struct st_context *st)
+{
+   util_destroy_blit(st->blit);
+   st->blit = NULL;
+}
+
+
+static void
+st_BlitFramebuffer(GLcontext *ctx,
+                   GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                   GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                   GLbitfield mask, GLenum filter)
+{
+   struct st_context *st = ctx->st;
+
+   const uint pFilter = ((filter == GL_NEAREST)
+                         ? PIPE_TEX_MIPFILTER_NEAREST
+                         : PIPE_TEX_MIPFILTER_LINEAR);
+
+   if (mask & GL_COLOR_BUFFER_BIT) {
+      struct st_renderbuffer *srcRb = 
+         st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer);
+      struct st_renderbuffer *dstRb = 
+         st_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
+      struct pipe_surface *srcSurf = srcRb->surface;
+      struct pipe_surface *dstSurf = dstRb->surface;
+
+      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+         /* invert Y */
+         srcY0 = srcRb->Base.Height - srcY0;
+         srcY1 = srcRb->Base.Height - srcY1;
+
+         dstY0 = dstRb->Base.Height - dstY0;
+         dstY1 = dstRb->Base.Height - dstY1;
+      }
+
+      util_blit_pixels(st->blit,
+                       srcSurf, srcX0, srcY0, srcX1, srcY1,
+                       dstSurf, dstX0, dstY0, dstX1, dstY1,
+                       0.0, pFilter);
+
+   }
+}
+
+
+
+void
+st_init_blit_functions(struct dd_function_table *functions)
+{
+#if FEATURE_EXT_framebuffer_blit
+   functions->BlitFramebuffer = st_BlitFramebuffer;
+#endif
+}
diff --git a/src/mesa/state_tracker/st_cb_blit.h b/src/mesa/state_tracker/st_cb_blit.h
new file mode 100644
index 0000000000..ed22986b53
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_blit.h
@@ -0,0 +1,46 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef ST_CB_BLIT_H
+#define ST_CB_BLIT_H
+
+
+#include "st_context.h"
+
+
+
+extern void
+st_init_blit(struct st_context *st);
+
+extern void
+st_destroy_blit(struct st_context *st);
+
+extern void
+st_init_blit_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_BLIT_H */
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c
new file mode 100644
index 0000000000..07fa2afce0
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.c
@@ -0,0 +1,240 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/imports.h"
+#include "main/mtypes.h"
+#include "main/bufferobj.h"
+
+#include "st_context.h"
+#include "st_cb_bufferobjects.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+
+
+/* Pixel buffers and Vertex/index buffers are handled through these
+ * mesa callbacks.  Framebuffer/Renderbuffer objects are
+ * created/managed elsewhere.
+ */
+
+
+
+/**
+ * There is some duplication between mesa's bufferobjects and our
+ * bufmgr buffers.  Both have an integer handle and a hashtable to
+ * lookup an opaque structure.  It would be nice if the handles and
+ * internal structure where somehow shared.
+ */
+static struct gl_buffer_object *
+st_bufferobj_alloc(GLcontext *ctx, GLuint name, GLenum target)
+{
+   struct st_buffer_object *st_obj = CALLOC_STRUCT(st_buffer_object);
+
+   if (!st_obj)
+      return NULL;
+
+   _mesa_initialize_buffer_object(&st_obj->Base, name, target);
+
+   return &st_obj->Base;
+}
+
+
+
+/**
+ * Deallocate/free a vertex/pixel buffer object.
+ * Called via glDeleteBuffersARB().
+ */
+static void
+st_bufferobj_free(GLcontext *ctx, struct gl_buffer_object *obj)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+
+   if (st_obj->buffer) 
+      pipe_buffer_reference(pipe->screen, &st_obj->buffer, NULL);
+
+   free(st_obj);
+}
+
+
+
+/**
+ * Replace data in a subrange of buffer object.  If the data range
+ * specified by size + offset extends beyond the end of the buffer or
+ * if data is NULL, no copy is performed.
+ * Called via glBufferSubDataARB().
+ */
+static void
+st_bufferobj_subdata(GLcontext *ctx,
+		     GLenum target,
+		     GLintptrARB offset,
+		     GLsizeiptrARB size,
+		     const GLvoid * data, struct gl_buffer_object *obj)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+   char *map;
+
+   if (offset >= st_obj->size || size > (st_obj->size - offset))
+      return;
+
+   map = pipe_buffer_map(pipe->screen, st_obj->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+   memcpy(map + offset, data, size);
+   pipe_buffer_unmap(pipe->screen, st_obj->buffer);
+}
+
+
+/**
+ * Called via glGetBufferSubDataARB().
+ */
+static void
+st_bufferobj_get_subdata(GLcontext *ctx,
+                         GLenum target,
+                         GLintptrARB offset,
+                         GLsizeiptrARB size,
+                         GLvoid * data, struct gl_buffer_object *obj)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+   char *map;
+
+   if (offset >= st_obj->size || size > (st_obj->size - offset))
+      return;
+
+   map = pipe_buffer_map(pipe->screen, st_obj->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+   memcpy(data, map + offset, size);
+   pipe_buffer_unmap(pipe->screen, st_obj->buffer);
+}
+
+
+/**
+ * Allocate space for and store data in a buffer object.  Any data that was
+ * previously stored in the buffer object is lost.  If data is NULL,
+ * memory will be allocated, but no copy will occur.
+ * Called via glBufferDataARB().
+ */
+static void
+st_bufferobj_data(GLcontext *ctx,
+		  GLenum target,
+		  GLsizeiptrARB size,
+		  const GLvoid * data,
+		  GLenum usage, 
+		  struct gl_buffer_object *obj)
+{
+   struct st_context *st = st_context(ctx);
+   struct pipe_context *pipe = st->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+   unsigned buffer_usage;
+
+   st_obj->Base.Size = size;
+   st_obj->Base.Usage = usage;
+   
+   switch(target) {
+   case GL_PIXEL_PACK_BUFFER_ARB:
+   case GL_PIXEL_UNPACK_BUFFER_ARB:
+      buffer_usage = PIPE_BUFFER_USAGE_PIXEL;
+      break;
+   case GL_ARRAY_BUFFER_ARB:
+      buffer_usage = PIPE_BUFFER_USAGE_VERTEX;
+      break;
+   case GL_ELEMENT_ARRAY_BUFFER_ARB:
+      buffer_usage = PIPE_BUFFER_USAGE_INDEX;
+      break;
+   default:
+      buffer_usage = 0;
+   }
+
+   pipe_buffer_reference( pipe->screen, &st_obj->buffer, NULL );
+
+   st_obj->buffer = pipe_buffer_create( pipe->screen, 32, buffer_usage, size );
+
+   st_obj->size = size;
+
+   if (data)
+      st_bufferobj_subdata(ctx, target, 0, size, data, obj);
+}
+
+
+/**
+ * Called via glMapBufferARB().
+ */
+static void *
+st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
+                 struct gl_buffer_object *obj)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+   GLuint flags;
+
+   switch (access) {
+   case GL_WRITE_ONLY:
+      flags = PIPE_BUFFER_USAGE_CPU_WRITE;
+      break;
+   case GL_READ_ONLY:
+      flags = PIPE_BUFFER_USAGE_CPU_READ;
+      break;
+   case GL_READ_WRITE:
+      /* fall-through */
+   default:
+      flags = PIPE_BUFFER_USAGE_CPU_READ | PIPE_BUFFER_USAGE_CPU_WRITE;
+      break;      
+   }
+
+   obj->Pointer = pipe_buffer_map(pipe->screen, st_obj->buffer, flags);
+   return obj->Pointer;
+}
+
+
+/**
+ * Called via glMapBufferARB().
+ */
+static GLboolean
+st_bufferobj_unmap(GLcontext *ctx, GLenum target, struct gl_buffer_object *obj)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *st_obj = st_buffer_object(obj);
+
+   pipe_buffer_unmap(pipe->screen, st_obj->buffer);
+   obj->Pointer = NULL;
+   return GL_TRUE;
+}
+
+
+void
+st_init_bufferobject_functions(struct dd_function_table *functions)
+{
+   functions->NewBufferObject = st_bufferobj_alloc;
+   functions->DeleteBuffer = st_bufferobj_free;
+   functions->BufferData = st_bufferobj_data;
+   functions->BufferSubData = st_bufferobj_subdata;
+   functions->GetBufferSubData = st_bufferobj_get_subdata;
+   functions->MapBuffer = st_bufferobj_map;
+   functions->UnmapBuffer = st_bufferobj_unmap;
+}
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.h b/src/mesa/state_tracker/st_cb_bufferobjects.h
new file mode 100644
index 0000000000..dcbb5a5233
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.h
@@ -0,0 +1,66 @@
+/**************************************************************************
+ * 
+ * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef ST_CB_BUFFEROBJECTS_H
+#define ST_CB_BUFFEROBJECTS_H
+
+struct st_context;
+struct gl_buffer_object;
+struct pipe_buffer;
+
+/**
+ * State_tracker vertex/pixel buffer object, derived from Mesa's
+ * gl_buffer_object.
+ */
+struct st_buffer_object
+{
+   struct gl_buffer_object Base;
+   struct pipe_buffer *buffer;  
+   GLsizeiptrARB size;
+};
+
+
+/* Are the obj->Name tests necessary?  Unfortunately yes, mesa
+ * allocates a couple of gl_buffer_object structs statically, and the
+ * Name == 0 test is the only way to identify them and avoid casting
+ * them erroneously to our structs.
+ */
+static INLINE struct st_buffer_object *
+st_buffer_object(struct gl_buffer_object *obj)
+{
+   if (obj->Name)
+      return (struct st_buffer_object *) obj;
+   else
+      return NULL;
+}
+
+
+extern void
+st_init_bufferobject_functions(struct dd_function_table *functions);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c
new file mode 100644
index 0000000000..668c3f9ebf
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_clear.c
@@ -0,0 +1,602 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+#include "main/glheader.h"
+#include "main/macros.h"
+#include "shader/prog_instruction.h"
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_accum.h"
+#include "st_cb_clear.h"
+#include "st_cb_fbo.h"
+#include "st_draw.h"
+#include "st_program.h"
+#include "st_public.h"
+#include "st_mesa_to_tgsi.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "util/u_pack_color.h"
+#include "util/u_simple_shaders.h"
+#include "util/u_draw_quad.h"
+
+#include "cso_cache/cso_context.h"
+
+
+void
+st_init_clear(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   /* rasterizer state: bypass clipping */
+   memset(&st->clear.raster, 0, sizeof(st->clear.raster));
+   st->clear.raster.gl_rasterization_rules = 1;
+   st->clear.raster.bypass_clipping = 1;
+
+   /* viewport state: identity since we're drawing in window coords */
+   st->clear.viewport.scale[0] = 1.0;
+   st->clear.viewport.scale[1] = 1.0;
+   st->clear.viewport.scale[2] = 1.0;
+   st->clear.viewport.scale[3] = 1.0;
+   st->clear.viewport.translate[0] = 0.0;
+   st->clear.viewport.translate[1] = 0.0;
+   st->clear.viewport.translate[2] = 0.0;
+   st->clear.viewport.translate[3] = 0.0;
+
+   /* fragment shader state: color pass-through program */
+   st->clear.fs =
+      util_make_fragment_passthrough_shader(pipe, &st->clear.frag_shader);
+
+   /* vertex shader state: color/position pass-through */
+   {
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_COLOR };
+      const uint semantic_indexes[] = { 0, 0 };
+      st->clear.vs = util_make_vertex_passthrough_shader(pipe, 2,
+                                                         semantic_names,
+                                                         semantic_indexes,
+                                                         &st->clear.vert_shader);
+   }
+}
+
+
+void
+st_destroy_clear(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   if (st->clear.vert_shader.tokens) {
+      FREE((void *) st->clear.vert_shader.tokens);
+      st->clear.vert_shader.tokens = NULL;
+   }
+
+   if (st->clear.frag_shader.tokens) {
+      FREE((void *) st->clear.frag_shader.tokens);
+      st->clear.frag_shader.tokens = NULL;
+   }
+
+   if (st->clear.fs) {
+      cso_delete_fragment_shader(st->cso_context, st->clear.fs);
+      st->clear.fs = NULL;
+   }
+   if (st->clear.vs) {
+      cso_delete_vertex_shader(st->cso_context, st->clear.vs);
+      st->clear.vs = NULL;
+   }
+   if (st->clear.vbuf) {
+      pipe_buffer_reference(pipe->screen, &st->clear.vbuf, NULL);
+      st->clear.vbuf = NULL;
+   }
+}
+
+
+static GLboolean
+is_depth_stencil_format(enum pipe_format pipeFormat)
+{
+   switch (pipeFormat) {
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_Z24S8_UNORM:
+      return GL_TRUE;
+   default:
+      return GL_FALSE;
+   }
+}
+
+
+
+/**
+ * Draw a screen-aligned quadrilateral.
+ * Coords are window coords with y=0=bottom.  These coords will be transformed
+ * by the vertex shader and viewport transform (which will flip Y if needed).
+ */
+static void
+draw_quad(GLcontext *ctx,
+          float x0, float y0, float x1, float y1, GLfloat z,
+          const GLfloat color[4])
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   const GLuint max_slots = 1024 / sizeof(st->clear.vertices);
+   GLuint i;
+   void *buf;
+
+   if (st->clear.vbuf_slot >= max_slots) {
+      pipe_buffer_reference(pipe->screen, &st->clear.vbuf, NULL);
+      st->clear.vbuf_slot = 0;
+   }
+
+   if (!st->clear.vbuf) {
+      st->clear.vbuf = pipe_buffer_create(pipe->screen, 32, PIPE_BUFFER_USAGE_VERTEX,
+                                          max_slots * sizeof(st->clear.vertices));
+   }
+
+   /* positions */
+   st->clear.vertices[0][0][0] = x0;
+   st->clear.vertices[0][0][1] = y0;
+
+   st->clear.vertices[1][0][0] = x1;
+   st->clear.vertices[1][0][1] = y0;
+
+   st->clear.vertices[2][0][0] = x1;
+   st->clear.vertices[2][0][1] = y1;
+
+   st->clear.vertices[3][0][0] = x0;
+   st->clear.vertices[3][0][1] = y1;
+
+   /* same for all verts: */
+   for (i = 0; i < 4; i++) {
+      st->clear.vertices[i][0][2] = z;
+      st->clear.vertices[i][0][3] = 1.0;
+      st->clear.vertices[i][1][0] = color[0];
+      st->clear.vertices[i][1][1] = color[1];
+      st->clear.vertices[i][1][2] = color[2];
+      st->clear.vertices[i][1][3] = color[3];
+   }
+
+   /* put vertex data into vbuf */
+   buf = pipe_buffer_map(pipe->screen, st->clear.vbuf, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   memcpy((char *)buf + st->clear.vbuf_slot * sizeof(st->clear.vertices), 
+          st->clear.vertices, 
+          sizeof(st->clear.vertices));
+
+   pipe_buffer_unmap(pipe->screen, st->clear.vbuf);
+
+   /* draw */
+   util_draw_vertex_buffer(pipe, 
+                           st->clear.vbuf, 
+                           st->clear.vbuf_slot * sizeof(st->clear.vertices),
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           2); /* attribs/vert */
+
+   /* Increment slot */
+   st->clear.vbuf_slot++;
+}
+
+
+
+/**
+ * Do glClear by drawing a quadrilateral.
+ * The vertices of the quad will be computed from the
+ * ctx->DrawBuffer->_X/Ymin/max fields.
+ */
+static void
+clear_with_quad(GLcontext *ctx,
+                GLboolean color, GLboolean depth, GLboolean stencil)
+{
+   struct st_context *st = ctx->st;
+   const GLfloat x0 = (GLfloat) ctx->DrawBuffer->_Xmin;
+   const GLfloat x1 = (GLfloat) ctx->DrawBuffer->_Xmax;
+   GLfloat y0, y1;
+
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+      y0 = (GLfloat) (ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax);
+      y1 = (GLfloat) (ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin);
+   }
+   else {
+      y0 = (GLfloat) ctx->DrawBuffer->_Ymin;
+      y1 = (GLfloat) ctx->DrawBuffer->_Ymax;
+   }
+
+   /*
+   printf("%s %s%s%s %f,%f %f,%f\n", __FUNCTION__, 
+	  color ? "color, " : "",
+	  depth ? "depth, " : "",
+	  stencil ? "stencil" : "",
+	  x0, y0,
+	  x1, y1);
+   */
+
+   cso_save_blend(st->cso_context);
+   cso_save_depth_stencil_alpha(st->cso_context);
+   cso_save_rasterizer(st->cso_context);
+   cso_save_viewport(st->cso_context);
+   cso_save_fragment_shader(st->cso_context);
+   cso_save_vertex_shader(st->cso_context);
+
+   /* blend state: RGBA masking */
+   {
+      struct pipe_blend_state blend;
+      memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      if (color) {
+         if (ctx->Color.ColorMask[0])
+            blend.colormask |= PIPE_MASK_R;
+         if (ctx->Color.ColorMask[1])
+            blend.colormask |= PIPE_MASK_G;
+         if (ctx->Color.ColorMask[2])
+            blend.colormask |= PIPE_MASK_B;
+         if (ctx->Color.ColorMask[3])
+            blend.colormask |= PIPE_MASK_A;
+         if (st->ctx->Color.DitherFlag)
+            blend.dither = 1;
+      }
+      cso_set_blend(st->cso_context, &blend);
+   }
+
+   /* depth_stencil state: always pass/set to ref value */
+   {
+      struct pipe_depth_stencil_alpha_state depth_stencil;
+      memset(&depth_stencil, 0, sizeof(depth_stencil));
+      if (depth) {
+         depth_stencil.depth.enabled = 1;
+         depth_stencil.depth.writemask = 1;
+         depth_stencil.depth.func = PIPE_FUNC_ALWAYS;
+      }
+
+      if (stencil) {
+         depth_stencil.stencil[0].enabled = 1;
+         depth_stencil.stencil[0].func = PIPE_FUNC_ALWAYS;
+         depth_stencil.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE;
+         depth_stencil.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
+         depth_stencil.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE;
+         depth_stencil.stencil[0].ref_value = ctx->Stencil.Clear;
+         depth_stencil.stencil[0].valuemask = 0xff;
+         depth_stencil.stencil[0].writemask = ctx->Stencil.WriteMask[0] & 0xff;
+      }
+
+      cso_set_depth_stencil_alpha(st->cso_context, &depth_stencil);
+   }
+
+   cso_set_rasterizer(st->cso_context, &st->clear.raster);
+   cso_set_viewport(st->cso_context, &st->clear.viewport);
+
+   cso_set_fragment_shader_handle(st->cso_context, st->clear.fs);
+   cso_set_vertex_shader_handle(st->cso_context, st->clear.vs);
+
+   /* draw quad matching scissor rect (XXX verify coord round-off) */
+   draw_quad(ctx, x0, y0, x1, y1, (GLfloat) ctx->Depth.Clear, ctx->Color.ClearColor);
+
+   /* Restore pipe state */
+   cso_restore_blend(st->cso_context);
+   cso_restore_depth_stencil_alpha(st->cso_context);
+   cso_restore_rasterizer(st->cso_context);
+   cso_restore_viewport(st->cso_context);
+   cso_restore_fragment_shader(st->cso_context);
+   cso_restore_vertex_shader(st->cso_context);
+}
+
+
+/**
+ * Determine if we need to clear the depth buffer by drawing a quad.
+ */
+static INLINE GLboolean
+check_clear_color_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   const struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   if (!ctx->Color.ColorMask[0] ||
+       !ctx->Color.ColorMask[1] ||
+       !ctx->Color.ColorMask[2] ||
+       !ctx->Color.ColorMask[3])
+      return TRUE;
+
+   return FALSE;
+}
+
+
+static INLINE GLboolean
+check_clear_depth_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   const struct st_renderbuffer *strb = st_renderbuffer(rb);
+   const GLuint stencilMax = (1 << rb->StencilBits) - 1;
+   GLboolean maskStencil
+      = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   if (maskStencil)
+      return TRUE;
+
+   return FALSE;
+}
+
+
+/**
+ * Determine if we need to clear the depth buffer by drawing a quad.
+ */
+static INLINE GLboolean
+check_clear_depth_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   const struct st_renderbuffer *strb = st_renderbuffer(rb);
+   const GLboolean isDS = is_depth_stencil_format(strb->surface->format);
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   if (isDS && 
+       strb->surface->status == PIPE_SURFACE_STATUS_DEFINED &&
+       ctx->DrawBuffer->Visual.stencilBits > 0)
+      return TRUE;
+
+   return FALSE;
+}
+
+
+/**
+ * Determine if we need to clear the stencil buffer by drawing a quad.
+ */
+static INLINE GLboolean
+check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   const struct st_renderbuffer *strb = st_renderbuffer(rb);
+   const GLboolean isDS = is_depth_stencil_format(strb->surface->format);
+   const GLuint stencilMax = (1 << rb->StencilBits) - 1;
+   const GLboolean maskStencil
+      = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
+
+   if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED)
+      return FALSE;
+
+   if (maskStencil) 
+      return TRUE;
+
+   if (ctx->Scissor.Enabled)
+      return TRUE;
+
+   /* This is correct, but it is necessary to look at the depth clear
+    * value held in the surface when it comes time to issue the clear,
+    * rather than taking depth and stencil clear values from the
+    * current state.
+    */
+   if (isDS && 
+       strb->surface->status == PIPE_SURFACE_STATUS_DEFINED &&
+       ctx->DrawBuffer->Visual.depthBits > 0)
+      return TRUE;
+
+   return FALSE;
+}
+
+
+
+static void
+clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (!strb->surface)
+      return;
+
+   if (check_clear_color_with_quad( ctx, rb )) {
+      /* masking or scissoring */
+      clear_with_quad(ctx, GL_TRUE, GL_FALSE, GL_FALSE);
+   }
+   else {
+      /* clear whole buffer w/out masking */
+      uint clearValue;
+      /* NOTE: we always pass the clear color as PIPE_FORMAT_A8R8G8B8_UNORM
+       * at this time!
+       */
+      util_pack_color(ctx->Color.ClearColor, PIPE_FORMAT_A8R8G8B8_UNORM, &clearValue);
+      ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue);
+   }
+}
+
+
+static void
+clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (!strb->surface)
+      return;
+
+   if (check_clear_depth_with_quad(ctx, rb)) {
+      /* scissoring or we have a combined depth/stencil buffer */
+      clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_FALSE);
+   }
+   else {
+      /* simple clear of whole buffer */
+      uint clearValue = util_pack_z(strb->surface->format, ctx->Depth.Clear);
+      ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue);
+   }
+}
+
+
+static void
+clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (!strb->surface)
+      return;
+
+   if (check_clear_stencil_with_quad(ctx, rb)) {
+      /* masking or scissoring or combined depth/stencil buffer */
+      clear_with_quad(ctx, GL_FALSE, GL_FALSE, GL_TRUE);
+   }
+   else {
+      /* simple clear of whole buffer */
+      GLuint clearValue = ctx->Stencil.Clear;
+
+      switch (strb->surface->format) {
+      case PIPE_FORMAT_S8Z24_UNORM:
+         clearValue <<= 24;
+         break;
+      default:
+         ; /* no-op, stencil value is in least significant bits */
+      }  
+
+      ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue);
+   }
+}
+
+
+static void
+clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+
+   if (!strb->surface)
+      return;
+
+   if (check_clear_depth_stencil_with_quad(ctx, rb)) {
+      /* masking or scissoring */
+      clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_TRUE);
+   }
+   else {
+      /* clear whole buffer w/out masking */
+      GLuint clearValue = util_pack_z(strb->surface->format, ctx->Depth.Clear);
+
+      switch (strb->surface->format) {
+      case PIPE_FORMAT_S8Z24_UNORM:
+         clearValue |= ctx->Stencil.Clear << 24;
+         break;
+      case PIPE_FORMAT_Z24S8_UNORM:
+         clearValue |= ctx->Stencil.Clear;
+         break;
+      default:
+         assert(0);
+      }  
+
+      ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue);
+   }
+}
+
+
+void st_flush_clear( struct st_context *st )
+{
+   /* Release vertex buffer to avoid synchronous rendering if we were
+    * to map it in the next frame.
+    */
+   pipe_buffer_reference(st->pipe->screen, &st->clear.vbuf, NULL);
+   st->clear.vbuf_slot = 0;
+}
+ 
+
+
+/**
+ * Called via ctx->Driver.Clear()
+ * XXX: doesn't pick up the differences between front/back/left/right
+ * clears.  Need to sort that out...
+ */
+static void st_clear(GLcontext *ctx, GLbitfield mask)
+{
+   static const GLbitfield BUFFER_BITS_DS
+      = (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL);
+   struct st_context *st = ctx->st;
+   struct gl_renderbuffer *depthRb
+      = ctx->DrawBuffer->Attachment[BUFFER_DEPTH].Renderbuffer;
+   struct gl_renderbuffer *stencilRb
+      = ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
+   GLbitfield cmask = mask & BUFFER_BITS_COLOR;
+
+   /* This makes sure the softpipe has the latest scissor, etc values */
+   st_validate_state( st );
+
+   /*
+    * XXX TO-DO:
+    * If we're going to use clear_with_quad() for any reason, use it to
+    * clear as many other buffers as possible.
+    * As it is now, we sometimes call clear_with_quad() three times to clear
+    * color/depth/stencil individually...
+    */
+
+   if (cmask) {
+      GLuint b;
+      for (b = 0; cmask; b++) {
+         if (cmask & (1 << b)) {
+            struct gl_renderbuffer *rb
+               = ctx->DrawBuffer->Attachment[b].Renderbuffer;
+            assert(rb);
+            clear_color_buffer(ctx, rb);
+            cmask &= ~(1 << b); /* turn off bit */
+         }
+         assert(b < BUFFER_COUNT);
+      }
+   }
+
+   if (mask & BUFFER_BIT_ACCUM) {
+      st_clear_accum_buffer(ctx,
+                       ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer);
+   }
+
+   if ((mask & BUFFER_BITS_DS) == BUFFER_BITS_DS && depthRb == stencilRb) {
+      /* clearing combined depth + stencil */
+      clear_depth_stencil_buffer(ctx, depthRb);
+   }
+   else {
+      /* separate depth/stencil clears */
+      if (mask & BUFFER_BIT_DEPTH) {
+         clear_depth_buffer(ctx, depthRb);
+      }
+      if (mask & BUFFER_BIT_STENCIL) {
+         clear_stencil_buffer(ctx, stencilRb);
+      }
+   }
+}
+
+
+void st_init_clear_functions(struct dd_function_table *functions)
+{
+   functions->Clear = st_clear;
+}
diff --git a/src/mesa/state_tracker/st_cb_clear.h b/src/mesa/state_tracker/st_cb_clear.h
new file mode 100644
index 0000000000..bc035ac25c
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_clear.h
@@ -0,0 +1,49 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_CLEAR_H
+#define ST_CB_CLEAR_H
+
+
+extern void
+st_init_clear(struct st_context *st);
+
+
+extern void
+st_destroy_clear(struct st_context *st);
+
+extern void
+st_flush_clear(struct st_context *st);
+
+
+extern void
+st_init_clear_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_CLEAR_H */
+
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
new file mode 100644
index 0000000000..32bf21411d
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -0,0 +1,1104 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/bufferobj.h"
+#include "main/macros.h"
+#include "main/texformat.h"
+#include "main/state.h"
+#include "shader/program.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_atom_constbuf.h"
+#include "st_draw.h"
+#include "st_program.h"
+#include "st_cb_drawpixels.h"
+#include "st_cb_readpixels.h"
+#include "st_cb_fbo.h"
+#include "st_cb_texture.h"
+#include "st_draw.h"
+#include "st_format.h"
+#include "st_mesa_to_tgsi.h"
+#include "st_texture.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+#include "util/u_draw_quad.h"
+#include "shader/prog_instruction.h"
+#include "cso_cache/cso_context.h"
+
+
+/**
+ * Check if the given program is:
+ * 0: MOVE result.color, fragment.color;
+ * 1: END;
+ */
+static GLboolean
+is_passthrough_program(const struct gl_fragment_program *prog)
+{
+   if (prog->Base.NumInstructions == 2) {
+      const struct prog_instruction *inst = prog->Base.Instructions;
+      if (inst[0].Opcode == OPCODE_MOV &&
+          inst[1].Opcode == OPCODE_END &&
+          inst[0].DstReg.File == PROGRAM_OUTPUT &&
+          inst[0].DstReg.Index == FRAG_RESULT_COLR &&
+          inst[0].DstReg.WriteMask == WRITEMASK_XYZW &&
+          inst[0].SrcReg[0].File == PROGRAM_INPUT &&
+          inst[0].SrcReg[0].Index == FRAG_ATTRIB_COL0 &&
+          inst[0].SrcReg[0].Swizzle == SWIZZLE_XYZW) {
+         return GL_TRUE;
+      }
+   }
+   return GL_FALSE;
+}
+
+
+
+/**
+ * Make fragment shader for glDraw/CopyPixels.  This shader is made
+ * by combining the pixel transfer shader with the user-defined shader.
+ */
+static struct st_fragment_program *
+combined_drawpix_fragment_program(GLcontext *ctx)
+{
+   struct st_context *st = ctx->st;
+   struct st_fragment_program *stfp;
+
+   if (st->pixel_xfer.program->serialNo == st->pixel_xfer.xfer_prog_sn
+       && st->fp->serialNo == st->pixel_xfer.user_prog_sn) {
+      /* the pixel tranfer program has not changed and the user-defined
+       * program has not changed, so re-use the combined program.
+       */
+      stfp = st->pixel_xfer.combined_prog;
+   }
+   else {
+      /* Concatenate the pixel transfer program with the current user-
+       * defined program.
+       */
+      if (is_passthrough_program(&st->fp->Base)) {
+         stfp = (struct st_fragment_program *)
+            _mesa_clone_program(ctx, &st->pixel_xfer.program->Base.Base);
+      }
+      else {
+#if 0
+         printf("Base program:\n");
+         _mesa_print_program(&st->fp->Base.Base);
+         printf("DrawPix program:\n");
+         _mesa_print_program(&st->pixel_xfer.program->Base.Base);
+#endif
+         stfp = (struct st_fragment_program *)
+            _mesa_combine_programs(ctx,
+                                   &st->pixel_xfer.program->Base.Base,
+                                   &st->fp->Base.Base);
+      }
+
+#if 0
+      {
+         struct gl_program *p = &stfp->Base.Base;
+         printf("Combined DrawPixels program:\n");
+         _mesa_print_program(p);
+         printf("InputsRead: 0x%x\n", p->InputsRead);
+         printf("OutputsWritten: 0x%x\n", p->OutputsWritten);
+         _mesa_print_parameter_list(p->Parameters);
+      }
+#endif
+
+      /* translate to TGSI tokens */
+      st_translate_fragment_program(st, stfp, NULL);
+
+      /* save new program, update serial numbers */
+      st->pixel_xfer.xfer_prog_sn = st->pixel_xfer.program->serialNo;
+      st->pixel_xfer.user_prog_sn = st->fp->serialNo;
+      st->pixel_xfer.combined_prog_sn = stfp->serialNo;
+      st->pixel_xfer.combined_prog = stfp;
+   }
+
+   /* Ideally we'd have updated the pipe constants during the normal
+    * st/atom mechanism.  But we can't since this is specific to glDrawPixels.
+    */
+   st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT);
+
+   return stfp;
+}
+
+
+/**
+ * Create fragment shader that does a TEX() instruction to get a Z
+ * value, then writes to FRAG_RESULT_DEPR.
+ * Pass fragment color through as-is.
+ */
+static struct st_fragment_program *
+make_fragment_shader_z(struct st_context *st)
+{
+   GLcontext *ctx = st->ctx;
+   struct gl_program *p;
+   GLuint ic = 0;
+
+   if (st->drawpix.z_shader) {
+      return st->drawpix.z_shader;
+   }
+
+   /*
+    * Create shader now
+    */
+   p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
+   if (!p)
+      return NULL;
+
+   p->NumInstructions = 3;
+
+   p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
+   if (!p->Instructions) {
+      ctx->Driver.DeleteProgram(ctx, p);
+      return NULL;
+   }
+   _mesa_init_instructions(p->Instructions, p->NumInstructions);
+
+   /* TEX result.depth, fragment.texcoord[0], texture[0], 2D; */
+   p->Instructions[ic].Opcode = OPCODE_TEX;
+   p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
+   p->Instructions[ic].DstReg.Index = FRAG_RESULT_DEPR;
+   p->Instructions[ic].DstReg.WriteMask = WRITEMASK_Z;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
+   p->Instructions[ic].TexSrcUnit = 0;
+   p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+   ic++;
+
+   /* MOV result.color, fragment.color */
+   p->Instructions[ic].Opcode = OPCODE_MOV;
+   p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
+   p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_COL0;
+   ic++;
+
+   /* END; */
+   p->Instructions[ic++].Opcode = OPCODE_END;
+
+   assert(ic == p->NumInstructions);
+
+   p->InputsRead = FRAG_BIT_TEX0 | FRAG_BIT_COL0;
+   p->OutputsWritten = (1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR);
+   p->SamplersUsed = 0x1;  /* sampler 0 (bit 0) is used */
+
+   st->drawpix.z_shader = (struct st_fragment_program *) p;
+   st_translate_fragment_program(st, st->drawpix.z_shader, NULL);
+
+   return st->drawpix.z_shader;
+}
+
+
+
+/**
+ * Create a simple vertex shader that just passes through the
+ * vertex position and texcoord (and optionally, color).
+ */
+static struct st_vertex_program *
+st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor)
+{
+   GLcontext *ctx = st->ctx;
+   struct st_vertex_program *stvp;
+   struct gl_program *p;
+   GLuint ic = 0;
+
+   if (st->drawpix.vert_shaders[passColor])
+      return st->drawpix.vert_shaders[passColor];
+
+   /*
+    * Create shader now
+    */
+   p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0);
+   if (!p)
+      return NULL;
+
+   if (passColor)
+      p->NumInstructions = 4;
+   else
+      p->NumInstructions = 3;
+
+   p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
+   if (!p->Instructions) {
+      ctx->Driver.DeleteProgram(ctx, p);
+      return NULL;
+   }
+   _mesa_init_instructions(p->Instructions, p->NumInstructions);
+   /* MOV result.pos, vertex.pos; */
+   p->Instructions[0].Opcode = OPCODE_MOV;
+   p->Instructions[0].DstReg.File = PROGRAM_OUTPUT;
+   p->Instructions[0].DstReg.Index = VERT_RESULT_HPOS;
+   p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[0].SrcReg[0].Index = VERT_ATTRIB_POS;
+   /* MOV result.texcoord0, vertex.texcoord0; */
+   p->Instructions[1].Opcode = OPCODE_MOV;
+   p->Instructions[1].DstReg.File = PROGRAM_OUTPUT;
+   p->Instructions[1].DstReg.Index = VERT_RESULT_TEX0;
+   p->Instructions[1].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[1].SrcReg[0].Index = VERT_ATTRIB_TEX0;
+   ic = 2;
+   if (passColor) {
+      /* MOV result.color0, vertex.color0; */
+      p->Instructions[ic].Opcode = OPCODE_MOV;
+      p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
+      p->Instructions[ic].DstReg.Index = VERT_RESULT_COL0;
+      p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
+      p->Instructions[ic].SrcReg[0].Index = VERT_ATTRIB_COLOR0;
+      ic++;
+   }
+
+   /* END; */
+   p->Instructions[ic].Opcode = OPCODE_END;
+   ic++;
+
+   assert(ic == p->NumInstructions);
+
+   p->InputsRead = VERT_BIT_POS | VERT_BIT_TEX0;
+   p->OutputsWritten = ((1 << VERT_RESULT_TEX0) |
+                        (1 << VERT_RESULT_HPOS));
+   if (passColor) {
+      p->InputsRead |= VERT_BIT_COLOR0;
+      p->OutputsWritten |= (1 << VERT_RESULT_COL0);
+   }
+
+   stvp = (struct st_vertex_program *) p;
+   st_translate_vertex_program(st, stvp, NULL, NULL, NULL);
+
+   st->drawpix.vert_shaders[passColor] = stvp;
+
+   return stvp;
+}
+
+
+static GLenum
+_mesa_base_format(GLenum format)
+{
+   switch (format) {
+   case GL_DEPTH_COMPONENT:
+      return GL_DEPTH_COMPONENT;
+   case GL_STENCIL_INDEX:
+      return GL_STENCIL_INDEX;
+   default:
+      return GL_RGBA;
+   }
+}
+
+
+/**
+ * Make texture containing an image for glDrawPixels image.
+ * If 'pixels' is NULL, leave the texture image data undefined.
+ */
+static struct pipe_texture *
+make_texture(struct st_context *st,
+	     GLsizei width, GLsizei height, GLenum format, GLenum type,
+	     const struct gl_pixelstore_attrib *unpack,
+	     const GLvoid *pixels)
+{
+   GLcontext *ctx = st->ctx;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const struct gl_texture_format *mformat;
+   struct pipe_texture *pt;
+   enum pipe_format pipeFormat;
+   GLuint cpp;
+   GLenum baseFormat;
+
+   baseFormat = _mesa_base_format(format);
+
+   mformat = st_ChooseTextureFormat(ctx, baseFormat, format, type);
+   assert(mformat);
+
+   pipeFormat = st_mesa_format_to_pipe_format(mformat->MesaFormat);
+   assert(pipeFormat);
+   cpp = st_sizeof_format(pipeFormat);
+
+   pixels = _mesa_map_drawpix_pbo(ctx, unpack, pixels);
+   if (!pixels)
+      return NULL;
+
+   pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height,
+			  1, 0,
+                          PIPE_TEXTURE_USAGE_SAMPLER);
+   if (!pt) {
+      _mesa_unmap_drawpix_pbo(ctx, unpack);
+      return NULL;
+   }
+
+   {
+      struct pipe_surface *surface;
+      static const GLuint dstImageOffsets = 0;
+      GLboolean success;
+      GLubyte *dest;
+      const GLbitfield imageTransferStateSave = ctx->_ImageTransferState;
+
+      /* we'll do pixel transfer in a fragment shader */
+      ctx->_ImageTransferState = 0x0;
+
+      surface = screen->get_tex_surface(screen, pt, 0, 0, 0,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      /* map texture surface */
+      dest = screen->surface_map(screen, surface,
+                                 PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      /* Put image into texture surface.
+       * Note that the image is actually going to be upside down in
+       * the texture.  We deal with that with texcoords.
+       */
+      success = mformat->StoreImage(ctx, 2,           /* dims */
+                                    baseFormat,       /* baseInternalFormat */
+                                    mformat,          /* gl_texture_format */
+                                    dest,             /* dest */
+                                    0, 0, 0,          /* dstX/Y/Zoffset */
+                                    surface->stride,  /* dstRowStride, bytes */
+                                    &dstImageOffsets, /* dstImageOffsets */
+                                    width, height, 1, /* size */
+                                    format, type,     /* src format/type */
+                                    pixels,           /* data source */
+                                    unpack);
+
+      /* unmap */
+      screen->surface_unmap(screen, surface);
+      pipe_surface_reference(&surface, NULL);
+
+      assert(success);
+
+      /* restore */
+      ctx->_ImageTransferState = imageTransferStateSave;
+   }
+
+   _mesa_unmap_drawpix_pbo(ctx, unpack);
+
+   return pt;
+}
+
+
+/**
+ * Draw quad with texcoords and optional color.
+ * Coords are window coords with y=0=bottom.
+ * \param color  may be null
+ * \param invertTex  if true, flip texcoords vertically
+ */
+static void
+draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z,
+          GLfloat x1, GLfloat y1, const GLfloat *color,
+          GLboolean invertTex)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = ctx->st->pipe;
+   GLfloat verts[4][3][4]; /* four verts, three attribs, XYZW */
+
+   /* setup vertex data */
+   {
+      const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+      const GLfloat fb_width = (GLfloat) fb->Width;
+      const GLfloat fb_height = (GLfloat) fb->Height;
+      const GLfloat clip_x0 = x0 / fb_width * 2.0f - 1.0f;
+      const GLfloat clip_y0 = y0 / fb_height * 2.0f - 1.0f;
+      const GLfloat clip_x1 = x1 / fb_width * 2.0f - 1.0f;
+      const GLfloat clip_y1 = y1 / fb_height * 2.0f - 1.0f;
+      const GLfloat sLeft = 0.0f, sRight = 1.0f;
+      const GLfloat tTop = invertTex, tBot = 1.0f - tTop;
+      GLuint tex, i;
+
+      /* upper-left */
+      verts[0][0][0] = clip_x0;    /* v[0].attr[0].x */
+      verts[0][0][1] = clip_y0;    /* v[0].attr[0].y */
+
+      /* upper-right */
+      verts[1][0][0] = clip_x1;
+      verts[1][0][1] = clip_y0;
+
+      /* lower-right */
+      verts[2][0][0] = clip_x1;
+      verts[2][0][1] = clip_y1;
+
+      /* lower-left */
+      verts[3][0][0] = clip_x0;
+      verts[3][0][1] = clip_y1;
+
+      tex = color ? 2 : 1;
+      verts[0][tex][0] = sLeft; /* v[0].attr[tex].s */
+      verts[0][tex][1] = tTop;  /* v[0].attr[tex].t */
+      verts[1][tex][0] = sRight;
+      verts[1][tex][1] = tTop;
+      verts[2][tex][0] = sRight;
+      verts[2][tex][1] = tBot;
+      verts[3][tex][0] = sLeft;
+      verts[3][tex][1] = tBot;
+
+      /* same for all verts: */
+      if (color) {
+         for (i = 0; i < 4; i++) {
+            verts[i][0][2] = z;   /*Z*/
+            verts[i][0][3] = 1.0f; /*W*/
+            verts[i][1][0] = color[0];
+            verts[i][1][1] = color[1];
+            verts[i][1][2] = color[2];
+            verts[i][1][3] = color[3];
+            verts[i][2][2] = 0.0f; /*R*/
+            verts[i][2][3] = 1.0f; /*Q*/
+         }
+      }
+      else {
+         for (i = 0; i < 4; i++) {
+            verts[i][0][2] = z;   /*Z*/
+            verts[i][0][3] = 1.0f; /*W*/
+            verts[i][1][2] = 0.0f; /*R*/
+            verts[i][1][3] = 1.0f; /*Q*/
+         }
+      }
+   }
+
+   {
+      struct pipe_buffer *buf;
+      ubyte *map;
+
+      /* allocate/load buffer object with vertex data */
+      buf = pipe_buffer_create(pipe->screen, 32, PIPE_BUFFER_USAGE_VERTEX,
+                               sizeof(verts));
+      map = pipe_buffer_map(pipe->screen, buf, PIPE_BUFFER_USAGE_CPU_WRITE);
+      memcpy(map, verts, sizeof(verts));
+      pipe_buffer_unmap(pipe->screen, buf);
+
+      util_draw_vertex_buffer(pipe, buf, 0,
+                              PIPE_PRIM_QUADS,
+                              4,  /* verts */
+                              3); /* attribs/vert */
+      pipe_buffer_reference(pipe->screen, &buf, NULL);
+   }
+}
+
+
+
+static void
+draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
+                   GLsizei width, GLsizei height,
+                   GLfloat zoomX, GLfloat zoomY,
+                   struct pipe_texture *pt,
+                   struct st_vertex_program *stvp,
+                   struct st_fragment_program *stfp,
+                   const GLfloat *color,
+                   GLboolean invertTex)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct cso_context *cso = ctx->st->cso_context;
+   GLfloat x0, y0, x1, y1;
+   GLsizei maxSize;
+
+   /* limit checks */
+   /* XXX if DrawPixels image is larger than max texture size, break
+    * it up into chunks.
+    */
+   maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
+   assert(width <= maxSize);
+   assert(height <= maxSize);
+
+   cso_save_rasterizer(cso);
+   cso_save_viewport(cso);
+   cso_save_samplers(cso);
+   cso_save_sampler_textures(cso);
+   cso_save_fragment_shader(cso);
+   cso_save_vertex_shader(cso);
+
+   /* rasterizer state: just scissor */
+   {
+      struct pipe_rasterizer_state rasterizer;
+      memset(&rasterizer, 0, sizeof(rasterizer));
+      rasterizer.gl_rasterization_rules = 1;
+      rasterizer.scissor = ctx->Scissor.Enabled;
+      cso_set_rasterizer(cso, &rasterizer);
+   }
+
+   /* fragment shader state: TEX lookup program */
+   cso_set_fragment_shader_handle(cso, stfp->driver_shader);
+
+   /* vertex shader state: position + texcoord pass-through */
+   cso_set_vertex_shader_handle(cso, stvp->driver_shader);
+
+
+   /* texture sampling state: */
+   {
+      struct pipe_sampler_state sampler;
+      memset(&sampler, 0, sizeof(sampler));
+      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP;
+      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP;
+      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP;
+      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.normalized_coords = 1;
+
+      cso_single_sampler(cso, 0, &sampler);
+      if (st->pixel_xfer.pixelmap_enabled) {
+         cso_single_sampler(cso, 1, &sampler);
+      }
+      cso_single_sampler_done(cso);
+   }
+
+   /* viewport state: viewport matching window dims */
+   {
+      const float width = (float) ctx->DrawBuffer->Width;
+      const float height = (float) ctx->DrawBuffer->Height;
+      struct pipe_viewport_state vp;
+      vp.scale[0] =  0.5f * width;
+      vp.scale[1] = -0.5f * height;
+      vp.scale[2] = 1.0f;
+      vp.scale[3] = 1.0f;
+      vp.translate[0] = 0.5f * width;
+      vp.translate[1] = 0.5f * height;
+      vp.translate[2] = 0.0f;
+      vp.translate[3] = 0.0f;
+      cso_set_viewport(cso, &vp);
+   }
+
+   /* texture state: */
+   if (st->pixel_xfer.pixelmap_enabled) {
+      struct pipe_texture *textures[2];
+      textures[0] = pt;
+      textures[1] = st->pixel_xfer.pixelmap_texture;
+      pipe->set_sampler_textures(pipe, 2, textures);
+   }
+   else {
+      pipe->set_sampler_textures(pipe, 1, &pt);
+   }
+
+   /* Compute window coords (y=0=bottom) with pixel zoom.
+    * Recall that these coords are transformed by the current
+    * vertex shader and viewport transformation.
+    */
+   x0 = (GLfloat) x;
+   x1 = x + width * ctx->Pixel.ZoomX;
+   y0 = (GLfloat) y;
+   y1 = y + height * ctx->Pixel.ZoomY;
+   //if(!color)
+   draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex);
+   //else
+   //printf("skip draw quad\n");
+   /* restore state */
+   cso_restore_rasterizer(cso);
+   cso_restore_viewport(cso);
+   cso_restore_samplers(cso);
+   cso_restore_sampler_textures(cso);
+   cso_restore_fragment_shader(cso);
+   cso_restore_vertex_shader(cso);
+}
+
+
+/**
+ * Check if a GL format/type combination is a match to the given pipe format.
+ * XXX probably move this to a re-usable place.
+ */
+static GLboolean
+compatible_formats(GLenum format, GLenum type, enum pipe_format pipeFormat)
+{
+   static const GLuint one = 1;
+   GLubyte littleEndian = *((GLubyte *) &one);
+
+   if (pipeFormat == PIPE_FORMAT_R8G8B8A8_UNORM &&
+       format == GL_RGBA &&
+       type == GL_UNSIGNED_BYTE &&
+       !littleEndian) {
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_R8G8B8A8_UNORM &&
+            format == GL_ABGR_EXT &&
+            type == GL_UNSIGNED_BYTE &&
+            littleEndian) {
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_A8R8G8B8_UNORM &&
+            format == GL_BGRA &&
+            type == GL_UNSIGNED_BYTE &&
+            littleEndian) {
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_R5G6B5_UNORM &&
+            format == GL_RGB &&
+            type == GL_UNSIGNED_SHORT_5_6_5) {
+      /* endian don't care */
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_R5G6B5_UNORM &&
+            format == GL_BGR &&
+            type == GL_UNSIGNED_SHORT_5_6_5_REV) {
+      /* endian don't care */
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_S8_UNORM &&
+            format == GL_STENCIL_INDEX &&
+            type == GL_UNSIGNED_BYTE) {
+      return GL_TRUE;
+   }
+   else if (pipeFormat == PIPE_FORMAT_Z32_UNORM &&
+            format == GL_DEPTH_COMPONENT &&
+            type == GL_UNSIGNED_INT) {
+      return GL_TRUE;
+   }
+   /* XXX add more cases */
+   else {
+      return GL_FALSE;
+   }
+}
+
+
+/**
+ * Check if any per-fragment ops are enabled.
+ * XXX probably move this to a re-usable place.
+ */
+static GLboolean
+any_fragment_ops(const struct st_context *st)
+{
+   if (st->state.depth_stencil.alpha.enabled ||
+       st->state.depth_stencil.depth.enabled ||
+       st->state.blend.blend_enable ||
+       st->state.blend.logicop_enable)
+      /* XXX more checks */
+      return GL_TRUE;
+   else
+      return GL_FALSE;
+}
+
+
+/**
+ * Check if any pixel transfer ops are enabled.
+ * XXX probably move this to a re-usable place.
+ */
+static GLboolean
+any_pixel_transfer_ops(const struct st_context *st)
+{
+   if (st->ctx->Pixel.RedScale != 1.0 ||
+       st->ctx->Pixel.RedBias != 0.0 ||
+       st->ctx->Pixel.GreenScale != 1.0 ||
+       st->ctx->Pixel.GreenBias != 0.0 ||
+       st->ctx->Pixel.BlueScale != 1.0 ||
+       st->ctx->Pixel.BlueBias != 0.0 ||
+       st->ctx->Pixel.AlphaScale != 1.0 ||
+       st->ctx->Pixel.AlphaBias != 0.0 ||
+       st->ctx->Pixel.MapColorFlag)
+      /* XXX more checks */
+      return GL_TRUE;
+   else
+      return GL_FALSE;
+}
+
+
+/**
+ * Draw image with a blit, or other non-textured quad method.
+ */
+static void
+draw_blit(struct st_context *st,
+          GLsizei width, GLsizei height,
+          GLenum format, GLenum type, const GLvoid *pixels)
+{
+
+
+}
+
+
+static void
+draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
+                    GLsizei width, GLsizei height, GLenum type,
+                    const struct gl_pixelstore_attrib *unpack,
+                    const GLvoid *pixels)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct st_renderbuffer *strb;
+   struct pipe_surface *ps;
+   const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0;
+   GLint skipPixels;
+   ubyte *stmap;
+
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   strb = st_renderbuffer(ctx->DrawBuffer->
+                          Attachment[BUFFER_STENCIL].Renderbuffer);
+   ps = screen->get_tex_surface(screen, strb->texture, 0, 0, 0,
+                                PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* map the stencil buffer */
+   stmap = screen->surface_map(screen, ps, 
+                               PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* if width > MAX_WIDTH, have to process image in chunks */
+   skipPixels = 0;
+   while (skipPixels < width) {
+      const GLint spanX = x + skipPixels;
+      const GLint spanWidth = MIN2(width - skipPixels, MAX_WIDTH);
+      GLint row;
+      for (row = 0; row < height; row++) {
+         GLint spanY = y + row;
+         GLubyte values[MAX_WIDTH];
+         GLenum destType = GL_UNSIGNED_BYTE;
+         const GLvoid *source = _mesa_image_address2d(unpack, pixels,
+                                                      width, height,
+                                                      GL_COLOR_INDEX, type,
+                                                      row, skipPixels);
+         _mesa_unpack_stencil_span(ctx, spanWidth, destType, values,
+                                   type, source, unpack,
+                                   ctx->_ImageTransferState);
+         if (zoom) {
+            /*
+            _swrast_write_zoomed_stencil_span(ctx, x, y, spanWidth,
+                                              spanX, spanY, values);
+            */
+         }
+         else {
+            if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+               spanY = ctx->DrawBuffer->Height - spanY - 1;
+            }
+
+            switch (ps->format) {
+            case PIPE_FORMAT_S8_UNORM:
+               {
+                  ubyte *dest = stmap + spanY * ps->stride + spanX;
+                  memcpy(dest, values, spanWidth);
+               }
+               break;
+            case PIPE_FORMAT_S8Z24_UNORM:
+               {
+                  uint *dest = (uint *) (stmap + spanY * ps->stride + spanX*4);
+                  GLint k;
+                  for (k = 0; k < spanWidth; k++) {
+                     uint p = dest[k];
+                     p = (p & 0xffffff) | (values[k] << 24);
+                     dest[k] = p;
+                  }
+               }
+               break;
+            default:
+               assert(0);
+            }
+         }
+      }
+      skipPixels += spanWidth;
+   }
+
+   /* unmap the stencil buffer */
+   screen->surface_unmap(screen, ps);
+   pipe_surface_reference(&ps, NULL);
+}
+
+
+/**
+ * Called via ctx->Driver.DrawPixels()
+ */
+static void
+st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
+              GLenum format, GLenum type,
+              const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels)
+{
+   struct st_fragment_program *stfp;
+   struct st_vertex_program *stvp;
+   struct st_context *st = ctx->st;
+   struct pipe_surface *ps;
+   GLuint bufferFormat;
+   const GLfloat *color;
+
+   if (format == GL_STENCIL_INDEX) {
+      draw_stencil_pixels(ctx, x, y, width, height, type, unpack, pixels);
+      return;
+   }
+
+   _mesa_set_vp_override( ctx, TRUE );
+   _mesa_update_state( ctx );
+
+   st_validate_state(st);
+
+   if (format == GL_DEPTH_COMPONENT) {
+      ps = st->state.framebuffer.zsbuf;
+      stfp = make_fragment_shader_z(ctx->st);
+      stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE);
+      color = ctx->Current.RasterColor;
+   }
+   else if (format == GL_STENCIL_INDEX) {
+      ps = st->state.framebuffer.zsbuf;
+      /* XXX special case - can't use texture map */
+      color = NULL;
+   }
+   else {
+      ps = st->state.framebuffer.cbufs[0];
+      stfp = combined_drawpix_fragment_program(ctx);
+      stvp = st_make_passthrough_vertex_shader(ctx->st, GL_FALSE);
+      color = NULL;
+   }
+
+   bufferFormat = ps->format;
+
+   if (1/*any_fragment_ops(st) ||
+       any_pixel_transfer_ops(st) ||
+       !compatible_formats(format, type, ps->format)*/) {
+      /* textured quad */
+      struct pipe_texture *pt
+         = make_texture(ctx->st, width, height, format, type, unpack, pixels);
+      if (pt) {
+         draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2],
+                            width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
+                            pt, stvp, stfp, color, GL_FALSE);
+         pipe_texture_reference(&pt, NULL);
+      }
+   }
+   else {
+      /* blit */
+      draw_blit(st, width, height, format, type, pixels);
+   }
+
+   _mesa_set_vp_override( ctx, FALSE );
+}
+
+
+
+static void
+copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
+                    GLsizei width, GLsizei height,
+                    GLint dstx, GLint dsty)
+{
+   struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer);
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   struct pipe_surface *psDraw;
+   ubyte *drawMap;
+   ubyte *buffer;
+   int i;
+
+   buffer = malloc(width * height * sizeof(ubyte));
+   if (!buffer) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels(stencil)");
+      return;
+   }
+
+   /* this will do stencil pixel transfer ops */
+   st_read_stencil_pixels(ctx, srcx, srcy, width, height, GL_UNSIGNED_BYTE,
+                          &ctx->DefaultPacking, buffer);
+
+   psDraw = screen->get_tex_surface(screen, rbDraw->texture, 0, 0, 0,
+                                    PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   assert(psDraw->block.width == 1);
+   assert(psDraw->block.height == 1);
+   
+   /* map the stencil buffer */
+   drawMap = screen->surface_map(screen, psDraw, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   /* draw */
+   /* XXX PixelZoom not handled yet */
+   for (i = 0; i < height; i++) {
+      ubyte *dst;
+      const ubyte *src;
+      int y;
+
+      y = dsty + i;
+
+      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+         y = ctx->DrawBuffer->Height - y - 1;
+      }
+
+      dst = drawMap + y * psDraw->stride + dstx * psDraw->block.size;
+      src = buffer + i * width;
+
+      switch (psDraw->format) {
+      case PIPE_FORMAT_S8Z24_UNORM:
+         {
+            uint *dst4 = (uint *) dst;
+            int j;
+            for (j = 0; j < width; j++) {
+               *dst4 = (*dst4 & 0xffffff) | (src[j] << 24);
+               dst4++;
+            }
+         }
+         break;
+      case PIPE_FORMAT_S8_UNORM:
+         memcpy(dst, src, width);
+         break;
+      default:
+         assert(0);
+      }
+   }
+
+   free(buffer);
+
+   /* unmap the stencil buffer */
+   screen->surface_unmap(screen, psDraw);
+   pipe_surface_reference(&psDraw, NULL);
+}
+
+
+static void
+st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
+              GLsizei width, GLsizei height,
+              GLint dstx, GLint dsty, GLenum type)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct st_renderbuffer *rbRead;
+   struct st_vertex_program *stvp;
+   struct st_fragment_program *stfp;
+   struct pipe_surface *psTex;
+   struct pipe_texture *pt;
+   GLfloat *color;
+   enum pipe_format srcFormat, texFormat;
+
+   /* make sure rendering has completed */
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   st_validate_state(st);
+
+   if (type == GL_STENCIL) {
+      /* can't use texturing to do stencil */
+      copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty);
+      return;
+   }
+
+   if (type == GL_COLOR) {
+      rbRead = st_get_color_read_renderbuffer(ctx);
+      color = NULL;
+      stfp = combined_drawpix_fragment_program(ctx);
+      stvp = st_make_passthrough_vertex_shader(ctx->st, GL_FALSE);
+   }
+   else {
+      assert(type == GL_DEPTH);
+      rbRead = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer);
+      color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
+      stfp = make_fragment_shader_z(ctx->st);
+      stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE);
+   }
+
+   srcFormat = rbRead->texture->format;
+
+   if (screen->is_format_supported(screen, srcFormat, PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      texFormat = srcFormat;
+   }
+   else {
+      /* srcFormat can't be used as a texture format */
+      if (type == GL_DEPTH) {
+         texFormat = st_choose_format(pipe, GL_DEPTH_COMPONENT, PIPE_TEXTURE_2D, 
+                                      PIPE_TEXTURE_USAGE_DEPTH_STENCIL);
+         assert(texFormat != PIPE_FORMAT_NONE); /* XXX no depth texture formats??? */
+      }
+      else {
+         /* default color format */
+         texFormat = st_choose_format(pipe, GL_RGBA, PIPE_TEXTURE_2D, 
+                                      PIPE_TEXTURE_USAGE_SAMPLER);
+         assert(texFormat != PIPE_FORMAT_NONE);
+      }
+   }
+
+   pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, texFormat, 0,
+                          width, height, 1, 0,
+                          PIPE_TEXTURE_USAGE_SAMPLER);
+   if (!pt)
+      return;
+
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+      srcy = ctx->DrawBuffer->Height - srcy - height;
+   }
+
+   if (srcFormat == texFormat) {
+      /* copy source framebuffer surface into mipmap/texture */
+      struct pipe_surface *psRead = screen->get_tex_surface(screen,
+                                       rbRead->texture, 0, 0, 0,
+                                       PIPE_BUFFER_USAGE_GPU_READ);
+      psTex = screen->get_tex_surface(screen, pt, 0, 0, 0, 
+                                      PIPE_BUFFER_USAGE_GPU_WRITE );
+      pipe->surface_copy(pipe,
+                         FALSE,
+			 psTex, /* dest */
+			 0, 0, /* destx/y */
+			 psRead,
+			 srcx, srcy, width, height);
+      pipe_surface_reference(&psRead, NULL);
+   }
+   else {
+      /* CPU-based fallback/conversion */
+      struct pipe_surface *psRead = screen->get_tex_surface(screen,
+                                       rbRead->texture, 0, 0, 0,
+                                       PIPE_BUFFER_USAGE_CPU_READ);
+
+      psTex = screen->get_tex_surface(screen, pt, 0, 0, 0, 
+                                      PIPE_BUFFER_USAGE_CPU_WRITE );
+
+      if (type == GL_COLOR) {
+         /* alternate path using get/put_tile() */
+         GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
+
+         pipe_get_tile_rgba(psRead, srcx, srcy, width, height, buf);
+         pipe_put_tile_rgba(psTex, 0, 0, width, height, buf);
+
+         free(buf);
+      }
+      else {
+         /* GL_DEPTH */
+         GLuint *buf = (GLuint *) malloc(width * height * sizeof(GLuint));
+         pipe_get_tile_z(psRead, srcx, srcy, width, height, buf);
+         pipe_put_tile_z(psTex, 0, 0, width, height, buf);
+         free(buf);
+      }
+      pipe_surface_reference(&psRead, NULL);
+   }
+
+   pipe_surface_reference(&psTex, NULL);
+
+   /* draw textured quad */
+   draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2],
+                      width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
+                      pt, stvp, stfp, color, GL_TRUE);
+
+   pipe_texture_reference(&pt, NULL);
+}
+
+
+
+void st_init_drawpixels_functions(struct dd_function_table *functions)
+{
+   functions->DrawPixels = st_DrawPixels;
+   functions->CopyPixels = st_CopyPixels;
+}
+
+
+void
+st_destroy_drawpix(struct st_context *st)
+{
+   st_reference_fragprog(st, &st->drawpix.z_shader, NULL);
+   st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
+   st_reference_vertprog(st, &st->drawpix.vert_shaders[0], NULL);
+   st_reference_vertprog(st, &st->drawpix.vert_shaders[1], NULL);
+}
+
+
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h
new file mode 100644
index 0000000000..26fe864d18
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_drawpixels.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_DRAWPIXELS_H
+#define ST_CB_DRAWPIXELS_H
+
+
+extern void st_init_drawpixels_functions(struct dd_function_table *functions);
+
+extern void
+st_destroy_drawpix(struct st_context *st);
+
+
+#endif /* ST_CB_DRAWPIXELS_H */
diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c
new file mode 100644
index 0000000000..9af648b8d0
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_fbo.c
@@ -0,0 +1,458 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Framebuffer/renderbuffer functions.
+ *
+ * \author Brian Paul
+ */
+
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/fbobject.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_screen.h"
+#include "st_context.h"
+#include "st_cb_fbo.h"
+#include "st_cb_texture.h"
+#include "st_format.h"
+#include "st_public.h"
+#include "st_texture.h"
+
+
+
+/**
+ * Compute the renderbuffer's Red/Green/EtcBit fields from the pipe format.
+ */
+static int
+init_renderbuffer_bits(struct st_renderbuffer *strb,
+                       enum pipe_format pipeFormat)
+{
+   struct pipe_format_info info;
+
+   if (!st_get_format_info( pipeFormat, &info )) {
+      assert( 0 );
+   }
+
+   strb->Base._ActualFormat = info.base_format;
+   strb->Base.RedBits = info.red_bits;
+   strb->Base.GreenBits = info.green_bits;
+   strb->Base.BlueBits = info.blue_bits;
+   strb->Base.AlphaBits = info.alpha_bits;
+   strb->Base.DepthBits = info.depth_bits;
+   strb->Base.StencilBits = info.stencil_bits;
+   strb->Base.DataType = st_format_datatype(pipeFormat);
+
+   return info.size;
+}
+
+/**
+ * gl_renderbuffer::AllocStorage()
+ * This is called to allocate the original drawing surface, and
+ * during window resize.
+ */
+static GLboolean
+st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+                              GLenum internalFormat,
+                              GLuint width, GLuint height)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+   struct pipe_texture template;
+   unsigned surface_usage;
+
+   /* Free the old surface and texture
+    */
+   pipe_surface_reference( &strb->surface, NULL );
+   pipe_texture_reference( &strb->texture, NULL );
+
+
+   memset(&template, 0, sizeof(template));
+
+   if (strb->format != PIPE_FORMAT_NONE) {
+      template.format = strb->format;
+   }
+   else {
+      template.format = st_choose_renderbuffer_format(pipe, internalFormat);
+   }
+
+   strb->Base.Width  = width;
+   strb->Base.Height = height;
+   init_renderbuffer_bits(strb, template.format);
+
+   template.target = PIPE_TEXTURE_2D;
+   template.compressed = 0;
+   pf_get_block(template.format, &template.block);
+   template.width[0] = width;
+   template.height[0] = height;
+   template.depth[0] = 1;
+   template.last_level = 0;
+   template.nr_samples = rb->NumSamples;
+
+   if (pf_is_depth_stencil(template.format)) {
+      template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+   }
+   else {
+      template.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                            PIPE_TEXTURE_USAGE_RENDER_TARGET);
+   }
+
+
+   /* Probably need dedicated flags for surface usage too: 
+    */
+   surface_usage = (PIPE_BUFFER_USAGE_GPU_READ |
+                    PIPE_BUFFER_USAGE_GPU_WRITE);
+#if 0
+                    PIPE_BUFFER_USAGE_CPU_READ |
+                    PIPE_BUFFER_USAGE_CPU_WRITE);
+#endif
+
+   strb->texture = pipe->screen->texture_create( pipe->screen,
+                                                 &template );
+
+   /* Special path for accum buffers.  
+    *
+    * Try a different surface format.  Since accum buffers are s/w
+    * only for now, the surface pixel format doesn't really matter,
+    * only that the buffer is large enough.
+    */
+   if (!strb->texture && template.format == DEFAULT_ACCUM_PIPE_FORMAT) 
+   {
+      /* Actually, just setting this usage value should be sufficient
+       * to tell the driver to go ahead and allocate the buffer, even
+       * if HW doesn't support the format.
+       */
+      template.tex_usage = 0;
+      surface_usage = (PIPE_BUFFER_USAGE_CPU_READ |
+                       PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      strb->texture = pipe->screen->texture_create( pipe->screen,
+                                                    &template );
+
+   }
+
+   if (!strb->texture) 
+      return FALSE;
+
+   strb->surface = pipe->screen->get_tex_surface( pipe->screen,
+                                                  strb->texture,
+                                                  0, 0, 0,
+                                                  surface_usage );
+
+   assert(strb->surface->texture);
+   assert(strb->surface->format);
+   assert(strb->surface->block.size);
+   assert(strb->surface->block.width);
+   assert(strb->surface->block.height);
+   assert(strb->surface->width == width);
+   assert(strb->surface->height == height);
+   assert(strb->surface->stride);
+
+
+   return strb->surface != NULL;
+}
+
+
+/**
+ * gl_renderbuffer::Delete()
+ */
+static void
+st_renderbuffer_delete(struct gl_renderbuffer *rb)
+{
+   struct st_renderbuffer *strb = st_renderbuffer(rb);
+   ASSERT(strb);
+   pipe_surface_reference(&strb->surface, NULL);
+   pipe_texture_reference(&strb->texture, NULL);
+   free(strb);
+}
+
+
+/**
+ * gl_renderbuffer::GetPointer()
+ */
+static void *
+null_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb,
+                 GLint x, GLint y)
+{
+   /* By returning NULL we force all software rendering to go through
+    * the span routines.
+    */
+#if 0
+   assert(0);  /* Should never get called with softpipe */
+#endif
+   return NULL;
+}
+
+
+/**
+ * Called via ctx->Driver.NewFramebuffer()
+ */
+static struct gl_framebuffer *
+st_new_framebuffer(GLcontext *ctx, GLuint name)
+{
+   /* XXX not sure we need to subclass gl_framebuffer for pipe */
+   return _mesa_new_framebuffer(ctx, name);
+}
+
+
+/**
+ * Called via ctx->Driver.NewRenderbuffer()
+ */
+static struct gl_renderbuffer *
+st_new_renderbuffer(GLcontext *ctx, GLuint name)
+{
+   struct st_renderbuffer *strb = CALLOC_STRUCT(st_renderbuffer);
+   if (strb) {
+      _mesa_init_renderbuffer(&strb->Base, name);
+      strb->Base.Delete = st_renderbuffer_delete;
+      strb->Base.AllocStorage = st_renderbuffer_alloc_storage;
+      strb->Base.GetPointer = null_get_pointer;
+      strb->format = PIPE_FORMAT_NONE;
+      return &strb->Base;
+   }
+   return NULL;
+}
+
+
+/**
+ * Allocate a renderbuffer for a an on-screen window (not a user-created
+ * renderbuffer).  The window system code determines the format.
+ */
+struct gl_renderbuffer *
+st_new_renderbuffer_fb(enum pipe_format format, int samples)
+{
+   struct st_renderbuffer *strb;
+
+   strb = CALLOC_STRUCT(st_renderbuffer);
+   if (!strb) {
+      _mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer");
+      return NULL;
+   }
+
+   _mesa_init_renderbuffer(&strb->Base, 0);
+   strb->Base.ClassID = 0x4242; /* just a unique value */
+   strb->Base.NumSamples = samples;
+   strb->format = format;
+
+   switch (format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      strb->Base.InternalFormat = GL_RGBA;
+      strb->Base._BaseFormat = GL_RGBA;
+      break;
+   case PIPE_FORMAT_Z16_UNORM:
+      strb->Base.InternalFormat = GL_DEPTH_COMPONENT16;
+      strb->Base._BaseFormat = GL_DEPTH_COMPONENT;
+      break;
+   case PIPE_FORMAT_Z32_UNORM:
+      strb->Base.InternalFormat = GL_DEPTH_COMPONENT32;
+      strb->Base._BaseFormat = GL_DEPTH_COMPONENT;
+      break;
+   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+   case PIPE_FORMAT_Z24X8_UNORM:
+      strb->Base.InternalFormat = GL_DEPTH24_STENCIL8_EXT;
+      strb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT;
+      break;
+   case PIPE_FORMAT_S8_UNORM:
+      strb->Base.InternalFormat = GL_STENCIL_INDEX8_EXT;
+      strb->Base._BaseFormat = GL_STENCIL_INDEX;
+      break;
+   case DEFAULT_ACCUM_PIPE_FORMAT: /*PIPE_FORMAT_R16G16B16A16_SNORM*/
+      strb->Base.InternalFormat = GL_RGBA16;
+      strb->Base._BaseFormat = GL_RGBA;
+      break;
+   default:
+      _mesa_problem(NULL,
+		    "Unexpected format in st_new_renderbuffer_fb");
+      return NULL;
+   }
+
+   /* st-specific methods */
+   strb->Base.Delete = st_renderbuffer_delete;
+   strb->Base.AllocStorage = st_renderbuffer_alloc_storage;
+   strb->Base.GetPointer = null_get_pointer;
+
+   /* surface is allocated in st_renderbuffer_alloc_storage() */
+   strb->surface = NULL;
+
+   return &strb->Base;
+}
+
+
+
+
+/**
+ * Called via ctx->Driver.BindFramebufferEXT().
+ */
+static void
+st_bind_framebuffer(GLcontext *ctx, GLenum target,
+                    struct gl_framebuffer *fb, struct gl_framebuffer *fbread)
+{
+
+}
+
+/**
+ * Called by ctx->Driver.FramebufferRenderbuffer
+ */
+static void
+st_framebuffer_renderbuffer(GLcontext *ctx, 
+                            struct gl_framebuffer *fb,
+                            GLenum attachment,
+                            struct gl_renderbuffer *rb)
+{
+   /* XXX no need for derivation? */
+   _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb);
+}
+
+
+/**
+ * Called by ctx->Driver.RenderTexture
+ */
+static void
+st_render_texture(GLcontext *ctx,
+                  struct gl_framebuffer *fb,
+                  struct gl_renderbuffer_attachment *att)
+{
+   struct st_renderbuffer *strb;
+   struct gl_renderbuffer *rb;
+   struct pipe_texture *pt = st_get_texobj_texture(att->Texture);
+   struct st_texture_object *stObj;
+   const struct gl_texture_image *texImage =
+      att->Texture->Image[att->CubeMapFace][att->TextureLevel];
+
+   if (!pt) 
+      return;
+
+   assert(!att->Renderbuffer);
+
+   /* create new renderbuffer which wraps the texture image */
+   rb = st_new_renderbuffer(ctx, 0);
+   if (!rb) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture()");
+      return;
+   }
+
+   _mesa_reference_renderbuffer(&att->Renderbuffer, rb);
+   assert(rb->RefCount == 1);
+   rb->AllocStorage = NULL; /* should not get called */
+   strb = st_renderbuffer(rb);
+
+   /* get the texture for the texture object */
+   stObj = st_texture_object(att->Texture);
+
+   /* point renderbuffer at texobject */
+   strb->rtt = stObj;
+   strb->rtt_level = att->TextureLevel;
+   strb->rtt_face = att->CubeMapFace;
+   strb->rtt_slice = att->Zoffset;
+
+   rb->Width = texImage->Width2;
+   rb->Height = texImage->Height2;
+   /*printf("***** render to texture level %d: %d x %d\n", att->TextureLevel, rb->Width, rb->Height);*/
+
+   /*printf("***** pipe texture %d x %d\n", pt->width[0], pt->height[0]);*/
+
+   pipe_texture_reference( &strb->texture, pt );
+
+   pipe_surface_reference(&strb->surface, NULL);
+
+   /* the new surface will be created during framebuffer validation */
+
+   init_renderbuffer_bits(strb, pt->format);
+
+   /*
+   printf("RENDER TO TEXTURE obj=%p pt=%p surf=%p  %d x %d\n",
+          att->Texture, pt, strb->surface, rb->Width, rb->Height);
+   */
+
+   /* Invalidate buffer state so that the pipe's framebuffer state
+    * gets updated.
+    * That's where the new renderbuffer (which we just created) gets
+    * passed to the pipe as a (color/depth) render target.
+    */
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
+/**
+ * Called via ctx->Driver.FinishRenderTexture.
+ */
+static void
+st_finish_render_texture(GLcontext *ctx,
+                         struct gl_renderbuffer_attachment *att)
+{
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   struct st_renderbuffer *strb = st_renderbuffer(att->Renderbuffer);
+
+   if (!strb)
+      return;
+
+   st_flush( ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL );
+
+   if (strb->surface)
+      screen->tex_surface_release( screen, &strb->surface );
+
+   strb->rtt = NULL;
+
+   /*
+   printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface);
+   */
+
+   _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
+
+   /* restore previous framebuffer state */
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
+
+void st_init_fbo_functions(struct dd_function_table *functions)
+{
+   functions->NewFramebuffer = st_new_framebuffer;
+   functions->NewRenderbuffer = st_new_renderbuffer;
+   functions->BindFramebuffer = st_bind_framebuffer;
+   functions->FramebufferRenderbuffer = st_framebuffer_renderbuffer;
+   functions->RenderTexture = st_render_texture;
+   functions->FinishRenderTexture = st_finish_render_texture;
+   /* no longer needed by core Mesa, drivers handle resizes...
+   functions->ResizeBuffers = st_resize_buffers;
+   */
+}
diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h
new file mode 100644
index 0000000000..44fa9fe9a4
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_fbo.h
@@ -0,0 +1,71 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_FBO_H
+#define ST_CB_FBO_H
+
+
+#define DEFAULT_ACCUM_PIPE_FORMAT PIPE_FORMAT_R16G16B16A16_SNORM
+
+
+
+/**
+ * Derived renderbuffer class.  Just need to add a pointer to the
+ * pipe surface.
+ */
+struct st_renderbuffer
+{
+   struct gl_renderbuffer Base;
+   struct pipe_texture *texture;
+   struct pipe_surface *surface; /* temporary view into texture */
+   enum pipe_format format;  /** preferred format, or PIPE_FORMAT_NONE */
+
+   struct st_texture_object *rtt;  /**< GL render to texture's texture */
+   int rtt_level, rtt_face, rtt_slice;
+
+   /** Render to texture state */
+   struct pipe_texture *texture_save;
+   struct pipe_surface *surface_save;
+};
+
+
+static INLINE struct st_renderbuffer *
+st_renderbuffer(struct gl_renderbuffer *rb)
+{
+   return (struct st_renderbuffer *) rb;
+}
+
+
+extern struct gl_renderbuffer *
+st_new_renderbuffer_fb(enum pipe_format format, int samples);
+
+extern void
+st_init_fbo_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_FBO_H */
diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c
new file mode 100644
index 0000000000..c7e8aa7cc5
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_feedback.c
@@ -0,0 +1,309 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * GL_SELECT and GL_FEEDBACK render modes.
+ * Basically, we use a private instance of the 'draw' module for doing
+ * selection/feedback.  It would be nice to use the transform_feedback
+ * hardware feature, but it's defined as happening pre-clip and we want
+ * post-clipped primitives.  Also, there's concerns about the efficiency
+ * of using the hardware for this anyway.
+ *
+ * Authors:
+ *   Brian Paul
+ */
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/feedback.h"
+#include "main/macros.h"
+
+#include "vbo/vbo.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_draw.h"
+#include "st_cb_feedback.h"
+#include "st_cb_bufferobjects.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "cso_cache/cso_cache.h"
+
+#include "draw/draw_context.h"
+#include "draw/draw_pipe.h"
+
+
+/**
+ * This is actually used for both feedback and selection.
+ */
+struct feedback_stage
+{
+   struct draw_stage stage;   /**< Base class */
+   GLcontext *ctx;            /**< Rendering context */
+   GLboolean reset_stipple_counter;
+};
+
+
+/**********************************************************************
+ * GL Feedback functions
+ **********************************************************************/
+
+static INLINE struct feedback_stage *
+feedback_stage( struct draw_stage *stage )
+{
+   return (struct feedback_stage *)stage;
+}
+
+
+static void
+feedback_vertex(GLcontext *ctx, const struct draw_context *draw,
+                const struct vertex_header *v)
+{
+   const struct st_context *st = ctx->st;
+   GLfloat win[4];
+   const GLfloat *color, *texcoord;
+   const GLfloat ci = 0;
+   GLuint slot;
+
+   /* Recall that Y=0=Top of window for Gallium wincoords */
+   win[0] = v->data[0][0];
+   win[1] = ctx->DrawBuffer->Height - v->data[0][1];
+   win[2] = v->data[0][2];
+   win[3] = 1.0F / v->data[0][3];
+
+   /* XXX
+    * When we compute vertex layout, save info about position of the
+    * color and texcoord attribs to use here.
+    */
+
+   slot = st->vertex_result_to_slot[VERT_RESULT_COL0];
+   if (slot != ~0U)
+      color = v->data[slot];
+   else
+      color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
+
+   slot = st->vertex_result_to_slot[VERT_RESULT_TEX0];
+   if (slot != ~0U)
+      texcoord = v->data[slot];
+   else
+      texcoord = ctx->Current.Attrib[VERT_ATTRIB_TEX0];
+
+   _mesa_feedback_vertex(ctx, win, color, ci, texcoord);
+}
+
+
+static void
+feedback_tri( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   struct draw_context *draw = stage->draw;
+   FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_POLYGON_TOKEN);
+   FEEDBACK_TOKEN(fs->ctx, (GLfloat) 3); /* three vertices */
+   feedback_vertex(fs->ctx, draw, prim->v[0]);
+   feedback_vertex(fs->ctx, draw, prim->v[1]);
+   feedback_vertex(fs->ctx, draw, prim->v[2]);
+}
+
+
+static void
+feedback_line( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   struct draw_context *draw = stage->draw;
+   if (fs->reset_stipple_counter) {
+      FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_LINE_RESET_TOKEN);
+      fs->reset_stipple_counter = GL_FALSE;
+   }
+   else {
+      FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_LINE_TOKEN);
+   }
+   feedback_vertex(fs->ctx, draw, prim->v[0]);
+   feedback_vertex(fs->ctx, draw, prim->v[1]);
+}
+
+
+static void
+feedback_point( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   struct draw_context *draw = stage->draw;
+   FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_POINT_TOKEN);
+   feedback_vertex(fs->ctx, draw, prim->v[0]);
+}
+
+
+static void
+feedback_flush( struct draw_stage *stage, unsigned flags )
+{
+   /* no-op */
+}
+
+
+static void
+feedback_reset_stipple_counter( struct draw_stage *stage )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   fs->reset_stipple_counter = GL_TRUE;
+}
+
+
+static void
+feedback_destroy( struct draw_stage *stage )
+{
+   /* no-op */
+}
+
+/**
+ * Create GL feedback drawing stage.
+ */
+static struct draw_stage *
+draw_glfeedback_stage(GLcontext *ctx, struct draw_context *draw)
+{
+   struct feedback_stage *fs = CALLOC_STRUCT(feedback_stage);
+
+   fs->stage.draw = draw;
+   fs->stage.next = NULL;
+   fs->stage.point = feedback_point;
+   fs->stage.line = feedback_line;
+   fs->stage.tri = feedback_tri;
+   fs->stage.flush = feedback_flush;
+   fs->stage.reset_stipple_counter = feedback_reset_stipple_counter;
+   fs->stage.destroy = feedback_destroy;
+   fs->ctx = ctx;
+
+   return &fs->stage;
+}
+
+
+
+/**********************************************************************
+ * GL Selection functions
+ **********************************************************************/
+
+static void
+select_tri( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   _mesa_update_hitflag( fs->ctx, prim->v[0]->data[0][2] );
+   _mesa_update_hitflag( fs->ctx, prim->v[1]->data[0][2] );
+   _mesa_update_hitflag( fs->ctx, prim->v[2]->data[0][2] );
+}
+
+static void
+select_line( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   _mesa_update_hitflag( fs->ctx, prim->v[0]->data[0][2] );
+   _mesa_update_hitflag( fs->ctx, prim->v[1]->data[0][2] );
+}
+
+
+static void
+select_point( struct draw_stage *stage, struct prim_header *prim )
+{
+   struct feedback_stage *fs = feedback_stage(stage);
+   _mesa_update_hitflag( fs->ctx, prim->v[0]->data[0][2] );
+}
+
+
+static void
+select_flush( struct draw_stage *stage, unsigned flags )
+{
+   /* no-op */
+}
+
+
+static void
+select_reset_stipple_counter( struct draw_stage *stage )
+{
+   /* no-op */
+}
+
+static void
+select_destroy( struct draw_stage *stage )
+{
+   /* no-op */
+}
+
+
+/**
+ * Create GL selection mode drawing stage.
+ */
+static struct draw_stage *
+draw_glselect_stage(GLcontext *ctx, struct draw_context *draw)
+{
+   struct feedback_stage *fs = CALLOC_STRUCT(feedback_stage);
+
+   fs->stage.draw = draw;
+   fs->stage.next = NULL;
+   fs->stage.point = select_point;
+   fs->stage.line = select_line;
+   fs->stage.tri = select_tri;
+   fs->stage.flush = select_flush;
+   fs->stage.reset_stipple_counter = select_reset_stipple_counter;
+   fs->stage.destroy = select_destroy;
+   fs->ctx = ctx;
+
+   return &fs->stage;
+}
+
+
+static void
+st_RenderMode(GLcontext *ctx, GLenum newMode )
+{
+   struct st_context *st = ctx->st;
+   struct draw_context *draw = st->draw;
+
+   if (newMode == GL_RENDER) {
+      /* restore normal VBO draw function */
+      vbo_set_draw_func(ctx, st_draw_vbo);
+   }
+   else if (newMode == GL_SELECT) {
+      if (!st->selection_stage)
+         st->selection_stage = draw_glselect_stage(ctx, draw);
+      draw_set_rasterize_stage(draw, st->selection_stage);
+      /* Plug in new vbo draw function */
+      vbo_set_draw_func(ctx, st_feedback_draw_vbo);
+   }
+   else {
+      if (!st->feedback_stage)
+         st->feedback_stage = draw_glfeedback_stage(ctx, draw);
+      draw_set_rasterize_stage(draw, st->feedback_stage);
+      /* Plug in new vbo draw function */
+      vbo_set_draw_func(ctx, st_feedback_draw_vbo);
+      /* need to generate/use a vertex program that emits pos/color/tex */
+      st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+   }
+}
+
+
+
+void st_init_feedback_functions(struct dd_function_table *functions)
+{
+   functions->RenderMode = st_RenderMode;
+}
diff --git a/src/mesa/state_tracker/st_cb_feedback.h b/src/mesa/state_tracker/st_cb_feedback.h
new file mode 100644
index 0000000000..2559ba3817
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_feedback.h
@@ -0,0 +1,37 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_FEEDBACK_H
+#define ST_CB_FEEDBACK_H
+
+
+extern void
+st_init_feedback_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_FEEDBACK_H */
diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c
new file mode 100644
index 0000000000..f8621ab125
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_flush.c
@@ -0,0 +1,149 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+#include "main/glheader.h"
+#include "main/macros.h"
+#include "main/context.h"
+#include "st_context.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_flush.h"
+#include "st_cb_clear.h"
+#include "st_cb_fbo.h"
+#include "st_public.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+#include "util/u_gen_mipmap.h"
+#include "util/u_blit.h"
+
+
+static INLINE GLboolean
+is_front_buffer_dirty(struct st_context *st)
+{
+   return st->frontbuffer_status == FRONT_STATUS_DIRTY;
+}
+
+
+/**
+ * Tell the screen to display the front color buffer on-screen.
+ */
+static void
+display_front_buffer(struct st_context *st)
+{
+   GLframebuffer *fb = st->ctx->DrawBuffer;
+   struct st_renderbuffer *strb
+      = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+   struct pipe_surface *front_surf = strb->surface;
+
+   /* Hook for copying "fake" frontbuffer if necessary:
+    */
+   st->pipe->screen->flush_frontbuffer( st->pipe->screen, front_surf,
+                                        st->pipe->priv );
+
+   /*
+   st->frontbuffer_status = FRONT_STATUS_UNDEFINED;
+   */
+}
+
+
+void st_flush( struct st_context *st, uint pipeFlushFlags,
+               struct pipe_fence_handle **fence )
+{
+   FLUSH_VERTICES(st->ctx, 0);
+
+   /* Release any vertex buffers that might potentially be accessed in
+    * successive frames:
+    */
+   st_flush_bitmap(st);
+   st_flush_clear(st);
+   util_blit_flush(st->blit);
+   util_gen_mipmap_flush(st->gen_mipmap);
+
+   st->pipe->flush( st->pipe, pipeFlushFlags, fence );
+}
+
+
+/**
+ * Flush, and wait for completion.
+ */
+void st_finish( struct st_context *st )
+{
+   struct pipe_fence_handle *fence = NULL;
+
+   st_flush(st, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, &fence);
+
+   if(fence) {
+      st->pipe->screen->fence_finish(st->pipe->screen, fence, 0);
+      st->pipe->screen->fence_reference(st->pipe->screen, &fence, NULL);
+   }
+}
+
+
+
+/**
+ * Called via ctx->Driver.Flush()
+ */
+static void st_glFlush(GLcontext *ctx)
+{
+   struct st_context *st = ctx->st;
+
+   if (is_front_buffer_dirty(st)) {
+      st_finish(st);
+      display_front_buffer(st);
+   }
+   else {
+      st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL);
+   }
+}
+
+
+/**
+ * Called via ctx->Driver.Finish()
+ */
+static void st_glFinish(GLcontext *ctx)
+{
+   struct st_context *st = ctx->st;
+
+   st_finish(st);
+
+   if (is_front_buffer_dirty(st)) {
+      display_front_buffer(st);
+   }
+}
+
+
+void st_init_flush_functions(struct dd_function_table *functions)
+{
+   functions->Flush = st_glFlush;
+   functions->Finish = st_glFinish;
+}
diff --git a/src/mesa/state_tracker/st_cb_flush.h b/src/mesa/state_tracker/st_cb_flush.h
new file mode 100644
index 0000000000..c26f779225
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_flush.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_FLUSH_H
+#define ST_CB_FLUSH_H
+
+
+extern void
+st_init_flush_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_FLUSH_H */
+
diff --git a/src/mesa/state_tracker/st_cb_get.c b/src/mesa/state_tracker/st_cb_get.c
new file mode 100644
index 0000000000..e7d7f03bc9
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_get.c
@@ -0,0 +1,97 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * glGet functions
+ *
+ * \author Brian Paul
+ */
+
+#include "main/imports.h"
+#include "main/context.h"
+
+#include "pipe/p_defines.h"
+
+#include "st_cb_fbo.h"
+#include "st_cb_get.h"
+
+
+
+/**
+ * Examine the current color read buffer format to determine
+ * which GL pixel format/type combo is the best match.
+ */
+static void
+get_preferred_read_format_type(GLcontext *ctx, GLint *format, GLint *type)
+{
+   struct gl_framebuffer *fb = ctx->ReadBuffer;
+   struct st_renderbuffer *strb = st_renderbuffer(fb->_ColorReadBuffer);
+
+   /* defaults */
+   *format = ctx->Const.ColorReadFormat;
+   *type = ctx->Const.ColorReadType;
+
+   if (strb) {
+      /* XXX could add more cases here... */
+      if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM) {
+         *format = GL_BGRA;
+         if (_mesa_little_endian())
+            *type = GL_UNSIGNED_INT_8_8_8_8_REV;
+         else
+            *type = GL_UNSIGNED_INT_8_8_8_8;
+      }
+   }
+}
+
+
+/**
+ * We only intercept the OES preferred ReadPixels format/type.
+ * Everything else goes to the default _mesa_GetIntegerv.
+ */
+static GLboolean 
+st_GetIntegerv(GLcontext *ctx, GLenum pname, GLint *params)
+{
+   GLint dummy;
+
+   switch (pname) {
+   case GL_IMPLEMENTATION_COLOR_READ_TYPE_OES:
+      get_preferred_read_format_type(ctx, &dummy, params);
+      return GL_TRUE;
+   case GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES:
+      get_preferred_read_format_type(ctx, params, &dummy);
+      return GL_TRUE;
+   default:
+      return GL_FALSE;
+   }
+}
+
+
+void st_init_get_functions(struct dd_function_table *functions)
+{
+   functions->GetIntegerv = st_GetIntegerv;
+}
diff --git a/src/mesa/state_tracker/st_cb_get.h b/src/mesa/state_tracker/st_cb_get.h
new file mode 100644
index 0000000000..8e9f3e9306
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_get.h
@@ -0,0 +1,37 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_GET_H
+#define ST_CB_GET_H
+
+
+extern void
+st_init_get_functions(struct dd_function_table *functions);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c
new file mode 100644
index 0000000000..ea0fa20012
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_program.c
@@ -0,0 +1,266 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "main/glheader.h"
+#include "main/macros.h"
+#include "main/enums.h"
+#include "shader/prog_instruction.h"
+#include "shader/prog_parameter.h"
+#include "shader/program.h"
+#include "shader/programopt.h"
+#include "shader/shader_api.h"
+
+#include "cso_cache/cso_context.h"
+#include "draw/draw_context.h"
+
+#include "st_context.h"
+#include "st_program.h"
+#include "st_atom_shader.h"
+#include "st_cb_program.h"
+
+
+static GLuint SerialNo = 1;
+
+
+/**
+ * Called via ctx->Driver.BindProgram() to bind an ARB vertex or
+ * fragment program.
+ */
+static void st_bind_program( GLcontext *ctx,
+			     GLenum target, 
+			     struct gl_program *prog )
+{
+   struct st_context *st = st_context(ctx);
+
+   switch (target) {
+   case GL_VERTEX_PROGRAM_ARB: 
+      st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+      break;
+   case GL_FRAGMENT_PROGRAM_ARB:
+      st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+      break;
+   }
+}
+
+
+/**
+ * Called via ctx->Driver.UseProgram() to bind a linked GLSL program
+ * (vertex shader + fragment shader).
+ */
+static void st_use_program( GLcontext *ctx,
+			    GLuint program )
+{
+   struct st_context *st = st_context(ctx);
+
+   st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+   st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+
+   _mesa_use_program(ctx, program);
+}
+
+
+
+/**
+ * Called via ctx->Driver.NewProgram() to allocate a new vertex or
+ * fragment program.
+ */
+static struct gl_program *st_new_program( GLcontext *ctx,
+					  GLenum target,
+					  GLuint id )
+{
+   switch (target) {
+   case GL_VERTEX_PROGRAM_ARB: {
+      struct st_vertex_program *prog = CALLOC_STRUCT(st_vertex_program);
+
+      prog->serialNo = SerialNo++;
+
+      return _mesa_init_vertex_program( ctx, 
+					&prog->Base,
+					target, 
+					id );
+   }
+
+   case GL_FRAGMENT_PROGRAM_ARB:
+   case GL_FRAGMENT_PROGRAM_NV: {
+      struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program);
+
+      prog->serialNo = SerialNo++;
+
+      return _mesa_init_fragment_program( ctx, 
+					  &prog->Base,
+					  target, 
+					  id );
+   }
+
+   default:
+      assert(0);
+      return NULL;
+   }
+}
+
+
+void
+st_delete_program(GLcontext *ctx, struct gl_program *prog)
+{
+   struct st_context *st = st_context(ctx);
+
+   switch( prog->Target ) {
+   case GL_VERTEX_PROGRAM_ARB:
+      {
+         struct st_vertex_program *stvp = (struct st_vertex_program *) prog;
+
+         if (stvp->driver_shader) {
+            cso_delete_vertex_shader(st->cso_context, stvp->driver_shader);
+            stvp->driver_shader = NULL;
+         }
+
+         if (stvp->draw_shader) {
+#if FEATURE_feedback || FEATURE_drawpix
+            /* this would only have been allocated for the RasterPos path */
+            draw_delete_vertex_shader(st->draw, stvp->draw_shader);
+            stvp->draw_shader = NULL;
+#endif
+         }
+
+         if (stvp->state.tokens) {
+            FREE((void *) stvp->state.tokens);
+            stvp->state.tokens = NULL;
+         }
+      }
+      break;
+   case GL_FRAGMENT_PROGRAM_ARB:
+      {
+         struct st_fragment_program *stfp = (struct st_fragment_program *) prog;
+
+         if (stfp->driver_shader) {
+            cso_delete_fragment_shader(st->cso_context, stfp->driver_shader);
+            stfp->driver_shader = NULL;
+         }
+         
+         if (stfp->state.tokens) {
+            FREE((void *) stfp->state.tokens);
+            stfp->state.tokens = NULL;
+         }
+
+         if (stfp->bitmap_program) {
+            struct gl_program *prg = &stfp->bitmap_program->Base.Base;
+            _mesa_reference_program(ctx, &prg, NULL);
+            stfp->bitmap_program = NULL;
+         }
+
+         st_free_translated_vertex_programs(st, stfp->vertex_programs);
+      }
+      break;
+   default:
+      assert(0); /* problem */
+   }
+
+   /* delete base class */
+   _mesa_delete_program( ctx, prog );
+}
+
+
+static GLboolean st_is_program_native( GLcontext *ctx,
+				       GLenum target, 
+				       struct gl_program *prog )
+{
+   return GL_TRUE;
+}
+
+
+static void st_program_string_notify( GLcontext *ctx,
+				      GLenum target,
+				      struct gl_program *prog )
+{
+   struct st_context *st = st_context(ctx);
+
+   if (target == GL_FRAGMENT_PROGRAM_ARB) {
+      struct st_fragment_program *stfp = (struct st_fragment_program *) prog;
+
+      stfp->serialNo++;
+
+      if (stfp->driver_shader) {
+         cso_delete_fragment_shader(st->cso_context, stfp->driver_shader);
+         stfp->driver_shader = NULL;
+      }
+
+      if (stfp->state.tokens) {
+         FREE((void *) stfp->state.tokens);
+         stfp->state.tokens = NULL;
+      }
+
+      stfp->param_state = stfp->Base.Base.Parameters->StateFlags;
+
+      if (st->fp == stfp)
+	 st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
+   }
+   else if (target == GL_VERTEX_PROGRAM_ARB) {
+      struct st_vertex_program *stvp = (struct st_vertex_program *) prog;
+
+      stvp->serialNo++;
+
+      if (stvp->driver_shader) {
+         cso_delete_vertex_shader(st->cso_context, stvp->driver_shader);
+         stvp->driver_shader = NULL;
+      }
+
+      if (stvp->draw_shader) {
+#if FEATURE_feedback || FEATURE_drawpix
+         /* this would only have been allocated for the RasterPos path */
+         draw_delete_vertex_shader(st->draw, stvp->draw_shader);
+         stvp->draw_shader = NULL;
+#endif
+      }
+
+      if (stvp->state.tokens) {
+         FREE((void *) stvp->state.tokens);
+         stvp->state.tokens = NULL;
+      }
+
+      stvp->param_state = stvp->Base.Base.Parameters->StateFlags;
+
+      if (st->vp == stvp)
+	 st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+   }
+}
+
+
+
+void st_init_program_functions(struct dd_function_table *functions)
+{
+   functions->BindProgram = st_bind_program;
+   functions->UseProgram = st_use_program;
+   functions->NewProgram = st_new_program;
+   functions->DeleteProgram = st_delete_program;
+   functions->IsProgramNative = st_is_program_native;
+   functions->ProgramStringNotify = st_program_string_notify;
+}
diff --git a/src/mesa/state_tracker/st_cb_program.h b/src/mesa/state_tracker/st_cb_program.h
new file mode 100644
index 0000000000..0de96f2fd2
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_program.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef ST_CB_PROGRAM_H
+#define ST_CB_PROGRAM_H
+
+
+extern void
+st_init_program_functions(struct dd_function_table *functions);
+
+extern void
+st_delete_program(GLcontext *ctx, struct gl_program *prog);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_queryobj.c b/src/mesa/state_tracker/st_cb_queryobj.c
new file mode 100644
index 0000000000..21c2c7dd9f
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_queryobj.c
@@ -0,0 +1,171 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * glBegin/EndQuery interface to pipe
+ *
+ * \author Brian Paul
+ */
+
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/image.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "st_context.h"
+#include "st_cb_queryobj.h"
+#include "st_public.h"
+
+
+struct st_query_object
+{
+   struct gl_query_object base;
+   struct pipe_query *pq;
+};
+
+
+/**
+ * Cast wrapper
+ */
+static struct st_query_object *
+st_query_object(struct gl_query_object *q)
+{
+   return (struct st_query_object *) q;
+}
+
+
+static struct gl_query_object *
+st_NewQueryObject(GLcontext *ctx, GLuint id)
+{
+   struct st_query_object *stq = CALLOC_STRUCT(st_query_object);
+   if (stq) {
+      stq->base.Id = id;
+      stq->base.Ready = GL_TRUE;
+      stq->pq = NULL;
+      return &stq->base;
+   }
+   return NULL;
+}
+
+
+
+static void
+st_DeleteQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   if (stq->pq) {
+      pipe->destroy_query(pipe, stq->pq);
+      stq->pq = NULL;
+   }
+
+   FREE(stq);
+}
+
+
+static void
+st_BeginQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   switch (q->Target) {
+   case GL_SAMPLES_PASSED_ARB:
+      if (!stq->pq)
+	 stq->pq = pipe->create_query( pipe, PIPE_QUERY_OCCLUSION_COUNTER );
+      break;
+   default:
+      assert(0);
+      return;
+   }
+
+   pipe->begin_query(pipe, stq->pq);
+}
+
+
+static void
+st_EndQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   pipe->end_query(pipe, stq->pq);
+}
+
+
+static void
+st_WaitQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   /* this function should only be called if we don't have a ready result */
+   assert(!stq->base.Ready);
+
+   while (!stq->base.Ready &&
+	  !pipe->get_query_result(pipe, 
+				  stq->pq,
+				  TRUE,
+				  &q->Result))
+   {
+      /* nothing */
+   }
+			    
+   q->Ready = GL_TRUE;
+}
+
+
+static void
+st_CheckQuery(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   if (!q->Ready) {
+      q->Ready = pipe->get_query_result(pipe, 
+					stq->pq,
+					FALSE,
+					&q->Result);
+   }
+}
+
+
+
+
+void st_init_query_functions(struct dd_function_table *functions)
+{
+   functions->NewQueryObject = st_NewQueryObject;
+   functions->DeleteQuery = st_DeleteQuery;
+   functions->BeginQuery = st_BeginQuery;
+   functions->EndQuery = st_EndQuery;
+   functions->WaitQuery = st_WaitQuery;
+   functions->CheckQuery = st_CheckQuery;
+}
diff --git a/src/mesa/state_tracker/st_cb_queryobj.h b/src/mesa/state_tracker/st_cb_queryobj.h
new file mode 100644
index 0000000000..9220a212b6
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_queryobj.h
@@ -0,0 +1,36 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef ST_CB_QUERYOBJ_H
+#define ST_CB_QUERYOBJ_H
+
+
+extern void
+st_init_query_functions(struct dd_function_table *functions);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c
new file mode 100644
index 0000000000..8867ca5652
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_rasterpos.c
@@ -0,0 +1,261 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * glRasterPos implementation.  Basically render a GL_POINT with our
+ * private draw module.  Plug in a special "rasterpos" stage at the end
+ * of the 'draw' pipeline to capture the results and update the current
+ * raster pos attributes.
+ *
+ * Authors:
+ *   Brian Paul
+ */
+
+
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/feedback.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_draw.h"
+#include "st_cb_rasterpos.h"
+#include "st_draw.h"
+#include "draw/draw_context.h"
+#include "draw/draw_pipe.h"
+#include "shader/prog_instruction.h"
+#include "vbo/vbo.h"
+
+
+
+/**
+ * Our special drawing pipeline stage (replaces rasterization).
+ */
+struct rastpos_stage
+{
+   struct draw_stage stage;   /**< Base class */
+   GLcontext *ctx;            /**< Rendering context */
+
+   /* vertex attrib info we can setup once and re-use */
+   struct gl_client_array array[VERT_ATTRIB_MAX];
+   const struct gl_client_array *arrays[VERT_ATTRIB_MAX];
+   struct _mesa_prim prim;
+};
+
+
+static INLINE struct rastpos_stage *
+rastpos_stage( struct draw_stage *stage )
+{
+   return (struct rastpos_stage *) stage;
+}
+
+static void
+rastpos_flush( struct draw_stage *stage, unsigned flags )
+{
+   /* no-op */
+}
+
+static void
+rastpos_reset_stipple_counter( struct draw_stage *stage )
+{
+   /* no-op */
+}
+
+static void
+rastpos_tri( struct draw_stage *stage, struct prim_header *prim )
+{
+   /* should never get here */
+   assert(0);
+}
+
+static void
+rastpos_line( struct draw_stage *stage, struct prim_header *prim )
+{
+   /* should never get here */
+   assert(0);
+}
+
+static void
+rastpos_destroy(struct draw_stage *stage)
+{
+   free(stage);
+}
+
+
+/**
+ * Update a raster pos attribute from the vertex result if it's present,
+ * else copy the current attrib.
+ */
+static void
+update_attrib(GLcontext *ctx, const GLuint *outputMapping,
+              const struct vertex_header *vert,
+              GLfloat *dest,
+              GLuint result, GLuint defaultAttrib)
+{
+   const GLfloat *src;
+   const GLuint k = outputMapping[result];
+   if (k != ~0U)
+      src = vert->data[k];
+   else
+      src = ctx->Current.Attrib[defaultAttrib];
+   COPY_4V(dest, src);
+}
+
+
+/**
+ * Normally, this function would render a GL_POINT.
+ */
+static void
+rastpos_point(struct draw_stage *stage, struct prim_header *prim)
+{
+   struct rastpos_stage *rs = rastpos_stage(stage);
+   GLcontext *ctx = rs->ctx;
+   struct st_context *st = ctx->st;
+   const GLfloat height = (GLfloat) ctx->DrawBuffer->Height;
+   const GLuint *outputMapping = st->vertex_result_to_slot;
+   const GLfloat *pos;
+   GLuint i;
+
+   /* if we get here, we didn't get clipped */
+   ctx->Current.RasterPosValid = GL_TRUE;
+
+   /* update raster pos */
+   pos = prim->v[0]->data[0];
+   ctx->Current.RasterPos[0] = pos[0];
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP)
+      ctx->Current.RasterPos[1] = height - pos[1]; /* invert Y */
+   else
+      ctx->Current.RasterPos[1] = pos[1];
+   ctx->Current.RasterPos[2] = pos[2];
+   ctx->Current.RasterPos[3] = pos[3];
+
+   /* update other raster attribs */
+   update_attrib(ctx, outputMapping, prim->v[0],
+                 ctx->Current.RasterColor,
+                 VERT_RESULT_COL0, VERT_ATTRIB_COLOR0);
+
+   update_attrib(ctx, outputMapping, prim->v[0],
+                 ctx->Current.RasterSecondaryColor,
+                 VERT_RESULT_COL1, VERT_ATTRIB_COLOR1);
+
+   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+      update_attrib(ctx, outputMapping, prim->v[0],
+                    ctx->Current.RasterTexCoords[i],
+                    VERT_RESULT_TEX0 + i, VERT_ATTRIB_TEX0 + i);
+   }
+
+   if (ctx->RenderMode == GL_SELECT) {
+      _mesa_update_hitflag( ctx, ctx->Current.RasterPos[2] );
+   }
+}
+
+
+/**
+ * Create rasterpos "drawing" stage.
+ */
+static struct rastpos_stage *
+new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw)
+{
+   struct rastpos_stage *rs = CALLOC_STRUCT(rastpos_stage);
+   GLuint i;
+
+   rs->stage.draw = draw;
+   rs->stage.next = NULL;
+   rs->stage.point = rastpos_point;
+   rs->stage.line = rastpos_line;
+   rs->stage.tri = rastpos_tri;
+   rs->stage.flush = rastpos_flush;
+   rs->stage.destroy = rastpos_destroy;
+   rs->stage.reset_stipple_counter = rastpos_reset_stipple_counter;
+   rs->stage.destroy = rastpos_destroy;
+   rs->ctx = ctx;
+
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      rs->array[i].Size = 4;
+      rs->array[i].Type = GL_FLOAT;
+      rs->array[i].Stride = 0;
+      rs->array[i].StrideB = 0;
+      rs->array[i].Ptr = (GLubyte *) ctx->Current.Attrib[i];
+      rs->array[i].Enabled = GL_TRUE;
+      rs->array[i].Normalized = GL_TRUE;
+      rs->array[i].BufferObj = NULL;
+      rs->arrays[i] = &rs->array[i];
+   }
+
+   rs->prim.mode = GL_POINTS;
+   rs->prim.indexed = 0;
+   rs->prim.begin = 1;
+   rs->prim.end = 1;
+   rs->prim.weak = 0;
+   rs->prim.start = 0;
+   rs->prim.count = 1;
+
+   return rs;
+}
+
+
+static void
+st_RasterPos(GLcontext *ctx, const GLfloat v[4])
+{
+   struct st_context *st = ctx->st;
+   struct draw_context *draw = st->draw;
+   struct rastpos_stage *rs;
+
+   if (st->rastpos_stage) {
+      /* get rastpos stage info */
+      rs = rastpos_stage(st->rastpos_stage);
+   }
+   else {
+      /* create rastpos draw stage */
+      rs = new_draw_rastpos_stage(ctx, draw);
+      st->rastpos_stage = &rs->stage;
+   }
+
+   /* plug our rastpos stage into the draw module */
+   draw_set_rasterize_stage(st->draw, st->rastpos_stage);
+
+   /* make sure everything's up to date */
+   st_validate_state(ctx->st);
+
+   /* This will get set only if rastpos_point(), above, gets called */
+   ctx->Current.RasterPosValid = GL_FALSE;
+
+   /* All vertex attribs but position were previously initialized above.
+    * Just plug in position pointer now.
+    */
+   rs->array[0].Ptr = (GLubyte *) v;
+
+   /* draw the point */
+   st_feedback_draw_vbo(ctx, rs->arrays, &rs->prim, 1, NULL, 0, 1);
+}
+
+
+
+void st_init_rasterpos_functions(struct dd_function_table *functions)
+{
+   functions->RasterPos = st_RasterPos;
+}
diff --git a/src/mesa/state_tracker/st_cb_rasterpos.h b/src/mesa/state_tracker/st_cb_rasterpos.h
new file mode 100644
index 0000000000..2b992e1405
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_rasterpos.h
@@ -0,0 +1,33 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef ST_CB_RASTERPOS_H
+#define ST_CB_RASTERPOS_H
+
+extern void st_init_rasterpos_functions(struct dd_function_table *functions);
+
+#endif
diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c
new file mode 100644
index 0000000000..646eaff190
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_readpixels.c
@@ -0,0 +1,464 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * glReadPixels interface to pipe
+ *
+ * \author Brian Paul
+ */
+
+
+#include "main/imports.h"
+#include "main/bufferobj.h"
+#include "main/context.h"
+#include "main/image.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+#include "st_context.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_readpixels.h"
+#include "st_cb_fbo.h"
+#include "st_format.h"
+#include "st_public.h"
+
+
+/**
+ * Special case for reading stencil buffer.
+ * For color/depth we use get_tile().  For stencil, map the stencil buffer.
+ */
+void
+st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
+                       GLsizei width, GLsizei height, GLenum type,
+                       const struct gl_pixelstore_attrib *packing,
+                       GLvoid *pixels)
+{
+   struct gl_framebuffer *fb = ctx->ReadBuffer;
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   struct st_renderbuffer *strb = st_renderbuffer(fb->_StencilBuffer);
+   struct pipe_surface *ps;
+   ubyte *stmap;
+   GLint j;
+
+   /* Create a CPU-READ surface/view into the renderbuffer's texture */
+   ps = screen->get_tex_surface(screen, strb->texture,  0, 0, 0,
+                                PIPE_BUFFER_USAGE_CPU_READ);
+
+   /* map the stencil buffer */
+   stmap = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ);
+
+   /* width should never be > MAX_WIDTH since we did clipping earlier */
+   ASSERT(width <= MAX_WIDTH);
+
+   /* process image row by row */
+   for (j = 0; j < height; j++, y++) {
+      GLvoid *dest;
+      GLstencil values[MAX_WIDTH];
+      GLint srcY;
+
+      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+         srcY = ctx->DrawBuffer->Height - y - 1;
+      }
+      else {
+         srcY = y;
+      }
+
+      /* get stencil values */
+      switch (ps->format) {
+      case PIPE_FORMAT_S8_UNORM:
+         {
+            const ubyte *src = stmap + srcY * ps->stride + x;
+            memcpy(values, src, width);
+         }
+         break;
+      case PIPE_FORMAT_S8Z24_UNORM:
+         {
+            const uint *src = (uint *) (stmap + srcY * ps->stride + x*4);
+            GLint k;
+            for (k = 0; k < width; k++) {
+               values[k] = src[k] >> 24;
+            }
+         }
+         break;
+      case PIPE_FORMAT_Z24S8_UNORM:
+         {
+            const uint *src = (uint *) (stmap + srcY * ps->stride + x*4);
+            GLint k;
+            for (k = 0; k < width; k++) {
+               values[k] = src[k] & 0xff;
+            }
+         }
+         break;
+      default:
+         assert(0);
+      }
+
+      /* store */
+      dest = _mesa_image_address2d(packing, pixels, width, height,
+                                   GL_STENCIL_INDEX, type, j, 0);
+
+      _mesa_pack_stencil_span(ctx, width, type, dest, values, packing);
+   }
+
+
+   /* unmap the stencil buffer */
+   screen->surface_unmap(screen, ps);
+   pipe_surface_reference(&ps, NULL);
+}
+
+
+/**
+ * Return renderbuffer to use for reading color pixels for glRead/CopyPixel
+ * commands.
+ * Special care is needed for the front buffer.
+ */
+struct st_renderbuffer *
+st_get_color_read_renderbuffer(GLcontext *ctx)
+{
+   struct gl_framebuffer *fb = ctx->ReadBuffer;
+   struct st_renderbuffer *strb =
+      st_renderbuffer(fb->_ColorReadBuffer);
+   struct st_renderbuffer *front = 
+      st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+
+   if (strb == front
+       && ctx->st->frontbuffer_status == FRONT_STATUS_COPY_OF_BACK) {
+      /* reading from front color buffer, which is a logical copy of the
+       * back color buffer.
+       */
+      struct st_renderbuffer *back = 
+         st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+      strb = back;
+   }
+
+   return strb;
+}
+
+
+/**
+ * Try to do glReadPixels in a fast manner for common cases.
+ * \return GL_TRUE for success, GL_FALSE for failure
+ */
+static GLboolean
+st_fast_readpixels(GLcontext *ctx, struct st_renderbuffer *strb,
+                   GLint x, GLint y, GLsizei width, GLsizei height,
+                   GLenum format, GLenum type,
+                   const struct gl_pixelstore_attrib *pack,
+                   GLvoid *dest)
+{
+   enum combination {
+      A8R8G8B8_UNORM_TO_RGBA_UBYTE,
+      A8R8G8B8_UNORM_TO_RGB_UBYTE,
+      A8R8G8B8_UNORM_TO_BGRA_UINT
+   } combo;
+
+   if (ctx->_ImageTransferState)
+      return GL_FALSE;
+
+   if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM &&
+       format == GL_RGBA && type == GL_UNSIGNED_BYTE) {
+      combo = A8R8G8B8_UNORM_TO_RGBA_UBYTE;
+   }
+   else if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM &&
+            format == GL_RGB && type == GL_UNSIGNED_BYTE) {
+      combo = A8R8G8B8_UNORM_TO_RGB_UBYTE;
+   }
+   else if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM &&
+            format == GL_BGRA && type == GL_UNSIGNED_INT_8_8_8_8_REV) {
+      combo = A8R8G8B8_UNORM_TO_BGRA_UINT;
+   }
+   else {
+      return GL_FALSE;
+   }
+
+   /*printf("st_fast_readpixels combo %d\n", (GLint) combo);*/
+
+   {
+      struct pipe_context *pipe = ctx->st->pipe;
+      struct pipe_screen *screen = pipe->screen;
+      struct pipe_surface *surf;
+      const GLubyte *map;
+      GLubyte *dst;
+      GLint row, col, dy, dstStride;
+
+      surf = screen->get_tex_surface(screen, strb->texture,  0, 0, 0,
+                                     PIPE_BUFFER_USAGE_CPU_READ);
+      if (!surf) {
+         return GL_FALSE;
+      }
+
+      map = screen->surface_map(screen, surf, PIPE_BUFFER_USAGE_CPU_READ);
+      if (!map) {
+         pipe_surface_reference(&surf, NULL);
+         return GL_FALSE;
+      }
+
+      if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
+         y = surf->height - y - 1;
+         dy = -1;
+      }
+      else {
+         dy = 1;
+      }
+
+      dst = _mesa_image_address2d(pack, dest, width, height,
+                                  format, type, 0, 0);
+      dstStride = _mesa_image_row_stride(pack, width, format, type);
+
+      switch (combo) {
+      case A8R8G8B8_UNORM_TO_RGBA_UBYTE:
+         for (row = 0; row < height; row++) {
+            const GLubyte *src = map + y * surf->stride + x * 4;
+            for (col = 0; col < width; col++) {
+               GLuint pixel = ((GLuint *) src)[col];
+               dst[col*4+0] = (pixel >> 16) & 0xff;
+               dst[col*4+1] = (pixel >>  8) & 0xff;
+               dst[col*4+2] = (pixel >>  0) & 0xff;
+               dst[col*4+3] = (pixel >> 24) & 0xff;
+            }
+            dst += dstStride;
+            y += dy;
+         }
+         break;
+      case A8R8G8B8_UNORM_TO_RGB_UBYTE:
+         for (row = 0; row < height; row++) {
+            const GLubyte *src = map + y * surf->stride + x * 4;
+            for (col = 0; col < width; col++) {
+               GLuint pixel = ((GLuint *) src)[col];
+               dst[col*3+0] = (pixel >> 16) & 0xff;
+               dst[col*3+1] = (pixel >>  8) & 0xff;
+               dst[col*3+2] = (pixel >>  0) & 0xff;
+            }
+            dst += dstStride;
+            y += dy;
+         }
+         break;
+      case A8R8G8B8_UNORM_TO_BGRA_UINT:
+         for (row = 0; row < height; row++) {
+            const GLubyte *src = map + y * surf->stride + x * 4;
+            memcpy(dst, src, 4 * width);
+            dst += dstStride;
+            y += dy;
+         }
+         break;
+      default:
+         ; /* nothing */
+      }
+
+      screen->surface_unmap(screen, surf);
+      pipe_surface_reference(&surf, NULL);
+   }
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Do glReadPixels by getting rows from the framebuffer surface with
+ * get_tile().  Convert to requested format/type with Mesa image routines.
+ * Image transfer ops are done in software too.
+ */
+static void
+st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
+              GLenum format, GLenum type,
+              const struct gl_pixelstore_attrib *pack,
+              GLvoid *dest)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   GLfloat temp[MAX_WIDTH][4];
+   const GLbitfield transferOps = ctx->_ImageTransferState;
+   GLsizei i, j;
+   GLint yStep, dfStride;
+   GLfloat *df;
+   struct st_renderbuffer *strb;
+   struct gl_pixelstore_attrib clippedPacking = *pack;
+   struct pipe_surface *surf;
+
+   assert(ctx->ReadBuffer->Width > 0);
+
+   /* XXX convolution not done yet */
+   assert((transferOps & IMAGE_CONVOLUTION_BIT) == 0);
+
+   /* Do all needed clipping here, so that we can forget about it later */
+   if (!_mesa_clip_readpixels(ctx, &x, &y, &width, &height, &clippedPacking)) {
+      /* The ReadPixels surface is totally outside the window bounds */
+      return;
+   }
+
+   dest = _mesa_map_readpix_pbo(ctx, &clippedPacking, dest);
+   if (!dest)
+      return;
+
+   /* make sure rendering has completed */
+   st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   if (format == GL_STENCIL_INDEX) {
+      st_read_stencil_pixels(ctx, x, y, width, height, type, pack, dest);
+      return;
+   }
+   else if (format == GL_DEPTH_COMPONENT) {
+      strb = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer);
+   }
+   else {
+      /* Read color buffer */
+      strb = st_get_color_read_renderbuffer(ctx);
+   }
+
+   if (!strb)
+      return;
+
+   /* try a fast-path readpixels before anything else */
+   if (st_fast_readpixels(ctx, strb, x, y, width, height,
+                          format, type, pack, dest)) {
+      /* success! */
+      _mesa_unmap_readpix_pbo(ctx, &clippedPacking);
+      return;
+   }
+
+   if (format == GL_RGBA && type == GL_FLOAT) {
+      /* write tile(row) directly into user's buffer */
+      df = (GLfloat *) _mesa_image_address2d(&clippedPacking, dest, width,
+                                             height, format, type, 0, 0);
+      dfStride = width * 4;
+   }
+   else {
+      /* write tile(row) into temp row buffer */
+      df = (GLfloat *) temp;
+      dfStride = 0;
+   }
+
+   /* determine bottom-to-top vs. top-to-bottom order */
+   if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
+      y = strb->Base.Height - 1 - y;
+      yStep = -1;
+   }
+   else {
+      yStep = 1;
+   }
+
+   /* Create a CPU-READ surface/view into the renderbuffer's texture */
+   surf = screen->get_tex_surface(screen, strb->texture,  0, 0, 0,
+                                  PIPE_BUFFER_USAGE_CPU_READ);
+
+   /*
+    * Copy pixels from pipe_surface to user memory
+    */
+   {
+      /* dest of first pixel in client memory */
+      GLubyte *dst = _mesa_image_address2d(&clippedPacking, dest, width,
+                                           height, format, type, 0, 0);
+      /* dest row stride */
+      const GLint dstStride = _mesa_image_row_stride(&clippedPacking, width,
+                                                     format, type);
+
+      if (surf->format == PIPE_FORMAT_S8Z24_UNORM ||
+          surf->format == PIPE_FORMAT_X8Z24_UNORM) {
+         if (format == GL_DEPTH_COMPONENT) {
+            for (i = 0; i < height; i++) {
+               GLuint ztemp[MAX_WIDTH];
+               GLfloat zfloat[MAX_WIDTH];
+               const double scale = 1.0 / ((1 << 24) - 1);
+               pipe_get_tile_raw(surf, x, y, width, 1, ztemp, 0);
+               y += yStep;
+               for (j = 0; j < width; j++) {
+                  zfloat[j] = (float) (scale * (ztemp[j] & 0xffffff));
+               }
+               _mesa_pack_depth_span(ctx, width, dst, type,
+                                     zfloat, &clippedPacking);
+               dst += dstStride;
+            }
+         }
+         else {
+            /* untested, but simple: */
+            assert(format == GL_DEPTH_STENCIL_EXT);
+            for (i = 0; i < height; i++) {
+               pipe_get_tile_raw(surf, x, y, width, 1, dst, 0);
+               y += yStep;
+               dst += dstStride;
+            }
+         }
+      }
+      else if (surf->format == PIPE_FORMAT_Z16_UNORM) {
+         for (i = 0; i < height; i++) {
+            GLushort ztemp[MAX_WIDTH];
+            GLfloat zfloat[MAX_WIDTH];
+            const double scale = 1.0 / 0xffff;
+            pipe_get_tile_raw(surf, x, y, width, 1, ztemp, 0);
+            y += yStep;
+            for (j = 0; j < width; j++) {
+               zfloat[j] = (float) (scale * ztemp[j]);
+            }
+            _mesa_pack_depth_span(ctx, width, dst, type,
+                                  zfloat, &clippedPacking);
+            dst += dstStride;
+         }
+      }
+      else if (surf->format == PIPE_FORMAT_Z32_UNORM) {
+         for (i = 0; i < height; i++) {
+            GLuint ztemp[MAX_WIDTH];
+            GLfloat zfloat[MAX_WIDTH];
+            const double scale = 1.0 / 0xffffffff;
+            pipe_get_tile_raw(surf, x, y, width, 1, ztemp, 0);
+            y += yStep;
+            for (j = 0; j < width; j++) {
+               zfloat[j] = (float) (scale * ztemp[j]);
+            }
+            _mesa_pack_depth_span(ctx, width, dst, type,
+                                  zfloat, &clippedPacking);
+            dst += dstStride;
+         }
+      }
+      else {
+         /* RGBA format */
+         /* Do a row at a time to flip image data vertically */
+         for (i = 0; i < height; i++) {
+            pipe_get_tile_rgba(surf, x, y, width, 1, df);
+            y += yStep;
+            df += dfStride;
+            if (!dfStride) {
+               _mesa_pack_rgba_span_float(ctx, width, temp, format, type, dst,
+                                          &clippedPacking, transferOps);
+               dst += dstStride;
+            }
+         }
+      }
+   }
+
+   pipe_surface_reference(&surf, NULL);
+
+   _mesa_unmap_readpix_pbo(ctx, &clippedPacking);
+}
+
+
+void st_init_readpixels_functions(struct dd_function_table *functions)
+{
+   functions->ReadPixels = st_readpixels;
+}
diff --git a/src/mesa/state_tracker/st_cb_readpixels.h b/src/mesa/state_tracker/st_cb_readpixels.h
new file mode 100644
index 0000000000..9e151be51f
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_readpixels.h
@@ -0,0 +1,45 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_READPIXELS_H
+#define ST_CB_READPIXELS_H
+
+extern struct st_renderbuffer *
+st_get_color_read_renderbuffer(GLcontext *ctx);
+
+extern void
+st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
+                       GLsizei width, GLsizei height, GLenum type,
+                       const struct gl_pixelstore_attrib *packing,
+                       GLvoid *pixels);
+
+extern void
+st_init_readpixels_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_READPIXELS_H */
diff --git a/src/mesa/state_tracker/st_cb_strings.c b/src/mesa/state_tracker/st_cb_strings.c
new file mode 100644
index 0000000000..2036ccafbf
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_strings.c
@@ -0,0 +1,84 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+#include "main/glheader.h"
+#include "main/macros.h"
+#include "main/version.h"
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "st_context.h"
+#include "st_cb_strings.h"
+
+#define ST_VERSION_STRING "0.2"
+
+static const GLubyte *
+st_get_string(GLcontext * ctx, GLenum name)
+{
+   struct st_context *st = st_context(ctx);
+   struct pipe_screen *screen = st->pipe->screen;
+
+   switch (name) {
+   case GL_VENDOR: {
+      const char *vendor = screen->get_vendor( screen );
+      const char *tungsten = "Tungsten Graphics, Inc.";
+
+      /* Tungsten Graphics, Inc. developed the state_tracker module
+       * (and much of Mesa), but the driver itself may come from elsewhere.
+       * The additional string allows "and XyzCorp" to reflect this.
+       */
+      if (vendor && strcmp(vendor, tungsten) != 0)
+	 util_snprintf(st->vendor, sizeof(st->vendor),
+                  "%s and %s", tungsten, vendor);
+      else
+	 util_snprintf(st->vendor, sizeof(st->vendor), "%s", tungsten);
+
+      return (GLubyte *) st->vendor;
+   }
+
+   case GL_RENDERER:
+      util_snprintf(st->renderer, sizeof(st->renderer), "Gallium %s on %s", 
+               ST_VERSION_STRING,
+	       screen->get_name( screen ));
+
+      return (GLubyte *) st->renderer;
+
+   default:
+      return NULL;
+   }
+}
+
+
+void st_init_string_functions(struct dd_function_table *functions)
+{
+   functions->GetString = st_get_string;
+}
diff --git a/src/mesa/state_tracker/st_cb_strings.h b/src/mesa/state_tracker/st_cb_strings.h
new file mode 100644
index 0000000000..3b765aaa59
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_strings.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_STRINGS_H
+#define ST_CB_STRINGS_H
+
+
+extern void
+st_init_string_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_CLEAR_H */
+
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
new file mode 100644
index 0000000000..9279a768c8
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -0,0 +1,1549 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "main/imports.h"
+#if FEATURE_convolve
+#include "main/convolve.h"
+#endif
+#include "main/enums.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "main/mipmap.h"
+#include "main/pixel.h"
+#include "main/texcompress.h"
+#include "main/texformat.h"
+#include "main/teximage.h"
+#include "main/texobj.h"
+#include "main/texstore.h"
+
+#include "state_tracker/st_context.h"
+#include "state_tracker/st_cb_fbo.h"
+#include "state_tracker/st_cb_texture.h"
+#include "state_tracker/st_format.h"
+#include "state_tracker/st_public.h"
+#include "state_tracker/st_texture.h"
+#include "state_tracker/st_gen_mipmap.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_tile.h"
+#include "util/u_blit.h"
+
+
+#define DBG if (0) printf
+
+
+static enum pipe_texture_target
+gl_target_to_pipe(GLenum target)
+{
+   switch (target) {
+   case GL_TEXTURE_1D:
+      return PIPE_TEXTURE_1D;
+
+   case GL_TEXTURE_2D:
+   case GL_TEXTURE_RECTANGLE_NV:
+      return PIPE_TEXTURE_2D;
+
+   case GL_TEXTURE_3D:
+      return PIPE_TEXTURE_3D;
+
+   case GL_TEXTURE_CUBE_MAP_ARB:
+      return PIPE_TEXTURE_CUBE;
+
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+
+/**
+ * Return nominal bytes per texel for a compressed format, 0 for non-compressed
+ * format.
+ */
+static int
+compressed_num_bytes(GLuint mesaFormat)
+{
+   switch(mesaFormat) {
+#if FEATURE_texture_fxt1
+   case MESA_FORMAT_RGB_FXT1:
+   case MESA_FORMAT_RGBA_FXT1:
+#endif
+#if FEATURE_texture_s3tc
+   case MESA_FORMAT_RGB_DXT1:
+   case MESA_FORMAT_RGBA_DXT1:
+      return 2;
+   case MESA_FORMAT_RGBA_DXT3:
+   case MESA_FORMAT_RGBA_DXT5:
+      return 4;
+#endif
+   default:
+      return 0;
+   }
+}
+
+
+/** called via ctx->Driver.NewTextureImage() */
+static struct gl_texture_image *
+st_NewTextureImage(GLcontext * ctx)
+{
+   DBG("%s\n", __FUNCTION__);
+   (void) ctx;
+   return (struct gl_texture_image *) CALLOC_STRUCT(st_texture_image);
+}
+
+
+/** called via ctx->Driver.NewTextureObject() */
+static struct gl_texture_object *
+st_NewTextureObject(GLcontext * ctx, GLuint name, GLenum target)
+{
+   struct st_texture_object *obj = CALLOC_STRUCT(st_texture_object);
+
+   DBG("%s\n", __FUNCTION__);
+   _mesa_initialize_texture_object(&obj->base, name, target);
+
+   return &obj->base;
+}
+
+/** called via ctx->Driver.DeleteTextureImage() */
+static void 
+st_DeleteTextureObject(GLcontext *ctx,
+                       struct gl_texture_object *texObj)
+{
+   struct st_texture_object *stObj = st_texture_object(texObj);
+   if (stObj->pt)
+      pipe_texture_reference(&stObj->pt, NULL);
+
+   _mesa_delete_texture_object(ctx, texObj);
+}
+
+
+/** called via ctx->Driver.FreeTexImageData() */
+static void
+st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage)
+{
+   struct st_texture_image *stImage = st_texture_image(texImage);
+
+   DBG("%s\n", __FUNCTION__);
+
+   if (stImage->pt) {
+      pipe_texture_reference(&stImage->pt, NULL);
+   }
+
+   if (texImage->Data) {
+      _mesa_align_free(texImage->Data);
+      texImage->Data = NULL;
+   }
+}
+
+
+/**
+ * From linux kernel i386 header files, copes with odd sizes better
+ * than COPY_DWORDS would:
+ * XXX Put this in src/mesa/main/imports.h ???
+ */
+#if defined(i386) || defined(__i386__)
+static INLINE void *
+__memcpy(void *to, const void *from, size_t n)
+{
+   int d0, d1, d2;
+   __asm__ __volatile__("rep ; movsl\n\t"
+                        "testb $2,%b4\n\t"
+                        "je 1f\n\t"
+                        "movsw\n"
+                        "1:\ttestb $1,%b4\n\t"
+                        "je 2f\n\t"
+                        "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2)
+                        :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from)
+                        :"memory");
+   return (to);
+}
+#else
+#define __memcpy(a,b,c) memcpy(a,b,c)
+#endif
+
+
+/**
+ * The system memcpy (at least on ubuntu 5.10) has problems copying
+ * to agp (writecombined) memory from a source which isn't 64-byte
+ * aligned - there is a 4x performance falloff.
+ *
+ * The x86 __memcpy is immune to this but is slightly slower
+ * (10%-ish) than the system memcpy.
+ *
+ * The sse_memcpy seems to have a slight cliff at 64/32 bytes, but
+ * isn't much faster than x86_memcpy for agp copies.
+ * 
+ * TODO: switch dynamically.
+ */
+static void *
+do_memcpy(void *dest, const void *src, size_t n)
+{
+   if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) {
+      return __memcpy(dest, src, n);
+   }
+   else
+      return memcpy(dest, src, n);
+}
+
+
+static int
+logbase2(int n)
+{
+   GLint i = 1, log2 = 0;
+   while (n > i) {
+      i *= 2;
+      log2++;
+   }
+   return log2;
+}
+
+
+/**
+ * Allocate a pipe_texture object for the given st_texture_object using
+ * the given st_texture_image to guess the mipmap size/levels.
+ *
+ * [comments...]
+ * Otherwise, store it in memory if (Border != 0) or (any dimension ==
+ * 1).
+ *    
+ * Otherwise, if max_level >= level >= min_level, create texture with
+ * space for images from min_level down to max_level.
+ *
+ * Otherwise, create texture with space for images from (level 0)..(1x1).
+ * Consider pruning this texture at a validation if the saving is worth it.
+ */
+static void
+guess_and_alloc_texture(struct st_context *st,
+			struct st_texture_object *stObj,
+			const struct st_texture_image *stImage)
+{
+   GLuint firstLevel;
+   GLuint lastLevel;
+   GLuint width = stImage->base.Width2;  /* size w/out border */
+   GLuint height = stImage->base.Height2;
+   GLuint depth = stImage->base.Depth2;
+   GLuint i, comp_byte = 0;
+   enum pipe_format fmt;
+
+   DBG("%s\n", __FUNCTION__);
+
+   assert(!stObj->pt);
+
+   if (stObj->pt &&
+       (GLint) stImage->level > stObj->base.BaseLevel &&
+       (stImage->base.Width == 1 ||
+        (stObj->base.Target != GL_TEXTURE_1D &&
+         stImage->base.Height == 1) ||
+        (stObj->base.Target == GL_TEXTURE_3D &&
+         stImage->base.Depth == 1)))
+      return;
+
+   /* If this image disrespects BaseLevel, allocate from level zero.
+    * Usually BaseLevel == 0, so it's unlikely to happen.
+    */
+   if ((GLint) stImage->level < stObj->base.BaseLevel)
+      firstLevel = 0;
+   else
+      firstLevel = stObj->base.BaseLevel;
+
+
+   /* Figure out image dimensions at start level. 
+    */
+   for (i = stImage->level; i > firstLevel; i--) {
+      if (width != 1)
+         width <<= 1;
+      if (height != 1)
+         height <<= 1;
+      if (depth != 1)
+         depth <<= 1;
+   }
+
+   if (width == 0 || height == 0 || depth == 0) {
+      /* no texture needed */
+      return;
+   }
+
+   /* Guess a reasonable value for lastLevel.  This is probably going
+    * to be wrong fairly often and might mean that we have to look at
+    * resizable buffers, or require that buffers implement lazy
+    * pagetable arrangements.
+    */
+   if ((stObj->base.MinFilter == GL_NEAREST ||
+        stObj->base.MinFilter == GL_LINEAR) &&
+       stImage->level == firstLevel) {
+      lastLevel = firstLevel;
+   }
+   else {
+      GLuint l2width = logbase2(width);
+      GLuint l2height = logbase2(height);
+      GLuint l2depth = logbase2(depth);
+      lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth);
+   }
+
+   if (stImage->base.IsCompressed)
+      comp_byte = compressed_num_bytes(stImage->base.TexFormat->MesaFormat);
+
+   fmt = st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat);
+   stObj->pt = st_texture_create(st,
+                                 gl_target_to_pipe(stObj->base.Target),
+                                 fmt,
+                                 lastLevel,
+                                 width,
+                                 height,
+                                 depth,
+                                 comp_byte,
+                                 ( (pf_is_depth_stencil(fmt) ?
+                                   PIPE_TEXTURE_USAGE_DEPTH_STENCIL :
+                                   PIPE_TEXTURE_USAGE_RENDER_TARGET) |
+                                   PIPE_TEXTURE_USAGE_SAMPLER ));
+
+   DBG("%s - success\n", __FUNCTION__);
+}
+
+
+/**
+ * Adjust pixel unpack params and image dimensions to strip off the
+ * texture border.
+ * Gallium doesn't support texture borders.  They've seldem been used
+ * and seldom been implemented correctly anyway.
+ * \param unpackNew  returns the new pixel unpack parameters
+ */
+static void
+strip_texture_border(GLint border,
+                     GLint *width, GLint *height, GLint *depth,
+                     const struct gl_pixelstore_attrib *unpack,
+                     struct gl_pixelstore_attrib *unpackNew)
+{
+   assert(border > 0);  /* sanity check */
+
+   *unpackNew = *unpack;
+
+   if (unpackNew->RowLength == 0)
+      unpackNew->RowLength = *width;
+
+   if (depth && unpackNew->ImageHeight == 0)
+      unpackNew->ImageHeight = *height;
+
+   unpackNew->SkipPixels += border;
+   if (height)
+      unpackNew->SkipRows += border;
+   if (depth)
+      unpackNew->SkipImages += border;
+
+   assert(*width >= 3);
+   *width = *width - 2 * border;
+   if (height && *height >= 3)
+      *height = *height - 2 * border;
+   if (depth && *depth >= 3)
+      *depth = *depth - 2 * border;
+}
+
+
+/**
+ * Do glTexImage1/2/3D().
+ */
+static void
+st_TexImage(GLcontext * ctx,
+            GLint dims,
+            GLenum target, GLint level,
+            GLint internalFormat,
+            GLint width, GLint height, GLint depth,
+            GLint border,
+            GLenum format, GLenum type, const void *pixels,
+            const struct gl_pixelstore_attrib *unpack,
+            struct gl_texture_object *texObj,
+            struct gl_texture_image *texImage,
+            GLsizei imageSize, int compressed)
+{
+   struct st_texture_object *stObj = st_texture_object(texObj);
+   struct st_texture_image *stImage = st_texture_image(texImage);
+   GLint postConvWidth, postConvHeight;
+   GLint texelBytes, sizeInBytes;
+   GLuint dstRowStride;
+   struct gl_pixelstore_attrib unpackNB;
+
+   DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__,
+       _mesa_lookup_enum_by_nr(target), level, width, height, depth, border);
+
+   /* gallium does not support texture borders, strip it off */
+   if (border) {
+      strip_texture_border(border, &width, &height, &depth,
+                           unpack, &unpackNB);
+      unpack = &unpackNB;
+      texImage->Width = width;
+      texImage->Height = height;
+      texImage->Depth = depth;
+      texImage->Border = 0;
+      border = 0;
+   }
+
+   postConvWidth = width;
+   postConvHeight = height;
+
+   stImage->face = _mesa_tex_target_to_face(target);
+   stImage->level = level;
+
+#if FEATURE_convolve
+   if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) {
+      _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth,
+                                         &postConvHeight);
+   }
+#endif
+
+   /* choose the texture format */
+   texImage->TexFormat = st_ChooseTextureFormat(ctx, internalFormat,
+                                                format, type);
+
+   _mesa_set_fetch_functions(texImage, dims);
+
+   if (texImage->TexFormat->TexelBytes == 0) {
+      /* must be a compressed format */
+      texelBytes = 0;
+      texImage->IsCompressed = GL_TRUE;
+      texImage->CompressedSize =
+	 ctx->Driver.CompressedTextureSize(ctx, texImage->Width,
+					   texImage->Height, texImage->Depth,
+					   texImage->TexFormat->MesaFormat);
+   }
+   else {
+      texelBytes = texImage->TexFormat->TexelBytes;
+      
+      /* Minimum pitch of 32 bytes */
+      if (postConvWidth * texelBytes < 32) {
+	 postConvWidth = 32 / texelBytes;
+	 texImage->RowStride = postConvWidth;
+      }
+      
+      /* we'll set RowStride elsewhere when the texture is a "mapped" state */
+      /*assert(texImage->RowStride == postConvWidth);*/
+   }
+
+   /* Release the reference to a potentially orphaned buffer.   
+    * Release any old malloced memory.
+    */
+   if (stImage->pt) {
+      pipe_texture_reference(&stImage->pt, NULL);
+      assert(!texImage->Data);
+   }
+   else if (texImage->Data) {
+      _mesa_align_free(texImage->Data);
+   }
+
+   if (width == 0 || height == 0 || depth == 0) {
+      /* stop after freeing old image */
+      return;
+   }
+
+   /* If this is the only mipmap level in the texture, could call
+    * bmBufferData with NULL data to free the old block and avoid
+    * waiting on any outstanding fences.
+    */
+   if (stObj->pt &&
+       (stObj->teximage_realloc ||
+        (/*stObj->pt->first_level == level &&*/
+         stObj->pt->last_level == level &&
+         stObj->pt->target != PIPE_TEXTURE_CUBE &&
+         !st_texture_match_image(stObj->pt, &stImage->base,
+                                 stImage->face, stImage->level)))) {
+
+      DBG("release it\n");
+      pipe_texture_reference(&stObj->pt, NULL);
+      assert(!stObj->pt);
+      stObj->teximage_realloc = FALSE;
+   }
+
+   if (!stObj->pt) {
+      guess_and_alloc_texture(ctx->st, stObj, stImage);
+      if (!stObj->pt) {
+         /* Probably out of memory.
+          * Try flushing any pending rendering, then retry.
+          */
+         st_finish(ctx->st);
+         guess_and_alloc_texture(ctx->st, stObj, stImage);
+         if (!stObj->pt) {
+            _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+            return;
+         }
+      }
+   }
+
+   assert(!stImage->pt);
+
+   if (stObj->pt &&
+       st_texture_match_image(stObj->pt, &stImage->base,
+                                 stImage->face, stImage->level)) {
+
+      pipe_texture_reference(&stImage->pt, stObj->pt);
+      assert(stImage->pt);
+   }
+
+   if (!stImage->pt)
+      DBG("XXX: Image did not fit into texture - storing in local memory!\n");
+
+   /* st_CopyTexImage calls this function with pixels == NULL, with
+    * the expectation that the texture will be set up but nothing
+    * more will be done.  This is where those calls return:
+    */
+   if (compressed) {
+      pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels,
+						      unpack,
+						      "glCompressedTexImage");
+   } else {
+      pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1,
+					   format, type,
+					   pixels, unpack, "glTexImage");
+   }
+   if (!pixels)
+      return;
+
+   if (stImage->pt) {
+      texImage->Data = st_texture_image_map(ctx->st, stImage, 0,
+                                            PIPE_BUFFER_USAGE_CPU_WRITE);
+      if (stImage->surface)
+         dstRowStride = stImage->surface->stride;
+   }
+   else {
+      /* Allocate regular memory and store the image there temporarily.   */
+      if (texImage->IsCompressed) {
+         sizeInBytes = texImage->CompressedSize;
+         dstRowStride =
+            _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width);
+         assert(dims != 3);
+      }
+      else {
+         dstRowStride = postConvWidth * texelBytes;
+         sizeInBytes = depth * dstRowStride * postConvHeight;
+      }
+
+      texImage->Data = _mesa_align_malloc(sizeInBytes, 16);
+   }
+
+   if (!texImage->Data) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+      return;
+   }
+
+   DBG("Upload image %dx%dx%d row_len %x pitch %x\n",
+       width, height, depth, width * texelBytes, dstRowStride);
+
+   /* Copy data.  Would like to know when it's ok for us to eg. use
+    * the blitter to copy.  Or, use the hardware to do the format
+    * conversion and copy:
+    */
+   if (compressed) {
+      memcpy(texImage->Data, pixels, imageSize);
+   }
+   else {
+      GLuint srcImageStride = _mesa_image_image_stride(unpack, width, height,
+						       format, type);
+      int i;
+      const GLubyte *src = (const GLubyte *) pixels;
+
+      for (i = 0; i++ < depth;) {
+	 if (!texImage->TexFormat->StoreImage(ctx, dims, 
+					      texImage->_BaseFormat, 
+					      texImage->TexFormat, 
+					      texImage->Data,
+					      0, 0, 0, /* dstX/Y/Zoffset */
+					      dstRowStride,
+					      texImage->ImageOffsets,
+					      width, height, 1,
+					      format, type, src, unpack)) {
+	    _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+	 }
+
+	 if (stImage->pt && i < depth) {
+	    st_texture_image_unmap(ctx->st, stImage);
+	    texImage->Data = st_texture_image_map(ctx->st, stImage, i,
+                                                  PIPE_BUFFER_USAGE_CPU_WRITE);
+	    src += srcImageStride;
+	 }
+      }
+   }
+
+   _mesa_unmap_teximage_pbo(ctx, unpack);
+
+   if (stImage->pt) {
+      st_texture_image_unmap(ctx->st, stImage);
+      texImage->Data = NULL;
+   }
+
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+   }
+}
+
+
+static void
+st_TexImage3D(GLcontext * ctx,
+              GLenum target, GLint level,
+              GLint internalFormat,
+              GLint width, GLint height, GLint depth,
+              GLint border,
+              GLenum format, GLenum type, const void *pixels,
+              const struct gl_pixelstore_attrib *unpack,
+              struct gl_texture_object *texObj,
+              struct gl_texture_image *texImage)
+{
+   st_TexImage(ctx, 3, target, level,
+                 internalFormat, width, height, depth, border,
+                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+}
+
+
+static void
+st_TexImage2D(GLcontext * ctx,
+              GLenum target, GLint level,
+              GLint internalFormat,
+              GLint width, GLint height, GLint border,
+              GLenum format, GLenum type, const void *pixels,
+              const struct gl_pixelstore_attrib *unpack,
+              struct gl_texture_object *texObj,
+              struct gl_texture_image *texImage)
+{
+   st_TexImage(ctx, 2, target, level,
+                 internalFormat, width, height, 1, border,
+                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+}
+
+
+static void
+st_TexImage1D(GLcontext * ctx,
+              GLenum target, GLint level,
+              GLint internalFormat,
+              GLint width, GLint border,
+              GLenum format, GLenum type, const void *pixels,
+              const struct gl_pixelstore_attrib *unpack,
+              struct gl_texture_object *texObj,
+              struct gl_texture_image *texImage)
+{
+   st_TexImage(ctx, 1, target, level,
+                 internalFormat, width, 1, 1, border,
+                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+}
+
+
+static void
+st_CompressedTexImage2D(GLcontext *ctx, GLenum target, GLint level,
+                        GLint internalFormat,
+                        GLint width, GLint height, GLint border,
+                        GLsizei imageSize, const GLvoid *data,
+                        struct gl_texture_object *texObj,
+                        struct gl_texture_image *texImage)
+{
+   st_TexImage(ctx, 2, target, level,
+		 internalFormat, width, height, 1, border,
+		 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1);
+}
+
+
+/**
+ * Need to map texture image into memory before copying image data,
+ * then unmap it.
+ */
+static void
+st_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
+                 GLenum format, GLenum type, GLvoid * pixels,
+                 struct gl_texture_object *texObj,
+                 struct gl_texture_image *texImage, int compressed)
+{
+   struct st_texture_image *stImage = st_texture_image(texImage);
+   GLuint dstImageStride = _mesa_image_image_stride(&ctx->Pack,
+                                                    texImage->Width,
+						    texImage->Height,
+                                                    format, type);
+   GLuint depth;
+   GLuint i;
+   GLubyte *dest;
+
+   /* Map */
+   if (stImage->pt) {
+      /* Image is stored in hardware format in a buffer managed by the
+       * kernel.  Need to explicitly map and unmap it.
+       */
+      texImage->Data = st_texture_image_map(ctx->st, stImage, 0,
+                                            PIPE_BUFFER_USAGE_CPU_READ);
+      texImage->RowStride = stImage->surface->stride / stImage->pt->block.size;
+   }
+   else {
+      /* Otherwise, the image should actually be stored in
+       * texImage->Data.  This is pretty confusing for
+       * everybody, I'd much prefer to separate the two functions of
+       * texImage->Data - storage for texture images in main memory
+       * and access (ie mappings) of images.  In other words, we'd
+       * create a new texImage->Map field and leave Data simply for
+       * storage.
+       */
+      assert(texImage->Data);
+   }
+
+   depth = texImage->Depth;
+   texImage->Depth = 1;
+
+   dest = (GLubyte *) pixels;
+
+   for (i = 0; i++ < depth;) {
+      if (compressed) {
+	 _mesa_get_compressed_teximage(ctx, target, level, dest,
+				       texObj, texImage);
+      } else {
+	 _mesa_get_teximage(ctx, target, level, format, type, dest,
+			    texObj, texImage);
+      }
+
+      if (stImage->pt && i < depth) {
+	 st_texture_image_unmap(ctx->st, stImage);
+	 texImage->Data = st_texture_image_map(ctx->st, stImage, i,
+                                               PIPE_BUFFER_USAGE_CPU_READ);
+	 dest += dstImageStride;
+      }
+   }
+
+   texImage->Depth = depth;
+
+   /* Unmap */
+   if (stImage->pt) {
+      st_texture_image_unmap(ctx->st, stImage);
+      texImage->Data = NULL;
+   }
+}
+
+
+static void
+st_GetTexImage(GLcontext * ctx, GLenum target, GLint level,
+               GLenum format, GLenum type, GLvoid * pixels,
+               struct gl_texture_object *texObj,
+               struct gl_texture_image *texImage)
+{
+   st_get_tex_image(ctx, target, level, format, type, pixels,
+                    texObj, texImage, 0);
+}
+
+
+static void
+st_GetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
+                         GLvoid *pixels,
+                         struct gl_texture_object *texObj,
+                         struct gl_texture_image *texImage)
+{
+   st_get_tex_image(ctx, target, level, 0, 0, pixels,
+                    (struct gl_texture_object *) texObj,
+                    (struct gl_texture_image *) texImage, 1);
+}
+
+
+
+static void
+st_TexSubimage(GLcontext * ctx,
+               GLint dims,
+               GLenum target, GLint level,
+               GLint xoffset, GLint yoffset, GLint zoffset,
+               GLint width, GLint height, GLint depth,
+               GLenum format, GLenum type, const void *pixels,
+               const struct gl_pixelstore_attrib *packing,
+               struct gl_texture_object *texObj,
+               struct gl_texture_image *texImage)
+{
+   struct st_texture_image *stImage = st_texture_image(texImage);
+   GLuint dstRowStride;
+   GLuint srcImageStride = _mesa_image_image_stride(packing, width, height,
+						    format, type);
+   int i;
+   const GLubyte *src;
+
+   DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__,
+       _mesa_lookup_enum_by_nr(target),
+       level, xoffset, yoffset, width, height);
+
+   pixels =
+      _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format,
+                                  type, pixels, packing, "glTexSubImage2D");
+   if (!pixels)
+      return;
+
+   /* Map buffer if necessary.  Need to lock to prevent other contexts
+    * from uploading the buffer under us.
+    */
+   if (stImage->pt) {
+      texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset, 
+                                            PIPE_BUFFER_USAGE_CPU_WRITE);
+      if (stImage->surface)
+         dstRowStride = stImage->surface->stride;
+   }
+
+   if (!texImage->Data) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage");
+      return;
+   }
+
+   src = (const GLubyte *) pixels;
+
+   for (i = 0; i++ < depth;) {
+      if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat,
+					   texImage->TexFormat,
+					   texImage->Data,
+					   xoffset, yoffset, 0,
+					   dstRowStride,
+					   texImage->ImageOffsets,
+					   width, height, 1,
+					   format, type, src, packing)) {
+	 _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage");
+      }
+
+      if (stImage->pt && i < depth) {
+         /* map next slice of 3D texture */
+	 st_texture_image_unmap(ctx->st, stImage);
+	 texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset + i,
+                                               PIPE_BUFFER_USAGE_CPU_WRITE);
+	 src += srcImageStride;
+      }
+   }
+
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+   }
+
+   _mesa_unmap_teximage_pbo(ctx, packing);
+
+   if (stImage->pt) {
+      st_texture_image_unmap(ctx->st, stImage);
+      texImage->Data = NULL;
+   }
+}
+
+
+
+static void
+st_TexSubImage3D(GLcontext * ctx,
+                   GLenum target,
+                   GLint level,
+                   GLint xoffset, GLint yoffset, GLint zoffset,
+                   GLsizei width, GLsizei height, GLsizei depth,
+                   GLenum format, GLenum type,
+                   const GLvoid * pixels,
+                   const struct gl_pixelstore_attrib *packing,
+                   struct gl_texture_object *texObj,
+                   struct gl_texture_image *texImage)
+{
+   st_TexSubimage(ctx, 3, target, level,
+                  xoffset, yoffset, zoffset,
+                  width, height, depth,
+                  format, type, pixels, packing, texObj, texImage);
+}
+
+
+static void
+st_TexSubImage2D(GLcontext * ctx,
+                   GLenum target,
+                   GLint level,
+                   GLint xoffset, GLint yoffset,
+                   GLsizei width, GLsizei height,
+                   GLenum format, GLenum type,
+                   const GLvoid * pixels,
+                   const struct gl_pixelstore_attrib *packing,
+                   struct gl_texture_object *texObj,
+                   struct gl_texture_image *texImage)
+{
+   st_TexSubimage(ctx, 2, target, level,
+                  xoffset, yoffset, 0,
+                  width, height, 1,
+                  format, type, pixels, packing, texObj, texImage);
+}
+
+
+static void
+st_TexSubImage1D(GLcontext * ctx,
+                   GLenum target,
+                   GLint level,
+                   GLint xoffset,
+                   GLsizei width,
+                   GLenum format, GLenum type,
+                   const GLvoid * pixels,
+                   const struct gl_pixelstore_attrib *packing,
+                   struct gl_texture_object *texObj,
+                   struct gl_texture_image *texImage)
+{
+   st_TexSubimage(ctx, 1, target, level,
+                  xoffset, 0, 0,
+                  width, 1, 1,
+                  format, type, pixels, packing, texObj, texImage);
+}
+
+
+
+/**
+ * Return 0 for GL_TEXTURE_CUBE_MAP_POSITIVE_X,
+ *        1 for GL_TEXTURE_CUBE_MAP_NEGATIVE_X,
+ *        etc.
+ * XXX duplicated from main/teximage.c
+ */
+static uint
+texture_face(GLenum target)
+{
+   if (target >= GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB &&
+       target <= GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB)
+      return (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
+   else
+      return 0;
+}
+
+
+
+/**
+ * Do a CopyTexSubImage operation by mapping the source surface and
+ * dest surface and using get_tile()/put_tile() to access the pixels/texels.
+ *
+ * Note: srcY=0=TOP of renderbuffer
+ */
+static void
+fallback_copy_texsubimage(GLcontext *ctx,
+                          GLenum target,
+                          GLint level,
+                          struct st_renderbuffer *strb,
+                          struct st_texture_image *stImage,
+                          GLenum baseFormat,
+                          GLint destX, GLint destY, GLint destZ,
+                          GLint srcX, GLint srcY,
+                          GLsizei width, GLsizei height)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const uint face = texture_face(target);
+   struct pipe_texture *pt = stImage->pt;
+   struct pipe_surface *src_surf, *dest_surf;
+
+   /* We'd use strb->surface, here but it's created for GPU read/write only */
+   src_surf = pipe->screen->get_tex_surface( pipe->screen,
+                                             strb->texture,
+                                             0, 0, 0,
+                                             PIPE_BUFFER_USAGE_CPU_READ);
+
+   dest_surf = screen->get_tex_surface(screen, pt, face, level, destZ,
+                                       PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   assert(width <= MAX_WIDTH);
+
+   if (baseFormat == GL_DEPTH_COMPONENT) {
+      const GLboolean scaleOrBias = (ctx->Pixel.DepthScale != 1.0F ||
+                                     ctx->Pixel.DepthBias != 0.0F);
+      GLint row, yStep;
+
+      /* determine bottom-to-top vs. top-to-bottom order for src buffer */
+      if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
+         srcY = strb->Base.Height - 1 - srcY;
+         yStep = -1;
+      }
+      else {
+         yStep = 1;
+      }
+
+      /* To avoid a large temp memory allocation, do copy row by row */
+      for (row = 0; row < height; row++, srcY += yStep, destY++) {
+         uint data[MAX_WIDTH];
+         pipe_get_tile_z(src_surf, srcX, srcY, width, 1, data);
+         if (scaleOrBias) {
+            _mesa_scale_and_bias_depth_uint(ctx, width, data);
+         }
+         pipe_put_tile_z(dest_surf, destX, destY, width, 1, data);
+      }
+   }
+   else {
+      /* RGBA format */
+      GLfloat *tempSrc =
+         (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
+      GLvoid *texDest =
+         st_texture_image_map(ctx->st, stImage, 0,PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      if (tempSrc && texDest) {
+         const GLint dims = 2;
+         struct gl_texture_image *texImage = &stImage->base;
+         GLint dstRowStride = stImage->surface->stride;
+         struct gl_pixelstore_attrib unpack = ctx->DefaultPacking;
+
+         if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
+            /* need to invert src */
+            srcY = strb->Base.Height - srcY - height;
+            unpack.Invert = GL_TRUE;
+         }
+
+         /* get float/RGBA image from framebuffer */
+         /* XXX this usually involves a lot of int/float conversion.
+          * try to avoid that someday.
+          */
+         pipe_get_tile_rgba(src_surf, srcX, srcY, width, height, tempSrc);
+
+         /* Store into texture memory.
+          * Note that this does some special things such as pixel transfer
+          * ops and format conversion.  In particular, if the dest tex format
+          * is actually RGBA but the user created the texture as GL_RGB we
+          * need to fill-in/override the alpha channel with 1.0.
+          */
+         texImage->TexFormat->StoreImage(ctx, dims,
+                                         texImage->_BaseFormat, 
+                                         texImage->TexFormat, 
+                                         texDest,
+                                         destX, destY, destZ,
+                                         dstRowStride,
+                                         texImage->ImageOffsets,
+                                         width, height, 1,
+                                         GL_RGBA, GL_FLOAT, tempSrc, /* src */
+                                         &unpack);
+      }
+      else {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage");
+      }
+
+      if (tempSrc)
+         _mesa_free(tempSrc);
+      if (texDest)
+         st_texture_image_unmap(ctx->st, stImage);
+   }
+
+   screen->tex_surface_release(screen, &dest_surf);
+   screen->tex_surface_release(screen, &src_surf);
+}
+
+
+/**
+ * Do a CopyTex[Sub]Image1/2/3D() using a hardware (blit) path if possible.
+ * Note that the region to copy has already been clipped so we know we
+ * won't read from outside the source renderbuffer's bounds.
+ *
+ * Note: srcY=0=Bottom of renderbuffer (GL convention)
+ */
+static void
+st_copy_texsubimage(GLcontext *ctx,
+                    GLenum target, GLint level,
+                    GLint destX, GLint destY, GLint destZ,
+                    GLint srcX, GLint srcY,
+                    GLsizei width, GLsizei height)
+{
+   struct gl_texture_unit *texUnit =
+      &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   struct gl_texture_object *texObj =
+      _mesa_select_tex_object(ctx, texUnit, target);
+   struct gl_texture_image *texImage =
+      _mesa_select_tex_image(ctx, texObj, target, level);
+   struct st_texture_image *stImage = st_texture_image(texImage);
+   const GLenum texBaseFormat = texImage->InternalFormat;
+   struct gl_framebuffer *fb = ctx->ReadBuffer;
+   struct st_renderbuffer *strb;
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   enum pipe_format dest_format, src_format;
+   GLboolean use_fallback = GL_TRUE;
+   GLboolean matching_base_formats;
+
+   /* any rendering in progress must complete before we grab the fb image */
+   st_finish(ctx->st);
+
+   /* determine if copying depth or color data */
+   if (texBaseFormat == GL_DEPTH_COMPONENT) {
+      strb = st_renderbuffer(fb->_DepthBuffer);
+   }
+   else if (texBaseFormat == GL_DEPTH_STENCIL_EXT) {
+      strb = st_renderbuffer(fb->_StencilBuffer);
+   }
+   else {
+      /* texBaseFormat == GL_RGB, GL_RGBA, GL_ALPHA, etc */
+      strb = st_renderbuffer(fb->_ColorReadBuffer);
+   }
+
+   assert(strb);
+   assert(strb->surface);
+   assert(stImage->pt);
+
+   src_format = strb->surface->format;
+   dest_format = stImage->pt->format;
+
+   /*
+    * Determine if the src framebuffer and dest texture have the same
+    * base format.  We need this to detect a case such as the framebuffer
+    * being GL_RGBA but the texture being GL_RGB.  If the actual hardware
+    * texture format stores RGBA we need to set A=1 (overriding the
+    * framebuffer's alpha values).  We can't do that with the blit or
+    * textured-quad paths.
+    */
+   matching_base_formats = (strb->Base._BaseFormat == texImage->_BaseFormat);
+
+   if (matching_base_formats && ctx->_ImageTransferState == 0x0) {
+      /* try potential hardware path */
+      struct pipe_surface *dest_surface = NULL;
+
+      if (src_format == dest_format) {
+         /* use surface_copy() / blit */
+         boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP);
+
+         dest_surface = screen->get_tex_surface(screen, stImage->pt,
+                                                stImage->face, stImage->level,
+                                                destZ,
+                                                PIPE_BUFFER_USAGE_GPU_WRITE);
+         if (do_flip)
+            srcY = strb->surface->height - srcY - height;
+
+         /* for surface_copy(), y=0=top, always */
+         pipe->surface_copy(pipe,
+                            do_flip,
+                            /* dest */
+                            dest_surface,
+                            destX, destY,
+                            /* src */
+                            strb->surface,
+                            srcX, srcY,
+                            /* size */
+                            width, height);
+         use_fallback = GL_FALSE;
+      }
+      else if (screen->is_format_supported(screen, src_format,
+                                           PIPE_TEXTURE_2D, 
+                                           PIPE_TEXTURE_USAGE_SAMPLER,
+                                           0) &&
+               screen->is_format_supported(screen, dest_format,
+                                           PIPE_TEXTURE_2D, 
+                                           PIPE_TEXTURE_USAGE_RENDER_TARGET,
+                                           0)) {
+         /* draw textured quad to do the copy */
+         boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP);
+         int srcY0, srcY1;
+
+         dest_surface = screen->get_tex_surface(screen, stImage->pt,
+                                                stImage->face, stImage->level,
+                                                destZ,
+                                                PIPE_BUFFER_USAGE_GPU_WRITE);
+
+         if (do_flip) {
+            srcY1 = strb->Base.Height - srcY - height;
+            srcY0 = srcY1 + height;
+         }
+         else {
+            srcY0 = srcY;
+            srcY1 = srcY0 + height;
+         }
+         util_blit_pixels(ctx->st->blit,
+                          strb->surface,
+                          srcX, srcY0,
+                          srcX + width, srcY1,
+                          dest_surface,
+                          destX, destY,
+                          destX + width, destY + height,
+                          0.0, PIPE_TEX_MIPFILTER_NEAREST);
+         use_fallback = GL_FALSE;
+      }
+
+      if (dest_surface)
+         pipe_surface_reference(&dest_surface, NULL);
+   }
+
+   if (use_fallback) {
+      /* software fallback */
+      fallback_copy_texsubimage(ctx, target, level,
+                                strb, stImage, texBaseFormat,
+                                destX, destY, destZ,
+                                srcX, srcY, width, height);
+   }
+
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+   }
+}
+
+
+
+static void
+st_CopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
+                  GLenum internalFormat,
+                  GLint x, GLint y, GLsizei width, GLint border)
+{
+   struct gl_texture_unit *texUnit =
+      &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   struct gl_texture_object *texObj =
+      _mesa_select_tex_object(ctx, texUnit, target);
+   struct gl_texture_image *texImage =
+      _mesa_select_tex_image(ctx, texObj, target, level);
+
+#if 0
+   if (border)
+      goto fail;
+#endif
+
+   /* Setup or redefine the texture object, texture and texture
+    * image.  Don't populate yet.  
+    */
+   ctx->Driver.TexImage1D(ctx, target, level, internalFormat,
+                          width, border,
+                          GL_RGBA, CHAN_TYPE, NULL,
+                          &ctx->DefaultPacking, texObj, texImage);
+
+   st_copy_texsubimage(ctx, target, level,
+                       0, 0, 0,  /* destX,Y,Z */
+                       x, y, width, 1);  /* src X, Y, size */
+}
+
+
+static void
+st_CopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
+                  GLenum internalFormat,
+                  GLint x, GLint y, GLsizei width, GLsizei height,
+                  GLint border)
+{
+   struct gl_texture_unit *texUnit =
+      &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+   struct gl_texture_object *texObj =
+      _mesa_select_tex_object(ctx, texUnit, target);
+   struct gl_texture_image *texImage =
+      _mesa_select_tex_image(ctx, texObj, target, level);
+
+   /* Setup or redefine the texture object, texture and texture
+    * image.  Don't populate yet.  
+    */
+   ctx->Driver.TexImage2D(ctx, target, level, internalFormat,
+                          width, height, border,
+                          GL_RGBA, CHAN_TYPE, NULL,
+                          &ctx->DefaultPacking, texObj, texImage);
+
+   st_copy_texsubimage(ctx, target, level,
+                       0, 0, 0,  /* destX,Y,Z */
+                       x, y, width, height);  /* src X, Y, size */
+}
+
+
+static void
+st_CopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
+                     GLint xoffset, GLint x, GLint y, GLsizei width)
+{
+   const GLint yoffset = 0, zoffset = 0;
+   const GLsizei height = 1;
+   st_copy_texsubimage(ctx, target, level,
+                       xoffset, yoffset, zoffset,  /* destX,Y,Z */
+                       x, y, width, height);  /* src X, Y, size */
+}
+
+
+static void
+st_CopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
+                     GLint xoffset, GLint yoffset,
+                     GLint x, GLint y, GLsizei width, GLsizei height)
+{
+   const GLint zoffset = 0;
+   st_copy_texsubimage(ctx, target, level,
+                       xoffset, yoffset, zoffset,  /* destX,Y,Z */
+                       x, y, width, height);  /* src X, Y, size */
+}
+
+
+static void
+st_CopyTexSubImage3D(GLcontext * ctx, GLenum target, GLint level,
+                     GLint xoffset, GLint yoffset, GLint zoffset,
+                     GLint x, GLint y, GLsizei width, GLsizei height)
+{
+   st_copy_texsubimage(ctx, target, level,
+                       xoffset, yoffset, zoffset,  /* destX,Y,Z */
+                       x, y, width, height);  /* src X, Y, size */
+}
+
+
+/**
+ * Compute which mipmap levels that really need to be sent to the hardware.
+ * This depends on the base image size, GL_TEXTURE_MIN_LOD,
+ * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL.
+ */
+static void
+calculate_first_last_level(struct st_texture_object *stObj)
+{
+   struct gl_texture_object *tObj = &stObj->base;
+
+   /* These must be signed values.  MinLod and MaxLod can be negative numbers,
+    * and having firstLevel and lastLevel as signed prevents the need for
+    * extra sign checks.
+    */
+   int firstLevel;
+   int lastLevel;
+
+   /* Yes, this looks overly complicated, but it's all needed.
+    */
+   switch (tObj->Target) {
+   case GL_TEXTURE_1D:
+   case GL_TEXTURE_2D:
+   case GL_TEXTURE_3D:
+   case GL_TEXTURE_CUBE_MAP:
+      if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) {
+         /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL.
+          */
+         firstLevel = lastLevel = tObj->BaseLevel;
+      }
+      else {
+         firstLevel = 0;
+         lastLevel = MIN2(tObj->MaxLevel,
+                          (int) tObj->Image[0][tObj->BaseLevel]->WidthLog2);
+      }
+      break;
+   case GL_TEXTURE_RECTANGLE_NV:
+   case GL_TEXTURE_4D_SGIS:
+      firstLevel = lastLevel = 0;
+      break;
+   default:
+      return;
+   }
+
+   stObj->lastLevel = lastLevel;
+}
+
+
+static void
+copy_image_data_to_texture(struct st_context *st,
+			   struct st_texture_object *stObj,
+                           GLuint dstLevel,
+			   struct st_texture_image *stImage)
+{
+   if (stImage->pt) {
+      /* Copy potentially with the blitter:
+       */
+      st_texture_image_copy(st->pipe,
+                            stObj->pt, dstLevel,  /* dest texture, level */
+                            stImage->pt, /* src texture */
+                            stImage->face
+                            );
+
+      pipe_texture_reference(&stImage->pt, NULL);
+   }
+   else if (stImage->base.Data) {
+      assert(stImage->base.Data != NULL);
+
+      /* More straightforward upload.  
+       */
+      st_texture_image_data(st->pipe,
+                               stObj->pt,
+                               stImage->face,
+                               dstLevel,
+                               stImage->base.Data,
+                               stImage->base.RowStride * 
+                               stObj->pt->block.size,
+                               stImage->base.RowStride *
+                               stImage->base.Height *
+                               stObj->pt->block.size);
+      _mesa_align_free(stImage->base.Data);
+      stImage->base.Data = NULL;
+   }
+
+   pipe_texture_reference(&stImage->pt, stObj->pt);
+}
+
+
+/**
+ * Called during state validation.  When this function is finished,
+ * the texture object should be ready for rendering.
+ * \return GL_TRUE for success, GL_FALSE for failure (out of mem)
+ */
+GLboolean
+st_finalize_texture(GLcontext *ctx,
+		    struct pipe_context *pipe,
+		    struct gl_texture_object *tObj,
+		    GLboolean *needFlush)
+{
+   struct st_texture_object *stObj = st_texture_object(tObj);
+   const GLuint nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+   int comp_byte = 0;
+   int cpp;
+   GLuint face;
+   struct st_texture_image *firstImage;
+
+   *needFlush = GL_FALSE;
+
+   /* We know/require this is true by now: 
+    */
+   assert(stObj->base._Complete);
+
+   /* What levels must the texture include at a minimum?
+    */
+   calculate_first_last_level(stObj);
+   firstImage = st_texture_image(stObj->base.Image[0][stObj->base.BaseLevel]);
+
+   /* If both firstImage and stObj point to a texture which can contain
+    * all active images, favour firstImage.  Note that because of the
+    * completeness requirement, we know that the image dimensions
+    * will match.
+    */
+   if (firstImage->pt &&
+       firstImage->pt != stObj->pt &&
+       firstImage->pt->last_level >= stObj->lastLevel) {
+
+      pipe_texture_reference(&stObj->pt, firstImage->pt);
+   }
+
+   /* FIXME: determine format block instead of cpp */
+   if (firstImage->base.IsCompressed) {
+      comp_byte = compressed_num_bytes(firstImage->base.TexFormat->MesaFormat);
+      cpp = comp_byte;
+   }
+   else {
+      cpp = firstImage->base.TexFormat->TexelBytes;
+   }
+
+   /* If we already have a gallium texture, check that it matches the texture
+    * object's format, target, size, num_levels, etc.
+    */
+   if (stObj->pt) {
+      const enum pipe_format fmt =
+         st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat);
+      if (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) ||
+          stObj->pt->format != fmt ||
+          stObj->pt->last_level < stObj->lastLevel ||
+          stObj->pt->width[0] != firstImage->base.Width2 ||
+          stObj->pt->height[0] != firstImage->base.Height2 ||
+          stObj->pt->depth[0] != firstImage->base.Depth2 ||
+          stObj->pt->block.size/stObj->pt->block.width != cpp || /* Nominal bytes per pixel */
+          stObj->pt->compressed != firstImage->base.IsCompressed) {
+         pipe_texture_release(&stObj->pt);
+         ctx->st->dirty.st |= ST_NEW_FRAMEBUFFER;
+      }
+   }
+
+   /* May need to create a new gallium texture:
+    */
+   if (!stObj->pt) {
+      const enum pipe_format fmt =
+         st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat);
+      stObj->pt = st_texture_create(ctx->st,
+                                    gl_target_to_pipe(stObj->base.Target),
+                                    fmt,
+                                    stObj->lastLevel,
+                                    firstImage->base.Width2,
+                                    firstImage->base.Height2,
+                                    firstImage->base.Depth2,
+                                    comp_byte,
+                                    ( (pf_is_depth_stencil(fmt) ?
+                                      PIPE_TEXTURE_USAGE_DEPTH_STENCIL :
+                                      PIPE_TEXTURE_USAGE_RENDER_TARGET) |
+                                      PIPE_TEXTURE_USAGE_SAMPLER ));
+
+      if (!stObj->pt) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+         return GL_FALSE;
+      }
+   }
+
+   /* Pull in any images not in the object's texture:
+    */
+   for (face = 0; face < nr_faces; face++) {
+      GLuint level;
+      for (level = 0; level <= stObj->lastLevel; level++) {
+         struct st_texture_image *stImage =
+            st_texture_image(stObj->base.Image[face][stObj->base.BaseLevel + level]);
+
+         /* Need to import images in main memory or held in other textures.
+          */
+         if (stImage && stObj->pt != stImage->pt) {
+            copy_image_data_to_texture(ctx->st, stObj, level, stImage);
+	    *needFlush = GL_TRUE;
+         }
+      }
+   }
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Returns pointer to a default/dummy texture.
+ * This is typically used when the current shader has tex/sample instructions
+ * but the user has not provided a (any) texture(s).
+ */
+struct gl_texture_object *
+st_get_default_texture(struct st_context *st)
+{
+   if (!st->default_texture) {
+      static const GLenum target = GL_TEXTURE_2D;
+      GLubyte pixels[16][16][4];
+      struct gl_texture_object *texObj;
+      struct gl_texture_image *texImg;
+
+      /* init image to gray */
+      memset(pixels, 127, sizeof(pixels));
+
+      texObj = st->ctx->Driver.NewTextureObject(st->ctx, 0, target);
+
+      texImg = _mesa_get_tex_image(st->ctx, texObj, target, 0);
+
+      _mesa_init_teximage_fields(st->ctx, target, texImg,
+                                 16, 16, 1, 0,  /* w, h, d, border */
+                                 GL_RGBA);
+
+      st_TexImage(st->ctx, 2, target,
+                  0, GL_RGBA,    /* level, intformat */
+                  16, 16, 1, 0,  /* w, h, d, border */
+                  GL_RGBA, GL_UNSIGNED_BYTE, pixels,
+                  &st->ctx->DefaultPacking,
+                  texObj, texImg,
+                  0, 0);
+
+      texObj->MinFilter = GL_NEAREST;
+      texObj->MagFilter = GL_NEAREST;
+      texObj->_Complete = GL_TRUE;
+
+      st->default_texture = texObj;
+   }
+   return st->default_texture;
+}
+
+
+void
+st_init_texture_functions(struct dd_function_table *functions)
+{
+   functions->ChooseTextureFormat = st_ChooseTextureFormat;
+   functions->TexImage1D = st_TexImage1D;
+   functions->TexImage2D = st_TexImage2D;
+   functions->TexImage3D = st_TexImage3D;
+   functions->TexSubImage1D = st_TexSubImage1D;
+   functions->TexSubImage2D = st_TexSubImage2D;
+   functions->TexSubImage3D = st_TexSubImage3D;
+   functions->CopyTexImage1D = st_CopyTexImage1D;
+   functions->CopyTexImage2D = st_CopyTexImage2D;
+   functions->CopyTexSubImage1D = st_CopyTexSubImage1D;
+   functions->CopyTexSubImage2D = st_CopyTexSubImage2D;
+   functions->CopyTexSubImage3D = st_CopyTexSubImage3D;
+   functions->GenerateMipmap = st_generate_mipmap;
+
+   functions->GetTexImage = st_GetTexImage;
+
+   /* compressed texture functions */
+   functions->CompressedTexImage2D = st_CompressedTexImage2D;
+   functions->GetCompressedTexImage = st_GetCompressedTexImage;
+   functions->CompressedTextureSize = _mesa_compressed_texture_size;
+
+   functions->NewTextureObject = st_NewTextureObject;
+   functions->NewTextureImage = st_NewTextureImage;
+   functions->DeleteTexture = st_DeleteTextureObject;
+   functions->FreeTexImageData = st_FreeTextureImageData;
+   functions->UpdateTexturePalette = 0;
+
+   functions->TextureMemCpy = do_memcpy;
+
+   /* XXX Temporary until we can query pipe's texture sizes */
+   functions->TestProxyTexImage = _mesa_test_proxy_teximage;
+}
diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h
new file mode 100644
index 0000000000..f1fe0339cd
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_texture.h
@@ -0,0 +1,48 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_CB_TEXTURE_H
+#define ST_CB_TEXTURE_H
+
+
+extern GLboolean
+st_finalize_texture(GLcontext *ctx,
+		    struct pipe_context *pipe, 
+		    struct gl_texture_object *tObj,
+		    GLboolean *needFlush);
+
+
+extern struct gl_texture_object *
+st_get_default_texture(struct st_context *st);
+
+
+extern void
+st_init_texture_functions(struct dd_function_table *functions);
+
+
+#endif /* ST_CB_TEXTURE_H */
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
new file mode 100644
index 0000000000..e584a6ceea
--- /dev/null
+++ b/src/mesa/state_tracker/st_context.c
@@ -0,0 +1,325 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/extensions.h"
+#include "main/matrix.h"
+#include "main/buffers.h"
+#include "main/scissor.h"
+#include "vbo/vbo.h"
+#include "shader/shader_api.h"
+#include "glapi/glapi.h"
+#include "st_public.h"
+#include "st_context.h"
+#include "st_cb_accum.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_blit.h"
+#include "st_cb_bufferobjects.h"
+#include "st_cb_clear.h"
+#if FEATURE_drawpix
+#include "st_cb_drawpixels.h"
+#include "st_cb_rasterpos.h"
+#endif
+#ifdef FEATURE_OES_draw_texture
+#include "st_cb_drawtex.h"
+#endif
+#include "st_cb_fbo.h"
+#include "st_cb_get.h"
+#if FEATURE_feedback
+#include "st_cb_feedback.h"
+#endif
+#include "st_cb_program.h"
+#include "st_cb_queryobj.h"
+#include "st_cb_readpixels.h"
+#include "st_cb_texture.h"
+#include "st_cb_flush.h"
+#include "st_cb_strings.h"
+#include "st_atom.h"
+#include "st_draw.h"
+#include "st_extensions.h"
+#include "st_gen_mipmap.h"
+#include "st_program.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+#include "draw/draw_context.h"
+#include "cso_cache/cso_cache.h"
+#include "cso_cache/cso_context.h"
+
+
+/**
+ * Called via ctx->Driver.UpdateState()
+ */
+void st_invalidate_state(GLcontext * ctx, GLuint new_state)
+{
+   struct st_context *st = st_context(ctx);
+
+   st->dirty.mesa |= new_state;
+   st->dirty.st |= ST_NEW_MESA;
+
+   /* This is the only core Mesa module we depend upon.
+    * No longer use swrast, swsetup, tnl.
+    */
+   _vbo_InvalidateState(ctx, new_state);
+}
+
+
+/**
+ * Check for multisample env var override.
+ */
+int
+st_get_msaa(void)
+{
+   const char *msaa = _mesa_getenv("__GL_FSAA_MODE");
+   if (msaa)
+      return atoi(msaa);
+   return 0;
+}
+
+
+static struct st_context *
+st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe )
+{
+   uint i;
+   struct st_context *st = CALLOC_STRUCT( st_context );
+   
+   ctx->st = st;
+
+   st->ctx = ctx;
+   st->pipe = pipe;
+
+   /* state tracker needs the VBO module */
+   _vbo_CreateContext(ctx);
+
+#if FEATURE_feedback || FEATURE_drawpix
+   st->draw = draw_create(); /* for selection/feedback */
+
+   /* Disable draw options that might convert points/lines to tris, etc.
+    * as that would foul-up feedback/selection mode.
+    */
+   draw_wide_line_threshold(st->draw, 1000.0f);
+   draw_wide_point_threshold(st->draw, 1000.0f);
+   draw_enable_line_stipple(st->draw, FALSE);
+   draw_enable_point_sprites(st->draw, FALSE);
+#endif
+
+   st->dirty.mesa = ~0;
+   st->dirty.st = ~0;
+
+   st->cso_context = cso_create_context(pipe);
+
+   st_init_atoms( st );
+   st_init_bitmap(st);
+   st_init_clear(st);
+   st_init_draw( st );
+   st_init_generate_mipmap(st);
+   st_init_blit(st);
+
+   for (i = 0; i < PIPE_MAX_SAMPLERS; i++)
+      st->state.sampler_list[i] = &st->state.samplers[i];
+
+   /* we want all vertex data to be placed in buffer objects */
+   vbo_use_buffer_objects(ctx);
+
+   /* Need these flags:
+    */
+   st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
+
+   st->ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
+
+   st->pixel_xfer.cache = _mesa_new_program_cache();
+
+   st->force_msaa = st_get_msaa();
+
+   /* GL limits and extensions */
+   st_init_limits(st);
+   st_init_extensions(st);
+
+   return st;
+}
+
+
+struct st_context *st_create_context(struct pipe_context *pipe,
+                                     const __GLcontextModes *visual,
+                                     struct st_context *share)
+{
+   GLcontext *ctx;
+   GLcontext *shareCtx = share ? share->ctx : NULL;
+   struct dd_function_table funcs;
+
+   memset(&funcs, 0, sizeof(funcs));
+   st_init_driver_functions(&funcs);
+
+   ctx = _mesa_create_context(visual, shareCtx, &funcs, NULL);
+
+   return st_create_context_priv(ctx, pipe);
+}
+
+
+static void st_destroy_context_priv( struct st_context *st )
+{
+   uint i;
+
+#if FEATURE_feedback || FEATURE_drawpix
+   draw_destroy(st->draw);
+#endif
+   st_destroy_atoms( st );
+   st_destroy_draw( st );
+   st_destroy_generate_mipmap(st);
+#if FEATURE_EXT_framebuffer_blit
+   st_destroy_blit(st);
+#endif
+   st_destroy_clear(st);
+#if FEATURE_drawpix
+   st_destroy_bitmap(st);
+   st_destroy_drawpix(st);
+#endif
+#ifdef FEATURE_OES_draw_texture
+   st_destroy_drawtex(st);
+#endif
+
+   for (i = 0; i < Elements(st->state.sampler_texture); i++) {
+      pipe_texture_reference(&st->state.sampler_texture[i], NULL);
+   }
+
+   for (i = 0; i < Elements(st->state.constants); i++) {
+      if (st->state.constants[i].buffer) {
+         pipe_buffer_reference(st->pipe->screen, &st->state.constants[i].buffer, NULL);
+      }
+   }
+
+   if (st->default_texture) {
+      st->ctx->Driver.DeleteTexture(st->ctx, st->default_texture);
+      st->default_texture = NULL;
+   }
+
+   free( st );
+}
+
+ 
+void st_destroy_context( struct st_context *st )
+{
+   struct pipe_context *pipe = st->pipe;
+   struct cso_context *cso = st->cso_context;
+   GLcontext *ctx = st->ctx;
+
+   /* need to unbind and destroy CSO objects before anything else */
+   cso_release_all(st->cso_context);
+
+   st_reference_fragprog(st, &st->fp, NULL);
+   st_reference_vertprog(st, &st->vp, NULL);
+
+   _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache);
+
+   _vbo_DestroyContext(st->ctx);
+
+   _mesa_free_context_data(ctx);
+
+   st_destroy_context_priv(st);
+
+   cso_destroy_context(cso);
+
+   pipe->destroy( pipe );
+
+   free(ctx);
+}
+
+
+void st_make_current(struct st_context *st,
+                     struct st_framebuffer *draw,
+                     struct st_framebuffer *read)
+{
+   if (st) {
+      GLboolean firstTime = st->ctx->FirstTimeCurrent;
+      _mesa_make_current(st->ctx, &draw->Base, &read->Base);
+      /* Need to initialize viewport here since draw->Base->Width/Height
+       * will still be zero at this point.
+       * This could be improved, but would require rather extensive work
+       * elsewhere (allocate rb surface storage sooner)
+       */
+      if (firstTime) {
+         GLuint w = draw->InitWidth, h = draw->InitHeight;
+         _mesa_set_viewport(st->ctx, 0, 0, w, h);
+         _mesa_set_scissor(st->ctx, 0, 0, w, h);
+
+      }
+   }
+   else {
+      _mesa_make_current(NULL, NULL, NULL);
+   }
+}
+
+
+void st_copy_context_state(struct st_context *dst,
+                           struct st_context *src,
+                           uint mask)
+{
+   _mesa_copy_context(dst->ctx, src->ctx, mask);
+}
+
+
+
+st_proc st_get_proc_address(const char *procname)
+{
+   return (st_proc) _glapi_get_proc_address(procname);
+}
+
+
+
+void st_init_driver_functions(struct dd_function_table *functions)
+{
+   _mesa_init_glsl_driver_functions(functions);
+
+#if FEATURE_accum
+   st_init_accum_functions(functions);
+#endif
+#if FEATURE_EXT_framebuffer_blit
+   st_init_blit_functions(functions);
+#endif
+   st_init_bufferobject_functions(functions);
+   st_init_clear_functions(functions);
+#if FEATURE_drawpix
+   st_init_bitmap_functions(functions);
+   st_init_drawpixels_functions(functions);
+   st_init_rasterpos_functions(functions);
+#endif
+   st_init_fbo_functions(functions);
+   st_init_get_functions(functions);
+#if FEATURE_feedback
+   st_init_feedback_functions(functions);
+#endif
+   st_init_program_functions(functions);
+#if FEATURE_ARB_occlusion_query
+   st_init_query_functions(functions);
+#endif
+   st_init_readpixels_functions(functions);
+   st_init_texture_functions(functions);
+   st_init_flush_functions(functions);
+   st_init_string_functions(functions);
+
+   functions->UpdateState = st_invalidate_state;
+}
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
new file mode 100644
index 0000000000..695ac4a96f
--- /dev/null
+++ b/src/mesa/state_tracker/st_context.h
@@ -0,0 +1,249 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef ST_CONTEXT_H
+#define ST_CONTEXT_H
+
+#include "main/mtypes.h"
+#include "shader/prog_cache.h"
+#include "pipe/p_state.h"
+
+
+struct st_context;
+struct st_texture_object;
+struct st_fragment_program;
+struct draw_context;
+struct draw_stage;
+struct cso_cache;
+struct cso_blend;
+struct gen_mipmap_state;
+struct blit_state;
+struct bitmap_cache;
+
+
+#define FRONT_STATUS_UNDEFINED    0
+#define FRONT_STATUS_DIRTY        1
+#define FRONT_STATUS_COPY_OF_BACK 2
+
+
+#define ST_NEW_MESA                    0x1 /* Mesa state has changed */
+#define ST_NEW_FRAGMENT_PROGRAM        0x2
+#define ST_NEW_VERTEX_PROGRAM          0x4
+#define ST_NEW_FRAMEBUFFER             0x8
+
+
+struct st_state_flags {
+   GLuint mesa;
+   GLuint st;
+};
+
+struct st_tracked_state {
+   const char *name;
+   struct st_state_flags dirty;
+   void (*update)( struct st_context *st );
+};
+
+
+
+struct st_context
+{
+   GLcontext *ctx;
+
+   struct pipe_context *pipe;
+
+   struct draw_context *draw;  /**< For selection/feedback/rastpos only */
+   struct draw_stage *feedback_stage;  /**< For GL_FEEDBACK rendermode */
+   struct draw_stage *selection_stage;  /**< For GL_SELECT rendermode */
+   struct draw_stage *rastpos_stage;  /**< For glRasterPos */
+
+   /* Some state is contained in constant objects.
+    * Other state is just parameter values.
+    */
+   struct {
+      struct pipe_blend_state               blend;
+      struct pipe_depth_stencil_alpha_state depth_stencil;
+      struct pipe_rasterizer_state          rasterizer;
+      struct pipe_sampler_state             samplers[PIPE_MAX_SAMPLERS];
+      struct pipe_sampler_state             *sampler_list[PIPE_MAX_SAMPLERS];
+      struct pipe_clip_state clip;
+      struct pipe_constant_buffer constants[2];
+      struct pipe_framebuffer_state framebuffer;
+      struct pipe_texture *sampler_texture[PIPE_MAX_SAMPLERS];
+      struct pipe_poly_stipple poly_stipple;
+      struct pipe_scissor_state scissor;
+      struct pipe_viewport_state viewport;
+
+      GLuint num_samplers;
+      GLuint num_textures;
+   } state;
+
+   struct {
+      struct st_tracked_state tracked_state[PIPE_SHADER_TYPES];
+   } constants;
+
+   /* XXX unused: */
+   struct {
+      struct gl_fragment_program *fragment_program;
+   } cb;
+
+   GLuint frontbuffer_status;  /**< one of FRONT_STATUS_ */
+
+   char vendor[100];
+   char renderer[100];
+
+   /* State to be validated:
+    */
+   struct st_tracked_state **atoms;
+   GLuint nr_atoms;
+
+   struct st_state_flags dirty;
+
+   GLboolean missing_textures;
+
+   GLfloat polygon_offset_scale; /* ?? */
+
+   /** Mapping from VERT_RESULT_x to post-transformed vertex slot */
+   const GLuint *vertex_result_to_slot;
+
+   struct st_vertex_program *vp;    /**< Currently bound vertex program */
+   struct st_fragment_program *fp;  /**< Currently bound fragment program */
+
+   struct gl_texture_object *default_texture;
+
+   struct {
+      struct gl_program_cache *cache;
+      struct st_fragment_program *program;  /**< cur pixel transfer prog */
+      GLuint xfer_prog_sn;  /**< pixel xfer program serial no. */
+      GLuint user_prog_sn;  /**< user fragment program serial no. */
+      struct st_fragment_program *combined_prog;
+      GLuint combined_prog_sn;
+      struct pipe_texture *pixelmap_texture;
+      boolean pixelmap_enabled;  /**< use the pixelmap texture? */
+   } pixel_xfer;
+
+   /** for glBitmap */
+   struct {
+      struct pipe_rasterizer_state rasterizer;
+      struct pipe_sampler_state sampler;
+      struct pipe_shader_state vert_shader;
+      enum pipe_format tex_format;
+      void *vs;
+      float vertices[4][3][4];  /**< vertex pos + color + texcoord */
+      struct pipe_buffer *vbuf;
+      unsigned vbuf_slot;       /* next free slot in vbuf */
+      struct bitmap_cache *cache;
+   } bitmap;
+
+   /** for glDraw/CopyPixels */
+   struct {
+      struct st_fragment_program *z_shader;
+      struct st_vertex_program *vert_shaders[2];
+   } drawpix;
+
+   /** for glClear */
+   struct {
+      struct pipe_shader_state vert_shader;
+      struct pipe_shader_state frag_shader;
+      struct pipe_rasterizer_state raster;
+      struct pipe_viewport_state viewport;
+      void *vs;
+      void *fs;
+      float vertices[4][2][4];  /**< vertex pos + color */
+      struct pipe_buffer *vbuf;
+      unsigned vbuf_slot;
+   } clear;
+
+   void *passthrough_fs;  /**< simple pass-through frag shader */
+
+   struct gen_mipmap_state *gen_mipmap;
+   struct blit_state *blit;
+
+   struct cso_context *cso_context;
+
+   int force_msaa;
+};
+
+
+/* Need this so that we can implement Mesa callbacks in this module.
+ */
+static INLINE struct st_context *st_context(GLcontext *ctx)
+{
+   return ctx->st;
+}
+
+
+/**
+ * Wrapper for GLframebuffer.
+ * This is an opaque type to the outside world.
+ */
+struct st_framebuffer
+{
+   GLframebuffer Base;
+   void *Private;
+   GLuint InitWidth, InitHeight;
+};
+
+
+extern void st_init_driver_functions(struct dd_function_table *functions);
+
+void st_invalidate_state(GLcontext * ctx, GLuint new_state);
+
+
+
+#define Y_0_TOP 1
+#define Y_0_BOTTOM 2
+
+static INLINE GLuint
+st_fb_orientation(const struct gl_framebuffer *fb)
+{
+   if (fb && fb->Name == 0) {
+      /* Drawing into a window (on-screen buffer).
+       *
+       * Negate Y scale to flip image vertically.
+       * The NDC Y coords prior to viewport transformation are in the range
+       * [y=-1=bottom, y=1=top]
+       * Hardware window coords are in the range [y=0=top, y=H-1=bottom] where
+       * H is the window height.
+       * Use the viewport transformation to invert Y.
+       */
+      return Y_0_TOP;
+   }
+   else {
+      /* Drawing into user-created FBO (very likely a texture).
+       *
+       * For textures, T=0=Bottom, so by extension Y=0=Bottom for rendering.
+       */
+      return Y_0_BOTTOM;
+   }
+}
+
+
+extern int
+st_get_msaa(void);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c
new file mode 100644
index 0000000000..c7d26ce33c
--- /dev/null
+++ b/src/mesa/state_tracker/st_debug.c
@@ -0,0 +1,70 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/context.h"
+#include "shader/prog_print.h"
+
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "cso_cache/cso_cache.h"
+
+#include "st_context.h"
+#include "st_debug.h"
+#include "st_program.h"
+
+
+
+/**
+ * Print current state.  May be called from inside gdb to see currently
+ * bound vertex/fragment shaders and associated constants.
+ */
+void
+st_print_current(void)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct st_context *st = ctx->st;
+
+#if 0
+   int i;
+
+   printf("Vertex Transform Inputs:\n");
+   for (i = 0; i < st->vp->state.num_inputs; i++) {
+      printf("  Slot %d:  VERT_ATTRIB_%d\n", i, st->vp->index_to_input[i]);
+   }
+#endif
+
+   tgsi_dump( st->vp->state.tokens, 0 );
+   if (st->vp->Base.Base.Parameters)
+      _mesa_print_parameter_list(st->vp->Base.Base.Parameters);
+
+   tgsi_dump( st->fp->state.tokens, 0 );
+   if (st->fp->Base.Base.Parameters)
+      _mesa_print_parameter_list(st->fp->Base.Base.Parameters);
+}
diff --git a/src/mesa/state_tracker/st_debug.h b/src/mesa/state_tracker/st_debug.h
new file mode 100644
index 0000000000..49d752e1b2
--- /dev/null
+++ b/src/mesa/state_tracker/st_debug.h
@@ -0,0 +1,36 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_DEBUG_H
+#define ST_DEBUG_H
+
+extern void
+st_print_current(void);
+
+
+#endif /* ST_DEBUG_H */
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
new file mode 100644
index 0000000000..ae71d586c2
--- /dev/null
+++ b/src/mesa/state_tracker/st_draw.c
@@ -0,0 +1,699 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "shader/prog_uniform.h"
+
+#include "vbo/vbo.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_bufferobjects.h"
+#include "st_draw.h"
+#include "st_program.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+
+static GLuint double_types[4] = {
+   PIPE_FORMAT_R64_FLOAT,
+   PIPE_FORMAT_R64G64_FLOAT,
+   PIPE_FORMAT_R64G64B64_FLOAT,
+   PIPE_FORMAT_R64G64B64A64_FLOAT
+};
+
+static GLuint float_types[4] = {
+   PIPE_FORMAT_R32_FLOAT,
+   PIPE_FORMAT_R32G32_FLOAT,
+   PIPE_FORMAT_R32G32B32_FLOAT,
+   PIPE_FORMAT_R32G32B32A32_FLOAT
+};
+
+static GLuint uint_types_norm[4] = {
+   PIPE_FORMAT_R32_UNORM,
+   PIPE_FORMAT_R32G32_UNORM,
+   PIPE_FORMAT_R32G32B32_UNORM,
+   PIPE_FORMAT_R32G32B32A32_UNORM
+};
+
+static GLuint uint_types_scale[4] = {
+   PIPE_FORMAT_R32_USCALED,
+   PIPE_FORMAT_R32G32_USCALED,
+   PIPE_FORMAT_R32G32B32_USCALED,
+   PIPE_FORMAT_R32G32B32A32_USCALED
+};
+
+static GLuint int_types_norm[4] = {
+   PIPE_FORMAT_R32_SNORM,
+   PIPE_FORMAT_R32G32_SNORM,
+   PIPE_FORMAT_R32G32B32_SNORM,
+   PIPE_FORMAT_R32G32B32A32_SNORM
+};
+
+static GLuint int_types_scale[4] = {
+   PIPE_FORMAT_R32_SSCALED,
+   PIPE_FORMAT_R32G32_SSCALED,
+   PIPE_FORMAT_R32G32B32_SSCALED,
+   PIPE_FORMAT_R32G32B32A32_SSCALED
+};
+
+static GLuint ushort_types_norm[4] = {
+   PIPE_FORMAT_R16_UNORM,
+   PIPE_FORMAT_R16G16_UNORM,
+   PIPE_FORMAT_R16G16B16_UNORM,
+   PIPE_FORMAT_R16G16B16A16_UNORM
+};
+
+static GLuint ushort_types_scale[4] = {
+   PIPE_FORMAT_R16_USCALED,
+   PIPE_FORMAT_R16G16_USCALED,
+   PIPE_FORMAT_R16G16B16_USCALED,
+   PIPE_FORMAT_R16G16B16A16_USCALED
+};
+
+static GLuint short_types_norm[4] = {
+   PIPE_FORMAT_R16_SNORM,
+   PIPE_FORMAT_R16G16_SNORM,
+   PIPE_FORMAT_R16G16B16_SNORM,
+   PIPE_FORMAT_R16G16B16A16_SNORM
+};
+
+static GLuint short_types_scale[4] = {
+   PIPE_FORMAT_R16_SSCALED,
+   PIPE_FORMAT_R16G16_SSCALED,
+   PIPE_FORMAT_R16G16B16_SSCALED,
+   PIPE_FORMAT_R16G16B16A16_SSCALED
+};
+
+static GLuint ubyte_types_norm[4] = {
+   PIPE_FORMAT_R8_UNORM,
+   PIPE_FORMAT_R8G8_UNORM,
+   PIPE_FORMAT_R8G8B8_UNORM,
+   PIPE_FORMAT_R8G8B8A8_UNORM
+};
+
+static GLuint ubyte_types_scale[4] = {
+   PIPE_FORMAT_R8_USCALED,
+   PIPE_FORMAT_R8G8_USCALED,
+   PIPE_FORMAT_R8G8B8_USCALED,
+   PIPE_FORMAT_R8G8B8A8_USCALED
+};
+
+static GLuint byte_types_norm[4] = {
+   PIPE_FORMAT_R8_SNORM,
+   PIPE_FORMAT_R8G8_SNORM,
+   PIPE_FORMAT_R8G8B8_SNORM,
+   PIPE_FORMAT_R8G8B8A8_SNORM
+};
+
+static GLuint byte_types_scale[4] = {
+   PIPE_FORMAT_R8_SSCALED,
+   PIPE_FORMAT_R8G8_SSCALED,
+   PIPE_FORMAT_R8G8B8_SSCALED,
+   PIPE_FORMAT_R8G8B8A8_SSCALED
+};
+
+static GLuint fixed_types[4] = {
+   PIPE_FORMAT_R32_FIXED,
+   PIPE_FORMAT_R32G32_FIXED,
+   PIPE_FORMAT_R32G32B32_FIXED,
+   PIPE_FORMAT_R32G32B32A32_FIXED
+};
+
+
+
+/**
+ * Return a PIPE_FORMAT_x for the given GL datatype and size.
+ */
+GLuint
+st_pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized)
+{
+   assert((type >= GL_BYTE && type <= GL_DOUBLE) ||
+          type == GL_FIXED);
+   assert(size >= 1);
+   assert(size <= 4);
+
+   if (normalized) {
+      switch (type) {
+      case GL_DOUBLE: return double_types[size-1];
+      case GL_FLOAT: return float_types[size-1];
+      case GL_INT: return int_types_norm[size-1];
+      case GL_SHORT: return short_types_norm[size-1];
+      case GL_BYTE: return byte_types_norm[size-1];
+      case GL_UNSIGNED_INT: return uint_types_norm[size-1];
+      case GL_UNSIGNED_SHORT: return ushort_types_norm[size-1];
+      case GL_UNSIGNED_BYTE: return ubyte_types_norm[size-1];
+      case GL_FIXED: return fixed_types[size-1];
+      default: assert(0); return 0;
+      }      
+   }
+   else {
+      switch (type) {
+      case GL_DOUBLE: return double_types[size-1];
+      case GL_FLOAT: return float_types[size-1];
+      case GL_INT: return int_types_scale[size-1];
+      case GL_SHORT: return short_types_scale[size-1];
+      case GL_BYTE: return byte_types_scale[size-1];
+      case GL_UNSIGNED_INT: return uint_types_scale[size-1];
+      case GL_UNSIGNED_SHORT: return ushort_types_scale[size-1];
+      case GL_UNSIGNED_BYTE: return ubyte_types_scale[size-1];
+      case GL_FIXED: return fixed_types[size-1];
+      default: assert(0); return 0;
+      }      
+   }
+   return 0; /* silence compiler warning */
+}
+
+
+/*
+ * If edge flags are needed, setup an bitvector of flags and call
+ * pipe->set_edgeflags().
+ * XXX memleak: need to free the returned pointer at some point
+ */
+static void *
+setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count,
+                const struct gl_client_array *array)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+
+   if ((primMode == GL_TRIANGLES ||
+        primMode == GL_QUADS ||
+        primMode == GL_POLYGON) &&
+       (ctx->Polygon.FrontMode != GL_FILL ||
+        ctx->Polygon.BackMode != GL_FILL)) {
+      /* need edge flags */
+      GLint i;
+      unsigned *vec;
+      struct st_buffer_object *stobj = st_buffer_object(array->BufferObj);
+      ubyte *map;
+
+      if (!stobj)
+         return NULL;
+
+      vec = (unsigned *) calloc(sizeof(unsigned), (count + 31) / 32);
+      if (!vec)
+         return NULL;
+
+      map = pipe_buffer_map(pipe->screen, stobj->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+      map = ADD_POINTERS(map, array->Ptr);
+
+      for (i = 0; i < count; i++) {
+         if (*((float *) map))
+            vec[i/32] |= 1 << (i % 32);
+
+         map += array->StrideB;
+      }
+
+      pipe_buffer_unmap(pipe->screen, stobj->buffer);
+
+      pipe->set_edgeflags(pipe, vec);
+
+      return vec;
+   }
+   else {
+      /* edge flags not needed */
+      pipe->set_edgeflags(pipe, NULL);
+      return NULL;
+   }
+}
+
+
+/**
+ * Examine the active arrays to determine if we have interleaved
+ * vertex arrays all living in one VBO, or all living in user space.
+ * \param userSpace  returns whether the arrays are in user space.
+ */
+static GLboolean
+is_interleaved_arrays(const struct st_vertex_program *vp,
+                      const struct gl_client_array **arrays,
+                      GLboolean *userSpace)
+{
+   GLuint attr;
+   const struct gl_buffer_object *firstBufObj = NULL;
+   GLint firstStride = -1;
+   GLuint num_client_arrays = 0;
+   const GLubyte *client_addr = NULL;
+
+   for (attr = 0; attr < vp->num_inputs; attr++) {
+      const GLuint mesaAttr = vp->index_to_input[attr];
+      const struct gl_buffer_object *bufObj = arrays[mesaAttr]->BufferObj;
+      const GLsizei stride = arrays[mesaAttr]->StrideB; /* in bytes */
+
+      if (firstStride < 0) {
+         firstStride = stride;
+      }
+      else if (firstStride != stride) {
+         return GL_FALSE;
+      }
+         
+      if (!bufObj || !bufObj->Name) {
+         num_client_arrays++;
+         /* Try to detect if the client-space arrays are
+          * "close" to each other.
+          */
+         if (!client_addr) {
+            client_addr = arrays[mesaAttr]->Ptr;
+         }
+         else if (abs(arrays[mesaAttr]->Ptr - client_addr) > firstStride) {
+            /* arrays start too far apart */
+            return GL_FALSE;
+         }
+      }
+      else if (!firstBufObj) {
+         firstBufObj = bufObj;
+      }
+      else if (bufObj != firstBufObj) {
+         return GL_FALSE;
+      }
+   }
+
+   *userSpace = (num_client_arrays == vp->num_inputs);
+   /* printf("user space: %d (%d %d)\n", (int) *userSpace,num_client_arrays,vp->num_inputs); */
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Compute the memory range occupied by the arrays.
+ */
+static void
+get_arrays_bounds(const struct st_vertex_program *vp,
+                       const struct gl_client_array **arrays,
+                       GLuint max_index,
+                       const GLubyte **low, const GLubyte **high)
+{
+   const GLubyte *low_addr = NULL;
+   GLuint attr;
+   GLint stride;
+
+   for (attr = 0; attr < vp->num_inputs; attr++) {
+      const GLuint mesaAttr = vp->index_to_input[attr];
+      const GLubyte *start = arrays[mesaAttr]->Ptr;
+      stride = arrays[mesaAttr]->StrideB;
+      if (attr == 0) {
+         low_addr = start;
+      }
+      else {
+         low_addr = MIN2(low_addr, start);
+      }
+   }
+
+   *low = low_addr;
+   *high = low_addr + (max_index + 1) * stride;
+}
+
+
+/**
+ * Set up for drawing interleaved arrays that all live in one VBO
+ * or all live in user space.
+ * \param vbuffer  returns vertex buffer info
+ * \param velements  returns vertex element info
+ */
+static void
+setup_interleaved_attribs(GLcontext *ctx,
+                          const struct st_vertex_program *vp,
+                          const struct gl_client_array **arrays,
+                          GLuint max_index,
+                          GLboolean userSpace,
+                          struct pipe_vertex_buffer *vbuffer,
+                          struct pipe_vertex_element velements[])
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   GLuint attr;
+   const GLubyte *offset0;
+
+   for (attr = 0; attr < vp->num_inputs; attr++) {
+      const GLuint mesaAttr = vp->index_to_input[attr];
+      struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
+      struct st_buffer_object *stobj = st_buffer_object(bufobj);
+      GLsizei stride = arrays[mesaAttr]->StrideB;
+
+      /*printf("stobj %u = %p\n", attr, (void*)stobj);*/
+
+      if (attr == 0) {
+         const GLubyte *low, *high;
+
+         get_arrays_bounds(vp, arrays, max_index, &low, &high);
+         /*printf("buffer range: %p %p  %d\n", low, high, high-low);*/
+
+         offset0 = low;
+         if (userSpace) {
+            vbuffer->buffer =
+               pipe_user_buffer_create(pipe->screen, (void *) low, high - low);
+            vbuffer->buffer_offset = 0;
+         }
+         else {
+            vbuffer->buffer = NULL;
+            pipe_buffer_reference(pipe->screen, &vbuffer->buffer, stobj->buffer);
+            vbuffer->buffer_offset = (unsigned) low;
+         }
+         vbuffer->stride = stride; /* in bytes */
+         vbuffer->max_index = max_index;
+      }
+
+      velements[attr].src_offset =
+         (unsigned) (arrays[mesaAttr]->Ptr - offset0);
+      velements[attr].vertex_buffer_index = 0;
+      velements[attr].nr_components = arrays[mesaAttr]->Size;
+      velements[attr].src_format =
+         st_pipe_vertex_format(arrays[mesaAttr]->Type,
+                               arrays[mesaAttr]->Size,
+                               arrays[mesaAttr]->Normalized);
+      assert(velements[attr].src_format);
+   }
+}
+
+
+/**
+ * Set up a separate pipe_vertex_buffer and pipe_vertex_element for each
+ * vertex attribute.
+ * \param vbuffer  returns vertex buffer info
+ * \param velements  returns vertex element info
+ */
+static void
+setup_non_interleaved_attribs(GLcontext *ctx,
+                              const struct st_vertex_program *vp,
+                              const struct gl_client_array **arrays,
+                              GLuint max_index,
+                              GLboolean *userSpace,
+                              struct pipe_vertex_buffer vbuffer[],
+                              struct pipe_vertex_element velements[])
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   GLuint attr;
+
+   for (attr = 0; attr < vp->num_inputs; attr++) {
+      const GLuint mesaAttr = vp->index_to_input[attr];
+      struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
+      GLsizei stride = arrays[mesaAttr]->StrideB;
+
+      *userSpace = GL_FALSE;
+
+      if (bufobj && bufobj->Name) {
+         /* Attribute data is in a VBO.
+          * Recall that for VBOs, the gl_client_array->Ptr field is
+          * really an offset from the start of the VBO, not a pointer.
+          */
+         struct st_buffer_object *stobj = st_buffer_object(bufobj);
+         assert(stobj->buffer);
+         /*printf("stobj %u = %p\n", attr, (void*) stobj);*/
+
+         vbuffer[attr].buffer = NULL;
+         pipe_buffer_reference(pipe->screen, &vbuffer[attr].buffer, stobj->buffer);
+         vbuffer[attr].buffer_offset = (unsigned) arrays[mesaAttr]->Ptr;
+         velements[attr].src_offset = 0;
+      }
+      else {
+         /* attribute data is in user-space memory, not a VBO */
+         uint bytes;
+         /*printf("user-space array %d stride %d\n", attr, stride);*/
+	
+         *userSpace = GL_TRUE;
+
+         /* wrap user data */
+         if (arrays[mesaAttr]->Ptr) {
+            /* user's vertex array */
+            if (arrays[mesaAttr]->StrideB) {
+               bytes = arrays[mesaAttr]->StrideB * (max_index + 1);
+            }
+            else {
+               bytes = arrays[mesaAttr]->Size
+                  * _mesa_sizeof_type(arrays[mesaAttr]->Type);
+            }
+            vbuffer[attr].buffer = pipe_user_buffer_create(pipe->screen,
+                           (void *) arrays[mesaAttr]->Ptr, bytes);
+         }
+         else {
+            /* no array, use ctx->Current.Attrib[] value */
+            bytes = sizeof(ctx->Current.Attrib[0]);
+            vbuffer[attr].buffer = pipe_user_buffer_create(pipe->screen,
+                           (void *) ctx->Current.Attrib[mesaAttr], bytes);
+            stride = 0;
+         }
+
+         vbuffer[attr].buffer_offset = 0;
+         velements[attr].src_offset = 0;
+      }
+
+      assert(velements[attr].src_offset <= 2048); /* 11-bit field */
+
+      /* common-case setup */
+      vbuffer[attr].stride = stride; /* in bytes */
+      vbuffer[attr].max_index = max_index;
+      velements[attr].vertex_buffer_index = attr;
+      velements[attr].nr_components = arrays[mesaAttr]->Size;
+      velements[attr].src_format
+         = st_pipe_vertex_format(arrays[mesaAttr]->Type,
+                                 arrays[mesaAttr]->Size,
+                                 arrays[mesaAttr]->Normalized);
+      assert(velements[attr].src_format);
+   }
+}
+
+
+
+/**
+ * Prior to drawing, check that any uniforms referenced by the
+ * current shader have been set.  If a uniform has not been set,
+ * issue a warning.
+ */
+static void
+check_uniforms(GLcontext *ctx)
+{
+   const struct gl_shader_program *shProg = ctx->Shader.CurrentProgram;
+   if (shProg && shProg->LinkStatus) {
+      GLuint i;
+      for (i = 0; i < shProg->Uniforms->NumUniforms; i++) {
+         const struct gl_uniform *u = &shProg->Uniforms->Uniforms[i];
+         if (!u->Initialized) {
+            _mesa_warning(ctx,
+                          "Using shader with uninitialized uniform: %s",
+                          u->Name);
+         }
+      }
+   }
+}
+
+
+/**
+ * This function gets plugged into the VBO module and is called when
+ * we have something to render.
+ * Basically, translate the information into the format expected by gallium.
+ */
+void
+st_draw_vbo(GLcontext *ctx,
+            const struct gl_client_array **arrays,
+            const struct _mesa_prim *prims,
+            GLuint nr_prims,
+            const struct _mesa_index_buffer *ib,
+            GLuint min_index,
+            GLuint max_index)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   const struct st_vertex_program *vp;
+   const struct pipe_shader_state *vs;
+   struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS];
+   GLuint attr;
+   struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS];
+   unsigned num_vbuffers, num_velements;
+   GLboolean userSpace;
+
+   /* sanity check for pointer arithmetic below */
+   assert(sizeof(arrays[0]->Ptr[0]) == 1);
+
+   st_validate_state(ctx->st);
+
+   /* must get these after state validation! */
+   vp = ctx->st->vp;
+   vs = &ctx->st->vp->state;
+
+#if 0
+   if (MESA_VERBOSE & VERBOSE_GLSL) {
+      check_uniforms(ctx);
+   }
+#endif
+
+   /*
+    * Setup the vbuffer[] and velements[] arrays.
+    */
+   if (is_interleaved_arrays(vp, arrays, &userSpace)) {
+      /*printf("Draw interleaved\n");*/
+      setup_interleaved_attribs(ctx, vp, arrays, max_index, userSpace,
+                                vbuffer, velements);
+      num_vbuffers = 1;
+      num_velements = vp->num_inputs;
+      if (num_velements == 0)
+         num_vbuffers = 0;
+   }
+   else {
+      /*printf("Draw non-interleaved\n");*/
+      setup_non_interleaved_attribs(ctx, vp, arrays, max_index,
+                                    &userSpace, vbuffer, velements);
+      num_vbuffers = vp->num_inputs;
+      num_velements = vp->num_inputs;
+   }
+
+#if 0
+   {
+      GLuint i;
+      for (i = 0; i < num_vbuffers; i++) {
+         printf("buffers[%d].stride = %u\n", i, vbuffer[i].stride);
+         printf("buffers[%d].max_index = %u\n", i, vbuffer[i].max_index);
+         printf("buffers[%d].buffer_offset = %u\n", i, vbuffer[i].buffer_offset);
+         printf("buffers[%d].buffer = %p\n", i, (void*) vbuffer[i].buffer);
+      }
+      for (i = 0; i < num_velements; i++) {
+         printf("vlements[%d].vbuffer_index = %u\n", i, velements[i].vertex_buffer_index);
+         printf("vlements[%d].src_offset = %u\n", i, velements[i].src_offset);
+         printf("vlements[%d].nr_comps = %u\n", i, velements[i].nr_components);
+         printf("vlements[%d].format = %s\n", i, pf_name(velements[i].src_format));
+      }
+   }
+#endif
+
+   pipe->set_vertex_buffers(pipe, num_vbuffers, vbuffer);
+   pipe->set_vertex_elements(pipe, num_velements, velements);
+
+   if (num_vbuffers == 0 || num_velements == 0)
+      return;
+
+   /* do actual drawing */
+   if (ib) {
+      /* indexed primitive */
+      struct gl_buffer_object *bufobj = ib->obj;
+      struct pipe_buffer *indexBuf = NULL;
+      unsigned indexSize, indexOffset, i;
+
+      switch (ib->type) {
+      case GL_UNSIGNED_INT:
+         indexSize = 4;
+         break;
+      case GL_UNSIGNED_SHORT:
+         indexSize = 2;
+         break;
+      case GL_UNSIGNED_BYTE:
+         indexSize = 1;
+         break;
+      default:
+         assert(0);
+	 return;
+      }
+
+      /* get/create the index buffer object */
+      if (bufobj && bufobj->Name) {
+         /* elements/indexes are in a real VBO */
+         struct st_buffer_object *stobj = st_buffer_object(bufobj);
+         pipe_buffer_reference(pipe->screen, &indexBuf, stobj->buffer);
+         indexOffset = (unsigned) ib->ptr / indexSize;
+      }
+      else {
+         /* element/indicies are in user space memory */
+         indexBuf = pipe_user_buffer_create(pipe->screen, (void *) ib->ptr,
+                                            ib->count * indexSize);
+         indexOffset = 0;
+      }
+
+      /* draw */
+      if (nr_prims == 1 && pipe->draw_range_elements != NULL) {
+         i = 0;
+
+         /* XXX: exercise temporary path to pass min/max directly
+          * through to driver & draw module.  These interfaces still
+          * need a bit of work...
+          */
+         setup_edgeflags(ctx, prims[i].mode,
+                         prims[i].start + indexOffset, prims[i].count,
+                         arrays[VERT_ATTRIB_EDGEFLAG]);
+
+         pipe->draw_range_elements(pipe, indexBuf, indexSize,
+                                   min_index,
+                                   max_index,
+                                   prims[i].mode,
+                                   prims[i].start + indexOffset, prims[i].count);
+      }
+      else {
+         for (i = 0; i < nr_prims; i++) {
+            setup_edgeflags(ctx, prims[i].mode,
+                            prims[i].start + indexOffset, prims[i].count,
+                            arrays[VERT_ATTRIB_EDGEFLAG]);
+            
+            pipe->draw_elements(pipe, indexBuf, indexSize,
+                                prims[i].mode,
+                                prims[i].start + indexOffset, prims[i].count);
+         }
+      }
+
+      pipe_buffer_reference(pipe->screen, &indexBuf, NULL);
+   }
+   else {
+      /* non-indexed */
+      GLuint i;
+      for (i = 0; i < nr_prims; i++) {
+         setup_edgeflags(ctx, prims[i].mode,
+                         prims[i].start, prims[i].count,
+                         arrays[VERT_ATTRIB_EDGEFLAG]);
+
+         pipe->draw_arrays(pipe, prims[i].mode, prims[i].start, prims[i].count);
+      }
+   }
+
+   /* unreference buffers (frees wrapped user-space buffer objects) */
+   for (attr = 0; attr < num_vbuffers; attr++) {
+      pipe_buffer_reference(pipe->screen, &vbuffer[attr].buffer, NULL);
+      assert(!vbuffer[attr].buffer);
+   }
+
+   if (userSpace) 
+   {
+      pipe->set_vertex_buffers(pipe, 0, NULL);
+   }
+}
+
+
+void st_init_draw( struct st_context *st )
+{
+   GLcontext *ctx = st->ctx;
+
+   vbo_set_draw_func(ctx, st_draw_vbo);
+}
+
+
+void st_destroy_draw( struct st_context *st )
+{
+}
+
+
diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h
new file mode 100644
index 0000000000..659c3ac834
--- /dev/null
+++ b/src/mesa/state_tracker/st_draw.h
@@ -0,0 +1,67 @@
+/**************************************************************************
+ * 
+ * Copyright 2004 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+    
+
+#ifndef ST_DRAW_H
+#define ST_DRAW_H
+
+struct _mesa_prim;
+struct _mesa_index_buffer;
+
+void st_init_draw( struct st_context *st );
+
+void st_destroy_draw( struct st_context *st );
+
+extern void
+st_draw_vbo(GLcontext *ctx,
+            const struct gl_client_array **arrays,
+            const struct _mesa_prim *prims,
+            GLuint nr_prims,
+            const struct _mesa_index_buffer *ib,
+            GLuint min_index,
+            GLuint max_index);
+
+extern void
+st_feedback_draw_vbo(GLcontext *ctx,
+                     const struct gl_client_array **arrays,
+                     const struct _mesa_prim *prims,
+                     GLuint nr_prims,
+                     const struct _mesa_index_buffer *ib,
+                     GLuint min_index,
+                     GLuint max_index);
+
+/* Internal function:
+ */
+extern GLuint
+st_pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized);
+
+#endif
diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c
new file mode 100644
index 0000000000..5c9c4506c2
--- /dev/null
+++ b/src/mesa/state_tracker/st_draw_feedback.c
@@ -0,0 +1,262 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/macros.h"
+#include "shader/prog_uniform.h"
+
+#include "vbo/vbo.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_cb_bufferobjects.h"
+#include "st_draw.h"
+#include "st_program.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+
+#include "draw/draw_private.h"
+#include "draw/draw_context.h"
+
+
+#if FEATURE_feedback || FEATURE_drawpix
+
+/**
+ * Set the (private) draw module's post-transformed vertex format when in
+ * GL_SELECT or GL_FEEDBACK mode or for glRasterPos.
+ */
+static void
+set_feedback_vertex_format(GLcontext *ctx)
+{
+#if 0
+   struct st_context *st = ctx->st;
+   struct vertex_info vinfo;
+   GLuint i;
+
+   memset(&vinfo, 0, sizeof(vinfo));
+
+   if (ctx->RenderMode == GL_SELECT) {
+      assert(ctx->RenderMode == GL_SELECT);
+      vinfo.num_attribs = 1;
+      vinfo.format[0] = FORMAT_4F;
+      vinfo.interp_mode[0] = INTERP_LINEAR;
+   }
+   else {
+      /* GL_FEEDBACK, or glRasterPos */
+      /* emit all attribs (pos, color, texcoord) as GLfloat[4] */
+      vinfo.num_attribs = st->state.vs->cso->state.num_outputs;
+      for (i = 0; i < vinfo.num_attribs; i++) {
+         vinfo.format[i] = FORMAT_4F;
+         vinfo.interp_mode[i] = INTERP_LINEAR;
+      }
+   }
+
+   draw_set_vertex_info(st->draw, &vinfo);
+#endif
+}
+
+
+/**
+ * Called by VBO to draw arrays when in selection or feedback mode and
+ * to implement glRasterPos.
+ * This is very much like the normal draw_vbo() function above.
+ * Look at code refactoring some day.
+ * Might move this into the failover module some day.
+ */
+void
+st_feedback_draw_vbo(GLcontext *ctx,
+                     const struct gl_client_array **arrays,
+                     const struct _mesa_prim *prims,
+                     GLuint nr_prims,
+                     const struct _mesa_index_buffer *ib,
+                     GLuint min_index,
+                     GLuint max_index)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   struct draw_context *draw = st->draw;
+   const struct st_vertex_program *vp;
+   const struct pipe_shader_state *vs;
+   struct pipe_buffer *index_buffer_handle = 0;
+   struct pipe_vertex_buffer vbuffers[PIPE_MAX_SHADER_INPUTS];
+   struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS];
+   GLuint attr, i;
+   ubyte *mapped_constants;
+
+   assert(draw);
+
+   st_validate_state(ctx->st);
+
+   /* must get these after state validation! */
+   vp = ctx->st->vp;
+   vs = &st->vp->state;
+
+   if (!st->vp->draw_shader) {
+      st->vp->draw_shader = draw_create_vertex_shader(draw, vs);
+   }
+
+   /*
+    * Set up the draw module's state.
+    *
+    * We'd like to do this less frequently, but the normal state-update
+    * code sends state updates to the pipe, not to our private draw module.
+    */
+   assert(draw);
+   draw_set_viewport_state(draw, &st->state.viewport);
+   draw_set_clip_state(draw, &st->state.clip);
+   draw_set_rasterizer_state(draw, &st->state.rasterizer);
+   draw_bind_vertex_shader(draw, st->vp->draw_shader);
+   set_feedback_vertex_format(ctx);
+
+   /* loop over TGSI shader inputs to determine vertex buffer
+    * and attribute info
+    */
+   for (attr = 0; attr < vp->num_inputs; attr++) {
+      const GLuint mesaAttr = vp->index_to_input[attr];
+      struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
+      void *map;
+
+      if (bufobj && bufobj->Name) {
+         /* Attribute data is in a VBO.
+          * Recall that for VBOs, the gl_client_array->Ptr field is
+          * really an offset from the start of the VBO, not a pointer.
+          */
+         struct st_buffer_object *stobj = st_buffer_object(bufobj);
+         assert(stobj->buffer);
+
+         vbuffers[attr].buffer = NULL;
+         pipe_buffer_reference(pipe->screen, &vbuffers[attr].buffer, stobj->buffer);
+         vbuffers[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */
+         velements[attr].src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr;
+      }
+      else {
+         /* attribute data is in user-space memory, not a VBO */
+         uint bytes = (arrays[mesaAttr]->Size
+                       * _mesa_sizeof_type(arrays[mesaAttr]->Type)
+                       * (max_index + 1));
+
+         /* wrap user data */
+         vbuffers[attr].buffer
+            = pipe_user_buffer_create(pipe->screen, (void *) arrays[mesaAttr]->Ptr,
+                                      bytes);
+         vbuffers[attr].buffer_offset = 0;
+         velements[attr].src_offset = 0;
+      }
+
+      /* common-case setup */
+      vbuffers[attr].stride = arrays[mesaAttr]->StrideB; /* in bytes */
+      vbuffers[attr].max_index = max_index;
+      velements[attr].vertex_buffer_index = attr;
+      velements[attr].nr_components = arrays[mesaAttr]->Size;
+      velements[attr].src_format = 
+         st_pipe_vertex_format(arrays[mesaAttr]->Type,
+                               arrays[mesaAttr]->Size,
+                               arrays[mesaAttr]->Normalized);
+      assert(velements[attr].src_format);
+
+      /* tell draw about this attribute */
+#if 0
+      draw_set_vertex_buffer(draw, attr, &vbuffer[attr]);
+#endif
+
+      /* map the attrib buffer */
+      map = pipe_buffer_map(pipe->screen, vbuffers[attr].buffer,
+                            PIPE_BUFFER_USAGE_CPU_READ);
+      draw_set_mapped_vertex_buffer(draw, attr, map);
+   }
+
+   draw_set_vertex_buffers(draw, vp->num_inputs, vbuffers);
+   draw_set_vertex_elements(draw, vp->num_inputs, velements);
+
+   if (ib) {
+      unsigned indexSize;
+      struct gl_buffer_object *bufobj = ib->obj;
+      struct st_buffer_object *stobj = st_buffer_object(bufobj);
+      void *map;
+
+      index_buffer_handle = stobj->buffer;
+
+      switch (ib->type) {
+      case GL_UNSIGNED_INT:
+         indexSize = 4;
+         break;
+      case GL_UNSIGNED_SHORT:
+         indexSize = 2;
+         break;
+      default:
+         assert(0);
+	 return;
+      }
+
+      map = pipe_buffer_map(pipe->screen, index_buffer_handle,
+                            PIPE_BUFFER_USAGE_CPU_READ);
+      draw_set_mapped_element_buffer(draw, indexSize, map);
+   }
+   else {
+      /* no index/element buffer */
+      draw_set_mapped_element_buffer(draw, 0, NULL);
+   }
+
+
+   /* map constant buffers */
+   mapped_constants = pipe_buffer_map(pipe->screen,
+                                      st->state.constants[PIPE_SHADER_VERTEX].buffer,
+                                      PIPE_BUFFER_USAGE_CPU_READ);
+   draw_set_mapped_constant_buffer(st->draw, mapped_constants,
+                                   st->state.constants[PIPE_SHADER_VERTEX].buffer->size);
+
+
+   /* draw here */
+   for (i = 0; i < nr_prims; i++) {
+      draw_arrays(draw, prims[i].mode, prims[i].start, prims[i].count);
+   }
+
+
+   /* unmap constant buffers */
+   pipe_buffer_unmap(pipe->screen, st->state.constants[PIPE_SHADER_VERTEX].buffer);
+
+   /*
+    * unmap vertex/index buffers
+    */
+   for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
+      if (draw->pt.vertex_buffer[i].buffer) {
+         pipe_buffer_unmap(pipe->screen, draw->pt.vertex_buffer[i].buffer);
+         pipe_buffer_reference(pipe->screen, &draw->pt.vertex_buffer[i].buffer, NULL);
+         draw_set_mapped_vertex_buffer(draw, i, NULL);
+      }
+   }
+   if (ib) {
+      pipe_buffer_unmap(pipe->screen, index_buffer_handle);
+      draw_set_mapped_element_buffer(draw, 0, NULL);
+   }
+}
+
+#endif /* FEATURE_feedback || FEATURE_drawpix */
+
diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c
new file mode 100644
index 0000000000..cd34529d38
--- /dev/null
+++ b/src/mesa/state_tracker/st_extensions.c
@@ -0,0 +1,279 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright (c) 2008 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/extensions.h"
+#include "main/macros.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+
+#include "st_context.h"
+#include "st_extensions.h"
+
+
+static int _min(int a, int b)
+{
+   return (a < b) ? a : b;
+}
+
+static float _maxf(float a, float b)
+{
+   return (a > b) ? a : b;
+}
+
+static int _clamp(int a, int min, int max)
+{
+   if (a < min)
+      return min;
+   else if (a > max)
+      return max;
+   else
+      return a;
+}
+
+
+/**
+ * Query driver to get implementation limits.
+ * Note that we have to limit/clamp against Mesa's internal limits too.
+ */
+void st_init_limits(struct st_context *st)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+   struct gl_constants *c = &st->ctx->Const;
+
+   c->MaxTextureLevels
+      = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS),
+            MAX_TEXTURE_LEVELS);
+
+   c->Max3DTextureLevels
+      = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_3D_LEVELS),
+            MAX_3D_TEXTURE_LEVELS);
+
+   c->MaxCubeTextureLevels
+      = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS),
+            MAX_CUBE_TEXTURE_LEVELS);
+
+   c->MaxTextureRectSize
+      = _min(1 << (c->MaxTextureLevels - 1), MAX_TEXTURE_RECT_SIZE);
+
+   c->MaxTextureImageUnits
+      = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS),
+            MAX_TEXTURE_IMAGE_UNITS);
+
+   c->MaxVertexTextureImageUnits
+      = _min(screen->get_param(screen, PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS),
+             MAX_VERTEX_TEXTURE_IMAGE_UNITS);
+
+   c->MaxTextureCoordUnits
+      = _min(c->MaxTextureImageUnits, MAX_TEXTURE_COORD_UNITS);
+
+   c->MaxTextureUnits = _min(c->MaxTextureImageUnits, c->MaxTextureCoordUnits);
+
+   c->MaxDrawBuffers
+      = _clamp(screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS),
+              1, MAX_DRAW_BUFFERS);
+
+   c->MaxLineWidth
+      = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH));
+   c->MaxLineWidthAA
+      = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH_AA));
+
+   c->MaxPointSize
+      = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH));
+   c->MaxPointSizeAA
+      = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH_AA));
+
+   c->MaxTextureMaxAnisotropy
+      = _maxf(2.0f, screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_ANISOTROPY));
+
+   c->MaxTextureLodBias
+      = screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_LOD_BIAS);
+
+   c->MaxDrawBuffers
+      = CLAMP(screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS),
+              1, MAX_DRAW_BUFFERS);
+}
+
+
+/**
+ * Use pipe_screen::get_param() to query PIPE_CAP_ values to determine
+ * which GL extensions are supported.
+ * Quite a few extensions are always supported because they are standard
+ * features or can be built on top of other gallium features.
+ * Some fine tuning may still be needed.
+ */
+void st_init_extensions(struct st_context *st)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+   GLcontext *ctx = st->ctx;
+
+   /*
+    * Extensions that are supported by all Gallium drivers:
+    */
+   ctx->Extensions.ARB_multisample = GL_TRUE;
+   ctx->Extensions.ARB_fragment_program = GL_TRUE;
+   ctx->Extensions.ARB_texture_border_clamp = GL_TRUE; /* XXX temp */
+   ctx->Extensions.ARB_texture_compression = GL_TRUE;
+   ctx->Extensions.ARB_texture_cube_map = GL_TRUE;
+   ctx->Extensions.ARB_texture_env_combine = GL_TRUE;
+   ctx->Extensions.ARB_texture_env_crossbar = GL_TRUE;
+   ctx->Extensions.ARB_texture_env_dot3 = GL_TRUE;
+   ctx->Extensions.ARB_vertex_program = GL_TRUE;
+   ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE;
+
+   ctx->Extensions.EXT_blend_color = GL_TRUE;
+   ctx->Extensions.EXT_blend_equation_separate = GL_TRUE;
+   ctx->Extensions.EXT_blend_func_separate = GL_TRUE;
+   ctx->Extensions.EXT_blend_logic_op = GL_TRUE;
+   ctx->Extensions.EXT_blend_minmax = GL_TRUE;
+   ctx->Extensions.EXT_blend_subtract = GL_TRUE;
+   ctx->Extensions.EXT_framebuffer_blit = GL_TRUE;
+   ctx->Extensions.EXT_framebuffer_object = GL_TRUE;
+   ctx->Extensions.EXT_fog_coord = GL_TRUE;
+   ctx->Extensions.EXT_multi_draw_arrays = GL_TRUE;
+   ctx->Extensions.EXT_pixel_buffer_object = GL_TRUE;
+   ctx->Extensions.EXT_point_parameters = GL_TRUE;
+   ctx->Extensions.EXT_secondary_color = GL_TRUE;
+   ctx->Extensions.EXT_stencil_wrap = GL_TRUE;
+   ctx->Extensions.EXT_texture_env_add = GL_TRUE;
+   ctx->Extensions.EXT_texture_env_combine = GL_TRUE;
+   ctx->Extensions.EXT_texture_env_dot3 = GL_TRUE;
+   ctx->Extensions.EXT_texture_lod_bias = GL_TRUE;
+
+   ctx->Extensions.NV_blend_square = GL_TRUE;
+   ctx->Extensions.NV_texgen_reflection = GL_TRUE;
+
+   ctx->Extensions.SGI_color_matrix = GL_TRUE;
+   ctx->Extensions.SGIS_generate_mipmap = GL_TRUE;
+
+   /*
+    * Extensions that depend on the driver/hardware:
+    */
+   if (screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS) > 0) {
+      ctx->Extensions.ARB_draw_buffers = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_GLSL)) {
+      ctx->Extensions.ARB_fragment_shader = GL_TRUE;
+      ctx->Extensions.ARB_vertex_shader = GL_TRUE;
+      ctx->Extensions.ARB_shader_objects = GL_TRUE;
+      ctx->Extensions.ARB_shading_language_100 = GL_TRUE;
+      ctx->Extensions.ARB_shading_language_120 = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_TEXTURE_MIRROR_REPEAT) > 0) {
+      ctx->Extensions.ARB_texture_mirrored_repeat = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_TEXTURE_MIRROR_CLAMP) > 0) {
+      ctx->Extensions.EXT_texture_mirror_clamp = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_NPOT_TEXTURES)) {
+      ctx->Extensions.ARB_texture_non_power_of_two = GL_TRUE;
+      ctx->Extensions.NV_texture_rectangle = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS) > 1) {
+      ctx->Extensions.ARB_multitexture = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_TWO_SIDED_STENCIL)) {
+      ctx->Extensions.ATI_separate_stencil = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_ANISOTROPIC_FILTER)) {
+      ctx->Extensions.EXT_texture_filter_anisotropic = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_POINT_SPRITE)) {
+      ctx->Extensions.ARB_point_sprite = GL_TRUE;
+      ctx->Extensions.NV_point_sprite = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_OCCLUSION_QUERY)) {
+      ctx->Extensions.ARB_occlusion_query = GL_TRUE;
+   }
+
+   if (screen->get_param(screen, PIPE_CAP_TEXTURE_SHADOW_MAP)) {
+      ctx->Extensions.ARB_depth_texture = GL_TRUE;
+      ctx->Extensions.ARB_shadow = GL_TRUE;
+      ctx->Extensions.EXT_shadow_funcs = GL_TRUE;
+      /*ctx->Extensions.ARB_shadow_ambient = GL_TRUE;*/
+   }
+
+   /* GL_EXT_packed_depth_stencil requires both the ability to render to
+    * a depth/stencil buffer and texture from depth/stencil source.
+    */
+   if (screen->is_format_supported(screen, PIPE_FORMAT_Z24S8_UNORM,
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0) &&
+       screen->is_format_supported(screen, PIPE_FORMAT_Z24S8_UNORM,
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      ctx->Extensions.EXT_packed_depth_stencil = GL_TRUE;
+   }
+   else if (screen->is_format_supported(screen, PIPE_FORMAT_S8Z24_UNORM,
+                                        PIPE_TEXTURE_2D, 
+                                        PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0) &&
+            screen->is_format_supported(screen, PIPE_FORMAT_S8Z24_UNORM,
+                                        PIPE_TEXTURE_2D, 
+                                        PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      ctx->Extensions.EXT_packed_depth_stencil = GL_TRUE;
+   }
+
+   /* sRGB support */
+   if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8B8A8_SRGB,
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0) ||
+      screen->is_format_supported(screen, PIPE_FORMAT_A8R8G8B8_SRGB,
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      ctx->Extensions.EXT_texture_sRGB = GL_TRUE;
+   }
+
+   /* s3tc support */
+   if (screen->is_format_supported(screen, PIPE_FORMAT_DXT5_RGBA,
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE;
+   }
+
+   /* ycbcr support */
+   if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR, 
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0) ||
+       screen->is_format_supported(screen, PIPE_FORMAT_YCBCR_REV, 
+                                   PIPE_TEXTURE_2D, 
+                                   PIPE_TEXTURE_USAGE_SAMPLER, 0)) {
+      ctx->Extensions.MESA_ycbcr_texture = GL_TRUE;
+   }
+}
diff --git a/src/mesa/state_tracker/st_extensions.h b/src/mesa/state_tracker/st_extensions.h
new file mode 100644
index 0000000000..2994f16dd3
--- /dev/null
+++ b/src/mesa/state_tracker/st_extensions.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_EXTENSIONS_H
+#define ST_EXTENSIONS_H
+
+
+extern void st_init_limits(struct st_context *st);
+
+extern void st_init_extensions(struct st_context *st);
+
+
+#endif /* ST_EXTENSIONS_H */
diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c
new file mode 100644
index 0000000000..9e2d60c926
--- /dev/null
+++ b/src/mesa/state_tracker/st_format.c
@@ -0,0 +1,718 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright (c) 2008 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+/**
+ * Texture Image-related functions.
+ * \author Brian Paul
+ */
+
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/texstore.h"
+#include "main/texformat.h"
+#include "main/enums.h"
+#include "main/macros.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+#include "st_context.h"
+#include "st_format.h"
+
+static GLuint
+format_bits(
+   pipe_format_rgbazs_t  info,
+   GLuint comp )
+{
+   return pf_get_component_bits( (enum pipe_format) info, comp );
+}
+
+static GLuint
+format_max_bits(
+   pipe_format_rgbazs_t  info )
+{
+   GLuint   size = format_bits( info, PIPE_FORMAT_COMP_R );
+
+   size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_G ) );
+   size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_B ) );
+   size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_A ) );
+   size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_Z ) );
+   size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_S ) );
+   return size;
+}
+
+static GLuint
+format_size(
+   pipe_format_rgbazs_t  info )
+{
+   return
+      format_bits( info, PIPE_FORMAT_COMP_R ) +
+      format_bits( info, PIPE_FORMAT_COMP_G ) +
+      format_bits( info, PIPE_FORMAT_COMP_B ) +
+      format_bits( info, PIPE_FORMAT_COMP_A ) +
+      format_bits( info, PIPE_FORMAT_COMP_Z ) +
+      format_bits( info, PIPE_FORMAT_COMP_S );
+}
+
+/*
+ * XXX temporary here
+ */
+GLboolean
+st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo)
+{
+   if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) {
+      pipe_format_rgbazs_t info;
+
+      info = format;
+
+#if 0
+      printf("%s\n", pf_name( format ) );
+#endif
+
+      /* Data type */
+      if (format == PIPE_FORMAT_A1R5G5B5_UNORM || format == PIPE_FORMAT_R5G6B5_UNORM) {
+         pinfo->datatype = GL_UNSIGNED_SHORT;
+      }
+      else {
+         GLuint size;
+
+         size = format_max_bits( info );
+         if (size == 8) {
+            if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM)
+               pinfo->datatype = GL_UNSIGNED_BYTE;
+            else
+               pinfo->datatype = GL_BYTE;
+         }
+         else if (size == 16) {
+            if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM)
+               pinfo->datatype = GL_UNSIGNED_SHORT;
+            else
+               pinfo->datatype = GL_SHORT;
+         }
+         else {
+            assert( size <= 32 );
+            if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM)
+               pinfo->datatype = GL_UNSIGNED_INT;
+            else
+               pinfo->datatype = GL_INT;
+         }
+      }
+
+      /* Component bits */
+      pinfo->red_bits = format_bits( info, PIPE_FORMAT_COMP_R );
+      pinfo->green_bits = format_bits( info, PIPE_FORMAT_COMP_G );
+      pinfo->blue_bits = format_bits( info, PIPE_FORMAT_COMP_B );
+      pinfo->alpha_bits = format_bits( info, PIPE_FORMAT_COMP_A );
+      pinfo->depth_bits = format_bits( info, PIPE_FORMAT_COMP_Z );
+      pinfo->stencil_bits = format_bits( info, PIPE_FORMAT_COMP_S );
+      pinfo->luminance_bits = 0;
+      pinfo->intensity_bits = 0;
+
+      /* Format size */
+      pinfo->size = format_size( info ) / 8;
+
+      /* Luminance & Intensity bits */
+      if( pf_swizzle_x(info) == PIPE_FORMAT_COMP_R &&
+          pf_swizzle_y(info) == PIPE_FORMAT_COMP_R &&
+          pf_swizzle_z(info) == PIPE_FORMAT_COMP_R ) {
+         if( pf_swizzle_w(info) == PIPE_FORMAT_COMP_R ) {
+            pinfo->intensity_bits = pinfo->red_bits;
+         }
+         else {
+            pinfo->luminance_bits = pinfo->red_bits;
+         }
+         pinfo->red_bits = 0;
+      }
+
+      /* Base format */
+      if (pinfo->depth_bits) {
+         if (pinfo->stencil_bits) {
+            pinfo->base_format = GL_DEPTH_STENCIL_EXT;
+         }
+         else {
+            pinfo->base_format = GL_DEPTH_COMPONENT;
+         }
+      }
+      else if (pinfo->stencil_bits) {
+         pinfo->base_format = GL_STENCIL_INDEX;
+      }
+      else {
+         pinfo->base_format = GL_RGBA;
+      }
+   }
+   else if (pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR) {
+      pinfo->base_format = GL_YCBCR_MESA;
+      pinfo->datatype = GL_UNSIGNED_SHORT;
+      pinfo->size = 2; /* two bytes per "texel" */
+   }
+   else {
+      /* compressed format? */
+      assert(0);
+   }
+
+#if 0
+   printf(
+      "ST_FORMAT: R(%u), G(%u), B(%u), A(%u), Z(%u), S(%u)\n",
+      pinfo->red_bits,
+      pinfo->green_bits,
+      pinfo->blue_bits,
+      pinfo->alpha_bits,
+      pinfo->depth_bits,
+      pinfo->stencil_bits );
+#endif
+
+   pinfo->format = format;
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Return bytes per pixel for the given format.
+ */
+GLuint
+st_sizeof_format(enum pipe_format format)
+{
+   struct pipe_format_info info;
+   if (!st_get_format_info( format, &info )) {
+      assert( 0 );
+      return 0;
+   }
+   return info.size;
+}
+
+
+/**
+ * Return bytes per pixel for the given format.
+ */
+GLenum
+st_format_datatype(enum pipe_format format)
+{
+   struct pipe_format_info info;
+   if (!st_get_format_info( format, &info )) {
+      assert( 0 );
+      return 0;
+   }
+   return info.datatype;
+}
+
+
+enum pipe_format
+st_mesa_format_to_pipe_format(GLuint mesaFormat)
+{
+   switch (mesaFormat) {
+      /* fix this */
+   case MESA_FORMAT_ARGB8888_REV:
+   case MESA_FORMAT_ARGB8888:
+      return PIPE_FORMAT_A8R8G8B8_UNORM;
+   case MESA_FORMAT_ARGB1555:
+      return PIPE_FORMAT_A1R5G5B5_UNORM;
+   case MESA_FORMAT_ARGB4444:
+      return PIPE_FORMAT_A4R4G4B4_UNORM;
+   case MESA_FORMAT_RGB565:
+      return PIPE_FORMAT_R5G6B5_UNORM;
+   case MESA_FORMAT_AL88:
+      return PIPE_FORMAT_A8L8_UNORM;
+   case MESA_FORMAT_A8:
+      return PIPE_FORMAT_A8_UNORM;
+   case MESA_FORMAT_L8:
+      return PIPE_FORMAT_L8_UNORM;
+   case MESA_FORMAT_I8:
+      return PIPE_FORMAT_I8_UNORM;
+   case MESA_FORMAT_Z16:
+      return PIPE_FORMAT_Z16_UNORM;
+   case MESA_FORMAT_Z32:
+      return PIPE_FORMAT_Z32_UNORM;
+   case MESA_FORMAT_Z24_S8:
+      return PIPE_FORMAT_Z24S8_UNORM;
+   case MESA_FORMAT_S8_Z24:
+      return PIPE_FORMAT_S8Z24_UNORM;
+   case MESA_FORMAT_YCBCR:
+      return PIPE_FORMAT_YCBCR;
+#if FEATURE_texture_s3tc
+   case MESA_FORMAT_RGB_DXT1:
+      return PIPE_FORMAT_DXT1_RGB;
+   case MESA_FORMAT_RGBA_DXT1:
+      return PIPE_FORMAT_DXT1_RGBA;
+   case MESA_FORMAT_RGBA_DXT3:
+      return PIPE_FORMAT_DXT3_RGBA;
+   case MESA_FORMAT_RGBA_DXT5:
+      return PIPE_FORMAT_DXT5_RGBA;
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SRGB_DXT1:
+      return PIPE_FORMAT_DXT1_SRGB;
+   case MESA_FORMAT_SRGBA_DXT1:
+      return PIPE_FORMAT_DXT1_SRGBA;
+   case MESA_FORMAT_SRGBA_DXT3:
+      return PIPE_FORMAT_DXT3_SRGBA;
+   case MESA_FORMAT_SRGBA_DXT5:
+      return PIPE_FORMAT_DXT5_SRGBA;
+#endif
+#endif
+#if FEATURE_EXT_texture_sRGB
+   case MESA_FORMAT_SLA8:
+      return PIPE_FORMAT_A8L8_SRGB;
+   case MESA_FORMAT_SL8:
+      return PIPE_FORMAT_L8_SRGB;
+   case MESA_FORMAT_SRGB8:
+      return PIPE_FORMAT_R8G8B8_SRGB;
+   case MESA_FORMAT_SRGBA8:
+      return PIPE_FORMAT_R8G8B8A8_SRGB;
+   case MESA_FORMAT_SARGB8:
+      return PIPE_FORMAT_A8R8G8B8_SRGB;
+#endif
+   default:
+      assert(0);
+      return 0;
+   }
+}
+
+/**
+ * Find an RGBA format supported by the context/winsys.
+ */
+static enum pipe_format
+default_rgba_format(struct pipe_screen *screen, 
+                    enum pipe_texture_target target,
+                    unsigned tex_usage, 
+                    unsigned geom_flags)
+{
+   static const enum pipe_format colorFormats[] = {
+      PIPE_FORMAT_A8R8G8B8_UNORM,
+      PIPE_FORMAT_B8G8R8A8_UNORM,
+      PIPE_FORMAT_R8G8B8A8_UNORM,
+      PIPE_FORMAT_R5G6B5_UNORM
+   };
+   uint i;
+   for (i = 0; i < Elements(colorFormats); i++) {
+      if (screen->is_format_supported( screen, colorFormats[i], target, tex_usage, geom_flags )) {
+         return colorFormats[i];
+      }
+   }
+   return PIPE_FORMAT_NONE;
+}
+
+/**
+ * Find an sRGBA format supported by the context/winsys.
+ */
+static enum pipe_format
+default_srgba_format(struct pipe_screen *screen, 
+                    enum pipe_texture_target target,
+                    unsigned tex_usage, 
+                    unsigned geom_flags)
+{
+   static const enum pipe_format colorFormats[] = {
+      PIPE_FORMAT_A8R8G8B8_SRGB,
+      PIPE_FORMAT_B8G8R8A8_SRGB,
+      PIPE_FORMAT_R8G8B8A8_SRGB,
+   };
+   uint i;
+   for (i = 0; i < Elements(colorFormats); i++) {
+      if (screen->is_format_supported( screen, colorFormats[i], target, tex_usage, geom_flags )) {
+         return colorFormats[i];
+      }
+   }
+   return PIPE_FORMAT_NONE;
+}
+
+/**
+ * Search list of formats for first RGBA format with >8 bits/channel.
+ */
+static enum pipe_format
+default_deep_rgba_format(struct pipe_screen *screen, 
+                         enum pipe_texture_target target,
+                         unsigned tex_usage, 
+                         unsigned geom_flags)
+{
+   if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_SNORM, target, tex_usage, geom_flags)) {
+      return PIPE_FORMAT_R16G16B16A16_SNORM;
+   }
+   if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET)
+      return default_rgba_format(screen, target, tex_usage, geom_flags);
+   else
+      return PIPE_FORMAT_NONE;
+}
+
+
+/**
+ * Find an Z format supported by the context/winsys.
+ */
+static enum pipe_format
+default_depth_format(struct pipe_screen *screen, 
+                     enum pipe_texture_target target,
+                     unsigned tex_usage, 
+                     unsigned geom_flags)
+{
+   static const enum pipe_format zFormats[] = {
+      PIPE_FORMAT_Z16_UNORM,
+      PIPE_FORMAT_Z32_UNORM,
+      PIPE_FORMAT_S8Z24_UNORM,
+      PIPE_FORMAT_Z24S8_UNORM
+   };
+   uint i;
+   for (i = 0; i < Elements(zFormats); i++) {
+      if (screen->is_format_supported( screen, zFormats[i], target, tex_usage, geom_flags )) {
+         return zFormats[i];
+      }
+   }
+   return PIPE_FORMAT_NONE;
+}
+
+
+/**
+ * Given an OpenGL internalFormat value for a texture or surface, return
+ * the best matching PIPE_FORMAT_x, or PIPE_FORMAT_NONE if there's no match.
+ * \param target  one of PIPE_TEXTURE_x
+ * \param tex_usage  either PIPE_TEXTURE_USAGE_RENDER_TARGET
+ *                   or PIPE_TEXTURE_USAGE_SAMPLER
+ */
+enum pipe_format
+st_choose_format(struct pipe_context *pipe, GLint internalFormat,
+                 enum pipe_texture_target target, unsigned tex_usage)
+{
+   struct pipe_screen *screen = pipe->screen;
+   unsigned geom_flags = 0;
+
+   switch (internalFormat) {
+   case 4:
+   case GL_RGBA:
+   case GL_COMPRESSED_RGBA:
+   case 3:
+   case GL_RGB:
+   case GL_COMPRESSED_RGB:
+   case GL_RGBA8:
+   case GL_RGB10_A2:
+   case GL_RGBA12:
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+   case GL_RGBA16:
+      if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET)
+         return default_deep_rgba_format( screen, target, tex_usage, geom_flags );
+      else
+         return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_RGBA4:
+   case GL_RGBA2:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A4R4G4B4_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A4R4G4B4_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_RGB5_A1:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A1R5G5B5_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_RGB8:
+   case GL_RGB10:
+   case GL_RGB12:
+   case GL_RGB16:
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_RGB5:
+   case GL_RGB4:
+   case GL_R3_G3_B2:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A1R5G5B5_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_R5G6B5_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_R5G6B5_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_ALPHA:
+   case GL_ALPHA4:
+   case GL_ALPHA8:
+   case GL_ALPHA12:
+   case GL_ALPHA16:
+   case GL_COMPRESSED_ALPHA:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A8_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case 1:
+   case GL_LUMINANCE:
+   case GL_LUMINANCE4:
+   case GL_LUMINANCE8:
+   case GL_LUMINANCE12:
+   case GL_LUMINANCE16:
+   case GL_COMPRESSED_LUMINANCE:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_L8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_L8_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case 2:
+   case GL_LUMINANCE_ALPHA:
+   case GL_LUMINANCE4_ALPHA4:
+   case GL_LUMINANCE6_ALPHA2:
+   case GL_LUMINANCE8_ALPHA8:
+   case GL_LUMINANCE12_ALPHA4:
+   case GL_LUMINANCE12_ALPHA12:
+   case GL_LUMINANCE16_ALPHA16:
+   case GL_COMPRESSED_LUMINANCE_ALPHA:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A8L8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A8L8_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_INTENSITY:
+   case GL_INTENSITY4:
+   case GL_INTENSITY8:
+   case GL_INTENSITY12:
+   case GL_INTENSITY16:
+   case GL_COMPRESSED_INTENSITY:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_I8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_I8_UNORM;
+      return default_rgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_YCBCR_MESA:
+      if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR,
+                                      target, tex_usage, geom_flags)) {
+         return PIPE_FORMAT_YCBCR;
+      }
+      if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR_REV,
+                                      target, tex_usage, geom_flags)) {
+         return PIPE_FORMAT_YCBCR_REV;
+      }
+      return PIPE_FORMAT_NONE;
+
+   case GL_RGB_S3TC:
+   case GL_RGB4_S3TC:
+   case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
+      return PIPE_FORMAT_DXT1_RGB;
+
+   case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
+      return PIPE_FORMAT_DXT1_RGBA;
+
+   case GL_RGBA_S3TC:
+   case GL_RGBA4_S3TC:
+   case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
+      return PIPE_FORMAT_DXT3_RGBA;
+
+   case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
+      return PIPE_FORMAT_DXT5_RGBA;
+
+#if 0
+   case GL_COMPRESSED_RGB_FXT1_3DFX:
+      return PIPE_FORMAT_RGB_FXT1;
+   case GL_COMPRESSED_RGBA_FXT1_3DFX:
+      return PIPE_FORMAT_RGB_FXT1;
+#endif
+
+   case GL_DEPTH_COMPONENT16:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_Z16_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_Z16_UNORM;
+      /* fall-through */
+   case GL_DEPTH_COMPONENT24:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_S8Z24_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_Z24S8_UNORM;
+      /* fall-through */
+   case GL_DEPTH_COMPONENT32:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_Z32_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_Z32_UNORM;
+      /* fall-through */
+   case GL_DEPTH_COMPONENT:
+      return default_depth_format( screen, target, tex_usage, geom_flags );
+
+   case GL_STENCIL_INDEX:
+   case GL_STENCIL_INDEX1_EXT:
+   case GL_STENCIL_INDEX4_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+   case GL_STENCIL_INDEX16_EXT:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_S8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_S8_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_S8Z24_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_Z24S8_UNORM;
+      return PIPE_FORMAT_NONE;
+
+   case GL_DEPTH_STENCIL_EXT:
+   case GL_DEPTH24_STENCIL8_EXT:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_S8Z24_UNORM;
+      if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_Z24S8_UNORM;
+      return PIPE_FORMAT_NONE;
+
+   case GL_SRGB_EXT:
+   case GL_SRGB8_EXT:
+   case GL_COMPRESSED_SRGB_EXT:
+   case GL_COMPRESSED_SRGB_ALPHA_EXT:
+   case GL_SRGB_ALPHA_EXT:
+   case GL_SRGB8_ALPHA8_EXT:
+      return default_srgba_format( screen, target, tex_usage, geom_flags );
+   case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
+      return PIPE_FORMAT_DXT1_SRGB;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
+      return PIPE_FORMAT_DXT1_SRGBA;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
+      return PIPE_FORMAT_DXT3_SRGBA;
+   case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
+      return PIPE_FORMAT_DXT5_SRGBA;
+
+   case GL_SLUMINANCE_ALPHA_EXT:
+   case GL_SLUMINANCE8_ALPHA8_EXT:
+   case GL_COMPRESSED_SLUMINANCE_EXT:
+   case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_A8L8_SRGB, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_A8L8_SRGB;
+      return default_srgba_format( screen, target, tex_usage, geom_flags );
+
+   case GL_SLUMINANCE_EXT:
+   case GL_SLUMINANCE8_EXT:
+      if (screen->is_format_supported( screen, PIPE_FORMAT_L8_SRGB, target, tex_usage, geom_flags ))
+         return PIPE_FORMAT_L8_SRGB;
+      return default_srgba_format( screen, target, tex_usage, geom_flags );
+
+   default:
+      return PIPE_FORMAT_NONE;
+   }
+}
+
+
+static GLboolean
+is_stencil_format(GLenum format)
+{
+   switch (format) {
+   case GL_STENCIL_INDEX:
+   case GL_STENCIL_INDEX1_EXT:
+   case GL_STENCIL_INDEX4_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+   case GL_STENCIL_INDEX16_EXT:
+   case GL_DEPTH_STENCIL_EXT:
+   case GL_DEPTH24_STENCIL8_EXT:
+      return GL_TRUE;
+   default:
+      return GL_FALSE;
+   }
+}
+
+/**
+ * Called by FBO code to choose a PIPE_FORMAT_ for drawing surfaces.
+ */
+enum pipe_format
+st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat)
+{
+   uint usage;
+   if (is_stencil_format(internalFormat))
+      usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+   else
+      usage = PIPE_TEXTURE_USAGE_RENDER_TARGET;
+   return st_choose_format(pipe, internalFormat, PIPE_TEXTURE_2D, usage);
+}
+
+
+static const struct gl_texture_format *
+translate_gallium_format_to_mesa_format(enum pipe_format format)
+{
+   switch (format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      return &_mesa_texformat_argb8888;
+   case PIPE_FORMAT_A1R5G5B5_UNORM:
+      return &_mesa_texformat_argb1555;
+   case PIPE_FORMAT_A4R4G4B4_UNORM:
+      return &_mesa_texformat_argb4444;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      return &_mesa_texformat_rgb565;
+   case PIPE_FORMAT_A8L8_UNORM:
+      return &_mesa_texformat_al88;
+   case PIPE_FORMAT_A8_UNORM:
+      return &_mesa_texformat_a8;
+   case PIPE_FORMAT_L8_UNORM:
+      return &_mesa_texformat_l8;
+   case PIPE_FORMAT_I8_UNORM:
+      return &_mesa_texformat_i8;
+   case PIPE_FORMAT_Z16_UNORM:
+      return &_mesa_texformat_z16;
+   case PIPE_FORMAT_Z32_UNORM:
+      return &_mesa_texformat_z32;
+   case PIPE_FORMAT_Z24S8_UNORM:
+      return &_mesa_texformat_z24_s8;
+   case PIPE_FORMAT_S8Z24_UNORM:
+      return &_mesa_texformat_s8_z24;
+   case PIPE_FORMAT_YCBCR:
+      return &_mesa_texformat_ycbcr;
+   case PIPE_FORMAT_YCBCR_REV:
+      return &_mesa_texformat_ycbcr_rev;
+#if FEATURE_texture_s3tc
+   case PIPE_FORMAT_DXT1_RGB:
+      return &_mesa_texformat_rgb_dxt1;
+   case PIPE_FORMAT_DXT1_RGBA:
+      return &_mesa_texformat_rgba_dxt1;
+   case PIPE_FORMAT_DXT3_RGBA:
+      return &_mesa_texformat_rgba_dxt3;
+   case PIPE_FORMAT_DXT5_RGBA:
+      return &_mesa_texformat_rgba_dxt5;
+#if FEATURE_EXT_texture_sRGB
+   case PIPE_FORMAT_DXT1_SRGB:
+      return &_mesa_texformat_srgb_dxt1;
+   case PIPE_FORMAT_DXT1_SRGBA:
+      return &_mesa_texformat_srgba_dxt1;
+   case PIPE_FORMAT_DXT3_SRGBA:
+      return &_mesa_texformat_srgba_dxt3;
+   case PIPE_FORMAT_DXT5_SRGBA:
+      return &_mesa_texformat_srgba_dxt5;
+#endif
+#endif
+#if FEATURE_EXT_texture_sRGB
+   case PIPE_FORMAT_A8L8_SRGB:
+      return &_mesa_texformat_sla8;
+   case PIPE_FORMAT_L8_SRGB:
+      return &_mesa_texformat_sl8;
+   case PIPE_FORMAT_R8G8B8_SRGB:
+      return &_mesa_texformat_srgb8;
+   case PIPE_FORMAT_R8G8B8A8_SRGB:
+      return &_mesa_texformat_srgba8;
+   case PIPE_FORMAT_A8R8G8B8_SRGB:
+      return &_mesa_texformat_sargb8;
+#endif
+   /* XXX add additional cases */
+   default:
+      assert(0);
+      return NULL;
+   }
+}
+
+
+/**
+ * Called via ctx->Driver.chooseTextureFormat().
+ */
+const struct gl_texture_format *
+st_ChooseTextureFormat(GLcontext *ctx, GLint internalFormat,
+                       GLenum format, GLenum type)
+{
+   enum pipe_format pFormat;
+
+   (void) format;
+   (void) type;
+
+   pFormat = st_choose_format(ctx->st->pipe, internalFormat, PIPE_TEXTURE_2D, 
+                              PIPE_TEXTURE_USAGE_SAMPLER);
+   if (pFormat == PIPE_FORMAT_NONE)
+      return NULL;
+
+   return translate_gallium_format_to_mesa_format(pFormat);
+}
diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h
new file mode 100644
index 0000000000..3f5ac3201b
--- /dev/null
+++ b/src/mesa/state_tracker/st_format.h
@@ -0,0 +1,79 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_FORMAT_H
+#define ST_FORMAT_H
+
+
+struct pipe_format_info
+{
+   enum pipe_format format;
+   GLenum base_format;
+   GLenum datatype;
+   GLubyte red_bits;
+   GLubyte green_bits;
+   GLubyte blue_bits;
+   GLubyte alpha_bits;
+   GLubyte luminance_bits;
+   GLubyte intensity_bits;
+   GLubyte depth_bits;
+   GLubyte stencil_bits;
+   GLubyte size;           /**< in bytes */
+};
+
+
+GLboolean
+st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo);
+
+
+extern GLuint
+st_sizeof_format(enum pipe_format format);
+
+
+extern GLenum
+st_format_datatype(enum pipe_format format);
+
+
+extern enum pipe_format
+st_mesa_format_to_pipe_format(GLuint mesaFormat);
+
+
+extern enum pipe_format
+st_choose_format(struct pipe_context *pipe, GLint internalFormat,
+                 enum pipe_texture_target target, unsigned tex_usage);
+
+extern enum pipe_format
+st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat);
+
+
+extern const struct gl_texture_format *
+st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
+                       GLenum format, GLenum type);
+
+
+#endif /* ST_CB_TEXIMAGE_H */
diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c
new file mode 100644
index 0000000000..0d9c7b97e3
--- /dev/null
+++ b/src/mesa/state_tracker/st_framebuffer.c
@@ -0,0 +1,333 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/imports.h"
+#include "main/buffers.h"
+#include "main/context.h"
+#include "main/framebuffer.h"
+#include "main/matrix.h"
+#include "main/renderbuffer.h"
+#include "main/scissor.h"
+#include "st_context.h"
+#include "st_cb_fbo.h"
+#include "st_public.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_inlines.h"
+
+
+struct st_framebuffer *
+st_create_framebuffer( const __GLcontextModes *visual,
+                       enum pipe_format colorFormat,
+                       enum pipe_format depthFormat,
+                       enum pipe_format stencilFormat,
+                       uint width, uint height,
+                       void *private)
+{
+   struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer);
+   if (stfb) {
+      int samples = st_get_msaa();
+
+      if (visual->sampleBuffers)
+         samples = visual->samples;
+
+      _mesa_initialize_framebuffer(&stfb->Base, visual);
+
+      {
+         /* fake frontbuffer */
+         /* XXX allocation should only happen in the unusual case
+            it's actually needed */
+         struct gl_renderbuffer *rb
+            = st_new_renderbuffer_fb(colorFormat, samples);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
+      }
+
+      if (visual->doubleBufferMode) {
+         struct gl_renderbuffer *rb
+            = st_new_renderbuffer_fb(colorFormat, samples);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb);
+      }
+
+      if (depthFormat == stencilFormat && depthFormat != PIPE_FORMAT_NONE) {
+         /* combined depth/stencil buffer */
+         struct gl_renderbuffer *depthStencilRb
+            = st_new_renderbuffer_fb(depthFormat, samples);
+         /* note: bind RB to two attachment points */
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, depthStencilRb);
+      }
+      else {
+         /* separate depth and/or stencil */
+
+         if (visual->depthBits == 32) {
+            /* 32-bit depth buffer */
+            struct gl_renderbuffer *depthRb
+               = st_new_renderbuffer_fb(depthFormat, samples);
+            _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
+         }
+         else if (visual->depthBits == 24) {
+            /* 24-bit depth buffer, ignore stencil bits */
+            struct gl_renderbuffer *depthRb
+               = st_new_renderbuffer_fb(depthFormat, samples);
+            _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
+         }
+         else if (visual->depthBits > 0) {
+            /* 16-bit depth buffer */
+            struct gl_renderbuffer *depthRb
+               = st_new_renderbuffer_fb(depthFormat, samples);
+            _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
+         }
+
+         if (visual->stencilBits > 0) {
+            /* 8-bit stencil */
+            struct gl_renderbuffer *stencilRb
+               = st_new_renderbuffer_fb(stencilFormat, samples);
+            _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, stencilRb);
+         }
+      }
+
+      if (visual->accumRedBits > 0) {
+         /* 16-bit/channel accum */
+         struct gl_renderbuffer *accumRb
+            = st_new_renderbuffer_fb(DEFAULT_ACCUM_PIPE_FORMAT, 0); /* XXX accum isn't multisampled right? */
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb);
+      }
+
+      stfb->Base.Initialized = GL_TRUE;
+      stfb->InitWidth = width;
+      stfb->InitHeight = height;
+      stfb->Private = private;
+   }
+   return stfb;
+}
+
+
+void st_resize_framebuffer( struct st_framebuffer *stfb,
+                            uint width, uint height )
+{
+   if (stfb->Base.Width != width || stfb->Base.Height != height) {
+      GET_CURRENT_CONTEXT(ctx);
+      if (ctx) {
+         if (stfb->InitWidth == 0 && stfb->InitHeight == 0) {
+            /* didn't have a valid size until now */
+            stfb->InitWidth = width;
+            stfb->InitHeight = height;
+            if (ctx->Viewport.Width <= 1) {
+               /* set context's initial viewport/scissor size */
+               _mesa_set_viewport(ctx, 0, 0, width, height);
+               _mesa_set_scissor(ctx, 0, 0, width, height);
+            }
+         }
+
+         _mesa_resize_framebuffer(ctx, &stfb->Base, width, height);
+
+         assert(stfb->Base.Width == width);
+         assert(stfb->Base.Height == height);
+      }
+   }
+}
+
+
+void st_unreference_framebuffer( struct st_framebuffer *stfb )
+{
+   _mesa_unreference_framebuffer((struct gl_framebuffer **) &stfb);
+}
+
+
+
+/**
+ * Set/replace a framebuffer surface.
+ * The user of the state tracker can use this instead of
+ * st_resize_framebuffer() to provide new surfaces when a window is resized.
+ */
+void
+st_set_framebuffer_surface(struct st_framebuffer *stfb,
+                           uint surfIndex, struct pipe_surface *surf)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   static const GLuint invalid_size = 9999999;
+   struct st_renderbuffer *strb;
+   GLuint width, height, i;
+
+   assert(surfIndex < BUFFER_COUNT);
+
+   strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer);
+
+   /* fail */
+   if (!strb) return;
+
+   /* replace the renderbuffer's surface/texture pointers */
+   pipe_surface_reference( &strb->surface, surf );
+   pipe_texture_reference( &strb->texture, surf->texture );
+
+   if (ctx) {
+      /* If ctx isn't set, we've likely not made current yet.
+       * But when we do, we need to start setting this dirty bit
+       * to ensure the renderbuffer attachements are up-to-date
+       * via update_framebuffer.
+       */
+      ctx->st->dirty.st |= ST_NEW_FRAMEBUFFER;
+   }
+
+   /* update renderbuffer's width/height */
+   strb->Base.Width = surf->width;
+   strb->Base.Height = surf->height;
+
+   /* Try to update the framebuffer's width/height from the renderbuffer
+    * sizes.  Before we start drawing, all the rbs _should_ be the same size.
+    */
+   width = height = invalid_size;
+   for (i = 0; i < BUFFER_COUNT; i++) {
+      if (stfb->Base.Attachment[i].Renderbuffer) {
+         if (width == invalid_size) {
+            width = stfb->Base.Attachment[i].Renderbuffer->Width;
+            height = stfb->Base.Attachment[i].Renderbuffer->Height;
+         }
+         else if (width != stfb->Base.Attachment[i].Renderbuffer->Width ||
+                  height != stfb->Base.Attachment[i].Renderbuffer->Height) {
+            /* inconsistant renderbuffer sizes, bail out */
+            return;
+         }
+      }
+   }
+
+   if (width != invalid_size) {
+      /* OK, the renderbuffers are of a consistant size, so update the
+       * parent framebuffer's size.
+       */
+      stfb->Base.Width = width;
+      stfb->Base.Height = height;
+   }
+}
+
+
+
+/**
+ * Return the pipe_surface for the given renderbuffer.
+ */
+int
+st_get_framebuffer_surface(struct st_framebuffer *stfb, uint surfIndex, struct pipe_surface **surface)
+{
+   struct st_renderbuffer *strb;
+
+   assert(surfIndex <= ST_SURFACE_DEPTH);
+
+   /* sanity checks, ST tokens should match Mesa tokens */
+   assert(ST_SURFACE_FRONT_LEFT == BUFFER_FRONT_LEFT);
+   assert(ST_SURFACE_BACK_RIGHT == BUFFER_BACK_RIGHT);
+
+   strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer);
+   if (strb) {
+      *surface = strb->surface;
+      return GL_TRUE;
+   }
+
+   *surface = NULL;
+   return GL_FALSE;
+}
+
+int
+st_get_framebuffer_texture(struct st_framebuffer *stfb, uint surfIndex, struct pipe_texture **texture)
+{
+   struct st_renderbuffer *strb;
+
+   assert(surfIndex <= ST_SURFACE_DEPTH);
+
+   /* sanity checks, ST tokens should match Mesa tokens */
+   assert(ST_SURFACE_FRONT_LEFT == BUFFER_FRONT_LEFT);
+   assert(ST_SURFACE_BACK_RIGHT == BUFFER_BACK_RIGHT);
+
+   strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer);
+   if (strb) {
+      *texture = strb->texture;
+      return GL_TRUE;
+   }
+
+   *texture = NULL;
+   return GL_FALSE;
+}
+
+/**
+ * This function is to be called prior to SwapBuffers on the given
+ * framebuffer.  It checks if the current context is bound to the framebuffer
+ * and flushes rendering if needed.
+ */
+void
+st_notify_swapbuffers(struct st_framebuffer *stfb)
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if (ctx && ctx->DrawBuffer == &stfb->Base) {
+      st_flush( ctx->st, 
+		PIPE_FLUSH_RENDER_CACHE | 
+		PIPE_FLUSH_SWAPBUFFERS |
+		PIPE_FLUSH_FRAME,
+                NULL );
+      ctx->st->frontbuffer_status = FRONT_STATUS_COPY_OF_BACK;
+   }
+}
+
+
+/** 
+ * Quick hack - allows the winsys to inform the driver that surface
+ * states are now undefined after a glXSwapBuffers or similar.
+ */
+void
+st_notify_swapbuffers_complete(struct st_framebuffer *stfb)
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   if (ctx && ctx->DrawBuffer == &stfb->Base) {
+      struct st_renderbuffer *strb;
+
+      /* Mark back color buffers as undefined */
+      strb = st_renderbuffer(stfb->Base.Attachment[BUFFER_BACK_LEFT].
+                             Renderbuffer);
+      if (strb && strb->surface)
+         strb->surface->status = PIPE_SURFACE_STATUS_UNDEFINED;
+
+      strb = st_renderbuffer(stfb->Base.Attachment[BUFFER_BACK_RIGHT].
+                             Renderbuffer);
+      if (strb && strb->surface)
+         strb->surface->status = PIPE_SURFACE_STATUS_UNDEFINED;
+   }
+}
+
+
+void *st_framebuffer_private( struct st_framebuffer *stfb )
+{
+   return stfb->Private;
+}
+
+void st_get_framebuffer_dimensions( struct st_framebuffer *stfb,
+				    uint *width,
+				    uint *height)
+{
+   *width = stfb->Base.Width;
+   *height = stfb->Base.Height;
+}
diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c
new file mode 100644
index 0000000000..6a3455aaba
--- /dev/null
+++ b/src/mesa/state_tracker/st_gen_mipmap.c
@@ -0,0 +1,209 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/imports.h"
+#include "main/mipmap.h"
+#include "main/teximage.h"
+#include "main/texformat.h"
+
+#include "shader/prog_instruction.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_gen_mipmap.h"
+
+#include "cso_cache/cso_cache.h"
+#include "cso_cache/cso_context.h"
+
+#include "st_context.h"
+#include "st_draw.h"
+#include "st_gen_mipmap.h"
+#include "st_program.h"
+#include "st_texture.h"
+#include "st_cb_texture.h"
+
+
+/**
+ * one-time init for generate mipmap
+ * XXX Note: there may be other times we need no-op/simple state like this.
+ * In that case, some code refactoring would be good.
+ */
+void
+st_init_generate_mipmap(struct st_context *st)
+{
+   st->gen_mipmap = util_create_gen_mipmap(st->pipe, st->cso_context);
+}
+
+
+void
+st_destroy_generate_mipmap(struct st_context *st)
+{
+   util_destroy_gen_mipmap(st->gen_mipmap);
+   st->gen_mipmap = NULL;
+}
+
+
+/**
+ * Generate mipmap levels using hardware rendering.
+ * \return TRUE if successful, FALSE if not possible
+ */
+static boolean
+st_render_mipmap(struct st_context *st,
+                 GLenum target,
+                 struct pipe_texture *pt,
+                 uint baseLevel, uint lastLevel)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const uint face = _mesa_tex_target_to_face(target);
+
+   assert(target != GL_TEXTURE_3D); /* not done yet */
+
+   /* check if we can render in the texture's format */
+   if (!screen->is_format_supported(screen, pt->format, target,
+                                    PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) {
+      return FALSE;
+   }
+
+   util_gen_mipmap(st->gen_mipmap, pt, face, baseLevel, lastLevel,
+                   PIPE_TEX_FILTER_LINEAR);
+
+   return TRUE;
+}
+
+
+static void
+fallback_generate_mipmap(GLcontext *ctx, GLenum target,
+                         struct gl_texture_object *texObj)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture *pt = st_get_texobj_texture(texObj);
+   const uint baseLevel = texObj->BaseLevel;
+   const uint lastLevel = pt->last_level;
+   const uint face = _mesa_tex_target_to_face(target), zslice = 0;
+   uint dstLevel;
+   GLenum datatype;
+   GLuint comps;
+
+   assert(target != GL_TEXTURE_3D); /* not done yet */
+
+   _mesa_format_to_type_and_comps(texObj->Image[face][baseLevel]->TexFormat,
+                                  &datatype, &comps);
+
+   for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
+      const uint srcLevel = dstLevel - 1;
+      struct pipe_surface *srcSurf, *dstSurf;
+      const ubyte *srcData;
+      ubyte *dstData;
+
+      srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_READ);
+      dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice,
+                                        PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      srcData = (ubyte *) pipe_surface_map(srcSurf,
+                                           PIPE_BUFFER_USAGE_CPU_READ)
+              + srcSurf->offset;
+      dstData = (ubyte *) pipe_surface_map(dstSurf,
+                                           PIPE_BUFFER_USAGE_CPU_WRITE)
+              + dstSurf->offset;
+
+      _mesa_generate_mipmap_level(target, datatype, comps,
+                   0 /*border*/,
+                   pt->width[srcLevel], pt->height[srcLevel], pt->depth[srcLevel],
+                   srcData,
+                   srcSurf->stride, /* stride in bytes */
+                   pt->width[dstLevel], pt->height[dstLevel], pt->depth[dstLevel],
+                   dstData,
+                   dstSurf->stride); /* stride in bytes */
+
+      pipe_surface_unmap(srcSurf);
+      pipe_surface_unmap(dstSurf);
+
+      pipe_surface_reference(&srcSurf, NULL);
+      pipe_surface_reference(&dstSurf, NULL);
+   }
+}
+
+
+void
+st_generate_mipmap(GLcontext *ctx, GLenum target,
+                   struct gl_texture_object *texObj)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_texture *pt = st_get_texobj_texture(texObj);
+   const uint baseLevel = texObj->BaseLevel;
+   uint lastLevel;
+   uint dstLevel;
+
+   if (!pt)
+      return;
+
+   lastLevel = pt->last_level;
+
+   if (!st_render_mipmap(st, target, pt, baseLevel, lastLevel)) {
+      fallback_generate_mipmap(ctx, target, texObj);
+   }
+
+   /* Fill in the Mesa gl_texture_image fields */
+   for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) {
+      const uint srcLevel = dstLevel - 1;
+      const struct gl_texture_image *srcImage
+         = _mesa_get_tex_image(ctx, texObj, target, srcLevel);
+      struct gl_texture_image *dstImage;
+      struct st_texture_image *stImage;
+      uint dstWidth = pt->width[dstLevel];
+      uint dstHeight = pt->height[dstLevel];
+      uint dstDepth = pt->depth[dstLevel];
+      uint border = srcImage->Border;
+
+      dstImage = _mesa_get_tex_image(ctx, texObj, target, dstLevel);
+      if (!dstImage) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "generating mipmaps");
+         return;
+      }
+
+      if (dstImage->ImageOffsets)
+         _mesa_free(dstImage->ImageOffsets);
+
+      /* Free old image data */
+      if (dstImage->Data)
+         ctx->Driver.FreeTexImageData(ctx, dstImage);
+
+      /* initialize new image */
+      _mesa_init_teximage_fields(ctx, target, dstImage, dstWidth, dstHeight,
+                                 dstDepth, border, srcImage->InternalFormat);
+
+      dstImage->TexFormat = srcImage->TexFormat;
+
+      stImage = (struct st_texture_image *) dstImage;
+      pipe_texture_reference(&stImage->pt, pt);
+   }
+}
diff --git a/src/mesa/state_tracker/st_gen_mipmap.h b/src/mesa/state_tracker/st_gen_mipmap.h
new file mode 100644
index 0000000000..00fbae9302
--- /dev/null
+++ b/src/mesa/state_tracker/st_gen_mipmap.h
@@ -0,0 +1,46 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_GEN_MIPMAP_H
+#define ST_GEN_MIPMAP_H
+
+
+extern void
+st_init_generate_mipmap(struct st_context *st);
+
+
+extern void
+st_destroy_generate_mipmap(struct st_context *st);
+
+
+extern void
+st_generate_mipmap(GLcontext *ctx, GLenum target,
+                   struct gl_texture_object *texObj);
+
+
+#endif /* ST_GEN_MIPMAP_H */
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
new file mode 100644
index 0000000000..8b9b18f94d
--- /dev/null
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -0,0 +1,1041 @@
+/**************************************************************************
+ * 
+ * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/*
+ * \author
+ * Michal Krol
+ */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_build.h"
+#include "tgsi/tgsi_util.h"
+#include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_sanity.h"
+#include "st_mesa_to_tgsi.h"
+#include "shader/prog_instruction.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+#include "pipe/p_debug.h"
+
+/*
+ * Map mesa register file to TGSI register file.
+ */
+static GLuint
+map_register_file(
+   enum register_file file,
+   GLuint index,
+   const GLuint immediateMapping[],
+   GLboolean indirectAccess )
+{
+   switch( file ) {
+   case PROGRAM_UNDEFINED:
+      return TGSI_FILE_NULL;
+   case PROGRAM_TEMPORARY:
+      return TGSI_FILE_TEMPORARY;
+   /*case PROGRAM_LOCAL_PARAM:*/
+   /*case PROGRAM_ENV_PARAM:*/
+
+      /* Because of the longstanding problem with mesa arb shaders
+       * where constants, immediates and state variables are all
+       * bundled together as PROGRAM_STATE_VAR, we can't tell from the
+       * mesa register file whether this is a CONSTANT or an
+       * IMMEDIATE, hence we need all the other information.
+       */
+   case PROGRAM_STATE_VAR:
+   case PROGRAM_NAMED_PARAM:
+   case PROGRAM_UNIFORM:
+      if (!indirectAccess && immediateMapping && immediateMapping[index] != ~0)
+         return TGSI_FILE_IMMEDIATE;
+      else
+	 return TGSI_FILE_CONSTANT;
+   case PROGRAM_CONSTANT:
+      if (indirectAccess)
+         return TGSI_FILE_CONSTANT;
+      assert(immediateMapping[index] != ~0);
+      return TGSI_FILE_IMMEDIATE;
+   case PROGRAM_INPUT:
+      return TGSI_FILE_INPUT;
+   case PROGRAM_OUTPUT:
+      return TGSI_FILE_OUTPUT;
+   case PROGRAM_ADDRESS:
+      return TGSI_FILE_ADDRESS;
+   default:
+      assert( 0 );
+      return TGSI_FILE_NULL;
+   }
+}
+
+/**
+ * Map mesa register file index to TGSI index.
+ * Take special care when processing input and output indices.
+ * \param file  one of TGSI_FILE_x
+ * \param index  the mesa register file index
+ * \param inputMapping  maps Mesa input indexes to TGSI input indexes
+ * \param outputMapping  maps Mesa output indexes to TGSI output indexes
+ */
+static GLuint
+map_register_file_index(
+   GLuint file,
+   GLuint index,
+   const GLuint inputMapping[],
+   const GLuint outputMapping[],
+   const GLuint immediateMapping[],
+   GLboolean indirectAccess )
+{
+   switch( file ) {
+   case TGSI_FILE_INPUT:
+      /* inputs are mapped according to the user-defined map */
+      return inputMapping[index];
+
+   case TGSI_FILE_OUTPUT:
+      return outputMapping[index];
+
+   case TGSI_FILE_IMMEDIATE:
+      if (indirectAccess)
+         return index;
+      assert(immediateMapping[index] != ~0);
+      return immediateMapping[index];
+
+   default:
+      return index;
+   }
+}
+
+/*
+ * Map mesa texture target to TGSI texture target.
+ */
+static GLuint
+map_texture_target(
+   GLuint textarget )
+{
+   switch( textarget ) {
+   case TEXTURE_1D_INDEX:
+      return TGSI_TEXTURE_1D;
+   case TEXTURE_2D_INDEX:
+      return TGSI_TEXTURE_2D;
+   case TEXTURE_3D_INDEX:
+      return TGSI_TEXTURE_3D;
+   case TEXTURE_CUBE_INDEX:
+      return TGSI_TEXTURE_CUBE;
+   case TEXTURE_RECT_INDEX:
+      return TGSI_TEXTURE_RECT;
+   default:
+      assert( 0 );
+   }
+
+   return TGSI_TEXTURE_1D;
+}
+
+static GLuint
+convert_sat(
+   GLuint sat )
+{
+   switch( sat ) {
+   case SATURATE_OFF:
+      return TGSI_SAT_NONE;
+   case SATURATE_ZERO_ONE:
+      return TGSI_SAT_ZERO_ONE;
+   case SATURATE_PLUS_MINUS_ONE:
+      return TGSI_SAT_MINUS_PLUS_ONE;
+   default:
+      assert( 0 );
+      return TGSI_SAT_NONE;
+   }
+}
+
+static GLuint
+convert_writemask(
+   GLuint writemask )
+{
+   assert( WRITEMASK_X == TGSI_WRITEMASK_X );
+   assert( WRITEMASK_Y == TGSI_WRITEMASK_Y );
+   assert( WRITEMASK_Z == TGSI_WRITEMASK_Z );
+   assert( WRITEMASK_W == TGSI_WRITEMASK_W );
+   assert( (writemask & ~TGSI_WRITEMASK_XYZW) == 0 );
+
+   return writemask;
+}
+
+static struct tgsi_full_immediate
+make_immediate(const float *value, uint size)
+{
+   struct tgsi_full_immediate imm;
+
+   imm = tgsi_default_full_immediate();
+   imm.Immediate.NrTokens += size;
+   imm.Immediate.DataType = TGSI_IMM_FLOAT32;
+   imm.u.Pointer = value;
+   return imm;
+}
+
+static void
+compile_instruction(
+   const struct prog_instruction *inst,
+   struct tgsi_full_instruction *fullinst,
+   const GLuint inputMapping[],
+   const GLuint outputMapping[],
+   const GLuint immediateMapping[],
+   GLboolean indirectAccess,
+   GLuint preamble_size,
+   GLuint processor,
+   GLboolean *insideSubroutine)
+{
+   GLuint i;
+   struct tgsi_full_dst_register *fulldst;
+   struct tgsi_full_src_register *fullsrc;
+
+   *fullinst = tgsi_default_full_instruction();
+
+   fullinst->Instruction.Saturate = convert_sat( inst->SaturateMode );
+   fullinst->Instruction.NumDstRegs = _mesa_num_inst_dst_regs( inst->Opcode );
+   fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs( inst->Opcode );
+
+   fulldst = &fullinst->FullDstRegisters[0];
+   fulldst->DstRegister.File = map_register_file( inst->DstReg.File, 0, NULL, GL_FALSE );
+   fulldst->DstRegister.Index = map_register_file_index(
+      fulldst->DstRegister.File,
+      inst->DstReg.Index,
+      inputMapping,
+      outputMapping,
+      NULL,
+      GL_FALSE );
+   fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask );
+
+   for (i = 0; i < fullinst->Instruction.NumSrcRegs; i++) {
+      GLuint j;
+
+      fullsrc = &fullinst->FullSrcRegisters[i];
+      fullsrc->SrcRegister.File = map_register_file(
+         inst->SrcReg[i].File,
+         inst->SrcReg[i].Index,
+         immediateMapping,
+         indirectAccess );
+      fullsrc->SrcRegister.Index = map_register_file_index(
+         fullsrc->SrcRegister.File,
+         inst->SrcReg[i].Index,
+         inputMapping,
+         outputMapping,
+         immediateMapping,
+         indirectAccess );
+
+
+      /* swizzle (ext swizzle also depends on negation) */
+      {
+         GLuint swz[4];
+         GLboolean extended = (inst->SrcReg[i].NegateBase != NEGATE_NONE &&
+                               inst->SrcReg[i].NegateBase != NEGATE_XYZW);
+         for( j = 0; j < 4; j++ ) {
+            swz[j] = GET_SWZ( inst->SrcReg[i].Swizzle, j );
+            if (swz[j] > SWIZZLE_W)
+               extended = GL_TRUE;
+         }
+         if (extended) {
+            for (j = 0; j < 4; j++) {
+               tgsi_util_set_src_register_extswizzle(&fullsrc->SrcRegisterExtSwz,
+                                                     swz[j], j);
+            }
+         }
+         else {
+            for (j = 0; j < 4; j++) {
+               tgsi_util_set_src_register_swizzle(&fullsrc->SrcRegister,
+                                                  swz[j], j);
+            }
+         }
+      }
+
+      if( inst->SrcReg[i].NegateBase == NEGATE_XYZW ) {
+         fullsrc->SrcRegister.Negate = 1;
+      }
+      else if( inst->SrcReg[i].NegateBase != NEGATE_NONE ) {
+         if( inst->SrcReg[i].NegateBase & NEGATE_X ) {
+            fullsrc->SrcRegisterExtSwz.NegateX = 1;
+         }
+         if( inst->SrcReg[i].NegateBase & NEGATE_Y ) {
+            fullsrc->SrcRegisterExtSwz.NegateY = 1;
+         }
+         if( inst->SrcReg[i].NegateBase & NEGATE_Z ) {
+            fullsrc->SrcRegisterExtSwz.NegateZ = 1;
+         }
+         if( inst->SrcReg[i].NegateBase & NEGATE_W ) {
+            fullsrc->SrcRegisterExtSwz.NegateW = 1;
+         }
+      }
+
+      if( inst->SrcReg[i].Abs ) {
+         fullsrc->SrcRegisterExtMod.Absolute = 1;
+      }
+
+      if( inst->SrcReg[i].NegateAbs ) {
+         fullsrc->SrcRegisterExtMod.Negate = 1;
+      }
+
+      if( inst->SrcReg[i].RelAddr ) {
+         fullsrc->SrcRegister.Indirect = 1;
+
+         fullsrc->SrcRegisterInd.File = TGSI_FILE_ADDRESS;
+         fullsrc->SrcRegisterInd.Index = 0;
+      }
+   }
+
+   switch( inst->Opcode ) {
+   case OPCODE_ARL:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ARL;
+      break;
+   case OPCODE_ABS:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ABS;
+      break;
+   case OPCODE_ADD:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ADD;
+      break;
+   case OPCODE_BGNLOOP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2;
+      fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
+      break;
+   case OPCODE_BGNSUB:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB;
+      *insideSubroutine = GL_TRUE;
+      break;
+   case OPCODE_BRA:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_BRA;
+      break;
+   case OPCODE_BRK:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_BRK;
+      break;
+   case OPCODE_CAL:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_CAL;
+      fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
+      break;
+   case OPCODE_CMP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_CMP;
+      break;
+   case OPCODE_CONT:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_CONT;
+      break;
+   case OPCODE_COS:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_COS;
+      break;
+   case OPCODE_DDX:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DDX;
+      break;
+   case OPCODE_DDY:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DDY;
+      break;
+   case OPCODE_DP2:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DP2;
+      break;
+   case OPCODE_DP2A:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DP2A;
+      break;
+   case OPCODE_DP3:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DP3;
+      break;
+   case OPCODE_DP4:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DP4;
+      break;
+   case OPCODE_DPH:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DPH;
+      break;
+   case OPCODE_DST:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_DST;
+      break;
+   case OPCODE_ELSE:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ELSE;
+      fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
+      break;
+   case OPCODE_ENDIF:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ENDIF;
+      break;
+   case OPCODE_ENDLOOP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2;
+      fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
+      break;
+   case OPCODE_ENDSUB:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB;
+      *insideSubroutine = GL_FALSE;
+      break;
+   case OPCODE_EX2:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_EX2;
+      break;
+   case OPCODE_EXP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_EXP;
+      break;
+   case OPCODE_FLR:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_FLR;
+      break;
+   case OPCODE_FRC:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_FRC;
+      break;
+   case OPCODE_IF:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_IF;
+      fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size;
+      break;
+   case OPCODE_TRUNC:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TRUNC;
+      break;
+   case OPCODE_KIL:
+      /* conditional */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_KIL;
+      break;
+   case OPCODE_KIL_NV:
+      /* predicated */
+      assert(inst->DstReg.CondMask == COND_TR);
+      fullinst->Instruction.Opcode = TGSI_OPCODE_KILP;
+      break;
+   case OPCODE_LG2:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_LG2;
+      break;
+   case OPCODE_LOG:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_LOG;
+      break;
+   case OPCODE_LIT:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_LIT;
+      break;
+   case OPCODE_LRP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_LRP;
+      break;
+   case OPCODE_MAD:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_MAD;
+      break;
+   case OPCODE_MAX:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_MAX;
+      break;
+   case OPCODE_MIN:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_MIN;
+      break;
+   case OPCODE_MOV:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_MOV;
+      break;
+   case OPCODE_MUL:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_MUL;
+      break;
+   case OPCODE_NOISE1:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE1;
+      break;
+   case OPCODE_NOISE2:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE2;
+      break;
+   case OPCODE_NOISE3:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE3;
+      break;
+   case OPCODE_NOISE4:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE4;
+      break;
+   case OPCODE_NOP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NOP;
+      break;
+   case OPCODE_NRM3:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NRM;
+      break;
+   case OPCODE_NRM4:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_NRM4;
+      break;
+   case OPCODE_POW:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_POW;
+      break;
+   case OPCODE_RCP:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_RCP;
+      break;
+   case OPCODE_RET:
+      /* If RET is used inside main (not a real subroutine) we may want
+       * to execute END instead of RET.  TBD...
+       */
+      if (1 /*  *insideSubroutine */) {
+         fullinst->Instruction.Opcode = TGSI_OPCODE_RET;
+      }
+      else {
+         /* inside main() pseudo-function */
+         fullinst->Instruction.Opcode = TGSI_OPCODE_END;
+      }
+      break;
+   case OPCODE_RSQ:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_RSQ;
+
+      /* KW: Don't do this here.  If particular hardware needs to do
+       * this, can do so in the driver..
+       */
+#if 0
+       tgsi_util_set_full_src_register_sign_mode(
+          &fullinst->FullSrcRegisters[0],
+          TGSI_UTIL_SIGN_CLEAR ); 
+#endif
+      break;
+   case OPCODE_SCS:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SCS;
+      fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XY;
+      break;
+   case OPCODE_SEQ:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SEQ;
+      break;
+   case OPCODE_SGE:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SGE;
+      break;
+   case OPCODE_SGT:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SGT;
+      break;
+   case OPCODE_SIN:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SIN;
+      break;
+   case OPCODE_SLE:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SLE;
+      break;
+   case OPCODE_SLT:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SLT;
+      break;
+   case OPCODE_SNE:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SNE;
+      break;
+   case OPCODE_SSG:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SSG;
+      break;
+   case OPCODE_SUB:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SUB;
+      break;
+   case OPCODE_SWZ:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_SWZ;
+      break;
+   case OPCODE_TEX:
+      /* ordinary texture lookup */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TEX;
+      fullinst->Instruction.NumSrcRegs = 2;
+      fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget );
+      fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit;
+      break;
+   case OPCODE_TXB:
+      /* texture lookup with LOD bias */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TXB;
+      fullinst->Instruction.NumSrcRegs = 2;
+      fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget );
+      fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit;
+      break;
+   case OPCODE_TXD:
+      /* texture lookup with explicit partial derivatives */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TXD;
+      fullinst->Instruction.NumSrcRegs = 4;
+      fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget );
+      /* src[0] = coord, src[1] = d[strq]/dx, src[2] = d[strq]/dy */
+      fullinst->FullSrcRegisters[3].SrcRegister.File = TGSI_FILE_SAMPLER;
+      fullinst->FullSrcRegisters[3].SrcRegister.Index = inst->TexSrcUnit;
+      break;
+   case OPCODE_TXL:
+      /* texture lookup with explicit LOD */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TXL;
+      fullinst->Instruction.NumSrcRegs = 2;
+      fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget );
+      fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit;
+      break;
+   case OPCODE_TXP:
+      /* texture lookup with divide by Q component */
+      /* convert to TEX w/ special flag for division */
+      fullinst->Instruction.Opcode = TGSI_OPCODE_TXP;
+      fullinst->Instruction.NumSrcRegs = 2;
+      fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget );
+      fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit;
+      break;
+   case OPCODE_XPD:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_XPD;
+      fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ;
+      break;
+   case OPCODE_END:
+      fullinst->Instruction.Opcode = TGSI_OPCODE_END;
+      break;
+   default:
+      assert( 0 );
+   }
+}
+
+/**
+ * \param usage_mask  bitfield of TGSI_WRITEMASK_{XYZW} tokens
+ */
+static struct tgsi_full_declaration
+make_input_decl(
+   GLuint index,
+   GLboolean interpolate_info,
+   GLuint interpolate,
+   GLuint usage_mask,
+   GLboolean semantic_info,
+   GLuint semantic_name,
+   GLbitfield semantic_index,
+   GLbitfield input_flags)
+{
+   struct tgsi_full_declaration decl;
+
+   assert(semantic_name < TGSI_SEMANTIC_COUNT);
+
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_INPUT;
+   decl.Declaration.UsageMask = usage_mask;
+   decl.Declaration.Semantic = semantic_info;
+   decl.DeclarationRange.First = index;
+   decl.DeclarationRange.Last = index;
+   if (semantic_info) {
+      decl.Semantic.SemanticName = semantic_name;
+      decl.Semantic.SemanticIndex = semantic_index;
+   }
+   if (interpolate_info) {
+      decl.Declaration.Interpolate = interpolate;
+   }
+   if (input_flags & PROG_PARAM_BIT_CENTROID)
+      decl.Declaration.Centroid = 1;
+   if (input_flags & PROG_PARAM_BIT_INVARIANT)
+      decl.Declaration.Invariant = 1;
+
+   return decl;
+}
+
+/**
+ * \param usage_mask  bitfield of TGSI_WRITEMASK_{XYZW} tokens
+ */
+static struct tgsi_full_declaration
+make_output_decl(
+   GLuint index,
+   GLuint semantic_name,
+   GLuint semantic_index,
+   GLuint usage_mask,
+   GLbitfield output_flags)
+{
+   struct tgsi_full_declaration decl;
+
+   assert(semantic_name < TGSI_SEMANTIC_COUNT);
+
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_OUTPUT;
+   decl.Declaration.UsageMask = usage_mask;
+   decl.Declaration.Semantic = 1;
+   decl.DeclarationRange.First = index;
+   decl.DeclarationRange.Last = index;
+   decl.Semantic.SemanticName = semantic_name;
+   decl.Semantic.SemanticIndex = semantic_index;
+   if (output_flags & PROG_PARAM_BIT_CENTROID)
+      decl.Declaration.Centroid = 1;
+   if (output_flags & PROG_PARAM_BIT_INVARIANT)
+      decl.Declaration.Invariant = 1;
+
+   return decl;
+}
+
+
+static struct tgsi_full_declaration
+make_temp_decl(
+   GLuint start_index,
+   GLuint end_index )
+{
+   struct tgsi_full_declaration decl;
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_TEMPORARY;
+   decl.DeclarationRange.First = start_index;
+   decl.DeclarationRange.Last = end_index;
+   return decl;
+}
+
+static struct tgsi_full_declaration
+make_addr_decl(
+   GLuint start_index,
+   GLuint end_index )
+{
+   struct tgsi_full_declaration decl;
+
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_ADDRESS;
+   decl.DeclarationRange.First = start_index;
+   decl.DeclarationRange.Last = end_index;
+   return decl;
+}
+
+static struct tgsi_full_declaration
+make_sampler_decl(GLuint index)
+{
+   struct tgsi_full_declaration decl;
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_SAMPLER;
+   decl.DeclarationRange.First = index;
+   decl.DeclarationRange.Last = index;
+   return decl;
+}
+
+/** Reference into a constant buffer */
+static struct tgsi_full_declaration
+make_constant_decl(GLuint first, GLuint last)
+{
+   struct tgsi_full_declaration decl;
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_CONSTANT;
+   decl.DeclarationRange.First = first;
+   decl.DeclarationRange.Last = last;
+   return decl;
+}
+
+
+
+/**
+ * Find the temporaries which are used in the given program.
+ */
+static void
+find_temporaries(const struct gl_program *program,
+                 GLboolean tempsUsed[MAX_PROGRAM_TEMPS])
+{
+   GLuint i, j;
+
+   for (i = 0; i < MAX_PROGRAM_TEMPS; i++)
+      tempsUsed[i] = GL_FALSE;
+
+   for (i = 0; i < program->NumInstructions; i++) {
+      const struct prog_instruction *inst = program->Instructions + i;
+      const GLuint n = _mesa_num_inst_src_regs( inst->Opcode );
+      for (j = 0; j < n; j++) {
+         if (inst->SrcReg[j].File == PROGRAM_TEMPORARY)
+            tempsUsed[inst->SrcReg[j].Index] = GL_TRUE;
+         if (inst->DstReg.File == PROGRAM_TEMPORARY)
+            tempsUsed[inst->DstReg.Index] = GL_TRUE;
+      }
+   }
+}
+
+
+
+
+/**
+ * Translate Mesa program to TGSI format.
+ * \param program  the program to translate
+ * \param numInputs  number of input registers used
+ * \param inputMapping  maps Mesa fragment program inputs to TGSI generic
+ *                      input indexes
+ * \param inputSemanticName  the TGSI_SEMANTIC flag for each input
+ * \param inputSemanticIndex  the semantic index (ex: which texcoord) for each input
+ * \param interpMode  the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input
+
+ * \param numOutputs  number of output registers used
+ * \param outputMapping  maps Mesa fragment program outputs to TGSI
+ *                       generic outputs
+ * \param outputSemanticName  the TGSI_SEMANTIC flag for each output
+ * \param outputSemanticIndex  the semantic index (ex: which texcoord) for each output
+ * \param tokens  array to store translated tokens in
+ * \param maxTokens  size of the tokens array
+ *
+ * \return number of tokens placed in 'tokens' buffer, or zero if error
+ */
+GLuint
+st_translate_mesa_program(
+   GLcontext *ctx,
+   uint procType,
+   const struct gl_program *program,
+   GLuint numInputs,
+   const GLuint inputMapping[],
+   const ubyte inputSemanticName[],
+   const ubyte inputSemanticIndex[],
+   const GLuint interpMode[],
+   const GLbitfield inputFlags[],
+   GLuint numOutputs,
+   const GLuint outputMapping[],
+   const ubyte outputSemanticName[],
+   const ubyte outputSemanticIndex[],
+   const GLbitfield outputFlags[],
+   struct tgsi_token *tokens,
+   GLuint maxTokens )
+{
+   GLuint i;
+   GLuint ti;  /* token index */
+   struct tgsi_header *header;
+   struct tgsi_processor *processor;
+   struct tgsi_full_instruction fullinst;
+   GLuint preamble_size = 0;
+   GLuint immediates[1000];
+   GLuint numImmediates = 0;
+   GLboolean insideSubroutine = GL_FALSE;
+   GLboolean indirectAccess = GL_FALSE;
+
+   assert(procType == TGSI_PROCESSOR_FRAGMENT ||
+          procType == TGSI_PROCESSOR_VERTEX);
+
+   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
+
+   header = (struct tgsi_header *) &tokens[1];
+   *header = tgsi_build_header();
+
+   processor = (struct tgsi_processor *) &tokens[2];
+   *processor = tgsi_build_processor( procType, header );
+
+   ti = 3;
+
+   /*
+    * Declare input attributes.
+    */
+   if (procType == TGSI_PROCESSOR_FRAGMENT) {
+      for (i = 0; i < numInputs; i++) {
+         struct tgsi_full_declaration fulldecl;
+         fulldecl = make_input_decl(i,
+                                    GL_TRUE, interpMode[i],
+                                    TGSI_WRITEMASK_XYZW,
+                                    GL_TRUE, inputSemanticName[i],
+                                    inputSemanticIndex[i],
+                                    inputFlags[i]);
+         ti += tgsi_build_full_declaration(&fulldecl,
+                                           &tokens[ti],
+                                           header,
+                                           maxTokens - ti );
+      }
+   }
+   else {
+      /* vertex prog */
+      /* XXX: this could probaby be merged with the clause above.
+       * the only difference is the semantic tags.
+       */
+      for (i = 0; i < numInputs; i++) {
+         struct tgsi_full_declaration fulldecl;
+         fulldecl = make_input_decl(i,
+                                    GL_FALSE, 0,
+                                    TGSI_WRITEMASK_XYZW,
+                                    GL_FALSE, 0, 0,
+                                    inputFlags[i]);
+         ti += tgsi_build_full_declaration(&fulldecl,
+                                           &tokens[ti],
+                                           header,
+                                           maxTokens - ti );
+      }
+   }
+
+   /*
+    * Declare output attributes.
+    */
+   if (procType == TGSI_PROCESSOR_FRAGMENT) {
+      for (i = 0; i < numOutputs; i++) {
+         struct tgsi_full_declaration fulldecl;
+         switch (outputSemanticName[i]) {
+         case TGSI_SEMANTIC_POSITION:
+            fulldecl = make_output_decl(i,
+                                        TGSI_SEMANTIC_POSITION, /* Z / Depth */
+                                        outputSemanticIndex[i],
+                                        TGSI_WRITEMASK_Z,
+                                        outputFlags[i]);
+            break;
+         case TGSI_SEMANTIC_COLOR:
+            fulldecl = make_output_decl(i,
+                                        TGSI_SEMANTIC_COLOR,
+                                        outputSemanticIndex[i],
+                                        TGSI_WRITEMASK_XYZW,
+                                        outputFlags[i]);
+            break;
+         default:
+            assert(0);
+            return 0;
+         }
+         ti += tgsi_build_full_declaration(&fulldecl,
+                                           &tokens[ti],
+                                           header,
+                                           maxTokens - ti );
+      }
+   }
+   else {
+      /* vertex prog */
+      for (i = 0; i < numOutputs; i++) {
+         struct tgsi_full_declaration fulldecl;
+         fulldecl = make_output_decl(i,
+                                     outputSemanticName[i],
+                                     outputSemanticIndex[i],
+                                     TGSI_WRITEMASK_XYZW,
+                                     outputFlags[i]);
+         ti += tgsi_build_full_declaration(&fulldecl,
+                                           &tokens[ti],
+                                           header,
+                                           maxTokens - ti );
+      }
+   }
+
+   /* temporary decls */
+   {
+      GLboolean tempsUsed[MAX_PROGRAM_TEMPS + 1];
+      GLboolean inside_range = GL_FALSE;
+      GLuint start_range = 0;
+
+      find_temporaries(program, tempsUsed);
+      tempsUsed[MAX_PROGRAM_TEMPS] = GL_FALSE;
+      for (i = 0; i < MAX_PROGRAM_TEMPS + 1; i++) {
+         if (tempsUsed[i] && !inside_range) {
+            inside_range = GL_TRUE;
+            start_range = i;
+         }
+         else if (!tempsUsed[i] && inside_range) {
+            struct tgsi_full_declaration fulldecl;
+
+            inside_range = GL_FALSE;
+            fulldecl = make_temp_decl( start_range, i - 1 );
+            ti += tgsi_build_full_declaration(
+               &fulldecl,
+               &tokens[ti],
+               header,
+               maxTokens - ti );
+         }
+      }
+   }
+
+   /* Declare address register.
+   */
+   if (program->NumAddressRegs > 0) {
+      struct tgsi_full_declaration fulldecl;
+
+      assert( program->NumAddressRegs == 1 );
+
+      fulldecl = make_addr_decl( 0, 0 );
+      ti += tgsi_build_full_declaration(
+         &fulldecl,
+         &tokens[ti],
+         header,
+         maxTokens - ti );
+
+      indirectAccess = GL_TRUE;
+   }
+
+   /* immediates/literals */
+   memset(immediates, ~0, sizeof(immediates));
+
+   /* Emit immediates only when there is no address register in use.
+    * FIXME: Be smarter and recognize param arrays -- indirect addressing is
+    *        only valid within the referenced array.
+    */
+   if (program->Parameters && !indirectAccess) {
+      for (i = 0; i < program->Parameters->NumParameters; i++) {
+         if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) {
+            struct tgsi_full_immediate fullimm;
+
+            fullimm = make_immediate( program->Parameters->ParameterValues[i], 4 );
+            ti += tgsi_build_full_immediate(
+               &fullimm,
+               &tokens[ti],
+               header,
+               maxTokens - ti );
+            immediates[i] = numImmediates;
+            numImmediates++;
+         }
+      }
+   }
+
+   /* constant buffer refs */
+   if (program->Parameters) {
+      GLint start = -1, end = -1;
+
+      for (i = 0; i < program->Parameters->NumParameters; i++) {
+         GLboolean emit = (i == program->Parameters->NumParameters - 1);
+         GLboolean matches;
+
+         switch (program->Parameters->Parameters[i].Type) {
+         case PROGRAM_ENV_PARAM:
+         case PROGRAM_STATE_VAR:
+         case PROGRAM_NAMED_PARAM:
+         case PROGRAM_UNIFORM:
+            matches = GL_TRUE;
+            break;
+         case PROGRAM_CONSTANT:
+            matches = indirectAccess;
+            break;
+         default:
+            matches = GL_FALSE;
+         }
+
+         if (matches) {
+            if (start == -1) {
+               /* begin a sequence */
+               start = i;
+               end = i;
+            }
+            else {
+               /* continue sequence */
+               end = i;
+            }
+         }
+         else {
+            if (start != -1) {
+               /* end of sequence */
+               emit = GL_TRUE;
+            }
+         }
+
+         if (emit && start >= 0) {
+            struct tgsi_full_declaration fulldecl;
+
+            fulldecl = make_constant_decl( start, end );
+            ti += tgsi_build_full_declaration(
+               &fulldecl,
+               &tokens[ti],
+               header,
+               maxTokens - ti );
+            start = end = -1;
+         }
+      }
+   }
+
+   /* texture samplers */
+   for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
+      if (program->SamplersUsed & (1 << i)) {
+         struct tgsi_full_declaration fulldecl;
+
+         fulldecl = make_sampler_decl( i );
+         ti += tgsi_build_full_declaration(
+            &fulldecl,
+            &tokens[ti],
+            header,
+            maxTokens - ti );
+      }
+   }
+
+   for (i = 0; i < program->NumInstructions; i++) {
+      compile_instruction(
+         &program->Instructions[i],
+         &fullinst,
+         inputMapping,
+         outputMapping,
+         immediates,
+         indirectAccess,
+         preamble_size,
+         procType,
+         &insideSubroutine );
+
+      ti += tgsi_build_full_instruction(
+         &fullinst,
+         &tokens[ti],
+         header,
+         maxTokens - ti );
+   }
+
+#if DEBUG
+   if(!tgsi_sanity_check(tokens)) {
+      debug_printf("Due to sanity check failure(s) above the following shader program is invalid:\n");
+      debug_printf("\nOriginal program:\n%s", program->String);
+      debug_printf("\nMesa program:\n");
+      _mesa_print_program(program);
+      debug_printf("\nTGSI program:\n");
+      tgsi_dump(tokens, 0);
+      assert(0);
+   }
+#endif
+
+   return ti;
+}
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h
new file mode 100644
index 0000000000..b465b3bddc
--- /dev/null
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h
@@ -0,0 +1,67 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#ifndef ST_MESA_TO_TGSI_H
+#define ST_MESA_TO_TGSI_H
+
+#include "main/mtypes.h"
+
+
+#if defined __cplusplus
+extern "C" {
+#endif
+
+struct tgsi_token;
+struct gl_program;
+
+GLuint
+st_translate_mesa_program(
+   GLcontext *ctx,
+   uint procType,
+   const struct gl_program *program,
+   GLuint numInputs,
+   const GLuint inputMapping[],
+   const ubyte inputSemanticName[],
+   const ubyte inputSemanticIndex[],
+   const GLuint interpMode[],
+   const GLbitfield inputFlags[],
+   GLuint numOutputs,
+   const GLuint outputMapping[],
+   const ubyte outputSemanticName[],
+   const ubyte outputSemanticIndex[],
+   const GLbitfield outputFlags[],
+   struct tgsi_token *tokens,
+   GLuint maxTokens );
+
+
+#if defined __cplusplus
+} /* extern "C" */
+#endif
+
+#endif /* ST_MESA_TO_TGSI_H */
+
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
new file mode 100644
index 0000000000..442eeed147
--- /dev/null
+++ b/src/mesa/state_tracker/st_program.c
@@ -0,0 +1,556 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Brian Paul
+  */
+
+
+#include "main/imports.h"
+#include "main/mtypes.h"
+#include "shader/prog_print.h"
+#include "shader/programopt.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+#include "draw/draw_context.h"
+#include "tgsi/tgsi_dump.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_program.h"
+#include "st_mesa_to_tgsi.h"
+#include "cso_cache/cso_context.h"
+
+
+#define ST_MAX_SHADER_TOKENS 4096
+
+
+#define TGSI_DEBUG 0
+
+
+/** XXX we should use the version of this from u_memory.h but including
+ * that header causes symbol collisions.
+ */
+static INLINE void *
+mem_dup(const void *src, uint size)
+{
+   void *dup = MALLOC(size);
+   if (dup)
+      memcpy(dup, src, size);
+   return dup;
+}
+
+
+
+/**
+ * Translate a Mesa vertex shader into a TGSI shader.
+ * \param outputMapping  to map vertex program output registers (VERT_RESULT_x)
+ *       to TGSI output slots
+ * \param tokensOut  destination for TGSI tokens
+ * \return  pointer to cached pipe_shader object.
+ */
+void
+st_translate_vertex_program(struct st_context *st,
+                            struct st_vertex_program *stvp,
+                            const GLuint outputMapping[],
+                            const ubyte *outputSemanticName,
+                            const ubyte *outputSemanticIndex)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct tgsi_token tokens[ST_MAX_SHADER_TOKENS];
+   GLuint defaultOutputMapping[VERT_RESULT_MAX];
+   struct pipe_shader_state vs;
+   GLuint attr, i;
+   GLuint num_generic = 0;
+   GLuint num_tokens;
+
+   ubyte vs_input_semantic_name[PIPE_MAX_SHADER_INPUTS];
+   ubyte vs_input_semantic_index[PIPE_MAX_SHADER_INPUTS];
+   uint vs_num_inputs = 0;
+
+   ubyte vs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS];
+   ubyte vs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS];
+   uint vs_num_outputs = 0;
+
+   GLbitfield input_flags[MAX_PROGRAM_INPUTS];
+   GLbitfield output_flags[MAX_PROGRAM_OUTPUTS];
+
+   memset(&vs, 0, sizeof(vs));
+   memset(input_flags, 0, sizeof(input_flags));
+   memset(output_flags, 0, sizeof(output_flags));
+
+   if (stvp->Base.IsPositionInvariant)
+      _mesa_insert_mvp_code(st->ctx, &stvp->Base);
+
+   /*
+    * Determine number of inputs, the mappings between VERT_ATTRIB_x
+    * and TGSI generic input indexes, plus input attrib semantic info.
+    */
+   for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) {
+      if (stvp->Base.Base.InputsRead & (1 << attr)) {
+         const GLuint slot = vs_num_inputs;
+
+         vs_num_inputs++;
+
+         stvp->input_to_index[attr] = slot;
+         stvp->index_to_input[slot] = attr;
+
+         switch (attr) {
+         case VERT_ATTRIB_POS:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_POSITION;
+            vs_input_semantic_index[slot] = 0;
+            break;
+         case VERT_ATTRIB_WEIGHT:
+            /* fall-through */
+         case VERT_ATTRIB_NORMAL:
+            /* just label as a generic */
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            vs_input_semantic_index[slot] = 0;
+            break;
+         case VERT_ATTRIB_COLOR0:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            vs_input_semantic_index[slot] = 0;
+            break;
+         case VERT_ATTRIB_COLOR1:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            vs_input_semantic_index[slot] = 1;
+            break;
+         case VERT_ATTRIB_FOG:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_FOG;
+            vs_input_semantic_index[slot] = 0;
+            break;
+         case VERT_ATTRIB_POINT_SIZE:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_PSIZE;
+            vs_input_semantic_index[slot] = 0;
+            break;
+         case VERT_ATTRIB_TEX0:
+         case VERT_ATTRIB_TEX1:
+         case VERT_ATTRIB_TEX2:
+         case VERT_ATTRIB_TEX3:
+         case VERT_ATTRIB_TEX4:
+         case VERT_ATTRIB_TEX5:
+         case VERT_ATTRIB_TEX6:
+         case VERT_ATTRIB_TEX7:
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            vs_input_semantic_index[slot] = num_generic++;
+            break;
+         case VERT_ATTRIB_GENERIC0:
+         case VERT_ATTRIB_GENERIC1:
+         case VERT_ATTRIB_GENERIC2:
+         case VERT_ATTRIB_GENERIC3:
+         case VERT_ATTRIB_GENERIC4:
+         case VERT_ATTRIB_GENERIC5:
+         case VERT_ATTRIB_GENERIC6:
+         case VERT_ATTRIB_GENERIC7:
+            assert(attr < VERT_ATTRIB_MAX);
+            vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            vs_input_semantic_index[slot] = num_generic++;
+            break;
+         default:
+            assert(0);
+         }
+
+         input_flags[slot] = stvp->Base.Base.InputFlags[attr];
+      }
+   }
+
+#if 0
+   if (outputMapping && outputSemanticName) {
+      printf("VERT_RESULT  written  out_slot  semantic_name  semantic_index\n");
+      for (attr = 0; attr < VERT_RESULT_MAX; attr++) {
+         printf("    %-2d          %c       %3d          %2d              %2d\n",
+                attr, 
+                ((stvp->Base.Base.OutputsWritten & (1 << attr)) ? 'Y' : ' '),
+                outputMapping[attr],
+                outputSemanticName[attr],
+                outputSemanticIndex[attr]);
+      }
+   }
+#endif
+
+   /* initialize output semantics to defaults */
+   for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) {
+      vs_output_semantic_name[i] = TGSI_SEMANTIC_GENERIC;
+      vs_output_semantic_index[i] = 0;
+      output_flags[i] = 0x0;
+   }
+
+   num_generic = 0;
+   /*
+    * Determine number of outputs, the (default) output register
+    * mapping and the semantic information for each output.
+    */
+   for (attr = 0; attr < VERT_RESULT_MAX; attr++) {
+      if (stvp->Base.Base.OutputsWritten & (1 << attr)) {
+         GLuint slot;
+
+         /* XXX
+          * Pass in the fragment program's input's semantic info.
+          * Use the generic semantic indexes from there, instead of
+          * guessing below.
+          */
+
+         if (outputMapping) {
+            slot = outputMapping[attr];
+            assert(slot != ~0);
+         }
+         else {
+            slot = vs_num_outputs;
+            vs_num_outputs++;
+            defaultOutputMapping[attr] = slot;
+         }
+
+         switch (attr) {
+         case VERT_RESULT_HPOS:
+            assert(slot == 0);
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_POSITION;
+            vs_output_semantic_index[slot] = 0;
+            break;
+         case VERT_RESULT_COL0:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            vs_output_semantic_index[slot] = 0;
+            break;
+         case VERT_RESULT_COL1:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            vs_output_semantic_index[slot] = 1;
+            break;
+         case VERT_RESULT_BFC0:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR;
+            vs_output_semantic_index[slot] = 0;
+            break;
+         case VERT_RESULT_BFC1:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR;
+            vs_output_semantic_index[slot] = 1;
+            break;
+         case VERT_RESULT_FOGC:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_FOG;
+            vs_output_semantic_index[slot] = 0;
+            break;
+         case VERT_RESULT_PSIZ:
+            vs_output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE;
+            vs_output_semantic_index[slot] = 0;
+            break;
+         case VERT_RESULT_EDGE:
+            assert(0);
+            break;
+         case VERT_RESULT_TEX0:
+         case VERT_RESULT_TEX1:
+         case VERT_RESULT_TEX2:
+         case VERT_RESULT_TEX3:
+         case VERT_RESULT_TEX4:
+         case VERT_RESULT_TEX5:
+         case VERT_RESULT_TEX6:
+         case VERT_RESULT_TEX7:
+            /* fall-through */
+         case VERT_RESULT_VAR0:
+            /* fall-through */
+         default:
+            if (outputSemanticName) {
+               /* use provided semantic into */
+               assert(outputSemanticName[attr] != TGSI_SEMANTIC_COUNT);
+               vs_output_semantic_name[slot] = outputSemanticName[attr];
+               vs_output_semantic_index[slot] = outputSemanticIndex[attr];
+            }
+            else {
+               /* use default semantic info */
+               vs_output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+               vs_output_semantic_index[slot] = num_generic++;
+            }
+         }
+
+         output_flags[slot] = stvp->Base.Base.OutputFlags[attr];
+      }
+   }
+
+   assert(vs_output_semantic_name[0] == TGSI_SEMANTIC_POSITION);
+
+
+   if (outputMapping) {
+      /* find max output slot referenced to compute vs_num_outputs */
+      GLuint maxSlot = 0;
+      for (attr = 0; attr < VERT_RESULT_MAX; attr++) {
+         if (outputMapping[attr] != ~0 && outputMapping[attr] > maxSlot)
+            maxSlot = outputMapping[attr];
+      }
+      vs_num_outputs = maxSlot + 1;
+   }
+   else {
+      outputMapping = defaultOutputMapping;
+   }
+
+   /* free old shader state, if any */
+   if (stvp->state.tokens) {
+      FREE((void *) stvp->state.tokens);
+      stvp->state.tokens = NULL;
+   }
+   if (stvp->driver_shader) {
+      cso_delete_vertex_shader(st->cso_context, stvp->driver_shader);
+      stvp->driver_shader = NULL;
+   }
+
+   /* XXX: fix static allocation of tokens:
+    */
+   num_tokens = st_translate_mesa_program(st->ctx,
+                                          TGSI_PROCESSOR_VERTEX,
+                                          &stvp->Base.Base,
+                                          /* inputs */
+                                          vs_num_inputs,
+                                          stvp->input_to_index,
+                                          vs_input_semantic_name,
+                                          vs_input_semantic_index,
+                                          NULL,
+                                          input_flags,
+                                          /* outputs */
+                                          vs_num_outputs,
+                                          outputMapping,
+                                          vs_output_semantic_name,
+                                          vs_output_semantic_index,
+                                          output_flags,
+                                          /* tokenized result */
+                                          tokens, ST_MAX_SHADER_TOKENS);
+
+   assert(num_tokens < ST_MAX_SHADER_TOKENS);
+
+   vs.tokens = (struct tgsi_token *)
+      mem_dup(tokens, num_tokens * sizeof(tokens[0]));
+
+   stvp->num_inputs = vs_num_inputs;
+   stvp->state = vs; /* struct copy */
+   stvp->driver_shader = pipe->create_vs_state(pipe, &vs);
+
+   if (0)
+      _mesa_print_program(&stvp->Base.Base);
+
+   if (TGSI_DEBUG)
+      tgsi_dump( vs.tokens, 0 );
+}
+
+
+
+/**
+ * Translate a Mesa fragment shader into a TGSI shader.
+ * \param inputMapping  to map fragment program input registers to TGSI
+ *                      input slots
+ * \param tokensOut  destination for TGSI tokens
+ * \return  pointer to cached pipe_shader object.
+ */
+void
+st_translate_fragment_program(struct st_context *st,
+                              struct st_fragment_program *stfp,
+                              const GLuint inputMapping[])
+{
+   struct pipe_context *pipe = st->pipe;
+   struct tgsi_token tokens[ST_MAX_SHADER_TOKENS];
+   GLuint outputMapping[FRAG_RESULT_MAX];
+   GLuint defaultInputMapping[FRAG_ATTRIB_MAX];
+   struct pipe_shader_state fs;
+   GLuint interpMode[16];  /* XXX size? */
+   GLuint attr;
+   const GLbitfield inputsRead = stfp->Base.Base.InputsRead;
+   GLuint vslot = 0;
+   GLuint num_generic = 0;
+   GLuint num_tokens;
+
+   uint fs_num_inputs = 0;
+
+   ubyte fs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS];
+   ubyte fs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS];
+   uint fs_num_outputs = 0;
+
+   GLbitfield input_flags[MAX_PROGRAM_INPUTS];
+   GLbitfield output_flags[MAX_PROGRAM_OUTPUTS];
+
+   memset(&fs, 0, sizeof(fs));
+   memset(input_flags, 0, sizeof(input_flags));
+   memset(output_flags, 0, sizeof(output_flags));
+
+   /* which vertex output goes to the first fragment input: */
+   if (inputsRead & FRAG_BIT_WPOS)
+      vslot = 0;
+   else
+      vslot = 1;
+
+   /*
+    * Convert Mesa program inputs to TGSI input register semantics.
+    */
+   for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) {
+      if (inputsRead & (1 << attr)) {
+         const GLuint slot = fs_num_inputs;
+
+         defaultInputMapping[attr] = slot;
+
+         stfp->input_map[slot] = vslot++;
+
+         fs_num_inputs++;
+
+         switch (attr) {
+         case FRAG_ATTRIB_WPOS:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_POSITION;
+            stfp->input_semantic_index[slot] = 0;
+            interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
+            break;
+         case FRAG_ATTRIB_COL0:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            stfp->input_semantic_index[slot] = 0;
+            interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
+            break;
+         case FRAG_ATTRIB_COL1:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;
+            stfp->input_semantic_index[slot] = 1;
+            interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
+            break;
+         case FRAG_ATTRIB_FOGC:
+            if (stfp->Base.UsesPointCoord)
+               stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            else
+               stfp->input_semantic_name[slot] = TGSI_SEMANTIC_FOG;
+            stfp->input_semantic_index[slot] = 0;
+            interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
+            break;
+         case FRAG_ATTRIB_TEX0:
+         case FRAG_ATTRIB_TEX1:
+         case FRAG_ATTRIB_TEX2:
+         case FRAG_ATTRIB_TEX3:
+         case FRAG_ATTRIB_TEX4:
+         case FRAG_ATTRIB_TEX5:
+         case FRAG_ATTRIB_TEX6:
+         case FRAG_ATTRIB_TEX7:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            stfp->input_semantic_index[slot] = num_generic++;
+            interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
+            break;
+         case FRAG_ATTRIB_VAR0:
+            /* fall-through */
+         default:
+            stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
+            stfp->input_semantic_index[slot] = num_generic++;
+            interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE;
+         }
+
+         input_flags[slot] = stfp->Base.Base.InputFlags[attr];
+      }
+   }
+
+   /*
+    * Semantics and mapping for outputs
+    */
+   {
+      uint numColors = 0;
+      GLbitfield outputsWritten = stfp->Base.Base.OutputsWritten;
+
+      /* if z is written, emit that first */
+      if (outputsWritten & (1 << FRAG_RESULT_DEPR)) {
+         fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_POSITION;
+         fs_output_semantic_index[fs_num_outputs] = 0;
+         outputMapping[FRAG_RESULT_DEPR] = fs_num_outputs;
+         fs_num_outputs++;
+         outputsWritten &= ~(1 << FRAG_RESULT_DEPR);
+      }
+
+      /* handle remaning outputs (color) */
+      for (attr = 0; attr < FRAG_RESULT_MAX; attr++) {
+         if (outputsWritten & (1 << attr)) {
+            switch (attr) {
+            case FRAG_RESULT_DEPR:
+               /* handled above */
+               assert(0);
+               break;
+            case FRAG_RESULT_COLR:
+               fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_COLOR;
+               fs_output_semantic_index[fs_num_outputs] = numColors;
+               outputMapping[attr] = fs_num_outputs;
+               numColors++;
+               break;
+            default:
+               assert(0);
+            }
+
+            output_flags[fs_num_outputs] = stfp->Base.Base.OutputFlags[attr];
+
+            fs_num_outputs++;
+         }
+      }
+   }
+
+   if (!inputMapping)
+      inputMapping = defaultInputMapping;
+
+   /* XXX: fix static allocation of tokens:
+    */
+   num_tokens = st_translate_mesa_program(st->ctx,
+                                          TGSI_PROCESSOR_FRAGMENT,
+                                          &stfp->Base.Base,
+                                          /* inputs */
+                                          fs_num_inputs,
+                                          inputMapping,
+                                          stfp->input_semantic_name,
+                                          stfp->input_semantic_index,
+                                          interpMode,
+                                          input_flags,
+                                          /* outputs */
+                                          fs_num_outputs,
+                                          outputMapping,
+                                          fs_output_semantic_name,
+                                          fs_output_semantic_index,
+                                          output_flags,
+                                          /* tokenized result */
+                                          tokens, ST_MAX_SHADER_TOKENS);
+
+   assert(num_tokens < ST_MAX_SHADER_TOKENS);
+
+   fs.tokens = (struct tgsi_token *)
+      mem_dup(tokens, num_tokens * sizeof(tokens[0]));
+
+   stfp->state = fs; /* struct copy */
+   stfp->driver_shader = pipe->create_fs_state(pipe, &fs);
+
+   if (0)
+      _mesa_print_program(&stfp->Base.Base);
+
+   if (TGSI_DEBUG)
+      tgsi_dump( fs.tokens, 0/*TGSI_DUMP_VERBOSE*/ );
+}
+
+
+/**
+ * Debug- print current shader text
+ */
+void
+st_print_shaders(GLcontext *ctx)
+{
+   struct gl_shader_program *shProg = ctx->Shader.CurrentProgram;
+   if (shProg) {
+      GLuint i;
+      for (i = 0; i < shProg->NumShaders; i++) {
+         printf("GLSL shader %u of %u:\n", i, shProg->NumShaders);
+         printf("%s\n", shProg->Shaders[i]->Source);
+      }
+   }
+}
diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h
new file mode 100644
index 0000000000..e2e5eddef2
--- /dev/null
+++ b/src/mesa/state_tracker/st_program.h
@@ -0,0 +1,158 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+    
+
+#ifndef ST_PROGRAM_H
+#define ST_PROGRAM_H
+
+#include "main/mtypes.h"
+#include "shader/program.h"
+#include "pipe/p_shader_tokens.h"
+
+
+struct cso_fragment_shader;
+struct cso_vertex_shader;
+struct translated_vertex_program;
+
+
+/**
+ * Derived from Mesa gl_fragment_program:
+ */
+struct st_fragment_program
+{
+   struct gl_fragment_program Base;
+   GLuint serialNo;
+
+   GLuint input_to_slot[FRAG_ATTRIB_MAX];  /**< Maps FRAG_ATTRIB_x to slot */
+   GLuint num_input_slots;
+
+   /** map FP input back to VP output */
+   GLuint input_map[PIPE_MAX_SHADER_INPUTS];
+
+   ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS];
+   ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS];
+
+   struct pipe_shader_state state;
+   void *driver_shader;
+
+   GLuint param_state;
+
+   /** List of vertex programs which have been translated such that their
+    * outputs match this fragment program's inputs.
+    */
+   struct translated_vertex_program *vertex_programs;
+
+   /** Program prefixed with glBitmap prologue */
+   struct st_fragment_program *bitmap_program;
+   uint bitmap_sampler;
+};
+
+
+/**
+ * Derived from Mesa gl_fragment_program:
+ */
+struct st_vertex_program
+{
+   struct gl_vertex_program Base;  /**< The Mesa vertex program */
+   GLuint serialNo;
+
+   /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */
+   GLuint input_to_index[VERT_ATTRIB_MAX];
+   /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */
+   GLuint index_to_input[PIPE_MAX_SHADER_INPUTS];
+
+   GLuint num_inputs;
+
+   struct pipe_shader_state state;
+   void *driver_shader;
+
+   /** For using our private draw module (glRasterPos) */
+   struct draw_vertex_shader *draw_shader;
+
+   GLuint param_state;
+};
+
+
+static INLINE struct st_fragment_program *
+st_fragment_program( struct gl_fragment_program *fp )
+{
+   return (struct st_fragment_program *)fp;
+}
+
+
+static INLINE struct st_vertex_program *
+st_vertex_program( struct gl_vertex_program *vp )
+{
+   return (struct st_vertex_program *)vp;
+}
+
+
+static INLINE void
+st_reference_vertprog(struct st_context *st,
+                      struct st_vertex_program **ptr,
+                      struct st_vertex_program *prog)
+{
+   _mesa_reference_program(st->ctx,
+                           (struct gl_program **) ptr,
+                           (struct gl_program *) prog);
+}
+
+static INLINE void
+st_reference_fragprog(struct st_context *st,
+                      struct st_fragment_program **ptr,
+                      struct st_fragment_program *prog)
+{
+   _mesa_reference_program(st->ctx,
+                           (struct gl_program **) ptr,
+                           (struct gl_program *) prog);
+}
+
+
+extern void
+st_translate_fragment_program(struct st_context *st,
+                              struct st_fragment_program *fp,
+                              const GLuint inputMapping[]);
+
+
+extern void
+st_translate_vertex_program(struct st_context *st,
+                            struct st_vertex_program *vp,
+                            const GLuint vert_output_to_slot[],
+                            const ubyte *fs_input_semantic_name,
+                            const ubyte *fs_input_semantic_index);
+
+
+extern void
+st_print_shaders(GLcontext *ctx);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h
new file mode 100644
index 0000000000..414218bb58
--- /dev/null
+++ b/src/mesa/state_tracker/st_public.h
@@ -0,0 +1,122 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef ST_PUBLIC_H
+#define ST_PUBLIC_H
+
+#include "GL/gl.h"
+#include "GL/internal/glcore.h"  /* for __GLcontextModes */
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+
+
+#define ST_SURFACE_FRONT_LEFT   0
+#define ST_SURFACE_BACK_LEFT    1
+#define ST_SURFACE_FRONT_RIGHT  2
+#define ST_SURFACE_BACK_RIGHT   3
+#define ST_SURFACE_DEPTH        8
+
+#define ST_TEXTURE_2D    0x2
+#define ST_TEXTURE_RECT  0x4
+
+#define ST_TEXTURE_RGB   0x1
+#define ST_TEXTURE_RGBA  0x2
+
+
+struct st_context;
+struct st_framebuffer;
+struct pipe_context;
+struct pipe_fence_handle;
+struct pipe_surface;
+struct pipe_texture;
+
+
+struct st_context *st_create_context(struct pipe_context *pipe,
+                                     const __GLcontextModes *visual,
+                                     struct st_context *share);
+
+void st_destroy_context( struct st_context *st );
+
+void st_copy_context_state(struct st_context *dst, struct st_context *src,
+                           uint mask);
+
+struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual,
+                                              enum pipe_format colorFormat,
+                                              enum pipe_format depthFormat,
+                                              enum pipe_format stencilFormat,
+                                              uint width, uint height,
+                                              void *privateData);
+
+void st_resize_framebuffer( struct st_framebuffer *stfb,
+                            uint width, uint height );
+
+void st_set_framebuffer_surface(struct st_framebuffer *stfb,
+                                uint surfIndex, struct pipe_surface *surf);
+
+void st_get_framebuffer_dimensions( struct st_framebuffer *stfb,
+				    uint *width, uint *height);
+
+int st_get_framebuffer_surface(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_surface **surface);
+
+int st_get_framebuffer_texture(struct st_framebuffer *stfb,
+                               uint surfIndex, struct pipe_texture **texture);
+
+void *st_framebuffer_private( struct st_framebuffer *stfb );
+
+void st_unreference_framebuffer( struct st_framebuffer *stfb );
+
+void st_make_current(struct st_context *st,
+                     struct st_framebuffer *draw,
+                     struct st_framebuffer *read);
+
+void st_flush( struct st_context *st, uint pipeFlushFlags,
+               struct pipe_fence_handle **fence );
+void st_finish( struct st_context *st );
+
+void st_notify_swapbuffers(struct st_framebuffer *stfb);
+void st_notify_swapbuffers_complete(struct st_framebuffer *stfb);
+
+int st_set_teximage(struct pipe_texture *pt, int target);
+
+/** Redirect rendering into stfb's surface to a texture image */
+int st_bind_teximage(struct st_framebuffer *stfb, uint surfIndex,
+                     int target, int format, int level);
+
+/** Undo surface-to-texture binding */
+int st_release_teximage(struct st_framebuffer *stfb, uint surfIndex,
+                        int target, int format, int level);
+
+
+/** Generic function type */
+typedef void (*st_proc)();
+
+st_proc st_get_proc_address(const char *procname);
+
+
+#endif
diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c
new file mode 100644
index 0000000000..63cfe5fc16
--- /dev/null
+++ b/src/mesa/state_tracker/st_texture.c
@@ -0,0 +1,488 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "st_context.h"
+#include "st_format.h"
+#include "st_public.h"
+#include "st_texture.h"
+#include "st_cb_fbo.h"
+#include "main/enums.h"
+#include "main/teximage.h"
+#include "main/texstore.h"
+
+#undef Elements  /* fix re-defined macro warning */
+
+#include "pipe/p_state.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_inlines.h"
+#include "util/u_rect.h"
+
+
+#define DBG if(0) printf
+
+#if 0
+static GLenum
+target_to_target(GLenum target)
+{
+   switch (target) {
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB:
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB:
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB:
+      return GL_TEXTURE_CUBE_MAP_ARB;
+   default:
+      return target;
+   }
+}
+#endif
+
+
+/**
+ * Allocate a new pipe_texture object
+ * width0, height0, depth0 are the dimensions of the level 0 image
+ * (the highest resolution).  last_level indicates how many mipmap levels
+ * to allocate storage for.  For non-mipmapped textures, this will be zero.
+ */
+struct pipe_texture *
+st_texture_create(struct st_context *st,
+                  enum pipe_texture_target target,
+		  enum pipe_format format,
+		  GLuint last_level,
+		  GLuint width0,
+		  GLuint height0,
+		  GLuint depth0,
+		  GLuint compress_byte,
+                  GLuint usage )
+{
+   struct pipe_texture pt, *newtex;
+   struct pipe_screen *screen = st->pipe->screen;
+
+   assert(target <= PIPE_TEXTURE_CUBE);
+
+   DBG("%s target %s format %s last_level %d\n", __FUNCTION__,
+       _mesa_lookup_enum_by_nr(target),
+       _mesa_lookup_enum_by_nr(format), last_level);
+
+   assert(format);
+   assert(screen->is_format_supported(screen, format, target, 
+                                      PIPE_TEXTURE_USAGE_SAMPLER, 0));
+
+   memset(&pt, 0, sizeof(pt));
+   pt.target = target;
+   pt.format = format;
+   pt.last_level = last_level;
+   pt.width[0] = width0;
+   pt.height[0] = height0;
+   pt.depth[0] = depth0;
+   pt.compressed = compress_byte ? 1 : 0;
+   pf_get_block(format, &pt.block);
+   pt.tex_usage = usage;
+
+   newtex = screen->texture_create(screen, &pt);
+
+   assert(!newtex || newtex->refcount == 1);
+
+   return newtex;
+}
+
+
+/**
+ * Check if a texture image be pulled into a unified mipmap texture.
+ * This mirrors the completeness test in a lot of ways.
+ *
+ * Not sure whether I want to pass gl_texture_image here.
+ */
+GLboolean
+st_texture_match_image(const struct pipe_texture *pt,
+                       const struct gl_texture_image *image,
+                       GLuint face, GLuint level)
+{
+   /* Images with borders are never pulled into mipmap textures. 
+    */
+   if (image->Border) 
+      return GL_FALSE;
+
+   if (st_mesa_format_to_pipe_format(image->TexFormat->MesaFormat) != pt->format ||
+       image->IsCompressed != pt->compressed)
+      return GL_FALSE;
+
+   /* Test image dimensions against the base level image adjusted for
+    * minification.  This will also catch images not present in the
+    * texture, changed targets, etc.
+    */
+   if (image->Width != pt->width[level] ||
+       image->Height != pt->height[level] ||
+       image->Depth != pt->depth[level])
+      return GL_FALSE;
+
+   return GL_TRUE;
+}
+
+
+#if 000
+/* Although we use the image_offset[] array to store relative offsets
+ * to cube faces, Mesa doesn't know anything about this and expects
+ * each cube face to be treated as a separate image.
+ *
+ * These functions present that view to mesa:
+ */
+const GLuint *
+st_texture_depth_offsets(struct pipe_texture *pt, GLuint level)
+{
+   static const GLuint zero = 0;
+
+   if (pt->target != PIPE_TEXTURE_3D || pt->level[level].nr_images == 1)
+      return &zero;
+   else
+      return pt->level[level].image_offset;
+}
+
+
+/**
+ * Return the offset to the given mipmap texture image within the
+ * texture memory buffer, in bytes.
+ */
+GLuint
+st_texture_image_offset(const struct pipe_texture * pt,
+                        GLuint face, GLuint level)
+{
+   if (pt->target == PIPE_TEXTURE_CUBE)
+      return (pt->level[level].level_offset +
+              pt->level[level].image_offset[face] * pt->cpp);
+   else
+      return pt->level[level].level_offset;
+}
+#endif
+
+
+/**
+ * Map a teximage in a mipmap texture.
+ * \param row_stride  returns row stride in bytes
+ * \param image_stride  returns image stride in bytes (for 3D textures).
+ * \return address of mapping
+ */
+GLubyte *
+st_texture_image_map(struct st_context *st, struct st_texture_image *stImage,
+		     GLuint zoffset,
+                     GLuint flags )
+{
+   struct pipe_screen *screen = st->pipe->screen;
+   struct pipe_texture *pt = stImage->pt;
+   DBG("%s \n", __FUNCTION__);
+
+   stImage->surface = screen->get_tex_surface(screen, pt, stImage->face,
+                                              stImage->level, zoffset, 
+                                              flags);
+
+   if (stImage->surface)
+      return screen->surface_map(screen, stImage->surface, flags);
+   else
+      return NULL;
+}
+
+
+void
+st_texture_image_unmap(struct st_context *st,
+                       struct st_texture_image *stImage)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+
+   DBG("%s\n", __FUNCTION__);
+
+   screen->surface_unmap(screen, stImage->surface);
+
+   pipe_surface_reference(&stImage->surface, NULL);
+}
+
+
+
+/**
+ * Upload data to a rectangular sub-region.  Lots of choices how to do this:
+ *
+ * - memcpy by span to current destination
+ * - upload data as new buffer and blit
+ *
+ * Currently always memcpy.
+ */
+static void
+st_surface_data(struct pipe_context *pipe,
+		struct pipe_surface *dst,
+		unsigned dstx, unsigned dsty,
+		const void *src, unsigned src_stride,
+		unsigned srcx, unsigned srcy, unsigned width, unsigned height)
+{
+   struct pipe_screen *screen = pipe->screen;
+   void *map = screen->surface_map(screen, dst, PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   pipe_copy_rect(map,
+                  &dst->block,
+                  dst->stride,
+                  dstx, dsty, 
+                  width, height, 
+                  src, src_stride, 
+                  srcx, srcy);
+
+   screen->surface_unmap(screen, dst);
+}
+
+
+/* Upload data for a particular image.
+ */
+void
+st_texture_image_data(struct pipe_context *pipe,
+                      struct pipe_texture *dst,
+                      GLuint face,
+                      GLuint level,
+                      void *src,
+                      GLuint src_row_stride, GLuint src_image_stride)
+{
+   struct pipe_screen *screen = pipe->screen;
+   GLuint depth = dst->depth[level];
+   GLuint i;
+   const GLubyte *srcUB = src;
+   struct pipe_surface *dst_surface;
+
+   DBG("%s\n", __FUNCTION__);
+   for (i = 0; i < depth; i++) {
+      dst_surface = screen->get_tex_surface(screen, dst, face, level, i,
+                                            PIPE_BUFFER_USAGE_CPU_WRITE);
+
+      st_surface_data(pipe, dst_surface,
+		      0, 0,                             /* dstx, dsty */
+		      srcUB,
+		      src_row_stride,
+		      0, 0,                             /* source x, y */
+		      dst->width[level], dst->height[level]);       /* width, height */
+
+      screen->tex_surface_release(screen, &dst_surface);
+
+      srcUB += src_image_stride;
+   }
+}
+
+
+/* Copy mipmap image between textures
+ */
+void
+st_texture_image_copy(struct pipe_context *pipe,
+                      struct pipe_texture *dst, GLuint dstLevel,
+                      struct pipe_texture *src,
+                      GLuint face)
+{
+   struct pipe_screen *screen = pipe->screen;
+   GLuint width = dst->width[dstLevel];
+   GLuint height = dst->height[dstLevel];
+   GLuint depth = dst->depth[dstLevel];
+   struct pipe_surface *src_surface;
+   struct pipe_surface *dst_surface;
+   GLuint i;
+
+   for (i = 0; i < depth; i++) {
+      GLuint srcLevel;
+
+      /* find src texture level of needed size */
+      for (srcLevel = 0; srcLevel <= src->last_level; srcLevel++) {
+         if (src->width[srcLevel] == width &&
+             src->height[srcLevel] == height) {
+            break;
+         }
+      }
+      assert(src->width[srcLevel] == width);
+      assert(src->height[srcLevel] == height);
+
+#if 0
+      {
+         src_surface = screen->get_tex_surface(screen, src, face, srcLevel, i,
+                                               PIPE_BUFFER_USAGE_CPU_READ);
+         ubyte *map = screen->surface_map(screen, src_surface, PIPE_BUFFER_USAGE_CPU_READ);
+         map += src_surface->width * src_surface->height * 4 / 2;
+         printf("%s center pixel: %d %d %d %d (pt %p[%d] -> %p[%d])\n",
+                __FUNCTION__,
+                map[0], map[1], map[2], map[3],
+                src, srcLevel, dst, dstLevel);
+
+         screen->surface_unmap(screen, src_surface);
+         pipe_surface_reference(&src_surface, NULL);
+      }
+#endif
+
+      dst_surface = screen->get_tex_surface(screen, dst, face, dstLevel, i,
+                                            PIPE_BUFFER_USAGE_GPU_WRITE);
+
+      src_surface = screen->get_tex_surface(screen, src, face, srcLevel, i,
+                                            PIPE_BUFFER_USAGE_GPU_READ);
+
+      pipe->surface_copy(pipe,
+                         FALSE,
+			 dst_surface,
+			 0, 0, /* destX, Y */
+			 src_surface,
+			 0, 0, /* srcX, Y */
+			 width, height);
+
+      screen->tex_surface_release(screen, &src_surface);
+      screen->tex_surface_release(screen, &dst_surface);
+   }
+}
+
+/** Bind a pipe surface for use as a texture image */
+int
+st_set_teximage(struct pipe_texture *pt, int target)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   const GLuint unit = ctx->Texture.CurrentUnit;
+   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+   struct gl_texture_object *texObj;
+   struct gl_texture_image *texImage;
+   struct st_texture_image *stImage;
+   int internalFormat;
+
+   switch (pt->format) {
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      internalFormat = GL_RGBA8;
+      break;
+   default:
+      return 0;
+   };
+
+   switch (target) {
+   case ST_TEXTURE_2D:
+      target = GL_TEXTURE_2D;
+      break;
+   case ST_TEXTURE_RECT:
+      target = GL_TEXTURE_RECTANGLE_ARB;
+      break;
+   default:
+      return 0;
+   }
+
+   texObj = _mesa_select_tex_object(ctx, texUnit, target);
+   texImage = _mesa_get_tex_image(ctx, texObj, target, 0);
+   stImage = st_texture_image(texImage);
+   
+   _mesa_init_teximage_fields(ctx, GL_TEXTURE_2D, texImage, pt->width[0],
+                              pt->height[0], 1, 0, internalFormat);
+
+   texImage->TexFormat = st_ChooseTextureFormat(ctx, internalFormat, GL_RGBA,
+                                                GL_UNSIGNED_BYTE);
+   _mesa_set_fetch_functions(texImage, 2);
+
+   pipe_texture_reference(&stImage->pt, pt);
+
+   return 1;
+}
+
+/** Redirect rendering into stfb's surface to a texture image */
+int
+st_bind_teximage(struct st_framebuffer *stfb, uint surfIndex,
+                 int target, int format, int level)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   const GLuint unit = ctx->Texture.CurrentUnit;
+   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+   struct gl_texture_object *texObj;
+   struct gl_texture_image *texImage;
+   struct st_texture_image *stImage;
+   struct st_renderbuffer *strb;
+   GLint face = 0, slice = 0;
+
+   assert(surfIndex <= ST_SURFACE_DEPTH);
+
+   strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer);
+
+   if (strb->texture_save || strb->surface_save) {
+      /* Error! */
+      return 0;
+   }
+
+   if (target == ST_TEXTURE_2D) {
+      texObj = texUnit->Current2D;
+      texImage = _mesa_get_tex_image(ctx, texObj, GL_TEXTURE_2D, level);
+      stImage = st_texture_image(texImage);
+   }
+   else {
+      /* unsupported target */
+      return 0;
+   }
+
+   st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   /* save the renderbuffer's surface/texture info */
+   pipe_texture_reference(&strb->texture_save, strb->texture);
+   pipe_surface_reference(&strb->surface_save, strb->surface);
+
+   /* plug in new surface/texture info */
+   pipe_texture_reference(&strb->texture, stImage->pt);
+   strb->surface = screen->get_tex_surface(screen, strb->texture,
+                                           face, level, slice,
+                                           (PIPE_BUFFER_USAGE_GPU_READ |
+                                            PIPE_BUFFER_USAGE_GPU_WRITE));
+
+   st->dirty.st |= ST_NEW_FRAMEBUFFER;
+
+   return 1;
+}
+
+
+/** Undo surface-to-texture binding */
+int
+st_release_teximage(struct st_framebuffer *stfb, uint surfIndex,
+                    int target, int format, int level)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct st_context *st = ctx->st;
+   struct st_renderbuffer *strb;
+
+   assert(surfIndex <= ST_SURFACE_DEPTH);
+
+   strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer);
+
+   if (!strb->texture_save || !strb->surface_save) {
+      /* Error! */
+      return 0;
+   }
+
+   st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   /* free tex surface, restore original */
+   pipe_surface_reference(&strb->surface, strb->surface_save);
+   pipe_texture_reference(&strb->texture, strb->texture_save);
+
+   pipe_surface_reference(&strb->surface_save, NULL);
+   pipe_texture_reference(&strb->texture_save, NULL);
+
+   st->dirty.st |= ST_NEW_FRAMEBUFFER;
+
+   return 1;
+}
diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h
new file mode 100644
index 0000000000..31f66ad52c
--- /dev/null
+++ b/src/mesa/state_tracker/st_texture.h
@@ -0,0 +1,179 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef ST_TEXTURE_H
+#define ST_TEXTURE_H
+
+
+#include "main/mtypes.h"
+
+struct pipe_context;
+struct pipe_texture;
+
+
+struct st_texture_image
+{
+   struct gl_texture_image base;
+
+   /* These aren't stored in gl_texture_image 
+    */
+   GLuint level;
+   GLuint face;
+
+   /* If stImage->pt != NULL, image data is stored here.
+    * Else if stImage->base.Data != NULL, image is stored there.
+    * Else there is no image data.
+    */
+   struct pipe_texture *pt;
+
+   struct pipe_surface *surface;
+};
+
+
+
+struct st_texture_object
+{
+   struct gl_texture_object base;       /* The "parent" object */
+
+   /* The texture must include at levels [0..lastLevel] once validated:
+    */
+   GLuint lastLevel;
+
+   /* On validation any active images held in main memory or in other
+    * textures will be copied to this texture and the old storage freed.
+    */
+   struct pipe_texture *pt;
+
+   GLboolean teximage_realloc;
+};
+
+
+static INLINE struct st_texture_image *
+st_texture_image(struct gl_texture_image *img)
+{
+   return (struct st_texture_image *) img;
+}
+
+static INLINE struct st_texture_object *
+st_texture_object(struct gl_texture_object *obj)
+{
+   return (struct st_texture_object *) obj;
+}
+
+
+static INLINE struct pipe_texture *
+st_get_texobj_texture(struct gl_texture_object *texObj)
+{
+   struct st_texture_object *stObj = st_texture_object(texObj);
+   return stObj ? stObj->pt : NULL;
+}
+
+
+static INLINE struct pipe_texture *
+st_get_stobj_texture(struct st_texture_object *stObj)
+{
+   return stObj ? stObj->pt : NULL;
+}
+
+static INLINE GLboolean pf_is_depth_stencil( enum pipe_format format )
+{
+   return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) +
+           pf_get_component_bits( format, PIPE_FORMAT_COMP_S )) != 0;
+}
+
+
+extern struct pipe_texture *
+st_texture_create(struct st_context *st,
+                  enum pipe_texture_target target,
+		  enum pipe_format format,
+                  GLuint last_level,
+                  GLuint width0,
+                  GLuint height0,
+                  GLuint depth0,
+                  GLuint compress_byte,
+                  GLuint tex_usage );
+
+
+/* Check if an image fits into an existing texture object.
+ */
+extern GLboolean
+st_texture_match_image(const struct pipe_texture *pt,
+                       const struct gl_texture_image *image,
+                       GLuint face, GLuint level);
+
+/* Return a pointer to an image within a texture.  Return image stride as
+ * well.
+ */
+extern GLubyte *
+st_texture_image_map(struct st_context *st,
+                     struct st_texture_image *stImage,
+		     GLuint zoffset,
+                     GLuint flags);
+
+extern void
+st_texture_image_unmap(struct st_context *st,
+                       struct st_texture_image *stImage);
+
+
+/* Return pointers to each 2d slice within an image.  Indexed by depth
+ * value.
+ */
+extern const GLuint *
+st_texture_depth_offsets(struct pipe_texture *pt, GLuint level);
+
+
+/* Return the linear offset of an image relative to the start of its region.
+ */
+extern GLuint
+st_texture_image_offset(const struct pipe_texture *pt,
+                        GLuint face, GLuint level);
+
+extern GLuint
+st_texture_texel_offset(const struct pipe_texture * pt,
+                        GLuint face, GLuint level,
+                        GLuint col, GLuint row, GLuint img);
+
+
+/* Upload an image into a texture
+ */
+extern void
+st_texture_image_data(struct pipe_context *pipe,
+                      struct pipe_texture *dst,
+                      GLuint face, GLuint level, void *src,
+                      GLuint src_row_pitch, GLuint src_image_pitch);
+
+
+/* Copy an image between two textures
+ */
+extern void
+st_texture_image_copy(struct pipe_context *pipe,
+                      struct pipe_texture *dst, GLuint dstLevel,
+                      struct pipe_texture *src,
+                      GLuint face);
+
+
+#endif
diff --git a/src/mesa/swrast/s_aalinetemp.h b/src/mesa/swrast/s_aalinetemp.h
index ca08203d83..42ffe9f20c 100644
--- a/src/mesa/swrast/s_aalinetemp.h
+++ b/src/mesa/swrast/s_aalinetemp.h
@@ -76,7 +76,7 @@ NAME(plot)(GLcontext *ctx, struct LineInfo *line, int ix, int iy)
    ATTRIB_LOOP_BEGIN
       GLfloat (*attribArray)[4] = line->span.array->attribs[attr];
       if (attr >= FRAG_ATTRIB_TEX0 && attr < FRAG_ATTRIB_VAR0
-          && !ctx->FragmentProgram._Active) {
+          && !ctx->FragmentProgram._Current) {
          /* texcoord w/ divide by Q */
          const GLuint unit = attr - FRAG_ATTRIB_TEX0;
          const GLfloat invQ = solve_plane_recip(fx, fy, line->attrPlane[attr][3]);
diff --git a/src/mesa/swrast/s_accum.c b/src/mesa/swrast/s_accum.c
index ff741777e7..c6c7dbf5cf 100644
--- a/src/mesa/swrast/s_accum.c
+++ b/src/mesa/swrast/s_accum.c
@@ -550,7 +550,7 @@ _swrast_Accum(GLcontext *ctx, GLenum op, GLfloat value)
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLint xpos, ypos, width, height;
 
-   if (SWRAST_CONTEXT(ctx)->NewState)
+   if (swrast->NewState)
       _swrast_validate_derived( ctx );
 
    if (!ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer) {
@@ -558,9 +558,9 @@ _swrast_Accum(GLcontext *ctx, GLenum op, GLfloat value)
       return;
    }
 
-   RENDER_START(swrast, ctx);
+   swrast_render_start(ctx);
 
-   /* Compute region after calling RENDER_START so that we know the
+   /* Compute region after calling swrast_render_start() so that we know the
     * drawbuffer's size/bounds are up to date.
     */
    xpos = ctx->DrawBuffer->_Xmin;
@@ -595,5 +595,5 @@ _swrast_Accum(GLcontext *ctx, GLenum op, GLfloat value)
          break;
    }
 
-   RENDER_FINISH(swrast, ctx);
+   swrast_render_finish(ctx);
 }
diff --git a/src/mesa/swrast/s_bitmap.c b/src/mesa/swrast/s_bitmap.c
index 35b34e654f..5e7822cf32 100644
--- a/src/mesa/swrast/s_bitmap.c
+++ b/src/mesa/swrast/s_bitmap.c
@@ -50,7 +50,6 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py,
 		const struct gl_pixelstore_attrib *unpack,
 		const GLubyte *bitmap )
 {
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLint row, col;
    GLuint count = 0;
    SWspan span;
@@ -61,7 +60,7 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py,
    if (!bitmap)
       return;
 
-   RENDER_START(swrast,ctx);
+   swrast_render_start(ctx);
 
    if (SWRAST_CONTEXT(ctx)->NewState)
       _swrast_validate_derived( ctx );
@@ -132,7 +131,7 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py,
       }
    }
 
-   RENDER_FINISH(swrast,ctx);
+   swrast_render_finish(ctx);
 
    _mesa_unmap_bitmap_pbo(ctx, unpack);
 }
@@ -157,7 +156,7 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py,
    ASSERT(ctx->RenderMode == GL_RENDER);
    ASSERT(bitmap);
 
-   RENDER_START(swrast,ctx);
+   swrast_render_start(ctx);
 
    if (SWRAST_CONTEXT(ctx)->NewState)
       _swrast_validate_derived( ctx );
@@ -224,6 +223,6 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py,
       }
    }
 
-   RENDER_FINISH(swrast,ctx);
+   swrast_render_finish(ctx);
 }
 #endif
diff --git a/src/mesa/swrast/s_blit.c b/src/mesa/swrast/s_blit.c
index bc4b2ac625..0e32cb8f65 100644
--- a/src/mesa/swrast/s_blit.c
+++ b/src/mesa/swrast/s_blit.c
@@ -737,7 +737,6 @@ _swrast_BlitFramebuffer(GLcontext *ctx,
                         GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
                         GLbitfield mask, GLenum filter)
 {
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    static const GLint buffers[3] = {
       GL_COLOR_BUFFER_BIT,
       GL_DEPTH_BUFFER_BIT,
@@ -753,7 +752,7 @@ _swrast_BlitFramebuffer(GLcontext *ctx,
       return;
    }
 
-   RENDER_START(swrast, ctx);
+   swrast_render_start(ctx);
 
    if (srcX1 - srcX0 == dstX1 - dstX0 &&
        srcY1 - srcY0 == dstY1 - dstY0 &&
@@ -789,5 +788,5 @@ _swrast_BlitFramebuffer(GLcontext *ctx,
       }
    }
 
-   RENDER_FINISH(swrast, ctx);
+   swrast_render_finish(ctx);
 }
diff --git a/src/mesa/swrast/s_buffers.c b/src/mesa/swrast/s_buffers.c
index 9e87d6d485..af475ad8cb 100644
--- a/src/mesa/swrast/s_buffers.c
+++ b/src/mesa/swrast/s_buffers.c
@@ -307,8 +307,6 @@ clear_color_buffers(GLcontext *ctx)
 void
 _swrast_Clear(GLcontext *ctx, GLbitfield buffers)
 {
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
-
 #ifdef DEBUG_FOO
    {
       const GLbitfield legalBits =
@@ -327,7 +325,7 @@ _swrast_Clear(GLcontext *ctx, GLbitfield buffers)
    }
 #endif
 
-   RENDER_START(swrast,ctx);
+   swrast_render_start(ctx);
 
    /* do software clearing here */
    if (buffers) {
@@ -347,5 +345,5 @@ _swrast_Clear(GLcontext *ctx, GLbitfield buffers)
       }
    }
 
-   RENDER_FINISH(swrast,ctx);
+   swrast_render_finish(ctx);
 }
diff --git a/src/mesa/swrast/s_context.c b/src/mesa/swrast/s_context.c
index 62857ddeb0..297940adbd 100644
--- a/src/mesa/swrast/s_context.c
+++ b/src/mesa/swrast/s_context.c
@@ -533,6 +533,9 @@ _swrast_update_texture_samplers(GLcontext *ctx)
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLuint u;
 
+   if (!swrast)
+      return; /* pipe hack */
+
    for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++) {
       const struct gl_texture_object *tObj = ctx->Texture.Unit[u]._Current;
       /* Note: If tObj is NULL, the sample function will be a simple
diff --git a/src/mesa/swrast/s_context.h b/src/mesa/swrast/s_context.h
index a511d1c9a1..cdd6fa5048 100644
--- a/src/mesa/swrast/s_context.h
+++ b/src/mesa/swrast/s_context.h
@@ -238,21 +238,43 @@ extern void
 _swrast_update_texture_samplers(GLcontext *ctx);
 
 
-#define SWRAST_CONTEXT(ctx) ((SWcontext *)ctx->swrast_context)
-
-#define RENDER_START(SWctx, GLctx)			\
-   do {							\
-      if ((SWctx)->Driver.SpanRenderStart) {		\
-         (*(SWctx)->Driver.SpanRenderStart)(GLctx);	\
-      }							\
-   } while (0)
-
-#define RENDER_FINISH(SWctx, GLctx)			\
-   do {							\
-      if ((SWctx)->Driver.SpanRenderFinish) {		\
-         (*(SWctx)->Driver.SpanRenderFinish)(GLctx);	\
-      }							\
-   } while (0)
+/** Return SWcontext for the given GLcontext */
+static INLINE SWcontext *
+SWRAST_CONTEXT(GLcontext *ctx)
+{
+   return (SWcontext *) ctx->swrast_context;
+}
+
+/** const version of above */
+static INLINE const SWcontext *
+CONST_SWRAST_CONTEXT(const GLcontext *ctx)
+{
+   return (const SWcontext *) ctx->swrast_context;
+}
+
+
+/**
+ * Called prior to framebuffer reading/writing.
+ * For drivers that rely on swrast for fallback rendering, this is the
+ * driver's opportunity to map renderbuffers and textures.
+ */
+static INLINE void
+swrast_render_start(GLcontext *ctx)
+{
+   SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   if (swrast->Driver.SpanRenderStart)
+      swrast->Driver.SpanRenderStart(ctx);
+}
+
+
+/** Called after framebuffer reading/writing */
+static INLINE void
+swrast_render_finish(GLcontext *ctx)
+{
+   SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   if (swrast->Driver.SpanRenderFinish)
+      swrast->Driver.SpanRenderFinish(ctx);
+}
 
 
 
diff --git a/src/mesa/swrast/s_copypix.c b/src/mesa/swrast/s_copypix.c
index fc5990b261..5ecfb1e90a 100644
--- a/src/mesa/swrast/s_copypix.c
+++ b/src/mesa/swrast/s_copypix.c
@@ -899,7 +899,7 @@ _swrast_CopyPixels( GLcontext *ctx,
 		    GLint destx, GLint desty, GLenum type )
 {
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
-   RENDER_START(swrast,ctx);
+   swrast_render_start(ctx);
       
    if (swrast->NewState)
       _swrast_validate_derived( ctx );
@@ -928,5 +928,5 @@ _swrast_CopyPixels( GLcontext *ctx,
       }
    }
 
-   RENDER_FINISH(swrast,ctx);
+   swrast_render_finish(ctx);
 }
diff --git a/src/mesa/swrast/s_drawpix.c b/src/mesa/swrast/s_drawpix.c
index 7af3e3dad1..700f76d4bc 100644
--- a/src/mesa/swrast/s_drawpix.c
+++ b/src/mesa/swrast/s_drawpix.c
@@ -830,7 +830,7 @@ _swrast_DrawPixels( GLcontext *ctx,
 {
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
 
-   RENDER_START(swrast,ctx);
+   swrast_render_start(ctx);
 
    if (ctx->NewState)
       _mesa_update_state(ctx);
@@ -840,7 +840,7 @@ _swrast_DrawPixels( GLcontext *ctx,
 
     pixels = _mesa_map_drawpix_pbo(ctx, unpack, pixels);
     if (!pixels) {
-       RENDER_FINISH(swrast,ctx);
+       swrast_render_finish(ctx);
        return;
     }
 
@@ -879,9 +879,9 @@ _swrast_DrawPixels( GLcontext *ctx,
       /* don't return yet, clean-up */
    }
 
-   RENDER_FINISH(swrast,ctx);
+   swrast_render_finish(ctx);
 
-   _mesa_unmap_drapix_pbo(ctx, unpack);
+   _mesa_unmap_drawpix_pbo(ctx, unpack);
 }
 
 
@@ -904,7 +904,7 @@ _swrast_DrawDepthPixelsMESA( GLcontext *ctx,
    if (swrast->NewState)
       _swrast_validate_derived( ctx );
 
-   RENDER_START(swrast,ctx);
+   swrast_render_start(ctx);
 
    switch (colorFormat) {
    case GL_COLOR_INDEX:
@@ -933,6 +933,6 @@ _swrast_DrawDepthPixelsMESA( GLcontext *ctx,
       _mesa_problem(ctx, "unexpected format in glDrawDepthPixelsMESA");
    }
 
-   RENDER_FINISH(swrast,ctx);
+   swrast_render_finish(ctx);
 }
 #endif
diff --git a/src/mesa/swrast/s_fog.c b/src/mesa/swrast/s_fog.c
index b9ba265db6..77ed0cfef9 100644
--- a/src/mesa/swrast/s_fog.c
+++ b/src/mesa/swrast/s_fog.c
@@ -164,7 +164,7 @@ else {										\
 void
 _swrast_fog_rgba_span( const GLcontext *ctx, SWspan *span )
 {
-   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   const SWcontext *swrast = CONST_SWRAST_CONTEXT(ctx);
    GLfloat rFog, gFog, bFog;
 
    ASSERT(swrast->_FogEnabled);
@@ -283,7 +283,7 @@ _swrast_fog_rgba_span( const GLcontext *ctx, SWspan *span )
 void
 _swrast_fog_ci_span( const GLcontext *ctx, SWspan *span )
 {
-   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   const SWcontext *swrast = CONST_SWRAST_CONTEXT(ctx);
    const GLuint fogIndex = (GLuint) ctx->Fog.Index;
    GLuint *index = span->array->index;
 
diff --git a/src/mesa/swrast/s_fragprog.c b/src/mesa/swrast/s_fragprog.c
index 525cf9d724..c6601f5593 100644
--- a/src/mesa/swrast/s_fragprog.c
+++ b/src/mesa/swrast/s_fragprog.c
@@ -33,6 +33,35 @@
 
 
 /**
+ * Apply texture object's swizzle (X/Y/Z/W/0/1) to incoming 'texel'
+ * and return results in 'colorOut'.
+ */
+static INLINE void
+swizzle_texel(const GLchan texel[4], GLfloat colorOut[4], GLuint swizzle)
+{
+   if (swizzle == SWIZZLE_NOOP) {
+      colorOut[0] = CHAN_TO_FLOAT(texel[0]);
+      colorOut[1] = CHAN_TO_FLOAT(texel[1]);
+      colorOut[2] = CHAN_TO_FLOAT(texel[2]);
+      colorOut[3] = CHAN_TO_FLOAT(texel[3]);
+   }
+   else {
+      GLfloat vector[6];
+      vector[SWIZZLE_X] = CHAN_TO_FLOAT(texel[0]);
+      vector[SWIZZLE_Y] = CHAN_TO_FLOAT(texel[1]);
+      vector[SWIZZLE_Z] = CHAN_TO_FLOAT(texel[2]);
+      vector[SWIZZLE_W] = CHAN_TO_FLOAT(texel[3]);
+      vector[SWIZZLE_ZERO] = 0.0F;
+      vector[SWIZZLE_ONE] = 1.0F;
+      colorOut[0] = vector[GET_SWZ(swizzle, 0)];
+      colorOut[1] = vector[GET_SWZ(swizzle, 1)];
+      colorOut[2] = vector[GET_SWZ(swizzle, 2)];
+      colorOut[3] = vector[GET_SWZ(swizzle, 3)];
+   }
+}
+
+
+/**
  * Fetch a texel with given lod.
  * Called via machine->FetchTexelLod()
  */
@@ -40,20 +69,23 @@ static void
 fetch_texel_lod( GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda,
                  GLuint unit, GLfloat color[4] )
 {
-   GLchan rgba[4];
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    const struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current;
 
-   if (texObj)
+   if (texObj) {
+      SWcontext *swrast = SWRAST_CONTEXT(ctx);
+      GLchan rgba[4];
+
       lambda = CLAMP(lambda, texObj->MinLod, texObj->MaxLod);
 
-   /* XXX use a float-valued TextureSample routine here!!! */
-   swrast->TextureSample[unit](ctx, texObj, 1, (const GLfloat (*)[4]) texcoord,
-                               &lambda, &rgba);
-   color[0] = CHAN_TO_FLOAT(rgba[0]);
-   color[1] = CHAN_TO_FLOAT(rgba[1]);
-   color[2] = CHAN_TO_FLOAT(rgba[2]);
-   color[3] = CHAN_TO_FLOAT(rgba[3]);
+      /* XXX use a float-valued TextureSample routine here!!! */
+      swrast->TextureSample[unit](ctx, texObj, 1,
+                                  (const GLfloat (*)[4]) texcoord,
+                                  &lambda, &rgba);
+      swizzle_texel(rgba, color, texObj->_Swizzle);
+   }
+   else {
+      ASSIGN_4V(color, 0.0F, 0.0F, 0.0F, 1.0F);
+   }
 }
 
 
@@ -69,13 +101,14 @@ fetch_texel_deriv( GLcontext *ctx, const GLfloat texcoord[4],
 {
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
    const struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current;
-   GLfloat lambda;
-   GLchan rgba[4];
 
    if (texObj) {
-      const struct gl_texture_image *texImg = texObj->Image[0][texObj->BaseLevel];
+      const struct gl_texture_image *texImg =
+         texObj->Image[0][texObj->BaseLevel];
       const GLfloat texW = (GLfloat) texImg->WidthScale;
       const GLfloat texH = (GLfloat) texImg->HeightScale;
+      GLfloat lambda;
+      GLchan rgba[4];
 
       lambda = _swrast_compute_lambda(texdx[0], texdy[0], /* ds/dx, ds/dy */
                                       texdx[1], texdy[1], /* dt/dx, dt/dy */
@@ -85,14 +118,16 @@ fetch_texel_deriv( GLcontext *ctx, const GLfloat texcoord[4],
                                       1.0F / texcoord[3]) + lodBias;
 
       lambda = CLAMP(lambda, texObj->MinLod, texObj->MaxLod);
-   }
 
-   swrast->TextureSample[unit](ctx, texObj, 1, (const GLfloat (*)[4]) texcoord,
-                               &lambda, &rgba);
-   color[0] = CHAN_TO_FLOAT(rgba[0]);
-   color[1] = CHAN_TO_FLOAT(rgba[1]);
-   color[2] = CHAN_TO_FLOAT(rgba[2]);
-   color[3] = CHAN_TO_FLOAT(rgba[3]);
+      /* XXX use a float-valued TextureSample routine here!!! */
+      swrast->TextureSample[unit](ctx, texObj, 1,
+                                  (const GLfloat (*)[4]) texcoord,
+                                  &lambda, &rgba);
+      swizzle_texel(rgba, color, texObj->_Swizzle);
+   }
+   else {
+      ASSIGN_4V(color, 0.0F, 0.0F, 0.0F, 1.0F);
+   }
 }
 
 
diff --git a/src/mesa/swrast/s_imaging.c b/src/mesa/swrast/s_imaging.c
index 591857c342..d6be3aa022 100644
--- a/src/mesa/swrast/s_imaging.c
+++ b/src/mesa/swrast/s_imaging.c
@@ -39,7 +39,6 @@ _swrast_CopyColorTable( GLcontext *ctx,
 			GLenum target, GLenum internalformat,
 			GLint x, GLint y, GLsizei width)
 {
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLchan data[MAX_WIDTH][4];
    struct gl_buffer_object *bufferSave;
 
@@ -51,13 +50,13 @@ _swrast_CopyColorTable( GLcontext *ctx,
    if (width > MAX_WIDTH)
       width = MAX_WIDTH;
 
-   RENDER_START( swrast, ctx );
+   swrast_render_start(ctx);
 
    /* read the data from framebuffer */
    _swrast_read_rgba_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
                            width, x, y, CHAN_TYPE, data );
 
-   RENDER_FINISH(swrast,ctx);
+   swrast_render_finish(ctx);
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
@@ -74,7 +73,6 @@ void
 _swrast_CopyColorSubTable( GLcontext *ctx,GLenum target, GLsizei start,
 			   GLint x, GLint y, GLsizei width)
 {
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLchan data[MAX_WIDTH][4];
    struct gl_buffer_object *bufferSave;
 
@@ -86,13 +84,13 @@ _swrast_CopyColorSubTable( GLcontext *ctx,GLenum target, GLsizei start,
    if (width > MAX_WIDTH)
       width = MAX_WIDTH;
 
-   RENDER_START( swrast, ctx );
+   swrast_render_start(ctx);
 
    /* read the data from framebuffer */
    _swrast_read_rgba_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
                            width, x, y, CHAN_TYPE, data );
 
-   RENDER_FINISH(swrast,ctx);
+   swrast_render_finish(ctx);
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
@@ -110,7 +108,6 @@ _swrast_CopyConvolutionFilter1D(GLcontext *ctx, GLenum target,
 				GLenum internalFormat, 
 				GLint x, GLint y, GLsizei width)
 {
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLchan rgba[MAX_CONVOLUTION_WIDTH][4];
    struct gl_buffer_object *bufferSave;
 
@@ -119,13 +116,13 @@ _swrast_CopyConvolutionFilter1D(GLcontext *ctx, GLenum target,
       return;
    }
 
-   RENDER_START( swrast, ctx );
+   swrast_render_start(ctx);
 
    /* read the data from framebuffer */
    _swrast_read_rgba_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
                            width, x, y, CHAN_TYPE, rgba );
    
-   RENDER_FINISH( swrast, ctx );
+   swrast_render_finish(ctx);
 
    /* save PBO binding */
    bufferSave = ctx->Unpack.BufferObj;
@@ -145,7 +142,6 @@ _swrast_CopyConvolutionFilter2D(GLcontext *ctx, GLenum target,
 				GLenum internalFormat, 
 				GLint x, GLint y, GLsizei width, GLsizei height)
 {
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    struct gl_pixelstore_attrib packSave;
    GLchan rgba[MAX_CONVOLUTION_HEIGHT][MAX_CONVOLUTION_WIDTH][4];
    GLint i;
@@ -156,7 +152,7 @@ _swrast_CopyConvolutionFilter2D(GLcontext *ctx, GLenum target,
       return;
    }
 
-   RENDER_START(swrast,ctx);
+   swrast_render_start(ctx);
    
    /* read pixels from framebuffer */
    for (i = 0; i < height; i++) {
@@ -164,7 +160,7 @@ _swrast_CopyConvolutionFilter2D(GLcontext *ctx, GLenum target,
                               width, x, y + i, CHAN_TYPE, rgba[i] );
    }
 
-   RENDER_FINISH(swrast,ctx);
+   swrast_render_finish(ctx);
 
    /*
     * HACK: save & restore context state so we can store this as a
diff --git a/src/mesa/swrast/s_readpix.c b/src/mesa/swrast/s_readpix.c
index f263045170..e901fc6b5d 100644
--- a/src/mesa/swrast/s_readpix.c
+++ b/src/mesa/swrast/s_readpix.c
@@ -555,11 +555,11 @@ _swrast_ReadPixels( GLcontext *ctx,
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
    struct gl_pixelstore_attrib clippedPacking = *packing;
 
-   /* Need to do RENDER_START before clipping or anything else since this
-    * is where a driver may grab the hw lock and get an updated window
-    * size.
+   /* Need to do swrast_render_start() before clipping or anything else
+    * since this is where a driver may grab the hw lock and get an updated
+    * window size.
     */
-   RENDER_START(swrast, ctx);
+   swrast_render_start(ctx);
 
    if (ctx->NewState)
       _mesa_update_state(ctx);
@@ -570,7 +570,7 @@ _swrast_ReadPixels( GLcontext *ctx,
    /* Do all needed clipping here, so that we can forget about it later */
    if (!_mesa_clip_readpixels(ctx, &x, &y, &width, &height, &clippedPacking)) {
       /* The ReadPixels region is totally outside the window bounds */
-      RENDER_FINISH(swrast, ctx);
+      swrast_render_finish(ctx);
       return;
    }
 
@@ -614,7 +614,7 @@ _swrast_ReadPixels( GLcontext *ctx,
          /* don't return yet, clean-up */
    }
 
-   RENDER_FINISH(swrast, ctx);
+   swrast_render_finish(ctx);
 
    _mesa_unmap_readpix_pbo(ctx, &clippedPacking);
 }
diff --git a/src/mesa/swrast/s_span.c b/src/mesa/swrast/s_span.c
index 214c2a1b6f..ab7b82b19d 100644
--- a/src/mesa/swrast/s_span.c
+++ b/src/mesa/swrast/s_span.c
@@ -1032,6 +1032,7 @@ add_specular(GLcontext *ctx, SWspan *span)
    ASSERT(!ctx->FragmentProgram._Current);
    ASSERT(span->arrayMask & SPAN_RGBA);
    ASSERT(swrast->_ActiveAttribMask & FRAG_BIT_COL1);
+   (void) swrast; /* silence warning */
 
    if (span->array->ChanType == GL_FLOAT) {
       if ((span->arrayAttribs & FRAG_BIT_COL0) == 0) {
@@ -1310,6 +1311,7 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
    /* Do the alpha test */
    if (ctx->Color.AlphaEnabled) {
       if (!_swrast_alpha_test(ctx, span)) {
+         /* all fragments failed test */
          goto end;
       }
    }
@@ -1322,6 +1324,7 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
       if (ctx->Stencil.Enabled && fb->Visual.stencilBits > 0) {
          /* Combined Z/stencil tests */
          if (!_swrast_stencil_and_ztest_span(ctx, span)) {
+            /* all fragments failed test */
             goto end;
          }
       }
@@ -1330,6 +1333,7 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
          ASSERT(ctx->Depth.Test);
          ASSERT(span->arrayMask & SPAN_Z);
          if (!_swrast_depth_test_span(ctx, span)) {
+            /* all fragments failed test */
             goto end;
          }
       }
@@ -1349,6 +1353,7 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
     * the occlusion test.
     */
    if (colorMask == 0x0) {
+      /* no colors to write */
       goto end;
    }
 
diff --git a/src/mesa/swrast/s_stencil.c b/src/mesa/swrast/s_stencil.c
index c925922463..2e84ddec71 100644
--- a/src/mesa/swrast/s_stencil.c
+++ b/src/mesa/swrast/s_stencil.c
@@ -997,10 +997,12 @@ stencil_and_ztest_pixels( GLcontext *ctx, SWspan *span, GLuint face )
 GLboolean
 _swrast_stencil_and_ztest_span(GLcontext *ctx, SWspan *span)
 {
+   const GLuint face = (span->facing == 0) ? 0 : ctx->Stencil._BackFace;
+
    if (span->arrayMask & SPAN_XY)
-      return stencil_and_ztest_pixels(ctx, span, span->facing);
+      return stencil_and_ztest_pixels(ctx, span, face);
    else
-      return stencil_and_ztest_span(ctx, span, span->facing);
+      return stencil_and_ztest_span(ctx, span, face);
 }
 
 
diff --git a/src/mesa/swrast/s_texcombine.c b/src/mesa/swrast/s_texcombine.c
index 632d650007..38b5633887 100644
--- a/src/mesa/swrast/s_texcombine.c
+++ b/src/mesa/swrast/s_texcombine.c
@@ -1,8 +1,9 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.1
+ * Version:  7.5
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -30,6 +31,7 @@
 #include "main/imports.h"
 #include "main/macros.h"
 #include "main/pixel.h"
+#include "shader/prog_instruction.h"
 
 #include "s_context.h"
 #include "s_texcombine.h"
@@ -66,9 +68,9 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n,
                  GLchan (*rgba)[4] )
 {
    const struct gl_texture_unit *textureUnit = &(ctx->Texture.Unit[unit]);
-   const GLchan (*argRGB [3])[4];
-   const GLchan (*argA [3])[4];
-   const GLuint RGBshift = textureUnit->_CurrentCombine->ScaleShiftRGB;
+   const GLchan (*argRGB [4])[4];
+   const GLchan (*argA [4])[4];
+   const GLint RGBshift = textureUnit->_CurrentCombine->ScaleShiftRGB;
    const GLuint Ashift   = textureUnit->_CurrentCombine->ScaleShiftA;
 #if CHAN_TYPE == GL_FLOAT
    const GLchan RGBmult = (GLfloat) (1 << RGBshift);
@@ -80,12 +82,12 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n,
    static const GLchan zero[4] = { 0, 0, 0, 0 };
    const GLuint numColorArgs = textureUnit->_CurrentCombine->_NumArgsRGB;
    const GLuint numAlphaArgs = textureUnit->_CurrentCombine->_NumArgsA;
-   GLchan ccolor[3][MAX_WIDTH][4];
+   GLchan ccolor[4][MAX_WIDTH][4];
    GLuint i, j;
 
    ASSERT(ctx->Extensions.EXT_texture_env_combine ||
           ctx->Extensions.ARB_texture_env_combine);
-   ASSERT(SWRAST_CONTEXT(ctx)->_AnyTextureCombine);
+   ASSERT(CONST_SWRAST_CONTEXT(ctx)->_AnyTextureCombine);
 
    /*
    printf("modeRGB 0x%x  modeA 0x%x  srcRGB1 0x%x  srcA1 0x%x  srcRGB2 0x%x  srcA2 0x%x\n",
@@ -98,7 +100,7 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n,
    */
 
    /*
-    * Do operand setup for up to 3 operands.  Loop over the terms.
+    * Do operand setup for up to 4 operands.  Loop over the terms.
     */
    for (j = 0; j < numColorArgs; j++) {
       const GLenum srcRGB = textureUnit->_CurrentCombine->SourceRGB[j];
@@ -295,7 +297,36 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n,
          }
          break;
       case GL_ADD:
-         {
+         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
+            /* (a * b) + (c * d) */
+            const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0];
+            const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1];
+            const GLchan (*arg2)[4] = (const GLchan (*)[4]) argRGB[2];
+            const GLchan (*arg3)[4] = (const GLchan (*)[4]) argRGB[3];
+            for (i = 0; i < n; i++) {
+#if CHAN_TYPE == GL_FLOAT
+               rgba[i][RCOMP] = (arg0[i][RCOMP] * arg1[i][RCOMP] +
+                                 arg2[i][RCOMP] * arg3[i][RCOMP]) * RGBmult;
+               rgba[i][GCOMP] = (arg0[i][GCOMP] * arg1[i][GCOMP] +
+                                 arg2[i][GCOMP] * arg3[i][GCOMP]) * RGBmult;
+               rgba[i][BCOMP] = (arg0[i][BCOMP] * arg1[i][BCOMP] +
+                                 arg2[i][BCOMP] * arg3[i][BCOMP]) * RGBmult;
+#else
+               const GLint shift = CHAN_BITS - RGBshift;
+               GLint r = (PROD(arg0[i][RCOMP], arg1[i][RCOMP]) >> shift) +
+                         (PROD(arg2[i][RCOMP], arg3[i][RCOMP]) >> shift);
+               GLint g = (PROD(arg0[i][GCOMP], arg1[i][GCOMP]) >> shift) +
+                         (PROD(arg2[i][GCOMP], arg3[i][GCOMP]) >> shift);
+               GLint b = (PROD(arg0[i][BCOMP], arg1[i][BCOMP]) >> shift) +
+                         (PROD(arg2[i][BCOMP], arg3[i][BCOMP]) >> shift);
+               rgba[i][RCOMP] = (GLchan) MIN2(r, CHAN_MAX);
+               rgba[i][GCOMP] = (GLchan) MIN2(g, CHAN_MAX);
+               rgba[i][BCOMP] = (GLchan) MIN2(b, CHAN_MAX);
+#endif
+            }
+         }
+         else {
+            /* 2-term addition */
             const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0];
             const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1];
             for (i = 0; i < n; i++) {
@@ -315,7 +346,37 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n,
          }
          break;
       case GL_ADD_SIGNED:
-         {
+         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
+            /* (a * b) + (c * d) - 0.5 */
+            const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0];
+            const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1];
+            const GLchan (*arg2)[4] = (const GLchan (*)[4]) argRGB[2];
+            const GLchan (*arg3)[4] = (const GLchan (*)[4]) argRGB[3];
+            for (i = 0; i < n; i++) {
+#if CHAN_TYPE == GL_FLOAT
+               rgba[i][RCOMP] = (arg0[i][RCOMP] + arg1[i][RCOMP] *
+                                 arg2[i][RCOMP] + arg3[i][RCOMP] - 0.5) * RGBmult;
+               rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP] *
+                                 arg2[i][GCOMP] + arg3[i][GCOMP] - 0.5) * RGBmult;
+               rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP] *
+                                 arg2[i][BCOMP] + arg3[i][BCOMP] - 0.5) * RGBmult;
+#else
+               GLint r = (((PROD(arg0[i][RCOMP], arg1[i][RCOMP]) +
+                            PROD(arg2[i][RCOMP], arg3[i][RCOMP])) >> CHAN_BITS) - half)
+                          << RGBshift;
+               GLint g = (((PROD(arg0[i][GCOMP], arg1[i][GCOMP]) +
+                            PROD(arg2[i][GCOMP], arg3[i][GCOMP])) >> CHAN_BITS) - half)
+                          << RGBshift;
+               GLint b = (((PROD(arg0[i][BCOMP], arg1[i][BCOMP]) +
+                            PROD(arg2[i][BCOMP], arg3[i][BCOMP])) >> CHAN_BITS) - half)
+                          << RGBshift;
+               rgba[i][RCOMP] = (GLchan) CLAMP(r, 0, CHAN_MAX);
+               rgba[i][GCOMP] = (GLchan) CLAMP(g, 0, CHAN_MAX);
+               rgba[i][BCOMP] = (GLchan) CLAMP(b, 0, CHAN_MAX);
+#endif
+            }
+         }
+         else {
             const GLchan (*arg0)[4] = (const GLchan (*)[4]) argRGB[0];
             const GLchan (*arg1)[4] = (const GLchan (*)[4]) argRGB[1];
             for (i = 0; i < n; i++) {
@@ -324,9 +385,9 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n,
                rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP] - 0.5) * RGBmult;
                rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP] - 0.5) * RGBmult;
 #else
-               GLint r = (GLint) arg0[i][RCOMP] + (GLint) arg1[i][RCOMP] -half;
-               GLint g = (GLint) arg0[i][GCOMP] + (GLint) arg1[i][GCOMP] -half;
-               GLint b = (GLint) arg0[i][BCOMP] + (GLint) arg1[i][BCOMP] -half;
+               GLint r = (GLint) arg0[i][RCOMP] + (GLint) arg1[i][RCOMP] - half;
+               GLint g = (GLint) arg0[i][GCOMP] + (GLint) arg1[i][GCOMP] - half;
+               GLint b = (GLint) arg0[i][BCOMP] + (GLint) arg1[i][BCOMP] - half;
                r = (r < 0) ? 0 : r << RGBshift;
                g = (g < 0) ? 0 : g << RGBshift;
                b = (b < 0) ? 0 : b << RGBshift;
@@ -573,9 +634,28 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n,
          }
          break;
       case GL_ADD:
-         {
+         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
+            /* (a * b) + (c * d) */
             const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0];
-            const GLchan  (*arg1)[4] = (const GLchan (*)[4]) argA[1];
+            const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1];
+            const GLchan (*arg2)[4] = (const GLchan (*)[4]) argA[2];
+            const GLchan (*arg3)[4] = (const GLchan (*)[4]) argA[3];
+            for (i = 0; i < n; i++) {
+#if CHAN_TYPE == GL_FLOAT
+               rgba[i][ACOMP] = (arg0[i][ACOMP] * arg1[i][ACOMP] +
+                                 arg2[i][ACOMP] * arg3[i][ACOMP]) * Amult;
+#else
+               const GLint shift = CHAN_BITS - Ashift;
+               GLint a = (PROD(arg0[i][ACOMP], arg1[i][ACOMP]) >> shift) +
+                         (PROD(arg2[i][ACOMP], arg3[i][ACOMP]) >> shift);
+               rgba[i][ACOMP] = (GLchan) MIN2(a, CHAN_MAX);
+#endif
+            }
+         }
+         else {
+            /* two-term add */
+            const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0];
+            const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1];
             for (i = 0; i < n; i++) {
 #if CHAN_TYPE == GL_FLOAT
                rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP]) * Amult;
@@ -587,7 +667,27 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n,
          }
          break;
       case GL_ADD_SIGNED:
-         {
+         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
+            /* (a * b) + (c * d) - 0.5 */
+            const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0];
+            const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1];
+            const GLchan (*arg2)[4] = (const GLchan (*)[4]) argA[2];
+            const GLchan (*arg3)[4] = (const GLchan (*)[4]) argA[3];
+            for (i = 0; i < n; i++) {
+#if CHAN_TYPE == GL_FLOAT
+               rgba[i][ACOMP] = (arg0[i][ACOMP] * arg1[i][ACOMP] +
+                                 arg2[i][ACOMP] * arg3[i][ACOMP] -
+                                 0.5) * Amult;
+#else
+               GLint a = (((PROD(arg0[i][ACOMP], arg1[i][ACOMP]) +
+                            PROD(arg2[i][ACOMP], arg3[i][ACOMP])) >> CHAN_BITS) - half)
+                          << Ashift;
+               rgba[i][ACOMP] = (GLchan) CLAMP(a, 0, CHAN_MAX);
+#endif
+            }
+         }
+         else {
+            /* a + b - 0.5 */
             const GLchan (*arg0)[4] = (const GLchan (*)[4]) argA[0];
             const GLchan (*arg1)[4] = (const GLchan (*)[4]) argA[1];
             for (i = 0; i < n; i++) {
@@ -596,7 +696,7 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n,
 #else
                GLint a = (GLint) arg0[i][ACOMP] + (GLint) arg1[i][ACOMP] -half;
                a = (a < 0) ? 0 : a << Ashift;
-               rgba[i][ACOMP] = (GLchan) MIN2(a, CHAN_MAX);
+               rgba[i][ACOMP] = (GLchan) CLAMP(a, 0, CHAN_MAX);
 #endif
             }
          }
@@ -717,6 +817,36 @@ texture_combine( const GLcontext *ctx, GLuint unit, GLuint n,
 
 
 /**
+ * Apply X/Y/Z/W/0/1 swizzle to an array of colors/texels.
+ * See GL_EXT_texture_swizzle.
+ */
+static void
+swizzle_texels(GLuint swizzle, GLuint count, GLchan (*texels)[4])
+{
+   const GLuint swzR = GET_SWZ(swizzle, 0);
+   const GLuint swzG = GET_SWZ(swizzle, 1);
+   const GLuint swzB = GET_SWZ(swizzle, 2);
+   const GLuint swzA = GET_SWZ(swizzle, 3);
+   GLchan vector[6];
+   GLuint i;
+
+   vector[SWIZZLE_ZERO] = 0;
+   vector[SWIZZLE_ONE] = CHAN_MAX;
+
+   for (i = 0; i < count; i++) {
+      vector[SWIZZLE_X] = texels[i][0];
+      vector[SWIZZLE_Y] = texels[i][1];
+      vector[SWIZZLE_Z] = texels[i][2];
+      vector[SWIZZLE_W] = texels[i][3];
+      texels[i][RCOMP] = vector[swzR];
+      texels[i][GCOMP] = vector[swzG];
+      texels[i][BCOMP] = vector[swzB];
+      texels[i][ACOMP] = vector[swzA];
+   }
+}
+
+
+/**
  * Apply a conventional OpenGL texture env mode (REPLACE, ADD, BLEND,
  * MODULATE, or DECAL) to an array of fragments.
  * Input:  textureUnit - pointer to texture unit to apply
@@ -1142,9 +1272,15 @@ _swrast_texture_span( GLcontext *ctx, SWspan *span )
             _mesa_lookup_rgba_float(&texUnit->ColorTable, span->end, texels);
 #endif
          }
+
+         /* GL_EXT_texture_swizzle */
+         if (curObj->_Swizzle != SWIZZLE_NOOP) {
+            swizzle_texels(curObj->_Swizzle, span->end, texels);
+         }
       }
    }
 
+
    /*
     * OK, now apply the texture (aka texture combine/blend).
     * We modify the span->color.rgba values.
@@ -1163,6 +1299,8 @@ _swrast_texture_span( GLcontext *ctx, SWspan *span )
             const GLchan (*texels)[4] = (const GLchan (*)[4])
                (swrast->TexelBuffer + unit *
                 (span->end * 4 * sizeof(GLchan)));
+
+
             texture_apply( ctx, texUnit, span->end,
                            (CONST GLchan (*)[4]) primary_rgba, texels,
                            span->array->rgba );
diff --git a/src/mesa/swrast/s_texfilter.c b/src/mesa/swrast/s_texfilter.c
index 9e44fba3da..8d72018cf4 100644
--- a/src/mesa/swrast/s_texfilter.c
+++ b/src/mesa/swrast/s_texfilter.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.0.3
+ * Version:  7.3
  *
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -33,6 +33,16 @@
 #include "s_texfilter.h"
 
 
+/*
+ * Note, the FRAC macro has to work perfectly.  Otherwise you'll sometimes
+ * see 1-pixel bands of improperly weighted linear-filtered textures.
+ * The tests/texwrap.c demo is a good test.
+ * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0.
+ * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x).
+ */
+#define FRAC(f)  ((f) - IFLOOR(f))
+
+
 /**
  * Constants for integer linear interpolation.
  */
@@ -223,257 +233,271 @@ lerp_rgba_3d(GLchan result[4], GLfloat a, GLfloat b, GLfloat c,
  * Used to compute texel locations for linear sampling.
  * Input:
  *    wrapMode = GL_REPEAT, GL_CLAMP, GL_CLAMP_TO_EDGE, GL_CLAMP_TO_BORDER
- *    S = texcoord in [0,1]
- *    SIZE = width (or height or depth) of texture
+ *    s = texcoord in [0,1]
+ *    size = width (or height or depth) of texture
  * Output:
- *    U = texcoord in [0, width]
- *    I0, I1 = two nearest texel indexes
+ *    i0, i1 = returns two nearest texel indexes
+ *    weight = returns blend factor between texels
  */
-#define COMPUTE_LINEAR_TEXEL_LOCATIONS(wrapMode, S, U, SIZE, I0, I1)	\
-{									\
-   switch (wrapMode) {							\
-   case GL_REPEAT:							\
-      U = S * SIZE - 0.5F;						\
-      if (img->_IsPowerOfTwo) {						\
-         I0 = IFLOOR(U) & (SIZE - 1);					\
-         I1 = (I0 + 1) & (SIZE - 1);					\
-      }									\
-      else {								\
-         I0 = REMAINDER(IFLOOR(U), SIZE);				\
-         I1 = REMAINDER(I0 + 1, SIZE);					\
-      }									\
-      break;								\
-   case GL_CLAMP_TO_EDGE:						\
-      if (S <= 0.0F)							\
-         U = 0.0F;							\
-      else if (S >= 1.0F)						\
-         U = (GLfloat) SIZE;						\
-      else								\
-         U = S * SIZE;							\
-      U -= 0.5F;							\
-      I0 = IFLOOR(U);							\
-      I1 = I0 + 1;							\
-      if (I0 < 0)							\
-         I0 = 0;							\
-      if (I1 >= (GLint) SIZE)						\
-         I1 = SIZE - 1;							\
-      break;								\
-   case GL_CLAMP_TO_BORDER:						\
-      {									\
-         const GLfloat min = -1.0F / (2.0F * SIZE);			\
-         const GLfloat max = 1.0F - min;				\
-         if (S <= min)							\
-            U = min * SIZE;						\
-         else if (S >= max)						\
-            U = max * SIZE;						\
-         else								\
-            U = S * SIZE;						\
-         U -= 0.5F;							\
-         I0 = IFLOOR(U);						\
-         I1 = I0 + 1;							\
-      }									\
-      break;								\
-   case GL_MIRRORED_REPEAT:						\
-      {									\
-         const GLint flr = IFLOOR(S);					\
-         if (flr & 1)							\
-            U = 1.0F - (S - (GLfloat) flr);	/* flr is odd */	\
-         else								\
-            U = S - (GLfloat) flr;		/* flr is even */	\
-         U = (U * SIZE) - 0.5F;						\
-         I0 = IFLOOR(U);						\
-         I1 = I0 + 1;							\
-         if (I0 < 0)							\
-            I0 = 0;							\
-         if (I1 >= (GLint) SIZE)					\
-            I1 = SIZE - 1;						\
-      }									\
-      break;								\
-   case GL_MIRROR_CLAMP_EXT:						\
-      U = FABSF(S);							\
-      if (U >= 1.0F)							\
-         U = (GLfloat) SIZE;						\
-      else								\
-         U *= SIZE;							\
-      U -= 0.5F;							\
-      I0 = IFLOOR(U);							\
-      I1 = I0 + 1;							\
-      break;								\
-   case GL_MIRROR_CLAMP_TO_EDGE_EXT:					\
-      U = FABSF(S);							\
-      if (U >= 1.0F)							\
-         U = (GLfloat) SIZE;						\
-      else								\
-         U *= SIZE;							\
-      U -= 0.5F;							\
-      I0 = IFLOOR(U);							\
-      I1 = I0 + 1;							\
-      if (I0 < 0)							\
-         I0 = 0;							\
-      if (I1 >= (GLint) SIZE)						\
-         I1 = SIZE - 1;							\
-      break;								\
-   case GL_MIRROR_CLAMP_TO_BORDER_EXT:					\
-      {									\
-         const GLfloat min = -1.0F / (2.0F * SIZE);			\
-         const GLfloat max = 1.0F - min;				\
-         U = FABSF(S);							\
-         if (U <= min)							\
-            U = min * SIZE;						\
-         else if (U >= max)						\
-            U = max * SIZE;						\
-         else								\
-            U *= SIZE;							\
-         U -= 0.5F;							\
-         I0 = IFLOOR(U);						\
-         I1 = I0 + 1;							\
-      }									\
-      break;								\
-   case GL_CLAMP:							\
-      if (S <= 0.0F)							\
-         U = 0.0F;							\
-      else if (S >= 1.0F)						\
-         U = (GLfloat) SIZE;						\
-      else								\
-         U = S * SIZE;							\
-      U -= 0.5F;							\
-      I0 = IFLOOR(U);							\
-      I1 = I0 + 1;							\
-      break;								\
-   default:								\
-      _mesa_problem(ctx, "Bad wrap mode");				\
-      return;								\
-   }									\
+static INLINE void
+linear_texel_locations(GLenum wrapMode,
+                       const struct gl_texture_image *img,
+                       GLint size, GLfloat s,
+                       GLint *i0, GLint *i1, GLfloat *weight)
+{
+   GLfloat u;
+   switch (wrapMode) {
+   case GL_REPEAT:
+      u = s * size - 0.5F;
+      if (img->_IsPowerOfTwo) {
+         *i0 = IFLOOR(u) & (size - 1);
+         *i1 = (*i0 + 1) & (size - 1);
+      }
+      else {
+         *i0 = REMAINDER(IFLOOR(u), size);
+         *i1 = REMAINDER(*i0 + 1, size);
+      }
+      break;
+   case GL_CLAMP_TO_EDGE:
+      if (s <= 0.0F)
+         u = 0.0F;
+      else if (s >= 1.0F)
+         u = (GLfloat) size;
+      else
+         u = s * size;
+      u -= 0.5F;
+      *i0 = IFLOOR(u);
+      *i1 = *i0 + 1;
+      if (*i0 < 0)
+         *i0 = 0;
+      if (*i1 >= (GLint) size)
+         *i1 = size - 1;
+      break;
+   case GL_CLAMP_TO_BORDER:
+      {
+         const GLfloat min = -1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         if (s <= min)
+            u = min * size;
+         else if (s >= max)
+            u = max * size;
+         else
+            u = s * size;
+         u -= 0.5F;
+         *i0 = IFLOOR(u);
+         *i1 = *i0 + 1;
+      }
+      break;
+   case GL_MIRRORED_REPEAT:
+      {
+         const GLint flr = IFLOOR(s);
+         if (flr & 1)
+            u = 1.0F - (s - (GLfloat) flr);
+         else
+            u = s - (GLfloat) flr;
+         u = (u * size) - 0.5F;
+         *i0 = IFLOOR(u);
+         *i1 = *i0 + 1;
+         if (*i0 < 0)
+            *i0 = 0;
+         if (*i1 >= (GLint) size)
+            *i1 = size - 1;
+      }
+      break;
+   case GL_MIRROR_CLAMP_EXT:
+      u = FABSF(s);
+      if (u >= 1.0F)
+         u = (GLfloat) size;
+      else
+         u *= size;
+      u -= 0.5F;
+      *i0 = IFLOOR(u);
+      *i1 = *i0 + 1;
+      break;
+   case GL_MIRROR_CLAMP_TO_EDGE_EXT:
+      u = FABSF(s);
+      if (u >= 1.0F)
+         u = (GLfloat) size;
+      else
+         u *= size;
+      u -= 0.5F;
+      *i0 = IFLOOR(u);
+      *i1 = *i0 + 1;
+      if (*i0 < 0)
+         *i0 = 0;
+      if (*i1 >= (GLint) size)
+         *i1 = size - 1;
+      break;
+   case GL_MIRROR_CLAMP_TO_BORDER_EXT:
+      {
+         const GLfloat min = -1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         u = FABSF(s);
+         if (u <= min)
+            u = min * size;
+         else if (u >= max)
+            u = max * size;
+         else
+            u *= size;
+         u -= 0.5F;
+         *i0 = IFLOOR(u);
+         *i1 = *i0 + 1;
+      }
+      break;
+   case GL_CLAMP:
+      if (s <= 0.0F)
+         u = 0.0F;
+      else if (s >= 1.0F)
+         u = (GLfloat) size;
+      else
+         u = s * size;
+      u -= 0.5F;
+      *i0 = IFLOOR(u);
+      *i1 = *i0 + 1;
+      break;
+   default:
+      _mesa_problem(NULL, "Bad wrap mode");
+      u = 0.0F;
+   }
+   *weight = FRAC(u);
 }
 
 
 /**
  * Used to compute texel location for nearest sampling.
  */
-#define COMPUTE_NEAREST_TEXEL_LOCATION(wrapMode, S, SIZE, I)		\
-{									\
-   switch (wrapMode) {							\
-   case GL_REPEAT:							\
-      /* s limited to [0,1) */						\
-      /* i limited to [0,size-1] */					\
-      I = IFLOOR(S * SIZE);						\
-      if (img->_IsPowerOfTwo)						\
-         I &= (SIZE - 1);						\
-      else								\
-         I = REMAINDER(I, SIZE);					\
-      break;								\
-   case GL_CLAMP_TO_EDGE:						\
-      {									\
-         /* s limited to [min,max] */					\
-         /* i limited to [0, size-1] */					\
-         const GLfloat min = 1.0F / (2.0F * SIZE);			\
-         const GLfloat max = 1.0F - min;				\
-         if (S < min)							\
-            I = 0;							\
-         else if (S > max)						\
-            I = SIZE - 1;						\
-         else								\
-            I = IFLOOR(S * SIZE);					\
-      }									\
-      break;								\
-   case GL_CLAMP_TO_BORDER:						\
-      {									\
-         /* s limited to [min,max] */					\
-         /* i limited to [-1, size] */					\
-         const GLfloat min = -1.0F / (2.0F * SIZE);			\
-         const GLfloat max = 1.0F - min;				\
-         if (S <= min)							\
-            I = -1;							\
-         else if (S >= max)						\
-            I = SIZE;							\
-         else								\
-            I = IFLOOR(S * SIZE);					\
-      }									\
-      break;								\
-   case GL_MIRRORED_REPEAT:						\
-      {									\
-         const GLfloat min = 1.0F / (2.0F * SIZE);			\
-         const GLfloat max = 1.0F - min;				\
-         const GLint flr = IFLOOR(S);					\
-         GLfloat u;							\
-         if (flr & 1)							\
-            u = 1.0F - (S - (GLfloat) flr);	/* flr is odd */	\
-         else								\
-            u = S - (GLfloat) flr;		/* flr is even */	\
-         if (u < min)							\
-            I = 0;							\
-         else if (u > max)						\
-            I = SIZE - 1;						\
-         else								\
-            I = IFLOOR(u * SIZE);					\
-      }									\
-      break;								\
-   case GL_MIRROR_CLAMP_EXT:						\
-      {									\
-         /* s limited to [0,1] */					\
-         /* i limited to [0,size-1] */					\
-         const GLfloat u = FABSF(S);					\
-         if (u <= 0.0F)							\
-            I = 0;							\
-         else if (u >= 1.0F)						\
-            I = SIZE - 1;						\
-         else								\
-            I = IFLOOR(u * SIZE);					\
-      }									\
-      break;								\
-   case GL_MIRROR_CLAMP_TO_EDGE_EXT:					\
-      {									\
-         /* s limited to [min,max] */					\
-         /* i limited to [0, size-1] */					\
-         const GLfloat min = 1.0F / (2.0F * SIZE);			\
-         const GLfloat max = 1.0F - min;				\
-         const GLfloat u = FABSF(S);					\
-         if (u < min)							\
-            I = 0;							\
-         else if (u > max)						\
-            I = SIZE - 1;						\
-         else								\
-            I = IFLOOR(u * SIZE);					\
-      }									\
-      break;								\
-   case GL_MIRROR_CLAMP_TO_BORDER_EXT:					\
-      {									\
-         /* s limited to [min,max] */					\
-         /* i limited to [0, size-1] */					\
-         const GLfloat min = -1.0F / (2.0F * SIZE);			\
-         const GLfloat max = 1.0F - min;				\
-         const GLfloat u = FABSF(S);					\
-         if (u < min)							\
-            I = -1;							\
-         else if (u > max)						\
-            I = SIZE;							\
-         else								\
-            I = IFLOOR(u * SIZE);					\
-      }									\
-      break;								\
-   case GL_CLAMP:							\
-      /* s limited to [0,1] */						\
-      /* i limited to [0,size-1] */					\
-      if (S <= 0.0F)							\
-         I = 0;								\
-      else if (S >= 1.0F)						\
-         I = SIZE - 1;							\
-      else								\
-         I = IFLOOR(S * SIZE);						\
-      break;								\
-   default:								\
-      _mesa_problem(ctx, "Bad wrap mode");				\
-      return;								\
-   }									\
+static INLINE GLint
+nearest_texel_location(GLenum wrapMode,
+                       const struct gl_texture_image *img,
+                       GLint size, GLfloat s)
+{
+   GLint i;
+
+   switch (wrapMode) {
+   case GL_REPEAT:
+      /* s limited to [0,1) */
+      /* i limited to [0,size-1] */
+      i = IFLOOR(s * size);
+      if (img->_IsPowerOfTwo)
+         i &= (size - 1);
+      else
+         i = REMAINDER(i, size);
+      return i;
+   case GL_CLAMP_TO_EDGE:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const GLfloat min = 1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         if (s < min)
+            i = 0;
+         else if (s > max)
+            i = size - 1;
+         else
+            i = IFLOOR(s * size);
+      }
+      return i;
+   case GL_CLAMP_TO_BORDER:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [-1, size] */
+         const GLfloat min = -1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         if (s <= min)
+            i = -1;
+         else if (s >= max)
+            i = size;
+         else
+            i = IFLOOR(s * size);
+      }
+      return i;
+   case GL_MIRRORED_REPEAT:
+      {
+         const GLfloat min = 1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         const GLint flr = IFLOOR(s);
+         GLfloat u;
+         if (flr & 1)
+            u = 1.0F - (s - (GLfloat) flr);
+         else
+            u = s - (GLfloat) flr;
+         if (u < min)
+            i = 0;
+         else if (u > max)
+            i = size - 1;
+         else
+            i = IFLOOR(u * size);
+      }
+      return i;
+   case GL_MIRROR_CLAMP_EXT:
+      {
+         /* s limited to [0,1] */
+         /* i limited to [0,size-1] */
+         const GLfloat u = FABSF(s);
+         if (u <= 0.0F)
+            i = 0;
+         else if (u >= 1.0F)
+            i = size - 1;
+         else
+            i = IFLOOR(u * size);
+      }
+      return i;
+   case GL_MIRROR_CLAMP_TO_EDGE_EXT:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const GLfloat min = 1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         const GLfloat u = FABSF(s);
+         if (u < min)
+            i = 0;
+         else if (u > max)
+            i = size - 1;
+         else
+            i = IFLOOR(u * size);
+      }
+      return i;
+   case GL_MIRROR_CLAMP_TO_BORDER_EXT:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const GLfloat min = -1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         const GLfloat u = FABSF(s);
+         if (u < min)
+            i = -1;
+         else if (u > max)
+            i = size;
+         else
+            i = IFLOOR(u * size);
+      }
+      return i;
+   case GL_CLAMP:
+      /* s limited to [0,1] */
+      /* i limited to [0,size-1] */
+      if (s <= 0.0F)
+         i = 0;
+      else if (s >= 1.0F)
+         i = size - 1;
+      else
+         i = IFLOOR(s * size);
+      return i;
+   default:
+      _mesa_problem(NULL, "Bad wrap mode");
+      return 0;
+   }
 }
 
 
 /* Power of two image sizes only */
-#define COMPUTE_LINEAR_REPEAT_TEXEL_LOCATION(S, U, SIZE, I0, I1)	\
-{									\
-   U = S * SIZE - 0.5F;							\
-   I0 = IFLOOR(U) & (SIZE - 1);						\
-   I1 = (I0 + 1) & (SIZE - 1);						\
+static INLINE void
+linear_repeat_texel_location(GLuint size, GLfloat s,
+                             GLint *i0, GLint *i1, GLfloat *weight)
+{
+   GLfloat u = s * size - 0.5F;
+   *i0 = IFLOOR(u) & (size - 1);
+   *i1 = (*i0 + 1) & (size - 1);
+   *weight = FRAC(u);
 }
 
 
@@ -516,17 +540,6 @@ nearest_mipmap_level(const struct gl_texture_object *tObj, GLfloat lambda)
 
 
 /*
- * Note, the FRAC macro has to work perfectly.  Otherwise you'll sometimes
- * see 1-pixel bands of improperly weighted linear-filtered textures.
- * The tests/texwrap.c demo is a good test.
- * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0.
- * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x).
- */
-#define FRAC(f)  ((f) - IFLOOR(f))
-
-
-
-/*
  * Bitflags for texture border color sampling.
  */
 #define I0BIT   1
@@ -538,7 +551,7 @@ nearest_mipmap_level(const struct gl_texture_object *tObj, GLfloat lambda)
 
 
 
-/*
+/**
  * The lambda[] array values are always monotonic.  Either the whole span
  * will be minified, magnified, or split between the two.  This function
  * determines the subranges in [0, n-1] that are to be minified or magnified.
@@ -651,10 +664,10 @@ compute_min_mag_ranges(const struct gl_texture_object *tObj,
 /*                    1-D Texture Sampling Functions                  */
 /**********************************************************************/
 
-/*
+/**
  * Return the texture sample for coordinate (s) using GL_NEAREST filter.
  */
-static void
+static INLINE void
 sample_1d_nearest(GLcontext *ctx,
                   const struct gl_texture_object *tObj,
                   const struct gl_texture_image *img,
@@ -662,7 +675,7 @@ sample_1d_nearest(GLcontext *ctx,
 {
    const GLint width = img->Width2;  /* without border, power of two */
    GLint i;
-   COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapS, texcoord[0], width, i);
+   i = nearest_texel_location(tObj->WrapS, img, width, texcoord[0]);
    /* skip over the border, if any */
    i += img->Border;
    if (i < 0 || i >= (GLint) img->Width) {
@@ -675,10 +688,10 @@ sample_1d_nearest(GLcontext *ctx,
 }
 
 
-/*
+/**
  * Return the texture sample for coordinate (s) using GL_LINEAR filter.
  */
-static void
+static INLINE void
 sample_1d_linear(GLcontext *ctx,
                  const struct gl_texture_object *tObj,
                  const struct gl_texture_image *img,
@@ -686,12 +699,11 @@ sample_1d_linear(GLcontext *ctx,
 {
    const GLint width = img->Width2;
    GLint i0, i1;
-   GLfloat u;
    GLbitfield useBorderColor = 0x0;
    GLfloat a;
    GLchan t0[4], t1[4];  /* texels */
 
-   COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapS, texcoord[0], u, width, i0, i1);
+   linear_texel_locations(tObj->WrapS, img, width, texcoord[0], &i0, &i1, &a);
 
    if (img->Border) {
       i0 += img->Border;
@@ -716,7 +728,6 @@ sample_1d_linear(GLcontext *ctx,
       img->FetchTexelc(img, i1, 0, 0, t1);
    }
 
-   a = FRAC(u);
    lerp_rgba(rgba, a, t0, t1);
 }
 
@@ -776,7 +787,6 @@ sample_1d_nearest_mipmap_linear(GLcontext *ctx,
 }
 
 
-
 static void
 sample_1d_linear_mipmap_linear(GLcontext *ctx,
                                const struct gl_texture_object *tObj,
@@ -802,7 +812,7 @@ sample_1d_linear_mipmap_linear(GLcontext *ctx,
 }
 
 
-
+/** Sample 1D texture, nearest filtering for both min/magnification */
 static void
 sample_nearest_1d( GLcontext *ctx,
                    const struct gl_texture_object *tObj, GLuint n,
@@ -812,13 +822,13 @@ sample_nearest_1d( GLcontext *ctx,
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
    (void) lambda;
-   for (i=0;i<n;i++) {
+   for (i = 0; i < n; i++) {
       sample_1d_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
    }
 }
 
 
-
+/** Sample 1D texture, linear filtering for both min/magnification */
 static void
 sample_linear_1d( GLcontext *ctx,
                   const struct gl_texture_object *tObj, GLuint n,
@@ -828,17 +838,13 @@ sample_linear_1d( GLcontext *ctx,
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
    (void) lambda;
-   for (i=0;i<n;i++) {
+   for (i = 0; i < n; i++) {
       sample_1d_linear(ctx, tObj, image, texcoords[i], rgba[i]);
    }
 }
 
 
-/*
- * Given an (s) texture coordinate and lambda (level of detail) value,
- * return a texture sample.
- *
- */
+/** Sample 1D texture, using lambda to choose between min/magnification */
 static void
 sample_lambda_1d( GLcontext *ctx,
                   const struct gl_texture_object *tObj, GLuint n,
@@ -915,7 +921,7 @@ sample_lambda_1d( GLcontext *ctx,
 /**********************************************************************/
 
 
-/*
+/**
  * Return the texture sample for coordinate (s,t) using GL_NEAREST filter.
  */
 static INLINE void
@@ -930,8 +936,8 @@ sample_2d_nearest(GLcontext *ctx,
    GLint i, j;
    (void) ctx;
 
-   COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapS, texcoord[0], width,  i);
-   COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapT, texcoord[1], height, j);
+   i = nearest_texel_location(tObj->WrapS, img, width, texcoord[0]);
+   j = nearest_texel_location(tObj->WrapT, img, height, texcoord[1]);
 
    /* skip over the border, if any */
    i += img->Border;
@@ -947,7 +953,6 @@ sample_2d_nearest(GLcontext *ctx,
 }
 
 
-
 /**
  * Return the texture sample for coordinate (s,t) using GL_LINEAR filter.
  * New sampling code contributed by Lynn Quam <quam@ai.sri.com>.
@@ -963,12 +968,11 @@ sample_2d_linear(GLcontext *ctx,
    const GLint height = img->Height2;
    GLint i0, j0, i1, j1;
    GLbitfield useBorderColor = 0x0;
-   GLfloat u, v;
    GLfloat a, b;
    GLchan t00[4], t10[4], t01[4], t11[4]; /* sampled texel colors */
 
-   COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapS, texcoord[0], u, width,  i0, i1);
-   COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapT, texcoord[1], v, height, j0, j1);
+   linear_texel_locations(tObj->WrapS, img, width, texcoord[0],  &i0, &i1, &a);
+   linear_texel_locations(tObj->WrapT, img, height, texcoord[1], &j0, &j1, &b);
 
    if (img->Border) {
       i0 += img->Border;
@@ -1009,13 +1013,11 @@ sample_2d_linear(GLcontext *ctx,
       img->FetchTexelc(img, i1, j1, 0, t11);
    }
 
-   a = FRAC(u);
-   b = FRAC(v);
    lerp_rgba_2d(rgba, a, b, t00, t10, t01, t11);
 }
 
 
-/*
+/**
  * As above, but we know WRAP_S == REPEAT and WRAP_T == REPEAT.
  * We don't have to worry about the texture border.
  */
@@ -1029,8 +1031,7 @@ sample_2d_linear_repeat(GLcontext *ctx,
    const GLint width = img->Width2;
    const GLint height = img->Height2;
    GLint i0, j0, i1, j1;
-   GLfloat u, v;
-   GLfloat a, b;
+   GLfloat wi, wj;
    GLchan t00[4], t10[4], t01[4], t11[4]; /* sampled texel colors */
 
    (void) ctx;
@@ -1041,21 +1042,18 @@ sample_2d_linear_repeat(GLcontext *ctx,
    ASSERT(img->TexFormat->BaseFormat != GL_COLOR_INDEX);
    ASSERT(img->_IsPowerOfTwo);
 
-   COMPUTE_LINEAR_REPEAT_TEXEL_LOCATION(texcoord[0], u, width,  i0, i1);
-   COMPUTE_LINEAR_REPEAT_TEXEL_LOCATION(texcoord[1], v, height, j0, j1);
+   linear_repeat_texel_location(width,  texcoord[0], &i0, &i1, &wi);
+   linear_repeat_texel_location(height, texcoord[1], &j0, &j1, &wj);
 
    img->FetchTexelc(img, i0, j0, 0, t00);
    img->FetchTexelc(img, i1, j0, 0, t10);
    img->FetchTexelc(img, i0, j1, 0, t01);
    img->FetchTexelc(img, i1, j1, 0, t11);
 
-   a = FRAC(u);
-   b = FRAC(v);
-   lerp_rgba_2d(rgba, a, b, t00, t10, t01, t11);
+   lerp_rgba_2d(rgba, wi, wj, t00, t10, t01, t11);
 }
 
 
-
 static void
 sample_2d_nearest_mipmap_nearest(GLcontext *ctx,
                                  const struct gl_texture_object *tObj,
@@ -1070,7 +1068,6 @@ sample_2d_nearest_mipmap_nearest(GLcontext *ctx,
 }
 
 
-
 static void
 sample_2d_linear_mipmap_nearest(GLcontext *ctx,
                                 const struct gl_texture_object *tObj,
@@ -1086,7 +1083,6 @@ sample_2d_linear_mipmap_nearest(GLcontext *ctx,
 }
 
 
-
 static void
 sample_2d_nearest_mipmap_linear(GLcontext *ctx,
                                 const struct gl_texture_object *tObj,
@@ -1112,8 +1108,6 @@ sample_2d_nearest_mipmap_linear(GLcontext *ctx,
 }
 
 
-
-/* Trilinear filtering */
 static void
 sample_2d_linear_mipmap_linear( GLcontext *ctx,
                                 const struct gl_texture_object *tObj,
@@ -1140,10 +1134,10 @@ sample_2d_linear_mipmap_linear( GLcontext *ctx,
 
 
 static void
-sample_2d_linear_mipmap_linear_repeat( GLcontext *ctx,
-                                       const struct gl_texture_object *tObj,
-                                       GLuint n, const GLfloat texcoord[][4],
-                                       const GLfloat lambda[], GLchan rgba[][4] )
+sample_2d_linear_mipmap_linear_repeat(GLcontext *ctx,
+                                      const struct gl_texture_object *tObj,
+                                      GLuint n, const GLfloat texcoord[][4],
+                                      const GLfloat lambda[], GLchan rgba[][4])
 {
    GLuint i;
    ASSERT(lambda != NULL);
@@ -1158,35 +1152,38 @@ sample_2d_linear_mipmap_linear_repeat( GLcontext *ctx,
       else {
          GLchan t0[4], t1[4];  /* texels */
          const GLfloat f = FRAC(lambda[i]);
-         sample_2d_linear_repeat(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_2d_linear_repeat(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         sample_2d_linear_repeat(ctx, tObj, tObj->Image[0][level  ],
+                                 texcoord[i], t0);
+         sample_2d_linear_repeat(ctx, tObj, tObj->Image[0][level+1],
+                                 texcoord[i], t1);
          lerp_rgba(rgba[i], f, t0, t1);
       }
    }
 }
 
 
+/** Sample 2D texture, nearest filtering for both min/magnification */
 static void
-sample_nearest_2d( GLcontext *ctx,
-                   const struct gl_texture_object *tObj, GLuint n,
-                   const GLfloat texcoords[][4],
-                   const GLfloat lambda[], GLchan rgba[][4] )
+sample_nearest_2d(GLcontext *ctx,
+                  const struct gl_texture_object *tObj, GLuint n,
+                  const GLfloat texcoords[][4],
+                  const GLfloat lambda[], GLchan rgba[][4])
 {
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
    (void) lambda;
-   for (i=0;i<n;i++) {
+   for (i = 0; i < n; i++) {
       sample_2d_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
    }
 }
 
 
-
+/** Sample 2D texture, linear filtering for both min/magnification */
 static void
-sample_linear_2d( GLcontext *ctx,
-                  const struct gl_texture_object *tObj, GLuint n,
-                  const GLfloat texcoords[][4],
-                  const GLfloat lambda[], GLchan rgba[][4] )
+sample_linear_2d(GLcontext *ctx,
+                 const struct gl_texture_object *tObj, GLuint n,
+                 const GLfloat texcoords[][4],
+                 const GLfloat lambda[], GLchan rgba[][4])
 {
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
@@ -1195,19 +1192,19 @@ sample_linear_2d( GLcontext *ctx,
        tObj->WrapT == GL_REPEAT &&
        image->_IsPowerOfTwo &&
        image->Border == 0) {
-      for (i=0;i<n;i++) {
+      for (i = 0; i < n; i++) {
          sample_2d_linear_repeat(ctx, tObj, image, texcoords[i], rgba[i]);
       }
    }
    else {
-      for (i=0;i<n;i++) {
+      for (i = 0; i < n; i++) {
          sample_2d_linear(ctx, tObj, image, texcoords[i], rgba[i]);
       }
    }
 }
 
 
-/*
+/**
  * Optimized 2-D texture sampling:
  *    S and T wrap mode == GL_REPEAT
  *    GL_NEAREST min/mag filter
@@ -1216,10 +1213,10 @@ sample_linear_2d( GLcontext *ctx,
  *    Format = GL_RGB
  */
 static void
-opt_sample_rgb_2d( GLcontext *ctx,
-                   const struct gl_texture_object *tObj,
-                   GLuint n, const GLfloat texcoords[][4],
-                   const GLfloat lambda[], GLchan rgba[][4] )
+opt_sample_rgb_2d(GLcontext *ctx,
+                  const struct gl_texture_object *tObj,
+                  GLuint n, const GLfloat texcoords[][4],
+                  const GLfloat lambda[], GLchan rgba[][4])
 {
    const struct gl_texture_image *img = tObj->Image[0][tObj->BaseLevel];
    const GLfloat width = (GLfloat) img->Width;
@@ -1248,7 +1245,7 @@ opt_sample_rgb_2d( GLcontext *ctx,
 }
 
 
-/*
+/**
  * Optimized 2-D texture sampling:
  *    S and T wrap mode == GL_REPEAT
  *    GL_NEAREST min/mag filter
@@ -1257,10 +1254,10 @@ opt_sample_rgb_2d( GLcontext *ctx,
  *    Format = GL_RGBA
  */
 static void
-opt_sample_rgba_2d( GLcontext *ctx,
-                    const struct gl_texture_object *tObj,
-                    GLuint n, const GLfloat texcoords[][4],
-                    const GLfloat lambda[], GLchan rgba[][4] )
+opt_sample_rgba_2d(GLcontext *ctx,
+                   const struct gl_texture_object *tObj,
+                   GLuint n, const GLfloat texcoords[][4],
+                   const GLfloat lambda[], GLchan rgba[][4])
 {
    const struct gl_texture_image *img = tObj->Image[0][tObj->BaseLevel];
    const GLfloat width = (GLfloat) img->Width;
@@ -1287,15 +1284,12 @@ opt_sample_rgba_2d( GLcontext *ctx,
 }
 
 
-/*
- * Given an array of texture coordinate and lambda (level of detail)
- * values, return an array of texture sample.
- */
+/** Sample 2D texture, using lambda to choose between min/magnification */
 static void
-sample_lambda_2d( GLcontext *ctx,
-                  const struct gl_texture_object *tObj,
-                  GLuint n, const GLfloat texcoords[][4],
-                  const GLfloat lambda[], GLchan rgba[][4] )
+sample_lambda_2d(GLcontext *ctx,
+                 const struct gl_texture_object *tObj,
+                 GLuint n, const GLfloat texcoords[][4],
+                 const GLfloat lambda[], GLchan rgba[][4])
 {
    const struct gl_texture_image *tImg = tObj->Image[0][tObj->BaseLevel];
    GLuint minStart, minEnd;  /* texels with minification */
@@ -1409,10 +1403,10 @@ sample_lambda_2d( GLcontext *ctx,
 /*                    3-D Texture Sampling Functions                  */
 /**********************************************************************/
 
-/*
+/**
  * Return the texture sample for coordinate (s,t,r) using GL_NEAREST filter.
  */
-static void
+static INLINE void
 sample_3d_nearest(GLcontext *ctx,
                   const struct gl_texture_object *tObj,
                   const struct gl_texture_image *img,
@@ -1425,9 +1419,9 @@ sample_3d_nearest(GLcontext *ctx,
    GLint i, j, k;
    (void) ctx;
 
-   COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapS, texcoord[0], width,  i);
-   COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapT, texcoord[1], height, j);
-   COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapR, texcoord[2], depth,  k);
+   i = nearest_texel_location(tObj->WrapS, img, width, texcoord[0]);
+   j = nearest_texel_location(tObj->WrapT, img, height, texcoord[1]);
+   k = nearest_texel_location(tObj->WrapR, img, depth, texcoord[2]);
 
    if (i < 0 || i >= (GLint) img->Width ||
        j < 0 || j >= (GLint) img->Height ||
@@ -1441,8 +1435,7 @@ sample_3d_nearest(GLcontext *ctx,
 }
 
 
-
-/*
+/**
  * Return the texture sample for coordinate (s,t,r) using GL_LINEAR filter.
  */
 static void
@@ -1457,14 +1450,13 @@ sample_3d_linear(GLcontext *ctx,
    const GLint depth = img->Depth2;
    GLint i0, j0, k0, i1, j1, k1;
    GLbitfield useBorderColor = 0x0;
-   GLfloat u, v, w;
    GLfloat a, b, c;
    GLchan t000[4], t010[4], t001[4], t011[4];
    GLchan t100[4], t110[4], t101[4], t111[4];
 
-   COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapS, texcoord[0], u, width,  i0, i1);
-   COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapT, texcoord[1], v, height, j0, j1);
-   COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapR, texcoord[2], w, depth,  k0, k1);
+   linear_texel_locations(tObj->WrapS, img, width, texcoord[0],  &i0, &i1, &a);
+   linear_texel_locations(tObj->WrapT, img, height, texcoord[1], &j0, &j1, &b);
+   linear_texel_locations(tObj->WrapR, img, depth, texcoord[2],  &k0, &k1, &c);
 
    if (img->Border) {
       i0 += img->Border;
@@ -1536,14 +1528,10 @@ sample_3d_linear(GLcontext *ctx,
    }
 
    /* trilinear interpolation of samples */
-   a = FRAC(u);
-   b = FRAC(v);
-   c = FRAC(w);
    lerp_rgba_3d(rgba, a, b, c, t000, t100, t010, t110, t001, t101, t011, t111);
 }
 
 
-
 static void
 sample_3d_nearest_mipmap_nearest(GLcontext *ctx,
                                  const struct gl_texture_object *tObj,
@@ -1623,6 +1611,7 @@ sample_3d_linear_mipmap_linear(GLcontext *ctx,
 }
 
 
+/** Sample 3D texture, nearest filtering for both min/magnification */
 static void
 sample_nearest_3d(GLcontext *ctx,
                   const struct gl_texture_object *tObj, GLuint n,
@@ -1632,37 +1621,34 @@ sample_nearest_3d(GLcontext *ctx,
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
    (void) lambda;
-   for (i=0;i<n;i++) {
+   for (i = 0; i < n; i++) {
       sample_3d_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
    }
 }
 
 
-
+/** Sample 3D texture, linear filtering for both min/magnification */
 static void
-sample_linear_3d( GLcontext *ctx,
-                  const struct gl_texture_object *tObj, GLuint n,
-                  const GLfloat texcoords[][4],
-		  const GLfloat lambda[], GLchan rgba[][4] )
+sample_linear_3d(GLcontext *ctx,
+                 const struct gl_texture_object *tObj, GLuint n,
+                 const GLfloat texcoords[][4],
+		 const GLfloat lambda[], GLchan rgba[][4])
 {
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
    (void) lambda;
-   for (i=0;i<n;i++) {
+   for (i = 0; i < n; i++) {
       sample_3d_linear(ctx, tObj, image, texcoords[i], rgba[i]);
    }
 }
 
 
-/*
- * Given an (s,t,r) texture coordinate and lambda (level of detail) value,
- * return a texture sample.
- */
+/** Sample 3D texture, using lambda to choose between min/magnification */
 static void
-sample_lambda_3d( GLcontext *ctx,
-                  const struct gl_texture_object *tObj, GLuint n,
-                  const GLfloat texcoords[][4], const GLfloat lambda[],
-                  GLchan rgba[][4] )
+sample_lambda_3d(GLcontext *ctx,
+                 const struct gl_texture_object *tObj, GLuint n,
+                 const GLfloat texcoords[][4], const GLfloat lambda[],
+                 GLchan rgba[][4])
 {
    GLuint minStart, minEnd;  /* texels with minification */
    GLuint magStart, magEnd;  /* texels with magnification */
@@ -1951,11 +1937,12 @@ sample_cube_linear_mipmap_linear(GLcontext *ctx,
 }
 
 
+/** Sample cube texture, using lambda to choose between min/magnification */
 static void
-sample_lambda_cube( GLcontext *ctx,
-		    const struct gl_texture_object *tObj, GLuint n,
-		    const GLfloat texcoords[][4], const GLfloat lambda[],
-		    GLchan rgba[][4])
+sample_lambda_cube(GLcontext *ctx,
+		   const struct gl_texture_object *tObj, GLuint n,
+		   const GLfloat texcoords[][4], const GLfloat lambda[],
+		   GLchan rgba[][4])
 {
    GLuint minStart, minEnd;  /* texels with minification */
    GLuint magStart, magEnd;  /* texels with magnification */
@@ -2031,50 +2018,57 @@ sample_lambda_cube( GLcontext *ctx,
 static INLINE GLint
 clamp_rect_coord_nearest(GLenum wrapMode, GLfloat coord, GLint max)
 {
-   if (wrapMode == GL_CLAMP) {
+   switch (wrapMode) {
+   case GL_CLAMP:
       return IFLOOR( CLAMP(coord, 0.0F, max - 1) );
-   }
-   else if (wrapMode == GL_CLAMP_TO_EDGE) {
+   case GL_CLAMP_TO_EDGE:
       return IFLOOR( CLAMP(coord, 0.5F, max - 0.5F) );
-   }
-   else {
+   case GL_CLAMP_TO_BORDER:
       return IFLOOR( CLAMP(coord, -0.5F, max + 0.5F) );
+   default:
+      _mesa_problem(NULL, "bad wrapMode in clamp_rect_coord_nearest");
+      return 0;
    }
 }
 
 
-/*
+/**
  * As above, but GL_LINEAR filtering.
  */
 static INLINE void
 clamp_rect_coord_linear(GLenum wrapMode, GLfloat coord, GLint max,
-                        GLint *i0out, GLint *i1out)
+                        GLint *i0out, GLint *i1out, GLfloat *weight)
 {
    GLfloat fcol;
    GLint i0, i1;
-   if (wrapMode == GL_CLAMP) {
+   switch (wrapMode) {
+   case GL_CLAMP:
       /* Not exactly what the spec says, but it matches NVIDIA output */
       fcol = CLAMP(coord - 0.5F, 0.0, max-1);
       i0 = IFLOOR(fcol);
       i1 = i0 + 1;
-   }
-   else if (wrapMode == GL_CLAMP_TO_EDGE) {
+      break;
+   case GL_CLAMP_TO_EDGE:
       fcol = CLAMP(coord, 0.5F, max - 0.5F);
       fcol -= 0.5F;
       i0 = IFLOOR(fcol);
       i1 = i0 + 1;
       if (i1 > max - 1)
          i1 = max - 1;
-   }
-   else {
-      ASSERT(wrapMode == GL_CLAMP_TO_BORDER);
+      break;
+   case GL_CLAMP_TO_BORDER:
       fcol = CLAMP(coord, -0.5F, max + 0.5F);
       fcol -= 0.5F;
       i0 = IFLOOR(fcol);
       i1 = i0 + 1;
+   default:
+      _mesa_problem(NULL, "bad wrapMode in clamp_rect_coord_linear");
+      i0 = i1 = 0;
+      fcol = 0.0F;
    }
    *i0out = i0;
    *i1out = i1;
+   *weight = FRAC(fcol);
 }
 
 
@@ -2085,10 +2079,8 @@ sample_nearest_rect(GLcontext *ctx,
                     GLchan rgba[][4])
 {
    const struct gl_texture_image *img = tObj->Image[0][0];
-   const GLfloat width = (GLfloat) img->Width;
-   const GLfloat height = (GLfloat) img->Height;
-   const GLint width_minus_1 = img->Width - 1;
-   const GLint height_minus_1 = img->Height - 1;
+   const GLint width = img->Width;
+   const GLint height = img->Height;
    GLuint i;
 
    (void) ctx;
@@ -2106,7 +2098,7 @@ sample_nearest_rect(GLcontext *ctx,
       GLint row, col;
       col = clamp_rect_coord_nearest(tObj->WrapS, texcoords[i][0], width);
       row = clamp_rect_coord_nearest(tObj->WrapT, texcoords[i][1], height);
-      if (col < 0 || col > width_minus_1 || row < 0 || row > height_minus_1)
+      if (col < 0 || col >= width || row < 0 || row >= height)
          COPY_CHAN4(rgba[i], tObj->_BorderChan);
       else
          img->FetchTexelc(img, col, row, 0, rgba[i]);
@@ -2121,10 +2113,8 @@ sample_linear_rect(GLcontext *ctx,
 		   const GLfloat lambda[], GLchan rgba[][4])
 {
    const struct gl_texture_image *img = tObj->Image[0][0];
-   const GLfloat width = (GLfloat) img->Width;
-   const GLfloat height = (GLfloat) img->Height;
-   const GLint width_minus_1 = img->Width - 1;
-   const GLint height_minus_1 = img->Height - 1;
+   const GLint width = img->Width;
+   const GLint height = img->Height;
    GLuint i;
 
    (void) ctx;
@@ -2138,64 +2128,22 @@ sample_linear_rect(GLcontext *ctx,
           tObj->WrapT == GL_CLAMP_TO_BORDER);
    ASSERT(img->TexFormat->BaseFormat != GL_COLOR_INDEX);
 
-   /* XXX lots of opportunity for optimization in this loop */
    for (i = 0; i < n; i++) {
-      GLfloat frow, fcol;
       GLint i0, j0, i1, j1;
       GLchan t00[4], t01[4], t10[4], t11[4];
       GLfloat a, b;
       GLbitfield useBorderColor = 0x0;
 
-      /* NOTE: we DO NOT use [0, 1] texture coordinates! */
-      if (tObj->WrapS == GL_CLAMP) {
-         /* Not exactly what the spec says, but it matches NVIDIA output */
-         fcol = CLAMP(texcoords[i][0] - 0.5F, 0.0, width_minus_1);
-         i0 = IFLOOR(fcol);
-         i1 = i0 + 1;
-      }
-      else if (tObj->WrapS == GL_CLAMP_TO_EDGE) {
-         fcol = CLAMP(texcoords[i][0], 0.5F, width - 0.5F);
-         fcol -= 0.5F;
-         i0 = IFLOOR(fcol);
-         i1 = i0 + 1;
-         if (i1 > width_minus_1)
-            i1 = width_minus_1;
-      }
-      else {
-         ASSERT(tObj->WrapS == GL_CLAMP_TO_BORDER);
-         fcol = CLAMP(texcoords[i][0], -0.5F, width + 0.5F);
-         fcol -= 0.5F;
-         i0 = IFLOOR(fcol);
-         i1 = i0 + 1;
-      }
-
-      if (tObj->WrapT == GL_CLAMP) {
-         /* Not exactly what the spec says, but it matches NVIDIA output */
-         frow = CLAMP(texcoords[i][1] - 0.5F, 0.0, width_minus_1);
-         j0 = IFLOOR(frow);
-         j1 = j0 + 1;
-      }
-      else if (tObj->WrapT == GL_CLAMP_TO_EDGE) {
-         frow = CLAMP(texcoords[i][1], 0.5F, height - 0.5F);
-         frow -= 0.5F;
-         j0 = IFLOOR(frow);
-         j1 = j0 + 1;
-         if (j1 > height_minus_1)
-            j1 = height_minus_1;
-      }
-      else {
-         ASSERT(tObj->WrapT == GL_CLAMP_TO_BORDER);
-         frow = CLAMP(texcoords[i][1], -0.5F, height + 0.5F);
-         frow -= 0.5F;
-         j0 = IFLOOR(frow);
-         j1 = j0 + 1;
-      }
+      clamp_rect_coord_linear(tObj->WrapS, texcoords[i][0], width,
+                              &i0, &i1, &a);
+      clamp_rect_coord_linear(tObj->WrapT, texcoords[i][1], height,
+                              &j0, &j1, &b);
 
       /* compute integer rows/columns */
-      if (i0 < 0 || i0 > width_minus_1)   useBorderColor |= I0BIT;
-      if (i1 < 0 || i1 > width_minus_1)   useBorderColor |= I1BIT;
-      if (j0 < 0 || j0 > height_minus_1)  useBorderColor |= J0BIT;
-      if (j1 < 0 || j1 > height_minus_1)  useBorderColor |= J1BIT;
+      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
+      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
+      if (j0 < 0 || j0 >= height)  useBorderColor |= J0BIT;
+      if (j1 < 0 || j1 >= height)  useBorderColor |= J1BIT;
 
       /* get four texel samples */
       if (useBorderColor & (I0BIT | J0BIT))
@@ -2218,20 +2166,17 @@ sample_linear_rect(GLcontext *ctx,
       else
          img->FetchTexelc(img, i1, j1, 0, t11);
 
-      /* compute interpolants */
-      a = FRAC(fcol);
-      b = FRAC(frow);
-
       lerp_rgba_2d(rgba[i], a, b, t00, t10, t01, t11);
    }
 }
 
 
+/** Sample Rect texture, using lambda to choose between min/magnification */
 static void
-sample_lambda_rect( GLcontext *ctx,
-		    const struct gl_texture_object *tObj, GLuint n,
-		    const GLfloat texcoords[][4], const GLfloat lambda[],
-		    GLchan rgba[][4])
+sample_lambda_rect(GLcontext *ctx,
+		   const struct gl_texture_object *tObj, GLuint n,
+		   const GLfloat texcoords[][4], const GLfloat lambda[],
+		   GLchan rgba[][4])
 {
    GLuint minStart, minEnd, magStart, magEnd;
 
@@ -2243,22 +2188,22 @@ sample_lambda_rect( GLcontext *ctx,
 
    if (minStart < minEnd) {
       if (tObj->MinFilter == GL_NEAREST) {
-         sample_nearest_rect( ctx, tObj, minEnd - minStart,
-                              texcoords + minStart, NULL, rgba + minStart);
+         sample_nearest_rect(ctx, tObj, minEnd - minStart,
+                             texcoords + minStart, NULL, rgba + minStart);
       }
       else {
-         sample_linear_rect( ctx, tObj, minEnd - minStart,
-                             texcoords + minStart, NULL, rgba + minStart);
+         sample_linear_rect(ctx, tObj, minEnd - minStart,
+                            texcoords + minStart, NULL, rgba + minStart);
       }
    }
    if (magStart < magEnd) {
       if (tObj->MagFilter == GL_NEAREST) {
-         sample_nearest_rect( ctx, tObj, magEnd - magStart,
-                              texcoords + magStart, NULL, rgba + magStart);
+         sample_nearest_rect(ctx, tObj, magEnd - magStart,
+                             texcoords + magStart, NULL, rgba + magStart);
       }
       else {
-         sample_linear_rect( ctx, tObj, magEnd - magStart,
-                             texcoords + magStart, NULL, rgba + magStart);
+         sample_linear_rect(ctx, tObj, magEnd - magStart,
+                            texcoords + magStart, NULL, rgba + magStart);
       }
    }
 }
@@ -2269,7 +2214,7 @@ sample_lambda_rect( GLcontext *ctx,
 /*                2D Texture Array Sampling Functions                 */
 /**********************************************************************/
 
-/*
+/**
  * Return the texture sample for coordinate (s,t,r) using GL_NEAREST filter.
  */
 static void
@@ -2286,8 +2231,8 @@ sample_2d_array_nearest(GLcontext *ctx,
    GLint array;
    (void) ctx;
 
-   COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapS, texcoord[0], width,  i);
-   COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapT, texcoord[1], height, j);
+   i = nearest_texel_location(tObj->WrapS, img, width, texcoord[0]);
+   j = nearest_texel_location(tObj->WrapT, img, height, texcoord[1]);
    array = clamp_rect_coord_nearest(tObj->WrapR, texcoord[2], depth);
 
    if (i < 0 || i >= (GLint) img->Width ||
@@ -2302,8 +2247,7 @@ sample_2d_array_nearest(GLcontext *ctx,
 }
 
 
-
-/*
+/**
  * Return the texture sample for coordinate (s,t,r) using GL_LINEAR filter.
  */
 static void
@@ -2319,12 +2263,11 @@ sample_2d_array_linear(GLcontext *ctx,
    GLint i0, j0, i1, j1;
    GLint array;
    GLbitfield useBorderColor = 0x0;
-   GLfloat u, v;
    GLfloat a, b;
    GLchan t00[4], t01[4], t10[4], t11[4];
 
-   COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapS, texcoord[0], u, width,  i0, i1);
-   COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapT, texcoord[1], v, height, j0, j1);
+   linear_texel_locations(tObj->WrapS, img, width,  texcoord[0], &i0, &i1, &a);
+   linear_texel_locations(tObj->WrapT, img, height, texcoord[1], &j0, &j1, &b);
    array = clamp_rect_coord_nearest(tObj->WrapR, texcoord[2], depth);
 
    if (array < 0 || array >= depth) {
@@ -2372,19 +2315,16 @@ sample_2d_array_linear(GLcontext *ctx,
       }
       
       /* trilinear interpolation of samples */
-      a = FRAC(u);
-      b = FRAC(v);
       lerp_rgba_2d(rgba, a, b, t00, t10, t01, t11);
    }
 }
 
 
-
 static void
 sample_2d_array_nearest_mipmap_nearest(GLcontext *ctx,
                                        const struct gl_texture_object *tObj,
                                        GLuint n, const GLfloat texcoord[][4],
-                                       const GLfloat lambda[], GLchan rgba[][4] )
+                                       const GLfloat lambda[], GLchan rgba[][4])
 {
    GLuint i;
    for (i = 0; i < n; i++) {
@@ -2428,8 +2368,10 @@ sample_2d_array_nearest_mipmap_linear(GLcontext *ctx,
       else {
          GLchan t0[4], t1[4];  /* texels */
          const GLfloat f = FRAC(lambda[i]);
-         sample_2d_array_nearest(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_2d_array_nearest(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         sample_2d_array_nearest(ctx, tObj, tObj->Image[0][level  ],
+                                 texcoord[i], t0);
+         sample_2d_array_nearest(ctx, tObj, tObj->Image[0][level+1],
+                                 texcoord[i], t1);
          lerp_rgba(rgba[i], f, t0, t1);
       }
    }
@@ -2438,9 +2380,9 @@ sample_2d_array_nearest_mipmap_linear(GLcontext *ctx,
 
 static void
 sample_2d_array_linear_mipmap_linear(GLcontext *ctx,
-                               const struct gl_texture_object *tObj,
-                               GLuint n, const GLfloat texcoord[][4],
-                               const GLfloat lambda[], GLchan rgba[][4])
+                                     const struct gl_texture_object *tObj,
+                                     GLuint n, const GLfloat texcoord[][4],
+                                     const GLfloat lambda[], GLchan rgba[][4])
 {
    GLuint i;
    ASSERT(lambda != NULL);
@@ -2453,30 +2395,34 @@ sample_2d_array_linear_mipmap_linear(GLcontext *ctx,
       else {
          GLchan t0[4], t1[4];  /* texels */
          const GLfloat f = FRAC(lambda[i]);
-         sample_2d_array_linear(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_2d_array_linear(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         sample_2d_array_linear(ctx, tObj, tObj->Image[0][level  ],
+                                texcoord[i], t0);
+         sample_2d_array_linear(ctx, tObj, tObj->Image[0][level+1],
+                                texcoord[i], t1);
          lerp_rgba(rgba[i], f, t0, t1);
       }
    }
 }
 
 
+/** Sample 2D Array texture, nearest filtering for both min/magnification */
 static void
 sample_nearest_2d_array(GLcontext *ctx,
-                  const struct gl_texture_object *tObj, GLuint n,
-                  const GLfloat texcoords[][4], const GLfloat lambda[],
-                  GLchan rgba[][4])
+                        const struct gl_texture_object *tObj, GLuint n,
+                        const GLfloat texcoords[][4], const GLfloat lambda[],
+                        GLchan rgba[][4])
 {
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
    (void) lambda;
-   for (i=0;i<n;i++) {
+   for (i = 0; i < n; i++) {
       sample_2d_array_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
    }
 }
 
 
 
+/** Sample 2D Array texture, linear filtering for both min/magnification */
 static void
 sample_linear_2d_array(GLcontext *ctx,
                        const struct gl_texture_object *tObj, GLuint n,
@@ -2486,16 +2432,13 @@ sample_linear_2d_array(GLcontext *ctx,
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
    (void) lambda;
-   for (i=0;i<n;i++) {
+   for (i = 0; i < n; i++) {
       sample_2d_array_linear(ctx, tObj, image, texcoords[i], rgba[i]);
    }
 }
 
 
-/*
- * Given an (s,t,r) texture coordinate and lambda (level of detail) value,
- * return a texture sample.
- */
+/** Sample 2D Array texture, using lambda to choose between min/magnification */
 static void
 sample_lambda_2d_array(GLcontext *ctx,
                        const struct gl_texture_object *tObj, GLuint n,
@@ -2525,8 +2468,10 @@ sample_lambda_2d_array(GLcontext *ctx,
                                    texcoords[i], rgba[i]);
          break;
       case GL_NEAREST_MIPMAP_NEAREST:
-         sample_2d_array_nearest_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
-                                                lambda + minStart, rgba + minStart);
+         sample_2d_array_nearest_mipmap_nearest(ctx, tObj, m,
+                                                texcoords + minStart,
+                                                lambda + minStart,
+                                                rgba + minStart);
          break;
       case GL_LINEAR_MIPMAP_NEAREST:
          sample_2d_array_linear_mipmap_nearest(ctx, tObj, m, 
@@ -2535,8 +2480,10 @@ sample_lambda_2d_array(GLcontext *ctx,
                                                rgba + minStart);
          break;
       case GL_NEAREST_MIPMAP_LINEAR:
-         sample_2d_array_nearest_mipmap_linear(ctx, tObj, m, texcoords + minStart,
-                                               lambda + minStart, rgba + minStart);
+         sample_2d_array_nearest_mipmap_linear(ctx, tObj, m,
+                                               texcoords + minStart,
+                                               lambda + minStart,
+                                               rgba + minStart);
          break;
       case GL_LINEAR_MIPMAP_LINEAR:
          sample_2d_array_linear_mipmap_linear(ctx, tObj, m, 
@@ -2577,7 +2524,7 @@ sample_lambda_2d_array(GLcontext *ctx,
 /*                1D Texture Array Sampling Functions                 */
 /**********************************************************************/
 
-/*
+/**
  * Return the texture sample for coordinate (s,t,r) using GL_NEAREST filter.
  */
 static void
@@ -2593,7 +2540,7 @@ sample_1d_array_nearest(GLcontext *ctx,
    GLint array;
    (void) ctx;
 
-   COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapS, texcoord[0], width,  i);
+   i = nearest_texel_location(tObj->WrapS, img, width, texcoord[0]);
    array = clamp_rect_coord_nearest(tObj->WrapT, texcoord[1], height);
 
    if (i < 0 || i >= (GLint) img->Width ||
@@ -2607,8 +2554,7 @@ sample_1d_array_nearest(GLcontext *ctx,
 }
 
 
-
-/*
+/**
  * Return the texture sample for coordinate (s,t,r) using GL_LINEAR filter.
  */
 static void
@@ -2623,11 +2569,10 @@ sample_1d_array_linear(GLcontext *ctx,
    GLint i0, i1;
    GLint array;
    GLbitfield useBorderColor = 0x0;
-   GLfloat u;
    GLfloat a;
    GLchan t0[4], t1[4];
 
-   COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapS, texcoord[0], u, width,  i0, i1);
+   linear_texel_locations(tObj->WrapS, img, width, texcoord[0], &i0, &i1, &a);
    array = clamp_rect_coord_nearest(tObj->WrapT, texcoord[1], height);
 
    if (img->Border) {
@@ -2657,17 +2602,15 @@ sample_1d_array_linear(GLcontext *ctx,
    }
 
    /* bilinear interpolation of samples */
-   a = FRAC(u);
    lerp_rgba(rgba, a, t0, t1);
 }
 
 
-
 static void
 sample_1d_array_nearest_mipmap_nearest(GLcontext *ctx,
                                        const struct gl_texture_object *tObj,
                                        GLuint n, const GLfloat texcoord[][4],
-                                       const GLfloat lambda[], GLchan rgba[][4] )
+                                       const GLfloat lambda[], GLchan rgba[][4])
 {
    GLuint i;
    for (i = 0; i < n; i++) {
@@ -2721,9 +2664,9 @@ sample_1d_array_nearest_mipmap_linear(GLcontext *ctx,
 
 static void
 sample_1d_array_linear_mipmap_linear(GLcontext *ctx,
-                               const struct gl_texture_object *tObj,
-                               GLuint n, const GLfloat texcoord[][4],
-                               const GLfloat lambda[], GLchan rgba[][4])
+                                     const struct gl_texture_object *tObj,
+                                     GLuint n, const GLfloat texcoord[][4],
+                                     const GLfloat lambda[], GLchan rgba[][4])
 {
    GLuint i;
    ASSERT(lambda != NULL);
@@ -2744,22 +2687,23 @@ sample_1d_array_linear_mipmap_linear(GLcontext *ctx,
 }
 
 
+/** Sample 1D Array texture, nearest filtering for both min/magnification */
 static void
 sample_nearest_1d_array(GLcontext *ctx,
-                  const struct gl_texture_object *tObj, GLuint n,
-                  const GLfloat texcoords[][4], const GLfloat lambda[],
-                  GLchan rgba[][4])
+                        const struct gl_texture_object *tObj, GLuint n,
+                        const GLfloat texcoords[][4], const GLfloat lambda[],
+                        GLchan rgba[][4])
 {
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
    (void) lambda;
-   for (i=0;i<n;i++) {
+   for (i = 0; i < n; i++) {
       sample_1d_array_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
    }
 }
 
 
-
+/** Sample 1D Array texture, linear filtering for both min/magnification */
 static void
 sample_linear_1d_array(GLcontext *ctx,
                        const struct gl_texture_object *tObj, GLuint n,
@@ -2769,16 +2713,13 @@ sample_linear_1d_array(GLcontext *ctx,
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
    (void) lambda;
-   for (i=0;i<n;i++) {
+   for (i = 0; i < n; i++) {
       sample_1d_array_linear(ctx, tObj, image, texcoords[i], rgba[i]);
    }
 }
 
 
-/*
- * Given an (s,t,r) texture coordinate and lambda (level of detail) value,
- * return a texture sample.
- */
+/** Sample 1D Array texture, using lambda to choose between min/magnification */
 static void
 sample_lambda_1d_array(GLcontext *ctx,
                        const struct gl_texture_object *tObj, GLuint n,
@@ -2854,9 +2795,7 @@ sample_lambda_1d_array(GLcontext *ctx,
 }
 
 
-
-
-/*
+/**
  * Sample a shadow/depth texture.
  */
 static void
@@ -2887,7 +2826,7 @@ sample_depth_texture( GLcontext *ctx,
           tObj->Target == GL_TEXTURE_1D_ARRAY_EXT ||
           tObj->Target == GL_TEXTURE_2D_ARRAY_EXT);
 
-   UNCLAMPED_FLOAT_TO_CHAN(ambient, tObj->ShadowAmbient);
+   UNCLAMPED_FLOAT_TO_CHAN(ambient, tObj->CompareFailValue);
 
    /* XXXX if tObj->MinFilter != tObj->MagFilter, we're ignoring lambda */
 
@@ -2906,33 +2845,36 @@ sample_depth_texture( GLcontext *ctx,
             break;
             
          case GL_TEXTURE_1D:
-            COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapS, texcoords[i][0],
-                                           width, col);
+            col = nearest_texel_location(tObj->WrapS, img, width,
+                                         texcoords[i][0]);
             row = 0;
             slice = 0;
             break;
 
          case GL_TEXTURE_2D:
-            COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapS, texcoords[i][0],
-                                           width, col);
-            COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapT, texcoords[i][1],
-                                           height, row);
+            col = nearest_texel_location(tObj->WrapS, img, width,
+                                         texcoords[i][0]);
+            row = nearest_texel_location(tObj->WrapT, img, height,
+                                         texcoords[i][1]);
             slice = 0;
             break;
 
          case GL_TEXTURE_1D_ARRAY_EXT:
-            COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapS, texcoords[i][0],
-                                           width, col);
+            col = nearest_texel_location(tObj->WrapS, img, width,
+                                         texcoords[i][0]);
             row = clamp_rect_coord_nearest(tObj->WrapT, texcoords[i][1], height);
             slice = 0;
+            break;
 
          case GL_TEXTURE_2D_ARRAY_EXT:
-            COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapS, texcoords[i][0],
-                                           width, col);
-            COMPUTE_NEAREST_TEXEL_LOCATION(tObj->WrapT, texcoords[i][1],
-                                           height, row);
+            col = nearest_texel_location(tObj->WrapS, img, width,
+                                         texcoords[i][0]);
+            row = nearest_texel_location(tObj->WrapT, img, height,
+                                         texcoords[i][1]);
             slice = clamp_rect_coord_nearest(tObj->WrapR, texcoords[i][2], depth);
             break;
+         default:
+            col = row = slice = 0;
          }
 
          if (col >= 0 && row >= 0 && col < width && row < height && 
@@ -3007,41 +2949,44 @@ sample_depth_texture( GLcontext *ctx,
          GLfloat depth00, depth01, depth10, depth11;
          GLint i0, i1, j0, j1;
          GLint slice;
-         GLfloat u, v;
+         GLfloat a, b;
          GLuint useBorderTexel;
 
          switch (tObj->Target) {
          case GL_TEXTURE_RECTANGLE_ARB:
             clamp_rect_coord_linear(tObj->WrapS, texcoords[i][0],
-                                    width, &i0, &i1);
+                                    width, &i0, &i1, &a);
             clamp_rect_coord_linear(tObj->WrapT, texcoords[i][1],
-                                    height, &j0, &j1);
+                                    height, &j0, &j1, &b);
             slice = 0;
             break;
 
          case GL_TEXTURE_1D:
          case GL_TEXTURE_2D:
-            COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapS, texcoords[i][0],
-                                           u, width, i0, i1);
-            COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapT, texcoords[i][1],
-                                           v, height,j0, j1);
+            linear_texel_locations(tObj->WrapS, img, width,
+                                   texcoords[i][0], &i0, &i1, &a);
+            linear_texel_locations(tObj->WrapT, img, height,
+                                   texcoords[i][1], &j0, &j1, &b);
             slice = 0;
             break;
 
          case GL_TEXTURE_1D_ARRAY_EXT:
-            COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapS, texcoords[i][0],
-                                           u, width, i0, i1);
+            linear_texel_locations(tObj->WrapS, img, width,
+                                   texcoords[i][0], &i0, &i1, &a);
             j0 = clamp_rect_coord_nearest(tObj->WrapT, texcoords[i][1], height);
             j1 = j0;
             slice = 0;
+            break;
 
          case GL_TEXTURE_2D_ARRAY_EXT:
-            COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapS, texcoords[i][0],
-                                           u, width, i0, i1);
-            COMPUTE_LINEAR_TEXEL_LOCATIONS(tObj->WrapT, texcoords[i][1],
-                                           v, height,j0, j1);
+            linear_texel_locations(tObj->WrapS, img, width,
+                                   texcoords[i][0], &i0, &i1, &a);
+            linear_texel_locations(tObj->WrapT, img, height,
+                                   texcoords[i][1], &j0, &j1, &b);
             slice = clamp_rect_coord_nearest(tObj->WrapR, texcoords[i][2], depth);
             break;
+         default:
+            slice = 0;
          }
 
          useBorderTexel = 0;
@@ -3103,8 +3048,6 @@ sample_depth_texture( GLcontext *ctx,
 
          if (0) {
             /* compute a single weighted depth sample and do one comparison */
-            const GLfloat a = FRAC(u + 1.0F);
-            const GLfloat b = FRAC(v + 1.0F);
             const GLfloat depthSample
                = lerp_2d(a, b, depth00, depth10, depth01, depth11);
             if ((depthSample <= texcoords[i][compare_coord] && function == GL_LEQUAL) ||
@@ -3175,8 +3118,6 @@ sample_depth_texture( GLcontext *ctx,
             case GL_NONE:
                /* ordinary bilinear filtering */
                {
-                  const GLfloat a = FRAC(u + 1.0F);
-                  const GLfloat b = FRAC(v + 1.0F);
                   const GLfloat depthSample
                      = lerp_2d(a, b, depth00, depth10, depth01, depth11);
                   CLAMPED_FLOAT_TO_CHAN(result, depthSample);
@@ -3215,103 +3156,6 @@ sample_depth_texture( GLcontext *ctx,
 }
 
 
-#if 0
-/*
- * Experimental depth texture sampling function.
- */
-static void
-sample_depth_texture2(const GLcontext *ctx,
-                     const struct gl_texture_unit *texUnit,
-                     GLuint n, const GLfloat texcoords[][4],
-                     GLchan texel[][4])
-{
-   const struct gl_texture_object *texObj = texUnit->_Current;
-   const GLint baseLevel = texObj->BaseLevel;
-   const struct gl_texture_image *texImage = texObj->Image[0][baseLevel];
-   const GLuint width = texImage->Width;
-   const GLuint height = texImage->Height;
-   GLchan ambient;
-   GLboolean lequal, gequal;
-
-   if (texObj->Target != GL_TEXTURE_2D) {
-      _mesa_problem(ctx, "only 2-D depth textures supported at this time");
-      return;
-   }
-
-   if (texObj->MinFilter != texObj->MagFilter) {
-      _mesa_problem(ctx, "mipmapped depth textures not supported at this time");
-      return;
-   }
-
-   /* XXX the GL_SGIX_shadow extension spec doesn't say what to do if
-    * GL_TEXTURE_COMPARE_SGIX == GL_TRUE but the current texture object
-    * isn't a depth texture.
-    */
-   if (texImage->TexFormat->BaseFormat != GL_DEPTH_COMPONENT) {
-      _mesa_problem(ctx,"GL_TEXTURE_COMPARE_SGIX enabled with non-depth texture");
-      return;
-   }
-
-   UNCLAMPED_FLOAT_TO_CHAN(ambient, tObj->ShadowAmbient);
-
-   if (texObj->CompareOperator == GL_TEXTURE_LEQUAL_R_SGIX) {
-      lequal = GL_TRUE;
-      gequal = GL_FALSE;
-   }
-   else {
-      lequal = GL_FALSE;
-      gequal = GL_TRUE;
-   }
-
-   {
-      GLuint i;
-      for (i = 0; i < n; i++) {
-         const GLint K = 3;
-         GLint col, row, ii, jj, imin, imax, jmin, jmax, samples, count;
-         GLfloat w;
-         GLchan lum;
-         COMPUTE_NEAREST_TEXEL_LOCATION(texObj->WrapS, texcoords[i][0],
-					width, col);
-         COMPUTE_NEAREST_TEXEL_LOCATION(texObj->WrapT, texcoords[i][1],
-					height, row);
-
-         imin = col - K;
-         imax = col + K;
-         jmin = row - K;
-         jmax = row + K;
-
-         if (imin < 0)  imin = 0;
-         if (imax >= width)  imax = width - 1;
-         if (jmin < 0)  jmin = 0;
-         if (jmax >= height) jmax = height - 1;
-
-         samples = (imax - imin + 1) * (jmax - jmin + 1);
-         count = 0;
-         for (jj = jmin; jj <= jmax; jj++) {
-            for (ii = imin; ii <= imax; ii++) {
-               GLfloat depthSample;
-               texImage->FetchTexelf(texImage, ii, jj, 0, &depthSample);
-               if ((depthSample <= r[i] && lequal) ||
-                   (depthSample >= r[i] && gequal)) {
-                  count++;
-               }
-            }
-         }
-
-         w = (GLfloat) count / (GLfloat) samples;
-         w = CHAN_MAXF - w * (CHAN_MAXF - (GLfloat) ambient);
-         lum = (GLint) w;
-
-         texel[i][RCOMP] = lum;
-         texel[i][GCOMP] = lum;
-         texel[i][BCOMP] = lum;
-         texel[i][ACOMP] = CHAN_MAX;
-      }
-   }
-}
-#endif
-
-
 /**
  * We use this function when a texture object is in an "incomplete" state.
  * When a fragment program attempts to sample an incomplete texture we
diff --git a/src/mesa/swrast/s_texstore.c b/src/mesa/swrast/s_texstore.c
index 16b00b9fa1..f9ff9ad6a4 100644
--- a/src/mesa/swrast/s_texstore.c
+++ b/src/mesa/swrast/s_texstore.c
@@ -67,7 +67,6 @@ static GLvoid *
 read_color_image( GLcontext *ctx, GLint x, GLint y, GLenum type,
                   GLsizei width, GLsizei height )
 {
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    struct gl_renderbuffer *rb = ctx->ReadBuffer->_ColorReadBuffer;
    const GLint pixelSize = _mesa_bytes_per_pixel(GL_RGBA, type);
    const GLint stride = width * pixelSize;
@@ -78,7 +77,7 @@ read_color_image( GLcontext *ctx, GLint x, GLint y, GLenum type,
    if (!image)
       return NULL;
 
-   RENDER_START(swrast, ctx);
+   swrast_render_start(ctx);
 
    dst = image;
    for (row = 0; row < height; row++) {
@@ -86,7 +85,7 @@ read_color_image( GLcontext *ctx, GLint x, GLint y, GLenum type,
       dst += stride;
    }
 
-   RENDER_FINISH(swrast, ctx);
+   swrast_render_finish(ctx);
 
    return image;
 }
@@ -101,7 +100,6 @@ read_depth_image( GLcontext *ctx, GLint x, GLint y,
                   GLsizei width, GLsizei height )
 {
    struct gl_renderbuffer *rb = ctx->ReadBuffer->_DepthBuffer;
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLuint *image, *dst;
    GLint i;
 
@@ -109,7 +107,7 @@ read_depth_image( GLcontext *ctx, GLint x, GLint y,
    if (!image)
       return NULL;
 
-   RENDER_START(swrast, ctx);
+   swrast_render_start(ctx);
 
    dst = image;
    for (i = 0; i < height; i++) {
@@ -117,7 +115,7 @@ read_depth_image( GLcontext *ctx, GLint x, GLint y,
       dst += width;
    }
 
-   RENDER_FINISH(swrast, ctx);
+   swrast_render_finish(ctx);
 
    return image;
 }
@@ -132,7 +130,6 @@ read_depth_stencil_image(GLcontext *ctx, GLint x, GLint y,
 {
    struct gl_renderbuffer *depthRb = ctx->ReadBuffer->_DepthBuffer;
    struct gl_renderbuffer *stencilRb = ctx->ReadBuffer->_StencilBuffer;
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLuint *image, *dst;
    GLint i;
 
@@ -143,7 +140,7 @@ read_depth_stencil_image(GLcontext *ctx, GLint x, GLint y,
    if (!image)
       return NULL;
 
-   RENDER_START(swrast, ctx);
+   swrast_render_start(ctx);
 
    /* read from depth buffer */
    dst = image;
@@ -205,7 +202,7 @@ read_depth_stencil_image(GLcontext *ctx, GLint x, GLint y,
       dst += width;
    }
 
-   RENDER_FINISH(swrast, ctx);
+   swrast_render_finish(ctx);
 
    return image;
 }
diff --git a/src/mesa/swrast/s_triangle.c b/src/mesa/swrast/s_triangle.c
index 2033ab5529..0598052f50 100644
--- a/src/mesa/swrast/s_triangle.c
+++ b/src/mesa/swrast/s_triangle.c
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.3
+ * Version:  7.3
  *
  * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
  *
@@ -35,6 +35,7 @@
 #include "main/imports.h"
 #include "main/macros.h"
 #include "main/texformat.h"
+#include "shader/prog_instruction.h"
 
 #include "s_aatriangle.h"
 #include "s_context.h"
@@ -263,6 +264,7 @@ affine_span(GLcontext *ctx, SWspan *span,
             struct affine_info *info)
 {
    GLchan sample[4];  /* the filtered texture sample */
+   const GLuint texEnableSave = ctx->Texture._EnabledUnits;
 
    /* Instead of defining a function for each mode, a test is done
     * between the outer and inner loops. This is to reduce code size
@@ -392,6 +394,9 @@ affine_span(GLcontext *ctx, SWspan *span,
    GLuint i;
    GLchan *dest = span->array->rgba[0];
 
+   /* Disable tex units so they're not re-applied in swrast_write_rgba_span */
+   ctx->Texture._EnabledUnits = 0x0;
+
    span->intTex[0] -= FIXED_HALF;
    span->intTex[1] -= FIXED_HALF;
    switch (info->filter) {
@@ -493,8 +498,12 @@ affine_span(GLcontext *ctx, SWspan *span,
    }
    span->interpMask &= ~SPAN_RGBA;
    ASSERT(span->arrayMask & SPAN_RGBA);
+
    _swrast_write_rgba_span(ctx, span);
 
+   /* re-enable texture units */
+   ctx->Texture._EnabledUnits = texEnableSave;
+
 #undef SPAN_NEAREST
 #undef SPAN_LINEAR
 }
@@ -1055,6 +1064,7 @@ _swrast_choose_triangle( GLcontext *ctx )
              && ctx->Texture.Unit[0]._ReallyEnabled == TEXTURE_2D_BIT
              && texObj2D->WrapS == GL_REPEAT
              && texObj2D->WrapT == GL_REPEAT
+             && texObj2D->_Swizzle == SWIZZLE_NOOP
              && texImg->_IsPowerOfTwo
              && texImg->Border == 0
              && texImg->Width == texImg->RowStride
@@ -1062,7 +1072,8 @@ _swrast_choose_triangle( GLcontext *ctx )
              && minFilter == magFilter
              && ctx->Light.Model.ColorControl == GL_SINGLE_COLOR
              && !swrast->_FogEnabled
-             && ctx->Texture.Unit[0].EnvMode != GL_COMBINE_EXT) {
+             && ctx->Texture.Unit[0].EnvMode != GL_COMBINE_EXT
+             && ctx->Texture.Unit[0].EnvMode != GL_COMBINE4_NV) {
 	    if (ctx->Hint.PerspectiveCorrection==GL_FASTEST) {
 	       if (minFilter == GL_NEAREST
 		   && format == MESA_FORMAT_RGB
diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c
index 8977fadcca..f0d31fdac3 100644
--- a/src/mesa/tnl/t_context.c
+++ b/src/mesa/tnl/t_context.c
@@ -32,6 +32,8 @@
 #include "main/macros.h"
 #include "main/mtypes.h"
 #include "main/light.h"
+#include "math/m_translate.h"
+#include "math/m_xform.h"
 
 #include "tnl.h"
 #include "t_context.h"
@@ -78,6 +80,12 @@ _tnl_CreateContext( GLcontext *ctx )
 
    tnl->nr_blocks = 0;
 
+   /* plug in the VBO drawing function */
+   vbo_set_draw_func(ctx, _tnl_draw_prims);
+
+   _math_init_transformation();
+   _math_init_translate();
+
    return GL_TRUE;
 }
 
@@ -109,24 +117,28 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state )
 
    tnl->pipeline.new_state |= new_state;
 
-   /* Calculate tnl->render_inputs:
+   /* Calculate tnl->render_inputs.  This bitmask indicates which vertex
+    * attributes need to be emitted to the rasterizer.
     */
    if (ctx->Visual.rgbMode) {
       GLuint i;
 
       RENDERINPUTS_ZERO( tnl->render_inputs_bitset );
       RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_POS );
+
       if (!fp || (fp->Base.InputsRead & FRAG_BIT_COL0)) {
          RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_COLOR0 );
       }
+
+      if (NEED_SECONDARY_COLOR(ctx))
+         RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_COLOR1 );
+
       for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-         if (ctx->Texture._EnabledCoordUnits & (1 << i)) {
+         if (ctx->Texture._EnabledCoordUnits & (1 << i) ||
+             (fp && fp->Base.InputsRead & FRAG_BIT_TEX(i))) {
             RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_TEX(i) );
          }
       }
-
-      if (NEED_SECONDARY_COLOR(ctx))
-         RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_COLOR1 );
    }
    else {
       RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_POS );
@@ -137,7 +149,7 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state )
       /* fixed-function fog */
       RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_FOG );
    }
-   else if (ctx->FragmentProgram._Active || ctx->FragmentProgram._Current) {
+   else if (ctx->FragmentProgram._Current) {
       struct gl_fragment_program *fp = ctx->FragmentProgram._Current;
       if (fp) {
          if (fp->FogOption != GL_NONE || (fp->Base.InputsRead & FRAG_BIT_FOGC)) {
diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c
index fd647c1f4a..2ec65d5323 100644
--- a/src/mesa/tnl/t_draw.c
+++ b/src/mesa/tnl/t_draw.c
@@ -88,6 +88,29 @@ static void free_space(GLcontext *ctx)
 } while (0)
 
 
+/**
+ * Convert array of BGRA/GLubyte[4] values to RGBA/float[4]
+ * \param ptr  input/ubyte array
+ * \param fptr  output/float array
+ */
+static void
+convert_bgra_to_float(const struct gl_client_array *input,
+                      const GLubyte *ptr, GLfloat *fptr,
+                      GLuint count )
+{
+   GLuint i;
+   assert(input->Normalized);
+   assert(input->Size == 4);
+   for (i = 0; i < count; i++) {
+      const GLubyte *in = (GLubyte *) ptr;  /* in is in BGRA order */
+      *fptr++ = UBYTE_TO_FLOAT(in[2]);  /* red */
+      *fptr++ = UBYTE_TO_FLOAT(in[1]);  /* green */
+      *fptr++ = UBYTE_TO_FLOAT(in[0]);  /* blue */
+      *fptr++ = UBYTE_TO_FLOAT(in[3]);  /* alpha */
+      ptr += input->StrideB;
+   }
+}
+
 
 /* Adjust pointer to point at first requested element, convert to
  * floating point, populate VB->AttribPtr[].
@@ -112,7 +135,13 @@ static void _tnl_import_array( GLcontext *ctx,
 	 CONVERT(GLbyte, BYTE_TO_FLOAT); 
 	 break;
       case GL_UNSIGNED_BYTE: 
-	 CONVERT(GLubyte, UBYTE_TO_FLOAT); 
+         if (input->Format == GL_BGRA) {
+            /* See GL_EXT_vertex_array_bgra */
+            convert_bgra_to_float(input, ptr, fptr, count);
+         }
+         else {
+            CONVERT(GLubyte, UBYTE_TO_FLOAT); 
+         }
 	 break;
       case GL_SHORT: 
 	 CONVERT(GLshort, SHORT_TO_FLOAT); 
diff --git a/src/mesa/tnl/t_rasterpos.c b/src/mesa/tnl/t_rasterpos.c
index dfbe0f1806..14d79eab54 100644
--- a/src/mesa/tnl/t_rasterpos.c
+++ b/src/mesa/tnl/t_rasterpos.c
@@ -26,11 +26,11 @@
 #include "main/glheader.h"
 #include "main/colormac.h"
 #include "main/context.h"
-#include "feedback.h"
-#include "light.h"
+#include "main/feedback.h"
+#include "main/light.h"
 #include "main/macros.h"
-#include "rastpos.h"
-#include "simple_list.h"
+#include "main/rastpos.h"
+#include "main/simple_list.h"
 #include "main/mtypes.h"
 
 #include "math/m_matrix.h"
diff --git a/src/mesa/tnl/t_vertex.c b/src/mesa/tnl/t_vertex.c
index b661524c87..10b78f820e 100644
--- a/src/mesa/tnl/t_vertex.c
+++ b/src/mesa/tnl/t_vertex.c
@@ -376,6 +376,22 @@ void _tnl_notify_pipeline_output_change( GLcontext *ctx )
    invalidate_funcs(vtx);
 }
 
+
+static void adjust_input_ptrs( GLcontext *ctx, GLint diff)
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   struct tnl_clipspace_attr *a = vtx->attr;
+   const GLuint count = vtx->attr_count;
+   int j;
+
+   diff -= 1;
+   for (j=0; j<count; ++j) {
+           register GLvector4f *vptr = VB->AttribPtr[a->attrib];
+	   (a++)->inputptr += diff*vptr->stride;
+   }
+}
+
 static void update_input_ptrs( GLcontext *ctx, GLuint start )
 {
    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
@@ -431,13 +447,40 @@ void *_tnl_emit_vertices_to_buffer( GLcontext *ctx,
    struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
 
    update_input_ptrs(ctx, start);
-
    /* Note: dest should not be adjusted for non-zero 'start' values:
     */
    vtx->emit( ctx, end - start, (GLubyte*) dest );	
    return (void *)((GLubyte *)dest + vtx->vertex_size * (end - start));
 }
 
+/* Emit indexed VB vertices start..end to dest.  Note that VB vertex at
+ * postion start will be emitted to dest at position zero.
+ */
+
+void *_tnl_emit_indexed_vertices_to_buffer( GLcontext *ctx,
+					    const GLuint *elts,
+					    GLuint start,
+					    GLuint end,
+					    void *dest )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   GLuint oldIndex;
+   GLubyte *cdest = dest;
+
+   update_input_ptrs(ctx, oldIndex = elts[start++]);
+   vtx->emit( ctx, 1, cdest );
+   cdest += vtx->vertex_size;
+
+   for (; start < end; ++start) {
+      adjust_input_ptrs(ctx, elts[start] - oldIndex);
+      oldIndex = elts[start];
+      vtx->emit( ctx, 1, cdest);
+      cdest += vtx->vertex_size;
+   }
+
+   return (void *) cdest;
+}
+
 
 void _tnl_init_vertices( GLcontext *ctx, 
 			GLuint vb_size,
diff --git a/src/mesa/tnl/t_vertex.h b/src/mesa/tnl/t_vertex.h
index 712311a146..2dfd7b57f0 100644
--- a/src/mesa/tnl/t_vertex.h
+++ b/src/mesa/tnl/t_vertex.h
@@ -119,6 +119,18 @@ extern void *_tnl_emit_vertices_to_buffer( GLcontext *ctx,
 					   GLuint end,
 					   void *dest );
 
+/* This function isn't optimal. Check out 
+ * gallium/auxilary/translate for a more comprehensive implementation of
+ * the same functionality.
+ */
+  
+extern void *_tnl_emit_indexed_vertices_to_buffer( GLcontext *ctx,
+						   const GLuint *elts,
+						   GLuint start,
+						   GLuint end,
+						   void *dest );
+
+
 extern void _tnl_build_vertices( GLcontext *ctx,
 				 GLuint start,
 				 GLuint end,
diff --git a/src/mesa/tnl/t_vertex_generic.c b/src/mesa/tnl/t_vertex_generic.c
index f763522f91..9812f8c808 100644
--- a/src/mesa/tnl/t_vertex_generic.c
+++ b/src/mesa/tnl/t_vertex_generic.c
@@ -113,7 +113,7 @@ static INLINE void insert_3f_viewport_2( const struct tnl_clipspace_attr *a, GLu
    DEBUG_INSERT;
    out[0] = vp[0] * in[0] + vp[12];
    out[1] = vp[5] * in[1] + vp[13];
-   out[2] = vp[10] * in[2] + vp[14];
+   out[2] = vp[14];
 }
 
 static INLINE void insert_3f_viewport_1( const struct tnl_clipspace_attr *a, GLubyte *v,
diff --git a/src/mesa/tnl/t_vertex_sse.c b/src/mesa/tnl/t_vertex_sse.c
index 76043bd1b5..7a255d680a 100644
--- a/src/mesa/tnl/t_vertex_sse.c
+++ b/src/mesa/tnl/t_vertex_sse.c
@@ -146,7 +146,8 @@ static void emit_load3f_1( struct x86_program *p,
 			   struct x86_reg dest,
 			   struct x86_reg arg0 )
 {
-   emit_load4f_1(p, dest, arg0);
+   /* Loading from memory erases the upper bits. */
+   sse_movss(&p->func, dest, arg0);
 }
 
 static void emit_load2f_2( struct x86_program *p, 
@@ -160,7 +161,8 @@ static void emit_load2f_1( struct x86_program *p,
 			   struct x86_reg dest,
 			   struct x86_reg arg0 )
 {
-   emit_load4f_1(p, dest, arg0);
+   /* Loading from memory erases the upper bits. */
+   sse_movss(&p->func, dest, arg0);
 }
 
 static void emit_load1f_1( struct x86_program *p, 
@@ -352,6 +354,7 @@ static GLboolean build_vertex_emit( struct x86_program *p )
    struct x86_reg temp = x86_make_reg(file_XMM, 0);
    struct x86_reg vp0 = x86_make_reg(file_XMM, 1);
    struct x86_reg vp1 = x86_make_reg(file_XMM, 2);
+   struct x86_reg temp2 = x86_make_reg(file_XMM, 3);
    GLubyte *fixup, *label;
 
    /* Push a few regs?
@@ -524,7 +527,8 @@ static GLboolean build_vertex_emit( struct x86_program *p )
 	    sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z));
 
 	    get_src_ptr(p, srcECX, vtxESI, &a[1]);
-	    emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize);
+	    emit_load(p, temp2, 1, x86_deref(srcECX), a[1].inputsize);
+	    sse_movss(&p->func, temp, temp2);
 	    update_src_ptr(p, srcECX, vtxESI, &a[1]);
 
 	    /* Rearrange and possibly do BGR conversion:
@@ -539,8 +543,8 @@ static GLboolean build_vertex_emit( struct x86_program *p )
 	 }
 	 else {
 	    _mesa_printf("Can't emit 3ub\n");
+	    return GL_FALSE;	/* add this later */
 	 }
-	 return GL_FALSE;	/* add this later */
 	 break;
 
       case EMIT_4UB_4F_RGBA:
diff --git a/src/mesa/tnl/t_vp_build.h b/src/mesa/tnl/t_vp_build.h
index d1c147ff03..d6ebc66c04 100644
--- a/src/mesa/tnl/t_vp_build.h
+++ b/src/mesa/tnl/t_vp_build.h
@@ -29,6 +29,13 @@
 
 #include "main/mtypes.h"
 
+#define TNL_FIXED_FUNCTION_STATE_FLAGS (_NEW_PROGRAM |		\
+					_NEW_LIGHT |		\
+					_NEW_TEXTURE |		\
+					_NEW_TEXTURE_MATRIX |		\
+					_NEW_TRANSFORM |	\
+					_NEW_FOG |		\
+					_NEW_POINT)
 
 extern void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx );
 
diff --git a/src/mesa/tnl/tnl.h b/src/mesa/tnl/tnl.h
index 9a1717bf89..4d628aa9a6 100644
--- a/src/mesa/tnl/tnl.h
+++ b/src/mesa/tnl/tnl.h
@@ -81,6 +81,8 @@ _tnl_draw_prims( GLcontext *ctx,
 		 GLuint min_index,
 		 GLuint max_index);
 
+extern void
+_mesa_load_tracked_matrices(GLcontext *ctx);
 
 extern void
 _tnl_RasterPos(GLcontext *ctx, const GLfloat vObj[4]);
diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h
index 4c51b44cda..5362226c2f 100644
--- a/src/mesa/vbo/vbo.h
+++ b/src/mesa/vbo/vbo.h
@@ -117,4 +117,22 @@ void vbo_rebase_prims( GLcontext *ctx,
 void vbo_use_buffer_objects(GLcontext *ctx);
 
 
+void vbo_set_draw_func(GLcontext *ctx, vbo_draw_func func);
+
+
+void GLAPIENTRY
+_vbo_Color4f(GLfloat r, GLfloat g, GLfloat b, GLfloat a);
+
+void GLAPIENTRY
+_vbo_Normal3f(GLfloat x, GLfloat y, GLfloat z);
+
+void GLAPIENTRY
+_vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q);
+
+void GLAPIENTRY
+_vbo_Materialfv(GLenum face, GLenum pname, const GLfloat *params);
+
+void GLAPIENTRY
+_vbo_VertexAttrib4f(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+
 #endif
diff --git a/src/mesa/vbo/vbo_context.c b/src/mesa/vbo/vbo_context.c
index a73a46b9c8..ca8190fd05 100644
--- a/src/mesa/vbo/vbo_context.c
+++ b/src/mesa/vbo/vbo_context.c
@@ -28,9 +28,11 @@
 #include "main/imports.h"
 #include "main/mtypes.h"
 #include "main/api_arrayelt.h"
+#include "math/m_eval.h"
 #include "vbo.h"
 #include "vbo_context.h"
 
+#if 0
 /* Reach out and grab this to use as the default:
  */
 extern void _tnl_draw_prims( GLcontext *ctx,
@@ -40,6 +42,7 @@ extern void _tnl_draw_prims( GLcontext *ctx,
 			     const struct _mesa_index_buffer *ib,
 			     GLuint min_index,
 			     GLuint max_index );
+#endif
 
 
 
@@ -76,6 +79,7 @@ static void init_legacy_currval(GLcontext *ctx)
       cl->StrideB = 0;
       cl->Enabled = 1;
       cl->Type = GL_FLOAT;
+      cl->Format = GL_RGBA;
       cl->Ptr = (const void *)ctx->Current.Attrib[i];
       cl->BufferObj = ctx->Array.NullBufferObj;
    }
@@ -97,6 +101,7 @@ static void init_generic_currval(GLcontext *ctx)
        */
       cl->Size = 1;
       cl->Type = GL_FLOAT;
+      cl->Format = GL_RGBA;
       cl->Ptr = (const void *)ctx->Current.Attrib[VERT_ATTRIB_GENERIC0 + i];
       cl->Stride = 0;
       cl->StrideB = 0;
@@ -141,6 +146,7 @@ static void init_mat_currval(GLcontext *ctx)
 
       cl->Ptr = (const void *)ctx->Light.Material.Attrib[i];
       cl->Type = GL_FLOAT;
+      cl->Format = GL_RGBA;
       cl->Stride = 0;
       cl->StrideB = 0;
       cl->Enabled = 1;
@@ -212,7 +218,9 @@ GLboolean _vbo_CreateContext( GLcontext *ctx )
 
    /* By default: 
     */
+#if 0 /* dead - see vbo_set_draw_func() */
    vbo->draw_prims = _tnl_draw_prims;
+#endif
 
    /* Hook our functions into exec and compile dispatch tables.  These
     * will pretty much be permanently installed, which means that the
@@ -223,6 +231,8 @@ GLboolean _vbo_CreateContext( GLcontext *ctx )
    vbo_save_init( ctx );
 #endif
 
+   _math_init_eval();
+
    return GL_TRUE;
 }
 
@@ -240,10 +250,20 @@ void _vbo_DestroyContext( GLcontext *ctx )
       ctx->aelt_context = NULL;
    }
 
-   vbo_exec_destroy(ctx);
+   if (vbo_context(ctx)) {
+      vbo_exec_destroy(ctx);
 #if FEATURE_dlist
-   vbo_save_destroy(ctx);
+      vbo_save_destroy(ctx);
 #endif
-   FREE(vbo_context(ctx));
-   ctx->swtnl_im = NULL;
+      FREE(vbo_context(ctx));
+      ctx->swtnl_im = NULL;
+   }
 }
+
+
+void vbo_set_draw_func(GLcontext *ctx, vbo_draw_func func)
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   vbo->draw_prims = func;
+}
+
diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c
index d48f5230cb..2743bf6b55 100644
--- a/src/mesa/vbo/vbo_exec_api.c
+++ b/src/mesa/vbo/vbo_exec_api.c
@@ -143,29 +143,36 @@ static void vbo_exec_copy_to_current( struct vbo_exec_context *exec )
 
    for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       if (exec->vtx.attrsz[i]) {
-	 GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
-
          /* Note: the exec->vtx.current[i] pointers point into the
           * ctx->Current.Attrib and ctx->Light.Material.Attrib arrays.
           */
-	 COPY_CLEAN_4V(current, 
-		       exec->vtx.attrsz[i], 
-		       exec->vtx.attrptr[i]);
-
+	 GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
+         GLfloat tmp[4];
+
+         COPY_CLEAN_4V(tmp, 
+                       exec->vtx.attrsz[i], 
+                       exec->vtx.attrptr[i]);
+         
+         if (memcmp(current, tmp, sizeof(tmp)) != 0)
+         { 
+            memcpy(current, tmp, sizeof(tmp));
 	 
-	 /* Given that we explicitly state size here, there is no need
-	  * for the COPY_CLEAN above, could just copy 16 bytes and be
-	  * done.  The only problem is when Mesa accesses ctx->Current
-	  * directly.
-	  */
-	 vbo->currval[i].Size = exec->vtx.attrsz[i];
-
-	 /* This triggers rather too much recalculation of Mesa state
-	  * that doesn't get used (eg light positions).
-	  */
-	 if (i >= VBO_ATTRIB_MAT_FRONT_AMBIENT &&
-	     i <= VBO_ATTRIB_MAT_BACK_INDEXES)
-	    ctx->NewState |= _NEW_LIGHT;
+            /* Given that we explicitly state size here, there is no need
+             * for the COPY_CLEAN above, could just copy 16 bytes and be
+             * done.  The only problem is when Mesa accesses ctx->Current
+             * directly.
+             */
+            vbo->currval[i].Size = exec->vtx.attrsz[i];
+
+            /* This triggers rather too much recalculation of Mesa state
+             * that doesn't get used (eg light positions).
+             */
+            if (i >= VBO_ATTRIB_MAT_FRONT_AMBIENT &&
+                i <= VBO_ATTRIB_MAT_BACK_INDEXES)
+               ctx->NewState |= _NEW_LIGHT;
+            
+            ctx->NewState |= _NEW_CURRENT_ATTRIB;
+         }
       }
    }
 
@@ -693,7 +700,8 @@ void vbo_exec_vtx_init( struct vbo_exec_context *exec )
    GLuint i;
 
    /* Allocate a buffer object.  Will just reuse this object
-    * continuously.
+    * continuously, unless vbo_use_buffer_objects() is called to enable
+    * use of real VBOs.
     */
    _mesa_reference_buffer_object(ctx,
                                  &exec->vtx.bufferobj,
@@ -772,3 +780,36 @@ static void reset_attrfv( struct vbo_exec_context *exec )
    exec->vtx.vertex_size = 0;
 }
       
+
+void GLAPIENTRY
+_vbo_Color4f(GLfloat r, GLfloat g, GLfloat b, GLfloat a)
+{
+   vbo_Color4f(r, g, b, a);
+}
+
+
+void GLAPIENTRY
+_vbo_Normal3f(GLfloat x, GLfloat y, GLfloat z)
+{
+   vbo_Normal3f(x, y, z);
+}
+
+
+void GLAPIENTRY
+_vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q)
+{
+   vbo_MultiTexCoord4f(target, s, t, r, q);
+}
+
+void GLAPIENTRY
+_vbo_Materialfv(GLenum face, GLenum pname, const GLfloat *params)
+{
+   vbo_Materialfv(face, pname, params);
+}
+
+
+void GLAPIENTRY
+_vbo_VertexAttrib4f(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w)
+{
+   vbo_VertexAttrib4fARB(index, x, y, z, w);
+}
diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c
index 82f4db17d1..8871e10cf6 100644
--- a/src/mesa/vbo/vbo_exec_array.c
+++ b/src/mesa/vbo/vbo_exec_array.c
@@ -30,6 +30,8 @@
 #include "main/state.h"
 #include "main/api_validate.h"
 #include "main/api_noop.h"
+#include "main/varray.h"
+#include "glapi/dispatch.h"
 
 #include "vbo_context.h"
 
@@ -125,6 +127,7 @@ static void recalculate_input_bindings( GLcontext *ctx )
    struct vbo_context *vbo = vbo_context(ctx);
    struct vbo_exec_context *exec = &vbo->exec;
    const struct gl_client_array **inputs = &exec->array.inputs[0];
+   GLbitfield const_inputs = 0x0;
    GLuint i;
 
    exec->array.program_mode = get_program_mode(ctx);
@@ -139,19 +142,24 @@ static void recalculate_input_bindings( GLcontext *ctx )
       for (i = 0; i <= VERT_ATTRIB_TEX7; i++) {
 	 if (exec->array.legacy_array[i]->Enabled)
 	    inputs[i] = exec->array.legacy_array[i];
-	 else
+	 else {
 	    inputs[i] = &vbo->legacy_currval[i];
+            const_inputs |= 1 << i;
+         }
       }
 
       for (i = 0; i < MAT_ATTRIB_MAX; i++) {
 	 inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->mat_currval[i];
+         const_inputs |= 1 << (VERT_ATTRIB_GENERIC0 + i);
       }
 
       /* Could use just about anything, just to fill in the empty
        * slots:
        */
-      for (i = MAT_ATTRIB_MAX; i < VERT_ATTRIB_MAX - VERT_ATTRIB_GENERIC0; i++)
+      for (i = MAT_ATTRIB_MAX; i < VERT_ATTRIB_MAX - VERT_ATTRIB_GENERIC0; i++) {
 	 inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->generic_currval[i];
+         const_inputs |= 1 << (VERT_ATTRIB_GENERIC0 + i);
+      }
 
       break;
    case VP_NV:
@@ -164,15 +172,19 @@ static void recalculate_input_bindings( GLcontext *ctx )
 	    inputs[i] = exec->array.generic_array[i];
 	 else if (exec->array.legacy_array[i]->Enabled)
 	    inputs[i] = exec->array.legacy_array[i];
-	 else
+	 else {
 	    inputs[i] = &vbo->legacy_currval[i];
+            const_inputs |= 1 << i;
+         }
       }
 
       /* Could use just about anything, just to fill in the empty
        * slots:
        */
-      for (i = VERT_ATTRIB_GENERIC0; i < VERT_ATTRIB_MAX; i++)
+      for (i = VERT_ATTRIB_GENERIC0; i < VERT_ATTRIB_MAX; i++) {
 	 inputs[i] = &vbo->generic_currval[i - VERT_ATTRIB_GENERIC0];
+         const_inputs |= 1 << i;
+      }
 
       break;
    case VP_ARB:
@@ -187,25 +199,34 @@ static void recalculate_input_bindings( GLcontext *ctx )
 	 inputs[0] = exec->array.generic_array[0];
       else if (exec->array.legacy_array[0]->Enabled)
 	 inputs[0] = exec->array.legacy_array[0];
-      else
+      else {
 	 inputs[0] = &vbo->legacy_currval[0];
+         const_inputs |= 1 << 0;
+      }
 
 
       for (i = 1; i <= VERT_ATTRIB_TEX7; i++) {
 	 if (exec->array.legacy_array[i]->Enabled)
 	    inputs[i] = exec->array.legacy_array[i];
-	 else
+	 else {
 	    inputs[i] = &vbo->legacy_currval[i];
+            const_inputs |= 1 << i;
+         }
       }
 
       for (i = 0; i < 16; i++) {
 	 if (exec->array.generic_array[i]->Enabled)
 	    inputs[VERT_ATTRIB_GENERIC0 + i] = exec->array.generic_array[i];
-	 else
+	 else {
 	    inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->generic_currval[i];
+            const_inputs |= 1 << (VERT_ATTRIB_GENERIC0 + i);
+         }
+
       }
       break;
    }
+
+   _mesa_set_varying_vp_inputs( ctx, ~const_inputs );
 }
 
 static void bind_arrays( GLcontext *ctx )
@@ -255,6 +276,11 @@ vbo_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
 
    bind_arrays( ctx );
 
+   /* Again...
+    */
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+
    prim[0].begin = 1;
    prim[0].end = 1;
    prim[0].weak = 0;
@@ -295,6 +321,9 @@ vbo_exec_DrawRangeElements(GLenum mode,
 
    bind_arrays( ctx );
 
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+
    ib.count = count;
    ib.type = type; 
    ib.obj = ctx->Array.ElementArrayBufferObj;
@@ -403,3 +432,29 @@ void vbo_exec_array_destroy( struct vbo_exec_context *exec )
 {
    /* nothing to do */
 }
+
+
+/* This API entrypoint is not ordinarily used */
+void GLAPIENTRY
+_mesa_DrawArrays(GLenum mode, GLint first, GLsizei count)
+{
+   vbo_exec_DrawArrays(mode, first, count);
+}
+
+
+/* This API entrypoint is not ordinarily used */
+void GLAPIENTRY
+_mesa_DrawElements(GLenum mode, GLsizei count, GLenum type,
+                   const GLvoid *indices)
+{
+   vbo_exec_DrawElements(mode, count, type, indices);
+}
+
+
+/* This API entrypoint is not ordinarily used */
+void GLAPIENTRY
+_mesa_DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count,
+                        GLenum type, const GLvoid *indices)
+{
+   vbo_exec_DrawRangeElements(mode, start, end, count, type, indices);
+}
diff --git a/src/mesa/vbo/vbo_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c
index 92356ba977..45133fa4f3 100644
--- a/src/mesa/vbo/vbo_exec_draw.c
+++ b/src/mesa/vbo/vbo_exec_draw.c
@@ -150,6 +150,7 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
    GLubyte *data = exec->vtx.buffer_map;
    const GLuint *map;
    GLuint attr;
+   GLbitfield varying_inputs = 0x0;
 
    /* Install the default (ie Current) attributes first, then overlay
     * all active ones.
@@ -175,7 +176,20 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
          exec->vtx.inputs[attr + 16] = &vbo->generic_currval[attr];
       }
       map = vbo->map_vp_arb;
+
+      /* check if VERT_ATTRIB_POS is not read but VERT_BIT_GENERIC0 is read.
+       * In that case we effectively need to route the data from
+       * glVertexAttrib(0, val) calls to feed into the GENERIC0 input.
+       */
+      if ((ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_POS) == 0 &&
+          (ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_GENERIC0)) {
+         exec->vtx.inputs[16] = exec->vtx.inputs[0];
+         exec->vtx.attrsz[16] = exec->vtx.attrsz[0];
+         exec->vtx.attrsz[0] = 0;
+      }
       break;
+   default:
+      assert(0);
    }
 
    /* Make all active attributes (including edgeflag) available as
@@ -204,6 +218,7 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
 	 arrays[attr].StrideB = exec->vtx.vertex_size * sizeof(GLfloat);
 	 arrays[attr].Stride = exec->vtx.vertex_size * sizeof(GLfloat);
 	 arrays[attr].Type = GL_FLOAT;
+         arrays[attr].Format = GL_RGBA;
 	 arrays[attr].Enabled = 1;
          _mesa_reference_buffer_object(ctx,
                                        &arrays[attr].BufferObj,
@@ -211,8 +226,11 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
 	 arrays[attr]._MaxElement = count; /* ??? */
 
 	 data += exec->vtx.attrsz[src] * sizeof(GLfloat);
+         varying_inputs |= 1<<attr;
       }
    }
+
+   _mesa_set_varying_vp_inputs( ctx, varying_inputs );
 }
 
 
@@ -242,6 +260,9 @@ void vbo_exec_vtx_flush( struct vbo_exec_context *exec )
 	  */
 	 vbo_exec_bind_arrays( ctx );
 
+         if (ctx->NewState)
+            _mesa_update_state( ctx );
+
          /* if using a real VBO, unmap it before drawing */
          if (exec->vtx.bufferobj->Name) {
             ctx->Driver.UnmapBuffer(ctx, target, exec->vtx.bufferobj);
diff --git a/src/mesa/vbo/vbo_save.h b/src/mesa/vbo/vbo_save.h
index b7e9baabf8..7dda54af48 100644
--- a/src/mesa/vbo/vbo_save.h
+++ b/src/mesa/vbo/vbo_save.h
@@ -166,7 +166,7 @@ void vbo_loopback_vertex_list( GLcontext *ctx,
 void vbo_save_EndList( GLcontext *ctx );
 void vbo_save_NewList( GLcontext *ctx, GLuint list, GLenum mode );
 void vbo_save_EndCallList( GLcontext *ctx );
-void vbo_save_BeginCallList( GLcontext *ctx, struct mesa_display_list *list );
+void vbo_save_BeginCallList( GLcontext *ctx, struct gl_display_list *list );
 void vbo_save_SaveFlushVertices( GLcontext *ctx );
 GLboolean vbo_save_NotifyBegin( GLcontext *ctx, GLenum mode );
 
diff --git a/src/mesa/vbo/vbo_save_api.c b/src/mesa/vbo/vbo_save_api.c
index f69a33d817..ddfd276577 100644
--- a/src/mesa/vbo/vbo_save_api.c
+++ b/src/mesa/vbo/vbo_save_api.c
@@ -293,7 +293,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
 	  node->count == 0);
 
    if (save->dangling_attr_ref)
-      ctx->ListState.CurrentList->flags |= MESA_DLIST_DANGLING_REFS;
+      ctx->ListState.CurrentList->Flags |= DLIST_DANGLING_REFS;
 
    save->vertex_store->used += save->vertex_size * node->count;
    save->prim_store->used += node->prim_count;
@@ -1076,10 +1076,10 @@ void vbo_save_EndList( GLcontext *ctx )
    assert(save->vertex_size == 0);
 }
  
-void vbo_save_BeginCallList( GLcontext *ctx, struct mesa_display_list *dlist )
+void vbo_save_BeginCallList( GLcontext *ctx, struct gl_display_list *dlist )
 {
    struct vbo_save_context *save = &vbo_context(ctx)->save;
-   save->replay_flags |= dlist->flags;
+   save->replay_flags |= dlist->Flags;
 }
 
 void vbo_save_EndCallList( GLcontext *ctx )
diff --git a/src/mesa/vbo/vbo_save_draw.c b/src/mesa/vbo/vbo_save_draw.c
index ed82f09958..617c5a77ba 100644
--- a/src/mesa/vbo/vbo_save_draw.c
+++ b/src/mesa/vbo/vbo_save_draw.c
@@ -64,18 +64,26 @@ static void _playback_copy_to_current( GLcontext *ctx,
    for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       if (node->attrsz[i]) {
 	 GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
+         GLfloat tmp[4];
 
-	 COPY_CLEAN_4V(current, 
-		       node->attrsz[i], 
-		       data);
+         COPY_CLEAN_4V(tmp, 
+                       node->attrsz[i], 
+                       data);
+         
+         if (memcmp(current, tmp, 4 * sizeof(GLfloat)) != 0)
+         {
+            memcpy(current, tmp, 4 * sizeof(GLfloat));
 
-	 vbo->currval[i].Size = node->attrsz[i];
+            vbo->currval[i].Size = node->attrsz[i];
 
-	 data += node->attrsz[i];
+            if (i >= VBO_ATTRIB_FIRST_MATERIAL &&
+                i <= VBO_ATTRIB_LAST_MATERIAL)
+               ctx->NewState |= _NEW_LIGHT;
+
+            ctx->NewState |= _NEW_CURRENT_ATTRIB;
+         }
 
-	 if (i >= VBO_ATTRIB_FIRST_MATERIAL &&
-	     i <= VBO_ATTRIB_LAST_MATERIAL)
-	    ctx->NewState |= _NEW_LIGHT;
+	 data += node->attrsz[i];
       }
    }
 
@@ -110,6 +118,10 @@ static void vbo_bind_vertex_list( GLcontext *ctx,
    GLuint data = node->buffer_offset;
    const GLuint *map;
    GLuint attr;
+   GLubyte node_attrsz[VBO_ATTRIB_MAX];  /* copy of node->attrsz[] */
+   GLbitfield varying_inputs = 0x0;
+
+   memcpy(node_attrsz, node->attrsz, sizeof(node->attrsz));
 
    /* Install the default (ie Current) attributes first, then overlay
     * all active ones.
@@ -135,13 +147,26 @@ static void vbo_bind_vertex_list( GLcontext *ctx,
          save->inputs[attr + 16] = &vbo->generic_currval[attr];
       }
       map = vbo->map_vp_arb;
+
+      /* check if VERT_ATTRIB_POS is not read but VERT_BIT_GENERIC0 is read.
+       * In that case we effectively need to route the data from
+       * glVertexAttrib(0, val) calls to feed into the GENERIC0 input.
+       */
+      if ((ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_POS) == 0 &&
+          (ctx->VertexProgram._Current->Base.InputsRead & VERT_BIT_GENERIC0)) {
+         save->inputs[16] = save->inputs[0];
+         node_attrsz[16] = node_attrsz[0];
+         node_attrsz[0] = 0;
+      }
       break;
+   default:
+      assert(0);
    }
 
    for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) {
       GLuint src = map[attr];
 
-      if (node->attrsz[src]) {
+      if (node_attrsz[src]) {
          /* override the default array set above */
          save->inputs[attr] = &arrays[attr];
 
@@ -150,6 +175,7 @@ static void vbo_bind_vertex_list( GLcontext *ctx,
 	 arrays[attr].StrideB = node->vertex_size * sizeof(GLfloat);
 	 arrays[attr].Stride = node->vertex_size * sizeof(GLfloat);
 	 arrays[attr].Type = GL_FLOAT;
+         arrays[attr].Format = GL_RGBA;
 	 arrays[attr].Enabled = 1;
          _mesa_reference_buffer_object(ctx,
                                        &arrays[attr].BufferObj,
@@ -159,8 +185,11 @@ static void vbo_bind_vertex_list( GLcontext *ctx,
 	 assert(arrays[attr].BufferObj->Name);
 
 	 data += node->attrsz[src] * sizeof(GLfloat);
+         varying_inputs |= 1<<attr;
       }
    }
+
+   _mesa_set_varying_vp_inputs( ctx, varying_inputs );
 }
 
 static void vbo_save_loopback_vertex_list( GLcontext *ctx,
@@ -229,6 +258,11 @@ void vbo_save_playback_vertex_list( GLcontext *ctx, void *data )
 
       vbo_bind_vertex_list( ctx, node );
 
+      /* Again...
+       */
+      if (ctx->NewState)
+	 _mesa_update_state( ctx );
+
       vbo_context(ctx)->draw_prims( ctx, 
 				    save->inputs, 
 				    node->prim, 
diff --git a/src/mesa/vbo/vbo_split_copy.c b/src/mesa/vbo/vbo_split_copy.c
index 685cc0fdf6..5fb66d3318 100644
--- a/src/mesa/vbo/vbo_split_copy.c
+++ b/src/mesa/vbo/vbo_split_copy.c
@@ -461,6 +461,7 @@ static void replay_init( struct copy_context *copy )
 
       dst->Size = src->Size;
       dst->Type = src->Type;
+      dst->Format = GL_RGBA;
       dst->Stride = copy->vertex_size;
       dst->StrideB = copy->vertex_size;
       dst->Ptr = copy->dstbuf + offset;
diff --git a/src/mesa/vf/vf.c b/src/mesa/vf/vf.c
new file mode 100644
index 0000000000..82f3d2b641
--- /dev/null
+++ b/src/mesa/vf/vf.c
@@ -0,0 +1,372 @@
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keithw@tungstengraphics.com>
+ */
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+
+#include "vf.h"
+
+#define DBG 0
+
+
+
+static GLboolean match_fastpath( struct vertex_fetch *vf,
+				 const struct vf_fastpath *fp)
+{
+   GLuint j;
+
+   if (vf->attr_count != fp->attr_count) 
+      return GL_FALSE;
+
+   for (j = 0; j < vf->attr_count; j++) 
+      if (vf->attr[j].format != fp->attr[j].format ||
+	  vf->attr[j].inputsize != fp->attr[j].size ||
+	  vf->attr[j].vertoffset != fp->attr[j].offset) 
+	 return GL_FALSE;
+      
+   if (fp->match_strides) {
+      if (vf->vertex_stride != fp->vertex_stride)
+	 return GL_FALSE;
+
+      for (j = 0; j < vf->attr_count; j++) 
+	 if (vf->attr[j].inputstride != fp->attr[j].stride) 
+	    return GL_FALSE;
+   }
+   
+   return GL_TRUE;
+}
+
+static GLboolean search_fastpath_emit( struct vertex_fetch *vf )
+{
+   struct vf_fastpath *fp = vf->fastpath;
+
+   for ( ; fp ; fp = fp->next) {
+      if (match_fastpath(vf, fp)) {
+         vf->emit = fp->func;
+	 return GL_TRUE;
+      }
+   }
+
+   return GL_FALSE;
+}
+
+void vf_register_fastpath( struct vertex_fetch *vf,
+			     GLboolean match_strides )
+{
+   struct vf_fastpath *fastpath = CALLOC_STRUCT(vf_fastpath);
+   GLuint i;
+
+   fastpath->vertex_stride = vf->vertex_stride;
+   fastpath->attr_count = vf->attr_count;
+   fastpath->match_strides = match_strides;
+   fastpath->func = vf->emit;
+   fastpath->attr = (struct vf_attr_type *)
+      _mesa_malloc(vf->attr_count * sizeof(fastpath->attr[0]));
+
+   for (i = 0; i < vf->attr_count; i++) {
+      fastpath->attr[i].format = vf->attr[i].format;
+      fastpath->attr[i].stride = vf->attr[i].inputstride;
+      fastpath->attr[i].size = vf->attr[i].inputsize;
+      fastpath->attr[i].offset = vf->attr[i].vertoffset;
+   }
+
+   fastpath->next = vf->fastpath;
+   vf->fastpath = fastpath;
+}
+
+
+
+
+/***********************************************************************
+ * Build codegen functions or return generic ones:
+ */
+static void choose_emit_func( struct vertex_fetch *vf, 
+			      GLuint count, 
+			      GLubyte *dest)
+{
+   vf->emit = NULL;
+   
+   /* Does this match an existing (hardwired, codegen or known-bad)
+    * fastpath?
+    */
+   if (search_fastpath_emit(vf)) {
+      /* Use this result.  If it is null, then it is already known
+       * that the current state will fail for codegen and there is no
+       * point trying again.
+       */
+   }
+   else if (vf->codegen_emit) {
+      vf->codegen_emit( vf );
+   }
+
+   if (!vf->emit) {
+      vf_generate_hardwired_emit(vf);
+   }
+
+   /* Otherwise use the generic version:
+    */
+   if (!vf->emit)
+      vf->emit = vf_generic_emit;
+
+   vf->emit( vf, count, dest );
+}
+
+
+
+
+
+/***********************************************************************
+ * Public entrypoints, mostly dispatch to the above:
+ */
+
+
+
+GLuint vf_set_vertex_attributes( struct vertex_fetch *vf, 
+				 const struct vf_attr_map *map,
+				 GLuint nr, 
+				 GLuint vertex_stride )
+{
+   GLuint offset = 0;
+   GLuint i, j;
+
+   assert(nr < VF_ATTRIB_MAX);
+
+   memset(vf->lookup, 0, sizeof(vf->lookup));
+
+   for (j = 0, i = 0; i < nr; i++) {
+      const GLuint format = map[i].format;
+      if (format == EMIT_PAD) {
+	 if (DBG)
+	    _mesa_printf("%d: pad %d, offset %d\n", i,  
+			 map[i].offset, offset);  
+
+	 offset += map[i].offset;
+
+      }
+      else {
+	 assert(vf->lookup[map[i].attrib] == 0);
+	 vf->lookup[map[i].attrib] = &vf->attr[j];
+
+	 vf->attr[j].attrib = map[i].attrib;
+	 vf->attr[j].format = format;
+	 vf->attr[j].insert = vf_format_info[format].insert;
+	 vf->attr[j].extract = vf_format_info[format].extract;
+	 vf->attr[j].vertattrsize = vf_format_info[format].attrsize;
+	 vf->attr[j].vertoffset = offset;
+	 
+	 if (DBG)
+	    _mesa_printf("%d: %s, offset %d\n", i,  
+			 vf_format_info[format].name,
+			 vf->attr[j].vertoffset);   
+
+	 offset += vf_format_info[format].attrsize;
+	 j++;
+      }
+   }
+
+   vf->attr_count = j;
+   vf->vertex_stride = vertex_stride ? vertex_stride : offset;
+   vf->emit = choose_emit_func;
+
+   assert(vf->vertex_stride >= offset);
+   return vf->vertex_stride;
+}
+
+
+
+void vf_set_vp_matrix( struct vertex_fetch *vf,
+		       const GLfloat *viewport )
+{
+   assert(vf->allow_viewport_emits);
+
+   /* scale */
+   vf->vp[0] = viewport[MAT_SX];
+   vf->vp[1] = viewport[MAT_SY];
+   vf->vp[2] = viewport[MAT_SZ];
+   vf->vp[3] = 1.0;
+
+   /* translate */
+   vf->vp[4] = viewport[MAT_TX];
+   vf->vp[5] = viewport[MAT_TY];
+   vf->vp[6] = viewport[MAT_TZ];
+   vf->vp[7] = 0.0;
+}
+
+void vf_set_vp_scale_translate( struct vertex_fetch *vf,
+				const GLfloat *scale,
+				const GLfloat *translate )
+{
+   assert(vf->allow_viewport_emits);
+
+   vf->vp[0] = scale[0];
+   vf->vp[1] = scale[1];
+   vf->vp[2] = scale[2];
+   vf->vp[3] = scale[3];
+
+   vf->vp[4] = translate[0];
+   vf->vp[5] = translate[1];
+   vf->vp[6] = translate[2];
+   vf->vp[7] = translate[3];
+}
+
+
+/* Set attribute pointers, adjusted for start position:
+ */
+void vf_set_sources( struct vertex_fetch *vf,
+		     GLvector4f * const sources[],
+		     GLuint start )
+{
+   struct vf_attr *a = vf->attr;
+   GLuint j;
+   
+   for (j = 0; j < vf->attr_count; j++) {
+      const GLvector4f *vptr = sources[a[j].attrib];
+      
+      if ((a[j].inputstride != vptr->stride) ||
+	  (a[j].inputsize != vptr->size))
+	 vf->emit = choose_emit_func;
+      
+      a[j].inputstride = vptr->stride;
+      a[j].inputsize = vptr->size;
+      a[j].do_insert = a[j].insert[vptr->size - 1]; 
+      a[j].inputptr = ((GLubyte *)vptr->data) + start * vptr->stride;
+   }
+}
+
+
+
+/* Emit count VB vertices to dest.  
+ */
+void vf_emit_vertices( struct vertex_fetch *vf,
+		       GLuint count,
+		       void *dest )
+{
+   vf->emit( vf, count, (GLubyte*) dest );	
+}
+
+
+/* Extract a named attribute from a hardware vertex.  Will have to
+ * reverse any viewport transformation, swizzling or other conversions
+ * which may have been applied.
+ *
+ * This is mainly required for on-the-fly vertex translations to
+ * swrast format.
+ */
+void vf_get_attr( struct vertex_fetch *vf,
+		  const void *vertex,
+		  GLenum attr, 
+		  const GLfloat *dflt,
+		  GLfloat *dest )
+{
+   const struct vf_attr *a = vf->attr;
+   const GLuint attr_count = vf->attr_count;
+   GLuint j;
+
+   for (j = 0; j < attr_count; j++) {
+      if (a[j].attrib == attr) {
+	 a[j].extract( &a[j], dest, (GLubyte *)vertex + a[j].vertoffset );
+	 return;
+      }
+   }
+
+   /* Else return the value from ctx->Current.
+    */
+   _mesa_memcpy( dest, dflt, 4*sizeof(GLfloat));
+}
+
+
+
+
+struct vertex_fetch *vf_create( GLboolean allow_viewport_emits )
+{
+   struct vertex_fetch *vf = CALLOC_STRUCT(vertex_fetch);
+   GLuint i;
+
+   for (i = 0; i < VF_ATTRIB_MAX; i++)
+      vf->attr[i].vf = vf;
+
+   vf->allow_viewport_emits = allow_viewport_emits;
+
+   switch(CHAN_TYPE) {
+   case GL_UNSIGNED_BYTE:
+      vf->chan_scale[0] = 255.0;
+      vf->chan_scale[1] = 255.0;
+      vf->chan_scale[2] = 255.0;
+      vf->chan_scale[3] = 255.0;
+      break;
+   case GL_UNSIGNED_SHORT:
+      vf->chan_scale[0] = 65535.0;
+      vf->chan_scale[1] = 65535.0;
+      vf->chan_scale[2] = 65535.0;
+      vf->chan_scale[3] = 65535.0;
+      break;
+   default:
+      vf->chan_scale[0] = 1.0;
+      vf->chan_scale[1] = 1.0;
+      vf->chan_scale[2] = 1.0;
+      vf->chan_scale[3] = 1.0;
+      break;
+   }
+
+   vf->identity[0] = 0.0;
+   vf->identity[1] = 0.0;
+   vf->identity[2] = 0.0;
+   vf->identity[3] = 1.0;
+
+   vf->codegen_emit = NULL;
+
+#ifdef USE_SSE_ASM
+   if (!_mesa_getenv("MESA_NO_CODEGEN"))
+      vf->codegen_emit = vf_generate_sse_emit;
+#endif
+
+   return vf;
+}
+
+
+void vf_destroy( struct vertex_fetch *vf )
+{
+   struct vf_fastpath *fp, *tmp;
+
+   for (fp = vf->fastpath ; fp ; fp = tmp) {
+      tmp = fp->next;
+      FREE(fp->attr);
+
+      /* KW: At the moment, fp->func is constrained to be allocated by
+       * _mesa_exec_alloc(), as the hardwired fastpaths in
+       * t_vertex_generic.c are handled specially.  It would be nice
+       * to unify them, but this probably won't change until this
+       * module gets another overhaul.
+       */
+      _mesa_exec_free((void *) fp->func);
+      FREE(fp);
+   }
+   
+   vf->fastpath = NULL;
+   FREE(vf);
+}
diff --git a/src/mesa/vf/vf.h b/src/mesa/vf/vf.h
new file mode 100644
index 0000000000..83d7547619
--- /dev/null
+++ b/src/mesa/vf/vf.h
@@ -0,0 +1,234 @@
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keithw@tungstengraphics.com>
+ */
+
+#ifndef VF_VERTEX_H
+#define VF_VERTEX_H
+
+#include "main/mtypes.h"
+#include "math/m_vector.h"
+
+enum {
+   VF_ATTRIB_POS = 0,
+   VF_ATTRIB_WEIGHT = 1,
+   VF_ATTRIB_NORMAL = 2,
+   VF_ATTRIB_COLOR0 = 3,
+   VF_ATTRIB_COLOR1 = 4,
+   VF_ATTRIB_FOG = 5,
+   VF_ATTRIB_COLOR_INDEX = 6,
+   VF_ATTRIB_EDGEFLAG = 7,
+   VF_ATTRIB_TEX0 = 8,
+   VF_ATTRIB_TEX1 = 9,
+   VF_ATTRIB_TEX2 = 10,
+   VF_ATTRIB_TEX3 = 11,
+   VF_ATTRIB_TEX4 = 12,
+   VF_ATTRIB_TEX5 = 13,
+   VF_ATTRIB_TEX6 = 14,
+   VF_ATTRIB_TEX7 = 15,
+   VF_ATTRIB_VAR0 = 16,
+   VF_ATTRIB_VAR1 = 17,
+   VF_ATTRIB_VAR2 = 18,
+   VF_ATTRIB_VAR3 = 19,
+   VF_ATTRIB_VAR4 = 20,
+   VF_ATTRIB_VAR5 = 21,
+   VF_ATTRIB_VAR6 = 22,
+   VF_ATTRIB_VAR7 = 23,
+   VF_ATTRIB_POINTSIZE = 24,
+   VF_ATTRIB_BFC0 = 25,
+   VF_ATTRIB_BFC1 = 26,
+   VF_ATTRIB_CLIP_POS = 27,
+   VF_ATTRIB_VERTEX_HEADER = 28,
+   VF_ATTRIB_MAX = 29
+};
+
+
+enum vf_attr_format {
+   EMIT_1F,
+   EMIT_2F,
+   EMIT_3F,
+   EMIT_4F,
+   EMIT_2F_VIEWPORT,		/* do viewport transform and emit */
+   EMIT_3F_VIEWPORT,		/* do viewport transform and emit */
+   EMIT_4F_VIEWPORT,		/* do viewport transform and emit */
+   EMIT_3F_XYW,			/* for projective texture */
+   EMIT_1UB_1F,			/* for fog coordinate */
+   EMIT_3UB_3F_RGB,		/* for specular color */
+   EMIT_3UB_3F_BGR,		/* for specular color */
+   EMIT_4UB_4F_RGBA,		/* for color */
+   EMIT_4UB_4F_BGRA,		/* for color */
+   EMIT_4UB_4F_ARGB,		/* for color */
+   EMIT_4UB_4F_ABGR,		/* for color */
+   EMIT_4CHAN_4F_RGBA,		/* for swrast color */
+   EMIT_PAD,			/* leave a hole of 'offset' bytes */
+   EMIT_MAX
+};
+
+struct vf_attr_map {
+   GLuint attrib;
+   enum vf_attr_format format;
+   GLuint offset;
+};
+
+struct vertex_fetch;
+
+void vf_set_vp_matrix( struct vertex_fetch *vf,
+		      const GLfloat *viewport );
+
+void vf_set_vp_scale_translate( struct vertex_fetch *vf,
+				const GLfloat *scale,
+				const GLfloat *translate );
+
+GLuint vf_set_vertex_attributes( struct vertex_fetch *vf,
+				 const struct vf_attr_map *map,
+				 GLuint nr, 
+				 GLuint vertex_stride );
+
+void vf_set_sources( struct vertex_fetch *vf,
+		     GLvector4f * const attrib[],
+		     GLuint start ); 
+
+void vf_emit_vertices( struct vertex_fetch *vf,
+		       GLuint count,
+		       void *dest );
+
+void vf_get_attr( struct vertex_fetch *vf,
+		  const void *vertex,
+		  GLenum attr, 
+		  const GLfloat *dflt,
+		  GLfloat *dest );
+
+struct vertex_fetch *vf_create( GLboolean allow_viewport_emits );
+
+void vf_destroy( struct vertex_fetch *vf );
+
+
+
+/***********************************************************************
+ * Internal functions and structs:
+ */
+
+struct vf_attr;
+
+typedef void (*vf_extract_func)( const struct vf_attr *a, 
+				 GLfloat *out, 
+				 const GLubyte *v );
+
+typedef void (*vf_insert_func)( const struct vf_attr *a, 
+				GLubyte *v, 
+				const GLfloat *in );
+
+typedef void (*vf_emit_func)( struct vertex_fetch *vf,
+			      GLuint count, 
+			      GLubyte *dest );
+
+
+
+/* Describes how to convert/move a vertex attribute from a vertex
+ * array to a vertex structure.
+ */
+struct vf_attr
+{
+   struct vertex_fetch *vf;
+
+   GLuint format;
+   GLuint inputsize;
+   GLuint inputstride;
+   GLuint vertoffset;      /* position of the attrib in the vertex struct */
+
+   GLuint attrib;          /* which vertex attrib (0=position, etc) */
+   GLuint vertattrsize;    /* size of the attribute in bytes */
+
+   GLubyte *inputptr;
+   const vf_insert_func *insert;
+   vf_insert_func do_insert;
+   vf_extract_func extract;
+};
+
+struct vertex_fetch
+{
+   struct vf_attr attr[VF_ATTRIB_MAX];
+   GLuint attr_count;
+   GLuint vertex_stride;
+
+   struct vf_attr *lookup[VF_ATTRIB_MAX];
+   
+   vf_emit_func emit;
+
+   /* Parameters and constants for codegen:
+    */
+   GLboolean allow_viewport_emits;
+   GLfloat vp[8];		
+   GLfloat chan_scale[4];
+   GLfloat identity[4];
+
+   struct vf_fastpath *fastpath;
+   
+   void (*codegen_emit)( struct vertex_fetch *vf );
+};
+
+
+struct vf_attr_type {
+   GLuint format;
+   GLuint size;
+   GLuint stride;
+   GLuint offset;
+};
+
+struct vf_fastpath {
+   GLuint vertex_stride;
+   GLuint attr_count;
+   GLboolean match_strides;
+
+   struct vf_attr_type *attr;
+
+   vf_emit_func func;
+   struct vf_fastpath *next;
+};
+
+
+void vf_register_fastpath( struct vertex_fetch *vtx,
+			     GLboolean match_strides );
+
+void vf_generic_emit( struct vertex_fetch *vf,
+			GLuint count,
+			GLubyte *v );
+
+void vf_generate_hardwired_emit( struct vertex_fetch *vf );
+
+void vf_generate_sse_emit( struct vertex_fetch *vf );
+
+
+struct vf_format_info {
+   const char *name;
+   vf_extract_func extract;
+   vf_insert_func insert[4];
+   const GLuint attrsize;
+};
+
+const struct vf_format_info vf_format_info[EMIT_MAX];
+
+
+#endif
diff --git a/src/mesa/vf/vf_generic.c b/src/mesa/vf/vf_generic.c
new file mode 100644
index 0000000000..baa00af29a
--- /dev/null
+++ b/src/mesa/vf/vf_generic.c
@@ -0,0 +1,981 @@
+
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keithw@tungstengraphics.com>
+ */
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+#include "main/simple_list.h"
+
+#include "vf/vf.h"
+
+
+/*
+ * These functions take the NDC coordinates pointed to by 'in', apply the
+ * NDC->Viewport mapping and store the results at 'v'.
+ */
+
+static INLINE void insert_4f_viewport_4( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+   out[2] = scale[2] * in[2] + trans[2];
+   out[3] = in[3];
+}
+
+static INLINE void insert_4f_viewport_3( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+   out[2] = scale[2] * in[2] + trans[2];
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_viewport_2( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+   out[2] =                    trans[2];
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_viewport_1( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] =                    trans[1];
+   out[2] =                    trans[2];
+   out[3] = 1;
+}
+
+static INLINE void insert_3f_viewport_3( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+   out[2] = scale[2] * in[2] + trans[2];
+}
+
+static INLINE void insert_3f_viewport_2( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+   out[2] = scale[2] * in[2] + trans[2];
+}
+
+static INLINE void insert_3f_viewport_1( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] =                    trans[1];
+   out[2] =                    trans[2];
+}
+
+static INLINE void insert_2f_viewport_2( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = scale[1] * in[1] + trans[1];
+}
+
+static INLINE void insert_2f_viewport_1( const struct vf_attr *a, GLubyte *v,
+					 const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = scale[0] * in[0] + trans[0];
+   out[1] = trans[1];
+}
+
+
+/*
+ * These functions do the same as above, except for the viewport mapping.
+ */
+
+static INLINE void insert_4f_4( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = in[3];
+}
+
+static INLINE void insert_4f_3( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_2( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = 0;
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static INLINE void insert_3f_xyw_4( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[3];
+}
+
+static INLINE void insert_3f_xyw_err( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   (void) a; (void) v; (void) in;
+   _mesa_exit(1);
+}
+
+static INLINE void insert_3f_3( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+}
+
+static INLINE void insert_3f_2( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+}
+
+static INLINE void insert_3f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = 0;
+   out[2] = 0;
+}
+
+
+static INLINE void insert_2f_2( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+}
+
+static INLINE void insert_2f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = 0;
+}
+
+static INLINE void insert_1f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+
+   out[0] = in[0];
+}
+
+static INLINE void insert_null( const struct vf_attr *a, GLubyte *v, const GLfloat *in )
+{
+   (void) a; (void) v; (void) in;
+}
+
+static INLINE void insert_4chan_4f_rgba_4( const struct vf_attr *a, GLubyte *v, 
+					   const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[3], in[3]);
+}
+
+static INLINE void insert_4chan_4f_rgba_3( const struct vf_attr *a, GLubyte *v, 
+					   const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); 
+   c[3] = CHAN_MAX;
+}
+
+static INLINE void insert_4chan_4f_rgba_2( const struct vf_attr *a, GLubyte *v, 
+					   const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
+   c[2] = 0;
+   c[3] = CHAN_MAX;
+}
+
+static INLINE void insert_4chan_4f_rgba_1( const struct vf_attr *a, GLubyte *v, 
+					   const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   c[1] = 0;
+   c[2] = 0;
+   c[3] = CHAN_MAX;
+}
+
+static INLINE void insert_4ub_4f_rgba_4( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]);
+}
+
+static INLINE void insert_4ub_4f_rgba_3( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_rgba_2( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[2] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_rgba_1( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   v[1] = 0;
+   v[2] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_bgra_4( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]);
+}
+
+static INLINE void insert_4ub_4f_bgra_3( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_bgra_2( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[0] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_bgra_1( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   v[1] = 0;
+   v[0] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_argb_4( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]);
+}
+
+static INLINE void insert_4ub_4f_argb_3( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]);
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_argb_2( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   v[3] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_argb_1( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   v[2] = 0x00;
+   v[3] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_abgr_4( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]);
+}
+
+static INLINE void insert_4ub_4f_abgr_3( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]);
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_abgr_2( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   v[1] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_abgr_1( const struct vf_attr *a, GLubyte *v, 
+					 const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   v[2] = 0x00;
+   v[1] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_3ub_3f_rgb_3( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
+}
+
+static INLINE void insert_3ub_3f_rgb_2( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[2] = 0;
+}
+
+static INLINE void insert_3ub_3f_rgb_1( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   v[1] = 0;
+   v[2] = 0;
+}
+
+static INLINE void insert_3ub_3f_bgr_3( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
+}
+
+static INLINE void insert_3ub_3f_bgr_2( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[0] = 0;
+}
+
+static INLINE void insert_3ub_3f_bgr_1( const struct vf_attr *a, GLubyte *v, 
+					const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   v[1] = 0;
+   v[0] = 0;
+}
+
+
+static INLINE void insert_1ub_1f_1( const struct vf_attr *a, GLubyte *v, 
+				    const GLfloat *in )
+{
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+}
+
+
+/***********************************************************************
+ * Functions to perform the reverse operations to the above, for
+ * swrast translation and clip-interpolation.
+ * 
+ * Currently always extracts a full 4 floats.
+ */
+
+static void extract_4f_viewport( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   /* Although included for completeness, the position coordinate is
+    * usually handled differently during clipping.
+    */
+   out[0] = (in[0] - trans[0]) / scale[0];
+   out[1] = (in[1] - trans[1]) / scale[1];
+   out[2] = (in[2] - trans[2]) / scale[2];
+   out[3] = in[3];
+}
+
+static void extract_3f_viewport( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = (in[0] - trans[0]) / scale[0];
+   out[1] = (in[1] - trans[1]) / scale[1];
+   out[2] = (in[2] - trans[2]) / scale[2];
+   out[3] = 1;
+}
+
+
+static void extract_2f_viewport( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   const GLfloat *scale = a->vf->vp;
+   const GLfloat *trans = a->vf->vp + 4;
+   
+   out[0] = (in[0] - trans[0]) / scale[0];
+   out[1] = (in[1] - trans[1]) / scale[1];
+   out[2] = 0;
+   out[3] = 1;
+}
+
+
+static void extract_4f( const struct vf_attr *a, GLfloat *out, const GLubyte *v  )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = in[3];
+}
+
+static void extract_3f_xyw( const struct vf_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+   out[3] = in[2];
+}
+
+
+static void extract_3f( const struct vf_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = 1;
+}
+
+
+static void extract_2f( const struct vf_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static void extract_1f( const struct vf_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = 0;
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static void extract_4chan_4f_rgba( const struct vf_attr *a, GLfloat *out, 
+				   const GLubyte *v )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+
+   out[0] = CHAN_TO_FLOAT(c[0]);
+   out[1] = CHAN_TO_FLOAT(c[1]);
+   out[2] = CHAN_TO_FLOAT(c[2]);
+   out[3] = CHAN_TO_FLOAT(c[3]);
+}
+
+static void extract_4ub_4f_rgba( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[0] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[2] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_4ub_4f_bgra( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[2] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[0] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_4ub_4f_argb( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[3] = UBYTE_TO_FLOAT(v[0]);
+   out[0] = UBYTE_TO_FLOAT(v[1]);
+   out[1] = UBYTE_TO_FLOAT(v[2]);
+   out[2] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_4ub_4f_abgr( const struct vf_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[3] = UBYTE_TO_FLOAT(v[0]);
+   out[2] = UBYTE_TO_FLOAT(v[1]);
+   out[1] = UBYTE_TO_FLOAT(v[2]);
+   out[0] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_3ub_3f_rgb( const struct vf_attr *a, GLfloat *out, 
+				const GLubyte *v )
+{
+   (void) a;
+   out[0] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[2] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = 1;
+}
+
+static void extract_3ub_3f_bgr( const struct vf_attr *a, GLfloat *out, 
+				const GLubyte *v )
+{
+   (void) a;
+   out[2] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[0] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = 1;
+}
+
+static void extract_1ub_1f( const struct vf_attr *a, GLfloat *out, const GLubyte *v )
+{
+   (void) a;
+   out[0] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = 0;
+   out[2] = 0;
+   out[3] = 1;
+}
+
+
+const struct vf_format_info vf_format_info[EMIT_MAX] = 
+{
+   { "1f",
+     extract_1f,
+     { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 },
+     sizeof(GLfloat) },
+
+   { "2f",
+     extract_2f,
+     { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 },
+     2 * sizeof(GLfloat) },
+
+   { "3f",
+     extract_3f,
+     { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 },
+     3 * sizeof(GLfloat) },
+
+   { "4f",
+     extract_4f,
+     { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 },
+     4 * sizeof(GLfloat) },
+
+   { "2f_viewport",
+     extract_2f_viewport,
+     { insert_2f_viewport_1, insert_2f_viewport_2, insert_2f_viewport_2,
+       insert_2f_viewport_2 },
+     2 * sizeof(GLfloat) },
+
+   { "3f_viewport",
+     extract_3f_viewport,
+     { insert_3f_viewport_1, insert_3f_viewport_2, insert_3f_viewport_3,
+       insert_3f_viewport_3 },
+     3 * sizeof(GLfloat) },
+
+   { "4f_viewport",
+     extract_4f_viewport,
+     { insert_4f_viewport_1, insert_4f_viewport_2, insert_4f_viewport_3,
+       insert_4f_viewport_4 }, 
+     4 * sizeof(GLfloat) },
+
+   { "3f_xyw",
+     extract_3f_xyw,
+     { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, 
+       insert_3f_xyw_4 },
+     3 * sizeof(GLfloat) },
+
+   { "1ub_1f",
+     extract_1ub_1f,
+     { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 },
+     sizeof(GLubyte) },
+
+   { "3ub_3f_rgb",
+     extract_3ub_3f_rgb,
+     { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3,
+       insert_3ub_3f_rgb_3 },
+     3 * sizeof(GLubyte) },
+
+   { "3ub_3f_bgr",
+     extract_3ub_3f_bgr,
+     { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3,
+       insert_3ub_3f_bgr_3 },
+     3 * sizeof(GLubyte) },
+
+   { "4ub_4f_rgba",
+     extract_4ub_4f_rgba,
+     { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, 
+       insert_4ub_4f_rgba_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4ub_4f_bgra",
+     extract_4ub_4f_bgra,
+     { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3,
+       insert_4ub_4f_bgra_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4ub_4f_argb",
+     extract_4ub_4f_argb,
+     { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3,
+       insert_4ub_4f_argb_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4ub_4f_abgr",
+     extract_4ub_4f_abgr,
+     { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3,
+       insert_4ub_4f_abgr_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4chan_4f_rgba",
+     extract_4chan_4f_rgba,
+     { insert_4chan_4f_rgba_1, insert_4chan_4f_rgba_2, insert_4chan_4f_rgba_3,
+       insert_4chan_4f_rgba_4 },
+     4 * sizeof(GLchan) },
+
+   { "pad",
+     NULL,
+     { NULL, NULL, NULL, NULL },
+     0 }
+
+};
+
+
+
+    
+/***********************************************************************
+ * Hardwired fastpaths for emitting whole vertices or groups of
+ * vertices
+ */
+#define EMIT5(NR, F0, F1, F2, F3, F4, NAME)				\
+static void NAME( struct vertex_fetch *vf,				\
+		  GLuint count,						\
+		  GLubyte *v )						\
+{									\
+   struct vf_attr *a = vf->attr;				\
+   GLuint i;								\
+									\
+   for (i = 0 ; i < count ; i++, v += vf->vertex_stride) {		\
+      if (NR > 0) {							\
+	 F0( &a[0], v + a[0].vertoffset, (GLfloat *)a[0].inputptr );	\
+	 a[0].inputptr += a[0].inputstride;				\
+      }									\
+      									\
+      if (NR > 1) {							\
+	 F1( &a[1], v + a[1].vertoffset, (GLfloat *)a[1].inputptr );	\
+	 a[1].inputptr += a[1].inputstride;				\
+      }									\
+      									\
+      if (NR > 2) {							\
+	 F2( &a[2], v + a[2].vertoffset, (GLfloat *)a[2].inputptr );	\
+	 a[2].inputptr += a[2].inputstride;				\
+      }									\
+      									\
+      if (NR > 3) {							\
+	 F3( &a[3], v + a[3].vertoffset, (GLfloat *)a[3].inputptr );	\
+	 a[3].inputptr += a[3].inputstride;				\
+      }									\
+									\
+      if (NR > 4) {							\
+	 F4( &a[4], v + a[4].vertoffset, (GLfloat *)a[4].inputptr );	\
+	 a[4].inputptr += a[4].inputstride;				\
+      }									\
+   }									\
+}
+
+   
+#define EMIT2(F0, F1, NAME) EMIT5(2, F0, F1, insert_null, \
+				  insert_null, insert_null, NAME)
+
+#define EMIT3(F0, F1, F2, NAME) EMIT5(3, F0, F1, F2, insert_null, \
+				      insert_null, NAME)
+   
+#define EMIT4(F0, F1, F2, F3, NAME) EMIT5(4, F0, F1, F2, F3, \
+				          insert_null, NAME)
+   
+
+EMIT2(insert_3f_viewport_3, insert_4ub_4f_rgba_4, emit_viewport3_rgba4)
+EMIT2(insert_3f_viewport_3, insert_4ub_4f_bgra_4, emit_viewport3_bgra4)
+EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4)
+
+EMIT3(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_viewport4_rgba4_st2)
+EMIT3(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2,  emit_viewport4_bgra4_st2)
+EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2)
+
+EMIT4(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_viewport4_rgba4_st2_st2)
+EMIT4(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2, insert_2f_2,  emit_viewport4_bgra4_st2_st2)
+EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2)
+
+
+/* Use the codegen paths to select one of a number of hardwired
+ * fastpaths.
+ */
+void vf_generate_hardwired_emit( struct vertex_fetch *vf )
+{
+   vf_emit_func func = NULL;
+
+   /* Does it fit a hardwired fastpath?  Help! this is growing out of
+    * control!
+    */
+   switch (vf->attr_count) {
+   case 2:
+      if (vf->attr[0].do_insert == insert_3f_viewport_3) {
+	 if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4) 
+	    func = emit_viewport3_bgra4;
+	 else if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) 
+	    func = emit_viewport3_rgba4;
+      }
+      else if (vf->attr[0].do_insert == insert_3f_3 &&
+	       vf->attr[1].do_insert == insert_4ub_4f_rgba_4) {
+ 	 func = emit_xyz3_rgba4; 
+      }
+      break;
+   case 3:
+      if (vf->attr[2].do_insert == insert_2f_2) {
+	 if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) {
+	    if (vf->attr[0].do_insert == insert_4f_viewport_4)
+	       func = emit_viewport4_rgba4_st2;
+	    else if (vf->attr[0].do_insert == insert_4f_4) 
+	       func = emit_xyzw4_rgba4_st2;
+	 }
+	 else if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4 &&
+		  vf->attr[0].do_insert == insert_4f_viewport_4)
+	    func = emit_viewport4_bgra4_st2;
+      }
+      break;
+   case 4:
+      if (vf->attr[2].do_insert == insert_2f_2 &&
+	  vf->attr[3].do_insert == insert_2f_2) {
+	 if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) {
+	    if (vf->attr[0].do_insert == insert_4f_viewport_4)
+	       func = emit_viewport4_rgba4_st2_st2;
+	    else if (vf->attr[0].do_insert == insert_4f_4) 
+	       func = emit_xyzw4_rgba4_st2_st2;
+	 }
+	 else if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4 &&
+		  vf->attr[0].do_insert == insert_4f_viewport_4)
+	    func = emit_viewport4_bgra4_st2_st2;
+      }
+      break;
+   }
+
+   vf->emit = func;
+}
+
+/***********************************************************************
+ * Generic (non-codegen) functions for whole vertices or groups of
+ * vertices
+ */
+
+void vf_generic_emit( struct vertex_fetch *vf,
+		      GLuint count,
+		      GLubyte *v )
+{
+   struct vf_attr *a = vf->attr;
+   const GLuint attr_count = vf->attr_count;
+   const GLuint stride = vf->vertex_stride;
+   GLuint i, j;
+
+   for (i = 0 ; i < count ; i++, v += stride) {
+      for (j = 0; j < attr_count; j++) {
+	 GLfloat *in = (GLfloat *)a[j].inputptr;
+	 a[j].inputptr += a[j].inputstride;
+	 a[j].do_insert( &a[j], v + a[j].vertoffset, in );
+      }
+   }
+}
+
+
diff --git a/src/mesa/vf/vf_sse.c b/src/mesa/vf/vf_sse.c
new file mode 100644
index 0000000000..4d70196ffe
--- /dev/null
+++ b/src/mesa/vf/vf_sse.c
@@ -0,0 +1,662 @@
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Keith Whitwell <keithw@tungstengraphics.com>
+ */
+
+#include "main/glheader.h"
+#include "main/colormac.h"
+#include "main/simple_list.h"
+#include "main/enums.h"
+
+#include "vf/vf.h"
+
+#if defined(USE_SSE_ASM)
+
+#include "x86/rtasm/x86sse.h"
+#include "x86/common_x86_asm.h"
+
+
+#define X    0
+#define Y    1
+#define Z    2
+#define W    3
+
+
+struct x86_program {
+   struct x86_function func;
+
+   struct vertex_fetch *vf;
+   GLboolean inputs_safe;
+   GLboolean outputs_safe;
+   GLboolean have_sse2;
+   
+   struct x86_reg identity;
+   struct x86_reg chan0;
+};
+
+
+static struct x86_reg get_identity( struct x86_program *p )
+{
+   return p->identity;
+}
+
+static void emit_load4f_4( struct x86_program *p, 			   
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movups(&p->func, dest, arg0);
+}
+
+static void emit_load4f_3( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Have to jump through some hoops:
+    *
+    * c 0 0 0
+    * c 0 0 1
+    * 0 0 c 1
+    * a b c 1
+    */
+   sse_movss(&p->func, dest, x86_make_disp(arg0, 8));
+   sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) );
+   sse_shufps(&p->func, dest, dest, SHUF(Y,Z,X,W) );
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_load4f_2( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Initialize from identity, then pull in low two words:
+    */
+   sse_movups(&p->func, dest, get_identity(p));
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_load4f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Pull in low word, then swizzle in identity */
+   sse_movss(&p->func, dest, arg0);
+   sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) );
+}
+
+
+
+static void emit_load3f_3( struct x86_program *p, 			   
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Over-reads by 1 dword - potential SEGV if input is a vertex
+    * array.
+    */
+   if (p->inputs_safe) {
+      sse_movups(&p->func, dest, arg0);
+   } 
+   else {
+      /* c 0 0 0
+       * c c c c
+       * a b c c 
+       */
+      sse_movss(&p->func, dest, x86_make_disp(arg0, 8));
+      sse_shufps(&p->func, dest, dest, SHUF(X,X,X,X));
+      sse_movlps(&p->func, dest, arg0);
+   }
+}
+
+static void emit_load3f_2( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   emit_load4f_2(p, dest, arg0);
+}
+
+static void emit_load3f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   emit_load4f_1(p, dest, arg0);
+}
+
+static void emit_load2f_2( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_load2f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   emit_load4f_1(p, dest, arg0);
+}
+
+static void emit_load1f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movss(&p->func, dest, arg0);
+}
+
+static void (*load[4][4])( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 ) = {
+   { emit_load1f_1, 
+     emit_load1f_1, 
+     emit_load1f_1, 
+     emit_load1f_1 },
+
+   { emit_load2f_1, 
+     emit_load2f_2, 
+     emit_load2f_2, 
+     emit_load2f_2 },
+
+   { emit_load3f_1, 
+     emit_load3f_2, 
+     emit_load3f_3, 
+     emit_load3f_3 },
+
+   { emit_load4f_1, 
+     emit_load4f_2, 
+     emit_load4f_3, 
+     emit_load4f_4 } 
+};
+
+static void emit_load( struct x86_program *p,
+		       struct x86_reg dest,
+		       GLuint sz,
+		       struct x86_reg src,
+		       GLuint src_sz)
+{
+   load[sz-1][src_sz-1](p, dest, src);
+}
+
+static void emit_store4f( struct x86_program *p, 			   
+			  struct x86_reg dest,
+			  struct x86_reg arg0 )
+{
+   sse_movups(&p->func, dest, arg0);
+}
+
+static void emit_store3f( struct x86_program *p, 
+			  struct x86_reg dest,
+			  struct x86_reg arg0 )
+{
+   if (p->outputs_safe) {
+      /* Emit the extra dword anyway.  This may hurt writecombining,
+       * may cause other problems.
+       */
+      sse_movups(&p->func, dest, arg0);
+   }
+   else {
+      /* Alternate strategy - emit two, shuffle, emit one.
+       */
+      sse_movlps(&p->func, dest, arg0);
+      sse_shufps(&p->func, arg0, arg0, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */
+      sse_movss(&p->func, x86_make_disp(dest,8), arg0);
+   }
+}
+
+static void emit_store2f( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_store1f( struct x86_program *p, 
+			  struct x86_reg dest,
+			  struct x86_reg arg0 )
+{
+   sse_movss(&p->func, dest, arg0);
+}
+
+
+static void (*store[4])( struct x86_program *p, 
+			 struct x86_reg dest,
+			 struct x86_reg arg0 ) = 
+{
+   emit_store1f, 
+   emit_store2f, 
+   emit_store3f, 
+   emit_store4f 
+};
+
+static void emit_store( struct x86_program *p,
+			struct x86_reg dest,
+			GLuint sz,
+			struct x86_reg temp )
+
+{
+   store[sz-1](p, dest, temp);
+}
+
+static void emit_pack_store_4ub( struct x86_program *p,
+				 struct x86_reg dest,
+				 struct x86_reg temp )
+{
+   /* Scale by 255.0
+    */
+   sse_mulps(&p->func, temp, p->chan0);
+
+   if (p->have_sse2) {
+      sse2_cvtps2dq(&p->func, temp, temp);
+      sse2_packssdw(&p->func, temp, temp);
+      sse2_packuswb(&p->func, temp, temp);
+      sse_movss(&p->func, dest, temp);
+   }
+   else {
+      struct x86_reg mmx0 = x86_make_reg(file_MMX, 0);
+      struct x86_reg mmx1 = x86_make_reg(file_MMX, 1);
+      sse_cvtps2pi(&p->func, mmx0, temp);
+      sse_movhlps(&p->func, temp, temp);
+      sse_cvtps2pi(&p->func, mmx1, temp);
+      mmx_packssdw(&p->func, mmx0, mmx1);
+      mmx_packuswb(&p->func, mmx0, mmx0);
+      mmx_movd(&p->func, dest, mmx0);
+   }
+}
+
+static GLint get_offset( const void *a, const void *b )
+{
+   return (const char *)b - (const char *)a;
+}
+
+/* Not much happens here.  Eventually use this function to try and
+ * avoid saving/reloading the source pointers each vertex (if some of
+ * them can fit in registers).
+ */
+static void get_src_ptr( struct x86_program *p,
+			 struct x86_reg srcREG,
+			 struct x86_reg vfREG,
+			 struct vf_attr *a )
+{
+   struct vertex_fetch *vf = p->vf;
+   struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr));
+
+   /* Load current a[j].inputptr
+    */
+   x86_mov(&p->func, srcREG, ptr_to_src);
+}
+
+static void update_src_ptr( struct x86_program *p,
+			 struct x86_reg srcREG,
+			 struct x86_reg vfREG,
+			 struct vf_attr *a )
+{
+   if (a->inputstride) {
+      struct vertex_fetch *vf = p->vf;
+      struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr));
+
+      /* add a[j].inputstride (hardcoded value - could just as easily
+       * pull the stride value from memory each time).
+       */
+      x86_lea(&p->func, srcREG, x86_make_disp(srcREG, a->inputstride));
+      
+      /* save new value of a[j].inputptr 
+       */
+      x86_mov(&p->func, ptr_to_src, srcREG);
+   }
+}
+
+
+/* Lots of hardcoding
+ *
+ * EAX -- pointer to current output vertex
+ * ECX -- pointer to current attribute 
+ * 
+ */
+static GLboolean build_vertex_emit( struct x86_program *p )
+{
+   struct vertex_fetch *vf = p->vf;
+   GLuint j = 0;
+
+   struct x86_reg vertexEAX = x86_make_reg(file_REG32, reg_AX);
+   struct x86_reg srcECX = x86_make_reg(file_REG32, reg_CX);
+   struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP);
+   struct x86_reg vfESI = x86_make_reg(file_REG32, reg_SI);
+   struct x86_reg temp = x86_make_reg(file_XMM, 0);
+   struct x86_reg vp0 = x86_make_reg(file_XMM, 1);
+   struct x86_reg vp1 = x86_make_reg(file_XMM, 2);
+   GLubyte *fixup, *label;
+
+   /* Push a few regs?
+    */
+   x86_push(&p->func, countEBP);
+   x86_push(&p->func, vfESI);
+
+
+   /* Get vertex count, compare to zero
+    */
+   x86_xor(&p->func, srcECX, srcECX);
+   x86_mov(&p->func, countEBP, x86_fn_arg(&p->func, 2));
+   x86_cmp(&p->func, countEBP, srcECX);
+   fixup = x86_jcc_forward(&p->func, cc_E);
+
+   /* Initialize destination register. 
+    */
+   x86_mov(&p->func, vertexEAX, x86_fn_arg(&p->func, 3));
+
+   /* Move argument 1 (vf) into a reg:
+    */
+   x86_mov(&p->func, vfESI, x86_fn_arg(&p->func, 1));
+
+   
+   /* Possibly load vp0, vp1 for viewport calcs:
+    */
+   if (vf->allow_viewport_emits) {
+      sse_movups(&p->func, vp0, x86_make_disp(vfESI, get_offset(vf, &vf->vp[0])));
+      sse_movups(&p->func, vp1, x86_make_disp(vfESI, get_offset(vf, &vf->vp[4])));
+   }
+
+   /* always load, needed or not:
+    */
+   sse_movups(&p->func, p->chan0, x86_make_disp(vfESI, get_offset(vf, &vf->chan_scale[0])));
+   sse_movups(&p->func, p->identity, x86_make_disp(vfESI, get_offset(vf, &vf->identity[0])));
+
+   /* Note address for loop jump */
+   label = x86_get_label(&p->func);
+
+   /* Emit code for each of the attributes.  Currently routes
+    * everything through SSE registers, even when it might be more
+    * efficient to stick with regular old x86.  No optimization or
+    * other tricks - enough new ground to cover here just getting
+    * things working.
+    */
+   while (j < vf->attr_count) {
+      struct vf_attr *a = &vf->attr[j];
+      struct x86_reg dest = x86_make_disp(vertexEAX, a->vertoffset);
+
+      /* Now, load an XMM reg from src, perhaps transform, then save.
+       * Could be shortcircuited in specific cases:
+       */
+      switch (a->format) {
+      case EMIT_1F:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize);
+	 emit_store(p, dest, 1, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_2F:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize);
+	 emit_store(p, dest, 2, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_3F:
+	 /* Potentially the worst case - hardcode 2+1 copying:
+	  */
+	 if (0) {
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	    emit_store(p, dest, 3, temp);
+	    update_src_ptr(p, srcECX, vfESI, a);
+	 }
+	 else {
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize);
+	    emit_store(p, dest, 2, temp);
+	    if (a->inputsize > 2) {
+	       emit_load(p, temp, 1, x86_make_disp(srcECX, 8), 1);
+	       emit_store(p, x86_make_disp(dest,8), 1, temp);
+	    }
+	    else {
+	       sse_movss(&p->func, x86_make_disp(dest,8), get_identity(p));
+	    }
+	    update_src_ptr(p, srcECX, vfESI, a);
+	 }
+	 break;
+      case EMIT_4F:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 emit_store(p, dest, 4, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_2F_VIEWPORT: 
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize);
+	 sse_mulps(&p->func, temp, vp0);
+	 sse_addps(&p->func, temp, vp1);
+	 emit_store(p, dest, 2, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_3F_VIEWPORT: 
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	 sse_mulps(&p->func, temp, vp0);
+	 sse_addps(&p->func, temp, vp1);
+	 emit_store(p, dest, 3, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_4F_VIEWPORT: 
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_mulps(&p->func, temp, vp0);
+	 sse_addps(&p->func, temp, vp1);
+	 emit_store(p, dest, 4, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_3F_XYW:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z));
+	 emit_store(p, dest, 3, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+
+      case EMIT_1UB_1F:	 
+	 /* Test for PAD3 + 1UB:
+	  */
+	 if (j > 0 &&
+	     a[-1].vertoffset + a[-1].vertattrsize <= a->vertoffset - 3)
+	 {
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize);
+	    sse_shufps(&p->func, temp, temp, SHUF(X,X,X,X));
+	    emit_pack_store_4ub(p, x86_make_disp(dest, -3), temp); /* overkill! */
+	    update_src_ptr(p, srcECX, vfESI, a);
+	 }
+	 else {
+	    _mesa_printf("Can't emit 1ub %x %x %d\n", a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize );
+	    return GL_FALSE;
+	 }
+	 break;
+      case EMIT_3UB_3F_RGB:
+      case EMIT_3UB_3F_BGR:
+	 /* Test for 3UB + PAD1:
+	  */
+	 if (j == vf->attr_count - 1 ||
+	     a[1].vertoffset >= a->vertoffset + 4) {
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	    if (a->format == EMIT_3UB_3F_BGR)
+	       sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W));
+	    emit_pack_store_4ub(p, dest, temp);
+	    update_src_ptr(p, srcECX, vfESI, a);
+	 }
+	 /* Test for 3UB + 1UB:
+	  */
+	 else if (j < vf->attr_count - 1 &&
+		  a[1].format == EMIT_1UB_1F &&
+		  a[1].vertoffset == a->vertoffset + 3) {
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	    update_src_ptr(p, srcECX, vfESI, a);
+
+	    /* Make room for incoming value:
+	     */
+	    sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z));
+
+	    get_src_ptr(p, srcECX, vfESI, &a[1]);
+	    emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize);
+	    update_src_ptr(p, srcECX, vfESI, &a[1]);
+
+	    /* Rearrange and possibly do BGR conversion:
+	     */
+	    if (a->format == EMIT_3UB_3F_BGR)
+	       sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X));
+	    else
+	       sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X));
+
+	    emit_pack_store_4ub(p, dest, temp);
+	    j++;		/* NOTE: two attrs consumed */
+	 }
+	 else {
+	    _mesa_printf("Can't emit 3ub\n");
+	 }
+	 return GL_FALSE;	/* add this later */
+	 break;
+
+      case EMIT_4UB_4F_RGBA:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_4UB_4F_BGRA:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W));
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_4UB_4F_ARGB:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z));
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_4UB_4F_ABGR:
+	 get_src_ptr(p, srcECX, vfESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X));
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vfESI, a);
+	 break;
+      case EMIT_4CHAN_4F_RGBA:
+	 switch (CHAN_TYPE) {
+	 case GL_UNSIGNED_BYTE:
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	    emit_pack_store_4ub(p, dest, temp);
+	    update_src_ptr(p, srcECX, vfESI, a);
+	    break;
+	 case GL_FLOAT:
+	    get_src_ptr(p, srcECX, vfESI, a);
+	    emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	    emit_store(p, dest, 4, temp);
+	    update_src_ptr(p, srcECX, vfESI, a);
+	    break;
+	 case GL_UNSIGNED_SHORT:
+	 default:
+	    _mesa_printf("unknown CHAN_TYPE %s\n", _mesa_lookup_enum_by_nr(CHAN_TYPE));
+	    return GL_FALSE;
+	 }
+	 break;
+      default:
+	 _mesa_printf("unknown a[%d].format %d\n", j, a->format);
+	 return GL_FALSE;	/* catch any new opcodes */
+      }
+      
+      /* Increment j by at least 1 - may have been incremented above also:
+       */
+      j++;
+   }
+
+   /* Next vertex:
+    */
+   x86_lea(&p->func, vertexEAX, x86_make_disp(vertexEAX, vf->vertex_stride));
+
+   /* decr count, loop if not zero
+    */
+   x86_dec(&p->func, countEBP);
+   x86_test(&p->func, countEBP, countEBP); 
+   x86_jcc(&p->func, cc_NZ, label);
+
+   /* Exit mmx state?
+    */
+   if (p->func.need_emms)
+      mmx_emms(&p->func);
+
+   /* Land forward jump here:
+    */
+   x86_fixup_fwd_jump(&p->func, fixup);
+
+   /* Pop regs and return
+    */
+   x86_pop(&p->func, x86_get_base_reg(vfESI));
+   x86_pop(&p->func, countEBP);
+   x86_ret(&p->func);
+
+   vf->emit = (vf_emit_func)x86_get_func(&p->func);
+   return GL_TRUE;
+}
+
+
+
+void vf_generate_sse_emit( struct vertex_fetch *vf )
+{
+   struct x86_program p;   
+
+   if (!cpu_has_xmm) {
+      vf->codegen_emit = NULL;
+      return;
+   }
+
+   _mesa_memset(&p, 0, sizeof(p));
+
+   p.vf = vf;
+   p.inputs_safe = 0;		/* for now */
+   p.outputs_safe = 0;		/* for now */
+   p.have_sse2 = cpu_has_xmm2;
+   p.identity = x86_make_reg(file_XMM, 6);
+   p.chan0 = x86_make_reg(file_XMM, 7);
+
+   x86_init_func(&p.func);
+
+   if (build_vertex_emit(&p)) {
+      vf_register_fastpath( vf, GL_TRUE );
+   }
+   else {
+      /* Note the failure so that we don't keep trying to codegen an
+       * impossible state:
+       */
+      vf_register_fastpath( vf, GL_FALSE );
+      x86_release_func(&p.func);
+   }
+}
+
+#else
+
+void vf_generate_sse_emit( struct vertex_fetch *vf )
+{
+   /* Dummy version for when USE_SSE_ASM not defined */
+}
+
+#endif
diff --git a/src/mesa/x86-64/glapi_x86-64.S b/src/mesa/x86-64/glapi_x86-64.S
index cb34061b36..1eb4358b1c 100644
--- a/src/mesa/x86-64/glapi_x86-64.S
+++ b/src/mesa/x86-64/glapi_x86-64.S
@@ -21112,21 +21112,25 @@ GL_PREFIX(DrawBuffersARB):
 	.size	GL_PREFIX(DrawBuffersARB), .-GL_PREFIX(DrawBuffersARB)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(PolygonOffsetEXT)
-	.type	GL_PREFIX(PolygonOffsetEXT), @function
-GL_PREFIX(PolygonOffsetEXT):
+	.globl	GL_PREFIX(RenderbufferStorageMultisample)
+	.type	GL_PREFIX(RenderbufferStorageMultisample), @function
+GL_PREFIX(RenderbufferStorageMultisample):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
 	movq	4488(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
-	subq	$24, %rsp
-	movq	%xmm0, (%rsp)
-	movq	%xmm1, 8(%rsp)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%r8
 	call	_x86_64_get_dispatch@PLT
-	movq	8(%rsp), %xmm1
-	movq	(%rsp), %xmm0
-	addq	$24, %rsp
+	popq	%r8
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
 	movq	4488(%rax), %r11
 	jmp	*%r11
 #else
@@ -21136,35 +21140,38 @@ GL_PREFIX(PolygonOffsetEXT):
 	movq	4488(%rax), %r11
 	jmp	*%r11
 1:
-	subq	$24, %rsp
-	movq	%xmm0, (%rsp)
-	movq	%xmm1, 8(%rsp)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%r8
 	call	_glapi_get_dispatch
-	movq	8(%rsp), %xmm1
-	movq	(%rsp), %xmm0
-	addq	$24, %rsp
+	popq	%r8
+	popq	%rcx
+	popq	%rdx
+	popq	%rsi
+	popq	%rdi
 	movq	4488(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(PolygonOffsetEXT), .-GL_PREFIX(PolygonOffsetEXT)
+	.size	GL_PREFIX(RenderbufferStorageMultisample), .-GL_PREFIX(RenderbufferStorageMultisample)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_562)
-	.type	GL_PREFIX(_dispatch_stub_562), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_562))
-GL_PREFIX(_dispatch_stub_562):
+	.globl	GL_PREFIX(PolygonOffsetEXT)
+	.type	GL_PREFIX(PolygonOffsetEXT), @function
+GL_PREFIX(PolygonOffsetEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
 	movq	4496(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
-	pushq	%rdi
-	pushq	%rsi
-	pushq	%rbp
+	subq	$24, %rsp
+	movq	%xmm0, (%rsp)
+	movq	%xmm1, 8(%rsp)
 	call	_x86_64_get_dispatch@PLT
-	popq	%rbp
-	popq	%rsi
-	popq	%rdi
+	movq	8(%rsp), %xmm1
+	movq	(%rsp), %xmm0
+	addq	$24, %rsp
 	movq	4496(%rax), %r11
 	jmp	*%r11
 #else
@@ -21174,17 +21181,17 @@ GL_PREFIX(_dispatch_stub_562):
 	movq	4496(%rax), %r11
 	jmp	*%r11
 1:
-	pushq	%rdi
-	pushq	%rsi
-	pushq	%rbp
+	subq	$24, %rsp
+	movq	%xmm0, (%rsp)
+	movq	%xmm1, 8(%rsp)
 	call	_glapi_get_dispatch
-	popq	%rbp
-	popq	%rsi
-	popq	%rdi
+	movq	8(%rsp), %xmm1
+	movq	(%rsp), %xmm0
+	addq	$24, %rsp
 	movq	4496(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_562), .-GL_PREFIX(_dispatch_stub_562)
+	.size	GL_PREFIX(PolygonOffsetEXT), .-GL_PREFIX(PolygonOffsetEXT)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(_dispatch_stub_563)
@@ -21234,13 +21241,13 @@ GL_PREFIX(_dispatch_stub_564):
 	movq	4512(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
-	subq	$24, %rsp
-	movq	%rdi, (%rsp)
-	movq	%xmm0, 8(%rsp)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
-	movq	8(%rsp), %xmm0
-	movq	(%rsp), %rdi
-	addq	$24, %rsp
+	popq	%rbp
+	popq	%rsi
+	popq	%rdi
 	movq	4512(%rax), %r11
 	jmp	*%r11
 #else
@@ -21250,13 +21257,13 @@ GL_PREFIX(_dispatch_stub_564):
 	movq	4512(%rax), %r11
 	jmp	*%r11
 1:
-	subq	$24, %rsp
-	movq	%rdi, (%rsp)
-	movq	%xmm0, 8(%rsp)
+	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
 	call	_glapi_get_dispatch
-	movq	8(%rsp), %xmm0
-	movq	(%rsp), %rdi
-	addq	$24, %rsp
+	popq	%rbp
+	popq	%rsi
+	popq	%rdi
 	movq	4512(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
@@ -21272,13 +21279,13 @@ GL_PREFIX(_dispatch_stub_565):
 	movq	4520(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
-	pushq	%rdi
-	pushq	%rsi
-	pushq	%rbp
+	subq	$24, %rsp
+	movq	%rdi, (%rsp)
+	movq	%xmm0, 8(%rsp)
 	call	_x86_64_get_dispatch@PLT
-	popq	%rbp
-	popq	%rsi
-	popq	%rdi
+	movq	8(%rsp), %xmm0
+	movq	(%rsp), %rdi
+	addq	$24, %rsp
 	movq	4520(%rax), %r11
 	jmp	*%r11
 #else
@@ -21288,13 +21295,13 @@ GL_PREFIX(_dispatch_stub_565):
 	movq	4520(%rax), %r11
 	jmp	*%r11
 1:
-	pushq	%rdi
-	pushq	%rsi
-	pushq	%rbp
+	subq	$24, %rsp
+	movq	%rdi, (%rsp)
+	movq	%xmm0, 8(%rsp)
 	call	_glapi_get_dispatch
-	popq	%rbp
-	popq	%rsi
-	popq	%rdi
+	movq	8(%rsp), %xmm0
+	movq	(%rsp), %rdi
+	addq	$24, %rsp
 	movq	4520(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
@@ -21425,7 +21432,11 @@ GL_PREFIX(_dispatch_stub_569):
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
+	popq	%rbp
+	popq	%rsi
 	popq	%rdi
 	movq	4552(%rax), %r11
 	jmp	*%r11
@@ -21437,7 +21448,11 @@ GL_PREFIX(_dispatch_stub_569):
 	jmp	*%r11
 1:
 	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
 	call	_glapi_get_dispatch
+	popq	%rbp
+	popq	%rsi
 	popq	%rdi
 	movq	4552(%rax), %r11
 	jmp	*%r11
@@ -21445,12 +21460,42 @@ GL_PREFIX(_dispatch_stub_569):
 	.size	GL_PREFIX(_dispatch_stub_569), .-GL_PREFIX(_dispatch_stub_569)
 
 	.p2align	4,,15
+	.globl	GL_PREFIX(_dispatch_stub_570)
+	.type	GL_PREFIX(_dispatch_stub_570), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_570))
+GL_PREFIX(_dispatch_stub_570):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	4560(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdi
+	movq	4560(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	4560(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	call	_glapi_get_dispatch
+	popq	%rdi
+	movq	4560(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(_dispatch_stub_570), .-GL_PREFIX(_dispatch_stub_570)
+
+	.p2align	4,,15
 	.globl	GL_PREFIX(ColorPointerEXT)
 	.type	GL_PREFIX(ColorPointerEXT), @function
 GL_PREFIX(ColorPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4560(%rax), %r11
+	movq	4568(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21464,13 +21509,13 @@ GL_PREFIX(ColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4560(%rax), %r11
+	movq	4568(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4560(%rax), %r11
+	movq	4568(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21484,7 +21529,7 @@ GL_PREFIX(ColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4560(%rax), %r11
+	movq	4568(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorPointerEXT), .-GL_PREFIX(ColorPointerEXT)
@@ -21495,7 +21540,7 @@ GL_PREFIX(ColorPointerEXT):
 GL_PREFIX(EdgeFlagPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4568(%rax), %r11
+	movq	4576(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21505,13 +21550,13 @@ GL_PREFIX(EdgeFlagPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4568(%rax), %r11
+	movq	4576(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4568(%rax), %r11
+	movq	4576(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21521,7 +21566,7 @@ GL_PREFIX(EdgeFlagPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4568(%rax), %r11
+	movq	4576(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(EdgeFlagPointerEXT), .-GL_PREFIX(EdgeFlagPointerEXT)
@@ -21532,7 +21577,7 @@ GL_PREFIX(EdgeFlagPointerEXT):
 GL_PREFIX(IndexPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4576(%rax), %r11
+	movq	4584(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21546,13 +21591,13 @@ GL_PREFIX(IndexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4576(%rax), %r11
+	movq	4584(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4576(%rax), %r11
+	movq	4584(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21566,7 +21611,7 @@ GL_PREFIX(IndexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4576(%rax), %r11
+	movq	4584(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IndexPointerEXT), .-GL_PREFIX(IndexPointerEXT)
@@ -21577,7 +21622,7 @@ GL_PREFIX(IndexPointerEXT):
 GL_PREFIX(NormalPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4584(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21591,13 +21636,13 @@ GL_PREFIX(NormalPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4584(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4584(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21611,7 +21656,7 @@ GL_PREFIX(NormalPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4584(%rax), %r11
+	movq	4592(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(NormalPointerEXT), .-GL_PREFIX(NormalPointerEXT)
@@ -21622,7 +21667,7 @@ GL_PREFIX(NormalPointerEXT):
 GL_PREFIX(TexCoordPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4592(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21636,13 +21681,13 @@ GL_PREFIX(TexCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4592(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4592(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21656,7 +21701,7 @@ GL_PREFIX(TexCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4592(%rax), %r11
+	movq	4600(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(TexCoordPointerEXT), .-GL_PREFIX(TexCoordPointerEXT)
@@ -21667,7 +21712,7 @@ GL_PREFIX(TexCoordPointerEXT):
 GL_PREFIX(VertexPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4600(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21681,13 +21726,13 @@ GL_PREFIX(VertexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4600(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4600(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21701,7 +21746,7 @@ GL_PREFIX(VertexPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4600(%rax), %r11
+	movq	4608(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexPointerEXT), .-GL_PREFIX(VertexPointerEXT)
@@ -21712,7 +21757,7 @@ GL_PREFIX(VertexPointerEXT):
 GL_PREFIX(PointParameterfEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4608(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -21722,13 +21767,13 @@ GL_PREFIX(PointParameterfEXT):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4608(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4608(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -21738,7 +21783,7 @@ GL_PREFIX(PointParameterfEXT):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4608(%rax), %r11
+	movq	4616(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameterfEXT), .-GL_PREFIX(PointParameterfEXT)
@@ -21749,7 +21794,7 @@ GL_PREFIX(PointParameterfEXT):
 GL_PREFIX(PointParameterfvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4616(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21759,13 +21804,13 @@ GL_PREFIX(PointParameterfvEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4616(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4616(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21775,7 +21820,7 @@ GL_PREFIX(PointParameterfvEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4616(%rax), %r11
+	movq	4624(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameterfvEXT), .-GL_PREFIX(PointParameterfvEXT)
@@ -21786,7 +21831,7 @@ GL_PREFIX(PointParameterfvEXT):
 GL_PREFIX(LockArraysEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4624(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21796,13 +21841,13 @@ GL_PREFIX(LockArraysEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4624(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4624(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21812,7 +21857,7 @@ GL_PREFIX(LockArraysEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4624(%rax), %r11
+	movq	4632(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(LockArraysEXT), .-GL_PREFIX(LockArraysEXT)
@@ -21823,37 +21868,37 @@ GL_PREFIX(LockArraysEXT):
 GL_PREFIX(UnlockArraysEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4632(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	4632(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4632(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	4632(%rax), %r11
+	movq	4640(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(UnlockArraysEXT), .-GL_PREFIX(UnlockArraysEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_580)
-	.type	GL_PREFIX(_dispatch_stub_580), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_580))
-GL_PREFIX(_dispatch_stub_580):
+	.globl	GL_PREFIX(_dispatch_stub_581)
+	.type	GL_PREFIX(_dispatch_stub_581), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_581))
+GL_PREFIX(_dispatch_stub_581):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4640(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21863,13 +21908,13 @@ GL_PREFIX(_dispatch_stub_580):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4640(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4640(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21879,19 +21924,19 @@ GL_PREFIX(_dispatch_stub_580):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4640(%rax), %r11
+	movq	4648(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_580), .-GL_PREFIX(_dispatch_stub_580)
+	.size	GL_PREFIX(_dispatch_stub_581), .-GL_PREFIX(_dispatch_stub_581)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_581)
-	.type	GL_PREFIX(_dispatch_stub_581), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_581))
-GL_PREFIX(_dispatch_stub_581):
+	.globl	GL_PREFIX(_dispatch_stub_582)
+	.type	GL_PREFIX(_dispatch_stub_582), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_582))
+GL_PREFIX(_dispatch_stub_582):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4648(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21901,13 +21946,13 @@ GL_PREFIX(_dispatch_stub_581):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4648(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4648(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21917,10 +21962,10 @@ GL_PREFIX(_dispatch_stub_581):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4648(%rax), %r11
+	movq	4656(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_581), .-GL_PREFIX(_dispatch_stub_581)
+	.size	GL_PREFIX(_dispatch_stub_582), .-GL_PREFIX(_dispatch_stub_582)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(SecondaryColor3bEXT)
@@ -21928,7 +21973,7 @@ GL_PREFIX(_dispatch_stub_581):
 GL_PREFIX(SecondaryColor3bEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4656(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -21938,13 +21983,13 @@ GL_PREFIX(SecondaryColor3bEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4656(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4656(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -21954,7 +21999,7 @@ GL_PREFIX(SecondaryColor3bEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4656(%rax), %r11
+	movq	4664(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3bEXT), .-GL_PREFIX(SecondaryColor3bEXT)
@@ -21965,25 +22010,25 @@ GL_PREFIX(SecondaryColor3bEXT):
 GL_PREFIX(SecondaryColor3bvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4664(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4664(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4664(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4664(%rax), %r11
+	movq	4672(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3bvEXT), .-GL_PREFIX(SecondaryColor3bvEXT)
@@ -21994,7 +22039,7 @@ GL_PREFIX(SecondaryColor3bvEXT):
 GL_PREFIX(SecondaryColor3dEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4672(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -22006,13 +22051,13 @@ GL_PREFIX(SecondaryColor3dEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4672(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4672(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -22024,7 +22069,7 @@ GL_PREFIX(SecondaryColor3dEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4672(%rax), %r11
+	movq	4680(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3dEXT), .-GL_PREFIX(SecondaryColor3dEXT)
@@ -22035,25 +22080,25 @@ GL_PREFIX(SecondaryColor3dEXT):
 GL_PREFIX(SecondaryColor3dvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4680(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4680(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4680(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4680(%rax), %r11
+	movq	4688(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3dvEXT), .-GL_PREFIX(SecondaryColor3dvEXT)
@@ -22064,7 +22109,7 @@ GL_PREFIX(SecondaryColor3dvEXT):
 GL_PREFIX(SecondaryColor3fEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4688(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -22076,13 +22121,13 @@ GL_PREFIX(SecondaryColor3fEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4688(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4688(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -22094,7 +22139,7 @@ GL_PREFIX(SecondaryColor3fEXT):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4688(%rax), %r11
+	movq	4696(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3fEXT), .-GL_PREFIX(SecondaryColor3fEXT)
@@ -22105,25 +22150,25 @@ GL_PREFIX(SecondaryColor3fEXT):
 GL_PREFIX(SecondaryColor3fvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4696(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4696(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4696(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4696(%rax), %r11
+	movq	4704(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3fvEXT), .-GL_PREFIX(SecondaryColor3fvEXT)
@@ -22134,7 +22179,7 @@ GL_PREFIX(SecondaryColor3fvEXT):
 GL_PREFIX(SecondaryColor3iEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4704(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22144,13 +22189,13 @@ GL_PREFIX(SecondaryColor3iEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4704(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4704(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22160,7 +22205,7 @@ GL_PREFIX(SecondaryColor3iEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4704(%rax), %r11
+	movq	4712(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3iEXT), .-GL_PREFIX(SecondaryColor3iEXT)
@@ -22171,25 +22216,25 @@ GL_PREFIX(SecondaryColor3iEXT):
 GL_PREFIX(SecondaryColor3ivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4712(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4712(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4712(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4712(%rax), %r11
+	movq	4720(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3ivEXT), .-GL_PREFIX(SecondaryColor3ivEXT)
@@ -22200,7 +22245,7 @@ GL_PREFIX(SecondaryColor3ivEXT):
 GL_PREFIX(SecondaryColor3sEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4720(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22210,13 +22255,13 @@ GL_PREFIX(SecondaryColor3sEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4720(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4720(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22226,7 +22271,7 @@ GL_PREFIX(SecondaryColor3sEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4720(%rax), %r11
+	movq	4728(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3sEXT), .-GL_PREFIX(SecondaryColor3sEXT)
@@ -22237,25 +22282,25 @@ GL_PREFIX(SecondaryColor3sEXT):
 GL_PREFIX(SecondaryColor3svEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4728(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4728(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4728(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4728(%rax), %r11
+	movq	4736(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3svEXT), .-GL_PREFIX(SecondaryColor3svEXT)
@@ -22266,7 +22311,7 @@ GL_PREFIX(SecondaryColor3svEXT):
 GL_PREFIX(SecondaryColor3ubEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4736(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22276,13 +22321,13 @@ GL_PREFIX(SecondaryColor3ubEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4736(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4736(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22292,7 +22337,7 @@ GL_PREFIX(SecondaryColor3ubEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4736(%rax), %r11
+	movq	4744(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3ubEXT), .-GL_PREFIX(SecondaryColor3ubEXT)
@@ -22303,25 +22348,25 @@ GL_PREFIX(SecondaryColor3ubEXT):
 GL_PREFIX(SecondaryColor3ubvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4744(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4744(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4744(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4744(%rax), %r11
+	movq	4752(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3ubvEXT), .-GL_PREFIX(SecondaryColor3ubvEXT)
@@ -22332,7 +22377,7 @@ GL_PREFIX(SecondaryColor3ubvEXT):
 GL_PREFIX(SecondaryColor3uiEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4752(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22342,13 +22387,13 @@ GL_PREFIX(SecondaryColor3uiEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4752(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4752(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22358,7 +22403,7 @@ GL_PREFIX(SecondaryColor3uiEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4752(%rax), %r11
+	movq	4760(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3uiEXT), .-GL_PREFIX(SecondaryColor3uiEXT)
@@ -22369,25 +22414,25 @@ GL_PREFIX(SecondaryColor3uiEXT):
 GL_PREFIX(SecondaryColor3uivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4760(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4760(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4760(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4760(%rax), %r11
+	movq	4768(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3uivEXT), .-GL_PREFIX(SecondaryColor3uivEXT)
@@ -22398,7 +22443,7 @@ GL_PREFIX(SecondaryColor3uivEXT):
 GL_PREFIX(SecondaryColor3usEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4768(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22408,13 +22453,13 @@ GL_PREFIX(SecondaryColor3usEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4768(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4768(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22424,7 +22469,7 @@ GL_PREFIX(SecondaryColor3usEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4768(%rax), %r11
+	movq	4776(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3usEXT), .-GL_PREFIX(SecondaryColor3usEXT)
@@ -22435,25 +22480,25 @@ GL_PREFIX(SecondaryColor3usEXT):
 GL_PREFIX(SecondaryColor3usvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4776(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4776(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4776(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4776(%rax), %r11
+	movq	4784(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColor3usvEXT), .-GL_PREFIX(SecondaryColor3usvEXT)
@@ -22464,7 +22509,7 @@ GL_PREFIX(SecondaryColor3usvEXT):
 GL_PREFIX(SecondaryColorPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4784(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22478,13 +22523,13 @@ GL_PREFIX(SecondaryColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4784(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4784(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22498,7 +22543,7 @@ GL_PREFIX(SecondaryColorPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4784(%rax), %r11
+	movq	4792(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SecondaryColorPointerEXT), .-GL_PREFIX(SecondaryColorPointerEXT)
@@ -22509,7 +22554,7 @@ GL_PREFIX(SecondaryColorPointerEXT):
 GL_PREFIX(MultiDrawArraysEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4792(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22523,13 +22568,13 @@ GL_PREFIX(MultiDrawArraysEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4792(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4792(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22543,7 +22588,7 @@ GL_PREFIX(MultiDrawArraysEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4792(%rax), %r11
+	movq	4800(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(MultiDrawArraysEXT), .-GL_PREFIX(MultiDrawArraysEXT)
@@ -22554,7 +22599,7 @@ GL_PREFIX(MultiDrawArraysEXT):
 GL_PREFIX(MultiDrawElementsEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4800(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22568,13 +22613,13 @@ GL_PREFIX(MultiDrawElementsEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4800(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4800(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22588,7 +22633,7 @@ GL_PREFIX(MultiDrawElementsEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4800(%rax), %r11
+	movq	4808(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(MultiDrawElementsEXT), .-GL_PREFIX(MultiDrawElementsEXT)
@@ -22599,7 +22644,7 @@ GL_PREFIX(MultiDrawElementsEXT):
 GL_PREFIX(FogCoordPointerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4808(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22609,13 +22654,13 @@ GL_PREFIX(FogCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4808(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4808(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22625,7 +22670,7 @@ GL_PREFIX(FogCoordPointerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4808(%rax), %r11
+	movq	4816(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoordPointerEXT), .-GL_PREFIX(FogCoordPointerEXT)
@@ -22636,7 +22681,7 @@ GL_PREFIX(FogCoordPointerEXT):
 GL_PREFIX(FogCoorddEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4816(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$8, %rsp
@@ -22644,13 +22689,13 @@ GL_PREFIX(FogCoorddEXT):
 	call	_x86_64_get_dispatch@PLT
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4816(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4816(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$8, %rsp
@@ -22658,7 +22703,7 @@ GL_PREFIX(FogCoorddEXT):
 	call	_glapi_get_dispatch
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4816(%rax), %r11
+	movq	4824(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoorddEXT), .-GL_PREFIX(FogCoorddEXT)
@@ -22669,25 +22714,25 @@ GL_PREFIX(FogCoorddEXT):
 GL_PREFIX(FogCoorddvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4824(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4824(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4824(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4824(%rax), %r11
+	movq	4832(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoorddvEXT), .-GL_PREFIX(FogCoorddvEXT)
@@ -22698,7 +22743,7 @@ GL_PREFIX(FogCoorddvEXT):
 GL_PREFIX(FogCoordfEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4832(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$8, %rsp
@@ -22706,13 +22751,13 @@ GL_PREFIX(FogCoordfEXT):
 	call	_x86_64_get_dispatch@PLT
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4832(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4832(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$8, %rsp
@@ -22720,7 +22765,7 @@ GL_PREFIX(FogCoordfEXT):
 	call	_glapi_get_dispatch
 	movq	(%rsp), %xmm0
 	addq	$8, %rsp
-	movq	4832(%rax), %r11
+	movq	4840(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoordfEXT), .-GL_PREFIX(FogCoordfEXT)
@@ -22731,58 +22776,58 @@ GL_PREFIX(FogCoordfEXT):
 GL_PREFIX(FogCoordfvEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4840(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4840(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4840(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4840(%rax), %r11
+	movq	4848(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FogCoordfvEXT), .-GL_PREFIX(FogCoordfvEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_606)
-	.type	GL_PREFIX(_dispatch_stub_606), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_606))
-GL_PREFIX(_dispatch_stub_606):
+	.globl	GL_PREFIX(_dispatch_stub_607)
+	.type	GL_PREFIX(_dispatch_stub_607), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_607))
+GL_PREFIX(_dispatch_stub_607):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4848(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	4848(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4848(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	4848(%rax), %r11
+	movq	4856(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_606), .-GL_PREFIX(_dispatch_stub_606)
+	.size	GL_PREFIX(_dispatch_stub_607), .-GL_PREFIX(_dispatch_stub_607)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(BlendFuncSeparateEXT)
@@ -22790,7 +22835,7 @@ GL_PREFIX(_dispatch_stub_606):
 GL_PREFIX(BlendFuncSeparateEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4856(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22804,13 +22849,13 @@ GL_PREFIX(BlendFuncSeparateEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4856(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4856(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22824,7 +22869,7 @@ GL_PREFIX(BlendFuncSeparateEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4856(%rax), %r11
+	movq	4864(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BlendFuncSeparateEXT), .-GL_PREFIX(BlendFuncSeparateEXT)
@@ -22835,25 +22880,25 @@ GL_PREFIX(BlendFuncSeparateEXT):
 GL_PREFIX(FlushVertexArrayRangeNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4864(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	4864(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4864(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	4864(%rax), %r11
+	movq	4872(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FlushVertexArrayRangeNV), .-GL_PREFIX(FlushVertexArrayRangeNV)
@@ -22864,7 +22909,7 @@ GL_PREFIX(FlushVertexArrayRangeNV):
 GL_PREFIX(VertexArrayRangeNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4872(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22874,13 +22919,13 @@ GL_PREFIX(VertexArrayRangeNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4872(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4872(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22890,7 +22935,7 @@ GL_PREFIX(VertexArrayRangeNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4872(%rax), %r11
+	movq	4880(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexArrayRangeNV), .-GL_PREFIX(VertexArrayRangeNV)
@@ -22901,7 +22946,7 @@ GL_PREFIX(VertexArrayRangeNV):
 GL_PREFIX(CombinerInputNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4880(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22919,13 +22964,13 @@ GL_PREFIX(CombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4880(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4880(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22943,7 +22988,7 @@ GL_PREFIX(CombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4880(%rax), %r11
+	movq	4888(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerInputNV), .-GL_PREFIX(CombinerInputNV)
@@ -22954,7 +22999,7 @@ GL_PREFIX(CombinerInputNV):
 GL_PREFIX(CombinerOutputNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4888(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -22972,13 +23017,13 @@ GL_PREFIX(CombinerOutputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4888(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4888(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -22996,7 +23041,7 @@ GL_PREFIX(CombinerOutputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4888(%rax), %r11
+	movq	4896(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerOutputNV), .-GL_PREFIX(CombinerOutputNV)
@@ -23007,7 +23052,7 @@ GL_PREFIX(CombinerOutputNV):
 GL_PREFIX(CombinerParameterfNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4896(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23017,13 +23062,13 @@ GL_PREFIX(CombinerParameterfNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4896(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4896(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23033,7 +23078,7 @@ GL_PREFIX(CombinerParameterfNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	4896(%rax), %r11
+	movq	4904(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameterfNV), .-GL_PREFIX(CombinerParameterfNV)
@@ -23044,7 +23089,7 @@ GL_PREFIX(CombinerParameterfNV):
 GL_PREFIX(CombinerParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4904(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23054,13 +23099,13 @@ GL_PREFIX(CombinerParameterfvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4904(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4904(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23070,7 +23115,7 @@ GL_PREFIX(CombinerParameterfvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4904(%rax), %r11
+	movq	4912(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameterfvNV), .-GL_PREFIX(CombinerParameterfvNV)
@@ -23081,7 +23126,7 @@ GL_PREFIX(CombinerParameterfvNV):
 GL_PREFIX(CombinerParameteriNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4912(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23091,13 +23136,13 @@ GL_PREFIX(CombinerParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4912(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4912(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23107,7 +23152,7 @@ GL_PREFIX(CombinerParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4912(%rax), %r11
+	movq	4920(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameteriNV), .-GL_PREFIX(CombinerParameteriNV)
@@ -23118,7 +23163,7 @@ GL_PREFIX(CombinerParameteriNV):
 GL_PREFIX(CombinerParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4920(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23128,13 +23173,13 @@ GL_PREFIX(CombinerParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4920(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4920(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23144,7 +23189,7 @@ GL_PREFIX(CombinerParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	4920(%rax), %r11
+	movq	4928(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CombinerParameterivNV), .-GL_PREFIX(CombinerParameterivNV)
@@ -23155,7 +23200,7 @@ GL_PREFIX(CombinerParameterivNV):
 GL_PREFIX(FinalCombinerInputNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4928(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23169,13 +23214,13 @@ GL_PREFIX(FinalCombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4928(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4928(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23189,7 +23234,7 @@ GL_PREFIX(FinalCombinerInputNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4928(%rax), %r11
+	movq	4936(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FinalCombinerInputNV), .-GL_PREFIX(FinalCombinerInputNV)
@@ -23200,7 +23245,7 @@ GL_PREFIX(FinalCombinerInputNV):
 GL_PREFIX(GetCombinerInputParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4936(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23214,13 +23259,13 @@ GL_PREFIX(GetCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4936(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4936(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23234,7 +23279,7 @@ GL_PREFIX(GetCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4936(%rax), %r11
+	movq	4944(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerInputParameterfvNV), .-GL_PREFIX(GetCombinerInputParameterfvNV)
@@ -23245,7 +23290,7 @@ GL_PREFIX(GetCombinerInputParameterfvNV):
 GL_PREFIX(GetCombinerInputParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4944(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23259,13 +23304,13 @@ GL_PREFIX(GetCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4944(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4944(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23279,7 +23324,7 @@ GL_PREFIX(GetCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4944(%rax), %r11
+	movq	4952(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerInputParameterivNV), .-GL_PREFIX(GetCombinerInputParameterivNV)
@@ -23290,7 +23335,7 @@ GL_PREFIX(GetCombinerInputParameterivNV):
 GL_PREFIX(GetCombinerOutputParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4952(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23304,13 +23349,13 @@ GL_PREFIX(GetCombinerOutputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4952(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4952(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23324,7 +23369,7 @@ GL_PREFIX(GetCombinerOutputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4952(%rax), %r11
+	movq	4960(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerOutputParameterfvNV), .-GL_PREFIX(GetCombinerOutputParameterfvNV)
@@ -23335,7 +23380,7 @@ GL_PREFIX(GetCombinerOutputParameterfvNV):
 GL_PREFIX(GetCombinerOutputParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4960(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23349,13 +23394,13 @@ GL_PREFIX(GetCombinerOutputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4960(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4960(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23369,7 +23414,7 @@ GL_PREFIX(GetCombinerOutputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4960(%rax), %r11
+	movq	4968(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetCombinerOutputParameterivNV), .-GL_PREFIX(GetCombinerOutputParameterivNV)
@@ -23380,7 +23425,7 @@ GL_PREFIX(GetCombinerOutputParameterivNV):
 GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4968(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23390,13 +23435,13 @@ GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4968(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4968(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23406,7 +23451,7 @@ GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4968(%rax), %r11
+	movq	4976(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetFinalCombinerInputParameterfvNV), .-GL_PREFIX(GetFinalCombinerInputParameterfvNV)
@@ -23417,7 +23462,7 @@ GL_PREFIX(GetFinalCombinerInputParameterfvNV):
 GL_PREFIX(GetFinalCombinerInputParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4976(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23427,13 +23472,13 @@ GL_PREFIX(GetFinalCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4976(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4976(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23443,7 +23488,7 @@ GL_PREFIX(GetFinalCombinerInputParameterivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	4976(%rax), %r11
+	movq	4984(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetFinalCombinerInputParameterivNV), .-GL_PREFIX(GetFinalCombinerInputParameterivNV)
@@ -23454,25 +23499,25 @@ GL_PREFIX(GetFinalCombinerInputParameterivNV):
 GL_PREFIX(ResizeBuffersMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4984(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	4984(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4984(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	4984(%rax), %r11
+	movq	4992(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ResizeBuffersMESA), .-GL_PREFIX(ResizeBuffersMESA)
@@ -23483,7 +23528,7 @@ GL_PREFIX(ResizeBuffersMESA):
 GL_PREFIX(WindowPos2dMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	4992(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23493,13 +23538,13 @@ GL_PREFIX(WindowPos2dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4992(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	4992(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23509,7 +23554,7 @@ GL_PREFIX(WindowPos2dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	4992(%rax), %r11
+	movq	5000(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2dMESA), .-GL_PREFIX(WindowPos2dMESA)
@@ -23520,25 +23565,25 @@ GL_PREFIX(WindowPos2dMESA):
 GL_PREFIX(WindowPos2dvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5000(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5000(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5000(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5000(%rax), %r11
+	movq	5008(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2dvMESA), .-GL_PREFIX(WindowPos2dvMESA)
@@ -23549,7 +23594,7 @@ GL_PREFIX(WindowPos2dvMESA):
 GL_PREFIX(WindowPos2fMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5008(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23559,13 +23604,13 @@ GL_PREFIX(WindowPos2fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5008(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5008(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23575,7 +23620,7 @@ GL_PREFIX(WindowPos2fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5008(%rax), %r11
+	movq	5016(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2fMESA), .-GL_PREFIX(WindowPos2fMESA)
@@ -23586,25 +23631,25 @@ GL_PREFIX(WindowPos2fMESA):
 GL_PREFIX(WindowPos2fvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5016(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5016(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5016(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5016(%rax), %r11
+	movq	5024(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2fvMESA), .-GL_PREFIX(WindowPos2fvMESA)
@@ -23615,7 +23660,7 @@ GL_PREFIX(WindowPos2fvMESA):
 GL_PREFIX(WindowPos2iMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5024(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23625,13 +23670,13 @@ GL_PREFIX(WindowPos2iMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5024(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5024(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23641,7 +23686,7 @@ GL_PREFIX(WindowPos2iMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5024(%rax), %r11
+	movq	5032(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2iMESA), .-GL_PREFIX(WindowPos2iMESA)
@@ -23652,25 +23697,25 @@ GL_PREFIX(WindowPos2iMESA):
 GL_PREFIX(WindowPos2ivMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5032(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5032(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5032(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5032(%rax), %r11
+	movq	5040(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2ivMESA), .-GL_PREFIX(WindowPos2ivMESA)
@@ -23681,7 +23726,7 @@ GL_PREFIX(WindowPos2ivMESA):
 GL_PREFIX(WindowPos2sMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5040(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23691,13 +23736,13 @@ GL_PREFIX(WindowPos2sMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5040(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5040(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23707,7 +23752,7 @@ GL_PREFIX(WindowPos2sMESA):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5040(%rax), %r11
+	movq	5048(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2sMESA), .-GL_PREFIX(WindowPos2sMESA)
@@ -23718,25 +23763,25 @@ GL_PREFIX(WindowPos2sMESA):
 GL_PREFIX(WindowPos2svMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5048(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5048(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5048(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5048(%rax), %r11
+	movq	5056(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos2svMESA), .-GL_PREFIX(WindowPos2svMESA)
@@ -23747,7 +23792,7 @@ GL_PREFIX(WindowPos2svMESA):
 GL_PREFIX(WindowPos3dMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5056(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23759,13 +23804,13 @@ GL_PREFIX(WindowPos3dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5056(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5056(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23777,7 +23822,7 @@ GL_PREFIX(WindowPos3dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5056(%rax), %r11
+	movq	5064(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3dMESA), .-GL_PREFIX(WindowPos3dMESA)
@@ -23788,25 +23833,25 @@ GL_PREFIX(WindowPos3dMESA):
 GL_PREFIX(WindowPos3dvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5064(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5064(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5064(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5064(%rax), %r11
+	movq	5072(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3dvMESA), .-GL_PREFIX(WindowPos3dvMESA)
@@ -23817,7 +23862,7 @@ GL_PREFIX(WindowPos3dvMESA):
 GL_PREFIX(WindowPos3fMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5072(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -23829,13 +23874,13 @@ GL_PREFIX(WindowPos3fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5072(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5072(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -23847,7 +23892,7 @@ GL_PREFIX(WindowPos3fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$24, %rsp
-	movq	5072(%rax), %r11
+	movq	5080(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3fMESA), .-GL_PREFIX(WindowPos3fMESA)
@@ -23858,25 +23903,25 @@ GL_PREFIX(WindowPos3fMESA):
 GL_PREFIX(WindowPos3fvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5080(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5080(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5080(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5080(%rax), %r11
+	movq	5088(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3fvMESA), .-GL_PREFIX(WindowPos3fvMESA)
@@ -23887,7 +23932,7 @@ GL_PREFIX(WindowPos3fvMESA):
 GL_PREFIX(WindowPos3iMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5088(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23897,13 +23942,13 @@ GL_PREFIX(WindowPos3iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5088(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5088(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23913,7 +23958,7 @@ GL_PREFIX(WindowPos3iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5088(%rax), %r11
+	movq	5096(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3iMESA), .-GL_PREFIX(WindowPos3iMESA)
@@ -23924,25 +23969,25 @@ GL_PREFIX(WindowPos3iMESA):
 GL_PREFIX(WindowPos3ivMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5096(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5096(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5096(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5096(%rax), %r11
+	movq	5104(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3ivMESA), .-GL_PREFIX(WindowPos3ivMESA)
@@ -23953,7 +23998,7 @@ GL_PREFIX(WindowPos3ivMESA):
 GL_PREFIX(WindowPos3sMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5104(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -23963,13 +24008,13 @@ GL_PREFIX(WindowPos3sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5104(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5104(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -23979,7 +24024,7 @@ GL_PREFIX(WindowPos3sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5104(%rax), %r11
+	movq	5112(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3sMESA), .-GL_PREFIX(WindowPos3sMESA)
@@ -23990,25 +24035,25 @@ GL_PREFIX(WindowPos3sMESA):
 GL_PREFIX(WindowPos3svMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5112(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5112(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5112(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5112(%rax), %r11
+	movq	5120(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos3svMESA), .-GL_PREFIX(WindowPos3svMESA)
@@ -24019,7 +24064,7 @@ GL_PREFIX(WindowPos3svMESA):
 GL_PREFIX(WindowPos4dMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5120(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -24033,13 +24078,13 @@ GL_PREFIX(WindowPos4dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5120(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5120(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -24053,7 +24098,7 @@ GL_PREFIX(WindowPos4dMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5120(%rax), %r11
+	movq	5128(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4dMESA), .-GL_PREFIX(WindowPos4dMESA)
@@ -24064,25 +24109,25 @@ GL_PREFIX(WindowPos4dMESA):
 GL_PREFIX(WindowPos4dvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5128(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5128(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5128(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5128(%rax), %r11
+	movq	5136(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4dvMESA), .-GL_PREFIX(WindowPos4dvMESA)
@@ -24093,7 +24138,7 @@ GL_PREFIX(WindowPos4dvMESA):
 GL_PREFIX(WindowPos4fMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5136(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -24107,13 +24152,13 @@ GL_PREFIX(WindowPos4fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5136(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5136(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -24127,7 +24172,7 @@ GL_PREFIX(WindowPos4fMESA):
 	movq	8(%rsp), %xmm1
 	movq	(%rsp), %xmm0
 	addq	$40, %rsp
-	movq	5136(%rax), %r11
+	movq	5144(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4fMESA), .-GL_PREFIX(WindowPos4fMESA)
@@ -24138,25 +24183,25 @@ GL_PREFIX(WindowPos4fMESA):
 GL_PREFIX(WindowPos4fvMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5144(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5144(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5144(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5144(%rax), %r11
+	movq	5152(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4fvMESA), .-GL_PREFIX(WindowPos4fvMESA)
@@ -24167,7 +24212,7 @@ GL_PREFIX(WindowPos4fvMESA):
 GL_PREFIX(WindowPos4iMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5152(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24181,13 +24226,13 @@ GL_PREFIX(WindowPos4iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5152(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5152(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24201,7 +24246,7 @@ GL_PREFIX(WindowPos4iMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5152(%rax), %r11
+	movq	5160(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4iMESA), .-GL_PREFIX(WindowPos4iMESA)
@@ -24212,25 +24257,25 @@ GL_PREFIX(WindowPos4iMESA):
 GL_PREFIX(WindowPos4ivMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5160(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5160(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5160(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5160(%rax), %r11
+	movq	5168(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4ivMESA), .-GL_PREFIX(WindowPos4ivMESA)
@@ -24241,7 +24286,7 @@ GL_PREFIX(WindowPos4ivMESA):
 GL_PREFIX(WindowPos4sMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5168(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24255,13 +24300,13 @@ GL_PREFIX(WindowPos4sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5168(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5168(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24275,7 +24320,7 @@ GL_PREFIX(WindowPos4sMESA):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5168(%rax), %r11
+	movq	5176(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(WindowPos4sMESA), .-GL_PREFIX(WindowPos4sMESA)
@@ -24286,49 +24331,11 @@ GL_PREFIX(WindowPos4sMESA):
 GL_PREFIX(WindowPos4svMESA):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5176(%rax), %r11
-	jmp	*%r11
-#elif defined(PTHREADS)
-	pushq	%rdi
-	call	_x86_64_get_dispatch@PLT
-	popq	%rdi
-	movq	5176(%rax), %r11
-	jmp	*%r11
-#else
-	movq	_glapi_Dispatch(%rip), %rax
-	testq	%rax, %rax
-	je	1f
-	movq	5176(%rax), %r11
-	jmp	*%r11
-1:
-	pushq	%rdi
-	call	_glapi_get_dispatch
-	popq	%rdi
-	movq	5176(%rax), %r11
-	jmp	*%r11
-#endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(WindowPos4svMESA), .-GL_PREFIX(WindowPos4svMESA)
-
-	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_648)
-	.type	GL_PREFIX(_dispatch_stub_648), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_648))
-GL_PREFIX(_dispatch_stub_648):
-#if defined(GLX_USE_TLS)
-	call	_x86_64_get_dispatch@PLT
 	movq	5184(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
-	pushq	%rsi
-	pushq	%rdx
-	pushq	%rcx
-	pushq	%r8
 	call	_x86_64_get_dispatch@PLT
-	popq	%r8
-	popq	%rcx
-	popq	%rdx
-	popq	%rsi
 	popq	%rdi
 	movq	5184(%rax), %r11
 	jmp	*%r11
@@ -24340,20 +24347,12 @@ GL_PREFIX(_dispatch_stub_648):
 	jmp	*%r11
 1:
 	pushq	%rdi
-	pushq	%rsi
-	pushq	%rdx
-	pushq	%rcx
-	pushq	%r8
 	call	_glapi_get_dispatch
-	popq	%r8
-	popq	%rcx
-	popq	%rdx
-	popq	%rsi
 	popq	%rdi
 	movq	5184(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_648), .-GL_PREFIX(_dispatch_stub_648)
+	.size	GL_PREFIX(WindowPos4svMESA), .-GL_PREFIX(WindowPos4svMESA)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(_dispatch_stub_649)
@@ -24370,11 +24369,7 @@ GL_PREFIX(_dispatch_stub_649):
 	pushq	%rdx
 	pushq	%rcx
 	pushq	%r8
-	pushq	%r9
-	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
-	popq	%rbp
-	popq	%r9
 	popq	%r8
 	popq	%rcx
 	popq	%rdx
@@ -24394,11 +24389,7 @@ GL_PREFIX(_dispatch_stub_649):
 	pushq	%rdx
 	pushq	%rcx
 	pushq	%r8
-	pushq	%r9
-	pushq	%rbp
 	call	_glapi_get_dispatch
-	popq	%rbp
-	popq	%r9
 	popq	%r8
 	popq	%rcx
 	popq	%rdx
@@ -24421,9 +24412,17 @@ GL_PREFIX(_dispatch_stub_650):
 #elif defined(PTHREADS)
 	pushq	%rdi
 	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%r8
+	pushq	%r9
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
+	popq	%r9
+	popq	%r8
+	popq	%rcx
+	popq	%rdx
 	popq	%rsi
 	popq	%rdi
 	movq	5200(%rax), %r11
@@ -24437,9 +24436,17 @@ GL_PREFIX(_dispatch_stub_650):
 1:
 	pushq	%rdi
 	pushq	%rsi
+	pushq	%rdx
+	pushq	%rcx
+	pushq	%r8
+	pushq	%r9
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
+	popq	%r9
+	popq	%r8
+	popq	%rcx
+	popq	%rdx
 	popq	%rsi
 	popq	%rdi
 	movq	5200(%rax), %r11
@@ -24458,7 +24465,11 @@ GL_PREFIX(_dispatch_stub_651):
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
+	popq	%rbp
+	popq	%rsi
 	popq	%rdi
 	movq	5208(%rax), %r11
 	jmp	*%r11
@@ -24470,7 +24481,11 @@ GL_PREFIX(_dispatch_stub_651):
 	jmp	*%r11
 1:
 	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
 	call	_glapi_get_dispatch
+	popq	%rbp
+	popq	%rsi
 	popq	%rdi
 	movq	5208(%rax), %r11
 	jmp	*%r11
@@ -24488,11 +24503,7 @@ GL_PREFIX(_dispatch_stub_652):
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
-	pushq	%rsi
-	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
-	popq	%rbp
-	popq	%rsi
 	popq	%rdi
 	movq	5216(%rax), %r11
 	jmp	*%r11
@@ -24504,11 +24515,7 @@ GL_PREFIX(_dispatch_stub_652):
 	jmp	*%r11
 1:
 	pushq	%rdi
-	pushq	%rsi
-	pushq	%rbp
 	call	_glapi_get_dispatch
-	popq	%rbp
-	popq	%rsi
 	popq	%rdi
 	movq	5216(%rax), %r11
 	jmp	*%r11
@@ -24527,9 +24534,9 @@ GL_PREFIX(_dispatch_stub_653):
 #elif defined(PTHREADS)
 	pushq	%rdi
 	pushq	%rsi
-	pushq	%rdx
+	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
-	popq	%rdx
+	popq	%rbp
 	popq	%rsi
 	popq	%rdi
 	movq	5224(%rax), %r11
@@ -24543,9 +24550,9 @@ GL_PREFIX(_dispatch_stub_653):
 1:
 	pushq	%rdi
 	pushq	%rsi
-	pushq	%rdx
+	pushq	%rbp
 	call	_glapi_get_dispatch
-	popq	%rdx
+	popq	%rbp
 	popq	%rsi
 	popq	%rdi
 	movq	5224(%rax), %r11
@@ -24564,7 +24571,11 @@ GL_PREFIX(_dispatch_stub_654):
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
 	call	_x86_64_get_dispatch@PLT
+	popq	%rdx
+	popq	%rsi
 	popq	%rdi
 	movq	5232(%rax), %r11
 	jmp	*%r11
@@ -24576,7 +24587,11 @@ GL_PREFIX(_dispatch_stub_654):
 	jmp	*%r11
 1:
 	pushq	%rdi
+	pushq	%rsi
+	pushq	%rdx
 	call	_glapi_get_dispatch
+	popq	%rdx
+	popq	%rsi
 	popq	%rdi
 	movq	5232(%rax), %r11
 	jmp	*%r11
@@ -24594,11 +24609,7 @@ GL_PREFIX(_dispatch_stub_655):
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
-	pushq	%rsi
-	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
-	popq	%rbp
-	popq	%rsi
 	popq	%rdi
 	movq	5240(%rax), %r11
 	jmp	*%r11
@@ -24610,11 +24621,7 @@ GL_PREFIX(_dispatch_stub_655):
 	jmp	*%r11
 1:
 	pushq	%rdi
-	pushq	%rsi
-	pushq	%rbp
 	call	_glapi_get_dispatch
-	popq	%rbp
-	popq	%rsi
 	popq	%rdi
 	movq	5240(%rax), %r11
 	jmp	*%r11
@@ -24632,7 +24639,11 @@ GL_PREFIX(_dispatch_stub_656):
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
+	popq	%rbp
+	popq	%rsi
 	popq	%rdi
 	movq	5248(%rax), %r11
 	jmp	*%r11
@@ -24644,7 +24655,11 @@ GL_PREFIX(_dispatch_stub_656):
 	jmp	*%r11
 1:
 	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
 	call	_glapi_get_dispatch
+	popq	%rbp
+	popq	%rsi
 	popq	%rdi
 	movq	5248(%rax), %r11
 	jmp	*%r11
@@ -24652,12 +24667,42 @@ GL_PREFIX(_dispatch_stub_656):
 	.size	GL_PREFIX(_dispatch_stub_656), .-GL_PREFIX(_dispatch_stub_656)
 
 	.p2align	4,,15
+	.globl	GL_PREFIX(_dispatch_stub_657)
+	.type	GL_PREFIX(_dispatch_stub_657), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_657))
+GL_PREFIX(_dispatch_stub_657):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	5256(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdi
+	movq	5256(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	5256(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	call	_glapi_get_dispatch
+	popq	%rdi
+	movq	5256(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(_dispatch_stub_657), .-GL_PREFIX(_dispatch_stub_657)
+
+	.p2align	4,,15
 	.globl	GL_PREFIX(AreProgramsResidentNV)
 	.type	GL_PREFIX(AreProgramsResidentNV), @function
 GL_PREFIX(AreProgramsResidentNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5256(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24667,13 +24712,13 @@ GL_PREFIX(AreProgramsResidentNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5256(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5256(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24683,7 +24728,7 @@ GL_PREFIX(AreProgramsResidentNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5256(%rax), %r11
+	movq	5264(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AreProgramsResidentNV), .-GL_PREFIX(AreProgramsResidentNV)
@@ -24694,7 +24739,7 @@ GL_PREFIX(AreProgramsResidentNV):
 GL_PREFIX(BindProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5264(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24704,13 +24749,13 @@ GL_PREFIX(BindProgramNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5264(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5264(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24720,7 +24765,7 @@ GL_PREFIX(BindProgramNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5264(%rax), %r11
+	movq	5272(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindProgramNV), .-GL_PREFIX(BindProgramNV)
@@ -24731,7 +24776,7 @@ GL_PREFIX(BindProgramNV):
 GL_PREFIX(DeleteProgramsNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5272(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24741,13 +24786,13 @@ GL_PREFIX(DeleteProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5272(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5272(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24757,7 +24802,7 @@ GL_PREFIX(DeleteProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5272(%rax), %r11
+	movq	5280(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteProgramsNV), .-GL_PREFIX(DeleteProgramsNV)
@@ -24768,7 +24813,7 @@ GL_PREFIX(DeleteProgramsNV):
 GL_PREFIX(ExecuteProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5280(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24778,13 +24823,13 @@ GL_PREFIX(ExecuteProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5280(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5280(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24794,7 +24839,7 @@ GL_PREFIX(ExecuteProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5280(%rax), %r11
+	movq	5288(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ExecuteProgramNV), .-GL_PREFIX(ExecuteProgramNV)
@@ -24805,7 +24850,7 @@ GL_PREFIX(ExecuteProgramNV):
 GL_PREFIX(GenProgramsNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5288(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24815,13 +24860,13 @@ GL_PREFIX(GenProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5288(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5288(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24831,7 +24876,7 @@ GL_PREFIX(GenProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5288(%rax), %r11
+	movq	5296(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenProgramsNV), .-GL_PREFIX(GenProgramsNV)
@@ -24842,7 +24887,7 @@ GL_PREFIX(GenProgramsNV):
 GL_PREFIX(GetProgramParameterdvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5296(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24856,13 +24901,13 @@ GL_PREFIX(GetProgramParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5296(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5296(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24876,7 +24921,7 @@ GL_PREFIX(GetProgramParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5296(%rax), %r11
+	movq	5304(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramParameterdvNV), .-GL_PREFIX(GetProgramParameterdvNV)
@@ -24887,7 +24932,7 @@ GL_PREFIX(GetProgramParameterdvNV):
 GL_PREFIX(GetProgramParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5304(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24901,13 +24946,13 @@ GL_PREFIX(GetProgramParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5304(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5304(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24921,7 +24966,7 @@ GL_PREFIX(GetProgramParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5304(%rax), %r11
+	movq	5312(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramParameterfvNV), .-GL_PREFIX(GetProgramParameterfvNV)
@@ -24932,7 +24977,7 @@ GL_PREFIX(GetProgramParameterfvNV):
 GL_PREFIX(GetProgramStringNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5312(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24942,13 +24987,13 @@ GL_PREFIX(GetProgramStringNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5312(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5312(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24958,7 +25003,7 @@ GL_PREFIX(GetProgramStringNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5312(%rax), %r11
+	movq	5320(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramStringNV), .-GL_PREFIX(GetProgramStringNV)
@@ -24969,7 +25014,7 @@ GL_PREFIX(GetProgramStringNV):
 GL_PREFIX(GetProgramivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5320(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -24979,13 +25024,13 @@ GL_PREFIX(GetProgramivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5320(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5320(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -24995,7 +25040,7 @@ GL_PREFIX(GetProgramivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5320(%rax), %r11
+	movq	5328(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramivNV), .-GL_PREFIX(GetProgramivNV)
@@ -25006,7 +25051,7 @@ GL_PREFIX(GetProgramivNV):
 GL_PREFIX(GetTrackMatrixivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5328(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25020,13 +25065,13 @@ GL_PREFIX(GetTrackMatrixivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5328(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5328(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25040,7 +25085,7 @@ GL_PREFIX(GetTrackMatrixivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5328(%rax), %r11
+	movq	5336(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetTrackMatrixivNV), .-GL_PREFIX(GetTrackMatrixivNV)
@@ -25051,7 +25096,7 @@ GL_PREFIX(GetTrackMatrixivNV):
 GL_PREFIX(GetVertexAttribPointervNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5336(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25061,13 +25106,13 @@ GL_PREFIX(GetVertexAttribPointervNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5336(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5336(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25077,7 +25122,7 @@ GL_PREFIX(GetVertexAttribPointervNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5336(%rax), %r11
+	movq	5344(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribPointervNV), .-GL_PREFIX(GetVertexAttribPointervNV)
@@ -25088,7 +25133,7 @@ GL_PREFIX(GetVertexAttribPointervNV):
 GL_PREFIX(GetVertexAttribdvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5344(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25098,13 +25143,13 @@ GL_PREFIX(GetVertexAttribdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5344(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5344(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25114,7 +25159,7 @@ GL_PREFIX(GetVertexAttribdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5344(%rax), %r11
+	movq	5352(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribdvNV), .-GL_PREFIX(GetVertexAttribdvNV)
@@ -25125,7 +25170,7 @@ GL_PREFIX(GetVertexAttribdvNV):
 GL_PREFIX(GetVertexAttribfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5352(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25135,13 +25180,13 @@ GL_PREFIX(GetVertexAttribfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5352(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5352(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25151,7 +25196,7 @@ GL_PREFIX(GetVertexAttribfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5352(%rax), %r11
+	movq	5360(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribfvNV), .-GL_PREFIX(GetVertexAttribfvNV)
@@ -25162,7 +25207,7 @@ GL_PREFIX(GetVertexAttribfvNV):
 GL_PREFIX(GetVertexAttribivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5360(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25172,13 +25217,13 @@ GL_PREFIX(GetVertexAttribivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5360(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5360(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25188,7 +25233,7 @@ GL_PREFIX(GetVertexAttribivNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5360(%rax), %r11
+	movq	5368(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetVertexAttribivNV), .-GL_PREFIX(GetVertexAttribivNV)
@@ -25199,25 +25244,25 @@ GL_PREFIX(GetVertexAttribivNV):
 GL_PREFIX(IsProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5368(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5368(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5368(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5368(%rax), %r11
+	movq	5376(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IsProgramNV), .-GL_PREFIX(IsProgramNV)
@@ -25228,7 +25273,7 @@ GL_PREFIX(IsProgramNV):
 GL_PREFIX(LoadProgramNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5376(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25242,13 +25287,13 @@ GL_PREFIX(LoadProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5376(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5376(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25262,7 +25307,7 @@ GL_PREFIX(LoadProgramNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5376(%rax), %r11
+	movq	5384(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(LoadProgramNV), .-GL_PREFIX(LoadProgramNV)
@@ -25273,7 +25318,7 @@ GL_PREFIX(LoadProgramNV):
 GL_PREFIX(ProgramParameters4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5384(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25287,13 +25332,13 @@ GL_PREFIX(ProgramParameters4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5384(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5384(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25307,7 +25352,7 @@ GL_PREFIX(ProgramParameters4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5384(%rax), %r11
+	movq	5392(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramParameters4dvNV), .-GL_PREFIX(ProgramParameters4dvNV)
@@ -25318,7 +25363,7 @@ GL_PREFIX(ProgramParameters4dvNV):
 GL_PREFIX(ProgramParameters4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5392(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25332,13 +25377,13 @@ GL_PREFIX(ProgramParameters4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5392(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5392(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25352,7 +25397,7 @@ GL_PREFIX(ProgramParameters4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5392(%rax), %r11
+	movq	5400(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramParameters4fvNV), .-GL_PREFIX(ProgramParameters4fvNV)
@@ -25363,7 +25408,7 @@ GL_PREFIX(ProgramParameters4fvNV):
 GL_PREFIX(RequestResidentProgramsNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5400(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25373,13 +25418,13 @@ GL_PREFIX(RequestResidentProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5400(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5400(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25389,7 +25434,7 @@ GL_PREFIX(RequestResidentProgramsNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5400(%rax), %r11
+	movq	5408(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(RequestResidentProgramsNV), .-GL_PREFIX(RequestResidentProgramsNV)
@@ -25400,7 +25445,7 @@ GL_PREFIX(RequestResidentProgramsNV):
 GL_PREFIX(TrackMatrixNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5408(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25414,13 +25459,13 @@ GL_PREFIX(TrackMatrixNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5408(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5408(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25434,7 +25479,7 @@ GL_PREFIX(TrackMatrixNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5408(%rax), %r11
+	movq	5416(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(TrackMatrixNV), .-GL_PREFIX(TrackMatrixNV)
@@ -25445,7 +25490,7 @@ GL_PREFIX(TrackMatrixNV):
 GL_PREFIX(VertexAttrib1dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5416(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25455,13 +25500,13 @@ GL_PREFIX(VertexAttrib1dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5416(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5416(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25471,7 +25516,7 @@ GL_PREFIX(VertexAttrib1dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5416(%rax), %r11
+	movq	5424(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1dNV), .-GL_PREFIX(VertexAttrib1dNV)
@@ -25482,7 +25527,7 @@ GL_PREFIX(VertexAttrib1dNV):
 GL_PREFIX(VertexAttrib1dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5424(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25492,13 +25537,13 @@ GL_PREFIX(VertexAttrib1dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5424(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5424(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25508,7 +25553,7 @@ GL_PREFIX(VertexAttrib1dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5424(%rax), %r11
+	movq	5432(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1dvNV), .-GL_PREFIX(VertexAttrib1dvNV)
@@ -25519,7 +25564,7 @@ GL_PREFIX(VertexAttrib1dvNV):
 GL_PREFIX(VertexAttrib1fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5432(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25529,13 +25574,13 @@ GL_PREFIX(VertexAttrib1fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5432(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5432(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25545,7 +25590,7 @@ GL_PREFIX(VertexAttrib1fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5432(%rax), %r11
+	movq	5440(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1fNV), .-GL_PREFIX(VertexAttrib1fNV)
@@ -25556,7 +25601,7 @@ GL_PREFIX(VertexAttrib1fNV):
 GL_PREFIX(VertexAttrib1fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5440(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25566,13 +25611,13 @@ GL_PREFIX(VertexAttrib1fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5440(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5440(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25582,7 +25627,7 @@ GL_PREFIX(VertexAttrib1fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5440(%rax), %r11
+	movq	5448(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1fvNV), .-GL_PREFIX(VertexAttrib1fvNV)
@@ -25593,7 +25638,7 @@ GL_PREFIX(VertexAttrib1fvNV):
 GL_PREFIX(VertexAttrib1sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5448(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25603,13 +25648,13 @@ GL_PREFIX(VertexAttrib1sNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5448(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5448(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25619,7 +25664,7 @@ GL_PREFIX(VertexAttrib1sNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5448(%rax), %r11
+	movq	5456(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1sNV), .-GL_PREFIX(VertexAttrib1sNV)
@@ -25630,7 +25675,7 @@ GL_PREFIX(VertexAttrib1sNV):
 GL_PREFIX(VertexAttrib1svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5456(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25640,13 +25685,13 @@ GL_PREFIX(VertexAttrib1svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5456(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5456(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25656,7 +25701,7 @@ GL_PREFIX(VertexAttrib1svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5456(%rax), %r11
+	movq	5464(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib1svNV), .-GL_PREFIX(VertexAttrib1svNV)
@@ -25667,7 +25712,7 @@ GL_PREFIX(VertexAttrib1svNV):
 GL_PREFIX(VertexAttrib2dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5464(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25679,13 +25724,13 @@ GL_PREFIX(VertexAttrib2dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5464(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5464(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25697,7 +25742,7 @@ GL_PREFIX(VertexAttrib2dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5464(%rax), %r11
+	movq	5472(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2dNV), .-GL_PREFIX(VertexAttrib2dNV)
@@ -25708,7 +25753,7 @@ GL_PREFIX(VertexAttrib2dNV):
 GL_PREFIX(VertexAttrib2dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5472(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25718,13 +25763,13 @@ GL_PREFIX(VertexAttrib2dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5472(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5472(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25734,7 +25779,7 @@ GL_PREFIX(VertexAttrib2dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5472(%rax), %r11
+	movq	5480(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2dvNV), .-GL_PREFIX(VertexAttrib2dvNV)
@@ -25745,7 +25790,7 @@ GL_PREFIX(VertexAttrib2dvNV):
 GL_PREFIX(VertexAttrib2fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5480(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$24, %rsp
@@ -25757,13 +25802,13 @@ GL_PREFIX(VertexAttrib2fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5480(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5480(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$24, %rsp
@@ -25775,7 +25820,7 @@ GL_PREFIX(VertexAttrib2fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$24, %rsp
-	movq	5480(%rax), %r11
+	movq	5488(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2fNV), .-GL_PREFIX(VertexAttrib2fNV)
@@ -25786,7 +25831,7 @@ GL_PREFIX(VertexAttrib2fNV):
 GL_PREFIX(VertexAttrib2fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5488(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25796,13 +25841,13 @@ GL_PREFIX(VertexAttrib2fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5488(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5488(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25812,7 +25857,7 @@ GL_PREFIX(VertexAttrib2fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5488(%rax), %r11
+	movq	5496(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2fvNV), .-GL_PREFIX(VertexAttrib2fvNV)
@@ -25823,7 +25868,7 @@ GL_PREFIX(VertexAttrib2fvNV):
 GL_PREFIX(VertexAttrib2sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5496(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25833,13 +25878,13 @@ GL_PREFIX(VertexAttrib2sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5496(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5496(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25849,7 +25894,7 @@ GL_PREFIX(VertexAttrib2sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5496(%rax), %r11
+	movq	5504(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2sNV), .-GL_PREFIX(VertexAttrib2sNV)
@@ -25860,7 +25905,7 @@ GL_PREFIX(VertexAttrib2sNV):
 GL_PREFIX(VertexAttrib2svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5504(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25870,13 +25915,13 @@ GL_PREFIX(VertexAttrib2svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5504(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5504(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25886,7 +25931,7 @@ GL_PREFIX(VertexAttrib2svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5504(%rax), %r11
+	movq	5512(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib2svNV), .-GL_PREFIX(VertexAttrib2svNV)
@@ -25897,7 +25942,7 @@ GL_PREFIX(VertexAttrib2svNV):
 GL_PREFIX(VertexAttrib3dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5512(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -25911,13 +25956,13 @@ GL_PREFIX(VertexAttrib3dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5512(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5512(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -25931,7 +25976,7 @@ GL_PREFIX(VertexAttrib3dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5512(%rax), %r11
+	movq	5520(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3dNV), .-GL_PREFIX(VertexAttrib3dNV)
@@ -25942,7 +25987,7 @@ GL_PREFIX(VertexAttrib3dNV):
 GL_PREFIX(VertexAttrib3dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5520(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -25952,13 +25997,13 @@ GL_PREFIX(VertexAttrib3dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5520(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5520(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -25968,7 +26013,7 @@ GL_PREFIX(VertexAttrib3dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5520(%rax), %r11
+	movq	5528(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3dvNV), .-GL_PREFIX(VertexAttrib3dvNV)
@@ -25979,7 +26024,7 @@ GL_PREFIX(VertexAttrib3dvNV):
 GL_PREFIX(VertexAttrib3fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5528(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -25993,13 +26038,13 @@ GL_PREFIX(VertexAttrib3fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5528(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5528(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -26013,7 +26058,7 @@ GL_PREFIX(VertexAttrib3fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5528(%rax), %r11
+	movq	5536(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3fNV), .-GL_PREFIX(VertexAttrib3fNV)
@@ -26024,7 +26069,7 @@ GL_PREFIX(VertexAttrib3fNV):
 GL_PREFIX(VertexAttrib3fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5536(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26034,13 +26079,13 @@ GL_PREFIX(VertexAttrib3fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5536(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5536(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26050,7 +26095,7 @@ GL_PREFIX(VertexAttrib3fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5536(%rax), %r11
+	movq	5544(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3fvNV), .-GL_PREFIX(VertexAttrib3fvNV)
@@ -26061,7 +26106,7 @@ GL_PREFIX(VertexAttrib3fvNV):
 GL_PREFIX(VertexAttrib3sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5544(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26075,13 +26120,13 @@ GL_PREFIX(VertexAttrib3sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5544(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5544(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26095,7 +26140,7 @@ GL_PREFIX(VertexAttrib3sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5544(%rax), %r11
+	movq	5552(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3sNV), .-GL_PREFIX(VertexAttrib3sNV)
@@ -26106,7 +26151,7 @@ GL_PREFIX(VertexAttrib3sNV):
 GL_PREFIX(VertexAttrib3svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5552(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26116,13 +26161,13 @@ GL_PREFIX(VertexAttrib3svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5552(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5552(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26132,7 +26177,7 @@ GL_PREFIX(VertexAttrib3svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5552(%rax), %r11
+	movq	5560(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib3svNV), .-GL_PREFIX(VertexAttrib3svNV)
@@ -26143,7 +26188,7 @@ GL_PREFIX(VertexAttrib3svNV):
 GL_PREFIX(VertexAttrib4dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5560(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -26159,13 +26204,13 @@ GL_PREFIX(VertexAttrib4dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5560(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5560(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -26181,7 +26226,7 @@ GL_PREFIX(VertexAttrib4dNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5560(%rax), %r11
+	movq	5568(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4dNV), .-GL_PREFIX(VertexAttrib4dNV)
@@ -26192,7 +26237,7 @@ GL_PREFIX(VertexAttrib4dNV):
 GL_PREFIX(VertexAttrib4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5568(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26202,13 +26247,13 @@ GL_PREFIX(VertexAttrib4dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5568(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5568(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26218,7 +26263,7 @@ GL_PREFIX(VertexAttrib4dvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5568(%rax), %r11
+	movq	5576(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4dvNV), .-GL_PREFIX(VertexAttrib4dvNV)
@@ -26229,7 +26274,7 @@ GL_PREFIX(VertexAttrib4dvNV):
 GL_PREFIX(VertexAttrib4fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5576(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$40, %rsp
@@ -26245,13 +26290,13 @@ GL_PREFIX(VertexAttrib4fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5576(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5576(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$40, %rsp
@@ -26267,7 +26312,7 @@ GL_PREFIX(VertexAttrib4fNV):
 	movq	8(%rsp), %xmm0
 	movq	(%rsp), %rdi
 	addq	$40, %rsp
-	movq	5576(%rax), %r11
+	movq	5584(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4fNV), .-GL_PREFIX(VertexAttrib4fNV)
@@ -26278,7 +26323,7 @@ GL_PREFIX(VertexAttrib4fNV):
 GL_PREFIX(VertexAttrib4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5584(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26288,13 +26333,13 @@ GL_PREFIX(VertexAttrib4fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5584(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5584(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26304,7 +26349,7 @@ GL_PREFIX(VertexAttrib4fvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5584(%rax), %r11
+	movq	5592(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4fvNV), .-GL_PREFIX(VertexAttrib4fvNV)
@@ -26315,7 +26360,7 @@ GL_PREFIX(VertexAttrib4fvNV):
 GL_PREFIX(VertexAttrib4sNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5592(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26329,13 +26374,13 @@ GL_PREFIX(VertexAttrib4sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5592(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5592(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26349,7 +26394,7 @@ GL_PREFIX(VertexAttrib4sNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5592(%rax), %r11
+	movq	5600(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4sNV), .-GL_PREFIX(VertexAttrib4sNV)
@@ -26360,7 +26405,7 @@ GL_PREFIX(VertexAttrib4sNV):
 GL_PREFIX(VertexAttrib4svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5600(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26370,13 +26415,13 @@ GL_PREFIX(VertexAttrib4svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5600(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5600(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26386,7 +26431,7 @@ GL_PREFIX(VertexAttrib4svNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5600(%rax), %r11
+	movq	5608(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4svNV), .-GL_PREFIX(VertexAttrib4svNV)
@@ -26397,7 +26442,7 @@ GL_PREFIX(VertexAttrib4svNV):
 GL_PREFIX(VertexAttrib4ubNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5608(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26411,13 +26456,13 @@ GL_PREFIX(VertexAttrib4ubNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5608(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5608(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26431,7 +26476,7 @@ GL_PREFIX(VertexAttrib4ubNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5608(%rax), %r11
+	movq	5616(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4ubNV), .-GL_PREFIX(VertexAttrib4ubNV)
@@ -26442,7 +26487,7 @@ GL_PREFIX(VertexAttrib4ubNV):
 GL_PREFIX(VertexAttrib4ubvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5616(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26452,13 +26497,13 @@ GL_PREFIX(VertexAttrib4ubvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5616(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5616(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26468,7 +26513,7 @@ GL_PREFIX(VertexAttrib4ubvNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5616(%rax), %r11
+	movq	5624(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttrib4ubvNV), .-GL_PREFIX(VertexAttrib4ubvNV)
@@ -26479,7 +26524,7 @@ GL_PREFIX(VertexAttrib4ubvNV):
 GL_PREFIX(VertexAttribPointerNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5624(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26493,13 +26538,13 @@ GL_PREFIX(VertexAttribPointerNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5624(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5624(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26513,7 +26558,7 @@ GL_PREFIX(VertexAttribPointerNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5624(%rax), %r11
+	movq	5632(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribPointerNV), .-GL_PREFIX(VertexAttribPointerNV)
@@ -26524,7 +26569,7 @@ GL_PREFIX(VertexAttribPointerNV):
 GL_PREFIX(VertexAttribs1dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5632(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26534,13 +26579,13 @@ GL_PREFIX(VertexAttribs1dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5632(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5632(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26550,7 +26595,7 @@ GL_PREFIX(VertexAttribs1dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5632(%rax), %r11
+	movq	5640(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs1dvNV), .-GL_PREFIX(VertexAttribs1dvNV)
@@ -26561,7 +26606,7 @@ GL_PREFIX(VertexAttribs1dvNV):
 GL_PREFIX(VertexAttribs1fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5640(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26571,13 +26616,13 @@ GL_PREFIX(VertexAttribs1fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5640(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5640(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26587,7 +26632,7 @@ GL_PREFIX(VertexAttribs1fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5640(%rax), %r11
+	movq	5648(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs1fvNV), .-GL_PREFIX(VertexAttribs1fvNV)
@@ -26598,7 +26643,7 @@ GL_PREFIX(VertexAttribs1fvNV):
 GL_PREFIX(VertexAttribs1svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5648(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26608,13 +26653,13 @@ GL_PREFIX(VertexAttribs1svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5648(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5648(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26624,7 +26669,7 @@ GL_PREFIX(VertexAttribs1svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5648(%rax), %r11
+	movq	5656(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs1svNV), .-GL_PREFIX(VertexAttribs1svNV)
@@ -26635,7 +26680,7 @@ GL_PREFIX(VertexAttribs1svNV):
 GL_PREFIX(VertexAttribs2dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5656(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26645,13 +26690,13 @@ GL_PREFIX(VertexAttribs2dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5656(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5656(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26661,7 +26706,7 @@ GL_PREFIX(VertexAttribs2dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5656(%rax), %r11
+	movq	5664(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs2dvNV), .-GL_PREFIX(VertexAttribs2dvNV)
@@ -26672,7 +26717,7 @@ GL_PREFIX(VertexAttribs2dvNV):
 GL_PREFIX(VertexAttribs2fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5664(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26682,13 +26727,13 @@ GL_PREFIX(VertexAttribs2fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5664(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5664(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26698,7 +26743,7 @@ GL_PREFIX(VertexAttribs2fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5664(%rax), %r11
+	movq	5672(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs2fvNV), .-GL_PREFIX(VertexAttribs2fvNV)
@@ -26709,7 +26754,7 @@ GL_PREFIX(VertexAttribs2fvNV):
 GL_PREFIX(VertexAttribs2svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5672(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26719,13 +26764,13 @@ GL_PREFIX(VertexAttribs2svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5672(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5672(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26735,7 +26780,7 @@ GL_PREFIX(VertexAttribs2svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5672(%rax), %r11
+	movq	5680(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs2svNV), .-GL_PREFIX(VertexAttribs2svNV)
@@ -26746,7 +26791,7 @@ GL_PREFIX(VertexAttribs2svNV):
 GL_PREFIX(VertexAttribs3dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5680(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26756,13 +26801,13 @@ GL_PREFIX(VertexAttribs3dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5680(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5680(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26772,7 +26817,7 @@ GL_PREFIX(VertexAttribs3dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5680(%rax), %r11
+	movq	5688(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs3dvNV), .-GL_PREFIX(VertexAttribs3dvNV)
@@ -26783,7 +26828,7 @@ GL_PREFIX(VertexAttribs3dvNV):
 GL_PREFIX(VertexAttribs3fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5688(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26793,13 +26838,13 @@ GL_PREFIX(VertexAttribs3fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5688(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5688(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26809,7 +26854,7 @@ GL_PREFIX(VertexAttribs3fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5688(%rax), %r11
+	movq	5696(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs3fvNV), .-GL_PREFIX(VertexAttribs3fvNV)
@@ -26820,7 +26865,7 @@ GL_PREFIX(VertexAttribs3fvNV):
 GL_PREFIX(VertexAttribs3svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5696(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26830,13 +26875,13 @@ GL_PREFIX(VertexAttribs3svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5696(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5696(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26846,7 +26891,7 @@ GL_PREFIX(VertexAttribs3svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5696(%rax), %r11
+	movq	5704(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs3svNV), .-GL_PREFIX(VertexAttribs3svNV)
@@ -26857,7 +26902,7 @@ GL_PREFIX(VertexAttribs3svNV):
 GL_PREFIX(VertexAttribs4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5704(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26867,13 +26912,13 @@ GL_PREFIX(VertexAttribs4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5704(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5704(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26883,7 +26928,7 @@ GL_PREFIX(VertexAttribs4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5704(%rax), %r11
+	movq	5712(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4dvNV), .-GL_PREFIX(VertexAttribs4dvNV)
@@ -26894,7 +26939,7 @@ GL_PREFIX(VertexAttribs4dvNV):
 GL_PREFIX(VertexAttribs4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5712(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26904,13 +26949,13 @@ GL_PREFIX(VertexAttribs4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5712(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5712(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26920,7 +26965,7 @@ GL_PREFIX(VertexAttribs4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5712(%rax), %r11
+	movq	5720(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4fvNV), .-GL_PREFIX(VertexAttribs4fvNV)
@@ -26931,7 +26976,7 @@ GL_PREFIX(VertexAttribs4fvNV):
 GL_PREFIX(VertexAttribs4svNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5720(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26941,13 +26986,13 @@ GL_PREFIX(VertexAttribs4svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5720(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5720(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26957,7 +27002,7 @@ GL_PREFIX(VertexAttribs4svNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5720(%rax), %r11
+	movq	5728(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4svNV), .-GL_PREFIX(VertexAttribs4svNV)
@@ -26968,7 +27013,7 @@ GL_PREFIX(VertexAttribs4svNV):
 GL_PREFIX(VertexAttribs4ubvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5728(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -26978,13 +27023,13 @@ GL_PREFIX(VertexAttribs4ubvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5728(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5728(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -26994,7 +27039,7 @@ GL_PREFIX(VertexAttribs4ubvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5728(%rax), %r11
+	movq	5736(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(VertexAttribs4ubvNV), .-GL_PREFIX(VertexAttribs4ubvNV)
@@ -27005,7 +27050,7 @@ GL_PREFIX(VertexAttribs4ubvNV):
 GL_PREFIX(AlphaFragmentOp1ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5736(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27023,13 +27068,13 @@ GL_PREFIX(AlphaFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5736(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5736(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27047,7 +27092,7 @@ GL_PREFIX(AlphaFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5736(%rax), %r11
+	movq	5744(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AlphaFragmentOp1ATI), .-GL_PREFIX(AlphaFragmentOp1ATI)
@@ -27058,7 +27103,7 @@ GL_PREFIX(AlphaFragmentOp1ATI):
 GL_PREFIX(AlphaFragmentOp2ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5744(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27076,13 +27121,13 @@ GL_PREFIX(AlphaFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5744(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5744(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27100,7 +27145,7 @@ GL_PREFIX(AlphaFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5744(%rax), %r11
+	movq	5752(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AlphaFragmentOp2ATI), .-GL_PREFIX(AlphaFragmentOp2ATI)
@@ -27111,7 +27156,7 @@ GL_PREFIX(AlphaFragmentOp2ATI):
 GL_PREFIX(AlphaFragmentOp3ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5752(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27129,13 +27174,13 @@ GL_PREFIX(AlphaFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5752(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5752(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27153,7 +27198,7 @@ GL_PREFIX(AlphaFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5752(%rax), %r11
+	movq	5760(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(AlphaFragmentOp3ATI), .-GL_PREFIX(AlphaFragmentOp3ATI)
@@ -27164,25 +27209,25 @@ GL_PREFIX(AlphaFragmentOp3ATI):
 GL_PREFIX(BeginFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5760(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	5760(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5760(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	5760(%rax), %r11
+	movq	5768(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BeginFragmentShaderATI), .-GL_PREFIX(BeginFragmentShaderATI)
@@ -27193,25 +27238,25 @@ GL_PREFIX(BeginFragmentShaderATI):
 GL_PREFIX(BindFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5768(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5768(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5768(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5768(%rax), %r11
+	movq	5776(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindFragmentShaderATI), .-GL_PREFIX(BindFragmentShaderATI)
@@ -27222,7 +27267,7 @@ GL_PREFIX(BindFragmentShaderATI):
 GL_PREFIX(ColorFragmentOp1ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5776(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27240,13 +27285,13 @@ GL_PREFIX(ColorFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5776(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5776(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27264,7 +27309,7 @@ GL_PREFIX(ColorFragmentOp1ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5776(%rax), %r11
+	movq	5784(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorFragmentOp1ATI), .-GL_PREFIX(ColorFragmentOp1ATI)
@@ -27275,7 +27320,7 @@ GL_PREFIX(ColorFragmentOp1ATI):
 GL_PREFIX(ColorFragmentOp2ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5784(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27293,13 +27338,13 @@ GL_PREFIX(ColorFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5784(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5784(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27317,7 +27362,7 @@ GL_PREFIX(ColorFragmentOp2ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5784(%rax), %r11
+	movq	5792(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorFragmentOp2ATI), .-GL_PREFIX(ColorFragmentOp2ATI)
@@ -27328,7 +27373,7 @@ GL_PREFIX(ColorFragmentOp2ATI):
 GL_PREFIX(ColorFragmentOp3ATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5792(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27346,13 +27391,13 @@ GL_PREFIX(ColorFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5792(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5792(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27370,7 +27415,7 @@ GL_PREFIX(ColorFragmentOp3ATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5792(%rax), %r11
+	movq	5800(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ColorFragmentOp3ATI), .-GL_PREFIX(ColorFragmentOp3ATI)
@@ -27381,25 +27426,25 @@ GL_PREFIX(ColorFragmentOp3ATI):
 GL_PREFIX(DeleteFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5800(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5800(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5800(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5800(%rax), %r11
+	movq	5808(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteFragmentShaderATI), .-GL_PREFIX(DeleteFragmentShaderATI)
@@ -27410,25 +27455,25 @@ GL_PREFIX(DeleteFragmentShaderATI):
 GL_PREFIX(EndFragmentShaderATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5808(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
 	popq	%rbp
-	movq	5808(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5808(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rbp
 	call	_glapi_get_dispatch
 	popq	%rbp
-	movq	5808(%rax), %r11
+	movq	5816(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(EndFragmentShaderATI), .-GL_PREFIX(EndFragmentShaderATI)
@@ -27439,25 +27484,25 @@ GL_PREFIX(EndFragmentShaderATI):
 GL_PREFIX(GenFragmentShadersATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5816(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5816(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5816(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5816(%rax), %r11
+	movq	5824(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenFragmentShadersATI), .-GL_PREFIX(GenFragmentShadersATI)
@@ -27468,7 +27513,7 @@ GL_PREFIX(GenFragmentShadersATI):
 GL_PREFIX(PassTexCoordATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5824(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27478,13 +27523,13 @@ GL_PREFIX(PassTexCoordATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5824(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5824(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27494,7 +27539,7 @@ GL_PREFIX(PassTexCoordATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5824(%rax), %r11
+	movq	5832(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PassTexCoordATI), .-GL_PREFIX(PassTexCoordATI)
@@ -27505,7 +27550,7 @@ GL_PREFIX(PassTexCoordATI):
 GL_PREFIX(SampleMapATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5832(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27515,13 +27560,13 @@ GL_PREFIX(SampleMapATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5832(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5832(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27531,7 +27576,7 @@ GL_PREFIX(SampleMapATI):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5832(%rax), %r11
+	movq	5840(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SampleMapATI), .-GL_PREFIX(SampleMapATI)
@@ -27542,7 +27587,7 @@ GL_PREFIX(SampleMapATI):
 GL_PREFIX(SetFragmentShaderConstantATI):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5840(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27552,13 +27597,13 @@ GL_PREFIX(SetFragmentShaderConstantATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5840(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5840(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27568,7 +27613,7 @@ GL_PREFIX(SetFragmentShaderConstantATI):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5840(%rax), %r11
+	movq	5848(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(SetFragmentShaderConstantATI), .-GL_PREFIX(SetFragmentShaderConstantATI)
@@ -27579,7 +27624,7 @@ GL_PREFIX(SetFragmentShaderConstantATI):
 GL_PREFIX(PointParameteriNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5848(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27589,13 +27634,13 @@ GL_PREFIX(PointParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5848(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5848(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27605,7 +27650,7 @@ GL_PREFIX(PointParameteriNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5848(%rax), %r11
+	movq	5856(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(PointParameteriNV), .-GL_PREFIX(PointParameteriNV)
@@ -27616,7 +27661,7 @@ GL_PREFIX(PointParameteriNV):
 GL_PREFIX(PointParameterivNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5856(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27626,13 +27671,13 @@ GL_PREFIX(PointParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5856(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5856(%rax), %r11
+	movq	5864(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27642,40 +27687,10 @@ GL_PREFIX(PointParameterivNV):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5856(%rax), %r11
-	jmp	*%r11
-#endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(PointParameterivNV), .-GL_PREFIX(PointParameterivNV)
-
-	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_733)
-	.type	GL_PREFIX(_dispatch_stub_733), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_733))
-GL_PREFIX(_dispatch_stub_733):
-#if defined(GLX_USE_TLS)
-	call	_x86_64_get_dispatch@PLT
-	movq	5864(%rax), %r11
-	jmp	*%r11
-#elif defined(PTHREADS)
-	pushq	%rdi
-	call	_x86_64_get_dispatch@PLT
-	popq	%rdi
-	movq	5864(%rax), %r11
-	jmp	*%r11
-#else
-	movq	_glapi_Dispatch(%rip), %rax
-	testq	%rax, %rax
-	je	1f
-	movq	5864(%rax), %r11
-	jmp	*%r11
-1:
-	pushq	%rdi
-	call	_glapi_get_dispatch
-	popq	%rdi
 	movq	5864(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_733), .-GL_PREFIX(_dispatch_stub_733)
+	.size	GL_PREFIX(PointParameterivNV), .-GL_PREFIX(PointParameterivNV)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(_dispatch_stub_734)
@@ -27718,11 +27733,7 @@ GL_PREFIX(_dispatch_stub_735):
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
-	pushq	%rsi
-	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
-	popq	%rbp
-	popq	%rsi
 	popq	%rdi
 	movq	5880(%rax), %r11
 	jmp	*%r11
@@ -27734,11 +27745,7 @@ GL_PREFIX(_dispatch_stub_735):
 	jmp	*%r11
 1:
 	pushq	%rdi
-	pushq	%rsi
-	pushq	%rbp
 	call	_glapi_get_dispatch
-	popq	%rbp
-	popq	%rsi
 	popq	%rdi
 	movq	5880(%rax), %r11
 	jmp	*%r11
@@ -27794,7 +27801,11 @@ GL_PREFIX(_dispatch_stub_737):
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
 	call	_x86_64_get_dispatch@PLT
+	popq	%rbp
+	popq	%rsi
 	popq	%rdi
 	movq	5896(%rax), %r11
 	jmp	*%r11
@@ -27806,7 +27817,11 @@ GL_PREFIX(_dispatch_stub_737):
 	jmp	*%r11
 1:
 	pushq	%rdi
+	pushq	%rsi
+	pushq	%rbp
 	call	_glapi_get_dispatch
+	popq	%rbp
+	popq	%rsi
 	popq	%rdi
 	movq	5896(%rax), %r11
 	jmp	*%r11
@@ -27814,12 +27829,42 @@ GL_PREFIX(_dispatch_stub_737):
 	.size	GL_PREFIX(_dispatch_stub_737), .-GL_PREFIX(_dispatch_stub_737)
 
 	.p2align	4,,15
+	.globl	GL_PREFIX(_dispatch_stub_738)
+	.type	GL_PREFIX(_dispatch_stub_738), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_738))
+GL_PREFIX(_dispatch_stub_738):
+#if defined(GLX_USE_TLS)
+	call	_x86_64_get_dispatch@PLT
+	movq	5904(%rax), %r11
+	jmp	*%r11
+#elif defined(PTHREADS)
+	pushq	%rdi
+	call	_x86_64_get_dispatch@PLT
+	popq	%rdi
+	movq	5904(%rax), %r11
+	jmp	*%r11
+#else
+	movq	_glapi_Dispatch(%rip), %rax
+	testq	%rax, %rax
+	je	1f
+	movq	5904(%rax), %r11
+	jmp	*%r11
+1:
+	pushq	%rdi
+	call	_glapi_get_dispatch
+	popq	%rdi
+	movq	5904(%rax), %r11
+	jmp	*%r11
+#endif /* defined(GLX_USE_TLS) */
+	.size	GL_PREFIX(_dispatch_stub_738), .-GL_PREFIX(_dispatch_stub_738)
+
+	.p2align	4,,15
 	.globl	GL_PREFIX(GetProgramNamedParameterdvNV)
 	.type	GL_PREFIX(GetProgramNamedParameterdvNV), @function
 GL_PREFIX(GetProgramNamedParameterdvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5904(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27833,13 +27878,13 @@ GL_PREFIX(GetProgramNamedParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5904(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5904(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27853,7 +27898,7 @@ GL_PREFIX(GetProgramNamedParameterdvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5904(%rax), %r11
+	movq	5912(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramNamedParameterdvNV), .-GL_PREFIX(GetProgramNamedParameterdvNV)
@@ -27864,7 +27909,7 @@ GL_PREFIX(GetProgramNamedParameterdvNV):
 GL_PREFIX(GetProgramNamedParameterfvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5912(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27878,13 +27923,13 @@ GL_PREFIX(GetProgramNamedParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5912(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5912(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -27898,7 +27943,7 @@ GL_PREFIX(GetProgramNamedParameterfvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5912(%rax), %r11
+	movq	5920(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetProgramNamedParameterfvNV), .-GL_PREFIX(GetProgramNamedParameterfvNV)
@@ -27909,7 +27954,7 @@ GL_PREFIX(GetProgramNamedParameterfvNV):
 GL_PREFIX(ProgramNamedParameter4dNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5920(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$56, %rsp
@@ -27929,13 +27974,13 @@ GL_PREFIX(ProgramNamedParameter4dNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5920(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5920(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$56, %rsp
@@ -27955,7 +28000,7 @@ GL_PREFIX(ProgramNamedParameter4dNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5920(%rax), %r11
+	movq	5928(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4dNV), .-GL_PREFIX(ProgramNamedParameter4dNV)
@@ -27966,7 +28011,7 @@ GL_PREFIX(ProgramNamedParameter4dNV):
 GL_PREFIX(ProgramNamedParameter4dvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5928(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -27980,13 +28025,13 @@ GL_PREFIX(ProgramNamedParameter4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5928(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5928(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28000,7 +28045,7 @@ GL_PREFIX(ProgramNamedParameter4dvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5928(%rax), %r11
+	movq	5936(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4dvNV), .-GL_PREFIX(ProgramNamedParameter4dvNV)
@@ -28011,7 +28056,7 @@ GL_PREFIX(ProgramNamedParameter4dvNV):
 GL_PREFIX(ProgramNamedParameter4fNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5936(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	subq	$56, %rsp
@@ -28031,13 +28076,13 @@ GL_PREFIX(ProgramNamedParameter4fNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5936(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5936(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 1:
 	subq	$56, %rsp
@@ -28057,7 +28102,7 @@ GL_PREFIX(ProgramNamedParameter4fNV):
 	movq	8(%rsp), %rsi
 	movq	(%rsp), %rdi
 	addq	$56, %rsp
-	movq	5936(%rax), %r11
+	movq	5944(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4fNV), .-GL_PREFIX(ProgramNamedParameter4fNV)
@@ -28068,7 +28113,7 @@ GL_PREFIX(ProgramNamedParameter4fNV):
 GL_PREFIX(ProgramNamedParameter4fvNV):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5944(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28082,13 +28127,13 @@ GL_PREFIX(ProgramNamedParameter4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5944(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5944(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28102,19 +28147,19 @@ GL_PREFIX(ProgramNamedParameter4fvNV):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	5944(%rax), %r11
+	movq	5952(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(ProgramNamedParameter4fvNV), .-GL_PREFIX(ProgramNamedParameter4fvNV)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_744)
-	.type	GL_PREFIX(_dispatch_stub_744), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_744))
-GL_PREFIX(_dispatch_stub_744):
+	.globl	GL_PREFIX(_dispatch_stub_745)
+	.type	GL_PREFIX(_dispatch_stub_745), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_745))
+GL_PREFIX(_dispatch_stub_745):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5952(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28124,13 +28169,13 @@ GL_PREFIX(_dispatch_stub_744):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5952(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5952(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28140,19 +28185,19 @@ GL_PREFIX(_dispatch_stub_744):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5952(%rax), %r11
+	movq	5960(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_744), .-GL_PREFIX(_dispatch_stub_744)
+	.size	GL_PREFIX(_dispatch_stub_745), .-GL_PREFIX(_dispatch_stub_745)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_745)
-	.type	GL_PREFIX(_dispatch_stub_745), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_745))
-GL_PREFIX(_dispatch_stub_745):
+	.globl	GL_PREFIX(_dispatch_stub_746)
+	.type	GL_PREFIX(_dispatch_stub_746), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_746))
+GL_PREFIX(_dispatch_stub_746):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5960(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28162,13 +28207,13 @@ GL_PREFIX(_dispatch_stub_745):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5960(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5960(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28178,10 +28223,10 @@ GL_PREFIX(_dispatch_stub_745):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5960(%rax), %r11
+	movq	5968(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_745), .-GL_PREFIX(_dispatch_stub_745)
+	.size	GL_PREFIX(_dispatch_stub_746), .-GL_PREFIX(_dispatch_stub_746)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(BindFramebufferEXT)
@@ -28189,7 +28234,7 @@ GL_PREFIX(_dispatch_stub_745):
 GL_PREFIX(BindFramebufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5968(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28199,13 +28244,13 @@ GL_PREFIX(BindFramebufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5968(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5968(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28215,7 +28260,7 @@ GL_PREFIX(BindFramebufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5968(%rax), %r11
+	movq	5976(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindFramebufferEXT), .-GL_PREFIX(BindFramebufferEXT)
@@ -28226,7 +28271,7 @@ GL_PREFIX(BindFramebufferEXT):
 GL_PREFIX(BindRenderbufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5976(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28236,13 +28281,13 @@ GL_PREFIX(BindRenderbufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5976(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5976(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28252,7 +28297,7 @@ GL_PREFIX(BindRenderbufferEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5976(%rax), %r11
+	movq	5984(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(BindRenderbufferEXT), .-GL_PREFIX(BindRenderbufferEXT)
@@ -28263,25 +28308,25 @@ GL_PREFIX(BindRenderbufferEXT):
 GL_PREFIX(CheckFramebufferStatusEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5984(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	5984(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5984(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	5984(%rax), %r11
+	movq	5992(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(CheckFramebufferStatusEXT), .-GL_PREFIX(CheckFramebufferStatusEXT)
@@ -28292,7 +28337,7 @@ GL_PREFIX(CheckFramebufferStatusEXT):
 GL_PREFIX(DeleteFramebuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	5992(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28302,13 +28347,13 @@ GL_PREFIX(DeleteFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5992(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	5992(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28318,7 +28363,7 @@ GL_PREFIX(DeleteFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	5992(%rax), %r11
+	movq	6000(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteFramebuffersEXT), .-GL_PREFIX(DeleteFramebuffersEXT)
@@ -28329,7 +28374,7 @@ GL_PREFIX(DeleteFramebuffersEXT):
 GL_PREFIX(DeleteRenderbuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6000(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28339,13 +28384,13 @@ GL_PREFIX(DeleteRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6000(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6000(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28355,7 +28400,7 @@ GL_PREFIX(DeleteRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6000(%rax), %r11
+	movq	6008(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(DeleteRenderbuffersEXT), .-GL_PREFIX(DeleteRenderbuffersEXT)
@@ -28366,7 +28411,7 @@ GL_PREFIX(DeleteRenderbuffersEXT):
 GL_PREFIX(FramebufferRenderbufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6008(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28380,13 +28425,13 @@ GL_PREFIX(FramebufferRenderbufferEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6008(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6008(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28400,7 +28445,7 @@ GL_PREFIX(FramebufferRenderbufferEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6008(%rax), %r11
+	movq	6016(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferRenderbufferEXT), .-GL_PREFIX(FramebufferRenderbufferEXT)
@@ -28411,7 +28456,7 @@ GL_PREFIX(FramebufferRenderbufferEXT):
 GL_PREFIX(FramebufferTexture1DEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6016(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28425,13 +28470,13 @@ GL_PREFIX(FramebufferTexture1DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6016(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6016(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28445,7 +28490,7 @@ GL_PREFIX(FramebufferTexture1DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6016(%rax), %r11
+	movq	6024(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTexture1DEXT), .-GL_PREFIX(FramebufferTexture1DEXT)
@@ -28456,7 +28501,7 @@ GL_PREFIX(FramebufferTexture1DEXT):
 GL_PREFIX(FramebufferTexture2DEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6024(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28470,13 +28515,13 @@ GL_PREFIX(FramebufferTexture2DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6024(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6024(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28490,7 +28535,7 @@ GL_PREFIX(FramebufferTexture2DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6024(%rax), %r11
+	movq	6032(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTexture2DEXT), .-GL_PREFIX(FramebufferTexture2DEXT)
@@ -28501,7 +28546,7 @@ GL_PREFIX(FramebufferTexture2DEXT):
 GL_PREFIX(FramebufferTexture3DEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6032(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28519,13 +28564,13 @@ GL_PREFIX(FramebufferTexture3DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6032(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6032(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28543,7 +28588,7 @@ GL_PREFIX(FramebufferTexture3DEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6032(%rax), %r11
+	movq	6040(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTexture3DEXT), .-GL_PREFIX(FramebufferTexture3DEXT)
@@ -28554,7 +28599,7 @@ GL_PREFIX(FramebufferTexture3DEXT):
 GL_PREFIX(GenFramebuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6040(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28564,13 +28609,13 @@ GL_PREFIX(GenFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6040(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6040(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28580,7 +28625,7 @@ GL_PREFIX(GenFramebuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6040(%rax), %r11
+	movq	6048(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenFramebuffersEXT), .-GL_PREFIX(GenFramebuffersEXT)
@@ -28591,7 +28636,7 @@ GL_PREFIX(GenFramebuffersEXT):
 GL_PREFIX(GenRenderbuffersEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6048(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28601,13 +28646,13 @@ GL_PREFIX(GenRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6048(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6048(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28617,7 +28662,7 @@ GL_PREFIX(GenRenderbuffersEXT):
 	popq	%rbp
 	popq	%rsi
 	popq	%rdi
-	movq	6048(%rax), %r11
+	movq	6056(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenRenderbuffersEXT), .-GL_PREFIX(GenRenderbuffersEXT)
@@ -28628,25 +28673,25 @@ GL_PREFIX(GenRenderbuffersEXT):
 GL_PREFIX(GenerateMipmapEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6056(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6056(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6056(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6056(%rax), %r11
+	movq	6064(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GenerateMipmapEXT), .-GL_PREFIX(GenerateMipmapEXT)
@@ -28657,7 +28702,7 @@ GL_PREFIX(GenerateMipmapEXT):
 GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6064(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28671,13 +28716,13 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6064(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6064(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28691,7 +28736,7 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6064(%rax), %r11
+	movq	6072(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetFramebufferAttachmentParameterivEXT), .-GL_PREFIX(GetFramebufferAttachmentParameterivEXT)
@@ -28702,7 +28747,7 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT):
 GL_PREFIX(GetRenderbufferParameterivEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6072(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28712,13 +28757,13 @@ GL_PREFIX(GetRenderbufferParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6072(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6072(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28728,7 +28773,7 @@ GL_PREFIX(GetRenderbufferParameterivEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6072(%rax), %r11
+	movq	6080(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(GetRenderbufferParameterivEXT), .-GL_PREFIX(GetRenderbufferParameterivEXT)
@@ -28739,25 +28784,25 @@ GL_PREFIX(GetRenderbufferParameterivEXT):
 GL_PREFIX(IsFramebufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6080(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6080(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6080(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6080(%rax), %r11
+	movq	6088(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IsFramebufferEXT), .-GL_PREFIX(IsFramebufferEXT)
@@ -28768,25 +28813,25 @@ GL_PREFIX(IsFramebufferEXT):
 GL_PREFIX(IsRenderbufferEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6088(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
 	call	_x86_64_get_dispatch@PLT
 	popq	%rdi
-	movq	6088(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6088(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
 	call	_glapi_get_dispatch
 	popq	%rdi
-	movq	6088(%rax), %r11
+	movq	6096(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(IsRenderbufferEXT), .-GL_PREFIX(IsRenderbufferEXT)
@@ -28797,7 +28842,7 @@ GL_PREFIX(IsRenderbufferEXT):
 GL_PREFIX(RenderbufferStorageEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6096(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28811,13 +28856,13 @@ GL_PREFIX(RenderbufferStorageEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6096(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6096(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28831,19 +28876,19 @@ GL_PREFIX(RenderbufferStorageEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6096(%rax), %r11
+	movq	6104(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(RenderbufferStorageEXT), .-GL_PREFIX(RenderbufferStorageEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_763)
-	.type	GL_PREFIX(_dispatch_stub_763), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_763))
-GL_PREFIX(_dispatch_stub_763):
+	.globl	GL_PREFIX(_dispatch_stub_764)
+	.type	GL_PREFIX(_dispatch_stub_764), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_764))
+GL_PREFIX(_dispatch_stub_764):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6104(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28861,13 +28906,13 @@ GL_PREFIX(_dispatch_stub_763):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6104(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6104(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28885,10 +28930,10 @@ GL_PREFIX(_dispatch_stub_763):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6104(%rax), %r11
+	movq	6112(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_763), .-GL_PREFIX(_dispatch_stub_763)
+	.size	GL_PREFIX(_dispatch_stub_764), .-GL_PREFIX(_dispatch_stub_764)
 
 	.p2align	4,,15
 	.globl	GL_PREFIX(FramebufferTextureLayerEXT)
@@ -28896,7 +28941,7 @@ GL_PREFIX(_dispatch_stub_763):
 GL_PREFIX(FramebufferTextureLayerEXT):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6112(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28910,13 +28955,13 @@ GL_PREFIX(FramebufferTextureLayerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6112(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6112(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28930,19 +28975,19 @@ GL_PREFIX(FramebufferTextureLayerEXT):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6112(%rax), %r11
+	movq	6120(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
 	.size	GL_PREFIX(FramebufferTextureLayerEXT), .-GL_PREFIX(FramebufferTextureLayerEXT)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_765)
-	.type	GL_PREFIX(_dispatch_stub_765), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_765))
-GL_PREFIX(_dispatch_stub_765):
+	.globl	GL_PREFIX(_dispatch_stub_766)
+	.type	GL_PREFIX(_dispatch_stub_766), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_766))
+GL_PREFIX(_dispatch_stub_766):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6120(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -28956,13 +29001,13 @@ GL_PREFIX(_dispatch_stub_765):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6120(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6120(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -28976,19 +29021,19 @@ GL_PREFIX(_dispatch_stub_765):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6120(%rax), %r11
+	movq	6128(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_765), .-GL_PREFIX(_dispatch_stub_765)
+	.size	GL_PREFIX(_dispatch_stub_766), .-GL_PREFIX(_dispatch_stub_766)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_766)
-	.type	GL_PREFIX(_dispatch_stub_766), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_766))
-GL_PREFIX(_dispatch_stub_766):
+	.globl	GL_PREFIX(_dispatch_stub_767)
+	.type	GL_PREFIX(_dispatch_stub_767), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_767))
+GL_PREFIX(_dispatch_stub_767):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6128(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29002,13 +29047,13 @@ GL_PREFIX(_dispatch_stub_766):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6128(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6128(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29022,19 +29067,19 @@ GL_PREFIX(_dispatch_stub_766):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6128(%rax), %r11
+	movq	6136(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_766), .-GL_PREFIX(_dispatch_stub_766)
+	.size	GL_PREFIX(_dispatch_stub_767), .-GL_PREFIX(_dispatch_stub_767)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_767)
-	.type	GL_PREFIX(_dispatch_stub_767), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_767))
-GL_PREFIX(_dispatch_stub_767):
+	.globl	GL_PREFIX(_dispatch_stub_768)
+	.type	GL_PREFIX(_dispatch_stub_768), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_768))
+GL_PREFIX(_dispatch_stub_768):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6136(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29048,13 +29093,13 @@ GL_PREFIX(_dispatch_stub_767):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6136(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6136(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29068,19 +29113,19 @@ GL_PREFIX(_dispatch_stub_767):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6136(%rax), %r11
+	movq	6144(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_767), .-GL_PREFIX(_dispatch_stub_767)
+	.size	GL_PREFIX(_dispatch_stub_768), .-GL_PREFIX(_dispatch_stub_768)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_768)
-	.type	GL_PREFIX(_dispatch_stub_768), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_768))
-GL_PREFIX(_dispatch_stub_768):
+	.globl	GL_PREFIX(_dispatch_stub_769)
+	.type	GL_PREFIX(_dispatch_stub_769), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_769))
+GL_PREFIX(_dispatch_stub_769):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6144(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29090,13 +29135,13 @@ GL_PREFIX(_dispatch_stub_768):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6144(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6144(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29106,19 +29151,19 @@ GL_PREFIX(_dispatch_stub_768):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6144(%rax), %r11
+	movq	6152(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_768), .-GL_PREFIX(_dispatch_stub_768)
+	.size	GL_PREFIX(_dispatch_stub_769), .-GL_PREFIX(_dispatch_stub_769)
 
 	.p2align	4,,15
-	.globl	GL_PREFIX(_dispatch_stub_769)
-	.type	GL_PREFIX(_dispatch_stub_769), @function
-	HIDDEN(GL_PREFIX(_dispatch_stub_769))
-GL_PREFIX(_dispatch_stub_769):
+	.globl	GL_PREFIX(_dispatch_stub_770)
+	.type	GL_PREFIX(_dispatch_stub_770), @function
+	HIDDEN(GL_PREFIX(_dispatch_stub_770))
+GL_PREFIX(_dispatch_stub_770):
 #if defined(GLX_USE_TLS)
 	call	_x86_64_get_dispatch@PLT
-	movq	6152(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 #elif defined(PTHREADS)
 	pushq	%rdi
@@ -29128,13 +29173,13 @@ GL_PREFIX(_dispatch_stub_769):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6152(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 #else
 	movq	_glapi_Dispatch(%rip), %rax
 	testq	%rax, %rax
 	je	1f
-	movq	6152(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 1:
 	pushq	%rdi
@@ -29144,10 +29189,10 @@ GL_PREFIX(_dispatch_stub_769):
 	popq	%rdx
 	popq	%rsi
 	popq	%rdi
-	movq	6152(%rax), %r11
+	movq	6160(%rax), %r11
 	jmp	*%r11
 #endif /* defined(GLX_USE_TLS) */
-	.size	GL_PREFIX(_dispatch_stub_769), .-GL_PREFIX(_dispatch_stub_769)
+	.size	GL_PREFIX(_dispatch_stub_770), .-GL_PREFIX(_dispatch_stub_770)
 
 	.globl GL_PREFIX(ArrayElementEXT) ; .set GL_PREFIX(ArrayElementEXT), GL_PREFIX(ArrayElement)
 	.globl GL_PREFIX(BindTextureEXT) ; .set GL_PREFIX(BindTextureEXT), GL_PREFIX(BindTexture)
@@ -29402,7 +29447,26 @@ GL_PREFIX(_dispatch_stub_769):
 	.globl GL_PREFIX(IsProgramARB) ; .set GL_PREFIX(IsProgramARB), GL_PREFIX(IsProgramNV)
 	.globl GL_PREFIX(PointParameteri) ; .set GL_PREFIX(PointParameteri), GL_PREFIX(PointParameteriNV)
 	.globl GL_PREFIX(PointParameteriv) ; .set GL_PREFIX(PointParameteriv), GL_PREFIX(PointParameterivNV)
-	.globl GL_PREFIX(BlendEquationSeparate) ; .set GL_PREFIX(BlendEquationSeparate), GL_PREFIX(_dispatch_stub_745)
+	.globl GL_PREFIX(BlendEquationSeparate) ; .set GL_PREFIX(BlendEquationSeparate), GL_PREFIX(_dispatch_stub_746)
+	.globl GL_PREFIX(BindFramebuffer) ; .set GL_PREFIX(BindFramebuffer), GL_PREFIX(BindFramebufferEXT)
+	.globl GL_PREFIX(BindRenderbuffer) ; .set GL_PREFIX(BindRenderbuffer), GL_PREFIX(BindRenderbufferEXT)
+	.globl GL_PREFIX(CheckFramebufferStatus) ; .set GL_PREFIX(CheckFramebufferStatus), GL_PREFIX(CheckFramebufferStatusEXT)
+	.globl GL_PREFIX(DeleteFramebuffers) ; .set GL_PREFIX(DeleteFramebuffers), GL_PREFIX(DeleteFramebuffersEXT)
+	.globl GL_PREFIX(DeleteRenderbuffers) ; .set GL_PREFIX(DeleteRenderbuffers), GL_PREFIX(DeleteRenderbuffersEXT)
+	.globl GL_PREFIX(FramebufferRenderbuffer) ; .set GL_PREFIX(FramebufferRenderbuffer), GL_PREFIX(FramebufferRenderbufferEXT)
+	.globl GL_PREFIX(FramebufferTexture1D) ; .set GL_PREFIX(FramebufferTexture1D), GL_PREFIX(FramebufferTexture1DEXT)
+	.globl GL_PREFIX(FramebufferTexture2D) ; .set GL_PREFIX(FramebufferTexture2D), GL_PREFIX(FramebufferTexture2DEXT)
+	.globl GL_PREFIX(FramebufferTexture3D) ; .set GL_PREFIX(FramebufferTexture3D), GL_PREFIX(FramebufferTexture3DEXT)
+	.globl GL_PREFIX(GenFramebuffers) ; .set GL_PREFIX(GenFramebuffers), GL_PREFIX(GenFramebuffersEXT)
+	.globl GL_PREFIX(GenRenderbuffers) ; .set GL_PREFIX(GenRenderbuffers), GL_PREFIX(GenRenderbuffersEXT)
+	.globl GL_PREFIX(GenerateMipmap) ; .set GL_PREFIX(GenerateMipmap), GL_PREFIX(GenerateMipmapEXT)
+	.globl GL_PREFIX(GetFramebufferAttachmentParameteriv) ; .set GL_PREFIX(GetFramebufferAttachmentParameteriv), GL_PREFIX(GetFramebufferAttachmentParameterivEXT)
+	.globl GL_PREFIX(GetRenderbufferParameteriv) ; .set GL_PREFIX(GetRenderbufferParameteriv), GL_PREFIX(GetRenderbufferParameterivEXT)
+	.globl GL_PREFIX(IsFramebuffer) ; .set GL_PREFIX(IsFramebuffer), GL_PREFIX(IsFramebufferEXT)
+	.globl GL_PREFIX(IsRenderbuffer) ; .set GL_PREFIX(IsRenderbuffer), GL_PREFIX(IsRenderbufferEXT)
+	.globl GL_PREFIX(RenderbufferStorage) ; .set GL_PREFIX(RenderbufferStorage), GL_PREFIX(RenderbufferStorageEXT)
+	.globl GL_PREFIX(BlitFramebuffer) ; .set GL_PREFIX(BlitFramebuffer), GL_PREFIX(_dispatch_stub_764)
+	.globl GL_PREFIX(FramebufferTextureLayer) ; .set GL_PREFIX(FramebufferTextureLayer), GL_PREFIX(FramebufferTextureLayerEXT)
 
 #if defined(GLX_USE_TLS) && defined(__linux__)
 	.section ".note.ABI-tag", "a"
diff --git a/src/mesa/x86/Makefile b/src/mesa/x86/Makefile
index dc8c7f355e..aa49a9134a 100644
--- a/src/mesa/x86/Makefile
+++ b/src/mesa/x86/Makefile
@@ -21,7 +21,7 @@ clean:
 
 
 gen_matypes: gen_matypes.c
-	$(CC) $(INCLUDE_DIRS) $(CFLAGS) gen_matypes.c -o gen_matypes
+	$(HOST_CC) $(INCLUDE_DIRS) $(HOST_CFLAGS) gen_matypes.c -o gen_matypes
 
 # need some special rules here, unfortunately
 matypes.h: ../main/mtypes.h ../tnl/t_context.h gen_matypes
diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S
index b99c2b6520..2e05a74ac6 100644
--- a/src/mesa/x86/glapi_x86.S
+++ b/src/mesa/x86/glapi_x86.S
@@ -712,23 +712,24 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(GetActiveAttribARB, _gloffset_GetActiveAttribARB, GetActiveAttribARB@28)
 	GL_STUB(GetAttribLocationARB, _gloffset_GetAttribLocationARB, GetAttribLocationARB@8)
 	GL_STUB(DrawBuffersARB, _gloffset_DrawBuffersARB, DrawBuffersARB@8)
+	GL_STUB(RenderbufferStorageMultisample, _gloffset_RenderbufferStorageMultisample, RenderbufferStorageMultisample@20)
 	GL_STUB(PolygonOffsetEXT, _gloffset_PolygonOffsetEXT, PolygonOffsetEXT@8)
-	GL_STUB(_dispatch_stub_562, _gloffset_GetPixelTexGenParameterfvSGIS, _dispatch_stub_562@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_562, _dispatch_stub_562@8))
-	GL_STUB(_dispatch_stub_563, _gloffset_GetPixelTexGenParameterivSGIS, _dispatch_stub_563@8)
+	GL_STUB(_dispatch_stub_563, _gloffset_GetPixelTexGenParameterfvSGIS, _dispatch_stub_563@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_563, _dispatch_stub_563@8))
-	GL_STUB(_dispatch_stub_564, _gloffset_PixelTexGenParameterfSGIS, _dispatch_stub_564@8)
+	GL_STUB(_dispatch_stub_564, _gloffset_GetPixelTexGenParameterivSGIS, _dispatch_stub_564@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_564, _dispatch_stub_564@8))
-	GL_STUB(_dispatch_stub_565, _gloffset_PixelTexGenParameterfvSGIS, _dispatch_stub_565@8)
+	GL_STUB(_dispatch_stub_565, _gloffset_PixelTexGenParameterfSGIS, _dispatch_stub_565@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_565, _dispatch_stub_565@8))
-	GL_STUB(_dispatch_stub_566, _gloffset_PixelTexGenParameteriSGIS, _dispatch_stub_566@8)
+	GL_STUB(_dispatch_stub_566, _gloffset_PixelTexGenParameterfvSGIS, _dispatch_stub_566@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_566, _dispatch_stub_566@8))
-	GL_STUB(_dispatch_stub_567, _gloffset_PixelTexGenParameterivSGIS, _dispatch_stub_567@8)
+	GL_STUB(_dispatch_stub_567, _gloffset_PixelTexGenParameteriSGIS, _dispatch_stub_567@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_567, _dispatch_stub_567@8))
-	GL_STUB(_dispatch_stub_568, _gloffset_SampleMaskSGIS, _dispatch_stub_568@8)
+	GL_STUB(_dispatch_stub_568, _gloffset_PixelTexGenParameterivSGIS, _dispatch_stub_568@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_568, _dispatch_stub_568@8))
-	GL_STUB(_dispatch_stub_569, _gloffset_SamplePatternSGIS, _dispatch_stub_569@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_569, _dispatch_stub_569@4))
+	GL_STUB(_dispatch_stub_569, _gloffset_SampleMaskSGIS, _dispatch_stub_569@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_569, _dispatch_stub_569@8))
+	GL_STUB(_dispatch_stub_570, _gloffset_SamplePatternSGIS, _dispatch_stub_570@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_570, _dispatch_stub_570@4))
 	GL_STUB(ColorPointerEXT, _gloffset_ColorPointerEXT, ColorPointerEXT@20)
 	GL_STUB(EdgeFlagPointerEXT, _gloffset_EdgeFlagPointerEXT, EdgeFlagPointerEXT@12)
 	GL_STUB(IndexPointerEXT, _gloffset_IndexPointerEXT, IndexPointerEXT@16)
@@ -739,10 +740,10 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(PointParameterfvEXT, _gloffset_PointParameterfvEXT, PointParameterfvEXT@8)
 	GL_STUB(LockArraysEXT, _gloffset_LockArraysEXT, LockArraysEXT@8)
 	GL_STUB(UnlockArraysEXT, _gloffset_UnlockArraysEXT, UnlockArraysEXT@0)
-	GL_STUB(_dispatch_stub_580, _gloffset_CullParameterdvEXT, _dispatch_stub_580@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_580, _dispatch_stub_580@8))
-	GL_STUB(_dispatch_stub_581, _gloffset_CullParameterfvEXT, _dispatch_stub_581@8)
+	GL_STUB(_dispatch_stub_581, _gloffset_CullParameterdvEXT, _dispatch_stub_581@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_581, _dispatch_stub_581@8))
+	GL_STUB(_dispatch_stub_582, _gloffset_CullParameterfvEXT, _dispatch_stub_582@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_582, _dispatch_stub_582@8))
 	GL_STUB(SecondaryColor3bEXT, _gloffset_SecondaryColor3bEXT, SecondaryColor3bEXT@12)
 	GL_STUB(SecondaryColor3bvEXT, _gloffset_SecondaryColor3bvEXT, SecondaryColor3bvEXT@4)
 	GL_STUB(SecondaryColor3dEXT, _gloffset_SecondaryColor3dEXT, SecondaryColor3dEXT@24)
@@ -767,8 +768,8 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(FogCoorddvEXT, _gloffset_FogCoorddvEXT, FogCoorddvEXT@4)
 	GL_STUB(FogCoordfEXT, _gloffset_FogCoordfEXT, FogCoordfEXT@4)
 	GL_STUB(FogCoordfvEXT, _gloffset_FogCoordfvEXT, FogCoordfvEXT@4)
-	GL_STUB(_dispatch_stub_606, _gloffset_PixelTexGenSGIX, _dispatch_stub_606@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_606, _dispatch_stub_606@4))
+	GL_STUB(_dispatch_stub_607, _gloffset_PixelTexGenSGIX, _dispatch_stub_607@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_607, _dispatch_stub_607@4))
 	GL_STUB(BlendFuncSeparateEXT, _gloffset_BlendFuncSeparateEXT, BlendFuncSeparateEXT@16)
 	GL_STUB(FlushVertexArrayRangeNV, _gloffset_FlushVertexArrayRangeNV, FlushVertexArrayRangeNV@0)
 	GL_STUB(VertexArrayRangeNV, _gloffset_VertexArrayRangeNV, VertexArrayRangeNV@8)
@@ -810,24 +811,24 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(WindowPos4ivMESA, _gloffset_WindowPos4ivMESA, WindowPos4ivMESA@4)
 	GL_STUB(WindowPos4sMESA, _gloffset_WindowPos4sMESA, WindowPos4sMESA@16)
 	GL_STUB(WindowPos4svMESA, _gloffset_WindowPos4svMESA, WindowPos4svMESA@4)
-	GL_STUB(_dispatch_stub_648, _gloffset_MultiModeDrawArraysIBM, _dispatch_stub_648@20)
-	HIDDEN(GL_PREFIX(_dispatch_stub_648, _dispatch_stub_648@20))
-	GL_STUB(_dispatch_stub_649, _gloffset_MultiModeDrawElementsIBM, _dispatch_stub_649@24)
-	HIDDEN(GL_PREFIX(_dispatch_stub_649, _dispatch_stub_649@24))
-	GL_STUB(_dispatch_stub_650, _gloffset_DeleteFencesNV, _dispatch_stub_650@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_650, _dispatch_stub_650@8))
-	GL_STUB(_dispatch_stub_651, _gloffset_FinishFenceNV, _dispatch_stub_651@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_651, _dispatch_stub_651@4))
-	GL_STUB(_dispatch_stub_652, _gloffset_GenFencesNV, _dispatch_stub_652@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_652, _dispatch_stub_652@8))
-	GL_STUB(_dispatch_stub_653, _gloffset_GetFenceivNV, _dispatch_stub_653@12)
-	HIDDEN(GL_PREFIX(_dispatch_stub_653, _dispatch_stub_653@12))
-	GL_STUB(_dispatch_stub_654, _gloffset_IsFenceNV, _dispatch_stub_654@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_654, _dispatch_stub_654@4))
-	GL_STUB(_dispatch_stub_655, _gloffset_SetFenceNV, _dispatch_stub_655@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_655, _dispatch_stub_655@8))
-	GL_STUB(_dispatch_stub_656, _gloffset_TestFenceNV, _dispatch_stub_656@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_656, _dispatch_stub_656@4))
+	GL_STUB(_dispatch_stub_649, _gloffset_MultiModeDrawArraysIBM, _dispatch_stub_649@20)
+	HIDDEN(GL_PREFIX(_dispatch_stub_649, _dispatch_stub_649@20))
+	GL_STUB(_dispatch_stub_650, _gloffset_MultiModeDrawElementsIBM, _dispatch_stub_650@24)
+	HIDDEN(GL_PREFIX(_dispatch_stub_650, _dispatch_stub_650@24))
+	GL_STUB(_dispatch_stub_651, _gloffset_DeleteFencesNV, _dispatch_stub_651@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_651, _dispatch_stub_651@8))
+	GL_STUB(_dispatch_stub_652, _gloffset_FinishFenceNV, _dispatch_stub_652@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_652, _dispatch_stub_652@4))
+	GL_STUB(_dispatch_stub_653, _gloffset_GenFencesNV, _dispatch_stub_653@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_653, _dispatch_stub_653@8))
+	GL_STUB(_dispatch_stub_654, _gloffset_GetFenceivNV, _dispatch_stub_654@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_654, _dispatch_stub_654@12))
+	GL_STUB(_dispatch_stub_655, _gloffset_IsFenceNV, _dispatch_stub_655@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_655, _dispatch_stub_655@4))
+	GL_STUB(_dispatch_stub_656, _gloffset_SetFenceNV, _dispatch_stub_656@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_656, _dispatch_stub_656@8))
+	GL_STUB(_dispatch_stub_657, _gloffset_TestFenceNV, _dispatch_stub_657@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_657, _dispatch_stub_657@4))
 	GL_STUB(AreProgramsResidentNV, _gloffset_AreProgramsResidentNV, AreProgramsResidentNV@12)
 	GL_STUB(BindProgramNV, _gloffset_BindProgramNV, BindProgramNV@8)
 	GL_STUB(DeleteProgramsNV, _gloffset_DeleteProgramsNV, DeleteProgramsNV@8)
@@ -904,26 +905,26 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(SetFragmentShaderConstantATI, _gloffset_SetFragmentShaderConstantATI, SetFragmentShaderConstantATI@8)
 	GL_STUB(PointParameteriNV, _gloffset_PointParameteriNV, PointParameteriNV@8)
 	GL_STUB(PointParameterivNV, _gloffset_PointParameterivNV, PointParameterivNV@8)
-	GL_STUB(_dispatch_stub_733, _gloffset_ActiveStencilFaceEXT, _dispatch_stub_733@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_733, _dispatch_stub_733@4))
-	GL_STUB(_dispatch_stub_734, _gloffset_BindVertexArrayAPPLE, _dispatch_stub_734@4)
+	GL_STUB(_dispatch_stub_734, _gloffset_ActiveStencilFaceEXT, _dispatch_stub_734@4)
 	HIDDEN(GL_PREFIX(_dispatch_stub_734, _dispatch_stub_734@4))
-	GL_STUB(_dispatch_stub_735, _gloffset_DeleteVertexArraysAPPLE, _dispatch_stub_735@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_735, _dispatch_stub_735@8))
-	GL_STUB(_dispatch_stub_736, _gloffset_GenVertexArraysAPPLE, _dispatch_stub_736@8)
+	GL_STUB(_dispatch_stub_735, _gloffset_BindVertexArrayAPPLE, _dispatch_stub_735@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_735, _dispatch_stub_735@4))
+	GL_STUB(_dispatch_stub_736, _gloffset_DeleteVertexArraysAPPLE, _dispatch_stub_736@8)
 	HIDDEN(GL_PREFIX(_dispatch_stub_736, _dispatch_stub_736@8))
-	GL_STUB(_dispatch_stub_737, _gloffset_IsVertexArrayAPPLE, _dispatch_stub_737@4)
-	HIDDEN(GL_PREFIX(_dispatch_stub_737, _dispatch_stub_737@4))
+	GL_STUB(_dispatch_stub_737, _gloffset_GenVertexArraysAPPLE, _dispatch_stub_737@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_737, _dispatch_stub_737@8))
+	GL_STUB(_dispatch_stub_738, _gloffset_IsVertexArrayAPPLE, _dispatch_stub_738@4)
+	HIDDEN(GL_PREFIX(_dispatch_stub_738, _dispatch_stub_738@4))
 	GL_STUB(GetProgramNamedParameterdvNV, _gloffset_GetProgramNamedParameterdvNV, GetProgramNamedParameterdvNV@16)
 	GL_STUB(GetProgramNamedParameterfvNV, _gloffset_GetProgramNamedParameterfvNV, GetProgramNamedParameterfvNV@16)
 	GL_STUB(ProgramNamedParameter4dNV, _gloffset_ProgramNamedParameter4dNV, ProgramNamedParameter4dNV@44)
 	GL_STUB(ProgramNamedParameter4dvNV, _gloffset_ProgramNamedParameter4dvNV, ProgramNamedParameter4dvNV@16)
 	GL_STUB(ProgramNamedParameter4fNV, _gloffset_ProgramNamedParameter4fNV, ProgramNamedParameter4fNV@28)
 	GL_STUB(ProgramNamedParameter4fvNV, _gloffset_ProgramNamedParameter4fvNV, ProgramNamedParameter4fvNV@16)
-	GL_STUB(_dispatch_stub_744, _gloffset_DepthBoundsEXT, _dispatch_stub_744@16)
-	HIDDEN(GL_PREFIX(_dispatch_stub_744, _dispatch_stub_744@16))
-	GL_STUB(_dispatch_stub_745, _gloffset_BlendEquationSeparateEXT, _dispatch_stub_745@8)
-	HIDDEN(GL_PREFIX(_dispatch_stub_745, _dispatch_stub_745@8))
+	GL_STUB(_dispatch_stub_745, _gloffset_DepthBoundsEXT, _dispatch_stub_745@16)
+	HIDDEN(GL_PREFIX(_dispatch_stub_745, _dispatch_stub_745@16))
+	GL_STUB(_dispatch_stub_746, _gloffset_BlendEquationSeparateEXT, _dispatch_stub_746@8)
+	HIDDEN(GL_PREFIX(_dispatch_stub_746, _dispatch_stub_746@8))
 	GL_STUB(BindFramebufferEXT, _gloffset_BindFramebufferEXT, BindFramebufferEXT@8)
 	GL_STUB(BindRenderbufferEXT, _gloffset_BindRenderbufferEXT, BindRenderbufferEXT@8)
 	GL_STUB(CheckFramebufferStatusEXT, _gloffset_CheckFramebufferStatusEXT, CheckFramebufferStatusEXT@4)
@@ -941,19 +942,19 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB(IsFramebufferEXT, _gloffset_IsFramebufferEXT, IsFramebufferEXT@4)
 	GL_STUB(IsRenderbufferEXT, _gloffset_IsRenderbufferEXT, IsRenderbufferEXT@4)
 	GL_STUB(RenderbufferStorageEXT, _gloffset_RenderbufferStorageEXT, RenderbufferStorageEXT@16)
-	GL_STUB(_dispatch_stub_763, _gloffset_BlitFramebufferEXT, _dispatch_stub_763@40)
-	HIDDEN(GL_PREFIX(_dispatch_stub_763, _dispatch_stub_763@40))
+	GL_STUB(_dispatch_stub_764, _gloffset_BlitFramebufferEXT, _dispatch_stub_764@40)
+	HIDDEN(GL_PREFIX(_dispatch_stub_764, _dispatch_stub_764@40))
 	GL_STUB(FramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT, FramebufferTextureLayerEXT@20)
-	GL_STUB(_dispatch_stub_765, _gloffset_StencilFuncSeparateATI, _dispatch_stub_765@16)
-	HIDDEN(GL_PREFIX(_dispatch_stub_765, _dispatch_stub_765@16))
-	GL_STUB(_dispatch_stub_766, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_766@16)
+	GL_STUB(_dispatch_stub_766, _gloffset_StencilFuncSeparateATI, _dispatch_stub_766@16)
 	HIDDEN(GL_PREFIX(_dispatch_stub_766, _dispatch_stub_766@16))
-	GL_STUB(_dispatch_stub_767, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_767@16)
+	GL_STUB(_dispatch_stub_767, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_767@16)
 	HIDDEN(GL_PREFIX(_dispatch_stub_767, _dispatch_stub_767@16))
-	GL_STUB(_dispatch_stub_768, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_768@12)
-	HIDDEN(GL_PREFIX(_dispatch_stub_768, _dispatch_stub_768@12))
-	GL_STUB(_dispatch_stub_769, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_769@12)
+	GL_STUB(_dispatch_stub_768, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_768@16)
+	HIDDEN(GL_PREFIX(_dispatch_stub_768, _dispatch_stub_768@16))
+	GL_STUB(_dispatch_stub_769, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_769@12)
 	HIDDEN(GL_PREFIX(_dispatch_stub_769, _dispatch_stub_769@12))
+	GL_STUB(_dispatch_stub_770, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_770@12)
+	HIDDEN(GL_PREFIX(_dispatch_stub_770, _dispatch_stub_770@12))
 	GL_STUB_ALIAS(ArrayElementEXT, _gloffset_ArrayElement, ArrayElementEXT@4, ArrayElement, ArrayElement@4)
 	GL_STUB_ALIAS(BindTextureEXT, _gloffset_BindTexture, BindTextureEXT@8, BindTexture, BindTexture@8)
 	GL_STUB_ALIAS(DrawArraysEXT, _gloffset_DrawArrays, DrawArraysEXT@12, DrawArrays, DrawArrays@12)
@@ -1269,6 +1270,24 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB_ALIAS(IsProgramARB, _gloffset_IsProgramNV, IsProgramARB@4, IsProgramNV, IsProgramNV@4)
 	GL_STUB_ALIAS(PointParameteri, _gloffset_PointParameteriNV, PointParameteri@8, PointParameteriNV, PointParameteriNV@8)
 	GL_STUB_ALIAS(PointParameteriv, _gloffset_PointParameterivNV, PointParameteriv@8, PointParameterivNV, PointParameterivNV@8)
+	GL_STUB_ALIAS(BindFramebuffer, _gloffset_BindFramebufferEXT, BindFramebuffer@8, BindFramebufferEXT, BindFramebufferEXT@8)
+	GL_STUB_ALIAS(BindRenderbuffer, _gloffset_BindRenderbufferEXT, BindRenderbuffer@8, BindRenderbufferEXT, BindRenderbufferEXT@8)
+	GL_STUB_ALIAS(CheckFramebufferStatus, _gloffset_CheckFramebufferStatusEXT, CheckFramebufferStatus@4, CheckFramebufferStatusEXT, CheckFramebufferStatusEXT@4)
+	GL_STUB_ALIAS(DeleteFramebuffers, _gloffset_DeleteFramebuffersEXT, DeleteFramebuffers@8, DeleteFramebuffersEXT, DeleteFramebuffersEXT@8)
+	GL_STUB_ALIAS(DeleteRenderbuffers, _gloffset_DeleteRenderbuffersEXT, DeleteRenderbuffers@8, DeleteRenderbuffersEXT, DeleteRenderbuffersEXT@8)
+	GL_STUB_ALIAS(FramebufferRenderbuffer, _gloffset_FramebufferRenderbufferEXT, FramebufferRenderbuffer@16, FramebufferRenderbufferEXT, FramebufferRenderbufferEXT@16)
+	GL_STUB_ALIAS(FramebufferTexture1D, _gloffset_FramebufferTexture1DEXT, FramebufferTexture1D@20, FramebufferTexture1DEXT, FramebufferTexture1DEXT@20)
+	GL_STUB_ALIAS(FramebufferTexture2D, _gloffset_FramebufferTexture2DEXT, FramebufferTexture2D@20, FramebufferTexture2DEXT, FramebufferTexture2DEXT@20)
+	GL_STUB_ALIAS(FramebufferTexture3D, _gloffset_FramebufferTexture3DEXT, FramebufferTexture3D@24, FramebufferTexture3DEXT, FramebufferTexture3DEXT@24)
+	GL_STUB_ALIAS(GenFramebuffers, _gloffset_GenFramebuffersEXT, GenFramebuffers@8, GenFramebuffersEXT, GenFramebuffersEXT@8)
+	GL_STUB_ALIAS(GenRenderbuffers, _gloffset_GenRenderbuffersEXT, GenRenderbuffers@8, GenRenderbuffersEXT, GenRenderbuffersEXT@8)
+	GL_STUB_ALIAS(GenerateMipmap, _gloffset_GenerateMipmapEXT, GenerateMipmap@4, GenerateMipmapEXT, GenerateMipmapEXT@4)
+	GL_STUB_ALIAS(GetFramebufferAttachmentParameteriv, _gloffset_GetFramebufferAttachmentParameterivEXT, GetFramebufferAttachmentParameteriv@16, GetFramebufferAttachmentParameterivEXT, GetFramebufferAttachmentParameterivEXT@16)
+	GL_STUB_ALIAS(GetRenderbufferParameteriv, _gloffset_GetRenderbufferParameterivEXT, GetRenderbufferParameteriv@12, GetRenderbufferParameterivEXT, GetRenderbufferParameterivEXT@12)
+	GL_STUB_ALIAS(IsFramebuffer, _gloffset_IsFramebufferEXT, IsFramebuffer@4, IsFramebufferEXT, IsFramebufferEXT@4)
+	GL_STUB_ALIAS(IsRenderbuffer, _gloffset_IsRenderbufferEXT, IsRenderbuffer@4, IsRenderbufferEXT, IsRenderbufferEXT@4)
+	GL_STUB_ALIAS(RenderbufferStorage, _gloffset_RenderbufferStorageEXT, RenderbufferStorage@16, RenderbufferStorageEXT, RenderbufferStorageEXT@16)
+	GL_STUB_ALIAS(FramebufferTextureLayer, _gloffset_FramebufferTextureLayerEXT, FramebufferTextureLayer@20, FramebufferTextureLayerEXT, FramebufferTextureLayerEXT@20)
 
 		GLOBL	GLNAME(gl_dispatch_functions_end)
 		HIDDEN(GLNAME(gl_dispatch_functions_end))
diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c
index e68f8dfec1..5aedf5b04b 100644
--- a/src/mesa/x86/rtasm/x86sse.c
+++ b/src/mesa/x86/rtasm/x86sse.c
@@ -1,3 +1,4 @@
+#ifdef USE_X86_ASM
 #if defined(__i386__) || defined(__386__)
 
 #include "main/imports.h"
@@ -290,7 +291,7 @@ void x86_call( struct x86_function *p, void (*label)())
 void x86_call( struct x86_function *p, struct x86_reg reg)
 {
    emit_1ub(p, 0xff);
-   emit_modrm(p, reg, reg);
+   emit_modrm_noreg(p, 2, reg);
 }
 #endif
 
@@ -1192,3 +1193,9 @@ void x86sse_dummy( void )
 }
 
 #endif
+
+#else  /* USE_X86_ASM */
+
+int x86sse_c_dummy_var; /* silence warning */
+
+#endif /* USE_X86_ASM */
diff --git a/src/xvmc/Makefile b/src/xvmc/Makefile
new file mode 100644
index 0000000000..7badcfd264
--- /dev/null
+++ b/src/xvmc/Makefile
@@ -0,0 +1,73 @@
+TARGET		= libXvMCg3dvl.so
+SONAME		= libXvMCg3dvl.so.1
+GALLIUMDIR	= ../gallium
+
+OBJECTS         = block.o surface.o context.o subpicture.o attributes.o
+
+ifeq (${DRIVER}, softpipe)
+OBJECTS         += ${GALLIUMDIR}/winsys/g3dvl/xsp_winsys.o
+endif
+
+CFLAGS	+= -g -fPIC -Wall -Werror=implicit-function-declaration	\
+	   -I${GALLIUMDIR}/state_trackers/g3dvl		\
+	   -I${GALLIUMDIR}/winsys/g3dvl			\
+	   -I${GALLIUMDIR}/include			\
+	   -I${GALLIUMDIR}/auxiliary			\
+	   -I${GALLIUMDIR}/drivers
+
+ifeq (${DRIVER}, softpipe)
+LDFLAGS	+= -L${GALLIUMDIR}/state_trackers/g3dvl		\
+	   -L${GALLIUMDIR}/drivers/softpipe		\
+	   -L${GALLIUMDIR}/auxiliary/tgsi		\
+	   -L${GALLIUMDIR}/auxiliary/draw		\
+	   -L${GALLIUMDIR}/auxiliary/translate		\
+	   -L${GALLIUMDIR}/auxiliary/cso_cache		\
+	   -L${GALLIUMDIR}/auxiliary/util		\
+	   -L${GALLIUMDIR}/auxiliary/rtasm
+else
+LDFLAGS	+= -L${GALLIUMDIR}/state_trackers/g3dvl		\
+   	   -L${GALLIUMDIR}/winsys/g3dvl/nouveau		\
+   	   -L${GALLIUMDIR}/auxiliary/util
+endif
+
+ifeq (${DRIVER}, softpipe)
+LIBS	+= -lg3dvl -lsoftpipe -ldraw -ltgsi -ltranslate -lrtasm -lcso_cache -lutil -lm
+else
+LIBS	+= -lg3dvl -lnouveau_dri -lutil
+endif
+
+#############################################
+
+ifeq (${DRIVER}, softpipe)
+.PHONY	= all clean g3dvl
+else
+.PHONY	= all clean g3dvl nouveau_winsys
+endif
+
+all: ${TARGET}
+
+ifeq (${DRIVER}, softpipe)
+${TARGET}: ${OBJECTS} g3dvl
+	$(CC) ${LDFLAGS} -shared -Wl,-soname,${SONAME} -o $@ ${OBJECTS} ${LIBS}
+
+g3dvl:
+	cd ${GALLIUMDIR}/state_trackers/g3dvl; ${MAKE}
+
+clean:
+	cd ${GALLIUMDIR}/state_trackers/g3dvl; ${MAKE} clean
+	rm -rf ${OBJECTS} ${TARGET}
+else
+${TARGET}: ${OBJECTS} g3dvl nouveau_winsys
+	$(CC) ${LDFLAGS} -shared -Wl,-soname,${SONAME} -o $@ ${OBJECTS} ${LIBS}
+
+g3dvl:
+	cd ${GALLIUMDIR}/state_trackers/g3dvl; ${MAKE}
+
+nouveau_winsys:
+	cd ${GALLIUMDIR}/winsys/g3dvl/nouveau; ${MAKE}
+
+clean:
+	cd ${GALLIUMDIR}/state_trackers/g3dvl; ${MAKE} clean
+	cd ${GALLIUMDIR}/winsys/g3dvl/nouveau; ${MAKE} clean
+	rm -rf ${OBJECTS} ${TARGET}
+endif
diff --git a/src/xvmc/attributes.c b/src/xvmc/attributes.c
new file mode 100644
index 0000000000..674524b8b8
--- /dev/null
+++ b/src/xvmc/attributes.c
@@ -0,0 +1,20 @@
+#include <assert.h>
+#include <X11/Xlib.h>
+#include <X11/extensions/Xvlib.h>
+#include <X11/extensions/XvMC.h>
+
+XvAttribute* XvMCQueryAttributes(Display *display, XvMCContext *context, int *number)
+{
+	return NULL;
+}
+
+Status XvMCSetAttribute(Display *display, XvMCContext *context, Atom attribute, int value)
+{
+	return BadImplementation;
+}
+
+Status XvMCGetAttribute(Display *display, XvMCContext *context, Atom attribute, int *value)
+{
+	return BadImplementation;
+}
+
diff --git a/src/xvmc/block.c b/src/xvmc/block.c
new file mode 100644
index 0000000000..b38a89be09
--- /dev/null
+++ b/src/xvmc/block.c
@@ -0,0 +1,79 @@
+#include <assert.h>
+#include <X11/Xlib.h>
+#include <X11/extensions/XvMC.h>
+#include <util/u_memory.h>
+#include <vl_display.h>
+#include <vl_screen.h>
+#include <vl_context.h>
+
+#define BLOCK_SIZE (64 * 2)
+
+Status XvMCCreateBlocks(Display *display, XvMCContext *context, unsigned int num_blocks, XvMCBlockArray *blocks)
+{
+	struct vlContext *vl_ctx;
+
+	assert(display);
+
+	if (!context)
+		return XvMCBadContext;
+	if (num_blocks == 0)
+		return BadValue;
+
+	assert(blocks);
+
+	vl_ctx = context->privData;
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlContextGetScreen(vl_ctx))));
+
+	blocks->context_id = context->context_id;
+	blocks->num_blocks = num_blocks;
+	blocks->blocks = MALLOC(BLOCK_SIZE * num_blocks);
+	/* Since we don't have a VL type for blocks, set privData to the display so we can catch mismatches */
+	blocks->privData = display;
+
+	return Success;
+}
+
+Status XvMCDestroyBlocks(Display *display, XvMCBlockArray *blocks)
+{
+	assert(display);
+	assert(blocks);
+	assert(display == blocks->privData);
+	FREE(blocks->blocks);
+
+	return Success;
+}
+
+Status XvMCCreateMacroBlocks(Display *display, XvMCContext *context, unsigned int num_blocks, XvMCMacroBlockArray *blocks)
+{
+	struct vlContext *vl_ctx;
+
+	assert(display);
+
+	if (!context)
+		return XvMCBadContext;
+	if (num_blocks == 0)
+		return BadValue;
+
+	assert(blocks);
+
+	vl_ctx = context->privData;
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlContextGetScreen(vl_ctx))));
+
+	blocks->context_id = context->context_id;
+	blocks->num_blocks = num_blocks;
+	blocks->macro_blocks = MALLOC(sizeof(XvMCMacroBlock) * num_blocks);
+	/* Since we don't have a VL type for blocks, set privData to the display so we can catch mismatches */
+	blocks->privData = display;
+
+	return Success;
+}
+
+Status XvMCDestroyMacroBlocks(Display *display, XvMCMacroBlockArray *blocks)
+{
+	assert(display);
+	assert(blocks);
+	assert(display == blocks->privData);
+	FREE(blocks->macro_blocks);
+
+	return Success;
+}
diff --git a/src/xvmc/context.c b/src/xvmc/context.c
new file mode 100644
index 0000000000..273f658029
--- /dev/null
+++ b/src/xvmc/context.c
@@ -0,0 +1,207 @@
+#include <assert.h>
+#include <X11/Xlib.h>
+#include <X11/extensions/XvMClib.h>
+#include <pipe/p_context.h>
+#include <vl_display.h>
+#include <vl_screen.h>
+#include <vl_context.h>
+#include <vl_winsys.h>
+
+static Status Validate
+(
+	Display *display,
+	XvPortID port,
+	int surface_type_id,
+	unsigned int width,
+	unsigned int height,
+	int flags,
+	int *found_port,
+	int *chroma_format,
+	int *mc_type
+)
+{
+	unsigned int	found_surface = 0;
+	XvAdaptorInfo	*adaptor_info;
+	unsigned int	num_adaptors;
+	int		num_types;
+	unsigned int	max_width, max_height;
+	Status		ret;
+	unsigned int	i, j, k;
+
+	assert(display && chroma_format);
+
+	*found_port = 0;
+
+	ret = XvQueryAdaptors(display, XDefaultRootWindow(display), &num_adaptors, &adaptor_info);
+	if (ret != Success)
+		return ret;
+
+	/* Scan through all adaptors looking for this port and surface */
+	for (i = 0; i < num_adaptors && !*found_port; ++i)
+	{
+		/* Scan through all ports of this adaptor looking for our port */
+		for (j = 0; j < adaptor_info[i].num_ports && !*found_port; ++j)
+		{
+			/* If this is our port, scan through all its surfaces looking for our surface */
+			if (adaptor_info[i].base_id + j == port)
+			{
+				XvMCSurfaceInfo *surface_info;
+
+				*found_port = 1;
+				surface_info = XvMCListSurfaceTypes(display, adaptor_info[i].base_id, &num_types);
+
+				if (surface_info)
+				{
+					for (k = 0; k < num_types && !found_surface; ++k)
+					{
+						if (surface_info[k].surface_type_id == surface_type_id)
+						{
+							found_surface = 1;
+							max_width = surface_info[k].max_width;
+							max_height = surface_info[k].max_height;
+							*chroma_format = surface_info[k].chroma_format;
+							*mc_type = surface_info[k].mc_type;
+						}
+					}
+
+					XFree(surface_info);
+				}
+				else
+				{
+					XvFreeAdaptorInfo(adaptor_info);
+					return BadAlloc;
+				}
+			}
+		}
+	}
+
+	XvFreeAdaptorInfo(adaptor_info);
+
+	if (!*found_port)
+		return XvBadPort;
+	if (!found_surface)
+		return BadMatch;
+	if (width > max_width || height > max_height)
+		return BadValue;
+	if (flags != XVMC_DIRECT && flags != 0)
+		return BadValue;
+
+	return Success;
+}
+
+static enum vlProfile ProfileToVL(int xvmc_profile)
+{
+	if (xvmc_profile & XVMC_MPEG_1)
+		assert(0);
+	else if (xvmc_profile & XVMC_MPEG_2)
+		return vlProfileMpeg2Main;
+	else if (xvmc_profile & XVMC_H263)
+		assert(0);
+	else if (xvmc_profile & XVMC_MPEG_4)
+		assert(0);
+	else
+		assert(0);
+
+	return -1;
+}
+
+static enum vlEntryPoint EntryToVL(int xvmc_entry)
+{
+	return xvmc_entry & XVMC_IDCT ? vlEntryPointIDCT : vlEntryPointMC;
+}
+
+static enum vlFormat FormatToVL(int xvmc_format)
+{
+	switch (xvmc_format)
+	{
+		case XVMC_CHROMA_FORMAT_420:
+			return vlFormatYCbCr420;
+		case XVMC_CHROMA_FORMAT_422:
+			return vlFormatYCbCr422;
+		case XVMC_CHROMA_FORMAT_444:
+			return vlFormatYCbCr444;
+		default:
+			assert(0);
+	}
+
+	return -1;
+}
+
+Status XvMCCreateContext(Display *display, XvPortID port, int surface_type_id, int width, int height, int flags, XvMCContext *context)
+{
+	int			found_port;
+	int			chroma_format;
+	int			mc_type;
+	Status			ret;
+	struct vlDisplay	*vl_dpy;
+	struct vlScreen		*vl_scrn;
+	struct vlContext	*vl_ctx;
+	struct pipe_context	*pipe;
+
+	assert(display);
+
+	if (!context)
+		return XvMCBadContext;
+
+	ret = Validate(display, port, surface_type_id, width, height, flags, &found_port, &chroma_format, &mc_type);
+
+	/* XXX: Success and XvBadPort have the same value */
+	if (ret != Success || !found_port)
+		return ret;
+
+	/* XXX: Assumes default screen, should check which screen port is on */
+	pipe = create_pipe_context(display, XDefaultScreen(display));
+
+	assert(pipe);
+
+	vlCreateDisplay(display, &vl_dpy);
+	vlCreateScreen(vl_dpy, XDefaultScreen(display), pipe->screen, &vl_scrn);
+	vlCreateContext
+	(
+		vl_scrn,
+		pipe,
+		width,
+		height,
+		FormatToVL(chroma_format),
+		ProfileToVL(mc_type),
+		EntryToVL(mc_type),
+		&vl_ctx
+	);
+
+	context->context_id = XAllocID(display);
+	context->surface_type_id = surface_type_id;
+	context->width = width;
+	context->height = height;
+	context->flags = flags;
+	context->port = port;
+	context->privData = vl_ctx;
+
+	return Success;
+}
+
+Status XvMCDestroyContext(Display *display, XvMCContext *context)
+{
+	struct vlContext	*vl_ctx;
+	struct vlScreen		*vl_screen;
+	struct vlDisplay	*vl_dpy;
+	struct pipe_context	*pipe;
+
+	assert(display);
+
+	if (!context)
+		return XvMCBadContext;
+
+	vl_ctx = context->privData;
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlContextGetScreen(vl_ctx))));
+
+	pipe = vlGetPipeContext(vl_ctx);
+	vl_screen = vlContextGetScreen(vl_ctx);
+	vl_dpy = vlGetDisplay(vl_screen);
+	vlDestroyContext(vl_ctx);
+	vlDestroyScreen(vl_screen);
+	vlDestroyDisplay(vl_dpy);
+	destroy_pipe_context(pipe);
+
+	return Success;
+}
diff --git a/src/xvmc/subpicture.c b/src/xvmc/subpicture.c
new file mode 100644
index 0000000000..c8f70c90d0
--- /dev/null
+++ b/src/xvmc/subpicture.c
@@ -0,0 +1,215 @@
+#include <assert.h>
+#include <X11/Xlib.h>
+#include <X11/extensions/Xvlib.h>
+#include <X11/extensions/XvMC.h>
+
+Status XvMCCreateSubpicture
+(
+	Display *display,
+	XvMCContext *context,
+	XvMCSubpicture *subpicture,
+	unsigned short width,
+	unsigned short height,
+	int xvimage_id
+)
+{
+	assert(display);
+	
+	if (!context)
+		return XvMCBadContext;
+	
+	assert(subpicture);
+	
+	if (width > 2048 || height > 2048)
+		return BadValue;
+	
+	if (xvimage_id != 123)
+		return BadMatch;
+	
+	subpicture->subpicture_id = XAllocID(display);
+	subpicture->context_id = context->context_id;
+	subpicture->xvimage_id = xvimage_id;
+	subpicture->width = width;
+	subpicture->height = height;
+	subpicture->num_palette_entries = 0;
+	subpicture->entry_bytes = 0;
+	subpicture->component_order[0] = 0;
+	subpicture->component_order[1] = 0;
+	subpicture->component_order[2] = 0;
+	subpicture->component_order[3] = 0;
+	/* TODO: subpicture->privData = ;*/
+	
+	return Success;
+}
+
+Status XvMCClearSubpicture
+(
+	Display *display,
+	XvMCSubpicture *subpicture,
+	short x,
+	short y,
+	unsigned short width,
+	unsigned short height,
+	unsigned int color
+)
+{
+	assert(display);
+	
+	if (!subpicture)
+		return XvMCBadSubpicture;
+	
+	/* TODO: Assert clear rect is within bounds? Or clip? */
+	
+	return Success;
+}
+
+Status XvMCCompositeSubpicture
+(
+	Display *display,
+	XvMCSubpicture *subpicture,
+	XvImage *image,
+	short srcx,
+	short srcy,
+	unsigned short width,
+	unsigned short height,
+	short dstx,
+	short dsty
+)
+{
+	assert(display);
+	
+	if (!subpicture)
+		return XvMCBadSubpicture;
+	
+	assert(image);
+	
+	if (subpicture->xvimage_id != image->id)
+		return BadMatch;
+	
+	/* TODO: Assert rects are within bounds? Or clip? */
+	
+	return Success;
+}
+
+Status XvMCDestroySubpicture(Display *display, XvMCSubpicture *subpicture)
+{
+	assert(display);
+	
+	if (!subpicture)
+		return XvMCBadSubpicture;
+	
+	return BadImplementation;
+}
+
+Status XvMCSetSubpicturePalette(Display *display, XvMCSubpicture *subpicture, unsigned char *palette)
+{
+	assert(display);
+	
+	if (!subpicture)
+		return XvMCBadSubpicture;
+	
+	assert(palette);
+	
+	/* We don't support paletted subpictures */
+	return BadMatch;
+}
+
+Status XvMCBlendSubpicture
+(
+	Display *display,
+	XvMCSurface *target_surface,
+	XvMCSubpicture *subpicture,
+	short subx,
+	short suby,
+	unsigned short subw,
+	unsigned short subh,
+	short surfx,
+	short surfy,
+	unsigned short surfw,
+	unsigned short surfh
+)
+{
+	assert(display);
+	
+	if (!target_surface)
+		return XvMCBadSurface;
+	
+	if (!subpicture)
+		return XvMCBadSubpicture;
+	
+	if (target_surface->context_id != subpicture->context_id)
+		return BadMatch;
+	
+	/* TODO: Assert rects are within bounds? Or clip? */
+	return Success;
+}
+
+Status XvMCBlendSubpicture2
+(
+	Display *display,
+	XvMCSurface *source_surface,
+	XvMCSurface *target_surface,
+	XvMCSubpicture *subpicture,
+	short subx,
+	short suby,
+	unsigned short subw,
+	unsigned short subh,
+	short surfx,
+	short surfy,
+	unsigned short surfw,
+	unsigned short surfh
+)
+{
+	assert(display);
+	
+	if (!source_surface || !target_surface)
+		return XvMCBadSurface;
+	
+	if (!subpicture)
+		return XvMCBadSubpicture;
+	
+	if (source_surface->context_id != subpicture->context_id)
+		return BadMatch;
+		
+	if (source_surface->context_id != subpicture->context_id)
+		return BadMatch;
+	
+	/* TODO: Assert rects are within bounds? Or clip? */
+	return Success;
+}
+
+Status XvMCSyncSubpicture(Display *display, XvMCSubpicture *subpicture)
+{
+	assert(display);
+	
+	if (!subpicture)
+		return XvMCBadSubpicture;
+	
+	return Success;
+}
+
+Status XvMCFlushSubpicture(Display *display, XvMCSubpicture *subpicture)
+{
+	assert(display);
+	
+	if (!subpicture)
+		return XvMCBadSubpicture;
+	
+	return Success;
+}
+
+Status XvMCGetSubpictureStatus(Display *display, XvMCSubpicture *subpicture, int *status)
+{
+	assert(display);
+	
+	if (!subpicture)
+		return XvMCBadSubpicture;
+	
+	assert(status);
+	
+	/* TODO */
+	*status = 0;
+	
+	return Success;
+}
+
diff --git a/src/xvmc/surface.c b/src/xvmc/surface.c
new file mode 100644
index 0000000000..7c5f45bd34
--- /dev/null
+++ b/src/xvmc/surface.c
@@ -0,0 +1,355 @@
+#include <assert.h>
+#include <X11/Xlib.h>
+#include <X11/extensions/XvMC.h>
+#include <vl_display.h>
+#include <vl_screen.h>
+#include <vl_context.h>
+#include <vl_surface.h>
+#include <vl_types.h>
+
+static enum vlMacroBlockType TypeToVL(int xvmc_mb_type)
+{
+	if (xvmc_mb_type & XVMC_MB_TYPE_INTRA)
+		return vlMacroBlockTypeIntra;
+	if ((xvmc_mb_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)) == XVMC_MB_TYPE_MOTION_FORWARD)
+		return vlMacroBlockTypeFwdPredicted;
+	if ((xvmc_mb_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)) == XVMC_MB_TYPE_MOTION_BACKWARD)
+		return vlMacroBlockTypeBkwdPredicted;
+	if ((xvmc_mb_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)) == (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD))
+		return vlMacroBlockTypeBiPredicted;
+
+	assert(0);
+
+	return -1;
+}
+
+static enum vlPictureType PictureToVL(int xvmc_pic)
+{
+	switch (xvmc_pic)
+	{
+		case XVMC_TOP_FIELD:
+			return vlPictureTypeTopField;
+		case XVMC_BOTTOM_FIELD:
+			return vlPictureTypeBottomField;
+		case XVMC_FRAME_PICTURE:
+			return vlPictureTypeFrame;
+		default:
+			assert(0);
+	}
+
+	return -1;
+}
+
+static enum vlMotionType MotionToVL(int xvmc_motion_type, int xvmc_dct_type)
+{
+	switch (xvmc_motion_type)
+	{
+		case XVMC_PREDICTION_FRAME:
+			return xvmc_dct_type == XVMC_DCT_TYPE_FIELD ? vlMotionType16x8 : vlMotionTypeFrame;
+		case XVMC_PREDICTION_FIELD:
+			return vlMotionTypeField;
+		case XVMC_PREDICTION_DUAL_PRIME:
+			return vlMotionTypeDualPrime;
+		default:
+			assert(0);
+	}
+
+	return -1;
+}
+
+Status XvMCCreateSurface(Display *display, XvMCContext *context, XvMCSurface *surface)
+{
+	struct vlContext *vl_ctx;
+	struct vlSurface *vl_sfc;
+
+	assert(display);
+
+	if (!context)
+		return XvMCBadContext;
+	if (!surface)
+		return XvMCBadSurface;
+
+	vl_ctx = context->privData;
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlContextGetScreen(vl_ctx))));
+
+	if (vlCreateSurface(vlContextGetScreen(vl_ctx),
+	                    context->width, context->height,
+	                    vlGetPictureFormat(vl_ctx),
+	                    &vl_sfc))
+	{
+		return BadAlloc;
+	}
+
+	vlBindToContext(vl_sfc, vl_ctx);
+
+	surface->surface_id = XAllocID(display);
+	surface->context_id = context->context_id;
+	surface->surface_type_id = context->surface_type_id;
+	surface->width = context->width;
+	surface->height = context->height;
+	surface->privData = vl_sfc;
+
+	return Success;
+}
+
+Status XvMCRenderSurface
+(
+	Display *display,
+	XvMCContext *context,
+	unsigned int picture_structure,
+	XvMCSurface *target_surface,
+	XvMCSurface *past_surface,
+	XvMCSurface *future_surface,
+	unsigned int flags,
+	unsigned int num_macroblocks,
+	unsigned int first_macroblock,
+	XvMCMacroBlockArray *macroblocks,
+	XvMCBlockArray *blocks
+)
+{
+	struct vlContext		*vl_ctx;
+	struct vlSurface		*target_vl_surface;
+	struct vlSurface		*past_vl_surface;
+	struct vlSurface		*future_vl_surface;
+	struct vlMpeg2MacroBlockBatch	batch;
+	struct vlMpeg2MacroBlock	vl_macroblocks[num_macroblocks];
+	unsigned int			i;
+
+	assert(display);
+
+	if (!context)
+		return XvMCBadContext;
+	if (!target_surface)
+		return XvMCBadSurface;
+
+	if
+	(
+		picture_structure != XVMC_TOP_FIELD &&
+		picture_structure != XVMC_BOTTOM_FIELD &&
+		picture_structure != XVMC_FRAME_PICTURE
+	)
+		return BadValue;
+	if (future_surface && !past_surface)
+		return BadMatch;
+
+	vl_ctx = context->privData;
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlContextGetScreen(vl_ctx))));
+
+	target_vl_surface = target_surface->privData;
+	past_vl_surface = past_surface ? past_surface->privData : NULL;
+	future_vl_surface = future_surface ? future_surface->privData : NULL;
+
+	assert(context->context_id == target_surface->context_id);
+	assert(!past_surface || context->context_id == past_surface->context_id);
+	assert(!future_surface || context->context_id == future_surface->context_id);
+
+	assert(macroblocks);
+	assert(blocks);
+
+	assert(macroblocks->context_id == context->context_id);
+	assert(blocks->context_id == context->context_id);
+
+	assert(flags == 0 || flags == XVMC_SECOND_FIELD);
+
+	batch.past_surface = past_vl_surface;
+	batch.future_surface = future_vl_surface;
+	batch.picture_type = PictureToVL(picture_structure);
+	batch.field_order = flags & XVMC_SECOND_FIELD ? vlFieldOrderSecond : vlFieldOrderFirst;
+	batch.num_macroblocks = num_macroblocks;
+	batch.macroblocks = vl_macroblocks;
+
+	for (i = 0; i < num_macroblocks; ++i)
+	{
+		unsigned int j = first_macroblock + i;
+
+		unsigned int k, l, m;
+
+		batch.macroblocks[i].mbx = macroblocks->macro_blocks[j].x;
+		batch.macroblocks[i].mby = macroblocks->macro_blocks[j].y;
+		batch.macroblocks[i].mb_type = TypeToVL(macroblocks->macro_blocks[j].macroblock_type);
+		if (batch.macroblocks[i].mb_type != vlMacroBlockTypeIntra)
+			batch.macroblocks[i].mo_type = MotionToVL(macroblocks->macro_blocks[j].motion_type, macroblocks->macro_blocks[j].dct_type);
+		batch.macroblocks[i].dct_type = macroblocks->macro_blocks[j].dct_type == XVMC_DCT_TYPE_FIELD ? vlDCTTypeFieldCoded : vlDCTTypeFrameCoded;
+
+		for (k = 0; k < 2; ++k)
+			for (l = 0; l < 2; ++l)
+				for (m = 0; m < 2; ++m)
+					batch.macroblocks[i].PMV[k][l][m] = macroblocks->macro_blocks[j].PMV[k][l][m];
+
+		batch.macroblocks[i].cbp = macroblocks->macro_blocks[j].coded_block_pattern;
+		batch.macroblocks[i].blocks = blocks->blocks + (macroblocks->macro_blocks[j].index * 64);
+	}
+
+	vlRenderMacroBlocksMpeg2(&batch, target_vl_surface);
+
+	return Success;
+}
+
+Status XvMCFlushSurface(Display *display, XvMCSurface *surface)
+{
+	struct vlSurface *vl_sfc;
+
+	assert(display);
+
+	if (!surface)
+		return XvMCBadSurface;
+
+	vl_sfc = surface->privData;
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlSurfaceGetScreen(vl_sfc))));
+
+	vlSurfaceFlush(vl_sfc);
+
+	return Success;
+}
+
+Status XvMCSyncSurface(Display *display, XvMCSurface *surface)
+{
+	struct vlSurface *vl_sfc;
+
+	assert(display);
+
+	if (!surface)
+		return XvMCBadSurface;
+
+	vl_sfc = surface->privData;
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlSurfaceGetScreen(vl_sfc))));
+
+	vlSurfaceSync(vl_sfc);
+
+	return Success;
+}
+
+Status XvMCPutSurface
+(
+	Display *display,
+	XvMCSurface *surface,
+	Drawable drawable,
+	short srcx,
+	short srcy,
+	unsigned short srcw,
+	unsigned short srch,
+	short destx,
+	short desty,
+	unsigned short destw,
+	unsigned short desth,
+	int flags
+)
+{
+	Window			root;
+	int			x, y;
+	unsigned int		width, height;
+	unsigned int		border_width;
+	unsigned int		depth;
+	struct vlSurface	*vl_sfc;
+
+	assert(display);
+
+	if (!surface)
+		return XvMCBadSurface;
+
+	if (XGetGeometry(display, drawable, &root, &x, &y, &width, &height, &border_width, &depth) == BadDrawable)
+		return BadDrawable;
+
+	assert(flags == XVMC_TOP_FIELD || flags == XVMC_BOTTOM_FIELD || flags == XVMC_FRAME_PICTURE);
+
+	/* TODO: Correct for negative srcx,srcy & destx,desty by clipping */
+
+	assert(srcx + srcw - 1 < surface->width);
+	assert(srcy + srch - 1 < surface->height);
+	/* XXX: Some apps (mplayer) hit these asserts because they call
+	 * this function after the window has been resized by the WM
+	 * but before they've handled the corresponding XEvent and
+	 * know about the new dimensions. The output will be clipped
+	 * for a few frames until the app updates destw and desth.
+	 */
+	/*assert(destx + destw - 1 < width);
+	assert(desty + desth - 1 < height);*/
+
+	vl_sfc = surface->privData;
+
+	vlPutPicture(vl_sfc, drawable, srcx, srcy, srcw, srch, destx, desty, destw, desth, width, height, PictureToVL(flags));
+
+	return Success;
+}
+
+Status XvMCGetSurfaceStatus(Display *display, XvMCSurface *surface, int *status)
+{
+	struct vlSurface	*vl_sfc;
+	enum vlResourceStatus	res_status;
+
+	assert(display);
+
+	if (!surface)
+		return XvMCBadSurface;
+
+	assert(status);
+
+	vl_sfc = surface->privData;
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlSurfaceGetScreen(vl_sfc))));
+
+	vlSurfaceGetStatus(vl_sfc, &res_status);
+
+	switch (res_status)
+	{
+		case vlResourceStatusFree:
+		{
+			*status = 0;
+			break;
+		}
+		case vlResourceStatusRendering:
+		{
+			*status = XVMC_RENDERING;
+			break;
+		}
+		case vlResourceStatusDisplaying:
+		{
+			*status = XVMC_DISPLAYING;
+			break;
+		}
+		default:
+			assert(0);
+	}
+
+	return Success;
+}
+
+Status XvMCDestroySurface(Display *display, XvMCSurface *surface)
+{
+	struct vlSurface *vl_sfc;
+
+	assert(display);
+
+	if (!surface)
+		return XvMCBadSurface;
+
+	vl_sfc = surface->privData;
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlSurfaceGetScreen(vl_sfc))));
+
+	vlDestroySurface(vl_sfc);
+
+	return Success;
+}
+
+Status XvMCHideSurface(Display *display, XvMCSurface *surface)
+{
+	struct vlSurface *vl_sfc;
+
+	assert(display);
+
+	if (!surface)
+		return XvMCBadSurface;
+
+	vl_sfc = surface->privData;
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlSurfaceGetScreen(vl_sfc))));
+
+	/* No op, only for overlaid rendering */
+
+	return Success;
+}
diff --git a/src/xvmc/tests/.gitignore b/src/xvmc/tests/.gitignore
new file mode 100644
index 0000000000..e1d2f9023d
--- /dev/null
+++ b/src/xvmc/tests/.gitignore
@@ -0,0 +1,5 @@
+test_context
+test_surface
+test_blocks
+test_rendering
+xvmc_bench
diff --git a/src/xvmc/tests/Makefile b/src/xvmc/tests/Makefile
new file mode 100644
index 0000000000..de095161d2
--- /dev/null
+++ b/src/xvmc/tests/Makefile
@@ -0,0 +1,27 @@
+CFLAGS	+= -g -Wall
+LDFLAGS	+=
+LIBS	+= -lXvMCW -lXvMC -lXv
+
+#############################################
+
+.PHONY	= all clean
+
+all: test_context test_surface test_blocks test_rendering xvmc_bench
+
+test_context: test_context.o testlib.o
+	$(CC) ${LDFLAGS} -o $@ $^ ${LIBS}
+
+test_surface: test_surface.o testlib.o
+	$(CC) ${LDFLAGS} -o $@ $^ ${LIBS}
+
+test_blocks: test_blocks.o testlib.o
+	$(CC) ${LDFLAGS} -o $@ $^ ${LIBS}
+
+test_rendering: test_rendering.o testlib.o
+	$(CC) ${LDFLAGS} -o $@ $^ ${LIBS}
+
+xvmc_bench: xvmc_bench.o testlib.o
+	$(CC) ${LDFLAGS} -o $@ $^ ${LIBS}
+
+clean:
+	rm -rf *.o test_context test_surface test_blocks test_rendering xvmc_bench
diff --git a/src/xvmc/tests/test_blocks.c b/src/xvmc/tests/test_blocks.c
new file mode 100644
index 0000000000..dc80adfa65
--- /dev/null
+++ b/src/xvmc/tests/test_blocks.c
@@ -0,0 +1,84 @@
+#include <assert.h>
+#include <error.h>
+#include "testlib.h"
+
+int main(int argc, char **argv)
+{
+	const unsigned int	width = 16, height = 16;
+	const unsigned int	min_required_blocks = 1, min_required_macroblocks = 1;
+	const unsigned int	mc_types[2] = {XVMC_MOCOMP | XVMC_MPEG_2, XVMC_IDCT | XVMC_MPEG_2};
+
+	Display			*display;
+	XvPortID		port_num;
+	int			surface_type_id;
+	unsigned int		is_overlay, intra_unsigned;
+	int			colorkey;
+	XvMCContext		context;
+	XvMCSurface		surface;
+	XvMCBlockArray		blocks = {0};
+	XvMCMacroBlockArray	macroblocks = {0};
+
+	display = XOpenDisplay(NULL);
+
+	if (!GetPort
+	(
+		display,
+		width,
+		height,
+		XVMC_CHROMA_FORMAT_420,
+    		mc_types,
+    		2,
+    		&port_num,
+    		&surface_type_id,
+    		&is_overlay,
+    		&intra_unsigned
+	))
+	{
+		XCloseDisplay(display);
+		error(1, 0, "Error, unable to find a good port.\n");
+	}
+
+	if (is_overlay)
+	{
+		Atom xv_colorkey = XInternAtom(display, "XV_COLORKEY", 0);
+		XvGetPortAttribute(display, port_num, xv_colorkey, &colorkey);
+	}
+
+	assert(XvMCCreateContext(display, port_num, surface_type_id, width, height, XVMC_DIRECT, &context) == Success);
+	assert(XvMCCreateSurface(display, &context, &surface) == Success);
+
+	/* Test NULL context */
+	assert(XvMCCreateBlocks(display, NULL, 1, &blocks) == XvMCBadContext);
+	/* Test 0 blocks */
+	assert(XvMCCreateBlocks(display, &context, 0, &blocks) == BadValue);
+	/* Test valid params */
+	assert(XvMCCreateBlocks(display, &context, min_required_blocks, &blocks) == Success);
+	/* Test context id assigned and correct */
+	assert(blocks.context_id == context.context_id);
+	/* Test number of blocks assigned and correct */
+	assert(blocks.num_blocks == min_required_blocks);
+	/* Test block pointer valid */
+	assert(blocks.blocks != NULL);
+	/* Test NULL context */
+	assert(XvMCCreateMacroBlocks(display, NULL, 1, &macroblocks) == XvMCBadContext);
+	/* Test 0 macroblocks */
+	assert(XvMCCreateMacroBlocks(display, &context, 0, &macroblocks) == BadValue);
+	/* Test valid params */
+	assert(XvMCCreateMacroBlocks(display, &context, min_required_macroblocks, &macroblocks) == Success);
+	/* Test context id assigned and correct */
+	assert(macroblocks.context_id == context.context_id);
+	/* Test macroblock pointer valid */
+	assert(macroblocks.macro_blocks != NULL);
+	/* Test valid params */
+	assert(XvMCDestroyMacroBlocks(display, &macroblocks) == Success);
+	/* Test valid params */
+	assert(XvMCDestroyBlocks(display, &blocks) == Success);
+
+	assert(XvMCDestroySurface(display, &surface) == Success);
+	assert(XvMCDestroyContext(display, &context) == Success);
+
+	XvUngrabPort(display, port_num, CurrentTime);
+	XCloseDisplay(display);
+
+	return 0;
+}
diff --git a/src/xvmc/tests/test_context.c b/src/xvmc/tests/test_context.c
new file mode 100644
index 0000000000..53f7449cd0
--- /dev/null
+++ b/src/xvmc/tests/test_context.c
@@ -0,0 +1,92 @@
+#include <assert.h>
+#include <error.h>
+#include "testlib.h"
+
+int main(int argc, char **argv)
+{
+	const unsigned int	width = 16, height = 16;
+	const unsigned int	mc_types[2] = {XVMC_MOCOMP | XVMC_MPEG_2, XVMC_IDCT | XVMC_MPEG_2};
+
+	Display			*display;
+	XvPortID		port_num;
+	int			surface_type_id;
+	unsigned int		is_overlay, intra_unsigned;
+	int			colorkey;
+	XvMCContext		context = {0};
+
+	display = XOpenDisplay(NULL);
+
+	if (!GetPort
+	(
+		display,
+		width,
+		height,
+		XVMC_CHROMA_FORMAT_420,
+    		mc_types,
+    		2,
+    		&port_num,
+    		&surface_type_id,
+    		&is_overlay,
+    		&intra_unsigned
+	))
+	{
+		XCloseDisplay(display);
+		error(1, 0, "Error, unable to find a good port.\n");
+	}
+
+	if (is_overlay)
+	{
+		Atom xv_colorkey = XInternAtom(display, "XV_COLORKEY", 0);
+		XvGetPortAttribute(display, port_num, xv_colorkey, &colorkey);
+	}
+
+	/* Test NULL context */
+	/* XXX: XvMCBadContext not a valid return for XvMCCreateContext in the XvMC API, but openChrome driver returns it */
+	assert(XvMCCreateContext(display, port_num, surface_type_id, width, height, XVMC_DIRECT, NULL) == XvMCBadContext);
+	/* Test invalid port */
+	/* XXX: Success and XvBadPort have the same value, if this call actually gets passed the validation step as of now we'll crash later */
+	assert(XvMCCreateContext(display, -1, surface_type_id, width, height, XVMC_DIRECT, &context) == XvBadPort);
+	/* Test invalid surface */
+	assert(XvMCCreateContext(display, port_num, -1, width, height, XVMC_DIRECT, &context) == BadMatch);
+	/* Test invalid flags */
+	assert(XvMCCreateContext(display, port_num, surface_type_id, width, height, -1, &context) == BadValue);
+	/* Test huge width */
+	assert(XvMCCreateContext(display, port_num, surface_type_id, 16384, height, XVMC_DIRECT, &context) == BadValue);
+	/* Test huge height */
+	assert(XvMCCreateContext(display, port_num, surface_type_id, width, 16384, XVMC_DIRECT, &context) == BadValue);
+	/* Test huge width & height */
+	assert(XvMCCreateContext(display, port_num, surface_type_id, 16384, 16384, XVMC_DIRECT, &context) == BadValue);
+	/* Test valid params */
+	assert(XvMCCreateContext(display, port_num, surface_type_id, width, height, XVMC_DIRECT, &context) == Success);
+	/* Test context id assigned */
+	assert(context.context_id != 0);
+	/* Test surface type id assigned and correct */
+	assert(context.surface_type_id == surface_type_id);
+	/* Test width & height assigned and correct */
+	assert(context.width == width && context.height == height);
+	/* Test port assigned and correct */
+	assert(context.port == port_num);
+	/* Test flags assigned and correct */
+	assert(context.flags == XVMC_DIRECT);
+	/* Test NULL context */
+	assert(XvMCDestroyContext(display, NULL) == XvMCBadContext);
+	/* Test valid params */
+	assert(XvMCDestroyContext(display, &context) == Success);
+	/* Test awkward but valid width */
+	assert(XvMCCreateContext(display, port_num, surface_type_id, width + 1, height, XVMC_DIRECT, &context) == Success);
+	assert(context.width >= width + 1);
+	assert(XvMCDestroyContext(display, &context) == Success);
+	/* Test awkward but valid height */
+	assert(XvMCCreateContext(display, port_num, surface_type_id, width, height + 1, XVMC_DIRECT, &context) == Success);
+	assert(context.height >= height + 1);
+	assert(XvMCDestroyContext(display, &context) == Success);
+	/* Test awkward but valid width & height */
+	assert(XvMCCreateContext(display, port_num, surface_type_id, width + 1, height + 1, XVMC_DIRECT, &context) == Success);
+	assert(context.width >= width + 1 && context.height >= height + 1);
+	assert(XvMCDestroyContext(display, &context) == Success);
+
+	XvUngrabPort(display, port_num, CurrentTime);
+	XCloseDisplay(display);
+
+	return 0;
+}
diff --git a/src/xvmc/tests/test_rendering.c b/src/xvmc/tests/test_rendering.c
new file mode 100644
index 0000000000..1e7467a3aa
--- /dev/null
+++ b/src/xvmc/tests/test_rendering.c
@@ -0,0 +1,287 @@
+#include <assert.h>
+#include <stdio.h>
+#include <string.h>
+#include <error.h>
+#include "testlib.h"
+
+#define BLOCK_WIDTH			8
+#define BLOCK_HEIGHT			8
+#define BLOCK_SIZE			(BLOCK_WIDTH * BLOCK_HEIGHT)
+#define MACROBLOCK_WIDTH		16
+#define MACROBLOCK_HEIGHT		16
+#define MACROBLOCK_WIDTH_IN_BLOCKS	(MACROBLOCK_WIDTH / BLOCK_WIDTH)
+#define MACROBLOCK_HEIGHT_IN_BLOCKS	(MACROBLOCK_HEIGHT / BLOCK_HEIGHT)
+#define BLOCKS_PER_MACROBLOCK		6
+
+#define INPUT_WIDTH			16
+#define INPUT_HEIGHT			16
+#define INPUT_WIDTH_IN_MACROBLOCKS	(INPUT_WIDTH / MACROBLOCK_WIDTH)
+#define INPUT_HEIGHT_IN_MACROBLOCKS	(INPUT_HEIGHT / MACROBLOCK_HEIGHT)
+#define NUM_MACROBLOCKS			(INPUT_WIDTH_IN_MACROBLOCKS * INPUT_HEIGHT_IN_MACROBLOCKS)
+
+#define DEFAULT_OUTPUT_WIDTH		INPUT_WIDTH
+#define DEFAULT_OUTPUT_HEIGHT		INPUT_HEIGHT
+#define DEFAULT_ACCEPTABLE_ERR		0.01
+
+void ParseArgs(int argc, char **argv, unsigned int *output_width, unsigned int *output_height, double *acceptable_error, int *prompt)
+{
+	int fail = 0;
+	int i;
+
+	*output_width = DEFAULT_OUTPUT_WIDTH;
+	*output_height = DEFAULT_OUTPUT_WIDTH;
+	*acceptable_error = DEFAULT_ACCEPTABLE_ERR;
+	*prompt = 1;
+
+	for (i = 1; i < argc && !fail; ++i)
+	{
+		if (!strcmp(argv[i], "-w"))
+		{
+			if (sscanf(argv[++i], "%u", output_width) != 1)
+				fail = 1;
+		}
+		else if (!strcmp(argv[i], "-h"))
+		{
+			if (sscanf(argv[++i], "%u", output_height) != 1)
+				fail = 1;
+		}
+		else if (!strcmp(argv[i], "-e"))
+		{
+			if (sscanf(argv[++i], "%lf", acceptable_error) != 1)
+				fail = 1;
+		}
+		else if (strcmp(argv[i], "-n"))
+			*prompt = 0;
+		else
+			fail = 1;
+	}
+
+	if (fail)
+		error
+		(
+			1, 0,
+			"Bad argument.\n"
+			"\n"
+			"Usage: %s [options]\n"
+			"\t-w <width>\tOutput width\n"
+			"\t-h <height>\tOutput height\n"
+			"\t-e <error>\tAcceptable margin of error per pixel, from 0 to 1\n"
+			"\t-n\tDon't prompt for quit\n",
+			argv[0]
+		);
+}
+
+void Gradient(short *block, unsigned int start, unsigned int stop, int horizontal)
+{
+	unsigned int x, y;
+	unsigned int range = stop - start;
+
+	if (horizontal)
+	{
+		for (y = 0; y < BLOCK_HEIGHT; ++y)
+			for (x = 0; x < BLOCK_WIDTH; ++x)
+				block[y * BLOCK_WIDTH + x] = (short)(start + range * (x / (float)(BLOCK_WIDTH - 1)));
+	}
+	else
+	{
+		for (y = 0; y < BLOCK_HEIGHT; ++y)
+			for (x = 0; x < BLOCK_WIDTH; ++x)
+				block[y * BLOCK_WIDTH + x] = (short)(start + range * (y / (float)(BLOCK_HEIGHT - 1)));
+	}
+}
+
+int main(int argc, char **argv)
+{
+	unsigned int		output_width;
+	unsigned int		output_height;
+	double			acceptable_error;
+	int			prompt;
+	Display			*display;
+	Window			root, window;
+	const unsigned int	mc_types[2] = {XVMC_MOCOMP | XVMC_MPEG_2, XVMC_IDCT | XVMC_MPEG_2};
+	XvPortID		port_num;
+	int			surface_type_id;
+	unsigned int		is_overlay, intra_unsigned;
+	int			colorkey;
+	XvMCContext		context;
+	XvMCSurface		surface;
+	XvMCBlockArray		block_array;
+	XvMCMacroBlockArray	mb_array;
+	int			mbx, mby, bx, by;
+	XvMCMacroBlock		*mb;
+	short			*blocks;
+	int			quit = 0;
+
+	ParseArgs(argc, argv, &output_width, &output_height, &acceptable_error, &prompt);
+
+	display = XOpenDisplay(NULL);
+
+	if (!GetPort
+	(
+		display,
+		INPUT_WIDTH,
+		INPUT_HEIGHT,
+		XVMC_CHROMA_FORMAT_420,
+    		mc_types,
+    		2,
+    		&port_num,
+    		&surface_type_id,
+    		&is_overlay,
+    		&intra_unsigned
+	))
+	{
+		XCloseDisplay(display);
+		error(1, 0, "Error, unable to find a good port.\n");
+	}
+
+	if (is_overlay)
+	{
+		Atom xv_colorkey = XInternAtom(display, "XV_COLORKEY", 0);
+		XvGetPortAttribute(display, port_num, xv_colorkey, &colorkey);
+	}
+
+	root = XDefaultRootWindow(display);
+	window = XCreateSimpleWindow(display, root, 0, 0, output_width, output_height, 0, 0, colorkey);
+
+	assert(XvMCCreateContext(display, port_num, surface_type_id, INPUT_WIDTH, INPUT_HEIGHT, XVMC_DIRECT, &context) == Success);
+	assert(XvMCCreateSurface(display, &context, &surface) == Success);
+	assert(XvMCCreateBlocks(display, &context, NUM_MACROBLOCKS * BLOCKS_PER_MACROBLOCK, &block_array) == Success);
+	assert(XvMCCreateMacroBlocks(display, &context, NUM_MACROBLOCKS, &mb_array) == Success);
+
+	mb = mb_array.macro_blocks;
+	blocks = block_array.blocks;
+
+	for (mby = 0; mby < INPUT_HEIGHT_IN_MACROBLOCKS; ++mby)
+		for (mbx = 0; mbx < INPUT_WIDTH_IN_MACROBLOCKS; ++mbx)
+		{
+			mb->x = mbx;
+			mb->y = mby;
+			mb->macroblock_type = XVMC_MB_TYPE_INTRA;
+			/*mb->motion_type = ;*/
+			/*mb->motion_vertical_field_select = ;*/
+			mb->dct_type = XVMC_DCT_TYPE_FRAME;
+			/*mb->PMV[0][0][0] = ;
+			mb->PMV[0][0][1] = ;
+			mb->PMV[0][1][0] = ;
+			mb->PMV[0][1][1] = ;
+			mb->PMV[1][0][0] = ;
+			mb->PMV[1][0][1] = ;
+			mb->PMV[1][1][0] = ;
+			mb->PMV[1][1][1] = ;*/
+			mb->index = (mby * INPUT_WIDTH_IN_MACROBLOCKS + mbx) * BLOCKS_PER_MACROBLOCK;
+			mb->coded_block_pattern = 0x3F;
+
+			mb++;
+
+			for (by = 0; by < MACROBLOCK_HEIGHT_IN_BLOCKS; ++by)
+				for (bx = 0; bx < MACROBLOCK_WIDTH_IN_BLOCKS; ++bx)
+				{
+					const int start = 16, stop = 235, range = stop - start;
+
+					Gradient
+					(
+						blocks,
+						(short)(start + range * ((mbx * MACROBLOCK_WIDTH + bx * BLOCK_WIDTH) / (float)(INPUT_WIDTH - 1))),
+						(short)(start + range * ((mbx * MACROBLOCK_WIDTH + bx * BLOCK_WIDTH + BLOCK_WIDTH - 1) / (float)(INPUT_WIDTH - 1))),
+						1
+					);
+
+					blocks += BLOCK_SIZE;
+				}
+
+			for (by = 0; by < MACROBLOCK_HEIGHT_IN_BLOCKS / 2; ++by)
+				for (bx = 0; bx < MACROBLOCK_WIDTH_IN_BLOCKS / 2; ++bx)
+				{
+					const int start = 16, stop = 240, range = stop - start;
+
+					Gradient
+					(
+						blocks,
+						(short)(start + range * ((mbx * MACROBLOCK_WIDTH + bx * BLOCK_WIDTH) / (float)(INPUT_WIDTH - 1))),
+						(short)(start + range * ((mbx * MACROBLOCK_WIDTH + bx * BLOCK_WIDTH + BLOCK_WIDTH - 1) / (float)(INPUT_WIDTH - 1))),
+						1
+					);
+
+					blocks += BLOCK_SIZE;
+
+					Gradient
+					(
+						blocks,
+						(short)(start + range * ((mbx * MACROBLOCK_WIDTH + bx * BLOCK_WIDTH) / (float)(INPUT_WIDTH - 1))),
+						(short)(start + range * ((mbx * MACROBLOCK_WIDTH + bx * BLOCK_WIDTH + BLOCK_WIDTH - 1) / (float)(INPUT_WIDTH - 1))),
+						1
+					);
+
+					blocks += BLOCK_SIZE;
+				}
+		}
+
+	XSelectInput(display, window, ExposureMask | KeyPressMask);
+	XMapWindow(display, window);
+	XSync(display, 0);
+
+	/* Test NULL context */
+	assert(XvMCRenderSurface(display, NULL, XVMC_FRAME_PICTURE, &surface, NULL, NULL, 0, NUM_MACROBLOCKS, 0, &mb_array, &block_array) == XvMCBadContext);
+	/* Test NULL surface */
+	assert(XvMCRenderSurface(display, &context, XVMC_FRAME_PICTURE, NULL, NULL, NULL, 0, NUM_MACROBLOCKS, 0, &mb_array, &block_array) == XvMCBadSurface);
+	/* Test bad picture structure */
+	assert(XvMCRenderSurface(display, &context, 0, &surface, NULL, NULL, 0, NUM_MACROBLOCKS, 0, &mb_array, &block_array) == BadValue);
+	/* Test valid params */
+	assert(XvMCRenderSurface(display, &context, XVMC_FRAME_PICTURE, &surface, NULL, NULL, 0, NUM_MACROBLOCKS, 0, &mb_array, &block_array) == Success);
+
+	/* Test NULL surface */
+	assert(XvMCPutSurface(display, NULL, window, 0, 0, INPUT_WIDTH, INPUT_HEIGHT, 0, 0, output_width, output_height, XVMC_FRAME_PICTURE) == XvMCBadSurface);
+	/* Test bad window */
+	/* XXX: X halts with a bad drawable for some reason, doesn't return BadDrawable as expected */
+	/*assert(XvMCPutSurface(display, &surface, 0, 0, 0, width, height, 0, 0, width, height, XVMC_FRAME_PICTURE) == BadDrawable);*/
+
+	if (prompt)
+	{
+		puts("Press any button to quit...");
+
+		while (!quit)
+		{
+			if (XPending(display) > 0)
+			{
+				XEvent event;
+
+				XNextEvent(display, &event);
+
+				switch (event.type)
+				{
+					case Expose:
+					{
+						/* Test valid params */
+						assert
+						(
+							XvMCPutSurface
+							(
+								display, &surface, window,
+								0, 0, INPUT_WIDTH, INPUT_HEIGHT,
+								0, 0, output_width, output_height,
+								XVMC_FRAME_PICTURE
+							) == Success
+						);
+						break;
+					}
+					case KeyPress:
+					{
+						quit = 1;
+						break;
+					}
+				}
+			}
+		}
+	}
+
+	assert(XvMCDestroyBlocks(display, &block_array) == Success);
+	assert(XvMCDestroyMacroBlocks(display, &mb_array) == Success);
+	assert(XvMCDestroySurface(display, &surface) == Success);
+	assert(XvMCDestroyContext(display, &context) == Success);
+
+	XvUngrabPort(display, port_num, CurrentTime);
+	XDestroyWindow(display, window);
+	XCloseDisplay(display);
+
+	return 0;
+}
diff --git a/src/xvmc/tests/test_surface.c b/src/xvmc/tests/test_surface.c
new file mode 100644
index 0000000000..25ebdcc4fc
--- /dev/null
+++ b/src/xvmc/tests/test_surface.c
@@ -0,0 +1,72 @@
+#include <assert.h>
+#include <error.h>
+#include "testlib.h"
+
+int main(int argc, char **argv)
+{
+	const unsigned int	width = 16, height = 16;
+	const unsigned int	mc_types[2] = {XVMC_MOCOMP | XVMC_MPEG_2, XVMC_IDCT | XVMC_MPEG_2};
+	
+	Display			*display;
+	XvPortID		port_num;
+	int			surface_type_id;
+	unsigned int		is_overlay, intra_unsigned;
+	int			colorkey;
+	XvMCContext		context;
+	XvMCSurface		surface = {0};
+	
+	display = XOpenDisplay(NULL);
+	
+	if (!GetPort
+	(
+		display,
+		width,
+		height,
+		XVMC_CHROMA_FORMAT_420,
+    		mc_types,
+    		2,
+    		&port_num,
+    		&surface_type_id,
+    		&is_overlay,
+    		&intra_unsigned
+	))
+	{
+		XCloseDisplay(display);
+		error(1, 0, "Error, unable to find a good port.\n");
+	}
+	
+	if (is_overlay)
+	{
+		Atom xv_colorkey = XInternAtom(display, "XV_COLORKEY", 0);
+		XvGetPortAttribute(display, port_num, xv_colorkey, &colorkey);
+	}
+	
+	assert(XvMCCreateContext(display, port_num, surface_type_id, width, height, XVMC_DIRECT, &context) == Success);
+	
+	/* Test NULL context */
+	assert(XvMCCreateSurface(display, NULL, &surface) == XvMCBadContext);
+	/* Test NULL surface */
+	assert(XvMCCreateSurface(display, &context, NULL) == XvMCBadSurface);
+	/* Test valid params */
+	assert(XvMCCreateSurface(display, &context, &surface) == Success);
+	/* Test surface id assigned */
+	assert(surface.surface_id != 0);
+	/* Test context id assigned and correct */
+	assert(surface.context_id == context.context_id);
+	/* Test surface type id assigned and correct */
+	assert(surface.surface_type_id == surface_type_id);
+	/* Test width & height assigned and correct */
+	assert(surface.width == width && surface.height == height);
+	/* Test valid params */
+	assert(XvMCDestroySurface(display, &surface) == Success);
+	/* Test NULL surface */
+	assert(XvMCDestroySurface(display, NULL) == XvMCBadSurface);
+	
+	assert(XvMCDestroyContext(display, &context) == Success);
+	
+	XvUngrabPort(display, port_num, CurrentTime);
+	XCloseDisplay(display);
+	
+	return 0;
+}
+
diff --git a/src/xvmc/tests/testlib.c b/src/xvmc/tests/testlib.c
new file mode 100644
index 0000000000..59a03ca813
--- /dev/null
+++ b/src/xvmc/tests/testlib.c
@@ -0,0 +1,119 @@
+#include "testlib.h"
+#include <stdio.h>
+
+/*
+void test(int pred, const char *pred_string, const char *doc_string, const char *file, unsigned int line)
+{
+	fputs(doc_string, stderr);
+	if (!pred)
+		fprintf(stderr, " FAIL!\n\t\"%s\" at %s:%u\n", pred_string, file, line);
+	else
+		fputs(" PASS!\n", stderr);
+}
+*/
+
+int GetPort
+(
+	Display *display,
+	unsigned int width,
+	unsigned int height,
+	unsigned int chroma_format,
+	const unsigned int *mc_types,
+	unsigned int num_mc_types,
+	XvPortID *port_id,
+	int *surface_type_id,
+	unsigned int *is_overlay,
+	unsigned int *intra_unsigned
+)
+{
+	unsigned int	found_port = 0;
+	XvAdaptorInfo	*adaptor_info;
+	unsigned int	num_adaptors;
+	int		num_types;
+	int		ev_base, err_base;
+	unsigned int	i, j, k, l;
+
+	if (!XvMCQueryExtension(display, &ev_base, &err_base))
+		return 0;
+	if (XvQueryAdaptors(display, XDefaultRootWindow(display), &num_adaptors, &adaptor_info) != Success)
+		return 0;
+
+	for (i = 0; i < num_adaptors && !found_port; ++i)
+	{
+		if (adaptor_info[i].type & XvImageMask)
+		{
+			XvMCSurfaceInfo *surface_info = XvMCListSurfaceTypes(display, adaptor_info[i].base_id, &num_types);
+
+			if (surface_info)
+			{
+				for (j = 0; j < num_types && !found_port; ++j)
+				{
+					if
+					(
+						surface_info[j].chroma_format == chroma_format &&
+						surface_info[j].max_width >= width &&
+						surface_info[j].max_height >= height
+					)
+					{
+						for (k = 0; k < num_mc_types && !found_port; ++k)
+						{
+							if (surface_info[j].mc_type == mc_types[k])
+							{
+								for (l = 0; l < adaptor_info[i].num_ports && !found_port; ++l)
+								{
+									if (XvGrabPort(display, adaptor_info[i].base_id + l, CurrentTime) == Success)
+									{
+										*port_id = adaptor_info[i].base_id + l;
+										*surface_type_id = surface_info[j].surface_type_id;
+										*is_overlay = surface_info[j].flags & XVMC_OVERLAID_SURFACE;
+										*intra_unsigned = surface_info[j].flags & XVMC_INTRA_UNSIGNED;
+										found_port = 1;
+									}
+								}
+							}
+						}
+					}
+				}
+
+				XFree(surface_info);
+			}
+		}
+	}
+
+	XvFreeAdaptorInfo(adaptor_info);
+
+	return found_port;
+}
+
+unsigned int align(unsigned int value, unsigned int alignment)
+{
+	return (value + alignment - 1) & ~(alignment - 1);
+}
+
+/* From the glibc manual */
+int timeval_subtract(struct timeval *result, struct timeval *x, struct timeval *y)
+{
+	/* Perform the carry for the later subtraction by updating y. */
+	if (x->tv_usec < y->tv_usec)
+	{
+		int nsec = (y->tv_usec - x->tv_usec) / 1000000 + 1;
+		y->tv_usec -= 1000000 * nsec;
+		y->tv_sec += nsec;
+	}
+	if (x->tv_usec - y->tv_usec > 1000000)
+	{
+		int nsec = (x->tv_usec - y->tv_usec) / 1000000;
+		y->tv_usec += 1000000 * nsec;
+		y->tv_sec -= nsec;
+	}
+
+	/*
+	 * Compute the time remaining to wait.
+	 * tv_usec is certainly positive.
+	 */
+	result->tv_sec = x->tv_sec - y->tv_sec;
+	result->tv_usec = x->tv_usec - y->tv_usec;
+
+	/* Return 1 if result is negative. */
+	return x->tv_sec < y->tv_sec;
+}
diff --git a/src/xvmc/tests/testlib.h b/src/xvmc/tests/testlib.h
new file mode 100644
index 0000000000..af71ad74e1
--- /dev/null
+++ b/src/xvmc/tests/testlib.h
@@ -0,0 +1,42 @@
+#ifndef testlib_h
+#define testlib_h
+
+/*
+#define TEST(pred, doc)	test(pred, #pred, doc, __FILE__, __LINE__)
+
+void test(int pred, const char *pred_string, const char *doc_string, const char *file, unsigned int line);
+*/
+
+#include <sys/time.h>
+#include <X11/Xlib.h>
+#include <X11/extensions/XvMClib.h>
+
+/*
+ * display: IN			A valid X display
+ * width, height: IN		Surface size that the port must display
+ * chroma_format: IN		Chroma format that the port must display
+ * mc_types, num_mc_types: IN	List of MC types that the port must support, first port that matches the first mc_type will be returned
+ * port_id: OUT			Your port's ID
+ * surface_type_id: OUT		Your port's surface ID
+ * is_overlay: OUT		If 1, port uses overlay surfaces, you need to set a colorkey
+ * intra_unsigned: OUT		If 1, port uses unsigned values for intra-coded blocks
+ */
+int GetPort
+(
+	Display *display,
+	unsigned int width,
+	unsigned int height,
+	unsigned int chroma_format,
+	const unsigned int *mc_types,
+	unsigned int num_mc_types,
+	XvPortID *port_id,
+	int *surface_type_id,
+	unsigned int *is_overlay,
+	unsigned int *intra_unsigned
+);
+
+unsigned int align(unsigned int value, unsigned int alignment);
+
+int timeval_subtract(struct timeval *result, struct timeval *x, struct timeval *y);
+
+#endif
diff --git a/src/xvmc/tests/xvmc_bench.c b/src/xvmc/tests/xvmc_bench.c
new file mode 100644
index 0000000000..d5a39ecf17
--- /dev/null
+++ b/src/xvmc/tests/xvmc_bench.c
@@ -0,0 +1,271 @@
+#include <assert.h>
+#include <stdio.h>
+#include <string.h>
+#include <error.h>
+#include <sys/time.h>
+#include "testlib.h"
+
+#define MACROBLOCK_WIDTH		16
+#define MACROBLOCK_HEIGHT		16
+#define BLOCKS_PER_MACROBLOCK		6
+
+#define DEFAULT_INPUT_WIDTH		720
+#define DEFAULT_INPUT_HEIGHT		480
+#define DEFAULT_REPS			100
+
+#define PIPELINE_STEP_MC		1
+#define PIPELINE_STEP_CSC		2
+#define PIPELINE_STEP_SWAP		4
+
+#define MB_TYPE_I			1
+#define MB_TYPE_P			2
+#define MB_TYPE_B			4
+
+struct Config
+{
+	unsigned int input_width;
+	unsigned int input_height;
+	unsigned int output_width;
+	unsigned int output_height;
+	unsigned int pipeline;
+	unsigned int mb_types;
+	unsigned int reps;
+};
+
+void ParseArgs(int argc, char **argv, struct Config *config)
+{
+	int fail = 0;
+	int i;
+
+	config->input_width = DEFAULT_INPUT_WIDTH;
+	config->input_height = DEFAULT_INPUT_HEIGHT;
+	config->output_width = 0;
+	config->output_height = 0;
+	config->pipeline = 0;
+	config->mb_types = 0;
+	config->reps = DEFAULT_REPS;
+
+	for (i = 1; i < argc && !fail; ++i)
+	{
+		if (!strcmp(argv[i], "-iw"))
+		{
+			if (sscanf(argv[++i], "%u", &config->input_width) != 1)
+				fail = 1;
+		}
+		else if (!strcmp(argv[i], "-ih"))
+		{
+			if (sscanf(argv[++i], "%u", &config->input_height) != 1)
+				fail = 1;
+		}
+		else if (!strcmp(argv[i], "-ow"))
+		{
+			if (sscanf(argv[++i], "%u", &config->output_width) != 1)
+				fail = 1;
+		}
+		else if (!strcmp(argv[i], "-oh"))
+		{
+			if (sscanf(argv[++i], "%u", &config->output_height) != 1)
+				fail = 1;
+		}
+		else if (!strcmp(argv[i], "-p"))
+		{
+			char *token = strtok(argv[++i], ",");
+
+			while (token && !fail)
+			{
+				if (!strcmp(token, "mc"))
+					config->pipeline |= PIPELINE_STEP_MC;
+				else if (!strcmp(token, "csc"))
+					config->pipeline |= PIPELINE_STEP_CSC;
+				else if (!strcmp(token, "swp"))
+					config->pipeline |= PIPELINE_STEP_SWAP;
+				else
+					fail = 1;
+
+				if (!fail)
+					token = strtok(NULL, ",");
+			}
+		}
+		else if (!strcmp(argv[i], "-mb"))
+		{
+			char *token = strtok(argv[++i], ",");
+
+			while (token && !fail)
+			{
+				if (strcmp(token, "i"))
+					config->mb_types |= MB_TYPE_I;
+				else if (strcmp(token, "p"))
+					config->mb_types |= MB_TYPE_P;
+				else if (strcmp(token, "b"))
+					config->mb_types |= MB_TYPE_B;
+				else
+					fail = 1;
+
+				if (!fail)
+					token = strtok(NULL, ",");
+			}
+		}
+		else if (!strcmp(argv[i], "-r"))
+		{
+			if (sscanf(argv[++i], "%u", &config->reps) != 1)
+				fail = 1;
+		}
+		else
+			fail = 1;
+	}
+
+	if (fail)
+		error
+		(
+			1, 0,
+			"Bad argument.\n"
+			"\n"
+			"Usage: %s [options]\n"
+			"\t-iw <width>\tInput width\n"
+			"\t-ih <height>\tInput height\n"
+			"\t-ow <width>\tOutput width\n"
+			"\t-oh <height>\tOutput height\n"
+			"\t-p <pipeline>\tPipeline to test\n"
+			"\t-mb <mb type>\tMacroBlock types to use\n"
+			"\t-r <reps>\tRepetitions\n\n"
+			"\tPipeline steps: mc,csc,swap\n"
+			"\tMB types: i,p,b\n",
+			argv[0]
+		);
+
+	if (config->output_width == 0)
+		config->output_width = config->input_width;
+	if (config->output_height == 0)
+		config->output_height = config->input_height;
+	if (!config->pipeline)
+		config->pipeline = PIPELINE_STEP_MC | PIPELINE_STEP_CSC | PIPELINE_STEP_SWAP;
+	if (!config->mb_types)
+		config->mb_types = MB_TYPE_I | MB_TYPE_P | MB_TYPE_B;
+}
+
+int main(int argc, char **argv)
+{
+	struct Config		config;
+	Display			*display;
+	Window			root, window;
+	const unsigned int	mc_types[2] = {XVMC_MOCOMP | XVMC_MPEG_2, XVMC_IDCT | XVMC_MPEG_2};
+	XvPortID		port_num;
+	int			surface_type_id;
+	unsigned int		is_overlay, intra_unsigned;
+	int			colorkey;
+	XvMCContext		context;
+	XvMCSurface		surface;
+	XvMCBlockArray		block_array;
+	XvMCMacroBlockArray	mb_array;
+	unsigned int		mbw, mbh;
+	unsigned int		mbx, mby;
+	unsigned int		reps;
+	struct timeval		start, stop, diff;
+	double			diff_secs;
+
+	ParseArgs(argc, argv, &config);
+
+	mbw = align(config.input_width, MACROBLOCK_WIDTH) / MACROBLOCK_WIDTH;
+	mbh = align(config.input_height, MACROBLOCK_HEIGHT) / MACROBLOCK_HEIGHT;
+
+	display = XOpenDisplay(NULL);
+
+	if (!GetPort
+	(
+		display,
+		config.input_width,
+		config.input_height,
+		XVMC_CHROMA_FORMAT_420,
+    		mc_types,
+    		2,
+    		&port_num,
+    		&surface_type_id,
+    		&is_overlay,
+    		&intra_unsigned
+	))
+	{
+		XCloseDisplay(display);
+		error(1, 0, "Error, unable to find a good port.\n");
+	}
+
+	if (is_overlay)
+	{
+		Atom xv_colorkey = XInternAtom(display, "XV_COLORKEY", 0);
+		XvGetPortAttribute(display, port_num, xv_colorkey, &colorkey);
+	}
+
+	root = XDefaultRootWindow(display);
+	window = XCreateSimpleWindow(display, root, 0, 0, config.output_width, config.output_height, 0, 0, colorkey);
+
+	assert(XvMCCreateContext(display, port_num, surface_type_id, config.input_width, config.input_height, XVMC_DIRECT, &context) == Success);
+	assert(XvMCCreateSurface(display, &context, &surface) == Success);
+	assert(XvMCCreateBlocks(display, &context, mbw * mbh * BLOCKS_PER_MACROBLOCK, &block_array) == Success);
+	assert(XvMCCreateMacroBlocks(display, &context, mbw * mbh, &mb_array) == Success);
+
+	for (mby = 0; mby < mbh; ++mby)
+		for (mbx = 0; mbx < mbw; ++mbx)
+		{
+			mb_array.macro_blocks[mby * mbw + mbx].x = mbx;
+			mb_array.macro_blocks[mby * mbw + mbx].y = mby;
+			mb_array.macro_blocks[mby * mbw + mbx].macroblock_type = XVMC_MB_TYPE_INTRA;
+			/*mb->motion_type = ;*/
+			/*mb->motion_vertical_field_select = ;*/
+			mb_array.macro_blocks[mby * mbw + mbx].dct_type = XVMC_DCT_TYPE_FRAME;
+			/*mb->PMV[0][0][0] = ;
+			mb->PMV[0][0][1] = ;
+			mb->PMV[0][1][0] = ;
+			mb->PMV[0][1][1] = ;
+			mb->PMV[1][0][0] = ;
+			mb->PMV[1][0][1] = ;
+			mb->PMV[1][1][0] = ;
+			mb->PMV[1][1][1] = ;*/
+			mb_array.macro_blocks[mby * mbw + mbx].index = (mby * mbw + mbx) * BLOCKS_PER_MACROBLOCK;
+			mb_array.macro_blocks[mby * mbw + mbx].coded_block_pattern = 0x3F;
+		}
+
+	XSelectInput(display, window, ExposureMask | KeyPressMask);
+	XMapWindow(display, window);
+	XSync(display, 0);
+
+	gettimeofday(&start, NULL);
+
+	for (reps = 0; reps < config.reps; ++reps)
+	{
+		if (config.pipeline & PIPELINE_STEP_MC)
+		{
+			assert(XvMCRenderSurface(display, &context, XVMC_FRAME_PICTURE, &surface, NULL, NULL, 0, mbw * mbh, 0, &mb_array, &block_array) == Success);
+			assert(XvMCFlushSurface(display, &surface) == Success);
+		}
+		if (config.pipeline & PIPELINE_STEP_CSC)
+			assert(XvMCPutSurface(display, &surface, window, 0, 0, config.input_width, config.input_height, 0, 0, config.output_width, config.output_height, XVMC_FRAME_PICTURE) == Success);
+	}
+
+	gettimeofday(&stop, NULL);
+
+	timeval_subtract(&diff, &stop, &start);
+	diff_secs = (double)diff.tv_sec + (double)diff.tv_usec / 1000000.0;
+
+	printf("XvMC Benchmark\n");
+	printf("Input: %u,%u\nOutput: %u,%u\n", config.input_width, config.input_height, config.output_width, config.output_height);
+	printf("Pipeline: ");
+	if (config.pipeline & PIPELINE_STEP_MC)
+		printf("|mc|");
+	if (config.pipeline & PIPELINE_STEP_CSC)
+		printf("|csc|");
+	if (config.pipeline & PIPELINE_STEP_SWAP)
+		printf("|swap|");
+	printf("\n");
+	printf("Reps: %u\n", config.reps);
+	printf("Total time: %.2lf (%.2lf reps / sec)\n", diff_secs, config.reps / diff_secs);
+
+	assert(XvMCDestroyBlocks(display, &block_array) == Success);
+	assert(XvMCDestroyMacroBlocks(display, &mb_array) == Success);
+	assert(XvMCDestroySurface(display, &surface) == Success);
+	assert(XvMCDestroyContext(display, &context) == Success);
+
+	XvUngrabPort(display, port_num, CurrentTime);
+	XDestroyWindow(display, window);
+	XCloseDisplay(display);
+
+	return 0;
+}